{"unit_id": "orig:1934748940152500427", "ticker_idx": 0, "ts": "2025-06-16T23:04:27+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes market sentiment is nearing a bottom, and the stock is likely to see a positive reaction after this event", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares Morgan Stanley Equal Weight on MRVL but with tactical bullish lean: sentiment near bottom, positive stock reaction expected post-ASIC webcast.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell: Focus on ASIC Updates, Optical Communications Business More Anticipated\nMorgan Stanley (J. Moore, June 16, 2025)\n\nMorgan Stanley judges that AI investments are still accelerating, even at current high levels – while Morgan Stanley has its own preferences, all AI beneficiaries should be in a favorable development environment. Marvell is no exception. Given that its performance in the past few quarters met expectations but fell short of higher market expectations, leading to negative market sentiment, Morgan Stanley believes the stock has some tactical appeal before this event.\n\nHowever, compared to ASICs, Morgan Stanley is more optimistic about the optical module business, and the company's decision to cancel Analyst Day and focus solely on ASICs seems somewhat surprising. Morgan Stanley analyzes that while the cooperation opportunity with Amazon is substantial, this collaboration model has a clear dilutive effect on Marvell's profit margins, as Amazon's Annapurna team undertakes most of the design work.\n\nNevertheless, Morgan Stanley has re-lowered its average selling price (ASP) forecast for Trainium 2 sales, and the current controversy surrounding Trainium 3 is unlikely to be a significant negative factor in 2025. With related revenue of approximately $2 billion in 2025, starting from a lower base will be beneficial for growth in 2026.\n\nRegarding expectations for this event, Morgan Stanley anticipates that the company will update its AI-related guidance during the webcast, introduce the latest progress of its ASIC customers, and disclose more information about its advanced packaging platforms and custom networking business. Marvell has historically been effective in boosting investor sentiment during Analyst Days, and Morgan Stanley expects an update to AI revenue guidance and TAM (Total Addressable Market) forecasts this time, especially given that its 2024 guidance has been over-achieved.\n\nRegarding the ASIC Total Addressable Market (TAM):\n\nMarvell stated during its AI Day last year that AI-related revenue would exceed $1.5 billion in 2024 and reach $2.5 billion in 2025. In addition, the company projected an overall AI accelerator TAM of $172 billion by 2028 (including standardized products), with custom chips expected to account for 16%, or $27.5 billion. Morgan Stanley points out that this forecast is significantly lower than Broadcom's (AVGO) projected customer SAM (Serviceable Addressable Market) of $60 billion to $90 billion by 2027 (including standard GPUs). Therefore, Morgan Stanley believes Marvell may raise its own forecast at this event. Marvell also once predicted that its overall data center revenue (covering ASICs, switching, interconnect, and storage) could reach $15 billion by 2028, implying a compound annual growth rate (CAGR) of 38% for this business from 2024 to 2028. Morgan Stanley believes this goal is quite challenging. Currently, Morgan Stanley's model projects Marvell's data center revenue to grow from $4.2 billion in 2024 to $6.2 billion in 2025, with AI revenue growing from $2.1 billion to $3.7 billion.\nMorgan Stanley does not expect any new AI accelerator customer announcements but remains focused on the latest progress of the two current major customers:\n\nRegarding the Trainium project: Morgan Stanley judges that it previously overestimated the potential of Trainium in terms of price or shipments. It currently expects Trainium 2 to grow linearly in 2025, contributing approximately $2 billion in revenue for the full year. There is still some uncertainty regarding the next-generation Trainium 3; although Alchip (covered by Charlie Chan) may capture a share of some orders, based on Marvell management and multiple channel information, Morgan Stanley believes the company has secured a significant ASIC project from Amazon that will drive growth in 2026. While MRVL needs to maintain confidentiality regarding customer roadmaps, Morgan Stanley hopes this event will reveal more details about related projects for 2026.\n\nRegarding the Microsoft Maia project: Marvell's second AI accelerator customer is believed to be an even greater opportunity than the Amazon Trainium project because Maia is a complete front-end design, utilizing more of Marvell's own IP. Morgan Stanley notes that company management is optimistic about on-schedule mass production in 2025, but due to the high uncertainty of ASIC projects, Morgan Stanley still considers it a \"wait-and-see\" situation. Although there are rumors that the third-generation chip might shift to AVGO, Morgan Stanley believes this is unlikely – partly due to high conversion costs, and partly because it is currently too early. However, Morgan Stanley believes Broadcom is also observing this project and would actively pursue it if an opportunity arises. Any details regarding project progress will be a market focus.\n\nRegarding stock price judgment: Morgan Stanley believes that the company's most optimistic growth expectation currently comes from its optical module business, which is why its decision to position this event as an \"ASIC webinar\" rather than an overall AI update is curious. Considering the negative impact from the Chinese market, single-digit growth in optical module revenue this quarter is reasonable. But looking to the second half of the year, as demand recovers, Marvell should see accelerating growth in its PAM4 and 400ZR businesses. Morgan Stanley also notes that while the launch of NVIDIA's Rubin-generation GPU might lead to some market share pressure, such a pullback is acceptable given its already high market share.\n\nOverall, Morgan Stanley maintains its Equal Weight rating on Marvell, not issuing a significantly bullish or bearish signal, as other AI-related stocks are currently more attractive in terms of valuation and sentiment. However, considering that Marvell remains a core member of the AI winners' camp, Morgan Stanley believes market sentiment is nearing a bottom, and the stock is likely to see a positive reaction after this event.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04586764182521108, "ret_signed_m1d": -0.04586764182521108, "alpha_spy_m1d": -0.0364431121733233, "alpha_c_m1d": -0.02160600448809946, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00610621991647331, "ret_signed_p1d": -0.00610621991647331, "alpha_spy_p1d": 0.0024388126201055904, "alpha_c_p1d": 0.0006900698042445752, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.005112108324812681, "ret_signed_p1w": 0.005112108324812681, "alpha_spy_p1w": 0.006366103629449449, "alpha_c_p1w": 0.010655443635985251, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.029100713031454895, "ret_signed_p1m": 0.029100713031454895, "alpha_spy_p1m": -0.006240197607602305, "alpha_c_p1m": -0.07500736184128831, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": -0.053613912304387834, "ret_signed_p3m": -0.053613912304387834, "alpha_spy_p3m": -0.14801593262788015, "alpha_c_p3m": -0.20757388168952773, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.2642953920412845, "ret_signed_p6m": 0.2642953920412845, "alpha_spy_p6m": 0.12445586287646981, "alpha_c_p6m": -0.13672208132617003, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.0098, -0.0001, -0.0016, 0.0321, 0.0166, -0.053, -0.08, -0.12, -0.1267, -0.1111, -0.1031, -0.2108, -0.2989, -0.2766, -0.2904, -0.1353, -0.2501, -0.2418, -0.2579, -0.2428, -0.2626, -0.2658, -0.2658, -0.2988, -0.2812, -0.2364, -0.1859, -0.1633, -0.1664, -0.1667, -0.1711, -0.1345, -0.1149, -0.1199, -0.1306, -0.2004, -0.1816, -0.1529, -0.0841, -0.0702, -0.0642, -0.0741, -0.0946, -0.1116, -0.1278, -0.1464, -0.1217, -0.1382, -0.1382, -0.0937, -0.0828, -0.095, -0.1453, -0.1271, -0.1145, -0.0585, -0.0747, -0.0294, -0.0182, -0.0224, -0.031, -0.0111, -0.0459, 0.0, -0.0061, 0.0643, 0.0643, 0.0439, 0.0051, 0.068, 0.0782, 0.1356, 0.0957, 0.0991, 0.0826, 0.0544, 0.0676, 0.0676, 0.016, 0.0217, 0.0261, 0.0417, 0.0334, 0.0305, 0.0291, 0.0069, 0.0234, 0.0609, 0.0383, 0.0231, 0.0413, 0.0523, 0.0547, 0.0788, 0.085, 0.1617, 0.1422, 0.0581, 0.0877, 0.0891, 0.0705, 0.078, 0.0992, 0.0983, 0.1058, 0.1273, 0.1233, 0.0828, 0.0906, 0.0243, 0.0122, 0.012, 0.0375, 0.0368, 0.0554, 0.0629, 0.0976, -0.1065, -0.1065, -0.0819, -0.1144, -0.089, -0.0999, -0.062, -0.0501, -0.0464, -0.0536, -0.0428, -0.0417, -0.0214, 0.0088, 0.055, 0.0554, 0.0734, 0.0605, 0.1382, 0.1911, 0.182, 0.1709, 0.1948, 0.1923, 0.2251, 0.2254, 0.2637, 0.236, 0.3146, 0.2888, 0.2175, 0.2713, 0.2262, 0.2642, 0.2548, 0.2508, 0.2208, 0.1983, 0.1527, 0.1771, 0.1965, 0.2616, 0.258, 0.2821, 0.2596, 0.3331, 0.2852, 0.2457, 0.3212, 0.3273, 0.293, 0.3259, 0.2704, 0.2704, 0.2447, 0.2295, 0.1868, 0.119, 0.1565, 0.0905, 0.1015, 0.1916, 0.1865, 0.2475, 0.2475, 0.2714, 0.2956, 0.321, 0.425, 0.3964, 0.4067, 0.3084, 0.2643]}
{"unit_id": "orig:1927860613444739179", "ticker_idx": 0, "ts": "2025-05-28T22:52:42+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sovereign AI is likely to begin contributing to NVIDIA's earnings soon. CAPEX spending from non-CSP entities is expected to drive reacceleration of NVIDIA's top-line growth.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish NVDA: Blackwell ramp at record pace, sovereign AI + Stargate expansion to drive top-line reacceleration, guidance beats even after $8B H20 export hit.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250529_1Q26 NVIDIA [Beat, $3.3tn, 24x]\n\n[Key Takeaways]\n\n(1) The Blackwell ramp-up is progressing at the fastest pace in history. Yield stabilization contributed to gross margins exceeding market expectations by +7%.\n\n(2) Sovereign AI is likely to begin contributing to NVIDIA’s earnings soon. CAPEX spending from non-CSP entities is expected to drive reacceleration of NVIDIA’s top-line growth.\n\n(3) The explosive growth in token throughput from services like ChatGPT and Gemini indicates an exponential surge in inference demand.\n\n(4) The GB300 system, architecturally optimized for Reasoning AI, has begun sampling at major CSPs and is expected to begin volume production and shipments by the end of this quarter.\n\n⸻\n\n[Contents]\n\n(1) Revenue beat by +2%, operating income beat by +7% versus consensus. Shares are up 4–5% in after-hours trading.\n\n(2) FY2Q26(F) Guidance\n• Revenue: $44.1–45.9 billion vs. Street estimate of $45.45 billion\n• Adjusted Gross Margin: 71.5–72.5% vs. 71.74%\n• Adjusted Operating Expenses: $4.0 billion vs. $3.86 billion\n\n(3) The FY2Q26 guidance reflects a revenue loss of ~$8.0 billion due to H20 export restrictions. Even so, the upper end of the guidance range is +1% above consensus.\n• Without the $8bn impact, implied revenue would be $53.9bn, which is +18.6% above expectations.\n• Street expectations had already been revised down by ~$4bn due to this issue, so adjusted guidance still beats by +9.8%.\n\n(4) In this quarter, a $4.5bn impairment charge was taken due to U.S. export restrictions to China. NVIDIA is also under regulatory investigation by Chinese authorities related to its Mellanox acquisition.\n• The impairment is $1.0bn lower than the originally guided $5.5bn.\n• H20 deliveries in H1 to China totaled $4.6bn, with $2.5bn undelivered. If fully delivered, the total would have been $7.0bn.\n(“Regarding H20, yes, we recognized $4.6 billion H20 in Q1. We were unable to ship $2.5 billion. So the total for Q1 should have been $7 billion.”)\n\n(5) Within the data center segment, networking revenue reached $4.97bn, a YoY increase of +56%, scaling up from $3bn–level in prior quarters to now exceeding $5bn.\n\n(6) While networking’s mix is still below the FY24 level of 18–22%, it has significantly improved from 8.5% last quarter. [Driven by NVLink scale-up, and enhanced connectivity via InfiniBand, Ethernet NICs, BlueField DPU, etc.]\n\n(7) This shift is attributed to a declining revenue mix from U.S. Big Tech hyperscalers and a rising contribution from Sovereign AI and second-tier players.\n\n(8) In fact, when comparing NVIDIA’s data center revenue to the combined CAPEX of the top 4 hyperscalers, the ratio rose to 51.1% this quarter from the mid-40% range in FY25.\n\n(9) With increasing buildouts of AI data center infrastructure across the Middle East and other regions, as well as the global expansion of OpenAI’s Stargate project, these trends are beginning to materialize in NVIDIA’s P&L. This is expected to mitigate China-related uncertainties and drive top-line reacceleration.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "thread:1930815488038511069", "ticker_idx": 0, "ts": "2025-06-06T02:34:19+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The market expects TSMC to continue hitting record highs in 2025 and 2026", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC expected to hit record highs in 2025–26 on AI monopoly; NVDA and AVGO also seen posting strong revenue growth as GPU and ASIC shipments expand 50%+ annually.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "ASICs “Ready to Duel” with AI GPUs – TSMC Dominates Orders and Raises Prices -Digitimes\n\nWith NVIDIA holding a near-monopoly in the AI GPU battlefield, numerous companies are accelerating the development and mass production of ASICs (application-specific integrated circuits), shaping an industry narrative of an impending “showdown” between NVIDIA GPUs and an expanding ASIC army. Notably, TSMC, which serves both camps, has raised prices for its 2nm and advanced packaging services, helping it grow revenue insulated from tariff and currency fluctuations. The market expects TSMC to continue hitting record highs in 2025 and 2026.\n\nA year ago, NVIDIA CEO Jensen Huang openly declared plans to enter the ASIC battlefield. However, he has since pivoted, now advocating “collaboration over competition” to grow the AI pie together.\n\nAccording to supply chain sources, AI demand is expected to remain robust over the long term. Shipments of both AI GPUs and ASICs are projected to grow steadily over the next three years. For ASICs in particular, shipments are conservatively estimated to grow at an annual rate of at least 50%. Leading players like NVIDIA and Broadcom are expected to see strong revenue growth, and TSMC—supplier to both—is profiting handsomely.\n\nThe rise of generative AI has ushered in the AI era and driven explosive demand for AI servers. NVIDIA, with its combination of GPUs and powerful CUDA software, quickly seized control of the ecosystem, securing dominance from customers to the broader supply chain, with everyone scrambling to collaborate.\n\nStill, supply chain insiders note that while NVIDIA’s AI GPUs are widely used for training and inference, demand has consistently outstripped supply since the boom in 2023. Moreover, the high prices and closed CUDA ecosystem have put cloud service providers (CSPs) and chipmakers at a disadvantage in price negotiations. To mitigate risk and ultimately “de-NVIDIA” their infrastructure, many are accelerating the development of in-house ASICs.\n\nSources explain that the collective goal of the ASIC players is to reduce dependency on NVIDIA’s expensive GPUs and CUDA stack. NVIDIA’s H100 and latest B200 chips cost tens of thousands of dollars each, and complete server systems can run into the millions. Though self-developing ASICs involves high initial costs, scaling production can dramatically reduce unit costs.\n\nThat’s why major NVIDIA clients like Google, AWS, Meta, and Microsoft are not only buying NVIDIA GPUs but also aggressively ramping up their in-house chip programs. They're placing direct wafer orders with fabs to avoid capacity constraints.\n\nMeanwhile, Broadcom—NVIDIA’s primary rival in the ASIC space—has also posted surprisingly strong growth.\n\nSupply chain sources note that while Jensen Huang initially announced an ASIC division to expand the customer base, he later pivoted to unveil the NVLink Fusion strategy. This strategy aims to establish a vast partner ecosystem built around NVLink, enabling semi-custom AI infrastructure development across industries.\n\nThis ecosystem includes companies like MediaTek, Marvell, SiFive, Astera Labs, Synopsys, Cadence, Fujitsu, and Qualcomm—but notably excludes Broadcom, raising speculation about the future dynamic between NVIDIA and Broadcom.\n\nOn the competition between ASICs and AI GPUs, TSMC Chairman Mark Liu plainly stated:\n\n“It doesn’t matter who wins—NVIDIA and the ASIC makers are all my customers. They all manufacture with TSMC.”\n\nAccording to supply chain insiders, with Samsung Electronics and Intel far behind TSMC in the sub-7nm race, TSMC has virtually monopolized major orders from 7nm down to the upcoming 2nm node. AI clients are choosing TSMC for its full-stack capabilities, including CoWoS advanced packaging.\n\nTSMC, which closely monitors customer trends and market demand, is aggressively expanding 2nm, CoWoS, and SoIC capacity—evidence that AI demand has not yet peaked.\n\nTSMC previously projected that revenue from AI accelerators will grow at a compound annual growth rate (CAGR) of nearly 45% from 2024 over the next five years. According to supply chain estimates, TSMC’s total CoWoS capacity in 2025 will reach 650,000 wafers, with NVIDIA securing over 370,000 wafers—more than half. ASIC clients are also ramping up, and by 2026, CoWoS capacity is expected to reach 730,000 wafers. From 2025 to 2027, ASIC chip shipments are forecast to grow more than 50% annually.\n\nMeta, xAI, and others will begin bringing ASIC server capacity online in the coming years, with expectations of a major ramp-up in 2026. AWS, Google, and Microsoft are also advancing their AI server platforms. As a result, the supply chain anticipates market expansion, unlocking even more opportunities for collaboration across the\n\n---\n\nhttps://t.co/NJACSC3oHQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00301505360250931, "ret_signed_m1d": 0.00301505360250931, "alpha_spy_m1d": 0.013179653390970048, "alpha_c_m1d": 0.008673279463507377, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010050240932224286, "ret_signed_p1d": 0.010050240932224286, "alpha_spy_p1d": 0.009148958044080002, "alpha_c_p1d": -0.006686905751161154, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.03956843064744797, "ret_signed_p1w": 0.03956843064744797, "alpha_spy_p1w": 0.043140177536053925, "alpha_c_p1w": 0.022712579077975636, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.09003292845197919, "ret_signed_p1m": 0.09003292845197919, "alpha_spy_p1m": 0.05101882703781935, "alpha_c_p1m": -0.016919204892853257, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.17077614984210632, "ret_signed_p3m": 0.17077614984210632, "alpha_spy_p3m": 0.09315979424185472, "alpha_c_p3m": 0.03743228649710928, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.42879100079745247, "ret_signed_p6m": 0.42879100079745247, "alpha_spy_p6m": 0.2868664058312209, "alpha_c_p6m": 0.03220173133154702, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.0286, -0.0116, -0.0347, -0.0407, -0.0296, -0.0241, -0.0432, -0.0101, -0.0231, -0.0231, -0.005, -0.0151, -0.0372, -0.0432, -0.0854, -0.0513, -0.0533, -0.0533, -0.0533, -0.1477, -0.1799, -0.2111, -0.1327, -0.1065, -0.1307, -0.1186, -0.1407, -0.1487, -0.1457, -0.1608, -0.1799, -0.1226, -0.1317, -0.1075, -0.0985, -0.0935, -0.0874, -0.0874, -0.0452, -0.0573, -0.0754, -0.0673, -0.0774, -0.0462, -0.0382, -0.0261, 0.004, -0.002, 0.003, -0.0111, -0.0171, -0.005, -0.0131, -0.0111, -0.0191, -0.0302, -0.0281, -0.0281, -0.0281, -0.0492, -0.0452, -0.005, 0.003, 0.0, 0.0101, 0.0503, 0.0704, 0.0547, 0.0396, 0.0345, 0.0547, 0.0648, 0.0446, 0.0648, 0.0295, 0.0598, 0.0799, 0.085, 0.09, 0.0698, 0.0951, 0.0951, 0.1001, 0.0951, 0.09, 0.09, 0.1001, 0.1102, 0.1102, 0.1052, 0.1203, 0.1405, 0.1405, 0.1657, 0.1607, 0.1405, 0.1556, 0.1556, 0.1556, 0.1556, 0.1455, 0.1657, 0.1708, 0.1708, 0.1455, 0.1607, 0.1355, 0.191, 0.1859, 0.191, 0.191, 0.2111, 0.1859, 0.191, 0.191, 0.196, 0.1455, 0.1607, 0.1455, 0.1809, 0.1859, 0.2011, 0.1708, 0.1708, 0.1758, 0.1708, 0.1708, 0.1708, 0.191, 0.191, 0.2111, 0.2364, 0.2515, 0.2717, 0.2667, 0.2971, 0.2819, 0.3021, 0.2819, 0.3123, 0.3579, 0.3579, 0.3376, 0.3173, 0.3173, 0.3224, 0.3427, 0.3832, 0.4187, 0.4187, 0.4541, 0.4339, 0.4592, 0.4592, 0.4339, 0.444, 0.4845, 0.5048, 0.4693, 0.4997, 0.4997, 0.4795, 0.4693, 0.4693, 0.4997, 0.4947, 0.5251, 0.5251, 0.52, 0.5301, 0.5251, 0.4795, 0.4845, 0.4795, 0.4947, 0.4845, 0.4947, 0.4795, 0.4491, 0.4643, 0.4237, 0.4136, 0.4744, 0.4035, 0.3933, 0.4339, 0.4592, 0.4541, 0.4592, 0.4288]}
{"unit_id": "thread:1930815488038511069", "ticker_idx": 1, "ts": "2025-06-06T02:34:19+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Leading players like NVIDIA and Broadcom are expected to see strong revenue growth", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC expected to hit record highs in 2025–26 on AI monopoly; NVDA and AVGO also seen posting strong revenue growth as GPU and ASIC shipments expand 50%+ annually.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "ASICs “Ready to Duel” with AI GPUs – TSMC Dominates Orders and Raises Prices -Digitimes\n\nWith NVIDIA holding a near-monopoly in the AI GPU battlefield, numerous companies are accelerating the development and mass production of ASICs (application-specific integrated circuits), shaping an industry narrative of an impending “showdown” between NVIDIA GPUs and an expanding ASIC army. Notably, TSMC, which serves both camps, has raised prices for its 2nm and advanced packaging services, helping it grow revenue insulated from tariff and currency fluctuations. The market expects TSMC to continue hitting record highs in 2025 and 2026.\n\nA year ago, NVIDIA CEO Jensen Huang openly declared plans to enter the ASIC battlefield. However, he has since pivoted, now advocating “collaboration over competition” to grow the AI pie together.\n\nAccording to supply chain sources, AI demand is expected to remain robust over the long term. Shipments of both AI GPUs and ASICs are projected to grow steadily over the next three years. For ASICs in particular, shipments are conservatively estimated to grow at an annual rate of at least 50%. Leading players like NVIDIA and Broadcom are expected to see strong revenue growth, and TSMC—supplier to both—is profiting handsomely.\n\nThe rise of generative AI has ushered in the AI era and driven explosive demand for AI servers. NVIDIA, with its combination of GPUs and powerful CUDA software, quickly seized control of the ecosystem, securing dominance from customers to the broader supply chain, with everyone scrambling to collaborate.\n\nStill, supply chain insiders note that while NVIDIA’s AI GPUs are widely used for training and inference, demand has consistently outstripped supply since the boom in 2023. Moreover, the high prices and closed CUDA ecosystem have put cloud service providers (CSPs) and chipmakers at a disadvantage in price negotiations. To mitigate risk and ultimately “de-NVIDIA” their infrastructure, many are accelerating the development of in-house ASICs.\n\nSources explain that the collective goal of the ASIC players is to reduce dependency on NVIDIA’s expensive GPUs and CUDA stack. NVIDIA’s H100 and latest B200 chips cost tens of thousands of dollars each, and complete server systems can run into the millions. Though self-developing ASICs involves high initial costs, scaling production can dramatically reduce unit costs.\n\nThat’s why major NVIDIA clients like Google, AWS, Meta, and Microsoft are not only buying NVIDIA GPUs but also aggressively ramping up their in-house chip programs. They're placing direct wafer orders with fabs to avoid capacity constraints.\n\nMeanwhile, Broadcom—NVIDIA’s primary rival in the ASIC space—has also posted surprisingly strong growth.\n\nSupply chain sources note that while Jensen Huang initially announced an ASIC division to expand the customer base, he later pivoted to unveil the NVLink Fusion strategy. This strategy aims to establish a vast partner ecosystem built around NVLink, enabling semi-custom AI infrastructure development across industries.\n\nThis ecosystem includes companies like MediaTek, Marvell, SiFive, Astera Labs, Synopsys, Cadence, Fujitsu, and Qualcomm—but notably excludes Broadcom, raising speculation about the future dynamic between NVIDIA and Broadcom.\n\nOn the competition between ASICs and AI GPUs, TSMC Chairman Mark Liu plainly stated:\n\n“It doesn’t matter who wins—NVIDIA and the ASIC makers are all my customers. They all manufacture with TSMC.”\n\nAccording to supply chain insiders, with Samsung Electronics and Intel far behind TSMC in the sub-7nm race, TSMC has virtually monopolized major orders from 7nm down to the upcoming 2nm node. AI clients are choosing TSMC for its full-stack capabilities, including CoWoS advanced packaging.\n\nTSMC, which closely monitors customer trends and market demand, is aggressively expanding 2nm, CoWoS, and SoIC capacity—evidence that AI demand has not yet peaked.\n\nTSMC previously projected that revenue from AI accelerators will grow at a compound annual growth rate (CAGR) of nearly 45% from 2024 over the next five years. According to supply chain estimates, TSMC’s total CoWoS capacity in 2025 will reach 650,000 wafers, with NVIDIA securing over 370,000 wafers—more than half. ASIC clients are also ramping up, and by 2026, CoWoS capacity is expected to reach 730,000 wafers. From 2025 to 2027, ASIC chip shipments are forecast to grow more than 50% annually.\n\nMeta, xAI, and others will begin bringing ASIC server capacity online in the coming years, with expectations of a major ramp-up in 2026. AWS, Google, and Microsoft are also advancing their AI server platforms. As a result, the supply chain anticipates market expansion, unlocking even more opportunities for collaboration across the\n\n---\n\nhttps://t.co/NJACSC3oHQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012207001118832439, "ret_signed_m1d": -0.012207001118832439, "alpha_spy_m1d": -0.0020424013303717015, "alpha_c_m1d": -0.006548775257834372, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0064212515479658006, "ret_signed_p1d": 0.0064212515479658006, "alpha_spy_p1d": 0.005519968659821517, "alpha_c_p1d": -0.01031589513541964, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.001833597129654807, "ret_signed_p1w": 0.001833597129654807, "alpha_spy_p1w": 0.00540534401826076, "alpha_c_p1w": -0.015022254439817528, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.11664557480728432, "ret_signed_p1m": 0.11664557480728432, "alpha_spy_p1m": 0.07763147339312448, "alpha_c_p1m": 0.00969344146245188, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.2040070117374555, "ret_signed_p3m": 0.2040070117374555, "alpha_spy_p3m": 0.1263906561372039, "alpha_c_p3m": 0.07066314839245846, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.2697054882881218, "ret_signed_p6m": 0.2697054882881218, "alpha_spy_p6m": 0.12778089332189024, "alpha_c_p6m": -0.12688378117778365, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.2326, -0.1833, -0.1844, -0.1415, -0.1566, -0.1855, -0.1708, -0.1636, -0.1695, -0.1433, -0.1484, -0.1973, -0.2137, -0.2262, -0.2353, -0.2228, -0.2209, -0.2817, -0.3345, -0.311, -0.3205, -0.1933, -0.241, -0.2173, -0.2188, -0.2083, -0.2627, -0.2839, -0.2839, -0.3162, -0.3022, -0.2753, -0.249, -0.2167, -0.2328, -0.2307, -0.2314, -0.2125, -0.1921, -0.1969, -0.1988, -0.174, -0.1718, -0.1769, -0.1321, -0.0832, -0.045, -0.0486, -0.0446, -0.0434, -0.0518, -0.07, -0.0627, -0.0736, -0.0736, -0.0439, -0.0488, -0.0179, -0.0465, -0.0306, -0.0035, 0.0014, -0.0122, 0.0, 0.0064, 0.0158, 0.0079, 0.0232, 0.0018, 0.021, 0.017, 0.0266, 0.0266, 0.0151, 0.0174, 0.0437, 0.0889, 0.0939, 0.1132, 0.1149, 0.0818, 0.1097, 0.1244, 0.1244, 0.1166, 0.1291, 0.1494, 0.158, 0.1638, 0.1578, 0.2046, 0.2093, 0.2208, 0.2166, 0.2094, 0.1787, 0.2051, 0.226, 0.2243, 0.2473, 0.2385, 0.265, 0.2552, 0.2259, 0.2702, 0.2579, 0.2661, 0.2756, 0.2893, 0.2847, 0.2925, 0.2814, 0.2845, 0.2734, 0.2844, 0.2394, 0.2377, 0.2348, 0.256, 0.2689, 0.2827, 0.2815, 0.2714, 0.2291, 0.2291, 0.2051, 0.204, 0.2113, 0.1786, 0.1877, 0.205, 0.2514, 0.2503, 0.2549, 0.2544, 0.2341, 0.2017, 0.2437, 0.2468, 0.2957, 0.2592, 0.2489, 0.254, 0.2575, 0.2833, 0.3167, 0.3214, 0.333, 0.324, 0.3094, 0.3058, 0.3346, 0.359, 0.2926, 0.329, 0.2705, 0.2691, 0.283, 0.293, 0.2889, 0.2785, 0.2722, 0.2855, 0.3144, 0.3514, 0.4187, 0.4611, 0.4318, 0.429, 0.46, 0.4022, 0.3776, 0.3273, 0.3278, 0.4047, 0.3631, 0.3677, 0.3187, 0.342, 0.3168, 0.2799, 0.3163, 0.2748, 0.2624, 0.2883, 0.2549, 0.2721, 0.2721, 0.2491, 0.2697]}
{"unit_id": "thread:1930815488038511069", "ticker_idx": 2, "ts": "2025-06-06T02:34:19+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Leading players like NVIDIA and Broadcom are expected to see strong revenue growth", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC expected to hit record highs in 2025–26 on AI monopoly; NVDA and AVGO also seen posting strong revenue growth as GPU and ASIC shipments expand 50%+ annually.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "ASICs “Ready to Duel” with AI GPUs – TSMC Dominates Orders and Raises Prices -Digitimes\n\nWith NVIDIA holding a near-monopoly in the AI GPU battlefield, numerous companies are accelerating the development and mass production of ASICs (application-specific integrated circuits), shaping an industry narrative of an impending “showdown” between NVIDIA GPUs and an expanding ASIC army. Notably, TSMC, which serves both camps, has raised prices for its 2nm and advanced packaging services, helping it grow revenue insulated from tariff and currency fluctuations. The market expects TSMC to continue hitting record highs in 2025 and 2026.\n\nA year ago, NVIDIA CEO Jensen Huang openly declared plans to enter the ASIC battlefield. However, he has since pivoted, now advocating “collaboration over competition” to grow the AI pie together.\n\nAccording to supply chain sources, AI demand is expected to remain robust over the long term. Shipments of both AI GPUs and ASICs are projected to grow steadily over the next three years. For ASICs in particular, shipments are conservatively estimated to grow at an annual rate of at least 50%. Leading players like NVIDIA and Broadcom are expected to see strong revenue growth, and TSMC—supplier to both—is profiting handsomely.\n\nThe rise of generative AI has ushered in the AI era and driven explosive demand for AI servers. NVIDIA, with its combination of GPUs and powerful CUDA software, quickly seized control of the ecosystem, securing dominance from customers to the broader supply chain, with everyone scrambling to collaborate.\n\nStill, supply chain insiders note that while NVIDIA’s AI GPUs are widely used for training and inference, demand has consistently outstripped supply since the boom in 2023. Moreover, the high prices and closed CUDA ecosystem have put cloud service providers (CSPs) and chipmakers at a disadvantage in price negotiations. To mitigate risk and ultimately “de-NVIDIA” their infrastructure, many are accelerating the development of in-house ASICs.\n\nSources explain that the collective goal of the ASIC players is to reduce dependency on NVIDIA’s expensive GPUs and CUDA stack. NVIDIA’s H100 and latest B200 chips cost tens of thousands of dollars each, and complete server systems can run into the millions. Though self-developing ASICs involves high initial costs, scaling production can dramatically reduce unit costs.\n\nThat’s why major NVIDIA clients like Google, AWS, Meta, and Microsoft are not only buying NVIDIA GPUs but also aggressively ramping up their in-house chip programs. They're placing direct wafer orders with fabs to avoid capacity constraints.\n\nMeanwhile, Broadcom—NVIDIA’s primary rival in the ASIC space—has also posted surprisingly strong growth.\n\nSupply chain sources note that while Jensen Huang initially announced an ASIC division to expand the customer base, he later pivoted to unveil the NVLink Fusion strategy. This strategy aims to establish a vast partner ecosystem built around NVLink, enabling semi-custom AI infrastructure development across industries.\n\nThis ecosystem includes companies like MediaTek, Marvell, SiFive, Astera Labs, Synopsys, Cadence, Fujitsu, and Qualcomm—but notably excludes Broadcom, raising speculation about the future dynamic between NVIDIA and Broadcom.\n\nOn the competition between ASICs and AI GPUs, TSMC Chairman Mark Liu plainly stated:\n\n“It doesn’t matter who wins—NVIDIA and the ASIC makers are all my customers. They all manufacture with TSMC.”\n\nAccording to supply chain insiders, with Samsung Electronics and Intel far behind TSMC in the sub-7nm race, TSMC has virtually monopolized major orders from 7nm down to the upcoming 2nm node. AI clients are choosing TSMC for its full-stack capabilities, including CoWoS advanced packaging.\n\nTSMC, which closely monitors customer trends and market demand, is aggressively expanding 2nm, CoWoS, and SoIC capacity—evidence that AI demand has not yet peaked.\n\nTSMC previously projected that revenue from AI accelerators will grow at a compound annual growth rate (CAGR) of nearly 45% from 2024 over the next five years. According to supply chain estimates, TSMC’s total CoWoS capacity in 2025 will reach 650,000 wafers, with NVIDIA securing over 370,000 wafers—more than half. ASIC clients are also ramping up, and by 2026, CoWoS capacity is expected to reach 730,000 wafers. From 2025 to 2027, ASIC chip shipments are forecast to grow more than 50% annually.\n\nMeta, xAI, and others will begin bringing ASIC server capacity online in the coming years, with expectations of a major ramp-up in 2026. AWS, Google, and Microsoft are also advancing their AI server platforms. As a result, the supply chain anticipates market expansion, unlocking even more opportunities for collaboration across the\n\n---\n\nhttps://t.co/NJACSC3oHQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.052646542785388384, "ret_signed_m1d": 0.052646542785388384, "alpha_spy_m1d": 0.06281114257384912, "alpha_c_m1d": 0.05830476864638645, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010731753810176325, "ret_signed_p1d": -0.010731753810176325, "alpha_spy_p1d": -0.011633036698320609, "alpha_c_p1d": -0.027468900493561765, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.007168124310281021, "ret_signed_p1w": 0.007168124310281021, "alpha_spy_p1w": 0.010739871198886974, "alpha_c_p1w": -0.009687727259191314, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.11296872464411822, "ret_signed_p1m": 0.11296872464411822, "alpha_spy_p1m": 0.07395462322995838, "alpha_c_p1m": 0.006016591299285778, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.22748067541138628, "ret_signed_p3m": 0.22748067541138628, "alpha_spy_p3m": 0.14986431981113468, "alpha_c_p3m": 0.09413681206638924, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.569885380902382, "ret_signed_p6m": 0.569885380902382, "alpha_spy_p6m": 0.4279607859361505, "alpha_c_p6m": 0.1732961114364766, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.2325, -0.2158, -0.2274, -0.2105, -0.2147, -0.2382, -0.2104, -0.2284, -0.2238, -0.2255, -0.2376, -0.274, -0.3035, -0.3151, -0.322, -0.3175, -0.3031, -0.3763, -0.4076, -0.3758, -0.3681, -0.2502, -0.3022, -0.2632, -0.2777, -0.2753, -0.2929, -0.3075, -0.3075, -0.3269, -0.3132, -0.2836, -0.238, -0.2212, -0.2205, -0.2258, -0.2205, -0.2009, -0.1753, -0.1871, -0.1897, -0.1706, -0.1586, -0.1568, -0.1027, -0.0588, -0.06, -0.0579, -0.0742, -0.066, -0.0618, -0.0697, -0.0664, -0.0737, -0.0737, -0.0457, -0.0304, -0.0201, -0.0197, 0.0072, 0.0402, 0.0573, 0.0526, 0.0, -0.0107, -0.0093, 0.0242, 0.037, 0.0072, 0.0209, 0.0099, 0.0175, 0.0175, 0.0148, 0.0301, 0.0707, 0.0743, 0.0967, 0.0934, 0.1189, 0.0746, 0.0956, 0.117, 0.117, 0.113, 0.1033, 0.1281, 0.1179, 0.1138, 0.1187, 0.1404, 0.1399, 0.1628, 0.1502, 0.1699, 0.1309, 0.1516, 0.1719, 0.1779, 0.1946, 0.2073, 0.2284, 0.1922, 0.1717, 0.2085, 0.1891, 0.2246, 0.233, 0.238, 0.2336, 0.2699, 0.2547, 0.2634, 0.2435, 0.2412, 0.1971, 0.1819, 0.1756, 0.1934, 0.1944, 0.2097, 0.2188, 0.2529, 0.2072, 0.2072, 0.2106, 0.2275, 0.2425, 0.3594, 0.4031, 0.3666, 0.5002, 0.4598, 0.4608, 0.4779, 0.4613, 0.4052, 0.4019, 0.4002, 0.3776, 0.3782, 0.3797, 0.3667, 0.3603, 0.3333, 0.3415, 0.3556, 0.3751, 0.3759, 0.3642, 0.3679, 0.4049, 0.4029, 0.32, 0.4504, 0.3993, 0.4286, 0.4401, 0.4205, 0.4201, 0.3933, 0.3837, 0.4, 0.44, 0.4722, 0.5166, 0.5695, 0.5308, 0.503, 0.4742, 0.4311, 0.4597, 0.4459, 0.4209, 0.4573, 0.4311, 0.4444, 0.3824, 0.3925, 0.3933, 0.3845, 0.4411, 0.4102, 0.3833, 0.5369, 0.5656, 0.6166, 0.6166, 0.6385, 0.5699]}
{"unit_id": "orig:1932924341320364460", "ticker_idx": 0, "ts": "2025-06-11T22:14:09+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we reiterate our Buy ratings on Hynix and Samsung", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi reiteration of Buy on Hynix and Samsung on memory supply shortage intensifying in 2H25 from GDDR7/LPDDR5X demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi on Memory: \n\nExpect Supply Shortage to Intensify in 2H25E Due to Strong Demand from GDDR7 and LPDDR5X: We expect the global memory supply shortage to further intensify due to:\n\n[1] rising demand for GDDR7 on aggressive roll-out of AI Inference models and \n\n[2] Apple’s upcoming 12GB DRAM upgrade for the iPhone 17 in 2H25E. \n\nWith the anticipation of edge AI devices emerging as a new driver for AI DRAM sales and the increase of mobile DRAM content materializing in 2H25E, we reiterate our Buy ratings on Hynix and Samsung.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03958327626533886, "ret_signed_m1d": -0.03958327626533886, "alpha_spy_m1d": -0.04244334753511192, "alpha_c_m1d": -0.04030884529697254, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0007255690316336771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0187499007098616, "ret_signed_p1d": -0.0187499007098616, "alpha_spy_p1d": -0.022724126807016076, "alpha_c_p1d": -0.02359941848758862, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.004849517777727019, "ret_p1w": 0.027083342458764426, "ret_signed_p1w": 0.027083342458764426, "alpha_spy_p1w": 0.03360186167688295, "alpha_c_p1w": 0.02437219012258285, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.002711152336181577, "ret_p1m": 0.23750011522559267, "ret_signed_p1m": 0.23750011522559267, "alpha_spy_p1m": 0.1937491348350462, "alpha_c_p1m": 0.13959271616320978, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.09790739906238288, "ret_p3m": 0.15583383743849155, "ret_signed_p3m": 0.15583383743849155, "alpha_spy_p3m": 0.0737065726849997, "alpha_c_p3m": 0.02355957538381137, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.13227426205468018, "ret_p6m": 1.2632152640928092, "ret_signed_p6m": 1.2632152640928092, "alpha_spy_p6m": 1.1186159738339696, "alpha_c_p6m": 0.8824301602927187, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.38078510380009045, "price_path": [-0.1495, -0.1432, -0.1557, -0.1453, -0.1266, -0.1037, -0.1203, -0.1349, -0.1099, -0.1391, -0.1711, -0.2068, -0.1806, -0.1769, -0.1906, -0.2422, -0.3146, -0.295, -0.3137, -0.238, -0.248, -0.2505, -0.2489, -0.2763, -0.2721, -0.2721, -0.2655, -0.2771, -0.2472, -0.2584, -0.2331, -0.243, -0.248, -0.2617, -0.2617, -0.2264, -0.2264, -0.2264, -0.2064, -0.2085, -0.2093, -0.189, -0.1744, -0.1432, -0.1661, -0.1495, -0.1707, -0.1599, -0.1661, -0.1811, -0.1682, -0.1557, -0.1578, -0.1349, -0.1167, -0.1479, -0.1354, -0.1354, -0.0937, -0.0646, -0.0646, -0.0458, -0.0396, 0.0, -0.0187, -0.0187, 0.0333, 0.0375, 0.0271, 0.025, 0.0708, 0.0813, 0.1604, 0.1917, 0.2208, 0.1833, 0.2167, 0.1896, 0.1625, 0.1604, 0.1271, 0.1292, 0.175, 0.1708, 0.2375, 0.2271, 0.25, 0.2438, 0.2333, 0.1229, 0.1208, 0.1354, 0.1187, 0.1208, 0.1229, 0.1083, 0.0917, 0.0938, 0.0979, 0.1396, 0.075, 0.075, 0.0979, 0.0771, 0.0917, 0.0687, 0.1125, 0.1208, 0.1583, 0.1521, 0.1521, 0.1146, 0.0958, 0.0646, 0.0208, 0.0458, 0.0813, 0.0896, 0.0833, 0.1204, 0.1225, 0.0682, 0.087, 0.0953, 0.1078, 0.1412, 0.1558, 0.2017, 0.2685, 0.281, 0.3707, 0.3812, 0.4521, 0.3916, 0.4834, 0.4834, 0.4646, 0.5063, 0.4917, 0.4876, 0.4041, 0.4563, 0.45, 0.5022, 0.6503, 0.6503, 0.6503, 0.6503, 0.6503, 0.6503, 0.7859, 0.7317, 0.7171, 0.763, 0.8881, 0.9424, 1.0258, 0.9987, 1.0091, 0.9966, 1.1281, 1.2324, 1.174, 1.3284, 1.3701, 1.3325, 1.5871, 1.4452, 1.416, 1.4744, 1.4202, 1.5286, 1.5829, 1.5745, 1.5537, 1.3367, 1.5286, 1.3784, 1.345, 1.3826, 1.174, 1.1698, 1.1656, 1.1865, 1.2716, 1.2131, 1.2465, 1.33, 1.305, 1.2632]}
{"unit_id": "orig:1932924341320364460", "ticker_idx": 1, "ts": "2025-06-11T22:14:09+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we reiterate our Buy ratings on Hynix and Samsung", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi reiteration of Buy on Hynix and Samsung on memory supply shortage intensifying in 2H25 from GDDR7/LPDDR5X demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Citi on Memory: \n\nExpect Supply Shortage to Intensify in 2H25E Due to Strong Demand from GDDR7 and LPDDR5X: We expect the global memory supply shortage to further intensify due to:\n\n[1] rising demand for GDDR7 on aggressive roll-out of AI Inference models and \n\n[2] Apple’s upcoming 12GB DRAM upgrade for the iPhone 17 in 2H25E. \n\nWith the anticipation of edge AI devices emerging as a new driver for AI DRAM sales and the increase of mobile DRAM content materializing in 2H25E, we reiterate our Buy ratings on Hynix and Samsung.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011686266327631145, "ret_signed_m1d": -0.011686266327631145, "alpha_spy_m1d": -0.014546337597404202, "alpha_c_m1d": -0.013891494346880195, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0022052280192490503, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006677847512062796, "ret_signed_p1d": -0.006677847512062796, "alpha_spy_p1d": -0.010652073609217272, "alpha_c_p1d": -0.015956444809025472, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.009278597296962676, "ret_p1w": -0.0016694950595369518, "ret_signed_p1w": -0.0016694950595369518, "alpha_spy_p1w": 0.004849024158581572, "alpha_c_p1w": -0.006953690572089166, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.005284195512552214, "ret_p1m": 0.02461026976045866, "ret_signed_p1m": 0.02461026976045866, "alpha_spy_p1m": -0.01914071063008782, "alpha_c_p1m": -0.046397216853621526, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.07100748661408018, "ret_p3m": 0.177462075512117, "ret_signed_p3m": 0.177462075512117, "alpha_spy_p3m": 0.09533481075862515, "alpha_c_p3m": 0.07373846017432428, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.10372361533779273, "ret_p6m": 0.7732293875726444, "ret_signed_p6m": 0.7732293875726444, "alpha_spy_p6m": 0.6286300973138048, "alpha_c_p6m": 0.5585870872804422, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.21464230029220221, "price_path": [-0.0922, -0.0441, -0.0441, -0.0291, -0.0009, 0.024, 0.0041, -0.0076, 0.019, 0.0256, 0.005, -0.0351, -0.0184, -0.0184, -0.0384, -0.0634, -0.1119, -0.1068, -0.1152, -0.0584, -0.0785, -0.0618, -0.0551, -0.0868, -0.0801, -0.0768, -0.0751, -0.0818, -0.0701, -0.0701, -0.0701, -0.0684, -0.0684, -0.0735, -0.0735, -0.0935, -0.0935, -0.0935, -0.0885, -0.0885, -0.0851, -0.0384, -0.0501, -0.0417, -0.0434, -0.0518, -0.0684, -0.0668, -0.0701, -0.0868, -0.0952, -0.0868, -0.1002, -0.0668, -0.0634, -0.0618, -0.0518, -0.0518, -0.0351, -0.0134, -0.0134, -0.0017, -0.0117, 0.0, -0.0067, -0.0267, -0.0451, -0.0301, -0.0017, -0.0117, -0.0067, -0.0317, 0.01, 0.0234, 0.005, 0.0213, 0.0045, 0.0112, 0.0213, 0.0716, 0.0632, 0.0364, 0.0313, 0.0145, 0.0246, 0.0515, 0.0498, 0.07, 0.0868, 0.1204, 0.1271, 0.1388, 0.1086, 0.1153, 0.1086, 0.1069, 0.1825, 0.1859, 0.2195, 0.1993, 0.1573, 0.1707, 0.1741, 0.1556, 0.1842, 0.206, 0.1926, 0.1943, 0.2077, 0.2027, 0.2027, 0.1758, 0.1758, 0.1842, 0.1859, 0.1993, 0.201, 0.1808, 0.1859, 0.1691, 0.1707, 0.1355, 0.1607, 0.1724, 0.1775, 0.1674, 0.1775, 0.201, 0.2195, 0.2329, 0.2665, 0.285, 0.3337, 0.3135, 0.3488, 0.3488, 0.4025, 0.4227, 0.4345, 0.4462, 0.3992, 0.4206, 0.4155, 0.451, 0.5142, 0.5142, 0.5142, 0.5142, 0.5142, 0.5142, 0.5927, 0.5741, 0.5455, 0.6028, 0.6484, 0.6518, 0.6551, 0.645, 0.6636, 0.6281, 0.6669, 0.7209, 0.6787, 0.6956, 0.7564, 0.8137, 0.8745, 0.7699, 0.6973, 0.6737, 0.6518, 0.6973, 0.7462, 0.7395, 0.7344, 0.6399, 0.6973, 0.6501, 0.6281, 0.6973, 0.5994, 0.6315, 0.6754, 0.7344, 0.7462, 0.6956, 0.7007, 0.7445, 0.7631, 0.7732]}
{"unit_id": "orig:1938018416247677136", "ticker_idx": 0, "ts": "2025-06-25T23:36:11+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Annualized revenue is over $6B. Currently, we expect to achieve this criterion faster. The yield improvement of 12H is particularly contributing significantly. Customer inventory levels are healthy. The demand environment is positive, and the impact of tariffs is considered limited. The Q4 guidance is entirely due to Micron's strong execution and AI growth.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: HBM TAM doubling to $35B, market share gains accelerating, healthy inventory, strong Q4 execution, and positive AI-driven demand outlook.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron FY3Q25 Key Q&A\n\nQ: HBM is expected to remain a market that is 15-20% the size of the overall accelerator (GPU, ASIC) market. Can the HBM TAM expand faster than the accelerator market?\n\nA: HBM TAM grew from $18B last year to $35B this year (CY2025). The HBM bit demand growth rate in 2026 will significantly exceed the overall DRAM bit growth rate. After 2027, technological advancements such as HBM4E and custom HBM are expected.\n\nQ: What is the approximate revenue scale of HBM?\n\nA: Annualized revenue is over $6B. Previously, we stated that we would achieve a market share similar to our DRAM market share (23-24%) by year-end. Currently, we expect to achieve this criterion faster. The yield improvement of 12H is particularly contributing significantly.\n\nQ: At the earnings announcement in June of last year, all 2025 HBM volumes were sold out. How are negotiations for 2026 HBM supply progressing now? Is customers' 2026 HBM demand forecast exceeding our supply capacity?\n\nA: We are focusing on preparing and executing to meet customer demand for 2026.\n\nQ: HBM4 TSV count has doubled. What is the approximate trade ratio of HBM4?\n\nA: HBM4 trade ratio is greater than 3, while HBM3E is approximately 3.\n\nQ: Does HBM bit demand or margin differ by processor customer?\n\nA: Both GPUs and ASICs are high-value-creating products. Both require capacities around 200-288GB. This figure is expected to increase further with the transition to 12-stack in the future. We don't get into the specifics or differentiating the two here regarding clear margin differences.\n\nQ: What is the progress of HBM4 certification?\n\nA: HBM4 products have been sampled to multiple customers and are progressing satisfactorily. There are strengths in both performance and power efficiency. We will be prepared to meet customer demand in 2026. It is manufactured based on the sufficiently verified 1-beta process node.\n\nQ: Capex is currently at about 35% of revenue. Is this level still reasonable in 2026 when the HBM proportion increases?\n\nA: Capex is expected to increase slightly in FY4Q, but the annual forecast of $14B Capex remains. Investments in new fab construction and new process node equipment installation will continue. The timing may be somewhat irregular.\n\nQ: What are the current NAND utilization rate and next year's outlook?\n\nA: Advanced process-based NAND production lines are operating at full capacity. NAND capacity is expected to decrease by approximately 10% by CY3Q25. Overall capacity reduction is expected to lead to higher utilization rates.\n\nQ: What is the reason for the overall margin increase, and can this be seen as a new baseline?\n\nA: Prices have fallen but sales were at better-than-expected prices. 1) Price effect in the current quarter, 2) Inventory reduction effect, and 3) Product mix improvement led to an upward revision of the next quarter's price baseline. Overall, the market is still favorable.\n\nQ: What is your perception of overall customer inventory levels and the outlook for bit demand in the second half?\n\nA: Customer inventory levels are healthy. Some pull-in may have occurred but is deemed minimal. The demand environment is positive, and the impact of tariffs is considered limited. The Q4 guidance is entirely due to Micron's strong execution and AI growth.\n\nQ: The correlation between AI server and SSD demand.\n\nA: Last year, SSD demand was strong, but from the end of last year to early this year, customers were undergoing inventory adjustments in the data center SSD market. The second half is expected to be better than the first half. We plan to further strengthen our SSD business by shifting to a portfolio focused on high-value NAND.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005186703787299907, "ret_signed_m1d": 0.005186703787299907, "alpha_spy_m1d": 0.005746754585245539, "alpha_c_m1d": 0.018486137889198506, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009823196998103967, "ret_signed_p1d": -0.009823196998103967, "alpha_spy_p1d": -0.017646976872591313, "alpha_c_p1d": -0.017090671033344718, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": -0.04330064082854557, "ret_signed_p1w": -0.04330064082854557, "alpha_spy_p1w": -0.06525678147023772, "alpha_c_p1w": -0.06528460239557365, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": -0.12113815880729972, "ret_signed_p1m": -0.12113815880729972, "alpha_spy_p1m": -0.16610454004474628, "alpha_c_p1m": -0.1669230338782275, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.2948915937920622, "ret_signed_p3m": 0.2948915937920622, "alpha_spy_p3m": 0.19348042596955572, "alpha_c_p3m": 0.12421568967381957, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.9563045658042233, "ret_signed_p6m": 0.9563045658042233, "alpha_spy_p6m": 0.838987711996035, "alpha_c_p6m": 0.6944949622109877, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3059, -0.3172, -0.3029, -0.3037, -0.4158, -0.4914, -0.4627, -0.485, -0.3881, -0.4495, -0.4534, -0.4419, -0.4417, -0.4552, -0.4593, -0.4593, -0.4755, -0.4483, -0.4269, -0.3916, -0.373, -0.3826, -0.3958, -0.3953, -0.3888, -0.3657, -0.368, -0.3673, -0.3507, -0.3308, -0.3253, -0.2747, -0.2383, -0.2509, -0.2499, -0.2299, -0.2248, -0.2291, -0.2468, -0.2548, -0.2662, -0.2662, -0.2426, -0.2442, -0.2393, -0.2577, -0.2284, -0.1965, -0.1886, -0.1647, -0.1469, -0.1281, -0.103, -0.0882, -0.087, -0.0916, -0.0582, -0.0543, -0.0427, -0.0427, -0.0287, -0.0406, 0.0052, 0.0, -0.0098, -0.0196, -0.0314, -0.05, -0.0433, -0.039, -0.039, -0.0567, -0.0213, -0.0385, -0.0316, -0.0205, -0.067, -0.0552, -0.0842, -0.1091, -0.1002, -0.1093, -0.1409, -0.1361, -0.1211, -0.1248, -0.1249, -0.1193, -0.0975, -0.1415, -0.175, -0.1523, -0.1421, -0.1443, -0.12, -0.0648, -0.0268, 0.0049, -0.0225, -0.0145, -0.0492, -0.0282, -0.04, -0.078, -0.0892, -0.0743, -0.0842, -0.0836, -0.0738, -0.0404, -0.0639, -0.0639, -0.068, -0.0662, -0.023, 0.0333, 0.0341, 0.0638, 0.1012, 0.1844, 0.2368, 0.241, 0.2493, 0.2585, 0.3285, 0.28, 0.2949, 0.309, 0.272, 0.2336, 0.2371, 0.2892, 0.3161, 0.4328, 0.4454, 0.4784, 0.503, 0.4615, 0.5469, 0.5138, 0.4293, 0.5173, 0.4723, 0.5107, 0.5941, 0.5929, 0.6275, 0.5922, 0.5621, 0.627, 0.7239, 0.7324, 0.7466, 0.7838, 0.7632, 0.7613, 0.8473, 0.7161, 0.8693, 0.8759, 0.8726, 0.9937, 0.8977, 0.9276, 0.865, 0.9428, 0.9044, 0.7985, 0.7782, 0.585, 0.6322, 0.7625, 0.7672, 0.8123, 0.8123, 0.8613, 0.8926, 0.885, 0.843, 0.7839, 0.8671, 0.9435, 0.9868, 1.0756, 1.0343, 0.898, 0.8693, 0.8301, 0.775, 0.9563]}
{"unit_id": "orig:1927519271128535506", "ticker_idx": 0, "ts": "2025-05-28T00:16:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley maintains its 'Top Pick' rating on Nvidia, as GB200 ramp improvements and potential H20 license reinstatement could reignite momentum in the second half of the year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Top Pick on NVDA; expects H2 reacceleration on GB200 ramp + potential H20 license reinstatement despite near-term headwinds.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: H20 Ban and Supply Bottlenecks Impacting Earnings\nMorgan Stanley (J. Moore, May 27, 2025)\n\nMorgan Stanley highlights three key variables to watch in this earnings report:\n\n1. The U.S. export ban on H20 chips to China created a revenue headwind of over 10% in the July quarter—an impact not fully anticipated by management.\n\n2. Shipments of Blackwell GB200 racks were delayed due to “critical obstacles,” but ODM capacity has recently recovered significantly. Multiple channel sources reported that 1,500 units were shipped in April alone. Meanwhile, non-rack form factors such as B200 and Hopper solutions are also fulfilling inference demand.\n\n3. An explosion in inference demand is driving purchases of bare dies and custom systems, confirming the commercial momentum behind AI applications.\n\nMorgan Stanley has revised its April quarter revenue forecast down to $42.2 billion (vs. company guidance of $43 billion), factoring in a roughly $1 billion loss from halted H20 sales. For the July quarter, revenue is projected at $43.5 billion, assuming a $4–5 billion impact from the H20 ban. However, excluding H20, quarter-over-quarter incremental growth would be approximately $6.5 billion, signaling strong underlying demand even amid supply constraints.\n\nDespite short-term limitations due to capacity and regulatory restrictions, Morgan Stanley expects gross margins to remain flat or slightly improve and maintains its “Top Pick” rating on Nvidia, as GB200 ramp improvements and potential H20 license reinstatement could reignite momentum in the second half of the year.\n\n⸻\n\nKey Takeaways:\n•H20 Ban Impact:\n•April and July quarters saw ~$1B and $4–5B in revenue loss due to the H20 ban.\n•Excluding H20, revenue increased ~$6.5B sequentially.\n•Rack Shipments Recovering:\n•Combined ODM shipments reached 1,500 units in April.\n•Non-rack solutions (B200, Hopper) are also meeting inference demand despite supply constraints.\n•Surging Inference Demand:\n•Hyperscalers globally are reporting unexpectedly strong growth in inference requests.\n•This is driving rapid restocking of GPUs and underscores the sustained commercial momentum of AI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "orig:1921891242239283623", "ticker_idx": 0, "ts": "2025-05-12T11:32:33+00:00", "symbol": "MSCI China", "kind": "index", "geography": "China", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "over the next 90 days, the market will face more upside risk", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan bullish on Chinese equities for next 90 days: MSCI China at 12.2x P/E below avg, $550bn underweight flows to close, and US-China Grand Bargain now base case.", "resolved_tickers": ["MCHI"], "resolution_method": "etf", "resolution_reason": "MSCI China index ETF MCHI", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "250512_Comment on the U.S.-China Joint Statement - JPMorgan\n\n[Key Positives from the Joint Statement]\n\n(1) The positives from the joint statement can be summarized in three points.\n\n(2) Tariff cuts exceeding expectations\n■ For 90 days, the U.S. will reduce tariffs on Chinese imports to 30%. China will also reduce tariffs on U.S. imports to 10%.\n■ A 30% tariff on Chinese goods is lower than the market expectation of \"50%\" and even below our house's \"Bull-Case\" scenario of \"40%\".\n\n(3) A stance against decoupling\n■ Both the U.S. and China stated that they do not wish for economic decoupling and have established a mechanism for future negotiations.\n(4) Willingness to solve joint issues\n■ Bessent stated that both countries will cooperate to resolve the fentanyl issue and strive for balanced trade. He also mentioned the possibility of China signing a purchase agreement for U.S. products.\n\n[The 3 Key Questions]\n\n(1) 1st Question: What is the next catalyst?\n(2) Beyond the tariff cuts, the next steps we expect in our bullish scenario are as follows:\n\n(3) First is the removal or relaxation of trade restrictions on advanced technologies such as AI semiconductors.\n(4) Second is an \"explicit commitment\" from China to purchase specific U.S. products to reduce the trade deficit.\n\n(5) Third is a positive statement regarding Chinese companies listed in the U.S. (ADR), and the withdrawal of delisting threats without valid reasons.\n\n(6) We believe these factors are likely to attract market attention during the 90-day negotiation period between the U.S. and China.\n\n(7) 2nd Question: Does a Total Reset imply a Grand Bargain between the U.S. and China?\n\n(8) Trump has so far mentioned a complete reset in U.S.-China relations.\n\n(9) Ultimately, what the market wanted to know was whether, like in past cases, the two sides would cooperate in a limited way only when mutually beneficial economically—or whether they would narrow differences not only economically but also politically.\n\n(10) Now, what was once the most optimistic scenario—Case 2, including political cooperation—is becoming the Base-Scenario.\n\n(11) We believe the U.S. and China are now on the path toward a Grand Bargain, which corresponds to the highest level on JPMorgan Global Research Chair Joyce Chang’s 5-stage U.S.-China decoupling scale.\n\n(12) This is a very, very important shift, as it could 1) raise EPS growth expectations for Chinese companies, and 2) significantly contribute to lowering the geopolitical risk premium.\n\n■ Base Case Scenario: The U.S. and China cooperate only when it is economically beneficial. Political confrontation continues, but trade is partially maintained. This was the realistic base case in the past.\n■ Bull Case Scenario: The two sides narrow differences not only economically but also politically, with a clear cooperation plan (e.g., in trade, technology, and security). This is the \"ideal vision\" the market dreamed of. Even if they remain adversaries, trade continues to some extent. This was previously seen as the Bull Case.\n\n(13) JPMorgan now states that what was previously considered the Bull Case is now becoming the new Base Case.\n\n(14) In other words, the market is beginning to realistically consider not just limited economic cooperation, but the possibility of a broader political and economic Grand Bargain.\n\n(15) Question 3: Can the rally in Chinese equities continue?\n\n(16) Today, the Hang Seng Index rose 3% following related headlines and closed with an additional 2% gain after the joint statement.\n\n(17) We believe that over the next 90 days, the market will face more upside risk, based on three grounds:\n\n(18) Attractive valuation\n■ MSCI China is currently trading at a P/E of 12.2x, which is 6% below the 10-year average. In contrast, MSCI EM/DM is trading at a 7%/11% premium to their respective historical averages.\n(19) Favorable global flows\n■ Most global long-only funds are currently underweight China. To return to a neutral positioning, about $550bn of net buying would be required.\n(20) Continued positive news flow expected\n■ Both the U.S. and China have incentives to proceed with mutually beneficial agreements, and we believe there is a high probability of continued positive developments.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03287330140326894, "ret_signed_m1d": -0.03287330140326894, "alpha_spy_m1d": -0.0008829902321326744, "alpha_c_m1d": -0.0008829902321326744, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.031990311171136265, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0065383283650205515, "ret_signed_p1d": -0.0065383283650205515, "alpha_spy_p1d": -0.013142268682190594, "alpha_c_p1d": -0.013142268682190594, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.006603940317170043, "ret_p1w": -0.0065383283650205515, "ret_signed_p1w": -0.0065383283650205515, "alpha_spy_p1w": -0.026881628638765753, "alpha_c_p1w": -0.026881628638765753, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.020343300273745202, "ret_p1m": 0.011442003816338309, "ret_signed_p1m": 0.011442003816338309, "alpha_spy_p1m": -0.02301816663190759, "alpha_c_p1m": -0.02301816663190759, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.0344601704482459, "ret_p3m": 0.06747537875313814, "ret_signed_p3m": 0.06747537875313814, "alpha_spy_p3m": -0.02022601606504959, "alpha_c_p3m": -0.02022601606504959, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.08770139481818773, "ret_p6m": 0.150562362104268, "ret_signed_p6m": 0.150562362104268, "alpha_spy_p6m": -0.014318322958706275, "alpha_c_p6m": -0.014318322958706275, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.16488068506297426, "price_path": [-0.0579, -0.0583, -0.0291, -0.0291, -0.0194, -0.0242, -0.006, 0.0129, -0.0267, -0.0182, 0.0062, -0.0027, -0.0269, -0.0409, -0.0265, 0.0211, 0.0267, 0.0262, -0.0049, 0.014, 0.0104, 0.0129, 0.042, 0.0652, 0.0599, 0.059, 0.0296, 0.0107, 0.01, -0.0011, -0.0011, 0.0136, -0.0089, -0.012, -0.0093, -0.0163, -0.0252, -0.0919, -0.1645, -0.1787, -0.1271, -0.1286, -0.0977, -0.0817, -0.0888, -0.1072, -0.1048, -0.1048, -0.103, -0.0756, -0.0618, -0.0561, -0.0572, -0.0583, -0.0614, -0.0618, -0.063, -0.0298, -0.0301, -0.0202, -0.043, -0.0292, -0.0329, 0.0, -0.0065, 0.0098, -0.0089, -0.0116, -0.0065, 0.0025, -0.0013, -0.0056, -0.0051, -0.0051, -0.0185, -0.0314, -0.0218, -0.0407, -0.0327, -0.0236, -0.0049, -0.0013, -0.0025, 0.0084, 0.0114, 0.0145, 0.0133, -0.0053, 0.009, -0.0015, -0.0108, -0.0108, -0.0187, -0.0072, 0.0192, 0.0185, 0.0187, 0.0121, 0.0106, 0.0137, 0.0088, 0.0036, 0.0036, 0.0018, 0.0112, 0.0011, 0.009, 0.0079, 0.0189, 0.0396, 0.0365, 0.044, 0.0537, 0.0603, 0.0761, 0.0867, 0.0855, 0.0809, 0.0766, 0.0741, 0.0618, 0.0534, 0.0367, 0.0541, 0.0612, 0.0704, 0.0675, 0.0673, 0.0605, 0.0772, 0.1108, 0.0877, 0.0946, 0.1012, 0.093, 0.099, 0.1005, 0.1293, 0.1287, 0.1302, 0.1009, 0.1087, 0.1245, 0.1245, 0.1306, 0.1249, 0.106, 0.1253, 0.1427, 0.155, 0.1471, 0.1777, 0.1729, 0.1817, 0.1896, 0.2069, 0.1876, 0.1819, 0.1817, 0.1676, 0.1845, 0.1878, 0.1797, 0.2019, 0.2078, 0.2151, 0.2287, 0.2206, 0.2223, 0.2071, 0.2129, 0.1944, 0.126, 0.161, 0.1478, 0.1601, 0.1629, 0.1612, 0.1761, 0.1636, 0.1561, 0.1759, 0.1832, 0.203, 0.1944, 0.2012, 0.1808, 0.166, 0.1656, 0.1506]}
{"unit_id": "orig:1921891242239283623", "ticker_idx": 1, "ts": "2025-05-12T11:32:33+00:00", "symbol": "Hang Seng Index", "kind": "index", "geography": "China", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Can the rally in Chinese equities continue?... We believe that over the next 90 days, the market will face more upside risk", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan bullish on Chinese equities for next 90 days: MSCI China at 12.2x P/E below avg, $550bn underweight flows to close, and US-China Grand Bargain now base case.", "resolved_tickers": ["EWH"], "resolution_method": "etf", "resolution_reason": "Hang Seng Index proxy ETF EWH (iShares MSCI Hong Kong)", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "250512_Comment on the U.S.-China Joint Statement - JPMorgan\n\n[Key Positives from the Joint Statement]\n\n(1) The positives from the joint statement can be summarized in three points.\n\n(2) Tariff cuts exceeding expectations\n■ For 90 days, the U.S. will reduce tariffs on Chinese imports to 30%. China will also reduce tariffs on U.S. imports to 10%.\n■ A 30% tariff on Chinese goods is lower than the market expectation of \"50%\" and even below our house's \"Bull-Case\" scenario of \"40%\".\n\n(3) A stance against decoupling\n■ Both the U.S. and China stated that they do not wish for economic decoupling and have established a mechanism for future negotiations.\n(4) Willingness to solve joint issues\n■ Bessent stated that both countries will cooperate to resolve the fentanyl issue and strive for balanced trade. He also mentioned the possibility of China signing a purchase agreement for U.S. products.\n\n[The 3 Key Questions]\n\n(1) 1st Question: What is the next catalyst?\n(2) Beyond the tariff cuts, the next steps we expect in our bullish scenario are as follows:\n\n(3) First is the removal or relaxation of trade restrictions on advanced technologies such as AI semiconductors.\n(4) Second is an \"explicit commitment\" from China to purchase specific U.S. products to reduce the trade deficit.\n\n(5) Third is a positive statement regarding Chinese companies listed in the U.S. (ADR), and the withdrawal of delisting threats without valid reasons.\n\n(6) We believe these factors are likely to attract market attention during the 90-day negotiation period between the U.S. and China.\n\n(7) 2nd Question: Does a Total Reset imply a Grand Bargain between the U.S. and China?\n\n(8) Trump has so far mentioned a complete reset in U.S.-China relations.\n\n(9) Ultimately, what the market wanted to know was whether, like in past cases, the two sides would cooperate in a limited way only when mutually beneficial economically—or whether they would narrow differences not only economically but also politically.\n\n(10) Now, what was once the most optimistic scenario—Case 2, including political cooperation—is becoming the Base-Scenario.\n\n(11) We believe the U.S. and China are now on the path toward a Grand Bargain, which corresponds to the highest level on JPMorgan Global Research Chair Joyce Chang’s 5-stage U.S.-China decoupling scale.\n\n(12) This is a very, very important shift, as it could 1) raise EPS growth expectations for Chinese companies, and 2) significantly contribute to lowering the geopolitical risk premium.\n\n■ Base Case Scenario: The U.S. and China cooperate only when it is economically beneficial. Political confrontation continues, but trade is partially maintained. This was the realistic base case in the past.\n■ Bull Case Scenario: The two sides narrow differences not only economically but also politically, with a clear cooperation plan (e.g., in trade, technology, and security). This is the \"ideal vision\" the market dreamed of. Even if they remain adversaries, trade continues to some extent. This was previously seen as the Bull Case.\n\n(13) JPMorgan now states that what was previously considered the Bull Case is now becoming the new Base Case.\n\n(14) In other words, the market is beginning to realistically consider not just limited economic cooperation, but the possibility of a broader political and economic Grand Bargain.\n\n(15) Question 3: Can the rally in Chinese equities continue?\n\n(16) Today, the Hang Seng Index rose 3% following related headlines and closed with an additional 2% gain after the joint statement.\n\n(17) We believe that over the next 90 days, the market will face more upside risk, based on three grounds:\n\n(18) Attractive valuation\n■ MSCI China is currently trading at a P/E of 12.2x, which is 6% below the 10-year average. In contrast, MSCI EM/DM is trading at a 7%/11% premium to their respective historical averages.\n(19) Favorable global flows\n■ Most global long-only funds are currently underweight China. To return to a neutral positioning, about $550bn of net buying would be required.\n(20) Continued positive news flow expected\n■ Both the U.S. and China have incentives to proceed with mutually beneficial agreements, and we believe there is a high probability of continued positive developments.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01924098965454968, "ret_signed_m1d": -0.01924098965454968, "alpha_spy_m1d": 0.012749321516586587, "alpha_c_m1d": 0.012749321516586587, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.031990311171136265, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0016033083081792299, "ret_signed_p1d": -0.0016033083081792299, "alpha_spy_p1d": -0.008207248625349273, "alpha_c_p1d": -0.008207248625349273, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.006603940317170043, "ret_p1w": 0.010689546193338728, "ret_signed_p1w": 0.010689546193338728, "alpha_spy_p1w": -0.009653754080406474, "alpha_c_p1w": -0.009653754080406474, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.020343300273745202, "ret_p1m": 0.06253343137001965, "ret_signed_p1m": 0.06253343137001965, "alpha_spy_p1m": 0.028073260921773757, "alpha_c_p1m": 0.028073260921773757, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.0344601704482459, "ret_p3m": 0.14999881753996713, "ret_signed_p3m": 0.14999881753996713, "alpha_spy_p3m": 0.062297422721779405, "alpha_c_p3m": 0.062297422721779405, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.08770139481818773, "ret_p6m": 0.1936835134972439, "ret_signed_p6m": 0.1936835134972439, "alpha_spy_p6m": 0.028802828434269623, "alpha_c_p6m": 0.028802828434269623, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.16488068506297426, "price_path": [-0.1096, -0.1005, -0.0925, -0.0925, -0.0973, -0.0962, -0.0919, -0.0812, -0.0796, -0.0711, -0.0567, -0.0567, -0.0599, -0.0748, -0.062, -0.0283, -0.0198, -0.0214, -0.0406, -0.0321, -0.0267, -0.0289, -0.0299, -0.0171, -0.0262, -0.0182, -0.0369, -0.0465, -0.0518, -0.0545, -0.0609, -0.0481, -0.0615, -0.0641, -0.0518, -0.0492, -0.0657, -0.1245, -0.1619, -0.1876, -0.1561, -0.1571, -0.1464, -0.1256, -0.1261, -0.1267, -0.1219, -0.1219, -0.1245, -0.0989, -0.0871, -0.0823, -0.0818, -0.0844, -0.0828, -0.0636, -0.0663, -0.0486, -0.0481, -0.0267, -0.0219, -0.0257, -0.0192, 0.0, -0.0016, 0.0102, 0.0048, 0.0069, 0.0107, 0.0166, 0.0059, 0.0037, 0.0032, 0.0032, 0.0123, 0.0059, 0.0176, 0.0123, 0.0251, 0.0358, 0.0385, 0.0412, 0.0449, 0.0551, 0.0625, 0.0668, 0.0652, 0.0556, 0.0708, 0.0588, 0.0495, 0.0495, 0.0375, 0.0577, 0.0867, 0.0954, 0.0877, 0.0872, 0.0845, 0.0872, 0.0981, 0.1014, 0.1014, 0.0823, 0.0872, 0.0883, 0.0888, 0.097, 0.0981, 0.097, 0.1003, 0.1063, 0.109, 0.1178, 0.1342, 0.1363, 0.1402, 0.1353, 0.1505, 0.1571, 0.1434, 0.1303, 0.1101, 0.1282, 0.1374, 0.1369, 0.15, 0.1495, 0.1462, 0.1571, 0.1773, 0.1675, 0.1571, 0.1516, 0.1516, 0.1549, 0.156, 0.1729, 0.1746, 0.1718, 0.1582, 0.1631, 0.1555, 0.1555, 0.1626, 0.1544, 0.1467, 0.1566, 0.1686, 0.1773, 0.1828, 0.2035, 0.197, 0.2002, 0.1871, 0.1904, 0.1789, 0.1724, 0.1675, 0.1566, 0.1511, 0.1467, 0.1527, 0.174, 0.1768, 0.1833, 0.1795, 0.1773, 0.1762, 0.1691, 0.1762, 0.1773, 0.1282, 0.1522, 0.15, 0.1516, 0.1462, 0.1571, 0.1697, 0.1637, 0.1484, 0.1587, 0.1691, 0.1806, 0.1948, 0.1981, 0.1707, 0.1762, 0.2013, 0.1937]}
{"unit_id": "orig:1935563763723489549", "ticker_idx": 0, "ts": "2025-06-19T05:02:16+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm increasingly leaning towards a bullish view on NVIDIA these days", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Increasingly bullish NVDA: dominant HBM/CoWoS allocation, CUDA moat, and Sovereign AI demand all favor NVIDIA over ASIC challengers.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Recently, a friend asked me, \"Why are you so interested in ASICs these days?\" I replied, \"Why am I interested in ASICs? Because NVIDIA is the emperor of this era, and ASICs are challenging the emperor.\"\n\nHowever, my friend retorted. He said that designing and manufacturing an ASIC takes 2 to 3 years, and no matter how good the specifications are when designed, it will be obsolete in two years. He also mentioned that many ASIC projects emerged due to Blackwell's delays, but a significant number of them have been paused as Blackwell's supply has eased.\n\nI found my friend's points quite plausible. If NVIDIA's roadmap holds true, a tremendous volume of NVIDIA chips will be released every year, and it might seem difficult for ASICs with a two-year cycle to beat them.\n\nHowever, I don't think this is necessarily the complete answer. For example, what's to stop the delays experienced with Blackwell from being repeated with Rubin?\n\nFurthermore, no matter how low NVIDIA's Total Cost of Ownership (TCO) may be, will big tech companies continue to accept NVIDIA's high margins? I still find it hard to shake the feeling that big tech companies will be dissatisfied with NVIDIA's gross margins.\n\nEven so, I'm increasingly leaning towards a bullish view on NVIDIA these days.\n\n1. Dominant Position in the AI Chip Market\n\nNVIDIA is the largest consumer of HBM, buying the most HBM from SK Hynix and Micron. According to my personal estimation, about 70% of SK Hynix's HBM is sold to NVIDIA. Also, thanks to its strong relationship with TSMC, NVIDIA has the largest CoWoS capacity allocation.\n\nWhat does this signify? It means they can sell AI chips faster and in greater quantities than anyone else.\n\nI fundamentally believe that AI is a race against time. Big tech companies are trying to develop faster, more powerful AI than their competitors. This is a chicken game where if you stop, you lose.\n\nConversely, this means that if big tech companies try to use ASICs for training, having to tailor software for them and work with limited supply and performance from ASICs, they are, by contrast to NVIDIA's case, losing out.\n\nIn other words, this was a battle NVIDIA won from the very beginning, when the foundation of AI technology began to be built on NVIDIA's chips and the CUDA software. (Of course, this is under the assumption that NVIDIA also continues to evolve.)\n\n2. The Reality of Sovereign AI\n\nThe same goes for Sovereign AI.\n\nI actually think ASICs are more suitable for Sovereign AI.\n\nFrom the perspective of each nation, instead of being dependent on an American company like NVIDIA, I've always believed that true Sovereign AI means fostering local fabless companies in each country to develop their own AI chips and create independent AI.\n\nHowever, even for countries rich in chip design technology like South Korea and Taiwan, fostering their own AI chips is an extremely arduous and difficult policy.\n\nEspecially in democratic countries with fixed terms, there's a strong tendency to formulate policies based on short-term results.\n\nThis implies that few democratic countries will embark on the path of true Sovereign AI, which involves long development times for proprietary chips and uncertain outcomes. Instead, most will buy NVIDIA chips and create knock-offs of ChatGPT.\n\nUnless a country is like China or Russia, where the US restricts the sale of NVIDIA chips, the majority of countries will eventually buy NVIDIA's chips. That's because it's the easiest way to see results in the shortest amount of time.\n\nIn this sense, Sovereign AI might be viewed as an \"NVIDIA Sovereign Policy\" – copying American AI running on American chips (chuckles).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011204192941437152, "ret_signed_p1d": -0.011204192941437152, "alpha_spy_p1d": -0.008855553791258242, "alpha_c_p1d": -0.0023691201295533704, "bench_spy_p1d": -0.0023486391501789106, "bench_c_p1d": -0.008835072811883782, "ret_p1w": 0.06557607836357615, "ret_signed_p1w": 0.06557607836357615, "alpha_spy_p1w": 0.038395454900062465, "alpha_c_p1w": 0.009937945860100328, "bench_spy_p1w": 0.027180623463513687, "bench_c_p1w": 0.05563813250347582, "ret_p1m": 0.18511156252150518, "ret_signed_p1m": 0.18511156252150518, "alpha_spy_p1m": 0.1315576506149514, "alpha_c_p1m": 0.07954767932362516, "bench_spy_p1m": 0.053553911906553786, "bench_c_p1m": 0.10556388319788002, "ret_p3m": 0.20215760833989327, "ret_signed_p3m": 0.20215760833989327, "alpha_spy_p3m": 0.09417838349931174, "alpha_c_p3m": 0.032234884042055656, "bench_spy_p3m": 0.10797922484058153, "bench_c_p3m": 0.1699227242978376, "ret_p6m": 0.20318691915063614, "ret_signed_p6m": 0.20318691915063614, "alpha_spy_p6m": 0.05550495644924536, "alpha_c_p6m": -0.14537935462074136, "bench_spy_p6m": 0.14768196270139078, "bench_c_p6m": 0.3485662737713775, "price_path": [-0.1655, -0.1705, -0.2181, -0.2341, -0.2462, -0.2551, -0.2429, -0.241, -0.3003, -0.3518, -0.3289, -0.3381, -0.2142, -0.2606, -0.2375, -0.2391, -0.2288, -0.2818, -0.3024, -0.3024, -0.3339, -0.3203, -0.294, -0.2685, -0.237, -0.2527, -0.2507, -0.2514, -0.2329, -0.213, -0.2177, -0.2196, -0.1954, -0.1933, -0.1982, -0.1546, -0.1069, -0.0698, -0.0733, -0.0694, -0.0682, -0.0764, -0.0941, -0.087, -0.0976, -0.0976, -0.0687, -0.0734, -0.0433, -0.0712, -0.0557, -0.0293, -0.0245, -0.0378, -0.0259, -0.0197, -0.0105, -0.0182, -0.0033, -0.0241, -0.0054, -0.0093, 0.0, 0.0, -0.0112, -0.009, 0.0166, 0.0607, 0.0656, 0.0843, 0.086, 0.0538, 0.0809, 0.0953, 0.0953, 0.0877, 0.0998, 0.1196, 0.128, 0.1336, 0.1278, 0.1734, 0.178, 0.1892, 0.1851, 0.178, 0.1481, 0.1739, 0.1943, 0.1926, 0.2149, 0.2064, 0.2323, 0.2226, 0.1941, 0.2373, 0.2253, 0.2333, 0.2426, 0.2558, 0.2514, 0.259, 0.2482, 0.2512, 0.2404, 0.2511, 0.2073, 0.2057, 0.2028, 0.2235, 0.236, 0.2495, 0.2483, 0.2385, 0.1973, 0.1973, 0.1739, 0.1728, 0.18, 0.1481, 0.1569, 0.1738, 0.2189, 0.2179, 0.2224, 0.2219, 0.2022, 0.1706, 0.2115, 0.2145, 0.2622, 0.2266, 0.2165, 0.2215, 0.2249, 0.2501, 0.2826, 0.2871, 0.2985, 0.2897, 0.2754, 0.272, 0.3, 0.3238, 0.2591, 0.2945, 0.2376, 0.2362, 0.2498, 0.2595, 0.2555, 0.2453, 0.2393, 0.2522, 0.2804, 0.3163, 0.3819, 0.4232, 0.3947, 0.392, 0.4221, 0.3658, 0.3419, 0.2929, 0.2934, 0.3683, 0.3278, 0.3322, 0.2845, 0.3073, 0.2827, 0.2467, 0.2822, 0.2418, 0.2297, 0.2549, 0.2224, 0.2391, 0.2391, 0.2167, 0.2368, 0.2474, 0.2345, 0.2607, 0.254, 0.2756, 0.2716, 0.2634, 0.2438, 0.2032]}
{"unit_id": "orig:1933340311327293796", "ticker_idx": 0, "ts": "2025-06-13T01:47:04+00:00", "symbol": "ORCL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JP Morgan believes that Oracle reported better-than-expected revenue and earnings in F4Q, raised its FY26 revenue guidance to 'at least $67 billion'", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPMorgan note on ORCL: F4Q beat, FY26 guidance raised to $67B+ (17% YoY), $25B capex, Stargate upside; Neutral rating, $185 PT.", "resolved_tickers": ["ORCL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Oracle: Strong Cloud Growth and Heavyweight Capex, Stargate Becomes an Implied Incremental Driver\nJP Morgan (M. Murphy, 12/06/25)\n\nJP Morgan believes that Oracle reported better-than-expected revenue and earnings in F4Q, raised its FY26 revenue guidance to \"at least $67 billion\" (equivalent to 17% YoY growth in USD), and significantly increased its CAPEX guidance to $25 billion, reflecting a high degree of confidence in OCI capacity and the potential business from Stargate.\n \nF4Q Beat Expectations: Q4 CC Revenue Grew 11%, Cloud Business Grew 27%\n       \nIn Q4, total revenue grew 11% in both USD and constant currency, outperforming the market consensus of 9%. This was thanks to a foreign exchange benefit of approximately $143 million and a $170 million beat, 60% of which came from License revenue. Cloud business (SaaS+IaaS) revenue in constant currency grew 27% YoY, landing at the midpoint of the company's 26–28% guidance range, indicating continued acceleration in the core cloud business. Adjusted operating margin was 44% ($7.0 billion), slightly higher than the market's expectation of $6.8 billion; EPS of $1.70 also surpassed the $1.64 consensus.\n\nFY26 Guidance Jumps: Both Revenue Growth and RPO Revised Upward\n   \nOracle raised its FY26 total revenue guidance to \"at least $67 billion,\" corresponding to 17% and 16% YoY growth in constant currency and USD, respectively, which is higher than the market consensus of 14%. The Total Cloud business is expected to grow by more than 40% YoY, with Infrastructure growing over 70%. Total RPO (Remaining Performance Obligations) is expected to double YoY, reflecting optimism about large-scale AI projects and the potential of the Stargate business. For Q1, revenue growth is guided at 12–14%, with Cloud growth expected to be between 26–30%.\n\nSurging CAPEX: OCI Capacity Expansion Accelerates\n   \nThe FY26 capital expenditure guidance of $25 billion exceeds market expectations by about 20% and is designated for the procurement of core equipment that directly generates revenue. Oracle emphasized that it only builds new centers when there are orders and is not constrained by GPU supply. While this strong CAPEX commitment will erode short-term free cash flow, it is a necessary measure to support the continued rapid growth of OCI.\n\nStargate Could Be an Additional Catalyst\n   \nOracle remains cautiously optimistic about the Stargate project, believing it is not fully factored into the FY26 guidance. Once the core AI supercomputing business lands as expected, RPO and Infrastructure revenue could see significant further upside.\n\nRisks and Outlook\n   \nJP Morgan points out that although a rising proportion of license revenue and a 250bp sequential decline in profit margins pose potential concerns, the market is more likely to focus on the accelerated revenue growth driven by higher-than-expected AI cloud demand. The firm maintains a \"Neutral\" rating and a $185 price target, balancing the short-term pressure on free cash flow against the long-term cloud expansion opportunities.\n\n$ORCL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07136877206166214, "ret_signed_m1d": -0.07136877206166214, "alpha_spy_m1d": -0.08267524353732425, "alpha_c_m1d": -0.08267524353732425, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.011306471475662105, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0191431834588619, "ret_signed_p1d": -0.0191431834588619, "alpha_spy_p1d": -0.02865737993757722, "alpha_c_p1d": -0.02865737993757722, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.00951419647871532, "ret_p1w": -0.04669637474253563, "ret_signed_p1w": -0.04669637474253563, "alpha_spy_p1w": -0.04508309974672631, "alpha_c_p1w": -0.04508309974672631, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": -0.0016132749958093218, "ret_p1m": 0.0675922100764701, "ret_signed_p1m": 0.0675922100764701, "alpha_spy_p1m": 0.017915418310759712, "alpha_c_p1m": 0.017915418310759712, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.04967679176571038, "ret_p3m": 0.5287968043087379, "ret_signed_p3m": 0.5287968043087379, "alpha_spy_p3m": 0.4330880578888672, "alpha_c_p3m": 0.4330880578888672, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.0957087464198707, "ret_p6m": 0.028678254148199356, "ret_signed_p6m": 0.028678254148199356, "alpha_spy_p6m": -0.12299999753779756, "alpha_c_p6m": -0.12299999753779756, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.1516782516859969, "price_path": [-0.3081, -0.2942, -0.2929, -0.2952, -0.283, -0.2873, -0.3157, -0.3251, -0.3478, -0.3527, -0.3428, -0.3247, -0.3647, -0.4061, -0.4113, -0.4236, -0.3533, -0.3804, -0.385, -0.3744, -0.3777, -0.3971, -0.4024, -0.4024, -0.4293, -0.4088, -0.3895, -0.3611, -0.3565, -0.3489, -0.3458, -0.3462, -0.324, -0.2996, -0.3063, -0.3137, -0.306, -0.3016, -0.3015, -0.2695, -0.246, -0.2429, -0.2594, -0.2543, -0.2582, -0.2551, -0.2697, -0.2691, -0.2753, -0.2753, -0.2477, -0.2387, -0.2431, -0.2309, -0.226, -0.2141, -0.2189, -0.2048, -0.1914, -0.1769, -0.1754, -0.1805, -0.0714, 0.0, -0.0191, -0.0327, -0.0202, -0.0202, -0.0467, -0.038, 0.0002, -0.0209, -0.0112, -0.0231, 0.0158, 0.0174, 0.0686, 0.1027, 0.1027, 0.0792, 0.0896, 0.0957, 0.0942, 0.0736, 0.0676, 0.094, 0.1236, 0.1582, 0.1429, 0.134, 0.1087, 0.1264, 0.1307, 0.1413, 0.1534, 0.164, 0.1669, 0.1816, 0.1381, 0.1759, 0.1905, 0.194, 0.1612, 0.1643, 0.1765, 0.182, 0.137, 0.1406, 0.1561, 0.1597, 0.0925, 0.0945, 0.0857, 0.1006, 0.0961, 0.0905, 0.098, 0.119, 0.0529, 0.0529, 0.0491, 0.0404, 0.0384, 0.084, 0.1104, 0.1245, 0.5288, 0.4335, 0.3605, 0.4068, 0.4278, 0.4034, 0.3811, 0.4372, 0.528, 0.4613, 0.4363, 0.3565, 0.3199, 0.3166, 0.3095, 0.3457, 0.3446, 0.3323, 0.3577, 0.3235, 0.3439, 0.3851, 0.3665, 0.4367, 0.3946, 0.4162, 0.4599, 0.3588, 0.2929, 0.2834, 0.2718, 0.3063, 0.3216, 0.3126, 0.3099, 0.2841, 0.1982, 0.2249, 0.2027, 0.1576, 0.1675, 0.1372, 0.116, 0.1233, 0.1015, 0.0588, 0.0148, 0.0395, 0.0255, 0.0284, 0.052, -0.0173, -0.0729, -0.0658, -0.081, -0.044, -0.044, -0.058, -0.0627, -0.062, -0.0311, -0.0003, 0.0149, 0.0287]}
{"unit_id": "reply:1931503367337267621", "ticker_idx": 0, "ts": "2025-06-08T00:07:43+00:00", "symbol": "CRWV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CRWV is a bubble. I'm certain of it.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Certain CRWV is a bubble; strongly bearish.", "resolved_tickers": ["CRWV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@qiushao87 CRWV is a bubble. I'm certain of it.", "reply_text": null, "parent_text": "@Jukanlosreve NV seems to have a son that becomes a top-performing stock every year, this year is CRWV😂  \n\nIn comparison, sovereign AI has failed to drive it to break through DeepSeek's resistance price at 143, while the risks from model iteration are increasing", "quoted_text": null, "parent_author": "qiushao87", "ret_m1d": -0.13534856024246888, "ret_signed_m1d": 0.13534856024246888, "alpha_spy_m1d": 0.1344480889337064, "alpha_c_m1d": 0.1344480889337064, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.0009004713087624738, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.044417100159967804, "ret_signed_p1d": 0.044417100159967804, "alpha_spy_p1d": 0.05008676482885854, "alpha_c_p1d": 0.05008676482885854, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005669664668890739, "ret_p1w": -0.022208550079983902, "ret_signed_p1w": 0.022208550079983902, "alpha_spy_p1w": 0.02721122565438827, "alpha_c_p1w": 0.02721122565438827, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.005002675574404369, "ret_p1m": -0.06570023907631706, "ret_signed_p1m": 0.06570023907631706, "alpha_spy_p1m": 0.10321016567354446, "alpha_c_p1m": 0.10321016567354446, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.037509926597227405, "ret_p3m": -0.4603331273098803, "ret_signed_p3m": 0.4603331273098803, "alpha_spy_p3m": 0.5459770389317149, "alpha_c_p3m": 0.5459770389317149, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.0856439116218346, "ret_p6m": -0.5309685631304369, "ret_signed_p6m": 0.5309685631304369, "alpha_spy_p6m": 0.6739780665154144, "alpha_c_p6m": 0.6739780665154144, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.14300950338497742, "price_path": [null, null, null, null, null, null, null, null, null, null, null, null, -0.7532, -0.7713, -0.6757, -0.6215, -0.6681, -0.705, -0.6925, -0.731, -0.6991, -0.7397, -0.7301, -0.731, -0.7484, -0.7487, -0.7589, -0.7589, -0.7815, -0.7624, -0.7432, -0.7423, -0.7437, -0.7408, -0.7358, -0.7452, -0.7266, -0.6819, -0.6853, -0.6651, -0.6693, -0.6607, -0.6831, -0.6392, -0.6097, -0.5838, -0.5943, -0.5046, -0.4658, -0.4433, -0.3375, -0.3821, -0.3662, -0.3662, -0.2352, -0.2835, -0.3489, -0.3133, -0.2585, -0.0717, 0.0062, -0.1669, -0.1353, 0.0, -0.0444, -0.0765, -0.081, -0.092, -0.0222, 0.0606, 0.0487, 0.0487, 0.1325, 0.0714, 0.0649, -0.016, -0.0248, -0.013, 0.0059, -0.038, -0.0637, 0.0191, 0.0191, -0.0148, -0.0657, -0.0558, -0.1469, -0.2237, -0.1834, -0.1327, -0.1176, -0.1844, -0.2411, -0.2294, -0.1994, -0.2224, -0.2597, -0.2867, -0.3197, -0.3292, -0.3653, -0.2959, -0.3576, -0.346, -0.3101, -0.3199, -0.2531, -0.2008, -0.1377, -0.0824, -0.2735, -0.3862, -0.3833, -0.4028, -0.427, -0.4354, -0.4399, -0.4202, -0.4301, -0.4362, -0.402, -0.3659, -0.3643, -0.3643, -0.4242, -0.4455, -0.4603, -0.4504, -0.4229, -0.3817, -0.2774, -0.3048, -0.3093, -0.2568, -0.2674, -0.2544, -0.2511, -0.2297, -0.1781, -0.1925, -0.1771, -0.2186, -0.2576, -0.2442, -0.1558, -0.1545, -0.1487, -0.1685, -0.1743, -0.2052, -0.1365, -0.1173, -0.146, -0.1263, -0.173, -0.141, -0.1256, -0.1556, -0.2162, -0.2285, -0.2503, -0.2391, -0.1823, -0.1606, -0.1684, -0.1368, -0.1915, -0.1751, -0.2207, -0.2859, -0.2941, -0.3403, -0.3584, -0.3485, -0.4547, -0.473, -0.5167, -0.5228, -0.5353, -0.5379, -0.5378, -0.573, -0.558, -0.546, -0.5602, -0.5417, -0.5417, -0.5489, -0.5246, -0.531]}
{"unit_id": "orig:1932928689827950676", "ticker_idx": 0, "ts": "2025-06-11T22:31:26+00:00", "symbol": "ORCL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Even the FY26 CapEx outlook of $25B could be a conservative estimate. Demand is astronomical, and supply constraints are due to equipment purchasing limitations.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish ORCL: FY26 IaaS >70% growth + RPO >100% guidance underpinned by astronomical demand, non-cancelable contracts, and conservative $25B CapEx estimate.", "resolved_tickers": ["ORCL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Key Q&A from Oracle Conference Call\n\nQ. Does FY26 growth guidance (IaaS revenue growth over 70%, RPO growth over 100%) include Stargate?\n• Stargate is still a developing project.\n• Some business with partners (e.g., OpenAI), despite the official launch not yet occurring, is included.\n• However, at this point, OpenAI’s share within the total scale remains small.\n• AI, database, and applications areas are all included in RPO.\n• RPO consists of non-cancelable contracts, clearly confirming actual demand.\n• Supply is not keeping pace with demand, leading to delays or suspensions for some customers.\n• Additionally, many recently secured large contracts are unrelated to AI.\n• Multicloud demand has also surged dramatically.\n\nQ. Reasons for FY25 CapEx significantly exceeding consensus and explanation of FY26’s $25B figure.\n• CapEx mostly goes toward equipment purchases.\n• Oracle does not directly construct buildings; construction partners complete them, and Oracle pays rent afterward.\n• Increased CapEx indicates Oracle is rapidly filling data centers, implying mass equipment procurement.\n• Expansion of capacity is proceeding as quickly as possible.\n• Even the FY26 CapEx outlook of $25B could be a conservative estimate.\n• Infrastructure orders and builds proceed strictly based on demand.\n• Recent orders emphasized “taking all available infrastructure, regardless of location.”\n• Demand is astronomical, and supply constraints are due to equipment purchasing limitations.\n• Efforts to reduce CapEx through technological improvements are ongoing, but with near-infinite demand, CapEx inevitably rises.\n• Currently, there are no issues with GPU supply.\n\nQ. Sustainability and profitability of Oracle’s AI business?\n• Oracle’s latest database is an AI-centric vector database, critical for enterprises to practically leverage AI.\n• Most current AI models are trained on public data.\n• Enterprises, however, require AI models leveraging their internal data.\n• Oracle seamlessly connects internal enterprise data and AI models for immediate usability.\n• Essential to store data in a format usable by AI models.\n• This involves vectorizing data, allowing AI models to effectively search and learn from it.\n\n$ORCL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006236489593748784, "ret_signed_m1d": 0.006236489593748784, "alpha_spy_m1d": 0.0033764183239757273, "alpha_c_m1d": 0.0033764183239757273, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.002860071269773057, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.13312174257164133, "ret_signed_p1d": 0.13312174257164133, "alpha_spy_p1d": 0.12914751647448686, "alpha_c_p1d": 0.12914751647448686, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.003974226097154476, "ret_p1w": 0.19554369688841344, "ret_signed_p1w": 0.19554369688841344, "alpha_spy_p1w": 0.20206221610653197, "alpha_c_p1w": 0.20206221610653197, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": -0.006518519218118524, "ret_p1m": 0.33518171608362124, "ret_signed_p1m": 0.33518171608362124, "alpha_spy_p1m": 0.29143073569307476, "alpha_c_p1m": 0.29143073569307476, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.04375098039054648, "ret_p3m": 0.35495371149902843, "ret_signed_p3m": 0.35495371149902843, "alpha_spy_p3m": 0.2728264467455366, "alpha_c_p3m": 0.2728264467455366, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.08212726475349186, "ret_p6m": 0.2198558091939835, "ret_signed_p6m": 0.2198558091939835, "alpha_spy_p6m": 0.07525651893514396, "alpha_c_p6m": 0.07525651893514396, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.14459929025883955, "price_path": [-0.1567, -0.13, -0.1557, -0.1388, -0.1372, -0.14, -0.1251, -0.1304, -0.165, -0.1764, -0.2042, -0.2102, -0.1981, -0.176, -0.2247, -0.2754, -0.2816, -0.2967, -0.2109, -0.244, -0.2496, -0.2366, -0.2406, -0.2643, -0.2708, -0.2708, -0.3037, -0.2786, -0.255, -0.2204, -0.2148, -0.2055, -0.2018, -0.2022, -0.1751, -0.1454, -0.1536, -0.1626, -0.1531, -0.1479, -0.1476, -0.1086, -0.08, -0.0761, -0.0963, -0.0901, -0.0949, -0.0911, -0.1089, -0.1081, -0.1157, -0.1157, -0.082, -0.071, -0.0764, -0.0615, -0.0556, -0.041, -0.0469, -0.0297, -0.0134, 0.0044, 0.0062, 0.0, 0.1331, 0.2202, 0.1968, 0.1803, 0.1955, 0.1955, 0.1632, 0.1738, 0.2205, 0.1947, 0.2066, 0.192, 0.2395, 0.2414, 0.3039, 0.3455, 0.3455, 0.3168, 0.3295, 0.3369, 0.3352, 0.31, 0.3027, 0.335, 0.371, 0.4133, 0.3946, 0.3837, 0.3529, 0.3744, 0.3797, 0.3927, 0.4074, 0.4203, 0.4238, 0.4418, 0.3887, 0.4348, 0.4526, 0.4569, 0.4169, 0.4207, 0.4356, 0.4423, 0.3873, 0.3918, 0.4106, 0.4151, 0.333, 0.3355, 0.3247, 0.343, 0.3375, 0.3307, 0.3398, 0.3654, 0.2848, 0.2848, 0.2801, 0.2696, 0.267, 0.3227, 0.355, 0.3722, 0.8654, 0.7491, 0.6601, 0.7166, 0.7423, 0.7125, 0.6853, 0.7537, 0.8644, 0.7831, 0.7526, 0.6552, 0.6105, 0.6065, 0.5979, 0.642, 0.6407, 0.6257, 0.6567, 0.6149, 0.6399, 0.6901, 0.6674, 0.753, 0.7018, 0.728, 0.7814, 0.658, 0.5776, 0.566, 0.5518, 0.594, 0.6126, 0.6016, 0.5983, 0.5669, 0.4621, 0.4946, 0.4675, 0.4125, 0.4246, 0.3876, 0.3617, 0.3707, 0.344, 0.2919, 0.2383, 0.2683, 0.2513, 0.2549, 0.2836, 0.1991, 0.1312, 0.1399, 0.1214, 0.1665, 0.1665, 0.1494, 0.1436, 0.1446, 0.1823, 0.2199]}
{"unit_id": "orig:1927520560000745912", "ticker_idx": 0, "ts": "2025-05-28T00:21:27+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if Nvidia reports around $44B in Q1 revenue, discusses margin normalization, and provides an honest Q2 guide—not necessarily matching the buy-side's more bearish $45.47B estimate—the stock could still perform well, according to JP Morgan", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "JP Morgan view: NVDA can still perform well into earnings even with conservative Q2 guide, despite H20 ban headwinds and margin compression.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Trading Desk Commentary: Nvidia Earnings Preview Analysis\nJP Morgan (May 23, 2025)\n\nOverall, if Nvidia reports around $44B in Q1 revenue, discusses margin normalization, and provides an honest Q2 guide—not necessarily matching the buy-side’s more bearish $45.47B estimate—the stock could still perform well, according to JP Morgan.\n\n⸻\n\n■ Sentiment: Market Confusion and Disbelief\n\nAs reflected in our very disappointing buy-side survey results (shared with participants), interpreting Nvidia’s guidance this quarter is extremely difficult.\n\nA key factor is the H20 export ban that took effect on April 9, which we estimate could remove $15–16B in revenue from this fiscal year (similar to AMD’s case, this implies ~$7.6B impact on each of the June and September quarters).\n\nVirtually no one on the sell-side has updated their models to reflect this new reality. I expected the buy-side to do better—but based on our high-response-rate buy-side survey, many have not adjusted their numbers either. For instance, the average buy-side estimate still expects the June quarter guide to come in just $920M below consensus, far from the ~$7B cut that would be needed (see my potential estimates below).\n\n⸻\n\n■ Q1 Gross Margin: Confusion Still Reigns\n\nThere’s also the issue of Q1 gross margin guidance, which is likely to take a significant hit from the inventory write-down.\n\nRoughly one-third of survey respondents expect gross margin to be ~58%, while the other two-thirds are still clustered around 65–72%, despite the sell-side consensus remaining at 71.1%.\n\n⸻\n\n■ Key Takeaways From Investor Conversations Today\n• Some investors say, “Don’t worry, stay positive.” They believe the company may “stuff the channel” before a strong second-half demand spike to smooth earnings. Today’s Oracle (ORCL) news highlighted the upside potential in demand. Doubts about slow ODM ramp yields have begun to fade (Albert Hong wrote a great note on this last week). And we’ve all seen the accounts receivable buildup—there’s no point in debating that anymore.\n• A few investors are still including that $15–16B in their full-year estimates, expecting Nvidia to resume some chip shipments to China this year.\n• One investor suggested that some of those who missed the recent 35% rally have gone short, and are now talking up expectations to manufacture a “surprise” guide and trap the market. We’ve indeed seen short interest ticking up over the past few weeks.\n• Many believe that if the second half improves and GB300 ramps smoothly, any pullback would attract strong buying interest.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "orig:1937894226991493454", "ticker_idx": 0, "ts": "2025-06-25T15:22:42+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASIC server revenue CAGR will reach 96% from 2023 to 2027, significantly higher than AI GPU's 47%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC note: ASIC CAGR 96% through 2027, GB300 mass production 4Q25, RTX6000D >1M units 2H25, general server orders rebounding; bullish on ASIC/GPU server supply chain.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Servers: Strong ASIC Growth Momentum, Blackwell and General Server Order Improvement Expected in 2H25\n\nHSBC (F. Lee, 25/06/24)\n\nHSBC points out that ASIC server revenue CAGR will reach 96% from 2023 to 2027, significantly higher than AI GPU's 47%. Concurrently, orders for the GB series, RTX6000D, and general servers are all expected to see an upward turning point in 2H25, highlighting the structural shifts in server market demand.\n\nSignificant Growth Momentum for ASIC Servers, Meta and AWS Continue to Drive High-Density Server Upgrades\n\nHSBC projects a 96% compound annual growth rate for ASIC server revenue from 2023 to 2027, far surpassing AI GPUs' 47%. Meta is set to launch its high-density Santa Cruz system with 256 ASICs per rack in 2027, utilizing an orthogonal layout to optimize heat dissipation and wiring. This is expected to drive upgrades in PCB, CCL, power supply, and network accessory specifications. AWS is extending the lifecycle of Trainium 2 and advancing its Teton series high-density racks, which feature liquid cooling designs.\n\nGB200 and GB300 Server Demand Continues to Strengthen, Capacity Yield Remains a Short-Term Key Variable\n\nGB300 is expected to enter mass production in 4Q25, with full-year shipments reaching 7,000 to 8,000 units. Overall GB order volume has been raised to around 30,000 units. Although rack yield in May fluctuated between 40-50%, it is projected to improve to 60-70% in 2H25. Microsoft, Oracle, and Dell are contributing to major new orders, with Quanta, as the core contract manufacturer, pushing forward with GB300 adoption.\n\nRTX6000D Shipments Exceed Expectations Driven by China Demand, Blackwell Transition Ongoing\n\nRTX6000D is expected to ship over 1 million units in 2H25, primarily driven by the Chinese market. \nBlackwell HGX (8 GPU) has not yet seen upward revisions in expectations, but the trend of transitioning from B200 to B300 is still underway, and the pace of this transition in 2H25 warrants continued observation.\n\nGeneral Server Orders Rebound in Second Half, AWS and Microsoft as Main Drivers\n\nAlthough Meta has not yet revised up general server orders, demand for general servers is strengthening from 3Q25 onwards, driven by AWS and Microsoft. Taiwanese ODMs and component manufacturers such as ASPEED and Wiwynn have already reflected the upward trend in 2H25 demand, with full-year cloud server shipments expected to grow by 11%. CSP full-year capital expenditure remains below 25%, indicating that spending will be more concentrated in the second half.\n\n$AVGO $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003325099929278874, "ret_signed_m1d": -0.003325099929278874, "alpha_spy_m1d": -0.0027650491313332415, "alpha_c_m1d": 0.009974334172619725, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02085763478710234, "ret_signed_p1d": 0.02085763478710234, "alpha_spy_p1d": 0.013033854912614995, "alpha_c_p1d": 0.01359016075186159, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": 0.01983740863252592, "ret_signed_p1w": 0.01983740863252592, "alpha_spy_p1w": -0.0021187320091662354, "alpha_c_p1w": -0.002146552934502166, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": 0.09091239090585002, "ret_signed_p1m": 0.09091239090585002, "alpha_spy_p1m": 0.045946009668403454, "alpha_c_p1m": 0.04512751583492225, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.28233694763457606, "ret_signed_p3m": 0.28233694763457606, "alpha_spy_p3m": 0.1809257798120696, "alpha_c_p3m": 0.11166104351633344, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.24861219321601213, "ret_signed_p6m": 0.24861219321601213, "alpha_spy_p6m": 0.13129533940782379, "alpha_c_p6m": -0.01319741037722344, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3625, -0.3688, -0.3647, -0.3513, -0.4194, -0.4485, -0.4189, -0.4118, -0.302, -0.3505, -0.3141, -0.3276, -0.3254, -0.3418, -0.3554, -0.3554, -0.3734, -0.3607, -0.3331, -0.2907, -0.275, -0.2744, -0.2793, -0.2744, -0.2561, -0.2323, -0.2433, -0.2457, -0.2279, -0.2168, -0.2151, -0.1647, -0.1238, -0.125, -0.123, -0.1382, -0.1306, -0.1266, -0.134, -0.131, -0.1378, -0.1378, -0.1117, -0.0974, -0.0878, -0.0875, -0.0624, -0.0318, -0.0158, -0.0201, -0.0691, -0.0791, -0.0778, -0.0466, -0.0347, -0.0625, -0.0497, -0.0599, -0.0528, -0.0528, -0.0554, -0.0411, -0.0033, 0.0, 0.0209, 0.0178, 0.0416, 0.0003, 0.0198, 0.0398, 0.0398, 0.036, 0.027, 0.0501, 0.0406, 0.0368, 0.0414, 0.0616, 0.0611, 0.0824, 0.0706, 0.089, 0.0527, 0.0719, 0.0909, 0.0965, 0.112, 0.1238, 0.1435, 0.1098, 0.0906, 0.125, 0.1069, 0.1399, 0.1478, 0.1524, 0.1483, 0.1821, 0.1679, 0.176, 0.1575, 0.1553, 0.1143, 0.1002, 0.0943, 0.1109, 0.1118, 0.1261, 0.1345, 0.1663, 0.1237, 0.1237, 0.1269, 0.1426, 0.1566, 0.2654, 0.3061, 0.2721, 0.3964, 0.3589, 0.3598, 0.3757, 0.3603, 0.308, 0.3049, 0.3034, 0.2823, 0.2829, 0.2843, 0.2722, 0.2662, 0.2411, 0.2487, 0.2619, 0.28, 0.2807, 0.2698, 0.2733, 0.3077, 0.3059, 0.2287, 0.3501, 0.3025, 0.3298, 0.3405, 0.3222, 0.3219, 0.297, 0.2881, 0.3032, 0.3404, 0.3704, 0.4117, 0.461, 0.425, 0.3991, 0.3723, 0.3321, 0.3588, 0.3459, 0.3226, 0.3565, 0.3322, 0.3445, 0.2868, 0.2962, 0.2969, 0.2888, 0.3415, 0.3127, 0.2877, 0.4306, 0.4574, 0.5048, 0.5048, 0.5252, 0.4613, 0.4443, 0.4406, 0.4422, 0.4771, 0.5182, 0.5378, 0.5631, 0.5381, 0.3624, 0.2862, 0.2918, 0.234, 0.2486]}
{"unit_id": "orig:1937894226991493454", "ticker_idx": 1, "ts": "2025-06-25T15:22:42+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB300 is expected to enter mass production in 4Q25, with full-year shipments reaching 7,000 to 8,000 units. Overall GB order volume has been raised to around 30,000 units", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC note: ASIC CAGR 96% through 2027, GB300 mass production 4Q25, RTX6000D >1M units 2H25, general server orders rebounding; bullish on ASIC/GPU server supply chain.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Servers: Strong ASIC Growth Momentum, Blackwell and General Server Order Improvement Expected in 2H25\n\nHSBC (F. Lee, 25/06/24)\n\nHSBC points out that ASIC server revenue CAGR will reach 96% from 2023 to 2027, significantly higher than AI GPU's 47%. Concurrently, orders for the GB series, RTX6000D, and general servers are all expected to see an upward turning point in 2H25, highlighting the structural shifts in server market demand.\n\nSignificant Growth Momentum for ASIC Servers, Meta and AWS Continue to Drive High-Density Server Upgrades\n\nHSBC projects a 96% compound annual growth rate for ASIC server revenue from 2023 to 2027, far surpassing AI GPUs' 47%. Meta is set to launch its high-density Santa Cruz system with 256 ASICs per rack in 2027, utilizing an orthogonal layout to optimize heat dissipation and wiring. This is expected to drive upgrades in PCB, CCL, power supply, and network accessory specifications. AWS is extending the lifecycle of Trainium 2 and advancing its Teton series high-density racks, which feature liquid cooling designs.\n\nGB200 and GB300 Server Demand Continues to Strengthen, Capacity Yield Remains a Short-Term Key Variable\n\nGB300 is expected to enter mass production in 4Q25, with full-year shipments reaching 7,000 to 8,000 units. Overall GB order volume has been raised to around 30,000 units. Although rack yield in May fluctuated between 40-50%, it is projected to improve to 60-70% in 2H25. Microsoft, Oracle, and Dell are contributing to major new orders, with Quanta, as the core contract manufacturer, pushing forward with GB300 adoption.\n\nRTX6000D Shipments Exceed Expectations Driven by China Demand, Blackwell Transition Ongoing\n\nRTX6000D is expected to ship over 1 million units in 2H25, primarily driven by the Chinese market. \nBlackwell HGX (8 GPU) has not yet seen upward revisions in expectations, but the trend of transitioning from B200 to B300 is still underway, and the pace of this transition in 2H25 warrants continued observation.\n\nGeneral Server Orders Rebound in Second Half, AWS and Microsoft as Main Drivers\n\nAlthough Meta has not yet revised up general server orders, demand for general servers is strengthening from 3Q25 onwards, driven by AWS and Microsoft. Taiwanese ODMs and component manufacturers such as ASPEED and Wiwynn have already reflected the upward trend in 2H25 demand, with full-year cloud server shipments expected to grow by 11%. CSP full-year capital expenditure remains below 25%, indicating that spending will be more concentrated in the second half.\n\n$AVGO $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04153985983831754, "ret_signed_m1d": -0.04153985983831754, "alpha_spy_m1d": -0.040979809040371906, "alpha_c_m1d": -0.02824042573641894, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004601045575293838, "ret_signed_p1d": 0.004601045575293838, "alpha_spy_p1d": -0.003222734299193508, "alpha_c_p1d": -0.0026664284599469124, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": 0.019052474587535784, "ret_signed_p1w": 0.019052474587535784, "alpha_spy_p1w": -0.002903666054156373, "alpha_c_p1w": -0.0029314869794923037, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": 0.12591537288427257, "ret_signed_p1m": 0.12591537288427257, "alpha_spy_p1m": 0.08094899164682601, "alpha_c_p1m": 0.0801304978133448, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.18994459598187352, "ret_signed_p3m": 0.18994459598187352, "alpha_spy_p3m": 0.08853342815936704, "alpha_c_p3m": 0.019268691863630893, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.1286339477687075, "ret_signed_p6m": 0.1286339477687075, "alpha_spy_p6m": 0.011317093960519164, "alpha_c_p6m": -0.13317565582452806, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.2893, -0.2977, -0.2862, -0.2845, -0.3403, -0.3889, -0.3673, -0.376, -0.2591, -0.3029, -0.2812, -0.2826, -0.2729, -0.3229, -0.3423, -0.3423, -0.372, -0.3592, -0.3344, -0.3103, -0.2807, -0.2954, -0.2935, -0.2942, -0.2768, -0.258, -0.2624, -0.2643, -0.2414, -0.2394, -0.2441, -0.203, -0.1581, -0.123, -0.1263, -0.1226, -0.1215, -0.1292, -0.1459, -0.1393, -0.1492, -0.1492, -0.122, -0.1264, -0.098, -0.1244, -0.1098, -0.0849, -0.0804, -0.0929, -0.0817, -0.0758, -0.0671, -0.0744, -0.0603, -0.08, -0.0623, -0.066, -0.0572, -0.0572, -0.0678, -0.0657, -0.0415, 0.0, 0.0046, 0.0223, 0.0238, -0.0065, 0.0191, 0.0326, 0.0326, 0.0255, 0.0369, 0.0555, 0.0634, 0.0688, 0.0632, 0.1062, 0.1106, 0.1211, 0.1173, 0.1106, 0.0824, 0.1067, 0.1259, 0.1244, 0.1454, 0.1374, 0.1618, 0.1527, 0.1258, 0.1665, 0.1552, 0.1627, 0.1715, 0.184, 0.1798, 0.187, 0.1768, 0.1796, 0.1694, 0.1795, 0.1382, 0.1367, 0.134, 0.1535, 0.1653, 0.178, 0.1769, 0.1676, 0.1288, 0.1288, 0.1067, 0.1057, 0.1124, 0.0824, 0.0907, 0.1066, 0.1492, 0.1482, 0.1524, 0.152, 0.1334, 0.1036, 0.1422, 0.145, 0.1899, 0.1564, 0.1469, 0.1516, 0.1548, 0.1785, 0.2092, 0.2135, 0.2242, 0.2159, 0.2025, 0.1992, 0.2256, 0.248, 0.187, 0.2205, 0.1667, 0.1654, 0.1783, 0.1874, 0.1837, 0.1741, 0.1684, 0.1805, 0.2071, 0.241, 0.3028, 0.3418, 0.3149, 0.3123, 0.3408, 0.2877, 0.2651, 0.2189, 0.2194, 0.29, 0.2518, 0.256, 0.211, 0.2325, 0.2093, 0.1754, 0.2088, 0.1707, 0.1593, 0.1831, 0.1524, 0.1682, 0.1682, 0.1471, 0.166, 0.176, 0.1639, 0.1885, 0.1822, 0.2026, 0.1988, 0.1911, 0.1726, 0.1343, 0.1426, 0.1518, 0.1079, 0.1286]}
{"unit_id": "orig:1937894226991493454", "ticker_idx": 2, "ts": "2025-06-25T15:22:42+00:00", "symbol": "Wiwynn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwanese ODMs and component manufacturers such as ASPEED and Wiwynn have already reflected the upward trend in 2H25 demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC note: ASIC CAGR 96% through 2027, GB300 mass production 4Q25, RTX6000D >1M units 2H25, general server orders rebounding; bullish on ASIC/GPU server supply chain.", "resolved_tickers": ["6669.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wiwynn Corp listed on TWSE as 6669.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Global Servers: Strong ASIC Growth Momentum, Blackwell and General Server Order Improvement Expected in 2H25\n\nHSBC (F. Lee, 25/06/24)\n\nHSBC points out that ASIC server revenue CAGR will reach 96% from 2023 to 2027, significantly higher than AI GPU's 47%. Concurrently, orders for the GB series, RTX6000D, and general servers are all expected to see an upward turning point in 2H25, highlighting the structural shifts in server market demand.\n\nSignificant Growth Momentum for ASIC Servers, Meta and AWS Continue to Drive High-Density Server Upgrades\n\nHSBC projects a 96% compound annual growth rate for ASIC server revenue from 2023 to 2027, far surpassing AI GPUs' 47%. Meta is set to launch its high-density Santa Cruz system with 256 ASICs per rack in 2027, utilizing an orthogonal layout to optimize heat dissipation and wiring. This is expected to drive upgrades in PCB, CCL, power supply, and network accessory specifications. AWS is extending the lifecycle of Trainium 2 and advancing its Teton series high-density racks, which feature liquid cooling designs.\n\nGB200 and GB300 Server Demand Continues to Strengthen, Capacity Yield Remains a Short-Term Key Variable\n\nGB300 is expected to enter mass production in 4Q25, with full-year shipments reaching 7,000 to 8,000 units. Overall GB order volume has been raised to around 30,000 units. Although rack yield in May fluctuated between 40-50%, it is projected to improve to 60-70% in 2H25. Microsoft, Oracle, and Dell are contributing to major new orders, with Quanta, as the core contract manufacturer, pushing forward with GB300 adoption.\n\nRTX6000D Shipments Exceed Expectations Driven by China Demand, Blackwell Transition Ongoing\n\nRTX6000D is expected to ship over 1 million units in 2H25, primarily driven by the Chinese market. \nBlackwell HGX (8 GPU) has not yet seen upward revisions in expectations, but the trend of transitioning from B200 to B300 is still underway, and the pace of this transition in 2H25 warrants continued observation.\n\nGeneral Server Orders Rebound in Second Half, AWS and Microsoft as Main Drivers\n\nAlthough Meta has not yet revised up general server orders, demand for general servers is strengthening from 3Q25 onwards, driven by AWS and Microsoft. Taiwanese ODMs and component manufacturers such as ASPEED and Wiwynn have already reflected the upward trend in 2H25 demand, with full-year cloud server shipments expected to grow by 11%. CSP full-year capital expenditure remains below 25%, indicating that spending will be more concentrated in the second half.\n\n$AVGO $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0719557195571956, "ret_signed_m1d": -0.0719557195571956, "alpha_spy_m1d": -0.07139566875924996, "alpha_c_m1d": -0.06349372330442904, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.008461996252766557, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03321033210332103, "ret_signed_p1d": -0.03321033210332103, "alpha_spy_p1d": -0.041034111977808374, "alpha_c_p1d": -0.04010825716528843, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.006897925061967403, "ret_p1w": -0.08118081180811809, "ret_signed_p1w": -0.08118081180811809, "alpha_spy_p1w": -0.10313695244981025, "alpha_c_p1w": -0.09830529002607968, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.017124478217961592, "ret_p1m": -0.07380073800738007, "ret_signed_p1m": -0.07380073800738007, "alpha_spy_p1m": -0.11876711924482664, "alpha_c_p1m": -0.12208639776758112, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.04828565976020105, "ret_p3m": 0.1531365313653137, "ret_signed_p3m": 0.1531365313653137, "alpha_spy_p3m": 0.05172536354280721, "alpha_c_p3m": 0.018092353850114984, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.1350441775151987, "ret_p6m": 0.5405904059040589, "ret_signed_p6m": 0.5405904059040589, "alpha_spy_p6m": 0.4232735520958706, "alpha_c_p6m": 0.4038597644493154, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.13673064145474356, "price_path": [-0.3712, -0.4142, -0.3819, -0.3766, -0.3766, -0.3766, -0.4375, -0.4124, -0.4249, -0.3676, -0.3192, -0.3157, -0.3336, -0.3407, -0.3264, -0.3246, -0.3407, -0.3658, -0.3192, -0.3318, -0.3174, -0.3192, -0.3085, -0.3157, -0.3157, -0.2601, -0.2691, -0.2386, -0.2512, -0.2297, -0.2118, -0.192, -0.2028, -0.1598, -0.1473, -0.1473, -0.1903, -0.1759, -0.1598, -0.1616, -0.1831, -0.1508, -0.167, -0.1473, -0.1329, -0.1329, -0.1365, -0.1114, -0.0917, -0.1096, -0.1168, -0.1168, -0.1096, -0.0953, -0.0917, -0.081, -0.0953, -0.0863, -0.0828, -0.0935, -0.0917, -0.0863, -0.072, 0.0, -0.0332, -0.0664, -0.0664, -0.0959, -0.0812, -0.0886, -0.0941, -0.0941, -0.107, -0.0554, -0.0424, -0.0387, -0.0904, -0.0701, -0.0756, -0.0941, -0.0443, -0.0683, -0.0886, -0.0978, -0.0738, -0.072, -0.0406, -0.0295, -0.0221, 0.0203, 0.0203, -0.024, 0.0166, 0.024, 0.0369, 0.0664, 0.1716, 0.2066, 0.2491, 0.2528, 0.2343, 0.2638, 0.2196, 0.1015, 0.1181, 0.0959, 0.1255, 0.1273, 0.1015, 0.0978, 0.0941, 0.0461, 0.0646, 0.0793, 0.083, 0.1162, 0.1531, 0.1107, 0.2214, 0.2214, 0.1956, 0.1882, 0.1808, 0.1661, 0.1808, 0.1919, 0.1531, 0.1587, 0.1365, 0.1421, 0.1255, 0.1255, 0.2232, 0.2177, 0.2232, 0.2786, 0.2786, 0.2952, 0.2675, 0.2435, 0.2435, 0.2657, 0.2952, 0.4004, 0.4244, 0.3967, 0.4317, 0.5, 0.5, 0.5148, 0.5148, 0.4963, 0.5074, 0.5166, 0.5867, 0.6107, 0.5923, 0.5867, 0.5369, 0.5941, 0.5037, 0.5849, 0.7417, 0.6827, 0.6458, 0.5793, 0.6052, 0.524, 0.5314, 0.5793, 0.5111, 0.4871, 0.5387, 0.6421, 0.6421, 0.6863, 0.6347, 0.6531, 0.6974, 0.6771, 0.7288, 0.7085, 0.6679, 0.6697, 0.6679, 0.6255, 0.583, 0.5351, 0.5406, 0.5406]}
{"unit_id": "orig:1937894226991493454", "ticker_idx": 3, "ts": "2025-06-25T15:22:42+00:00", "symbol": "ASPEED Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwanese ODMs and component manufacturers such as ASPEED and Wiwynn have already reflected the upward trend in 2H25 demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC note: ASIC CAGR 96% through 2027, GB300 mass production 4Q25, RTX6000D >1M units 2H25, general server orders rebounding; bullish on ASIC/GPU server supply chain.", "resolved_tickers": ["6674.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "ASPEED Technology listed on TWSE as 6674.TW (BMC supplier)", "bench_c": "EWT", "bench_c_reason": "country_fallback (suffix='TW')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Global Servers: Strong ASIC Growth Momentum, Blackwell and General Server Order Improvement Expected in 2H25\n\nHSBC (F. Lee, 25/06/24)\n\nHSBC points out that ASIC server revenue CAGR will reach 96% from 2023 to 2027, significantly higher than AI GPU's 47%. Concurrently, orders for the GB series, RTX6000D, and general servers are all expected to see an upward turning point in 2H25, highlighting the structural shifts in server market demand.\n\nSignificant Growth Momentum for ASIC Servers, Meta and AWS Continue to Drive High-Density Server Upgrades\n\nHSBC projects a 96% compound annual growth rate for ASIC server revenue from 2023 to 2027, far surpassing AI GPUs' 47%. Meta is set to launch its high-density Santa Cruz system with 256 ASICs per rack in 2027, utilizing an orthogonal layout to optimize heat dissipation and wiring. This is expected to drive upgrades in PCB, CCL, power supply, and network accessory specifications. AWS is extending the lifecycle of Trainium 2 and advancing its Teton series high-density racks, which feature liquid cooling designs.\n\nGB200 and GB300 Server Demand Continues to Strengthen, Capacity Yield Remains a Short-Term Key Variable\n\nGB300 is expected to enter mass production in 4Q25, with full-year shipments reaching 7,000 to 8,000 units. Overall GB order volume has been raised to around 30,000 units. Although rack yield in May fluctuated between 40-50%, it is projected to improve to 60-70% in 2H25. Microsoft, Oracle, and Dell are contributing to major new orders, with Quanta, as the core contract manufacturer, pushing forward with GB300 adoption.\n\nRTX6000D Shipments Exceed Expectations Driven by China Demand, Blackwell Transition Ongoing\n\nRTX6000D is expected to ship over 1 million units in 2H25, primarily driven by the Chinese market. \nBlackwell HGX (8 GPU) has not yet seen upward revisions in expectations, but the trend of transitioning from B200 to B300 is still underway, and the pace of this transition in 2H25 warrants continued observation.\n\nGeneral Server Orders Rebound in Second Half, AWS and Microsoft as Main Drivers\n\nAlthough Meta has not yet revised up general server orders, demand for general servers is strengthening from 3Q25 onwards, driven by AWS and Microsoft. Taiwanese ODMs and component manufacturers such as ASPEED and Wiwynn have already reflected the upward trend in 2H25 demand, with full-year cloud server shipments expected to grow by 11%. CSP full-year capital expenditure remains below 25%, indicating that spending will be more concentrated in the second half.\n\n$AVGO $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011214924749926913, "ret_signed_m1d": 0.011214924749926913, "alpha_spy_m1d": 0.011774975547872546, "alpha_c_m1d": 0.01976495554309854, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.008550030793171626, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.007823779874487347, "alpha_c_p1d": -0.009422376352849149, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.009422376352849149, "ret_p1w": -0.009345794392523366, "ret_signed_p1w": -0.009345794392523366, "alpha_spy_p1w": -0.03130193503421552, "alpha_c_p1w": -0.03447227057456559, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.025126476182042223, "ret_p1m": -0.024299051160010188, "ret_signed_p1m": -0.024299051160010188, "alpha_spy_p1m": -0.06926543239745675, "alpha_c_p1m": -0.060592873296478555, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.03629382213646837, "ret_p3m": -0.2186916030455972, "ret_signed_p3m": -0.2186916030455972, "alpha_spy_p3m": -0.3201027708681037, "alpha_c_p3m": -0.3357740924037471, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.11708248935814991, "ret_p6m": -0.2598131126332506, "ret_signed_p6m": -0.2598131126332506, "alpha_spy_p6m": -0.3771299664414389, "alpha_c_p6m": -0.37072777774375854, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.11091466511050796, "price_path": [-0.1607, -0.2187, -0.2168, -0.215, -0.215, -0.215, -0.2935, -0.3626, -0.4262, -0.372, -0.3832, -0.3907, -0.3664, -0.3701, -0.3664, -0.3458, -0.3514, -0.357, -0.3682, -0.3607, -0.3495, -0.286, -0.2206, -0.2075, -0.2075, -0.1327, -0.1271, -0.1234, -0.1477, -0.1458, -0.1047, -0.1159, -0.1589, -0.1888, -0.172, -0.1794, -0.1944, -0.2056, -0.2168, -0.2224, -0.2299, -0.2523, -0.2617, -0.2673, -0.1944, -0.1944, -0.114, -0.0991, -0.1065, -0.0187, -0.0056, -0.0131, -0.015, -0.0467, 0.0486, 0.028, 0.0224, 0.0037, -0.0093, -0.0486, 0.0093, -0.0187, 0.0112, 0.0, 0.0, 0.0056, -0.0187, -0.015, -0.0093, -0.0093, -0.0336, -0.0411, -0.0523, -0.0636, -0.0617, -0.0243, -0.0112, 0.0206, 0.0411, 0.0561, 0.0654, 0.0037, -0.0093, 0.015, -0.0243, -0.0056, -0.0037, 0.0093, 0.0131, 0.0037, 0.0037, 0.0393, 0.0355, 0.0262, 0.0, 0.0206, 0.0822, 0.086, 0.1308, 0.1028, 0.0617, 0.0336, 0.0654, 0.0355, 0.0636, 0.0206, 0.0075, -0.0449, -0.0579, -0.0953, -0.1196, -0.1477, -0.185, -0.1813, -0.1794, -0.1944, -0.2056, -0.2075, -0.1944, -0.114, -0.2, -0.2393, -0.2336, -0.2336, -0.2336, -0.2299, -0.2187, -0.2112, -0.1963, -0.1944, -0.2056, -0.2056, -0.1832, -0.2, -0.2056, -0.2112, -0.2112, -0.2168, -0.215, -0.2093, -0.2093, -0.2037, -0.2262, -0.2187, -0.215, -0.2224, -0.2449, -0.2318, -0.2224, -0.2374, -0.2374, -0.2355, -0.2393, -0.2449, -0.2056, -0.2243, -0.1477, -0.2168, -0.2112, -0.2262, -0.2262, -0.2318, -0.2168, -0.2056, -0.2075, -0.2112, -0.215, -0.2206, -0.2299, -0.228, -0.2411, -0.243, -0.243, -0.2318, -0.2411, -0.2355, -0.2355, -0.2299, -0.2318, -0.2336, -0.2318, -0.2393, -0.2355, -0.2318, -0.2449, -0.2467, -0.2505, -0.2579, -0.2598, -0.2598]}
{"unit_id": "orig:1930405301787996326", "ticker_idx": 0, "ts": "2025-06-04T23:24:23+00:00", "symbol": "SNPS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan: EDA export controls short-term headwind for Synopsys/Cadence, indirect benefit to Taiwan fabless (MediaTek, Novatek) and SK Hynix via CXMT yield delays.", "resolved_tickers": ["SNPS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "EDA Export Controls: Short-Term Disruption to China’s Fabless Sector, Long-Term Catalyst for Localization and Onshore Substitution\n\nJP Morgan (G. Hariharan, 2025/06/02)\n\nThe U.S. Bureau of Industry and Security (BIS) has imposed stricter licensing requirements on exports of EDA tools and IP to China, which may cause short-term disruptions for China’s advanced digital IC design sector. However, these measures are likely to accelerate the localization of domestic EDA tools and foster the rise of alternative suppliers, potentially reshaping the industry over the long term.\n\n⸻\n\n• Export Control Tightening and Targeted Entities\n\nSince 2022, the U.S. BIS has restricted exports of EDA software (ECAD) for advanced 3nm GAAFET process design.\nIn late May 2025, further measures were introduced requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers, especially those linked to the defense sector.\n\nThese three companies account for over 70% of the global EDA market, meaning that license delays or denials could severely limit access to critical tools and IP for Chinese logic IC design firms.\n\n⸻\n\n• Analog/Power IC Design and Domestic Substitution Potential\n\nThis round of export controls mainly targets digital chip design and IP licensing.\nThe impact on analog and power IC firms is relatively limited, with local Chinese EDA players like Empyrean, Primarius, and Semitronix expected to provide adequate tool support in these segments.\n\nIndustry research suggests that local substitution is less challenging in analog and power domains, whereas companies reliant on advanced digital design tools from Synopsys, Cadence, and Siemens will be more severely impacted.\n\n⸻\n\n• Accelerated Localization and Supply Chain Restructuring\n\nIn the short term, failure to secure U.S. licenses could result in project delays and development constraints for China’s fabless design firms.\n\nOver the medium to long term, however, the certification and adoption of domestic EDA tools is expected to accelerate.\nChinese firms like Empyrean and Primarius have seen 30%+ revenue CAGR and over 50% R&D CAGR in recent years, and increased support from the government and capital markets will further fuel the maturation of China’s indigenous EDA ecosystem, reducing reliance on imported software.\n\n⸻\n\n• TSMC Advanced Nodes and Design Service Intermediaries\n\nChina’s leading fabless firms—especially in non-AI domains such as mobile SoCs, ADAS, and crypto mining—remain heavily reliant on TSMC’s advanced nodes and corresponding PDKs.\n\nHowever, TSMC primarily supports mainstream U.S. EDA tools and does not fully integrate with Chinese domestic EDA tools.\nIf licenses for Synopsys, Cadence, and Siemens are denied, these firms may be locked out of leading-edge process access.\n\nDesign service providers such as Alchip and GUC could act as intermediaries to obtain licenses and maintain TSMC access on behalf of Chinese clients, but the sustainability of this workaround remains uncertain.\n\nReduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek.\n\n⸻\n\n• Impact on Memory Sector and Supply-Demand Dynamics\n\nIf EDA tools face similar restrictions as equipment exports, Chinese memory manufacturers could struggle to improve DDR5 yields.\n\nFor instance, ChangXin Memory Technologies (CXMT) is reportedly operating a 10Z4nm process comparable to global 1y/1znm nodes.\nHowever, intensified EDA controls could delay yield ramp-up in DDR5, sustaining a tight supply-demand environment for these products.\n\nIn the short term, DDR4 demand remains supported, while EDA-related delays in China’s capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014645264354754328, "ret_signed_m1d": 0.014645264354754328, "alpha_spy_m1d": 0.01491388389480064, "alpha_c_m1d": 0.01491388389480064, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": 0.00026861954004631095, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019040997736930576, "ret_signed_p1d": -0.019040997736930576, "alpha_spy_p1d": -0.02387376679311981, "alpha_c_p1d": -0.02387376679311981, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0048327690561892345, "ret_p1w": 0.05820074051190893, "ret_signed_p1w": -0.05820074051190893, "alpha_spy_p1w": -0.04908880934978388, "alpha_c_p1w": -0.04908880934978388, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.009111931162125053, "ret_p1m": 0.15966099749185614, "ret_signed_p1m": -0.15966099749185614, "alpha_spy_p1m": -0.10720725692591326, "alpha_c_p1m": -0.10720725692591326, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.052453740565942875, "ret_p3m": 0.27542850951501685, "ret_signed_p3m": -0.27542850951501685, "alpha_spy_p3m": -0.18980269021506024, "alpha_c_p3m": -0.18980269021506024, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.08562581929995661, "ret_p6m": -0.13421671936843016, "ret_signed_p6m": 0.13421671936843016, "alpha_spy_p6m": 0.28129663953466133, "alpha_c_p6m": 0.28129663953466133, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.14707992016623117, "price_path": [-0.0471, -0.0916, -0.0898, -0.0861, -0.0943, -0.057, -0.0416, -0.0533, -0.0484, -0.049, -0.052, -0.0326, -0.0337, -0.0492, -0.057, -0.0745, -0.0937, -0.0787, -0.0732, -0.1171, -0.1798, -0.195, -0.1936, -0.094, -0.1401, -0.1099, -0.1082, -0.1005, -0.1216, -0.1265, -0.1265, -0.1497, -0.1264, -0.1077, -0.0725, -0.0562, -0.0618, -0.0262, -0.03, -0.0292, -0.0019, 0.0089, 0.0008, 0.0155, 0.0239, 0.0205, 0.0685, 0.0899, 0.094, 0.0873, 0.0872, 0.0905, 0.0896, 0.0597, 0.064, 0.0542, 0.0542, 0.0816, -0.0227, -0.0385, -0.0195, -0.0149, -0.0146, 0.0, 0.019, 0.0271, 0.0477, 0.066, 0.0582, 0.0461, 0.012, 0.0131, 0.0047, 0.0009, 0.0009, -0.0056, -0.0047, 0.0122, 0.0298, 0.0476, 0.0622, 0.0835, 0.1027, 0.1055, 0.1597, 0.1597, 0.1338, 0.1655, 0.1643, 0.1965, 0.1819, 0.1613, 0.2027, 0.2071, 0.2448, 0.2358, 0.2499, 0.3, 0.2985, 0.2884, 0.2713, 0.2524, 0.3437, 0.3638, 0.3387, 0.3074, 0.3441, 0.3282, 0.3227, 0.311, 0.309, 0.3021, 0.3225, 0.3072, 0.3026, 0.3058, 0.3215, 0.295, 0.2704, 0.2641, 0.2818, 0.2616, 0.2595, 0.2747, 0.2937, 0.2754, 0.2754, 0.2511, 0.2513, 0.2721, 0.2641, 0.2872, 0.2772, -0.1805, -0.0742, -0.1009, -0.1141, -0.0998, -0.101, 0.0146, 0.0471, 0.0879, 0.0362, -0.0108, 0.0296, 0.0308, 0.0178, 0.0427, 0.0329, -0.0043, -0.0085, 0.0106, 0.0098, 0.0342, 0.0237, -0.0724, -0.0532, -0.0622, -0.0788, -0.0697, -0.054, -0.0419, -0.0299, -0.0399, -0.0361, -0.019, -0.0164, -0.0346, -0.0377, -0.0639, -0.0409, -0.0581, -0.1201, -0.1354, -0.1628, -0.1686, -0.153, -0.164, -0.1572, -0.1675, -0.1762, -0.1753, -0.1889, -0.1836, -0.1851, -0.1793, -0.1449, -0.1513, -0.1342, -0.1342]}
{"unit_id": "orig:1930405301787996326", "ticker_idx": 1, "ts": "2025-06-04T23:24:23+00:00", "symbol": "CDNS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan: EDA export controls short-term headwind for Synopsys/Cadence, indirect benefit to Taiwan fabless (MediaTek, Novatek) and SK Hynix via CXMT yield delays.", "resolved_tickers": ["CDNS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Application", "unit_kind": "original", "lang": "en", "tweet_text": "EDA Export Controls: Short-Term Disruption to China’s Fabless Sector, Long-Term Catalyst for Localization and Onshore Substitution\n\nJP Morgan (G. Hariharan, 2025/06/02)\n\nThe U.S. Bureau of Industry and Security (BIS) has imposed stricter licensing requirements on exports of EDA tools and IP to China, which may cause short-term disruptions for China’s advanced digital IC design sector. However, these measures are likely to accelerate the localization of domestic EDA tools and foster the rise of alternative suppliers, potentially reshaping the industry over the long term.\n\n⸻\n\n• Export Control Tightening and Targeted Entities\n\nSince 2022, the U.S. BIS has restricted exports of EDA software (ECAD) for advanced 3nm GAAFET process design.\nIn late May 2025, further measures were introduced requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers, especially those linked to the defense sector.\n\nThese three companies account for over 70% of the global EDA market, meaning that license delays or denials could severely limit access to critical tools and IP for Chinese logic IC design firms.\n\n⸻\n\n• Analog/Power IC Design and Domestic Substitution Potential\n\nThis round of export controls mainly targets digital chip design and IP licensing.\nThe impact on analog and power IC firms is relatively limited, with local Chinese EDA players like Empyrean, Primarius, and Semitronix expected to provide adequate tool support in these segments.\n\nIndustry research suggests that local substitution is less challenging in analog and power domains, whereas companies reliant on advanced digital design tools from Synopsys, Cadence, and Siemens will be more severely impacted.\n\n⸻\n\n• Accelerated Localization and Supply Chain Restructuring\n\nIn the short term, failure to secure U.S. licenses could result in project delays and development constraints for China’s fabless design firms.\n\nOver the medium to long term, however, the certification and adoption of domestic EDA tools is expected to accelerate.\nChinese firms like Empyrean and Primarius have seen 30%+ revenue CAGR and over 50% R&D CAGR in recent years, and increased support from the government and capital markets will further fuel the maturation of China’s indigenous EDA ecosystem, reducing reliance on imported software.\n\n⸻\n\n• TSMC Advanced Nodes and Design Service Intermediaries\n\nChina’s leading fabless firms—especially in non-AI domains such as mobile SoCs, ADAS, and crypto mining—remain heavily reliant on TSMC’s advanced nodes and corresponding PDKs.\n\nHowever, TSMC primarily supports mainstream U.S. EDA tools and does not fully integrate with Chinese domestic EDA tools.\nIf licenses for Synopsys, Cadence, and Siemens are denied, these firms may be locked out of leading-edge process access.\n\nDesign service providers such as Alchip and GUC could act as intermediaries to obtain licenses and maintain TSMC access on behalf of Chinese clients, but the sustainability of this workaround remains uncertain.\n\nReduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek.\n\n⸻\n\n• Impact on Memory Sector and Supply-Demand Dynamics\n\nIf EDA tools face similar restrictions as equipment exports, Chinese memory manufacturers could struggle to improve DDR5 yields.\n\nFor instance, ChangXin Memory Technologies (CXMT) is reportedly operating a 10Z4nm process comparable to global 1y/1znm nodes.\nHowever, intensified EDA controls could delay yield ramp-up in DDR5, sustaining a tight supply-demand environment for these products.\n\nIn the short term, DDR4 demand remains supported, while EDA-related delays in China’s capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0036345358233770275, "ret_signed_m1d": 0.0036345358233770275, "alpha_spy_m1d": 0.0039031553634233385, "alpha_c_m1d": 0.0039031553634233385, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": 0.00026861954004631095, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006419886846539935, "ret_signed_p1d": -0.006419886846539935, "alpha_spy_p1d": -0.01125265590272917, "alpha_c_p1d": -0.01125265590272917, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0048327690561892345, "ret_p1w": 0.04347832396711104, "ret_signed_p1w": -0.04347832396711104, "alpha_spy_p1w": -0.03436639280498599, "alpha_c_p1w": -0.03436639280498599, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.009111931162125053, "ret_p1m": 0.11008832993897832, "ret_signed_p1m": -0.11008832993897832, "alpha_spy_p1m": -0.05763458937303545, "alpha_c_p1m": -0.05763458937303545, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.052453740565942875, "ret_p3m": 0.19031929327755992, "ret_signed_p3m": -0.19031929327755992, "alpha_spy_p3m": -0.10469347397760331, "alpha_c_p3m": -0.10469347397760331, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.08562581929995661, "ret_p6m": 0.04059107491433256, "ret_signed_p6m": -0.04059107491433256, "alpha_spy_p6m": 0.10648884525189861, "alpha_c_p6m": 0.10648884525189861, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.14707992016623117, "price_path": [-0.1739, -0.2053, -0.2043, -0.1874, -0.1795, -0.1595, -0.1335, -0.1376, -0.1213, -0.1219, -0.1086, -0.0926, -0.0859, -0.1053, -0.1082, -0.1281, -0.1361, -0.121, -0.1001, -0.1545, -0.209, -0.2058, -0.2132, -0.1011, -0.1372, -0.1186, -0.1149, -0.1022, -0.1194, -0.1161, -0.1161, -0.1455, -0.1231, -0.0984, -0.0429, -0.0162, -0.0294, 0.0266, 0.0113, 0.0181, 0.0467, 0.049, 0.0387, 0.0474, 0.0461, 0.0387, 0.059, 0.0824, 0.0849, 0.0828, 0.0923, 0.088, 0.0888, 0.0786, 0.0755, 0.0717, 0.0717, 0.0974, -0.0197, -0.0336, -0.0249, -0.0064, -0.0036, 0.0, 0.0064, 0.0087, 0.0247, 0.0383, 0.0435, 0.0457, 0.019, 0.0218, 0.0156, 0.0083, 0.0083, 0.0034, -0.0029, 0.0082, 0.0106, 0.0316, 0.0367, 0.0467, 0.0512, 0.0562, 0.1101, 0.1101, 0.088, 0.0981, 0.0968, 0.096, 0.089, 0.0784, 0.0803, 0.0685, 0.0899, 0.0719, 0.0745, 0.0856, 0.1089, 0.1027, 0.1284, 0.1337, 0.2441, 0.2603, 0.2383, 0.2125, 0.2394, 0.2245, 0.2234, 0.2036, 0.1959, 0.1858, 0.2011, 0.1859, 0.1852, 0.1885, 0.2101, 0.1783, 0.1734, 0.1806, 0.1876, 0.1725, 0.1686, 0.1783, 0.2034, 0.1903, 0.1903, 0.1644, 0.1798, 0.1866, 0.1923, 0.225, 0.2288, 0.1499, 0.2048, 0.1667, 0.194, 0.1855, 0.1796, 0.2394, 0.2682, 0.2682, 0.2442, 0.2125, 0.1925, 0.1892, 0.1838, 0.1931, 0.1956, 0.1795, 0.1796, 0.2003, 0.1735, 0.1889, 0.1843, 0.1107, 0.1285, 0.1065, 0.1009, 0.1014, 0.1077, 0.1197, 0.1326, 0.1227, 0.1456, 0.1722, 0.1936, 0.1593, 0.1597, 0.1416, 0.1504, 0.1393, 0.1319, 0.1129, 0.1021, 0.1041, 0.1173, 0.0819, 0.0718, 0.0733, 0.0697, 0.0574, 0.0299, 0.042, 0.0263, 0.021, 0.0342, 0.0315, 0.0406, 0.0406]}
{"unit_id": "orig:1930405301787996326", "ticker_idx": 2, "ts": "2025-06-04T23:24:23+00:00", "symbol": "MediaTek", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan: EDA export controls short-term headwind for Synopsys/Cadence, indirect benefit to Taiwan fabless (MediaTek, Novatek) and SK Hynix via CXMT yield delays.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "EDA Export Controls: Short-Term Disruption to China’s Fabless Sector, Long-Term Catalyst for Localization and Onshore Substitution\n\nJP Morgan (G. Hariharan, 2025/06/02)\n\nThe U.S. Bureau of Industry and Security (BIS) has imposed stricter licensing requirements on exports of EDA tools and IP to China, which may cause short-term disruptions for China’s advanced digital IC design sector. However, these measures are likely to accelerate the localization of domestic EDA tools and foster the rise of alternative suppliers, potentially reshaping the industry over the long term.\n\n⸻\n\n• Export Control Tightening and Targeted Entities\n\nSince 2022, the U.S. BIS has restricted exports of EDA software (ECAD) for advanced 3nm GAAFET process design.\nIn late May 2025, further measures were introduced requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers, especially those linked to the defense sector.\n\nThese three companies account for over 70% of the global EDA market, meaning that license delays or denials could severely limit access to critical tools and IP for Chinese logic IC design firms.\n\n⸻\n\n• Analog/Power IC Design and Domestic Substitution Potential\n\nThis round of export controls mainly targets digital chip design and IP licensing.\nThe impact on analog and power IC firms is relatively limited, with local Chinese EDA players like Empyrean, Primarius, and Semitronix expected to provide adequate tool support in these segments.\n\nIndustry research suggests that local substitution is less challenging in analog and power domains, whereas companies reliant on advanced digital design tools from Synopsys, Cadence, and Siemens will be more severely impacted.\n\n⸻\n\n• Accelerated Localization and Supply Chain Restructuring\n\nIn the short term, failure to secure U.S. licenses could result in project delays and development constraints for China’s fabless design firms.\n\nOver the medium to long term, however, the certification and adoption of domestic EDA tools is expected to accelerate.\nChinese firms like Empyrean and Primarius have seen 30%+ revenue CAGR and over 50% R&D CAGR in recent years, and increased support from the government and capital markets will further fuel the maturation of China’s indigenous EDA ecosystem, reducing reliance on imported software.\n\n⸻\n\n• TSMC Advanced Nodes and Design Service Intermediaries\n\nChina’s leading fabless firms—especially in non-AI domains such as mobile SoCs, ADAS, and crypto mining—remain heavily reliant on TSMC’s advanced nodes and corresponding PDKs.\n\nHowever, TSMC primarily supports mainstream U.S. EDA tools and does not fully integrate with Chinese domestic EDA tools.\nIf licenses for Synopsys, Cadence, and Siemens are denied, these firms may be locked out of leading-edge process access.\n\nDesign service providers such as Alchip and GUC could act as intermediaries to obtain licenses and maintain TSMC access on behalf of Chinese clients, but the sustainability of this workaround remains uncertain.\n\nReduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek.\n\n⸻\n\n• Impact on Memory Sector and Supply-Demand Dynamics\n\nIf EDA tools face similar restrictions as equipment exports, Chinese memory manufacturers could struggle to improve DDR5 yields.\n\nFor instance, ChangXin Memory Technologies (CXMT) is reportedly operating a 10Z4nm process comparable to global 1y/1znm nodes.\nHowever, intensified EDA controls could delay yield ramp-up in DDR5, sustaining a tight supply-demand environment for these products.\n\nIn the short term, DDR4 demand remains supported, while EDA-related delays in China’s capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.00026861954004631095, "alpha_c_m1d": 0.011757706789659927, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02352947850070075, "ret_signed_p1d": -0.02352947850070075, "alpha_spy_p1d": -0.018696709444511517, "alpha_c_p1d": -0.02174197388754051, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.03137243896375663, "ret_signed_p1w": 0.03137243896375663, "alpha_spy_p1w": 0.02226050780163158, "alpha_c_p1w": -0.008865870664874098, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.015686418519055945, "ret_signed_p1m": 0.015686418519055945, "alpha_spy_p1m": -0.03676732204688693, "alpha_c_p1m": -0.11090745496088106, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.08766413352050795, "ret_signed_p3m": 0.08766413352050795, "alpha_spy_p3m": 0.0020383142205513405, "alpha_c_p3m": -0.06542432789401609, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.07166910785764946, "ret_signed_p6m": 0.07166910785764946, "alpha_spy_p6m": -0.07541081230858171, "alpha_c_p6m": -0.30950263181601545, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.149, 0.0941, 0.0863, 0.1059, 0.0902, 0.098, 0.1176, 0.1569, 0.149, 0.1647, 0.1608, 0.1608, 0.1961, 0.1882, 0.1647, 0.149, 0.0902, 0.1451, 0.1255, 0.1255, 0.1255, 0.0157, -0.0118, -0.0706, 0.0196, 0.0863, 0.098, 0.0863, 0.0706, 0.051, 0.0706, 0.0549, 0.0196, 0.0745, 0.0431, 0.0824, 0.0588, 0.0745, 0.0588, 0.0588, 0.0196, 0.0157, 0.0039, 0.0, 0.0118, 0.0392, 0.0314, 0.0627, 0.0667, 0.0745, 0.0706, 0.0275, 0.0314, 0.051, 0.0392, 0.0314, 0.0039, 0.0039, 0.0039, -0.0118, -0.0118, -0.0118, 0.0, 0.0, -0.0235, 0.0039, 0.0078, 0.0471, 0.0314, 0.0078, -0.0196, -0.0157, 0.0, 0.0078, -0.0078, -0.0196, -0.0157, -0.0039, 0.0157, 0.0118, 0.0078, -0.0196, 0.0, 0.0157, 0.0157, 0.0317, 0.0237, 0.0157, 0.0837, 0.1197, 0.1356, 0.1077, 0.1157, 0.1277, 0.1117, 0.1277, 0.1396, 0.1436, 0.1436, 0.1436, 0.1436, 0.1197, 0.0877, 0.1037, 0.0957, 0.0957, 0.0637, 0.0717, 0.0557, 0.0837, 0.0797, 0.0837, 0.0877, 0.0957, 0.1117, 0.0957, 0.1277, 0.1117, 0.0837, 0.0917, 0.0917, 0.1277, 0.1197, 0.1197, 0.1077, 0.0957, 0.0877, 0.0877, 0.1157, 0.1077, 0.1476, 0.1436, 0.2076, 0.1916, 0.1836, 0.1876, 0.1876, 0.2276, 0.2076, 0.2076, 0.1516, 0.1237, 0.0997, 0.0637, 0.0757, 0.0477, 0.0477, 0.0517, 0.0277, 0.0237, 0.0557, 0.0557, 0.0597, 0.0677, 0.0757, 0.0757, 0.0517, 0.0437, 0.0517, 0.0637, 0.0637, 0.0717, 0.0717, 0.0637, 0.0357, 0.0357, 0.0677, 0.0517, 0.0397, 0.0477, 0.0477, 0.0397, 0.0517, 0.0637, 0.0317, 0.0077, -0.0003, -0.0083, -0.0043, -0.0043, -0.0163, -0.0163, -0.0643, -0.0723, -0.0523, -0.0843, -0.0803, -0.0523, 0.0397, 0.0717]}
{"unit_id": "orig:1930405301787996326", "ticker_idx": 3, "ts": "2025-06-04T23:24:23+00:00", "symbol": "Novatek", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan: EDA export controls short-term headwind for Synopsys/Cadence, indirect benefit to Taiwan fabless (MediaTek, Novatek) and SK Hynix via CXMT yield delays.", "resolved_tickers": ["3034.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Novatek Microelectronics listed on TWSE as 3034.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "EDA Export Controls: Short-Term Disruption to China’s Fabless Sector, Long-Term Catalyst for Localization and Onshore Substitution\n\nJP Morgan (G. Hariharan, 2025/06/02)\n\nThe U.S. Bureau of Industry and Security (BIS) has imposed stricter licensing requirements on exports of EDA tools and IP to China, which may cause short-term disruptions for China’s advanced digital IC design sector. However, these measures are likely to accelerate the localization of domestic EDA tools and foster the rise of alternative suppliers, potentially reshaping the industry over the long term.\n\n⸻\n\n• Export Control Tightening and Targeted Entities\n\nSince 2022, the U.S. BIS has restricted exports of EDA software (ECAD) for advanced 3nm GAAFET process design.\nIn late May 2025, further measures were introduced requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers, especially those linked to the defense sector.\n\nThese three companies account for over 70% of the global EDA market, meaning that license delays or denials could severely limit access to critical tools and IP for Chinese logic IC design firms.\n\n⸻\n\n• Analog/Power IC Design and Domestic Substitution Potential\n\nThis round of export controls mainly targets digital chip design and IP licensing.\nThe impact on analog and power IC firms is relatively limited, with local Chinese EDA players like Empyrean, Primarius, and Semitronix expected to provide adequate tool support in these segments.\n\nIndustry research suggests that local substitution is less challenging in analog and power domains, whereas companies reliant on advanced digital design tools from Synopsys, Cadence, and Siemens will be more severely impacted.\n\n⸻\n\n• Accelerated Localization and Supply Chain Restructuring\n\nIn the short term, failure to secure U.S. licenses could result in project delays and development constraints for China’s fabless design firms.\n\nOver the medium to long term, however, the certification and adoption of domestic EDA tools is expected to accelerate.\nChinese firms like Empyrean and Primarius have seen 30%+ revenue CAGR and over 50% R&D CAGR in recent years, and increased support from the government and capital markets will further fuel the maturation of China’s indigenous EDA ecosystem, reducing reliance on imported software.\n\n⸻\n\n• TSMC Advanced Nodes and Design Service Intermediaries\n\nChina’s leading fabless firms—especially in non-AI domains such as mobile SoCs, ADAS, and crypto mining—remain heavily reliant on TSMC’s advanced nodes and corresponding PDKs.\n\nHowever, TSMC primarily supports mainstream U.S. EDA tools and does not fully integrate with Chinese domestic EDA tools.\nIf licenses for Synopsys, Cadence, and Siemens are denied, these firms may be locked out of leading-edge process access.\n\nDesign service providers such as Alchip and GUC could act as intermediaries to obtain licenses and maintain TSMC access on behalf of Chinese clients, but the sustainability of this workaround remains uncertain.\n\nReduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek.\n\n⸻\n\n• Impact on Memory Sector and Supply-Demand Dynamics\n\nIf EDA tools face similar restrictions as equipment exports, Chinese memory manufacturers could struggle to improve DDR5 yields.\n\nFor instance, ChangXin Memory Technologies (CXMT) is reportedly operating a 10Z4nm process comparable to global 1y/1znm nodes.\nHowever, intensified EDA controls could delay yield ramp-up in DDR5, sustaining a tight supply-demand environment for these products.\n\nIn the short term, DDR4 demand remains supported, while EDA-related delays in China’s capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0010090873105889075, "ret_signed_m1d": -0.0010090873105889075, "alpha_spy_m1d": -0.0012777068506352185, "alpha_c_m1d": 0.01074861947907102, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0048327690561892345, "alpha_c_p1d": 0.0017875046131602401, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.03733603549907416, "ret_signed_p1w": 0.03733603549907416, "alpha_spy_p1w": 0.028224104336949107, "alpha_c_p1w": -0.00290227412955657, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.10393540801251078, "ret_signed_p1m": 0.10393540801251078, "alpha_spy_p1m": 0.0514816674465679, "alpha_c_p1m": -0.022658465467426225, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": -0.08842828720347296, "ret_signed_p3m": -0.08842828720347296, "alpha_spy_p3m": -0.17405410650342956, "alpha_c_p3m": -0.241516748617997, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": -0.1619077827443145, "ret_signed_p6m": -0.1619077827443145, "alpha_spy_p6m": -0.30898770291054567, "alpha_c_p6m": -0.5430795224179794, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.0979, 0.0938, 0.0777, 0.0979, 0.0979, 0.0878, 0.0898, 0.0999, 0.0898, 0.112, 0.1181, 0.1221, 0.1282, 0.1261, 0.1322, 0.1302, 0.0999, 0.114, 0.0716, 0.0716, 0.0716, -0.0353, -0.0414, -0.0747, 0.0172, 0.0293, 0.0151, 0.0091, -0.0091, -0.0091, 0.0071, -0.003, -0.0353, -0.0141, -0.0272, 0.0151, 0.0353, 0.0252, 0.0474, 0.0474, 0.0394, 0.0494, 0.0535, 0.0474, 0.0272, 0.0333, 0.0595, 0.0858, 0.0878, 0.0696, 0.0535, 0.0394, 0.0474, 0.0575, 0.0474, 0.0394, 0.0575, 0.0353, 0.0434, 0.0414, 0.0414, -0.0061, -0.001, 0.0, 0.0, 0.005, 0.003, 0.0212, 0.0373, 0.0232, 0.0293, 0.0434, 0.0454, 0.0595, 0.0333, 0.0474, 0.0595, 0.0656, 0.0797, 0.0716, 0.0878, 0.0999, 0.0898, 0.0878, 0.1039, 0.0938, 0.0918, 0.0777, 0.0841, 0.0564, 0.0564, 0.0543, 0.0607, 0.0639, 0.0639, 0.0564, 0.05, 0.0447, 0.0585, 0.0585, 0.0521, 0.0394, 0.034, 0.017, 0.0117, 0.0117, 0.0053, 0.0053, -0.0245, -0.0522, -0.0629, -0.0799, -0.0778, -0.065, -0.0735, -0.0842, -0.0959, -0.0884, -0.0927, -0.0938, -0.0991, -0.0884, -0.0831, -0.0725, -0.0735, -0.0735, -0.0884, -0.0799, -0.0778, -0.0671, -0.0661, -0.0693, -0.0799, -0.0959, -0.1129, -0.1119, -0.1023, -0.1012, -0.0969, -0.0661, -0.0448, -0.0703, -0.0565, -0.0597, -0.0618, -0.0746, -0.0746, -0.0916, -0.114, -0.1097, -0.1087, -0.1087, -0.0948, -0.0948, -0.0969, -0.0969, -0.1182, -0.1321, -0.1342, -0.1257, -0.1363, -0.1182, -0.1119, -0.1044, -0.1151, -0.1151, -0.1182, -0.1353, -0.1417, -0.1449, -0.163, -0.1608, -0.1566, -0.1619, -0.1779, -0.1896, -0.1832, -0.1704, -0.1683, -0.1651, -0.1694, -0.1619, -0.1651, -0.1651, -0.163, -0.1715, -0.1704, -0.163, -0.1566, -0.1619]}
{"unit_id": "orig:1930405301787996326", "ticker_idx": 4, "ts": "2025-06-04T23:24:23+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "EDA-related delays in China's capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan: EDA export controls short-term headwind for Synopsys/Cadence, indirect benefit to Taiwan fabless (MediaTek, Novatek) and SK Hynix via CXMT yield delays.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "EDA Export Controls: Short-Term Disruption to China’s Fabless Sector, Long-Term Catalyst for Localization and Onshore Substitution\n\nJP Morgan (G. Hariharan, 2025/06/02)\n\nThe U.S. Bureau of Industry and Security (BIS) has imposed stricter licensing requirements on exports of EDA tools and IP to China, which may cause short-term disruptions for China’s advanced digital IC design sector. However, these measures are likely to accelerate the localization of domestic EDA tools and foster the rise of alternative suppliers, potentially reshaping the industry over the long term.\n\n⸻\n\n• Export Control Tightening and Targeted Entities\n\nSince 2022, the U.S. BIS has restricted exports of EDA software (ECAD) for advanced 3nm GAAFET process design.\nIn late May 2025, further measures were introduced requiring Synopsys, Cadence, and Siemens EDA to obtain additional licenses for exporting tools and IP to Chinese customers, especially those linked to the defense sector.\n\nThese three companies account for over 70% of the global EDA market, meaning that license delays or denials could severely limit access to critical tools and IP for Chinese logic IC design firms.\n\n⸻\n\n• Analog/Power IC Design and Domestic Substitution Potential\n\nThis round of export controls mainly targets digital chip design and IP licensing.\nThe impact on analog and power IC firms is relatively limited, with local Chinese EDA players like Empyrean, Primarius, and Semitronix expected to provide adequate tool support in these segments.\n\nIndustry research suggests that local substitution is less challenging in analog and power domains, whereas companies reliant on advanced digital design tools from Synopsys, Cadence, and Siemens will be more severely impacted.\n\n⸻\n\n• Accelerated Localization and Supply Chain Restructuring\n\nIn the short term, failure to secure U.S. licenses could result in project delays and development constraints for China’s fabless design firms.\n\nOver the medium to long term, however, the certification and adoption of domestic EDA tools is expected to accelerate.\nChinese firms like Empyrean and Primarius have seen 30%+ revenue CAGR and over 50% R&D CAGR in recent years, and increased support from the government and capital markets will further fuel the maturation of China’s indigenous EDA ecosystem, reducing reliance on imported software.\n\n⸻\n\n• TSMC Advanced Nodes and Design Service Intermediaries\n\nChina’s leading fabless firms—especially in non-AI domains such as mobile SoCs, ADAS, and crypto mining—remain heavily reliant on TSMC’s advanced nodes and corresponding PDKs.\n\nHowever, TSMC primarily supports mainstream U.S. EDA tools and does not fully integrate with Chinese domestic EDA tools.\nIf licenses for Synopsys, Cadence, and Siemens are denied, these firms may be locked out of leading-edge process access.\n\nDesign service providers such as Alchip and GUC could act as intermediaries to obtain licenses and maintain TSMC access on behalf of Chinese clients, but the sustainability of this workaround remains uncertain.\n\nReduced competition may benefit Taiwanese fabless companies such as MediaTek and Novatek.\n\n⸻\n\n• Impact on Memory Sector and Supply-Demand Dynamics\n\nIf EDA tools face similar restrictions as equipment exports, Chinese memory manufacturers could struggle to improve DDR5 yields.\n\nFor instance, ChangXin Memory Technologies (CXMT) is reportedly operating a 10Z4nm process comparable to global 1y/1znm nodes.\nHowever, intensified EDA controls could delay yield ramp-up in DDR5, sustaining a tight supply-demand environment for these products.\n\nIn the short term, DDR4 demand remains supported, while EDA-related delays in China’s capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04597709938807859, "ret_signed_m1d": -0.04597709938807859, "alpha_spy_m1d": -0.0462457189281249, "alpha_c_m1d": -0.03421939259841866, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032183782009547635, "ret_signed_p1d": 0.032183782009547635, "alpha_spy_p1d": 0.03701655106573687, "alpha_c_p1d": 0.033971286622707875, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.1034481489964525, "ret_signed_p1w": 0.1034481489964525, "alpha_spy_p1w": 0.09433621783432744, "alpha_c_p1w": 0.06320983936782176, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.2804596714416854, "ret_signed_p1m": 0.2804596714416854, "alpha_spy_p1m": 0.2280059308757425, "alpha_c_p1m": 0.15386579796174837, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.17871148439176676, "ret_signed_p3m": 0.17871148439176676, "alpha_spy_p3m": 0.09308566509181015, "alpha_c_p3m": 0.025623022977242726, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 1.506555764559058, "ret_signed_p6m": 1.506555764559058, "alpha_spy_p6m": 1.3594758443928268, "alpha_c_p6m": 1.125384024885393, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.117, -0.1376, -0.1381, -0.0872, -0.0835, -0.0615, -0.0546, -0.0683, -0.0569, -0.0362, -0.011, -0.0293, -0.0454, -0.0179, -0.05, -0.0853, -0.1248, -0.0959, -0.0918, -0.1069, -0.1638, -0.2437, -0.2221, -0.2427, -0.1592, -0.1702, -0.173, -0.1712, -0.2014, -0.1969, -0.1969, -0.1895, -0.2024, -0.1693, -0.1817, -0.1537, -0.1647, -0.1702, -0.1854, -0.1854, -0.1464, -0.1464, -0.1464, -0.1243, -0.1266, -0.1276, -0.1051, -0.089, -0.0546, -0.0798, -0.0615, -0.0849, -0.0729, -0.0798, -0.0963, -0.0821, -0.0683, -0.0706, -0.0454, -0.0253, -0.0598, -0.046, -0.046, 0.0, 0.0322, 0.0322, 0.0529, 0.0598, 0.1034, 0.0828, 0.0828, 0.1402, 0.1448, 0.1333, 0.131, 0.1816, 0.1931, 0.2805, 0.3149, 0.3471, 0.3057, 0.3425, 0.3126, 0.2828, 0.2805, 0.2437, 0.246, 0.2966, 0.292, 0.3655, 0.354, 0.3793, 0.3724, 0.3609, 0.2391, 0.2368, 0.2529, 0.2345, 0.2368, 0.2391, 0.223, 0.2046, 0.2069, 0.2115, 0.2575, 0.1862, 0.1862, 0.2115, 0.1885, 0.2046, 0.1793, 0.2276, 0.2368, 0.2782, 0.2713, 0.2713, 0.2299, 0.2092, 0.1747, 0.1264, 0.154, 0.1931, 0.2023, 0.1954, 0.2363, 0.2386, 0.1787, 0.1994, 0.2086, 0.2225, 0.2593, 0.2754, 0.3261, 0.3997, 0.4135, 0.5125, 0.524, 0.6023, 0.5355, 0.6368, 0.6368, 0.6161, 0.6622, 0.6461, 0.6414, 0.5494, 0.6069, 0.6, 0.6576, 0.821, 0.821, 0.821, 0.821, 0.821, 0.821, 0.9707, 0.9108, 0.8947, 0.9453, 1.0835, 1.1433, 1.2354, 1.2055, 1.217, 1.2032, 1.3482, 1.4633, 1.3989, 1.5692, 1.6153, 1.5738, 1.8547, 1.6981, 1.6659, 1.7304, 1.6705, 1.7902, 1.8501, 1.8409, 1.8179, 1.5784, 1.7902, 1.6245, 1.5876, 1.6291, 1.3989, 1.3943, 1.3897, 1.4127, 1.5066]}
{"unit_id": "orig:1938004447101391036", "ticker_idx": 0, "ts": "2025-06-25T22:40:41+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a cautious approach seems necessary regarding stock price reaction", "tweet_type": "analysis", "timeline": "quarters", "summary": "Post-MU earnings: cautious on MU near-term reaction but bullish sequentially on NVDA/AVGO first, then memory IDMs (incl. SK Hynix), then materials/components.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250626_3Q25 Micron [Beat, $142bn, 11.5x]\n\n[Key Takeaways]\n\n(1) While strong earnings and guidance were presented, a cautious approach seems necessary regarding stock price reaction.\n(2) In the Q&A, management maintained a firm stance and, for the first time in an earnings conference call, officially mentioned HBM margins and bit demand for ASICs comparable to GPUs.\n(3) The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components.\n\n[Contents]\n\n(1) Revenue exceeded consensus by +5% and operating profit by +17%. After-hours trading saw a rise of up to +7.4% before paring gains. Currently at +0.6%.\n\n(2) FY4Q25 Guidance\n\nRevenue: $10.4~$11.0bn [$10.7 ± 0.3bn] vs. $9,890.6mn\n\nGPM: 35.5~37.5% vs. 39.16%\n\n(3) FY25 Guidance\nCAPEX: $14bn vs. $14,016mn\n\nDRAM: High-teens vs. +11.15%\nNAND: Low-double digits vs. +10.87%\nS/D: Bit supply growth to be below industry bit demand growth for non-HBM DRAM and NAND\n\n(4) The reduction in after-hours gains is believed to be due to: 1) Some profit-taking given the over 30% performance in the past month, and 2) While FY4Q25 revenue guidance exceeded market expectations, the GPM upper range was more than 150 basis points below market expectations.\n\n(5) HBM revenue reached an all-time high, increasing by approximately +50% QoQ. Estimated at around $1.5bn, this is slightly above 1/4 of SK Hynix's estimated 2Q25 HBM revenue (mid-to-high 6 trillion KRW).\n\n(6) The company mentioned that it expects to achieve HBM market share equivalent to its legacy DRAM share in the second half of this year, driven by the smooth mass production of HBM3E 12-hi.\n\n(7) In the Q&A session, the company projected the CY25 HBM TAM to be approximately $35bn, and reiterated its views on: 1) Custom HBM, 2) The launch of next-generation products HBM4/HBM4E, and 3) Maintaining the excess demand period.\n\n(8) It was also stated that the margins for HBM going into ASICs are not significantly different from those for GPUs, and the bit content for future ASICs is expected to be similar to GPUs.\n\n(9) CAPEX for the current quarter was somewhat reduced at $2.7bn, but the annual total of $14bn remains. Specific CAPEX guidance for FY26 has not yet been provided.\n\n(10) For HBM4, samples have already been provided to multiple customers, with mass production planned starting next year. The main node is 1-b.\n\n(11) Furthermore, the company stated that DRAM inventory will remain very tight in the next quarter (FY4Q25 = June-August 2025 by calendar year), and NAND inventory, which continues to suffer from market weakness, will also improve. However, NAND shows a disappointing performance consistent with guidance vs. DRAM High-teens vs. +11.15%.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005186703787299907, "ret_signed_m1d": 0.005186703787299907, "alpha_spy_m1d": 0.005746754585245539, "alpha_c_m1d": 0.018486137889198506, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009823196998103967, "ret_signed_p1d": -0.009823196998103967, "alpha_spy_p1d": -0.017646976872591313, "alpha_c_p1d": -0.017090671033344718, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": -0.04330064082854557, "ret_signed_p1w": -0.04330064082854557, "alpha_spy_p1w": -0.06525678147023772, "alpha_c_p1w": -0.06528460239557365, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": -0.12113815880729972, "ret_signed_p1m": -0.12113815880729972, "alpha_spy_p1m": -0.16610454004474628, "alpha_c_p1m": -0.1669230338782275, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.2948915937920622, "ret_signed_p3m": 0.2948915937920622, "alpha_spy_p3m": 0.19348042596955572, "alpha_c_p3m": 0.12421568967381957, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.9563045658042233, "ret_signed_p6m": 0.9563045658042233, "alpha_spy_p6m": 0.838987711996035, "alpha_c_p6m": 0.6944949622109877, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3059, -0.3172, -0.3029, -0.3037, -0.4158, -0.4914, -0.4627, -0.485, -0.3881, -0.4495, -0.4534, -0.4419, -0.4417, -0.4552, -0.4593, -0.4593, -0.4755, -0.4483, -0.4269, -0.3916, -0.373, -0.3826, -0.3958, -0.3953, -0.3888, -0.3657, -0.368, -0.3673, -0.3507, -0.3308, -0.3253, -0.2747, -0.2383, -0.2509, -0.2499, -0.2299, -0.2248, -0.2291, -0.2468, -0.2548, -0.2662, -0.2662, -0.2426, -0.2442, -0.2393, -0.2577, -0.2284, -0.1965, -0.1886, -0.1647, -0.1469, -0.1281, -0.103, -0.0882, -0.087, -0.0916, -0.0582, -0.0543, -0.0427, -0.0427, -0.0287, -0.0406, 0.0052, 0.0, -0.0098, -0.0196, -0.0314, -0.05, -0.0433, -0.039, -0.039, -0.0567, -0.0213, -0.0385, -0.0316, -0.0205, -0.067, -0.0552, -0.0842, -0.1091, -0.1002, -0.1093, -0.1409, -0.1361, -0.1211, -0.1248, -0.1249, -0.1193, -0.0975, -0.1415, -0.175, -0.1523, -0.1421, -0.1443, -0.12, -0.0648, -0.0268, 0.0049, -0.0225, -0.0145, -0.0492, -0.0282, -0.04, -0.078, -0.0892, -0.0743, -0.0842, -0.0836, -0.0738, -0.0404, -0.0639, -0.0639, -0.068, -0.0662, -0.023, 0.0333, 0.0341, 0.0638, 0.1012, 0.1844, 0.2368, 0.241, 0.2493, 0.2585, 0.3285, 0.28, 0.2949, 0.309, 0.272, 0.2336, 0.2371, 0.2892, 0.3161, 0.4328, 0.4454, 0.4784, 0.503, 0.4615, 0.5469, 0.5138, 0.4293, 0.5173, 0.4723, 0.5107, 0.5941, 0.5929, 0.6275, 0.5922, 0.5621, 0.627, 0.7239, 0.7324, 0.7466, 0.7838, 0.7632, 0.7613, 0.8473, 0.7161, 0.8693, 0.8759, 0.8726, 0.9937, 0.8977, 0.9276, 0.865, 0.9428, 0.9044, 0.7985, 0.7782, 0.585, 0.6322, 0.7625, 0.7672, 0.8123, 0.8123, 0.8613, 0.8926, 0.885, 0.843, 0.7839, 0.8671, 0.9435, 0.9868, 1.0756, 1.0343, 0.898, 0.8693, 0.8301, 0.775, 0.9563]}
{"unit_id": "orig:1938004447101391036", "ticker_idx": 1, "ts": "2025-06-25T22:40:41+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components", "tweet_type": "analysis", "timeline": "quarters", "summary": "Post-MU earnings: cautious on MU near-term reaction but bullish sequentially on NVDA/AVGO first, then memory IDMs (incl. SK Hynix), then materials/components.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250626_3Q25 Micron [Beat, $142bn, 11.5x]\n\n[Key Takeaways]\n\n(1) While strong earnings and guidance were presented, a cautious approach seems necessary regarding stock price reaction.\n(2) In the Q&A, management maintained a firm stance and, for the first time in an earnings conference call, officially mentioned HBM margins and bit demand for ASICs comparable to GPUs.\n(3) The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components.\n\n[Contents]\n\n(1) Revenue exceeded consensus by +5% and operating profit by +17%. After-hours trading saw a rise of up to +7.4% before paring gains. Currently at +0.6%.\n\n(2) FY4Q25 Guidance\n\nRevenue: $10.4~$11.0bn [$10.7 ± 0.3bn] vs. $9,890.6mn\n\nGPM: 35.5~37.5% vs. 39.16%\n\n(3) FY25 Guidance\nCAPEX: $14bn vs. $14,016mn\n\nDRAM: High-teens vs. +11.15%\nNAND: Low-double digits vs. +10.87%\nS/D: Bit supply growth to be below industry bit demand growth for non-HBM DRAM and NAND\n\n(4) The reduction in after-hours gains is believed to be due to: 1) Some profit-taking given the over 30% performance in the past month, and 2) While FY4Q25 revenue guidance exceeded market expectations, the GPM upper range was more than 150 basis points below market expectations.\n\n(5) HBM revenue reached an all-time high, increasing by approximately +50% QoQ. Estimated at around $1.5bn, this is slightly above 1/4 of SK Hynix's estimated 2Q25 HBM revenue (mid-to-high 6 trillion KRW).\n\n(6) The company mentioned that it expects to achieve HBM market share equivalent to its legacy DRAM share in the second half of this year, driven by the smooth mass production of HBM3E 12-hi.\n\n(7) In the Q&A session, the company projected the CY25 HBM TAM to be approximately $35bn, and reiterated its views on: 1) Custom HBM, 2) The launch of next-generation products HBM4/HBM4E, and 3) Maintaining the excess demand period.\n\n(8) It was also stated that the margins for HBM going into ASICs are not significantly different from those for GPUs, and the bit content for future ASICs is expected to be similar to GPUs.\n\n(9) CAPEX for the current quarter was somewhat reduced at $2.7bn, but the annual total of $14bn remains. Specific CAPEX guidance for FY26 has not yet been provided.\n\n(10) For HBM4, samples have already been provided to multiple customers, with mass production planned starting next year. The main node is 1-b.\n\n(11) Furthermore, the company stated that DRAM inventory will remain very tight in the next quarter (FY4Q25 = June-August 2025 by calendar year), and NAND inventory, which continues to suffer from market weakness, will also improve. However, NAND shows a disappointing performance consistent with guidance vs. DRAM High-teens vs. +11.15%.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04153985983831754, "ret_signed_m1d": -0.04153985983831754, "alpha_spy_m1d": -0.040979809040371906, "alpha_c_m1d": -0.02824042573641894, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004601045575293838, "ret_signed_p1d": 0.004601045575293838, "alpha_spy_p1d": -0.003222734299193508, "alpha_c_p1d": -0.0026664284599469124, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": 0.019052474587535784, "ret_signed_p1w": 0.019052474587535784, "alpha_spy_p1w": -0.002903666054156373, "alpha_c_p1w": -0.0029314869794923037, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": 0.12591537288427257, "ret_signed_p1m": 0.12591537288427257, "alpha_spy_p1m": 0.08094899164682601, "alpha_c_p1m": 0.0801304978133448, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.18994459598187352, "ret_signed_p3m": 0.18994459598187352, "alpha_spy_p3m": 0.08853342815936704, "alpha_c_p3m": 0.019268691863630893, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.1286339477687075, "ret_signed_p6m": 0.1286339477687075, "alpha_spy_p6m": 0.011317093960519164, "alpha_c_p6m": -0.13317565582452806, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.2893, -0.2977, -0.2862, -0.2845, -0.3403, -0.3889, -0.3673, -0.376, -0.2591, -0.3029, -0.2812, -0.2826, -0.2729, -0.3229, -0.3423, -0.3423, -0.372, -0.3592, -0.3344, -0.3103, -0.2807, -0.2954, -0.2935, -0.2942, -0.2768, -0.258, -0.2624, -0.2643, -0.2414, -0.2394, -0.2441, -0.203, -0.1581, -0.123, -0.1263, -0.1226, -0.1215, -0.1292, -0.1459, -0.1393, -0.1492, -0.1492, -0.122, -0.1264, -0.098, -0.1244, -0.1098, -0.0849, -0.0804, -0.0929, -0.0817, -0.0758, -0.0671, -0.0744, -0.0603, -0.08, -0.0623, -0.066, -0.0572, -0.0572, -0.0678, -0.0657, -0.0415, 0.0, 0.0046, 0.0223, 0.0238, -0.0065, 0.0191, 0.0326, 0.0326, 0.0255, 0.0369, 0.0555, 0.0634, 0.0688, 0.0632, 0.1062, 0.1106, 0.1211, 0.1173, 0.1106, 0.0824, 0.1067, 0.1259, 0.1244, 0.1454, 0.1374, 0.1618, 0.1527, 0.1258, 0.1665, 0.1552, 0.1627, 0.1715, 0.184, 0.1798, 0.187, 0.1768, 0.1796, 0.1694, 0.1795, 0.1382, 0.1367, 0.134, 0.1535, 0.1653, 0.178, 0.1769, 0.1676, 0.1288, 0.1288, 0.1067, 0.1057, 0.1124, 0.0824, 0.0907, 0.1066, 0.1492, 0.1482, 0.1524, 0.152, 0.1334, 0.1036, 0.1422, 0.145, 0.1899, 0.1564, 0.1469, 0.1516, 0.1548, 0.1785, 0.2092, 0.2135, 0.2242, 0.2159, 0.2025, 0.1992, 0.2256, 0.248, 0.187, 0.2205, 0.1667, 0.1654, 0.1783, 0.1874, 0.1837, 0.1741, 0.1684, 0.1805, 0.2071, 0.241, 0.3028, 0.3418, 0.3149, 0.3123, 0.3408, 0.2877, 0.2651, 0.2189, 0.2194, 0.29, 0.2518, 0.256, 0.211, 0.2325, 0.2093, 0.1754, 0.2088, 0.1707, 0.1593, 0.1831, 0.1524, 0.1682, 0.1682, 0.1471, 0.166, 0.176, 0.1639, 0.1885, 0.1822, 0.2026, 0.1988, 0.1911, 0.1726, 0.1343, 0.1426, 0.1518, 0.1079, 0.1286]}
{"unit_id": "orig:1938004447101391036", "ticker_idx": 2, "ts": "2025-06-25T22:40:41+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components", "tweet_type": "analysis", "timeline": "quarters", "summary": "Post-MU earnings: cautious on MU near-term reaction but bullish sequentially on NVDA/AVGO first, then memory IDMs (incl. SK Hynix), then materials/components.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250626_3Q25 Micron [Beat, $142bn, 11.5x]\n\n[Key Takeaways]\n\n(1) While strong earnings and guidance were presented, a cautious approach seems necessary regarding stock price reaction.\n(2) In the Q&A, management maintained a firm stance and, for the first time in an earnings conference call, officially mentioned HBM margins and bit demand for ASICs comparable to GPUs.\n(3) The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components.\n\n[Contents]\n\n(1) Revenue exceeded consensus by +5% and operating profit by +17%. After-hours trading saw a rise of up to +7.4% before paring gains. Currently at +0.6%.\n\n(2) FY4Q25 Guidance\n\nRevenue: $10.4~$11.0bn [$10.7 ± 0.3bn] vs. $9,890.6mn\n\nGPM: 35.5~37.5% vs. 39.16%\n\n(3) FY25 Guidance\nCAPEX: $14bn vs. $14,016mn\n\nDRAM: High-teens vs. +11.15%\nNAND: Low-double digits vs. +10.87%\nS/D: Bit supply growth to be below industry bit demand growth for non-HBM DRAM and NAND\n\n(4) The reduction in after-hours gains is believed to be due to: 1) Some profit-taking given the over 30% performance in the past month, and 2) While FY4Q25 revenue guidance exceeded market expectations, the GPM upper range was more than 150 basis points below market expectations.\n\n(5) HBM revenue reached an all-time high, increasing by approximately +50% QoQ. Estimated at around $1.5bn, this is slightly above 1/4 of SK Hynix's estimated 2Q25 HBM revenue (mid-to-high 6 trillion KRW).\n\n(6) The company mentioned that it expects to achieve HBM market share equivalent to its legacy DRAM share in the second half of this year, driven by the smooth mass production of HBM3E 12-hi.\n\n(7) In the Q&A session, the company projected the CY25 HBM TAM to be approximately $35bn, and reiterated its views on: 1) Custom HBM, 2) The launch of next-generation products HBM4/HBM4E, and 3) Maintaining the excess demand period.\n\n(8) It was also stated that the margins for HBM going into ASICs are not significantly different from those for GPUs, and the bit content for future ASICs is expected to be similar to GPUs.\n\n(9) CAPEX for the current quarter was somewhat reduced at $2.7bn, but the annual total of $14bn remains. Specific CAPEX guidance for FY26 has not yet been provided.\n\n(10) For HBM4, samples have already been provided to multiple customers, with mass production planned starting next year. The main node is 1-b.\n\n(11) Furthermore, the company stated that DRAM inventory will remain very tight in the next quarter (FY4Q25 = June-August 2025 by calendar year), and NAND inventory, which continues to suffer from market weakness, will also improve. However, NAND shows a disappointing performance consistent with guidance vs. DRAM High-teens vs. +11.15%.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003325099929278874, "ret_signed_m1d": -0.003325099929278874, "alpha_spy_m1d": -0.0027650491313332415, "alpha_c_m1d": 0.009974334172619725, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02085763478710234, "ret_signed_p1d": 0.02085763478710234, "alpha_spy_p1d": 0.013033854912614995, "alpha_c_p1d": 0.01359016075186159, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": 0.01983740863252592, "ret_signed_p1w": 0.01983740863252592, "alpha_spy_p1w": -0.0021187320091662354, "alpha_c_p1w": -0.002146552934502166, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": 0.09091239090585002, "ret_signed_p1m": 0.09091239090585002, "alpha_spy_p1m": 0.045946009668403454, "alpha_c_p1m": 0.04512751583492225, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.28233694763457606, "ret_signed_p3m": 0.28233694763457606, "alpha_spy_p3m": 0.1809257798120696, "alpha_c_p3m": 0.11166104351633344, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.24861219321601213, "ret_signed_p6m": 0.24861219321601213, "alpha_spy_p6m": 0.13129533940782379, "alpha_c_p6m": -0.01319741037722344, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3625, -0.3688, -0.3647, -0.3513, -0.4194, -0.4485, -0.4189, -0.4118, -0.302, -0.3505, -0.3141, -0.3276, -0.3254, -0.3418, -0.3554, -0.3554, -0.3734, -0.3607, -0.3331, -0.2907, -0.275, -0.2744, -0.2793, -0.2744, -0.2561, -0.2323, -0.2433, -0.2457, -0.2279, -0.2168, -0.2151, -0.1647, -0.1238, -0.125, -0.123, -0.1382, -0.1306, -0.1266, -0.134, -0.131, -0.1378, -0.1378, -0.1117, -0.0974, -0.0878, -0.0875, -0.0624, -0.0318, -0.0158, -0.0201, -0.0691, -0.0791, -0.0778, -0.0466, -0.0347, -0.0625, -0.0497, -0.0599, -0.0528, -0.0528, -0.0554, -0.0411, -0.0033, 0.0, 0.0209, 0.0178, 0.0416, 0.0003, 0.0198, 0.0398, 0.0398, 0.036, 0.027, 0.0501, 0.0406, 0.0368, 0.0414, 0.0616, 0.0611, 0.0824, 0.0706, 0.089, 0.0527, 0.0719, 0.0909, 0.0965, 0.112, 0.1238, 0.1435, 0.1098, 0.0906, 0.125, 0.1069, 0.1399, 0.1478, 0.1524, 0.1483, 0.1821, 0.1679, 0.176, 0.1575, 0.1553, 0.1143, 0.1002, 0.0943, 0.1109, 0.1118, 0.1261, 0.1345, 0.1663, 0.1237, 0.1237, 0.1269, 0.1426, 0.1566, 0.2654, 0.3061, 0.2721, 0.3964, 0.3589, 0.3598, 0.3757, 0.3603, 0.308, 0.3049, 0.3034, 0.2823, 0.2829, 0.2843, 0.2722, 0.2662, 0.2411, 0.2487, 0.2619, 0.28, 0.2807, 0.2698, 0.2733, 0.3077, 0.3059, 0.2287, 0.3501, 0.3025, 0.3298, 0.3405, 0.3222, 0.3219, 0.297, 0.2881, 0.3032, 0.3404, 0.3704, 0.4117, 0.461, 0.425, 0.3991, 0.3723, 0.3321, 0.3588, 0.3459, 0.3226, 0.3565, 0.3322, 0.3445, 0.2868, 0.2962, 0.2969, 0.2888, 0.3415, 0.3127, 0.2877, 0.4306, 0.4574, 0.5048, 0.5048, 0.5252, 0.4613, 0.4443, 0.4406, 0.4422, 0.4771, 0.5182, 0.5378, 0.5631, 0.5381, 0.3624, 0.2862, 0.2918, 0.234, 0.2486]}
{"unit_id": "orig:1938004447101391036", "ticker_idx": 3, "ts": "2025-06-25T22:40:41+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components", "tweet_type": "analysis", "timeline": "quarters", "summary": "Post-MU earnings: cautious on MU near-term reaction but bullish sequentially on NVDA/AVGO first, then memory IDMs (incl. SK Hynix), then materials/components.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250626_3Q25 Micron [Beat, $142bn, 11.5x]\n\n[Key Takeaways]\n\n(1) While strong earnings and guidance were presented, a cautious approach seems necessary regarding stock price reaction.\n(2) In the Q&A, management maintained a firm stance and, for the first time in an earnings conference call, officially mentioned HBM margins and bit demand for ASICs comparable to GPUs.\n(3) The sequential positive alignment is expected to continue in the order of NVDA/AVGO > Memory IDM > Materials & Components.\n\n[Contents]\n\n(1) Revenue exceeded consensus by +5% and operating profit by +17%. After-hours trading saw a rise of up to +7.4% before paring gains. Currently at +0.6%.\n\n(2) FY4Q25 Guidance\n\nRevenue: $10.4~$11.0bn [$10.7 ± 0.3bn] vs. $9,890.6mn\n\nGPM: 35.5~37.5% vs. 39.16%\n\n(3) FY25 Guidance\nCAPEX: $14bn vs. $14,016mn\n\nDRAM: High-teens vs. +11.15%\nNAND: Low-double digits vs. +10.87%\nS/D: Bit supply growth to be below industry bit demand growth for non-HBM DRAM and NAND\n\n(4) The reduction in after-hours gains is believed to be due to: 1) Some profit-taking given the over 30% performance in the past month, and 2) While FY4Q25 revenue guidance exceeded market expectations, the GPM upper range was more than 150 basis points below market expectations.\n\n(5) HBM revenue reached an all-time high, increasing by approximately +50% QoQ. Estimated at around $1.5bn, this is slightly above 1/4 of SK Hynix's estimated 2Q25 HBM revenue (mid-to-high 6 trillion KRW).\n\n(6) The company mentioned that it expects to achieve HBM market share equivalent to its legacy DRAM share in the second half of this year, driven by the smooth mass production of HBM3E 12-hi.\n\n(7) In the Q&A session, the company projected the CY25 HBM TAM to be approximately $35bn, and reiterated its views on: 1) Custom HBM, 2) The launch of next-generation products HBM4/HBM4E, and 3) Maintaining the excess demand period.\n\n(8) It was also stated that the margins for HBM going into ASICs are not significantly different from those for GPUs, and the bit content for future ASICs is expected to be similar to GPUs.\n\n(9) CAPEX for the current quarter was somewhat reduced at $2.7bn, but the annual total of $14bn remains. Specific CAPEX guidance for FY26 has not yet been provided.\n\n(10) For HBM4, samples have already been provided to multiple customers, with mass production planned starting next year. The main node is 1-b.\n\n(11) Furthermore, the company stated that DRAM inventory will remain very tight in the next quarter (FY4Q25 = June-August 2025 by calendar year), and NAND inventory, which continues to suffer from market weakness, will also improve. However, NAND shows a disappointing performance consistent with guidance vs. DRAM High-teens vs. +11.15%.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026223746874970866, "ret_signed_m1d": -0.026223746874970866, "alpha_spy_m1d": -0.025663696077025233, "alpha_c_m1d": -0.012924312773072266, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024475540972238585, "ret_signed_p1d": 0.024475540972238585, "alpha_spy_p1d": 0.01665176109775124, "alpha_c_p1d": 0.017208066936997835, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": -0.024475540972238585, "ret_signed_p1w": -0.024475540972238585, "alpha_spy_p1w": -0.04643168161393074, "alpha_c_p1w": -0.04645950253926667, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": -0.05769233090280079, "ret_signed_p1m": -0.05769233090280079, "alpha_spy_p1m": -0.10265871214024735, "alpha_c_p1m": -0.10347720597372856, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.22904547606599368, "ret_signed_p3m": 0.22904547606599368, "alpha_spy_p3m": 0.1276343082434872, "alpha_c_p3m": 0.05836957194775105, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.934242090254298, "ret_signed_p6m": 0.934242090254298, "alpha_spy_p6m": 0.8169252364461097, "alpha_c_p6m": 0.6724324866610625, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3044, -0.3344, -0.3124, -0.3093, -0.3208, -0.3641, -0.4248, -0.4084, -0.4241, -0.3606, -0.369, -0.3711, -0.3697, -0.3927, -0.3892, -0.3892, -0.3836, -0.3934, -0.3683, -0.3777, -0.3564, -0.3648, -0.369, -0.3805, -0.3805, -0.3508, -0.3508, -0.3508, -0.3341, -0.3358, -0.3365, -0.3194, -0.3072, -0.281, -0.3002, -0.2863, -0.3041, -0.295, -0.3002, -0.3128, -0.302, -0.2915, -0.2932, -0.274, -0.2587, -0.285, -0.2745, -0.2745, -0.2395, -0.215, -0.215, -0.1993, -0.1941, -0.1608, -0.1766, -0.1766, -0.1329, -0.1294, -0.1381, -0.1399, -0.1014, -0.0927, -0.0262, 0.0, 0.0245, -0.007, 0.021, -0.0017, -0.0245, -0.0262, -0.0542, -0.0524, -0.014, -0.0175, 0.0385, 0.0297, 0.049, 0.0437, 0.035, -0.0577, -0.0594, -0.0472, -0.0612, -0.0594, -0.0577, -0.0699, -0.0839, -0.0822, -0.0787, -0.0437, -0.0979, -0.0979, -0.0787, -0.0962, -0.0839, -0.1031, -0.0664, -0.0594, -0.028, -0.0332, -0.0332, -0.0647, -0.0804, -0.1066, -0.1434, -0.1224, -0.0927, -0.0857, -0.0909, -0.0598, -0.0581, -0.1036, -0.0878, -0.0808, -0.0703, -0.0423, -0.0301, 0.0084, 0.0645, 0.075, 0.1503, 0.159, 0.2185, 0.1678, 0.2448, 0.2448, 0.229, 0.2641, 0.2518, 0.2483, 0.1783, 0.222, 0.2168, 0.2606, 0.3849, 0.3849, 0.3849, 0.3849, 0.3849, 0.3849, 0.4987, 0.4531, 0.4409, 0.4794, 0.5845, 0.63, 0.7, 0.6772, 0.686, 0.6755, 0.7858, 0.8733, 0.8243, 0.9539, 0.9889, 0.9574, 1.171, 1.0519, 1.0274, 1.0764, 1.0309, 1.1219, 1.1675, 1.1605, 1.143, 0.9609, 1.1219, 0.9959, 0.9679, 0.9994, 0.8243, 0.8208, 0.8173, 0.8348, 0.9062, 0.8572, 0.8852, 0.9553, 0.9342, 0.8992, 0.9062, 1.0218, 0.9833, 1.0569, 0.9798, 1.0008, 0.9413, 0.8572, 0.9307, 0.9342]}
{"unit_id": "orig:1930041494783832179", "ticker_idx": 0, "ts": "2025-06-03T23:18:45+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": "AI networking", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Customers are not just buying chips — they are adopting end-to-end AI platforms from NVIDIA", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Citi bus tour note endorses NVDA full-stack AI infrastructure moat: networking revenue +64% QoQ, Spectrum-X expanding, Infiniband dominance, and end-to-end platform adoption.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250604_NVIDIA: Silicon Valley Bus Tour – CITI\n\n[Key Takeaways]\n1. NVIDIA’s AI infrastructure advantage lies in its comprehensive full-stack AI solution.\n2. Especially in the networking division, NVIDIA shows strength in both AI server scale-up and scale-out. Its Infiniband and Ethernet offerings each provide significant value depending on customer needs and use cases.\n\n⸻\n\n[Contents]\n1. On June 3rd, during our Silicon Valley Bus Tour, we held a meeting with Kevin Deierling, VP of NVIDIA’s Networking Division.\n2. NVIDIA is not merely a GPU company — it vertically integrates all layers of AI infrastructure, including:\n• Compute (e.g., CPUs),\n• Networking (e.g., Infiniband, Ethernet),\n• Storage (e.g., DPU, CUDA-based software).\n3. NVIDIA’s architecture is optimized for AI workloads, and its powerful software ecosystem is a key differentiator over competitors.\n4. For AI cluster networking infrastructure, Infiniband provides:\n• Up to 20x greater network efficiency than traditional Ethernet,\n• Significantly lower latency.\n5. In large-scale AI training/inference environments, Infiniband remains the preferred choice, although Ethernet adoption is steadily increasing.\n6. Spectrum-X is NVIDIA’s solution for Ethernet-based AI networking and is already being adopted by several hyperscalers.\n7. AI infrastructure is no longer just about compute power — network bandwidth and latency optimization are now essential.\n8. NVIDIA leverages DPUs (e.g., BlueField) to offload network processing and security tasks, thereby maximizing system efficiency.\n9. The integration with NVIDIA’s software stack — CUDA, NCCL, Magnum IO — enables real performance realization of the hardware.\n10.Customers are not just buying chips — they are adopting end-to-end AI platforms from NVIDIA.\n11.NVIDIA’s AI networking strategy focuses on AI-optimized execution, using a unified networking operating system to streamline compute-storage communication.\n12.The ultimate goal is to minimize Total Cost of Ownership (TCO) per token per second per user.\n13.This networking OS is called “Dynamo”, and is referred to as the OS for AI factories (AI server datacenters).\n14.After acquiring Mellanox, NVIDIA discontinued traditional campus switch development and focused solely on AI-optimized switches, recognizing that AI’s “agentic flow” demands fundamentally different networking requirements than telecom or web services.\nhttps://t.co/aWtKgnHJmj a result, NVIDIA’s networking revenue grew 64% QoQ, nearing $5 billion.\n16.From a scale-up perspective, NVLink (a direct GPU-to-GPU interconnect) generated over $1 billion in revenue.\n17. From a scale-out perspective, the Spectrum-X product line (Ethernet-based) is rapidly expanding, with two new customer wins. The run rate of Spectrum-X — including NICs and switches — is estimated at $2 billion per quarter.\n18.Infiniband vs. Ethernet:\n• Infiniband remains the gold standard in AI and HPC.\n• Ethernet is evolving, and hyperscalers prefer it for its familiarity.\n• vs. Arista (EOS OS): Arista is attempting to shift from traditional DC switching to AI workloads, but lacks NVIDIA’s end-to-end stack.\n• vs. Broadcom: While Jericho offers programmable features, it introduces latency, which is detrimental to AI performance.\n19. NVLink currently uses up to 3 miles (4.8km) of copper cable. On a PCB, even 10–15 inch copper traces can introduce signal noise, requiring DSPs and retimers — increasing complexity and cost.\n20.The solution is CPO (Co-Packaged Optics) — by placing optical transceivers directly next to the ASIC, electrical signals are immediately converted to optical, eliminating the need for DSPs and retimers and enabling efficient large-scale AI expansion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02719151462371494, "ret_signed_m1d": -0.02719151462371494, "alpha_spy_m1d": -0.021521259971801765, "alpha_c_m1d": -0.005124746957045678, "bench_spy_m1d": -0.005670254651913176, "bench_c_m1d": -0.022066767666669262, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00495687095373043, "ret_signed_p1d": 0.00495687095373043, "alpha_spy_p1d": 0.005225418356696876, "alpha_c_p1d": -0.00694072427209691, "bench_spy_p1d": -0.00026854740296644586, "bench_c_p1d": 0.01189759522582734, "ret_p1w": 0.01940242388990243, "ret_signed_p1w": 0.01940242388990243, "alpha_spy_p1w": 0.0076761301371559565, "alpha_c_p1w": -0.0339759646730009, "bench_spy_p1w": 0.011726293752746475, "bench_c_p1w": 0.053378388562903334, "ret_p1m": 0.11358815962237645, "ret_signed_p1m": 0.11358815962237645, "alpha_spy_p1m": 0.06964477351614695, "alpha_c_p1m": -0.01688327455202554, "bench_spy_p1m": 0.0439433861062295, "bench_c_p1m": 0.130471434174402, "ret_p3m": 0.23348034808441498, "ret_signed_p3m": 0.23348034808441498, "alpha_spy_p3m": 0.1481460707788247, "alpha_c_p3m": 0.06667290689640892, "bench_spy_p3m": 0.08533427730559029, "bench_c_p3m": 0.16680744118800606, "ret_p6m": 0.2766087459024622, "ret_signed_p6m": 0.2766087459024622, "alpha_spy_p6m": 0.12983687106978659, "alpha_c_p6m": -0.12099561606719167, "bench_spy_p6m": 0.14677187483267562, "bench_c_p6m": 0.3976043619696539, "price_path": [-0.2171, -0.2021, -0.2425, -0.2299, -0.1804, -0.1816, -0.1384, -0.1536, -0.1826, -0.1678, -0.1607, -0.1665, -0.1403, -0.1454, -0.1944, -0.2109, -0.2234, -0.2325, -0.22, -0.2181, -0.2791, -0.3322, -0.3086, -0.3181, -0.1904, -0.2383, -0.2145, -0.216, -0.2055, -0.2601, -0.2813, -0.2813, -0.3138, -0.2997, -0.2727, -0.2464, -0.2139, -0.2301, -0.228, -0.2287, -0.2097, -0.1892, -0.194, -0.196, -0.1711, -0.1689, -0.174, -0.129, -0.0799, -0.0416, -0.0452, -0.0412, -0.04, -0.0484, -0.0667, -0.0594, -0.0703, -0.0703, -0.0405, -0.0454, -0.0144, -0.0431, -0.0272, 0.0, 0.005, -0.0087, 0.0035, 0.01, 0.0194, 0.0115, 0.0268, 0.0054, 0.0246, 0.0206, 0.0302, 0.0302, 0.0187, 0.021, 0.0474, 0.0928, 0.0978, 0.1171, 0.1188, 0.0856, 0.1136, 0.1284, 0.1284, 0.1206, 0.1331, 0.1535, 0.1621, 0.1679, 0.1619, 0.2088, 0.2136, 0.2251, 0.2209, 0.2137, 0.1828, 0.2094, 0.2304, 0.2287, 0.2517, 0.2429, 0.2695, 0.2596, 0.2302, 0.2747, 0.2624, 0.2706, 0.2801, 0.2938, 0.2893, 0.2971, 0.286, 0.289, 0.2779, 0.2889, 0.2438, 0.2421, 0.2391, 0.2605, 0.2734, 0.2872, 0.286, 0.2759, 0.2335, 0.2335, 0.2094, 0.2083, 0.2156, 0.1828, 0.1919, 0.2093, 0.2558, 0.2547, 0.2593, 0.2588, 0.2385, 0.206, 0.2481, 0.2512, 0.3003, 0.2636, 0.2533, 0.2584, 0.2619, 0.2879, 0.3214, 0.326, 0.3377, 0.3287, 0.314, 0.3105, 0.3393, 0.3638, 0.2971, 0.3337, 0.275, 0.2736, 0.2876, 0.2976, 0.2935, 0.283, 0.2768, 0.2901, 0.3191, 0.3561, 0.4237, 0.4663, 0.4369, 0.434, 0.4651, 0.4071, 0.3825, 0.332, 0.3325, 0.4097, 0.368, 0.3725, 0.3234, 0.3468, 0.3215, 0.2844, 0.3209, 0.2793, 0.2668, 0.2928, 0.2593, 0.2766]}
{"unit_id": "thread:1932315591727329435", "ticker_idx": 0, "ts": "2025-06-10T05:55:12+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We view the success of the iPhone 17 as highly likely to fail", "tweet_type": "analysis", "timeline": "months", "summary": "iPhone 17 likely to disappoint on AI; bearish AAPL shipments and negative read-through for memory suppliers MU and SK Hynix in H2.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Korean Memory Industry Worried About Memory Demand Amid Potential iPhone 17 Weakness\n\nAccording to South Korean media reports, Apple unveiled new AI features at its annual event, \"WWDC 2025,\" but some are calling it \"an empty shell.\"\n\nApple announced features such as real-time call translation and automatic spam call and message classification for the iPhone.\n\nHowever, these features appear to be merely echoing existing AI functions in the market. Furthermore, real-time call translation is a technology that Samsung Electronics already introduced in its Galaxy S24 series last year.\n\nIndustry experts commented that this event could further exacerbate Apple's disadvantage in the AI market.\n\nGuo Ming-chi, a TF International analyst from Taiwan and a recognized Apple expert, stated, \"The most important part of WWDC 2025 for investors is Apple's AI. In a situation where the AI industry is accelerating, if Apple does not significantly increase its investment, it risks falling behind competitors.\"\n\nThe negative impact on the memory industry could also grow.\n\nUntil now, there were high expectations that Apple would aggressively enter the AI smartphone competition by significantly expanding memory capacity in the iPhone 17 series, leading to substantial benefits for the memory industry.\n\nFor example, AI smartphones that perform AI functions on the device essentially require larger memory capacities. Consequently, predictions emerged that all iPhone 17 series models would be equipped with 12GB of DRAM.\n\nNotably, the average selling price (ASP) of 12GB DRAM is 50% higher compared to the previous 8GB models.\n\nThis led to increased revenue expectations for major memory manufacturers like SK hynix and Micron, which supply low-power DRAM (LPDDR) and NAND flash to Apple.\n\nHowever, if there's no innovation in AI features and Apple's smartphone shipments underperform compared to previous years, it could negatively impact memory performance in the second half of the year.\n\nAn industry insider stated, \"It is highly probable that Apple's market share in China will continue to decline, and there's a high possibility of extended deferred consumer demand into 2026. We view the success of the iPhone 17 as highly likely to fail.\"\n\nNotably, Apple, along with Samsung Electronics, is a major purchaser of semiconductors globally. There are also concerns that Apple might push for lower supply prices, citing slowing smartphone demand.\n\n$AAPL\n\n---\n\nhttps://t.co/CaaZL06hgK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006019488733489342, "ret_signed_m1d": 0.006019488733489342, "alpha_spy_m1d": 0.0003817879376147859, "alpha_c_m1d": 0.0004563360335200217, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.00556315269996932, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019193715350202756, "ret_signed_p1d": 0.019193715350202756, "alpha_spy_p1d": 0.016341800759414604, "alpha_c_p1d": 0.01699333966107619, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0022003756891265658, "ret_p1w": -0.03468681628714798, "ret_signed_p1w": 0.03468681628714798, "alpha_spy_p1w": 0.025484222252788236, "alpha_c_p1w": 0.03655513192207971, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": 0.0018683156349317276, "ret_p1m": 0.041792195511688, "ret_signed_p1m": -0.041792195511688, "alpha_spy_p1m": -0.003944954948904789, "alpha_c_p1m": 0.030393404396976464, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.07218559990866447, "ret_p3m": 0.1840037128595935, "ret_signed_p3m": -0.1840037128595935, "alpha_spy_p3m": -0.10760679677631013, "alpha_c_p3m": -0.09090100538513313, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.09310270747446037, "ret_p6m": 0.40498477100190144, "ret_signed_p6m": -0.40498477100190144, "alpha_spy_p6m": -0.26448355059537776, "alpha_c_p6m": -0.19751209460041297, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.20747267640148848, "price_path": [0.0332, 0.052, 0.0545, 0.0481, 0.0606, 0.055, 0.0756, 0.0877, 0.1026, 0.0916, 0.1031, 0.0737, 0.0946, 0.0998, 0.1033, 0.0013, -0.0717, -0.1058, -0.1504, -0.0201, -0.0617, -0.0236, -0.002, -0.0039, -0.0427, -0.0293, -0.0293, -0.0482, -0.0157, 0.0082, 0.0268, 0.0313, 0.0355, 0.0408, 0.0471, 0.0512, 0.0119, -0.0199, -0.0218, -0.0329, -0.0268, -0.0217, 0.0401, 0.0506, 0.0477, 0.0433, 0.0424, 0.0301, 0.0207, -0.0029, -0.0065, -0.0365, -0.0365, -0.0121, -0.0111, -0.0134, -0.009, -0.0048, 0.003, 0.0007, -0.0101, 0.0062, -0.006, 0.0, -0.0192, -0.0171, -0.0307, -0.021, -0.0347, -0.03, -0.03, -0.0082, -0.0058, -0.0117, -0.0055, -0.0082, -0.0078, 0.0123, 0.0254, 0.0482, 0.0537, 0.0537, 0.0359, 0.0362, 0.0418, 0.0481, 0.0419, 0.0294, 0.0318, 0.037, 0.0363, 0.042, 0.0484, 0.0579, 0.0566, 0.0547, 0.0553, 0.0562, 0.0424, 0.0315, 0.0242, -0.0014, 0.0034, 0.0012, 0.0522, 0.0857, 0.1316, 0.1222, 0.1344, 0.1526, 0.1499, 0.144, 0.1405, 0.1389, 0.1164, 0.1109, 0.1251, 0.1221, 0.1327, 0.1386, 0.1488, 0.1467, 0.1467, 0.1348, 0.178, 0.1844, 0.184, 0.1751, 0.1576, 0.1203, 0.1363, 0.1562, 0.1692, 0.1764, 0.1805, 0.1751, 0.2127, 0.265, 0.2568, 0.2463, 0.2689, 0.2619, 0.2568, 0.2578, 0.2619, 0.2702, 0.2745, 0.268, 0.2669, 0.2747, 0.2549, 0.2116, 0.2234, 0.2239, 0.2317, 0.2223, 0.2462, 0.2954, 0.298, 0.2767, 0.2823, 0.2983, 0.3278, 0.3288, 0.3322, 0.3406, 0.3356, 0.329, 0.3339, 0.3344, 0.3326, 0.3262, 0.3322, 0.361, 0.3522, 0.3496, 0.3469, 0.3225, 0.3224, 0.3279, 0.3165, 0.3424, 0.3643, 0.3695, 0.3724, 0.3724, 0.3788, 0.3998, 0.4151, 0.405]}
{"unit_id": "thread:1932315591727329435", "ticker_idx": 1, "ts": "2025-06-10T05:55:12+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if there's no innovation in AI features and Apple's smartphone shipments underperform compared to previous years, it could negatively impact memory performance in the second half of the year", "tweet_type": "analysis", "timeline": "months", "summary": "iPhone 17 likely to disappoint on AI; bearish AAPL shipments and negative read-through for memory suppliers MU and SK Hynix in H2.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Korean Memory Industry Worried About Memory Demand Amid Potential iPhone 17 Weakness\n\nAccording to South Korean media reports, Apple unveiled new AI features at its annual event, \"WWDC 2025,\" but some are calling it \"an empty shell.\"\n\nApple announced features such as real-time call translation and automatic spam call and message classification for the iPhone.\n\nHowever, these features appear to be merely echoing existing AI functions in the market. Furthermore, real-time call translation is a technology that Samsung Electronics already introduced in its Galaxy S24 series last year.\n\nIndustry experts commented that this event could further exacerbate Apple's disadvantage in the AI market.\n\nGuo Ming-chi, a TF International analyst from Taiwan and a recognized Apple expert, stated, \"The most important part of WWDC 2025 for investors is Apple's AI. In a situation where the AI industry is accelerating, if Apple does not significantly increase its investment, it risks falling behind competitors.\"\n\nThe negative impact on the memory industry could also grow.\n\nUntil now, there were high expectations that Apple would aggressively enter the AI smartphone competition by significantly expanding memory capacity in the iPhone 17 series, leading to substantial benefits for the memory industry.\n\nFor example, AI smartphones that perform AI functions on the device essentially require larger memory capacities. Consequently, predictions emerged that all iPhone 17 series models would be equipped with 12GB of DRAM.\n\nNotably, the average selling price (ASP) of 12GB DRAM is 50% higher compared to the previous 8GB models.\n\nThis led to increased revenue expectations for major memory manufacturers like SK hynix and Micron, which supply low-power DRAM (LPDDR) and NAND flash to Apple.\n\nHowever, if there's no innovation in AI features and Apple's smartphone shipments underperform compared to previous years, it could negatively impact memory performance in the second half of the year.\n\nAn industry insider stated, \"It is highly probable that Apple's market share in China will continue to decline, and there's a high possibility of extended deferred consumer demand into 2026. We view the success of the iPhone 17 as highly likely to fail.\"\n\nNotably, Apple, along with Samsung Electronics, is a major purchaser of semiconductors globally. There are also concerns that Apple might push for lower supply prices, citing slowing smartphone demand.\n\n$AAPL\n\n---\n\nhttps://t.co/CaaZL06hgK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027948118414457324, "ret_signed_m1d": 0.027948118414457324, "alpha_spy_m1d": 0.022310417618582767, "alpha_c_m1d": 0.008449403797535116, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.019498714616922208, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016558644240776355, "ret_signed_p1d": -0.016558644240776355, "alpha_spy_p1d": -0.019410558831564506, "alpha_c_p1d": -0.017283687203689446, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0007250429629130917, "ret_p1w": 0.05431920075755614, "ret_signed_p1w": -0.05431920075755614, "alpha_spy_p1w": -0.06352179479191589, "alpha_c_p1w": -0.056150789876047824, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": -0.0018315891184916833, "ret_p1m": 0.07197360265270647, "ret_signed_p1m": -0.07197360265270647, "alpha_spy_p1m": -0.03412636208992326, "alpha_c_p1m": 0.017124797603716102, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.08909840025642257, "ret_p3m": 0.15203821711103904, "ret_signed_p3m": -0.15203821711103904, "alpha_spy_p3m": -0.07564130102775568, "alpha_c_p3m": -0.03294789547090704, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.119090321640132, "ret_p6m": 1.0547328067255424, "ret_signed_p6m": -1.0547328067255424, "alpha_spy_p6m": -0.9142315863190187, "alpha_c_p6m": -0.6643791137839823, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.3903536929415601, "price_path": [-0.1698, -0.1181, -0.0978, -0.11, -0.107, -0.0988, -0.1712, -0.1518, -0.1759, -0.1939, -0.2024, -0.2262, -0.2387, -0.2228, -0.2238, -0.3487, -0.433, -0.401, -0.4258, -0.3178, -0.3863, -0.3907, -0.3778, -0.3776, -0.3926, -0.3972, -0.3972, -0.4153, -0.3849, -0.361, -0.3217, -0.301, -0.3117, -0.3264, -0.3258, -0.3186, -0.2928, -0.2954, -0.2946, -0.2762, -0.254, -0.2478, -0.1914, -0.1508, -0.1649, -0.1637, -0.1414, -0.1357, -0.1405, -0.1603, -0.1692, -0.182, -0.182, -0.1556, -0.1574, -0.1519, -0.1724, -0.1398, -0.1042, -0.0954, -0.0688, -0.0489, -0.0279, 0.0, 0.0166, 0.0179, 0.0128, 0.0499, 0.0543, 0.0673, 0.0673, 0.0829, 0.0696, 0.1206, 0.1149, 0.1039, 0.093, 0.0798, 0.0591, 0.0666, 0.0714, 0.0714, 0.0516, 0.0911, 0.072, 0.0796, 0.0921, 0.0401, 0.0533, 0.021, -0.0068, 0.0031, -0.007, -0.0422, -0.0369, -0.0202, -0.0243, -0.0244, -0.0182, 0.0062, -0.0429, -0.0803, -0.0549, -0.0436, -0.0461, -0.019, 0.0426, 0.085, 0.1203, 0.0898, 0.0987, 0.06, 0.0835, 0.0703, 0.0279, 0.0154, 0.032, 0.0209, 0.0216, 0.0326, 0.0699, 0.0436, 0.0436, 0.039, 0.0411, 0.0892, 0.152, 0.1528, 0.186, 0.2277, 0.3204, 0.3788, 0.3836, 0.3928, 0.403, 0.4811, 0.427, 0.4436, 0.4593, 0.4181, 0.3753, 0.3792, 0.4373, 0.4673, 0.5973, 0.6114, 0.6482, 0.6757, 0.6294, 0.7246, 0.6877, 0.5935, 0.6915, 0.6414, 0.6843, 0.7772, 0.7759, 0.8144, 0.7751, 0.7416, 0.8139, 0.9219, 0.9314, 0.9472, 0.9887, 0.9657, 0.9636, 1.0595, 0.9132, 1.084, 1.0913, 1.0877, 1.2227, 1.1157, 1.149, 1.0792, 1.1659, 1.1231, 1.0051, 0.9824, 0.767, 0.8197, 0.965, 0.9702, 1.0205, 1.0205, 1.0751, 1.11, 1.1015, 1.0547]}
{"unit_id": "thread:1932315591727329435", "ticker_idx": 2, "ts": "2025-06-10T05:55:12+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "could negatively impact memory performance in the second half of the year", "tweet_type": "analysis", "timeline": "months", "summary": "iPhone 17 likely to disappoint on AI; bearish AAPL shipments and negative read-through for memory suppliers MU and SK Hynix in H2.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Korean Memory Industry Worried About Memory Demand Amid Potential iPhone 17 Weakness\n\nAccording to South Korean media reports, Apple unveiled new AI features at its annual event, \"WWDC 2025,\" but some are calling it \"an empty shell.\"\n\nApple announced features such as real-time call translation and automatic spam call and message classification for the iPhone.\n\nHowever, these features appear to be merely echoing existing AI functions in the market. Furthermore, real-time call translation is a technology that Samsung Electronics already introduced in its Galaxy S24 series last year.\n\nIndustry experts commented that this event could further exacerbate Apple's disadvantage in the AI market.\n\nGuo Ming-chi, a TF International analyst from Taiwan and a recognized Apple expert, stated, \"The most important part of WWDC 2025 for investors is Apple's AI. In a situation where the AI industry is accelerating, if Apple does not significantly increase its investment, it risks falling behind competitors.\"\n\nThe negative impact on the memory industry could also grow.\n\nUntil now, there were high expectations that Apple would aggressively enter the AI smartphone competition by significantly expanding memory capacity in the iPhone 17 series, leading to substantial benefits for the memory industry.\n\nFor example, AI smartphones that perform AI functions on the device essentially require larger memory capacities. Consequently, predictions emerged that all iPhone 17 series models would be equipped with 12GB of DRAM.\n\nNotably, the average selling price (ASP) of 12GB DRAM is 50% higher compared to the previous 8GB models.\n\nThis led to increased revenue expectations for major memory manufacturers like SK hynix and Micron, which supply low-power DRAM (LPDDR) and NAND flash to Apple.\n\nHowever, if there's no innovation in AI features and Apple's smartphone shipments underperform compared to previous years, it could negatively impact memory performance in the second half of the year.\n\nAn industry insider stated, \"It is highly probable that Apple's market share in China will continue to decline, and there's a high possibility of extended deferred consumer demand into 2026. We view the success of the iPhone 17 as highly likely to fail.\"\n\nNotably, Apple, along with Samsung Electronics, is a major purchaser of semiconductors globally. There are also concerns that Apple might push for lower supply prices, citing slowing smartphone demand.\n\n$AAPL\n\n---\n\nhttps://t.co/CaaZL06hgK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006507625414107987, "ret_signed_m1d": 0.006507625414107987, "alpha_spy_m1d": 0.0008699246182334308, "alpha_c_m1d": -0.012991089202814221, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.019498714616922208, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04121468867328337, "ret_signed_p1d": -0.04121468867328337, "alpha_spy_p1d": -0.044066603264071524, "alpha_c_p1d": -0.041939731636196464, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0007250429629130917, "ret_p1w": 0.08026030501656423, "ret_signed_p1w": -0.08026030501656423, "alpha_spy_p1w": -0.08946289905092397, "alpha_c_p1w": -0.08209189413505591, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": -0.0018315891184916833, "ret_p1m": 0.21908889840668322, "ret_signed_p1m": -0.21908889840668322, "alpha_spy_p1m": -0.1812416578439, "alpha_c_p1m": -0.12999049815026065, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.08909840025642257, "ret_p3m": 0.18826472307620667, "ret_signed_p3m": -0.18826472307620667, "alpha_spy_p3m": -0.11186780699292331, "alpha_c_p3m": -0.06917440143607467, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.119090321640132, "ret_p6m": 1.3999705390081774, "ret_signed_p6m": -1.3999705390081774, "alpha_spy_p6m": -1.2594693186016537, "alpha_c_p6m": -1.0096168460666173, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.3903536929415601, "price_path": [-0.1352, -0.1144, -0.1079, -0.1209, -0.1101, -0.0906, -0.0668, -0.0841, -0.0992, -0.0733, -0.1036, -0.1369, -0.1742, -0.1469, -0.143, -0.1573, -0.211, -0.2863, -0.266, -0.2855, -0.2066, -0.217, -0.2196, -0.2179, -0.2465, -0.2421, -0.2421, -0.2352, -0.2473, -0.2162, -0.2279, -0.2014, -0.2118, -0.217, -0.2313, -0.2313, -0.1945, -0.1945, -0.1945, -0.1737, -0.1759, -0.1768, -0.1555, -0.1404, -0.1079, -0.1317, -0.1144, -0.1365, -0.1252, -0.1317, -0.1473, -0.1339, -0.1209, -0.1231, -0.0992, -0.0803, -0.1128, -0.0998, -0.0998, -0.0564, -0.026, -0.026, -0.0065, 0.0, 0.0412, 0.0217, 0.0217, 0.0759, 0.0803, 0.0694, 0.0672, 0.115, 0.1258, 0.2082, 0.2408, 0.2711, 0.2321, 0.2668, 0.2386, 0.2104, 0.2082, 0.1735, 0.1757, 0.2234, 0.2191, 0.2885, 0.2777, 0.3015, 0.295, 0.2842, 0.1692, 0.167, 0.1822, 0.1649, 0.167, 0.1692, 0.154, 0.1367, 0.1388, 0.1432, 0.1866, 0.1193, 0.1193, 0.1432, 0.1215, 0.1367, 0.1128, 0.1584, 0.167, 0.2061, 0.1996, 0.1996, 0.1605, 0.141, 0.1085, 0.0629, 0.0889, 0.1258, 0.1345, 0.128, 0.1665, 0.1687, 0.1122, 0.1318, 0.1405, 0.1535, 0.1883, 0.2035, 0.2513, 0.3208, 0.3338, 0.4272, 0.4381, 0.5119, 0.4489, 0.5445, 0.5445, 0.525, 0.5684, 0.5532, 0.5489, 0.462, 0.5163, 0.5098, 0.5641, 0.7183, 0.7183, 0.7183, 0.7183, 0.7183, 0.7183, 0.8595, 0.803, 0.7878, 0.8356, 0.966, 1.0224, 1.1093, 1.0811, 1.092, 1.0789, 1.2158, 1.3244, 1.2636, 1.4243, 1.4678, 1.4287, 1.6937, 1.546, 1.5156, 1.5764, 1.5199, 1.6329, 1.6893, 1.6807, 1.6589, 1.433, 1.6329, 1.4765, 1.4417, 1.4808, 1.2636, 1.2592, 1.2549, 1.2766, 1.3652, 1.3043, 1.3391, 1.4261, 1.4]}
{"unit_id": "orig:1943316584975733072", "ticker_idx": 0, "ts": "2025-07-10T14:29:13+00:00", "symbol": "Wiwynn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley has raised Wiwynn's 2025-2027 earnings forecasts by 34%, 9%, and 12%, respectively, and increased its price target to NT$3,150", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley upgrades Wiwynn and Wistron EPS+PT on strong AI server demand; maintains positive rating on Quanta despite Q2 GB200 miss.", "resolved_tickers": ["6669.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wiwynn Corp listed on TWSE as 6669.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwanese ODM Manufacturers: AI Server Shipments Remain Strong, Maintaining Positive Outlook for Wiwynn, Wistron, and Quanta\nMorgan Stanley (S. Shih, July 08, 2025)\n\nMorgan Stanley states that despite unfavorable exchange rates, Taiwanese ODM manufacturers' June revenues generally exceeded market expectations, driven by strong AI server shipment momentum (especially for Trainium 2 and GB200 racks). Morgan Stanley has upgraded Wiwynn's and Wistron's earnings forecasts and maintained positive ratings on the three major ODM firms.\n\nWiwynn's AI Server Contribution Reaches 40%, Driving Earnings and Price Target Upgrades\nMorgan Stanley estimates Wiwynn shipped approximately 8,000 ASIC servers in Q2, contributing nearly 40% to its total Q2 revenue, which reached a record high of NT220.7billion.Q3revenueisexpectedtofurtherincreaseby5232.6 billion. Based on sustained Trainium 2 demand, Morgan Stanley has raised Wiwynn's 2025-2027 earnings forecasts by 34%, 9%, and 12%, respectively, and increased its price target to NT$3,150.\n\nWistron Benefits from Multiple Client and Regional Expansion, Earnings Upgraded\nWistron benefits from the growth of its subsidiary Wiwynn, expanded orders for enterprise clients' L10 compute trays and L6 board-level assembly (GB300 and Rubin), and potential business expansion in the Middle East market. Morgan Stanley has raised its 2025-2027 net profit forecasts by 11%, 2%, and 8%, respectively, and increased its price target to NT$150.\n\nQuanta's GB200 Shipments Below Expectation, Still Expected to Deliver 10K Units Annually\nQuanta's Q2 GB200 shipments were only about 1,300 units, lower than the expected 2,200 units, leading to lower-than-expected Q2 revenue. Morgan Stanley has thus lowered its 2025-2027 earnings forecasts by 3%, 3%, and 1%, respectively, but still expects to deliver approximately 10,000 GB200 racks for the full year. Meta's ASIC project could become a growth driver after 2026.\n\nMiddle East AI Factory Investment Brings Long-Term Growth Potential\nThe Middle East region plans to invest over $500 billion to build 8GW of AI computing power, which is expected to drive demand for 150,000 GB300-level racks over the next five years. Brand manufacturers (e.g., Dell, SMCI, Lenovo) will be major beneficiaries. ODM manufacturers like Foxconn, FII, Wistron, and its subsidiary Wiwynn may also benefit, but this has not yet been incorporated into the earnings models.\n\nGB300 Rack Unit Price Slightly Up, But Assembly Gross Margin Slightly Down\nThe average unit price for a GB300 NVL72 rack is expected to be $3.2 million to $3.3 million, an increase of 6-10% compared to GB200. ODM manufacturers' assembly gross margin is projected to decrease from 1.7% for GB200 to 1.6% for GB300, still resulting in a pure assembly profit of approximately $50,000 per unit. The overall added value from assembly remains largely unchanged, with major upgrades concentrated at the chip level.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013487475915221592, "ret_signed_m1d": -0.013487475915221592, "alpha_spy_m1d": -0.010675097476527395, "alpha_c_m1d": -0.016795134419587243, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003853564547206201, "ret_signed_p1d": 0.003853564547206201, "alpha_spy_p1d": 0.007368988417801758, "alpha_c_p1d": 0.008250858683306772, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": -0.05394990366088637, "ret_signed_p1w": -0.05394990366088637, "alpha_spy_p1w": -0.0574971947278915, "alpha_c_p1w": -0.06982643793397214, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.11368015414258181, "ret_signed_p1m": 0.11368015414258181, "alpha_spy_p1m": 0.09552804827660788, "alpha_c_p1m": 0.07889140377049175, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.35260115606936426, "ret_signed_p3m": 0.35260115606936426, "alpha_spy_p3m": 0.28044772535420637, "alpha_c_p3m": 0.23949075339066228, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.7263969171483622, "ret_signed_p6m": 0.7263969171483622, "alpha_spy_p6m": 0.6284960703563021, "alpha_c_p6m": 0.6002034942741696, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.2853, -0.304, -0.3115, -0.2966, -0.2947, -0.3115, -0.3377, -0.2891, -0.3022, -0.2872, -0.2891, -0.2778, -0.2853, -0.2853, -0.2273, -0.2367, -0.2049, -0.218, -0.1955, -0.1768, -0.1562, -0.1675, -0.1226, -0.1095, -0.1095, -0.1544, -0.1394, -0.1226, -0.1244, -0.1469, -0.1132, -0.13, -0.1095, -0.0945, -0.0945, -0.0982, -0.072, -0.0515, -0.0702, -0.0777, -0.0777, -0.0702, -0.0552, -0.0515, -0.0402, -0.0552, -0.0459, -0.0421, -0.0533, -0.0515, -0.0459, -0.0308, 0.0443, 0.0096, -0.025, -0.025, -0.0559, -0.0405, -0.0482, -0.0539, -0.0539, -0.0674, -0.0135, 0.0, 0.0039, -0.0501, -0.0289, -0.0347, -0.0539, -0.0019, -0.027, -0.0482, -0.0578, -0.0328, -0.0308, 0.0019, 0.0135, 0.0212, 0.0655, 0.0655, 0.0193, 0.0617, 0.0694, 0.0829, 0.1137, 0.2235, 0.2601, 0.3044, 0.3083, 0.289, 0.3198, 0.2736, 0.1503, 0.1676, 0.1445, 0.1753, 0.1773, 0.1503, 0.1464, 0.1426, 0.0925, 0.1118, 0.1272, 0.131, 0.1657, 0.2042, 0.1599, 0.2755, 0.2755, 0.2486, 0.2408, 0.2331, 0.2177, 0.2331, 0.2447, 0.2042, 0.21, 0.1869, 0.1927, 0.1753, 0.1753, 0.2775, 0.2717, 0.2775, 0.3353, 0.3353, 0.3526, 0.3237, 0.2987, 0.2987, 0.3218, 0.3526, 0.4624, 0.4875, 0.4586, 0.4952, 0.5665, 0.5665, 0.5819, 0.5819, 0.5626, 0.5742, 0.5838, 0.657, 0.6821, 0.6628, 0.657, 0.605, 0.6647, 0.5703, 0.6551, 0.8189, 0.7572, 0.7187, 0.6493, 0.6763, 0.5915, 0.5992, 0.6493, 0.578, 0.553, 0.6069, 0.7148, 0.7148, 0.7611, 0.7071, 0.7264, 0.7726, 0.7514, 0.8054, 0.7842, 0.7418, 0.7437, 0.7418, 0.6975, 0.6532, 0.6031, 0.6089, 0.6089, 0.6513, 0.6378, 0.6397, 0.6339, 0.6339, 0.657, 0.6763, 0.6936, 0.7283, 0.7283, 0.7264]}
{"unit_id": "orig:1943316584975733072", "ticker_idx": 1, "ts": "2025-07-10T14:29:13+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley has raised its 2025-2027 net profit forecasts by 11%, 2%, and 8%, respectively, and increased its price target to NT$150", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley upgrades Wiwynn and Wistron EPS+PT on strong AI server demand; maintains positive rating on Quanta despite Q2 GB200 miss.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwanese ODM Manufacturers: AI Server Shipments Remain Strong, Maintaining Positive Outlook for Wiwynn, Wistron, and Quanta\nMorgan Stanley (S. Shih, July 08, 2025)\n\nMorgan Stanley states that despite unfavorable exchange rates, Taiwanese ODM manufacturers' June revenues generally exceeded market expectations, driven by strong AI server shipment momentum (especially for Trainium 2 and GB200 racks). Morgan Stanley has upgraded Wiwynn's and Wistron's earnings forecasts and maintained positive ratings on the three major ODM firms.\n\nWiwynn's AI Server Contribution Reaches 40%, Driving Earnings and Price Target Upgrades\nMorgan Stanley estimates Wiwynn shipped approximately 8,000 ASIC servers in Q2, contributing nearly 40% to its total Q2 revenue, which reached a record high of NT220.7billion.Q3revenueisexpectedtofurtherincreaseby5232.6 billion. Based on sustained Trainium 2 demand, Morgan Stanley has raised Wiwynn's 2025-2027 earnings forecasts by 34%, 9%, and 12%, respectively, and increased its price target to NT$3,150.\n\nWistron Benefits from Multiple Client and Regional Expansion, Earnings Upgraded\nWistron benefits from the growth of its subsidiary Wiwynn, expanded orders for enterprise clients' L10 compute trays and L6 board-level assembly (GB300 and Rubin), and potential business expansion in the Middle East market. Morgan Stanley has raised its 2025-2027 net profit forecasts by 11%, 2%, and 8%, respectively, and increased its price target to NT$150.\n\nQuanta's GB200 Shipments Below Expectation, Still Expected to Deliver 10K Units Annually\nQuanta's Q2 GB200 shipments were only about 1,300 units, lower than the expected 2,200 units, leading to lower-than-expected Q2 revenue. Morgan Stanley has thus lowered its 2025-2027 earnings forecasts by 3%, 3%, and 1%, respectively, but still expects to deliver approximately 10,000 GB200 racks for the full year. Meta's ASIC project could become a growth driver after 2026.\n\nMiddle East AI Factory Investment Brings Long-Term Growth Potential\nThe Middle East region plans to invest over $500 billion to build 8GW of AI computing power, which is expected to drive demand for 150,000 GB300-level racks over the next five years. Brand manufacturers (e.g., Dell, SMCI, Lenovo) will be major beneficiaries. ODM manufacturers like Foxconn, FII, Wistron, and its subsidiary Wiwynn may also benefit, but this has not yet been incorporated into the earnings models.\n\nGB300 Rack Unit Price Slightly Up, But Assembly Gross Margin Slightly Down\nThe average unit price for a GB300 NVL72 rack is expected to be $3.2 million to $3.3 million, an increase of 6-10% compared to GB200. ODM manufacturers' assembly gross margin is projected to decrease from 1.7% for GB200 to 1.6% for GB300, still resulting in a pure assembly profit of approximately $50,000 per unit. The overall added value from assembly remains largely unchanged, with major upgrades concentrated at the chip level.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004166666666666652, "ret_signed_m1d": 0.004166666666666652, "alpha_spy_m1d": 0.006979045105360848, "alpha_c_m1d": 0.0008590081623010004, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004166666666666652, "ret_signed_p1d": -0.004166666666666652, "alpha_spy_p1d": -0.0006512427960710943, "alpha_c_p1d": 0.00023062746943391943, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": -0.01666666666666672, "ret_signed_p1w": -0.01666666666666672, "alpha_spy_p1w": -0.02021395773367185, "alpha_c_p1w": -0.03254320093975249, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.03750000000000009, "ret_signed_p1m": 0.03750000000000009, "alpha_spy_p1m": 0.019347894134026156, "alpha_c_p1m": 0.0027112496279100284, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.30000000000000004, "ret_signed_p3m": 0.30000000000000004, "alpha_spy_p3m": 0.22784656928484215, "alpha_c_p3m": 0.18688959732129806, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.29166666666666674, "ret_signed_p6m": 0.29166666666666674, "alpha_spy_p6m": 0.19376581987460662, "alpha_c_p6m": 0.1654732437924742, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.216, -0.1741, -0.1966, -0.1902, -0.1539, -0.17, -0.2039, -0.1741, -0.1861, -0.1741, -0.1741, -0.162, -0.1821, -0.1821, -0.1499, -0.166, -0.1217, -0.1418, -0.1378, -0.1378, -0.1257, -0.1257, -0.0774, -0.1015, -0.1056, -0.1217, -0.1176, -0.0894, -0.0894, -0.0894, -0.0653, -0.0733, -0.0693, -0.0612, -0.0612, -0.0733, -0.0875, -0.05, -0.0417, -0.0833, -0.0542, -0.0458, -0.0417, -0.0292, -0.0125, -0.0083, -0.0167, -0.0333, -0.0375, -0.0125, -0.0125, 0.0, 0.0458, 0.0417, 0.0083, 0.0208, 0.0, 0.0, 0.0083, -0.0083, -0.0167, -0.0125, 0.0042, 0.0, -0.0042, -0.025, 0.0, -0.0167, -0.0167, -0.0083, -0.0125, -0.0458, -0.0333, -0.025, -0.0292, 0.0, 0.0042, 0.0042, 0.025, 0.025, -0.0042, 0.0333, 0.0208, 0.0292, 0.0375, 0.0625, 0.05, -0.0208, -0.0292, -0.0292, -0.0333, -0.0375, -0.0542, -0.0708, -0.0667, -0.0542, -0.0583, -0.0417, -0.0417, -0.0583, -0.0875, -0.0833, -0.0833, -0.0583, -0.0375, -0.0458, -0.0333, 0.0, -0.0083, 0.0, -0.0042, -0.0167, -0.0083, 0.0, 0.0083, 0.0125, 0.0208, 0.0125, 0.0458, 0.0667, 0.0667, 0.1708, 0.2167, 0.2458, 0.2792, 0.2792, 0.3, 0.2542, 0.2583, 0.2583, 0.2125, 0.15, 0.15, 0.1625, 0.1375, 0.1833, 0.1875, 0.1958, 0.1917, 0.1917, 0.2208, 0.2333, 0.2708, 0.2917, 0.2542, 0.1958, 0.1583, 0.1458, 0.1833, 0.1458, 0.1542, 0.2042, 0.1583, 0.2, 0.1708, 0.15, 0.1417, 0.1417, 0.1917, 0.1583, 0.1583, 0.1625, 0.1875, 0.2125, 0.2042, 0.175, 0.1917, 0.2083, 0.2167, 0.2583, 0.2417, 0.2458, 0.2292, 0.2, 0.1958, 0.1792, 0.1583, 0.15, 0.1625, 0.2, 0.2208, 0.2083, 0.2125, 0.2125, 0.2208, 0.2167, 0.225, 0.2542, 0.2542, 0.2917]}
{"unit_id": "orig:1943316584975733072", "ticker_idx": 2, "ts": "2025-07-10T14:29:13+00:00", "symbol": "Quanta", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "still expects to deliver approximately 10,000 GB200 racks for the full year. Meta's ASIC project could become a growth driver after 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley upgrades Wiwynn and Wistron EPS+PT on strong AI server demand; maintains positive rating on Quanta despite Q2 GB200 miss.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta in server/ODM context = Quanta Computer, TWSE 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwanese ODM Manufacturers: AI Server Shipments Remain Strong, Maintaining Positive Outlook for Wiwynn, Wistron, and Quanta\nMorgan Stanley (S. Shih, July 08, 2025)\n\nMorgan Stanley states that despite unfavorable exchange rates, Taiwanese ODM manufacturers' June revenues generally exceeded market expectations, driven by strong AI server shipment momentum (especially for Trainium 2 and GB200 racks). Morgan Stanley has upgraded Wiwynn's and Wistron's earnings forecasts and maintained positive ratings on the three major ODM firms.\n\nWiwynn's AI Server Contribution Reaches 40%, Driving Earnings and Price Target Upgrades\nMorgan Stanley estimates Wiwynn shipped approximately 8,000 ASIC servers in Q2, contributing nearly 40% to its total Q2 revenue, which reached a record high of NT220.7billion.Q3revenueisexpectedtofurtherincreaseby5232.6 billion. Based on sustained Trainium 2 demand, Morgan Stanley has raised Wiwynn's 2025-2027 earnings forecasts by 34%, 9%, and 12%, respectively, and increased its price target to NT$3,150.\n\nWistron Benefits from Multiple Client and Regional Expansion, Earnings Upgraded\nWistron benefits from the growth of its subsidiary Wiwynn, expanded orders for enterprise clients' L10 compute trays and L6 board-level assembly (GB300 and Rubin), and potential business expansion in the Middle East market. Morgan Stanley has raised its 2025-2027 net profit forecasts by 11%, 2%, and 8%, respectively, and increased its price target to NT$150.\n\nQuanta's GB200 Shipments Below Expectation, Still Expected to Deliver 10K Units Annually\nQuanta's Q2 GB200 shipments were only about 1,300 units, lower than the expected 2,200 units, leading to lower-than-expected Q2 revenue. Morgan Stanley has thus lowered its 2025-2027 earnings forecasts by 3%, 3%, and 1%, respectively, but still expects to deliver approximately 10,000 GB200 racks for the full year. Meta's ASIC project could become a growth driver after 2026.\n\nMiddle East AI Factory Investment Brings Long-Term Growth Potential\nThe Middle East region plans to invest over $500 billion to build 8GW of AI computing power, which is expected to drive demand for 150,000 GB300-level racks over the next five years. Brand manufacturers (e.g., Dell, SMCI, Lenovo) will be major beneficiaries. ODM manufacturers like Foxconn, FII, Wistron, and its subsidiary Wiwynn may also benefit, but this has not yet been incorporated into the earnings models.\n\nGB300 Rack Unit Price Slightly Up, But Assembly Gross Margin Slightly Down\nThe average unit price for a GB300 NVL72 rack is expected to be $3.2 million to $3.3 million, an increase of 6-10% compared to GB200. ODM manufacturers' assembly gross margin is projected to decrease from 1.7% for GB200 to 1.6% for GB300, still resulting in a pure assembly profit of approximately $50,000 per unit. The overall added value from assembly remains largely unchanged, with major upgrades concentrated at the chip level.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01098901098901095, "ret_signed_m1d": 0.01098901098901095, "alpha_spy_m1d": 0.013801389427705146, "alpha_c_m1d": 0.0076813524846452985, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0035154238705955576, "alpha_c_p1d": 0.004397294136100571, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": -0.0073260073260073, "ret_signed_p1w": -0.0073260073260073, "alpha_spy_p1w": -0.010873298393012432, "alpha_c_p1w": -0.023202541599093074, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.0641025641025641, "ret_signed_p1m": 0.0641025641025641, "alpha_spy_p1m": 0.045950458236590164, "alpha_c_p1m": 0.029313813730474036, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.13003663003663002, "ret_signed_p3m": 0.13003663003663002, "alpha_spy_p3m": 0.05788319932147212, "alpha_c_p3m": 0.016926227357928036, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.0146520146520146, "ret_signed_p6m": 0.0146520146520146, "alpha_spy_p6m": -0.08324883214004553, "alpha_c_p6m": -0.11154140822217795, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.2029, -0.2029, -0.2064, -0.1959, -0.1941, -0.1889, -0.2186, -0.1784, -0.1959, -0.1784, -0.1434, -0.1522, -0.1662, -0.1662, -0.1225, -0.1382, -0.112, -0.112, -0.1207, -0.1032, -0.0962, -0.084, -0.0333, -0.0298, -0.0385, -0.07, -0.0683, -0.049, -0.0648, -0.077, -0.0718, -0.07, -0.0595, -0.0508, -0.0508, -0.0525, -0.0385, -0.0211, -0.0158, -0.035, -0.035, -0.0088, -0.0106, 0.0017, 0.0052, -0.0001, -0.0106, -0.0106, -0.0246, -0.0228, -0.0141, 0.0122, 0.0244, 0.0139, -0.0018, 0.0055, 0.0293, 0.033, 0.0256, 0.0183, 0.0147, -0.0037, 0.011, 0.0, 0.0, -0.0128, -0.0092, -0.0018, -0.0073, 0.0018, 0.0018, -0.0385, -0.0311, -0.0147, -0.0165, -0.0037, -0.0201, -0.011, 0.0311, 0.0311, 0.0293, 0.0549, 0.0568, 0.0586, 0.0641, 0.0623, 0.0366, -0.0055, -0.0092, -0.0092, -0.0147, -0.0183, -0.0495, -0.0513, -0.0531, -0.0183, -0.0293, -0.0147, -0.0256, -0.0385, -0.0623, -0.0714, -0.0714, -0.0586, -0.0531, -0.0476, -0.0385, 0.0, -0.011, 0.0037, 0.011, 0.0018, 0.0, 0.0238, 0.0293, 0.0293, 0.0495, 0.0385, 0.044, 0.044, 0.044, 0.0623, 0.0495, 0.0769, 0.0989, 0.0989, 0.13, 0.1209, 0.0916, 0.0916, 0.0788, 0.0403, 0.0476, 0.0678, 0.0568, 0.0861, 0.0659, 0.0604, 0.0751, 0.0751, 0.1337, 0.1136, 0.1282, 0.1209, 0.1007, 0.0897, 0.0531, 0.0604, 0.0733, 0.0513, 0.0678, 0.0678, 0.0549, 0.044, 0.0238, 0.011, -0.011, -0.022, 0.0037, -0.011, -0.0037, 0.0073, 0.0238, 0.0311, 0.033, 0.0513, 0.0568, 0.0714, 0.0788, 0.0897, 0.0897, 0.0842, 0.0714, 0.0549, 0.0495, 0.044, 0.0, -0.0293, -0.0476, -0.0293, -0.022, -0.0311, -0.033, -0.033, -0.0366, -0.0366, -0.0366, -0.0037, -0.0037, 0.0147]}
{"unit_id": "orig:1940902022779490515", "ticker_idx": 0, "ts": "2025-07-03T22:34:37+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix continues to strengthen its position among AI ASIC customers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS sees HBM supply-demand stable; SK Hynix leads AI ASIC wins while Samsung HBM forecasts are cut; DDR/NAND face potential downturn from Q4 2025 on tariff risks.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Memory Review: HBM Supply and Demand Remain Balanced, SK Hynix Takes Lead Among AI Customers\n\nUBS (T. Arcuri, 25/07/01)\n\nUBS believes that despite statistical differences between HBM shipments and end-demand, the current supply-demand relationship remains healthy. SK Hynix continues to strengthen its position among AI ASIC customers, and industry short-term shipments and pricing are also showing positive trends.\n\nHBM Supply-Demand Discrepancy Driven by Inventory Cycle, Overall Relationship Stable\n\nUBS points out that there are multi-month inventory and packaging lead time differences in HBM shipments. It can take three to four months from HBM manufacturers' shipments to final server deployment, thus the current supply-demand gap does not pose a risk. Considering an inventory factor of about 3.5 months, the HBM supply-demand gap for 2025 and 2026 is projected to be 12% and 6%, respectively. Additionally, UBS has lowered Samsung's HBM shipment forecast to 5.6 billion Gb for 2025 and 8.1 billion Gb for 2026.\n\nSK Hynix Continues to Lead in HBM Customer Acquisition\n\nIn terms of AI chip customers, SK Hynix has secured 100% of the HBM3E 8-Hi order for AWS Trainium 2.5 (manufactured by Marvell) and has become the sole HBM supplier for Trainium 3 (manufactured by Alchip). In the Google TPU 7p project, SK Hynix is the primary supplier, with Samsung as a secondary supplier; for TPU 7e, SK Hynix may exclusively secure the HBM3E 12-Hi order. Its HBM negotiations with Nvidia for 2026 are still ongoing, with rising demand for the Blackwell platform likely being a key driver.\n\nQ2 2025 Shipment Expectations Upgraded, Q3 NAND Prices Revised Upwards\n\nUBS has raised its industry shipment expectations: Q2/Q3 2025 DRAM bit shipments are projected to increase by +17%/+11% QoQ (previously +15%/+9%), and NAND by +15%/+7% QoQ (previously +12%/+13%). In terms of pricing, Q2 DDR contract prices are expected to rise by 7% QoQ, and Q3 by 3%; NAND contract prices have been revised down for Q2 to flat (previously +3%), and up for Q3 from -3% to +3%. Considering tariff risks, UBS expects DDR and NAND to enter a three-quarter downturn cycle starting from Q4 2025.\n\nTariff Policies May Accelerate Implementation, Affecting Future Price Trends\n\nRegarding the implementation timeline for the US Section 232 tariffs, UBS mentions that the policy might be introduced earlier, by late July to early August. If implemented, it could exert some downward pressure on DDR and NAND pricing. UBS makes cautious assumptions for future market trends based on this scenario, demonstrating high sensitivity to industry cycle changes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001795285002198277, "ret_signed_m1d": 0.001795285002198277, "alpha_spy_m1d": 0.009615040446345202, "alpha_c_m1d": 0.010151615694433125, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028725348775210713, "ret_signed_p1d": -0.028725348775210713, "alpha_spy_p1d": -0.028725348775210713, "alpha_c_p1d": -0.028725348775210713, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.06642734791570937, "ret_signed_p1w": 0.06642734791570937, "alpha_spy_p1w": 0.06565978267537087, "alpha_c_p1w": 0.05267663986185589, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.073608656940225, "ret_signed_p1m": -0.073608656940225, "alpha_spy_p1m": -0.06781963225696874, "alpha_c_p1m": -0.0747721966568653, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.24955813656335146, "ret_signed_p3m": 0.24955813656335146, "alpha_spy_p3m": 0.1812962111754568, "alpha_c_p3m": 0.09886366093767363, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 1.1554572810508361, "ret_signed_p6m": 1.1554572810508361, "alpha_spy_p6m": 1.0452304634552643, "alpha_c_p6m": 0.8614780702027602, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.4093, -0.3925, -0.4086, -0.3434, -0.352, -0.3541, -0.3527, -0.3764, -0.3728, -0.3728, -0.367, -0.3771, -0.3513, -0.3609, -0.3391, -0.3477, -0.352, -0.3638, -0.3638, -0.3333, -0.3333, -0.3333, -0.3161, -0.3179, -0.3186, -0.3011, -0.2885, -0.2617, -0.2814, -0.267, -0.2853, -0.276, -0.2814, -0.2943, -0.2832, -0.2724, -0.2742, -0.2545, -0.2388, -0.2657, -0.2549, -0.2549, -0.219, -0.1939, -0.1939, -0.1777, -0.1724, -0.1382, -0.1544, -0.1544, -0.1095, -0.1059, -0.1149, -0.1167, -0.0772, -0.0682, 0.0, 0.0269, 0.0521, 0.0197, 0.0485, 0.0251, 0.0018, 0.0, -0.0287, -0.0269, 0.0126, 0.009, 0.0664, 0.0575, 0.0772, 0.0718, 0.0628, -0.0323, -0.0341, -0.0215, -0.0359, -0.0341, -0.0323, -0.0449, -0.0592, -0.0575, -0.0539, -0.018, -0.0736, -0.0736, -0.0539, -0.0718, -0.0592, -0.079, -0.0413, -0.0341, -0.0018, -0.0072, -0.0072, -0.0395, -0.0557, -0.0826, -0.1203, -0.0987, -0.0682, -0.061, -0.0664, -0.0345, -0.0327, -0.0795, -0.0633, -0.0561, -0.0453, -0.0165, -0.0039, 0.0356, 0.0931, 0.1039, 0.1812, 0.1902, 0.2514, 0.1992, 0.2783, 0.2783, 0.2621, 0.2981, 0.2855, 0.2819, 0.21, 0.255, 0.2496, 0.2945, 0.4222, 0.4222, 0.4222, 0.4222, 0.4222, 0.4222, 0.539, 0.4923, 0.4797, 0.5192, 0.6271, 0.6739, 0.7458, 0.7224, 0.7314, 0.7206, 0.8339, 0.9238, 0.8734, 1.0065, 1.0424, 1.0101, 1.2294, 1.1072, 1.082, 1.1323, 1.0856, 1.1791, 1.2258, 1.2186, 1.2007, 1.0137, 1.1791, 1.0496, 1.0209, 1.0532, 0.8734, 0.8698, 0.8662, 0.8842, 0.9575, 0.9072, 0.936, 1.0079, 0.9863, 0.9503, 0.9575, 1.0763, 1.0367, 1.1123, 1.0331, 1.0547, 0.9935, 0.9072, 0.9827, 0.9863, 0.9683, 1.0871, 1.1015, 1.1159, 1.1159, 1.1555]}
{"unit_id": "orig:1940902022779490515", "ticker_idx": 1, "ts": "2025-07-03T22:34:37+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS has lowered Samsung's HBM shipment forecast to 5.6 billion Gb for 2025 and 8.1 billion Gb for 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS sees HBM supply-demand stable; SK Hynix leads AI ASIC wins while Samsung HBM forecasts are cut; DDR/NAND face potential downturn from Q4 2025 on tariff risks.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Memory Review: HBM Supply and Demand Remain Balanced, SK Hynix Takes Lead Among AI Customers\n\nUBS (T. Arcuri, 25/07/01)\n\nUBS believes that despite statistical differences between HBM shipments and end-demand, the current supply-demand relationship remains healthy. SK Hynix continues to strengthen its position among AI ASIC customers, and industry short-term shipments and pricing are also showing positive trends.\n\nHBM Supply-Demand Discrepancy Driven by Inventory Cycle, Overall Relationship Stable\n\nUBS points out that there are multi-month inventory and packaging lead time differences in HBM shipments. It can take three to four months from HBM manufacturers' shipments to final server deployment, thus the current supply-demand gap does not pose a risk. Considering an inventory factor of about 3.5 months, the HBM supply-demand gap for 2025 and 2026 is projected to be 12% and 6%, respectively. Additionally, UBS has lowered Samsung's HBM shipment forecast to 5.6 billion Gb for 2025 and 8.1 billion Gb for 2026.\n\nSK Hynix Continues to Lead in HBM Customer Acquisition\n\nIn terms of AI chip customers, SK Hynix has secured 100% of the HBM3E 8-Hi order for AWS Trainium 2.5 (manufactured by Marvell) and has become the sole HBM supplier for Trainium 3 (manufactured by Alchip). In the Google TPU 7p project, SK Hynix is the primary supplier, with Samsung as a secondary supplier; for TPU 7e, SK Hynix may exclusively secure the HBM3E 12-Hi order. Its HBM negotiations with Nvidia for 2026 are still ongoing, with rising demand for the Blackwell platform likely being a key driver.\n\nQ2 2025 Shipment Expectations Upgraded, Q3 NAND Prices Revised Upwards\n\nUBS has raised its industry shipment expectations: Q2/Q3 2025 DRAM bit shipments are projected to increase by +17%/+11% QoQ (previously +15%/+9%), and NAND by +15%/+7% QoQ (previously +12%/+13%). In terms of pricing, Q2 DDR contract prices are expected to rise by 7% QoQ, and Q3 by 3%; NAND contract prices have been revised down for Q2 to flat (previously +3%), and up for Q3 from -3% to +3%. Considering tariff risks, UBS expects DDR and NAND to enter a three-quarter downturn cycle starting from Q4 2025.\n\nTariff Policies May Accelerate Implementation, Affecting Future Price Trends\n\nRegarding the implementation timeline for the US Section 232 tariffs, UBS mentions that the policy might be introduced earlier, by late July to early August. If implemented, it could exert some downward pressure on DDR and NAND pricing. UBS makes cautious assumptions for future market trends based on this scenario, demonstrating high sensitivity to industry cycle changes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.047021928431757676, "ret_signed_m1d": 0.047021928431757676, "alpha_spy_m1d": 0.03920217298761075, "alpha_c_m1d": 0.034023745904501745, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.01299818252725593, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00783698807195965, "ret_signed_p1d": 0.00783698807195965, "alpha_spy_p1d": 0.00783698807195965, "alpha_c_p1d": 0.00783698807195965, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.043887145588279775, "ret_signed_p1w": 0.043887145588279775, "alpha_spy_p1w": 0.04465471082861827, "alpha_c_p1w": 0.04396501257655816, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 7.786698827838556e-05, "ret_p1m": 0.07993724117807055, "ret_signed_p1m": -0.07993724117807055, "alpha_spy_p1m": -0.08572626586132681, "alpha_c_p1m": -0.0795092112325495, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0004280299455210468, "ret_p3m": 0.3209142981420554, "ret_signed_p3m": -0.3209142981420554, "alpha_spy_p3m": -0.25265237275416075, "alpha_c_p3m": -0.22260233639537597, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.09831196174667944, "ret_p6m": 0.8420378088709506, "ret_signed_p6m": -0.8420378088709506, "alpha_spy_p6m": -0.7318109912753787, "alpha_c_p6m": -0.6983512147938857, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.14368659407706486, "price_path": [-0.1712, -0.1666, -0.1743, -0.1214, -0.1401, -0.1245, -0.1183, -0.1479, -0.1416, -0.1385, -0.137, -0.1432, -0.1323, -0.1323, -0.1323, -0.1307, -0.1307, -0.1354, -0.1354, -0.1541, -0.1541, -0.1541, -0.1494, -0.1494, -0.1463, -0.1027, -0.1136, -0.1058, -0.1074, -0.1151, -0.1307, -0.1292, -0.1323, -0.1479, -0.1556, -0.1479, -0.1603, -0.1292, -0.1261, -0.1245, -0.1151, -0.1151, -0.0996, -0.0793, -0.0793, -0.0684, -0.0778, -0.0669, -0.0731, -0.0918, -0.1089, -0.0949, -0.0684, -0.0778, -0.0731, -0.0965, -0.0575, -0.045, -0.0622, -0.047, -0.0627, -0.0564, -0.047, 0.0, -0.0078, -0.0329, -0.0376, -0.0533, -0.0439, -0.0188, -0.0204, -0.0016, 0.0141, 0.0455, 0.0517, 0.0627, 0.0345, 0.0408, 0.0345, 0.0329, 0.1034, 0.1066, 0.1379, 0.1191, 0.0799, 0.0925, 0.0956, 0.0784, 0.105, 0.1254, 0.1129, 0.1144, 0.127, 0.1223, 0.1223, 0.0972, 0.0972, 0.105, 0.1066, 0.1191, 0.1207, 0.1019, 0.1066, 0.0909, 0.0925, 0.0596, 0.0831, 0.094, 0.0987, 0.0893, 0.0987, 0.1207, 0.1379, 0.1505, 0.1818, 0.1991, 0.2445, 0.2257, 0.2586, 0.2586, 0.3088, 0.3276, 0.3386, 0.3495, 0.3056, 0.3256, 0.3209, 0.354, 0.413, 0.413, 0.413, 0.413, 0.413, 0.413, 0.4862, 0.4689, 0.4421, 0.4957, 0.5382, 0.5413, 0.5445, 0.535, 0.5523, 0.5193, 0.5555, 0.6059, 0.5665, 0.5823, 0.6389, 0.6925, 0.7491, 0.6515, 0.5838, 0.5618, 0.5413, 0.5838, 0.6295, 0.6232, 0.6185, 0.5303, 0.5838, 0.5398, 0.5193, 0.5838, 0.4925, 0.5224, 0.5634, 0.6185, 0.6295, 0.5823, 0.587, 0.6279, 0.6452, 0.6547, 0.7066, 0.724, 0.7066, 0.7003, 0.6893, 0.7145, 0.65, 0.6185, 0.6988, 0.694, 0.6736, 0.7397, 0.7554, 0.7491, 0.7491, 0.842]}
{"unit_id": "quote:1925562523606294638", "ticker_idx": 0, "ts": "2025-05-22T14:40:55+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bullish for NVIDIA", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Author sees JPM's AI ASIC growth forecast as net bullish for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Bullish for NVIDIA\n\n$NVDA https://t.co/tm4fdCaXT8", "reply_text": null, "parent_text": null, "quoted_text": "Mixed Feedback on AI ASICs, but Meta’s MTIA Represents Potential Upside in the Next Phase-JPM\n\nWe forecast that AI ASIC chip shipments will grow by approximately 40% YoY in 2025, reaching around 4 million units. This follows a high base in 2024 (70% YoY growth), largely driven by", "parent_author": null, "ret_m1d": -0.007754252050075072, "ret_signed_m1d": -0.007754252050075072, "alpha_spy_m1d": -0.007359972366447476, "alpha_c_m1d": -0.010495028149897223, "bench_spy_m1d": -0.00039427968362759636, "bench_c_m1d": 0.002740776099822151, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01159380523896325, "ret_signed_p1d": -0.01159380523896325, "alpha_spy_p1d": -0.004768129373284302, "alpha_c_p1d": 0.002400615323976396, "bench_spy_p1d": -0.0068256758656789485, "bench_c_p1d": -0.013994420562939647, "ret_p1w": 0.047880777139537534, "ret_signed_p1w": 0.047880777139537534, "alpha_spy_p1w": 0.03594439380651471, "alpha_c_p1w": 0.03430160449543229, "bench_spy_p1w": 0.011936383333022826, "bench_c_p1w": 0.013579172644105242, "ret_p1m": 0.08303838082447901, "ret_signed_p1m": 0.08303838082447901, "alpha_spy_p1m": 0.06083442091819147, "alpha_c_p1m": 0.00222781790389992, "bench_spy_p1m": 0.02220395990628754, "bench_c_p1m": 0.08081056292057909, "ret_p3m": 0.322383436882691, "ret_signed_p3m": 0.322383436882691, "alpha_spy_p3m": 0.22186466900034985, "alpha_c_p3m": 0.11450167162457059, "bench_spy_p3m": 0.10051876788234115, "bench_c_p3m": 0.2078817652581204, "ret_p6m": 0.4318597657661136, "ret_signed_p6m": 0.4318597657661136, "alpha_spy_p6m": 0.2728879760328009, "alpha_c_p6m": -0.0007220364212952024, "bench_spy_p6m": 0.15897178973331272, "bench_c_p6m": 0.4325818021874088, "price_path": [-0.0193, -0.0468, -0.0118, -0.0955, -0.0596, -0.1414, -0.1269, -0.117, -0.1677, -0.1517, -0.1947, -0.1813, -0.1287, -0.1299, -0.084, -0.1001, -0.131, -0.1153, -0.1077, -0.1139, -0.086, -0.0914, -0.1436, -0.1611, -0.1744, -0.1841, -0.1707, -0.1687, -0.2336, -0.29, -0.2649, -0.275, -0.1393, -0.1902, -0.1649, -0.1665, -0.1553, -0.2134, -0.2359, -0.2359, -0.2704, -0.2555, -0.2268, -0.1988, -0.1643, -0.1814, -0.1793, -0.18, -0.1598, -0.138, -0.1431, -0.1452, -0.1187, -0.1164, -0.1218, -0.074, -0.0218, 0.0189, 0.0151, 0.0193, 0.0206, 0.0117, -0.0078, 0.0, -0.0116, -0.0116, 0.0201, 0.0149, 0.0479, 0.0173, 0.0343, 0.0632, 0.0684, 0.0539, 0.0669, 0.0738, 0.0838, 0.0754, 0.0917, 0.0689, 0.0894, 0.0851, 0.0953, 0.0953, 0.083, 0.0854, 0.1135, 0.1618, 0.1671, 0.1877, 0.1895, 0.1542, 0.1839, 0.1997, 0.1997, 0.1914, 0.2046, 0.2263, 0.2355, 0.2417, 0.2353, 0.2852, 0.2902, 0.3025, 0.2981, 0.2903, 0.2576, 0.2858, 0.3081, 0.3063, 0.3307, 0.3214, 0.3497, 0.3392, 0.3079, 0.3552, 0.3421, 0.3508, 0.361, 0.3755, 0.3707, 0.379, 0.3672, 0.3704, 0.3586, 0.3703, 0.3224, 0.3206, 0.3174, 0.3401, 0.3538, 0.3685, 0.3673, 0.3565, 0.3114, 0.3114, 0.2858, 0.2846, 0.2924, 0.2575, 0.2672, 0.2856, 0.3351, 0.334, 0.3389, 0.3383, 0.3167, 0.2822, 0.327, 0.3302, 0.3825, 0.3435, 0.3325, 0.3379, 0.3417, 0.3692, 0.4048, 0.4098, 0.4222, 0.4127, 0.397, 0.3932, 0.4239, 0.4499, 0.3791, 0.4179, 0.3555, 0.354, 0.3689, 0.3795, 0.3752, 0.364, 0.3574, 0.3715, 0.4024, 0.4418, 0.5136, 0.5589, 0.5276, 0.5246, 0.5577, 0.496, 0.4698, 0.4161, 0.4167, 0.4987, 0.4544, 0.4592, 0.4069, 0.4319]}
{"unit_id": "thread:1930444962774299047", "ticker_idx": 0, "ts": "2025-06-05T02:01:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix's leadership—already having secured the market—is strongly reflected in recent earnings, suggesting that profit improvement will continue for quite a long time", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SK hynix on durable HBM leadership and continuing upcycle; bearish NAND on absent AI tailwind.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "250605_Semiconductor Market and Outlook for SK hynix (with Meritz Securities Analyst Sunwoo Kim) – SK hynix\n\n⸻\n\n[Key Takeaways]\n\n(1) The AI upcycle will continue, and SK hynix will maintain its leadership in the memory industry based on its technological competitiveness.\n\n(2) The main drivers of AI demand will expand from enterprises to nation-states, and the utility of AI in the multimodal era is only beginning to be discovered.\n\n⸻\n\n[Contents]\n\n(1) SK hynix’s 1Q25 earnings exceeded market expectations significantly.\n\n(2) The title of the Meritz Securities report was “Differentiated Leadership Within the Upcycle.”\n\n(3) This title carries two meanings.\n\n(4) First, that the upcycle is ongoing and will continue into next year. Second—and more importantly—is “leadership.”\n\n(5) The HBM market is unfolding as a completely different one from traditional DRAM.\n\n(6) In the past, competition in memory was all about cost. Neither SK hynix nor other companies could command a premium on product prices.\n\n(7) That era of commodity memory has come to an end due to AI, and we are now in the age of specialized purpose semiconductors [like HBM].\n\n(8) The key difference now is that who makes the product determines its fundamental characteristics.\n\n(9) In this regard, SK hynix’s leadership—already having secured the market—is strongly reflected in recent earnings, suggesting that profit improvement will continue for quite a long time.\n\n(10) The risk to watch out for is that nothing lasts forever. Once second and third vendors enter the market and reach similar quality levels, profit margins will inevitably decline.\n\n(11) It’s true that SK hynix was well prepared when partners and customers within the value chain first approached them, which helped seize the opportunity. But going forward, the company must continue to ask what products can maintain overwhelming competitiveness—and pursue bold investments and strategic direction accordingly.\n\n(12) The most fundamental strength of SK hynix in terms of technology lies in front-end process competitiveness. The company’s leadership in the much more high-end HBM market has been made possible thanks to its 1-b process technology edge.\n\n(13) The 1-c process is also expected to produce products soon, and within the industry, it is seen as having the best energy efficiency [low power, high performance], likely leading to a successful market adoption.\n\n(14) The future task will be to accurately foresee trends and solidify partnerships within the value chain. If customers accept the products proposed by the company, then technological leadership can be sufficiently maintained.\n\n(15) Furthermore, future AI memory demand drivers will move to the nation-state level. Until now, AI investment was mainly driven by corporate competition, but going forward, some countries are expected to position AI as a national growth engine.\n\n(16) By product:\n1. DRAM will serve as a computational assistant and core differentiator, commanding a high price.\n2. NAND, on the other hand, does not see rising conventional demand and will not significantly benefit from AI, so strategic direction must be considered carefully.\n\n(17) AI has only just moved from text to image. Now, we are entering a full-fledged multimodal era—including video, audio, and more. The real utility of AI that stimulates primal human instincts is only now being discovered.\n\n(18) Always pay attention to the speed and nature of technological change. Especially:\n1. What technology is used for computation?\n2. Has memory become closer to the computation?\n3. Which company is leading at the forefront of that domain?\n\nThese will be the most crucial questions going forward.\n\n---\n\nOriginal: \nhttps://t.co/30O6gHrBbr", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031180282591622754, "ret_signed_m1d": -0.031180282591622754, "alpha_spy_m1d": -0.036036520725193366, "alpha_c_m1d": -0.032970988099138454, "bench_spy_m1d": 0.004856238133570612, "bench_c_m1d": 0.0017907055075156997, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.010268979858962801, "alpha_c_p1d": -0.005690423562760838, "bench_spy_p1d": 0.010268979858962801, "bench_c_p1d": 0.005690423562760838, "ret_p1w": 0.04899788645804648, "ret_signed_p1w": 0.04899788645804648, "alpha_spy_p1w": 0.030955552836329225, "alpha_c_p1w": 0.0018431286936022584, "bench_spy_p1w": 0.018042333621717255, "bench_c_p1w": 0.04715475776444422, "ret_p1m": 0.2048997886458046, "ret_signed_p1m": 0.2048997886458046, "alpha_spy_p1m": 0.14733508209110635, "alpha_c_p1m": 0.07628851731189346, "bench_spy_p1m": 0.05756470655469825, "bench_c_p1m": 0.12861127133391115, "ret_p3m": 0.16203243886079655, "ret_signed_p3m": 0.16203243886079655, "alpha_spy_p3m": 0.07921843774674842, "alpha_c_p3m": 0.022040320401363722, "bench_spy_p3m": 0.08281400111404813, "bench_c_p3m": 0.13999211845943282, "ret_p6m": 1.365905054923366, "ret_signed_p6m": 1.365905054923366, "alpha_spy_p6m": 1.206963007183769, "alpha_c_p6m": 0.9640745153545607, "bench_spy_p6m": 0.15894204773959708, "bench_c_p6m": 0.4018305395688053, "price_path": [-0.1645, -0.165, -0.1156, -0.1121, -0.0907, -0.0841, -0.0974, -0.0863, -0.0663, -0.0418, -0.0596, -0.0752, -0.0485, -0.0796, -0.1138, -0.1521, -0.1241, -0.1201, -0.1347, -0.1899, -0.2672, -0.2463, -0.2664, -0.1854, -0.1961, -0.1988, -0.197, -0.2263, -0.2219, -0.2219, -0.2148, -0.2272, -0.1952, -0.2072, -0.1801, -0.1908, -0.1961, -0.2108, -0.2108, -0.173, -0.173, -0.173, -0.1516, -0.1539, -0.1548, -0.133, -0.1174, -0.0841, -0.1085, -0.0907, -0.1134, -0.1018, -0.1085, -0.1245, -0.1107, -0.0974, -0.0996, -0.0752, -0.0557, -0.0891, -0.0757, -0.0757, -0.0312, 0.0, 0.0, 0.02, 0.0267, 0.069, 0.049, 0.049, 0.1047, 0.1091, 0.098, 0.0958, 0.1448, 0.1559, 0.2405, 0.2739, 0.3051, 0.265, 0.3007, 0.2717, 0.2428, 0.2405, 0.2049, 0.2071, 0.2561, 0.2517, 0.3229, 0.3118, 0.3363, 0.3296, 0.3185, 0.2004, 0.1982, 0.2138, 0.196, 0.1982, 0.2004, 0.1849, 0.167, 0.1693, 0.1737, 0.2183, 0.1492, 0.1492, 0.1737, 0.1514, 0.167, 0.1425, 0.1893, 0.1982, 0.2383, 0.2316, 0.2316, 0.1915, 0.1715, 0.1381, 0.0913, 0.118, 0.1559, 0.1648, 0.1581, 0.1977, 0.1999, 0.142, 0.162, 0.171, 0.1843, 0.22, 0.2356, 0.2847, 0.3561, 0.3695, 0.4654, 0.4765, 0.5524, 0.4877, 0.5858, 0.5858, 0.5657, 0.6103, 0.5947, 0.5903, 0.5011, 0.5568, 0.5501, 0.6059, 0.7642, 0.7642, 0.7642, 0.7642, 0.7642, 0.7642, 0.9092, 0.8512, 0.8356, 0.8847, 1.0185, 1.0765, 1.1657, 1.1367, 1.1479, 1.1345, 1.275, 1.3865, 1.3241, 1.4891, 1.5337, 1.4936, 1.7657, 1.614, 1.5828, 1.6452, 1.5873, 1.7032, 1.7612, 1.7523, 1.73, 1.498, 1.7032, 1.5426, 1.507, 1.5471, 1.3241, 1.3196, 1.3151, 1.3374, 1.4284, 1.3659]}
{"unit_id": "thread:1930444962774299047", "ticker_idx": 1, "ts": "2025-06-05T02:01:59+00:00", "symbol": "HBM", "kind": "sector", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The HBM market is unfolding as a completely different one from traditional DRAM… who makes the product determines its fundamental characteristics", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SK hynix on durable HBM leadership and continuing upcycle; bearish NAND on absent AI tailwind.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "250605_Semiconductor Market and Outlook for SK hynix (with Meritz Securities Analyst Sunwoo Kim) – SK hynix\n\n⸻\n\n[Key Takeaways]\n\n(1) The AI upcycle will continue, and SK hynix will maintain its leadership in the memory industry based on its technological competitiveness.\n\n(2) The main drivers of AI demand will expand from enterprises to nation-states, and the utility of AI in the multimodal era is only beginning to be discovered.\n\n⸻\n\n[Contents]\n\n(1) SK hynix’s 1Q25 earnings exceeded market expectations significantly.\n\n(2) The title of the Meritz Securities report was “Differentiated Leadership Within the Upcycle.”\n\n(3) This title carries two meanings.\n\n(4) First, that the upcycle is ongoing and will continue into next year. Second—and more importantly—is “leadership.”\n\n(5) The HBM market is unfolding as a completely different one from traditional DRAM.\n\n(6) In the past, competition in memory was all about cost. Neither SK hynix nor other companies could command a premium on product prices.\n\n(7) That era of commodity memory has come to an end due to AI, and we are now in the age of specialized purpose semiconductors [like HBM].\n\n(8) The key difference now is that who makes the product determines its fundamental characteristics.\n\n(9) In this regard, SK hynix’s leadership—already having secured the market—is strongly reflected in recent earnings, suggesting that profit improvement will continue for quite a long time.\n\n(10) The risk to watch out for is that nothing lasts forever. Once second and third vendors enter the market and reach similar quality levels, profit margins will inevitably decline.\n\n(11) It’s true that SK hynix was well prepared when partners and customers within the value chain first approached them, which helped seize the opportunity. But going forward, the company must continue to ask what products can maintain overwhelming competitiveness—and pursue bold investments and strategic direction accordingly.\n\n(12) The most fundamental strength of SK hynix in terms of technology lies in front-end process competitiveness. The company’s leadership in the much more high-end HBM market has been made possible thanks to its 1-b process technology edge.\n\n(13) The 1-c process is also expected to produce products soon, and within the industry, it is seen as having the best energy efficiency [low power, high performance], likely leading to a successful market adoption.\n\n(14) The future task will be to accurately foresee trends and solidify partnerships within the value chain. If customers accept the products proposed by the company, then technological leadership can be sufficiently maintained.\n\n(15) Furthermore, future AI memory demand drivers will move to the nation-state level. Until now, AI investment was mainly driven by corporate competition, but going forward, some countries are expected to position AI as a national growth engine.\n\n(16) By product:\n1. DRAM will serve as a computational assistant and core differentiator, commanding a high price.\n2. NAND, on the other hand, does not see rising conventional demand and will not significantly benefit from AI, so strategic direction must be considered carefully.\n\n(17) AI has only just moved from text to image. Now, we are entering a full-fledged multimodal era—including video, audio, and more. The real utility of AI that stimulates primal human instincts is only now being discovered.\n\n(18) Always pay attention to the speed and nature of technological change. Especially:\n1. What technology is used for computation?\n2. Has memory become closer to the computation?\n3. Which company is leading at the forefront of that domain?\n\nThese will be the most crucial questions going forward.\n\n---\n\nOriginal: \nhttps://t.co/30O6gHrBbr", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02725929025536659, "ret_signed_m1d": -0.02725929025536659, "alpha_spy_m1d": -0.032115528388937206, "alpha_c_m1d": -0.02904999576288229, "bench_spy_m1d": 0.004856238133570612, "bench_c_m1d": 0.0017907055075156997, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007118865814648601, "ret_signed_p1d": 0.007118865814648601, "alpha_spy_p1d": -0.0031501140443142006, "alpha_c_p1d": 0.0014284422518877626, "bench_spy_p1d": 0.010268979858962801, "bench_c_p1d": 0.005690423562760838, "ret_p1w": 0.049604493735145784, "ret_signed_p1w": 0.049604493735145784, "alpha_spy_p1w": 0.03156216011342853, "alpha_c_p1w": 0.0024497359707015623, "bench_spy_p1w": 0.018042333621717255, "bench_c_p1w": 0.04715475776444422, "ret_p1m": 0.14435567799352186, "ret_signed_p1m": 0.14435567799352186, "alpha_spy_p1m": 0.08679097143882361, "alpha_c_p1m": 0.015744406659610716, "bench_spy_p1m": 0.05756470655469825, "bench_c_p1m": 0.12861127133391115, "ret_p3m": 0.15138145541431847, "ret_signed_p3m": 0.15138145541431847, "alpha_spy_p3m": 0.06856745430027034, "alpha_c_p3m": 0.011389336954885643, "bench_spy_p3m": 0.08281400111404813, "bench_c_p3m": 0.13999211845943282, "ret_p6m": 1.1042740595892786, "ret_signed_p6m": 1.1042740595892786, "alpha_spy_p6m": 0.9453320118496815, "alpha_c_p6m": 0.7024435200204733, "bench_spy_p6m": 0.15894204773959708, "bench_c_p6m": 0.4018305395688053, "price_path": [-0.1477, -0.1422, -0.0978, -0.1002, -0.0745, -0.0488, -0.0576, -0.0478, -0.0286, -0.038, -0.0437, -0.0615, -0.05, -0.0612, -0.0881, -0.1189, -0.0982, -0.0972, -0.1536, -0.2106, -0.2413, -0.2415, -0.2123, -0.1907, -0.2026, -0.1932, -0.1903, -0.2162, -0.2141, -0.213, -0.2165, -0.212, -0.1889, -0.1788, -0.1623, -0.1692, -0.1762, -0.1826, -0.18, -0.1649, -0.1659, -0.1656, -0.1502, -0.143, -0.1399, -0.0967, -0.0809, -0.072, -0.0803, -0.0692, -0.0804, -0.0777, -0.0881, -0.1023, -0.1051, -0.0978, -0.0936, -0.0748, -0.0652, -0.0832, -0.0636, -0.0509, -0.0273, 0.0, 0.0071, 0.0252, 0.0341, 0.0581, 0.0496, 0.041, 0.0667, 0.0748, 0.0853, 0.0812, 0.1048, 0.0953, 0.1559, 0.1695, 0.1697, 0.158, 0.1594, 0.1446, 0.1411, 0.1591, 0.1444, 0.1289, 0.1577, 0.1437, 0.1736, 0.1834, 0.1724, 0.1817, 0.1721, 0.1342, 0.1393, 0.1448, 0.116, 0.121, 0.1254, 0.1182, 0.1377, 0.1418, 0.1634, 0.1539, 0.1033, 0.1169, 0.1302, 0.1157, 0.1402, 0.1615, 0.1877, 0.2039, 0.2109, 0.2101, 0.1963, 0.1822, 0.1708, 0.1474, 0.1279, 0.1473, 0.1565, 0.1529, 0.1563, 0.1772, 0.1691, 0.1378, 0.1514, 0.1591, 0.1825, 0.2134, 0.2223, 0.2585, 0.3035, 0.3457, 0.4099, 0.4215, 0.4666, 0.4419, 0.5144, 0.4951, 0.5125, 0.5398, 0.5238, 0.511, 0.4667, 0.5134, 0.5202, 0.5973, 0.6765, 0.6896, 0.6995, 0.6829, 0.717, 0.7038, 0.7449, 0.7544, 0.7216, 0.7726, 0.8659, 0.8859, 0.9305, 0.9034, 0.9014, 0.9108, 1.0094, 1.0682, 1.0389, 1.1144, 1.1416, 1.1468, 1.2924, 1.1541, 1.1803, 1.1958, 1.1678, 1.2701, 1.2677, 1.2743, 1.2402, 1.162, 1.2345, 1.1227, 1.0953, 1.055, 0.9664, 1.0278, 1.043, 1.0884, 1.1227, 1.1043]}
{"unit_id": "thread:1930444962774299047", "ticker_idx": 2, "ts": "2025-06-05T02:01:59+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": "NAND", "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND, on the other hand, does not see rising conventional demand and will not significantly benefit from AI, so strategic direction must be considered carefully", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SK hynix on durable HBM leadership and continuing upcycle; bearish NAND on absent AI tailwind.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket: dominant producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "250605_Semiconductor Market and Outlook for SK hynix (with Meritz Securities Analyst Sunwoo Kim) – SK hynix\n\n⸻\n\n[Key Takeaways]\n\n(1) The AI upcycle will continue, and SK hynix will maintain its leadership in the memory industry based on its technological competitiveness.\n\n(2) The main drivers of AI demand will expand from enterprises to nation-states, and the utility of AI in the multimodal era is only beginning to be discovered.\n\n⸻\n\n[Contents]\n\n(1) SK hynix’s 1Q25 earnings exceeded market expectations significantly.\n\n(2) The title of the Meritz Securities report was “Differentiated Leadership Within the Upcycle.”\n\n(3) This title carries two meanings.\n\n(4) First, that the upcycle is ongoing and will continue into next year. Second—and more importantly—is “leadership.”\n\n(5) The HBM market is unfolding as a completely different one from traditional DRAM.\n\n(6) In the past, competition in memory was all about cost. Neither SK hynix nor other companies could command a premium on product prices.\n\n(7) That era of commodity memory has come to an end due to AI, and we are now in the age of specialized purpose semiconductors [like HBM].\n\n(8) The key difference now is that who makes the product determines its fundamental characteristics.\n\n(9) In this regard, SK hynix’s leadership—already having secured the market—is strongly reflected in recent earnings, suggesting that profit improvement will continue for quite a long time.\n\n(10) The risk to watch out for is that nothing lasts forever. Once second and third vendors enter the market and reach similar quality levels, profit margins will inevitably decline.\n\n(11) It’s true that SK hynix was well prepared when partners and customers within the value chain first approached them, which helped seize the opportunity. But going forward, the company must continue to ask what products can maintain overwhelming competitiveness—and pursue bold investments and strategic direction accordingly.\n\n(12) The most fundamental strength of SK hynix in terms of technology lies in front-end process competitiveness. The company’s leadership in the much more high-end HBM market has been made possible thanks to its 1-b process technology edge.\n\n(13) The 1-c process is also expected to produce products soon, and within the industry, it is seen as having the best energy efficiency [low power, high performance], likely leading to a successful market adoption.\n\n(14) The future task will be to accurately foresee trends and solidify partnerships within the value chain. If customers accept the products proposed by the company, then technological leadership can be sufficiently maintained.\n\n(15) Furthermore, future AI memory demand drivers will move to the nation-state level. Until now, AI investment was mainly driven by corporate competition, but going forward, some countries are expected to position AI as a national growth engine.\n\n(16) By product:\n1. DRAM will serve as a computational assistant and core differentiator, commanding a high price.\n2. NAND, on the other hand, does not see rising conventional demand and will not significantly benefit from AI, so strategic direction must be considered carefully.\n\n(17) AI has only just moved from text to image. Now, we are entering a full-fledged multimodal era—including video, audio, and more. The real utility of AI that stimulates primal human instincts is only now being discovered.\n\n(18) Always pay attention to the speed and nature of technological change. Especially:\n1. What technology is used for computation?\n2. Has memory become closer to the computation?\n3. Which company is leading at the forefront of that domain?\n\nThese will be the most crucial questions going forward.\n\n---\n\nOriginal: \nhttps://t.co/30O6gHrBbr", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01806984371973732, "ret_signed_m1d": 0.01806984371973732, "alpha_spy_m1d": 0.022926081853307932, "alpha_c_m1d": 0.01986054922725302, "bench_spy_m1d": 0.004856238133570612, "bench_c_m1d": 0.0017907055075156997, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009056086950027575, "ret_signed_p1d": -0.009056086950027575, "alpha_spy_p1d": 0.0012128929089352258, "alpha_c_p1d": -0.0033656633872667374, "bench_spy_p1d": 0.010268979858962801, "bench_c_p1d": 0.005690423562760838, "ret_p1w": 0.03826138954408427, "ret_signed_p1w": -0.03826138954408427, "alpha_spy_p1w": -0.020219055922367016, "alpha_c_p1w": 0.008893368220359951, "bench_spy_p1w": 0.018042333621717255, "bench_c_p1w": 0.04715475776444422, "ret_p1m": 0.1488360921654261, "ret_signed_p1m": -0.1488360921654261, "alpha_spy_p1m": -0.09127138561072784, "alpha_c_p1m": -0.020224820831514945, "bench_spy_p1m": 0.05756470655469825, "bench_c_p1m": 0.12861127133391115, "ret_p3m": 0.19918192755288672, "ret_signed_p3m": -0.19918192755288672, "alpha_spy_p3m": -0.11636792643883859, "alpha_c_p3m": -0.05918980909345389, "bench_spy_p3m": 0.08281400111404813, "bench_c_p3m": 0.13999211845943282, "ret_p6m": 2.2931136699378643, "ret_signed_p6m": -2.2931136699378643, "alpha_spy_p6m": -2.1341716221982674, "alpha_c_p6m": -1.891283130369059, "bench_spy_p6m": 0.15894204773959708, "bench_c_p6m": 0.4018305395688053, "price_path": [0.0426, 0.0316, 0.0562, 0.0468, 0.1153, 0.1506, 0.1535, 0.1457, 0.1566, 0.1338, 0.1088, 0.0829, 0.0904, 0.0736, 0.0399, -0.0022, 0.0002, 0.0043, -0.1014, -0.1944, -0.2403, -0.2094, -0.1816, -0.161, -0.1708, -0.1605, -0.1606, -0.1958, -0.1951, -0.1994, -0.207, -0.209, -0.1842, -0.1708, -0.1527, -0.1583, -0.1591, -0.1701, -0.1681, -0.1506, -0.1501, -0.1539, -0.1299, -0.1157, -0.0976, -0.0371, -0.0252, -0.0143, -0.0316, -0.0249, -0.0426, -0.0346, -0.0501, -0.0718, -0.0779, -0.0718, -0.0651, -0.0493, -0.0358, -0.0588, -0.0588, -0.0456, -0.0181, 0.0, 0.0091, 0.0286, 0.0346, 0.0476, 0.0383, 0.0319, 0.0559, 0.0673, 0.0829, 0.0858, 0.1169, 0.1386, 0.1765, 0.1824, 0.1835, 0.1743, 0.1641, 0.14, 0.1489, 0.1735, 0.1488, 0.1411, 0.1721, 0.1784, 0.1859, 0.1808, 0.151, 0.159, 0.1465, 0.1209, 0.1216, 0.122, 0.1077, 0.132, 0.1328, 0.1203, 0.1287, 0.1341, 0.1516, 0.146, 0.1024, 0.1126, 0.1165, 0.1052, 0.1096, 0.147, 0.1578, 0.2092, 0.2105, 0.2062, 0.1874, 0.1729, 0.1642, 0.1383, 0.1272, 0.1525, 0.1591, 0.1638, 0.1697, 0.1999, 0.2151, 0.2039, 0.1992, 0.2144, 0.2772, 0.3682, 0.3915, 0.4157, 0.4894, 0.6169, 0.7059, 0.7332, 0.7927, 0.7652, 0.8414, 0.8478, 0.8893, 0.9234, 0.8925, 0.8508, 0.792, 0.9302, 0.9465, 1.0223, 1.1519, 1.2544, 1.2365, 1.1858, 1.2602, 1.2731, 1.2269, 1.323, 1.2508, 1.3987, 1.5074, 1.4681, 1.591, 1.5707, 1.583, 1.6971, 1.9873, 2.0705, 2.0073, 2.2662, 2.3163, 2.3303, 2.4569, 2.3042, 2.4117, 2.4662, 2.663, 2.9909, 2.9962, 3.0634, 2.8205, 2.5441, 2.7553, 2.4956, 2.5388, 2.3062, 2.1517, 2.325, 2.2844, 2.145, 2.2283, 2.2931]}
{"unit_id": "orig:1927959720259715557", "ticker_idx": 0, "ts": "2025-05-29T05:26:31+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "J.P. Morgan maintains NVIDIA's target price at $170", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares JPM maintaining NVDA PT at $170; implied continued bullish stance.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan maintains NVIDIA’s target price at $170.\n\n$NVDA https://t.co/2RcYsZ31HH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031467771192964955, "ret_signed_m1d": -0.031467771192964955, "alpha_spy_m1d": -0.027536013589865016, "alpha_c_m1d": -0.02442091775738242, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029168709381678815, "ret_signed_p1d": -0.029168709381678815, "alpha_spy_p1d": -0.028050357189775443, "alpha_c_p1d": -0.011428601389960602, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.005747599702493611, "ret_signed_p1w": 0.005747599702493611, "alpha_spy_p1w": 0.0006632770562644552, "alpha_c_p1w": -0.023832855030635036, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.13342157350479011, "ret_signed_p1m": 0.13342157350479011, "alpha_spy_p1m": 0.08820871516440842, "alpha_c_p1m": -0.007270048081713831, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.30600340555452554, "ret_signed_p3m": 0.30600340555452554, "alpha_spy_p3m": 0.20937208449485212, "alpha_c_p3m": 0.09021349223987807, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.2853115202986203, "ret_signed_p6m": 0.2853115202986203, "alpha_spy_p6m": 0.16199846496754655, "alpha_c_p6m": -0.05084882377385047, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [-0.1806, -0.1668, -0.1573, -0.2057, -0.1905, -0.2315, -0.2187, -0.1685, -0.1696, -0.1259, -0.1412, -0.1707, -0.1557, -0.1484, -0.1544, -0.1277, -0.1329, -0.1827, -0.1994, -0.2121, -0.2214, -0.2086, -0.2067, -0.2686, -0.3224, -0.2985, -0.3081, -0.1786, -0.2272, -0.203, -0.2046, -0.1939, -0.2493, -0.2709, -0.2709, -0.3038, -0.2895, -0.2621, -0.2354, -0.2025, -0.2188, -0.2168, -0.2175, -0.1981, -0.1774, -0.1823, -0.1843, -0.159, -0.1568, -0.1619, -0.1163, -0.0665, -0.0277, -0.0313, -0.0272, -0.026, -0.0346, -0.0531, -0.0457, -0.0568, -0.0568, -0.0265, -0.0315, 0.0, -0.0292, -0.013, 0.0146, 0.0196, 0.0057, 0.0182, 0.0247, 0.0343, 0.0262, 0.0418, 0.02, 0.0396, 0.0355, 0.0453, 0.0453, 0.0336, 0.0359, 0.0627, 0.1087, 0.1138, 0.1334, 0.1351, 0.1014, 0.1298, 0.1448, 0.1448, 0.1369, 0.1496, 0.1703, 0.179, 0.1849, 0.1788, 0.2265, 0.2313, 0.243, 0.2388, 0.2314, 0.2001, 0.227, 0.2483, 0.2466, 0.2699, 0.261, 0.288, 0.278, 0.2482, 0.2933, 0.2808, 0.2891, 0.2988, 0.3127, 0.3081, 0.316, 0.3047, 0.3078, 0.2965, 0.3077, 0.262, 0.2602, 0.2572, 0.2788, 0.2919, 0.306, 0.3048, 0.2945, 0.2515, 0.2515, 0.227, 0.2259, 0.2334, 0.2, 0.2093, 0.2269, 0.2741, 0.273, 0.2777, 0.2772, 0.2566, 0.2236, 0.2663, 0.2694, 0.3193, 0.2821, 0.2716, 0.2768, 0.2804, 0.3067, 0.3406, 0.3454, 0.3572, 0.3481, 0.3332, 0.3296, 0.3588, 0.3837, 0.3161, 0.3531, 0.2936, 0.2921, 0.3064, 0.3165, 0.3123, 0.3017, 0.2954, 0.3089, 0.3383, 0.3759, 0.4445, 0.4877, 0.4578, 0.455, 0.4865, 0.4277, 0.4026, 0.3514, 0.3519, 0.4302, 0.3879, 0.3925, 0.3427, 0.3664, 0.3408, 0.3031, 0.3402, 0.298, 0.2853]}
{"unit_id": "orig:1934944268810129479", "ticker_idx": 0, "ts": "2025-06-17T12:00:37+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "FY26 valuation downgrade pressure remains... significant pressure from declining DRAM revenue in FY26... future valuation and earnings forecast adjustments are likely. Maintain 'In-Line' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley In-Line on AMAT: FY26 under pressure from -12% DRAM YoY and TSMC share uncertainty; valuation downgrade risk if earnings not revised up.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "AMAT: Valuation Still Attractive, but FY26 Outlook Under Pressure Due to Weak DRAM and TSMC Share Uncertainty\n\nMorgan Stanley (S. Brett, 2025/06/16)\n\nWhile the short-term outlook is slightly positive, market pessimism for FY26 is increasing, primarily due to a lack of evidence for TSMC's share growth and the high base effect on DRAM revenue. Simultaneously, valuation remains polarized, but the possibility of future valuation downgrades cannot be ruled out.\n\nMore conservative view on TSMC, despite technological catalysts.\n\nRegarding TSMC, optimists believe GAA (Gate-All-Around) and backside power delivery technologies will drive AMAT's market share up in 2026. However, Morgan Stanley is more conservative, noting no clear signs of market share improvement in 2024. Short-term forecasts may remain unchanged, but if actual order growth occurs, an upward revision to forecasts will be considered.\n\nExpected FY26 DRAM weakness leads to significant divergence from market expectations.\n\nMorgan Stanley predicts AMAT's DRAM revenue in FY26 will decline by 12% year-over-year, significantly lower than the market consensus (-1%). They believe prior demand from China and Samsung created a high base for FY24-25, leading to more severe year-over-year pressure in FY26, especially for non-China customers starting from 1Q26.\n\nNeutral on overall SPE valuation, but earnings forecast upgrades needed to support stock upside.\n\nMarket views are divided: one side believes the SPE sector should command a higher valuation premium, while the other points out that current valuation is already one standard deviation above the historical average, and even higher when factoring in a discount for China business. The firm holds a neutral view, believing that if future earnings expectations are not revised upward, stock upside will be limited.\n\nShort-term TSMC revenue growth to support stock, but FY26 valuation downgrade pressure remains.\n\nAlthough TSMC is expected to show quarter-over-quarter growth in 2H FY25, alleviating short-term market concerns, significant pressure from declining DRAM revenue in FY26, coupled with current market expectations not fully reflecting a pessimistic scenario, suggests future valuation and earnings forecast adjustments are likely. Maintain \"In-Line\" rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01413066626075965, "ret_signed_m1d": -0.01413066626075965, "alpha_spy_m1d": -0.005511986828005577, "alpha_c_m1d": -0.007287870920409167, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007180156657963455, "ret_signed_p1d": 0.007180156657963455, "alpha_spy_p1d": 0.007029520903792275, "alpha_c_p1d": 0.011002891126656422, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.034981930193652255, "ret_signed_p1w": -0.034981930193652255, "alpha_spy_p1w": -0.01649960504710002, "alpha_c_p1w": 0.0030546639519211105, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.11901897885905055, "ret_signed_p1m": -0.11901897885905055, "alpha_spy_p1m": -0.07126362206876169, "alpha_c_p1m": -0.013166354365651278, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": -0.03336813208740608, "ret_signed_p3m": 0.03336813208740608, "alpha_spy_p3m": 0.13683309537846455, "alpha_c_p3m": 0.1967154272681345, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 0.5881410296028386, "ret_signed_p6m": -0.5881410296028386, "alpha_spy_p6m": -0.4308527128726718, "alpha_c_p6m": -0.1580404537468485, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [-0.1157, -0.1287, -0.1125, -0.12, -0.1369, -0.1541, -0.1691, -0.1688, -0.1657, -0.1537, -0.2238, -0.2728, -0.2391, -0.2613, -0.1424, -0.2082, -0.1698, -0.1722, -0.1669, -0.2085, -0.2126, -0.2126, -0.2235, -0.2078, -0.1776, -0.1398, -0.1319, -0.1363, -0.1434, -0.1368, -0.1468, -0.1116, -0.1144, -0.1235, -0.1075, -0.1065, -0.1087, -0.0378, -0.0089, -0.0025, 0.0009, -0.0516, -0.0495, -0.0507, -0.0708, -0.0779, -0.0952, -0.0952, -0.0704, -0.0718, -0.0839, -0.0996, -0.0966, -0.0709, -0.0698, -0.0569, -0.0422, -0.0247, -0.0018, -0.0075, 0.0052, -0.0201, 0.0141, 0.0, -0.0072, -0.0072, -0.0266, -0.0122, 0.035, 0.0516, 0.0542, 0.0524, 0.0516, 0.0555, 0.0914, 0.0974, 0.0974, 0.0959, 0.1201, 0.1224, 0.1375, 0.1369, 0.1322, 0.1448, 0.119, 0.1059, 0.0939, 0.1064, 0.075, 0.0742, 0.0806, 0.0666, 0.0929, 0.0823, 0.0879, 0.0343, 0.0339, 0.0501, 0.0291, 0.0233, 0.052, 0.0619, 0.0591, 0.0825, 0.0916, 0.0813, -0.0708, -0.0607, -0.0682, -0.0754, -0.0792, -0.064, -0.0668, -0.0523, -0.053, -0.0479, -0.0739, -0.0739, -0.0923, -0.0999, -0.0884, -0.0625, -0.0665, -0.0581, -0.0586, -0.0198, -0.0334, -0.0153, -0.0003, 0.0261, 0.0931, 0.0951, 0.1551, 0.1571, 0.1604, 0.1498, 0.1747, 0.1806, 0.1794, 0.2543, 0.288, 0.2531, 0.2899, 0.2187, 0.253, 0.2691, 0.2094, 0.2643, 0.2569, 0.311, 0.3118, 0.2961, 0.3142, 0.3019, 0.2706, 0.3161, 0.3177, 0.3326, 0.3113, 0.3581, 0.3396, 0.3428, 0.3694, 0.326, 0.3877, 0.3453, 0.3253, 0.3542, 0.3173, 0.3291, 0.2859, 0.302, 0.3175, 0.2968, 0.3545, 0.2711, 0.293, 0.3328, 0.3995, 0.4428, 0.4428, 0.456, 0.4704, 0.5315, 0.5505, 0.5552, 0.5469, 0.5478, 0.5419, 0.5881]}
{"unit_id": "reply:1937751047113765269", "ticker_idx": 0, "ts": "2025-06-25T05:53:45+00:00", "symbol": "Doosan Enerbility", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Keep a close eye on a Korean company called Doosan Enerbility — it's the TSMC of the nuclear power industry", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Doosan Enerbility as the dominant, must-watch nuclear power components play — 'the TSMC of nuclear'.", "resolved_tickers": ["034020.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Enerbility 034020.KS Korea nuclear/power", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@RihardJarc Keep a close eye on a Korean company called Doosan Enerbility — it’s the TSMC of the nuclear power industry.", "reply_text": null, "parent_text": "@Jukanlosreve I agree, wrote about this in december last year: https://t.co/bpFkGGekRt", "quoted_text": null, "parent_author": "RihardJarc", "ret_m1d": 0.04711246200607899, "ret_signed_m1d": 0.04711246200607899, "alpha_spy_m1d": 0.047672512804024625, "alpha_c_m1d": 0.038209917752585865, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": 0.008902544253493128, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016717325227963542, "ret_signed_p1d": -0.016717325227963542, "alpha_spy_p1d": -0.02454110510245089, "alpha_c_p1d": -0.02756727087851152, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.010849945650547976, "ret_p1w": -0.06382978723404253, "ret_signed_p1w": -0.06382978723404253, "alpha_spy_p1w": -0.08578592787573469, "alpha_c_p1w": -0.09429303843681303, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.03046325120277049, "ret_p1m": -0.006079027355623046, "ret_signed_p1m": -0.006079027355623046, "alpha_spy_p1m": -0.05104540859306961, "alpha_c_p1m": -0.07347363461609224, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.0673946072604692, "ret_p3m": -0.05167173252279633, "ret_signed_p3m": -0.05167173252279633, "alpha_spy_p3m": -0.1530829003453028, "alpha_c_p3m": -0.12019396893867695, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.06852223641588062, "ret_p6m": 0.13221884498480252, "ret_signed_p6m": 0.13221884498480252, "alpha_spy_p6m": 0.014901991176614171, "alpha_c_p6m": 0.05301147520554261, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.0792073697792599, "price_path": [-0.6299, -0.6436, -0.6375, -0.636, -0.6413, -0.6497, -0.6862, -0.6847, -0.6847, -0.6634, -0.6444, -0.6451, -0.6337, -0.6451, -0.6284, -0.6056, -0.6041, -0.6102, -0.6087, -0.5912, -0.576, -0.5699, -0.5608, -0.56, -0.56, -0.579, -0.579, -0.579, -0.5851, -0.5783, -0.5767, -0.5623, -0.5494, -0.5243, -0.5144, -0.4787, -0.4742, -0.4347, -0.4058, -0.4187, -0.3799, -0.3693, -0.3564, -0.3951, -0.3587, -0.3868, -0.3777, -0.3777, -0.3518, -0.3024, -0.3024, -0.2781, -0.2705, -0.2234, -0.1702, -0.1702, -0.0942, -0.0973, -0.0729, -0.0729, -0.0851, 0.0426, 0.0471, 0.0, -0.0167, 0.0, 0.0395, -0.0502, -0.0638, -0.0593, -0.0881, -0.0441, 0.0122, -0.0213, -0.041, -0.0805, -0.0851, -0.0076, -0.0334, -0.038, -0.0152, 0.0395, -0.0486, -0.0365, -0.0061, 0.0015, -0.035, -0.0106, -0.0319, -0.003, -0.0669, -0.0228, -0.0152, 0.0076, -0.0061, 0.0091, 0.0547, -0.0046, 0.0106, -0.0046, -0.0046, -0.0106, -0.0957, -0.1277, -0.0653, -0.0547, 0.0015, -0.038, -0.041, -0.0441, -0.0623, -0.0942, -0.0821, -0.0517, -0.0456, -0.0608, -0.0745, -0.0547, -0.0486, -0.0729, -0.0729, -0.1064, -0.038, -0.0684, -0.076, -0.076, -0.0517, -0.041, 0.0015, -0.0274, -0.041, -0.0441, -0.0471, -0.0182, -0.0091, -0.0091, -0.0091, -0.0091, -0.0091, -0.0091, 0.1322, 0.1793, 0.152, 0.2599, 0.272, 0.2188, 0.2204, 0.2158, 0.2006, 0.1596, 0.2295, 0.2447, 0.3131, 0.465, 0.3632, 0.348, 0.3602, 0.3602, 0.2705, 0.2052, 0.1839, 0.2097, 0.1884, 0.1915, 0.2629, 0.1915, 0.1976, 0.1459, 0.1307, 0.1809, 0.1109, 0.1018, 0.117, 0.1809, 0.1793, 0.1611, 0.1383, 0.1398, 0.1915, 0.2036, 0.2219, 0.1672, 0.1657, 0.1672, 0.1748, 0.2112, 0.1717, 0.1748, 0.1474, 0.1322]}
{"unit_id": "orig:1939450639488356630", "ticker_idx": 0, "ts": "2025-06-29T22:27:20+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM margins are also expected to remain strong, suggesting upside potential for earnings compared to market expectations", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman investor meeting: SK Hynix top pick on HBM upside; Samsung re-rating potential on HBM4 catalyst + book value; Micron in contention but secondary preference.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Global Investor Meeting Feedback on Korean Memory Sector (Goldman Sachs)\n\n1. HBM Pricing and SK Hynix Earnings Expectations\n\n- HBM4 price negotiations are understood to be in their final stages for 2026.\n- Investors are concerned that the HBM4 price increase rate might be lower than the cost increase rate.\n- However, SK Hynix anticipates an increase in HBM shipments due to strong HBM demand and enhanced competitiveness.\n- HBM margins are also expected to remain strong, suggesting upside potential for earnings compared to market expectations.\n\n2. Low Concerns Regarding Samsung Electronics' HBM Qualification Delays, but HBM4 Expectations Exist\n\n- SK Hynix is already shipping the majority of its 12-layer HBM3E.\n- Samsung Electronics is experiencing shipment delays due to qualification delays with a major GPU customer.\n- Investors did not express significant concerns about Samsung Electronics' qualification delays, judging it more important to fo- cus resources on the future 3rd generation HBM4 and 1c nm.\n- If news of HBM4 progress emerges in Q3, it could act as a positive catalyst compared to current low expectations.\n\n3. Traditional Memory Price Increases Are Positive but Impact Is Limited\n\n- Recent memory spot price increases have been notable, leading to expectations of ASP upside.\n- Strong price trends are expected to continue until Q3, but some price increases might be attributed to preemptive demand due to tariff issues.\n- The price increase is centered on DDR4, where Korean companies have less than 10% exposure, thus limiting the actual impact on earnings.\n\n4. SK Hynix Expected to Increase Capex, Samsung Electronics Opinions Divided\n\n- SK Hynix is expected to see capital expenditure of approximately ₩23 trillion (+$40% YoY) this year due to strong HBM, with potential for further increases if demand exceeds expectations and competitor expansions are delayed.\n- Opinions among investors are divided regarding Samsung Electronics' memory Capex. While it's generally expected to remain at last year's level, there's room for an increase depending on HBM qualification progress in Q3.\n\n5. Investor Preference: SK Hynix Maintains Lead, Interest in Samsung Electronics Growing\n\n- At the beginning of the meeting, many questions were about choosing between SK Hynix and Micron, but as time passed, inquiries regarding Samsung Electronics increased.\n- Investor preference still favored SK Hynix, with many investors contemplating entry timing due to the recent sharp rise in stock price.\n- Samsung Electronics is trading below its book value, and with limited negative catalysts expected after the Q2 earnings announcement, its valuation appeal is highlighted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027397398657104444, "ret_signed_m1d": -0.027397398657104444, "alpha_spy_m1d": -0.02263884459205001, "alpha_c_m1d": -0.02574800112189224, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0222603326750217, "ret_signed_p1d": -0.0222603326750217, "alpha_spy_p1d": -0.021936653366796843, "alpha_c_p1d": -0.011359555912624608, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.07191784907164522, "ret_signed_p1w": -0.07191784907164522, "alpha_spy_p1w": -0.07649817528738789, "alpha_c_p1w": -0.07507339485526976, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.10102747767706732, "ret_signed_p1m": -0.10102747767706732, "alpha_spy_p1m": -0.12920570911285345, "alpha_c_p1m": -0.1512999641652567, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.22265372912559833, "ret_signed_p3m": 0.22265372912559833, "alpha_spy_p3m": 0.1546366419664018, "alpha_c_p3m": 0.07140586318322062, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.0043232127011406, "ret_signed_p6m": 1.0043232127011406, "alpha_spy_p6m": 0.8844627424924387, "alpha_c_p6m": 0.6980623846192748, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.3235, -0.3348, -0.3772, -0.4366, -0.4206, -0.436, -0.3737, -0.3819, -0.384, -0.3826, -0.4052, -0.4018, -0.4018, -0.3963, -0.4059, -0.3813, -0.3905, -0.3696, -0.3778, -0.3819, -0.3932, -0.3932, -0.3642, -0.3642, -0.3642, -0.3478, -0.3495, -0.3501, -0.3334, -0.3214, -0.2958, -0.3146, -0.3009, -0.3184, -0.3095, -0.3146, -0.3269, -0.3163, -0.306, -0.3078, -0.289, -0.274, -0.2997, -0.2894, -0.2894, -0.2551, -0.2312, -0.2312, -0.2158, -0.2106, -0.1781, -0.1935, -0.1935, -0.1507, -0.1473, -0.1558, -0.1575, -0.1199, -0.1113, -0.0462, -0.0205, 0.0034, -0.0274, 0.0, -0.0223, -0.0445, -0.0462, -0.0736, -0.0719, -0.0342, -0.0377, 0.0171, 0.0086, 0.0274, 0.0223, 0.0137, -0.0771, -0.0788, -0.0668, -0.0805, -0.0788, -0.0771, -0.089, -0.1027, -0.101, -0.0976, -0.0634, -0.1164, -0.1164, -0.0976, -0.1147, -0.1027, -0.1216, -0.0856, -0.0788, -0.0479, -0.0531, -0.0531, -0.0839, -0.0993, -0.125, -0.161, -0.1404, -0.1113, -0.1045, -0.1096, -0.0792, -0.0774, -0.122, -0.1066, -0.0997, -0.0894, -0.062, -0.05, -0.0123, 0.0426, 0.0529, 0.1266, 0.1352, 0.1935, 0.1438, 0.2192, 0.2192, 0.2038, 0.2381, 0.2261, 0.2227, 0.1541, 0.1969, 0.1918, 0.2347, 0.3564, 0.3564, 0.3564, 0.3564, 0.3564, 0.3564, 0.4679, 0.4233, 0.4113, 0.449, 0.5519, 0.5965, 0.6651, 0.6428, 0.6514, 0.6411, 0.7491, 0.8348, 0.7868, 0.9137, 0.948, 0.9171, 1.1264, 1.0097, 0.9857, 1.0338, 0.9892, 1.0783, 1.1229, 1.1161, 1.0989, 0.9206, 1.0783, 0.9549, 0.9274, 0.9583, 0.7868, 0.7834, 0.78, 0.7971, 0.867, 0.819, 0.8464, 0.9151, 0.8945, 0.8602, 0.867, 0.9803, 0.9425, 1.0146, 0.9391, 0.9597, 0.9014, 0.819, 0.8911, 0.8945, 0.8773, 0.9906, 1.0043]}
{"unit_id": "orig:1939450639488356630", "ticker_idx": 1, "ts": "2025-06-29T22:27:20+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics is trading below its book value, and with limited negative catalysts expected after the Q2 earnings announcement, its valuation appeal is highlighted", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman investor meeting: SK Hynix top pick on HBM upside; Samsung re-rating potential on HBM4 catalyst + book value; Micron in contention but secondary preference.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Global Investor Meeting Feedback on Korean Memory Sector (Goldman Sachs)\n\n1. HBM Pricing and SK Hynix Earnings Expectations\n\n- HBM4 price negotiations are understood to be in their final stages for 2026.\n- Investors are concerned that the HBM4 price increase rate might be lower than the cost increase rate.\n- However, SK Hynix anticipates an increase in HBM shipments due to strong HBM demand and enhanced competitiveness.\n- HBM margins are also expected to remain strong, suggesting upside potential for earnings compared to market expectations.\n\n2. Low Concerns Regarding Samsung Electronics' HBM Qualification Delays, but HBM4 Expectations Exist\n\n- SK Hynix is already shipping the majority of its 12-layer HBM3E.\n- Samsung Electronics is experiencing shipment delays due to qualification delays with a major GPU customer.\n- Investors did not express significant concerns about Samsung Electronics' qualification delays, judging it more important to fo- cus resources on the future 3rd generation HBM4 and 1c nm.\n- If news of HBM4 progress emerges in Q3, it could act as a positive catalyst compared to current low expectations.\n\n3. Traditional Memory Price Increases Are Positive but Impact Is Limited\n\n- Recent memory spot price increases have been notable, leading to expectations of ASP upside.\n- Strong price trends are expected to continue until Q3, but some price increases might be attributed to preemptive demand due to tariff issues.\n- The price increase is centered on DDR4, where Korean companies have less than 10% exposure, thus limiting the actual impact on earnings.\n\n4. SK Hynix Expected to Increase Capex, Samsung Electronics Opinions Divided\n\n- SK Hynix is expected to see capital expenditure of approximately ₩23 trillion (+$40% YoY) this year due to strong HBM, with potential for further increases if demand exceeds expectations and competitor expansions are delayed.\n- Opinions among investors are divided regarding Samsung Electronics' memory Capex. While it's generally expected to remain at last year's level, there's room for an increase depending on HBM qualification progress in Q3.\n\n5. Investor Preference: SK Hynix Maintains Lead, Interest in Samsung Electronics Growing\n\n- At the beginning of the meeting, many questions were about choosing between SK Hynix and Micron, but as time passed, inquiries regarding Samsung Electronics increased.\n- Investor preference still favored SK Hynix, with many investors contemplating entry timing due to the recent sharp rise in stock price.\n- Samsung Electronics is trading below its book value, and with limited negative catalysts expected after the Q2 earnings announcement, its valuation appeal is highlighted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016722402281645454, "ret_signed_m1d": 0.016722402281645454, "alpha_spy_m1d": 0.021480956346699887, "alpha_c_m1d": 0.026436822514976144, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.00971442023333069, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0066889344851499555, "ret_signed_p1d": 0.0066889344851499555, "alpha_spy_p1d": 0.007012613793374811, "alpha_c_p1d": 0.015613638728858459, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.008924704243708503, "ret_p1w": 0.031772537907617915, "ret_signed_p1w": 0.031772537907617915, "alpha_spy_p1w": 0.027192211691875245, "alpha_c_p1w": 0.025098702711477294, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.006673835196140621, "ret_p1m": 0.18060189178675334, "ret_signed_p1m": 0.18060189178675334, "alpha_spy_p1m": 0.15242366035096722, "alpha_c_p1m": 0.1377160119485783, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.042885879838175045, "ret_p3m": 0.43979932535856836, "ret_signed_p3m": 0.43979932535856836, "alpha_spy_p3m": 0.37178223819937184, "alpha_c_p3m": 0.3405721434951374, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.09922718186343094, "ret_p6m": 0.8728673000811853, "ret_signed_p6m": 0.8728673000811853, "alpha_spy_p6m": 0.7530068298724835, "alpha_c_p6m": 0.7169281887388539, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.1559391113423314, "price_path": [-0.0227, -0.0427, -0.0676, -0.1158, -0.1108, -0.1191, -0.0626, -0.0826, -0.0659, -0.0593, -0.0909, -0.0842, -0.0809, -0.0792, -0.0859, -0.0742, -0.0742, -0.0742, -0.0726, -0.0726, -0.0776, -0.0776, -0.0975, -0.0975, -0.0975, -0.0925, -0.0925, -0.0892, -0.0427, -0.0543, -0.046, -0.0476, -0.056, -0.0726, -0.0709, -0.0742, -0.0909, -0.0992, -0.0909, -0.1042, -0.0709, -0.0676, -0.0659, -0.056, -0.056, -0.0393, -0.0177, -0.0177, -0.0061, -0.0161, -0.0044, -0.0111, -0.031, -0.0493, -0.0344, -0.0061, -0.0161, -0.0111, -0.036, 0.0055, 0.0188, 0.0006, 0.0167, 0.0, 0.0067, 0.0167, 0.0669, 0.0585, 0.0318, 0.0268, 0.01, 0.0201, 0.0468, 0.0452, 0.0652, 0.0819, 0.1154, 0.1221, 0.1338, 0.1037, 0.1104, 0.1037, 0.102, 0.1773, 0.1806, 0.214, 0.194, 0.1522, 0.1656, 0.1689, 0.1505, 0.1789, 0.2007, 0.1873, 0.189, 0.2023, 0.1973, 0.1973, 0.1706, 0.1706, 0.1789, 0.1806, 0.194, 0.1957, 0.1756, 0.1806, 0.1639, 0.1656, 0.1304, 0.1555, 0.1672, 0.1722, 0.1622, 0.1722, 0.1957, 0.214, 0.2274, 0.2609, 0.2793, 0.3278, 0.3077, 0.3428, 0.3428, 0.3963, 0.4164, 0.4281, 0.4398, 0.393, 0.4143, 0.4093, 0.4445, 0.5075, 0.5075, 0.5075, 0.5075, 0.5075, 0.5075, 0.5856, 0.5672, 0.5386, 0.5957, 0.6411, 0.6444, 0.6478, 0.6377, 0.6562, 0.6209, 0.6595, 0.7133, 0.6713, 0.6881, 0.7486, 0.8057, 0.8661, 0.762, 0.6898, 0.6663, 0.6444, 0.6898, 0.7385, 0.7318, 0.7267, 0.6327, 0.6898, 0.6427, 0.6209, 0.6898, 0.5924, 0.6243, 0.6679, 0.7267, 0.7385, 0.6881, 0.6931, 0.7368, 0.7553, 0.7654, 0.8208, 0.8393, 0.8208, 0.8141, 0.8023, 0.8292, 0.7603, 0.7267, 0.8124, 0.8074, 0.7855, 0.8561, 0.8729]}
{"unit_id": "orig:1939450639488356630", "ticker_idx": 2, "ts": "2025-06-29T22:27:20+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "many questions were about choosing between SK Hynix and Micron", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman investor meeting: SK Hynix top pick on HBM upside; Samsung re-rating potential on HBM4 catalyst + book value; Micron in contention but secondary preference.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Global Investor Meeting Feedback on Korean Memory Sector (Goldman Sachs)\n\n1. HBM Pricing and SK Hynix Earnings Expectations\n\n- HBM4 price negotiations are understood to be in their final stages for 2026.\n- Investors are concerned that the HBM4 price increase rate might be lower than the cost increase rate.\n- However, SK Hynix anticipates an increase in HBM shipments due to strong HBM demand and enhanced competitiveness.\n- HBM margins are also expected to remain strong, suggesting upside potential for earnings compared to market expectations.\n\n2. Low Concerns Regarding Samsung Electronics' HBM Qualification Delays, but HBM4 Expectations Exist\n\n- SK Hynix is already shipping the majority of its 12-layer HBM3E.\n- Samsung Electronics is experiencing shipment delays due to qualification delays with a major GPU customer.\n- Investors did not express significant concerns about Samsung Electronics' qualification delays, judging it more important to fo- cus resources on the future 3rd generation HBM4 and 1c nm.\n- If news of HBM4 progress emerges in Q3, it could act as a positive catalyst compared to current low expectations.\n\n3. Traditional Memory Price Increases Are Positive but Impact Is Limited\n\n- Recent memory spot price increases have been notable, leading to expectations of ASP upside.\n- Strong price trends are expected to continue until Q3, but some price increases might be attributed to preemptive demand due to tariff issues.\n- The price increase is centered on DDR4, where Korean companies have less than 10% exposure, thus limiting the actual impact on earnings.\n\n4. SK Hynix Expected to Increase Capex, Samsung Electronics Opinions Divided\n\n- SK Hynix is expected to see capital expenditure of approximately ₩23 trillion (+$40% YoY) this year due to strong HBM, with potential for further increases if demand exceeds expectations and competitor expansions are delayed.\n- Opinions among investors are divided regarding Samsung Electronics' memory Capex. While it's generally expected to remain at last year's level, there's room for an increase depending on HBM qualification progress in Q3.\n\n5. Investor Preference: SK Hynix Maintains Lead, Interest in Samsung Electronics Growing\n\n- At the beginning of the meeting, many questions were about choosing between SK Hynix and Micron, but as time passed, inquiries regarding Samsung Electronics increased.\n- Investor preference still favored SK Hynix, with many investors contemplating entry timing due to the recent sharp rise in stock price.\n- Samsung Electronics is trading below its book value, and with limited negative catalysts expected after the Q2 earnings announcement, its valuation appeal is highlighted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012251577573582306, "ret_signed_m1d": 0.012251577573582306, "alpha_spy_m1d": 0.01701013163863674, "alpha_c_m1d": 0.013900975108794511, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01914799578333326, "ret_signed_p1d": -0.01914799578333326, "alpha_spy_p1d": -0.018824316475108405, "alpha_c_p1d": -0.00824721902093617, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.02610236865156823, "ret_signed_p1w": -0.02610236865156823, "alpha_spy_p1w": -0.0306826948673109, "alpha_c_p1w": -0.029257914435192767, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.09074737744311612, "ret_signed_p1m": -0.09074737744311612, "alpha_spy_p1m": -0.11892560887890224, "alpha_c_p1m": -0.1410198639313055, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.2736522796191896, "ret_signed_p3m": 0.2736522796191896, "alpha_spy_p3m": 0.20563519245999307, "alpha_c_p3m": 0.12240441367681187, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.2450564955537793, "ret_signed_p6m": 1.2450564955537793, "alpha_spy_p6m": 1.1251960253450775, "alpha_c_p6m": 0.9387956674719136, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2811, -0.3968, -0.4749, -0.4453, -0.4682, -0.3682, -0.4316, -0.4357, -0.4238, -0.4236, -0.4375, -0.4418, -0.4418, -0.4585, -0.4303, -0.4083, -0.3718, -0.3527, -0.3626, -0.3762, -0.3757, -0.369, -0.3451, -0.3475, -0.3468, -0.3297, -0.3091, -0.3034, -0.2512, -0.2135, -0.2266, -0.2256, -0.2049, -0.1996, -0.2041, -0.2224, -0.2306, -0.2424, -0.2424, -0.218, -0.2196, -0.2146, -0.2336, -0.2034, -0.1704, -0.1623, -0.1376, -0.1192, -0.0998, -0.0739, -0.0586, -0.0574, -0.0621, -0.0277, -0.0236, -0.0116, -0.0116, 0.0028, -0.0095, 0.0378, 0.0325, 0.0223, 0.0123, 0.0, -0.0191, -0.0123, -0.0078, -0.0078, -0.0261, 0.0104, -0.0073, -0.0002, 0.0113, -0.0367, -0.0246, -0.0544, -0.0802, -0.071, -0.0804, -0.113, -0.108, -0.0926, -0.0964, -0.0965, -0.0907, -0.0682, -0.1136, -0.1482, -0.1248, -0.1143, -0.1166, -0.0915, -0.0345, 0.0048, 0.0375, 0.0092, 0.0175, -0.0184, 0.0034, -0.0088, -0.0481, -0.0596, -0.0443, -0.0545, -0.0539, -0.0437, -0.0092, -0.0335, -0.0335, -0.0378, -0.0358, 0.0087, 0.0669, 0.0676, 0.0983, 0.137, 0.2228, 0.2769, 0.2813, 0.2898, 0.2993, 0.3716, 0.3216, 0.3369, 0.3515, 0.3133, 0.2737, 0.2772, 0.3311, 0.3588, 0.4793, 0.4923, 0.5264, 0.5518, 0.509, 0.5971, 0.5629, 0.4757, 0.5665, 0.5201, 0.5598, 0.6458, 0.6446, 0.6803, 0.6439, 0.6128, 0.6798, 0.7798, 0.7886, 0.8033, 0.8417, 0.8204, 0.8184, 0.9072, 0.7718, 0.93, 0.9367, 0.9334, 1.0584, 0.9593, 0.9901, 0.9255, 1.0058, 0.9662, 0.8569, 0.8359, 0.6364, 0.6852, 0.8197, 0.8246, 0.8712, 0.8712, 0.9217, 0.9541, 0.9462, 0.9029, 0.8418, 0.9277, 1.0065, 1.0512, 1.143, 1.1003, 0.9596, 0.93, 0.8894, 0.8326, 1.0198, 1.1609, 1.2477, 1.2451]}
{"unit_id": "orig:1934943220842283082", "ticker_idx": 0, "ts": "2025-06-17T11:56:28+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom Holds #1 Market Share, Leads in 2nm Era, with FY26 Revenue Expected to Reach $31 Billion", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM bullish on AVGO and MRVL as ASIC market dominators; also flags SNPS, CDNS, ARM as EDA/IP beneficiaries of 30%+ CAGR custom AI chip growth.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "ASIC: Custom AI Chip Market Expands Rapidly, Broadcom and Marvell Continue to Dominate the Landscape\nJPM (H. Sur, 06/15/25)\n\nJPM points out that with the surge in AI training and inference workloads, the custom AI ASIC chip market is set to reach $30 billion in 2025 and will expand at a compound annual growth rate of over 30%. Broadcom and Marvell will continue to dominate this track with market shares of 55-60% and 15%, respectively.\n\nAI Acceleration Drives Custom ASIC Market Growth, Customers Advance Multi-Generational Product Roadmaps\n\nThe high-end custom ASIC chip market is expected to reach $30 billion in 2025, with an annual growth rate of 30%. Cloud service providers and hyperscale customers (such as Google, Meta, Microsoft, Amazon) are accelerating the adoption of custom AI accelerators (XPUs) to gain advantages in performance, power consumption, and cost. They are extending this to multi-generational product plans, forming a basis for long-term growth.\n\nBroadcom Holds #1 Market Share, Leads in 2nm Era, with FY26 Revenue Expected to Reach $31 Billion\n\nBroadcom has achieved breakthroughs in multiple projects, including Google's TPU v6/v7/v8, Meta's MTIA, and a 2nm project with OpenAI and ARM/Softbank. AI-related revenue is projected to be $19.5 billion in FY25 and reach $31 billion in FY26. The company's integrated capabilities in high-performance computing, networking, and advanced packaging constitute its core advantages.\n\nMarvell Continuously Expands Customer Base, Amazon and Microsoft 2nm Projects to Ramp Up from CY27\n\nMarvell is currently mass-producing its first AI ASIC projects (Trainium 2 and Axion CPU) and will mass-produce subsequent projects (Trainium 3, Maia 3nm) in 2026. The company has secured 2nm ASIC design wins from Amazon and Microsoft, which are expected to contribute to revenue from 2027 to 2028. It also has a broad presence in sub-markets like SmartNICs.\n\nRising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM\n\nThe boom in ASIC customization is driving rapid growth in demand for EDA tools and IP cores. As customer roadmaps become more complex and the number of chip design engineers surges, companies that provide design tools and core IP, such as Synopsys, Cadence, and ARM, stand to benefit. The EDA refresh cycle is accelerating, indicating strong future revenue growth prospects.\n\nCustom ASICs Are a Core Growth Engine for Broadcom and Marvell in the Coming Years\n\nOver the past decade, both companies have built leading IP and chip design capabilities through organic growth and acquisitions, creating technological moats in key areas like computing, acceleration, networking, and storage. As AI applications deepen, custom ASICs have become a core growth strategy, driving continuous increases in revenue and valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010947570774294757, "ret_signed_m1d": 0.010947570774294757, "alpha_spy_m1d": 0.0023288913415406842, "alpha_c_m1d": 0.004104775433944274, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007579068479083073, "ret_signed_p1d": 0.007579068479083073, "alpha_spy_p1d": 0.007729704233254253, "alpha_c_p1d": 0.0037563340103901055, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.06023523382586249, "ret_signed_p1w": 0.06023523382586249, "alpha_spy_p1w": 0.04175290867931025, "alpha_c_p1w": 0.022198639680289123, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.12872807270507058, "ret_signed_p1m": 0.12872807270507058, "alpha_spy_p1m": 0.08097271591478172, "alpha_c_p1m": 0.022875448211671312, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": 0.4465132750437517, "ret_signed_p3m": 0.4465132750437517, "alpha_spy_p3m": 0.34304831175269324, "alpha_c_p3m": 0.2831659798630233, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 0.6627952042762115, "ret_signed_p6m": 0.6627952042762115, "alpha_spy_p6m": 0.5055068875460447, "alpha_c_p6m": 0.23269462842022137, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [-0.2359, -0.2314, -0.2331, -0.2451, -0.2811, -0.3103, -0.3218, -0.3286, -0.3242, -0.3099, -0.3824, -0.4134, -0.3819, -0.3743, -0.2575, -0.3091, -0.2704, -0.2848, -0.2824, -0.2998, -0.3143, -0.3143, -0.3335, -0.32, -0.2906, -0.2455, -0.2288, -0.2282, -0.2334, -0.2282, -0.2087, -0.1834, -0.1951, -0.1976, -0.1787, -0.1668, -0.1651, -0.1114, -0.068, -0.0692, -0.0671, -0.0832, -0.0751, -0.0709, -0.0788, -0.0756, -0.0828, -0.0828, -0.055, -0.0399, -0.0297, -0.0293, -0.0026, 0.03, 0.047, 0.0423, -0.0098, -0.0204, -0.019, 0.0142, 0.0269, -0.0027, 0.0109, 0.0, 0.0076, 0.0076, 0.0048, 0.02, 0.0602, 0.0638, 0.086, 0.0827, 0.108, 0.0641, 0.0849, 0.1061, 0.1061, 0.1021, 0.0925, 0.117, 0.107, 0.1029, 0.1078, 0.1293, 0.1287, 0.1514, 0.1389, 0.1585, 0.1198, 0.1403, 0.1605, 0.1664, 0.183, 0.1955, 0.2164, 0.1805, 0.1602, 0.1967, 0.1774, 0.2126, 0.221, 0.2258, 0.2215, 0.2574, 0.2424, 0.251, 0.2313, 0.229, 0.1854, 0.1704, 0.1641, 0.1817, 0.1827, 0.1979, 0.2069, 0.2406, 0.1954, 0.1954, 0.1988, 0.2155, 0.2304, 0.3461, 0.3894, 0.3533, 0.4855, 0.4455, 0.4465, 0.4635, 0.447, 0.3914, 0.3881, 0.3865, 0.3641, 0.3647, 0.3662, 0.3533, 0.347, 0.3203, 0.3284, 0.3424, 0.3617, 0.3624, 0.3508, 0.3545, 0.3911, 0.3892, 0.3071, 0.4362, 0.3856, 0.4146, 0.426, 0.4066, 0.4062, 0.3797, 0.3702, 0.3863, 0.4259, 0.4578, 0.5017, 0.5541, 0.5158, 0.4883, 0.4598, 0.4171, 0.4454, 0.4318, 0.407, 0.443, 0.4171, 0.4303, 0.3689, 0.3789, 0.3797, 0.371, 0.427, 0.3964, 0.3698, 0.5218, 0.5503, 0.6008, 0.6008, 0.6225, 0.5545, 0.5364, 0.5325, 0.5342, 0.5713, 0.615, 0.6359, 0.6628]}
{"unit_id": "orig:1934943220842283082", "ticker_idx": 1, "ts": "2025-06-17T11:56:28+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell Continuously Expands Customer Base, Amazon and Microsoft 2nm Projects to Ramp Up from CY27", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM bullish on AVGO and MRVL as ASIC market dominators; also flags SNPS, CDNS, ARM as EDA/IP beneficiaries of 30%+ CAGR custom AI chip growth.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "ASIC: Custom AI Chip Market Expands Rapidly, Broadcom and Marvell Continue to Dominate the Landscape\nJPM (H. Sur, 06/15/25)\n\nJPM points out that with the surge in AI training and inference workloads, the custom AI ASIC chip market is set to reach $30 billion in 2025 and will expand at a compound annual growth rate of over 30%. Broadcom and Marvell will continue to dominate this track with market shares of 55-60% and 15%, respectively.\n\nAI Acceleration Drives Custom ASIC Market Growth, Customers Advance Multi-Generational Product Roadmaps\n\nThe high-end custom ASIC chip market is expected to reach $30 billion in 2025, with an annual growth rate of 30%. Cloud service providers and hyperscale customers (such as Google, Meta, Microsoft, Amazon) are accelerating the adoption of custom AI accelerators (XPUs) to gain advantages in performance, power consumption, and cost. They are extending this to multi-generational product plans, forming a basis for long-term growth.\n\nBroadcom Holds #1 Market Share, Leads in 2nm Era, with FY26 Revenue Expected to Reach $31 Billion\n\nBroadcom has achieved breakthroughs in multiple projects, including Google's TPU v6/v7/v8, Meta's MTIA, and a 2nm project with OpenAI and ARM/Softbank. AI-related revenue is projected to be $19.5 billion in FY25 and reach $31 billion in FY26. The company's integrated capabilities in high-performance computing, networking, and advanced packaging constitute its core advantages.\n\nMarvell Continuously Expands Customer Base, Amazon and Microsoft 2nm Projects to Ramp Up from CY27\n\nMarvell is currently mass-producing its first AI ASIC projects (Trainium 2 and Axion CPU) and will mass-produce subsequent projects (Trainium 3, Maia 3nm) in 2026. The company has secured 2nm ASIC design wins from Amazon and Microsoft, which are expected to contribute to revenue from 2027 to 2028. It also has a broad presence in sub-markets like SmartNICs.\n\nRising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM\n\nThe boom in ASIC customization is driving rapid growth in demand for EDA tools and IP cores. As customer roadmaps become more complex and the number of chip design engineers surges, companies that provide design tools and core IP, such as Synopsys, Cadence, and ARM, stand to benefit. The EDA refresh cycle is accelerating, indicating strong future revenue growth prospects.\n\nCustom ASICs Are a Core Growth Engine for Broadcom and Marvell in the Coming Years\n\nOver the past decade, both companies have built leading IP and chip design capabilities through organic growth and acquisitions, creating technological moats in key areas like computing, acceleration, networking, and storage. As AI applications deepen, custom ASICs have become a core growth strategy, driving continuous increases in revenue and valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006143734912960275, "ret_signed_m1d": 0.006143734912960275, "alpha_spy_m1d": -0.002474944519793798, "alpha_c_m1d": -0.0006990604273902079, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07086723733043909, "ret_signed_p1d": 0.07086723733043909, "alpha_spy_p1d": 0.07101787308461027, "alpha_c_p1d": 0.06704450286174612, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.07458218088434854, "ret_signed_p1w": 0.07458218088434854, "alpha_spy_p1w": 0.0560998557377963, "alpha_c_p1w": 0.03654558673877517, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.013115976126339524, "ret_signed_p1m": 0.013115976126339524, "alpha_spy_p1m": -0.03463938066394934, "alpha_c_p1m": -0.09273664836705975, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": -0.03693203461145189, "ret_signed_p3m": -0.03693203461145189, "alpha_spy_p3m": -0.14039699790251037, "alpha_c_p3m": -0.20027932979218033, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 0.3231456569819964, "ret_signed_p6m": 0.3231456569819964, "alpha_spy_p6m": 0.16585734025182952, "alpha_c_p6m": -0.10695491887399378, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [0.006, 0.0046, 0.0384, 0.0228, -0.0472, -0.0743, -0.1146, -0.1213, -0.1056, -0.0976, -0.2059, -0.2946, -0.2722, -0.286, -0.13, -0.2455, -0.2372, -0.2533, -0.2382, -0.258, -0.2613, -0.2613, -0.2945, -0.2768, -0.2317, -0.1809, -0.1582, -0.1613, -0.1616, -0.166, -0.1292, -0.1094, -0.1144, -0.1253, -0.1955, -0.1766, -0.1477, -0.0784, -0.0644, -0.0584, -0.0684, -0.089, -0.1062, -0.1224, -0.1412, -0.1163, -0.1329, -0.1329, -0.0882, -0.0772, -0.0894, -0.14, -0.1217, -0.109, -0.0527, -0.069, -0.0234, -0.0121, -0.0164, -0.025, -0.005, -0.04, 0.0061, 0.0, 0.0709, 0.0709, 0.0503, 0.0113, 0.0746, 0.0849, 0.1426, 0.1024, 0.1059, 0.0893, 0.0609, 0.0742, 0.0742, 0.0223, 0.028, 0.0324, 0.0481, 0.0397, 0.0369, 0.0354, 0.0131, 0.0297, 0.0675, 0.0447, 0.0294, 0.0477, 0.0587, 0.0612, 0.0855, 0.0916, 0.1688, 0.1492, 0.0646, 0.0943, 0.0958, 0.077, 0.0846, 0.1059, 0.1051, 0.1126, 0.1342, 0.1302, 0.0895, 0.0973, 0.0306, 0.0184, 0.0183, 0.0439, 0.0431, 0.0619, 0.0695, 0.1043, -0.101, -0.101, -0.0763, -0.109, -0.0834, -0.0944, -0.0562, -0.0442, -0.0405, -0.0478, -0.0369, -0.0358, -0.0153, 0.015, 0.0614, 0.0619, 0.08, 0.067, 0.1452, 0.1984, 0.1893, 0.1781, 0.2022, 0.1996, 0.2326, 0.2329, 0.2715, 0.2436, 0.3227, 0.2967, 0.225, 0.2791, 0.2337, 0.2719, 0.2625, 0.2585, 0.2283, 0.2057, 0.1597, 0.1843, 0.2038, 0.2693, 0.2658, 0.2899, 0.2673, 0.3413, 0.2931, 0.2533, 0.3293, 0.3355, 0.301, 0.334, 0.2782, 0.2782, 0.2523, 0.237, 0.1941, 0.1258, 0.1636, 0.0972, 0.1082, 0.1989, 0.1938, 0.2552, 0.2552, 0.2792, 0.3035, 0.3292, 0.4338, 0.405, 0.4153, 0.3164, 0.2721, 0.3231]}
{"unit_id": "orig:1934943220842283082", "ticker_idx": 2, "ts": "2025-06-17T11:56:28+00:00", "symbol": "SNPS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM bullish on AVGO and MRVL as ASIC market dominators; also flags SNPS, CDNS, ARM as EDA/IP beneficiaries of 30%+ CAGR custom AI chip growth.", "resolved_tickers": ["SNPS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "ASIC: Custom AI Chip Market Expands Rapidly, Broadcom and Marvell Continue to Dominate the Landscape\nJPM (H. Sur, 06/15/25)\n\nJPM points out that with the surge in AI training and inference workloads, the custom AI ASIC chip market is set to reach $30 billion in 2025 and will expand at a compound annual growth rate of over 30%. Broadcom and Marvell will continue to dominate this track with market shares of 55-60% and 15%, respectively.\n\nAI Acceleration Drives Custom ASIC Market Growth, Customers Advance Multi-Generational Product Roadmaps\n\nThe high-end custom ASIC chip market is expected to reach $30 billion in 2025, with an annual growth rate of 30%. Cloud service providers and hyperscale customers (such as Google, Meta, Microsoft, Amazon) are accelerating the adoption of custom AI accelerators (XPUs) to gain advantages in performance, power consumption, and cost. They are extending this to multi-generational product plans, forming a basis for long-term growth.\n\nBroadcom Holds #1 Market Share, Leads in 2nm Era, with FY26 Revenue Expected to Reach $31 Billion\n\nBroadcom has achieved breakthroughs in multiple projects, including Google's TPU v6/v7/v8, Meta's MTIA, and a 2nm project with OpenAI and ARM/Softbank. AI-related revenue is projected to be $19.5 billion in FY25 and reach $31 billion in FY26. The company's integrated capabilities in high-performance computing, networking, and advanced packaging constitute its core advantages.\n\nMarvell Continuously Expands Customer Base, Amazon and Microsoft 2nm Projects to Ramp Up from CY27\n\nMarvell is currently mass-producing its first AI ASIC projects (Trainium 2 and Axion CPU) and will mass-produce subsequent projects (Trainium 3, Maia 3nm) in 2026. The company has secured 2nm ASIC design wins from Amazon and Microsoft, which are expected to contribute to revenue from 2027 to 2028. It also has a broad presence in sub-markets like SmartNICs.\n\nRising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM\n\nThe boom in ASIC customization is driving rapid growth in demand for EDA tools and IP cores. As customer roadmaps become more complex and the number of chip design engineers surges, companies that provide design tools and core IP, such as Synopsys, Cadence, and ARM, stand to benefit. The EDA refresh cycle is accelerating, indicating strong future revenue growth prospects.\n\nCustom ASICs Are a Core Growth Engine for Broadcom and Marvell in the Coming Years\n\nOver the past decade, both companies have built leading IP and chip design capabilities through organic growth and acquisitions, creating technological moats in key areas like computing, acceleration, networking, and storage. As AI applications deepen, custom ASICs have become a core growth strategy, driving continuous increases in revenue and valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008435022753698451, "ret_signed_m1d": 0.008435022753698451, "alpha_spy_m1d": -0.00018365667905562155, "alpha_c_m1d": -0.00018365667905562155, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.008618679432754073, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0037442128778917283, "ret_signed_p1d": -0.0037442128778917283, "alpha_spy_p1d": -0.0035935771237205483, "alpha_c_p1d": -0.0035935771237205483, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": -0.00015063575417118003, "ret_p1w": 0.0075305366760758385, "ret_signed_p1w": 0.0075305366760758385, "alpha_spy_p1w": -0.010951788470476398, "alpha_c_p1w": -0.010951788470476398, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.018482325146552236, "ret_p1m": 0.20151455519666417, "ret_signed_p1m": 0.20151455519666417, "alpha_spy_p1m": 0.1537591984063753, "alpha_c_p1m": 0.1537591984063753, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.04775535679028886, "ret_p3m": -0.10506937806214922, "ret_signed_p3m": -0.10506937806214922, "alpha_spy_p3m": -0.2085343413532077, "alpha_c_p3m": -0.2085343413532077, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.10346496329105848, "ret_p6m": 0.0009044860776226127, "ret_signed_p6m": 0.0009044860776226127, "alpha_spy_p6m": -0.15638383065254424, "alpha_c_p6m": -0.15638383065254424, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.15728831673016685, "price_path": [-0.0534, -0.0564, -0.0371, -0.0382, -0.0536, -0.0614, -0.0788, -0.0979, -0.083, -0.0775, -0.1212, -0.1836, -0.1988, -0.1973, -0.0982, -0.1441, -0.114, -0.1124, -0.1046, -0.1257, -0.1306, -0.1306, -0.1536, -0.1304, -0.1118, -0.0768, -0.0605, -0.0661, -0.0308, -0.0345, -0.0337, -0.0066, 0.0042, -0.0039, 0.0108, 0.0191, 0.0158, 0.0635, 0.0848, 0.0889, 0.0822, 0.0821, 0.0854, 0.0846, 0.0548, 0.0591, 0.0493, 0.0493, 0.0765, -0.0273, -0.0429, -0.024, -0.0195, -0.0192, -0.0046, 0.0143, 0.0223, 0.0428, 0.061, 0.0533, 0.0412, 0.0073, 0.0084, 0.0, -0.0037, -0.0037, -0.0102, -0.0093, 0.0075, 0.025, 0.0427, 0.0573, 0.0784, 0.0976, 0.1004, 0.1543, 0.1543, 0.1286, 0.1601, 0.1589, 0.191, 0.1764, 0.1559, 0.1971, 0.2015, 0.239, 0.23, 0.2441, 0.2939, 0.2924, 0.2825, 0.2654, 0.2466, 0.3374, 0.3575, 0.3325, 0.3013, 0.3379, 0.322, 0.3165, 0.3049, 0.3029, 0.296, 0.3164, 0.3012, 0.2965, 0.2998, 0.3154, 0.289, 0.2645, 0.2582, 0.2758, 0.2558, 0.2537, 0.2688, 0.2877, 0.2695, 0.2695, 0.2453, 0.2455, 0.2662, 0.2582, 0.2812, 0.2713, -0.1843, -0.0785, -0.1051, -0.1182, -0.104, -0.1052, 0.0099, 0.0423, 0.0829, 0.0314, -0.0154, 0.0248, 0.026, 0.0131, 0.0378, 0.0281, -0.0089, -0.0131, 0.0059, 0.0051, 0.0294, 0.019, -0.0767, -0.0576, -0.0666, -0.0831, -0.074, -0.0584, -0.0464, -0.0344, -0.0444, -0.0406, -0.0236, -0.021, -0.0391, -0.0422, -0.0683, -0.0454, -0.0624, -0.1242, -0.1394, -0.1667, -0.1724, -0.157, -0.1679, -0.1611, -0.1714, -0.18, -0.1791, -0.1926, -0.1874, -0.1889, -0.1831, -0.1489, -0.1552, -0.1382, -0.1382, -0.1207, -0.0781, -0.0548, -0.0188, -0.0245, -0.0182, -0.0203, -0.0201, 0.0009]}
{"unit_id": "orig:1934943220842283082", "ticker_idx": 3, "ts": "2025-06-17T11:56:28+00:00", "symbol": "CDNS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM bullish on AVGO and MRVL as ASIC market dominators; also flags SNPS, CDNS, ARM as EDA/IP beneficiaries of 30%+ CAGR custom AI chip growth.", "resolved_tickers": ["CDNS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Application", "unit_kind": "original", "lang": "en", "tweet_text": "ASIC: Custom AI Chip Market Expands Rapidly, Broadcom and Marvell Continue to Dominate the Landscape\nJPM (H. Sur, 06/15/25)\n\nJPM points out that with the surge in AI training and inference workloads, the custom AI ASIC chip market is set to reach $30 billion in 2025 and will expand at a compound annual growth rate of over 30%. Broadcom and Marvell will continue to dominate this track with market shares of 55-60% and 15%, respectively.\n\nAI Acceleration Drives Custom ASIC Market Growth, Customers Advance Multi-Generational Product Roadmaps\n\nThe high-end custom ASIC chip market is expected to reach $30 billion in 2025, with an annual growth rate of 30%. Cloud service providers and hyperscale customers (such as Google, Meta, Microsoft, Amazon) are accelerating the adoption of custom AI accelerators (XPUs) to gain advantages in performance, power consumption, and cost. They are extending this to multi-generational product plans, forming a basis for long-term growth.\n\nBroadcom Holds #1 Market Share, Leads in 2nm Era, with FY26 Revenue Expected to Reach $31 Billion\n\nBroadcom has achieved breakthroughs in multiple projects, including Google's TPU v6/v7/v8, Meta's MTIA, and a 2nm project with OpenAI and ARM/Softbank. AI-related revenue is projected to be $19.5 billion in FY25 and reach $31 billion in FY26. The company's integrated capabilities in high-performance computing, networking, and advanced packaging constitute its core advantages.\n\nMarvell Continuously Expands Customer Base, Amazon and Microsoft 2nm Projects to Ramp Up from CY27\n\nMarvell is currently mass-producing its first AI ASIC projects (Trainium 2 and Axion CPU) and will mass-produce subsequent projects (Trainium 3, Maia 3nm) in 2026. The company has secured 2nm ASIC design wins from Amazon and Microsoft, which are expected to contribute to revenue from 2027 to 2028. It also has a broad presence in sub-markets like SmartNICs.\n\nRising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM\n\nThe boom in ASIC customization is driving rapid growth in demand for EDA tools and IP cores. As customer roadmaps become more complex and the number of chip design engineers surges, companies that provide design tools and core IP, such as Synopsys, Cadence, and ARM, stand to benefit. The EDA refresh cycle is accelerating, indicating strong future revenue growth prospects.\n\nCustom ASICs Are a Core Growth Engine for Broadcom and Marvell in the Coming Years\n\nOver the past decade, both companies have built leading IP and chip design capabilities through organic growth and acquisitions, creating technological moats in key areas like computing, acceleration, networking, and storage. As AI applications deepen, custom ASICs have become a core growth strategy, driving continuous increases in revenue and valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006053503540447425, "ret_signed_m1d": 0.006053503540447425, "alpha_spy_m1d": -0.002565175892306648, "alpha_c_m1d": -0.002565175892306648, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.008618679432754073, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007224092515416003, "ret_signed_p1d": -0.007224092515416003, "alpha_spy_p1d": -0.007073456761244823, "alpha_c_p1d": -0.007073456761244823, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": -0.00015063575417118003, "ret_p1w": -0.00735790035796402, "ret_signed_p1w": -0.00735790035796402, "alpha_spy_p1w": -0.025840225504516257, "alpha_c_p1w": -0.025840225504516257, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.018482325146552236, "ret_p1m": 0.05210697850256052, "ret_signed_p1m": 0.05210697850256052, "alpha_spy_p1m": 0.004351621712271658, "alpha_c_p1m": 0.004351621712271658, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.04775535679028886, "ret_p3m": 0.1487625785495923, "ret_signed_p3m": 0.1487625785495923, "alpha_spy_p3m": 0.04529761525853382, "alpha_c_p3m": 0.04529761525853382, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.10346496329105848, "ret_p6m": 0.1306354433397785, "ret_signed_p6m": 0.1306354433397785, "alpha_spy_p6m": -0.026652873390388354, "alpha_c_p6m": -0.026652873390388354, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.15728831673016685, "price_path": [-0.1355, -0.1223, -0.1065, -0.1, -0.119, -0.1219, -0.1415, -0.1494, -0.1345, -0.1139, -0.1675, -0.2211, -0.218, -0.2253, -0.1149, -0.1504, -0.1322, -0.1286, -0.116, -0.1329, -0.1297, -0.1297, -0.1587, -0.1366, -0.1123, -0.0576, -0.0313, -0.0443, 0.0108, -0.0042, 0.0024, 0.0306, 0.0329, 0.0227, 0.0312, 0.03, 0.0227, 0.0427, 0.0658, 0.0682, 0.0661, 0.0755, 0.0712, 0.0721, 0.062, 0.059, 0.0552, 0.0552, 0.0805, -0.0347, -0.0485, -0.0399, -0.0216, -0.019, -0.0154, -0.0091, -0.0069, 0.0089, 0.0224, 0.0274, 0.0296, 0.0033, 0.0061, 0.0, -0.0072, -0.0072, -0.012, -0.0182, -0.0074, -0.0049, 0.0157, 0.0207, 0.0306, 0.035, 0.04, 0.093, 0.093, 0.0712, 0.0812, 0.08, 0.0791, 0.0722, 0.0618, 0.0637, 0.0521, 0.0731, 0.0554, 0.0579, 0.0689, 0.0918, 0.0858, 0.111, 0.1163, 0.2249, 0.2409, 0.2193, 0.1939, 0.2203, 0.2057, 0.2045, 0.1851, 0.1775, 0.1675, 0.1826, 0.1676, 0.167, 0.1702, 0.1915, 0.1601, 0.1554, 0.1624, 0.1693, 0.1545, 0.1506, 0.1601, 0.1849, 0.172, 0.172, 0.1465, 0.1616, 0.1683, 0.1739, 0.2062, 0.2099, 0.1322, 0.1863, 0.1488, 0.1757, 0.1673, 0.1614, 0.2203, 0.2487, 0.2487, 0.2251, 0.1938, 0.1741, 0.1709, 0.1656, 0.1748, 0.1772, 0.1613, 0.1614, 0.1818, 0.1555, 0.1706, 0.1661, 0.0936, 0.1111, 0.0895, 0.0839, 0.0844, 0.0907, 0.1025, 0.1152, 0.1054, 0.128, 0.1542, 0.1753, 0.1415, 0.1419, 0.124, 0.1327, 0.1218, 0.1144, 0.0958, 0.0851, 0.0871, 0.1001, 0.0653, 0.0553, 0.0568, 0.0533, 0.0411, 0.0141, 0.026, 0.0105, 0.0053, 0.0183, 0.0156, 0.0246, 0.0246, 0.0429, 0.0355, 0.0633, 0.1241, 0.1281, 0.1289, 0.1281, 0.1206, 0.1306]}
{"unit_id": "orig:1934943220842283082", "ticker_idx": 4, "ts": "2025-06-17T11:56:28+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM bullish on AVGO and MRVL as ASIC market dominators; also flags SNPS, CDNS, ARM as EDA/IP beneficiaries of 30%+ CAGR custom AI chip growth.", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "ASIC: Custom AI Chip Market Expands Rapidly, Broadcom and Marvell Continue to Dominate the Landscape\nJPM (H. Sur, 06/15/25)\n\nJPM points out that with the surge in AI training and inference workloads, the custom AI ASIC chip market is set to reach $30 billion in 2025 and will expand at a compound annual growth rate of over 30%. Broadcom and Marvell will continue to dominate this track with market shares of 55-60% and 15%, respectively.\n\nAI Acceleration Drives Custom ASIC Market Growth, Customers Advance Multi-Generational Product Roadmaps\n\nThe high-end custom ASIC chip market is expected to reach $30 billion in 2025, with an annual growth rate of 30%. Cloud service providers and hyperscale customers (such as Google, Meta, Microsoft, Amazon) are accelerating the adoption of custom AI accelerators (XPUs) to gain advantages in performance, power consumption, and cost. They are extending this to multi-generational product plans, forming a basis for long-term growth.\n\nBroadcom Holds #1 Market Share, Leads in 2nm Era, with FY26 Revenue Expected to Reach $31 Billion\n\nBroadcom has achieved breakthroughs in multiple projects, including Google's TPU v6/v7/v8, Meta's MTIA, and a 2nm project with OpenAI and ARM/Softbank. AI-related revenue is projected to be $19.5 billion in FY25 and reach $31 billion in FY26. The company's integrated capabilities in high-performance computing, networking, and advanced packaging constitute its core advantages.\n\nMarvell Continuously Expands Customer Base, Amazon and Microsoft 2nm Projects to Ramp Up from CY27\n\nMarvell is currently mass-producing its first AI ASIC projects (Trainium 2 and Axion CPU) and will mass-produce subsequent projects (Trainium 3, Maia 3nm) in 2026. The company has secured 2nm ASIC design wins from Amazon and Microsoft, which are expected to contribute to revenue from 2027 to 2028. It also has a broad presence in sub-markets like SmartNICs.\n\nRising Demand for IP and EDA Benefits Synopsys, Cadence, and ARM\n\nThe boom in ASIC customization is driving rapid growth in demand for EDA tools and IP cores. As customer roadmaps become more complex and the number of chip design engineers surges, companies that provide design tools and core IP, such as Synopsys, Cadence, and ARM, stand to benefit. The EDA refresh cycle is accelerating, indicating strong future revenue growth prospects.\n\nCustom ASICs Are a Core Growth Engine for Broadcom and Marvell in the Coming Years\n\nOver the past decade, both companies have built leading IP and chip design capabilities through organic growth and acquisitions, creating technological moats in key areas like computing, acceleration, networking, and storage. As AI applications deepen, custom ASICs have become a core growth strategy, driving continuous increases in revenue and valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01851857318936312, "ret_signed_m1d": -0.01851857318936312, "alpha_spy_m1d": -0.02713725262211719, "alpha_c_m1d": -0.0253613685297136, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00919017288444035, "ret_signed_p1d": 0.00919017288444035, "alpha_spy_p1d": 0.00934080863861153, "alpha_c_p1d": 0.005367438415747383, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.0807766883830976, "ret_signed_p1w": 0.0807766883830976, "alpha_spy_p1w": 0.06229436323654536, "alpha_c_p1w": 0.04274009423752423, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.06343278467626212, "ret_signed_p1m": 0.06343278467626212, "alpha_spy_p1m": 0.01567742788597326, "alpha_c_p1m": -0.04241983981713715, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": 0.04090656522257152, "ret_signed_p3m": 0.04090656522257152, "alpha_spy_p3m": -0.06255839806848695, "alpha_c_p3m": -0.12244072995815691, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": -0.022111642628872552, "ret_signed_p6m": -0.022111642628872552, "alpha_spy_p6m": -0.1793999593590394, "alpha_c_p6m": -0.4522122184848627, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [-0.1772, -0.1772, -0.1373, -0.1412, -0.2058, -0.2228, -0.2551, -0.2621, -0.2608, -0.2528, -0.3248, -0.3939, -0.3876, -0.407, -0.2635, -0.3058, -0.2814, -0.274, -0.2828, -0.3021, -0.304, -0.304, -0.3309, -0.3052, -0.2734, -0.226, -0.2168, -0.2255, -0.2285, -0.2119, -0.2026, -0.1482, -0.1573, -0.154, -0.1419, -0.1949, -0.1998, -0.1376, -0.1247, -0.0779, -0.0805, -0.0605, -0.0875, -0.0945, -0.0986, -0.1067, -0.1212, -0.1212, -0.0744, -0.0634, -0.1148, -0.1394, -0.129, -0.1101, -0.0992, -0.1048, -0.0802, -0.0422, -0.0283, -0.03, -0.0422, -0.0634, -0.0185, 0.0, 0.0092, 0.0092, 0.0022, 0.0319, 0.0808, 0.087, 0.0928, 0.1433, 0.1176, 0.0802, 0.0685, 0.0717, 0.0717, 0.0149, 0.0212, 0.0228, 0.0265, 0.0084, -0.0012, 0.0165, 0.0634, 0.0861, 0.0831, 0.1189, 0.0814, 0.1006, 0.1055, 0.1275, 0.1358, 0.1296, 0.1286, -0.0231, -0.0493, -0.0323, -0.0518, -0.0594, -0.0632, -0.043, -0.0254, -0.0161, -0.0216, -0.0288, -0.0401, -0.0253, -0.074, -0.0937, -0.079, -0.047, -0.048, -0.0308, -0.0281, -0.015, -0.0443, -0.0443, -0.0855, -0.0919, -0.0638, -0.0453, -0.0386, -0.0271, 0.0651, 0.069, 0.0409, 0.0632, 0.0631, 0.0598, 0.0126, -0.0125, -0.0017, -0.0258, -0.0029, -0.0281, -0.0352, -0.034, -0.0223, 0.0391, 0.0513, 0.0547, 0.0795, 0.1011, 0.1524, 0.1792, 0.0697, 0.1881, 0.162, 0.1793, 0.1829, 0.1443, 0.185, 0.1704, 0.145, 0.1512, 0.1794, 0.2342, 0.196, 0.1774, 0.1432, 0.1734, 0.1656, 0.1106, 0.1069, 0.0935, 0.0529, 0.0699, 0.0347, 0.0278, -0.0305, -0.0342, -0.0308, -0.06, -0.0534, -0.0842, -0.0909, -0.0692, -0.0918, -0.0837, -0.0837, -0.0633, -0.0671, -0.0569, -0.0382, -0.0292, -0.0236, -0.0341, -0.0193, -0.0221]}
{"unit_id": "quote:1933413076101181845", "ticker_idx": 0, "ts": "2025-06-13T06:36:13+00:00", "symbol": "Sony", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Uncertainty is bound to grow... Apple, Sony's biggest customer, is exploring a switch to Samsung. This poses a far greater risk than any of the previous ones listed above.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Sony semiconductor business: Chinese clients shifting to smaller sensors, Chinese CMOS rivals gaining share, and Apple potentially switching to Samsung sensors.", "resolved_tickers": ["SONY"], "resolution_method": "exchange_ticker", "resolution_reason": "Sony Group Corp listed on NYSE as SONY (ADR)", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Of course, the uncertainty is bound to grow.\n\n1. Key Chinese clients are shifting back to smaller sensors from 1-inch sensors.\n→ This naturally leads to lower unit prices and shrinking margins.\n\n2. Rapid rise of domestic Chinese sensor makers.\n→ Chinese CMOS players like OmniVision and Huawei’s in-house sensors have made significant progress, and more Chinese smartphones are adopting them.\n\n3. Apple, Sony’s biggest customer, is exploring a switch to Samsung.\n→ Apple, the largest player in the smartphone market and Sony’s most important source of semiconductor revenue, is reportedly considering adopting Samsung’s CMOS sensors.\n→ This poses a far greater risk than any of the previous ones listed above.", "reply_text": null, "parent_text": null, "quoted_text": "Sony: Uncertainty is increasing regarding the outlook for its semiconductor business.", "parent_author": null, "ret_m1d": 0.01812569640659345, "ret_signed_m1d": -0.01812569640659345, "alpha_spy_m1d": -0.006819224930931345, "alpha_c_m1d": -0.003909882780070539, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.01421581362652291, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006170452970653217, "ret_signed_p1d": -0.006170452970653217, "alpha_spy_p1d": 0.0033437435080621025, "alpha_c_p1d": 0.010010432728509988, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.016180885699163206, "ret_p1w": -0.04550714582676829, "ret_signed_p1w": 0.04550714582676829, "alpha_spy_p1w": 0.04389387083095897, "alpha_c_p1w": 0.0512352160658075, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.005728070239039207, "ret_p1m": -0.05900503928417278, "ret_signed_p1m": 0.05900503928417278, "alpha_spy_p1m": 0.10868183104988316, "alpha_c_p1m": 0.1296396466890668, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.07063460740489402, "ret_p3m": 0.10489777405870537, "ret_signed_p3m": -0.10489777405870537, "alpha_spy_p3m": -0.009189027638834668, "alpha_c_p3m": 0.02650546487746608, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.13140323893617145, "ret_p6m": 0.06247585471726369, "ret_signed_p6m": -0.06247585471726369, "alpha_spy_p6m": 0.08920239696873322, "alpha_c_p6m": 0.17567867658455838, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.23815453130182207, "price_path": [-0.057, -0.0397, -0.0428, -0.0297, -0.0282, -0.0286, -0.0193, -0.0139, -0.0351, -0.0208, -0.0293, -0.042, -0.0983, -0.1585, -0.1809, -0.1778, -0.1142, -0.1161, -0.1153, -0.096, -0.0868, -0.0991, -0.074, -0.074, -0.0841, -0.0617, -0.0401, -0.0312, -0.037, -0.0251, -0.0069, 0.0027, -0.0251, -0.0231, -0.0266, -0.027, -0.0351, -0.0571, -0.0474, -0.0459, -0.0536, -0.0405, -0.0409, -0.0482, -0.0343, -0.005, -0.0401, -0.0231, -0.0154, -0.0154, 0.0243, 0.0143, 0.0266, 0.0174, 0.0366, 0.0324, 0.0285, 0.0154, 0.0282, 0.0255, 0.0104, -0.0008, 0.0181, 0.0, 0.0062, -0.0104, -0.0019, -0.0019, -0.0455, -0.0582, -0.0363, -0.039, -0.015, 0.0143, 0.0039, -0.0177, -0.0085, -0.0135, -0.0135, -0.0243, -0.0262, -0.0293, -0.0494, -0.0578, -0.059, -0.0713, -0.0752, -0.0586, -0.0775, -0.0717, -0.071, -0.0258, -0.0212, -0.0405, -0.0605, -0.0686, -0.0667, -0.0621, -0.0602, -0.0451, -0.054, -0.039, 0.0039, 0.047, 0.0521, 0.0486, 0.0791, 0.0698, 0.1084, 0.0933, 0.0891, 0.0845, 0.0656, 0.1034, 0.0845, 0.0752, 0.0737, 0.0798, 0.0613, 0.0613, 0.0393, 0.0424, 0.0737, 0.0632, 0.1057, 0.1072, 0.1049, 0.1257, 0.1265, 0.1315, 0.1115, 0.1246, 0.1597, 0.1323, 0.1465, 0.1477, 0.1315, 0.1504, 0.1485, 0.1223, 0.1103, 0.1084, 0.0829, 0.1169, 0.1608, 0.1469, 0.1473, 0.1508, 0.0895, 0.1134, 0.1118, 0.1142, 0.0991, 0.1145, 0.1257, 0.1238, 0.1088, 0.1072, 0.1049, 0.1068, 0.118, 0.0868, 0.0702, 0.0756, 0.0779, 0.0764, 0.0829, 0.091, 0.0686, 0.0787, 0.1246, 0.167, 0.1354, 0.1616, 0.1219, 0.1057, 0.101, 0.0702, 0.1003, 0.1188, 0.103, 0.1215, 0.1215, 0.1319, 0.1099, 0.1018, 0.0976, 0.091, 0.081, 0.0625]}
{"unit_id": "orig:1935243346680889672", "ticker_idx": 0, "ts": "2025-06-18T07:49:03+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I'm not sure if Morgan Stanley really knows what they're talking about when it comes to Intel. From what I've heard, the yield for 18A has already surpassed 50%.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Implicitly bullish Intel; disputes Morgan Stanley's bearish view, citing 18A yield already above 50%.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I’m not sure if Morgan Stanley really knows what they’re talking about when it comes to Intel. From what I’ve heard, the yield for 18A has already surpassed 50%. https://t.co/0STpfAH2Xv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03210798238280832, "ret_signed_m1d": -0.03210798238280832, "alpha_spy_m1d": -0.03225864083152841, "alpha_c_m1d": -0.028299805562907565, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.03303866911974129, "ret_signed_p1w": 0.03303866911974129, "alpha_spy_p1w": 0.013832093550114388, "alpha_c_p1w": -0.014982993164712521, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": 0.060958561186351234, "ret_signed_p1m": 0.060958561186351234, "alpha_spy_p1m": 0.0066324472309045035, "alpha_c_p1m": -0.04993672292770146, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": 0.15262916340528387, "ret_signed_p3m": 0.15262916340528387, "alpha_spy_p3m": 0.04312237762682147, "alpha_c_p3m": -0.0171413084304588, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.8385294901041978, "ret_signed_p6m": 0.8385294901041978, "alpha_spy_p6m": 0.6783734625107618, "alpha_c_p6m": 0.4260993706035019, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [0.1289, 0.127, 0.1261, 0.0898, 0.0991, 0.0568, 0.0568, 0.0261, 0.0228, 0.0437, -0.0763, -0.0893, -0.1564, 0.0019, -0.0749, -0.0814, -0.0549, -0.0763, -0.1052, -0.1191, -0.1191, -0.1233, -0.0921, -0.0419, 0.0, -0.067, -0.0456, -0.0535, -0.0647, -0.0703, -0.0405, -0.0568, -0.0721, -0.0549, -0.0228, -0.0033, 0.0321, 0.0498, 0.0014, 0.0028, 0.0079, -0.0051, -0.0102, -0.0372, -0.0437, -0.067, -0.067, -0.0437, -0.0521, -0.0577, -0.0903, -0.0814, -0.0558, -0.0577, -0.0698, -0.0665, -0.047, 0.0275, -0.0377, -0.0335, -0.0628, -0.0349, -0.0321, 0.0, 0.0, -0.0191, -0.014, 0.0493, 0.033, 0.047, 0.0558, 0.0423, 0.0633, 0.0181, 0.0465, 0.0465, 0.0237, 0.0977, 0.0907, 0.1084, 0.0903, 0.0842, 0.0665, 0.0558, 0.061, 0.0749, 0.0824, 0.0814, 0.0931, 0.053, -0.0368, -0.0377, -0.0503, -0.0535, -0.0786, -0.1014, -0.0926, -0.0605, -0.0503, -0.08, -0.0717, -0.0391, 0.0149, 0.034, 0.1103, 0.1429, 0.101, 0.1778, 0.0954, 0.0935, 0.154, 0.1424, 0.1331, 0.1564, 0.1601, 0.1331, 0.1331, 0.1266, 0.1168, 0.1452, 0.1396, 0.1391, 0.1373, 0.1526, 0.1452, 0.1205, 0.1526, 0.1759, 0.1587, 0.4225, 0.3765, 0.3383, 0.3653, 0.4528, 0.5817, 0.6519, 0.6045, 0.5612, 0.6724, 0.7357, 0.7138, 0.7027, 0.7296, 0.7417, 0.759, 0.6924, 0.732, 0.658, 0.7287, 0.7143, 0.7222, 0.7729, 0.7738, 0.718, 0.7757, 0.7813, 0.8399, 0.9325, 0.9237, 0.8688, 0.8609, 0.8381, 0.7231, 0.7859, 0.7329, 0.7743, 0.7892, 0.7627, 0.7631, 0.671, 0.6529, 0.6152, 0.5975, 0.6338, 0.5644, 0.6054, 0.6654, 0.6673, 0.7129, 0.7129, 0.8874, 0.8618, 1.0228, 1.0363, 0.8846, 0.9269, 0.8753, 0.8846, 0.8976, 0.8385]}
{"unit_id": "quote:1938185727613002108", "ticker_idx": 0, "ts": "2025-06-26T10:41:01+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm hoping that SK Hynix will announce around July that all of its 2026 HBM supply has been sold out. NVIDIA can't keep delaying forever either.", "tweet_type": "prediction", "timeline": "months", "summary": "Bullish SK Hynix; expects 2026 HBM supply to be fully sold out by ~July as NVIDIA can't delay contracts indefinitely.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Of course, I’m hoping that SK Hynix will announce around July that all of its 2026 HBM supply has been sold out.\nNVIDIA can’t keep delaying forever either.", "reply_text": null, "parent_text": null, "quoted_text": "The reason why Micron CEO SJ did not announce HBM sold-out for 2026 on the conference call is simple: Nvidia is delaying the contract. Nvidia is hoping that Samsung will pass HBM3E validation, allowing them to buy Micron's HBM at a lower price. That's why they're dragging out the", "parent_author": null, "ret_m1d": -0.023890800700825943, "ret_signed_m1d": -0.023890800700825943, "alpha_spy_m1d": -0.016127757170079193, "alpha_c_m1d": -0.01667576177365171, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.030716774358483856, "ret_signed_p1d": -0.030716774358483856, "alpha_spy_p1d": -0.035685106122489096, "alpha_c_p1d": -0.035117933747531915, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.04948804126557793, "ret_signed_p1w": -0.04948804126557793, "alpha_spy_p1w": -0.07150259577759732, "alpha_c_p1w": -0.07264821351741257, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": -0.09215032307545168, "ret_signed_p1m": -0.09215032307545168, "alpha_spy_p1m": -0.13338461973479476, "alpha_c_p1m": -0.12927144601217, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.2338615160807298, "ret_signed_p3m": 0.2338615160807298, "alpha_spy_p3m": 0.1469496117212168, "alpha_c_p3m": 0.07332759016880863, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.8709296139877303, "ret_signed_p6m": 0.8709296139877303, "alpha_spy_p6m": 0.7522384134919835, "alpha_c_p6m": 0.5858286202288445, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3503, -0.3289, -0.3258, -0.337, -0.3793, -0.4386, -0.4225, -0.4379, -0.3759, -0.384, -0.3861, -0.3847, -0.4072, -0.4038, -0.4038, -0.3984, -0.4079, -0.3834, -0.3926, -0.3718, -0.38, -0.384, -0.3953, -0.3953, -0.3663, -0.3663, -0.3663, -0.35, -0.3517, -0.3524, -0.3357, -0.3237, -0.2982, -0.3169, -0.3033, -0.3207, -0.3118, -0.3169, -0.3292, -0.3186, -0.3084, -0.3101, -0.2914, -0.2765, -0.302, -0.2918, -0.2918, -0.2577, -0.2338, -0.2338, -0.2184, -0.2133, -0.1809, -0.1962, -0.1962, -0.1536, -0.1502, -0.1587, -0.1604, -0.1229, -0.1143, -0.0495, -0.0239, 0.0, -0.0307, -0.0034, -0.0256, -0.0478, -0.0495, -0.0768, -0.0751, -0.0375, -0.041, 0.0137, 0.0051, 0.0239, 0.0188, 0.0102, -0.0802, -0.0819, -0.07, -0.0836, -0.0819, -0.0802, -0.0922, -0.1058, -0.1041, -0.1007, -0.0666, -0.1195, -0.1195, -0.1007, -0.1177, -0.1058, -0.1246, -0.0887, -0.0819, -0.0512, -0.0563, -0.0563, -0.087, -0.1024, -0.128, -0.1638, -0.1433, -0.1143, -0.1075, -0.1126, -0.0823, -0.0806, -0.125, -0.1096, -0.1028, -0.0925, -0.0652, -0.0532, -0.0156, 0.039, 0.0493, 0.1228, 0.1313, 0.1894, 0.1399, 0.2151, 0.2151, 0.1997, 0.2339, 0.2219, 0.2185, 0.1501, 0.1928, 0.1877, 0.2304, 0.3518, 0.3518, 0.3518, 0.3518, 0.3518, 0.3518, 0.4629, 0.4184, 0.4065, 0.4441, 0.5466, 0.591, 0.6594, 0.6372, 0.6457, 0.6355, 0.7431, 0.8286, 0.7807, 0.9072, 0.9414, 0.9106, 1.1191, 1.0029, 0.979, 1.0268, 0.9824, 1.0712, 1.1157, 1.1088, 1.0918, 0.914, 1.0712, 0.9482, 0.9209, 0.9516, 0.7807, 0.7773, 0.7739, 0.791, 0.8607, 0.8128, 0.8401, 0.9086, 0.888, 0.8538, 0.8607, 0.9735, 0.9359, 1.0077, 0.9325, 0.953, 0.8949, 0.8128, 0.8846, 0.888, 0.8709]}
{"unit_id": "quote:1928438243307262201", "ticker_idx": 0, "ts": "2025-05-30T13:08:00+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Lowered its shipment forecast... The ASIC segment has also been revised down by 20%... weak execution has raised concerns", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish MediaTek: shipment forecast cut across smartphones (Qualcomm pricing, Xiaomi N3 shift) and ASIC (down 20%), raising execution concerns.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• 2454 MediaTek\n→ Lowered its shipment forecast.\n→ In the smartphone segment, the reduction is driven by Qualcomm’s price cuts and Xiaomi’s adoption of the N3 process.\n→ The ASIC segment has also been revised down by 20%.\n→ This segment was originally expected to contribute one-third of the group’s revenue, but weak execution has raised concerns.", "reply_text": null, "parent_text": null, "quoted_text": "5/27 ：\n。6498久禾光: 微型鏡頭廠商，產品以NB窄邊框的鏡頭為主，佔公司30%營收以上，今年還會導入AIOT & 人形機器人視覺鏡頭，公司派認為今年會逐季成長，去年EPS5.6,今年6-7.\n。8299群聯: 本月56-58億，下月近60億，Q2會有匯損.\n。組裝廠目前財務開始回報匯損大概7-10%，5月業績公布前要留意.", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.0011196043038306236, "alpha_c_m1d": 0.01806050327011488, "bench_spy_m1d": 0.0011196043038306236, "bench_c_m1d": 0.01806050327011488, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.005632901136813695, "alpha_c_p1d": 0.014807147079420213, "bench_spy_p1d": 0.005632901136813695, "bench_c_p1d": 0.014807147079420213, "ret_p1w": 0.015873062228522583, "ret_signed_p1w": -0.015873062228522583, "alpha_spy_p1w": 0.0006693034663574071, "alpha_c_p1w": 0.03826669450682618, "bench_spy_p1w": 0.01654236569487999, "bench_c_p1w": 0.054139756735348765, "ret_p1m": -0.007936531114261292, "ret_signed_p1m": 0.007936531114261292, "alpha_spy_p1m": 0.059322692159845425, "alpha_c_p1m": 0.1711482159485166, "bench_spy_p1m": 0.05138616104558413, "bench_c_p1m": 0.1632116848342553, "ret_p3m": 0.13298359733889398, "ret_signed_p3m": -0.13298359733889398, "alpha_spy_p3m": -0.03262299655895218, "alpha_c_p3m": 0.10855966526298522, "bench_spy_p3m": 0.1003606007799418, "bench_c_p3m": 0.2415432626018792, "ret_p6m": -0.06933495613412755, "ret_signed_p6m": 0.06933495613412755, "alpha_spy_p6m": 0.21045767970226426, "alpha_c_p6m": 0.48380834675259454, "bench_spy_p6m": 0.1411227235681367, "bench_c_p6m": 0.414473390618467, "price_path": [0.1944, 0.1944, 0.1865, 0.1627, 0.1071, 0.0992, 0.119, 0.1032, 0.1111, 0.131, 0.1706, 0.1627, 0.1786, 0.1746, 0.1746, 0.2103, 0.2024, 0.1786, 0.1627, 0.1032, 0.1587, 0.1389, 0.1389, 0.1389, 0.0278, 0.0, -0.0595, 0.0317, 0.0992, 0.1111, 0.0992, 0.0833, 0.0635, 0.0833, 0.0675, 0.0317, 0.0873, 0.0556, 0.0952, 0.0714, 0.0873, 0.0714, 0.0714, 0.0317, 0.0278, 0.0159, 0.0119, 0.0238, 0.0516, 0.0437, 0.0754, 0.0794, 0.0873, 0.0833, 0.0397, 0.0437, 0.0635, 0.0516, 0.0437, 0.0159, 0.0159, 0.0159, 0.0, 0.0, 0.0, 0.0119, 0.0119, -0.0119, 0.0159, 0.0198, 0.0595, 0.0437, 0.0198, -0.0079, -0.004, 0.0119, 0.0198, 0.004, -0.0079, -0.004, 0.0079, 0.0278, 0.0238, 0.0198, -0.0079, 0.0119, 0.0278, 0.0278, 0.044, 0.0359, 0.0278, 0.0966, 0.133, 0.1492, 0.1208, 0.1289, 0.1411, 0.1249, 0.1411, 0.1532, 0.1573, 0.1573, 0.1573, 0.1573, 0.133, 0.1006, 0.1168, 0.1087, 0.1087, 0.0763, 0.0844, 0.0682, 0.0966, 0.0925, 0.0966, 0.1006, 0.1087, 0.1249, 0.1087, 0.1411, 0.1249, 0.0966, 0.1047, 0.1047, 0.1411, 0.133, 0.133, 0.1208, 0.1087, 0.1006, 0.1006, 0.1289, 0.1208, 0.1613, 0.1573, 0.222, 0.2058, 0.1977, 0.2018, 0.2018, 0.2422, 0.222, 0.222, 0.1654, 0.137, 0.1128, 0.0763, 0.0885, 0.0601, 0.0601, 0.0642, 0.0399, 0.0359, 0.0682, 0.0682, 0.0723, 0.0804, 0.0885, 0.0885, 0.0642, 0.0561, 0.0642, 0.0763, 0.0763, 0.0844, 0.0844, 0.0763, 0.048, 0.048, 0.0804, 0.0642, 0.0521, 0.0601, 0.0601, 0.0521, 0.0642, 0.0763, 0.044, 0.0197, 0.0116, 0.0035, 0.0075, 0.0075, -0.0046, -0.0046, -0.0531, -0.0612, -0.041, -0.0734, -0.0693]}
{"unit_id": "orig:1927140222598345058", "ticker_idx": 0, "ts": "2025-05-26T23:10:08+00:00", "symbol": "Elite Material", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Elite Material also continues to operate at a tight utilization rate. It is expected to add 600K units of annual capacity at its Kunshan plant, which will bring total capacity to a +15% YoY increase in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA bullish on PCB/CCL value chain (Elite Material, Gold Circuit, Accton) on ASIC-driven supply constraints; Celestica recovery 2H25; Marvell benefits from Tranium 2 surge.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Elite Material Co. 2383.TW Taiwan CCL", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250527_All Eyes on ASIC – CLSA\n\n[Key Takeaways]\n    \n1. Based on our latest CoWoS capacity update, we expect strong ASIC demand in 2025–2026, driven primarily by:\n    •  (1) AWS Tranium\n    •  (2) Google TPU\n    \n2. As a result, supply constraints will likely intensify across the value chain, especially in PCBs and CCL (including mainboards), which should lead to strong earnings for the related suppliers.\n\n⸻\n\n[Contents]\n\n1. According to recent updates on TSMC’s CoWoS capacity, strong ASIC demand is expected to persist through 2025 and 2026.\n    \n2. CoWoS capacity is projected to grow from 67K wafers/month in Q4 2025 to 85K wafers/month by Q4 2026, a YoY increase of +27%, led by:\n    \n• (1) NVIDIA\n• (2) Expanding ASIC demand\n    \n3. We estimate that CoWoS demand for AWS Tranium 2 (manufactured by Marvell) will surge by +371% YoY in 2025.\n    \n4. Broadcom’s CoWoS demand is also expected to grow +29% YoY in 2026, mainly due to:\n    \n• (1) Google TPU\n• (2) Meta MTIA v3\n    \n5. Demand for Tranium 2 and future next-generation versions is expected to provide strong momentum across the ASIC-related value chain in 2025.\n\n6. Amazon executives have cited bottlenecks in the supply of mainboards, CCL, PCB, and other components essential to AI infrastructure. As component/raw material capacity expands, Elite Material, Gold Circuit, and Accton are expected to benefit.\n    \n7. Beginning in 2H25, the ramp-up in mass production of Google TPU v6/v7e/v7p will also drive value chain demand and support strong earnings across related sectors.\n   \n8. Celestica previously attributed revenue declines in its Enterprise segment—beginning 4Q25—to a technology transition in AI/ML compute programs for hyperscaler clients, especially related to Google TPU:\n    \n• Celestica’s Enterprise segment revenue declined by -10% in 4Q24, -39% in 1Q25, and is expected to decline by -40% in 2Q25(F).\n• A recovery is expected from 2H25 onward.\n    \n9. Our CoWoS demand outlook assumes Broadcom’s CoWoS demand will grow +29% YoY in 2026, supported by:\n    \n• The launch of TPU v6e from 2Q25\n• The full-scale ramp of TPU v7p in 2026\n    \n10. Thus, the mass production of TPUs starting in 2H25 will positively impact earnings across lower-tier value chain players, especially given continued industry-wide supply constraints.\n    \n11. Given the ongoing supply shortage across the sector, we maintain a preference for PCB and CCL companies.\n    \n12. Gold Circuit is currently operating at full capacity. The company is expanding capacity at its Suzhou plant in China and completing the remaining half of Phase 1 at its Thailand facility.\n    \n13. Elite Material also continues to operate at a tight utilization rate. It is expected to add 600K units of annual capacity at its Kunshan plant, which will bring total capacity to a +15% YoY increase in 2026.\n    \n14. GCEis also expected to increase its share of AWS ASIC orders starting from 2H25. Thanks to its proactive capacity expansion strategy, GCE is projected to gain market share over Chinese competitors.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0042432374305405585, "ret_signed_m1d": 0.0042432374305405585, "alpha_spy_m1d": 0.0042432374305405585, "alpha_c_m1d": 0.0042432374305405585, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0014144416309740393, "ret_signed_p1d": -0.0014144416309740393, "alpha_spy_p1d": -0.022205109352441355, "alpha_c_p1d": -0.025178770097267433, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.023764328466293394, "ret_p1w": 0.07213573601824175, "ret_signed_p1w": 0.07213573601824175, "alpha_spy_p1w": 0.04865131822548374, "alpha_c_p1w": 0.04506465494431078, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02707108107393097, "ret_p1m": 0.2164072955171068, "ret_signed_p1m": 0.2164072955171068, "alpha_spy_p1m": 0.16552954277562515, "alpha_c_p1m": 0.12452527066378893, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.09188202485331787, "ret_p3m": 0.7468173970023317, "ret_signed_p3m": 0.7468173970023317, "alpha_spy_p3m": 0.6461131032591683, "alpha_c_p3m": 0.6033972446953622, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.14342015230696958, "ret_p6m": 0.9345279353534053, "ret_signed_p6m": 0.9345279353534053, "alpha_spy_p6m": 0.7881708195048824, "alpha_c_p6m": 0.7006685424012749, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.23385939295213043, "price_path": [-0.1047, -0.1061, -0.1061, -0.1146, -0.1457, -0.1429, -0.1471, -0.1796, -0.1924, -0.2037, -0.1895, -0.1796, -0.1796, -0.1711, -0.1429, -0.1726, -0.1754, -0.1584, -0.1457, -0.1358, -0.1414, -0.1584, -0.1726, -0.2291, -0.2291, -0.2546, -0.2546, -0.2546, -0.3289, -0.3607, -0.4243, -0.367, -0.3041, -0.2702, -0.2306, -0.2687, -0.2716, -0.2617, -0.2702, -0.2942, -0.2617, -0.2716, -0.1994, -0.2122, -0.2008, -0.2136, -0.2136, -0.1358, -0.1188, -0.1174, -0.1344, -0.0636, -0.0396, -0.0453, -0.0467, -0.0283, -0.0311, -0.0325, -0.0566, -0.0438, 0.0184, 0.0099, 0.0042, 0.0, -0.0014, 0.0198, 0.0594, 0.0594, 0.0721, 0.1089, 0.1033, 0.116, 0.1146, 0.1598, 0.2588, 0.2348, 0.2136, 0.2164, 0.2433, 0.2263, 0.1952, 0.1952, 0.1853, 0.1796, 0.2164, 0.2136, 0.2093, 0.2249, 0.2475, 0.273, 0.2772, 0.3126, 0.3055, 0.2744, 0.3479, 0.3536, 0.3536, 0.331, 0.3126, 0.372, 0.3904, 0.3762, 0.4498, 0.4356, 0.4003, 0.3975, 0.3989, 0.4059, 0.4498, 0.4356, 0.4215, 0.5629, 0.5629, 0.57, 0.5912, 0.57, 0.6195, 0.6266, 0.6337, 0.662, 0.6973, 0.7115, 0.8105, 0.8458, 0.7397, 0.6266, 0.7468, 0.7044, 0.768, 0.768, 0.8105, 0.7912, 0.7482, 0.6551, 0.6479, 0.6694, 0.5978, 0.6694, 0.6408, 0.7554, 0.7984, 0.8987, 0.9274, 0.8844, 0.8414, 0.8844, 0.913, 0.8485, 0.8629, 0.9274, 0.8557, 0.7984, 0.7554, 0.7554, 0.7554, 0.7626, 0.8557, 0.87, 0.87, 0.7697, 0.7124, 0.7697, 0.7697, 0.7196, 0.6766, 0.6336, 0.6336, 0.5834, 0.6121, 0.6193, 0.6049, 0.5978, 0.5978, 0.6694, 0.7984, 0.7912, 0.8271, 0.9489, 1.0205, 1.1065, 1.0635, 1.0778, 0.9632, 0.9489, 0.8915, 0.9274, 0.9704, 0.9775, 1.0778, 0.9345]}
{"unit_id": "orig:1927140222598345058", "ticker_idx": 1, "ts": "2025-05-26T23:10:08+00:00", "symbol": "Gold Circuit", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Gold Circuit is currently operating at full capacity. The company is expanding capacity at its Suzhou plant in China and completing the remaining half of Phase 1 at its Thailand facility", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA bullish on PCB/CCL value chain (Elite Material, Gold Circuit, Accton) on ASIC-driven supply constraints; Celestica recovery 2H25; Marvell benefits from Tranium 2 surge.", "resolved_tickers": ["2368.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Gold Circuit Electronics 2368.TW Taiwan PCB", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250527_All Eyes on ASIC – CLSA\n\n[Key Takeaways]\n    \n1. Based on our latest CoWoS capacity update, we expect strong ASIC demand in 2025–2026, driven primarily by:\n    •  (1) AWS Tranium\n    •  (2) Google TPU\n    \n2. As a result, supply constraints will likely intensify across the value chain, especially in PCBs and CCL (including mainboards), which should lead to strong earnings for the related suppliers.\n\n⸻\n\n[Contents]\n\n1. According to recent updates on TSMC’s CoWoS capacity, strong ASIC demand is expected to persist through 2025 and 2026.\n    \n2. CoWoS capacity is projected to grow from 67K wafers/month in Q4 2025 to 85K wafers/month by Q4 2026, a YoY increase of +27%, led by:\n    \n• (1) NVIDIA\n• (2) Expanding ASIC demand\n    \n3. We estimate that CoWoS demand for AWS Tranium 2 (manufactured by Marvell) will surge by +371% YoY in 2025.\n    \n4. Broadcom’s CoWoS demand is also expected to grow +29% YoY in 2026, mainly due to:\n    \n• (1) Google TPU\n• (2) Meta MTIA v3\n    \n5. Demand for Tranium 2 and future next-generation versions is expected to provide strong momentum across the ASIC-related value chain in 2025.\n\n6. Amazon executives have cited bottlenecks in the supply of mainboards, CCL, PCB, and other components essential to AI infrastructure. As component/raw material capacity expands, Elite Material, Gold Circuit, and Accton are expected to benefit.\n    \n7. Beginning in 2H25, the ramp-up in mass production of Google TPU v6/v7e/v7p will also drive value chain demand and support strong earnings across related sectors.\n   \n8. Celestica previously attributed revenue declines in its Enterprise segment—beginning 4Q25—to a technology transition in AI/ML compute programs for hyperscaler clients, especially related to Google TPU:\n    \n• Celestica’s Enterprise segment revenue declined by -10% in 4Q24, -39% in 1Q25, and is expected to decline by -40% in 2Q25(F).\n• A recovery is expected from 2H25 onward.\n    \n9. Our CoWoS demand outlook assumes Broadcom’s CoWoS demand will grow +29% YoY in 2026, supported by:\n    \n• The launch of TPU v6e from 2Q25\n• The full-scale ramp of TPU v7p in 2026\n    \n10. Thus, the mass production of TPUs starting in 2H25 will positively impact earnings across lower-tier value chain players, especially given continued industry-wide supply constraints.\n    \n11. Given the ongoing supply shortage across the sector, we maintain a preference for PCB and CCL companies.\n    \n12. Gold Circuit is currently operating at full capacity. The company is expanding capacity at its Suzhou plant in China and completing the remaining half of Phase 1 at its Thailand facility.\n    \n13. Elite Material also continues to operate at a tight utilization rate. It is expected to add 600K units of annual capacity at its Kunshan plant, which will bring total capacity to a +15% YoY increase in 2026.\n    \n14. GCEis also expected to increase its share of AWS ASIC orders starting from 2H25. Thanks to its proactive capacity expansion strategy, GCE is projected to gain market share over Chinese competitors.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0020575829071749485, "ret_signed_m1d": 0.0020575829071749485, "alpha_spy_m1d": 0.0020575829071749485, "alpha_c_m1d": 0.0020575829071749485, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0020575829071749485, "ret_signed_p1d": 0.0020575829071749485, "alpha_spy_p1d": -0.018733084814292367, "alpha_c_p1d": -0.021706745559118445, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.023764328466293394, "ret_p1w": 0.0493826962274071, "ret_signed_p1w": 0.0493826962274071, "alpha_spy_p1w": 0.02589827843464909, "alpha_c_p1w": 0.022311615153476128, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02707108107393097, "ret_p1m": 0.15324728237920238, "ret_signed_p1m": 0.15324728237920238, "alpha_spy_p1m": 0.10236952963772072, "alpha_c_p1m": 0.06136525752588451, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.09188202485331787, "ret_p3m": 0.8898103276943865, "ret_signed_p3m": 0.8898103276943865, "alpha_spy_p3m": 0.789106033951223, "alpha_c_p3m": 0.7463901753874169, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.14342015230696958, "ret_p6m": 1.1339283655702759, "ret_signed_p6m": 1.1339283655702759, "alpha_spy_p6m": 0.987571249721753, "alpha_c_p6m": 0.9000689726181454, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.23385939295213043, "price_path": [-0.0864, -0.0864, -0.0864, -0.107, -0.0844, -0.0782, -0.0617, -0.0617, -0.0947, -0.1399, -0.107, -0.107, -0.1091, -0.0905, -0.0823, -0.1111, -0.0967, -0.0844, -0.1193, -0.1091, -0.0988, -0.1029, -0.1296, -0.1728, -0.1811, -0.1975, -0.1975, -0.1975, -0.2778, -0.3498, -0.4136, -0.356, -0.2963, -0.2572, -0.2099, -0.2469, -0.2449, -0.2407, -0.2387, -0.2675, -0.2037, -0.2202, -0.1975, -0.2078, -0.1831, -0.2119, -0.2119, -0.1379, -0.1111, -0.0905, -0.1173, -0.0782, -0.07, -0.0535, -0.0782, -0.07, -0.0556, -0.072, -0.0926, -0.0864, 0.0041, 0.0288, 0.0021, 0.0, 0.0021, 0.0144, 0.07, 0.07, 0.0494, 0.1049, 0.1029, 0.1481, 0.1132, 0.1461, 0.1584, 0.1626, 0.1461, 0.144, 0.1502, 0.107, 0.0844, 0.0864, 0.0967, 0.1091, 0.1532, 0.1448, 0.1343, 0.1617, 0.2416, 0.2395, 0.2227, 0.2206, 0.2648, 0.2543, 0.3279, 0.3595, 0.3216, 0.3069, 0.3048, 0.3469, 0.3363, 0.3448, 0.3889, 0.3995, 0.3742, 0.3742, 0.3932, 0.4205, 0.4521, 0.471, 0.4879, 0.5762, 0.5762, 0.5236, 0.5426, 0.5489, 0.5657, 0.7215, 0.854, 0.9403, 0.9361, 0.9614, 0.9593, 1.014, 0.9698, 0.8667, 0.8898, 0.9319, 0.9593, 0.9277, 1.0035, 1.0455, 1.1255, 0.9214, 0.8309, 0.8603, 0.8372, 0.8477, 0.8898, 0.9572, 1.0582, 1.0771, 1.0497, 0.974, 0.9319, 0.8856, 0.8667, 0.833, 0.8919, 0.9151, 0.9319, 0.8435, 0.7362, 0.7362, 0.833, 0.8309, 0.8309, 0.8582, 0.8582, 0.814, 0.833, 0.8267, 0.8267, 0.7909, 0.732, 0.7383, 0.7509, 0.7004, 0.7425, 0.7699, 0.6857, 0.6794, 0.6794, 0.7172, 0.7783, 0.7509, 0.7825, 0.9593, 1.075, 1.0329, 0.955, 1.014, 0.955, 1.0287, 1.0982, 1.3065, 1.3865, 1.2897, 1.2602, 1.1339]}
{"unit_id": "orig:1927140222598345058", "ticker_idx": 2, "ts": "2025-05-26T23:10:08+00:00", "symbol": "Accton", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Elite Material, Gold Circuit, and Accton are expected to benefit", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA bullish on PCB/CCL value chain (Elite Material, Gold Circuit, Accton) on ASIC-driven supply constraints; Celestica recovery 2H25; Marvell benefits from Tranium 2 surge.", "resolved_tickers": ["2345.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Accton Technology 2345.TW Taiwan", "bench_c": "EWT", "bench_c_reason": "country_fallback (suffix='TW')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "250527_All Eyes on ASIC – CLSA\n\n[Key Takeaways]\n    \n1. Based on our latest CoWoS capacity update, we expect strong ASIC demand in 2025–2026, driven primarily by:\n    •  (1) AWS Tranium\n    •  (2) Google TPU\n    \n2. As a result, supply constraints will likely intensify across the value chain, especially in PCBs and CCL (including mainboards), which should lead to strong earnings for the related suppliers.\n\n⸻\n\n[Contents]\n\n1. According to recent updates on TSMC’s CoWoS capacity, strong ASIC demand is expected to persist through 2025 and 2026.\n    \n2. CoWoS capacity is projected to grow from 67K wafers/month in Q4 2025 to 85K wafers/month by Q4 2026, a YoY increase of +27%, led by:\n    \n• (1) NVIDIA\n• (2) Expanding ASIC demand\n    \n3. We estimate that CoWoS demand for AWS Tranium 2 (manufactured by Marvell) will surge by +371% YoY in 2025.\n    \n4. Broadcom’s CoWoS demand is also expected to grow +29% YoY in 2026, mainly due to:\n    \n• (1) Google TPU\n• (2) Meta MTIA v3\n    \n5. Demand for Tranium 2 and future next-generation versions is expected to provide strong momentum across the ASIC-related value chain in 2025.\n\n6. Amazon executives have cited bottlenecks in the supply of mainboards, CCL, PCB, and other components essential to AI infrastructure. As component/raw material capacity expands, Elite Material, Gold Circuit, and Accton are expected to benefit.\n    \n7. Beginning in 2H25, the ramp-up in mass production of Google TPU v6/v7e/v7p will also drive value chain demand and support strong earnings across related sectors.\n   \n8. Celestica previously attributed revenue declines in its Enterprise segment—beginning 4Q25—to a technology transition in AI/ML compute programs for hyperscaler clients, especially related to Google TPU:\n    \n• Celestica’s Enterprise segment revenue declined by -10% in 4Q24, -39% in 1Q25, and is expected to decline by -40% in 2Q25(F).\n• A recovery is expected from 2H25 onward.\n    \n9. Our CoWoS demand outlook assumes Broadcom’s CoWoS demand will grow +29% YoY in 2026, supported by:\n    \n• The launch of TPU v6e from 2Q25\n• The full-scale ramp of TPU v7p in 2026\n    \n10. Thus, the mass production of TPUs starting in 2H25 will positively impact earnings across lower-tier value chain players, especially given continued industry-wide supply constraints.\n    \n11. Given the ongoing supply shortage across the sector, we maintain a preference for PCB and CCL companies.\n    \n12. Gold Circuit is currently operating at full capacity. The company is expanding capacity at its Suzhou plant in China and completing the remaining half of Phase 1 at its Thailand facility.\n    \n13. Elite Material also continues to operate at a tight utilization rate. It is expected to add 600K units of annual capacity at its Kunshan plant, which will bring total capacity to a +15% YoY increase in 2026.\n    \n14. GCEis also expected to increase its share of AWS ASIC orders starting from 2H25. Thanks to its proactive capacity expansion strategy, GCE is projected to gain market share over Chinese competitors.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.001375475924280778, "ret_signed_m1d": -0.001375475924280778, "alpha_spy_m1d": -0.001375475924280778, "alpha_c_m1d": -0.001375475924280778, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008253025651537294, "ret_signed_p1d": -0.008253025651537294, "alpha_spy_p1d": -0.02904369337300461, "alpha_c_p1d": -0.009374193819456589, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.0011211681679192953, "ret_p1w": -0.002750951848561556, "ret_signed_p1w": -0.002750951848561556, "alpha_spy_p1w": -0.02623536964131956, "alpha_c_p1w": 0.009954577011212162, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": -0.012705528859773718, "ret_p1m": -0.00550198875004948, "ret_signed_p1m": -0.00550198875004948, "alpha_spy_p1m": -0.05637974149153113, "alpha_c_p1m": -0.0671611685371446, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.06165917978709512, "ret_p3m": 0.3238317874255203, "ret_signed_p3m": 0.3238317874255203, "alpha_spy_p3m": 0.22312749368235685, "alpha_c_p3m": 0.22947451595774804, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.09435727146777229, "ret_p6m": 0.2834200802304254, "ret_signed_p6m": 0.2834200802304254, "alpha_spy_p6m": 0.13706296438190257, "alpha_c_p6m": 0.11245598743208385, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.17096409279834157, "price_path": [-0.044, -0.0743, -0.0743, -0.0977, -0.0633, -0.055, -0.044, -0.0894, -0.1073, -0.1279, -0.121, -0.1059, -0.121, -0.1018, -0.0825, -0.1142, -0.1169, -0.088, -0.1155, -0.0935, -0.099, -0.1279, -0.1609, -0.2077, -0.1898, -0.1898, -0.1898, -0.1898, -0.2696, -0.3425, -0.4017, -0.3418, -0.2999, -0.2779, -0.249, -0.2834, -0.2806, -0.2779, -0.304, -0.3343, -0.3026, -0.2902, -0.2407, -0.2366, -0.1774, -0.1829, -0.1829, -0.1458, -0.1582, -0.1224, -0.1307, -0.1087, -0.0839, -0.0729, -0.0633, -0.0124, -0.0041, -0.0151, -0.0303, -0.0316, 0.011, 0.022, -0.0014, 0.0, -0.0083, 0.0124, 0.0248, 0.0248, -0.0028, -0.0041, 0.0083, 0.0083, -0.0055, 0.0138, 0.0812, 0.1073, 0.0935, 0.0963, 0.0825, 0.0303, 0.0041, -0.0165, -0.0193, -0.0179, -0.0055, 0.022, 0.022, 0.0069, 0.0041, 0.0495, 0.0564, 0.0743, 0.0564, 0.0151, 0.0454, 0.0949, 0.1183, 0.0963, 0.0468, 0.0674, 0.0578, 0.0509, 0.0977, 0.0908, 0.0578, 0.0688, 0.0633, 0.0908, 0.1706, 0.1719, 0.1747, 0.243, 0.243, 0.2333, 0.2555, 0.2555, 0.3587, 0.3782, 0.4214, 0.4214, 0.4562, 0.4423, 0.4423, 0.505, 0.4144, 0.3322, 0.3238, 0.3099, 0.3406, 0.367, 0.3698, 0.3712, 0.4144, 0.3155, 0.3419, 0.3517, 0.3587, 0.4214, 0.4214, 0.4214, 0.5189, 0.5607, 0.5398, 0.4911, 0.4911, 0.4841, 0.505, 0.4144, 0.3782, 0.4005, 0.4005, 0.3601, 0.3712, 0.3712, 0.4562, 0.4492, 0.498, 0.5398, 0.5398, 0.4283, 0.4562, 0.4632, 0.4632, 0.4423, 0.3935, 0.4214, 0.3935, 0.3322, 0.3642, 0.3378, 0.2597, 0.3015, 0.3015, 0.3406, 0.3447, 0.3559, 0.4214, 0.505, 0.498, 0.6025, 0.4562, 0.4702, 0.4005, 0.3907, 0.3642, 0.3698, 0.3656, 0.3698, 0.3907, 0.2834]}
{"unit_id": "orig:1927140222598345058", "ticker_idx": 3, "ts": "2025-05-26T23:10:08+00:00", "symbol": "Celestica", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "A recovery is expected from 2H25 onward", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA bullish on PCB/CCL value chain (Elite Material, Gold Circuit, Accton) on ASIC-driven supply constraints; Celestica recovery 2H25; Marvell benefits from Tranium 2 surge.", "resolved_tickers": ["CLS"], "resolution_method": "exchange_ticker", "resolution_reason": "Celestica Inc. CLS US/NYSE", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250527_All Eyes on ASIC – CLSA\n\n[Key Takeaways]\n    \n1. Based on our latest CoWoS capacity update, we expect strong ASIC demand in 2025–2026, driven primarily by:\n    •  (1) AWS Tranium\n    •  (2) Google TPU\n    \n2. As a result, supply constraints will likely intensify across the value chain, especially in PCBs and CCL (including mainboards), which should lead to strong earnings for the related suppliers.\n\n⸻\n\n[Contents]\n\n1. According to recent updates on TSMC’s CoWoS capacity, strong ASIC demand is expected to persist through 2025 and 2026.\n    \n2. CoWoS capacity is projected to grow from 67K wafers/month in Q4 2025 to 85K wafers/month by Q4 2026, a YoY increase of +27%, led by:\n    \n• (1) NVIDIA\n• (2) Expanding ASIC demand\n    \n3. We estimate that CoWoS demand for AWS Tranium 2 (manufactured by Marvell) will surge by +371% YoY in 2025.\n    \n4. Broadcom’s CoWoS demand is also expected to grow +29% YoY in 2026, mainly due to:\n    \n• (1) Google TPU\n• (2) Meta MTIA v3\n    \n5. Demand for Tranium 2 and future next-generation versions is expected to provide strong momentum across the ASIC-related value chain in 2025.\n\n6. Amazon executives have cited bottlenecks in the supply of mainboards, CCL, PCB, and other components essential to AI infrastructure. As component/raw material capacity expands, Elite Material, Gold Circuit, and Accton are expected to benefit.\n    \n7. Beginning in 2H25, the ramp-up in mass production of Google TPU v6/v7e/v7p will also drive value chain demand and support strong earnings across related sectors.\n   \n8. Celestica previously attributed revenue declines in its Enterprise segment—beginning 4Q25—to a technology transition in AI/ML compute programs for hyperscaler clients, especially related to Google TPU:\n    \n• Celestica’s Enterprise segment revenue declined by -10% in 4Q24, -39% in 1Q25, and is expected to decline by -40% in 2Q25(F).\n• A recovery is expected from 2H25 onward.\n    \n9. Our CoWoS demand outlook assumes Broadcom’s CoWoS demand will grow +29% YoY in 2026, supported by:\n    \n• The launch of TPU v6e from 2Q25\n• The full-scale ramp of TPU v7p in 2026\n    \n10. Thus, the mass production of TPUs starting in 2H25 will positively impact earnings across lower-tier value chain players, especially given continued industry-wide supply constraints.\n    \n11. Given the ongoing supply shortage across the sector, we maintain a preference for PCB and CCL companies.\n    \n12. Gold Circuit is currently operating at full capacity. The company is expanding capacity at its Suzhou plant in China and completing the remaining half of Phase 1 at its Thailand facility.\n    \n13. Elite Material also continues to operate at a tight utilization rate. It is expected to add 600K units of annual capacity at its Kunshan plant, which will bring total capacity to a +15% YoY increase in 2026.\n    \n14. GCEis also expected to increase its share of AWS ASIC orders starting from 2H25. Thanks to its proactive capacity expansion strategy, GCE is projected to gain market share over Chinese competitors.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004938511435981452, "ret_signed_p1d": -0.004938511435981452, "alpha_spy_p1d": -0.025729179157448767, "alpha_c_p1d": -0.028702839902274846, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.023764328466293394, "ret_p1w": -0.023687970219625876, "ret_signed_p1w": -0.023687970219625876, "alpha_spy_p1w": -0.04717238801238388, "alpha_c_p1w": -0.050759051293556845, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02707108107393097, "ret_p1m": 0.20398422087706813, "ret_signed_p1m": 0.20398422087706813, "alpha_spy_p1m": 0.15310646813558648, "alpha_c_p1m": 0.11210219602375027, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.09188202485331787, "ret_p3m": 0.5178705489664459, "ret_signed_p3m": 0.5178705489664459, "alpha_spy_p3m": 0.41716625522328243, "alpha_c_p3m": 0.3744503966594763, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.14342015230696958, "ret_p6m": 1.5104211129426255, "ret_signed_p6m": 1.5104211129426255, "alpha_spy_p6m": 1.3640639970941026, "alpha_c_p6m": 1.276561719990495, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.23385939295213043, "price_path": [-0.0652, -0.0971, -0.104, -0.2232, -0.2448, -0.2387, -0.3168, -0.2673, -0.3207, -0.2752, -0.2502, -0.2767, -0.2334, -0.2318, -0.2678, -0.2306, -0.1994, -0.1969, -0.1631, -0.185, -0.2663, -0.2936, -0.3079, -0.3403, -0.3382, -0.3107, -0.4209, -0.4463, -0.4108, -0.4237, -0.3131, -0.3598, -0.3377, -0.323, -0.3089, -0.3257, -0.2966, -0.2966, -0.3254, -0.3072, -0.2618, -0.2288, -0.2546, -0.2654, -0.2714, -0.2856, -0.2533, -0.2126, -0.2269, -0.2032, -0.1918, -0.1937, -0.2211, -0.1603, -0.0804, -0.0504, -0.0668, -0.0521, -0.0659, -0.057, -0.0455, -0.0096, 0.0, 0.0, -0.0049, -0.0113, -0.0259, -0.0342, -0.0237, 0.015, 0.0036, -0.0013, 0.0105, 0.0266, 0.0171, 0.0688, 0.0942, 0.049, 0.0734, 0.1064, 0.1405, 0.1405, 0.1319, 0.1412, 0.204, 0.2475, 0.2568, 0.2779, 0.3067, 0.2464, 0.2966, 0.3331, 0.3331, 0.289, 0.3033, 0.3468, 0.3295, 0.3479, 0.358, 0.3433, 0.3401, 0.3634, 0.3403, 0.3645, 0.3134, 0.4128, 0.3726, 0.4248, 0.4512, 0.6908, 0.6942, 0.6729, 0.63, 0.6862, 0.6448, 0.6626, 0.6842, 0.7382, 0.7126, 0.7866, 0.7094, 0.6459, 0.6362, 0.6312, 0.5462, 0.5309, 0.5179, 0.5813, 0.6147, 0.6625, 0.6916, 0.7773, 0.6301, 0.6301, 0.6793, 0.7734, 0.8527, 1.0313, 1.0451, 1.043, 1.1179, 1.0569, 1.0237, 1.0841, 1.1293, 1.073, 1.1097, 1.1171, 1.1327, 1.1565, 1.0127, 1.0477, 1.0406, 1.0587, 1.0623, 1.1067, 1.1002, 0.9559, 0.9699, 0.9904, 1.1334, 1.1744, 1.04, 1.1852, 1.0737, 1.2596, 1.3425, 1.3141, 1.2893, 1.2155, 1.2674, 1.3665, 1.4828, 1.5263, 1.7325, 1.8272, 1.847, 1.8834, 1.9213, 1.8107, 1.9515, 1.8523, 1.6961, 1.8845, 1.7778, 1.8005, 1.4504, 1.6022, 1.5895, 1.5104]}
{"unit_id": "orig:1927140222598345058", "ticker_idx": 4, "ts": "2025-05-26T23:10:08+00:00", "symbol": "Marvell", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CoWoS demand for AWS Tranium 2 (manufactured by Marvell) will surge by +371% YoY in 2025", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA bullish on PCB/CCL value chain (Elite Material, Gold Circuit, Accton) on ASIC-driven supply constraints; Celestica recovery 2H25; Marvell benefits from Tranium 2 surge.", "resolved_tickers": ["MRVL"], "resolution_method": "exchange_ticker", "resolution_reason": "Marvell Technology MRVL US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250527_All Eyes on ASIC – CLSA\n\n[Key Takeaways]\n    \n1. Based on our latest CoWoS capacity update, we expect strong ASIC demand in 2025–2026, driven primarily by:\n    •  (1) AWS Tranium\n    •  (2) Google TPU\n    \n2. As a result, supply constraints will likely intensify across the value chain, especially in PCBs and CCL (including mainboards), which should lead to strong earnings for the related suppliers.\n\n⸻\n\n[Contents]\n\n1. According to recent updates on TSMC’s CoWoS capacity, strong ASIC demand is expected to persist through 2025 and 2026.\n    \n2. CoWoS capacity is projected to grow from 67K wafers/month in Q4 2025 to 85K wafers/month by Q4 2026, a YoY increase of +27%, led by:\n    \n• (1) NVIDIA\n• (2) Expanding ASIC demand\n    \n3. We estimate that CoWoS demand for AWS Tranium 2 (manufactured by Marvell) will surge by +371% YoY in 2025.\n    \n4. Broadcom’s CoWoS demand is also expected to grow +29% YoY in 2026, mainly due to:\n    \n• (1) Google TPU\n• (2) Meta MTIA v3\n    \n5. Demand for Tranium 2 and future next-generation versions is expected to provide strong momentum across the ASIC-related value chain in 2025.\n\n6. Amazon executives have cited bottlenecks in the supply of mainboards, CCL, PCB, and other components essential to AI infrastructure. As component/raw material capacity expands, Elite Material, Gold Circuit, and Accton are expected to benefit.\n    \n7. Beginning in 2H25, the ramp-up in mass production of Google TPU v6/v7e/v7p will also drive value chain demand and support strong earnings across related sectors.\n   \n8. Celestica previously attributed revenue declines in its Enterprise segment—beginning 4Q25—to a technology transition in AI/ML compute programs for hyperscaler clients, especially related to Google TPU:\n    \n• Celestica’s Enterprise segment revenue declined by -10% in 4Q24, -39% in 1Q25, and is expected to decline by -40% in 2Q25(F).\n• A recovery is expected from 2H25 onward.\n    \n9. Our CoWoS demand outlook assumes Broadcom’s CoWoS demand will grow +29% YoY in 2026, supported by:\n    \n• The launch of TPU v6e from 2Q25\n• The full-scale ramp of TPU v7p in 2026\n    \n10. Thus, the mass production of TPUs starting in 2H25 will positively impact earnings across lower-tier value chain players, especially given continued industry-wide supply constraints.\n    \n11. Given the ongoing supply shortage across the sector, we maintain a preference for PCB and CCL companies.\n    \n12. Gold Circuit is currently operating at full capacity. The company is expanding capacity at its Suzhou plant in China and completing the remaining half of Phase 1 at its Thailand facility.\n    \n13. Elite Material also continues to operate at a tight utilization rate. It is expected to add 600K units of annual capacity at its Kunshan plant, which will bring total capacity to a +15% YoY increase in 2026.\n    \n14. GCEis also expected to increase its share of AWS ASIC orders starting from 2H25. Thanks to its proactive capacity expansion strategy, GCE is projected to gain market share over Chinese competitors.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.051573605071879136, "ret_signed_p1d": 0.051573605071879136, "alpha_spy_p1d": 0.03078293735041182, "alpha_c_p1d": 0.01986035057467972, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.03171325449719942, "ret_p1w": 0.012852261451459457, "ret_signed_p1w": 0.012852261451459457, "alpha_spy_p1w": -0.010632156341298549, "alpha_c_p1w": -0.011827677738614373, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02467993919007383, "ret_p1m": 0.23924875260168776, "ret_signed_p1m": 0.23924875260168776, "alpha_spy_p1m": 0.1883709998602061, "alpha_c_p1m": 0.09563355311126065, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.14361519949042711, "ret_p3m": 0.1743003426563836, "ret_signed_p3m": 0.1743003426563836, "alpha_spy_p3m": 0.07359604891322014, "alpha_c_p3m": -0.036363324985640766, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.21066366764202438, "ret_p6m": 0.2983447518514335, "ret_signed_p6m": 0.2983447518514335, "alpha_spy_p6m": 0.15198763600291065, "alpha_c_p6m": -0.10533621740458865, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.40368096925602215, "price_path": [0.5619, 0.4426, 0.5112, 0.4126, 0.4538, 0.4836, 0.1896, 0.1659, 0.0808, 0.1054, 0.1496, 0.1317, 0.1314, 0.1587, 0.1238, 0.149, 0.1602, 0.1585, 0.1975, 0.1796, 0.0988, 0.0675, 0.0211, 0.0133, 0.0315, 0.0407, -0.0843, -0.1865, -0.1606, -0.1766, 0.0033, -0.1298, -0.1203, -0.1389, -0.1214, -0.1443, -0.1481, -0.1481, -0.1864, -0.1659, -0.114, -0.0554, -0.0292, -0.0328, -0.0331, -0.0382, 0.0043, 0.027, 0.0213, 0.0087, -0.0722, -0.0504, -0.0171, 0.0628, 0.0789, 0.0858, 0.0743, 0.0506, 0.0308, 0.012, -0.0096, 0.0191, 0.0, 0.0, 0.0516, 0.0643, 0.0501, -0.0082, 0.0129, 0.0275, 0.0924, 0.0737, 0.1262, 0.1392, 0.1343, 0.1244, 0.1475, 0.1071, 0.1603, 0.1532, 0.235, 0.235, 0.2112, 0.1663, 0.2392, 0.2511, 0.3177, 0.2714, 0.2753, 0.2562, 0.2234, 0.2388, 0.2388, 0.1789, 0.1855, 0.1906, 0.2088, 0.199, 0.1957, 0.1941, 0.1684, 0.1875, 0.231, 0.2048, 0.1872, 0.2083, 0.221, 0.2238, 0.2518, 0.2589, 0.3479, 0.3254, 0.2277, 0.262, 0.2637, 0.2421, 0.2508, 0.2754, 0.2744, 0.2831, 0.308, 0.3034, 0.2564, 0.2655, 0.1885, 0.1745, 0.1743, 0.2038, 0.203, 0.2246, 0.2333, 0.2736, 0.0368, 0.0368, 0.0653, 0.0275, 0.0571, 0.0444, 0.0884, 0.1022, 0.1065, 0.0981, 0.1106, 0.112, 0.1355, 0.1705, 0.2241, 0.2246, 0.2455, 0.2305, 0.3207, 0.3821, 0.3715, 0.3587, 0.3864, 0.3834, 0.4215, 0.4218, 0.4664, 0.4342, 0.5254, 0.4954, 0.4127, 0.4751, 0.4228, 0.4668, 0.4559, 0.4513, 0.4165, 0.3904, 0.3375, 0.3658, 0.3883, 0.4639, 0.4597, 0.4876, 0.4615, 0.5469, 0.4912, 0.4454, 0.533, 0.5401, 0.5003, 0.5384, 0.4741, 0.4741, 0.4442, 0.4266, 0.3771, 0.2983]}
{"unit_id": "thread:1942730666082459939", "ticker_idx": 0, "ts": "2025-07-08T23:40:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The industry expects Q2 to be the bottom, with a reversal in Q3 potentially leading to an operating profit in the ₩8 trillion range", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung Q2 seen as the earnings bottom; inventory allowance reversal on HBM shipment recovery expected to lift Q3 operating profit to ~₩8T range.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Samsung Electronics Records ₩1.5 Trillion in HBM Inventory Asset Reversal\n\nSamsung Electronics' operating profit for the second quarter of this year was more than halved compared to a year ago. This significant reduction in operating profit was due to a large-scale reflection of inventory asset valuation allowances in the Semiconductor (DS) division.\n\nHowever, this allowance is largely characterized as a pre-emptive, one-time cost, and there is potential for partial reversal if shipments recover in the second half of the year. The industry expects Q2 to be the bottom, with a reversal in Q3 potentially leading to an operating profit in the ₩8 trillion range.\n\nOn the 8th, Samsung Electronics announced that its consolidated operating profit for the second quarter of this year was provisionally tallied at ₩4.6 trillion. This represents a 55.9% decrease compared to the same period last year (₩10.44 trillion) and a 31.2% decrease compared to the previous quarter (₩6.69 trillion). This figure is also more than ₩1.5 trillion lower than the initial securities firm consensus (₩6.1833 trillion).\n\nThis earnings result is provisional, and specific figures by division were not disclosed. However, Samsung Electronics stated in its explanation material that the reasons for the decrease in operating profit included inventory asset valuation allowances in the DS division, export sanctions on artificial intelligence (AI) semiconductors, and reduced line utilization rates.\n\nThe industry estimates that Samsung Electronics recognized an inventory asset valuation allowance of ₩1 trillion to ₩1.5 trillion in its DS division. This is analyzed as a result of significantly reflecting unsold inventory of high-performance DRAM, including High Bandwidth Memory (HBM), and the impact of restrictions on AI chip exports to China in the Foundry division.\n\nAn inventory asset valuation allowance is an accounting treatment that reduces the book value of unsold inventory to its actual market price. The depreciated amount is recognized as a loss and is deducted from operating profit as an expense.\n\nThe product estimated to account for the largest portion of this amount is HBM. Samsung Electronics significantly increased HBM production in response to expanding AI demand, but shipments were not made on time due to customer qualification delays, leaving production as inventory. It appears that a portion of this was processed as an inventory asset valuation allowance. The fact that HBM3E 12-high products have not passed NVIDIA's qual (quality) tests is also a regrettable point for Samsung Electronics.\n\nOn this day, Kim Sun-woo, a researcher at Meritz Securities, stated in a report titled 'Intersection of Structural and Lagging Costs' that \"Samsung Electronics' Q2 provisional operating profit of ₩4.6 trillion is disappointing, despite strong smartphone performance, due to decreased HBM sales, the reflection of lagging costs from advanced processes, and continued low Foundry utilization rates.\"\n\nResearcher Kim analyzed, \"Around mid-last year, when Samsung Electronics planned for HBM advance production (HBM3E 8-high), I judged that the decision to pre-invest wafers before quality certification results would cause some inventory burden. According to the ambitious production plan that has been underway since last year, it is likely that ₩1 trillion to ₩1.5 trillion in lagging costs, including sluggish inventory, has been recognized this year.\"\n\nHowever, the allowance processed as an expense this time will be reversed back into profit upon future product shipments.\n\nOn the same day, market research firm Counterpoint Research stated in its report 'SK Hynix, Competing for Leadership with Samsung in the Memory Market' that \"Although Samsung Electronics' operating profit for this quarter decreased by 56% compared to the same period last year, performance improvement is expected in the second half of the year as it supplies HBM3E products to AMD and Broadcom.\" However, it added, \"Shipments to NVIDIA remain uncertain,\" and \"Due to intensified sales regulations to China, the increase in HBM sales this year will be limited compared to last year.\"\n\n---\n\nhttps://t.co/WbUi0Oq7CP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004885940247018272, "ret_signed_m1d": 0.004885940247018272, "alpha_spy_m1d": 0.00433792387181664, "alpha_c_m1d": 0.010231441586980305, "bench_spy_m1d": 0.000548016375201632, "bench_c_m1d": -0.005345501339962033, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016286703429456373, "ret_signed_p1d": -0.016286703429456373, "alpha_spy_p1d": -0.022283330196667728, "alpha_c_p1d": -0.022295565333592227, "bench_spy_p1d": 0.005996626767211355, "bench_c_p1d": 0.006008861904135854, "ret_p1w": 0.037459282758823154, "ret_signed_p1w": 0.037459282758823154, "alpha_spy_p1w": 0.03455766355548784, "alpha_c_p1w": 0.031138258210178238, "bench_spy_p1w": 0.002901619203335315, "bench_c_p1w": 0.006321024548644916, "ret_p1m": 0.12052110347053557, "ret_signed_p1m": 0.12052110347053557, "alpha_spy_p1m": 0.10046765326937801, "alpha_c_p1m": 0.09414479469039039, "bench_spy_p1m": 0.020053450201157563, "bench_c_p1m": 0.026376308780145186, "ret_p3m": 0.4682479839729372, "ret_signed_p3m": 0.4682479839729372, "alpha_spy_p3m": 0.38647774919014033, "alpha_c_p3m": 0.3558912510545016, "bench_spy_p3m": 0.08177023478279688, "bench_c_p3m": 0.11235673291843562, "ret_p6m": 0.9710162609002393, "ret_signed_p6m": 0.9710162609002393, "alpha_spy_p6m": 0.8654432826420919, "alpha_c_p6m": 0.8443732150584198, "bench_spy_p6m": 0.10557297825814738, "bench_c_p6m": 0.1266430458418195, "price_path": [-0.087, -0.1065, -0.0903, -0.0838, -0.1146, -0.1081, -0.1048, -0.1032, -0.1097, -0.0984, -0.0984, -0.0984, -0.0967, -0.0967, -0.1016, -0.1016, -0.121, -0.121, -0.121, -0.1162, -0.1162, -0.1129, -0.0676, -0.0789, -0.0708, -0.0725, -0.0806, -0.0967, -0.0951, -0.0984, -0.1146, -0.1226, -0.1146, -0.1275, -0.0951, -0.0919, -0.0903, -0.0806, -0.0806, -0.0644, -0.0433, -0.0433, -0.032, -0.0417, -0.0304, -0.0369, -0.0563, -0.0741, -0.0595, -0.032, -0.0417, -0.0369, -0.0611, -0.0207, -0.0077, -0.0255, -0.0098, -0.0261, -0.0195, -0.0098, 0.0391, 0.0309, 0.0049, 0.0, -0.0163, -0.0065, 0.0195, 0.0179, 0.0375, 0.0537, 0.0863, 0.0928, 0.1042, 0.0749, 0.0814, 0.0749, 0.0733, 0.1466, 0.1498, 0.1824, 0.1629, 0.1221, 0.1352, 0.1384, 0.1205, 0.1482, 0.1694, 0.1564, 0.158, 0.171, 0.1661, 0.1661, 0.1401, 0.1401, 0.1482, 0.1498, 0.1629, 0.1645, 0.145, 0.1498, 0.1336, 0.1352, 0.101, 0.1254, 0.1368, 0.1417, 0.1319, 0.1417, 0.1645, 0.1824, 0.1954, 0.228, 0.2459, 0.2932, 0.2736, 0.3078, 0.3078, 0.3599, 0.3795, 0.3909, 0.4023, 0.3567, 0.3775, 0.3725, 0.4069, 0.4682, 0.4682, 0.4682, 0.4682, 0.4682, 0.4682, 0.5443, 0.5263, 0.4985, 0.5541, 0.5983, 0.6016, 0.6048, 0.595, 0.613, 0.5787, 0.6163, 0.6686, 0.6278, 0.6441, 0.703, 0.7586, 0.8175, 0.7161, 0.6457, 0.6228, 0.6016, 0.6457, 0.6932, 0.6866, 0.6817, 0.5901, 0.6457, 0.5999, 0.5787, 0.6457, 0.5509, 0.5819, 0.6245, 0.6817, 0.6932, 0.6441, 0.649, 0.6916, 0.7095, 0.7194, 0.7733, 0.7913, 0.7733, 0.7668, 0.7554, 0.7815, 0.7145, 0.6817, 0.7652, 0.7603, 0.739, 0.8077, 0.8241, 0.8175, 0.8175, 0.914, 0.9644, 0.971, 0.971]}
{"unit_id": "orig:1943317565473984632", "ticker_idx": 0, "ts": "2025-07-10T14:33:07+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reiterating its \"Buy\" rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi Buy reiteration on MRVL; XPU custom projects and interconnect TAM expansion drive long-term growth into 2026.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell: XPU Custom Projects Drive Long-term Growth, Interconnect TAM Continues to Expand\nCiti (A. Malik, July 07, 2025)\n\nCiti states that Marvell's stock price has risen by approximately 50% since its low in April. Management effectively boosted market confidence in long-term XPU growth opportunities through its Custom AI Day. Despite variations in project advancement pace, the firm maintains its expectations for XPU penetration and interconnect TAM expansion in 2026, reiterating its \"Buy\" rating.\n\nXPU Attach Definition and Representative Projects\nXPU Attach broadly refers to all custom projects surrounding customized computing engines, such as security offload chips, custom NICs with prefetch capabilities, CXL-enabled DRAM access chips, custom PCIe switches, and scale-up fabrics. These projects help the main chip focus on AI training/inference. Marvell is a primary supplier of such ancillary custom chips, with Meta's custom NIC being a representative example.\n\nLong-term Gross Margin Structure and Profit Model Optimization\nAlthough an increased proportion of ASIC revenue will pull down the overall gross margin to 55-60%, Marvell's operating margin (OM%) is still expected to continue improving, as the majority of R&D costs for custom ASICs are borne by clients. The ASP (Average Selling Price) per chip will also see generational growth due to process iteration and increased functional integration.\n\nInterconnect Market TAM Significantly Expands\nMarvell's interconnect product TAM (Total Addressable Market) has increased by 37%, primarily benefiting from DCI (Data Center Interconnect) opportunities driven by expanding AI cluster sizes, the trend of network opticalization (e.g., AEC replacing copper cables), and growing demand for PCIe retimers. Although data center growth is mainly driven by ASICs, the interconnect business still holds structural growth potential.\n\nProject Progress Pace Varies but 2026 Expectations Maintained\nCiti notes that the Trainium2 lifecycle might extend into 2026, and Microsoft's XPU project might be delayed by half a year. However, the overall XPU attach opportunity is still expected to ramp up as originally anticipated in 2026. Therefore, long-term forecasts remain unchanged, with Marvell expected to continue benefiting from expansion in both custom computing and interconnect segments.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014994601397067098, "ret_signed_m1d": -0.014994601397067098, "alpha_spy_m1d": -0.012182222958372901, "alpha_c_m1d": -0.0076909046179208795, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00804930967333295, "ret_signed_p1d": -0.00804930967333295, "alpha_spy_p1d": -0.004533885802737392, "alpha_c_p1d": -0.007944968105888406, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.017599007583071424, "ret_signed_p1w": -0.017599007583071424, "alpha_spy_p1w": -0.021146298650076556, "alpha_c_p1w": -0.03217190454151475, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.0551158045691138, "ret_signed_p1m": 0.0551158045691138, "alpha_spy_p1m": 0.03696369870313987, "alpha_c_p1m": 0.03421288800936173, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.1864937320103055, "ret_signed_p3m": 0.1864937320103055, "alpha_spy_p3m": 0.1143403012951476, "alpha_c_p3m": 0.014227401224113034, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.22031618273202058, "ret_signed_p6m": 0.22031618273202058, "alpha_spy_p6m": 0.12241533593996046, "alpha_c_p6m": -0.08206823502606242, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.2876, -0.2732, -0.2921, -0.2953, -0.2953, -0.3269, -0.31, -0.267, -0.2185, -0.1968, -0.1998, -0.2001, -0.2043, -0.1692, -0.1504, -0.1551, -0.1655, -0.2324, -0.2144, -0.1869, -0.1208, -0.1074, -0.1017, -0.1112, -0.1309, -0.1472, -0.1628, -0.1806, -0.1569, -0.1727, -0.1727, -0.13, -0.1195, -0.1313, -0.1795, -0.1621, -0.1499, -0.0962, -0.1118, -0.0683, -0.0575, -0.0616, -0.0698, -0.0507, -0.0841, -0.0401, -0.0459, 0.0217, 0.0217, 0.002, -0.0352, 0.0252, 0.035, 0.0901, 0.0518, 0.0551, 0.0393, 0.0121, 0.0248, 0.0248, -0.0247, -0.0192, -0.015, 0.0, -0.008, -0.0108, -0.0121, -0.0334, -0.0176, 0.0184, -0.0033, -0.0179, -0.0004, 0.0101, 0.0124, 0.0356, 0.0415, 0.1151, 0.0965, 0.0157, 0.0441, 0.0454, 0.0276, 0.0348, 0.0551, 0.0543, 0.0615, 0.0821, 0.0783, 0.0394, 0.0469, -0.0168, -0.0284, -0.0285, -0.0041, -0.0048, 0.0131, 0.0203, 0.0536, -0.1423, -0.1423, -0.1187, -0.1499, -0.1255, -0.136, -0.0996, -0.0881, -0.0846, -0.0915, -0.0812, -0.0801, -0.0606, -0.0317, 0.0127, 0.0131, 0.0304, 0.018, 0.0926, 0.1434, 0.1347, 0.124, 0.1469, 0.1445, 0.176, 0.1763, 0.2131, 0.1865, 0.2619, 0.2371, 0.1687, 0.2203, 0.177, 0.2135, 0.2045, 0.2007, 0.1719, 0.1503, 0.1065, 0.1299, 0.1485, 0.211, 0.2076, 0.2307, 0.2091, 0.2797, 0.2337, 0.1957, 0.2682, 0.2741, 0.2412, 0.2727, 0.2195, 0.2195, 0.1948, 0.1802, 0.1392, 0.0741, 0.1101, 0.0468, 0.0573, 0.1439, 0.139, 0.1975, 0.1975, 0.2205, 0.2437, 0.2681, 0.3679, 0.3405, 0.3503, 0.2559, 0.2136, 0.2624, 0.2209, 0.1526, 0.1503, 0.1477, 0.1153, 0.1532, 0.148, 0.1577, 0.197, 0.1807, 0.1807, 0.1787, 0.1708, 0.1844, 0.1601, 0.1601, 0.2203]}
{"unit_id": "orig:1930405942392434997", "ticker_idx": 0, "ts": "2025-06-04T23:26:56+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS remains constructive on SK Hynix, reiterating its Buy rating and raising the target price from KRW 262,000 to KRW 277,000", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS reiterates Buy on SK Hynix, raises PT to KRW 277K on DDR pricing uptrend, inventory drawdown, and HBM leadership.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Memory: DDR Pricing Rebound Amid Inventory and Tariff Uncertainty in 2H; HBM Demand Revised Down, but DRAM Outlook Remains Positive\n\nUBS (Nicolas Gaudois, 2025/06/02)\n\nDriven by synchronized demand from servers and smartphones, DDR demand has exceeded expectations, leading to short-term pricing momentum. However, inventory digestion and tariff uncertainties in the second half remain key concerns.\nAlthough HBM demand is being revised downward due to a slower ramp of AI ASICs, SK Hynix maintains strong market share and technological leadership in DRAM, and UBS reiterates its Buy rating.\n\n⸻\n\n• Short-Term Upside in DDR Pricing\n• Industry checks indicate that not only smartphone and PC OEMs but also cloud service providers (CSPs) are replenishing DDR5 server memory, leading to bit supply falling short of demand and continued inventory drawdowns across customers.\n• UBS now forecasts 2Q25 DDR contract prices to rise 7% QoQ, up from the previous estimate of 5%.\nSigns suggest that some DDR4 and LPDDR5 products are seeing price hikes exceeding 10%.\n• The 3Q25 DDR contract price forecast is also raised slightly from 2% to 3%.\n• NAND Flash contract pricing is unchanged, with expectations of +3% QoQ in 2Q25 and -3% QoQ in 3Q25.\n\n⸻\n\n• Inventory & Tariff Risks in 2H25\n• As of May 2025, DDR inventory levels stand at:\n• Hyperscaler data centers: ~13 weeks\n• Smartphone OEMs: ~10 weeks\n• PC OEMs: ~12 weeks\n— each down by ~3 weeks from the beginning of the year.\n• SK Hynix’s DDR5 finished goods inventory has dropped to ~2 weeks, one of the lowest levels in the industry.\n• However, part of the bit shipment growth in 2Q25 is believed to be front-loaded pre-stocking, and\nmemory shipment growth for PCs and smartphones is likely to moderate in 2H25.\n• In addition, the U.S. is considering tariffs of 10%–25% on both Chinese and non-Chinese smartphones,\ncreating downstream demand risk and pressure in 2H25 contract negotiations.\n\n⸻\n\n• HBM Demand Forecast Lowered\n• Due to delays in ramping up AI ASICs like Google TPU v6e/v7 and AWS Trainium 3,\nUBS lowers its 2025 HBM bit shipment forecast from 189B Gb to 163B Gb (+89% YoY),\nand the 2026 estimate from 291B Gb to 254B Gb (+56% YoY).\n• AWS Trainium 2/3 and Google TPU v7/v7e are all configured with HBM3E 12-Hi (144GB),\nand contrary to earlier expectations, have not yet transitioned to HBM4.\n• SK Hynix continues to expand its market share in cloud AI HBM, underscoring its technological leadership in the space.\n\n⸻\n\n• DRAM Sector Remains Attractive\n• While HBM demand faces headwinds, the uptrend in mainstream DRAM pricing remains intact.\nWith declining inventory and recovering demand, DDR supply-demand tightness is expected to persist into 2H25.\n• UBS remains constructive on SK Hynix, reiterating its Buy rating and raising the target price from KRW 262,000 to KRW 277,000,\nreflecting the company’s leadership in both DDR and HBM markets and its long-term earnings growth potential.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04597709938807859, "ret_signed_m1d": -0.04597709938807859, "alpha_spy_m1d": -0.0462457189281249, "alpha_c_m1d": -0.03421939259841866, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032183782009547635, "ret_signed_p1d": 0.032183782009547635, "alpha_spy_p1d": 0.03701655106573687, "alpha_c_p1d": 0.033971286622707875, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.1034481489964525, "ret_signed_p1w": 0.1034481489964525, "alpha_spy_p1w": 0.09433621783432744, "alpha_c_p1w": 0.06320983936782176, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.2804596714416854, "ret_signed_p1m": 0.2804596714416854, "alpha_spy_p1m": 0.2280059308757425, "alpha_c_p1m": 0.15386579796174837, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.17871148439176676, "ret_signed_p3m": 0.17871148439176676, "alpha_spy_p3m": 0.09308566509181015, "alpha_c_p3m": 0.025623022977242726, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 1.506555764559058, "ret_signed_p6m": 1.506555764559058, "alpha_spy_p6m": 1.3594758443928268, "alpha_c_p6m": 1.125384024885393, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.117, -0.1376, -0.1381, -0.0872, -0.0835, -0.0615, -0.0546, -0.0683, -0.0569, -0.0362, -0.011, -0.0293, -0.0454, -0.0179, -0.05, -0.0853, -0.1248, -0.0959, -0.0918, -0.1069, -0.1638, -0.2437, -0.2221, -0.2427, -0.1592, -0.1702, -0.173, -0.1712, -0.2014, -0.1969, -0.1969, -0.1895, -0.2024, -0.1693, -0.1817, -0.1537, -0.1647, -0.1702, -0.1854, -0.1854, -0.1464, -0.1464, -0.1464, -0.1243, -0.1266, -0.1276, -0.1051, -0.089, -0.0546, -0.0798, -0.0615, -0.0849, -0.0729, -0.0798, -0.0963, -0.0821, -0.0683, -0.0706, -0.0454, -0.0253, -0.0598, -0.046, -0.046, 0.0, 0.0322, 0.0322, 0.0529, 0.0598, 0.1034, 0.0828, 0.0828, 0.1402, 0.1448, 0.1333, 0.131, 0.1816, 0.1931, 0.2805, 0.3149, 0.3471, 0.3057, 0.3425, 0.3126, 0.2828, 0.2805, 0.2437, 0.246, 0.2966, 0.292, 0.3655, 0.354, 0.3793, 0.3724, 0.3609, 0.2391, 0.2368, 0.2529, 0.2345, 0.2368, 0.2391, 0.223, 0.2046, 0.2069, 0.2115, 0.2575, 0.1862, 0.1862, 0.2115, 0.1885, 0.2046, 0.1793, 0.2276, 0.2368, 0.2782, 0.2713, 0.2713, 0.2299, 0.2092, 0.1747, 0.1264, 0.154, 0.1931, 0.2023, 0.1954, 0.2363, 0.2386, 0.1787, 0.1994, 0.2086, 0.2225, 0.2593, 0.2754, 0.3261, 0.3997, 0.4135, 0.5125, 0.524, 0.6023, 0.5355, 0.6368, 0.6368, 0.6161, 0.6622, 0.6461, 0.6414, 0.5494, 0.6069, 0.6, 0.6576, 0.821, 0.821, 0.821, 0.821, 0.821, 0.821, 0.9707, 0.9108, 0.8947, 0.9453, 1.0835, 1.1433, 1.2354, 1.2055, 1.217, 1.2032, 1.3482, 1.4633, 1.3989, 1.5692, 1.6153, 1.5738, 1.8547, 1.6981, 1.6659, 1.7304, 1.6705, 1.7902, 1.8501, 1.8409, 1.8179, 1.5784, 1.7902, 1.6245, 1.5876, 1.6291, 1.3989, 1.3943, 1.3897, 1.4127, 1.5066]}
{"unit_id": "orig:1938233963790930222", "ticker_idx": 0, "ts": "2025-06-26T13:52:42+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We expect Nvidia (NVDA) to maintain approximately 80% or more of the overall AI accelerator market over time", "tweet_type": "analysis", "timeline": "years", "summary": "BofA adopted: AI accelerator TAM grows to $650B by 2030; NVDA holds 80%+ share, AMD ~5%, AVGO/MRVL ASIC ecosystem 10-15%; consensus estimates conservative with upside for most suppliers.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AI Accelerators (GPU, ASIC, XPU) 2\n\nWhile not all computers (including servers) need accelerators, all AI servers must be equipped with accelerators like GPUs, ASICs, and XPUs to be competitive in AI workloads.\n\nTypes of Accelerators\n\nThere are two main types of accelerators/XPUs: GPUs and ASICs.\n\nGPUs refer to general-purpose or commercial accelerator solutions, designed by Nvidia (NVDA) and AMD (and to some extent, Intel (INTC)).\n\nConsidering the continuously evolving nature of AI workloads and the excellent product roadmaps/cycles of the two GPU suppliers, we expect GPUs to maintain 80-85% of the total AI accelerator market over time.\n\nASICs generally refer to semi-custom-made accelerators designed by AVGO (Broadcom) and MRVL (Marvell) (and Asia-based suppliers like Alchip, GUC). Many hyperscalers with the resources to develop their own accelerators often take this path, collaborating with existing SerDes/IP/design suppliers like AVGO/MRVL. We believe a market for semi-custom accelerators (10-15% of the total market over time) exists for more specialized and repetitive workloads, such as Gmail search and Instagram algorithms, because they can be processed more efficiently on ASICs than on power-intensive GPUs.\n\nTotal Addressable Market (TAM)\n\nWe believe the AI accelerator market can grow from $126 billion in 2024 to over $400 billion in 2027, over $500 billion in 2028, and over $650 billion in 2030, with a Compound Annual Growth Rate (CAGR) of +26% from 2025-30. For reference, this is the total system market size (including accelerator silicon, HBM, software, packaging, etc.), and custom silicon suppliers like AVGO/MRVL only report their silicon revenue.\n\nWe expect Nvidia (NVDA) to maintain approximately 80% or more of the overall AI accelerator market over time, with AMD holding around 5%, and the broader ASIC ecosystem accounting for 10-15%. Therefore, we believe that the consensus enterprise accelerator revenue forecasts (below) are generally conservative and have upside potential for most suppliers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004579972911195829, "ret_signed_m1d": -0.004579972911195829, "alpha_spy_m1d": 0.0031830706195509206, "alpha_c_m1d": 0.0026350660159784045, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017610676115523027, "ret_signed_p1d": 0.017610676115523027, "alpha_spy_p1d": 0.012642344351517787, "alpha_c_p1d": 0.013209516726474968, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": 0.027867425132180434, "ret_signed_p1w": 0.027867425132180434, "alpha_spy_p1w": 0.005852870620161044, "alpha_c_p1w": 0.004707252880345791, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": 0.11921052616034622, "ret_signed_p1m": 0.11921052616034622, "alpha_spy_p1m": 0.07797622950100314, "alpha_c_p1m": 0.0820894032236279, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.15107767485204815, "ret_signed_p3m": 0.15107767485204815, "alpha_spy_p3m": 0.06416577049253513, "alpha_c_p3m": -0.009456251059873022, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.16765770446898332, "ret_signed_p6m": 0.16765770446898332, "alpha_spy_p6m": 0.04896650397323654, "alpha_c_p6m": -0.11744328928990244, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3009, -0.2895, -0.2878, -0.3434, -0.3917, -0.3702, -0.3788, -0.2625, -0.3061, -0.2845, -0.2859, -0.2763, -0.326, -0.3454, -0.3454, -0.3749, -0.3621, -0.3375, -0.3135, -0.2839, -0.2987, -0.2968, -0.2974, -0.2801, -0.2614, -0.2658, -0.2676, -0.2449, -0.2429, -0.2476, -0.2066, -0.1619, -0.127, -0.1303, -0.1266, -0.1255, -0.1332, -0.1498, -0.1432, -0.1531, -0.1531, -0.126, -0.1304, -0.1022, -0.1284, -0.1139, -0.0891, -0.0846, -0.097, -0.0859, -0.08, -0.0714, -0.0786, -0.0646, -0.0842, -0.0666, -0.0703, -0.0615, -0.0615, -0.0721, -0.07, -0.0459, -0.0046, 0.0, 0.0176, 0.0192, -0.0111, 0.0144, 0.0279, 0.0279, 0.0208, 0.0321, 0.0507, 0.0586, 0.0639, 0.0584, 0.1011, 0.1055, 0.116, 0.1122, 0.1055, 0.0775, 0.1017, 0.1208, 0.1192, 0.1402, 0.1322, 0.1564, 0.1474, 0.1206, 0.1611, 0.1499, 0.1574, 0.1661, 0.1786, 0.1744, 0.1815, 0.1714, 0.1742, 0.164, 0.1741, 0.133, 0.1315, 0.1288, 0.1482, 0.1599, 0.1726, 0.1715, 0.1622, 0.1236, 0.1236, 0.1017, 0.1006, 0.1073, 0.0774, 0.0857, 0.1015, 0.1439, 0.1429, 0.1471, 0.1467, 0.1282, 0.0986, 0.1369, 0.1397, 0.1845, 0.1511, 0.1417, 0.1463, 0.1495, 0.1731, 0.2037, 0.2079, 0.2186, 0.2104, 0.1969, 0.1937, 0.22, 0.2423, 0.1816, 0.2149, 0.1614, 0.1601, 0.1729, 0.182, 0.1782, 0.1687, 0.163, 0.1751, 0.2016, 0.2353, 0.2969, 0.3356, 0.3089, 0.3063, 0.3346, 0.2818, 0.2593, 0.2133, 0.2138, 0.2841, 0.2461, 0.2502, 0.2055, 0.2268, 0.2038, 0.17, 0.2033, 0.1653, 0.154, 0.1777, 0.1471, 0.1629, 0.1629, 0.1419, 0.1607, 0.1706, 0.1586, 0.1831, 0.1768, 0.1971, 0.1933, 0.1857, 0.1673, 0.1291, 0.1373, 0.1466, 0.1028, 0.1235, 0.1677]}
{"unit_id": "orig:1938233963790930222", "ticker_idx": 1, "ts": "2025-06-26T13:52:42+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD holding around 5%", "tweet_type": "analysis", "timeline": "years", "summary": "BofA adopted: AI accelerator TAM grows to $650B by 2030; NVDA holds 80%+ share, AMD ~5%, AVGO/MRVL ASIC ecosystem 10-15%; consensus estimates conservative with upside for most suppliers.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AI Accelerators (GPU, ASIC, XPU) 2\n\nWhile not all computers (including servers) need accelerators, all AI servers must be equipped with accelerators like GPUs, ASICs, and XPUs to be competitive in AI workloads.\n\nTypes of Accelerators\n\nThere are two main types of accelerators/XPUs: GPUs and ASICs.\n\nGPUs refer to general-purpose or commercial accelerator solutions, designed by Nvidia (NVDA) and AMD (and to some extent, Intel (INTC)).\n\nConsidering the continuously evolving nature of AI workloads and the excellent product roadmaps/cycles of the two GPU suppliers, we expect GPUs to maintain 80-85% of the total AI accelerator market over time.\n\nASICs generally refer to semi-custom-made accelerators designed by AVGO (Broadcom) and MRVL (Marvell) (and Asia-based suppliers like Alchip, GUC). Many hyperscalers with the resources to develop their own accelerators often take this path, collaborating with existing SerDes/IP/design suppliers like AVGO/MRVL. We believe a market for semi-custom accelerators (10-15% of the total market over time) exists for more specialized and repetitive workloads, such as Gmail search and Instagram algorithms, because they can be processed more efficiently on ASICs than on power-intensive GPUs.\n\nTotal Addressable Market (TAM)\n\nWe believe the AI accelerator market can grow from $126 billion in 2024 to over $400 billion in 2027, over $500 billion in 2028, and over $650 billion in 2030, with a Compound Annual Growth Rate (CAGR) of +26% from 2025-30. For reference, this is the total system market size (including accelerator silicon, HBM, software, packaging, etc.), and custom silicon suppliers like AVGO/MRVL only report their silicon revenue.\n\nWe expect Nvidia (NVDA) to maintain approximately 80% or more of the overall AI accelerator market over time, with AMD holding around 5%, and the broader ASIC ecosystem accounting for 10-15%. Therefore, we believe that the consensus enterprise accelerator revenue forecasts (below) are generally conservative and have upside potential for most suppliers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0019487666590344288, "ret_signed_m1d": -0.0019487666590344288, "alpha_spy_m1d": 0.0058142768717123205, "alpha_c_m1d": 0.005266272268139804, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0009048224487724355, "ret_signed_p1d": 0.0009048224487724355, "alpha_spy_p1d": -0.004063509315232805, "alpha_c_p1d": -0.003496336940275624, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.04015861155207634, "ret_signed_p1w": -0.04015861155207634, "alpha_spy_p1w": -0.06217316606409573, "alpha_c_p1w": -0.06331878380391098, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": 0.15861643726797992, "ret_signed_p1m": 0.15861643726797992, "alpha_spy_p1m": 0.11738214060863683, "alpha_c_p1m": 0.1214953143312616, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.11984968052970779, "ret_signed_p3m": 0.11984968052970779, "alpha_spy_p3m": 0.03293777617019478, "alpha_c_p3m": -0.04068424538221338, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.48545381093798645, "ret_signed_p6m": 0.48545381093798645, "alpha_spy_p6m": 0.3667626104422397, "alpha_c_p6m": 0.2003528171791007, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.2849, -0.2847, -0.2834, -0.3472, -0.4031, -0.4179, -0.4557, -0.326, -0.3827, -0.3499, -0.3423, -0.3368, -0.3855, -0.391, -0.391, -0.4045, -0.3996, -0.3709, -0.3425, -0.3273, -0.3291, -0.3314, -0.3225, -0.3273, -0.3124, -0.2999, -0.3136, -0.3015, -0.2922, -0.2842, -0.2475, -0.2173, -0.1807, -0.1997, -0.1845, -0.2014, -0.21, -0.2201, -0.2295, -0.2323, -0.2323, -0.2027, -0.2145, -0.2133, -0.2293, -0.2022, -0.1835, -0.1747, -0.1948, -0.1913, -0.1528, -0.1423, -0.1569, -0.1753, -0.1915, -0.1203, -0.1154, -0.1176, -0.1176, -0.1075, -0.0981, -0.0365, -0.0019, 0.0, 0.0009, -0.0124, -0.0527, -0.0359, -0.0402, -0.0402, -0.0618, -0.0408, -0.0367, 0.0033, 0.0191, 0.0178, 0.083, 0.1141, 0.1164, 0.0926, 0.0927, 0.0768, 0.1042, 0.1283, 0.1586, 0.2087, 0.235, 0.2494, 0.2271, 0.195, 0.2304, 0.2132, 0.1353, 0.1999, 0.2024, 0.1991, 0.2176, 0.2835, 0.2594, 0.2355, 0.2259, 0.1592, 0.1498, 0.1394, 0.1676, 0.137, 0.1597, 0.1632, 0.1733, 0.1319, 0.1319, 0.1297, 0.1284, 0.126, 0.0519, 0.0538, 0.0845, 0.1104, 0.0834, 0.1036, 0.1217, 0.1168, 0.1077, 0.0991, 0.0954, 0.1121, 0.1198, 0.1197, 0.1224, 0.1098, 0.1231, 0.126, 0.1415, 0.1813, 0.1461, 0.4178, 0.4721, 0.6395, 0.6209, 0.4957, 0.5063, 0.5179, 0.6606, 0.6325, 0.6222, 0.6743, 0.6567, 0.6024, 0.6355, 0.7603, 0.8073, 0.7957, 0.8397, 0.7737, 0.7826, 0.8071, 0.7403, 0.784, 0.6544, 0.6254, 0.6981, 0.6531, 0.8019, 0.7258, 0.7178, 0.674, 0.6028, 0.5559, 0.4339, 0.4183, 0.4967, 0.4346, 0.4911, 0.4911, 0.514, 0.5295, 0.4981, 0.5145, 0.5032, 0.5171, 0.5389, 0.5425, 0.5411, 0.5411, 0.467, 0.4447, 0.4558, 0.3788, 0.3994, 0.4855]}
{"unit_id": "orig:1938233963790930222", "ticker_idx": 2, "ts": "2025-06-26T13:52:42+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "consensus enterprise accelerator revenue forecasts (below) are generally conservative and have upside potential for most suppliers", "tweet_type": "analysis", "timeline": "years", "summary": "BofA adopted: AI accelerator TAM grows to $650B by 2030; NVDA holds 80%+ share, AMD ~5%, AVGO/MRVL ASIC ecosystem 10-15%; consensus estimates conservative with upside for most suppliers.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AI Accelerators (GPU, ASIC, XPU) 2\n\nWhile not all computers (including servers) need accelerators, all AI servers must be equipped with accelerators like GPUs, ASICs, and XPUs to be competitive in AI workloads.\n\nTypes of Accelerators\n\nThere are two main types of accelerators/XPUs: GPUs and ASICs.\n\nGPUs refer to general-purpose or commercial accelerator solutions, designed by Nvidia (NVDA) and AMD (and to some extent, Intel (INTC)).\n\nConsidering the continuously evolving nature of AI workloads and the excellent product roadmaps/cycles of the two GPU suppliers, we expect GPUs to maintain 80-85% of the total AI accelerator market over time.\n\nASICs generally refer to semi-custom-made accelerators designed by AVGO (Broadcom) and MRVL (Marvell) (and Asia-based suppliers like Alchip, GUC). Many hyperscalers with the resources to develop their own accelerators often take this path, collaborating with existing SerDes/IP/design suppliers like AVGO/MRVL. We believe a market for semi-custom accelerators (10-15% of the total market over time) exists for more specialized and repetitive workloads, such as Gmail search and Instagram algorithms, because they can be processed more efficiently on ASICs than on power-intensive GPUs.\n\nTotal Addressable Market (TAM)\n\nWe believe the AI accelerator market can grow from $126 billion in 2024 to over $400 billion in 2027, over $500 billion in 2028, and over $650 billion in 2030, with a Compound Annual Growth Rate (CAGR) of +26% from 2025-30. For reference, this is the total system market size (including accelerator silicon, HBM, software, packaging, etc.), and custom silicon suppliers like AVGO/MRVL only report their silicon revenue.\n\nWe expect Nvidia (NVDA) to maintain approximately 80% or more of the overall AI accelerator market over time, with AMD holding around 5%, and the broader ASIC ecosystem accounting for 10-15%. Therefore, we believe that the consensus enterprise accelerator revenue forecasts (below) are generally conservative and have upside potential for most suppliers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020431482389267952, "ret_signed_m1d": -0.020431482389267952, "alpha_spy_m1d": -0.012668438858521203, "alpha_c_m1d": -0.013216443462093719, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0030349522206614132, "ret_signed_p1d": -0.0030349522206614132, "alpha_spy_p1d": -0.008003283984666654, "alpha_c_p1d": -0.007436111609709473, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": 0.018543938644319802, "ret_signed_p1w": 0.018543938644319802, "alpha_spy_p1w": -0.003470615867699589, "alpha_c_p1w": -0.004616233607514841, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": 0.07406442187369522, "ret_signed_p1m": 0.07406442187369522, "alpha_spy_p1m": 0.032830125214352135, "alpha_c_p1m": 0.0369432989369769, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.25669299836579507, "ret_signed_p3m": 0.25669299836579507, "alpha_spy_p3m": 0.16978109400628205, "alpha_c_p3m": 0.0961590724538739, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.26195789919732526, "ret_signed_p6m": 0.26195789919732526, "alpha_spy_p6m": 0.1432666987015785, "alpha_c_p6m": -0.023143094561560495, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3817, -0.3777, -0.3645, -0.4313, -0.4598, -0.4308, -0.4238, -0.3163, -0.3638, -0.3282, -0.3414, -0.3392, -0.3552, -0.3686, -0.3686, -0.3862, -0.3738, -0.3467, -0.3052, -0.2899, -0.2893, -0.2941, -0.2893, -0.2713, -0.248, -0.2588, -0.2611, -0.2437, -0.2328, -0.2312, -0.1818, -0.1417, -0.1429, -0.1409, -0.1558, -0.1484, -0.1445, -0.1517, -0.1487, -0.1554, -0.1554, -0.1298, -0.1159, -0.1065, -0.1061, -0.0816, -0.0515, -0.0359, -0.0402, -0.0882, -0.098, -0.0967, -0.0661, -0.0544, -0.0816, -0.0691, -0.0792, -0.0722, -0.0722, -0.0747, -0.0607, -0.0237, -0.0204, 0.0, -0.003, 0.0203, -0.0201, -0.001, 0.0185, 0.0185, 0.0148, 0.006, 0.0286, 0.0194, 0.0156, 0.0201, 0.0399, 0.0394, 0.0603, 0.0487, 0.0668, 0.0312, 0.05, 0.0686, 0.0741, 0.0893, 0.1009, 0.1201, 0.0871, 0.0684, 0.102, 0.0842, 0.1166, 0.1243, 0.1288, 0.1248, 0.1579, 0.1441, 0.152, 0.1339, 0.1317, 0.0916, 0.0777, 0.0719, 0.0882, 0.0891, 0.103, 0.1113, 0.1424, 0.1008, 0.1008, 0.1039, 0.1193, 0.133, 0.2396, 0.2794, 0.2461, 0.3679, 0.3311, 0.332, 0.3476, 0.3325, 0.2813, 0.2783, 0.2768, 0.2561, 0.2567, 0.2581, 0.2462, 0.2403, 0.2158, 0.2232, 0.2361, 0.2539, 0.2546, 0.2439, 0.2473, 0.281, 0.2792, 0.2036, 0.3225, 0.2759, 0.3026, 0.3131, 0.2952, 0.2949, 0.2705, 0.2617, 0.2765, 0.313, 0.3424, 0.3829, 0.4311, 0.3958, 0.3705, 0.3442, 0.3049, 0.331, 0.3184, 0.2956, 0.3288, 0.305, 0.3171, 0.2605, 0.2697, 0.2704, 0.2625, 0.3141, 0.2859, 0.2614, 0.4014, 0.4276, 0.4741, 0.4741, 0.4941, 0.4315, 0.4148, 0.4112, 0.4128, 0.4469, 0.4872, 0.5064, 0.5312, 0.5067, 0.3345, 0.2599, 0.2654, 0.2088, 0.2231, 0.262]}
{"unit_id": "orig:1938233963790930222", "ticker_idx": 3, "ts": "2025-06-26T13:52:42+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "consensus enterprise accelerator revenue forecasts (below) are generally conservative and have upside potential for most suppliers", "tweet_type": "analysis", "timeline": "years", "summary": "BofA adopted: AI accelerator TAM grows to $650B by 2030; NVDA holds 80%+ share, AMD ~5%, AVGO/MRVL ASIC ecosystem 10-15%; consensus estimates conservative with upside for most suppliers.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AI Accelerators (GPU, ASIC, XPU) 2\n\nWhile not all computers (including servers) need accelerators, all AI servers must be equipped with accelerators like GPUs, ASICs, and XPUs to be competitive in AI workloads.\n\nTypes of Accelerators\n\nThere are two main types of accelerators/XPUs: GPUs and ASICs.\n\nGPUs refer to general-purpose or commercial accelerator solutions, designed by Nvidia (NVDA) and AMD (and to some extent, Intel (INTC)).\n\nConsidering the continuously evolving nature of AI workloads and the excellent product roadmaps/cycles of the two GPU suppliers, we expect GPUs to maintain 80-85% of the total AI accelerator market over time.\n\nASICs generally refer to semi-custom-made accelerators designed by AVGO (Broadcom) and MRVL (Marvell) (and Asia-based suppliers like Alchip, GUC). Many hyperscalers with the resources to develop their own accelerators often take this path, collaborating with existing SerDes/IP/design suppliers like AVGO/MRVL. We believe a market for semi-custom accelerators (10-15% of the total market over time) exists for more specialized and repetitive workloads, such as Gmail search and Instagram algorithms, because they can be processed more efficiently on ASICs than on power-intensive GPUs.\n\nTotal Addressable Market (TAM)\n\nWe believe the AI accelerator market can grow from $126 billion in 2024 to over $400 billion in 2027, over $500 billion in 2028, and over $650 billion in 2030, with a Compound Annual Growth Rate (CAGR) of +26% from 2025-30. For reference, this is the total system market size (including accelerator silicon, HBM, software, packaging, etc.), and custom silicon suppliers like AVGO/MRVL only report their silicon revenue.\n\nWe expect Nvidia (NVDA) to maintain approximately 80% or more of the overall AI accelerator market over time, with AMD holding around 5%, and the broader ASIC ecosystem accounting for 10-15%. Therefore, we believe that the consensus enterprise accelerator revenue forecasts (below) are generally conservative and have upside potential for most suppliers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05051892813237813, "ret_signed_m1d": -0.05051892813237813, "alpha_spy_m1d": -0.04275588460163138, "alpha_c_m1d": -0.043303889205203894, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03513814949480876, "ret_signed_p1d": -0.03513814949480876, "alpha_spy_p1d": -0.040106481258814, "alpha_c_p1d": -0.039539308883856816, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.05989744935867347, "ret_signed_p1w": -0.05989744935867347, "alpha_spy_p1w": -0.08191200387069286, "alpha_c_p1w": -0.08305762161050811, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": -0.07126744392607809, "ret_signed_p1m": -0.07126744392607809, "alpha_spy_p1m": -0.11250174058542117, "alpha_c_p1m": -0.10838856686279641, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": -0.06613637239805326, "ret_signed_p3m": -0.06613637239805326, "alpha_spy_p3m": -0.15304827675756627, "alpha_c_p3m": -0.22667029830997443, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.053076811006233626, "ret_signed_p6m": 0.053076811006233626, "alpha_spy_p6m": -0.06561438948951315, "alpha_c_p6m": -0.23202418275265213, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.231, -0.2172, -0.2102, -0.305, -0.3826, -0.363, -0.3751, -0.2386, -0.3396, -0.3324, -0.3465, -0.3333, -0.3506, -0.3535, -0.3535, -0.3825, -0.367, -0.3276, -0.2831, -0.2632, -0.266, -0.2662, -0.2701, -0.2378, -0.2206, -0.225, -0.2345, -0.2959, -0.2794, -0.2541, -0.1934, -0.1812, -0.1759, -0.1847, -0.2027, -0.2177, -0.232, -0.2483, -0.2266, -0.2411, -0.2411, -0.202, -0.1923, -0.2031, -0.2473, -0.2313, -0.2202, -0.1709, -0.1852, -0.1453, -0.1354, -0.1392, -0.1467, -0.1292, -0.1598, -0.1194, -0.1248, -0.0628, -0.0628, -0.0808, -0.1149, -0.0595, -0.0505, 0.0, -0.0351, -0.0321, -0.0466, -0.0715, -0.0599, -0.0599, -0.1053, -0.1003, -0.0964, -0.0827, -0.09, -0.0925, -0.0938, -0.1133, -0.0988, -0.0658, -0.0857, -0.0991, -0.083, -0.0734, -0.0713, -0.05, -0.0446, 0.023, 0.0058, -0.0683, -0.0422, -0.041, -0.0574, -0.0507, -0.0321, -0.0328, -0.0262, -0.0073, -0.0108, -0.0465, -0.0396, -0.098, -0.1087, -0.1088, -0.0864, -0.087, -0.0706, -0.064, -0.0335, -0.2132, -0.2132, -0.1915, -0.2202, -0.1978, -0.2074, -0.174, -0.1635, -0.1602, -0.1666, -0.1571, -0.1561, -0.1382, -0.1117, -0.071, -0.0706, -0.0547, -0.0661, 0.0023, 0.0489, 0.0409, 0.0311, 0.0521, 0.0499, 0.0788, 0.079, 0.1128, 0.0884, 0.1576, 0.1349, 0.0721, 0.1194, 0.0798, 0.1132, 0.1049, 0.1014, 0.075, 0.0552, 0.015, 0.0365, 0.0536, 0.1109, 0.1078, 0.129, 0.1092, 0.1739, 0.1317, 0.0969, 0.1634, 0.1688, 0.1386, 0.1675, 0.1187, 0.1187, 0.096, 0.0826, 0.0451, -0.0147, 0.0184, -0.0397, -0.0301, 0.0493, 0.0448, 0.0985, 0.0985, 0.1196, 0.1409, 0.1633, 0.2548, 0.2297, 0.2387, 0.1521, 0.1133, 0.158, 0.12, 0.0573, 0.0552, 0.0528, 0.0231, 0.0578, 0.0531]}
{"unit_id": "orig:1941270466683077002", "ticker_idx": 0, "ts": "2025-07-04T22:58:40+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our 'Buy' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs reiterates Buy on ASML ahead of Q2 2025 earnings; bullish on 2026 growth drivers including leading-edge logic, HBM demand, and DUV recovery.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) ASML: Q2 2025 Earnings Preview\n\nASML is scheduled to announce its Q2 2025 earnings on July 16, and we've summarized key themes and our latest outlook on the company. Our main areas of focus include: (1) key drivers supporting ASML's growth expectations for 2026; (2) the extent to which artificial intelligence (AI) will continue to drive demand for cutting-edge logic/memory chips; (3) the order intake run-rate needed in the coming quarters to meet 2026 market consensus estimates; and (4) the future direction of lithography technology advancements beyond the Gate-All-Around (GAA) architecture transition. We maintain our 'Buy' rating.\n\n1. Key Drivers Supporting ASML's Growth Expectations for 2026\n\nWe believe investors are currently focused on the key positive and negative factors supporting ASML's growth expectations for 2026 (for reference, Visible Alpha Consensus Data currently anticipates 8% year-over-year revenue growth). In a recent roadshow we hosted, ASML indicated potential for growth in 2026 driven by: (1) leading-edge logic; (2) advanced memory applications such as High Bandwidth Memory (HBM); and (3) a recovery in investments for trailing-edge processes.\n\n1. Accordingly, we maintain a positive view on ASML's 2026 growth opportunities. This is based on our expectation that leading-edge demand will remain robust due to TSMC's 2nm node expansion, which will be a critical node in multi-year capacity build-outs. Furthermore, we see upside potential to 2026 estimates if other major leading-edge players resolve technical issues and resume capacity expansions.\n\n2. Separately, demand for advanced memory, such as HBM and DDR5, remains robust and could lead to further capacity expansion next year, driving front-end semiconductor equipment demand. In this context, Micron recently stated in its earnings call that a significant portion of its FY25 CapEx is allocated to HBM capacity expansion and that it will continue to expand greenfield capacity by installing equipment for new nodes in the future.\n\n3. In addition, recent data from analog/power semiconductor companies suggest relatively healthy demand outlooks and inventory levels for Western companies, while demand from China continues to be strong. For example, Microchip raised its Q1 FY26 guidance based on better-than-expected performance, with May order intake marking the highest in the past two years. We expect a recovery in investments in trailing-edge capacity to support demand for ASML's (higher-margin) Deep Ultraviolet (DUV) equipment going forward.\n\n4. Finally, investors will likely seek further clarification on potential risk factors related to China revenue. However, we note that current investments by major companies in China are now diversified across a greater number of customers than in the past.\n\n2. To What Extent Will AI Continue to Drive Demand for Cutting-Edge Logic/Memory Chips?\n\nIn our view, after very strong build-outs over the past two years, investors are now keen to understand the sustainability of AI-related semiconductor demand. While acknowledging the significant investments in leading-edge semiconductors in recent years, we believe AI can provide a long-term tailwind for cutting-edge semiconductor chips.\n\n1. Firstly, our conviction in the sustainability of AI demand is supported by positive data from the AI ecosystem (see Chart 1), which highlights the diversification of AI applications and the broader adoption of AI. For instance, our TSMC analyst recently made positive revisions to wafer revenue forecasts for 3nm/5nm processes, reflecting a reduction in near-term AI order-related risks.\n\n2. Furthermore, Nvidia stated during its Q1 FY26 earnings call that Microsoft processed approximately 100 trillion tokens in that quarter, a five-fold increase year-over-year. We view the aforementioned as largely consistent with our view that the proliferation of inference models (compared to traditional single inference approaches) will spur computing demand and, consequently, drive demand for cutting-edge semiconductors to power these models.\n\n3. More broadly, we maintain a positive outlook on leading-edge demand for AI applications and believe ASML is well-positioned to benefit from this trend, given its status as the sole supplier of lithography equipment enabling the incremental node shrinkage necessary to make chip performance even more powerful for complex AI workloads.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026842579922585763, "ret_signed_m1d": 0.026842579922585763, "alpha_spy_m1d": 0.026842579922585763, "alpha_c_m1d": 0.026842579922585763, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018501606746225585, "ret_signed_p1d": 0.018501606746225585, "alpha_spy_p1d": 0.02595362382952293, "alpha_c_p1d": 0.03211135163960299, "bench_spy_p1d": -0.0074520170832973465, "bench_c_p1d": -0.013609744893377407, "ret_p1w": 0.0371550151363127, "ret_signed_p1w": 0.0371550151363127, "alpha_spy_p1w": 0.039905572083737795, "alpha_c_p1w": 0.023510083420335448, "bench_spy_p1w": -0.002750556947425098, "bench_c_p1w": 0.013644931715977249, "ret_p1m": -0.08401936417970102, "ret_signed_p1m": -0.08401936417970102, "alpha_spy_p1m": -0.09334216158588571, "alpha_c_p1m": -0.1071841663045533, "bench_spy_p1m": 0.00932279740618469, "bench_c_p1m": 0.02316480212485228, "ret_p3m": 0.2793926516061771, "ret_signed_p3m": 0.2793926516061771, "alpha_spy_p3m": 0.2074905495424424, "alpha_c_p3m": 0.10285367065109874, "bench_spy_p3m": 0.07190210206373471, "bench_c_p3m": 0.17653898095507836, "ret_p6m": 0.3812703082515716, "ret_signed_p6m": 0.3812703082515716, "alpha_spy_p6m": 0.27499992625941205, "alpha_c_p6m": 0.09312682113815174, "bench_spy_p6m": 0.10627038199215955, "bench_c_p6m": 0.28814348711341986, "price_path": [-0.1325, -0.162, -0.1268, -0.1272, -0.108, -0.0847, -0.1322, -0.147, -0.147, -0.147, -0.1517, -0.1218, -0.1055, -0.1069, -0.1136, -0.1128, -0.1166, -0.1166, -0.0766, -0.0798, -0.0878, -0.083, -0.0438, -0.0494, 0.0114, 0.0369, 0.0412, 0.0346, 0.0136, 0.0045, 0.0009, 0.008, -0.0039, -0.0255, -0.0082, 0.0102, 0.0021, 0.0093, -0.0083, -0.02, -0.0096, -0.0106, -0.0006, 0.0124, 0.0314, 0.0381, 0.0467, 0.0261, 0.0079, 0.0162, 0.01, 0.0085, -0.0096, -0.0136, 0.0141, 0.0485, 0.0532, 0.0273, 0.035, 0.0276, 0.0133, 0.0276, 0.0268, 0.0, 0.0185, 0.0296, 0.0309, 0.0464, 0.0372, 0.0423, 0.0708, -0.051, -0.014, -0.0393, -0.0535, -0.0864, -0.0801, -0.066, -0.0827, -0.0446, -0.0527, -0.0459, -0.0677, -0.0953, -0.084, -0.0921, -0.0985, -0.0714, -0.062, -0.0466, -0.0361, -0.0259, -0.0224, -0.0319, -0.0288, -0.0233, -0.0296, -0.0352, -0.0174, -0.0167, -0.0107, 0.0039, -0.0051, -0.032, -0.0328, -0.0615, -0.0492, -0.0153, 0.0039, 0.0288, 0.0388, 0.0343, 0.0431, 0.0489, 0.1091, 0.1144, 0.1206, 0.2073, 0.2069, 0.2329, 0.2465, 0.2286, 0.2342, 0.2359, 0.2552, 0.2592, 0.2794, 0.3344, 0.3382, 0.3644, 0.3302, 0.2952, 0.2922, 0.2376, 0.2833, 0.2873, 0.3274, 0.3341, 0.3315, 0.367, 0.353, 0.323, 0.3498, 0.36, 0.3808, 0.3822, 0.4005, 0.428, 0.3985, 0.4113, 0.3983, 0.381, 0.3638, 0.3264, 0.3507, 0.351, 0.3584, 0.3427, 0.3315, 0.335, 0.3185, 0.3505, 0.3554, 0.2704, 0.3082, 0.3104, 0.3851, 0.3666, 0.3761, 0.4116, 0.4305, 0.4675, 0.4582, 0.4495, 0.4672, 0.4515, 0.441, 0.4332, 0.4095, 0.4181, 0.384, 0.3313, 0.3598, 0.3733, 0.3648, 0.3758, 0.3694, 0.3694, 0.3694, 0.3813]}
{"unit_id": "orig:1927707944944066900", "ticker_idx": 0, "ts": "2025-05-28T12:46:03+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "more investors see any pullback as a buying opportunity, noting smoother 2H dynamics and a ramp in GB300 production", "tweet_type": "analysis", "timeline": "months", "summary": "Relays JPM investor sentiment on NVDA: buy-side views any pullback as opportunity on 2H/GB300 ramp, though miss risk and slowing flows flag potential 15–20% downside.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan Investor Sentiment Analysis: The Difficulty of Interpreting NVIDIA (NVDA) Guidance\n\nInvestor sentiment: As evidenced by somewhat confusing buy-side (institutional investor) survey results, this is a particularly difficult quarter to interpret NVIDIA (NVDA)’s guidance. In principle, the China export restrictions on the H20 chip that took effect on April 9 should result in an annual revenue exclusion of approximately $15–16 billion. (For reference, AMD’s case suggests a $7.6 billion impact on both the July and October quarters respectively.) Nevertheless, it appears that most sell-side analysts have not updated their models to reflect the H20-related loss, and many investors we’ve spoken with also indicate they have not incorporated it yet. This is highlighted as a potential risk factor that could drive volatility around the earnings release.\n\nIn fact, buy-side consensus is that July quarter guidance will come in approximately $920 million below prior expectations—not the previously cited $7 billion impact. Nearly everyone I’ve spoken to expects Q2 (July quarter) revenue guidance in the range of $45–46 billion. There is also concern over Q1 margin guidance, which will be heavily affected by amortization. About one-third of our survey respondents expect Q1 gross margin (GM) to be around 58% (vs. the sell-side consensus of 71.1%), while the remaining two-thirds are clustered between 65–72%. Reconciling these views is challenging, but here are a few key takeaways from our investor conversations:\n\n1. Many investors are comforted by the view that NVIDIA will ship inventory in anticipation of a strong 2H (i.e., “fill the channel to smooth volatility”). Last Friday’s news involving Oracle (ORCL) highlighted the possibility of unexpected demand upside. In addition, there is growing optimism around ODM yield improvements (Albert Wang put out a great note on this last week). There’s no dispute here—everyone is watching those receivables, and backend yields are improving quickly.\n\n2. Some investors expect NVIDIA to sell some form of chips to China this year—such as the B30 or a custom RTX Pro 6000 server SKU—retaining all or part of the $15–16 billion China revenue as a temporary item. Others suggest that not all of the H20 revenue is lost—some say “none of it”—and that the company may recover a portion of it.\n\n3. One investor proposed that some investors who missed the recent 35% rally are now taking short positions and intentionally circulating higher guidance expectations to set the stage for a disappointment. Short interest has indeed risen in recent weeks (although the name still has a skewed long bias). However, more investors see any pullback as a buying opportunity, noting smoother 2H dynamics and a ramp in GB300 production. Gokul’s sourcing from Computex was positive for both the GB200 ramp and what appears to be a more seamless, higher-demand GB300 ramp.\n\n4. Lastly, one risk flagged by our Delta One desk is that beta appetite in the market appears to be waning, and both retail and institutional flows into NVIDIA are slowing. Historically, when both types of flows and sentiment stall or reverse, NVIDIA’s stock has experienced a 15–20% decline. Our team believes that if NVIDIA misses earnings tomorrow, this could be a meaningful downside risk.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "orig:1935314449495150943", "ticker_idx": 0, "ts": "2025-06-18T12:31:35+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "major investments focused on Blackwell (especially the GB200) and its supporting infrastructure", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi/Dell'Oro data: datacenter capex +50% YoY, 5M+ AI accelerators in 2025, Blackwell-focused; raises 2025 overall capex growth forecast to 30%.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Datacenter Capex: AI Drives Significant Growth in Datacenter Investment, High-End Accelerator Shipments Expected to Exceed 5 Million Units in 2025\n\nCiti (A. Malik, 06/17/25)\n\nThe institution, citing a Dell’Oro report, points out that global datacenter Capex is expected to continue its rapid growth in 2025. AI training remains the primary investment focus, with the top four cloud providers driving the deployment of millions of accelerators.\n\nGlobal Datacenter Capex in 1Q25 Grew Over 50% YoY, Totaling $134 Billion\n\nThe growth was mainly driven by accelerated investment in AI servers, which accounted for over half of the total Capex, surpassing general-purpose infrastructure. Due to anticipated tariff risks, system vendors and hyperscale cloud service providers are stockpiling in advance, leading to price increases for components like NICs and memory. The enterprise market also saw a 21% YoY growth, driven by the server replacement cycle.\n\nThe Top 11 Cloud Providers Hold 60% of the Global Market Share, with the Top 4 Growing 73% YoY\n\nIn 2025, the Capex for the Top 4 (US-based) providers is expected to grow 39% YoY, leading the entire industry into a multi-year investment cycle expansion. The top four cloud providers plan to deploy over 5 million AI accelerators in 2025, with major investments focused on Blackwell (especially the GB200) and its supporting infrastructure.\n\nMajor Cloud Providers are Actively Expanding Datacenters with a Clear Cadence for New Platform Deployment\n\nMicrosoft plans a large-scale deployment of its Maia platform in the second half of 2025. META expects to expand to 14 new regions within the next 2-4 years. Amazon and Google are benefiting from cloud service growth, which is driving their general-purpose server deployments. Oracle plans to add datacenters in 7 new regions within the next 12 to 18 months, launching a four-year investment plan totaling $100 billion.\n\nASICs and Custom Accelerators Gain Attention; Apple Plans to Launch \"Baltra\" AI Training Chip in 2026\n\nShipments of custom accelerators like Google's TPU and Amazon's Trainium are expected to reach 2 million units in 2025. Apple is maintaining a hybrid Capex strategy and is rumored to be developing an AI training chip named \"Baltra.\" Enterprise customers, due to the lack of supporting infrastructure like liquid cooling for Blackwell and NVIDIA's one-year update cycle, are increasingly turning to public cloud AI solutions.\n\nHigh-End Accelerated Server Capex in 2025 is Expected to Reach $205 Billion, Accounting for 34% of Total Capex\n\nDell’Oro forecasts that high-end AI servers will ship with 5 million accelerators in 2025, corresponding to a server Capex of $205 billion. This represents 34% of the total datacenter Capex and excludes spending on networking and physical infrastructure. The institution (Citi) has raised its 2025 overall Capex growth forecast to 30%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009348349326185335, "ret_signed_m1d": -0.009348349326185335, "alpha_spy_m1d": -0.009499007774905421, "alpha_c_m1d": -0.00554017250628458, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.06069576878985283, "ret_signed_p1w": 0.06069576878985283, "alpha_spy_p1w": 0.04148919322022593, "alpha_c_p1w": 0.012674106505399019, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": 0.1891670177429181, "ret_signed_p1m": 0.1891670177429181, "alpha_spy_p1m": 0.13484090378747138, "alpha_c_p1m": 0.07827173362886541, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": 0.22188645019109177, "ret_signed_p3m": 0.22188645019109177, "alpha_spy_p3m": 0.11237966441262937, "alpha_c_p3m": 0.05211597835534909, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.24381553306802606, "ret_signed_p6m": 0.24381553306802606, "alpha_spy_p6m": 0.08365950547459011, "alpha_c_p6m": -0.1686145864326698, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [-0.191, -0.1655, -0.1705, -0.2181, -0.2341, -0.2462, -0.2551, -0.2429, -0.241, -0.3003, -0.3518, -0.3289, -0.3381, -0.2142, -0.2606, -0.2375, -0.2391, -0.2288, -0.2818, -0.3024, -0.3024, -0.3339, -0.3203, -0.294, -0.2685, -0.237, -0.2527, -0.2507, -0.2514, -0.2329, -0.213, -0.2177, -0.2196, -0.1954, -0.1933, -0.1982, -0.1546, -0.1069, -0.0698, -0.0733, -0.0694, -0.0682, -0.0764, -0.0941, -0.087, -0.0976, -0.0976, -0.0687, -0.0734, -0.0433, -0.0712, -0.0557, -0.0293, -0.0245, -0.0378, -0.0259, -0.0197, -0.0105, -0.0182, -0.0033, -0.0241, -0.0054, -0.0093, 0.0, 0.0, -0.0112, -0.009, 0.0166, 0.0607, 0.0656, 0.0843, 0.086, 0.0538, 0.0809, 0.0953, 0.0953, 0.0877, 0.0998, 0.1196, 0.128, 0.1336, 0.1278, 0.1734, 0.178, 0.1892, 0.1851, 0.178, 0.1481, 0.1739, 0.1943, 0.1926, 0.2149, 0.2064, 0.2323, 0.2226, 0.1941, 0.2373, 0.2253, 0.2333, 0.2426, 0.2558, 0.2514, 0.259, 0.2482, 0.2512, 0.2404, 0.2511, 0.2073, 0.2057, 0.2028, 0.2235, 0.236, 0.2495, 0.2483, 0.2385, 0.1973, 0.1973, 0.1739, 0.1728, 0.18, 0.1481, 0.1569, 0.1738, 0.2189, 0.2179, 0.2224, 0.2219, 0.2022, 0.1706, 0.2115, 0.2145, 0.2622, 0.2266, 0.2165, 0.2215, 0.2249, 0.2501, 0.2826, 0.2871, 0.2985, 0.2897, 0.2754, 0.272, 0.3, 0.3238, 0.2591, 0.2945, 0.2376, 0.2362, 0.2498, 0.2595, 0.2555, 0.2453, 0.2393, 0.2522, 0.2804, 0.3163, 0.3819, 0.4232, 0.3947, 0.392, 0.4221, 0.3658, 0.3419, 0.2929, 0.2934, 0.3683, 0.3278, 0.3322, 0.2845, 0.3073, 0.2827, 0.2467, 0.2822, 0.2418, 0.2297, 0.2549, 0.2224, 0.2391, 0.2391, 0.2167, 0.2368, 0.2474, 0.2345, 0.2607, 0.254, 0.2756, 0.2716, 0.2634, 0.2438]}
{"unit_id": "orig:1935314449495150943", "ticker_idx": 1, "ts": "2025-06-18T12:31:35+00:00", "symbol": "datacenter", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "high-end AI servers will ship with 5 million accelerators in 2025, corresponding to a server Capex of $205 billion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi/Dell'Oro data: datacenter capex +50% YoY, 5M+ AI accelerators in 2025, Blackwell-focused; raises 2025 overall capex growth forecast to 30%.", "resolved_tickers": ["DTCR"], "resolution_method": "etf", "resolution_reason": "US datacenter sector proxied by DTCR ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Datacenter Capex: AI Drives Significant Growth in Datacenter Investment, High-End Accelerator Shipments Expected to Exceed 5 Million Units in 2025\n\nCiti (A. Malik, 06/17/25)\n\nThe institution, citing a Dell’Oro report, points out that global datacenter Capex is expected to continue its rapid growth in 2025. AI training remains the primary investment focus, with the top four cloud providers driving the deployment of millions of accelerators.\n\nGlobal Datacenter Capex in 1Q25 Grew Over 50% YoY, Totaling $134 Billion\n\nThe growth was mainly driven by accelerated investment in AI servers, which accounted for over half of the total Capex, surpassing general-purpose infrastructure. Due to anticipated tariff risks, system vendors and hyperscale cloud service providers are stockpiling in advance, leading to price increases for components like NICs and memory. The enterprise market also saw a 21% YoY growth, driven by the server replacement cycle.\n\nThe Top 11 Cloud Providers Hold 60% of the Global Market Share, with the Top 4 Growing 73% YoY\n\nIn 2025, the Capex for the Top 4 (US-based) providers is expected to grow 39% YoY, leading the entire industry into a multi-year investment cycle expansion. The top four cloud providers plan to deploy over 5 million AI accelerators in 2025, with major investments focused on Blackwell (especially the GB200) and its supporting infrastructure.\n\nMajor Cloud Providers are Actively Expanding Datacenters with a Clear Cadence for New Platform Deployment\n\nMicrosoft plans a large-scale deployment of its Maia platform in the second half of 2025. META expects to expand to 14 new regions within the next 2-4 years. Amazon and Google are benefiting from cloud service growth, which is driving their general-purpose server deployments. Oracle plans to add datacenters in 7 new regions within the next 12 to 18 months, launching a four-year investment plan totaling $100 billion.\n\nASICs and Custom Accelerators Gain Attention; Apple Plans to Launch \"Baltra\" AI Training Chip in 2026\n\nShipments of custom accelerators like Google's TPU and Amazon's Trainium are expected to reach 2 million units in 2025. Apple is maintaining a hybrid Capex strategy and is rumored to be developing an AI training chip named \"Baltra.\" Enterprise customers, due to the lack of supporting infrastructure like liquid cooling for Blackwell and NVIDIA's one-year update cycle, are increasingly turning to public cloud AI solutions.\n\nHigh-End Accelerated Server Capex in 2025 is Expected to Reach $205 Billion, Accounting for 34% of Total Capex\n\nDell’Oro forecasts that high-end AI servers will ship with 5 million accelerators in 2025, corresponding to a server Capex of $205 billion. This represents 34% of the total datacenter Capex and excludes spending on networking and physical infrastructure. The institution (Citi) has raised its 2025 overall Capex growth forecast to 30%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003298739044428678, "ret_signed_m1d": -0.003298739044428678, "alpha_spy_m1d": -0.0034493974931487648, "alpha_c_m1d": -0.0034493974931487648, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": 0.00015065844872008682, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.017042176628836003, "ret_signed_p1w": 0.017042176628836003, "alpha_spy_p1w": -0.0021643989407909014, "alpha_c_p1w": -0.0021643989407909014, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.019206575569626905, "ret_p1m": 0.06387601731631132, "ret_signed_p1m": 0.06387601731631132, "alpha_spy_p1m": 0.009549903360864587, "alpha_c_p1m": 0.009549903360864587, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.05432611395544673, "ret_p3m": 0.09648318389955635, "ret_signed_p3m": 0.09648318389955635, "alpha_spy_p3m": -0.013023601878906055, "alpha_c_p3m": -0.013023601878906055, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.1095067857784624, "ret_p6m": 0.19706777808590803, "ret_signed_p6m": 0.19706777808590803, "alpha_spy_p6m": 0.036911750492472084, "alpha_c_p6m": 0.036911750492472084, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.16015602759343595, "price_path": [-0.0495, -0.0434, -0.0632, -0.0726, -0.0863, -0.105, -0.1061, -0.0951, -0.0874, -0.1083, -0.1666, -0.1781, -0.2067, -0.155, -0.1677, -0.1539, -0.1418, -0.1413, -0.1567, -0.1429, -0.1429, -0.1572, -0.1396, -0.1319, -0.1231, -0.1138, -0.1089, -0.0962, -0.0852, -0.0786, -0.0572, -0.0555, -0.055, -0.0693, -0.0709, -0.0682, -0.0544, -0.0583, -0.0555, -0.0517, -0.0341, -0.0319, -0.0341, -0.0401, -0.0473, -0.0484, -0.0484, -0.0374, -0.0484, -0.0434, -0.0533, -0.0357, -0.0308, -0.0104, -0.0093, -0.0077, -0.0049, 0.0033, 0.0011, 0.006, -0.0154, 0.0055, -0.0033, 0.0, 0.0, -0.0027, 0.0071, 0.0313, 0.017, 0.0126, 0.0224, 0.0302, 0.0202, 0.0324, 0.039, 0.039, 0.0175, 0.0197, 0.0235, 0.023, 0.0241, 0.0318, 0.0462, 0.0561, 0.0639, 0.0675, 0.0661, 0.0694, 0.0794, 0.0794, 0.0777, 0.07, 0.0639, 0.0545, 0.0534, 0.0357, 0.0528, 0.049, 0.0401, 0.0456, 0.0385, 0.0401, 0.0501, 0.0473, 0.0362, 0.0351, 0.0379, 0.0335, 0.0351, 0.0274, 0.0517, 0.0445, 0.0462, 0.0467, 0.0506, 0.0451, 0.0451, 0.0241, 0.013, 0.0125, 0.0252, 0.0235, 0.0396, 0.0595, 0.0854, 0.0899, 0.0965, 0.097, 0.1059, 0.118, 0.1153, 0.1407, 0.1357, 0.1319, 0.1324, 0.1258, 0.1225, 0.1319, 0.1506, 0.1672, 0.1683, 0.1805, 0.175, 0.1893, 0.1932, 0.1694, 0.1943, 0.1877, 0.2153, 0.2175, 0.2031, 0.2291, 0.2059, 0.1893, 0.2131, 0.2313, 0.2518, 0.2418, 0.238, 0.2324, 0.2358, 0.233, 0.1998, 0.2203, 0.2009, 0.2015, 0.2264, 0.2159, 0.1987, 0.1484, 0.1523, 0.1412, 0.1291, 0.1236, 0.0921, 0.0904, 0.1225, 0.123, 0.1352, 0.1352, 0.1556, 0.1473, 0.1484, 0.1589, 0.1645, 0.1799, 0.2009, 0.2037, 0.1965, 0.1971]}
{"unit_id": "thread:1927848466983833646", "ticker_idx": 0, "ts": "2025-05-28T22:04:26+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Both companies are likely to benefit significantly in terms of revenue.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Relays supply-chain report that NVDA and AMD will launch China-compliant AI GPUs from July 2025, with both set to benefit significantly in revenue; AMD seen as potentially bigger beneficiary.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD to Follow NVIDIA in Launching Downgraded AI Chips for China – Digitimes\n\nAccording to sources in the semiconductor supply chain, NVIDIA and AMD have swiftly adjusted their downgraded chip designs and are expected to launch a new series of AI GPUs that comply with export regulations to China starting from July 2025.\n\nReportedly, NVIDIA is preparing a product tentatively named the B20, while AMD plans to compete in the market with products such as the Radeon AI PRO R9700. These chips are capable of running models like DeepSeek, and interest from Chinese customers is said to be high. Both companies are likely to benefit significantly in terms of revenue.\n\nThe U.S. government has strengthened AI chip export restrictions citing national security concerns. Before leaving office, President Biden introduced a three-tier export control policy. Following his return to the presidency, Donald Trump immediately resumed trade wars, strictly enforced origin tracing rules, and pushed for a “Made in America” agenda, further tightening bans including the export of NVIDIA’s H20 chip to China.\n\nAlthough Trump recently withdrew the three-tier AI chip export control policy and warned that any company using Huawei’s Ascend AI chips would be in violation of U.S. export controls, NVIDIA—once holding over 90% market share in AI GPUs—has seen its share in China drop significantly following the H20 ban.\n\nIn response, NVIDIA CEO Jensen Huang has publicly stated multiple times that the Trump administration should reconsider its export control policies, arguing that these measures have not achieved their intended goals.\n\nFour years ago, when the Biden administration first took office, NVIDIA's market share in China was 90%, but it has now fallen to around 50%. The original intent of slowing down China's AI development has, in reality, backfired.\n\nChina currently holds 50% of the world’s AI researchers, making it the largest concentration of AI talent globally. The country remains steadfast in its commitment to advancing AI.\n\nJensen Huang recently emphasized the innovations and contributions of China’s DeepSeek, saying the model provides tangible global value and praising Alibaba’s Qwen3 model as well. He stated confidently that China’s AI research will continue to thrive.\n\nHe also warned that if NVIDIA were forced out of the Chinese market, other companies like Huawei would quickly fill the void.\n\nAll countries desire to lead in AI, and even Trump is reportedly aware that the previous administration’s AI containment policies have failed. Jensen Huang believes that export control and AI containment policies must be reversed together, and while that hasn’t happened yet, he expressed hope that changes will come.\n\nNVIDIA’s repeated emphasis on the importance of the Chinese market has fueled speculation that it will once again launch new downgraded chips designed specifically for China.\n\nJensen Huang stated that it is no longer feasible to redesign the H20 or existing products to comply with the restrictions, and as a result, NVIDIA decided to write down $5.5 billion worth of chip inventory. This figure surpasses the annual revenue of many semiconductor firms, representing not only a major revenue loss for NVIDIA but also a significant tax loss for the U.S. government.\n\nNVIDIA is currently working on new strategies. While there is no clear direction yet, the company’s commitment to the Chinese market remains unchanged.\n\nSources in the supply chain say that the Chinese market remains crucial for both NVIDIA and AMD. Although AMD has historically had a much smaller presence in the AI GPU space compared to NVIDIA, its shipments have recently grown noticeably.\n\nThe launch of downgraded chips could thus represent an even bigger opportunity for AMD, potentially allowing it to compete head-on with NVIDIA starting in July, with NVIDIA’s B20 (based on the Blackwell architecture) and AMD’s Radeon AI PRO R9700 both entering the Chinese market.\n\nThese chips will feature significant adjustments to memory and compute performance, and are designed for AI workstations, optimized for local inference, model fine-tuning, and other data-intensive workflows.\n\n---\n\nhttps://t.co/g3YPmrRr6K", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "thread:1927848466983833646", "ticker_idx": 1, "ts": "2025-05-28T22:04:26+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The launch of downgraded chips could thus represent an even bigger opportunity for AMD", "tweet_type": "news_reaction", "timeline": "months", "summary": "Relays supply-chain report that NVDA and AMD will launch China-compliant AI GPUs from July 2025, with both set to benefit significantly in revenue; AMD seen as potentially bigger beneficiary.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD to Follow NVIDIA in Launching Downgraded AI Chips for China – Digitimes\n\nAccording to sources in the semiconductor supply chain, NVIDIA and AMD have swiftly adjusted their downgraded chip designs and are expected to launch a new series of AI GPUs that comply with export regulations to China starting from July 2025.\n\nReportedly, NVIDIA is preparing a product tentatively named the B20, while AMD plans to compete in the market with products such as the Radeon AI PRO R9700. These chips are capable of running models like DeepSeek, and interest from Chinese customers is said to be high. Both companies are likely to benefit significantly in terms of revenue.\n\nThe U.S. government has strengthened AI chip export restrictions citing national security concerns. Before leaving office, President Biden introduced a three-tier export control policy. Following his return to the presidency, Donald Trump immediately resumed trade wars, strictly enforced origin tracing rules, and pushed for a “Made in America” agenda, further tightening bans including the export of NVIDIA’s H20 chip to China.\n\nAlthough Trump recently withdrew the three-tier AI chip export control policy and warned that any company using Huawei’s Ascend AI chips would be in violation of U.S. export controls, NVIDIA—once holding over 90% market share in AI GPUs—has seen its share in China drop significantly following the H20 ban.\n\nIn response, NVIDIA CEO Jensen Huang has publicly stated multiple times that the Trump administration should reconsider its export control policies, arguing that these measures have not achieved their intended goals.\n\nFour years ago, when the Biden administration first took office, NVIDIA's market share in China was 90%, but it has now fallen to around 50%. The original intent of slowing down China's AI development has, in reality, backfired.\n\nChina currently holds 50% of the world’s AI researchers, making it the largest concentration of AI talent globally. The country remains steadfast in its commitment to advancing AI.\n\nJensen Huang recently emphasized the innovations and contributions of China’s DeepSeek, saying the model provides tangible global value and praising Alibaba’s Qwen3 model as well. He stated confidently that China’s AI research will continue to thrive.\n\nHe also warned that if NVIDIA were forced out of the Chinese market, other companies like Huawei would quickly fill the void.\n\nAll countries desire to lead in AI, and even Trump is reportedly aware that the previous administration’s AI containment policies have failed. Jensen Huang believes that export control and AI containment policies must be reversed together, and while that hasn’t happened yet, he expressed hope that changes will come.\n\nNVIDIA’s repeated emphasis on the importance of the Chinese market has fueled speculation that it will once again launch new downgraded chips designed specifically for China.\n\nJensen Huang stated that it is no longer feasible to redesign the H20 or existing products to comply with the restrictions, and as a result, NVIDIA decided to write down $5.5 billion worth of chip inventory. This figure surpasses the annual revenue of many semiconductor firms, representing not only a major revenue loss for NVIDIA but also a significant tax loss for the U.S. government.\n\nNVIDIA is currently working on new strategies. While there is no clear direction yet, the company’s commitment to the Chinese market remains unchanged.\n\nSources in the supply chain say that the Chinese market remains crucial for both NVIDIA and AMD. Although AMD has historically had a much smaller presence in the AI GPU space compared to NVIDIA, its shipments have recently grown noticeably.\n\nThe launch of downgraded chips could thus represent an even bigger opportunity for AMD, potentially allowing it to compete head-on with NVIDIA starting in July, with NVIDIA’s B20 (based on the Blackwell architecture) and AMD’s Radeon AI PRO R9700 both entering the Chinese market.\n\nThese chips will feature significant adjustments to memory and compute performance, and are designed for AI workstations, optimized for local inference, model fine-tuning, and other data-intensive workflows.\n\n---\n\nhttps://t.co/g3YPmrRr6K", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015062882678083866, "ret_signed_m1d": 0.015062882678083866, "alpha_spy_m1d": 0.009243802409775403, "alpha_c_m1d": 0.004293802703780436, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0015062747477048166, "ret_signed_p1d": 0.0015062747477048166, "alpha_spy_p1d": -0.0024410025930914347, "alpha_c_p1d": -0.005590589248309907, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.050682271750568164, "ret_signed_p1w": 0.050682271750568164, "alpha_spy_p1w": 0.036730407337165705, "alpha_c_p1w": 0.01193826485302174, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.2730816223529495, "ret_signed_p1m": 0.2730816223529495, "alpha_spy_p1m": 0.22893070713900476, "alpha_c_p1m": 0.12932850722803768, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.4474570239845286, "ret_signed_p3m": 0.4474570239845286, "alpha_spy_p3m": 0.3510875338680577, "alpha_c_p3m": 0.2347155599636055, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.8254474850105991, "ret_signed_p6m": 0.8254474850105991, "alpha_spy_p6m": 0.7088233435636409, "alpha_c_p6m": 0.4840545514591028, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.1152, -0.1296, -0.1073, -0.0991, -0.1241, -0.1112, -0.1438, -0.1427, -0.1069, -0.1307, -0.1054, -0.0733, -0.0828, -0.0587, -0.0507, -0.0569, 0.0088, 0.0173, -0.0237, -0.055, -0.0854, -0.0897, -0.0893, -0.0877, -0.1689, -0.2401, -0.2589, -0.307, -0.1419, -0.2141, -0.1724, -0.1627, -0.1557, -0.2177, -0.2247, -0.2247, -0.2419, -0.2357, -0.1991, -0.1629, -0.1436, -0.1459, -0.1489, -0.1374, -0.1436, -0.1246, -0.1087, -0.1262, -0.1108, -0.0989, -0.0888, -0.042, -0.0035, 0.0431, 0.0189, 0.0382, 0.0167, 0.0058, -0.0071, -0.0191, -0.0226, -0.0226, 0.0151, 0.0, 0.0015, -0.0189, 0.0157, 0.0394, 0.0507, 0.0251, 0.0295, 0.0786, 0.092, 0.0734, 0.05, 0.0292, 0.1199, 0.1262, 0.1234, 0.1234, 0.1363, 0.1481, 0.2266, 0.2706, 0.2731, 0.2742, 0.2573, 0.206, 0.2274, 0.222, 0.222, 0.1944, 0.2212, 0.2264, 0.2773, 0.2974, 0.2958, 0.3788, 0.4184, 0.4213, 0.391, 0.3911, 0.3709, 0.4057, 0.4365, 0.475, 0.5387, 0.5722, 0.5906, 0.5622, 0.5214, 0.5664, 0.5445, 0.4453, 0.5276, 0.5307, 0.5265, 0.5502, 0.6341, 0.6033, 0.5728, 0.5607, 0.4757, 0.4638, 0.4506, 0.4864, 0.4475, 0.4763, 0.4809, 0.4937, 0.441, 0.441, 0.4382, 0.4366, 0.4335, 0.3392, 0.3416, 0.3806, 0.4136, 0.3793, 0.405, 0.428, 0.4218, 0.4102, 0.3993, 0.3946, 0.4158, 0.4257, 0.4255, 0.4289, 0.4129, 0.4297, 0.4335, 0.4532, 0.5039, 0.4591, 0.805, 0.8741, 1.0872, 1.0635, 0.9041, 0.9176, 0.9324, 1.1141, 1.0783, 1.0652, 1.1315, 1.1091, 1.04, 1.0821, 1.241, 1.3008, 1.2861, 1.3421, 1.258, 1.2694, 1.3006, 1.2156, 1.2712, 1.1061, 1.0693, 1.1618, 1.1046, 1.2939, 1.1971, 1.1869, 1.1311, 1.0405, 0.9808, 0.8254]}
{"unit_id": "orig:1939870805758026112", "ticker_idx": 0, "ts": "2025-07-01T02:16:55+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "companies like NVDA might secure more exemptions on computing power export restrictions in the future", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS: NVDA may benefit from loosened export controls as Trump trades restrictions for China concessions on critical minerals.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Tariffs: Section 232 Investigation Results Potentially in July-August, Electronic Product Tariffs Most Likely, Export Control Relaxation and Material Restrictions as Potential Variables\nUBS (T. Arcuri, 2025/06/29)\n\nUBS points out that tariff policies will become a blind spot for investors in the next 4-6 weeks. The Section 232 investigation may conclude earlier, in July or August, with the focus likely shifting to imposing a 25% tariff on finished electronic products and introducing a mechanism for deducting the content of U.S.-produced products. Concurrently, export controls might see marginal relaxation, but restrictions on materials could become a new focal point in the U.S.-China competition.\n\nSection 232 Tariff Investigation Pace Significantly Accelerated, Final Measures Expected in July-August\n\nExperts indicate that the Trump administration has recently accelerated the semiconductor-related Section 232 investigation. The originally anticipated 9-month investigation period has been compressed, and final policies might be released in late July to early August. This accelerated pace is notably earlier than market expectations and could trigger a short-term policy shock reaction in the market.\n\n25% Tariff on Finished Electronic Products Expected as Main Outcome, No Product Content Exemption Initially\n\nExperts emphasize that the most probable outcome of the investigation is a uniform 25% tariff on finished electronic products, rather than on semiconductors themselves. This policy will be superimposed on existing country-specific tariffs and will initially not include exemptions for U.S.-produced product content or exceptions for Taiwan, due to high administrative enforcement difficulty. In the long run, a deduction mechanism might be gradually introduced, but no exemptions are expected in the short term.\n\nExport Controls Might Loosen, NVDA Potentially a Beneficiary\n\nExperts believe that the Trump administration might be more willing to relax export controls in exchange for China's concessions in areas like critical minerals. For instance, EDA and aerospace components have already become bargaining chips, suggesting that companies like NVDA might secure more exemptions on computing power export restrictions in the future, especially under a larger trade agreement framework.\n\nAI Diffusion Rules Potentially Replaced, Bilateral AI Cooperation Framework More Likely to Land\n\nExperts note that the government has not strongly pushed for AI Diffusion rules. The Trump administration is more likely to replace relevant regulations with bilateral AI cooperation agreements between countries to support AI infrastructure collaboration with the UAE and Saudi Arabia. However, these agreements currently resemble frameworks of intent rather than substantive supply agreements, leaving uncertainty about their actual implementation.\n\nTotal Ban on Semiconductor Equipment Unlikely, Material Segment Becomes New Potential Chokepoint\n\nExperts indicate that a comprehensive ban on U.S. semiconductor equipment exports to China is extremely difficult and would require multi-country coordination, making it unrealistic. A more feasible approach is to expand the scope of the export controls from October 7, 2022, to include 40-65nm legacy process equipment. However, in comparison, export restrictions on semiconductor materials hold greater potential, possibly gaining policy attention due to Huawei's involvement, thus becoming a key direction for the next round of U.S.-China competition.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030593628670980477, "ret_signed_m1d": 0.030593628670980477, "alpha_spy_m1d": 0.030269844560538717, "alpha_c_m1d": 0.019572715393224804, "bench_spy_m1d": 0.0003237841104417605, "bench_c_m1d": 0.011020913277755673, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025766351418616162, "ret_signed_p1d": 0.025766351418616162, "alpha_spy_p1d": 0.021233074902800242, "alpha_c_p1d": 0.006153529443998718, "bench_spy_p1d": 0.0045332765158159205, "bench_c_p1d": 0.019612821974617445, "ret_p1w": 0.043705126928820226, "ret_signed_p1w": 0.043705126928820226, "alpha_spy_p1w": 0.0393499366565746, "alpha_c_p1w": 0.016007793751005917, "bench_spy_p1w": 0.004355190272245624, "bench_c_p1w": 0.02769733317781431, "ret_p1m": 0.16940637132901948, "ret_signed_p1m": 0.16940637132901948, "alpha_spy_p1m": 0.14219036856095935, "alpha_c_p1m": 0.09537783166273206, "bench_spy_p1m": 0.027216002768060132, "bench_c_p1m": 0.07402853966628742, "ret_p3m": 0.16242687904430064, "ret_signed_p3m": 0.16242687904430064, "alpha_spy_p3m": 0.08794317966583853, "alpha_c_p3m": -0.004264073598932994, "bench_spy_p3m": 0.07448369937846211, "bench_c_p3m": 0.16669095264323364, "ret_p6m": 0.23047048282727722, "ret_signed_p6m": 0.23047048282727722, "alpha_spy_p6m": 0.10630690019844913, "alpha_c_p6m": -0.09385950444228386, "bench_spy_p6m": 0.1241635826288281, "bench_c_p6m": 0.3243299872695611, "price_path": [-0.336, -0.3848, -0.3631, -0.3719, -0.2543, -0.2984, -0.2764, -0.2779, -0.2682, -0.3184, -0.338, -0.338, -0.3679, -0.355, -0.3301, -0.3058, -0.2759, -0.2908, -0.2889, -0.2895, -0.272, -0.2532, -0.2576, -0.2594, -0.2365, -0.2344, -0.2391, -0.1977, -0.1525, -0.1172, -0.1205, -0.1168, -0.1157, -0.1235, -0.1403, -0.1336, -0.1436, -0.1436, -0.1162, -0.1207, -0.0921, -0.1186, -0.1039, -0.0789, -0.0743, -0.0869, -0.0756, -0.0697, -0.061, -0.0683, -0.0541, -0.0739, -0.0562, -0.0599, -0.051, -0.051, -0.0616, -0.0596, -0.0352, 0.0066, 0.0112, 0.029, 0.0306, 0.0, 0.0258, 0.0394, 0.0394, 0.0322, 0.0437, 0.0625, 0.0705, 0.0758, 0.0703, 0.1135, 0.1179, 0.1285, 0.1247, 0.1179, 0.0896, 0.114, 0.1333, 0.1318, 0.153, 0.1449, 0.1694, 0.1603, 0.1332, 0.1742, 0.1628, 0.1704, 0.1792, 0.1918, 0.1876, 0.1948, 0.1845, 0.1873, 0.1771, 0.1873, 0.1457, 0.1442, 0.1414, 0.1611, 0.1729, 0.1857, 0.1846, 0.1753, 0.1362, 0.1362, 0.114, 0.113, 0.1198, 0.0895, 0.0979, 0.1139, 0.1568, 0.1558, 0.16, 0.1596, 0.1408, 0.1109, 0.1497, 0.1525, 0.1978, 0.164, 0.1545, 0.1592, 0.1624, 0.1863, 0.2172, 0.2215, 0.2322, 0.2239, 0.2104, 0.2071, 0.2337, 0.2562, 0.1948, 0.2285, 0.1744, 0.1731, 0.186, 0.1952, 0.1915, 0.1818, 0.1761, 0.1883, 0.2151, 0.2492, 0.3114, 0.3506, 0.3236, 0.3209, 0.3496, 0.2962, 0.2735, 0.2269, 0.2274, 0.2985, 0.2601, 0.2643, 0.219, 0.2406, 0.2173, 0.1831, 0.2168, 0.1784, 0.1669, 0.1909, 0.16, 0.1759, 0.1759, 0.1547, 0.1737, 0.1838, 0.1716, 0.1964, 0.19, 0.2105, 0.2067, 0.199, 0.1804, 0.1418, 0.1501, 0.1594, 0.1152, 0.1361, 0.1808, 0.1984, 0.2344, 0.2305]}
{"unit_id": "orig:1935116506536493535", "ticker_idx": 0, "ts": "2025-06-17T23:25:02+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Daishin Securities Raises Target Price to 320,000 Won", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Daishin Securities raises SK Hynix target to KRW 320,000, calling it the beginning of a new era; author shares/endorses bullish call.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix: The Beginning of a New Era – Daishin Securities Raises Target Price to 320,000 Won https://t.co/eHwWP5KeeA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00401608475530546, "ret_signed_m1d": -0.00401608475530546, "alpha_spy_m1d": -0.012634764188059533, "alpha_c_m1d": -0.010858880095655943, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01004021188826365, "ret_signed_p1d": -0.01004021188826365, "alpha_spy_p1d": -0.00988957613409247, "alpha_c_p1d": -0.013862946356956618, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.11847387014928157, "ret_signed_p1w": 0.11847387014928157, "alpha_spy_p1w": 0.09999154500272933, "alpha_c_p1w": 0.0804372760037082, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.18875503830101237, "ret_signed_p1m": 0.18875503830101237, "alpha_spy_p1m": 0.1409996815107235, "alpha_c_p1m": 0.0829024138076131, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": 0.321182745593249, "ret_signed_p3m": 0.321182745593249, "alpha_spy_p3m": 0.21771778230219052, "alpha_c_p3m": 0.15783545041252056, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 1.3625255321702983, "ret_signed_p6m": 1.3625255321702983, "alpha_spy_p6m": 1.2052372154401314, "alpha_c_p6m": 0.9324249563143081, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [-0.1581, -0.1361, -0.1521, -0.1662, -0.1421, -0.1702, -0.201, -0.2355, -0.2103, -0.2067, -0.2199, -0.2696, -0.3393, -0.3205, -0.3385, -0.2656, -0.2752, -0.2776, -0.276, -0.3025, -0.2985, -0.2985, -0.292, -0.3033, -0.2744, -0.2852, -0.2608, -0.2704, -0.2752, -0.2884, -0.2884, -0.2544, -0.2544, -0.2544, -0.2351, -0.2371, -0.2379, -0.2183, -0.2042, -0.1742, -0.1962, -0.1802, -0.2006, -0.1902, -0.1962, -0.2107, -0.1982, -0.1862, -0.1882, -0.1662, -0.1486, -0.1787, -0.1667, -0.1667, -0.1265, -0.0984, -0.0984, -0.0803, -0.0743, -0.0361, -0.0542, -0.0542, -0.004, 0.0, -0.01, -0.012, 0.0321, 0.0422, 0.1185, 0.1486, 0.1767, 0.1406, 0.1727, 0.1466, 0.1205, 0.1185, 0.0863, 0.0884, 0.1325, 0.1285, 0.1928, 0.1827, 0.2048, 0.1988, 0.1888, 0.0823, 0.0803, 0.0944, 0.0783, 0.0803, 0.0823, 0.0683, 0.0522, 0.0542, 0.0582, 0.0984, 0.0361, 0.0361, 0.0582, 0.0382, 0.0522, 0.0301, 0.0723, 0.0803, 0.1165, 0.1104, 0.1104, 0.0743, 0.0562, 0.0261, -0.0161, 0.008, 0.0422, 0.0502, 0.0442, 0.0799, 0.0819, 0.0296, 0.0477, 0.0557, 0.0678, 0.1, 0.1141, 0.1583, 0.2226, 0.2347, 0.3212, 0.3312, 0.3996, 0.3413, 0.4298, 0.4298, 0.4117, 0.4519, 0.4378, 0.4338, 0.3534, 0.4036, 0.3976, 0.4479, 0.5906, 0.5906, 0.5906, 0.5906, 0.5906, 0.5906, 0.7214, 0.6691, 0.655, 0.6992, 0.8199, 0.8722, 0.9526, 0.9265, 0.9365, 0.9245, 1.0512, 1.1517, 1.0954, 1.2442, 1.2844, 1.2482, 1.4936, 1.3568, 1.3287, 1.385, 1.3327, 1.4373, 1.4895, 1.4815, 1.4614, 1.2522, 1.4373, 1.2925, 1.2603, 1.2965, 1.0954, 1.0914, 1.0873, 1.1075, 1.1895, 1.1331, 1.1653, 1.2458, 1.2217, 1.1814, 1.1895, 1.3223, 1.278, 1.3625]}
{"unit_id": "orig:1943318860352098565", "ticker_idx": 0, "ts": "2025-07-10T14:38:16+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes MediaTek's revenue certainty for the Google TPU project has strengthened, maintaining a target price of NT1,888. If MediaTek successfully secures the Meta MTIA ASIC order, it could potentially reach a bull-case scenario valuation of NT2,600", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley maintains Buy on MediaTek NT$1,888 PT; bull case NT$2,600 if Meta MTIA order secured alongside Google TPU ramp.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "MediaTek: Benefiting from Google TPU Project Progress, Could Reach Bull Case Target Price if Meta Order Secured\nMorgan Stanley (C. Chan, July 09, 2025)\n\nMorgan Stanley believes MediaTek's revenue certainty for the Google TPU project has strengthened, maintaining a target price of NT1,888.If MediaTek successfully secures the Meta MTIA ASIC order,it could potentially reach a bull−case scenario valuation of NT2,600.\n\nGoogle TPU Project Expected to Tape-Out Mid-September, Significant Revenue Contribution in 2026\nMediaTek's involvement in Google's next-generation TPU is expected to complete tape-out by mid-September 2025. The primary revenue contribution is anticipated in 2026, projected to exceed $1 billion for the full year, with most of it concentrated in Q4 2026.\n\nMeta MTIA Project Competition Fierce, MediaTek and Broadcom Are Final Candidates\nIn Meta's AI accelerator MTIA project, MediaTek and Broadcom are the two final design service candidates. Broadcom holds an advantage in complex packaging, while MediaTek is more attractive in terms of Total Cost of Ownership (TCO) due to its operational efficiency and cost control capabilities. A decision is expected by the end of July or early August, with the current winning probability at 50%.\n\nMaintaining Optimistic Outlook for MediaTek's AI ASIC Business\nMorgan Stanley considers the Meta MTIA project as part of its bull case scenario. The firm believes that if MediaTek wins orders from both Google and Meta, two major CSPs (Cloud Service Providers), it will significantly enhance its market position and revenue flexibility in the AI ASIC domain.\n\nSmartphone Business Stable in Short Term, Q3 2025 Revenue Slightly Down\nAlthough MediaTek previously canceled some wafer foundry orders, it has recently re-ordered, reflecting stable demand for smartphones in China. Q3 2025 revenue is expected to decrease by 8.5% QoQ, but the overall short-term fundamentals remain stable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03214287460208054, "ret_signed_m1d": -0.03214287460208054, "alpha_spy_m1d": -0.029330496163386344, "alpha_c_m1d": -0.024839177822934322, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01428568254167173, "ret_signed_p1d": 0.01428568254167173, "alpha_spy_p1d": 0.017801106412267287, "alpha_c_p1d": 0.014390024109116273, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.007142841270835865, "ret_signed_p1w": -0.007142841270835865, "alpha_spy_p1w": -0.010690132337840996, "alpha_c_p1w": -0.02171573822927919, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": -0.03571429523749847, "ret_signed_p1m": -0.03571429523749847, "alpha_spy_p1m": -0.053866401103472406, "alpha_c_p1m": -0.056617211797250544, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": -0.053571398414588245, "ret_signed_p3m": -0.053571398414588245, "alpha_spy_p3m": -0.12572482912974614, "alpha_c_p3m": -0.2258377292007807, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.0499999777791702, "ret_signed_p6m": 0.0499999777791702, "alpha_spy_p6m": -0.047900869012889924, "alpha_c_p6m": -0.2523844399789128, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.0193, -0.0298, -0.0438, -0.0613, -0.0438, -0.0578, -0.0894, -0.0403, -0.0683, -0.0333, -0.0543, -0.0403, -0.0543, -0.0543, -0.0894, -0.0929, -0.1034, -0.1069, -0.0964, -0.0718, -0.0788, -0.0508, -0.0473, -0.0403, -0.0438, -0.0823, -0.0788, -0.0613, -0.0718, -0.0788, -0.1034, -0.1034, -0.1034, -0.1174, -0.1174, -0.1174, -0.1069, -0.1069, -0.1279, -0.1034, -0.0999, -0.0648, -0.0788, -0.0999, -0.1244, -0.1209, -0.1069, -0.0999, -0.1139, -0.1244, -0.1209, -0.1104, -0.0929, -0.0964, -0.0999, -0.1244, -0.1069, -0.0929, -0.0929, -0.0786, -0.0857, -0.0929, -0.0321, 0.0, 0.0143, -0.0107, -0.0036, 0.0071, -0.0071, 0.0071, 0.0179, 0.0214, 0.0214, 0.0214, 0.0214, 0.0, -0.0286, -0.0143, -0.0214, -0.0214, -0.05, -0.0429, -0.0571, -0.0321, -0.0357, -0.0321, -0.0286, -0.0214, -0.0071, -0.0214, 0.0071, -0.0071, -0.0321, -0.025, -0.025, 0.0071, 0.0, 0.0, -0.0107, -0.0214, -0.0286, -0.0286, -0.0036, -0.0107, 0.025, 0.0214, 0.0786, 0.0643, 0.0571, 0.0607, 0.0607, 0.0964, 0.0786, 0.0786, 0.0286, 0.0036, -0.0179, -0.05, -0.0393, -0.0643, -0.0643, -0.0607, -0.0821, -0.0857, -0.0571, -0.0571, -0.0536, -0.0464, -0.0393, -0.0393, -0.0607, -0.0679, -0.0607, -0.05, -0.05, -0.0429, -0.0429, -0.05, -0.075, -0.075, -0.0464, -0.0607, -0.0714, -0.0643, -0.0643, -0.0714, -0.0607, -0.05, -0.0786, -0.1, -0.1071, -0.1143, -0.1107, -0.1107, -0.1214, -0.1214, -0.1643, -0.1714, -0.1536, -0.1821, -0.1786, -0.1536, -0.0714, -0.0429, -0.0036, 0.0321, 0.0107, 0.0, 0.0036, 0.0179, 0.0286, 0.0143, 0.0429, -0.0036, 0.0036, 0.0143, 0.0143, 0.0179, 0.0143, 0.0071, 0.0, -0.0071, -0.0143, -0.0143, -0.0107, 0.0143, 0.0143, 0.0214, 0.0214, 0.05]}
{"unit_id": "orig:1938958664058183901", "ticker_idx": 0, "ts": "2025-06-28T13:52:24+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EPS and Target Price Upgraded, Positive on GenAI Cycle Driving Overall Growth… the target price is raised to $530", "tweet_type": "news_reaction", "timeline": "years", "summary": "Morgan Stanley raises MSFT PT to $530 on Azure AI revenue upside, 33% CAGR to $200B+ by 2028, and mid-double-digit EPS CAGR from GenAI cycle.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Microsoft: Azure AI Business Accelerating ROI from GenAI Investments, Opening Up Valuation Upside\n\nMorgan Stanley (K. Weiss, 2025/06/26)\n\nMorgan Stanley believes that Microsoft's capital expenditures in the generative AI (GenAI) sector are yielding rapid returns. Both Azure AI's revenue and profitability have upside potential. Under more optimistic model assumptions, Azure revenue could exceed $200 billion by 2028.\n\nCapex Derivation Model Shows Upside for Both Azure AI Revenue and Profitability\n\nMorgan Stanley's capital expenditure derivation model, used to estimate Azure AI revenue, suggests that current model assumptions are conservative. The firm believes that Azure AI's gross margin has already surpassed 40% and is likely to reach or even exceed 50% in the future. In a 60% gross margin scenario, Azure revenue would be significantly higher than current expectations, implying higher returns per unit of capital expenditure.\n\nInvestor Views on OpenAI Dependence Diverge, But Growth Prospect Remains Attractive\n\nMorgan Stanley notes that some investors believe OpenAI's contribution is underestimated and that Azure's growth sustainability is strong. Others, however, are concerned about Microsoft's reliance on OpenAI and the risks associated with changes in their relationship. In response, the firm emphasizes that Azure AI's core business itself exhibits a triple-digit growth trend, which is far from peaking.\n\nUpdated OpenAI Model Still Indicates Conservative Azure Forecasts\n\nAfter incorporating new information and assumptions into the model, Morgan Stanley found that if OpenAI revenue is calculated solely based on current assumptions, the growth of the \"other AI customers\" (non-OpenAI) segment is underestimated in the model, even showing a negative growth trend. This clearly contradicts the reality of current triple-digit growth. If a more reasonable assumption of slowing growth is applied, overall Azure revenue could maintain a 33% compound annual growth rate (CAGR), exceeding $200 billion by 2028.\n\nEPS and Target Price Upgraded, Positive on GenAI Cycle Driving Overall Growth\n\nMorgan Stanley expects Microsoft to benefit from the structural shift in IT spending driven by GenAI and to achieve mid-double-digit EPS CAGR growth through cost control and business synergy. Based on a FY27 EPS of $17.71, the target price is raised to $530, corresponding to a 30x price-to-earnings (P/E) ratio, reflecting its dominant position and strong execution in the AI cycle.\n\n$MSFT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0029552926869632623, "ret_signed_m1d": -0.0029552926869632623, "alpha_spy_m1d": 0.00180326137809117, "alpha_c_m1d": 0.00180326137809117, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.004758554065054432, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010775838078136424, "ret_signed_p1d": -0.010775838078136424, "alpha_spy_p1d": -0.010452158769911568, "alpha_c_p1d": -0.010452158769911568, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.0003236793082248557, "ret_p1w": 0.0006232540167663281, "ret_signed_p1w": 0.0006232540167663281, "alpha_spy_p1w": -0.003957072198976341, "alpha_c_p1w": -0.003957072198976341, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.004580326215742669, "ret_p1m": 0.03047791301061009, "ret_signed_p1m": 0.03047791301061009, "alpha_spy_p1m": 0.002299681574823964, "alpha_c_p1m": 0.002299681574823964, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.028178231435786127, "ret_p3m": 0.02101593002140012, "ret_signed_p3m": 0.02101593002140012, "alpha_spy_p3m": -0.0470011571377964, "alpha_c_p3m": -0.0470011571377964, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.06801708715919652, "ret_p6m": -0.017785992455645605, "ret_signed_p6m": -0.017785992455645605, "alpha_spy_p6m": -0.13764646266434744, "alpha_c_p6m": -0.13764646266434744, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.11986047020870183, "price_path": [-0.2331, -0.2513, -0.2779, -0.2819, -0.2885, -0.2164, -0.2347, -0.2205, -0.2218, -0.2259, -0.2543, -0.262, -0.262, -0.2793, -0.2639, -0.2487, -0.2228, -0.2137, -0.215, -0.2093, -0.2068, -0.1463, -0.1265, -0.1247, -0.1305, -0.1304, -0.1207, -0.1196, -0.0985, -0.0987, -0.0911, -0.089, -0.0867, -0.0775, -0.0789, -0.0901, -0.0855, -0.095, -0.095, -0.0738, -0.0805, -0.0779, -0.0745, -0.0712, -0.0692, -0.0674, -0.0598, -0.0543, -0.0496, -0.0533, -0.0498, -0.0373, -0.0451, -0.0367, -0.0389, -0.0345, -0.0345, -0.0402, -0.0229, -0.0147, -0.0103, 0.0001, -0.003, 0.0, -0.0108, -0.0127, 0.0029, 0.0029, 0.0006, -0.0016, 0.0123, 0.0082, 0.0119, 0.0113, 0.0169, 0.0165, 0.0287, 0.0254, 0.0254, 0.0158, 0.017, 0.0271, 0.0328, 0.0303, 0.0305, 0.0318, 0.0726, 0.0537, 0.0769, 0.061, 0.0553, 0.0471, 0.0495, 0.049, 0.064, 0.0466, 0.0504, 0.0458, 0.0396, 0.0248, 0.0167, 0.0154, 0.0214, 0.0154, 0.011, 0.0204, 0.0263, 0.0203, 0.0203, 0.0172, 0.0176, 0.0229, -0.0032, 0.0032, 0.0037, 0.0076, 0.0089, 0.0268, 0.0378, 0.0251, 0.027, 0.0239, 0.043, 0.036, 0.0254, 0.0273, 0.021, 0.0299, 0.0363, 0.043, 0.0465, 0.0386, 0.0418, 0.0644, 0.0551, 0.0569, 0.052, 0.0289, 0.0352, 0.0342, 0.0339, 0.0302, 0.0342, 0.0407, 0.0424, 0.0482, 0.0483, 0.0544, 0.0703, 0.0916, 0.0905, 0.0587, 0.0427, 0.0412, 0.0357, 0.0213, 0.001, 0.0005, 0.0189, 0.0243, 0.0293, 0.0135, 0.0274, 0.0219, -0.0056, -0.0191, -0.0348, -0.0475, -0.0437, -0.0377, -0.0205, -0.0205, -0.0074, -0.018, -0.0114, -0.0362, -0.0299, -0.0252, -0.0094, -0.0074, -0.0345, -0.0246, -0.0346, -0.0421, -0.0389, -0.0394, -0.0236, -0.0197, -0.0217, -0.0178]}
{"unit_id": "orig:1932074720536752602", "ticker_idx": 0, "ts": "2025-06-09T13:58:04+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS favors companies competitively positioned in both N2 expansions and mature process pricing recovery", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS bullish foundry note: TSMC favored on N2 ramp + mature node pricing recovery; Intel continues outsourcing to TSMC supporting demand.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Foundry: N2 Growth and Mature Nodes Advancing in Parallel, Capacity Expansion and Pricing Resilience Expected\nUBS (S. Lin, 25/06/06)\n\nOver the next three years, driven by multiple platform migrations including smartphones, PCs, and servers, the N2 node is expected to become TSMC’s second-largest process node. Concurrently, utilization and pricing improvements in mature processes will underpin the broader foundry industry.\n\n• Robust N2 Node Demand Expansion\nUBS’s TAM analysis suggests Qualcomm and MediaTek’s flagship SoCs will migrate concurrently with Apple’s A-series to N2, marking the first simultaneous generation upgrade. From H226E through 2027, AMD server platforms and Intel’s PC outsourcing will further drive N2 demand. Accelerated chip upgrades driven by AI applications for mobile and PC platforms, alongside future server expansions, are expected to result in approximately 200K wafers/month demand for N2 by 2028. Strong demand supports TSMC’s capacity expansion plans to 80K wafers/month by year-end 2026 and 140K wafers/month by year-end 2027, despite potential tariff-related impacts on iPhones.\n\n• Intel 18A In-house Production and Continued Outsourcing\nIntel plans to start volume production of 18A (Panther Lake) by late 2025, with an initial yield estimated around 50% and production reaching 20K–30K wafers/month by 2027. To maintain time-to-market advantages and supply flexibility, Intel will continue outsourcing PC CPUs, such as Nova Lake, to TSMC’s N3/N2 nodes, with potential partial outsourcing of Diamond Rapids server CPUs as well. Regarding its foundry service (18AP), final PDK availability may not materialize until Q1 2026, with external customers’ mass production likely to begin in H2 2028 at the earliest.\n\n• Mature Process Utilization and Pricing Recovery\nUBS believes the worst has passed for mature nodes, forecasting utilization to recover from 60%+ in 2024 to over 70% by mid-2025. China domestic 28/40nm utilization remains above 80%, averaging around 85% throughout the year. Driven by production line restarts and distributor restocking, utilization could see stability or slight improvement through Q3 2025, with a seasonal decline expected from Q4 2025 to early 2026. Pricing power for mature nodes is expected to improve in the second half, as leading manufacturers like Hua Hong have already started raising prices, reflecting a better supply-demand balance.\n\n• Risks and Implications\nKey points to watch include: TSMC’s N2 capacity expansion timeline and potential demand forecast deviation due to iPhone shipment fluctuations; the balance between Intel’s outsourcing and internal 18A production impacting foundry revenue structure; and potential oversupply resulting from increased Chinese WFE investments in the 28/40nm domain. Overall, driven by both N2 node expansion and mature process improvements, leading foundries are expected to continue benefiting from high utilization rates and pricing strength. UBS favors companies competitively positioned in both N2 expansions and mature process pricing recovery.\n\n$TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00879232261884455, "ret_signed_m1d": -0.00879232261884455, "alpha_spy_m1d": -0.007891851310082076, "alpha_c_m1d": 0.007669303414965967, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026376893229590026, "ret_signed_p1d": 0.026376893229590026, "alpha_spy_p1d": 0.020707228560699287, "alpha_c_p1d": 0.006490417905588686, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.045742211969475255, "ret_signed_p1w": 0.045742211969475255, "alpha_spy_p1w": 0.040739536395070886, "alpha_c_p1w": 0.02075765762518955, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.10479794531128617, "ret_signed_p1m": 0.10479794531128617, "alpha_spy_p1m": 0.06728801871405876, "alpha_c_p1m": 0.0015910626365585046, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.14043507209634387, "ret_signed_p3m": 0.14043507209634387, "alpha_spy_p3m": 0.05479116047450927, "alpha_c_p3m": 0.012399580802975985, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.42069014101955027, "ret_signed_p6m": 0.42069014101955027, "alpha_spy_p6m": 0.27768063763457285, "alpha_c_p6m": 0.021912021646373514, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1474, -0.1743, -0.1622, -0.1519, -0.1636, -0.1606, -0.1427, -0.1462, -0.1248, -0.1261, -0.1618, -0.1872, -0.2017, -0.1981, -0.1853, -0.1769, -0.2397, -0.2908, -0.2939, -0.3171, -0.2331, -0.2699, -0.2412, -0.2471, -0.24, -0.2673, -0.267, -0.267, -0.2857, -0.2686, -0.2376, -0.2069, -0.2024, -0.2106, -0.2054, -0.1947, -0.1656, -0.1339, -0.1478, -0.1677, -0.1568, -0.1535, -0.1472, -0.0967, -0.0629, -0.0591, -0.0617, -0.0617, -0.0652, -0.0655, -0.0736, -0.0522, -0.0726, -0.0726, -0.045, -0.0525, -0.0476, -0.0661, -0.0587, -0.0454, -0.0222, -0.0177, -0.0088, 0.0, 0.0264, 0.0343, 0.0445, 0.0235, 0.0457, 0.0371, 0.0352, 0.0352, 0.0158, 0.0198, 0.0671, 0.08, 0.0861, 0.1082, 0.0982, 0.0894, 0.1326, 0.1384, 0.1384, 0.1111, 0.1048, 0.1241, 0.114, 0.1171, 0.1087, 0.1489, 0.1518, 0.1908, 0.1656, 0.1581, 0.1375, 0.1653, 0.1714, 0.1908, 0.177, 0.1701, 0.1778, 0.1715, 0.1404, 0.1588, 0.1271, 0.1218, 0.1764, 0.1725, 0.1738, 0.1845, 0.1706, 0.1685, 0.1582, 0.1705, 0.1283, 0.1084, 0.1022, 0.1297, 0.1423, 0.1575, 0.1602, 0.1553, 0.1194, 0.1194, 0.1074, 0.1219, 0.1404, 0.1802, 0.1985, 0.2166, 0.2628, 0.2553, 0.2574, 0.2673, 0.2746, 0.2782, 0.3066, 0.2883, 0.326, 0.3751, 0.3653, 0.3456, 0.3296, 0.329, 0.3584, 0.4031, 0.4013, 0.4212, 0.4708, 0.4301, 0.4811, 0.4586, 0.3651, 0.4732, 0.4394, 0.4821, 0.4584, 0.4352, 0.448, 0.4325, 0.4051, 0.4141, 0.4346, 0.4507, 0.4666, 0.4839, 0.4748, 0.4613, 0.4828, 0.4302, 0.4282, 0.4068, 0.3935, 0.4362, 0.4162, 0.4135, 0.3726, 0.3853, 0.3717, 0.3517, 0.3734, 0.3497, 0.3379, 0.3845, 0.3846, 0.4103, 0.4103, 0.4179, 0.3992, 0.4207]}
{"unit_id": "orig:1932074720536752602", "ticker_idx": 1, "ts": "2025-06-09T13:58:04+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel will continue outsourcing PC CPUs, such as Nova Lake, to TSMC's N3/N2 nodes, with potential partial outsourcing of Diamond Rapids server CPUs as well", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS bullish foundry note: TSMC favored on N2 ramp + mature node pricing recovery; Intel continues outsourcing to TSMC supporting demand.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Foundry: N2 Growth and Mature Nodes Advancing in Parallel, Capacity Expansion and Pricing Resilience Expected\nUBS (S. Lin, 25/06/06)\n\nOver the next three years, driven by multiple platform migrations including smartphones, PCs, and servers, the N2 node is expected to become TSMC’s second-largest process node. Concurrently, utilization and pricing improvements in mature processes will underpin the broader foundry industry.\n\n• Robust N2 Node Demand Expansion\nUBS’s TAM analysis suggests Qualcomm and MediaTek’s flagship SoCs will migrate concurrently with Apple’s A-series to N2, marking the first simultaneous generation upgrade. From H226E through 2027, AMD server platforms and Intel’s PC outsourcing will further drive N2 demand. Accelerated chip upgrades driven by AI applications for mobile and PC platforms, alongside future server expansions, are expected to result in approximately 200K wafers/month demand for N2 by 2028. Strong demand supports TSMC’s capacity expansion plans to 80K wafers/month by year-end 2026 and 140K wafers/month by year-end 2027, despite potential tariff-related impacts on iPhones.\n\n• Intel 18A In-house Production and Continued Outsourcing\nIntel plans to start volume production of 18A (Panther Lake) by late 2025, with an initial yield estimated around 50% and production reaching 20K–30K wafers/month by 2027. To maintain time-to-market advantages and supply flexibility, Intel will continue outsourcing PC CPUs, such as Nova Lake, to TSMC’s N3/N2 nodes, with potential partial outsourcing of Diamond Rapids server CPUs as well. Regarding its foundry service (18AP), final PDK availability may not materialize until Q1 2026, with external customers’ mass production likely to begin in H2 2028 at the earliest.\n\n• Mature Process Utilization and Pricing Recovery\nUBS believes the worst has passed for mature nodes, forecasting utilization to recover from 60%+ in 2024 to over 70% by mid-2025. China domestic 28/40nm utilization remains above 80%, averaging around 85% throughout the year. Driven by production line restarts and distributor restocking, utilization could see stability or slight improvement through Q3 2025, with a seasonal decline expected from Q4 2025 to early 2026. Pricing power for mature nodes is expected to improve in the second half, as leading manufacturers like Hua Hong have already started raising prices, reflecting a better supply-demand balance.\n\n• Risks and Implications\nKey points to watch include: TSMC’s N2 capacity expansion timeline and potential demand forecast deviation due to iPhone shipment fluctuations; the balance between Intel’s outsourcing and internal 18A production impacting foundry revenue structure; and potential oversupply resulting from increased Chinese WFE investments in the 28/40nm domain. Overall, driven by both N2 node expansion and mature process improvements, leading foundries are expected to continue benefiting from high utilization rates and pricing strength. UBS favors companies competitively positioned in both N2 expansions and mature process pricing recovery.\n\n$TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02050781668367574, "ret_signed_m1d": -0.02050781668367574, "alpha_spy_m1d": -0.019607345374913265, "alpha_c_m1d": -0.004046190649865222, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07812502037268176, "ret_signed_p1d": 0.07812502037268176, "alpha_spy_p1d": 0.07245535570379102, "alpha_c_p1d": 0.058238545048680423, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.012695323959633464, "ret_signed_p1w": 0.012695323959633464, "alpha_spy_p1w": 0.007692648385229095, "alpha_c_p1w": -0.012289230384652239, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.15185550194655728, "ret_signed_p1m": 0.15185550194655728, "alpha_spy_p1m": 0.11434557534932988, "alpha_c_p1m": 0.04864861927182962, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.20166021291152125, "ret_signed_p3m": 0.20166021291152125, "alpha_spy_p3m": 0.11601630128968665, "alpha_c_p3m": 0.07362472161815337, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 1.1225587007975286, "ret_signed_p6m": 1.1225587007975286, "alpha_spy_p6m": 0.9795491974125512, "alpha_c_p6m": 0.7237805814243519, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [0.0098, 0.1572, 0.1743, 0.2544, 0.2656, 0.1777, 0.1699, 0.1846, 0.1826, 0.1816, 0.1436, 0.1533, 0.1089, 0.1089, 0.0767, 0.0732, 0.0952, -0.0308, -0.0444, -0.1147, 0.0513, -0.0293, -0.0361, -0.0083, -0.0308, -0.061, -0.0757, -0.0757, -0.0801, -0.0474, 0.0054, 0.0493, -0.021, 0.0015, -0.0068, -0.0186, -0.0244, 0.0068, -0.0103, -0.0264, -0.0083, 0.0254, 0.0459, 0.083, 0.1016, 0.0508, 0.0522, 0.0576, 0.0439, 0.0386, 0.0103, 0.0034, -0.021, -0.021, 0.0034, -0.0054, -0.0112, -0.0454, -0.0361, -0.0093, -0.0112, -0.0239, -0.0205, 0.0, 0.0781, 0.0098, 0.0142, -0.0166, 0.0127, 0.0156, 0.0493, 0.0493, 0.0293, 0.0347, 0.1011, 0.084, 0.0986, 0.1079, 0.0938, 0.1157, 0.0684, 0.0981, 0.0981, 0.0742, 0.1519, 0.1445, 0.1631, 0.144, 0.1377, 0.1191, 0.1079, 0.1133, 0.1279, 0.1357, 0.1348, 0.147, 0.105, 0.0107, 0.0098, -0.0034, -0.0068, -0.0332, -0.0571, -0.0479, -0.0142, -0.0034, -0.0347, -0.0259, 0.0083, 0.0649, 0.085, 0.165, 0.1992, 0.1553, 0.2358, 0.1494, 0.1475, 0.2109, 0.1987, 0.189, 0.2134, 0.2173, 0.189, 0.189, 0.1821, 0.1719, 0.2017, 0.1958, 0.1953, 0.1934, 0.2095, 0.2017, 0.1758, 0.2095, 0.2339, 0.2158, 0.4927, 0.4443, 0.4043, 0.4326, 0.5244, 0.6597, 0.7334, 0.6836, 0.6382, 0.7549, 0.8213, 0.7983, 0.7866, 0.8149, 0.8276, 0.8457, 0.7759, 0.8174, 0.7397, 0.814, 0.7988, 0.8071, 0.8604, 0.8613, 0.8027, 0.8633, 0.8691, 0.9307, 1.0278, 1.0186, 0.9609, 0.9526, 0.9287, 0.8081, 0.874, 0.8184, 0.8618, 0.8774, 0.8496, 0.8501, 0.7534, 0.7344, 0.6948, 0.6763, 0.7144, 0.6416, 0.6846, 0.7476, 0.7495, 0.7974, 0.7974, 0.9805, 0.9536, 1.1226]}
{"unit_id": "thread:1936924773508133342", "ticker_idx": 0, "ts": "2025-06-22T23:10:26+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "analysts have now revised their forecast for TSMC's gross profit margin in the second half to approximately 55%… the 'dual ratios,' including gross profit margin and operating profit margin, may fall to the low end of the guidance, making it difficult to surpass the guidance targets… TSMC may not achieve the market's expectation of profit growth this quarter compared to Q1, at most only matching the previous quarter or even facing a single-digit percentage sequential decline", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish TSMC near-term: NTD 12% appreciation crushing margins ~5pp, H2 GPM revised down to ~55%, Q2 profit likely misses growth expectations vs Q1.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: TSMC Faces Two Headwinds, Q2 Profit Under Pressure; Analysts Forecast \"Dual Ratios\" May Dip to Low End of Guidance\n\nTSMC (2330) has recently encountered two major external headwinds. Not only are there reports that the White House intends to revoke the waivers allowing its mainland China fabs to acquire U.S. technology, but as Q2 rapidly approaches its end, the New Taiwan Dollar's strong appreciation of 12% against the U.S. dollar this quarter appears irreversible. This severe appreciation is significantly impacting TSMC's gross profit margin and operating profit margin by nearly 5 percentage points each, severely eroding profits. The company's net profit for this quarter may fall below that of the previous quarter, failing to meet market expectations for continuous profit growth.\n\nFurthermore, the exchange rate has a substantial impact on the gross profit margin of export-oriented TSMC, which is heavily reliant on U.S. dollar revenue. Analysts had initially expected TSMC's gross profit margin in the second half of the year to challenge the 58% range. However, due to exchange rate interference, analysts have now revised their forecast for TSMC's gross profit margin in the second half to approximately 55%. In addition, initial yield adjustments during the mass production of 2nm in the second half of the year may also dilute the gross profit margin.\n\nRegarding the exchange rate issue, TSMC Chairman C.C. Wei stated directly at the June shareholders' meeting that exchange rate fluctuations would affect TSMC's gross profit margin and operating profit margin. He projected that for every 1% appreciation of the New Taiwan Dollar, the \"dual ratios\" would each be impacted by approximately 0.4 percentage points, calling it \"a very significant number.\" TSMC can only continue to strive to maintain technological leadership and sell at \"the value it deserves.\"\n\nAt its April investor conference, TSMC released its Q2 U.S. dollar revenue guidance of between $28.4 billion and $29.2 billion, with an average sequential increase of nearly 13%. These figures were higher than analysts' average sequential growth estimate of around 6% at the time. However, TSMC's average exchange rate assumption in April was NT$32.5 to US$1. With the New Taiwan Dollar appreciating more than 2% against the U.S. dollar in the past month and a strong 12% appreciation this quarter, closing at 29.529 last Friday (June 20), the average exchange rate has significantly diverged from TSMC's April investor conference assumption.\n\nBased on C.C. Wei's previous statements regarding the impact of exchange rate fluctuations on TSMC's \"dual ratios,\" the New Taiwan Dollar's strong 12% appreciation this quarter has had a significant impact, affecting both the gross profit margin and operating profit margin by nearly 5 percentage points each.\n\nConsidering that only six trading days remain in June, making it difficult to reverse the strong appreciation of the New Taiwan Dollar, analysts estimate that TSMC's Q2 U.S. dollar revenue guidance is still likely to be met. However, the \"dual ratios,\" including gross profit margin and operating profit margin, may fall to the low end of the guidance, making it difficult to surpass the guidance targets. Based on this, TSMC may not achieve the market's expectation of profit growth this quarter compared to Q1, at most only matching the previous quarter or even facing a single-digit percentage sequential decline, though it will still be better than the same period last year, setting a new record for the same period in history, with estimated earnings per share of NT$13 for the quarter.\n\nData from the past eight years show that TSMC's Q2 profit has outperformed Q1 approximately 50% of the time. In Q2 2024, TSMC's profit surpassed Q1 due to the rapid growth of AI and HPC applications. However, in Q2 of 2023, 2021, 2018, and 2017, profits all declined compared to Q1.\n\nForeign media have reported that the U.S. may revoke the waivers enjoyed by major semiconductor manufacturers such as TSMC, Samsung, and SK Hynix for acquiring U.S. technology in their mainland China fabs. The impact of such policy changes on TSMC's orders in mainland China remains to be seen. TSMC is scheduled to hold an investor conference in July to announce its Q1 financial report and operational outlook. With these two major headwinds emerging recently, the views of the company's senior management are garnering significant attention.\n\n$TSM\n\n---\n\nhttps://t.co/YFlRixS1TW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0038513868739630253, "ret_signed_m1d": 0.0038513868739630253, "alpha_spy_m1d": -0.00592948327630638, "alpha_c_m1d": -0.002448306486417251, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04645287181090407, "ret_signed_p1d": -0.04645287181090407, "alpha_spy_p1d": -0.03540566281781632, "alpha_c_p1d": -0.009724140113324786, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.07688276527060123, "ret_signed_p1w": -0.07688276527060123, "alpha_spy_p1w": -0.04739010592313453, "alpha_c_p1w": -0.012130215823267898, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": 0.11544310415283543, "ret_signed_p1m": -0.11544310415283543, "alpha_spy_p1m": -0.06760509542374349, "alpha_c_p1m": -0.025911955617052396, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.2813211688404449, "ret_signed_p3m": -0.2813211688404449, "alpha_spy_p3m": -0.17783040382183746, "alpha_c_p3m": -0.07114247727857448, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 0.37196590218525305, "ret_signed_p6m": -0.37196590218525305, "alpha_spy_p6m": -0.23766260136593265, "alpha_c_p6m": -0.028273215953928288, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [-0.1781, -0.203, -0.2172, -0.2136, -0.2011, -0.1928, -0.2544, -0.3046, -0.3075, -0.3303, -0.2479, -0.284, -0.2559, -0.2617, -0.2547, -0.2815, -0.2812, -0.2812, -0.2995, -0.2828, -0.2524, -0.2222, -0.2179, -0.2259, -0.2208, -0.2103, -0.1818, -0.1507, -0.1643, -0.1839, -0.1731, -0.1699, -0.1638, -0.1142, -0.081, -0.0774, -0.0799, -0.0799, -0.0833, -0.0836, -0.0916, -0.0706, -0.0905, -0.0905, -0.0635, -0.0708, -0.066, -0.0842, -0.077, -0.0639, -0.0412, -0.0367, -0.028, -0.0194, 0.0065, 0.0143, 0.0243, 0.0037, 0.0255, 0.017, 0.0151, 0.0151, -0.0039, 0.0, 0.0465, 0.0591, 0.0651, 0.0868, 0.0769, 0.0683, 0.1107, 0.1164, 0.1164, 0.0896, 0.0834, 0.1023, 0.0924, 0.0955, 0.0872, 0.1266, 0.1295, 0.1677, 0.143, 0.1357, 0.1154, 0.1427, 0.1487, 0.1677, 0.1542, 0.1474, 0.155, 0.1488, 0.1183, 0.1364, 0.1053, 0.1001, 0.1536, 0.1498, 0.1511, 0.1615, 0.148, 0.1459, 0.1358, 0.1478, 0.1064, 0.0869, 0.0809, 0.1078, 0.1202, 0.135, 0.1377, 0.1329, 0.0977, 0.0977, 0.0859, 0.1002, 0.1183, 0.1573, 0.1753, 0.193, 0.2383, 0.231, 0.233, 0.2428, 0.2499, 0.2534, 0.2813, 0.2633, 0.3004, 0.3484, 0.3389, 0.3196, 0.3038, 0.3032, 0.3321, 0.3759, 0.3742, 0.3936, 0.4423, 0.4024, 0.4525, 0.4303, 0.3387, 0.4447, 0.4115, 0.4534, 0.4301, 0.4074, 0.4199, 0.4047, 0.3779, 0.3867, 0.4069, 0.4226, 0.4382, 0.4552, 0.4463, 0.4329, 0.4541, 0.4025, 0.4006, 0.3796, 0.3665, 0.4083, 0.3888, 0.3862, 0.346, 0.3585, 0.3451, 0.3255, 0.3468, 0.3236, 0.3119, 0.3576, 0.3578, 0.383, 0.383, 0.3904, 0.3721, 0.3932, 0.4092, 0.3972, 0.4057, 0.4398, 0.4472, 0.4793, 0.458, 0.3967, 0.3761, 0.372]}
{"unit_id": "orig:1932967520908292430", "ticker_idx": 0, "ts": "2025-06-12T01:05:44+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains its \"Buy\" rating for Wistron, citing the rapid ramp-up of its AI server business and better-than-expected overall revenue.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GS Buy on Wistron (AI server ramp) and GS Neutral-to-positive on Inventec (stable margins, AI baseboard); adopting directional stances.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Wistron/Inventec: Wistron's Strong Growth, Inventec's Moderate Differentiation\nGoldman Sachs (V. Jeng, 2025/06/08)\n\nGoldman Sachs believes that Wistron benefited from the ASIC AI server and general-purpose server new product cycles, with May revenue increasing by 56% month-on-month. In contrast, Inventec experienced a 1% month-on-month decline due to exchange rate and product mix impacts. Both companies will face high base challenges in Q3, with Wistron showing stronger growth momentum in the AI server domain.\n\nWistron's Revenue and Shipments Soar\n\nGoldman Sachs points out that Wistron's May revenue was NT$208.4 billion, a month-on-month increase of 56% and a year-on-year increase of 162%, exceeding estimates by 11%. This growth primarily stemmed from: 1) Increased penetration of Wiwynn ASIC AI servers, driving its revenue up by 187% month-on-month; 2) Wistron's own rack-level AI server shipments reaching approximately 800–900 units in May (compared to only 100–200 units in March–April); 3) The new product cycle for general-purpose servers; 4) PC replacement and AI notebooks driving NB shipments up by 6% and desktop shipments up by 38%. Goldman Sachs maintains a cautious stance on rack-level AI servers but is optimistic about the flexibility and superior gross margins of ASIC solutions.\n\nInventec's Revenue and AI Strategy Balance\n\nGoldman Sachs estimates that Inventec's May revenue decreased by 1% month-on-month, mainly due to the appreciation of the NT dollar, limited exposure to rack-level AI servers, and a slowdown in PC and general-purpose server pull-ins in Q2. Looking ahead to Q3 2025, a high base, exchange rates, and earlier pull-ins are expected to lead to an approximately 8% month-on-month decline. Nevertheless, Inventec possesses cost and gross margin advantages in the H200/B200 baseboard-level AI server domain. With its comprehensive AI server product line covering both GPU and ASIC engines, and extensive coverage of US and Chinese CSP customers, Goldman Sachs holds a neutral-to-positive view on its long-term performance.\n\nAI Server Business Comparison and Strategy\nGoldman Sachs emphasizes that Wistron, leveraging Wiwynn's deep expertise in ASIC AI servers, has expanded from the component level to the system level, enabling it to quickly capture demand from hyperscale cloud service providers. Inventec, on the other hand, with its low budget threshold and high gross margin characteristics for baseboard-level solutions, possesses differentiated competitiveness in specialized AI scenarios. Both companies are actively deploying AI servers, but Wistron's system delivery and shipment pace are clearly faster, leading to more significant short-term growth.\n\nRatings and Outlook\nGoldman Sachs maintains its \"Buy\" rating for Wistron, citing the rapid ramp-up of its AI server business and better-than-expected overall revenue. It maintains a \"Neutral\" rating for Inventec due to its lower revenue base, relatively controllable month-on-month fluctuation risk in Q3, and its ability to achieve stable gross margins and cash flow through AI server baseboard solutions.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012875536480686733, "ret_signed_m1d": -0.012875536480686733, "alpha_spy_m1d": -0.008917042334274261, "alpha_c_m1d": -0.003682240078868193, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01716738197424883, "ret_signed_p1d": 0.01716738197424883, "alpha_spy_p1d": 0.02834744637071629, "alpha_c_p1d": 0.031183938840673164, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.008583690987124415, "ret_signed_p1w": -0.008583690987124415, "alpha_spy_p1w": 0.0018675188572383483, "alpha_c_p1w": -0.0046260109222659995, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": 0.025751072961373467, "ret_signed_p1m": 0.025751072961373467, "alpha_spy_p1m": -0.010213522785604834, "alpha_c_p1m": -0.030744085589712533, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": -0.0042918454935622075, "ret_signed_p3m": -0.0042918454935622075, "alpha_spy_p3m": -0.0846274048940322, "alpha_c_p3m": -0.09997454024361663, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 0.296137339055794, "ret_signed_p6m": 0.296137339055794, "alpha_spy_p6m": 0.15390333645715337, "alpha_c_p6m": 0.08385485314077101, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [-0.1202, -0.1077, -0.1451, -0.1285, -0.1243, -0.1409, -0.1409, -0.1243, -0.1326, -0.1658, -0.209, -0.1575, -0.1617, -0.1617, -0.1617, -0.2455, -0.3202, -0.38, -0.3186, -0.2505, -0.1924, -0.1492, -0.1725, -0.1658, -0.1285, -0.1451, -0.1799, -0.1492, -0.1617, -0.1492, -0.1492, -0.1368, -0.1575, -0.1575, -0.1243, -0.1409, -0.0953, -0.116, -0.1119, -0.1119, -0.0994, -0.0994, -0.0496, -0.0745, -0.0787, -0.0953, -0.0911, -0.0621, -0.0621, -0.0621, -0.0372, -0.0455, -0.0413, -0.033, -0.033, -0.0455, -0.0601, -0.0215, -0.0129, -0.0558, -0.0258, -0.0172, -0.0129, 0.0, 0.0172, 0.0215, 0.0129, -0.0043, -0.0086, 0.0172, 0.0172, 0.03, 0.0773, 0.073, 0.0386, 0.0515, 0.03, 0.03, 0.0386, 0.0215, 0.0129, 0.0172, 0.0343, 0.03, 0.0258, 0.0043, 0.03, 0.0129, 0.0129, 0.0215, 0.0172, -0.0172, -0.0043, 0.0043, 0.0, 0.03, 0.0343, 0.0343, 0.0558, 0.0558, 0.0258, 0.0644, 0.0515, 0.0601, 0.0687, 0.0944, 0.0815, 0.0086, 0.0, 0.0, -0.0043, -0.0086, -0.0258, -0.0429, -0.0386, -0.0258, -0.03, -0.0129, -0.0129, -0.03, -0.0601, -0.0558, -0.0558, -0.03, -0.0086, -0.0172, -0.0043, 0.03, 0.0215, 0.03, 0.0258, 0.0129, 0.0215, 0.03, 0.0386, 0.0429, 0.0515, 0.0429, 0.0773, 0.0987, 0.0987, 0.206, 0.2532, 0.2833, 0.3176, 0.3176, 0.3391, 0.2918, 0.2961, 0.2961, 0.2489, 0.1845, 0.1845, 0.1974, 0.1717, 0.2189, 0.2232, 0.2318, 0.2275, 0.2275, 0.2575, 0.2704, 0.309, 0.3305, 0.2918, 0.2318, 0.1931, 0.1803, 0.2189, 0.1803, 0.1888, 0.2403, 0.1931, 0.2361, 0.206, 0.1845, 0.176, 0.176, 0.2275, 0.1931, 0.1931, 0.1974, 0.2232, 0.2489, 0.2403, 0.2103, 0.2275, 0.2446, 0.2532, 0.2961]}
{"unit_id": "orig:1932967520908292430", "ticker_idx": 1, "ts": "2025-06-12T01:05:44+00:00", "symbol": "Inventec", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs holds a neutral-to-positive view on its long-term performance", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GS Buy on Wistron (AI server ramp) and GS Neutral-to-positive on Inventec (stable margins, AI baseboard); adopting directional stances.", "resolved_tickers": ["2356.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Inventec Corporation 2356.TW Taiwan server ODM", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Wistron/Inventec: Wistron's Strong Growth, Inventec's Moderate Differentiation\nGoldman Sachs (V. Jeng, 2025/06/08)\n\nGoldman Sachs believes that Wistron benefited from the ASIC AI server and general-purpose server new product cycles, with May revenue increasing by 56% month-on-month. In contrast, Inventec experienced a 1% month-on-month decline due to exchange rate and product mix impacts. Both companies will face high base challenges in Q3, with Wistron showing stronger growth momentum in the AI server domain.\n\nWistron's Revenue and Shipments Soar\n\nGoldman Sachs points out that Wistron's May revenue was NT$208.4 billion, a month-on-month increase of 56% and a year-on-year increase of 162%, exceeding estimates by 11%. This growth primarily stemmed from: 1) Increased penetration of Wiwynn ASIC AI servers, driving its revenue up by 187% month-on-month; 2) Wistron's own rack-level AI server shipments reaching approximately 800–900 units in May (compared to only 100–200 units in March–April); 3) The new product cycle for general-purpose servers; 4) PC replacement and AI notebooks driving NB shipments up by 6% and desktop shipments up by 38%. Goldman Sachs maintains a cautious stance on rack-level AI servers but is optimistic about the flexibility and superior gross margins of ASIC solutions.\n\nInventec's Revenue and AI Strategy Balance\n\nGoldman Sachs estimates that Inventec's May revenue decreased by 1% month-on-month, mainly due to the appreciation of the NT dollar, limited exposure to rack-level AI servers, and a slowdown in PC and general-purpose server pull-ins in Q2. Looking ahead to Q3 2025, a high base, exchange rates, and earlier pull-ins are expected to lead to an approximately 8% month-on-month decline. Nevertheless, Inventec possesses cost and gross margin advantages in the H200/B200 baseboard-level AI server domain. With its comprehensive AI server product line covering both GPU and ASIC engines, and extensive coverage of US and Chinese CSP customers, Goldman Sachs holds a neutral-to-positive view on its long-term performance.\n\nAI Server Business Comparison and Strategy\nGoldman Sachs emphasizes that Wistron, leveraging Wiwynn's deep expertise in ASIC AI servers, has expanded from the component level to the system level, enabling it to quickly capture demand from hyperscale cloud service providers. Inventec, on the other hand, with its low budget threshold and high gross margin characteristics for baseboard-level solutions, possesses differentiated competitiveness in specialized AI scenarios. Both companies are actively deploying AI servers, but Wistron's system delivery and shipment pace are clearly faster, leading to more significant short-term growth.\n\nRatings and Outlook\nGoldman Sachs maintains its \"Buy\" rating for Wistron, citing the rapid ramp-up of its AI server business and better-than-expected overall revenue. It maintains a \"Neutral\" rating for Inventec due to its lower revenue base, relatively controllable month-on-month fluctuation risk in Q3, and its ability to achieve stable gross margins and cash flow through AI server baseboard solutions.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02206737113149304, "ret_signed_m1d": -0.02206737113149304, "alpha_spy_m1d": -0.01810887698508057, "alpha_c_m1d": -0.0128740747296745, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017421623341502324, "ret_signed_p1d": 0.017421623341502324, "alpha_spy_p1d": 0.028601687737969783, "alpha_c_p1d": 0.03143818020792666, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.0429732822724761, "ret_signed_p1w": -0.0429732822724761, "alpha_spy_p1w": -0.032522072428113336, "alpha_c_p1w": -0.03901560220761768, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": 0.014818684675364402, "ret_signed_p1m": 0.014818684675364402, "alpha_spy_p1m": -0.0211459110716139, "alpha_c_p1m": -0.0416764738757216, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": 0.019651173036799863, "ret_signed_p3m": 0.019651173036799863, "alpha_spy_p3m": -0.06068438636367013, "alpha_c_p3m": -0.07603152171325456, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 0.13804663094569625, "ret_signed_p6m": 0.13804663094569625, "alpha_spy_p6m": -0.0041873716529443605, "alpha_c_p6m": -0.07423585496932672, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [0.0836, 0.1057, 0.0941, 0.0976, 0.1034, 0.0685, 0.0581, 0.065, 0.0383, 0.0232, -0.0197, 0.0058, -0.0046, -0.0046, -0.0046, -0.1034, -0.1475, -0.2125, -0.1347, -0.0708, -0.0325, -0.0395, -0.0616, -0.0616, -0.0755, -0.0743, -0.0859, -0.0639, -0.0732, -0.0662, -0.0546, -0.036, -0.0557, -0.0557, -0.0314, -0.0523, -0.0407, -0.0639, -0.0418, -0.0383, -0.0105, -0.0035, -0.0197, -0.0302, -0.0407, -0.0372, -0.0407, -0.0139, -0.0221, -0.036, -0.0105, -0.0337, -0.0279, -0.0209, -0.0209, -0.0616, -0.0581, -0.0337, -0.0337, -0.0453, -0.0383, -0.0279, -0.0221, 0.0, 0.0174, 0.0197, -0.0058, -0.0186, -0.043, -0.0627, -0.0662, -0.0465, -0.0407, -0.0221, -0.0128, -0.0174, -0.0058, 0.0221, 0.0302, 0.0221, 0.0139, -0.0105, 0.0221, 0.0076, 0.0148, 0.0064, 0.0583, 0.0269, 0.0257, 0.0438, 0.0486, 0.0233, 0.0329, 0.0317, 0.0462, 0.0571, 0.0547, 0.0668, 0.0752, 0.0752, 0.0547, 0.0752, 0.0619, 0.068, 0.0341, 0.0354, 0.0172, 0.0124, 0.0088, -0.0057, 0.0076, -0.0118, -0.0045, -0.0118, -0.0118, -0.0093, 0.0003, 0.0003, -0.0009, -0.0057, -0.0118, -0.0202, -0.0202, -0.0021, 0.0076, 0.0112, 0.0197, 0.0148, 0.0076, 0.0486, 0.0535, 0.045, 0.0329, 0.0462, 0.0764, 0.0788, 0.0873, 0.0788, 0.138, 0.0849, 0.0849, 0.1054, 0.0885, 0.0994, 0.1054, 0.1054, 0.1235, 0.1332, 0.1199, 0.1199, 0.0933, 0.0402, 0.0486, 0.0583, 0.0595, 0.0897, 0.097, 0.1115, 0.1078, 0.1078, 0.1296, 0.1356, 0.1489, 0.1284, 0.1091, 0.1006, 0.0704, 0.0523, 0.0438, 0.0233, 0.0462, 0.0148, 0.0341, 0.0547, 0.0209, 0.0052, -0.013, -0.0178, 0.0076, -0.0142, -0.0081, -0.0081, 0.0486, 0.0511, 0.0535, 0.0644, 0.0668, 0.0897, 0.1127, 0.138]}
{"unit_id": "orig:1927959450784014724", "ticker_idx": 0, "ts": "2025-05-29T05:25:27+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raises NVIDIA's target price from $160 to $170", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Morgan Stanley raises NVDA PT from $160 to $170; author shares/adopts bullish stance.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley raises NVIDIA’s target price from $160 to $170. https://t.co/v1H3XCeChg", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031467771192964955, "ret_signed_m1d": -0.031467771192964955, "alpha_spy_m1d": -0.027536013589865016, "alpha_c_m1d": -0.02442091775738242, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029168709381678815, "ret_signed_p1d": -0.029168709381678815, "alpha_spy_p1d": -0.028050357189775443, "alpha_c_p1d": -0.011428601389960602, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.005747599702493611, "ret_signed_p1w": 0.005747599702493611, "alpha_spy_p1w": 0.0006632770562644552, "alpha_c_p1w": -0.023832855030635036, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.13342157350479011, "ret_signed_p1m": 0.13342157350479011, "alpha_spy_p1m": 0.08820871516440842, "alpha_c_p1m": -0.007270048081713831, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.30600340555452554, "ret_signed_p3m": 0.30600340555452554, "alpha_spy_p3m": 0.20937208449485212, "alpha_c_p3m": 0.09021349223987807, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.2853115202986203, "ret_signed_p6m": 0.2853115202986203, "alpha_spy_p6m": 0.16199846496754655, "alpha_c_p6m": -0.05084882377385047, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [-0.1806, -0.1668, -0.1573, -0.2057, -0.1905, -0.2315, -0.2187, -0.1685, -0.1696, -0.1259, -0.1412, -0.1707, -0.1557, -0.1484, -0.1544, -0.1277, -0.1329, -0.1827, -0.1994, -0.2121, -0.2214, -0.2086, -0.2067, -0.2686, -0.3224, -0.2985, -0.3081, -0.1786, -0.2272, -0.203, -0.2046, -0.1939, -0.2493, -0.2709, -0.2709, -0.3038, -0.2895, -0.2621, -0.2354, -0.2025, -0.2188, -0.2168, -0.2175, -0.1981, -0.1774, -0.1823, -0.1843, -0.159, -0.1568, -0.1619, -0.1163, -0.0665, -0.0277, -0.0313, -0.0272, -0.026, -0.0346, -0.0531, -0.0457, -0.0568, -0.0568, -0.0265, -0.0315, 0.0, -0.0292, -0.013, 0.0146, 0.0196, 0.0057, 0.0182, 0.0247, 0.0343, 0.0262, 0.0418, 0.02, 0.0396, 0.0355, 0.0453, 0.0453, 0.0336, 0.0359, 0.0627, 0.1087, 0.1138, 0.1334, 0.1351, 0.1014, 0.1298, 0.1448, 0.1448, 0.1369, 0.1496, 0.1703, 0.179, 0.1849, 0.1788, 0.2265, 0.2313, 0.243, 0.2388, 0.2314, 0.2001, 0.227, 0.2483, 0.2466, 0.2699, 0.261, 0.288, 0.278, 0.2482, 0.2933, 0.2808, 0.2891, 0.2988, 0.3127, 0.3081, 0.316, 0.3047, 0.3078, 0.2965, 0.3077, 0.262, 0.2602, 0.2572, 0.2788, 0.2919, 0.306, 0.3048, 0.2945, 0.2515, 0.2515, 0.227, 0.2259, 0.2334, 0.2, 0.2093, 0.2269, 0.2741, 0.273, 0.2777, 0.2772, 0.2566, 0.2236, 0.2663, 0.2694, 0.3193, 0.2821, 0.2716, 0.2768, 0.2804, 0.3067, 0.3406, 0.3454, 0.3572, 0.3481, 0.3332, 0.3296, 0.3588, 0.3837, 0.3161, 0.3531, 0.2936, 0.2921, 0.3064, 0.3165, 0.3123, 0.3017, 0.2954, 0.3089, 0.3383, 0.3759, 0.4445, 0.4877, 0.4578, 0.455, 0.4865, 0.4277, 0.4026, 0.3514, 0.3519, 0.4302, 0.3879, 0.3925, 0.3427, 0.3664, 0.3408, 0.3031, 0.3402, 0.298, 0.2853]}
{"unit_id": "orig:1935880040354660439", "ticker_idx": 0, "ts": "2025-06-20T01:59:03+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If Meta successfully transitions to lower-cost SRAM-based ASICs, Meta's own EPS is highly likely to improve", "tweet_type": "analysis", "timeline": "months", "summary": "MTIA v2 mass production by year-end: bullish META on EPS improvement, bullish Broadcom and Celestica as ASIC/server beneficiaries.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Meta Completes Development of Its Own Chip, MTIA v2 - By BK Tech Insight\n\nMeta's Self-Developed Chip (MTIA v2)\n- Meta's inference chip test has been successfully completed.\n- Meta plans to begin mass production of its self-developed AI chip, MTIA v2, by the end of this year.\n- Broadcom, a U.S. company, is primarily responsible for the chip's design.\n\nTechnical Features\n- MTIA v2 is for AI inference and accelerating ad recommendation algorithms.\n- It uses an 8x8 PE (Processing Element) structure internally.\n- Adopts TSMC's 5nm process, with significantly increased SRAM capacity and LPDDR5 memory.\n- Maximizes efficiency by incorporating a large amount of SRAM instead of costly HBM.\n\nImplications\n- Meta's completion of MTIA v2 development is very important and has significant implications.\n\n1. Expansion of SRAM-centric Inference Chips Without HBM\n\n- Tesla and Grok have successfully developed and are operating HBM-less SRAM-centric inference chips, but Grok's low market share has kept them from gaining much attention.\n- However, the largest AI workload currently is Meta's advertising/recommendation algorithms.\n- Therefore, Meta's shift from GPU-based inference workloads to its own SRAM-based ASICs is highly significant.\n- It is crucial to closely observe whether SRAM-based ASICs can successfully establish themselves.\n- Attention to the ASIC and ASIC server value chains is warranted.\n\n2. Re-evaluating Meta's Internal Margin Improvement\n\n- Meta's cost of revenue for advertising/recommendation algorithms is essentially the GPU price and GPU operating costs.\n- If Meta successfully transitions to lower-cost SRAM-based ASICs, Meta's own EPS is highly likely to improve.\n- The estimated cost of Meta's inference chip is around $5,000, significantly lower than general GPU prices. However, for Meta's advertising/recommendation algorithm inference, its performance is not significantly inferior, suggesting excellent cost-effectiveness.\n\nConclusion\n- It is estimated that Meta, with the single largest AI workload, has successfully tested its SRAM-based ASIC.\n- Mass production is expected to begin by the end of the year. If the transition is successful, the beneficiaries in the semiconductor sector are likely to change.\n- Furthermore, if the proliferation of inexpensive inference ASICs succeeds, Broadcom and Celestica (a big tech ASIC server assembly company) are expected to benefit.\n- The EPS improvement of big tech companies and solution providers transitioning to inference chips should also be followed up on.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019667497711689963, "ret_signed_m1d": 0.019667497711689963, "alpha_spy_m1d": 0.01731332946981201, "alpha_c_m1d": 0.013541397920367126, "bench_spy_m1d": 0.002354168241877952, "bench_c_m1d": 0.0061260997913228366, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023712353162844035, "ret_signed_p1d": 0.023712353162844035, "alpha_spy_p1d": 0.013834872658348374, "alpha_c_p1d": 0.011050294859408227, "bench_spy_p1d": 0.009877480504495662, "bench_c_p1d": 0.012662058303435808, "ret_p1w": 0.07515223796237702, "ret_signed_p1w": 0.07515223796237702, "alpha_spy_p1w": 0.040438070176231466, "alpha_c_p1w": 0.025547315631994083, "bench_spy_p1w": 0.034714167786145556, "bench_c_p1w": 0.04960492233038294, "ret_p1m": 0.04487436910646392, "ret_signed_p1m": 0.04487436910646392, "alpha_spy_p1m": -0.013162211699540238, "alpha_c_p1m": -0.004048131318641568, "bench_spy_p1m": 0.05803658080600416, "bench_c_p1m": 0.048922500425105486, "ret_p3m": 0.13683605944612354, "ret_signed_p3m": 0.13683605944612354, "alpha_spy_p3m": 0.027628298324442024, "alpha_c_p3m": -0.021452850619592123, "bench_spy_p3m": 0.10920776112168151, "bench_c_p3m": 0.15828891006571566, "ret_p6m": -0.049643755233257036, "ret_signed_p6m": -0.049643755233257036, "alpha_spy_p6m": -0.19828955695447525, "alpha_c_p6m": -0.18986605983947713, "bench_spy_p6m": 0.14864580172121822, "bench_c_p6m": 0.1402223046062201, "price_path": [-0.0828, -0.1053, -0.1176, -0.1554, -0.156, -0.1419, -0.1449, -0.2215, -0.2609, -0.244, -0.2525, -0.1422, -0.2, -0.204, -0.2217, -0.2363, -0.2644, -0.2656, -0.2656, -0.2903, -0.2674, -0.2381, -0.2193, -0.1986, -0.195, -0.1881, -0.196, -0.1621, -0.1257, -0.1224, -0.1399, -0.126, -0.1243, -0.1324, -0.0636, -0.0393, -0.0344, -0.0571, -0.0623, -0.0622, -0.067, -0.0694, -0.0678, -0.0817, -0.0817, -0.0594, -0.0575, -0.0554, -0.0518, -0.0175, -0.0235, 0.0074, 0.0026, 0.0217, 0.0164, 0.0286, 0.0165, 0.0154, -0.0, 0.029, 0.0218, 0.0197, 0.0197, 0.0, 0.0237, 0.0437, 0.0386, 0.0641, 0.0752, 0.0817, 0.054, 0.0458, 0.0537, 0.0537, 0.0528, 0.0562, 0.0739, 0.0658, 0.0515, 0.0565, 0.0411, 0.0301, 0.0279, 0.0321, 0.0449, 0.0329, 0.0458, 0.0476, 0.0444, 0.0517, 0.0259, 0.0188, 0.1335, 0.0992, 0.1378, 0.1189, 0.1314, 0.1165, 0.1274, 0.1224, 0.1578, 0.1432, 0.1462, 0.1508, 0.1246, 0.1013, 0.0958, 0.0832, 0.1062, 0.104, 0.1052, 0.0953, 0.1008, 0.0826, 0.0826, 0.0773, 0.0802, 0.0972, 0.1027, 0.1025, 0.1222, 0.102, 0.1005, 0.1073, 0.1207, 0.1416, 0.1368, 0.1435, 0.1407, 0.1221, 0.1078, 0.1155, 0.0983, 0.0907, 0.0902, 0.077, 0.052, 0.0662, 0.042, 0.0495, 0.0457, 0.0527, 0.0757, 0.0343, 0.0496, 0.0392, 0.0523, 0.0443, 0.0514, 0.0737, 0.0753, 0.0756, 0.0764, 0.0828, 0.1011, 0.102, 0.1023, -0.0226, -0.0492, -0.0648, -0.08, -0.0674, -0.0923, -0.0883, -0.0735, -0.0804, -0.1069, -0.1056, -0.1062, -0.1171, -0.1235, -0.1343, -0.136, -0.1285, -0.101, -0.067, -0.0708, -0.0708, -0.0498, -0.0602, -0.051, -0.062, -0.0299, -0.0124, -0.0221, -0.0366, -0.0466, -0.0428, -0.0552, -0.0496]}
{"unit_id": "orig:1935880040354660439", "ticker_idx": 1, "ts": "2025-06-20T01:59:03+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom and Celestica (a big tech ASIC server assembly company) are expected to benefit", "tweet_type": "analysis", "timeline": "months", "summary": "MTIA v2 mass production by year-end: bullish META on EPS improvement, bullish Broadcom and Celestica as ASIC/server beneficiaries.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Meta Completes Development of Its Own Chip, MTIA v2 - By BK Tech Insight\n\nMeta's Self-Developed Chip (MTIA v2)\n- Meta's inference chip test has been successfully completed.\n- Meta plans to begin mass production of its self-developed AI chip, MTIA v2, by the end of this year.\n- Broadcom, a U.S. company, is primarily responsible for the chip's design.\n\nTechnical Features\n- MTIA v2 is for AI inference and accelerating ad recommendation algorithms.\n- It uses an 8x8 PE (Processing Element) structure internally.\n- Adopts TSMC's 5nm process, with significantly increased SRAM capacity and LPDDR5 memory.\n- Maximizes efficiency by incorporating a large amount of SRAM instead of costly HBM.\n\nImplications\n- Meta's completion of MTIA v2 development is very important and has significant implications.\n\n1. Expansion of SRAM-centric Inference Chips Without HBM\n\n- Tesla and Grok have successfully developed and are operating HBM-less SRAM-centric inference chips, but Grok's low market share has kept them from gaining much attention.\n- However, the largest AI workload currently is Meta's advertising/recommendation algorithms.\n- Therefore, Meta's shift from GPU-based inference workloads to its own SRAM-based ASICs is highly significant.\n- It is crucial to closely observe whether SRAM-based ASICs can successfully establish themselves.\n- Attention to the ASIC and ASIC server value chains is warranted.\n\n2. Re-evaluating Meta's Internal Margin Improvement\n\n- Meta's cost of revenue for advertising/recommendation algorithms is essentially the GPU price and GPU operating costs.\n- If Meta successfully transitions to lower-cost SRAM-based ASICs, Meta's own EPS is highly likely to improve.\n- The estimated cost of Meta's inference chip is around $5,000, significantly lower than general GPU prices. However, for Meta's advertising/recommendation algorithm inference, its performance is not significantly inferior, suggesting excellent cost-effectiveness.\n\nConclusion\n- It is estimated that Meta, with the single largest AI workload, has successfully tested its SRAM-based ASIC.\n- Mass production is expected to begin by the end of the year. If the transition is successful, the beneficiaries in the semiconductor sector are likely to change.\n- Furthermore, if the proliferation of inexpensive inference ASICs succeeds, Broadcom and Celestica (a big tech ASIC server assembly company) are expected to benefit.\n- The EPS improvement of big tech companies and solution providers transitioning to inference chips should also be followed up on.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0027200261364130895, "ret_signed_m1d": 0.0027200261364130895, "alpha_spy_m1d": 0.0003658578945351376, "alpha_c_m1d": -0.006193800987139841, "bench_spy_m1d": 0.002354168241877952, "bench_c_m1d": 0.008913827123552931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015120580417935736, "ret_signed_p1d": 0.015120580417935736, "alpha_spy_p1d": 0.005243099913440075, "alpha_c_p1d": 0.008780949325656096, "bench_spy_p1d": 0.009877480504495662, "bench_c_p1d": 0.006339631092279641, "ret_p1w": 0.07744318157053831, "ret_signed_p1w": 0.07744318157053831, "alpha_spy_p1w": 0.042729013784392755, "alpha_c_p1w": 0.0077078276474005225, "bench_spy_p1w": 0.034714167786145556, "bench_c_p1w": 0.06973535392313779, "ret_p1m": 0.1528859232370552, "ret_signed_p1m": 0.1528859232370552, "alpha_spy_p1m": 0.09484934243105103, "alpha_c_p1m": 0.036698833784037044, "bench_spy_p1m": 0.05803658080600416, "bench_c_p1m": 0.11618708945301814, "ret_p3m": 0.3847353550051851, "ret_signed_p3m": 0.3847353550051851, "alpha_spy_p3m": 0.27552759388350356, "alpha_c_p3m": 0.21191477668027803, "bench_spy_p3m": 0.10920776112168151, "bench_c_p3m": 0.17282057832490705, "ret_p6m": 0.3616232667115493, "ret_signed_p6m": 0.3616232667115493, "alpha_spy_p6m": 0.2129774649903311, "alpha_c_p6m": 0.005723551912615532, "bench_spy_p6m": 0.14864580172121822, "bench_c_p6m": 0.3558997147989338, "price_path": [-0.2487, -0.2846, -0.3136, -0.3251, -0.3318, -0.3275, -0.3132, -0.3854, -0.4162, -0.3849, -0.3773, -0.2611, -0.3124, -0.2739, -0.2882, -0.2859, -0.3032, -0.3176, -0.3176, -0.3367, -0.3232, -0.294, -0.2491, -0.2325, -0.2319, -0.2371, -0.2319, -0.2125, -0.1873, -0.199, -0.2015, -0.1827, -0.1708, -0.1691, -0.1157, -0.0725, -0.0737, -0.0716, -0.0877, -0.0796, -0.0754, -0.0832, -0.08, -0.0872, -0.0872, -0.0596, -0.0445, -0.0344, -0.034, -0.0075, 0.025, 0.0419, 0.0373, -0.0146, -0.0251, -0.0237, 0.0093, 0.0219, -0.0075, 0.0061, -0.0048, 0.0027, 0.0027, 0.0, 0.0151, 0.0551, 0.0586, 0.0807, 0.0774, 0.1026, 0.059, 0.0796, 0.1008, 0.1008, 0.0968, 0.0872, 0.1116, 0.1016, 0.0976, 0.1024, 0.1238, 0.1233, 0.1458, 0.1334, 0.1529, 0.1144, 0.1348, 0.1549, 0.1608, 0.1772, 0.1897, 0.2105, 0.1748, 0.1546, 0.1909, 0.1718, 0.2067, 0.2151, 0.2199, 0.2156, 0.2514, 0.2364, 0.245, 0.2254, 0.2231, 0.1797, 0.1647, 0.1584, 0.176, 0.177, 0.1921, 0.201, 0.2346, 0.1896, 0.1896, 0.193, 0.2096, 0.2244, 0.3396, 0.3827, 0.3467, 0.4783, 0.4386, 0.4395, 0.4564, 0.4401, 0.3847, 0.3815, 0.3798, 0.3575, 0.3581, 0.3596, 0.3468, 0.3405, 0.3139, 0.322, 0.3359, 0.3551, 0.3559, 0.3443, 0.348, 0.3844, 0.3825, 0.3008, 0.4293, 0.3789, 0.4078, 0.4191, 0.3998, 0.3994, 0.373, 0.3636, 0.3796, 0.419, 0.4507, 0.4945, 0.5466, 0.5085, 0.4811, 0.4527, 0.4102, 0.4384, 0.4249, 0.4002, 0.4361, 0.4103, 0.4234, 0.3623, 0.3722, 0.373, 0.3644, 0.4202, 0.3897, 0.3632, 0.5145, 0.5428, 0.5931, 0.5931, 0.6147, 0.547, 0.529, 0.5251, 0.5268, 0.5637, 0.6072, 0.628, 0.6548, 0.6283, 0.4422, 0.3616]}
{"unit_id": "orig:1935880040354660439", "ticker_idx": 2, "ts": "2025-06-20T01:59:03+00:00", "symbol": "Celestica", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom and Celestica (a big tech ASIC server assembly company) are expected to benefit", "tweet_type": "analysis", "timeline": "months", "summary": "MTIA v2 mass production by year-end: bullish META on EPS improvement, bullish Broadcom and Celestica as ASIC/server beneficiaries.", "resolved_tickers": ["CLS"], "resolution_method": "exchange_ticker", "resolution_reason": "Celestica Inc. CLS US/NYSE", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Meta Completes Development of Its Own Chip, MTIA v2 - By BK Tech Insight\n\nMeta's Self-Developed Chip (MTIA v2)\n- Meta's inference chip test has been successfully completed.\n- Meta plans to begin mass production of its self-developed AI chip, MTIA v2, by the end of this year.\n- Broadcom, a U.S. company, is primarily responsible for the chip's design.\n\nTechnical Features\n- MTIA v2 is for AI inference and accelerating ad recommendation algorithms.\n- It uses an 8x8 PE (Processing Element) structure internally.\n- Adopts TSMC's 5nm process, with significantly increased SRAM capacity and LPDDR5 memory.\n- Maximizes efficiency by incorporating a large amount of SRAM instead of costly HBM.\n\nImplications\n- Meta's completion of MTIA v2 development is very important and has significant implications.\n\n1. Expansion of SRAM-centric Inference Chips Without HBM\n\n- Tesla and Grok have successfully developed and are operating HBM-less SRAM-centric inference chips, but Grok's low market share has kept them from gaining much attention.\n- However, the largest AI workload currently is Meta's advertising/recommendation algorithms.\n- Therefore, Meta's shift from GPU-based inference workloads to its own SRAM-based ASICs is highly significant.\n- It is crucial to closely observe whether SRAM-based ASICs can successfully establish themselves.\n- Attention to the ASIC and ASIC server value chains is warranted.\n\n2. Re-evaluating Meta's Internal Margin Improvement\n\n- Meta's cost of revenue for advertising/recommendation algorithms is essentially the GPU price and GPU operating costs.\n- If Meta successfully transitions to lower-cost SRAM-based ASICs, Meta's own EPS is highly likely to improve.\n- The estimated cost of Meta's inference chip is around $5,000, significantly lower than general GPU prices. However, for Meta's advertising/recommendation algorithm inference, its performance is not significantly inferior, suggesting excellent cost-effectiveness.\n\nConclusion\n- It is estimated that Meta, with the single largest AI workload, has successfully tested its SRAM-based ASIC.\n- Mass production is expected to begin by the end of the year. If the transition is successful, the beneficiaries in the semiconductor sector are likely to change.\n- Furthermore, if the proliferation of inexpensive inference ASICs succeeds, Broadcom and Celestica (a big tech ASIC server assembly company) are expected to benefit.\n- The EPS improvement of big tech companies and solution providers transitioning to inference chips should also be followed up on.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007542736853412668, "ret_signed_m1d": 0.007542736853412668, "alpha_spy_m1d": 0.005188568611534716, "alpha_c_m1d": 0.003094415470310974, "bench_spy_m1d": 0.002354168241877952, "bench_c_m1d": 0.004448321383101694, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008208242589817116, "ret_signed_p1d": 0.008208242589817116, "alpha_spy_p1d": -0.0016692379146785452, "alpha_c_p1d": -0.0031934319312443282, "bench_spy_p1d": 0.009877480504495662, "bench_c_p1d": 0.011401674521061445, "ret_p1w": 0.12896548837096988, "ret_signed_p1w": 0.12896548837096988, "alpha_spy_p1w": 0.09425132058482433, "alpha_c_p1w": 0.08470824735227267, "bench_spy_p1w": 0.034714167786145556, "bench_c_p1w": 0.04425724101869721, "ret_p1m": 0.20550180746077285, "ret_signed_p1m": 0.20550180746077285, "alpha_spy_p1m": 0.1474652266547687, "alpha_c_p1m": 0.1176870628049469, "bench_spy_p1m": 0.05803658080600416, "bench_c_p1m": 0.08781474465582595, "ret_p3m": 0.8313984432942962, "ret_signed_p3m": 0.8313984432942962, "alpha_spy_p3m": 0.7221906821726147, "alpha_c_p3m": 0.7011089760771967, "bench_spy_p3m": 0.10920776112168151, "bench_c_p3m": 0.13028946721709955, "ret_p6m": 1.2404793356406758, "ret_signed_p6m": 1.2404793356406758, "alpha_spy_p6m": 1.0918335339194576, "alpha_c_p6m": 1.0538240063187958, "bench_spy_p6m": 0.14864580172121822, "bench_c_p6m": 0.18665532932188, "price_path": [-0.28, -0.3518, -0.376, -0.3885, -0.4172, -0.4154, -0.391, -0.4884, -0.5108, -0.4795, -0.4909, -0.3932, -0.4344, -0.4148, -0.4019, -0.3894, -0.4043, -0.3785, -0.3785, -0.4041, -0.3879, -0.3479, -0.3186, -0.3415, -0.351, -0.3563, -0.3689, -0.3403, -0.3044, -0.317, -0.2961, -0.286, -0.2877, -0.3119, -0.2582, -0.1875, -0.1611, -0.1756, -0.1625, -0.1747, -0.1669, -0.1568, -0.125, -0.1165, -0.1165, -0.1209, -0.1265, -0.1395, -0.1468, -0.1375, -0.1033, -0.1134, -0.1177, -0.1072, -0.093, -0.1015, -0.0558, -0.0333, -0.0732, -0.0517, -0.0226, 0.0075, 0.0075, 0.0, 0.0082, 0.0637, 0.1021, 0.1103, 0.129, 0.1544, 0.1012, 0.1455, 0.1777, 0.1777, 0.1388, 0.1514, 0.1898, 0.1745, 0.1908, 0.1997, 0.1868, 0.1839, 0.2045, 0.1841, 0.2055, 0.1603, 0.2482, 0.2126, 0.2587, 0.282, 0.4938, 0.4968, 0.4779, 0.4401, 0.4897, 0.4532, 0.4688, 0.4879, 0.5356, 0.513, 0.5784, 0.5102, 0.4541, 0.4455, 0.4411, 0.366, 0.3525, 0.341, 0.397, 0.4265, 0.4688, 0.4945, 0.5701, 0.4401, 0.4401, 0.4836, 0.5667, 0.6368, 0.7946, 0.8068, 0.8049, 0.8711, 0.8172, 0.7878, 0.8412, 0.8812, 0.8314, 0.8638, 0.8704, 0.8841, 0.9052, 0.7782, 0.8091, 0.8028, 0.8188, 0.8219, 0.8612, 0.8554, 0.7279, 0.7403, 0.7584, 0.8848, 0.921, 0.8023, 0.9305, 0.8321, 0.9963, 1.0695, 1.0444, 1.0225, 0.9573, 1.0032, 1.0907, 1.1934, 1.2319, 1.414, 1.4977, 1.5152, 1.5474, 1.5809, 1.4831, 1.6075, 1.5199, 1.3819, 1.5483, 1.454, 1.4741, 1.1648, 1.2989, 1.2877, 1.2179, 1.3107, 1.0874, 1.071, 1.3851, 1.4236, 1.4569, 1.4569, 1.5468, 1.3543, 1.2502, 1.2806, 1.387, 1.404, 1.5199, 1.5292, 1.5783, 1.5986, 1.2665, 1.2405]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 0, "ts": "2025-07-10T13:38:03+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy: Meta", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007617953523383347, "ret_signed_m1d": 0.007617953523383347, "alpha_spy_m1d": 0.010430331962077544, "alpha_c_m1d": 0.004244014879513092, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033739386438702557, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013379201227002313, "ret_signed_p1d": -0.013379201227002313, "alpha_spy_p1d": -0.009863777356406755, "alpha_c_p1d": -0.006912527618104791, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.006466673608897522, "ret_p1w": -0.035517837617023096, "ret_signed_p1w": -0.035517837617023096, "alpha_spy_p1w": -0.03906512868402823, "alpha_c_p1w": -0.03533051292700795, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": -0.00018732469001514485, "ret_p1m": 0.057835153865735345, "ret_signed_p1m": 0.057835153865735345, "alpha_spy_p1m": 0.03968304799976141, "alpha_c_p1m": 0.04911913968451964, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.008716014181215703, "ret_p3m": -0.018808999475233068, "ret_signed_p3m": -0.018808999475233068, "alpha_spy_p3m": -0.09096243019039096, "alpha_c_p3m": -0.11032505415440019, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.09151605467916712, "ret_p6m": -0.10431244343697266, "ret_signed_p6m": -0.10431244343697266, "alpha_spy_p6m": -0.20221329022903278, "alpha_c_p6m": -0.2072878007801222, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.10297535734314955, "price_path": [-0.2697, -0.2834, -0.3098, -0.311, -0.311, -0.3341, -0.3126, -0.2851, -0.2674, -0.248, -0.2447, -0.2382, -0.2457, -0.2138, -0.1797, -0.1766, -0.193, -0.18, -0.1783, -0.1859, -0.1214, -0.0986, -0.094, -0.1153, -0.1202, -0.12, -0.1246, -0.1268, -0.1253, -0.1384, -0.1384, -0.1174, -0.1157, -0.1137, -0.1103, -0.0782, -0.0837, -0.0548, -0.0593, -0.0413, -0.0464, -0.0349, -0.0462, -0.0473, -0.0617, -0.0345, -0.0413, -0.0433, -0.0433, -0.0617, -0.0395, -0.0207, -0.0255, -0.0016, 0.0088, 0.0149, -0.011, -0.0188, -0.0113, -0.0113, -0.0122, -0.009, 0.0076, 0.0, -0.0134, -0.0087, -0.0232, -0.0335, -0.0355, -0.0316, -0.0196, -0.0308, -0.0188, -0.0171, -0.02, -0.0132, -0.0375, -0.044, 0.0635, 0.0313, 0.0676, 0.0498, 0.0615, 0.0476, 0.0578, 0.0531, 0.0863, 0.0727, 0.0755, 0.0797, 0.0552, 0.0333, 0.0282, 0.0163, 0.0379, 0.0358, 0.0369, 0.0277, 0.0328, 0.0158, 0.0158, 0.0108, 0.0135, 0.0294, 0.0347, 0.0345, 0.0529, 0.034, 0.0325, 0.039, 0.0515, 0.0712, 0.0667, 0.0729, 0.0703, 0.0529, 0.0394, 0.0467, 0.0305, 0.0234, 0.0229, 0.0105, -0.0129, 0.0004, -0.0223, -0.0153, -0.0188, -0.0123, 0.0093, -0.0295, -0.0152, -0.0249, -0.0127, -0.0202, -0.0135, 0.0075, 0.009, 0.0092, 0.01, 0.016, 0.0331, 0.034, 0.0343, -0.0829, -0.1079, -0.1225, -0.1368, -0.1249, -0.1483, -0.1445, -0.1307, -0.1371, -0.162, -0.1608, -0.1614, -0.1716, -0.1776, -0.1877, -0.1893, -0.1823, -0.1564, -0.1246, -0.1282, -0.1282, -0.1084, -0.1182, -0.1096, -0.1199, -0.0897, -0.0734, -0.0825, -0.096, -0.1054, -0.1019, -0.1135, -0.1083, -0.095, -0.1056, -0.085, -0.0928, -0.089, -0.0843, -0.0807, -0.0807, -0.0866, -0.0929, -0.0829, -0.091, -0.091, -0.1043]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 1, "ts": "2025-07-10T13:38:03+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Buy: Google", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005629858925053188, "ret_signed_m1d": -0.005629858925053188, "alpha_spy_m1d": -0.002817480486358992, "alpha_c_m1d": -0.009003797568923444, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033739386438702557, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014469203241549211, "ret_signed_p1d": 0.014469203241549211, "alpha_spy_p1d": 0.01798462711214477, "alpha_c_p1d": 0.020935876850446733, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.006466673608897522, "ret_p1w": 0.03355494762802791, "ret_signed_p1w": 0.03355494762802791, "alpha_spy_p1w": 0.030007656561022777, "alpha_c_p1w": 0.033742272318043054, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": -0.00018732469001514485, "ret_p1m": 0.13399403477929783, "ret_signed_p1m": 0.13399403477929783, "alpha_spy_p1m": 0.1158419289133239, "alpha_c_p1m": 0.12527802059808213, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.008716014181215703, "ret_p3m": 0.3848655664217364, "ret_signed_p3m": 0.3848655664217364, "alpha_spy_p3m": 0.3127121357065785, "alpha_c_p3m": 0.2933495117425693, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.09151605467916712, "ret_p6m": 0.7770421917833696, "ret_signed_p6m": 0.7770421917833696, "alpha_spy_p6m": 0.6791413449913095, "alpha_c_p6m": 0.6740668344402201, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.10297535734314955, "price_path": [-0.1055, -0.121, -0.1378, -0.15, -0.15, -0.1696, -0.1483, -0.1264, -0.1043, -0.0893, -0.0969, -0.0994, -0.107, -0.093, -0.0776, -0.0766, -0.0821, -0.1488, -0.1325, -0.1411, -0.1089, -0.1029, -0.0701, -0.078, -0.0655, -0.0635, -0.0779, -0.0522, -0.0392, -0.0527, -0.0527, -0.0278, -0.0308, -0.0336, -0.0343, -0.0495, -0.0655, -0.055, -0.0541, -0.0234, -0.0086, 0.0055, -0.0015, -0.0108, -0.0166, -0.0048, -0.0094, -0.0242, -0.0242, -0.0618, -0.07, -0.0611, -0.0391, -0.023, 0.0051, -0.0078, -0.01, 0.0057, 0.0108, 0.0108, -0.0047, -0.0184, -0.0056, 0.0, 0.0145, 0.0222, 0.0247, 0.0301, 0.0336, 0.0419, 0.0703, 0.0772, 0.071, 0.0819, 0.0876, 0.0842, 0.1021, 0.1065, 0.0804, 0.0648, 0.0981, 0.096, 0.104, 0.1064, 0.134, 0.1316, 0.1448, 0.137, 0.1426, 0.148, 0.1457, 0.1348, 0.1222, 0.1246, 0.1603, 0.1738, 0.1662, 0.1681, 0.1915, 0.1987, 0.1987, 0.1899, 0.2986, 0.3078, 0.323, 0.3188, 0.3503, 0.3477, 0.3545, 0.3569, 0.4178, 0.4153, 0.4061, 0.4202, 0.4354, 0.423, 0.4181, 0.3926, 0.385, 0.3893, 0.3752, 0.3699, 0.38, 0.3845, 0.3826, 0.4112, 0.3849, 0.3784, 0.361, 0.3331, 0.3758, 0.3831, 0.4146, 0.417, 0.4274, 0.4457, 0.4114, 0.4183, 0.4261, 0.4647, 0.5173, 0.5072, 0.5472, 0.5861, 0.5845, 0.5988, 0.5639, 0.6021, 0.6046, 0.5712, 0.6347, 0.6415, 0.6156, 0.5698, 0.5576, 0.6061, 0.6019, 0.65, 0.6311, 0.6886, 0.7952, 0.8226, 0.8029, 0.8029, 0.8042, 0.7744, 0.7796, 0.8011, 0.7898, 0.8104, 0.769, 0.7879, 0.8056, 0.7617, 0.744, 0.738, 0.7287, 0.6731, 0.7055, 0.732, 0.7468, 0.7725, 0.7711, 0.7711, 0.7678, 0.7681, 0.7697, 0.7649, 0.7649, 0.777]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 2, "ts": "2025-07-10T13:38:03+00:00", "symbol": "V", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy: Visa", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["V"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLF", "bench_c_reason": "industry='Credit Services'", "bench_c_industry": "Credit Services", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00528284796524825, "ret_signed_m1d": 0.00528284796524825, "alpha_spy_m1d": 0.008095226403942446, "alpha_c_m1d": 0.0111640769189435, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.00588122895369525, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02233890243675829, "ret_signed_p1d": -0.02233890243675829, "alpha_spy_p1d": -0.018823478566162732, "alpha_c_p1d": -0.011904395457582262, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.010434506979176028, "ret_p1w": -0.017056227331087026, "ret_signed_p1w": -0.017056227331087026, "alpha_spy_p1w": -0.020603518398092158, "alpha_c_p1w": -0.013261841182848566, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": -0.00379438614823846, "ret_p1m": -0.05366978721418214, "ret_signed_p1m": -0.05366978721418214, "alpha_spy_p1m": -0.07182189308015607, "alpha_c_p1m": -0.03735405128131086, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": -0.016315735932871278, "ret_p3m": -0.007979803087012738, "ret_signed_p3m": -0.007979803087012738, "alpha_spy_p3m": -0.08013323380217063, "alpha_c_p3m": -0.03144795947264256, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.023468156385629824, "ret_p6m": -0.02276837205603932, "ret_signed_p6m": -0.02276837205603932, "alpha_spy_p6m": -0.12066921884809945, "alpha_c_p6m": -0.07197037476306123, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.04920200270702191, "price_path": [-0.0597, -0.058, -0.0705, -0.0754, -0.0754, -0.102, -0.0703, -0.062, -0.0582, -0.0598, -0.0532, -0.0419, -0.0308, -0.0393, -0.0249, -0.022, -0.0246, -0.0186, -0.0146, -0.011, -0.0017, 0.0007, 0.0016, 0.018, 0.026, 0.0338, 0.0308, 0.0068, 0.0059, -0.0066, -0.0066, 0.0096, 0.0108, 0.0183, 0.0262, 0.0265, 0.028, 0.0341, 0.0306, 0.0403, 0.0303, 0.0416, 0.049, 0.0436, -0.0085, -0.0011, 0.0055, -0.0436, -0.0436, -0.0486, -0.0341, -0.0119, -0.0298, -0.0277, -0.0204, -0.0023, -0.0012, -0.0047, 0.0084, 0.0084, 0.0021, -0.0037, 0.0053, 0.0, -0.0223, -0.0151, -0.0249, -0.0168, -0.0171, -0.0192, -0.0139, -0.0113, -0.0017, -0.0054, 0.0033, -0.0012, -0.0129, -0.014, -0.0293, -0.0464, -0.0386, -0.0518, -0.0454, -0.0668, -0.0537, -0.0561, -0.0521, -0.0358, -0.0275, -0.0304, -0.0352, -0.036, -0.0319, -0.0326, -0.0147, -0.0181, -0.0115, -0.0138, -0.0152, -0.0098, -0.0098, -0.0146, -0.0123, -0.012, -0.0339, -0.0365, -0.0317, -0.0482, -0.0331, -0.0445, -0.0456, -0.0429, -0.0255, -0.0481, -0.0384, -0.0307, -0.0466, -0.0467, -0.0572, -0.0503, -0.0425, -0.0391, -0.0209, -0.0262, -0.0152, -0.0168, -0.008, -0.011, -0.0231, -0.0327, -0.0337, -0.0194, -0.0269, -0.0559, -0.0376, -0.0305, -0.0226, -0.0279, -0.0262, -0.0222, -0.0209, -0.0235, -0.0393, -0.0288, -0.0409, -0.0517, -0.0421, -0.0428, -0.0515, -0.0541, -0.0574, -0.046, -0.0442, -0.0521, -0.0692, -0.0812, -0.0941, -0.0858, -0.0868, -0.0749, -0.0712, -0.0565, -0.0586, -0.0586, -0.0567, -0.0681, -0.0703, -0.0703, -0.0774, -0.0658, -0.0782, -0.0791, -0.0813, -0.0252, -0.019, -0.0216, -0.0266, -0.0286, -0.0241, -0.015, -0.0069, -0.0033, 0.0017, 0.0017, 0.0013, 0.0002, -0.0026, -0.0108, -0.0108, -0.0228]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 3, "ts": "2025-07-10T13:38:03+00:00", "symbol": "WMT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy: Walmart", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["WMT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLP", "bench_c_reason": "industry='Discount Stores'", "bench_c_industry": "Discount Stores", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02055651407197079, "ret_signed_m1d": 0.02055651407197079, "alpha_spy_m1d": 0.023368892510664985, "alpha_c_m1d": 0.024131930620671316, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.0035754165487005274, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004849235199058444, "ret_signed_p1d": -0.004849235199058444, "alpha_spy_p1d": -0.001333811328462886, "alpha_c_p1d": -0.001150452282679404, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.0036987829163790398, "ret_p1w": 0.0024244959489547746, "ret_signed_p1w": 0.0024244959489547746, "alpha_spy_p1w": -0.001122795118050357, "alpha_c_p1w": 0.0024244959489547746, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.0, "ret_p1m": 0.09350622745510728, "ret_signed_p1m": 0.09350622745510728, "alpha_spy_p1m": 0.07535412158913335, "alpha_c_p1m": 0.07661559562825482, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.016890631826852465, "ret_p3m": 0.09088263006595576, "ret_signed_p3m": 0.09088263006595576, "alpha_spy_p3m": 0.018729199350797865, "alpha_c_p3m": 0.12039224923445879, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": -0.029509619168503032, "ret_p6m": 0.19390428435481155, "ret_signed_p6m": 0.19390428435481155, "alpha_spy_p6m": 0.09600343756275143, "alpha_c_p6m": 0.2217593541220927, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": -0.027855069767281138, "price_path": [-0.0038, -0.0118, -0.041, -0.0197, -0.0197, -0.0282, -0.0025, -0.0014, 0.0079, 0.0, 0.0014, 0.01, 0.0227, 0.0244, 0.0385, 0.0446, 0.0364, 0.0393, 0.0246, 0.0196, 0.0199, 0.0108, 0.0208, 0.0157, 0.0356, 0.0344, 0.031, 0.0166, 0.0113, 0.0156, 0.0156, 0.0287, 0.0251, 0.0236, 0.0407, 0.0518, 0.054, 0.0473, 0.0327, 0.0275, 0.0273, 0.0259, 0.0099, -0.0003, -0.0044, -0.006, -0.0064, 0.0024, 0.0024, 0.0133, 0.0317, 0.0331, 0.0253, 0.0122, 0.0254, 0.0308, 0.0356, 0.029, 0.0369, 0.0369, 0.0473, 0.0235, 0.0206, 0.0, -0.0048, 0.0097, 0.0056, 0.0031, 0.0024, 0.002, 0.0085, 0.0104, 0.0086, 0.0183, 0.0275, 0.029, 0.0366, 0.0288, 0.0329, 0.0383, 0.0499, 0.0469, 0.0896, 0.0871, 0.0935, 0.0956, 0.0923, 0.0646, 0.0631, 0.0566, 0.064, 0.0703, 0.0838, 0.0351, 0.0232, 0.0151, 0.0149, 0.0152, 0.0155, 0.0247, 0.0247, 0.0339, 0.0507, 0.0665, 0.062, 0.0807, 0.0808, 0.061, 0.0846, 0.0935, 0.0956, 0.0928, 0.1018, 0.0947, 0.0813, 0.0864, 0.0833, 0.0855, 0.0889, 0.09, 0.0892, 0.089, 0.0774, 0.0746, 0.0785, 0.0852, 0.0909, 0.0873, 0.0753, 0.0761, 0.079, 0.1328, 0.1521, 0.125, 0.1383, 0.1311, 0.1224, 0.1321, 0.1291, 0.1218, 0.1039, 0.0901, 0.0826, 0.0802, 0.0691, 0.0734, 0.0806, 0.0722, 0.0744, 0.084, 0.0822, 0.093, 0.093, 0.0835, 0.0829, 0.0878, 0.0713, 0.0631, 0.1318, 0.1129, 0.0995, 0.1306, 0.1528, 0.1528, 0.1677, 0.1785, 0.1878, 0.2089, 0.2135, 0.2163, 0.1999, 0.2158, 0.1959, 0.2206, 0.2356, 0.2366, 0.2221, 0.2246, 0.2158, 0.2108, 0.1922, 0.1742, 0.1817, 0.1817, 0.1831, 0.1915, 0.185, 0.1796, 0.1796, 0.1939]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 4, "ts": "2025-07-10T13:38:03+00:00", "symbol": "ETN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy: Eaton", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["ETN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0059835781776784724, "ret_signed_m1d": 0.0059835781776784724, "alpha_spy_m1d": 0.008795956616372669, "alpha_c_m1d": 0.011556238726504464, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.005572660548825992, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008332364383492719, "ret_signed_p1d": 0.008332364383492719, "alpha_spy_p1d": 0.011847788254088276, "alpha_c_p1d": 0.011914781716394929, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00358241733290221, "ret_p1w": 0.06453421739539889, "ret_signed_p1w": 0.06453421739539889, "alpha_spy_p1w": 0.060986926328393753, "alpha_c_p1w": 0.05750224289391337, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.007031974501485516, "ret_p1m": 0.01749427067943965, "ret_signed_p1m": 0.01749427067943965, "alpha_spy_p1m": -0.0006578351865342835, "alpha_c_p1m": 0.01689726930982416, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.0005970013696154908, "ret_p3m": 0.04020878559984298, "ret_signed_p3m": 0.04020878559984298, "alpha_spy_p3m": -0.03194464511531492, "alpha_c_p3m": 0.012696094590344043, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.027512691009498935, "ret_p6m": -0.0796065882374064, "ret_signed_p6m": -0.0796065882374064, "alpha_spy_p6m": -0.17750743502946653, "alpha_c_p6m": -0.13557048054723708, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.055963892309830676, "price_path": [-0.2259, -0.2278, -0.25, -0.2524, -0.2524, -0.277, -0.2563, -0.2319, -0.2002, -0.1952, -0.1982, -0.19, -0.1798, -0.1595, -0.1649, -0.1651, -0.1665, -0.155, -0.1415, -0.1336, -0.0994, -0.0752, -0.0798, -0.0862, -0.0799, -0.0793, -0.0822, -0.0987, -0.1023, -0.1023, -0.1023, -0.0824, -0.0894, -0.0918, -0.1047, -0.1084, -0.0889, -0.0854, -0.0871, -0.0732, -0.089, -0.0934, -0.0893, -0.0763, -0.095, -0.0549, -0.0759, -0.0637, -0.0637, -0.0738, -0.069, -0.0402, -0.0428, -0.0266, -0.0123, -0.0018, -0.0073, 0.0015, 0.0128, 0.0128, 0.0024, -0.0018, 0.006, 0.0, 0.0083, 0.0074, 0.0125, 0.0147, 0.0645, 0.0587, 0.0448, 0.042, 0.0632, 0.0762, 0.0965, 0.0982, 0.0905, 0.0907, 0.0757, 0.0661, 0.0758, -0.0033, 0.0015, 0.01, 0.0175, 0.0098, 0.0188, 0.0025, -0.0042, -0.0156, -0.0087, -0.0213, -0.0291, -0.0315, -0.0252, -0.0304, -0.0128, -0.0146, -0.0035, -0.0209, -0.0209, -0.036, -0.0382, -0.0235, -0.0212, -0.0199, -0.0235, 0.0158, 0.0098, 0.0261, 0.0531, 0.0409, 0.0189, 0.0411, 0.0502, 0.0609, 0.0334, 0.0438, 0.0228, 0.0252, 0.0296, 0.0495, 0.0483, 0.0565, 0.0473, 0.0657, 0.0402, 0.0564, 0.0577, 0.035, 0.0526, 0.0498, 0.0704, 0.0532, 0.0468, 0.0591, 0.0473, 0.0112, 0.0443, 0.0552, 0.0649, 0.0544, 0.0873, 0.0743, 0.07, 0.084, 0.0592, 0.0839, 0.0612, 0.051, 0.0673, 0.0346, 0.0388, -0.0044, -0.0091, -0.0362, -0.0487, -0.028, -0.0771, -0.0672, -0.0708, -0.0533, -0.0392, -0.0392, -0.0274, -0.0447, -0.0633, -0.0564, -0.0469, -0.0505, -0.0344, -0.039, -0.0061, -0.0148, -0.0665, -0.063, -0.0722, -0.1119, -0.1116, -0.1063, -0.0991, -0.0923, -0.0898, -0.0898, -0.0941, -0.0961, -0.0977, -0.1044, -0.1044, -0.0796]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 5, "ts": "2025-07-10T13:38:03+00:00", "symbol": "CMI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy: CMI", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["CMI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015969110974857537, "ret_signed_m1d": -0.015969110974857537, "alpha_spy_m1d": -0.01315673253616334, "alpha_c_m1d": -0.010396450426031545, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.005572660548825992, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0035583875378836227, "ret_signed_p1d": -0.0035583875378836227, "alpha_spy_p1d": -4.296366728806511e-05, "alpha_c_p1d": 2.4029795018587485e-05, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00358241733290221, "ret_p1w": 0.02393921333400928, "ret_signed_p1w": 0.02393921333400928, "alpha_spy_p1w": 0.020391922267004148, "alpha_c_p1w": 0.016907238832523763, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.007031974501485516, "ret_p1m": 0.1382231533826157, "ret_signed_p1m": 0.1382231533826157, "alpha_spy_p1m": 0.12007104751664177, "alpha_c_p1m": 0.1376261520130002, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.0005970013696154908, "ret_p3m": 0.28336940080339335, "ret_signed_p3m": 0.28336940080339335, "alpha_spy_p3m": 0.21121597008823545, "alpha_c_p3m": 0.2558567097938944, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.027512691009498935, "ret_p6m": 0.5497388223974065, "ret_signed_p6m": 0.5497388223974065, "alpha_spy_p6m": 0.45183797560534633, "alpha_c_p6m": 0.4937749300875758, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.055963892309830676, "price_path": [-0.1561, -0.1597, -0.1796, -0.1759, -0.1759, -0.1981, -0.1838, -0.1705, -0.1404, -0.1424, -0.1396, -0.1427, -0.1407, -0.1377, -0.1232, -0.116, -0.1276, -0.1287, -0.104, -0.1031, -0.0333, -0.0216, -0.0267, -0.0288, -0.0219, -0.0155, -0.0305, -0.0548, -0.059, -0.0671, -0.0671, -0.0405, -0.05, -0.0446, -0.0546, -0.067, -0.0458, -0.0506, -0.0497, -0.0396, -0.0422, -0.0383, -0.0484, -0.045, -0.0603, -0.0521, -0.0742, -0.0736, -0.0736, -0.0758, -0.0697, -0.0476, -0.0626, -0.0466, -0.0421, -0.0368, -0.0306, -0.0194, -0.0241, -0.0241, -0.0358, -0.0273, -0.016, 0.0, -0.0036, 0.0006, 0.0056, 0.0163, 0.0239, 0.0305, 0.0135, 0.0506, 0.0646, 0.0742, 0.0867, 0.0902, 0.0774, 0.0749, 0.0811, 0.0465, 0.0634, 0.0976, 0.1225, 0.1311, 0.1382, 0.142, 0.1736, 0.1964, 0.1905, 0.1747, 0.1821, 0.1829, 0.1656, 0.1593, 0.1892, 0.1708, 0.1778, 0.1865, 0.188, 0.1778, 0.1778, 0.1683, 0.1591, 0.1851, 0.1819, 0.1747, 0.1669, 0.1979, 0.215, 0.2127, 0.2243, 0.228, 0.2172, 0.2604, 0.2522, 0.2558, 0.2538, 0.2355, 0.2239, 0.2357, 0.2398, 0.2485, 0.2552, 0.2787, 0.2793, 0.2973, 0.2834, 0.2898, 0.2613, 0.221, 0.209, 0.2316, 0.2515, 0.2494, 0.2173, 0.2238, 0.2355, 0.206, 0.2348, 0.2458, 0.2446, 0.2246, 0.3043, 0.295, 0.2937, 0.2938, 0.2673, 0.2982, 0.368, 0.3992, 0.4317, 0.407, 0.4173, 0.3819, 0.3674, 0.3478, 0.3743, 0.3989, 0.3751, 0.4027, 0.4416, 0.4693, 0.4758, 0.4758, 0.4783, 0.4729, 0.4888, 0.5075, 0.5162, 0.516, 0.5072, 0.4848, 0.5572, 0.5538, 0.5142, 0.5355, 0.5275, 0.4767, 0.4861, 0.5043, 0.5416, 0.5351, 0.5366, 0.5366, 0.5411, 0.5278, 0.5247, 0.5154, 0.5154, 0.5497]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 6, "ts": "2025-07-10T13:38:03+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sell: Apple", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005978871432844146, "ret_signed_m1d": 0.005978871432844146, "alpha_spy_m1d": 0.0031664929941499498, "alpha_c_m1d": 0.009286529937209798, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00588468076829296, "ret_signed_p1d": 0.00588468076829296, "alpha_spy_p1d": 0.0023692568976974027, "alpha_c_p1d": 0.001487386632192389, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": -0.011251673488998915, "ret_signed_p1w": 0.011251673488998915, "alpha_spy_p1w": 0.014798964556004046, "alpha_c_p1w": 0.02712820776208469, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.07975165398013617, "ret_signed_p1m": -0.07975165398013617, "alpha_spy_p1m": -0.06159954811416224, "alpha_c_p1m": -0.04496290360804611, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.20884659492453683, "ret_signed_p3m": -0.20884659492453683, "alpha_spy_p3m": -0.13669316420937894, "alpha_c_p3m": -0.09573619224583485, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.27856788070932637, "ret_signed_p6m": -0.27856788070932637, "alpha_spy_p6m": -0.18066703391726624, "alpha_c_p6m": -0.15237445783513381, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.0478, -0.0496, -0.0866, -0.0739, -0.0739, -0.0918, -0.0609, -0.038, -0.0203, -0.016, -0.012, -0.007, -0.0009, 0.003, -0.0345, -0.0649, -0.0667, -0.0773, -0.0715, -0.0666, -0.0076, 0.0024, -0.0004, -0.0045, -0.0054, -0.0171, -0.0261, -0.0486, -0.052, -0.0807, -0.0807, -0.0574, -0.0564, -0.0587, -0.0544, -0.0504, -0.043, -0.0451, -0.0555, -0.04, -0.0516, -0.0459, -0.0642, -0.0622, -0.0751, -0.0659, -0.079, -0.0745, -0.0745, -0.0537, -0.0514, -0.057, -0.0511, -0.0537, -0.0533, -0.0341, -0.0216, 0.0001, 0.0054, 0.0054, -0.0116, -0.0113, -0.006, 0.0, -0.0059, -0.0178, -0.0155, -0.0106, -0.0113, -0.0058, 0.0003, 0.0094, 0.0082, 0.0064, 0.0069, 0.0077, -0.0054, -0.0158, -0.0228, -0.0472, -0.0427, -0.0447, 0.004, 0.0359, 0.0798, 0.0707, 0.0824, 0.0997, 0.0971, 0.0915, 0.0882, 0.0867, 0.0652, 0.06, 0.0735, 0.0707, 0.0808, 0.0864, 0.0961, 0.0941, 0.0941, 0.0827, 0.124, 0.1301, 0.1297, 0.1212, 0.1045, 0.0689, 0.0842, 0.1032, 0.1156, 0.1225, 0.1264, 0.1212, 0.1571, 0.207, 0.1992, 0.1892, 0.2107, 0.204, 0.1992, 0.2001, 0.204, 0.2119, 0.2161, 0.2098, 0.2088, 0.2163, 0.1973, 0.156, 0.1673, 0.1678, 0.1752, 0.1663, 0.1891, 0.236, 0.2385, 0.2181, 0.2235, 0.2387, 0.267, 0.2679, 0.2712, 0.2792, 0.2743, 0.2681, 0.2728, 0.2732, 0.2715, 0.2654, 0.2711, 0.2986, 0.2902, 0.2877, 0.2852, 0.2618, 0.2617, 0.267, 0.2561, 0.2808, 0.3017, 0.3067, 0.3094, 0.3094, 0.3156, 0.3356, 0.3502, 0.3406, 0.3243, 0.3152, 0.311, 0.3077, 0.3152, 0.3117, 0.3129, 0.2932, 0.2956, 0.2825, 0.2841, 0.2911, 0.2784, 0.2849, 0.2918, 0.2918, 0.2898, 0.2915, 0.2883, 0.2826, 0.2826, 0.2786]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 7, "ts": "2025-07-10T13:38:03+00:00", "symbol": "F", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sell: Ford", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["F"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Auto Manufacturers'", "bench_c_industry": "Auto Manufacturers", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006717067075641681, "ret_signed_m1d": 0.006717067075641681, "alpha_spy_m1d": 0.0039046886369474842, "alpha_c_m1d": -0.004214789330177826, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.010931856405819507, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010915233997917717, "ret_signed_p1d": 0.010915233997917717, "alpha_spy_p1d": 0.0073998101273221595, "alpha_c_p1d": 0.011186369160633536, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": 0.0002711351627158187, "ret_p1w": -0.0638118011299803, "ret_signed_p1w": 0.0638118011299803, "alpha_spy_p1w": 0.06735909219698544, "alpha_c_p1w": 0.058616973426635854, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": -0.005194827703344451, "ret_p1m": -0.049538117616395705, "ret_signed_p1m": 0.049538117616395705, "alpha_spy_p1m": 0.06769022348236964, "alpha_c_p1m": 0.06042491192664745, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.010886794310251746, "ret_p3m": 0.014279733108669257, "ret_signed_p3m": -0.014279733108669257, "alpha_spy_p3m": 0.05787369760648864, "alpha_c_p3m": 0.05233154184359967, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.06661127495226893, "ret_p6m": 0.1482360851120894, "ret_signed_p6m": -0.1482360851120894, "alpha_spy_p6m": -0.050335238320029285, "alpha_c_p6m": -0.07497676453756852, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.07325932057452089, "price_path": [-0.1964, -0.218, -0.2221, -0.2031, -0.2031, -0.2163, -0.2014, -0.1907, -0.1675, -0.1691, -0.1708, -0.16, -0.1716, -0.1575, -0.1493, -0.1584, -0.136, -0.1493, -0.1493, -0.1369, -0.1142, -0.11, -0.1058, -0.0974, -0.0932, -0.0974, -0.0999, -0.1201, -0.1234, -0.1301, -0.1301, -0.1343, -0.1461, -0.1419, -0.1285, -0.162, -0.1444, -0.1402, -0.152, -0.1385, -0.1318, -0.11, -0.105, -0.1159, -0.1243, -0.1083, -0.1251, -0.1243, -0.1243, -0.1108, -0.0974, -0.0991, -0.1201, -0.1075, -0.0932, -0.089, -0.047, -0.0118, -0.0084, -0.0084, -0.0269, -0.0193, -0.0067, 0.0, -0.0109, -0.0025, -0.0285, -0.0563, -0.0638, -0.063, -0.0504, -0.0605, -0.0445, -0.0546, -0.0369, -0.0529, -0.0697, -0.0873, -0.0705, -0.0915, -0.0806, -0.0714, -0.0588, -0.0529, -0.0495, -0.0521, -0.0427, -0.0266, -0.0266, -0.0266, -0.0249, -0.0147, -0.0223, -0.0359, -0.001, 0.0058, 0.0075, 0.0134, -0.0036, 0.0015, 0.0015, -0.0027, -0.0112, -0.0061, -0.001, -0.0053, -0.0223, -0.0283, 0.0007, -0.0061, -0.0061, -0.0121, -0.0078, -0.001, -0.0112, -0.0078, -0.0019, -0.0104, -0.0112, 0.0219, 0.0287, 0.0177, 0.0441, 0.0398, 0.0781, 0.0807, 0.0143, -0.001, -0.0215, -0.0291, -0.0181, -0.0087, 0.0007, -0.001, 0.0143, 0.0202, 0.0687, 0.0577, 0.05, 0.1777, 0.1283, 0.1172, 0.1266, 0.1113, 0.1172, 0.107, 0.0883, 0.1155, 0.1164, 0.137, 0.1327, 0.1448, 0.1577, 0.1431, 0.1353, 0.1086, 0.1207, 0.1104, 0.0682, 0.1043, 0.1155, 0.1336, 0.1353, 0.1353, 0.1431, 0.1327, 0.1155, 0.1267, 0.131, 0.1216, 0.131, 0.1259, 0.1543, 0.1732, 0.1844, 0.1749, 0.1766, 0.1457, 0.1465, 0.1594, 0.1586, 0.1439, 0.15, 0.15, 0.1457, 0.1431, 0.1388, 0.1293, 0.1293, 0.1482]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 8, "ts": "2025-07-10T13:38:03+00:00", "symbol": "LUV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sell: Southwest", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["LUV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Airlines'", "bench_c_industry": "Airlines", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07523410554215892, "ret_signed_m1d": 0.07523410554215892, "alpha_spy_m1d": 0.07242172710346473, "alpha_c_m1d": 0.06966144499333293, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.005572660548825992, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00589019365377863, "ret_signed_p1d": 0.00589019365377863, "alpha_spy_p1d": 0.0023747697831830727, "alpha_c_p1d": 0.00230777632087642, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00358241733290221, "ret_p1w": -0.02356087821538999, "ret_signed_p1w": 0.02356087821538999, "alpha_spy_p1w": 0.027108169282395123, "alpha_c_p1w": 0.030592852716875507, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.007031974501485516, "ret_p1m": -0.20696107861976254, "ret_signed_p1m": 0.20696107861976254, "alpha_spy_p1m": 0.22511318448573647, "alpha_c_p1m": 0.20755807998937803, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.0005970013696154908, "ret_p3m": -0.12793048918806804, "ret_signed_p3m": 0.12793048918806804, "alpha_spy_p3m": 0.20008391990322594, "alpha_c_p3m": 0.15544318019756698, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.027512691009498935, "ret_p6m": 0.11680899963159752, "ret_signed_p6m": -0.11680899963159752, "alpha_spy_p6m": -0.018908152839537395, "alpha_c_p6m": -0.060845107321766845, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.055963892309830676, "price_path": [-0.3196, -0.3129, -0.3382, -0.3414, -0.3414, -0.36, -0.3411, -0.3206, -0.2956, -0.2948, -0.2906, -0.2743, -0.2557, -0.2392, -0.2054, -0.1993, -0.2056, -0.1939, -0.1729, -0.17, -0.1676, -0.167, -0.1646, -0.1521, -0.1303, -0.1324, -0.1348, -0.1545, -0.1692, -0.1761, -0.1761, -0.1306, -0.1245, -0.1026, -0.1114, -0.1186, -0.1322, -0.141, -0.1359, -0.1087, -0.1064, -0.0853, -0.1085, -0.1327, -0.1553, -0.1428, -0.1631, -0.1657, -0.1657, -0.1596, -0.1465, -0.147, -0.1609, -0.1561, -0.139, -0.1315, -0.1055, -0.09, -0.0876, -0.0876, -0.0948, -0.0902, -0.0752, 0.0, -0.0059, 0.0094, -0.0059, 0.0016, -0.0236, -0.0225, -0.0131, -0.0048, 0.0024, -0.1095, -0.1108, -0.1293, -0.1673, -0.1778, -0.1719, -0.2021, -0.2035, -0.1989, -0.2046, -0.2086, -0.207, -0.222, -0.1775, -0.1708, -0.178, -0.1566, -0.1617, -0.1542, -0.1652, -0.1681, -0.1226, -0.1293, -0.1058, -0.1114, -0.117, -0.1191, -0.1191, -0.1365, -0.1295, -0.1621, -0.1549, -0.1462, -0.1656, -0.1624, -0.146, -0.1476, -0.1538, -0.15, -0.1352, -0.1312, -0.1244, -0.1304, -0.1024, -0.1099, -0.1403, -0.1242, -0.118, -0.1409, -0.1309, -0.1301, -0.1244, -0.1228, -0.1279, -0.1263, -0.147, -0.1575, -0.1411, -0.1002, -0.0964, -0.1037, -0.0975, -0.073, -0.0663, -0.091, -0.1479, -0.133, -0.1317, -0.1629, -0.1745, -0.1899, -0.1842, -0.1807, -0.2012, -0.1487, -0.1516, -0.1263, -0.1207, -0.1387, -0.1024, -0.1104, -0.1164, -0.1425, -0.1395, -0.1406, -0.161, -0.1169, -0.104, -0.0749, -0.056, -0.056, -0.0628, -0.0639, -0.0461, -0.0394, -0.0359, 0.0191, 0.0218, 0.0272, 0.0724, 0.0969, 0.1085, 0.1233, 0.1354, 0.1033, 0.0966, 0.1109, 0.1373, 0.119, 0.1168, 0.1168, 0.1163, 0.1119, 0.116, 0.1176, 0.1176, 0.1168]}
{"unit_id": "orig:1943303706285678793", "ticker_idx": 9, "ts": "2025-07-10T13:38:03+00:00", "symbol": "HSY", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sell: Hershey", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs Q3 2025 Top Ideas: Buy META, GOOGL, V, WMT, ETN, CMI; Sell AAPL, F, LUV, HSY.", "resolved_tickers": ["HSY"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLP", "bench_c_reason": "industry='Confectioners'", "bench_c_industry": "Confectioners", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Q3 2025 Top Ideas\n\nBuy:\nMeta\nGoogle\nVisa\nWalmart\nEaton\nCMI\n\nSell:\nApple\nFord\nSouthwest\nHershey https://t.co/zdghjL1MSv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0163386779632122, "ret_signed_m1d": 0.0163386779632122, "alpha_spy_m1d": 0.013526299524518004, "alpha_c_m1d": 0.012763261414511673, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.0035754165487005274, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0007287046935207098, "ret_signed_p1d": -0.0007287046935207098, "alpha_spy_p1d": -0.004244128564116267, "alpha_c_p1d": -0.0044274876098997495, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.0036987829163790398, "ret_p1w": 0.040026639626128846, "ret_signed_p1w": -0.040026639626128846, "alpha_spy_p1w": -0.036479348559123714, "alpha_c_p1w": -0.040026639626128846, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.0, "ret_p1m": 0.1192297952209993, "ret_signed_p1m": -0.1192297952209993, "alpha_spy_p1m": -0.10107768935502537, "alpha_c_p1m": -0.10233916339414684, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.016890631826852465, "ret_p3m": 0.1897394675492068, "ret_signed_p3m": -0.1897394675492068, "alpha_spy_p3m": -0.11758603683404889, "alpha_c_p3m": -0.21924908671770982, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": -0.029509619168503032, "ret_p6m": 0.1250181006847666, "ret_signed_p6m": -0.1250181006847666, "alpha_spy_p6m": -0.027117253892706472, "alpha_c_p6m": -0.15287317045204774, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": -0.027855069767281138, "price_path": [0.0257, 0.0116, -0.0108, 0.0034, 0.0034, 0.0027, 0.0085, 0.0034, -0.0123, -0.0166, -0.0179, -0.0058, 0.007, 0.0066, -0.0161, 0.0038, 0.024, 0.0159, 0.0246, 0.0293, 0.0093, -0.0164, -0.026, -0.0141, -0.0374, -0.0491, -0.0543, -0.0811, -0.0655, -0.0564, -0.0564, -0.0389, -0.0531, -0.0202, -0.024, -0.0194, -0.0095, -0.0192, -0.0113, -0.0176, 0.0138, 0.0237, 0.0207, 0.0454, 0.0272, 0.035, 0.0257, 0.0309, 0.0309, 0.0414, 0.0516, 0.045, 0.0176, 0.0143, 0.0143, 0.008, 0.0678, 0.0756, 0.0719, 0.0719, 0.0663, 0.0321, -0.0163, 0.0, 0.0007, 0.0016, -0.0105, -0.0022, 0.04, 0.0404, 0.0736, 0.1036, 0.1129, 0.117, 0.1365, 0.121, 0.1311, 0.1469, 0.1305, 0.148, 0.1565, 0.1476, 0.1455, 0.1388, 0.1192, 0.0655, 0.0719, 0.0937, 0.0949, 0.0894, 0.0919, 0.1064, 0.1114, 0.0978, 0.1104, 0.1058, 0.1225, 0.1118, 0.1016, 0.1246, 0.1246, 0.137, 0.1304, 0.1341, 0.1352, 0.1478, 0.1487, 0.1359, 0.1356, 0.1323, 0.1357, 0.184, 0.1695, 0.1572, 0.164, 0.1772, 0.1674, 0.1683, 0.1339, 0.1513, 0.1268, 0.1448, 0.1557, 0.1569, 0.1943, 0.1946, 0.1897, 0.1969, 0.1787, 0.1759, 0.158, 0.1594, 0.1727, 0.1433, 0.149, 0.1397, 0.1392, 0.1274, 0.1113, 0.0979, 0.1105, 0.1032, 0.0728, 0.0476, 0.0382, -0.0066, 0.0157, 0.0397, 0.026, 0.0442, 0.0338, 0.0571, 0.091, 0.0903, 0.0968, 0.104, 0.1282, 0.1254, 0.1266, 0.1472, 0.1459, 0.1518, 0.1607, 0.1607, 0.16, 0.1415, 0.1272, 0.114, 0.1253, 0.1243, 0.115, 0.1026, 0.1105, 0.1199, 0.1214, 0.1602, 0.1604, 0.1653, 0.1605, 0.1673, 0.1495, 0.1263, 0.1398, 0.1398, 0.1361, 0.1149, 0.1237, 0.1224, 0.1224, 0.125]}
{"unit_id": "thread:1932673938624311511", "ticker_idx": 0, "ts": "2025-06-11T05:39:08+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Upgraded to \"Hold\" given stronger-than-expected memory price increases in CY3Q25 and rapid yield improvements for Micron's HBM3e-12Hi", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF Securities upgrade of MU to Hold; near-term beat on pricing/HBM yields but cautious from 4Q25 on China competition and cycle risk.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[GF Securities Overseas Electronics and Communications] Micron (MU US): Upgraded to \"Hold\" on Price Increase and Rapid HBM Yield Improvement\n\nUpgraded to \"Hold\": We upgraded Micron to \"Hold\" given stronger-than-expected memory price increases in CY3Q25 and rapid yield improvements for Micron's HBM3e-12Hi. Our target price is based on FY25E 2.5x P/B. Despite short-term strength, we remain cautious about the memory cycle starting from CY4Q25 due to rising influence from Chinese memory suppliers and sluggish end-market demand. On the other hand, memory manufacturers will achieve revenue growth for 10 consecutive quarters through CY3Q, marking the longest upcycle. Considering Micron's elevated valuation amid the AI-driven boom, we see an unfavorable risk-reward profile.\n\nStrong Earnings Expectations: We forecast better-than-expected memory price increases for the third fiscal quarter, driven by restocking in PCs/servers and AI demand. We predict the contract average selling price (ASP) of non-HBM DRAM to rise by 7% QoQ (previously expected 4%), with server DDR5 prices up by 8–10%. Hence, we estimate Micron’s FY3Q25 revenue to reach USD 9.2 billion, exceeding the upper end of guidance (USD 9 billion). For FY4Q25, we expect revenue to reach USD 10.6 billion (above consensus of USD 9.8 billion). Profit margins are also anticipated to beat expectations, benefiting from favorable pricing and increased HBM revenue share.\n\nRapid HBM Yield Improvements; Samsung Faces Further Setbacks: Micron significantly improved its HBM yields using Hanmi’s TCB tool in CY2Q25, achieving a 70% yield for HBM3e-12Hi and 75% yield for HBM3e-8Hi. With increased market share among NVIDIA, AWS, and AMD, we raise our 2025 HBM revenue forecast for Micron to USD 7.3 billion, above consensus expectations of USD 6.0–7.0 billion. Conversely, Samsung failed to pass its third round of HBM certification in June, with the next round expected in September. We expect HBM pricing to remain stable in 2026.\n\nSlowing Price Momentum and Rising Competition from China in 4Q: We expect price momentum in CY4Q25 to slow significantly due to weak end-market demand and potential PC memory destocking. Additionally, Chinese memory manufacturers' influence on the DDR5 market will significantly increase in 2026. ChangXin Memory (CXMT) is expected to completely exit DDR4 and shift to DDR5 by 1H26, increasing its DDR5 bit shipment share from 2–3% in 2025 to 8–9% in 2026, thereby exerting significant downward pressure on DDR5 prices.\n\n$MU\n\n---\n\nSource: https://t.co/OCPbVURCrN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016288921779956334, "ret_signed_m1d": -0.016288921779956334, "alpha_spy_m1d": -0.01914899304972939, "alpha_c_m1d": -0.01701449081159001, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0007255690316336771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0012927788932852735, "ret_signed_p1d": 0.0012927788932852735, "alpha_spy_p1d": -0.0026814472038692028, "alpha_c_p1d": -0.003556738884441746, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.004849517777727019, "ret_p1w": 0.04990090936126412, "ret_signed_p1w": 0.04990090936126412, "alpha_spy_p1w": 0.05641942857938265, "alpha_c_p1w": 0.047189757025082546, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.002711152336181577, "ret_p1m": 0.062017399456682165, "ret_signed_p1m": 0.062017399456682165, "alpha_spy_p1m": 0.018266419066135686, "alpha_c_p1m": -0.03588999960570072, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.09790739906238288, "ret_p3m": 0.13404945225299714, "ret_signed_p3m": 0.13404945225299714, "alpha_spy_p3m": 0.05192218749950528, "alpha_c_p3m": 0.001775190198316956, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.13227426205468018, "ret_p6m": 0.9564371609536746, "ret_signed_p6m": 0.9564371609536746, "alpha_spy_p6m": 0.811837870694835, "alpha_c_p6m": 0.5756520571535841, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.38078510380009045, "price_path": [-0.1325, -0.1125, -0.1245, -0.1215, -0.1135, -0.1847, -0.1656, -0.1894, -0.207, -0.2154, -0.2388, -0.2511, -0.2355, -0.2364, -0.3593, -0.4422, -0.4108, -0.4351, -0.3289, -0.3963, -0.4006, -0.3879, -0.3877, -0.4025, -0.407, -0.407, -0.4248, -0.3949, -0.3715, -0.3328, -0.3124, -0.3229, -0.3374, -0.3368, -0.3297, -0.3043, -0.3069, -0.3061, -0.2879, -0.2661, -0.26, -0.2046, -0.1646, -0.1785, -0.1774, -0.1554, -0.1498, -0.1545, -0.174, -0.1827, -0.1953, -0.1953, -0.1694, -0.1711, -0.1657, -0.1859, -0.1538, -0.1188, -0.1101, -0.0839, -0.0644, -0.0438, -0.0163, 0.0, 0.0013, -0.0037, 0.0328, 0.0371, 0.0499, 0.0499, 0.0652, 0.0521, 0.1024, 0.0967, 0.0859, 0.0752, 0.0622, 0.0419, 0.0492, 0.054, 0.054, 0.0345, 0.0733, 0.0545, 0.062, 0.0743, 0.0232, 0.0361, 0.0044, -0.023, -0.0132, -0.0232, -0.0578, -0.0525, -0.0362, -0.0402, -0.0403, -0.0342, -0.0102, -0.0585, -0.0952, -0.0703, -0.0592, -0.0616, -0.0349, 0.0256, 0.0673, 0.102, 0.072, 0.0808, 0.0427, 0.0658, 0.0529, 0.0111, -0.0011, 0.0152, 0.0043, 0.005, 0.0158, 0.0524, 0.0266, 0.0266, 0.0221, 0.0241, 0.0715, 0.1333, 0.134, 0.1667, 0.2077, 0.2989, 0.3564, 0.361, 0.3701, 0.3802, 0.4569, 0.4038, 0.4201, 0.4355, 0.395, 0.3529, 0.3567, 0.4139, 0.4434, 0.5713, 0.5851, 0.6213, 0.6484, 0.6029, 0.6965, 0.6602, 0.5676, 0.664, 0.6147, 0.6568, 0.7482, 0.7469, 0.7848, 0.7462, 0.7132, 0.7843, 0.8906, 0.8999, 0.9155, 0.9563, 0.9336, 0.9316, 1.0259, 0.882, 1.0501, 1.0573, 1.0537, 1.1865, 1.0813, 1.114, 1.0453, 1.1306, 1.0885, 0.9724, 0.9501, 0.7382, 0.79, 0.933, 0.9381, 0.9876, 0.9876, 1.0413, 1.0756, 1.0673, 1.0213, 0.9564]}
{"unit_id": "orig:1938183627545973006", "ticker_idx": 0, "ts": "2025-06-26T10:32:41+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I judge that the related concerns are unfounded", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish MU: post-earnings dip driven by unfounded concerns over blended ASP mix shift and 2026 HBM lock-in timing, not fundamental weakness.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Why did Micron's stock price turn downward despite strong earnings? - Korea Daishin Securities\n\nIf there's market concern, I believe it's because Micron's stock price turned downward after its earnings announcement. I judge that the stock price's downward turn stemmed from three main factors.\n\n1. Decline in DRAM Blended ASP\n\nWhile DRAM sales significantly exceeded expectations, the Blended ASP (Average Selling Price) slightly decreased from the previous quarter (ASP: estimated to decline 1-2% QoQ). The market seems to be questioning this ASP decline, despite HBM revenue, which has a higher ASP than general-purpose products, increasing by 50%.However, I don't believe the decline in Blended ASP was due to weakness in the ASP of leading-edge products like HBM and high-value DDR5. It is estimated that leading-edge products continue to see strong ASPs amidst a tight supply-demand environment. I judge that the Blended ASP decline this quarter was largely due to changes in product mix (accelerated sales growth of mobile and industrial products, which have lower ASPs compared to servers).\n\n2. When will the 2026 HBM volume lock-in occur?\n\nI believe that in this earnings announcement, Micron Technology provided sufficient outlook on the direction of HBM demand growth, related technology and product competitiveness, and market share expansion. However, there was no mention of the 2026 Nvidia-bound volume lock-in, which the market is focused on, and I judge that this partly served as a catalyst for profit-taking.I believe it would have been difficult to materialize the 2026 volume lock-in in this earnings announcement. I judge that the volume lock-in will become concrete after priority negotiations with leading companies are completed. As HBM4, a new product for 2026, is a leading-edge product that can contribute to AI technology innovation, related demand will be sufficiently strong, and I believe that market expectations can be met during this process. If so, I judge that the related concerns are unfounded.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00992064949241711, "ret_signed_m1d": 0.00992064949241711, "alpha_spy_m1d": 0.01768369302316386, "alpha_c_m1d": 0.017135688419591344, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009841259532637148, "ret_signed_p1d": -0.009841259532637148, "alpha_spy_p1d": -0.014809591296642388, "alpha_c_p1d": -0.014242418921685207, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.02944444933381929, "ret_signed_p1w": -0.02944444933381929, "alpha_spy_p1w": -0.05145900384583868, "alpha_c_p1w": -0.05260462158565393, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": -0.11615291307436704, "ret_signed_p1m": -0.11615291307436704, "alpha_spy_p1m": -0.15738720973371012, "alpha_c_p1m": -0.15327403601108536, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.32195742379580516, "ret_signed_p3m": 0.32195742379580516, "alpha_spy_p3m": 0.23504551943629215, "alpha_c_p3m": 0.161423497883884, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 1.1137855991590007, "ret_signed_p6m": 1.1137855991590007, "alpha_spy_p6m": 0.9950943986632539, "alpha_c_p6m": 0.8286846054001149, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3104, -0.296, -0.2968, -0.41, -0.4863, -0.4574, -0.4798, -0.382, -0.444, -0.448, -0.4363, -0.4362, -0.4498, -0.454, -0.454, -0.4703, -0.4428, -0.4212, -0.3856, -0.3668, -0.3765, -0.3898, -0.3893, -0.3828, -0.3594, -0.3617, -0.361, -0.3443, -0.3242, -0.3186, -0.2675, -0.2307, -0.2435, -0.2425, -0.2222, -0.2171, -0.2214, -0.2394, -0.2474, -0.259, -0.259, -0.2351, -0.2367, -0.2317, -0.2503, -0.2208, -0.1885, -0.1806, -0.1564, -0.1384, -0.1194, -0.0941, -0.0791, -0.0779, -0.0825, -0.0489, -0.0449, -0.0332, -0.0332, -0.019, -0.0311, 0.0152, 0.0099, 0.0, -0.0098, -0.0218, -0.0406, -0.0338, -0.0294, -0.0294, -0.0474, -0.0116, -0.0289, -0.022, -0.0107, -0.0578, -0.0458, -0.0751, -0.1003, -0.0913, -0.1005, -0.1324, -0.1275, -0.1124, -0.1162, -0.1162, -0.1106, -0.0885, -0.133, -0.1668, -0.1439, -0.1336, -0.1359, -0.1113, -0.0555, -0.0172, 0.0148, -0.0128, -0.0047, -0.0398, -0.0185, -0.0304, -0.0689, -0.0802, -0.0652, -0.0752, -0.0745, -0.0646, -0.0308, -0.0546, -0.0546, -0.0588, -0.0569, -0.0133, 0.0436, 0.0443, 0.0743, 0.1122, 0.1961, 0.249, 0.2533, 0.2617, 0.271, 0.3417, 0.2927, 0.3077, 0.322, 0.2846, 0.2459, 0.2493, 0.302, 0.3292, 0.447, 0.4597, 0.4931, 0.5179, 0.476, 0.5623, 0.5288, 0.4435, 0.5323, 0.4869, 0.5257, 0.6099, 0.6087, 0.6436, 0.608, 0.5776, 0.6431, 0.741, 0.7496, 0.764, 0.8015, 0.7806, 0.7787, 0.8656, 0.7331, 0.8879, 0.8945, 0.8912, 1.0135, 0.9166, 0.9467, 0.8835, 0.962, 0.9232, 0.8163, 0.7958, 0.6007, 0.6484, 0.78, 0.7848, 0.8303, 0.8303, 0.8798, 0.9114, 0.9037, 0.8613, 0.8016, 0.8857, 0.9628, 1.0065, 1.0962, 1.0545, 0.9168, 0.8879, 0.8482, 0.7926, 0.9757, 1.1138]}
{"unit_id": "orig:1931575664026832996", "ticker_idx": 0, "ts": "2025-06-08T04:54:59+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we were impressed by NVIDIA's \"full-stack moat\" in AI, built across computing, networking, and storage", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Citi impressed by NVIDIA's full-stack AI moat; networking business up 64% sequentially to $5B with broad-based growth across scale-out and scale-up.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi’s Take on NVIDIA\n\nYesterday, during Citi's annual Silicon Valley bus tour, we held an investor meeting with Kevin Deierling, SVP of NVIDIA's Networking Business. Overall, we were impressed by NVIDIA's \"full-stack moat\" in AI, built across computing, networking, and storage. Here are the key takeaways:\n\nEverything Revolves Around AI\n\nThe company focuses on the networking functions required for AI. NVIDIA has a unified network operating system, focused on optimizing compute and storage, with the core goal of achieving the lowest user token cost per second (TCO). Dynamo is the operating system for AI factories. Since Mellanox was acquired by NVIDIA, the team stopped developing campus switches and shifted focus to AI switches, because the response logic for telecom/large-scale web differs from that of AI agent flows.\n\nSpectrum-X Sales\n\nIn the recently announced Q1 earnings, the networking business achieved robust sequential growth of 64%, reaching $5 billion. This growth appears to be broad-based, covering products from \"scale out\" to \"scale up.\" On the \"scale up\" front, the company achieved over $1 billion in sales with its NVLink solution; on the \"scale out\" front, the company continues to make progress with its Spectrum-X (Ethernet) product portfolio, adding two new customers. Spectrum-X's quarterly revenue run rate is $2 billion, encompassing both NICs (Network Interface Cards) and switches.\n\nInfiniband vs. Ethernet\n\nEthernet is still evolving, but Infiniband remains the de facto standard. Hyperscale customers request Ethernet due to familiarity, and NVIDIA has strong capabilities in Spectrum-X. Although general vendors like Arista with its EOS system are trying to transition from traditional data center switching business to AI applications, NVIDIA's advantage lies in its complete technology stack. Broadcom's main product in the market is Tomahawk, not the Jericho platform, which offers programmability but adds latency.\n\nNVLink\n\nNVLink is an in-rack \"scale-up\" platform custom-built for AI. It connects multiple GPUs (up to 72) through a cache-coherent interface. Larger LLM models benefit from massive GPU scale-up, and sometimes developers build models based on the hardware GPU stack. Customers don't need to buy the entire technology stack from NVIDIA; they can purchase as needed, such as SuperNICs, switches, and NVLink Fusion for semi-custom AI infrastructure.\n\nCo-Packaged Optics (CPO)\n\nNVLink uses over three miles of copper cables. Copper wires have low power consumption and are inexpensive. However, the reason for using CPO when scaling out is that 10 to 15 inches of copper wire on a PCB can generate significant noise, requiring additional DSPs and retimers to handle it. To avoid this issue, optical components can be directly packaged next to the ASIC, converting signals into optical signals, thus eliminating the need for DSPs and retimers.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006380282151325245, "ret_signed_m1d": -0.006380282151325245, "alpha_spy_m1d": -0.005479810842562771, "alpha_c_m1d": 0.010081343882485272, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009324789052088667, "ret_signed_p1d": 0.009324789052088667, "alpha_spy_p1d": 0.0036551243831979274, "alpha_c_p1d": -0.010561686271912674, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.014513336154033274, "ret_signed_p1w": 0.014513336154033274, "alpha_spy_p1w": 0.009510660579628905, "alpha_c_p1w": -0.01047121819025243, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.12186151647081211, "ret_signed_p1m": 0.12186151647081211, "alpha_spy_p1m": 0.08435158987358471, "alpha_c_p1m": 0.018654633796084452, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.20361719462765793, "ret_signed_p3m": 0.20361719462765793, "alpha_spy_p3m": 0.11797328300582333, "alpha_c_p3m": 0.07558170333429004, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.27240289611841284, "ret_signed_p6m": 0.27240289611841284, "alpha_spy_p6m": 0.12939339273343542, "alpha_c_p6m": -0.1263752232547639, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1885, -0.1897, -0.147, -0.162, -0.1907, -0.176, -0.169, -0.1748, -0.1488, -0.1538, -0.2024, -0.2187, -0.2311, -0.2401, -0.2277, -0.2258, -0.2863, -0.3388, -0.3154, -0.3248, -0.1984, -0.2458, -0.2223, -0.2238, -0.2133, -0.2674, -0.2884, -0.2884, -0.3205, -0.3067, -0.2799, -0.2538, -0.2217, -0.2377, -0.2356, -0.2363, -0.2175, -0.1972, -0.202, -0.204, -0.1793, -0.1771, -0.1821, -0.1376, -0.089, -0.0511, -0.0547, -0.0507, -0.0495, -0.0578, -0.0759, -0.0687, -0.0795, -0.0795, -0.05, -0.0548, -0.0241, -0.0526, -0.0368, -0.0099, -0.005, -0.0185, -0.0064, 0.0, 0.0093, 0.0015, 0.0167, -0.0046, 0.0145, 0.0105, 0.0201, 0.0201, 0.0086, 0.0109, 0.037, 0.082, 0.0869, 0.1061, 0.1078, 0.0749, 0.1026, 0.1172, 0.1172, 0.1095, 0.1219, 0.1421, 0.1506, 0.1564, 0.1504, 0.1969, 0.2016, 0.213, 0.2089, 0.2017, 0.1712, 0.1974, 0.2182, 0.2165, 0.2393, 0.2306, 0.257, 0.2472, 0.2181, 0.2621, 0.2499, 0.258, 0.2675, 0.281, 0.2765, 0.2843, 0.2732, 0.2763, 0.2652, 0.2762, 0.2315, 0.2298, 0.2269, 0.248, 0.2608, 0.2745, 0.2733, 0.2633, 0.2213, 0.2213, 0.1974, 0.1963, 0.2036, 0.1711, 0.1801, 0.1973, 0.2434, 0.2423, 0.2469, 0.2464, 0.2263, 0.1941, 0.2358, 0.2388, 0.2875, 0.2512, 0.2409, 0.246, 0.2495, 0.2751, 0.3083, 0.3129, 0.3245, 0.3156, 0.301, 0.2975, 0.326, 0.3503, 0.2843, 0.3205, 0.2624, 0.261, 0.2749, 0.2847, 0.2807, 0.2703, 0.2641, 0.2773, 0.3061, 0.3427, 0.4096, 0.4518, 0.4227, 0.4199, 0.4506, 0.3932, 0.3688, 0.3188, 0.3193, 0.3957, 0.3544, 0.3589, 0.3103, 0.3335, 0.3084, 0.2717, 0.3079, 0.2667, 0.2543, 0.28, 0.2469, 0.264, 0.264, 0.2411, 0.2616, 0.2724]}
{"unit_id": "orig:1937052726694994130", "ticker_idx": 0, "ts": "2025-06-23T07:38:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our \"buy\" and \"overweight\" opinion on the semiconductor sector", "tweet_type": "analysis", "timeline": "months", "summary": "Korea Investment & Securities reiterates Buy on Samsung/SK Hynix and Overweight on semis; VEU revocation unlikely to be implemented, short-term selloff overdone.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Don't always respond to US fear marketing - Korea Investment & Securities\n\nSamsung Electronics, SK Hynix Reportedly Reviewing VEU Revocation for Chinese Fabs\n\nAccording to a recent Wall Street Journal report, the U.S. government is reportedly considering revoking Validated End-User (VEU) authorizations previously granted to Samsung Electronics, SK Hynix, and TSMC for their semiconductor fabs in China. VEU is a system that allows the supply of U.S. equipment and technology without separate export licenses. If VEU is revoked, Samsung Electronics and SK Hynix would need to obtain individual approvals from the U.S. Department of Commerce for each piece of U.S. equipment brought into their Chinese factories.\n\nLow Probability of Actual Implementation Considering Supply Chain Impact\n\nSimilar to semiconductor reciprocal tariffs, from a supply chain perspective, we believe the actual implementation of this measure is unlikely. Accordingly, we assess that the impact on domestic semiconductor-related companies will be limited. This is because if the introduction of equipment and technology to Samsung Electronics and SK Hynix's fabs is blocked, it would disrupt the global memory supply chain and cause memory prices to rise, ultimately harming major U.S. customers through increased component procurement costs. Samsung Electronics' Xi'an fab accounts for approximately 40% of its NAND capacity, while SK Hynix's Wuxi fab accounts for 40% of its DRAM capacity, and its Dalian fab accounts for 30% of its NAND capacity.\n\nVEU Revocation Likely a Strategic Pressure Tactic by the U.S.\n\nThe mention of possible VEU revocation can be interpreted as a strategic pressure tactic by the U.S., rather than an immediate restrictive measure. There is also significant opposition within the U.S. government. According to the Wall Street Journal report, opponents within the administration are concerned that \"revoking VEU could actually help Chinese companies grow and strengthen the Chinese government's control over those factories.\"\n\nShort-Term Stock Price Adjustment is Excessive, Mid-to-Long Term Memory Investment Momentum Remains Valid\n\nFollowing this report over the weekend in Korean time, stock prices fell, mainly for U.S. technology and semiconductor equipment stocks. Today, in the Korean market, stock prices are declining, primarily for semiconductor front-end equipment stocks. However, we believe this stock price adjustment is excessive. While intensified U.S. sanctions against China and China's efforts to increase semiconductor self-sufficiency are mid-to-long term risk factors, the likelihood of immediate damage to the operation and investment of Samsung Electronics and SK Hynix's Chinese factories in the short term is limited. We believe that an excessive stock price decline due to short-term policy noise is likely to see a rebound, and we maintain our \"buy\" and \"overweight\" opinion on the semiconductor sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025862097916495408, "ret_signed_m1d": 0.025862097916495408, "alpha_spy_m1d": 0.03564296806676481, "alpha_c_m1d": 0.03713523974361421, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.011273141827118804, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043103473681822546, "ret_signed_p1d": 0.043103473681822546, "alpha_spy_p1d": 0.03205626468873479, "alpha_c_p1d": 0.025152234033618548, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.017951239648203998, "ret_p1w": 0.03735863299370279, "ret_signed_p1w": 0.03735863299370279, "alpha_spy_p1w": 0.00786597364623609, "alpha_c_p1w": -0.005254934377409981, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.04261356737111277, "ret_p1m": 0.14491088372511252, "ret_signed_p1m": 0.14491088372511252, "alpha_spy_p1m": 0.09707287499602058, "alpha_c_p1m": 0.07932291634925637, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.06558796737585615, "ret_p3m": 0.3929748331415097, "ret_signed_p3m": 0.3929748331415097, "alpha_spy_p3m": 0.2894840681229023, "alpha_c_p3m": 0.25611719453844706, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.13685763860306266, "ret_p6m": 0.7912417514852141, "ret_signed_p6m": 0.7912417514852141, "alpha_spy_p6m": 0.6569384506658937, "alpha_c_p6m": 0.6158200833730949, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.17542166811211923, "price_path": [0.0524, 0.0592, 0.0379, -0.0034, 0.0138, 0.0138, -0.0069, -0.0328, -0.0828, -0.0776, -0.0862, -0.0276, -0.0483, -0.031, -0.0241, -0.0569, -0.05, -0.0466, -0.0448, -0.0517, -0.0397, -0.0397, -0.0397, -0.0379, -0.0379, -0.0431, -0.0431, -0.0638, -0.0638, -0.0638, -0.0586, -0.0586, -0.0552, -0.0069, -0.019, -0.0103, -0.0121, -0.0207, -0.0379, -0.0362, -0.0397, -0.0569, -0.0655, -0.0569, -0.0707, -0.0362, -0.0328, -0.031, -0.0207, -0.0207, -0.0034, 0.019, 0.019, 0.031, 0.0207, 0.0328, 0.0259, 0.0052, -0.0138, 0.0017, 0.031, 0.0207, 0.0259, 0.0, 0.0431, 0.0569, 0.0379, 0.0547, 0.0374, 0.0443, 0.0547, 0.1067, 0.0981, 0.0703, 0.0651, 0.0478, 0.0582, 0.0859, 0.0842, 0.105, 0.1224, 0.1571, 0.164, 0.1761, 0.1449, 0.1518, 0.1449, 0.1432, 0.2212, 0.2247, 0.2594, 0.2386, 0.1952, 0.2091, 0.2126, 0.1935, 0.223, 0.2455, 0.2316, 0.2334, 0.2473, 0.2421, 0.2421, 0.2143, 0.2143, 0.223, 0.2247, 0.2386, 0.2403, 0.2195, 0.2247, 0.2074, 0.2091, 0.1727, 0.1987, 0.2108, 0.216, 0.2056, 0.216, 0.2403, 0.2594, 0.2733, 0.308, 0.3271, 0.3774, 0.3565, 0.393, 0.393, 0.4485, 0.4693, 0.4814, 0.4936, 0.445, 0.4671, 0.4619, 0.4985, 0.5639, 0.5639, 0.5639, 0.5639, 0.5639, 0.5639, 0.6449, 0.6257, 0.5961, 0.6553, 0.7024, 0.7059, 0.7093, 0.6989, 0.7181, 0.6815, 0.7215, 0.7773, 0.7337, 0.7512, 0.8139, 0.8731, 0.9359, 0.8278, 0.7529, 0.7285, 0.7059, 0.7529, 0.8034, 0.7965, 0.7912, 0.6937, 0.7529, 0.7041, 0.6815, 0.7529, 0.6518, 0.685, 0.7303, 0.7912, 0.8034, 0.7512, 0.7564, 0.8017, 0.8209, 0.8313, 0.8888, 0.908, 0.8888, 0.8818, 0.8697, 0.8975, 0.8261, 0.7912]}
{"unit_id": "orig:1937052726694994130", "ticker_idx": 1, "ts": "2025-06-23T07:38:53+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our \"buy\" and \"overweight\" opinion on the semiconductor sector", "tweet_type": "analysis", "timeline": "months", "summary": "Korea Investment & Securities reiterates Buy on Samsung/SK Hynix and Overweight on semis; VEU revocation unlikely to be implemented, short-term selloff overdone.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Don't always respond to US fear marketing - Korea Investment & Securities\n\nSamsung Electronics, SK Hynix Reportedly Reviewing VEU Revocation for Chinese Fabs\n\nAccording to a recent Wall Street Journal report, the U.S. government is reportedly considering revoking Validated End-User (VEU) authorizations previously granted to Samsung Electronics, SK Hynix, and TSMC for their semiconductor fabs in China. VEU is a system that allows the supply of U.S. equipment and technology without separate export licenses. If VEU is revoked, Samsung Electronics and SK Hynix would need to obtain individual approvals from the U.S. Department of Commerce for each piece of U.S. equipment brought into their Chinese factories.\n\nLow Probability of Actual Implementation Considering Supply Chain Impact\n\nSimilar to semiconductor reciprocal tariffs, from a supply chain perspective, we believe the actual implementation of this measure is unlikely. Accordingly, we assess that the impact on domestic semiconductor-related companies will be limited. This is because if the introduction of equipment and technology to Samsung Electronics and SK Hynix's fabs is blocked, it would disrupt the global memory supply chain and cause memory prices to rise, ultimately harming major U.S. customers through increased component procurement costs. Samsung Electronics' Xi'an fab accounts for approximately 40% of its NAND capacity, while SK Hynix's Wuxi fab accounts for 40% of its DRAM capacity, and its Dalian fab accounts for 30% of its NAND capacity.\n\nVEU Revocation Likely a Strategic Pressure Tactic by the U.S.\n\nThe mention of possible VEU revocation can be interpreted as a strategic pressure tactic by the U.S., rather than an immediate restrictive measure. There is also significant opposition within the U.S. government. According to the Wall Street Journal report, opponents within the administration are concerned that \"revoking VEU could actually help Chinese companies grow and strengthen the Chinese government's control over those factories.\"\n\nShort-Term Stock Price Adjustment is Excessive, Mid-to-Long Term Memory Investment Momentum Remains Valid\n\nFollowing this report over the weekend in Korean time, stock prices fell, mainly for U.S. technology and semiconductor equipment stocks. Today, in the Korean market, stock prices are declining, primarily for semiconductor front-end equipment stocks. However, we believe this stock price adjustment is excessive. While intensified U.S. sanctions against China and China's efforts to increase semiconductor self-sufficiency are mid-to-long term risk factors, the likelihood of immediate damage to the operation and investment of Samsung Electronics and SK Hynix's Chinese factories in the short term is limited. We believe that an excessive stock price decline due to short-term policy noise is likely to see a rebound, and we maintain our \"buy\" and \"overweight\" opinion on the semiconductor sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009633960880562231, "ret_signed_m1d": -0.009633960880562231, "alpha_spy_m1d": 0.00014690926970717388, "alpha_c_m1d": -0.003334267520181955, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07321773991092995, "ret_signed_p1d": 0.07321773991092995, "alpha_spy_p1d": 0.06217053091784219, "alpha_c_p1d": 0.03648900821335066, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.12524094727529445, "ret_signed_p1w": 0.12524094727529445, "alpha_spy_p1w": 0.09574828792782775, "alpha_c_p1w": 0.060488397827961116, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": 0.034682017315795344, "ret_signed_p1m": 0.034682017315795344, "alpha_spy_p1m": -0.013155991413296597, "alpha_c_p1m": -0.05484913121998769, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.37192101427859003, "ret_signed_p3m": 0.37192101427859003, "alpha_spy_p3m": 0.2684302492599826, "alpha_c_p3m": 0.16174232271671962, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 1.046804114037171, "ret_signed_p6m": 1.046804114037171, "alpha_spy_p6m": 0.9125008132178507, "alpha_c_p6m": 0.7031114278058463, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [-0.1768, -0.2038, -0.2334, -0.2665, -0.2422, -0.2388, -0.2514, -0.2991, -0.3661, -0.348, -0.3653, -0.2953, -0.3045, -0.3068, -0.3053, -0.3307, -0.3268, -0.3268, -0.3207, -0.3315, -0.3038, -0.3141, -0.2907, -0.2999, -0.3045, -0.3172, -0.3172, -0.2845, -0.2845, -0.2845, -0.2661, -0.268, -0.2688, -0.2499, -0.2364, -0.2076, -0.2288, -0.2134, -0.233, -0.223, -0.2288, -0.2426, -0.2307, -0.2191, -0.2211, -0.1999, -0.183, -0.2119, -0.2004, -0.2004, -0.1618, -0.1349, -0.1349, -0.1175, -0.1118, -0.0751, -0.0925, -0.0925, -0.0443, -0.0405, -0.0501, -0.052, -0.0096, 0.0, 0.0732, 0.1021, 0.1291, 0.0944, 0.1252, 0.1002, 0.0751, 0.0732, 0.0424, 0.0443, 0.0867, 0.0829, 0.1445, 0.1349, 0.1561, 0.1503, 0.1407, 0.0385, 0.0366, 0.0501, 0.0347, 0.0366, 0.0385, 0.025, 0.0096, 0.0116, 0.0154, 0.0539, -0.0058, -0.0058, 0.0154, -0.0039, 0.0096, -0.0116, 0.0289, 0.0366, 0.0713, 0.0655, 0.0655, 0.0308, 0.0135, -0.0154, -0.0559, -0.0328, 0.0, 0.0077, 0.0019, 0.0362, 0.0381, -0.0121, 0.0053, 0.013, 0.0246, 0.0555, 0.069, 0.1114, 0.1732, 0.1848, 0.2677, 0.2774, 0.343, 0.287, 0.3719, 0.3719, 0.3546, 0.3931, 0.3796, 0.3758, 0.2986, 0.3468, 0.341, 0.3893, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.6517, 0.6015, 0.588, 0.6305, 0.7463, 0.7964, 0.8736, 0.8485, 0.8582, 0.8466, 0.9682, 1.0646, 1.0106, 1.1534, 1.192, 1.1573, 1.3927, 1.2615, 1.2344, 1.2885, 1.2383, 1.3386, 1.3888, 1.3811, 1.3618, 1.1611, 1.3386, 1.1997, 1.1688, 1.2036, 1.0106, 1.0067, 1.0029, 1.0222, 1.1009, 1.0468, 1.0777, 1.1549, 1.1318, 1.0931, 1.1009, 1.2283, 1.1858, 1.2669, 1.182, 1.2051, 1.1395, 1.0468]}
{"unit_id": "orig:1937052726694994130", "ticker_idx": 2, "ts": "2025-06-23T07:38:53+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "an excessive stock price decline due to short-term policy noise is likely to see a rebound, and we maintain our \"buy\" and \"overweight\" opinion on the semiconductor sector", "tweet_type": "analysis", "timeline": "months", "summary": "Korea Investment & Securities reiterates Buy on Samsung/SK Hynix and Overweight on semis; VEU revocation unlikely to be implemented, short-term selloff overdone.", "resolved_tickers": ["EWY"], "resolution_method": "etf", "resolution_reason": "Korea semiconductor sector proxied by EWY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Don't always respond to US fear marketing - Korea Investment & Securities\n\nSamsung Electronics, SK Hynix Reportedly Reviewing VEU Revocation for Chinese Fabs\n\nAccording to a recent Wall Street Journal report, the U.S. government is reportedly considering revoking Validated End-User (VEU) authorizations previously granted to Samsung Electronics, SK Hynix, and TSMC for their semiconductor fabs in China. VEU is a system that allows the supply of U.S. equipment and technology without separate export licenses. If VEU is revoked, Samsung Electronics and SK Hynix would need to obtain individual approvals from the U.S. Department of Commerce for each piece of U.S. equipment brought into their Chinese factories.\n\nLow Probability of Actual Implementation Considering Supply Chain Impact\n\nSimilar to semiconductor reciprocal tariffs, from a supply chain perspective, we believe the actual implementation of this measure is unlikely. Accordingly, we assess that the impact on domestic semiconductor-related companies will be limited. This is because if the introduction of equipment and technology to Samsung Electronics and SK Hynix's fabs is blocked, it would disrupt the global memory supply chain and cause memory prices to rise, ultimately harming major U.S. customers through increased component procurement costs. Samsung Electronics' Xi'an fab accounts for approximately 40% of its NAND capacity, while SK Hynix's Wuxi fab accounts for 40% of its DRAM capacity, and its Dalian fab accounts for 30% of its NAND capacity.\n\nVEU Revocation Likely a Strategic Pressure Tactic by the U.S.\n\nThe mention of possible VEU revocation can be interpreted as a strategic pressure tactic by the U.S., rather than an immediate restrictive measure. There is also significant opposition within the U.S. government. According to the Wall Street Journal report, opponents within the administration are concerned that \"revoking VEU could actually help Chinese companies grow and strengthen the Chinese government's control over those factories.\"\n\nShort-Term Stock Price Adjustment is Excessive, Mid-to-Long Term Memory Investment Momentum Remains Valid\n\nFollowing this report over the weekend in Korean time, stock prices fell, mainly for U.S. technology and semiconductor equipment stocks. Today, in the Korean market, stock prices are declining, primarily for semiconductor front-end equipment stocks. However, we believe this stock price adjustment is excessive. While intensified U.S. sanctions against China and China's efforts to increase semiconductor self-sufficiency are mid-to-long term risk factors, the likelihood of immediate damage to the operation and investment of Samsung Electronics and SK Hynix's Chinese factories in the short term is limited. We believe that an excessive stock price decline due to short-term policy noise is likely to see a rebound, and we maintain our \"buy\" and \"overweight\" opinion on the semiconductor sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012820564683778457, "ret_signed_m1d": -0.012820564683778457, "alpha_spy_m1d": -0.003039694533509052, "alpha_c_m1d": -0.003039694533509052, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.009780870150269405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04652842858972894, "ret_signed_p1d": 0.04652842858972894, "alpha_spy_p1d": 0.03548121959664119, "alpha_c_p1d": 0.03548121959664119, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.011047208993087754, "ret_p1w": 0.03399598163418038, "ret_signed_p1w": 0.03399598163418038, "alpha_spy_p1w": 0.004503322286713685, "alpha_c_p1w": 0.004503322286713685, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.0294926593474667, "ret_p1m": 0.0499856165876702, "ret_signed_p1m": 0.0499856165876702, "alpha_spy_p1m": 0.00214760785857826, "alpha_c_p1m": 0.00214760785857826, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.04783800872909194, "ret_p3m": 0.16320947866872526, "ret_signed_p3m": 0.16320947866872526, "alpha_spy_p3m": 0.05971871365011783, "alpha_c_p3m": 0.05971871365011783, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.10349076501860743, "ret_p6m": 0.31850549603669576, "ret_signed_p6m": 0.31850549603669576, "alpha_spy_p6m": 0.18420219521737535, "alpha_c_p6m": 0.18420219521737535, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.1343033008193204, "price_path": [-0.1816, -0.1863, -0.2107, -0.2215, -0.2099, -0.2131, -0.2345, -0.2617, -0.2685, -0.2953, -0.2325, -0.2535, -0.2177, -0.2171, -0.2167, -0.2213, -0.2177, -0.2177, -0.2165, -0.2105, -0.1991, -0.194, -0.1961, -0.1984, -0.1919, -0.1906, -0.1991, -0.1687, -0.159, -0.1573, -0.1675, -0.176, -0.169, -0.1639, -0.1628, -0.1557, -0.153, -0.1531, -0.1527, -0.1587, -0.1487, -0.1603, -0.1507, -0.1507, -0.136, -0.1249, -0.1098, -0.1266, -0.1106, -0.1085, -0.0784, -0.0585, -0.0555, -0.0424, -0.047, -0.035, -0.0249, -0.0432, -0.0177, -0.0402, -0.0223, -0.0223, -0.0128, 0.0, 0.0465, 0.0425, 0.0363, 0.019, 0.034, 0.032, 0.0364, 0.0472, 0.0472, 0.0094, 0.0383, 0.0383, 0.0536, 0.0452, 0.0545, 0.0526, 0.0527, 0.0524, 0.0428, 0.0641, 0.05, 0.0638, 0.056, 0.0546, 0.0474, 0.0586, 0.0614, 0.0488, 0.0177, 0.0419, 0.05, 0.0542, 0.0579, 0.0588, 0.052, 0.0674, 0.0763, 0.0611, 0.0572, 0.0487, 0.0291, 0.0279, 0.0259, 0.0589, 0.0477, 0.0428, 0.0439, 0.0608, 0.0406, 0.0406, 0.036, 0.0503, 0.0526, 0.0606, 0.0687, 0.0792, 0.1041, 0.1216, 0.1331, 0.1407, 0.1623, 0.15, 0.1632, 0.1531, 0.1749, 0.1733, 0.153, 0.1508, 0.1361, 0.1559, 0.1538, 0.1796, 0.2033, 0.2079, 0.2213, 0.2007, 0.2128, 0.2018, 0.1628, 0.2046, 0.1912, 0.2269, 0.2659, 0.2751, 0.3027, 0.2769, 0.2885, 0.2975, 0.326, 0.3585, 0.3606, 0.3959, 0.3722, 0.3973, 0.4517, 0.3794, 0.3918, 0.347, 0.3224, 0.3781, 0.3727, 0.3817, 0.3551, 0.3559, 0.34, 0.3194, 0.3045, 0.2753, 0.2852, 0.3008, 0.2934, 0.3188, 0.3188, 0.309, 0.3054, 0.3263, 0.3397, 0.3234, 0.3584, 0.3695, 0.3767, 0.3937, 0.3679, 0.3349, 0.3358, 0.3185]}
{"unit_id": "quote:1934402598712562003", "ticker_idx": 0, "ts": "2025-06-16T00:08:13+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "US Firms Reiterate Positive Outlook... revenue contribution is expected to continue until at least 2028. Brokers estimate that TPU will contribute $1 billion to MediaTek's revenue next year, doubling to $2-3 billion the year after. The positive rating remains unchanged.", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/endorses US banks' reiterated positive on MediaTek; TPU revenue $1B→$2-3B, 2nm orders potentially through 2028.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "2454 MediaTek: US Firms Reiterate Positive Outlook\nMajor US investment banks believe the market is overly concerned about a US client's TPU (Tensor Processing Unit) progress. Brokerage channel checks indicate that the TPU compute die tape-out is expected in mid-August, not late September as market rumors suggest. Furthermore, the I/O die has already taped out. The client will use this for training, specifically the v8p version, not the v7e as rumored in the market.\n\nIt's also believed that MediaTek will secure 2nm TPU orders (potentially named v9). While it's uncertain if this will be an exclusive supply, the revenue contribution is expected to continue until at least 2028. Brokers estimate that TPU will contribute $1 billion to MediaTek's revenue next year, doubling to $2-3 billion the year after. The positive rating remains unchanged.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/16 外電綜合整理\n\n- 2454聯發科: 美系重申正向\n美系大行認為市場過度擔憂美系客戶TPU 進度，券商通路訪查認為TPU compute die 的tape out時間落在8月中，而非市場傳言的9月底，而且I/O die 已經taped out 了。客戶會用在訓練，而且是v8p, 而非市場傳的v7e, 接下來認為聯發科也會取得2nm", "parent_author": null, "ret_m1d": -0.0039841257772917915, "ret_signed_m1d": -0.0039841257772917915, "alpha_spy_m1d": 0.005440403874595989, "alpha_c_m1d": 0.020277511559819827, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01593630090007303, "ret_signed_p1d": 0.01593630090007303, "alpha_spy_p1d": 0.02448133343665193, "alpha_c_p1d": 0.022732590620790916, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": 0.001253995304636768, "alpha_c_p1w": 0.00554333531117257, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.13343495014332585, "ret_signed_p1m": 0.13343495014332585, "alpha_spy_p1m": 0.09809403950426865, "alpha_c_p1m": 0.029326875270582642, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.20249724186795914, "ret_signed_p3m": 0.20249724186795914, "alpha_spy_p3m": 0.10809522154446682, "alpha_c_p3m": 0.04853727248281925, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.15374735914923354, "ret_signed_p6m": 0.15374735914923354, "alpha_spy_p6m": 0.013907829984418862, "alpha_c_p6m": -0.24727011421822098, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [0.1673, 0.1833, 0.1793, 0.1793, 0.2151, 0.2072, 0.1833, 0.1673, 0.1076, 0.1633, 0.1434, 0.1434, 0.1434, 0.0319, 0.004, -0.0558, 0.0359, 0.1036, 0.1155, 0.1036, 0.0876, 0.0677, 0.0876, 0.0717, 0.0359, 0.0916, 0.0598, 0.0996, 0.0757, 0.0916, 0.0757, 0.0757, 0.0359, 0.0319, 0.0199, 0.0159, 0.0279, 0.0558, 0.0478, 0.0797, 0.0837, 0.0916, 0.0876, 0.0438, 0.0478, 0.0677, 0.0558, 0.0478, 0.0199, 0.0199, 0.0199, 0.004, 0.004, 0.004, 0.0159, 0.0159, -0.008, 0.0199, 0.0239, 0.0637, 0.0478, 0.0239, -0.004, 0.0, 0.0159, 0.0239, 0.008, -0.004, 0.0, 0.012, 0.0319, 0.0279, 0.0239, -0.004, 0.0159, 0.0319, 0.0319, 0.0481, 0.04, 0.0319, 0.1009, 0.1375, 0.1537, 0.1253, 0.1334, 0.1456, 0.1294, 0.1456, 0.1578, 0.1619, 0.1619, 0.1619, 0.1619, 0.1375, 0.105, 0.1212, 0.1131, 0.1131, 0.0806, 0.0887, 0.0725, 0.1009, 0.0969, 0.1009, 0.105, 0.1131, 0.1294, 0.1131, 0.1456, 0.1294, 0.1009, 0.1091, 0.1091, 0.1456, 0.1375, 0.1375, 0.1253, 0.1131, 0.105, 0.105, 0.1334, 0.1253, 0.1659, 0.1619, 0.2269, 0.2106, 0.2025, 0.2066, 0.2066, 0.2472, 0.2269, 0.2269, 0.17, 0.1416, 0.1172, 0.0806, 0.0928, 0.0644, 0.0644, 0.0684, 0.0441, 0.04, 0.0725, 0.0725, 0.0766, 0.0847, 0.0928, 0.0928, 0.0684, 0.0603, 0.0684, 0.0806, 0.0806, 0.0887, 0.0887, 0.0806, 0.0522, 0.0522, 0.0847, 0.0684, 0.0562, 0.0644, 0.0644, 0.0562, 0.0684, 0.0806, 0.0481, 0.0237, 0.0156, 0.0075, 0.0116, 0.0116, -0.0006, -0.0006, -0.0494, -0.0575, -0.0372, -0.0697, -0.0656, -0.0372, 0.0562, 0.0887, 0.1334, 0.1741, 0.1497, 0.1375, 0.1416, 0.1578, 0.17, 0.1537]}
{"unit_id": "orig:1943319586654556403", "ticker_idx": 0, "ts": "2025-07-10T14:41:09+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": "data center", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "June's better-than-expected export performance initially suggests that AI server demand may remain strong", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Relays UBS note flagging Taiwan ADP export beat as positive signal for NVDA Q2 data center upside; TSMC sales cited as next verify.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwan Export Data: June ADP Exports Show Significant Increase, Potentially Indicating Upside for NVDA Q2 Data Center Revenue\nUBS (T. Arcuri, July 08, 2025)\n\nUBS notes that despite fluctuations in the correlation between Taiwan's ADP data and NVIDIA's revenue, June's better-than-expected export performance initially suggests that AI server demand may remain strong.\n\nJune ADP Exports Significantly Exceed Seasonal Average\nTaiwan's Ministry of Finance announced that June exports of automatic data processing (ADP) equipment (excluding notebooks) reached $14.18 billion, a 7.5% month-over-month increase. This is significantly higher than the seasonal average of -4.5%, making it the second strongest June performance in over a decade, second only to the historical high.\n\nFQ2 Overall Export Projections Far Exceed Current NVDA Revenue Expectations\nIf July's proportion of FQ2 exports is historically consistent, UBS estimates that total ADP exports for this quarter will reach $44 billion, a 24% quarter-over-quarter increase. This would mark a new high after FQ1's 63% growth, far exceeding UBS's forecast of only a 1.5% quarter-over-quarter growth for NVDA's data center revenue.\n\nCurrent Data, Though Affected by Tariffs, Still Holds Positive Reference Value\nWhile some of the export growth may be driven by accelerated shipments due to US tariff policies, and the correlation between ADP and NVDA's data center revenue has been unstable over the past 8 to 10 quarters, UBS believes this data can still be regarded as a positive signal in the short term.\n\nADP Data Coverage Differs from NVDA's Actual Revenue Structure\nADP export statistics only cover complete servers and do not include components like AI chip exports assembled in China or Malaysia, nor do they include networking equipment. Additionally, they comprise traditional products from non-NVDA manufacturers, leading to structural differences compared to NVDA's revenue.\n\nFuture Focus on TSMC Sales to Verify Supply/Demand Strength\nUBS advises investors to closely monitor TSMC's June sales data to further verify whether this surge in ADP exports is due to a structural increase in AI demand or merely a short-term accelerated shipment behavior. This will help provide a clearer judgment of NVDA's fundamentals.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007434524872936277, "ret_signed_m1d": -0.007434524872936277, "alpha_spy_m1d": -0.00462214643424208, "alpha_c_m1d": -0.00013082809379005855, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0049968979800461, "ret_signed_p1d": 0.0049968979800461, "alpha_spy_p1d": 0.008512321850641658, "alpha_c_p1d": 0.005101239547490644, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": 0.05423517561357549, "ret_signed_p1w": 0.05423517561357549, "alpha_spy_p1w": 0.05068788454657036, "alpha_c_p1w": 0.039662278655132166, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.11334541901263417, "ret_signed_p1m": 0.11334541901263417, "alpha_spy_p1m": 0.09519331314666024, "alpha_c_p1m": 0.0924425024528821, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.12766865088683543, "ret_signed_p3m": 0.12766865088683543, "alpha_spy_p3m": 0.05551522017167754, "alpha_c_p3m": -0.044597679899357034, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.15095161006972124, "ret_signed_p6m": 0.15095161006972124, "alpha_spy_p6m": 0.05305076327766112, "alpha_c_p6m": -0.15143280768836176, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.3254, -0.3163, -0.3633, -0.3816, -0.3816, -0.4095, -0.3974, -0.3741, -0.3515, -0.3236, -0.3375, -0.3357, -0.3363, -0.3199, -0.3023, -0.3064, -0.3082, -0.2867, -0.2848, -0.2892, -0.2505, -0.2083, -0.1753, -0.1784, -0.175, -0.1739, -0.1812, -0.1969, -0.1906, -0.2, -0.2, -0.1743, -0.1785, -0.1519, -0.1766, -0.1629, -0.1395, -0.1352, -0.147, -0.1364, -0.1309, -0.1228, -0.1296, -0.1164, -0.1349, -0.1183, -0.1218, -0.1135, -0.1135, -0.1234, -0.1215, -0.0987, -0.0597, -0.0553, -0.0387, -0.0372, -0.0658, -0.0417, -0.029, -0.029, -0.0357, -0.025, -0.0074, 0.0, 0.005, -0.0002, 0.0402, 0.0443, 0.0542, 0.0506, 0.0444, 0.0179, 0.0407, 0.0587, 0.0573, 0.0771, 0.0695, 0.0924, 0.0839, 0.0586, 0.0969, 0.0863, 0.0934, 0.1016, 0.1133, 0.1094, 0.1161, 0.1066, 0.1092, 0.0996, 0.1091, 0.0703, 0.0689, 0.0663, 0.0846, 0.0957, 0.1077, 0.1066, 0.0979, 0.0614, 0.0614, 0.0407, 0.0397, 0.0461, 0.0178, 0.0257, 0.0406, 0.0806, 0.0797, 0.0837, 0.0832, 0.0658, 0.0378, 0.074, 0.0767, 0.119, 0.0874, 0.0785, 0.0829, 0.0859, 0.1082, 0.1371, 0.1411, 0.1511, 0.1434, 0.1307, 0.1277, 0.1525, 0.1736, 0.1162, 0.1477, 0.0971, 0.0959, 0.108, 0.1166, 0.113, 0.104, 0.0987, 0.1101, 0.1351, 0.167, 0.2251, 0.2617, 0.2364, 0.234, 0.2608, 0.2109, 0.1896, 0.1462, 0.1466, 0.213, 0.1772, 0.1811, 0.1388, 0.1589, 0.1372, 0.1052, 0.1367, 0.1009, 0.0901, 0.1125, 0.0837, 0.0985, 0.0985, 0.0787, 0.0965, 0.1059, 0.0945, 0.1176, 0.1117, 0.1308, 0.1273, 0.1201, 0.1027, 0.0667, 0.0744, 0.0831, 0.0418, 0.0613, 0.103, 0.1195, 0.1531, 0.1495, 0.1495, 0.1612, 0.1471, 0.143, 0.1366, 0.1366, 0.151]}
{"unit_id": "orig:1943319586654556403", "ticker_idx": 1, "ts": "2025-07-10T14:41:09+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS advises investors to closely monitor TSMC's June sales data to further verify whether this surge in ADP exports is due to a structural increase in AI demand", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Relays UBS note flagging Taiwan ADP export beat as positive signal for NVDA Q2 data center upside; TSMC sales cited as next verify.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwan Export Data: June ADP Exports Show Significant Increase, Potentially Indicating Upside for NVDA Q2 Data Center Revenue\nUBS (T. Arcuri, July 08, 2025)\n\nUBS notes that despite fluctuations in the correlation between Taiwan's ADP data and NVIDIA's revenue, June's better-than-expected export performance initially suggests that AI server demand may remain strong.\n\nJune ADP Exports Significantly Exceed Seasonal Average\nTaiwan's Ministry of Finance announced that June exports of automatic data processing (ADP) equipment (excluding notebooks) reached $14.18 billion, a 7.5% month-over-month increase. This is significantly higher than the seasonal average of -4.5%, making it the second strongest June performance in over a decade, second only to the historical high.\n\nFQ2 Overall Export Projections Far Exceed Current NVDA Revenue Expectations\nIf July's proportion of FQ2 exports is historically consistent, UBS estimates that total ADP exports for this quarter will reach $44 billion, a 24% quarter-over-quarter increase. This would mark a new high after FQ1's 63% growth, far exceeding UBS's forecast of only a 1.5% quarter-over-quarter growth for NVDA's data center revenue.\n\nCurrent Data, Though Affected by Tariffs, Still Holds Positive Reference Value\nWhile some of the export growth may be driven by accelerated shipments due to US tariff policies, and the correlation between ADP and NVDA's data center revenue has been unstable over the past 8 to 10 quarters, UBS believes this data can still be regarded as a positive signal in the short term.\n\nADP Data Coverage Differs from NVDA's Actual Revenue Structure\nADP export statistics only cover complete servers and do not include components like AI chip exports assembled in China or Malaysia, nor do they include networking equipment. Additionally, they comprise traditional products from non-NVDA manufacturers, leading to structural differences compared to NVDA's revenue.\n\nFuture Focus on TSMC Sales to Verify Supply/Demand Strength\nUBS advises investors to closely monitor TSMC's June sales data to further verify whether this surge in ADP exports is due to a structural increase in AI demand or merely a short-term accelerated shipment behavior. This will help provide a clearer judgment of NVDA's fundamentals.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00909101104819332, "ret_signed_m1d": -0.00909101104819332, "alpha_spy_m1d": -0.006278632609499124, "alpha_c_m1d": -0.0017873142690471022, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0035154238705955576, "alpha_c_p1d": 0.00010434156744454359, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": 0.027272584532529454, "ret_signed_p1w": 0.027272584532529454, "alpha_spy_p1w": 0.023725293465524322, "alpha_c_p1w": 0.012699687574086127, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.06818174171385505, "ret_signed_p1m": 0.06818174171385505, "alpha_spy_p1m": 0.050029635847881115, "alpha_c_p1m": 0.04727882515410298, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.3097635926647444, "ret_signed_p3m": 0.3097635926647444, "alpha_spy_p3m": 0.2376101619495865, "alpha_c_p3m": 0.13749726187855194, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.45149494358142706, "ret_signed_p6m": 0.45149494358142706, "alpha_spy_p6m": 0.35359409678936693, "alpha_c_p6m": 0.14911052582334405, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.217, -0.2061, -0.226, -0.2333, -0.2305, -0.2441, -0.2613, -0.2097, -0.2179, -0.1961, -0.188, -0.1835, -0.178, -0.178, -0.14, -0.1509, -0.1672, -0.1599, -0.169, -0.1409, -0.1337, -0.1228, -0.0957, -0.1011, -0.0966, -0.1092, -0.1147, -0.1038, -0.111, -0.1092, -0.1165, -0.1264, -0.1246, -0.1246, -0.1246, -0.1436, -0.14, -0.1038, -0.0966, -0.0993, -0.0902, -0.054, -0.0359, -0.05, -0.0636, -0.0682, -0.05, -0.0409, -0.0591, -0.0409, -0.0727, -0.0455, -0.0273, -0.0227, -0.0182, -0.0364, -0.0136, -0.0136, -0.0091, -0.0136, -0.0182, -0.0182, -0.0091, 0.0, 0.0, -0.0045, 0.0091, 0.0273, 0.0273, 0.05, 0.0455, 0.0273, 0.0409, 0.0409, 0.0409, 0.0409, 0.0318, 0.05, 0.0545, 0.0545, 0.0318, 0.0455, 0.0227, 0.0727, 0.0682, 0.0727, 0.0727, 0.0909, 0.0682, 0.0727, 0.0727, 0.0773, 0.0318, 0.0455, 0.0318, 0.0636, 0.0682, 0.0818, 0.0545, 0.0545, 0.0591, 0.0545, 0.0545, 0.0545, 0.0727, 0.0727, 0.0909, 0.1136, 0.1273, 0.1455, 0.1409, 0.1683, 0.1546, 0.1729, 0.1546, 0.182, 0.2231, 0.2231, 0.2048, 0.1865, 0.1865, 0.1911, 0.2094, 0.2459, 0.2778, 0.2778, 0.3098, 0.2915, 0.3143, 0.3143, 0.2915, 0.3006, 0.3371, 0.3554, 0.3235, 0.3508, 0.3508, 0.3326, 0.3235, 0.3235, 0.3508, 0.3463, 0.3737, 0.3737, 0.3691, 0.3782, 0.3737, 0.3326, 0.3371, 0.3326, 0.3463, 0.3371, 0.3463, 0.3326, 0.3052, 0.3189, 0.2824, 0.2733, 0.328, 0.2641, 0.255, 0.2915, 0.3143, 0.3098, 0.3143, 0.2869, 0.3052, 0.3235, 0.3189, 0.3326, 0.3645, 0.3508, 0.3737, 0.3462, 0.3553, 0.3279, 0.3141, 0.3096, 0.3096, 0.3096, 0.3416, 0.3645, 0.3691, 0.3691, 0.3828, 0.4011, 0.392, 0.4194, 0.4194, 0.4515]}
{"unit_id": "orig:1938096047936700725", "ticker_idx": 0, "ts": "2025-06-26T04:44:40+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect Nvidia rack revenue for 2026 to more than double YoY, driven by both units (off a low base) and ASP (from mainly NVL36 to all NVL72)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Macquarie bullish NVDA call: 2025 GB rack shipments 10k racks, 2026 rack revenue more than doubling YoY on units and ASP uplift.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Macquarie Securities: With no major transition gap between GB200 and GB300 in late-3Q to early-4Q, we expect 2025 GB rack shipments of 10k (or 6.5k NVL72 equivalent racks), contributing 25-30% of total revenue. We also expect Nvidia rack revenue for 2026 to more than double YoY, driven by both units (off a low base) and ASP (from mainly NVL36 to all NVL72).\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004579972911195829, "ret_signed_m1d": -0.004579972911195829, "alpha_spy_m1d": 0.0031830706195509206, "alpha_c_m1d": 0.0026350660159784045, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017610676115523027, "ret_signed_p1d": 0.017610676115523027, "alpha_spy_p1d": 0.012642344351517787, "alpha_c_p1d": 0.013209516726474968, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": 0.027867425132180434, "ret_signed_p1w": 0.027867425132180434, "alpha_spy_p1w": 0.005852870620161044, "alpha_c_p1w": 0.004707252880345791, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": 0.11921052616034622, "ret_signed_p1m": 0.11921052616034622, "alpha_spy_p1m": 0.07797622950100314, "alpha_c_p1m": 0.0820894032236279, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.15107767485204815, "ret_signed_p3m": 0.15107767485204815, "alpha_spy_p3m": 0.06416577049253513, "alpha_c_p3m": -0.009456251059873022, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.16765770446898332, "ret_signed_p6m": 0.16765770446898332, "alpha_spy_p6m": 0.04896650397323654, "alpha_c_p6m": -0.11744328928990244, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3009, -0.2895, -0.2878, -0.3434, -0.3917, -0.3702, -0.3788, -0.2625, -0.3061, -0.2845, -0.2859, -0.2763, -0.326, -0.3454, -0.3454, -0.3749, -0.3621, -0.3375, -0.3135, -0.2839, -0.2987, -0.2968, -0.2974, -0.2801, -0.2614, -0.2658, -0.2676, -0.2449, -0.2429, -0.2476, -0.2066, -0.1619, -0.127, -0.1303, -0.1266, -0.1255, -0.1332, -0.1498, -0.1432, -0.1531, -0.1531, -0.126, -0.1304, -0.1022, -0.1284, -0.1139, -0.0891, -0.0846, -0.097, -0.0859, -0.08, -0.0714, -0.0786, -0.0646, -0.0842, -0.0666, -0.0703, -0.0615, -0.0615, -0.0721, -0.07, -0.0459, -0.0046, 0.0, 0.0176, 0.0192, -0.0111, 0.0144, 0.0279, 0.0279, 0.0208, 0.0321, 0.0507, 0.0586, 0.0639, 0.0584, 0.1011, 0.1055, 0.116, 0.1122, 0.1055, 0.0775, 0.1017, 0.1208, 0.1192, 0.1402, 0.1322, 0.1564, 0.1474, 0.1206, 0.1611, 0.1499, 0.1574, 0.1661, 0.1786, 0.1744, 0.1815, 0.1714, 0.1742, 0.164, 0.1741, 0.133, 0.1315, 0.1288, 0.1482, 0.1599, 0.1726, 0.1715, 0.1622, 0.1236, 0.1236, 0.1017, 0.1006, 0.1073, 0.0774, 0.0857, 0.1015, 0.1439, 0.1429, 0.1471, 0.1467, 0.1282, 0.0986, 0.1369, 0.1397, 0.1845, 0.1511, 0.1417, 0.1463, 0.1495, 0.1731, 0.2037, 0.2079, 0.2186, 0.2104, 0.1969, 0.1937, 0.22, 0.2423, 0.1816, 0.2149, 0.1614, 0.1601, 0.1729, 0.182, 0.1782, 0.1687, 0.163, 0.1751, 0.2016, 0.2353, 0.2969, 0.3356, 0.3089, 0.3063, 0.3346, 0.2818, 0.2593, 0.2133, 0.2138, 0.2841, 0.2461, 0.2502, 0.2055, 0.2268, 0.2038, 0.17, 0.2033, 0.1653, 0.154, 0.1777, 0.1471, 0.1629, 0.1629, 0.1419, 0.1607, 0.1706, 0.1586, 0.1831, 0.1768, 0.1971, 0.1933, 0.1857, 0.1673, 0.1291, 0.1373, 0.1466, 0.1028, 0.1235, 0.1677]}
{"unit_id": "thread:1933436468296626528", "ticker_idx": 0, "ts": "2025-06-13T08:09:10+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Apple's very strengths might, ironically, be hindering its fundamental transformation towards AI innovation, creating a crisis situation that warrants a serious internal review... the iPhone could ultimately become a device like a BlackBerry—beautiful, but with a nonsensical Siri that responds as slowly as a flickering fluorescent light", "tweet_type": "analysis", "timeline": "years", "summary": "Bearish AAPL: ecosystem lock-in blocks AI progress, privacy policy limits data use, M-chips poorly suited for AI; risks iPhone becoming irrelevant like BlackBerry over next 5–10 years.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Apple WWDC25 Commentary by Professor Seokjun Kwon of Sungkyunkwan University, South Korea\n\nAt WWDC on June 9, contrary to many people's expectations, Apple showed that there was nothing new that could truly be called an advancement in Apple's unique AI. As one of the leading IT giants of our time, Apple naturally invested in AI at a corporate level, and in 2024, they even announced an AI service called 'Apple Intelligence.' Therefore, many investors had anticipated the disclosure of an AI at a comparable level in the rapidly changing global AI innovation environment. However, reports from domestic and international media suggest that when the lid was finally opened, the situation was such that one might wonder if Apple had, in fact, given up on AI.\n\nIt's not that Apple lacks capital, personnel, or an ecosystem. If there's a reason for their significant lag in AI innovation compared to other companies, it's paradoxically due to the characteristics of their robust ecosystem. In particular, the device ecosystem, spearheaded by the iPhone (+iPad) and MacBook, serves as a technological moat that allows Apple to outpace competitors. Yet, it simultaneously acts as a moat that works against their progression towards AI and AGI.\n\nApple will likely continue to upgrade the iPhone series, and to do so, they will invest enormous capital to produce Apple Silicon, specifically the M-series chips, using TSMC's most advanced processes, such as 2-nanometer, 18A, and 14A, which are the most expensive foundry and packaging processes of their time. However, at least the M-series chips are not suitable architectures for current AI model computations. Furthermore, Apple's blind spot in advancing AI innovation stems from the limitations in utilizing user information due to its device-centric ecosystem.\n\nMy understanding of Apple's basic policy is to excessively prioritize user privacy for the competitiveness of its device ecosystem. This policy will fundamentally provide users with a sense of security. Indeed, most of the features of Apple Intelligence, which Apple announced last year, run only on the device, and cloud-based AI is limited to a private cloud compute approach. However, precisely this policy makes real-time, large-scale LLM integration, which relies on cloud resources, almost impossible. Apple explicitly states externally that it does not use user personal data for foundation model training. While this might be an excellent approach for managing Apple's loyal customers, it severely reduces the inherent usability of the data embedded in Apple's touted ecosystem.\n\nThe M-series chips and AI are also not a good match. While they are undeniably optimized hardware for individual devices, they are fundamentally designed for mobile and standalone devices, revealing limitations in memory capacity and bandwidth for large-scale data parallel-high-speed processing. This is because they were not originally optimized architectures designed for large-scale matrix data. For instance, the M4 Pro includes a 20-core GPU, showing a computational power of about 9 TFLOPS in FP32 operations. In terms of computational power alone, it's excellent. However, even the outdated Nvidia A100 GPU showed 312 TFLOPS in FP16 operations and was already allocating nearly 1 TB/s of bandwidth with its HBM2, so in terms of overcoming the memory barrier, there's no comparison. Also, as far as I know, M-series chips do not natively support open-source frameworks like PyTorch, which leads to the cumbersome process of repeated compilation and conversion to run models on M-series chips. Apple might develop solutions to optimize so-called on-device AI on M-series devices. However, unless the aforementioned memory capacity and bandwidth issues are fundamentally resolved at the hardware level, there will be limitations on the size of data that can be processed on-device. This, in turn, acts as a fundamental limit to on-device inference performance.\n\nThe realistic approach Apple could take to overcome this would naturally be to adopt third-party models like Meta Llama, Google Gemini, or Anthropic's Claude and adapt them to their device environment. Apple has not shied away from attempting this. Siri is based on ChatGPT, and iOS utilizes Google Gemini resources. However, this is technically difficult due to user data privacy barriers and also challenging from a policy standpoint. It's not that Apple lacks technology that could circumvent this. For example, they could establish a separate channel for user-selected anonymized telemetry collection to control data collection. Also, for cloud integration, only embedding-level generation tasks could be performed on the device, while large-scale inference logic would be processed in an encrypted form in the cloud. Furthermore, well-known methodologies like distillation, quantization, and pruning could be attempted for computational optimization tailored to the M-series architecture.\n\nFrom a hardware perspective, a more fundamental approach would be to develop AI-specific server chips. Apple's hardware engineers, with their accumulated experience in optimizing the M-series, would likely be capable of such work. The problem, however, lies in finding HBM suitable for the M-series. On the surface, it might seem like Apple could do as Nvidia, TSMC, and SK Hynix have done in forming an alliance to find sufficiently good GPU+HBM+logic die packaging combinations. But technically, this is by no means an easy task. For instance, Apple's M-chips use TSMC's InFO (integrated Fan-Out) packaging technology, where LPDDR memory is matched, and integration proceeds in SiP form. HBM, however, does not use the SiP method; it employs TSV-based silicon interposers from the outset, with bandwidth expansion and latency minimization in mind, leading to entirely different packaging. This requires TSMC's 2.5D packaging solutions like CoWoS and SoW-X. Since Apple did not design the M-series chips with these fundamental differences in packaging stages in mind, it's difficult to directly leverage SK Hynix's or TSMC's packaging expertise. If they were to utilize it, there's a method of separating the SoC and HBM into chiplets to reduce the interposer area and simultaneously improve yield and cost. However, in this case, Apple would need to negotiate with TSMC to completely convert one fab line for Apple's exclusive use. This means not just bearing the fab costs, but also investing in the construction of a new fab itself. Whether TSMC would accept such an investment is a separate matter.\n\nThere are further anticipated difficulties in Apple's process of utilizing M-chips as AI accelerators. Due to the large chip die area, even with maximum utilization of the interposer process, it's difficult to avoid issues of reduced manufacturing yield, which inevitably leads to increased costs. Furthermore, the CoWoS packaging, optimized for Nvidia dies, would need various modifications. Additionally, M-chips are matched with LPDDR4X/5, not HBM, so the memory controller would need to be redesigned to match with HBM. For example, HBM3E is based on a 1024-bit interface and supports up to 1.18 TB/s, and HBM4 is expected to be enhanced to 1.638 TB/s, while the M1 Max is 512-bit based with a bandwidth of only 400 GB/s. If they were to reuse the current matching method, they would either have to forcibly adjust HBM interface timings to match LPDDR4/5, or optimize by separating memory chiplets and logic chiplets. The problem with this approach is that it makes it difficult to fully leverage the original advantages of HBM.\n\nWhile there are technically feasible solutions, it's hard to guarantee that any single direction currently offers a definitive answer. As stated repeatedly, Apple's unique ecosystem, especially the technical ecosystem barrier where user information is tied to devices, is now acting as an internal barrier for Apple as it expands into AI and opens up its ecosystem.\n\nIt might be a coincidence, but as if to offer an excuse for this, a few days before WWDC, Apple released a paper online titled 'The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity.'\n**https://t.co/LNQ1jpTlu0\nThe main content of this paper analyzes the limitations of large reasoning models (LRMs) and reasoning (thinking) token usage patterns, utilizing several well-known puzzles to test the performance of publicly available LRMs. Examples include Towers of Hanoi, Checker Jump, River Crossing, and Block World. Apple's researchers also showed test results indicating that the correlation between AI reasoning models, particularly the intelligence index derived as a function of the number of parameters, does not necessarily scale proportionally, contrary to popular belief. Apple's researchers claimed to have observed that AI reasoning models, despite their immense size, tend to copy and memorize similar patterns from existing training data rather than truly understanding the problems, and then produce results. This might imply that AI reasoning models, contrary to what is known, do not necessarily function by effectively utilizing a massive parameter space to solve more complex problems more accurately. A particularly interesting phenomenon observed was that these reasoning models experience a kind of 'critical phenomenon.' That is, beyond a certain apparent threshold, they exhibit paradoxical behavior, seemingly reducing their reasoning effort. This suggests that it's not true reasoning but rather a way of avoiding problems by combining plausible patterns from already learned data within the limits of computational and energy resources.\n\nIf we take the claims and observations in this paper at face value, Apple's AI researchers are, at the very least, sending a warning message that the development of massive reasoning models, which Meta, Google, OpenAI, and DeepMind are competitively developing and pouring huge capital into daily, will not directly lead to AGI. What's interesting is that this message came out a few days before WWDC (submitted June 7, 2025). While one should certainly be wary of over-interpreting coincidences, it makes one suspect that behind Apple's WWDC presentation, which barely touched on artificial intelligence, i.e., Apple Intelligence, and primarily focused on what seemed to be only incremental upgrades (I didn't even expect them to include naming conventions aligned with the year), there might be a case of Apple's \"sour grapes\" phenomenon regarding AI.\nIn other words, from Apple's perspective, they might have wanted to exclaim:\n'We know we seem to be lagging in AI innovation on the surface. Investors, loyal Apple customers, don't worry! Of course, we have it all figured out! And it's not that we don't know how to use that massive personal information; we're just trying our best not to touch our loyal customers' data because it's precious! And AI reasoning might not be as grand as you think! AGI is a distant dream, and it turns out that scaling law was an illusion after all? We will approach AI more cautiously, more efficiently, and by spending money properly, and we will eventually achieve a safe, beautiful Apple Intelligence with a strong, unique Apple identity! So just wait a little longer!'\n(Of course, this is just my speculation.)\nIn fact, this Apple paper has various points of criticism scattered throughout, which supports the idea that this hunch might not just be mere speculation. For example, the mathematical puzzles tested by Apple's researchers are somewhat disconnected from real-world, multimodal reasoning environment performance tests. This is because such puzzles are often used to test the efficiency of computations or the scalability of algorithms. The reason these types of puzzles are frequently used for testing algorithm efficiency and scaling characteristics is that they have relatively well-defined rules and state spaces, and algorithms can also be regularized. If Apple had intended to point out the limitations of existing LRMs, it would have been better if they had added various benchmarks to these puzzle tests and also demonstrated results by specifying a few more hybrid tasks leading from puzzles to coding and mathematical problem-solving. There's another aspect of Apple's researchers' paper that can be criticized. One of the core arguments of this paper is that 'even if the token resources required for reasoning are increased, if the complexity exceeds a certain level (i.e., passes a threshold), the model's reasoning ability stagnates or even decreases.' If misinterpreted, this could lead to the conclusion that 'the scaling law of reasoning models is an illusion!' However, this directly contradicts observations from actual language model scaling laws. Hundreds of scaling-related studies to date have consistently shown that performance improves in a power-law manner as the number of parameters increases, and beyond a certain size, performance is observed to move towards saturation. At the very least, performance might reach saturation, but it doesn't decrease. Indeed, as Apple's researchers showed in their paper, there is no evidence of scaling experiments for reasoning computational power in this paper. This might be because Apple does not have a GPU-based AIDC large enough to test a parameter space big enough to confirm scaling trends. (That is, Apple might not have hundreds of thousands of Nvidia GPUs like Microsoft, Google, or Meta.) If they had intended to conduct scaling tests, it would have been better to compare various models like ChatGPT's o1, o3, etc., in the same puzzle environment. Verifying the scaling law is similar to verifying the scaling law of large language models, and for this, Apple's researchers should have tested combinations of training data, parameters, and computational load and shown the performance curve. Also, if they intended to test saturation for the size of the parameter space, it would have been better to fix the number of parameters and test performance degradation by varying the degree of fine-tuning. The Tower of Hanoi problem tested by Apple's researchers is a case in point. This is a classic problem with a precisely known algorithm, and Apple's researchers explicitly provided it to the computer, yet they reported no performance improvement. However, in that process, there was no analysis of internal token allocation, attention patterns, or memory resource usage patterns. In other words, it is unclear how the explicit algorithm affected the reallocation of internal computational resources.\n\nGiven Apple's accumulated innovation know-how, its robust ecosystem, and the still powerful optimization technology of its M-chips, it is premature to easily conclude that Apple will fall behind in AI innovation based on a few signals. However, it's worth considering whether Apple's very strengths might, ironically, be hindering its fundamental transformation towards AI innovation, creating a crisis situation that warrants a serious internal review. Apple will continue to invest heavily in beautiful designs and devices with an inviting aura, but that will likely come after the current AI innovation frenzy has matured sufficiently. One thing is certain: Apple cannot afford to sit back and merely observe this maturation from the sidelines while other competitors invest heavily in AI. The smartphone platform, including the iPhone, has been around for almost 20 years, and given the technology replacement cycle, the next 10 years, or even the next 5 years, are uncertain. When that time comes, a company that continues to focus solely on beautiful and aura-filled devices will likely fall behind in the competition.\n\nApple's strong cash reserves, hardware engineers, and technological ecosystem can still be utilized as unique assets. However, for them to truly become helpful assets, Apple might need to transform into an entirely different identity than it is now. If they miss that timing, the iPhone could ultimately become a device like a BlackBerry—beautiful, but with a nonsensical Siri that responds as slowly as a flickering fluorescent light.\n\n$AAPL\n\n---\n\nhttps://t.co/t8wWUy9nZ5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013998564529722923, "ret_signed_m1d": -0.013998564529722923, "alpha_spy_m1d": -0.002692093054060818, "alpha_c_m1d": 0.00021724909679998738, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.01421581362652291, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010028094916962882, "ret_signed_p1d": -0.010028094916962882, "alpha_spy_p1d": -0.000513898438247562, "alpha_c_p1d": 0.006152790782200324, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.016180885699163206, "ret_p1w": 0.023161228702667902, "ret_signed_p1w": -0.023161228702667902, "alpha_spy_p1w": -0.024774503698477224, "alpha_c_p1w": -0.017433158463628695, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.005728070239039207, "ret_p1m": 0.06194965061489399, "ret_signed_p1m": -0.06194965061489399, "alpha_spy_p1m": -0.01227285884918361, "alpha_c_p1m": 0.008684956790000031, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.07063460740489402, "ret_p3m": 0.15575156632203324, "ret_signed_p3m": -0.15575156632203324, "alpha_spy_p3m": -0.060042819902162536, "alpha_c_p3m": -0.02434832738586179, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.13140323893617145, "ret_p6m": 0.4175367242705521, "ret_signed_p6m": -0.4175367242705521, "alpha_spy_p6m": -0.2658584725845552, "alpha_c_p6m": -0.17938219296873004, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.23815453130182207, "price_path": [0.0812, 0.0942, 0.0884, 0.1096, 0.1221, 0.1375, 0.1262, 0.138, 0.1077, 0.1292, 0.1346, 0.1382, 0.033, -0.0423, -0.0775, -0.1235, 0.0109, -0.032, 0.0073, 0.0295, 0.0276, -0.0124, 0.0014, 0.0014, -0.018, 0.0154, 0.0401, 0.0593, 0.0639, 0.0683, 0.0737, 0.0803, 0.0845, 0.0439, 0.0111, 0.0092, -0.0023, 0.004, 0.0093, 0.073, 0.0839, 0.0808, 0.0764, 0.0754, 0.0628, 0.053, 0.0287, 0.025, -0.006, -0.006, 0.0191, 0.0202, 0.0178, 0.0224, 0.0267, 0.0347, 0.0324, 0.0213, 0.038, 0.0255, 0.0317, 0.0119, 0.014, 0.0, 0.01, -0.0041, 0.0007, 0.0007, 0.0232, 0.0257, 0.0196, 0.026, 0.0232, 0.0236, 0.0444, 0.0579, 0.0814, 0.087, 0.087, 0.0687, 0.069, 0.0748, 0.0812, 0.0749, 0.0619, 0.0644, 0.0698, 0.0691, 0.075, 0.0816, 0.0914, 0.0901, 0.0881, 0.0887, 0.0896, 0.0754, 0.0641, 0.0566, 0.0302, 0.0351, 0.0329, 0.0855, 0.12, 0.1675, 0.1577, 0.1703, 0.1891, 0.1863, 0.1802, 0.1766, 0.175, 0.1518, 0.1461, 0.1607, 0.1576, 0.1686, 0.1746, 0.1852, 0.183, 0.183, 0.1707, 0.2153, 0.222, 0.2215, 0.2123, 0.1943, 0.1558, 0.1723, 0.1929, 0.2063, 0.2136, 0.2179, 0.2123, 0.2511, 0.305, 0.2966, 0.2858, 0.309, 0.3019, 0.2966, 0.2976, 0.3018, 0.3104, 0.3149, 0.3081, 0.3071, 0.3151, 0.2946, 0.2499, 0.2621, 0.2627, 0.2707, 0.261, 0.2857, 0.3364, 0.3391, 0.3171, 0.3229, 0.3394, 0.3699, 0.3709, 0.3744, 0.3831, 0.3778, 0.3711, 0.3762, 0.3767, 0.3748, 0.3682, 0.3744, 0.4041, 0.395, 0.3923, 0.3896, 0.3643, 0.3642, 0.3699, 0.3582, 0.3849, 0.4075, 0.4128, 0.4158, 0.4158, 0.4224, 0.4441, 0.4599, 0.4495, 0.4319, 0.4221, 0.4175]}
{"unit_id": "orig:1938000316521431123", "ticker_idx": 0, "ts": "2025-06-25T22:24:16+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Overweight investment opinion. Raise target price from KRW 285,000 to KRW 360,000.", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM reiterates Overweight on SK Hynix, raises PT to KRW 360K on HBM leadership, 4-year earnings upcycle, and AI-driven ASP expansion.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix: Target Price Raised to KRW 360,000 (JPM)\n\n1. Investment Opinion and Target Price\n\n- Maintain Overweight investment opinion.\n- Raise target price from KRW 285,000 to KRW 360,000.\n\n2. Continued Leadership in HBM Market\n\n- Proactive response compared to competitors (pre-emptive supply of HBM4 samples to NVIDIA in March).\n- Maintain technological lead through exclusive material contracts.\n- Expected expansion of HBM4 market share due to competitors' certification delays (expected 37% production share by 2026).\n\n3. Structural ASP Increase Driven by Expanding HBM Revenue Share\n\n- HBM revenue share expected to exceed 50% after 2026.\n- HBM4 ASP has a 35% premium structure compared to HBM3E.\n- Transition to a highly profitable structure centered on HBM, anticipating an extraordinary earnings upcycle for the next four years.\n\n4. Strengthening Link with AI Ecosystem\n\n- Enhance HBM4/4E performance competitiveness through logic die collaboration with TSMC.\n- Expected to benefit in conjunction with ASIC market growth.\n\n5. Economics of HBM Business\n\n- HBM wafer cost at ~$22,000, approximately 3 times that of DRAM.\n- Investment of 10,000 wafers/month can generate $2.6 billion in annual revenue.\n- OPM (Operating Profit Margin) is expected to remain above 60%.\n\n6. Earnings Upcycle from 2024 to 2027\n\n- Operating profit projected to increase for four consecutive years from 2025-2027 (recovering after 2023 deficit).\n- 2025E EPS: KRW 43,005 → 2026E: KRW 47,632 → 2027E: KRW 63,591.\n- ROE: 2025E 35.2%, 2026E 28.7%, 2027E 28.9%.\n\n7. DRAM ASP Increase and Share Expansion\n\n- DRAM ASP expected to continue increasing from 2025-2026 (2025E $4.4 → 2026E $5.0 → 2027E $5.9).\n- DRAM revenue accounts for 75-84% of total revenue, contributing 96-97% to operating profit.\n\n8. Limited Recovery Speed in NAND Segment\n\n- Profitability limited by intensified competition and delayed eSSD price recovery.\n- NAND operating margin expected to remain in the single digits, with structural improvement taking time.\n\n9. Q2 2025 Earnings Preview\n\n- Revenue: KRW 19.9 trillion, Operating Profit: KRW 9.0 trillion (OPM 45%).\n- DRAM ASP expected to increase 2% QoQ, while NAND remains weak.\n- Overall earnings expected to meet consensus.\n\n10. Rerating Justification\n\n- Enhanced earnings visibility based on HBM-centric structural profitability improvement and predictable AI demand.\n- Expectation of shortened memory cycles and a shift to an AI demand-driven ASP cycle.\n- Application of a premium based on HBM technology (P/B 1.9x, applying upcycle valuation).\n\n11. Financial Stability and Cash Flow\n\n- FCFF (Free Cash Flow to Firm) projected at KRW 1.2 trillion, KRW 2.4 trillion, and KRW 3.0 trillion for 2025, 2026, and 2027, respectively.\n- Expected to convert to net cash after 2026 (Net cash of KRW 5.2 trillion in 2027).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026223746874970866, "ret_signed_m1d": -0.026223746874970866, "alpha_spy_m1d": -0.025663696077025233, "alpha_c_m1d": -0.012924312773072266, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024475540972238585, "ret_signed_p1d": 0.024475540972238585, "alpha_spy_p1d": 0.01665176109775124, "alpha_c_p1d": 0.017208066936997835, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": -0.024475540972238585, "ret_signed_p1w": -0.024475540972238585, "alpha_spy_p1w": -0.04643168161393074, "alpha_c_p1w": -0.04645950253926667, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": -0.05769233090280079, "ret_signed_p1m": -0.05769233090280079, "alpha_spy_p1m": -0.10265871214024735, "alpha_c_p1m": -0.10347720597372856, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.22904547606599368, "ret_signed_p3m": 0.22904547606599368, "alpha_spy_p3m": 0.1276343082434872, "alpha_c_p3m": 0.05836957194775105, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.934242090254298, "ret_signed_p6m": 0.934242090254298, "alpha_spy_p6m": 0.8169252364461097, "alpha_c_p6m": 0.6724324866610625, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3044, -0.3344, -0.3124, -0.3093, -0.3208, -0.3641, -0.4248, -0.4084, -0.4241, -0.3606, -0.369, -0.3711, -0.3697, -0.3927, -0.3892, -0.3892, -0.3836, -0.3934, -0.3683, -0.3777, -0.3564, -0.3648, -0.369, -0.3805, -0.3805, -0.3508, -0.3508, -0.3508, -0.3341, -0.3358, -0.3365, -0.3194, -0.3072, -0.281, -0.3002, -0.2863, -0.3041, -0.295, -0.3002, -0.3128, -0.302, -0.2915, -0.2932, -0.274, -0.2587, -0.285, -0.2745, -0.2745, -0.2395, -0.215, -0.215, -0.1993, -0.1941, -0.1608, -0.1766, -0.1766, -0.1329, -0.1294, -0.1381, -0.1399, -0.1014, -0.0927, -0.0262, 0.0, 0.0245, -0.007, 0.021, -0.0017, -0.0245, -0.0262, -0.0542, -0.0524, -0.014, -0.0175, 0.0385, 0.0297, 0.049, 0.0437, 0.035, -0.0577, -0.0594, -0.0472, -0.0612, -0.0594, -0.0577, -0.0699, -0.0839, -0.0822, -0.0787, -0.0437, -0.0979, -0.0979, -0.0787, -0.0962, -0.0839, -0.1031, -0.0664, -0.0594, -0.028, -0.0332, -0.0332, -0.0647, -0.0804, -0.1066, -0.1434, -0.1224, -0.0927, -0.0857, -0.0909, -0.0598, -0.0581, -0.1036, -0.0878, -0.0808, -0.0703, -0.0423, -0.0301, 0.0084, 0.0645, 0.075, 0.1503, 0.159, 0.2185, 0.1678, 0.2448, 0.2448, 0.229, 0.2641, 0.2518, 0.2483, 0.1783, 0.222, 0.2168, 0.2606, 0.3849, 0.3849, 0.3849, 0.3849, 0.3849, 0.3849, 0.4987, 0.4531, 0.4409, 0.4794, 0.5845, 0.63, 0.7, 0.6772, 0.686, 0.6755, 0.7858, 0.8733, 0.8243, 0.9539, 0.9889, 0.9574, 1.171, 1.0519, 1.0274, 1.0764, 1.0309, 1.1219, 1.1675, 1.1605, 1.143, 0.9609, 1.1219, 0.9959, 0.9679, 0.9994, 0.8243, 0.8208, 0.8173, 0.8348, 0.9062, 0.8572, 0.8852, 0.9553, 0.9342, 0.8992, 0.9062, 1.0218, 0.9833, 1.0569, 0.9798, 1.0008, 0.9413, 0.8572, 0.9307, 0.9342]}
{"unit_id": "thread:1923248967787348083", "ticker_idx": 0, "ts": "2025-05-16T05:27:40+00:00", "symbol": "HBM memory", "kind": "sector", "geography": "global", "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We're not just heading for a memory supercycle—an ultra cycle is coming. For AI, HBM is now absolutely essential.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish HBM into an 'ultra cycle': LLM demand exponentially outpaces supply; HBM4+ stacks/SoW are inevitable; China structurally behind, boosting Western HBM dominance.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "Global HBM memory basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DeepSeek paper review: 1. Why is HBM demand skyrocketing?\n\n(1) Exponential growth in model size and context length\n\nThe paper states that the number of LLM parameters, context window length, and batch size are increasing by an order of magnitude nearly every year. As a result, memory requirements are growing dramatically.\nIn other words, the pace of LLM development far outstrips the advancement of HBM technology.\n\n---\n\n(2) The pace and physical limits of HBM supply\n\nIn contrast, the increase in HBM capacity is less than 50% per year.\n\nThis means that HBM manufacturers can never keep up with the speed demanded by models and data. Naturally, the bottleneck in large-scale AI systems is shifting to memory.\n\n---\n\n(3) Fundamental limitations of compression techniques like MLA/FP8\n\nThe paper introduces various methods such as MLA (Multi-head Latent Attention) for KV cache, low-precision (FP8) quantization, and activation quantization to significantly reduce the HBM usage per token.\n\nMLA: Reduces memory usage by 4–5 times compared to previous methods\n\nFP8: Cuts memory usage in half compared to BF16\n\n---\n\nHowever, because the model size, number of tokens, and context length are all growing exponentially, the overall demand for HBM inevitably continues to increase.\n\n---\n\n(4) HBM bottlenecks in real data center environments\n\nThe paper emphasizes that, in real data centers, HBM needs to be much faster than it is today in order to fully saturate 160 lanes of PCIe 6.0 (over 1 TB/s).\n\n---\n\nIn other words, meeting these requirements in the future will only be possible by increasing the number of HBM stacks, expanding the size of interposers, or adopting system-on-wafer (SoW) architectures.\n\n---\n\n(5) The need for next-generation HBM (3D DRAM stacking, SoW, etc.)\n\nThe paper clearly states that \"DRAM-stacked accelerators\" and \"system-on-wafer (SoW)\" represent the future of LLM infrastructure.\n\nDRAM-stacked: HBM4/4E and beyond, with 3D stacked DRAM delivering over 1.2 TB/s per stack\n\nSoW: Integration of DRAM and logic at the wafer level, enabling memory bandwidth of several TB/s\n\n---\n\nMy thoughts after reading this paper:\n\nWe’re not just heading for a memory supercycle—an ultra cycle is coming.\n\nFor AI, HBM is now absolutely essential.\n\nWith NVIDIA Blackwell, the most expensive component is no longer the GPU chip, but the HBM chips.\n\nIn the short term, software techniques like MLA and FP8 can temporarily reduce HBM demand, but in the medium to long term, LLM trends (such as ever-larger models and longer context windows) will drive an “ultra cycle” after HBM4 (>1.2 TB/s), with more stacks, larger interposers, and highly integrated packaging like SoW.\n\nIn this respect, China is still behind. HBM packaging capabilities within China are only at about the HBM3E level.\n\nIn my view, the U.S. has also realized that memory is the alpha and omega of the AI era. The U.S. is sanctioning CXMT to prevent China from independently sourcing HBM.\n\nEven if DeepSeek successfully develops its own ASIC, there’s a real possibility that HBM will remain the biggest bottleneck for China’s AI development.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012259166571098786, "ret_signed_m1d": -0.012259166571098786, "alpha_spy_m1d": -0.005965059588630658, "alpha_c_m1d": -0.015140412181541127, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": 0.00288124561044234, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011970560875108172, "ret_signed_p1d": -0.011970560875108172, "alpha_spy_p1d": -0.013064478975146177, "alpha_c_p1d": -0.010306698369892842, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.00166386250521533, "ret_p1w": -0.03834146492295809, "ret_signed_p1w": -0.03834146492295809, "alpha_spy_p1w": -0.012946024802480638, "alpha_c_p1w": -0.001899602621851182, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03644186230110691, "ret_p1m": 0.1482678887473127, "ret_signed_p1m": 0.1482678887473127, "alpha_spy_p1m": 0.13399656401966245, "alpha_c_p1m": 0.0794422941728449, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.0688255945744678, "ret_p3m": 0.30157748310745425, "ret_signed_p3m": 0.30157748310745425, "alpha_spy_p3m": 0.21306092186897563, "alpha_c_p3m": 0.07894942510712166, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.22262805800033258, "ret_p6m": 1.4505510201616207, "ret_signed_p6m": 1.4505510201616207, "alpha_spy_p6m": 1.2971524091435829, "alpha_c_p6m": 0.9945409496535293, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.4560100705080914, "price_path": [0.0348, 0.0508, 0.0345, 0.0146, -0.0096, -0.0249, -0.0091, -0.035, -0.054, -0.0645, -0.0689, -0.0498, -0.0664, -0.0573, -0.0846, -0.0787, -0.0306, -0.0331, -0.0052, 0.0221, 0.0125, 0.0229, 0.0434, 0.033, 0.027, 0.0078, 0.02, 0.0076, -0.0214, -0.0544, -0.0321, -0.031, -0.0919, -0.1537, -0.1866, -0.1868, -0.1552, -0.1321, -0.1448, -0.1349, -0.1318, -0.1596, -0.1574, -0.1562, -0.16, -0.1551, -0.13, -0.1192, -0.1012, -0.1087, -0.1163, -0.1232, -0.1204, -0.1036, -0.1046, -0.1043, -0.0876, -0.0798, -0.0765, -0.0302, -0.0128, -0.0031, -0.0123, 0.0, -0.012, -0.009, -0.0203, -0.0355, -0.0383, -0.0305, -0.0258, -0.0058, 0.0047, -0.015, 0.0061, 0.02, 0.0456, 0.075, 0.0827, 0.1023, 0.112, 0.1381, 0.1289, 0.1199, 0.1483, 0.1569, 0.1678, 0.1635, 0.1893, 0.1794, 0.2449, 0.2596, 0.2603, 0.2471, 0.2491, 0.2328, 0.2287, 0.2474, 0.2314, 0.2151, 0.2466, 0.2317, 0.2643, 0.2745, 0.2627, 0.2725, 0.2618, 0.2195, 0.2249, 0.2308, 0.2, 0.2052, 0.2102, 0.2023, 0.2223, 0.2267, 0.2496, 0.2398, 0.1852, 0.1998, 0.2143, 0.1987, 0.225, 0.2476, 0.2764, 0.294, 0.3016, 0.3008, 0.2858, 0.2708, 0.2583, 0.2325, 0.2111, 0.2321, 0.2423, 0.2387, 0.2423, 0.2654, 0.2566, 0.223, 0.2374, 0.2456, 0.271, 0.3049, 0.3144, 0.3536, 0.4024, 0.4481, 0.5177, 0.5301, 0.5786, 0.5518, 0.6305, 0.6095, 0.6275, 0.657, 0.6394, 0.6253, 0.5776, 0.6281, 0.6355, 0.7189, 0.8046, 0.8189, 0.8296, 0.8116, 0.8486, 0.8343, 0.8784, 0.8887, 0.8534, 0.9082, 1.0094, 1.0313, 1.0802, 1.0507, 1.0483, 1.059, 1.1661, 1.2296, 1.1981, 1.2806, 1.3094, 1.314, 1.4724, 1.3232, 1.3526, 1.37, 1.3396, 1.4506]}
{"unit_id": "orig:1934408453679911374", "ticker_idx": 0, "ts": "2025-06-16T00:31:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We anticipate mid-to-long-term benefits for domestic memory companies such as SK Hynix", "tweet_type": "analysis", "timeline": "years", "summary": "Meritz bullish on SK Hynix and Samsung long-term: AI DC era drives stair-step DRAM demand surges, conservative supply = severe shortage, commodity memory upgrading to HBM.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Two Implications of the 'SK Ulsan AI DC Project' - Meritz Securities, Kim Sun-woo\n\nSK Group to Build Ulsan AI Data Center (DC) Project. Collaboration with Amazon is a Surprise.\n\nAccording to several media reports released today, SK Group is entering the AI data center (DC) business in Ulsan and plans to hold a groundbreaking ceremony in August. The project will be built in the Mipo National Industrial Complex in Ulsan Nam-gu, with a 40MW capacity expected to be operational by November 2027, expanding to a total of 103MW by February 2029. If this plan proceeds, the project will be the largest DC project in Korea (Figure 1).\n\nThe most surprising aspects of this project are its partners and scale. 1) It will involve collaboration with Amazon (AWS), marking the first time a major US tech company's investment and participation are planned in Korea. 2) It plans to start with a 103MW capacity and has future expansion plans to an astounding 1GW level. Considering that the Stargate Project in Texas, USA, is starting with 200MW and has long-term expansion plans to 1.2GW, this project is comparable in its gigantic scale.\n\nImplication 1. The Front-End Industry for the Next 20 Years (PC → Smartphone → AI DC)\n\nOur recent report, 'Computex 2025 Meritz Field Visit: AI NEXT,' emphasized the importance of AI DC in the mid-to-long-term direction of change in the semiconductor industry. Currently, society is entering a new era of IT (Intelligence Technology) from traditional IT (Information Technology) amidst an exponential increase in information, and the integration and scaling up of semiconductors are enabling this. In the past, during the first and second industrial revolutions, improvements in thermal efficiency led to productivity enhancements, which in turn led to the creation of large-scale factories and industrial complexes. Similarly, in the upcoming 'Intelligence Technology' era, large-scale factories and industrial complexes, namely AI DCs, will emerge as key and most promising industries, serving as 'intelligence' production complexes for the future.\n\nThis aligns with the concept of 'AI Factory' emphasized by NVIDIA's Jensen Huang. As the era where 'information' creates added value wanes, and 'information' transforms into 'intelligence,' large-scale intelligence industrial complexes are expected to become the front-end industry responsible for the next 15-20 years. After 20 years of the PC era from 1980, followed by 20 years of the smartphone era in the 2000s, AI DC is now expected to emerge as the most crucial front-end industry for at least the next 20 years.\n\nSuccess in a front-end industry determines the success or failure of not only that company but also the entire group. In Korea, for example, the success of 'smartphones' in the 2000s led to revenue growth for over a decade for not only related affiliates like camera modules, substrates, displays, batteries, and semiconductor components, but also construction affiliates involved in component production.\n\nSimilarly, if the business viability of SK AI DC is proven, it will lead to performance improvements for all related member companies. Overall benefits are expected through vertical integration, including not only DC operation businesses but also energy suppliers, semiconductor/material/component suppliers, construction companies, and security firms. Notably, it should be emphasized that SK Group can now move beyond being merely a memory semiconductor supplier to becoming a key semiconductor procurement entity, including GPUs, thereby potentially building stronger negotiation power.\n\nAlready, AI DC investments are accelerating worldwide, not only from companies but also from sovereign funds (Figure 2). Companies newly entering the AI DC business often lack accumulated operational know-how, making various partnerships essential. According to our channel checks, SK Group has been preparing to establish a technological alliance for several years for this Ulsan AI DC investment. Technical collaborations with European power and cooling technology companies and US DC operating companies are expected to be announced.\n\nImplication 2. Shift in DRAM Supply-Demand Cycle: Demand Curve Transforms into Stair-Step Surges\n\nIn the past, the DRAM industry, due to its commodity characteristics, primarily competed on cost rather than price. As homogeneous goods were produced, prices for all products were similar, and ultimately, large-scale production capacity (operating leverage) and technological leadership (cost reduction) were key competitive advantages.\n\nFurthermore, discrepancies in supply and demand timing were the main drivers of DRAM cycles. In contrast to 'demand,' which fluctuated sharply in a curved manner depending on the economy, 'supply,' which had a production lag, needed to be predicted and executed 1-2 years in advance as a result of prior investment to ensure timely delivery.\n\nAfter the 20-year PC era and the subsequent 20-year smartphone era, the new 20-year AI DC era is expected to form a new semiconductor demand curve. In the past, PCs and smartphones generated demand with annual sales of 300 million and 1.5 billion units, respectively. Accordingly, suppliers projected product sales for the next 12 months, considering economic and technological conditions, and executed production investments based on estimated supply availability given competitive situations.\n\nNow, the front-end market, composed of billions of devices, is transforming into tens to hundreds of large-scale DC investments. Ultimately, the demand generated by the investment in a single AI DC site will transform from a marginal demand increase in the past to a massive, step-function surge. Simply put, the decision to invest in one AI DC facility can be described as a situation where demand for millions of smartphones suddenly emerges. The problem is that in this process, the accuracy of memory suppliers' demand forecasts will significantly decrease compared to the past, and the risk of a misforecast will become immense. Consequently, suppliers are likely to lean more towards a collective risk-aversion approach rather than the past volume-competition structure. Ultimately, the large-scale demand increase centered around AI DCs will transform the demand curve into stair-step surges, and a slow, minimum-scale preparation of supply will highly likely lead to a severe supply shortage.\n\nWe have been expecting an improvement in the DRAM cycle not only this year but also through 2026. This is because, in the new era of intelligent technology, commodity memory is transforming into specialty memory (such as HBM), leading to an increase in chip size, while suppliers continue to engage in conservative investment and face capacity constraints. With China's pursuit continuing as a supply constraint factor for some time, there is also a strong possibility that stair-step demand surges from AI DCs next year could create an unprecedented supply shortage. We anticipate mid-to-long-term benefits for domestic memory companies such as SK Hynix and Samsung Electronics.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05040316776889919, "ret_signed_m1d": -0.05040316776889919, "alpha_spy_m1d": -0.04097863811701141, "alpha_c_m1d": -0.026141530431787574, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004032278728435923, "ret_signed_p1d": 0.004032278728435923, "alpha_spy_p1d": 0.012577311265014823, "alpha_c_p1d": 0.010828568449153808, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.046370952307773106, "ret_signed_p1w": 0.046370952307773106, "alpha_spy_p1w": 0.047624947612409874, "alpha_c_p1w": 0.051914287618945676, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.20362919004367397, "ret_signed_p1m": 0.20362919004367397, "alpha_spy_p1m": 0.16828827940461677, "alpha_c_p1m": 0.09952111517093076, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.2396913036108994, "ret_signed_p3m": 0.2396913036108994, "alpha_spy_p3m": 0.1452892832874071, "alpha_c_p3m": 0.08573133422575951, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 1.287191450865563, "ret_signed_p6m": 1.287191450865563, "alpha_spy_p6m": 1.1473519217007484, "alpha_c_p6m": 0.8861739774981086, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.1729, -0.1548, -0.1326, -0.1487, -0.1628, -0.1387, -0.1668, -0.1978, -0.2324, -0.2071, -0.2035, -0.2167, -0.2666, -0.3367, -0.3178, -0.3359, -0.2626, -0.2723, -0.2747, -0.2731, -0.2997, -0.2956, -0.2956, -0.2892, -0.3005, -0.2715, -0.2823, -0.2578, -0.2675, -0.2723, -0.2856, -0.2856, -0.2514, -0.2514, -0.2514, -0.232, -0.234, -0.2348, -0.2151, -0.201, -0.1709, -0.193, -0.1769, -0.1974, -0.187, -0.193, -0.2075, -0.195, -0.1829, -0.1849, -0.1628, -0.1452, -0.1754, -0.1633, -0.1633, -0.123, -0.0948, -0.0948, -0.0766, -0.0706, -0.0323, -0.0504, -0.0504, 0.0, 0.004, -0.006, -0.0081, 0.0363, 0.0464, 0.123, 0.1532, 0.1815, 0.1452, 0.1774, 0.1512, 0.125, 0.123, 0.0907, 0.0927, 0.1371, 0.1331, 0.1976, 0.1875, 0.2097, 0.2036, 0.1935, 0.0867, 0.0847, 0.0988, 0.0827, 0.0847, 0.0867, 0.0726, 0.0565, 0.0585, 0.0625, 0.1028, 0.0403, 0.0403, 0.0625, 0.0423, 0.0565, 0.0343, 0.0766, 0.0847, 0.121, 0.1149, 0.1149, 0.0786, 0.0605, 0.0302, -0.0121, 0.0121, 0.0464, 0.0544, 0.0484, 0.0842, 0.0862, 0.0337, 0.0519, 0.06, 0.0721, 0.1044, 0.1185, 0.163, 0.2276, 0.2397, 0.3265, 0.3366, 0.4053, 0.3467, 0.4355, 0.4355, 0.4174, 0.4577, 0.4436, 0.4396, 0.3588, 0.4093, 0.4032, 0.4537, 0.5971, 0.5971, 0.5971, 0.5971, 0.5971, 0.5971, 0.7283, 0.6758, 0.6617, 0.7061, 0.8272, 0.8797, 0.9605, 0.9342, 0.9443, 0.9322, 1.0594, 1.1604, 1.1038, 1.2532, 1.2936, 1.2573, 1.5036, 1.3663, 1.338, 1.3946, 1.3421, 1.4471, 1.4996, 1.4915, 1.4713, 1.2613, 1.4471, 1.3017, 1.2694, 1.3057, 1.1038, 1.0998, 1.0958, 1.116, 1.1983, 1.1417, 1.174, 1.2549, 1.2306, 1.1902, 1.1983, 1.3316, 1.2872]}
{"unit_id": "orig:1934408453679911374", "ticker_idx": 1, "ts": "2025-06-16T00:31:29+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We anticipate mid-to-long-term benefits for domestic memory companies such as SK Hynix and Samsung Electronics", "tweet_type": "analysis", "timeline": "years", "summary": "Meritz bullish on SK Hynix and Samsung long-term: AI DC era drives stair-step DRAM demand surges, conservative supply = severe shortage, commodity memory upgrading to HBM.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Two Implications of the 'SK Ulsan AI DC Project' - Meritz Securities, Kim Sun-woo\n\nSK Group to Build Ulsan AI Data Center (DC) Project. Collaboration with Amazon is a Surprise.\n\nAccording to several media reports released today, SK Group is entering the AI data center (DC) business in Ulsan and plans to hold a groundbreaking ceremony in August. The project will be built in the Mipo National Industrial Complex in Ulsan Nam-gu, with a 40MW capacity expected to be operational by November 2027, expanding to a total of 103MW by February 2029. If this plan proceeds, the project will be the largest DC project in Korea (Figure 1).\n\nThe most surprising aspects of this project are its partners and scale. 1) It will involve collaboration with Amazon (AWS), marking the first time a major US tech company's investment and participation are planned in Korea. 2) It plans to start with a 103MW capacity and has future expansion plans to an astounding 1GW level. Considering that the Stargate Project in Texas, USA, is starting with 200MW and has long-term expansion plans to 1.2GW, this project is comparable in its gigantic scale.\n\nImplication 1. The Front-End Industry for the Next 20 Years (PC → Smartphone → AI DC)\n\nOur recent report, 'Computex 2025 Meritz Field Visit: AI NEXT,' emphasized the importance of AI DC in the mid-to-long-term direction of change in the semiconductor industry. Currently, society is entering a new era of IT (Intelligence Technology) from traditional IT (Information Technology) amidst an exponential increase in information, and the integration and scaling up of semiconductors are enabling this. In the past, during the first and second industrial revolutions, improvements in thermal efficiency led to productivity enhancements, which in turn led to the creation of large-scale factories and industrial complexes. Similarly, in the upcoming 'Intelligence Technology' era, large-scale factories and industrial complexes, namely AI DCs, will emerge as key and most promising industries, serving as 'intelligence' production complexes for the future.\n\nThis aligns with the concept of 'AI Factory' emphasized by NVIDIA's Jensen Huang. As the era where 'information' creates added value wanes, and 'information' transforms into 'intelligence,' large-scale intelligence industrial complexes are expected to become the front-end industry responsible for the next 15-20 years. After 20 years of the PC era from 1980, followed by 20 years of the smartphone era in the 2000s, AI DC is now expected to emerge as the most crucial front-end industry for at least the next 20 years.\n\nSuccess in a front-end industry determines the success or failure of not only that company but also the entire group. In Korea, for example, the success of 'smartphones' in the 2000s led to revenue growth for over a decade for not only related affiliates like camera modules, substrates, displays, batteries, and semiconductor components, but also construction affiliates involved in component production.\n\nSimilarly, if the business viability of SK AI DC is proven, it will lead to performance improvements for all related member companies. Overall benefits are expected through vertical integration, including not only DC operation businesses but also energy suppliers, semiconductor/material/component suppliers, construction companies, and security firms. Notably, it should be emphasized that SK Group can now move beyond being merely a memory semiconductor supplier to becoming a key semiconductor procurement entity, including GPUs, thereby potentially building stronger negotiation power.\n\nAlready, AI DC investments are accelerating worldwide, not only from companies but also from sovereign funds (Figure 2). Companies newly entering the AI DC business often lack accumulated operational know-how, making various partnerships essential. According to our channel checks, SK Group has been preparing to establish a technological alliance for several years for this Ulsan AI DC investment. Technical collaborations with European power and cooling technology companies and US DC operating companies are expected to be announced.\n\nImplication 2. Shift in DRAM Supply-Demand Cycle: Demand Curve Transforms into Stair-Step Surges\n\nIn the past, the DRAM industry, due to its commodity characteristics, primarily competed on cost rather than price. As homogeneous goods were produced, prices for all products were similar, and ultimately, large-scale production capacity (operating leverage) and technological leadership (cost reduction) were key competitive advantages.\n\nFurthermore, discrepancies in supply and demand timing were the main drivers of DRAM cycles. In contrast to 'demand,' which fluctuated sharply in a curved manner depending on the economy, 'supply,' which had a production lag, needed to be predicted and executed 1-2 years in advance as a result of prior investment to ensure timely delivery.\n\nAfter the 20-year PC era and the subsequent 20-year smartphone era, the new 20-year AI DC era is expected to form a new semiconductor demand curve. In the past, PCs and smartphones generated demand with annual sales of 300 million and 1.5 billion units, respectively. Accordingly, suppliers projected product sales for the next 12 months, considering economic and technological conditions, and executed production investments based on estimated supply availability given competitive situations.\n\nNow, the front-end market, composed of billions of devices, is transforming into tens to hundreds of large-scale DC investments. Ultimately, the demand generated by the investment in a single AI DC site will transform from a marginal demand increase in the past to a massive, step-function surge. Simply put, the decision to invest in one AI DC facility can be described as a situation where demand for millions of smartphones suddenly emerges. The problem is that in this process, the accuracy of memory suppliers' demand forecasts will significantly decrease compared to the past, and the risk of a misforecast will become immense. Consequently, suppliers are likely to lean more towards a collective risk-aversion approach rather than the past volume-competition structure. Ultimately, the large-scale demand increase centered around AI DCs will transform the demand curve into stair-step surges, and a slow, minimum-scale preparation of supply will highly likely lead to a severe supply shortage.\n\nWe have been expecting an improvement in the DRAM cycle not only this year but also through 2026. This is because, in the new era of intelligent technology, commodity memory is transforming into specialty memory (such as HBM), leading to an increase in chip size, while suppliers continue to engage in conservative investment and face capacity constraints. With China's pursuit continuing as a supply constraint factor for some time, there is also a strong possibility that stair-step demand surges from AI DCs next year could create an unprecedented supply shortage. We anticipate mid-to-long-term benefits for domestic memory companies such as SK Hynix and Samsung Electronics.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01923074383815826, "ret_signed_m1d": 0.01923074383815826, "alpha_spy_m1d": 0.02865527349004604, "alpha_c_m1d": 0.035153977513263546, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.015923233675105286, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015734213369598793, "ret_signed_p1d": 0.015734213369598793, "alpha_spy_p1d": 0.024279245906177693, "alpha_c_p1d": 0.022811149988016477, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.007076936618417684, "ret_p1w": 0.013985982883102821, "ret_signed_p1w": 0.013985982883102821, "alpha_spy_p1w": 0.015239978187739589, "alpha_c_p1w": 0.012987963074120756, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": 0.0009980198089820647, "ret_p1m": 0.1204671603214782, "ret_signed_p1m": 0.1204671603214782, "alpha_spy_p1m": 0.085126249682421, "alpha_c_p1m": 0.05752491887102451, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.06294224145045368, "ret_p3m": 0.2910876730361438, "ret_signed_p3m": 0.2910876730361438, "alpha_spy_p3m": 0.19668565271265148, "alpha_c_p3m": 0.17394953367062693, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.11713813936551687, "ret_p6m": 0.9152361177913726, "ret_signed_p6m": 0.9152361177913726, "alpha_spy_p6m": 0.7753965886265579, "alpha_c_p6m": 0.6935780397820375, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.22165807800933512, "price_path": [0.0167, 0.0462, 0.0723, 0.0514, 0.0393, 0.0671, 0.074, 0.0524, 0.0105, 0.028, 0.028, 0.007, -0.0192, -0.0699, -0.0647, -0.0734, -0.014, -0.035, -0.0175, -0.0105, -0.0437, -0.0367, -0.0332, -0.0315, -0.0385, -0.0262, -0.0262, -0.0262, -0.0245, -0.0245, -0.0297, -0.0297, -0.0507, -0.0507, -0.0507, -0.0455, -0.0455, -0.042, 0.007, -0.0052, 0.0035, 0.0017, -0.007, -0.0245, -0.0227, -0.0262, -0.0437, -0.0524, -0.0437, -0.0577, -0.0227, -0.0192, -0.0175, -0.007, -0.007, 0.0105, 0.0332, 0.0332, 0.0455, 0.035, 0.0472, 0.0402, 0.0192, 0.0, 0.0157, 0.0455, 0.035, 0.0402, 0.014, 0.0577, 0.0717, 0.0524, 0.0695, 0.0519, 0.0589, 0.0695, 0.1222, 0.1134, 0.0853, 0.08, 0.0624, 0.073, 0.1011, 0.0994, 0.1205, 0.1381, 0.1732, 0.1803, 0.1926, 0.1609, 0.168, 0.1609, 0.1592, 0.2383, 0.2418, 0.277, 0.2559, 0.2119, 0.226, 0.2295, 0.2102, 0.2401, 0.2629, 0.2489, 0.2506, 0.2647, 0.2594, 0.2594, 0.2313, 0.2313, 0.2401, 0.2418, 0.2559, 0.2577, 0.2366, 0.2418, 0.2242, 0.226, 0.1891, 0.2155, 0.2278, 0.233, 0.2225, 0.233, 0.2577, 0.277, 0.2911, 0.3263, 0.3456, 0.3966, 0.3755, 0.4125, 0.4125, 0.4687, 0.4899, 0.5022, 0.5145, 0.4652, 0.4877, 0.4824, 0.5195, 0.5857, 0.5857, 0.5857, 0.5857, 0.5857, 0.5857, 0.6679, 0.6484, 0.6184, 0.6785, 0.7262, 0.7297, 0.7333, 0.7227, 0.7421, 0.705, 0.7456, 0.8022, 0.758, 0.7757, 0.8393, 0.8993, 0.9629, 0.8534, 0.7774, 0.7527, 0.7297, 0.7774, 0.8287, 0.8216, 0.8163, 0.7174, 0.7774, 0.728, 0.705, 0.7774, 0.6749, 0.7085, 0.7545, 0.8163, 0.8287, 0.7757, 0.781, 0.8269, 0.8463, 0.8569, 0.9152, 0.9347, 0.9152]}
{"unit_id": "orig:1934408453679911374", "ticker_idx": 2, "ts": "2025-06-16T00:31:29+00:00", "symbol": "DRAM", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "stair-step demand surges from AI DCs next year could create an unprecedented supply shortage", "tweet_type": "analysis", "timeline": "years", "summary": "Meritz bullish on SK Hynix and Samsung long-term: AI DC era drives stair-step DRAM demand surges, conservative supply = severe shortage, commodity memory upgrading to HBM.", "resolved_tickers": ["005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: Samsung Electronics + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Two Implications of the 'SK Ulsan AI DC Project' - Meritz Securities, Kim Sun-woo\n\nSK Group to Build Ulsan AI Data Center (DC) Project. Collaboration with Amazon is a Surprise.\n\nAccording to several media reports released today, SK Group is entering the AI data center (DC) business in Ulsan and plans to hold a groundbreaking ceremony in August. The project will be built in the Mipo National Industrial Complex in Ulsan Nam-gu, with a 40MW capacity expected to be operational by November 2027, expanding to a total of 103MW by February 2029. If this plan proceeds, the project will be the largest DC project in Korea (Figure 1).\n\nThe most surprising aspects of this project are its partners and scale. 1) It will involve collaboration with Amazon (AWS), marking the first time a major US tech company's investment and participation are planned in Korea. 2) It plans to start with a 103MW capacity and has future expansion plans to an astounding 1GW level. Considering that the Stargate Project in Texas, USA, is starting with 200MW and has long-term expansion plans to 1.2GW, this project is comparable in its gigantic scale.\n\nImplication 1. The Front-End Industry for the Next 20 Years (PC → Smartphone → AI DC)\n\nOur recent report, 'Computex 2025 Meritz Field Visit: AI NEXT,' emphasized the importance of AI DC in the mid-to-long-term direction of change in the semiconductor industry. Currently, society is entering a new era of IT (Intelligence Technology) from traditional IT (Information Technology) amidst an exponential increase in information, and the integration and scaling up of semiconductors are enabling this. In the past, during the first and second industrial revolutions, improvements in thermal efficiency led to productivity enhancements, which in turn led to the creation of large-scale factories and industrial complexes. Similarly, in the upcoming 'Intelligence Technology' era, large-scale factories and industrial complexes, namely AI DCs, will emerge as key and most promising industries, serving as 'intelligence' production complexes for the future.\n\nThis aligns with the concept of 'AI Factory' emphasized by NVIDIA's Jensen Huang. As the era where 'information' creates added value wanes, and 'information' transforms into 'intelligence,' large-scale intelligence industrial complexes are expected to become the front-end industry responsible for the next 15-20 years. After 20 years of the PC era from 1980, followed by 20 years of the smartphone era in the 2000s, AI DC is now expected to emerge as the most crucial front-end industry for at least the next 20 years.\n\nSuccess in a front-end industry determines the success or failure of not only that company but also the entire group. In Korea, for example, the success of 'smartphones' in the 2000s led to revenue growth for over a decade for not only related affiliates like camera modules, substrates, displays, batteries, and semiconductor components, but also construction affiliates involved in component production.\n\nSimilarly, if the business viability of SK AI DC is proven, it will lead to performance improvements for all related member companies. Overall benefits are expected through vertical integration, including not only DC operation businesses but also energy suppliers, semiconductor/material/component suppliers, construction companies, and security firms. Notably, it should be emphasized that SK Group can now move beyond being merely a memory semiconductor supplier to becoming a key semiconductor procurement entity, including GPUs, thereby potentially building stronger negotiation power.\n\nAlready, AI DC investments are accelerating worldwide, not only from companies but also from sovereign funds (Figure 2). Companies newly entering the AI DC business often lack accumulated operational know-how, making various partnerships essential. According to our channel checks, SK Group has been preparing to establish a technological alliance for several years for this Ulsan AI DC investment. Technical collaborations with European power and cooling technology companies and US DC operating companies are expected to be announced.\n\nImplication 2. Shift in DRAM Supply-Demand Cycle: Demand Curve Transforms into Stair-Step Surges\n\nIn the past, the DRAM industry, due to its commodity characteristics, primarily competed on cost rather than price. As homogeneous goods were produced, prices for all products were similar, and ultimately, large-scale production capacity (operating leverage) and technological leadership (cost reduction) were key competitive advantages.\n\nFurthermore, discrepancies in supply and demand timing were the main drivers of DRAM cycles. In contrast to 'demand,' which fluctuated sharply in a curved manner depending on the economy, 'supply,' which had a production lag, needed to be predicted and executed 1-2 years in advance as a result of prior investment to ensure timely delivery.\n\nAfter the 20-year PC era and the subsequent 20-year smartphone era, the new 20-year AI DC era is expected to form a new semiconductor demand curve. In the past, PCs and smartphones generated demand with annual sales of 300 million and 1.5 billion units, respectively. Accordingly, suppliers projected product sales for the next 12 months, considering economic and technological conditions, and executed production investments based on estimated supply availability given competitive situations.\n\nNow, the front-end market, composed of billions of devices, is transforming into tens to hundreds of large-scale DC investments. Ultimately, the demand generated by the investment in a single AI DC site will transform from a marginal demand increase in the past to a massive, step-function surge. Simply put, the decision to invest in one AI DC facility can be described as a situation where demand for millions of smartphones suddenly emerges. The problem is that in this process, the accuracy of memory suppliers' demand forecasts will significantly decrease compared to the past, and the risk of a misforecast will become immense. Consequently, suppliers are likely to lean more towards a collective risk-aversion approach rather than the past volume-competition structure. Ultimately, the large-scale demand increase centered around AI DCs will transform the demand curve into stair-step surges, and a slow, minimum-scale preparation of supply will highly likely lead to a severe supply shortage.\n\nWe have been expecting an improvement in the DRAM cycle not only this year but also through 2026. This is because, in the new era of intelligent technology, commodity memory is transforming into specialty memory (such as HBM), leading to an increase in chip size, while suppliers continue to engage in conservative investment and face capacity constraints. With China's pursuit continuing as a supply constraint factor for some time, there is also a strong possibility that stair-step demand surges from AI DCs next year could create an unprecedented supply shortage. We anticipate mid-to-long-term benefits for domestic memory companies such as SK Hynix and Samsung Electronics.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015586211965370467, "ret_signed_m1d": -0.015586211965370467, "alpha_spy_m1d": -0.006161682313482686, "alpha_c_m1d": 0.0003370217097348194, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.015923233675105286, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009883246049017358, "ret_signed_p1d": 0.009883246049017358, "alpha_spy_p1d": 0.01842827858559626, "alpha_c_p1d": 0.016960182667435042, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.007076936618417684, "ret_p1w": 0.030178467595437963, "ret_signed_p1w": 0.030178467595437963, "alpha_spy_p1w": 0.03143246290007473, "alpha_c_p1w": 0.0291804477864559, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": 0.0009980198089820647, "ret_p1m": 0.16204817518257608, "ret_signed_p1m": 0.16204817518257608, "alpha_spy_p1m": 0.12670726454351888, "alpha_c_p1m": 0.0991059337321224, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.06294224145045368, "ret_p3m": 0.2653894883235216, "ret_signed_p3m": 0.2653894883235216, "alpha_spy_p3m": 0.17098746800002929, "alpha_c_p3m": 0.14825134895800474, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.11713813936551687, "ret_p6m": 1.101213784328468, "ret_signed_p6m": 1.101213784328468, "alpha_spy_p6m": 0.9613742551636533, "alpha_c_p6m": 0.8795557063191328, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.22165807800933512, "price_path": [-0.0781, -0.0543, -0.0302, -0.0486, -0.0618, -0.0358, -0.0464, -0.0727, -0.111, -0.0896, -0.0877, -0.1049, -0.1429, -0.2033, -0.1912, -0.2047, -0.1383, -0.1536, -0.1461, -0.1418, -0.1717, -0.1662, -0.1644, -0.1603, -0.1695, -0.1489, -0.1543, -0.142, -0.146, -0.1484, -0.1576, -0.1576, -0.151, -0.151, -0.151, -0.1387, -0.1397, -0.1384, -0.1041, -0.1031, -0.0837, -0.0956, -0.0919, -0.1109, -0.1048, -0.1096, -0.1256, -0.1237, -0.1133, -0.1213, -0.0928, -0.0822, -0.0964, -0.0851, -0.0851, -0.0562, -0.0308, -0.0308, -0.0156, -0.0178, 0.0075, -0.0051, -0.0156, 0.0, 0.0099, 0.0197, 0.0135, 0.0383, 0.0302, 0.0903, 0.1125, 0.1169, 0.1073, 0.1146, 0.1051, 0.0972, 0.1226, 0.1021, 0.089, 0.1086, 0.0977, 0.1353, 0.1443, 0.1545, 0.162, 0.1658, 0.13, 0.1325, 0.1457, 0.1218, 0.1263, 0.1238, 0.1159, 0.1474, 0.1502, 0.1698, 0.1794, 0.1261, 0.1332, 0.146, 0.1263, 0.1483, 0.1486, 0.1627, 0.1677, 0.1928, 0.1872, 0.1872, 0.155, 0.1459, 0.1352, 0.1149, 0.134, 0.152, 0.1455, 0.1451, 0.1542, 0.1561, 0.1114, 0.1337, 0.1439, 0.1526, 0.1635, 0.1758, 0.2103, 0.2523, 0.2654, 0.3264, 0.3411, 0.4009, 0.3611, 0.424, 0.424, 0.4431, 0.4738, 0.4729, 0.477, 0.412, 0.4485, 0.4428, 0.4866, 0.5914, 0.5914, 0.5914, 0.5914, 0.5914, 0.5914, 0.6981, 0.6621, 0.64, 0.6923, 0.7767, 0.8047, 0.8469, 0.8284, 0.8432, 0.8186, 0.9025, 0.9813, 0.9309, 1.0145, 1.0664, 1.0783, 1.2333, 1.1099, 1.0577, 1.0736, 1.0359, 1.1123, 1.1641, 1.1565, 1.1438, 0.9893, 1.1123, 1.0148, 0.9872, 1.0416, 0.8894, 0.9042, 0.9251, 0.9661, 1.0135, 0.9587, 0.9775, 1.0409, 1.0385, 1.0236, 1.0568, 1.1332, 1.1012]}
{"unit_id": "orig:1922520474690740721", "ticker_idx": 0, "ts": "2025-05-14T05:12:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintaining Optimism, 2H Outlook Is Key Catalyst… management confidence and improved rack shipments in the second half are expected to be stock catalysts", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley Buy-equivalent note on NVDA: GB200 rack shipments recovering, full-year Blackwell growth on track, 2H catalysts intact.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: Improvement in Rack Shipment Data Eases Market Concerns; Full-Year Blackwell Growth Likely to Materialize\n\nMorgan Stanley (J. Moore, May 12, 2025)\n\n(1) GB200 Rack Deployment Initially Complex, Market Expectations Were Overly Optimistic\n\nMorgan Stanley notes that NVIDIA’s GB200-based full rack solution faced a slow initial rollout due to new technologies such as ARM architecture and liquid cooling. Previously, the Taiwan market had projected 70,000–90,000 rack shipments in 2025, far exceeding Morgan Stanley’s channel-based estimate of 25,000–30,000 units. Some in the market were concerned that NVIDIA’s shipments could far outpace actual assembly capacity, leading to inventory buildup.\n\n(2) Taiwan Shipment Data Significantly Improves, Annual Target Achievable\n\nLatest data shows that Foxconn, Wistron, and Quanta shipped 1,500 GB200 racks in April, bringing the total for Q1 and April to 2,500 units. Even if the monthly pace remains at 1,500 units, the full-year target of 15,000 units is attainable, easing concerns about a potential “shipment cliff.” Morgan Stanley expects monthly shipments to rise further, laying a solid foundation for Blackwell GPU shipments for the full year.\n\n(3) HGX Platform Supplements Supply, Providing a Complete Shipment Path\nIn addition to the NVL72 architecture, the HGX B200 platform is being deployed through customers like Coreweave in 48–64 GPU configurations. Wistron alone could drive shipments of 500,000 Blackwell GPUs in Q2 through motherboard deliveries. Combined with the estimated 320,000 units from NVL72, the total surpasses 820,000 units, approaching the quarterly 1 million-unit CoWoS capacity, dispelling concerns over a “capacity mismatch.”\n\n(4) Maintaining Optimism, 2H Outlook Is Key Catalyst\nMorgan Stanley emphasizes that its current view is more reliant on industry channel feedback rather than Taiwan shipment numbers. Channels broadly remain optimistic about demand over the next few quarters. Despite lowered guidance due to the H20 export ban and slightly below-consensus revenue, management confidence and improved rack shipments in the second half are expected to be stock catalysts. Ongoing recovery in Taiwan shipment data is a positive signal for market sentiment.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.039973377094377316, "ret_signed_m1d": -0.039973377094377316, "alpha_spy_m1d": -0.038696948916340124, "alpha_c_m1d": -0.030174563266902377, "bench_spy_m1d": -0.0012764281780371922, "bench_c_m1d": -0.00979881382747494, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0037681199952365185, "ret_signed_p1d": -0.0037681199952365185, "alpha_spy_p1d": -0.008652356816422846, "alpha_c_p1d": -0.0003002463395035404, "bench_spy_p1d": 0.004884236821186327, "bench_c_p1d": -0.003467873655732978, "ret_p1w": -0.02615626048767461, "ret_signed_p1w": -0.02615626048767461, "alpha_spy_p1w": -0.01810643135401302, "alpha_c_p1w": 0.00013510521301440814, "bench_spy_p1w": -0.00804982913366159, "bench_c_p1w": -0.026291365700689018, "ret_p1m": 0.07145027173226604, "ret_signed_p1m": 0.07145027173226604, "alpha_spy_p1m": 0.043948282336632616, "alpha_c_p1m": 0.010318829963762877, "bench_spy_p1m": 0.02750198939563342, "bench_c_p1m": 0.06113144176850316, "ret_p3m": 0.3452981160978814, "ret_signed_p3m": 0.3452981160978814, "alpha_spy_p3m": 0.2598476112316106, "alpha_c_p3m": 0.16218598063243883, "bench_spy_p3m": 0.08545050486627082, "bench_c_p3m": 0.18311213546544258, "ret_p6m": 0.3898602455333784, "ret_signed_p6m": 0.3898602455333784, "alpha_spy_p6m": 0.2425373826443955, "alpha_c_p6m": -0.02588235536198802, "bench_spy_p6m": 0.1473228628889829, "bench_c_p6m": 0.4157426008953664, "price_path": [0.0258, 0.0258, 0.0299, 0.0286, 0.0351, -0.0068, -0.0375, -0.0644, -0.0301, -0.1123, -0.0771, -0.1573, -0.1431, -0.1334, -0.1831, -0.1674, -0.2096, -0.1965, -0.1448, -0.146, -0.101, -0.1168, -0.1471, -0.1317, -0.1242, -0.1303, -0.1029, -0.1082, -0.1595, -0.1767, -0.1897, -0.1992, -0.1861, -0.1841, -0.2478, -0.3032, -0.2786, -0.2885, -0.1552, -0.2052, -0.1804, -0.182, -0.171, -0.2279, -0.2501, -0.2501, -0.284, -0.2693, -0.2411, -0.2136, -0.1798, -0.1966, -0.1945, -0.1952, -0.1753, -0.154, -0.159, -0.1611, -0.1351, -0.1328, -0.1381, -0.0912, -0.04, 0.0, -0.0038, 0.0004, 0.0017, -0.0071, -0.0262, -0.0185, -0.0299, -0.0299, 0.0012, -0.0039, 0.0284, -0.0016, 0.0151, 0.0434, 0.0486, 0.0344, 0.0471, 0.0539, 0.0637, 0.0554, 0.0715, 0.0491, 0.0692, 0.0649, 0.075, 0.075, 0.063, 0.0653, 0.0929, 0.1402, 0.1455, 0.1657, 0.1674, 0.1328, 0.162, 0.1774, 0.1774, 0.1693, 0.1823, 0.2036, 0.2126, 0.2186, 0.2124, 0.2614, 0.2663, 0.2784, 0.274, 0.2664, 0.2342, 0.2619, 0.2838, 0.282, 0.3061, 0.2969, 0.3247, 0.3143, 0.2837, 0.3301, 0.3172, 0.3258, 0.3358, 0.35, 0.3453, 0.3534, 0.3418, 0.345, 0.3334, 0.3449, 0.2979, 0.2961, 0.293, 0.3152, 0.3287, 0.3432, 0.3419, 0.3313, 0.2871, 0.2871, 0.2619, 0.2608, 0.2684, 0.2342, 0.2437, 0.2618, 0.3103, 0.3092, 0.314, 0.3135, 0.2923, 0.2584, 0.3024, 0.3055, 0.3568, 0.3185, 0.3078, 0.3131, 0.3168, 0.3438, 0.3788, 0.3837, 0.3958, 0.3865, 0.3711, 0.3674, 0.3975, 0.423, 0.3535, 0.3916, 0.3304, 0.3289, 0.3435, 0.3539, 0.3497, 0.3387, 0.3322, 0.3461, 0.3764, 0.4151, 0.4856, 0.53, 0.4993, 0.4963, 0.5288, 0.4683, 0.4425, 0.3899]}
{"unit_id": "orig:1934486017916150007", "ticker_idx": 0, "ts": "2025-06-16T05:39:42+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AVGO and MRVL Continue to Dominate ASICs - JPMorgan", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPMorgan's bullish AVGO/MRVL thesis on ASIC TAM raised to $30B; SK Hynix also benefits as ASIC HBM mix rises to 20-30% with improving ASP and margins.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250616_AVGO and MRVL Continue to Dominate ASICs - JPMorgan\n\n[Key Takeaways]\n\n(1) JPMorgan has raised its ASIC TAM (Total Addressable Market) forecast from the previous $25bn to $30bn.\n\n(2) This upward revision is due to the expectation that the ASIC market growth rate will accelerate as hyperscalers advance their next-generation ASIC development roadmaps.\n\n(3) Both Broadcom and Marvell Technology are presumed to have secured next-generation 2nm ASIC projects, and they are expected to maintain their overwhelming dominance in the market.\n\n[Implication]\n\n(1) Earlier this year, SK Hynix's ASIC revenue mix was communicated as 'double-digit, early 10%' levels.\n\n(2) According to JPMorgan, the shipment mix accounted for by ASICs on a unit basis has already reached 40%.\n\n(3) Currently, 1) the perception that ASICs primarily sell middle-to-low end products with lower ASPs compared to HBM integrated into AI GPUs is receding, and 2) expectations for Q (Quantity) growth are increasingly rising.\n\n(4) In fact, the memory integrated into ASICs is also being upgraded to high-end products [HBM3E~4] comparable to the level integrated into GPU systems, and in some cases, the simple profit margin is estimated to be equivalent to or even higher than that of NVIDIA-bound products.\n\n(5) Recently, SK Hynix has also started communicating its ASIC mix within HBM revenue at 20-30% from the previous double-digit level.\n\n(6) As CSPs (Cloud Service Providers) increase their adoption rate of ASICs, the growth in SK Hynix's ASIC-related revenue mix is expected to 1) alleviate concerns about P (Price) [higher-than-expected ASP and excellent margins], 2) drive company-wide top-line growth due to strong Q growth, and 3) be used as a card to sustain a tight HBM supply-demand balance [used as a negotiation tool with NVIDIA].\n\n(7) In conclusion, it is judged that the time when the performance contribution of ASICs can be visibly observed is being brought forward.\n\n(8) An external environment that can stimulate multiples due to strong annual EPS growth or expectations of additional EPS upside seems to be unfolding.\n\n[Contents]\n\n(1) Broadcom's Google TPU v6 [3nm] ramp-up has begun, and the TPU v7 tape-out was recently completed.\n\n(2) Furthermore, revenue recognition from various clients, including Meta's MTIA [3nm], OpenAI, and Softbank/ARM, is expected to fully materialize from 2026.\n\n(3) The cycle for new ASIC design and release is becoming shorter, and the complexity of projects is increasing, with the number of transistors per chip/chiplet rising to over 150-200 billion.\n\n(4) We estimate that Broadcom will record over $15bn in revenue this year from 3nm projects alone, and its total AI revenue [ASIC+Networking] is projected to reach $19-20bn.\n\n(5) In addition, recent channel checks indicate that Broadcom has already successfully secured 2nm projects, including TPU v8 and the next-generation MTIA.\n\n(6) Accordingly, full-scale mass production of new projects such as OpenAI's and Softbank/ARM's 2-3nm ASICs and Meta's next-generation MTIA is expected to begin contributing to revenue from 2H26.\n\n(7) In Marvell Technology's case, it has also reportedly succeeded in securing 2nm projects targeting ramp-up from 2027, including Amazon Trainium 4 and Microsoft Maia Gen 3.\n\n(8) The mass production of its first two ASIC projects, Trainium 2 [5nm] and Axion ARM CPU, is imminent, and Trainium/Maia [3nm] mass production is scheduled for 2026.\n\n(9) Meanwhile, investors often tend to roughly consider high-speed I/O SERDES circuit technology as the sole differentiator between Broadcom/Marvell Technology and other ASIC/COT competitors.\n\n(10) We agree that SERDES technology is a core competency, but we believe it is not the only one.\n\n(11) This is because, in addition to SERDES IP, factors such as 1) high-performance memory architecture interfaces, 2) advanced cell libraries, 3) advanced packaging integration [2.5D, 3D SOIC, organic interposer], and 4) HBM4 logic die design capabilities broadly contribute.\n\n(12) In fact, the only large CSPs or OEMs capable of fully undertaking cutting-edge chip design [including manufacturing → testing → packaging] internally are essentially Apple, Amazon [limited to CPU, DPU], and Cisco [switch/routing chips].\n\n(13) Although hyperscalers are continuously increasing ASIC adoption in pursuit of 1) performance differentiation and 2) lower power consumption, the current bottleneck is their lack of capability to 100% internalize large-scale and complex SoC designs and immediately reflect them in current AI server operations.\n\n(14) In this environment, all hyperscalers [Google/Microsoft/Meta/Amazon] and large OEMs [Nokia/Ericsson/Cisco/Samsung Electronics/Juniper/ZTE] developing ASICs have significant and minor relationships with Broadcom/Marvell Technology in the areas of 1) design IP, 2) full-chip design expertise [integrating over 100 billion transistors on a single silicon], and 3) advanced packaging integration.\n\n(15) These factors are the driving force behind Broadcom/Marvell Technology's dominance of the entire market, with their ASIC market share nearing 70-75% [Broadcom 55-60%, Marvell Technology 15%].\n\n(16) As of this year, ASIC XPUs are estimated to already account for a '40% mix' on a unit basis in the XPU/GPU market.\n\n(17) The ASIC market growth has accelerated over the past few years due to the slowdown of Moore's Law and the increasing complexity of AI computing workloads, and the adoption rate of ASICs by hyperscalers is also structurally expected to rise in the future.\n\n(18) Accordingly, we are raising our ASIC TAM from the previous $25bn to $30bn [CAGR +30%] and maintaining a positive opinion on the related value chain.\n\n$AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013486682457092392, "ret_signed_m1d": -0.013486682457092392, "alpha_spy_m1d": -0.004062152805204611, "alpha_c_m1d": 0.010774954880019227, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010829019318885025, "ret_signed_p1d": -0.010829019318885025, "alpha_spy_p1d": -0.0022839867823061244, "alpha_c_p1d": -0.00403272959816714, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.008993684358509002, "ret_signed_p1w": 0.008993684358509002, "alpha_spy_p1w": 0.01024767966314577, "alpha_c_p1w": 0.014537019669681572, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.11702199192450169, "ret_signed_p1m": 0.11702199192450169, "alpha_spy_p1m": 0.08168108128544449, "alpha_c_p1m": 0.012913917051758483, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.4298947685967165, "ret_signed_p3m": 0.4298947685967165, "alpha_spy_p3m": 0.3354927482732242, "alpha_c_p3m": 0.2759347992115766, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.6181834534415322, "ret_signed_p6m": 0.6181834534415322, "alpha_spy_p6m": 0.47834392427671757, "alpha_c_p6m": 0.21716598007407772, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.2266, -0.2442, -0.2397, -0.2414, -0.2532, -0.2889, -0.3178, -0.3292, -0.3359, -0.3315, -0.3174, -0.3891, -0.4197, -0.3886, -0.3811, -0.2656, -0.3165, -0.2783, -0.2925, -0.2902, -0.3074, -0.3217, -0.3217, -0.3407, -0.3273, -0.2983, -0.2537, -0.2372, -0.2365, -0.2417, -0.2365, -0.2173, -0.1922, -0.2038, -0.2063, -0.1876, -0.1758, -0.1741, -0.1211, -0.0781, -0.0793, -0.0772, -0.0932, -0.0852, -0.081, -0.0887, -0.0856, -0.0927, -0.0927, -0.0653, -0.0503, -0.0402, -0.0398, -0.0134, 0.0188, 0.0356, 0.0311, -0.0205, -0.031, -0.0296, 0.0032, 0.0157, -0.0135, 0.0, -0.0108, -0.0033, -0.0033, -0.006, 0.009, 0.0488, 0.0523, 0.0742, 0.0709, 0.096, 0.0526, 0.0731, 0.0941, 0.0941, 0.0901, 0.0807, 0.1049, 0.095, 0.0909, 0.0958, 0.117, 0.1165, 0.1389, 0.1266, 0.1459, 0.1077, 0.128, 0.1479, 0.1538, 0.1701, 0.1825, 0.2032, 0.1678, 0.1476, 0.1837, 0.1647, 0.1994, 0.2078, 0.2126, 0.2083, 0.2438, 0.2289, 0.2375, 0.218, 0.2157, 0.1726, 0.1577, 0.1515, 0.1689, 0.1699, 0.1849, 0.1938, 0.2272, 0.1824, 0.1824, 0.1858, 0.2023, 0.2171, 0.3315, 0.3743, 0.3386, 0.4694, 0.4299, 0.4308, 0.4476, 0.4314, 0.3764, 0.3731, 0.3715, 0.3493, 0.3499, 0.3514, 0.3386, 0.3324, 0.306, 0.314, 0.3278, 0.3469, 0.3477, 0.3362, 0.3399, 0.3761, 0.3742, 0.2929, 0.4207, 0.3706, 0.3993, 0.4105, 0.3913, 0.391, 0.3648, 0.3554, 0.3712, 0.4104, 0.442, 0.4855, 0.5373, 0.4994, 0.4722, 0.444, 0.4017, 0.4298, 0.4163, 0.3917, 0.4274, 0.4018, 0.4148, 0.3541, 0.364, 0.3647, 0.3562, 0.4116, 0.3813, 0.355, 0.5053, 0.5335, 0.5835, 0.5835, 0.6049, 0.5377, 0.5197, 0.5159, 0.5176, 0.5543, 0.5975, 0.6182]}
{"unit_id": "orig:1934486017916150007", "ticker_idx": 1, "ts": "2025-06-16T05:39:42+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AVGO and MRVL Continue to Dominate ASICs - JPMorgan", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPMorgan's bullish AVGO/MRVL thesis on ASIC TAM raised to $30B; SK Hynix also benefits as ASIC HBM mix rises to 20-30% with improving ASP and margins.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250616_AVGO and MRVL Continue to Dominate ASICs - JPMorgan\n\n[Key Takeaways]\n\n(1) JPMorgan has raised its ASIC TAM (Total Addressable Market) forecast from the previous $25bn to $30bn.\n\n(2) This upward revision is due to the expectation that the ASIC market growth rate will accelerate as hyperscalers advance their next-generation ASIC development roadmaps.\n\n(3) Both Broadcom and Marvell Technology are presumed to have secured next-generation 2nm ASIC projects, and they are expected to maintain their overwhelming dominance in the market.\n\n[Implication]\n\n(1) Earlier this year, SK Hynix's ASIC revenue mix was communicated as 'double-digit, early 10%' levels.\n\n(2) According to JPMorgan, the shipment mix accounted for by ASICs on a unit basis has already reached 40%.\n\n(3) Currently, 1) the perception that ASICs primarily sell middle-to-low end products with lower ASPs compared to HBM integrated into AI GPUs is receding, and 2) expectations for Q (Quantity) growth are increasingly rising.\n\n(4) In fact, the memory integrated into ASICs is also being upgraded to high-end products [HBM3E~4] comparable to the level integrated into GPU systems, and in some cases, the simple profit margin is estimated to be equivalent to or even higher than that of NVIDIA-bound products.\n\n(5) Recently, SK Hynix has also started communicating its ASIC mix within HBM revenue at 20-30% from the previous double-digit level.\n\n(6) As CSPs (Cloud Service Providers) increase their adoption rate of ASICs, the growth in SK Hynix's ASIC-related revenue mix is expected to 1) alleviate concerns about P (Price) [higher-than-expected ASP and excellent margins], 2) drive company-wide top-line growth due to strong Q growth, and 3) be used as a card to sustain a tight HBM supply-demand balance [used as a negotiation tool with NVIDIA].\n\n(7) In conclusion, it is judged that the time when the performance contribution of ASICs can be visibly observed is being brought forward.\n\n(8) An external environment that can stimulate multiples due to strong annual EPS growth or expectations of additional EPS upside seems to be unfolding.\n\n[Contents]\n\n(1) Broadcom's Google TPU v6 [3nm] ramp-up has begun, and the TPU v7 tape-out was recently completed.\n\n(2) Furthermore, revenue recognition from various clients, including Meta's MTIA [3nm], OpenAI, and Softbank/ARM, is expected to fully materialize from 2026.\n\n(3) The cycle for new ASIC design and release is becoming shorter, and the complexity of projects is increasing, with the number of transistors per chip/chiplet rising to over 150-200 billion.\n\n(4) We estimate that Broadcom will record over $15bn in revenue this year from 3nm projects alone, and its total AI revenue [ASIC+Networking] is projected to reach $19-20bn.\n\n(5) In addition, recent channel checks indicate that Broadcom has already successfully secured 2nm projects, including TPU v8 and the next-generation MTIA.\n\n(6) Accordingly, full-scale mass production of new projects such as OpenAI's and Softbank/ARM's 2-3nm ASICs and Meta's next-generation MTIA is expected to begin contributing to revenue from 2H26.\n\n(7) In Marvell Technology's case, it has also reportedly succeeded in securing 2nm projects targeting ramp-up from 2027, including Amazon Trainium 4 and Microsoft Maia Gen 3.\n\n(8) The mass production of its first two ASIC projects, Trainium 2 [5nm] and Axion ARM CPU, is imminent, and Trainium/Maia [3nm] mass production is scheduled for 2026.\n\n(9) Meanwhile, investors often tend to roughly consider high-speed I/O SERDES circuit technology as the sole differentiator between Broadcom/Marvell Technology and other ASIC/COT competitors.\n\n(10) We agree that SERDES technology is a core competency, but we believe it is not the only one.\n\n(11) This is because, in addition to SERDES IP, factors such as 1) high-performance memory architecture interfaces, 2) advanced cell libraries, 3) advanced packaging integration [2.5D, 3D SOIC, organic interposer], and 4) HBM4 logic die design capabilities broadly contribute.\n\n(12) In fact, the only large CSPs or OEMs capable of fully undertaking cutting-edge chip design [including manufacturing → testing → packaging] internally are essentially Apple, Amazon [limited to CPU, DPU], and Cisco [switch/routing chips].\n\n(13) Although hyperscalers are continuously increasing ASIC adoption in pursuit of 1) performance differentiation and 2) lower power consumption, the current bottleneck is their lack of capability to 100% internalize large-scale and complex SoC designs and immediately reflect them in current AI server operations.\n\n(14) In this environment, all hyperscalers [Google/Microsoft/Meta/Amazon] and large OEMs [Nokia/Ericsson/Cisco/Samsung Electronics/Juniper/ZTE] developing ASICs have significant and minor relationships with Broadcom/Marvell Technology in the areas of 1) design IP, 2) full-chip design expertise [integrating over 100 billion transistors on a single silicon], and 3) advanced packaging integration.\n\n(15) These factors are the driving force behind Broadcom/Marvell Technology's dominance of the entire market, with their ASIC market share nearing 70-75% [Broadcom 55-60%, Marvell Technology 15%].\n\n(16) As of this year, ASIC XPUs are estimated to already account for a '40% mix' on a unit basis in the XPU/GPU market.\n\n(17) The ASIC market growth has accelerated over the past few years due to the slowdown of Moore's Law and the increasing complexity of AI computing workloads, and the adoption rate of ASICs by hyperscalers is also structurally expected to rise in the future.\n\n(18) Accordingly, we are raising our ASIC TAM from the previous $25bn to $30bn [CAGR +30%] and maintaining a positive opinion on the related value chain.\n\n$AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04586764182521108, "ret_signed_m1d": -0.04586764182521108, "alpha_spy_m1d": -0.0364431121733233, "alpha_c_m1d": -0.02160600448809946, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00610621991647331, "ret_signed_p1d": -0.00610621991647331, "alpha_spy_p1d": 0.0024388126201055904, "alpha_c_p1d": 0.0006900698042445752, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.005112108324812681, "ret_signed_p1w": 0.005112108324812681, "alpha_spy_p1w": 0.006366103629449449, "alpha_c_p1w": 0.010655443635985251, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.029100713031454895, "ret_signed_p1m": 0.029100713031454895, "alpha_spy_p1m": -0.006240197607602305, "alpha_c_p1m": -0.07500736184128831, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": -0.053613912304387834, "ret_signed_p3m": -0.053613912304387834, "alpha_spy_p3m": -0.14801593262788015, "alpha_c_p3m": -0.20757388168952773, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.2642953920412845, "ret_signed_p6m": 0.2642953920412845, "alpha_spy_p6m": 0.12445586287646981, "alpha_c_p6m": -0.13672208132617003, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.0098, -0.0001, -0.0016, 0.0321, 0.0166, -0.053, -0.08, -0.12, -0.1267, -0.1111, -0.1031, -0.2108, -0.2989, -0.2766, -0.2904, -0.1353, -0.2501, -0.2418, -0.2579, -0.2428, -0.2626, -0.2658, -0.2658, -0.2988, -0.2812, -0.2364, -0.1859, -0.1633, -0.1664, -0.1667, -0.1711, -0.1345, -0.1149, -0.1199, -0.1306, -0.2004, -0.1816, -0.1529, -0.0841, -0.0702, -0.0642, -0.0741, -0.0946, -0.1116, -0.1278, -0.1464, -0.1217, -0.1382, -0.1382, -0.0937, -0.0828, -0.095, -0.1453, -0.1271, -0.1145, -0.0585, -0.0747, -0.0294, -0.0182, -0.0224, -0.031, -0.0111, -0.0459, 0.0, -0.0061, 0.0643, 0.0643, 0.0439, 0.0051, 0.068, 0.0782, 0.1356, 0.0957, 0.0991, 0.0826, 0.0544, 0.0676, 0.0676, 0.016, 0.0217, 0.0261, 0.0417, 0.0334, 0.0305, 0.0291, 0.0069, 0.0234, 0.0609, 0.0383, 0.0231, 0.0413, 0.0523, 0.0547, 0.0788, 0.085, 0.1617, 0.1422, 0.0581, 0.0877, 0.0891, 0.0705, 0.078, 0.0992, 0.0983, 0.1058, 0.1273, 0.1233, 0.0828, 0.0906, 0.0243, 0.0122, 0.012, 0.0375, 0.0368, 0.0554, 0.0629, 0.0976, -0.1065, -0.1065, -0.0819, -0.1144, -0.089, -0.0999, -0.062, -0.0501, -0.0464, -0.0536, -0.0428, -0.0417, -0.0214, 0.0088, 0.055, 0.0554, 0.0734, 0.0605, 0.1382, 0.1911, 0.182, 0.1709, 0.1948, 0.1923, 0.2251, 0.2254, 0.2637, 0.236, 0.3146, 0.2888, 0.2175, 0.2713, 0.2262, 0.2642, 0.2548, 0.2508, 0.2208, 0.1983, 0.1527, 0.1771, 0.1965, 0.2616, 0.258, 0.2821, 0.2596, 0.3331, 0.2852, 0.2457, 0.3212, 0.3273, 0.293, 0.3259, 0.2704, 0.2704, 0.2447, 0.2295, 0.1868, 0.119, 0.1565, 0.0905, 0.1015, 0.1916, 0.1865, 0.2475, 0.2475, 0.2714, 0.2956, 0.321, 0.425, 0.3964, 0.4067, 0.3084, 0.2643]}
{"unit_id": "orig:1934486017916150007", "ticker_idx": 2, "ts": "2025-06-16T05:39:42+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the time when the performance contribution of ASICs can be visibly observed is being brought forward... An external environment that can stimulate multiples due to strong annual EPS growth or expectations of additional EPS upside seems to be unfolding", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPMorgan's bullish AVGO/MRVL thesis on ASIC TAM raised to $30B; SK Hynix also benefits as ASIC HBM mix rises to 20-30% with improving ASP and margins.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250616_AVGO and MRVL Continue to Dominate ASICs - JPMorgan\n\n[Key Takeaways]\n\n(1) JPMorgan has raised its ASIC TAM (Total Addressable Market) forecast from the previous $25bn to $30bn.\n\n(2) This upward revision is due to the expectation that the ASIC market growth rate will accelerate as hyperscalers advance their next-generation ASIC development roadmaps.\n\n(3) Both Broadcom and Marvell Technology are presumed to have secured next-generation 2nm ASIC projects, and they are expected to maintain their overwhelming dominance in the market.\n\n[Implication]\n\n(1) Earlier this year, SK Hynix's ASIC revenue mix was communicated as 'double-digit, early 10%' levels.\n\n(2) According to JPMorgan, the shipment mix accounted for by ASICs on a unit basis has already reached 40%.\n\n(3) Currently, 1) the perception that ASICs primarily sell middle-to-low end products with lower ASPs compared to HBM integrated into AI GPUs is receding, and 2) expectations for Q (Quantity) growth are increasingly rising.\n\n(4) In fact, the memory integrated into ASICs is also being upgraded to high-end products [HBM3E~4] comparable to the level integrated into GPU systems, and in some cases, the simple profit margin is estimated to be equivalent to or even higher than that of NVIDIA-bound products.\n\n(5) Recently, SK Hynix has also started communicating its ASIC mix within HBM revenue at 20-30% from the previous double-digit level.\n\n(6) As CSPs (Cloud Service Providers) increase their adoption rate of ASICs, the growth in SK Hynix's ASIC-related revenue mix is expected to 1) alleviate concerns about P (Price) [higher-than-expected ASP and excellent margins], 2) drive company-wide top-line growth due to strong Q growth, and 3) be used as a card to sustain a tight HBM supply-demand balance [used as a negotiation tool with NVIDIA].\n\n(7) In conclusion, it is judged that the time when the performance contribution of ASICs can be visibly observed is being brought forward.\n\n(8) An external environment that can stimulate multiples due to strong annual EPS growth or expectations of additional EPS upside seems to be unfolding.\n\n[Contents]\n\n(1) Broadcom's Google TPU v6 [3nm] ramp-up has begun, and the TPU v7 tape-out was recently completed.\n\n(2) Furthermore, revenue recognition from various clients, including Meta's MTIA [3nm], OpenAI, and Softbank/ARM, is expected to fully materialize from 2026.\n\n(3) The cycle for new ASIC design and release is becoming shorter, and the complexity of projects is increasing, with the number of transistors per chip/chiplet rising to over 150-200 billion.\n\n(4) We estimate that Broadcom will record over $15bn in revenue this year from 3nm projects alone, and its total AI revenue [ASIC+Networking] is projected to reach $19-20bn.\n\n(5) In addition, recent channel checks indicate that Broadcom has already successfully secured 2nm projects, including TPU v8 and the next-generation MTIA.\n\n(6) Accordingly, full-scale mass production of new projects such as OpenAI's and Softbank/ARM's 2-3nm ASICs and Meta's next-generation MTIA is expected to begin contributing to revenue from 2H26.\n\n(7) In Marvell Technology's case, it has also reportedly succeeded in securing 2nm projects targeting ramp-up from 2027, including Amazon Trainium 4 and Microsoft Maia Gen 3.\n\n(8) The mass production of its first two ASIC projects, Trainium 2 [5nm] and Axion ARM CPU, is imminent, and Trainium/Maia [3nm] mass production is scheduled for 2026.\n\n(9) Meanwhile, investors often tend to roughly consider high-speed I/O SERDES circuit technology as the sole differentiator between Broadcom/Marvell Technology and other ASIC/COT competitors.\n\n(10) We agree that SERDES technology is a core competency, but we believe it is not the only one.\n\n(11) This is because, in addition to SERDES IP, factors such as 1) high-performance memory architecture interfaces, 2) advanced cell libraries, 3) advanced packaging integration [2.5D, 3D SOIC, organic interposer], and 4) HBM4 logic die design capabilities broadly contribute.\n\n(12) In fact, the only large CSPs or OEMs capable of fully undertaking cutting-edge chip design [including manufacturing → testing → packaging] internally are essentially Apple, Amazon [limited to CPU, DPU], and Cisco [switch/routing chips].\n\n(13) Although hyperscalers are continuously increasing ASIC adoption in pursuit of 1) performance differentiation and 2) lower power consumption, the current bottleneck is their lack of capability to 100% internalize large-scale and complex SoC designs and immediately reflect them in current AI server operations.\n\n(14) In this environment, all hyperscalers [Google/Microsoft/Meta/Amazon] and large OEMs [Nokia/Ericsson/Cisco/Samsung Electronics/Juniper/ZTE] developing ASICs have significant and minor relationships with Broadcom/Marvell Technology in the areas of 1) design IP, 2) full-chip design expertise [integrating over 100 billion transistors on a single silicon], and 3) advanced packaging integration.\n\n(15) These factors are the driving force behind Broadcom/Marvell Technology's dominance of the entire market, with their ASIC market share nearing 70-75% [Broadcom 55-60%, Marvell Technology 15%].\n\n(16) As of this year, ASIC XPUs are estimated to already account for a '40% mix' on a unit basis in the XPU/GPU market.\n\n(17) The ASIC market growth has accelerated over the past few years due to the slowdown of Moore's Law and the increasing complexity of AI computing workloads, and the adoption rate of ASICs by hyperscalers is also structurally expected to rise in the future.\n\n(18) Accordingly, we are raising our ASIC TAM from the previous $25bn to $30bn [CAGR +30%] and maintaining a positive opinion on the related value chain.\n\n$AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05040316776889919, "ret_signed_m1d": -0.05040316776889919, "alpha_spy_m1d": -0.04097863811701141, "alpha_c_m1d": -0.026141530431787574, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004032278728435923, "ret_signed_p1d": 0.004032278728435923, "alpha_spy_p1d": 0.012577311265014823, "alpha_c_p1d": 0.010828568449153808, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.046370952307773106, "ret_signed_p1w": 0.046370952307773106, "alpha_spy_p1w": 0.047624947612409874, "alpha_c_p1w": 0.051914287618945676, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.20362919004367397, "ret_signed_p1m": 0.20362919004367397, "alpha_spy_p1m": 0.16828827940461677, "alpha_c_p1m": 0.09952111517093076, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.2396913036108994, "ret_signed_p3m": 0.2396913036108994, "alpha_spy_p3m": 0.1452892832874071, "alpha_c_p3m": 0.08573133422575951, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 1.287191450865563, "ret_signed_p6m": 1.287191450865563, "alpha_spy_p6m": 1.1473519217007484, "alpha_c_p6m": 0.8861739774981086, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.1729, -0.1548, -0.1326, -0.1487, -0.1628, -0.1387, -0.1668, -0.1978, -0.2324, -0.2071, -0.2035, -0.2167, -0.2666, -0.3367, -0.3178, -0.3359, -0.2626, -0.2723, -0.2747, -0.2731, -0.2997, -0.2956, -0.2956, -0.2892, -0.3005, -0.2715, -0.2823, -0.2578, -0.2675, -0.2723, -0.2856, -0.2856, -0.2514, -0.2514, -0.2514, -0.232, -0.234, -0.2348, -0.2151, -0.201, -0.1709, -0.193, -0.1769, -0.1974, -0.187, -0.193, -0.2075, -0.195, -0.1829, -0.1849, -0.1628, -0.1452, -0.1754, -0.1633, -0.1633, -0.123, -0.0948, -0.0948, -0.0766, -0.0706, -0.0323, -0.0504, -0.0504, 0.0, 0.004, -0.006, -0.0081, 0.0363, 0.0464, 0.123, 0.1532, 0.1815, 0.1452, 0.1774, 0.1512, 0.125, 0.123, 0.0907, 0.0927, 0.1371, 0.1331, 0.1976, 0.1875, 0.2097, 0.2036, 0.1935, 0.0867, 0.0847, 0.0988, 0.0827, 0.0847, 0.0867, 0.0726, 0.0565, 0.0585, 0.0625, 0.1028, 0.0403, 0.0403, 0.0625, 0.0423, 0.0565, 0.0343, 0.0766, 0.0847, 0.121, 0.1149, 0.1149, 0.0786, 0.0605, 0.0302, -0.0121, 0.0121, 0.0464, 0.0544, 0.0484, 0.0842, 0.0862, 0.0337, 0.0519, 0.06, 0.0721, 0.1044, 0.1185, 0.163, 0.2276, 0.2397, 0.3265, 0.3366, 0.4053, 0.3467, 0.4355, 0.4355, 0.4174, 0.4577, 0.4436, 0.4396, 0.3588, 0.4093, 0.4032, 0.4537, 0.5971, 0.5971, 0.5971, 0.5971, 0.5971, 0.5971, 0.7283, 0.6758, 0.6617, 0.7061, 0.8272, 0.8797, 0.9605, 0.9342, 0.9443, 0.9322, 1.0594, 1.1604, 1.1038, 1.2532, 1.2936, 1.2573, 1.5036, 1.3663, 1.338, 1.3946, 1.3421, 1.4471, 1.4996, 1.4915, 1.4713, 1.2613, 1.4471, 1.3017, 1.2694, 1.3057, 1.1038, 1.0998, 1.0958, 1.116, 1.1983, 1.1417, 1.174, 1.2549, 1.2306, 1.1902, 1.1983, 1.3316, 1.2872]}
{"unit_id": "quote:1930425919904330233", "ticker_idx": 0, "ts": "2025-06-05T00:46:19+00:00", "symbol": "3231.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Brokers Reiterate Positive Outlook... Most brokers see upside potential for 2Q earnings revisions... AI server profitability has already surpassed that of its NB (notebook) business", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares broker reiteration of positive outlook on Wistron: May revenue beat, 2Q upside revisions expected, AI server margins ahead of notebook business.", "resolved_tickers": ["3231.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "🔹 Wistron (3231): Brokers Reiterate Positive Outlook\n• May revenue exceeded NT$200 billion, bringing QTD performance to 85% of U.S. broker estimates, significantly ahead of expectations.\n• Most brokers see upside potential for 2Q earnings revisions.\n• However, some U.S. investment banks noted that peer companies may not replicate the same month-on-month growth because:\n• Wistron is shipping L10-level compute trays,\n• while peers are delivering L11-level racks,\n• which require longer assembly and testing time.\n• Regarding Wistron itself, brokers observe that U.S. customer pull-ins remain smooth,\nwith an original forecast of 500–1,000 GB200 units to be shipped during May–June, which is tracking in line with expectations.\n• New projects with Oracle and Cisco are also in the pipeline.\n• Brokers emphasize that AI infrastructure build-out is not a linear process,\nbut for Wistron, AI server profitability has already surpassed that of its NB (notebook) business.", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/5 外電綜合整理\n\n-3231緯創: 券商重申正向\n5月營收一舉超過2000億，QTD達成美系券商預估數85%，表現大幅優於預期，券商們都認為2Q還有上修空間。不過美系大行認為同業未必有如此月增幅，原因在於緯創出是L10的compute tray，而同業出L11的 rack，後者的組裝跟測試需要花更多時間。", "parent_author": null, "ret_m1d": -0.008695652173912993, "ret_signed_m1d": -0.008695652173912993, "alpha_spy_m1d": -0.013551890307483605, "alpha_c_m1d": -0.01153316242033342, "bench_spy_m1d": 0.004856238133570612, "bench_c_m1d": 0.002837510246420427, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04347826086956519, "ret_signed_p1d": -0.04347826086956519, "alpha_spy_p1d": -0.05374724072852799, "alpha_c_p1d": -0.05292222233308119, "bench_spy_p1d": 0.010268979858962801, "bench_c_p1d": 0.009443961463516004, "ret_p1w": 0.01304347826086949, "ret_signed_p1w": 0.01304347826086949, "alpha_spy_p1w": -0.004998855360847765, "alpha_c_p1w": -0.014229660809497391, "bench_spy_p1w": 0.018042333621717255, "bench_c_p1w": 0.02727313907036688, "ret_p1m": 0.034782608695652195, "ret_signed_p1m": 0.034782608695652195, "alpha_spy_p1m": -0.022782097859046058, "alpha_c_p1m": -0.05523511461392161, "bench_spy_p1m": 0.05756470655469825, "bench_c_p1m": 0.0900177233095738, "ret_p3m": -0.04347826086956519, "ret_signed_p3m": -0.04347826086956519, "alpha_spy_p3m": -0.12629226198361332, "alpha_c_p3m": -0.145543038457985, "bench_spy_p3m": 0.08281400111404813, "bench_c_p3m": 0.10206477758841981, "ret_p6m": 0.25652173913043486, "ret_signed_p6m": 0.25652173913043486, "alpha_spy_p6m": 0.09757969139083777, "alpha_c_p6m": 0.040823582995427765, "bench_spy_p6m": 0.15894204773959708, "bench_c_p6m": 0.2156981561350071, "price_path": [-0.1087, -0.1213, -0.1213, -0.1255, -0.1213, -0.1087, -0.0961, -0.1339, -0.1171, -0.1129, -0.1297, -0.1297, -0.1129, -0.1213, -0.155, -0.1987, -0.1466, -0.1508, -0.1508, -0.1508, -0.2357, -0.3114, -0.3719, -0.3097, -0.2407, -0.1819, -0.1381, -0.1617, -0.155, -0.1171, -0.1339, -0.1693, -0.1381, -0.1508, -0.1381, -0.1381, -0.1255, -0.1466, -0.1466, -0.1129, -0.1297, -0.0835, -0.1045, -0.1003, -0.1003, -0.0877, -0.0877, -0.0372, -0.0625, -0.0667, -0.0835, -0.0793, -0.0499, -0.0499, -0.0499, -0.0246, -0.033, -0.0288, -0.0204, -0.0204, -0.033, -0.0478, -0.0087, 0.0, -0.0435, -0.013, -0.0043, 0.0, 0.013, 0.0304, 0.0348, 0.0261, 0.0087, 0.0043, 0.0304, 0.0304, 0.0435, 0.0913, 0.087, 0.0522, 0.0652, 0.0435, 0.0435, 0.0522, 0.0348, 0.0261, 0.0304, 0.0478, 0.0435, 0.0391, 0.0174, 0.0435, 0.0261, 0.0261, 0.0348, 0.0304, -0.0043, 0.0087, 0.0174, 0.013, 0.0435, 0.0478, 0.0478, 0.0696, 0.0696, 0.0391, 0.0783, 0.0652, 0.0739, 0.0826, 0.1087, 0.0957, 0.0217, 0.013, 0.013, 0.0087, 0.0043, -0.013, -0.0304, -0.0261, -0.013, -0.0174, 0.0, 0.0, -0.0174, -0.0478, -0.0435, -0.0435, -0.0174, 0.0043, -0.0043, 0.0087, 0.0435, 0.0348, 0.0435, 0.0391, 0.0261, 0.0348, 0.0435, 0.0522, 0.0565, 0.0652, 0.0565, 0.0913, 0.113, 0.113, 0.2217, 0.2696, 0.3, 0.3348, 0.3348, 0.3565, 0.3087, 0.313, 0.313, 0.2652, 0.2, 0.2, 0.213, 0.187, 0.2348, 0.2391, 0.2478, 0.2435, 0.2435, 0.2739, 0.287, 0.3261, 0.3478, 0.3087, 0.2478, 0.2087, 0.1957, 0.2348, 0.1957, 0.2043, 0.2565, 0.2087, 0.2522, 0.2217, 0.2, 0.1913, 0.1913, 0.2435, 0.2087, 0.2087, 0.213, 0.2391, 0.2652, 0.2565]}
{"unit_id": "orig:1923353151337570341", "ticker_idx": 0, "ts": "2025-05-16T12:21:40+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix, Samsung Electronics, both rated Buy", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman reiterates Buy on SK Hynix (PT 290,000 KRW) and Samsung Electronics (PT 74,000 KRW) on improving NAND industry conditions and stronger 2Q25 outlook vs Kioxia.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) APAC Tech: Kioxia’s 1Q25 Results and Outlook & Korean Memory Companies\n\nOn May 15, Japanese NAND manufacturer Kioxia announced its results for the first quarter of 2025 (the fourth quarter of its fiscal year 2024, March quarter). Here, we provide implications for our covered Korean memory companies (SK Hynix, Samsung Electronics, both rated Buy). In summary, Kioxia’s 1Q25 results were similar to those of the Korean players, while its 2Q25 guidance (due to company-specific reasons) was somewhat weaker than that of its Korean peers. Kioxia indicated that tariff impacts will be limited in 2Q25; however, we expect Korean companies to see a certain level of front-loaded demand. Kioxia’s future outlook suggests improving NAND industry conditions.\n\n1Q25 Results Similar to Korean Players (see Table 1):\nKioxia’s NAND pricing, bit shipments, and margin directions were similar to those of Samsung Electronics and SK Hynix. For all three, sharp declines in ASP (average selling price) and bit shipments quarter-on-quarter, along with customer inventory adjustments, led to deteriorating operating profit margins (OPM).\n2Q25 Guidance Weaker than Korean Peers (see Table 1):\nKioxia guided for 2Q25 sales to change by -8% to +1% quarter-on-quarter (in USD terms, excluding FX impact), which is weaker than Samsung Electronics’ and SK Hynix’s NAND revenue growth forecasts (+14% and +21% qoq, respectively, in USD terms). We believe Kioxia’s weak guidance is due to “company-specific factors,” as Kioxia noted that customers are transitioning from BiCS Flash Gen 5 to Gen 8, resulting in some demand softness.\n\nTariff Impact Likely to Be Limited:\nThe company stated there was no tariff impact in 1Q, and while there may be some front-loaded demand in 2Q due to partial tariff issues, the tariff impact is expected to be minor, with a margin impact of about 1%. We believe there was no special tariff-related front-loaded demand for Korean players in 1Q (actual NAND shipments were similar to guidance), but we expect to see some front-loaded demand in 2Q.\nMaintaining Strict Capex Plans:\nThe company continues to maintain a strict capex plan and will flexibly adjust production according to market conditions. This year, there will be no large-scale infrastructure capex, and equipment investment will focus on ramping up BiCS Flash Gen 8 mass production and responding to demand for large-capacity SSDs for AI.\n\nOutlook for NAND Industry Improvement:\n\n1. Kioxia stated that NAND prices have bottomed out and some segments are showing signs of improvement.\n\n2. Smartphone and PC inventories have begun to stabilize, but the pace of inventory normalization in 2025 is difficult to predict due to macro uncertainty.\n\n3. Enterprise and data center demand remains solid thanks to expanding AI adoption.\n\nSK Hynix 12M target price: 290,000 KRW (based on 2.0x target PBR on average 2025 BVPS)\nKey downside risks: severe deterioration in memory supply-demand / delay in technology migration\n\nSamsung Electronics 12M target price: 74,000 KRW / Preferred stock target price: 61,000 KRW (18% discount to common stock)\nTarget price based on 2025E SOTP using EV/EBITDA\n\nKey downside risks: severe deterioration in memory supply-demand / severe deterioration in smartphone profitability / decline in mobile OLED market share", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019560048974488042, "ret_signed_m1d": -0.019560048974488042, "alpha_spy_m1d": -0.013265941992019914, "alpha_c_m1d": -0.022441294584930382, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": 0.00288124561044234, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02493889526388049, "ret_signed_p1d": -0.02493889526388049, "alpha_spy_p1d": -0.026032813363918494, "alpha_c_p1d": -0.02327503275866516, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.00166386250521533, "ret_p1w": -0.022004930192753402, "ret_signed_p1w": -0.022004930192753402, "alpha_spy_p1w": 0.0033905099277240502, "alpha_c_p1w": 0.014436932108353506, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03644186230110691, "ret_p1m": 0.2149041982118578, "ret_signed_p1m": 0.2149041982118578, "alpha_spy_p1m": 0.20063287348420755, "alpha_c_p1m": 0.14607860363739, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.0688255945744678, "ret_p3m": 0.36186832337704744, "ret_signed_p3m": 0.36186832337704744, "alpha_spy_p3m": 0.2733517621385688, "alpha_c_p3m": 0.13924026537671486, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.22262805800033258, "ret_p6m": 1.9729651079447188, "ret_signed_p6m": 1.9729651079447188, "alpha_spy_p6m": 1.819566496926681, "alpha_c_p6m": 1.5169550374366274, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.4560100705080914, "price_path": [0.0203, 0.0616, 0.03, 0.0179, -0.004, -0.0259, -0.0137, -0.0259, -0.0699, -0.0699, -0.09, -0.0557, -0.0592, -0.0592, -0.0812, -0.0817, -0.0274, -0.0235, 0.0, 0.0073, -0.0073, 0.0049, 0.0269, 0.0538, 0.0342, 0.0171, 0.0465, 0.0122, -0.0254, -0.0675, -0.0367, -0.0323, -0.0484, -0.109, -0.1941, -0.1711, -0.1932, -0.1042, -0.1159, -0.1188, -0.1169, -0.1491, -0.1443, -0.1443, -0.1364, -0.1501, -0.1149, -0.1281, -0.0983, -0.11, -0.1159, -0.132, -0.132, -0.0905, -0.0905, -0.0905, -0.067, -0.0694, -0.0704, -0.0465, -0.0293, 0.0073, -0.0196, 0.0, -0.0249, -0.0122, -0.0196, -0.0372, -0.022, -0.0073, -0.0098, 0.0171, 0.0385, 0.0018, 0.0165, 0.0165, 0.0655, 0.0998, 0.0998, 0.1218, 0.1292, 0.1757, 0.1537, 0.1537, 0.2149, 0.2198, 0.2076, 0.2051, 0.259, 0.2712, 0.3643, 0.4011, 0.4354, 0.3913, 0.4305, 0.3986, 0.3668, 0.3643, 0.3251, 0.3276, 0.3815, 0.3766, 0.4549, 0.4427, 0.4696, 0.4623, 0.45, 0.3202, 0.3178, 0.3349, 0.3153, 0.3178, 0.3202, 0.3031, 0.2835, 0.2859, 0.2908, 0.3398, 0.2639, 0.2639, 0.2908, 0.2663, 0.2835, 0.2565, 0.308, 0.3178, 0.3619, 0.3545, 0.3545, 0.3104, 0.2884, 0.2516, 0.2002, 0.2296, 0.2712, 0.281, 0.2737, 0.3172, 0.3197, 0.2559, 0.278, 0.2878, 0.3025, 0.3418, 0.3589, 0.4129, 0.4914, 0.5061, 0.6116, 0.6238, 0.7072, 0.6361, 0.744, 0.744, 0.722, 0.771, 0.7539, 0.7489, 0.6508, 0.7122, 0.7048, 0.7661, 0.9403, 0.9403, 0.9403, 0.9403, 0.9403, 0.9403, 1.0997, 1.0359, 1.0188, 1.0727, 1.2199, 1.2837, 1.3818, 1.3499, 1.3622, 1.3475, 1.502, 1.6246, 1.556, 1.7375, 1.7865, 1.7424, 2.0416, 1.8748, 1.8405, 1.9092, 1.8454, 1.973]}
{"unit_id": "orig:1923353151337570341", "ticker_idx": 1, "ts": "2025-05-16T12:21:40+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix, Samsung Electronics, both rated Buy", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman reiterates Buy on SK Hynix (PT 290,000 KRW) and Samsung Electronics (PT 74,000 KRW) on improving NAND industry conditions and stronger 2Q25 outlook vs Kioxia.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) APAC Tech: Kioxia’s 1Q25 Results and Outlook & Korean Memory Companies\n\nOn May 15, Japanese NAND manufacturer Kioxia announced its results for the first quarter of 2025 (the fourth quarter of its fiscal year 2024, March quarter). Here, we provide implications for our covered Korean memory companies (SK Hynix, Samsung Electronics, both rated Buy). In summary, Kioxia’s 1Q25 results were similar to those of the Korean players, while its 2Q25 guidance (due to company-specific reasons) was somewhat weaker than that of its Korean peers. Kioxia indicated that tariff impacts will be limited in 2Q25; however, we expect Korean companies to see a certain level of front-loaded demand. Kioxia’s future outlook suggests improving NAND industry conditions.\n\n1Q25 Results Similar to Korean Players (see Table 1):\nKioxia’s NAND pricing, bit shipments, and margin directions were similar to those of Samsung Electronics and SK Hynix. For all three, sharp declines in ASP (average selling price) and bit shipments quarter-on-quarter, along with customer inventory adjustments, led to deteriorating operating profit margins (OPM).\n2Q25 Guidance Weaker than Korean Peers (see Table 1):\nKioxia guided for 2Q25 sales to change by -8% to +1% quarter-on-quarter (in USD terms, excluding FX impact), which is weaker than Samsung Electronics’ and SK Hynix’s NAND revenue growth forecasts (+14% and +21% qoq, respectively, in USD terms). We believe Kioxia’s weak guidance is due to “company-specific factors,” as Kioxia noted that customers are transitioning from BiCS Flash Gen 5 to Gen 8, resulting in some demand softness.\n\nTariff Impact Likely to Be Limited:\nThe company stated there was no tariff impact in 1Q, and while there may be some front-loaded demand in 2Q due to partial tariff issues, the tariff impact is expected to be minor, with a margin impact of about 1%. We believe there was no special tariff-related front-loaded demand for Korean players in 1Q (actual NAND shipments were similar to guidance), but we expect to see some front-loaded demand in 2Q.\nMaintaining Strict Capex Plans:\nThe company continues to maintain a strict capex plan and will flexibly adjust production according to market conditions. This year, there will be no large-scale infrastructure capex, and equipment investment will focus on ramping up BiCS Flash Gen 8 mass production and responding to demand for large-capacity SSDs for AI.\n\nOutlook for NAND Industry Improvement:\n\n1. Kioxia stated that NAND prices have bottomed out and some segments are showing signs of improvement.\n\n2. Smartphone and PC inventories have begun to stabilize, but the pace of inventory normalization in 2025 is difficult to predict due to macro uncertainty.\n\n3. Enterprise and data center demand remains solid thanks to expanding AI adoption.\n\nSK Hynix 12M target price: 290,000 KRW (based on 2.0x target PBR on average 2025 BVPS)\nKey downside risks: severe deterioration in memory supply-demand / delay in technology migration\n\nSamsung Electronics 12M target price: 74,000 KRW / Preferred stock target price: 61,000 KRW (18% discount to common stock)\nTarget price based on 2025E SOTP using EV/EBITDA\n\nKey downside risks: severe deterioration in memory supply-demand / severe deterioration in smartphone profitability / decline in mobile OLED market share", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008802920277104054, "ret_signed_m1d": 0.008802920277104054, "alpha_spy_m1d": 0.015097027259572182, "alpha_c_m1d": 0.010676138573544214, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": -0.00187321829644016, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017605490629326592, "ret_signed_p1d": -0.017605490629326592, "alpha_spy_p1d": -0.018699408729364597, "alpha_c_p1d": -0.016158110671899983, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.0014473799574266089, "ret_p1w": -0.04577458357014497, "ret_signed_p1w": -0.04577458357014497, "alpha_spy_p1w": -0.020379143449667514, "alpha_c_p1w": -0.011375561155943403, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03439902241420156, "ret_p1m": 0.007042308227692695, "ret_signed_p1m": 0.007042308227692695, "alpha_spy_p1m": -0.007229016499957552, "alpha_c_p1m": -0.027654761824216667, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.03469707005190936, "ret_p3m": 0.27360941426703933, "ret_signed_p3m": 0.27360941426703933, "alpha_spy_p3m": 0.1850928530285607, "alpha_c_p3m": 0.1294369381105236, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.14417247615651574, "ret_p6m": 0.789941073349784, "ret_signed_p6m": 0.789941073349784, "alpha_spy_p6m": 0.6365424623317462, "alpha_c_p6m": 0.528072838629364, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.26186823472042, "price_path": [-0.0042, 0.0273, 0.0221, 0.0186, 0.0028, 0.0011, -0.0094, -0.0147, -0.0462, -0.0462, -0.0462, -0.0549, -0.0497, -0.0602, -0.0602, -0.0619, -0.0392, -0.0427, -0.0427, 0.0081, 0.0081, 0.0238, 0.0536, 0.0799, 0.0589, 0.0466, 0.0746, 0.0816, 0.0599, 0.0176, 0.0352, 0.0352, 0.0141, -0.0123, -0.0634, -0.0581, -0.0669, -0.007, -0.0282, -0.0106, -0.0035, -0.037, -0.0299, -0.0264, -0.0246, -0.0317, -0.0194, -0.0194, -0.0194, -0.0176, -0.0176, -0.0229, -0.0229, -0.044, -0.044, -0.044, -0.0387, -0.0387, -0.0352, 0.0141, 0.0018, 0.0106, 0.0088, 0.0, -0.0176, -0.0158, -0.0194, -0.037, -0.0458, -0.037, -0.0511, -0.0158, -0.0123, -0.0106, 0.0, 0.0, 0.0176, 0.0405, 0.0405, 0.0528, 0.0423, 0.0546, 0.0475, 0.0264, 0.007, 0.0229, 0.0528, 0.0423, 0.0475, 0.0211, 0.0651, 0.0792, 0.0599, 0.077, 0.0593, 0.0664, 0.077, 0.1301, 0.1213, 0.0929, 0.0876, 0.0699, 0.0805, 0.1089, 0.1071, 0.1284, 0.1461, 0.1815, 0.1886, 0.201, 0.1691, 0.1762, 0.1691, 0.1673, 0.247, 0.2506, 0.286, 0.2648, 0.2205, 0.2346, 0.2382, 0.2187, 0.2488, 0.2718, 0.2577, 0.2594, 0.2736, 0.2683, 0.2683, 0.24, 0.24, 0.2488, 0.2506, 0.2648, 0.2665, 0.2453, 0.2506, 0.2329, 0.2346, 0.1974, 0.224, 0.2364, 0.2417, 0.2311, 0.2417, 0.2665, 0.286, 0.3002, 0.3356, 0.3551, 0.4065, 0.3852, 0.4224, 0.4224, 0.4791, 0.5003, 0.5127, 0.5251, 0.4755, 0.4981, 0.4928, 0.5302, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.6796, 0.6601, 0.6298, 0.6903, 0.7383, 0.7419, 0.7455, 0.7348, 0.7544, 0.717, 0.7579, 0.8149, 0.7704, 0.7882, 0.8522, 0.9127, 0.9768, 0.8664, 0.7899, 0.765, 0.7419, 0.7899]}
{"unit_id": "thread:1936958783395426624", "ticker_idx": 0, "ts": "2025-06-23T01:25:35+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hynix shares are still undervalued.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Author calls Hynix undervalued and cites Macquarie's KRW 360,000 PT implying 42% upside.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Hynix shares are still undervalued.\"\n\nMacquarie has set the most aggressive price target for Hynix at the moment — KRW 360,000.\n\n---\n\nHynix's current share price is KRW 252,000, which means Macquarie sees up to 42% upside from here.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009633960880562231, "ret_signed_m1d": -0.009633960880562231, "alpha_spy_m1d": 0.00014690926970717388, "alpha_c_m1d": -0.003334267520181955, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07321773991092995, "ret_signed_p1d": 0.07321773991092995, "alpha_spy_p1d": 0.06217053091784219, "alpha_c_p1d": 0.03648900821335066, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.12524094727529445, "ret_signed_p1w": 0.12524094727529445, "alpha_spy_p1w": 0.09574828792782775, "alpha_c_p1w": 0.060488397827961116, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": 0.034682017315795344, "ret_signed_p1m": 0.034682017315795344, "alpha_spy_p1m": -0.013155991413296597, "alpha_c_p1m": -0.05484913121998769, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.37192101427859003, "ret_signed_p3m": 0.37192101427859003, "alpha_spy_p3m": 0.2684302492599826, "alpha_c_p3m": 0.16174232271671962, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 1.046804114037171, "ret_signed_p6m": 1.046804114037171, "alpha_spy_p6m": 0.9125008132178507, "alpha_c_p6m": 0.7031114278058463, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [-0.1768, -0.2038, -0.2334, -0.2665, -0.2422, -0.2388, -0.2514, -0.2991, -0.3661, -0.348, -0.3653, -0.2953, -0.3045, -0.3068, -0.3053, -0.3307, -0.3268, -0.3268, -0.3207, -0.3315, -0.3038, -0.3141, -0.2907, -0.2999, -0.3045, -0.3172, -0.3172, -0.2845, -0.2845, -0.2845, -0.2661, -0.268, -0.2688, -0.2499, -0.2364, -0.2076, -0.2288, -0.2134, -0.233, -0.223, -0.2288, -0.2426, -0.2307, -0.2191, -0.2211, -0.1999, -0.183, -0.2119, -0.2004, -0.2004, -0.1618, -0.1349, -0.1349, -0.1175, -0.1118, -0.0751, -0.0925, -0.0925, -0.0443, -0.0405, -0.0501, -0.052, -0.0096, 0.0, 0.0732, 0.1021, 0.1291, 0.0944, 0.1252, 0.1002, 0.0751, 0.0732, 0.0424, 0.0443, 0.0867, 0.0829, 0.1445, 0.1349, 0.1561, 0.1503, 0.1407, 0.0385, 0.0366, 0.0501, 0.0347, 0.0366, 0.0385, 0.025, 0.0096, 0.0116, 0.0154, 0.0539, -0.0058, -0.0058, 0.0154, -0.0039, 0.0096, -0.0116, 0.0289, 0.0366, 0.0713, 0.0655, 0.0655, 0.0308, 0.0135, -0.0154, -0.0559, -0.0328, 0.0, 0.0077, 0.0019, 0.0362, 0.0381, -0.0121, 0.0053, 0.013, 0.0246, 0.0555, 0.069, 0.1114, 0.1732, 0.1848, 0.2677, 0.2774, 0.343, 0.287, 0.3719, 0.3719, 0.3546, 0.3931, 0.3796, 0.3758, 0.2986, 0.3468, 0.341, 0.3893, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.6517, 0.6015, 0.588, 0.6305, 0.7463, 0.7964, 0.8736, 0.8485, 0.8582, 0.8466, 0.9682, 1.0646, 1.0106, 1.1534, 1.192, 1.1573, 1.3927, 1.2615, 1.2344, 1.2885, 1.2383, 1.3386, 1.3888, 1.3811, 1.3618, 1.1611, 1.3386, 1.1997, 1.1688, 1.2036, 1.0106, 1.0067, 1.0029, 1.0222, 1.1009, 1.0468, 1.0777, 1.1549, 1.1318, 1.0931, 1.1009, 1.2283, 1.1858, 1.2669, 1.182, 1.2051, 1.1395, 1.0468]}
{"unit_id": "orig:1926211761541206524", "ticker_idx": 0, "ts": "2025-05-24T09:40:45+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe NVIDIA is likely to maintain a one to two generation lead in AI scale-up networking", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA sees NVDA maintaining 1-2 gen lead in AI scale-up networking while UALink completes AMD's system portfolio; NVLink Fusion unlocks ~$12B TAM.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AI Interconnects Post-2026: NVLink Fusion vs. UALink\n\nUALink aims to challenge NVLink’s dominance in AI scale-up\n\nWe review two major AI scale-up interconnect protocols recently announced: 1) NVIDIA’s NVLink Fusion and 2) Ultra Accelerator Link (UALink), an open standard supported by AMD, Broadcom, and other ecosystem players. These two technologies are core scale-up technologies that allow AI accelerators (GPUs, ASICs, etc.) to connect via fast, low-latency shared memory networking and scale to large rack-scale systems. Importantly, UALink represents the industry’s latest effort to catch up to NVIDIA’s proprietary and closed NVLink technology, which has been a key competitive advantage for NVIDIA in building large-scale rack-level systems. Although UALink is generally one to two generations behind NVLink, we see it as a significant development that could enable the non-NVIDIA ecosystem to challenge NVIDIA’s dominance in AI scale-up starting in 2027. Until then, we expect NVIDIA to maintain its leadership across chips, software, and system/rack-level design. Meanwhile, NVIDIA’s NVLink Fusion partially opens NVLink IP to selected ecosystem partners (notably excluding Broadcom and AMD), unlocking a ~$12B networking TAM linked to custom CPUs/accelerators that NVIDIA previously could not access.\n\nAMD’s system-level portfolio completed with UALink\n\nUALink leverages AMD’s Infinity Fabric technology as a standard protocol, supporting features like shared memory coherency between CPU and GPU. AMD was a founding member of the UALink consortium and has contributed to building the supporting ecosystem, including related silicon IP, validation, and software. Importantly, UALink completes AMD’s system-level portfolio consisting of hardware (MI Instinct), software (ROCm), rack solutions (ZT Systems), scale-up (UALink), and scale-out (Ultra Ethernet). With the MI400 generation targeted for the second half of 2026, AMD plans to begin designing rack-level AI systems using ZT Systems technology, and UALink will provide much-needed intra-rack accelerator interconnect and switching functionality.\n\nNVIDIA and NVLink remain ahead\n\nHowever, the latest UALink 1.0 standard (expected to be adopted in systems in 2027) supports up to 800GB/s of bandwidth per accelerator—less than half of the 1.8TB/s offered by the latest NVLink 5.0 (see Table 6 for details). Even in terms of total aggregated bandwidth (per pod), UALink 1.0 supports a maximum of 819TB/s, which is only comparable to NVLink 5.0’s 1,000TB/s expected in Q4 2024. UALink-compatible systems are expected to roll out in earnest starting in 2027—coinciding with the expected launch of NVIDIA’s NVLink 6.0-based Rubin systems. Therefore, we believe NVIDIA is likely to maintain a one to two generation lead in AI scale-up networking.\n\nNVLink Fusion unlocks ~$12B TAM in custom systems\n\nMeanwhile, NVLink Fusion can expand NVIDIA’s TAM in the networking segment connected to custom accelerators. We assume that custom silicon will account for about 15% (~$75B) of the $500B AI computing TAM by 2029-30, with AI networking comprising around 15% ($11B) of that segment. Similarly, NVLink Fusion opens opportunities for NVIDIA-based networking in ARM CPU systems (e.g., Qualcomm/Fujitsu), offering an additional ~$1B networking TAM (15% of the ~$5-6B ARM CPU TAM). Through NVLink Fusion, all end customers (hyperscalers) will be able to choose between NVLink and UALink for their silicon, providing NVIDIA with further opportunities.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03206635423474524, "ret_signed_p1d": 0.03206635423474524, "alpha_spy_p1d": 0.011275686513277927, "alpha_c_p1d": 0.0003530997375458256, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.03171325449719942, "ret_p1w": 0.04638595688495273, "ret_signed_p1w": 0.04638595688495273, "alpha_spy_p1w": 0.022901539092194723, "alpha_c_p1w": 0.0217060176948789, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02467993919007383, "ret_p1m": 0.12659200664388637, "ret_signed_p1m": 0.12659200664388637, "alpha_spy_p1m": 0.07571425390240472, "alpha_c_p1m": -0.017023192846540747, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.14361519949042711, "ret_p3m": 0.33286728943802, "ret_signed_p3m": 0.33286728943802, "alpha_spy_p3m": 0.23216299569485654, "alpha_c_p3m": 0.12220362179599564, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.21066366764202438, "ret_p6m": 0.3815433597521749, "ret_signed_p6m": 0.3815433597521749, "alpha_spy_p6m": 0.23518624390365206, "alpha_c_p6m": -0.022137609503847244, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.40368096925602215, "price_path": [-0.0002, -0.0849, -0.0486, -0.1313, -0.1166, -0.1066, -0.1579, -0.1417, -0.1852, -0.1717, -0.1184, -0.1197, -0.0733, -0.0896, -0.1208, -0.1049, -0.0972, -0.1035, -0.0753, -0.0807, -0.1335, -0.1513, -0.1647, -0.1745, -0.161, -0.159, -0.2246, -0.2817, -0.2563, -0.2665, -0.1292, -0.1807, -0.1551, -0.1568, -0.1454, -0.2041, -0.227, -0.227, -0.2619, -0.2468, -0.2177, -0.1894, -0.1545, -0.1718, -0.1696, -0.1704, -0.1499, -0.1279, -0.1331, -0.1352, -0.1084, -0.106, -0.1115, -0.0631, -0.0104, 0.0308, 0.027, 0.0313, 0.0326, 0.0235, 0.0039, 0.0117, 0.0, 0.0, 0.0321, 0.0268, 0.0602, 0.0292, 0.0464, 0.0756, 0.081, 0.0663, 0.0794, 0.0864, 0.0965, 0.088, 0.1045, 0.0814, 0.1021, 0.0978, 0.1082, 0.1082, 0.0957, 0.0982, 0.1266, 0.1754, 0.1808, 0.2016, 0.2035, 0.1677, 0.1978, 0.2137, 0.2137, 0.2054, 0.2188, 0.2407, 0.25, 0.2562, 0.2498, 0.3003, 0.3054, 0.3178, 0.3133, 0.3054, 0.2723, 0.3009, 0.3234, 0.3216, 0.3463, 0.3369, 0.3655, 0.3549, 0.3233, 0.3711, 0.3579, 0.3667, 0.377, 0.3917, 0.3868, 0.3952, 0.3832, 0.3865, 0.3745, 0.3864, 0.3379, 0.3361, 0.3329, 0.3558, 0.3697, 0.3846, 0.3833, 0.3724, 0.3268, 0.3268, 0.3009, 0.2997, 0.3076, 0.2722, 0.2821, 0.3007, 0.3508, 0.3496, 0.3546, 0.354, 0.3322, 0.2972, 0.3425, 0.3458, 0.3987, 0.3592, 0.3481, 0.3536, 0.3574, 0.3853, 0.4213, 0.4263, 0.4389, 0.4292, 0.4134, 0.4096, 0.4406, 0.4669, 0.3953, 0.4346, 0.3714, 0.3699, 0.385, 0.3957, 0.3913, 0.38, 0.3733, 0.3876, 0.4189, 0.4587, 0.5314, 0.5772, 0.5456, 0.5425, 0.5759, 0.5136, 0.487, 0.4327, 0.4333, 0.5163, 0.4714, 0.4763, 0.4234, 0.4487, 0.4215, 0.3815]}
{"unit_id": "orig:1926211761541206524", "ticker_idx": 1, "ts": "2025-05-24T09:40:45+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UALink completes AMD's system-level portfolio consisting of hardware (MI Instinct), software (ROCm), rack solutions (ZT Systems), scale-up (UALink), and scale-out (Ultra Ethernet)", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA sees NVDA maintaining 1-2 gen lead in AI scale-up networking while UALink completes AMD's system portfolio; NVLink Fusion unlocks ~$12B TAM.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AI Interconnects Post-2026: NVLink Fusion vs. UALink\n\nUALink aims to challenge NVLink’s dominance in AI scale-up\n\nWe review two major AI scale-up interconnect protocols recently announced: 1) NVIDIA’s NVLink Fusion and 2) Ultra Accelerator Link (UALink), an open standard supported by AMD, Broadcom, and other ecosystem players. These two technologies are core scale-up technologies that allow AI accelerators (GPUs, ASICs, etc.) to connect via fast, low-latency shared memory networking and scale to large rack-scale systems. Importantly, UALink represents the industry’s latest effort to catch up to NVIDIA’s proprietary and closed NVLink technology, which has been a key competitive advantage for NVIDIA in building large-scale rack-level systems. Although UALink is generally one to two generations behind NVLink, we see it as a significant development that could enable the non-NVIDIA ecosystem to challenge NVIDIA’s dominance in AI scale-up starting in 2027. Until then, we expect NVIDIA to maintain its leadership across chips, software, and system/rack-level design. Meanwhile, NVIDIA’s NVLink Fusion partially opens NVLink IP to selected ecosystem partners (notably excluding Broadcom and AMD), unlocking a ~$12B networking TAM linked to custom CPUs/accelerators that NVIDIA previously could not access.\n\nAMD’s system-level portfolio completed with UALink\n\nUALink leverages AMD’s Infinity Fabric technology as a standard protocol, supporting features like shared memory coherency between CPU and GPU. AMD was a founding member of the UALink consortium and has contributed to building the supporting ecosystem, including related silicon IP, validation, and software. Importantly, UALink completes AMD’s system-level portfolio consisting of hardware (MI Instinct), software (ROCm), rack solutions (ZT Systems), scale-up (UALink), and scale-out (Ultra Ethernet). With the MI400 generation targeted for the second half of 2026, AMD plans to begin designing rack-level AI systems using ZT Systems technology, and UALink will provide much-needed intra-rack accelerator interconnect and switching functionality.\n\nNVIDIA and NVLink remain ahead\n\nHowever, the latest UALink 1.0 standard (expected to be adopted in systems in 2027) supports up to 800GB/s of bandwidth per accelerator—less than half of the 1.8TB/s offered by the latest NVLink 5.0 (see Table 6 for details). Even in terms of total aggregated bandwidth (per pod), UALink 1.0 supports a maximum of 819TB/s, which is only comparable to NVLink 5.0’s 1,000TB/s expected in Q4 2024. UALink-compatible systems are expected to roll out in earnest starting in 2027—coinciding with the expected launch of NVIDIA’s NVLink 6.0-based Rubin systems. Therefore, we believe NVIDIA is likely to maintain a one to two generation lead in AI scale-up networking.\n\nNVLink Fusion unlocks ~$12B TAM in custom systems\n\nMeanwhile, NVLink Fusion can expand NVIDIA’s TAM in the networking segment connected to custom accelerators. We assume that custom silicon will account for about 15% (~$75B) of the $500B AI computing TAM by 2029-30, with AI networking comprising around 15% ($11B) of that segment. Similarly, NVLink Fusion opens opportunities for NVIDIA-based networking in ARM CPU systems (e.g., Qualcomm/Fujitsu), offering an additional ~$1B networking TAM (15% of the ~$5-6B ARM CPU TAM). Through NVLink Fusion, all end customers (hyperscalers) will be able to choose between NVLink and UALink for their silicon, providing NVIDIA with further opportunities.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03852778618495134, "ret_signed_p1d": 0.03852778618495134, "alpha_spy_p1d": 0.017737118463484025, "alpha_c_p1d": 0.0068145316877519235, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.03171325449719942, "ret_p1w": 0.03916235872029228, "ret_signed_p1w": 0.03916235872029228, "alpha_spy_p1w": 0.015677940927534273, "alpha_c_p1w": 0.014482419530218449, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02467993919007383, "ret_p1m": 0.2549179198580882, "ret_signed_p1m": 0.2549179198580882, "alpha_spy_p1m": 0.20404016711660655, "alpha_c_p1m": 0.11130272036766109, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.14361519949042711, "ret_p3m": 0.48409038470795696, "ret_signed_p3m": 0.48409038470795696, "alpha_spy_p3m": 0.3833860909647935, "alpha_c_p3m": 0.2734267170659326, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.21066366764202438, "ret_p6m": 1.0876620286734107, "ret_signed_p6m": 1.0876620286734107, "alpha_spy_p6m": 0.9413049128248878, "alpha_c_p6m": 0.6839810594173885, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.40368096925602215, "price_path": [-0.0505, -0.0979, -0.0947, -0.1095, -0.0867, -0.0783, -0.1039, -0.0907, -0.124, -0.1228, -0.0863, -0.1106, -0.0847, -0.0519, -0.0616, -0.037, -0.0287, -0.0351, 0.0321, 0.0408, -0.0011, -0.0332, -0.0643, -0.0686, -0.0683, -0.0666, -0.1497, -0.2226, -0.2418, -0.291, -0.1221, -0.1959, -0.1533, -0.1433, -0.1362, -0.1996, -0.2068, -0.2068, -0.2244, -0.218, -0.1806, -0.1436, -0.1238, -0.1262, -0.1292, -0.1175, -0.1238, -0.1043, -0.0881, -0.106, -0.0902, -0.0781, -0.0677, -0.0199, 0.0195, 0.0672, 0.0424, 0.0622, 0.0402, 0.029, 0.0159, 0.0036, 0.0, 0.0, 0.0385, 0.0231, 0.0247, 0.0038, 0.0392, 0.0635, 0.075, 0.0488, 0.0533, 0.1035, 0.1172, 0.0982, 0.0742, 0.053, 0.1458, 0.1522, 0.1494, 0.1494, 0.1625, 0.1747, 0.2549, 0.3, 0.3025, 0.3037, 0.2864, 0.2339, 0.2557, 0.2502, 0.2502, 0.222, 0.2494, 0.2547, 0.3069, 0.3274, 0.3257, 0.4107, 0.4512, 0.4542, 0.4232, 0.4233, 0.4026, 0.4382, 0.4697, 0.5091, 0.5743, 0.6086, 0.6273, 0.5983, 0.5565, 0.6026, 0.5802, 0.4787, 0.5629, 0.5661, 0.5618, 0.586, 0.6718, 0.6404, 0.6092, 0.5968, 0.5098, 0.4976, 0.4841, 0.5208, 0.4809, 0.5105, 0.5151, 0.5282, 0.4743, 0.4743, 0.4715, 0.4698, 0.4667, 0.3701, 0.3726, 0.4126, 0.4463, 0.4112, 0.4375, 0.461, 0.4546, 0.4428, 0.4316, 0.4268, 0.4486, 0.4586, 0.4584, 0.462, 0.4456, 0.4628, 0.4667, 0.4868, 0.5387, 0.4928, 0.8467, 0.9174, 1.1354, 1.1112, 0.9481, 0.9619, 0.9771, 1.163, 1.1264, 1.113, 1.1808, 1.1578, 1.0871, 1.1303, 1.2928, 1.354, 1.339, 1.3962, 1.3102, 1.3218, 1.3538, 1.2668, 1.3237, 1.1548, 1.1171, 1.2118, 1.1532, 1.3469, 1.2478, 1.2374, 1.1804, 1.0877]}
{"unit_id": "orig:1921545333601096021", "ticker_idx": 0, "ts": "2025-05-11T12:38:02+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Short-term (3–6 months): Prefer Samsung Electronics (SEC) — strong downside protection in case of recession", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares JP Morgan note: prefer Samsung near-term (3–6m) for downside protection, SK Hynix mid-to-long-term (6–12m+) on HBM/AI leadership.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Memory TAM Revised Downward\n\n• Memory TAM Revision for 2025–2026: Adjusted downward by -1% for 2025 and -5% for 2026 to reflect U.S.-China tariff risks\n• Preference for DRAM and HBM: AI demand driving higher share of DRAM (especially HBM); AI-related demand expected to account for 40% of total DRAM TAM\n• Relatively Weak NAND: Limited AI impact and weak consumer electronics demand leading to reduced content growth\n\n⸻\n\n🧠 HBM Market Outlook\n\n• HBM Demand Growth Adjustment: 2025 forecast revised slightly from +122% → +122%, and 2026 from +52% → +47%\n• 2026 Supply Adjustment: TSV capacity cut due to CoWoS bottlenecks and slower ASIC demand → 7% reduction in HBM supply\n• Expected Inventory Build in 2026: About 6 weeks’ worth of inventory expected to accumulate, but at manageable levels\n\n⸻\n\n💾 DRAM/NAND Pricing Outlook\n\n• DRAM: Strong in 1H25 due to pull-in from server and PC demand, but ASP expected to decline starting 4Q25\n• NAND: Prices to continue falling through 2Q25, rebound in 3Q25, and stabilize in 4Q25 due to supply adjustments\n\n⸻\n\n📉 Demand Outlook (2025–2026E)\n\n• Smartphone Shipments: -3% in 2025, -5% in 2026\n• Notebook Shipments: -3% in 2025, -10% in 2026\n• Server Shipments: -2% in 2025, -6% in 2026\n\n⸻\n\n🏗 Capex\n\n• SK Hynix: DRAM Capex raised (W19T → W22T), reflecting M15X line and HBM backend conversion\n• Samsung Electronics: DRAM +10%, NAND +30% increase; continued aggressive NAND investment\n\n⸻\n\n📊 Stock & Investment Strategy\n\n• Short-term (3–6 months): Prefer Samsung Electronics (SEC) — strong downside protection in case of recession\n• Mid-to-Long-term (6–12+ months): Prefer SK Hynix — sustained AI upside based on HBM leadership\n• Ultra-short-term (1–3 months): NYT may rebound — benefits from strong commodity DRAM pricing\n\n⸻\n\n📌 Conclusion & Risk Factors\n\n• Positive Factors: Resilient AI demand, limited DRAM/HBM supply, focus on high-margin products\n• Negative Factors: Potential demand shock and ASP decline if tariffs intensify, weak NAND demand\n• Neutral View: Memory fundamentals attractive in the long term, but short-term correction risks remain", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.048611076125371966, "ret_signed_m1d": -0.048611076125371966, "alpha_spy_m1d": -0.0166207649542357, "alpha_c_m1d": -0.004254556057067815, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.04435652006830415, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012152700018378204, "ret_signed_p1d": -0.012152700018378204, "alpha_spy_p1d": -0.018756640335548247, "alpha_c_p1d": -0.033760152522467224, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.02160745250408902, "ret_p1w": -0.031249829624243053, "ret_signed_p1w": -0.031249829624243053, "alpha_spy_p1w": -0.051593129897988255, "alpha_c_p1w": -0.06133343191680074, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.030083602292557687, "ret_p1m": 0.0277777873629117, "ret_signed_p1m": 0.0277777873629117, "alpha_spy_p1m": -0.006682383085334198, "alpha_c_p1m": -0.03006150378033512, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.05783929114324682, "ret_p3m": 0.23146582640161362, "ret_signed_p3m": 0.23146582640161362, "alpha_spy_p3m": 0.1437644315834259, "alpha_c_p3m": 0.07324717183597218, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.15821865456564144, "ret_p6m": 0.8405266351742104, "ret_signed_p6m": 0.8405266351742104, "alpha_spy_p6m": 0.6756459501112362, "alpha_c_p6m": 0.5458161362583034, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.29471049891590706, "price_path": [-0.037, -0.037, -0.0335, -0.0335, -0.018, 0.0131, 0.0079, 0.0044, -0.0111, -0.0128, -0.0232, -0.0283, -0.0594, -0.0594, -0.0594, -0.068, -0.0629, -0.0732, -0.0732, -0.0749, -0.0525, -0.056, -0.056, -0.0059, -0.0059, 0.0096, 0.039, 0.0649, 0.0441, 0.0321, 0.0597, 0.0666, 0.0451, 0.0035, 0.0208, 0.0208, 0.0, -0.026, -0.0764, -0.0712, -0.0799, -0.0208, -0.0417, -0.0243, -0.0174, -0.0503, -0.0434, -0.0399, -0.0382, -0.0451, -0.033, -0.033, -0.033, -0.0312, -0.0312, -0.0365, -0.0365, -0.0573, -0.0573, -0.0573, -0.0521, -0.0521, -0.0486, 0.0, -0.0122, -0.0035, -0.0052, -0.0139, -0.0312, -0.0295, -0.033, -0.0503, -0.059, -0.0503, -0.0642, -0.0295, -0.026, -0.0243, -0.0139, -0.0139, 0.0035, 0.026, 0.026, 0.0382, 0.0278, 0.0399, 0.033, 0.0122, -0.0069, 0.0087, 0.0382, 0.0278, 0.033, 0.0069, 0.0503, 0.0642, 0.0451, 0.062, 0.0446, 0.0515, 0.062, 0.1144, 0.1057, 0.0778, 0.0725, 0.055, 0.0655, 0.0935, 0.0917, 0.1127, 0.1302, 0.1651, 0.1721, 0.1843, 0.1529, 0.1598, 0.1529, 0.1511, 0.2297, 0.2332, 0.2681, 0.2472, 0.2035, 0.2175, 0.221, 0.2018, 0.2315, 0.2542, 0.2402, 0.2419, 0.2559, 0.2507, 0.2507, 0.2227, 0.2227, 0.2315, 0.2332, 0.2472, 0.2489, 0.228, 0.2332, 0.2157, 0.2175, 0.1808, 0.207, 0.2192, 0.2245, 0.214, 0.2245, 0.2489, 0.2681, 0.2821, 0.3171, 0.3363, 0.3869, 0.366, 0.4026, 0.4026, 0.4585, 0.4795, 0.4917, 0.504, 0.4551, 0.4773, 0.4721, 0.5089, 0.5747, 0.5747, 0.5747, 0.5747, 0.5747, 0.5747, 0.6563, 0.637, 0.6072, 0.6668, 0.7142, 0.7177, 0.7212, 0.7107, 0.73, 0.6931, 0.7335, 0.7896, 0.7458, 0.7633, 0.8265, 0.8861, 0.9493, 0.8405]}
{"unit_id": "orig:1921545333601096021", "ticker_idx": 1, "ts": "2025-05-11T12:38:02+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Mid-to-Long-term (6–12+ months): Prefer SK Hynix — sustained AI upside based on HBM leadership", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares JP Morgan note: prefer Samsung near-term (3–6m) for downside protection, SK Hynix mid-to-long-term (6–12m+) on HBM/AI leadership.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Memory TAM Revised Downward\n\n• Memory TAM Revision for 2025–2026: Adjusted downward by -1% for 2025 and -5% for 2026 to reflect U.S.-China tariff risks\n• Preference for DRAM and HBM: AI demand driving higher share of DRAM (especially HBM); AI-related demand expected to account for 40% of total DRAM TAM\n• Relatively Weak NAND: Limited AI impact and weak consumer electronics demand leading to reduced content growth\n\n⸻\n\n🧠 HBM Market Outlook\n\n• HBM Demand Growth Adjustment: 2025 forecast revised slightly from +122% → +122%, and 2026 from +52% → +47%\n• 2026 Supply Adjustment: TSV capacity cut due to CoWoS bottlenecks and slower ASIC demand → 7% reduction in HBM supply\n• Expected Inventory Build in 2026: About 6 weeks’ worth of inventory expected to accumulate, but at manageable levels\n\n⸻\n\n💾 DRAM/NAND Pricing Outlook\n\n• DRAM: Strong in 1H25 due to pull-in from server and PC demand, but ASP expected to decline starting 4Q25\n• NAND: Prices to continue falling through 2Q25, rebound in 3Q25, and stabilize in 4Q25 due to supply adjustments\n\n⸻\n\n📉 Demand Outlook (2025–2026E)\n\n• Smartphone Shipments: -3% in 2025, -5% in 2026\n• Notebook Shipments: -3% in 2025, -10% in 2026\n• Server Shipments: -2% in 2025, -6% in 2026\n\n⸻\n\n🏗 Capex\n\n• SK Hynix: DRAM Capex raised (W19T → W22T), reflecting M15X line and HBM backend conversion\n• Samsung Electronics: DRAM +10%, NAND +30% increase; continued aggressive NAND investment\n\n⸻\n\n📊 Stock & Investment Strategy\n\n• Short-term (3–6 months): Prefer Samsung Electronics (SEC) — strong downside protection in case of recession\n• Mid-to-Long-term (6–12+ months): Prefer SK Hynix — sustained AI upside based on HBM leadership\n• Ultra-short-term (1–3 months): NYT may rebound — benefits from strong commodity DRAM pricing\n\n⸻\n\n📌 Conclusion & Risk Factors\n\n• Positive Factors: Resilient AI demand, limited DRAM/HBM supply, focus on high-margin products\n• Negative Factors: Potential demand shock and ASP decline if tariffs intensify, weak NAND demand\n• Neutral View: Memory fundamentals attractive in the long term, but short-term correction risks remain", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02512820760846257, "ret_signed_m1d": -0.02512820760846257, "alpha_spy_m1d": 0.006862103562673694, "alpha_c_m1d": 0.03388367449388019, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017948823359660082, "ret_signed_p1d": 0.017948823359660082, "alpha_spy_p1d": 0.01134488304249004, "alpha_c_p1d": -0.01637993640323554, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.022564136047578343, "ret_signed_p1w": 0.022564136047578343, "alpha_spy_p1w": 0.002220835773833141, "alpha_c_p1w": -0.013660096880542705, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.18418625046954373, "ret_signed_p1m": 0.18418625046954373, "alpha_spy_p1m": 0.14972608002129784, "alpha_c_p1m": 0.08031528809333222, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.3460164956963192, "ret_signed_p3m": 0.3460164956963192, "alpha_spy_p3m": 0.2583151008781315, "alpha_c_p3m": 0.11978345655460498, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 2.014904164711914, "ret_signed_p6m": 2.014904164711914, "alpha_spy_p6m": 1.8500234796489397, "alpha_c_p6m": 1.529035787529001, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0134, 0.0624, 0.07, 0.0802, 0.07, 0.1133, 0.0802, 0.0675, 0.0445, 0.0216, 0.0343, 0.0215, -0.0246, -0.0246, -0.0456, -0.0097, -0.0133, -0.0133, -0.0364, -0.0369, 0.02, 0.0241, 0.0487, 0.0564, 0.041, 0.0538, 0.0769, 0.1051, 0.0846, 0.0667, 0.0974, 0.0615, 0.0221, -0.0221, 0.0103, 0.0149, -0.0021, -0.0656, -0.1549, -0.1308, -0.1538, -0.0605, -0.0728, -0.0759, -0.0738, -0.1077, -0.1026, -0.1026, -0.0944, -0.1087, -0.0718, -0.0856, -0.0544, -0.0667, -0.0728, -0.0897, -0.0897, -0.0462, -0.0462, -0.0462, -0.0215, -0.0241, -0.0251, 0.0, 0.0179, 0.0564, 0.0282, 0.0487, 0.0226, 0.0359, 0.0282, 0.0097, 0.0256, 0.041, 0.0385, 0.0667, 0.0891, 0.0506, 0.066, 0.066, 0.1174, 0.1534, 0.1534, 0.1765, 0.1842, 0.233, 0.2099, 0.2099, 0.2741, 0.2792, 0.2664, 0.2638, 0.3203, 0.3332, 0.4308, 0.4693, 0.5053, 0.459, 0.5001, 0.4667, 0.4334, 0.4308, 0.3897, 0.3923, 0.4488, 0.4436, 0.5258, 0.513, 0.5412, 0.5335, 0.5207, 0.3845, 0.382, 0.4, 0.3794, 0.382, 0.3845, 0.3666, 0.346, 0.3486, 0.3537, 0.4051, 0.3255, 0.3255, 0.3537, 0.328, 0.346, 0.3178, 0.3717, 0.382, 0.4282, 0.4205, 0.4205, 0.3743, 0.3512, 0.3126, 0.2587, 0.2895, 0.3332, 0.3434, 0.3357, 0.3814, 0.384, 0.3171, 0.3402, 0.3505, 0.366, 0.4071, 0.4251, 0.4817, 0.564, 0.5795, 0.6901, 0.703, 0.7904, 0.7158, 0.829, 0.829, 0.8059, 0.8573, 0.8393, 0.8342, 0.7313, 0.7956, 0.7878, 0.8522, 1.0348, 1.0348, 1.0348, 1.0348, 1.0348, 1.0348, 1.202, 1.1351, 1.1171, 1.1737, 1.3281, 1.3949, 1.4978, 1.4644, 1.4773, 1.4618, 1.6239, 1.7525, 1.6805, 1.8708, 1.9223, 1.876, 2.1898, 2.0149]}
{"unit_id": "orig:1943278932582883376", "ticker_idx": 0, "ts": "2025-07-10T11:59:36+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "J.P. Morgan projects Broadcom's EPS to increase by 38% year-over-year to $6.71 in fiscal year 2025, followed by a further 25% growth to $8.38 in 2026", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares J.P. Morgan's bullish Broadcom projections: 38% EPS growth in FY2025, 25% in FY2026, FCF $39B, net debt declining sharply.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan projects Broadcom's EPS to increase by 38% year-over-year to $6.71 in fiscal year 2025, followed by a further 25% growth to $8.38 in 2026. Free cash flow is expected to reach $39 billion in 2026, and the net debt-to-equity ratio is forecast to decline from 0.9 in 2024 to 0.2 in 2026, indicating a robust financial position.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009077835299488024, "ret_signed_m1d": 0.009077835299488024, "alpha_spy_m1d": 0.01189021373818222, "alpha_c_m1d": 0.016381532078634242, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003703619499012767, "ret_signed_p1d": -0.003703619499012767, "alpha_spy_p1d": -0.00018819562841720927, "alpha_c_p1d": -0.0035992779315682233, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": 0.04012362189736063, "ret_signed_p1w": 0.04012362189736063, "alpha_spy_p1w": 0.0365763308303555, "alpha_c_p1w": 0.025550724938917302, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.10737130836109499, "ret_signed_p1m": 0.10737130836109499, "alpha_spy_p1m": 0.08921920249512105, "alpha_c_p1m": 0.08646839180134291, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.22362530765050637, "ret_signed_p3m": 0.22362530765050637, "alpha_spy_p3m": 0.15147187693534847, "alpha_c_p3m": 0.0513589768643139, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.2668187477682662, "ret_signed_p6m": 0.2668187477682662, "alpha_spy_p6m": 0.16891790097620607, "alpha_c_p6m": -0.035565669989816806, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.3539, -0.3517, -0.3675, -0.3806, -0.3806, -0.3979, -0.3857, -0.3591, -0.3184, -0.3033, -0.3028, -0.3075, -0.3028, -0.2852, -0.2623, -0.2729, -0.2752, -0.2581, -0.2473, -0.2458, -0.1973, -0.158, -0.1591, -0.1572, -0.1718, -0.1645, -0.1607, -0.1678, -0.1649, -0.1714, -0.1714, -0.1463, -0.1327, -0.1235, -0.1231, -0.099, -0.0695, -0.0542, -0.0584, -0.1055, -0.1151, -0.1138, -0.0838, -0.0724, -0.0991, -0.0868, -0.0966, -0.0898, -0.0898, -0.0923, -0.0785, -0.0422, -0.039, -0.019, -0.022, 0.0009, -0.0387, -0.02, -0.0008, -0.0008, -0.0044, -0.0131, 0.0091, 0.0, -0.0037, 0.0007, 0.0201, 0.0196, 0.0401, 0.0288, 0.0465, 0.0116, 0.0301, 0.0483, 0.0537, 0.0686, 0.08, 0.0988, 0.0664, 0.0481, 0.081, 0.0637, 0.0954, 0.103, 0.1074, 0.1035, 0.1359, 0.1223, 0.1301, 0.1123, 0.1102, 0.0708, 0.0573, 0.0516, 0.0675, 0.0684, 0.0821, 0.0902, 0.1207, 0.0798, 0.0798, 0.0829, 0.098, 0.1115, 0.216, 0.2551, 0.2225, 0.3419, 0.3058, 0.3067, 0.322, 0.3072, 0.257, 0.254, 0.2525, 0.2323, 0.2328, 0.2342, 0.2225, 0.2168, 0.1927, 0.2, 0.2126, 0.2301, 0.2308, 0.2203, 0.2236, 0.2567, 0.2549, 0.1808, 0.2974, 0.2517, 0.2779, 0.2882, 0.2706, 0.2703, 0.2464, 0.2378, 0.2523, 0.2881, 0.3169, 0.3566, 0.4039, 0.3693, 0.3445, 0.3187, 0.2801, 0.3057, 0.2934, 0.271, 0.3036, 0.2802, 0.292, 0.2366, 0.2456, 0.2463, 0.2385, 0.2891, 0.2615, 0.2374, 0.3748, 0.4005, 0.4461, 0.4461, 0.4657, 0.4043, 0.3879, 0.3844, 0.3859, 0.4194, 0.4589, 0.4778, 0.5021, 0.4781, 0.3092, 0.236, 0.2414, 0.1858, 0.1999, 0.238, 0.2443, 0.273, 0.2763, 0.2763, 0.2833, 0.2733, 0.2749, 0.2613, 0.2613, 0.2668]}
{"unit_id": "orig:1933031934005031273", "ticker_idx": 0, "ts": "2025-06-12T05:21:41+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "demand for high-performance computing chips driven by AI applications remains strong. Advanced process technologies and advanced packaging are the main drivers of global foundry industry growth. The industry's annual growth rate is expected to reach 19.1% in 2025", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares TrendForce bullish foundry outlook: 19.1% industry growth in 2025, 2nm mass production H2, advanced packaging up 76%; implicitly long TSM.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "【TrendForce: Global Foundry Industry to Grow 19.1% in 2025】\n\nAt the recently held “TSS 2025 Semiconductor Industry Executive Forum” hosted by TrendForce, Guo Zuorong, Senior Deputy General Manager of Research at TrendForce, stated that demand for high-performance computing chips driven by AI applications remains strong. Advanced process technologies and advanced packaging are the main drivers of global foundry industry growth. The industry’s annual growth rate is expected to reach 19.1% in 2025.\nIn addition, 2nm advanced process technology will officially enter mass production in the second half of this year, and advanced packaging capacity will continue to expand, with an expected annual growth of 76%.\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009794317055577761, "ret_signed_m1d": -0.009794317055577761, "alpha_spy_m1d": -0.005835822909165289, "alpha_c_m1d": -0.00496820360084671, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.004826113454731051, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020099254909447195, "ret_signed_p1d": -0.020099254909447195, "alpha_spy_p1d": -0.008919190512979736, "alpha_c_p1d": 0.003309477673309913, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.023408732582757108, "ret_p1w": -0.008958833689798662, "ret_signed_p1w": -0.008958833689798662, "alpha_spy_p1w": 0.0014923761545641012, "alpha_c_p1w": -0.006830788242481622, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.0021280454473170396, "ret_p1m": 0.06948893909759213, "ret_signed_p1m": 0.06948893909759213, "alpha_spy_p1m": 0.03352434335061383, "alpha_c_p1m": -0.023005829782107634, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.09249476887969976, "ret_p3m": 0.1647403299980681, "ret_signed_p3m": 0.1647403299980681, "alpha_spy_p3m": 0.0844047705975981, "alpha_c_p3m": 0.03302835797710202, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.13171197202096607, "ret_p6m": 0.37237052079749144, "ret_signed_p6m": 0.37237052079749144, "alpha_spy_p6m": 0.23013651819885084, "alpha_c_p6m": -0.012467181564879803, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.38483770236237125, "price_path": [-0.1881, -0.1993, -0.1964, -0.1792, -0.1826, -0.1621, -0.1633, -0.1976, -0.2219, -0.2357, -0.2323, -0.22, -0.212, -0.2721, -0.3211, -0.324, -0.3462, -0.2658, -0.301, -0.2735, -0.2792, -0.2724, -0.2985, -0.2982, -0.2982, -0.3162, -0.2998, -0.2701, -0.2407, -0.2364, -0.2443, -0.2393, -0.2291, -0.2012, -0.1708, -0.1842, -0.2032, -0.1928, -0.1896, -0.1836, -0.1352, -0.1028, -0.0992, -0.1017, -0.1017, -0.1051, -0.1053, -0.1131, -0.0926, -0.1121, -0.1121, -0.0857, -0.0929, -0.0882, -0.1059, -0.0989, -0.0861, -0.0639, -0.0596, -0.051, -0.0426, -0.0174, -0.0098, 0.0, -0.0201, 0.0012, -0.0071, -0.009, -0.009, -0.0275, -0.0237, 0.0216, 0.0339, 0.0398, 0.061, 0.0513, 0.0429, 0.0843, 0.0899, 0.0899, 0.0638, 0.0577, 0.0762, 0.0665, 0.0695, 0.0615, 0.0999, 0.1027, 0.14, 0.1159, 0.1087, 0.089, 0.1156, 0.1215, 0.14, 0.1268, 0.1202, 0.1276, 0.1216, 0.0918, 0.1094, 0.0791, 0.074, 0.1262, 0.1225, 0.1238, 0.134, 0.1207, 0.1187, 0.1089, 0.1206, 0.0802, 0.0611, 0.0552, 0.0815, 0.0936, 0.1081, 0.1108, 0.106, 0.0717, 0.0717, 0.0602, 0.0741, 0.0918, 0.1299, 0.1474, 0.1647, 0.2089, 0.2018, 0.2038, 0.2133, 0.2203, 0.2237, 0.2509, 0.2334, 0.2695, 0.3164, 0.3071, 0.2883, 0.2729, 0.2723, 0.3005, 0.3433, 0.3416, 0.3606, 0.4081, 0.3692, 0.418, 0.3964, 0.3069, 0.4104, 0.3781, 0.4189, 0.3962, 0.374, 0.3862, 0.3714, 0.3452, 0.3538, 0.3735, 0.3888, 0.4041, 0.4207, 0.412, 0.399, 0.4196, 0.3693, 0.3673, 0.3469, 0.3341, 0.3749, 0.3558, 0.3533, 0.3141, 0.3263, 0.3132, 0.2941, 0.3149, 0.2922, 0.2808, 0.3254, 0.3256, 0.3502, 0.3502, 0.3574, 0.3396, 0.3601, 0.3758, 0.364, 0.3724]}
{"unit_id": "orig:1932052448010739928", "ticker_idx": 0, "ts": "2025-06-09T12:29:34+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our 'Buy' rating on TSMC, with a 12-month target price of NT$1,145... For ADR (American Depositary Receipt, TSM), our 12-month target price is US$229", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Sachs Buy on TSMC/TSM, PT US$229 ADR / NT$1,145, on CoWoS growth, advanced node transition, and 26% revenue growth in 2025.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs TSMC Analysis\n\n1. Improving Sentiment and Limited Downside Risk to Earnings\n\nEasing supply chain mismatches, lower probability of AI order cuts\n\n1. Limited downside risk to earnings at this juncture – We believe the likelihood of further earnings downside for TSMC is limited following our last earnings estimate revision, which accounted for unfavorable foreign exchange trends and revised CoWoS (Chip-on-Wafer-on-Substrate) outlook.\n\n2. Easing concerns on further AI order cuts – Supply chain mismatches between upstream/downstream are easing as assembly yields of downstream ODM (Original Design Manufacturer) players improve.\n\n3. Sustained strength in CoWoS shipments – We expect TSMC's CoWoS shipments to grow strongly by 52% year-on-year from 585k in 2025 to 923k in 2026.\n\n2. Strong Growth Outlook Heading into 2026\n\n1. Advanced node transition to drive revenue growthWe forecast TSMC's revenue (in USD) to exhibit strong growth of 26.3%/14.1% in 2025/26, respectively.\n\n2. Transition to more advanced nodesAI customers will transition from N4 (4nm) to N3 (3nm).Non-AI customers (mainly smartphone and PC) will transition to N2 (2nm), driving blended ASP upside.\n\n3. The proportion of advanced nodes (7nm and below) in TSMC's wafer revenue will reach 79.3% in 2026, up from 68.8% in 2024.\n\n3. Expect Further Price Increases in 2026\n\n1. Advanced node transition to drive revenue growthPrice increases for advanced nodes and CoWoS to offset rising costs\n\n2. We expect the average selling price (ASP) of 5nm and below nodes/CoWoS to increase by 3%/5% year-on-year, respectively, in 2026.\n\n・This will offset rising costs due to higher public utility charges in Taiwan and accelerated fab expansion in the US.\n\n3. Gross margin (GM) will remain at 56.8%/54.7% in 2025/26, respectively, which is above the long-term profitability target of >53%.\n\n4. Recommendation, Target Price Risks, and Valuation Methodology\nTarget Price\n\nWe maintain our 'Buy' rating on TSMC, with a 12-month target price of NT$1,145, derived by applying a target P/E (price-to-earnings) multiple of 20x to our 2026 estimated EPS (earnings per share).\n\nThe target P/E of 20x represents +0.5 standard deviation above the 5-year average trading P/E.\nFor ADR (American Depositary Receipt, TSM), our 12-month target price is US$229, derived by applying a USD/TWD exchange rate of 30.0 and a 20% ADR premium.\n\n5. Key Risks\n\n1. Further deterioration in end-demand recovery impacting utilization rates.\n2. Slower-than-expected adoption of process transitions by customers.\n3. Unfavorable foreign exchange trends or higher-than-expected cost increases weighing on margin outlook.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00879232261884455, "ret_signed_m1d": -0.00879232261884455, "alpha_spy_m1d": -0.007891851310082076, "alpha_c_m1d": 0.007669303414965967, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026376893229590026, "ret_signed_p1d": 0.026376893229590026, "alpha_spy_p1d": 0.020707228560699287, "alpha_c_p1d": 0.006490417905588686, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.045742211969475255, "ret_signed_p1w": 0.045742211969475255, "alpha_spy_p1w": 0.040739536395070886, "alpha_c_p1w": 0.02075765762518955, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.10479794531128617, "ret_signed_p1m": 0.10479794531128617, "alpha_spy_p1m": 0.06728801871405876, "alpha_c_p1m": 0.0015910626365585046, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.14043507209634387, "ret_signed_p3m": 0.14043507209634387, "alpha_spy_p3m": 0.05479116047450927, "alpha_c_p3m": 0.012399580802975985, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.42069014101955027, "ret_signed_p6m": 0.42069014101955027, "alpha_spy_p6m": 0.27768063763457285, "alpha_c_p6m": 0.021912021646373514, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1474, -0.1743, -0.1622, -0.1519, -0.1636, -0.1606, -0.1427, -0.1462, -0.1248, -0.1261, -0.1618, -0.1872, -0.2017, -0.1981, -0.1853, -0.1769, -0.2397, -0.2908, -0.2939, -0.3171, -0.2331, -0.2699, -0.2412, -0.2471, -0.24, -0.2673, -0.267, -0.267, -0.2857, -0.2686, -0.2376, -0.2069, -0.2024, -0.2106, -0.2054, -0.1947, -0.1656, -0.1339, -0.1478, -0.1677, -0.1568, -0.1535, -0.1472, -0.0967, -0.0629, -0.0591, -0.0617, -0.0617, -0.0652, -0.0655, -0.0736, -0.0522, -0.0726, -0.0726, -0.045, -0.0525, -0.0476, -0.0661, -0.0587, -0.0454, -0.0222, -0.0177, -0.0088, 0.0, 0.0264, 0.0343, 0.0445, 0.0235, 0.0457, 0.0371, 0.0352, 0.0352, 0.0158, 0.0198, 0.0671, 0.08, 0.0861, 0.1082, 0.0982, 0.0894, 0.1326, 0.1384, 0.1384, 0.1111, 0.1048, 0.1241, 0.114, 0.1171, 0.1087, 0.1489, 0.1518, 0.1908, 0.1656, 0.1581, 0.1375, 0.1653, 0.1714, 0.1908, 0.177, 0.1701, 0.1778, 0.1715, 0.1404, 0.1588, 0.1271, 0.1218, 0.1764, 0.1725, 0.1738, 0.1845, 0.1706, 0.1685, 0.1582, 0.1705, 0.1283, 0.1084, 0.1022, 0.1297, 0.1423, 0.1575, 0.1602, 0.1553, 0.1194, 0.1194, 0.1074, 0.1219, 0.1404, 0.1802, 0.1985, 0.2166, 0.2628, 0.2553, 0.2574, 0.2673, 0.2746, 0.2782, 0.3066, 0.2883, 0.326, 0.3751, 0.3653, 0.3456, 0.3296, 0.329, 0.3584, 0.4031, 0.4013, 0.4212, 0.4708, 0.4301, 0.4811, 0.4586, 0.3651, 0.4732, 0.4394, 0.4821, 0.4584, 0.4352, 0.448, 0.4325, 0.4051, 0.4141, 0.4346, 0.4507, 0.4666, 0.4839, 0.4748, 0.4613, 0.4828, 0.4302, 0.4282, 0.4068, 0.3935, 0.4362, 0.4162, 0.4135, 0.3726, 0.3853, 0.3717, 0.3517, 0.3734, 0.3497, 0.3379, 0.3845, 0.3846, 0.4103, 0.4103, 0.4179, 0.3992, 0.4207]}
{"unit_id": "orig:1940901032256852150", "ticker_idx": 0, "ts": "2025-07-03T22:30:40+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's CoWoS capacity is expected to grow by over 30% in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish survey: TSMC CoWoS +33% in 2026 supports NVDA/AI ASICs; Alchip sole Trainium3 provider with strong 2026 revenue growth; Marvell growth tied to Trainium2 lifecycle.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Survey: CoWoS Expansion Drives Computing Chip Growth, B30 Export to China a Key Variable\n\nMorgan Stanley (C. Chan, 25/06/30)\n\nMorgan Stanley's survey indicates that TSMC's CoWoS capacity is expected to grow by over 30% in 2026, which will support the NVIDIA and AI ASIC supply chains. Chinese AI demand is strong but constrained by hardware supply, and whether B30 GPUs can be exported to China remains the biggest uncertain factor. Meanwhile, Asian ASIC design service providers like Alchip are showing positive growth prospects due to AWS custom chip orders.\n\nTSMC CoWoS Capacity to Grow 33% in 2026, Driving GPU and ASIC Chip Demand\n\nMorgan Stanley has upgraded its forecast for TSMC's CoWoS capacity in 2026 to 90,000-95,000 wafers, an increase of approximately 33% from the 70,000 wafers at the end of 2025. Among these, CoWoS-L capacity may increase to 68,000 wafers, reflecting strong demand for Blackwell or Rubin chips. Concurrently, the testing time for Blackwell chips is expected to increase to 800-900 seconds, verifying the rising complexity of testing, with KYEC maintaining its leading share.\n\nStrong Demand from Chinese AI Developers, But Limited by B30 GPU Supply and Training Capacity Bottlenecks\n\nAlthough Chinese developers are aware of NVIDIA's B30 chip (RTX 6000 Pro), there is currently no official order confirmation. Taiwanese supply chain sources indicate that 2 million B30 chips are expected to begin tape-out in the second half of 2025. If export to China is not possible, some developers plan to switch to Huawei's 910C chip, but the latter is not yet readily available. Some developers are delaying capital expenditures until the availability of B30 is clear, while local ASICs can currently only be used for inference tasks.\n\nDelay in China's Next-Generation Large Model Training Attributed to Restricted GPU Resources\n\nMorgan Stanley's survey found that progress in training large AI models in China is lagging, mainly due to a shortage of GPU training compute power. While inference demand is more active, training resources remain a bottleneck, impacting the overall pace of AI deployment.\n\nBullish on Asian ASIC Design Service Providers, Alchip Expected to Benefit from AWS Trainium3 Orders\n\nFor AWS's custom chip Trainium3 project, Alchip is the sole service provider, and is expected to see significant revenue growth in 2026. The actual scale will depend on TSMC's 3nm and CoWoS capacity allocation. Marvell, on the other hand, is focusing on \"XPU-attach\" chips, and its growth will depend on the lifespan of Trainium2 and the shipment volume of its attached chips. Morgan Stanley expects the design service provider for Trainium4 to be selected in July, with Alchip having a higher chance of winning.\n\nInvestors Generally Positive on AI, But Focus Shifts to Demand Sustainability and Alternative Investments\n\nIn this survey, investors are generally optimistic about the AI sector, but their focus has shifted to the sustainability of future demand and the selection of investment targets beyond NVIDIA. This reflects the market's further discernment and positioning regarding the return on AI capital expenditure and beneficiaries in the mid-stream of the industry chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004587151290653346, "ret_signed_m1d": -0.004587151290653346, "alpha_spy_m1d": 0.003232604153493579, "alpha_c_m1d": 0.0037691794015815017, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004587151290653346, "ret_signed_p1d": -0.004587151290653346, "alpha_spy_p1d": -0.004587151290653346, "alpha_c_p1d": -0.004587151290653346, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.009174415763257837, "ret_signed_p1w": 0.009174415763257837, "alpha_spy_p1w": 0.00840685052291934, "alpha_c_p1w": -0.0045762922905956405, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.06422023125109866, "ret_signed_p1m": 0.06422023125109866, "alpha_spy_p1m": 0.07000925593435492, "alpha_c_p1m": 0.06305669153445836, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.2020367997531729, "ret_signed_p3m": 0.2020367997531729, "alpha_spy_p3m": 0.13377487436527824, "alpha_c_p3m": 0.051342324127495065, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.3954987085373485, "ret_signed_p6m": 0.3954987085373485, "alpha_spy_p6m": 0.28527189094177663, "alpha_c_p6m": 0.1015194976892726, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.2253, -0.2545, -0.2829, -0.2116, -0.1878, -0.2098, -0.1988, -0.2189, -0.2262, -0.2235, -0.2372, -0.2545, -0.2025, -0.2107, -0.1888, -0.1805, -0.176, -0.1705, -0.1705, -0.1321, -0.1431, -0.1595, -0.1522, -0.1614, -0.133, -0.1257, -0.1148, -0.0874, -0.0928, -0.0883, -0.1011, -0.1065, -0.0956, -0.1029, -0.1011, -0.1084, -0.1184, -0.1166, -0.1166, -0.1166, -0.1358, -0.1321, -0.0956, -0.0883, -0.091, -0.0819, -0.0453, -0.0271, -0.0413, -0.055, -0.0596, -0.0413, -0.0321, -0.0505, -0.0321, -0.0642, -0.0367, -0.0183, -0.0138, -0.0092, -0.0275, -0.0046, -0.0046, 0.0, -0.0046, -0.0092, -0.0092, 0.0, 0.0092, 0.0092, 0.0046, 0.0183, 0.0367, 0.0367, 0.0596, 0.055, 0.0367, 0.0505, 0.0505, 0.0505, 0.0505, 0.0413, 0.0596, 0.0642, 0.0642, 0.0413, 0.055, 0.0321, 0.0826, 0.078, 0.0826, 0.0826, 0.1009, 0.078, 0.0826, 0.0826, 0.0872, 0.0413, 0.055, 0.0413, 0.0734, 0.078, 0.0917, 0.0642, 0.0642, 0.0688, 0.0642, 0.0642, 0.0642, 0.0826, 0.0826, 0.1009, 0.1239, 0.1376, 0.156, 0.1514, 0.179, 0.1652, 0.1836, 0.1652, 0.1928, 0.2343, 0.2343, 0.2159, 0.1974, 0.1974, 0.202, 0.2205, 0.2573, 0.2895, 0.2895, 0.3218, 0.3034, 0.3264, 0.3264, 0.3034, 0.3126, 0.3494, 0.3678, 0.3356, 0.3632, 0.3632, 0.3448, 0.3356, 0.3356, 0.3632, 0.3586, 0.3863, 0.3863, 0.3817, 0.3909, 0.3863, 0.3448, 0.3494, 0.3448, 0.3586, 0.3494, 0.3586, 0.3448, 0.3172, 0.331, 0.2941, 0.2849, 0.3402, 0.2757, 0.2665, 0.3034, 0.3264, 0.3218, 0.3264, 0.2988, 0.3172, 0.3356, 0.331, 0.3448, 0.377, 0.3632, 0.3863, 0.3585, 0.3678, 0.34, 0.3262, 0.3216, 0.3216, 0.3216, 0.3539, 0.377, 0.3816, 0.3816, 0.3955]}
{"unit_id": "orig:1940901032256852150", "ticker_idx": 1, "ts": "2025-07-03T22:30:40+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "will support the NVIDIA and AI ASIC supply chains", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish survey: TSMC CoWoS +33% in 2026 supports NVDA/AI ASICs; Alchip sole Trainium3 provider with strong 2026 revenue growth; Marvell growth tied to Trainium2 lifecycle.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Survey: CoWoS Expansion Drives Computing Chip Growth, B30 Export to China a Key Variable\n\nMorgan Stanley (C. Chan, 25/06/30)\n\nMorgan Stanley's survey indicates that TSMC's CoWoS capacity is expected to grow by over 30% in 2026, which will support the NVIDIA and AI ASIC supply chains. Chinese AI demand is strong but constrained by hardware supply, and whether B30 GPUs can be exported to China remains the biggest uncertain factor. Meanwhile, Asian ASIC design service providers like Alchip are showing positive growth prospects due to AWS custom chip orders.\n\nTSMC CoWoS Capacity to Grow 33% in 2026, Driving GPU and ASIC Chip Demand\n\nMorgan Stanley has upgraded its forecast for TSMC's CoWoS capacity in 2026 to 90,000-95,000 wafers, an increase of approximately 33% from the 70,000 wafers at the end of 2025. Among these, CoWoS-L capacity may increase to 68,000 wafers, reflecting strong demand for Blackwell or Rubin chips. Concurrently, the testing time for Blackwell chips is expected to increase to 800-900 seconds, verifying the rising complexity of testing, with KYEC maintaining its leading share.\n\nStrong Demand from Chinese AI Developers, But Limited by B30 GPU Supply and Training Capacity Bottlenecks\n\nAlthough Chinese developers are aware of NVIDIA's B30 chip (RTX 6000 Pro), there is currently no official order confirmation. Taiwanese supply chain sources indicate that 2 million B30 chips are expected to begin tape-out in the second half of 2025. If export to China is not possible, some developers plan to switch to Huawei's 910C chip, but the latter is not yet readily available. Some developers are delaying capital expenditures until the availability of B30 is clear, while local ASICs can currently only be used for inference tasks.\n\nDelay in China's Next-Generation Large Model Training Attributed to Restricted GPU Resources\n\nMorgan Stanley's survey found that progress in training large AI models in China is lagging, mainly due to a shortage of GPU training compute power. While inference demand is more active, training resources remain a bottleneck, impacting the overall pace of AI deployment.\n\nBullish on Asian ASIC Design Service Providers, Alchip Expected to Benefit from AWS Trainium3 Orders\n\nFor AWS's custom chip Trainium3 project, Alchip is the sole service provider, and is expected to see significant revenue growth in 2026. The actual scale will depend on TSMC's 3nm and CoWoS capacity allocation. Marvell, on the other hand, is focusing on \"XPU-attach\" chips, and its growth will depend on the lifespan of Trainium2 and the shipment volume of its attached chips. Morgan Stanley expects the design service provider for Trainium4 to be selected in July, with Alchip having a higher chance of winning.\n\nInvestors Generally Positive on AI, But Focus Shifts to Demand Sustainability and Alternative Investments\n\nIn this survey, investors are generally optimistic about the AI sector, but their focus has shifted to the sustainability of future demand and the selection of investment targets beyond NVIDIA. This reflects the market's further discernment and positioning regarding the return on AI capital expenditure and beneficiaries in the mid-stream of the industry chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01311665584850008, "ret_signed_m1d": -0.01311665584850008, "alpha_spy_m1d": -0.005296900404353155, "alpha_c_m1d": -0.004760325156265233, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.029873272768840442, "ret_signed_p1w": 0.029873272768840442, "alpha_spy_p1w": 0.029105707528501945, "alpha_c_p1w": 0.016122564714986964, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.09024726239087122, "ret_signed_p1m": 0.09024726239087122, "alpha_spy_p1m": 0.09603628707412748, "alpha_c_p1m": 0.08908372267423093, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.17102118224793927, "ret_signed_p3m": 0.17102118224793927, "alpha_spy_p3m": 0.10275925686004461, "alpha_c_p3m": 0.020326706622261437, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.19587892503340987, "ret_signed_p6m": 0.19587892503340987, "alpha_spy_p6m": 0.08565210743783802, "alpha_c_p6m": -0.09810028581466601, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.3873, -0.3957, -0.2825, -0.3249, -0.3039, -0.3052, -0.2959, -0.3443, -0.3631, -0.3631, -0.3918, -0.3794, -0.3554, -0.3321, -0.3034, -0.3177, -0.3159, -0.3165, -0.2996, -0.2815, -0.2857, -0.2875, -0.2654, -0.2635, -0.268, -0.2281, -0.1846, -0.1507, -0.1539, -0.1503, -0.1492, -0.1567, -0.1729, -0.1664, -0.1761, -0.1761, -0.1497, -0.154, -0.1265, -0.152, -0.1379, -0.1138, -0.1094, -0.1215, -0.1106, -0.1049, -0.0966, -0.1036, -0.09, -0.109, -0.0919, -0.0955, -0.087, -0.087, -0.0972, -0.0952, -0.0718, -0.0316, -0.0271, -0.01, -0.0085, -0.0379, -0.0131, 0.0, 0.0, -0.0069, 0.0041, 0.0222, 0.0299, 0.035, 0.0297, 0.0713, 0.0755, 0.0857, 0.082, 0.0756, 0.0483, 0.0718, 0.0904, 0.0889, 0.1093, 0.1015, 0.1251, 0.1163, 0.0902, 0.1297, 0.1187, 0.126, 0.1345, 0.1466, 0.1426, 0.1495, 0.1396, 0.1423, 0.1325, 0.1423, 0.1023, 0.1008, 0.0982, 0.117, 0.1285, 0.1408, 0.1397, 0.1307, 0.0931, 0.0931, 0.0718, 0.0708, 0.0773, 0.0482, 0.0563, 0.0717, 0.1129, 0.112, 0.116, 0.1156, 0.0976, 0.0688, 0.1061, 0.1088, 0.1524, 0.1199, 0.1107, 0.1152, 0.1184, 0.1413, 0.171, 0.1752, 0.1855, 0.1775, 0.1645, 0.1614, 0.1869, 0.2086, 0.1496, 0.1819, 0.1299, 0.1287, 0.1411, 0.1499, 0.1463, 0.137, 0.1315, 0.1433, 0.169, 0.2018, 0.2617, 0.2994, 0.2734, 0.2709, 0.2984, 0.247, 0.2252, 0.1804, 0.1809, 0.2493, 0.2123, 0.2163, 0.1728, 0.1936, 0.1711, 0.1383, 0.1706, 0.1337, 0.1227, 0.1457, 0.116, 0.1314, 0.1314, 0.1109, 0.1292, 0.1389, 0.1272, 0.151, 0.1449, 0.1646, 0.161, 0.1535, 0.1356, 0.0985, 0.1065, 0.1155, 0.0729, 0.093, 0.136, 0.1529, 0.1876, 0.1838, 0.1838, 0.1959]}
{"unit_id": "orig:1940901032256852150", "ticker_idx": 2, "ts": "2025-07-03T22:30:40+00:00", "symbol": "Alchip", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Alchip is the sole service provider, and is expected to see significant revenue growth in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish survey: TSMC CoWoS +33% in 2026 supports NVDA/AI ASICs; Alchip sole Trainium3 provider with strong 2026 revenue growth; Marvell growth tied to Trainium2 lifecycle.", "resolved_tickers": ["3661.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Alchip Technologies 3661.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Survey: CoWoS Expansion Drives Computing Chip Growth, B30 Export to China a Key Variable\n\nMorgan Stanley (C. Chan, 25/06/30)\n\nMorgan Stanley's survey indicates that TSMC's CoWoS capacity is expected to grow by over 30% in 2026, which will support the NVIDIA and AI ASIC supply chains. Chinese AI demand is strong but constrained by hardware supply, and whether B30 GPUs can be exported to China remains the biggest uncertain factor. Meanwhile, Asian ASIC design service providers like Alchip are showing positive growth prospects due to AWS custom chip orders.\n\nTSMC CoWoS Capacity to Grow 33% in 2026, Driving GPU and ASIC Chip Demand\n\nMorgan Stanley has upgraded its forecast for TSMC's CoWoS capacity in 2026 to 90,000-95,000 wafers, an increase of approximately 33% from the 70,000 wafers at the end of 2025. Among these, CoWoS-L capacity may increase to 68,000 wafers, reflecting strong demand for Blackwell or Rubin chips. Concurrently, the testing time for Blackwell chips is expected to increase to 800-900 seconds, verifying the rising complexity of testing, with KYEC maintaining its leading share.\n\nStrong Demand from Chinese AI Developers, But Limited by B30 GPU Supply and Training Capacity Bottlenecks\n\nAlthough Chinese developers are aware of NVIDIA's B30 chip (RTX 6000 Pro), there is currently no official order confirmation. Taiwanese supply chain sources indicate that 2 million B30 chips are expected to begin tape-out in the second half of 2025. If export to China is not possible, some developers plan to switch to Huawei's 910C chip, but the latter is not yet readily available. Some developers are delaying capital expenditures until the availability of B30 is clear, while local ASICs can currently only be used for inference tasks.\n\nDelay in China's Next-Generation Large Model Training Attributed to Restricted GPU Resources\n\nMorgan Stanley's survey found that progress in training large AI models in China is lagging, mainly due to a shortage of GPU training compute power. While inference demand is more active, training resources remain a bottleneck, impacting the overall pace of AI deployment.\n\nBullish on Asian ASIC Design Service Providers, Alchip Expected to Benefit from AWS Trainium3 Orders\n\nFor AWS's custom chip Trainium3 project, Alchip is the sole service provider, and is expected to see significant revenue growth in 2026. The actual scale will depend on TSMC's 3nm and CoWoS capacity allocation. Marvell, on the other hand, is focusing on \"XPU-attach\" chips, and its growth will depend on the lifespan of Trainium2 and the shipment volume of its attached chips. Morgan Stanley expects the design service provider for Trainium4 to be selected in July, with Alchip having a higher chance of winning.\n\nInvestors Generally Positive on AI, But Focus Shifts to Demand Sustainability and Alternative Investments\n\nIn this survey, investors are generally optimistic about the AI sector, but their focus has shifted to the sustainability of future demand and the selection of investment targets beyond NVIDIA. This reflects the market's further discernment and positioning regarding the return on AI capital expenditure and beneficiaries in the mid-stream of the industry chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00472437253202318, "ret_signed_m1d": 0.00472437253202318, "alpha_spy_m1d": 0.012544127976170105, "alpha_c_m1d": 0.013080703224258028, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00472437253202318, "ret_signed_p1d": 0.00472437253202318, "alpha_spy_p1d": 0.00472437253202318, "alpha_c_p1d": 0.00472437253202318, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.0897637767160151, "ret_signed_p1w": 0.0897637767160151, "alpha_spy_p1w": 0.0889962114756766, "alpha_c_p1w": 0.07601306866216162, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.2236220334308563, "ret_signed_p1m": 0.2236220334308563, "alpha_spy_p1m": 0.22941105811411255, "alpha_c_p1m": 0.222458493714216, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.10167620020199064, "ret_signed_p3m": 0.10167620020199064, "alpha_spy_p3m": 0.03341427481409598, "alpha_c_p3m": -0.049018275423687196, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.022190182871399733, "ret_signed_p6m": 0.022190182871399733, "alpha_spy_p6m": -0.08803663472417211, "alpha_c_p6m": -0.27178902797667615, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.2457, -0.3197, -0.3874, -0.3276, -0.3307, -0.2976, -0.2567, -0.3165, -0.3024, -0.3055, -0.337, -0.3717, -0.3433, -0.3354, -0.3197, -0.3102, -0.315, -0.3307, -0.3307, -0.2756, -0.2835, -0.2835, -0.2976, -0.2614, -0.2756, -0.2031, -0.1512, -0.1323, -0.1433, -0.1402, -0.1795, -0.0976, -0.0614, -0.0866, -0.0646, -0.1008, -0.0992, -0.1008, -0.115, -0.115, -0.1433, -0.1433, -0.1165, -0.126, -0.1197, -0.1417, -0.1323, -0.1055, -0.1244, -0.1701, -0.1701, -0.1134, -0.1165, -0.0961, -0.1181, -0.1087, -0.1276, -0.0929, -0.0929, -0.0551, -0.0252, 0.0031, 0.0047, 0.0, 0.0047, 0.0142, 0.0126, 0.0598, 0.0898, 0.0803, 0.0425, 0.0614, 0.1339, 0.1827, 0.2378, 0.1906, 0.1717, 0.2047, 0.252, 0.2236, 0.222, 0.2016, 0.1811, 0.2236, 0.2236, 0.1827, 0.1811, 0.1685, 0.1858, 0.2094, 0.2315, 0.2835, 0.2976, 0.3386, 0.3559, 0.3323, 0.2709, 0.1528, 0.1921, 0.1622, 0.2268, 0.2535, 0.285, 0.2598, 0.2772, 0.2331, 0.2225, 0.2654, 0.2241, 0.2622, 0.2638, 0.2448, 0.2527, 0.2479, 0.205, 0.1843, 0.17, 0.1955, 0.1828, 0.1923, 0.1955, 0.1764, 0.0842, 0.0715, 0.0413, 0.0413, 0.1017, 0.0794, 0.0444, 0.0524, 0.0524, 0.1001, 0.0953, 0.1176, 0.1176, 0.0715, -0.0223, -0.0207, -0.016, -0.0191, 0.0285, 0.0667, 0.0238, 0.0015, 0.0015, -0.0255, -0.0255, 0.0333, 0.0667, 0.1112, 0.0969, 0.0953, 0.0619, 0.1669, 0.1335, 0.1812, 0.1557, 0.1192, 0.1208, 0.116, 0.1478, 0.0333, 0.0556, -0.0223, -0.0493, -0.0557, -0.0621, -0.0112, 0.0444, 0.0508, 0.046, 0.0874, 0.0142, 0.0079, 0.0031, -0.008, 0.0365, 0.0715, 0.0476, 0.0238, 0.0079, -0.0414, 0.0174, 0.0079, 0.0222, 0.0254, 0.0238, 0.0111, 0.0111, 0.0222]}
{"unit_id": "orig:1940901032256852150", "ticker_idx": 3, "ts": "2025-07-03T22:30:40+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell, on the other hand, is focusing on 'XPU-attach' chips, and its growth will depend on the lifespan of Trainium2 and the shipment volume of its attached chips", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish survey: TSMC CoWoS +33% in 2026 supports NVDA/AI ASICs; Alchip sole Trainium3 provider with strong 2026 revenue growth; Marvell growth tied to Trainium2 lifecycle.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Survey: CoWoS Expansion Drives Computing Chip Growth, B30 Export to China a Key Variable\n\nMorgan Stanley (C. Chan, 25/06/30)\n\nMorgan Stanley's survey indicates that TSMC's CoWoS capacity is expected to grow by over 30% in 2026, which will support the NVIDIA and AI ASIC supply chains. Chinese AI demand is strong but constrained by hardware supply, and whether B30 GPUs can be exported to China remains the biggest uncertain factor. Meanwhile, Asian ASIC design service providers like Alchip are showing positive growth prospects due to AWS custom chip orders.\n\nTSMC CoWoS Capacity to Grow 33% in 2026, Driving GPU and ASIC Chip Demand\n\nMorgan Stanley has upgraded its forecast for TSMC's CoWoS capacity in 2026 to 90,000-95,000 wafers, an increase of approximately 33% from the 70,000 wafers at the end of 2025. Among these, CoWoS-L capacity may increase to 68,000 wafers, reflecting strong demand for Blackwell or Rubin chips. Concurrently, the testing time for Blackwell chips is expected to increase to 800-900 seconds, verifying the rising complexity of testing, with KYEC maintaining its leading share.\n\nStrong Demand from Chinese AI Developers, But Limited by B30 GPU Supply and Training Capacity Bottlenecks\n\nAlthough Chinese developers are aware of NVIDIA's B30 chip (RTX 6000 Pro), there is currently no official order confirmation. Taiwanese supply chain sources indicate that 2 million B30 chips are expected to begin tape-out in the second half of 2025. If export to China is not possible, some developers plan to switch to Huawei's 910C chip, but the latter is not yet readily available. Some developers are delaying capital expenditures until the availability of B30 is clear, while local ASICs can currently only be used for inference tasks.\n\nDelay in China's Next-Generation Large Model Training Attributed to Restricted GPU Resources\n\nMorgan Stanley's survey found that progress in training large AI models in China is lagging, mainly due to a shortage of GPU training compute power. While inference demand is more active, training resources remain a bottleneck, impacting the overall pace of AI deployment.\n\nBullish on Asian ASIC Design Service Providers, Alchip Expected to Benefit from AWS Trainium3 Orders\n\nFor AWS's custom chip Trainium3 project, Alchip is the sole service provider, and is expected to see significant revenue growth in 2026. The actual scale will depend on TSMC's 3nm and CoWoS capacity allocation. Marvell, on the other hand, is focusing on \"XPU-attach\" chips, and its growth will depend on the lifespan of Trainium2 and the shipment volume of its attached chips. Morgan Stanley expects the design service provider for Trainium4 to be selected in July, with Alchip having a higher chance of winning.\n\nInvestors Generally Positive on AI, But Focus Shifts to Demand Sustainability and Alternative Investments\n\nIn this survey, investors are generally optimistic about the AI sector, but their focus has shifted to the sustainability of future demand and the selection of investment targets beyond NVIDIA. This reflects the market's further discernment and positioning regarding the return on AI capital expenditure and beneficiaries in the mid-stream of the industry chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01237028010898078, "ret_signed_m1d": -0.01237028010898078, "alpha_spy_m1d": -0.004550524664833855, "alpha_c_m1d": -0.004013949416745932, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.02420847780573243, "ret_signed_p1w": -0.02420847780573243, "alpha_spy_p1w": -0.024976043046070928, "alpha_c_p1w": -0.03795918585958591, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.008899584580905007, "ret_signed_p1m": -0.008899584580905007, "alpha_spy_p1m": -0.003110559897648746, "alpha_c_p1m": -0.0100631242975453, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.11916478629956484, "ret_signed_p3m": 0.11916478629956484, "alpha_spy_p3m": 0.05090286091167018, "alpha_c_p3m": -0.031529689326113, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.15014472254939615, "ret_signed_p6m": 0.15014472254939615, "alpha_spy_p6m": 0.0399179049538243, "alpha_c_p6m": -0.14383448829867973, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.3224, -0.3353, -0.1901, -0.2976, -0.2898, -0.3049, -0.2908, -0.3093, -0.3123, -0.3123, -0.3432, -0.3267, -0.2848, -0.2374, -0.2163, -0.2192, -0.2195, -0.2236, -0.1893, -0.1709, -0.1756, -0.1857, -0.251, -0.2334, -0.2066, -0.1421, -0.129, -0.1234, -0.1327, -0.1519, -0.1679, -0.183, -0.2005, -0.1773, -0.1927, -0.1927, -0.1511, -0.1409, -0.1523, -0.1994, -0.1824, -0.1705, -0.1181, -0.1333, -0.0908, -0.0803, -0.0843, -0.0923, -0.0737, -0.1063, -0.0633, -0.069, -0.0031, -0.0031, -0.0222, -0.0585, 0.0004, 0.01, 0.0637, 0.0263, 0.0295, 0.0141, -0.0124, 0.0, 0.0, -0.0483, -0.043, -0.0388, -0.0242, -0.0321, -0.0347, -0.0361, -0.0568, -0.0414, -0.0062, -0.0274, -0.0416, -0.0246, -0.0144, -0.0121, 0.0105, 0.0163, 0.0881, 0.0699, -0.0089, 0.0188, 0.0201, 0.0027, 0.0097, 0.0296, 0.0288, 0.0358, 0.0559, 0.0522, 0.0143, 0.0216, -0.0406, -0.0519, -0.052, -0.0282, -0.0289, -0.0114, -0.0044, 0.0281, -0.1631, -0.1631, -0.14, -0.1705, -0.1467, -0.1569, -0.1214, -0.1102, -0.1067, -0.1135, -0.1034, -0.1024, -0.0833, -0.0551, -0.0118, -0.0114, 0.0055, -0.0066, 0.0662, 0.1157, 0.1072, 0.0968, 0.1192, 0.1168, 0.1475, 0.1478, 0.1837, 0.1578, 0.2314, 0.2072, 0.1404, 0.1908, 0.1485, 0.1841, 0.1753, 0.1716, 0.1435, 0.1224, 0.0797, 0.1026, 0.1207, 0.1817, 0.1784, 0.2009, 0.1799, 0.2487, 0.2038, 0.1668, 0.2375, 0.2433, 0.2112, 0.2419, 0.19, 0.19, 0.1659, 0.1516, 0.1116, 0.0481, 0.0833, 0.0215, 0.0317, 0.1162, 0.1114, 0.1685, 0.1685, 0.1909, 0.2136, 0.2374, 0.3348, 0.308, 0.3176, 0.2255, 0.1842, 0.2318, 0.1913, 0.1247, 0.1224, 0.1199, 0.0883, 0.1252, 0.1202, 0.1296, 0.168, 0.1521, 0.1521, 0.1501]}
{"unit_id": "orig:1940900646758334844", "ticker_idx": 0, "ts": "2025-07-03T22:29:09+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintaining KLA as the Biggest Beneficiary of GAA, Shifting to a Positive Stance on AMAT for 2025", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley upgrades 2025 WFE to +6%, raises KLA/AMAT/LAM revenue forecasts; KLA is top GAA beneficiary, AMAT stance turns positive on TSMC share recovery and China demand.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: H2 Equipment Demand Rebounds Beyond Expectations, Driving Up Global WFE Forecast\n\nMorgan Stanley (S. Brett, 25/07/01)\n\nMorgan Stanley has upgraded its 2025 global WFE (Wafer Fab Equipment) forecast, primarily due to strong demand from China in the second half of the year, especially in the Foundry/Logic sector. Concurrently, it has raised its 2025 revenue forecasts for U.S. SPE (Semiconductor Production Equipment) manufacturers (KLA, AMAT, LAM), benefiting from TSMC's GAA production lines and increased equipment imports by China.\n\n2025 Global WFE Forecast Upgraded from +2% YoY to +6%, China is the Main Contributor\n\nMorgan Stanley has raised its 2025 global WFE market size forecast from $104 billion to $109 billion, increasing the year-on-year growth from 2% to 6%. The 2026 forecast has been slightly increased to $110 billion. This upgrade is primarily attributed to the Chinese market, with approximately $5 billion in new equipment spending, particularly in Foundry/Logic.\n\nChina's H2 WFE Spending Expected to Improve Significantly, Equipment Imports to Resume from July\n\nDespite the lower capacity utilization rates of most Chinese manufacturers, Morgan Stanley notes that policy-driven investment momentum will be further unleashed in the second half of 2025. From January to May, China's semiconductor equipment imports decreased by 4% year-on-year, mainly from Japan, the US, Netherlands, and Singapore, with only Malaysia showing positive growth. However, a rebound is expected starting in July, and the second half trend will replicate the H/H (half-on-half) structure of 2024 (19% growth in 2H).\n\nUpgrading KLA and AMAT's 2025 Revenue Forecasts for TSMC Projects and the Chinese Market\n\nMorgan Stanley has greater confidence in KLA's leading position in GAA (Gate-All-Around) processes, raising its 2025/26 revenue forecasts from TSMC to $2.76 billion and $3.05 billion, respectively. AMAT also gained back market share at TSMC, with its 2025/26 revenue forecasts upgraded from $5.6/$5.26 billion to $5.93/$5.38 billion. During the same period, the expected share of revenue from the Chinese market for KLA/AMAT in 2025 was also slightly raised to 32%/28%.\n\nMaintaining KLA as the Biggest Beneficiary of GAA, Shifting to a Positive Stance on AMAT for 2025\n\nWhile still cautious about AMAT's DRAM-related prospects for 2026, Morgan Stanley's outlook for AMAT in 2025 has turned more positive, driven by the regaining of TSMC share and the recovery of equipment demand from China. At the same time, it continues to emphasize that KLA possesses the strongest competitiveness in the GAA development trend and will be the biggest beneficiary of this process evolution.\n\nOverall Semiconductor Equipment Spending Momentum Driven by China, U.S. Manufacturers Directly Benefiting\n\nMorgan Stanley points out that the key to the global WFE rebound in 2025 comes from China's push for inventory replenishment in the second half and new project drives. U.S. SPE manufacturers will be the most direct beneficiaries, with TSMC and mainland China forming a dual pulling factor. The firm still notes that the uncertainty of the future DRAM cycle could cause disruptions in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0037639243675235967, "ret_signed_m1d": -0.0037639243675235967, "alpha_spy_m1d": 0.0040558310766233285, "alpha_c_m1d": 0.004592406324711251, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.004369452975262256, "ret_signed_p1w": 0.004369452975262256, "alpha_spy_p1w": 0.003601887734923759, "alpha_c_p1w": -0.009381255078591222, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.04103482806017178, "ret_signed_p1m": -0.04103482806017178, "alpha_spy_p1m": -0.03524580337691552, "alpha_c_p1m": -0.042198367776812074, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.1691225323081469, "ret_signed_p3m": 0.1691225323081469, "alpha_spy_p3m": 0.10086060692025223, "alpha_c_p3m": 0.01842805668246905, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.389318751747356, "ret_signed_p6m": 0.389318751747356, "alpha_spy_p6m": 0.2790919341517841, "alpha_c_p6m": 0.0953395408992801, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.3476, -0.3531, -0.2411, -0.2914, -0.2768, -0.2771, -0.2704, -0.3071, -0.3155, -0.3155, -0.3302, -0.3134, -0.291, -0.2561, -0.2505, -0.2515, -0.2577, -0.2418, -0.2703, -0.2467, -0.252, -0.2669, -0.2508, -0.2411, -0.2434, -0.1795, -0.1412, -0.1314, -0.1303, -0.1463, -0.1488, -0.1452, -0.1575, -0.1699, -0.1811, -0.1811, -0.1466, -0.1591, -0.1629, -0.1814, -0.1754, -0.1609, -0.1541, -0.1435, -0.1261, -0.1031, -0.0735, -0.0569, -0.0536, -0.0616, -0.0348, -0.0342, -0.0578, -0.0578, -0.0807, -0.0739, -0.0384, -0.0337, -0.0234, -0.0375, -0.0312, -0.0278, -0.0038, 0.0, 0.0, -0.0129, -0.0058, -0.0015, 0.0044, 0.0, -0.003, 0.0129, 0.0096, 0.0135, 0.0071, 0.0143, -0.035, -0.0297, -0.0221, -0.0243, -0.0015, -0.0092, 0.0004, -0.0493, -0.041, -0.0097, -0.0445, -0.0393, -0.0135, -0.0106, -0.0156, 0.0118, 0.0269, 0.0333, -0.0537, -0.0425, -0.0504, -0.0478, -0.0544, -0.0567, -0.0466, -0.0374, -0.0365, -0.031, -0.0548, -0.0548, -0.0826, -0.0853, -0.0534, -0.0189, -0.0147, -0.0052, 0.0109, 0.0398, 0.0449, 0.0719, 0.0737, 0.0729, 0.1345, 0.1325, 0.1612, 0.1611, 0.1584, 0.148, 0.1536, 0.1534, 0.1691, 0.2236, 0.2349, 0.194, 0.2354, 0.1758, 0.1518, 0.1419, 0.0652, 0.111, 0.1118, 0.1782, 0.191, 0.1995, 0.2496, 0.2437, 0.2078, 0.2563, 0.2821, 0.3171, 0.3073, 0.339, 0.3163, 0.3102, 0.3215, 0.2937, 0.3301, 0.3076, 0.2935, 0.3202, 0.2908, 0.2996, 0.2592, 0.2295, 0.2309, 0.2194, 0.2676, 0.197, 0.1912, 0.2342, 0.2441, 0.2585, 0.2585, 0.2763, 0.2564, 0.2919, 0.3157, 0.3117, 0.3186, 0.3296, 0.3307, 0.3451, 0.353, 0.2963, 0.3302, 0.3283, 0.2725, 0.3272, 0.3525, 0.3742, 0.3775, 0.3865, 0.3865, 0.3893]}
{"unit_id": "orig:1940900646758334844", "ticker_idx": 1, "ts": "2025-07-03T22:29:09+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley's outlook for AMAT in 2025 has turned more positive, driven by the regaining of TSMC share and the recovery of equipment demand from China", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley upgrades 2025 WFE to +6%, raises KLA/AMAT/LAM revenue forecasts; KLA is top GAA beneficiary, AMAT stance turns positive on TSMC share recovery and China demand.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: H2 Equipment Demand Rebounds Beyond Expectations, Driving Up Global WFE Forecast\n\nMorgan Stanley (S. Brett, 25/07/01)\n\nMorgan Stanley has upgraded its 2025 global WFE (Wafer Fab Equipment) forecast, primarily due to strong demand from China in the second half of the year, especially in the Foundry/Logic sector. Concurrently, it has raised its 2025 revenue forecasts for U.S. SPE (Semiconductor Production Equipment) manufacturers (KLA, AMAT, LAM), benefiting from TSMC's GAA production lines and increased equipment imports by China.\n\n2025 Global WFE Forecast Upgraded from +2% YoY to +6%, China is the Main Contributor\n\nMorgan Stanley has raised its 2025 global WFE market size forecast from $104 billion to $109 billion, increasing the year-on-year growth from 2% to 6%. The 2026 forecast has been slightly increased to $110 billion. This upgrade is primarily attributed to the Chinese market, with approximately $5 billion in new equipment spending, particularly in Foundry/Logic.\n\nChina's H2 WFE Spending Expected to Improve Significantly, Equipment Imports to Resume from July\n\nDespite the lower capacity utilization rates of most Chinese manufacturers, Morgan Stanley notes that policy-driven investment momentum will be further unleashed in the second half of 2025. From January to May, China's semiconductor equipment imports decreased by 4% year-on-year, mainly from Japan, the US, Netherlands, and Singapore, with only Malaysia showing positive growth. However, a rebound is expected starting in July, and the second half trend will replicate the H/H (half-on-half) structure of 2024 (19% growth in 2H).\n\nUpgrading KLA and AMAT's 2025 Revenue Forecasts for TSMC Projects and the Chinese Market\n\nMorgan Stanley has greater confidence in KLA's leading position in GAA (Gate-All-Around) processes, raising its 2025/26 revenue forecasts from TSMC to $2.76 billion and $3.05 billion, respectively. AMAT also gained back market share at TSMC, with its 2025/26 revenue forecasts upgraded from $5.6/$5.26 billion to $5.93/$5.38 billion. During the same period, the expected share of revenue from the Chinese market for KLA/AMAT in 2025 was also slightly raised to 32%/28%.\n\nMaintaining KLA as the Biggest Beneficiary of GAA, Shifting to a Positive Stance on AMAT for 2025\n\nWhile still cautious about AMAT's DRAM-related prospects for 2026, Morgan Stanley's outlook for AMAT in 2025 has turned more positive, driven by the regaining of TSMC share and the recovery of equipment demand from China. At the same time, it continues to emphasize that KLA possesses the strongest competitiveness in the GAA development trend and will be the biggest beneficiary of this process evolution.\n\nOverall Semiconductor Equipment Spending Momentum Driven by China, U.S. Manufacturers Directly Benefiting\n\nMorgan Stanley points out that the key to the global WFE rebound in 2025 comes from China's push for inventory replenishment in the second half and new project drives. U.S. SPE manufacturers will be the most direct beneficiaries, with TSMC and mainland China forming a dual pulling factor. The firm still notes that the uncertainty of the future DRAM cycle could cause disruptions in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005443754096110531, "ret_signed_m1d": -0.005443754096110531, "alpha_spy_m1d": 0.002376001348036394, "alpha_c_m1d": 0.0029125765961243166, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.036534772107009506, "ret_signed_p1w": 0.036534772107009506, "alpha_spy_p1w": 0.03576720686667101, "alpha_c_p1w": 0.022784064053156028, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.0578906235292016, "ret_signed_p1m": -0.0578906235292016, "alpha_spy_p1m": -0.05210159884594534, "alpha_c_p1m": -0.059054163245841895, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.0747279410214785, "ret_signed_p3m": 0.0747279410214785, "alpha_spy_p3m": 0.006466015633583844, "alpha_c_p3m": -0.07596653460419933, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.3774689097865285, "ret_signed_p6m": 0.3774689097865285, "alpha_spy_p6m": 0.2672420921909566, "alpha_c_p6m": 0.0834896989384526, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.3066, -0.3269, -0.2185, -0.2785, -0.2435, -0.2456, -0.2408, -0.2787, -0.2825, -0.2825, -0.2925, -0.2782, -0.2506, -0.2162, -0.209, -0.213, -0.2194, -0.2134, -0.2225, -0.1905, -0.193, -0.2013, -0.1868, -0.1858, -0.1878, -0.1232, -0.0969, -0.0911, -0.0879, -0.1358, -0.1339, -0.135, -0.1533, -0.1598, -0.1756, -0.1756, -0.1529, -0.1542, -0.1652, -0.1795, -0.1768, -0.1534, -0.1524, -0.1406, -0.1272, -0.1113, -0.0904, -0.0956, -0.084, -0.1071, -0.0759, -0.0888, -0.0953, -0.0953, -0.113, -0.0999, -0.0569, -0.0418, -0.0394, -0.041, -0.0418, -0.0382, -0.0054, 0.0, 0.0, -0.0014, 0.0206, 0.0227, 0.0365, 0.036, 0.0317, 0.0431, 0.0197, 0.0077, -0.0032, 0.0082, -0.0205, -0.0211, -0.0153, -0.0281, -0.0041, -0.0138, -0.0087, -0.0575, -0.0579, -0.0431, -0.0623, -0.0676, -0.0414, -0.0323, -0.0349, -0.0136, -0.0053, -0.0147, -0.1533, -0.144, -0.1509, -0.1575, -0.161, -0.1471, -0.1497, -0.1364, -0.1371, -0.1325, -0.1561, -0.1561, -0.1729, -0.1798, -0.1694, -0.1457, -0.1494, -0.1418, -0.1422, -0.1068, -0.1192, -0.1027, -0.089, -0.065, -0.0039, -0.0021, 0.0526, 0.0544, 0.0574, 0.0477, 0.0704, 0.0758, 0.0747, 0.143, 0.1737, 0.1419, 0.1754, 0.1105, 0.1418, 0.1564, 0.1021, 0.1521, 0.1453, 0.1946, 0.1954, 0.181, 0.1975, 0.1863, 0.1578, 0.1993, 0.2008, 0.2143, 0.1949, 0.2375, 0.2207, 0.2236, 0.2478, 0.2083, 0.2645, 0.2259, 0.2077, 0.234, 0.2003, 0.2112, 0.1718, 0.1864, 0.2006, 0.1817, 0.2343, 0.1583, 0.1782, 0.2145, 0.2752, 0.3147, 0.3147, 0.3267, 0.3399, 0.3955, 0.4129, 0.4171, 0.4096, 0.4104, 0.405, 0.4472, 0.4206, 0.3633, 0.3742, 0.3614, 0.3058, 0.3333, 0.3486, 0.3623, 0.3687, 0.3716, 0.3716, 0.3775]}
{"unit_id": "orig:1940900646758334844", "ticker_idx": 2, "ts": "2025-07-03T22:29:09+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it has raised its 2025 revenue forecasts for U.S. SPE (Semiconductor Production Equipment) manufacturers (KLA, AMAT, LAM)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley upgrades 2025 WFE to +6%, raises KLA/AMAT/LAM revenue forecasts; KLA is top GAA beneficiary, AMAT stance turns positive on TSMC share recovery and China demand.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: H2 Equipment Demand Rebounds Beyond Expectations, Driving Up Global WFE Forecast\n\nMorgan Stanley (S. Brett, 25/07/01)\n\nMorgan Stanley has upgraded its 2025 global WFE (Wafer Fab Equipment) forecast, primarily due to strong demand from China in the second half of the year, especially in the Foundry/Logic sector. Concurrently, it has raised its 2025 revenue forecasts for U.S. SPE (Semiconductor Production Equipment) manufacturers (KLA, AMAT, LAM), benefiting from TSMC's GAA production lines and increased equipment imports by China.\n\n2025 Global WFE Forecast Upgraded from +2% YoY to +6%, China is the Main Contributor\n\nMorgan Stanley has raised its 2025 global WFE market size forecast from $104 billion to $109 billion, increasing the year-on-year growth from 2% to 6%. The 2026 forecast has been slightly increased to $110 billion. This upgrade is primarily attributed to the Chinese market, with approximately $5 billion in new equipment spending, particularly in Foundry/Logic.\n\nChina's H2 WFE Spending Expected to Improve Significantly, Equipment Imports to Resume from July\n\nDespite the lower capacity utilization rates of most Chinese manufacturers, Morgan Stanley notes that policy-driven investment momentum will be further unleashed in the second half of 2025. From January to May, China's semiconductor equipment imports decreased by 4% year-on-year, mainly from Japan, the US, Netherlands, and Singapore, with only Malaysia showing positive growth. However, a rebound is expected starting in July, and the second half trend will replicate the H/H (half-on-half) structure of 2024 (19% growth in 2H).\n\nUpgrading KLA and AMAT's 2025 Revenue Forecasts for TSMC Projects and the Chinese Market\n\nMorgan Stanley has greater confidence in KLA's leading position in GAA (Gate-All-Around) processes, raising its 2025/26 revenue forecasts from TSMC to $2.76 billion and $3.05 billion, respectively. AMAT also gained back market share at TSMC, with its 2025/26 revenue forecasts upgraded from $5.6/$5.26 billion to $5.93/$5.38 billion. During the same period, the expected share of revenue from the Chinese market for KLA/AMAT in 2025 was also slightly raised to 32%/28%.\n\nMaintaining KLA as the Biggest Beneficiary of GAA, Shifting to a Positive Stance on AMAT for 2025\n\nWhile still cautious about AMAT's DRAM-related prospects for 2026, Morgan Stanley's outlook for AMAT in 2025 has turned more positive, driven by the regaining of TSMC share and the recovery of equipment demand from China. At the same time, it continues to emphasize that KLA possesses the strongest competitiveness in the GAA development trend and will be the biggest beneficiary of this process evolution.\n\nOverall Semiconductor Equipment Spending Momentum Driven by China, U.S. Manufacturers Directly Benefiting\n\nMorgan Stanley points out that the key to the global WFE rebound in 2025 comes from China's push for inventory replenishment in the second half and new project drives. U.S. SPE manufacturers will be the most direct beneficiaries, with TSMC and mainland China forming a dual pulling factor. The firm still notes that the uncertainty of the future DRAM cycle could cause disruptions in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0002024204646235983, "ret_signed_m1d": 0.0002024204646235983, "alpha_spy_m1d": 0.008022175908770524, "alpha_c_m1d": 0.008558751156858446, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.022771022110071648, "ret_signed_p1w": 0.022771022110071648, "alpha_spy_p1w": 0.02200345686973315, "alpha_c_p1w": 0.00902031405621817, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.02469382256034658, "ret_signed_p1m": -0.02469382256034658, "alpha_spy_p1m": -0.018904797877090318, "alpha_c_p1m": -0.025857362276986873, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.35780187358023463, "ret_signed_p3m": 0.35780187358023463, "alpha_spy_p3m": 0.28953994819233997, "alpha_c_p3m": 0.2071073979545568, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.8086770492046209, "ret_signed_p6m": 0.8086770492046209, "alpha_spy_p6m": 0.6984502316090491, "alpha_c_p6m": 0.514697838356545, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.3721, -0.3918, -0.2821, -0.3294, -0.3188, -0.3149, -0.3091, -0.3424, -0.3563, -0.3563, -0.3702, -0.3591, -0.3263, -0.2842, -0.279, -0.2775, -0.2842, -0.2765, -0.2759, -0.2477, -0.2539, -0.2614, -0.2422, -0.2439, -0.2392, -0.1697, -0.1394, -0.1422, -0.144, -0.1477, -0.1538, -0.1491, -0.1642, -0.1665, -0.1817, -0.1817, -0.1526, -0.152, -0.1504, -0.1844, -0.1673, -0.1554, -0.1442, -0.1428, -0.1283, -0.1086, -0.0813, -0.0818, -0.0747, -0.0963, -0.057, -0.0646, -0.0665, -0.0665, -0.0842, -0.0729, -0.0322, -0.0282, -0.0199, -0.0163, -0.0149, -0.0202, 0.0002, 0.0, 0.0, -0.0068, 0.0103, 0.0101, 0.0228, 0.0296, 0.0082, 0.0229, 0.0158, 0.02, 0.0187, 0.0297, -0.0113, -0.0173, -0.0104, -0.0187, -0.0019, 0.0013, 0.0028, -0.0402, -0.0247, -0.004, -0.0216, -0.029, 0.0034, 0.0298, 0.0323, 0.0655, 0.0803, 0.0867, 0.0071, 0.0007, 0.0154, 0.0034, -0.004, 0.0129, 0.025, 0.0488, 0.0492, 0.0534, 0.0136, 0.0136, -0.018, -0.0108, 0.0163, 0.0419, 0.0634, 0.0684, 0.0865, 0.1697, 0.1837, 0.2065, 0.2192, 0.2337, 0.2784, 0.2845, 0.3379, 0.3352, 0.3012, 0.2993, 0.3013, 0.3293, 0.3578, 0.448, 0.4905, 0.4786, 0.5124, 0.4232, 0.4454, 0.4298, 0.3321, 0.3975, 0.4025, 0.4681, 0.4437, 0.435, 0.4607, 0.4708, 0.4323, 0.4961, 0.5381, 0.591, 0.5781, 0.6293, 0.6327, 0.5967, 0.635, 0.5797, 0.6737, 0.6447, 0.6159, 0.6871, 0.6142, 0.6369, 0.5547, 0.5034, 0.4953, 0.4525, 0.5089, 0.4155, 0.4465, 0.5249, 0.5406, 0.5732, 0.5732, 0.5819, 0.5696, 0.6041, 0.6226, 0.5956, 0.6119, 0.653, 0.6842, 0.709, 0.7136, 0.6304, 0.6688, 0.6583, 0.5741, 0.6729, 0.7498, 0.7801, 0.7791, 0.8012, 0.8012, 0.8087]}
{"unit_id": "orig:1934792510901231713", "ticker_idx": 0, "ts": "2025-06-17T01:57:35+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "additional investments primarily in large-cap memory stocks are more attractive", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish large-cap Korean memory stocks (Samsung, Hynix) into 3Q25 on HBM demand extension, commodity DRAM shortage, and foreign buying momentum through 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Buying Semiconductors: Expectation Extended to 2026\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\nForeigners' semiconductor buying spree has been formidable recently. Last night, as tensions in macro-level conflicts eased, the Philadelphia Semiconductor Index surged by 3%, and this momentum is driving up the stock prices of Samsung Electronics and SK Hynix.\n\nWhat should we focus on and respond to?\n\n1. Reflecting Next Year's HBM Expectations\n\nThe outlook for HBM demand next year is becoming even more optimistic. Negotiations for volume between NVIDIA and SK Hynix are further strengthening confidence in strong demand for 2026. Based on the launch of AMD MI350 and performance expectations for MI400, which is anticipated to rival Rubin, AMD and Samsung Electronics' stock prices are rejoining the AI rally. We believe that upward revisions to 2026 operating profits will be more positively reflected in stock prices. Focusing on 2026 further alleviates valuation burdens for investors.\n\n2. Extension of Commodity Supply Shortages\n\nRecent consecutive discontinuations of DDR4 by CXMT and Micron, Samsung Electronics' aggressive D1c capacity expansion, and memory manufacturers' limited Capex are further strengthening the belief in the sustained shortage of commodity memory. Threat factors to supply expansion are being removed one by one. We project mobile DRAM prices to rise by an average of 10% in 3Q25, following 2Q.\n\n3. How Long Will This Last?\n\nThe core lies in belief in the next year and foreign buying momentum. Until recently, concerns about a \"peak-out\" dominated the market, but investment fever in data centers remains unabated. The period for investors to confirm relief for next year extends from the confirmation of full sales and tight supply from TSMC and SK Hynix in July to the strong demand from cloud companies to be verified in October. Sovereign AI is a catalyst stimulating the market in both Korea and globally. During this period, the market will adopt an even more optimistic view on future AI growth.\n\n4. On AI Peak-out and the 2-Year Infrastructure Cycle\n\nWe do not believe AI is a bubble. However, infrastructure investment is only the beginning of a long-term AI cycle we will face. Various 'other' AI themes will stimulate the stock market going forward. While we feel this 2-year cycle is being extended due to sovereign AI, tariffs, and conservative investment trends in an uncertain macro environment, we still believe that the infrastructure investment cycle fundamentally considers variables based on a 2-year cycle.\n\n5. Conclusion\n\nIt seems appropriate to adopt a more optimistic view for 3Q25. It's the same product, the same narrative, but for an extended period. The party continues until dawn. Considering foreign buying momentum and conservative investment, additional investments primarily in large-cap memory stocks are more attractive.\n\nThank you.\n\n(Published on 2025/06/17)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015490482807901329, "ret_signed_m1d": -0.015490482807901329, "alpha_spy_m1d": -0.024109162240655402, "alpha_c_m1d": -0.022617859418850306, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.007127376610948977, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029259952901696362, "ret_signed_p1d": 0.029259952901696362, "alpha_spy_p1d": 0.029410588655867542, "alpha_c_p1d": 0.0280583209562959, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.001201631945400461, "ret_p1w": 0.041308136605195944, "ret_signed_p1w": 0.041308136605195944, "alpha_spy_p1w": 0.022825811458643708, "alpha_c_p1w": 0.015078398643756596, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.026229737961439348, "ret_p1m": 0.12042785440673054, "ret_signed_p1m": 0.12042785440673054, "alpha_spy_p1m": 0.07267249761644168, "alpha_c_p1m": 0.04658911028508328, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.07383874412164726, "ret_p3m": 0.3057226443715446, "ret_signed_p3m": 0.3057226443715446, "alpha_spy_p3m": 0.20225768108048614, "alpha_c_p3m": 0.1813691908466124, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.12435345352493221, "ret_p6m": 0.8786103790463109, "ret_signed_p6m": 0.8786103790463109, "alpha_spy_p6m": 0.721322062316144, "alpha_c_p6m": 0.6423436312533779, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.236266747792933, "price_path": [0.03, 0.0557, 0.0352, 0.0232, 0.0506, 0.0574, 0.0361, -0.0052, 0.012, 0.012, -0.0086, -0.0344, -0.0843, -0.0792, -0.0878, -0.0293, -0.0499, -0.0327, -0.0258, -0.0585, -0.0516, -0.0482, -0.0465, -0.0534, -0.0413, -0.0413, -0.0413, -0.0396, -0.0396, -0.0448, -0.0448, -0.0654, -0.0654, -0.0654, -0.0602, -0.0602, -0.0568, -0.0086, -0.0207, -0.012, -0.0138, -0.0224, -0.0396, -0.0379, -0.0413, -0.0585, -0.0671, -0.0585, -0.0723, -0.0379, -0.0344, -0.0327, -0.0224, -0.0224, -0.0052, 0.0172, 0.0172, 0.0293, 0.0189, 0.031, 0.0241, 0.0034, -0.0155, 0.0, 0.0293, 0.0189, 0.0241, -0.0017, 0.0413, 0.0551, 0.0361, 0.0529, 0.0356, 0.0425, 0.0529, 0.1048, 0.0962, 0.0685, 0.0633, 0.046, 0.0564, 0.0841, 0.0823, 0.1031, 0.1204, 0.1551, 0.162, 0.1741, 0.1429, 0.1499, 0.1429, 0.1412, 0.2191, 0.2226, 0.2572, 0.2365, 0.1932, 0.207, 0.2105, 0.1914, 0.2209, 0.2434, 0.2295, 0.2313, 0.2451, 0.2399, 0.2399, 0.2122, 0.2122, 0.2209, 0.2226, 0.2365, 0.2382, 0.2174, 0.2226, 0.2053, 0.207, 0.1706, 0.1966, 0.2087, 0.2139, 0.2036, 0.2139, 0.2382, 0.2572, 0.2711, 0.3057, 0.3248, 0.375, 0.3542, 0.3906, 0.3906, 0.446, 0.4668, 0.4789, 0.491, 0.4425, 0.4646, 0.4594, 0.4959, 0.5612, 0.5612, 0.5612, 0.5612, 0.5612, 0.5612, 0.642, 0.6229, 0.5933, 0.6525, 0.6994, 0.7029, 0.7064, 0.696, 0.7151, 0.6786, 0.7186, 0.7742, 0.7308, 0.7482, 0.8108, 0.8699, 0.9325, 0.8247, 0.7499, 0.7255, 0.7029, 0.7499, 0.8003, 0.7934, 0.7882, 0.6907, 0.7499, 0.7012, 0.6786, 0.7499, 0.649, 0.6821, 0.7273, 0.7882, 0.8003, 0.7482, 0.7534, 0.7986, 0.8177, 0.8282, 0.8856, 0.9047, 0.8856, 0.8786]}
{"unit_id": "orig:1934792510901231713", "ticker_idx": 1, "ts": "2025-06-17T01:57:35+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Negotiations for volume between NVIDIA and SK Hynix are further strengthening confidence in strong demand for 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish large-cap Korean memory stocks (Samsung, Hynix) into 3Q25 on HBM demand extension, commodity DRAM shortage, and foreign buying momentum through 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Buying Semiconductors: Expectation Extended to 2026\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\nForeigners' semiconductor buying spree has been formidable recently. Last night, as tensions in macro-level conflicts eased, the Philadelphia Semiconductor Index surged by 3%, and this momentum is driving up the stock prices of Samsung Electronics and SK Hynix.\n\nWhat should we focus on and respond to?\n\n1. Reflecting Next Year's HBM Expectations\n\nThe outlook for HBM demand next year is becoming even more optimistic. Negotiations for volume between NVIDIA and SK Hynix are further strengthening confidence in strong demand for 2026. Based on the launch of AMD MI350 and performance expectations for MI400, which is anticipated to rival Rubin, AMD and Samsung Electronics' stock prices are rejoining the AI rally. We believe that upward revisions to 2026 operating profits will be more positively reflected in stock prices. Focusing on 2026 further alleviates valuation burdens for investors.\n\n2. Extension of Commodity Supply Shortages\n\nRecent consecutive discontinuations of DDR4 by CXMT and Micron, Samsung Electronics' aggressive D1c capacity expansion, and memory manufacturers' limited Capex are further strengthening the belief in the sustained shortage of commodity memory. Threat factors to supply expansion are being removed one by one. We project mobile DRAM prices to rise by an average of 10% in 3Q25, following 2Q.\n\n3. How Long Will This Last?\n\nThe core lies in belief in the next year and foreign buying momentum. Until recently, concerns about a \"peak-out\" dominated the market, but investment fever in data centers remains unabated. The period for investors to confirm relief for next year extends from the confirmation of full sales and tight supply from TSMC and SK Hynix in July to the strong demand from cloud companies to be verified in October. Sovereign AI is a catalyst stimulating the market in both Korea and globally. During this period, the market will adopt an even more optimistic view on future AI growth.\n\n4. On AI Peak-out and the 2-Year Infrastructure Cycle\n\nWe do not believe AI is a bubble. However, infrastructure investment is only the beginning of a long-term AI cycle we will face. Various 'other' AI themes will stimulate the stock market going forward. While we feel this 2-year cycle is being extended due to sovereign AI, tariffs, and conservative investment trends in an uncertain macro environment, we still believe that the infrastructure investment cycle fundamentally considers variables based on a 2-year cycle.\n\n5. Conclusion\n\nIt seems appropriate to adopt a more optimistic view for 3Q25. It's the same product, the same narrative, but for an extended period. The party continues until dawn. Considering foreign buying momentum and conservative investment, additional investments primarily in large-cap memory stocks are more attractive.\n\nThank you.\n\n(Published on 2025/06/17)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00401608475530546, "ret_signed_m1d": -0.00401608475530546, "alpha_spy_m1d": -0.012634764188059533, "alpha_c_m1d": -0.010858880095655943, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01004021188826365, "ret_signed_p1d": -0.01004021188826365, "alpha_spy_p1d": -0.00988957613409247, "alpha_c_p1d": -0.013862946356956618, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.11847387014928157, "ret_signed_p1w": 0.11847387014928157, "alpha_spy_p1w": 0.09999154500272933, "alpha_c_p1w": 0.0804372760037082, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.18875503830101237, "ret_signed_p1m": 0.18875503830101237, "alpha_spy_p1m": 0.1409996815107235, "alpha_c_p1m": 0.0829024138076131, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": 0.321182745593249, "ret_signed_p3m": 0.321182745593249, "alpha_spy_p3m": 0.21771778230219052, "alpha_c_p3m": 0.15783545041252056, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 1.3625255321702983, "ret_signed_p6m": 1.3625255321702983, "alpha_spy_p6m": 1.2052372154401314, "alpha_c_p6m": 0.9324249563143081, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [-0.1581, -0.1361, -0.1521, -0.1662, -0.1421, -0.1702, -0.201, -0.2355, -0.2103, -0.2067, -0.2199, -0.2696, -0.3393, -0.3205, -0.3385, -0.2656, -0.2752, -0.2776, -0.276, -0.3025, -0.2985, -0.2985, -0.292, -0.3033, -0.2744, -0.2852, -0.2608, -0.2704, -0.2752, -0.2884, -0.2884, -0.2544, -0.2544, -0.2544, -0.2351, -0.2371, -0.2379, -0.2183, -0.2042, -0.1742, -0.1962, -0.1802, -0.2006, -0.1902, -0.1962, -0.2107, -0.1982, -0.1862, -0.1882, -0.1662, -0.1486, -0.1787, -0.1667, -0.1667, -0.1265, -0.0984, -0.0984, -0.0803, -0.0743, -0.0361, -0.0542, -0.0542, -0.004, 0.0, -0.01, -0.012, 0.0321, 0.0422, 0.1185, 0.1486, 0.1767, 0.1406, 0.1727, 0.1466, 0.1205, 0.1185, 0.0863, 0.0884, 0.1325, 0.1285, 0.1928, 0.1827, 0.2048, 0.1988, 0.1888, 0.0823, 0.0803, 0.0944, 0.0783, 0.0803, 0.0823, 0.0683, 0.0522, 0.0542, 0.0582, 0.0984, 0.0361, 0.0361, 0.0582, 0.0382, 0.0522, 0.0301, 0.0723, 0.0803, 0.1165, 0.1104, 0.1104, 0.0743, 0.0562, 0.0261, -0.0161, 0.008, 0.0422, 0.0502, 0.0442, 0.0799, 0.0819, 0.0296, 0.0477, 0.0557, 0.0678, 0.1, 0.1141, 0.1583, 0.2226, 0.2347, 0.3212, 0.3312, 0.3996, 0.3413, 0.4298, 0.4298, 0.4117, 0.4519, 0.4378, 0.4338, 0.3534, 0.4036, 0.3976, 0.4479, 0.5906, 0.5906, 0.5906, 0.5906, 0.5906, 0.5906, 0.7214, 0.6691, 0.655, 0.6992, 0.8199, 0.8722, 0.9526, 0.9265, 0.9365, 0.9245, 1.0512, 1.1517, 1.0954, 1.2442, 1.2844, 1.2482, 1.4936, 1.3568, 1.3287, 1.385, 1.3327, 1.4373, 1.4895, 1.4815, 1.4614, 1.2522, 1.4373, 1.2925, 1.2603, 1.2965, 1.0954, 1.0914, 1.0873, 1.1075, 1.1895, 1.1331, 1.1653, 1.2458, 1.2217, 1.1814, 1.1895, 1.3223, 1.278, 1.3625]}
{"unit_id": "orig:1934792510901231713", "ticker_idx": 2, "ts": "2025-06-17T01:57:35+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "It seems appropriate to adopt a more optimistic view for 3Q25. The party continues until dawn.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish large-cap Korean memory stocks (Samsung, Hynix) into 3Q25 on HBM demand extension, commodity DRAM shortage, and foreign buying momentum through 2026.", "resolved_tickers": ["EWY"], "resolution_method": "etf", "resolution_reason": "Korea semiconductor sector proxied by EWY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Buying Semiconductors: Expectation Extended to 2026\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\nForeigners' semiconductor buying spree has been formidable recently. Last night, as tensions in macro-level conflicts eased, the Philadelphia Semiconductor Index surged by 3%, and this momentum is driving up the stock prices of Samsung Electronics and SK Hynix.\n\nWhat should we focus on and respond to?\n\n1. Reflecting Next Year's HBM Expectations\n\nThe outlook for HBM demand next year is becoming even more optimistic. Negotiations for volume between NVIDIA and SK Hynix are further strengthening confidence in strong demand for 2026. Based on the launch of AMD MI350 and performance expectations for MI400, which is anticipated to rival Rubin, AMD and Samsung Electronics' stock prices are rejoining the AI rally. We believe that upward revisions to 2026 operating profits will be more positively reflected in stock prices. Focusing on 2026 further alleviates valuation burdens for investors.\n\n2. Extension of Commodity Supply Shortages\n\nRecent consecutive discontinuations of DDR4 by CXMT and Micron, Samsung Electronics' aggressive D1c capacity expansion, and memory manufacturers' limited Capex are further strengthening the belief in the sustained shortage of commodity memory. Threat factors to supply expansion are being removed one by one. We project mobile DRAM prices to rise by an average of 10% in 3Q25, following 2Q.\n\n3. How Long Will This Last?\n\nThe core lies in belief in the next year and foreign buying momentum. Until recently, concerns about a \"peak-out\" dominated the market, but investment fever in data centers remains unabated. The period for investors to confirm relief for next year extends from the confirmation of full sales and tight supply from TSMC and SK Hynix in July to the strong demand from cloud companies to be verified in October. Sovereign AI is a catalyst stimulating the market in both Korea and globally. During this period, the market will adopt an even more optimistic view on future AI growth.\n\n4. On AI Peak-out and the 2-Year Infrastructure Cycle\n\nWe do not believe AI is a bubble. However, infrastructure investment is only the beginning of a long-term AI cycle we will face. Various 'other' AI themes will stimulate the stock market going forward. While we feel this 2-year cycle is being extended due to sovereign AI, tariffs, and conservative investment trends in an uncertain macro environment, we still believe that the infrastructure investment cycle fundamentally considers variables based on a 2-year cycle.\n\n5. Conclusion\n\nIt seems appropriate to adopt a more optimistic view for 3Q25. It's the same product, the same narrative, but for an extended period. The party continues until dawn. Considering foreign buying momentum and conservative investment, additional investments primarily in large-cap memory stocks are more attractive.\n\nThank you.\n\n(Published on 2025/06/17)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.023412915931918965, "ret_signed_m1d": 0.023412915931918965, "alpha_spy_m1d": 0.014794236499164892, "alpha_c_m1d": 0.014794236499164892, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.008618679432754073, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01861035356627805, "ret_signed_p1d": 0.01861035356627805, "alpha_spy_p1d": 0.01876098932044923, "alpha_c_p1d": 0.01876098932044923, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": -0.00015063575417118003, "ret_p1w": 0.09034977770836483, "ret_signed_p1w": 0.09034977770836483, "alpha_spy_p1w": 0.0718674525618126, "alpha_c_p1w": 0.0718674525618126, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.018482325146552236, "ret_p1m": 0.0968033416215126, "ret_signed_p1m": 0.0968033416215126, "alpha_spy_p1m": 0.049047984831223745, "alpha_c_p1m": 0.049047984831223745, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.04775535679028886, "ret_p3m": 0.1805494643669565, "ret_signed_p3m": 0.1805494643669565, "alpha_spy_p3m": 0.07708450107589804, "alpha_c_p3m": 0.07708450107589804, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.10346496329105848, "ret_p6m": 0.4520487194142655, "ret_signed_p6m": 0.4520487194142655, "alpha_spy_p6m": 0.29476040268409864, "alpha_c_p6m": 0.29476040268409864, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.15728831673016685, "price_path": [-0.1486, -0.1466, -0.1399, -0.1435, -0.1474, -0.1522, -0.1777, -0.189, -0.1768, -0.1801, -0.2025, -0.2308, -0.2379, -0.2658, -0.2004, -0.2223, -0.1849, -0.1843, -0.1839, -0.1887, -0.1849, -0.1849, -0.1837, -0.1774, -0.1655, -0.1603, -0.1624, -0.1648, -0.158, -0.1567, -0.1655, -0.1339, -0.1238, -0.122, -0.1327, -0.1415, -0.1342, -0.1289, -0.1277, -0.1204, -0.1175, -0.1177, -0.1172, -0.1235, -0.113, -0.1252, -0.1151, -0.1151, -0.0998, -0.0882, -0.0725, -0.09, -0.0734, -0.0711, -0.0398, -0.0191, -0.0159, -0.0023, -0.0071, 0.0054, 0.0159, -0.0032, 0.0234, 0.0, 0.0186, 0.0186, 0.0285, 0.0419, 0.0903, 0.0861, 0.0797, 0.0617, 0.0773, 0.0752, 0.0798, 0.0911, 0.0911, 0.0516, 0.0818, 0.0818, 0.0977, 0.089, 0.0986, 0.0967, 0.0968, 0.0965, 0.0864, 0.1087, 0.094, 0.1084, 0.1003, 0.0988, 0.0913, 0.103, 0.1058, 0.0928, 0.0603, 0.0855, 0.094, 0.0983, 0.1022, 0.1031, 0.0961, 0.1121, 0.1214, 0.1055, 0.1015, 0.0926, 0.0722, 0.071, 0.0689, 0.1033, 0.0916, 0.0864, 0.0876, 0.1052, 0.0842, 0.0842, 0.0794, 0.0943, 0.0967, 0.1051, 0.1135, 0.1244, 0.1504, 0.1685, 0.1805, 0.1885, 0.211, 0.1981, 0.2119, 0.2014, 0.2241, 0.2224, 0.2013, 0.199, 0.1837, 0.2043, 0.2022, 0.229, 0.2536, 0.2584, 0.2724, 0.2509, 0.2635, 0.2521, 0.2115, 0.255, 0.241, 0.2783, 0.3189, 0.3285, 0.3572, 0.3303, 0.3425, 0.3518, 0.3815, 0.4154, 0.4175, 0.4543, 0.4297, 0.4558, 0.5125, 0.4372, 0.4501, 0.4034, 0.3778, 0.4358, 0.4301, 0.4396, 0.4118, 0.4127, 0.3961, 0.3746, 0.3591, 0.3287, 0.339, 0.3552, 0.3476, 0.374, 0.374, 0.3638, 0.36, 0.3818, 0.3958, 0.3788, 0.4153, 0.4268, 0.4343, 0.452]}
{"unit_id": "orig:1928052951089827864", "ticker_idx": 0, "ts": "2025-05-29T11:36:59+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The investment rating remains Buy. The target price is raised to $180, maintaining the prior P/E multiple of 30x.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi Buy reiteration on NVDA with PT raised to $180; expects margin expansion, trading-range breakout to new 52-week highs on Blackwell demand and upcoming catalysts.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi) NVIDIA Corp Analysis: Positive AI Outlook (“AI is A-OK”), Target Price Raised to $180\n\nCiti’s View\nNVIDIA reported its April quarter earnings in line with market expectations, and guided July quarter revenue of $45 billion, which is approximately $1 billion above expectations. This suggests that NVIDIA has passed the final transitional quarter following China’s H20 GPU export restrictions. Additionally, Blackwell product revenue reached $24 billion, surpassing Citi’s estimate of $20 billion. Management reaffirmed its target gross margin in the mid-70% range, supported by improving profitability in Blackwell and the absence of significant tariff impacts.\n\nAccordingly, Citi expects NVIDIA’s margins to continue expanding through the January quarter and believes the stock is likely to break out of the trading range it has been stuck in since mid-last year and reach new 52-week highs. Citi also expects upcoming events—including its Silicon Valley Bus Tour next week and NVIDIA’s GTC event in Paris on June 11–12—to serve as positive catalysts. GTC, in particular, is expected to feature announcements on sovereign AI initiatives across European nations. NVIDIA emphasized that sovereign AI and AI factories will require tens of gigawatts in power, with a data center market opportunity of approximately $50 billion per gigawatt.\nCiti maintains its FY2026 EPS forecast but raises FY2027 and FY2028 EPS estimates by 9% and 5%, respectively, to reflect improved gross margin assumptions. The target price is raised to $180, maintaining the prior P/E multiple of 30x. The investment rating remains Buy.\n\nCiti’s In-House AI GPU Model Update\n\nShipments:\nCiti maintains its prior GPU shipment estimates of 6 million units in FY2026 (+8% YoY) and 6.4 million units in FY2027 (+6% YoY). Based on stronger-than-expected Blackwell demand, findings from Citi’s Asia channel checks, and commentary from NVIDIA’s suppliers, Citi reiterates its expectation that Blackwell shipments in FY2026 will be 10x higher than the prior year. The full-year shipment estimate is based on TSMC’s CoWoS monthly capacity of ~70,000 wafers and NVIDIA being allocated more than 50%.\n\nAI GPU Revenue:\nDriven by shipment growth and a favorable product mix, Citi estimates FY2026 and FY2027 AI GPU revenues will reach $168 billion (+60% YoY) and $217 billion (+29% YoY), respectively. AI GPU sales are expected to account for the mid-to-high 80% range of NVIDIA’s overall Data Center revenue during FY2026–2027. The remaining contribution will come from networking (Ethernet and InfiniBand), standalone CPUs, and software/services.\n\nEarnings & Guidance\nIn the April quarter, NVIDIA reported revenue of $44.1 billion and non-GAAP EPS of $0.96 (excluding stock-based compensation). EPS including H20-related costs and taxes was $0.81. This exceeded Citi’s forecast of $43.6 billion / $0.94 (ex-SBC) and also beat the FactSet consensus of $43.3 billion / $0.74 (inclusive of costs).\n\nNVIDIA incurred $4.5 billion in costs related to excess H20 inventory and purchase obligations stemming from China’s export restrictions. This was below the previously expected $5.5 billion. H20 sales were $4.6 billion in the April quarter, prior to the restrictions, and an additional $2.5 billion in shipments were not fulfilled.\nFor the July quarter, NVIDIA guided revenue of $45 billion (midpoint), which is slightly below Citi’s estimate of $44.2 billion and the market consensus of $45.9 billion. This guidance reflects an ~$8 billion revenue loss from the H20 export controls.\n\nLooking at other metrics, April quarter non-GAAP gross margin (excluding SBC) was 71.3%, slightly above Citi’s estimate of 71%. July quarter non-GAAP gross margin guidance is 72.0% (midpoint), in line with the market and 100bps above Citi’s expectations. NVIDIA reiterated its goal to maintain gross margins in the mid-70% range by year-end.\n\nSegment Breakdown\n\nData Center (DC):\nThanks to the Blackwell launch, Data Center revenue rose 10% QoQ (vs. Citi/Street est. of +17%/+11%). With the transition from Hopper to Blackwell nearly complete, Blackwell accounted for ~70% of quarterly compute revenue of $34 billion. Citi estimates that the $24 billion Blackwell revenue corresponds to roughly 700,000 units sold.\n\nNetworking revenue re-accelerated, growing 64% QoQ and 56% YoY to reach $8 billion annually, driven by Blackwell and Spectrum-X. Notably, GCP and Meta are now Spectrum-X customers. Hyperscalers accounted for less than 50% of DC revenue. Citi sees the high mix of enterprise, sovereign AI, and consumer internet clients as positive and indicative of NVIDIA's efforts to expand beyond hyperscalers, especially in inference. According to management, inference workloads are growing rapidly. OpenAI, Microsoft, and Google have shown explosive growth in token generation. For example, Microsoft processed over 100 trillion tokens in Q1, up 5x YoY.\nNVIDIA’s inference platform continues to attract customers due to its processing throughput and efficiency, especially as test-time scaling increases compute demand for inference workloads.\nIn networking, NVIDIA achieved $5 billion in Q1 revenue, up 64% QoQ. Growth spanned both scale-up and scale-out infrastructure.\n\n• On the scale-up side, NVIDIA achieved $1 billion in NVLink revenue.\n\n• On the scale-out side, it added two new customers for its Spectrum-X (Ethernet) portfolio.\n\nBlackwell Detailed Analysis\nNVIDIA continues to emphasize that Blackwell is the fastest product ramp in company history. While management didn’t disclose exact platform contribution for the quarter, commentary implies revenue was at least $24 billion, up from $11 billion last quarter.\n\nCiti had forecast $20 billion for Blackwell this quarter and views the results positively. Citi believes demand for Blackwell significantly exceeds supply.\nIn fact, major suppliers like Wistron expect to ship 100–200 NVL servers per week in Q1, potentially tripling by June. Given the faster-than-expected adoption and deployment of Blackwell, Citi projects that July quarter shipments may exceed 900,000 units.\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031467771192964955, "ret_signed_m1d": -0.031467771192964955, "alpha_spy_m1d": -0.027536013589865016, "alpha_c_m1d": -0.02442091775738242, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029168709381678815, "ret_signed_p1d": -0.029168709381678815, "alpha_spy_p1d": -0.028050357189775443, "alpha_c_p1d": -0.011428601389960602, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.005747599702493611, "ret_signed_p1w": 0.005747599702493611, "alpha_spy_p1w": 0.0006632770562644552, "alpha_c_p1w": -0.023832855030635036, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.13342157350479011, "ret_signed_p1m": 0.13342157350479011, "alpha_spy_p1m": 0.08820871516440842, "alpha_c_p1m": -0.007270048081713831, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.30600340555452554, "ret_signed_p3m": 0.30600340555452554, "alpha_spy_p3m": 0.20937208449485212, "alpha_c_p3m": 0.09021349223987807, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.2853115202986203, "ret_signed_p6m": 0.2853115202986203, "alpha_spy_p6m": 0.16199846496754655, "alpha_c_p6m": -0.05084882377385047, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [-0.1806, -0.1668, -0.1573, -0.2057, -0.1905, -0.2315, -0.2187, -0.1685, -0.1696, -0.1259, -0.1412, -0.1707, -0.1557, -0.1484, -0.1544, -0.1277, -0.1329, -0.1827, -0.1994, -0.2121, -0.2214, -0.2086, -0.2067, -0.2686, -0.3224, -0.2985, -0.3081, -0.1786, -0.2272, -0.203, -0.2046, -0.1939, -0.2493, -0.2709, -0.2709, -0.3038, -0.2895, -0.2621, -0.2354, -0.2025, -0.2188, -0.2168, -0.2175, -0.1981, -0.1774, -0.1823, -0.1843, -0.159, -0.1568, -0.1619, -0.1163, -0.0665, -0.0277, -0.0313, -0.0272, -0.026, -0.0346, -0.0531, -0.0457, -0.0568, -0.0568, -0.0265, -0.0315, 0.0, -0.0292, -0.013, 0.0146, 0.0196, 0.0057, 0.0182, 0.0247, 0.0343, 0.0262, 0.0418, 0.02, 0.0396, 0.0355, 0.0453, 0.0453, 0.0336, 0.0359, 0.0627, 0.1087, 0.1138, 0.1334, 0.1351, 0.1014, 0.1298, 0.1448, 0.1448, 0.1369, 0.1496, 0.1703, 0.179, 0.1849, 0.1788, 0.2265, 0.2313, 0.243, 0.2388, 0.2314, 0.2001, 0.227, 0.2483, 0.2466, 0.2699, 0.261, 0.288, 0.278, 0.2482, 0.2933, 0.2808, 0.2891, 0.2988, 0.3127, 0.3081, 0.316, 0.3047, 0.3078, 0.2965, 0.3077, 0.262, 0.2602, 0.2572, 0.2788, 0.2919, 0.306, 0.3048, 0.2945, 0.2515, 0.2515, 0.227, 0.2259, 0.2334, 0.2, 0.2093, 0.2269, 0.2741, 0.273, 0.2777, 0.2772, 0.2566, 0.2236, 0.2663, 0.2694, 0.3193, 0.2821, 0.2716, 0.2768, 0.2804, 0.3067, 0.3406, 0.3454, 0.3572, 0.3481, 0.3332, 0.3296, 0.3588, 0.3837, 0.3161, 0.3531, 0.2936, 0.2921, 0.3064, 0.3165, 0.3123, 0.3017, 0.2954, 0.3089, 0.3383, 0.3759, 0.4445, 0.4877, 0.4578, 0.455, 0.4865, 0.4277, 0.4026, 0.3514, 0.3519, 0.4302, 0.3879, 0.3925, 0.3427, 0.3664, 0.3408, 0.3031, 0.3402, 0.298, 0.2853]}
{"unit_id": "orig:1927485332565680281", "ticker_idx": 0, "ts": "2025-05-27T22:01:28+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "our conviction has only grown stronger... inference demand driven by token growth under conditions of supply shortage is deeply significant... alleviating the biggest concern around NVDA stock", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long NVDA: explosive token/inference demand is revenue-generating (not VC speculation), alleviating the key bear case; conviction has grown stronger.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: NVDA Commentary (Explosive Token Growth)\n\nWhat truly matters in the long run is the explosive growth in tokens. We’ve written about this before, and our conviction has only grown stronger. Every hyperscaler has reported unexpectedly strong token growth. But our confidence doesn’t stem solely from these reports. Everyone we speak to in this space is surprised by the level of inference demand and says they are competing to secure GPUs.\n\nLLM cloud customers, facing constrained GB200 supply, are asking cloud partners to expand capacity for Hopper and B200. Some cloud customers are listening, which explains the rising demand for those form factors. That said, due to multi-year ROIC (Return on Invested Capital) targets, there’s still some hesitation about adding more Hopper capacity right now.\n\nThis strength is unlikely to translate directly into a beat for the July quarter due to the constraints mentioned earlier. However, we believe the inference demand driven by token growth under conditions of supply shortage is deeply significant. This is not the result of speculative VC-funded mega-training clusters. It’s evidence that models are actually being used and are generating revenue—alleviating the biggest concern around NVDA stock, which is that all of this demand is being driven by eventually-exhaustible venture capital.\n\nThe strength we’re emphasizing here stems from usage—revenue-generating parts of the business—and shows clear signs that inference at scale is becoming a reality as the market shifts toward inference-oriented models.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031070050974117636, "ret_signed_m1d": -0.031070050974117636, "alpha_spy_m1d": -0.010702831348300745, "alpha_c_m1d": -0.0003316124021961464, "bench_spy_m1d": -0.02036721962581689, "bench_c_m1d": -0.03073843857192149, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005092117794083917, "ret_signed_p1d": -0.005092117794083917, "alpha_spy_p1d": 0.0006932966829971132, "alpha_c_p1d": 0.005562224713679842, "bench_spy_p1d": -0.00578541447708103, "bench_c_p1d": -0.010654342507763759, "ret_p1w": 0.04221407101909369, "ret_signed_p1w": 0.04221407101909369, "alpha_spy_p1w": 0.03385754687530862, "alpha_c_p1w": 0.02662032188993968, "bench_spy_p1w": 0.008356524143785071, "bench_c_p1w": 0.015593749129154011, "ret_p1m": 0.13889841614726683, "ret_signed_p1m": 0.13889841614726683, "alpha_spy_p1m": 0.10884724041286109, "alpha_c_p1m": 0.015495539690978166, "bench_spy_p1m": 0.030051175734405744, "bench_c_p1m": 0.12340287645628867, "ret_p3m": 0.3136707616043415, "ret_signed_p3m": 0.3136707616043415, "alpha_spy_p3m": 0.2188256380657594, "alpha_c_p3m": 0.11499331017395487, "bench_spy_p3m": 0.0948451235385821, "bench_c_p3m": 0.19867745143038662, "ret_p6m": 0.3767046424124256, "ret_signed_p6m": 0.3767046424124256, "alpha_spy_p6m": 0.24935722555262974, "alpha_c_p6m": -0.008934470055188326, "bench_spy_p6m": 0.1273474168597959, "bench_c_p6m": 0.38563911246761395, "price_path": [-0.1134, -0.0782, -0.1583, -0.1441, -0.1344, -0.1841, -0.1684, -0.2106, -0.1974, -0.1458, -0.147, -0.1021, -0.1179, -0.1481, -0.1327, -0.1252, -0.1314, -0.104, -0.1093, -0.1604, -0.1776, -0.1906, -0.2001, -0.1871, -0.1851, -0.2487, -0.304, -0.2794, -0.2893, -0.1562, -0.2061, -0.1813, -0.183, -0.172, -0.2289, -0.251, -0.251, -0.2848, -0.2702, -0.242, -0.2145, -0.1807, -0.1976, -0.1954, -0.1962, -0.1763, -0.155, -0.16, -0.1621, -0.1361, -0.1338, -0.1391, -0.0923, -0.0411, -0.0012, -0.0049, -0.0007, 0.0005, -0.0083, -0.0273, -0.0197, -0.0311, -0.0311, 0.0, -0.0051, 0.0272, -0.0027, 0.0139, 0.0422, 0.0474, 0.0331, 0.0459, 0.0526, 0.0624, 0.0542, 0.0702, 0.0478, 0.0679, 0.0637, 0.0737, 0.0737, 0.0617, 0.0641, 0.0916, 0.1389, 0.1441, 0.1643, 0.1661, 0.1314, 0.1606, 0.176, 0.176, 0.1679, 0.1809, 0.2022, 0.2112, 0.2172, 0.2109, 0.2599, 0.2648, 0.2768, 0.2725, 0.2649, 0.2328, 0.2605, 0.2823, 0.2805, 0.3045, 0.2954, 0.3231, 0.3128, 0.2822, 0.3285, 0.3157, 0.3242, 0.3342, 0.3484, 0.3437, 0.3518, 0.3402, 0.3434, 0.3318, 0.3433, 0.2963, 0.2946, 0.2915, 0.3137, 0.3271, 0.3416, 0.3403, 0.3298, 0.2856, 0.2856, 0.2605, 0.2593, 0.267, 0.2327, 0.2422, 0.2603, 0.3088, 0.3077, 0.3125, 0.312, 0.2908, 0.2569, 0.3008, 0.304, 0.3552, 0.317, 0.3062, 0.3115, 0.3152, 0.3422, 0.3771, 0.382, 0.3942, 0.3848, 0.3695, 0.3658, 0.3958, 0.4214, 0.3519, 0.39, 0.3288, 0.3273, 0.3419, 0.3523, 0.3481, 0.3371, 0.3306, 0.3445, 0.3748, 0.4134, 0.4838, 0.5282, 0.4975, 0.4946, 0.527, 0.4665, 0.4408, 0.3882, 0.3887, 0.4692, 0.4257, 0.4304, 0.3792, 0.4036, 0.3773, 0.3386, 0.3767]}
{"unit_id": "orig:1932081421834944884", "ticker_idx": 0, "ts": "2025-06-09T14:24:41+00:00", "symbol": "NDX", "kind": "index", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect Megacap Tech and the Nasdaq 100 to reach new All-Time Highs (ATH) again", "tweet_type": "analysis", "timeline": "weeks", "summary": "JPM bullish on Nasdaq 100 to new ATH, NVDA on Jensen keynotes + strong AI demand, and AVGO on AI revenue well above consensus for FY25/26.", "resolved_tickers": ["QQQ"], "resolution_method": "etf", "resolution_reason": "Nasdaq-100 index best proxied by QQQ ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan) AI / MEGACAP TECH – Last Week Nasdaq 100 (NDX) and Philadelphia Semiconductor Index (SOX) Rose 2.0% and 5.9% Respectively\n\nThe broad AI theme rally was boosted by positive earnings announcements from NVIDIA (NVDA) and Broadcom (AVGO), as well as comments from TSMC, all indicating strong AI demand.\n\n1. Future Outlook?\n\nThe past two weeks of semiconductor company earnings should support the floor for a short-term AI theme rally. The Nasdaq 100 index is currently 1.9% below its previous DeepSeek highs, and as macroeconomic indicators remain robust, we expect Megacap Tech and the Nasdaq 100 to reach new All-Time Highs (ATH) again. This week, NVIDIA CEO Jensen Huang has two keynote speeches (June 9th, June 11th) where he could potentially announce unexpected positive news. A re-escalation of trade wars (including semiconductor bans/sanctions) could dampen this optimistic outlook, but it's important to note that NVIDIA's $45 billion revenue guidance already includes an estimated $8 billion loss in China revenue due to existing sanctions. Another consideration is positioning, where we've seen increased buying from Hedge Funds (HF) as retail investor interest wanes.\n\n2. Insights into AI Demand (Josh Meyers; Post-Broadcom Earnings)\n\n– Persistent AI growth stands out (investors were initially skeptical when Harlan mentioned $5.1 billion in AI revenue for the July quarter). Our team projects +60% YoY AI growth for the fiscal year, implying over $6.0 billion in AI revenue for Q4 and $19.5 billion for the full year, surpassing the consensus of $18.5 billion. Even more exciting is that CEO Hock Tan endorsed Harlan's projection that this figure would grow another 60% in fiscal year 2026 to over $31.0 billion (vs. consensus of $25.6 billion). This will be supported by TPU V6 3nm, TPU V7 3nm, Meta 3nm and other projects, and Tomahawk 6. The networking segment's share of AI revenue is expected to normalize going forward. Cloud/hyperscaler capital expenditure (capex) remains strong, driven by anticipated AI training workloads and unexpected AI inference workload demand.\n\nTSMC CEO Comment (Reuters) – \"US tariffs are having some impact on the company and have been discussed with Washington, but Artificial Intelligence (AI) demand remains strong and continues to outstrip supply.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0014698771677770228, "ret_signed_m1d": -0.0014698771677770228, "alpha_spy_m1d": -0.000569405859014549, "alpha_c_m1d": -0.000569405859014549, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.0009004713087624738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006614042775864082, "ret_signed_p1d": 0.006614042775864082, "alpha_spy_p1d": 0.0009443781069733426, "alpha_c_p1d": 0.0009443781069733426, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005669664668890739, "ret_p1w": 0.006764624578648348, "ret_signed_p1w": 0.006764624578648348, "alpha_spy_p1w": 0.001761949004243979, "alpha_c_p1w": 0.001761949004243979, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.005002675574404369, "ret_p1m": 0.04194529500627753, "ret_signed_p1m": 0.04194529500627753, "alpha_spy_p1m": 0.004435368409050122, "alpha_c_p1m": 0.004435368409050122, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.037509926597227405, "ret_p3m": 0.08512529145610648, "ret_signed_p3m": 0.08512529145610648, "alpha_spy_p3m": -0.0005186201657281231, "alpha_c_p3m": -0.0005186201657281231, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.0856439116218346, "ret_p6m": 0.17471337453011082, "ret_signed_p6m": 0.17471337453011082, "alpha_spy_p6m": 0.0317038711451334, "alpha_c_p6m": 0.0317038711451334, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.14300950338497742, "price_path": [-0.1027, -0.1188, -0.0975, -0.0917, -0.1072, -0.0952, -0.0983, -0.0953, -0.0754, -0.0702, -0.0873, -0.0925, -0.1164, -0.1164, -0.1093, -0.1028, -0.1508, -0.2036, -0.2016, -0.216, -0.1219, -0.1593, -0.1438, -0.138, -0.137, -0.163, -0.1632, -0.1632, -0.1839, -0.1625, -0.1435, -0.1194, -0.1096, -0.1098, -0.104, -0.1041, -0.0924, -0.0789, -0.0844, -0.0929, -0.0893, -0.0799, -0.0805, -0.0431, -0.0285, -0.0226, -0.0216, -0.0173, -0.0164, -0.0197, -0.0333, -0.0315, -0.0404, -0.0404, -0.0179, -0.0222, -0.0203, -0.0218, -0.0141, -0.0064, -0.0036, -0.0111, -0.0015, 0.0, 0.0066, 0.0032, 0.0056, -0.007, 0.0068, -0.0031, -0.0032, -0.0032, -0.0073, 0.0029, 0.0183, 0.0209, 0.0304, 0.0339, 0.0406, 0.0319, 0.039, 0.0493, 0.0493, 0.0414, 0.0419, 0.0493, 0.0478, 0.0455, 0.0492, 0.0502, 0.0513, 0.0598, 0.0588, 0.0643, 0.0588, 0.0636, 0.0658, 0.0684, 0.0718, 0.0701, 0.0715, 0.0658, 0.0449, 0.0641, 0.0569, 0.0702, 0.0738, 0.0838, 0.0806, 0.0942, 0.0948, 0.0939, 0.0891, 0.0887, 0.0739, 0.0675, 0.0626, 0.079, 0.0759, 0.0802, 0.0818, 0.0886, 0.076, 0.076, 0.067, 0.0754, 0.0851, 0.0867, 0.092, 0.0951, 0.0954, 0.1018, 0.1067, 0.1162, 0.1152, 0.113, 0.123, 0.1306, 0.1373, 0.1298, 0.1258, 0.1209, 0.1256, 0.1308, 0.1339, 0.1393, 0.144, 0.1392, 0.1477, 0.1417, 0.1548, 0.1534, 0.1133, 0.137, 0.1294, 0.1374, 0.1331, 0.1406, 0.155, 0.1547, 0.1435, 0.1531, 0.1655, 0.1862, 0.1953, 0.2007, 0.1824, 0.1881, 0.1938, 0.1695, 0.1771, 0.1552, 0.1516, 0.177, 0.1739, 0.173, 0.149, 0.1499, 0.1401, 0.1262, 0.1329, 0.1061, 0.1144, 0.1429, 0.15, 0.1601, 0.1601, 0.1695, 0.1656, 0.1747]}
{"unit_id": "orig:1932081421834944884", "ticker_idx": 1, "ts": "2025-06-09T14:24:41+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The past two weeks of semiconductor company earnings should support the floor for a short-term AI theme rally", "tweet_type": "analysis", "timeline": "weeks", "summary": "JPM bullish on Nasdaq 100 to new ATH, NVDA on Jensen keynotes + strong AI demand, and AVGO on AI revenue well above consensus for FY25/26.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan) AI / MEGACAP TECH – Last Week Nasdaq 100 (NDX) and Philadelphia Semiconductor Index (SOX) Rose 2.0% and 5.9% Respectively\n\nThe broad AI theme rally was boosted by positive earnings announcements from NVIDIA (NVDA) and Broadcom (AVGO), as well as comments from TSMC, all indicating strong AI demand.\n\n1. Future Outlook?\n\nThe past two weeks of semiconductor company earnings should support the floor for a short-term AI theme rally. The Nasdaq 100 index is currently 1.9% below its previous DeepSeek highs, and as macroeconomic indicators remain robust, we expect Megacap Tech and the Nasdaq 100 to reach new All-Time Highs (ATH) again. This week, NVIDIA CEO Jensen Huang has two keynote speeches (June 9th, June 11th) where he could potentially announce unexpected positive news. A re-escalation of trade wars (including semiconductor bans/sanctions) could dampen this optimistic outlook, but it's important to note that NVIDIA's $45 billion revenue guidance already includes an estimated $8 billion loss in China revenue due to existing sanctions. Another consideration is positioning, where we've seen increased buying from Hedge Funds (HF) as retail investor interest wanes.\n\n2. Insights into AI Demand (Josh Meyers; Post-Broadcom Earnings)\n\n– Persistent AI growth stands out (investors were initially skeptical when Harlan mentioned $5.1 billion in AI revenue for the July quarter). Our team projects +60% YoY AI growth for the fiscal year, implying over $6.0 billion in AI revenue for Q4 and $19.5 billion for the full year, surpassing the consensus of $18.5 billion. Even more exciting is that CEO Hock Tan endorsed Harlan's projection that this figure would grow another 60% in fiscal year 2026 to over $31.0 billion (vs. consensus of $25.6 billion). This will be supported by TPU V6 3nm, TPU V7 3nm, Meta 3nm and other projects, and Tomahawk 6. The networking segment's share of AI revenue is expected to normalize going forward. Cloud/hyperscaler capital expenditure (capex) remains strong, driven by anticipated AI training workloads and unexpected AI inference workload demand.\n\nTSMC CEO Comment (Reuters) – \"US tariffs are having some impact on the company and have been discussed with Washington, but Artificial Intelligence (AI) demand remains strong and continues to outstrip supply.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006380282151325245, "ret_signed_m1d": -0.006380282151325245, "alpha_spy_m1d": -0.005479810842562771, "alpha_c_m1d": 0.010081343882485272, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009324789052088667, "ret_signed_p1d": 0.009324789052088667, "alpha_spy_p1d": 0.0036551243831979274, "alpha_c_p1d": -0.010561686271912674, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.014513336154033274, "ret_signed_p1w": 0.014513336154033274, "alpha_spy_p1w": 0.009510660579628905, "alpha_c_p1w": -0.01047121819025243, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.12186151647081211, "ret_signed_p1m": 0.12186151647081211, "alpha_spy_p1m": 0.08435158987358471, "alpha_c_p1m": 0.018654633796084452, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.20361719462765793, "ret_signed_p3m": 0.20361719462765793, "alpha_spy_p3m": 0.11797328300582333, "alpha_c_p3m": 0.07558170333429004, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.27240289611841284, "ret_signed_p6m": 0.27240289611841284, "alpha_spy_p6m": 0.12939339273343542, "alpha_c_p6m": -0.1263752232547639, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1885, -0.1897, -0.147, -0.162, -0.1907, -0.176, -0.169, -0.1748, -0.1488, -0.1538, -0.2024, -0.2187, -0.2311, -0.2401, -0.2277, -0.2258, -0.2863, -0.3388, -0.3154, -0.3248, -0.1984, -0.2458, -0.2223, -0.2238, -0.2133, -0.2674, -0.2884, -0.2884, -0.3205, -0.3067, -0.2799, -0.2538, -0.2217, -0.2377, -0.2356, -0.2363, -0.2175, -0.1972, -0.202, -0.204, -0.1793, -0.1771, -0.1821, -0.1376, -0.089, -0.0511, -0.0547, -0.0507, -0.0495, -0.0578, -0.0759, -0.0687, -0.0795, -0.0795, -0.05, -0.0548, -0.0241, -0.0526, -0.0368, -0.0099, -0.005, -0.0185, -0.0064, 0.0, 0.0093, 0.0015, 0.0167, -0.0046, 0.0145, 0.0105, 0.0201, 0.0201, 0.0086, 0.0109, 0.037, 0.082, 0.0869, 0.1061, 0.1078, 0.0749, 0.1026, 0.1172, 0.1172, 0.1095, 0.1219, 0.1421, 0.1506, 0.1564, 0.1504, 0.1969, 0.2016, 0.213, 0.2089, 0.2017, 0.1712, 0.1974, 0.2182, 0.2165, 0.2393, 0.2306, 0.257, 0.2472, 0.2181, 0.2621, 0.2499, 0.258, 0.2675, 0.281, 0.2765, 0.2843, 0.2732, 0.2763, 0.2652, 0.2762, 0.2315, 0.2298, 0.2269, 0.248, 0.2608, 0.2745, 0.2733, 0.2633, 0.2213, 0.2213, 0.1974, 0.1963, 0.2036, 0.1711, 0.1801, 0.1973, 0.2434, 0.2423, 0.2469, 0.2464, 0.2263, 0.1941, 0.2358, 0.2388, 0.2875, 0.2512, 0.2409, 0.246, 0.2495, 0.2751, 0.3083, 0.3129, 0.3245, 0.3156, 0.301, 0.2975, 0.326, 0.3503, 0.2843, 0.3205, 0.2624, 0.261, 0.2749, 0.2847, 0.2807, 0.2703, 0.2641, 0.2773, 0.3061, 0.3427, 0.4096, 0.4518, 0.4227, 0.4199, 0.4506, 0.3932, 0.3688, 0.3188, 0.3193, 0.3957, 0.3544, 0.3589, 0.3103, 0.3335, 0.3084, 0.2717, 0.3079, 0.2667, 0.2543, 0.28, 0.2469, 0.264, 0.264, 0.2411, 0.2616, 0.2724]}
{"unit_id": "orig:1932081421834944884", "ticker_idx": 2, "ts": "2025-06-09T14:24:41+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Our team projects +60% YoY AI growth for the fiscal year, implying over $6.0 billion in AI revenue for Q4 and $19.5 billion for the full year, surpassing the consensus of $18.5 billion", "tweet_type": "analysis", "timeline": "weeks", "summary": "JPM bullish on Nasdaq 100 to new ATH, NVDA on Jensen keynotes + strong AI demand, and AVGO on AI revenue well above consensus for FY25/26.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan) AI / MEGACAP TECH – Last Week Nasdaq 100 (NDX) and Philadelphia Semiconductor Index (SOX) Rose 2.0% and 5.9% Respectively\n\nThe broad AI theme rally was boosted by positive earnings announcements from NVIDIA (NVDA) and Broadcom (AVGO), as well as comments from TSMC, all indicating strong AI demand.\n\n1. Future Outlook?\n\nThe past two weeks of semiconductor company earnings should support the floor for a short-term AI theme rally. The Nasdaq 100 index is currently 1.9% below its previous DeepSeek highs, and as macroeconomic indicators remain robust, we expect Megacap Tech and the Nasdaq 100 to reach new All-Time Highs (ATH) again. This week, NVIDIA CEO Jensen Huang has two keynote speeches (June 9th, June 11th) where he could potentially announce unexpected positive news. A re-escalation of trade wars (including semiconductor bans/sanctions) could dampen this optimistic outlook, but it's important to note that NVIDIA's $45 billion revenue guidance already includes an estimated $8 billion loss in China revenue due to existing sanctions. Another consideration is positioning, where we've seen increased buying from Hedge Funds (HF) as retail investor interest wanes.\n\n2. Insights into AI Demand (Josh Meyers; Post-Broadcom Earnings)\n\n– Persistent AI growth stands out (investors were initially skeptical when Harlan mentioned $5.1 billion in AI revenue for the July quarter). Our team projects +60% YoY AI growth for the fiscal year, implying over $6.0 billion in AI revenue for Q4 and $19.5 billion for the full year, surpassing the consensus of $18.5 billion. Even more exciting is that CEO Hock Tan endorsed Harlan's projection that this figure would grow another 60% in fiscal year 2026 to over $31.0 billion (vs. consensus of $25.6 billion). This will be supported by TPU V6 3nm, TPU V7 3nm, Meta 3nm and other projects, and Tomahawk 6. The networking segment's share of AI revenue is expected to normalize going forward. Cloud/hyperscaler capital expenditure (capex) remains strong, driven by anticipated AI training workloads and unexpected AI inference workload demand.\n\nTSMC CEO Comment (Reuters) – \"US tariffs are having some impact on the company and have been discussed with Washington, but Artificial Intelligence (AI) demand remains strong and continues to outstrip supply.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010848173740044542, "ret_signed_m1d": 0.010848173740044542, "alpha_spy_m1d": 0.011748645048807016, "alpha_c_m1d": 0.02730979977385506, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0014328929116100841, "ret_signed_p1d": 0.0014328929116100841, "alpha_spy_p1d": -0.004236771757280655, "alpha_c_p1d": -0.018453582412391256, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.032012483768575795, "ret_signed_p1w": 0.032012483768575795, "alpha_spy_p1w": 0.027009808194171425, "alpha_c_p1w": 0.007027929424290091, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.11527643939303767, "ret_signed_p1m": 0.11527643939303767, "alpha_spy_p1m": 0.07776651279581026, "alpha_c_p1m": 0.012069556718310004, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.25601983468867817, "ret_signed_p3m": 0.25601983468867817, "alpha_spy_p3m": 0.17037592306684357, "alpha_c_p3m": 0.12798434339531028, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.5683782731496272, "ret_signed_p6m": 0.5683782731496272, "alpha_spy_p6m": 0.42536876976464977, "alpha_c_p6m": 0.16960015377645044, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.2073, -0.219, -0.2019, -0.2062, -0.23, -0.2018, -0.22, -0.2154, -0.2171, -0.2293, -0.2661, -0.2959, -0.3077, -0.3146, -0.3101, -0.2955, -0.3695, -0.4011, -0.369, -0.3613, -0.2421, -0.2947, -0.2552, -0.2699, -0.2674, -0.2852, -0.3, -0.3, -0.3196, -0.3058, -0.2758, -0.2298, -0.2127, -0.2121, -0.2174, -0.2121, -0.1922, -0.1664, -0.1783, -0.1809, -0.1616, -0.1495, -0.1477, -0.0929, -0.0486, -0.0498, -0.0477, -0.0641, -0.0559, -0.0516, -0.0596, -0.0563, -0.0637, -0.0637, -0.0353, -0.0199, -0.0095, -0.009, 0.0181, 0.0515, 0.0688, 0.0641, 0.0108, 0.0, 0.0014, 0.0353, 0.0483, 0.0181, 0.032, 0.0208, 0.0286, 0.0286, 0.0258, 0.0413, 0.0823, 0.0859, 0.1086, 0.1052, 0.1311, 0.0863, 0.1075, 0.1291, 0.1291, 0.125, 0.1153, 0.1403, 0.13, 0.1259, 0.1309, 0.1528, 0.1522, 0.1754, 0.1626, 0.1826, 0.1431, 0.1641, 0.1847, 0.1907, 0.2076, 0.2204, 0.2417, 0.2051, 0.1844, 0.2216, 0.202, 0.2378, 0.2464, 0.2514, 0.247, 0.2836, 0.2683, 0.2771, 0.257, 0.2546, 0.2101, 0.1948, 0.1883, 0.2064, 0.2073, 0.2228, 0.232, 0.2665, 0.2203, 0.2203, 0.2238, 0.2408, 0.256, 0.3742, 0.4183, 0.3815, 0.5165, 0.4757, 0.4767, 0.494, 0.4772, 0.4204, 0.4171, 0.4154, 0.3925, 0.3932, 0.3947, 0.3815, 0.375, 0.3478, 0.356, 0.3703, 0.39, 0.3908, 0.379, 0.3828, 0.4201, 0.4181, 0.3343, 0.4662, 0.4145, 0.4441, 0.4557, 0.4359, 0.4355, 0.4084, 0.3987, 0.4151, 0.4556, 0.4881, 0.533, 0.5865, 0.5474, 0.5193, 0.4902, 0.4466, 0.4755, 0.4616, 0.4363, 0.4731, 0.4467, 0.4601, 0.3974, 0.4076, 0.4084, 0.3996, 0.4568, 0.4255, 0.3983, 0.5535, 0.5826, 0.6341, 0.6341, 0.6563, 0.5869, 0.5684]}
{"unit_id": "orig:1933011120765472840", "ticker_idx": 0, "ts": "2025-06-12T03:58:59+00:00", "symbol": "GOOG", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It can promote the growth of the Google Cloud business, which is not being fully valued at its current level of an estimated 2026 Price-to-Earnings Ratio (PER) of about 18x", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish thesis: Google Cloud undervalued at 18x 2026 PE, OpenAI partnership accelerates growth and signals confidence in search dominance.", "resolved_tickers": ["GOOG"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Google Cloud and OpenAI's Collaboration, Morgan Stanley: \"This holds a very important meaning for Google investors\"\n\nAccording to previous reports from foreign media such as The Wall Street Journal (WSJ), OpenAI signed a partnership agreement with Google last month to receive the necessary computing power for model training and operation from Google Cloud, it was reported. Until then, OpenAI could not collaborate with other cloud services due to its exclusive contract with Microsoft, and until before January of this year, Microsoft was OpenAI's sole cloud service provider.\n\nAccording to the latest report published by Morgan Stanley analyst Brian Nowak and his team, the partnership agreement between Google Cloud and OpenAI holds very significant meaning in the following two aspects:\n\n1. Expectation of accelerating the growth of the Google Cloud business: It can promote the growth of the Google Cloud business, which is not being fully valued at its current level of an estimated 2026 Price-to-Earnings Ratio (PER) of about 18x.\n\n2. A demonstration of Google's confidence: It sends an important signal to the market that Google has strong confidence in its long-term position in the search market.\nThe report speculated that while it is not confirmed whether OpenAI will use Google's TPUs or GPUs, it is more likely that they will rely on Google's GPUs to meet their computing capacity demands. Through this, OpenAI can maintain consistency across its entire infrastructure, avoid 'vendor lock-in' where it becomes dependent on a specific provider, and avoid the hassle of engineers having to familiarize themselves with a new chip architecture.\n\nGoogle Cloud, Welcoming a New Growth Engine?\n\nThe report pointed out that this collaboration could be an important turning point for the Google Cloud business.\n\nCurrently, Google Cloud's annualized revenue is approaching nearly $50 billion, and it has maintained a growth rate of 20% to 30% over the past two years. However, compared to Amazon's AWS or Microsoft's Azure, it has not shown a sustained dominant position in terms of market share and growth. If a significant-sized OpenAI joins the fold, it could become a catalyst to accelerate Google Cloud's growth.\n\nThe report analyzed that Google's current estimated 2026 PER of about 18x, corresponding to its current stock price, means the market has not yet assigned a proper value premium to Google Cloud.\n\nIt also added that as Google's chip production capacity constraints ease in the second half of the year, this could become a driving force for faster and more sustained growth for Google Cloud. This is a potential positive factor for investors that has not yet been fully priced in.\n\nAlthough the specific terms and scale of this collaboration have not been disclosed, analysts expect that as the inference stage of generative AI continues to develop, OpenAI will require increasingly more computing capacity. For reference, early this year, OpenAI signed a $11.9 billion deal with CoreWeave (including a $350 million equity investment), and recently extended this deal by an additional $4 billion.\n\nGoogle, Strong Confidence in its Search Position\n\nThe deeper meaning of this collaboration is that it reflects internal confidence in the long-term competitiveness of Google's core search business.\n\nFor a long time, investors have been concerned about whether products like OpenAI's ChatGPT would dismantle Google Search's dominant position.\n\nThe report sharply pointed out, \"If Google were truly concerned about its search business being threatened, would it willingly provide key computing power to a potential competitor?\" and analyzed that the answer would be 'no.' This suggests that Google knows the data on the health of its business better than any outsider, and has strong confidence in its user base, data assets, and pace of innovation.\n\nThe report added that Meta's AI products could pose a greater structural threat to Google's search business than OpenAI, but Meta AI still has the challenge of proving its commercialization capabilities.\n\n$GOOG", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010284188118019566, "ret_signed_m1d": 0.010284188118019566, "alpha_spy_m1d": 0.014242682264432038, "alpha_c_m1d": 0.002081901164162847, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": 0.008202286953856719, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00615917500429064, "ret_signed_p1d": -0.00615917500429064, "alpha_spy_p1d": 0.005020889392176819, "alpha_c_p1d": 0.003973031130427018, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.010132206134717658, "ret_p1w": -0.016895581533173387, "ret_signed_p1w": -0.016895581533173387, "alpha_spy_p1w": -0.006444371688810624, "alpha_c_p1w": -0.015351601467845533, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.0015439800653278546, "ret_p1m": 0.024523839851944063, "ret_signed_p1m": 0.024523839851944063, "alpha_spy_p1m": -0.011440755895034238, "alpha_c_p1m": -0.0009253342472927972, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.02544917409923686, "ret_p3m": 0.35703477458472044, "ret_signed_p3m": 0.35703477458472044, "alpha_spy_p3m": 0.27669921518425045, "alpha_c_p3m": 0.24007770699454278, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.11695706759017765, "ret_p6m": 0.8216525942887918, "ret_signed_p6m": 0.8216525942887918, "alpha_spy_p6m": 0.6794185916901512, "alpha_c_p6m": 0.6846866928817017, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.13696590140709008, "price_path": [-0.0599, -0.0819, -0.0615, -0.0685, -0.0617, -0.0409, -0.0248, -0.0567, -0.074, -0.1192, -0.1183, -0.1033, -0.1034, -0.1386, -0.1662, -0.1577, -0.1727, -0.091, -0.1231, -0.1004, -0.0887, -0.1044, -0.1224, -0.1345, -0.1345, -0.1542, -0.1314, -0.1098, -0.0887, -0.0752, -0.0833, -0.0854, -0.092, -0.0812, -0.0642, -0.0628, -0.0676, -0.1376, -0.121, -0.1287, -0.0993, -0.092, -0.0585, -0.0665, -0.055, -0.0526, -0.067, -0.0402, -0.0294, -0.0429, -0.0429, -0.0181, -0.0215, -0.0238, -0.0245, -0.0385, -0.0535, -0.044, -0.0416, -0.0128, 0.0037, 0.0172, 0.0103, 0.0, -0.0062, 0.0055, 0.0015, -0.0169, -0.0169, -0.0522, -0.0619, -0.0522, -0.031, -0.0144, 0.0073, 0.0024, -0.0003, 0.0158, 0.0202, 0.0202, 0.0033, -0.0102, 0.0039, 0.0098, 0.0245, 0.033, 0.0346, 0.0384, 0.0437, 0.0507, 0.0801, 0.0856, 0.0822, 0.0917, 0.0967, 0.093, 0.11, 0.1157, 0.0898, 0.0733, 0.1061, 0.1037, 0.1127, 0.1148, 0.1419, 0.1393, 0.1536, 0.1473, 0.1517, 0.1579, 0.1544, 0.1442, 0.1312, 0.1336, 0.1681, 0.1819, 0.1751, 0.1765, 0.2, 0.2066, 0.2066, 0.1979, 0.3059, 0.3147, 0.3289, 0.3243, 0.357, 0.3549, 0.3618, 0.3652, 0.4239, 0.422, 0.4131, 0.4271, 0.4436, 0.4302, 0.4272, 0.4017, 0.3945, 0.398, 0.382, 0.3775, 0.3887, 0.3937, 0.3939, 0.4225, 0.3977, 0.3883, 0.3699, 0.3432, 0.3836, 0.3924, 0.4236, 0.4246, 0.4354, 0.4536, 0.4215, 0.4282, 0.435, 0.4734, 0.5266, 0.5182, 0.5563, 0.5943, 0.5939, 0.6069, 0.5726, 0.6105, 0.6138, 0.5819, 0.6435, 0.65, 0.6256, 0.5786, 0.5665, 0.6153, 0.6117, 0.6571, 0.64, 0.6947, 0.8012, 0.8304, 0.8114, 0.8114, 0.8105, 0.7822, 0.7873, 0.8133, 0.8007, 0.8217]}
{"unit_id": "thread:1927987037921235258", "ticker_idx": 0, "ts": "2025-05-29T07:15:04+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the era of having 'Nvidia as the only choice' is ending. A diversified, open, and distinctly Chinese computing ecosystem is taking shape.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish NVDA: China's de-Nvidia-ization is becoming engineering reality, shrinking its Chinese market share and eroding its once-unassailable position.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Commentary: China's AI computing system ditches Nvidia—Jensen Huang faces new reality-Digitimes\n\nAmid the escalating US-China tech war and supply chain risks, the trend of \"de-Nvidia-ization\" is rapidly spreading within China's computing power industry. This shift represents not only a technological transformation but also a comprehensive change encompassing policy, technology, ecosystem, and market dynamics.\n\nThe long march toward computing self-reliance involves domestic AI chip and solution companies such as Infinigence AI and Lisuan Tech joining forces to challenge Nvidia's dominant position in the Chinese market. Their goal is to establish a computing power system free from external sanctions.\n\nStrategic shift becomes engineering reality\nOn May 28, 2024, the State Council Information Office of the People's Republic of China held a press conference that clearly signaled its commitment to advancing indigenous technological autonomy.\n\nThe attending companies were among China's top-tier representatives of cutting-edge strength. One startup founded by a Tsinghua University-affiliated team, AI computing newcomer Infinigence AI, became a focal point.\n\nThrough its Infini-AI heterogeneous hybrid training platform, it integrates multiple domestic and overseas chips, including Huawei Ascend, AMD, Iluvatar CoreX, Enflame, MetaX, and Moore Threads, breaking the longstanding dependence on Nvidia's CUDA ecosystem.\n\nThe Infini-AI platform supports heterogeneous hybrid training with computing capacity at the scale of thousands of GPUs, achieving a utilization rate of 97.6% and possessing scalability potential up to tens of thousands of GPUs. This indicates that China is not only seeking breakthroughs in chips but also realizing cross-platform integration at the AI training platform level, providing alternative solutions for local large models.\n\nAccording to available data, no fewer than 100 computing clusters in China have shifted to heterogeneous system architectures, meaning that \"de-Nvidia-ization\" has moved from an option to an engineering practice. Large model enterprises such as Zhipu AI, Moonshot AI, and ShengShu are already operating on the Infini-AI platform. The clear message behind this is that China is establishing a new foundational order for homegrown AI.\n\nDomestic GPU breakthrough emerges\nMeanwhile, China's domestic GPUs are beginning to make substantial progress.\n\nFounded in 2021, Lisuan Tech recently announced the official testing launch of its first self-developed architecture GPU, the \"G100,\" after three years of research and development. It positions itself as China's first high-end GPU manufactured using a 6nm process. The G100 reportedly targets performance comparable to Nvidia's RTX 4060 series, supporting mainstream graphics APIs while featuring high computing power, large memory capacity, and low power consumption. It can be applied across diverse scenarios such as gaming, industrial design, and smart vehicles.\n\nReports suggest that the G100 has secured orders exceeding CNY100 million and attracted partnership talks with several major end manufacturers. In a GPU field long dominated by Nvidia and AMD, these domestic breakthroughs carry strategic significance in filling China's high-end GPU gap. More importantly, the G100 claims to adopt a fully independent IP architecture, free from constraints imposed by CUDA or US-based ecosystems.\n\nDriven by geopolitical pressures and export controls, China's push for autonomous control over computing infrastructure is intensifying rapidly. According to SEMI and IDC data, global AI training computing demand will double in 2025. As the world's second-largest market, China's move to build its own computing power is inevitable.\n\nIndustrial transformation beyond chip replacement\nAlthough Nvidia still holds a 50% market share in China, generating revenue of US$17.1 billion in fiscal year 2025—the second largest after the US domestic market—it has faced challenges since the US Department of Commerce imposed export controls on H20 chips. Nvidia reported inventory losses exceeding US$5.5 billion and adjusted its China strategy by planning to release updated, heavily restricted GPUs at significantly reduced prices.\n\nThis approach carries risks. The downgraded GPUs reportedly omit high-bandwidth memory (HBM) used for advanced AI computations and do not employ TSMC's CoWoS packaging technology, lowering technical thresholds. Consequently, their appeal to Chinese customers diminishes, as they increasingly view Nvidia products as optional rather than essential, shaking Nvidia's once unassailable position.\n\nNvidia CEO Jensen Huang has stated that US restrictions on technology exports to China were a \"failure,\" inadvertently accelerating China's research and development and innovation pace. China is moving toward open-source models and platforms, while hardware and middleware are following paths of independence. Once this decentralized, de-Nvidia-ized process matures, China's AI industry will possess counterattack capabilities.\n\n\"De-Nvidia-ization\" is not simply abandoning a brand or platform; it is a comprehensive reorganization spanning architecture, toolchains, and entire ecosystems. As one Chinese tech leader noted, supply chain choices are no longer straightforward. Although Nvidia chips offer high performance, they create heavy dependencies. Using Chinese chips may incur short-term switching costs but secure long-term independence.\n\nThis wave of technological autonomy is driving Chinese chipmakers to build complete ecosystems covering chip design, foundries, software platforms, and training models. Emerging forces like Infinigence AI and Lisuan Tech are constructing a fully independent, entirely \"decoupled-from-US\" AI supply system.\n\nFuture implications for regional competition\nCurrently, China's AI computing power stands at a turning point. With domestic chips filling gaps and heterogeneous hybrid training systems maturing, the era of having \"Nvidia as the only choice\" is ending. A diversified, open, and distinctly Chinese computing ecosystem is taking shape.\n\nFor Taiwan, observing China's de-Nvidia-ization process offers more than just insight into geopolitical rivalry; it presents an opportunity to rethink AI industry chain autonomy, supply chain resilience, and cross-platform collaboration potential. After all, in the future driven by computing power-enabled AI, whoever controls more flexible, secure, and scalable computing platforms will command the next wave of technological competition.\n$NVDA\n\n---\n\nhttps://t.co/UKtHIRo6nh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031467771192964955, "ret_signed_m1d": 0.031467771192964955, "alpha_spy_m1d": 0.027536013589865016, "alpha_c_m1d": 0.02442091775738242, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029168709381678815, "ret_signed_p1d": 0.029168709381678815, "alpha_spy_p1d": 0.028050357189775443, "alpha_c_p1d": 0.011428601389960602, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.005747599702493611, "ret_signed_p1w": -0.005747599702493611, "alpha_spy_p1w": -0.0006632770562644552, "alpha_c_p1w": 0.023832855030635036, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.13342157350479011, "ret_signed_p1m": -0.13342157350479011, "alpha_spy_p1m": -0.08820871516440842, "alpha_c_p1m": 0.007270048081713831, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.30600340555452554, "ret_signed_p3m": -0.30600340555452554, "alpha_spy_p3m": -0.20937208449485212, "alpha_c_p3m": -0.09021349223987807, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.2853115202986203, "ret_signed_p6m": -0.2853115202986203, "alpha_spy_p6m": -0.16199846496754655, "alpha_c_p6m": 0.05084882377385047, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [-0.1806, -0.1668, -0.1573, -0.2057, -0.1905, -0.2315, -0.2187, -0.1685, -0.1696, -0.1259, -0.1412, -0.1707, -0.1557, -0.1484, -0.1544, -0.1277, -0.1329, -0.1827, -0.1994, -0.2121, -0.2214, -0.2086, -0.2067, -0.2686, -0.3224, -0.2985, -0.3081, -0.1786, -0.2272, -0.203, -0.2046, -0.1939, -0.2493, -0.2709, -0.2709, -0.3038, -0.2895, -0.2621, -0.2354, -0.2025, -0.2188, -0.2168, -0.2175, -0.1981, -0.1774, -0.1823, -0.1843, -0.159, -0.1568, -0.1619, -0.1163, -0.0665, -0.0277, -0.0313, -0.0272, -0.026, -0.0346, -0.0531, -0.0457, -0.0568, -0.0568, -0.0265, -0.0315, 0.0, -0.0292, -0.013, 0.0146, 0.0196, 0.0057, 0.0182, 0.0247, 0.0343, 0.0262, 0.0418, 0.02, 0.0396, 0.0355, 0.0453, 0.0453, 0.0336, 0.0359, 0.0627, 0.1087, 0.1138, 0.1334, 0.1351, 0.1014, 0.1298, 0.1448, 0.1448, 0.1369, 0.1496, 0.1703, 0.179, 0.1849, 0.1788, 0.2265, 0.2313, 0.243, 0.2388, 0.2314, 0.2001, 0.227, 0.2483, 0.2466, 0.2699, 0.261, 0.288, 0.278, 0.2482, 0.2933, 0.2808, 0.2891, 0.2988, 0.3127, 0.3081, 0.316, 0.3047, 0.3078, 0.2965, 0.3077, 0.262, 0.2602, 0.2572, 0.2788, 0.2919, 0.306, 0.3048, 0.2945, 0.2515, 0.2515, 0.227, 0.2259, 0.2334, 0.2, 0.2093, 0.2269, 0.2741, 0.273, 0.2777, 0.2772, 0.2566, 0.2236, 0.2663, 0.2694, 0.3193, 0.2821, 0.2716, 0.2768, 0.2804, 0.3067, 0.3406, 0.3454, 0.3572, 0.3481, 0.3332, 0.3296, 0.3588, 0.3837, 0.3161, 0.3531, 0.2936, 0.2921, 0.3064, 0.3165, 0.3123, 0.3017, 0.2954, 0.3089, 0.3383, 0.3759, 0.4445, 0.4877, 0.4578, 0.455, 0.4865, 0.4277, 0.4026, 0.3514, 0.3519, 0.4302, 0.3879, 0.3925, 0.3427, 0.3664, 0.3408, 0.3031, 0.3402, 0.298, 0.2853]}
{"unit_id": "orig:1938152881817063453", "ticker_idx": 0, "ts": "2025-06-26T08:30:30+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Structural Bull Thesis Starting To Finally Play Out", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses UBS Buy on Micron with $155 SOTP target (5x HBM, 3x Core DRAM+NAND) on structural HBM-driven profitability transformation.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron: Structural Bull Thesis Starting To Finally Play Out\n\n📌 UBS Report Key Summary (June 26, 2025)\n\n🧠 Investment Core Logic: Realization of Structural Bull Thesis\n\n- HBM  is driving a structural transformation of the DRAM business, leading to improved profitability and stability.\n\n- HBM accounts for ~6-7% of overall DRAM bits, and ~19-20% based on DRAM production capacity.\n\n- Limited DRAM production capacity is preferentially allocated to high-margin product lines (HBM), maximizing profitability.\n\n- HBM revenue and profit allow for application of high multiples, changing the valuation framework.\n\n- For FY26~27, 5x P/S is applied to HBM, and 3x P/S to Core DRAM+NAND.\n\n- Accordingly, a target price of $155 is presented based on the SOTP (Sum-of-the-Parts) method.\n\n📊 FY25 Q3 Earnings Summary (May Quarter 2025)\n- Revenue: $9.30B. This exceeded the consensus of $8.86B and UBS's estimate of $9.25B, going above the upper end of guidance.\n- Non-GAAP EPS: $1.91. This surpassed the consensus of $1.60 and UBS's estimate of $1.78, also exceeding the upper end of guidance.\n- Gross Margin: 39.0%. This was higher than the consensus of 36.7% and UBS's estimate of 38.0%, influenced by cost improvements.\n- FCF (Free Cash Flow): $1.95B. This was better than the consensus of $1.72B and UBS's estimate of $1.64B, due to the effect of reduced CapEx.\n\nHBM revenue is estimated to grow +50% YoY, with half of the DRAM revenue increase attributable to HBM.\nThe mass production yield of HBM3E 12-Hi products is superior to last year's 8-Hi, positively impacting the cost structure.\n\n🔮 FY25 Q4 Guidance (August Quarter 2025)\n- Revenue: Guidance is $10.7B ± $300M. This significantly surpasses the consensus of $9.89B and UBS's estimate of $10.74B, indicating robust AI/data center demand.\n- EPS: Guidance is $2.50 ± $0.15. This substantially exceeds the consensus of $2.01 and UBS's estimate of $2.50.\n- Gross Margin: Guidance is 42% ± 1%. This is higher than the consensus of 39.9% and UBS's estimate of 42.3%, due to a stronger mix of high-value products.\n\n🧩 Product-Specific Insights\nHBM\n\n- FY25 Q3 revenue estimated at approximately $1.68B, with Q/Q growth of ~+50%.\n- FY25 Q4 estimated revenue around $2.3B, with Q/Q growth projected at over +40%.\n- Gross Margin: Estimated to be mid-50% or higher.\n\nCore DRAM\n\n- ASP Q/Q -4% (due to increased proportion of consumer products).\n- DRAM bit shipment growth for full FY25 upgraded to high-teens %.\n\nNAND\n- FY25 Q3: Bit +24% Q/Q, ASP -6.5% Q/Q.\n- NAND bit growth for full FY25 maintained at +20%.\n\n🔩 Industry/Demand Outlook\n- AI data center demand is driving the market: CNBU (Compute & Networking Business Unit) accounts for 55% of total revenue.\n\n- PC·Mobile: Demand recovery is slow, with slight improvements in mobile DRAM/NAND.\n\n- Industrial/Automotive Memory: Signs of demand recovery are beginning.\n\n📈 Valuation Details\n\n- HBM: For HBM, the applied Revenue (C26E+C27E average) is $16.6B, with a 5x P/S multiple, resulting in a per-share value of $68.\n- Core DRAM+NAND: For Core DRAM+NAND, the applied Revenue (C26E+C27E average) is $36.6B, with a 3x P/S multiple, resulting in a per-share value of $87.\n- Total: The total combined per-share value is $155.\n\n📉 Upside/Downside Scenarios\n\n- Upside Scenario: Target price of $195. Key assumptions include HBM ASP +16%, and DRAM/NAND bit growth of +20% or more.\n- Base Scenario: Target price of $155. Key assumptions maintain current estimates (HBM 5x, DRAM/NAND 3x P/S).\n- Downside Scenario: Target price of $95. Key assumptions include HBM ASP -14%, DRAM ASP -16%, and slowing NAND growth.\n\n🔎 Other Noteworthy Points\n- FY25 CapEx is maintained at $14B, with greenfield capacity increasing from Q4.\n- Share buybacks are limited due to CHIPS Act restrictions.\n- UBS considers the upcoming separate disclosure of HBM quarterly margins (expected) as a strong momentum factor.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00992064949241711, "ret_signed_m1d": 0.00992064949241711, "alpha_spy_m1d": 0.01768369302316386, "alpha_c_m1d": 0.017135688419591344, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009841259532637148, "ret_signed_p1d": -0.009841259532637148, "alpha_spy_p1d": -0.014809591296642388, "alpha_c_p1d": -0.014242418921685207, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.02944444933381929, "ret_signed_p1w": -0.02944444933381929, "alpha_spy_p1w": -0.05145900384583868, "alpha_c_p1w": -0.05260462158565393, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": -0.11615291307436704, "ret_signed_p1m": -0.11615291307436704, "alpha_spy_p1m": -0.15738720973371012, "alpha_c_p1m": -0.15327403601108536, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.32195742379580516, "ret_signed_p3m": 0.32195742379580516, "alpha_spy_p3m": 0.23504551943629215, "alpha_c_p3m": 0.161423497883884, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 1.1137855991590007, "ret_signed_p6m": 1.1137855991590007, "alpha_spy_p6m": 0.9950943986632539, "alpha_c_p6m": 0.8286846054001149, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3104, -0.296, -0.2968, -0.41, -0.4863, -0.4574, -0.4798, -0.382, -0.444, -0.448, -0.4363, -0.4362, -0.4498, -0.454, -0.454, -0.4703, -0.4428, -0.4212, -0.3856, -0.3668, -0.3765, -0.3898, -0.3893, -0.3828, -0.3594, -0.3617, -0.361, -0.3443, -0.3242, -0.3186, -0.2675, -0.2307, -0.2435, -0.2425, -0.2222, -0.2171, -0.2214, -0.2394, -0.2474, -0.259, -0.259, -0.2351, -0.2367, -0.2317, -0.2503, -0.2208, -0.1885, -0.1806, -0.1564, -0.1384, -0.1194, -0.0941, -0.0791, -0.0779, -0.0825, -0.0489, -0.0449, -0.0332, -0.0332, -0.019, -0.0311, 0.0152, 0.0099, 0.0, -0.0098, -0.0218, -0.0406, -0.0338, -0.0294, -0.0294, -0.0474, -0.0116, -0.0289, -0.022, -0.0107, -0.0578, -0.0458, -0.0751, -0.1003, -0.0913, -0.1005, -0.1324, -0.1275, -0.1124, -0.1162, -0.1162, -0.1106, -0.0885, -0.133, -0.1668, -0.1439, -0.1336, -0.1359, -0.1113, -0.0555, -0.0172, 0.0148, -0.0128, -0.0047, -0.0398, -0.0185, -0.0304, -0.0689, -0.0802, -0.0652, -0.0752, -0.0745, -0.0646, -0.0308, -0.0546, -0.0546, -0.0588, -0.0569, -0.0133, 0.0436, 0.0443, 0.0743, 0.1122, 0.1961, 0.249, 0.2533, 0.2617, 0.271, 0.3417, 0.2927, 0.3077, 0.322, 0.2846, 0.2459, 0.2493, 0.302, 0.3292, 0.447, 0.4597, 0.4931, 0.5179, 0.476, 0.5623, 0.5288, 0.4435, 0.5323, 0.4869, 0.5257, 0.6099, 0.6087, 0.6436, 0.608, 0.5776, 0.6431, 0.741, 0.7496, 0.764, 0.8015, 0.7806, 0.7787, 0.8656, 0.7331, 0.8879, 0.8945, 0.8912, 1.0135, 0.9166, 0.9467, 0.8835, 0.962, 0.9232, 0.8163, 0.7958, 0.6007, 0.6484, 0.78, 0.7848, 0.8303, 0.8303, 0.8798, 0.9114, 0.9037, 0.8613, 0.8016, 0.8857, 0.9628, 1.0065, 1.0962, 1.0545, 0.9168, 0.8879, 0.8482, 0.7926, 0.9757, 1.1138]}
{"unit_id": "thread:1939489230071439466", "ticker_idx": 0, "ts": "2025-06-30T01:00:41+00:00", "symbol": "Lens Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I judge that Lens Technology has significant potential. It is particularly attractive in that the possibility of its H-shares trading at a premium to its A-shares upon listing cannot be ruled out.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish on Lens Technology as Apple's foldable hinge/UTG turnkey supplier; H-shares seen as having significant upside potential on HK listing.", "resolved_tickers": ["300433.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Lens Technology 300433.SZ China Shenzhen glass covers", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "What is the Significance of Lens Technology's Hong Kong Listing?\n\nWritten by: Jukan C\n\nThe Alpha and Omega of Apple's Foldable SCM\n\nIt is an open secret within the industry that Lens Technology is virtually confirmed to become the largest supplier, or at the very least the first vendor, for the internal hinges of Apple's foldable products.\n\nApple is reportedly highly concerned that using Samsung's foldable supply chain could lead to the transfer of foldable-related know-how and technology to Samsung. Consequently, it is understood that Apple prefers Lens Technology, which has no prior business dealings with Samsung.\n\nLens Technology is also said to be responding favorably to Apple by most actively pursuing capacity (CAPA) expansion.\n\nThis Hong Kong IPO can also be interpreted as a move to raise funds for increasing CAPA.\n\nFurthermore, it can be seen as a strategic decision to facilitate easier investment for foreign investors looking to participate in the Apple value chain.\n\nAccording to recent supply chain sources, China's Lens Technology has reportedly proposed a turnkey solution to Apple. This proposal includes supplying the protective film for the top of the UTG (Ultra-Thin Glass) and handling the BM (Black Matrix) printing process.\n\nBM is the process of applying a black masking to the edges of the protective film. It serves to visually conceal the module components behind the OLED screen. For foldable OLED protective films, the durability of the BM is critical, as it must not detach despite repeated bending.\n\nPreviously, Apple held Sekyung Hi-tech's BM printing technology in the highest regard. Sekyung Hi-tech has an exclusive supply agreement with Japan's Alps Alpine for BM thermal transfer equipment. Because of this, the industry anticipated that Sekyung Hi-tech would easily join Apple's foldable SCM, despite being a partner of Samsung Electronics.\n\nHowever, a Chinese component industry expert explained, “Lens Technology made a turnkey proposal to Apple using a non-thermal transfer method for BM printing.” He added, “They offered a significantly lower unit price in exchange for securing the entire volume for both UTG and protective films, which has essentially swayed Apple’s decision-makers.”\n\nPersonally, I judge that Lens Technology has significant potential. It is particularly attractive in that the possibility of its H-shares trading at a premium to its A-shares upon listing cannot be ruled out.\n\nThis is not a recommendation to buy or sell.\n\n---\n\nhttps://t.co/PGKuXUbu5I", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010762332146434295, "ret_signed_m1d": -0.010762332146434295, "alpha_spy_m1d": -0.0060037780813798625, "alpha_c_m1d": -0.0010479119131036052, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.00971442023333069, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0013452056997588224, "ret_signed_p1d": -0.0013452056997588224, "alpha_spy_p1d": -0.0010215263915339667, "alpha_c_p1d": 0.007579498543949681, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.008924704243708503, "ret_p1w": 0.02780271998447037, "ret_signed_p1w": 0.02780271998447037, "alpha_spy_p1w": 0.0232223937687277, "alpha_c_p1w": 0.02112888478832975, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.006673835196140621, "ret_p1m": 0.042152524785897816, "ret_signed_p1m": 0.042152524785897816, "alpha_spy_p1m": 0.013974293350111688, "alpha_c_p1m": -0.0007333550522772292, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.042885879838175045, "ret_p3m": 0.5076233043006333, "ret_signed_p3m": 0.5076233043006333, "alpha_spy_p3m": 0.4396062171414368, "alpha_c_p3m": 0.40839612243720236, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.09922718186343094, "ret_p6m": 0.2796189622255083, "ret_signed_p6m": 0.2796189622255083, "alpha_spy_p6m": 0.15975849201680647, "alpha_c_p6m": 0.1236798508831769, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.1559391113423314, "price_path": [0.113, -0.0058, -0.0058, -0.2048, -0.2128, -0.2299, -0.1617, -0.1124, -0.1194, -0.1388, -0.167, -0.1621, -0.1252, -0.1075, -0.1384, -0.0939, -0.104, -0.0807, -0.0939, -0.1053, -0.0838, -0.0838, -0.0838, -0.0838, -0.0534, -0.0688, -0.0516, -0.0767, 0.0122, -0.0072, -0.0208, -0.0463, -0.0345, -0.0543, -0.0503, -0.0587, -0.0688, -0.0767, -0.0758, -0.0965, -0.0908, -0.0758, -0.1075, -0.1075, -0.1146, -0.0952, -0.0367, -0.0375, -0.0296, -0.0503, -0.049, -0.0521, -0.0723, -0.0477, -0.0305, -0.0121, -0.03, -0.0413, -0.0493, -0.0179, -0.0112, -0.0314, -0.0108, 0.0, -0.0013, -0.0247, 0.0865, 0.0646, 0.0278, 0.0462, 0.0175, 0.0085, 0.0175, 0.043, 0.0404, 0.0332, 0.0816, 0.0601, 0.078, 0.0735, 0.057, 0.0664, 0.0592, 0.0502, 0.0422, 0.0215, 0.0229, 0.0036, 0.004, 0.026, 0.0422, 0.0534, 0.026, 0.0879, 0.0803, 0.0942, 0.0996, 0.104, 0.1556, 0.1489, 0.2112, 0.2709, 0.3067, 0.3049, 0.3587, 0.4081, 0.4269, 0.3951, 0.4502, 0.3466, 0.3377, 0.3543, 0.4081, 0.4798, 0.3377, 0.3251, 0.3502, 0.3224, 0.3619, 0.3587, 0.4395, 0.4085, 0.3682, 0.4821, 0.439, 0.4942, 0.5076, 0.4341, 0.4619, 0.5013, 0.5013, 0.5013, 0.5013, 0.5013, 0.5013, 0.5013, 0.5188, 0.4305, 0.3556, 0.3135, 0.3202, 0.2821, 0.2265, 0.2345, 0.2973, 0.3206, 0.2973, 0.335, 0.3399, 0.3354, 0.3881, 0.3377, 0.3242, 0.3305, 0.2864, 0.2837, 0.407, 0.4002, 0.3431, 0.2958, 0.3422, 0.312, 0.2553, 0.2252, 0.2139, 0.1946, 0.1955, 0.1806, 0.1923, 0.2211, 0.236, 0.2229, 0.2373, 0.2895, 0.3035, 0.2697, 0.2742, 0.2985, 0.3219, 0.3165, 0.3003, 0.3791, 0.3498, 0.3048, 0.2954, 0.3341, 0.2864, 0.2918, 0.2994, 0.2796]}
{"unit_id": "orig:1923525477349888055", "ticker_idx": 0, "ts": "2025-05-16T23:46:25+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Despite market rumors suggesting a possible reduction in Blackwell packaging volume to 340–350K units, Morgan Stanley's supply chain checks indicate otherwise. It is projected that NVIDIA will consume 160K CoWoS-L substrates in H1 2025, and over 100K more in Q3", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS supply chain note: NVDA CoWoS-L demand intact at 390K vs rumored cuts; TSMC wafer output ~5.1M; AMD named Humain partner; MediaTek WoA PC chip unlikely to launch near term.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Key Points on the AI Supply Chain: Large-Scale Middle Eastern AI Project, CoWoS-L Capacity Forecast, and Developments in China and Taiwan\nMorgan Stanley (C. Chan, 2025/05/16) \n(1) Launch of the Middle East Humain AI Project – 500MW Compute Deployment\nSaudi Arabia has announced government-backed support for the establishment of the AI company \"Humain,\" with a plan to invest $10 billion over the next five years to deploy 500 megawatts of AI computing power. The project will be primarily led by SMI as the ODM and executed in two phases. The first phase will involve the deployment of 18,000 NVIDIA GB300 Grace Blackwell supercomputing nodes, equipped with InfiniBand networks. Channel checks reveal that the nodes will come in both NVL72 and HGX configurations, with deployment expected to continue until Q1 2026. AMD’s GB300 has also been listed as a strategic partner.\n(2) 2025 CoWoS-L Packaging Forecast Maintained at 390K Units\nDespite market rumors suggesting a possible reduction in Blackwell packaging volume to 340–350K units, Morgan Stanley's supply chain checks indicate otherwise. It is projected that NVIDIA will consume 160K CoWoS-L substrates in H1 2025, and over 100K more in Q3. Approximately 94% of packaging volume is dedicated to the dual-die B200/B300 series, translating into an estimated annual wafer output of about 5.1 million units at TSMC. The Greater China downstream team expects 6,000–9,000 GB200 NVL72 units to be shipped in H1 2025, consuming roughly 500 B200 chips. Additionally, 45,000–50,000 HGX systems (8 GPUs per unit) are projected to be shipped, consuming another ~400 chips. With rack integration improving and capacity utilization increasing, further inventory digestion is expected in Q3 2025. The key uncertainty remains whether rack-level integration and assembly can accelerate in parallel.\n\n(3) China AI GPU Sanctions and Inventory Preparation Trends\nOn May 13, the U.S. Department of Commerce issued new regulations, signaling growing demand for the B30 chip designed for inference applications in China. Morgan Stanley estimates a small volume of wafers may be produced in Q2 2025, rising to as many as 500,000 chips by Q3. Due to the H20/HBM version ban and Amkor’s cancellation of approximately 30,000 CoWoS-S packaging orders, inventories are expected to shift toward the H200 model, and the surplus will need to be redistributed within alternative supply chains.\n(4) Preview of Taipei Computex\nNext week’s Taipei Computex will unveil several new AI hardware products. However, Morgan Stanley believes the “WoA AI PC” chip jointly developed by NVIDIA and MediaTek still lacks sufficient software ecosystem support and is unlikely to make a formal debut in the near term. The event will mainly focus on integration of technologies across the AI server and PC ecosystems and the practical applications of these technologies.\n\n(5) Steady Demand, Rack Integration as a Core Bottleneck\nOverall, demand for AI servers remains robust—from large-scale orders in the Middle East to ongoing global rack deployment, compute power applications continue to expand. The real bottleneck in the supply chain is no longer chip production capacity but the complexity of hardware-software integration and system testing timelines. Key areas of focus going forward will be whether rack assembly can scale as scheduled and the pace of inventory digestion across various supply chain stages.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004209667739918155, "ret_signed_m1d": -0.004209667739918155, "alpha_spy_m1d": 0.0020844392425499736, "alpha_c_m1d": -0.007090913350360495, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": 0.00288124561044234, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0012554832990736564, "ret_signed_p1d": 0.0012554832990736564, "alpha_spy_p1d": 0.00016156519903565147, "alpha_c_p1d": 0.0029193458042889864, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.00166386250521533, "ret_p1w": -0.030354560747671955, "ret_signed_p1w": -0.030354560747671955, "alpha_spy_p1w": -0.004959120627194502, "alpha_c_p1w": 0.006087301553434954, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03644186230110691, "ret_p1m": 0.06868569001533453, "ret_signed_p1m": 0.06868569001533453, "alpha_spy_p1m": 0.054414365287684285, "alpha_c_p1m": -0.00013990455913326372, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.0688255945744678, "ret_p3m": 0.34123039617723494, "ret_signed_p3m": 0.34123039617723494, "alpha_spy_p3m": 0.2527138349387563, "alpha_c_p3m": 0.11860233817690236, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.22262805800033258, "ret_p6m": 0.47027363178471604, "ret_signed_p6m": 0.47027363178471604, "alpha_spy_p6m": 0.3168750207666782, "alpha_c_p6m": 0.014263561276624648, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.4560100705080914, "price_path": [0.0294, 0.0282, 0.0347, -0.0073, -0.0379, -0.0649, -0.0305, -0.1127, -0.0775, -0.1577, -0.1434, -0.1338, -0.1835, -0.1678, -0.21, -0.1968, -0.1452, -0.1464, -0.1014, -0.1172, -0.1475, -0.1321, -0.1246, -0.1307, -0.1033, -0.1086, -0.1598, -0.177, -0.19, -0.1996, -0.1865, -0.1845, -0.2482, -0.3035, -0.2789, -0.2888, -0.1556, -0.2055, -0.1807, -0.1823, -0.1713, -0.2283, -0.2504, -0.2504, -0.2843, -0.2696, -0.2414, -0.214, -0.1801, -0.197, -0.1948, -0.1956, -0.1757, -0.1544, -0.1594, -0.1614, -0.1355, -0.1332, -0.1385, -0.0916, -0.0404, -0.0004, -0.0042, 0.0, 0.0013, -0.0075, -0.0266, -0.019, -0.0304, -0.0304, 0.0007, -0.0044, 0.028, -0.002, 0.0146, 0.043, 0.0482, 0.0339, 0.0467, 0.0534, 0.0632, 0.0549, 0.071, 0.0486, 0.0687, 0.0645, 0.0745, 0.0745, 0.0625, 0.0648, 0.0924, 0.1397, 0.145, 0.1651, 0.1669, 0.1323, 0.1615, 0.1769, 0.1769, 0.1688, 0.1818, 0.203, 0.212, 0.2181, 0.2118, 0.2608, 0.2657, 0.2778, 0.2734, 0.2658, 0.2337, 0.2614, 0.2833, 0.2815, 0.3055, 0.2963, 0.3241, 0.3138, 0.2831, 0.3295, 0.3166, 0.3252, 0.3352, 0.3494, 0.3447, 0.3528, 0.3412, 0.3444, 0.3328, 0.3443, 0.2973, 0.2955, 0.2924, 0.3146, 0.3281, 0.3426, 0.3413, 0.3307, 0.2865, 0.2865, 0.2614, 0.2602, 0.2679, 0.2336, 0.2431, 0.2612, 0.3098, 0.3087, 0.3135, 0.3129, 0.2917, 0.2578, 0.3018, 0.305, 0.3562, 0.318, 0.3072, 0.3125, 0.3162, 0.3432, 0.3782, 0.383, 0.3952, 0.3858, 0.3705, 0.3668, 0.3969, 0.4224, 0.3529, 0.391, 0.3298, 0.3283, 0.3429, 0.3533, 0.3491, 0.3381, 0.3316, 0.3455, 0.3758, 0.4144, 0.4849, 0.5293, 0.4986, 0.4957, 0.5281, 0.4676, 0.4419, 0.3892, 0.3898, 0.4703]}
{"unit_id": "orig:1923525477349888055", "ticker_idx": 1, "ts": "2025-05-16T23:46:25+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Approximately 94% of packaging volume is dedicated to the dual-die B200/B300 series, translating into an estimated annual wafer output of about 5.1 million units at TSMC", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS supply chain note: NVDA CoWoS-L demand intact at 390K vs rumored cuts; TSMC wafer output ~5.1M; AMD named Humain partner; MediaTek WoA PC chip unlikely to launch near term.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Key Points on the AI Supply Chain: Large-Scale Middle Eastern AI Project, CoWoS-L Capacity Forecast, and Developments in China and Taiwan\nMorgan Stanley (C. Chan, 2025/05/16) \n(1) Launch of the Middle East Humain AI Project – 500MW Compute Deployment\nSaudi Arabia has announced government-backed support for the establishment of the AI company \"Humain,\" with a plan to invest $10 billion over the next five years to deploy 500 megawatts of AI computing power. The project will be primarily led by SMI as the ODM and executed in two phases. The first phase will involve the deployment of 18,000 NVIDIA GB300 Grace Blackwell supercomputing nodes, equipped with InfiniBand networks. Channel checks reveal that the nodes will come in both NVL72 and HGX configurations, with deployment expected to continue until Q1 2026. AMD’s GB300 has also been listed as a strategic partner.\n(2) 2025 CoWoS-L Packaging Forecast Maintained at 390K Units\nDespite market rumors suggesting a possible reduction in Blackwell packaging volume to 340–350K units, Morgan Stanley's supply chain checks indicate otherwise. It is projected that NVIDIA will consume 160K CoWoS-L substrates in H1 2025, and over 100K more in Q3. Approximately 94% of packaging volume is dedicated to the dual-die B200/B300 series, translating into an estimated annual wafer output of about 5.1 million units at TSMC. The Greater China downstream team expects 6,000–9,000 GB200 NVL72 units to be shipped in H1 2025, consuming roughly 500 B200 chips. Additionally, 45,000–50,000 HGX systems (8 GPUs per unit) are projected to be shipped, consuming another ~400 chips. With rack integration improving and capacity utilization increasing, further inventory digestion is expected in Q3 2025. The key uncertainty remains whether rack-level integration and assembly can accelerate in parallel.\n\n(3) China AI GPU Sanctions and Inventory Preparation Trends\nOn May 13, the U.S. Department of Commerce issued new regulations, signaling growing demand for the B30 chip designed for inference applications in China. Morgan Stanley estimates a small volume of wafers may be produced in Q2 2025, rising to as many as 500,000 chips by Q3. Due to the H20/HBM version ban and Amkor’s cancellation of approximately 30,000 CoWoS-S packaging orders, inventories are expected to shift toward the H200 model, and the surplus will need to be redistributed within alternative supply chains.\n(4) Preview of Taipei Computex\nNext week’s Taipei Computex will unveil several new AI hardware products. However, Morgan Stanley believes the “WoA AI PC” chip jointly developed by NVIDIA and MediaTek still lacks sufficient software ecosystem support and is unlikely to make a formal debut in the near term. The event will mainly focus on integration of technologies across the AI server and PC ecosystems and the practical applications of these technologies.\n\n(5) Steady Demand, Rack Integration as a Core Bottleneck\nOverall, demand for AI servers remains robust—from large-scale orders in the Middle East to ongoing global rack deployment, compute power applications continue to expand. The real bottleneck in the supply chain is no longer chip production capacity but the complexity of hardware-software integration and system testing timelines. Key areas of focus going forward will be whether rack assembly can scale as scheduled and the pace of inventory digestion across various supply chain stages.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005009983967306453, "ret_signed_m1d": -0.005009983967306453, "alpha_spy_m1d": 0.0012841230151616756, "alpha_c_m1d": -0.007891229577748793, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": 0.00288124561044234, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014028017178506658, "ret_signed_p1d": -0.014028017178506658, "alpha_spy_p1d": -0.015121935278544663, "alpha_c_p1d": -0.012364154673291328, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.00166386250521533, "ret_p1w": -0.014028017178506658, "ret_signed_p1w": -0.014028017178506658, "alpha_spy_p1w": 0.011367422941970795, "alpha_c_p1w": 0.02241384512260025, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03644186230110691, "ret_p1m": 0.031412286021266445, "ret_signed_p1m": 0.031412286021266445, "alpha_spy_p1m": 0.017140961293616197, "alpha_c_p1m": -0.03741330855320135, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.0688255945744678, "ret_p3m": 0.2075071240898203, "ret_signed_p3m": 0.2075071240898203, "alpha_spy_p3m": 0.11899056285134169, "alpha_c_p3m": -0.01512093391051228, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.22262805800033258, "ret_p6m": 0.49016431803948035, "ret_signed_p6m": 0.49016431803948035, "alpha_spy_p6m": 0.3367657070214425, "alpha_c_p6m": 0.03415424753138896, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.4560100705080914, "price_path": [0.0971, 0.0871, 0.0771, 0.0921, 0.0722, 0.0522, 0.0572, 0.0372, 0.0372, 0.0173, -0.0026, 0.0173, 0.0023, 0.0023, -0.0046, -0.0316, -0.0146, -0.0376, -0.0435, -0.0326, -0.0271, -0.0461, -0.013, -0.0261, -0.0261, -0.008, -0.018, -0.0401, -0.0461, -0.0882, -0.0541, -0.0561, -0.0561, -0.0561, -0.1503, -0.1824, -0.2134, -0.1353, -0.1092, -0.1333, -0.1212, -0.1433, -0.1513, -0.1483, -0.1633, -0.1824, -0.1253, -0.1343, -0.1102, -0.1012, -0.0962, -0.0902, -0.0902, -0.0481, -0.0601, -0.0782, -0.0701, -0.0802, -0.0491, -0.0411, -0.0291, 0.001, -0.005, 0.0, -0.014, -0.02, -0.008, -0.016, -0.014, -0.022, -0.0331, -0.0311, -0.0311, -0.0311, -0.0521, -0.0481, -0.008, 0.0, -0.003, 0.007, 0.0471, 0.0671, 0.0515, 0.0364, 0.0314, 0.0515, 0.0616, 0.0415, 0.0616, 0.0264, 0.0566, 0.0767, 0.0817, 0.0868, 0.0666, 0.0918, 0.0918, 0.0968, 0.0918, 0.0868, 0.0868, 0.0968, 0.1069, 0.1069, 0.1019, 0.1169, 0.1371, 0.1371, 0.1622, 0.1572, 0.1371, 0.1522, 0.1522, 0.1522, 0.1522, 0.1421, 0.1622, 0.1673, 0.1673, 0.1421, 0.1572, 0.132, 0.1874, 0.1824, 0.1874, 0.1874, 0.2075, 0.1824, 0.1874, 0.1874, 0.1924, 0.1421, 0.1572, 0.1421, 0.1773, 0.1824, 0.1974, 0.1673, 0.1673, 0.1723, 0.1673, 0.1673, 0.1673, 0.1874, 0.1874, 0.2075, 0.2327, 0.2478, 0.2679, 0.2629, 0.2932, 0.278, 0.2982, 0.278, 0.3083, 0.3538, 0.3538, 0.3336, 0.3134, 0.3134, 0.3184, 0.3386, 0.379, 0.4144, 0.4144, 0.4498, 0.4295, 0.4548, 0.4548, 0.4295, 0.4397, 0.4801, 0.5003, 0.4649, 0.4952, 0.4952, 0.475, 0.4649, 0.4649, 0.4952, 0.4902, 0.5205, 0.5205, 0.5154, 0.5255, 0.5205, 0.475, 0.4801, 0.475, 0.4902]}
{"unit_id": "orig:1923525477349888055", "ticker_idx": 2, "ts": "2025-05-16T23:46:25+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's GB300 has also been listed as a strategic partner", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS supply chain note: NVDA CoWoS-L demand intact at 390K vs rumored cuts; TSMC wafer output ~5.1M; AMD named Humain partner; MediaTek WoA PC chip unlikely to launch near term.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Key Points on the AI Supply Chain: Large-Scale Middle Eastern AI Project, CoWoS-L Capacity Forecast, and Developments in China and Taiwan\nMorgan Stanley (C. Chan, 2025/05/16) \n(1) Launch of the Middle East Humain AI Project – 500MW Compute Deployment\nSaudi Arabia has announced government-backed support for the establishment of the AI company \"Humain,\" with a plan to invest $10 billion over the next five years to deploy 500 megawatts of AI computing power. The project will be primarily led by SMI as the ODM and executed in two phases. The first phase will involve the deployment of 18,000 NVIDIA GB300 Grace Blackwell supercomputing nodes, equipped with InfiniBand networks. Channel checks reveal that the nodes will come in both NVL72 and HGX configurations, with deployment expected to continue until Q1 2026. AMD’s GB300 has also been listed as a strategic partner.\n(2) 2025 CoWoS-L Packaging Forecast Maintained at 390K Units\nDespite market rumors suggesting a possible reduction in Blackwell packaging volume to 340–350K units, Morgan Stanley's supply chain checks indicate otherwise. It is projected that NVIDIA will consume 160K CoWoS-L substrates in H1 2025, and over 100K more in Q3. Approximately 94% of packaging volume is dedicated to the dual-die B200/B300 series, translating into an estimated annual wafer output of about 5.1 million units at TSMC. The Greater China downstream team expects 6,000–9,000 GB200 NVL72 units to be shipped in H1 2025, consuming roughly 500 B200 chips. Additionally, 45,000–50,000 HGX systems (8 GPUs per unit) are projected to be shipped, consuming another ~400 chips. With rack integration improving and capacity utilization increasing, further inventory digestion is expected in Q3 2025. The key uncertainty remains whether rack-level integration and assembly can accelerate in parallel.\n\n(3) China AI GPU Sanctions and Inventory Preparation Trends\nOn May 13, the U.S. Department of Commerce issued new regulations, signaling growing demand for the B30 chip designed for inference applications in China. Morgan Stanley estimates a small volume of wafers may be produced in Q2 2025, rising to as many as 500,000 chips by Q3. Due to the H20/HBM version ban and Amkor’s cancellation of approximately 30,000 CoWoS-S packaging orders, inventories are expected to shift toward the H200 model, and the surplus will need to be redistributed within alternative supply chains.\n(4) Preview of Taipei Computex\nNext week’s Taipei Computex will unveil several new AI hardware products. However, Morgan Stanley believes the “WoA AI PC” chip jointly developed by NVIDIA and MediaTek still lacks sufficient software ecosystem support and is unlikely to make a formal debut in the near term. The event will mainly focus on integration of technologies across the AI server and PC ecosystems and the practical applications of these technologies.\n\n(5) Steady Demand, Rack Integration as a Core Bottleneck\nOverall, demand for AI servers remains robust—from large-scale orders in the Middle East to ongoing global rack deployment, compute power applications continue to expand. The real bottleneck in the supply chain is no longer chip production capacity but the complexity of hardware-software integration and system testing timelines. Key areas of focus going forward will be whether rack assembly can scale as scheduled and the pace of inventory digestion across various supply chain stages.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01860544797510777, "ret_signed_m1d": -0.01860544797510777, "alpha_spy_m1d": -0.012311340992639641, "alpha_c_m1d": -0.02148669358555011, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": 0.00288124561044234, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020739099967143515, "ret_signed_p1d": -0.020739099967143515, "alpha_spy_p1d": -0.02183301806718152, "alpha_c_p1d": -0.019075237461928185, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.00166386250521533, "ret_p1w": -0.05854741587057333, "ret_signed_p1w": -0.05854741587057333, "alpha_spy_p1w": -0.033151975750095874, "alpha_c_p1w": -0.02210555356946642, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03644186230110691, "ret_p1m": 0.07868909588450257, "ret_signed_p1m": 0.07868909588450257, "alpha_spy_p1m": 0.06441777115685232, "alpha_c_p1m": 0.009863501310034772, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.0688255945744678, "ret_p3m": 0.5739523858576274, "ret_signed_p3m": 0.5739523858576274, "alpha_spy_p3m": 0.4854358246191488, "alpha_c_p3m": 0.35132432785729484, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.22262805800033258, "ret_p6m": 1.0822736156037887, "ret_signed_p6m": 1.0822736156037887, "alpha_spy_p6m": 0.9288750045857508, "alpha_c_p6m": 0.6262635450956973, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.4560100705080914, "price_path": [-0.0247, -0.0212, -0.0256, -0.054, -0.0773, -0.1127, -0.1061, -0.1507, -0.1477, -0.1616, -0.1401, -0.1323, -0.1564, -0.1439, -0.1753, -0.1742, -0.1398, -0.1627, -0.1383, -0.1074, -0.1166, -0.0934, -0.0856, -0.0916, -0.0283, -0.0201, -0.0596, -0.0898, -0.1191, -0.1232, -0.1228, -0.1213, -0.1995, -0.2681, -0.2862, -0.3325, -0.1735, -0.243, -0.2029, -0.1935, -0.1867, -0.2465, -0.2532, -0.2532, -0.2698, -0.2638, -0.2286, -0.1937, -0.1751, -0.1773, -0.1802, -0.1692, -0.1751, -0.1568, -0.1415, -0.1583, -0.1435, -0.132, -0.1223, -0.0772, -0.0402, 0.0047, -0.0186, 0.0, -0.0207, -0.0312, -0.0436, -0.0551, -0.0585, -0.0585, -0.0223, -0.0368, -0.0353, -0.055, -0.0217, 0.0012, 0.012, -0.0126, -0.0084, 0.0389, 0.0518, 0.0339, 0.0114, -0.0086, 0.0787, 0.0847, 0.0821, 0.0821, 0.0945, 0.1059, 0.1814, 0.2239, 0.2263, 0.2274, 0.2111, 0.1616, 0.1822, 0.177, 0.177, 0.1505, 0.1762, 0.1813, 0.2303, 0.2496, 0.2481, 0.3281, 0.3662, 0.369, 0.3398, 0.3399, 0.3205, 0.354, 0.3836, 0.4208, 0.4821, 0.5144, 0.532, 0.5047, 0.4654, 0.5087, 0.4877, 0.3922, 0.4714, 0.4744, 0.4703, 0.4931, 0.574, 0.5443, 0.515, 0.5033, 0.4214, 0.4099, 0.3972, 0.4318, 0.3942, 0.422, 0.4264, 0.4388, 0.388, 0.388, 0.3853, 0.3837, 0.3808, 0.2899, 0.2922, 0.3299, 0.3616, 0.3286, 0.3533, 0.3754, 0.3695, 0.3584, 0.3478, 0.3433, 0.3637, 0.3732, 0.373, 0.3764, 0.3609, 0.3771, 0.3808, 0.3998, 0.4486, 0.4054, 0.7386, 0.8052, 1.0104, 0.9876, 0.8341, 0.8471, 0.8613, 1.0364, 1.0019, 0.9892, 1.0531, 1.0315, 0.9649, 1.0055, 1.1586, 1.2162, 1.202, 1.256, 1.175, 1.1859, 1.216, 1.1341, 1.1877, 1.0287, 0.9932, 1.0823]}
{"unit_id": "orig:1923525477349888055", "ticker_idx": 3, "ts": "2025-05-16T23:46:25+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes the 'WoA AI PC' chip jointly developed by NVIDIA and MediaTek still lacks sufficient software ecosystem support and is unlikely to make a formal debut in the near term", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS supply chain note: NVDA CoWoS-L demand intact at 390K vs rumored cuts; TSMC wafer output ~5.1M; AMD named Humain partner; MediaTek WoA PC chip unlikely to launch near term.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Key Points on the AI Supply Chain: Large-Scale Middle Eastern AI Project, CoWoS-L Capacity Forecast, and Developments in China and Taiwan\nMorgan Stanley (C. Chan, 2025/05/16) \n(1) Launch of the Middle East Humain AI Project – 500MW Compute Deployment\nSaudi Arabia has announced government-backed support for the establishment of the AI company \"Humain,\" with a plan to invest $10 billion over the next five years to deploy 500 megawatts of AI computing power. The project will be primarily led by SMI as the ODM and executed in two phases. The first phase will involve the deployment of 18,000 NVIDIA GB300 Grace Blackwell supercomputing nodes, equipped with InfiniBand networks. Channel checks reveal that the nodes will come in both NVL72 and HGX configurations, with deployment expected to continue until Q1 2026. AMD’s GB300 has also been listed as a strategic partner.\n(2) 2025 CoWoS-L Packaging Forecast Maintained at 390K Units\nDespite market rumors suggesting a possible reduction in Blackwell packaging volume to 340–350K units, Morgan Stanley's supply chain checks indicate otherwise. It is projected that NVIDIA will consume 160K CoWoS-L substrates in H1 2025, and over 100K more in Q3. Approximately 94% of packaging volume is dedicated to the dual-die B200/B300 series, translating into an estimated annual wafer output of about 5.1 million units at TSMC. The Greater China downstream team expects 6,000–9,000 GB200 NVL72 units to be shipped in H1 2025, consuming roughly 500 B200 chips. Additionally, 45,000–50,000 HGX systems (8 GPUs per unit) are projected to be shipped, consuming another ~400 chips. With rack integration improving and capacity utilization increasing, further inventory digestion is expected in Q3 2025. The key uncertainty remains whether rack-level integration and assembly can accelerate in parallel.\n\n(3) China AI GPU Sanctions and Inventory Preparation Trends\nOn May 13, the U.S. Department of Commerce issued new regulations, signaling growing demand for the B30 chip designed for inference applications in China. Morgan Stanley estimates a small volume of wafers may be produced in Q2 2025, rising to as many as 500,000 chips by Q3. Due to the H20/HBM version ban and Amkor’s cancellation of approximately 30,000 CoWoS-S packaging orders, inventories are expected to shift toward the H200 model, and the surplus will need to be redistributed within alternative supply chains.\n(4) Preview of Taipei Computex\nNext week’s Taipei Computex will unveil several new AI hardware products. However, Morgan Stanley believes the “WoA AI PC” chip jointly developed by NVIDIA and MediaTek still lacks sufficient software ecosystem support and is unlikely to make a formal debut in the near term. The event will mainly focus on integration of technologies across the AI server and PC ecosystems and the practical applications of these technologies.\n\n(5) Steady Demand, Rack Integration as a Core Bottleneck\nOverall, demand for AI servers remains robust—from large-scale orders in the Middle East to ongoing global rack deployment, compute power applications continue to expand. The real bottleneck in the supply chain is no longer chip production capacity but the complexity of hardware-software integration and system testing timelines. Key areas of focus going forward will be whether rack assembly can scale as scheduled and the pace of inventory digestion across various supply chain stages.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003662967910337489, "ret_signed_m1d": -0.003662967910337489, "alpha_spy_m1d": -0.009957074892805617, "alpha_c_m1d": -0.000781722299895149, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": 0.00288124561044234, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04029301884144054, "ret_signed_p1d": 0.04029301884144054, "alpha_spy_p1d": 0.04138693694147855, "alpha_c_p1d": 0.03862915633622521, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.00166386250521533, "ret_p1w": -0.036630050931103164, "ret_signed_p1w": 0.036630050931103164, "alpha_spy_p1w": 0.011234610810625711, "alpha_c_p1w": 0.0001881886299962554, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03644186230110691, "ret_p1m": -0.0805860376828812, "ret_signed_p1m": 0.0805860376828812, "alpha_spy_p1m": 0.09485736241053144, "alpha_c_p1m": 0.149411632257349, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.0688255945744678, "ret_p3m": 0.023420313134721482, "ret_signed_p3m": -0.023420313134721482, "alpha_spy_p3m": 0.06509624810375714, "alpha_c_p3m": 0.1992077448656111, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.22262805800033258, "ret_p6m": -0.06622233253111498, "ret_signed_p6m": 0.06622233253111498, "alpha_spy_p6m": 0.21962094354915285, "alpha_c_p6m": 0.5222324030392064, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.4560100705080914, "price_path": [0.1245, 0.1282, 0.0989, 0.1062, 0.1319, 0.1355, 0.1429, 0.1099, 0.1099, 0.0769, 0.1026, 0.1026, 0.0952, 0.0733, 0.022, 0.0147, 0.033, 0.0183, 0.0256, 0.044, 0.0806, 0.0733, 0.0879, 0.0842, 0.0842, 0.1172, 0.1099, 0.0879, 0.0733, 0.0183, 0.0696, 0.0513, 0.0513, 0.0513, -0.0513, -0.0769, -0.1319, -0.0476, 0.0147, 0.0256, 0.0147, 0.0, -0.0183, 0.0, -0.0147, -0.0476, 0.0037, -0.0256, 0.011, -0.011, 0.0037, -0.011, -0.011, -0.0476, -0.0513, -0.0623, -0.0659, -0.0549, -0.0293, -0.0366, -0.0073, -0.0037, 0.0037, 0.0, -0.0403, -0.0366, -0.0183, -0.0293, -0.0366, -0.0623, -0.0623, -0.0623, -0.0769, -0.0769, -0.0769, -0.0659, -0.0659, -0.0879, -0.0623, -0.0586, -0.022, -0.0366, -0.0586, -0.0842, -0.0806, -0.0659, -0.0586, -0.0733, -0.0842, -0.0806, -0.0696, -0.0513, -0.0549, -0.0586, -0.0842, -0.0659, -0.0513, -0.0513, -0.0363, -0.0438, -0.0513, 0.0122, 0.0458, 0.0608, 0.0346, 0.0421, 0.0533, 0.0384, 0.0533, 0.0645, 0.0682, 0.0682, 0.0682, 0.0682, 0.0458, 0.016, 0.0309, 0.0234, 0.0234, -0.0065, 0.001, -0.0139, 0.0122, 0.0085, 0.0122, 0.016, 0.0234, 0.0384, 0.0234, 0.0533, 0.0384, 0.0122, 0.0197, 0.0197, 0.0533, 0.0458, 0.0458, 0.0346, 0.0234, 0.016, 0.016, 0.0421, 0.0346, 0.072, 0.0682, 0.128, 0.1131, 0.1056, 0.1093, 0.1093, 0.1467, 0.128, 0.128, 0.0757, 0.0496, 0.0272, -0.0065, 0.0047, -0.0214, -0.0214, -0.0177, -0.0401, -0.0438, -0.0139, -0.0139, -0.0102, -0.0027, 0.0047, 0.0047, -0.0177, -0.0251, -0.0177, -0.0065, -0.0065, 0.001, 0.001, -0.0065, -0.0326, -0.0326, -0.0027, -0.0177, -0.0289, -0.0214, -0.0214, -0.0289, -0.0177, -0.0065, -0.0363, -0.0588, -0.0662]}
{"unit_id": "orig:1930078782918144411", "ticker_idx": 0, "ts": "2025-06-04T01:46:55+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Revenue guidance could approach JP Morgan's $16.1 billion estimate, above consensus... AI Business Expected to Grow 60% in FY2025... FCF expected to grow at double-digit rates over the coming years", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JP Morgan Buy-equivalent preview on AVGO: Q2 in-line/beat, July guide above consensus, 60% AI revenue growth FY2025, double-digit FCF growth.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Broadcom: Q2 Earnings Preview – Continued Strong ASIC Growth Expected\nJP Morgan (H. Sur, June 3, 2025)\n\nBroadcom is scheduled to report its F2Q25 earnings this Thursday, and revenue, earnings, and free cash flow (FCF) are expected to be in line with or slightly above market expectations.\nAI-related products (custom ASICs and networking) continue to show strong growth, with April quarter AI revenue potentially exceeding $5 billion, driving July quarter guidance to increase 5–7% QoQ, close to JP Morgan’s forecast of $16.1 billion.\nGoogle’s 3nm TPU v6 ASIC has entered mass production, with a lifecycle revenue estimate exceeding $15 billion.\nOn the AI networking side, demand remains strong for Tomahawk 5 and Jericho 3 AI, and the next-gen Tomahawk 6 is set to begin mass production in the second half of the year.\nFull-year AI revenue for FY2025 is projected at $19–20 billion, representing 60% YoY growth.\nMeanwhile, non-AI semiconductor businesses are recovering with the broader industry cycle, and wireless has upside potential.\nVMware software renewals remain strong, with customers shifting toward VCF full-stack premium products, continuing to unlock synergy benefits.\n\n⸻\n\n• April Quarter Results Expected In Line or Slightly Better:\n• Strong demand for AI networking and custom ASICs\n• Non-AI segments (telecom, servers, broadband) are stabilizing\n• Healthy VMware renewal momentum supporting overall order strength\n\n⸻\n\n• July Quarter Guidance Likely In Line or Above Expectations:\n• Continued growth in AI business\n• 3nm TPU ASIC project entering production ramp\n• Revenue guidance could approach JP Morgan’s $16.1 billion estimate, above consensus\n• Gradual improvement in non-AI segments expected\n\n⸻\n\n• AI Business Expected to Grow 60% in FY2025:\n• Projected revenue: $19–20 billion\n• Key drivers:\n• Google TPU v6\n• Meta and ByteDance custom ASIC programs\n• Strong demand for AI networking products\n• TPU v6 alone could generate over $15 billion in lifecycle revenue\n\n⸻\n\n• Strong Free Cash Flow Growth & Enhanced Capital Returns:\n• FCF expected to grow at double-digit rates over the coming years\n• Supporting dividend increases and deleveraging\n• Enhancing overall profitability", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016201804129663877, "ret_signed_m1d": -0.016201804129663877, "alpha_spy_m1d": -0.016470423669710188, "alpha_c_m1d": -0.00444409734000395, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004404742321395871, "ret_signed_p1d": -0.004404742321395871, "alpha_spy_p1d": 0.00042802673479336306, "alpha_c_p1d": -0.0026172377082356313, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": -0.031292995729120365, "ret_signed_p1w": -0.031292995729120365, "alpha_spy_p1w": -0.04040492689124542, "alpha_c_p1w": -0.0715313053577511, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.05648734408414113, "ret_signed_p1m": 0.05648734408414113, "alpha_spy_p1m": 0.004033603518198259, "alpha_c_p1m": -0.07010652939579587, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.14175740498695077, "ret_signed_p3m": 0.14175740498695077, "alpha_spy_p3m": 0.05613158568699417, "alpha_c_p3m": -0.01133105642757326, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.5289895205311899, "ret_signed_p6m": 0.5289895205311899, "alpha_spy_p6m": 0.3819096003649587, "alpha_c_p6m": 0.14781778085752495, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.2555, -0.2956, -0.2741, -0.2583, -0.2693, -0.2533, -0.2573, -0.2795, -0.2532, -0.2702, -0.2659, -0.2675, -0.2789, -0.3134, -0.3412, -0.3522, -0.3587, -0.3545, -0.3409, -0.4101, -0.4397, -0.4096, -0.4024, -0.2908, -0.34, -0.3031, -0.3168, -0.3146, -0.3312, -0.3451, -0.3451, -0.3634, -0.3505, -0.3224, -0.2793, -0.2634, -0.2628, -0.2678, -0.2628, -0.2442, -0.22, -0.2312, -0.2336, -0.2155, -0.2042, -0.2025, -0.1513, -0.1098, -0.1109, -0.1089, -0.1244, -0.1166, -0.1126, -0.1201, -0.117, -0.1239, -0.1239, -0.0974, -0.0829, -0.0732, -0.0728, -0.0474, -0.0162, 0.0, -0.0044, -0.0542, -0.0643, -0.063, -0.0313, -0.0192, -0.0474, -0.0344, -0.0449, -0.0376, -0.0376, -0.0402, -0.0257, 0.0127, 0.0161, 0.0373, 0.0341, 0.0583, 0.0164, 0.0362, 0.0565, 0.0565, 0.0526, 0.0435, 0.0669, 0.0573, 0.0534, 0.0581, 0.0786, 0.0781, 0.0998, 0.0878, 0.1065, 0.0696, 0.0892, 0.1084, 0.1141, 0.1299, 0.1419, 0.1618, 0.1276, 0.1082, 0.143, 0.1246, 0.1582, 0.1662, 0.1709, 0.1668, 0.201, 0.1867, 0.1949, 0.1761, 0.1739, 0.1322, 0.1179, 0.1118, 0.1287, 0.1296, 0.1441, 0.1527, 0.185, 0.1418, 0.1418, 0.145, 0.161, 0.1752, 0.2857, 0.327, 0.2926, 0.4189, 0.3807, 0.3816, 0.3978, 0.3821, 0.329, 0.3259, 0.3243, 0.3029, 0.3035, 0.3049, 0.2926, 0.2865, 0.2611, 0.2688, 0.2822, 0.3006, 0.3013, 0.2902, 0.2938, 0.3287, 0.3269, 0.2485, 0.3718, 0.3235, 0.3512, 0.362, 0.3435, 0.3431, 0.3178, 0.3087, 0.3241, 0.3619, 0.3924, 0.4344, 0.4844, 0.4478, 0.4215, 0.3943, 0.3535, 0.3806, 0.3675, 0.3439, 0.3783, 0.3536, 0.3661, 0.3075, 0.317, 0.3178, 0.3095, 0.363, 0.3338, 0.3084, 0.4536, 0.4808, 0.529, 0.529]}
{"unit_id": "orig:1934749411198058523", "ticker_idx": 0, "ts": "2025-06-16T23:06:20+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Qualcomm's acquisition of Alphawave for $2.4 billion aims to complement its I/O and SerDes IP and penetrate the advanced chiplet ecosystem, thereby assisting its expansion from core smartphone business into the data center and AI inference domains", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley sees QCOM's Alphawave deal bolstering chiplet/IP capabilities and opening data center/AI inference revenue, with Amazon Trainium2 and Microsoft Maia as key 2026 catalysts.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Qualcomm: Bolstering IP and Chiplet Design, Betting on Data Center and AI Inference Markets\nMorgan Stanley (L. Simpson, June 10, 2025)\n\nQualcomm's acquisition of Alphawave for $2.4 billion aims to complement its I/O and SerDes IP and penetrate the advanced chiplet ecosystem, thereby assisting its expansion from core smartphone business into the data center and AI inference domains. The interconnect IP gained from Alphawave, coupled with the logic block capabilities acquired from OpenFive previously, is expected to reduce Qualcomm's reliance on Arm licenses in the future and accelerate chiplet design and integration. Alphawave's chiplet interconnect solution can meet the high-bandwidth requirements of backplane and fly-over interconnects for high-density AI racks. This capability, the firm believes, is crucial for data center and AI hardware iteration. Qualcomm is deeply collaborating with the Arm ecosystem through its HUMAIN project, utilizing new IP resources to gain an early advantage in the low-power, inference acceleration market, but it still faces fierce competition from NVIDIA, AMD, and cloud vendors' in-house designed chips. The firm believes that if Qualcomm can materialize its follow-up cooperation with Amazon's Annapurna team on Trainium 2 in 2026 and drive projects like Microsoft Maia to fruition, it will significantly boost its data center revenue. Considering the high barriers in the x86 market and past experience with Centriq, Qualcomm's entry into the data center CPU/NPU domain still requires \"performance validation,\" and the market will closely monitor its product roadmap, approval progress, and initial mass production developments.\n\n$QCOM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013705559695721403, "ret_signed_m1d": -0.013705559695721403, "alpha_spy_m1d": -0.004281030043833622, "alpha_c_m1d": 0.010556077641390216, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015362956426071217, "ret_signed_p1d": -0.015362956426071217, "alpha_spy_p1d": -0.006817923889492317, "alpha_c_p1d": -0.008566666705353332, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": -0.023777591984331048, "ret_signed_p1w": -0.023777591984331048, "alpha_spy_p1w": -0.02252359667969428, "alpha_c_p1w": -0.018234256673158478, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": -0.01638298421856299, "ret_signed_p1m": -0.01638298421856299, "alpha_spy_p1m": -0.05172389485762019, "alpha_c_p1m": -0.1204910590913062, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.03543785148185563, "ret_signed_p3m": 0.03543785148185563, "alpha_spy_p3m": -0.05896416884163669, "alpha_c_p3m": -0.11852211790328426, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.13409842674467742, "ret_signed_p6m": 0.13409842674467742, "alpha_spy_p6m": -0.005741102420137256, "alpha_c_p6m": -0.2669190466227771, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [0.0028, 0.0009, -0.0063, 0.0144, 0.0148, 0.0037, 0.001, -0.0323, -0.0266, -0.0302, -0.0237, -0.1165, -0.1923, -0.178, -0.2101, -0.0901, -0.1484, -0.1176, -0.118, -0.1217, -0.1399, -0.134, -0.134, -0.1373, -0.1209, -0.11, -0.0672, -0.0586, -0.0646, -0.0693, -0.0593, -0.1432, -0.1141, -0.1164, -0.1135, -0.0856, -0.0806, -0.08, -0.0361, -0.041, -0.0306, -0.033, -0.0337, -0.0261, -0.0253, -0.0412, -0.0662, -0.0788, -0.0788, -0.0582, -0.0647, -0.06, -0.0799, -0.0709, -0.0562, -0.0555, -0.0593, -0.0486, -0.0093, 0.0144, 0.0166, 0.0117, -0.0137, 0.0, -0.0154, -0.0207, -0.0207, -0.0354, -0.0238, -0.0074, -0.006, 0.0084, 0.0106, 0.0152, 0.0161, 0.0347, 0.034, 0.034, 0.0078, 0.0164, 0.0158, 0.0142, 0.0038, -0.0164, -0.0164, -0.0178, -0.0272, -0.0132, 0.0134, 0.0071, 0.0192, 0.0126, 0.0098, 0.0266, 0.0332, 0.014, -0.0644, -0.0553, -0.0597, -0.0648, -0.0703, -0.0699, -0.0593, -0.0567, -0.02, -0.0018, 0.0078, 0.0062, 0.0129, -0.004, -0.0091, -0.0175, 0.0073, -0.0029, 0.0147, 0.0185, 0.0251, 0.0246, 0.0246, 0.0122, 0.0026, 0.0239, 0.0247, 0.0273, 0.0172, 0.019, 0.0354, 0.0375, 0.0336, 0.0523, 0.0595, 0.0779, 0.0697, 0.0881, 0.0869, 0.1126, 0.0878, 0.0847, 0.0597, 0.0665, 0.0674, 0.0825, 0.0846, 0.081, 0.0608, 0.0756, 0.062, -0.0153, 0.0372, 0.0369, 0.0448, 0.0519, 0.0479, 0.0709, 0.0824, 0.0852, 0.0901, 0.0831, 0.2032, 0.1606, 0.1455, 0.1364, 0.1597, 0.1586, 0.1081, 0.1522, 0.1104, 0.0956, 0.0999, 0.1154, 0.1326, 0.1187, 0.1154, 0.069, 0.0582, 0.0649, 0.0231, 0.0469, 0.0582, 0.0469, 0.0587, 0.0587, 0.0776, 0.0773, 0.0944, 0.1224, 0.1235, 0.1264, 0.1297, 0.1341]}
{"unit_id": "orig:1932405153933795395", "ticker_idx": 0, "ts": "2025-06-10T11:51:05+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs) SK Hynix Buy, Target Price KRW 290,000", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Buy on SK Hynix (PT KRW 290K) on HBM leadership + AI memory; Samsung also flagged as beneficiary of new government's pro-semiconductor policies.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) SK Hynix Buy, Target Price KRW 290,000: South Korea Semiconductor Industry Policy and SK Hynix Investment Analysis\n\nHistorically, the Korean government's semiconductor policies have focused on 1) building mega semiconductor manufacturing hubs, 2) tax credits for domestic facility investments, and 3) R&D support. However, the impact of government policies on large companies like Samsung Electronics (SEC) and SK Hynix has often been limited for the following reasons: 1) Construction projects typically span long periods (e.g., 10 or 20-year plans), which has limited impact on the short- to medium-term capital expenditure (capex) plans of both SEC and Hynix. 2) Project durations typically exceed the presidential term of four years, which cannot guarantee the continuity of project execution. 3) In the past, neither SEC nor Hynix has significantly increased their capex plans due to tax credits; rather, changes in capex plans have been driven by changes in industry supply and demand conditions and outlook.\n\nThe new President-elect Lee Jae-myung's key semiconductor policy agenda includes: 1) swift enactment of the pending Semiconductor Special Act, 2) 10% tax credit for domestic chip production/sales, 3) acceleration of the Yongin Semiconductor Manufacturing Hub project timeline, 4) expanded support for semiconductor R&D, and 5) establishment of national AI data centers. While it remains to be seen whether the proposed policies will be implemented as planned, if realized, both SEC and Hynix could potentially benefit from higher tax credits. Furthermore, given that the new President-elect has pledged to prioritize support for the semiconductor industry and that his first action after being confirmed as a presidential candidate on April 28 was to visit SK Hynix's Icheon campus, we believe the new government's pro-semiconductor stance could provide positive investment sentiment for stocks we cover, such as SEC and SK Hynix.\n\nWe view SK Hynix as a strong beneficiary of artificial intelligence (AI) based on its solid High Bandwidth Memory (HBM) market leadership, and we continue to highlight it as our top pick within our Korean technology coverage. With a robust HBM outlook maintained, the company's earnings stability, stemming from its AI memory leadership, has significantly improved compared to past cycles, with HBM expected to account for nearly half of its DRAM revenue and 60% of its DRAM profit this year. Considering our expectation for its Return on Equity (ROE) to exceed 30% this year and the current stock trading near its historical average valuation multiple, we believe the risk-reward is attractive, and we maintain our Buy rating on SK Hynix with a 12-month target price of KRW 290,000.\n\nInvestment Overview - SK Hynix\n\nSK Hynix is the world's second-largest DRAM supplier and one of the leading NAND suppliers. We believe the company's earnings stability, stemming from its AI memory leadership, has significantly improved compared to past cycles. However, given its still unburdensome valuation level, we assess that this point is not yet fully appreciated, leading us to assign a Buy rating to the stock. Key risks include a significant deterioration in memory supply and demand and delays in technology transitions.\n\nTarget Price Risks & Methodology - SK Hynix\n\nValuation Methodology: Our 12-month target price of KRW 290,000 is derived by applying a target Price-to-Book (P/B) ratio of 2.0x to the average of our estimated 2025 Book Value Per Share (BVPS).\n\nKey Risks: 1) Significant deterioration in memory supply and demand, and 2) delays in technology transitions.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006507625414107987, "ret_signed_m1d": -0.006507625414107987, "alpha_spy_m1d": -0.0008699246182334308, "alpha_c_m1d": 0.012991089202814221, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.019498714616922208, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04121468867328337, "ret_signed_p1d": 0.04121468867328337, "alpha_spy_p1d": 0.044066603264071524, "alpha_c_p1d": 0.041939731636196464, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0007250429629130917, "ret_p1w": 0.08026030501656423, "ret_signed_p1w": 0.08026030501656423, "alpha_spy_p1w": 0.08946289905092397, "alpha_c_p1w": 0.08209189413505591, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": -0.0018315891184916833, "ret_p1m": 0.21908889840668322, "ret_signed_p1m": 0.21908889840668322, "alpha_spy_p1m": 0.1812416578439, "alpha_c_p1m": 0.12999049815026065, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.08909840025642257, "ret_p3m": 0.18826472307620667, "ret_signed_p3m": 0.18826472307620667, "alpha_spy_p3m": 0.11186780699292331, "alpha_c_p3m": 0.06917440143607467, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.119090321640132, "ret_p6m": 1.3999705390081774, "ret_signed_p6m": 1.3999705390081774, "alpha_spy_p6m": 1.2594693186016537, "alpha_c_p6m": 1.0096168460666173, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.3903536929415601, "price_path": [-0.1352, -0.1144, -0.1079, -0.1209, -0.1101, -0.0906, -0.0668, -0.0841, -0.0992, -0.0733, -0.1036, -0.1369, -0.1742, -0.1469, -0.143, -0.1573, -0.211, -0.2863, -0.266, -0.2855, -0.2066, -0.217, -0.2196, -0.2179, -0.2465, -0.2421, -0.2421, -0.2352, -0.2473, -0.2162, -0.2279, -0.2014, -0.2118, -0.217, -0.2313, -0.2313, -0.1945, -0.1945, -0.1945, -0.1737, -0.1759, -0.1768, -0.1555, -0.1404, -0.1079, -0.1317, -0.1144, -0.1365, -0.1252, -0.1317, -0.1473, -0.1339, -0.1209, -0.1231, -0.0992, -0.0803, -0.1128, -0.0998, -0.0998, -0.0564, -0.026, -0.026, -0.0065, 0.0, 0.0412, 0.0217, 0.0217, 0.0759, 0.0803, 0.0694, 0.0672, 0.115, 0.1258, 0.2082, 0.2408, 0.2711, 0.2321, 0.2668, 0.2386, 0.2104, 0.2082, 0.1735, 0.1757, 0.2234, 0.2191, 0.2885, 0.2777, 0.3015, 0.295, 0.2842, 0.1692, 0.167, 0.1822, 0.1649, 0.167, 0.1692, 0.154, 0.1367, 0.1388, 0.1432, 0.1866, 0.1193, 0.1193, 0.1432, 0.1215, 0.1367, 0.1128, 0.1584, 0.167, 0.2061, 0.1996, 0.1996, 0.1605, 0.141, 0.1085, 0.0629, 0.0889, 0.1258, 0.1345, 0.128, 0.1665, 0.1687, 0.1122, 0.1318, 0.1405, 0.1535, 0.1883, 0.2035, 0.2513, 0.3208, 0.3338, 0.4272, 0.4381, 0.5119, 0.4489, 0.5445, 0.5445, 0.525, 0.5684, 0.5532, 0.5489, 0.462, 0.5163, 0.5098, 0.5641, 0.7183, 0.7183, 0.7183, 0.7183, 0.7183, 0.7183, 0.8595, 0.803, 0.7878, 0.8356, 0.966, 1.0224, 1.1093, 1.0811, 1.092, 1.0789, 1.2158, 1.3244, 1.2636, 1.4243, 1.4678, 1.4287, 1.6937, 1.546, 1.5156, 1.5764, 1.5199, 1.6329, 1.6893, 1.6807, 1.6589, 1.433, 1.6329, 1.4765, 1.4417, 1.4808, 1.2636, 1.2592, 1.2549, 1.2766, 1.3652, 1.3043, 1.3391, 1.4261, 1.4]}
{"unit_id": "orig:1932405153933795395", "ticker_idx": 1, "ts": "2025-06-10T11:51:05+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the new government's pro-semiconductor stance could provide positive investment sentiment for stocks we cover, such as SEC and SK Hynix", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Buy on SK Hynix (PT KRW 290K) on HBM leadership + AI memory; Samsung also flagged as beneficiary of new government's pro-semiconductor policies.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) SK Hynix Buy, Target Price KRW 290,000: South Korea Semiconductor Industry Policy and SK Hynix Investment Analysis\n\nHistorically, the Korean government's semiconductor policies have focused on 1) building mega semiconductor manufacturing hubs, 2) tax credits for domestic facility investments, and 3) R&D support. However, the impact of government policies on large companies like Samsung Electronics (SEC) and SK Hynix has often been limited for the following reasons: 1) Construction projects typically span long periods (e.g., 10 or 20-year plans), which has limited impact on the short- to medium-term capital expenditure (capex) plans of both SEC and Hynix. 2) Project durations typically exceed the presidential term of four years, which cannot guarantee the continuity of project execution. 3) In the past, neither SEC nor Hynix has significantly increased their capex plans due to tax credits; rather, changes in capex plans have been driven by changes in industry supply and demand conditions and outlook.\n\nThe new President-elect Lee Jae-myung's key semiconductor policy agenda includes: 1) swift enactment of the pending Semiconductor Special Act, 2) 10% tax credit for domestic chip production/sales, 3) acceleration of the Yongin Semiconductor Manufacturing Hub project timeline, 4) expanded support for semiconductor R&D, and 5) establishment of national AI data centers. While it remains to be seen whether the proposed policies will be implemented as planned, if realized, both SEC and Hynix could potentially benefit from higher tax credits. Furthermore, given that the new President-elect has pledged to prioritize support for the semiconductor industry and that his first action after being confirmed as a presidential candidate on April 28 was to visit SK Hynix's Icheon campus, we believe the new government's pro-semiconductor stance could provide positive investment sentiment for stocks we cover, such as SEC and SK Hynix.\n\nWe view SK Hynix as a strong beneficiary of artificial intelligence (AI) based on its solid High Bandwidth Memory (HBM) market leadership, and we continue to highlight it as our top pick within our Korean technology coverage. With a robust HBM outlook maintained, the company's earnings stability, stemming from its AI memory leadership, has significantly improved compared to past cycles, with HBM expected to account for nearly half of its DRAM revenue and 60% of its DRAM profit this year. Considering our expectation for its Return on Equity (ROE) to exceed 30% this year and the current stock trading near its historical average valuation multiple, we believe the risk-reward is attractive, and we maintain our Buy rating on SK Hynix with a 12-month target price of KRW 290,000.\n\nInvestment Overview - SK Hynix\n\nSK Hynix is the world's second-largest DRAM supplier and one of the leading NAND suppliers. We believe the company's earnings stability, stemming from its AI memory leadership, has significantly improved compared to past cycles. However, given its still unburdensome valuation level, we assess that this point is not yet fully appreciated, leading us to assign a Buy rating to the stock. Key risks include a significant deterioration in memory supply and demand and delays in technology transitions.\n\nTarget Price Risks & Methodology - SK Hynix\n\nValuation Methodology: Our 12-month target price of KRW 290,000 is derived by applying a target Price-to-Book (P/B) ratio of 2.0x to the average of our estimated 2025 Book Value Per Share (BVPS).\n\nKey Risks: 1) Significant deterioration in memory supply and demand, and 2) delays in technology transitions.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010135214079110222, "ret_signed_m1d": 0.010135214079110222, "alpha_spy_m1d": 0.015772914874984778, "alpha_c_m1d": 0.015698366779079542, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.00556315269996932, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01182444999950305, "ret_signed_p1d": 0.01182444999950305, "alpha_spy_p1d": 0.014676364590291202, "alpha_c_p1d": 0.014024825688629616, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0022003756891265658, "ret_p1w": -0.01858105794232978, "ret_signed_p1w": -0.01858105794232978, "alpha_spy_p1w": -0.009378463907970036, "alpha_c_p1w": -0.02044937357726151, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": 0.0018683156349317276, "ret_p1m": 0.026528396917327335, "ret_signed_p1m": 0.026528396917327335, "alpha_spy_p1m": -0.011318843645455878, "alpha_c_p1m": -0.04565720299133713, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.07218559990866447, "ret_p3m": 0.18118759120342998, "ret_signed_p3m": 0.18118759120342998, "alpha_spy_p3m": 0.10479067512014661, "alpha_c_p3m": 0.08808488372896961, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.09310270747446037, "ret_p6m": 0.7839541321784638, "ret_signed_p6m": 0.7839541321784638, "alpha_spy_p6m": 0.6434529117719401, "alpha_c_p6m": 0.5764814557769753, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.20747267640148848, "price_path": [-0.0815, -0.0815, -0.0328, -0.0328, -0.0177, 0.0109, 0.0361, 0.0159, 0.0042, 0.031, 0.0378, 0.0169, -0.0236, -0.0068, -0.0068, -0.027, -0.0524, -0.1014, -0.0963, -0.1047, -0.0473, -0.0676, -0.0507, -0.0439, -0.076, -0.0693, -0.0659, -0.0642, -0.0709, -0.0591, -0.0591, -0.0591, -0.0574, -0.0574, -0.0625, -0.0625, -0.0828, -0.0828, -0.0828, -0.0777, -0.0777, -0.0743, -0.027, -0.0389, -0.0304, -0.0321, -0.0405, -0.0574, -0.0557, -0.0591, -0.076, -0.0845, -0.076, -0.0895, -0.0557, -0.0524, -0.0507, -0.0405, -0.0405, -0.0236, -0.0017, -0.0017, 0.0101, 0.0, 0.0118, 0.0051, -0.0152, -0.0338, -0.0186, 0.0101, 0.0, 0.0051, -0.0203, 0.022, 0.0355, 0.0169, 0.0333, 0.0163, 0.0231, 0.0333, 0.0843, 0.0758, 0.0486, 0.0435, 0.0265, 0.0367, 0.0639, 0.0622, 0.0826, 0.0996, 0.1336, 0.1404, 0.1523, 0.1217, 0.1285, 0.1217, 0.12, 0.1965, 0.1999, 0.2339, 0.2135, 0.171, 0.1846, 0.188, 0.1693, 0.1982, 0.2203, 0.2067, 0.2084, 0.222, 0.2169, 0.2169, 0.1897, 0.1897, 0.1982, 0.1999, 0.2135, 0.2152, 0.1948, 0.1999, 0.1829, 0.1846, 0.1489, 0.1744, 0.1863, 0.1914, 0.1812, 0.1914, 0.2152, 0.2339, 0.2475, 0.2815, 0.3002, 0.3494, 0.329, 0.3647, 0.3647, 0.4191, 0.4395, 0.4514, 0.4633, 0.4157, 0.4374, 0.4323, 0.4681, 0.5322, 0.5322, 0.5322, 0.5322, 0.5322, 0.5322, 0.6115, 0.5928, 0.5637, 0.6218, 0.6679, 0.6713, 0.6747, 0.6645, 0.6832, 0.6474, 0.6866, 0.7413, 0.6986, 0.7157, 0.7771, 0.8352, 0.8966, 0.7908, 0.7174, 0.6935, 0.6713, 0.7174, 0.7669, 0.7601, 0.7549, 0.6593, 0.7174, 0.6696, 0.6474, 0.7174, 0.6184, 0.6508, 0.6952, 0.7549, 0.7669, 0.7157, 0.7208, 0.7652, 0.784]}
{"unit_id": "quote:1939552638238106087", "ticker_idx": 0, "ts": "2025-06-30T05:12:38+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD could also sufficiently benefit from inference", "tweet_type": "analysis", "timeline": "unspecified", "summary": "AMD flagged as a beneficiary of inference demand given Microsoft chip delays and OpenAI's GPU leasing activity.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Here are my takeaways after reading this report:\n\n1. The migration of MediaTek's SerDes personnel to Nvidia significantly impacted TPU delays (as far as I know, at least half, if not the majority, of the SerDes personnel transferred to Nvidia).\n\n2. The Microsoft-OpenAI collaboration might be stronger than expected?The report mentions that OpenAI is providing significant feedback on Microsoft's custom chip.This suggests that, for now, the collaboration between Microsoft and OAI isn't at a concerning level. (Of course, this report deals with past information, and it's possible that OAI has already decided to part ways with Microsoft.)\n\n3. In the comments section, BK Tech Insights pointed to AMD as a beneficiary for inference. I've reconsidered this, and given that OAI reportedly leased Google TPUs, I think AMD could also sufficiently benefit from inference.", "reply_text": null, "parent_text": null, "quoted_text": "[Microsoft's AI Chip Development Status and Problems] - BK Tech insight\n\n📌 Executive Summary\n\n- Microsoft's self-developed AI chip performance and schedule, aimed at reducing Nvidia dependence, are significantly delayed more than expected.\n- The latest AI chip (Braga) launch is", "parent_author": null, "ret_m1d": 0.013460209614263308, "ret_signed_m1d": 0.013460209614263308, "alpha_spy_m1d": 0.01821876367931774, "alpha_c_m1d": 0.015109607149475512, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.040803337104838744, "ret_signed_p1d": -0.040803337104838744, "alpha_spy_p1d": -0.04047965779661389, "alpha_c_p1d": -0.029902560342441653, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.05003517371470778, "ret_signed_p1w": -0.05003517371470778, "alpha_spy_p1w": -0.05461549993045045, "alpha_c_p1w": -0.05319071949833232, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": 0.25045814005353795, "ret_signed_p1m": 0.25045814005353795, "alpha_spy_p1m": 0.22227990861775182, "alpha_c_p1m": 0.20018565356534856, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.13650465968382575, "ret_signed_p3m": 0.13650465968382575, "alpha_spy_p3m": 0.06848757252462923, "alpha_c_p3m": -0.014743206258551966, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.5144468156443565, "ret_signed_p6m": 0.5144468156443565, "alpha_spy_p6m": 0.39458634543565463, "alpha_c_p6m": 0.20818598756249074, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2744, -0.339, -0.3956, -0.4106, -0.4488, -0.3175, -0.3749, -0.3418, -0.334, -0.3285, -0.3778, -0.3834, -0.3834, -0.397, -0.3921, -0.363, -0.3342, -0.3189, -0.3207, -0.323, -0.314, -0.3189, -0.3037, -0.2911, -0.305, -0.2927, -0.2833, -0.2753, -0.2381, -0.2075, -0.1704, -0.1896, -0.1743, -0.1914, -0.2001, -0.2103, -0.2198, -0.2226, -0.2226, -0.1927, -0.2047, -0.2035, -0.2197, -0.1922, -0.1733, -0.1643, -0.1847, -0.1812, -0.1421, -0.1315, -0.1463, -0.1649, -0.1814, -0.1093, -0.1043, -0.1065, -0.1065, -0.0963, -0.0868, -0.0245, 0.0106, 0.0125, 0.0135, 0.0, -0.0408, -0.0238, -0.0281, -0.0281, -0.05, -0.0288, -0.0246, 0.0159, 0.0319, 0.0306, 0.0966, 0.1281, 0.1304, 0.1063, 0.1064, 0.0903, 0.118, 0.1425, 0.1732, 0.2238, 0.2505, 0.265, 0.2425, 0.21, 0.2458, 0.2284, 0.1495, 0.2149, 0.2175, 0.2141, 0.2329, 0.2996, 0.2752, 0.251, 0.2413, 0.1737, 0.1642, 0.1537, 0.1822, 0.1512, 0.1742, 0.1778, 0.188, 0.1461, 0.1461, 0.1439, 0.1426, 0.1402, 0.0651, 0.067, 0.0981, 0.1243, 0.097, 0.1175, 0.1357, 0.1308, 0.1216, 0.1129, 0.1092, 0.1261, 0.1339, 0.1338, 0.1365, 0.1237, 0.1371, 0.1402, 0.1558, 0.1961, 0.1605, 0.4356, 0.4906, 0.66, 0.6412, 0.5144, 0.5252, 0.5369, 0.6815, 0.653, 0.6426, 0.6953, 0.6774, 0.6225, 0.656, 0.7824, 0.83, 0.8183, 0.8628, 0.7959, 0.8049, 0.8298, 0.7622, 0.8064, 0.6751, 0.6458, 0.7194, 0.6739, 0.8245, 0.7474, 0.7393, 0.695, 0.6229, 0.5754, 0.4519, 0.4361, 0.5155, 0.4526, 0.5098, 0.5098, 0.533, 0.5487, 0.5168, 0.5335, 0.5221, 0.5361, 0.5582, 0.5618, 0.5604, 0.5605, 0.4854, 0.4629, 0.4741, 0.3961, 0.4169, 0.5041, 0.5148, 0.5144]}
{"unit_id": "orig:1930966384340787682", "ticker_idx": 0, "ts": "2025-06-06T12:33:56+00:00", "symbol": "ADI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks Are ADI and TXN", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi relays top picks ADI and TXN as most defensive; Buy on AVGO, MCHP, MU, NXPI; maintains 8% YoY 2025 semi revenue growth outlook.", "resolved_tickers": ["ADI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi – U.S. Semiconductors: April Revenue in Line with Our Forecast, but Below Seasonal Average; Maintain 8% YoY Growth Outlook for 2025, Top Picks Are ADI and TXN\n\nOn Thursday (local time), the Semiconductor Industry Association (SIA) reported that U.S. semiconductor industry revenue for April was $55.0 billion, down 11.1% MoM. While this decline is larger than the typical seasonal average of -10.0%, it is exactly in line with our forecast of $55.0 billion (-11.1% MoM), reflecting weakness in microprocessor and DRAM sales.\n\nWe expect further moderation in the second half of 2025 due to tariff-related adjustments, resulting in monthly sales figures continuing to fall below seasonal norms. Nonetheless, we maintain our full-year 2025 semiconductor revenue growth forecast of +8% YoY.\n\nOur top picks remain Analog Devices (ADI) and Texas Instruments (TXN), which we view as the most defensive names during economic downturns. Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP), Micron Technology (MU), and NXP Semiconductors (NXPI).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018986765618083212, "ret_signed_m1d": -0.018986765618083212, "alpha_spy_m1d": -0.008822165829622475, "alpha_c_m1d": -0.013328539757085145, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024295921723286762, "ret_signed_p1d": 0.024295921723286762, "alpha_spy_p1d": 0.02339463883514248, "alpha_c_p1d": 0.007558775039901322, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.01246289668800471, "ret_signed_p1w": 0.01246289668800471, "alpha_spy_p1w": 0.016034643576610663, "alpha_c_p1w": -0.004392954881467626, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.0879600706610808, "ret_signed_p1m": 0.0879600706610808, "alpha_spy_p1m": 0.04894596924692096, "alpha_c_p1m": -0.018992062683751643, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.10463952718689251, "ret_signed_p3m": 0.10463952718689251, "alpha_spy_p3m": 0.027023171586640915, "alpha_c_p3m": -0.028704336158104526, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.20383348852312522, "ret_signed_p6m": 0.20383348852312522, "alpha_spy_p6m": 0.06190889355689366, "alpha_c_p6m": -0.19275578094278023, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.0636, -0.0746, -0.0872, -0.065, -0.0519, -0.0632, -0.0596, -0.0778, -0.0845, -0.049, -0.0434, -0.0571, -0.0642, -0.0972, -0.0968, -0.1079, -0.1061, -0.1899, -0.2628, -0.2326, -0.2561, -0.1193, -0.2012, -0.1976, -0.2001, -0.1959, -0.212, -0.2105, -0.2105, -0.2162, -0.2005, -0.1799, -0.1256, -0.1285, -0.1324, -0.1371, -0.127, -0.1323, -0.1105, -0.1144, -0.1239, -0.1049, -0.0918, -0.0706, -0.0005, 0.0153, 0.0159, 0.0172, 0.0172, 0.0057, 0.0055, -0.0047, -0.0508, -0.0573, -0.0573, -0.0286, -0.0343, -0.0337, -0.0416, -0.035, -0.0217, -0.0169, -0.019, 0.0, 0.0243, 0.0499, 0.0463, 0.0444, 0.0125, 0.0429, 0.0233, 0.0332, 0.0332, 0.0274, 0.0392, 0.0572, 0.0559, 0.0677, 0.0661, 0.0709, 0.0827, 0.103, 0.1054, 0.1054, 0.088, 0.103, 0.0921, 0.1029, 0.1009, 0.0954, 0.0817, 0.0826, 0.0842, 0.0881, 0.082, 0.0596, 0.0262, 0.0185, 0.025, 0.0383, 0.0382, 0.0398, 0.0107, -0.0025, 0.0006, -0.0071, -0.0071, 0.0039, 0.0076, 0.0081, 0.044, 0.0692, 0.0628, 0.0422, 0.0418, 0.0368, 0.1017, 0.1111, 0.1347, 0.145, 0.1501, 0.1496, 0.1439, 0.1307, 0.1307, 0.1217, 0.1046, 0.1117, 0.116, 0.1246, 0.121, 0.1167, 0.1213, 0.1076, 0.1063, 0.1026, 0.1126, 0.125, 0.1082, 0.1172, 0.1147, 0.123, 0.1181, 0.1182, 0.1057, 0.1098, 0.0808, 0.0916, 0.0931, 0.0954, 0.0559, 0.0747, 0.0745, 0.0178, 0.06, 0.0633, 0.0757, 0.0914, 0.0971, 0.1122, 0.1129, 0.0857, 0.0989, 0.0751, 0.0977, 0.0812, 0.0617, 0.052, 0.0576, 0.0552, 0.0361, 0.066, 0.0519, 0.0321, 0.048, 0.0543, 0.0906, 0.0729, 0.061, 0.0386, 0.0395, 0.0489, 0.0172, 0.0494, 0.0814, 0.1384, 0.165, 0.165, 0.1985, 0.2038]}
{"unit_id": "orig:1930966384340787682", "ticker_idx": 1, "ts": "2025-06-06T12:33:56+00:00", "symbol": "TXN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks Are ADI and TXN", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi relays top picks ADI and TXN as most defensive; Buy on AVGO, MCHP, MU, NXPI; maintains 8% YoY 2025 semi revenue growth outlook.", "resolved_tickers": ["TXN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi – U.S. Semiconductors: April Revenue in Line with Our Forecast, but Below Seasonal Average; Maintain 8% YoY Growth Outlook for 2025, Top Picks Are ADI and TXN\n\nOn Thursday (local time), the Semiconductor Industry Association (SIA) reported that U.S. semiconductor industry revenue for April was $55.0 billion, down 11.1% MoM. While this decline is larger than the typical seasonal average of -10.0%, it is exactly in line with our forecast of $55.0 billion (-11.1% MoM), reflecting weakness in microprocessor and DRAM sales.\n\nWe expect further moderation in the second half of 2025 due to tariff-related adjustments, resulting in monthly sales figures continuing to fall below seasonal norms. Nonetheless, we maintain our full-year 2025 semiconductor revenue growth forecast of +8% YoY.\n\nOur top picks remain Analog Devices (ADI) and Texas Instruments (TXN), which we view as the most defensive names during economic downturns. Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP), Micron Technology (MU), and NXP Semiconductors (NXPI).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010030095258716787, "ret_signed_m1d": -0.010030095258716787, "alpha_spy_m1d": 0.00013450452974395066, "alpha_c_m1d": -0.0043718693977187195, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03528733488206459, "ret_signed_p1d": 0.03528733488206459, "alpha_spy_p1d": 0.03438605199392031, "alpha_c_p1d": 0.01855018819867915, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.01340810788248592, "ret_signed_p1w": 0.01340810788248592, "alpha_spy_p1w": 0.016979854771091873, "alpha_c_p1w": -0.0034477436869864153, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.10908430716012152, "ret_signed_p1m": 0.10908430716012152, "alpha_spy_p1m": 0.07007020574596168, "alpha_c_p1m": 0.0021321738152890735, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.024606599630339288, "ret_signed_p3m": 0.024606599630339288, "alpha_spy_p3m": -0.05300975596991231, "alpha_c_p3m": -0.10873726371465775, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": -0.11190494280955166, "ret_signed_p6m": -0.11190494280955166, "alpha_spy_p6m": -0.2538295377757832, "alpha_c_p6m": -0.5084942122754571, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.0862, -0.0946, -0.1054, -0.0909, -0.0725, -0.0753, -0.0635, -0.069, -0.0776, -0.0441, -0.052, -0.0493, -0.0685, -0.0914, -0.074, -0.0828, -0.0818, -0.1539, -0.2199, -0.2064, -0.2476, -0.1266, -0.193, -0.2394, -0.222, -0.221, -0.2448, -0.2351, -0.2351, -0.2497, -0.2437, -0.216, -0.1645, -0.1608, -0.1629, -0.1716, -0.1682, -0.1775, -0.145, -0.1559, -0.1628, -0.1436, -0.1392, -0.1047, -0.0267, -0.0219, -0.0264, -0.0239, -0.0204, -0.0212, -0.0193, -0.0364, -0.0635, -0.0838, -0.0838, -0.0477, -0.043, -0.0386, -0.0497, -0.0427, -0.0226, -0.0088, -0.01, 0.0, 0.0353, 0.0513, 0.0378, 0.0376, 0.0134, 0.0353, 0.0274, 0.0308, 0.0308, 0.03, 0.0466, 0.0696, 0.0674, 0.0722, 0.0762, 0.079, 0.0937, 0.1204, 0.1226, 0.1226, 0.1091, 0.1258, 0.1246, 0.1416, 0.1498, 0.1436, 0.1348, 0.1259, 0.1256, 0.1258, 0.1151, 0.1169, -0.0321, -0.035, -0.0386, -0.0165, -0.0054, -0.0151, -0.0522, -0.0533, -0.0435, -0.0295, -0.0268, -0.0268, -0.02, -0.0384, 0.0101, 0.0118, 0.014, 0.0185, 0.0172, 0.0257, 0.0509, 0.0506, 0.0786, 0.0782, 0.0782, 0.0755, 0.0683, 0.0599, 0.0599, 0.0459, 0.0246, -0.0196, -0.0163, -0.0273, -0.0315, -0.0368, -0.035, -0.0442, -0.0672, -0.0702, -0.0562, -0.0493, -0.0611, -0.0598, -0.0471, -0.0345, -0.0471, -0.034, -0.0409, -0.0383, -0.0557, -0.0456, -0.0561, -0.0483, -0.0732, -0.0494, -0.0632, -0.1012, -0.0834, -0.0895, -0.0825, -0.0814, -0.0757, -0.0599, -0.0534, -0.1064, -0.0987, -0.1147, -0.1132, -0.1263, -0.1611, -0.1598, -0.1473, -0.1473, -0.1584, -0.1361, -0.1477, -0.1521, -0.1519, -0.1564, -0.1387, -0.1432, -0.1585, -0.1815, -0.1692, -0.1704, -0.1902, -0.1582, -0.1483, -0.1457, -0.1267, -0.1267, -0.1113, -0.1119]}
{"unit_id": "orig:1930966384340787682", "ticker_idx": 2, "ts": "2025-06-06T12:33:56+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Other Buy-rated names include Broadcom (AVGO)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi relays top picks ADI and TXN as most defensive; Buy on AVGO, MCHP, MU, NXPI; maintains 8% YoY 2025 semi revenue growth outlook.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi – U.S. Semiconductors: April Revenue in Line with Our Forecast, but Below Seasonal Average; Maintain 8% YoY Growth Outlook for 2025, Top Picks Are ADI and TXN\n\nOn Thursday (local time), the Semiconductor Industry Association (SIA) reported that U.S. semiconductor industry revenue for April was $55.0 billion, down 11.1% MoM. While this decline is larger than the typical seasonal average of -10.0%, it is exactly in line with our forecast of $55.0 billion (-11.1% MoM), reflecting weakness in microprocessor and DRAM sales.\n\nWe expect further moderation in the second half of 2025 due to tariff-related adjustments, resulting in monthly sales figures continuing to fall below seasonal norms. Nonetheless, we maintain our full-year 2025 semiconductor revenue growth forecast of +8% YoY.\n\nOur top picks remain Analog Devices (ADI) and Texas Instruments (TXN), which we view as the most defensive names during economic downturns. Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP), Micron Technology (MU), and NXP Semiconductors (NXPI).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.052646542785388384, "ret_signed_m1d": 0.052646542785388384, "alpha_spy_m1d": 0.06281114257384912, "alpha_c_m1d": 0.05830476864638645, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010731753810176325, "ret_signed_p1d": -0.010731753810176325, "alpha_spy_p1d": -0.011633036698320609, "alpha_c_p1d": -0.027468900493561765, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.007168124310281021, "ret_signed_p1w": 0.007168124310281021, "alpha_spy_p1w": 0.010739871198886974, "alpha_c_p1w": -0.009687727259191314, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.11296872464411822, "ret_signed_p1m": 0.11296872464411822, "alpha_spy_p1m": 0.07395462322995838, "alpha_c_p1m": 0.006016591299285778, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.22748067541138628, "ret_signed_p3m": 0.22748067541138628, "alpha_spy_p3m": 0.14986431981113468, "alpha_c_p3m": 0.09413681206638924, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.569885380902382, "ret_signed_p6m": 0.569885380902382, "alpha_spy_p6m": 0.4279607859361505, "alpha_c_p6m": 0.1732961114364766, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.2325, -0.2158, -0.2274, -0.2105, -0.2147, -0.2382, -0.2104, -0.2284, -0.2238, -0.2255, -0.2376, -0.274, -0.3035, -0.3151, -0.322, -0.3175, -0.3031, -0.3763, -0.4076, -0.3758, -0.3681, -0.2502, -0.3022, -0.2632, -0.2777, -0.2753, -0.2929, -0.3075, -0.3075, -0.3269, -0.3132, -0.2836, -0.238, -0.2212, -0.2205, -0.2258, -0.2205, -0.2009, -0.1753, -0.1871, -0.1897, -0.1706, -0.1586, -0.1568, -0.1027, -0.0588, -0.06, -0.0579, -0.0742, -0.066, -0.0618, -0.0697, -0.0664, -0.0737, -0.0737, -0.0457, -0.0304, -0.0201, -0.0197, 0.0072, 0.0402, 0.0573, 0.0526, 0.0, -0.0107, -0.0093, 0.0242, 0.037, 0.0072, 0.0209, 0.0099, 0.0175, 0.0175, 0.0148, 0.0301, 0.0707, 0.0743, 0.0967, 0.0934, 0.1189, 0.0746, 0.0956, 0.117, 0.117, 0.113, 0.1033, 0.1281, 0.1179, 0.1138, 0.1187, 0.1404, 0.1399, 0.1628, 0.1502, 0.1699, 0.1309, 0.1516, 0.1719, 0.1779, 0.1946, 0.2073, 0.2284, 0.1922, 0.1717, 0.2085, 0.1891, 0.2246, 0.233, 0.238, 0.2336, 0.2699, 0.2547, 0.2634, 0.2435, 0.2412, 0.1971, 0.1819, 0.1756, 0.1934, 0.1944, 0.2097, 0.2188, 0.2529, 0.2072, 0.2072, 0.2106, 0.2275, 0.2425, 0.3594, 0.4031, 0.3666, 0.5002, 0.4598, 0.4608, 0.4779, 0.4613, 0.4052, 0.4019, 0.4002, 0.3776, 0.3782, 0.3797, 0.3667, 0.3603, 0.3333, 0.3415, 0.3556, 0.3751, 0.3759, 0.3642, 0.3679, 0.4049, 0.4029, 0.32, 0.4504, 0.3993, 0.4286, 0.4401, 0.4205, 0.4201, 0.3933, 0.3837, 0.4, 0.44, 0.4722, 0.5166, 0.5695, 0.5308, 0.503, 0.4742, 0.4311, 0.4597, 0.4459, 0.4209, 0.4573, 0.4311, 0.4444, 0.3824, 0.3925, 0.3933, 0.3845, 0.4411, 0.4102, 0.3833, 0.5369, 0.5656, 0.6166, 0.6166, 0.6385, 0.5699]}
{"unit_id": "orig:1930966384340787682", "ticker_idx": 3, "ts": "2025-06-06T12:33:56+00:00", "symbol": "MCHP", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi relays top picks ADI and TXN as most defensive; Buy on AVGO, MCHP, MU, NXPI; maintains 8% YoY 2025 semi revenue growth outlook.", "resolved_tickers": ["MCHP"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi – U.S. Semiconductors: April Revenue in Line with Our Forecast, but Below Seasonal Average; Maintain 8% YoY Growth Outlook for 2025, Top Picks Are ADI and TXN\n\nOn Thursday (local time), the Semiconductor Industry Association (SIA) reported that U.S. semiconductor industry revenue for April was $55.0 billion, down 11.1% MoM. While this decline is larger than the typical seasonal average of -10.0%, it is exactly in line with our forecast of $55.0 billion (-11.1% MoM), reflecting weakness in microprocessor and DRAM sales.\n\nWe expect further moderation in the second half of 2025 due to tariff-related adjustments, resulting in monthly sales figures continuing to fall below seasonal norms. Nonetheless, we maintain our full-year 2025 semiconductor revenue growth forecast of +8% YoY.\n\nOur top picks remain Analog Devices (ADI) and Texas Instruments (TXN), which we view as the most defensive names during economic downturns. Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP), Micron Technology (MU), and NXP Semiconductors (NXPI).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013486498129425706, "ret_signed_m1d": -0.013486498129425706, "alpha_spy_m1d": -0.0033218983409649683, "alpha_c_m1d": -0.007828272268427638, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04291205988720237, "ret_signed_p1d": 0.04291205988720237, "alpha_spy_p1d": 0.042010776999058086, "alpha_c_p1d": 0.02617491320381693, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.007356435475903922, "ret_signed_p1w": 0.007356435475903922, "alpha_spy_p1w": 0.010928182364509875, "alpha_c_p1w": -0.009499416093568414, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.09547885586255211, "ret_signed_p1m": 0.09547885586255211, "alpha_spy_p1m": 0.056464754448392274, "alpha_c_p1m": -0.01147327748228033, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": -0.023469580033211646, "ret_signed_p3m": -0.023469580033211646, "alpha_spy_p3m": -0.10108593563346324, "alpha_c_p3m": -0.15681344337820868, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": -0.16803673581985035, "ret_signed_p6m": -0.16803673581985035, "alpha_spy_p6m": -0.3099613307860819, "alpha_c_p6m": -0.5646260052857558, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.2055, -0.2107, -0.2109, -0.1864, -0.1596, -0.1757, -0.1701, -0.2244, -0.2134, -0.1804, -0.2037, -0.2066, -0.2276, -0.2594, -0.2638, -0.2624, -0.2559, -0.3809, -0.4492, -0.4207, -0.4626, -0.3172, -0.4098, -0.4113, -0.4087, -0.411, -0.4229, -0.4136, -0.4136, -0.4003, -0.3873, -0.3623, -0.2834, -0.2869, -0.2851, -0.3016, -0.2992, -0.3007, -0.2714, -0.2734, -0.2816, -0.2688, -0.2527, -0.1585, -0.0729, -0.0525, -0.0776, -0.0754, -0.0691, -0.0789, -0.0754, -0.0989, -0.1103, -0.1389, -0.1389, -0.1007, -0.1194, -0.1094, -0.1105, -0.0805, -0.0216, -0.0089, -0.0135, 0.0, 0.0429, 0.0665, 0.0667, 0.0411, 0.0074, 0.0451, 0.0374, 0.0425, 0.0425, 0.057, 0.051, 0.0794, 0.0973, 0.0848, 0.0803, 0.0785, 0.0985, 0.1212, 0.1197, 0.1197, 0.0955, 0.1427, 0.1445, 0.1507, 0.1427, 0.1349, 0.1205, 0.1407, 0.1387, 0.1461, 0.1318, 0.1534, 0.0766, 0.0392, 0.0607, 0.0809, 0.0832, 0.0772, 0.0359, 0.017, 0.0205, 0.0288, 0.0141, 0.0149, -0.0518, -0.0659, -0.0115, 0.0077, 0.0113, 0.007, 0.0048, -0.0083, 0.0231, 0.013, 0.067, 0.0579, 0.0435, 0.0285, 0.0069, 0.0031, 0.0031, -0.0185, -0.0235, -0.0057, 0.0173, 0.0077, -0.0006, -0.0009, 0.0034, -0.0016, -0.0252, -0.0054, 0.0151, 0.0225, 0.0054, 0.0092, -0.0014, 0.0162, 0.0006, -0.0059, -0.0113, -0.009, -0.0107, 0.0205, 0.0268, 0.0276, 0.0025, 0.0327, 0.0163, -0.0678, -0.0063, -0.0031, 0.0063, 0.0085, 0.0052, 0.035, 0.042, -0.0046, 0.0045, -0.0252, -0.0039, -0.0179, -0.0349, -0.0421, -0.0367, -0.0369, -0.0818, -0.0617, -0.0841, -0.1315, -0.1449, -0.1557, -0.1415, -0.1542, -0.1747, -0.2022, -0.215, -0.2161, -0.2435, -0.2145, -0.202, -0.193, -0.1814, -0.1814, -0.1657, -0.168]}
{"unit_id": "orig:1930966384340787682", "ticker_idx": 4, "ts": "2025-06-06T12:33:56+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron Technology (MU)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi relays top picks ADI and TXN as most defensive; Buy on AVGO, MCHP, MU, NXPI; maintains 8% YoY 2025 semi revenue growth outlook.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi – U.S. Semiconductors: April Revenue in Line with Our Forecast, but Below Seasonal Average; Maintain 8% YoY Growth Outlook for 2025, Top Picks Are ADI and TXN\n\nOn Thursday (local time), the Semiconductor Industry Association (SIA) reported that U.S. semiconductor industry revenue for April was $55.0 billion, down 11.1% MoM. While this decline is larger than the typical seasonal average of -10.0%, it is exactly in line with our forecast of $55.0 billion (-11.1% MoM), reflecting weakness in microprocessor and DRAM sales.\n\nWe expect further moderation in the second half of 2025 due to tariff-related adjustments, resulting in monthly sales figures continuing to fall below seasonal norms. Nonetheless, we maintain our full-year 2025 semiconductor revenue growth forecast of +8% YoY.\n\nOur top picks remain Analog Devices (ADI) and Texas Instruments (TXN), which we view as the most defensive names during economic downturns. Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP), Micron Technology (MU), and NXP Semiconductors (NXPI).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02091003034336192, "ret_signed_m1d": -0.02091003034336192, "alpha_spy_m1d": -0.01074543055490118, "alpha_c_m1d": -0.015251804482363851, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022015545117126667, "ret_signed_p1d": 0.022015545117126667, "alpha_spy_p1d": 0.021114262228982383, "alpha_c_p1d": 0.005278398433741227, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.06484897673487722, "ret_signed_p1w": 0.06484897673487722, "alpha_spy_p1w": 0.06842072362348317, "alpha_c_p1w": 0.04799312516540488, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.10568243610849493, "ret_signed_p1m": 0.10568243610849493, "alpha_spy_p1m": 0.06666833469433509, "alpha_c_p1m": -0.0012696972363375192, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.0946182007900278, "ret_signed_p3m": 0.0946182007900278, "alpha_spy_p3m": 0.01700184518977621, "alpha_c_p3m": -0.03872566255496923, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 1.218469810598613, "ret_signed_p6m": 1.218469810598613, "alpha_spy_p6m": 1.0765452156323814, "alpha_c_p6m": 0.8218805411327075, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.1808, -0.1202, -0.1271, -0.0728, -0.0514, -0.0642, -0.0611, -0.0524, -0.1286, -0.1082, -0.1336, -0.1524, -0.1614, -0.1864, -0.1996, -0.1828, -0.1839, -0.3152, -0.4038, -0.3702, -0.3963, -0.2827, -0.3547, -0.3593, -0.3458, -0.3456, -0.3614, -0.3662, -0.3662, -0.3852, -0.3533, -0.3282, -0.2868, -0.2651, -0.2763, -0.2918, -0.2912, -0.2836, -0.2564, -0.2592, -0.2584, -0.2389, -0.2156, -0.2091, -0.1499, -0.1071, -0.122, -0.1208, -0.0973, -0.0913, -0.0964, -0.1172, -0.1265, -0.1399, -0.1399, -0.1122, -0.114, -0.1083, -0.1299, -0.0956, -0.0581, -0.0489, -0.0209, 0.0, 0.022, 0.0514, 0.0688, 0.0702, 0.0648, 0.1039, 0.1085, 0.1221, 0.1221, 0.1385, 0.1245, 0.1782, 0.1722, 0.1606, 0.1492, 0.1353, 0.1136, 0.1214, 0.1265, 0.1265, 0.1057, 0.1472, 0.1271, 0.1351, 0.1482, 0.0936, 0.1074, 0.0735, 0.0443, 0.0547, 0.044, 0.007, 0.0127, 0.0302, 0.0258, 0.0257, 0.0323, 0.0579, 0.0063, -0.033, -0.0063, 0.0056, 0.003, 0.0315, 0.0962, 0.1407, 0.1779, 0.1458, 0.1552, 0.1144, 0.1392, 0.1253, 0.0807, 0.0676, 0.085, 0.0734, 0.0741, 0.0857, 0.1249, 0.0973, 0.0973, 0.0924, 0.0946, 0.1452, 0.2113, 0.2121, 0.2469, 0.2908, 0.3883, 0.4497, 0.4547, 0.4643, 0.4751, 0.5572, 0.5004, 0.5178, 0.5343, 0.491, 0.446, 0.4501, 0.5112, 0.5427, 0.6795, 0.6942, 0.7329, 0.7618, 0.7132, 0.8133, 0.7744, 0.6754, 0.7785, 0.7258, 0.7708, 0.8685, 0.8671, 0.9076, 0.8663, 0.8311, 0.9071, 1.0207, 1.0306, 1.0473, 1.0909, 1.0667, 1.0645, 1.1653, 1.0115, 1.1912, 1.1988, 1.195, 1.3369, 1.2245, 1.2594, 1.1861, 1.2772, 1.2322, 1.1081, 1.0843, 0.8578, 0.9132, 1.066, 1.0715, 1.1244, 1.1244, 1.1818, 1.2185]}
{"unit_id": "orig:1930966384340787682", "ticker_idx": 5, "ts": "2025-06-06T12:33:56+00:00", "symbol": "NXPI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NXP Semiconductors (NXPI)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi relays top picks ADI and TXN as most defensive; Buy on AVGO, MCHP, MU, NXPI; maintains 8% YoY 2025 semi revenue growth outlook.", "resolved_tickers": ["NXPI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi – U.S. Semiconductors: April Revenue in Line with Our Forecast, but Below Seasonal Average; Maintain 8% YoY Growth Outlook for 2025, Top Picks Are ADI and TXN\n\nOn Thursday (local time), the Semiconductor Industry Association (SIA) reported that U.S. semiconductor industry revenue for April was $55.0 billion, down 11.1% MoM. While this decline is larger than the typical seasonal average of -10.0%, it is exactly in line with our forecast of $55.0 billion (-11.1% MoM), reflecting weakness in microprocessor and DRAM sales.\n\nWe expect further moderation in the second half of 2025 due to tariff-related adjustments, resulting in monthly sales figures continuing to fall below seasonal norms. Nonetheless, we maintain our full-year 2025 semiconductor revenue growth forecast of +8% YoY.\n\nOur top picks remain Analog Devices (ADI) and Texas Instruments (TXN), which we view as the most defensive names during economic downturns. Other Buy-rated names include Broadcom (AVGO), Microchip Technology (MCHP), Micron Technology (MU), and NXP Semiconductors (NXPI).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0037558622875951686, "ret_signed_m1d": -0.0037558622875951686, "alpha_spy_m1d": 0.006408737500865569, "alpha_c_m1d": 0.0019023635734028987, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02600165125265419, "ret_signed_p1d": 0.02600165125265419, "alpha_spy_p1d": 0.025100368364509906, "alpha_c_p1d": 0.00926450456926875, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.015504584071738714, "ret_signed_p1w": 0.015504584071738714, "alpha_spy_p1w": 0.019076330960344667, "alpha_c_p1w": -0.0013512674977336214, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.09328604144933172, "ret_signed_p1m": 0.09328604144933172, "alpha_spy_p1m": 0.054271940035171884, "alpha_c_p1m": -0.01366609189550072, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.10392866205183449, "ret_signed_p3m": 0.10392866205183449, "alpha_spy_p3m": 0.02631230645158289, "alpha_c_p3m": -0.02941520129316255, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": -0.03050886569335054, "ret_signed_p6m": -0.03050886569335054, "alpha_spy_p6m": -0.1724334606595821, "alpha_c_p6m": -0.427098135159256, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.0139, -0.0204, -0.0368, -0.0122, 0.0054, -0.0077, -0.0014, -0.0243, -0.0339, 0.0166, 0.0176, -0.0093, -0.0367, -0.0852, -0.0848, -0.0851, -0.0675, -0.1724, -0.2257, -0.2137, -0.2609, -0.105, -0.2027, -0.1873, -0.1788, -0.1736, -0.1882, -0.1779, -0.1779, -0.1742, -0.159, -0.1291, -0.0679, -0.068, -0.0551, -0.1207, -0.1125, -0.1244, -0.0941, -0.1094, -0.1217, -0.1002, -0.092, -0.0761, -0.0031, 0.0227, 0.0176, 0.0147, 0.0235, 0.0129, 0.0116, -0.0112, -0.0538, -0.0748, -0.0748, -0.0423, -0.0552, -0.0568, -0.0797, -0.0716, -0.042, 0.0113, -0.0038, 0.0, 0.026, 0.0532, 0.0469, 0.0468, 0.0155, 0.0474, 0.0248, 0.0182, 0.0182, 0.0064, 0.0153, 0.0521, 0.0463, 0.056, 0.0499, 0.057, 0.0701, 0.1182, 0.1228, 0.1228, 0.0933, 0.124, 0.1147, 0.1281, 0.1074, 0.0866, 0.0694, 0.0671, 0.086, 0.0928, 0.1043, 0.103, 0.087, 0.0857, 0.0802, 0.1053, 0.0969, 0.0688, 0.0341, 0.0155, 0.0255, 0.0085, -0.0039, -0.0039, 0.0021, -0.0075, 0.0645, 0.1152, 0.1201, 0.1067, 0.1224, 0.1091, 0.1067, 0.0833, 0.136, 0.1449, 0.1505, 0.1497, 0.1565, 0.1361, 0.1361, 0.1255, 0.1039, 0.0903, 0.0969, 0.0909, 0.0821, 0.0608, 0.0798, 0.0586, 0.0607, 0.069, 0.0784, 0.1008, 0.0888, 0.097, 0.0965, 0.1064, 0.1023, 0.0985, 0.0989, 0.1067, 0.093, 0.1066, 0.1124, 0.1247, 0.0671, 0.0966, 0.0761, -0.0019, 0.0531, 0.0503, 0.0557, 0.0566, 0.0417, 0.0683, 0.0805, 0.0554, 0.0727, 0.0651, 0.0767, 0.035, -0.0051, 0.003, 0.0163, 0.0225, -0.0065, 0.0227, 0.0033, -0.0059, -0.0031, -0.0141, -0.0082, -0.0221, -0.0421, -0.0742, -0.0835, -0.0763, -0.1049, -0.0701, -0.069, -0.0717, -0.0584, -0.0584, -0.0526, -0.0305]}
{"unit_id": "thread:1932623415355322845", "ticker_idx": 0, "ts": "2025-06-11T02:18:23+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung is expected to be the biggest beneficiary, making volume, price, and profit margins all positive variables", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung seen as biggest DDR4 price-surge beneficiary in Q2, with volume/price/margins all positive; partially offsets HBM weakness.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Samsung Benefits from Surprise DDR4 Demand... A 'Welcome Rain in a Drought' for Q2 Earnings\n\nAccording to reports from Korean media, Samsung Electronics is expected to deplete its DDR4 and LPDDR4 inventory in the middle of this year by a much larger margin than originally anticipated.\n\nDDR4 and LPDDR4 are considered legacy (mature) products in the DRAM industry. Major DRAM manufacturers are focusing on producing DDR5 and LPDDR5X and have decided to significantly reduce their DDR4 and LPDDR4 production by the end of this year. In Samsung's case, the revenue share of DDR4 and LPDDR4 in its total DRAM sales reached the 30% range last year. The plan is to reduce this to a single-digit percentage this year.\n\nIn contrast, customers are reportedly moving actively to secure inventory in advance to counter the production cuts of DDR4 and LPDDR4. Consequently, the prices for DDR4 and LPDDR4 are showing a sharp short-term increase.\n\nAccording to market research firm TrendForce, the price increase for PC DDR4 modules in the second quarter of this year is projected to be 13-18% quarter-over-quarter, far exceeding the initial forecast of 3-8%. The forecast for the third quarter has also been revised upward from 3-8% to 8-13%.\n\nThis trend is a positive factor for Samsung Electronics' memory business in the second quarter. Although legacy DRAM is a non-core business, the surge in demand allows Samsung to quickly clear out its existing inventory.\n\nA semiconductor industry official explained, \"DDR4 and LPDDR4 products have reached a point where they 'can't be sold because there's no stock,' and the price premium over DDR5 has narrowed to an abnormal level. In particular, Samsung is known to have proposed price hikes to its mobile and consumer electronics clients that are above the industry average.\"\n\nAnother official commented, \"In the first quarter, DDR4 shipments decreased slightly due to the off-season, but they are expected to rebound in the second quarter, leading to a meaningful expansion in revenue. There's a possibility that the Q2 DRAM bit growth (supply growth rate) could exceed the low 10% range that Samsung initially guided.\"\n\nThanks to this, Samsung is expected to be able to partially offset the impact of its sluggish HBM business in the second quarter. Previously, it was estimated that Samsung supplied around 600-800 million Gb (Gigabits) of HBM in the first quarter, a significant decrease from the previous quarter (estimated at 2 billion Gb). A similar supply volume is anticipated for the second quarter.\n\nKim Un-ho, a researcher at IBK Investment & Securities, stated, \"DDR4 prices rose most sharply during the second quarter, and Samsung is expected to be the biggest beneficiary, making volume, price, and profit margins all positive variables. On the other hand, HBM volume is significantly weaker than previous forecasts, so reflecting this, the memory division's second-quarter operating profit will not be significantly different from the previous forecast.\"\n\n---\n\nhttps://t.co/1rMl7xqEAK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011686266327631145, "ret_signed_m1d": -0.011686266327631145, "alpha_spy_m1d": -0.014546337597404202, "alpha_c_m1d": -0.013891494346880195, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0022052280192490503, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006677847512062796, "ret_signed_p1d": -0.006677847512062796, "alpha_spy_p1d": -0.010652073609217272, "alpha_c_p1d": -0.015956444809025472, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.009278597296962676, "ret_p1w": -0.0016694950595369518, "ret_signed_p1w": -0.0016694950595369518, "alpha_spy_p1w": 0.004849024158581572, "alpha_c_p1w": -0.006953690572089166, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.005284195512552214, "ret_p1m": 0.02461026976045866, "ret_signed_p1m": 0.02461026976045866, "alpha_spy_p1m": -0.01914071063008782, "alpha_c_p1m": -0.046397216853621526, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.07100748661408018, "ret_p3m": 0.177462075512117, "ret_signed_p3m": 0.177462075512117, "alpha_spy_p3m": 0.09533481075862515, "alpha_c_p3m": 0.07373846017432428, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.10372361533779273, "ret_p6m": 0.7732293875726444, "ret_signed_p6m": 0.7732293875726444, "alpha_spy_p6m": 0.6286300973138048, "alpha_c_p6m": 0.5585870872804422, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.21464230029220221, "price_path": [-0.0922, -0.0441, -0.0441, -0.0291, -0.0009, 0.024, 0.0041, -0.0076, 0.019, 0.0256, 0.005, -0.0351, -0.0184, -0.0184, -0.0384, -0.0634, -0.1119, -0.1068, -0.1152, -0.0584, -0.0785, -0.0618, -0.0551, -0.0868, -0.0801, -0.0768, -0.0751, -0.0818, -0.0701, -0.0701, -0.0701, -0.0684, -0.0684, -0.0735, -0.0735, -0.0935, -0.0935, -0.0935, -0.0885, -0.0885, -0.0851, -0.0384, -0.0501, -0.0417, -0.0434, -0.0518, -0.0684, -0.0668, -0.0701, -0.0868, -0.0952, -0.0868, -0.1002, -0.0668, -0.0634, -0.0618, -0.0518, -0.0518, -0.0351, -0.0134, -0.0134, -0.0017, -0.0117, 0.0, -0.0067, -0.0267, -0.0451, -0.0301, -0.0017, -0.0117, -0.0067, -0.0317, 0.01, 0.0234, 0.005, 0.0213, 0.0045, 0.0112, 0.0213, 0.0716, 0.0632, 0.0364, 0.0313, 0.0145, 0.0246, 0.0515, 0.0498, 0.07, 0.0868, 0.1204, 0.1271, 0.1388, 0.1086, 0.1153, 0.1086, 0.1069, 0.1825, 0.1859, 0.2195, 0.1993, 0.1573, 0.1707, 0.1741, 0.1556, 0.1842, 0.206, 0.1926, 0.1943, 0.2077, 0.2027, 0.2027, 0.1758, 0.1758, 0.1842, 0.1859, 0.1993, 0.201, 0.1808, 0.1859, 0.1691, 0.1707, 0.1355, 0.1607, 0.1724, 0.1775, 0.1674, 0.1775, 0.201, 0.2195, 0.2329, 0.2665, 0.285, 0.3337, 0.3135, 0.3488, 0.3488, 0.4025, 0.4227, 0.4345, 0.4462, 0.3992, 0.4206, 0.4155, 0.451, 0.5142, 0.5142, 0.5142, 0.5142, 0.5142, 0.5142, 0.5927, 0.5741, 0.5455, 0.6028, 0.6484, 0.6518, 0.6551, 0.645, 0.6636, 0.6281, 0.6669, 0.7209, 0.6787, 0.6956, 0.7564, 0.8137, 0.8745, 0.7699, 0.6973, 0.6737, 0.6518, 0.6973, 0.7462, 0.7395, 0.7344, 0.6399, 0.6973, 0.6501, 0.6281, 0.6973, 0.5994, 0.6315, 0.6754, 0.7344, 0.7462, 0.6956, 0.7007, 0.7445, 0.7631, 0.7732]}
{"unit_id": "thread:1932625670359962103", "ticker_idx": 0, "ts": "2025-06-11T02:27:20+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "He assessed that TSMC's continuous price increases could become the new 'normal' in the future, which is a potential positive for TSMC.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bullish TSMC: multiple banks see sustained price hikes (3-10%), 2nm ramp, and structural pricing power as durable positives.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Global investment banks' views on TSMC:\n\n- TSMC's 3nm and 4nm process lines are operating at full capacity, leading to overall May revenue exceeding expectations.\n\n- According to the latest research from Morgan Stanley semiconductor analyst Jame Wu (詹家鴻), U.S. customers are prepared to accept higher prices to secure 4nm process capacity at TSMC's U.S. factory.\n\n- Consequently, Taiwanese media reported that TSMC is exploring the possibility of raising foundry prices at its U.S. factory by at least 10%.\n\n- KGI Securities expects TSMC's 2nm process to begin production in the fourth quarter of this year and contribute significantly to revenue starting next year.\n\n- TSMC's 2nm monthly production capacity is projected to expand from 25,000 wafers by the end of this year to 70,000-75,000 wafers by the end of next year.\n\n- Goldman Sachs anticipates that TSMC will increase prices for processes below 5nm by approximately 3% and CoWoS prices by 5% in 2025.\n\n- This is interpreted as a measure to reflect rising costs, such as investments in the U.S. factory and electricity rate hikes in Taiwan.\n\n- Gokul Hariharan, Head of Taiwan Research at J.P. Morgan, pointed out that while most investors currently view TSMC's price increases as a one-off event, they overlook that TSMC has already established a continuous price adjustment system for several years, which has been a win-win strategy for both customers and TSMC.\n\n- He assessed that TSMC's continuous price increases could become the new \"normal\" in the future, which is a potential positive for TSMC.\n\n$TSM\n\n---\n\nhttps://t.co/7QhNkEV26N", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007659961326445197, "ret_signed_m1d": -0.007659961326445197, "alpha_spy_m1d": -0.010520032596218254, "alpha_c_m1d": -0.008385530358078874, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0007255690316336771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009891194551069349, "ret_signed_p1d": 0.009891194551069349, "alpha_spy_p1d": 0.005916968453914873, "alpha_c_p1d": 0.0050416767733423296, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.004849517777727019, "ret_p1w": 0.0008437472942941504, "ret_signed_p1w": 0.0008437472942941504, "alpha_spy_p1w": 0.007362266512412674, "alpha_c_p1w": -0.0018674050418874266, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.002711152336181577, "ret_p1m": 0.07706723137752891, "ret_signed_p1m": 0.07706723137752891, "alpha_spy_p1m": 0.03331625098698243, "alpha_c_p1m": -0.020840167684853972, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.09790739906238288, "ret_p3m": 0.1587755057866027, "ret_signed_p3m": 0.1587755057866027, "alpha_spy_p3m": 0.07664824103311085, "alpha_c_p3m": 0.02650124373192253, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.13227426205468018, "ret_p6m": 0.37752734566654156, "ret_signed_p6m": 0.37752734566654156, "alpha_spy_p6m": 0.232928055407702, "alpha_c_p6m": -0.0032577581335488937, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.38078510380009045, "price_path": [-0.19, -0.18, -0.1914, -0.1884, -0.1711, -0.1745, -0.1539, -0.1551, -0.1896, -0.2142, -0.2282, -0.2247, -0.2123, -0.2042, -0.2649, -0.3143, -0.3173, -0.3397, -0.2585, -0.2941, -0.2663, -0.2721, -0.2652, -0.2916, -0.2913, -0.2913, -0.3094, -0.2929, -0.2629, -0.2332, -0.2289, -0.2368, -0.2318, -0.2214, -0.1933, -0.1626, -0.1761, -0.1953, -0.1848, -0.1816, -0.1755, -0.1267, -0.0939, -0.0903, -0.0929, -0.0929, -0.0962, -0.0965, -0.1043, -0.0837, -0.1033, -0.1033, -0.0767, -0.0839, -0.0792, -0.0971, -0.09, -0.077, -0.0546, -0.0503, -0.0417, -0.0332, -0.0077, 0.0, 0.0099, -0.0104, 0.0111, 0.0027, 0.0008, 0.0008, -0.0179, -0.0141, 0.0317, 0.0442, 0.0501, 0.0715, 0.0617, 0.0533, 0.0951, 0.1007, 0.1007, 0.0743, 0.0682, 0.0868, 0.0771, 0.0801, 0.072, 0.1108, 0.1136, 0.1513, 0.1269, 0.1197, 0.0998, 0.1266, 0.1326, 0.1513, 0.138, 0.1313, 0.1387, 0.1327, 0.1026, 0.1204, 0.0898, 0.0846, 0.1374, 0.1336, 0.1349, 0.1452, 0.1318, 0.1298, 0.1198, 0.1317, 0.0908, 0.0716, 0.0657, 0.0922, 0.1044, 0.1191, 0.1217, 0.117, 0.0823, 0.0823, 0.0706, 0.0847, 0.1026, 0.1411, 0.1588, 0.1763, 0.2209, 0.2137, 0.2157, 0.2253, 0.2324, 0.2358, 0.2633, 0.2456, 0.2821, 0.3295, 0.3201, 0.301, 0.2855, 0.2849, 0.3134, 0.3566, 0.3549, 0.374, 0.4221, 0.3827, 0.432, 0.4102, 0.3198, 0.4244, 0.3917, 0.4329, 0.41, 0.3876, 0.4, 0.385, 0.3585, 0.3672, 0.3871, 0.4025, 0.418, 0.4347, 0.4259, 0.4128, 0.4336, 0.3828, 0.3809, 0.3602, 0.3473, 0.3885, 0.3693, 0.3667, 0.3271, 0.3394, 0.3262, 0.3069, 0.3279, 0.305, 0.2935, 0.3385, 0.3387, 0.3636, 0.3636, 0.3708, 0.3528, 0.3736, 0.3894, 0.3775]}
{"unit_id": "orig:1932073627807264938", "ticker_idx": 0, "ts": "2025-06-09T13:53:43+00:00", "symbol": "ADI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley prefers companies that have conservatively guided inventory-driven growth, such as ADI and NXPI", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley prefers ADI and NXPI on conservative inventory guidance; positive on memory leaders; cautious on cyclicals, favors AI growth names.", "resolved_tickers": ["ADI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Inventory Tracker: Overall Above 10-Year Median, De-stocking Momentum Slowing but Structural High Inventory May Persist\nMorgan Stanley (J. Moore, 25/06/09)\n\nAlthough inventory levels across segments remain above historical medians and saw a slight increase in Q1, the inventory destocking cycle has passed its inflection point. Subsequent tariff uncertainties may lead supply chains to maintain structurally elevated inventory levels.\n\n• Supply Chain Overview: Aggregate Days of Inventory (DOI) Rise Below Seasonal Norm\nIn Q1, total inventory days across manufacturers, distributors, and customers rose by 12 days quarter-over-quarter (QoQ), below the typical seasonal increase of 19 days. Distributor DOI rose for the second consecutive quarter, and customer inventory saw the largest increase, primarily due to the fastest decline in customer Cost of Goods Sold (COGS), coupled with incomplete digestion of incremental inventory. Overall, supply chain destocking momentum weakened, with inventory levels still significantly above the 10-year median.\n\n• Manufacturers and Segment Analysis: Divergent Inventory Contributions from Memory and Compute\nManufacturers’ DOI remained flat QoQ, below trend expectations. Within segments, DOI in the computing sector dropped by 15.5% QoQ, primarily driven by shipment recovery after the alleviation of NVIDIA Blackwell supply constraints. Memory segment DOI surged by 24.9%, reflecting both seasonal effects and substantial DRAM production increases to meet computing demand. Analog and MCU segments saw only a slight 2.7% rise, indicating a slow and moderate recovery phase.\n\n• Distributor Dynamics: Divergent Inventory and Shipment Trends\nOverall distributor DOI increased by 3 days, aligning with the seasonal expectation of 4 days, and currently stands 21 days above the historical median. Arrow and Avnet experienced DOI increases of 6–7 days due to a 7% drop in COGS, whereas WPG’s DOI fell by 2 days due to a 6% rise in COGS. Channel checks indicated that distributors highlighted volume growth in their Q1 results, though overall data still depict a state of “stalled destocking.”\n\n• End-Customers: Inventory Replenishment and Segment Differences\nEnd-customer DOI increased by 9 days QoQ, aligning with seasonal expectations. Manufacturers and consumer market sectors saw the largest increases, primarily due to sharp declines in COGS (48.6% and 40.6%, respectively). Conversely, storage (-43.1%), automotive supply chains (-7.2%), computing/mobile (-4.3%), and consumer sectors (-2.9%) experienced inventory reductions. The industry overall has passed inventory troughs, but macroeconomic factors and tariff-related noise may sustain elevated inventory levels longer than usual.\n\n• Implications\nInventory’s short-term fundamental impact is neutral to slightly positive, as moderate replenishment supports demand. However, ongoing tariff uncertainty could increase corporate willingness to restock, suppressing normal inventory cycle fluctuations. Morgan Stanley prefers companies that have conservatively guided inventory-driven growth, such as ADI and NXPI, and remains positive on memory leaders benefiting from the need to maintain higher inventory levels. The firm maintains a cautious stance on cyclical sectors, favoring high-growth AI narratives instead.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02371963141511968, "ret_signed_m1d": -0.02371963141511968, "alpha_spy_m1d": -0.022819160106357206, "alpha_c_m1d": -0.007258005381309163, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0249494940736823, "ret_signed_p1d": 0.0249494940736823, "alpha_spy_p1d": 0.01927982940479156, "alpha_c_p1d": 0.005063018749680959, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.018184978497047588, "ret_signed_p1w": 0.018184978497047588, "alpha_spy_p1w": 0.013182302922643219, "alpha_c_p1w": -0.006799575847238115, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.07682509559276185, "ret_signed_p1m": 0.07682509559276185, "alpha_spy_p1m": 0.039315168995534444, "alpha_c_p1m": -0.026381787081965813, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.08531728505718217, "ret_signed_p3m": 0.08531728505718217, "alpha_spy_p3m": -0.00032662656465243245, "alpha_c_p3m": -0.04271820623618572, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.2037668913060442, "ret_signed_p6m": 0.2037668913060442, "alpha_spy_p6m": 0.060757387921066774, "alpha_c_p6m": -0.19501122806713256, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.0965, -0.1089, -0.0872, -0.0744, -0.0854, -0.0819, -0.0997, -0.1062, -0.0715, -0.0661, -0.0794, -0.0864, -0.1187, -0.1182, -0.1291, -0.1273, -0.2091, -0.2803, -0.2508, -0.2738, -0.1402, -0.2201, -0.2166, -0.2191, -0.2149, -0.2307, -0.2292, -0.2292, -0.2348, -0.2195, -0.1993, -0.1463, -0.1491, -0.1529, -0.1575, -0.1477, -0.1529, -0.1316, -0.1354, -0.1447, -0.1261, -0.1133, -0.0926, -0.0243, -0.0088, -0.0082, -0.0069, -0.0069, -0.0182, -0.0184, -0.0283, -0.0733, -0.0797, -0.0797, -0.0516, -0.0572, -0.0566, -0.0643, -0.0579, -0.0449, -0.0402, -0.0423, -0.0237, 0.0, 0.0249, 0.0214, 0.0196, -0.0116, 0.0182, -0.001, 0.0087, 0.0087, 0.003, 0.0146, 0.0322, 0.0308, 0.0423, 0.0409, 0.0455, 0.057, 0.0768, 0.0792, 0.0792, 0.0622, 0.0768, 0.0662, 0.0767, 0.0748, 0.0694, 0.056, 0.0569, 0.0585, 0.0623, 0.0563, 0.0344, 0.0018, -0.0057, 0.0007, 0.0137, 0.0136, 0.0152, -0.0133, -0.0261, -0.0231, -0.0307, -0.0306, -0.0199, -0.0163, -0.0158, 0.0192, 0.0438, 0.0376, 0.0174, 0.0171, 0.0122, 0.0756, 0.0847, 0.1078, 0.1179, 0.1229, 0.1223, 0.1168, 0.1039, 0.1039, 0.0951, 0.0784, 0.0853, 0.0896, 0.098, 0.0944, 0.0902, 0.0947, 0.0813, 0.08, 0.0765, 0.0862, 0.0983, 0.0819, 0.0907, 0.0883, 0.0963, 0.0916, 0.0917, 0.0795, 0.0835, 0.0552, 0.0657, 0.0671, 0.0694, 0.0308, 0.0492, 0.049, -0.0064, 0.0349, 0.0381, 0.0502, 0.0655, 0.071, 0.0858, 0.0865, 0.06, 0.0729, 0.0496, 0.0716, 0.0555, 0.0365, 0.0271, 0.0325, 0.0302, 0.0115, 0.0407, 0.027, 0.0076, 0.0231, 0.0293, 0.0647, 0.0475, 0.0358, 0.014, 0.0148, 0.024, -0.0069, 0.0245, 0.0557, 0.1114, 0.1374, 0.1374, 0.1701, 0.1753, 0.2038]}
{"unit_id": "orig:1932073627807264938", "ticker_idx": 1, "ts": "2025-06-09T13:53:43+00:00", "symbol": "NXPI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley prefers companies that have conservatively guided inventory-driven growth, such as ADI and NXPI", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley prefers ADI and NXPI on conservative inventory guidance; positive on memory leaders; cautious on cyclicals, favors AI growth names.", "resolved_tickers": ["NXPI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Inventory Tracker: Overall Above 10-Year Median, De-stocking Momentum Slowing but Structural High Inventory May Persist\nMorgan Stanley (J. Moore, 25/06/09)\n\nAlthough inventory levels across segments remain above historical medians and saw a slight increase in Q1, the inventory destocking cycle has passed its inflection point. Subsequent tariff uncertainties may lead supply chains to maintain structurally elevated inventory levels.\n\n• Supply Chain Overview: Aggregate Days of Inventory (DOI) Rise Below Seasonal Norm\nIn Q1, total inventory days across manufacturers, distributors, and customers rose by 12 days quarter-over-quarter (QoQ), below the typical seasonal increase of 19 days. Distributor DOI rose for the second consecutive quarter, and customer inventory saw the largest increase, primarily due to the fastest decline in customer Cost of Goods Sold (COGS), coupled with incomplete digestion of incremental inventory. Overall, supply chain destocking momentum weakened, with inventory levels still significantly above the 10-year median.\n\n• Manufacturers and Segment Analysis: Divergent Inventory Contributions from Memory and Compute\nManufacturers’ DOI remained flat QoQ, below trend expectations. Within segments, DOI in the computing sector dropped by 15.5% QoQ, primarily driven by shipment recovery after the alleviation of NVIDIA Blackwell supply constraints. Memory segment DOI surged by 24.9%, reflecting both seasonal effects and substantial DRAM production increases to meet computing demand. Analog and MCU segments saw only a slight 2.7% rise, indicating a slow and moderate recovery phase.\n\n• Distributor Dynamics: Divergent Inventory and Shipment Trends\nOverall distributor DOI increased by 3 days, aligning with the seasonal expectation of 4 days, and currently stands 21 days above the historical median. Arrow and Avnet experienced DOI increases of 6–7 days due to a 7% drop in COGS, whereas WPG’s DOI fell by 2 days due to a 6% rise in COGS. Channel checks indicated that distributors highlighted volume growth in their Q1 results, though overall data still depict a state of “stalled destocking.”\n\n• End-Customers: Inventory Replenishment and Segment Differences\nEnd-customer DOI increased by 9 days QoQ, aligning with seasonal expectations. Manufacturers and consumer market sectors saw the largest increases, primarily due to sharp declines in COGS (48.6% and 40.6%, respectively). Conversely, storage (-43.1%), automotive supply chains (-7.2%), computing/mobile (-4.3%), and consumer sectors (-2.9%) experienced inventory reductions. The industry overall has passed inventory troughs, but macroeconomic factors and tariff-related noise may sustain elevated inventory levels longer than usual.\n\n• Implications\nInventory’s short-term fundamental impact is neutral to slightly positive, as moderate replenishment supports demand. However, ongoing tariff uncertainty could increase corporate willingness to restock, suppressing normal inventory cycle fluctuations. Morgan Stanley prefers companies that have conservatively guided inventory-driven growth, such as ADI and NXPI, and remains positive on memory leaders benefiting from the need to maintain higher inventory levels. The firm maintains a cautious stance on cyclical sectors, favoring high-growth AI narratives instead.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025342699225589516, "ret_signed_m1d": -0.025342699225589516, "alpha_spy_m1d": -0.024442227916827042, "alpha_c_m1d": -0.008881073191778999, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026468859785055976, "ret_signed_p1d": 0.026468859785055976, "alpha_spy_p1d": 0.020799195116165237, "alpha_c_p1d": 0.006582384461054636, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.020884177611983468, "ret_signed_p1w": 0.020884177611983468, "alpha_spy_p1w": 0.0158815020375791, "alpha_c_p1w": -0.0041003767323022355, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.09547202744002004, "ret_signed_p1m": 0.09547202744002004, "alpha_spy_p1m": 0.05796210084279263, "alpha_c_p1m": -0.007734855234707627, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.06270317795183122, "ret_signed_p3m": 0.06270317795183122, "alpha_spy_p3m": -0.022940733670003377, "alpha_c_p3m": -0.06533231334153666, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.020045395246087505, "ret_signed_p6m": 0.020045395246087505, "alpha_spy_p6m": -0.12296410813888992, "alpha_c_p6m": -0.37873272412708925, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.0452, -0.0612, -0.0372, -0.0201, -0.0328, -0.0268, -0.049, -0.0584, -0.0092, -0.0082, -0.0344, -0.0611, -0.1084, -0.108, -0.1083, -0.0911, -0.1934, -0.2453, -0.2337, -0.2796, -0.1277, -0.2229, -0.2079, -0.1996, -0.1945, -0.2088, -0.1987, -0.1987, -0.1951, -0.1803, -0.1512, -0.0916, -0.0917, -0.079, -0.143, -0.135, -0.1466, -0.117, -0.132, -0.1439, -0.123, -0.115, -0.0995, -0.0284, -0.0032, -0.0082, -0.011, -0.0024, -0.0128, -0.0141, -0.0363, -0.0778, -0.0982, -0.0982, -0.0665, -0.0792, -0.0807, -0.103, -0.0951, -0.0663, -0.0144, -0.029, -0.0253, 0.0, 0.0265, 0.0203, 0.0203, -0.0102, 0.0209, -0.0012, -0.0076, -0.0076, -0.0191, -0.0104, 0.0255, 0.0197, 0.0293, 0.0233, 0.0302, 0.043, 0.0899, 0.0943, 0.0943, 0.0656, 0.0955, 0.0864, 0.0995, 0.0793, 0.059, 0.0423, 0.04, 0.0585, 0.0651, 0.0763, 0.075, 0.0595, 0.0582, 0.0528, 0.0773, 0.0691, 0.0417, 0.0079, -0.0102, -0.0005, -0.0171, -0.0291, -0.0291, -0.0233, -0.0327, 0.0375, 0.0869, 0.0917, 0.0787, 0.0939, 0.081, 0.0786, 0.0558, 0.1072, 0.1159, 0.1213, 0.1206, 0.1272, 0.1073, 0.1073, 0.097, 0.076, 0.0627, 0.0691, 0.0632, 0.0547, 0.0339, 0.0524, 0.0317, 0.0338, 0.042, 0.051, 0.0729, 0.0613, 0.0692, 0.0687, 0.0784, 0.0743, 0.0707, 0.071, 0.0787, 0.0653, 0.0786, 0.0842, 0.0962, 0.0401, 0.0688, 0.0488, -0.0272, 0.0264, 0.0236, 0.029, 0.0298, 0.0153, 0.0412, 0.0532, 0.0286, 0.0455, 0.0381, 0.0495, 0.0087, -0.0304, -0.0224, -0.0095, -0.0034, -0.0317, -0.0032, -0.0221, -0.0311, -0.0284, -0.0391, -0.0333, -0.0469, -0.0664, -0.0976, -0.1067, -0.0997, -0.1275, -0.0936, -0.0926, -0.0952, -0.0822, -0.0822, -0.0766, -0.0551, 0.02]}
{"unit_id": "orig:1932073627807264938", "ticker_idx": 2, "ts": "2025-06-09T13:53:43+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "remains positive on memory leaders benefiting from the need to maintain higher inventory levels", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley prefers ADI and NXPI on conservative inventory guidance; positive on memory leaders; cautious on cyclicals, favors AI growth names.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Inventory Tracker: Overall Above 10-Year Median, De-stocking Momentum Slowing but Structural High Inventory May Persist\nMorgan Stanley (J. Moore, 25/06/09)\n\nAlthough inventory levels across segments remain above historical medians and saw a slight increase in Q1, the inventory destocking cycle has passed its inflection point. Subsequent tariff uncertainties may lead supply chains to maintain structurally elevated inventory levels.\n\n• Supply Chain Overview: Aggregate Days of Inventory (DOI) Rise Below Seasonal Norm\nIn Q1, total inventory days across manufacturers, distributors, and customers rose by 12 days quarter-over-quarter (QoQ), below the typical seasonal increase of 19 days. Distributor DOI rose for the second consecutive quarter, and customer inventory saw the largest increase, primarily due to the fastest decline in customer Cost of Goods Sold (COGS), coupled with incomplete digestion of incremental inventory. Overall, supply chain destocking momentum weakened, with inventory levels still significantly above the 10-year median.\n\n• Manufacturers and Segment Analysis: Divergent Inventory Contributions from Memory and Compute\nManufacturers’ DOI remained flat QoQ, below trend expectations. Within segments, DOI in the computing sector dropped by 15.5% QoQ, primarily driven by shipment recovery after the alleviation of NVIDIA Blackwell supply constraints. Memory segment DOI surged by 24.9%, reflecting both seasonal effects and substantial DRAM production increases to meet computing demand. Analog and MCU segments saw only a slight 2.7% rise, indicating a slow and moderate recovery phase.\n\n• Distributor Dynamics: Divergent Inventory and Shipment Trends\nOverall distributor DOI increased by 3 days, aligning with the seasonal expectation of 4 days, and currently stands 21 days above the historical median. Arrow and Avnet experienced DOI increases of 6–7 days due to a 7% drop in COGS, whereas WPG’s DOI fell by 2 days due to a 6% rise in COGS. Channel checks indicated that distributors highlighted volume growth in their Q1 results, though overall data still depict a state of “stalled destocking.”\n\n• End-Customers: Inventory Replenishment and Segment Differences\nEnd-customer DOI increased by 9 days QoQ, aligning with seasonal expectations. Manufacturers and consumer market sectors saw the largest increases, primarily due to sharp declines in COGS (48.6% and 40.6%, respectively). Conversely, storage (-43.1%), automotive supply chains (-7.2%), computing/mobile (-4.3%), and consumer sectors (-2.9%) experienced inventory reductions. The industry overall has passed inventory troughs, but macroeconomic factors and tariff-related noise may sustain elevated inventory levels longer than usual.\n\n• Implications\nInventory’s short-term fundamental impact is neutral to slightly positive, as moderate replenishment supports demand. However, ongoing tariff uncertainty could increase corporate willingness to restock, suppressing normal inventory cycle fluctuations. Morgan Stanley prefers companies that have conservatively guided inventory-driven growth, such as ADI and NXPI, and remains positive on memory leaders benefiting from the need to maintain higher inventory levels. The firm maintains a cautious stance on cyclical sectors, favoring high-growth AI narratives instead.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01763257724948845, "ret_signed_m1d": -0.01763257724948845, "alpha_spy_m1d": -0.016732105940725977, "alpha_c_m1d": -0.0011709512156779338, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008422801136441152, "ret_signed_p1d": 0.008422801136441152, "alpha_spy_p1d": 0.002753136467550413, "alpha_c_p1d": -0.011463674187560188, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.03987246510246708, "ret_signed_p1w": 0.03987246510246708, "alpha_spy_p1w": 0.03486978952806271, "alpha_c_p1w": 0.014887910758181376, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.12898545698485306, "ret_signed_p1m": 0.12898545698485306, "alpha_spy_p1m": 0.09147553038762565, "alpha_c_p1m": 0.025778574310125396, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.153687167442089, "ret_signed_p3m": 0.153687167442089, "alpha_spy_p3m": 0.06804325582025439, "alpha_c_p3m": 0.025651676148721103, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 1.1171126185425055, "ret_signed_p6m": 1.1171126185425055, "alpha_spy_p6m": 0.9741031151575281, "alpha_c_p6m": 0.7183344991693288, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1198, -0.1221, -0.0973, -0.0721, -0.0807, -0.071, -0.0522, -0.0608, -0.0666, -0.0838, -0.0724, -0.0833, -0.1095, -0.1397, -0.1195, -0.1185, -0.1728, -0.2281, -0.2586, -0.2586, -0.2309, -0.209, -0.2207, -0.2115, -0.2087, -0.234, -0.2319, -0.2308, -0.2341, -0.23, -0.2074, -0.1979, -0.1819, -0.1885, -0.1953, -0.2015, -0.1991, -0.1846, -0.1855, -0.1852, -0.1702, -0.1633, -0.1604, -0.1183, -0.1032, -0.0944, -0.1025, -0.0918, -0.1029, -0.1002, -0.1103, -0.1241, -0.1268, -0.1196, -0.1158, -0.0972, -0.0879, -0.1053, -0.0864, -0.0742, -0.051, -0.0245, -0.0176, 0.0, 0.0084, 0.0318, 0.0235, 0.0151, 0.0399, 0.0478, 0.0581, 0.0541, 0.0771, 0.0678, 0.1269, 0.1403, 0.1406, 0.1292, 0.1307, 0.1164, 0.1129, 0.1306, 0.1162, 0.1011, 0.129, 0.1154, 0.1446, 0.1542, 0.1439, 0.1529, 0.1438, 0.107, 0.1119, 0.1174, 0.0894, 0.0942, 0.0984, 0.0914, 0.1107, 0.1147, 0.1358, 0.1267, 0.0774, 0.0905, 0.1035, 0.0892, 0.1132, 0.1336, 0.1589, 0.1745, 0.1816, 0.1808, 0.1675, 0.1535, 0.1424, 0.1198, 0.1008, 0.1197, 0.1288, 0.1252, 0.1285, 0.1486, 0.1409, 0.1102, 0.1236, 0.1311, 0.1537, 0.1835, 0.1923, 0.2275, 0.2713, 0.312, 0.3745, 0.386, 0.4302, 0.4058, 0.4764, 0.4579, 0.475, 0.5017, 0.4864, 0.4742, 0.4306, 0.4759, 0.4824, 0.557, 0.6347, 0.6473, 0.6567, 0.6409, 0.6735, 0.6608, 0.7021, 0.7106, 0.6787, 0.7286, 0.8194, 0.839, 0.8825, 0.8562, 0.8545, 0.8631, 0.9591, 1.0168, 0.9877, 1.0615, 1.0885, 1.0938, 1.2359, 1.1012, 1.1254, 1.1404, 1.1129, 1.2123, 1.2109, 1.2171, 1.1842, 1.1066, 1.1781, 1.0694, 1.0427, 1.005, 0.9175, 0.9766, 0.9916, 1.0358, 1.0695, 1.0509, 1.0762, 1.1171]}
{"unit_id": "thread:1943166978493157565", "ticker_idx": 0, "ts": "2025-07-10T04:34:44+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn (Hon Hai), Wistron, and Quanta were given an 'Outperform' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rates Foxconn, Wistron, and Quanta Outperform on GB200 rack shipment ramp; author relays and adopts the directional call.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "• Morgan Stanley: Nvidia GB200 NVL72 server rack shipments in Q3 are expected to increase by more than 60% QoQ to 10,000-11,000 units. Among related Taiwanese companies, Foxconn (Hon Hai), Wistron, and Quanta were given an 'Outperform' rating.\n\n• GB200 server rack shipments are estimated to have increased to approximately 2,000-2,500 units in May and approximately 2,500 units in June, with total Q2 shipments reaching approximately 6,000 units.\n\n---\n\nhttps://t.co/Z03KRGEDww", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003095975232198178, "ret_signed_m1d": -0.003095975232198178, "alpha_spy_m1d": -0.0002835967935039818, "alpha_c_m1d": -0.00640363373656383, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0035154238705955576, "alpha_c_p1d": 0.004397294136100571, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": 0.015479876160990669, "ret_signed_p1w": 0.015479876160990669, "alpha_spy_p1w": 0.011932585093985537, "alpha_c_p1w": -0.00039665811209510515, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.20433436532507732, "ret_signed_p1m": 0.20433436532507732, "alpha_spy_p1m": 0.18618225945910338, "alpha_c_p1m": 0.16954561495298726, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.414860681114551, "ret_signed_p3m": 0.414860681114551, "alpha_spy_p3m": 0.3427072503993931, "alpha_c_p3m": 0.301750278435849, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.436532507739938, "ret_signed_p6m": 0.436532507739938, "alpha_spy_p6m": 0.3386316609478779, "alpha_c_p6m": 0.31033908486574546, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.1725, -0.1665, -0.1874, -0.1964, -0.1904, -0.1844, -0.2113, -0.1695, -0.1844, -0.1695, -0.1486, -0.1456, -0.1545, -0.1545, -0.1187, -0.1486, -0.1277, -0.1336, -0.1277, -0.1217, -0.0858, -0.056, -0.0261, -0.047, -0.056, -0.0799, -0.0799, -0.0739, -0.0858, -0.0799, -0.0858, -0.0918, -0.0918, -0.0679, -0.0679, -0.0888, -0.0948, -0.0679, -0.0709, -0.0858, -0.0769, -0.0709, -0.0679, -0.0589, -0.0649, -0.0739, -0.0739, -0.0679, -0.0799, -0.0709, -0.0679, -0.041, -0.0321, -0.0261, -0.0141, -0.038, -0.0111, 0.0031, 0.0155, -0.0031, -0.0031, -0.0155, -0.0031, 0.0, 0.0, -0.0062, 0.0093, 0.0062, 0.0155, 0.0248, 0.0217, 0.0031, 0.0279, 0.0774, 0.0805, 0.0898, 0.0619, 0.0619, 0.1022, 0.1022, 0.1146, 0.1424, 0.1517, 0.2043, 0.2043, 0.226, 0.2198, 0.2291, 0.2353, 0.2817, 0.3003, 0.2879, 0.2415, 0.2755, 0.2539, 0.2848, 0.2941, 0.2879, 0.2755, 0.2601, 0.2291, 0.2291, 0.2415, 0.257, 0.2693, 0.2601, 0.2848, 0.3003, 0.3282, 0.3467, 0.3375, 0.3344, 0.3127, 0.3313, 0.3251, 0.3375, 0.3684, 0.3808, 0.4241, 0.3591, 0.3591, 0.3375, 0.356, 0.3963, 0.4025, 0.4025, 0.4149, 0.3932, 0.3715, 0.3715, 0.3189, 0.2755, 0.2786, 0.3808, 0.4025, 0.4768, 0.4799, 0.5015, 0.4799, 0.4799, 0.5263, 0.548, 0.6099, 0.6223, 0.5944, 0.5573, 0.5108, 0.5294, 0.5356, 0.5108, 0.5449, 0.5263, 0.5542, 0.5604, 0.4923, 0.4613, 0.4241, 0.418, 0.4644, 0.3932, 0.3622, 0.356, 0.4056, 0.4303, 0.3963, 0.3746, 0.3746, 0.4056, 0.4149, 0.4303, 0.4365, 0.4551, 0.4458, 0.3994, 0.4056, 0.3715, 0.3498, 0.3406, 0.3375, 0.3715, 0.387, 0.387, 0.387, 0.387, 0.3963, 0.4303, 0.4118, 0.4272, 0.4272, 0.4365]}
{"unit_id": "thread:1943166978493157565", "ticker_idx": 1, "ts": "2025-07-10T04:34:44+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn (Hon Hai), Wistron, and Quanta were given an 'Outperform' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rates Foxconn, Wistron, and Quanta Outperform on GB200 rack shipment ramp; author relays and adopts the directional call.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "• Morgan Stanley: Nvidia GB200 NVL72 server rack shipments in Q3 are expected to increase by more than 60% QoQ to 10,000-11,000 units. Among related Taiwanese companies, Foxconn (Hon Hai), Wistron, and Quanta were given an 'Outperform' rating.\n\n• GB200 server rack shipments are estimated to have increased to approximately 2,000-2,500 units in May and approximately 2,500 units in June, with total Q2 shipments reaching approximately 6,000 units.\n\n---\n\nhttps://t.co/Z03KRGEDww", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004166666666666652, "ret_signed_m1d": 0.004166666666666652, "alpha_spy_m1d": 0.006979045105360848, "alpha_c_m1d": 0.0008590081623010004, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004166666666666652, "ret_signed_p1d": -0.004166666666666652, "alpha_spy_p1d": -0.0006512427960710943, "alpha_c_p1d": 0.00023062746943391943, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": -0.01666666666666672, "ret_signed_p1w": -0.01666666666666672, "alpha_spy_p1w": -0.02021395773367185, "alpha_c_p1w": -0.03254320093975249, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.03750000000000009, "ret_signed_p1m": 0.03750000000000009, "alpha_spy_p1m": 0.019347894134026156, "alpha_c_p1m": 0.0027112496279100284, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.30000000000000004, "ret_signed_p3m": 0.30000000000000004, "alpha_spy_p3m": 0.22784656928484215, "alpha_c_p3m": 0.18688959732129806, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.29166666666666674, "ret_signed_p6m": 0.29166666666666674, "alpha_spy_p6m": 0.19376581987460662, "alpha_c_p6m": 0.1654732437924742, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.216, -0.1741, -0.1966, -0.1902, -0.1539, -0.17, -0.2039, -0.1741, -0.1861, -0.1741, -0.1741, -0.162, -0.1821, -0.1821, -0.1499, -0.166, -0.1217, -0.1418, -0.1378, -0.1378, -0.1257, -0.1257, -0.0774, -0.1015, -0.1056, -0.1217, -0.1176, -0.0894, -0.0894, -0.0894, -0.0653, -0.0733, -0.0693, -0.0612, -0.0612, -0.0733, -0.0875, -0.05, -0.0417, -0.0833, -0.0542, -0.0458, -0.0417, -0.0292, -0.0125, -0.0083, -0.0167, -0.0333, -0.0375, -0.0125, -0.0125, 0.0, 0.0458, 0.0417, 0.0083, 0.0208, 0.0, 0.0, 0.0083, -0.0083, -0.0167, -0.0125, 0.0042, 0.0, -0.0042, -0.025, 0.0, -0.0167, -0.0167, -0.0083, -0.0125, -0.0458, -0.0333, -0.025, -0.0292, 0.0, 0.0042, 0.0042, 0.025, 0.025, -0.0042, 0.0333, 0.0208, 0.0292, 0.0375, 0.0625, 0.05, -0.0208, -0.0292, -0.0292, -0.0333, -0.0375, -0.0542, -0.0708, -0.0667, -0.0542, -0.0583, -0.0417, -0.0417, -0.0583, -0.0875, -0.0833, -0.0833, -0.0583, -0.0375, -0.0458, -0.0333, 0.0, -0.0083, 0.0, -0.0042, -0.0167, -0.0083, 0.0, 0.0083, 0.0125, 0.0208, 0.0125, 0.0458, 0.0667, 0.0667, 0.1708, 0.2167, 0.2458, 0.2792, 0.2792, 0.3, 0.2542, 0.2583, 0.2583, 0.2125, 0.15, 0.15, 0.1625, 0.1375, 0.1833, 0.1875, 0.1958, 0.1917, 0.1917, 0.2208, 0.2333, 0.2708, 0.2917, 0.2542, 0.1958, 0.1583, 0.1458, 0.1833, 0.1458, 0.1542, 0.2042, 0.1583, 0.2, 0.1708, 0.15, 0.1417, 0.1417, 0.1917, 0.1583, 0.1583, 0.1625, 0.1875, 0.2125, 0.2042, 0.175, 0.1917, 0.2083, 0.2167, 0.2583, 0.2417, 0.2458, 0.2292, 0.2, 0.1958, 0.1792, 0.1583, 0.15, 0.1625, 0.2, 0.2208, 0.2083, 0.2125, 0.2125, 0.2208, 0.2167, 0.225, 0.2542, 0.2542, 0.2917]}
{"unit_id": "thread:1943166978493157565", "ticker_idx": 2, "ts": "2025-07-10T04:34:44+00:00", "symbol": "Quanta", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn (Hon Hai), Wistron, and Quanta were given an 'Outperform' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rates Foxconn, Wistron, and Quanta Outperform on GB200 rack shipment ramp; author relays and adopts the directional call.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta in server/ODM context = Quanta Computer, TWSE 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "• Morgan Stanley: Nvidia GB200 NVL72 server rack shipments in Q3 are expected to increase by more than 60% QoQ to 10,000-11,000 units. Among related Taiwanese companies, Foxconn (Hon Hai), Wistron, and Quanta were given an 'Outperform' rating.\n\n• GB200 server rack shipments are estimated to have increased to approximately 2,000-2,500 units in May and approximately 2,500 units in June, with total Q2 shipments reaching approximately 6,000 units.\n\n---\n\nhttps://t.co/Z03KRGEDww", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01098901098901095, "ret_signed_m1d": 0.01098901098901095, "alpha_spy_m1d": 0.013801389427705146, "alpha_c_m1d": 0.0076813524846452985, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0035154238705955576, "alpha_c_p1d": 0.004397294136100571, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": -0.0073260073260073, "ret_signed_p1w": -0.0073260073260073, "alpha_spy_p1w": -0.010873298393012432, "alpha_c_p1w": -0.023202541599093074, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.0641025641025641, "ret_signed_p1m": 0.0641025641025641, "alpha_spy_p1m": 0.045950458236590164, "alpha_c_p1m": 0.029313813730474036, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.13003663003663002, "ret_signed_p3m": 0.13003663003663002, "alpha_spy_p3m": 0.05788319932147212, "alpha_c_p3m": 0.016926227357928036, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.0146520146520146, "ret_signed_p6m": 0.0146520146520146, "alpha_spy_p6m": -0.08324883214004553, "alpha_c_p6m": -0.11154140822217795, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.2029, -0.2029, -0.2064, -0.1959, -0.1941, -0.1889, -0.2186, -0.1784, -0.1959, -0.1784, -0.1434, -0.1522, -0.1662, -0.1662, -0.1225, -0.1382, -0.112, -0.112, -0.1207, -0.1032, -0.0962, -0.084, -0.0333, -0.0298, -0.0385, -0.07, -0.0683, -0.049, -0.0648, -0.077, -0.0718, -0.07, -0.0595, -0.0508, -0.0508, -0.0525, -0.0385, -0.0211, -0.0158, -0.035, -0.035, -0.0088, -0.0106, 0.0017, 0.0052, -0.0001, -0.0106, -0.0106, -0.0246, -0.0228, -0.0141, 0.0122, 0.0244, 0.0139, -0.0018, 0.0055, 0.0293, 0.033, 0.0256, 0.0183, 0.0147, -0.0037, 0.011, 0.0, 0.0, -0.0128, -0.0092, -0.0018, -0.0073, 0.0018, 0.0018, -0.0385, -0.0311, -0.0147, -0.0165, -0.0037, -0.0201, -0.011, 0.0311, 0.0311, 0.0293, 0.0549, 0.0568, 0.0586, 0.0641, 0.0623, 0.0366, -0.0055, -0.0092, -0.0092, -0.0147, -0.0183, -0.0495, -0.0513, -0.0531, -0.0183, -0.0293, -0.0147, -0.0256, -0.0385, -0.0623, -0.0714, -0.0714, -0.0586, -0.0531, -0.0476, -0.0385, 0.0, -0.011, 0.0037, 0.011, 0.0018, 0.0, 0.0238, 0.0293, 0.0293, 0.0495, 0.0385, 0.044, 0.044, 0.044, 0.0623, 0.0495, 0.0769, 0.0989, 0.0989, 0.13, 0.1209, 0.0916, 0.0916, 0.0788, 0.0403, 0.0476, 0.0678, 0.0568, 0.0861, 0.0659, 0.0604, 0.0751, 0.0751, 0.1337, 0.1136, 0.1282, 0.1209, 0.1007, 0.0897, 0.0531, 0.0604, 0.0733, 0.0513, 0.0678, 0.0678, 0.0549, 0.044, 0.0238, 0.011, -0.011, -0.022, 0.0037, -0.011, -0.0037, 0.0073, 0.0238, 0.0311, 0.033, 0.0513, 0.0568, 0.0714, 0.0788, 0.0897, 0.0897, 0.0842, 0.0714, 0.0549, 0.0495, 0.044, 0.0, -0.0293, -0.0476, -0.0293, -0.022, -0.0311, -0.033, -0.033, -0.0366, -0.0366, -0.0366, -0.0037, -0.0037, 0.0147]}
{"unit_id": "orig:1942351335887888602", "ticker_idx": 0, "ts": "2025-07-07T22:33:40+00:00", "symbol": "ASMPT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If next-generation CoWoS adopts the Fluxless TCB process, K&S and ASMPT will benefit the most", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS advanced packaging report: K&S and ASMPT benefit most if CoWoS shifts to Fluxless TCB; Besi positioned if flip-chip maintained; OSAT recovery could trigger equipment orders.", "resolved_tickers": ["0522.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "ASMPT listed on Hong Kong Exchange 0522.HK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Advanced Packaging: New Architectures Drive Supply Chain Restructuring, HBM and CoWoS Create Opportunities for Vendor Differentiation\n\nUBS (M. Jenkins, 25/07/07)\n\nDuring its inaugural global advanced packaging conference, UBS noted that as AI demand continues to drive packaging technology evolution, the supply chain structure is undergoing restructuring due to new design solutions. Key changes are centered around the WMCM architecture, divergences in HBM packaging pathways, and uncertainties regarding next-generation CoWoS processes.\n\nWMCM to Reshape High-End Smartphone Packaging Architecture, Benefiting Multi-Die Heterogeneous Integration\n\nApple's plan to adopt WMCM (Wafer-Level Multi-Chip Module) has drawn significant attention across the supply chain. Compared to InFO, which only connects SoC and DRAM, WMCM supports the direct connection of multiple dies such as CPU/GPU/memory to a fan-out substrate without an organic substrate. Although flip-chip remains dominant in the short term, UBS expects a gradual shift towards Hybrid Bonding in the future, which is likely to spur demand for more chiplet architectures.\n\nHBM Packaging Path Divergences Emerge, Thermal Bottlenecks May Accelerate Shift to Advanced Technologies\n\nSamsung and Micron tend to adopt Fluxless TCB (Thermosonic Compression Bonding) and Hybrid Bonding to reduce thermal resistance and enhance integration, while SK Hynix may continue with its MR-MUF (Mass Reflow-Molded Underfill) process. UBS points out that 16Hi stacks are starting to encounter thermal bottlenecks, which are difficult to alleviate even with reflow materials. This will drive the penetration of higher yield, lower temperature/pressure packaging technologies. The base scenario predicts that HBM4E 16Hi will begin transitioning to Fluxless TCB, and 20Hi stages will see large-scale adoption of Hybrid Bonding.\n\nHigh Uncertainty in CoWoS Evolution, Equipment Supplier Landscape May See a Reshuffle\n\nCurrently, the TSMC CoW (Chip-on-Wafer) portion is mainly supplied by Shibaura, while ASMPT and Besi are more advanced in WoS (Wafer-on-Substrate). If next-generation CoWoS adopts the Fluxless TCB process, K&S and ASMPT will benefit the most; the former's formic acid process offers high yield but lower capacity. If flip-chip solutions are maintained, Besi plans to launch its new-generation 8800 Chameo tool in 2026 to capture the CoW market, with placement accuracy reaching 1μm.\n\nOSAT Capacity Utilization Recovers to 75%, Still Needs to Exceed 80% to Trigger New Equipment Orders\n\nMost OSAT (Outsourced Semiconductor Assembly and Test) vendors have seen their utilization rates in traditional applications recover to approximately 75%. If this further increases to above 80%, it is expected to lead to a new round of equipment order growth, benefiting upstream equipment manufacturers.\n\nSignificant Divergence in Views Among Global Manufacturers, Supply Chain on the Eve of Transformation\n\nDespite a general optimism for future growth driven by the AI trend, significant divergences still exist among manufacturers regarding key packaging roadmaps, technology selections, and equipment pathways. UBS believes this provides space for equipment vendors to diversify competition and make breakthroughs, with the industry potentially reaching a critical evolution inflection point within the next two to three years.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013793071826071035, "ret_signed_m1d": 0.013793071826071035, "alpha_spy_m1d": 0.0062851052475698665, "alpha_c_m1d": -4.453872219922772e-06, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018965531906749655, "ret_signed_p1d": 0.018965531906749655, "alpha_spy_p1d": 0.019513248124494953, "alpha_c_p1d": 0.005668403418577217, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.04482756650030506, "ret_signed_p1w": 0.04482756650030506, "alpha_spy_p1w": 0.03817365244167026, "alpha_c_p1w": 0.024953382531836654, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": 0.16551726062760896, "ret_signed_p1m": 0.16551726062760896, "alpha_spy_p1m": 0.15377215424838608, "alpha_c_p1m": 0.13928049275807775, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.4710603526552277, "ret_signed_p3m": 0.4710603526552277, "alpha_spy_p3m": 0.38986665459492453, "alpha_c_p3m": 0.2620240960159099, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.3369341705988802, "ret_signed_p6m": 0.3369341705988802, "alpha_spy_p6m": 0.22371895739664827, "alpha_c_p6m": 0.03424452191645799, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.1832, -0.1772, -0.1497, -0.1257, -0.1429, -0.1609, -0.166, -0.166, -0.166, -0.1626, -0.1446, -0.1446, -0.136, -0.1232, -0.1112, -0.1052, -0.1052, -0.0846, -0.0846, -0.0872, -0.0966, -0.0863, -0.0812, -0.0452, -0.0397, -0.025, -0.025, -0.0371, -0.0793, -0.0724, -0.069, -0.081, -0.0966, -0.0853, -0.0974, -0.0871, -0.0716, -0.0879, -0.0767, -0.0612, -0.0629, -0.0414, -0.0474, -0.0172, -0.0103, -0.0397, -0.0629, -0.0629, -0.0595, -0.0655, -0.0853, -0.0974, -0.0853, -0.0543, -0.0431, -0.0284, -0.0336, -0.0026, -0.0078, -0.0078, -0.0129, -0.0086, 0.0138, 0.0, 0.019, 0.0121, 0.0129, 0.0405, 0.0448, 0.0672, 0.0569, 0.0793, 0.0733, 0.0897, 0.0552, 0.0897, 0.2103, 0.2172, 0.1871, 0.1957, 0.1526, 0.1534, 0.1241, 0.1647, 0.1655, 0.1664, 0.1974, 0.1897, 0.2149, 0.2175, 0.2426, 0.2097, 0.2417, 0.2132, 0.1543, 0.1518, 0.1431, 0.2063, 0.298, 0.2582, 0.2115, 0.2582, 0.2201, 0.2288, 0.1777, 0.1682, 0.1604, 0.176, 0.182, 0.1855, 0.1916, 0.2288, 0.2279, 0.2158, 0.2409, 0.2331, 0.3231, 0.3317, 0.3715, 0.3672, 0.452, 0.4096, 0.3646, 0.3767, 0.42, 0.42, 0.4711, 0.452, 0.5455, 0.5455, 0.5593, 0.6502, 0.6061, 0.555, 0.4529, 0.5195, 0.5282, 0.4572, 0.5065, 0.4858, 0.446, 0.4408, 0.5576, 0.6225, 0.51, 0.51, 0.4512, 0.4191, 0.4295, 0.4131, 0.4114, 0.4243, 0.4044, 0.4252, 0.4001, 0.3724, 0.3863, 0.3482, 0.349, 0.3153, 0.2989, 0.3127, 0.2322, 0.2262, 0.2686, 0.2547, 0.2816, 0.3032, 0.324, 0.3231, 0.3387, 0.3508, 0.3603, 0.3473, 0.369, 0.3343, 0.3646, 0.3629, 0.3015, 0.2937, 0.3144, 0.3058, 0.2867, 0.3387, 0.3404, 0.3395, 0.3395, 0.3395, 0.3179, 0.3369]}
{"unit_id": "orig:1942351335887888602", "ticker_idx": 1, "ts": "2025-07-07T22:33:40+00:00", "symbol": "Kulicke and Soffa", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "If next-generation CoWoS adopts the Fluxless TCB process, K&S and ASMPT will benefit the most", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS advanced packaging report: K&S and ASMPT benefit most if CoWoS shifts to Fluxless TCB; Besi positioned if flip-chip maintained; OSAT recovery could trigger equipment orders.", "resolved_tickers": ["KLIC"], "resolution_method": "exchange_ticker", "resolution_reason": "Kulicke and Soffa Industries KLIC US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Advanced Packaging: New Architectures Drive Supply Chain Restructuring, HBM and CoWoS Create Opportunities for Vendor Differentiation\n\nUBS (M. Jenkins, 25/07/07)\n\nDuring its inaugural global advanced packaging conference, UBS noted that as AI demand continues to drive packaging technology evolution, the supply chain structure is undergoing restructuring due to new design solutions. Key changes are centered around the WMCM architecture, divergences in HBM packaging pathways, and uncertainties regarding next-generation CoWoS processes.\n\nWMCM to Reshape High-End Smartphone Packaging Architecture, Benefiting Multi-Die Heterogeneous Integration\n\nApple's plan to adopt WMCM (Wafer-Level Multi-Chip Module) has drawn significant attention across the supply chain. Compared to InFO, which only connects SoC and DRAM, WMCM supports the direct connection of multiple dies such as CPU/GPU/memory to a fan-out substrate without an organic substrate. Although flip-chip remains dominant in the short term, UBS expects a gradual shift towards Hybrid Bonding in the future, which is likely to spur demand for more chiplet architectures.\n\nHBM Packaging Path Divergences Emerge, Thermal Bottlenecks May Accelerate Shift to Advanced Technologies\n\nSamsung and Micron tend to adopt Fluxless TCB (Thermosonic Compression Bonding) and Hybrid Bonding to reduce thermal resistance and enhance integration, while SK Hynix may continue with its MR-MUF (Mass Reflow-Molded Underfill) process. UBS points out that 16Hi stacks are starting to encounter thermal bottlenecks, which are difficult to alleviate even with reflow materials. This will drive the penetration of higher yield, lower temperature/pressure packaging technologies. The base scenario predicts that HBM4E 16Hi will begin transitioning to Fluxless TCB, and 20Hi stages will see large-scale adoption of Hybrid Bonding.\n\nHigh Uncertainty in CoWoS Evolution, Equipment Supplier Landscape May See a Reshuffle\n\nCurrently, the TSMC CoW (Chip-on-Wafer) portion is mainly supplied by Shibaura, while ASMPT and Besi are more advanced in WoS (Wafer-on-Substrate). If next-generation CoWoS adopts the Fluxless TCB process, K&S and ASMPT will benefit the most; the former's formic acid process offers high yield but lower capacity. If flip-chip solutions are maintained, Besi plans to launch its new-generation 8800 Chameo tool in 2026 to capture the CoW market, with placement accuracy reaching 1μm.\n\nOSAT Capacity Utilization Recovers to 75%, Still Needs to Exceed 80% to Trigger New Equipment Orders\n\nMost OSAT (Outsourced Semiconductor Assembly and Test) vendors have seen their utilization rates in traditional applications recover to approximately 75%. If this further increases to above 80%, it is expected to lead to a new round of equipment order growth, benefiting upstream equipment manufacturers.\n\nSignificant Divergence in Views Among Global Manufacturers, Supply Chain on the Eve of Transformation\n\nDespite a general optimism for future growth driven by the AI trend, significant divergences still exist among manufacturers regarding key packaging roadmaps, technology selections, and equipment pathways. UBS believes this provides space for equipment vendors to diversify competition and make breakthroughs, with the industry potentially reaching a critical evolution inflection point within the next two to three years.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029980401280719438, "ret_signed_m1d": 0.029980401280719438, "alpha_spy_m1d": 0.02247243470221827, "alpha_c_m1d": 0.01618287558242848, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03250203028097909, "ret_signed_p1d": 0.03250203028097909, "alpha_spy_p1d": 0.033049746498724386, "alpha_c_p1d": 0.01920490179280665, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.0022416163710081705, "ret_signed_p1w": 0.0022416163710081705, "alpha_spy_p1w": -0.004412297687626632, "alpha_c_p1w": -0.017632567597460236, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": -0.09330335780277865, "ret_signed_p1m": -0.09330335780277865, "alpha_spy_p1m": -0.10504846418200153, "alpha_c_p1m": -0.11954012567230987, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.15789097128663698, "ret_signed_p3m": 0.15789097128663698, "alpha_spy_p3m": 0.0766972732263338, "alpha_c_p3m": -0.051145285352680814, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.3118835216734863, "ret_signed_p6m": 0.3118835216734863, "alpha_spy_p6m": 0.19866830847125438, "alpha_c_p6m": 0.009193872991064111, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.1054, -0.1811, -0.168, -0.1616, -0.1513, -0.1839, -0.1895, -0.1895, -0.1806, -0.1616, -0.1472, -0.1087, -0.1065, -0.0998, -0.0914, -0.1023, -0.1026, -0.0747, -0.0875, -0.116, -0.1407, -0.1165, -0.1134, -0.0541, -0.0363, -0.046, -0.0491, -0.058, -0.0786, -0.0736, -0.0937, -0.1015, -0.1115, -0.1115, -0.0895, -0.097, -0.0887, -0.1045, -0.1026, -0.0817, -0.0694, -0.0836, -0.0761, -0.0486, -0.0196, -0.0327, -0.0313, -0.0541, -0.0154, -0.0382, -0.0462, -0.0462, -0.0625, -0.0538, -0.0171, -0.0213, -0.0087, -0.0263, -0.0305, 0.0003, 0.0305, 0.03, 0.03, 0.0, 0.0325, 0.0249, 0.0353, 0.016, 0.0022, -0.0148, -0.0143, 0.0059, -0.0076, -0.0008, -0.0132, -0.0255, -0.0403, -0.0474, -0.0129, -0.0233, -0.044, -0.0821, -0.0947, -0.0796, -0.0933, -0.1003, -0.0199, -0.0076, -0.0177, 0.0409, 0.0572, 0.0479, 0.0118, 0.0255, 0.0163, 0.0056, 0.0193, 0.0555, 0.0482, 0.0535, 0.0541, 0.0591, 0.0507, 0.0507, 0.0233, 0.0191, 0.0586, 0.0706, 0.0689, 0.0636, 0.0639, 0.0933, 0.0821, 0.0972, 0.1524, 0.1572, 0.2117, 0.1824, 0.2091, 0.1982, 0.1686, 0.1559, 0.1531, 0.1421, 0.1444, 0.1616, 0.1579, 0.1359, 0.1717, 0.1278, 0.1469, 0.1511, 0.0751, 0.1275, 0.1278, 0.128, 0.119, 0.0943, 0.1151, 0.1235, 0.0844, 0.1311, 0.1492, 0.1497, 0.1542, 0.1182, 0.1142, 0.1244, 0.1325, 0.081, 0.1123, 0.0819, 0.0821, 0.1038, 0.0883, 0.0982, 0.0593, 0.0464, 0.0036, 0.0013, -0.0063, 0.099, 0.1511, 0.17, 0.2266, 0.266, 0.266, 0.2702, 0.261, 0.2708, 0.3235, 0.3395, 0.3654, 0.3665, 0.3798, 0.3947, 0.3829, 0.3539, 0.3463, 0.317, 0.2885, 0.2901, 0.2915, 0.3085, 0.311, 0.3139, 0.3139, 0.3223, 0.3093, 0.3119]}
{"unit_id": "orig:1942351335887888602", "ticker_idx": 2, "ts": "2025-07-07T22:33:40+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If flip-chip solutions are maintained, Besi plans to launch its new-generation 8800 Chameo tool in 2026 to capture the CoW market", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS advanced packaging report: K&S and ASMPT benefit most if CoWoS shifts to Fluxless TCB; Besi positioned if flip-chip maintained; OSAT recovery could trigger equipment orders.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Advanced Packaging: New Architectures Drive Supply Chain Restructuring, HBM and CoWoS Create Opportunities for Vendor Differentiation\n\nUBS (M. Jenkins, 25/07/07)\n\nDuring its inaugural global advanced packaging conference, UBS noted that as AI demand continues to drive packaging technology evolution, the supply chain structure is undergoing restructuring due to new design solutions. Key changes are centered around the WMCM architecture, divergences in HBM packaging pathways, and uncertainties regarding next-generation CoWoS processes.\n\nWMCM to Reshape High-End Smartphone Packaging Architecture, Benefiting Multi-Die Heterogeneous Integration\n\nApple's plan to adopt WMCM (Wafer-Level Multi-Chip Module) has drawn significant attention across the supply chain. Compared to InFO, which only connects SoC and DRAM, WMCM supports the direct connection of multiple dies such as CPU/GPU/memory to a fan-out substrate without an organic substrate. Although flip-chip remains dominant in the short term, UBS expects a gradual shift towards Hybrid Bonding in the future, which is likely to spur demand for more chiplet architectures.\n\nHBM Packaging Path Divergences Emerge, Thermal Bottlenecks May Accelerate Shift to Advanced Technologies\n\nSamsung and Micron tend to adopt Fluxless TCB (Thermosonic Compression Bonding) and Hybrid Bonding to reduce thermal resistance and enhance integration, while SK Hynix may continue with its MR-MUF (Mass Reflow-Molded Underfill) process. UBS points out that 16Hi stacks are starting to encounter thermal bottlenecks, which are difficult to alleviate even with reflow materials. This will drive the penetration of higher yield, lower temperature/pressure packaging technologies. The base scenario predicts that HBM4E 16Hi will begin transitioning to Fluxless TCB, and 20Hi stages will see large-scale adoption of Hybrid Bonding.\n\nHigh Uncertainty in CoWoS Evolution, Equipment Supplier Landscape May See a Reshuffle\n\nCurrently, the TSMC CoW (Chip-on-Wafer) portion is mainly supplied by Shibaura, while ASMPT and Besi are more advanced in WoS (Wafer-on-Substrate). If next-generation CoWoS adopts the Fluxless TCB process, K&S and ASMPT will benefit the most; the former's formic acid process offers high yield but lower capacity. If flip-chip solutions are maintained, Besi plans to launch its new-generation 8800 Chameo tool in 2026 to capture the CoW market, with placement accuracy reaching 1μm.\n\nOSAT Capacity Utilization Recovers to 75%, Still Needs to Exceed 80% to Trigger New Equipment Orders\n\nMost OSAT (Outsourced Semiconductor Assembly and Test) vendors have seen their utilization rates in traditional applications recover to approximately 75%. If this further increases to above 80%, it is expected to lead to a new round of equipment order growth, benefiting upstream equipment manufacturers.\n\nSignificant Divergence in Views Among Global Manufacturers, Supply Chain on the Eve of Transformation\n\nDespite a general optimism for future growth driven by the AI trend, significant divergences still exist among manufacturers regarding key packaging roadmaps, technology selections, and equipment pathways. UBS believes this provides space for equipment vendors to diversify competition and make breakthroughs, with the industry potentially reaching a critical evolution inflection point within the next two to three years.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0028830439601594815, "ret_signed_m1d": -0.0028830439601594815, "alpha_spy_m1d": -0.01039101053866065, "alpha_c_m1d": -0.01668056965845044, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0065897424903180735, "ret_signed_p1d": 0.0065897424903180735, "alpha_spy_p1d": 0.007137458708063371, "alpha_c_p1d": -0.006707385997854365, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.026770986970144017, "ret_signed_p1w": 0.026770986970144017, "alpha_spy_p1w": 0.020117072911509215, "alpha_c_p1w": 0.006896803001675611, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": -0.050247102695128776, "ret_signed_p1m": -0.050247102695128776, "alpha_spy_p1m": -0.061992209074351656, "alpha_c_p1m": -0.07648387056466, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.100082378105258, "ret_signed_p3m": 0.100082378105258, "alpha_spy_p3m": 0.01888868004495481, "alpha_c_p3m": -0.1089538785340598, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.0963756795750994, "ret_signed_p6m": 0.0963756795750994, "alpha_spy_p6m": -0.01683953362713253, "alpha_c_p6m": -0.2063139691073228, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.3591, -0.3349, -0.3303, -0.3082, -0.2161, -0.2412, -0.2435, -0.2435, -0.2435, -0.2499, -0.2437, -0.2293, -0.2007, -0.1951, -0.1878, -0.2237, -0.2237, -0.1787, -0.1755, -0.1858, -0.1433, -0.126, -0.1112, -0.0527, -0.0222, -0.0103, -0.0486, -0.0667, -0.0869, -0.1063, -0.1034, -0.1096, -0.1326, -0.1145, -0.0984, -0.103, -0.0898, -0.1223, -0.1437, -0.1207, -0.1108, -0.089, -0.0643, -0.054, -0.0091, -0.0243, 0.0111, 0.0371, 0.0461, 0.0449, 0.0358, 0.0272, 0.0144, 0.0354, 0.0618, 0.0643, 0.0692, 0.0745, 0.0465, 0.0062, 0.0152, 0.0045, -0.0029, 0.0, 0.0066, 0.0045, 0.0494, 0.0474, 0.0268, 0.0404, -0.0119, 0.0502, 0.054, 0.0721, 0.0362, 0.0185, 0.0461, -0.0354, 0.0115, -0.0115, 0.0082, -0.0189, -0.035, -0.0247, -0.0502, -0.0544, -0.0119, -0.0054, 0.0012, 0.0194, 0.0173, 0.0103, -0.0235, -0.0297, -0.0165, -0.0354, -0.049, -0.0198, -0.0185, -0.0058, -0.0132, -0.0136, -0.0519, -0.0486, -0.1194, -0.1285, -0.1174, -0.1046, -0.0869, -0.082, -0.0745, -0.0638, -0.0787, -0.0268, -0.0395, -0.0474, 0.028, 0.0008, 0.0115, 0.0453, 0.0589, 0.0437, 0.0082, 0.0412, 0.0449, 0.0527, 0.1001, 0.075, 0.2088, 0.2092, 0.2055, 0.1977, 0.1557, 0.1759, 0.1573, 0.1985, 0.1952, 0.1672, 0.196, 0.1886, 0.1409, 0.1882, 0.2113, 0.2191, 0.2076, 0.2162, 0.2113, 0.2158, 0.201, 0.1742, 0.1656, 0.1347, 0.105, 0.124, 0.1351, 0.1149, 0.1021, 0.0824, 0.0725, 0.0474, 0.0634, 0.0713, 0.0115, 0.0247, 0.0264, 0.0675, 0.07, 0.07, 0.0688, 0.082, 0.1199, 0.1326, 0.1495, 0.2096, 0.1804, 0.1569, 0.1277, 0.0881, 0.091, 0.1005, 0.0651, 0.0828, 0.077, 0.0807, 0.0865, 0.0836, 0.0836, 0.0836, 0.0881, 0.0964]}
{"unit_id": "orig:1927328220350669026", "ticker_idx": 0, "ts": "2025-05-27T11:37:10+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe buy-side expectations for June quarter guidance may fall short of consensus by close to $7 billion… Margins are likely to be significantly impacted due to necessary write-downs", "tweet_type": "analysis", "timeline": "weeks", "summary": "JPM preview flags NVDA June-quarter guidance ~$7B below consensus and gross margin risk from H20 export-ban write-downs; bearish into print.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan NVDA Preview Summary\n\n• Positioning Score (1 = fully short/underweight, 10 = fully long/overweight): 7\n\n• Implied Move: 6.8% (as of May 23, 2025)\n\n⸻\n\nMarket Sentiment:\n\nAs seen in our somewhat frustrating buy-side survey (already shared with participants), interpreting this quarter’s NVDA guidance is extremely challenging.\n\nAssuming all else equal, we estimate that the export ban on H20 as of April 9 will remove approximately $15–16 billion from Nvidia’s FY revenue.\n(Using AMD as a benchmark, this implies ~$7.6 billion impact for each of the June and September quarters.)\n\nWhile most sell-side analysts haven’t fully reflected this issue in their forecasts, we had expected the buy-side to have priced it in to some extent.\nHowever, based on actual survey results (with high participation), many investors still hadn’t accounted for this impact.\n\nIn fact, we believe buy-side expectations for June quarter guidance may fall short of consensus by close to $7 billion, yet the average shortfall from the survey came in at only $920 million.\n\n⸻\n\nAnother Key Concern: April Quarter Margin Guidance\n\nMargins are likely to be significantly impacted due to necessary write-downs.\n\n• One-third of survey respondents expect F1Q gross margin (GM) to be around 58% (vs. current sell-side consensus of 71.1%).\n\n• The remaining two-thirds of responses were concentrated in the 65–72% range.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031070050974117636, "ret_signed_m1d": 0.031070050974117636, "alpha_spy_m1d": 0.010702831348300745, "alpha_c_m1d": 0.0003316124021961464, "bench_spy_m1d": -0.02036721962581689, "bench_c_m1d": -0.03073843857192149, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005092117794083917, "ret_signed_p1d": 0.005092117794083917, "alpha_spy_p1d": -0.0006932966829971132, "alpha_c_p1d": -0.005562224713679842, "bench_spy_p1d": -0.00578541447708103, "bench_c_p1d": -0.010654342507763759, "ret_p1w": 0.04221407101909369, "ret_signed_p1w": -0.04221407101909369, "alpha_spy_p1w": -0.03385754687530862, "alpha_c_p1w": -0.02662032188993968, "bench_spy_p1w": 0.008356524143785071, "bench_c_p1w": 0.015593749129154011, "ret_p1m": 0.13889841614726683, "ret_signed_p1m": -0.13889841614726683, "alpha_spy_p1m": -0.10884724041286109, "alpha_c_p1m": -0.015495539690978166, "bench_spy_p1m": 0.030051175734405744, "bench_c_p1m": 0.12340287645628867, "ret_p3m": 0.3136707616043415, "ret_signed_p3m": -0.3136707616043415, "alpha_spy_p3m": -0.2188256380657594, "alpha_c_p3m": -0.11499331017395487, "bench_spy_p3m": 0.0948451235385821, "bench_c_p3m": 0.19867745143038662, "ret_p6m": 0.3767046424124256, "ret_signed_p6m": -0.3767046424124256, "alpha_spy_p6m": -0.24935722555262974, "alpha_c_p6m": 0.008934470055188326, "bench_spy_p6m": 0.1273474168597959, "bench_c_p6m": 0.38563911246761395, "price_path": [-0.1134, -0.0782, -0.1583, -0.1441, -0.1344, -0.1841, -0.1684, -0.2106, -0.1974, -0.1458, -0.147, -0.1021, -0.1179, -0.1481, -0.1327, -0.1252, -0.1314, -0.104, -0.1093, -0.1604, -0.1776, -0.1906, -0.2001, -0.1871, -0.1851, -0.2487, -0.304, -0.2794, -0.2893, -0.1562, -0.2061, -0.1813, -0.183, -0.172, -0.2289, -0.251, -0.251, -0.2848, -0.2702, -0.242, -0.2145, -0.1807, -0.1976, -0.1954, -0.1962, -0.1763, -0.155, -0.16, -0.1621, -0.1361, -0.1338, -0.1391, -0.0923, -0.0411, -0.0012, -0.0049, -0.0007, 0.0005, -0.0083, -0.0273, -0.0197, -0.0311, -0.0311, 0.0, -0.0051, 0.0272, -0.0027, 0.0139, 0.0422, 0.0474, 0.0331, 0.0459, 0.0526, 0.0624, 0.0542, 0.0702, 0.0478, 0.0679, 0.0637, 0.0737, 0.0737, 0.0617, 0.0641, 0.0916, 0.1389, 0.1441, 0.1643, 0.1661, 0.1314, 0.1606, 0.176, 0.176, 0.1679, 0.1809, 0.2022, 0.2112, 0.2172, 0.2109, 0.2599, 0.2648, 0.2768, 0.2725, 0.2649, 0.2328, 0.2605, 0.2823, 0.2805, 0.3045, 0.2954, 0.3231, 0.3128, 0.2822, 0.3285, 0.3157, 0.3242, 0.3342, 0.3484, 0.3437, 0.3518, 0.3402, 0.3434, 0.3318, 0.3433, 0.2963, 0.2946, 0.2915, 0.3137, 0.3271, 0.3416, 0.3403, 0.3298, 0.2856, 0.2856, 0.2605, 0.2593, 0.267, 0.2327, 0.2422, 0.2603, 0.3088, 0.3077, 0.3125, 0.312, 0.2908, 0.2569, 0.3008, 0.304, 0.3552, 0.317, 0.3062, 0.3115, 0.3152, 0.3422, 0.3771, 0.382, 0.3942, 0.3848, 0.3695, 0.3658, 0.3958, 0.4214, 0.3519, 0.39, 0.3288, 0.3273, 0.3419, 0.3523, 0.3481, 0.3371, 0.3306, 0.3445, 0.3748, 0.4134, 0.4838, 0.5282, 0.4975, 0.4946, 0.527, 0.4665, 0.4408, 0.3882, 0.3887, 0.4692, 0.4257, 0.4304, 0.3792, 0.4036, 0.3773, 0.3386, 0.3767]}
{"unit_id": "quote:1930407504951009697", "ticker_idx": 0, "ts": "2025-06-04T23:33:09+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EDA-related delays in China's capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix", "tweet_type": "analysis", "timeline": "quarters", "summary": "EDA export controls could delay CXMT yield ramp, sustaining tight DDR5 supply and indirectly benefiting SK Hynix.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"If EDA tools face similar restrictions as equipment exports, Chinese memory manufacturers could struggle to improve DDR5 yields.  For instance, ChangXin Memory Technologies (CXMT) is reportedly operating a 10Z4nm process comparable to global 1y/1znm nodes. However, intensified EDA controls could delay yield ramp-up in DDR5, sustaining a tight supply-demand environment for these products.  In the short term, DDR4 demand remains supported, while EDA-related delays in China’s capacity expansion—CXMT plans to ramp up to 300K wafers/month by 4Q25—could indirectly benefit global memory suppliers like SK Hynix.\"", "reply_text": null, "parent_text": null, "quoted_text": "EDA Export Controls: Short-Term Disruption to China’s Fabless Sector, Long-Term Catalyst for Localization and Onshore Substitution\n\nJP Morgan (G. Hariharan, 2025/06/02)\n\nThe U.S. Bureau of Industry and Security (BIS) has imposed stricter licensing requirements on exports of EDA", "parent_author": null, "ret_m1d": -0.04597709938807859, "ret_signed_m1d": -0.04597709938807859, "alpha_spy_m1d": -0.0462457189281249, "alpha_c_m1d": -0.03421939259841866, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032183782009547635, "ret_signed_p1d": 0.032183782009547635, "alpha_spy_p1d": 0.03701655106573687, "alpha_c_p1d": 0.033971286622707875, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.1034481489964525, "ret_signed_p1w": 0.1034481489964525, "alpha_spy_p1w": 0.09433621783432744, "alpha_c_p1w": 0.06320983936782176, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.2804596714416854, "ret_signed_p1m": 0.2804596714416854, "alpha_spy_p1m": 0.2280059308757425, "alpha_c_p1m": 0.15386579796174837, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.17871148439176676, "ret_signed_p3m": 0.17871148439176676, "alpha_spy_p3m": 0.09308566509181015, "alpha_c_p3m": 0.025623022977242726, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 1.506555764559058, "ret_signed_p6m": 1.506555764559058, "alpha_spy_p6m": 1.3594758443928268, "alpha_c_p6m": 1.125384024885393, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.117, -0.1376, -0.1381, -0.0872, -0.0835, -0.0615, -0.0546, -0.0683, -0.0569, -0.0362, -0.011, -0.0293, -0.0454, -0.0179, -0.05, -0.0853, -0.1248, -0.0959, -0.0918, -0.1069, -0.1638, -0.2437, -0.2221, -0.2427, -0.1592, -0.1702, -0.173, -0.1712, -0.2014, -0.1969, -0.1969, -0.1895, -0.2024, -0.1693, -0.1817, -0.1537, -0.1647, -0.1702, -0.1854, -0.1854, -0.1464, -0.1464, -0.1464, -0.1243, -0.1266, -0.1276, -0.1051, -0.089, -0.0546, -0.0798, -0.0615, -0.0849, -0.0729, -0.0798, -0.0963, -0.0821, -0.0683, -0.0706, -0.0454, -0.0253, -0.0598, -0.046, -0.046, 0.0, 0.0322, 0.0322, 0.0529, 0.0598, 0.1034, 0.0828, 0.0828, 0.1402, 0.1448, 0.1333, 0.131, 0.1816, 0.1931, 0.2805, 0.3149, 0.3471, 0.3057, 0.3425, 0.3126, 0.2828, 0.2805, 0.2437, 0.246, 0.2966, 0.292, 0.3655, 0.354, 0.3793, 0.3724, 0.3609, 0.2391, 0.2368, 0.2529, 0.2345, 0.2368, 0.2391, 0.223, 0.2046, 0.2069, 0.2115, 0.2575, 0.1862, 0.1862, 0.2115, 0.1885, 0.2046, 0.1793, 0.2276, 0.2368, 0.2782, 0.2713, 0.2713, 0.2299, 0.2092, 0.1747, 0.1264, 0.154, 0.1931, 0.2023, 0.1954, 0.2363, 0.2386, 0.1787, 0.1994, 0.2086, 0.2225, 0.2593, 0.2754, 0.3261, 0.3997, 0.4135, 0.5125, 0.524, 0.6023, 0.5355, 0.6368, 0.6368, 0.6161, 0.6622, 0.6461, 0.6414, 0.5494, 0.6069, 0.6, 0.6576, 0.821, 0.821, 0.821, 0.821, 0.821, 0.821, 0.9707, 0.9108, 0.8947, 0.9453, 1.0835, 1.1433, 1.2354, 1.2055, 1.217, 1.2032, 1.3482, 1.4633, 1.3989, 1.5692, 1.6153, 1.5738, 1.8547, 1.6981, 1.6659, 1.7304, 1.6705, 1.7902, 1.8501, 1.8409, 1.8179, 1.5784, 1.7902, 1.6245, 1.5876, 1.6291, 1.3989, 1.3943, 1.3897, 1.4127, 1.5066]}
{"unit_id": "orig:1932209341224071255", "ticker_idx": 0, "ts": "2025-06-09T22:53:00+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our existing view that the pressure to de-rate the entire Apple value chain is likely to continue", "tweet_type": "analysis", "timeline": "months", "summary": "Bearish AAPL: WWDC disappointed on AI, iPhone 17 likely to underperform, China share erosion, and de-rating pressure on entire Apple value chain expected to continue.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities Electronics/IT Components Analyst Yang Seung-soo]\n\n▶ Apple WWDC 2025 Review: An Empty Promise\n\nApple held its annual event, 'WWDC 2025,' on the 10th (KST), announcing content focused on OS reorganization rather than AI.\n\nThis is the first time in 12 years since 2013 that Apple has fundamentally redesigned the visual composition of its OS, and the \"Liquid Glass\" unveiled on this day, true to Apple's style, showed high design completeness.\n\nHowever, at a time when new AI-related strategies needed to be revealed, we believe this announcement was a somewhat disappointing strategic judgment.\n\nOf course, some AI-related features were disclosed, but they merely followed existing AI functionalities on the market rather than demonstrating differentiated innovation.\n\nCompared to Google's annual developer event, 'I/O 2025,' last month, Apple's relative inferiority in both technical impact and direction became even more apparent.\n\nOverall, this WWDC served to further highlight market concerns about Apple's lack of AI competitiveness, and we judge that Apple is highly likely to gradually fall behind compared to major competitors who are concentrating billions of dollars in AI investments in the mid to long term.\n\nWe also see a high possibility of the iPhone 17, scheduled to be released this year, failing to achieve success.\n\nThis is due to: 1) the rapid improvement in the hardware completeness of Chinese smartphone manufacturers, which is likely to lead to a continued structural decline in Apple's market share in China, and\n\n2. the high possibility of expanded demand deferral based on consumer expectations, given that comprehensive changes across both software and hardware are anticipated from 2026, including full-scale Siri updates, foldable iPhones, and full-screen display adoption.\n\nIt is true that the prolonged sluggishness in Apple's stock price and its entire value chain since the beginning of the year has highlighted some valuation appeal.\n\nHowever, at this point, where fundamental questions about Apple's AI competitiveness and uncertainties regarding demand recovery have not been resolved, we maintain our existing view that the pressure to de-rate the entire Apple value chain is likely to continue.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01226100962740917, "ret_signed_m1d": -0.01226100962740917, "alpha_spy_m1d": -0.013161480936171643, "alpha_c_m1d": -0.017145548873212624, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.004884539245803454, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00605594241060059, "ret_signed_p1d": -0.00605594241060059, "alpha_spy_p1d": -0.00038627774170985063, "alpha_c_p1d": -0.00046166790732393714, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005594274503276653, "ret_p1w": -0.015040997154461389, "ret_signed_p1w": 0.015040997154461389, "alpha_spy_p1w": 0.020043672728865758, "alpha_c_p1w": 0.029694678958964937, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.014653681804503549, "ret_p1m": 0.04249182456139544, "ret_signed_p1m": -0.04249182456139544, "alpha_spy_p1m": -0.004981897964168036, "alpha_c_p1m": 0.029251915779288984, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.07174374034068443, "ret_p3m": 0.19162134557723287, "ret_signed_p3m": -0.19162134557723287, "alpha_spy_p3m": -0.10597743395539827, "alpha_c_p3m": -0.09332339681975976, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.09829794875747311, "ret_p6m": 0.42364121712305636, "ret_signed_p6m": -0.42364121712305636, "alpha_spy_p6m": -0.28063171373807894, "alpha_c_p6m": -0.21230277674546283, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.21133844037759353, "price_path": [0.0757, 0.0395, 0.0584, 0.0609, 0.0544, 0.0671, 0.0614, 0.0821, 0.0943, 0.1092, 0.0982, 0.1097, 0.0802, 0.1012, 0.1065, 0.1099, 0.0073, -0.0661, -0.1004, -0.1452, -0.0142, -0.056, -0.0177, 0.004, 0.0021, -0.0369, -0.0235, -0.0235, -0.0424, -0.0098, 0.0143, 0.033, 0.0375, 0.0418, 0.0471, 0.0535, 0.0575, 0.018, -0.014, -0.0159, -0.0271, -0.0209, -0.0158, 0.0464, 0.057, 0.054, 0.0496, 0.0487, 0.0364, 0.0269, 0.0032, -0.0004, -0.0307, -0.0307, -0.0062, -0.0051, -0.0074, -0.003, 0.0012, 0.009, 0.0068, -0.0041, 0.0123, 0.0, 0.0061, -0.0133, -0.0112, -0.0248, -0.015, -0.0288, -0.0242, -0.0242, -0.0022, 0.0002, -0.0057, 0.0005, -0.0022, -0.0018, 0.0185, 0.0316, 0.0546, 0.0601, 0.0601, 0.0422, 0.0425, 0.0481, 0.0544, 0.0482, 0.0356, 0.038, 0.0432, 0.0425, 0.0483, 0.0548, 0.0643, 0.063, 0.0611, 0.0617, 0.0625, 0.0487, 0.0377, 0.0304, 0.0046, 0.0094, 0.0073, 0.0586, 0.0922, 0.1385, 0.129, 0.1413, 0.1596, 0.1568, 0.1509, 0.1474, 0.1458, 0.1232, 0.1177, 0.1319, 0.1289, 0.1396, 0.1455, 0.1557, 0.1537, 0.1537, 0.1416, 0.1851, 0.1916, 0.1912, 0.1822, 0.1646, 0.1271, 0.1432, 0.1632, 0.1763, 0.1835, 0.1877, 0.1822, 0.22, 0.2726, 0.2644, 0.2539, 0.2766, 0.2695, 0.2644, 0.2654, 0.2695, 0.2778, 0.2823, 0.2757, 0.2746, 0.2825, 0.2625, 0.2189, 0.2308, 0.2313, 0.2391, 0.2297, 0.2538, 0.3032, 0.3059, 0.2844, 0.29, 0.3061, 0.3359, 0.3368, 0.3403, 0.3488, 0.3436, 0.3371, 0.342, 0.3425, 0.3407, 0.3342, 0.3403, 0.3692, 0.3604, 0.3578, 0.3551, 0.3305, 0.3304, 0.3359, 0.3245, 0.3505, 0.3726, 0.3778, 0.3807, 0.3807, 0.3871, 0.4083, 0.4236]}
{"unit_id": "orig:1936210658628190434", "ticker_idx": 0, "ts": "2025-06-20T23:52:48+00:00", "symbol": "ANET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Arista maintains a leading share in the AI Ethernet front-end... some competitors' gains from short-term client deployments are not sustainable", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares JPM note reiterating Arista's leadership in AI Ethernet; competitor gains seen as unsustainable, share stable or rising.", "resolved_tickers": ["ANET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Arista: AI Ethernet Network Landscape Stabilizes, Market Share Remains Leading but Faces Short-Term Competitive Volatility\nJPM (S. Chatterjee, 06/19/25)\n\nThe firm notes that despite market doubts about the competitive progress of Nvidia and Celestica, Arista maintains a leading share in the AI Ethernet front-end. The overall market structure is stable with some changes, and some competitors' gains from short-term client deployments are not sustainable.\n\n- Arista Maintains Leadership in AI Ethernet Network Market, Forms Triad with Nvidia and Celestica\n\nIn the 2024 AI Ethernet network market, Arista held a 26% market share, Celestica 22%, and Nvidia 18%; Q1 2025 data shows this structure is largely stable. Cisco's market share is 5%. The firm points out that although Nvidia achieved strong quarterly revenue growth with Spectrum-X, its share growth mainly stemmed from the cannibalization of InfiniBand demand and concentrated short-term client purchases.\n\n- AI Switching Market Shows Strong Growth, Arista's Share Stable or Slightly Increasing\n\nThe AI Switching (Ethernet + InfiniBand) market is expected to maintain a 35% CAGR until 2029, with growth rates potentially exceeding 60% in 2025 and 2026. Against this backdrop, Arista's total AI Switching share rose from 15% in 2023 to 18% in Q1 2025, outperforming market expectations.\nArista Leads in AI Front-End, Still in Early Stages of Client Adoption in Back-EndIn the AI front-end, Arista ranks first with a 38% share, while Celestica and Nvidia hold 16% and 13% respectively. In the AI back-end market, Celestica leads with a 28% share due to its deployment of 800G white-box switches at Amazon, Google, and Meta, whereas Arista's current back-end deployments at Tier 1 clients like Meta are still in early stages.\n\n- Nvidia's Overall AI Switching Share Declines, Partly Due to InfiniBand Weakness\n\nDespite Nvidia's strong performance in Ethernet products, its InfiniBand business has seen two consecutive quarters of year-over-year decline. Its overall AI Switching share dropped from 59% in 2023 to 46% in Q1 2025. The firm believes this indicates that Spectrum-X's growth partly comes from an internal migration from InfiniBand.\n\n- Spectrum-X Revenue Structure is Complex, Short-Term Growth Potentially Driven by NICs and xAI Clients\n\nSpectrum-X revenue is not purely switching revenue; it includes 55% Switching, 22% optical modules, 19% NICs, and 4% copper cabling. Recent growth is believed to stem from concentrated deployments by large client xAI, where a single client might contribute $400 million to $600 million in AI networking revenue, driving Nvidia's short-term revenue performance and creating competitive pressure for Arista.\n\n$ANET", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04626084479732784, "ret_signed_m1d": 0.04626084479732784, "alpha_spy_m1d": 0.043906676555449886, "alpha_c_m1d": 0.041812523414226144, "bench_spy_m1d": 0.002354168241877952, "bench_c_m1d": 0.004448321383101694, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0660869211390398, "ret_signed_p1d": 0.0660869211390398, "alpha_spy_p1d": 0.05620944063454414, "alpha_c_p1d": 0.054685246617978356, "bench_spy_p1d": 0.009877480504495662, "bench_c_p1d": 0.011401674521061445, "ret_p1w": 0.1523478190104166, "ret_signed_p1w": 0.1523478190104166, "alpha_spy_p1w": 0.11763365122427105, "alpha_c_p1w": 0.1080905779917194, "bench_spy_p1w": 0.034714167786145556, "bench_c_p1w": 0.04425724101869721, "ret_p1m": 0.29402899258378623, "ret_signed_p1m": 0.29402899258378623, "alpha_spy_p1m": 0.23599241177778207, "alpha_c_p1m": 0.20621424792796028, "bench_spy_p1m": 0.05803658080600416, "bench_c_p1m": 0.08781474465582595, "ret_p3m": 0.6561158995697465, "ret_signed_p3m": 0.6561158995697465, "alpha_spy_p3m": 0.546908138448065, "alpha_c_p3m": 0.5258264323526469, "bench_spy_p3m": 0.10920776112168151, "bench_c_p3m": 0.13028946721709955, "ret_p6m": 0.4595941958220109, "ret_signed_p6m": 0.4595941958220109, "alpha_spy_p6m": 0.31094839410079267, "alpha_c_p6m": 0.2729388665001309, "bench_spy_p6m": 0.14864580172121822, "bench_c_p6m": 0.18665532932188, "price_path": [0.008, -0.0532, -0.0845, -0.0963, -0.1017, -0.09, -0.0706, -0.1737, -0.2537, -0.2096, -0.1942, -0.1188, -0.1577, -0.1574, -0.1468, -0.1513, -0.1667, -0.1745, -0.1745, -0.2154, -0.2038, -0.1796, -0.1271, -0.0967, -0.0921, -0.064, -0.0461, 0.0184, 0.0553, 0.0479, 0.0524, 0.0023, 0.0118, 0.0031, 0.0704, 0.1275, 0.1388, 0.1139, 0.1179, 0.1212, 0.1092, 0.0717, 0.0732, 0.0574, 0.0574, 0.0723, 0.0758, 0.0014, 0.0045, 0.0409, 0.0957, 0.1008, 0.1035, 0.1275, 0.1223, 0.0864, 0.0923, 0.1104, 0.0707, 0.1025, 0.0398, 0.0463, 0.0463, 0.0, 0.0661, 0.1011, 0.1166, 0.1779, 0.1523, 0.1862, 0.1468, 0.1725, 0.1886, 0.1886, 0.1765, 0.1987, 0.2322, 0.2323, 0.2588, 0.2565, 0.2449, 0.2557, 0.2983, 0.296, 0.294, 0.2728, 0.3106, 0.3222, 0.325, 0.3629, 0.3753, 0.4155, 0.4286, 0.3631, 0.3954, 0.3695, 0.609, 0.6148, 0.6137, 0.5959, 0.6377, 0.6001, 0.5824, 0.5919, 0.6005, 0.5395, 0.5243, 0.5308, 0.5449, 0.5425, 0.5568, 0.5452, 0.5795, 0.5832, 0.5832, 0.5753, 0.5928, 0.6368, 0.6562, 0.6233, 0.6453, 0.7475, 0.7744, 0.6161, 0.6861, 0.6482, 0.6561, 0.7004, 0.7346, 0.6858, 0.6706, 0.6538, 0.6587, 0.6522, 0.6623, 0.6894, 0.7307, 0.6749, 0.687, 0.7333, 0.6845, 0.8245, 0.8346, 0.7867, 0.7096, 0.6092, 0.6624, 0.6929, 0.6591, 0.6983, 0.6921, 0.6996, 0.7711, 0.7834, 0.8181, 0.8176, 0.8786, 0.837, 0.8283, 0.8271, 0.7803, 0.6281, 0.5539, 0.5612, 0.5914, 0.5644, 0.565, 0.5107, 0.5231, 0.4755, 0.4313, 0.4471, 0.3866, 0.3615, 0.4165, 0.4497, 0.48, 0.48, 0.5151, 0.4853, 0.475, 0.4817, 0.4904, 0.4909, 0.4969, 0.5077, 0.5346, 0.5581, 0.4465, 0.4596]}
{"unit_id": "orig:1940193710635196697", "ticker_idx": 0, "ts": "2025-07-01T23:40:02+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price upgraded to KRW 71,000 (applying P/B 0.94x based on FY26E), current stock price is at 0.8x, more than 30% discounted compared to 10-year average", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares JPMorgan's upgraded PT of KRW 71,000 on Samsung Electronics, noting stock trades at 30%+ discount to 10-year average P/B.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electronics Preview\n\n1. Goldman Sachs (GS)\n\n1) Q2 2025 Operating Profit Outlook Downgraded (KRW 6.6 Trillion → KRW 5.5 Trillion)\n\n- Primary reasons: Decrease in HBM shipments, expanded foundry losses, and foreign exchange losses due to stronger won.\n\n- Strong volume and prices in general DRAM partially offset these factors.\n\n2) HBM Related Issues\n\n- Delay in HBM3E 12Hi certification by a major GPU customer → Shipment timing pushed back by more than one quarter.\n- Q2 2025 HBM Revenue estimate: US1.1billion(downgraded from previous US1.6 billion).\n- 2025 HBM Revenue estimate downgraded by 20% → Around US$8 billion for the year (+16% YoY).\n- HBM3E and HBM4 supply expected to fully commence in 2026-27 → 2026 HBM shipments +63% YoY.\n\n3) General DRAM and NAND Outlook\n\n- DRAM price outlook upgraded: Expecting DDR5 spillover effect → ASP improvement in H2.\n- Q2 2025 DRAM ASP: Slightly flat, recovery from Q3 2025.\n- NAND similar pattern: Weak in Q2 2025, eSSD demand recovery leading from Q3 2025.\n\n4) Foundry and System LSI\n- Q2 2025 Operating Profit (OP): KRW -2.4 trillion (expanded loss from previous KRW -1.7 trillion).\n- Consecutive 2 trillion won level losses for three quarters.\n- Meaningful utilization rate recovery difficult without securing new customers.\n\n5) Target Price: KRW 74,000 (unchanged)\n\n2. JPMorgan (JPM)\n\n1) HBM Business Progress but Technology Gap Remains\n\n- HBM3E 12Hi NVIDIA certification delay led to a downgrade of 2025 HBM bit growth guidance from 50% to 12%.\n- ASIC demand from AMD, AVGO, etc., expected to contribute meaningfully from Q4 2025.\n- 1cnm technology has passed PRA and is entering risk production in Q4 2025, but mass production (MP) is more realistic by late Q1 2026.\n- Recovery of technology leadership will take time (challenges include yield, lack of customer track record, and competitors' pre-contracts).\n\n2) Foundry Division Showing Progress but Profit Improvement Not Yet\n\n- Secured new orders for Exynos 2500 (foldable) and 8nm (Nintendo Switch 2).\n- However, due to the absence of large advanced process orders for mobile/AI (e.g., QCOM 2nm), the timing of profitability turnaround will be delayed.\n\n3) DX Division Maintains Healthy Margins Amidst Cost Pressures\n\n- Despite rising component prices and tariff uncertainties, flagship model sales and economies of scale are expected to improve margins by +50bps year-over-year.\n- New foldable product launch scheduled for July 9, consumer response in H2 is key.\n\n4) Earnings Outlook and Target Price\n\n- Q2 OP forecast: KRW 6 trillion.\n- FY25E EPS: KRW 4,922 (down 8%), FY26E EPS: KRW 5,939 (up 1%), FY27E EPS: KRW 6,759 (newly introduced).\n- Target price upgraded to KRW 71,000 (applying P/B 0.94x based on FY26E), current stock price is at 0.8x, more than 30% discounted compared to 10-year average.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0066444899273386815, "ret_signed_m1d": -0.0066444899273386815, "alpha_spy_m1d": -0.006968274037780442, "alpha_c_m1d": -0.0156495617739717, "bench_spy_m1d": 0.0003237841104417605, "bench_c_m1d": 0.009005071846633017, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009966800520785357, "ret_signed_p1d": 0.009966800520785357, "alpha_spy_p1d": 0.005433524004969437, "alpha_c_p1d": -0.000592246602281854, "bench_spy_p1d": 0.0045332765158159205, "bench_c_p1d": 0.010559047123067211, "ret_p1w": 0.019933601041570714, "ret_signed_p1w": 0.019933601041570714, "alpha_spy_p1w": 0.01557841076932509, "alpha_c_p1w": -0.0012642186803266675, "bench_spy_p1w": 0.004355190272245624, "bench_c_p1w": 0.021197819721897382, "ret_p1m": 0.20598010656438226, "ret_signed_p1m": 0.20598010656438226, "alpha_spy_p1m": 0.17876410379632213, "alpha_c_p1m": 0.1513519785985229, "bench_spy_p1m": 0.027216002768060132, "bench_c_p1m": 0.05462812796585936, "ret_p3m": 0.38372090123335423, "ret_signed_p3m": 0.38372090123335423, "alpha_spy_p3m": 0.3092372018548921, "alpha_c_p3m": 0.27124382154456694, "bench_spy_p3m": 0.07448369937846211, "bench_c_p3m": 0.11247707968878728, "ret_p6m": 0.8537488975838134, "ret_signed_p6m": 0.8537488975838134, "alpha_spy_p6m": 0.7295853149549854, "alpha_c_p6m": 0.6846033648323622, "bench_spy_p6m": 0.1241635826288281, "bench_c_p6m": 0.16914553275145128, "price_path": [-0.049, -0.0738, -0.1217, -0.1167, -0.125, -0.0688, -0.0886, -0.0721, -0.0655, -0.0969, -0.0903, -0.087, -0.0853, -0.0919, -0.0804, -0.0804, -0.0804, -0.0787, -0.0787, -0.0837, -0.0837, -0.1035, -0.1035, -0.1035, -0.0986, -0.0986, -0.0953, -0.049, -0.0606, -0.0523, -0.054, -0.0622, -0.0787, -0.0771, -0.0804, -0.0969, -0.1052, -0.0969, -0.1101, -0.0771, -0.0738, -0.0721, -0.0622, -0.0622, -0.0457, -0.0243, -0.0243, -0.0127, -0.0226, -0.011, -0.0177, -0.0375, -0.0556, -0.0408, -0.0127, -0.0226, -0.0177, -0.0424, -0.0011, 0.0121, -0.0061, 0.01, -0.0066, 0.0, 0.01, 0.0598, 0.0515, 0.0249, 0.0199, 0.0033, 0.0133, 0.0399, 0.0382, 0.0581, 0.0748, 0.108, 0.1146, 0.1262, 0.0963, 0.103, 0.0963, 0.0947, 0.1694, 0.1728, 0.206, 0.186, 0.1445, 0.1578, 0.1611, 0.1429, 0.1711, 0.1927, 0.1794, 0.1811, 0.1944, 0.1894, 0.1894, 0.1628, 0.1628, 0.1711, 0.1728, 0.186, 0.1877, 0.1678, 0.1728, 0.1561, 0.1578, 0.1229, 0.1478, 0.1595, 0.1645, 0.1545, 0.1645, 0.1877, 0.206, 0.2193, 0.2525, 0.2708, 0.3189, 0.299, 0.3339, 0.3339, 0.387, 0.407, 0.4186, 0.4302, 0.3837, 0.4049, 0.3999, 0.4349, 0.4975, 0.4975, 0.4975, 0.4975, 0.4975, 0.4975, 0.5751, 0.5567, 0.5284, 0.5851, 0.6302, 0.6335, 0.6368, 0.6268, 0.6452, 0.6101, 0.6485, 0.7019, 0.6602, 0.6769, 0.737, 0.7937, 0.8537, 0.7503, 0.6786, 0.6552, 0.6335, 0.6786, 0.7269, 0.7203, 0.7153, 0.6218, 0.6786, 0.6318, 0.6101, 0.6786, 0.5818, 0.6135, 0.6569, 0.7153, 0.7269, 0.6769, 0.6819, 0.7253, 0.7436, 0.7536, 0.8087, 0.8271, 0.8087, 0.802, 0.7903, 0.817, 0.7486, 0.7153, 0.8004, 0.7953, 0.7737, 0.8437, 0.8604, 0.8537]}
{"unit_id": "orig:1933341111109759135", "ticker_idx": 0, "ts": "2025-06-13T01:50:15+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains a \"Buy\" rating on Besi with a 12-month price target of €146", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Sachs Buy on Besi, PT €146, on raised long-term guidance 20-30% above consensus driven by AI packaging demand.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Besi: Long-Term Guidance Raised, Growth Driven by AI and Next-Gen Packaging\nGoldman Sachs (A. Duval, 12/06/25)\n\nGoldman Sachs notes that Besi, during its 2025 Investor Day, raised its long-term revenue target from €1.0 billion to €1.5–1.9 billion, gross margin from 62–66% to 64–68%, and operating profit margin from 35–50% to 40–55%. This reflects its strong confidence in packaging demand from AI data centers, edge computing, and consumer electronics, and highlights that the accelerated adoption of advanced packaging technologies such as 2.5D/3D chipsets and wafer-level bonding will be the primary growth driver.\n \nLong-Term Metrics Exceed Expectations\n   \nThe new guidance midpoint revenue of €1.7 billion is about 20% higher than the Visible Alpha market consensus of €1.4 billion. The implied profit, calculated at a 48% EBIT rate, is also approximately 30% ahead of consensus, suggesting there is upside potential in the market's valuation of Besi's prospects.\n\nDual Drivers of AI Deployment and Advanced Packaging\n   \nBesi's management emphasized that the continued proliferation of AI in data centers and consumer applications, along with strong demand for next-generation packaging for logic/memory, will bring sustained incremental business for its wafer-level packaging and mainstream die attach operations.\n \nTopics of Investor Focus\n   \nGoldman Sachs believes that investors will focus on the following four points: 1) Demand drivers in the logic and memory markets; 2) The adoption pace of technologies such as Hybrid Bonding and TCB (Thermo-Compression Bonding); 3) Changes in the industry's competitive landscape; 4) The recovery timeline for the mainstream die attach business.\n \nValuation and Risks\n   \nGoldman Sachs maintains a \"Buy\" rating on Besi with a 12-month price target of €146, based on a 27x CY26E EV/EBITDA multiple. Key risks include cyclical fluctuations in customer spending, delays in the rollout of hybrid bonding technology, and increased industry competition.\n\n$BESI", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025019846228167775, "ret_signed_m1d": -0.025019846228167775, "alpha_spy_m1d": -0.03632631770382988, "alpha_c_m1d": -0.04898968229342915, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.023969836065261374, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00873709690264679, "ret_signed_p1d": 0.00873709690264679, "alpha_spy_p1d": -0.0007770995760685295, "alpha_c_p1d": -0.01612780363370092, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.02486490053634771, "ret_p1w": -0.021842681275734743, "ret_signed_p1w": -0.021842681275734743, "alpha_spy_p1w": -0.02022940627992542, "alpha_c_p1w": -0.03460586703818225, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.012763185762447504, "ret_p1m": -0.009928480401537154, "ret_signed_p1m": -0.009928480401537154, "alpha_spy_p1m": -0.059605272167247536, "alpha_c_p1m": -0.12016637886544257, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.11023789846390541, "ret_p3m": -0.10762509831642764, "ret_signed_p3m": -0.10762509831642764, "alpha_spy_p3m": -0.20333384473629834, "alpha_c_p3m": -0.27821540488970775, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.17059030657328011, "ret_p6m": 0.16640194865629243, "ret_signed_p6m": 0.16640194865629243, "alpha_spy_p6m": 0.014723696970295519, "alpha_c_p6m": -0.2677006271138811, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.4341025757701735, "price_path": [-0.1757, -0.1636, -0.1679, -0.1924, -0.178, -0.1594, -0.1858, -0.2157, -0.2351, -0.2609, -0.2404, -0.2432, -0.3022, -0.3104, -0.3413, -0.3337, -0.382, -0.3587, -0.3542, -0.3329, -0.2441, -0.2684, -0.2705, -0.2705, -0.2705, -0.2767, -0.2707, -0.2569, -0.2292, -0.2238, -0.2168, -0.2515, -0.2515, -0.2081, -0.2049, -0.2149, -0.1739, -0.1573, -0.143, -0.0866, -0.0572, -0.0457, -0.0826, -0.1001, -0.1195, -0.1382, -0.1354, -0.1414, -0.1636, -0.1461, -0.1307, -0.135, -0.1223, -0.1537, -0.1743, -0.1521, -0.1426, -0.1215, -0.0977, -0.0878, -0.0445, -0.0592, -0.025, 0.0, 0.0087, 0.0075, -0.0012, -0.0095, -0.0218, -0.0016, 0.0238, 0.0262, 0.031, 0.0361, 0.0091, -0.0298, -0.021, -0.0314, -0.0385, -0.0357, -0.0294, -0.0314, 0.0119, 0.0099, -0.0099, 0.0032, -0.0473, 0.0127, 0.0163, 0.0338, -0.0008, -0.0179, 0.0087, -0.0699, -0.0246, -0.0469, -0.0278, -0.054, -0.0695, -0.0596, -0.0842, -0.0882, -0.0473, -0.0409, -0.0346, -0.0171, -0.0191, -0.0258, -0.0584, -0.0643, -0.0516, -0.0699, -0.083, -0.0548, -0.0536, -0.0413, -0.0485, -0.0488, -0.0858, -0.0826, -0.1509, -0.1597, -0.1489, -0.1366, -0.1195, -0.1148, -0.1076, -0.0973, -0.1116, -0.0616, -0.0739, -0.0814, -0.0087, -0.0349, -0.0246, 0.0079, 0.021, 0.0064, -0.0278, 0.004, 0.0075, 0.0151, 0.0608, 0.0365, 0.1656, 0.166, 0.1624, 0.1549, 0.1144, 0.1338, 0.116, 0.1557, 0.1525, 0.1255, 0.1533, 0.1461, 0.1001, 0.1458, 0.168, 0.1755, 0.1644, 0.1728, 0.168, 0.1724, 0.1581, 0.1322, 0.1239, 0.0941, 0.0655, 0.0838, 0.0945, 0.0751, 0.0627, 0.0437, 0.0342, 0.0099, 0.0254, 0.033, -0.0246, -0.0119, -0.0103, 0.0294, 0.0318, 0.0318, 0.0306, 0.0433, 0.0798, 0.0921, 0.1084, 0.1664]}
{"unit_id": "orig:1921770409927549247", "ticker_idx": 0, "ts": "2025-05-12T03:32:25+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We downgrade the entire memory sector and assign an Underperform rating on Micron (MU US)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities downgrades entire memory sector to Underperform; memory downturn from 4Q25 on tariffs, weak demand, HBM pricing pressure; SK Hynix least preferred.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities – Memory Sector: Downturn Approaching, HBM Pricing Pressured\n\n– We downgrade the entire memory sector and assign an Underperform rating on Micron (MU US), SK Hynix (000660 KS), and SanDisk (SNDK US).\n\n◾+ Sector slowdown and demand weakening intensifying\n– Following a bottom in 1Q23, the memory sector is projected to post 10 consecutive quarters of improvement through 3Q25 — marking the longest upcycle since 2016.\n– However, a downturn is increasingly likely starting from 4Q25. Key reasons include:\n 1) Risk of a 25% tariff on electronics/semiconductors by the Trump administration\n 2) Deteriorating global macro conditions → smartphone shipment growth to turn negative, PC shipments limited to +1~2%\n 3) HBM pricing could also become a headwind\n\n◾+ HBM and DRAM prices: Potential weakening from 3Q25\n– As of CYQ2, memory pricing was stronger than expected (thanks to low smartphone inventories and preemptive restocking during the 90-day tariff waiver)\n– 2Q25 DRAM/NAND contract pricing: QoQ +67% expected\n– However, from 3Q25, price momentum is likely to slow due to weak demand and inventory pressure from servers and PCs:\n - Pricing increase of only around 0~5% or potentially flat\n - This flat trend may persist through 4Q25\n◾+ Samsung’s entry to intensify HBM market competition\n– Samsung is preparing to qualify HBM3e-12Hi with NVIDIA (possibly by June)\n– NVIDIA is delaying 2026 HBM3e pricing negotiations pending Samsung’s qualification\n– Samsung’s market entry in 2026 may trigger HBM4 price declines\n - Currently, HBM4 is priced at a 20% premium to HBM3e-12Hi\n - CXMT’s DDR5 production ratio also expected to reach 20% by end-2025\n◾+ Stock preference\n– 1st: Micron > 2nd: SanDisk > 3rd: SK Hynix\n - Micron: U.S.-based production → potential beneficiary of tariffs; gaining share in AMD/AWS\n - SanDisk: Focused on NAND → less exposed to HBM, relatively defensive\n - SK Hynix: Over 40% of DRAM revenue from HBM, but high exposure to HBM pricing pressure creates downside risk\n\n✅ Conclusion: The memory sector is likely to peak in 3Q25, with a potential slowdown thereafter. HBM pricing headwinds and renewed U.S. tariff risk are expected to weigh on medium- to long-term valuations. Among covered names, Micron is viewed most favorably, while SK Hynix remains relatively conservative due to its large HBM exposure.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06967173414031069, "ret_signed_m1d": 0.06967173414031069, "alpha_spy_m1d": 0.03768142296917443, "alpha_c_m1d": 0.010659852037967932, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05027631913789388, "ret_signed_p1d": -0.05027631913789388, "alpha_spy_p1d": -0.043672378820723834, "alpha_c_p1d": -0.015947559374998255, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.068913266260344, "ret_signed_p1w": -0.068913266260344, "alpha_spy_p1w": -0.048569965986598795, "alpha_c_p1w": -0.03268903333222295, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.23675368123748686, "ret_signed_p1m": -0.23675368123748686, "alpha_spy_p1m": -0.20229351078924096, "alpha_c_p1m": -0.13288271886127534, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.21329835415038878, "ret_signed_p3m": -0.21329835415038878, "alpha_spy_p3m": -0.12559695933220105, "alpha_c_p3m": 0.01293468499132544, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 1.366148981056865, "ret_signed_p6m": -1.366148981056865, "alpha_spy_p6m": -1.2012682959938907, "alpha_c_p6m": -0.8802806038739521, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [-0.0079, 0.0352, 0.0769, 0.0769, 0.1556, 0.1293, 0.1165, 0.0696, 0.0325, 0.0088, 0.0574, -0.0064, 0.0132, -0.0202, -0.0131, 0.0209, -0.034, 0.0059, -0.0577, -0.0364, 0.035, 0.0267, 0.0907, 0.1158, 0.1007, 0.1044, 0.1146, 0.025, 0.049, 0.0192, -0.003, -0.0135, -0.043, -0.0585, -0.0388, -0.04, -0.1945, -0.2987, -0.2592, -0.2898, -0.1562, -0.241, -0.2464, -0.2305, -0.2303, -0.2488, -0.2545, -0.2545, -0.2768, -0.2392, -0.2098, -0.1611, -0.1356, -0.1488, -0.167, -0.1662, -0.1573, -0.1254, -0.1286, -0.1276, -0.1048, -0.0774, -0.0697, 0.0, 0.0503, 0.0328, 0.0342, 0.0619, 0.0689, 0.063, 0.0385, 0.0275, 0.0117, 0.0117, 0.0443, 0.0421, 0.0489, 0.0235, 0.0638, 0.1079, 0.1188, 0.1517, 0.1763, 0.2022, 0.2368, 0.2572, 0.2589, 0.2526, 0.2985, 0.3039, 0.32, 0.32, 0.3393, 0.3228, 0.386, 0.3788, 0.3653, 0.3518, 0.3355, 0.3099, 0.3191, 0.3251, 0.3251, 0.3006, 0.3494, 0.3258, 0.3352, 0.3506, 0.2864, 0.3027, 0.2628, 0.2284, 0.2406, 0.228, 0.1846, 0.1912, 0.2118, 0.2067, 0.2066, 0.2143, 0.2444, 0.1837, 0.1375, 0.1688, 0.1828, 0.1798, 0.2133, 0.2894, 0.3418, 0.3855, 0.3478, 0.3588, 0.3109, 0.34, 0.3237, 0.2712, 0.2558, 0.2763, 0.2626, 0.2635, 0.2771, 0.3232, 0.2907, 0.2907, 0.285, 0.2876, 0.3471, 0.4248, 0.4258, 0.4668, 0.5184, 0.633, 0.7053, 0.7111, 0.7225, 0.7352, 0.8317, 0.7649, 0.7854, 0.8048, 0.7538, 0.7009, 0.7057, 0.7776, 0.8147, 0.9755, 0.9929, 1.0384, 1.0724, 1.0152, 1.1329, 1.0872, 0.9708, 1.092, 1.0301, 1.083, 1.1979, 1.1963, 1.244, 1.1953, 1.1539, 1.2433, 1.3769, 1.3886, 1.4083, 1.4595, 1.431, 1.4284, 1.5471, 1.3661]}
{"unit_id": "orig:1921770409927549247", "ticker_idx": 1, "ts": "2025-05-12T03:32:25+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We downgrade the entire memory sector and assign an Underperform rating on… SK Hynix (000660 KS)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities downgrades entire memory sector to Underperform; memory downturn from 4Q25 on tariffs, weak demand, HBM pricing pressure; SK Hynix least preferred.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities – Memory Sector: Downturn Approaching, HBM Pricing Pressured\n\n– We downgrade the entire memory sector and assign an Underperform rating on Micron (MU US), SK Hynix (000660 KS), and SanDisk (SNDK US).\n\n◾+ Sector slowdown and demand weakening intensifying\n– Following a bottom in 1Q23, the memory sector is projected to post 10 consecutive quarters of improvement through 3Q25 — marking the longest upcycle since 2016.\n– However, a downturn is increasingly likely starting from 4Q25. Key reasons include:\n 1) Risk of a 25% tariff on electronics/semiconductors by the Trump administration\n 2) Deteriorating global macro conditions → smartphone shipment growth to turn negative, PC shipments limited to +1~2%\n 3) HBM pricing could also become a headwind\n\n◾+ HBM and DRAM prices: Potential weakening from 3Q25\n– As of CYQ2, memory pricing was stronger than expected (thanks to low smartphone inventories and preemptive restocking during the 90-day tariff waiver)\n– 2Q25 DRAM/NAND contract pricing: QoQ +67% expected\n– However, from 3Q25, price momentum is likely to slow due to weak demand and inventory pressure from servers and PCs:\n - Pricing increase of only around 0~5% or potentially flat\n - This flat trend may persist through 4Q25\n◾+ Samsung’s entry to intensify HBM market competition\n– Samsung is preparing to qualify HBM3e-12Hi with NVIDIA (possibly by June)\n– NVIDIA is delaying 2026 HBM3e pricing negotiations pending Samsung’s qualification\n– Samsung’s market entry in 2026 may trigger HBM4 price declines\n - Currently, HBM4 is priced at a 20% premium to HBM3e-12Hi\n - CXMT’s DDR5 production ratio also expected to reach 20% by end-2025\n◾+ Stock preference\n– 1st: Micron > 2nd: SanDisk > 3rd: SK Hynix\n - Micron: U.S.-based production → potential beneficiary of tariffs; gaining share in AMD/AWS\n - SanDisk: Focused on NAND → less exposed to HBM, relatively defensive\n - SK Hynix: Over 40% of DRAM revenue from HBM, but high exposure to HBM pricing pressure creates downside risk\n\n✅ Conclusion: The memory sector is likely to peak in 3Q25, with a potential slowdown thereafter. HBM pricing headwinds and renewed U.S. tariff risk are expected to weigh on medium- to long-term valuations. Among covered names, Micron is viewed most favorably, while SK Hynix remains relatively conservative due to its large HBM exposure.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02512820760846257, "ret_signed_m1d": 0.02512820760846257, "alpha_spy_m1d": -0.006862103562673694, "alpha_c_m1d": -0.03388367449388019, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017948823359660082, "ret_signed_p1d": -0.017948823359660082, "alpha_spy_p1d": -0.01134488304249004, "alpha_c_p1d": 0.01637993640323554, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.022564136047578343, "ret_signed_p1w": -0.022564136047578343, "alpha_spy_p1w": -0.002220835773833141, "alpha_c_p1w": 0.013660096880542705, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.18418625046954373, "ret_signed_p1m": -0.18418625046954373, "alpha_spy_p1m": -0.14972608002129784, "alpha_c_p1m": -0.08031528809333222, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.3460164956963192, "ret_signed_p3m": -0.3460164956963192, "alpha_spy_p3m": -0.2583151008781315, "alpha_c_p3m": -0.11978345655460498, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 2.014904164711914, "ret_signed_p6m": -2.014904164711914, "alpha_spy_p6m": -1.8500234796489397, "alpha_c_p6m": -1.529035787529001, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0134, 0.0624, 0.07, 0.0802, 0.07, 0.1133, 0.0802, 0.0675, 0.0445, 0.0216, 0.0343, 0.0215, -0.0246, -0.0246, -0.0456, -0.0097, -0.0133, -0.0133, -0.0364, -0.0369, 0.02, 0.0241, 0.0487, 0.0564, 0.041, 0.0538, 0.0769, 0.1051, 0.0846, 0.0667, 0.0974, 0.0615, 0.0221, -0.0221, 0.0103, 0.0149, -0.0021, -0.0656, -0.1549, -0.1308, -0.1538, -0.0605, -0.0728, -0.0759, -0.0738, -0.1077, -0.1026, -0.1026, -0.0944, -0.1087, -0.0718, -0.0856, -0.0544, -0.0667, -0.0728, -0.0897, -0.0897, -0.0462, -0.0462, -0.0462, -0.0215, -0.0241, -0.0251, 0.0, 0.0179, 0.0564, 0.0282, 0.0487, 0.0226, 0.0359, 0.0282, 0.0097, 0.0256, 0.041, 0.0385, 0.0667, 0.0891, 0.0506, 0.066, 0.066, 0.1174, 0.1534, 0.1534, 0.1765, 0.1842, 0.233, 0.2099, 0.2099, 0.2741, 0.2792, 0.2664, 0.2638, 0.3203, 0.3332, 0.4308, 0.4693, 0.5053, 0.459, 0.5001, 0.4667, 0.4334, 0.4308, 0.3897, 0.3923, 0.4488, 0.4436, 0.5258, 0.513, 0.5412, 0.5335, 0.5207, 0.3845, 0.382, 0.4, 0.3794, 0.382, 0.3845, 0.3666, 0.346, 0.3486, 0.3537, 0.4051, 0.3255, 0.3255, 0.3537, 0.328, 0.346, 0.3178, 0.3717, 0.382, 0.4282, 0.4205, 0.4205, 0.3743, 0.3512, 0.3126, 0.2587, 0.2895, 0.3332, 0.3434, 0.3357, 0.3814, 0.384, 0.3171, 0.3402, 0.3505, 0.366, 0.4071, 0.4251, 0.4817, 0.564, 0.5795, 0.6901, 0.703, 0.7904, 0.7158, 0.829, 0.829, 0.8059, 0.8573, 0.8393, 0.8342, 0.7313, 0.7956, 0.7878, 0.8522, 1.0348, 1.0348, 1.0348, 1.0348, 1.0348, 1.0348, 1.202, 1.1351, 1.1171, 1.1737, 1.3281, 1.3949, 1.4978, 1.4644, 1.4773, 1.4618, 1.6239, 1.7525, 1.6805, 1.8708, 1.9223, 1.876, 2.1898, 2.0149]}
{"unit_id": "orig:1921770409927549247", "ticker_idx": 2, "ts": "2025-05-12T03:32:25+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We downgrade the entire memory sector and assign an Underperform rating on… SanDisk (SNDK US)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities downgrades entire memory sector to Underperform; memory downturn from 4Q25 on tariffs, weak demand, HBM pricing pressure; SK Hynix least preferred.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities – Memory Sector: Downturn Approaching, HBM Pricing Pressured\n\n– We downgrade the entire memory sector and assign an Underperform rating on Micron (MU US), SK Hynix (000660 KS), and SanDisk (SNDK US).\n\n◾+ Sector slowdown and demand weakening intensifying\n– Following a bottom in 1Q23, the memory sector is projected to post 10 consecutive quarters of improvement through 3Q25 — marking the longest upcycle since 2016.\n– However, a downturn is increasingly likely starting from 4Q25. Key reasons include:\n 1) Risk of a 25% tariff on electronics/semiconductors by the Trump administration\n 2) Deteriorating global macro conditions → smartphone shipment growth to turn negative, PC shipments limited to +1~2%\n 3) HBM pricing could also become a headwind\n\n◾+ HBM and DRAM prices: Potential weakening from 3Q25\n– As of CYQ2, memory pricing was stronger than expected (thanks to low smartphone inventories and preemptive restocking during the 90-day tariff waiver)\n– 2Q25 DRAM/NAND contract pricing: QoQ +67% expected\n– However, from 3Q25, price momentum is likely to slow due to weak demand and inventory pressure from servers and PCs:\n - Pricing increase of only around 0~5% or potentially flat\n - This flat trend may persist through 4Q25\n◾+ Samsung’s entry to intensify HBM market competition\n– Samsung is preparing to qualify HBM3e-12Hi with NVIDIA (possibly by June)\n– NVIDIA is delaying 2026 HBM3e pricing negotiations pending Samsung’s qualification\n– Samsung’s market entry in 2026 may trigger HBM4 price declines\n - Currently, HBM4 is priced at a 20% premium to HBM3e-12Hi\n - CXMT’s DDR5 production ratio also expected to reach 20% by end-2025\n◾+ Stock preference\n– 1st: Micron > 2nd: SanDisk > 3rd: SK Hynix\n - Micron: U.S.-based production → potential beneficiary of tariffs; gaining share in AMD/AWS\n - SanDisk: Focused on NAND → less exposed to HBM, relatively defensive\n - SK Hynix: Over 40% of DRAM revenue from HBM, but high exposure to HBM pricing pressure creates downside risk\n\n✅ Conclusion: The memory sector is likely to peak in 3Q25, with a potential slowdown thereafter. HBM pricing headwinds and renewed U.S. tariff risk are expected to weigh on medium- to long-term valuations. Among covered names, Micron is viewed most favorably, while SK Hynix remains relatively conservative due to its large HBM exposure.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08553425587938934, "ret_signed_m1d": 0.08553425587938934, "alpha_spy_m1d": 0.05354394470825308, "alpha_c_m1d": 0.04117773581108519, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.04435652006830415, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017688382218057352, "ret_signed_p1d": -0.017688382218057352, "alpha_spy_p1d": -0.01108444190088731, "alpha_c_p1d": 0.003919070286031667, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.02160745250408902, "ret_p1w": -0.060334423330455245, "ret_signed_p1w": 0.060334423330455245, "alpha_spy_p1w": 0.08067772360420045, "alpha_c_p1w": 0.09041802562301293, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.030083602292557687, "ret_p1m": 0.007753808850855526, "ret_signed_p1m": -0.007753808850855526, "alpha_spy_p1m": 0.02670636159739037, "alpha_c_p1m": 0.05008548229239129, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.05783929114324682, "ret_p3m": -0.014053836312608481, "ret_signed_p3m": 0.014053836312608481, "alpha_spy_p3m": 0.10175523113079621, "alpha_c_p3m": 0.17227249087824992, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.15821865456564144, "ret_p6m": 3.714562761474756, "ret_signed_p6m": -3.714562761474756, "alpha_spy_p6m": -3.5496820764117816, "alpha_c_p6m": -3.419852262558849, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.29471049891590706, "price_path": [null, -0.1277, -0.1119, -0.1119, 0.1425, 0.1146, 0.1694, 0.2205, 0.1776, 0.1669, 0.1287, 0.0943, 0.1352, 0.219, 0.1798, 0.1745, 0.1267, 0.2358, 0.2842, 0.2639, 0.2222, 0.1737, 0.3283, 0.3484, 0.3671, 0.3649, 0.362, 0.3075, 0.284, 0.2648, 0.275, 0.2813, 0.1866, 0.1536, 0.1657, 0.1551, -0.0729, -0.2704, -0.2372, -0.2161, -0.1136, -0.247, -0.2253, -0.1871, -0.1948, -0.2239, -0.2418, -0.2418, -0.2777, -0.2823, -0.2644, -0.2176, -0.204, -0.2181, -0.2018, -0.222, -0.2103, -0.1665, -0.1614, -0.1803, -0.1527, -0.1117, -0.0855, 0.0, 0.0177, 0.016, -0.0177, -0.0201, -0.0603, -0.0264, -0.0647, -0.0831, -0.0967, -0.0967, -0.0749, -0.0652, -0.0647, -0.0867, -0.0955, -0.0637, -0.0351, -0.0521, -0.0514, 0.0133, 0.0078, -0.0252, 0.0007, 0.0298, 0.0712, 0.0683, 0.1296, 0.1296, 0.1287, 0.1376, 0.1471, 0.1449, 0.1495, 0.1425, 0.0989, 0.0894, 0.1197, 0.1245, 0.1245, 0.0957, 0.1187, 0.1195, 0.1376, 0.1168, 0.0293, 0.0351, 0.0022, 0.0061, 0.0223, 0.0082, 0.0022, 0.0419, 0.0191, 0.0293, 0.015, 0.0402, 0.0514, 0.04, 0.0015, 0.03, 0.016, 0.0201, -0.0141, 0.0744, 0.0509, 0.1347, 0.1391, 0.1311, 0.0792, 0.103, 0.0802, 0.0758, 0.1025, 0.1236, 0.1335, 0.1473, 0.1737, 0.2326, 0.2714, 0.2714, 0.2375, 0.2845, 0.5144, 0.661, 0.7081, 0.7085, 0.7911, 1.0426, 1.087, 1.1829, 1.2183, 1.277, 1.3957, 1.4766, 1.4938, 1.5781, 1.4189, 1.2847, 1.3533, 1.7502, 1.7187, 1.9348, 2.008, 2.1115, 1.936, 1.9307, 2.1955, 2.1422, 1.8328, 2.2617, 2.0843, 2.4965, 2.4958, 2.3962, 2.5871, 2.6174, 2.5607, 3.0477, 3.5108, 3.2765, 3.2532, 3.9518, 3.7449, 3.8299, 4.016, 3.7146]}
{"unit_id": "orig:1932076453639884868", "ticker_idx": 0, "ts": "2025-06-09T14:04:57+00:00", "symbol": "ANET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes these concerns have been overly magnified, and the opportunity in Ethernet-based AI will be strongly validated in the second half... Base-case Valuation Suggests Market Expectations Can Be Surpassed", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley Buy thesis on ANET: 2H backend AI networking revenue catalyst, conservative $2.25B estimate surpasses consensus, architectural debate overblown.", "resolved_tickers": ["ANET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Arista Networks: Architectural Debate Unresolved, 2H AI Networking Revenue Will Act as a Stock Catalyst\nMorgan Stanley (M. Marshall, 25/06/09)\n\nDespite intense discussions in the first half regarding the adaptability of backend AI network architectures, Morgan Stanley believes these concerns have been overly magnified, and the opportunity in Ethernet-based AI will be strongly validated in the second half.\n\n• 2H Backend Networking Revenue Is a Key Catalyst\nMorgan Stanley expects the “Demystifying AI” technology briefing on June 12 to mainly focus on architectural details with limited stock impact; the true catalyst is the actual backend network revenue performance in the second half, driven by increased Blackwell deployment. If growth exceeds expectations, sentiment could quickly shift towards a bullish scenario.\n\n• Base-case Valuation Suggests Market Expectations Can Be Surpassed\nUnder conservative assumptions, Morgan Stanley estimates that in 2025, Arista’s frontend, backend, and campus networking businesses will each generate approximately $750 million, totaling $2.25 billion. This maps to approximately $3 EPS in 2026, surpassing market EPS forecasts for 2025/2026, reflecting the feasibility of outperformance.\n\n• Architectural Debate Has Been Overemphasized\nMorgan Stanley argues the market’s skepticism about Arista’s backend networking share ignores the sustained growth and architectural advantages of its frontend business. Even if backend share is only half that of frontend, Arista could still meet or slightly exceed market revenue and earnings expectations. Thus, short-term concerns around technological adaptability are not decisive.\n\n• Risks and Considerations\nMorgan Stanley cautions that an unexpected slowdown in cloud capital expenditures could disrupt revenue momentum. Investors should closely monitor market feedback following the AI briefing and the realization of backend revenue in the second half, to assess whether to further increase positions or guard against cyclical risks.\n\n$ANET", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004648728657597134, "ret_signed_m1d": 0.004648728657597134, "alpha_spy_m1d": 0.005549199966359608, "alpha_c_m1d": 0.009533267903400588, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.004884539245803454, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0320248554316479, "ret_signed_p1d": -0.0320248554316479, "alpha_spy_p1d": -0.03769452010053864, "alpha_c_p1d": -0.037619129934924556, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005594274503276653, "ret_p1w": -0.01766535805740488, "ret_signed_p1w": -0.01766535805740488, "alpha_spy_p1w": -0.022668033631809248, "alpha_c_p1w": -0.03231903986190843, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.014653681804503549, "ret_p1m": 0.06807847241871512, "ret_signed_p1m": 0.06807847241871512, "alpha_spy_p1m": 0.030568545821487714, "alpha_c_p1m": -0.0036652679219693063, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.07174374034068443, "ret_p3m": 0.45836770370206903, "ret_signed_p3m": 0.45836770370206903, "alpha_spy_p3m": 0.37272379208023443, "alpha_c_p3m": 0.3600697549445959, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.09829794875747311, "ret_p6m": 0.31425616952388014, "ret_signed_p6m": 0.31425616952388014, "alpha_spy_p6m": 0.17124666613890271, "alpha_c_p6m": 0.10291772914628661, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.21133844037759353, "price_path": [-0.171, -0.1721, -0.1373, -0.121, -0.1323, -0.1279, -0.1429, -0.1412, -0.096, -0.1019, -0.1564, -0.1843, -0.1948, -0.1996, -0.1892, -0.1719, -0.2637, -0.335, -0.2958, -0.282, -0.2149, -0.2495, -0.2493, -0.2398, -0.2438, -0.2575, -0.2645, -0.2645, -0.3009, -0.2906, -0.269, -0.2222, -0.1951, -0.191, -0.166, -0.1501, -0.0926, -0.0597, -0.0663, -0.0623, -0.1069, -0.0985, -0.1062, -0.0463, 0.0046, 0.0147, -0.0075, -0.0039, -0.001, -0.0117, -0.0451, -0.0438, -0.0579, -0.0579, -0.0445, -0.0414, -0.1077, -0.105, -0.0725, -0.0238, -0.0192, -0.0167, 0.0046, 0.0, -0.032, -0.0268, -0.0106, -0.046, -0.0177, -0.0736, -0.0678, -0.0678, -0.109, -0.0501, -0.0189, -0.0051, 0.0495, 0.0268, 0.0569, 0.0218, 0.0447, 0.0591, 0.0591, 0.0482, 0.0681, 0.0979, 0.098, 0.1216, 0.1195, 0.1092, 0.1188, 0.1568, 0.1548, 0.153, 0.1341, 0.1678, 0.1781, 0.1806, 0.2144, 0.2254, 0.2613, 0.2729, 0.2146, 0.2433, 0.2202, 0.4337, 0.4388, 0.4378, 0.422, 0.4592, 0.4257, 0.4099, 0.4184, 0.426, 0.3717, 0.3582, 0.3639, 0.3765, 0.3744, 0.3871, 0.3768, 0.4073, 0.4106, 0.4106, 0.4036, 0.4192, 0.4584, 0.4757, 0.4464, 0.466, 0.557, 0.581, 0.44, 0.5024, 0.4686, 0.4756, 0.5151, 0.5456, 0.5021, 0.4885, 0.4736, 0.4779, 0.4721, 0.4811, 0.5053, 0.542, 0.4924, 0.5031, 0.5444, 0.5009, 0.6256, 0.6346, 0.5919, 0.5232, 0.4338, 0.4812, 0.5084, 0.4783, 0.5132, 0.5076, 0.5144, 0.5781, 0.589, 0.6199, 0.6195, 0.6739, 0.6368, 0.629, 0.628, 0.5863, 0.4506, 0.3845, 0.391, 0.418, 0.3939, 0.3944, 0.3461, 0.3571, 0.3147, 0.2753, 0.2894, 0.2354, 0.2131, 0.2621, 0.2917, 0.3187, 0.3187, 0.35, 0.3235, 0.3143]}
{"unit_id": "orig:1928441675019632769", "ticker_idx": 0, "ts": "2025-05-30T13:21:38+00:00", "symbol": "ANET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Weakness Presents a Buying Opportunity; Ethernet TAM Expanding", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Citi views ANET weakness as a buying opportunity; Ethernet AI backend TAM expanding and ANET doubled market share through FY2024.", "resolved_tickers": ["ANET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Citi) Arista Networks (ANET.N) – Weakness Presents a Buying Opportunity; Ethernet TAM Expanding\n\nIn its Q1 earnings report last night, NVIDIA (NVDA) revealed that its Spectrum-X revenue has reached an annualized run rate of $8 billion and that it has added Google Cloud and Meta as new Spectrum-X cloud customers. These two announcements have raised concerns that Arista and other players may be losing share in the Ethernet-based AI backend switch market.\n\nHowever, we continue to believe that the total addressable market (TAM) for Ethernet AI backend switching is rapidly expanding, and we have noted that Arista Networks (ANET) has doubled its market share through FY2024. While switching likely accounts for a relatively small portion of the overall Spectrum-X revenue (with Dell’Oro estimating it to be less than 30%), we view NVIDIA’s Ethernet switch revenue as benefiting from the same TAM expansion that is also driving growth for Arista and others.\n\nLastly, we believe that hyperscalers continue to pursue diversification in networking. Google appears to be moving away from a single-vendor strategy, and Meta seems to be taking a diversified approach to AI networking, including InfiniBand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0031163047664871657, "ret_signed_m1d": -0.0031163047664871657, "alpha_spy_m1d": -0.004235909070317789, "alpha_c_m1d": -0.006234349858481925, "bench_spy_m1d": 0.0011196043038306236, "bench_c_m1d": 0.003118045091994759, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.036241913801578374, "ret_signed_p1d": 0.036241913801578374, "alpha_spy_p1d": 0.03060901266476468, "alpha_c_p1d": 0.027407353019470504, "bench_spy_p1d": 0.005632901136813695, "bench_c_p1d": 0.00883456078210787, "ret_p1w": 0.12246076506343129, "ret_signed_p1w": 0.12246076506343129, "alpha_spy_p1w": 0.1059183993685513, "alpha_c_p1w": 0.09019716946438283, "bench_spy_p1w": 0.01654236569487999, "bench_c_p1w": 0.032263595599048456, "ret_p1m": 0.180863322707012, "ret_signed_p1m": 0.180863322707012, "alpha_spy_p1m": 0.12947716166142786, "alpha_c_p1m": 0.0823848159978684, "bench_spy_p1m": 0.05138616104558413, "bench_c_p1m": 0.0984785067091436, "ret_p3m": 0.5382041229375141, "ret_signed_p3m": 0.5382041229375141, "alpha_spy_p3m": 0.4378435221575723, "alpha_c_p3m": 0.3910547513182705, "bench_spy_p3m": 0.1003606007799418, "bench_c_p3m": 0.14714937161924357, "ret_p6m": 0.41008770809781336, "ret_signed_p6m": 0.41008770809781336, "alpha_spy_p6m": 0.26896498452967665, "alpha_c_p6m": 0.19518311096011187, "bench_spy_p6m": 0.1411227235681367, "bench_c_p6m": 0.2149045971377015, "price_path": [-0.0129, 0.0136, -0.0265, -0.0379, -0.1048, -0.0867, -0.0738, -0.075, -0.0361, -0.0179, -0.0306, -0.0256, -0.0424, -0.0405, 0.01, 0.0035, -0.0575, -0.0886, -0.1004, -0.1057, -0.0941, -0.0748, -0.1774, -0.257, -0.2132, -0.1978, -0.1228, -0.1615, -0.1612, -0.1506, -0.1551, -0.1705, -0.1782, -0.1782, -0.219, -0.2074, -0.1833, -0.131, -0.1008, -0.0961, -0.0682, -0.0504, 0.0139, 0.0506, 0.0432, 0.0477, -0.0022, 0.0073, -0.0014, 0.0656, 0.1225, 0.1337, 0.1088, 0.1129, 0.1161, 0.1042, 0.0668, 0.0683, 0.0526, 0.0526, 0.0675, 0.071, -0.0031, 0.0, 0.0362, 0.0907, 0.0958, 0.0986, 0.1225, 0.1173, 0.0815, 0.0874, 0.1054, 0.0659, 0.0975, 0.0351, 0.0416, 0.0416, -0.0045, 0.0613, 0.0961, 0.1116, 0.1726, 0.1472, 0.1809, 0.1416, 0.1672, 0.1833, 0.1833, 0.1712, 0.1933, 0.2267, 0.2268, 0.2531, 0.2508, 0.2393, 0.25, 0.2925, 0.2902, 0.2882, 0.2671, 0.3047, 0.3163, 0.319, 0.3568, 0.3691, 0.4092, 0.4222, 0.357, 0.3891, 0.3633, 0.6018, 0.6076, 0.6064, 0.5888, 0.6303, 0.5929, 0.5753, 0.5847, 0.5933, 0.5325, 0.5174, 0.5239, 0.538, 0.5355, 0.5497, 0.5382, 0.5724, 0.5761, 0.5761, 0.5682, 0.5856, 0.6294, 0.6488, 0.616, 0.6379, 0.7396, 0.7664, 0.6088, 0.6786, 0.6408, 0.6487, 0.6928, 0.7268, 0.6782, 0.6631, 0.6464, 0.6512, 0.6447, 0.6548, 0.6818, 0.7229, 0.6674, 0.6794, 0.7255, 0.6769, 0.8163, 0.8263, 0.7786, 0.7019, 0.6019, 0.6549, 0.6852, 0.6517, 0.6907, 0.6844, 0.6919, 0.7632, 0.7754, 0.8099, 0.8094, 0.8702, 0.8287, 0.8201, 0.8189, 0.7723, 0.6207, 0.5469, 0.5541, 0.5843, 0.5574, 0.5579, 0.5039, 0.5163, 0.4688, 0.4249, 0.4406, 0.3803, 0.3554, 0.4101]}
{"unit_id": "quote:1930135799733072027", "ticker_idx": 0, "ts": "2025-06-04T05:33:29+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Trump allegedly opened a backdoor for NVIDIA to sell a new downgraded AI chip called the 'B30' to China. Major Chinese tech giants like ByteDance, Alibaba, and Tencent have already placed orders.", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares bullish report: NVIDIA's B30 chip gains China orders; Gigabyte is the primary beneficiary with guidance raise and May revenue +100% YoY.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Trump allegedly opened a backdoor for NVIDIA to sell a new downgraded AI chip called the ‘B30’ to China. Major Chinese tech giants like ByteDance, Alibaba, and Tencent have already placed orders. The B30 chip is based on the latest Blackwell architecture but deliberately avoids using high-bandwidth memory (HBM) and TSMC’s (2330) advanced packaging technology. \n\nThe bulk of the B30 orders reportedly went to Gigabyte (2376), which is expected to raise its financial guidance during its upcoming earnings call. Gigabyte’s May revenue is said to have more than doubled year-on-year.", "reply_text": null, "parent_text": null, "quoted_text": "轉\n川普給輝達走後門，開放新款降規版AI晶片「B30」，已經先拿到中國科技巨頭如字節跳動、阿里巴巴與騰訊的買單，B30晶片採用最新一代Blackwell架構，刻意不使用高頻寬記憶體(HBM)與台積電(2330)的先進封裝技術，B30大訂單主要的訂單是由技嘉(2376)吃到，技嘉法說會將調高財測，5月營收YoY 100%起跳", "parent_author": null, "ret_m1d": -0.004932421576486434, "ret_signed_m1d": -0.004932421576486434, "alpha_spy_m1d": -0.005201041116532745, "alpha_c_m1d": 0.006825285213173493, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013599137425771679, "ret_signed_p1d": -0.013599137425771679, "alpha_spy_p1d": -0.008766368369582445, "alpha_c_p1d": -0.011811632812611439, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.006481889145962372, "ret_signed_p1w": 0.006481889145962372, "alpha_spy_p1w": -0.0026300420161626814, "alpha_c_p1w": -0.03375642048266836, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.12282315830271795, "ret_signed_p1m": 0.12282315830271795, "alpha_spy_p1m": 0.07036941773677508, "alpha_c_p1m": -0.003770715177219053, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.22739630300135127, "ret_signed_p3m": 0.22739630300135127, "alpha_spy_p3m": 0.14177048370139467, "alpha_c_p3m": 0.07430784158682724, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.2703119733794417, "ret_signed_p6m": 0.2703119733794417, "alpha_spy_p6m": 0.12323205321321051, "alpha_c_p6m": -0.11085976629422323, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.206, -0.2463, -0.2337, -0.1845, -0.1856, -0.1427, -0.1578, -0.1867, -0.1719, -0.1648, -0.1707, -0.1445, -0.1496, -0.1984, -0.2148, -0.2272, -0.2363, -0.2239, -0.222, -0.2827, -0.3355, -0.312, -0.3214, -0.1944, -0.242, -0.2184, -0.2199, -0.2094, -0.2637, -0.2849, -0.2849, -0.3172, -0.3032, -0.2763, -0.2501, -0.2178, -0.2339, -0.2318, -0.2325, -0.2136, -0.1932, -0.198, -0.2, -0.1752, -0.173, -0.1781, -0.1333, -0.0845, -0.0464, -0.05, -0.0459, -0.0447, -0.0531, -0.0713, -0.0641, -0.0749, -0.0749, -0.0452, -0.0501, -0.0192, -0.0478, -0.032, -0.0049, 0.0, -0.0136, -0.0014, 0.005, 0.0144, 0.0065, 0.0218, 0.0004, 0.0196, 0.0156, 0.0252, 0.0252, 0.0137, 0.0159, 0.0422, 0.0874, 0.0924, 0.1116, 0.1133, 0.0803, 0.1081, 0.1228, 0.1228, 0.1151, 0.1275, 0.1478, 0.1564, 0.1621, 0.1562, 0.2029, 0.2076, 0.2191, 0.2149, 0.2077, 0.177, 0.2034, 0.2243, 0.2226, 0.2455, 0.2368, 0.2633, 0.2534, 0.2242, 0.2684, 0.2561, 0.2643, 0.2738, 0.2874, 0.2829, 0.2907, 0.2796, 0.2826, 0.2716, 0.2826, 0.2377, 0.236, 0.233, 0.2542, 0.2671, 0.2809, 0.2797, 0.2696, 0.2274, 0.2274, 0.2034, 0.2023, 0.2096, 0.1769, 0.186, 0.2033, 0.2496, 0.2485, 0.2531, 0.2526, 0.2324, 0.2001, 0.242, 0.245, 0.2939, 0.2574, 0.2471, 0.2522, 0.2557, 0.2815, 0.3148, 0.3195, 0.3311, 0.3222, 0.3075, 0.304, 0.3327, 0.3571, 0.2907, 0.3271, 0.2687, 0.2673, 0.2812, 0.2912, 0.2871, 0.2767, 0.2705, 0.2837, 0.3126, 0.3495, 0.4167, 0.459, 0.4298, 0.427, 0.4579, 0.4002, 0.3757, 0.3254, 0.3259, 0.4027, 0.3612, 0.3657, 0.3168, 0.3401, 0.315, 0.2781, 0.3144, 0.273, 0.2606, 0.2864, 0.2531, 0.2703, 0.2703]}
{"unit_id": "quote:1930135799733072027", "ticker_idx": 1, "ts": "2025-06-04T05:33:29+00:00", "symbol": "2376.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The bulk of the B30 orders reportedly went to Gigabyte (2376), which is expected to raise its financial guidance during its upcoming earnings call. Gigabyte's May revenue is said to have more than doubled year-on-year.", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares bullish report: NVIDIA's B30 chip gains China orders; Gigabyte is the primary beneficiary with guidance raise and May revenue +100% YoY.", "resolved_tickers": ["2376.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "Trump allegedly opened a backdoor for NVIDIA to sell a new downgraded AI chip called the ‘B30’ to China. Major Chinese tech giants like ByteDance, Alibaba, and Tencent have already placed orders. The B30 chip is based on the latest Blackwell architecture but deliberately avoids using high-bandwidth memory (HBM) and TSMC’s (2330) advanced packaging technology. \n\nThe bulk of the B30 orders reportedly went to Gigabyte (2376), which is expected to raise its financial guidance during its upcoming earnings call. Gigabyte’s May revenue is said to have more than doubled year-on-year.", "reply_text": null, "parent_text": null, "quoted_text": "轉\n川普給輝達走後門，開放新款降規版AI晶片「B30」，已經先拿到中國科技巨頭如字節跳動、阿里巴巴與騰訊的買單，B30晶片採用最新一代Blackwell架構，刻意不使用高頻寬記憶體(HBM)與台積電(2330)的先進封裝技術，B30大訂單主要的訂單是由技嘉(2376)吃到，技嘉法說會將調高財測，5月營收YoY 100%起跳", "parent_author": null, "ret_m1d": -0.01567940794019429, "ret_signed_m1d": -0.01567940794019429, "alpha_spy_m1d": -0.0159480274802406, "alpha_c_m1d": -0.014201254783333783, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.0014781531568605066, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0017421564377992915, "ret_signed_p1d": 0.0017421564377992915, "alpha_spy_p1d": 0.006574925493988526, "alpha_c_p1d": 0.004571638001291234, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0028294815634919424, "ret_p1w": -0.010452938626796193, "ret_signed_p1w": -0.010452938626796193, "alpha_spy_p1w": -0.019564869788921246, "alpha_c_p1w": -0.025402122535126637, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.014949183908330443, "ret_p1m": -0.03658539542894479, "ret_signed_p1m": -0.03658539542894479, "alpha_spy_p1m": -0.08903913599488766, "alpha_c_p1m": -0.12351893368653477, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.08693353825758998, "ret_p3m": -0.03735252189380478, "ret_signed_p3m": -0.03735252189380478, "alpha_spy_p3m": -0.12297834119376139, "alpha_c_p3m": -0.14750872825121786, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.11015620635741308, "ret_p6m": -0.10779014224303851, "ret_signed_p6m": -0.10779014224303851, "alpha_spy_p6m": -0.2548700624092697, "alpha_c_p6m": -0.30971419748289675, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.20192405523985824, "price_path": [-0.0767, -0.0697, -0.0801, -0.0871, -0.108, -0.0941, -0.0993, -0.0767, -0.0801, -0.0662, -0.0697, -0.0714, -0.0645, -0.0767, -0.0906, -0.1045, -0.1568, -0.1411, -0.1394, -0.1394, -0.1394, -0.2247, -0.2997, -0.3693, -0.3066, -0.2962, -0.2596, -0.2352, -0.2509, -0.2352, -0.2369, -0.2596, -0.2718, -0.2334, -0.2369, -0.1951, -0.1969, -0.1882, -0.2038, -0.2038, -0.1777, -0.1899, -0.1655, -0.1725, -0.1585, -0.1481, -0.1324, -0.1341, -0.0836, -0.0749, -0.0923, -0.1063, -0.1045, -0.0941, -0.101, -0.108, -0.0993, -0.1063, -0.0592, -0.0383, -0.0383, -0.0732, -0.0157, 0.0, 0.0017, 0.0035, 0.0087, 0.0, -0.0105, -0.0139, 0.0, -0.0139, -0.0139, -0.007, -0.0244, -0.0244, -0.0348, -0.0105, -0.0087, -0.0105, -0.0192, -0.0139, -0.0017, -0.0192, -0.0366, -0.0488, -0.0505, -0.0453, -0.0296, -0.0314, -0.0226, -0.0366, -0.0122, -0.0087, 0.0, 0.0087, 0.0122, -0.0244, 0.0035, 0.0122, 0.0017, 0.007, -0.0157, -0.0301, -0.0066, -0.0066, -0.0121, 0.0042, 0.0042, 0.0277, 0.0331, 0.0457, 0.0313, 0.0367, 0.0186, 0.0078, 0.0186, 0.0259, -0.0247, -0.0157, -0.0211, 0.0024, 0.0132, 0.0277, 0.015, 0.0006, -0.0374, -0.0482, -0.0464, -0.0428, -0.0283, -0.0085, 0.0277, 0.0006, -0.0103, -0.0103, -0.0193, -0.0012, -0.0247, -0.0048, 0.0186, 0.015, 0.0222, 0.0042, 0.0584, 0.0584, 0.0584, 0.08, 0.0584, 0.08, 0.1252, 0.1252, 0.1234, 0.1071, 0.08, 0.08, 0.0439, 0.0114, 0.0078, 0.0222, 0.006, 0.0295, 0.0403, 0.0186, 0.0096, 0.0096, 0.0295, 0.0349, 0.0277, 0.0168, 0.0132, 0.0006, -0.0355, -0.0355, 0.006, -0.0211, -0.003, -0.0066, -0.0211, -0.0247, -0.0229, -0.0753, -0.1114, -0.1349, -0.1042, -0.1349, -0.1204, -0.124, -0.1114, -0.1078]}
{"unit_id": "thread:1933295951604797556", "ticker_idx": 0, "ts": "2025-06-12T22:50:48+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron anticipates that prices will continue to rise... Micron plans to push for robust and aggressive pricing capabilities in the coming months... gross margins and pricing still have significant upside potential", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Micron confirms DDR4 EOL, rising DRAM prices from Q2 2025, aggressive pricing push in mobile, and significant margin upside — bullish for MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive Interview with Micron by DigiTimes: Prices to Rise Again! Micron Confirms DDR4 End-of-Life and Significant Shortages\n\nThe world's three major DRAM manufacturers are definitively shifting from DDR4 specifications to advanced process products. Following announcements from the two major Korean memory companies regarding their DDR4 discontinuation schedules, Micron has now confirmed it sent letters to customers notifying them of DDR4's End-of-Life (EOL). Shipments are expected to gradually cease over the next 2-3 quarters.\n\nSumit Sadana, Micron's Executive Vice President and Chief Business Officer, stated in an interview with DIGITIMES that DDR4 will continue to experience a \"severe shortage.\" In the future, Micron's DDR4/LPDDR4 DRAM will only be supplied strategically to long-term customers in three key areas. Meanwhile, DDR5/LPDDR5 products are entering a \"sweet spot\" in market pricing.\n\nMicron simultaneously released its future market strategy, indicating it will increase pricing power in segments with lower profitability compared to other market areas, ensuring increased margins. Sadana specifically pointed out that product pricing and profitability in the mobile device market have lagged behind other markets, and Micron plans to push for robust and aggressive pricing capabilities in the coming months.\n\nWith major memory manufacturers poised to transition DRAM process capacity, Micron, which had not previously announced its DDR4 discontinuation plans, has finally made an official statement.\n\nSadana said that DDR4/LPDDR4 EOL notices have recently been delivered to customers, primarily targeting the PC and data center segments. Over the next three quarters, production of DDR4 DRAM for consumer, PC, and data center use will be scaled down or reduced.\n\nIn the future, supply will primarily be for long-term partners in the \"automotive, industrial, and networking\" markets, mainly because these three market segments place a higher emphasis on quality and require long-term contract guarantees.\n\nThe memory market cycle has been rapidly changing in recent years, especially for DDR4 and LPDDR4. Sadana noted that the recent market is experiencing a supply-demand imbalance, leading to sharp price increases. He even suggested that the prices of LPDDR4 and DDR4 might exceed those of DDR5/LPDDR5.\n\nAlthough the DDR4 series is an older product and accounts for a very small portion of Micron's revenue, Micron acknowledges that it does pose some challenges for certain customers and is committed to addressing their pain points.\n\nDespite Micron's commitment to maintain stable multi-year supply for specialized customers, the increasing demand for High Bandwidth Memory (HBM) has led to a significant surge in demand for DRAM wafers, thereby limiting the supply of non-HBM products. Market prices have shown a noticeable upward trend starting from Q2 2025.\n\nMicron anticipates that prices will continue to rise.\n\nSadana also urged customers to consider that DDR5/LPDDR5 is entering a market sweet spot, and for mainstream application customers like mobile devices, PCs, and data centers, upgrading will be more advantageous, as DDR5/LPDDR5 supply will also face tightness.\n\nWith data center demand remaining strong and Cloud Service Providers (CSPs) making large-scale capital investments, AI-related capital expenditure will continue to increase in 2025 and remain very substantial in 2026.\n\nFurthermore, different countries are now investing in sovereign AI initiatives, implying capital expenditure plans will extend for many years. Micron believes this timing presents an exciting and tremendous opportunity for the memory industry.\n\nBoth AI training and inference require a significant amount of memory, and the current development bottleneck is also in memory. Simultaneously increasing memory capacity and bandwidth is key to enhancing system performance.\n\nSadana pointed out that PC and mobile device demand is currently stable, and inventory levels have returned to a healthy level as of early 2025.\n\nWhile PC and mobile device shipments are flat, they are expected to see low single-digit growth in 2025. However, the introduction of AI features requires more DRAM for PC, laptop, and mobile products. For example, smartphones are already moving from 8GB to 12GB or 16GB.\n\nNevertheless, prices in the mobile device market are generally lower than in other applications. Micron stated that it will further focus on and enhance its more robust and aggressive pricing capabilities in this segment.\n\nSadana emphasized that compared to other semiconductor industries, memory's gross margins and pricing still have significant upside potential, not yet reaching the general level of the semiconductor industry.\n\nThe key focus is on improving overall profitability to \"ensure ideal returns on investment\" and reinvesting capital into R&D and new fab investments to capitalize on the overall industry growth momentum.\n\n---\n\nhttps://t.co/nD0g38gyGo", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0012911097738207422, "ret_signed_m1d": -0.0012911097738207422, "alpha_spy_m1d": 0.00266738437259173, "alpha_c_m1d": 0.003535003680910309, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.004826113454731051, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004992304290626248, "ret_signed_p1d": -0.004992304290626248, "alpha_spy_p1d": 0.006187760105841211, "alpha_c_p1d": 0.01841642829213086, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.023408732582757108, "ret_p1w": 0.048545372035644574, "ret_signed_p1w": 0.048545372035644574, "alpha_spy_p1w": 0.05899658188000734, "alpha_c_p1w": 0.050673417482961614, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.0021280454473170396, "ret_p1m": 0.0728801648175943, "ret_signed_p1m": 0.0728801648175943, "alpha_spy_p1m": 0.036915569070615994, "alpha_c_p1m": -0.019614604062105467, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.09249476887969976, "ret_p3m": 0.16515159513999267, "ret_signed_p3m": 0.16515159513999267, "alpha_spy_p3m": 0.08481603573952268, "alpha_c_p3m": 0.033439623119026596, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.13171197202096607, "ret_p6m": 1.0450333612391387, "ret_signed_p6m": 1.0450333612391387, "alpha_spy_p6m": 0.9027993586404981, "alpha_c_p6m": 0.6601956588767675, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.38483770236237125, "price_path": [-0.1137, -0.1256, -0.1227, -0.1146, -0.1858, -0.1667, -0.1904, -0.208, -0.2164, -0.2398, -0.2521, -0.2364, -0.2374, -0.3601, -0.4429, -0.4115, -0.4359, -0.3297, -0.3971, -0.4014, -0.3887, -0.3885, -0.4033, -0.4078, -0.4078, -0.4255, -0.3957, -0.3723, -0.3336, -0.3133, -0.3238, -0.3383, -0.3377, -0.3306, -0.3052, -0.3078, -0.307, -0.2889, -0.2671, -0.261, -0.2056, -0.1657, -0.1795, -0.1784, -0.1565, -0.1509, -0.1556, -0.1751, -0.1838, -0.1963, -0.1963, -0.1704, -0.1721, -0.1668, -0.187, -0.1549, -0.1199, -0.1113, -0.0851, -0.0656, -0.045, -0.0176, -0.0013, 0.0, -0.005, 0.0315, 0.0358, 0.0485, 0.0485, 0.0639, 0.0508, 0.101, 0.0953, 0.0845, 0.0739, 0.0609, 0.0405, 0.0479, 0.0526, 0.0526, 0.0332, 0.0719, 0.0532, 0.0606, 0.0729, 0.0219, 0.0348, 0.0031, -0.0242, -0.0145, -0.0245, -0.059, -0.0538, -0.0374, -0.0414, -0.0415, -0.0354, -0.0115, -0.0597, -0.0964, -0.0715, -0.0604, -0.0628, -0.0362, 0.0243, 0.0659, 0.1006, 0.0706, 0.0794, 0.0413, 0.0644, 0.0515, 0.0098, -0.0024, 0.0139, 0.003, 0.0037, 0.0145, 0.0511, 0.0253, 0.0253, 0.0208, 0.0228, 0.0701, 0.1318, 0.1326, 0.1652, 0.2062, 0.2972, 0.3546, 0.3593, 0.3683, 0.3784, 0.4551, 0.402, 0.4183, 0.4337, 0.3932, 0.3512, 0.3549, 0.4121, 0.4415, 0.5693, 0.5831, 0.6193, 0.6462, 0.6008, 0.6943, 0.658, 0.5655, 0.6618, 0.6126, 0.6547, 0.746, 0.7447, 0.7825, 0.7439, 0.711, 0.782, 0.8881, 0.8974, 0.913, 0.9537, 0.9312, 0.9291, 1.0233, 0.8796, 1.0474, 1.0546, 1.0511, 1.1837, 1.0786, 1.1112, 1.0427, 1.1279, 1.0858, 0.9699, 0.9476, 0.736, 0.7877, 0.9305, 0.9356, 0.985, 0.985, 1.0387, 1.073, 1.0646, 1.0187, 0.9539, 1.045]}
{"unit_id": "orig:1933343961730724035", "ticker_idx": 0, "ts": "2025-06-13T02:01:34+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVDA leads 800 G port growth of over 100 %... NVDA has become the sales champion for 800 G ports and equipment", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi note: long NVDA and CLS as 800G leaders, cautious CSCO on minimal AI backend exposure, ANET stable with 800G acceleration in 2H25.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "US Communications Equipment: Ethernet Accelerates Replacement of IB, 800G Deployment Shows Strong Growth, NVDA and CLS Lead, ANET and CSCO Diverge\n\nCiti (A. Malik, 25/06/12)\n\nEthernet’s pace in displacing InfiniBand in the data-center switch market has exceeded expectations, lifting 2025 switch-market growth from 14 % to 23 %. In the 800 G era, market-share reshuffling highlights NVDA and CLS as the most striking performers.\n\n• Ethernet’s replacement of InfiniBand drives accelerated switch growth\n\nResearch indicates that Ethernet’s displacement of InfiniBand has surpassed Dell’Oro’s earlier expectations, resulting in significantly better-than-forecast market growth in 1Q25. Dell’Oro has raised its 2025 data-center-switch growth forecast from 14 % to 23 %.\n\n• Arista backend deployment stable; 800 G to accelerate in 2H25\n\nArista’s backend-switch deployment is estimated at US$1.6 billion, about an 11 % market share, flat versus FY24. Although 800 G shipments are “still slowly ramping up,” they are expected to accelerate markedly in 2H25, with enterprise networks standing out thanks to competitive displacement and share gains.\n\n• NVDA leads 800 G port growth of over 100 %\n\nNVDA has become the sales champion for 800 G ports and equipment, with port shipments surging over 100 % QoQ. Drivers include Blackwell server read acceleration (about 70 % of DC compute sales) and the migration from InfiniBand to Ethernet. Its SpectrumX platform has seen significant revenue uplift from large wins with xAI, Azure, Oracle, Google Cloud, and CoreWeave.\n\n• Cisco frontend networking solid; backend AI deployment still modest\n\nCisco’s data-center-switch sales are driven mainly by Tier-2/3 cloud providers and enterprise frontend networks. AI backend-switch shipments are “not yet significant,” as related AI-networking products are still booked under the router segment, making large near-term growth unlikely.\n\n• CLS 800 G shipments nearly triple QoQ\n\nCLS (Celestica) saw May 800 G equipment shipments nearly triple QoQ, driven by accelerated deployments at Google and AMZN and Meta’s first large-scale adoption. Its AI-networking gear is also showing eye-catching momentum across the industry.\n\n$NVDA $CLS $ANET $CSCO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02134263768652178, "ret_signed_m1d": 0.02134263768652178, "alpha_spy_m1d": 0.010036166210859676, "alpha_c_m1d": -0.0026271983787395925, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.023969836065261374, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019159054367494432, "ret_signed_p1d": 0.019159054367494432, "alpha_spy_p1d": 0.009644857888779113, "alpha_c_p1d": -0.005705846168853279, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.02486490053634771, "ret_p1w": 0.013242320775086691, "ret_signed_p1w": 0.013242320775086691, "alpha_spy_p1w": 0.014855595770896013, "alpha_c_p1w": 0.0004791350126391869, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.012763185762447504, "ret_p1m": 0.155666819563006, "ret_signed_p1m": 0.155666819563006, "alpha_spy_p1m": 0.1059900277972956, "alpha_c_p1m": 0.04542892109910057, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.11023789846390541, "ret_p3m": 0.2490668468027053, "ret_signed_p3m": 0.2490668468027053, "alpha_spy_p3m": 0.1533581003828346, "alpha_c_p3m": 0.07847654022942518, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.17059030657328011, "ret_p6m": 0.3071128831845811, "ret_signed_p6m": 0.3071128831845811, "alpha_spy_p6m": 0.15543463149858416, "alpha_c_p6m": -0.12698969258559245, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.4341025757701735, "price_path": [-0.187, -0.1723, -0.1652, -0.171, -0.1449, -0.1499, -0.1988, -0.2152, -0.2276, -0.2367, -0.2242, -0.2223, -0.283, -0.3358, -0.3123, -0.3217, -0.1947, -0.2424, -0.2187, -0.2202, -0.2097, -0.2641, -0.2852, -0.2852, -0.3174, -0.3035, -0.2766, -0.2504, -0.2181, -0.2342, -0.2321, -0.2328, -0.2139, -0.1935, -0.1983, -0.2003, -0.1755, -0.1733, -0.1784, -0.1337, -0.0849, -0.0468, -0.0504, -0.0463, -0.0451, -0.0535, -0.0717, -0.0644, -0.0753, -0.0753, -0.0456, -0.0505, -0.0196, -0.0482, -0.0324, -0.0054, -0.0004, -0.014, -0.0018, 0.0046, 0.0139, 0.0061, 0.0213, 0.0, 0.0192, 0.0151, 0.0247, 0.0247, 0.0132, 0.0155, 0.0418, 0.0869, 0.0919, 0.1112, 0.1128, 0.0798, 0.1076, 0.1223, 0.1223, 0.1146, 0.127, 0.1473, 0.1559, 0.1617, 0.1557, 0.2024, 0.2071, 0.2186, 0.2144, 0.2072, 0.1765, 0.2029, 0.2238, 0.2221, 0.245, 0.2362, 0.2627, 0.2529, 0.2236, 0.2679, 0.2556, 0.2638, 0.2733, 0.2869, 0.2824, 0.2901, 0.2791, 0.2821, 0.271, 0.282, 0.2372, 0.2355, 0.2325, 0.2537, 0.2665, 0.2803, 0.2791, 0.2691, 0.2269, 0.2269, 0.2029, 0.2018, 0.2091, 0.1764, 0.1855, 0.2028, 0.2491, 0.248, 0.2526, 0.2521, 0.2319, 0.1995, 0.2415, 0.2445, 0.2934, 0.2569, 0.2466, 0.2517, 0.2552, 0.281, 0.3143, 0.3189, 0.3306, 0.3216, 0.307, 0.3034, 0.3321, 0.3565, 0.2902, 0.3266, 0.2682, 0.2667, 0.2807, 0.2906, 0.2865, 0.2761, 0.2699, 0.2832, 0.312, 0.3489, 0.4161, 0.4584, 0.4292, 0.4264, 0.4573, 0.3996, 0.3751, 0.3249, 0.3254, 0.4021, 0.3606, 0.3652, 0.3163, 0.3396, 0.3144, 0.2775, 0.3139, 0.2725, 0.2601, 0.2859, 0.2526, 0.2698, 0.2698, 0.2468, 0.2674, 0.2782, 0.2651, 0.2918, 0.285, 0.3071]}
{"unit_id": "orig:1933343961730724035", "ticker_idx": 1, "ts": "2025-06-13T02:01:34+00:00", "symbol": "CLS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CLS 800 G shipments nearly triple QoQ... Its AI-networking gear is also showing eye-catching momentum across the industry", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi note: long NVDA and CLS as 800G leaders, cautious CSCO on minimal AI backend exposure, ANET stable with 800G acceleration in 2H25.", "resolved_tickers": ["CLS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "US Communications Equipment: Ethernet Accelerates Replacement of IB, 800G Deployment Shows Strong Growth, NVDA and CLS Lead, ANET and CSCO Diverge\n\nCiti (A. Malik, 25/06/12)\n\nEthernet’s pace in displacing InfiniBand in the data-center switch market has exceeded expectations, lifting 2025 switch-market growth from 14 % to 23 %. In the 800 G era, market-share reshuffling highlights NVDA and CLS as the most striking performers.\n\n• Ethernet’s replacement of InfiniBand drives accelerated switch growth\n\nResearch indicates that Ethernet’s displacement of InfiniBand has surpassed Dell’Oro’s earlier expectations, resulting in significantly better-than-forecast market growth in 1Q25. Dell’Oro has raised its 2025 data-center-switch growth forecast from 14 % to 23 %.\n\n• Arista backend deployment stable; 800 G to accelerate in 2H25\n\nArista’s backend-switch deployment is estimated at US$1.6 billion, about an 11 % market share, flat versus FY24. Although 800 G shipments are “still slowly ramping up,” they are expected to accelerate markedly in 2H25, with enterprise networks standing out thanks to competitive displacement and share gains.\n\n• NVDA leads 800 G port growth of over 100 %\n\nNVDA has become the sales champion for 800 G ports and equipment, with port shipments surging over 100 % QoQ. Drivers include Blackwell server read acceleration (about 70 % of DC compute sales) and the migration from InfiniBand to Ethernet. Its SpectrumX platform has seen significant revenue uplift from large wins with xAI, Azure, Oracle, Google Cloud, and CoreWeave.\n\n• Cisco frontend networking solid; backend AI deployment still modest\n\nCisco’s data-center-switch sales are driven mainly by Tier-2/3 cloud providers and enterprise frontend networks. AI backend-switch shipments are “not yet significant,” as related AI-networking products are still booked under the router segment, making large near-term growth unlikely.\n\n• CLS 800 G shipments nearly triple QoQ\n\nCLS (Celestica) saw May 800 G equipment shipments nearly triple QoQ, driven by accelerated deployments at Google and AMZN and Meta’s first large-scale adoption. Its AI-networking gear is also showing eye-catching momentum across the industry.\n\n$NVDA $CLS $ANET $CSCO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.043086202964902176, "ret_signed_m1d": 0.043086202964902176, "alpha_spy_m1d": 0.03177973148924007, "alpha_c_m1d": 0.028870389338379265, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.01421581362652291, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02321873150558207, "ret_signed_p1d": 0.02321873150558207, "alpha_spy_p1d": 0.013704535026866749, "alpha_c_p1d": 0.007037845806418863, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.016180885699163206, "ret_p1w": 0.07899141268528509, "ret_signed_p1w": 0.07899141268528509, "alpha_spy_p1w": 0.08060468768109441, "alpha_c_p1w": 0.07326334244624588, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.005728070239039207, "ret_p1m": 0.29450253828180895, "ret_signed_p1m": 0.29450253828180895, "alpha_spy_p1m": 0.24482574651609856, "alpha_c_p1m": 0.22386793087691492, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.07063460740489402, "ret_p3m": 1.0189100381597558, "ret_signed_p3m": 1.0189100381597558, "alpha_spy_p3m": 0.9232012917398851, "alpha_c_p3m": 0.8875067992235843, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.13140323893617145, "ret_p6m": 1.7189021366565589, "ret_signed_p6m": 1.7189021366565589, "alpha_spy_p6m": 1.567223884970562, "alpha_c_p6m": 1.4807476053547368, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.23815453130182207, "price_path": [-0.3021, -0.2666, -0.2368, -0.2344, -0.2023, -0.2231, -0.3006, -0.3267, -0.3402, -0.3712, -0.3692, -0.3429, -0.4479, -0.4722, -0.4384, -0.4507, -0.3452, -0.3898, -0.3686, -0.3547, -0.3412, -0.3572, -0.3295, -0.3295, -0.357, -0.3396, -0.2963, -0.2648, -0.2895, -0.2998, -0.3054, -0.319, -0.2882, -0.2494, -0.2631, -0.2405, -0.2296, -0.2314, -0.2576, -0.1996, -0.1234, -0.0948, -0.1104, -0.0964, -0.1096, -0.1011, -0.0902, -0.0559, -0.0468, -0.0468, -0.0515, -0.0575, -0.0715, -0.0794, -0.0693, -0.0325, -0.0433, -0.048, -0.0367, -0.0214, -0.0305, 0.0188, 0.0431, 0.0, 0.0232, 0.0547, 0.0871, 0.0871, 0.079, 0.0878, 0.1477, 0.1892, 0.198, 0.2181, 0.2456, 0.1881, 0.2359, 0.2707, 0.2707, 0.2288, 0.2423, 0.2838, 0.2673, 0.2848, 0.2945, 0.2805, 0.2774, 0.2997, 0.2776, 0.3007, 0.252, 0.3468, 0.3084, 0.3582, 0.3833, 0.6117, 0.615, 0.5947, 0.5538, 0.6074, 0.5679, 0.5849, 0.6054, 0.6569, 0.6325, 0.7031, 0.6295, 0.569, 0.5597, 0.5549, 0.4739, 0.4593, 0.4469, 0.5074, 0.5392, 0.5848, 0.6125, 0.6942, 0.5539, 0.5539, 0.6008, 0.6905, 0.7661, 0.9363, 0.9495, 0.9475, 1.0189, 0.9607, 0.9291, 0.9867, 1.0298, 0.9761, 1.011, 1.0181, 1.033, 1.0557, 0.9186, 0.952, 0.9452, 0.9624, 0.9659, 1.0082, 1.002, 0.8644, 0.8778, 0.8973, 1.0337, 1.0728, 0.9446, 1.083, 0.9768, 1.154, 1.233, 1.2059, 1.1822, 1.1119, 1.1614, 1.2559, 1.3667, 1.4082, 1.6047, 1.695, 1.7139, 1.7486, 1.7847, 1.6792, 1.8135, 1.7189, 1.57, 1.7496, 1.6479, 1.6695, 1.3358, 1.4805, 1.4684, 1.393, 1.4933, 1.2523, 1.2346, 1.5735, 1.6151, 1.6509, 1.6509, 1.748, 1.5403, 1.4279, 1.4608, 1.5755, 1.5939, 1.7189]}
{"unit_id": "orig:1933343961730724035", "ticker_idx": 2, "ts": "2025-06-13T02:01:34+00:00", "symbol": "ANET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Arista backend deployment stable; 800 G to accelerate in 2H25... expected to accelerate markedly in 2H25, with enterprise networks standing out thanks to competitive displacement and share gains", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi note: long NVDA and CLS as 800G leaders, cautious CSCO on minimal AI backend exposure, ANET stable with 800G acceleration in 2H25.", "resolved_tickers": ["ANET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "US Communications Equipment: Ethernet Accelerates Replacement of IB, 800G Deployment Shows Strong Growth, NVDA and CLS Lead, ANET and CSCO Diverge\n\nCiti (A. Malik, 25/06/12)\n\nEthernet’s pace in displacing InfiniBand in the data-center switch market has exceeded expectations, lifting 2025 switch-market growth from 14 % to 23 %. In the 800 G era, market-share reshuffling highlights NVDA and CLS as the most striking performers.\n\n• Ethernet’s replacement of InfiniBand drives accelerated switch growth\n\nResearch indicates that Ethernet’s displacement of InfiniBand has surpassed Dell’Oro’s earlier expectations, resulting in significantly better-than-forecast market growth in 1Q25. Dell’Oro has raised its 2025 data-center-switch growth forecast from 14 % to 23 %.\n\n• Arista backend deployment stable; 800 G to accelerate in 2H25\n\nArista’s backend-switch deployment is estimated at US$1.6 billion, about an 11 % market share, flat versus FY24. Although 800 G shipments are “still slowly ramping up,” they are expected to accelerate markedly in 2H25, with enterprise networks standing out thanks to competitive displacement and share gains.\n\n• NVDA leads 800 G port growth of over 100 %\n\nNVDA has become the sales champion for 800 G ports and equipment, with port shipments surging over 100 % QoQ. Drivers include Blackwell server read acceleration (about 70 % of DC compute sales) and the migration from InfiniBand to Ethernet. Its SpectrumX platform has seen significant revenue uplift from large wins with xAI, Azure, Oracle, Google Cloud, and CoreWeave.\n\n• Cisco frontend networking solid; backend AI deployment still modest\n\nCisco’s data-center-switch sales are driven mainly by Tier-2/3 cloud providers and enterprise frontend networks. AI backend-switch shipments are “not yet significant,” as related AI-networking products are still booked under the router segment, making large near-term growth unlikely.\n\n• CLS 800 G shipments nearly triple QoQ\n\nCLS (Celestica) saw May 800 G equipment shipments nearly triple QoQ, driven by accelerated deployments at Google and AMZN and Meta’s first large-scale adoption. Its AI-networking gear is also showing eye-catching momentum across the industry.\n\n$NVDA $CLS $ANET $CSCO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03703300731405745, "ret_signed_m1d": 0.03703300731405745, "alpha_spy_m1d": 0.025726535838395348, "alpha_c_m1d": 0.022817193687534543, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.01421581362652291, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029669712063258524, "ret_signed_p1d": 0.029669712063258524, "alpha_spy_p1d": 0.020155515584543204, "alpha_c_p1d": 0.013488826364095319, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.016180885699163206, "ret_p1w": -0.06605304358321651, "ret_signed_p1w": -0.06605304358321651, "alpha_spy_p1w": -0.06443976858740719, "alpha_c_p1w": -0.07178111382225572, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.005728070239039207, "ret_p1m": 0.1734705418208553, "ret_signed_p1m": 0.1734705418208553, "alpha_spy_p1m": 0.12379375005514492, "alpha_c_p1m": 0.10283593441596128, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.07063460740489402, "ret_p3m": 0.6320520163618502, "ret_signed_p3m": 0.6320520163618502, "alpha_spy_p3m": 0.5363432699419795, "alpha_c_p3m": 0.5006487774256787, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.13140323893617145, "ret_p6m": 0.3980509230493554, "ret_signed_p6m": 0.3980509230493554, "alpha_spy_p6m": 0.2463726713633585, "alpha_c_p6m": 0.15989639174753334, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.23815453130182207, "price_path": [-0.0905, -0.0859, -0.1016, -0.0998, -0.0524, -0.0586, -0.1158, -0.145, -0.156, -0.161, -0.1501, -0.132, -0.2283, -0.303, -0.2618, -0.2474, -0.177, -0.2133, -0.2131, -0.2031, -0.2074, -0.2218, -0.229, -0.229, -0.2672, -0.2564, -0.2338, -0.1847, -0.1564, -0.152, -0.1258, -0.1091, -0.0488, -0.0144, -0.0213, -0.0171, -0.0639, -0.055, -0.0631, -0.0003, 0.0531, 0.0636, 0.0403, 0.0441, 0.0471, 0.036, 0.0009, 0.0023, -0.0125, -0.0125, 0.0015, 0.0048, -0.0648, -0.0618, -0.0278, 0.0233, 0.028, 0.0306, 0.0531, 0.0482, 0.0146, 0.0201, 0.037, 0.0, 0.0297, -0.0289, -0.0228, -0.0228, -0.0661, -0.0043, 0.0284, 0.0429, 0.1001, 0.0762, 0.1079, 0.071, 0.0951, 0.1101, 0.1101, 0.0988, 0.1195, 0.1508, 0.1509, 0.1756, 0.1735, 0.1626, 0.1727, 0.2126, 0.2104, 0.2086, 0.1887, 0.224, 0.2349, 0.2375, 0.2729, 0.2845, 0.322, 0.3343, 0.2731, 0.3032, 0.279, 0.5028, 0.5082, 0.5071, 0.4905, 0.5295, 0.4944, 0.4779, 0.4867, 0.4947, 0.4378, 0.4236, 0.4297, 0.4429, 0.4406, 0.4539, 0.4431, 0.4751, 0.4786, 0.4786, 0.4713, 0.4876, 0.5286, 0.5468, 0.5161, 0.5367, 0.6321, 0.6572, 0.5094, 0.5748, 0.5394, 0.5467, 0.5881, 0.62, 0.5744, 0.5603, 0.5446, 0.5491, 0.543, 0.5525, 0.5778, 0.6164, 0.5643, 0.5755, 0.6188, 0.5733, 0.704, 0.7134, 0.6687, 0.5966, 0.5029, 0.5526, 0.5811, 0.5495, 0.5861, 0.5803, 0.5873, 0.6541, 0.6656, 0.698, 0.6976, 0.7545, 0.7156, 0.7075, 0.7064, 0.6627, 0.5205, 0.4512, 0.458, 0.4863, 0.4611, 0.4616, 0.4109, 0.4225, 0.378, 0.3368, 0.3515, 0.295, 0.2716, 0.3229, 0.354, 0.3822, 0.3822, 0.4151, 0.3872, 0.3776, 0.3839, 0.392, 0.3924, 0.3981]}
{"unit_id": "orig:1933343961730724035", "ticker_idx": 3, "ts": "2025-06-13T02:01:34+00:00", "symbol": "CSCO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI backend-switch shipments are 'not yet significant,' as related AI-networking products are still booked under the router segment, making large near-term growth unlikely", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi note: long NVDA and CLS as 800G leaders, cautious CSCO on minimal AI backend exposure, ANET stable with 800G acceleration in 2H25.", "resolved_tickers": ["CSCO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "US Communications Equipment: Ethernet Accelerates Replacement of IB, 800G Deployment Shows Strong Growth, NVDA and CLS Lead, ANET and CSCO Diverge\n\nCiti (A. Malik, 25/06/12)\n\nEthernet’s pace in displacing InfiniBand in the data-center switch market has exceeded expectations, lifting 2025 switch-market growth from 14 % to 23 %. In the 800 G era, market-share reshuffling highlights NVDA and CLS as the most striking performers.\n\n• Ethernet’s replacement of InfiniBand drives accelerated switch growth\n\nResearch indicates that Ethernet’s displacement of InfiniBand has surpassed Dell’Oro’s earlier expectations, resulting in significantly better-than-forecast market growth in 1Q25. Dell’Oro has raised its 2025 data-center-switch growth forecast from 14 % to 23 %.\n\n• Arista backend deployment stable; 800 G to accelerate in 2H25\n\nArista’s backend-switch deployment is estimated at US$1.6 billion, about an 11 % market share, flat versus FY24. Although 800 G shipments are “still slowly ramping up,” they are expected to accelerate markedly in 2H25, with enterprise networks standing out thanks to competitive displacement and share gains.\n\n• NVDA leads 800 G port growth of over 100 %\n\nNVDA has become the sales champion for 800 G ports and equipment, with port shipments surging over 100 % QoQ. Drivers include Blackwell server read acceleration (about 70 % of DC compute sales) and the migration from InfiniBand to Ethernet. Its SpectrumX platform has seen significant revenue uplift from large wins with xAI, Azure, Oracle, Google Cloud, and CoreWeave.\n\n• Cisco frontend networking solid; backend AI deployment still modest\n\nCisco’s data-center-switch sales are driven mainly by Tier-2/3 cloud providers and enterprise frontend networks. AI backend-switch shipments are “not yet significant,” as related AI-networking products are still booked under the router segment, making large near-term growth unlikely.\n\n• CLS 800 G shipments nearly triple QoQ\n\nCLS (Celestica) saw May 800 G equipment shipments nearly triple QoQ, driven by accelerated deployments at Google and AMZN and Meta’s first large-scale adoption. Its AI-networking gear is also showing eye-catching momentum across the industry.\n\n$NVDA $CLS $ANET $CSCO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015759251255480278, "ret_signed_m1d": -0.015759251255480278, "alpha_spy_m1d": -0.004452779779818172, "alpha_c_m1d": -0.004452779779818172, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.011306471475662105, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02215651681278663, "ret_signed_p1d": -0.02215651681278663, "alpha_spy_p1d": -0.01264232033407131, "alpha_c_p1d": -0.01264232033407131, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.00951419647871532, "ret_p1w": 0.034794801714056245, "ret_signed_p1w": -0.034794801714056245, "alpha_spy_p1w": -0.03640807670986557, "alpha_c_p1w": -0.03640807670986557, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": -0.0016132749958093218, "ret_p1m": 0.0645630543589466, "ret_signed_p1m": -0.0645630543589466, "alpha_spy_p1m": -0.014886262593236221, "alpha_c_p1m": -0.014886262593236221, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.04967679176571038, "ret_p3m": 0.06942909484729709, "ret_signed_p3m": -0.06942909484729709, "alpha_spy_p3m": 0.026279651572573615, "alpha_c_p3m": 0.026279651572573615, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.0957087464198707, "ret_p6m": 0.24533115489820911, "ret_signed_p6m": -0.24533115489820911, "alpha_spy_p6m": -0.0936529032122122, "alpha_c_p6m": -0.0936529032122122, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.1516782516859969, "price_path": [-0.0654, -0.0527, -0.0604, -0.0654, -0.055, -0.0547, -0.0421, -0.0483, -0.0567, -0.0435, -0.0421, -0.0418, -0.1058, -0.149, -0.151, -0.1701, -0.093, -0.117, -0.1055, -0.1042, -0.1075, -0.13, -0.13, -0.13, -0.1517, -0.1412, -0.132, -0.1217, -0.1152, -0.1131, -0.1053, -0.0992, -0.0932, -0.0743, -0.0744, -0.076, -0.0705, -0.0683, -0.0674, -0.0378, -0.036, -0.0437, 0.0027, -0.0073, -0.0027, -0.0105, -0.0139, -0.0114, -0.0153, -0.0153, -0.0053, -0.0117, -0.0162, -0.0164, -0.0037, 0.0042, 0.0047, 0.0083, 0.0307, 0.0282, 0.0165, 0.0016, 0.0158, 0.0, 0.0222, 0.0193, 0.0273, 0.0273, 0.0348, 0.0513, 0.0593, 0.064, 0.0758, 0.0711, 0.0825, 0.0782, 0.0702, 0.0889, 0.0889, 0.082, 0.0766, 0.0873, 0.0793, 0.0666, 0.0646, 0.0545, 0.0575, 0.0721, 0.0682, 0.0727, 0.0699, 0.0771, 0.0719, 0.0782, 0.0661, 0.0668, 0.0718, 0.0686, 0.0534, 0.0751, 0.0599, 0.0864, 0.0972, 0.1269, 0.1093, 0.1204, 0.1051, 0.0878, 0.0391, 0.0509, 0.0479, 0.0534, 0.052, 0.0567, 0.0539, 0.0735, 0.0743, 0.0898, 0.0845, 0.0845, 0.0642, 0.0622, 0.0672, 0.0501, 0.0498, 0.057, 0.0694, 0.0625, 0.0443, 0.052, 0.0506, 0.063, 0.0781, 0.0707, 0.0635, 0.0602, 0.0567, 0.065, 0.0551, 0.063, 0.074, 0.0792, 0.0723, 0.0726, 0.0882, 0.0895, 0.1106, 0.1048, 0.0729, 0.0653, 0.0843, 0.0978, 0.091, 0.1075, 0.1157, 0.1168, 0.1158, 0.1097, 0.1154, 0.1274, 0.1468, 0.1264, 0.1514, 0.1545, 0.1757, 0.1421, 0.1386, 0.1218, 0.1223, 0.1384, 0.1324, 0.168, 0.222, 0.2318, 0.2283, 0.2218, 0.2379, 0.1913, 0.2017, 0.204, 0.2052, 0.2013, 0.2013, 0.215, 0.2008, 0.2139, 0.2278, 0.228, 0.2313, 0.2453]}
{"unit_id": "orig:1930079711650284029", "ticker_idx": 0, "ts": "2025-06-04T01:50:37+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "multiple drivers — including further share gains at TSMC and in DRAM, as well as Intel-related upside — are expected to support both KLA's valuation premium and earnings growth", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS raises WFE outlook but stays neutral on SPE sector; bullish KLA on TSMC/DRAM share gains + Intel upside into 2026; AMAT/ASML face 45-50% earnings downside risk from full export ban.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: Forecast Raised on China Spend, but Industry Faces Continued Headwinds\nUBS (T. Arcuri, June 2, 2025)\n\nKLA, despite facing concerns about potential growth slowdown in 2H25 due to China’s domestic substitution push, has gained market recognition for its continued share gains in TSMC, DRAM, and advanced packaging.\nMoreover, Intel’s transition from post-test inspection to in-line inspection could present additional growth opportunities for KLA.\n\nLooking ahead to 2026, multiple drivers — including further share gains at TSMC and in DRAM, as well as Intel-related upside — are expected to support both KLA’s valuation premium and earnings growth.\n\n⸻\n\n• Global WFE Forecast Raised as China Spending Declines Less Than Expected:\n• Based on recent updates from chipmakers and equipment vendors, UBS has revised its 2025/2026 global WFE (wafer fab equipment) outlook upward, mainly due to stronger-than-expected domestic project activity in China.\n• China’s WFE capex is now expected to reach $33 billion in 2025 (YoY -8%), with a rebound anticipated in 2026.\n• Globally, WFE spending is forecast at $107 billion in 2025 (+10%) and $110 billion in 2026 (+3%).\n• While annual data points to moderate growth, quarterly trends remain too soft to justify a major short-term upgrade, so UBS maintains a neutral rating on the semiconductor production equipment (SPE) sector.\n\n⸻\n\n• 2026 Unlikely to Be a Down Year for WFE, but Geopolitical Risk Remains Central:\n• Despite some investor concerns about a WFE downturn in 2026, UBS sees such forecasts as overly conservative.\n• The market currently expects system revenue for key equipment makers (e.g., AMAT, ASML) to grow by ~7% YoY in 2026.\n• However, if U.S.–China trade tensions escalate into a full export ban,\n→ U.S. equipment vendors could face earnings downside of up to 45–50%.\n• Geopolitics remains the single greatest source of uncertainty for the industry.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008055323305087625, "ret_signed_m1d": -0.008055323305087625, "alpha_spy_m1d": -0.008323942845133936, "alpha_c_m1d": 0.0037023834845723025, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012530686188534812, "ret_signed_p1d": 0.012530686188534812, "alpha_spy_p1d": 0.017363455244724046, "alpha_c_p1d": 0.014318190801695052, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.11496119333584942, "ret_signed_p1w": 0.11496119333584942, "alpha_spy_p1w": 0.10584926217372437, "alpha_c_p1w": 0.07472288370721869, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.1821913875635668, "ret_signed_p1m": 0.1821913875635668, "alpha_spy_p1m": 0.1297376469976239, "alpha_c_p1m": 0.055597514083629784, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.11738765546327068, "ret_signed_p3m": 0.11738765546327068, "alpha_spy_p3m": 0.03176183616331407, "alpha_c_p3m": -0.035700805951253356, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.487733496809736, "ret_signed_p6m": 0.487733496809736, "alpha_spy_p6m": 0.3406535766435048, "alpha_c_p6m": 0.10656175713607108, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.0927, -0.1347, -0.1528, -0.1321, -0.1307, -0.0905, -0.0838, -0.0944, -0.0821, -0.0861, -0.0988, -0.0717, -0.0792, -0.1, -0.1071, -0.1408, -0.1329, -0.1295, -0.1247, -0.2081, -0.2646, -0.2288, -0.2353, -0.1029, -0.1623, -0.1451, -0.1454, -0.1375, -0.1809, -0.1907, -0.1907, -0.2081, -0.1883, -0.1618, -0.1206, -0.114, -0.1151, -0.1225, -0.1037, -0.1373, -0.1095, -0.1157, -0.1334, -0.1143, -0.1029, -0.1056, -0.03, 0.0152, 0.0268, 0.0281, 0.0092, 0.0063, 0.0105, -0.004, -0.0186, -0.0319, -0.0319, 0.0089, -0.0059, -0.0104, -0.0322, -0.0251, -0.0081, 0.0, 0.0125, 0.0331, 0.0604, 0.0953, 0.115, 0.1188, 0.1094, 0.141, 0.1418, 0.1139, 0.1139, 0.0868, 0.0949, 0.1367, 0.1424, 0.1545, 0.1378, 0.1453, 0.1493, 0.1777, 0.1822, 0.1822, 0.1669, 0.1753, 0.1804, 0.1874, 0.1822, 0.1787, 0.1975, 0.1936, 0.1982, 0.1906, 0.199, 0.1408, 0.147, 0.1561, 0.1534, 0.1804, 0.1713, 0.1827, 0.124, 0.1337, 0.1707, 0.1296, 0.1358, 0.1662, 0.1697, 0.1638, 0.1962, 0.214, 0.2216, 0.1187, 0.132, 0.1226, 0.1256, 0.1179, 0.1152, 0.1271, 0.138, 0.139, 0.1456, 0.1174, 0.1174, 0.0845, 0.0814, 0.119, 0.1598, 0.1648, 0.176, 0.1951, 0.2292, 0.2353, 0.2672, 0.2693, 0.2684, 0.3412, 0.3388, 0.3727, 0.3726, 0.3694, 0.3571, 0.3638, 0.3636, 0.3821, 0.4465, 0.4599, 0.4115, 0.4604, 0.39, 0.3616, 0.3499, 0.2593, 0.3134, 0.3144, 0.3929, 0.408, 0.4181, 0.4773, 0.4703, 0.4279, 0.4852, 0.5157, 0.5571, 0.5454, 0.5829, 0.5562, 0.5489, 0.5622, 0.5293, 0.5724, 0.5459, 0.5292, 0.5607, 0.526, 0.5364, 0.4886, 0.4535, 0.4552, 0.4416, 0.4985, 0.4151, 0.4082, 0.4591, 0.4708, 0.4877, 0.4877]}
{"unit_id": "orig:1930079711650284029", "ticker_idx": 1, "ts": "2025-06-04T01:50:37+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": "equipment", "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "UBS maintains a neutral rating on the semiconductor production equipment (SPE) sector", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS raises WFE outlook but stays neutral on SPE sector; bullish KLA on TSMC/DRAM share gains + Intel upside into 2026; AMAT/ASML face 45-50% earnings downside risk from full export ban.", "resolved_tickers": ["AMAT", "LRCX", "KLAC"], "resolution_method": "basket", "resolution_reason": "US semiconductor equipment sub-sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: Forecast Raised on China Spend, but Industry Faces Continued Headwinds\nUBS (T. Arcuri, June 2, 2025)\n\nKLA, despite facing concerns about potential growth slowdown in 2H25 due to China’s domestic substitution push, has gained market recognition for its continued share gains in TSMC, DRAM, and advanced packaging.\nMoreover, Intel’s transition from post-test inspection to in-line inspection could present additional growth opportunities for KLA.\n\nLooking ahead to 2026, multiple drivers — including further share gains at TSMC and in DRAM, as well as Intel-related upside — are expected to support both KLA’s valuation premium and earnings growth.\n\n⸻\n\n• Global WFE Forecast Raised as China Spending Declines Less Than Expected:\n• Based on recent updates from chipmakers and equipment vendors, UBS has revised its 2025/2026 global WFE (wafer fab equipment) outlook upward, mainly due to stronger-than-expected domestic project activity in China.\n• China’s WFE capex is now expected to reach $33 billion in 2025 (YoY -8%), with a rebound anticipated in 2026.\n• Globally, WFE spending is forecast at $107 billion in 2025 (+10%) and $110 billion in 2026 (+3%).\n• While annual data points to moderate growth, quarterly trends remain too soft to justify a major short-term upgrade, so UBS maintains a neutral rating on the semiconductor production equipment (SPE) sector.\n\n⸻\n\n• 2026 Unlikely to Be a Down Year for WFE, but Geopolitical Risk Remains Central:\n• Despite some investor concerns about a WFE downturn in 2026, UBS sees such forecasts as overly conservative.\n• The market currently expects system revenue for key equipment makers (e.g., AMAT, ASML) to grow by ~7% YoY in 2026.\n• However, if U.S.–China trade tensions escalate into a full export ban,\n→ U.S. equipment vendors could face earnings downside of up to 45–50%.\n• Geopolitics remains the single greatest source of uncertainty for the industry.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007440950229617327, "ret_signed_m1d": 0.007440950229617327, "alpha_spy_m1d": 0.007709569769663638, "alpha_c_m1d": -0.0043167565600426006, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00937963886000559, "ret_signed_p1d": -0.00937963886000559, "alpha_spy_p1d": -0.014212407916194825, "alpha_c_p1d": -0.01116714347316583, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.08497681973739664, "ret_signed_p1w": -0.08497681973739664, "alpha_spy_p1w": -0.07586488857527159, "alpha_c_p1w": -0.04473851010876591, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.1768491748326606, "ret_signed_p1m": -0.1768491748326606, "alpha_spy_p1m": -0.12439543426671773, "alpha_c_p1m": -0.0502553013527236, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.09912656954673997, "ret_signed_p3m": -0.09912656954673997, "alpha_spy_p3m": -0.013500750246783363, "alpha_c_p3m": 0.053961891867784065, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.6257299147471799, "ret_signed_p6m": -0.6257299147471799, "alpha_spy_p6m": -0.4786499945809487, "alpha_c_p6m": -0.24455817507351496, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.0673, -0.1135, -0.1274, -0.1036, -0.1065, -0.0679, -0.0676, -0.0788, -0.0705, -0.0735, -0.0891, -0.0658, -0.0723, -0.0913, -0.1051, -0.1303, -0.1272, -0.1245, -0.1148, -0.2014, -0.2619, -0.2256, -0.2435, -0.114, -0.1758, -0.1522, -0.1516, -0.1448, -0.1871, -0.1974, -0.1974, -0.2125, -0.1959, -0.1635, -0.1198, -0.1127, -0.1141, -0.1217, -0.11, -0.1246, -0.0918, -0.0972, -0.1093, -0.0898, -0.0862, -0.0861, -0.0084, 0.0288, 0.0338, 0.0348, 0.0082, 0.0056, 0.0085, -0.0095, -0.0178, -0.0343, -0.0343, -0.0005, -0.0057, -0.0109, -0.0371, -0.027, -0.0074, 0.0, 0.0094, 0.0272, 0.0502, 0.0806, 0.085, 0.0936, 0.073, 0.1111, 0.1033, 0.0907, 0.0907, 0.0678, 0.0801, 0.1268, 0.1362, 0.1444, 0.1396, 0.1423, 0.143, 0.1733, 0.1768, 0.1768, 0.1686, 0.1867, 0.1891, 0.2018, 0.2025, 0.1913, 0.2078, 0.1945, 0.193, 0.1857, 0.1972, 0.1506, 0.1501, 0.1581, 0.1489, 0.1739, 0.1683, 0.1747, 0.1192, 0.1283, 0.1545, 0.1264, 0.1235, 0.1566, 0.1715, 0.1696, 0.2017, 0.2166, 0.218, 0.0982, 0.1037, 0.1036, 0.0974, 0.0905, 0.1017, 0.1093, 0.1275, 0.1277, 0.1334, 0.0991, 0.0991, 0.0693, 0.0683, 0.0955, 0.1284, 0.137, 0.1457, 0.1589, 0.2166, 0.2192, 0.2452, 0.2563, 0.2711, 0.3368, 0.339, 0.3927, 0.3923, 0.3792, 0.3705, 0.3824, 0.3954, 0.4122, 0.4957, 0.5288, 0.4955, 0.5382, 0.4544, 0.4659, 0.4617, 0.3721, 0.4352, 0.4348, 0.506, 0.5018, 0.4961, 0.5324, 0.5295, 0.4892, 0.5495, 0.5766, 0.6163, 0.5997, 0.6489, 0.6347, 0.6194, 0.6483, 0.6003, 0.6733, 0.638, 0.6141, 0.6627, 0.6095, 0.626, 0.5626, 0.5367, 0.5396, 0.511, 0.5726, 0.4786, 0.4962, 0.5579, 0.5919, 0.6257, 0.6257]}
{"unit_id": "orig:1930079711650284029", "ticker_idx": 2, "ts": "2025-06-04T01:50:37+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if U.S.–China trade tensions escalate into a full export ban, U.S. equipment vendors could face earnings downside of up to 45–50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS raises WFE outlook but stays neutral on SPE sector; bullish KLA on TSMC/DRAM share gains + Intel upside into 2026; AMAT/ASML face 45-50% earnings downside risk from full export ban.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: Forecast Raised on China Spend, but Industry Faces Continued Headwinds\nUBS (T. Arcuri, June 2, 2025)\n\nKLA, despite facing concerns about potential growth slowdown in 2H25 due to China’s domestic substitution push, has gained market recognition for its continued share gains in TSMC, DRAM, and advanced packaging.\nMoreover, Intel’s transition from post-test inspection to in-line inspection could present additional growth opportunities for KLA.\n\nLooking ahead to 2026, multiple drivers — including further share gains at TSMC and in DRAM, as well as Intel-related upside — are expected to support both KLA’s valuation premium and earnings growth.\n\n⸻\n\n• Global WFE Forecast Raised as China Spending Declines Less Than Expected:\n• Based on recent updates from chipmakers and equipment vendors, UBS has revised its 2025/2026 global WFE (wafer fab equipment) outlook upward, mainly due to stronger-than-expected domestic project activity in China.\n• China’s WFE capex is now expected to reach $33 billion in 2025 (YoY -8%), with a rebound anticipated in 2026.\n• Globally, WFE spending is forecast at $107 billion in 2025 (+10%) and $110 billion in 2026 (+3%).\n• While annual data points to moderate growth, quarterly trends remain too soft to justify a major short-term upgrade, so UBS maintains a neutral rating on the semiconductor production equipment (SPE) sector.\n\n⸻\n\n• 2026 Unlikely to Be a Down Year for WFE, but Geopolitical Risk Remains Central:\n• Despite some investor concerns about a WFE downturn in 2026, UBS sees such forecasts as overly conservative.\n• The market currently expects system revenue for key equipment makers (e.g., AMAT, ASML) to grow by ~7% YoY in 2026.\n• However, if U.S.–China trade tensions escalate into a full export ban,\n→ U.S. equipment vendors could face earnings downside of up to 45–50%.\n• Geopolitics remains the single greatest source of uncertainty for the industry.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0011733312374548888, "ret_signed_m1d": 0.0011733312374548888, "alpha_spy_m1d": 0.0014419507775011997, "alpha_c_m1d": -0.010584375552205039, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01395657620786861, "ret_signed_p1d": -0.01395657620786861, "alpha_spy_p1d": -0.018789345264057844, "alpha_c_p1d": -0.01574408082102885, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.06706602263966399, "ret_signed_p1w": -0.06706602263966399, "alpha_spy_p1w": -0.05795409147753894, "alpha_c_p1w": -0.02682771301103326, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.17983092867349693, "ret_signed_p1m": -0.17983092867349693, "alpha_spy_p1m": -0.12737718810755405, "alpha_c_p1m": -0.05323705519355992, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": -0.004379987322213452, "ret_signed_p3m": 0.004379987322213452, "alpha_spy_p3m": 0.09000580662217006, "alpha_c_p3m": 0.1574684487367375, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.5511506353796056, "ret_signed_p6m": -0.5511506353796056, "alpha_spy_p6m": -0.40407071521337445, "alpha_c_p6m": -0.1699788957059407, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.0413, -0.0764, -0.0948, -0.0821, -0.0766, -0.0414, -0.0427, -0.0547, -0.0488, -0.0493, -0.0632, -0.0458, -0.0539, -0.0721, -0.0906, -0.1067, -0.1064, -0.103, -0.0902, -0.1655, -0.2182, -0.1819, -0.2059, -0.078, -0.1487, -0.1075, -0.11, -0.1043, -0.149, -0.1535, -0.1535, -0.1652, -0.1483, -0.1158, -0.0753, -0.0668, -0.0714, -0.0791, -0.0719, -0.0827, -0.0449, -0.0479, -0.0576, -0.0405, -0.0394, -0.0418, 0.0345, 0.0655, 0.0724, 0.0761, 0.0196, 0.0219, 0.0206, -0.001, -0.0087, -0.0273, -0.0273, -0.0006, -0.0021, -0.0151, -0.032, -0.0288, -0.0012, 0.0, 0.014, 0.0297, 0.0485, 0.0731, 0.0671, 0.0807, 0.0535, 0.0903, 0.0751, 0.0674, 0.0674, 0.0465, 0.0619, 0.1127, 0.1306, 0.1333, 0.1314, 0.1306, 0.1348, 0.1734, 0.1798, 0.1798, 0.1782, 0.2042, 0.2066, 0.2229, 0.2223, 0.2172, 0.2307, 0.2031, 0.1889, 0.1761, 0.1895, 0.1557, 0.1549, 0.1617, 0.1467, 0.175, 0.1635, 0.1696, 0.112, 0.1115, 0.129, 0.1063, 0.1001, 0.131, 0.1417, 0.1386, 0.1638, 0.1735, 0.1625, -0.001, 0.0099, 0.0018, -0.006, -0.0101, 0.0063, 0.0032, 0.0188, 0.0181, 0.0236, -0.0044, -0.0044, -0.0241, -0.0323, -0.02, 0.0079, 0.0036, 0.0126, 0.0121, 0.0538, 0.0392, 0.0586, 0.0748, 0.1032, 0.1752, 0.1773, 0.2419, 0.244, 0.2476, 0.2362, 0.2629, 0.2693, 0.268, 0.3485, 0.3847, 0.3472, 0.3867, 0.3102, 0.3471, 0.3644, 0.3003, 0.3593, 0.3513, 0.4095, 0.4103, 0.3934, 0.4129, 0.3997, 0.366, 0.415, 0.4167, 0.4327, 0.4098, 0.46, 0.4402, 0.4436, 0.4722, 0.4256, 0.4919, 0.4463, 0.4249, 0.4559, 0.4162, 0.429, 0.3825, 0.3997, 0.4164, 0.3942, 0.4562, 0.3666, 0.3901, 0.4329, 0.5045, 0.5512, 0.5512]}
{"unit_id": "orig:1930079711650284029", "ticker_idx": 3, "ts": "2025-06-04T01:50:37+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if U.S.–China trade tensions escalate into a full export ban, U.S. equipment vendors could face earnings downside of up to 45–50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS raises WFE outlook but stays neutral on SPE sector; bullish KLA on TSMC/DRAM share gains + Intel upside into 2026; AMAT/ASML face 45-50% earnings downside risk from full export ban.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Global Semiconductor Equipment: Forecast Raised on China Spend, but Industry Faces Continued Headwinds\nUBS (T. Arcuri, June 2, 2025)\n\nKLA, despite facing concerns about potential growth slowdown in 2H25 due to China’s domestic substitution push, has gained market recognition for its continued share gains in TSMC, DRAM, and advanced packaging.\nMoreover, Intel’s transition from post-test inspection to in-line inspection could present additional growth opportunities for KLA.\n\nLooking ahead to 2026, multiple drivers — including further share gains at TSMC and in DRAM, as well as Intel-related upside — are expected to support both KLA’s valuation premium and earnings growth.\n\n⸻\n\n• Global WFE Forecast Raised as China Spending Declines Less Than Expected:\n• Based on recent updates from chipmakers and equipment vendors, UBS has revised its 2025/2026 global WFE (wafer fab equipment) outlook upward, mainly due to stronger-than-expected domestic project activity in China.\n• China’s WFE capex is now expected to reach $33 billion in 2025 (YoY -8%), with a rebound anticipated in 2026.\n• Globally, WFE spending is forecast at $107 billion in 2025 (+10%) and $110 billion in 2026 (+3%).\n• While annual data points to moderate growth, quarterly trends remain too soft to justify a major short-term upgrade, so UBS maintains a neutral rating on the semiconductor production equipment (SPE) sector.\n\n⸻\n\n• 2026 Unlikely to Be a Down Year for WFE, but Geopolitical Risk Remains Central:\n• Despite some investor concerns about a WFE downturn in 2026, UBS sees such forecasts as overly conservative.\n• The market currently expects system revenue for key equipment makers (e.g., AMAT, ASML) to grow by ~7% YoY in 2026.\n• However, if U.S.–China trade tensions escalate into a full export ban,\n→ U.S. equipment vendors could face earnings downside of up to 45–50%.\n• Geopolitics remains the single greatest source of uncertainty for the industry.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0010729755390244122, "ret_signed_m1d": -0.0010729755390244122, "alpha_spy_m1d": -0.0008043559989781013, "alpha_c_m1d": -0.01283068232868434, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010116316659241953, "ret_signed_p1d": -0.010116316659241953, "alpha_spy_p1d": -0.014949085715431187, "alpha_c_p1d": -0.011903821272402193, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.057939830383483715, "ret_signed_p1w": -0.057939830383483715, "alpha_spy_p1w": -0.04882789922135866, "alpha_c_p1w": -0.017701520754852984, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.037860155961492836, "ret_signed_p1m": -0.037860155961492836, "alpha_spy_p1m": 0.014593584604450038, "alpha_c_p1m": 0.08873371751844417, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": -0.022398953768692165, "ret_signed_p3m": 0.022398953768692165, "alpha_spy_p3m": 0.10802477306864877, "alpha_c_p3m": 0.1754874151832162, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.3813107804147733, "ret_signed_p6m": -0.3813107804147733, "alpha_spy_p6m": -0.23423086024854212, "alpha_c_p6m": -0.00013904074110837605, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.0177, -0.0236, -0.0343, -0.0137, -0.0184, 0.0019, 0.0105, 0.0175, 0.0285, 0.0276, 0.0096, 0.0308, 0.031, 0.0042, -0.0182, -0.0441, -0.074, -0.0531, -0.0583, -0.1157, -0.138, -0.1596, -0.1232, -0.153, -0.1174, -0.1179, -0.0985, -0.0749, -0.1229, -0.1379, -0.1379, -0.1379, -0.1426, -0.1124, -0.0959, -0.0973, -0.1041, -0.1033, -0.1071, -0.1071, -0.0667, -0.0699, -0.078, -0.0731, -0.0336, -0.0392, 0.0222, 0.048, 0.0524, 0.0457, 0.0245, 0.0153, 0.0116, 0.0189, 0.0067, -0.015, 0.0025, 0.021, 0.0129, 0.0201, 0.0023, -0.0095, 0.0011, 0.0, 0.0101, 0.0233, 0.0425, 0.0492, 0.0579, 0.0371, 0.0187, 0.0271, 0.0208, 0.0193, 0.0011, -0.0031, 0.025, 0.0598, 0.0645, 0.0383, 0.0461, 0.0386, 0.0242, 0.0386, 0.0379, 0.0107, 0.0294, 0.0406, 0.042, 0.0576, 0.0483, 0.0535, 0.0823, -0.0408, -0.0034, -0.029, -0.0434, -0.0766, -0.0702, -0.0559, -0.0728, -0.0344, -0.0425, -0.0356, -0.0577, -0.0856, -0.0742, -0.0823, -0.0888, -0.0614, -0.0519, -0.0364, -0.0258, -0.0155, -0.0119, -0.0215, -0.0184, -0.0129, -0.0192, -0.0249, -0.0069, -0.0061, -0.0001, 0.0146, 0.0056, -0.0216, -0.0224, -0.0514, -0.039, -0.0047, 0.0146, 0.0398, 0.05, 0.0454, 0.0543, 0.0601, 0.121, 0.1264, 0.1327, 0.2203, 0.2198, 0.2461, 0.2599, 0.2418, 0.2475, 0.2492, 0.2687, 0.2727, 0.2931, 0.3488, 0.3526, 0.379, 0.3445, 0.3091, 0.306, 0.2509, 0.2971, 0.3011, 0.3417, 0.3484, 0.3458, 0.3816, 0.3675, 0.3372, 0.3643, 0.3746, 0.3956, 0.397, 0.4155, 0.4434, 0.4135, 0.4264, 0.4133, 0.3958, 0.3784, 0.3407, 0.3651, 0.3655, 0.373, 0.3571, 0.3457, 0.3493, 0.3327, 0.365, 0.3699, 0.284, 0.3222, 0.3245, 0.3999, 0.3813]}
{"unit_id": "orig:1943318257118908791", "ticker_idx": 0, "ts": "2025-07-10T14:35:52+00:00", "symbol": "TER", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Teradyne may receive a short-term stock price catalyst from the RTX testing news", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: TER gets short-term NVDA RTX testing catalyst but Q2/Q3 revenue below Street; raises PT $68→$74 with cautious stance. Advantest capacity constrained.", "resolved_tickers": ["TER"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Teradyne: Short-Term Catalyst from NVDA RTX Testing Involvement, Overall Market Share Expansion Limited, Q2 Outlook Still Under Pressure\nMorgan Stanley (S. Brett, July 09, 2025)\n\nMorgan Stanley notes that Teradyne may gain opportunities in NVDA RTX testing due to Advantest's capacity bottlenecks. While unlikely to lead to broad market share expansion, the news could bring a positive stock price reaction. Concurrently, the Q2 earnings guidance is expected to fall short of market expectations, maintaining a relatively cautious stance.\n\nAdvantest Bottleneck Drives Teradyne's Entry into RTX Testing\nFeedback from Asian supply chain channels indicates that Advantest's V93000 test platform has become a capacity bottleneck for NVDA Blackwell/Rubin production. In this context, NVDA is considering using Teradyne's Ultra Flex Plus platform for RTX chip testing. Teradyne has previously participated in test solution deployment for NVSwitch, CPO, and Mellanox modules, and the addition of the RTX segment is expected to bring marginal revenue contribution.\n\nLimited Short-Term Market Share Expansion in NVDA, System-Level Testing Still Dominated by Advantest\nDespite Teradyne potentially entering more high-value test nodes, Morgan Stanley judges that its position in NVDA's testing ecosystem will not significantly expand. The main reasons are: CPO and other components can be tested independently, system-level test resources remain highly concentrated with Advantest, and transferring core GPU test procedures is difficult. Teradyne's overall revenue contribution from NVDA is expected to be only mid-single-digit percentages in 2025-2026, with limited impact on the company's overall performance.\n\nRevenue and Price Target Upgraded, Valuation Reflects Overall Test Equipment Manufacturer Expansion Trend\nMorgan Stanley slightly raised its 2025/26 revenue forecasts to $2.82 billion/$3.19 billion, and EPS forecasts to $3.04/$3.90, to reflect marginal improvements from new NVDA test equipment orders. Concurrently, the target valuation multiple was increased by 1x to 19x, consistent with the valuation expansion trend among peers, raising the price target from $68 to $74.\n\nJune Quarter and September Quarter Expectations Remain Conservative, Street Needs to Downgrade Towards MS Estimates\nMorgan Stanley believes the current Street expectation of 18% QoQ revenue growth for the June quarter is too high, with its model projecting 11%. Similarly, for the September quarter, Street's growth expectation is 18%, while MS's estimate is only 12%. Morgan Stanley maintains its long-term view that the Street's future earnings expectations for the company are overly optimistic, predicting that the market will gradually lower its expectations and converge towards MS's estimates starting from the June quarter.\n\nNVDA Association Limited, But May Bring Valuation Sentiment Premium\nAlthough NVDA's contribution is relatively small, Morgan Stanley notes that the market typically assigns higher valuation premiums to companies associated with NVDA. Therefore, Teradyne may receive a short-term stock price catalyst from the RTX testing news, especially as the topic of AI testing demand continues to heat up.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05921720358319049, "ret_signed_m1d": -0.05921720358319049, "alpha_spy_m1d": -0.056404825144496296, "alpha_c_m1d": -0.051913506804044274, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015919646415188504, "ret_signed_p1d": -0.015919646415188504, "alpha_spy_p1d": -0.012404222544592947, "alpha_c_p1d": -0.01581530484774396, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.0590143852852727, "ret_signed_p1w": -0.0590143852852727, "alpha_spy_p1w": -0.06256167635227783, "alpha_c_p1w": -0.07358728224371602, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.08973844893562655, "ret_signed_p1m": 0.08973844893562655, "alpha_spy_p1m": 0.07158634306965261, "alpha_c_p1m": 0.06883553237587448, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.42243972341013336, "ret_signed_p3m": 0.42243972341013336, "alpha_spy_p3m": 0.35028629269497547, "alpha_c_p3m": 0.2501733926239409, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 1.1083538336226049, "ret_signed_p6m": 1.1083538336226049, "alpha_spy_p6m": 1.0104529868305447, "alpha_c_p6m": 0.8059694158645219, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.2522, -0.2502, -0.2795, -0.2786, -0.2786, -0.2932, -0.2782, -0.2591, -0.2174, -0.2192, -0.2221, -0.2418, -0.2487, -0.253, -0.2318, -0.2362, -0.2501, -0.2397, -0.222, -0.2162, -0.1591, -0.1404, -0.1468, -0.1598, -0.1624, -0.1752, -0.1757, -0.2004, -0.2041, -0.2194, -0.2194, -0.1712, -0.1863, -0.1789, -0.203, -0.194, -0.1741, -0.1679, -0.1712, -0.1358, -0.1199, -0.1006, -0.12, -0.1289, -0.1576, -0.1109, -0.1236, -0.1253, -0.1253, -0.125, -0.1142, -0.0835, -0.0854, -0.0768, -0.0839, -0.0882, -0.0679, -0.045, -0.0564, -0.0564, -0.0814, -0.0649, -0.0592, 0.0, -0.0159, -0.0345, -0.0567, -0.0673, -0.059, -0.0477, -0.0385, -0.0499, -0.0567, -0.0739, -0.0859, -0.0758, -0.0818, 0.0916, 0.0893, 0.0562, 0.072, 0.0843, 0.0635, 0.0795, 0.0897, 0.0614, 0.1319, 0.1561, 0.1338, 0.1095, 0.1195, 0.1217, 0.1105, 0.1083, 0.1691, 0.1871, 0.2027, 0.1863, 0.1954, 0.1989, 0.1989, 0.223, 0.213, 0.2106, 0.22, 0.2069, 0.168, 0.1959, 0.1729, 0.1393, 0.1578, 0.1564, 0.1604, 0.206, 0.2165, 0.3721, 0.367, 0.3521, 0.3486, 0.3734, 0.3635, 0.3971, 0.4324, 0.467, 0.4783, 0.5009, 0.4224, 0.4677, 0.4737, 0.3406, 0.4122, 0.3903, 0.4315, 0.4139, 0.4023, 0.4128, 0.4548, 0.4092, 0.4722, 0.4645, 0.4971, 0.4655, 0.7655, 0.7954, 0.8449, 0.8582, 0.7829, 0.9041, 0.878, 0.8502, 0.8684, 0.7989, 0.8196, 0.7214, 0.7252, 0.6995, 0.6618, 0.7076, 0.5824, 0.613, 0.6859, 0.7032, 0.8221, 0.8221, 0.8476, 0.8246, 0.9294, 0.9816, 1.0176, 1.0394, 1.0619, 1.0313, 1.0723, 1.0719, 0.9642, 0.9777, 0.9532, 0.8813, 0.9346, 0.9826, 1.0033, 1.0163, 1.0166, 1.0166, 1.0204, 1.0047, 0.9977, 0.9661, 0.9661, 1.1084]}
{"unit_id": "orig:1943318257118908791", "ticker_idx": 1, "ts": "2025-07-10T14:35:52+00:00", "symbol": "TER", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Street expectation of 18% QoQ revenue growth for the June quarter is too high, with its model projecting 11%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: TER gets short-term NVDA RTX testing catalyst but Q2/Q3 revenue below Street; raises PT $68→$74 with cautious stance. Advantest capacity constrained.", "resolved_tickers": ["TER"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Teradyne: Short-Term Catalyst from NVDA RTX Testing Involvement, Overall Market Share Expansion Limited, Q2 Outlook Still Under Pressure\nMorgan Stanley (S. Brett, July 09, 2025)\n\nMorgan Stanley notes that Teradyne may gain opportunities in NVDA RTX testing due to Advantest's capacity bottlenecks. While unlikely to lead to broad market share expansion, the news could bring a positive stock price reaction. Concurrently, the Q2 earnings guidance is expected to fall short of market expectations, maintaining a relatively cautious stance.\n\nAdvantest Bottleneck Drives Teradyne's Entry into RTX Testing\nFeedback from Asian supply chain channels indicates that Advantest's V93000 test platform has become a capacity bottleneck for NVDA Blackwell/Rubin production. In this context, NVDA is considering using Teradyne's Ultra Flex Plus platform for RTX chip testing. Teradyne has previously participated in test solution deployment for NVSwitch, CPO, and Mellanox modules, and the addition of the RTX segment is expected to bring marginal revenue contribution.\n\nLimited Short-Term Market Share Expansion in NVDA, System-Level Testing Still Dominated by Advantest\nDespite Teradyne potentially entering more high-value test nodes, Morgan Stanley judges that its position in NVDA's testing ecosystem will not significantly expand. The main reasons are: CPO and other components can be tested independently, system-level test resources remain highly concentrated with Advantest, and transferring core GPU test procedures is difficult. Teradyne's overall revenue contribution from NVDA is expected to be only mid-single-digit percentages in 2025-2026, with limited impact on the company's overall performance.\n\nRevenue and Price Target Upgraded, Valuation Reflects Overall Test Equipment Manufacturer Expansion Trend\nMorgan Stanley slightly raised its 2025/26 revenue forecasts to $2.82 billion/$3.19 billion, and EPS forecasts to $3.04/$3.90, to reflect marginal improvements from new NVDA test equipment orders. Concurrently, the target valuation multiple was increased by 1x to 19x, consistent with the valuation expansion trend among peers, raising the price target from $68 to $74.\n\nJune Quarter and September Quarter Expectations Remain Conservative, Street Needs to Downgrade Towards MS Estimates\nMorgan Stanley believes the current Street expectation of 18% QoQ revenue growth for the June quarter is too high, with its model projecting 11%. Similarly, for the September quarter, Street's growth expectation is 18%, while MS's estimate is only 12%. Morgan Stanley maintains its long-term view that the Street's future earnings expectations for the company are overly optimistic, predicting that the market will gradually lower its expectations and converge towards MS's estimates starting from the June quarter.\n\nNVDA Association Limited, But May Bring Valuation Sentiment Premium\nAlthough NVDA's contribution is relatively small, Morgan Stanley notes that the market typically assigns higher valuation premiums to companies associated with NVDA. Therefore, Teradyne may receive a short-term stock price catalyst from the RTX testing news, especially as the topic of AI testing demand continues to heat up.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05921720358319049, "ret_signed_m1d": 0.05921720358319049, "alpha_spy_m1d": 0.056404825144496296, "alpha_c_m1d": 0.051913506804044274, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015919646415188504, "ret_signed_p1d": 0.015919646415188504, "alpha_spy_p1d": 0.012404222544592947, "alpha_c_p1d": 0.01581530484774396, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.0590143852852727, "ret_signed_p1w": 0.0590143852852727, "alpha_spy_p1w": 0.06256167635227783, "alpha_c_p1w": 0.07358728224371602, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.08973844893562655, "ret_signed_p1m": -0.08973844893562655, "alpha_spy_p1m": -0.07158634306965261, "alpha_c_p1m": -0.06883553237587448, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.42243972341013336, "ret_signed_p3m": -0.42243972341013336, "alpha_spy_p3m": -0.35028629269497547, "alpha_c_p3m": -0.2501733926239409, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 1.1083538336226049, "ret_signed_p6m": -1.1083538336226049, "alpha_spy_p6m": -1.0104529868305447, "alpha_c_p6m": -0.8059694158645219, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.2522, -0.2502, -0.2795, -0.2786, -0.2786, -0.2932, -0.2782, -0.2591, -0.2174, -0.2192, -0.2221, -0.2418, -0.2487, -0.253, -0.2318, -0.2362, -0.2501, -0.2397, -0.222, -0.2162, -0.1591, -0.1404, -0.1468, -0.1598, -0.1624, -0.1752, -0.1757, -0.2004, -0.2041, -0.2194, -0.2194, -0.1712, -0.1863, -0.1789, -0.203, -0.194, -0.1741, -0.1679, -0.1712, -0.1358, -0.1199, -0.1006, -0.12, -0.1289, -0.1576, -0.1109, -0.1236, -0.1253, -0.1253, -0.125, -0.1142, -0.0835, -0.0854, -0.0768, -0.0839, -0.0882, -0.0679, -0.045, -0.0564, -0.0564, -0.0814, -0.0649, -0.0592, 0.0, -0.0159, -0.0345, -0.0567, -0.0673, -0.059, -0.0477, -0.0385, -0.0499, -0.0567, -0.0739, -0.0859, -0.0758, -0.0818, 0.0916, 0.0893, 0.0562, 0.072, 0.0843, 0.0635, 0.0795, 0.0897, 0.0614, 0.1319, 0.1561, 0.1338, 0.1095, 0.1195, 0.1217, 0.1105, 0.1083, 0.1691, 0.1871, 0.2027, 0.1863, 0.1954, 0.1989, 0.1989, 0.223, 0.213, 0.2106, 0.22, 0.2069, 0.168, 0.1959, 0.1729, 0.1393, 0.1578, 0.1564, 0.1604, 0.206, 0.2165, 0.3721, 0.367, 0.3521, 0.3486, 0.3734, 0.3635, 0.3971, 0.4324, 0.467, 0.4783, 0.5009, 0.4224, 0.4677, 0.4737, 0.3406, 0.4122, 0.3903, 0.4315, 0.4139, 0.4023, 0.4128, 0.4548, 0.4092, 0.4722, 0.4645, 0.4971, 0.4655, 0.7655, 0.7954, 0.8449, 0.8582, 0.7829, 0.9041, 0.878, 0.8502, 0.8684, 0.7989, 0.8196, 0.7214, 0.7252, 0.6995, 0.6618, 0.7076, 0.5824, 0.613, 0.6859, 0.7032, 0.8221, 0.8221, 0.8476, 0.8246, 0.9294, 0.9816, 1.0176, 1.0394, 1.0619, 1.0313, 1.0723, 1.0719, 0.9642, 0.9777, 0.9532, 0.8813, 0.9346, 0.9826, 1.0033, 1.0163, 1.0166, 1.0166, 1.0204, 1.0047, 0.9977, 0.9661, 0.9661, 1.1084]}
{"unit_id": "orig:1943318257118908791", "ticker_idx": 2, "ts": "2025-07-10T14:35:52+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Advantest's V93000 test platform has become a capacity bottleneck for NVDA Blackwell/Rubin production", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: TER gets short-term NVDA RTX testing catalyst but Q2/Q3 revenue below Street; raises PT $68→$74 with cautious stance. Advantest capacity constrained.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Teradyne: Short-Term Catalyst from NVDA RTX Testing Involvement, Overall Market Share Expansion Limited, Q2 Outlook Still Under Pressure\nMorgan Stanley (S. Brett, July 09, 2025)\n\nMorgan Stanley notes that Teradyne may gain opportunities in NVDA RTX testing due to Advantest's capacity bottlenecks. While unlikely to lead to broad market share expansion, the news could bring a positive stock price reaction. Concurrently, the Q2 earnings guidance is expected to fall short of market expectations, maintaining a relatively cautious stance.\n\nAdvantest Bottleneck Drives Teradyne's Entry into RTX Testing\nFeedback from Asian supply chain channels indicates that Advantest's V93000 test platform has become a capacity bottleneck for NVDA Blackwell/Rubin production. In this context, NVDA is considering using Teradyne's Ultra Flex Plus platform for RTX chip testing. Teradyne has previously participated in test solution deployment for NVSwitch, CPO, and Mellanox modules, and the addition of the RTX segment is expected to bring marginal revenue contribution.\n\nLimited Short-Term Market Share Expansion in NVDA, System-Level Testing Still Dominated by Advantest\nDespite Teradyne potentially entering more high-value test nodes, Morgan Stanley judges that its position in NVDA's testing ecosystem will not significantly expand. The main reasons are: CPO and other components can be tested independently, system-level test resources remain highly concentrated with Advantest, and transferring core GPU test procedures is difficult. Teradyne's overall revenue contribution from NVDA is expected to be only mid-single-digit percentages in 2025-2026, with limited impact on the company's overall performance.\n\nRevenue and Price Target Upgraded, Valuation Reflects Overall Test Equipment Manufacturer Expansion Trend\nMorgan Stanley slightly raised its 2025/26 revenue forecasts to $2.82 billion/$3.19 billion, and EPS forecasts to $3.04/$3.90, to reflect marginal improvements from new NVDA test equipment orders. Concurrently, the target valuation multiple was increased by 1x to 19x, consistent with the valuation expansion trend among peers, raising the price target from $68 to $74.\n\nJune Quarter and September Quarter Expectations Remain Conservative, Street Needs to Downgrade Towards MS Estimates\nMorgan Stanley believes the current Street expectation of 18% QoQ revenue growth for the June quarter is too high, with its model projecting 11%. Similarly, for the September quarter, Street's growth expectation is 18%, while MS's estimate is only 12%. Morgan Stanley maintains its long-term view that the Street's future earnings expectations for the company are overly optimistic, predicting that the market will gradually lower its expectations and converge towards MS's estimates starting from the June quarter.\n\nNVDA Association Limited, But May Bring Valuation Sentiment Premium\nAlthough NVDA's contribution is relatively small, Morgan Stanley notes that the market typically assigns higher valuation premiums to companies associated with NVDA. Therefore, Teradyne may receive a short-term stock price catalyst from the RTX testing news, especially as the topic of AI testing demand continues to heat up.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023566027757840335, "ret_signed_m1d": 0.023566027757840335, "alpha_spy_m1d": 0.02075364931914614, "alpha_c_m1d": 0.016262330978694117, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00711430048027295, "ret_signed_p1d": -0.00711430048027295, "alpha_spy_p1d": -0.010629724350868508, "alpha_c_p1d": -0.007218642047717494, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": 0.04090707527181725, "ret_signed_p1w": -0.04090707527181725, "alpha_spy_p1w": -0.03735978420481212, "alpha_c_p1w": -0.026334178313373924, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": -0.06758554958308871, "ret_signed_p1m": 0.06758554958308871, "alpha_spy_p1m": 0.08573765544906264, "alpha_c_p1m": 0.08848846614284078, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.6043715239416318, "ret_signed_p3m": -0.6043715239416318, "alpha_spy_p3m": -0.532218093226474, "alpha_c_p3m": -0.4321051931554394, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.7496160090190283, "ret_signed_p6m": -0.7496160090190283, "alpha_spy_p6m": -0.6517151622269681, "alpha_c_p6m": -0.44723159126094525, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.4745, -0.4839, -0.5177, -0.5009, -0.5122, -0.5166, -0.5206, -0.5057, -0.4901, -0.4664, -0.4922, -0.4922, -0.4786, -0.4427, -0.4365, -0.4365, -0.4365, -0.4353, -0.4142, -0.4046, -0.3508, -0.3643, -0.3334, -0.3408, -0.3532, -0.3716, -0.353, -0.3766, -0.3967, -0.394, -0.3685, -0.365, -0.3548, -0.3203, -0.3449, -0.3695, -0.3529, -0.3404, -0.3103, -0.2949, -0.2606, -0.2615, -0.2582, -0.2616, -0.2438, -0.1709, -0.1507, -0.1407, -0.1613, -0.1257, -0.1365, -0.1229, -0.0938, -0.0485, -0.0587, -0.0525, -0.0427, -0.0729, -0.0636, -0.0507, -0.0498, -0.0258, -0.0236, 0.0, 0.0071, 0.024, 0.0422, 0.0489, 0.0409, -0.0053, -0.0053, 0.0089, 0.0147, 0.008, 0.0222, -0.0694, -0.0791, -0.0889, -0.0796, -0.0916, -0.1063, -0.1067, -0.1036, -0.1009, -0.0676, -0.0676, -0.0084, 0.0454, -0.0031, 0.0044, 0.0191, 0.0147, -0.0427, -0.0267, -0.0427, -0.0325, -0.0302, 0.0133, 0.028, 0.0382, -0.044, -0.0516, -0.0489, -0.004, 0.0173, 0.0618, 0.1303, 0.1672, 0.2183, 0.2406, 0.2406, 0.2566, 0.2312, 0.2921, 0.3357, 0.3784, 0.3784, 0.3699, 0.3366, 0.2868, 0.3495, 0.3054, 0.3076, 0.3406, 0.3981, 0.5941, 0.6044, 0.6008, 0.6044, 0.5892, 0.5892, 0.5148, 0.5473, 0.5585, 0.5032, 0.5603, 0.5505, 0.5237, 0.4671, 0.5219, 0.6213, 0.6146, 0.971, 0.9844, 1.0615, 1.0615, 0.9408, 0.8254, 0.8828, 0.7786, 0.8463, 0.771, 0.767, 0.8418, 0.7403, 0.7821, 0.7162, 0.7064, 0.8565, 0.632, 0.632, 0.7002, 0.734, 0.8187, 0.8334, 0.7554, 0.7652, 0.8588, 0.8445, 0.8, 0.8044, 0.8049, 0.7955, 0.8748, 0.8521, 0.7331, 0.7086, 0.7331, 0.6757, 0.7104, 0.7866, 0.7532, 0.7964, 0.7639, 0.804, 0.7625, 0.7496, 0.7496, 0.7496, 0.7496]}
{"unit_id": "orig:1939871531506815071", "ticker_idx": 0, "ts": "2025-07-01T02:19:48+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "TPU Could Be a Hidden Catalyst for Google Cloud Growth... their contribution to Google Cloud revenue might be underestimated", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS: OpenAI TPU adoption is a hidden catalyst for Google Cloud; NVDA revenue triples at Google; AWS/Trainium face negative interpretive pressure.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Google's Partnership with OpenAI: TPU Adoption Becomes Key Endorsement for Google's Infrastructure\nMorgan Stanley (B. Nowak, 2025/06/30)\n\nIf OpenAI officially adopts Google TPUs for inference, it will serve as a significant endorsement of Google's ASIC ecosystem's technical capabilities, reflecting Google's leading position in AI chip infrastructure. It also reveals the practical impact of NVIDIA's capacity limitations on industry decisions.\n\nOpenAI's Adoption of TPU Strengthens Google's Position in AI Chip Infrastructure\n\nOpenAI and Google have reached a cooperation agreement: in addition to utilizing Google Cloud computing resources, OpenAI also plans to use Google TPUs for inference tasks, although they will not receive the most advanced TPU versions. Morgan Stanley believes this reflects OpenAI's recognition of Google's infrastructure capabilities and highlights Google's long-term accumulation in the ASIC ecosystem. This will also mark OpenAI's first use of non-NVIDIA chips for large-scale tasks, potentially leading to higher confidence in Google's AI infrastructure.\n\nGoogle TPU Could Be a Hidden Catalyst for Google Cloud Growth\n\nAlthough the current market is still dominated by NVIDIA GPUs, Morgan Stanley points out that as developers become more familiar with TPUs and as TPUs expand to customers beyond Google, their contribution to Google Cloud revenue might be underestimated. Morgan Stanley's model forecasts TPU expenditure of only $21 billion / $24 billion in 2027/28, far lower than NVIDIA GPU expenditure ($243 billion / $258 billion) in the same period. This discrepancy indicates significant room for Google to gain market share.\n\nNVIDIA Supply Bottleneck May Be One of the Inducements for OpenAI's Shift to TPU\n\nMorgan Stanley estimates that NVIDIA's revenue from Google will more than triple year-over-year to $20 billion, accounting for approximately 65% of the processor market share. However, since NVIDIA's capacity, especially for rack-scale products, is sold out, part of Google's motivation for shifting to non-NVIDIA architectures also stems from the shortage of inference computing power, rather than a pure shift in competitive landscape. This passive demand-side push further highlights the differentiated value of Google's product ecosystem.\n\nAWS and Trainium Face Negative Interpretive Pressure\n\nOpenAI currently deploys its AI workloads on multiple cloud platforms, including Google Cloud, Azure, Oracle, and CoreWeave, but AWS is not yet among them. Morgan Stanley believes this may reflect AWS's persistent capacity bottlenecks and also casts a negative light on its custom Trainium chips. Especially given that TPUs are an older generation, OpenAI still choosing them over Trainium further amplifies market attention on AWS's growth and Trainium's market position.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002217876900135085, "ret_signed_m1d": 0.002217876900135085, "alpha_spy_m1d": 0.0018940927896933246, "alpha_c_m1d": -0.0049276861679752315, "bench_spy_m1d": 0.0003237841104417605, "bench_c_m1d": 0.007145563068110317, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015923470766441294, "ret_signed_p1d": 0.015923470766441294, "alpha_spy_p1d": 0.011390194250625374, "alpha_c_p1d": 0.018800229896211662, "bench_spy_p1d": 0.0045332765158159205, "bench_c_p1d": -0.0028767591297703676, "ret_p1w": -0.008416817179226932, "ret_signed_p1w": -0.008416817179226932, "alpha_spy_p1w": -0.012772007451472556, "alpha_c_p1w": 0.0006775228157763769, "bench_spy_p1w": 0.004355190272245624, "bench_c_p1w": -0.009094339995003309, "ret_p1m": 0.11766375747710489, "ret_signed_p1m": 0.11766375747710489, "alpha_spy_p1m": 0.09044775470904476, "alpha_c_p1m": 0.12982041013990187, "bench_spy_p1m": 0.027216002768060132, "bench_c_p1m": -0.012156652662796974, "ret_p3m": 0.4033239713607655, "ret_signed_p3m": 0.4033239713607655, "alpha_spy_p3m": 0.32884027198230337, "alpha_c_p3m": 0.30289596119972506, "bench_spy_p3m": 0.07448369937846211, "bench_c_p3m": 0.10042801016104042, "ret_p6m": 0.7889930303124919, "ret_signed_p6m": 0.7889930303124919, "alpha_spy_p6m": 0.6648294476836638, "alpha_c_p6m": 0.6857498976882614, "bench_spy_p6m": 0.1241635826288281, "bench_c_p6m": 0.10324313262423046, "price_path": [-0.1439, -0.173, -0.1664, -0.1781, -0.0985, -0.132, -0.1074, -0.0965, -0.1121, -0.1291, -0.1414, -0.1414, -0.1612, -0.1396, -0.1176, -0.0953, -0.08, -0.0877, -0.0903, -0.098, -0.0838, -0.0683, -0.0673, -0.0728, -0.1401, -0.1237, -0.1324, -0.0999, -0.0939, -0.0607, -0.0687, -0.056, -0.054, -0.0686, -0.0426, -0.0294, -0.0431, -0.0431, -0.0179, -0.021, -0.0238, -0.0245, -0.0399, -0.0561, -0.0455, -0.0445, -0.0135, 0.0014, 0.0157, 0.0086, -0.0008, -0.0067, 0.0053, 0.0006, -0.0143, -0.0143, -0.0523, -0.0606, -0.0516, -0.0293, -0.0131, 0.0153, 0.0022, 0.0, 0.0159, 0.021, 0.021, 0.0054, -0.0084, 0.0044, 0.0101, 0.0247, 0.0325, 0.035, 0.0405, 0.044, 0.0524, 0.0811, 0.0881, 0.0818, 0.0929, 0.0986, 0.0952, 0.1132, 0.1177, 0.0913, 0.0756, 0.1092, 0.1071, 0.1152, 0.1176, 0.1455, 0.1431, 0.1564, 0.1485, 0.1541, 0.1596, 0.1573, 0.1463, 0.1335, 0.136, 0.172, 0.1857, 0.178, 0.1799, 0.2036, 0.2108, 0.2108, 0.2019, 0.3118, 0.3211, 0.3364, 0.3322, 0.364, 0.3614, 0.3682, 0.3707, 0.4322, 0.4296, 0.4203, 0.4346, 0.4499, 0.4374, 0.4325, 0.4067, 0.3991, 0.4033, 0.3892, 0.3837, 0.394, 0.3985, 0.3966, 0.4255, 0.3989, 0.3924, 0.3748, 0.3466, 0.3897, 0.3971, 0.4289, 0.4313, 0.4418, 0.4603, 0.4256, 0.4326, 0.4406, 0.4795, 0.5327, 0.5225, 0.5629, 0.6022, 0.6006, 0.615, 0.5798, 0.6183, 0.6208, 0.5871, 0.6513, 0.6582, 0.632, 0.5856, 0.5733, 0.6224, 0.6181, 0.6667, 0.6476, 0.7057, 0.8134, 0.841, 0.8212, 0.8212, 0.8225, 0.7924, 0.7976, 0.8194, 0.8079, 0.8287, 0.7869, 0.806, 0.8239, 0.7795, 0.7617, 0.7556, 0.7462, 0.6901, 0.7228, 0.7495, 0.7644, 0.7905, 0.789]}
{"unit_id": "orig:1939871531506815071", "ticker_idx": 1, "ts": "2025-07-01T02:19:48+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's revenue from Google will more than triple year-over-year to $20 billion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS: OpenAI TPU adoption is a hidden catalyst for Google Cloud; NVDA revenue triples at Google; AWS/Trainium face negative interpretive pressure.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Google's Partnership with OpenAI: TPU Adoption Becomes Key Endorsement for Google's Infrastructure\nMorgan Stanley (B. Nowak, 2025/06/30)\n\nIf OpenAI officially adopts Google TPUs for inference, it will serve as a significant endorsement of Google's ASIC ecosystem's technical capabilities, reflecting Google's leading position in AI chip infrastructure. It also reveals the practical impact of NVIDIA's capacity limitations on industry decisions.\n\nOpenAI's Adoption of TPU Strengthens Google's Position in AI Chip Infrastructure\n\nOpenAI and Google have reached a cooperation agreement: in addition to utilizing Google Cloud computing resources, OpenAI also plans to use Google TPUs for inference tasks, although they will not receive the most advanced TPU versions. Morgan Stanley believes this reflects OpenAI's recognition of Google's infrastructure capabilities and highlights Google's long-term accumulation in the ASIC ecosystem. This will also mark OpenAI's first use of non-NVIDIA chips for large-scale tasks, potentially leading to higher confidence in Google's AI infrastructure.\n\nGoogle TPU Could Be a Hidden Catalyst for Google Cloud Growth\n\nAlthough the current market is still dominated by NVIDIA GPUs, Morgan Stanley points out that as developers become more familiar with TPUs and as TPUs expand to customers beyond Google, their contribution to Google Cloud revenue might be underestimated. Morgan Stanley's model forecasts TPU expenditure of only $21 billion / $24 billion in 2027/28, far lower than NVIDIA GPU expenditure ($243 billion / $258 billion) in the same period. This discrepancy indicates significant room for Google to gain market share.\n\nNVIDIA Supply Bottleneck May Be One of the Inducements for OpenAI's Shift to TPU\n\nMorgan Stanley estimates that NVIDIA's revenue from Google will more than triple year-over-year to $20 billion, accounting for approximately 65% of the processor market share. However, since NVIDIA's capacity, especially for rack-scale products, is sold out, part of Google's motivation for shifting to non-NVIDIA architectures also stems from the shortage of inference computing power, rather than a pure shift in competitive landscape. This passive demand-side push further highlights the differentiated value of Google's product ecosystem.\n\nAWS and Trainium Face Negative Interpretive Pressure\n\nOpenAI currently deploys its AI workloads on multiple cloud platforms, including Google Cloud, Azure, Oracle, and CoreWeave, but AWS is not yet among them. Morgan Stanley believes this may reflect AWS's persistent capacity bottlenecks and also casts a negative light on its custom Trainium chips. Especially given that TPUs are an older generation, OpenAI still choosing them over Trainium further amplifies market attention on AWS's growth and Trainium's market position.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030593628670980477, "ret_signed_m1d": 0.030593628670980477, "alpha_spy_m1d": 0.030269844560538717, "alpha_c_m1d": 0.019572715393224804, "bench_spy_m1d": 0.0003237841104417605, "bench_c_m1d": 0.011020913277755673, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025766351418616162, "ret_signed_p1d": 0.025766351418616162, "alpha_spy_p1d": 0.021233074902800242, "alpha_c_p1d": 0.006153529443998718, "bench_spy_p1d": 0.0045332765158159205, "bench_c_p1d": 0.019612821974617445, "ret_p1w": 0.043705126928820226, "ret_signed_p1w": 0.043705126928820226, "alpha_spy_p1w": 0.0393499366565746, "alpha_c_p1w": 0.016007793751005917, "bench_spy_p1w": 0.004355190272245624, "bench_c_p1w": 0.02769733317781431, "ret_p1m": 0.16940637132901948, "ret_signed_p1m": 0.16940637132901948, "alpha_spy_p1m": 0.14219036856095935, "alpha_c_p1m": 0.09537783166273206, "bench_spy_p1m": 0.027216002768060132, "bench_c_p1m": 0.07402853966628742, "ret_p3m": 0.16242687904430064, "ret_signed_p3m": 0.16242687904430064, "alpha_spy_p3m": 0.08794317966583853, "alpha_c_p3m": -0.004264073598932994, "bench_spy_p3m": 0.07448369937846211, "bench_c_p3m": 0.16669095264323364, "ret_p6m": 0.23047048282727722, "ret_signed_p6m": 0.23047048282727722, "alpha_spy_p6m": 0.10630690019844913, "alpha_c_p6m": -0.09385950444228386, "bench_spy_p6m": 0.1241635826288281, "bench_c_p6m": 0.3243299872695611, "price_path": [-0.336, -0.3848, -0.3631, -0.3719, -0.2543, -0.2984, -0.2764, -0.2779, -0.2682, -0.3184, -0.338, -0.338, -0.3679, -0.355, -0.3301, -0.3058, -0.2759, -0.2908, -0.2889, -0.2895, -0.272, -0.2532, -0.2576, -0.2594, -0.2365, -0.2344, -0.2391, -0.1977, -0.1525, -0.1172, -0.1205, -0.1168, -0.1157, -0.1235, -0.1403, -0.1336, -0.1436, -0.1436, -0.1162, -0.1207, -0.0921, -0.1186, -0.1039, -0.0789, -0.0743, -0.0869, -0.0756, -0.0697, -0.061, -0.0683, -0.0541, -0.0739, -0.0562, -0.0599, -0.051, -0.051, -0.0616, -0.0596, -0.0352, 0.0066, 0.0112, 0.029, 0.0306, 0.0, 0.0258, 0.0394, 0.0394, 0.0322, 0.0437, 0.0625, 0.0705, 0.0758, 0.0703, 0.1135, 0.1179, 0.1285, 0.1247, 0.1179, 0.0896, 0.114, 0.1333, 0.1318, 0.153, 0.1449, 0.1694, 0.1603, 0.1332, 0.1742, 0.1628, 0.1704, 0.1792, 0.1918, 0.1876, 0.1948, 0.1845, 0.1873, 0.1771, 0.1873, 0.1457, 0.1442, 0.1414, 0.1611, 0.1729, 0.1857, 0.1846, 0.1753, 0.1362, 0.1362, 0.114, 0.113, 0.1198, 0.0895, 0.0979, 0.1139, 0.1568, 0.1558, 0.16, 0.1596, 0.1408, 0.1109, 0.1497, 0.1525, 0.1978, 0.164, 0.1545, 0.1592, 0.1624, 0.1863, 0.2172, 0.2215, 0.2322, 0.2239, 0.2104, 0.2071, 0.2337, 0.2562, 0.1948, 0.2285, 0.1744, 0.1731, 0.186, 0.1952, 0.1915, 0.1818, 0.1761, 0.1883, 0.2151, 0.2492, 0.3114, 0.3506, 0.3236, 0.3209, 0.3496, 0.2962, 0.2735, 0.2269, 0.2274, 0.2985, 0.2601, 0.2643, 0.219, 0.2406, 0.2173, 0.1831, 0.2168, 0.1784, 0.1669, 0.1909, 0.16, 0.1759, 0.1759, 0.1547, 0.1737, 0.1838, 0.1716, 0.1964, 0.19, 0.2105, 0.2067, 0.199, 0.1804, 0.1418, 0.1501, 0.1594, 0.1152, 0.1361, 0.1808, 0.1984, 0.2344, 0.2305]}
{"unit_id": "orig:1939871531506815071", "ticker_idx": 2, "ts": "2025-07-01T02:19:48+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "casts a negative light on its custom Trainium chips... further amplifies market attention on AWS's growth and Trainium's market position", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MS: OpenAI TPU adoption is a hidden catalyst for Google Cloud; NVDA revenue triples at Google; AWS/Trainium face negative interpretive pressure.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Google's Partnership with OpenAI: TPU Adoption Becomes Key Endorsement for Google's Infrastructure\nMorgan Stanley (B. Nowak, 2025/06/30)\n\nIf OpenAI officially adopts Google TPUs for inference, it will serve as a significant endorsement of Google's ASIC ecosystem's technical capabilities, reflecting Google's leading position in AI chip infrastructure. It also reveals the practical impact of NVIDIA's capacity limitations on industry decisions.\n\nOpenAI's Adoption of TPU Strengthens Google's Position in AI Chip Infrastructure\n\nOpenAI and Google have reached a cooperation agreement: in addition to utilizing Google Cloud computing resources, OpenAI also plans to use Google TPUs for inference tasks, although they will not receive the most advanced TPU versions. Morgan Stanley believes this reflects OpenAI's recognition of Google's infrastructure capabilities and highlights Google's long-term accumulation in the ASIC ecosystem. This will also mark OpenAI's first use of non-NVIDIA chips for large-scale tasks, potentially leading to higher confidence in Google's AI infrastructure.\n\nGoogle TPU Could Be a Hidden Catalyst for Google Cloud Growth\n\nAlthough the current market is still dominated by NVIDIA GPUs, Morgan Stanley points out that as developers become more familiar with TPUs and as TPUs expand to customers beyond Google, their contribution to Google Cloud revenue might be underestimated. Morgan Stanley's model forecasts TPU expenditure of only $21 billion / $24 billion in 2027/28, far lower than NVIDIA GPU expenditure ($243 billion / $258 billion) in the same period. This discrepancy indicates significant room for Google to gain market share.\n\nNVIDIA Supply Bottleneck May Be One of the Inducements for OpenAI's Shift to TPU\n\nMorgan Stanley estimates that NVIDIA's revenue from Google will more than triple year-over-year to $20 billion, accounting for approximately 65% of the processor market share. However, since NVIDIA's capacity, especially for rack-scale products, is sold out, part of Google's motivation for shifting to non-NVIDIA architectures also stems from the shortage of inference computing power, rather than a pure shift in competitive landscape. This passive demand-side push further highlights the differentiated value of Google's product ecosystem.\n\nAWS and Trainium Face Negative Interpretive Pressure\n\nOpenAI currently deploys its AI workloads on multiple cloud platforms, including Google Cloud, Azure, Oracle, and CoreWeave, but AWS is not yet among them. Morgan Stanley believes this may reflect AWS's persistent capacity bottlenecks and also casts a negative light on its custom Trainium chips. Especially given that TPUs are an older generation, OpenAI still choosing them over Trainium further amplifies market attention on AWS's growth and Trainium's market position.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004853521235747338, "ret_signed_m1d": 0.004853521235747338, "alpha_spy_m1d": 0.005177305346189098, "alpha_c_m1d": 0.0005925554135463029, "bench_spy_m1d": 0.0003237841104417605, "bench_c_m1d": -0.004260965822201035, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002449462616694653, "ret_signed_p1d": 0.002449462616694653, "alpha_spy_p1d": 0.006982739132510574, "alpha_c_p1d": 0.01046738259401403, "bench_spy_p1d": 0.0045332765158159205, "bench_c_p1d": 0.008017919977319377, "ret_p1w": -0.004989594801851349, "ret_signed_p1w": 0.004989594801851349, "alpha_spy_p1w": 0.009344785074096973, "alpha_c_p1w": 0.002240523352642687, "bench_spy_p1w": 0.004355190272245624, "bench_c_p1w": -0.002749071449208662, "ret_p1m": 0.04413496975062481, "ret_signed_p1m": -0.04413496975062481, "alpha_spy_p1m": -0.016918966982564676, "alpha_c_p1m": -0.021501484883315047, "bench_spy_p1m": 0.027216002768060132, "bench_c_p1m": 0.02263348486730976, "ret_p3m": -0.003084495662983855, "ret_signed_p3m": 0.003084495662983855, "alpha_spy_p3m": 0.07756819504144596, "alpha_c_p3m": 0.10329886234398766, "bench_spy_p3m": 0.07448369937846211, "bench_c_p3m": 0.10021436668100381, "ret_p6m": 0.05406875535668543, "ret_signed_p6m": -0.05406875535668543, "alpha_spy_p6m": 0.07009482727214267, "alpha_c_p6m": 0.07348177536380307, "bench_spy_p6m": 0.1241635826288281, "bench_c_p6m": 0.1275505307204885, "price_path": [-0.1907, -0.2243, -0.205, -0.2259, -0.1332, -0.178, -0.1614, -0.1739, -0.1854, -0.2092, -0.217, -0.217, -0.241, -0.2145, -0.1808, -0.1539, -0.1427, -0.1486, -0.15, -0.1635, -0.1373, -0.1383, -0.1547, -0.1608, -0.144, -0.1287, -0.1243, -0.0536, -0.0412, -0.0463, -0.0694, -0.0674, -0.0649, -0.0743, -0.0877, -0.0787, -0.0883, -0.0883, -0.0655, -0.0714, -0.067, -0.0701, -0.0626, -0.0669, -0.06, -0.0569, -0.0313, -0.0158, -0.0129, -0.0329, -0.0327, -0.0379, -0.0198, -0.0256, -0.036, -0.036, -0.0489, -0.0544, -0.0349, -0.0384, -0.0152, 0.0129, -0.0049, 0.0, -0.0024, 0.0134, 0.0134, 0.0137, -0.005, 0.0094, 0.0082, 0.0207, 0.0237, 0.0267, 0.0124, 0.0155, 0.0257, 0.0401, 0.0318, 0.0355, 0.0534, 0.0498, 0.0559, 0.0479, 0.0441, 0.0619, -0.0259, -0.04, -0.0304, 0.0084, 0.0121, 0.0101, 0.0038, 0.0046, 0.0186, 0.0477, 0.0479, 0.05, 0.0342, 0.0152, 0.0068, 0.038, 0.0339, 0.0374, 0.0393, 0.0505, 0.0387, 0.0387, 0.0221, 0.0251, 0.069, 0.0538, 0.0698, 0.0806, 0.0448, 0.043, 0.0349, 0.0498, 0.0616, 0.0506, 0.0489, 0.05, 0.0325, 0.0011, -0.0011, -0.0105, -0.0031, 0.0078, -0.004, 0.0008, 0.0088, -0.0043, 0.002, 0.006, 0.0216, 0.033, -0.0186, -0.0018, -0.0185, -0.0222, -0.0272, -0.0337, -0.0181, 0.0071, -0.0114, 0.0029, 0.017, 0.0295, 0.0399, 0.0446, 0.0109, 0.1078, 0.1521, 0.1309, 0.1349, 0.1024, 0.1086, 0.1267, 0.1299, 0.1077, 0.0777, 0.0645, 0.0563, 0.0095, 0.0101, -0.0151, 0.001, 0.0264, 0.0418, 0.0395, 0.0395, 0.0579, 0.0609, 0.0633, 0.0541, 0.0392, 0.0411, 0.0292, 0.0338, 0.0513, 0.0445, 0.026, 0.0094, 0.0095, 0.0037, 0.0286, 0.0313, 0.0362, 0.053, 0.0541]}
{"unit_id": "quote:1922491308913971339", "ticker_idx": 0, "ts": "2025-05-14T03:17:00+00:00", "symbol": "Shin Zu Shing", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I bought shares of the Taiwanese stock Shin Zu Shing, making up one-third of my portfolio", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Initiated a large position in Shin Zu Shing (Taiwan), one-third of portfolio, on Apple foldable supply chain thesis.", "resolved_tickers": ["3376.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Shin Zu Shing listed on TWSE as 3376.TW", "bench_c": "EWT", "bench_c_reason": "country_fallback (suffix='TW')", "bench_c_industry": "Metal Fabrication", "unit_kind": "quote", "lang": "en", "tweet_text": "Today, I bought shares of the Taiwanese stock Shin Zu Shing, making up one-third of my portfolio.\n\n*Not investment advice.", "reply_text": null, "parent_text": null, "quoted_text": "I did some research on Apple’s foldable supply chain and bought a Taiwanese stock today—for the first time in my life.", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0012764281780371922, "alpha_c_m1d": 0.00911293016667869, "bench_spy_m1d": -0.0012764281780371922, "bench_c_m1d": -0.00911293016667869, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020460400278729307, "ret_signed_p1d": -0.020460400278729307, "alpha_spy_p1d": -0.025344637099915635, "alpha_c_p1d": -0.024551939632201036, "bench_spy_p1d": 0.004884236821186327, "bench_c_p1d": 0.004091539353471729, "ret_p1w": -0.023017930278396714, "ret_signed_p1w": -0.023017930278396714, "alpha_spy_p1w": -0.014968101144735124, "alpha_c_p1w": -0.02208807881866126, "bench_spy_p1w": -0.00804982913366159, "bench_c_p1w": -0.0009298514597354535, "ret_p1m": -0.02813299027773164, "ret_signed_p1m": -0.02813299027773164, "alpha_spy_p1m": -0.05563497967336506, "alpha_c_p1m": -0.08336828922300943, "bench_spy_p1m": 0.02750198939563342, "bench_c_p1m": 0.05523529894527779, "ret_p3m": 0.20535708489824733, "ret_signed_p3m": 0.20535708489824733, "alpha_spy_p3m": 0.11990658003197652, "alpha_c_p3m": 0.08428591323198109, "bench_spy_p3m": 0.08545050486627082, "bench_c_p3m": 0.12107117166626624, "ret_p6m": 0.07668062049734514, "ret_signed_p6m": 0.07668062049734514, "alpha_spy_p6m": -0.07064224239163774, "alpha_c_p6m": -0.1310560315595255, "bench_spy_p6m": 0.1473228628889829, "bench_c_p6m": 0.20773665205687064, "price_path": [0.4476, 0.4859, 0.491, 0.4706, 0.4783, 0.468, 0.5166, 0.491, 0.4655, 0.4169, 0.4169, 0.3939, 0.4143, 0.3913, 0.3555, 0.3299, 0.2941, 0.2711, 0.3427, 0.3171, 0.3043, 0.2583, 0.2992, 0.2864, 0.2737, 0.2532, 0.2353, 0.3069, 0.2992, 0.2788, 0.2506, 0.1509, 0.2404, 0.2506, 0.2506, 0.2506, 0.1279, 0.0153, -0.0844, -0.0486, -0.0921, -0.0307, -0.0102, -0.0588, -0.0179, -0.046, -0.1407, -0.1381, -0.0793, -0.0742, -0.0486, -0.0563, -0.0486, -0.0486, -0.0486, -0.0102, -0.0767, -0.0512, -0.0895, -0.023, -0.0281, -0.0281, 0.0, 0.0, -0.0205, -0.023, -0.0486, -0.0384, -0.023, -0.0435, -0.0537, -0.0614, -0.0793, -0.0972, -0.0793, -0.0793, -0.1125, -0.11, -0.1023, -0.0435, -0.0588, -0.0895, -0.0639, -0.0205, -0.0281, -0.0409, -0.0563, -0.0665, 0.0256, 0.1279, 0.1151, 0.1023, 0.1535, 0.1202, 0.1355, 0.1228, 0.0972, 0.0895, 0.0818, 0.0895, 0.0639, 0.0563, 0.0537, 0.0767, 0.0844, 0.0921, 0.0537, 0.0716, 0.0614, 0.0972, 0.2046, 0.2123, 0.179, 0.2353, 0.2481, 0.2174, 0.1995, 0.1969, 0.202, 0.2046, 0.2046, 0.1714, 0.1765, 0.2631, 0.2474, 0.2264, 0.2054, 0.187, 0.2132, 0.25, 0.229, 0.229, 0.2027, 0.1555, 0.1607, 0.1423, 0.2553, 0.25, 0.2868, 0.3603, 0.3866, 0.25, 0.2159, 0.2106, 0.2106, 0.2316, 0.2264, 0.2185, 0.2106, 0.2054, 0.187, 0.166, 0.1817, 0.1555, 0.1581, 0.2264, 0.2526, 0.2447, 0.2553, 0.2579, 0.2421, 0.2421, 0.3655, 0.3761, 0.3655, 0.3262, 0.3262, 0.3472, 0.3393, 0.3235, 0.3235, 0.1922, 0.1633, 0.1581, 0.1607, 0.1528, 0.1581, 0.1817, 0.1423, 0.1003, 0.1003, 0.1318, 0.1029, 0.0977, 0.1161, 0.1108, 0.1056, 0.0636, 0.0741, 0.0767]}
{"unit_id": "orig:1934941584690459105", "ticker_idx": 0, "ts": "2025-06-17T11:49:57+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek's stock price may remain under pressure in the near term due to a slowdown in the mobile/IoT business, continuous delays in AI ASIC projects, and the impact of TWD appreciation on profitability. Full-year revenue and earnings forecasts face downside risks.", "tweet_type": "news_reaction", "timeline": "months", "summary": "JPM bearish on MediaTek: 2H25 mobile/IoT slowdown, ASIC project delays cut 2026 target to $600M vs $1B, TWD headwinds, full-year revenue and earnings at downside risk.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "MediaTek: 2H25 Growth Momentum Slows, ASIC Project Delays Add to Downside Pressure\nJPM (G. Hariharan, 06/16/2025)\n\nJPM notes that MediaTek's stock price may remain under pressure in the near term due to a slowdown in the mobile/IoT business, continuous delays in AI ASIC projects, and the impact of TWD appreciation on profitability. Full-year revenue and earnings forecasts face downside risks.\n\n Mobile and Smart Edge businesses to slow significantly in 2H25, with weakening YoY growth momentum\n\nJPM forecasts that after achieving 16% YoY growth in the first half of the year, driven by tariff front-loading and Chinese consumer subsidies, MediaTek will face inventory destocking pressure and unfavorable currency effects in 2H25. Revenue from the Mobile and Smart Edge businesses is expected to decline by 13% and 10% QoQ, respectively. China's smartphone end-demand remains weak, and the full-year YoY growth for the mobile business is now expected to be only 2%, down from the previous forecast of 10%.\n\n ASIC project delayed again, putting the $1 billion revenue target for 2026 at downside risk\n\nThe tape-out of the 3nm data center ASIC project with a U.S. CSP client is now expected to be delayed to 3Q or 4Q25, which could push mass production to late 2026. JPM has lowered its 2026 revenue forecast for this project to $600 million, making the original $1 billion target likely unattainable. Furthermore, while a follow-up 2nm project may be won, Broadcom will continue to dominate the client's high-performance products, and MediaTek's 2027 revenue forecast also faces downward revisions.\n\nLimited breakthrough in the high-end phone market, with no significant progress in Samsung collaboration\n   \nAlthough flagship market share is expected to grow by 20%, JPM believes MediaTek will struggle to make a breakthrough in 2025, expecting only one new Honor flagship model to adopt its SoC. Samsung has not yet opened a window for collaboration, and high-end growth expectations for 2026 are weak.\n\nXiaomi's self-developed SoC has limited short-term impact, but the long-term competitive threat should not be ignored\n   \nXiaomi's launch of its self-developed SoC, the X-Ring O1, has a limited impact on MediaTek's fundamentals in the short term (shipment estimate of only 1 million units). However, if Xiaomi achieves a breakthrough in its self-developed 5G modem, it could compress MediaTek's market space in commercial SoCs in the long run, posing a potential competitive threat, especially in the flagship market.\n \nSlow progress in ASIC customer expansion, with no major breakthroughs expected in the short term\n   \nAlthough MediaTek has engaged with several potential ASIC clients, including Chinese and U.S. CSPs and automotive manufacturers, JPM expects it will be difficult to secure large-scale orders within the next 6 months. Only some small-scale custom chip projects (such as video processors) have entered the initial stages of collaboration, with limited overall project scale. In the future, the company may secure orders from Japanese automakers for automotive ADAS-related ASICs, but mass production might not occur until 2027-28.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01568631900046713, "ret_signed_m1d": 0.01568631900046713, "alpha_spy_m1d": 0.024304998433221203, "alpha_c_m1d": 0.022529114340817613, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007843059981644807, "ret_signed_p1d": -0.007843059981644807, "alpha_spy_p1d": -0.007993695735815987, "alpha_c_p1d": -0.004020325512951839, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": -0.0039216295094112175, "ret_signed_p1w": 0.0039216295094112175, "alpha_spy_p1w": 0.022403954655963454, "alpha_c_p1w": 0.04195822365498458, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.12765179719624298, "ret_signed_p1m": -0.12765179719624298, "alpha_spy_p1m": -0.07989644040595412, "alpha_c_p1m": -0.021799172702843705, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": 0.18763324295055095, "ret_signed_p3m": -0.18763324295055095, "alpha_spy_p3m": -0.08416827965949247, "alpha_c_p3m": -0.024285947769822513, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 0.16763946087197779, "ret_signed_p6m": -0.16763946087197779, "alpha_spy_p6m": -0.01035114414181093, "alpha_c_p6m": 0.26246111498401237, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [0.1647, 0.1608, 0.1608, 0.1961, 0.1882, 0.1647, 0.149, 0.0902, 0.1451, 0.1255, 0.1255, 0.1255, 0.0157, -0.0118, -0.0706, 0.0196, 0.0863, 0.098, 0.0863, 0.0706, 0.051, 0.0706, 0.0549, 0.0196, 0.0745, 0.0431, 0.0824, 0.0588, 0.0745, 0.0588, 0.0588, 0.0196, 0.0157, 0.0039, 0.0, 0.0118, 0.0392, 0.0314, 0.0627, 0.0667, 0.0745, 0.0706, 0.0275, 0.0314, 0.051, 0.0392, 0.0314, 0.0039, 0.0039, 0.0039, -0.0118, -0.0118, -0.0118, 0.0, 0.0, -0.0235, 0.0039, 0.0078, 0.0471, 0.0314, 0.0078, -0.0196, -0.0157, 0.0, 0.0078, -0.0078, -0.0196, -0.0157, -0.0039, 0.0157, 0.0118, 0.0078, -0.0196, 0.0, 0.0157, 0.0157, 0.0317, 0.0237, 0.0157, 0.0837, 0.1197, 0.1356, 0.1077, 0.1157, 0.1277, 0.1117, 0.1277, 0.1396, 0.1436, 0.1436, 0.1436, 0.1436, 0.1197, 0.0877, 0.1037, 0.0957, 0.0957, 0.0637, 0.0717, 0.0557, 0.0837, 0.0797, 0.0837, 0.0877, 0.0957, 0.1117, 0.0957, 0.1277, 0.1117, 0.0837, 0.0917, 0.0917, 0.1277, 0.1197, 0.1197, 0.1077, 0.0957, 0.0877, 0.0877, 0.1157, 0.1077, 0.1476, 0.1436, 0.2076, 0.1916, 0.1836, 0.1876, 0.1876, 0.2276, 0.2076, 0.2076, 0.1516, 0.1237, 0.0997, 0.0637, 0.0757, 0.0477, 0.0477, 0.0517, 0.0277, 0.0237, 0.0557, 0.0557, 0.0597, 0.0677, 0.0757, 0.0757, 0.0517, 0.0437, 0.0517, 0.0637, 0.0637, 0.0717, 0.0717, 0.0637, 0.0357, 0.0357, 0.0677, 0.0517, 0.0397, 0.0477, 0.0477, 0.0397, 0.0517, 0.0637, 0.0317, 0.0077, -0.0003, -0.0083, -0.0043, -0.0043, -0.0163, -0.0163, -0.0643, -0.0723, -0.0523, -0.0843, -0.0803, -0.0523, 0.0397, 0.0717, 0.1157, 0.1556, 0.1317, 0.1197, 0.1237, 0.1396, 0.1516, 0.1356, 0.1676]}
{"unit_id": "quote:1928787309731729598", "ticker_idx": 0, "ts": "2025-05-31T12:15:04+00:00", "symbol": "CPNG", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "As a Korean, I really do use Coupang a lot. I even ordered delivery food through Coupang today", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Korean user endorses CPNG as a heavily-used platform, implicitly backing the quoted bullish thesis.", "resolved_tickers": ["CPNG"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "quote", "lang": "en", "tweet_text": "As a Korean, I really do use Coupang a lot.\n\nI even ordered delivery food through Coupang today, haha.", "reply_text": null, "parent_text": null, "quoted_text": "$CPNG is my first purchase of 2025. \n\nIt is entering the profitability inflection phase, margins are expanding and is set for a strong decade ahead.\n\nHere is my $CPNG investment thesis: 🧵 https://t.co/36XahOuiJr", "parent_author": null, "ret_m1d": -0.010232918629707988, "ret_signed_m1d": -0.010232918629707988, "alpha_spy_m1d": -0.004631569339677788, "alpha_c_m1d": -0.00789974358761203, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.0023331750420959585, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004234327402515614, "ret_signed_p1d": -0.004234327402515614, "alpha_spy_p1d": -0.00993691719070311, "alpha_c_p1d": -0.009554032522948841, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.005319705120433227, "ret_p1w": 0.004587126325570345, "ret_signed_p1w": 0.004587126325570345, "alpha_spy_p1w": -0.007172290837464734, "alpha_c_p1w": 0.0065936956152570225, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": -0.002006569289686677, "ret_p1m": 0.0561044510948141, "ret_signed_p1m": 0.0561044510948141, "alpha_spy_p1m": 0.010945876915161579, "alpha_c_p1m": 0.035460183032945514, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.020644268061868587, "ret_p3m": 0.010232851327357073, "ret_signed_p3m": 0.010232851327357073, "alpha_spy_p3m": -0.0878391942025365, "alpha_c_p3m": -0.08471741898178298, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.09495027030914005, "ret_p6m": -0.020465769957065172, "ret_signed_p6m": -0.020465769957065172, "alpha_spy_p6m": -0.1658698315133711, "alpha_c_p6m": -0.1138338216278032, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.09336805167073803, "price_path": [-0.1538, -0.1824, -0.1972, -0.2269, -0.2244, -0.2142, -0.2174, -0.2057, -0.1701, -0.1934, -0.1694, -0.1775, -0.1764, -0.1676, -0.1743, -0.168, -0.1701, -0.211, -0.2262, -0.2174, -0.2064, -0.2428, -0.2922, -0.3028, -0.3006, -0.2294, -0.259, -0.2361, -0.2403, -0.2421, -0.2509, -0.2452, -0.2452, -0.2544, -0.235, -0.2036, -0.18, -0.1729, -0.1736, -0.169, -0.1754, -0.1718, -0.1496, -0.1535, -0.1531, -0.0614, -0.0956, -0.0826, -0.0776, -0.0628, -0.0505, -0.0529, -0.0388, -0.0416, -0.043, -0.0339, -0.0374, -0.0314, -0.0314, 0.0039, -0.0014, -0.0159, -0.0102, 0.0, -0.0042, 0.0004, 0.0071, 0.0039, 0.0046, -0.0028, -0.0021, -0.0011, -0.0085, 0.0, -0.0032, 0.0021, 0.0021, 0.0074, 0.0014, 0.0215, 0.0176, 0.0519, 0.0872, 0.0572, 0.0561, 0.0596, 0.0688, 0.0688, 0.0586, 0.0582, 0.0646, 0.0745, 0.0614, 0.0801, 0.0984, 0.1016, 0.0992, 0.1112, 0.1076, 0.0861, 0.0776, 0.0635, 0.0674, 0.0607, 0.0526, 0.0392, 0.0385, 0.0141, 0.0455, 0.055, -0.0088, -0.0187, -0.018, -0.0236, -0.0071, 0.0046, -0.0004, -0.0071, 0.0226, 0.018, 0.0176, -0.0042, 0.0145, 0.0071, 0.0028, -0.0035, 0.0102, 0.0085, 0.0085, 0.0025, -0.0004, 0.0138, 0.0201, 0.1115, 0.1408, 0.1334, 0.145, 0.1436, 0.1821, 0.1754, 0.1831, 0.1793, 0.1595, 0.1404, 0.1376, 0.1306, 0.1239, 0.1359, 0.1584, 0.1362, 0.1443, 0.1489, 0.1418, 0.1443, 0.1422, 0.1344, 0.1457, 0.102, 0.1214, 0.1207, 0.1133, 0.1115, 0.1013, 0.1182, 0.109, 0.0984, 0.0995, 0.0992, 0.12, 0.1263, 0.1316, 0.1274, 0.1281, 0.1284, 0.1337, 0.0663, 0.0233, 0.0191, 0.0205, 0.0109, 0.0173, 0.0067, -0.0042, -0.0187, -0.0247, -0.0332, -0.0628, -0.0593, -0.0293, -0.0205]}
{"unit_id": "orig:1928776593784320348", "ticker_idx": 0, "ts": "2025-05-31T11:32:29+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain NVIDIA as our top pick with a Buy rating. We revise our FY26/27/28 pro forma EPS estimates by +6%/+2%/+12% to $4.21/$5.87/$7.23, respectively, and raise our PO from $160 to $180", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares BofA Buy on NVDA, PT raised to $180 on Blackwell ramp + EPS upside; flags gradual headwind to AVGO from NVDA networking share gains.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) NVIDIA – Best-in-Class Opportunity Capture and Execution; Raising Estimates and Price Objective (PO) to $180\n\n1. Blackwell Back on Track; Validated Long-Term $10 EPS Potential\nKey Takeaways from 1Q Earnings:\n1. China Risk Mitigated: $15B in 1H revenue from H20 products already factored into models.\n2. Blackwell Rack Ramp Reaching Full Production: All major hyperscalers are currently ramping production at ~1,000 racks per week, or ~13,000 racks per quarter (assuming ASP of $2.5M per rack, this translates to over $30B per quarter, or over $100B from top hyperscalers combined; NVIDIA did not disclose incremental figures).\n3. Gross Margin (GM) Recovery Confidence: NVIDIA expressed confidence in gross margin returning to mid-70% range within the year — signaling both demand recovery and improved rack-scale execution capabilities.\n\nThe networking segment also returned to a low/mid-teens share of 1Q data center revenue. This indicates that NVIDIA’s full rack-scale portfolio — including NVLink (scale-up), Quantum/Spectrum (scale-out), and BlueField (DPU/NIC) — is now expanding meaningfully across several hyperscalers (including new Spectrum-X wins from Google and Meta). This poses a gradual headwind to AVGO (Broadcom). We maintain NVIDIA as our top pick with a Buy rating. We revise our FY26/27/28 pro forma EPS estimates by +6%/+2%/+12% to $4.21/$5.87/$7.23, respectively, and raise our PO from $160 to $180 by applying a 30x P/E (previously 28x), reflecting faster EPS growth.\n\n⸻\n\n2. Positive Factors: Long-Term $10 EPS Potential, AI Capex Diversification\nOur FY28 (i.e., CY27) pro forma EPS estimate of $7.23 aligns with $277B in data center revenue and a $325B–$350B TAM (assuming 80–85% NVIDIA share). If the TAM estimate is raised to $450B–$500B while maintaining NVIDIA’s share, conceptually, this opens a path to $10+ EPS in CY27/FY28.\nSecond, we highlight NVIDIA’s 50% FCF margin (2.6x the Mag-6 avg of 19%) and attractive 0.9x PEG ratio (vs. Mag-6 avg of 3x).\nThird, AI Capex is diversifying beyond US cloud names — with multiple sovereign-led builds (e.g., Saudi Arabia; see related report).\n\n⸻\n\n3. Risks: Accelerated Product Cadence Increases Execution Risk\nWhile NVIDIA’s annual (October) next-gen product cadence is admirable, it increases execution risk — similar to the ~2-quarter delay seen with Blackwell rack rollout.\nSecond, (uneven) deployment of power-stable data centers may pose a bottleneck to AI builds — not just chip/system availability.\nFinally, headlines or real risks from use/restriction of AI products as bargaining tools in global trade negotiations cannot be ignored.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016377900094598385, "ret_signed_m1d": -0.016377900094598385, "alpha_spy_m1d": -0.010776550804568186, "alpha_c_m1d": -0.0017868054989061966, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027951559872750442, "ret_signed_p1d": 0.027951559872750442, "alpha_spy_p1d": 0.022248970084562947, "alpha_c_p1d": 0.005386862266536285, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": 0.03821517428480492, "ret_signed_p1w": 0.03821517428480492, "alpha_spy_p1w": 0.02645575712176984, "alpha_c_p1w": -0.017929386337546704, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.11596045644949049, "ret_signed_p1m": 0.11596045644949049, "alpha_spy_p1m": 0.07080188226983797, "alpha_c_p1m": -0.01778379953220588, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.3115628279071214, "ret_signed_p3m": 0.3115628279071214, "alpha_spy_p3m": 0.21349078237722785, "alpha_c_p3m": 0.08312064926844309, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 0.29452879416834277, "ret_signed_p6m": 0.29452879416834277, "alpha_spy_p6m": 0.14912473261203685, "alpha_c_p6m": -0.10296403538305898, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [-0.1462, -0.1952, -0.1798, -0.2214, -0.2084, -0.1575, -0.1587, -0.1144, -0.1299, -0.1598, -0.1446, -0.1372, -0.1433, -0.1162, -0.1215, -0.1719, -0.1889, -0.2017, -0.2111, -0.1982, -0.1962, -0.259, -0.3135, -0.2893, -0.299, -0.1678, -0.217, -0.1925, -0.1941, -0.1833, -0.2394, -0.2612, -0.2612, -0.2946, -0.2802, -0.2524, -0.2253, -0.1919, -0.2085, -0.2064, -0.2072, -0.1876, -0.1665, -0.1715, -0.1735, -0.1479, -0.1457, -0.1509, -0.1047, -0.0542, -0.0148, -0.0186, -0.0144, -0.0132, -0.0218, -0.0406, -0.0331, -0.0443, -0.0443, -0.0137, -0.0187, 0.0132, -0.0164, 0.0, 0.028, 0.033, 0.019, 0.0316, 0.0382, 0.0479, 0.0397, 0.0555, 0.0335, 0.0533, 0.0491, 0.059, 0.059, 0.0472, 0.0495, 0.0767, 0.1233, 0.1285, 0.1484, 0.1501, 0.116, 0.1447, 0.1599, 0.1599, 0.1519, 0.1647, 0.1857, 0.1946, 0.2005, 0.1944, 0.2426, 0.2475, 0.2594, 0.2551, 0.2476, 0.2159, 0.2432, 0.2648, 0.263, 0.2867, 0.2776, 0.305, 0.2948, 0.2646, 0.3103, 0.2977, 0.3061, 0.3159, 0.33, 0.3253, 0.3333, 0.3219, 0.325, 0.3136, 0.325, 0.2786, 0.2768, 0.2738, 0.2957, 0.3089, 0.3232, 0.322, 0.3116, 0.268, 0.268, 0.2432, 0.242, 0.2496, 0.2158, 0.2252, 0.2431, 0.2909, 0.2898, 0.2945, 0.294, 0.2731, 0.2397, 0.283, 0.2862, 0.3367, 0.299, 0.2883, 0.2936, 0.2972, 0.3239, 0.3583, 0.3631, 0.3751, 0.3659, 0.3507, 0.3471, 0.3767, 0.4019, 0.3334, 0.371, 0.3106, 0.3092, 0.3236, 0.3338, 0.3296, 0.3188, 0.3124, 0.3261, 0.356, 0.394, 0.4635, 0.5073, 0.477, 0.4741, 0.5061, 0.4465, 0.4211, 0.3692, 0.3697, 0.4491, 0.4062, 0.4109, 0.3603, 0.3844, 0.3584, 0.3203, 0.3579, 0.3151, 0.3022, 0.329, 0.2945]}
{"unit_id": "orig:1928776593784320348", "ticker_idx": 1, "ts": "2025-05-31T11:32:29+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This poses a gradual headwind to AVGO (Broadcom)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares BofA Buy on NVDA, PT raised to $180 on Blackwell ramp + EPS upside; flags gradual headwind to AVGO from NVDA networking share gains.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) NVIDIA – Best-in-Class Opportunity Capture and Execution; Raising Estimates and Price Objective (PO) to $180\n\n1. Blackwell Back on Track; Validated Long-Term $10 EPS Potential\nKey Takeaways from 1Q Earnings:\n1. China Risk Mitigated: $15B in 1H revenue from H20 products already factored into models.\n2. Blackwell Rack Ramp Reaching Full Production: All major hyperscalers are currently ramping production at ~1,000 racks per week, or ~13,000 racks per quarter (assuming ASP of $2.5M per rack, this translates to over $30B per quarter, or over $100B from top hyperscalers combined; NVIDIA did not disclose incremental figures).\n3. Gross Margin (GM) Recovery Confidence: NVIDIA expressed confidence in gross margin returning to mid-70% range within the year — signaling both demand recovery and improved rack-scale execution capabilities.\n\nThe networking segment also returned to a low/mid-teens share of 1Q data center revenue. This indicates that NVIDIA’s full rack-scale portfolio — including NVLink (scale-up), Quantum/Spectrum (scale-out), and BlueField (DPU/NIC) — is now expanding meaningfully across several hyperscalers (including new Spectrum-X wins from Google and Meta). This poses a gradual headwind to AVGO (Broadcom). We maintain NVIDIA as our top pick with a Buy rating. We revise our FY26/27/28 pro forma EPS estimates by +6%/+2%/+12% to $4.21/$5.87/$7.23, respectively, and raise our PO from $160 to $180 by applying a 30x P/E (previously 28x), reflecting faster EPS growth.\n\n⸻\n\n2. Positive Factors: Long-Term $10 EPS Potential, AI Capex Diversification\nOur FY28 (i.e., CY27) pro forma EPS estimate of $7.23 aligns with $277B in data center revenue and a $325B–$350B TAM (assuming 80–85% NVIDIA share). If the TAM estimate is raised to $450B–$500B while maintaining NVIDIA’s share, conceptually, this opens a path to $10+ EPS in CY27/FY28.\nSecond, we highlight NVIDIA’s 50% FCF margin (2.6x the Mag-6 avg of 19%) and attractive 0.9x PEG ratio (vs. Mag-6 avg of 3x).\nThird, AI Capex is diversifying beyond US cloud names — with multiple sovereign-led builds (e.g., Saudi Arabia; see related report).\n\n⸻\n\n3. Risks: Accelerated Product Cadence Increases Execution Risk\nWhile NVIDIA’s annual (October) next-gen product cadence is admirable, it increases execution risk — similar to the ~2-quarter delay seen with Blackwell rack rollout.\nSecond, (uneven) deployment of power-stable data centers may pose a bottleneck to AI builds — not just chip/system availability.\nFinally, headlines or real risks from use/restriction of AI products as bargaining tools in global trade negotiations cannot be ignored.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02669779229960967, "ret_signed_m1d": 0.02669779229960967, "alpha_spy_m1d": 0.02109644300957947, "alpha_c_m1d": 0.01210669770391748, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03272891970555247, "ret_signed_p1d": -0.03272891970555247, "alpha_spy_p1d": -0.027026329917364977, "alpha_c_p1d": -0.010164222099338316, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": -0.01781197001345769, "ret_signed_p1w": 0.01781197001345769, "alpha_spy_p1w": 0.02957138717649277, "alpha_c_p1w": 0.07395653063580931, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.06695785923365705, "ret_signed_p1m": -0.06695785923365705, "alpha_spy_p1m": -0.021799285054004525, "alpha_c_p1m": 0.06678639674803932, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.24392444966750992, "ret_signed_p3m": -0.24392444966750992, "alpha_spy_p3m": -0.14585240413761635, "alpha_c_p3m": -0.015482271028831596, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 0.5544108023198013, "ret_signed_p6m": -0.5544108023198013, "alpha_spy_p6m": -0.4090067407634954, "alpha_c_p6m": -0.15691797276839958, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [-0.232, -0.2807, -0.2185, -0.2606, -0.238, -0.2214, -0.2329, -0.2162, -0.2203, -0.2437, -0.216, -0.2339, -0.2294, -0.231, -0.2431, -0.2792, -0.3085, -0.32, -0.3268, -0.3224, -0.3081, -0.3808, -0.4118, -0.3802, -0.3726, -0.2556, -0.3072, -0.2685, -0.2829, -0.2805, -0.2979, -0.3125, -0.3125, -0.3317, -0.3182, -0.2887, -0.2435, -0.2268, -0.2261, -0.2314, -0.2261, -0.2066, -0.1812, -0.193, -0.1955, -0.1765, -0.1646, -0.1629, -0.1091, -0.0655, -0.0667, -0.0646, -0.0808, -0.0727, -0.0685, -0.0763, -0.0731, -0.0804, -0.0804, -0.0525, -0.0373, -0.0271, -0.0267, 0.0, 0.0327, 0.0497, 0.0451, -0.0072, -0.0178, -0.0164, 0.0169, 0.0296, -0.0, 0.0136, 0.0027, 0.0103, 0.0103, 0.0075, 0.0227, 0.063, 0.0666, 0.0888, 0.0855, 0.1109, 0.067, 0.0878, 0.109, 0.109, 0.105, 0.0954, 0.12, 0.1099, 0.1058, 0.1107, 0.1322, 0.1317, 0.1545, 0.1419, 0.1615, 0.1228, 0.1433, 0.1636, 0.1695, 0.1861, 0.1987, 0.2196, 0.1837, 0.1633, 0.1999, 0.1806, 0.2158, 0.2242, 0.2291, 0.2248, 0.2608, 0.2457, 0.2543, 0.2346, 0.2323, 0.1885, 0.1735, 0.1671, 0.1849, 0.1858, 0.201, 0.2101, 0.2439, 0.1985, 0.1985, 0.202, 0.2187, 0.2336, 0.3497, 0.393, 0.3569, 0.4894, 0.4494, 0.4504, 0.4674, 0.4509, 0.3951, 0.3918, 0.3902, 0.3677, 0.3683, 0.3698, 0.3569, 0.3505, 0.3238, 0.3319, 0.3459, 0.3653, 0.366, 0.3544, 0.3581, 0.3948, 0.3929, 0.3106, 0.44, 0.3893, 0.4184, 0.4297, 0.4103, 0.4099, 0.3834, 0.3738, 0.3899, 0.4297, 0.4616, 0.5057, 0.5582, 0.5199, 0.4922, 0.4637, 0.4208, 0.4492, 0.4356, 0.4107, 0.4469, 0.4209, 0.4341, 0.3725, 0.3826, 0.3833, 0.3746, 0.4308, 0.4002, 0.3734, 0.5259, 0.5544]}
{"unit_id": "orig:1938021999420575914", "ticker_idx": 0, "ts": "2025-06-25T23:50:26+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix target price raised from ₩290,000 to ₩350,000", "tweet_type": "news_reaction", "timeline": "months", "summary": "CLSA raises SK Hynix TP to ₩350K on HBM leadership; Samsung positive on HBM3E approval and governance restructuring catalysts.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Korea Memory Sector Outlook: Micron Call Read-Across-CLSA\n\n📌 Overall Summary (Korea Memory Sector Outlook)\n\n- Micron's strong earnings and guidance provide positive implications for SK Hynix and Samsung Electronics.\n- Strong demand for AI server DRAM and HBM continues.\n- SK Hynix target price raised from ₩290,000 to ₩350,000.\n- For Samsung Electronics, HBM3E approval/governance restructuring could act as short-term stock catalysts.\n\n💡 Micron Earnings and Guidance Summary\n\nFY3Q25 Earnings (Micron):\n- Revenue: $9.3B, +5% vs. consensus\n- EPS: $1.91, +18% vs. consensus\n- DRAM Revenue: $7.1B, Bit shipments +20%↑\n- HBM Revenue: +50% QoQ\n- NAND Revenue: $2.1B, Bit shipments +25%↑\n\nFY4Q25 Guidance (Micron):\n- Revenue: $10.7B (+8% vs. consensus)\n- EPS: $2.5 (+25% vs. consensus)\n- GPM: 42%\n- FY25 Capex: Maintain $14B\n📊 HBM and AI Demand Outlook\n- HBM shipment customers: Expanded from 2 companies to 4 companies (GPU and ASIC).\n- Micron's HBM market share: Expected to reach 24% in 2H25 (previously forecast for end of 2025).\n- 2025 HBM TAM: Expected to be $35B (rapid growth from $18B in 2024).\n- Enterprise eSSD demand also expected to recover from 2H25.\n\n🧠 SK Hynix: Strong HBM Leadership & Earnings Estimates\n- 70% of HBM supplied to Nvidia.\n- 2025 HBM shipments expected to grow +95%, 2026 +40%.\n- HBM proportion of DRAM revenue expected to reach 55-60% by 2027-2028.\n- 2025 global cloud Capex: Expected to grow +35%.\n\n💵 SK Hynix Earnings Estimates (CLSA Standard)\n\n📈 Annual Outlook\n- 2025E Revenue: ₩86.2 trillion (+30% YoY)\n- 2025E Operating Profit: ₩38.6 trillion (OPM 45%)\n- 2025E Net Income: ₩33.4 trillion\n- DRAM ASP: +24%, NAND ASP: -11%\n🔮 Multiples & Target Price\n- BVPS (as of 4Q 2026): ₩202,531\n- Target P/B: 1.7x (average of past +2σ)\n- → Target Price: ₩350,000\n\n🏭 Samsung Electronics: HBM3E & Non-Memory Strategy are Key\n\n🧾 Investment Points\n- Global No. 1 DRAM/NAND market share.\n- Increasing demand for high-value products related to HBM and AI.\n- Growth potential in non-memory (Foundry, S.LSI), OLED, CE, and other divisions.\n\n🎯 Catalysts\n- ₩10 trillion share buyback.\n- HBM3E approval (Nvidia) expected.\n- Emergence of governance restructuring scenarios.\n\n💲 Valuation\n\n- SOTP (Sum-of-the-parts) methodMemory: Average of Hynix and Micron EV/EBITDA.\n- Foundry: 50% discount to TSMC.\n- Mobile: 50% discount to Apple.\n- Display: Premium granted.\n- CE: 10% premium compared to peers.\n\n⚠️ Risk Factors\n\nSK Hynix faces key risks such as U.S.-China trade and geopolitical tensions, weak NAND demand, and the expansion of Chinese competitors.\n\nSamsung Electronics is exposed to risks like delayed HBM qualification and a slow recovery in memory demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026223746874970866, "ret_signed_m1d": -0.026223746874970866, "alpha_spy_m1d": -0.025663696077025233, "alpha_c_m1d": -0.012924312773072266, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024475540972238585, "ret_signed_p1d": 0.024475540972238585, "alpha_spy_p1d": 0.01665176109775124, "alpha_c_p1d": 0.017208066936997835, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": -0.024475540972238585, "ret_signed_p1w": -0.024475540972238585, "alpha_spy_p1w": -0.04643168161393074, "alpha_c_p1w": -0.04645950253926667, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": -0.05769233090280079, "ret_signed_p1m": -0.05769233090280079, "alpha_spy_p1m": -0.10265871214024735, "alpha_c_p1m": -0.10347720597372856, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.22904547606599368, "ret_signed_p3m": 0.22904547606599368, "alpha_spy_p3m": 0.1276343082434872, "alpha_c_p3m": 0.05836957194775105, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.934242090254298, "ret_signed_p6m": 0.934242090254298, "alpha_spy_p6m": 0.8169252364461097, "alpha_c_p6m": 0.6724324866610625, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3044, -0.3344, -0.3124, -0.3093, -0.3208, -0.3641, -0.4248, -0.4084, -0.4241, -0.3606, -0.369, -0.3711, -0.3697, -0.3927, -0.3892, -0.3892, -0.3836, -0.3934, -0.3683, -0.3777, -0.3564, -0.3648, -0.369, -0.3805, -0.3805, -0.3508, -0.3508, -0.3508, -0.3341, -0.3358, -0.3365, -0.3194, -0.3072, -0.281, -0.3002, -0.2863, -0.3041, -0.295, -0.3002, -0.3128, -0.302, -0.2915, -0.2932, -0.274, -0.2587, -0.285, -0.2745, -0.2745, -0.2395, -0.215, -0.215, -0.1993, -0.1941, -0.1608, -0.1766, -0.1766, -0.1329, -0.1294, -0.1381, -0.1399, -0.1014, -0.0927, -0.0262, 0.0, 0.0245, -0.007, 0.021, -0.0017, -0.0245, -0.0262, -0.0542, -0.0524, -0.014, -0.0175, 0.0385, 0.0297, 0.049, 0.0437, 0.035, -0.0577, -0.0594, -0.0472, -0.0612, -0.0594, -0.0577, -0.0699, -0.0839, -0.0822, -0.0787, -0.0437, -0.0979, -0.0979, -0.0787, -0.0962, -0.0839, -0.1031, -0.0664, -0.0594, -0.028, -0.0332, -0.0332, -0.0647, -0.0804, -0.1066, -0.1434, -0.1224, -0.0927, -0.0857, -0.0909, -0.0598, -0.0581, -0.1036, -0.0878, -0.0808, -0.0703, -0.0423, -0.0301, 0.0084, 0.0645, 0.075, 0.1503, 0.159, 0.2185, 0.1678, 0.2448, 0.2448, 0.229, 0.2641, 0.2518, 0.2483, 0.1783, 0.222, 0.2168, 0.2606, 0.3849, 0.3849, 0.3849, 0.3849, 0.3849, 0.3849, 0.4987, 0.4531, 0.4409, 0.4794, 0.5845, 0.63, 0.7, 0.6772, 0.686, 0.6755, 0.7858, 0.8733, 0.8243, 0.9539, 0.9889, 0.9574, 1.171, 1.0519, 1.0274, 1.0764, 1.0309, 1.1219, 1.1675, 1.1605, 1.143, 0.9609, 1.1219, 0.9959, 0.9679, 0.9994, 0.8243, 0.8208, 0.8173, 0.8348, 0.9062, 0.8572, 0.8852, 0.9553, 0.9342, 0.8992, 0.9062, 1.0218, 0.9833, 1.0569, 0.9798, 1.0008, 0.9413, 0.8572, 0.9307, 0.9342]}
{"unit_id": "orig:1938021999420575914", "ticker_idx": 1, "ts": "2025-06-25T23:50:26+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "For Samsung Electronics, HBM3E approval/governance restructuring could act as short-term stock catalysts", "tweet_type": "news_reaction", "timeline": "months", "summary": "CLSA raises SK Hynix TP to ₩350K on HBM leadership; Samsung positive on HBM3E approval and governance restructuring catalysts.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Korea Memory Sector Outlook: Micron Call Read-Across-CLSA\n\n📌 Overall Summary (Korea Memory Sector Outlook)\n\n- Micron's strong earnings and guidance provide positive implications for SK Hynix and Samsung Electronics.\n- Strong demand for AI server DRAM and HBM continues.\n- SK Hynix target price raised from ₩290,000 to ₩350,000.\n- For Samsung Electronics, HBM3E approval/governance restructuring could act as short-term stock catalysts.\n\n💡 Micron Earnings and Guidance Summary\n\nFY3Q25 Earnings (Micron):\n- Revenue: $9.3B, +5% vs. consensus\n- EPS: $1.91, +18% vs. consensus\n- DRAM Revenue: $7.1B, Bit shipments +20%↑\n- HBM Revenue: +50% QoQ\n- NAND Revenue: $2.1B, Bit shipments +25%↑\n\nFY4Q25 Guidance (Micron):\n- Revenue: $10.7B (+8% vs. consensus)\n- EPS: $2.5 (+25% vs. consensus)\n- GPM: 42%\n- FY25 Capex: Maintain $14B\n📊 HBM and AI Demand Outlook\n- HBM shipment customers: Expanded from 2 companies to 4 companies (GPU and ASIC).\n- Micron's HBM market share: Expected to reach 24% in 2H25 (previously forecast for end of 2025).\n- 2025 HBM TAM: Expected to be $35B (rapid growth from $18B in 2024).\n- Enterprise eSSD demand also expected to recover from 2H25.\n\n🧠 SK Hynix: Strong HBM Leadership & Earnings Estimates\n- 70% of HBM supplied to Nvidia.\n- 2025 HBM shipments expected to grow +95%, 2026 +40%.\n- HBM proportion of DRAM revenue expected to reach 55-60% by 2027-2028.\n- 2025 global cloud Capex: Expected to grow +35%.\n\n💵 SK Hynix Earnings Estimates (CLSA Standard)\n\n📈 Annual Outlook\n- 2025E Revenue: ₩86.2 trillion (+30% YoY)\n- 2025E Operating Profit: ₩38.6 trillion (OPM 45%)\n- 2025E Net Income: ₩33.4 trillion\n- DRAM ASP: +24%, NAND ASP: -11%\n🔮 Multiples & Target Price\n- BVPS (as of 4Q 2026): ₩202,531\n- Target P/B: 1.7x (average of past +2σ)\n- → Target Price: ₩350,000\n\n🏭 Samsung Electronics: HBM3E & Non-Memory Strategy are Key\n\n🧾 Investment Points\n- Global No. 1 DRAM/NAND market share.\n- Increasing demand for high-value products related to HBM and AI.\n- Growth potential in non-memory (Foundry, S.LSI), OLED, CE, and other divisions.\n\n🎯 Catalysts\n- ₩10 trillion share buyback.\n- HBM3E approval (Nvidia) expected.\n- Emergence of governance restructuring scenarios.\n\n💲 Valuation\n\n- SOTP (Sum-of-the-parts) methodMemory: Average of Hynix and Micron EV/EBITDA.\n- Foundry: 50% discount to TSMC.\n- Mobile: 50% discount to Apple.\n- Display: Premium granted.\n- CE: 10% premium compared to peers.\n\n⚠️ Risk Factors\n\nSK Hynix faces key risks such as U.S.-China trade and geopolitical tensions, weak NAND demand, and the expansion of Chinese competitors.\n\nSamsung Electronics is exposed to risks like delayed HBM qualification and a slow recovery in memory demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013050606401113884, "ret_signed_m1d": -0.013050606401113884, "alpha_spy_m1d": -0.012490555603168252, "alpha_c_m1d": -0.004588610148347327, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.008461996252766557, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017944510848193818, "ret_signed_p1d": -0.017944510848193818, "alpha_spy_p1d": -0.025768290722681164, "alpha_c_p1d": -0.02484243591016122, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.006897925061967403, "ret_p1w": -0.0020729123543138384, "ret_signed_p1w": -0.0020729123543138384, "alpha_spy_p1w": -0.024029052996005995, "alpha_c_p1w": -0.01919739057227543, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.017124478217961592, "ret_p1m": 0.08327613791643707, "ret_signed_p1m": 0.08327613791643707, "alpha_spy_p1m": 0.0383097566789905, "alpha_c_p1m": 0.034990478156236016, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.04828565976020105, "ret_p3m": 0.3705084088935371, "ret_signed_p3m": 0.3705084088935371, "alpha_spy_p3m": 0.2690972410710306, "alpha_c_p3m": 0.23546423137833838, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.1350441775151987, "ret_p6m": 0.773947759962897, "ret_signed_p6m": 0.773947759962897, "alpha_spy_p6m": 0.6566309061547086, "alpha_c_p6m": 0.6372171185081534, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.13673064145474356, "price_path": [-0.0179, -0.0571, -0.0408, -0.0408, -0.0604, -0.0848, -0.1321, -0.1272, -0.1354, -0.0799, -0.0995, -0.0832, -0.0767, -0.1077, -0.1011, -0.0979, -0.0962, -0.1028, -0.0914, -0.0914, -0.0914, -0.0897, -0.0897, -0.0946, -0.0946, -0.1142, -0.1142, -0.1142, -0.1093, -0.1093, -0.106, -0.0604, -0.0718, -0.0636, -0.0653, -0.0734, -0.0897, -0.0881, -0.0914, -0.1077, -0.1158, -0.1077, -0.1207, -0.0881, -0.0848, -0.0832, -0.0734, -0.0734, -0.0571, -0.0359, -0.0359, -0.0245, -0.0343, -0.0228, -0.0294, -0.0489, -0.0669, -0.0522, -0.0245, -0.0343, -0.0294, -0.0538, -0.0131, 0.0, -0.0179, -0.0021, -0.0185, -0.0119, -0.0021, 0.0472, 0.039, 0.0127, 0.0078, -0.0086, 0.0012, 0.0275, 0.0258, 0.0455, 0.0619, 0.0948, 0.1013, 0.1128, 0.0833, 0.0898, 0.0833, 0.0816, 0.1555, 0.1588, 0.1916, 0.1719, 0.1309, 0.144, 0.1473, 0.1292, 0.1571, 0.1785, 0.1653, 0.167, 0.1801, 0.1752, 0.1752, 0.1489, 0.1489, 0.1571, 0.1588, 0.1719, 0.1735, 0.1539, 0.1588, 0.1424, 0.144, 0.1095, 0.1342, 0.1456, 0.1506, 0.1407, 0.1506, 0.1735, 0.1916, 0.2047, 0.2376, 0.2556, 0.3032, 0.2835, 0.318, 0.318, 0.3705, 0.3902, 0.4017, 0.4132, 0.3672, 0.3882, 0.3832, 0.4178, 0.4797, 0.4797, 0.4797, 0.4797, 0.4797, 0.4797, 0.5563, 0.5382, 0.5102, 0.5662, 0.6107, 0.614, 0.6173, 0.6074, 0.6256, 0.5909, 0.6289, 0.6816, 0.6404, 0.6569, 0.7162, 0.7723, 0.8317, 0.7294, 0.6585, 0.6355, 0.614, 0.6585, 0.7064, 0.6998, 0.6948, 0.6025, 0.6585, 0.6124, 0.5909, 0.6585, 0.5629, 0.5942, 0.6371, 0.6948, 0.7064, 0.6569, 0.6618, 0.7047, 0.7228, 0.7327, 0.7871, 0.8053, 0.7871, 0.7805, 0.769, 0.7954, 0.7278, 0.6948, 0.7789, 0.7739]}
{"unit_id": "orig:1929669558341980456", "ticker_idx": 0, "ts": "2025-06-02T22:40:48+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley considers this a 'Show-me Story'— significant uncertainty remains until actual tape-out and successful testing are achieved... optical module business may face some headwinds due to China's export restrictions... a risk that has not been fully priced in by the market", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley's cautious MRVL view: ASIC a 'show-me story' with unresolved uncertainty; optical segment China export risk not yet priced in.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Regarding Marvell (MRVL) – Morgan Stanley’s View:\n• The market has already fully priced in the weak AWS Trainium shipments and lower ASPs.\n• Expectations for AWS have been reset, and the stock is now trading based on a new baseline.\n• ASIC Segment:\nAccording to Morgan Stanley’s analysis:\nTrainium 2 has delivered disappointing results compared to initial expectations.\nThe current tracked scale is around $2 billion,\nwhile earlier projections called for 1.5 to 2 million units at $2,500 per unit.\nHowever, this was already reflected in market expectations earlier this year, so it’s not a surprise.\nTrainium 3 is expected to have more clarity on orders by the end of the year,\nwith AlChip confirmed to secure part of the orders.\nAnother ASIC customer is presumed to be Microsoft,\nand that project is expected to ramp up in 2026.\nStill, Morgan Stanley considers this a “Show-me Story”—\nsignificant uncertainty remains until actual tape-out and successful testing are achieved.\n• Optical Module Segment:\nMRVL’s optical module business may face some headwinds due to China’s export restrictions,\nparticularly around products like NVIDIA H20 and AMD Mi308.\nThis could cap upside potential,\nand Morgan Stanley notes this is a risk that has not been fully priced in by the market.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020823250187019138, "ret_signed_m1d": 0.020823250187019138, "alpha_spy_m1d": 0.015221900896988938, "alpha_c_m1d": 0.006232155591326949, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014478475093503418, "ret_signed_p1d": -0.014478475093503418, "alpha_spy_p1d": -0.008775885305315922, "alpha_c_p1d": 0.008086222512710739, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": 0.12477634549566652, "ret_signed_p1w": -0.12477634549566652, "alpha_spy_p1w": -0.11301692833263144, "alpha_c_p1w": -0.0686317848733149, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.2402797017599776, "ret_signed_p1m": -0.2402797017599776, "alpha_spy_p1m": -0.19512112758032507, "alpha_c_p1m": -0.10653544577828122, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.2574135315626489, "ret_signed_p3m": -0.2574135315626489, "alpha_spy_p3m": -0.15934148603275533, "alpha_c_p3m": -0.028971352923970572, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 0.35925773772714376, "ret_signed_p6m": -0.35925773772714376, "alpha_spy_p6m": -0.21385367617083784, "alpha_c_p6m": 0.038235091824257994, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [0.4647, 0.1745, 0.1511, 0.0671, 0.0913, 0.135, 0.1173, 0.117, 0.144, 0.1095, 0.1344, 0.1454, 0.1438, 0.1823, 0.1646, 0.0848, 0.054, 0.0081, 0.0005, 0.0184, 0.0275, -0.0959, -0.1968, -0.1713, -0.187, -0.0094, -0.1409, -0.1314, -0.1498, -0.1326, -0.1552, -0.1589, -0.1589, -0.1967, -0.1765, -0.1253, -0.0673, -0.0415, -0.0451, -0.0454, -0.0504, -0.0085, 0.014, 0.0083, -0.0041, -0.0839, -0.0625, -0.0296, 0.0493, 0.0652, 0.0721, 0.0607, 0.0373, 0.0177, -0.0008, -0.0221, 0.0062, -0.0127, -0.0127, 0.0382, 0.0508, 0.0368, -0.0208, 0.0, 0.0145, 0.0786, 0.06, 0.1119, 0.1248, 0.1199, 0.1101, 0.1329, 0.0931, 0.1456, 0.1386, 0.2193, 0.2193, 0.1959, 0.1515, 0.2235, 0.2352, 0.301, 0.2552, 0.2592, 0.2403, 0.2079, 0.223, 0.223, 0.164, 0.1705, 0.1755, 0.1934, 0.1838, 0.1806, 0.1789, 0.1535, 0.1724, 0.2154, 0.1895, 0.1721, 0.1929, 0.2055, 0.2082, 0.2359, 0.2429, 0.3308, 0.3085, 0.2122, 0.246, 0.2476, 0.2263, 0.2349, 0.2592, 0.2582, 0.2669, 0.2914, 0.2869, 0.2405, 0.2494, 0.1734, 0.1596, 0.1594, 0.1885, 0.1877, 0.2091, 0.2177, 0.2574, 0.0236, 0.0236, 0.0518, 0.0145, 0.0436, 0.0311, 0.0746, 0.0882, 0.0925, 0.0842, 0.0966, 0.0979, 0.1211, 0.1557, 0.2086, 0.2091, 0.2297, 0.2149, 0.304, 0.3645, 0.3541, 0.3414, 0.3688, 0.3658, 0.4035, 0.4038, 0.4477, 0.416, 0.506, 0.4764, 0.3948, 0.4564, 0.4047, 0.4482, 0.4375, 0.4329, 0.3985, 0.3728, 0.3205, 0.3485, 0.3707, 0.4453, 0.4412, 0.4687, 0.443, 0.5272, 0.4723, 0.427, 0.5135, 0.5206, 0.4813, 0.5189, 0.4554, 0.4554, 0.4259, 0.4085, 0.3596, 0.2819, 0.3249, 0.2493, 0.2618, 0.3651, 0.3593]}
{"unit_id": "orig:1939550029838246243", "ticker_idx": 0, "ts": "2025-06-30T05:02:16+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Microsoft's self-developed AI chip performance and schedule, aimed at reducing Nvidia dependence, are significantly delayed more than expected… unlikely to reduce its dependence on Nvidia in the short term", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVDA long as MSFT chip delays entrench dependence; MRVL short on ASIC delays/missed targets; AMD long as likely beneficiary of big-tech/Marvell schedule slippage.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[Microsoft's AI Chip Development Status and Problems] - BK Tech insight\n\n📌 Executive Summary\n\n- Microsoft's self-developed AI chip performance and schedule, aimed at reducing Nvidia dependence, are significantly delayed more than expected.\n- The latest AI chip (Braga) launch is delayed by at least 6 months, shifting from 2025 to 2026, due to design changes and a shortage of personnel.\n- Both performance and power efficiency are expected to significantly lag behind Nvidia's latest chip (Blackwell).\n\n🔖 Details\n\n1️⃣ Background of Microsoft's AI Chip Development\n\n- Microsoft is highly dependent on Nvidia chips and is considered one of Nvidia's largest customers.\n- Self-development of AI chips began in 2019, and the first AI chip (Maia 100) was unveiled in 2023 but has seen little actual service utilization.\n\n2️⃣ Problems with the Next-Generation Chip (Braga)\n\n- The next-generation chip (Braga, Maia 200), originally planned for mass production in 2025, has been delayed to 2026.\n- Design changes requested by OpenAI in the middle of development caused chip stability issues, and additional work to resolve these issues significantly delayed the schedule.\n- Around 20% of the related development personnel left due to schedule pressure and high work intensity.\n\n3️⃣ Performance and Efficiency Disadvantages\n- Braga is projected to have significantly lower competitiveness in terms of performance and power efficiency compared to Nvidia's latest AI chip, Blackwell (released in 2024).\n- Microsoft plans subsequent chips like Braga-R and Clea after Braga, but Braga's delay makes the schedule for subsequent chips uncertain.\n\n4️⃣ Comparison with Other Companies\n- Amazon and Google are also actively developing their own AI chips; Amazon's Trainium 3 chip development is proceeding smoothly, targeting a release by the end of this year as planned.\n- Google has been developing its own AI chips (TPU) for over 10 years, and its next-generation chip (Ironwood) is also on schedule for release. However, Google also faced difficulties with key personnel leaving for Nvidia during its collaboration with MediaTek.\n\n5️⃣ Nvidia's Response\n- Nvidia CEO Jensen Huang recently stated at a developer event that most competitor AI chip development projects would eventually be abandoned.\n- Nvidia has set very high performance targets for its latest AI system (GB200) to make it difficult for customers to switch to their own chips.\n\n📌 Microsoft AI Chip Roadmap\n- Maia 100 (Unveiled 2023): AI chip focused on image processing, not yet applied to actual services, only internal testing.\n- Braga (Maia 200): Targets mass production in 2026 (delayed by at least 6 months), faces performance issues.\n- Braga-R: Successor to Braga, scheduled for 2026.\n- Clea (Maia 300): Planned with an entirely new design, considered the first model with potential to truly compete with Nvidia chips (scheduled for 2027).\n\n🧩 Conclusion and Outlook\n- Microsoft's self-developed AI chip project faces significant challenges in achieving performance competitiveness against Nvidia, and due to development delays and design issues, it is unlikely to reduce its dependence on Nvidia in the short term.\n- The upcoming Clea chip (Maia 300) is expected to be the first model truly capable of competing with Nvidia, but overcoming schedule delays and technical issues will be a key challenge.\n\n// Short comment\n\n- Microsoft's next-generation ASIC, speculated to be primarily involving Marvell and Alchip, is expected to be delayed due to development issues.\n- This led to Marvell's stock price weakening on Friday.\n- Overall, ASIC projects handled by Marvell seem to be experiencing delays and failing to meet performance targets.\n- Due to schedule delays for big tech companies collaborating with Marvell, the possibility of AMD receiving orders for inference purposes is increasingly high for next year.\n- It is necessary to follow up on AMD's potential gains from the schedule delays of big tech companies cooperating with Marvell.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0015191774050881124, "ret_signed_m1d": -0.0015191774050881124, "alpha_spy_m1d": 0.00323937665996632, "alpha_c_m1d": 0.0001302201301240924, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02968544324345679, "ret_signed_p1d": -0.02968544324345679, "alpha_spy_p1d": -0.029361763935231933, "alpha_c_p1d": -0.018784666481059697, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": 0.001582351692805295, "ret_signed_p1w": 0.001582351692805295, "alpha_spy_p1w": -0.0029979745229373744, "alpha_c_p1w": -0.001573194090819241, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": 0.11089309220074783, "ret_signed_p1m": 0.11089309220074783, "alpha_spy_p1m": 0.0827148607649617, "alpha_c_p1m": 0.06062060571255845, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.1247548253226114, "ret_signed_p3m": 0.1247548253226114, "alpha_spy_p3m": 0.056737738163414875, "alpha_c_p3m": -0.026493040619766317, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.19774160615120495, "ret_signed_p6m": 0.19774160615120495, "alpha_spy_p6m": 0.07788113594250312, "alpha_c_p6m": -0.10851922193066077, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.3011, -0.3557, -0.4031, -0.382, -0.3905, -0.2764, -0.3192, -0.2979, -0.2993, -0.2899, -0.3387, -0.3577, -0.3577, -0.3866, -0.3741, -0.3499, -0.3264, -0.2974, -0.3118, -0.31, -0.3106, -0.2936, -0.2753, -0.2796, -0.2814, -0.2591, -0.2572, -0.2617, -0.2215, -0.1777, -0.1434, -0.1467, -0.143, -0.142, -0.1495, -0.1658, -0.1593, -0.1691, -0.1691, -0.1424, -0.1468, -0.1191, -0.1448, -0.1305, -0.1062, -0.1018, -0.114, -0.103, -0.0973, -0.0889, -0.096, -0.0822, -0.1014, -0.0842, -0.0878, -0.0792, -0.0792, -0.0895, -0.0875, -0.0639, -0.0233, -0.0188, -0.0015, 0.0, -0.0297, -0.0047, 0.0085, 0.0085, 0.0016, 0.0127, 0.031, 0.0387, 0.0439, 0.0385, 0.0804, 0.0847, 0.095, 0.0913, 0.0848, 0.0572, 0.081, 0.0997, 0.0982, 0.1187, 0.1109, 0.1347, 0.1258, 0.0996, 0.1393, 0.1283, 0.1356, 0.1442, 0.1564, 0.1524, 0.1593, 0.1494, 0.1521, 0.1422, 0.152, 0.1117, 0.1102, 0.1075, 0.1266, 0.1381, 0.1505, 0.1494, 0.1404, 0.1025, 0.1025, 0.081, 0.0799, 0.0865, 0.0572, 0.0653, 0.0808, 0.1224, 0.1215, 0.1256, 0.1251, 0.107, 0.0779, 0.1156, 0.1183, 0.1622, 0.1294, 0.1202, 0.1248, 0.1279, 0.1511, 0.181, 0.1852, 0.1956, 0.1876, 0.1744, 0.1713, 0.197, 0.2189, 0.1594, 0.192, 0.1396, 0.1383, 0.1508, 0.1598, 0.1561, 0.1467, 0.1411, 0.153, 0.179, 0.2121, 0.2725, 0.3105, 0.2843, 0.2817, 0.3095, 0.2577, 0.2357, 0.1905, 0.191, 0.26, 0.2227, 0.2267, 0.1828, 0.2038, 0.1812, 0.148, 0.1806, 0.1434, 0.1323, 0.1555, 0.1256, 0.141, 0.141, 0.1204, 0.1389, 0.1486, 0.1368, 0.1608, 0.1547, 0.1746, 0.1709, 0.1634, 0.1453, 0.1079, 0.116, 0.125, 0.0821, 0.1023, 0.1457, 0.1628, 0.1977]}
{"unit_id": "orig:1939550029838246243", "ticker_idx": 1, "ts": "2025-06-30T05:02:16+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Overall, ASIC projects handled by Marvell seem to be experiencing delays and failing to meet performance targets", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVDA long as MSFT chip delays entrench dependence; MRVL short on ASIC delays/missed targets; AMD long as likely beneficiary of big-tech/Marvell schedule slippage.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[Microsoft's AI Chip Development Status and Problems] - BK Tech insight\n\n📌 Executive Summary\n\n- Microsoft's self-developed AI chip performance and schedule, aimed at reducing Nvidia dependence, are significantly delayed more than expected.\n- The latest AI chip (Braga) launch is delayed by at least 6 months, shifting from 2025 to 2026, due to design changes and a shortage of personnel.\n- Both performance and power efficiency are expected to significantly lag behind Nvidia's latest chip (Blackwell).\n\n🔖 Details\n\n1️⃣ Background of Microsoft's AI Chip Development\n\n- Microsoft is highly dependent on Nvidia chips and is considered one of Nvidia's largest customers.\n- Self-development of AI chips began in 2019, and the first AI chip (Maia 100) was unveiled in 2023 but has seen little actual service utilization.\n\n2️⃣ Problems with the Next-Generation Chip (Braga)\n\n- The next-generation chip (Braga, Maia 200), originally planned for mass production in 2025, has been delayed to 2026.\n- Design changes requested by OpenAI in the middle of development caused chip stability issues, and additional work to resolve these issues significantly delayed the schedule.\n- Around 20% of the related development personnel left due to schedule pressure and high work intensity.\n\n3️⃣ Performance and Efficiency Disadvantages\n- Braga is projected to have significantly lower competitiveness in terms of performance and power efficiency compared to Nvidia's latest AI chip, Blackwell (released in 2024).\n- Microsoft plans subsequent chips like Braga-R and Clea after Braga, but Braga's delay makes the schedule for subsequent chips uncertain.\n\n4️⃣ Comparison with Other Companies\n- Amazon and Google are also actively developing their own AI chips; Amazon's Trainium 3 chip development is proceeding smoothly, targeting a release by the end of this year as planned.\n- Google has been developing its own AI chips (TPU) for over 10 years, and its next-generation chip (Ironwood) is also on schedule for release. However, Google also faced difficulties with key personnel leaving for Nvidia during its collaboration with MediaTek.\n\n5️⃣ Nvidia's Response\n- Nvidia CEO Jensen Huang recently stated at a developer event that most competitor AI chip development projects would eventually be abandoned.\n- Nvidia has set very high performance targets for its latest AI system (GB200) to make it difficult for customers to switch to their own chips.\n\n📌 Microsoft AI Chip Roadmap\n- Maia 100 (Unveiled 2023): AI chip focused on image processing, not yet applied to actual services, only internal testing.\n- Braga (Maia 200): Targets mass production in 2026 (delayed by at least 6 months), faces performance issues.\n- Braga-R: Successor to Braga, scheduled for 2026.\n- Clea (Maia 300): Planned with an entirely new design, considered the first model with potential to truly compete with Nvidia chips (scheduled for 2027).\n\n🧩 Conclusion and Outlook\n- Microsoft's self-developed AI chip project faces significant challenges in achieving performance competitiveness against Nvidia, and due to development delays and design issues, it is unlikely to reduce its dependence on Nvidia in the short term.\n- The upcoming Clea chip (Maia 300) is expected to be the first model truly capable of competing with Nvidia, but overcoming schedule delays and technical issues will be a key challenge.\n\n// Short comment\n\n- Microsoft's next-generation ASIC, speculated to be primarily involving Marvell and Alchip, is expected to be delayed due to development issues.\n- This led to Marvell's stock price weakening on Friday.\n- Overall, ASIC projects handled by Marvell seem to be experiencing delays and failing to meet performance targets.\n- Due to schedule delays for big tech companies collaborating with Marvell, the possibility of AMD receiving orders for inference purposes is increasingly high for next year.\n- It is necessary to follow up on AMD's potential gains from the schedule delays of big tech companies cooperating with Marvell.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0031006306918653914, "ret_signed_m1d": 0.0031006306918653914, "alpha_spy_m1d": -0.0016579233731890408, "alpha_c_m1d": 0.0014512331566531866, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014987024120991288, "ret_signed_p1d": 0.014987024120991288, "alpha_spy_p1d": 0.014663344812766432, "alpha_c_p1d": 0.004086247358594197, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.07558122123396815, "ret_signed_p1w": 0.07558122123396815, "alpha_spy_p1w": 0.08016154744971082, "alpha_c_p1w": 0.07873676701759269, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.012887715973491365, "ret_signed_p1m": 0.012887715973491365, "alpha_spy_p1m": 0.04106594740927749, "alpha_c_p1m": 0.06316020246168075, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.08370289492023764, "ret_signed_p3m": -0.08370289492023764, "alpha_spy_p3m": -0.015685807761041115, "alpha_c_p3m": 0.06754497102214008, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.13449458810417747, "ret_signed_p6m": -0.13449458810417747, "alpha_spy_p6m": -0.01463411789547564, "alpha_c_p6m": 0.17176623997768825, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.184, -0.282, -0.3621, -0.3418, -0.3544, -0.2133, -0.3177, -0.3102, -0.3248, -0.3111, -0.3291, -0.332, -0.332, -0.362, -0.346, -0.3053, -0.2593, -0.2388, -0.2416, -0.2419, -0.2459, -0.2125, -0.1947, -0.1992, -0.209, -0.2725, -0.2554, -0.2293, -0.1667, -0.154, -0.1486, -0.1576, -0.1762, -0.1917, -0.2065, -0.2234, -0.2009, -0.2159, -0.2159, -0.1755, -0.1655, -0.1766, -0.2224, -0.2058, -0.1943, -0.1434, -0.1581, -0.1169, -0.1067, -0.1106, -0.1183, -0.1003, -0.1319, -0.0902, -0.0957, -0.0317, -0.0317, -0.0503, -0.0855, -0.0283, -0.019, 0.0332, -0.0031, 0.0, -0.015, -0.0407, -0.0287, -0.0287, -0.0756, -0.0704, -0.0664, -0.0522, -0.0598, -0.0624, -0.0637, -0.0839, -0.0689, -0.0347, -0.0553, -0.0691, -0.0526, -0.0426, -0.0404, -0.0184, -0.0129, 0.0569, 0.0392, -0.0373, -0.0104, -0.0091, -0.0261, -0.0192, 0.0, -0.0007, 0.0061, 0.0256, 0.022, -0.0148, -0.0077, -0.0681, -0.0791, -0.0792, -0.0561, -0.0567, -0.0398, -0.0329, -0.0014, -0.1871, -0.1871, -0.1647, -0.1943, -0.1712, -0.1811, -0.1466, -0.1357, -0.1324, -0.139, -0.1291, -0.1281, -0.1096, -0.0822, -0.0402, -0.0398, -0.0234, -0.0351, 0.0356, 0.0837, 0.0754, 0.0653, 0.0871, 0.0847, 0.1146, 0.1149, 0.1498, 0.1246, 0.1961, 0.1725, 0.1077, 0.1566, 0.1156, 0.1502, 0.1416, 0.138, 0.1107, 0.0902, 0.0487, 0.071, 0.0886, 0.1478, 0.1446, 0.1665, 0.146, 0.2129, 0.1693, 0.1333, 0.202, 0.2076, 0.1764, 0.2063, 0.1558, 0.1558, 0.1324, 0.1186, 0.0798, 0.018, 0.0522, -0.0078, 0.0021, 0.0842, 0.0795, 0.135, 0.135, 0.1567, 0.1787, 0.2019, 0.2965, 0.2705, 0.2798, 0.1904, 0.1503, 0.1965, 0.1571, 0.0924, 0.0902, 0.0878, 0.0571, 0.093, 0.088, 0.0972, 0.1345]}
{"unit_id": "orig:1939550029838246243", "ticker_idx": 2, "ts": "2025-06-30T05:02:16+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the possibility of AMD receiving orders for inference purposes is increasingly high for next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVDA long as MSFT chip delays entrench dependence; MRVL short on ASIC delays/missed targets; AMD long as likely beneficiary of big-tech/Marvell schedule slippage.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[Microsoft's AI Chip Development Status and Problems] - BK Tech insight\n\n📌 Executive Summary\n\n- Microsoft's self-developed AI chip performance and schedule, aimed at reducing Nvidia dependence, are significantly delayed more than expected.\n- The latest AI chip (Braga) launch is delayed by at least 6 months, shifting from 2025 to 2026, due to design changes and a shortage of personnel.\n- Both performance and power efficiency are expected to significantly lag behind Nvidia's latest chip (Blackwell).\n\n🔖 Details\n\n1️⃣ Background of Microsoft's AI Chip Development\n\n- Microsoft is highly dependent on Nvidia chips and is considered one of Nvidia's largest customers.\n- Self-development of AI chips began in 2019, and the first AI chip (Maia 100) was unveiled in 2023 but has seen little actual service utilization.\n\n2️⃣ Problems with the Next-Generation Chip (Braga)\n\n- The next-generation chip (Braga, Maia 200), originally planned for mass production in 2025, has been delayed to 2026.\n- Design changes requested by OpenAI in the middle of development caused chip stability issues, and additional work to resolve these issues significantly delayed the schedule.\n- Around 20% of the related development personnel left due to schedule pressure and high work intensity.\n\n3️⃣ Performance and Efficiency Disadvantages\n- Braga is projected to have significantly lower competitiveness in terms of performance and power efficiency compared to Nvidia's latest AI chip, Blackwell (released in 2024).\n- Microsoft plans subsequent chips like Braga-R and Clea after Braga, but Braga's delay makes the schedule for subsequent chips uncertain.\n\n4️⃣ Comparison with Other Companies\n- Amazon and Google are also actively developing their own AI chips; Amazon's Trainium 3 chip development is proceeding smoothly, targeting a release by the end of this year as planned.\n- Google has been developing its own AI chips (TPU) for over 10 years, and its next-generation chip (Ironwood) is also on schedule for release. However, Google also faced difficulties with key personnel leaving for Nvidia during its collaboration with MediaTek.\n\n5️⃣ Nvidia's Response\n- Nvidia CEO Jensen Huang recently stated at a developer event that most competitor AI chip development projects would eventually be abandoned.\n- Nvidia has set very high performance targets for its latest AI system (GB200) to make it difficult for customers to switch to their own chips.\n\n📌 Microsoft AI Chip Roadmap\n- Maia 100 (Unveiled 2023): AI chip focused on image processing, not yet applied to actual services, only internal testing.\n- Braga (Maia 200): Targets mass production in 2026 (delayed by at least 6 months), faces performance issues.\n- Braga-R: Successor to Braga, scheduled for 2026.\n- Clea (Maia 300): Planned with an entirely new design, considered the first model with potential to truly compete with Nvidia chips (scheduled for 2027).\n\n🧩 Conclusion and Outlook\n- Microsoft's self-developed AI chip project faces significant challenges in achieving performance competitiveness against Nvidia, and due to development delays and design issues, it is unlikely to reduce its dependence on Nvidia in the short term.\n- The upcoming Clea chip (Maia 300) is expected to be the first model truly capable of competing with Nvidia, but overcoming schedule delays and technical issues will be a key challenge.\n\n// Short comment\n\n- Microsoft's next-generation ASIC, speculated to be primarily involving Marvell and Alchip, is expected to be delayed due to development issues.\n- This led to Marvell's stock price weakening on Friday.\n- Overall, ASIC projects handled by Marvell seem to be experiencing delays and failing to meet performance targets.\n- Due to schedule delays for big tech companies collaborating with Marvell, the possibility of AMD receiving orders for inference purposes is increasingly high for next year.\n- It is necessary to follow up on AMD's potential gains from the schedule delays of big tech companies cooperating with Marvell.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013460209614263308, "ret_signed_m1d": 0.013460209614263308, "alpha_spy_m1d": 0.01821876367931774, "alpha_c_m1d": 0.015109607149475512, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.040803337104838744, "ret_signed_p1d": -0.040803337104838744, "alpha_spy_p1d": -0.04047965779661389, "alpha_c_p1d": -0.029902560342441653, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.05003517371470778, "ret_signed_p1w": -0.05003517371470778, "alpha_spy_p1w": -0.05461549993045045, "alpha_c_p1w": -0.05319071949833232, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": 0.25045814005353795, "ret_signed_p1m": 0.25045814005353795, "alpha_spy_p1m": 0.22227990861775182, "alpha_c_p1m": 0.20018565356534856, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.13650465968382575, "ret_signed_p3m": 0.13650465968382575, "alpha_spy_p3m": 0.06848757252462923, "alpha_c_p3m": -0.014743206258551966, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.5144468156443565, "ret_signed_p6m": 0.5144468156443565, "alpha_spy_p6m": 0.39458634543565463, "alpha_c_p6m": 0.20818598756249074, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2744, -0.339, -0.3956, -0.4106, -0.4488, -0.3175, -0.3749, -0.3418, -0.334, -0.3285, -0.3778, -0.3834, -0.3834, -0.397, -0.3921, -0.363, -0.3342, -0.3189, -0.3207, -0.323, -0.314, -0.3189, -0.3037, -0.2911, -0.305, -0.2927, -0.2833, -0.2753, -0.2381, -0.2075, -0.1704, -0.1896, -0.1743, -0.1914, -0.2001, -0.2103, -0.2198, -0.2226, -0.2226, -0.1927, -0.2047, -0.2035, -0.2197, -0.1922, -0.1733, -0.1643, -0.1847, -0.1812, -0.1421, -0.1315, -0.1463, -0.1649, -0.1814, -0.1093, -0.1043, -0.1065, -0.1065, -0.0963, -0.0868, -0.0245, 0.0106, 0.0125, 0.0135, 0.0, -0.0408, -0.0238, -0.0281, -0.0281, -0.05, -0.0288, -0.0246, 0.0159, 0.0319, 0.0306, 0.0966, 0.1281, 0.1304, 0.1063, 0.1064, 0.0903, 0.118, 0.1425, 0.1732, 0.2238, 0.2505, 0.265, 0.2425, 0.21, 0.2458, 0.2284, 0.1495, 0.2149, 0.2175, 0.2141, 0.2329, 0.2996, 0.2752, 0.251, 0.2413, 0.1737, 0.1642, 0.1537, 0.1822, 0.1512, 0.1742, 0.1778, 0.188, 0.1461, 0.1461, 0.1439, 0.1426, 0.1402, 0.0651, 0.067, 0.0981, 0.1243, 0.097, 0.1175, 0.1357, 0.1308, 0.1216, 0.1129, 0.1092, 0.1261, 0.1339, 0.1338, 0.1365, 0.1237, 0.1371, 0.1402, 0.1558, 0.1961, 0.1605, 0.4356, 0.4906, 0.66, 0.6412, 0.5144, 0.5252, 0.5369, 0.6815, 0.653, 0.6426, 0.6953, 0.6774, 0.6225, 0.656, 0.7824, 0.83, 0.8183, 0.8628, 0.7959, 0.8049, 0.8298, 0.7622, 0.8064, 0.6751, 0.6458, 0.7194, 0.6739, 0.8245, 0.7474, 0.7393, 0.695, 0.6229, 0.5754, 0.4519, 0.4361, 0.5155, 0.4526, 0.5098, 0.5098, 0.533, 0.5487, 0.5168, 0.5335, 0.5221, 0.5361, 0.5582, 0.5618, 0.5604, 0.5605, 0.4854, 0.4629, 0.4741, 0.3961, 0.4169, 0.5041, 0.5148, 0.5144]}
{"unit_id": "orig:1927520049117442210", "ticker_idx": 0, "ts": "2025-05-28T00:19:25+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HSBC has lowered its FY26/FY27 EPS forecasts by 18%/17%, maintaining a \"Hold\" rating and $120 target price", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC Hold on NVDA with $120 PT; EPS cut 18%/17% on H20 ban, margin pressure, and 2H supply-demand imbalance.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Sales Guidance Remains Steady, But Outlook Grows More Uncertain\nHSBC (F. Lee, May 15, 2025)\n\nNvidia has guided for FY1Q26 and FY2Q26 revenues of $43.5B and $45.0B, respectively—slightly below market consensus but in line with management’s projections. However, the China H20 chip export ban and a $550M inventory write-down have already weighed on margins, and H20-related revenue losses will persist starting in Q2.\n\nMoreover, a mismatch between upstream AI GPU shipments and downstream ODM rack installations is expected to intensify supply-demand imbalances in the second half, potentially pressuring EPS performance for FY26 and FY27. Despite improving sentiment around AI investment, HSBC has lowered its FY26/FY27 EPS forecasts by 18%/17%, maintaining a “Hold” rating and $120 target price.\n\n⸻\n\n• Sales Pace in Line With Expectations\n• FY1Q26 revenue of $43.5B is broadly in line with management guidance ($43.0B) and market consensus ($43.1B).\n• FY2Q26 revenue guidance of $45.0B is below the $46.7B consensus, but better than some pessimistic estimates around $44.0B, thus offering relative stability.\n\n⸻\n\n• Rising Uncertainty in 2H\n• Although U.S. cloud service providers are expected to increase CAPEX by 43% in 2025,\n• There has been no significant AI GPU supply chain ramp-up since February.\n• The disconnect between AI GPU supply and ODM rack installation demand could result in lower-than-expected shipments in 2H FY26.\n\n⸻\n\n• Margin Pressure and Valuation Adjustment\n• Due to the H20 export ban and executive-level inventory write-down,\n• HSBC has revised FY26/FY27 EPS down to $3.89 and $4.78, respectively—11% and 15% below consensus.\n• Maintaining a $120 price target based on a 31x FY26 PE multiple,\n• HSBC reiterates its “Hold” rating on the stock.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": -0.005118180170402997, "alpha_spy_m1d": 0.0007009000979054658, "alpha_c_m1d": 0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": -0.03249016424752793, "alpha_spy_p1d": -0.02854288690673168, "alpha_c_p1d": -0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": -0.05274087220320278, "alpha_spy_p1w": -0.03878900778980032, "alpha_c_p1w": -0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": -0.14999444685150065, "alpha_spy_p1m": -0.10584353163755589, "alpha_c_p1m": -0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": -0.3338957051915541, "alpha_spy_p3m": -0.2375262150750832, "alpha_c_p3m": -0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": -0.3401284775531095, "alpha_spy_p6m": -0.22350433610615128, "alpha_c_p6m": 0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "orig:1921433573565083836", "ticker_idx": 0, "ts": "2025-05-11T05:13:57+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Our top AI picks are NVIDIA (NVDA) and Broadcom (AVGO), leaders in commercial/custom AI silicon, respectively", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA top AI picks: NVDA and AVGO as leaders; AMD and MRVL as Tier-2 beneficiaries with PEG below 1x by 2027 on strong EPS CAGRs.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Q1 Cloud Capex Review: AI Investment Remains Robust\n\n1. AI Capex Outlook for 2025/2026: +44% / +4% YoY Growth\n\nDespite a range of concerns since the beginning of the year—such as Capex intensity (sustainability), DeepSeek (AI efficiency), ongoing China regulations, and the recent rollback of AI diffusion initiatives—demand for data center and AI-related Capex remains strong. In Q1, global hyperscaler Capex totaled $92 billion, slightly below the consensus of $93.8 billion. However, most hyperscalers indicated continued strong AI investment throughout 2025, with expectations for growth to persist into 2026.\n\nOverall, our consensus model shows 2025/2026 Capex reaching $414 billion / $432 billion, representing +44% / +4% YoY growth. This is 300bps/100bps higher than pre-earnings expectations ($402B/$429B), and 5,090bps/4,900bps higher compared to May 2024 projections.\n\nWhile CSPs (cloud service providers) have not provided extensive guidance on 2026 spending yet, the rollback of AI diffusion regulations has allowed for sovereign AI infrastructure development in most countries, supporting a solid investment environment.\n\nMeanwhile, China’s GPU regulations (which affect ~10% of global AI TAM) are now considered to have fully played out. Residual revenues (e.g., B20 deployments) are expected to be incremental. AMD has stated that its $500 billion AI TAM outlook already factors in the impact of China’s regulations, suggesting limited long-term implications for U.S. accelerator vendors.\n\nThe next major catalyst for AI is expected to be Computex (May 19–21 in Taiwan), where NVIDIA (NVDA) CEO is scheduled to deliver a keynote on its roadmap and new products.\n\n⸻\n\n2. 2025 Capex Outlook: Strong Despite Multiple Concerns\n\nCloud Capex is expected to remain strong in 2025. Google and Microsoft have reaffirmed their 2025 (FY26 for Microsoft) guidance. Meta has raised its 2025 Capex guidance from $60–65B to $64–72B. While most companies avoided commenting on 2026 forecasts, Microsoft indicated YoY Capex growth, noting a shift toward short-lived assets (servers, CPUs, GPUs) over long-term assets (buildings, land).\n\nAmazon has not updated its $105 billion 2025 Capex forecast from the previous call, but its custom silicon Trainium2 (MRVL) is being deployed across more instances.\n\n⸻\n\n3. Questions Around Capex Intensity and AI Sustainability\n\nCloud Capex intensity (Capex as a percentage of revenue) is rising rapidly, expected to reach 20–22% in 2025–2027 (vs. 17% in 2024 and 10–12% during 2018–2023). We expect continued AI expansion, supported by the following:\n1. Leading model sizes double every six months pre- and post-training\n2. Derivative models like DeepSeek\n3. Inference scaling during testing (token generation 100x vs standard inference)\n4. Still early stages of sovereign/enterprise AI adoption\n\n⸻\n\n4. Top AI Investment Picks: NVIDIA $NVDA, Broadcom $AVGO, Plus $AMD / $MRVL\n\nOur top AI picks are NVIDIA (NVDA) and Broadcom (AVGO), leaders in commercial/custom AI silicon, respectively. Additionally, AMD (AI GPUs) and MRVL (AI ASICs, optical/DSP) are expected to benefit from AI growth as Tier-2 players.\n\nAMD is projected to maintain a 3–4% share in the accelerator market by 2027 ($12.6B revenue), while MRVL is expected to maintain a 10–15% share in the custom ASIC market ($4B revenue).\n\nOverall, AMD/MRVL EPS is projected to grow at a +27% / +39% CAGR from 2024 to 2027, and with the AI market growing at +45% (based on TSMC), their PEG ratios are expected to fall below 1.0x.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.051626008542716484, "ret_signed_m1d": -0.051626008542716484, "alpha_spy_m1d": -0.01963569737158022, "alpha_c_m1d": 0.007385873559626277, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0563414029159095, "ret_signed_p1d": 0.0563414029159095, "alpha_spy_p1d": 0.04973746259873946, "alpha_c_p1d": 0.022012643153013878, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.10219506593164018, "ret_signed_p1w": 0.10219506593164018, "alpha_spy_p1w": 0.08185176565789498, "alpha_c_p1w": 0.06597083300351914, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.1704065092705338, "ret_signed_p1m": 0.1704065092705338, "alpha_spy_p1m": 0.1359463388222879, "alpha_c_p1m": 0.06653554689432228, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.46977681802609217, "ret_signed_p3m": 0.46977681802609217, "alpha_spy_p3m": 0.38207542320790444, "alpha_c_p3m": 0.24354377888437795, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.6155690716586328, "ret_signed_p6m": 0.6155690716586328, "alpha_spy_p6m": 0.4506883865956586, "alpha_c_p6m": 0.12970069447571997, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0661, 0.0998, 0.1288, 0.1288, 0.1332, 0.1318, 0.139, 0.0928, 0.0591, 0.0294, 0.0672, -0.0233, 0.0155, -0.0728, -0.0571, -0.0464, -0.1011, -0.0839, -0.1303, -0.1159, -0.059, -0.0603, -0.0108, -0.0282, -0.0615, -0.0446, -0.0363, -0.0431, -0.0129, -0.0188, -0.0751, -0.0941, -0.1084, -0.1189, -0.1045, -0.1023, -0.1724, -0.2333, -0.2062, -0.2171, -0.0705, -0.1254, -0.0981, -0.0999, -0.0878, -0.1505, -0.1749, -0.1749, -0.2121, -0.196, -0.165, -0.1347, -0.0975, -0.116, -0.1137, -0.1145, -0.0926, -0.0691, -0.0746, -0.0769, -0.0483, -0.0458, -0.0516, 0.0, 0.0563, 0.1003, 0.0962, 0.1008, 0.1022, 0.0925, 0.0715, 0.0799, 0.0674, 0.0674, 0.1016, 0.096, 0.1316, 0.0986, 0.1169, 0.1481, 0.1538, 0.1381, 0.1522, 0.1596, 0.1704, 0.1613, 0.1789, 0.1543, 0.1764, 0.1718, 0.1828, 0.1828, 0.1696, 0.1722, 0.2025, 0.2546, 0.2604, 0.2826, 0.2846, 0.2464, 0.2785, 0.2955, 0.2955, 0.2866, 0.3009, 0.3243, 0.3342, 0.3409, 0.334, 0.3879, 0.3933, 0.4066, 0.4018, 0.3934, 0.3581, 0.3886, 0.4126, 0.4107, 0.4371, 0.427, 0.4576, 0.4462, 0.4125, 0.4635, 0.4494, 0.4588, 0.4698, 0.4855, 0.4803, 0.4892, 0.4764, 0.4799, 0.4672, 0.4799, 0.4281, 0.4261, 0.4227, 0.4472, 0.462, 0.4779, 0.4765, 0.4649, 0.4162, 0.4162, 0.3886, 0.3873, 0.3957, 0.358, 0.3685, 0.3884, 0.4418, 0.4406, 0.4459, 0.4453, 0.422, 0.3846, 0.433, 0.4365, 0.493, 0.4508, 0.439, 0.4448, 0.4489, 0.4786, 0.5171, 0.5225, 0.5359, 0.5256, 0.5086, 0.5046, 0.5377, 0.5658, 0.4893, 0.5312, 0.4638, 0.4622, 0.4783, 0.4898, 0.4851, 0.473, 0.4659, 0.4812, 0.5145, 0.557, 0.6346, 0.6835, 0.6497, 0.6465, 0.6822, 0.6156]}
{"unit_id": "orig:1921433573565083836", "ticker_idx": 1, "ts": "2025-05-11T05:13:57+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Our top AI picks are NVIDIA (NVDA) and Broadcom (AVGO), leaders in commercial/custom AI silicon, respectively", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA top AI picks: NVDA and AVGO as leaders; AMD and MRVL as Tier-2 beneficiaries with PEG below 1x by 2027 on strong EPS CAGRs.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Q1 Cloud Capex Review: AI Investment Remains Robust\n\n1. AI Capex Outlook for 2025/2026: +44% / +4% YoY Growth\n\nDespite a range of concerns since the beginning of the year—such as Capex intensity (sustainability), DeepSeek (AI efficiency), ongoing China regulations, and the recent rollback of AI diffusion initiatives—demand for data center and AI-related Capex remains strong. In Q1, global hyperscaler Capex totaled $92 billion, slightly below the consensus of $93.8 billion. However, most hyperscalers indicated continued strong AI investment throughout 2025, with expectations for growth to persist into 2026.\n\nOverall, our consensus model shows 2025/2026 Capex reaching $414 billion / $432 billion, representing +44% / +4% YoY growth. This is 300bps/100bps higher than pre-earnings expectations ($402B/$429B), and 5,090bps/4,900bps higher compared to May 2024 projections.\n\nWhile CSPs (cloud service providers) have not provided extensive guidance on 2026 spending yet, the rollback of AI diffusion regulations has allowed for sovereign AI infrastructure development in most countries, supporting a solid investment environment.\n\nMeanwhile, China’s GPU regulations (which affect ~10% of global AI TAM) are now considered to have fully played out. Residual revenues (e.g., B20 deployments) are expected to be incremental. AMD has stated that its $500 billion AI TAM outlook already factors in the impact of China’s regulations, suggesting limited long-term implications for U.S. accelerator vendors.\n\nThe next major catalyst for AI is expected to be Computex (May 19–21 in Taiwan), where NVIDIA (NVDA) CEO is scheduled to deliver a keynote on its roadmap and new products.\n\n⸻\n\n2. 2025 Capex Outlook: Strong Despite Multiple Concerns\n\nCloud Capex is expected to remain strong in 2025. Google and Microsoft have reaffirmed their 2025 (FY26 for Microsoft) guidance. Meta has raised its 2025 Capex guidance from $60–65B to $64–72B. While most companies avoided commenting on 2026 forecasts, Microsoft indicated YoY Capex growth, noting a shift toward short-lived assets (servers, CPUs, GPUs) over long-term assets (buildings, land).\n\nAmazon has not updated its $105 billion 2025 Capex forecast from the previous call, but its custom silicon Trainium2 (MRVL) is being deployed across more instances.\n\n⸻\n\n3. Questions Around Capex Intensity and AI Sustainability\n\nCloud Capex intensity (Capex as a percentage of revenue) is rising rapidly, expected to reach 20–22% in 2025–2027 (vs. 17% in 2024 and 10–12% during 2018–2023). We expect continued AI expansion, supported by the following:\n1. Leading model sizes double every six months pre- and post-training\n2. Derivative models like DeepSeek\n3. Inference scaling during testing (token generation 100x vs standard inference)\n4. Still early stages of sovereign/enterprise AI adoption\n\n⸻\n\n4. Top AI Investment Picks: NVIDIA $NVDA, Broadcom $AVGO, Plus $AMD / $MRVL\n\nOur top AI picks are NVIDIA (NVDA) and Broadcom (AVGO), leaders in commercial/custom AI silicon, respectively. Additionally, AMD (AI GPUs) and MRVL (AI ASICs, optical/DSP) are expected to benefit from AI growth as Tier-2 players.\n\nAMD is projected to maintain a 3–4% share in the accelerator market by 2027 ($12.6B revenue), while MRVL is expected to maintain a 10–15% share in the custom ASIC market ($4B revenue).\n\nOverall, AMD/MRVL EPS is projected to grow at a +27% / +39% CAGR from 2024 to 2027, and with the AI market growing at +45% (based on TSMC), their PEG ratios are expected to fall below 1.0x.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.060384586975187116, "ret_signed_m1d": -0.060384586975187116, "alpha_spy_m1d": -0.02839427580405085, "alpha_c_m1d": -0.0013727048728443547, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04892135595253566, "ret_signed_p1d": 0.04892135595253566, "alpha_spy_p1d": 0.04231741563536562, "alpha_c_p1d": 0.014592596189640039, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.040843049571419154, "ret_signed_p1w": 0.040843049571419154, "alpha_spy_p1w": 0.020499749297673953, "alpha_c_p1w": 0.004618816643298107, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.10402567152839559, "ret_signed_p1m": 0.10402567152839559, "alpha_spy_p1m": 0.0695655010801497, "alpha_c_p1m": 0.00015470915218407377, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.37410848864319224, "ret_signed_p3m": 0.37410848864319224, "alpha_spy_p3m": 0.2864070938250045, "alpha_c_p3m": 0.14787544950147802, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.59478635939029, "ret_signed_p6m": 0.59478635939029, "alpha_spy_p6m": 0.42990567432731575, "alpha_c_p6m": 0.10891798220737714, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0634, 0.061, 0.0485, 0.0485, 0.0283, 0.0292, 0.0202, -0.0162, -0.0644, -0.0887, -0.0419, -0.11, -0.1027, -0.1569, -0.1564, -0.138, -0.1926, -0.1228, -0.1701, -0.1447, -0.1261, -0.139, -0.1202, -0.1249, -0.1511, -0.12, -0.1401, -0.135, -0.1369, -0.1504, -0.1909, -0.2238, -0.2368, -0.2444, -0.2395, -0.2234, -0.3049, -0.3398, -0.3044, -0.2958, -0.1644, -0.2224, -0.1789, -0.1951, -0.1924, -0.212, -0.2283, -0.2283, -0.2499, -0.2347, -0.2016, -0.1509, -0.1321, -0.1314, -0.1372, -0.1314, -0.1094, -0.081, -0.0941, -0.097, -0.0757, -0.0623, -0.0604, 0.0, 0.0489, 0.0476, 0.0499, 0.0317, 0.0408, 0.0456, 0.0368, 0.0404, 0.0322, 0.0322, 0.0635, 0.0806, 0.092, 0.0925, 0.1224, 0.1592, 0.1783, 0.1731, 0.1144, 0.1024, 0.104, 0.1414, 0.1557, 0.1224, 0.1377, 0.1254, 0.1339, 0.1339, 0.1309, 0.148, 0.1932, 0.1972, 0.2222, 0.2184, 0.2469, 0.1976, 0.2209, 0.2448, 0.2448, 0.2403, 0.2295, 0.2571, 0.2458, 0.2412, 0.2467, 0.2709, 0.2703, 0.2958, 0.2817, 0.3038, 0.2602, 0.2833, 0.306, 0.3127, 0.3313, 0.3454, 0.369, 0.3286, 0.3057, 0.3468, 0.3251, 0.3647, 0.3741, 0.3796, 0.3747, 0.4151, 0.3982, 0.4079, 0.3858, 0.3832, 0.3341, 0.3172, 0.3101, 0.33, 0.331, 0.3481, 0.3582, 0.3962, 0.3453, 0.3453, 0.3491, 0.3679, 0.3847, 0.5149, 0.5636, 0.523, 0.6718, 0.6268, 0.6279, 0.647, 0.6285, 0.566, 0.5622, 0.5604, 0.5352, 0.5359, 0.5376, 0.523, 0.5159, 0.4859, 0.495, 0.5107, 0.5324, 0.5333, 0.5202, 0.5244, 0.5656, 0.5634, 0.471, 0.6164, 0.5594, 0.592, 0.6048, 0.583, 0.5826, 0.5527, 0.542, 0.5601, 0.6047, 0.6406, 0.6901, 0.749, 0.7059, 0.6749, 0.6429, 0.5948]}
{"unit_id": "orig:1921433573565083836", "ticker_idx": 2, "ts": "2025-05-11T05:13:57+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD/MRVL EPS is projected to grow at a +27% / +39% CAGR from 2024 to 2027, and with the AI market growing at +45% (based on TSMC), their PEG ratios are expected to fall below 1.0x", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA top AI picks: NVDA and AVGO as leaders; AMD and MRVL as Tier-2 beneficiaries with PEG below 1x by 2027 on strong EPS CAGRs.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Q1 Cloud Capex Review: AI Investment Remains Robust\n\n1. AI Capex Outlook for 2025/2026: +44% / +4% YoY Growth\n\nDespite a range of concerns since the beginning of the year—such as Capex intensity (sustainability), DeepSeek (AI efficiency), ongoing China regulations, and the recent rollback of AI diffusion initiatives—demand for data center and AI-related Capex remains strong. In Q1, global hyperscaler Capex totaled $92 billion, slightly below the consensus of $93.8 billion. However, most hyperscalers indicated continued strong AI investment throughout 2025, with expectations for growth to persist into 2026.\n\nOverall, our consensus model shows 2025/2026 Capex reaching $414 billion / $432 billion, representing +44% / +4% YoY growth. This is 300bps/100bps higher than pre-earnings expectations ($402B/$429B), and 5,090bps/4,900bps higher compared to May 2024 projections.\n\nWhile CSPs (cloud service providers) have not provided extensive guidance on 2026 spending yet, the rollback of AI diffusion regulations has allowed for sovereign AI infrastructure development in most countries, supporting a solid investment environment.\n\nMeanwhile, China’s GPU regulations (which affect ~10% of global AI TAM) are now considered to have fully played out. Residual revenues (e.g., B20 deployments) are expected to be incremental. AMD has stated that its $500 billion AI TAM outlook already factors in the impact of China’s regulations, suggesting limited long-term implications for U.S. accelerator vendors.\n\nThe next major catalyst for AI is expected to be Computex (May 19–21 in Taiwan), where NVIDIA (NVDA) CEO is scheduled to deliver a keynote on its roadmap and new products.\n\n⸻\n\n2. 2025 Capex Outlook: Strong Despite Multiple Concerns\n\nCloud Capex is expected to remain strong in 2025. Google and Microsoft have reaffirmed their 2025 (FY26 for Microsoft) guidance. Meta has raised its 2025 Capex guidance from $60–65B to $64–72B. While most companies avoided commenting on 2026 forecasts, Microsoft indicated YoY Capex growth, noting a shift toward short-lived assets (servers, CPUs, GPUs) over long-term assets (buildings, land).\n\nAmazon has not updated its $105 billion 2025 Capex forecast from the previous call, but its custom silicon Trainium2 (MRVL) is being deployed across more instances.\n\n⸻\n\n3. Questions Around Capex Intensity and AI Sustainability\n\nCloud Capex intensity (Capex as a percentage of revenue) is rising rapidly, expected to reach 20–22% in 2025–2027 (vs. 17% in 2024 and 10–12% during 2018–2023). We expect continued AI expansion, supported by the following:\n1. Leading model sizes double every six months pre- and post-training\n2. Derivative models like DeepSeek\n3. Inference scaling during testing (token generation 100x vs standard inference)\n4. Still early stages of sovereign/enterprise AI adoption\n\n⸻\n\n4. Top AI Investment Picks: NVIDIA $NVDA, Broadcom $AVGO, Plus $AMD / $MRVL\n\nOur top AI picks are NVIDIA (NVDA) and Broadcom (AVGO), leaders in commercial/custom AI silicon, respectively. Additionally, AMD (AI GPUs) and MRVL (AI ASICs, optical/DSP) are expected to benefit from AI growth as Tier-2 players.\n\nAMD is projected to maintain a 3–4% share in the accelerator market by 2027 ($12.6B revenue), while MRVL is expected to maintain a 10–15% share in the custom ASIC market ($4B revenue).\n\nOverall, AMD/MRVL EPS is projected to grow at a +27% / +39% CAGR from 2024 to 2027, and with the AI market growing at +45% (based on TSMC), their PEG ratios are expected to fall below 1.0x.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04883468622422249, "ret_signed_m1d": -0.04883468622422249, "alpha_spy_m1d": -0.016844375053086225, "alpha_c_m1d": 0.010177195878120271, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04014054964522118, "ret_signed_p1d": 0.04014054964522118, "alpha_spy_p1d": 0.03353660932805114, "alpha_c_p1d": 0.005811789882325558, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.06122821817442814, "ret_signed_p1w": 0.06122821817442814, "alpha_spy_p1w": 0.04088491790068294, "alpha_c_p1w": 0.025003985246307092, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.13984456837858783, "ret_signed_p1m": 0.13984456837858783, "alpha_spy_p1m": 0.10538439793034193, "alpha_c_p1m": 0.03597360600237631, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.5945245053365891, "ret_signed_p3m": 0.5945245053365891, "alpha_spy_p3m": 0.5068231105184013, "alpha_c_p3m": 0.36829146619487485, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 1.3127080715844928, "ret_signed_p6m": 1.3127080715844928, "alpha_spy_p6m": 1.1478273865215185, "alpha_c_p6m": 0.8268396944015799, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0333, 0.0341, 0.0461, 0.0461, 0.057, 0.0608, 0.056, 0.0252, -0.0001, -0.0385, -0.0313, -0.0796, -0.0764, -0.0915, -0.0682, -0.0597, -0.0857, -0.0722, -0.1063, -0.1051, -0.0678, -0.0926, -0.0661, -0.0326, -0.0426, -0.0175, -0.0091, -0.0155, 0.053, 0.0619, 0.0191, -0.0136, -0.0453, -0.0498, -0.0494, -0.0477, -0.1324, -0.2068, -0.2264, -0.2766, -0.1043, -0.1796, -0.1361, -0.126, -0.1187, -0.1834, -0.1907, -0.1907, -0.2087, -0.2022, -0.164, -0.1262, -0.1061, -0.1085, -0.1115, -0.0996, -0.1061, -0.0862, -0.0696, -0.0879, -0.0718, -0.0594, -0.0488, 0.0, 0.0401, 0.0888, 0.0635, 0.0837, 0.0612, 0.0499, 0.0364, 0.024, 0.0203, 0.0203, 0.0596, 0.0438, 0.0454, 0.0241, 0.0602, 0.085, 0.0967, 0.07, 0.0746, 0.1259, 0.1398, 0.1204, 0.096, 0.0744, 0.169, 0.1755, 0.1727, 0.1727, 0.1861, 0.1985, 0.2803, 0.3263, 0.3289, 0.3301, 0.3124, 0.2589, 0.2812, 0.2755, 0.2755, 0.2468, 0.2747, 0.2802, 0.3333, 0.3542, 0.3526, 0.4392, 0.4806, 0.4836, 0.452, 0.4521, 0.431, 0.4674, 0.4994, 0.5397, 0.6062, 0.6411, 0.6603, 0.6307, 0.5881, 0.635, 0.6122, 0.5087, 0.5945, 0.5979, 0.5934, 0.6181, 0.7057, 0.6736, 0.6418, 0.6291, 0.5404, 0.5279, 0.5142, 0.5516, 0.5109, 0.5411, 0.5458, 0.5592, 0.5042, 0.5042, 0.5013, 0.4995, 0.4964, 0.3979, 0.4004, 0.4412, 0.4756, 0.4398, 0.4666, 0.4906, 0.4841, 0.4721, 0.4606, 0.4557, 0.4779, 0.4882, 0.488, 0.4916, 0.4748, 0.4924, 0.4964, 0.5169, 0.5698, 0.523, 0.8841, 0.9563, 1.1787, 1.154, 0.9876, 1.0017, 1.0171, 1.2068, 1.1694, 1.1558, 1.2249, 1.2015, 1.1294, 1.1734, 1.3393, 1.4017, 1.3863, 1.4448, 1.357, 1.3688, 1.4015, 1.3127]}
{"unit_id": "orig:1921433573565083836", "ticker_idx": 3, "ts": "2025-05-11T05:13:57+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD/MRVL EPS is projected to grow at a +27% / +39% CAGR from 2024 to 2027, and with the AI market growing at +45% (based on TSMC), their PEG ratios are expected to fall below 1.0x", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA top AI picks: NVDA and AVGO as leaders; AMD and MRVL as Tier-2 beneficiaries with PEG below 1x by 2027 on strong EPS CAGRs.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Q1 Cloud Capex Review: AI Investment Remains Robust\n\n1. AI Capex Outlook for 2025/2026: +44% / +4% YoY Growth\n\nDespite a range of concerns since the beginning of the year—such as Capex intensity (sustainability), DeepSeek (AI efficiency), ongoing China regulations, and the recent rollback of AI diffusion initiatives—demand for data center and AI-related Capex remains strong. In Q1, global hyperscaler Capex totaled $92 billion, slightly below the consensus of $93.8 billion. However, most hyperscalers indicated continued strong AI investment throughout 2025, with expectations for growth to persist into 2026.\n\nOverall, our consensus model shows 2025/2026 Capex reaching $414 billion / $432 billion, representing +44% / +4% YoY growth. This is 300bps/100bps higher than pre-earnings expectations ($402B/$429B), and 5,090bps/4,900bps higher compared to May 2024 projections.\n\nWhile CSPs (cloud service providers) have not provided extensive guidance on 2026 spending yet, the rollback of AI diffusion regulations has allowed for sovereign AI infrastructure development in most countries, supporting a solid investment environment.\n\nMeanwhile, China’s GPU regulations (which affect ~10% of global AI TAM) are now considered to have fully played out. Residual revenues (e.g., B20 deployments) are expected to be incremental. AMD has stated that its $500 billion AI TAM outlook already factors in the impact of China’s regulations, suggesting limited long-term implications for U.S. accelerator vendors.\n\nThe next major catalyst for AI is expected to be Computex (May 19–21 in Taiwan), where NVIDIA (NVDA) CEO is scheduled to deliver a keynote on its roadmap and new products.\n\n⸻\n\n2. 2025 Capex Outlook: Strong Despite Multiple Concerns\n\nCloud Capex is expected to remain strong in 2025. Google and Microsoft have reaffirmed their 2025 (FY26 for Microsoft) guidance. Meta has raised its 2025 Capex guidance from $60–65B to $64–72B. While most companies avoided commenting on 2026 forecasts, Microsoft indicated YoY Capex growth, noting a shift toward short-lived assets (servers, CPUs, GPUs) over long-term assets (buildings, land).\n\nAmazon has not updated its $105 billion 2025 Capex forecast from the previous call, but its custom silicon Trainium2 (MRVL) is being deployed across more instances.\n\n⸻\n\n3. Questions Around Capex Intensity and AI Sustainability\n\nCloud Capex intensity (Capex as a percentage of revenue) is rising rapidly, expected to reach 20–22% in 2025–2027 (vs. 17% in 2024 and 10–12% during 2018–2023). We expect continued AI expansion, supported by the following:\n1. Leading model sizes double every six months pre- and post-training\n2. Derivative models like DeepSeek\n3. Inference scaling during testing (token generation 100x vs standard inference)\n4. Still early stages of sovereign/enterprise AI adoption\n\n⸻\n\n4. Top AI Investment Picks: NVIDIA $NVDA, Broadcom $AVGO, Plus $AMD / $MRVL\n\nOur top AI picks are NVIDIA (NVDA) and Broadcom (AVGO), leaders in commercial/custom AI silicon, respectively. Additionally, AMD (AI GPUs) and MRVL (AI ASICs, optical/DSP) are expected to benefit from AI growth as Tier-2 players.\n\nAMD is projected to maintain a 3–4% share in the accelerator market by 2027 ($12.6B revenue), while MRVL is expected to maintain a 10–15% share in the custom ASIC market ($4B revenue).\n\nOverall, AMD/MRVL EPS is projected to grow at a +27% / +39% CAGR from 2024 to 2027, and with the AI market growing at +45% (based on TSMC), their PEG ratios are expected to fall below 1.0x.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07519371156457744, "ret_signed_m1d": -0.07519371156457744, "alpha_spy_m1d": -0.04320340039344117, "alpha_c_m1d": -0.016181829462234676, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015193770867061529, "ret_signed_p1d": 0.015193770867061529, "alpha_spy_p1d": 0.008589830549891486, "alpha_c_p1d": -0.019134988895834093, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": -0.03007744904434051, "ret_signed_p1w": -0.03007744904434051, "alpha_spy_p1w": -0.05042074931808571, "alpha_c_p1w": -0.06630168197246156, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.06728685204623219, "ret_signed_p1m": 0.06728685204623219, "alpha_spy_p1m": 0.032826681597986296, "alpha_c_p1m": -0.036584110329979325, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.17693157086350353, "ret_signed_p3m": 0.17693157086350353, "alpha_spy_p3m": 0.0892301760453158, "alpha_c_p3m": -0.04930146827821069, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.35999584882610725, "ret_signed_p6m": 0.35999584882610725, "alpha_spy_p6m": 0.19511516376313298, "alpha_c_p6m": -0.12587252835680562, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.6326, 0.6028, 0.6494, 0.6494, 0.6615, 0.7196, 0.6835, 0.6076, 0.5163, 0.4402, 0.4696, 0.3574, 0.422, 0.3292, 0.3679, 0.3959, 0.1193, 0.097, 0.017, 0.0401, 0.0817, 0.0648, 0.0645, 0.0902, 0.0574, 0.0811, 0.0916, 0.0901, 0.1268, 0.1099, 0.0339, 0.0044, -0.0392, -0.0465, -0.0295, -0.0208, -0.1383, -0.2345, -0.2102, -0.2252, -0.056, -0.1812, -0.1722, -0.1898, -0.1733, -0.1949, -0.1984, -0.1984, -0.2344, -0.2152, -0.1664, -0.1112, -0.0865, -0.0899, -0.0902, -0.095, -0.055, -0.0336, -0.0391, -0.0509, -0.127, -0.1065, -0.0752, 0.0, 0.0152, 0.0217, 0.0109, -0.0115, -0.0301, -0.0478, -0.0681, -0.0411, -0.0591, -0.0591, -0.0105, 0.0014, -0.0119, -0.0668, -0.047, -0.0332, 0.0279, 0.0102, 0.0597, 0.0719, 0.0673, 0.058, 0.0797, 0.0417, 0.0918, 0.0851, 0.162, 0.162, 0.1397, 0.0974, 0.166, 0.1772, 0.2398, 0.1963, 0.2, 0.182, 0.1512, 0.1656, 0.1656, 0.1093, 0.1155, 0.1203, 0.1374, 0.1282, 0.1251, 0.1236, 0.0993, 0.1173, 0.1583, 0.1336, 0.117, 0.1369, 0.1488, 0.1515, 0.1779, 0.1845, 0.2683, 0.2471, 0.1552, 0.1875, 0.189, 0.1687, 0.1769, 0.2001, 0.1991, 0.2073, 0.2308, 0.2264, 0.1822, 0.1907, 0.1183, 0.1051, 0.1049, 0.1327, 0.1319, 0.1523, 0.1605, 0.1983, -0.0245, -0.0245, 0.0024, -0.0332, -0.0054, -0.0173, 0.0241, 0.0371, 0.0412, 0.0332, 0.045, 0.0463, 0.0685, 0.1014, 0.1518, 0.1523, 0.172, 0.1578, 0.2427, 0.3004, 0.2905, 0.2784, 0.3045, 0.3017, 0.3375, 0.3378, 0.3797, 0.3495, 0.4353, 0.407, 0.3293, 0.3879, 0.3387, 0.3802, 0.3699, 0.3656, 0.3328, 0.3083, 0.2585, 0.2852, 0.3063, 0.3774, 0.3735, 0.3997, 0.3752, 0.4555, 0.4032, 0.36]}
{"unit_id": "orig:1929341166409130146", "ticker_idx": 0, "ts": "2025-06-02T00:55:54+00:00", "symbol": "memory semiconductors", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the profit cycle will be prolonged, and the logic for re-rating memory stocks will gain strength", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory sector re-rating thesis: DDR4 exit by all major players signals shift to high-margin/high-ASP products, chronic undersupply, prolonged profit cycle — bullish memory stocks.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T"], "resolution_method": "basket", "resolution_reason": "Global memory semicon basket: Micron, SK Hynix, Samsung, Kioxia", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Background of CXMT’s Early DDR4 Production Termination\n[Samsung Securities – Semiconductors, IT / Lee Jong-wook]\n\nOn Tuesday (May 27), CXMT announced a phased termination of DDR4 production by 2026. On the same night, Micron also announced the end of LPDDR4 production. Following Samsung Electronics’ declaration in April, memory makers are now rushing to abandon DDR4 production.\n\nThis raises two key questions:\n1. Why is DDR4 being phased out so quickly, when DDR3 continued to be produced for more than five years even after its demand share fell below 10%?\n2. If the early retirement of DDR4 was due to the threat from China, why is even CXMT—China’s own DRAM maker—quickly exiting the DDR4 market?\n\nThe rules of the game have changed.\n1. The era of cost is over. In traditional business, technology was synonymous with cost. Securing MS (market share) and expanding capacity helped reduce costs, but that no longer holds meaning.\n2. The era of performance has arrived. Volume growth has halted and the limits of cost reduction have been reached, so the focus must shift to high-margin segments—even if they come at higher prices.\n3. There is no longer a reason to expand capacity. The immediate cause of DDR4’s early termination is indeed the threat from China, but the fundamental reason is the declining importance of capacity and market share. Even for CXMT, having the largest capacity in DDR4 now means nothing.\n\nTo win in this new game, memory makers must abandon stable cash cow products—even if they are high-yield and customer-secured—and instead prioritize securing high-end, custom products with uncertain yields and demand.\nAmid these efforts, downside risks in supply have increased, and the market will suffer from chronic undersupply.\nAs a result, the profit cycle will be prolonged, and the logic for re-rating memory stocks will gain strength.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0026950946610285187, "ret_signed_m1d": -0.0026950946610285187, "alpha_spy_m1d": 0.002906254629001681, "alpha_c_m1d": 0.01189599993466367, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008734561798307672, "ret_signed_p1d": 0.008734561798307672, "alpha_spy_p1d": 0.0030319720101201764, "alpha_c_p1d": -0.013830135807906485, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": 0.08628632590856977, "ret_signed_p1w": 0.08628632590856977, "alpha_spy_p1w": 0.07452690874553469, "alpha_c_p1w": 0.030141765286218147, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.21450870823939938, "ret_signed_p1m": 0.21450870823939938, "alpha_spy_p1m": 0.16935013405974686, "alpha_c_p1m": 0.08076445225770301, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.2527737292106815, "ret_signed_p3m": 0.2527737292106815, "alpha_spy_p3m": 0.1547016836807879, "alpha_c_p3m": 0.024331550572003147, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 1.8752459088008115, "ret_signed_p6m": 1.8752459088008115, "alpha_spy_p6m": 1.7298418472445056, "alpha_c_p6m": 1.4777530792494098, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [0.013, -0.0048, 0.0524, 0.0275, 0.0187, 0.0634, 0.0645, 0.1125, 0.1536, 0.1521, 0.1423, 0.1576, 0.1422, 0.1157, 0.0866, 0.0936, 0.0693, 0.0509, 0.0042, 0.0042, 0.0126, -0.0644, -0.1332, -0.2033, -0.1682, -0.1595, -0.0952, -0.1141, -0.1112, -0.1092, -0.1479, -0.142, -0.1476, -0.1478, -0.1491, -0.1209, -0.1161, -0.0956, -0.0991, -0.1049, -0.1139, -0.1145, -0.1031, -0.1038, -0.1036, -0.0794, -0.0718, -0.0552, 0.0014, 0.0125, 0.0276, 0.0139, 0.0238, 0.0115, 0.0127, 0.0026, -0.0209, -0.0253, -0.0171, -0.0142, 0.004, 0.022, -0.0027, 0.0, 0.0087, 0.0375, 0.0664, 0.0783, 0.0863, 0.0961, 0.1228, 0.1031, 0.0866, 0.1075, 0.1233, 0.1268, 0.1307, 0.1725, 0.1988, 0.247, 0.2554, 0.2559, 0.2453, 0.2441, 0.2145, 0.2176, 0.2488, 0.2159, 0.2137, 0.2488, 0.2569, 0.2622, 0.2612, 0.2459, 0.2549, 0.2472, 0.2113, 0.2077, 0.212, 0.1946, 0.2157, 0.2233, 0.2038, 0.2185, 0.2185, 0.2387, 0.2344, 0.187, 0.1927, 0.202, 0.1857, 0.2012, 0.2263, 0.2477, 0.2926, 0.2931, 0.2897, 0.2791, 0.2532, 0.2478, 0.2143, 0.1918, 0.2197, 0.2258, 0.2282, 0.2286, 0.2528, 0.2619, 0.2469, 0.2501, 0.2571, 0.2764, 0.3567, 0.3744, 0.4067, 0.4818, 0.5811, 0.6874, 0.6966, 0.7659, 0.7129, 0.7814, 0.767, 0.8169, 0.8388, 0.842, 0.8239, 0.7265, 0.7992, 0.8298, 0.8708, 1.0227, 1.1299, 1.1554, 1.0893, 1.1145, 1.1464, 1.1714, 1.1793, 1.1329, 1.2138, 1.3592, 1.3353, 1.4455, 1.4098, 1.4418, 1.4573, 1.7136, 1.8899, 1.8126, 1.9614, 2.0856, 2.0798, 2.1973, 2.0781, 2.0732, 2.2057, 2.2508, 2.4938, 2.4757, 2.4868, 2.4322, 1.9938, 2.1948, 1.9912, 2.0413, 2.0704, 1.8361, 1.8855, 1.8752]}
{"unit_id": "orig:1929341166409130146", "ticker_idx": 1, "ts": "2025-06-02T00:55:54+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron also announced the end of LPDDR4 production", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory sector re-rating thesis: DDR4 exit by all major players signals shift to high-margin/high-ASP products, chronic undersupply, prolonged profit cycle — bullish memory stocks.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Background of CXMT’s Early DDR4 Production Termination\n[Samsung Securities – Semiconductors, IT / Lee Jong-wook]\n\nOn Tuesday (May 27), CXMT announced a phased termination of DDR4 production by 2026. On the same night, Micron also announced the end of LPDDR4 production. Following Samsung Electronics’ declaration in April, memory makers are now rushing to abandon DDR4 production.\n\nThis raises two key questions:\n1. Why is DDR4 being phased out so quickly, when DDR3 continued to be produced for more than five years even after its demand share fell below 10%?\n2. If the early retirement of DDR4 was due to the threat from China, why is even CXMT—China’s own DRAM maker—quickly exiting the DDR4 market?\n\nThe rules of the game have changed.\n1. The era of cost is over. In traditional business, technology was synonymous with cost. Securing MS (market share) and expanding capacity helped reduce costs, but that no longer holds meaning.\n2. The era of performance has arrived. Volume growth has halted and the limits of cost reduction have been reached, so the focus must shift to high-margin segments—even if they come at higher prices.\n3. There is no longer a reason to expand capacity. The immediate cause of DDR4’s early termination is indeed the threat from China, but the fundamental reason is the declining importance of capacity and market share. Even for CXMT, having the largest capacity in DDR4 now means nothing.\n\nTo win in this new game, memory makers must abandon stable cash cow products—even if they are high-yield and customer-secured—and instead prioritize securing high-end, custom products with uncertain yields and demand.\nAmid these efforts, downside risks in supply have increased, and the market will suffer from chronic undersupply.\nAs a result, the profit cycle will be prolonged, and the logic for re-rating memory stocks will gain strength.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.037889552037211005, "ret_signed_m1d": -0.037889552037211005, "alpha_spy_m1d": -0.032288202747180805, "alpha_c_m1d": -0.023298457441518816, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04145453792004772, "ret_signed_p1d": 0.04145453792004772, "alpha_spy_p1d": 0.03575194813186022, "alpha_c_p1d": 0.01888984031383356, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": 0.13006726401685498, "ret_signed_p1w": 0.13006726401685498, "alpha_spy_p1w": 0.1183078468538199, "alpha_c_p1w": 0.07392270339450335, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.23130991868697715, "ret_signed_p1m": 0.23130991868697715, "alpha_spy_p1m": 0.18615134450732462, "alpha_c_p1m": 0.09756566270528078, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.2437853050794292, "ret_signed_p3m": 0.2437853050794292, "alpha_spy_p3m": 0.14571325954953562, "alpha_c_p3m": 0.015343126440750865, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 1.2905082981128015, "ret_signed_p6m": 1.2905082981128015, "alpha_spy_p6m": 1.1451042365564956, "alpha_c_p6m": 0.8930154685613998, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [-0.0404, -0.0919, -0.0544, -0.1142, -0.0942, -0.0271, -0.0349, 0.0252, 0.0488, 0.0347, 0.0382, 0.0477, -0.0365, -0.0139, -0.042, -0.0628, -0.0727, -0.1004, -0.115, -0.0965, -0.0976, -0.2428, -0.3408, -0.3036, -0.3325, -0.2069, -0.2865, -0.2916, -0.2766, -0.2764, -0.2938, -0.2992, -0.2992, -0.3202, -0.2849, -0.2572, -0.2114, -0.1874, -0.1998, -0.2169, -0.2162, -0.2079, -0.1778, -0.1809, -0.18, -0.1585, -0.1327, -0.1255, -0.06, -0.0127, -0.0291, -0.0278, -0.0018, 0.0048, -0.0008, -0.0238, -0.0341, -0.049, -0.049, -0.0183, -0.0204, -0.0141, -0.0379, 0.0, 0.0415, 0.0516, 0.0826, 0.1057, 0.1301, 0.1626, 0.1818, 0.1833, 0.1774, 0.2206, 0.2257, 0.2408, 0.2408, 0.2589, 0.2434, 0.3028, 0.2961, 0.2834, 0.2707, 0.2553, 0.2313, 0.24, 0.2456, 0.2456, 0.2226, 0.2685, 0.2462, 0.2551, 0.2696, 0.2092, 0.2245, 0.187, 0.1547, 0.1662, 0.1544, 0.1135, 0.1197, 0.1391, 0.1343, 0.1342, 0.1414, 0.1698, 0.1127, 0.0692, 0.0987, 0.1119, 0.109, 0.1405, 0.2121, 0.2613, 0.3024, 0.2669, 0.2773, 0.2323, 0.2596, 0.2443, 0.195, 0.1805, 0.1997, 0.1869, 0.1877, 0.2005, 0.2438, 0.2133, 0.2133, 0.2079, 0.2103, 0.2663, 0.3393, 0.3402, 0.3788, 0.4273, 0.5351, 0.603, 0.6085, 0.6192, 0.6311, 0.7218, 0.659, 0.6783, 0.6965, 0.6486, 0.5989, 0.6034, 0.671, 0.7058, 0.857, 0.8733, 0.9161, 0.948, 0.8943, 1.005, 0.962, 0.8526, 0.9665, 0.9083, 0.958, 1.0661, 1.0645, 1.1093, 1.0636, 1.0247, 1.1087, 1.2343, 1.2453, 1.2638, 1.3119, 1.2852, 1.2828, 1.3943, 1.2242, 1.4228, 1.4313, 1.4271, 1.584, 1.4596, 1.4983, 1.4172, 1.518, 1.4682, 1.331, 1.3047, 1.0542, 1.1155, 1.2844, 1.2905]}
{"unit_id": "orig:1929341166409130146", "ticker_idx": 2, "ts": "2025-06-02T00:55:54+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Following Samsung Electronics' declaration in April, memory makers are now rushing to abandon DDR4 production", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory sector re-rating thesis: DDR4 exit by all major players signals shift to high-margin/high-ASP products, chronic undersupply, prolonged profit cycle — bullish memory stocks.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Background of CXMT’s Early DDR4 Production Termination\n[Samsung Securities – Semiconductors, IT / Lee Jong-wook]\n\nOn Tuesday (May 27), CXMT announced a phased termination of DDR4 production by 2026. On the same night, Micron also announced the end of LPDDR4 production. Following Samsung Electronics’ declaration in April, memory makers are now rushing to abandon DDR4 production.\n\nThis raises two key questions:\n1. Why is DDR4 being phased out so quickly, when DDR3 continued to be produced for more than five years even after its demand share fell below 10%?\n2. If the early retirement of DDR4 was due to the threat from China, why is even CXMT—China’s own DRAM maker—quickly exiting the DDR4 market?\n\nThe rules of the game have changed.\n1. The era of cost is over. In traditional business, technology was synonymous with cost. Securing MS (market share) and expanding capacity helped reduce costs, but that no longer holds meaning.\n2. The era of performance has arrived. Volume growth has halted and the limits of cost reduction have been reached, so the focus must shift to high-margin segments—even if they come at higher prices.\n3. There is no longer a reason to expand capacity. The immediate cause of DDR4’s early termination is indeed the threat from China, but the fundamental reason is the declining importance of capacity and market share. Even for CXMT, having the largest capacity in DDR4 now means nothing.\n\nTo win in this new game, memory makers must abandon stable cash cow products—even if they are high-yield and customer-secured—and instead prioritize securing high-end, custom products with uncertain yields and demand.\nAmid these efforts, downside risks in supply have increased, and the market will suffer from chronic undersupply.\nAs a result, the profit cycle will be prolonged, and the logic for re-rating memory stocks will gain strength.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010563322371586525, "ret_signed_m1d": -0.010563322371586525, "alpha_spy_m1d": -0.004961973081556326, "alpha_c_m1d": -0.0018061275593284076, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.008757194812258118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.005702589788187495, "alpha_c_p1d": -0.015024643506287738, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.015024643506287738, "ret_p1w": 0.05281696178281403, "ret_signed_p1w": 0.05281696178281403, "alpha_spy_p1w": 0.04105754461977895, "alpha_c_p1w": 0.02457058982986937, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.02824637195294466, "ret_p1m": 0.06636010447637197, "ret_signed_p1m": 0.06636010447637197, "alpha_spy_p1m": 0.02120153029671945, "alpha_c_p1m": -0.01278106816018898, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.07914117263656095, "ret_p3m": 0.23286822017917408, "ret_signed_p3m": 0.23286822017917408, "alpha_spy_p3m": 0.1347961746492805, "alpha_c_p3m": 0.08677775202534832, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.14609046815382576, "ret_p6m": 0.7668106187644699, "ret_signed_p6m": 0.7668106187644699, "alpha_spy_p6m": 0.621406557208164, "alpha_c_p6m": 0.5593160310340552, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.20749458773041463, "price_path": [-0.0549, -0.0497, -0.0602, -0.0602, -0.0619, -0.0392, -0.0427, -0.0427, 0.0081, 0.0081, 0.0238, 0.0536, 0.0799, 0.0589, 0.0466, 0.0746, 0.0816, 0.0599, 0.0176, 0.0352, 0.0352, 0.0141, -0.0123, -0.0634, -0.0581, -0.0669, -0.007, -0.0282, -0.0106, -0.0035, -0.037, -0.0299, -0.0264, -0.0246, -0.0317, -0.0194, -0.0194, -0.0194, -0.0176, -0.0176, -0.0229, -0.0229, -0.044, -0.044, -0.044, -0.0387, -0.0387, -0.0352, 0.0141, 0.0018, 0.0106, 0.0088, 0.0, -0.0176, -0.0158, -0.0194, -0.037, -0.0458, -0.037, -0.0511, -0.0158, -0.0123, -0.0106, 0.0, 0.0, 0.0176, 0.0405, 0.0405, 0.0528, 0.0423, 0.0546, 0.0475, 0.0264, 0.007, 0.0229, 0.0528, 0.0423, 0.0475, 0.0211, 0.0651, 0.0792, 0.0599, 0.077, 0.0593, 0.0664, 0.077, 0.1301, 0.1213, 0.0929, 0.0876, 0.0699, 0.0805, 0.1089, 0.1071, 0.1284, 0.1461, 0.1815, 0.1886, 0.201, 0.1691, 0.1762, 0.1691, 0.1673, 0.247, 0.2506, 0.286, 0.2648, 0.2205, 0.2346, 0.2382, 0.2187, 0.2488, 0.2718, 0.2577, 0.2594, 0.2736, 0.2683, 0.2683, 0.24, 0.24, 0.2488, 0.2506, 0.2648, 0.2665, 0.2453, 0.2506, 0.2329, 0.2346, 0.1974, 0.224, 0.2364, 0.2417, 0.2311, 0.2417, 0.2665, 0.286, 0.3002, 0.3356, 0.3551, 0.4065, 0.3852, 0.4224, 0.4224, 0.4791, 0.5003, 0.5127, 0.5251, 0.4755, 0.4981, 0.4928, 0.5302, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.6796, 0.6601, 0.6298, 0.6903, 0.7383, 0.7419, 0.7455, 0.7348, 0.7544, 0.717, 0.7579, 0.8149, 0.7704, 0.7882, 0.8522, 0.9127, 0.9768, 0.8664, 0.7899, 0.765, 0.7419, 0.7899, 0.8415, 0.8344, 0.8291, 0.7294, 0.7899, 0.7401, 0.717, 0.7899, 0.6867, 0.7205, 0.7668]}
{"unit_id": "thread:1927485785143448001", "ticker_idx": 0, "ts": "2025-05-27T22:03:16+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "With shipments normalized ahead of Nvidia's quarterly earnings release on Wednesday, investors should take note that despite early technical hiccups, Blackwell shipments are proceeding at a fast pace.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish NVDA: GB200 rack issues resolved, Blackwell shipments accelerating, GB300 launching Q3 to offset China headwinds and support revenue targets.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Nvidia’s Suppliers Resolve AI ‘Rack’ Issues, Boosting Sales – FT\n\n• Nvidia’s suppliers have resolved technical issues that had delayed shipments, accelerating production of the company’s flagship AI data center product, the GB200 rack.\n\n• At last week’s Computex conference in Taipei, Nvidia’s Taiwanese partners—Foxconn, Inventec, and Wistron—emphasized that GB200 rack shipments began at the end of Q1 and that production capacity is now rapidly scaling up.\n\n• Each GB200 AI rack includes 36 “Grace” central processing units and 72 Blackwell graphics processing units, all connected via Nvidia’s NVLink communication system.\n\n• Nvidia’s supply chain partners spent several months addressing a range of challenges with the GB200 racks, including overheating from the 72 high-performance GPUs, coolant leaks in the liquid-cooling system, and software bugs and chip-to-chip connectivity issues stemming from the complexity of synchronizing such a large number of processors.\n\n• With shipments normalized ahead of Nvidia’s quarterly earnings release on Wednesday, investors should take note that despite early technical hiccups, Blackwell shipments are proceeding at a fast pace.\n\n• Nvidia has also signaled that its next-generation GB300 AI rack—designed to handle more complex inference models like OpenAI’s 01 and DeepSeek’s R1 with enhanced memory capacity—will launch in Q3.\n\n• To accelerate deployment, Nvidia has made compromises in several aspects of the GB300’s design.\n\n• Notably, Nvidia had initially planned to introduce a new chipboard layout called “Cordelia,” which would allow for individual GPU replacement. However, due to installation issues encountered in April, the company has reverted to the current “Bianca” design used in the GB200 racks.\n\n• This decision is expected to help offset declining China sales in the second half of the year and support Nvidia’s revenue targets through an earlier-than-expected product launch.\n\n• However, according to supply chain sources, Cordelia has the potential to offer better margins and make maintenance easier for customers.\n\n• Nvidia has not abandoned Cordelia and has informed suppliers that it intends to implement the redesign in its next-generation AI chips.\n\n---\n\nhttps://t.co/RNwM9bJskf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031070050974117636, "ret_signed_m1d": -0.031070050974117636, "alpha_spy_m1d": -0.010702831348300745, "alpha_c_m1d": -0.0003316124021961464, "bench_spy_m1d": -0.02036721962581689, "bench_c_m1d": -0.03073843857192149, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005092117794083917, "ret_signed_p1d": -0.005092117794083917, "alpha_spy_p1d": 0.0006932966829971132, "alpha_c_p1d": 0.005562224713679842, "bench_spy_p1d": -0.00578541447708103, "bench_c_p1d": -0.010654342507763759, "ret_p1w": 0.04221407101909369, "ret_signed_p1w": 0.04221407101909369, "alpha_spy_p1w": 0.03385754687530862, "alpha_c_p1w": 0.02662032188993968, "bench_spy_p1w": 0.008356524143785071, "bench_c_p1w": 0.015593749129154011, "ret_p1m": 0.13889841614726683, "ret_signed_p1m": 0.13889841614726683, "alpha_spy_p1m": 0.10884724041286109, "alpha_c_p1m": 0.015495539690978166, "bench_spy_p1m": 0.030051175734405744, "bench_c_p1m": 0.12340287645628867, "ret_p3m": 0.3136707616043415, "ret_signed_p3m": 0.3136707616043415, "alpha_spy_p3m": 0.2188256380657594, "alpha_c_p3m": 0.11499331017395487, "bench_spy_p3m": 0.0948451235385821, "bench_c_p3m": 0.19867745143038662, "ret_p6m": 0.3767046424124256, "ret_signed_p6m": 0.3767046424124256, "alpha_spy_p6m": 0.24935722555262974, "alpha_c_p6m": -0.008934470055188326, "bench_spy_p6m": 0.1273474168597959, "bench_c_p6m": 0.38563911246761395, "price_path": [-0.1134, -0.0782, -0.1583, -0.1441, -0.1344, -0.1841, -0.1684, -0.2106, -0.1974, -0.1458, -0.147, -0.1021, -0.1179, -0.1481, -0.1327, -0.1252, -0.1314, -0.104, -0.1093, -0.1604, -0.1776, -0.1906, -0.2001, -0.1871, -0.1851, -0.2487, -0.304, -0.2794, -0.2893, -0.1562, -0.2061, -0.1813, -0.183, -0.172, -0.2289, -0.251, -0.251, -0.2848, -0.2702, -0.242, -0.2145, -0.1807, -0.1976, -0.1954, -0.1962, -0.1763, -0.155, -0.16, -0.1621, -0.1361, -0.1338, -0.1391, -0.0923, -0.0411, -0.0012, -0.0049, -0.0007, 0.0005, -0.0083, -0.0273, -0.0197, -0.0311, -0.0311, 0.0, -0.0051, 0.0272, -0.0027, 0.0139, 0.0422, 0.0474, 0.0331, 0.0459, 0.0526, 0.0624, 0.0542, 0.0702, 0.0478, 0.0679, 0.0637, 0.0737, 0.0737, 0.0617, 0.0641, 0.0916, 0.1389, 0.1441, 0.1643, 0.1661, 0.1314, 0.1606, 0.176, 0.176, 0.1679, 0.1809, 0.2022, 0.2112, 0.2172, 0.2109, 0.2599, 0.2648, 0.2768, 0.2725, 0.2649, 0.2328, 0.2605, 0.2823, 0.2805, 0.3045, 0.2954, 0.3231, 0.3128, 0.2822, 0.3285, 0.3157, 0.3242, 0.3342, 0.3484, 0.3437, 0.3518, 0.3402, 0.3434, 0.3318, 0.3433, 0.2963, 0.2946, 0.2915, 0.3137, 0.3271, 0.3416, 0.3403, 0.3298, 0.2856, 0.2856, 0.2605, 0.2593, 0.267, 0.2327, 0.2422, 0.2603, 0.3088, 0.3077, 0.3125, 0.312, 0.2908, 0.2569, 0.3008, 0.304, 0.3552, 0.317, 0.3062, 0.3115, 0.3152, 0.3422, 0.3771, 0.382, 0.3942, 0.3848, 0.3695, 0.3658, 0.3958, 0.4214, 0.3519, 0.39, 0.3288, 0.3273, 0.3419, 0.3523, 0.3481, 0.3371, 0.3306, 0.3445, 0.3748, 0.4134, 0.4838, 0.5282, 0.4975, 0.4946, 0.527, 0.4665, 0.4408, 0.3882, 0.3887, 0.4692, 0.4257, 0.4304, 0.3792, 0.4036, 0.3773, 0.3386, 0.3767]}
{"unit_id": "quote:1932398326680588538", "ticker_idx": 0, "ts": "2025-06-10T11:23:57+00:00", "symbol": "廣達", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "maintaining a positive outlook on Quanta", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Maintains positive/bullish stance on Quanta on improved GB200 yields, ahead-of-schedule GB300/Meta ASIC, and MSFT ASIC taking shape.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "廣達 = Quanta Computer, TWSE 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "Meta ASIC Project Enters Testing Phase; Quanta and Chenming Electric Collaborate\n\nThe Meta ASIC project has now entered its testing phase. Quanta (廣達) and Chenming Electric (晟銘電) are currently working together on the production and assembly of components for this project. Finalization is expected to occur progressively in Q3 2025, with small-scale production commencing in Q4 2025.\n\nImproved Yields and Accelerated Project Timelines:\n\nGB200 production yield has improved.\n\nThe progress of GB300 and the Meta ASIC project is better than anticipated.\n\nA Microsoft (MSFT) ASIC project is also taking shape.\nThese positive developments lead to maintaining a positive outlook on Quanta.", "reply_text": null, "parent_text": null, "quoted_text": "Meta ASIC專案已經入測試階段，目前廣達及晟銘電正攜手生產及組裝該專案之零組件，預計3Q25會陸續定案，並於4Q25開始小量生產\n\nGB200生產良率轉佳，GB300及Meta ASIC進度優於預期，MSFT ASIC專案成形，維持廣達正向看法。", "parent_author": null, "ret_m1d": -0.026454967379172967, "ret_signed_m1d": -0.026454967379172967, "alpha_spy_m1d": -0.02081726658329841, "alpha_c_m1d": -0.020891814679203646, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.00556315269996932, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0017636720106898318, "ret_signed_p1d": -0.0017636720106898318, "alpha_spy_p1d": 0.0010882425800983198, "alpha_c_p1d": 0.00043670367843673397, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0022003756891265658, "ret_p1w": -0.0017636720106898318, "ret_signed_p1w": -0.0017636720106898318, "alpha_spy_p1w": 0.007438922023669914, "alpha_c_p1w": -0.0036319876456215594, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": 0.0018683156349317276, "ret_p1m": 0.019989335412115894, "ret_signed_p1m": 0.019989335412115894, "alpha_spy_p1m": -0.01785790515066732, "alpha_c_p1m": -0.05219626449654857, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.07218559990866447, "ret_p3m": -0.04468390143466672, "ret_signed_p3m": -0.04468390143466672, "alpha_spy_p3m": -0.12108081751795008, "alpha_c_p3m": -0.13778660890912708, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.09310270747446037, "ret_p6m": 0.08096695872479676, "ret_signed_p6m": 0.08096695872479676, "alpha_spy_p6m": -0.05953426168172693, "alpha_c_p6m": -0.12650571767669172, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.20747267640148848, "price_path": [-0.0917, -0.0952, -0.1005, -0.0847, -0.1058, -0.0935, -0.097, -0.1199, -0.1199, -0.1111, -0.1323, -0.1499, -0.2081, -0.1746, -0.1781, -0.1781, -0.1781, -0.2593, -0.3245, -0.3704, -0.3086, -0.2399, -0.1958, -0.1958, -0.1993, -0.1887, -0.1869, -0.1817, -0.2116, -0.1711, -0.1887, -0.1711, -0.1358, -0.1446, -0.1587, -0.1587, -0.1146, -0.1305, -0.1041, -0.1041, -0.1129, -0.0952, -0.0882, -0.0758, -0.0247, -0.0212, -0.03, -0.0617, -0.06, -0.0406, -0.0564, -0.0688, -0.0635, -0.0617, -0.0511, -0.0423, -0.0423, -0.0441, -0.03, -0.0123, -0.0071, -0.0265, -0.0265, 0.0, -0.0018, 0.0106, 0.0141, 0.0088, -0.0018, -0.0018, -0.0159, -0.0141, -0.0053, 0.0212, 0.0335, 0.0229, 0.0071, 0.0144, 0.0385, 0.0422, 0.0348, 0.0274, 0.0237, 0.0052, 0.02, 0.0089, 0.0089, -0.004, -0.0003, 0.0071, 0.0015, 0.0108, 0.0108, -0.0299, -0.0225, -0.0059, -0.0077, 0.0052, -0.0114, -0.0022, 0.0403, 0.0403, 0.0385, 0.0643, 0.0662, 0.068, 0.0736, 0.0717, 0.0459, 0.0034, -0.0003, -0.0003, -0.0059, -0.0096, -0.041, -0.0428, -0.0447, -0.0096, -0.0207, -0.0059, -0.017, -0.0299, -0.0539, -0.0632, -0.0632, -0.0502, -0.0447, -0.0391, -0.0299, 0.0089, -0.0022, 0.0126, 0.02, 0.0108, 0.0089, 0.0329, 0.0385, 0.0385, 0.0588, 0.0477, 0.0532, 0.0532, 0.0532, 0.0717, 0.0588, 0.0865, 0.1087, 0.1087, 0.1401, 0.1309, 0.1013, 0.1013, 0.0884, 0.0496, 0.0569, 0.0773, 0.0662, 0.0957, 0.0754, 0.0699, 0.0847, 0.0847, 0.1438, 0.1235, 0.1382, 0.1309, 0.1105, 0.0994, 0.0625, 0.0699, 0.0828, 0.0606, 0.0773, 0.0773, 0.0643, 0.0532, 0.0329, 0.02, -0.0022, -0.0133, 0.0126, -0.0022, 0.0052, 0.0163, 0.0329, 0.0403, 0.0422, 0.0606, 0.0662, 0.081]}
{"unit_id": "quote:1928596252637454469", "ticker_idx": 0, "ts": "2025-05-30T23:35:52+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung is going all-in on HBM4. In fact, notable progress has already been confirmed in their HBM4 development.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Cautiously bullish Samsung: HBM3E volumes limited even if qualified, but HBM4 development showing notable progress and is the real bet.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Samsung HBM Update:\n\n1) Internally, there's strong confidence at Samsung that they will pass HBM3E qualification. However, even if they succeed, the expected shipment volume is likely to be limited.\n\n2) Instead, Samsung is going all-in on HBM4. In fact, notable progress has already been confirmed in their HBM4 development.", "reply_text": null, "parent_text": null, "quoted_text": "Samsung's HBM status update\n\n1) They are now guiding for $nvda HBM3e 12-hi qualification from late Q2 to late Q3. Clearly a delay from Q2\n\n2) Non-NVIDIA qualification is smoother than NVIDIA\n\n3) Have to guide down HBM revenue growth, unable to achieve 100% revenue growth in 2025", "parent_author": null, "ret_m1d": -0.0017794085151879857, "ret_signed_m1d": -0.0017794085151879857, "alpha_spy_m1d": -0.0028990128190186093, "alpha_c_m1d": -0.004897453607182745, "bench_spy_m1d": 0.0011196043038306236, "bench_c_m1d": 0.003118045091994759, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010676097430414533, "ret_signed_p1d": 0.010676097430414533, "alpha_spy_p1d": 0.005043196293600838, "alpha_c_p1d": 0.001841536648306663, "bench_spy_p1d": 0.005632901136813695, "bench_c_p1d": 0.00883456078210787, "ret_p1w": 0.05160136156545514, "ret_signed_p1w": 0.05160136156545514, "alpha_spy_p1w": 0.03505899587057515, "alpha_c_p1w": 0.019337765966406684, "bench_spy_p1w": 0.01654236569487999, "bench_c_p1w": 0.032263595599048456, "ret_p1m": 0.07058360525126739, "ret_signed_p1m": 0.07058360525126739, "alpha_spy_p1m": 0.019197444205683256, "alpha_c_p1m": -0.02789490145787621, "bench_spy_p1m": 0.05138616104558413, "bench_c_p1m": 0.0984785067091436, "ret_p3m": 0.2639330296755291, "ret_signed_p3m": 0.2639330296755291, "alpha_spy_p3m": 0.16357242889558732, "alpha_c_p3m": 0.11678365805628554, "bench_spy_p3m": 0.1003606007799418, "bench_c_p3m": 0.14714937161924357, "ret_p6m": 0.7389184659273367, "ret_signed_p6m": 0.7389184659273367, "alpha_spy_p6m": 0.5977957423592, "alpha_c_p6m": 0.5240138687896352, "bench_spy_p6m": 0.1411227235681367, "bench_c_p6m": 0.2149045971377015, "price_path": [-0.036, -0.0448, -0.0395, -0.0501, -0.0501, -0.0519, -0.0289, -0.0324, -0.0324, 0.0189, 0.0189, 0.0348, 0.0648, 0.0914, 0.0702, 0.0578, 0.0861, 0.0931, 0.0712, 0.0285, 0.0463, 0.0463, 0.0249, -0.0018, -0.0534, -0.048, -0.0569, 0.0036, -0.0178, 0.0, 0.0071, -0.0267, -0.0196, -0.016, -0.0142, -0.0214, -0.0089, -0.0089, -0.0089, -0.0071, -0.0071, -0.0125, -0.0125, -0.0338, -0.0338, -0.0338, -0.0285, -0.0285, -0.0249, 0.0249, 0.0125, 0.0214, 0.0196, 0.0107, -0.0071, -0.0053, -0.0089, -0.0267, -0.0356, -0.0267, -0.0409, -0.0053, -0.0018, 0.0, 0.0107, 0.0107, 0.0285, 0.0516, 0.0516, 0.0641, 0.0534, 0.0658, 0.0587, 0.0374, 0.0178, 0.0338, 0.0641, 0.0534, 0.0587, 0.032, 0.0765, 0.0907, 0.0712, 0.0885, 0.0706, 0.0777, 0.0885, 0.1422, 0.1332, 0.1046, 0.0992, 0.0813, 0.0921, 0.1207, 0.1189, 0.1404, 0.1583, 0.1941, 0.2013, 0.2138, 0.1816, 0.1887, 0.1816, 0.1798, 0.2604, 0.2639, 0.2997, 0.2783, 0.2335, 0.2478, 0.2514, 0.2317, 0.2621, 0.2854, 0.2711, 0.2729, 0.2872, 0.2818, 0.2818, 0.2532, 0.2532, 0.2621, 0.2639, 0.2783, 0.28, 0.2586, 0.2639, 0.246, 0.2478, 0.2102, 0.2371, 0.2496, 0.255, 0.2442, 0.255, 0.28, 0.2997, 0.3141, 0.3499, 0.3696, 0.4215, 0.4, 0.4376, 0.4376, 0.4949, 0.5164, 0.5289, 0.5414, 0.4913, 0.5141, 0.5087, 0.5465, 0.6139, 0.6139, 0.6139, 0.6139, 0.6139, 0.6139, 0.6976, 0.6778, 0.6472, 0.7083, 0.7569, 0.7605, 0.7641, 0.7533, 0.7731, 0.7353, 0.7767, 0.8342, 0.7893, 0.8073, 0.872, 0.9331, 0.9979, 0.8864, 0.8091, 0.7839, 0.7605, 0.8091, 0.8612, 0.854, 0.8486, 0.7479, 0.8091, 0.7587, 0.7353, 0.8091, 0.7048, 0.7389]}
{"unit_id": "orig:1924467607379546278", "ticker_idx": 0, "ts": "2025-05-19T14:10:07+00:00", "symbol": "Eoptolink", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks: Eoptolink, Innolight", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's top picks Eoptolink and Innolight on 800G optical module demand surge to 37M units in 2026 and sector P/E re-rating toward 15–20x.", "resolved_tickers": ["300502.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Eoptolink Technology 300502.SZ China", "bench_c": "MCHI", "bench_c_reason": "country_fallback (suffix='SZ')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Modules: Strong 800G Demand Expected in 2026 – 37 Million Units Globally\nCiti Research (L. Tsang, May 16, 2025)\n\n• Global 800G Optical Module Shipments to Reach 37 Million Units in 2026 (+85% YoY)\nThis surge is driven by a higher number of optical modules per ASIC and the transition of overseas cloud infrastructure from 100G/200G/400G to 800G. However, the transition to 1.6T Ethernet switches has been delayed due to capacity conversion and extended customer readiness timelines, pushing out 1.6T module demand to 2027. Cloud service providers (CSPs) are increasingly open to second-tier suppliers, and production capacity and delivery capability will determine market share.\n\n⸻\n\nKey Takeaways\n• Explosive 800G Demand\n• Industry demand for 800G modules has been revised upward from 20M/23M to 20M in 2025 and 37M in 2026\n• This is driven by the adoption of mainstream 800G switches such as Tomahawk 5 and a high attach rate\n• CSPs are still dominated by large overseas players, but second-tier vendors are getting qualified at a faster pace\n• 1.6T Transition Delayed\n• Due to capacity yield issues and customer migration preparation, 1.6T demand forecasts have been revised down from 2.3M/12M to 2M/5M for 2025/26\n• The actual ramp-up is expected to begin in 2H26 or 4Q26, with more meaningful demand shifting to 2027\n• Middle East AI Projects May Catalyze 1.6T Demand\n• Aside from the increase in rack yield rates for GB200 deployments in 2025,\nrising AI infrastructure investments by Middle Eastern nations could drive incremental 1.6T demand starting in 2026\n• Sector Valuation Re-Rating Underway\n• As concerns over peak AI capex ease, and with improving sentiment around US-China tariff negotiations and demand outlook,\nthe optical module sector has rebounded from FY25 P/E multiples of 8–10x and is expected to re-rate toward the 15–20x range\n\n• Top Picks: Eoptolink, Innolight\n\n⸻\n\nMajor CSPs and Their Optical Module Suppliers\n\n• Amazon: Innolight, Eoptolink, AAOI, Fabrinet (ramping)\n\n• Google: Innolight, Lumentum, Coherent, TBD (possibly Eoptolink)\n\n• Meta: Innolight, Coherent, Eoptolink, Celestica (ramping)\n\n• Microsoft: Innolight, Coherent, AAOI, Lumentum (ramping)\n\n• NVIDIA: Mellanox (TFC), Coherent, Eoptolink\n\n• Oracle: Innolight, Coherent, AAOI (ramping), Lumentum (ramping)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.022590147966634166, "ret_signed_m1d": 0.022590147966634166, "alpha_spy_m1d": 0.023682870717476767, "alpha_c_m1d": 0.027709022702447683, "bench_spy_m1d": -0.0010927227508426007, "bench_c_m1d": -0.005118874735813517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004742203810832235, "ret_signed_p1d": 0.004742203810832235, "alpha_spy_p1d": 0.008104331860627756, "alpha_c_p1d": -0.004398557367017331, "bench_spy_p1d": -0.003362128049795521, "bench_c_p1d": 0.009140761177849566, "ret_p1w": -0.0018968445379911048, "ret_signed_p1w": -0.0018968445379911048, "alpha_spy_p1w": 0.024563568158141647, "alpha_c_p1w": -0.003359329256398147, "bench_spy_p1w": -0.02646041269613275, "bench_c_p1w": 0.0014624847184070422, "ret_p1m": 0.2982606340587064, "ret_signed_p1m": 0.2982606340587064, "alpha_spy_p1m": 0.2937551375459957, "alpha_c_p1m": 0.2931777600434775, "bench_spy_p1m": 0.004505496512710705, "bench_c_p1m": 0.005082874015228889, "ret_p3m": 1.800374264792433, "ret_signed_p3m": 1.800374264792433, "alpha_spy_p3m": 1.7129460665053373, "alpha_c_p3m": 1.7055650482942937, "bench_spy_p3m": 0.0874281982870957, "bench_c_p3m": 0.09480921649813934, "ret_p6m": 2.951017138268723, "ret_signed_p6m": 2.951017138268723, "alpha_spy_p6m": 2.7962414551846813, "alpha_c_p6m": 2.7596505673116947, "bench_spy_p6m": 0.15477568308404166, "bench_c_p6m": 0.19136657095702825, "price_path": [0.0466, 0.0278, 0.0436, -0.0214, -0.0455, -0.0257, -0.0829, -0.1719, -0.2085, -0.1968, -0.187, -0.178, -0.1993, -0.1861, -0.1869, -0.1907, -0.2022, -0.191, -0.1752, -0.1059, -0.1539, -0.1512, -0.1803, -0.143, -0.1474, -0.1568, -0.1793, -0.1794, -0.154, -0.1621, -0.1418, -0.2, -0.2, -0.3601, -0.3933, -0.3929, -0.3381, -0.3297, -0.3195, -0.3226, -0.3526, -0.3506, -0.3287, -0.298, -0.2999, -0.2125, -0.221, -0.1795, -0.2136, -0.221, -0.2258, -0.2258, -0.2258, -0.2258, -0.1927, -0.1878, -0.0692, -0.0584, -0.0271, -0.0009, 0.0247, -0.0008, 0.0226, 0.0, 0.0047, 0.0067, 0.015, -0.0004, -0.0019, -0.0332, 0.0256, 0.0564, 0.0698, 0.0698, 0.0467, 0.1217, 0.1744, 0.2087, 0.2225, 0.2071, 0.1921, 0.2218, 0.2126, 0.2932, 0.2983, 0.3601, 0.3695, 0.3333, 0.3002, 0.2933, 0.338, 0.4156, 0.4929, 0.5394, 0.5416, 0.5147, 0.5714, 0.5556, 0.533, 0.6359, 0.6428, 0.6125, 0.5896, 0.5865, 0.9038, 1.0579, 1.224, 1.2097, 1.1605, 1.1417, 1.1112, 1.0943, 1.0963, 1.1549, 1.3438, 1.307, 1.2932, 1.2194, 1.2361, 1.2112, 1.2591, 1.2254, 1.2092, 1.3378, 1.4833, 1.867, 1.8004, 1.8263, 1.9607, 2.1122, 2.0922, 2.1252, 2.3185, 2.5889, 2.4182, 2.7369, 3.3019, 3.317, 3.7085, 3.3413, 3.4808, 2.7826, 3.2328, 2.8116, 2.8216, 3.0595, 3.6043, 3.3386, 3.2455, 3.1904, 3.2135, 3.1875, 3.2419, 3.1571, 3.2782, 3.2981, 3.557, 3.5328, 3.7058, 3.433, 3.433, 3.433, 3.433, 3.433, 3.433, 3.433, 3.2558, 3.1958, 3.2249, 2.8359, 2.9097, 2.936, 2.8336, 2.9903, 3.4288, 3.3974, 3.2147, 3.5203, 3.8957, 3.9766, 3.9218, 3.5328, 3.1729, 3.2026, 3.2181, 3.1946, 3.3232, 3.2393, 3.0722, 2.951]}
{"unit_id": "orig:1930396244050137098", "ticker_idx": 0, "ts": "2025-06-04T22:48:24+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We believe the upside potential in Neocloud, enterprise, and sovereign AI demand remains underappreciated by the market. From 2H25 onwards, an increase in market share from a new U.S. CSP is expected to further lift Wistron's performance.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares CLSA Buy-equivalent thesis: Wistron's AI server ramp underappreciated; 100K incremental Blackwell GPUs from 4Q25 also bullish NVDA.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "250605_Wistron: GB Production Ramping Up for Real – CLSA\n\n1. Wistron’s May revenue rose +56% MoM and +162% YoY.\nQuarter-to-date (QTD) sales are significantly outperforming market expectations.\n■ As of May, Wistron has achieved approx. 78% of our 2Q25 forecast and 86% of market consensus.\n2. Management attributed the revenue surge to strong AI demand, and expects another sequential increase in June.\n3. These results reaffirm our prior view from the “Golden Oil on Troubled Waters” report, which highlighted a strong Blackwell Rack production ramp-up.\n4. We believe the upside potential in Neocloud, enterprise, and sovereign AI demand remains underappreciated by the market.\n5. From 2H25 onwards, an increase in market share from a new U.S. CSP is expected to further lift Wistron’s performance.\n6. Meanwhile, Dell recently disclosed a record-high AI backlog of $14.4bn as of 1Q25, and stated that its AI pipeline over the next five quarters is several times larger than its current backlog.\n7. Multiple sovereign AI project wins were referenced across the U.S., U.K., France, Germany, and Japan, alongside a large-scale recent project in the Middle East, all of which are positive for Wistron.\n8. CoreWeave recently announced lease agreements for two new AI data centers, triggering a sharp rally in its share price.\n9. Based on this, we now expect up to 100K additional Blackwell GPU demand to materialize from 4Q25 through mid-2026.\n10.xAI is also building a new cluster in Memphis with 1,000K GPUs, separate from its existing 200K-GPU cluster.\n11. The project is now expected to be completed by late 2025 to early 2026, much earlier than the previously assumed early 2027 timeline.\n12. As a result, Wistron is positioned to structurally benefit from the accelerating infrastructure build-out by players like Dell, CoreWeave, and xAI.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03947368421052633, "ret_signed_m1d": -0.03947368421052633, "alpha_spy_m1d": -0.03974230375057264, "alpha_c_m1d": -0.03799553105366582, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.0014781531568605066, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00877192982456143, "ret_signed_p1d": 0.00877192982456143, "alpha_spy_p1d": 0.013604698880750665, "alpha_c_p1d": 0.011601411388053373, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0028294815634919424, "ret_p1w": 0.00877192982456143, "ret_signed_p1w": 0.00877192982456143, "alpha_spy_p1w": -0.0003400013375636224, "alpha_c_p1w": -0.006177254083769013, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.014949183908330443, "ret_p1m": 0.06140350877192979, "ret_signed_p1m": 0.06140350877192979, "alpha_spy_p1m": 0.008949768205986919, "alpha_c_p1m": -0.02553002948566019, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.08693353825758998, "ret_p3m": -0.03947368421052633, "ret_signed_p3m": -0.03947368421052633, "alpha_spy_p3m": -0.12509950351048293, "alpha_c_p3m": -0.1496298905679394, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.11015620635741308, "ret_p6m": 0.2763157894736843, "ret_signed_p6m": 0.2763157894736843, "alpha_spy_p6m": 0.12923586930745312, "alpha_c_p6m": 0.07439173423382606, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.20192405523985824, "price_path": [-0.0924, -0.1009, -0.1136, -0.1136, -0.1179, -0.1136, -0.1009, -0.0882, -0.1263, -0.1094, -0.1051, -0.1221, -0.1221, -0.1051, -0.1136, -0.1475, -0.1917, -0.1391, -0.1433, -0.1433, -0.1433, -0.229, -0.3053, -0.3664, -0.3036, -0.2341, -0.1747, -0.1306, -0.1543, -0.1475, -0.1094, -0.1263, -0.162, -0.1306, -0.1433, -0.1306, -0.1306, -0.1179, -0.1391, -0.1391, -0.1051, -0.1221, -0.0754, -0.0967, -0.0924, -0.0924, -0.0797, -0.0797, -0.0288, -0.0542, -0.0585, -0.0754, -0.0712, -0.0415, -0.0415, -0.0415, -0.0161, -0.0246, -0.0203, -0.0118, -0.0118, -0.0246, -0.0395, 0.0, 0.0088, -0.0351, -0.0044, 0.0044, 0.0088, 0.0219, 0.0395, 0.0439, 0.0351, 0.0175, 0.0132, 0.0395, 0.0395, 0.0526, 0.1009, 0.0965, 0.0614, 0.0746, 0.0526, 0.0526, 0.0614, 0.0439, 0.0351, 0.0395, 0.057, 0.0526, 0.0482, 0.0263, 0.0526, 0.0351, 0.0351, 0.0439, 0.0395, 0.0044, 0.0175, 0.0263, 0.0219, 0.0526, 0.057, 0.057, 0.0789, 0.0789, 0.0482, 0.0877, 0.0746, 0.0833, 0.0921, 0.1184, 0.1053, 0.0307, 0.0219, 0.0219, 0.0175, 0.0132, -0.0044, -0.0219, -0.0175, -0.0044, -0.0088, 0.0088, 0.0088, -0.0088, -0.0395, -0.0351, -0.0351, -0.0088, 0.0132, 0.0044, 0.0175, 0.0526, 0.0439, 0.0526, 0.0482, 0.0351, 0.0439, 0.0526, 0.0614, 0.0658, 0.0746, 0.0658, 0.1009, 0.1228, 0.1228, 0.2325, 0.2807, 0.3114, 0.3465, 0.3465, 0.3684, 0.3202, 0.3246, 0.3246, 0.2763, 0.2105, 0.2105, 0.2237, 0.1974, 0.2456, 0.25, 0.2588, 0.2544, 0.2544, 0.2851, 0.2982, 0.3377, 0.3596, 0.3202, 0.2588, 0.2193, 0.2061, 0.2456, 0.2061, 0.2149, 0.2675, 0.2193, 0.2632, 0.2325, 0.2105, 0.2018, 0.2018, 0.2544, 0.2193, 0.2193, 0.2237, 0.25, 0.2763]}
{"unit_id": "orig:1930396244050137098", "ticker_idx": 1, "ts": "2025-06-04T22:48:24+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we now expect up to 100K additional Blackwell GPU demand to materialize from 4Q25 through mid-2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares CLSA Buy-equivalent thesis: Wistron's AI server ramp underappreciated; 100K incremental Blackwell GPUs from 4Q25 also bullish NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250605_Wistron: GB Production Ramping Up for Real – CLSA\n\n1. Wistron’s May revenue rose +56% MoM and +162% YoY.\nQuarter-to-date (QTD) sales are significantly outperforming market expectations.\n■ As of May, Wistron has achieved approx. 78% of our 2Q25 forecast and 86% of market consensus.\n2. Management attributed the revenue surge to strong AI demand, and expects another sequential increase in June.\n3. These results reaffirm our prior view from the “Golden Oil on Troubled Waters” report, which highlighted a strong Blackwell Rack production ramp-up.\n4. We believe the upside potential in Neocloud, enterprise, and sovereign AI demand remains underappreciated by the market.\n5. From 2H25 onwards, an increase in market share from a new U.S. CSP is expected to further lift Wistron’s performance.\n6. Meanwhile, Dell recently disclosed a record-high AI backlog of $14.4bn as of 1Q25, and stated that its AI pipeline over the next five quarters is several times larger than its current backlog.\n7. Multiple sovereign AI project wins were referenced across the U.S., U.K., France, Germany, and Japan, alongside a large-scale recent project in the Middle East, all of which are positive for Wistron.\n8. CoreWeave recently announced lease agreements for two new AI data centers, triggering a sharp rally in its share price.\n9. Based on this, we now expect up to 100K additional Blackwell GPU demand to materialize from 4Q25 through mid-2026.\n10.xAI is also building a new cluster in Memphis with 1,000K GPUs, separate from its existing 200K-GPU cluster.\n11. The project is now expected to be completed by late 2025 to early 2026, much earlier than the previously assumed early 2027 timeline.\n12. As a result, Wistron is positioned to structurally benefit from the accelerating infrastructure build-out by players like Dell, CoreWeave, and xAI.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004932421576486434, "ret_signed_m1d": -0.004932421576486434, "alpha_spy_m1d": -0.005201041116532745, "alpha_c_m1d": 0.006825285213173493, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013599137425771679, "ret_signed_p1d": -0.013599137425771679, "alpha_spy_p1d": -0.008766368369582445, "alpha_c_p1d": -0.011811632812611439, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.006481889145962372, "ret_signed_p1w": 0.006481889145962372, "alpha_spy_p1w": -0.0026300420161626814, "alpha_c_p1w": -0.03375642048266836, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.12282315830271795, "ret_signed_p1m": 0.12282315830271795, "alpha_spy_p1m": 0.07036941773677508, "alpha_c_p1m": -0.003770715177219053, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.22739630300135127, "ret_signed_p3m": 0.22739630300135127, "alpha_spy_p3m": 0.14177048370139467, "alpha_c_p3m": 0.07430784158682724, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.2703119733794417, "ret_signed_p6m": 0.2703119733794417, "alpha_spy_p6m": 0.12323205321321051, "alpha_c_p6m": -0.11085976629422323, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.206, -0.2463, -0.2337, -0.1845, -0.1856, -0.1427, -0.1578, -0.1867, -0.1719, -0.1648, -0.1707, -0.1445, -0.1496, -0.1984, -0.2148, -0.2272, -0.2363, -0.2239, -0.222, -0.2827, -0.3355, -0.312, -0.3214, -0.1944, -0.242, -0.2184, -0.2199, -0.2094, -0.2637, -0.2849, -0.2849, -0.3172, -0.3032, -0.2763, -0.2501, -0.2178, -0.2339, -0.2318, -0.2325, -0.2136, -0.1932, -0.198, -0.2, -0.1752, -0.173, -0.1781, -0.1333, -0.0845, -0.0464, -0.05, -0.0459, -0.0447, -0.0531, -0.0713, -0.0641, -0.0749, -0.0749, -0.0452, -0.0501, -0.0192, -0.0478, -0.032, -0.0049, 0.0, -0.0136, -0.0014, 0.005, 0.0144, 0.0065, 0.0218, 0.0004, 0.0196, 0.0156, 0.0252, 0.0252, 0.0137, 0.0159, 0.0422, 0.0874, 0.0924, 0.1116, 0.1133, 0.0803, 0.1081, 0.1228, 0.1228, 0.1151, 0.1275, 0.1478, 0.1564, 0.1621, 0.1562, 0.2029, 0.2076, 0.2191, 0.2149, 0.2077, 0.177, 0.2034, 0.2243, 0.2226, 0.2455, 0.2368, 0.2633, 0.2534, 0.2242, 0.2684, 0.2561, 0.2643, 0.2738, 0.2874, 0.2829, 0.2907, 0.2796, 0.2826, 0.2716, 0.2826, 0.2377, 0.236, 0.233, 0.2542, 0.2671, 0.2809, 0.2797, 0.2696, 0.2274, 0.2274, 0.2034, 0.2023, 0.2096, 0.1769, 0.186, 0.2033, 0.2496, 0.2485, 0.2531, 0.2526, 0.2324, 0.2001, 0.242, 0.245, 0.2939, 0.2574, 0.2471, 0.2522, 0.2557, 0.2815, 0.3148, 0.3195, 0.3311, 0.3222, 0.3075, 0.304, 0.3327, 0.3571, 0.2907, 0.3271, 0.2687, 0.2673, 0.2812, 0.2912, 0.2871, 0.2767, 0.2705, 0.2837, 0.3126, 0.3495, 0.4167, 0.459, 0.4298, 0.427, 0.4579, 0.4002, 0.3757, 0.3254, 0.3259, 0.4027, 0.3612, 0.3657, 0.3168, 0.3401, 0.315, 0.2781, 0.3144, 0.273, 0.2606, 0.2864, 0.2531, 0.2703, 0.2703]}
{"unit_id": "orig:1938032856535863588", "ticker_idx": 0, "ts": "2025-06-26T00:33:34+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "At the peak of HBM growth, another profit cycle for generic DRAM leverage has begun.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bullish MU: beat on revenue/profit/guidance, HBM share accelerating to 23% by CY3Q, generic DRAM supply shortage + healthy inventory signals new profit cycle.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron Records All-Time High DRAM Revenue, Three Key Takeaways\n\n[Samsung Securities Semiconductor, IT/Jongwook Lee]\n\nMicron's FY3Q25 revenue and operating profit exceeded market consensus (Bloomberg) by 5% and 17%, respectively. The revenue guidance for the next quarter also surpassed consensus by 9%. We believe the following three points are crucial to note from Micron's record-breaking revenue.\n\nHBM Market Share Growth Rate\n\nMicron mentioned that the period for its HBM market share to approach its DRAM market share of 23% has been pulled forward from year-end (CY4Q) to CY3Q. This indicates the fastest growth rate in the HBM market, attributed to its successful penetration into the NVIDIA and Broadcom markets with HBM3E 12-high products.\n\nChronic Generic DRAM Supply Shortage\n\nDRAM capacity expansion is primarily for HBM. Micron also noted that the per-unit capacity cannibalization effect of HBM is steadily increasing. In the future, there is a high probability that the company will reduce market share competition in generic DRAM and adopt aggressive pricing policies.\n\nHealthy Customer Inventory\n\nThe pre-order effect due to tariffs is negligible, and customer inventory levels are healthy.\nThis case directly demonstrates that HBM demand growth has begun to be accompanied by profits in generic DRAM. While generic DRAM customers differ from HBM customers, the industry conditions are shared, and profits are being leveraged by HBM. At the peak of HBM growth, another profit cycle for generic DRAM leverage has begun.\n\nThank you.\n\n(Public Disclosure Date: 2025/06/26)\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00992064949241711, "ret_signed_m1d": 0.00992064949241711, "alpha_spy_m1d": 0.01768369302316386, "alpha_c_m1d": 0.017135688419591344, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009841259532637148, "ret_signed_p1d": -0.009841259532637148, "alpha_spy_p1d": -0.014809591296642388, "alpha_c_p1d": -0.014242418921685207, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.02944444933381929, "ret_signed_p1w": -0.02944444933381929, "alpha_spy_p1w": -0.05145900384583868, "alpha_c_p1w": -0.05260462158565393, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": -0.11615291307436704, "ret_signed_p1m": -0.11615291307436704, "alpha_spy_p1m": -0.15738720973371012, "alpha_c_p1m": -0.15327403601108536, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.32195742379580516, "ret_signed_p3m": 0.32195742379580516, "alpha_spy_p3m": 0.23504551943629215, "alpha_c_p3m": 0.161423497883884, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 1.1137855991590007, "ret_signed_p6m": 1.1137855991590007, "alpha_spy_p6m": 0.9950943986632539, "alpha_c_p6m": 0.8286846054001149, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3104, -0.296, -0.2968, -0.41, -0.4863, -0.4574, -0.4798, -0.382, -0.444, -0.448, -0.4363, -0.4362, -0.4498, -0.454, -0.454, -0.4703, -0.4428, -0.4212, -0.3856, -0.3668, -0.3765, -0.3898, -0.3893, -0.3828, -0.3594, -0.3617, -0.361, -0.3443, -0.3242, -0.3186, -0.2675, -0.2307, -0.2435, -0.2425, -0.2222, -0.2171, -0.2214, -0.2394, -0.2474, -0.259, -0.259, -0.2351, -0.2367, -0.2317, -0.2503, -0.2208, -0.1885, -0.1806, -0.1564, -0.1384, -0.1194, -0.0941, -0.0791, -0.0779, -0.0825, -0.0489, -0.0449, -0.0332, -0.0332, -0.019, -0.0311, 0.0152, 0.0099, 0.0, -0.0098, -0.0218, -0.0406, -0.0338, -0.0294, -0.0294, -0.0474, -0.0116, -0.0289, -0.022, -0.0107, -0.0578, -0.0458, -0.0751, -0.1003, -0.0913, -0.1005, -0.1324, -0.1275, -0.1124, -0.1162, -0.1162, -0.1106, -0.0885, -0.133, -0.1668, -0.1439, -0.1336, -0.1359, -0.1113, -0.0555, -0.0172, 0.0148, -0.0128, -0.0047, -0.0398, -0.0185, -0.0304, -0.0689, -0.0802, -0.0652, -0.0752, -0.0745, -0.0646, -0.0308, -0.0546, -0.0546, -0.0588, -0.0569, -0.0133, 0.0436, 0.0443, 0.0743, 0.1122, 0.1961, 0.249, 0.2533, 0.2617, 0.271, 0.3417, 0.2927, 0.3077, 0.322, 0.2846, 0.2459, 0.2493, 0.302, 0.3292, 0.447, 0.4597, 0.4931, 0.5179, 0.476, 0.5623, 0.5288, 0.4435, 0.5323, 0.4869, 0.5257, 0.6099, 0.6087, 0.6436, 0.608, 0.5776, 0.6431, 0.741, 0.7496, 0.764, 0.8015, 0.7806, 0.7787, 0.8656, 0.7331, 0.8879, 0.8945, 0.8912, 1.0135, 0.9166, 0.9467, 0.8835, 0.962, 0.9232, 0.8163, 0.7958, 0.6007, 0.6484, 0.78, 0.7848, 0.8303, 0.8303, 0.8798, 0.9114, 0.9037, 0.8613, 0.8016, 0.8857, 0.9628, 1.0065, 1.0962, 1.0545, 0.9168, 0.8879, 0.8482, 0.7926, 0.9757, 1.1138]}
{"unit_id": "thread:1922799707539341701", "ticker_idx": 0, "ts": "2025-05-14T23:42:28+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Strong DRAM and NAND demand: Recovery is progressing as expected across both data center and consumer markets. Pricing momentum sustained: DRAM prices are on the rise. HBM3E 12-High: Production ramping up. Strong and sustained demand for AI and high-performance computing; no signs of CapEx decline.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: DRAM/NAND pricing momentum, HBM3E ramp, improving yields, AI demand strength, and margin recovery expected ahead.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Micron at JPMorgan Conference – May 14, 2025\n\n1. Earnings & Demand Outlook\n\n• Strong DRAM and NAND demand: Recovery is progressing as expected across both data center and consumer markets (PCs, smartphones).\n\n• Pricing momentum sustained: DRAM prices are on the rise, while the decline in NAND prices is slowing. Micron sent letters in March indicating intent to raise prices.\n\n• May quarter progress: No update to guidance, but overall quarterly trend is positive; price increases are being well executed.\n\n• Q4 (August quarter) margin: Margin is currently lower due to a higher share of consumer products, but margin improvement is expected moving forward.\n\n2. HBM (High Bandwidth Memory)\n\n• HBM3E 12-High: Production ramping up. From the August quarter, 12-High is expected to surpass 8-High in shipment volume.\n\n• Yield improving rapidly, with stable yield expected by the end of Q3.\n\n• Continued strong demand for HBM, with positive feedback from key cloud customers.\n\n• Already engaged in 2026 HBM supply discussions; early market entry helping secure customer trust.\n\n3. Trade & Tariff Impact\n\n• While some front-loaded demand due to tariff expectations may exist, the overall demand recovery is driven by cyclical recovery.\n\n• DRAM and NAND are exempt from U.S. tariffs; no current price pass-through to U.S. customers.\n\n• Growing customer interest in U.S. production investments in Boise, NY, and Virginia; ESG factors (green energy, etc.) also viewed positively.\n\n4. AI & Data Center Demand\n\n• Strong and sustained demand for AI and high-performance computing; no signs of CapEx decline.\n\n• In addition to HBM, demand is increasing for high-capacity DRAM, low-power DRAM, and SSDs.\n\n• HBM4 scheduled for release in 2026; customized platform roadmaps being discussed with customers.\n\n5. NAND & SSD Strategy\n\n• Data center SSD share expansion: 7% in 2022 → 12% in 2023 → projected 15% in 2024.\n\n• Preparing to launch next-generation Gen9-based SSDs, optimized for high-performance markets.\n\n6. Technology Transition & Manufacturing Strategy\n\n• DRAM: Mass production of One Gamma (1γ) node, the first to adopt EUV, has begun; 30% bit density improvement and stable yield progress.\n\n• NAND: Ramping up Gen9 technology, achieving performance optimized for high-end SSDs.\n\n7. Competition & Market Landscape\n\n• Chinese players (DRAM/NAND): Primarily focused on legacy markets like DDR4. Some are attempting DDR5, but Micron remains focused on the premium segment.\n\n• On Samsung’s HBM entry: Emphasized that delayed entry and shorter platform lifecycles at customers maintain differentiation opportunities.\n\n8. Organizational Restructure\n\n• Transition to four business units:\n\n• High-Performance Computing (HBM, high-capacity DRAM, etc.)\n\n• Data Center (traditional server/enterprise)\n\n• Client (PCs, smartphones)\n\n• Automotive & Embedded\n\n• From Q4 onward, Micron will disclose revenue, gross margin, and operating margin by business unit — aimed at enhancing transparency and responsiveness to market conditions.\n\n---\n\nSource:", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016890391890211465, "ret_signed_m1d": 0.016890391890211465, "alpha_spy_m1d": 0.018166820068248657, "alpha_c_m1d": 0.026689205717686404, "bench_spy_m1d": -0.0012764281780371922, "bench_c_m1d": -0.00979881382747494, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0013637731216828985, "ret_signed_p1d": 0.0013637731216828985, "alpha_spy_p1d": -0.003520463699503429, "alpha_c_p1d": 0.004831646777415877, "bench_spy_p1d": 0.004884236821186327, "bench_c_p1d": -0.003467873655732978, "ret_p1w": 0.005455172718119128, "ret_signed_p1w": 0.005455172718119128, "alpha_spy_p1w": 0.013505001851780718, "alpha_c_p1w": 0.031746538418808146, "bench_spy_p1w": -0.00804982913366159, "bench_c_p1w": -0.026291365700689018, "ret_p1m": 0.21884169946124632, "ret_signed_p1m": 0.21884169946124632, "alpha_spy_p1m": 0.1913397100656129, "alpha_c_p1m": 0.15771025769274316, "bench_spy_p1m": 0.02750198939563342, "bench_c_p1m": 0.06113144176850316, "ret_p3m": 0.2991653528776992, "ret_signed_p3m": 0.2991653528776992, "alpha_spy_p3m": 0.21371484801142837, "alpha_c_p3m": 0.1160532174122566, "bench_spy_p3m": 0.08545050486627082, "bench_c_p3m": 0.18311213546544258, "ret_p6m": 1.5042353347052901, "ret_signed_p6m": 1.5042353347052901, "alpha_spy_p6m": 1.3569124718163073, "alpha_c_p6m": 1.0884927338099237, "bench_spy_p6m": 0.1473228628889829, "bench_c_p6m": 0.4157426008953664, "price_path": [0.0427, 0.0427, 0.1189, 0.0934, 0.0811, 0.0356, -0.0004, -0.0233, 0.0237, -0.038, -0.019, -0.0514, -0.0445, -0.0116, -0.0647, -0.026, -0.0876, -0.067, 0.0021, -0.0059, 0.056, 0.0803, 0.0658, 0.0693, 0.0792, -0.0076, 0.0157, -0.0132, -0.0347, -0.0449, -0.0734, -0.0884, -0.0693, -0.0705, -0.2201, -0.321, -0.2827, -0.3124, -0.1831, -0.2651, -0.2704, -0.2549, -0.2547, -0.2727, -0.2782, -0.2782, -0.2998, -0.2634, -0.2349, -0.1878, -0.163, -0.1758, -0.1935, -0.1927, -0.1841, -0.1532, -0.1563, -0.1554, -0.1332, -0.1067, -0.0992, -0.0318, 0.0169, 0.0, 0.0014, 0.0281, 0.0349, 0.0292, 0.0055, -0.0051, -0.0205, -0.0205, 0.0111, 0.009, 0.0155, -0.009, 0.03, 0.0727, 0.0832, 0.1151, 0.1389, 0.164, 0.1974, 0.2173, 0.2188, 0.2128, 0.2572, 0.2625, 0.278, 0.278, 0.2967, 0.2807, 0.3419, 0.335, 0.3219, 0.3089, 0.293, 0.2683, 0.2772, 0.2829, 0.2829, 0.2593, 0.3065, 0.2836, 0.2928, 0.3077, 0.2455, 0.2613, 0.2226, 0.1893, 0.2012, 0.189, 0.1469, 0.1533, 0.1733, 0.1683, 0.1682, 0.1757, 0.2049, 0.1461, 0.1013, 0.1317, 0.1452, 0.1423, 0.1747, 0.2484, 0.2992, 0.3415, 0.3049, 0.3157, 0.2692, 0.2974, 0.2816, 0.2308, 0.2159, 0.2357, 0.2225, 0.2233, 0.2365, 0.2811, 0.2497, 0.2497, 0.2441, 0.2467, 0.3043, 0.3795, 0.3804, 0.4201, 0.4701, 0.5811, 0.651, 0.6567, 0.6677, 0.68, 0.7735, 0.7088, 0.7287, 0.7474, 0.6981, 0.6468, 0.6515, 0.7211, 0.757, 0.9127, 0.9295, 0.9736, 1.0065, 0.9511, 1.0651, 1.0209, 0.9081, 1.0255, 0.9655, 1.0168, 1.1281, 1.1265, 1.1726, 1.1255, 1.0854, 1.172, 1.3013, 1.3127, 1.3317, 1.3813, 1.3538, 1.3512, 1.4661, 1.2909, 1.4955, 1.5042]}
{"unit_id": "thread:1937305935137136852", "ticker_idx": 0, "ts": "2025-06-24T00:25:02+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia is less optimistic about its business performance beyond 2026... Nvidia plans to reduce wafer orders for 2026... Nvidia's absolute lead in the cloud AI market may be wavering", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish NVDA on wavering dominance, reduced wafer orders post-2026; bullish AMD on strong AI chips and rising procurement by Microsoft/OpenAI.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: UALink camp's first chip aims for tape-out, poised to challenge Nvidia\n\nFollowing Nvidia's announcement of NVLink Fusion, the competition between NVLink and UALink has resurfaced.\n\nSince Nvidia already has a relatively mature product line and ecosystem, it has attracted many customers to develop related chips. This helps Nvidia avoid losing its voice in the ASIC domain and ensures it remains competitive against UALink in the future.\n\nHowever, UALink's overall development pace currently lags behind NVLink, which already has actual products on the market. UALink only officially launched its 1.0 specification several months ago. Still, rumors suggest that UALink's first product could tape out as early as the end of this year, with more related products expected to join the lineup in 2026.\n\nOverall, UALink's product launch schedule is not too far behind that of NVLink. According to industry sources discussing ongoing projects, although NVLink has attracted some customers to start investing, the UALink camp reportedly has \"dozens of projects\" under R&D. Over time, UALink stands a strong chance to leapfrog and challenge Nvidia's dominance.\n\nInsiders note that despite UALink's slower progress, most customers still prefer open standard specifications.\n\nLooking further ahead, it is quite possible that UALink will surpass NVLink in market share. Moreover, market whispers are suggesting that while NVLink Fusion claims to offer an open ecosystem for developing customized products, the actual openness falls short of what industry players initially expected.\n\nCustomers willing to adopt NVLink-based development do so partly as a technical experiment. Currently, Nvidia remains the mainstream brand for AI general-purpose computing, so launching products compatible with the market majority is advantageous. But in the long run, if the level of openness remains limited, expanding UALink's ecosystem becomes an inevitable trend.\n\nAdditionally, industry experts mention that AMD's latest AI chips and integrated rack solutions have performed impressively. This marks a genuine opportunity to challenge Nvidia's leadership. Major companies like Microsoft and OpenAI are likely to increase their procurement of AMD solutions next year.\n\nThis means that starting next year, the UALink specification supported by AMD's compute cores could gradually expand its market influence. With such a foundation, subsequent ASIC product development and market acceptance are expected to grow steadily. The growth potential and ceiling for UALink clearly exceed those of NVLink.\n\nRumors have emerged indicating that Nvidia is less optimistic about its business performance beyond 2026.\n\nThis can be inferred from CEO Jensen Huang's recent global speeches promoting national-level orders, as well as Nvidia's reactions to potential further bans on H20 exports to China.\n\nThere are even reports that Nvidia plans to reduce wafer orders for 2026. Meanwhile, ASIC vendors have recently revised upward their forecasts for the market size going forward. These shifts signal that Nvidia's absolute lead in the cloud AI market may be wavering, presenting a prime opportunity for UALink to expand its influence.\n\nMany industry observers welcome these changes optimistically, believing that increased competition will foster better technological products and accelerate AI development.\n\n---\n\nhttps://t.co/53W9THmNdC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025219676936607094, "ret_signed_m1d": 0.025219676936607094, "alpha_spy_m1d": 0.014293175290764237, "alpha_c_m1d": -0.01020784674882036, "bench_spy_m1d": -0.010926501645842857, "bench_c_m1d": -0.03542752368542745, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0433402059174941, "ret_signed_p1d": -0.0433402059174941, "alpha_spy_p1d": -0.04277984128688983, "alpha_c_p1d": -0.029861512825032577, "bench_spy_p1d": 0.0005603646306042709, "bench_c_p1d": 0.013478693092461524, "ret_p1w": 0.03651130416442272, "ret_signed_p1w": -0.03651130416442272, "alpha_spy_p1w": -0.018596982536833284, "alpha_c_p1w": -0.020675736576599046, "bench_spy_p1w": 0.017914321627589436, "bench_c_p1w": 0.015835567587823673, "ret_p1m": 0.15469918316305287, "ret_signed_p1m": -0.15469918316305287, "alpha_spy_p1m": -0.10949325778386076, "alpha_c_p1m": -0.09920091040279955, "bench_spy_p1m": 0.0452059253791921, "bench_c_p1m": 0.05549827276025332, "ret_p3m": 0.194590680189777, "ret_signed_p3m": -0.194590680189777, "alpha_spy_p3m": -0.09775147552207342, "alpha_c_p3m": -0.031925867516542805, "bench_spy_p3m": 0.09683920466770357, "bench_c_p3m": 0.16266481267323418, "ret_p6m": 0.15591059607603341, "ret_signed_p6m": -0.15591059607603341, "alpha_spy_p6m": -0.04634632623956669, "alpha_c_p6m": 0.09340807937538154, "bench_spy_p6m": 0.10956426983646672, "bench_c_p6m": 0.24931867545141495, "price_path": [-0.2466, -0.2585, -0.2673, -0.2553, -0.2535, -0.3117, -0.3624, -0.3399, -0.3489, -0.227, -0.2727, -0.25, -0.2515, -0.2414, -0.2936, -0.3138, -0.3138, -0.3448, -0.3314, -0.3056, -0.2804, -0.2495, -0.2649, -0.2629, -0.2636, -0.2454, -0.2259, -0.2305, -0.2324, -0.2086, -0.2065, -0.2113, -0.1684, -0.1216, -0.085, -0.0884, -0.0846, -0.0834, -0.0915, -0.1089, -0.102, -0.1124, -0.1124, -0.0839, -0.0886, -0.059, -0.0864, -0.0712, -0.0452, -0.0405, -0.0535, -0.0419, -0.0357, -0.0267, -0.0343, -0.0196, -0.0401, -0.0217, -0.0256, -0.0164, -0.0164, -0.0274, -0.0252, 0.0, 0.0433, 0.0481, 0.0666, 0.0682, 0.0365, 0.0632, 0.0773, 0.0773, 0.0699, 0.0818, 0.1013, 0.1095, 0.1151, 0.1093, 0.1542, 0.1587, 0.1697, 0.1657, 0.1588, 0.1293, 0.1547, 0.1747, 0.1731, 0.1951, 0.1867, 0.2121, 0.2026, 0.1746, 0.217, 0.2053, 0.2131, 0.2222, 0.2353, 0.231, 0.2384, 0.2278, 0.2307, 0.2201, 0.2306, 0.1876, 0.1859, 0.1831, 0.2034, 0.2158, 0.229, 0.2279, 0.2182, 0.1777, 0.1777, 0.1547, 0.1536, 0.1606, 0.1293, 0.138, 0.1546, 0.199, 0.198, 0.2024, 0.2019, 0.1825, 0.1515, 0.1917, 0.1946, 0.2415, 0.2065, 0.1966, 0.2015, 0.2049, 0.2296, 0.2616, 0.2661, 0.2772, 0.2686, 0.2546, 0.2512, 0.2787, 0.3021, 0.2385, 0.2734, 0.2173, 0.216, 0.2293, 0.2389, 0.235, 0.225, 0.219, 0.2317, 0.2594, 0.2948, 0.3593, 0.3999, 0.3719, 0.3692, 0.3989, 0.3435, 0.32, 0.2717, 0.2722, 0.3459, 0.3061, 0.3104, 0.2635, 0.2859, 0.2617, 0.2263, 0.2612, 0.2214, 0.2095, 0.2343, 0.2024, 0.2189, 0.2189, 0.1968, 0.2166, 0.227, 0.2143, 0.24, 0.2335, 0.2547, 0.2508, 0.2427, 0.2235, 0.1835, 0.1921, 0.2018, 0.1559]}
{"unit_id": "thread:1937305935137136852", "ticker_idx": 1, "ts": "2025-06-24T00:25:02+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's latest AI chips and integrated rack solutions have performed impressively. This marks a genuine opportunity to challenge Nvidia's leadership. Major companies like Microsoft and OpenAI are likely to increase their procurement of AMD solutions next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish NVDA on wavering dominance, reduced wafer orders post-2026; bullish AMD on strong AI chips and rising procurement by Microsoft/OpenAI.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: UALink camp's first chip aims for tape-out, poised to challenge Nvidia\n\nFollowing Nvidia's announcement of NVLink Fusion, the competition between NVLink and UALink has resurfaced.\n\nSince Nvidia already has a relatively mature product line and ecosystem, it has attracted many customers to develop related chips. This helps Nvidia avoid losing its voice in the ASIC domain and ensures it remains competitive against UALink in the future.\n\nHowever, UALink's overall development pace currently lags behind NVLink, which already has actual products on the market. UALink only officially launched its 1.0 specification several months ago. Still, rumors suggest that UALink's first product could tape out as early as the end of this year, with more related products expected to join the lineup in 2026.\n\nOverall, UALink's product launch schedule is not too far behind that of NVLink. According to industry sources discussing ongoing projects, although NVLink has attracted some customers to start investing, the UALink camp reportedly has \"dozens of projects\" under R&D. Over time, UALink stands a strong chance to leapfrog and challenge Nvidia's dominance.\n\nInsiders note that despite UALink's slower progress, most customers still prefer open standard specifications.\n\nLooking further ahead, it is quite possible that UALink will surpass NVLink in market share. Moreover, market whispers are suggesting that while NVLink Fusion claims to offer an open ecosystem for developing customized products, the actual openness falls short of what industry players initially expected.\n\nCustomers willing to adopt NVLink-based development do so partly as a technical experiment. Currently, Nvidia remains the mainstream brand for AI general-purpose computing, so launching products compatible with the market majority is advantageous. But in the long run, if the level of openness remains limited, expanding UALink's ecosystem becomes an inevitable trend.\n\nAdditionally, industry experts mention that AMD's latest AI chips and integrated rack solutions have performed impressively. This marks a genuine opportunity to challenge Nvidia's leadership. Major companies like Microsoft and OpenAI are likely to increase their procurement of AMD solutions next year.\n\nThis means that starting next year, the UALink specification supported by AMD's compute cores could gradually expand its market influence. With such a foundation, subsequent ASIC product development and market acceptance are expected to grow steadily. The growth potential and ceiling for UALink clearly exceed those of NVLink.\n\nRumors have emerged indicating that Nvidia is less optimistic about its business performance beyond 2026.\n\nThis can be inferred from CEO Jensen Huang's recent global speeches promoting national-level orders, as well as Nvidia's reactions to potential further bans on H20 exports to China.\n\nThere are even reports that Nvidia plans to reduce wafer orders for 2026. Meanwhile, ASIC vendors have recently revised upward their forecasts for the market size going forward. These shifts signal that Nvidia's absolute lead in the cloud AI market may be wavering, presenting a prime opportunity for UALink to expand its influence.\n\nMany industry observers welcome these changes optimistically, believing that increased competition will foster better technological products and accelerate AI development.\n\n---\n\nhttps://t.co/53W9THmNdC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06393116602595106, "ret_signed_m1d": -0.06393116602595106, "alpha_spy_m1d": -0.0530046643801082, "alpha_c_m1d": -0.028503642340523605, "bench_spy_m1d": -0.010926501645842857, "bench_c_m1d": -0.03542752368542745, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.035902632981737126, "ret_signed_p1d": 0.035902632981737126, "alpha_spy_p1d": 0.035342268351132855, "alpha_c_p1d": 0.0224239398892756, "bench_spy_p1d": 0.0005603646306042709, "bench_c_p1d": 0.013478693092461524, "ret_p1w": -0.016759316536723134, "ret_signed_p1w": -0.016759316536723134, "alpha_spy_p1w": -0.03467363816431257, "alpha_c_p1w": -0.03259488412454681, "bench_spy_p1w": 0.017914321627589436, "bench_c_p1w": 0.015835567587823673, "ret_p1m": 0.1460666206062775, "ret_signed_p1m": 0.1460666206062775, "alpha_spy_p1m": 0.1008606952270854, "alpha_c_p1m": 0.09056834784602419, "bench_spy_p1m": 0.0452059253791921, "bench_c_p1m": 0.05549827276025332, "ret_p3m": 0.1369645865565685, "ret_signed_p3m": 0.1369645865565685, "alpha_spy_p3m": 0.04012538188886494, "alpha_c_p3m": -0.02570022611666567, "bench_spy_p3m": 0.09683920466770357, "bench_c_p3m": 0.16266481267323418, "ret_p6m": 0.43112050200239227, "ret_signed_p6m": 0.43112050200239227, "alpha_spy_p6m": 0.32155623216592555, "alpha_c_p6m": 0.18180182655097732, "bench_spy_p6m": 0.10956426983646672, "bench_c_p6m": 0.24931867545141495, "price_path": [-0.2296, -0.2544, -0.2578, -0.2575, -0.2562, -0.3224, -0.3805, -0.3958, -0.435, -0.3004, -0.3592, -0.3253, -0.3173, -0.3116, -0.3622, -0.3679, -0.3679, -0.3819, -0.3769, -0.347, -0.3176, -0.3018, -0.3037, -0.3061, -0.2968, -0.3018, -0.2863, -0.2734, -0.2876, -0.275, -0.2653, -0.2571, -0.219, -0.1876, -0.1496, -0.1693, -0.1536, -0.1711, -0.18, -0.1905, -0.2002, -0.2031, -0.2031, -0.1724, -0.1847, -0.1835, -0.2001, -0.1719, -0.1526, -0.1434, -0.1643, -0.1607, -0.1206, -0.1097, -0.1249, -0.144, -0.1609, -0.087, -0.0818, -0.0841, -0.0841, -0.0736, -0.0639, 0.0, 0.0359, 0.0379, 0.0389, 0.0251, -0.0168, 0.0007, -0.0038, -0.0038, -0.0262, -0.0044, -0.0001, 0.0414, 0.0577, 0.0564, 0.1241, 0.1564, 0.1588, 0.1341, 0.1341, 0.1177, 0.1461, 0.1711, 0.2026, 0.2545, 0.2818, 0.2968, 0.2736, 0.2403, 0.277, 0.2592, 0.1784, 0.2454, 0.248, 0.2445, 0.2638, 0.3322, 0.3072, 0.2823, 0.2724, 0.2031, 0.1934, 0.1826, 0.2119, 0.1801, 0.2036, 0.2073, 0.2178, 0.1748, 0.1748, 0.1726, 0.1712, 0.1687, 0.0918, 0.0938, 0.1256, 0.1525, 0.1245, 0.1455, 0.1642, 0.1591, 0.1498, 0.1408, 0.137, 0.1543, 0.1623, 0.1622, 0.165, 0.1519, 0.1656, 0.1687, 0.1848, 0.2261, 0.1896, 0.4716, 0.5279, 0.7017, 0.6824, 0.5524, 0.5634, 0.5755, 0.7236, 0.6944, 0.6837, 0.7378, 0.7195, 0.6632, 0.6975, 0.8271, 0.8758, 0.8638, 0.9095, 0.8409, 0.8502, 0.8757, 0.8063, 0.8517, 0.7171, 0.6871, 0.7625, 0.7158, 0.8702, 0.7912, 0.7829, 0.7375, 0.6636, 0.6149, 0.4883, 0.4721, 0.5535, 0.4891, 0.5476, 0.5476, 0.5714, 0.5875, 0.5549, 0.5719, 0.5602, 0.5746, 0.5973, 0.601, 0.5995, 0.5996, 0.5226, 0.4995, 0.511, 0.4311]}
{"unit_id": "orig:1930844781535772857", "ticker_idx": 0, "ts": "2025-06-06T04:30:43+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the stock was overcrowded and at elevated levels… a ~5% pullback despite an in-line print appears to be a healthy technical correction", "tweet_type": "analysis", "timeline": "days", "summary": "Near-term bearish on AVGO: overcrowded longs and 90% rally from bottom set up a healthy technical pullback despite in-line results.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Investor Note: What Do Broadcom’s Earnings Tell Us?\n\nMeets Expectations, But Overcrowded Longs Lead to Pullback\n• After-hours: -4%. Results were in line with expectations, but the stock was overcrowded and at elevated levels.\n• Broadcom’s stock had rallied 30% in the month leading up to earnings, and is up nearly 90% from the bottom, suggesting excessive long positioning.\n• At these levels, a ~5% pullback despite an in-line print appears to be a healthy technical correction.\n\n⸻\n\nBroadcom FY2Q / 3Q Revenue and Guidance\n• Total Revenue: $15.0B in F2Q / $15.8B guidance for F3Q\n• AI Revenue: $4.4B in F2Q / $5.1B guidance for F3Q\n\nThis is broadly in line with foreign institutional expectations (F2Q: $15.0B / F3Q: $15.9B; AI: $4.5B / $5.0B).\nHowever, EBITDA guidance was slightly weak.\n\n⸻\n\n🟢 Bull Case\n• Continued growth in AI XPU shipments over the next few quarters\n• Non-AI businesses appear to have bottomed out\n• New demand in AI inference monetization is emerging\n• During the call, Hock Tan was asked:\n“You previously mentioned a ~60% CAGR in AI-related SAM opportunities. Is that a fair way to think about your AI business growth this year and next?”\n\nHe responded concisely: “Yes.”\n• This implies AI revenue will grow:\n• $12.2B in 2024\n• $19.5B in 2025\n• $31.2B in 2026\n\n→ This supports bullish AI sentiment across the market.\n\n⸻\n\n🔴 Bear Case\n• Broadcom may be experiencing over-earning in networking segments\n• AI XPU sales were underwhelming this quarter, as inferred from stronger EBIT margins\n• Some investors were already modeling $30B+ in AI ASIC revenue by 2026\n• Hock’s response on the call was reasonable but not overly enthusiastic\n• The current quarter appears to be a transition phase, with greater scrutiny likely in H2 2025", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.052646542785388384, "ret_signed_m1d": -0.052646542785388384, "alpha_spy_m1d": -0.06281114257384912, "alpha_c_m1d": -0.05830476864638645, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010731753810176325, "ret_signed_p1d": 0.010731753810176325, "alpha_spy_p1d": 0.011633036698320609, "alpha_c_p1d": 0.027468900493561765, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.007168124310281021, "ret_signed_p1w": -0.007168124310281021, "alpha_spy_p1w": -0.010739871198886974, "alpha_c_p1w": 0.009687727259191314, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.11296872464411822, "ret_signed_p1m": -0.11296872464411822, "alpha_spy_p1m": -0.07395462322995838, "alpha_c_p1m": -0.006016591299285778, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.22748067541138628, "ret_signed_p3m": -0.22748067541138628, "alpha_spy_p3m": -0.14986431981113468, "alpha_c_p3m": -0.09413681206638924, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.569885380902382, "ret_signed_p6m": -0.569885380902382, "alpha_spy_p6m": -0.4279607859361505, "alpha_c_p6m": -0.1732961114364766, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.2325, -0.2158, -0.2274, -0.2105, -0.2147, -0.2382, -0.2104, -0.2284, -0.2238, -0.2255, -0.2376, -0.274, -0.3035, -0.3151, -0.322, -0.3175, -0.3031, -0.3763, -0.4076, -0.3758, -0.3681, -0.2502, -0.3022, -0.2632, -0.2777, -0.2753, -0.2929, -0.3075, -0.3075, -0.3269, -0.3132, -0.2836, -0.238, -0.2212, -0.2205, -0.2258, -0.2205, -0.2009, -0.1753, -0.1871, -0.1897, -0.1706, -0.1586, -0.1568, -0.1027, -0.0588, -0.06, -0.0579, -0.0742, -0.066, -0.0618, -0.0697, -0.0664, -0.0737, -0.0737, -0.0457, -0.0304, -0.0201, -0.0197, 0.0072, 0.0402, 0.0573, 0.0526, 0.0, -0.0107, -0.0093, 0.0242, 0.037, 0.0072, 0.0209, 0.0099, 0.0175, 0.0175, 0.0148, 0.0301, 0.0707, 0.0743, 0.0967, 0.0934, 0.1189, 0.0746, 0.0956, 0.117, 0.117, 0.113, 0.1033, 0.1281, 0.1179, 0.1138, 0.1187, 0.1404, 0.1399, 0.1628, 0.1502, 0.1699, 0.1309, 0.1516, 0.1719, 0.1779, 0.1946, 0.2073, 0.2284, 0.1922, 0.1717, 0.2085, 0.1891, 0.2246, 0.233, 0.238, 0.2336, 0.2699, 0.2547, 0.2634, 0.2435, 0.2412, 0.1971, 0.1819, 0.1756, 0.1934, 0.1944, 0.2097, 0.2188, 0.2529, 0.2072, 0.2072, 0.2106, 0.2275, 0.2425, 0.3594, 0.4031, 0.3666, 0.5002, 0.4598, 0.4608, 0.4779, 0.4613, 0.4052, 0.4019, 0.4002, 0.3776, 0.3782, 0.3797, 0.3667, 0.3603, 0.3333, 0.3415, 0.3556, 0.3751, 0.3759, 0.3642, 0.3679, 0.4049, 0.4029, 0.32, 0.4504, 0.3993, 0.4286, 0.4401, 0.4205, 0.4201, 0.3933, 0.3837, 0.4, 0.44, 0.4722, 0.5166, 0.5695, 0.5308, 0.503, 0.4742, 0.4311, 0.4597, 0.4459, 0.4209, 0.4573, 0.4311, 0.4444, 0.3824, 0.3925, 0.3933, 0.3845, 0.4411, 0.4102, 0.3833, 0.5369, 0.5656, 0.6166, 0.6166, 0.6385, 0.5699]}
{"unit_id": "orig:1930760368324047155", "ticker_idx": 0, "ts": "2025-06-05T22:55:18+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AVGO remains his top pick, and he is bullish into the print. He expects July-quarter revenues to grow QoQ by +5~7%... guidance could come in close to $16.1B (vs. consensus at $15.75B)", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares/adopts JP Morgan Harlan Sur's bullish AVGO preview: AI revenue >$5B, guidance ~$16.1B above $15.75B consensus, FY25 AI revenue $19-20B, top pick into earnings.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan AVGO Earnings Preview (by Joshua Myers)\n\n1. Investor Sentiment:\nAVGO has been the most crowded long in the semiconductor space over the past five months and remains heavily crowded (though slightly less so heading into earnings, as short interest has ticked up a bit). This stems from high expectations around “Harlan,” the AI ASIC reference platform first mentioned last fall, which has solidified AVGO’s position as the undisputed leader in the space. There’s also a broad perception that this stock remains hard to short. According to our buyside survey, there’s a mildly optimistic—but not irrational—view that AI semis will provide a modest boost to earnings (excluding margins). Investors are bullish on the TPUv6p ramp, and in addition to noise around SKU losses dying down, optimism is also growing around new projects. Minor watchpoints include recovery in networking, wireless, and broadband. The key investor question is: How should I be positioned for this “transition quarter”? Harlan makes clear in today’s note that the ramp has already begun, so if he’s right, this setup could be primed for upside surprise.\n2. Positioning Score (1 = max short/underweight, 10 = max long/overweight):\n9 (The risk here is that there are very few shorts left.)\n3. Implied Volatility:\n6.7%\n4. Recent Analyst Commentary (Harlan Sur):\nYesterday, Harlan Sur published a well-received note elaborating on the underappreciated (at least based on same-day investor calls) growth drivers from AVGO’s 3nm TPU ASIC and MTIA ramp, as well as the teased launch of Tomahawk 6 (including the Tomahawk 6 F1 stand-up chip), which was hinted at on LinkedIn. Today, he officially previews the April-quarter earnings set to be announced Thursday morning (I’ve previewed them separately here).\n\nHarlan sees continued strength in AI products (custom ASICs and networking) as well as stabilization in non-AI semis (enterprise, server/storage, broadband, and wireless). Combined with improving VMware profitability, these factors are expected to deliver slightly better-than-consensus results across revenue, EPS, and FCF. He expects July-quarter revenues to grow QoQ by +5~7%.\n\nHarlan’s expectations for revenue and guidance are even higher than our own bullish buyside earnings bar. On revenue guidance, he thinks it could come in close to $16.1B (vs. consensus at $15.75B, and our buyside bar at $15.93B). This is driven by his expectation that AI semiconductor revenue (driven by the 3nm TPU6 ramp for GOOGL) could exceed $5B (vs. consensus at $4.8B and our buyside bar at $4.95B). He believes this single TPU6 SKU—described as “the world’s most powerful custom XPU AI accelerator”—could generate over $15B in lifetime revenue.\n\nIn addition, driven by strong networking demand and the launch of the next-gen Tomahawk 6, Harlan expects FY25 AI revenue to reach $19–20B, up 60% YoY.\n\nBeyond AI, cyclical trends are improving (with potential upside in wireless), and VMware renewals remain strong (driven by higher ASP VCF full-stack solution upgrades/upsells). AVGO remains his top pick, and he is bullish into the print.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004424229914138156, "ret_signed_m1d": 0.004424229914138156, "alpha_spy_m1d": -0.00043200821943245593, "alpha_c_m1d": 0.002633524406622456, "bench_spy_m1d": 0.004856238133570612, "bench_c_m1d": 0.0017907055075156997, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.050013504671835296, "ret_signed_p1d": -0.050013504671835296, "alpha_spy_p1d": -0.0602824845307981, "alpha_c_p1d": -0.055703928234596134, "bench_spy_p1d": 0.010268979858962801, "bench_c_p1d": 0.005690423562760838, "ret_p1w": -0.014850108794867256, "ret_signed_p1w": -0.014850108794867256, "alpha_spy_p1w": -0.03289244241658451, "alpha_c_p1w": -0.06200486655931148, "bench_spy_p1w": 0.018042333621717255, "bench_c_p1w": 0.04715475776444422, "ret_p1m": 0.06116148699574664, "ret_signed_p1m": 0.06116148699574664, "alpha_spy_p1m": 0.003596780441048386, "alpha_c_p1m": -0.06744978433816451, "bench_spy_p1m": 0.05756470655469825, "bench_c_p1m": 0.12861127133391115, "ret_p3m": 0.15008649618034386, "ret_signed_p3m": 0.15008649618034386, "alpha_spy_p3m": 0.06727249506629573, "alpha_c_p3m": 0.01009437772091104, "bench_spy_p3m": 0.08281400111404813, "bench_c_p3m": 0.13999211845943282, "ret_p6m": 0.5565748696165951, "ret_signed_p6m": 0.5565748696165951, "alpha_spy_p6m": 0.397632821876998, "alpha_c_p6m": 0.15474433004778976, "bench_spy_p6m": 0.15894204773959708, "bench_c_p6m": 0.4018305395688053, "price_path": [-0.2925, -0.2709, -0.255, -0.266, -0.25, -0.254, -0.2763, -0.2499, -0.267, -0.2626, -0.2642, -0.2757, -0.3103, -0.3383, -0.3494, -0.3559, -0.3517, -0.3379, -0.4075, -0.4372, -0.407, -0.3997, -0.2877, -0.3371, -0.3, -0.3138, -0.3115, -0.3282, -0.3422, -0.3422, -0.3606, -0.3476, -0.3194, -0.2762, -0.2601, -0.2595, -0.2645, -0.2595, -0.2408, -0.2166, -0.2278, -0.2302, -0.2121, -0.2007, -0.199, -0.1475, -0.1058, -0.107, -0.105, -0.1205, -0.1127, -0.1087, -0.1162, -0.1131, -0.1201, -0.1201, -0.0934, -0.0789, -0.0691, -0.0687, -0.0432, -0.0118, 0.0044, 0.0, -0.05, -0.0602, -0.0589, -0.027, -0.0149, -0.0432, -0.0301, -0.0406, -0.0334, -0.0334, -0.036, -0.0214, 0.0172, 0.0206, 0.0418, 0.0387, 0.063, 0.0209, 0.0408, 0.0612, 0.0612, 0.0573, 0.0481, 0.0717, 0.062, 0.0581, 0.0628, 0.0834, 0.0829, 0.1046, 0.0926, 0.1114, 0.0743, 0.094, 0.1133, 0.119, 0.1349, 0.1469, 0.167, 0.1326, 0.1131, 0.1481, 0.1296, 0.1633, 0.1714, 0.176, 0.1719, 0.2063, 0.1919, 0.2002, 0.1813, 0.1791, 0.1372, 0.1228, 0.1168, 0.1337, 0.1346, 0.1492, 0.1578, 0.1902, 0.1468, 0.1468, 0.1501, 0.1661, 0.1804, 0.2914, 0.3329, 0.2983, 0.4252, 0.3868, 0.3877, 0.404, 0.3882, 0.3349, 0.3318, 0.3302, 0.3087, 0.3093, 0.3107, 0.2983, 0.2922, 0.2666, 0.2744, 0.2878, 0.3063, 0.3071, 0.2959, 0.2995, 0.3346, 0.3328, 0.254, 0.3779, 0.3293, 0.3571, 0.368, 0.3494, 0.3491, 0.3236, 0.3145, 0.3299, 0.368, 0.3985, 0.4407, 0.491, 0.4542, 0.4278, 0.4005, 0.3595, 0.3867, 0.3736, 0.3498, 0.3844, 0.3596, 0.3722, 0.3133, 0.3229, 0.3236, 0.3153, 0.3691, 0.3397, 0.3141, 0.46, 0.4873, 0.5358, 0.5358, 0.5566]}
{"unit_id": "orig:1924622189376193013", "ticker_idx": 0, "ts": "2025-05-20T00:24:22+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "CITI maintains a favorable view on SK Hynix", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: HBM supply deficit deepens to -10%/-6% in 2025/2026; maintains favorable view on SK Hynix; NVDA and MU tagged as beneficiaries of tight HBM supply and surging demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250520_Tight HBM Supply to Continue in 2025E & 2026E on Continued HBM Quality Issues – CITI\n\n1. CITI projects that the HBM supply shortage will intensify further, extending beyond 2025 into 2026.\n\n2. This is due to persistent quality issues with HBM3E 12hi and lower-than-expected yield rates (especially in TSV and bonding processes), which are likely to limit the growth rate of HBM production.\n\n3. While demand remains solid, supply shortages are expected to worsen.\n\n4. Accordingly, the supply/demand ratio forecast has been revised from -8%/-3% to -10%/-6% for 2025E/2026E, respectively.\n\n5. The 2025 HBM supply growth forecast has been lowered by 11 percentage points, from YoY +97% to YoY +86%.\n\n6. While the market is concerned that Samsung Electronics entering the HBM3E 12hi market after passing qualification tests could lead to oversupply, CITI believes Samsung’s entry would have only a limited impact on overall market balance.\n\n7. CITI expects Samsung’s HBM supply ramp-up to fall significantly short of market expectations and, as a result, has revised down the 2025 HBM bit shipment forecast by 6%.\n\n8. In contrast, HBM demand is projected to grow strongly, with YoY +106% in 2025 and YoY +57% in 2026. This growth is being driven not only by NVIDIA but also by rising ASIC demand.\n\n•2025 Supply: YoY +86%\n•2026 Supply: YoY +48%\n\n9. CITI expects the spread of AI supercomputing to accelerate content per system, driving HBM demand growth in both 2025 and 2026.\n\n10. Stacking is expected to increase from 8hi for HBM3 to 12hi for HBM3E, and eventually to 16hi for HBM4. Correspondingly, memory capacity per stack is projected to rise from 16GB → 24GB → 32GB.\n\n11. As a result, given the increasingly tight HBM supply-demand balance through 2025 and 2026, CITI maintains a favorable view on SK Hynix.\n\n$MU $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01287130207535725, "ret_signed_m1d": -0.01287130207535725, "alpha_spy_m1d": -0.01624477216356146, "alpha_c_m1d": -0.014785489597911594, "bench_spy_m1d": 0.0033734700882042112, "bench_c_m1d": 0.0019141875225543448, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007425885877136995, "ret_signed_p1d": -0.007425885877136995, "alpha_spy_p1d": 0.00942480339540308, "alpha_c_p1d": 0.009149755012880734, "bench_spy_p1d": -0.016850689272540076, "bench_c_p1d": -0.01657564089001773, "ret_p1w": 0.0024752174773745583, "ret_signed_p1w": 0.0024752174773745583, "alpha_spy_p1w": 0.005342604572966314, "alpha_c_p1w": 0.004796591403593453, "bench_spy_p1w": -0.0028673870955917558, "bench_c_p1w": -0.002321373926218895, "ret_p1m": 0.2225009555682549, "ret_signed_p1m": 0.2225009555682549, "alpha_spy_p1m": 0.21475861470740565, "alpha_c_p1m": 0.15306214657726502, "bench_spy_p1m": 0.0077423408608492394, "bench_c_p1m": 0.06943880899098986, "ret_p3m": 0.3712838666462479, "ret_signed_p3m": 0.3712838666462479, "alpha_spy_p3m": 0.28274173138421776, "alpha_c_p3m": 0.1672030011220511, "bench_spy_p3m": 0.08854213526203014, "bench_c_p3m": 0.2040808655241968, "ret_p6m": 2.064391916203823, "ret_signed_p6m": 2.064391916203823, "alpha_spy_p6m": 1.9050759488063975, "alpha_c_p6m": 1.616440430980195, "bench_spy_p6m": 0.1593159673974256, "bench_c_p6m": 0.44795148522362815, "price_path": [0.0428, 0.0305, 0.0083, -0.0138, -0.0015, -0.0139, -0.0584, -0.0584, -0.0787, -0.0441, -0.0475, -0.0475, -0.0698, -0.0703, -0.0153, -0.0114, 0.0124, 0.0198, 0.005, 0.0173, 0.0396, 0.0668, 0.047, 0.0297, 0.0594, 0.0248, -0.0134, -0.0559, -0.0248, -0.0203, -0.0366, -0.098, -0.1842, -0.1609, -0.1832, -0.0931, -0.105, -0.1079, -0.1059, -0.1386, -0.1337, -0.1337, -0.1257, -0.1396, -0.104, -0.1173, -0.0871, -0.099, -0.105, -0.1213, -0.1213, -0.0792, -0.0792, -0.0792, -0.0554, -0.0579, -0.0589, -0.0347, -0.0173, 0.0198, -0.0074, 0.0124, -0.0129, 0.0, -0.0074, -0.0252, -0.0099, 0.005, 0.0025, 0.0297, 0.0514, 0.0142, 0.0291, 0.0291, 0.0787, 0.1134, 0.1134, 0.1357, 0.1432, 0.1903, 0.1679, 0.1679, 0.2299, 0.2349, 0.2225, 0.22, 0.2746, 0.287, 0.3812, 0.4184, 0.4531, 0.4085, 0.4482, 0.4159, 0.3837, 0.3812, 0.3415, 0.344, 0.3986, 0.3936, 0.473, 0.4606, 0.4878, 0.4804, 0.468, 0.3366, 0.3341, 0.3514, 0.3316, 0.3341, 0.3366, 0.3192, 0.2994, 0.3019, 0.3068, 0.3564, 0.2795, 0.2795, 0.3068, 0.282, 0.2994, 0.2721, 0.3242, 0.3341, 0.3787, 0.3713, 0.3713, 0.3266, 0.3043, 0.2671, 0.2151, 0.2448, 0.287, 0.2969, 0.2895, 0.3335, 0.336, 0.2714, 0.2938, 0.3037, 0.3186, 0.3584, 0.3757, 0.4304, 0.5098, 0.5247, 0.6315, 0.6439, 0.7284, 0.6564, 0.7656, 0.7656, 0.7433, 0.7929, 0.7756, 0.7706, 0.6713, 0.7333, 0.7259, 0.788, 0.9643, 0.9643, 0.9643, 0.9643, 0.9643, 0.9643, 1.1257, 1.0611, 1.0438, 1.0984, 1.2474, 1.312, 1.4113, 1.379, 1.3914, 1.3765, 1.533, 1.6571, 1.5876, 1.7714, 1.821, 1.7763, 2.0793, 1.9104, 1.8757, 1.9452, 1.8806, 2.0098, 2.0743, 2.0644]}
{"unit_id": "orig:1924622189376193013", "ticker_idx": 1, "ts": "2025-05-20T00:24:22+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU $NVDA", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: HBM supply deficit deepens to -10%/-6% in 2025/2026; maintains favorable view on SK Hynix; NVDA and MU tagged as beneficiaries of tight HBM supply and surging demand.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250520_Tight HBM Supply to Continue in 2025E & 2026E on Continued HBM Quality Issues – CITI\n\n1. CITI projects that the HBM supply shortage will intensify further, extending beyond 2025 into 2026.\n\n2. This is due to persistent quality issues with HBM3E 12hi and lower-than-expected yield rates (especially in TSV and bonding processes), which are likely to limit the growth rate of HBM production.\n\n3. While demand remains solid, supply shortages are expected to worsen.\n\n4. Accordingly, the supply/demand ratio forecast has been revised from -8%/-3% to -10%/-6% for 2025E/2026E, respectively.\n\n5. The 2025 HBM supply growth forecast has been lowered by 11 percentage points, from YoY +97% to YoY +86%.\n\n6. While the market is concerned that Samsung Electronics entering the HBM3E 12hi market after passing qualification tests could lead to oversupply, CITI believes Samsung’s entry would have only a limited impact on overall market balance.\n\n7. CITI expects Samsung’s HBM supply ramp-up to fall significantly short of market expectations and, as a result, has revised down the 2025 HBM bit shipment forecast by 6%.\n\n8. In contrast, HBM demand is projected to grow strongly, with YoY +106% in 2025 and YoY +57% in 2026. This growth is being driven not only by NVIDIA but also by rising ASIC demand.\n\n•2025 Supply: YoY +86%\n•2026 Supply: YoY +48%\n\n9. CITI expects the spread of AI supercomputing to accelerate content per system, driving HBM demand growth in both 2025 and 2026.\n\n10. Stacking is expected to increase from 8hi for HBM3 to 12hi for HBM3E, and eventually to 16hi for HBM4. Correspondingly, memory capacity per stack is projected to rise from 16GB → 24GB → 32GB.\n\n11. As a result, given the increasingly tight HBM supply-demand balance through 2025 and 2026, CITI maintains a favorable view on SK Hynix.\n\n$MU $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005606566351137765, "ret_signed_m1d": 0.005606566351137765, "alpha_spy_m1d": 0.002233096262933554, "alpha_c_m1d": 0.0036923788285834203, "bench_spy_m1d": 0.0033734700882042112, "bench_c_m1d": 0.0019141875225543448, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023037766092255718, "ret_signed_p1d": -0.023037766092255718, "alpha_spy_p1d": -0.006187076819715642, "alpha_c_p1d": -0.006462125202237989, "bench_spy_p1d": -0.016850689272540076, "bench_c_p1d": -0.01657564089001773, "ret_p1w": -0.017533168494243467, "ret_signed_p1w": -0.017533168494243467, "alpha_spy_p1w": -0.014665781398651712, "alpha_c_p1w": -0.015211794568024573, "bench_spy_p1w": -0.0028673870955917558, "bench_c_p1w": -0.002321373926218895, "ret_p1m": 0.2417941074689851, "ret_signed_p1m": 0.2417941074689851, "alpha_spy_p1m": 0.23405176660813587, "alpha_c_p1m": 0.17235529847799524, "bench_spy_p1m": 0.0077423408608492394, "bench_c_p1m": 0.06943880899098986, "ret_p3m": 0.23326981594718021, "ret_signed_p3m": 0.23326981594718021, "alpha_spy_p3m": 0.14472768068515007, "alpha_c_p3m": 0.029188950422983417, "bench_spy_p3m": 0.08854213526203014, "bench_c_p3m": 0.2040808655241968, "ret_p6m": 1.5003467951057181, "ret_signed_p6m": 1.5003467951057181, "alpha_spy_p6m": 1.3410308277082925, "alpha_c_p6m": 1.05239530988209, "bench_spy_p6m": 0.1593159673974256, "bench_c_p6m": 0.44795148522362815, "price_path": [0.0504, 0.0062, -0.0287, -0.051, -0.0053, -0.0652, -0.0468, -0.0783, -0.0715, -0.0396, -0.0912, -0.0536, -0.1135, -0.0934, -0.0263, -0.0341, 0.0261, 0.0497, 0.0356, 0.039, 0.0486, -0.0357, -0.0131, -0.0412, -0.0621, -0.072, -0.0996, -0.1143, -0.0957, -0.0968, -0.2422, -0.3403, -0.3031, -0.3319, -0.2062, -0.2859, -0.291, -0.276, -0.2758, -0.2933, -0.2987, -0.2987, -0.3197, -0.2843, -0.2566, -0.2108, -0.1867, -0.1992, -0.2163, -0.2156, -0.2072, -0.1772, -0.1802, -0.1793, -0.1578, -0.132, -0.1248, -0.0592, -0.0119, -0.0283, -0.027, -0.001, 0.0056, 0.0, -0.023, -0.0333, -0.0482, -0.0482, -0.0175, -0.0196, -0.0133, -0.0371, 0.0008, 0.0423, 0.0525, 0.0835, 0.1066, 0.131, 0.1635, 0.1828, 0.1843, 0.1784, 0.2216, 0.2267, 0.2418, 0.2418, 0.2599, 0.2444, 0.3039, 0.2971, 0.2844, 0.2718, 0.2564, 0.2323, 0.241, 0.2466, 0.2466, 0.2236, 0.2695, 0.2472, 0.2561, 0.2706, 0.2102, 0.2255, 0.188, 0.1556, 0.1672, 0.1553, 0.1144, 0.1206, 0.14, 0.1352, 0.1351, 0.1424, 0.1707, 0.1136, 0.0701, 0.0996, 0.1128, 0.1099, 0.1414, 0.2131, 0.2623, 0.3035, 0.268, 0.2784, 0.2333, 0.2606, 0.2453, 0.1959, 0.1814, 0.2007, 0.1879, 0.1887, 0.2014, 0.2448, 0.2143, 0.2143, 0.2089, 0.2113, 0.2673, 0.3404, 0.3413, 0.3799, 0.4285, 0.5363, 0.6043, 0.6098, 0.6205, 0.6324, 0.7232, 0.6604, 0.6797, 0.6979, 0.65, 0.6002, 0.6047, 0.6723, 0.7072, 0.8585, 0.8749, 0.9177, 0.9496, 0.8958, 1.0066, 0.9636, 0.8541, 0.9681, 0.9098, 0.9596, 1.0678, 1.0662, 1.1111, 1.0653, 1.0263, 1.1104, 1.2361, 1.2471, 1.2656, 1.3138, 1.2871, 1.2846, 1.3962, 1.226, 1.4248, 1.4333, 1.4291, 1.5861, 1.4617, 1.5003]}
{"unit_id": "orig:1924622189376193013", "ticker_idx": 2, "ts": "2025-05-20T00:24:22+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM demand is projected to grow strongly, with YoY +106% in 2025 and YoY +57% in 2026. This growth is being driven not only by NVIDIA but also by rising ASIC demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: HBM supply deficit deepens to -10%/-6% in 2025/2026; maintains favorable view on SK Hynix; NVDA and MU tagged as beneficiaries of tight HBM supply and surging demand.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250520_Tight HBM Supply to Continue in 2025E & 2026E on Continued HBM Quality Issues – CITI\n\n1. CITI projects that the HBM supply shortage will intensify further, extending beyond 2025 into 2026.\n\n2. This is due to persistent quality issues with HBM3E 12hi and lower-than-expected yield rates (especially in TSV and bonding processes), which are likely to limit the growth rate of HBM production.\n\n3. While demand remains solid, supply shortages are expected to worsen.\n\n4. Accordingly, the supply/demand ratio forecast has been revised from -8%/-3% to -10%/-6% for 2025E/2026E, respectively.\n\n5. The 2025 HBM supply growth forecast has been lowered by 11 percentage points, from YoY +97% to YoY +86%.\n\n6. While the market is concerned that Samsung Electronics entering the HBM3E 12hi market after passing qualification tests could lead to oversupply, CITI believes Samsung’s entry would have only a limited impact on overall market balance.\n\n7. CITI expects Samsung’s HBM supply ramp-up to fall significantly short of market expectations and, as a result, has revised down the 2025 HBM bit shipment forecast by 6%.\n\n8. In contrast, HBM demand is projected to grow strongly, with YoY +106% in 2025 and YoY +57% in 2026. This growth is being driven not only by NVIDIA but also by rising ASIC demand.\n\n•2025 Supply: YoY +86%\n•2026 Supply: YoY +48%\n\n9. CITI expects the spread of AI supercomputing to accelerate content per system, driving HBM demand growth in both 2025 and 2026.\n\n10. Stacking is expected to increase from 8hi for HBM3 to 12hi for HBM3E, and eventually to 16hi for HBM4. Correspondingly, memory capacity per stack is projected to rise from 16GB → 24GB → 32GB.\n\n11. As a result, given the increasingly tight HBM supply-demand balance through 2025 and 2026, CITI maintains a favorable view on SK Hynix.\n\n$MU $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008855431327049468, "ret_signed_m1d": 0.008855431327049468, "alpha_spy_m1d": 0.0054819612388452565, "alpha_c_m1d": 0.006941243804495123, "bench_spy_m1d": 0.0033734700882042112, "bench_c_m1d": 0.0019141875225543448, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019199280198776703, "ret_signed_p1d": -0.019199280198776703, "alpha_spy_p1d": -0.0023485909262366267, "alpha_c_p1d": -0.0026236393087589738, "bench_spy_p1d": -0.016850689272540076, "bench_c_p1d": -0.01657564089001773, "ret_p1w": 0.008334456792334688, "ret_signed_p1w": 0.008334456792334688, "alpha_spy_p1w": 0.011201843887926444, "alpha_c_p1w": 0.010655830718553583, "bench_spy_p1w": -0.0028673870955917558, "bench_c_p1w": -0.002321373926218895, "ret_p1m": 0.08267662564328204, "ret_signed_p1m": 0.08267662564328204, "alpha_spy_p1m": 0.0749342847824328, "alpha_c_p1m": 0.013237816652292178, "bench_spy_p1m": 0.0077423408608492394, "bench_c_p1m": 0.06943880899098986, "ret_p3m": 0.34292697281766804, "ret_signed_p3m": 0.34292697281766804, "alpha_spy_p3m": 0.2543848375556379, "alpha_c_p3m": 0.13884610729347124, "bench_spy_p3m": 0.08854213526203014, "bench_c_p3m": 0.2040808655241968, "ret_p6m": 0.4423604201096851, "ret_signed_p6m": 0.4423604201096851, "alpha_spy_p6m": 0.28304445271225953, "alpha_c_p6m": -0.005591065113943028, "bench_spy_p6m": 0.1593159673974256, "bench_c_p6m": 0.44795148522362815, "price_path": [0.0425, 0.0003, -0.0306, -0.0578, -0.0232, -0.106, -0.0705, -0.1513, -0.1369, -0.1272, -0.1773, -0.1615, -0.204, -0.1907, -0.1387, -0.1399, -0.0946, -0.1105, -0.141, -0.1255, -0.1179, -0.1241, -0.0965, -0.1019, -0.1534, -0.1708, -0.1839, -0.1935, -0.1803, -0.1783, -0.2424, -0.2982, -0.2734, -0.2834, -0.1492, -0.1995, -0.1745, -0.1761, -0.1651, -0.2224, -0.2448, -0.2448, -0.2788, -0.2641, -0.2357, -0.208, -0.1739, -0.1909, -0.1887, -0.1895, -0.1694, -0.1479, -0.153, -0.1551, -0.1289, -0.1266, -0.1319, -0.0847, -0.0331, 0.0071, 0.0033, 0.0076, 0.0089, 0.0, -0.0192, -0.0115, -0.023, -0.023, 0.0083, 0.0032, 0.0358, 0.0056, 0.0223, 0.0509, 0.0561, 0.0417, 0.0546, 0.0614, 0.0713, 0.063, 0.0791, 0.0566, 0.0768, 0.0726, 0.0827, 0.0827, 0.0705, 0.0729, 0.1007, 0.1484, 0.1537, 0.174, 0.1758, 0.1409, 0.1703, 0.1858, 0.1858, 0.1776, 0.1907, 0.2122, 0.2212, 0.2274, 0.221, 0.2704, 0.2754, 0.2875, 0.2831, 0.2754, 0.2431, 0.271, 0.293, 0.2912, 0.3154, 0.3062, 0.3341, 0.3237, 0.2928, 0.3396, 0.3266, 0.3353, 0.3453, 0.3597, 0.3549, 0.3631, 0.3514, 0.3546, 0.3429, 0.3545, 0.3071, 0.3053, 0.3022, 0.3246, 0.3382, 0.3528, 0.3515, 0.3408, 0.2963, 0.2963, 0.271, 0.2698, 0.2775, 0.243, 0.2526, 0.2708, 0.3197, 0.3186, 0.3234, 0.3229, 0.3015, 0.2674, 0.3117, 0.3149, 0.3665, 0.328, 0.3171, 0.3225, 0.3262, 0.3534, 0.3886, 0.3935, 0.4058, 0.3964, 0.3809, 0.3772, 0.4075, 0.4332, 0.3632, 0.4016, 0.3399, 0.3384, 0.3531, 0.3636, 0.3593, 0.3483, 0.3417, 0.3557, 0.3862, 0.4252, 0.4962, 0.5409, 0.51, 0.507, 0.5397, 0.4788, 0.4529, 0.3998, 0.4003, 0.4814, 0.4376, 0.4424]}
{"unit_id": "quote:1937842290539798659", "ticker_idx": 0, "ts": "2025-06-25T11:56:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "These people think Nvidia could go up another 70% from here", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Loop Capital's Buy reiteration and PT raise to $250, implying ~70% upside from current levels.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Wow! These people think Nvidia could go up another 70% from here.", "reply_text": null, "parent_text": null, "quoted_text": "Loop Capital Reiterates Buy on $NVDA, Raises PT to $250 from $175; says Gen AI compute spending outlook supports long-term valuation\n\nAnalyst comments: \"Reiterating our Buy rating as our work strongly suggests that 'the math just works', as fantastic as that may sound.", "parent_author": null, "ret_m1d": -0.04153985983831754, "ret_signed_m1d": -0.04153985983831754, "alpha_spy_m1d": -0.040979809040371906, "alpha_c_m1d": -0.02824042573641894, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004601045575293838, "ret_signed_p1d": 0.004601045575293838, "alpha_spy_p1d": -0.003222734299193508, "alpha_c_p1d": -0.0026664284599469124, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": 0.019052474587535784, "ret_signed_p1w": 0.019052474587535784, "alpha_spy_p1w": -0.002903666054156373, "alpha_c_p1w": -0.0029314869794923037, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": 0.12591537288427257, "ret_signed_p1m": 0.12591537288427257, "alpha_spy_p1m": 0.08094899164682601, "alpha_c_p1m": 0.0801304978133448, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.18994459598187352, "ret_signed_p3m": 0.18994459598187352, "alpha_spy_p3m": 0.08853342815936704, "alpha_c_p3m": 0.019268691863630893, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.1286339477687075, "ret_signed_p6m": 0.1286339477687075, "alpha_spy_p6m": 0.011317093960519164, "alpha_c_p6m": -0.13317565582452806, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.2893, -0.2977, -0.2862, -0.2845, -0.3403, -0.3889, -0.3673, -0.376, -0.2591, -0.3029, -0.2812, -0.2826, -0.2729, -0.3229, -0.3423, -0.3423, -0.372, -0.3592, -0.3344, -0.3103, -0.2807, -0.2954, -0.2935, -0.2942, -0.2768, -0.258, -0.2624, -0.2643, -0.2414, -0.2394, -0.2441, -0.203, -0.1581, -0.123, -0.1263, -0.1226, -0.1215, -0.1292, -0.1459, -0.1393, -0.1492, -0.1492, -0.122, -0.1264, -0.098, -0.1244, -0.1098, -0.0849, -0.0804, -0.0929, -0.0817, -0.0758, -0.0671, -0.0744, -0.0603, -0.08, -0.0623, -0.066, -0.0572, -0.0572, -0.0678, -0.0657, -0.0415, 0.0, 0.0046, 0.0223, 0.0238, -0.0065, 0.0191, 0.0326, 0.0326, 0.0255, 0.0369, 0.0555, 0.0634, 0.0688, 0.0632, 0.1062, 0.1106, 0.1211, 0.1173, 0.1106, 0.0824, 0.1067, 0.1259, 0.1244, 0.1454, 0.1374, 0.1618, 0.1527, 0.1258, 0.1665, 0.1552, 0.1627, 0.1715, 0.184, 0.1798, 0.187, 0.1768, 0.1796, 0.1694, 0.1795, 0.1382, 0.1367, 0.134, 0.1535, 0.1653, 0.178, 0.1769, 0.1676, 0.1288, 0.1288, 0.1067, 0.1057, 0.1124, 0.0824, 0.0907, 0.1066, 0.1492, 0.1482, 0.1524, 0.152, 0.1334, 0.1036, 0.1422, 0.145, 0.1899, 0.1564, 0.1469, 0.1516, 0.1548, 0.1785, 0.2092, 0.2135, 0.2242, 0.2159, 0.2025, 0.1992, 0.2256, 0.248, 0.187, 0.2205, 0.1667, 0.1654, 0.1783, 0.1874, 0.1837, 0.1741, 0.1684, 0.1805, 0.2071, 0.241, 0.3028, 0.3418, 0.3149, 0.3123, 0.3408, 0.2877, 0.2651, 0.2189, 0.2194, 0.29, 0.2518, 0.256, 0.211, 0.2325, 0.2093, 0.1754, 0.2088, 0.1707, 0.1593, 0.1831, 0.1524, 0.1682, 0.1682, 0.1471, 0.166, 0.176, 0.1639, 0.1885, 0.1822, 0.2026, 0.1988, 0.1911, 0.1726, 0.1343, 0.1426, 0.1518, 0.1079, 0.1286]}
{"unit_id": "orig:1942352197834137683", "ticker_idx": 0, "ts": "2025-07-07T22:37:05+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM maintains its \"Overweight\" rating on Foxconn given the improving GB200 yield and the company's leading position in the AI server market", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM reiterates Overweight on Foxconn; GB200 ramp + AI server leadership to drive growth and offset H2 seasonality.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Foxconn: GB200 Shipments Continue to Advance, AI Server Business is the Main Growth Driver\nJPM (G. Hariharan, 25/07/07)\n\nJPM notes that despite weaker June sales due to exchange rates and prior pull-ins, Foxconn's overall 2Q25 revenue was in line with expectations. AI server shipments, especially for GB200, are steadily progressing, becoming the core driver of the company's future revenue growth.\n\nJune Sales Weakened by NTD Appreciation and Prior Pull-in Effects\nFoxconn's June revenue was NT$540 billion, a 12% month-over-month decrease but a 10% year-over-year increase. The decline was primarily due to the 1-2% appreciation of the New Taiwan Dollar (NTD) in June and the fading effect of customers pulling in orders earlier due to tariff expectations, leading to a sales decline in computing, smart consumer electronics, and components. Cloud and networking products remained stable, and AI revenue continued to grow but was partially offset by weakness in general servers and a decline in NVIDIA HGX shipments.\n\nSecond Quarter Revenue Performance in Line with Expectations, GB200 Shipment Pace Normal\nPreliminary 2Q25 revenue increased by 9% quarter-over-quarter and 16% year-over-year, aligning with market expectations. JPM estimates that Foxconn shipped approximately 2,000 GB200 racks in 2Q25 and maintains its full-year expectation of around 10,000 units, believing that GB200 is the core driver of overall growth. Third-quarter revenue is expected to increase by approximately 7% quarter-over-quarter, slightly lower than the 18-20% growth seen in the same periods of 2023 and 2024. This is mainly due to the fading of pull-in effects, but the continuous ramp-up of GB200 and the launch of new mobile phones will provide some support.\n\nExchange Rate Factors Limit Gross Margin Improvement, But AI Shipment Outlook Supports Valuation\nThe NTD exchange rate remains a key factor influencing gross margin. JPM currently expects Foxconn's 2Q/3Q gross margins to improve by +0.3 percentage points and remain flat quarter-over-quarter, respectively. Although there are market doubts about the pace of GB200 rack shipments, and Foxconn's stock price has lagged behind the overall AI industry's rally compared to other ODM manufacturers, JPM maintains its \"Overweight\" rating on Foxconn given the improving GB200 yield and the company's leading position in the AI server market.\n\nAI Server Business Becomes Key to Mitigating Seasonal Weakness in H2 Demand\nJPM believes that as investors' expectations for 2026 AI demand strengthen, AI server-related companies are receiving more attention. Foxconn's AI-related revenue is expected to continue growing in the coming quarters, which will offset seasonal demand weakening and market uncertainties in the second half of the year, supporting overall revenue performance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.0075079665785011684, "alpha_c_m1d": -0.008002506395051512, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.008002506395051512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012422360248447228, "ret_signed_p1d": -0.012422360248447228, "alpha_spy_p1d": -0.01187464403070193, "alpha_c_p1d": -0.01779658953827923, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.005374229289832, "ret_p1w": -0.0031055900621117516, "ret_signed_p1w": -0.0031055900621117516, "alpha_spy_p1w": -0.009759504120746554, "alpha_c_p1w": -0.00592997226644254, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.0028243822043307887, "ret_p1m": 0.14596273291925477, "ret_signed_p1m": 0.14596273291925477, "alpha_spy_p1m": 0.1342176265400319, "alpha_c_p1m": 0.12556425673340343, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.020398476185851333, "ret_p3m": 0.4006211180124224, "ret_signed_p3m": 0.4006211180124224, "alpha_spy_p3m": 0.3194274199521192, "alpha_c_p3m": 0.27663022839543316, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.12399088961698923, "ret_p6m": 0.4161490683229814, "ret_signed_p6m": 0.4161490683229814, "alpha_spy_p6m": 0.30293385512074944, "alpha_c_p6m": 0.2721218028027472, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.14402726552023415, "price_path": [-0.3257, -0.2598, -0.1939, -0.1699, -0.1639, -0.1849, -0.1939, -0.1879, -0.1819, -0.2089, -0.1669, -0.1819, -0.1669, -0.1459, -0.1429, -0.1519, -0.1519, -0.116, -0.1459, -0.1249, -0.1309, -0.1249, -0.119, -0.083, -0.053, -0.0231, -0.044, -0.053, -0.077, -0.077, -0.071, -0.083, -0.077, -0.083, -0.089, -0.089, -0.065, -0.065, -0.086, -0.092, -0.065, -0.068, -0.083, -0.074, -0.068, -0.065, -0.056, -0.062, -0.071, -0.071, -0.065, -0.077, -0.068, -0.065, -0.038, -0.0291, -0.0231, -0.0111, -0.035, -0.0081, 0.0062, 0.0186, 0.0, 0.0, -0.0124, 0.0, 0.0031, 0.0031, -0.0031, 0.0124, 0.0093, 0.0186, 0.028, 0.0248, 0.0062, 0.0311, 0.0807, 0.0839, 0.0932, 0.0652, 0.0652, 0.1056, 0.1056, 0.118, 0.146, 0.1553, 0.2081, 0.2081, 0.2298, 0.2236, 0.2329, 0.2391, 0.2857, 0.3043, 0.2919, 0.2453, 0.2795, 0.2578, 0.2888, 0.2981, 0.2919, 0.2795, 0.264, 0.2329, 0.2329, 0.2453, 0.2609, 0.2733, 0.264, 0.2888, 0.3043, 0.3323, 0.3509, 0.3416, 0.3385, 0.3168, 0.3354, 0.3292, 0.3416, 0.3727, 0.3851, 0.4286, 0.3634, 0.3634, 0.3416, 0.3602, 0.4006, 0.4068, 0.4068, 0.4193, 0.3975, 0.3758, 0.3758, 0.323, 0.2795, 0.2826, 0.3851, 0.4068, 0.4814, 0.4845, 0.5062, 0.4845, 0.4845, 0.5311, 0.5528, 0.6149, 0.6273, 0.5994, 0.5621, 0.5155, 0.5342, 0.5404, 0.5155, 0.5497, 0.5311, 0.559, 0.5652, 0.4969, 0.4658, 0.4286, 0.4224, 0.4689, 0.3975, 0.3665, 0.3602, 0.4099, 0.4348, 0.4006, 0.3789, 0.3789, 0.4099, 0.4193, 0.4348, 0.441, 0.4596, 0.4503, 0.4037, 0.4099, 0.3758, 0.354, 0.3447, 0.3416, 0.3758, 0.3913, 0.3913, 0.3913, 0.3913, 0.4006, 0.4348, 0.4161]}
{"unit_id": "orig:1942728247462223907", "ticker_idx": 0, "ts": "2025-07-08T23:31:22+00:00", "symbol": "005930.KS", "kind": "equity", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "already lowered valuations and shareholder return policies are viewed positively", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Foreign brokerages broadly cautious on HBM competitiveness but see Q2 as earnings bottom; lowered valuations and shareholder returns viewed positively — net constructive on Samsung.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250709_Q2 2025 Samsung Electronics - Foreign Brokerage Comments\n\n[Key Takeaways]\n\n(1) Q2 2025 earnings are attributed to HBM revaluation losses and a sluggish Foundry business. \n(2) There is a common cautious stance regarding HBM business competitiveness.\n(3) However, already lowered valuations and shareholder return policies are viewed positively.\n\n[House Comments]\n(1) Goldman Sachs\n\nAdjusted HBM shipments and DRAM/NAND ASP estimates downwards, while smartphone margin estimates were revised upwards.\n\nLowered 2025(F) ~ 2027(F) EPS by -3.6%, -1.4%, and -0.9% respectively.\n\n(2) JPMorgan\n\nFocusing on HBM4 and the possibility of 1-c node entering Rubin GPU in the future. However, sample shipments are delayed, so market share gains will likely be limited for now.\n\nExpects HBM shipment growth of +12% YoY in '25 and +74% YoY in '26 driven by NVDA/AMD & AI ASIC volume growth.\n\nHowever, during the earnings call on July 31st, management is highly likely to lower the '25 HBM guidance and is expected to share detailed updates on: 1) Qual status, 2) DRAM capacity allocation strategy, and 3) Changes in packaging technology.\n\n(3) Nomura\n\nJudged that the impact of tariff factors was minimal in Q2, but potential impact on US exports and profitability depending on future tariff negotiations.\n\nQ3 2025(F) operating profit is expected to recover to around ₩9 trillion.\n\nROE is expected to bottom out at around 8% this year and gradually rise to 11% in '26 and 13% in '27 with HBM competitiveness recovery.\n\n(4) CLSA\n\nHBM order delays from AMD and Broadcom were reflected in Q2 2025, leading to stagnation.\n\nForecasts an earnings rebound with Q2 2025 as the bottom, and expects progress in the HBM business and recovery in the Foundry business in 2H25 and '26.\n\n(5) CITI\n\nQ2 2025 earnings likely marked a bottom, and increasing GDDR7 demand could drive DS division recovery. This is because GDDR adoption is expected to spread as an HBM alternative in AI Inference.\n\nHowever, reflecting the weaker-than-expected earnings, lowered '25 and '26 earnings by -7% and -8% respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004885940247018272, "ret_signed_m1d": 0.004885940247018272, "alpha_spy_m1d": 0.00433792387181664, "alpha_c_m1d": 0.010231441586980305, "bench_spy_m1d": 0.000548016375201632, "bench_c_m1d": -0.005345501339962033, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016286703429456373, "ret_signed_p1d": -0.016286703429456373, "alpha_spy_p1d": -0.022283330196667728, "alpha_c_p1d": -0.022295565333592227, "bench_spy_p1d": 0.005996626767211355, "bench_c_p1d": 0.006008861904135854, "ret_p1w": 0.037459282758823154, "ret_signed_p1w": 0.037459282758823154, "alpha_spy_p1w": 0.03455766355548784, "alpha_c_p1w": 0.031138258210178238, "bench_spy_p1w": 0.002901619203335315, "bench_c_p1w": 0.006321024548644916, "ret_p1m": 0.12052110347053557, "ret_signed_p1m": 0.12052110347053557, "alpha_spy_p1m": 0.10046765326937801, "alpha_c_p1m": 0.09414479469039039, "bench_spy_p1m": 0.020053450201157563, "bench_c_p1m": 0.026376308780145186, "ret_p3m": 0.4682479839729372, "ret_signed_p3m": 0.4682479839729372, "alpha_spy_p3m": 0.38647774919014033, "alpha_c_p3m": 0.3558912510545016, "bench_spy_p3m": 0.08177023478279688, "bench_c_p3m": 0.11235673291843562, "ret_p6m": 0.9710162609002393, "ret_signed_p6m": 0.9710162609002393, "alpha_spy_p6m": 0.8654432826420919, "alpha_c_p6m": 0.8443732150584198, "bench_spy_p6m": 0.10557297825814738, "bench_c_p6m": 0.1266430458418195, "price_path": [-0.087, -0.1065, -0.0903, -0.0838, -0.1146, -0.1081, -0.1048, -0.1032, -0.1097, -0.0984, -0.0984, -0.0984, -0.0967, -0.0967, -0.1016, -0.1016, -0.121, -0.121, -0.121, -0.1162, -0.1162, -0.1129, -0.0676, -0.0789, -0.0708, -0.0725, -0.0806, -0.0967, -0.0951, -0.0984, -0.1146, -0.1226, -0.1146, -0.1275, -0.0951, -0.0919, -0.0903, -0.0806, -0.0806, -0.0644, -0.0433, -0.0433, -0.032, -0.0417, -0.0304, -0.0369, -0.0563, -0.0741, -0.0595, -0.032, -0.0417, -0.0369, -0.0611, -0.0207, -0.0077, -0.0255, -0.0098, -0.0261, -0.0195, -0.0098, 0.0391, 0.0309, 0.0049, 0.0, -0.0163, -0.0065, 0.0195, 0.0179, 0.0375, 0.0537, 0.0863, 0.0928, 0.1042, 0.0749, 0.0814, 0.0749, 0.0733, 0.1466, 0.1498, 0.1824, 0.1629, 0.1221, 0.1352, 0.1384, 0.1205, 0.1482, 0.1694, 0.1564, 0.158, 0.171, 0.1661, 0.1661, 0.1401, 0.1401, 0.1482, 0.1498, 0.1629, 0.1645, 0.145, 0.1498, 0.1336, 0.1352, 0.101, 0.1254, 0.1368, 0.1417, 0.1319, 0.1417, 0.1645, 0.1824, 0.1954, 0.228, 0.2459, 0.2932, 0.2736, 0.3078, 0.3078, 0.3599, 0.3795, 0.3909, 0.4023, 0.3567, 0.3775, 0.3725, 0.4069, 0.4682, 0.4682, 0.4682, 0.4682, 0.4682, 0.4682, 0.5443, 0.5263, 0.4985, 0.5541, 0.5983, 0.6016, 0.6048, 0.595, 0.613, 0.5787, 0.6163, 0.6686, 0.6278, 0.6441, 0.703, 0.7586, 0.8175, 0.7161, 0.6457, 0.6228, 0.6016, 0.6457, 0.6932, 0.6866, 0.6817, 0.5901, 0.6457, 0.5999, 0.5787, 0.6457, 0.5509, 0.5819, 0.6245, 0.6817, 0.6932, 0.6441, 0.649, 0.6916, 0.7095, 0.7194, 0.7733, 0.7913, 0.7733, 0.7668, 0.7554, 0.7815, 0.7145, 0.6817, 0.7652, 0.7603, 0.739, 0.8077, 0.8241, 0.8175, 0.8175, 0.914, 0.9644, 0.971, 0.971]}
{"unit_id": "orig:1931865924925268398", "ticker_idx": 0, "ts": "2025-06-09T00:08:23+00:00", "symbol": "semiconductors", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Overweight Rating on Semiconductor Sector", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "SK Securities reiterates Overweight on semiconductors + SK Hynix as top pick; HBM4 cost structure drives industry tightness and structural supply constraint.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "Semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "[SK Securities – Donghee Han, Jemin Park, Semiconductor Sector]\n\n⸻\n\nHBM4 Price Premium Estimation\n\nHBM4 price premium expected at 30–40%\n\nAnalysis of Cost Structure Changes:\n\n1. Core Die Penalty: Cost increase of approximately 22%\n• TSV vias (power, signal, I/O) are expanded by more than 300%\n• → Die penalty estimated at 17–19%\n• Additional factors: Increase in metal layers due to RDL, micro bump, power and ground layout\n\n2. Introduction of Logic Die: Cost increase of 250–300%\n• Logic die introduced to improve power efficiency\n• Base die cost transitions from “memory process cost” to “external foundry pricing,” leading to unavoidable cost surge\n\n3. Cost Offset Factors:\n• Allocation of core die capacity from basic die + added value differential\n• Potential 8% (1/13) increase in core die production capacity\n• Core die provides higher added value than base die, which more than offsets the cost increase\n\n⸻\n\nMaintain Overweight Rating on Semiconductor Sector\n• HBM process complexity is rising sharply → continued decline in yield\n• High expected premium on HBM4 acts as a structural constraint on both HBM and commodity leading-edge supply\n→ Industry tightness expected to persist\n• Reaffirm strong fundamentals of the semiconductor sector\n\n⸻\n\nReaffirmation of Competitive Edge in HBM4\n→ Maintain SK Hynix as top pick", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016461626033810517, "ret_signed_m1d": -0.016461626033810517, "alpha_spy_m1d": -0.015561154725048043, "alpha_c_m1d": -0.015561154725048043, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.0009004713087624738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01988647532400134, "ret_signed_p1d": 0.01988647532400134, "alpha_spy_p1d": 0.014216810655110601, "alpha_c_p1d": 0.014216810655110601, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005669664668890739, "ret_p1w": 0.024984554344285703, "ret_signed_p1w": 0.024984554344285703, "alpha_spy_p1w": 0.019981878769881334, "alpha_c_p1w": 0.019981878769881334, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.005002675574404369, "ret_p1m": 0.10320688267472766, "ret_signed_p1m": 0.10320688267472766, "alpha_spy_p1m": 0.06569695607750026, "alpha_c_p1m": 0.06569695607750026, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.037509926597227405, "ret_p3m": 0.1280354912933679, "ret_signed_p3m": 0.1280354912933679, "alpha_spy_p3m": 0.042391579671533286, "alpha_c_p3m": 0.042391579671533286, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.0856439116218346, "ret_p6m": 0.39877811937317675, "ret_signed_p6m": 0.39877811937317675, "alpha_spy_p6m": 0.25576861598819933, "alpha_c_p6m": 0.25576861598819933, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.14300950338497742, "price_path": [-0.1418, -0.1456, -0.1183, -0.109, -0.1227, -0.115, -0.1164, -0.1253, -0.1025, -0.1063, -0.136, -0.1526, -0.1743, -0.177, -0.1731, -0.1668, -0.2389, -0.2964, -0.2804, -0.2997, -0.1795, -0.2364, -0.2166, -0.2154, -0.2103, -0.2437, -0.2507, -0.2507, -0.269, -0.2532, -0.2259, -0.1868, -0.1751, -0.1802, -0.1816, -0.1778, -0.1738, -0.1476, -0.1531, -0.161, -0.1438, -0.1362, -0.1306, -0.0761, -0.0444, -0.0349, -0.0383, -0.041, -0.0426, -0.0444, -0.0603, -0.0629, -0.076, -0.076, -0.0467, -0.0568, -0.0501, -0.067, -0.0532, -0.0318, -0.0203, -0.022, -0.0165, 0.0, 0.0199, 0.0191, 0.0241, 0.0001, 0.025, 0.018, 0.0219, 0.0219, 0.0129, 0.0193, 0.0567, 0.071, 0.0788, 0.0835, 0.0853, 0.0735, 0.0945, 0.1038, 0.1038, 0.0887, 0.1032, 0.1108, 0.1189, 0.1188, 0.1104, 0.1317, 0.1258, 0.1352, 0.1298, 0.1306, 0.1106, 0.1154, 0.12, 0.1188, 0.1341, 0.1399, 0.1529, 0.1238, 0.105, 0.1293, 0.1173, 0.1154, 0.1329, 0.1423, 0.1418, 0.1679, 0.1725, 0.175, 0.1506, 0.1553, 0.132, 0.1245, 0.1187, 0.1427, 0.1438, 0.1548, 0.1584, 0.1631, 0.1297, 0.1297, 0.1149, 0.1147, 0.128, 0.1413, 0.154, 0.159, 0.1707, 0.1828, 0.1843, 0.1954, 0.1956, 0.1879, 0.2335, 0.2286, 0.2538, 0.2519, 0.2496, 0.2495, 0.2524, 0.2557, 0.2701, 0.2986, 0.3163, 0.3102, 0.3363, 0.3117, 0.3468, 0.3434, 0.2661, 0.3222, 0.2979, 0.3301, 0.336, 0.3342, 0.3513, 0.3438, 0.3177, 0.3418, 0.3665, 0.4008, 0.4131, 0.4345, 0.4155, 0.4127, 0.4247, 0.3728, 0.3991, 0.3663, 0.3548, 0.3963, 0.3662, 0.3836, 0.3419, 0.3425, 0.3243, 0.2971, 0.321, 0.2652, 0.2692, 0.3197, 0.3232, 0.3532, 0.3532, 0.371, 0.3736, 0.3988]}
{"unit_id": "orig:1931865924925268398", "ticker_idx": 1, "ts": "2025-06-09T00:08:23+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain SK Hynix as top pick", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "SK Securities reiterates Overweight on semiconductors + SK Hynix as top pick; HBM4 cost structure drives industry tightness and structural supply constraint.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[SK Securities – Donghee Han, Jemin Park, Semiconductor Sector]\n\n⸻\n\nHBM4 Price Premium Estimation\n\nHBM4 price premium expected at 30–40%\n\nAnalysis of Cost Structure Changes:\n\n1. Core Die Penalty: Cost increase of approximately 22%\n• TSV vias (power, signal, I/O) are expanded by more than 300%\n• → Die penalty estimated at 17–19%\n• Additional factors: Increase in metal layers due to RDL, micro bump, power and ground layout\n\n2. Introduction of Logic Die: Cost increase of 250–300%\n• Logic die introduced to improve power efficiency\n• Base die cost transitions from “memory process cost” to “external foundry pricing,” leading to unavoidable cost surge\n\n3. Cost Offset Factors:\n• Allocation of core die capacity from basic die + added value differential\n• Potential 8% (1/13) increase in core die production capacity\n• Core die provides higher added value than base die, which more than offsets the cost increase\n\n⸻\n\nMaintain Overweight Rating on Semiconductor Sector\n• HBM process complexity is rising sharply → continued decline in yield\n• High expected premium on HBM4 acts as a structural constraint on both HBM and commodity leading-edge supply\n→ Industry tightness expected to persist\n• Reaffirm strong fundamentals of the semiconductor sector\n\n⸻\n\nReaffirmation of Competitive Edge in HBM4\n→ Maintain SK Hynix as top pick", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019650687484922047, "ret_signed_m1d": -0.019650687484922047, "alpha_spy_m1d": -0.018750216176159573, "alpha_c_m1d": -0.00318906145111153, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006550252000495194, "ret_signed_p1d": 0.006550252000495194, "alpha_spy_p1d": 0.0008805873316044543, "alpha_c_p1d": -0.013336223323506147, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.08296944757355784, "ret_signed_p1w": 0.08296944757355784, "alpha_spy_p1w": 0.07796677199915347, "alpha_c_p1w": 0.05798489322927214, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.23144107256924995, "ret_signed_p1m": 0.23144107256924995, "alpha_spy_p1m": 0.19393114597202255, "alpha_c_p1m": 0.1282341898945223, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.1610633249950959, "ret_signed_p3m": 0.1610633249950959, "alpha_spy_p3m": 0.0754194133732613, "alpha_c_p3m": 0.033027833701728015, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 1.441948416482865, "ret_signed_p6m": 1.441948416482865, "alpha_spy_p6m": 1.2989389130978877, "alpha_c_p6m": 1.0431702971096883, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.133, -0.1295, -0.1086, -0.1021, -0.1151, -0.1042, -0.0846, -0.0606, -0.0781, -0.0933, -0.0672, -0.0977, -0.1313, -0.1688, -0.1413, -0.1374, -0.1518, -0.2058, -0.2816, -0.2612, -0.2808, -0.2014, -0.2119, -0.2145, -0.2128, -0.2415, -0.2372, -0.2372, -0.2302, -0.2424, -0.211, -0.2228, -0.1962, -0.2067, -0.2119, -0.2263, -0.2263, -0.1892, -0.1892, -0.1892, -0.1683, -0.1705, -0.1714, -0.15, -0.1348, -0.1021, -0.126, -0.1086, -0.1308, -0.1195, -0.126, -0.1417, -0.1282, -0.1151, -0.1173, -0.0933, -0.0742, -0.107, -0.0939, -0.0939, -0.0502, -0.0197, -0.0197, 0.0, 0.0066, 0.048, 0.0284, 0.0284, 0.083, 0.0873, 0.0764, 0.0742, 0.1223, 0.1332, 0.2162, 0.2489, 0.2795, 0.2402, 0.2751, 0.2467, 0.2183, 0.2162, 0.1812, 0.1834, 0.2314, 0.2271, 0.2969, 0.286, 0.31, 0.3035, 0.2926, 0.1769, 0.1747, 0.19, 0.1725, 0.1747, 0.1769, 0.1616, 0.1441, 0.1463, 0.1507, 0.1943, 0.1266, 0.1266, 0.1507, 0.1288, 0.1441, 0.1201, 0.1659, 0.1747, 0.214, 0.2074, 0.2074, 0.1681, 0.1485, 0.1157, 0.0699, 0.0961, 0.1332, 0.1419, 0.1354, 0.1742, 0.1764, 0.1195, 0.1392, 0.1479, 0.1611, 0.196, 0.2114, 0.2595, 0.3294, 0.3425, 0.4366, 0.4475, 0.5218, 0.4584, 0.5546, 0.5546, 0.535, 0.5787, 0.5634, 0.559, 0.4716, 0.5262, 0.5197, 0.5743, 0.7296, 0.7296, 0.7296, 0.7296, 0.7296, 0.7296, 0.8717, 0.8148, 0.7995, 0.8476, 0.9788, 1.0357, 1.1231, 1.0947, 1.1057, 1.0925, 1.2303, 1.3396, 1.2784, 1.4402, 1.4839, 1.4446, 1.7113, 1.5626, 1.532, 1.5933, 1.5364, 1.6501, 1.707, 1.6982, 1.6763, 1.4489, 1.6501, 1.4927, 1.4577, 1.4971, 1.2784, 1.274, 1.2696, 1.2915, 1.3807, 1.3194, 1.3544, 1.4419]}
{"unit_id": "quote:1929882305046499820", "ticker_idx": 0, "ts": "2025-06-03T12:46:11+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's DRAM production bit growth, previously at 14% y/y, is likely to be revised upward again", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung ramping DRAM supply back up in 2H 2025 is bearish for MU via increased competition and pricing pressure.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Memory Semiconductor Supply: Weak First Half, Strong Second Half – DRAM Production Ramping Up Again (Samsung Policy Shift)\n\nSamsung's DRAM production bit growth, previously at 14% y/y, is likely to be revised upward again.\n\n- In the first half of the year, industry-wide supply growth forecasts were lowered due to Samsung’s restructuring (1bnm redesign, P4 fab delay, etc.),\n- However, Samsung's 2025 DRAM bit growth guidance, which had been adjusted from the low teens → to 5% → is now expected to be revised back up to around 8–10%.", "reply_text": null, "parent_text": null, "quoted_text": "Rumor: Samsung, with improved yields from redesigned 1bnm process and lowered HBM production targets, expected to ramp up DDR5/LPDDR5 output in the second half of 2025 to recover market share.\n\n$MU", "parent_author": null, "ret_m1d": -0.03980446232712087, "ret_signed_m1d": 0.03980446232712087, "alpha_spy_m1d": 0.034134207675207695, "alpha_c_m1d": 0.01773769466045161, "bench_spy_m1d": -0.005670254651913176, "bench_c_m1d": -0.022066767666669262, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009779950185895947, "ret_signed_p1d": -0.009779950185895947, "alpha_spy_p1d": -0.010048497588862393, "alpha_c_p1d": 0.002117645039931393, "bench_spy_p1d": -0.00026854740296644586, "bench_c_p1d": 0.01189759522582734, "ret_p1w": 0.11628356956551356, "ret_signed_p1w": -0.11628356956551356, "alpha_spy_p1w": -0.10455727581276708, "alpha_c_p1w": -0.06290518100261022, "bench_spy_p1w": 0.011726293752746475, "bench_c_p1w": 0.053378388562903334, "ret_p1m": 0.19061116106084097, "ret_signed_p1m": -0.19061116106084097, "alpha_spy_p1m": -0.14666777495461147, "alpha_c_p1m": -0.06013972688643898, "bench_spy_p1m": 0.0439433861062295, "bench_c_p1m": 0.130471434174402, "ret_p3m": 0.16500755528619848, "ret_signed_p3m": -0.16500755528619848, "alpha_spy_p3m": -0.07967327798060819, "alpha_c_p3m": 0.0017998859018075741, "bench_spy_p3m": 0.08533427730559029, "bench_c_p3m": 0.16680744118800606, "ret_p6m": 1.2554626891972234, "ret_signed_p6m": -1.2554626891972234, "alpha_spy_p6m": -1.1086908143645477, "alpha_c_p6m": -0.8578583272275695, "bench_spy_p6m": 0.14677187483267562, "bench_c_p6m": 0.3976043619696539, "price_path": [-0.1281, -0.092, -0.1495, -0.1302, -0.0659, -0.0733, -0.0156, 0.0071, -0.0065, -0.0032, 0.006, -0.0748, -0.0532, -0.0801, -0.1001, -0.1096, -0.1362, -0.1502, -0.1324, -0.1335, -0.273, -0.367, -0.3313, -0.359, -0.2384, -0.3149, -0.3198, -0.3054, -0.3052, -0.322, -0.3271, -0.3271, -0.3473, -0.3133, -0.2867, -0.2428, -0.2198, -0.2317, -0.2481, -0.2474, -0.2394, -0.2106, -0.2135, -0.2126, -0.192, -0.1672, -0.1603, -0.0974, -0.052, -0.0678, -0.0665, -0.0416, -0.0352, -0.0406, -0.0627, -0.0726, -0.0868, -0.0868, -0.0574, -0.0594, -0.0533, -0.0762, -0.0398, 0.0, 0.0098, 0.0395, 0.0617, 0.0851, 0.1163, 0.1348, 0.1362, 0.1306, 0.172, 0.1769, 0.1914, 0.1914, 0.2088, 0.1939, 0.251, 0.2445, 0.2323, 0.2201, 0.2054, 0.1823, 0.1906, 0.196, 0.196, 0.1739, 0.218, 0.1966, 0.2051, 0.219, 0.1611, 0.1758, 0.1398, 0.1087, 0.1198, 0.1084, 0.0692, 0.0751, 0.0937, 0.0891, 0.089, 0.096, 0.1232, 0.0684, 0.0267, 0.055, 0.0676, 0.0649, 0.0951, 0.1638, 0.2111, 0.2506, 0.2165, 0.2265, 0.1832, 0.2095, 0.1948, 0.1474, 0.1335, 0.152, 0.1397, 0.1404, 0.1527, 0.1943, 0.165, 0.165, 0.1598, 0.1622, 0.2159, 0.286, 0.2869, 0.3239, 0.3705, 0.474, 0.5391, 0.5444, 0.5547, 0.5662, 0.6533, 0.593, 0.6115, 0.629, 0.583, 0.5352, 0.5395, 0.6044, 0.6379, 0.7831, 0.7988, 0.8398, 0.8705, 0.8189, 0.9252, 0.8839, 0.7788, 0.8882, 0.8323, 0.8801, 0.9838, 0.9824, 1.0254, 0.9815, 0.9441, 1.0248, 1.1454, 1.1559, 1.1737, 1.2199, 1.1942, 1.1919, 1.299, 1.1357, 1.3264, 1.3345, 1.3305, 1.4811, 1.3617, 1.3989, 1.321, 1.4178, 1.37, 1.2382, 1.213, 0.9725, 1.0312, 1.1935, 1.1993, 1.2555]}
{"unit_id": "orig:1938191290447147364", "ticker_idx": 0, "ts": "2025-06-26T11:03:08+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we are raising SK Hynix's target price to KRW 330,000 (Target P/B raised from 1.9x to 2.0x, reflecting changes in profit forecasts)", "tweet_type": "analysis", "timeline": "months", "summary": "Daishin raises SK Hynix TP to KRW 330K; bullish Korean semis on HBM4 supply tightness, DRAM beat, ASIC upside, and lower cycle volatility.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Can we buy more Korean semiconductor stocks? - Korea Daishin Securities\n\nIt is too early to say that the rally in memory semiconductor stocks has ended. If Micron's DRAM sales exceeded expectations, I judge that SK Hynix's DRAM sales will be even better. Our firm expects SK Hynix's 2Q25 DRAM sales to reach the mid-20% range, surpassing the guidance (low 10% growth QoQ), and we anticipate that the industry trend will continue to move towards SK Hynix consistently approaching the forefront in terms of overall memory semiconductor revenue. If so, I believe there will be sufficient short-term catalysts to continue the stock rally.\n\nFrom a valuation perspective, I believe there is still upside potential as the two changes presented in our previous report (250618_Semiconductor: Reasons for further upside) are materializing.\n\n1. Expansion Opportunities in AI\n\nI judge that these remain valid. Given the tight supply-demand environment for HBM4, the bargaining power of suppliers will not easily disappear, and the possibility of increased orders in ASICs is gradually becoming visible. If the growth in volume expands, the absolute profit growth that can be achieved will also be significant, which can lead to changes highlighting the price attractiveness of the stock.\n\n2. Change in Course in the General-Purpose Market\n\nThe strategic change in direction among suppliers is materializing. Even if there is a short-term loss in sales, the industry is developing in a direction that maintains the gap with China through strategic changes in product lines, and strategic production control is being maintained to reduce recessionary periods where new entry is easy.\n\n- In terms of CapEx, it is being executed restrictively only for the most advanced products where strong demand is visible. The Capital Intensity (CapEx as a percentage of revenue) across the industry is judged to remain in the mid-to-high 20% range, lower than in past boom periods.\n\n- As the industry's response strategies become more flexible, I believe the volatility of the cycle will continue to decrease. I judge that the market will begin to assign higher value to memory semiconductor stocks as volatility decreases.\n\n- Our Conclusion As SK Hynix is leading AI technology from a customer-centric perspective, I believe it can receive a higher valuation than before. Although the stock price has risen rapidly, there are still changes to anticipate, and considering this, we are raising SK Hynix's target price to KRW 330,000 (Target P/B raised from 1.9x to 2.0x, reflecting changes in profit forecasts).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023890800700825943, "ret_signed_m1d": -0.023890800700825943, "alpha_spy_m1d": -0.016127757170079193, "alpha_c_m1d": -0.01667576177365171, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.030716774358483856, "ret_signed_p1d": -0.030716774358483856, "alpha_spy_p1d": -0.035685106122489096, "alpha_c_p1d": -0.035117933747531915, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": -0.04948804126557793, "ret_signed_p1w": -0.04948804126557793, "alpha_spy_p1w": -0.07150259577759732, "alpha_c_p1w": -0.07264821351741257, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": -0.09215032307545168, "ret_signed_p1m": -0.09215032307545168, "alpha_spy_p1m": -0.13338461973479476, "alpha_c_p1m": -0.12927144601217, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.2338615160807298, "ret_signed_p3m": 0.2338615160807298, "alpha_spy_p3m": 0.1469496117212168, "alpha_c_p3m": 0.07332759016880863, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.8709296139877303, "ret_signed_p6m": 0.8709296139877303, "alpha_spy_p6m": 0.7522384134919835, "alpha_c_p6m": 0.5858286202288445, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.3503, -0.3289, -0.3258, -0.337, -0.3793, -0.4386, -0.4225, -0.4379, -0.3759, -0.384, -0.3861, -0.3847, -0.4072, -0.4038, -0.4038, -0.3984, -0.4079, -0.3834, -0.3926, -0.3718, -0.38, -0.384, -0.3953, -0.3953, -0.3663, -0.3663, -0.3663, -0.35, -0.3517, -0.3524, -0.3357, -0.3237, -0.2982, -0.3169, -0.3033, -0.3207, -0.3118, -0.3169, -0.3292, -0.3186, -0.3084, -0.3101, -0.2914, -0.2765, -0.302, -0.2918, -0.2918, -0.2577, -0.2338, -0.2338, -0.2184, -0.2133, -0.1809, -0.1962, -0.1962, -0.1536, -0.1502, -0.1587, -0.1604, -0.1229, -0.1143, -0.0495, -0.0239, 0.0, -0.0307, -0.0034, -0.0256, -0.0478, -0.0495, -0.0768, -0.0751, -0.0375, -0.041, 0.0137, 0.0051, 0.0239, 0.0188, 0.0102, -0.0802, -0.0819, -0.07, -0.0836, -0.0819, -0.0802, -0.0922, -0.1058, -0.1041, -0.1007, -0.0666, -0.1195, -0.1195, -0.1007, -0.1177, -0.1058, -0.1246, -0.0887, -0.0819, -0.0512, -0.0563, -0.0563, -0.087, -0.1024, -0.128, -0.1638, -0.1433, -0.1143, -0.1075, -0.1126, -0.0823, -0.0806, -0.125, -0.1096, -0.1028, -0.0925, -0.0652, -0.0532, -0.0156, 0.039, 0.0493, 0.1228, 0.1313, 0.1894, 0.1399, 0.2151, 0.2151, 0.1997, 0.2339, 0.2219, 0.2185, 0.1501, 0.1928, 0.1877, 0.2304, 0.3518, 0.3518, 0.3518, 0.3518, 0.3518, 0.3518, 0.4629, 0.4184, 0.4065, 0.4441, 0.5466, 0.591, 0.6594, 0.6372, 0.6457, 0.6355, 0.7431, 0.8286, 0.7807, 0.9072, 0.9414, 0.9106, 1.1191, 1.0029, 0.979, 1.0268, 0.9824, 1.0712, 1.1157, 1.1088, 1.0918, 0.914, 1.0712, 0.9482, 0.9209, 0.9516, 0.7807, 0.7773, 0.7739, 0.791, 0.8607, 0.8128, 0.8401, 0.9086, 0.888, 0.8538, 0.8607, 0.9735, 0.9359, 1.0077, 0.9325, 0.953, 0.8949, 0.8128, 0.8846, 0.888, 0.8709]}
{"unit_id": "orig:1938191290447147364", "ticker_idx": 1, "ts": "2025-06-26T11:03:08+00:00", "symbol": "Korean semiconductors", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Can we buy more Korean semiconductor stocks?", "tweet_type": "analysis", "timeline": "months", "summary": "Daishin raises SK Hynix TP to KRW 330K; bullish Korean semis on HBM4 supply tightness, DRAM beat, ASIC upside, and lower cycle volatility.", "resolved_tickers": ["EWY", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korean semis: EWY ETF + Samsung + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Can we buy more Korean semiconductor stocks? - Korea Daishin Securities\n\nIt is too early to say that the rally in memory semiconductor stocks has ended. If Micron's DRAM sales exceeded expectations, I judge that SK Hynix's DRAM sales will be even better. Our firm expects SK Hynix's 2Q25 DRAM sales to reach the mid-20% range, surpassing the guidance (low 10% growth QoQ), and we anticipate that the industry trend will continue to move towards SK Hynix consistently approaching the forefront in terms of overall memory semiconductor revenue. If so, I believe there will be sufficient short-term catalysts to continue the stock rally.\n\nFrom a valuation perspective, I believe there is still upside potential as the two changes presented in our previous report (250618_Semiconductor: Reasons for further upside) are materializing.\n\n1. Expansion Opportunities in AI\n\nI judge that these remain valid. Given the tight supply-demand environment for HBM4, the bargaining power of suppliers will not easily disappear, and the possibility of increased orders in ASICs is gradually becoming visible. If the growth in volume expands, the absolute profit growth that can be achieved will also be significant, which can lead to changes highlighting the price attractiveness of the stock.\n\n2. Change in Course in the General-Purpose Market\n\nThe strategic change in direction among suppliers is materializing. Even if there is a short-term loss in sales, the industry is developing in a direction that maintains the gap with China through strategic changes in product lines, and strategic production control is being maintained to reduce recessionary periods where new entry is easy.\n\n- In terms of CapEx, it is being executed restrictively only for the most advanced products where strong demand is visible. The Capital Intensity (CapEx as a percentage of revenue) across the industry is judged to remain in the mid-to-high 20% range, lower than in past boom periods.\n\n- As the industry's response strategies become more flexible, I believe the volatility of the cycle will continue to decrease. I judge that the market will begin to assign higher value to memory semiconductor stocks as volatility decreases.\n\n- Our Conclusion As SK Hynix is leading AI technology from a customer-centric perspective, I believe it can receive a higher valuation than before. Although the stock price has risen rapidly, there are still changes to anticipate, and considering this, we are raising SK Hynix's target price to KRW 330,000 (Target P/B raised from 1.9x to 2.0x, reflecting changes in profit forecasts).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00011960522586492346, "ret_signed_m1d": 0.00011960522586492346, "alpha_spy_m1d": 0.007882648756611673, "alpha_c_m1d": 0.006970274881920664, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.006850669656055741, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010411935832070266, "ret_signed_p1d": -0.010411935832070266, "alpha_spy_p1d": -0.015380267596075506, "alpha_c_p1d": -0.009217079262150133, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": -0.001194856569920133, "ret_p1w": 0.009125804065709583, "ret_signed_p1w": 0.009125804065709583, "alpha_spy_p1w": -0.012888750446309808, "alpha_c_p1w": -0.014333805162920524, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023459609228630107, "ret_p1m": 0.008967405061862147, "ret_signed_p1m": 0.008967405061862147, "alpha_spy_p1m": -0.03226689159748094, "alpha_c_p1m": -0.03452626499308938, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.04349367005495153, "ret_p3m": 0.2605538856678893, "ret_signed_p3m": 0.2605538856678893, "alpha_spy_p3m": 0.17364198130837627, "alpha_c_p3m": 0.14301645326643753, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.11753743240145176, "ret_p6m": 0.6512179760392051, "ret_signed_p6m": 0.6512179760392051, "alpha_spy_p6m": 0.5325267755434583, "alpha_c_p6m": 0.497867553915741, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.15335042212346406, "price_path": [-0.213, -0.1966, -0.1966, -0.2139, -0.245, -0.283, -0.2846, -0.2723, -0.2396, -0.2374, -0.2323, -0.2295, -0.249, -0.2445, -0.2434, -0.2407, -0.2441, -0.2284, -0.2299, -0.2236, -0.2265, -0.2258, -0.2308, -0.2335, -0.2207, -0.2176, -0.2171, -0.2132, -0.2165, -0.2134, -0.1907, -0.1902, -0.1767, -0.1826, -0.1809, -0.192, -0.1905, -0.1901, -0.2034, -0.1996, -0.1934, -0.1937, -0.1728, -0.1618, -0.1752, -0.1634, -0.1627, -0.1361, -0.1145, -0.1135, -0.1003, -0.1034, -0.0849, -0.089, -0.1015, -0.0852, -0.0863, -0.074, -0.0779, -0.0606, -0.062, -0.0115, 0.0001, 0.0, -0.0104, -0.0021, -0.0079, -0.0105, 0.0091, -0.0028, -0.0233, -0.0031, -0.0098, 0.0166, 0.02, 0.0287, 0.033, 0.0358, 0.0167, 0.0153, 0.03, 0.0109, 0.0181, 0.014, 0.009, 0.0272, 0.0325, 0.0456, 0.0463, 0.0047, 0.017, 0.0269, 0.0165, 0.0311, 0.0324, 0.0377, 0.0455, 0.063, 0.0548, 0.0535, 0.0316, 0.0202, 0.0141, 0.002, 0.0239, 0.0305, 0.0246, 0.0249, 0.0349, 0.0295, 0.003, 0.015, 0.0258, 0.0316, 0.04, 0.0499, 0.0736, 0.106, 0.1194, 0.1588, 0.1702, 0.2127, 0.1855, 0.2265, 0.2233, 0.243, 0.2606, 0.2539, 0.256, 0.2129, 0.2406, 0.2365, 0.2708, 0.3399, 0.3413, 0.3457, 0.339, 0.3429, 0.3394, 0.3899, 0.3824, 0.3646, 0.4076, 0.4694, 0.4883, 0.5211, 0.502, 0.5148, 0.5025, 0.5604, 0.6173, 0.588, 0.6471, 0.671, 0.6879, 0.795, 0.6983, 0.6703, 0.664, 0.634, 0.6967, 0.7259, 0.7243, 0.7084, 0.6181, 0.6844, 0.6211, 0.5999, 0.6237, 0.5375, 0.552, 0.563, 0.5964, 0.6236, 0.5877, 0.5973, 0.6414, 0.645, 0.6318, 0.6638, 0.7111, 0.6947, 0.7219, 0.6846, 0.6898, 0.6477, 0.6036, 0.6517, 0.6582, 0.6512]}
{"unit_id": "orig:1935312303462715490", "ticker_idx": 0, "ts": "2025-06-18T12:23:04+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "maintaining an 'Overweight' rating for the company... 2028 EPS is expected to reach $8-9", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/endorses JPM Overweight on MRVL; TAM revised up to $94B by 2028, 20% AI ASIC share target, EPS $8-9.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell: AI ASIC Platform Continues to Expand, Targeting 20% Data Center Market Share by 2028\nJPM (H. Sur, 2025/06/18)\n\nMarvell, leveraging its end-to-end custom chip development platform, strong customer base, and technological accumulation, is poised to continuously expand its market share in the AI ASIC field and achieve significant EPS growth before 2028.\n\n• Data Center and AI ASIC Market Size Revised Upward, Marvell's Target Share Increased to 20%\n\nThe overall data center TAM (Total Addressable Market) has been revised upward to $94 billion by 2028 (previously $75 billion), with the AI Custom Chip (ASIC) TAM revised upward to $54 billion (previously $43 billion). The diversification of AI workloads and the increasing demand for customized XPU (Accelerated Processing Unit) are the main drivers. Opportunities related to customized XPU (including SmartNIC, DPU, security acceleration, co-processors, etc.) collectively amount to a TAM of $15 billion. The institution believes that if Marvell successfully achieves its target of 20% AI data center market share, its 2028 EPS is expected to reach $8-9, maintaining an \"Overweight\" rating for the company.\n\n• Secured Multiple Generations of ASIC Design Wins with Three Major Cloud Vendors and Added Two New Clients, Covering Five Major Customers in Total\n\nMarvell has achieved multi-generational AI ASIC design wins with three hyperscale clients: Amazon, Google, and Microsoft, covering 18 multi-generational product slots. Additionally, it has added two new AI XPU ASIC clients, bringing the total number of major clients to five. Some design wins have extended to the 2nm node (expected to include Amazon Trainium 4 and Microsoft Maia Gen 3), which is anticipated to help drive its market share to 20% by 2028. The current project pipeline still has 50+ opportunities and 10+ clients, with a total potential revenue of $75 billion.\n\n• Full-Process Custom Chip Platform Offers Significant Advantages, Technology Covers 2nm to 14a Process and Multi-Generational Packaging Capabilities\n\nMarvell has built a strong IP portfolio and delivery capability through its leading technologies in system architecture, high-speed SerDes (up to 448G), SRAM, high-bandwidth memory interfaces, 2.5D/3D/4D packaging, silicon photonics, and co-packaged optics. To date, the company has 20 products deployed at 5nm/3nm nodes, is progressing with 2nm chip testing, and is initiating 16a/14a process development. This platform capability helps clients accelerate time-to-market and deepens collaboration with cloud vendors.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06617742597775589, "ret_signed_m1d": -0.06617742597775589, "alpha_spy_m1d": -0.06632808442647598, "alpha_c_m1d": -0.06236924915785513, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.013075462479225308, "ret_signed_p1w": 0.013075462479225308, "alpha_spy_p1w": -0.0061311130904015965, "alpha_c_p1w": -0.034946199805228506, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": -0.038439644806319784, "ret_signed_p1m": -0.038439644806319784, "alpha_spy_p1m": -0.09276575876176651, "alpha_c_p1m": -0.14933492892037248, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": -0.09959714653714336, "ret_signed_p3m": -0.09959714653714336, "alpha_spy_p3m": -0.20910393231560576, "alpha_c_p3m": -0.26936761837288603, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.19496294868851893, "ret_signed_p6m": 0.19496294868851893, "alpha_spy_p6m": 0.03480692109508299, "alpha_c_p6m": -0.21746717081217692, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [-0.0619, -0.0303, -0.0448, -0.1103, -0.1356, -0.1732, -0.1795, -0.1648, -0.1573, -0.2585, -0.3412, -0.3203, -0.3332, -0.1876, -0.2954, -0.2877, -0.3027, -0.2886, -0.3071, -0.3102, -0.3102, -0.3412, -0.3246, -0.2826, -0.2351, -0.2139, -0.2168, -0.2171, -0.2212, -0.1868, -0.1684, -0.173, -0.1832, -0.2487, -0.2311, -0.2041, -0.1394, -0.1264, -0.1207, -0.1301, -0.1493, -0.1653, -0.1805, -0.198, -0.1748, -0.1903, -0.1903, -0.1485, -0.1382, -0.1497, -0.1969, -0.1799, -0.168, -0.1154, -0.1306, -0.0881, -0.0775, -0.0815, -0.0895, -0.0708, -0.1035, -0.0604, -0.0662, 0.0, 0.0, -0.0192, -0.0556, 0.0035, 0.0131, 0.067, 0.0295, 0.0327, 0.0172, -0.0093, 0.0031, 0.0031, -0.0454, -0.04, -0.0359, -0.0212, -0.0291, -0.0318, -0.0331, -0.0539, -0.0384, -0.0032, -0.0244, -0.0387, -0.0216, -0.0113, -0.0091, 0.0136, 0.0194, 0.0915, 0.0732, -0.0059, 0.0219, 0.0233, 0.0058, 0.0128, 0.0327, 0.0319, 0.039, 0.0592, 0.0554, 0.0174, 0.0247, -0.0376, -0.049, -0.0491, -0.0252, -0.0259, -0.0084, -0.0013, 0.0313, -0.1605, -0.1605, -0.1374, -0.168, -0.1441, -0.1543, -0.1187, -0.1075, -0.104, -0.1108, -0.1007, -0.0996, -0.0805, -0.0522, -0.0088, -0.0084, 0.0086, -0.0036, 0.0695, 0.1191, 0.1106, 0.1002, 0.1226, 0.1202, 0.151, 0.1513, 0.1874, 0.1613, 0.2352, 0.2109, 0.1439, 0.1944, 0.1521, 0.1877, 0.1789, 0.1752, 0.147, 0.1259, 0.083, 0.106, 0.1241, 0.1853, 0.182, 0.2046, 0.1835, 0.2526, 0.2075, 0.1704, 0.2413, 0.2471, 0.2149, 0.2457, 0.1936, 0.1936, 0.1694, 0.1551, 0.1151, 0.0513, 0.0866, 0.0246, 0.0349, 0.1196, 0.1148, 0.1721, 0.1721, 0.1946, 0.2173, 0.2412, 0.3389, 0.312, 0.3216, 0.2293, 0.1879, 0.2356, 0.195]}
{"unit_id": "orig:1943316150512947225", "ticker_idx": 0, "ts": "2025-07-10T14:27:30+00:00", "symbol": "CSCO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We recommend prioritizing related targets such as CSCO, ANET, and FFIV", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley Buy recommendation on CSCO, ANET, FFIV, HPE as AI/network equipment spending remains stable and AI tops CIO priorities.", "resolved_tickers": ["CSCO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "US Optical Modules: Network Equipment Spending Relatively Stable, AI Remains CIO's Top Investment Priority\n\nMorgan Stanley (M. Marshall, July 08, 2025)\n\nMorgan Stanley's Q2 2025 CIO survey indicates stable network equipment spending and continued strong enthusiasm for AI investment, despite persistent macroeconomic and tariff uncertainties. Compared to other sub-sectors like communications software, network equipment demonstrates greater defensiveness. We recommend prioritizing related targets such as CSCO, ANET, and FFIV.\n\nNetwork Equipment Spending Expectations Stable, Ranking Slightly Down\nThe communications sector's 2025 growth forecast is 2.9%, a 7bps decrease from the previous quarter, but still 13bps higher than 2024. Network equipment's priority in CIO spending shifted from 7th to 8th, but spending resilience remains strong, making it one of the few sectors whose expectations have not been downgraded since early 2025.\n\nAI Continues to Be CIO's Top Priority, Budget Source Leaning Towards New Investment\nAI remains the top priority for CIOs in 2025, with 60% of CIOs expecting to enter the production phase in 2025 or earlier. 56% of AI spending comes from new budgets, a higher proportion than the 41% in Q1 2025, indicating enterprises' willingness for incremental investment in AI. Related beneficiaries include CSCO, ANET, FFIV, HPE, etc.\n\nPublic Cloud Penetration Continues to Increase, Equipment Upgrade Demand Rises\nThe Q2 2025 survey shows enterprises expect 68% of workloads to migrate to public clouds by the end of 2027, up from the current 44% and 50% by the end of 2025. With clearer distinctions drawn between cloud and on-premise deployments, some enterprises are beginning to refresh on-premise equipment that hasn't been replaced in years, benefiting demand for network-related equipment.\n\nCommunications Software Shows Some Improvement But Remains Under Pressure\nCommunications software spending remains less defensive but has shown marginal improvement compared to the last quarter, especially in collaboration and customer service software. Spending intention for collaboration software slightly rose to the edge of the Top 10, and customer service software's defensiveness climbed from 25th in Q1 to 10th, indicating the emergence of structural highlights.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0074170307482666065, "ret_signed_m1d": 0.0074170307482666065, "alpha_spy_m1d": 0.010229409186960803, "alpha_c_m1d": 0.010229409186960803, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.0028123784386941963, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011780142667407345, "ret_signed_p1d": -0.011780142667407345, "alpha_spy_p1d": -0.008264718796811787, "alpha_c_p1d": -0.008264718796811787, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.0035154238705955576, "ret_p1w": -0.0066899394553201486, "ret_signed_p1w": -0.0066899394553201486, "alpha_spy_p1w": -0.01023723052232528, "alpha_c_p1w": -0.01023723052232528, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.0035472910670051316, "ret_p1m": 0.044066313343625874, "ret_signed_p1m": 0.044066313343625874, "alpha_spy_p1m": 0.02591420747765194, "alpha_c_p1m": 0.02591420747765194, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.018152105865973933, "ret_p3m": 0.009403443257937738, "ret_signed_p3m": 0.009403443257937738, "alpha_spy_p3m": -0.06274998745722016, "alpha_c_p3m": -0.06274998745722016, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.0721534307151579, "ret_p6m": 0.1185065156699796, "ret_signed_p6m": 0.1185065156699796, "alpha_spy_p6m": 0.02060566887791948, "alpha_c_p6m": 0.02060566887791948, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.09790084679206013, "price_path": [-0.1701, -0.1731, -0.1939, -0.1939, -0.1939, -0.214, -0.2043, -0.1958, -0.1862, -0.1802, -0.1783, -0.1711, -0.1654, -0.1598, -0.1423, -0.1424, -0.1439, -0.1388, -0.1368, -0.1359, -0.1085, -0.1069, -0.114, -0.071, -0.0803, -0.0759, -0.0832, -0.0864, -0.084, -0.0877, -0.0877, -0.0784, -0.0843, -0.0885, -0.0887, -0.077, -0.0696, -0.0692, -0.0658, -0.045, -0.0473, -0.0582, -0.072, -0.0589, -0.0735, -0.053, -0.0556, -0.0482, -0.0482, -0.0413, -0.0259, -0.0186, -0.0142, -0.0032, -0.0076, 0.003, -0.0011, -0.0084, 0.0089, 0.0089, 0.0025, -0.0025, 0.0074, 0.0, -0.0118, -0.0137, -0.023, -0.0202, -0.0067, -0.0103, -0.0061, -0.0087, -0.002, -0.0068, -0.001, -0.0122, -0.0116, -0.007, -0.0099, -0.024, -0.0039, -0.018, 0.0065, 0.0166, 0.0441, 0.0278, 0.0381, 0.0239, 0.0079, -0.0372, -0.0263, -0.0291, -0.024, -0.0253, -0.0209, -0.0236, -0.0054, -0.0047, 0.0097, 0.0048, 0.0048, -0.014, -0.0159, -0.0112, -0.0271, -0.0273, -0.0207, -0.0092, -0.0156, -0.0324, -0.0253, -0.0266, -0.0151, -0.0012, -0.008, -0.0147, -0.0177, -0.0209, -0.0132, -0.0224, -0.0151, -0.0049, -0.0001, -0.0065, -0.0063, 0.0082, 0.0094, 0.029, 0.0236, -0.006, -0.013, 0.0046, 0.0172, 0.0109, 0.0261, 0.0337, 0.0347, 0.0338, 0.0281, 0.0334, 0.0445, 0.0625, 0.0436, 0.0668, 0.0697, 0.0893, 0.0581, 0.0549, 0.0394, 0.0398, 0.0548, 0.0492, 0.0821, 0.1322, 0.1412, 0.138, 0.132, 0.1469, 0.1038, 0.1134, 0.1155, 0.1166, 0.113, 0.113, 0.1257, 0.1126, 0.1247, 0.1376, 0.1377, 0.1408, 0.1538, 0.1633, 0.1742, 0.1598, 0.1383, 0.1449, 0.1346, 0.112, 0.1259, 0.1474, 0.1428, 0.1415, 0.1415, 0.1415, 0.1436, 0.1382, 0.1326, 0.127, 0.127, 0.1185]}
{"unit_id": "orig:1943316150512947225", "ticker_idx": 1, "ts": "2025-07-10T14:27:30+00:00", "symbol": "ANET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We recommend prioritizing related targets such as CSCO, ANET, and FFIV", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley Buy recommendation on CSCO, ANET, FFIV, HPE as AI/network equipment spending remains stable and AI tops CIO priorities.", "resolved_tickers": ["ANET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "US Optical Modules: Network Equipment Spending Relatively Stable, AI Remains CIO's Top Investment Priority\n\nMorgan Stanley (M. Marshall, July 08, 2025)\n\nMorgan Stanley's Q2 2025 CIO survey indicates stable network equipment spending and continued strong enthusiasm for AI investment, despite persistent macroeconomic and tariff uncertainties. Compared to other sub-sectors like communications software, network equipment demonstrates greater defensiveness. We recommend prioritizing related targets such as CSCO, ANET, and FFIV.\n\nNetwork Equipment Spending Expectations Stable, Ranking Slightly Down\nThe communications sector's 2025 growth forecast is 2.9%, a 7bps decrease from the previous quarter, but still 13bps higher than 2024. Network equipment's priority in CIO spending shifted from 7th to 8th, but spending resilience remains strong, making it one of the few sectors whose expectations have not been downgraded since early 2025.\n\nAI Continues to Be CIO's Top Priority, Budget Source Leaning Towards New Investment\nAI remains the top priority for CIOs in 2025, with 60% of CIOs expecting to enter the production phase in 2025 or earlier. 56% of AI spending comes from new budgets, a higher proportion than the 41% in Q1 2025, indicating enterprises' willingness for incremental investment in AI. Related beneficiaries include CSCO, ANET, FFIV, HPE, etc.\n\nPublic Cloud Penetration Continues to Increase, Equipment Upgrade Demand Rises\nThe Q2 2025 survey shows enterprises expect 68% of workloads to migrate to public clouds by the end of 2027, up from the current 44% and 50% by the end of 2025. With clearer distinctions drawn between cloud and on-premise deployments, some enterprises are beginning to refresh on-premise equipment that hasn't been replaced in years, benefiting demand for network-related equipment.\n\nCommunications Software Shows Some Improvement But Remains Under Pressure\nCommunications software spending remains less defensive but has shown marginal improvement compared to the last quarter, especially in collaboration and customer service software. Spending intention for collaboration software slightly rose to the edge of the Top 10, and customer service software's defensiveness climbed from 25th in Q1 to 10th, indicating the emergence of structural highlights.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -9.410232518880512e-05, "ret_signed_m1d": -9.410232518880512e-05, "alpha_spy_m1d": 0.0027182761135053912, "alpha_c_m1d": -0.0034017608295544566, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": 0.0033076585043656515, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021450736284299055, "ret_signed_p1d": 0.021450736284299055, "alpha_spy_p1d": 0.024966160154894612, "alpha_c_p1d": 0.025848030420399626, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.004397294136100571, "ret_p1w": 0.05353281016460154, "ret_signed_p1w": 0.05353281016460154, "alpha_spy_p1w": 0.04998551909759641, "alpha_c_p1w": 0.03765627589151577, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.015876534273085774, "ret_p1m": 0.3094363672683833, "ret_signed_p1m": 0.3094363672683833, "alpha_spy_p1m": 0.2912842614024094, "alpha_c_p1m": 0.27464761689629325, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.03478875037209006, "ret_p3m": 0.36692061374239926, "ret_signed_p3m": 0.36692061374239926, "alpha_spy_p3m": 0.29476718302724136, "alpha_c_p3m": 0.2538102110636973, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.11311040267870198, "ret_p6m": 0.25693861090181547, "ret_signed_p6m": 0.25693861090181547, "alpha_spy_p6m": 0.15903776410975534, "alpha_c_p6m": 0.13074518802762292, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.12619342287419255, "price_path": [-0.3076, -0.3113, -0.3238, -0.3301, -0.3301, -0.3633, -0.3539, -0.3343, -0.2917, -0.267, -0.2632, -0.2405, -0.226, -0.1736, -0.1437, -0.1497, -0.146, -0.1867, -0.1789, -0.186, -0.1314, -0.0851, -0.0759, -0.0962, -0.0929, -0.0902, -0.0999, -0.1304, -0.1292, -0.142, -0.142, -0.1298, -0.127, -0.1874, -0.1849, -0.1553, -0.1109, -0.1068, -0.1045, -0.0851, -0.0893, -0.1184, -0.1137, -0.099, -0.1312, -0.1054, -0.1563, -0.151, -0.151, -0.1885, -0.1349, -0.1065, -0.0939, -0.0442, -0.0649, -0.0374, -0.0694, -0.0485, -0.0355, -0.0355, -0.0453, -0.0273, -0.0001, 0.0, 0.0215, 0.0196, 0.0102, 0.0189, 0.0535, 0.0517, 0.0501, 0.0328, 0.0635, 0.0729, 0.0752, 0.1059, 0.116, 0.1486, 0.1593, 0.1061, 0.1323, 0.1113, 0.3057, 0.3104, 0.3094, 0.295, 0.3289, 0.2984, 0.284, 0.2917, 0.2987, 0.2492, 0.2369, 0.2422, 0.2536, 0.2517, 0.2632, 0.2538, 0.2817, 0.2847, 0.2847, 0.2783, 0.2925, 0.3282, 0.344, 0.3172, 0.3351, 0.418, 0.4398, 0.3114, 0.3682, 0.3375, 0.3439, 0.3798, 0.4076, 0.368, 0.3556, 0.342, 0.3459, 0.3407, 0.3489, 0.3709, 0.4044, 0.3591, 0.3689, 0.4065, 0.3669, 0.4805, 0.4887, 0.4498, 0.3872, 0.3058, 0.349, 0.3737, 0.3463, 0.3781, 0.373, 0.3792, 0.4372, 0.4472, 0.4753, 0.4749, 0.5244, 0.4906, 0.4836, 0.4826, 0.4446, 0.3211, 0.2609, 0.2668, 0.2914, 0.2695, 0.2699, 0.2259, 0.236, 0.1973, 0.1614, 0.1742, 0.1251, 0.1048, 0.1494, 0.1764, 0.201, 0.201, 0.2295, 0.2053, 0.1969, 0.2024, 0.2094, 0.2098, 0.2147, 0.2234, 0.2453, 0.2644, 0.1738, 0.1844, 0.1867, 0.1512, 0.1725, 0.2336, 0.2299, 0.2355, 0.2303, 0.2303, 0.2404, 0.2621, 0.246, 0.2328, 0.2328, 0.2569]}
{"unit_id": "orig:1943316150512947225", "ticker_idx": 2, "ts": "2025-07-10T14:27:30+00:00", "symbol": "FFIV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We recommend prioritizing related targets such as CSCO, ANET, and FFIV", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley Buy recommendation on CSCO, ANET, FFIV, HPE as AI/network equipment spending remains stable and AI tops CIO priorities.", "resolved_tickers": ["FFIV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "US Optical Modules: Network Equipment Spending Relatively Stable, AI Remains CIO's Top Investment Priority\n\nMorgan Stanley (M. Marshall, July 08, 2025)\n\nMorgan Stanley's Q2 2025 CIO survey indicates stable network equipment spending and continued strong enthusiasm for AI investment, despite persistent macroeconomic and tariff uncertainties. Compared to other sub-sectors like communications software, network equipment demonstrates greater defensiveness. We recommend prioritizing related targets such as CSCO, ANET, and FFIV.\n\nNetwork Equipment Spending Expectations Stable, Ranking Slightly Down\nThe communications sector's 2025 growth forecast is 2.9%, a 7bps decrease from the previous quarter, but still 13bps higher than 2024. Network equipment's priority in CIO spending shifted from 7th to 8th, but spending resilience remains strong, making it one of the few sectors whose expectations have not been downgraded since early 2025.\n\nAI Continues to Be CIO's Top Priority, Budget Source Leaning Towards New Investment\nAI remains the top priority for CIOs in 2025, with 60% of CIOs expecting to enter the production phase in 2025 or earlier. 56% of AI spending comes from new budgets, a higher proportion than the 41% in Q1 2025, indicating enterprises' willingness for incremental investment in AI. Related beneficiaries include CSCO, ANET, FFIV, HPE, etc.\n\nPublic Cloud Penetration Continues to Increase, Equipment Upgrade Demand Rises\nThe Q2 2025 survey shows enterprises expect 68% of workloads to migrate to public clouds by the end of 2027, up from the current 44% and 50% by the end of 2025. With clearer distinctions drawn between cloud and on-premise deployments, some enterprises are beginning to refresh on-premise equipment that hasn't been replaced in years, benefiting demand for network-related equipment.\n\nCommunications Software Shows Some Improvement But Remains Under Pressure\nCommunications software spending remains less defensive but has shown marginal improvement compared to the last quarter, especially in collaboration and customer service software. Spending intention for collaboration software slightly rose to the edge of the Top 10, and customer service software's defensiveness climbed from 25th in Q1 to 10th, indicating the emergence of structural highlights.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007969535543426387, "ret_signed_m1d": 0.007969535543426387, "alpha_spy_m1d": 0.010781913982120583, "alpha_c_m1d": 0.010781913982120583, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.0028123784386941963, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026676446727906034, "ret_signed_p1d": -0.026676446727906034, "alpha_spy_p1d": -0.023161022857310476, "alpha_c_p1d": -0.023161022857310476, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.0035154238705955576, "ret_p1w": 0.00150048907090361, "ret_signed_p1w": 0.00150048907090361, "alpha_spy_p1w": -0.0020468019961015216, "alpha_c_p1w": -0.0020468019961015216, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.0035472910670051316, "ret_p1m": 0.07295999632026562, "ret_signed_p1m": 0.07295999632026562, "alpha_spy_p1m": 0.054807890454291686, "alpha_c_p1m": 0.054807890454291686, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.018152105865973933, "ret_p3m": 0.10293772252929645, "ret_signed_p3m": 0.10293772252929645, "alpha_spy_p3m": 0.030784291814138554, "alpha_c_p3m": 0.030784291814138554, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.0721534307151579, "ret_p6m": -0.14425291824517372, "ret_signed_p6m": -0.14425291824517372, "alpha_spy_p6m": -0.24215376503723385, "alpha_c_p6m": -0.24215376503723385, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.09790084679206013, "price_path": [-0.124, -0.1212, -0.1364, -0.1377, -0.1377, -0.1637, -0.1505, -0.1248, -0.1058, -0.0996, -0.1161, -0.1239, -0.1172, -0.1138, -0.1036, -0.109, -0.1169, -0.1081, -0.0892, -0.0994, -0.0652, -0.0556, -0.061, -0.0574, -0.0499, -0.0473, -0.0433, -0.0455, -0.0514, -0.0593, -0.0593, -0.0443, -0.0554, -0.0644, -0.0484, -0.0463, -0.0308, -0.0202, -0.0198, -0.0149, -0.0172, -0.0208, -0.0233, -0.0274, -0.0443, -0.0318, -0.0377, -0.0447, -0.0447, -0.0425, -0.016, -0.0133, -0.0183, -0.036, -0.0138, -0.0186, -0.0131, -0.0126, 0.0008, 0.0008, -0.0021, 0.0084, 0.008, 0.0, -0.0267, -0.0191, -0.0304, -0.0226, 0.0015, 0.0076, 0.0008, -0.0069, 0.0024, 0.0068, -0.0017, 0.0058, -0.0022, -0.003, 0.0451, 0.0251, 0.0704, 0.063, 0.0712, 0.0578, 0.073, 0.0608, 0.0886, 0.0864, 0.0621, 0.0515, 0.0546, 0.052, 0.053, 0.0478, 0.0578, 0.0527, 0.0523, 0.0595, 0.0617, 0.0442, 0.0442, 0.0228, 0.0305, 0.0482, 0.0534, 0.0692, 0.0851, 0.0894, 0.1157, 0.0726, 0.0807, 0.0795, 0.0798, 0.1064, 0.096, 0.101, 0.0827, 0.082, 0.0804, 0.0833, 0.0759, 0.0777, 0.0914, 0.091, 0.0837, 0.0971, 0.1029, 0.1424, 0.1416, 0.1062, 0.1359, 0.1443, 0.1029, -0.0151, 0.0034, -0.0113, -0.0023, -0.0068, 0.0066, -0.0052, -0.0316, -0.1077, -0.1372, -0.1441, -0.1562, -0.1696, -0.1951, -0.1781, -0.1727, -0.1806, -0.1971, -0.2047, -0.1991, -0.2035, -0.2197, -0.2376, -0.2429, -0.2531, -0.247, -0.2187, -0.2126, -0.1997, -0.2056, -0.2056, -0.2025, -0.2045, -0.2046, -0.1976, -0.1886, -0.1726, -0.1723, -0.1398, -0.1394, -0.1182, -0.1241, -0.123, -0.1388, -0.1461, -0.136, -0.1433, -0.1317, -0.1315, -0.1252, -0.1252, -0.1225, -0.1349, -0.1351, -0.1488, -0.1488, -0.1443]}
{"unit_id": "orig:1943316150512947225", "ticker_idx": 3, "ts": "2025-07-10T14:27:30+00:00", "symbol": "HPE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Related beneficiaries include CSCO, ANET, FFIV, HPE, etc.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley Buy recommendation on CSCO, ANET, FFIV, HPE as AI/network equipment spending remains stable and AI tops CIO priorities.", "resolved_tickers": ["HPE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "US Optical Modules: Network Equipment Spending Relatively Stable, AI Remains CIO's Top Investment Priority\n\nMorgan Stanley (M. Marshall, July 08, 2025)\n\nMorgan Stanley's Q2 2025 CIO survey indicates stable network equipment spending and continued strong enthusiasm for AI investment, despite persistent macroeconomic and tariff uncertainties. Compared to other sub-sectors like communications software, network equipment demonstrates greater defensiveness. We recommend prioritizing related targets such as CSCO, ANET, and FFIV.\n\nNetwork Equipment Spending Expectations Stable, Ranking Slightly Down\nThe communications sector's 2025 growth forecast is 2.9%, a 7bps decrease from the previous quarter, but still 13bps higher than 2024. Network equipment's priority in CIO spending shifted from 7th to 8th, but spending resilience remains strong, making it one of the few sectors whose expectations have not been downgraded since early 2025.\n\nAI Continues to Be CIO's Top Priority, Budget Source Leaning Towards New Investment\nAI remains the top priority for CIOs in 2025, with 60% of CIOs expecting to enter the production phase in 2025 or earlier. 56% of AI spending comes from new budgets, a higher proportion than the 41% in Q1 2025, indicating enterprises' willingness for incremental investment in AI. Related beneficiaries include CSCO, ANET, FFIV, HPE, etc.\n\nPublic Cloud Penetration Continues to Increase, Equipment Upgrade Demand Rises\nThe Q2 2025 survey shows enterprises expect 68% of workloads to migrate to public clouds by the end of 2027, up from the current 44% and 50% by the end of 2025. With clearer distinctions drawn between cloud and on-premise deployments, some enterprises are beginning to refresh on-premise equipment that hasn't been replaced in years, benefiting demand for network-related equipment.\n\nCommunications Software Shows Some Improvement But Remains Under Pressure\nCommunications software spending remains less defensive but has shown marginal improvement compared to the last quarter, especially in collaboration and customer service software. Spending intention for collaboration software slightly rose to the edge of the Top 10, and customer service software's defensiveness climbed from 25th in Q1 to 10th, indicating the emergence of structural highlights.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023015551975301807, "ret_signed_m1d": -0.023015551975301807, "alpha_spy_m1d": -0.02020317353660761, "alpha_c_m1d": -0.02020317353660761, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.0028123784386941963, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02677323777776397, "ret_signed_p1d": -0.02677323777776397, "alpha_spy_p1d": -0.023257813907168412, "alpha_c_p1d": -0.023257813907168412, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.0035154238705955576, "ret_p1w": -0.021606419799378496, "ret_signed_p1w": -0.021606419799378496, "alpha_spy_p1w": -0.025153710866383627, "alpha_c_p1w": -0.025153710866383627, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.0035472910670051316, "ret_p1m": -0.022545841249994036, "ret_signed_p1m": -0.022545841249994036, "alpha_spy_p1m": -0.04069794711596797, "alpha_c_p1m": -0.04069794711596797, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.018152105865973933, "ret_p3m": 0.17621593161026428, "ret_signed_p3m": 0.17621593161026428, "alpha_spy_p3m": 0.10406250089510638, "alpha_c_p3m": 0.10406250089510638, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.0721534307151579, "ret_p6m": 0.14813806340210256, "ret_signed_p6m": 0.14813806340210256, "alpha_spy_p6m": 0.050237216610042434, "alpha_c_p6m": 0.050237216610042434, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.09790084679206013, "price_path": [-0.3341, -0.3001, -0.3062, -0.294, -0.294, -0.309, -0.2936, -0.2842, -0.2455, -0.2427, -0.2311, -0.2395, -0.2437, -0.2348, -0.2138, -0.2227, -0.2311, -0.2278, -0.2134, -0.2124, -0.1728, -0.1644, -0.1677, -0.1784, -0.1723, -0.1859, -0.1737, -0.1821, -0.1817, -0.1877, -0.1877, -0.1635, -0.1751, -0.177, -0.1942, -0.1914, -0.1751, -0.1681, -0.1719, -0.1555, -0.1462, -0.1448, -0.1504, -0.1546, -0.177, -0.1495, -0.1653, -0.1649, -0.1649, -0.1564, -0.1635, -0.1475, -0.1315, -0.1372, -0.1353, -0.0395, -0.038, -0.0019, 0.0023, 0.0023, -0.0178, -0.0221, -0.023, 0.0, -0.0268, -0.0291, -0.0474, -0.0442, -0.0216, -0.0085, -0.0296, -0.0413, -0.0249, -0.0366, -0.0263, -0.0193, -0.0207, -0.0188, -0.0282, -0.0695, -0.0597, -0.0667, -0.0545, -0.0418, -0.0225, -0.0329, 0.0014, 0.0127, 0.0014, -0.0061, 0.016, 0.0023, -0.0117, 0.0249, 0.0545, 0.0597, 0.0568, 0.0676, 0.0878, 0.0601, 0.0601, 0.0653, 0.0719, 0.0878, 0.1047, 0.1038, 0.1217, 0.148, 0.1677, 0.1616, 0.1667, 0.1475, 0.1616, 0.1918, 0.18, 0.188, 0.1663, 0.155, 0.1432, 0.1332, 0.1399, 0.1597, 0.1772, 0.1559, 0.1536, 0.1705, 0.1762, 0.2315, 0.2395, 0.1521, 0.171, 0.1805, 0.1824, 0.0624, 0.0841, 0.0799, 0.0988, 0.0903, 0.095, 0.1096, 0.1417, 0.1451, 0.146, 0.1526, 0.1531, 0.1569, 0.1125, 0.1196, 0.1049, 0.1073, 0.1276, 0.0808, 0.1087, 0.0827, 0.078, 0.0025, -0.0131, -0.0273, -0.0547, -0.0311, -0.0042, 0.0086, 0.0166, 0.0166, 0.0327, 0.036, 0.035, 0.0511, 0.0813, 0.1016, 0.1266, 0.1696, 0.1927, 0.1587, 0.1271, 0.1356, 0.1474, 0.1342, 0.1295, 0.1605, 0.1733, 0.1676, 0.161, 0.161, 0.1633, 0.1557, 0.1434, 0.141, 0.141, 0.1481]}
{"unit_id": "orig:1928087036545880194", "ticker_idx": 0, "ts": "2025-05-29T13:52:26+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "maintain our price target of $170, based on ~30–35x multiple on CY26 EPS of $5.39. We reiterate our Overweight rating on NVIDIA", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/adopts J.P. Morgan's Overweight on NVDA with $170 Dec-2025 PT on Blackwell ramp, strong AI demand, and CY26 EPS of $5.39.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan | NVIDIA Analysis: Solid Results/Guidance Despite China Regulations; Smooth Blackwell Production/Ramp; GB300 Mass Production Preparation This Quarter; Maintain Overweight (OW) Rating\n\nDespite approximately $2.5 billion in lost H20 GPU shipments to China due to newly enacted regulations (effective April 9), NVIDIA reported solid revenue of $44 billion for the April quarter, beating both our and consensus estimates of $43 billion. However, gross margin (61% vs. prior guidance of 71%) and EPS ($0.81 vs. consensus of $0.93) were hit due to a $4.5 billion H20 inventory write-down and purchase commitment.\n\nFor the July quarter, the company guided revenue of $45 billion. While this falls short of consensus estimates ($46 billion), which we believe do not fully reflect H20 shipment restrictions, it exceeds our estimate that takes the restrictions into account ($41–43 billion). Excluding H20 shipments, July quarter data center revenue is expected to grow approximately 16% QoQ, supported by continued investments in AI/accelerated computing (training and inference) and a strong ramp of the Blackwell data center platform.\n\nBlackwell shipments are expected to increase throughout the year, and the company is on track to begin mass production of the next-generation Blackwell Ultra (B300/GB300) by the end of the quarter. We still expect that ~$5 billion worth of H20 shipments originally planned for 2H will not be fulfilled due to the April restrictions.\n\nIn networking, the April quarter saw significant improvement both QoQ and YoY, driven by strong NVLink switch content within the NVL72 rackscale platform. That said, over the past 12 months, networking revenue has only grown 37% compared to 113% for compute, so we will closely monitor the sustainability of the improvement.\n\nDemand for Blackwell remains very strong and is expected to exceed supply for several quarters. However, the company and its supply chain partners are currently shipping 1,000 NVL72 racks per week, demonstrating strong execution. Gross margin guidance (72%) came in 30bps above consensus, reflecting the company’s aggressive ramp of the high-margin Blackwell platform, supported by improved yields and cost structures.\n\nGiven the recent easing of export controls and large-scale data center announcements in regions like the UAE, Saudi Arabia, and Taiwan, we believe the company has strong visibility into continued high growth through 2026.\n\nConclusion: NVIDIA remains 1–2 steps ahead of its peers through its comprehensive silicon/hardware/software platform and robust ecosystem. With an aggressive product rollout cadence and increasing segmentation, it continues to widen the gap. We adjust our estimates and introduce our first CY2026 projections, maintaining our December 2025 price target of $170 and Overweight (OW) rating.\n\n⸻\n\n1. April Quarter Revenue Beats Consensus Despite H20 GPU Shipment Restrictions to China\n\nNVIDIA reported April quarter revenue (FQ1 2026) of $44 billion (+12% QoQ), beating consensus ($43.31 billion), despite a $2.4 billion revenue loss from H20 GPU shipment restrictions to China (effective April 9). However, gross margin (61% vs. prior guidance of 71%) and EPS ($0.81 vs. consensus of $0.93) were impacted by a $4.5 billion inventory write-down and purchase obligations related to H20.\n 1. Data center revenue grew ~10% QoQ, supported by increased Blackwell production (~5% QoQ growth in data center compute), though partially offset by lower H20 shipments.\n 2. Networking revenue rose ~64% QoQ due to growth in Spectrum-X Ethernet and NVLink switch deployments.\n 3. Gaming revenue increased ~48% QoQ driven by supply improvements and expansion of the RTX 50 series.\n 4. Pro Visualization revenue remained flat QoQ.\n 5. Automotive revenue declined ~1% QoQ.\n 6. Gross margin fell to 61% (consensus 71.1%) due to H20-related write-downs.\n 7. EPS came in at $0.81, below consensus ($0.93), but excluding H20 charges, EPS would be $0.96.\n\n⸻\n\n2.July Quarter Revenue Guidance at $45B — Above Our Estimate Despite $8B in H20 Losses\n\nWe estimate an additional $5 billion of H20 shipments expected for 2H will not occur due to restrictions, but expect this to be offset by a strong Blackwell/Blackwell Ultra ramp in 2H. For the July quarter, the company guided $45 billion in revenue. While this falls short of the $46 billion consensus (due to underestimation of H20 impact by sell-side analysts), it is above our more conservative estimate of $41–43 billion.\n\nExcluding H20, July data center revenue is projected to grow ~16% QoQ, driven by continued investment in AI/accelerated computing and the ramp of the Blackwell platform. Gross margin is expected at 72%, ~30bps above consensus, as NVIDIA increases shipment volumes and improves yield/costs. Operating expenses (OpEx) are forecast at $4 billion. EPS is projected at $0.99, slightly below consensus of $1.01.\n\n⸻\n\n3. Smooth Blackwell Ramp Continues — Strong Execution Against Model Innovation like DeepSeek\n\nWe view NVIDIA’s response to innovations like DeepSeek as highly effective. As we’ve noted, such model innovations increase overall computing complexity (e.g., test-time inference, reinforcement learning using AI/synthetic data, agent AI, WFM systems, multimodal variants), sustaining demand for greater and more powerful infrastructure. NVIDIA plans to launch the Blackwell Ultra platform in 2H and expects supply shortages to persist for several quarters.\n\n⸻\n\n4. Revised Estimates and Maintained $170 Dec 2025 Price Target\n\nDespite solid topline results, we revise our estimates to reflect lower margins in 1H due to the H20 write-down. We introduce our CY26 EPS forecast for the first time and maintain our price target of $170, based on ~30–35x multiple on CY26 EPS of $5.39. We reiterate our Overweight rating on NVIDIA.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031467771192964955, "ret_signed_m1d": -0.031467771192964955, "alpha_spy_m1d": -0.027536013589865016, "alpha_c_m1d": -0.02442091775738242, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029168709381678815, "ret_signed_p1d": -0.029168709381678815, "alpha_spy_p1d": -0.028050357189775443, "alpha_c_p1d": -0.011428601389960602, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.005747599702493611, "ret_signed_p1w": 0.005747599702493611, "alpha_spy_p1w": 0.0006632770562644552, "alpha_c_p1w": -0.023832855030635036, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.13342157350479011, "ret_signed_p1m": 0.13342157350479011, "alpha_spy_p1m": 0.08820871516440842, "alpha_c_p1m": -0.007270048081713831, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.30600340555452554, "ret_signed_p3m": 0.30600340555452554, "alpha_spy_p3m": 0.20937208449485212, "alpha_c_p3m": 0.09021349223987807, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.2853115202986203, "ret_signed_p6m": 0.2853115202986203, "alpha_spy_p6m": 0.16199846496754655, "alpha_c_p6m": -0.05084882377385047, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [-0.1806, -0.1668, -0.1573, -0.2057, -0.1905, -0.2315, -0.2187, -0.1685, -0.1696, -0.1259, -0.1412, -0.1707, -0.1557, -0.1484, -0.1544, -0.1277, -0.1329, -0.1827, -0.1994, -0.2121, -0.2214, -0.2086, -0.2067, -0.2686, -0.3224, -0.2985, -0.3081, -0.1786, -0.2272, -0.203, -0.2046, -0.1939, -0.2493, -0.2709, -0.2709, -0.3038, -0.2895, -0.2621, -0.2354, -0.2025, -0.2188, -0.2168, -0.2175, -0.1981, -0.1774, -0.1823, -0.1843, -0.159, -0.1568, -0.1619, -0.1163, -0.0665, -0.0277, -0.0313, -0.0272, -0.026, -0.0346, -0.0531, -0.0457, -0.0568, -0.0568, -0.0265, -0.0315, 0.0, -0.0292, -0.013, 0.0146, 0.0196, 0.0057, 0.0182, 0.0247, 0.0343, 0.0262, 0.0418, 0.02, 0.0396, 0.0355, 0.0453, 0.0453, 0.0336, 0.0359, 0.0627, 0.1087, 0.1138, 0.1334, 0.1351, 0.1014, 0.1298, 0.1448, 0.1448, 0.1369, 0.1496, 0.1703, 0.179, 0.1849, 0.1788, 0.2265, 0.2313, 0.243, 0.2388, 0.2314, 0.2001, 0.227, 0.2483, 0.2466, 0.2699, 0.261, 0.288, 0.278, 0.2482, 0.2933, 0.2808, 0.2891, 0.2988, 0.3127, 0.3081, 0.316, 0.3047, 0.3078, 0.2965, 0.3077, 0.262, 0.2602, 0.2572, 0.2788, 0.2919, 0.306, 0.3048, 0.2945, 0.2515, 0.2515, 0.227, 0.2259, 0.2334, 0.2, 0.2093, 0.2269, 0.2741, 0.273, 0.2777, 0.2772, 0.2566, 0.2236, 0.2663, 0.2694, 0.3193, 0.2821, 0.2716, 0.2768, 0.2804, 0.3067, 0.3406, 0.3454, 0.3572, 0.3481, 0.3332, 0.3296, 0.3588, 0.3837, 0.3161, 0.3531, 0.2936, 0.2921, 0.3064, 0.3165, 0.3123, 0.3017, 0.2954, 0.3089, 0.3383, 0.3759, 0.4445, 0.4877, 0.4578, 0.455, 0.4865, 0.4277, 0.4026, 0.3514, 0.3519, 0.4302, 0.3879, 0.3925, 0.3427, 0.3664, 0.3408, 0.3031, 0.3402, 0.298, 0.2853]}
{"unit_id": "orig:1940903893556514936", "ticker_idx": 0, "ts": "2025-07-03T22:42:03+00:00", "symbol": "Alchip", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley indicates that Alchip may have become the exclusive supplier for AWS Trainium3 and has a high probability of securing the Trainium4 project... the company's CoWoS capacity is expected to significantly ramp up in 2026, driving accelerated revenue growth", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Buy-equivalent thesis: Alchip exclusive on Trainium3, high win-rate on T4, 2026 ramp; Marvell exits XPU foundry, loses share.", "resolved_tickers": ["3661.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Alchip Technologies 3661.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AIchip: Expected to Exclusively Supply Trainium3 and Likely to Secure Trainium4, Significant Growth Potential in 2026\n\nMorgan Stanley (C. Chan, 25/07/02)\n\nMorgan Stanley indicates that Alchip may have become the exclusive supplier for AWS Trainium3 and has a high probability of securing the Trainium4 project. Considering its smooth progression to 3nm and potential 2nm project roadmap, the company's CoWoS capacity is expected to significantly ramp up in 2026, driving accelerated revenue growth.\n\nTrainium3 Expected to be Exclusively Contracted by Alchip, Mass Production to Begin in 2026\n\nAccording to industry surveys, Alchip is highly likely to be the sole XPU supplier for AWS Trainium3. Its 3nm chip successfully taped out in May 2025, demonstrating mature backend design and no need to change service providers. Morgan Stanley has raised its 2026 CoWoS wafer start forecast to 40,000 wafers, corresponding to approximately 600,000 to 650,000 units shipped, contributing significantly to revenue.\n\nMarvell Exits XPU Foundry Competition, Shifts to Peripheral Chip Support\n\nIn contrast, Marvell recently canceled some 2026 CoWoS-R orders, indicating its failure to enter the 3nm XPU foundry project and its focus solely on XPU auxiliary chips. This further solidifies Alchip's dominant position in Trainium3 foundry.\n\nTrainium4 Project Also Has High Win Rate, Collaboration with Astera Drives 2nm Layout\n\nAlchip hinted at positive progress on its 2nm test chip project during its Q1 2025 earnings call. Coupled with its recent collaboration with Astera Labs on chiplet I/O design capabilities, this increases its likelihood of securing the Trainium4 project. The market expects a decision on this project by the end of July, and Morgan Stanley has upgraded this expectation to its base case.\n\nAlthough AWS Builds Small-Scale Production Lines, Alchip Still Holds Nearly 100% of Foundry Orders\n\nWhile the client operates a small CoWoS production line, it is primarily for supplemental capacity, to be utilized if Alchip's capacity is insufficient. Overall, Alchip still handles virtually all Turnkey foundry business. The estimated unit price for its 3nm chip is US$3,000 (excluding HBM), and clients generally use an HBM consignment model to reduce working capital pressure.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00472437253202318, "ret_signed_m1d": 0.00472437253202318, "alpha_spy_m1d": 0.012544127976170105, "alpha_c_m1d": 0.013080703224258028, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00472437253202318, "ret_signed_p1d": 0.00472437253202318, "alpha_spy_p1d": 0.00472437253202318, "alpha_c_p1d": 0.00472437253202318, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.0897637767160151, "ret_signed_p1w": 0.0897637767160151, "alpha_spy_p1w": 0.0889962114756766, "alpha_c_p1w": 0.07601306866216162, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.2236220334308563, "ret_signed_p1m": 0.2236220334308563, "alpha_spy_p1m": 0.22941105811411255, "alpha_c_p1m": 0.222458493714216, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.10167620020199064, "ret_signed_p3m": 0.10167620020199064, "alpha_spy_p3m": 0.03341427481409598, "alpha_c_p3m": -0.049018275423687196, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.022190182871399733, "ret_signed_p6m": 0.022190182871399733, "alpha_spy_p6m": -0.08803663472417211, "alpha_c_p6m": -0.27178902797667615, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.2457, -0.3197, -0.3874, -0.3276, -0.3307, -0.2976, -0.2567, -0.3165, -0.3024, -0.3055, -0.337, -0.3717, -0.3433, -0.3354, -0.3197, -0.3102, -0.315, -0.3307, -0.3307, -0.2756, -0.2835, -0.2835, -0.2976, -0.2614, -0.2756, -0.2031, -0.1512, -0.1323, -0.1433, -0.1402, -0.1795, -0.0976, -0.0614, -0.0866, -0.0646, -0.1008, -0.0992, -0.1008, -0.115, -0.115, -0.1433, -0.1433, -0.1165, -0.126, -0.1197, -0.1417, -0.1323, -0.1055, -0.1244, -0.1701, -0.1701, -0.1134, -0.1165, -0.0961, -0.1181, -0.1087, -0.1276, -0.0929, -0.0929, -0.0551, -0.0252, 0.0031, 0.0047, 0.0, 0.0047, 0.0142, 0.0126, 0.0598, 0.0898, 0.0803, 0.0425, 0.0614, 0.1339, 0.1827, 0.2378, 0.1906, 0.1717, 0.2047, 0.252, 0.2236, 0.222, 0.2016, 0.1811, 0.2236, 0.2236, 0.1827, 0.1811, 0.1685, 0.1858, 0.2094, 0.2315, 0.2835, 0.2976, 0.3386, 0.3559, 0.3323, 0.2709, 0.1528, 0.1921, 0.1622, 0.2268, 0.2535, 0.285, 0.2598, 0.2772, 0.2331, 0.2225, 0.2654, 0.2241, 0.2622, 0.2638, 0.2448, 0.2527, 0.2479, 0.205, 0.1843, 0.17, 0.1955, 0.1828, 0.1923, 0.1955, 0.1764, 0.0842, 0.0715, 0.0413, 0.0413, 0.1017, 0.0794, 0.0444, 0.0524, 0.0524, 0.1001, 0.0953, 0.1176, 0.1176, 0.0715, -0.0223, -0.0207, -0.016, -0.0191, 0.0285, 0.0667, 0.0238, 0.0015, 0.0015, -0.0255, -0.0255, 0.0333, 0.0667, 0.1112, 0.0969, 0.0953, 0.0619, 0.1669, 0.1335, 0.1812, 0.1557, 0.1192, 0.1208, 0.116, 0.1478, 0.0333, 0.0556, -0.0223, -0.0493, -0.0557, -0.0621, -0.0112, 0.0444, 0.0508, 0.046, 0.0874, 0.0142, 0.0079, 0.0031, -0.008, 0.0365, 0.0715, 0.0476, 0.0238, 0.0079, -0.0414, 0.0174, 0.0079, 0.0222, 0.0254, 0.0238, 0.0111, 0.0111, 0.0222]}
{"unit_id": "orig:1940903893556514936", "ticker_idx": 1, "ts": "2025-07-03T22:42:03+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell recently canceled some 2026 CoWoS-R orders, indicating its failure to enter the 3nm XPU foundry project", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Buy-equivalent thesis: Alchip exclusive on Trainium3, high win-rate on T4, 2026 ramp; Marvell exits XPU foundry, loses share.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AIchip: Expected to Exclusively Supply Trainium3 and Likely to Secure Trainium4, Significant Growth Potential in 2026\n\nMorgan Stanley (C. Chan, 25/07/02)\n\nMorgan Stanley indicates that Alchip may have become the exclusive supplier for AWS Trainium3 and has a high probability of securing the Trainium4 project. Considering its smooth progression to 3nm and potential 2nm project roadmap, the company's CoWoS capacity is expected to significantly ramp up in 2026, driving accelerated revenue growth.\n\nTrainium3 Expected to be Exclusively Contracted by Alchip, Mass Production to Begin in 2026\n\nAccording to industry surveys, Alchip is highly likely to be the sole XPU supplier for AWS Trainium3. Its 3nm chip successfully taped out in May 2025, demonstrating mature backend design and no need to change service providers. Morgan Stanley has raised its 2026 CoWoS wafer start forecast to 40,000 wafers, corresponding to approximately 600,000 to 650,000 units shipped, contributing significantly to revenue.\n\nMarvell Exits XPU Foundry Competition, Shifts to Peripheral Chip Support\n\nIn contrast, Marvell recently canceled some 2026 CoWoS-R orders, indicating its failure to enter the 3nm XPU foundry project and its focus solely on XPU auxiliary chips. This further solidifies Alchip's dominant position in Trainium3 foundry.\n\nTrainium4 Project Also Has High Win Rate, Collaboration with Astera Drives 2nm Layout\n\nAlchip hinted at positive progress on its 2nm test chip project during its Q1 2025 earnings call. Coupled with its recent collaboration with Astera Labs on chiplet I/O design capabilities, this increases its likelihood of securing the Trainium4 project. The market expects a decision on this project by the end of July, and Morgan Stanley has upgraded this expectation to its base case.\n\nAlthough AWS Builds Small-Scale Production Lines, Alchip Still Holds Nearly 100% of Foundry Orders\n\nWhile the client operates a small CoWoS production line, it is primarily for supplemental capacity, to be utilized if Alchip's capacity is insufficient. Overall, Alchip still handles virtually all Turnkey foundry business. The estimated unit price for its 3nm chip is US$3,000 (excluding HBM), and clients generally use an HBM consignment model to reduce working capital pressure.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01237028010898078, "ret_signed_m1d": 0.01237028010898078, "alpha_spy_m1d": 0.004550524664833855, "alpha_c_m1d": 0.004013949416745932, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.02420847780573243, "ret_signed_p1w": 0.02420847780573243, "alpha_spy_p1w": 0.024976043046070928, "alpha_c_p1w": 0.03795918585958591, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.008899584580905007, "ret_signed_p1m": 0.008899584580905007, "alpha_spy_p1m": 0.003110559897648746, "alpha_c_p1m": 0.0100631242975453, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.11916478629956484, "ret_signed_p3m": -0.11916478629956484, "alpha_spy_p3m": -0.05090286091167018, "alpha_c_p3m": 0.031529689326113, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.15014472254939615, "ret_signed_p6m": -0.15014472254939615, "alpha_spy_p6m": -0.0399179049538243, "alpha_c_p6m": 0.14383448829867973, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.3224, -0.3353, -0.1901, -0.2976, -0.2898, -0.3049, -0.2908, -0.3093, -0.3123, -0.3123, -0.3432, -0.3267, -0.2848, -0.2374, -0.2163, -0.2192, -0.2195, -0.2236, -0.1893, -0.1709, -0.1756, -0.1857, -0.251, -0.2334, -0.2066, -0.1421, -0.129, -0.1234, -0.1327, -0.1519, -0.1679, -0.183, -0.2005, -0.1773, -0.1927, -0.1927, -0.1511, -0.1409, -0.1523, -0.1994, -0.1824, -0.1705, -0.1181, -0.1333, -0.0908, -0.0803, -0.0843, -0.0923, -0.0737, -0.1063, -0.0633, -0.069, -0.0031, -0.0031, -0.0222, -0.0585, 0.0004, 0.01, 0.0637, 0.0263, 0.0295, 0.0141, -0.0124, 0.0, 0.0, -0.0483, -0.043, -0.0388, -0.0242, -0.0321, -0.0347, -0.0361, -0.0568, -0.0414, -0.0062, -0.0274, -0.0416, -0.0246, -0.0144, -0.0121, 0.0105, 0.0163, 0.0881, 0.0699, -0.0089, 0.0188, 0.0201, 0.0027, 0.0097, 0.0296, 0.0288, 0.0358, 0.0559, 0.0522, 0.0143, 0.0216, -0.0406, -0.0519, -0.052, -0.0282, -0.0289, -0.0114, -0.0044, 0.0281, -0.1631, -0.1631, -0.14, -0.1705, -0.1467, -0.1569, -0.1214, -0.1102, -0.1067, -0.1135, -0.1034, -0.1024, -0.0833, -0.0551, -0.0118, -0.0114, 0.0055, -0.0066, 0.0662, 0.1157, 0.1072, 0.0968, 0.1192, 0.1168, 0.1475, 0.1478, 0.1837, 0.1578, 0.2314, 0.2072, 0.1404, 0.1908, 0.1485, 0.1841, 0.1753, 0.1716, 0.1435, 0.1224, 0.0797, 0.1026, 0.1207, 0.1817, 0.1784, 0.2009, 0.1799, 0.2487, 0.2038, 0.1668, 0.2375, 0.2433, 0.2112, 0.2419, 0.19, 0.19, 0.1659, 0.1516, 0.1116, 0.0481, 0.0833, 0.0215, 0.0317, 0.1162, 0.1114, 0.1685, 0.1685, 0.1909, 0.2136, 0.2374, 0.3348, 0.308, 0.3176, 0.2255, 0.1842, 0.2318, 0.1913, 0.1247, 0.1224, 0.1199, 0.0883, 0.1252, 0.1202, 0.1296, 0.168, 0.1521, 0.1521, 0.1501]}
{"unit_id": "reply:1928597275372036460", "ticker_idx": 0, "ts": "2025-05-30T23:39:56+00:00", "symbol": "NVTS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GaN(질화갈륨) 기반 고전압 제품이 엔비디아 GB300에 채택되었으며, 현재 시장에서 이보다 더 잘하는 경쟁사는 없다. 800V급 주문도 지속적으로 들어오고 있다.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish NVTS: GaN high-voltage product adopted in NVIDIA GB300, no competitor does it better, 800V orders continuing.", "resolved_tickers": ["NVTS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@Semicon_player \"미국 주식 NVTS 나노비 반도체: GaN(질화갈륨) 기반 고전압 제품이 엔비디아 GB300에 채택되었으며, 현재 시장에서 이보다 더 잘하는 경쟁사는 없다. 800V급 주문도 지속적으로 들어오고 있다.\"  \n\nhttps://t.co/4KFDmKeq6r", "reply_text": null, "parent_text": "3월 10일의 코멘터리 \nCES 4.5$ 이후 2.25불 까지 하락 할 것을 예상 \n\n하지만 트럼프가 관세로 더 매수기회를 줌 1.8$ 까지 떨어짐 \n그래도 계속 이 악물고 매수 \n\n결국 모두 수익 전환 https://t.co/DGMdMVOFIT", "quoted_text": null, "parent_author": "Semicon_player", "ret_m1d": 0.05686273868154523, "ret_signed_m1d": 0.05686273868154523, "alpha_spy_m1d": 0.05574313437771461, "alpha_c_m1d": 0.03880223541143035, "bench_spy_m1d": 0.0011196043038306236, "bench_c_m1d": 0.01806050327011488, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08235295767602646, "ret_signed_p1d": 0.08235295767602646, "alpha_spy_p1d": 0.07672005653921277, "alpha_c_p1d": 0.06754581059660625, "bench_spy_p1d": 0.005632901136813695, "bench_c_p1d": 0.014807147079420213, "ret_p1w": 0.21176469488264904, "ret_signed_p1w": 0.21176469488264904, "alpha_spy_p1w": 0.19522232918776905, "alpha_c_p1w": 0.15762493814730028, "bench_spy_p1w": 0.01654236569487999, "bench_c_p1w": 0.054139756735348765, "ret_p1m": 0.2843137869052099, "ret_signed_p1m": 0.2843137869052099, "alpha_spy_p1m": 0.2329276258596258, "alpha_c_p1m": 0.12110210207095462, "bench_spy_p1m": 0.05138616104558413, "bench_c_p1m": 0.1632116848342553, "ret_p3m": 0.19019613434653437, "ret_signed_p3m": 0.19019613434653437, "alpha_spy_p3m": 0.08983553356659257, "alpha_c_p3m": -0.05134712825534482, "bench_spy_p3m": 0.1003606007799418, "bench_c_p3m": 0.2415432626018792, "ret_p6m": 0.6254902189944813, "ret_signed_p6m": 0.6254902189944813, "alpha_spy_p6m": 0.4843674954263446, "alpha_c_p6m": 0.21101682837601432, "bench_spy_p6m": 0.1411227235681367, "bench_c_p6m": 0.414473390618467, "price_path": [-0.5216, -0.4824, -0.5118, -0.502, -0.5618, -0.502, -0.5216, -0.5608, -0.5098, -0.4804, -0.5137, -0.5039, -0.5431, -0.5431, -0.5569, -0.5667, -0.5824, -0.6, -0.6412, -0.598, -0.6078, -0.6255, -0.6853, -0.6451, -0.6569, -0.6804, -0.6275, -0.6549, -0.6471, -0.6588, -0.6588, -0.6608, -0.6431, -0.6431, -0.6471, -0.6471, -0.6353, -0.602, -0.598, -0.6059, -0.6157, -0.6216, -0.6255, -0.598, -0.6078, -0.6265, -0.6167, -0.6275, -0.6314, -0.5941, -0.5804, -0.5902, -0.5941, -0.5922, -0.602, -0.6078, -0.6255, -0.0098, -0.1353, -0.1353, 0.2745, 0.2078, 0.0569, 0.0, 0.0824, 0.1647, 0.3294, 0.1863, 0.2118, 0.4314, 0.5863, 0.5608, 0.5314, 0.4353, 0.4098, 0.3451, 0.3725, 0.3725, 0.3745, 0.2784, 0.4412, 0.3765, 0.2961, 0.3078, 0.2843, 0.2294, 0.2392, 0.2627, 0.2627, 0.2059, 0.2902, 0.2608, 0.2275, 0.1451, 0.151, 0.2216, 0.2059, 0.2294, 0.3314, 0.6608, 0.6902, 0.7294, 0.7608, 0.7235, 0.6784, 0.4745, 0.4137, 0.4373, 0.5608, 0.5784, 0.3275, 0.2451, 0.3314, 0.3059, 0.2941, 0.3647, 0.4431, 0.4216, 0.3569, 0.3275, 0.2608, 0.2255, 0.2176, 0.2608, 0.2157, 0.1941, 0.1902, 0.1843, 0.149, 0.149, 0.1294, 0.1235, 0.0863, 0.098, 0.1, 0.1294, 0.1059, 0.1922, 0.2098, 0.198, 0.1549, 0.2667, 0.3196, 0.3745, 0.4039, 0.349, 0.3451, 0.2765, 0.2608, 0.4098, 0.4157, 0.4314, 0.6069, 0.5333, 0.5941, 0.5255, 0.5176, 0.5686, 0.6137, 0.9549, 1.4647, 1.9725, 2.0647, 1.8745, 2.3529, 1.9471, 1.6676, 1.6686, 1.7588, 1.7275, 1.5157, 1.6608, 1.4627, 1.6392, 1.402, 1.051, 0.9333, 0.7324, 0.5373, 0.8824, 0.7882, 0.6941, 0.551, 0.5902, 0.5157, 0.5255, 0.5137, 0.4804, 0.4824, 0.6255]}
{"unit_id": "orig:1938956440607285263", "ticker_idx": 0, "ts": "2025-06-28T13:43:34+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley anticipates this improving trend to continue into 2025 and has raised its target price to $135", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley Equal-Weight on MU with PT raised to $135; neutral risk-reward due to high valuation and HBM supply-demand risks despite AI-driven earnings strength.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron: Performance Continues to Improve Driven by AI, But Valuation Pressure and HBM Supply-Demand Risks Warrant Attention\n\nMorgan Stanley (J. Moore, 2025/06/26)\n\nMorgan Stanley believes that Micron has shown strong performance in recent quarters, benefiting from AI-related demand and an expanding share in HBM. However, the current high valuation, coupled with the uncertainty in HBM growth trajectory and instability in consumer products, renders the overall risk-reward neutral.\n\nMicron's May quarter results were robust, with EPS reaching $2.50, and guidance exceeding market expectations. This was primarily driven by improved DRAM pricing and increased shipments fueled by AI demand. Although DRAM core pricing has reached an inflection point and the consumer product mix still poses a drag, the AI-related product mix (HBM, DC LPDDR, 96+GB RDIMM) supported the gross margin performance. Morgan Stanley anticipates this improving trend to continue into 2025 and has raised its target price to $135.\n\nThe HBM business continues to expand, with annualized revenue surpassing $6 billion in May, and Micron maintaining its leading position alongside SK Hynix. Nevertheless, Morgan Stanley cautions that Samsung's HBM4 validation on the Rubin platform may begin in the September quarter, which, if successful, could lead to price pressure. At the same time, there's a possibility that the overall HBM TAM expectation is overestimated, with a discrepancy in the revenue ratio compared to GPU/ASIC, thus requiring continued attention to the actual demand fulfillment in the future.\n\nAI inference demand is driving both volume and price increases for DRAM, supporting Micron's increased share in data center-related products. However, Morgan Stanley points out that the DRAM industry's pricing inflection point has already been factored into the market. Micron's current 3x PB valuation is considered quite expensive, and its free cash flow remains limited. Furthermore, despite the positive AI trend, the rapid growth trajectory of HBM could still be affected by inventory, changes in customer pull-in pace, and a structural overestimation of TAM.\n\nConsumer-side shipments (e.g., Mobile) showed significant quarter-over-quarter improvement, though still negative year-over-year. Morgan Stanley believes that the recent rebound in shipments more reflects inventory replenishment rather than pull-in. The negative impact of the consumer mix is expected to gradually dissipate in the second half of the year, and DRAM prices may continue to rise in the future. Although tariff uncertainties persist, their current impact does not constitute a primary downside risk.\n\nMorgan Stanley expects the positive trend to continue into the second half of 2025, but most driving factors are already reflected in the current stock price. The valuation model is based on a cyclical EPS calculation of $7.5, with the target price raised from 14x PE to 18x PE, considering the structural improvement in profitability driven by AI. However, due to the low correlation between free cash flow and profitability, Micron maintains an \"Equal-Weight\" (EW) rating. Overall, Micron still has structural highlights, but further stock price appreciation requires stronger validation from the AI industry chain.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012251577573582306, "ret_signed_m1d": 0.012251577573582306, "alpha_spy_m1d": 0.01701013163863674, "alpha_c_m1d": 0.013900975108794511, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01914799578333326, "ret_signed_p1d": -0.01914799578333326, "alpha_spy_p1d": -0.018824316475108405, "alpha_c_p1d": -0.00824721902093617, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.02610236865156823, "ret_signed_p1w": -0.02610236865156823, "alpha_spy_p1w": -0.0306826948673109, "alpha_c_p1w": -0.029257914435192767, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.09074737744311612, "ret_signed_p1m": -0.09074737744311612, "alpha_spy_p1m": -0.11892560887890224, "alpha_c_p1m": -0.1410198639313055, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.2736522796191896, "ret_signed_p3m": 0.2736522796191896, "alpha_spy_p3m": 0.20563519245999307, "alpha_c_p3m": 0.12240441367681187, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.2450564955537793, "ret_signed_p6m": 1.2450564955537793, "alpha_spy_p6m": 1.1251960253450775, "alpha_c_p6m": 0.9387956674719136, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2811, -0.3968, -0.4749, -0.4453, -0.4682, -0.3682, -0.4316, -0.4357, -0.4238, -0.4236, -0.4375, -0.4418, -0.4418, -0.4585, -0.4303, -0.4083, -0.3718, -0.3527, -0.3626, -0.3762, -0.3757, -0.369, -0.3451, -0.3475, -0.3468, -0.3297, -0.3091, -0.3034, -0.2512, -0.2135, -0.2266, -0.2256, -0.2049, -0.1996, -0.2041, -0.2224, -0.2306, -0.2424, -0.2424, -0.218, -0.2196, -0.2146, -0.2336, -0.2034, -0.1704, -0.1623, -0.1376, -0.1192, -0.0998, -0.0739, -0.0586, -0.0574, -0.0621, -0.0277, -0.0236, -0.0116, -0.0116, 0.0028, -0.0095, 0.0378, 0.0325, 0.0223, 0.0123, 0.0, -0.0191, -0.0123, -0.0078, -0.0078, -0.0261, 0.0104, -0.0073, -0.0002, 0.0113, -0.0367, -0.0246, -0.0544, -0.0802, -0.071, -0.0804, -0.113, -0.108, -0.0926, -0.0964, -0.0965, -0.0907, -0.0682, -0.1136, -0.1482, -0.1248, -0.1143, -0.1166, -0.0915, -0.0345, 0.0048, 0.0375, 0.0092, 0.0175, -0.0184, 0.0034, -0.0088, -0.0481, -0.0596, -0.0443, -0.0545, -0.0539, -0.0437, -0.0092, -0.0335, -0.0335, -0.0378, -0.0358, 0.0087, 0.0669, 0.0676, 0.0983, 0.137, 0.2228, 0.2769, 0.2813, 0.2898, 0.2993, 0.3716, 0.3216, 0.3369, 0.3515, 0.3133, 0.2737, 0.2772, 0.3311, 0.3588, 0.4793, 0.4923, 0.5264, 0.5518, 0.509, 0.5971, 0.5629, 0.4757, 0.5665, 0.5201, 0.5598, 0.6458, 0.6446, 0.6803, 0.6439, 0.6128, 0.6798, 0.7798, 0.7886, 0.8033, 0.8417, 0.8204, 0.8184, 0.9072, 0.7718, 0.93, 0.9367, 0.9334, 1.0584, 0.9593, 0.9901, 0.9255, 1.0058, 0.9662, 0.8569, 0.8359, 0.6364, 0.6852, 0.8197, 0.8246, 0.8712, 0.8712, 0.9217, 0.9541, 0.9462, 0.9029, 0.8418, 0.9277, 1.0065, 1.0512, 1.143, 1.1003, 0.9596, 0.93, 0.8894, 0.8326, 1.0198, 1.1609, 1.2477, 1.2451]}
{"unit_id": "orig:1942351599789302254", "ticker_idx": 0, "ts": "2025-07-07T22:34:43+00:00", "symbol": "Infineon Technologies", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "PSMC's entry into GaN manufacturing could lower the industry's entry barrier, creating price competition similar to the SiC market, thereby exerting some pressure on Infineon", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley flags strategic and pricing headwinds for Infineon as TSMC exits GaN/Si and PSMC enters, lowering barriers and intensifying competition.", "resolved_tickers": ["IFX.DE"], "resolution_method": "exchange_ticker", "resolution_reason": "Infineon Technologies IFX.DE Germany", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Infineon: TSMC's Exit from GaN/Si Capacity May Impact Its Supply Chain and Pricing Environment\n\nMorgan Stanley (L. Simpson, 25/07/03)\n\nMorgan Stanley notes that TSMC's plan to exit its GaN on Silicon production line by 2027, prompting Navitas to shift to PSMC, could lower the entry barrier for the GaN/Si industry. This might have strategic and pricing implications for Infineon.\n\nTSMC Plans to Exit GaN/Si Business by 2027, Navitas Gradually Transfers Capacity to PSMC\n\nNavitas has announced that TSMC will cease its GaN/Si capacity by July 2027, signaling TSMC's gradual divestment from certain non-core businesses, especially in areas facing significant competition from low-cost manufacturing in China. Navitas will shift to PSMC, with device qualification expected to be completed by Q4 2025 and mass production of 100V products beginning in H1 2026. The transition for 650V products will be completed within the next 12 to 24 months.\n\nInfineon May Benefit from Potential GaN Strategic Cooperation Opportunities, But Also Faces Price Pressure\n\nMorgan Stanley points out that after TSMC's exit, other GaN device manufacturers may seek to collaborate with Infineon on its 300mm production lines or process IP. Concurrently, PSMC's entry into GaN manufacturing could lower the industry's entry barrier, creating price competition similar to the SiC market, thereby exerting some pressure on Infineon.\n\nInfineon May Need to Enhance Internal GaN Device Reliability Testing Capabilities\n\nPreviously, Infineon and GaN Systems utilized TSMC for GaN device reliability testing. In the future, as TSMC gradually exits, Infineon may need to build more of its own reliability testing capabilities to ensure product quality and consistency with customer requirements.\n\nInfineon Still Holds Leading Patent Reserves in GaN/Si Field\n\nDespite the changing competitive landscape, Infineon still possesses an extensive intellectual property portfolio in the GaN/Si field, encompassing 350 patent families. This provides a technological barrier for maintaining its long-term competitive advantage.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008219876596697762, "ret_signed_m1d": 0.008219876596697762, "alpha_spy_m1d": 0.01572784317519893, "alpha_c_m1d": 0.02201740229498872, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01926967075531527, "ret_signed_p1d": -0.01926967075531527, "alpha_spy_p1d": -0.01981738697306057, "alpha_c_p1d": -0.005972542267142833, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.0067376818918221115, "ret_signed_p1w": -0.0067376818918221115, "alpha_spy_p1w": -8.376783318730929e-05, "alpha_c_p1w": 0.013136502076646295, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": -0.04864576074909732, "ret_signed_p1m": 0.04864576074909732, "alpha_spy_p1m": 0.0603908671283202, "alpha_c_p1m": 0.07488252861862854, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": -0.07734804111399052, "ret_signed_p3m": 0.07734804111399052, "alpha_spy_p3m": 0.1585417391742937, "alpha_c_p3m": 0.2863842977533083, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.01684410113028867, "ret_signed_p6m": -0.01684410113028867, "alpha_spy_p6m": 0.09637111207194327, "alpha_c_p6m": 0.28584554755213354, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.3439, -0.3052, -0.3059, -0.2826, -0.2707, -0.2777, -0.2826, -0.2826, -0.2826, -0.2814, -0.2556, -0.2029, -0.188, -0.1946, -0.2101, -0.2214, -0.2214, -0.1908, -0.1878, -0.2078, -0.1847, -0.1648, -0.1523, -0.0829, -0.0659, -0.0659, -0.0884, -0.1, -0.0974, -0.0866, -0.066, -0.0647, -0.1001, -0.0823, -0.065, -0.0718, -0.065, -0.0763, -0.0883, -0.074, -0.035, -0.0372, -0.0325, -0.0325, -0.0032, -0.0181, -0.0415, -0.0454, -0.0426, -0.0585, -0.0783, -0.0823, -0.0778, -0.0675, -0.0407, -0.0581, -0.0447, -0.0389, -0.0267, -0.0437, -0.0243, -0.0051, -0.0082, 0.0, 0.0193, 0.0217, 0.031, 0.0236, 0.0067, 0.015, 0.015, 0.0282, 0.0272, 0.0272, 0.0013, -0.034, -0.0749, -0.0702, -0.0569, -0.0438, -0.033, -0.0666, -0.0891, -0.0903, -0.0486, -0.0752, -0.0703, -0.0443, -0.0491, -0.0089, -0.0132, -0.0132, -0.0121, -0.012, -0.0028, -0.019, -0.0257, -0.0069, -0.0121, -0.022, -0.0325, -0.0213, -0.0578, -0.0658, -0.1075, -0.1507, -0.1571, -0.149, -0.1356, -0.1353, -0.1422, -0.1397, -0.1442, -0.1201, -0.1315, -0.1201, -0.0922, -0.1109, -0.1109, -0.0891, -0.097, -0.1021, -0.1197, -0.1096, -0.1096, -0.0958, -0.0773, -0.0799, -0.0815, -0.0961, -0.109, -0.1058, -0.137, -0.1218, -0.1342, -0.1238, -0.1036, -0.1252, -0.081, -0.0503, -0.078, -0.0993, -0.0973, -0.0759, -0.0764, -0.0736, -0.0722, -0.0751, -0.0672, -0.0786, -0.0769, -0.0858, -0.0858, -0.1013, -0.0869, -0.0237, -0.0237, -0.0488, -0.0734, -0.111, -0.1167, -0.1131, -0.1458, -0.1143, -0.0958, -0.0624, -0.0376, -0.0209, -0.046, -0.0357, -0.0097, -0.0163, 0.0108, 0.0136, 0.0108, -0.0106, -0.0144, -0.022, -0.022, -0.0298, -0.0298, -0.0298, -0.0349, -0.0117, -0.0104, -0.0104, -0.0104, -0.0104, -0.0098, 0.0168]}
{"unit_id": "orig:1932939577234501921", "ticker_idx": 0, "ts": "2025-06-11T23:14:42+00:00", "symbol": "CRWV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "shareholder returns will essentially be wiped out", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Shares bearish sell-side report on CoreWeave: debt financing wipes shareholder returns and GPU residual value assumptions are unrealistic.", "resolved_tickers": ["CRWV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "I’d like to share a sell-side report on CoreWeave.\nThere are many points I agree with, so I thought you might find it worth reading.\n\nBrannin McBee, Chief Development Officer (CDO), explained in an interview with Barron’s last month, “Our model is not one where we purchase infrastructure first and wait for customers to come. We sign ‘take-or-pay’ contracts with customers and then invest in GPUs, data centers, and other infrastructure based on those contracts.”  \n\nHowever, analysts Gil Luria and Alexander Flat pointed out, “The example contract presented by CoreWeave actually highlights how this model makes it difficult to generate shareholder returns.”  \n\nThey also analyzed that “In the contract example, CoreWeave assumes it will finance 15% of the infrastructure purchase cost with its own equity. But this is an unrealistic scenario that does not reflect the company’s actual capital capacity.”  \n\n“In reality, the company will inevitably have to rely heavily on other forms of debt financing. Considering the average interest rate of 9.5%, shareholder returns will essentially be wiped out,” they explained.  \n\nThey further raised doubts about the residual value CoreWeave assigned to its GPU assets after the contracts expire. The company estimated that the residual value would amount to 75% of the total contract revenue, but the analysts criticized this figure as highly unrealistic. They added, “Four years from now, the revenue per GPU will likely be less than 25% of today’s level.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03473611926685649, "ret_signed_m1d": -0.03473611926685649, "alpha_spy_m1d": -0.031876047997083434, "alpha_c_m1d": -0.031876047997083434, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.002860071269773057, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004876391064934094, "ret_signed_p1d": 0.004876391064934094, "alpha_spy_p1d": 0.00885061716208857, "alpha_c_p1d": 0.00885061716208857, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.003974226097154476, "ret_p1w": 0.13560456556840417, "ret_signed_p1w": -0.13560456556840417, "alpha_spy_p1w": -0.1421230847865227, "alpha_c_p1w": -0.1421230847865227, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": -0.006518519218118524, "ret_p1m": -0.07621913089319776, "ret_signed_p1m": 0.07621913089319776, "alpha_spy_p1m": 0.11997011128374424, "alpha_c_p1m": 0.11997011128374424, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.04375098039054648, "ret_p3m": -0.3750834672087393, "ret_signed_p3m": 0.3750834672087393, "alpha_spy_p3m": 0.45721073196223116, "alpha_c_p3m": 0.45721073196223116, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.08212726475349186, "ret_p6m": -0.4271876970735844, "ret_signed_p6m": 0.4271876970735844, "alpha_spy_p6m": 0.571786987332424, "alpha_c_p6m": 0.571786987332424, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.14459929025883955, "price_path": [null, null, null, null, null, null, null, null, null, null, -0.7328, -0.7523, -0.6488, -0.5901, -0.6406, -0.6806, -0.667, -0.7087, -0.6742, -0.7181, -0.7077, -0.7087, -0.7276, -0.7279, -0.7389, -0.7389, -0.7634, -0.7428, -0.722, -0.721, -0.7225, -0.7193, -0.714, -0.7241, -0.7039, -0.6555, -0.6593, -0.6373, -0.642, -0.6326, -0.6568, -0.6093, -0.5774, -0.5494, -0.5607, -0.4636, -0.4216, -0.3972, -0.2826, -0.3309, -0.3137, -0.3137, -0.1719, -0.2241, -0.2949, -0.2564, -0.1971, 0.0052, 0.0895, -0.0979, -0.0637, 0.0828, 0.0347, 0.0, -0.0049, -0.0168, 0.0588, 0.1485, 0.1356, 0.1356, 0.2263, 0.1602, 0.1531, 0.0655, 0.056, 0.0687, 0.0892, 0.0417, 0.0138, 0.1035, 0.1035, 0.0668, 0.0117, 0.0224, -0.0762, -0.1594, -0.1157, -0.0609, -0.0445, -0.1168, -0.1782, -0.1656, -0.1331, -0.158, -0.1984, -0.2277, -0.2633, -0.2736, -0.3127, -0.2376, -0.3043, -0.2919, -0.2529, -0.2636, -0.1912, -0.1346, -0.0663, -0.0063, -0.2134, -0.3353, -0.3322, -0.3534, -0.3795, -0.3886, -0.3935, -0.3721, -0.3829, -0.3895, -0.3525, -0.3134, -0.3117, -0.3117, -0.3765, -0.3996, -0.4156, -0.4049, -0.3751, -0.3305, -0.2175, -0.2472, -0.2521, -0.1953, -0.2067, -0.1927, -0.1891, -0.1659, -0.11, -0.1257, -0.1089, -0.1539, -0.1961, -0.1816, -0.0858, -0.0845, -0.0782, -0.0996, -0.1059, -0.1394, -0.0649, -0.0442, -0.0753, -0.054, -0.1045, -0.0699, -0.0532, -0.0857, -0.1512, -0.1646, -0.1882, -0.1761, -0.1146, -0.0911, -0.0995, -0.0653, -0.1245, -0.1068, -0.1562, -0.2268, -0.2357, -0.2857, -0.3052, -0.2945, -0.4096, -0.4293, -0.4767, -0.4832, -0.4968, -0.4997, -0.4995, -0.5377, -0.5214, -0.5084, -0.5238, -0.5037, -0.5037, -0.5116, -0.4852, -0.4921, -0.4699, -0.4272]}
{"unit_id": "orig:1943319836609876036", "ticker_idx": 0, "ts": "2025-07-10T14:42:08+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "recommends the Big 3 equipment stocks, with a rating order of KLAC > LRCX > AMAT. KLAC's FY26 EPS estimate is raised by 3% due to increased market share, and its target price is raised to $1,035", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi upgrades WFE outlook and Big 3 equipment stocks KLAC ($1,035 PT), LRCX ($108 PT), AMAT ($220 PT) on eased geopolitical risk, AI capex, and memory demand recovery.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "US WFE: Geopolitical Risks Temporarily Eased, WFE Outlook Upgraded, AI and Memory Demand Jointly Drive Upturn Cycle\nCiti (A. Malik, July 07, 2025)\n\nCiti notes that recent consensus reached between the U.S. and China has unusually alleviated concerns about equipment embargoes and raw material export controls, bringing temporary certainty to WFE investment. Driven by both AI capital expenditure and memory demand recovery, Citi has upgraded its 2025/26 WFE outlook and recommends the Big 3 equipment stocks, with a rating order of KLAC > LRCX > AMAT.\n\nChina Restriction Risks Lifted, Industry's Biggest Uncertainty Temporarily Eased\nCiti states that one of the biggest risks suppressing valuations in 1H 2025, namely the escalation of U.S.-China policies, has been temporarily removed. China's reinstatement of rare earth magnet export licenses for six months and the U.S.'s lifting of EDA sanctions, among other \"countermeasures,\" have temporarily cleared the risk of broad equipment embargoes. Even if the U.S. were to revoke equipment exemptions for Samsung and TSMC in China in the future, the impact on WFE would be relatively controllable, as U.S. equipment suppliers have a higher dependence on Chinese clients. Based on this, Citi has upgraded the industry framework from Phase 1 to Phase 2, representing a release of medium-term risk.\n\nWFE Forecast Upgraded, 2025 Stabilizes, 2026 Enters Upturn Cycle\nCiti has raised its 2025/26 global WFE forecast to $100 billion/$110 billion, with annual growth rates of +1%/+10% respectively, a significant improvement from previous expectations of -5%/+2%. The main drivers are the rebound in memory prices (upward revision of 3Q25 DRAM/NAND price estimates) and a high capital expenditure cycle driven by AI, particularly an upward revision in NAND equipment capital expenditure. In the medium term, to support the continued growth of the AI semiconductor TAM until 2030, the industry needs an additional $15 billion to $20 billion in WFE investment annually.\n\nBig 3 Ranking Adjusted, KLAC's Leading Logic Stronger, Price Targets Fully Upgraded\nCiti believes that manufacturers with the strongest drivers from leading-edge processes and advanced packaging will have the greatest upside elasticity, in the order of KLAC > LRCX > AMAT. Specific adjustments are as follows: KLAC's FY26 EPS estimate is raised by 3% due to increased market share, and its target price is raised to $1,035 (28x P/E); LRCX's FY25/26 EPS is raised by 2%/16% respectively, and its target price is raised to $108 (23x P/E); AMAT's FY25/26 EPS is raised by 2%/8%, and its target price is raised to $220 (20.5x P/E), fully reflecting a more positive WFE outlook.\n\nStrong Correlation with Cloud CapEx, 2026 WFE Growth Logic Supported by Data\nOver the past decade, WFE has shown a high positive correlation (correlation coefficient ~0.88) with the capital expenditure of top cloud providers. Citi forecasts that the Big 4 cloud providers' capital expenditure will grow by 35%/15% in 2025/26 respectively, driving WFE's synchronized upward trend. Current Chinese customs data shows a significant increase in imports of deposition and etch equipment, mainly from Malaysia (presumed to be Lam) and South Korea, indicating that memory capital expenditure has gradually started.\n\nMacroeconomic and Policy Factors Remain Key Risks\nCiti cautions that a macroeconomic slowdown and policy reversals remain major risks. Current forecasts for global GDP growth are 2.4% in 2025, lower than 2.8% in 2024, with a modest recovery to 2.5% in 2026. Furthermore, if a future Trump administration takes office and tightens policies towards China/Taiwan again, it could also impact the medium-to-long term WFE trend and equipment stock valuations.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005858010390632384, "ret_signed_m1d": -0.005858010390632384, "alpha_spy_m1d": -0.003045631951938188, "alpha_c_m1d": 0.0014456863885138338, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00435044391515349, "ret_signed_p1d": -0.00435044391515349, "alpha_spy_p1d": -0.0008350200445579326, "alpha_c_p1d": -0.004246102347708947, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": 0.00911029234617966, "ret_signed_p1w": 0.00911029234617966, "alpha_spy_p1w": 0.005563001279174529, "alpha_c_p1w": -0.005462604612263666, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": -0.014882174981170282, "ret_signed_p1m": -0.014882174981170282, "alpha_spy_p1m": -0.033034280847144215, "alpha_c_p1m": -0.035785091540922354, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.17066274315711016, "ret_signed_p3m": 0.17066274315711016, "alpha_spy_p3m": 0.09850931244195227, "alpha_c_p3m": -0.0016035876290823037, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.37772894999946427, "ret_signed_p6m": 0.37772894999946427, "alpha_spy_p6m": 0.27982810320740414, "alpha_c_p6m": 0.07534453224138127, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.2803, -0.2736, -0.3101, -0.3184, -0.3184, -0.3331, -0.3164, -0.2941, -0.2594, -0.2538, -0.2548, -0.2609, -0.2451, -0.2735, -0.25, -0.2552, -0.2701, -0.2541, -0.2444, -0.2467, -0.183, -0.145, -0.1352, -0.1341, -0.1501, -0.1525, -0.149, -0.1612, -0.1735, -0.1846, -0.1846, -0.1503, -0.1627, -0.1665, -0.1849, -0.179, -0.1646, -0.1578, -0.1472, -0.1299, -0.107, -0.0776, -0.061, -0.0577, -0.0656, -0.039, -0.0384, -0.0619, -0.0619, -0.0847, -0.0779, -0.0426, -0.0379, -0.0277, -0.0417, -0.0354, -0.0321, -0.0081, -0.0044, -0.0044, -0.0172, -0.0101, -0.0059, 0.0, -0.0044, -0.0073, 0.0085, 0.0052, 0.0091, 0.0027, 0.0098, -0.0392, -0.034, -0.0263, -0.0286, -0.0059, -0.0135, -0.0039, -0.0534, -0.0452, -0.014, -0.0487, -0.0434, -0.0178, -0.0149, -0.0199, 0.0074, 0.0225, 0.0288, -0.0578, -0.0466, -0.0545, -0.052, -0.0585, -0.0608, -0.0508, -0.0416, -0.0407, -0.0352, -0.0589, -0.0589, -0.0866, -0.0893, -0.0575, -0.0232, -0.019, -0.0096, 0.0065, 0.0353, 0.0404, 0.0672, 0.069, 0.0683, 0.1296, 0.1276, 0.1561, 0.1561, 0.1533, 0.143, 0.1486, 0.1484, 0.164, 0.2183, 0.2295, 0.1888, 0.23, 0.1707, 0.1468, 0.1369, 0.0606, 0.1062, 0.107, 0.1731, 0.1858, 0.1943, 0.2442, 0.2383, 0.2026, 0.2508, 0.2765, 0.3114, 0.3016, 0.3331, 0.3106, 0.3045, 0.3157, 0.288, 0.3243, 0.302, 0.2879, 0.3144, 0.2852, 0.2939, 0.2537, 0.2242, 0.2256, 0.2141, 0.262, 0.1918, 0.186, 0.2288, 0.2387, 0.253, 0.253, 0.2707, 0.2509, 0.2863, 0.3099, 0.306, 0.3129, 0.3238, 0.3249, 0.3393, 0.3471, 0.2907, 0.3244, 0.3225, 0.267, 0.3214, 0.3466, 0.3682, 0.3715, 0.3805, 0.3805, 0.3833, 0.3625, 0.3444, 0.3135, 0.3135, 0.3777]}
{"unit_id": "orig:1943319836609876036", "ticker_idx": 1, "ts": "2025-07-10T14:42:08+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "LRCX's FY25/26 EPS is raised by 2%/16% respectively, and its target price is raised to $108", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi upgrades WFE outlook and Big 3 equipment stocks KLAC ($1,035 PT), LRCX ($108 PT), AMAT ($220 PT) on eased geopolitical risk, AI capex, and memory demand recovery.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "US WFE: Geopolitical Risks Temporarily Eased, WFE Outlook Upgraded, AI and Memory Demand Jointly Drive Upturn Cycle\nCiti (A. Malik, July 07, 2025)\n\nCiti notes that recent consensus reached between the U.S. and China has unusually alleviated concerns about equipment embargoes and raw material export controls, bringing temporary certainty to WFE investment. Driven by both AI capital expenditure and memory demand recovery, Citi has upgraded its 2025/26 WFE outlook and recommends the Big 3 equipment stocks, with a rating order of KLAC > LRCX > AMAT.\n\nChina Restriction Risks Lifted, Industry's Biggest Uncertainty Temporarily Eased\nCiti states that one of the biggest risks suppressing valuations in 1H 2025, namely the escalation of U.S.-China policies, has been temporarily removed. China's reinstatement of rare earth magnet export licenses for six months and the U.S.'s lifting of EDA sanctions, among other \"countermeasures,\" have temporarily cleared the risk of broad equipment embargoes. Even if the U.S. were to revoke equipment exemptions for Samsung and TSMC in China in the future, the impact on WFE would be relatively controllable, as U.S. equipment suppliers have a higher dependence on Chinese clients. Based on this, Citi has upgraded the industry framework from Phase 1 to Phase 2, representing a release of medium-term risk.\n\nWFE Forecast Upgraded, 2025 Stabilizes, 2026 Enters Upturn Cycle\nCiti has raised its 2025/26 global WFE forecast to $100 billion/$110 billion, with annual growth rates of +1%/+10% respectively, a significant improvement from previous expectations of -5%/+2%. The main drivers are the rebound in memory prices (upward revision of 3Q25 DRAM/NAND price estimates) and a high capital expenditure cycle driven by AI, particularly an upward revision in NAND equipment capital expenditure. In the medium term, to support the continued growth of the AI semiconductor TAM until 2030, the industry needs an additional $15 billion to $20 billion in WFE investment annually.\n\nBig 3 Ranking Adjusted, KLAC's Leading Logic Stronger, Price Targets Fully Upgraded\nCiti believes that manufacturers with the strongest drivers from leading-edge processes and advanced packaging will have the greatest upside elasticity, in the order of KLAC > LRCX > AMAT. Specific adjustments are as follows: KLAC's FY26 EPS estimate is raised by 3% due to increased market share, and its target price is raised to $1,035 (28x P/E); LRCX's FY25/26 EPS is raised by 2%/16% respectively, and its target price is raised to $108 (23x P/E); AMAT's FY25/26 EPS is raised by 2%/8%, and its target price is raised to $220 (20.5x P/E), fully reflecting a more positive WFE outlook.\n\nStrong Correlation with Cloud CapEx, 2026 WFE Growth Logic Supported by Data\nOver the past decade, WFE has shown a high positive correlation (correlation coefficient ~0.88) with the capital expenditure of top cloud providers. Citi forecasts that the Big 4 cloud providers' capital expenditure will grow by 35%/15% in 2025/26 respectively, driving WFE's synchronized upward trend. Current Chinese customs data shows a significant increase in imports of deposition and etch equipment, mainly from Malaysia (presumed to be Lam) and South Korea, indicating that memory capital expenditure has gradually started.\n\nMacroeconomic and Policy Factors Remain Key Risks\nCiti cautions that a macroeconomic slowdown and policy reversals remain major risks. Current forecasts for global GDP growth are 2.4% in 2025, lower than 2.8% in 2024, with a modest recovery to 2.5% in 2026. Furthermore, if a future Trump administration takes office and tightens policies towards China/Taiwan again, it could also impact the medium-to-long term WFE trend and equipment stock valuations.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012368965574004132, "ret_signed_m1d": -0.012368965574004132, "alpha_spy_m1d": -0.009556587135309935, "alpha_c_m1d": -0.005065268794857913, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0066296939376260244, "ret_signed_p1d": 0.0066296939376260244, "alpha_spy_p1d": 0.010145117808221582, "alpha_c_p1d": 0.006734035505070568, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.002671646194210786, "ret_signed_p1w": -0.002671646194210786, "alpha_spy_p1w": -0.006218937261215918, "alpha_c_p1w": -0.017244543152654113, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.006827607703510763, "ret_signed_p1m": 0.006827607703510763, "alpha_spy_p1m": -0.01132449816246317, "alpha_c_p1m": -0.014075308856241309, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.3915212953770788, "ret_signed_p3m": 0.3915212953770788, "alpha_spy_p3m": 0.3193678646619209, "alpha_c_p3m": 0.21925496459088634, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.8378260162765676, "ret_signed_p6m": 0.8378260162765676, "alpha_spy_p6m": 0.7399251694845075, "alpha_c_p6m": 0.5354415985184846, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.3302, -0.3245, -0.357, -0.3707, -0.3707, -0.3842, -0.3734, -0.3413, -0.3002, -0.295, -0.2936, -0.3002, -0.2926, -0.292, -0.2644, -0.2705, -0.2779, -0.2591, -0.2607, -0.2562, -0.1881, -0.1585, -0.1613, -0.1631, -0.1666, -0.1727, -0.168, -0.1828, -0.1851, -0.1999, -0.1999, -0.1715, -0.1709, -0.1693, -0.2026, -0.1859, -0.1742, -0.1633, -0.1619, -0.1477, -0.1284, -0.1018, -0.1023, -0.0953, -0.1164, -0.078, -0.0854, -0.0873, -0.0873, -0.1046, -0.0935, -0.0537, -0.0499, -0.0418, -0.0382, -0.0368, -0.0421, -0.0221, -0.0223, -0.0223, -0.0289, -0.0122, -0.0124, 0.0, 0.0066, -0.0142, 0.0001, -0.0068, -0.0027, -0.004, 0.0067, -0.0333, -0.0392, -0.0325, -0.0406, -0.0241, -0.021, -0.0195, -0.0615, -0.0464, -0.0262, -0.0433, -0.0507, -0.0189, 0.0068, 0.0093, 0.0418, 0.0562, 0.0625, -0.0153, -0.0216, -0.0072, -0.0189, -0.0262, -0.0097, 0.0022, 0.0254, 0.0258, 0.03, -0.009, -0.009, -0.0399, -0.0329, -0.0063, 0.0187, 0.0397, 0.0446, 0.0623, 0.1437, 0.1573, 0.1796, 0.1921, 0.2062, 0.25, 0.2559, 0.3081, 0.3055, 0.2722, 0.2704, 0.2723, 0.2997, 0.3276, 0.4157, 0.4574, 0.4457, 0.4788, 0.3915, 0.4132, 0.398, 0.3025, 0.3663, 0.3713, 0.4354, 0.4115, 0.403, 0.4282, 0.438, 0.4004, 0.4628, 0.5039, 0.5556, 0.5429, 0.593, 0.5964, 0.5612, 0.5986, 0.5445, 0.6364, 0.6081, 0.5799, 0.6495, 0.5782, 0.6004, 0.5201, 0.4699, 0.462, 0.4202, 0.4753, 0.384, 0.4143, 0.491, 0.5063, 0.5382, 0.5382, 0.5467, 0.5347, 0.5684, 0.5865, 0.5601, 0.576, 0.6162, 0.6467, 0.671, 0.6755, 0.5941, 0.6317, 0.6213, 0.5391, 0.6356, 0.7108, 0.7405, 0.7395, 0.7611, 0.7611, 0.7684, 0.7466, 0.7258, 0.7, 0.7, 0.8378]}
{"unit_id": "orig:1943319836609876036", "ticker_idx": 2, "ts": "2025-07-10T14:42:08+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMAT's FY25/26 EPS is raised by 2%/8%, and its target price is raised to $220", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi upgrades WFE outlook and Big 3 equipment stocks KLAC ($1,035 PT), LRCX ($108 PT), AMAT ($220 PT) on eased geopolitical risk, AI capex, and memory demand recovery.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "US WFE: Geopolitical Risks Temporarily Eased, WFE Outlook Upgraded, AI and Memory Demand Jointly Drive Upturn Cycle\nCiti (A. Malik, July 07, 2025)\n\nCiti notes that recent consensus reached between the U.S. and China has unusually alleviated concerns about equipment embargoes and raw material export controls, bringing temporary certainty to WFE investment. Driven by both AI capital expenditure and memory demand recovery, Citi has upgraded its 2025/26 WFE outlook and recommends the Big 3 equipment stocks, with a rating order of KLAC > LRCX > AMAT.\n\nChina Restriction Risks Lifted, Industry's Biggest Uncertainty Temporarily Eased\nCiti states that one of the biggest risks suppressing valuations in 1H 2025, namely the escalation of U.S.-China policies, has been temporarily removed. China's reinstatement of rare earth magnet export licenses for six months and the U.S.'s lifting of EDA sanctions, among other \"countermeasures,\" have temporarily cleared the risk of broad equipment embargoes. Even if the U.S. were to revoke equipment exemptions for Samsung and TSMC in China in the future, the impact on WFE would be relatively controllable, as U.S. equipment suppliers have a higher dependence on Chinese clients. Based on this, Citi has upgraded the industry framework from Phase 1 to Phase 2, representing a release of medium-term risk.\n\nWFE Forecast Upgraded, 2025 Stabilizes, 2026 Enters Upturn Cycle\nCiti has raised its 2025/26 global WFE forecast to $100 billion/$110 billion, with annual growth rates of +1%/+10% respectively, a significant improvement from previous expectations of -5%/+2%. The main drivers are the rebound in memory prices (upward revision of 3Q25 DRAM/NAND price estimates) and a high capital expenditure cycle driven by AI, particularly an upward revision in NAND equipment capital expenditure. In the medium term, to support the continued growth of the AI semiconductor TAM until 2030, the industry needs an additional $15 billion to $20 billion in WFE investment annually.\n\nBig 3 Ranking Adjusted, KLAC's Leading Logic Stronger, Price Targets Fully Upgraded\nCiti believes that manufacturers with the strongest drivers from leading-edge processes and advanced packaging will have the greatest upside elasticity, in the order of KLAC > LRCX > AMAT. Specific adjustments are as follows: KLAC's FY26 EPS estimate is raised by 3% due to increased market share, and its target price is raised to $1,035 (28x P/E); LRCX's FY25/26 EPS is raised by 2%/16% respectively, and its target price is raised to $108 (23x P/E); AMAT's FY25/26 EPS is raised by 2%/8%, and its target price is raised to $220 (20.5x P/E), fully reflecting a more positive WFE outlook.\n\nStrong Correlation with Cloud CapEx, 2026 WFE Growth Logic Supported by Data\nOver the past decade, WFE has shown a high positive correlation (correlation coefficient ~0.88) with the capital expenditure of top cloud providers. Citi forecasts that the Big 4 cloud providers' capital expenditure will grow by 35%/15% in 2025/26 respectively, driving WFE's synchronized upward trend. Current Chinese customs data shows a significant increase in imports of deposition and etch equipment, mainly from Malaysia (presumed to be Lam) and South Korea, indicating that memory capital expenditure has gradually started.\n\nMacroeconomic and Policy Factors Remain Key Risks\nCiti cautions that a macroeconomic slowdown and policy reversals remain major risks. Current forecasts for global GDP growth are 2.4% in 2025, lower than 2.8% in 2024, with a modest recovery to 2.5% in 2026. Furthermore, if a future Trump administration takes office and tightens policies towards China/Taiwan again, it could also impact the medium-to-long term WFE trend and equipment stock valuations.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01333123039850248, "ret_signed_m1d": -0.01333123039850248, "alpha_spy_m1d": -0.010518851959808284, "alpha_c_m1d": -0.006027533619356262, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0005049061666235755, "ret_signed_p1d": -0.0005049061666235755, "alpha_spy_p1d": 0.003010517703971982, "alpha_c_p1d": -0.00040056459917903187, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.027824016643818505, "ret_signed_p1w": -0.027824016643818505, "alpha_spy_p1w": -0.03137130771082364, "alpha_c_p1w": -0.04239691360226183, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": -0.06645449999852526, "ret_signed_p1m": -0.06645449999852526, "alpha_spy_p1m": -0.0846066058644992, "alpha_c_p1m": -0.08735741655827733, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.0713848834897004, "ret_signed_p3m": 0.0713848834897004, "alpha_spy_p3m": -0.0007685472254574943, "alpha_c_p3m": -0.10088144729649207, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.3642840942837171, "ret_signed_p6m": 0.3642840942837171, "alpha_spy_p6m": 0.266383247491657, "alpha_c_p6m": 0.061899676525634106, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.2722, -0.2676, -0.3042, -0.3078, -0.3078, -0.3174, -0.3036, -0.277, -0.2438, -0.2369, -0.2407, -0.247, -0.2411, -0.2499, -0.219, -0.2215, -0.2294, -0.2154, -0.2145, -0.2164, -0.1541, -0.1287, -0.1231, -0.1201, -0.1663, -0.1644, -0.1655, -0.1831, -0.1894, -0.2046, -0.2046, -0.1828, -0.184, -0.1947, -0.2085, -0.2058, -0.1833, -0.1823, -0.1709, -0.158, -0.1426, -0.1225, -0.1275, -0.1163, -0.1386, -0.1085, -0.1209, -0.1272, -0.1272, -0.1443, -0.1316, -0.0901, -0.0755, -0.0733, -0.0748, -0.0755, -0.0721, -0.0405, -0.0352, -0.0352, -0.0366, -0.0154, -0.0133, 0.0, -0.0005, -0.0047, 0.0064, -0.0163, -0.0278, -0.0383, -0.0274, -0.055, -0.0556, -0.05, -0.0623, -0.0392, -0.0486, -0.0436, -0.0907, -0.0911, -0.0768, -0.0953, -0.1004, -0.0751, -0.0665, -0.0689, -0.0484, -0.0404, -0.0494, -0.1832, -0.1742, -0.1808, -0.1872, -0.1905, -0.1771, -0.1796, -0.1669, -0.1675, -0.163, -0.1859, -0.1859, -0.202, -0.2087, -0.1986, -0.1758, -0.1793, -0.172, -0.1724, -0.1383, -0.1502, -0.1344, -0.1212, -0.0979, -0.039, -0.0373, 0.0155, 0.0172, 0.0201, 0.0108, 0.0327, 0.0379, 0.0368, 0.1027, 0.1323, 0.1016, 0.1339, 0.0714, 0.1015, 0.1156, 0.0632, 0.1115, 0.105, 0.1525, 0.1532, 0.1394, 0.1553, 0.1445, 0.117, 0.157, 0.1584, 0.1715, 0.1528, 0.1939, 0.1777, 0.1805, 0.2038, 0.1657, 0.2199, 0.1826, 0.1651, 0.1905, 0.158, 0.1685, 0.1305, 0.1446, 0.1582, 0.1401, 0.1907, 0.1175, 0.1367, 0.1717, 0.2303, 0.2684, 0.2684, 0.28, 0.2926, 0.3463, 0.3631, 0.3672, 0.3599, 0.3607, 0.3555, 0.3961, 0.3706, 0.3153, 0.3257, 0.3134, 0.2598, 0.2863, 0.3011, 0.3143, 0.3204, 0.3232, 0.3232, 0.3289, 0.3348, 0.3191, 0.304, 0.304, 0.3643]}
{"unit_id": "orig:1927959902258950159", "ticker_idx": 0, "ts": "2025-05-29T05:27:15+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS maintains NVIDIA's target price at $175", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "UBS maintains NVDA Buy/target at $175; author shares the bullish rating.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS maintains NVIDIA’s target price at $175 https://t.co/HjcdluQPyP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031467771192964955, "ret_signed_m1d": -0.031467771192964955, "alpha_spy_m1d": -0.027536013589865016, "alpha_c_m1d": -0.02442091775738242, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029168709381678815, "ret_signed_p1d": -0.029168709381678815, "alpha_spy_p1d": -0.028050357189775443, "alpha_c_p1d": -0.011428601389960602, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.005747599702493611, "ret_signed_p1w": 0.005747599702493611, "alpha_spy_p1w": 0.0006632770562644552, "alpha_c_p1w": -0.023832855030635036, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.13342157350479011, "ret_signed_p1m": 0.13342157350479011, "alpha_spy_p1m": 0.08820871516440842, "alpha_c_p1m": -0.007270048081713831, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.30600340555452554, "ret_signed_p3m": 0.30600340555452554, "alpha_spy_p3m": 0.20937208449485212, "alpha_c_p3m": 0.09021349223987807, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.2853115202986203, "ret_signed_p6m": 0.2853115202986203, "alpha_spy_p6m": 0.16199846496754655, "alpha_c_p6m": -0.05084882377385047, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [-0.1806, -0.1668, -0.1573, -0.2057, -0.1905, -0.2315, -0.2187, -0.1685, -0.1696, -0.1259, -0.1412, -0.1707, -0.1557, -0.1484, -0.1544, -0.1277, -0.1329, -0.1827, -0.1994, -0.2121, -0.2214, -0.2086, -0.2067, -0.2686, -0.3224, -0.2985, -0.3081, -0.1786, -0.2272, -0.203, -0.2046, -0.1939, -0.2493, -0.2709, -0.2709, -0.3038, -0.2895, -0.2621, -0.2354, -0.2025, -0.2188, -0.2168, -0.2175, -0.1981, -0.1774, -0.1823, -0.1843, -0.159, -0.1568, -0.1619, -0.1163, -0.0665, -0.0277, -0.0313, -0.0272, -0.026, -0.0346, -0.0531, -0.0457, -0.0568, -0.0568, -0.0265, -0.0315, 0.0, -0.0292, -0.013, 0.0146, 0.0196, 0.0057, 0.0182, 0.0247, 0.0343, 0.0262, 0.0418, 0.02, 0.0396, 0.0355, 0.0453, 0.0453, 0.0336, 0.0359, 0.0627, 0.1087, 0.1138, 0.1334, 0.1351, 0.1014, 0.1298, 0.1448, 0.1448, 0.1369, 0.1496, 0.1703, 0.179, 0.1849, 0.1788, 0.2265, 0.2313, 0.243, 0.2388, 0.2314, 0.2001, 0.227, 0.2483, 0.2466, 0.2699, 0.261, 0.288, 0.278, 0.2482, 0.2933, 0.2808, 0.2891, 0.2988, 0.3127, 0.3081, 0.316, 0.3047, 0.3078, 0.2965, 0.3077, 0.262, 0.2602, 0.2572, 0.2788, 0.2919, 0.306, 0.3048, 0.2945, 0.2515, 0.2515, 0.227, 0.2259, 0.2334, 0.2, 0.2093, 0.2269, 0.2741, 0.273, 0.2777, 0.2772, 0.2566, 0.2236, 0.2663, 0.2694, 0.3193, 0.2821, 0.2716, 0.2768, 0.2804, 0.3067, 0.3406, 0.3454, 0.3572, 0.3481, 0.3332, 0.3296, 0.3588, 0.3837, 0.3161, 0.3531, 0.2936, 0.2921, 0.3064, 0.3165, 0.3123, 0.3017, 0.2954, 0.3089, 0.3383, 0.3759, 0.4445, 0.4877, 0.4578, 0.455, 0.4865, 0.4277, 0.4026, 0.3514, 0.3519, 0.4302, 0.3879, 0.3925, 0.3427, 0.3664, 0.3408, 0.3031, 0.3402, 0.298, 0.2853]}
{"unit_id": "orig:1938958037278175282", "ticker_idx": 0, "ts": "2025-06-28T13:49:54+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "overall supply and demand will balance out in the second half of 2026, with CoWoS-L accounting for 64% in 2026 and a significant increase in ASP", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM bullish TSMC analysis: CoWoS/WMCM capacity ramp, ASP uplift, and Apple WMCM adoption drive multi-year revenue growth.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC: CoWoS and WMCM Capacity Expansion Rebalancing, AI Demand Upgrade Drives Packaging Diversification\n\nJPM (G. Hariharan, 2025/06/26)\n\nTSMC's CoWoS and WMCM capacity plans for 2025-2027 show dynamic adjustments in response to AI platform evolution, customer diversification, and packaging process optimization. Overall supply and demand are expected to balance out in 2026.\n\nCoWoS Capacity Continues to Grow Through 2027, But Short-Term Tightness in 2025 Due to Weak China Demand\n\nThe agency expects TSMC's CoWoS capacity to reach 69k wfpm by the end of 2025, rising to 84k/102k wfpm in 2026/2027, slightly lower than previous expectations. This is primarily due to U.S.-China export restrictions and declining demand from Chinese customers (e.g., NVDA H20). Nevertheless, driven by CoWoS-L migration and customer structure diversification, overall supply and demand will balance out in the second half of 2026, with CoWoS-L accounting for 64% in 2026 and a significant increase in ASP.\n\nNVDA Dominates AI GPU Packaging Demand, But Unit Shipment Growth Slows in 2026\n\nThe introduction of the Rubin platform and Vera CPU will drive a 25% increase in NVDA's CoWoS demand in 2026. However, due to Blackwell inventory adjustments and a void in the Chinese market, GPU shipments are not expected to grow year-over-year. Starting in 2027, demand will be boosted by Rubin Ultra (80% increase in package area) and AI ASIC demand. AMD's momentum is weaker in 2025-2026, with the MI450 series possibly beginning larger-scale shipments in the second half of 2026.\n\nASIC Customer Structure Diversifies, Broadcom and AWS Drive 2026 Growth\n\nBroadcom continues to benefit from its dominant position in the Google TPU project, while Meta will begin its first large-scale mass production of CoWoS accelerators in 2026. AWS's Trainium 3, supported by Alchip, will ramp up in 2026 and continue to scale in 2027. MediaTek's 2026 outlook is impacted by delays in the TPUv7e project, but TPUv8e is still a collaboration with Google, potentially offering future upside.\n\nNon-AI Processors Gradually Adopt CoWoS, Boosting Packaging Demand for CPUs and Networking Products\n\nCoWoS applications have expanded to non-AI processors, such as NVDA Vera CPU (using CoWoS-R) and AMD Venice Server CPU (combining HBM), which will adopt high-density packaging solutions. As interconnection bottlenecks between logic and memory intensify, more processor categories are expected to adopt CoWoS and its derivatives in the next 2 to 3 years.\n\nApple's WMCM Packaging to Ramp Up Significantly in 2026, Driving InFO to WMCM Conversion\n\nApple will adopt WMCM packaging (featuring the A20 Pro chip, N2 process) in all high-end iPhones (including foldables) starting in 2026. Estimated WMCM capacity is 27k wfpm by the end of 2026, reaching 40k wfpm in 2027, with long-term potential demand of up to 90k wfpm. WMCM is more expensive than InFO (approximately $2k-$2.5k per wafer) but offers simplified processes, positively impacting TSMC's revenue. InFO capacity will be reduced by more than half in the future, with some production lines being converted to WMCM.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00918365939265664, "ret_signed_m1d": 0.00918365939265664, "alpha_spy_m1d": 0.013942213457711072, "alpha_c_m1d": 0.010833056927868845, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007991564135397589, "ret_signed_p1d": -0.007991564135397589, "alpha_spy_p1d": -0.007667884827172733, "alpha_c_p1d": 0.0029092126269995022, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": 0.01183267690623624, "ret_signed_p1w": 0.01183267690623624, "alpha_spy_p1w": 0.00725235069049357, "alpha_c_p1w": 0.008677131122611703, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": 0.06552167894744954, "ret_signed_p1m": 0.06552167894744954, "alpha_spy_p1m": 0.03734344751166341, "alpha_c_p1m": 0.015249192459260152, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.22536451428302318, "ret_signed_p3m": 0.22536451428302318, "alpha_spy_p3m": 0.15734742712382666, "alpha_c_p3m": 0.07411664834064546, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.3187823231477893, "ret_signed_p6m": 0.3187823231477893, "alpha_spy_p6m": 0.19892185293908748, "alpha_c_p6m": 0.012521495065923594, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2504, -0.3077, -0.3542, -0.357, -0.3781, -0.3016, -0.3352, -0.309, -0.3144, -0.3079, -0.3328, -0.3325, -0.3325, -0.3495, -0.334, -0.3058, -0.2778, -0.2737, -0.2812, -0.2764, -0.2667, -0.2402, -0.2113, -0.224, -0.2421, -0.2322, -0.2292, -0.2235, -0.1775, -0.1466, -0.1432, -0.1456, -0.1456, -0.1488, -0.149, -0.1564, -0.1369, -0.1555, -0.1555, -0.1304, -0.1372, -0.1327, -0.1496, -0.1429, -0.1307, -0.1096, -0.1055, -0.0974, -0.0894, -0.0654, -0.0581, -0.0488, -0.0679, -0.0477, -0.0556, -0.0574, -0.0574, -0.075, -0.0714, -0.0283, -0.0166, -0.0109, 0.0092, 0.0, -0.008, 0.0314, 0.0367, 0.0367, 0.0118, 0.006, 0.0236, 0.0144, 0.0173, 0.0096, 0.0462, 0.0489, 0.0844, 0.0614, 0.0546, 0.0358, 0.0611, 0.0667, 0.0844, 0.0718, 0.0655, 0.0725, 0.0668, 0.0385, 0.0552, 0.0264, 0.0215, 0.0712, 0.0677, 0.0689, 0.0786, 0.066, 0.0641, 0.0547, 0.0659, 0.0274, 0.0093, 0.0037, 0.0287, 0.0402, 0.054, 0.0565, 0.052, 0.0193, 0.0193, 0.0084, 0.0216, 0.0385, 0.0747, 0.0914, 0.1079, 0.1499, 0.1431, 0.145, 0.154, 0.1607, 0.1639, 0.1898, 0.1731, 0.2075, 0.2522, 0.2433, 0.2254, 0.2107, 0.2102, 0.237, 0.2777, 0.2761, 0.2941, 0.3394, 0.3023, 0.3488, 0.3282, 0.2431, 0.3415, 0.3108, 0.3496, 0.328, 0.3069, 0.3186, 0.3044, 0.2795, 0.2877, 0.3064, 0.321, 0.3355, 0.3513, 0.343, 0.3306, 0.3503, 0.3024, 0.3006, 0.2811, 0.2689, 0.3078, 0.2896, 0.2872, 0.2499, 0.2615, 0.2491, 0.2309, 0.2507, 0.2291, 0.2183, 0.2607, 0.2609, 0.2843, 0.2843, 0.2911, 0.2742, 0.2937, 0.3086, 0.2974, 0.3054, 0.337, 0.3438, 0.3737, 0.3539, 0.297, 0.2779, 0.274, 0.23, 0.2643, 0.2833, 0.3025, 0.3188]}
{"unit_id": "orig:1935684272184258683", "ticker_idx": 0, "ts": "2025-06-19T13:01:08+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA has effectively secured 'visibility' into annual data center revenue exceeding approximately $300 billion. This figure is more than 50% higher than Wall Street's consensus for NVIDIA's FY2026 data center revenue", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA: $1T pipeline implies $300B+ annual DC revenue, 50%+ above Wall St consensus, backed by multi-GW AI infrastructure projects.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bullish Outlook on NVIDIA by Mirae Asset Securities\n\nMeanwhile, NVIDIA announced during its Q1 FY2026 earnings call that it has \"visibility into tens of gigawatts of AI infrastructure projects in the not-too-distant future.\"\n\nAssuming this project pipeline is a \"minimum\" of 20GW, and applying NVIDIA's directly mentioned revenue per GW of approximately $40 billion to $50 billion, we can see that the minimum revenue opportunity from this pipeline alone is about $1 trillion. Jensen Huang's assertion that there is a \"trillion-dollar opportunity\" when advocating for accelerated computing and AI factory concepts does not sound like an exaggeration.\n\nNVIDIA did not specify the exact timing for the execution of these project pipelines, but it is highly likely that these projects will be pushed forward within the next three years. Even assuming a three-year timeframe from now, NVIDIA has effectively secured \"visibility\" into annual data center revenue exceeding approximately $300 billion. This figure is more than 50% higher than Wall Street's consensus for NVIDIA's FY2026 data center revenue, which stands at $200 billion.\n\nOf course, this may seem like a very optimistic forecast. However, considering that emerging cloud-related companies like Crusoe alone possess a project pipeline of about 20GW, it is not entirely unrealistic. Furthermore, it should be noted that Crusoe is just one of many development companies.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011204192941437152, "ret_signed_p1d": -0.011204192941437152, "alpha_spy_p1d": -0.008855553791258242, "alpha_c_p1d": -0.0023691201295533704, "bench_spy_p1d": -0.0023486391501789106, "bench_c_p1d": -0.008835072811883782, "ret_p1w": 0.06557607836357615, "ret_signed_p1w": 0.06557607836357615, "alpha_spy_p1w": 0.038395454900062465, "alpha_c_p1w": 0.009937945860100328, "bench_spy_p1w": 0.027180623463513687, "bench_c_p1w": 0.05563813250347582, "ret_p1m": 0.18511156252150518, "ret_signed_p1m": 0.18511156252150518, "alpha_spy_p1m": 0.1315576506149514, "alpha_c_p1m": 0.07954767932362516, "bench_spy_p1m": 0.053553911906553786, "bench_c_p1m": 0.10556388319788002, "ret_p3m": 0.20215760833989327, "ret_signed_p3m": 0.20215760833989327, "alpha_spy_p3m": 0.09417838349931174, "alpha_c_p3m": 0.032234884042055656, "bench_spy_p3m": 0.10797922484058153, "bench_c_p3m": 0.1699227242978376, "ret_p6m": 0.20318691915063614, "ret_signed_p6m": 0.20318691915063614, "alpha_spy_p6m": 0.05550495644924536, "alpha_c_p6m": -0.14537935462074136, "bench_spy_p6m": 0.14768196270139078, "bench_c_p6m": 0.3485662737713775, "price_path": [-0.1655, -0.1705, -0.2181, -0.2341, -0.2462, -0.2551, -0.2429, -0.241, -0.3003, -0.3518, -0.3289, -0.3381, -0.2142, -0.2606, -0.2375, -0.2391, -0.2288, -0.2818, -0.3024, -0.3024, -0.3339, -0.3203, -0.294, -0.2685, -0.237, -0.2527, -0.2507, -0.2514, -0.2329, -0.213, -0.2177, -0.2196, -0.1954, -0.1933, -0.1982, -0.1546, -0.1069, -0.0698, -0.0733, -0.0694, -0.0682, -0.0764, -0.0941, -0.087, -0.0976, -0.0976, -0.0687, -0.0734, -0.0433, -0.0712, -0.0557, -0.0293, -0.0245, -0.0378, -0.0259, -0.0197, -0.0105, -0.0182, -0.0033, -0.0241, -0.0054, -0.0093, 0.0, 0.0, -0.0112, -0.009, 0.0166, 0.0607, 0.0656, 0.0843, 0.086, 0.0538, 0.0809, 0.0953, 0.0953, 0.0877, 0.0998, 0.1196, 0.128, 0.1336, 0.1278, 0.1734, 0.178, 0.1892, 0.1851, 0.178, 0.1481, 0.1739, 0.1943, 0.1926, 0.2149, 0.2064, 0.2323, 0.2226, 0.1941, 0.2373, 0.2253, 0.2333, 0.2426, 0.2558, 0.2514, 0.259, 0.2482, 0.2512, 0.2404, 0.2511, 0.2073, 0.2057, 0.2028, 0.2235, 0.236, 0.2495, 0.2483, 0.2385, 0.1973, 0.1973, 0.1739, 0.1728, 0.18, 0.1481, 0.1569, 0.1738, 0.2189, 0.2179, 0.2224, 0.2219, 0.2022, 0.1706, 0.2115, 0.2145, 0.2622, 0.2266, 0.2165, 0.2215, 0.2249, 0.2501, 0.2826, 0.2871, 0.2985, 0.2897, 0.2754, 0.272, 0.3, 0.3238, 0.2591, 0.2945, 0.2376, 0.2362, 0.2498, 0.2595, 0.2555, 0.2453, 0.2393, 0.2522, 0.2804, 0.3163, 0.3819, 0.4232, 0.3947, 0.392, 0.4221, 0.3658, 0.3419, 0.2929, 0.2934, 0.3683, 0.3278, 0.3322, 0.2845, 0.3073, 0.2827, 0.2467, 0.2822, 0.2418, 0.2297, 0.2549, 0.2224, 0.2391, 0.2391, 0.2167, 0.2368, 0.2474, 0.2345, 0.2607, 0.254, 0.2756, 0.2716, 0.2634, 0.2438, 0.2032]}
{"unit_id": "orig:1943315877878993231", "ticker_idx": 1, "ts": "2025-07-10T14:26:25+00:00", "symbol": "Eoptolink", "kind": "equity", "geography": "China", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "raised its net profit forecasts for covered optical module manufacturers for 2025-27 by up to 42%, and their 12-month target prices by 23%-82%", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman adopts bullish stance on optical module names; raises PTs 23-82%, bullish on Innolight/Eoptolink valuation re-rating and HG Tech order overflow opportunity.", "resolved_tickers": ["300502.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Eoptolink Technology 300502.SZ China optical modules", "bench_c": "MCHI", "bench_c_reason": "country_fallback (suffix='SZ')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Modules: Rising Attach Ratio Drives Significant Increase in 800G/1.6T Optical Module Demand, Second-Tier Manufacturers Expected to Benefit\nGoldman Sachs (J. Guo, July 07, 2025)\n\nGoldman Sachs points out that AI server reference designs show a significant increase in the attach ratio of GPUs/ASICs to optical modules, leading to a substantial upward revision in demand expectations for 800G and 1.6T optical modules. Concurrently, second-tier manufacturers are likely to benefit from order overflow due to capacity constraints at leading manufacturers.\n\nRising Attach Ratio Boosts Optical Module Demand\nBased on the latest AI server reference designs, Goldman Sachs estimates the attach ratio for H100/GB200 at 1:3, rising to 1:4.5 for B300, and as high as 1:8 for some ASIC architectures (such as DSF). This is primarily driven by system bandwidth upgrades and the deployment of high-speed NICs like ConnectX-8. The firm has thus raised its 2025/26 800G shipment forecasts to 19.9/33.5 million units (previously 18.1/21.2 million units) and its 1.6T forecast to 7 million units in 2026.\n\nUpward Revision of Profit Forecasts and Convergence of Valuation Spreads Highlighted\nBuilding on the upward revision of shipment forecasts, Goldman Sachs has raised its net profit forecasts for covered optical module manufacturers for 2025-27 by up to 42%, and their 12-month target prices by 23%-82%. The report highlights two key investment themes: first, the trend of valuation convergence between Eoptolink (新易盛) and Innolight (中际旭创); second, the opportunity for second-tier manufacturers like HG Tech to gain order overflow amidst capacity tightness.\n\nSecond-Tier Manufacturer HG Tech Expected to Benefit from Order Overflow\nGoldman Sachs notes that with rapidly growing 800G capacity demand, leading manufacturers need to significantly expand production. Their reliance on manual processes also limits the pace of expansion. Second-tier manufacturers with existing capacity and product capabilities are well-positioned to take on excess orders. HG Tech has a capacity of 100,000 units/month in Thailand and has passed testing and commenced small-batch deliveries of 800G products. If mass shipments can be achieved in the second half of 2025, it could bring up to 24% upside to its 2026 profits.\n\nEoptolink and Innolight Valuation Gap Expected to Further Narrow\nDespite similar projected profits and growth rates for both companies from 2025-27, Eoptolink (新易盛) currently has a market capitalization that is only 75% to 80% of Innolight (中际旭创)'s, and its 2026 P/E still trades at approximately a 20% discount. Goldman Sachs expects that if Eoptolink's Q2 2025 earnings report shows continued matching or surpassing of profits, it will drive a further narrowing of the valuation gap. Furthermore, increased shipments of its 1.6T self-branded optical modules are expected to enhance its profit margin and valuation center.\n\nMid-to-Long Term Growth Resilience Superior to Cyclical Concerns\nInvestors often focus on industry cyclicality, but Goldman Sachs believes that even if AI chip shipment growth slows, the increasing attach ratio can still support optical module shipment growth, forming a more robust \"through-cycle\" growth logic. Coupled with Innolight (中际旭创) and Eoptolink (新易盛) currently trading at 15x and 12x their 2026 P/E respectively, which are at historical valuation lows, Goldman Sachs believes the market is still underestimating the mid-to-long term growth of optical modules.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018789928167288084, "ret_signed_m1d": 0.018789928167288084, "alpha_spy_m1d": 0.02160230660598228, "alpha_c_m1d": 0.026606631527396618, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007816703360108535, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014205181107085019, "ret_signed_p1d": -0.014205181107085019, "alpha_spy_p1d": -0.010689757236489461, "alpha_c_p1d": -0.013114450561891755, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.0010907305451932636, "ret_p1w": 0.37918072747895115, "ret_signed_p1w": 0.37918072747895115, "alpha_spy_p1w": 0.375633436411946, "alpha_c_p1w": 0.3444597818936934, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.034720945585257734, "ret_p1m": 0.3700112333585448, "ret_signed_p1m": 0.3700112333585448, "alpha_spy_p1m": 0.35185912749257087, "alpha_c_p1m": 0.3122035673597654, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.0578076659987794, "ret_p3m": 1.74911672772667, "ret_signed_p3m": 1.74911672772667, "alpha_spy_p3m": 1.676963297011512, "alpha_c_p3m": 1.5527888094398588, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.19632791828681118, "ret_p6m": 2.2384817361875275, "ret_signed_p6m": 2.2384817361875275, "alpha_spy_p6m": 2.1405808893954674, "alpha_c_p6m": 2.093222174396017, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.14525956179151045, "price_path": [-0.578, -0.5799, -0.5985, -0.5973, -0.5837, -0.5646, -0.5658, -0.5117, -0.5169, -0.4912, -0.5123, -0.5169, -0.5199, -0.5199, -0.5199, -0.5199, -0.4994, -0.4963, -0.4228, -0.4161, -0.3966, -0.3804, -0.3645, -0.3803, -0.3658, -0.3799, -0.3769, -0.3757, -0.3706, -0.3801, -0.381, -0.4004, -0.364, -0.3449, -0.3366, -0.3366, -0.3509, -0.3044, -0.2717, -0.2504, -0.2419, -0.2514, -0.2607, -0.2423, -0.248, -0.198, -0.1949, -0.1566, -0.1507, -0.1732, -0.1937, -0.198, -0.1702, -0.1221, -0.0742, -0.0453, -0.044, -0.0607, -0.0255, -0.0353, -0.0493, 0.0145, 0.0188, 0.0, -0.0142, -0.0162, 0.1806, 0.2762, 0.3792, 0.3703, 0.3398, 0.3281, 0.3093, 0.2988, 0.3, 0.3363, 0.4535, 0.4307, 0.4221, 0.3763, 0.3867, 0.3713, 0.401, 0.3801, 0.37, 0.4498, 0.54, 0.778, 0.7366, 0.7527, 0.8361, 0.93, 0.9176, 0.9381, 1.0579, 1.2256, 1.1198, 1.3174, 1.6678, 1.6772, 1.92, 1.6922, 1.7787, 1.3457, 1.625, 1.3638, 1.3699, 1.5175, 1.8553, 1.6906, 1.6328, 1.5986, 1.613, 1.5968, 1.6306, 1.578, 1.6531, 1.6655, 1.826, 1.811, 1.9183, 1.7491, 1.7491, 1.7491, 1.7491, 1.7491, 1.7491, 1.7491, 1.6392, 1.602, 1.6201, 1.3788, 1.4246, 1.4409, 1.3774, 1.4746, 1.7465, 1.727, 1.6138, 1.8032, 2.0361, 2.0862, 2.0522, 1.811, 1.5878, 1.6062, 1.6159, 1.6013, 1.681, 1.629, 1.5254, 1.4502, 1.4735, 1.4491, 1.3343, 1.3427, 1.3815, 1.4231, 1.4633, 1.2548, 1.2022, 1.2903, 1.4887, 1.5614, 1.6141, 1.6728, 1.7433, 1.8044, 1.8129, 1.8374, 2.0404, 2.1502, 2.2489, 2.1305, 2.2168, 2.2079, 2.0605, 2.3529, 2.198, 2.2642, 2.4798, 2.4423, 2.4086, 2.3671, 2.3469, 2.3408, 2.3521, 2.2385, 2.2385, 2.2385]}
{"unit_id": "orig:1938956754735505796", "ticker_idx": 0, "ts": "2025-06-28T13:44:49+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS believes that HBM has transitioned from a peripheral business to a core driver of Micron's profit and valuation", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS Buy-equivalent thesis on MU: HBM now core profit/valuation driver, supply constraints boosting margins, HBM4 ramp ahead.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron: HBM Growth Becomes Core Performance Pillar, Supply Constraints Boost Profitability Stability\n\nUBS (T. Arcuri, 2025/06/26)\n\nUBS believes that HBM has transitioned from a peripheral business to a core driver of Micron's profit and valuation. Current supply constraints in high-value products are prompting the company to reallocate resources, enhancing overall gross margin and sustainability.\n\nPerformance Exceeds Market Expectations, HBM Shows Strong Results\n\nMicron's May quarter revenue exceeded both the upper end of its guidance and market expectations. Gross margin stood at 39%, higher than UBS estimates and market consensus, with EPS at $1.91, surpassing expectations by $0.24. HBM revenue grew by nearly 50% year-over-year, estimated at $1.68 billion, accounting for roughly half of the incremental DRAM revenue, highlighting HBM's importance to the company's revenue structure.\n\nHBM Becomes Mainstream DRAM Product, Non-AI Demand Marginalized\n\nHBM accounts for approximately 6-7% of DRAM bits but commands 19-20% of capacity allocation, effectively \"squeezing\" the traditional DRAM market. Due to tight capacity, the company is prioritizing resources towards high-value-added markets, a trend expected to continue until new capacity comes online around 2026, with expansions remaining restrained.\n\nYield Improvement and Product Mix Optimization Drive Gross Margin\n\nGross margin improvement stemmed more from cost reductions than price increases, with both DRAM and NAND costs declining by high single-digit percentages. HBM3E 12-Hi yield exceeded expectations, outperforming last year's 8-Hi performance. While the product mix leaned towards the consumer market in the recent quarter, it is expected to refocus on data centers in FQ4, leading to ASP improvement.\n\nHBM Share Increases Ahead of Schedule, Potentially Shifting to HBM4 in 2026\n\nMicron anticipates reaching its 20% DRAM equivalent share target in the second half of 2024 (originally planned by year-end), suggesting FQ4 HBM revenue could reach $2.3 billion, a sequential increase of approximately 40%. HBM4 samples have begun shipping, with mass production expected to commence in 2026, further solidifying its technological leadership.\n\nFull-Year DRAM and NAND Growth Forecasts Both Upgraded\n\nThe company has raised its 2025 DRAM bit growth forecast from mid-to-high single digits to high single digits, reflecting stronger data center demand driven by AI. Full-year NAND growth is still expected to exceed 20%, maintaining a growth trend despite slightly weaker shipments in the second half due to inventory reduction.\n\nData Center Contribution Continues to Rise, Capital Expenditures Proceeding as Planned\n\nData center revenue contribution has increased to approximately 55%, with HBM being a key driver. Full-year capital expenditures remain at the $14 billion target, and FQ3 free cash flow reached $1.95 billion. The share repurchase program remains limited by the CHIPS Act, but is expected to gradually resume once restrictions ease in the next 3-5 years.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012251577573582306, "ret_signed_m1d": 0.012251577573582306, "alpha_spy_m1d": 0.01701013163863674, "alpha_c_m1d": 0.013900975108794511, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01914799578333326, "ret_signed_p1d": -0.01914799578333326, "alpha_spy_p1d": -0.018824316475108405, "alpha_c_p1d": -0.00824721902093617, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.02610236865156823, "ret_signed_p1w": -0.02610236865156823, "alpha_spy_p1w": -0.0306826948673109, "alpha_c_p1w": -0.029257914435192767, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.09074737744311612, "ret_signed_p1m": -0.09074737744311612, "alpha_spy_p1m": -0.11892560887890224, "alpha_c_p1m": -0.1410198639313055, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.2736522796191896, "ret_signed_p3m": 0.2736522796191896, "alpha_spy_p3m": 0.20563519245999307, "alpha_c_p3m": 0.12240441367681187, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.2450564955537793, "ret_signed_p6m": 1.2450564955537793, "alpha_spy_p6m": 1.1251960253450775, "alpha_c_p6m": 0.9387956674719136, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2811, -0.3968, -0.4749, -0.4453, -0.4682, -0.3682, -0.4316, -0.4357, -0.4238, -0.4236, -0.4375, -0.4418, -0.4418, -0.4585, -0.4303, -0.4083, -0.3718, -0.3527, -0.3626, -0.3762, -0.3757, -0.369, -0.3451, -0.3475, -0.3468, -0.3297, -0.3091, -0.3034, -0.2512, -0.2135, -0.2266, -0.2256, -0.2049, -0.1996, -0.2041, -0.2224, -0.2306, -0.2424, -0.2424, -0.218, -0.2196, -0.2146, -0.2336, -0.2034, -0.1704, -0.1623, -0.1376, -0.1192, -0.0998, -0.0739, -0.0586, -0.0574, -0.0621, -0.0277, -0.0236, -0.0116, -0.0116, 0.0028, -0.0095, 0.0378, 0.0325, 0.0223, 0.0123, 0.0, -0.0191, -0.0123, -0.0078, -0.0078, -0.0261, 0.0104, -0.0073, -0.0002, 0.0113, -0.0367, -0.0246, -0.0544, -0.0802, -0.071, -0.0804, -0.113, -0.108, -0.0926, -0.0964, -0.0965, -0.0907, -0.0682, -0.1136, -0.1482, -0.1248, -0.1143, -0.1166, -0.0915, -0.0345, 0.0048, 0.0375, 0.0092, 0.0175, -0.0184, 0.0034, -0.0088, -0.0481, -0.0596, -0.0443, -0.0545, -0.0539, -0.0437, -0.0092, -0.0335, -0.0335, -0.0378, -0.0358, 0.0087, 0.0669, 0.0676, 0.0983, 0.137, 0.2228, 0.2769, 0.2813, 0.2898, 0.2993, 0.3716, 0.3216, 0.3369, 0.3515, 0.3133, 0.2737, 0.2772, 0.3311, 0.3588, 0.4793, 0.4923, 0.5264, 0.5518, 0.509, 0.5971, 0.5629, 0.4757, 0.5665, 0.5201, 0.5598, 0.6458, 0.6446, 0.6803, 0.6439, 0.6128, 0.6798, 0.7798, 0.7886, 0.8033, 0.8417, 0.8204, 0.8184, 0.9072, 0.7718, 0.93, 0.9367, 0.9334, 1.0584, 0.9593, 0.9901, 0.9255, 1.0058, 0.9662, 0.8569, 0.8359, 0.6364, 0.6852, 0.8197, 0.8246, 0.8712, 0.8712, 0.9217, 0.9541, 0.9462, 0.9029, 0.8418, 0.9277, 1.0065, 1.0512, 1.143, 1.1003, 0.9596, 0.93, 0.8894, 0.8326, 1.0198, 1.1609, 1.2477, 1.2451]}
{"unit_id": "orig:1937485629618844018", "ticker_idx": 0, "ts": "2025-06-24T12:19:05+00:00", "symbol": "GOOG", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reiterate Buy. We estimate Gemini-powered subscription revenue could grow to $12 billion by 2027", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares BofA Buy reiteration on Google with $200 PT; bullish on AI monetization, Gemini subscriptions, and Workspace pricing.", "resolved_tickers": ["GOOG"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Google Price Target $200: The Evolution of Google's Monetization in the AI Era - Bullish and Bearish Cases for Google\n\nAI Transforms Google Search Monetization Landscape\nAI capabilities are changing search interactions, enhancing utility and functionality. Based on these new capabilities, a multi-tiered product monetization structure, including subscriptions, is evolving. This report examines the following search business drivers: 1) Growth in the revenue pie driven by a step-function increase in search utility, 2) A shift from traditional ad-based search to paid tiers, 3) A decrease in search clicks and a transition from CPC (cost-per-click) to CPA (cost-per-action) transactions, and 4) Increased traffic to OpenAI and competitive catalysts.\n\nIs it a New Era for Metrics to Evaluate Search Monetization?\nIncreased adoption of Google's AI Overviews and AI modes may lead to fewer clicks, but it will enable Google to better interpret search query intent and deliver more targeted ads. Additionally, Google's updated Smart Bidding framework will further support AI monetization by expanding ad eligibility across a wider range of search query types. As Google's monetization evolves, the importance of traditional search metrics like paid clicks may decrease, and monetization KPIs like ARPU (average revenue per user) may better reflect search performance.\n\nBullish Case: Favorable Position Through Massive Product Reach\nWe have noted Google management's increased focus on disclosing product reach (15 products with over 500 million users). Google is favorably positioned with its technical talent, vast data for building AI models, and multi-product deployment capabilities to drive AI usage and monetization. Beyond search, Google will secure growth through Google One subscriptions, Workspace price increases, and system-level integration across Android and Chrome OS. Reiterate Buy. We estimate Gemini-powered subscription revenue could grow to $12 billion by 2027, and Workspace stands to gain over $1 billion annually from price increases this year.\n\nBearish Case: A New Phase of Internet Distribution\nThe starting point for accessing information and transactions will shift from browser-based to mobile OS and app-based, with at least five major tech companies (Amazon, Apple, Google, Meta, OpenAI) vying for this opportunity. This shift will erode Google's long-standing competitive moats and decrease user share. In the short term, OpenAI has secured significant funding and is rapidly expanding, with a high likelihood of introducing advertising. Meanwhile, Meta could roll out Agentic AI capabilities to 1 billion assistant users, and Amazon has built agents that can shop across other sites.\n\n$GOOG", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01031362733928598, "ret_signed_m1d": -0.01031362733928598, "alpha_spy_m1d": 0.0006128743065568765, "alpha_c_m1d": 0.002420360772309671, "bench_spy_m1d": -0.010926501645842857, "bench_c_m1d": -0.012733988111595651, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022355963146170943, "ret_signed_p1d": 0.022355963146170943, "alpha_spy_p1d": 0.021795598515566672, "alpha_c_p1d": 0.0225460510852743, "bench_spy_p1d": 0.0005603646306042709, "bench_c_p1d": -0.00019008793910335609, "ret_p1w": 0.05466797783868871, "ret_signed_p1w": 0.05466797783868871, "alpha_spy_p1w": 0.03675365621109927, "alpha_c_p1w": 0.030625441488580707, "bench_spy_p1w": 0.017914321627589436, "bench_c_p1w": 0.024042536350108, "ret_p1m": 0.14170732321058943, "ret_signed_p1m": 0.14170732321058943, "alpha_spy_p1m": 0.09650139783139733, "alpha_c_p1m": 0.11110770486558019, "bench_spy_p1m": 0.0452059253791921, "bench_c_p1m": 0.03059961834500924, "ret_p3m": 0.5230006389319626, "ret_signed_p3m": 0.5230006389319626, "alpha_spy_p3m": 0.426161434264259, "alpha_c_p3m": 0.3882481336483452, "bench_spy_p3m": 0.09683920466770357, "bench_c_p3m": 0.13475250528361737, "ret_p6m": 0.779665115044222, "ret_signed_p6m": 0.779665115044222, "alpha_spy_p6m": 0.6701008452077553, "alpha_c_p6m": 0.6752853087977244, "bench_spy_p6m": 0.10956426983646672, "bench_c_p6m": 0.10437980624649756, "price_path": [-0.023, -0.0707, -0.0697, -0.054, -0.0541, -0.0912, -0.1203, -0.1114, -0.1272, -0.041, -0.0749, -0.0509, -0.0385, -0.0551, -0.0741, -0.0868, -0.0868, -0.1077, -0.0836, -0.0609, -0.0385, -0.0244, -0.0329, -0.035, -0.042, -0.0307, -0.0127, -0.0113, -0.0163, -0.0902, -0.0726, -0.0808, -0.0498, -0.042, -0.0067, -0.0151, -0.003, -0.0004, -0.0156, 0.0126, 0.024, 0.0098, 0.0098, 0.036, 0.0324, 0.0299, 0.0292, 0.0145, -0.0014, 0.0086, 0.0111, 0.0416, 0.059, 0.0731, 0.0659, 0.055, 0.0485, 0.0608, 0.0566, 0.0372, 0.0372, -0.0001, -0.0103, 0.0, 0.0224, 0.0399, 0.0628, 0.0575, 0.0547, 0.0717, 0.0764, 0.0764, 0.0585, 0.0442, 0.0591, 0.0653, 0.0809, 0.0898, 0.0916, 0.0956, 0.1011, 0.1085, 0.1396, 0.1453, 0.1417, 0.1518, 0.157, 0.1531, 0.171, 0.1771, 0.1498, 0.1324, 0.167, 0.1644, 0.174, 0.1761, 0.2048, 0.202, 0.2171, 0.2104, 0.2151, 0.2216, 0.2179, 0.2072, 0.1935, 0.196, 0.2324, 0.2469, 0.2397, 0.2413, 0.2661, 0.273, 0.273, 0.2638, 0.3777, 0.387, 0.402, 0.3972, 0.4317, 0.4294, 0.4367, 0.4403, 0.5022, 0.5002, 0.4908, 0.5056, 0.523, 0.5089, 0.5057, 0.4788, 0.4713, 0.4749, 0.4581, 0.4532, 0.4651, 0.4704, 0.4706, 0.5007, 0.4746, 0.4646, 0.4453, 0.4171, 0.4598, 0.469, 0.5019, 0.503, 0.5143, 0.5336, 0.4997, 0.5068, 0.514, 0.5544, 0.6107, 0.6017, 0.6419, 0.6821, 0.6816, 0.6953, 0.6592, 0.6991, 0.7026, 0.669, 0.7339, 0.7408, 0.7151, 0.6655, 0.6527, 0.7042, 0.7003, 0.7483, 0.7303, 0.788, 0.9003, 0.9311, 0.9111, 0.9111, 0.9101, 0.8803, 0.8857, 0.9131, 0.8998, 0.9219, 0.8775, 0.8972, 0.9166, 0.873, 0.8541, 0.8469, 0.8374, 0.7797]}
{"unit_id": "orig:1926890991224549795", "ticker_idx": 0, "ts": "2025-05-26T06:39:46+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The construction of AI Factories has just begun and is expected to continue for several years. Not only CSPs but also sovereign AI initiatives are fully investing in AI Factory CAPEX. Physical AI/Robotics is anticipated to emerge within 3–5 years, driving GPU demand. Surprisingly Strong Demand for GB300", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan Computex takeaways: NVDA AI factory cycle multi-year + strong GB300 demand; Vertiv HVDC TAM expansion; ASPEED BMC content headwind; MediaTek ASIC roadmap on track.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250526_Asian Tech: Key Takeaways from 2025 Computex - JPMorgan\n\n(1) Continued Strong AI Demand with Visibility for Growth in Coming Years\n • NVDA: The construction of AI Factories has just begun and is expected to continue for several years.\n • Not only CSPs but also sovereign AI initiatives are fully investing in AI Factory CAPEX.\n • Physical AI/Robotics is anticipated to emerge within 3–5 years, driving GPU demand.\n\n(2) Short-Term Bottlenecks in GB200 Supply; Significant Production Increase Expected from 2H25\n • Multiple server ODM channel checks indicate short-term production bottlenecks in GB200 NVL72 Rack.\n • This is due to 1) low test pass rates and 2) low backplane assembly yields.\n • However, many companies expect rack shipments to increase significantly from 2Q25. The forecast for 25K NVL72 Rack deliveries in 2025 remains unchanged. \n\n(3) Surprisingly Strong Demand for GB300\n • Recent ODM channel checks confirm strong demand for GB300 NVL72.\n • This is because racks with more HBM are suitable for large-scale inference tasks.\n • This trend is more optimistic than the market’s base scenario, as many investors view GB300 as a transitional product between GB200 and Rubin. \n\n(4) GB300 Compute Tray Equipped with One DPU; Increased Adoption of BMC in Non-Compute Areas\n • GB200 Bianca includes two BF3 DPUs, while GB300 Cordelia includes one BF3 DPU.\n • Despite the GB300 GPU board design reverting to Bianca, it still includes one DPU (additional possible in MGX).\n • This is negative for content growth in BMC (Board Management Controller) suppliers like ASPEED.\n • However, as overall rack structures become more complex, BMC adoption is structurally increasing in non-compute areas such as switches, CDUs, and power racks.\n\n(5) Ongoing Competition in Liquid Cooling\n • All value chain players showcased liquid cooling solutions.\n • This technology encompasses multiple layers, including 1) chip, 2) rack/system, and 3) infrastructure levels, with different players and competitive dynamics at each layer.\n • Most CSPs designate suppliers (consign), limiting ODMs’ influence and opportunities to adopt their own components.\n • In chip-level cooling, AVC is positioned as a top-tier cold plate module supplier for GB200/GB300 in 2025 and Rubin in 2H26, expected to maintain solid market share within CSPs. \n\n(6) GB300 Reverts to Bianca Design\n • Several liquid cooling companies exhibited NVQD (NVIDIA Qualified Design) designs based on Cordelia GPU boards but reported reverting to Bianca boards.\n • Such design changes are negative for cooling component suppliers, as Bianca designs reduce TAM compared to NVQD.\n • Although GB300’s TDP is approximately 15% higher than GB200, companies believe existing QD designs can adequately handle the increased heat.\n\n(7) Delay in ODMs’ Adoption of In-House Liquid Cooling Components\n • Several server ODMs showcased their own liquid cooling components, including 1) cold plates, 2) manifolds, and 3) side carts.\n • Some companies also presented next-generation products like two-phase DLC (Direct Liquid Cooling), liquid metal heatsinks, and immersion cooling.\n • By increasing the adoption of in-house components, companies aim to improve margins and shorten product development cycles.\n • However, due to GB300 delays, CSPs prioritize the release schedules of existing GB200 products, leading to a temporary delay in the adoption of in-house components.\n\n(8) Data Center Power: HVDC Emerging as a Next-Generation Game Changer\n • Rubin is expected to consume approximately 200kW, while the NVL576 Kyber structure is anticipated to consume around 600kW.\n • When rack power consumption exceeds 200kW, issues arise, including rack space constraints, copper busbar overload, and inefficient power conversion.\n • To address this, proposals include increasing voltage from the existing 230V to 800V in Rubin Ultra and Kyber servers.\n • Implementing 800V HVDC (High Voltage Direct Current) power racks eliminates AC/DC conversion devices within IT racks and provides space for high-density PSUs.\n• The adoption of HVDC is expected to significantly expand the TAM for power-related infrastructure, benefiting companies like Vertiv, Delta, Flextronics, Lite-On Tech, Infineon, and MPS. \n\n(9) Separate Power Rack Structure Integrating BBU and Supercapacitors\n • Rubin is anticipated to initially adopt independent power racks.\n • While NVIDIA recommends centralized power racks supporting multiple computing racks within data centers, CSPs prefer placing individual power racks next to each computing rack to enhance power efficiency.\n • Given the ample space within power racks, the adoption of BBUs (Battery Backup Units) and supercapacitors is likely to increase, potentially doubling or tripling component content compared to existing PSUs.\n\n(10) RTX Pro 6000 Blackwell Server: Targeting Enterprise Market, Potential H20 Replacement?\n • The RTX Pro 6000 Blackwell features up to eight PCIe-based GPUs (each with 600W TDP and 96GB GDDR7 memory), air cooling, CX8 smart NIC, BF-3 DPU, and a 4U height.\n • Using GDDR7 memory reduces the burden of adopting CoWoS-based HBM, which is positive.\n • However, the RTX Pro 6000’s compute performance is over 3–4 times higher than H20 (FP32: 117TFLOPs vs. 44TFLOPs), presenting practical challenges in replacing H20.\n\n(11) NVLink Fusion\n • NVLink, previously a proprietary interconnect between NVIDIA GPUs, is now open to integration with third-party chips (ASICs, Qualcomm/Fujitsu CPUs).\n • NVIDIA aims to establish its technology as the default interconnect in custom ASIC environments.\n • CSPs express concerns about increased dependency on NVIDIA (lock-in). Monitoring actual adoption over the next 2–3 quarters is necessary.\n • Currently, initial interest is observed among some CSPs and second-tier ASIC customers.\n\n(12) CPO: NPO and CPC Preferred for Now\n • The goal is to introduce Co-Packaged Optics (CPO) in Rubin Ultra and NVL576 Kyber platforms between 2026 and 2027.\n • However, thermal domain separation issues affect signal transmission, posing technical barriers.\n • Consequently, the market currently favors Near-Packaged Optics (NPO) and Co-Packaged Copper (CPC) as transitional solutions.\n\n(13) AI PC Adoption Continues, but Innovative AI Applications Still Lacking\n • Several PC brands are attempting to differentiate their AI PC products through proprietary AI solutions.\n • However, there is still a lack of innovative AI applications compelling enough to drive replacement demand, and AI PC penetration remains lower than expected.\n • In the medium to long term, reduced AI inference costs and improved local AI performance are expected to enrich locally executed AI apps, promoting AI PC adoption in both enterprise and consumer markets.\n\n(14) MediaTek’s ASIC and Automotive Roadmap Progressing Smoothly\n • MediaTek has been producing large networking chips measuring 91x91mm for over a year and continues to strengthen its AI ASIC technology roadmap.\n • Developments include A16 compute die, 448 SerDes, 3D packaging, CPO, and custom HBM.\n • Mass production of the first data center ASIC (GOOGL TPU v7e) is expected in 2H26, with potential orders for the 2nm process TPU v8e project in 2027–28.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03206635423474524, "ret_signed_p1d": 0.03206635423474524, "alpha_spy_p1d": 0.011275686513277927, "alpha_c_p1d": 0.0003530997375458256, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.03171325449719942, "ret_p1w": 0.04638595688495273, "ret_signed_p1w": 0.04638595688495273, "alpha_spy_p1w": 0.022901539092194723, "alpha_c_p1w": 0.0217060176948789, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02467993919007383, "ret_p1m": 0.12659200664388637, "ret_signed_p1m": 0.12659200664388637, "alpha_spy_p1m": 0.07571425390240472, "alpha_c_p1m": -0.017023192846540747, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.14361519949042711, "ret_p3m": 0.33286728943802, "ret_signed_p3m": 0.33286728943802, "alpha_spy_p3m": 0.23216299569485654, "alpha_c_p3m": 0.12220362179599564, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.21066366764202438, "ret_p6m": 0.3815433597521749, "ret_signed_p6m": 0.3815433597521749, "alpha_spy_p6m": 0.23518624390365206, "alpha_c_p6m": -0.022137609503847244, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.40368096925602215, "price_path": [-0.0002, -0.0849, -0.0486, -0.1313, -0.1166, -0.1066, -0.1579, -0.1417, -0.1852, -0.1717, -0.1184, -0.1197, -0.0733, -0.0896, -0.1208, -0.1049, -0.0972, -0.1035, -0.0753, -0.0807, -0.1335, -0.1513, -0.1647, -0.1745, -0.161, -0.159, -0.2246, -0.2817, -0.2563, -0.2665, -0.1292, -0.1807, -0.1551, -0.1568, -0.1454, -0.2041, -0.227, -0.227, -0.2619, -0.2468, -0.2177, -0.1894, -0.1545, -0.1718, -0.1696, -0.1704, -0.1499, -0.1279, -0.1331, -0.1352, -0.1084, -0.106, -0.1115, -0.0631, -0.0104, 0.0308, 0.027, 0.0313, 0.0326, 0.0235, 0.0039, 0.0117, 0.0, 0.0, 0.0321, 0.0268, 0.0602, 0.0292, 0.0464, 0.0756, 0.081, 0.0663, 0.0794, 0.0864, 0.0965, 0.088, 0.1045, 0.0814, 0.1021, 0.0978, 0.1082, 0.1082, 0.0957, 0.0982, 0.1266, 0.1754, 0.1808, 0.2016, 0.2035, 0.1677, 0.1978, 0.2137, 0.2137, 0.2054, 0.2188, 0.2407, 0.25, 0.2562, 0.2498, 0.3003, 0.3054, 0.3178, 0.3133, 0.3054, 0.2723, 0.3009, 0.3234, 0.3216, 0.3463, 0.3369, 0.3655, 0.3549, 0.3233, 0.3711, 0.3579, 0.3667, 0.377, 0.3917, 0.3868, 0.3952, 0.3832, 0.3865, 0.3745, 0.3864, 0.3379, 0.3361, 0.3329, 0.3558, 0.3697, 0.3846, 0.3833, 0.3724, 0.3268, 0.3268, 0.3009, 0.2997, 0.3076, 0.2722, 0.2821, 0.3007, 0.3508, 0.3496, 0.3546, 0.354, 0.3322, 0.2972, 0.3425, 0.3458, 0.3987, 0.3592, 0.3481, 0.3536, 0.3574, 0.3853, 0.4213, 0.4263, 0.4389, 0.4292, 0.4134, 0.4096, 0.4406, 0.4669, 0.3953, 0.4346, 0.3714, 0.3699, 0.385, 0.3957, 0.3913, 0.38, 0.3733, 0.3876, 0.4189, 0.4587, 0.5314, 0.5772, 0.5456, 0.5425, 0.5759, 0.5136, 0.487, 0.4327, 0.4333, 0.5163, 0.4714, 0.4763, 0.4234, 0.4487, 0.4215, 0.3815]}
{"unit_id": "orig:1926890991224549795", "ticker_idx": 1, "ts": "2025-05-26T06:39:46+00:00", "symbol": "Vertiv", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The adoption of HVDC is expected to significantly expand the TAM for power-related infrastructure, benefiting companies like Vertiv", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan Computex takeaways: NVDA AI factory cycle multi-year + strong GB300 demand; Vertiv HVDC TAM expansion; ASPEED BMC content headwind; MediaTek ASIC roadmap on track.", "resolved_tickers": ["VRT"], "resolution_method": "exchange_ticker", "resolution_reason": "Vertiv Holdings listed on NYSE as VRT", "bench_c": "XLI", "bench_c_reason": "industry='Electrical Equipment & Parts'", "bench_c_industry": "Electrical Equipment & Parts", "unit_kind": "original", "lang": "en", "tweet_text": "250526_Asian Tech: Key Takeaways from 2025 Computex - JPMorgan\n\n(1) Continued Strong AI Demand with Visibility for Growth in Coming Years\n • NVDA: The construction of AI Factories has just begun and is expected to continue for several years.\n • Not only CSPs but also sovereign AI initiatives are fully investing in AI Factory CAPEX.\n • Physical AI/Robotics is anticipated to emerge within 3–5 years, driving GPU demand.\n\n(2) Short-Term Bottlenecks in GB200 Supply; Significant Production Increase Expected from 2H25\n • Multiple server ODM channel checks indicate short-term production bottlenecks in GB200 NVL72 Rack.\n • This is due to 1) low test pass rates and 2) low backplane assembly yields.\n • However, many companies expect rack shipments to increase significantly from 2Q25. The forecast for 25K NVL72 Rack deliveries in 2025 remains unchanged. \n\n(3) Surprisingly Strong Demand for GB300\n • Recent ODM channel checks confirm strong demand for GB300 NVL72.\n • This is because racks with more HBM are suitable for large-scale inference tasks.\n • This trend is more optimistic than the market’s base scenario, as many investors view GB300 as a transitional product between GB200 and Rubin. \n\n(4) GB300 Compute Tray Equipped with One DPU; Increased Adoption of BMC in Non-Compute Areas\n • GB200 Bianca includes two BF3 DPUs, while GB300 Cordelia includes one BF3 DPU.\n • Despite the GB300 GPU board design reverting to Bianca, it still includes one DPU (additional possible in MGX).\n • This is negative for content growth in BMC (Board Management Controller) suppliers like ASPEED.\n • However, as overall rack structures become more complex, BMC adoption is structurally increasing in non-compute areas such as switches, CDUs, and power racks.\n\n(5) Ongoing Competition in Liquid Cooling\n • All value chain players showcased liquid cooling solutions.\n • This technology encompasses multiple layers, including 1) chip, 2) rack/system, and 3) infrastructure levels, with different players and competitive dynamics at each layer.\n • Most CSPs designate suppliers (consign), limiting ODMs’ influence and opportunities to adopt their own components.\n • In chip-level cooling, AVC is positioned as a top-tier cold plate module supplier for GB200/GB300 in 2025 and Rubin in 2H26, expected to maintain solid market share within CSPs. \n\n(6) GB300 Reverts to Bianca Design\n • Several liquid cooling companies exhibited NVQD (NVIDIA Qualified Design) designs based on Cordelia GPU boards but reported reverting to Bianca boards.\n • Such design changes are negative for cooling component suppliers, as Bianca designs reduce TAM compared to NVQD.\n • Although GB300’s TDP is approximately 15% higher than GB200, companies believe existing QD designs can adequately handle the increased heat.\n\n(7) Delay in ODMs’ Adoption of In-House Liquid Cooling Components\n • Several server ODMs showcased their own liquid cooling components, including 1) cold plates, 2) manifolds, and 3) side carts.\n • Some companies also presented next-generation products like two-phase DLC (Direct Liquid Cooling), liquid metal heatsinks, and immersion cooling.\n • By increasing the adoption of in-house components, companies aim to improve margins and shorten product development cycles.\n • However, due to GB300 delays, CSPs prioritize the release schedules of existing GB200 products, leading to a temporary delay in the adoption of in-house components.\n\n(8) Data Center Power: HVDC Emerging as a Next-Generation Game Changer\n • Rubin is expected to consume approximately 200kW, while the NVL576 Kyber structure is anticipated to consume around 600kW.\n • When rack power consumption exceeds 200kW, issues arise, including rack space constraints, copper busbar overload, and inefficient power conversion.\n • To address this, proposals include increasing voltage from the existing 230V to 800V in Rubin Ultra and Kyber servers.\n • Implementing 800V HVDC (High Voltage Direct Current) power racks eliminates AC/DC conversion devices within IT racks and provides space for high-density PSUs.\n• The adoption of HVDC is expected to significantly expand the TAM for power-related infrastructure, benefiting companies like Vertiv, Delta, Flextronics, Lite-On Tech, Infineon, and MPS. \n\n(9) Separate Power Rack Structure Integrating BBU and Supercapacitors\n • Rubin is anticipated to initially adopt independent power racks.\n • While NVIDIA recommends centralized power racks supporting multiple computing racks within data centers, CSPs prefer placing individual power racks next to each computing rack to enhance power efficiency.\n • Given the ample space within power racks, the adoption of BBUs (Battery Backup Units) and supercapacitors is likely to increase, potentially doubling or tripling component content compared to existing PSUs.\n\n(10) RTX Pro 6000 Blackwell Server: Targeting Enterprise Market, Potential H20 Replacement?\n • The RTX Pro 6000 Blackwell features up to eight PCIe-based GPUs (each with 600W TDP and 96GB GDDR7 memory), air cooling, CX8 smart NIC, BF-3 DPU, and a 4U height.\n • Using GDDR7 memory reduces the burden of adopting CoWoS-based HBM, which is positive.\n • However, the RTX Pro 6000’s compute performance is over 3–4 times higher than H20 (FP32: 117TFLOPs vs. 44TFLOPs), presenting practical challenges in replacing H20.\n\n(11) NVLink Fusion\n • NVLink, previously a proprietary interconnect between NVIDIA GPUs, is now open to integration with third-party chips (ASICs, Qualcomm/Fujitsu CPUs).\n • NVIDIA aims to establish its technology as the default interconnect in custom ASIC environments.\n • CSPs express concerns about increased dependency on NVIDIA (lock-in). Monitoring actual adoption over the next 2–3 quarters is necessary.\n • Currently, initial interest is observed among some CSPs and second-tier ASIC customers.\n\n(12) CPO: NPO and CPC Preferred for Now\n • The goal is to introduce Co-Packaged Optics (CPO) in Rubin Ultra and NVL576 Kyber platforms between 2026 and 2027.\n • However, thermal domain separation issues affect signal transmission, posing technical barriers.\n • Consequently, the market currently favors Near-Packaged Optics (NPO) and Co-Packaged Copper (CPC) as transitional solutions.\n\n(13) AI PC Adoption Continues, but Innovative AI Applications Still Lacking\n • Several PC brands are attempting to differentiate their AI PC products through proprietary AI solutions.\n • However, there is still a lack of innovative AI applications compelling enough to drive replacement demand, and AI PC penetration remains lower than expected.\n • In the medium to long term, reduced AI inference costs and improved local AI performance are expected to enrich locally executed AI apps, promoting AI PC adoption in both enterprise and consumer markets.\n\n(14) MediaTek’s ASIC and Automotive Roadmap Progressing Smoothly\n • MediaTek has been producing large networking chips measuring 91x91mm for over a year and continues to strengthen its AI ASIC technology roadmap.\n • Developments include A16 compute die, 448 SerDes, 3D packaging, CPO, and custom HBM.\n • Mass production of the first data center ASIC (GOOGL TPU v7e) is expected in 2H26, with potential orders for the 2nm process TPU v8e project in 2027–28.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.054254002333311346, "ret_signed_p1d": 0.054254002333311346, "alpha_spy_p1d": 0.03346333461184403, "alpha_c_p1d": 0.036644006707012844, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.017609995626298502, "ret_p1w": 0.04887660332641741, "ret_signed_p1w": 0.04887660332641741, "alpha_spy_p1w": 0.025392185533659406, "alpha_c_p1w": 0.036024092379261186, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.012852510947156226, "ret_p1m": 0.17497510633217894, "ret_signed_p1m": 0.17497510633217894, "alpha_spy_p1m": 0.12409735359069729, "alpha_c_p1m": 0.1418038257929981, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.033171280539180836, "ret_p3m": 0.2158955650578358, "ret_signed_p3m": 0.2158955650578358, "alpha_spy_p3m": 0.11519127131467233, "alpha_c_p3m": 0.14269651596180188, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.07319904909603392, "ret_p6m": 0.5840500081236795, "ret_signed_p6m": 0.5840500081236795, "alpha_spy_p6m": 0.4376928922751566, "alpha_c_p6m": 0.5123530394183204, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.07169696870535902, "price_path": [-0.0535, -0.113, -0.0865, -0.1802, -0.1708, -0.1564, -0.2126, -0.1838, -0.2511, -0.2022, -0.1805, -0.2025, -0.1606, -0.145, -0.1871, -0.1437, -0.1446, -0.1489, -0.1003, -0.1206, -0.2162, -0.2671, -0.287, -0.3067, -0.2772, -0.2596, -0.3521, -0.4295, -0.3872, -0.396, -0.3007, -0.3496, -0.3316, -0.3128, -0.3018, -0.312, -0.297, -0.297, -0.3512, -0.3104, -0.251, -0.1945, -0.1651, -0.1727, -0.1724, -0.1801, -0.1113, -0.0878, -0.0894, -0.1024, -0.0828, -0.0807, -0.0968, -0.0297, 0.0138, 0.0513, 0.0119, 0.0182, 0.0217, 0.0216, -0.0051, 0.0006, 0.0, 0.0, 0.0543, 0.0488, 0.0416, 0.0364, 0.0489, 0.0787, 0.0836, 0.0776, 0.1077, 0.0755, 0.0416, 0.0619, 0.0995, 0.0656, 0.1186, 0.1201, 0.1439, 0.1439, 0.1387, 0.1195, 0.175, 0.1684, 0.1892, 0.2215, 0.2335, 0.1771, 0.1943, 0.228, 0.228, 0.2128, 0.2093, 0.2331, 0.1596, 0.1844, 0.198, 0.2235, 0.2046, 0.2595, 0.2397, 0.2123, 0.2035, 0.2506, 0.2571, 0.3205, 0.3693, 0.3707, 0.3849, 0.3986, 0.3601, 0.3467, 0.3329, 0.3424, 0.3389, 0.3441, 0.3432, 0.3805, 0.3198, 0.273, 0.2782, 0.3034, 0.2396, 0.2251, 0.2159, 0.21, 0.2009, 0.2289, 0.2421, 0.2894, 0.2252, 0.2252, 0.1912, 0.2064, 0.2074, 0.1911, 0.1702, 0.2063, 0.3135, 0.3036, 0.2952, 0.3285, 0.313, 0.3147, 0.364, 0.3798, 0.4601, 0.3703, 0.3607, 0.343, 0.3319, 0.377, 0.4495, 0.5526, 0.5531, 0.5393, 0.5643, 0.5265, 0.6083, 0.6227, 0.6239, 0.7199, 0.6829, 0.7344, 0.7086, 0.6719, 0.6885, 0.6796, 0.6487, 0.7603, 0.7877, 0.8535, 0.8311, 0.9147, 0.8617, 0.8531, 0.8391, 0.7374, 0.8324, 0.7585, 0.7276, 0.8049, 0.7204, 0.6658, 0.5723, 0.6428, 0.6012, 0.5841]}
{"unit_id": "orig:1926890991224549795", "ticker_idx": 2, "ts": "2025-05-26T06:39:46+00:00", "symbol": "ASPEED", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This is negative for content growth in BMC (Board Management Controller) suppliers like ASPEED", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan Computex takeaways: NVDA AI factory cycle multi-year + strong GB300 demand; Vertiv HVDC TAM expansion; ASPEED BMC content headwind; MediaTek ASIC roadmap on track.", "resolved_tickers": ["6674.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "ASPEED = ASPEED Technology, TWSE 6674.TW BMC controller", "bench_c": "EWT", "bench_c_reason": "country_fallback (suffix='TW')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "250526_Asian Tech: Key Takeaways from 2025 Computex - JPMorgan\n\n(1) Continued Strong AI Demand with Visibility for Growth in Coming Years\n • NVDA: The construction of AI Factories has just begun and is expected to continue for several years.\n • Not only CSPs but also sovereign AI initiatives are fully investing in AI Factory CAPEX.\n • Physical AI/Robotics is anticipated to emerge within 3–5 years, driving GPU demand.\n\n(2) Short-Term Bottlenecks in GB200 Supply; Significant Production Increase Expected from 2H25\n • Multiple server ODM channel checks indicate short-term production bottlenecks in GB200 NVL72 Rack.\n • This is due to 1) low test pass rates and 2) low backplane assembly yields.\n • However, many companies expect rack shipments to increase significantly from 2Q25. The forecast for 25K NVL72 Rack deliveries in 2025 remains unchanged. \n\n(3) Surprisingly Strong Demand for GB300\n • Recent ODM channel checks confirm strong demand for GB300 NVL72.\n • This is because racks with more HBM are suitable for large-scale inference tasks.\n • This trend is more optimistic than the market’s base scenario, as many investors view GB300 as a transitional product between GB200 and Rubin. \n\n(4) GB300 Compute Tray Equipped with One DPU; Increased Adoption of BMC in Non-Compute Areas\n • GB200 Bianca includes two BF3 DPUs, while GB300 Cordelia includes one BF3 DPU.\n • Despite the GB300 GPU board design reverting to Bianca, it still includes one DPU (additional possible in MGX).\n • This is negative for content growth in BMC (Board Management Controller) suppliers like ASPEED.\n • However, as overall rack structures become more complex, BMC adoption is structurally increasing in non-compute areas such as switches, CDUs, and power racks.\n\n(5) Ongoing Competition in Liquid Cooling\n • All value chain players showcased liquid cooling solutions.\n • This technology encompasses multiple layers, including 1) chip, 2) rack/system, and 3) infrastructure levels, with different players and competitive dynamics at each layer.\n • Most CSPs designate suppliers (consign), limiting ODMs’ influence and opportunities to adopt their own components.\n • In chip-level cooling, AVC is positioned as a top-tier cold plate module supplier for GB200/GB300 in 2025 and Rubin in 2H26, expected to maintain solid market share within CSPs. \n\n(6) GB300 Reverts to Bianca Design\n • Several liquid cooling companies exhibited NVQD (NVIDIA Qualified Design) designs based on Cordelia GPU boards but reported reverting to Bianca boards.\n • Such design changes are negative for cooling component suppliers, as Bianca designs reduce TAM compared to NVQD.\n • Although GB300’s TDP is approximately 15% higher than GB200, companies believe existing QD designs can adequately handle the increased heat.\n\n(7) Delay in ODMs’ Adoption of In-House Liquid Cooling Components\n • Several server ODMs showcased their own liquid cooling components, including 1) cold plates, 2) manifolds, and 3) side carts.\n • Some companies also presented next-generation products like two-phase DLC (Direct Liquid Cooling), liquid metal heatsinks, and immersion cooling.\n • By increasing the adoption of in-house components, companies aim to improve margins and shorten product development cycles.\n • However, due to GB300 delays, CSPs prioritize the release schedules of existing GB200 products, leading to a temporary delay in the adoption of in-house components.\n\n(8) Data Center Power: HVDC Emerging as a Next-Generation Game Changer\n • Rubin is expected to consume approximately 200kW, while the NVL576 Kyber structure is anticipated to consume around 600kW.\n • When rack power consumption exceeds 200kW, issues arise, including rack space constraints, copper busbar overload, and inefficient power conversion.\n • To address this, proposals include increasing voltage from the existing 230V to 800V in Rubin Ultra and Kyber servers.\n • Implementing 800V HVDC (High Voltage Direct Current) power racks eliminates AC/DC conversion devices within IT racks and provides space for high-density PSUs.\n• The adoption of HVDC is expected to significantly expand the TAM for power-related infrastructure, benefiting companies like Vertiv, Delta, Flextronics, Lite-On Tech, Infineon, and MPS. \n\n(9) Separate Power Rack Structure Integrating BBU and Supercapacitors\n • Rubin is anticipated to initially adopt independent power racks.\n • While NVIDIA recommends centralized power racks supporting multiple computing racks within data centers, CSPs prefer placing individual power racks next to each computing rack to enhance power efficiency.\n • Given the ample space within power racks, the adoption of BBUs (Battery Backup Units) and supercapacitors is likely to increase, potentially doubling or tripling component content compared to existing PSUs.\n\n(10) RTX Pro 6000 Blackwell Server: Targeting Enterprise Market, Potential H20 Replacement?\n • The RTX Pro 6000 Blackwell features up to eight PCIe-based GPUs (each with 600W TDP and 96GB GDDR7 memory), air cooling, CX8 smart NIC, BF-3 DPU, and a 4U height.\n • Using GDDR7 memory reduces the burden of adopting CoWoS-based HBM, which is positive.\n • However, the RTX Pro 6000’s compute performance is over 3–4 times higher than H20 (FP32: 117TFLOPs vs. 44TFLOPs), presenting practical challenges in replacing H20.\n\n(11) NVLink Fusion\n • NVLink, previously a proprietary interconnect between NVIDIA GPUs, is now open to integration with third-party chips (ASICs, Qualcomm/Fujitsu CPUs).\n • NVIDIA aims to establish its technology as the default interconnect in custom ASIC environments.\n • CSPs express concerns about increased dependency on NVIDIA (lock-in). Monitoring actual adoption over the next 2–3 quarters is necessary.\n • Currently, initial interest is observed among some CSPs and second-tier ASIC customers.\n\n(12) CPO: NPO and CPC Preferred for Now\n • The goal is to introduce Co-Packaged Optics (CPO) in Rubin Ultra and NVL576 Kyber platforms between 2026 and 2027.\n • However, thermal domain separation issues affect signal transmission, posing technical barriers.\n • Consequently, the market currently favors Near-Packaged Optics (NPO) and Co-Packaged Copper (CPC) as transitional solutions.\n\n(13) AI PC Adoption Continues, but Innovative AI Applications Still Lacking\n • Several PC brands are attempting to differentiate their AI PC products through proprietary AI solutions.\n • However, there is still a lack of innovative AI applications compelling enough to drive replacement demand, and AI PC penetration remains lower than expected.\n • In the medium to long term, reduced AI inference costs and improved local AI performance are expected to enrich locally executed AI apps, promoting AI PC adoption in both enterprise and consumer markets.\n\n(14) MediaTek’s ASIC and Automotive Roadmap Progressing Smoothly\n • MediaTek has been producing large networking chips measuring 91x91mm for over a year and continues to strengthen its AI ASIC technology roadmap.\n • Developments include A16 compute die, 448 SerDes, 3D packaging, CPO, and custom HBM.\n • Mass production of the first data center ASIC (GOOGL TPU v7e) is expected in 2H26, with potential orders for the 2nm process TPU v8e project in 2027–28.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030000019073486373, "ret_signed_m1d": -0.030000019073486373, "alpha_spy_m1d": -0.030000019073486373, "alpha_c_m1d": -0.030000019073486373, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012499999999999956, "ret_signed_p1d": 0.012499999999999956, "alpha_spy_p1d": 0.03329066772146727, "alpha_c_p1d": 0.013621168167919251, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.0011211681679192953, "ret_p1w": 0.18500003814697275, "ret_signed_p1w": -0.18500003814697275, "alpha_spy_p1w": -0.16151562035421474, "alpha_c_p1w": -0.19770556700674646, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": -0.012705528859773718, "ret_p1m": 0.35249996185302734, "ret_signed_p1m": -0.35249996185302734, "alpha_spy_p1m": -0.3016222091115457, "alpha_c_p1m": -0.2908407820659322, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.06165917978709512, "ret_p3m": 0.42250003814697257, "ret_signed_p3m": -0.42250003814697257, "alpha_spy_p3m": -0.3217957444038091, "alpha_c_p3m": -0.3281427666792003, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.09435727146777229, "ret_p6m": 0.04250001907348633, "ret_signed_p6m": -0.04250001907348633, "alpha_spy_p6m": 0.10385709677503652, "alpha_c_p6m": 0.12846407372485524, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.17096409279834157, "price_path": [0.18, 0.2, 0.2, 0.185, 0.2075, 0.205, 0.1975, 0.175, 0.185, 0.18, 0.1725, 0.17, 0.1725, 0.19, 0.1925, 0.185, 0.1825, 0.1725, 0.175, 0.1775, 0.175, 0.1525, 0.1225, 0.045, 0.0475, 0.05, 0.05, 0.05, -0.055, -0.1475, -0.2325, -0.16, -0.175, -0.185, -0.1525, -0.1575, -0.1525, -0.125, -0.1325, -0.14, -0.155, -0.145, -0.13, -0.045, 0.0425, 0.06, 0.06, 0.16, 0.1675, 0.1725, 0.14, 0.1425, 0.1975, 0.1825, 0.125, 0.085, 0.1075, 0.0975, 0.0775, 0.0625, 0.0475, 0.04, 0.03, 0.0, -0.0125, -0.02, 0.0775, 0.0775, 0.185, 0.205, 0.195, 0.3125, 0.33, 0.32, 0.3175, 0.275, 0.4025, 0.375, 0.3675, 0.3425, 0.325, 0.2725, 0.35, 0.3125, 0.3525, 0.3375, 0.3375, 0.345, 0.3125, 0.3175, 0.325, 0.325, 0.2925, 0.2825, 0.2675, 0.2525, 0.255, 0.305, 0.3225, 0.365, 0.3925, 0.4125, 0.425, 0.3425, 0.325, 0.3575, 0.305, 0.33, 0.3325, 0.35, 0.355, 0.3425, 0.3425, 0.39, 0.385, 0.3725, 0.3375, 0.365, 0.4475, 0.4525, 0.5125, 0.475, 0.42, 0.3825, 0.425, 0.385, 0.4225, 0.365, 0.3475, 0.2775, 0.26, 0.21, 0.1775, 0.14, 0.09, 0.095, 0.0975, 0.0775, 0.0625, 0.06, 0.0775, 0.185, 0.07, 0.0175, 0.025, 0.025, 0.025, 0.03, 0.045, 0.055, 0.075, 0.0775, 0.0625, 0.0625, 0.0925, 0.07, 0.0625, 0.055, 0.055, 0.0475, 0.05, 0.0575, 0.0575, 0.065, 0.035, 0.045, 0.05, 0.04, 0.01, 0.0275, 0.04, 0.02, 0.02, 0.0225, 0.0175, 0.01, 0.0625, 0.0375, 0.14, 0.0475, 0.055, 0.035, 0.035, 0.0275, 0.0475, 0.0625, 0.06, 0.055, 0.05, 0.0425]}
{"unit_id": "orig:1926890991224549795", "ticker_idx": 3, "ts": "2025-05-26T06:39:46+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek has been producing large networking chips measuring 91x91mm for over a year and continues to strengthen its AI ASIC technology roadmap. Mass production of the first data center ASIC (GOOGL TPU v7e) is expected in 2H26", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan Computex takeaways: NVDA AI factory cycle multi-year + strong GB300 demand; Vertiv HVDC TAM expansion; ASPEED BMC content headwind; MediaTek ASIC roadmap on track.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250526_Asian Tech: Key Takeaways from 2025 Computex - JPMorgan\n\n(1) Continued Strong AI Demand with Visibility for Growth in Coming Years\n • NVDA: The construction of AI Factories has just begun and is expected to continue for several years.\n • Not only CSPs but also sovereign AI initiatives are fully investing in AI Factory CAPEX.\n • Physical AI/Robotics is anticipated to emerge within 3–5 years, driving GPU demand.\n\n(2) Short-Term Bottlenecks in GB200 Supply; Significant Production Increase Expected from 2H25\n • Multiple server ODM channel checks indicate short-term production bottlenecks in GB200 NVL72 Rack.\n • This is due to 1) low test pass rates and 2) low backplane assembly yields.\n • However, many companies expect rack shipments to increase significantly from 2Q25. The forecast for 25K NVL72 Rack deliveries in 2025 remains unchanged. \n\n(3) Surprisingly Strong Demand for GB300\n • Recent ODM channel checks confirm strong demand for GB300 NVL72.\n • This is because racks with more HBM are suitable for large-scale inference tasks.\n • This trend is more optimistic than the market’s base scenario, as many investors view GB300 as a transitional product between GB200 and Rubin. \n\n(4) GB300 Compute Tray Equipped with One DPU; Increased Adoption of BMC in Non-Compute Areas\n • GB200 Bianca includes two BF3 DPUs, while GB300 Cordelia includes one BF3 DPU.\n • Despite the GB300 GPU board design reverting to Bianca, it still includes one DPU (additional possible in MGX).\n • This is negative for content growth in BMC (Board Management Controller) suppliers like ASPEED.\n • However, as overall rack structures become more complex, BMC adoption is structurally increasing in non-compute areas such as switches, CDUs, and power racks.\n\n(5) Ongoing Competition in Liquid Cooling\n • All value chain players showcased liquid cooling solutions.\n • This technology encompasses multiple layers, including 1) chip, 2) rack/system, and 3) infrastructure levels, with different players and competitive dynamics at each layer.\n • Most CSPs designate suppliers (consign), limiting ODMs’ influence and opportunities to adopt their own components.\n • In chip-level cooling, AVC is positioned as a top-tier cold plate module supplier for GB200/GB300 in 2025 and Rubin in 2H26, expected to maintain solid market share within CSPs. \n\n(6) GB300 Reverts to Bianca Design\n • Several liquid cooling companies exhibited NVQD (NVIDIA Qualified Design) designs based on Cordelia GPU boards but reported reverting to Bianca boards.\n • Such design changes are negative for cooling component suppliers, as Bianca designs reduce TAM compared to NVQD.\n • Although GB300’s TDP is approximately 15% higher than GB200, companies believe existing QD designs can adequately handle the increased heat.\n\n(7) Delay in ODMs’ Adoption of In-House Liquid Cooling Components\n • Several server ODMs showcased their own liquid cooling components, including 1) cold plates, 2) manifolds, and 3) side carts.\n • Some companies also presented next-generation products like two-phase DLC (Direct Liquid Cooling), liquid metal heatsinks, and immersion cooling.\n • By increasing the adoption of in-house components, companies aim to improve margins and shorten product development cycles.\n • However, due to GB300 delays, CSPs prioritize the release schedules of existing GB200 products, leading to a temporary delay in the adoption of in-house components.\n\n(8) Data Center Power: HVDC Emerging as a Next-Generation Game Changer\n • Rubin is expected to consume approximately 200kW, while the NVL576 Kyber structure is anticipated to consume around 600kW.\n • When rack power consumption exceeds 200kW, issues arise, including rack space constraints, copper busbar overload, and inefficient power conversion.\n • To address this, proposals include increasing voltage from the existing 230V to 800V in Rubin Ultra and Kyber servers.\n • Implementing 800V HVDC (High Voltage Direct Current) power racks eliminates AC/DC conversion devices within IT racks and provides space for high-density PSUs.\n• The adoption of HVDC is expected to significantly expand the TAM for power-related infrastructure, benefiting companies like Vertiv, Delta, Flextronics, Lite-On Tech, Infineon, and MPS. \n\n(9) Separate Power Rack Structure Integrating BBU and Supercapacitors\n • Rubin is anticipated to initially adopt independent power racks.\n • While NVIDIA recommends centralized power racks supporting multiple computing racks within data centers, CSPs prefer placing individual power racks next to each computing rack to enhance power efficiency.\n • Given the ample space within power racks, the adoption of BBUs (Battery Backup Units) and supercapacitors is likely to increase, potentially doubling or tripling component content compared to existing PSUs.\n\n(10) RTX Pro 6000 Blackwell Server: Targeting Enterprise Market, Potential H20 Replacement?\n • The RTX Pro 6000 Blackwell features up to eight PCIe-based GPUs (each with 600W TDP and 96GB GDDR7 memory), air cooling, CX8 smart NIC, BF-3 DPU, and a 4U height.\n • Using GDDR7 memory reduces the burden of adopting CoWoS-based HBM, which is positive.\n • However, the RTX Pro 6000’s compute performance is over 3–4 times higher than H20 (FP32: 117TFLOPs vs. 44TFLOPs), presenting practical challenges in replacing H20.\n\n(11) NVLink Fusion\n • NVLink, previously a proprietary interconnect between NVIDIA GPUs, is now open to integration with third-party chips (ASICs, Qualcomm/Fujitsu CPUs).\n • NVIDIA aims to establish its technology as the default interconnect in custom ASIC environments.\n • CSPs express concerns about increased dependency on NVIDIA (lock-in). Monitoring actual adoption over the next 2–3 quarters is necessary.\n • Currently, initial interest is observed among some CSPs and second-tier ASIC customers.\n\n(12) CPO: NPO and CPC Preferred for Now\n • The goal is to introduce Co-Packaged Optics (CPO) in Rubin Ultra and NVL576 Kyber platforms between 2026 and 2027.\n • However, thermal domain separation issues affect signal transmission, posing technical barriers.\n • Consequently, the market currently favors Near-Packaged Optics (NPO) and Co-Packaged Copper (CPC) as transitional solutions.\n\n(13) AI PC Adoption Continues, but Innovative AI Applications Still Lacking\n • Several PC brands are attempting to differentiate their AI PC products through proprietary AI solutions.\n • However, there is still a lack of innovative AI applications compelling enough to drive replacement demand, and AI PC penetration remains lower than expected.\n • In the medium to long term, reduced AI inference costs and improved local AI performance are expected to enrich locally executed AI apps, promoting AI PC adoption in both enterprise and consumer markets.\n\n(14) MediaTek’s ASIC and Automotive Roadmap Progressing Smoothly\n • MediaTek has been producing large networking chips measuring 91x91mm for over a year and continues to strengthen its AI ASIC technology roadmap.\n • Developments include A16 compute die, 448 SerDes, 3D packaging, CPO, and custom HBM.\n • Mass production of the first data center ASIC (GOOGL TPU v7e) is expected in 2H26, with potential orders for the 2nm process TPU v8e project in 2027–28.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.027343679912099628, "ret_signed_m1d": 0.027343679912099628, "alpha_spy_m1d": 0.027343679912099628, "alpha_c_m1d": 0.027343679912099628, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.020790667721467315, "alpha_c_p1d": -0.03171325449719942, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.03171325449719942, "ret_p1w": -0.015625044918212372, "ret_signed_p1w": -0.015625044918212372, "alpha_spy_p1w": -0.03910946271097038, "alpha_c_p1w": -0.0403049841082862, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.02467993919007383, "ret_p1m": -0.007812522459106241, "ret_signed_p1m": -0.007812522459106241, "alpha_spy_p1m": -0.058690275200587894, "alpha_c_p1m": -0.15142772194953336, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.14361519949042711, "ret_p3m": 0.08739862362171102, "ret_signed_p3m": 0.08739862362171102, "alpha_spy_p3m": -0.013305670121452451, "alpha_c_p3m": -0.12326504402031335, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.21066366764202438, "ret_p6m": -0.06794399441145571, "ret_signed_p6m": -0.06794399441145571, "alpha_spy_p6m": -0.21430111025997856, "alpha_c_p6m": -0.47162496366747786, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.40368096925602215, "price_path": [0.2188, 0.1836, 0.1836, 0.1484, 0.1758, 0.1758, 0.168, 0.1445, 0.0898, 0.082, 0.1016, 0.0859, 0.0937, 0.1133, 0.1523, 0.1445, 0.1602, 0.1562, 0.1562, 0.1914, 0.1836, 0.1602, 0.1445, 0.0859, 0.1406, 0.1211, 0.1211, 0.1211, 0.0117, -0.0156, -0.0742, 0.0156, 0.082, 0.0937, 0.082, 0.0664, 0.0469, 0.0664, 0.0508, 0.0156, 0.0703, 0.0391, 0.0781, 0.0547, 0.0703, 0.0547, 0.0547, 0.0156, 0.0117, 0.0, -0.0039, 0.0078, 0.0352, 0.0273, 0.0586, 0.0625, 0.0703, 0.0664, 0.0234, 0.0273, 0.0469, 0.0352, 0.0273, 0.0, 0.0, 0.0, -0.0156, -0.0156, -0.0156, -0.0039, -0.0039, -0.0273, 0.0, 0.0039, 0.043, 0.0273, 0.0039, -0.0234, -0.0195, -0.0039, 0.0039, -0.0117, -0.0234, -0.0195, -0.0078, 0.0117, 0.0078, 0.0039, -0.0234, -0.0039, 0.0117, 0.0117, 0.0277, 0.0197, 0.0117, 0.0794, 0.1153, 0.1312, 0.1033, 0.1113, 0.1232, 0.1073, 0.1232, 0.1352, 0.1392, 0.1392, 0.1392, 0.1392, 0.1153, 0.0834, 0.0993, 0.0914, 0.0914, 0.0595, 0.0675, 0.0516, 0.0794, 0.0754, 0.0794, 0.0834, 0.0914, 0.1073, 0.0914, 0.1232, 0.1073, 0.0794, 0.0874, 0.0874, 0.1232, 0.1153, 0.1153, 0.1033, 0.0914, 0.0834, 0.0834, 0.1113, 0.1033, 0.1432, 0.1392, 0.2029, 0.187, 0.179, 0.183, 0.183, 0.2228, 0.2029, 0.2029, 0.1471, 0.1193, 0.0954, 0.0595, 0.0715, 0.0436, 0.0436, 0.0476, 0.0237, 0.0197, 0.0516, 0.0516, 0.0555, 0.0635, 0.0715, 0.0715, 0.0476, 0.0396, 0.0476, 0.0595, 0.0595, 0.0675, 0.0675, 0.0595, 0.0316, 0.0316, 0.0635, 0.0476, 0.0356, 0.0436, 0.0436, 0.0356, 0.0476, 0.0595, 0.0277, 0.0038, -0.0042, -0.0122, -0.0082, -0.0082, -0.0201, -0.0201, -0.0679]}
{"unit_id": "thread:1938072049198764472", "ticker_idx": 0, "ts": "2025-06-26T03:09:18+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs Raises TSMC Target Price to NT$1,210, Recommends as Conviction Buy", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses Goldman Sachs Conviction Buy on TSMC with PT raised to NT$1,210 on AI demand recovery, CoWoS expansion, and 28.7%/17.1% revenue growth forecasts for 2025/2026.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Goldman Sachs Raises TSMC Target Price to NT$1,210, Recommends as Conviction Buy\n\nTaiwan Semiconductor Manufacturing Company (TSMC, 2330) has seen its growth momentum further strengthened. According to the latest report from Goldman Sachs, market concerns over AI order cuts have significantly eased. This, coupled with the rapid expansion of CoWoS advanced packaging technology adoption, has driven an upward revision in overall revenue forecasts. Goldman Sachs has raised its target price from NT$1,145 to NT$1,210, maintained its \"Buy\" rating, and placed it on its Conviction Buy list.\n\nGoldman Sachs has raised its earnings forecasts for TSMC by 2% to 6% for 2025 to 2027, primarily based on two key factors. First, an upward revision of wafer revenue forecasts for the 3nm and 5nm processes, reflecting a significant easing of market concerns about AI order reductions. Second, as more customers adopt CoWoS technology, there's a comprehensive upward revision in demand for advanced packaging. With the gradual alleviation of supply chain bottlenecks, the risk of large-scale AI order cuts in the short term has significantly decreased.\n\nKelvin Lu, Goldman Sachs' technology industry analyst, commented that TSMC's multiple growth drivers are firing up simultaneously, indicating a continuously optimistic operational outlook. Looking ahead to 2025 and 2026, revenue is expected to increase by 28.7% and 17.1% year-on-year, respectively, higher than the previous estimates of 26.3% and 14.1% (in USD terms).\n\nThe primary reasons for this growth come from two major catalysts: First, as process nodes continue to advance, the average selling price (ASP) increases concurrently. AI customers are accelerating their upgrades from N4 to N3, while non-AI customers, such as smartphone and PC manufacturers, will gradually adopt the N2 process. Second, advanced processes and CoWoS packaging are expected to undergo a new round of price adjustments, further boosting overall revenue momentum and helping to offset negative factors such as increased costs from US fab construction and rising energy costs in Taiwan.\n\nLooking ahead, Goldman Sachs reiterates its \"Buy\" rating on TSMC (including ADRs) and remains optimistic about its long-term growth potential. As structural demand from 5G, AI, High-Performance Computing (HPC), and Electric Vehicles (EV) continues to expand, TSMC, with its technological leadership and excellent execution, demonstrates a strong competitive advantage in advanced process and packaging fields, possessing superior growth momentum compared to peers.\n\n$TSM\n\n---\n\nhttps://t.co/bNWxEhyjMB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005669383808963202, "ret_signed_m1d": -0.005669383808963202, "alpha_spy_m1d": 0.0020936597217835473, "alpha_c_m1d": 0.0015456551182110312, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020356271161076522, "ret_signed_p1d": 0.020356271161076522, "alpha_spy_p1d": 0.015387939397071282, "alpha_c_p1d": 0.015955111772028463, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": 0.04816746631319102, "ret_signed_p1w": 0.04816746631319102, "alpha_spy_p1w": 0.02615291180117163, "alpha_c_p1w": 0.025007294061356378, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": 0.09637959346135139, "ret_signed_p1m": 0.09637959346135139, "alpha_spy_p1m": 0.0551452968020083, "alpha_c_p1m": 0.05925847052463307, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.2660233628385713, "ret_signed_p3m": 0.2660233628385713, "alpha_spy_p3m": 0.1791114584790583, "alpha_c_p3m": 0.10548943692665014, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.2974603307004282, "ret_signed_p6m": 0.2974603307004282, "alpha_spy_p6m": 0.17876913020468144, "alpha_c_p6m": 0.012359336941542454, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.2617, -0.2499, -0.2421, -0.3, -0.3471, -0.3499, -0.3712, -0.2939, -0.3278, -0.3013, -0.3069, -0.3002, -0.3254, -0.3251, -0.3251, -0.3423, -0.3266, -0.2981, -0.2698, -0.2657, -0.2732, -0.2684, -0.2586, -0.2318, -0.2026, -0.2154, -0.2337, -0.2237, -0.2207, -0.2149, -0.1683, -0.1372, -0.1337, -0.1361, -0.1361, -0.1393, -0.1396, -0.1471, -0.1274, -0.1461, -0.1461, -0.1208, -0.1276, -0.1231, -0.1401, -0.1334, -0.1211, -0.0998, -0.0956, -0.0874, -0.0793, -0.055, -0.0477, -0.0383, -0.0576, -0.0372, -0.0451, -0.0469, -0.0469, -0.0647, -0.0611, -0.0175, -0.0057, 0.0, 0.0204, 0.0111, 0.003, 0.0428, 0.0482, 0.0482, 0.023, 0.0172, 0.035, 0.0257, 0.0285, 0.0208, 0.0578, 0.0605, 0.0964, 0.0732, 0.0662, 0.0473, 0.0729, 0.0785, 0.0964, 0.0837, 0.0773, 0.0844, 0.0786, 0.05, 0.0669, 0.0378, 0.0329, 0.0831, 0.0795, 0.0807, 0.0905, 0.0778, 0.0758, 0.0664, 0.0777, 0.0388, 0.0205, 0.0148, 0.0401, 0.0517, 0.0657, 0.0682, 0.0637, 0.0306, 0.0306, 0.0196, 0.0329, 0.05, 0.0866, 0.1035, 0.1201, 0.1626, 0.1558, 0.1577, 0.1668, 0.1735, 0.1768, 0.203, 0.1861, 0.2209, 0.266, 0.2571, 0.2389, 0.2242, 0.2236, 0.2507, 0.2918, 0.2902, 0.3085, 0.3542, 0.3167, 0.3637, 0.3429, 0.2568, 0.3564, 0.3253, 0.3645, 0.3427, 0.3214, 0.3332, 0.3189, 0.2937, 0.3019, 0.3209, 0.3356, 0.3503, 0.3662, 0.3579, 0.3454, 0.3652, 0.3168, 0.315, 0.2953, 0.283, 0.3223, 0.3039, 0.3014, 0.2637, 0.2755, 0.2629, 0.2445, 0.2645, 0.2427, 0.2318, 0.2747, 0.2748, 0.2985, 0.2985, 0.3054, 0.2883, 0.308, 0.3231, 0.3118, 0.3198, 0.3518, 0.3587, 0.3889, 0.3689, 0.3113, 0.292, 0.2881, 0.2436, 0.2783, 0.2975]}
{"unit_id": "thread:1938072049198764472", "ticker_idx": 1, "ts": "2025-06-26T03:09:18+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs has raised its target price from NT$1,145 to NT$1,210, maintained its \"Buy\" rating, and placed it on its Conviction Buy list", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses Goldman Sachs Conviction Buy on TSMC with PT raised to NT$1,210 on AI demand recovery, CoWoS expansion, and 28.7%/17.1% revenue growth forecasts for 2025/2026.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Goldman Sachs Raises TSMC Target Price to NT$1,210, Recommends as Conviction Buy\n\nTaiwan Semiconductor Manufacturing Company (TSMC, 2330) has seen its growth momentum further strengthened. According to the latest report from Goldman Sachs, market concerns over AI order cuts have significantly eased. This, coupled with the rapid expansion of CoWoS advanced packaging technology adoption, has driven an upward revision in overall revenue forecasts. Goldman Sachs has raised its target price from NT$1,145 to NT$1,210, maintained its \"Buy\" rating, and placed it on its Conviction Buy list.\n\nGoldman Sachs has raised its earnings forecasts for TSMC by 2% to 6% for 2025 to 2027, primarily based on two key factors. First, an upward revision of wafer revenue forecasts for the 3nm and 5nm processes, reflecting a significant easing of market concerns about AI order reductions. Second, as more customers adopt CoWoS technology, there's a comprehensive upward revision in demand for advanced packaging. With the gradual alleviation of supply chain bottlenecks, the risk of large-scale AI order cuts in the short term has significantly decreased.\n\nKelvin Lu, Goldman Sachs' technology industry analyst, commented that TSMC's multiple growth drivers are firing up simultaneously, indicating a continuously optimistic operational outlook. Looking ahead to 2025 and 2026, revenue is expected to increase by 28.7% and 17.1% year-on-year, respectively, higher than the previous estimates of 26.3% and 14.1% (in USD terms).\n\nThe primary reasons for this growth come from two major catalysts: First, as process nodes continue to advance, the average selling price (ASP) increases concurrently. AI customers are accelerating their upgrades from N4 to N3, while non-AI customers, such as smartphone and PC manufacturers, will gradually adopt the N2 process. Second, advanced processes and CoWoS packaging are expected to undergo a new round of price adjustments, further boosting overall revenue momentum and helping to offset negative factors such as increased costs from US fab construction and rising energy costs in Taiwan.\n\nLooking ahead, Goldman Sachs reiterates its \"Buy\" rating on TSMC (including ADRs) and remains optimistic about its long-term growth potential. As structural demand from 5G, AI, High-Performance Computing (HPC), and Electric Vehicles (EV) continues to expand, TSMC, with its technological leadership and excellent execution, demonstrates a strong competitive advantage in advanced process and packaging fields, possessing superior growth momentum compared to peers.\n\n$TSM\n\n---\n\nhttps://t.co/bNWxEhyjMB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004651043759266282, "ret_signed_m1d": -0.004651043759266282, "alpha_spy_m1d": 0.003111999771480467, "alpha_c_m1d": 0.002563995167907951, "bench_spy_m1d": -0.007763043530746749, "bench_c_m1d": -0.007215038927174233, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0046512732817574065, "ret_signed_p1d": 0.0046512732817574065, "alpha_spy_p1d": -0.00031705848224783395, "alpha_c_p1d": 0.00025011389270934714, "bench_spy_p1d": 0.0049683317640052405, "bench_c_p1d": 0.004401159389048059, "ret_p1w": 0.013953590322780984, "ret_signed_p1w": 0.013953590322780984, "alpha_spy_p1w": -0.008060964189238407, "alpha_c_p1w": -0.009206581929053659, "bench_spy_p1w": 0.02201455451201939, "bench_c_p1w": 0.023160172251834643, "ret_p1m": 0.06511644880965628, "ret_signed_p1m": 0.06511644880965628, "alpha_spy_p1m": 0.023882152150313196, "alpha_c_p1m": 0.02799532587293796, "bench_spy_p1m": 0.041234296659343084, "bench_c_p1m": 0.03712112293671832, "ret_p3m": 0.2514978924671063, "ret_signed_p3m": 0.2514978924671063, "alpha_spy_p3m": 0.16458598810759328, "alpha_c_p3m": 0.09096396655518513, "bench_spy_p3m": 0.08691190435951301, "bench_c_p3m": 0.16053392591192117, "ret_p6m": 0.34000555214905925, "ret_signed_p6m": 0.34000555214905925, "alpha_spy_p6m": 0.22131435165331248, "alpha_c_p6m": 0.0549045583901735, "bench_spy_p6m": 0.11869120049574677, "bench_c_p6m": 0.28510099375888576, "price_path": [-0.1571, -0.1256, -0.1274, -0.1274, -0.1274, -0.2145, -0.2441, -0.2729, -0.2006, -0.1765, -0.1987, -0.1876, -0.208, -0.2154, -0.2126, -0.2265, -0.2441, -0.1913, -0.1997, -0.1774, -0.1691, -0.1645, -0.1589, -0.1589, -0.12, -0.1311, -0.1478, -0.1404, -0.1497, -0.1209, -0.1135, -0.1024, -0.0746, -0.0802, -0.0756, -0.0885, -0.0941, -0.083, -0.0904, -0.0885, -0.0959, -0.1061, -0.1043, -0.1043, -0.1043, -0.1237, -0.12, -0.083, -0.0756, -0.0783, -0.0691, -0.032, -0.0135, -0.0279, -0.0419, -0.0465, -0.0279, -0.0186, -0.0372, -0.0186, -0.0512, -0.0233, -0.0047, 0.0, 0.0047, -0.014, 0.0093, 0.0093, 0.014, 0.0093, 0.0047, 0.0047, 0.014, 0.0233, 0.0233, 0.0186, 0.0326, 0.0512, 0.0512, 0.0744, 0.0698, 0.0512, 0.0651, 0.0651, 0.0651, 0.0651, 0.0558, 0.0744, 0.0791, 0.0791, 0.0558, 0.0698, 0.0465, 0.0977, 0.093, 0.0977, 0.0977, 0.1163, 0.093, 0.0977, 0.0977, 0.1023, 0.0558, 0.0698, 0.0558, 0.0884, 0.093, 0.107, 0.0791, 0.0791, 0.0837, 0.0791, 0.0791, 0.0791, 0.0977, 0.0977, 0.1163, 0.1395, 0.1535, 0.1721, 0.1674, 0.1955, 0.1815, 0.2001, 0.1815, 0.2095, 0.2515, 0.2515, 0.2328, 0.2141, 0.2141, 0.2188, 0.2375, 0.2748, 0.3075, 0.3075, 0.3402, 0.3215, 0.3449, 0.3449, 0.3215, 0.3309, 0.3682, 0.3869, 0.3542, 0.3823, 0.3823, 0.3636, 0.3542, 0.3542, 0.3823, 0.3776, 0.4056, 0.4056, 0.4009, 0.4103, 0.4056, 0.3636, 0.3682, 0.3636, 0.3776, 0.3682, 0.3776, 0.3636, 0.3356, 0.3496, 0.3122, 0.3029, 0.3589, 0.2935, 0.2842, 0.3215, 0.3449, 0.3402, 0.3449, 0.3169, 0.3356, 0.3542, 0.3496, 0.3636, 0.3963, 0.3823, 0.4056, 0.3775, 0.3869, 0.3587, 0.3447, 0.34, 0.34, 0.34]}
{"unit_id": "orig:1928233497979707415", "ticker_idx": 0, "ts": "2025-05-29T23:34:25+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI demand now accounts for the majority of data center revenue, and continued growth is expected... AI-related revenue expected to further increase going forward... Demand for AI and data centers is expected to persist", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MRVL: AI majority of data center revenue, XPU mass production underway, new hyperscale customers emerging, optics and enterprise networking recovering.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell FY1Q26 Earnings Release\n(Shares down 3% after-hours, concerns about Taiwanese competitors resurfacing)\n\n▶️ Key Comments – FY4Q25\n• AI demand now accounts for the majority of data center revenue, and continued growth is expected (Data center accounts for 76% of total revenue).\n• Ramp-up in mass production of AI silicon programs and strong shipments of electro-optical products for AI and cloud applications.\n• AI-related revenue expected to further increase going forward.\n• NVLink Fusion partnership is accelerating the development of custom expansion solutions.\n• The XPU program for a major U.S. hyperscaler is emerging as a core revenue driver.\n\n⸻\n\n▶️ Key Q&A – FY4Q25\n\nQ: Direction of content for the next-gen XPU program and whether Marvell has exclusive rights to 3nm XPU?\nA: Information from Asian supply chains is incomplete, and Marvell strictly manages customer confidentiality. We are currently supplying this generation of AI XPUs and have successfully moved from A0 (initial success) to mass production. Follow-up XPU programs are underway. We’ve secured 3nm wafer and advanced packaging capacity, and production is scheduled to begin in 2026.\n\nQ: Can Marvell support a broader customer base beyond the initial 3–4 customers?\nA: Marvell has sufficient capabilities to expand its portfolio and customer collaborations in terms of engineering capacity. We’re intensifying R&D investments and reallocating talent to focus on data center and AI opportunities. New hyperscale customers are emerging who can utilize custom technologies.\n\nQ: On the partnership with NVIDIA?\nA: There is strong confidence in the partnership with NVIDIA. The NVLink Fusion collaboration demonstrates Marvell’s deep engagement with key ecosystem partners, and it proves that even NVIDIA recognizes the complementary role of custom XPUs.\n\nQ: Data center revenue breakdown (AI share) and full-year AI revenue outlook?\nA: As of Q4, AI accounts for more than half of data center revenue, and we are on a path where AI could soon account for more than half of total company revenue. The mix changes quarter to quarter. The custom business inherently carries lower gross margins.\n\nQ: While there is no official guidance for the second half yet, what variables are being considered?\nA: We’re closely monitoring macroeconomic volatility. Demand for AI and data centers is expected to persist. Core businesses—especially enterprise networking and telco—are showing a strong recovery and are expected to grow throughout the year.\n\nQ: Status of the optics business and impact of Amazon collaboration beyond custom silicon?\nA: The Amazon contract includes a range of opportunities in networking and connectivity, and it’s progressing smoothly. The optics business is performing very well and is expected to grow throughout the year.\n\nQ: Major AI GPU suppliers are preparing to ship next-gen platforms. Is Marvell’s 1.6T solution expected to ramp up strongly starting now?\nA: We’ve started shipping our 5nm-based 1.6T products. There is also very strong demand for 3nm products, and development is proceeding successfully.\n\nQ: Marvell previously indicated a target gross margin for ASICs. Is this still valid, and what is the sustainable gross margin for AI ASICs?\nA: The overall gross margin for the custom business remains within the previously mentioned range. While larger volumes tend to lower gross margin percentages, operating profit increases significantly due to scale.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013494309059614018, "ret_signed_m1d": 0.013494309059614018, "alpha_spy_m1d": 0.017426066662713957, "alpha_c_m1d": 0.020541162495196552, "bench_spy_m1d": -0.003931757603099939, "bench_c_m1d": -0.007046853435582534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.055546912821340166, "ret_signed_p1d": -0.055546912821340166, "alpha_spy_p1d": -0.054428560629436795, "alpha_c_p1d": -0.037806804829621954, "bench_spy_p1d": -0.0011183521919033712, "bench_c_p1d": -0.017740107991718213, "ret_p1w": 0.022438338592797313, "ret_signed_p1w": 0.022438338592797313, "alpha_spy_p1w": 0.017354015946568158, "alpha_c_p1w": -0.007142116140331334, "bench_spy_p1w": 0.005084322646229156, "bench_c_p1w": 0.029580454733128647, "ret_p1m": 0.21073278742462964, "ret_signed_p1m": 0.21073278742462964, "alpha_spy_p1m": 0.16551992908424795, "alpha_c_p1m": 0.0700411658381257, "bench_spy_p1m": 0.04521285834038169, "bench_c_p1m": 0.14069162158650395, "ret_p3m": 0.166182133379716, "ret_signed_p3m": 0.166182133379716, "alpha_spy_p3m": 0.06955081232004257, "alpha_c_p3m": -0.049607779934931484, "bench_spy_p3m": 0.09663132105967343, "bench_c_p3m": 0.21578991331464747, "ret_p6m": 0.21708327646534253, "ret_signed_p6m": 0.21708327646534253, "alpha_spy_p6m": 0.0937702211342688, "alpha_c_p6m": -0.11907706760712822, "bench_spy_p6m": 0.12331305533107373, "bench_c_p6m": 0.33616034407247075, "price_path": [0.3452, 0.3844, 0.4128, 0.1329, 0.1103, 0.0293, 0.0526, 0.0948, 0.0777, 0.0774, 0.1034, 0.0702, 0.0942, 0.1048, 0.1032, 0.1404, 0.1233, 0.0464, 0.0166, -0.0276, -0.035, -0.0177, -0.009, -0.1279, -0.2253, -0.2007, -0.2159, -0.0446, -0.1713, -0.1622, -0.18, -0.1633, -0.1852, -0.1888, -0.1888, -0.2252, -0.2057, -0.1563, -0.1004, -0.0755, -0.0789, -0.0792, -0.0841, -0.0436, -0.022, -0.0275, -0.0394, -0.1164, -0.0957, -0.064, 0.0121, 0.0275, 0.034, 0.0231, 0.0005, -0.0184, -0.0362, -0.0568, -0.0295, -0.0477, -0.0477, 0.0014, 0.0135, 0.0, -0.0555, -0.0355, -0.0215, 0.0403, 0.0224, 0.0725, 0.0849, 0.0802, 0.0708, 0.0927, 0.0543, 0.105, 0.0982, 0.1761, 0.1761, 0.1535, 0.1106, 0.1801, 0.1914, 0.2548, 0.2107, 0.2145, 0.1963, 0.1651, 0.1797, 0.1797, 0.1227, 0.129, 0.1338, 0.1511, 0.1418, 0.1387, 0.1371, 0.1126, 0.1308, 0.1723, 0.1473, 0.1305, 0.1506, 0.1627, 0.1654, 0.1921, 0.1988, 0.2836, 0.2621, 0.1692, 0.2018, 0.2034, 0.1828, 0.1912, 0.2146, 0.2136, 0.2219, 0.2456, 0.2412, 0.1965, 0.2051, 0.1318, 0.1184, 0.1183, 0.1464, 0.1456, 0.1662, 0.1745, 0.2128, -0.0127, -0.0127, 0.0145, -0.0215, 0.0066, -0.0055, 0.0365, 0.0497, 0.0537, 0.0457, 0.0577, 0.0589, 0.0814, 0.1147, 0.1657, 0.1662, 0.1861, 0.1718, 0.2577, 0.3162, 0.3061, 0.2939, 0.3202, 0.3174, 0.3537, 0.354, 0.3964, 0.3658, 0.4526, 0.424, 0.3453, 0.4047, 0.3549, 0.3969, 0.3865, 0.3821, 0.3489, 0.3241, 0.2737, 0.3007, 0.3221, 0.394, 0.3901, 0.4167, 0.3918, 0.4731, 0.4201, 0.3764, 0.4599, 0.4666, 0.4288, 0.4651, 0.4038, 0.4038, 0.3753, 0.3585, 0.3114, 0.2364, 0.2779, 0.205, 0.2171]}
{"unit_id": "orig:1930780932652584999", "ticker_idx": 0, "ts": "2025-06-06T00:17:01+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS raises its price target to $120 and maintains a Buy rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS Buy on MU, PT raised to $120; DDR ASP recovery, HBM execution lead, and gross margin expansion to ~40% in FQ4.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron: DDR Demand Recovery Drives Earnings; HBM Outlook Revised but Long-Term Growth Intact\nUBS (T. Arcuri, June 4, 2025)\n\nUBS believes the DDR demand environment is stabilizing, with potential sequential pricing increases in 2Q and 3Q. While HBM demand forecasts have been revised downward, Micron remains a leader in the high-end segment thanks to strong execution, and its long-term profitability outlook remains positive. UBS raises its price target to $120 and maintains a Buy rating.\n\n• DDR Demand Stabilizing and ASPs Trending Upward\n\nRecent industry checks indicate that inventory build-up from smartphone and PC customers has eased, while cloud service providers have resumed restocking DDR5 for servers, pushing customer inventories lower. UBS raises its 2Q25 DDR contract ASP estimate to +7% QoQ and 3Q25 ASP to +3% QoQ. For NAND, although suppliers continue to maintain tight production discipline, UBS keeps its existing estimates unchanged at +3% QoQ for 2Q25 and -3% QoQ for 3Q25.\n\n• Inventory Replenishment Pace and Tariff Uncertainty\n\nSo far, DRAM customer inventory levels have fallen by about three weeks, supporting resilient 2Q25 shipments. UBS expects FQ3:25 (May) revenue/EPS to be $9.25B/$1.78, with DRAM revenue of $7.17B (bit growth +15% QoQ, ASP +2% QoQ) and NAND revenue of $2.04B (bit growth +12% QoQ, ASP -2% QoQ).\n\nLooking ahead to FQ4:25 (August), front-loaded demand due to tariff concerns will begin to fade. UBS forecasts revenue of $10.1B and EPS of $2.09, with DRAM revenue of $8.12B (bit growth +9% QoQ, ASP +4% QoQ) and NAND revenue of $1.91B (bit -5% QoQ, ASP -2% QoQ).\n\n• HBM Demand Revised Down but Execution Advantage Remains\n\nDue to delays in volume production of AI ASICs like Google’s TPU v6e/v7 and AWS’s Trainium 3, UBS revises its 2025 HBM bit demand forecast down from 18.9B GB to 16.3B GB, and 2026 from 29.1B GB to 25.4B GB.\n\nHowever, with Samsung’s certification of its HBM3E 12-Hi for NVIDIA postponed to 4Q25, UBS believes Micron maintains a clear execution lead in high-end HBM3E and upcoming HBM4 technologies. UBS revises Micron’s HBM shipment forecasts to 75.9 million GB in 2025 and 124.5 million GB in 2026, and believes the pricing premium for HBM4 can offset revenue impacts from lower bit demand.\n\n• FQ3 and FQ4 Profitability Expected to Outperform\n\nBased on the above demand and pricing assumptions, UBS raises its FQ3:25 revenue/non-GAAP EPS forecast to $9.25B/$1.78, a significant upward revision. UBS expects FQ3:25 gross margin to exceed guidance (36.5%) by about 150 basis points.\n\nFor FQ4:25, with management indicating that tariff-related uncertainty is easing, UBS expects gross margin to rise another 200bps QoQ to around 40%, far exceeding prior market expectations.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02091003034336192, "ret_signed_m1d": -0.02091003034336192, "alpha_spy_m1d": -0.01074543055490118, "alpha_c_m1d": -0.015251804482363851, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022015545117126667, "ret_signed_p1d": 0.022015545117126667, "alpha_spy_p1d": 0.021114262228982383, "alpha_c_p1d": 0.005278398433741227, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.06484897673487722, "ret_signed_p1w": 0.06484897673487722, "alpha_spy_p1w": 0.06842072362348317, "alpha_c_p1w": 0.04799312516540488, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.10568243610849493, "ret_signed_p1m": 0.10568243610849493, "alpha_spy_p1m": 0.06666833469433509, "alpha_c_p1m": -0.0012696972363375192, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.0946182007900278, "ret_signed_p3m": 0.0946182007900278, "alpha_spy_p3m": 0.01700184518977621, "alpha_c_p3m": -0.03872566255496923, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 1.218469810598613, "ret_signed_p6m": 1.218469810598613, "alpha_spy_p6m": 1.0765452156323814, "alpha_c_p6m": 0.8218805411327075, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.1808, -0.1202, -0.1271, -0.0728, -0.0514, -0.0642, -0.0611, -0.0524, -0.1286, -0.1082, -0.1336, -0.1524, -0.1614, -0.1864, -0.1996, -0.1828, -0.1839, -0.3152, -0.4038, -0.3702, -0.3963, -0.2827, -0.3547, -0.3593, -0.3458, -0.3456, -0.3614, -0.3662, -0.3662, -0.3852, -0.3533, -0.3282, -0.2868, -0.2651, -0.2763, -0.2918, -0.2912, -0.2836, -0.2564, -0.2592, -0.2584, -0.2389, -0.2156, -0.2091, -0.1499, -0.1071, -0.122, -0.1208, -0.0973, -0.0913, -0.0964, -0.1172, -0.1265, -0.1399, -0.1399, -0.1122, -0.114, -0.1083, -0.1299, -0.0956, -0.0581, -0.0489, -0.0209, 0.0, 0.022, 0.0514, 0.0688, 0.0702, 0.0648, 0.1039, 0.1085, 0.1221, 0.1221, 0.1385, 0.1245, 0.1782, 0.1722, 0.1606, 0.1492, 0.1353, 0.1136, 0.1214, 0.1265, 0.1265, 0.1057, 0.1472, 0.1271, 0.1351, 0.1482, 0.0936, 0.1074, 0.0735, 0.0443, 0.0547, 0.044, 0.007, 0.0127, 0.0302, 0.0258, 0.0257, 0.0323, 0.0579, 0.0063, -0.033, -0.0063, 0.0056, 0.003, 0.0315, 0.0962, 0.1407, 0.1779, 0.1458, 0.1552, 0.1144, 0.1392, 0.1253, 0.0807, 0.0676, 0.085, 0.0734, 0.0741, 0.0857, 0.1249, 0.0973, 0.0973, 0.0924, 0.0946, 0.1452, 0.2113, 0.2121, 0.2469, 0.2908, 0.3883, 0.4497, 0.4547, 0.4643, 0.4751, 0.5572, 0.5004, 0.5178, 0.5343, 0.491, 0.446, 0.4501, 0.5112, 0.5427, 0.6795, 0.6942, 0.7329, 0.7618, 0.7132, 0.8133, 0.7744, 0.6754, 0.7785, 0.7258, 0.7708, 0.8685, 0.8671, 0.9076, 0.8663, 0.8311, 0.9071, 1.0207, 1.0306, 1.0473, 1.0909, 1.0667, 1.0645, 1.1653, 1.0115, 1.1912, 1.1988, 1.195, 1.3369, 1.2245, 1.2594, 1.1861, 1.2772, 1.2322, 1.1081, 1.0843, 0.8578, 0.9132, 1.066, 1.0715, 1.1244, 1.1244, 1.1818, 1.2185]}
{"unit_id": "reply:1926245222825984345", "ticker_idx": 0, "ts": "2025-05-24T11:53:43+00:00", "symbol": "Korean nuclear", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "That's exactly why I want to invest in Korean nuclear-related stocks. If they're not going to accept Chinese reactors, Korea is the only real alternative.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long Korean nuclear-related stocks as the only non-Chinese reactor alternative for allies.", "resolved_tickers": ["034020.KS", "006360.KS"], "resolution_method": "basket", "resolution_reason": "Korea nuclear basket: Doosan Enerbility, GS E&C; 017890 mismatch removed", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@AngelicaOung That’s exactly why I want to invest in Korean nuclear-related stocks.\nIf they’re not going to accept Chinese reactors, Korea is the only real alternative.", "reply_text": null, "parent_text": "The problem with Europe (and Taiwan) is they are amongst the American allies who genuinely did very well for themselves thanks to America’s largesse since time out of mind.\n\nThey can see with their own eyeballs, of course, America betraying other allies on the mildest", "quoted_text": null, "parent_author": "AngelicaOung", "ret_m1d": -0.019173592537026385, "ret_signed_m1d": -0.019173592537026385, "alpha_spy_m1d": -0.019173592537026385, "alpha_c_m1d": -0.019173592537026385, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011434254917417008, "ret_signed_p1d": 0.011434254917417008, "alpha_spy_p1d": -0.009356412804050307, "alpha_c_p1d": -0.006175740708881494, "bench_spy_p1d": 0.020790667721467315, "bench_c_p1d": 0.017609995626298502, "ret_p1w": 0.0315596649012681, "ret_signed_p1w": 0.0315596649012681, "alpha_spy_p1w": 0.008075247108510097, "alpha_c_p1w": 0.018707153954111877, "bench_spy_p1w": 0.023484417792758006, "bench_c_p1w": 0.012852510947156226, "ret_p1m": 0.3539868268540981, "ret_signed_p1m": 0.3539868268540981, "alpha_spy_p1m": 0.30310907411261645, "alpha_c_p1m": 0.32081554631491727, "bench_spy_p1m": 0.05087775274148165, "bench_c_p1m": 0.033171280539180836, "ret_p3m": 0.1917991686866558, "ret_signed_p3m": 0.1917991686866558, "alpha_spy_p3m": 0.09109487494349233, "alpha_c_p3m": 0.11860011959062189, "bench_spy_p3m": 0.10070429374316348, "bench_c_p3m": 0.07319904909603392, "ret_p6m": 0.3487678488198597, "ret_signed_p6m": 0.3487678488198597, "alpha_spy_p6m": 0.20241073297133688, "alpha_c_p6m": 0.2770708801145007, "bench_spy_p6m": 0.14635711584852285, "bench_c_p6m": 0.07169696870535902, "price_path": [-0.2267, -0.2316, -0.2746, -0.2746, -0.2761, -0.2415, -0.2444, -0.2436, -0.2396, -0.2633, -0.2522, -0.2652, -0.2534, -0.2632, -0.2639, -0.2542, -0.2753, -0.2551, -0.2633, -0.2701, -0.2805, -0.2836, -0.298, -0.3127, -0.3076, -0.3124, -0.3183, -0.3225, -0.3801, -0.3753, -0.3839, -0.3475, -0.3291, -0.3285, -0.3161, -0.3237, -0.3131, -0.29, -0.2902, -0.2948, -0.2886, -0.2673, -0.2524, -0.2299, -0.2258, -0.2156, -0.2156, -0.2429, -0.2429, -0.2429, -0.2367, -0.2401, -0.2375, -0.2003, -0.1938, -0.1709, -0.1728, -0.1285, -0.1244, -0.095, -0.0669, -0.085, -0.0192, 0.0, 0.0114, 0.0189, 0.0633, 0.0267, 0.0316, 0.0316, 0.0652, 0.1139, 0.1139, 0.1308, 0.1511, 0.192, 0.2223, 0.2068, 0.2861, 0.2694, 0.2743, 0.2648, 0.2468, 0.3384, 0.354, 0.3274, 0.2962, 0.3011, 0.3193, 0.2506, 0.2457, 0.2541, 0.2193, 0.2483, 0.3024, 0.2866, 0.2579, 0.2277, 0.2325, 0.2891, 0.2567, 0.2531, 0.2724, 0.3074, 0.2327, 0.2483, 0.2665, 0.2725, 0.2328, 0.2664, 0.2345, 0.2522, 0.188, 0.2258, 0.2437, 0.2651, 0.2543, 0.2649, 0.291, 0.2419, 0.2437, 0.2324, 0.2324, 0.2292, 0.1605, 0.1355, 0.1918, 0.2045, 0.2498, 0.2118, 0.2113, 0.2089, 0.189, 0.1584, 0.1707, 0.1831, 0.2053, 0.1918, 0.1913, 0.2076, 0.2146, 0.1956, 0.1987, 0.1729, 0.2297, 0.192, 0.1877, 0.1877, 0.2003, 0.212, 0.2422, 0.2148, 0.196, 0.1986, 0.1946, 0.2198, 0.2254, 0.2254, 0.2254, 0.2254, 0.2254, 0.2254, 0.3463, 0.3791, 0.3572, 0.4592, 0.4519, 0.4021, 0.4088, 0.4085, 0.4031, 0.3668, 0.4289, 0.4402, 0.4908, 0.6173, 0.5208, 0.5147, 0.5175, 0.5237, 0.4409, 0.4008, 0.3708, 0.4032, 0.3901, 0.3957, 0.4542, 0.389, 0.4017, 0.3488]}
{"unit_id": "orig:1930224984196296865", "ticker_idx": 0, "ts": "2025-06-04T11:27:52+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I remain positive on companies like Samsung Electronics and SK Hynix, which are broadly involved across the memory value chain", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung and SK Hynix on broad memory value chain exposure; avoids SSD-pure-plays Kioxia and Western Digital and HDD-only companies.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Personally, I don’t invest in companies that only make SSDs—such as Kioxia or Western Digital.\nThe same goes for companies that focus only on HDDs.\nHowever, I remain positive on companies like Samsung Electronics and SK Hynix, which are broadly involved across the memory value chain.\nAs of now, SSDs are still more expensive than HDDs. In fact, over 90% of storage infrastructure in data centers is still based on HDDs.\nSSDs are used in specific workloads that require high throughput, but HDDs remain dominant due to their cost advantage.\nSome market research firms forecast that SSDs will become cheaper than HDDs by around 2029.\nI also believe that only around then will SSD-focused companies begin to generate meaningful profits that satisfy investors.\n\nStill, despite this outlook, I don’t take long positions on SSDs, nor do I go short on HDDs just based on that expectation.\nOn the surface, SSDs appear to have the long-term edge, and HDDs may look like a declining asset. But I have my reasons for staying cautious.\n\nHere's why:\n\n1. “SSDs will be cheaper by 2029” is just a forecast, not a fact.\n\nBack in 2014–2015, major players like Samsung and SK Hynix boldly claimed SSDs would surpass HDDs in price competitiveness within a few years.\nBut what happened?\n SSD prices dropped—but so did HDD prices.\nEven now, in 2025, the price gap remains substantial.\nThis is exactly why I can’t fully trust optimistic projections about SSDs.\n\n2. Will the AI datacenter boom still be alive in 2029?\n\nToday, HDDs, HBM, and GPUs are flying off the shelves. But that’s not because of intrinsic demand—it’s because hyperscalers are throwing massive CAPEX into infrastructure.\nWill CSPs still be investing at that same pace in 2029?\nAccording to my estimates, about 60–70% of HDD companies' profits currently come from AI-related demand.\nNow suppose SSDs truly replace HDDs by 2029.\n Can SSD makers really expect 60–70% of their earnings to come from AI infrastructure as well?\n\nI seriously doubt it.\n\nConclusion\nI don’t have a crystal ball, but I do know how to invest in what I can reasonably predict over the next 1–2 years.\nFor me, betting on outcomes 4–5 years into the future is simply too risky.\nSure, some people might make great returns by betting on SSDs.\nBut I believe there are plenty of other companies in the market that offer much higher conviction with far less uncertainty.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01730103496863622, "ret_signed_m1d": -0.01730103496863622, "alpha_spy_m1d": -0.01756965450868253, "alpha_c_m1d": -0.015822881811775713, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.0014781531568605066, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022491352336643455, "ret_signed_p1d": 0.022491352336643455, "alpha_spy_p1d": 0.02732412139283269, "alpha_c_p1d": 0.025320833900135398, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0028294815634919424, "ret_p1w": 0.03633229035021457, "ret_signed_p1w": 0.03633229035021457, "alpha_spy_p1w": 0.027220359188089516, "alpha_c_p1w": 0.021383106441884125, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.014949183908330443, "ret_p1m": 0.11057675114151366, "ret_signed_p1m": 0.11057675114151366, "alpha_spy_p1m": 0.05812301057557079, "alpha_c_p1m": 0.02364321288392368, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.08693353825758998, "ret_p3m": 0.17672401698336704, "ret_signed_p3m": 0.17672401698336704, "alpha_spy_p3m": 0.09109819768341043, "alpha_c_p3m": 0.06656781062595396, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.11015620635741308, "ret_p6m": 0.8096791932845606, "ret_signed_p6m": 0.8096791932845606, "alpha_spy_p6m": 0.6625992731183294, "alpha_c_p6m": 0.6077551380447024, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.20192405523985824, "price_path": [-0.0764, -0.0764, -0.0781, -0.0558, -0.0592, -0.0592, -0.0093, -0.0093, 0.0061, 0.0354, 0.0612, 0.0405, 0.0285, 0.056, 0.0629, 0.0415, 0.0, 0.0173, 0.0173, -0.0035, -0.0294, -0.0796, -0.0744, -0.083, -0.0242, -0.045, -0.0277, -0.0208, -0.0536, -0.0467, -0.0433, -0.0415, -0.0484, -0.0363, -0.0363, -0.0363, -0.0346, -0.0346, -0.0398, -0.0398, -0.0606, -0.0606, -0.0606, -0.0554, -0.0554, -0.0519, -0.0035, -0.0156, -0.0069, -0.0087, -0.0173, -0.0346, -0.0329, -0.0363, -0.0536, -0.0623, -0.0536, -0.0675, -0.0329, -0.0294, -0.0277, -0.0173, -0.0173, 0.0, 0.0225, 0.0225, 0.0346, 0.0242, 0.0363, 0.0294, 0.0087, -0.0104, 0.0052, 0.0346, 0.0242, 0.0294, 0.0035, 0.0467, 0.0606, 0.0415, 0.0584, 0.0409, 0.0479, 0.0584, 0.1106, 0.1019, 0.074, 0.0688, 0.0514, 0.0618, 0.0897, 0.0879, 0.1088, 0.1262, 0.1611, 0.168, 0.1802, 0.1489, 0.1558, 0.1489, 0.1471, 0.2255, 0.2289, 0.2638, 0.2429, 0.1994, 0.2133, 0.2168, 0.1976, 0.2272, 0.2498, 0.2359, 0.2376, 0.2516, 0.2464, 0.2464, 0.2185, 0.2185, 0.2272, 0.2289, 0.2429, 0.2446, 0.2237, 0.2289, 0.2115, 0.2133, 0.1767, 0.2028, 0.215, 0.2202, 0.2098, 0.2202, 0.2446, 0.2638, 0.2777, 0.3125, 0.3316, 0.3821, 0.3612, 0.3978, 0.3978, 0.4535, 0.4744, 0.4866, 0.4988, 0.45, 0.4722, 0.467, 0.5037, 0.5693, 0.5693, 0.5693, 0.5693, 0.5693, 0.5693, 0.6506, 0.6313, 0.6016, 0.6611, 0.7083, 0.7118, 0.7153, 0.7048, 0.724, 0.6873, 0.7275, 0.7835, 0.7397, 0.7572, 0.8202, 0.8796, 0.9426, 0.8342, 0.759, 0.7345, 0.7118, 0.759, 0.8097, 0.8027, 0.7974, 0.6995, 0.759, 0.71, 0.6873, 0.759, 0.6576, 0.6908, 0.7362, 0.7974, 0.8097]}
{"unit_id": "orig:1930224984196296865", "ticker_idx": 1, "ts": "2025-06-04T11:27:52+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I remain positive on companies like Samsung Electronics and SK Hynix, which are broadly involved across the memory value chain", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung and SK Hynix on broad memory value chain exposure; avoids SSD-pure-plays Kioxia and Western Digital and HDD-only companies.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Personally, I don’t invest in companies that only make SSDs—such as Kioxia or Western Digital.\nThe same goes for companies that focus only on HDDs.\nHowever, I remain positive on companies like Samsung Electronics and SK Hynix, which are broadly involved across the memory value chain.\nAs of now, SSDs are still more expensive than HDDs. In fact, over 90% of storage infrastructure in data centers is still based on HDDs.\nSSDs are used in specific workloads that require high throughput, but HDDs remain dominant due to their cost advantage.\nSome market research firms forecast that SSDs will become cheaper than HDDs by around 2029.\nI also believe that only around then will SSD-focused companies begin to generate meaningful profits that satisfy investors.\n\nStill, despite this outlook, I don’t take long positions on SSDs, nor do I go short on HDDs just based on that expectation.\nOn the surface, SSDs appear to have the long-term edge, and HDDs may look like a declining asset. But I have my reasons for staying cautious.\n\nHere's why:\n\n1. “SSDs will be cheaper by 2029” is just a forecast, not a fact.\n\nBack in 2014–2015, major players like Samsung and SK Hynix boldly claimed SSDs would surpass HDDs in price competitiveness within a few years.\nBut what happened?\n SSD prices dropped—but so did HDD prices.\nEven now, in 2025, the price gap remains substantial.\nThis is exactly why I can’t fully trust optimistic projections about SSDs.\n\n2. Will the AI datacenter boom still be alive in 2029?\n\nToday, HDDs, HBM, and GPUs are flying off the shelves. But that’s not because of intrinsic demand—it’s because hyperscalers are throwing massive CAPEX into infrastructure.\nWill CSPs still be investing at that same pace in 2029?\nAccording to my estimates, about 60–70% of HDD companies' profits currently come from AI-related demand.\nNow suppose SSDs truly replace HDDs by 2029.\n Can SSD makers really expect 60–70% of their earnings to come from AI infrastructure as well?\n\nI seriously doubt it.\n\nConclusion\nI don’t have a crystal ball, but I do know how to invest in what I can reasonably predict over the next 1–2 years.\nFor me, betting on outcomes 4–5 years into the future is simply too risky.\nSure, some people might make great returns by betting on SSDs.\nBut I believe there are plenty of other companies in the market that offer much higher conviction with far less uncertainty.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04597709938807859, "ret_signed_m1d": -0.04597709938807859, "alpha_spy_m1d": -0.0462457189281249, "alpha_c_m1d": -0.03421939259841866, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032183782009547635, "ret_signed_p1d": 0.032183782009547635, "alpha_spy_p1d": 0.03701655106573687, "alpha_c_p1d": 0.033971286622707875, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.1034481489964525, "ret_signed_p1w": 0.1034481489964525, "alpha_spy_p1w": 0.09433621783432744, "alpha_c_p1w": 0.06320983936782176, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.2804596714416854, "ret_signed_p1m": 0.2804596714416854, "alpha_spy_p1m": 0.2280059308757425, "alpha_c_p1m": 0.15386579796174837, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.17871148439176676, "ret_signed_p3m": 0.17871148439176676, "alpha_spy_p3m": 0.09308566509181015, "alpha_c_p3m": 0.025623022977242726, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 1.506555764559058, "ret_signed_p6m": 1.506555764559058, "alpha_spy_p6m": 1.3594758443928268, "alpha_c_p6m": 1.125384024885393, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.117, -0.1376, -0.1381, -0.0872, -0.0835, -0.0615, -0.0546, -0.0683, -0.0569, -0.0362, -0.011, -0.0293, -0.0454, -0.0179, -0.05, -0.0853, -0.1248, -0.0959, -0.0918, -0.1069, -0.1638, -0.2437, -0.2221, -0.2427, -0.1592, -0.1702, -0.173, -0.1712, -0.2014, -0.1969, -0.1969, -0.1895, -0.2024, -0.1693, -0.1817, -0.1537, -0.1647, -0.1702, -0.1854, -0.1854, -0.1464, -0.1464, -0.1464, -0.1243, -0.1266, -0.1276, -0.1051, -0.089, -0.0546, -0.0798, -0.0615, -0.0849, -0.0729, -0.0798, -0.0963, -0.0821, -0.0683, -0.0706, -0.0454, -0.0253, -0.0598, -0.046, -0.046, 0.0, 0.0322, 0.0322, 0.0529, 0.0598, 0.1034, 0.0828, 0.0828, 0.1402, 0.1448, 0.1333, 0.131, 0.1816, 0.1931, 0.2805, 0.3149, 0.3471, 0.3057, 0.3425, 0.3126, 0.2828, 0.2805, 0.2437, 0.246, 0.2966, 0.292, 0.3655, 0.354, 0.3793, 0.3724, 0.3609, 0.2391, 0.2368, 0.2529, 0.2345, 0.2368, 0.2391, 0.223, 0.2046, 0.2069, 0.2115, 0.2575, 0.1862, 0.1862, 0.2115, 0.1885, 0.2046, 0.1793, 0.2276, 0.2368, 0.2782, 0.2713, 0.2713, 0.2299, 0.2092, 0.1747, 0.1264, 0.154, 0.1931, 0.2023, 0.1954, 0.2363, 0.2386, 0.1787, 0.1994, 0.2086, 0.2225, 0.2593, 0.2754, 0.3261, 0.3997, 0.4135, 0.5125, 0.524, 0.6023, 0.5355, 0.6368, 0.6368, 0.6161, 0.6622, 0.6461, 0.6414, 0.5494, 0.6069, 0.6, 0.6576, 0.821, 0.821, 0.821, 0.821, 0.821, 0.821, 0.9707, 0.9108, 0.8947, 0.9453, 1.0835, 1.1433, 1.2354, 1.2055, 1.217, 1.2032, 1.3482, 1.4633, 1.3989, 1.5692, 1.6153, 1.5738, 1.8547, 1.6981, 1.6659, 1.7304, 1.6705, 1.7902, 1.8501, 1.8409, 1.8179, 1.5784, 1.7902, 1.6245, 1.5876, 1.6291, 1.3989, 1.3943, 1.3897, 1.4127, 1.5066]}
{"unit_id": "orig:1930224984196296865", "ticker_idx": 2, "ts": "2025-06-04T11:27:52+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I don't invest in companies that only make SSDs—such as Kioxia or Western Digital", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung and SK Hynix on broad memory value chain exposure; avoids SSD-pure-plays Kioxia and Western Digital and HDD-only companies.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Personally, I don’t invest in companies that only make SSDs—such as Kioxia or Western Digital.\nThe same goes for companies that focus only on HDDs.\nHowever, I remain positive on companies like Samsung Electronics and SK Hynix, which are broadly involved across the memory value chain.\nAs of now, SSDs are still more expensive than HDDs. In fact, over 90% of storage infrastructure in data centers is still based on HDDs.\nSSDs are used in specific workloads that require high throughput, but HDDs remain dominant due to their cost advantage.\nSome market research firms forecast that SSDs will become cheaper than HDDs by around 2029.\nI also believe that only around then will SSD-focused companies begin to generate meaningful profits that satisfy investors.\n\nStill, despite this outlook, I don’t take long positions on SSDs, nor do I go short on HDDs just based on that expectation.\nOn the surface, SSDs appear to have the long-term edge, and HDDs may look like a declining asset. But I have my reasons for staying cautious.\n\nHere's why:\n\n1. “SSDs will be cheaper by 2029” is just a forecast, not a fact.\n\nBack in 2014–2015, major players like Samsung and SK Hynix boldly claimed SSDs would surpass HDDs in price competitiveness within a few years.\nBut what happened?\n SSD prices dropped—but so did HDD prices.\nEven now, in 2025, the price gap remains substantial.\nThis is exactly why I can’t fully trust optimistic projections about SSDs.\n\n2. Will the AI datacenter boom still be alive in 2029?\n\nToday, HDDs, HBM, and GPUs are flying off the shelves. But that’s not because of intrinsic demand—it’s because hyperscalers are throwing massive CAPEX into infrastructure.\nWill CSPs still be investing at that same pace in 2029?\nAccording to my estimates, about 60–70% of HDD companies' profits currently come from AI-related demand.\nNow suppose SSDs truly replace HDDs by 2029.\n Can SSD makers really expect 60–70% of their earnings to come from AI infrastructure as well?\n\nI seriously doubt it.\n\nConclusion\nI don’t have a crystal ball, but I do know how to invest in what I can reasonably predict over the next 1–2 years.\nFor me, betting on outcomes 4–5 years into the future is simply too risky.\nSure, some people might make great returns by betting on SSDs.\nBut I believe there are plenty of other companies in the market that offer much higher conviction with far less uncertainty.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03786407766990296, "ret_signed_m1d": 0.03786407766990296, "alpha_spy_m1d": 0.03813269720994927, "alpha_c_m1d": 0.026106370880243035, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027184466019417375, "ret_signed_p1d": -0.027184466019417375, "alpha_spy_p1d": -0.03201723507560661, "alpha_c_p1d": -0.028971970632577615, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.06359223300970873, "ret_signed_p1w": -0.06359223300970873, "alpha_spy_p1w": -0.05448030184758368, "alpha_c_p1w": -0.023353923381078, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.23689320388349522, "ret_signed_p1m": -0.23689320388349522, "alpha_spy_p1m": -0.18443946331755234, "alpha_c_p1m": -0.11029933040355822, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.29902912621359223, "ret_signed_p3m": -0.29902912621359223, "alpha_spy_p3m": -0.21340330691363563, "alpha_c_p3m": -0.1459406647990682, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 3.3927184466019416, "ret_signed_p6m": -3.3927184466019416, "alpha_spy_p6m": -3.2456385264357106, "alpha_c_p6m": -3.011546706928277, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.3544, 0.3369, 0.2854, 0.3199, 0.3316, 0.4369, 0.517, 0.5388, 0.4709, 0.4709, 0.4417, 0.3563, 0.299, 0.2913, 0.2325, 0.2447, 0.1592, 0.0947, 0.1243, 0.0024, -0.0859, -0.2631, -0.1262, -0.1845, -0.001, -0.0379, -0.0553, -0.0563, -0.1262, -0.1097, -0.135, -0.1243, -0.1437, -0.1068, -0.1199, -0.0922, -0.0845, -0.0845, -0.0995, -0.1102, -0.1141, -0.1141, -0.1141, -0.0684, -0.0617, -0.0068, 0.0782, 0.0714, 0.1019, 0.0752, 0.0782, 0.065, 0.0612, 0.0549, 0.0073, -0.001, 0.0083, 0.0058, 0.0184, 0.0583, 0.0189, -0.0316, -0.0379, 0.0, 0.0272, 0.051, 0.0252, 0.035, 0.0636, 0.0136, -0.0243, -0.0248, 0.0117, -0.0068, 0.0209, 0.1087, 0.2398, 0.2374, 0.2282, 0.2282, 0.2248, 0.215, 0.1471, 0.1709, 0.2369, 0.1553, 0.1942, 0.2398, 0.3146, 0.2413, 0.2078, 0.183, 0.1893, 0.1903, 0.1718, 0.1422, 0.1422, 0.1641, 0.2306, 0.2456, 0.1927, 0.1913, 0.1786, 0.1903, 0.2029, 0.1767, 0.1563, 0.1505, 0.1325, 0.1165, 0.1476, 0.1476, 0.2709, 0.2515, 0.2403, 0.2427, 0.1854, 0.2005, 0.1447, 0.1189, 0.1665, 0.1612, 0.1811, 0.1723, 0.1995, 0.2602, 0.299, 0.2699, 0.2733, 0.2748, 0.4879, 0.5291, 0.5413, 0.6917, 0.9442, 1.1553, 1.1553, 1.284, 1.1553, 1.1942, 1.199, 1.3398, 1.3398, 1.4029, 1.3738, 1.1335, 1.2694, 1.3665, 1.284, 1.6262, 2.0, 2.068, 1.8641, 1.8544, 2.0194, 1.9903, 1.9903, 1.9126, 2.068, 2.3398, 2.1845, 2.4709, 2.4175, 2.5485, 2.5777, 3.2621, 3.7621, 3.5534, 3.8932, 4.2913, 4.2549, 4.2549, 4.2233, 4.1189, 4.5825, 4.8447, 5.466, 5.4053, 5.4272, 5.3228, 3.8665, 4.4199, 3.9806, 4.2597, 4.5024, 3.8689, 3.8689, 3.783, 3.0709, 3.3927]}
{"unit_id": "orig:1930224984196296865", "ticker_idx": 3, "ts": "2025-06-04T11:27:52+00:00", "symbol": "WDC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I don't invest in companies that only make SSDs—such as Kioxia or Western Digital", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung and SK Hynix on broad memory value chain exposure; avoids SSD-pure-plays Kioxia and Western Digital and HDD-only companies.", "resolved_tickers": ["WDC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Personally, I don’t invest in companies that only make SSDs—such as Kioxia or Western Digital.\nThe same goes for companies that focus only on HDDs.\nHowever, I remain positive on companies like Samsung Electronics and SK Hynix, which are broadly involved across the memory value chain.\nAs of now, SSDs are still more expensive than HDDs. In fact, over 90% of storage infrastructure in data centers is still based on HDDs.\nSSDs are used in specific workloads that require high throughput, but HDDs remain dominant due to their cost advantage.\nSome market research firms forecast that SSDs will become cheaper than HDDs by around 2029.\nI also believe that only around then will SSD-focused companies begin to generate meaningful profits that satisfy investors.\n\nStill, despite this outlook, I don’t take long positions on SSDs, nor do I go short on HDDs just based on that expectation.\nOn the surface, SSDs appear to have the long-term edge, and HDDs may look like a declining asset. But I have my reasons for staying cautious.\n\nHere's why:\n\n1. “SSDs will be cheaper by 2029” is just a forecast, not a fact.\n\nBack in 2014–2015, major players like Samsung and SK Hynix boldly claimed SSDs would surpass HDDs in price competitiveness within a few years.\nBut what happened?\n SSD prices dropped—but so did HDD prices.\nEven now, in 2025, the price gap remains substantial.\nThis is exactly why I can’t fully trust optimistic projections about SSDs.\n\n2. Will the AI datacenter boom still be alive in 2029?\n\nToday, HDDs, HBM, and GPUs are flying off the shelves. But that’s not because of intrinsic demand—it’s because hyperscalers are throwing massive CAPEX into infrastructure.\nWill CSPs still be investing at that same pace in 2029?\nAccording to my estimates, about 60–70% of HDD companies' profits currently come from AI-related demand.\nNow suppose SSDs truly replace HDDs by 2029.\n Can SSD makers really expect 60–70% of their earnings to come from AI infrastructure as well?\n\nI seriously doubt it.\n\nConclusion\nI don’t have a crystal ball, but I do know how to invest in what I can reasonably predict over the next 1–2 years.\nFor me, betting on outcomes 4–5 years into the future is simply too risky.\nSure, some people might make great returns by betting on SSDs.\nBut I believe there are plenty of other companies in the market that offer much higher conviction with far less uncertainty.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013962903666833282, "ret_signed_m1d": 0.013962903666833282, "alpha_spy_m1d": 0.014231523206879593, "alpha_c_m1d": 0.012484750509972775, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.0014781531568605066, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011390758403118761, "ret_signed_p1d": -0.011390758403118761, "alpha_spy_p1d": -0.016223527459307996, "alpha_c_p1d": -0.014220239966610704, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0028294815634919424, "ret_p1w": 0.022781587062315145, "ret_signed_p1w": -0.022781587062315145, "alpha_spy_p1w": -0.013669655900190092, "alpha_c_p1w": -0.007832403153984702, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.014949183908330443, "ret_p1m": 0.21403640557540982, "ret_signed_p1m": -0.21403640557540982, "alpha_spy_p1m": -0.16158266500946694, "alpha_c_p1m": -0.12710286731781983, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.08693353825758998, "ret_p3m": 0.4760242703836066, "ret_signed_p3m": -0.4760242703836066, "alpha_spy_p3m": -0.39039845108365, "alpha_c_p3m": -0.36586806402619354, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.11015620635741308, "ret_p6m": 1.9014082409255142, "ret_signed_p6m": -1.9014082409255142, "alpha_spy_p6m": -1.754328320759283, "alpha_c_p6m": -1.699484185685656, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.20192405523985824, "price_path": [-0.2353, -0.2566, -0.2335, -0.2122, -0.2293, -0.1834, -0.1794, -0.1869, -0.1766, -0.1838, -0.1994, -0.1874, -0.2043, -0.2184, -0.2315, -0.2551, -0.2586, -0.2502, -0.2338, -0.3738, -0.44, -0.4286, -0.4214, -0.3343, -0.3606, -0.3692, -0.3431, -0.3411, -0.3468, -0.3305, -0.3305, -0.3406, -0.3274, -0.3087, -0.2634, -0.2522, -0.2498, -0.2551, -0.1957, -0.194, -0.1805, -0.1742, -0.1878, -0.1876, -0.1876, -0.1913, -0.146, -0.1013, -0.0978, -0.0877, -0.0833, -0.0699, -0.0715, -0.0924, -0.086, -0.0798, -0.0798, -0.0497, -0.04, -0.0363, -0.0547, -0.0429, -0.014, 0.0, 0.0114, 0.0187, 0.0476, 0.0279, 0.0228, 0.0248, 0.0233, 0.0547, 0.0761, 0.0875, 0.0875, 0.0893, 0.1093, 0.1404, 0.1492, 0.1668, 0.1628, 0.1756, 0.1729, 0.2085, 0.214, 0.214, 0.1982, 0.1762, 0.1876, 0.1953, 0.2151, 0.2297, 0.2407, 0.2223, 0.2313, 0.2493, 0.2629, 0.232, 0.2736, 0.2681, 0.2644, 0.2675, 0.2973, 0.3123, 0.4457, 0.4064, 0.42, 0.3933, 0.3555, 0.3676, 0.3774, 0.3713, 0.3946, 0.3976, 0.4007, 0.379, 0.4016, 0.3937, 0.3897, 0.3717, 0.4141, 0.4554, 0.4644, 0.4821, 0.5073, 0.476, 0.476, 0.5049, 0.58, 0.6644, 0.6929, 0.7159, 0.7389, 0.7478, 0.7685, 0.7963, 0.8833, 0.8962, 0.8567, 0.9341, 0.9613, 1.0676, 1.0279, 1.0176, 0.972, 0.9659, 1.1473, 1.2083, 1.402, 1.4143, 1.4153, 1.3044, 1.2059, 1.2289, 1.2017, 1.123, 1.1863, 1.0809, 1.2153, 1.3161, 1.3213, 1.2354, 1.2332, 1.2159, 1.3124, 1.3807, 1.3299, 1.2977, 1.6005, 1.5407, 1.7629, 1.9066, 1.7991, 1.9448, 2.0092, 1.9974, 2.2045, 2.1267, 2.0554, 1.8907, 1.9031, 1.988, 1.8116, 1.8321, 1.5793, 1.5602, 1.7761, 1.8586, 1.9014, 1.9014]}
{"unit_id": "orig:1929819499160916064", "ticker_idx": 0, "ts": "2025-06-03T08:36:37+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs added TSMC to its APAC Conviction List... we view the current valuation as a compelling buying opportunity... TSMC's stock is likely to continue outperforming", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman's Conviction Buy on TSMC: 26%/14% revenue growth FY25/26, trades below 5-yr avg P/E at 16.9x, compelling buying opportunity on AI/HPC demand + ASP expansion.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs added TSMC to its APAC Conviction List with the following commentary:\n\nTSMC- We project TSMC’s revenue to grow by 26% and 14% in FY2025 and FY2026, respectively, driven by accelerating demand for AI and high-performance computing (HPC), as well as rising ASP (average selling prices) in advanced nodes. TSMC is currently trading at 16.9x FY2026 P/E, which is below its 5-year historical average of 18.8x. Given TSMC’s year-to-date share price decline of 10%, in contrast to the 8% gain in the MXAP (Asia Pacific Index), we view the current valuation as a compelling buying opportunity. Bruce (analyst) believes that if the market gains greater confidence in the outlook for AI demand in H2 2025 and 2026, as well as ASP expansion from AI clients migrating to advanced nodes, TSMC’s stock is likely to continue outperforming.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004210495840462847, "ret_signed_m1d": -0.004210495840462847, "alpha_spy_m1d": 0.001459758811450329, "alpha_c_m1d": 0.017856271826206416, "bench_spy_m1d": -0.005670254651913176, "bench_c_m1d": -0.022066767666669262, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04210528443571282, "ret_signed_p1d": 0.04210528443571282, "alpha_spy_p1d": 0.042373831838679266, "alpha_c_p1d": 0.03020768920988548, "bench_spy_p1d": -0.00026854740296644586, "bench_c_p1d": 0.01189759522582734, "ret_p1w": 0.10000005868559514, "ret_signed_p1w": 0.10000005868559514, "alpha_spy_p1w": 0.08827376493284866, "alpha_c_p1w": 0.046621670122691805, "bench_spy_p1w": 0.011726293752746475, "bench_c_p1w": 0.053378388562903334, "ret_p1m": 0.14695153437075126, "ret_signed_p1m": 0.14695153437075126, "alpha_spy_p1m": 0.10300814826452176, "alpha_c_p1m": 0.016480100196349268, "bench_spy_p1m": 0.0439433861062295, "bench_c_p1m": 0.130471434174402, "ret_p3m": 0.22623394777803618, "ret_signed_p3m": 0.22623394777803618, "alpha_spy_p3m": 0.1408996704724459, "alpha_c_p3m": 0.059426506590030126, "bench_spy_p3m": 0.08533427730559029, "bench_c_p3m": 0.16680744118800606, "ret_p6m": 0.5283104854009475, "ret_signed_p6m": 0.5283104854009475, "alpha_spy_p6m": 0.3815386105682719, "alpha_c_p6m": 0.13070612343129362, "bench_spy_p6m": 0.14677187483267562, "bench_c_p6m": 0.3976043619696539, "price_path": [0.053, 0.053, 0.0457, 0.0174, 0.0352, 0.0111, 0.0048, 0.0163, 0.0221, 0.0021, 0.0368, 0.0232, 0.0232, 0.0421, 0.0316, 0.0084, 0.0021, -0.0421, -0.0063, -0.0084, -0.0084, -0.0084, -0.1074, -0.1411, -0.1737, -0.0916, -0.0642, -0.0895, -0.0768, -0.1, -0.1084, -0.1053, -0.1211, -0.1411, -0.0811, -0.0905, -0.0653, -0.0558, -0.0505, -0.0442, -0.0442, 0.0, -0.0126, -0.0316, -0.0232, -0.0337, -0.0011, 0.0074, 0.02, 0.0516, 0.0453, 0.0505, 0.0358, 0.0295, 0.0421, 0.0337, 0.0358, 0.0274, 0.0158, 0.0179, 0.0179, 0.0179, -0.0042, 0.0, 0.0421, 0.0505, 0.0474, 0.0579, 0.1, 0.1211, 0.1047, 0.0888, 0.0835, 0.1047, 0.1152, 0.0941, 0.1152, 0.0782, 0.11, 0.1311, 0.1364, 0.1417, 0.1205, 0.147, 0.147, 0.1522, 0.147, 0.1417, 0.1417, 0.1522, 0.1628, 0.1628, 0.1575, 0.1734, 0.1945, 0.1945, 0.2209, 0.2157, 0.1945, 0.2104, 0.2104, 0.2104, 0.2104, 0.1998, 0.2209, 0.2262, 0.2262, 0.1998, 0.2157, 0.1892, 0.2474, 0.2421, 0.2474, 0.2474, 0.2685, 0.2421, 0.2474, 0.2474, 0.2527, 0.1998, 0.2157, 0.1998, 0.2368, 0.2421, 0.2579, 0.2262, 0.2262, 0.2315, 0.2262, 0.2262, 0.2262, 0.2474, 0.2474, 0.2685, 0.2949, 0.3108, 0.3319, 0.3267, 0.3585, 0.3426, 0.3638, 0.3426, 0.3744, 0.4222, 0.4222, 0.401, 0.3797, 0.3797, 0.385, 0.4063, 0.4487, 0.4859, 0.4859, 0.523, 0.5018, 0.5283, 0.5283, 0.5018, 0.5124, 0.5548, 0.5761, 0.5389, 0.5708, 0.5708, 0.5495, 0.5389, 0.5389, 0.5708, 0.5655, 0.5973, 0.5973, 0.592, 0.6026, 0.5973, 0.5495, 0.5548, 0.5495, 0.5655, 0.5548, 0.5655, 0.5495, 0.5177, 0.5336, 0.4912, 0.4806, 0.5442, 0.4699, 0.4593, 0.5018, 0.5283]}
{"unit_id": "thread:1931520720049422343", "ticker_idx": 0, "ts": "2025-06-08T01:16:40+00:00", "symbol": "BYD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the stock price could still underperform, right? Especially if they keep clearing out inventory through heavy discounts like this", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish BYD: heavy discounting to clear inventory and rising protectionism risks will weigh on the stock even without bankruptcy risk.", "resolved_tickers": ["1211.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "BYD Chinese EV context; HK listing 1211.HK", "bench_c": "XLY", "bench_c_reason": "industry='Auto Manufacturers'", "bench_c_industry": "Auto Manufacturers", "unit_kind": "self_thread", "lang": "en", "tweet_text": "@JayGatsby0 Of course, I don’t think they’ll go bankrupt either.\nBut the stock price could still underperform, right?\nEspecially if they keep clearing out inventory through heavy discounts like this.\n\n---\n\n@JayGatsby0 There’s a growing discussion in Korea as well about whether BYD vehicles should be excluded from receiving subsidies.\nCountries are already starting to lean toward protectionism — especially those with their own domestic automakers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016137093820044246, "ret_signed_m1d": -0.016137093820044246, "alpha_spy_m1d": -0.01703756512880672, "alpha_c_m1d": -0.024085809906477462, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.007948716086433216, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03705658176393212, "ret_signed_p1d": -0.03705658176393212, "alpha_spy_p1d": -0.03138691709504138, "alpha_c_p1d": -0.025086730176831606, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01196985158710051, "ret_p1w": -0.007301316970118044, "ret_signed_p1w": 0.007301316970118044, "alpha_spy_p1w": 0.012303992544522413, "alpha_c_p1w": 0.008002655236861167, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.0007013382667431234, "ret_p1m": -0.06160152145974063, "ret_signed_p1m": 0.06160152145974063, "alpha_spy_p1m": 0.09911144805696803, "alpha_c_p1m": 0.08148643525151134, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.01988491379177071, "ret_p3m": -0.20079345194423848, "ret_signed_p3m": 0.20079345194423848, "alpha_spy_p3m": 0.2864373635660731, "alpha_c_p3m": 0.3026777744919582, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.10188432254771973, "ret_p6m": -0.23444426563742282, "ret_signed_p6m": 0.23444426563742282, "alpha_spy_p6m": 0.37745376902240024, "alpha_c_p6m": 0.34428032449173385, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.10983605885431103, "price_path": [-0.1049, -0.0903, -0.0272, -0.0277, 0.0121, 0.0514, 0.0696, -0.0126, 0.0171, -0.0187, 0.0045, 0.0272, 0.0257, -0.0096, -0.0262, -0.0383, -0.055, -0.055, -0.2052, -0.1669, -0.1548, -0.1321, -0.0701, -0.0519, -0.0535, -0.0827, -0.0772, -0.0772, -0.0772, -0.0524, -0.0156, -0.0156, 0.001, -0.0388, -0.0635, -0.0666, -0.0666, -0.0358, -0.0358, -0.0298, -0.0308, -0.0272, -0.0242, 0.0479, -0.002, 0.0454, 0.06, 0.0948, 0.0938, 0.121, 0.1664, 0.1503, 0.173, 0.0721, 0.0545, 0.0262, 0.0237, -0.0096, -0.0282, 0.0101, 0.0242, 0.0388, 0.0161, 0.0, 0.0371, 0.0768, 0.0279, 0.0026, -0.0073, -0.0134, -0.0249, -0.0478, -0.0379, -0.0356, -0.0042, -0.0065, -0.0387, -0.0501, -0.0631, -0.0631, -0.0647, -0.0593, -0.0708, -0.0708, -0.0616, -0.0723, -0.0861, -0.0815, -0.0792, -0.0562, -0.0631, -0.0532, -0.0333, -0.0234, 0.0263, 0.0202, 0.0103, -0.0088, -0.0203, -0.0188, -0.0754, -0.1174, -0.1113, -0.1243, -0.1427, -0.1465, -0.1373, -0.1496, -0.1473, -0.1488, -0.1205, -0.1297, -0.1373, -0.1304, -0.1281, -0.1327, -0.148, -0.1258, -0.1144, -0.1006, -0.1197, -0.1434, -0.1251, -0.171, -0.1526, -0.174, -0.2008, -0.1924, -0.1924, -0.1901, -0.1924, -0.1954, -0.2008, -0.1733, -0.1595, -0.1373, -0.1373, -0.132, -0.161, -0.187, -0.1916, -0.1725, -0.1855, -0.1694, -0.1572, -0.1572, -0.1289, -0.1633, -0.1733, -0.1733, -0.161, -0.1572, -0.1725, -0.187, -0.1939, -0.1694, -0.1771, -0.2107, -0.1985, -0.2054, -0.213, -0.2054, -0.2046, -0.1916, -0.1985, -0.1985, -0.2031, -0.2306, -0.2421, -0.2574, -0.2704, -0.2532, -0.2513, -0.2214, -0.2253, -0.2314, -0.2138, -0.2306, -0.2299, -0.2589, -0.2654, -0.2719, -0.2907, -0.2746, -0.2677, -0.2478, -0.2582, -0.2543, -0.2509, -0.2344]}
{"unit_id": "quote:1927665581802578242", "ticker_idx": 0, "ts": "2025-05-28T09:57:43+00:00", "symbol": "data center", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TD Cowen has turned bullish on the data center sector", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses TD Cowen's bullish turn on data centers, citing OpenAI/Stargate activity and MSFT expected to resume leasing in Q4 2025.", "resolved_tickers": ["DTCR"], "resolution_method": "etf", "resolution_reason": "US data center sector proxied by DTCR ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "TD Cowen has turned bullish on the data center sector, citing the following reasons:\n\n1. OpenAI remains highly active, primarily due to the Stargate projects, and has recently started leasing data center space in colocation (colo) facilities.\n\n2. Amazon has been consistently building its own data centers, and colo contracts have slowed down mainly because of their higher operating costs.\n\n3. Microsoft (MSFT) is expected to resume data center leasing in Q4 2025.", "reply_text": null, "parent_text": null, "quoted_text": "TD Cowen翻多Data center，1） Open AI持續活躍，主要因爲stargate projects，同時近期有向託管(colo)租賃數據中心；2）amazon最近持續自行建置數據中心，colo合約放緩主要因爲其較高的運營費用；3）MSFT預計4Q25會重新開始租賃數據中心\n\n #零級分", "parent_author": null, "ret_m1d": 0.01155394346785088, "ret_signed_m1d": 0.01155394346785088, "alpha_spy_m1d": 0.005734863199542417, "alpha_c_m1d": 0.005734863199542417, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.005819080268308463, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005199213245111967, "ret_signed_p1d": 0.005199213245111967, "alpha_spy_p1d": 0.0012519359043157152, "alpha_c_p1d": 0.0012519359043157152, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.003947277340796251, "ret_p1w": 0.03986126661383138, "ret_signed_p1w": 0.03986126661383138, "alpha_spy_p1w": 0.02590940220042892, "alpha_c_p1w": 0.02590940220042892, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.01395186441340246, "ret_p1m": 0.06412478200974325, "ret_signed_p1m": 0.06412478200974325, "alpha_spy_p1m": 0.01997386679579849, "alpha_c_p1m": 0.01997386679579849, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.044150915213944764, "ret_p3m": 0.09763455139575639, "ret_signed_p3m": 0.09763455139575639, "alpha_spy_p3m": 0.0012650612792854954, "alpha_c_p3m": 0.0012650612792854954, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.09636949011647089, "ret_p6m": 0.14757986361443698, "ret_signed_p6m": 0.14757986361443698, "alpha_spy_p6m": 0.030955722167478772, "alpha_c_p6m": 0.030955722167478772, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.11662414144695821, "price_path": [0.0202, -0.004, 0.0058, 0.0347, 0.0023, 0.0069, -0.0121, -0.0029, 0.004, -0.011, 0.026, 0.0381, 0.022, 0.0075, 0.0006, -0.0012, 0.0052, -0.0156, -0.0254, -0.0399, -0.0595, -0.0607, -0.0491, -0.041, -0.063, -0.1242, -0.1363, -0.1664, -0.1121, -0.1254, -0.1109, -0.0982, -0.0976, -0.1138, -0.0994, -0.0994, -0.1144, -0.0959, -0.0878, -0.0786, -0.0687, -0.0635, -0.0503, -0.0387, -0.0318, -0.0092, -0.0075, -0.0069, -0.022, -0.0237, -0.0208, -0.0064, -0.0104, -0.0075, -0.0035, 0.015, 0.0173, 0.015, 0.0087, 0.0012, 0.0, 0.0, 0.0116, 0.0, 0.0052, -0.0052, 0.0133, 0.0185, 0.0399, 0.041, 0.0427, 0.0456, 0.0543, 0.052, 0.0572, 0.0347, 0.0566, 0.0474, 0.0508, 0.0508, 0.0479, 0.0583, 0.0838, 0.0687, 0.0641, 0.0744, 0.0825, 0.0721, 0.0849, 0.0918, 0.0918, 0.0692, 0.0715, 0.0756, 0.075, 0.0761, 0.0843, 0.0994, 0.1098, 0.118, 0.1217, 0.1203, 0.1238, 0.1342, 0.1342, 0.1325, 0.1243, 0.118, 0.1081, 0.1069, 0.0883, 0.1063, 0.1023, 0.093, 0.0988, 0.0912, 0.093, 0.1034, 0.1005, 0.0889, 0.0878, 0.0907, 0.086, 0.0878, 0.0796, 0.1052, 0.0976, 0.0994, 0.1, 0.104, 0.0982, 0.0982, 0.0761, 0.0645, 0.064, 0.0773, 0.0756, 0.0924, 0.1133, 0.1406, 0.1453, 0.1522, 0.1528, 0.1621, 0.1749, 0.172, 0.1987, 0.1935, 0.1894, 0.19, 0.183, 0.1795, 0.1894, 0.2091, 0.2266, 0.2277, 0.2405, 0.2347, 0.2498, 0.2539, 0.2289, 0.255, 0.2481, 0.2771, 0.2794, 0.2643, 0.2916, 0.2672, 0.2498, 0.2748, 0.2939, 0.3154, 0.305, 0.3009, 0.2951, 0.2986, 0.2957, 0.2608, 0.2823, 0.262, 0.2626, 0.2887, 0.2777, 0.2597, 0.2068, 0.2109, 0.1993, 0.1865, 0.1807, 0.1476]}
{"unit_id": "orig:1922798390112551056", "ticker_idx": 0, "ts": "2025-05-14T23:37:14+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB200 rack bottlenecks have largely been resolved, and shipments between March and May are expected to increase by 3~4x", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Nomura's bullish Foxconn note: AI server rev to double YoY, GB200 ramp resolving, improved 2H25 visibility; positive read-through for semiconductor upcycle.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250515_Foxconn: Improved Visibility on AI Servers in 2025 – Nomura\n\n(1) 1Q25 Operating Profit was in line with market consensus, partially driven by Sharp’s contribution.\n\n(2) 2Q25(F) Guidance\n\nRevenue expected to grow YoY by +5~15%\nConsumer: Slight QoQ decline anticipated due to base effect and temporary gaps from new product launches\nCloud: Strong YoY and QoQ growth expected, driven by stabilized AI server production (estimated by Nomura to be GB200) and solid demand for general-purpose servers\nComputing: Expected to remain flat\n\n(3) AI server revenue is expected to more than double YoY, and in 2Q25(F), it is projected to make up about 50% of total server revenue mix\n\n(4) This growth became possible as mass shipments of GB200 racks ramped up in April, following small-scale shipments that began in March\n\n(5) Management mentioned that visibility for AI server orders in 2H25 has significantly improved, in line with recent comments from Quanta Computer executives\n\n(6) However, for the consumer and computing segments, the company maintains a conservative outlook due to ongoing macroeconomic uncertainties\n\n(7) Due to the strong TWD, full-year 2025 revenue growth guidance was widened to +5~15% from +15% previously (= downward revision)\n* According to Nomura estimates, a +1% TWD appreciation has an estimated -3% impact on revenue and -10bps on gross margin\n(8) FY25(F) CAPEX guidance remains at +20% YoY, with 60~70% allocated for capacity expansion in the U.S., India, Vietnam, and Mexico\n\n(9) In our view, as in April, GB200 rack bottlenecks have largely been resolved, and shipments between March and May are expected to increase by 3~4x\n\n(10) Thanks to improved GB200 productivity and computing tray redesigns for GB300, major CSPs are expected to resume large-scale orders\n\n(11) This trend aligns with continued aggressive annual CAPEX by global Big Tech firms and could signal a positive semiconductor upcycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030674809725492858, "ret_signed_m1d": -0.030674809725492858, "alpha_spy_m1d": -0.029398381547455665, "alpha_c_m1d": -0.02422533865513743, "bench_spy_m1d": -0.0012764281780371922, "bench_c_m1d": -0.006449471070355428, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021472347405442904, "ret_signed_p1d": -0.021472347405442904, "alpha_spy_p1d": -0.02635658422662923, "alpha_c_p1d": -0.022839020724900405, "bench_spy_p1d": 0.004884236821186327, "bench_c_p1d": 0.001366673319457501, "ret_p1w": -0.049079734365592875, "ret_signed_p1w": -0.049079734365592875, "alpha_spy_p1w": -0.041029905231931285, "alpha_c_p1w": -0.028364713379806616, "bench_spy_p1w": -0.00804982913366159, "bench_c_p1w": -0.02071502098578626, "ret_p1m": -0.03374232950284639, "ret_signed_p1m": -0.03374232950284639, "alpha_spy_p1m": -0.06124431889847981, "alpha_c_p1m": -0.06979062417625503, "bench_spy_p1m": 0.02750198939563342, "bench_c_p1m": 0.03604829467340864, "ret_p3m": 0.25884026797821735, "ret_signed_p3m": 0.25884026797821735, "alpha_spy_p3m": 0.17338976311194654, "alpha_c_p3m": 0.12887940801815367, "bench_spy_p3m": 0.08545050486627082, "bench_c_p3m": 0.12996085996006368, "ret_p6m": 0.5767292245383733, "ret_signed_p6m": 0.5767292245383733, "alpha_spy_p6m": 0.4294063616493904, "alpha_c_p6m": 0.3380520615374598, "bench_spy_p6m": 0.1473228628889829, "bench_c_p6m": 0.23867716300091346, "price_path": [0.092, 0.1196, 0.1258, 0.1227, 0.1166, 0.1227, 0.1288, 0.089, 0.1074, 0.0675, 0.0675, 0.0429, 0.046, 0.0644, 0.0644, 0.0552, 0.0552, 0.0337, 0.046, 0.0307, 0.0429, 0.0307, 0.0276, 0.0061, 0.0184, 0.0123, 0.0, 0.0123, 0.0184, -0.0153, -0.0552, -0.1043, -0.0675, -0.0583, -0.0583, -0.0583, -0.1503, -0.2331, -0.3098, -0.2423, -0.1748, -0.1503, -0.1442, -0.1656, -0.1748, -0.1687, -0.1626, -0.1902, -0.1472, -0.1626, -0.1472, -0.1258, -0.1227, -0.1319, -0.1319, -0.0951, -0.1258, -0.1043, -0.1104, -0.1043, -0.0982, -0.0613, -0.0307, 0.0, -0.0215, -0.0307, -0.0552, -0.0552, -0.0491, -0.0613, -0.0552, -0.0613, -0.0675, -0.0675, -0.0429, -0.0429, -0.0644, -0.0706, -0.0429, -0.046, -0.0613, -0.0521, -0.046, -0.0429, -0.0337, -0.0399, -0.0491, -0.0491, -0.0429, -0.0552, -0.046, -0.0429, -0.0153, -0.0061, 0.0, 0.0123, -0.0123, 0.0153, 0.03, 0.0427, 0.0236, 0.0236, 0.0109, 0.0236, 0.0268, 0.0268, 0.0204, 0.0363, 0.0331, 0.0427, 0.0522, 0.049, 0.03, 0.0554, 0.1063, 0.1094, 0.119, 0.0904, 0.0904, 0.1317, 0.1317, 0.1444, 0.173, 0.1825, 0.2366, 0.2366, 0.2588, 0.2525, 0.262, 0.2684, 0.3161, 0.3351, 0.3224, 0.2747, 0.3097, 0.2875, 0.3192, 0.3288, 0.3224, 0.3097, 0.2938, 0.262, 0.262, 0.2747, 0.2906, 0.3033, 0.2938, 0.3192, 0.3351, 0.3637, 0.3828, 0.3733, 0.3701, 0.3478, 0.3669, 0.3606, 0.3733, 0.4051, 0.4178, 0.4623, 0.3955, 0.3955, 0.3733, 0.3924, 0.4337, 0.44, 0.44, 0.4528, 0.4305, 0.4082, 0.4082, 0.3542, 0.3097, 0.3129, 0.4178, 0.44, 0.5163, 0.5195, 0.5418, 0.5195, 0.5195, 0.5672, 0.5894, 0.653, 0.6657, 0.6371, 0.599, 0.5513, 0.5704, 0.5767]}
{"unit_id": "orig:1922798390112551056", "ticker_idx": 1, "ts": "2025-05-14T23:37:14+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "could signal a positive semiconductor upcycle", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Nomura's bullish Foxconn note: AI server rev to double YoY, GB200 ramp resolving, improved 2H25 visibility; positive read-through for semiconductor upcycle.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "Global semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "250515_Foxconn: Improved Visibility on AI Servers in 2025 – Nomura\n\n(1) 1Q25 Operating Profit was in line with market consensus, partially driven by Sharp’s contribution.\n\n(2) 2Q25(F) Guidance\n\nRevenue expected to grow YoY by +5~15%\nConsumer: Slight QoQ decline anticipated due to base effect and temporary gaps from new product launches\nCloud: Strong YoY and QoQ growth expected, driven by stabilized AI server production (estimated by Nomura to be GB200) and solid demand for general-purpose servers\nComputing: Expected to remain flat\n\n(3) AI server revenue is expected to more than double YoY, and in 2Q25(F), it is projected to make up about 50% of total server revenue mix\n\n(4) This growth became possible as mass shipments of GB200 racks ramped up in April, following small-scale shipments that began in March\n\n(5) Management mentioned that visibility for AI server orders in 2H25 has significantly improved, in line with recent comments from Quanta Computer executives\n\n(6) However, for the consumer and computing segments, the company maintains a conservative outlook due to ongoing macroeconomic uncertainties\n\n(7) Due to the strong TWD, full-year 2025 revenue growth guidance was widened to +5~15% from +15% previously (= downward revision)\n* According to Nomura estimates, a +1% TWD appreciation has an estimated -3% impact on revenue and -10bps on gross margin\n(8) FY25(F) CAPEX guidance remains at +20% YoY, with 60~70% allocated for capacity expansion in the U.S., India, Vietnam, and Mexico\n\n(9) In our view, as in April, GB200 rack bottlenecks have largely been resolved, and shipments between March and May are expected to increase by 3~4x\n\n(10) Thanks to improved GB200 productivity and computing tray redesigns for GB300, major CSPs are expected to resume large-scale orders\n\n(11) This trend aligns with continued aggressive annual CAPEX by global Big Tech firms and could signal a positive semiconductor upcycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00979881382747494, "ret_signed_m1d": -0.00979881382747494, "alpha_spy_m1d": -0.008522385649437747, "alpha_c_m1d": -0.008522385649437747, "bench_spy_m1d": -0.0012764281780371922, "bench_c_m1d": -0.0012764281780371922, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003467873655732978, "ret_signed_p1d": -0.003467873655732978, "alpha_spy_p1d": -0.008352110476919306, "alpha_c_p1d": -0.008352110476919306, "bench_spy_p1d": 0.004884236821186327, "bench_c_p1d": 0.004884236821186327, "ret_p1w": -0.026291365700689018, "ret_signed_p1w": -0.026291365700689018, "alpha_spy_p1w": -0.018241536567027428, "alpha_c_p1w": -0.018241536567027428, "bench_spy_p1w": -0.00804982913366159, "bench_c_p1w": -0.00804982913366159, "ret_p1m": 0.06113144176850316, "ret_signed_p1m": 0.06113144176850316, "alpha_spy_p1m": 0.03362945237286974, "alpha_c_p1m": 0.03362945237286974, "bench_spy_p1m": 0.02750198939563342, "bench_c_p1m": 0.02750198939563342, "ret_p3m": 0.18311213546544258, "ret_signed_p3m": 0.18311213546544258, "alpha_spy_p3m": 0.09766163059917177, "alpha_c_p3m": 0.09766163059917177, "bench_spy_p3m": 0.08545050486627082, "bench_c_p3m": 0.08545050486627082, "ret_p6m": 0.4157426008953664, "ret_signed_p6m": 0.4157426008953664, "alpha_spy_p6m": 0.2684197380063835, "alpha_c_p6m": 0.2684197380063835, "bench_spy_p6m": 0.1473228628889829, "bench_c_p6m": 0.1473228628889829, "price_path": [0.0185, 0.0185, 0.0332, 0.0377, 0.0396, 0.0081, -0.0184, -0.0393, -0.0168, -0.0774, -0.0614, -0.1007, -0.0918, -0.075, -0.1138, -0.0923, -0.1349, -0.1362, -0.1108, -0.1147, -0.0864, -0.0767, -0.091, -0.083, -0.0844, -0.0936, -0.07, -0.074, -0.1048, -0.1219, -0.1444, -0.1473, -0.1432, -0.1367, -0.2114, -0.2709, -0.2544, -0.2744, -0.1498, -0.2088, -0.1882, -0.187, -0.1817, -0.2163, -0.2236, -0.2236, -0.2426, -0.2262, -0.1979, -0.1573, -0.1452, -0.1506, -0.152, -0.148, -0.1439, -0.1168, -0.1225, -0.1307, -0.1128, -0.105, -0.0992, -0.0427, -0.0098, 0.0, -0.0035, -0.0063, -0.008, -0.0099, -0.0263, -0.029, -0.0425, -0.0425, -0.0122, -0.0227, -0.0158, -0.0332, -0.0189, 0.0032, 0.0152, 0.0133, 0.0191, 0.0362, 0.0568, 0.056, 0.0611, 0.0363, 0.0621, 0.0548, 0.0589, 0.0589, 0.0495, 0.0562, 0.095, 0.1097, 0.1178, 0.1227, 0.1246, 0.1123, 0.1341, 0.1437, 0.1437, 0.1281, 0.1431, 0.1509, 0.1594, 0.1593, 0.1505, 0.1726, 0.1665, 0.1763, 0.1707, 0.1715, 0.1507, 0.1557, 0.1605, 0.1593, 0.1751, 0.1811, 0.1946, 0.1645, 0.145, 0.1702, 0.1577, 0.1558, 0.1739, 0.1836, 0.1831, 0.2101, 0.2149, 0.2175, 0.1922, 0.1971, 0.1729, 0.1652, 0.1592, 0.1841, 0.1852, 0.1966, 0.2003, 0.2052, 0.1706, 0.1706, 0.1552, 0.155, 0.1688, 0.1826, 0.1957, 0.2009, 0.2131, 0.2256, 0.2271, 0.2386, 0.2388, 0.2309, 0.2782, 0.2731, 0.2991, 0.2972, 0.2949, 0.2946, 0.2977, 0.3011, 0.316, 0.3456, 0.3639, 0.3576, 0.3847, 0.3591, 0.3955, 0.392, 0.3119, 0.37, 0.3448, 0.3782, 0.3843, 0.3824, 0.4001, 0.3924, 0.3654, 0.3904, 0.4159, 0.4514, 0.4642, 0.4864, 0.4667, 0.4638, 0.4763, 0.4225, 0.4497, 0.4157]}
{"unit_id": "thread:1939451768011980829", "ticker_idx": 0, "ts": "2025-06-29T22:31:49+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is known as a main supplier of low-power DRAM not only to Samsung Electronics but also to Apple's iPhones... Samsung Electronics' MX Business Division has been confirmed to be testing Micron's LPDDR5X samples in anticipation of the Galaxy S26 series launch", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares report that Micron wins 60% LPDDR5X share in Galaxy S25 and is already being tested for S26, implying continued supply gains over Samsung DS.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Samsung MX Business Division to Equip S25 with 60% Micron DRAM, 40% In-house Memory\n\nIt is predicted that Samsung Electronics' Mobile eXperience (MX) Business Division will equip the Galaxy S25 series with more of Micron's LPDDR5X than products from Samsung Electronics' Device Solutions (DS) Business Division. Previously, after Micron was selected as the primary supplier for LPDDR5X, industry insiders had expected that the DS Business Division would secure a significant portion of the total volume. However, the current situation suggests a movement in the opposite direction.\n\nOf course, the Galaxy S25 series is still only in its 27th week before launch, and considering that a new product's active sales cycle typically lasts about two years, there is still variability. This is because Samsung Electronics often determines order volumes through weekly discussions with suppliers. Nevertheless, this assessment is based on the forecast (estimated order volume) derived by the MX Business Division after comparing the yield and performance between Micron and the DS Business Division.\n\nA source familiar with the matter stated, \"Samsung Electronics' MX Business Division will ultimately conclude with Micron's LPDDR5X being installed in the Galaxy S25 series at a 60% ratio. Samsung Electronics' DS Business Division is expected to supply only 40%. The supply ratio from initial volumes to date is also 6 to 4.\"\n\nAnalysis based on Samsung Electronics' MX Business Division's forecast (estimated order volume) suggests this development. Typically, for a company to secure the necessary volume from suppliers at the desired time, a forecast must be prepared. This pre-discussion is even more crucial for Micron, which currently has limited DRAM capacity.\n\nThere is also analysis that Samsung's DS Business Division partially agreed to the 40% supply ratio, considering production efficiency. The Galaxy S25 series requires numerous testing processes for selecting good products. The LPDDR5X from the DS Business Division is reportedly still facing cost burdens due to lower yields. Typically, low-power DRAM needs to secure a yield of 80% or more for mass production, but the DS Business Division's products are reportedly falling short of this standard.\n\nAnother industry insider said, \"The Galaxy S25 series needs to achieve a data processing speed of at least 9.6 Gbps. Therefore, the DRAM installed in it must meet this specification. However, in the case of Samsung Electronics' DS Business Division's LPDDR5X, the yield is low, creating a cost burden as more are produced.\"\n\nThey added, \"From the MX Business Division's perspective, Samsung Electronics' DS Business Division's LPDDR5X is inferior to Micron's product in terms of heat generation and power efficiency, making it difficult to adopt a significant portion. The DS Business Division also likely decided that supplying only 40% would be better, considering yield and cost burdens.\"\n\nMicron is known as a main supplier of low-power DRAM not only to Samsung Electronics but also to Apple's iPhones. According to industry sources, Micron has stably performed wafer masking processes using DUV equipment up to 1-beta (β). The LPDDR5X used in the Galaxy S25 also applies the 1β process.\n\nEarlier this year, when Micron was selected as the primary supplier of LPDDR5X for the Galaxy S25 series, industry insiders had expected the DS Business Division to supply more than half of the total volume. It was explained that there would have been internal pressure, as this was the first time in 10 years that the primary supplier position shifted from the DS Business Division to Micron.\n\nConsequently, there was speculation that high-end products and initial volumes would be primarily from Micron, while the DS Business Division would account for a higher proportion of the overall volume. As a result, predictions were made that the installation ratio between the DS Business Division and Micron would adjust from 7:3 to at most 6:4. However, with DS products falling behind in terms of performance and yield, the current situation is reportedly moving in an unexpected direction.\n\nRegarding this, a Samsung Electronics DS Business Division official stated, \"We cannot confirm information related to our partners.\"\n\nMeanwhile, Samsung Electronics' MX Business Division has been confirmed to be testing Micron's LPDDR5X samples in anticipation of the Galaxy S26 series launch in February next year. This is the same product as the LPDDR5X supplied to the Galaxy S25, with samples submitted after slightly improving the circuit linewidth. Another industry official explained, \"The sample Micron submitted this time is a version of the LPDDR5X product previously supplied to the S25, with a slightly reduced circuit linewidth. The number of dies that can be produced per wafer will increase by about 15%.\"\n\n$MU\n\n---\n\nhttps://t.co/t0CYHrehRs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012251577573582306, "ret_signed_m1d": 0.012251577573582306, "alpha_spy_m1d": 0.01701013163863674, "alpha_c_m1d": 0.013900975108794511, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01914799578333326, "ret_signed_p1d": -0.01914799578333326, "alpha_spy_p1d": -0.018824316475108405, "alpha_c_p1d": -0.00824721902093617, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.02610236865156823, "ret_signed_p1w": -0.02610236865156823, "alpha_spy_p1w": -0.0306826948673109, "alpha_c_p1w": -0.029257914435192767, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.09074737744311612, "ret_signed_p1m": -0.09074737744311612, "alpha_spy_p1m": -0.11892560887890224, "alpha_c_p1m": -0.1410198639313055, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.2736522796191896, "ret_signed_p3m": 0.2736522796191896, "alpha_spy_p3m": 0.20563519245999307, "alpha_c_p3m": 0.12240441367681187, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.2450564955537793, "ret_signed_p6m": 1.2450564955537793, "alpha_spy_p6m": 1.1251960253450775, "alpha_c_p6m": 0.9387956674719136, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2811, -0.3968, -0.4749, -0.4453, -0.4682, -0.3682, -0.4316, -0.4357, -0.4238, -0.4236, -0.4375, -0.4418, -0.4418, -0.4585, -0.4303, -0.4083, -0.3718, -0.3527, -0.3626, -0.3762, -0.3757, -0.369, -0.3451, -0.3475, -0.3468, -0.3297, -0.3091, -0.3034, -0.2512, -0.2135, -0.2266, -0.2256, -0.2049, -0.1996, -0.2041, -0.2224, -0.2306, -0.2424, -0.2424, -0.218, -0.2196, -0.2146, -0.2336, -0.2034, -0.1704, -0.1623, -0.1376, -0.1192, -0.0998, -0.0739, -0.0586, -0.0574, -0.0621, -0.0277, -0.0236, -0.0116, -0.0116, 0.0028, -0.0095, 0.0378, 0.0325, 0.0223, 0.0123, 0.0, -0.0191, -0.0123, -0.0078, -0.0078, -0.0261, 0.0104, -0.0073, -0.0002, 0.0113, -0.0367, -0.0246, -0.0544, -0.0802, -0.071, -0.0804, -0.113, -0.108, -0.0926, -0.0964, -0.0965, -0.0907, -0.0682, -0.1136, -0.1482, -0.1248, -0.1143, -0.1166, -0.0915, -0.0345, 0.0048, 0.0375, 0.0092, 0.0175, -0.0184, 0.0034, -0.0088, -0.0481, -0.0596, -0.0443, -0.0545, -0.0539, -0.0437, -0.0092, -0.0335, -0.0335, -0.0378, -0.0358, 0.0087, 0.0669, 0.0676, 0.0983, 0.137, 0.2228, 0.2769, 0.2813, 0.2898, 0.2993, 0.3716, 0.3216, 0.3369, 0.3515, 0.3133, 0.2737, 0.2772, 0.3311, 0.3588, 0.4793, 0.4923, 0.5264, 0.5518, 0.509, 0.5971, 0.5629, 0.4757, 0.5665, 0.5201, 0.5598, 0.6458, 0.6446, 0.6803, 0.6439, 0.6128, 0.6798, 0.7798, 0.7886, 0.8033, 0.8417, 0.8204, 0.8184, 0.9072, 0.7718, 0.93, 0.9367, 0.9334, 1.0584, 0.9593, 0.9901, 0.9255, 1.0058, 0.9662, 0.8569, 0.8359, 0.6364, 0.6852, 0.8197, 0.8246, 0.8712, 0.8712, 0.9217, 0.9541, 0.9462, 0.9029, 0.8418, 0.9277, 1.0065, 1.0512, 1.143, 1.1003, 0.9596, 0.93, 0.8894, 0.8326, 1.0198, 1.1609, 1.2477, 1.2451]}
{"unit_id": "quote:1930055777525608790", "ticker_idx": 0, "ts": "2025-06-04T00:15:30+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the firm maintains a positive stance on its covered companies, including TSMC", "tweet_type": "news_reaction", "timeline": "months", "summary": "U.S. bank reiterates positive stance on TSMC, MediaTek, and raises targets on Auras (3017) on improved AI supply chain yields; CoWoS capacity/shipment forecasts raised.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "- U.S. Investment Bank’s Outlook on Tech Stocks:\n\nLast week, a U.S.-based securities firm held a roadshow with over 80 foreign institutional investors and concluded that sentiment toward AI has become more optimistic compared to March.\nWhile concerns in March centered around potential demand cuts and order cancellations, the current outlook reflects improved yields across the supply chain, which are now seen as sufficient to meet AI-related demand.\n\n(The firm reaffirmed its forecast of a 58% increase in CoWoS production capacity and a 52% increase in shipments in 2025.)\n\nAs a result, the firm maintains a positive stance on its covered companies, including TSMC, MediaTek, Winstek, and MPI (Chunghwa Precision Test).", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/4 外電綜合整理\n\n- 3017奇鋐: 美系升目標\n美系券商調升25-26年EPS至$37.4/$41，給予明年20xPE評價，同步升目標。\n\n券商觀察隨著水冷散熱在GB系列良率提升，再加上ASIC server 正向需求不變，券商同時認為匯率影響可控，所以在財務模型調高毛利率及OPM預估。重申正向評等不變。\n\n- 3665貿聯:", "parent_author": null, "ret_m1d": -0.040404059997174224, "ret_signed_m1d": -0.040404059997174224, "alpha_spy_m1d": -0.040672679537220535, "alpha_c_m1d": -0.028646353207514297, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008080749427813894, "ret_signed_p1d": 0.008080749427813894, "alpha_spy_p1d": 0.012913518484003128, "alpha_c_p1d": 0.009868254040974134, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.07575758903034391, "ret_signed_p1w": 0.07575758903034391, "alpha_spy_p1w": 0.06664565786821885, "alpha_c_p1w": 0.03551927940171318, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.10568196622058523, "ret_signed_p1m": 0.10568196622058523, "alpha_spy_p1m": 0.05322822565464236, "alpha_c_p1m": -0.02091190725935177, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.18176104823984152, "ret_signed_p3m": 0.18176104823984152, "alpha_spy_p3m": 0.09613522893988491, "alpha_c_p3m": 0.02867258682531748, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.46146820804763644, "ret_signed_p6m": 0.46146820804763644, "alpha_spy_p6m": 0.31438828788140527, "alpha_c_p6m": 0.08029646837397153, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.0104, 0.0034, -0.0237, -0.0067, -0.0298, -0.0358, -0.0247, -0.0192, -0.0384, -0.0051, -0.0182, -0.0182, 0.0, -0.0101, -0.0323, -0.0384, -0.0808, -0.0465, -0.0485, -0.0485, -0.0485, -0.1434, -0.1758, -0.2071, -0.1283, -0.102, -0.1263, -0.1141, -0.1364, -0.1444, -0.1414, -0.1566, -0.1758, -0.1182, -0.1273, -0.103, -0.0939, -0.0889, -0.0828, -0.0828, -0.0404, -0.0525, -0.0707, -0.0626, -0.0727, -0.0414, -0.0333, -0.0212, 0.0091, 0.003, 0.0081, -0.0061, -0.0121, 0.0, -0.0081, -0.0061, -0.0141, -0.0253, -0.0232, -0.0232, -0.0232, -0.0444, -0.0404, 0.0, 0.0081, 0.0051, 0.0152, 0.0556, 0.0758, 0.06, 0.0448, 0.0397, 0.06, 0.0702, 0.0499, 0.0702, 0.0347, 0.0651, 0.0854, 0.0905, 0.0955, 0.0753, 0.1006, 0.1006, 0.1057, 0.1006, 0.0955, 0.0955, 0.1057, 0.1158, 0.1158, 0.1108, 0.126, 0.1463, 0.1463, 0.1716, 0.1665, 0.1463, 0.1615, 0.1615, 0.1615, 0.1615, 0.1513, 0.1716, 0.1767, 0.1767, 0.1513, 0.1665, 0.1412, 0.197, 0.1919, 0.197, 0.197, 0.2173, 0.1919, 0.197, 0.197, 0.202, 0.1513, 0.1665, 0.1513, 0.1868, 0.1919, 0.2071, 0.1767, 0.1767, 0.1818, 0.1767, 0.1767, 0.1767, 0.197, 0.197, 0.2173, 0.2426, 0.2578, 0.2781, 0.2731, 0.3036, 0.2883, 0.3087, 0.2883, 0.3189, 0.3647, 0.3647, 0.3443, 0.324, 0.324, 0.3291, 0.3494, 0.3902, 0.4258, 0.4258, 0.4615, 0.4411, 0.4666, 0.4666, 0.4411, 0.4513, 0.492, 0.5124, 0.4767, 0.5073, 0.5073, 0.4869, 0.4767, 0.4767, 0.5073, 0.5022, 0.5328, 0.5328, 0.5277, 0.5379, 0.5328, 0.4869, 0.492, 0.4869, 0.5022, 0.492, 0.5022, 0.4869, 0.4564, 0.4717, 0.4309, 0.4207, 0.4818, 0.4105, 0.4004, 0.4411, 0.4666, 0.4615]}
{"unit_id": "quote:1930055777525608790", "ticker_idx": 1, "ts": "2025-06-04T00:15:30+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the firm maintains a positive stance on its covered companies, including TSMC, MediaTek", "tweet_type": "news_reaction", "timeline": "months", "summary": "U.S. bank reiterates positive stance on TSMC, MediaTek, and raises targets on Auras (3017) on improved AI supply chain yields; CoWoS capacity/shipment forecasts raised.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "- U.S. Investment Bank’s Outlook on Tech Stocks:\n\nLast week, a U.S.-based securities firm held a roadshow with over 80 foreign institutional investors and concluded that sentiment toward AI has become more optimistic compared to March.\nWhile concerns in March centered around potential demand cuts and order cancellations, the current outlook reflects improved yields across the supply chain, which are now seen as sufficient to meet AI-related demand.\n\n(The firm reaffirmed its forecast of a 58% increase in CoWoS production capacity and a 52% increase in shipments in 2025.)\n\nAs a result, the firm maintains a positive stance on its covered companies, including TSMC, MediaTek, Winstek, and MPI (Chunghwa Precision Test).", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/4 外電綜合整理\n\n- 3017奇鋐: 美系升目標\n美系券商調升25-26年EPS至$37.4/$41，給予明年20xPE評價，同步升目標。\n\n券商觀察隨著水冷散熱在GB系列良率提升，再加上ASIC server 正向需求不變，券商同時認為匯率影響可控，所以在財務模型調高毛利率及OPM預估。重申正向評等不變。\n\n- 3665貿聯:", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.00026861954004631095, "alpha_c_m1d": 0.011757706789659927, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02352947850070075, "ret_signed_p1d": -0.02352947850070075, "alpha_spy_p1d": -0.018696709444511517, "alpha_c_p1d": -0.02174197388754051, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.03137243896375663, "ret_signed_p1w": 0.03137243896375663, "alpha_spy_p1w": 0.02226050780163158, "alpha_c_p1w": -0.008865870664874098, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.015686418519055945, "ret_signed_p1m": 0.015686418519055945, "alpha_spy_p1m": -0.03676732204688693, "alpha_c_p1m": -0.11090745496088106, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.08766413352050795, "ret_signed_p3m": 0.08766413352050795, "alpha_spy_p3m": 0.0020383142205513405, "alpha_c_p3m": -0.06542432789401609, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.07166910785764946, "ret_signed_p6m": 0.07166910785764946, "alpha_spy_p6m": -0.07541081230858171, "alpha_c_p6m": -0.30950263181601545, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.149, 0.0941, 0.0863, 0.1059, 0.0902, 0.098, 0.1176, 0.1569, 0.149, 0.1647, 0.1608, 0.1608, 0.1961, 0.1882, 0.1647, 0.149, 0.0902, 0.1451, 0.1255, 0.1255, 0.1255, 0.0157, -0.0118, -0.0706, 0.0196, 0.0863, 0.098, 0.0863, 0.0706, 0.051, 0.0706, 0.0549, 0.0196, 0.0745, 0.0431, 0.0824, 0.0588, 0.0745, 0.0588, 0.0588, 0.0196, 0.0157, 0.0039, 0.0, 0.0118, 0.0392, 0.0314, 0.0627, 0.0667, 0.0745, 0.0706, 0.0275, 0.0314, 0.051, 0.0392, 0.0314, 0.0039, 0.0039, 0.0039, -0.0118, -0.0118, -0.0118, 0.0, 0.0, -0.0235, 0.0039, 0.0078, 0.0471, 0.0314, 0.0078, -0.0196, -0.0157, 0.0, 0.0078, -0.0078, -0.0196, -0.0157, -0.0039, 0.0157, 0.0118, 0.0078, -0.0196, 0.0, 0.0157, 0.0157, 0.0317, 0.0237, 0.0157, 0.0837, 0.1197, 0.1356, 0.1077, 0.1157, 0.1277, 0.1117, 0.1277, 0.1396, 0.1436, 0.1436, 0.1436, 0.1436, 0.1197, 0.0877, 0.1037, 0.0957, 0.0957, 0.0637, 0.0717, 0.0557, 0.0837, 0.0797, 0.0837, 0.0877, 0.0957, 0.1117, 0.0957, 0.1277, 0.1117, 0.0837, 0.0917, 0.0917, 0.1277, 0.1197, 0.1197, 0.1077, 0.0957, 0.0877, 0.0877, 0.1157, 0.1077, 0.1476, 0.1436, 0.2076, 0.1916, 0.1836, 0.1876, 0.1876, 0.2276, 0.2076, 0.2076, 0.1516, 0.1237, 0.0997, 0.0637, 0.0757, 0.0477, 0.0477, 0.0517, 0.0277, 0.0237, 0.0557, 0.0557, 0.0597, 0.0677, 0.0757, 0.0757, 0.0517, 0.0437, 0.0517, 0.0637, 0.0637, 0.0717, 0.0717, 0.0637, 0.0357, 0.0357, 0.0677, 0.0517, 0.0397, 0.0477, 0.0477, 0.0397, 0.0517, 0.0637, 0.0317, 0.0077, -0.0003, -0.0083, -0.0043, -0.0043, -0.0163, -0.0163, -0.0643, -0.0723, -0.0523, -0.0843, -0.0803, -0.0523, 0.0397, 0.0717]}
{"unit_id": "quote:1930055777525608790", "ticker_idx": 2, "ts": "2025-06-04T00:15:30+00:00", "symbol": "3017.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "美系券商調升25-26年EPS至$37.4/$41，給予明年20xPE評價，同步升目標。重申正向評等不變", "tweet_type": "news_reaction", "timeline": "months", "summary": "U.S. bank reiterates positive stance on TSMC, MediaTek, and raises targets on Auras (3017) on improved AI supply chain yields; CoWoS capacity/shipment forecasts raised.", "resolved_tickers": ["3017.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "3017.TW is Yuanta Financial on TWSE", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "quote", "lang": "en", "tweet_text": "- U.S. Investment Bank’s Outlook on Tech Stocks:\n\nLast week, a U.S.-based securities firm held a roadshow with over 80 foreign institutional investors and concluded that sentiment toward AI has become more optimistic compared to March.\nWhile concerns in March centered around potential demand cuts and order cancellations, the current outlook reflects improved yields across the supply chain, which are now seen as sufficient to meet AI-related demand.\n\n(The firm reaffirmed its forecast of a 58% increase in CoWoS production capacity and a 52% increase in shipments in 2025.)\n\nAs a result, the firm maintains a positive stance on its covered companies, including TSMC, MediaTek, Winstek, and MPI (Chunghwa Precision Test).", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/4 外電綜合整理\n\n- 3017奇鋐: 美系升目標\n美系券商調升25-26年EPS至$37.4/$41，給予明年20xPE評價，同步升目標。\n\n券商觀察隨著水冷散熱在GB系列良率提升，再加上ASIC server 正向需求不變，券商同時認為匯率影響可控，所以在財務模型調高毛利率及OPM預估。重申正向評等不變。\n\n- 3665貿聯:", "parent_author": null, "ret_m1d": -0.032305442214203284, "ret_signed_m1d": -0.032305442214203284, "alpha_spy_m1d": -0.032574061754249595, "alpha_c_m1d": -0.030827289057342777, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.0014781531568605066, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007342187053177351, "ret_signed_p1d": -0.007342187053177351, "alpha_spy_p1d": -0.002509417996988117, "alpha_c_p1d": -0.004512705489685409, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0028294815634919424, "ret_p1w": 0.06314244701777105, "ret_signed_p1w": 0.06314244701777105, "alpha_spy_p1w": 0.054030515855646, "alpha_c_p1w": 0.04819326310944061, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.014949183908330443, "ret_p1m": 0.14977971178593252, "ret_signed_p1m": 0.14977971178593252, "alpha_spy_p1m": 0.09732597121998965, "alpha_c_p1m": 0.06284617352834254, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.08693353825758998, "ret_p3m": 0.3938803535424449, "ret_signed_p3m": 0.3938803535424449, "alpha_spy_p3m": 0.3082545342424883, "alpha_c_p3m": 0.2837241471850318, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.11015620635741308, "ret_p6m": 1.066379482669193, "ret_signed_p6m": 1.066379482669193, "alpha_spy_p6m": 0.9192995625029619, "alpha_c_p6m": 0.8644554274293348, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.20192405523985824, "price_path": [-0.1791, -0.1718, -0.1777, -0.1836, -0.1689, -0.1542, -0.1762, -0.1718, -0.2173, -0.1909, -0.2144, -0.2364, -0.2276, -0.2188, -0.2217, -0.2599, -0.3333, -0.3267, -0.3157, -0.3157, -0.3157, -0.384, -0.4449, -0.5, -0.4501, -0.3957, -0.3803, -0.3612, -0.3899, -0.3855, -0.3877, -0.4082, -0.4148, -0.3715, -0.3524, -0.3084, -0.2944, -0.2996, -0.3186, -0.3186, -0.2555, -0.2628, -0.2467, -0.2408, -0.2276, -0.2232, -0.2056, -0.1733, -0.1057, -0.1116, -0.1175, -0.1263, -0.1219, -0.091, -0.0969, -0.1145, -0.1145, -0.1072, -0.0984, -0.0896, -0.0896, -0.0764, -0.0323, 0.0, -0.0073, 0.0, 0.0015, 0.0294, 0.0631, 0.1145, 0.1204, 0.1013, 0.1116, 0.072, 0.0881, 0.0778, 0.0705, 0.0822, 0.094, 0.0837, 0.0822, 0.091, 0.1087, 0.1145, 0.1498, 0.141, 0.1189, 0.1674, 0.1909, 0.2408, 0.2144, 0.2173, 0.2526, 0.2452, 0.2394, 0.3084, 0.2805, 0.2511, 0.2394, 0.257, 0.2467, 0.2819, 0.2893, 0.2805, 0.3554, 0.3554, 0.3715, 0.4347, 0.4758, 0.5051, 0.6006, 0.5859, 0.6006, 0.6446, 0.5786, 0.5712, 0.6814, 0.585, 0.4887, 0.5553, 0.5257, 0.5702, 0.5183, 0.5553, 0.5035, 0.4961, 0.3939, 0.4057, 0.3865, 0.385, 0.3983, 0.4279, 0.4235, 0.4502, 0.5183, 0.5553, 0.5479, 0.5405, 0.4887, 0.5627, 0.5405, 0.5035, 0.5035, 0.4724, 0.422, 0.3939, 0.3939, 0.4531, 0.5183, 0.6294, 0.7257, 0.7257, 0.7035, 0.6961, 0.7331, 0.7331, 0.7183, 0.7109, 0.8072, 0.7701, 0.7035, 0.6887, 0.7109, 0.7405, 0.7331, 0.7331, 0.7627, 0.9257, 0.9701, 1.0886, 1.1108, 1.2145, 1.1256, 1.1404, 1.2145, 1.1701, 1.296, 1.2812, 1.2293, 1.2145, 1.1034, 1.1034, 1.0293, 0.9849, 0.9775, 0.8368, 0.8442, 0.9479, 1.0516, 1.0664]}
{"unit_id": "quote:1930055777525608790", "ticker_idx": 3, "ts": "2025-06-04T00:15:30+00:00", "symbol": "CoWoS", "kind": "sector", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the firm reaffirmed its forecast of a 58% increase in CoWoS production capacity and a 52% increase in shipments in 2025", "tweet_type": "news_reaction", "timeline": "months", "summary": "U.S. bank reiterates positive stance on TSMC, MediaTek, and raises targets on Auras (3017) on improved AI supply chain yields; CoWoS capacity/shipment forecasts raised.", "resolved_tickers": ["TSM", "ASX", "3034.TW"], "resolution_method": "basket", "resolution_reason": "CoWoS advanced packaging basket: TSMC, ASE, Unimicron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "- U.S. Investment Bank’s Outlook on Tech Stocks:\n\nLast week, a U.S.-based securities firm held a roadshow with over 80 foreign institutional investors and concluded that sentiment toward AI has become more optimistic compared to March.\nWhile concerns in March centered around potential demand cuts and order cancellations, the current outlook reflects improved yields across the supply chain, which are now seen as sufficient to meet AI-related demand.\n\n(The firm reaffirmed its forecast of a 58% increase in CoWoS production capacity and a 52% increase in shipments in 2025.)\n\nAs a result, the firm maintains a positive stance on its covered companies, including TSMC, MediaTek, Winstek, and MPI (Chunghwa Precision Test).", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/4 外電綜合整理\n\n- 3017奇鋐: 美系升目標\n美系券商調升25-26年EPS至$37.4/$41，給予明年20xPE評價，同步升目標。\n\n券商觀察隨著水冷散熱在GB系列良率提升，再加上ASIC server 正向需求不變，券商同時認為匯率影響可控，所以在財務模型調高毛利率及OPM預估。重申正向評等不變。\n\n- 3665貿聯:", "parent_author": null, "ret_m1d": -0.014774511357090403, "ret_signed_m1d": -0.014774511357090403, "alpha_spy_m1d": -0.015043130897136714, "alpha_c_m1d": -0.0030168045674304755, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.000169274177924413, "ret_signed_p1d": 0.000169274177924413, "alpha_spy_p1d": 0.005002043234113647, "alpha_c_p1d": 0.0019567787910846532, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.04963897887272887, "ret_signed_p1w": 0.04963897887272887, "alpha_spy_p1w": 0.040527047710603815, "alpha_c_p1w": 0.009400669244098138, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.12639738880076115, "ret_signed_p1m": 0.12639738880076115, "alpha_spy_p1m": 0.07394364823481828, "alpha_c_p1m": -0.00019648467917585122, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.039353558506326713, "ret_signed_p3m": 0.039353558506326713, "alpha_spy_p3m": -0.046272260793629894, "alpha_c_p3m": -0.11373490290819732, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.27438139970531483, "ret_signed_p6m": 0.27438139970531483, "alpha_spy_p6m": 0.12730147953908366, "alpha_c_p6m": -0.10679033996835008, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.0029, -0.0255, -0.0338, -0.0083, -0.024, -0.0188, -0.0108, -0.0121, -0.0138, 0.0018, -0.0002, 0.013, 0.0104, -0.0114, -0.0256, -0.0478, -0.058, -0.041, -0.0509, -0.1009, -0.1366, -0.1761, -0.1919, -0.1423, -0.1411, -0.1121, -0.1216, -0.1139, -0.1369, -0.1413, -0.1359, -0.1481, -0.1478, -0.1178, -0.1068, -0.0878, -0.0845, -0.0875, -0.0754, -0.0665, -0.0446, -0.047, -0.0549, -0.0497, -0.0526, -0.0457, 0.002, 0.0289, 0.0346, 0.0266, 0.0147, 0.0057, 0.0093, 0.0065, 0.0128, -0.0051, 0.001, 0.0105, 0.0031, -0.0, -0.0112, -0.0227, -0.0148, 0.0, 0.0002, 0.0073, 0.0165, 0.0398, 0.0496, 0.0498, 0.0385, 0.0614, 0.0578, 0.0642, 0.0555, 0.0446, 0.0552, 0.0844, 0.0938, 0.0946, 0.1055, 0.0971, 0.0883, 0.1165, 0.1264, 0.123, 0.1045, 0.1012, 0.1145, 0.1019, 0.1047, 0.0936, 0.1123, 0.1194, 0.1384, 0.1252, 0.119, 0.1035, 0.1279, 0.1296, 0.1316, 0.1252, 0.1178, 0.1094, 0.0758, 0.0835, 0.0865, 0.0668, 0.0547, 0.0708, 0.0681, 0.0643, 0.0729, 0.0689, 0.0621, 0.0512, 0.0611, 0.0395, 0.0263, 0.0217, 0.0375, 0.0461, 0.0587, 0.0625, 0.0587, 0.0443, 0.0394, 0.0513, 0.0588, 0.0679, 0.0875, 0.1299, 0.121, 0.1354, 0.1204, 0.1193, 0.1312, 0.1377, 0.1403, 0.1663, 0.1668, 0.1801, 0.2011, 0.191, 0.1807, 0.1606, 0.159, 0.1619, 0.1711, 0.1727, 0.1773, 0.2053, 0.1956, 0.2302, 0.2164, 0.1621, 0.215, 0.1842, 0.2199, 0.2378, 0.2314, 0.2514, 0.2372, 0.2225, 0.2349, 0.2476, 0.2609, 0.2653, 0.3073, 0.3249, 0.3489, 0.3634, 0.3176, 0.3315, 0.3025, 0.2887, 0.3111, 0.2943, 0.2944, 0.2687, 0.2709, 0.2576, 0.2451, 0.2489, 0.2302, 0.2258, 0.2541, 0.2574, 0.2762, 0.2744]}
{"unit_id": "orig:1925324063494545538", "ticker_idx": 0, "ts": "2025-05-21T22:53:22+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI server revenue is expected to reach nearly $20 billion this year, but growth is largely back-loaded, with significant acceleration expected in the second half following large-scale shipments of GB200 racks beginning in late Q3", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish view on Dell: AI server revenue approaching $20B with significant 2H acceleration on GB200 rack ramp.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Dell: AI Server Growth to Accelerate in 2H\nMorgan Stanley (E. Woodring, 2025/05/20)\n\nDell is expected to slightly beat expectations for fiscal 1Q25 results, while maintaining its full-year revenue guidance of $102 billion and EPS of $9.30. Driven by strong AI infrastructure demand and steady performance in traditional hardware markets, AI server revenue for the year is expected to approach $20 billion, though growth will be mainly concentrated in the second half following GB200 rack capacity ramp-up.\n\n• 1Q Slightly Beats Expectations, 2Q EPS Guidance\nAI servers contributed approximately $2.3 billion in revenue during the first quarter, with overall results slightly exceeding expectations. The company issued a conservative 2Q EPS guidance of approximately $2.00 (vs. Morgan Stanley’s estimate of $2.10), implying revenue will remain flat or grow slightly quarter-over-quarter.\n\n• Full-Year Guidance Maintained\nManagement reaffirmed full-year revenue guidance at $102 billion and EPS at $9.30, and did not raise its outlook for fiscal 2026 during this earnings release.\n\n• AI Server Revenue to Surge in 2H\nAI server revenue is expected to reach nearly $20 billion this year, but growth is largely back-loaded, with significant acceleration expected in the second half following large-scale shipments of GB200 racks beginning in late Q3.\n\n• Steady Progress in Traditional Hardware\nDemand for storage and general-purpose servers remains stable. Dell continues to gain share in the enterprise storage and server markets, maintaining margin levels through cost optimization.\n\n• Risks and Opportunities Coexist\nUncertainties in the global tariff environment and component supply chain pace may still impact second-half performance. However, partnerships with sovereign projects like Abu Dhabi’s G42 and Dell’s leading position in the AI infrastructure market provide ongoing growth momentum.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02054852770612059, "ret_signed_m1d": 0.02054852770612059, "alpha_spy_m1d": 0.0034090260163195385, "alpha_c_m1d": 0.001401714679618804, "bench_spy_m1d": 0.01713950168980105, "bench_c_m1d": 0.019146813026501786, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00017853241785426022, "ret_signed_p1d": 0.00017853241785426022, "alpha_spy_p1d": -0.00021590278355976267, "alpha_c_p1d": -8.310171722802906e-05, "bench_spy_p1d": 0.0003944352014140229, "bench_c_p1d": 0.0002616341350822893, "ret_p1w": 0.016438752966285053, "ret_signed_p1w": 0.016438752966285053, "alpha_spy_p1w": 0.0080834842059343, "alpha_c_p1w": 0.00789029767523508, "bench_spy_p1w": 0.008355268760350754, "bench_c_p1w": 0.008548455291049972, "ret_p1m": 0.041365061634857314, "ret_signed_p1m": 0.041365061634857314, "alpha_spy_p1m": 0.016350519219939352, "alpha_c_p1m": -0.012411897172524666, "bench_spy_p1m": 0.025014542414917962, "bench_c_p1m": 0.05377695880738198, "ret_p3m": 0.23911946981130994, "ret_signed_p3m": 0.23911946981130994, "alpha_spy_p3m": 0.13216122613678372, "alpha_c_p3m": 0.07586371846185935, "bench_spy_p3m": 0.10695824367452622, "bench_c_p3m": 0.16325575134945058, "ret_p6m": 0.20581334767866677, "ret_signed_p6m": 0.20581334767866677, "alpha_spy_p6m": 0.04619472118592438, "alpha_c_p6m": -0.047823644563536494, "bench_spy_p6m": 0.1596186264927424, "bench_c_p6m": 0.25363699224220326, "price_path": [0.044, 0.0132, -0.0177, 0.0267, -0.0428, -0.0878, -0.1517, -0.1649, -0.1367, -0.1742, -0.1881, -0.198, -0.1871, -0.1604, -0.1793, -0.1507, -0.1331, -0.1439, -0.1236, -0.1219, -0.1338, -0.114, -0.1224, -0.1455, -0.138, -0.1807, -0.1908, -0.1837, -0.1537, -0.3144, -0.3641, -0.3342, -0.3556, -0.2526, -0.2992, -0.2727, -0.2437, -0.2411, -0.26, -0.2472, -0.2472, -0.2686, -0.2488, -0.2034, -0.1566, -0.1522, -0.156, -0.1622, -0.1802, -0.1785, -0.1549, -0.157, -0.1606, -0.1442, -0.1404, -0.1431, -0.076, -0.0365, -0.0081, -0.0095, 0.0202, 0.021, 0.0205, 0.0, 0.0002, 0.0016, 0.0016, 0.0184, 0.0164, 0.0152, -0.0059, -0.0344, -0.0003, 0.0069, -0.001, 0.0163, 0.0205, 0.0118, -0.0062, 0.0131, -0.0212, 0.0162, 0.0357, 0.0414, 0.0414, 0.0665, 0.0544, 0.0774, 0.0762, 0.1255, 0.1077, 0.0953, 0.0876, 0.1032, 0.1187, 0.1187, 0.1147, 0.1113, 0.1313, 0.1428, 0.1331, 0.1228, 0.1229, 0.104, 0.1068, 0.1725, 0.1521, 0.1153, 0.1412, 0.1514, 0.1771, 0.2008, 0.1977, 0.1979, 0.1903, 0.1421, 0.1683, 0.1705, 0.1495, 0.2014, 0.2345, 0.2408, 0.2706, 0.2482, 0.2457, 0.2405, 0.2391, 0.2128, 0.1526, 0.1467, 0.1737, 0.1752, 0.1751, 0.1886, 0.2025, 0.0958, 0.0958, 0.0851, 0.1125, 0.1363, 0.1198, 0.1034, 0.0881, 0.1164, 0.1247, 0.1217, 0.1375, 0.1454, 0.1692, 0.1851, 0.1836, 0.2172, 0.2051, 0.1849, 0.1748, 0.173, 0.2012, 0.2718, 0.3427, 0.322, 0.2625, 0.3076, 0.3534, 0.4759, 0.399, 0.3507, 0.3761, 0.3346, 0.3788, 0.3574, 0.3419, 0.3265, 0.3453, 0.3516, 0.3885, 0.4282, 0.4601, 0.4844, 0.4728, 0.4495, 0.4585, 0.4414, 0.3922, 0.3721, 0.343, 0.3207, 0.2846, 0.2492, 0.2668, 0.2058]}
{"unit_id": "orig:1940901566707560654", "ticker_idx": 0, "ts": "2025-07-03T22:32:48+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Focus on Traditional Products and Organizational Strategy Optimization… Intel still has the potential to expand market share and increase pricing power… achieving CPU product leadership is key to valuation recovery", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley note: INTC foundry delay to 14A is manageable; CPU product recovery is the core path to valuation re-rating.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Intel: Foundry Delay to 14A May Be Rational Adjustment, Focus on CPU Products Benefits Recovery\n\nMorgan Stanley (J. Moore, 25/07/03)\n\nMorgan Stanley points out that if Intel shifts its Foundry focus from 18A to the 14A node in 2027, while not officially confirmed, the short-term financial impact would be controllable due to already limited customer expectations for 18A. Instead, this move could help the company focus on its core business and rebuild the competitiveness of its traditional CPU products.\n\nFoundry Focus Delayed to 14A, Short-Term Economic Impact Controllable\n\nThe 18A node was already cooling down, and even if Foundry-related projects are delayed to 14A, customer contributions for 2025-2026 would be marginal. Significant asset impairment or profit impact is not expected. Morgan Stanley believes that most 18A node Foundry customers were primarily engaged in small, exploratory projects with limited capital expenditure involved.\n\nPast Foundry Attempts Lacked Sustainability, Organizational Mechanism Still Immature\n\nIntel's previous Foundry model failed to effectively serve diverse customers because its processes were optimized for its own high-performance CPUs, leading to high costs and narrow process applicability. Although the current Foundry spin-off into an independent business unit helps enhance market orientation, it remains questionable whether this will improve the overall existing product lines.\n\n18A Wafer Manufacturing Costs May Inhibit Capacity Utilization Due to Internal Pricing Adjustments\n\nIntel recently disclosed that some Intel 7 products will face long-term shortages. One reason might be the excessively high costs of the Intel 3/4 nodes, limiting Intel's own transition. Morgan Stanley speculates that the company might be raising the internal pricing for its own wafers to pursue Foundry cost competitiveness, which could in turn lead to an imbalance in Fab resource allocation.\n\nFocus on Traditional Products and Organizational Strategy Optimization, Weakening AI and Foundry Interference\n\nCanceling edge AI projects like Falcon Shores and delaying Foundry progress can help management refocus on the quantifiable, high-market-share client and server CPU core businesses. Driven by product refresh cycles, Intel still has the potential to expand market share and increase pricing power.\n\nShort-Term Market Sentiment Favors Foundry, But Product Recovery is the Core Path\n\nThe market still hopes for a breakthrough from Intel Foundry, but Morgan Stanley emphasizes that achieving CPU product leadership is key to valuation recovery. Although recent rumors of business spin-offs or joint ventures have sparked speculation, if the traditional CPU product line lacks a clear roadmap, financial performance will continue to be under pressure.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0271231932663214, "ret_signed_m1d": -0.0271231932663214, "alpha_spy_m1d": -0.019303437822174474, "alpha_c_m1d": -0.01876686257408655, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.059137391607004375, "ret_signed_p1w": 0.059137391607004375, "alpha_spy_p1w": 0.05836982636666588, "alpha_c_p1w": 0.0453866835531509, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": -0.1413961910866519, "ret_signed_p1m": -0.1413961910866519, "alpha_spy_p1m": -0.13560716640339565, "alpha_c_p1m": -0.1425597308032922, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.49177410309028646, "ret_signed_p3m": 0.49177410309028646, "alpha_spy_p3m": 0.4235121777023918, "alpha_c_p3m": 0.3410796274646086, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.6096043188683258, "ret_signed_p6m": 0.6096043188683258, "alpha_spy_p6m": 0.49937750127275393, "alpha_c_p6m": 0.3156251080202499, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.1298, -0.1939, -0.0427, -0.1161, -0.1223, -0.0969, -0.1174, -0.145, -0.1583, -0.1583, -0.1623, -0.1325, -0.0845, -0.0445, -0.1085, -0.088, -0.0956, -0.1063, -0.1116, -0.0831, -0.0987, -0.1134, -0.0969, -0.0663, -0.0476, -0.0138, 0.0031, -0.0431, -0.0418, -0.0369, -0.0494, -0.0542, -0.08, -0.0863, -0.1085, -0.1085, -0.0863, -0.0943, -0.0996, -0.1307, -0.1223, -0.0978, -0.0996, -0.1112, -0.108, -0.0894, -0.0182, -0.0805, -0.0765, -0.1045, -0.0778, -0.0751, -0.0445, -0.0445, -0.0627, -0.0578, 0.0027, -0.0129, 0.0004, 0.0089, -0.004, 0.016, -0.0271, 0.0, 0.0, -0.0218, 0.0489, 0.0422, 0.0591, 0.0418, 0.036, 0.0191, 0.0089, 0.0138, 0.0271, 0.0342, 0.0333, 0.0445, 0.0062, -0.0796, -0.0805, -0.0925, -0.0956, -0.1196, -0.1414, -0.1329, -0.1023, -0.0925, -0.1209, -0.1129, -0.0818, -0.0302, -0.012, 0.0609, 0.092, 0.052, 0.1254, 0.0467, 0.0449, 0.1027, 0.0916, 0.0827, 0.1049, 0.1085, 0.0827, 0.0827, 0.0765, 0.0671, 0.0943, 0.0889, 0.0885, 0.0867, 0.1014, 0.0943, 0.0707, 0.1014, 0.1236, 0.1072, 0.3593, 0.3153, 0.2788, 0.3046, 0.3882, 0.5113, 0.5785, 0.5331, 0.4918, 0.598, 0.6585, 0.6376, 0.6269, 0.6527, 0.6643, 0.6807, 0.6172, 0.655, 0.5843, 0.6518, 0.6381, 0.6456, 0.6941, 0.695, 0.6416, 0.6968, 0.7021, 0.7581, 0.8466, 0.8382, 0.7857, 0.7781, 0.7563, 0.6465, 0.7065, 0.6558, 0.6954, 0.7096, 0.6843, 0.6847, 0.5967, 0.5794, 0.5434, 0.5265, 0.5611, 0.4949, 0.534, 0.5914, 0.5932, 0.6367, 0.6367, 0.8035, 0.779, 0.9329, 0.9458, 0.8008, 0.8413, 0.7919, 0.8008, 0.8133, 0.7568, 0.6812, 0.6679, 0.659, 0.6029, 0.6132, 0.6372, 0.6172, 0.6163, 0.6078, 0.6078, 0.6096]}
{"unit_id": "thread:1936383013321113890", "ticker_idx": 0, "ts": "2025-06-21T11:17:41+00:00", "symbol": "Chinese semiconductor equipment", "kind": "sector", "geography": "China", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "it would essentially be a huge gift to Chinese domestic equipment suppliers", "tweet_type": "analysis", "timeline": "years", "summary": "US export restrictions will benefit Chinese domestic equipment makers; TSMC/Samsung/SK Hynix China fabs will transition to Chinese equipment.", "resolved_tickers": ["688012.SS", "002371.SZ", "300475.SZ"], "resolution_method": "basket", "resolution_reason": "China semi equip basket: AMEC, Naura, Advanced Micro", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "An industry source cited by Reuters pointed out that if it becomes more difficult to export U.S. equipment to foreign companies operating in China, it would essentially be a huge gift to Chinese domestic equipment suppliers.\n\nAnd that’s completely expected. At the end of the day, every company is driven by profit. You can’t just lift an entire fab like Aladdin’s lamp and move it to another country. So as long as these fabs were built in China, they’ll inevitably have to find ways to keep them running, regardless of U.S. sanctions.\n\nI believe that, in the end, the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment.\n\n---\n\nhttps://t.co/tBW57ynWJH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022253209427582132, "ret_signed_m1d": -0.022253209427582132, "alpha_spy_m1d": -0.012472339277312727, "alpha_c_m1d": -0.015953516067201856, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0010086530461819547, "ret_signed_p1d": -0.0010086530461819547, "alpha_spy_p1d": -0.012055862039269708, "alpha_c_p1d": -0.03773738474376124, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.005822901635315329, "ret_signed_p1w": 0.005822901635315329, "alpha_spy_p1w": -0.02366975771215137, "alpha_c_p1w": -0.058929647812018, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": -0.012705235160119713, "ret_signed_p1m": -0.012705235160119713, "alpha_spy_p1m": -0.06054324388921165, "alpha_c_p1m": -0.10223638369590275, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.5909593228295971, "ret_signed_p3m": 0.5909593228295971, "alpha_spy_p3m": 0.48746855781098963, "alpha_c_p3m": 0.38078063126772665, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 1.1574025165848056, "ret_signed_p6m": 1.1574025165848056, "alpha_spy_p6m": 1.0230992157654852, "alpha_c_p6m": 0.8137098303534809, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [0.0149, 0.0257, -0.003, 0.0041, 0.0129, 0.0079, -0.0039, -0.0039, -0.0918, -0.0997, -0.0576, -0.0501, -0.0304, -0.0259, -0.0388, -0.028, -0.0274, -0.0388, -0.0282, -0.0258, -0.0216, -0.0308, -0.0187, -0.0261, -0.0185, -0.0229, -0.0229, -0.0229, -0.0229, 0.0007, -0.0121, -0.0082, -0.0431, -0.0332, -0.0307, -0.0222, -0.0507, -0.0632, -0.0612, -0.0625, -0.0747, -0.081, -0.0889, -0.0789, -0.0873, -0.105, -0.0813, -0.0961, -0.0961, -0.0824, -0.0698, -0.0525, -0.0409, -0.0373, -0.0663, -0.0612, -0.0702, -0.0483, -0.0527, -0.0439, -0.0212, -0.0166, -0.0223, 0.0, -0.001, 0.0115, -0.0028, -0.0051, 0.0058, 0.0025, -0.0179, 0.0017, 0.0116, 0.006, 0.0134, -0.0035, -0.0157, -0.0072, -0.0211, -0.0177, -0.0177, -0.0142, -0.0134, -0.017, -0.0127, 0.0201, 0.0297, 0.0388, 0.0412, 0.0572, 0.0545, 0.0264, 0.0085, 0.0208, 0.0208, 0.0321, 0.0412, 0.0261, 0.0267, 0.0441, 0.0338, 0.036, 0.0488, 0.0645, 0.0487, 0.075, 0.1079, 0.1639, 0.1619, 0.1963, 0.1586, 0.2135, 0.1732, 0.2396, 0.1928, 0.188, 0.0946, 0.1389, 0.245, 0.2509, 0.2602, 0.3808, 0.466, 0.4912, 0.4838, 0.5503, 0.591, 0.6053, 0.7149, 0.725, 0.803, 0.7502, 0.7278, 0.8215, 0.8682, 0.8682, 0.8682, 0.8682, 0.8682, 0.8682, 0.8682, 0.9805, 0.9324, 0.8556, 0.7402, 0.7659, 0.9035, 0.8304, 0.7927, 0.8944, 0.9108, 0.8978, 1.1324, 1.2297, 1.1772, 1.1997, 1.1776, 1.1668, 1.2604, 1.2941, 1.3287, 1.4848, 1.5761, 1.8306, 1.6331, 1.7954, 1.7781, 1.5839, 1.5337, 1.4846, 1.4829, 1.2964, 1.0664, 1.1089, 1.1474, 1.173, 1.165, 1.2539, 1.203, 1.1403, 1.103, 1.0982, 1.126, 1.3189, 1.3366, 1.3062, 1.2915, 1.3042, 1.2124, 1.1574]}
{"unit_id": "thread:1936383013321113890", "ticker_idx": 1, "ts": "2025-06-21T11:17:41+00:00", "symbol": "TSMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment", "tweet_type": "analysis", "timeline": "years", "summary": "US export restrictions will benefit Chinese domestic equipment makers; TSMC/Samsung/SK Hynix China fabs will transition to Chinese equipment.", "resolved_tickers": ["TSM"], "resolution_method": "exchange_ticker", "resolution_reason": "TSMC listed on NYSE as TSM (ADR); also 2330.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "An industry source cited by Reuters pointed out that if it becomes more difficult to export U.S. equipment to foreign companies operating in China, it would essentially be a huge gift to Chinese domestic equipment suppliers.\n\nAnd that’s completely expected. At the end of the day, every company is driven by profit. You can’t just lift an entire fab like Aladdin’s lamp and move it to another country. So as long as these fabs were built in China, they’ll inevitably have to find ways to keep them running, regardless of U.S. sanctions.\n\nI believe that, in the end, the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment.\n\n---\n\nhttps://t.co/tBW57ynWJH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0038513868739630253, "ret_signed_m1d": 0.0038513868739630253, "alpha_spy_m1d": -0.00592948327630638, "alpha_c_m1d": -0.002448306486417251, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04645287181090407, "ret_signed_p1d": -0.04645287181090407, "alpha_spy_p1d": -0.03540566281781632, "alpha_c_p1d": -0.009724140113324786, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.07688276527060123, "ret_signed_p1w": -0.07688276527060123, "alpha_spy_p1w": -0.04739010592313453, "alpha_c_p1w": -0.012130215823267898, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": 0.11544310415283543, "ret_signed_p1m": -0.11544310415283543, "alpha_spy_p1m": -0.06760509542374349, "alpha_c_p1m": -0.025911955617052396, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.2813211688404449, "ret_signed_p3m": -0.2813211688404449, "alpha_spy_p3m": -0.17783040382183746, "alpha_c_p3m": -0.07114247727857448, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 0.37196590218525305, "ret_signed_p6m": -0.37196590218525305, "alpha_spy_p6m": -0.23766260136593265, "alpha_c_p6m": -0.028273215953928288, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [-0.1781, -0.203, -0.2172, -0.2136, -0.2011, -0.1928, -0.2544, -0.3046, -0.3075, -0.3303, -0.2479, -0.284, -0.2559, -0.2617, -0.2547, -0.2815, -0.2812, -0.2812, -0.2995, -0.2828, -0.2524, -0.2222, -0.2179, -0.2259, -0.2208, -0.2103, -0.1818, -0.1507, -0.1643, -0.1839, -0.1731, -0.1699, -0.1638, -0.1142, -0.081, -0.0774, -0.0799, -0.0799, -0.0833, -0.0836, -0.0916, -0.0706, -0.0905, -0.0905, -0.0635, -0.0708, -0.066, -0.0842, -0.077, -0.0639, -0.0412, -0.0367, -0.028, -0.0194, 0.0065, 0.0143, 0.0243, 0.0037, 0.0255, 0.017, 0.0151, 0.0151, -0.0039, 0.0, 0.0465, 0.0591, 0.0651, 0.0868, 0.0769, 0.0683, 0.1107, 0.1164, 0.1164, 0.0896, 0.0834, 0.1023, 0.0924, 0.0955, 0.0872, 0.1266, 0.1295, 0.1677, 0.143, 0.1357, 0.1154, 0.1427, 0.1487, 0.1677, 0.1542, 0.1474, 0.155, 0.1488, 0.1183, 0.1364, 0.1053, 0.1001, 0.1536, 0.1498, 0.1511, 0.1615, 0.148, 0.1459, 0.1358, 0.1478, 0.1064, 0.0869, 0.0809, 0.1078, 0.1202, 0.135, 0.1377, 0.1329, 0.0977, 0.0977, 0.0859, 0.1002, 0.1183, 0.1573, 0.1753, 0.193, 0.2383, 0.231, 0.233, 0.2428, 0.2499, 0.2534, 0.2813, 0.2633, 0.3004, 0.3484, 0.3389, 0.3196, 0.3038, 0.3032, 0.3321, 0.3759, 0.3742, 0.3936, 0.4423, 0.4024, 0.4525, 0.4303, 0.3387, 0.4447, 0.4115, 0.4534, 0.4301, 0.4074, 0.4199, 0.4047, 0.3779, 0.3867, 0.4069, 0.4226, 0.4382, 0.4552, 0.4463, 0.4329, 0.4541, 0.4025, 0.4006, 0.3796, 0.3665, 0.4083, 0.3888, 0.3862, 0.346, 0.3585, 0.3451, 0.3255, 0.3468, 0.3236, 0.3119, 0.3576, 0.3578, 0.383, 0.383, 0.3904, 0.3721, 0.3932, 0.4092, 0.3972, 0.4057, 0.4398, 0.4472, 0.4793, 0.458, 0.3967, 0.3761, 0.372]}
{"unit_id": "thread:1936383013321113890", "ticker_idx": 2, "ts": "2025-06-21T11:17:41+00:00", "symbol": "Samsung", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment", "tweet_type": "analysis", "timeline": "years", "summary": "US export restrictions will benefit Chinese domestic equipment makers; TSMC/Samsung/SK Hynix China fabs will transition to Chinese equipment.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung in semiconductor context = Samsung Electronics, 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "An industry source cited by Reuters pointed out that if it becomes more difficult to export U.S. equipment to foreign companies operating in China, it would essentially be a huge gift to Chinese domestic equipment suppliers.\n\nAnd that’s completely expected. At the end of the day, every company is driven by profit. You can’t just lift an entire fab like Aladdin’s lamp and move it to another country. So as long as these fabs were built in China, they’ll inevitably have to find ways to keep them running, regardless of U.S. sanctions.\n\nI believe that, in the end, the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment.\n\n---\n\nhttps://t.co/tBW57ynWJH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025862097916495408, "ret_signed_m1d": -0.025862097916495408, "alpha_spy_m1d": -0.03564296806676481, "alpha_c_m1d": -0.03713523974361421, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.011273141827118804, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043103473681822546, "ret_signed_p1d": -0.043103473681822546, "alpha_spy_p1d": -0.03205626468873479, "alpha_c_p1d": -0.025152234033618548, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.017951239648203998, "ret_p1w": 0.03735863299370279, "ret_signed_p1w": -0.03735863299370279, "alpha_spy_p1w": -0.00786597364623609, "alpha_c_p1w": 0.005254934377409981, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.04261356737111277, "ret_p1m": 0.14491088372511252, "ret_signed_p1m": -0.14491088372511252, "alpha_spy_p1m": -0.09707287499602058, "alpha_c_p1m": -0.07932291634925637, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.06558796737585615, "ret_p3m": 0.3929748331415097, "ret_signed_p3m": -0.3929748331415097, "alpha_spy_p3m": -0.2894840681229023, "alpha_c_p3m": -0.25611719453844706, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.13685763860306266, "ret_p6m": 0.7912417514852141, "ret_signed_p6m": -0.7912417514852141, "alpha_spy_p6m": -0.6569384506658937, "alpha_c_p6m": -0.6158200833730949, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.17542166811211923, "price_path": [0.0524, 0.0592, 0.0379, -0.0034, 0.0138, 0.0138, -0.0069, -0.0328, -0.0828, -0.0776, -0.0862, -0.0276, -0.0483, -0.031, -0.0241, -0.0569, -0.05, -0.0466, -0.0448, -0.0517, -0.0397, -0.0397, -0.0397, -0.0379, -0.0379, -0.0431, -0.0431, -0.0638, -0.0638, -0.0638, -0.0586, -0.0586, -0.0552, -0.0069, -0.019, -0.0103, -0.0121, -0.0207, -0.0379, -0.0362, -0.0397, -0.0569, -0.0655, -0.0569, -0.0707, -0.0362, -0.0328, -0.031, -0.0207, -0.0207, -0.0034, 0.019, 0.019, 0.031, 0.0207, 0.0328, 0.0259, 0.0052, -0.0138, 0.0017, 0.031, 0.0207, 0.0259, 0.0, 0.0431, 0.0569, 0.0379, 0.0547, 0.0374, 0.0443, 0.0547, 0.1067, 0.0981, 0.0703, 0.0651, 0.0478, 0.0582, 0.0859, 0.0842, 0.105, 0.1224, 0.1571, 0.164, 0.1761, 0.1449, 0.1518, 0.1449, 0.1432, 0.2212, 0.2247, 0.2594, 0.2386, 0.1952, 0.2091, 0.2126, 0.1935, 0.223, 0.2455, 0.2316, 0.2334, 0.2473, 0.2421, 0.2421, 0.2143, 0.2143, 0.223, 0.2247, 0.2386, 0.2403, 0.2195, 0.2247, 0.2074, 0.2091, 0.1727, 0.1987, 0.2108, 0.216, 0.2056, 0.216, 0.2403, 0.2594, 0.2733, 0.308, 0.3271, 0.3774, 0.3565, 0.393, 0.393, 0.4485, 0.4693, 0.4814, 0.4936, 0.445, 0.4671, 0.4619, 0.4985, 0.5639, 0.5639, 0.5639, 0.5639, 0.5639, 0.5639, 0.6449, 0.6257, 0.5961, 0.6553, 0.7024, 0.7059, 0.7093, 0.6989, 0.7181, 0.6815, 0.7215, 0.7773, 0.7337, 0.7512, 0.8139, 0.8731, 0.9359, 0.8278, 0.7529, 0.7285, 0.7059, 0.7529, 0.8034, 0.7965, 0.7912, 0.6937, 0.7529, 0.7041, 0.6815, 0.7529, 0.6518, 0.685, 0.7303, 0.7912, 0.8034, 0.7512, 0.7564, 0.8017, 0.8209, 0.8313, 0.8888, 0.908, 0.8888, 0.8818, 0.8697, 0.8975, 0.8261, 0.7912]}
{"unit_id": "thread:1936383013321113890", "ticker_idx": 3, "ts": "2025-06-21T11:17:41+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment", "tweet_type": "analysis", "timeline": "years", "summary": "US export restrictions will benefit Chinese domestic equipment makers; TSMC/Samsung/SK Hynix China fabs will transition to Chinese equipment.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "An industry source cited by Reuters pointed out that if it becomes more difficult to export U.S. equipment to foreign companies operating in China, it would essentially be a huge gift to Chinese domestic equipment suppliers.\n\nAnd that’s completely expected. At the end of the day, every company is driven by profit. You can’t just lift an entire fab like Aladdin’s lamp and move it to another country. So as long as these fabs were built in China, they’ll inevitably have to find ways to keep them running, regardless of U.S. sanctions.\n\nI believe that, in the end, the fabs operated by TSMC, Samsung, and SK Hynix in China will end up using Chinese-made semiconductor equipment.\n\n---\n\nhttps://t.co/tBW57ynWJH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009633960880562231, "ret_signed_m1d": 0.009633960880562231, "alpha_spy_m1d": -0.00014690926970717388, "alpha_c_m1d": 0.003334267520181955, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07321773991092995, "ret_signed_p1d": -0.07321773991092995, "alpha_spy_p1d": -0.06217053091784219, "alpha_c_p1d": -0.03648900821335066, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.12524094727529445, "ret_signed_p1w": -0.12524094727529445, "alpha_spy_p1w": -0.09574828792782775, "alpha_c_p1w": -0.060488397827961116, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": 0.034682017315795344, "ret_signed_p1m": -0.034682017315795344, "alpha_spy_p1m": 0.013155991413296597, "alpha_c_p1m": 0.05484913121998769, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.37192101427859003, "ret_signed_p3m": -0.37192101427859003, "alpha_spy_p3m": -0.2684302492599826, "alpha_c_p3m": -0.16174232271671962, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 1.046804114037171, "ret_signed_p6m": -1.046804114037171, "alpha_spy_p6m": -0.9125008132178507, "alpha_c_p6m": -0.7031114278058463, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [-0.1768, -0.2038, -0.2334, -0.2665, -0.2422, -0.2388, -0.2514, -0.2991, -0.3661, -0.348, -0.3653, -0.2953, -0.3045, -0.3068, -0.3053, -0.3307, -0.3268, -0.3268, -0.3207, -0.3315, -0.3038, -0.3141, -0.2907, -0.2999, -0.3045, -0.3172, -0.3172, -0.2845, -0.2845, -0.2845, -0.2661, -0.268, -0.2688, -0.2499, -0.2364, -0.2076, -0.2288, -0.2134, -0.233, -0.223, -0.2288, -0.2426, -0.2307, -0.2191, -0.2211, -0.1999, -0.183, -0.2119, -0.2004, -0.2004, -0.1618, -0.1349, -0.1349, -0.1175, -0.1118, -0.0751, -0.0925, -0.0925, -0.0443, -0.0405, -0.0501, -0.052, -0.0096, 0.0, 0.0732, 0.1021, 0.1291, 0.0944, 0.1252, 0.1002, 0.0751, 0.0732, 0.0424, 0.0443, 0.0867, 0.0829, 0.1445, 0.1349, 0.1561, 0.1503, 0.1407, 0.0385, 0.0366, 0.0501, 0.0347, 0.0366, 0.0385, 0.025, 0.0096, 0.0116, 0.0154, 0.0539, -0.0058, -0.0058, 0.0154, -0.0039, 0.0096, -0.0116, 0.0289, 0.0366, 0.0713, 0.0655, 0.0655, 0.0308, 0.0135, -0.0154, -0.0559, -0.0328, 0.0, 0.0077, 0.0019, 0.0362, 0.0381, -0.0121, 0.0053, 0.013, 0.0246, 0.0555, 0.069, 0.1114, 0.1732, 0.1848, 0.2677, 0.2774, 0.343, 0.287, 0.3719, 0.3719, 0.3546, 0.3931, 0.3796, 0.3758, 0.2986, 0.3468, 0.341, 0.3893, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.6517, 0.6015, 0.588, 0.6305, 0.7463, 0.7964, 0.8736, 0.8485, 0.8582, 0.8466, 0.9682, 1.0646, 1.0106, 1.1534, 1.192, 1.1573, 1.3927, 1.2615, 1.2344, 1.2885, 1.2383, 1.3386, 1.3888, 1.3811, 1.3618, 1.1611, 1.3386, 1.1997, 1.1688, 1.2036, 1.0106, 1.0067, 1.0029, 1.0222, 1.1009, 1.0468, 1.0777, 1.1549, 1.1318, 1.0931, 1.1009, 1.2283, 1.1858, 1.2669, 1.182, 1.2051, 1.1395, 1.0468]}
{"unit_id": "orig:1924825424515010661", "ticker_idx": 0, "ts": "2025-05-20T13:51:57+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This downward revision of 60–70 billion Gb, compared to the original plan, is viewed as a positive development in terms of short-term supply-demand dynamics", "tweet_type": "analysis", "timeline": "months", "summary": "CXMT's 2025 DRAM output revised down to ~240Gb from 300-310Gb on process delays; viewed as positive for near-term DRAM supply-demand.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities (Korea): CXMT’s 2025 Production Target Revised Downward, 15nm Process Delayed\n\nCompared to the beginning of the year, CXMT's production target is expected to be revised downward. While it was initially expected that CXMT's 2025 production would double year-on-year to reach around 300–310 billion Gb, recent expectations suggest a more modest growth of approximately 60% year-on-year, down to 240 billion Gb.\n\n- The demand landscape is shifting due to AI. The intrinsic technological characteristics of DRAM—low power consumption and high speed—are being further emphasized by AI, accelerating demand migration toward high-value products such as DDR5 and LPDDR5/5x. We believe demand from Chinese customers will not deviate from this trend.\n\n- In order to respond more swiftly to customer demand, CXMT likely needed to strengthen its capabilities in high-value product segments. During the process of shifting its production mix toward more advanced, higher-margin products—which are more difficult to mass-produce—the overall production target may have been reduced as a first consequence.\n\n- An additional constraint on production could be the delay in the development of the Gen 4b process. Gen 4b applies High-K Metal Gate (HKMG) technology to further improve speed over the existing Gen 4 process, and is essential for meeting high-end demand. While CXMT initially aimed to complete process development by the end of March 2025, this timeline has reportedly been pushed back to the end of September. Although there are attempts underway to pull the schedule forward again, given the difficulty in procuring equipment due to U.S. sanctions and the time required to optimize the technology, it is more reasonable to assume that Gen 4b’s contribution to mass production will be delayed compared to earlier expectations. If development is not completed by July or August, considering wafer lead times, volume production within the year will likely be limited.\n\nAs China continues to aggressively pursue semiconductor self-sufficiency, CXMT’s rise undeniably poses a strong threat to the DRAM industry. Given the Chinese government’s forceful push for higher self-sufficiency rates, there is also a clear tendency among Chinese customers to increase adoption of CXMT products. However, the product quality has not yet reached a mature level, and there are still real-world challenges to achieving stable mass production of high-value products. Taking this into account, it seems unlikely that short-term production will reach the earlier target of 300–310 billion Gb.\n\nConsidering various real-world constraints, CXMT’s 2025 production target is expected to be around 240 billion Gb. This downward revision of 60–70 billion Gb, compared to the original plan, is viewed as a positive development in terms of short-term supply-demand dynamics.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003017850541601752, "ret_signed_m1d": -0.003017850541601752, "alpha_spy_m1d": -0.006391320629805963, "alpha_c_m1d": -0.004932038064156097, "bench_spy_m1d": 0.0033734700882042112, "bench_c_m1d": 0.0019141875225543448, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011347141998016896, "ret_signed_p1d": -0.011347141998016896, "alpha_spy_p1d": 0.0055035472745231795, "alpha_c_p1d": 0.005228498892000832, "bench_spy_p1d": -0.016850689272540076, "bench_c_p1d": -0.01657564089001773, "ret_p1w": -0.016945372644636063, "ret_signed_p1w": -0.016945372644636063, "alpha_spy_p1w": -0.014077985549044308, "alpha_c_p1w": -0.014623998718417169, "bench_spy_p1w": -0.0028673870955917558, "bench_c_p1w": -0.002321373926218895, "ret_p1m": 0.17802083661966317, "ret_signed_p1m": 0.17802083661966317, "alpha_spy_p1m": 0.17027849575881393, "alpha_c_p1m": 0.1085820276286733, "bench_spy_p1m": 0.0077423408608492394, "bench_c_p1m": 0.06943880899098986, "ret_p3m": 0.29775628446498215, "ret_signed_p3m": 0.29775628446498215, "alpha_spy_p3m": 0.209214149202952, "alpha_c_p3m": 0.09367541894078535, "bench_spy_p3m": 0.08854213526203014, "bench_c_p3m": 0.2040808655241968, "ret_p6m": 1.4762319436427942, "ret_signed_p6m": 1.4762319436427942, "alpha_spy_p6m": 1.3169159762453686, "alpha_c_p6m": 1.028280458419166, "bench_spy_p6m": 0.1593159673974256, "bench_c_p6m": 0.44795148522362815, "price_path": [0.0439, 0.0239, -0.0005, -0.0159, -0.0001, -0.026, -0.0453, -0.0558, -0.0604, -0.0411, -0.0577, -0.0487, -0.0761, -0.0702, -0.0218, -0.0242, 0.0037, 0.0313, 0.0216, 0.0322, 0.0529, 0.0428, 0.0366, 0.0173, 0.0297, 0.0173, -0.012, -0.0454, -0.0229, -0.0218, -0.0828, -0.1449, -0.1785, -0.1786, -0.1471, -0.1234, -0.1362, -0.1262, -0.1231, -0.1511, -0.1489, -0.1477, -0.1515, -0.1467, -0.1214, -0.1106, -0.0925, -0.1, -0.1077, -0.1147, -0.1119, -0.095, -0.096, -0.0957, -0.0788, -0.0711, -0.0678, -0.0212, -0.0038, 0.0061, -0.0031, 0.0092, -0.003, 0.0, -0.0113, -0.0267, -0.0295, -0.0216, -0.0169, 0.0034, 0.0139, -0.0058, 0.0153, 0.0292, 0.0551, 0.0847, 0.0924, 0.1122, 0.1219, 0.1482, 0.1389, 0.1298, 0.1583, 0.167, 0.178, 0.1736, 0.1996, 0.1897, 0.2558, 0.2707, 0.2715, 0.2582, 0.2603, 0.2439, 0.2397, 0.2587, 0.2425, 0.226, 0.2577, 0.2427, 0.2757, 0.286, 0.2743, 0.2841, 0.2735, 0.2309, 0.2363, 0.2424, 0.2113, 0.2166, 0.2215, 0.2135, 0.2339, 0.2383, 0.2614, 0.2517, 0.1966, 0.2112, 0.2259, 0.2101, 0.2366, 0.2592, 0.2881, 0.3058, 0.3136, 0.3128, 0.2978, 0.2824, 0.2699, 0.244, 0.2224, 0.2436, 0.2539, 0.2503, 0.2539, 0.277, 0.2683, 0.2342, 0.2488, 0.2571, 0.2826, 0.3166, 0.3263, 0.3657, 0.415, 0.4607, 0.531, 0.5435, 0.5927, 0.5654, 0.6447, 0.6238, 0.6419, 0.6718, 0.6542, 0.6402, 0.5917, 0.6426, 0.65, 0.7338, 0.8206, 0.8349, 0.8455, 0.8276, 0.8645, 0.8502, 0.8955, 0.9053, 0.8699, 0.9252, 1.0272, 1.0494, 1.0986, 1.069, 1.0668, 1.0772, 1.1851, 1.2494, 1.2174, 1.3007, 1.33, 1.3348, 1.4947, 1.3443, 1.3731, 1.3906, 1.3599, 1.4715, 1.4691, 1.4762]}
{"unit_id": "orig:1939169224758567204", "ticker_idx": 0, "ts": "2025-06-29T03:49:05+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I project that SK Hynix will maintain its strength for the next month", "tweet_type": "analysis", "timeline": "months", "summary": "Short Hynix from Aug on Samsung HBM4 sample threat; short Hanmi on lost Micron/Hynix sockets; long BESI on expanding customer references.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "2025 H2 Memory Note \n\nTaker Jukan C\n\nSamsung reportedly changed its plan for Phase 4 (Ph4) of Pyeongtaek Campus Plant 4 (P4) from a foundry production line to a DRAM production line.\n\nThis seems consistent with Samsung's recent trend of reducing/delaying foundry investments. Recently, Korean media reported that Samsung decided to delay 1.4nm and focus intensively on 2nm investment and yield improvement.\n\nSamsung Electronics plans to install a 10nm-class 6th-generation (1c) DRAM production line in Ph4. The scale will be 80,000 wafers per month, which is 40% of P4's production capacity (200,000 12-inch wafers per month).\n\nSamsung was originally expected to apply HCB starting from HBM4, but it is now presumed that HCB application has been delayed to HBM4E.\n\nInstead, there's a possibility that Samsung might use Fluxless for HBM4. However, it seems unlikely that external companies like Hanmi or Hanwha Semitec, as expected by the market, will supply equipment to Samsung. Samsung is likely to procure most of the bonding equipment for HBM4 and HBM4E from its semiconductor equipment subsidiary, SEMES. BESI might supply some to Samsung, but BESI's goal seems to be to build references through collaboration with Samsung and then supply to SK Hynix and Micron.\n\nI don't think the market is seriously considering Samsung's HBM4 risk yet.\n\nSamsung has recently significantly improved the yield of its 1c DRAM itself, and the yield of 1c DRAM for HBM4 12-stack has also significantly increased compared to early this year.\n\nIn particular, according to supply chain sources, Samsung's 1c DRAM performance is definitely better than SK Hynix's 1b DRAM.\n\nThe reason NVIDIA has not yet finalized its HBM4 contract with SK Hynix is also understood to be due to Samsung's strides in HBM4.\n\nI project that SK Hynix will maintain its strength for the next month, but SK Hynix could weaken starting from August, when Samsung is expected to provide HBM4 samples.\n\nRegarding Hanmi Semiconductor:\n\nIt's almost certain that Micron will order Fluxless bonding equipment from BESI.\nThis is likely to be a considerable blow to Hanmi Semiconductor, which previously exclusively supplied HBM3E equipment to Micron.\n\nFurthermore, it's understood that even SK Hynix will increase the proportion of ASMPT and Hanwha Semitec in HBM4.\n\nTherefore, Hanmi Semiconductor is practically expected to have no other source of future revenue except for China-bound sales. Especially considering that its technology in hybrid bonding significantly lags behind competitors, placing expectations on Hanmi Semiconductor can be described as 'foolish'.\n\nThis note is not a buy/sell recommendation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027397398657104444, "ret_signed_m1d": -0.027397398657104444, "alpha_spy_m1d": -0.02263884459205001, "alpha_c_m1d": -0.02574800112189224, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0222603326750217, "ret_signed_p1d": -0.0222603326750217, "alpha_spy_p1d": -0.021936653366796843, "alpha_c_p1d": -0.011359555912624608, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.07191784907164522, "ret_signed_p1w": -0.07191784907164522, "alpha_spy_p1w": -0.07649817528738789, "alpha_c_p1w": -0.07507339485526976, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.10102747767706732, "ret_signed_p1m": -0.10102747767706732, "alpha_spy_p1m": -0.12920570911285345, "alpha_c_p1m": -0.1512999641652567, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.22265372912559833, "ret_signed_p3m": 0.22265372912559833, "alpha_spy_p3m": 0.1546366419664018, "alpha_c_p3m": 0.07140586318322062, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.0043232127011406, "ret_signed_p6m": 1.0043232127011406, "alpha_spy_p6m": 0.8844627424924387, "alpha_c_p6m": 0.6980623846192748, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.3235, -0.3348, -0.3772, -0.4366, -0.4206, -0.436, -0.3737, -0.3819, -0.384, -0.3826, -0.4052, -0.4018, -0.4018, -0.3963, -0.4059, -0.3813, -0.3905, -0.3696, -0.3778, -0.3819, -0.3932, -0.3932, -0.3642, -0.3642, -0.3642, -0.3478, -0.3495, -0.3501, -0.3334, -0.3214, -0.2958, -0.3146, -0.3009, -0.3184, -0.3095, -0.3146, -0.3269, -0.3163, -0.306, -0.3078, -0.289, -0.274, -0.2997, -0.2894, -0.2894, -0.2551, -0.2312, -0.2312, -0.2158, -0.2106, -0.1781, -0.1935, -0.1935, -0.1507, -0.1473, -0.1558, -0.1575, -0.1199, -0.1113, -0.0462, -0.0205, 0.0034, -0.0274, 0.0, -0.0223, -0.0445, -0.0462, -0.0736, -0.0719, -0.0342, -0.0377, 0.0171, 0.0086, 0.0274, 0.0223, 0.0137, -0.0771, -0.0788, -0.0668, -0.0805, -0.0788, -0.0771, -0.089, -0.1027, -0.101, -0.0976, -0.0634, -0.1164, -0.1164, -0.0976, -0.1147, -0.1027, -0.1216, -0.0856, -0.0788, -0.0479, -0.0531, -0.0531, -0.0839, -0.0993, -0.125, -0.161, -0.1404, -0.1113, -0.1045, -0.1096, -0.0792, -0.0774, -0.122, -0.1066, -0.0997, -0.0894, -0.062, -0.05, -0.0123, 0.0426, 0.0529, 0.1266, 0.1352, 0.1935, 0.1438, 0.2192, 0.2192, 0.2038, 0.2381, 0.2261, 0.2227, 0.1541, 0.1969, 0.1918, 0.2347, 0.3564, 0.3564, 0.3564, 0.3564, 0.3564, 0.3564, 0.4679, 0.4233, 0.4113, 0.449, 0.5519, 0.5965, 0.6651, 0.6428, 0.6514, 0.6411, 0.7491, 0.8348, 0.7868, 0.9137, 0.948, 0.9171, 1.1264, 1.0097, 0.9857, 1.0338, 0.9892, 1.0783, 1.1229, 1.1161, 1.0989, 0.9206, 1.0783, 0.9549, 0.9274, 0.9583, 0.7868, 0.7834, 0.78, 0.7971, 0.867, 0.819, 0.8464, 0.9151, 0.8945, 0.8602, 0.867, 0.9803, 0.9425, 1.0146, 0.9391, 0.9597, 0.9014, 0.819, 0.8911, 0.8945, 0.8773, 0.9906, 1.0043]}
{"unit_id": "orig:1939169224758567204", "ticker_idx": 1, "ts": "2025-06-29T03:49:05+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "SK Hynix could weaken starting from August, when Samsung is expected to provide HBM4 samples", "tweet_type": "analysis", "timeline": "months", "summary": "Short Hynix from Aug on Samsung HBM4 sample threat; short Hanmi on lost Micron/Hynix sockets; long BESI on expanding customer references.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "2025 H2 Memory Note \n\nTaker Jukan C\n\nSamsung reportedly changed its plan for Phase 4 (Ph4) of Pyeongtaek Campus Plant 4 (P4) from a foundry production line to a DRAM production line.\n\nThis seems consistent with Samsung's recent trend of reducing/delaying foundry investments. Recently, Korean media reported that Samsung decided to delay 1.4nm and focus intensively on 2nm investment and yield improvement.\n\nSamsung Electronics plans to install a 10nm-class 6th-generation (1c) DRAM production line in Ph4. The scale will be 80,000 wafers per month, which is 40% of P4's production capacity (200,000 12-inch wafers per month).\n\nSamsung was originally expected to apply HCB starting from HBM4, but it is now presumed that HCB application has been delayed to HBM4E.\n\nInstead, there's a possibility that Samsung might use Fluxless for HBM4. However, it seems unlikely that external companies like Hanmi or Hanwha Semitec, as expected by the market, will supply equipment to Samsung. Samsung is likely to procure most of the bonding equipment for HBM4 and HBM4E from its semiconductor equipment subsidiary, SEMES. BESI might supply some to Samsung, but BESI's goal seems to be to build references through collaboration with Samsung and then supply to SK Hynix and Micron.\n\nI don't think the market is seriously considering Samsung's HBM4 risk yet.\n\nSamsung has recently significantly improved the yield of its 1c DRAM itself, and the yield of 1c DRAM for HBM4 12-stack has also significantly increased compared to early this year.\n\nIn particular, according to supply chain sources, Samsung's 1c DRAM performance is definitely better than SK Hynix's 1b DRAM.\n\nThe reason NVIDIA has not yet finalized its HBM4 contract with SK Hynix is also understood to be due to Samsung's strides in HBM4.\n\nI project that SK Hynix will maintain its strength for the next month, but SK Hynix could weaken starting from August, when Samsung is expected to provide HBM4 samples.\n\nRegarding Hanmi Semiconductor:\n\nIt's almost certain that Micron will order Fluxless bonding equipment from BESI.\nThis is likely to be a considerable blow to Hanmi Semiconductor, which previously exclusively supplied HBM3E equipment to Micron.\n\nFurthermore, it's understood that even SK Hynix will increase the proportion of ASMPT and Hanwha Semitec in HBM4.\n\nTherefore, Hanmi Semiconductor is practically expected to have no other source of future revenue except for China-bound sales. Especially considering that its technology in hybrid bonding significantly lags behind competitors, placing expectations on Hanmi Semiconductor can be described as 'foolish'.\n\nThis note is not a buy/sell recommendation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027397398657104444, "ret_signed_m1d": 0.027397398657104444, "alpha_spy_m1d": 0.02263884459205001, "alpha_c_m1d": 0.02574800112189224, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0222603326750217, "ret_signed_p1d": 0.0222603326750217, "alpha_spy_p1d": 0.021936653366796843, "alpha_c_p1d": 0.011359555912624608, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.07191784907164522, "ret_signed_p1w": 0.07191784907164522, "alpha_spy_p1w": 0.07649817528738789, "alpha_c_p1w": 0.07507339485526976, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.10102747767706732, "ret_signed_p1m": 0.10102747767706732, "alpha_spy_p1m": 0.12920570911285345, "alpha_c_p1m": 0.1512999641652567, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.22265372912559833, "ret_signed_p3m": -0.22265372912559833, "alpha_spy_p3m": -0.1546366419664018, "alpha_c_p3m": -0.07140586318322062, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 1.0043232127011406, "ret_signed_p6m": -1.0043232127011406, "alpha_spy_p6m": -0.8844627424924387, "alpha_c_p6m": -0.6980623846192748, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.3235, -0.3348, -0.3772, -0.4366, -0.4206, -0.436, -0.3737, -0.3819, -0.384, -0.3826, -0.4052, -0.4018, -0.4018, -0.3963, -0.4059, -0.3813, -0.3905, -0.3696, -0.3778, -0.3819, -0.3932, -0.3932, -0.3642, -0.3642, -0.3642, -0.3478, -0.3495, -0.3501, -0.3334, -0.3214, -0.2958, -0.3146, -0.3009, -0.3184, -0.3095, -0.3146, -0.3269, -0.3163, -0.306, -0.3078, -0.289, -0.274, -0.2997, -0.2894, -0.2894, -0.2551, -0.2312, -0.2312, -0.2158, -0.2106, -0.1781, -0.1935, -0.1935, -0.1507, -0.1473, -0.1558, -0.1575, -0.1199, -0.1113, -0.0462, -0.0205, 0.0034, -0.0274, 0.0, -0.0223, -0.0445, -0.0462, -0.0736, -0.0719, -0.0342, -0.0377, 0.0171, 0.0086, 0.0274, 0.0223, 0.0137, -0.0771, -0.0788, -0.0668, -0.0805, -0.0788, -0.0771, -0.089, -0.1027, -0.101, -0.0976, -0.0634, -0.1164, -0.1164, -0.0976, -0.1147, -0.1027, -0.1216, -0.0856, -0.0788, -0.0479, -0.0531, -0.0531, -0.0839, -0.0993, -0.125, -0.161, -0.1404, -0.1113, -0.1045, -0.1096, -0.0792, -0.0774, -0.122, -0.1066, -0.0997, -0.0894, -0.062, -0.05, -0.0123, 0.0426, 0.0529, 0.1266, 0.1352, 0.1935, 0.1438, 0.2192, 0.2192, 0.2038, 0.2381, 0.2261, 0.2227, 0.1541, 0.1969, 0.1918, 0.2347, 0.3564, 0.3564, 0.3564, 0.3564, 0.3564, 0.3564, 0.4679, 0.4233, 0.4113, 0.449, 0.5519, 0.5965, 0.6651, 0.6428, 0.6514, 0.6411, 0.7491, 0.8348, 0.7868, 0.9137, 0.948, 0.9171, 1.1264, 1.0097, 0.9857, 1.0338, 0.9892, 1.0783, 1.1229, 1.1161, 1.0989, 0.9206, 1.0783, 0.9549, 0.9274, 0.9583, 0.7868, 0.7834, 0.78, 0.7971, 0.867, 0.819, 0.8464, 0.9151, 0.8945, 0.8602, 0.867, 0.9803, 0.9425, 1.0146, 0.9391, 0.9597, 0.9014, 0.819, 0.8911, 0.8945, 0.8773, 0.9906, 1.0043]}
{"unit_id": "orig:1939169224758567204", "ticker_idx": 3, "ts": "2025-06-29T03:49:05+00:00", "symbol": "BESI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BESI's goal seems to be to build references through collaboration with Samsung and then supply to SK Hynix and Micron", "tweet_type": "analysis", "timeline": "months", "summary": "Short Hynix from Aug on Samsung HBM4 sample threat; short Hanmi on lost Micron/Hynix sockets; long BESI on expanding customer references.", "resolved_tickers": ["BESI.AS"], "resolution_method": "exchange_ticker", "resolution_reason": "BE Semiconductor Industries Amsterdam BESI.AS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "2025 H2 Memory Note \n\nTaker Jukan C\n\nSamsung reportedly changed its plan for Phase 4 (Ph4) of Pyeongtaek Campus Plant 4 (P4) from a foundry production line to a DRAM production line.\n\nThis seems consistent with Samsung's recent trend of reducing/delaying foundry investments. Recently, Korean media reported that Samsung decided to delay 1.4nm and focus intensively on 2nm investment and yield improvement.\n\nSamsung Electronics plans to install a 10nm-class 6th-generation (1c) DRAM production line in Ph4. The scale will be 80,000 wafers per month, which is 40% of P4's production capacity (200,000 12-inch wafers per month).\n\nSamsung was originally expected to apply HCB starting from HBM4, but it is now presumed that HCB application has been delayed to HBM4E.\n\nInstead, there's a possibility that Samsung might use Fluxless for HBM4. However, it seems unlikely that external companies like Hanmi or Hanwha Semitec, as expected by the market, will supply equipment to Samsung. Samsung is likely to procure most of the bonding equipment for HBM4 and HBM4E from its semiconductor equipment subsidiary, SEMES. BESI might supply some to Samsung, but BESI's goal seems to be to build references through collaboration with Samsung and then supply to SK Hynix and Micron.\n\nI don't think the market is seriously considering Samsung's HBM4 risk yet.\n\nSamsung has recently significantly improved the yield of its 1c DRAM itself, and the yield of 1c DRAM for HBM4 12-stack has also significantly increased compared to early this year.\n\nIn particular, according to supply chain sources, Samsung's 1c DRAM performance is definitely better than SK Hynix's 1b DRAM.\n\nThe reason NVIDIA has not yet finalized its HBM4 contract with SK Hynix is also understood to be due to Samsung's strides in HBM4.\n\nI project that SK Hynix will maintain its strength for the next month, but SK Hynix could weaken starting from August, when Samsung is expected to provide HBM4 samples.\n\nRegarding Hanmi Semiconductor:\n\nIt's almost certain that Micron will order Fluxless bonding equipment from BESI.\nThis is likely to be a considerable blow to Hanmi Semiconductor, which previously exclusively supplied HBM3E equipment to Micron.\n\nFurthermore, it's understood that even SK Hynix will increase the proportion of ASMPT and Hanwha Semitec in HBM4.\n\nTherefore, Hanmi Semiconductor is practically expected to have no other source of future revenue except for China-bound sales. Especially considering that its technology in hybrid bonding significantly lags behind competitors, placing expectations on Hanmi Semiconductor can be described as 'foolish'.\n\nThis note is not a buy/sell recommendation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026761123496687356, "ret_signed_m1d": 0.026761123496687356, "alpha_spy_m1d": 0.03151967756174179, "alpha_c_m1d": 0.02841052103189956, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.038567481959402516, "ret_signed_p1d": -0.038567481959402516, "alpha_spy_p1d": -0.03824380265117766, "alpha_c_p1d": -0.027666705197005426, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": -0.04447066119076015, "ret_signed_p1w": -0.04447066119076015, "alpha_spy_p1w": -0.04905098740650282, "alpha_c_p1w": -0.04762620697438469, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": -0.055489993546796845, "ret_signed_p1m": -0.055489993546796845, "alpha_spy_p1m": -0.08366822498258297, "alpha_c_p1m": -0.10576248003498623, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": -0.0027548935179179113, "ret_signed_p3m": -0.0027548935179179113, "alpha_spy_p3m": -0.07077198067711443, "alpha_c_p3m": -0.15400275946029562, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.03817384804824586, "ret_signed_p6m": 0.03817384804824586, "alpha_spy_p6m": -0.08168662216045597, "alpha_c_p6m": -0.26808698003361986, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2501, -0.3085, -0.3167, -0.3473, -0.3397, -0.3876, -0.3645, -0.3601, -0.339, -0.251, -0.275, -0.2771, -0.2771, -0.2771, -0.2833, -0.2773, -0.2636, -0.2362, -0.2309, -0.2239, -0.2582, -0.2582, -0.2153, -0.2121, -0.222, -0.1814, -0.1649, -0.1507, -0.0948, -0.0657, -0.0543, -0.0909, -0.1082, -0.1275, -0.146, -0.1433, -0.1492, -0.1712, -0.1539, -0.1385, -0.1429, -0.1303, -0.1614, -0.1818, -0.1598, -0.1503, -0.1295, -0.1059, -0.096, -0.0531, -0.0677, -0.0338, -0.0091, -0.0004, -0.0016, -0.0102, -0.0185, -0.0307, -0.0106, 0.0146, 0.0169, 0.0216, 0.0268, 0.0, -0.0386, -0.0299, -0.0401, -0.0472, -0.0445, -0.0382, -0.0401, 0.0028, 0.0008, -0.0189, -0.0059, -0.0559, 0.0035, 0.0071, 0.0244, -0.0098, -0.0268, -0.0004, -0.0783, -0.0335, -0.0555, -0.0366, -0.0626, -0.0779, -0.0681, -0.0925, -0.0964, -0.0559, -0.0496, -0.0433, -0.026, -0.0279, -0.0346, -0.0669, -0.0728, -0.0602, -0.0783, -0.0913, -0.0634, -0.0622, -0.05, -0.0571, -0.0575, -0.0941, -0.0909, -0.1586, -0.1673, -0.1566, -0.1444, -0.1275, -0.1228, -0.1157, -0.1055, -0.1196, -0.0701, -0.0823, -0.0897, -0.0177, -0.0437, -0.0335, -0.0012, 0.0118, -0.0028, -0.0366, -0.0051, -0.0016, 0.0059, 0.0512, 0.0272, 0.1551, 0.1555, 0.1519, 0.1444, 0.1043, 0.1236, 0.1059, 0.1452, 0.1421, 0.1153, 0.1429, 0.1358, 0.0901, 0.1354, 0.1574, 0.1649, 0.1539, 0.1621, 0.1574, 0.1617, 0.1476, 0.122, 0.1137, 0.0842, 0.0559, 0.074, 0.0846, 0.0653, 0.0531, 0.0342, 0.0248, 0.0008, 0.0161, 0.0236, -0.0335, -0.0209, -0.0193, 0.0201, 0.0224, 0.0224, 0.0213, 0.0338, 0.0701, 0.0823, 0.0984, 0.1558, 0.1279, 0.1055, 0.0775, 0.0397, 0.0425, 0.0516, 0.0177, 0.0346, 0.0291, 0.0327, 0.0382]}
{"unit_id": "orig:1924280667539673104", "ticker_idx": 0, "ts": "2025-05-19T01:47:17+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Positive for Samsung Electronics and SK Hynix", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays KB Securities call: HBM4 price premium ~30% over HBM3E positive for both Samsung and SK Hynix into 2H 2026 mainstream ramp.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Analysis: South Korea KB Securities: “HBM4 Price Premium Expected to Reach 30%, Positive for Samsung Electronics and SK Hynix”\n\nThe price of sixth-generation High Bandwidth Memory (HBM4) is expected to increase by around 30% compared to the previous generation.\n\nDongwon Kim, an analyst at KB Securities, stated on the 19th, “HBM4 is expected to see a 30% price increase compared to its predecessor, HBM3E,” adding, “The price hike is due to increased manufacturing costs. While the wafer cost for existing 1b DRAM memory is about $2,000 to $2,500, the logic wafer for HBM4 produced at TSMC is estimated at $7,000 to $8,000—three to four times higher.”\n\nFurthermore, the increase in I/O channels (from 1,024 to 2,048) is expected to raise the complexity of bump and stacking processes, and the HBM4 die size is projected to be 15–20% larger than that of HBM3E.\n\nAccordingly, while the average selling price (ASP) rose by 20% at the time of HBM3E’s launch, HBM4 is expected to see a price increase of more than 30%.\n\nTotal annual HBM shipments are expected to exceed 3,000 billion gigabytes (Gb) in 2026, with HBM4 anticipated to replace HBM3E and become the mainstream product in the second half of 2026.\n\nhttps://t.co/2iZfiAPTit", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01792099860228724, "ret_signed_m1d": 0.01792099860228724, "alpha_spy_m1d": 0.019013721353129842, "alpha_c_m1d": 0.016471520699595565, "bench_spy_m1d": -0.0010927227508426007, "bench_c_m1d": 0.001449477902691676, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0017920214971340886, "ret_signed_p1d": 0.0017920214971340886, "alpha_spy_p1d": 0.0051541495469296095, "alpha_c_p1d": 0.005544019321804927, "bench_spy_p1d": -0.003362128049795521, "bench_c_p1d": -0.0037519978246708385, "ret_p1w": -0.01971337629530634, "ret_signed_p1w": -0.01971337629530634, "alpha_spy_p1w": 0.006747036400826412, "alpha_c_p1w": 0.013286028839067043, "bench_spy_p1w": -0.02646041269613275, "bench_c_p1w": -0.03299940513437338, "ret_p1m": 0.04121848913103876, "ret_signed_p1m": 0.04121848913103876, "alpha_spy_p1m": 0.036712992618328055, "alpha_c_p1m": 0.012354747905175989, "bench_spy_p1m": 0.004505496512710705, "bench_c_p1m": 0.02886374122586277, "ret_p3m": 0.2910245760912773, "ret_signed_p3m": 0.2910245760912773, "alpha_spy_p3m": 0.2035963778041816, "alpha_c_p3m": 0.14767033478834102, "bench_spy_p3m": 0.0874281982870957, "bench_c_p3m": 0.14335424130293628, "ret_p6m": 0.8745419677115553, "ret_signed_p6m": 0.8745419677115553, "alpha_spy_p6m": 0.7197662846275137, "alpha_c_p6m": 0.6216201777548656, "bench_spy_p6m": 0.15477568308404166, "bench_c_p6m": 0.2529217899566898, "price_path": [0.0458, 0.0404, 0.0369, 0.0208, 0.019, 0.0083, 0.003, -0.0291, -0.0291, -0.0291, -0.038, -0.0326, -0.0433, -0.0433, -0.0451, -0.0219, -0.0255, -0.0255, 0.0262, 0.0262, 0.0422, 0.0725, 0.0992, 0.0778, 0.0654, 0.0939, 0.101, 0.0789, 0.0358, 0.0538, 0.0538, 0.0323, 0.0054, -0.0466, -0.0412, -0.0502, 0.0108, -0.0108, 0.0072, 0.0143, -0.0197, -0.0125, -0.009, -0.0072, -0.0143, -0.0018, -0.0018, -0.0018, 0.0, 0.0, -0.0054, -0.0054, -0.0269, -0.0269, -0.0269, -0.0215, -0.0215, -0.0179, 0.0323, 0.0197, 0.0287, 0.0269, 0.0179, 0.0, 0.0018, -0.0018, -0.0197, -0.0287, -0.0197, -0.0341, 0.0018, 0.0054, 0.0072, 0.0179, 0.0179, 0.0358, 0.0591, 0.0591, 0.0717, 0.0609, 0.0735, 0.0663, 0.0448, 0.0251, 0.0412, 0.0717, 0.0609, 0.0663, 0.0394, 0.0842, 0.0986, 0.0789, 0.0963, 0.0783, 0.0855, 0.0963, 0.1504, 0.1414, 0.1125, 0.1071, 0.0891, 0.0999, 0.1287, 0.1269, 0.1486, 0.1666, 0.2027, 0.2099, 0.2225, 0.1901, 0.1973, 0.1901, 0.1882, 0.2694, 0.273, 0.3091, 0.2874, 0.2423, 0.2568, 0.2604, 0.2405, 0.2712, 0.2946, 0.2802, 0.282, 0.2964, 0.291, 0.291, 0.2622, 0.2622, 0.2712, 0.273, 0.2874, 0.2892, 0.2676, 0.273, 0.255, 0.2568, 0.2189, 0.2459, 0.2586, 0.264, 0.2532, 0.264, 0.2892, 0.3091, 0.3235, 0.3595, 0.3794, 0.4317, 0.41, 0.4479, 0.4479, 0.5056, 0.5272, 0.5399, 0.5525, 0.502, 0.525, 0.5196, 0.5576, 0.6255, 0.6255, 0.6255, 0.6255, 0.6255, 0.6255, 0.7097, 0.6898, 0.659, 0.7206, 0.7695, 0.7731, 0.7767, 0.7659, 0.7858, 0.7478, 0.7894, 0.8474, 0.8021, 0.8202, 0.8854, 0.947, 1.0122, 0.8999, 0.822, 0.7967, 0.7731, 0.822, 0.8745]}
{"unit_id": "orig:1924280667539673104", "ticker_idx": 1, "ts": "2025-05-19T01:47:17+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Positive for Samsung Electronics and SK Hynix", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays KB Securities call: HBM4 price premium ~30% over HBM3E positive for both Samsung and SK Hynix into 2H 2026 mainstream ramp.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Analysis: South Korea KB Securities: “HBM4 Price Premium Expected to Reach 30%, Positive for Samsung Electronics and SK Hynix”\n\nThe price of sixth-generation High Bandwidth Memory (HBM4) is expected to increase by around 30% compared to the previous generation.\n\nDongwon Kim, an analyst at KB Securities, stated on the 19th, “HBM4 is expected to see a 30% price increase compared to its predecessor, HBM3E,” adding, “The price hike is due to increased manufacturing costs. While the wafer cost for existing 1b DRAM memory is about $2,000 to $2,500, the logic wafer for HBM4 produced at TSMC is estimated at $7,000 to $8,000—three to four times higher.”\n\nFurthermore, the increase in I/O channels (from 1,024 to 2,048) is expected to raise the complexity of bump and stacking processes, and the HBM4 die size is projected to be 15–20% larger than that of HBM3E.\n\nAccordingly, while the average selling price (ASP) rose by 20% at the time of HBM3E’s launch, HBM4 is expected to see a price increase of more than 30%.\n\nTotal annual HBM shipments are expected to exceed 3,000 billion gigabytes (Gb) in 2026, with HBM4 anticipated to replace HBM3E and become the mainstream product in the second half of 2026.\n\nhttps://t.co/2iZfiAPTit", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025576751182819102, "ret_signed_m1d": 0.025576751182819102, "alpha_spy_m1d": 0.026669473933661703, "alpha_c_m1d": 0.02391011562518952, "bench_spy_m1d": -0.0010927227508426007, "bench_c_m1d": 0.001666635557629581, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013039132691024191, "ret_signed_p1d": 0.013039132691024191, "alpha_spy_p1d": 0.016401260740819712, "alpha_c_p1d": 0.014949663100108035, "bench_spy_p1d": -0.003362128049795521, "bench_c_p1d": -0.0019105304090838438, "ret_p1w": 0.018054117024026484, "ret_signed_p1w": 0.018054117024026484, "alpha_spy_p1w": 0.044514529720159235, "alpha_c_p1w": 0.052890079071001184, "bench_spy_p1w": -0.02646041269613275, "bench_c_p1w": -0.0348359620469747, "ret_p1m": 0.25100162917226587, "ret_signed_p1m": 0.25100162917226587, "alpha_spy_p1m": 0.24649613265955517, "alpha_c_p1m": 0.1876708467799808, "bench_spy_p1m": 0.004505496512710705, "bench_c_p1m": 0.06333078239228507, "ret_p3m": 0.3891642189405091, "ret_signed_p3m": 0.3891642189405091, "alpha_spy_p3m": 0.30173602065341343, "alpha_c_p3m": 0.1618970234202104, "bench_spy_p3m": 0.0874281982870957, "bench_c_p3m": 0.22726719552029873, "ret_p6m": 2.1144113754961045, "ret_signed_p6m": 2.1144113754961045, "alpha_spy_p6m": 1.9596356924120628, "alpha_c_p6m": 1.687436853797551, "bench_spy_p6m": 0.15477568308404166, "bench_c_p6m": 0.42697452169855343, "price_path": [0.0887, 0.0564, 0.0439, 0.0215, -0.0009, 0.0115, -0.001, -0.0461, -0.0461, -0.0667, -0.0316, -0.0351, -0.0351, -0.0577, -0.0582, -0.0025, 0.0015, 0.0256, 0.0331, 0.0181, 0.0306, 0.0532, 0.0807, 0.0607, 0.0431, 0.0732, 0.0381, -0.0005, -0.0436, -0.012, -0.0075, -0.0241, -0.0863, -0.1735, -0.1499, -0.1725, -0.0812, -0.0933, -0.0963, -0.0943, -0.1274, -0.1224, -0.1224, -0.1143, -0.1284, -0.0923, -0.1058, -0.0752, -0.0873, -0.0933, -0.1098, -0.1098, -0.0672, -0.0672, -0.0672, -0.0431, -0.0456, -0.0466, -0.0221, -0.0045, 0.0331, 0.0055, 0.0256, 0.0, 0.013, 0.0055, -0.0125, 0.003, 0.0181, 0.0155, 0.0431, 0.0651, 0.0274, 0.0425, 0.0425, 0.0927, 0.1279, 0.1279, 0.1505, 0.1581, 0.2058, 0.1832, 0.1832, 0.246, 0.251, 0.2384, 0.2359, 0.2912, 0.3038, 0.3992, 0.4369, 0.4721, 0.4268, 0.467, 0.4344, 0.4017, 0.3992, 0.359, 0.3615, 0.4168, 0.4118, 0.4922, 0.4796, 0.5072, 0.4997, 0.4871, 0.354, 0.3515, 0.3691, 0.349, 0.3515, 0.354, 0.3364, 0.3163, 0.3188, 0.3239, 0.3741, 0.2962, 0.2962, 0.3239, 0.2987, 0.3163, 0.2887, 0.3414, 0.3515, 0.3967, 0.3892, 0.3892, 0.3439, 0.3213, 0.2837, 0.2309, 0.261, 0.3038, 0.3138, 0.3063, 0.3509, 0.3534, 0.288, 0.3107, 0.3207, 0.3358, 0.3761, 0.3937, 0.449, 0.5295, 0.5446, 0.6528, 0.6654, 0.7509, 0.678, 0.7886, 0.7886, 0.766, 0.8163, 0.7987, 0.7937, 0.6931, 0.7559, 0.7484, 0.8113, 0.9899, 0.9899, 0.9899, 0.9899, 0.9899, 0.9899, 1.1534, 1.088, 1.0704, 1.1257, 1.2767, 1.3421, 1.4427, 1.41, 1.4226, 1.4075, 1.566, 1.6918, 1.6213, 1.8075, 1.8578, 1.8125, 2.1194, 1.9484, 1.9132, 1.9836, 1.9182, 2.049, 2.1144]}
{"unit_id": "thread:1927656927602868352", "ticker_idx": 0, "ts": "2025-05-28T09:23:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA is expected to release a low-power, downscaled version of the RTX PRO 6000 (previously known as B40) targeting the Chinese market", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVIDIA China RTX PRO 6000 to use GDDR7, boosting future demand; Samsung projected to capture 70% of GDDR7 supply, prices stable-to-rising.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶️ NVIDIA Expected to Launch RTX PRO 6000 Special Edition for China’s AI Market, Potentially Boosting Future GDDR7 Demand, Says TrendForce\n\n• According to TrendForce, in response to the new U.S. export restrictions announced in April, NVIDIA is expected to release a low-power, downscaled version of the RTX PRO 6000 (previously known as B40) targeting the Chinese market.\n\n• This model will reportedly switch from the originally planned HBM memory to GDDR7 and is expected to launch as early as the second half of 2025.\n\n• Based on recent interviews with Chinese industry insiders, TrendForce reports that following the implementation of the new export rules, Chinese CSP customers have not been able to obtain the H20, and the B30 model—initially planned for release in the second half of this year—is now likely to be delayed or altered.\n\n• TrendForce predicts that the performance of the China-specific RTX PRO 6000 will fall between that of the previous-generation L40S and the L20 China Edition.\n\n• Demand for the L20 has remained steady among local CSPs, especially for applications such as small-scale model training and AI inference, which suggests that the RTX PRO 6000 China Edition is also expected to achieve similar sales levels upon release.\n\n• However, with high-end models such as the H20 and B30—equipped with HBM—blocked from the Chinese market, NVIDIA now faces intense competition from domestic players like Huawei and Cambricon, who are developing their own AI solutions.\n\n• To comply with U.S. export controls, the RTX PRO 6000 for the Chinese market will be downgraded from HBM3e to GDDR7 memory.\n\n• Major memory suppliers have begun expanding GDDR7 production following the recent launch of NVIDIA’s RTX 50 series consumer GPUs.\n\n• However, since gaming consoles and other consumer applications have yet to adopt GDDR7, the primary demand through the first half of 2026 is expected to come from NVIDIA-driven PC, workstation GPU, and AI server applications.\n\n• Samsung is projected to lead GDDR7 production throughout 2025 and could capture up to 70% of the market share.\n\n• NVIDIA remains heavily reliant on GDDR7 produced by Samsung. TrendForce expects GDDR7 pricing to remain relatively stable going forward, with prices holding steady or increasing slightly.\n\n---\n\nhttps://t.co/746a6ns215", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "thread:1927656927602868352", "ticker_idx": 1, "ts": "2025-05-28T09:23:20+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung is projected to lead GDDR7 production throughout 2025 and could capture up to 70% of the market share", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVIDIA China RTX PRO 6000 to use GDDR7, boosting future demand; Samsung projected to capture 70% of GDDR7 supply, prices stable-to-rising.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶️ NVIDIA Expected to Launch RTX PRO 6000 Special Edition for China’s AI Market, Potentially Boosting Future GDDR7 Demand, Says TrendForce\n\n• According to TrendForce, in response to the new U.S. export restrictions announced in April, NVIDIA is expected to release a low-power, downscaled version of the RTX PRO 6000 (previously known as B40) targeting the Chinese market.\n\n• This model will reportedly switch from the originally planned HBM memory to GDDR7 and is expected to launch as early as the second half of 2025.\n\n• Based on recent interviews with Chinese industry insiders, TrendForce reports that following the implementation of the new export rules, Chinese CSP customers have not been able to obtain the H20, and the B30 model—initially planned for release in the second half of this year—is now likely to be delayed or altered.\n\n• TrendForce predicts that the performance of the China-specific RTX PRO 6000 will fall between that of the previous-generation L40S and the L20 China Edition.\n\n• Demand for the L20 has remained steady among local CSPs, especially for applications such as small-scale model training and AI inference, which suggests that the RTX PRO 6000 China Edition is also expected to achieve similar sales levels upon release.\n\n• However, with high-end models such as the H20 and B30—equipped with HBM—blocked from the Chinese market, NVIDIA now faces intense competition from domestic players like Huawei and Cambricon, who are developing their own AI solutions.\n\n• To comply with U.S. export controls, the RTX PRO 6000 for the Chinese market will be downgraded from HBM3e to GDDR7 memory.\n\n• Major memory suppliers have begun expanding GDDR7 production following the recent launch of NVIDIA’s RTX 50 series consumer GPUs.\n\n• However, since gaming consoles and other consumer applications have yet to adopt GDDR7, the primary demand through the first half of 2026 is expected to come from NVIDIA-driven PC, workstation GPU, and AI server applications.\n\n• Samsung is projected to lead GDDR7 production throughout 2025 and could capture up to 70% of the market share.\n\n• NVIDIA remains heavily reliant on GDDR7 produced by Samsung. TrendForce expects GDDR7 pricing to remain relatively stable going forward, with prices holding steady or increasing slightly.\n\n---\n\nhttps://t.co/746a6ns215", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03577816691703928, "ret_signed_m1d": -0.03577816691703928, "alpha_spy_m1d": -0.04159724718534774, "alpha_c_m1d": -0.039929694040201524, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.004151527123162246, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003577774024658087, "ret_signed_p1d": 0.003577774024658087, "alpha_spy_p1d": -0.00036950331613816445, "alpha_c_p1d": 0.00189120161376799, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.001686572410890097, "ret_p1w": 0.03398920879296696, "ret_signed_p1w": 0.03398920879296696, "alpha_spy_p1w": 0.020037344379564503, "alpha_c_p1w": 0.009944834166305538, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.024044374626661424, "ret_p1m": 0.07692313709455201, "ret_signed_p1m": 0.07692313709455201, "alpha_spy_p1m": 0.032772221880607244, "alpha_c_p1m": -0.010631433119161704, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.08755457021371371, "ret_p3m": 0.2869153325807341, "ret_signed_p3m": 0.2869153325807341, "alpha_spy_p3m": 0.1905458424642632, "alpha_c_p3m": 0.152751994672494, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.1341633379082401, "ret_p6m": 0.8187593489719909, "ret_signed_p6m": 0.8187593489719909, "alpha_spy_p6m": 0.7021352075250327, "alpha_c_p6m": 0.6383779277078925, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.1803814212640984, "price_path": [-0.0308, -0.0308, -0.0308, -0.0397, -0.0344, -0.045, -0.045, -0.0468, -0.0237, -0.0272, -0.0272, 0.0243, 0.0243, 0.0403, 0.0706, 0.0972, 0.0759, 0.0634, 0.0919, 0.099, 0.0769, 0.034, 0.0519, 0.0519, 0.0304, 0.0036, -0.0483, -0.0429, -0.0519, 0.0089, -0.0125, 0.0054, 0.0125, -0.0215, -0.0143, -0.0107, -0.0089, -0.0161, -0.0036, -0.0036, -0.0036, -0.0018, -0.0018, -0.0072, -0.0072, -0.0286, -0.0286, -0.0286, -0.0233, -0.0233, -0.0197, 0.0304, 0.0179, 0.0268, 0.025, 0.0161, -0.0018, 0.0, -0.0036, -0.0215, -0.0304, -0.0215, -0.0358, 0.0, 0.0036, 0.0054, 0.0161, 0.0161, 0.034, 0.0572, 0.0572, 0.0698, 0.059, 0.0716, 0.0644, 0.0429, 0.0233, 0.0394, 0.0698, 0.059, 0.0644, 0.0376, 0.0823, 0.0966, 0.0769, 0.0943, 0.0763, 0.0835, 0.0943, 0.1483, 0.1393, 0.1105, 0.1051, 0.0871, 0.0979, 0.1267, 0.1249, 0.1465, 0.1645, 0.2005, 0.2077, 0.2203, 0.1879, 0.1951, 0.1879, 0.1861, 0.2671, 0.2707, 0.3067, 0.2851, 0.2401, 0.2545, 0.2581, 0.2383, 0.2689, 0.2923, 0.2779, 0.2797, 0.2941, 0.2887, 0.2887, 0.2599, 0.2599, 0.2689, 0.2707, 0.2851, 0.2869, 0.2653, 0.2707, 0.2527, 0.2545, 0.2167, 0.2437, 0.2563, 0.2617, 0.2509, 0.2617, 0.2869, 0.3067, 0.3211, 0.3571, 0.3769, 0.4291, 0.4075, 0.4453, 0.4453, 0.5029, 0.5245, 0.5371, 0.5497, 0.4993, 0.5223, 0.5168, 0.5548, 0.6226, 0.6226, 0.6226, 0.6226, 0.6226, 0.6226, 0.7067, 0.6868, 0.656, 0.7175, 0.7663, 0.7699, 0.7736, 0.7627, 0.7826, 0.7446, 0.7862, 0.8441, 0.7989, 0.817, 0.882, 0.9435, 1.0086, 0.8965, 0.8188, 0.7934, 0.7699, 0.8188, 0.8712, 0.864, 0.8585, 0.7573, 0.8188, 0.7681, 0.7446, 0.8188]}
{"unit_id": "quote:1921134349803827413", "ticker_idx": 0, "ts": "2025-05-10T09:24:56+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This implies continued equipment Capex growth, especially for EUV infrastructure.", "tweet_type": "analysis", "timeline": "years", "summary": "Long ASML on sustained EUV Capex growth and High-NA ramp; long ASMI as the structural ALD leader with unique vertical integration into precursor supply.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Additional Reading for Context (by the Jukanlosreve)\n\n1. EUV Source Power and Increasing Lithographic Intensity Expected to Continue\n\nShuegraf acknowledged that the transition from N3 to N2 nodes does not significantly shrink feature sizes due to the Gate-All-Around (GAA) structure. However, he projected that lithographic intensity will continue to increase for nodes beyond that. He expressed confidence in ASML achieving 1000W source power from the current 600W, but noted that reaching 1500W would approach technical limitations.\n\nCommentary: This highlights that the end of Moore's Law scaling does not mean the end of lithographic demand. Even as geometric scaling slows, lithographic complexity—especially in terms of multi-patterning and total mask layers—is expected to intensify. If ASML succeeds in scaling source power to 1000W, it would substantially improve EUV throughput. However, 1500W may be near the physical and thermal limits of current source technology. This implies continued equipment Capex growth, especially for EUV infrastructure.\n\n2. 18A Node Outlook and High-NA Adoption Timing\n\nIn line with industry trends, Shuegraf stated that Intel must first prove the viability of its 18A node through its own products before attracting potential customers such as AVGO (Broadcom), NVDA (Nvidia), and AAPL (Apple M-series). He predicted that 2026 would be the mainstream ramp-up period for 18A, which would be about one year behind TSMC’s 2nm node production schedule. Regarding High-NA EUV, he noted that TSMC’s full adoption would likely be in the late 2020s, as it will take time for ASML to scale up production capacity.\n\nCommentary: Intel’s IFS strategy hinges on internal product validation of 18A before winning over external foundry customers. A 2026 ramp suggests Intel remains about a year behind TSMC’s roadmap, maintaining TSMC's lead for bleeding-edge nodes. As for High-NA EUV, its high cost, uncertain productivity, and ecosystem immaturity imply that industry-wide adoption may not occur until 2027–2029. ASML’s production capacity for High-NA tools will be a key bottleneck.\n\n3. Potential Expansion of EUV Adoption in DRAM\n\nShuegraf suggested that if leading HBM suppliers adopt EUV, it could drive broader EUV adoption across the DRAM industry. This is due to trends such as AI PCs and advanced packaging making DRAM a more specialized, rather than purely commoditized, product segment. While some customers may be able to maintain current lithographic intensity levels through 4F2 architecture, others are expected to adopt additional litho layers alongside 4F2.\n\nCommentary: DRAM, especially high-value products like HBM, is evolving from a cost-optimized commodity into a customized, performance-sensitive component. If HBM leaders (e.g., SK Hynix, Micron, Samsung) move forward with EUV, others may follow to stay competitive. 4F2 cell scaling may soon hit practical limits, and EUV patterning—especially for logic-like peripheral circuitry—could become standard. Expect increased DRAM Capex and technology bifurcation between leading-edge and trailing-edge suppliers.\n\n4. Additional Notes on China and ALD\n\n(A) Regarding China’s development of EUV using DPP (Discharge Produced Plasma)\nShuegraf acknowledged that for AI use cases, high wafer throughput is not always essential. However, he commented that DPP is economically inefficient and presents challenges in supply availability.\n\nCommentary: China’s DPP-based EUV efforts signal a desire to reduce reliance on ASML, but the technology is limited by low throughput, unstable plasma generation, and poor ecosystem maturity. Even if usable in narrow-edge AI inference, DPP remains economically unviable for volume manufacturing. The barriers to achieving ASML-equivalent tool maturity remain steep, reinforcing the durability of Western EUV hegemony for the foreseeable future.\n\n(B) On the topic of ALD (Atomic Layer Deposition)\nHe emphasized that ASMI remains the leading player due to its position as the only vertically integrated vendor with an in-house chemical supply chain.\n\nCommentary: As device architectures transition to GAA transistors and higher-layer 3D NAND, the precision required in deposition rises significantly. ASMI’s leadership in ALD is strengthened by its integrated control over precursor supply chains. This vertical integration is increasingly critical in securing performance, purity, and delivery lead times—making ALD a structural growth area, not just in logic, but also in memory and advanced packaging nodes.", "reply_text": null, "parent_text": null, "quoted_text": "BofA) Expert Call with Former Intel Lithography and Foundry Executive\n\nYesterday, we hosted an expert call with Klaus Shuegraf, currently VP at Photonic and formerly Director of Lithography Sourcing and VP of Foundry Strategy & Planning at Intel.\n\n1. EUV Source Power and https://t.co/DWm9As6BBs", "parent_author": null, "ret_m1d": -0.06012902404463538, "ret_signed_m1d": -0.06012902404463538, "alpha_spy_m1d": -0.028138712873499117, "alpha_c_m1d": -0.001117141942292621, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025191146732793213, "ret_signed_p1d": 0.025191146732793213, "alpha_spy_p1d": 0.01858720641562317, "alpha_c_p1d": -0.00913761303010241, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": -0.006747701073078538, "ret_signed_p1w": -0.006747701073078538, "alpha_spy_p1w": -0.02709100134682374, "alpha_c_p1w": -0.042971934001199585, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.02639070318381087, "ret_signed_p1m": 0.02639070318381087, "alpha_spy_p1m": -0.008069467264435026, "alpha_c_p1m": -0.07748025919240065, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": -0.08184198290749678, "ret_signed_p3m": -0.08184198290749678, "alpha_spy_p3m": -0.1695433777256845, "alpha_c_p3m": -0.308075022049211, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.38260481704527005, "ret_signed_p6m": 0.38260481704527005, "alpha_spy_p6m": 0.2177241319822958, "alpha_c_p6m": -0.10326356013764282, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0739, 0.103, 0.0945, 0.0879, 0.0706, 0.0725, 0.0612, 0.0615, 0.0665, 0.0431, 0.068, 0.045, 0.0144, 0.0319, -0.0127, 0.0056, 0.0015, -0.0045, -0.0448, -0.0553, -0.0351, -0.0397, -0.0199, -0.0115, -0.0046, 0.0062, 0.0053, -0.0124, 0.0084, 0.0085, -0.0176, -0.0396, -0.0648, -0.0941, -0.0737, -0.0787, -0.135, -0.1568, -0.1779, -0.1423, -0.1714, -0.1366, -0.137, -0.1181, -0.095, -0.142, -0.1566, -0.1566, -0.1566, -0.1613, -0.1317, -0.1155, -0.117, -0.1236, -0.1228, -0.1266, -0.1266, -0.087, -0.0901, -0.0981, -0.0933, -0.0546, -0.0601, 0.0, 0.0252, 0.0295, 0.0229, 0.0022, -0.0067, -0.0103, -0.0033, -0.0151, -0.0364, -0.0193, -0.0012, -0.0091, -0.0021, -0.0195, -0.031, -0.0207, -0.0217, -0.0118, 0.001, 0.0198, 0.0264, 0.0349, 0.0145, -0.0034, 0.0048, -0.0013, -0.0028, -0.0207, -0.0247, 0.0027, 0.0367, 0.0414, 0.0157, 0.0234, 0.016, 0.0019, 0.016, 0.0153, -0.0112, 0.007, 0.018, 0.0193, 0.0346, 0.0255, 0.0306, 0.0588, -0.0616, -0.025, -0.0501, -0.0642, -0.0967, -0.0904, -0.0765, -0.093, -0.0554, -0.0633, -0.0566, -0.0782, -0.1054, -0.0943, -0.1023, -0.1086, -0.0818, -0.0725, -0.0573, -0.047, -0.0369, -0.0334, -0.0428, -0.0397, -0.0343, -0.0405, -0.0461, -0.0285, -0.0277, -0.0219, -0.0074, -0.0163, -0.0429, -0.0437, -0.0721, -0.0599, -0.0264, -0.0074, 0.0172, 0.0272, 0.0226, 0.0314, 0.0371, 0.0966, 0.1019, 0.108, 0.1937, 0.1933, 0.219, 0.2325, 0.2148, 0.2204, 0.222, 0.2411, 0.245, 0.265, 0.3194, 0.3232, 0.349, 0.3152, 0.2806, 0.2776, 0.2237, 0.2689, 0.2728, 0.3125, 0.3191, 0.3166, 0.3516, 0.3378, 0.3082, 0.3346, 0.3447, 0.3653, 0.3666, 0.3847, 0.412, 0.3828, 0.3954, 0.3826]}
{"unit_id": "quote:1921134349803827413", "ticker_idx": 1, "ts": "2025-05-10T09:24:56+00:00", "symbol": "ASM International", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASMI's leadership in ALD is strengthened by its integrated control over precursor supply chains. This vertical integration is increasingly critical in securing performance, purity, and delivery lead times—making ALD a structural growth area", "tweet_type": "analysis", "timeline": "years", "summary": "Long ASML on sustained EUV Capex growth and High-NA ramp; long ASMI as the structural ALD leader with unique vertical integration into precursor supply.", "resolved_tickers": ["ASMIY"], "resolution_method": "exchange_ticker", "resolution_reason": "ASM International Amsterdam; US OTC ADR ASMIY", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Additional Reading for Context (by the Jukanlosreve)\n\n1. EUV Source Power and Increasing Lithographic Intensity Expected to Continue\n\nShuegraf acknowledged that the transition from N3 to N2 nodes does not significantly shrink feature sizes due to the Gate-All-Around (GAA) structure. However, he projected that lithographic intensity will continue to increase for nodes beyond that. He expressed confidence in ASML achieving 1000W source power from the current 600W, but noted that reaching 1500W would approach technical limitations.\n\nCommentary: This highlights that the end of Moore's Law scaling does not mean the end of lithographic demand. Even as geometric scaling slows, lithographic complexity—especially in terms of multi-patterning and total mask layers—is expected to intensify. If ASML succeeds in scaling source power to 1000W, it would substantially improve EUV throughput. However, 1500W may be near the physical and thermal limits of current source technology. This implies continued equipment Capex growth, especially for EUV infrastructure.\n\n2. 18A Node Outlook and High-NA Adoption Timing\n\nIn line with industry trends, Shuegraf stated that Intel must first prove the viability of its 18A node through its own products before attracting potential customers such as AVGO (Broadcom), NVDA (Nvidia), and AAPL (Apple M-series). He predicted that 2026 would be the mainstream ramp-up period for 18A, which would be about one year behind TSMC’s 2nm node production schedule. Regarding High-NA EUV, he noted that TSMC’s full adoption would likely be in the late 2020s, as it will take time for ASML to scale up production capacity.\n\nCommentary: Intel’s IFS strategy hinges on internal product validation of 18A before winning over external foundry customers. A 2026 ramp suggests Intel remains about a year behind TSMC’s roadmap, maintaining TSMC's lead for bleeding-edge nodes. As for High-NA EUV, its high cost, uncertain productivity, and ecosystem immaturity imply that industry-wide adoption may not occur until 2027–2029. ASML’s production capacity for High-NA tools will be a key bottleneck.\n\n3. Potential Expansion of EUV Adoption in DRAM\n\nShuegraf suggested that if leading HBM suppliers adopt EUV, it could drive broader EUV adoption across the DRAM industry. This is due to trends such as AI PCs and advanced packaging making DRAM a more specialized, rather than purely commoditized, product segment. While some customers may be able to maintain current lithographic intensity levels through 4F2 architecture, others are expected to adopt additional litho layers alongside 4F2.\n\nCommentary: DRAM, especially high-value products like HBM, is evolving from a cost-optimized commodity into a customized, performance-sensitive component. If HBM leaders (e.g., SK Hynix, Micron, Samsung) move forward with EUV, others may follow to stay competitive. 4F2 cell scaling may soon hit practical limits, and EUV patterning—especially for logic-like peripheral circuitry—could become standard. Expect increased DRAM Capex and technology bifurcation between leading-edge and trailing-edge suppliers.\n\n4. Additional Notes on China and ALD\n\n(A) Regarding China’s development of EUV using DPP (Discharge Produced Plasma)\nShuegraf acknowledged that for AI use cases, high wafer throughput is not always essential. However, he commented that DPP is economically inefficient and presents challenges in supply availability.\n\nCommentary: China’s DPP-based EUV efforts signal a desire to reduce reliance on ASML, but the technology is limited by low throughput, unstable plasma generation, and poor ecosystem maturity. Even if usable in narrow-edge AI inference, DPP remains economically unviable for volume manufacturing. The barriers to achieving ASML-equivalent tool maturity remain steep, reinforcing the durability of Western EUV hegemony for the foreseeable future.\n\n(B) On the topic of ALD (Atomic Layer Deposition)\nHe emphasized that ASMI remains the leading player due to its position as the only vertically integrated vendor with an in-house chemical supply chain.\n\nCommentary: As device architectures transition to GAA transistors and higher-layer 3D NAND, the precision required in deposition rises significantly. ASMI’s leadership in ALD is strengthened by its integrated control over precursor supply chains. This vertical integration is increasingly critical in securing performance, purity, and delivery lead times—making ALD a structural growth area, not just in logic, but also in memory and advanced packaging nodes.", "reply_text": null, "parent_text": null, "quoted_text": "BofA) Expert Call with Former Intel Lithography and Foundry Executive\n\nYesterday, we hosted an expert call with Klaus Shuegraf, currently VP at Photonic and formerly Director of Lithography Sourcing and VP of Foundry Strategy & Planning at Intel.\n\n1. EUV Source Power and https://t.co/DWm9As6BBs", "parent_author": null, "ret_m1d": -0.06790809825691857, "ret_signed_m1d": -0.06790809825691857, "alpha_spy_m1d": -0.035917787085782305, "alpha_c_m1d": -0.008896216154575809, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.031092876584287055, "ret_signed_p1d": 0.031092876584287055, "alpha_spy_p1d": 0.024488936267117012, "alpha_c_p1d": -0.0032358831786085673, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.002619418047691724, "ret_signed_p1w": 0.002619418047691724, "alpha_spy_p1w": -0.017723882226053478, "alpha_c_p1w": -0.03360481488042932, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.13492400932527793, "ret_signed_p1m": 0.13492400932527793, "alpha_spy_p1m": 0.10046383887703203, "alpha_c_p1m": 0.03105304694906641, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": -0.09880165280822972, "ret_signed_p3m": -0.09880165280822972, "alpha_spy_p3m": -0.18650304762641745, "alpha_c_p3m": -0.32503469194994394, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.17431006571731333, "ret_signed_p6m": 0.17431006571731333, "alpha_spy_p6m": 0.00942938065433907, "alpha_c_p6m": -0.31155831146559954, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0889, 0.1024, 0.1047, 0.1047, 0.0951, 0.0952, 0.0951, 0.0826, 0.0814, 0.045, 0.0616, -0.0232, -0.0233, -0.0484, -0.072, -0.0307, -0.1089, -0.0835, -0.16, -0.1341, -0.105, -0.1274, -0.0891, -0.0768, -0.0953, -0.0723, -0.0884, -0.1117, -0.0889, -0.0949, -0.1102, -0.1267, -0.1386, -0.1625, -0.1718, -0.1723, -0.2207, -0.2629, -0.2622, -0.3071, -0.2024, -0.2458, -0.1896, -0.1985, -0.1529, -0.1857, -0.1559, -0.1559, -0.1873, -0.1794, -0.1618, -0.1219, -0.0947, -0.0924, -0.0635, -0.1029, -0.1025, -0.072, -0.0869, -0.0991, -0.0652, -0.0716, -0.0679, 0.0, 0.0311, 0.0295, 0.0174, 0.0107, 0.0026, 0.0152, -0.0119, 0.0018, -0.0091, -0.0091, 0.026, 0.0121, 0.0434, 0.0072, 0.0108, 0.0274, 0.0381, 0.0376, 0.067, 0.0939, 0.1349, 0.1343, 0.1253, 0.106, 0.116, 0.1042, 0.0992, 0.0992, 0.0885, 0.1209, 0.1533, 0.1723, 0.1695, 0.1745, 0.1863, 0.1496, 0.1323, 0.1182, 0.1182, 0.1081, 0.1181, 0.1158, 0.1417, 0.1346, 0.1349, 0.1383, 0.1013, 0.1136, 0.1173, 0.1232, 0.0365, -0.0062, -0.0403, -0.073, -0.0384, -0.0518, -0.0486, -0.104, -0.1082, -0.1013, -0.1079, -0.1212, -0.0988, -0.091, -0.102, -0.0692, -0.0711, -0.0746, -0.1042, -0.1106, -0.1138, -0.1192, -0.1357, -0.1148, -0.1215, -0.1127, -0.1127, -0.0743, -0.1116, -0.1116, -0.1251, -0.1307, -0.1141, -0.0946, -0.0814, -0.0825, -0.0893, -0.0732, -0.076, -0.0179, 0.0039, 0.0124, 0.0869, 0.0739, 0.1013, 0.0857, 0.0889, 0.0778, 0.0695, 0.0906, 0.1124, 0.142, 0.2019, 0.1855, 0.2111, 0.1544, 0.1822, 0.1689, 0.1257, 0.1656, 0.1665, 0.1727, 0.1751, 0.1729, 0.1781, 0.1759, 0.1505, 0.1792, 0.1966, 0.2176, 0.1291, 0.1829, 0.2071, 0.2004, 0.2086, 0.1743]}
{"unit_id": "orig:1942350138342850827", "ticker_idx": 0, "ts": "2025-07-07T22:28:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi has raised NVIDIA's target price from $180 to $190, based on a significant upward revision of the AI accelerator and networking Total Addressable Market (TAM)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/endorses Citi Buy on NVDA with PT raised to $190 on sovereign AI demand, networking TAM expansion, Blackwell ramp, and raised FY27/FY28 estimates.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: Target Price Raised to $190 on Strong Sovereign AI Demand and Significant Network TAM Increase\nCiti (A. Malik, 25/07/07)\n\nCiti has raised NVIDIA's target price from $180 to $190, based on a significant upward revision of the AI accelerator and networking Total Addressable Market (TAM). This revision is driven by strengthening sovereign AI demand, the scaling up of AI training clusters, and the expansion of network connectivity. Concurrently, the Blackwell product line is progressing smoothly, and the trend of improving gross margins continues.\n\nData Center Semiconductor TAM Raised to $563 Billion, Network Contribution Growing Most Significantly\nCiti projects the AI TAM for data center semiconductors to reach $563 billion by 2028, a 13% increase from previous forecasts. Within this, the GPU and custom ASIC market is expected to grow by 4%, primarily driven by better-than-expected sovereign AI projects. The networking TAM has been significantly revised upwards from $90 billion to $119 billion, a 32% increase, propelled by larger-scale AI clusters and amplified enterprise internal network demand. Networking's share of the overall TAM is projected to increase to 21%.\n\nFY27/FY28 Sales and Earnings Forecasts Upwardly Revised, Networking Growth Particularly Prominent\nCiti has raised NVIDIA's FY27/FY28 data center revenue forecasts by +5% and +11% respectively. Networking sales forecasts have been raised by +12% and +27%, and are expected to account for approximately 20% of total data center revenue. Earnings per share (EPS) forecasts have increased by 6% and 21% for FY27 and FY28, respectively. The networking business is projected to grow by 112% in FY27, becoming the second growth engine driven by the opening of the NVLink platform and the expansion of AI switches.\n\nBlackwell Shipments Exceed Expectations, GB300 Expected to Maintain Smooth Delivery Pace\nGB200 shipments are accelerating, and Citi believes that previous concerns about supply chain bottlenecks have largely been resolved. The company is well-positioned to smoothly transition to the GB300 product leveraging its experience from the Hopper to B200 transition. NVIDIA plays a core role in multiple sovereign AI projects, and sovereign AI deployments are expected to reach tens of GW scale in the coming years, further enhancing long-term revenue certainty.\n\nGross Margin Rebounds to Mid-to-High Levels, Export Restriction Risks Limited\nThe gross margin is expected to rebound to over 70% in FY25, primarily benefiting from the mass production progress of Blackwell GB200. Although some countries (e.g., Malaysia, Thailand) may face future export regulations, Citi believes NVIDIA possesses tracking mechanisms to control risk and considers the impact of China export bans to be largely reflected in current expectations.\n\nAI Accelerators to Grow at 22% CAGR, GPUs Still Dominate the Market\nBetween 2025 and 2028, the overall AI accelerator Compound Annual Growth Rate (CAGR) is projected to be 22%. GPUs are expected to grow at 16% annually, while ASICs are projected to grow at 32%. However, GPU prices are significantly higher than ASICs (3 to 8 times higher). By 2028, GPU and ASIC sales are expected to reach $338 billion and $59 billion, respectively, totaling $396 billion, accounting for over 85% of the accelerator TAM. NVIDIA's technological leadership and strong user base advantages remain central to its market share.\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00695140583761078, "ret_signed_m1d": 0.00695140583761078, "alpha_spy_m1d": -0.0005565607408903883, "alpha_c_m1d": -0.006846119860680178, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011122326495506485, "ret_signed_p1d": 0.011122326495506485, "alpha_spy_p1d": 0.011670042713251783, "alpha_c_p1d": -0.0021748019926659534, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.03684273062740595, "ret_signed_p1w": 0.03684273062740595, "alpha_spy_p1w": 0.03018881656877115, "alpha_c_p1w": 0.016968546658937544, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": 0.1265166085526297, "ret_signed_p1m": 0.1265166085526297, "alpha_spy_p1m": 0.11477150217340681, "alpha_c_p1m": 0.10027984068309848, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.19376037136403101, "ret_signed_p3m": 0.19376037136403101, "alpha_spy_p3m": 0.11256667330372783, "alpha_c_p3m": -0.015275885275286782, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.18529450285638682, "ret_signed_p6m": 0.18529450285638682, "alpha_spy_p6m": 0.07207928965415489, "alpha_c_p6m": -0.11739514582603539, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.2775, -0.3203, -0.299, -0.3004, -0.291, -0.3397, -0.3587, -0.3587, -0.3876, -0.3751, -0.351, -0.3275, -0.2985, -0.3129, -0.3111, -0.3117, -0.2947, -0.2765, -0.2808, -0.2825, -0.2603, -0.2583, -0.2629, -0.2228, -0.179, -0.1448, -0.148, -0.1444, -0.1433, -0.1508, -0.1671, -0.1606, -0.1704, -0.1704, -0.1438, -0.1481, -0.1204, -0.1461, -0.1319, -0.1076, -0.1032, -0.1154, -0.1045, -0.0987, -0.0903, -0.0974, -0.0837, -0.1028, -0.0856, -0.0892, -0.0806, -0.0806, -0.0909, -0.0889, -0.0653, -0.0248, -0.0203, -0.0031, -0.0016, -0.0312, -0.0063, 0.007, 0.007, 0.0, 0.0111, 0.0293, 0.037, 0.0422, 0.0368, 0.0787, 0.083, 0.0933, 0.0895, 0.083, 0.0555, 0.0792, 0.098, 0.0964, 0.117, 0.1091, 0.1329, 0.1241, 0.0978, 0.1375, 0.1265, 0.1338, 0.1424, 0.1546, 0.1505, 0.1575, 0.1476, 0.1503, 0.1404, 0.1502, 0.11, 0.1084, 0.1058, 0.1248, 0.1363, 0.1487, 0.1476, 0.1386, 0.1007, 0.1007, 0.0792, 0.0782, 0.0848, 0.0555, 0.0636, 0.0791, 0.1206, 0.1197, 0.1238, 0.1234, 0.1052, 0.0762, 0.1138, 0.1165, 0.1604, 0.1277, 0.1184, 0.123, 0.1261, 0.1493, 0.1792, 0.1833, 0.1938, 0.1857, 0.1726, 0.1694, 0.1952, 0.217, 0.1575, 0.1902, 0.1378, 0.1365, 0.149, 0.1579, 0.1543, 0.1449, 0.1393, 0.1512, 0.1771, 0.2102, 0.2705, 0.3085, 0.2822, 0.2797, 0.3075, 0.2557, 0.2337, 0.1886, 0.1891, 0.258, 0.2207, 0.2248, 0.1809, 0.2018, 0.1793, 0.1462, 0.1788, 0.1416, 0.1305, 0.1537, 0.1238, 0.1392, 0.1392, 0.1186, 0.1371, 0.1468, 0.135, 0.159, 0.1529, 0.1727, 0.1691, 0.1615, 0.1435, 0.1062, 0.1142, 0.1232, 0.0804, 0.1006, 0.1439, 0.161, 0.1958, 0.1921, 0.1921, 0.2042, 0.1896, 0.1853]}
{"unit_id": "orig:1928780115363959281", "ticker_idx": 0, "ts": "2025-05-31T11:46:29+00:00", "symbol": "CDNS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Buy ratings on CDNS and SNPS, based on their unique long-term positioning in semiconductor R&D intensity and growing chip complexity", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA reiterates Buy on CDNS and SNPS post China export restriction selloff; downside seen largely priced in, valuations within historical ranges.", "resolved_tickers": ["CDNS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Application", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) EDA Update: Assessing Impact of China Export Restrictions – Maintain ‘Buy’ Ratings on CDNS and SNPS\n\nBase/Downside Scenarios Suggest ~2%/10% Revenue, ~5%/20% EPS Impact\nWe recently held brief discussions with Cadence Design Systems (CDNS) and Synopsys (SNPS) to assess the potential impact of the China export restrictions outlined in a recent letter from the U.S. Bureau of Industry and Security (BIS). Both companies are still evaluating the full implications and working with BIS for regulatory clarity. For now, we present three potential scenarios:\n1. Base Case: No impact\n2. Moderate Case: 25% reduction in China revenue, similar to patterns observed in semiconductor equipment firms\n3. Worst Case: 100% loss of China revenue\n\nGiven that China accounted for ~12% of CDNS and ~14% of SNPS revenue over the past four quarters, we estimate that the moderate/worst-case scenarios could translate to a ~2%/~10% revenue and ~5%/~20% pro forma EPS impact on an annualized basis. Valuation multiples remain within historical trading ranges. We maintain Buy ratings on CDNS and SNPS, based on their unique long-term positioning in semiconductor R&D intensity and growing chip complexity.\n\n⸻\n\nModerate Case: ~2% Revenue / ~5% EPS Decline\nWhile not exact, we reference the ~25% China revenue decline seen in semi-cap equipment under similar restrictions as a comparable framework.\n• CDNS: In CY24, CDNS derived 13% of revenue from China. Management expects China revenue to remain flat YoY in CY25, accounting for ~11%.\n• SNPS: Generated 15% of CY24 revenue from China but has underperformed in 1H. Management anticipates a YoY decline, in line with its typical back-weighted seasonality (45% H1 / 55% H2).\n\nAssuming a 25% decline in China revenue for the remaining CY25 and all of CY26, we estimate ~2% revenue and ~5% EPS impact for both companies.\n\n⸻\n\nWorst Case: ~10% Revenue / ~20% EPS Decline\nThis scenario assumes broad and immediate enforcement of restrictions, leading to a full halt in China revenue during the remainder of CY25 and all of CY26.\n• CDNS: -10% revenue, -20% pf-EPS\n• SNPS: -10% revenue, -19% pf-EPS\n\n⸻\n\nValuation Pressures Elevated but Still Within Historical Ranges\nFollowing recent EDA stock selloffs, our worst-case scenario implies FY26E P/E of 34x for SNPS (within its 5-year trading range of 24x–47x) and 46x for CDNS (within its 34x–59x range).\n• Moderate Case: FY26E P/E of 29x (SNPS) / 39x (CDNS)\n• Base Case: FY26E P/E of 27x (SNPS) / 37x (CDNS)\n\nWe expect the BIS action to remain an overhang in the near term, but in our view, potential downside is largely priced in. Both SNPS and CDNS remain key long-term beneficiaries of secular semiconductor design tailwinds, supporting our Buy thesis.\n\n⸻\n\nBIS Letter Triggers Sell-Off and SNPS Guidance Withdrawal\nThe BIS notified SNPS and CDNS that licenses are now required to export, reexport, or transfer EDA software/technology to Chinese customers, citing national security concerns.\n\nFollowing the announcement, shares of SNPS and CDNS dropped -11% and -12%, respectively, as of Thursday’s close. SNPS withdrew its 3Q25 and FY25 guidance. CDNS canceled its scheduled fireside chat at next week’s BofA Tech Conference. Both companies are actively assessing the potential ramifications of these new restrictions.\n\n$CDNS $SNPS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018664723200568223, "ret_signed_m1d": -0.018664723200568223, "alpha_spy_m1d": -0.013063373910538023, "alpha_c_m1d": -0.013063373910538023, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.0056013492900302, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002734720528858592, "ret_signed_p1d": 0.002734720528858592, "alpha_spy_p1d": -0.0029678692593289036, "alpha_c_p1d": -0.0029678692593289036, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.005702589788187495, "ret_p1w": 0.031210490961307347, "ret_signed_p1w": 0.031210490961307347, "alpha_spy_p1w": 0.019451073798272267, "alpha_c_p1w": 0.019451073798272267, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.01175941716303508, "ret_p1m": 0.057874381247833284, "ret_signed_p1m": 0.057874381247833284, "alpha_spy_p1m": 0.012715807068180762, "alpha_c_p1m": 0.012715807068180762, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.04515857417965252, "ret_p3m": 0.2111236797024043, "ret_signed_p3m": 0.2111236797024043, "alpha_spy_p3m": 0.11305163417251074, "alpha_c_p3m": 0.11305163417251074, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.09807204552989357, "ret_p6m": 0.038047396606355255, "ret_signed_p6m": 0.038047396606355255, "alpha_spy_p6m": -0.10735666494995066, "alpha_c_p6m": -0.10735666494995066, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.14540406155630592, "price_path": [-0.1635, -0.1688, -0.1686, -0.2002, -0.1992, -0.1822, -0.1742, -0.1542, -0.1279, -0.1321, -0.1156, -0.1163, -0.1029, -0.0868, -0.0801, -0.0995, -0.1025, -0.1225, -0.1306, -0.1153, -0.0943, -0.1491, -0.2039, -0.2007, -0.2081, -0.0953, -0.1316, -0.113, -0.1093, -0.0965, -0.1137, -0.1105, -0.1105, -0.1401, -0.1175, -0.0927, -0.0367, -0.0099, -0.0232, 0.0331, 0.0178, 0.0246, 0.0534, 0.0557, 0.0453, 0.054, 0.0527, 0.0453, 0.0658, 0.0893, 0.0918, 0.0897, 0.0992, 0.0949, 0.0958, 0.0855, 0.0824, 0.0786, 0.0786, 0.1044, -0.0134, -0.0275, -0.0187, 0.0, 0.0027, 0.0064, 0.0129, 0.0151, 0.0312, 0.045, 0.0501, 0.0524, 0.0255, 0.0283, 0.0221, 0.0147, 0.0147, 0.0098, 0.0035, 0.0146, 0.0171, 0.0381, 0.0433, 0.0534, 0.0579, 0.063, 0.1172, 0.1172, 0.0949, 0.1051, 0.1039, 0.103, 0.096, 0.0853, 0.0872, 0.0754, 0.0968, 0.0788, 0.0813, 0.0925, 0.116, 0.1098, 0.1356, 0.1409, 0.252, 0.2683, 0.2463, 0.2203, 0.2473, 0.2324, 0.2312, 0.2113, 0.2035, 0.1933, 0.2088, 0.1935, 0.1928, 0.196, 0.2178, 0.1858, 0.1809, 0.1882, 0.1952, 0.18, 0.1761, 0.1858, 0.2111, 0.1979, 0.1979, 0.1719, 0.1873, 0.1942, 0.1999, 0.2328, 0.2367, 0.1572, 0.2125, 0.1742, 0.2017, 0.1931, 0.1871, 0.2473, 0.2763, 0.2763, 0.2522, 0.2203, 0.2001, 0.1968, 0.1914, 0.2008, 0.2032, 0.187, 0.1871, 0.2079, 0.181, 0.1965, 0.1919, 0.1178, 0.1357, 0.1136, 0.1079, 0.1084, 0.1148, 0.1269, 0.1399, 0.1298, 0.1529, 0.1797, 0.2012, 0.1667, 0.1672, 0.1489, 0.1578, 0.1466, 0.1391, 0.12, 0.1091, 0.1112, 0.1245, 0.0888, 0.0787, 0.0802, 0.0766, 0.0641, 0.0365, 0.0486, 0.0329, 0.0275, 0.0408, 0.038]}
{"unit_id": "orig:1928780115363959281", "ticker_idx": 1, "ts": "2025-05-31T11:46:29+00:00", "symbol": "SNPS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Buy ratings on CDNS and SNPS, based on their unique long-term positioning in semiconductor R&D intensity and growing chip complexity", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA reiterates Buy on CDNS and SNPS post China export restriction selloff; downside seen largely priced in, valuations within historical ranges.", "resolved_tickers": ["SNPS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) EDA Update: Assessing Impact of China Export Restrictions – Maintain ‘Buy’ Ratings on CDNS and SNPS\n\nBase/Downside Scenarios Suggest ~2%/10% Revenue, ~5%/20% EPS Impact\nWe recently held brief discussions with Cadence Design Systems (CDNS) and Synopsys (SNPS) to assess the potential impact of the China export restrictions outlined in a recent letter from the U.S. Bureau of Industry and Security (BIS). Both companies are still evaluating the full implications and working with BIS for regulatory clarity. For now, we present three potential scenarios:\n1. Base Case: No impact\n2. Moderate Case: 25% reduction in China revenue, similar to patterns observed in semiconductor equipment firms\n3. Worst Case: 100% loss of China revenue\n\nGiven that China accounted for ~12% of CDNS and ~14% of SNPS revenue over the past four quarters, we estimate that the moderate/worst-case scenarios could translate to a ~2%/~10% revenue and ~5%/~20% pro forma EPS impact on an annualized basis. Valuation multiples remain within historical trading ranges. We maintain Buy ratings on CDNS and SNPS, based on their unique long-term positioning in semiconductor R&D intensity and growing chip complexity.\n\n⸻\n\nModerate Case: ~2% Revenue / ~5% EPS Decline\nWhile not exact, we reference the ~25% China revenue decline seen in semi-cap equipment under similar restrictions as a comparable framework.\n• CDNS: In CY24, CDNS derived 13% of revenue from China. Management expects China revenue to remain flat YoY in CY25, accounting for ~11%.\n• SNPS: Generated 15% of CY24 revenue from China but has underperformed in 1H. Management anticipates a YoY decline, in line with its typical back-weighted seasonality (45% H1 / 55% H2).\n\nAssuming a 25% decline in China revenue for the remaining CY25 and all of CY26, we estimate ~2% revenue and ~5% EPS impact for both companies.\n\n⸻\n\nWorst Case: ~10% Revenue / ~20% EPS Decline\nThis scenario assumes broad and immediate enforcement of restrictions, leading to a full halt in China revenue during the remainder of CY25 and all of CY26.\n• CDNS: -10% revenue, -20% pf-EPS\n• SNPS: -10% revenue, -19% pf-EPS\n\n⸻\n\nValuation Pressures Elevated but Still Within Historical Ranges\nFollowing recent EDA stock selloffs, our worst-case scenario implies FY26E P/E of 34x for SNPS (within its 5-year trading range of 24x–47x) and 46x for CDNS (within its 34x–59x range).\n• Moderate Case: FY26E P/E of 29x (SNPS) / 39x (CDNS)\n• Base Case: FY26E P/E of 27x (SNPS) / 37x (CDNS)\n\nWe expect the BIS action to remain an overhang in the near term, but in our view, potential downside is largely priced in. Both SNPS and CDNS remain key long-term beneficiaries of secular semiconductor design tailwinds, supporting our Buy thesis.\n\n⸻\n\nBIS Letter Triggers Sell-Off and SNPS Guidance Withdrawal\nThe BIS notified SNPS and CDNS that licenses are now required to export, reexport, or transfer EDA software/technology to Chinese customers, citing national security concerns.\n\nFollowing the announcement, shares of SNPS and CDNS dropped -11% and -12%, respectively, as of Thursday’s close. SNPS withdrew its 3Q25 and FY25 guidance. CDNS canceled its scheduled fireside chat at next week’s BofA Tech Conference. Both companies are actively assessing the potential ramifications of these new restrictions.\n\n$CDNS $SNPS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004655117319680047, "ret_signed_m1d": -0.004655117319680047, "alpha_spy_m1d": 0.0009462319703501532, "alpha_c_m1d": 0.0009462319703501532, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.0056013492900302, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0002360095904345716, "ret_signed_p1d": 0.0002360095904345716, "alpha_spy_p1d": -0.005466580197752924, "alpha_c_p1d": -0.005466580197752924, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.005702589788187495, "ret_p1w": 0.06354176875472639, "ret_signed_p1w": 0.06354176875472639, "alpha_spy_p1w": 0.05178235159169131, "alpha_c_p1w": 0.05178235159169131, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.01175941716303508, "ret_p1m": 0.11933934598043594, "ret_signed_p1m": 0.11933934598043594, "alpha_spy_p1m": 0.07418077180078342, "alpha_c_p1m": 0.07418077180078342, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.04515857417965252, "ret_p3m": 0.31324678950032947, "ret_signed_p3m": 0.31324678950032947, "alpha_spy_p3m": 0.2151747439704359, "alpha_c_p3m": 0.2151747439704359, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.09807204552989357, "ret_p6m": -0.13845330771205366, "ret_signed_p6m": -0.13845330771205366, "alpha_spy_p6m": -0.2838573692683596, "alpha_c_p6m": -0.2838573692683596, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.14540406155630592, "price_path": [-0.039, -0.0438, -0.0328, -0.0779, -0.0761, -0.0723, -0.0806, -0.0427, -0.0271, -0.039, -0.034, -0.0347, -0.0376, -0.018, -0.0191, -0.0349, -0.0427, -0.0605, -0.08, -0.0648, -0.0592, -0.1038, -0.1674, -0.1829, -0.1814, -0.0803, -0.1271, -0.0964, -0.0948, -0.0869, -0.1083, -0.1133, -0.1133, -0.1368, -0.1132, -0.0942, -0.0585, -0.0419, -0.0476, -0.0115, -0.0153, -0.0146, 0.0132, 0.0242, 0.0159, 0.0309, 0.0394, 0.0359, 0.0846, 0.1063, 0.1105, 0.1037, 0.1036, 0.107, 0.1061, 0.0757, 0.0801, 0.0701, 0.0701, 0.0979, -0.008, -0.0239, -0.0047, 0.0, 0.0002, 0.0151, 0.0344, 0.0426, 0.0635, 0.0821, 0.0742, 0.0619, 0.0273, 0.0284, 0.0198, 0.016, 0.016, 0.0094, 0.0104, 0.0275, 0.0453, 0.0634, 0.0783, 0.0998, 0.1193, 0.1222, 0.1772, 0.1772, 0.151, 0.1831, 0.1819, 0.2146, 0.1998, 0.1789, 0.2209, 0.2254, 0.2636, 0.2544, 0.2688, 0.3196, 0.3181, 0.3079, 0.2905, 0.2713, 0.364, 0.3844, 0.3589, 0.3271, 0.3644, 0.3483, 0.3427, 0.3308, 0.3288, 0.3217, 0.3425, 0.327, 0.3222, 0.3256, 0.3415, 0.3146, 0.2896, 0.2832, 0.3011, 0.2807, 0.2786, 0.2939, 0.3132, 0.2947, 0.2947, 0.27, 0.2702, 0.2913, 0.2831, 0.3066, 0.2965, -0.1681, -0.0602, -0.0873, -0.1007, -0.0862, -0.0874, 0.0299, 0.063, 0.1043, 0.0519, 0.0042, 0.0452, 0.0464, 0.0332, 0.0584, 0.0485, 0.0107, 0.0065, 0.0259, 0.0251, 0.0498, 0.0392, -0.0584, -0.0389, -0.048, -0.0649, -0.0557, -0.0397, -0.0275, -0.0153, -0.0254, -0.0216, -0.0042, -0.0016, -0.02, -0.0232, -0.0498, -0.0265, -0.0438, -0.1068, -0.1224, -0.1502, -0.156, -0.1403, -0.1513, -0.1445, -0.1549, -0.1637, -0.1628, -0.1766, -0.1713, -0.1728, -0.1669, -0.132, -0.1385]}
{"unit_id": "orig:1943318029368266820", "ticker_idx": 0, "ts": "2025-07-10T14:34:57+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "maintains an \"Equal-weight\" rating and has raised its price target to €660", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Equal-weight on ASML (PT €660); near-term orders strong but long-term growth constrained, valuation fully reflects medium-term expectations.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "ASML: Short-term Order Performance Optimistic, But Long-term Growth Expectations Tend Towards Conservative\n\nMorgan Stanley (L. Simpson, July 09, 2025)\n\nMorgan Stanley expects ASML's Q2 order performance to be strong, primarily driven by demand from China and TSMC. However, long-term growth is constrained by the slowing growth in EUV layer counts and a weak end market. The firm maintains an \"Equal-weight\" rating and has raised its price target to €660.\n\nQ2 2025 Orders May Exceed €4 Billion, Driven by China and TSMC\n\nMorgan Stanley anticipates ASML will achieve order volumes exceeding €4 billion in Q2 2025, mainly benefiting from a rebound in Chinese demand and the potential resumption of EUV orders from TSMC. Furthermore, China's revenue contribution in 2025 is expected to increase from the mid-20% range to the high-20% range, and the company is likely to meet the midpoint of its full-year revenue guidance.\n\nSlowing EUV Layer Growth Restricts Long-term Growth Potential\n\nASML benefited from rapid growth driven by EUV insertion during the N7 to N3 node period. However, it appears that device performance improvements are now shifting towards the system level, and lithography layer counts are approaching physical limits. This is expected to constrain the drivers for future revenue increases.\n\nFuture Growth Pace May Differ from Previous Cycle\n\nWhile current AI infrastructure investment is strong, its impact on spending for leading-edge process equipment is limited. Insufficient support from traditional end markets like smartphones and PCs, coupled with increasing concentration in the wafer foundry industry, suggests ASML's revenue CAGR from 2024 to 2027 will be in the low double-digits. It is unlikely to replicate the large-scale capital expenditure seen from TSMC in FY22.\n\nValuation Fully Reflects Medium-Term Expectations, Maintaining Neutral Rating\n\nMorgan Stanley believes ASML's current valuation is already quite fully reflected. The company also faces external challenges such as tariff risks, a weak mobile market, and subdued consumer demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01478265242470378, "ret_signed_m1d": 0.01478265242470378, "alpha_spy_m1d": 0.011970273986009583, "alpha_c_m1d": 0.007478955645557561, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008840506208121934, "ret_signed_p1d": 0.008840506208121934, "alpha_spy_p1d": 0.005325082337526377, "alpha_c_p1d": 0.00873616464067739, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.057681162211505876, "ret_signed_p1w": 0.057681162211505876, "alpha_spy_p1w": 0.06122845327851101, "alpha_c_p1w": 0.0722540591699492, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": -0.10357076239011898, "ret_signed_p1m": 0.10357076239011898, "alpha_spy_p1m": 0.12172286825609291, "alpha_c_p1m": 0.12447367894987105, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.2711884506426394, "ret_signed_p3m": -0.2711884506426394, "alpha_spy_p3m": -0.19903501992748152, "alpha_c_p3m": -0.09892211985644694, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.43574095334697893, "ret_signed_p6m": -0.43574095334697893, "alpha_spy_p6m": -0.3378401065549188, "alpha_c_p6m": -0.13335653558889593, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.1476, -0.1253, -0.1707, -0.1849, -0.1849, -0.1849, -0.1893, -0.1607, -0.1451, -0.1465, -0.1529, -0.1522, -0.1558, -0.1558, -0.1175, -0.1206, -0.1283, -0.1236, -0.0862, -0.0916, -0.0335, -0.0091, -0.0049, -0.0113, -0.0313, -0.04, -0.0435, -0.0367, -0.0481, -0.0687, -0.0522, -0.0346, -0.0423, -0.0355, -0.0523, -0.0635, -0.0535, -0.0545, -0.0449, -0.0325, -0.0143, -0.008, 0.0003, -0.0194, -0.0368, -0.0288, -0.0348, -0.0362, -0.0535, -0.0574, -0.0309, 0.002, 0.0065, -0.0183, -0.0109, -0.018, -0.0316, -0.018, -0.0187, -0.0443, -0.0267, -0.0161, -0.0148, 0.0, -0.0088, -0.0039, 0.0233, -0.093, -0.0577, -0.0819, -0.0955, -0.127, -0.1209, -0.1074, -0.1233, -0.087, -0.0947, -0.0882, -0.1091, -0.1354, -0.1246, -0.1323, -0.1384, -0.1126, -0.1036, -0.0889, -0.0789, -0.0691, -0.0658, -0.0748, -0.0719, -0.0667, -0.0726, -0.078, -0.061, -0.0603, -0.0546, -0.0407, -0.0492, -0.0749, -0.0757, -0.1031, -0.0914, -0.059, -0.0407, -0.0168, -0.0072, -0.0116, -0.0032, 0.0024, 0.0599, 0.065, 0.0709, 0.1538, 0.1533, 0.1782, 0.1913, 0.1741, 0.1795, 0.1811, 0.1995, 0.2033, 0.2227, 0.2753, 0.2789, 0.3039, 0.2712, 0.2378, 0.2349, 0.1827, 0.2264, 0.2302, 0.2686, 0.275, 0.2725, 0.3064, 0.293, 0.2644, 0.2899, 0.2997, 0.3196, 0.3209, 0.3384, 0.3647, 0.3365, 0.3487, 0.3363, 0.3197, 0.3033, 0.2676, 0.2908, 0.291, 0.2982, 0.2832, 0.2724, 0.2758, 0.26, 0.2906, 0.2953, 0.214, 0.2501, 0.2523, 0.3237, 0.306, 0.3151, 0.349, 0.367, 0.4024, 0.3935, 0.3852, 0.4021, 0.3871, 0.3771, 0.3697, 0.3469, 0.3552, 0.3226, 0.2723, 0.2995, 0.3124, 0.3043, 0.3148, 0.3087, 0.3087, 0.3087, 0.32, 0.3369, 0.3413, 0.3413, 0.4357]}
{"unit_id": "quote:1940201963817509063", "ticker_idx": 0, "ts": "2025-07-02T00:12:50+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's CoWoS capacity for next year will be in the range of 90-95k, aligning with previous estimates and representing a year-on-year increase of over 30%", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC CoWoS capacity up 30%+ YoY; Alchip sole Trainium3 supplier with T4 win likely; Broadcom and Marvell also beneficiaries of rising ASIC/CoWoS demand.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report\n\nMajor U.S. banks have confirmed that TSMC's CoWoS capacity for next year will be in the range of 90-95k, aligning with previous estimates and representing a year-on-year increase of over 30%. Notably, CoWoS-L capacity is expected to expand to 68k (up from a previous estimate of 55k), indicating strong demand for Blackwell or Rubin. Furthermore, the testing time for Blackwell chips will revert to 800-900 seconds from the current 600-700 seconds, as new testing machines begin operation.\n\nRegarding ASICs: It's observed that Alchip is the sole supplier for Trainium3 XPU, with revenue contributions set to begin next year. Marvell will focus on \"XPU-attach\" chips. The decision for T4 is expected by the end of this month, and brokers anticipate Alchip still has a higher chance of winning.\n\nTurning to TSMC's CoWoS customers for next year:\n\nNvidia: Total demand for 2026 remains unchanged at 580k units. CoWoS-L estimates have been revised upwards to 550k from 540k, while CoWoS-S estimated orders have been revised downwards to 0, as the H20 chip has not yet received U.S. export permission.\n\nBroadcom: CoWoS usage has been revised upwards to 110k units, driven by increased demand for Meta's MTIAv3 chips.\n\nAWS + Alchip: CoWoS usage has been revised upwards to 40k from 30k, due to the ramp-up of Trainium3.", "reply_text": null, "parent_text": null, "quoted_text": "早安！7/2 外電綜合整理\n\n- 緯創/緯穎、廣達：\n美系大行維持6月初的看法不變：針對美系CSP的GB200訂單，緯創提供L10 compute tray, 廣達是L11 組裝與測試。券商預期明年該客戶會下更多訂單，而商業模式也會改變成大家自行處理compute tray, 組裝與測試，此舉可以減少L10 與 L11", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.004512818660959983, "alpha_c_m1d": 0.01923555839228719, "bench_spy_m1d": -0.004512818660959983, "bench_c_m1d": -0.01923555839228719, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004608290215061039, "ret_signed_p1d": 0.004608290215061039, "alpha_spy_p1d": -0.003273095739813936, "alpha_c_p1d": -0.0038184571649411136, "bench_spy_p1d": 0.007881385954874975, "bench_c_p1d": 0.008426747380002153, "ret_p1w": 0.004608290215061039, "ret_signed_p1w": 0.004608290215061039, "alpha_spy_p1w": -0.0012099908820830674, "alpha_c_p1w": -0.01021851846620847, "bench_spy_p1w": 0.005818281097144107, "bench_c_p1w": 0.01482680868126951, "ret_p1m": 0.06912446692944307, "ret_signed_p1m": 0.06912446692944307, "alpha_spy_p1m": 0.050379935084899685, "alpha_c_p1m": 0.04235109484163413, "bench_spy_p1m": 0.01874453184454339, "bench_c_p1m": 0.026773372087808944, "ret_p3m": 0.20294931028578023, "ret_signed_p3m": 0.20294931028578023, "alpha_spy_p3m": 0.13030849539309175, "alpha_c_p3m": 0.055713723158523853, "bench_spy_p3m": 0.07264081489268848, "bench_c_p3m": 0.14723558712725637, "ret_p6m": 0.38800304998338797, "ret_signed_p6m": 0.38800304998338797, "alpha_spy_p6m": 0.26891261374821895, "alpha_c_p6m": 0.08914728951460749, "bench_spy_p6m": 0.11909043623516902, "bench_c_p6m": 0.2988557604687805, "price_path": [-0.1355, -0.2217, -0.2511, -0.2796, -0.208, -0.1841, -0.2061, -0.1951, -0.2153, -0.2227, -0.2199, -0.2337, -0.2511, -0.1988, -0.2071, -0.185, -0.1768, -0.1722, -0.1667, -0.1667, -0.1281, -0.1391, -0.1557, -0.1483, -0.1575, -0.129, -0.1217, -0.1107, -0.0832, -0.0887, -0.0841, -0.0969, -0.1024, -0.0914, -0.0988, -0.0969, -0.1043, -0.1144, -0.1125, -0.1125, -0.1125, -0.1318, -0.1281, -0.0914, -0.0841, -0.0868, -0.0776, -0.0409, -0.0226, -0.0369, -0.0507, -0.0553, -0.0369, -0.0276, -0.0461, -0.0276, -0.0599, -0.0323, -0.0138, -0.0092, -0.0046, -0.023, 0.0, 0.0, 0.0046, 0.0, -0.0046, -0.0046, 0.0046, 0.0138, 0.0138, 0.0092, 0.023, 0.0415, 0.0415, 0.0645, 0.0599, 0.0415, 0.0553, 0.0553, 0.0553, 0.0553, 0.0461, 0.0645, 0.0691, 0.0691, 0.0461, 0.0599, 0.0369, 0.0876, 0.0829, 0.0876, 0.0876, 0.106, 0.0829, 0.0876, 0.0876, 0.0922, 0.0461, 0.0599, 0.0461, 0.0783, 0.0829, 0.0968, 0.0691, 0.0691, 0.0737, 0.0691, 0.0691, 0.0691, 0.0876, 0.0876, 0.106, 0.129, 0.1429, 0.1613, 0.1567, 0.1844, 0.1706, 0.1891, 0.1706, 0.1983, 0.24, 0.24, 0.2215, 0.2029, 0.2029, 0.2076, 0.2261, 0.2631, 0.2955, 0.2955, 0.3279, 0.3094, 0.3325, 0.3325, 0.3094, 0.3186, 0.3556, 0.3741, 0.3418, 0.3695, 0.3695, 0.351, 0.3418, 0.3418, 0.3695, 0.3649, 0.3926, 0.3926, 0.388, 0.3973, 0.3926, 0.351, 0.3556, 0.351, 0.3649, 0.3556, 0.3649, 0.351, 0.3232, 0.3371, 0.3001, 0.2909, 0.3464, 0.2816, 0.2724, 0.3094, 0.3325, 0.3279, 0.3325, 0.3047, 0.3232, 0.3418, 0.3371, 0.351, 0.3834, 0.3695, 0.3926, 0.3648, 0.3741, 0.3462, 0.3323, 0.3277, 0.3277, 0.3277, 0.3602, 0.3834, 0.388, 0.388]}
{"unit_id": "quote:1940201963817509063", "ticker_idx": 1, "ts": "2025-07-02T00:12:50+00:00", "symbol": "Alchip", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Alchip is the sole supplier for Trainium3 XPU, with revenue contributions set to begin next year... brokers anticipate Alchip still has a higher chance of winning", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC CoWoS capacity up 30%+ YoY; Alchip sole Trainium3 supplier with T4 win likely; Broadcom and Marvell also beneficiaries of rising ASIC/CoWoS demand.", "resolved_tickers": ["3661.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Alchip Technologies 3661.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report\n\nMajor U.S. banks have confirmed that TSMC's CoWoS capacity for next year will be in the range of 90-95k, aligning with previous estimates and representing a year-on-year increase of over 30%. Notably, CoWoS-L capacity is expected to expand to 68k (up from a previous estimate of 55k), indicating strong demand for Blackwell or Rubin. Furthermore, the testing time for Blackwell chips will revert to 800-900 seconds from the current 600-700 seconds, as new testing machines begin operation.\n\nRegarding ASICs: It's observed that Alchip is the sole supplier for Trainium3 XPU, with revenue contributions set to begin next year. Marvell will focus on \"XPU-attach\" chips. The decision for T4 is expected by the end of this month, and brokers anticipate Alchip still has a higher chance of winning.\n\nTurning to TSMC's CoWoS customers for next year:\n\nNvidia: Total demand for 2026 remains unchanged at 580k units. CoWoS-L estimates have been revised upwards to 550k from 540k, while CoWoS-S estimated orders have been revised downwards to 0, as the H20 chip has not yet received U.S. export permission.\n\nBroadcom: CoWoS usage has been revised upwards to 110k units, driven by increased demand for Meta's MTIAv3 chips.\n\nAWS + Alchip: CoWoS usage has been revised upwards to 40k from 30k, due to the ramp-up of Trainium3.", "reply_text": null, "parent_text": null, "quoted_text": "早安！7/2 外電綜合整理\n\n- 緯創/緯穎、廣達：\n美系大行維持6月初的看法不變：針對美系CSP的GB200訂單，緯創提供L10 compute tray, 廣達是L11 組裝與測試。券商預期明年該客戶會下更多訂單，而商業模式也會改變成大家自行處理compute tray, 組裝與測試，此舉可以減少L10 與 L11", "parent_author": null, "ret_m1d": -0.0015674116816670747, "ret_signed_m1d": -0.0015674116816670747, "alpha_spy_m1d": 0.0029454069792929083, "alpha_c_m1d": 0.017668146710620114, "bench_spy_m1d": -0.004512818660959983, "bench_c_m1d": -0.01923555839228719, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004702157786933392, "ret_signed_p1d": -0.004702157786933392, "alpha_spy_p1d": -0.012583543741808367, "alpha_c_p1d": -0.013128905166935545, "bench_spy_p1d": 0.007881385954874975, "bench_c_p1d": 0.008426747380002153, "ret_p1w": 0.05485894530994084, "ret_signed_p1w": 0.05485894530994084, "alpha_spy_p1w": 0.049040664212796736, "alpha_c_p1w": 0.04003213662867133, "bench_spy_p1w": 0.005818281097144107, "bench_c_p1w": 0.01482680868126951, "ret_p1m": 0.2178683695580963, "ret_signed_p1m": 0.2178683695580963, "alpha_spy_p1m": 0.1991238377135529, "alpha_c_p1m": 0.19109499747028735, "bench_spy_p1m": 0.01874453184454339, "bench_c_p1m": 0.026773372087808944, "ret_p3m": 0.036370626111743576, "ret_signed_p3m": 0.036370626111743576, "alpha_spy_p3m": -0.0362701887809449, "alpha_c_p3m": -0.1108649610155128, "bench_spy_p3m": 0.07264081489268848, "bench_c_p3m": 0.14723558712725637, "ret_p6m": 0.0063079667283494345, "ret_signed_p6m": 0.0063079667283494345, "alpha_spy_p6m": -0.11278246950681958, "alpha_c_p6m": -0.29254779374043105, "bench_spy_p6m": 0.11909043623516902, "bench_c_p6m": 0.2988557604687805, "price_path": [-0.1661, -0.2492, -0.3229, -0.3903, -0.3307, -0.3339, -0.3009, -0.2602, -0.3197, -0.3056, -0.3088, -0.3401, -0.3746, -0.3464, -0.3386, -0.3229, -0.3135, -0.3182, -0.3339, -0.3339, -0.279, -0.2868, -0.2868, -0.3009, -0.2649, -0.279, -0.2069, -0.1552, -0.1364, -0.1473, -0.1442, -0.1834, -0.1019, -0.0658, -0.0909, -0.069, -0.105, -0.1034, -0.105, -0.1191, -0.1191, -0.1473, -0.1473, -0.1207, -0.1301, -0.1238, -0.1458, -0.1364, -0.1097, -0.1285, -0.174, -0.174, -0.1176, -0.1207, -0.1003, -0.1223, -0.1129, -0.1317, -0.0972, -0.0972, -0.0596, -0.0298, -0.0016, 0.0, -0.0047, 0.0, 0.0094, 0.0078, 0.0549, 0.0846, 0.0752, 0.0376, 0.0564, 0.1285, 0.1771, 0.232, 0.185, 0.1661, 0.1991, 0.2461, 0.2179, 0.2163, 0.1959, 0.1755, 0.2179, 0.2179, 0.1771, 0.1755, 0.163, 0.1803, 0.2038, 0.2257, 0.2774, 0.2915, 0.3323, 0.3495, 0.326, 0.2649, 0.1473, 0.1865, 0.1567, 0.221, 0.2476, 0.279, 0.2539, 0.2712, 0.2273, 0.2167, 0.2595, 0.2183, 0.2563, 0.2579, 0.2389, 0.2468, 0.2421, 0.1993, 0.1788, 0.1645, 0.1898, 0.1772, 0.1867, 0.1898, 0.1709, 0.0791, 0.0664, 0.0364, 0.0364, 0.0965, 0.0743, 0.0395, 0.0474, 0.0474, 0.0949, 0.0902, 0.1123, 0.1123, 0.0664, -0.0269, -0.0253, -0.0206, -0.0238, 0.0237, 0.0617, 0.019, -0.0032, -0.0032, -0.0301, -0.0301, 0.0285, 0.0617, 0.106, 0.0917, 0.0902, 0.0569, 0.1614, 0.1281, 0.1756, 0.1503, 0.1139, 0.1155, 0.1107, 0.1424, 0.0285, 0.0506, -0.0269, -0.0538, -0.0601, -0.0665, -0.0158, 0.0395, 0.0459, 0.0411, 0.0823, 0.0095, 0.0031, -0.0016, -0.0127, 0.0316, 0.0664, 0.0427, 0.019, 0.0031, -0.0459, 0.0126, 0.0031, 0.0174, 0.0205, 0.019, 0.0063, 0.0063]}
{"unit_id": "quote:1940201963817509063", "ticker_idx": 2, "ts": "2025-07-02T00:12:50+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell will focus on 'XPU-attach' chips", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC CoWoS capacity up 30%+ YoY; Alchip sole Trainium3 supplier with T4 win likely; Broadcom and Marvell also beneficiaries of rising ASIC/CoWoS demand.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report\n\nMajor U.S. banks have confirmed that TSMC's CoWoS capacity for next year will be in the range of 90-95k, aligning with previous estimates and representing a year-on-year increase of over 30%. Notably, CoWoS-L capacity is expected to expand to 68k (up from a previous estimate of 55k), indicating strong demand for Blackwell or Rubin. Furthermore, the testing time for Blackwell chips will revert to 800-900 seconds from the current 600-700 seconds, as new testing machines begin operation.\n\nRegarding ASICs: It's observed that Alchip is the sole supplier for Trainium3 XPU, with revenue contributions set to begin next year. Marvell will focus on \"XPU-attach\" chips. The decision for T4 is expected by the end of this month, and brokers anticipate Alchip still has a higher chance of winning.\n\nTurning to TSMC's CoWoS customers for next year:\n\nNvidia: Total demand for 2026 remains unchanged at 580k units. CoWoS-L estimates have been revised upwards to 550k from 540k, while CoWoS-S estimated orders have been revised downwards to 0, as the H20 chip has not yet received U.S. export permission.\n\nBroadcom: CoWoS usage has been revised upwards to 110k units, driven by increased demand for Meta's MTIAv3 chips.\n\nAWS + Alchip: CoWoS usage has been revised upwards to 40k from 30k, due to the ramp-up of Trainium3.", "reply_text": null, "parent_text": null, "quoted_text": "早安！7/2 外電綜合整理\n\n- 緯創/緯穎、廣達：\n美系大行維持6月初的看法不變：針對美系CSP的GB200訂單，緯創提供L10 compute tray, 廣達是L11 組裝與測試。券商預期明年該客戶會下更多訂單，而商業模式也會改變成大家自行處理compute tray, 組裝與測試，此舉可以減少L10 與 L11", "parent_author": null, "ret_m1d": 0.026801239705056545, "ret_signed_m1d": 0.026801239705056545, "alpha_spy_m1d": 0.03131405836601653, "alpha_c_m1d": 0.046036798097343734, "bench_spy_m1d": -0.004512818660959983, "bench_c_m1d": -0.01923555839228719, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01252522059618233, "ret_signed_p1d": 0.01252522059618233, "alpha_spy_p1d": 0.004643834641307354, "alpha_c_p1d": 0.004098473216180176, "bench_spy_p1d": 0.007881385954874975, "bench_c_p1d": 0.008426747380002153, "ret_p1w": -0.02680134273562862, "ret_signed_p1w": -0.02680134273562862, "alpha_spy_p1w": -0.032619623832772726, "alpha_c_p1w": -0.04162815141689813, "bench_spy_p1w": 0.005818281097144107, "bench_c_p1w": 0.01482680868126951, "ret_p1m": 0.08331031458633698, "ret_signed_p1m": 0.08331031458633698, "alpha_spy_p1m": 0.06456578274179359, "alpha_c_p1m": 0.05653694249852803, "bench_spy_p1m": 0.01874453184454339, "bench_c_p1m": 0.026773372087808944, "ret_p3m": 0.1105378947989859, "ret_signed_p3m": 0.1105378947989859, "alpha_spy_p3m": 0.037897079906297426, "alpha_c_p3m": -0.03669769232827047, "bench_spy_p3m": 0.07264081489268848, "bench_c_p3m": 0.14723558712725637, "ret_p6m": 0.16657375026234167, "ret_signed_p6m": 0.16657375026234167, "alpha_spy_p6m": 0.04748331402717265, "alpha_c_p6m": -0.13228201020643882, "bench_spy_p6m": 0.11909043623516902, "bench_c_p6m": 0.2988557604687805, "price_path": [-0.335, -0.3139, -0.327, -0.1799, -0.2888, -0.2809, -0.2962, -0.2819, -0.3006, -0.3037, -0.3037, -0.3349, -0.3182, -0.2758, -0.2279, -0.2065, -0.2094, -0.2097, -0.2139, -0.1791, -0.1605, -0.1653, -0.1755, -0.2416, -0.2238, -0.1966, -0.1313, -0.1181, -0.1125, -0.1219, -0.1413, -0.1574, -0.1728, -0.1904, -0.167, -0.1826, -0.1826, -0.1405, -0.1301, -0.1417, -0.1894, -0.1721, -0.1601, -0.1071, -0.1224, -0.0795, -0.0688, -0.0729, -0.0809, -0.0621, -0.0951, -0.0516, -0.0574, 0.0094, 0.0094, -0.01, -0.0467, 0.0129, 0.0226, 0.077, 0.0392, 0.0424, 0.0268, 0.0, 0.0125, 0.0125, -0.0364, -0.031, -0.0268, -0.012, -0.0199, -0.0226, -0.024, -0.045, -0.0294, 0.0062, -0.0152, -0.0296, -0.0124, -0.002, 0.0003, 0.0232, 0.029, 0.1018, 0.0833, 0.0035, 0.0316, 0.0329, 0.0152, 0.0224, 0.0425, 0.0417, 0.0488, 0.0692, 0.0654, 0.027, 0.0344, -0.0286, -0.04, -0.0402, -0.016, -0.0167, 0.001, 0.0081, 0.041, -0.1526, -0.1526, -0.1293, -0.1601, -0.136, -0.1464, -0.1104, -0.0991, -0.0956, -0.1024, -0.0922, -0.0911, -0.0718, -0.0433, 0.0005, 0.001, 0.0181, 0.0058, 0.0795, 0.1297, 0.1211, 0.1105, 0.1332, 0.1308, 0.1619, 0.1622, 0.1986, 0.1723, 0.2468, 0.2223, 0.1547, 0.2057, 0.1629, 0.1989, 0.19, 0.1863, 0.1578, 0.1365, 0.0932, 0.1164, 0.1347, 0.1965, 0.1931, 0.2159, 0.1946, 0.2644, 0.2189, 0.1814, 0.253, 0.2588, 0.2263, 0.2575, 0.2049, 0.2049, 0.1805, 0.166, 0.1256, 0.0612, 0.0968, 0.0343, 0.0446, 0.1302, 0.1253, 0.1832, 0.1832, 0.2058, 0.2288, 0.2529, 0.3515, 0.3244, 0.3341, 0.2409, 0.1991, 0.2472, 0.2062, 0.1388, 0.1365, 0.1339, 0.102, 0.1393, 0.1342, 0.1438, 0.1826, 0.1666, 0.1666]}
{"unit_id": "quote:1940201963817509063", "ticker_idx": 3, "ts": "2025-07-02T00:12:50+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom: CoWoS usage has been revised upwards to 110k units, driven by increased demand for Meta's MTIAv3 chips", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC CoWoS capacity up 30%+ YoY; Alchip sole Trainium3 supplier with T4 win likely; Broadcom and Marvell also beneficiaries of rising ASIC/CoWoS demand.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report\n\nMajor U.S. banks have confirmed that TSMC's CoWoS capacity for next year will be in the range of 90-95k, aligning with previous estimates and representing a year-on-year increase of over 30%. Notably, CoWoS-L capacity is expected to expand to 68k (up from a previous estimate of 55k), indicating strong demand for Blackwell or Rubin. Furthermore, the testing time for Blackwell chips will revert to 800-900 seconds from the current 600-700 seconds, as new testing machines begin operation.\n\nRegarding ASICs: It's observed that Alchip is the sole supplier for Trainium3 XPU, with revenue contributions set to begin next year. Marvell will focus on \"XPU-attach\" chips. The decision for T4 is expected by the end of this month, and brokers anticipate Alchip still has a higher chance of winning.\n\nTurning to TSMC's CoWoS customers for next year:\n\nNvidia: Total demand for 2026 remains unchanged at 580k units. CoWoS-L estimates have been revised upwards to 550k from 540k, while CoWoS-S estimated orders have been revised downwards to 0, as the H20 chip has not yet received U.S. export permission.\n\nBroadcom: CoWoS usage has been revised upwards to 110k units, driven by increased demand for Meta's MTIAv3 chips.\n\nAWS + Alchip: CoWoS usage has been revised upwards to 40k from 30k, due to the ramp-up of Trainium3.", "reply_text": null, "parent_text": null, "quoted_text": "早安！7/2 外電綜合整理\n\n- 緯創/緯穎、廣達：\n美系大行維持6月初的看法不變：針對美系CSP的GB200訂單，緯創提供L10 compute tray, 廣達是L11 組裝與測試。券商預期明年該客戶會下更多訂單，而商業模式也會改變成大家自行處理compute tray, 組裝與測試，此舉可以減少L10 與 L11", "parent_author": null, "ret_m1d": -0.019118231705431032, "ret_signed_m1d": -0.019118231705431032, "alpha_spy_m1d": -0.01460541304447105, "alpha_c_m1d": 0.00011732668685615621, "bench_spy_m1d": -0.004512818660959983, "bench_c_m1d": -0.01923555839228719, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01956287083585795, "ret_signed_p1d": 0.01956287083585795, "alpha_spy_p1d": 0.011681484880982973, "alpha_c_p1d": 0.011136123455855795, "bench_spy_p1d": 0.007881385954874975, "bench_c_p1d": 0.008426747380002153, "ret_p1w": 0.029640713387663498, "ret_signed_p1w": 0.029640713387663498, "alpha_spy_p1w": 0.02382243229051939, "alpha_c_p1w": 0.014813904706393988, "bench_spy_p1w": 0.005818281097144107, "bench_c_p1w": 0.01482680868126951, "ret_p1m": 0.08818092332101535, "ret_signed_p1m": 0.08818092332101535, "alpha_spy_p1m": 0.06943639147647196, "alpha_c_p1m": 0.0614075512332064, "bench_spy_p1m": 0.01874453184454339, "bench_c_p1m": 0.026773372087808944, "ret_p3m": 0.21697604907542356, "ret_signed_p3m": 0.21697604907542356, "alpha_spy_p3m": 0.14433523418273508, "alpha_c_p3m": 0.06974046194816719, "bench_spy_p3m": 0.07264081489268848, "bench_c_p3m": 0.14723558712725637, "ret_p6m": 0.30230207076744486, "ret_signed_p6m": 0.30230207076744486, "alpha_spy_p6m": 0.18321163453227585, "alpha_c_p6m": 0.003446310298664379, "bench_spy_p6m": 0.11909043623516902, "bench_c_p6m": 0.2988557604687805, "price_path": [-0.4593, -0.4302, -0.4233, -0.3156, -0.3631, -0.3275, -0.3407, -0.3385, -0.3546, -0.368, -0.368, -0.3856, -0.3732, -0.3461, -0.3045, -0.2891, -0.2886, -0.2934, -0.2886, -0.2706, -0.2473, -0.2581, -0.2604, -0.2429, -0.232, -0.2304, -0.181, -0.1409, -0.142, -0.1401, -0.155, -0.1475, -0.1436, -0.1508, -0.1479, -0.1546, -0.1546, -0.1289, -0.115, -0.1056, -0.1052, -0.0807, -0.0506, -0.035, -0.0392, -0.0873, -0.097, -0.0958, -0.0651, -0.0535, -0.0807, -0.0681, -0.0782, -0.0712, -0.0712, -0.0738, -0.0598, -0.0227, -0.0195, 0.001, -0.002, 0.0213, -0.0191, 0.0, 0.0196, 0.0196, 0.0159, 0.007, 0.0296, 0.0204, 0.0166, 0.0211, 0.0409, 0.0404, 0.0613, 0.0498, 0.0678, 0.0322, 0.0511, 0.0697, 0.0751, 0.0904, 0.102, 0.1212, 0.0882, 0.0694, 0.1031, 0.0853, 0.1177, 0.1255, 0.1299, 0.126, 0.1591, 0.1452, 0.1531, 0.135, 0.1329, 0.0927, 0.0788, 0.073, 0.0893, 0.0901, 0.1041, 0.1124, 0.1436, 0.1019, 0.1019, 0.105, 0.1204, 0.1341, 0.2408, 0.2807, 0.2474, 0.3693, 0.3325, 0.3333, 0.349, 0.3338, 0.2826, 0.2795, 0.278, 0.2574, 0.258, 0.2593, 0.2474, 0.2416, 0.217, 0.2244, 0.2374, 0.2551, 0.2558, 0.2451, 0.2486, 0.2823, 0.2805, 0.2048, 0.3239, 0.2772, 0.3039, 0.3144, 0.2965, 0.2962, 0.2718, 0.263, 0.2778, 0.3143, 0.3437, 0.3842, 0.4325, 0.3972, 0.3719, 0.3456, 0.3062, 0.3323, 0.3197, 0.2969, 0.3301, 0.3063, 0.3184, 0.2618, 0.271, 0.2717, 0.2637, 0.3154, 0.2872, 0.2626, 0.4028, 0.429, 0.4756, 0.4756, 0.4956, 0.4329, 0.4162, 0.4126, 0.4142, 0.4483, 0.4887, 0.5079, 0.5327, 0.5082, 0.3359, 0.2612, 0.2667, 0.21, 0.2243, 0.2632, 0.2697, 0.299, 0.3023, 0.3023]}
{"unit_id": "orig:1923162956398985547", "ticker_idx": 0, "ts": "2025-05-15T23:45:54+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential", "tweet_type": "analysis", "timeline": "quarters", "summary": "Morgan Stanley W-shaped memory cycle: long Samsung/Hynix/Phison/Longsys/Winbond/GigaDevice; DRAM sector faces 2H correction with prices declining from 4Q25.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "「Asia Technology: Memory Outlook – This Time It’s W-shaped」 (Morgan Stanley)\n\n📌 Memory Market Outlook: “This Time It’s a W-shaped Recovery”\n\n⸻\n\n1. Industry Structure and Cycle Overview\n\n• Traditionally a supply-driven cycle, the memory industry is now shifting toward a demand-driven cycle.\n\n• Expectation of pre-purchasing (inventory build-up) in 1H followed by demand rebound in 2H.\n\n• Stock prices have rebounded, but the recovery is mostly based on lagging indicators. Real demand remains uncertain.\n\n⸻\n\n2. Price Trends and Outlook\n\n• 2Q25 DRAM: Currently up +5%, DDR4 surging +20% due to inventory build-up.\n\n• 3Q25: Prices expected to flatten; 4Q25 may see a downturn.\n\n• NAND: +6% in 2Q, +4% in 3Q, flat in 4Q. Production cuts help support pricing.\n\n⸻\n\n3. HBM Market Outlook (2025)\n\n• 2025 HBM TAM revised down from 180B Gb to 155B Gb.\n\n• Key reasons:\n\n• Export controls impacting products like NVIDIA H20\n\n• HBM3E inventory overhang and weakening customer demand\n\n• Product lineup adjustments by NVIDIA and AMD (shifting to GDDR7)\n\n• Samsung may convert HBM4 capacity to commodity DRAM if it fails to secure orders — raising oversupply concerns in DRAM.\n\n⸻\n\n4. DRAM Segment Analysis\n\n• Smartphone/PC DRAM prices: Expected to rise +3~8% QoQ in 2Q25.\n\n• Server DRAM: Up +3% in 2Q25, with strong demand from China.\n\n• DDR4/LPDDR4x: Samsung and SK Hynix plan to phase out production, causing a supply squeeze and sharp price increases (+20%).\n\n⸻\n\n5. NAND Market Analysis\n\n• 3Q25 wafer prices: Up +4% due to production cuts by major players.\n\n• Post-4Q25: Rising price pressure anticipated from increased competition by Chinese players.\n\n• eSSD: Solid demand from U.S. CSPs; China still focused on HDD.\n\n• Smartphones: UFS/eMMC up +5% in 2Q; channel inventory up +10%+.\n\n⸻\n\n6. Strategy and Stock-Specific Investment Highlights\n\n• DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H.\n\n• Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential.\n\n• SK Hynix: HBM leader, reflected in premium P/B of 1.9x.\n\n• NAND: Prefer self-sustaining growth stories like Phison, Longsys.\n\n• NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions.\n\n• HBM: Strategically preferred on the back of AI-driven demand.\n\n⸻\n\n7. Key Reasons Behind the W-Shaped Cycle\n\n• Inventory build-up in 1H may lead to a demand gap in 2H.\n\n• Current rebound is temporary, driven by price hikes without real demand.\n\n• Prices expected to decline again starting in 4Q25, with the bottom likely around 2Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0017451100151208987, "ret_signed_m1d": 0.0017451100151208987, "alpha_spy_m1d": 0.0066056070178771575, "alpha_c_m1d": 0.003109918087799257, "bench_spy_m1d": -0.004860497002756259, "bench_c_m1d": -0.0013648080726783585, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008726105069844614, "ret_signed_p1d": -0.008726105069844614, "alpha_spy_p1d": -0.015060078760344986, "alpha_c_p1d": -0.010602838898430056, "bench_spy_p1d": 0.006333973690500372, "bench_c_p1d": 0.0018767338285854418, "ret_p1w": -0.0453752884929437, "ret_signed_p1w": -0.0453752884929437, "alpha_spy_p1w": -0.03289344687396578, "alpha_c_p1w": -0.02357959615737748, "bench_spy_p1w": -0.012481841618977918, "bench_c_p1w": -0.02179569233556622, "ret_p1m": 0.017451932642569057, "ret_signed_p1m": 0.017451932642569057, "alpha_spy_p1m": 0.006375816849655047, "alpha_c_p1m": -0.002680344626526665, "bench_spy_p1m": 0.01107611579291401, "bench_c_p1m": 0.020132277269095722, "ret_p3m": 0.2484485829748302, "ret_signed_p3m": 0.2484485829748302, "alpha_spy_p3m": 0.15677424380188842, "alpha_c_p3m": 0.1025559607930564, "bench_spy_p3m": 0.09167433917294177, "bench_c_p3m": 0.1458926221817738, "ret_p6m": 0.7267009949104253, "ret_signed_p6m": 0.7267009949104253, "alpha_spy_p6m": 0.5838306334658514, "alpha_c_p6m": 0.49399241478340805, "bench_spy_p6m": 0.14287036144457388, "bench_c_p6m": 0.23270858012701723, "price_path": [-0.0285, -0.0128, 0.0184, 0.0132, 0.0097, -0.0059, -0.0076, -0.0181, -0.0233, -0.0545, -0.0545, -0.0545, -0.0632, -0.058, -0.0684, -0.0684, -0.0701, -0.0475, -0.051, -0.051, -0.0007, -0.0007, 0.0149, 0.0444, 0.0704, 0.0496, 0.0375, 0.0652, 0.0722, 0.0506, 0.0087, 0.0262, 0.0262, 0.0052, -0.0209, -0.0716, -0.0663, -0.075, -0.0157, -0.0366, -0.0192, -0.0122, -0.0454, -0.0384, -0.0349, -0.0332, -0.0401, -0.0279, -0.0279, -0.0279, -0.0262, -0.0262, -0.0314, -0.0314, -0.0524, -0.0524, -0.0524, -0.0471, -0.0471, -0.0436, 0.0052, -0.007, 0.0017, 0.0, -0.0087, -0.0262, -0.0244, -0.0279, -0.0454, -0.0541, -0.0454, -0.0593, -0.0244, -0.0209, -0.0192, -0.0087, -0.0087, 0.0087, 0.0314, 0.0314, 0.0436, 0.0332, 0.0454, 0.0384, 0.0175, -0.0017, 0.014, 0.0436, 0.0332, 0.0384, 0.0122, 0.0558, 0.0698, 0.0506, 0.0676, 0.05, 0.0571, 0.0676, 0.1203, 0.1115, 0.0834, 0.0781, 0.0606, 0.0711, 0.0992, 0.0974, 0.1185, 0.1361, 0.1712, 0.1782, 0.1905, 0.1589, 0.1659, 0.1589, 0.1571, 0.2362, 0.2397, 0.2748, 0.2537, 0.2098, 0.2239, 0.2274, 0.2081, 0.2379, 0.2607, 0.2467, 0.2484, 0.2625, 0.2572, 0.2572, 0.2291, 0.2291, 0.2379, 0.2397, 0.2537, 0.2555, 0.2344, 0.2397, 0.2221, 0.2239, 0.187, 0.2133, 0.2256, 0.2309, 0.2204, 0.2309, 0.2555, 0.2748, 0.2888, 0.324, 0.3433, 0.3942, 0.3731, 0.41, 0.41, 0.4662, 0.4873, 0.4995, 0.5118, 0.4627, 0.4851, 0.4798, 0.5168, 0.583, 0.583, 0.583, 0.583, 0.583, 0.583, 0.665, 0.6456, 0.6156, 0.6756, 0.7232, 0.7267, 0.7302, 0.7196, 0.739, 0.702, 0.7426, 0.799, 0.7549, 0.7726, 0.8361, 0.896, 0.9595, 0.8502, 0.7743, 0.7496, 0.7267]}
{"unit_id": "orig:1923162956398985547", "ticker_idx": 1, "ts": "2025-05-15T23:45:54+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix: HBM leader, reflected in premium P/B of 1.9x", "tweet_type": "analysis", "timeline": "quarters", "summary": "Morgan Stanley W-shaped memory cycle: long Samsung/Hynix/Phison/Longsys/Winbond/GigaDevice; DRAM sector faces 2H correction with prices declining from 4Q25.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "「Asia Technology: Memory Outlook – This Time It’s W-shaped」 (Morgan Stanley)\n\n📌 Memory Market Outlook: “This Time It’s a W-shaped Recovery”\n\n⸻\n\n1. Industry Structure and Cycle Overview\n\n• Traditionally a supply-driven cycle, the memory industry is now shifting toward a demand-driven cycle.\n\n• Expectation of pre-purchasing (inventory build-up) in 1H followed by demand rebound in 2H.\n\n• Stock prices have rebounded, but the recovery is mostly based on lagging indicators. Real demand remains uncertain.\n\n⸻\n\n2. Price Trends and Outlook\n\n• 2Q25 DRAM: Currently up +5%, DDR4 surging +20% due to inventory build-up.\n\n• 3Q25: Prices expected to flatten; 4Q25 may see a downturn.\n\n• NAND: +6% in 2Q, +4% in 3Q, flat in 4Q. Production cuts help support pricing.\n\n⸻\n\n3. HBM Market Outlook (2025)\n\n• 2025 HBM TAM revised down from 180B Gb to 155B Gb.\n\n• Key reasons:\n\n• Export controls impacting products like NVIDIA H20\n\n• HBM3E inventory overhang and weakening customer demand\n\n• Product lineup adjustments by NVIDIA and AMD (shifting to GDDR7)\n\n• Samsung may convert HBM4 capacity to commodity DRAM if it fails to secure orders — raising oversupply concerns in DRAM.\n\n⸻\n\n4. DRAM Segment Analysis\n\n• Smartphone/PC DRAM prices: Expected to rise +3~8% QoQ in 2Q25.\n\n• Server DRAM: Up +3% in 2Q25, with strong demand from China.\n\n• DDR4/LPDDR4x: Samsung and SK Hynix plan to phase out production, causing a supply squeeze and sharp price increases (+20%).\n\n⸻\n\n5. NAND Market Analysis\n\n• 3Q25 wafer prices: Up +4% due to production cuts by major players.\n\n• Post-4Q25: Rising price pressure anticipated from increased competition by Chinese players.\n\n• eSSD: Solid demand from U.S. CSPs; China still focused on HDD.\n\n• Smartphones: UFS/eMMC up +5% in 2Q; channel inventory up +10%+.\n\n⸻\n\n6. Strategy and Stock-Specific Investment Highlights\n\n• DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H.\n\n• Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential.\n\n• SK Hynix: HBM leader, reflected in premium P/B of 1.9x.\n\n• NAND: Prefer self-sustaining growth stories like Phison, Longsys.\n\n• NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions.\n\n• HBM: Strategically preferred on the back of AI-driven demand.\n\n⸻\n\n7. Key Reasons Behind the W-Shaped Cycle\n\n• Inventory build-up in 1H may lead to a demand gap in 2H.\n\n• Current rebound is temporary, driven by price hikes without real demand.\n\n• Prices expected to decline again starting in 4Q25, with the bottom likely around 2Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02743156279590031, "ret_signed_m1d": 0.02743156279590031, "alpha_spy_m1d": 0.03229205979865657, "alpha_c_m1d": 0.023951621142182855, "bench_spy_m1d": -0.004860497002756259, "bench_c_m1d": 0.0034799416537174555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019950277377037473, "ret_signed_p1d": 0.019950277377037473, "alpha_spy_p1d": 0.013616303686537101, "alpha_c_p1d": 0.022823245261374026, "bench_spy_p1d": 0.006333973690500372, "bench_c_p1d": -0.0028729678843365525, "ret_p1w": -0.017954912531490175, "ret_signed_p1w": -0.017954912531490175, "alpha_spy_p1w": -0.005473070912512257, "alpha_c_p1w": 0.0076186885193919585, "bench_spy_p1w": -0.012481841618977918, "bench_c_p1w": -0.025573601050882133, "ret_p1m": 0.17668519819040518, "ret_signed_p1m": 0.17668519819040518, "alpha_spy_p1m": 0.16560908239749117, "alpha_c_p1m": 0.1367872639265315, "bench_spy_p1m": 0.01107611579291401, "bench_c_p1m": 0.03989793426387367, "ret_p3m": 0.34406904219508316, "ret_signed_p3m": 0.34406904219508316, "alpha_spy_p3m": 0.2523947030221414, "alpha_c_p3m": 0.12972845694973612, "bench_spy_p3m": 0.09167433917294177, "bench_c_p3m": 0.21434058524534705, "ret_p6m": 1.9021789867039964, "ret_signed_p6m": 1.9021789867039964, "alpha_spy_p6m": 1.7593086252594226, "alpha_c_p6m": 1.4935276409223743, "bench_spy_p6m": 0.14287036144457388, "bench_c_p6m": 0.4086513457816221, "price_path": [0.0506, 0.0407, 0.0828, 0.0506, 0.0382, 0.0159, -0.0064, 0.006, -0.0065, -0.0514, -0.0514, -0.0718, -0.0369, -0.0404, -0.0404, -0.0628, -0.0633, -0.008, -0.004, 0.02, 0.0274, 0.0125, 0.0249, 0.0474, 0.0748, 0.0549, 0.0374, 0.0673, 0.0324, -0.006, -0.0489, -0.0175, -0.013, -0.0294, -0.0913, -0.1781, -0.1546, -0.1771, -0.0863, -0.0983, -0.1012, -0.0993, -0.1322, -0.1272, -0.1272, -0.1192, -0.1332, -0.0973, -0.1107, -0.0803, -0.0923, -0.0983, -0.1147, -0.1147, -0.0723, -0.0723, -0.0723, -0.0484, -0.0509, -0.0519, -0.0274, -0.01, 0.0274, 0.0, 0.02, -0.0055, 0.0075, 0.0, -0.018, -0.0025, 0.0125, 0.01, 0.0374, 0.0593, 0.0218, 0.0368, 0.0368, 0.0867, 0.1217, 0.1217, 0.1442, 0.1517, 0.1992, 0.1767, 0.1767, 0.2391, 0.2441, 0.2316, 0.2291, 0.2841, 0.2966, 0.3915, 0.429, 0.464, 0.419, 0.459, 0.4265, 0.394, 0.3915, 0.3516, 0.3541, 0.409, 0.404, 0.484, 0.4715, 0.499, 0.4915, 0.479, 0.3466, 0.3441, 0.3616, 0.3416, 0.3441, 0.3466, 0.3291, 0.3091, 0.3116, 0.3166, 0.3666, 0.2891, 0.2891, 0.3166, 0.2916, 0.3091, 0.2816, 0.3341, 0.3441, 0.389, 0.3815, 0.3815, 0.3366, 0.3141, 0.2766, 0.2242, 0.2541, 0.2966, 0.3066, 0.2991, 0.3435, 0.346, 0.281, 0.3035, 0.3135, 0.3285, 0.3685, 0.386, 0.4411, 0.5211, 0.5362, 0.6437, 0.6562, 0.7413, 0.6688, 0.7788, 0.7788, 0.7563, 0.8064, 0.7888, 0.7838, 0.6838, 0.7463, 0.7388, 0.8014, 0.979, 0.979, 0.979, 0.979, 0.979, 0.979, 1.1416, 1.0766, 1.059, 1.1141, 1.2642, 1.3292, 1.4293, 1.3968, 1.4093, 1.3943, 1.5519, 1.677, 1.607, 1.7921, 1.8421, 1.7971, 2.1023, 1.9322, 1.8972, 1.9672, 1.9022]}
{"unit_id": "orig:1923162956398985547", "ticker_idx": 3, "ts": "2025-05-15T23:45:54+00:00", "symbol": "Longsys", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND: Prefer self-sustaining growth stories like Phison, Longsys", "tweet_type": "analysis", "timeline": "quarters", "summary": "Morgan Stanley W-shaped memory cycle: long Samsung/Hynix/Phison/Longsys/Winbond/GigaDevice; DRAM sector faces 2H correction with prices declining from 4Q25.", "resolved_tickers": ["301308.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Longsys Electronics 301308.SZ China NAND storage", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "「Asia Technology: Memory Outlook – This Time It’s W-shaped」 (Morgan Stanley)\n\n📌 Memory Market Outlook: “This Time It’s a W-shaped Recovery”\n\n⸻\n\n1. Industry Structure and Cycle Overview\n\n• Traditionally a supply-driven cycle, the memory industry is now shifting toward a demand-driven cycle.\n\n• Expectation of pre-purchasing (inventory build-up) in 1H followed by demand rebound in 2H.\n\n• Stock prices have rebounded, but the recovery is mostly based on lagging indicators. Real demand remains uncertain.\n\n⸻\n\n2. Price Trends and Outlook\n\n• 2Q25 DRAM: Currently up +5%, DDR4 surging +20% due to inventory build-up.\n\n• 3Q25: Prices expected to flatten; 4Q25 may see a downturn.\n\n• NAND: +6% in 2Q, +4% in 3Q, flat in 4Q. Production cuts help support pricing.\n\n⸻\n\n3. HBM Market Outlook (2025)\n\n• 2025 HBM TAM revised down from 180B Gb to 155B Gb.\n\n• Key reasons:\n\n• Export controls impacting products like NVIDIA H20\n\n• HBM3E inventory overhang and weakening customer demand\n\n• Product lineup adjustments by NVIDIA and AMD (shifting to GDDR7)\n\n• Samsung may convert HBM4 capacity to commodity DRAM if it fails to secure orders — raising oversupply concerns in DRAM.\n\n⸻\n\n4. DRAM Segment Analysis\n\n• Smartphone/PC DRAM prices: Expected to rise +3~8% QoQ in 2Q25.\n\n• Server DRAM: Up +3% in 2Q25, with strong demand from China.\n\n• DDR4/LPDDR4x: Samsung and SK Hynix plan to phase out production, causing a supply squeeze and sharp price increases (+20%).\n\n⸻\n\n5. NAND Market Analysis\n\n• 3Q25 wafer prices: Up +4% due to production cuts by major players.\n\n• Post-4Q25: Rising price pressure anticipated from increased competition by Chinese players.\n\n• eSSD: Solid demand from U.S. CSPs; China still focused on HDD.\n\n• Smartphones: UFS/eMMC up +5% in 2Q; channel inventory up +10%+.\n\n⸻\n\n6. Strategy and Stock-Specific Investment Highlights\n\n• DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H.\n\n• Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential.\n\n• SK Hynix: HBM leader, reflected in premium P/B of 1.9x.\n\n• NAND: Prefer self-sustaining growth stories like Phison, Longsys.\n\n• NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions.\n\n• HBM: Strategically preferred on the back of AI-driven demand.\n\n⸻\n\n7. Key Reasons Behind the W-Shaped Cycle\n\n• Inventory build-up in 1H may lead to a demand gap in 2H.\n\n• Current rebound is temporary, driven by price hikes without real demand.\n\n• Prices expected to decline again starting in 4Q25, with the bottom likely around 2Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03914176952542081, "ret_signed_m1d": 0.03914176952542081, "alpha_spy_m1d": 0.04400226652817707, "alpha_c_m1d": 0.04050657759809917, "bench_spy_m1d": -0.004860497002756259, "bench_c_m1d": -0.0013648080726783585, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004941545842292561, "ret_signed_p1d": -0.004941545842292561, "alpha_spy_p1d": -0.011275519532792933, "alpha_c_p1d": -0.0068182796708780025, "bench_spy_p1d": 0.006333973690500372, "bench_c_p1d": 0.0018767338285854418, "ret_p1w": -0.03224971791579234, "ret_signed_p1w": -0.03224971791579234, "alpha_spy_p1w": -0.019767876296814424, "alpha_c_p1w": -0.010454025580226123, "bench_spy_p1w": -0.012481841618977918, "bench_c_p1w": -0.02179569233556622, "ret_p1m": -0.07373209436201644, "ret_signed_p1m": -0.07373209436201644, "alpha_spy_p1m": -0.08480821015493045, "alpha_c_p1m": -0.09386437163111216, "bench_spy_p1m": 0.01107611579291401, "bench_c_p1m": 0.020132277269095722, "ret_p3m": 0.18478542005867982, "ret_signed_p3m": 0.18478542005867982, "alpha_spy_p3m": 0.09311108088573805, "alpha_c_p3m": 0.03889279787690603, "bench_spy_p3m": 0.09167433917294177, "bench_c_p3m": 0.1458926221817738, "ret_p6m": 2.6174250605465366, "ret_signed_p6m": 2.6174250605465366, "alpha_spy_p6m": 2.4745546991019625, "alpha_c_p6m": 2.3847164804195193, "bench_spy_p6m": 0.14287036144457388, "bench_c_p6m": 0.23270858012701723, "price_path": [0.2241, 0.1629, 0.2441, 0.2436, 0.2653, 0.3454, 0.3228, 0.342, 0.3594, 0.3303, 0.3198, 0.3776, 0.3437, 0.3476, 0.4016, 0.4291, 0.3957, 0.3862, 0.3293, 0.4092, 0.4173, 0.3926, 0.3767, 0.3377, 0.2967, 0.2642, 0.2289, 0.2346, 0.2341, 0.199, 0.2036, 0.1962, 0.2072, 0.1782, 0.1782, -0.0575, -0.0649, -0.0411, -0.0091, 0.0356, 0.0397, 0.0293, 0.013, 0.0068, 0.0, 0.0285, 0.013, 0.03, 0.001, 0.0072, -0.0423, -0.0195, 0.0131, 0.0131, 0.0131, 0.0131, 0.0414, 0.0338, 0.0378, 0.0086, 0.0306, 0.0324, 0.0391, 0.0, -0.0049, -0.0088, -0.0086, -0.0215, -0.0322, -0.0489, -0.0499, -0.068, -0.065, -0.0507, -0.0744, -0.0744, -0.08, -0.0767, -0.0585, -0.0446, -0.039, -0.0624, -0.0464, -0.0466, -0.0737, -0.0688, -0.0052, 0.0447, 0.0332, 0.0403, 0.0573, 0.065, 0.0753, 0.0761, 0.1047, 0.1377, 0.1065, 0.0819, 0.1157, 0.0878, 0.0728, 0.0882, 0.0661, 0.0472, 0.061, 0.058, 0.0644, 0.0828, 0.0822, 0.0579, 0.0583, 0.056, 0.0622, 0.1156, 0.1339, 0.1177, 0.1837, 0.1793, 0.161, 0.1425, 0.1482, 0.1573, 0.1603, 0.151, 0.1506, 0.1714, 0.1848, 0.1637, 0.1737, 0.1914, 0.1899, 0.1778, 0.2065, 0.2095, 0.2562, 0.2579, 0.2836, 0.2717, 0.2827, 0.2414, 0.3022, 0.2224, 0.1973, 0.1298, 0.217, 0.2607, 0.2101, 0.1973, 0.2679, 0.4432, 0.4475, 0.4984, 0.4919, 0.4823, 0.6638, 0.8296, 0.8765, 0.9618, 0.8713, 0.8336, 0.9293, 1.3151, 1.3151, 1.3151, 1.3151, 1.3151, 1.3151, 1.3151, 1.4213, 1.3408, 1.3785, 1.1847, 1.2042, 1.3619, 1.3134, 1.3328, 1.3655, 1.3527, 1.4732, 1.8869, 2.459, 2.3458, 2.5477, 2.6801, 2.398, 2.628, 2.43, 2.368, 2.4954, 2.6174]}
{"unit_id": "orig:1923162956398985547", "ticker_idx": 4, "ts": "2025-05-15T23:45:54+00:00", "symbol": "Winbond", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions", "tweet_type": "analysis", "timeline": "quarters", "summary": "Morgan Stanley W-shaped memory cycle: long Samsung/Hynix/Phison/Longsys/Winbond/GigaDevice; DRAM sector faces 2H correction with prices declining from 4Q25.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "「Asia Technology: Memory Outlook – This Time It’s W-shaped」 (Morgan Stanley)\n\n📌 Memory Market Outlook: “This Time It’s a W-shaped Recovery”\n\n⸻\n\n1. Industry Structure and Cycle Overview\n\n• Traditionally a supply-driven cycle, the memory industry is now shifting toward a demand-driven cycle.\n\n• Expectation of pre-purchasing (inventory build-up) in 1H followed by demand rebound in 2H.\n\n• Stock prices have rebounded, but the recovery is mostly based on lagging indicators. Real demand remains uncertain.\n\n⸻\n\n2. Price Trends and Outlook\n\n• 2Q25 DRAM: Currently up +5%, DDR4 surging +20% due to inventory build-up.\n\n• 3Q25: Prices expected to flatten; 4Q25 may see a downturn.\n\n• NAND: +6% in 2Q, +4% in 3Q, flat in 4Q. Production cuts help support pricing.\n\n⸻\n\n3. HBM Market Outlook (2025)\n\n• 2025 HBM TAM revised down from 180B Gb to 155B Gb.\n\n• Key reasons:\n\n• Export controls impacting products like NVIDIA H20\n\n• HBM3E inventory overhang and weakening customer demand\n\n• Product lineup adjustments by NVIDIA and AMD (shifting to GDDR7)\n\n• Samsung may convert HBM4 capacity to commodity DRAM if it fails to secure orders — raising oversupply concerns in DRAM.\n\n⸻\n\n4. DRAM Segment Analysis\n\n• Smartphone/PC DRAM prices: Expected to rise +3~8% QoQ in 2Q25.\n\n• Server DRAM: Up +3% in 2Q25, with strong demand from China.\n\n• DDR4/LPDDR4x: Samsung and SK Hynix plan to phase out production, causing a supply squeeze and sharp price increases (+20%).\n\n⸻\n\n5. NAND Market Analysis\n\n• 3Q25 wafer prices: Up +4% due to production cuts by major players.\n\n• Post-4Q25: Rising price pressure anticipated from increased competition by Chinese players.\n\n• eSSD: Solid demand from U.S. CSPs; China still focused on HDD.\n\n• Smartphones: UFS/eMMC up +5% in 2Q; channel inventory up +10%+.\n\n⸻\n\n6. Strategy and Stock-Specific Investment Highlights\n\n• DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H.\n\n• Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential.\n\n• SK Hynix: HBM leader, reflected in premium P/B of 1.9x.\n\n• NAND: Prefer self-sustaining growth stories like Phison, Longsys.\n\n• NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions.\n\n• HBM: Strategically preferred on the back of AI-driven demand.\n\n⸻\n\n7. Key Reasons Behind the W-Shaped Cycle\n\n• Inventory build-up in 1H may lead to a demand gap in 2H.\n\n• Current rebound is temporary, driven by price hikes without real demand.\n\n• Prices expected to decline again starting in 4Q25, with the bottom likely around 2Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0027624449852057253, "ret_signed_m1d": 0.0027624449852057253, "alpha_spy_m1d": 0.007622941987961984, "alpha_c_m1d": -0.0007174966685117301, "bench_spy_m1d": -0.004860497002756259, "bench_c_m1d": 0.0034799416537174555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00828722902552359, "ret_signed_p1d": -0.00828722902552359, "alpha_spy_p1d": -0.014621202716023962, "alpha_c_p1d": -0.005414261141187038, "bench_spy_p1d": 0.006333973690500372, "bench_c_p1d": -0.0028729678843365525, "ret_p1w": -0.013812118995935263, "ret_signed_p1w": -0.013812118995935263, "alpha_spy_p1w": -0.001330277376957345, "alpha_c_p1w": 0.01176148205494687, "bench_spy_p1w": -0.012481841618977918, "bench_c_p1w": -0.025573601050882133, "ret_p1m": 0.019337008966346936, "ret_signed_p1m": 0.019337008966346936, "alpha_spy_p1m": 0.008260893173432926, "alpha_c_p1m": -0.020560925297526733, "bench_spy_p1m": 0.01107611579291401, "bench_c_p1m": 0.03989793426387367, "ret_p3m": 0.019337008966346936, "ret_signed_p3m": 0.019337008966346936, "alpha_spy_p3m": -0.07233733020659483, "alpha_c_p3m": -0.1950035762790001, "bench_spy_p3m": 0.09167433917294177, "bench_c_p3m": 0.21434058524534705, "ret_p6m": 2.2099445477145157, "ret_signed_p6m": 2.2099445477145157, "alpha_spy_p6m": 2.0670741862699415, "alpha_c_p6m": 1.8012932019328936, "bench_spy_p6m": 0.14287036144457388, "bench_c_p6m": 0.4086513457816221, "price_path": [0.0249, 0.0359, 0.0608, 0.047, 0.0414, 0.0414, 0.0552, 0.0276, 0.0331, 0.0331, 0.0331, 0.0249, 0.0387, 0.0028, 0.0083, 0.0497, 0.0552, 0.1602, 0.1436, 0.1575, 0.1298, 0.1243, 0.1326, 0.1657, 0.163, 0.1022, 0.0691, 0.0746, 0.0635, 0.0331, -0.0304, 0.0, 0.0331, 0.0331, 0.0331, -0.0691, -0.1602, -0.2431, -0.1685, -0.1547, -0.1215, -0.0994, -0.1409, -0.1492, -0.1215, -0.1519, -0.1492, -0.1381, -0.1492, -0.1271, -0.1326, -0.1188, -0.1298, -0.1298, -0.1188, -0.1464, -0.1326, -0.1409, -0.1381, -0.1409, -0.0856, -0.011, 0.0028, 0.0, -0.0083, -0.0193, 0.0249, 0.0028, -0.0138, -0.0193, -0.0359, -0.0414, -0.0387, -0.0221, -0.0221, -0.0663, -0.0773, -0.0249, 0.0, 0.0138, 0.0166, 0.0387, 0.0387, 0.0387, 0.0193, 0.0166, 0.0359, 0.0856, 0.0829, 0.047, 0.0829, 0.1077, 0.1188, 0.1519, 0.1685, 0.1133, 0.0967, 0.0801, 0.0912, 0.058, 0.0552, 0.0442, 0.047, 0.0221, 0.0193, 0.0083, -0.011, -0.0249, -0.011, -0.0249, -0.0387, -0.0525, -0.0276, -0.0414, -0.0359, -0.0331, -0.0387, -0.0138, -0.0414, -0.0414, -0.0414, -0.0276, -0.058, -0.0884, -0.0359, -0.0387, 0.0193, 0.0, -0.0055, 0.0166, 0.0552, 0.0276, -0.0055, 0.0166, 0.0055, 0.0166, 0.0359, 0.0939, 0.0829, 0.0884, 0.1215, 0.1022, 0.1105, 0.1326, 0.2459, 0.3591, 0.3564, 0.3785, 0.3812, 0.4061, 0.4807, 0.5442, 0.6243, 0.7845, 0.8508, 0.8923, 1.0138, 0.9448, 0.8453, 0.7348, 0.7348, 0.8812, 0.8674, 0.989, 1.1436, 1.1436, 1.3564, 1.279, 1.4006, 1.4006, 1.3757, 1.2486, 1.2099, 1.4309, 1.4282, 1.5608, 1.4641, 1.511, 1.558, 1.558, 1.8122, 1.9834, 1.9834, 2.0497, 1.9945, 2.1271, 1.9448, 2.0939, 2.2597, 2.2099]}
{"unit_id": "orig:1923162956398985547", "ticker_idx": 5, "ts": "2025-05-15T23:45:54+00:00", "symbol": "GigaDevice", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions", "tweet_type": "analysis", "timeline": "quarters", "summary": "Morgan Stanley W-shaped memory cycle: long Samsung/Hynix/Phison/Longsys/Winbond/GigaDevice; DRAM sector faces 2H correction with prices declining from 4Q25.", "resolved_tickers": ["603986.SS"], "resolution_method": "exchange_ticker", "resolution_reason": "GigaDevice Semiconductor 603986.SS Shanghai NOR flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "「Asia Technology: Memory Outlook – This Time It’s W-shaped」 (Morgan Stanley)\n\n📌 Memory Market Outlook: “This Time It’s a W-shaped Recovery”\n\n⸻\n\n1. Industry Structure and Cycle Overview\n\n• Traditionally a supply-driven cycle, the memory industry is now shifting toward a demand-driven cycle.\n\n• Expectation of pre-purchasing (inventory build-up) in 1H followed by demand rebound in 2H.\n\n• Stock prices have rebounded, but the recovery is mostly based on lagging indicators. Real demand remains uncertain.\n\n⸻\n\n2. Price Trends and Outlook\n\n• 2Q25 DRAM: Currently up +5%, DDR4 surging +20% due to inventory build-up.\n\n• 3Q25: Prices expected to flatten; 4Q25 may see a downturn.\n\n• NAND: +6% in 2Q, +4% in 3Q, flat in 4Q. Production cuts help support pricing.\n\n⸻\n\n3. HBM Market Outlook (2025)\n\n• 2025 HBM TAM revised down from 180B Gb to 155B Gb.\n\n• Key reasons:\n\n• Export controls impacting products like NVIDIA H20\n\n• HBM3E inventory overhang and weakening customer demand\n\n• Product lineup adjustments by NVIDIA and AMD (shifting to GDDR7)\n\n• Samsung may convert HBM4 capacity to commodity DRAM if it fails to secure orders — raising oversupply concerns in DRAM.\n\n⸻\n\n4. DRAM Segment Analysis\n\n• Smartphone/PC DRAM prices: Expected to rise +3~8% QoQ in 2Q25.\n\n• Server DRAM: Up +3% in 2Q25, with strong demand from China.\n\n• DDR4/LPDDR4x: Samsung and SK Hynix plan to phase out production, causing a supply squeeze and sharp price increases (+20%).\n\n⸻\n\n5. NAND Market Analysis\n\n• 3Q25 wafer prices: Up +4% due to production cuts by major players.\n\n• Post-4Q25: Rising price pressure anticipated from increased competition by Chinese players.\n\n• eSSD: Solid demand from U.S. CSPs; China still focused on HDD.\n\n• Smartphones: UFS/eMMC up +5% in 2Q; channel inventory up +10%+.\n\n⸻\n\n6. Strategy and Stock-Specific Investment Highlights\n\n• DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H.\n\n• Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential.\n\n• SK Hynix: HBM leader, reflected in premium P/B of 1.9x.\n\n• NAND: Prefer self-sustaining growth stories like Phison, Longsys.\n\n• NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions.\n\n• HBM: Strategically preferred on the back of AI-driven demand.\n\n⸻\n\n7. Key Reasons Behind the W-Shaped Cycle\n\n• Inventory build-up in 1H may lead to a demand gap in 2H.\n\n• Current rebound is temporary, driven by price hikes without real demand.\n\n• Prices expected to decline again starting in 4Q25, with the bottom likely around 2Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020409793595915326, "ret_signed_m1d": 0.020409793595915326, "alpha_spy_m1d": 0.025270290598671585, "alpha_c_m1d": 0.01692985194219787, "bench_spy_m1d": -0.004860497002756259, "bench_c_m1d": 0.0034799416537174555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0023770540707734877, "ret_signed_p1d": 0.0023770540707734877, "alpha_spy_p1d": -0.003956919619726884, "alpha_c_p1d": 0.00525002195511004, "bench_spy_p1d": 0.006333973690500372, "bench_c_p1d": -0.0028729678843365525, "ret_p1w": -0.03557378953880164, "ret_signed_p1w": -0.03557378953880164, "alpha_spy_p1w": -0.02309194791982372, "alpha_c_p1w": -0.010000188487919504, "bench_spy_p1w": -0.012481841618977918, "bench_c_p1w": -0.025573601050882133, "ret_p1m": -0.025409840698627395, "ret_signed_p1m": -0.025409840698627395, "alpha_spy_p1m": -0.036485956491541405, "alpha_c_p1m": -0.06530777496250106, "bench_spy_p1m": 0.01107611579291401, "bench_c_p1m": 0.03989793426387367, "ret_p3m": -0.009758049068465335, "ret_signed_p3m": -0.009758049068465335, "alpha_spy_p3m": -0.1014323882414071, "alpha_c_p3m": -0.22409863431381238, "bench_spy_p3m": 0.09167433917294177, "bench_c_p3m": 0.21434058524534705, "ret_p6m": 0.8161695446423805, "ret_signed_p6m": 0.8161695446423805, "alpha_spy_p6m": 0.6732991831978066, "alpha_c_p6m": 0.40751819886075835, "bench_spy_p6m": 0.14287036144457388, "bench_c_p6m": 0.4086513457816221, "price_path": [0.045, 0.012, 0.0452, 0.0524, 0.1316, 0.1721, 0.123, 0.1211, 0.0962, 0.0806, 0.0621, 0.1025, 0.1138, 0.1289, 0.1039, 0.0862, 0.0507, 0.0531, 0.0187, 0.0647, 0.0494, 0.0648, 0.0734, 0.0525, 0.0132, 0.0256, -0.022, -0.0413, -0.0358, -0.0409, -0.042, -0.037, -0.0361, -0.0512, -0.0512, -0.1459, -0.1518, -0.1397, -0.1156, -0.0793, -0.0842, -0.0984, -0.105, -0.1049, -0.1123, -0.0782, -0.0791, -0.0607, -0.0889, -0.0768, -0.0557, 0.0387, 0.0548, 0.0548, 0.0548, 0.0548, 0.0807, 0.0684, 0.0452, 0.0197, 0.0285, 0.0288, 0.0204, 0.0, 0.0024, 0.0352, 0.034, -0.0316, -0.0356, -0.0541, -0.0489, -0.0746, -0.0697, -0.0676, -0.0795, -0.0795, -0.0748, -0.0526, -0.0336, 0.0118, 0.0052, -0.0213, -0.0234, -0.0168, -0.0254, -0.0081, 0.0026, 0.0051, 0.0099, -0.0249, -0.0147, 0.0157, 0.0287, 0.0102, 0.0356, 0.0371, 0.0166, -0.0084, -0.0159, -0.0231, -0.007, 0.0004, -0.018, -0.0296, -0.0271, -0.0334, -0.0547, -0.061, -0.0301, -0.037, -0.0374, -0.0476, -0.0384, -0.031, 0.0023, -0.0115, -0.0113, 0.0192, -0.0043, -0.0357, -0.0232, -0.0301, -0.0278, -0.0328, -0.0432, -0.0305, -0.0098, -0.0016, -0.0053, 0.0102, 0.0332, 0.0132, 0.044, 0.1274, 0.2402, 0.2971, 0.2691, 0.3039, 0.3517, 0.3068, 0.4375, 0.4467, 0.4063, 0.2657, 0.3116, 0.3467, 0.2917, 0.3258, 0.4584, 0.5635, 0.5695, 0.5549, 0.5338, 0.569, 0.5223, 0.6029, 0.5952, 0.6423, 0.6227, 0.5635, 0.6204, 0.7533, 0.7533, 0.7533, 0.7533, 0.7533, 0.7533, 0.7533, 0.8535, 0.7674, 0.7085, 0.6111, 0.6605, 0.7108, 0.6539, 0.7126, 0.8002, 0.7338, 0.7012, 0.8172, 0.999, 1.0299, 0.9822, 0.907, 0.8084, 0.8914, 0.8494, 0.7951, 0.8675, 0.8162]}
{"unit_id": "orig:1923162956398985547", "ticker_idx": 6, "ts": "2025-05-15T23:45:54+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H", "tweet_type": "analysis", "timeline": "quarters", "summary": "Morgan Stanley W-shaped memory cycle: long Samsung/Hynix/Phison/Longsys/Winbond/GigaDevice; DRAM sector faces 2H correction with prices declining from 4Q25.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "「Asia Technology: Memory Outlook – This Time It’s W-shaped」 (Morgan Stanley)\n\n📌 Memory Market Outlook: “This Time It’s a W-shaped Recovery”\n\n⸻\n\n1. Industry Structure and Cycle Overview\n\n• Traditionally a supply-driven cycle, the memory industry is now shifting toward a demand-driven cycle.\n\n• Expectation of pre-purchasing (inventory build-up) in 1H followed by demand rebound in 2H.\n\n• Stock prices have rebounded, but the recovery is mostly based on lagging indicators. Real demand remains uncertain.\n\n⸻\n\n2. Price Trends and Outlook\n\n• 2Q25 DRAM: Currently up +5%, DDR4 surging +20% due to inventory build-up.\n\n• 3Q25: Prices expected to flatten; 4Q25 may see a downturn.\n\n• NAND: +6% in 2Q, +4% in 3Q, flat in 4Q. Production cuts help support pricing.\n\n⸻\n\n3. HBM Market Outlook (2025)\n\n• 2025 HBM TAM revised down from 180B Gb to 155B Gb.\n\n• Key reasons:\n\n• Export controls impacting products like NVIDIA H20\n\n• HBM3E inventory overhang and weakening customer demand\n\n• Product lineup adjustments by NVIDIA and AMD (shifting to GDDR7)\n\n• Samsung may convert HBM4 capacity to commodity DRAM if it fails to secure orders — raising oversupply concerns in DRAM.\n\n⸻\n\n4. DRAM Segment Analysis\n\n• Smartphone/PC DRAM prices: Expected to rise +3~8% QoQ in 2Q25.\n\n• Server DRAM: Up +3% in 2Q25, with strong demand from China.\n\n• DDR4/LPDDR4x: Samsung and SK Hynix plan to phase out production, causing a supply squeeze and sharp price increases (+20%).\n\n⸻\n\n5. NAND Market Analysis\n\n• 3Q25 wafer prices: Up +4% due to production cuts by major players.\n\n• Post-4Q25: Rising price pressure anticipated from increased competition by Chinese players.\n\n• eSSD: Solid demand from U.S. CSPs; China still focused on HDD.\n\n• Smartphones: UFS/eMMC up +5% in 2Q; channel inventory up +10%+.\n\n⸻\n\n6. Strategy and Stock-Specific Investment Highlights\n\n• DRAM: Valuation attractive, but post-1H strength, a correction may follow in 2H.\n\n• Samsung: Trading near cycle-bottom P/B of 1.0x, with upside potential.\n\n• SK Hynix: HBM leader, reflected in premium P/B of 1.9x.\n\n• NAND: Prefer self-sustaining growth stories like Phison, Longsys.\n\n• NOR: Favor Winbond, GigaDevice; beneficiaries of tech migration and supply reductions.\n\n• HBM: Strategically preferred on the back of AI-driven demand.\n\n⸻\n\n7. Key Reasons Behind the W-Shaped Cycle\n\n• Inventory build-up in 1H may lead to a demand gap in 2H.\n\n• Current rebound is temporary, driven by price hikes without real demand.\n\n• Prices expected to decline again starting in 4Q25, with the bottom likely around 2Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009271585677823232, "ret_signed_m1d": -0.009271585677823232, "alpha_spy_m1d": -0.014132082680579491, "alpha_c_m1d": -0.005791644024105777, "bench_spy_m1d": -0.004860497002756259, "bench_c_m1d": 0.0034799416537174555, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012646563677404243, "ret_signed_p1d": -0.012646563677404243, "alpha_spy_p1d": -0.0063125899869038715, "alpha_c_p1d": -0.015519531561740796, "bench_spy_p1d": 0.006333973690500372, "bench_c_p1d": -0.0028729678843365525, "ret_p1w": -0.02327523373533021, "ret_signed_p1w": 0.02327523373533021, "alpha_spy_p1w": 0.010793392116352291, "alpha_c_p1w": -0.002298367315551924, "bench_spy_p1w": -0.012481841618977918, "bench_c_p1w": -0.025573601050882133, "ret_p1m": 0.13508077631056947, "ret_signed_p1m": -0.13508077631056947, "alpha_spy_p1m": -0.12400466051765546, "alpha_c_p1m": -0.0951828420466958, "bench_spy_p1m": 0.01107611579291401, "bench_c_p1m": 0.03989793426387367, "ret_p3m": 0.31072481216515446, "ret_signed_p3m": -0.31072481216515446, "alpha_spy_p3m": -0.2190504729922127, "alpha_c_p3m": -0.09638422691980741, "bench_spy_p3m": 0.09167433917294177, "bench_c_p3m": 0.21434058524534705, "ret_p6m": 1.3751341203881455, "ret_signed_p6m": -1.3751341203881455, "alpha_spy_p6m": -1.2322637589435717, "alpha_c_p6m": -0.9664827746065234, "bench_spy_p6m": 0.14287036144457388, "bench_c_p6m": 0.4086513457816221, "price_path": [0.0211, 0.0484, 0.0644, 0.0478, 0.0274, 0.0028, -0.0129, 0.0034, -0.023, -0.0421, -0.0528, -0.0574, -0.0377, -0.0548, -0.0454, -0.0734, -0.0672, -0.0183, -0.0208, 0.0078, 0.0352, 0.0254, 0.0359, 0.0565, 0.0454, 0.0396, 0.0201, 0.0322, 0.0195, -0.01, -0.0433, -0.0206, -0.0195, -0.0818, -0.1447, -0.1778, -0.1781, -0.1454, -0.1227, -0.1354, -0.1255, -0.1224, -0.1504, -0.1483, -0.1471, -0.151, -0.1459, -0.1204, -0.1092, -0.0908, -0.0985, -0.1063, -0.1133, -0.1105, -0.093, -0.094, -0.0937, -0.0766, -0.0686, -0.0653, -0.0184, -0.0005, 0.0093, 0.0, 0.0126, 0.0006, 0.0036, -0.0079, -0.0233, -0.0261, -0.0182, -0.0132, 0.0069, 0.0175, -0.0026, 0.0189, 0.0331, 0.0591, 0.0889, 0.0968, 0.1167, 0.1269, 0.1534, 0.1441, 0.1351, 0.1643, 0.173, 0.1838, 0.1795, 0.2058, 0.1959, 0.2625, 0.2773, 0.2782, 0.2646, 0.2668, 0.25, 0.2457, 0.2643, 0.2481, 0.2317, 0.264, 0.2488, 0.282, 0.2922, 0.2801, 0.2898, 0.2787, 0.2352, 0.2406, 0.2465, 0.2153, 0.2206, 0.2257, 0.2177, 0.2373, 0.2418, 0.2649, 0.2549, 0.1996, 0.2144, 0.2292, 0.2135, 0.24, 0.263, 0.2927, 0.3107, 0.3182, 0.3175, 0.3021, 0.2871, 0.2744, 0.2479, 0.226, 0.2473, 0.2576, 0.2542, 0.2579, 0.2817, 0.2726, 0.2387, 0.2531, 0.2614, 0.2873, 0.3222, 0.3318, 0.3716, 0.4213, 0.468, 0.5388, 0.5513, 0.6003, 0.5732, 0.6533, 0.6318, 0.6496, 0.6796, 0.6614, 0.6468, 0.5986, 0.65, 0.6577, 0.7428, 0.8296, 0.8443, 0.8552, 0.8368, 0.8748, 0.86, 0.904, 0.915, 0.8792, 0.9346, 1.0375, 1.0598, 1.1097, 1.0797, 1.077, 1.0884, 1.1976, 1.2619, 1.2301, 1.3142, 1.3429, 1.3471, 1.5082, 1.3567, 1.3879, 1.4059, 1.3751]}
{"unit_id": "orig:1933306111157023219", "ticker_idx": 0, "ts": "2025-06-12T23:31:10+00:00", "symbol": "Doosan Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "earnings momentum for the entire NVIDIA-centric supply chain could strengthen again in the second half of the year (Our covered stock: Doosan)", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz bullish on Doosan Electronics (NVIDIA CCL supply chain, H2 momentum) and ISU PETASYS (ASIC PCB, long-term uptrend) as AI hardware supply chain concerns dissipate.", "resolved_tickers": ["006280.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Electronics 006280.KS Korea CCL maker", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities Electronics/IT Components Yang Seung-soo]\n\n▶️ Review of Key Trends in AI Hardware\n\n<NVIDIA Supply Chain Concerns Dissipate>\n\n- In the first quarter, NVIDIA's ODMs such as Foxconn, Wistron, and Quanta experienced yield issues, leading to sluggish shipments of the GB200 NVL72 server racks. This continuously raised the possibility that the shipment forecast for the entire supply chain in the second half of the year could be significantly revised downwards.\n\n- This was because, although NVIDIA's business structure recognizes revenue at the time of shipping UBB (for HGX) or Bianca (for GB) boards under contracts with CSPs, regardless of the final GB200 NVL72 server rack shipment, the potential for Blackwell GPU production adjustments in the second half was highlighted due to the continuous inventory burden on downstream ODM companies (Refer to Figure 1).\n\n- However, concerns about production cuts are expected to be resolved from both top-down and bottom-up perspectives.\n\nTop-down: The expansion of inference and AI agents into the B2C domain signifies a surge in the number of tokens generated. This implies that AI hardware demand is structured to continuously expand, contrary to concerns about a reduction in AI capex. Furthermore, national-level infrastructure investments, such as for sovereign AI, are also beginning in earnest.\n\nBottom-up: The ramp-up speed of the GB200 NVL72 by Taiwanese ODM companies has been gradually improving through April and May, which is alleviating the inventory burden contrary to market concerns. On an annual basis, NVL72 shipments are also expected to be better than the market's feared range (10,000-15,000 units).\n\n- It is noteworthy that the earnings momentum for the entire NVIDIA-centric supply chain could strengthen again in the second half of the year (Our covered stock: Doosan).\n\n<ASIC Supply Chain Also Exceeds Expectations>\n- The entire Taiwanese ASIC supply chain is demonstrating performance that exceeds expectations (e.g., Wiwynn, and GCE, FHT, EMC which supply PCBs and CCLs for AWS Trainium).\n\n- The growth drivers for the ASIC market are 1) the full-scale expansion of the inference market and 2) the strengthening of CSPs' negotiation power against NVIDIA.\n\n- Consequently, in addition to the currently mass-produced Google TPU and AWS Trainium, custom-designed ASICs by U.S. CSPs are scheduled to enter mass production starting from the second half of 2025.\n\n- As a result, we maintain a positive view on the ASIC supply chain, as it is expected to continue a stable, long-term upward trajectory (Our covered stock: ISU PETASYS).\n\n<PCB, CCL Hardware Supply Chain Diversification – Expectation of NVIDIA Supply Chain Premium>\n\n- The trend of supply chain diversification for PCBs and CCLs within the AI supply chain is accelerating. This is because 1) the supply shortage of high-specification glass fiber with low dielectric constant (Low Dk) and low coefficient of thermal expansion (Low CTE) characteristics is intensifying due to the surge in AI demand, and 2) moves to reduce supply chain dependency on China are expanding due to the intensifying U.S.-China tech hegemony competition.\n\n- For CCLs, supply chain diversification is occurring within the relatively cost-sensitive ASIC supply chain. Representatively, Taiwan's TUC is presumed to have newly joined AWS's Trainium supply chain, and EMC is increasing its share within the Google TPU supply chain, which was virtually monopolized by Panasonic. Additionally, Korea's Doosan Electronics is also conducting sample tests for both clients.\n\n- However, in NVIDIA's case, it prioritizes delivery time and quality based on stable demand, and its attempts at supply chain diversification are understood to be limited at present. A premium for the NVIDIA supply chain is expected to be maintained for CCLs (Refer to Figures 13, 16).\n\n- In the PCB sector, the trend of diversifying production bases to countries like Thailand and Malaysia, away from China, continues. However, for high-difficulty PCBs, the skill level of the workforce is a key variable in securing yield and quality. In this aspect, the relative competitiveness of ISU PETASYS, which is building its 5th factory in Korea, is expected to be highlighted in the mid-to-long term.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008058646682434611, "ret_signed_m1d": -0.008058646682434611, "alpha_spy_m1d": -0.004100152536022139, "alpha_c_m1d": 0.002282623889617974, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.010341270572052585, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024175824429928272, "ret_signed_p1d": -0.024175824429928272, "alpha_spy_p1d": -0.012995760033460813, "alpha_c_p1d": -0.005413796731895015, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.018762027698033257, "ret_p1w": -0.06007328522956812, "ret_signed_p1w": -0.06007328522956812, "alpha_spy_p1w": -0.049622075385205355, "alpha_c_p1w": -0.06273250201733338, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": 0.0026592167877652617, "ret_p1m": -0.026373710736194766, "ret_signed_p1m": -0.026373710736194766, "alpha_spy_p1m": -0.06233830648317307, "alpha_c_p1m": -0.09831926749321562, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.07194555675702086, "ret_p3m": -0.02857148142508592, "ret_signed_p3m": -0.02857148142508592, "alpha_spy_p3m": -0.10890704082555591, "alpha_c_p3m": -0.13538187679040825, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.10681039536532233, "ret_p6m": -0.008058646682434611, "ret_signed_p6m": -0.008058646682434611, "alpha_spy_p6m": -0.15029264928107522, "alpha_c_p6m": -0.4011743233617775, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.3931156766793429, "price_path": [-0.041, -0.0081, -0.0022, -0.022, -0.033, -0.033, -0.0227, -0.0381, -0.0667, -0.085, -0.1055, -0.0535, -0.0762, -0.0769, -0.0828, -0.1385, -0.1355, -0.1736, -0.1414, -0.1502, -0.1363, -0.137, -0.1392, -0.1253, -0.1238, -0.1253, -0.1245, -0.1172, -0.1106, -0.1084, -0.1121, -0.0952, -0.0901, -0.0901, -0.0784, -0.0784, -0.0784, -0.1245, -0.1275, -0.1282, -0.1348, -0.1216, -0.1114, -0.0894, -0.0974, -0.085, -0.0886, -0.0835, -0.063, -0.0762, -0.0527, -0.0769, -0.0857, -0.0821, -0.0681, -0.0659, -0.0659, -0.0586, -0.0161, -0.0161, -0.011, -0.0059, -0.0081, 0.0, -0.0242, -0.0366, -0.0491, -0.0425, -0.0601, -0.0527, -0.0725, -0.0623, -0.0593, -0.0601, -0.0769, -0.0747, -0.0571, -0.0425, -0.0117, -0.0374, -0.0462, -0.0256, -0.022, -0.0132, -0.0264, -0.0374, -0.0227, -0.0425, 0.0462, 0.1245, 0.1297, 0.1436, 0.1216, 0.126, 0.1187, 0.0659, 0.0674, 0.0652, 0.0645, 0.0066, -0.0051, 0.0125, 0.0044, 0.0088, -0.0139, -0.0161, -0.0198, -0.0081, -0.0139, -0.0139, -0.0242, -0.0403, -0.0403, -0.0454, -0.0432, -0.0168, -0.0249, -0.0249, -0.0344, -0.0498, -0.0615, -0.0557, -0.0396, -0.0337, -0.0403, -0.0198, -0.0286, -0.0256, -0.0234, -0.0183, -0.0132, -0.0242, -0.0352, -0.0381, -0.0381, 0.022, 0.0344, 0.0103, 0.0029, -0.0352, -0.0418, -0.0535, -0.0527, -0.0308, -0.0308, -0.0308, -0.0308, -0.0308, -0.0308, -0.0462, -0.0513, -0.0571, -0.0432, -0.052, -0.0608, -0.0425, -0.0425, -0.0315, -0.0564, -0.0564, -0.0278, -0.0234, -0.0396, -0.0549, -0.0498, -0.0425, -0.0381, -0.0359, -0.0278, -0.0593, -0.0337, -0.0366, 0.0029, 0.0205, 0.0212, 0.0147, -0.0183, -0.03, -0.019, -0.0432, -0.044, -0.0498, -0.0161, -0.0161, -0.0066, 0.0051, 0.0073, 0.011, 0.0051, -0.0081]}
{"unit_id": "orig:1933306111157023219", "ticker_idx": 1, "ts": "2025-06-12T23:31:10+00:00", "symbol": "ISU PETASYS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain a positive view on the ASIC supply chain, as it is expected to continue a stable, long-term upward trajectory (Our covered stock: ISU PETASYS)", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz bullish on Doosan Electronics (NVIDIA CCL supply chain, H2 momentum) and ISU PETASYS (ASIC PCB, long-term uptrend) as AI hardware supply chain concerns dissipate.", "resolved_tickers": ["007660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "ISU Petasys 007660.KS Korea PCB ASIC supply chain", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities Electronics/IT Components Yang Seung-soo]\n\n▶️ Review of Key Trends in AI Hardware\n\n<NVIDIA Supply Chain Concerns Dissipate>\n\n- In the first quarter, NVIDIA's ODMs such as Foxconn, Wistron, and Quanta experienced yield issues, leading to sluggish shipments of the GB200 NVL72 server racks. This continuously raised the possibility that the shipment forecast for the entire supply chain in the second half of the year could be significantly revised downwards.\n\n- This was because, although NVIDIA's business structure recognizes revenue at the time of shipping UBB (for HGX) or Bianca (for GB) boards under contracts with CSPs, regardless of the final GB200 NVL72 server rack shipment, the potential for Blackwell GPU production adjustments in the second half was highlighted due to the continuous inventory burden on downstream ODM companies (Refer to Figure 1).\n\n- However, concerns about production cuts are expected to be resolved from both top-down and bottom-up perspectives.\n\nTop-down: The expansion of inference and AI agents into the B2C domain signifies a surge in the number of tokens generated. This implies that AI hardware demand is structured to continuously expand, contrary to concerns about a reduction in AI capex. Furthermore, national-level infrastructure investments, such as for sovereign AI, are also beginning in earnest.\n\nBottom-up: The ramp-up speed of the GB200 NVL72 by Taiwanese ODM companies has been gradually improving through April and May, which is alleviating the inventory burden contrary to market concerns. On an annual basis, NVL72 shipments are also expected to be better than the market's feared range (10,000-15,000 units).\n\n- It is noteworthy that the earnings momentum for the entire NVIDIA-centric supply chain could strengthen again in the second half of the year (Our covered stock: Doosan).\n\n<ASIC Supply Chain Also Exceeds Expectations>\n- The entire Taiwanese ASIC supply chain is demonstrating performance that exceeds expectations (e.g., Wiwynn, and GCE, FHT, EMC which supply PCBs and CCLs for AWS Trainium).\n\n- The growth drivers for the ASIC market are 1) the full-scale expansion of the inference market and 2) the strengthening of CSPs' negotiation power against NVIDIA.\n\n- Consequently, in addition to the currently mass-produced Google TPU and AWS Trainium, custom-designed ASICs by U.S. CSPs are scheduled to enter mass production starting from the second half of 2025.\n\n- As a result, we maintain a positive view on the ASIC supply chain, as it is expected to continue a stable, long-term upward trajectory (Our covered stock: ISU PETASYS).\n\n<PCB, CCL Hardware Supply Chain Diversification – Expectation of NVIDIA Supply Chain Premium>\n\n- The trend of supply chain diversification for PCBs and CCLs within the AI supply chain is accelerating. This is because 1) the supply shortage of high-specification glass fiber with low dielectric constant (Low Dk) and low coefficient of thermal expansion (Low CTE) characteristics is intensifying due to the surge in AI demand, and 2) moves to reduce supply chain dependency on China are expanding due to the intensifying U.S.-China tech hegemony competition.\n\n- For CCLs, supply chain diversification is occurring within the relatively cost-sensitive ASIC supply chain. Representatively, Taiwan's TUC is presumed to have newly joined AWS's Trainium supply chain, and EMC is increasing its share within the Google TPU supply chain, which was virtually monopolized by Panasonic. Additionally, Korea's Doosan Electronics is also conducting sample tests for both clients.\n\n- However, in NVIDIA's case, it prioritizes delivery time and quality based on stable demand, and its attempts at supply chain diversification are understood to be limited at present. A premium for the NVIDIA supply chain is expected to be maintained for CCLs (Refer to Figures 13, 16).\n\n- In the PCB sector, the trend of diversifying production bases to countries like Thailand and Malaysia, away from China, continues. However, for high-difficulty PCBs, the skill level of the workforce is a key variable in securing yield and quality. In this aspect, the relative competitiveness of ISU PETASYS, which is building its 5th factory in Korea, is expected to be highlighted in the mid-to-long term.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02988511145854922, "ret_signed_m1d": 0.02988511145854922, "alpha_spy_m1d": 0.03384360560496169, "alpha_c_m1d": 0.03907840786036776, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018390858584986125, "ret_signed_p1d": 0.018390858584986125, "alpha_spy_p1d": 0.029570922981453585, "alpha_c_p1d": 0.03240741545141046, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": 0.08390811796005804, "ret_signed_p1w": 0.08390811796005804, "alpha_spy_p1w": 0.0943593278044208, "alpha_c_p1w": 0.08786579802491645, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": 0.27816093753599147, "ret_signed_p1m": 0.27816093753599147, "alpha_spy_p1m": 0.24219634178901317, "alpha_c_p1m": 0.22166577898490547, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": 0.662068929525718, "ret_signed_p3m": 0.662068929525718, "alpha_spy_p3m": 0.5817333701252481, "alpha_c_p3m": 0.5663862347756636, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 2.108046192960634, "ret_signed_p6m": 2.108046192960634, "alpha_spy_p6m": 1.9658121903619934, "alpha_c_p6m": 1.895763707045611, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [-0.0599, -0.0576, -0.0794, -0.0817, -0.0885, -0.1309, -0.1492, -0.0633, -0.0931, -0.1253, -0.1816, -0.1667, -0.1885, -0.1839, -0.2494, -0.3414, -0.3046, -0.331, -0.2448, -0.2483, -0.2448, -0.2448, -0.2839, -0.2437, -0.2379, -0.2092, -0.2379, -0.2046, -0.2322, -0.2195, -0.2402, -0.2184, -0.2678, -0.2678, -0.1563, -0.1563, -0.1563, -0.1448, -0.1218, -0.0908, -0.0736, -0.0552, 0.0, -0.046, -0.0483, -0.0782, -0.0678, -0.0839, -0.0966, -0.0874, -0.0402, -0.0356, -0.0678, -0.054, -0.1172, -0.1241, -0.1241, -0.0793, -0.0655, -0.0655, -0.0529, -0.0552, 0.0299, 0.0, 0.0184, 0.0184, 0.054, 0.0414, 0.0839, 0.0678, 0.0483, 0.1299, 0.123, 0.1609, 0.1563, 0.2023, 0.1655, 0.1299, 0.2069, 0.2276, 0.1977, 0.3517, 0.3517, 0.3241, 0.2782, 0.292, 0.3586, 0.3839, 0.3885, 0.3862, 0.4069, 0.3609, 0.3609, 0.3839, 0.3908, 0.377, 0.4322, 0.4644, 0.4943, 0.4299, 0.5011, 0.5241, 0.492, 0.4782, 0.4851, 0.508, 0.4391, 0.446, 0.4138, 0.4138, 0.3724, 0.3471, 0.2713, 0.3034, 0.3586, 0.4529, 0.4644, 0.5011, 0.5011, 0.531, 0.4874, 0.4943, 0.4736, 0.4759, 0.4529, 0.5885, 0.6621, 0.7011, 0.7149, 0.6529, 0.7701, 0.7632, 0.777, 0.8149, 0.8149, 0.7632, 0.7241, 0.7379, 0.731, 0.6667, 0.6828, 0.6391, 0.6506, 0.7724, 0.7724, 0.7724, 0.7724, 0.7724, 0.7724, 0.7471, 0.6874, 0.708, 0.7885, 0.9011, 0.9517, 0.892, 0.8644, 0.9678, 0.9862, 1.1517, 1.2023, 1.3172, 1.4828, 1.4736, 1.6184, 1.6575, 1.6713, 1.4253, 1.5425, 1.3908, 1.692, 1.6759, 1.7494, 1.7448, 1.554, 2.023, 1.8437, 1.9333, 2.0644, 1.8069, 1.8759, 2.2345, 2.2552, 2.3632, 2.3103, 2.1103, 2.1632, 2.2115, 2.1379, 2.108]}
{"unit_id": "thread:1933729517937569954", "ticker_idx": 0, "ts": "2025-06-14T03:33:38+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD is demonstrating formidable ambition in the AI sector... putting significant pressure on NVIDIA... effectively putting NVIDIA in a hardware-catching-up role", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish AMD on MI500/N2P roadmap + RX 9060 XT value proposition; bearish NVDA as AMD's competitive push forces hardware catch-up.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD MI500 to be mass-produced using TSMC's N2P process, signaling direct competition with NVIDIA\n\nAMD is demonstrating formidable ambition in the AI sector. The MI500 series, slated for a 2027 launch, is already mapped out in its product roadmap, featuring the EPYC “Verano” CPUs and an AI rack-level solution, putting significant pressure on NVIDIA. In the consumer segment, AMD has recently launched its mainstream graphics card—the RX 9060 XT—which offers a better price-to-performance ratio compared to NVIDIA’s RTX 5060 Ti. Supply-chain sources estimate that the Verano CPU will be manufactured using TSMC’s second-generation 2 nm process (N2P), fully embracing TSMC by leveraging advanced process nodes and the latest packaging technologies to confront NVIDIA head-on with the strongest hardware.\n\nAt the “Advancing AI 2025” event, AMD Chair and CEO Dr. Lisa Su unveiled a clear product lineup. In the AI accelerator space, AMD is presenting a complete solution—from next year’s “Helios” AI cabinet system to the MI500 series GPU in the following year—continuing its strong push in the AI field, with a significant leap expected in 2026.\n\nIn contrast to NVIDIA’s upcoming “Vera” CPU, AMD’s Helios AI rack-level product, equipped with the Venice CPU, already employs TSMC’s 2 nm process technology for high-performance computing (HPC) and features up to 256 cores—a 33 percent increase over the previous generation. Supply-chain analysts believe that the next-generation Verano will continue to use the N2P process, and it is not ruled out that the MI500 GPU will further advance its process node, effectively putting NVIDIA in a hardware-catching-up role.\n\nConsumer-grade products are advancing in parallel. AMD’s mainstream market card, the RX 9060 XT, starts at just USD 299—approximately 8 percent cheaper than competing products—without compromising performance.\n\n$AMD $NVDA\n\n---\n\nhttps://t.co/zzJW4MPnxh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08093991436775749, "ret_signed_m1d": -0.08093991436775749, "alpha_spy_m1d": -0.07151538471586971, "alpha_c_m1d": -0.056678277030645874, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0056175258161350605, "ret_signed_p1d": 0.0056175258161350605, "alpha_spy_p1d": 0.014162558352713961, "alpha_c_p1d": 0.012413815536852946, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": 0.02523935799360011, "ret_signed_p1w": 0.02523935799360011, "alpha_spy_p1w": 0.026493353298236877, "alpha_c_p1w": 0.03078269330477268, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.2311891871335534, "ret_signed_p1m": 0.2311891871335534, "alpha_spy_p1m": 0.1958482764944962, "alpha_c_p1m": 0.1270811122608102, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.23166388892074763, "ret_signed_p3m": 0.23166388892074763, "alpha_spy_p3m": 0.1372618685972553, "alpha_c_p3m": 0.07770391953560774, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.7534614778654598, "ret_signed_p6m": 0.7534614778654598, "alpha_spy_p6m": 0.6136219487006451, "alpha_c_p6m": 0.3524440044980053, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.1595, -0.1523, -0.1578, -0.0992, -0.0916, -0.1282, -0.1562, -0.1833, -0.1871, -0.1868, -0.1854, -0.2579, -0.3215, -0.3382, -0.3812, -0.2338, -0.2982, -0.261, -0.2523, -0.2461, -0.3014, -0.3077, -0.3077, -0.323, -0.3175, -0.2848, -0.2526, -0.2353, -0.2374, -0.24, -0.2298, -0.2353, -0.2183, -0.2041, -0.2197, -0.2059, -0.1953, -0.1863, -0.1446, -0.1102, -0.0686, -0.0902, -0.0729, -0.0922, -0.1019, -0.1134, -0.1241, -0.1272, -0.1272, -0.0936, -0.107, -0.1057, -0.1239, -0.093, -0.0718, -0.0618, -0.0847, -0.0807, -0.0369, -0.0249, -0.0415, -0.0624, -0.0809, 0.0, 0.0056, 0.0032, 0.0032, 0.0146, 0.0252, 0.0953, 0.1346, 0.1368, 0.1378, 0.1227, 0.0769, 0.096, 0.0911, 0.0911, 0.0665, 0.0904, 0.0951, 0.1406, 0.1585, 0.1571, 0.2312, 0.2666, 0.2692, 0.2421, 0.2422, 0.2241, 0.2552, 0.2827, 0.3171, 0.374, 0.4039, 0.4203, 0.395, 0.3585, 0.3987, 0.3791, 0.2906, 0.364, 0.3669, 0.3631, 0.3842, 0.4591, 0.4317, 0.4045, 0.3936, 0.3177, 0.3071, 0.2953, 0.3273, 0.2925, 0.3183, 0.3223, 0.3338, 0.2867, 0.2867, 0.2843, 0.2828, 0.2801, 0.1958, 0.198, 0.2329, 0.2623, 0.2317, 0.2546, 0.2751, 0.2696, 0.2593, 0.2495, 0.2453, 0.2643, 0.273, 0.2729, 0.276, 0.2617, 0.2767, 0.2801, 0.2977, 0.3429, 0.3029, 0.6118, 0.6735, 0.8638, 0.8426, 0.7003, 0.7123, 0.7255, 0.8878, 0.8558, 0.8441, 0.9033, 0.8833, 0.8216, 0.8592, 1.0011, 1.0545, 1.0414, 1.0914, 1.0163, 1.0264, 1.0544, 0.9784, 1.0281, 0.8807, 0.8478, 0.9304, 0.8793, 1.0483, 0.9619, 0.9528, 0.903, 0.8221, 0.7687, 0.63, 0.6123, 0.7015, 0.6309, 0.6951, 0.6951, 0.7211, 0.7387, 0.703, 0.7217, 0.7088, 0.7246, 0.7494, 0.7535]}
{"unit_id": "thread:1933729517937569954", "ticker_idx": 1, "ts": "2025-06-14T03:33:38+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "effectively putting NVIDIA in a hardware-catching-up role", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish AMD on MI500/N2P roadmap + RX 9060 XT value proposition; bearish NVDA as AMD's competitive push forces hardware catch-up.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD MI500 to be mass-produced using TSMC's N2P process, signaling direct competition with NVIDIA\n\nAMD is demonstrating formidable ambition in the AI sector. The MI500 series, slated for a 2027 launch, is already mapped out in its product roadmap, featuring the EPYC “Verano” CPUs and an AI rack-level solution, putting significant pressure on NVIDIA. In the consumer segment, AMD has recently launched its mainstream graphics card—the RX 9060 XT—which offers a better price-to-performance ratio compared to NVIDIA’s RTX 5060 Ti. Supply-chain sources estimate that the Verano CPU will be manufactured using TSMC’s second-generation 2 nm process (N2P), fully embracing TSMC by leveraging advanced process nodes and the latest packaging technologies to confront NVIDIA head-on with the strongest hardware.\n\nAt the “Advancing AI 2025” event, AMD Chair and CEO Dr. Lisa Su unveiled a clear product lineup. In the AI accelerator space, AMD is presenting a complete solution—from next year’s “Helios” AI cabinet system to the MI500 series GPU in the following year—continuing its strong push in the AI field, with a significant leap expected in 2026.\n\nIn contrast to NVIDIA’s upcoming “Vera” CPU, AMD’s Helios AI rack-level product, equipped with the Venice CPU, already employs TSMC’s 2 nm process technology for high-performance computing (HPC) and features up to 256 cores—a 33 percent increase over the previous generation. Supply-chain analysts believe that the next-generation Verano will continue to use the N2P process, and it is not ruled out that the MI500 GPU will further advance its process node, effectively putting NVIDIA in a hardware-catching-up role.\n\nConsumer-grade products are advancing in parallel. AMD’s mainstream market card, the RX 9060 XT, starts at just USD 299—approximately 8 percent cheaper than competing products—without compromising performance.\n\n$AMD $NVDA\n\n---\n\nhttps://t.co/zzJW4MPnxh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018798885498186424, "ret_signed_m1d": 0.018798885498186424, "alpha_spy_m1d": 0.009374355846298643, "alpha_c_m1d": -0.005462751838925195, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.02426163733711162, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003939529332248171, "ret_signed_p1d": 0.003939529332248171, "alpha_spy_p1d": -0.00460550320433073, "alpha_c_p1d": -0.0028567603884697146, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.006796289720717885, "ret_p1w": -0.003593884416002502, "ret_signed_p1w": 0.003593884416002502, "alpha_spy_p1w": 0.002339889111365734, "alpha_c_p1w": -0.0019494508951700684, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.00554333531117257, "ret_p1m": 0.1797636240210243, "ret_signed_p1m": -0.1797636240210243, "alpha_spy_p1m": -0.1444227133819671, "alpha_c_p1m": -0.0756555491482811, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.1041080748727432, "ret_p3m": 0.22454905837338268, "ret_signed_p3m": -0.22454905837338268, "alpha_spy_p3m": -0.13014703804989036, "alpha_c_p3m": -0.07058908898824279, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.1539599693851399, "ret_p6m": 0.2785315797312218, "ret_signed_p6m": -0.2785315797312218, "alpha_spy_p6m": -0.1386920505664071, "alpha_c_p6m": 0.12248589363623275, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.4010174733674545, "price_path": [-0.1878, -0.1809, -0.1866, -0.161, -0.1659, -0.2138, -0.2299, -0.2421, -0.251, -0.2388, -0.2369, -0.2965, -0.3482, -0.3252, -0.3345, -0.2099, -0.2566, -0.2334, -0.2349, -0.2246, -0.2779, -0.2986, -0.2986, -0.3303, -0.3166, -0.2902, -0.2645, -0.2328, -0.2486, -0.2466, -0.2473, -0.2287, -0.2087, -0.2134, -0.2153, -0.191, -0.1889, -0.1938, -0.15, -0.1021, -0.0647, -0.0682, -0.0643, -0.0631, -0.0713, -0.0892, -0.082, -0.0927, -0.0927, -0.0636, -0.0683, -0.0381, -0.0661, -0.0506, -0.0241, -0.0192, -0.0326, -0.0206, -0.0143, -0.0051, -0.0129, 0.0021, -0.0188, 0.0, -0.0039, 0.0055, 0.0055, -0.0058, -0.0036, 0.0222, 0.0665, 0.0714, 0.0903, 0.0919, 0.0595, 0.0868, 0.1013, 0.1013, 0.0936, 0.1058, 0.1257, 0.1341, 0.1398, 0.1339, 0.1798, 0.1844, 0.1957, 0.1916, 0.1845, 0.1544, 0.1803, 0.2008, 0.1991, 0.2216, 0.213, 0.239, 0.2293, 0.2006, 0.244, 0.232, 0.24, 0.2494, 0.2627, 0.2583, 0.2659, 0.255, 0.258, 0.2471, 0.2579, 0.2139, 0.2122, 0.2093, 0.2301, 0.2427, 0.2563, 0.2551, 0.2452, 0.2038, 0.2038, 0.1803, 0.1792, 0.1864, 0.1543, 0.1632, 0.1802, 0.2256, 0.2245, 0.229, 0.2286, 0.2087, 0.177, 0.2181, 0.2211, 0.2691, 0.2333, 0.2232, 0.2281, 0.2316, 0.2569, 0.2896, 0.2942, 0.3056, 0.2968, 0.2824, 0.2789, 0.3071, 0.331, 0.266, 0.3016, 0.2443, 0.2429, 0.2566, 0.2664, 0.2624, 0.2521, 0.246, 0.259, 0.2874, 0.3235, 0.3895, 0.431, 0.4023, 0.3996, 0.4299, 0.3733, 0.3492, 0.3, 0.3004, 0.3758, 0.3351, 0.3395, 0.2915, 0.3144, 0.2897, 0.2535, 0.2892, 0.2485, 0.2364, 0.2617, 0.229, 0.2459, 0.2459, 0.2234, 0.2436, 0.2542, 0.2413, 0.2675, 0.2608, 0.2825, 0.2785]}
{"unit_id": "orig:1930780115576984004", "ticker_idx": 0, "ts": "2025-06-06T00:13:46+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MICRON TO BENEFIT", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Samsung HBM3E 12Hi cert delayed to Q4 2025; Micron seen as beneficiary.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "* UBS: SAMSUNG'S HBM3E 12HI NVIDIA CERTIFICATION DELAYED TO Q4 2025, MICRON TO BENEFIT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02091003034336192, "ret_signed_m1d": -0.02091003034336192, "alpha_spy_m1d": -0.01074543055490118, "alpha_c_m1d": -0.015251804482363851, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022015545117126667, "ret_signed_p1d": 0.022015545117126667, "alpha_spy_p1d": 0.021114262228982383, "alpha_c_p1d": 0.005278398433741227, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.06484897673487722, "ret_signed_p1w": 0.06484897673487722, "alpha_spy_p1w": 0.06842072362348317, "alpha_c_p1w": 0.04799312516540488, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.10568243610849493, "ret_signed_p1m": 0.10568243610849493, "alpha_spy_p1m": 0.06666833469433509, "alpha_c_p1m": -0.0012696972363375192, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.0946182007900278, "ret_signed_p3m": 0.0946182007900278, "alpha_spy_p3m": 0.01700184518977621, "alpha_c_p3m": -0.03872566255496923, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 1.218469810598613, "ret_signed_p6m": 1.218469810598613, "alpha_spy_p6m": 1.0765452156323814, "alpha_c_p6m": 0.8218805411327075, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.1808, -0.1202, -0.1271, -0.0728, -0.0514, -0.0642, -0.0611, -0.0524, -0.1286, -0.1082, -0.1336, -0.1524, -0.1614, -0.1864, -0.1996, -0.1828, -0.1839, -0.3152, -0.4038, -0.3702, -0.3963, -0.2827, -0.3547, -0.3593, -0.3458, -0.3456, -0.3614, -0.3662, -0.3662, -0.3852, -0.3533, -0.3282, -0.2868, -0.2651, -0.2763, -0.2918, -0.2912, -0.2836, -0.2564, -0.2592, -0.2584, -0.2389, -0.2156, -0.2091, -0.1499, -0.1071, -0.122, -0.1208, -0.0973, -0.0913, -0.0964, -0.1172, -0.1265, -0.1399, -0.1399, -0.1122, -0.114, -0.1083, -0.1299, -0.0956, -0.0581, -0.0489, -0.0209, 0.0, 0.022, 0.0514, 0.0688, 0.0702, 0.0648, 0.1039, 0.1085, 0.1221, 0.1221, 0.1385, 0.1245, 0.1782, 0.1722, 0.1606, 0.1492, 0.1353, 0.1136, 0.1214, 0.1265, 0.1265, 0.1057, 0.1472, 0.1271, 0.1351, 0.1482, 0.0936, 0.1074, 0.0735, 0.0443, 0.0547, 0.044, 0.007, 0.0127, 0.0302, 0.0258, 0.0257, 0.0323, 0.0579, 0.0063, -0.033, -0.0063, 0.0056, 0.003, 0.0315, 0.0962, 0.1407, 0.1779, 0.1458, 0.1552, 0.1144, 0.1392, 0.1253, 0.0807, 0.0676, 0.085, 0.0734, 0.0741, 0.0857, 0.1249, 0.0973, 0.0973, 0.0924, 0.0946, 0.1452, 0.2113, 0.2121, 0.2469, 0.2908, 0.3883, 0.4497, 0.4547, 0.4643, 0.4751, 0.5572, 0.5004, 0.5178, 0.5343, 0.491, 0.446, 0.4501, 0.5112, 0.5427, 0.6795, 0.6942, 0.7329, 0.7618, 0.7132, 0.8133, 0.7744, 0.6754, 0.7785, 0.7258, 0.7708, 0.8685, 0.8671, 0.9076, 0.8663, 0.8311, 0.9071, 1.0207, 1.0306, 1.0473, 1.0909, 1.0667, 1.0645, 1.1653, 1.0115, 1.1912, 1.1988, 1.195, 1.3369, 1.2245, 1.2594, 1.1861, 1.2772, 1.2322, 1.1081, 1.0843, 0.8578, 0.9132, 1.066, 1.0715, 1.1244, 1.1244, 1.1818, 1.2185]}
{"unit_id": "orig:1930780115576984004", "ticker_idx": 1, "ts": "2025-06-06T00:13:46+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SAMSUNG'S HBM3E 12HI NVIDIA CERTIFICATION DELAYED TO Q4 2025", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Samsung HBM3E 12Hi cert delayed to Q4 2025; Micron seen as beneficiary.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "* UBS: SAMSUNG'S HBM3E 12HI NVIDIA CERTIFICATION DELAYED TO Q4 2025, MICRON TO BENEFIT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.010164599788460738, "alpha_c_m1d": -0.009355607467128713, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.009355607467128713, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011844390025516294, "ret_signed_p1d": -0.011844390025516294, "alpha_spy_p1d": -0.01094310713737201, "alpha_c_p1d": -0.006935874945164233, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.0049085150803520605, "ret_p1w": -0.013536349655783209, "ret_signed_p1w": 0.013536349655783209, "alpha_spy_p1w": 0.009964602767177255, "alpha_c_p1w": 0.01693460192800189, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.0033982522722186825, "ret_p1m": 0.050396855504293514, "ret_signed_p1m": -0.050396855504293514, "alpha_spy_p1m": -0.011382754090133673, "alpha_c_p1m": 0.020850426627838647, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.07124728213213216, "ret_p3m": 0.1882933477972173, "ret_signed_p3m": -0.1882933477972173, "alpha_spy_p3m": -0.1106769921969657, "alpha_c_p3m": -0.09002544373399357, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.09826790406322372, "ret_p6m": 0.7237018825648858, "ret_signed_p6m": -0.7237018825648858, "alpha_spy_p6m": -0.5817772875986542, "alpha_c_p6m": -0.5188302066174015, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.20487167594748423, "price_path": [-0.0984, -0.0766, -0.0799, -0.0799, -0.0311, -0.0311, -0.016, 0.0126, 0.0378, 0.0176, 0.0059, 0.0328, 0.0395, 0.0186, -0.022, -0.0051, -0.0051, -0.0254, -0.0508, -0.0998, -0.0948, -0.1032, -0.0457, -0.066, -0.0491, -0.0423, -0.0745, -0.0677, -0.0643, -0.0626, -0.0694, -0.0575, -0.0575, -0.0575, -0.0558, -0.0558, -0.0609, -0.0609, -0.0812, -0.0812, -0.0812, -0.0761, -0.0761, -0.0728, -0.0254, -0.0372, -0.0288, -0.0305, -0.0389, -0.0558, -0.0541, -0.0575, -0.0745, -0.0829, -0.0745, -0.088, -0.0541, -0.0508, -0.0491, -0.0389, -0.0389, -0.022, 0.0, 0.0, 0.0118, 0.0017, 0.0135, 0.0068, -0.0135, -0.0321, -0.0169, 0.0118, 0.0017, 0.0068, -0.0186, 0.0237, 0.0372, 0.0186, 0.0351, 0.0181, 0.0249, 0.0351, 0.0861, 0.0776, 0.0504, 0.0453, 0.0283, 0.0385, 0.0657, 0.064, 0.0844, 0.1015, 0.1355, 0.1423, 0.1542, 0.1236, 0.1304, 0.1236, 0.1219, 0.1985, 0.2019, 0.236, 0.2155, 0.173, 0.1866, 0.19, 0.1713, 0.2002, 0.2223, 0.2087, 0.2104, 0.224, 0.2189, 0.2189, 0.1917, 0.1917, 0.2002, 0.2019, 0.2155, 0.2172, 0.1968, 0.2019, 0.1849, 0.1866, 0.1508, 0.1764, 0.1883, 0.1934, 0.1832, 0.1934, 0.2172, 0.236, 0.2496, 0.2836, 0.3024, 0.3517, 0.3313, 0.367, 0.367, 0.4215, 0.442, 0.4539, 0.4658, 0.4181, 0.4398, 0.4347, 0.4706, 0.5347, 0.5347, 0.5347, 0.5347, 0.5347, 0.5347, 0.6143, 0.5955, 0.5664, 0.6245, 0.6707, 0.6741, 0.6775, 0.6673, 0.6861, 0.6502, 0.6895, 0.7442, 0.7015, 0.7186, 0.7801, 0.8383, 0.8998, 0.7938, 0.7203, 0.6963, 0.6741, 0.7203, 0.7699, 0.763, 0.7579, 0.6621, 0.7203, 0.6724, 0.6502, 0.7203, 0.6211, 0.6536, 0.6981, 0.7579, 0.7699, 0.7186, 0.7237]}
{"unit_id": "quote:1936776745115668548", "ticker_idx": 0, "ts": "2025-06-22T13:22:14+00:00", "symbol": "CL=F", "kind": "commodity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I'm preparing to take a 2x short position on crude oil :)", "tweet_type": "prediction", "timeline": "days", "summary": "Author plans 2x leveraged short on crude oil despite Hormuz closure news and JPM $120–130 projection.", "resolved_tickers": ["CL=F"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "JPMorgan has projected that if the Strait of Hormuz is closed, oil prices could rise to $120–130 per barrel.\n\nI'm preparing to take a 2x short position on crude oil :)", "reply_text": null, "parent_text": null, "quoted_text": "&gt; Iran's parliament has approved the closure of the Strait of Hormuz, with the final decision resting with the Supreme National Security Council, according to Press TV.\n\nWTF?", "parent_author": null, "ret_m1d": 0.09370891795010161, "ret_signed_m1d": -0.09370891795010161, "alpha_spy_m1d": -0.10348978810037102, "alpha_c_m1d": -0.10348978810037102, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.009780870150269405, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.060429123639729965, "ret_signed_p1d": 0.060429123639729965, "alpha_spy_p1d": 0.07147633263281772, "alpha_c_p1d": 0.07147633263281772, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.011047208993087754, "ret_p1w": -0.04962781228816904, "ret_signed_p1w": 0.04962781228816904, "alpha_spy_p1w": 0.07912047163563574, "alpha_c_p1w": 0.07912047163563574, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.0294926593474667, "ret_p1m": -0.033571784849521835, "ret_signed_p1m": 0.033571784849521835, "alpha_spy_p1m": 0.08140979357861378, "alpha_c_p1m": 0.08140979357861378, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.04783800872909194, "ret_p3m": -0.07210629524697976, "ret_signed_p3m": 0.07210629524697976, "alpha_spy_p3m": 0.1755970602655872, "alpha_c_p3m": 0.1755970602655872, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.10349076501860743, "ret_p6m": -0.19325647738470908, "ret_signed_p6m": 0.19325647738470908, "alpha_spy_p6m": 0.3275597782040295, "alpha_c_p6m": 0.3275597782040295, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.1343033008193204, "price_path": [0.0166, 0.0206, 0.0124, 0.0434, 0.0393, 0.0467, -0.0228, -0.0952, -0.114, -0.1303, -0.0899, -0.1232, -0.1023, -0.1019, -0.1048, -0.0882, -0.0559, -0.0559, -0.0793, -0.0613, -0.0911, -0.0835, -0.0801, -0.0943, -0.1181, -0.1503, -0.1353, -0.1492, -0.1661, -0.1375, -0.1524, -0.1255, -0.1093, -0.0958, -0.0706, -0.0782, -0.1006, -0.0879, -0.085, -0.0868, -0.1013, -0.1067, -0.1019, -0.1019, -0.1112, -0.0974, -0.1105, -0.1127, -0.0874, -0.0744, -0.0826, -0.075, -0.0574, -0.047, -0.0515, -0.0053, -0.0069, 0.0652, 0.0476, 0.0924, 0.0968, 0.0968, 0.0937, 0.0, -0.0604, -0.0524, -0.0477, -0.0436, -0.0496, -0.0447, -0.0155, -0.022, -0.0293, -0.0085, -0.0026, -0.0019, -0.0283, -0.0009, -0.0223, -0.029, -0.0311, -0.0142, -0.0171, -0.0191, -0.0336, -0.0476, -0.0362, -0.0489, -0.0263, 0.0102, 0.0217, 0.0109, -0.0172, -0.0324, -0.0489, -0.0607, -0.0676, -0.0676, -0.0664, -0.0779, -0.0855, -0.0664, -0.0833, -0.0743, -0.0899, -0.0774, -0.0728, -0.0708, -0.0542, -0.0768, -0.0636, -0.0571, -0.0657, -0.0657, -0.0426, -0.0663, -0.0734, -0.0969, -0.0912, -0.0858, -0.0706, -0.0896, -0.085, -0.076, -0.0582, -0.0651, -0.0721, -0.0851, -0.0857, -0.0744, -0.0514, -0.0515, -0.0407, -0.0739, -0.0896, -0.0982, -0.1172, -0.1114, -0.0995, -0.099, -0.087, -0.1022, -0.1403, -0.1317, -0.1432, -0.1495, -0.1613, -0.1601, -0.1604, -0.156, -0.1461, -0.0981, -0.1023, -0.1051, -0.122, -0.1172, -0.1159, -0.1099, -0.1089, -0.116, -0.1301, -0.1325, -0.1279, -0.1223, -0.109, -0.1463, -0.1433, -0.1229, -0.1255, -0.1134, -0.1324, -0.1368, -0.1525, -0.1411, -0.1541, -0.1439, -0.1439, -0.1454, -0.1341, -0.1441, -0.1395, -0.129, -0.123, -0.1406, -0.1498, -0.1467, -0.1592, -0.1616, -0.1706, -0.1933]}
{"unit_id": "orig:1930243929485820222", "ticker_idx": 0, "ts": "2025-06-04T12:43:09+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC (Buy-rated, on Conviction List)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman reiterates Buy/CL on TSMC and MediaTek; AI sentiment improving as ODM yield gains reduce risk of further NVDA GB200/GB300 order cuts.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Generative AI / Broader AI Theme Back in Focus, Stocks on the Rise\n\nOur analyst Bruce, currently on a trip related to TSMC (and Asian semiconductor equipment names), reports signs of improving investor sentiment, particularly in the AI space. In contrast, investors remain cautious about non-AI demand amid continued macroeconomic and geopolitical uncertainty.\n\nPlease also refer to Bruce’s recent investor marketing sessions conducted in both the U.S. and Asia (TSMC analyst). According to him, while investor sentiment is showing signs of improvement in the AI segment, there remains a cautious stance toward non-AI demand due to persistent macro and geopolitical risks. Key discussions focused on TSMC (Buy-rated, on Conviction List), MediaTek (Buy-rated, on Conviction List), and UMC, a second-tier foundry.\n\nOn AI, Bruce noted:\n\n“We believe investor sentiment around AI demand has become increasingly positive. With improved assembly yields at downstream ODMs for AI server racks, the mismatch between upstream and downstream supply chains is easing. This reduces the likelihood of further and larger-scale AI order cuts in the near term. Previously, during our investor meetings in Hong Kong in March 2025 (see note), investors had broadly expressed concerns about potential downside risks to AI demand — particularly the risk of further order reductions for NVIDIA’s GB200/GB300 NVL72 AI server racks. However, as ODMs are now achieving better assembly yields, we believe the probability of significant further AI order cuts has diminished.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.040404059997174224, "ret_signed_m1d": -0.040404059997174224, "alpha_spy_m1d": -0.040672679537220535, "alpha_c_m1d": -0.028646353207514297, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008080749427813894, "ret_signed_p1d": 0.008080749427813894, "alpha_spy_p1d": 0.012913518484003128, "alpha_c_p1d": 0.009868254040974134, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.07575758903034391, "ret_signed_p1w": 0.07575758903034391, "alpha_spy_p1w": 0.06664565786821885, "alpha_c_p1w": 0.03551927940171318, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.10568196622058523, "ret_signed_p1m": 0.10568196622058523, "alpha_spy_p1m": 0.05322822565464236, "alpha_c_p1m": -0.02091190725935177, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.18176104823984152, "ret_signed_p3m": 0.18176104823984152, "alpha_spy_p3m": 0.09613522893988491, "alpha_c_p3m": 0.02867258682531748, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.46146820804763644, "ret_signed_p6m": 0.46146820804763644, "alpha_spy_p6m": 0.31438828788140527, "alpha_c_p6m": 0.08029646837397153, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.0104, 0.0034, -0.0237, -0.0067, -0.0298, -0.0358, -0.0247, -0.0192, -0.0384, -0.0051, -0.0182, -0.0182, 0.0, -0.0101, -0.0323, -0.0384, -0.0808, -0.0465, -0.0485, -0.0485, -0.0485, -0.1434, -0.1758, -0.2071, -0.1283, -0.102, -0.1263, -0.1141, -0.1364, -0.1444, -0.1414, -0.1566, -0.1758, -0.1182, -0.1273, -0.103, -0.0939, -0.0889, -0.0828, -0.0828, -0.0404, -0.0525, -0.0707, -0.0626, -0.0727, -0.0414, -0.0333, -0.0212, 0.0091, 0.003, 0.0081, -0.0061, -0.0121, 0.0, -0.0081, -0.0061, -0.0141, -0.0253, -0.0232, -0.0232, -0.0232, -0.0444, -0.0404, 0.0, 0.0081, 0.0051, 0.0152, 0.0556, 0.0758, 0.06, 0.0448, 0.0397, 0.06, 0.0702, 0.0499, 0.0702, 0.0347, 0.0651, 0.0854, 0.0905, 0.0955, 0.0753, 0.1006, 0.1006, 0.1057, 0.1006, 0.0955, 0.0955, 0.1057, 0.1158, 0.1158, 0.1108, 0.126, 0.1463, 0.1463, 0.1716, 0.1665, 0.1463, 0.1615, 0.1615, 0.1615, 0.1615, 0.1513, 0.1716, 0.1767, 0.1767, 0.1513, 0.1665, 0.1412, 0.197, 0.1919, 0.197, 0.197, 0.2173, 0.1919, 0.197, 0.197, 0.202, 0.1513, 0.1665, 0.1513, 0.1868, 0.1919, 0.2071, 0.1767, 0.1767, 0.1818, 0.1767, 0.1767, 0.1767, 0.197, 0.197, 0.2173, 0.2426, 0.2578, 0.2781, 0.2731, 0.3036, 0.2883, 0.3087, 0.2883, 0.3189, 0.3647, 0.3647, 0.3443, 0.324, 0.324, 0.3291, 0.3494, 0.3902, 0.4258, 0.4258, 0.4615, 0.4411, 0.4666, 0.4666, 0.4411, 0.4513, 0.492, 0.5124, 0.4767, 0.5073, 0.5073, 0.4869, 0.4767, 0.4767, 0.5073, 0.5022, 0.5328, 0.5328, 0.5277, 0.5379, 0.5328, 0.4869, 0.492, 0.4869, 0.5022, 0.492, 0.5022, 0.4869, 0.4564, 0.4717, 0.4309, 0.4207, 0.4818, 0.4105, 0.4004, 0.4411, 0.4666, 0.4615]}
{"unit_id": "orig:1930243929485820222", "ticker_idx": 1, "ts": "2025-06-04T12:43:09+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek (Buy-rated, on Conviction List)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman reiterates Buy/CL on TSMC and MediaTek; AI sentiment improving as ODM yield gains reduce risk of further NVDA GB200/GB300 order cuts.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Generative AI / Broader AI Theme Back in Focus, Stocks on the Rise\n\nOur analyst Bruce, currently on a trip related to TSMC (and Asian semiconductor equipment names), reports signs of improving investor sentiment, particularly in the AI space. In contrast, investors remain cautious about non-AI demand amid continued macroeconomic and geopolitical uncertainty.\n\nPlease also refer to Bruce’s recent investor marketing sessions conducted in both the U.S. and Asia (TSMC analyst). According to him, while investor sentiment is showing signs of improvement in the AI segment, there remains a cautious stance toward non-AI demand due to persistent macro and geopolitical risks. Key discussions focused on TSMC (Buy-rated, on Conviction List), MediaTek (Buy-rated, on Conviction List), and UMC, a second-tier foundry.\n\nOn AI, Bruce noted:\n\n“We believe investor sentiment around AI demand has become increasingly positive. With improved assembly yields at downstream ODMs for AI server racks, the mismatch between upstream and downstream supply chains is easing. This reduces the likelihood of further and larger-scale AI order cuts in the near term. Previously, during our investor meetings in Hong Kong in March 2025 (see note), investors had broadly expressed concerns about potential downside risks to AI demand — particularly the risk of further order reductions for NVIDIA’s GB200/GB300 NVL72 AI server racks. However, as ODMs are now achieving better assembly yields, we believe the probability of significant further AI order cuts has diminished.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.00026861954004631095, "alpha_c_m1d": 0.011757706789659927, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02352947850070075, "ret_signed_p1d": -0.02352947850070075, "alpha_spy_p1d": -0.018696709444511517, "alpha_c_p1d": -0.02174197388754051, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.03137243896375663, "ret_signed_p1w": 0.03137243896375663, "alpha_spy_p1w": 0.02226050780163158, "alpha_c_p1w": -0.008865870664874098, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.015686418519055945, "ret_signed_p1m": 0.015686418519055945, "alpha_spy_p1m": -0.03676732204688693, "alpha_c_p1m": -0.11090745496088106, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.08766413352050795, "ret_signed_p3m": 0.08766413352050795, "alpha_spy_p3m": 0.0020383142205513405, "alpha_c_p3m": -0.06542432789401609, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.07166910785764946, "ret_signed_p6m": 0.07166910785764946, "alpha_spy_p6m": -0.07541081230858171, "alpha_c_p6m": -0.30950263181601545, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.149, 0.0941, 0.0863, 0.1059, 0.0902, 0.098, 0.1176, 0.1569, 0.149, 0.1647, 0.1608, 0.1608, 0.1961, 0.1882, 0.1647, 0.149, 0.0902, 0.1451, 0.1255, 0.1255, 0.1255, 0.0157, -0.0118, -0.0706, 0.0196, 0.0863, 0.098, 0.0863, 0.0706, 0.051, 0.0706, 0.0549, 0.0196, 0.0745, 0.0431, 0.0824, 0.0588, 0.0745, 0.0588, 0.0588, 0.0196, 0.0157, 0.0039, 0.0, 0.0118, 0.0392, 0.0314, 0.0627, 0.0667, 0.0745, 0.0706, 0.0275, 0.0314, 0.051, 0.0392, 0.0314, 0.0039, 0.0039, 0.0039, -0.0118, -0.0118, -0.0118, 0.0, 0.0, -0.0235, 0.0039, 0.0078, 0.0471, 0.0314, 0.0078, -0.0196, -0.0157, 0.0, 0.0078, -0.0078, -0.0196, -0.0157, -0.0039, 0.0157, 0.0118, 0.0078, -0.0196, 0.0, 0.0157, 0.0157, 0.0317, 0.0237, 0.0157, 0.0837, 0.1197, 0.1356, 0.1077, 0.1157, 0.1277, 0.1117, 0.1277, 0.1396, 0.1436, 0.1436, 0.1436, 0.1436, 0.1197, 0.0877, 0.1037, 0.0957, 0.0957, 0.0637, 0.0717, 0.0557, 0.0837, 0.0797, 0.0837, 0.0877, 0.0957, 0.1117, 0.0957, 0.1277, 0.1117, 0.0837, 0.0917, 0.0917, 0.1277, 0.1197, 0.1197, 0.1077, 0.0957, 0.0877, 0.0877, 0.1157, 0.1077, 0.1476, 0.1436, 0.2076, 0.1916, 0.1836, 0.1876, 0.1876, 0.2276, 0.2076, 0.2076, 0.1516, 0.1237, 0.0997, 0.0637, 0.0757, 0.0477, 0.0477, 0.0517, 0.0277, 0.0237, 0.0557, 0.0557, 0.0597, 0.0677, 0.0757, 0.0757, 0.0517, 0.0437, 0.0517, 0.0637, 0.0637, 0.0717, 0.0717, 0.0637, 0.0357, 0.0357, 0.0677, 0.0517, 0.0397, 0.0477, 0.0477, 0.0397, 0.0517, 0.0637, 0.0317, 0.0077, -0.0003, -0.0083, -0.0043, -0.0043, -0.0163, -0.0163, -0.0643, -0.0723, -0.0523, -0.0843, -0.0803, -0.0523, 0.0397, 0.0717]}
{"unit_id": "orig:1930243929485820222", "ticker_idx": 2, "ts": "2025-06-04T12:43:09+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the probability of significant further AI order cuts has diminished", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman reiterates Buy/CL on TSMC and MediaTek; AI sentiment improving as ODM yield gains reduce risk of further NVDA GB200/GB300 order cuts.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Generative AI / Broader AI Theme Back in Focus, Stocks on the Rise\n\nOur analyst Bruce, currently on a trip related to TSMC (and Asian semiconductor equipment names), reports signs of improving investor sentiment, particularly in the AI space. In contrast, investors remain cautious about non-AI demand amid continued macroeconomic and geopolitical uncertainty.\n\nPlease also refer to Bruce’s recent investor marketing sessions conducted in both the U.S. and Asia (TSMC analyst). According to him, while investor sentiment is showing signs of improvement in the AI segment, there remains a cautious stance toward non-AI demand due to persistent macro and geopolitical risks. Key discussions focused on TSMC (Buy-rated, on Conviction List), MediaTek (Buy-rated, on Conviction List), and UMC, a second-tier foundry.\n\nOn AI, Bruce noted:\n\n“We believe investor sentiment around AI demand has become increasingly positive. With improved assembly yields at downstream ODMs for AI server racks, the mismatch between upstream and downstream supply chains is easing. This reduces the likelihood of further and larger-scale AI order cuts in the near term. Previously, during our investor meetings in Hong Kong in March 2025 (see note), investors had broadly expressed concerns about potential downside risks to AI demand — particularly the risk of further order reductions for NVIDIA’s GB200/GB300 NVL72 AI server racks. However, as ODMs are now achieving better assembly yields, we believe the probability of significant further AI order cuts has diminished.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004932421576486434, "ret_signed_m1d": -0.004932421576486434, "alpha_spy_m1d": -0.005201041116532745, "alpha_c_m1d": 0.006825285213173493, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013599137425771679, "ret_signed_p1d": -0.013599137425771679, "alpha_spy_p1d": -0.008766368369582445, "alpha_c_p1d": -0.011811632812611439, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.006481889145962372, "ret_signed_p1w": 0.006481889145962372, "alpha_spy_p1w": -0.0026300420161626814, "alpha_c_p1w": -0.03375642048266836, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.12282315830271795, "ret_signed_p1m": 0.12282315830271795, "alpha_spy_p1m": 0.07036941773677508, "alpha_c_p1m": -0.003770715177219053, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.22739630300135127, "ret_signed_p3m": 0.22739630300135127, "alpha_spy_p3m": 0.14177048370139467, "alpha_c_p3m": 0.07430784158682724, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.2703119733794417, "ret_signed_p6m": 0.2703119733794417, "alpha_spy_p6m": 0.12323205321321051, "alpha_c_p6m": -0.11085976629422323, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.206, -0.2463, -0.2337, -0.1845, -0.1856, -0.1427, -0.1578, -0.1867, -0.1719, -0.1648, -0.1707, -0.1445, -0.1496, -0.1984, -0.2148, -0.2272, -0.2363, -0.2239, -0.222, -0.2827, -0.3355, -0.312, -0.3214, -0.1944, -0.242, -0.2184, -0.2199, -0.2094, -0.2637, -0.2849, -0.2849, -0.3172, -0.3032, -0.2763, -0.2501, -0.2178, -0.2339, -0.2318, -0.2325, -0.2136, -0.1932, -0.198, -0.2, -0.1752, -0.173, -0.1781, -0.1333, -0.0845, -0.0464, -0.05, -0.0459, -0.0447, -0.0531, -0.0713, -0.0641, -0.0749, -0.0749, -0.0452, -0.0501, -0.0192, -0.0478, -0.032, -0.0049, 0.0, -0.0136, -0.0014, 0.005, 0.0144, 0.0065, 0.0218, 0.0004, 0.0196, 0.0156, 0.0252, 0.0252, 0.0137, 0.0159, 0.0422, 0.0874, 0.0924, 0.1116, 0.1133, 0.0803, 0.1081, 0.1228, 0.1228, 0.1151, 0.1275, 0.1478, 0.1564, 0.1621, 0.1562, 0.2029, 0.2076, 0.2191, 0.2149, 0.2077, 0.177, 0.2034, 0.2243, 0.2226, 0.2455, 0.2368, 0.2633, 0.2534, 0.2242, 0.2684, 0.2561, 0.2643, 0.2738, 0.2874, 0.2829, 0.2907, 0.2796, 0.2826, 0.2716, 0.2826, 0.2377, 0.236, 0.233, 0.2542, 0.2671, 0.2809, 0.2797, 0.2696, 0.2274, 0.2274, 0.2034, 0.2023, 0.2096, 0.1769, 0.186, 0.2033, 0.2496, 0.2485, 0.2531, 0.2526, 0.2324, 0.2001, 0.242, 0.245, 0.2939, 0.2574, 0.2471, 0.2522, 0.2557, 0.2815, 0.3148, 0.3195, 0.3311, 0.3222, 0.3075, 0.304, 0.3327, 0.3571, 0.2907, 0.3271, 0.2687, 0.2673, 0.2812, 0.2912, 0.2871, 0.2767, 0.2705, 0.2837, 0.3126, 0.3495, 0.4167, 0.459, 0.4298, 0.427, 0.4579, 0.4002, 0.3757, 0.3254, 0.3259, 0.4027, 0.3612, 0.3657, 0.3168, 0.3401, 0.315, 0.2781, 0.3144, 0.273, 0.2606, 0.2864, 0.2531, 0.2703, 0.2703]}
{"unit_id": "thread:1921391686539780149", "ticker_idx": 0, "ts": "2025-05-11T02:27:30+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell will likely have a hard time avoiding weakness", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Marvell faces weakness; MediaTek flagged as a positive watch.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Marvell will likely have a hard time avoiding weakness. However, MediaTek is worth keeping an eye on. https://t.co/Qceg6TIlKe\n\n---\n\nWhat will happen to non-TSMC OSAT companies? I think their growth is slow. https://t.co/ezfJXtMGw8\n\n---\n\nThis table reveals two key points:\n\n1. The B300A was thought to be canceled, but it still exists. However, its shipment volume is limited.\n\n2. Chinese ASIC companies have no access to nodes below 7nm. https://t.co/StdjJn6jpT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07519371156457744, "ret_signed_m1d": 0.07519371156457744, "alpha_spy_m1d": 0.04320340039344117, "alpha_c_m1d": 0.016181829462234676, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015193770867061529, "ret_signed_p1d": -0.015193770867061529, "alpha_spy_p1d": -0.008589830549891486, "alpha_c_p1d": 0.019134988895834093, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": -0.03007744904434051, "ret_signed_p1w": 0.03007744904434051, "alpha_spy_p1w": 0.05042074931808571, "alpha_c_p1w": 0.06630168197246156, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.06728685204623219, "ret_signed_p1m": -0.06728685204623219, "alpha_spy_p1m": -0.032826681597986296, "alpha_c_p1m": 0.036584110329979325, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.17693157086350353, "ret_signed_p3m": -0.17693157086350353, "alpha_spy_p3m": -0.0892301760453158, "alpha_c_p3m": 0.04930146827821069, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.35999584882610725, "ret_signed_p6m": -0.35999584882610725, "alpha_spy_p6m": -0.19511516376313298, "alpha_c_p6m": 0.12587252835680562, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.6326, 0.6028, 0.6494, 0.6494, 0.6615, 0.7196, 0.6835, 0.6076, 0.5163, 0.4402, 0.4696, 0.3574, 0.422, 0.3292, 0.3679, 0.3959, 0.1193, 0.097, 0.017, 0.0401, 0.0817, 0.0648, 0.0645, 0.0902, 0.0574, 0.0811, 0.0916, 0.0901, 0.1268, 0.1099, 0.0339, 0.0044, -0.0392, -0.0465, -0.0295, -0.0208, -0.1383, -0.2345, -0.2102, -0.2252, -0.056, -0.1812, -0.1722, -0.1898, -0.1733, -0.1949, -0.1984, -0.1984, -0.2344, -0.2152, -0.1664, -0.1112, -0.0865, -0.0899, -0.0902, -0.095, -0.055, -0.0336, -0.0391, -0.0509, -0.127, -0.1065, -0.0752, 0.0, 0.0152, 0.0217, 0.0109, -0.0115, -0.0301, -0.0478, -0.0681, -0.0411, -0.0591, -0.0591, -0.0105, 0.0014, -0.0119, -0.0668, -0.047, -0.0332, 0.0279, 0.0102, 0.0597, 0.0719, 0.0673, 0.058, 0.0797, 0.0417, 0.0918, 0.0851, 0.162, 0.162, 0.1397, 0.0974, 0.166, 0.1772, 0.2398, 0.1963, 0.2, 0.182, 0.1512, 0.1656, 0.1656, 0.1093, 0.1155, 0.1203, 0.1374, 0.1282, 0.1251, 0.1236, 0.0993, 0.1173, 0.1583, 0.1336, 0.117, 0.1369, 0.1488, 0.1515, 0.1779, 0.1845, 0.2683, 0.2471, 0.1552, 0.1875, 0.189, 0.1687, 0.1769, 0.2001, 0.1991, 0.2073, 0.2308, 0.2264, 0.1822, 0.1907, 0.1183, 0.1051, 0.1049, 0.1327, 0.1319, 0.1523, 0.1605, 0.1983, -0.0245, -0.0245, 0.0024, -0.0332, -0.0054, -0.0173, 0.0241, 0.0371, 0.0412, 0.0332, 0.045, 0.0463, 0.0685, 0.1014, 0.1518, 0.1523, 0.172, 0.1578, 0.2427, 0.3004, 0.2905, 0.2784, 0.3045, 0.3017, 0.3375, 0.3378, 0.3797, 0.3495, 0.4353, 0.407, 0.3293, 0.3879, 0.3387, 0.3802, 0.3699, 0.3656, 0.3328, 0.3083, 0.2585, 0.2852, 0.3063, 0.3774, 0.3735, 0.3997, 0.3752, 0.4555, 0.4032, 0.36]}
{"unit_id": "thread:1921391686539780149", "ticker_idx": 1, "ts": "2025-05-11T02:27:30+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek is worth keeping an eye on", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Marvell faces weakness; MediaTek flagged as a positive watch.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Marvell will likely have a hard time avoiding weakness. However, MediaTek is worth keeping an eye on. https://t.co/Qceg6TIlKe\n\n---\n\nWhat will happen to non-TSMC OSAT companies? I think their growth is slow. https://t.co/ezfJXtMGw8\n\n---\n\nThis table reveals two key points:\n\n1. The B300A was thought to be canceled, but it still exists. However, its shipment volume is limited.\n\n2. Chinese ASIC companies have no access to nodes below 7nm. https://t.co/StdjJn6jpT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007604681609199027, "ret_signed_m1d": 0.007604681609199027, "alpha_spy_m1d": 0.03959499278033529, "alpha_c_m1d": 0.06661656371154179, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.030418340471125127, "ret_signed_p1d": 0.030418340471125127, "alpha_spy_p1d": 0.023814400153955084, "alpha_c_p1d": -0.003910419291770495, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": -0.0038022443131817685, "ret_signed_p1w": -0.0038022443131817685, "alpha_spy_p1w": -0.02414554458692697, "alpha_c_p1w": -0.040026477241302816, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.015209170235562564, "ret_signed_p1m": 0.015209170235562564, "alpha_spy_p1m": -0.019251000212683334, "alpha_c_p1m": -0.08866179214064895, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.05070228383606645, "ret_signed_p3m": 0.05070228383606645, "alpha_spy_p3m": -0.036999110982121275, "alpha_c_p3m": -0.17553075530564777, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.019685214118315475, "ret_signed_p6m": 0.019685214118315475, "alpha_spy_p6m": -0.1451954709446588, "alpha_c_p6m": -0.4661831630645974, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.1293, 0.1559, 0.1445, 0.1749, 0.1673, 0.1711, 0.1407, 0.1483, 0.1749, 0.1787, 0.1863, 0.1521, 0.1521, 0.1179, 0.1445, 0.1445, 0.1369, 0.1141, 0.0608, 0.0532, 0.0722, 0.057, 0.0646, 0.0837, 0.1217, 0.1141, 0.1293, 0.1255, 0.1255, 0.1597, 0.1521, 0.1293, 0.1141, 0.057, 0.1103, 0.0913, 0.0913, 0.0913, -0.0152, -0.0418, -0.0989, -0.0114, 0.0532, 0.0646, 0.0532, 0.038, 0.019, 0.038, 0.0228, -0.0114, 0.0418, 0.0114, 0.0494, 0.0266, 0.0418, 0.0266, 0.0266, -0.0114, -0.0152, -0.0266, -0.0304, -0.019, 0.0076, 0.0, 0.0304, 0.0342, 0.0418, 0.038, -0.0038, 0.0, 0.019, 0.0076, 0.0, -0.0266, -0.0266, -0.0266, -0.0418, -0.0418, -0.0418, -0.0304, -0.0304, -0.0532, -0.0266, -0.0228, 0.0152, 0.0, -0.0228, -0.0494, -0.0456, -0.0304, -0.0228, -0.038, -0.0494, -0.0456, -0.0342, -0.0152, -0.019, -0.0228, -0.0494, -0.0304, -0.0152, -0.0152, 0.0003, -0.0075, -0.0152, 0.0507, 0.0856, 0.1011, 0.074, 0.0817, 0.0934, 0.0778, 0.0934, 0.105, 0.1089, 0.1089, 0.1089, 0.1089, 0.0856, 0.0546, 0.0701, 0.0623, 0.0623, 0.0313, 0.0391, 0.0236, 0.0507, 0.0468, 0.0507, 0.0546, 0.0623, 0.0778, 0.0623, 0.0934, 0.0778, 0.0507, 0.0585, 0.0585, 0.0934, 0.0856, 0.0856, 0.074, 0.0623, 0.0546, 0.0546, 0.0817, 0.074, 0.1127, 0.1089, 0.1709, 0.1554, 0.1476, 0.1515, 0.1515, 0.1903, 0.1709, 0.1709, 0.1166, 0.0895, 0.0662, 0.0313, 0.0429, 0.0158, 0.0158, 0.0197, -0.0036, -0.0075, 0.0236, 0.0236, 0.0274, 0.0352, 0.0429, 0.0429, 0.0197, 0.0119, 0.0197, 0.0313, 0.0313, 0.0391, 0.0391, 0.0313, 0.0042, 0.0042, 0.0352, 0.0197, 0.0081, 0.0158, 0.0158, 0.0081, 0.0197]}
{"unit_id": "quote:1932270898666721666", "ticker_idx": 0, "ts": "2025-06-10T02:57:36+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "shipments of NVIDIA's NVL72 racks are increasing noticeably", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "NVL72 rack shipments rising sharply MoM (1-1.15k→2-2.5k) with June expected higher; bullish demand signal for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "A major U.S. investment bank (Morgan Stanley) observed that shipments of NVIDIA's NVL72 racks are increasing noticeably.\n\n- Shipments in May are estimated at around 2,000–2,500 units, compared to approximately 1,000–1,150 units in April.\n- Analysts expect the shipment volume to continue rising in June.\n\nBreakdown by supplier:\n- Hon Hai (Foxconn): ~1,000 units\n- Wistron: Equivalent to ~900–1,000 racks based on compute trays\n- Quanta: ~400 units", "reply_text": null, "parent_text": null, "quoted_text": "早安！6/10 外電綜合整理\n\n- AI供應鏈追蹤報導：\n美系大行觀察NVL72 rack出貨量有感增加，訪查5月應有2-2.5k(~四月1-1.15k), 券商估計6月出貨量續增。\n其中鴻海1k, 緯創compute tray相當於900-1k rack, 廣達400。\n\n- 2376技嘉: 月營收靚\n美系大行重申Top Pick,", "parent_author": null, "ret_m1d": -0.009238640676650811, "ret_signed_m1d": -0.009238640676650811, "alpha_spy_m1d": -0.003600939880776255, "alpha_c_m1d": 0.010260073940271397, "bench_spy_m1d": -0.005637700795874556, "bench_c_m1d": -0.019498714616922208, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007780571947151982, "ret_signed_p1d": -0.007780571947151982, "alpha_spy_p1d": -0.00492865735636383, "alpha_c_p1d": -0.00705552898423889, "bench_spy_p1d": -0.0028519145907881516, "bench_c_p1d": -0.0007250429629130917, "ret_p1w": 0.0011808310557084312, "ret_signed_p1w": 0.0011808310557084312, "alpha_spy_p1w": 0.010383425090068177, "alpha_c_p1w": 0.0030124201742001144, "bench_spy_p1w": -0.009202594034359746, "bench_c_p1w": -0.0018315891184916833, "ret_p1m": 0.13150395341983012, "ret_signed_p1m": 0.13150395341983012, "alpha_spy_p1m": 0.09365671285704691, "alpha_c_p1m": 0.04240555316340755, "bench_spy_p1m": 0.03784724056278321, "bench_c_p1m": 0.08909840025642257, "ret_p3m": 0.1602640490058671, "ret_signed_p3m": 0.1602640490058671, "alpha_spy_p3m": 0.08386713292258374, "alpha_c_p3m": 0.0411737273657351, "bench_spy_p3m": 0.07639691608328336, "bench_c_p3m": 0.119090321640132, "ret_p6m": 0.24765625496960464, "ret_signed_p6m": 0.24765625496960464, "alpha_spy_p6m": 0.10715503456308095, "alpha_c_p6m": -0.14269743797195544, "bench_spy_p6m": 0.14050122040652369, "bench_c_p6m": 0.3903536929415601, "price_path": [-0.1971, -0.1548, -0.1697, -0.1982, -0.1837, -0.1766, -0.1824, -0.1566, -0.1616, -0.2098, -0.226, -0.2382, -0.2472, -0.2349, -0.233, -0.2929, -0.3449, -0.3218, -0.3311, -0.2058, -0.2528, -0.2294, -0.231, -0.2206, -0.2742, -0.295, -0.295, -0.3268, -0.3131, -0.2865, -0.2607, -0.2289, -0.2447, -0.2427, -0.2434, -0.2247, -0.2046, -0.2094, -0.2113, -0.1869, -0.1847, -0.1897, -0.1456, -0.0975, -0.0599, -0.0634, -0.0595, -0.0583, -0.0665, -0.0845, -0.0773, -0.088, -0.088, -0.0588, -0.0636, -0.0331, -0.0613, -0.0457, -0.019, -0.0142, -0.0276, -0.0156, -0.0092, 0.0, -0.0078, 0.0073, -0.0138, 0.0051, 0.0012, 0.0106, 0.0106, -0.0007, 0.0015, 0.0274, 0.072, 0.0769, 0.0959, 0.0975, 0.065, 0.0924, 0.1069, 0.1069, 0.0993, 0.1115, 0.1315, 0.14, 0.1457, 0.1398, 0.1858, 0.1905, 0.2018, 0.1977, 0.1906, 0.1603, 0.1864, 0.2069, 0.2053, 0.2279, 0.2192, 0.2454, 0.2356, 0.2068, 0.2504, 0.2383, 0.2464, 0.2558, 0.2692, 0.2647, 0.2724, 0.2615, 0.2645, 0.2536, 0.2644, 0.2201, 0.2185, 0.2156, 0.2365, 0.2491, 0.2627, 0.2615, 0.2516, 0.21, 0.21, 0.1864, 0.1853, 0.1925, 0.1603, 0.1692, 0.1862, 0.2319, 0.2308, 0.2354, 0.2349, 0.2149, 0.183, 0.2244, 0.2274, 0.2756, 0.2396, 0.2295, 0.2345, 0.2379, 0.2634, 0.2962, 0.3008, 0.3123, 0.3034, 0.289, 0.2855, 0.3138, 0.3378, 0.2725, 0.3083, 0.2507, 0.2493, 0.2631, 0.2729, 0.2688, 0.2586, 0.2524, 0.2655, 0.294, 0.3303, 0.3966, 0.4384, 0.4095, 0.4067, 0.4372, 0.3803, 0.3562, 0.3066, 0.3071, 0.3828, 0.3419, 0.3464, 0.2982, 0.3212, 0.2964, 0.26, 0.2958, 0.255, 0.2427, 0.2682, 0.2354, 0.2523, 0.2523, 0.2297, 0.2499, 0.2606, 0.2477]}
{"unit_id": "orig:1923523891491897650", "ticker_idx": 0, "ts": "2025-05-16T23:40:07+00:00", "symbol": "Hon Hai Precision", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB200 ramping up, long-term visibility in AI servers, diversified growth from smartphones and EV expansion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan reiterates bullish stance on Hon Hai: GB200/AI server revenue doubling QoQ, two-year datacenter visibility, smartphone/EV diversification offsetting 2H tariff/FX caution.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan (Foxconn)", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Hon Hai: GB200 ramping up, long-term visibility in AI servers, diversified growth from smartphones and EV expansion\nJPMorgan (G. Hariharan, 25/05/16)\n\n(1) GB200 server rack shipments accelerating, AI server revenue doubling\n\nHon Hai’s AI server revenue share rose from 40–42% in 2024 to 50% in Q1, driven primarily by the large-scale ramp-up of GB200 production starting in late March. Management expects AI server revenue to double quarter-over-quarter in Q2 and maintains its full-year growth guidance. The 2025 target is over 30,000 server rack shipments, including around 10,000 units of GB200/GB300 NVL72. U.S. capacity has already begun shipments, and global capacity is sufficient to meet demand for the next 1–2 years.\n\n(2) Limited integration bottlenecks, manageable margin impact\nThe main bottlenecks for AI servers are hardware-software integration and final testing. While the initial ramp-up is expected to take two quarters, margins were not significantly eroded due to pricing and cost hedging in contracts. The transition to GB300, which shares a similar architecture with GB200, will be smoother, and future modular GPU add-in card designs are expected to improve yields and maintainability.\n\n(3) Inventory remains high, reflecting long-cycle demand\nQ1 inventory increased over 20% YoY due to pre-build in March, mainly resulting from the long 6–8 week lead time for GB200 servers (compared to just 2 weeks for generic servers). Management expects high inventory levels to continue for several quarters to accommodate long-cycle demand driven by liquid cooling, networking, and integration complexity.\n\n(4) Dual growth drivers from AI and traditional servers, two-year visibility\nHon Hai’s datacenter business has two-year visibility. Management remains confident in achieving double-digit growth for both AI and general-purpose servers in 2025. Beyond AI servers, the company will also benefit from synchronized growth in smart consumer electronics and computing products.\n\n(5) Smartphone and EV geographic expansion diversifies risk and boosts momentum\nSmartphone production capacity in India and Brazil is ramping up quickly and is expected to account for 15–20% of group capacity, helping Hon Hai gain greater share in high-end smartphone assembly. The recent EV design project with Mitsubishi signals entry into traditional OEMs. If more design wins are secured, the 5% EV market share target becomes attainable.\n\n(6) Tariff and FX headwinds make 2H outlook more cautious\nThe company plans to increase 2025 capex by over 20% YoY, with 60–70% allocated to expansion in India, the U.S., and Mexico. Due to front-loaded shipments and demand uncertainty in 1H, management adopts a more conservative outlook for 2H. Every NT$1 appreciation in the TWD leads to a ~3% decline in local revenue and a 0.1ppt drop in gross margin, cited as a key reason for the downward revision of Q1 guidance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00949367912067145, "ret_signed_m1d": 0.00949367912067145, "alpha_spy_m1d": 0.01578778610313958, "alpha_c_m1d": 0.01136689741711161, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": -0.00187321829644016, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0253164442944549, "ret_signed_p1d": -0.0253164442944549, "alpha_spy_p1d": -0.026410362394492903, "alpha_c_p1d": -0.02386906433702829, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.0014473799574266089, "ret_p1w": -0.0253164442944549, "ret_signed_p1w": -0.0253164442944549, "alpha_spy_p1w": 7.899582602255428e-05, "alpha_c_p1w": 0.009082578119746665, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03439902241420156, "ret_p1m": -0.01898735824134301, "ret_signed_p1m": -0.01898735824134301, "alpha_spy_p1m": -0.03325868296899326, "alpha_c_p1m": -0.05368442829325237, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.03469707005190936, "ret_p3m": 0.30195642309308246, "ret_signed_p3m": 0.30195642309308246, "alpha_spy_p3m": 0.21343986185460384, "alpha_c_p3m": 0.15778394693656672, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.14417247615651574, "ret_p6m": 0.6364641186988618, "ret_signed_p6m": 0.6364641186988618, "alpha_spy_p6m": 0.4830655076808239, "alpha_c_p6m": 0.3745958839784418, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.26186823472042, "price_path": [0.1614, 0.1582, 0.1519, 0.1582, 0.1646, 0.1234, 0.1424, 0.1013, 0.1013, 0.0759, 0.0791, 0.0981, 0.0981, 0.0886, 0.0886, 0.0665, 0.0791, 0.0633, 0.0759, 0.0633, 0.0601, 0.038, 0.0506, 0.0443, 0.0316, 0.0443, 0.0506, 0.0158, -0.0253, -0.0759, -0.038, -0.0285, -0.0285, -0.0285, -0.1234, -0.2089, -0.288, -0.2184, -0.1487, -0.1234, -0.1171, -0.1392, -0.1487, -0.1424, -0.1361, -0.1646, -0.1203, -0.1361, -0.1203, -0.0981, -0.0949, -0.1044, -0.1044, -0.0665, -0.0981, -0.0759, -0.0823, -0.0759, -0.0696, -0.0316, 0.0, 0.0316, 0.0095, 0.0, -0.0253, -0.0253, -0.019, -0.0316, -0.0253, -0.0316, -0.038, -0.038, -0.0127, -0.0127, -0.0348, -0.0411, -0.0127, -0.0158, -0.0316, -0.0222, -0.0158, -0.0127, -0.0032, -0.0095, -0.019, -0.019, -0.0127, -0.0253, -0.0158, -0.0127, 0.0158, 0.0253, 0.0316, 0.0443, 0.019, 0.0475, 0.0626, 0.0757, 0.056, 0.056, 0.0429, 0.056, 0.0593, 0.0593, 0.0527, 0.0691, 0.0658, 0.0757, 0.0855, 0.0822, 0.0626, 0.0888, 0.1413, 0.1445, 0.1544, 0.1249, 0.1249, 0.1675, 0.1675, 0.1806, 0.2101, 0.22, 0.2757, 0.2757, 0.2987, 0.2921, 0.302, 0.3085, 0.3577, 0.3774, 0.3643, 0.3151, 0.3511, 0.3282, 0.361, 0.3708, 0.3643, 0.3511, 0.3348, 0.302, 0.302, 0.3151, 0.3315, 0.3446, 0.3348, 0.361, 0.3774, 0.4069, 0.4266, 0.4167, 0.4135, 0.3905, 0.4102, 0.4036, 0.4167, 0.4495, 0.4627, 0.5086, 0.4397, 0.4397, 0.4167, 0.4364, 0.479, 0.4856, 0.4856, 0.4987, 0.4758, 0.4528, 0.4528, 0.3971, 0.3511, 0.3544, 0.4627, 0.4856, 0.5643, 0.5676, 0.5906, 0.5676, 0.5676, 0.6168, 0.6397, 0.7053, 0.7185, 0.6889, 0.6496, 0.6004, 0.6201, 0.6266, 0.6004, 0.6365]}
{"unit_id": "orig:1942350980340932825", "ticker_idx": 0, "ts": "2025-07-07T22:32:15+00:00", "symbol": "memory", "kind": "sector", "geography": "US", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley has raised its 2025 growth forecast from 12.9% to 14.8%. The full-year upward revision is primarily attributed to the strong performance of the memory industry", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley raising 2025 semiconductor revenue growth forecast to 14.8% on strong AI-driven memory; bullish on memory and AI/analog/equipment cyclicals.", "resolved_tickers": ["MU"], "resolution_method": "exchange_ticker", "resolution_reason": "US DRAM memory = Micron Technology (MU)", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SIA Semiconductor Revenue: May Overall Performance Exceeds Expectations, Strong Memory Drives Upward Revision of Full-Year Growth Forecast\nMorgan Stanley (J. Moore, 25/07/07)\n\nMorgan Stanley notes that despite significant month-over-month declines in analog, discrete, and MCU segments, the May SIA monthly data was better than overall expectations, primarily driven by strong AI-led memory demand. The full-year revenue growth forecast has been raised from 12.9% to 14.8%.\n\nMemory Market Shines, DRAM and NAND Both Significantly Outperform Seasonality\nDRAM saw a 48.4% month-over-month increase, exceeding Morgan Stanley's 33% forecast and the five-year average of 40.7%, mainly driven by strong bit shipments. NAND increased by 37.4% month-over-month, also significantly outperforming expectations, with both bit shipments and Average Selling Price (ASP) exceeding forecasts. Morgan Stanley believes this reflects the sustained pull from AI for memory demand, potentially accompanied by some pull-in effects.\n\nAnalog, Discrete, and MCU MoM Data Significantly Below Expectations, Broad Market Softens in Short Term\nIn May, analog chips declined by 3.4% month-over-month, discrete devices by 10.2%, and MCUs by 5.4%, all falling below seasonal averages and estimates. This indicates that the broad market weakness was worse than expected. Despite this, Morgan Stanley still judges this trend to be a short-term anomaly, believing that relevant sectors are in the process of channel recovery for inventory replenishment.\n\nOverall SIA Revenue MoM Growth of 9.5%, Significantly Higher Than Expected and Historical Average\nMorgan Stanley's estimate was 6.2%, and the historical 10-year average was 6.3%. Regional performance showed significant divergence, with Asia Pacific and the Americas leading the growth (at +29.1% and +23.8% respectively). China and Europe showed moderate performance, while Japan experienced negative growth (-5.4%).\n\n2025 Growth Forecast Raised to 14.8%, Primarily Due to Improved Memory Shipment and Pricing Expectations\nMorgan Stanley has raised its 2025 growth forecast from 12.9% to 14.8%. The full-year upward revision is primarily attributed to the strong performance of the memory industry. Although strong performance is expected in the short term, given the record-high DRAM Wafer Fab Equipment (WFE) investment, the long-term memory growth path still requires caution regarding potential challenges.\n\nRecommended Portfolio Remains Unchanged, AI-Driven and Quality Cyclical Beneficiaries Remain Key Allocation Focus\nIn the phase where overall inventory shifts from destocking to restocking, Morgan Stanley maintains a positive view on AI-related companies and analog/equipment cyclical beneficiaries. Although some areas showed below-seasonal data in May, the three-month average trend remains near its annual high, supporting the recovery judgment.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01880419588439408, "ret_signed_m1d": 0.01880419588439408, "alpha_spy_m1d": 0.011296229305892913, "alpha_c_m1d": 0.005006670186103124, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.037524991430599286, "ret_signed_p1d": 0.037524991430599286, "alpha_spy_p1d": 0.038072707648344584, "alpha_c_p1d": 0.024227862942426848, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": -0.010923978903346065, "ret_signed_p1w": -0.010923978903346065, "alpha_spy_p1w": -0.017577892961980868, "alpha_c_p1w": -0.030798162871814472, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": -0.090560406947777, "ret_signed_p1m": -0.090560406947777, "alpha_spy_p1m": -0.10230551332699989, "alpha_c_p1m": -0.11679717481730822, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.5322715346809062, "ret_signed_p3m": 0.5322715346809062, "alpha_spy_p3m": 0.45107783662060297, "alpha_c_p3m": 0.32323527804158836, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 1.442724674394631, "ret_signed_p6m": 1.442724674394631, "alpha_spy_p6m": 1.329509461192399, "alpha_c_p6m": 1.1400350257122087, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.3513, -0.4164, -0.4206, -0.4083, -0.4082, -0.4224, -0.4268, -0.4268, -0.444, -0.4151, -0.3924, -0.355, -0.3353, -0.3455, -0.3595, -0.3589, -0.3521, -0.3275, -0.33, -0.3293, -0.3117, -0.2906, -0.2847, -0.2311, -0.1925, -0.2059, -0.2048, -0.1836, -0.1781, -0.1827, -0.2016, -0.21, -0.2221, -0.2221, -0.1971, -0.1987, -0.1936, -0.213, -0.1821, -0.1482, -0.1398, -0.1145, -0.0956, -0.0757, -0.0491, -0.0333, -0.0321, -0.0369, -0.0016, 0.0026, 0.0149, 0.0149, 0.0297, 0.0171, 0.0656, 0.0601, 0.0497, 0.0394, 0.0268, 0.0071, 0.0142, 0.0188, 0.0188, 0.0, 0.0375, 0.0193, 0.0266, 0.0384, -0.0109, 0.0016, -0.0291, -0.0555, -0.0461, -0.0558, -0.0892, -0.0841, -0.0683, -0.0722, -0.0723, -0.0664, -0.0432, -0.0899, -0.1254, -0.1013, -0.0906, -0.0929, -0.0671, -0.0086, 0.0317, 0.0653, 0.0363, 0.0448, 0.0079, 0.0303, 0.0178, -0.0226, -0.0344, -0.0187, -0.0292, -0.0285, -0.0181, 0.0173, -0.0076, -0.0076, -0.012, -0.01, 0.0358, 0.0955, 0.0962, 0.1278, 0.1674, 0.2556, 0.3111, 0.3156, 0.3244, 0.3341, 0.4084, 0.357, 0.3727, 0.3877, 0.3485, 0.3078, 0.3115, 0.3667, 0.3953, 0.5189, 0.5323, 0.5673, 0.5934, 0.5494, 0.64, 0.6048, 0.5153, 0.6085, 0.5609, 0.6016, 0.6899, 0.6887, 0.7253, 0.6879, 0.6561, 0.7248, 0.8275, 0.8365, 0.8516, 0.891, 0.8692, 0.8672, 0.9584, 0.8193, 0.9817, 0.9887, 0.9852, 1.1136, 1.0118, 1.0435, 0.9771, 1.0596, 1.0189, 0.9066, 0.8851, 0.6803, 0.7303, 0.8685, 0.8735, 0.9213, 0.9213, 0.9732, 1.0064, 0.9983, 0.9539, 0.8912, 0.9794, 1.0603, 1.1062, 1.2004, 1.1566, 1.0121, 0.9817, 0.9401, 0.8818, 1.0739, 1.2189, 1.3079, 1.3052, 1.3921, 1.3921, 1.3763, 1.4572, 1.4427]}
{"unit_id": "orig:1942350980340932825", "ticker_idx": 1, "ts": "2025-07-07T22:32:15+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "Morgan Stanley maintains a positive view on AI-related companies and analog/equipment cyclical beneficiaries", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley raising 2025 semiconductor revenue growth forecast to 14.8% on strong AI-driven memory; bullish on memory and AI/analog/equipment cyclicals.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "US semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "SIA Semiconductor Revenue: May Overall Performance Exceeds Expectations, Strong Memory Drives Upward Revision of Full-Year Growth Forecast\nMorgan Stanley (J. Moore, 25/07/07)\n\nMorgan Stanley notes that despite significant month-over-month declines in analog, discrete, and MCU segments, the May SIA monthly data was better than overall expectations, primarily driven by strong AI-led memory demand. The full-year revenue growth forecast has been raised from 12.9% to 14.8%.\n\nMemory Market Shines, DRAM and NAND Both Significantly Outperform Seasonality\nDRAM saw a 48.4% month-over-month increase, exceeding Morgan Stanley's 33% forecast and the five-year average of 40.7%, mainly driven by strong bit shipments. NAND increased by 37.4% month-over-month, also significantly outperforming expectations, with both bit shipments and Average Selling Price (ASP) exceeding forecasts. Morgan Stanley believes this reflects the sustained pull from AI for memory demand, potentially accompanied by some pull-in effects.\n\nAnalog, Discrete, and MCU MoM Data Significantly Below Expectations, Broad Market Softens in Short Term\nIn May, analog chips declined by 3.4% month-over-month, discrete devices by 10.2%, and MCUs by 5.4%, all falling below seasonal averages and estimates. This indicates that the broad market weakness was worse than expected. Despite this, Morgan Stanley still judges this trend to be a short-term anomaly, believing that relevant sectors are in the process of channel recovery for inventory replenishment.\n\nOverall SIA Revenue MoM Growth of 9.5%, Significantly Higher Than Expected and Historical Average\nMorgan Stanley's estimate was 6.2%, and the historical 10-year average was 6.3%. Regional performance showed significant divergence, with Asia Pacific and the Americas leading the growth (at +29.1% and +23.8% respectively). China and Europe showed moderate performance, while Japan experienced negative growth (-5.4%).\n\n2025 Growth Forecast Raised to 14.8%, Primarily Due to Improved Memory Shipment and Pricing Expectations\nMorgan Stanley has raised its 2025 growth forecast from 12.9% to 14.8%. The full-year upward revision is primarily attributed to the strong performance of the memory industry. Although strong performance is expected in the short term, given the record-high DRAM Wafer Fab Equipment (WFE) investment, the long-term memory growth path still requires caution regarding potential challenges.\n\nRecommended Portfolio Remains Unchanged, AI-Driven and Quality Cyclical Beneficiaries Remain Key Allocation Focus\nIn the phase where overall inventory shifts from destocking to restocking, Morgan Stanley maintains a positive view on AI-related companies and analog/equipment cyclical beneficiaries. Although some areas showed below-seasonal data in May, the three-month average trend remains near its annual high, supporting the recovery judgment.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013797525698290958, "ret_signed_m1d": 0.013797525698290958, "alpha_spy_m1d": 0.006289559119789789, "alpha_c_m1d": 0.006289559119789789, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.0075079665785011684, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013297128488172438, "ret_signed_p1d": 0.013297128488172438, "alpha_spy_p1d": 0.013844844705917736, "alpha_c_p1d": 0.013844844705917736, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": -0.0005477162177452977, "ret_p1w": 0.019874183968468406, "ret_signed_p1w": 0.019874183968468406, "alpha_spy_p1w": 0.013220269909833604, "alpha_c_p1w": 0.013220269909833604, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.006653914058634802, "ret_p1m": 0.02623676786953122, "ret_signed_p1m": 0.02623676786953122, "alpha_spy_p1m": 0.014491661490308339, "alpha_c_p1m": 0.014491661490308339, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.01174510637922288, "ret_p3m": 0.2090362566393178, "ret_signed_p3m": 0.2090362566393178, "alpha_spy_p3m": 0.1278425585790146, "alpha_c_p3m": 0.1278425585790146, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.08119369806030319, "ret_p6m": 0.3026896486824222, "ret_signed_p6m": 0.3026896486824222, "alpha_spy_p6m": 0.18947443548019027, "alpha_c_p6m": 0.18947443548019027, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.11321521320223193, "price_path": [-0.2464, -0.2986, -0.2804, -0.2793, -0.2747, -0.3053, -0.3118, -0.3118, -0.3286, -0.3141, -0.289, -0.253, -0.2423, -0.247, -0.2483, -0.2448, -0.2411, -0.2171, -0.2222, -0.2294, -0.2135, -0.2066, -0.2015, -0.1514, -0.1222, -0.1136, -0.1166, -0.1192, -0.1206, -0.1223, -0.1369, -0.1392, -0.1513, -0.1513, -0.1244, -0.1337, -0.1275, -0.143, -0.1303, -0.1107, -0.1001, -0.1017, -0.0966, -0.0815, -0.0632, -0.0639, -0.0594, -0.0814, -0.0586, -0.0649, -0.0614, -0.0614, -0.0697, -0.0638, -0.0294, -0.0163, -0.0092, -0.0048, -0.0031, -0.014, 0.0053, 0.0138, 0.0138, 0.0, 0.0133, 0.0202, 0.0277, 0.0276, 0.0199, 0.0395, 0.034, 0.0427, 0.0377, 0.0384, 0.0201, 0.0245, 0.0287, 0.0276, 0.0416, 0.047, 0.059, 0.0322, 0.015, 0.0373, 0.0262, 0.0245, 0.0406, 0.0492, 0.0488, 0.0727, 0.0769, 0.0792, 0.0568, 0.0611, 0.0397, 0.0329, 0.0275, 0.0496, 0.0506, 0.0607, 0.064, 0.0683, 0.0376, 0.0376, 0.024, 0.0238, 0.0361, 0.0483, 0.0599, 0.0645, 0.0753, 0.0864, 0.0878, 0.098, 0.0981, 0.0911, 0.133, 0.1285, 0.1516, 0.1499, 0.1478, 0.1476, 0.1503, 0.1533, 0.1666, 0.1928, 0.209, 0.2035, 0.2274, 0.2048, 0.237, 0.234, 0.1629, 0.2144, 0.1921, 0.2217, 0.2271, 0.2254, 0.2411, 0.2343, 0.2103, 0.2325, 0.2551, 0.2866, 0.2979, 0.3176, 0.3002, 0.2976, 0.3086, 0.2609, 0.285, 0.255, 0.2444, 0.2825, 0.2548, 0.2708, 0.2326, 0.2331, 0.2164, 0.1913, 0.2133, 0.1621, 0.1657, 0.2122, 0.2154, 0.2429, 0.2429, 0.2592, 0.2617, 0.2848, 0.3024, 0.2925, 0.3026, 0.3174, 0.319, 0.3372, 0.3257, 0.2658, 0.2614, 0.258, 0.2126, 0.2412, 0.2733, 0.2897, 0.3022, 0.3058, 0.3058, 0.3118, 0.3059, 0.3027]}
{"unit_id": "reply:1921343815513673867", "ticker_idx": 0, "ts": "2025-05-10T23:17:17+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I personally think a short-term rebound in the stock price won't come easily", "tweet_type": "prediction", "timeline": "weeks", "summary": "Skeptical a short-term rebound in Micron is coming; implicitly bearish near-term.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@OmerCheeema I enjoyed reading your post! That said, I personally think a short-term rebound in the stock price won’t come easily.", "reply_text": null, "parent_text": "@Jukanlosreve I wrote an article on $MU a couple of months ago. In my model, I was expecting a ramp up to $1.5B for Micron HBM memory. Morgan Stanley is more bullish at $2.5B. \n\nOur fair stock price estimates are similar. \n\nFull article is here: \n\nhttps://t.co/VjKCAJXeBP", "quoted_text": null, "parent_author": "OmerCheeema", "ret_m1d": -0.06967173414031069, "ret_signed_m1d": 0.06967173414031069, "alpha_spy_m1d": 0.03768142296917443, "alpha_c_m1d": 0.010659852037967932, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05027631913789388, "ret_signed_p1d": -0.05027631913789388, "alpha_spy_p1d": -0.043672378820723834, "alpha_c_p1d": -0.015947559374998255, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.068913266260344, "ret_signed_p1w": -0.068913266260344, "alpha_spy_p1w": -0.048569965986598795, "alpha_c_p1w": -0.03268903333222295, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.23675368123748686, "ret_signed_p1m": -0.23675368123748686, "alpha_spy_p1m": -0.20229351078924096, "alpha_c_p1m": -0.13288271886127534, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.21329835415038878, "ret_signed_p3m": -0.21329835415038878, "alpha_spy_p3m": -0.12559695933220105, "alpha_c_p3m": 0.01293468499132544, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 1.366148981056865, "ret_signed_p6m": -1.366148981056865, "alpha_spy_p6m": -1.2012682959938907, "alpha_c_p6m": -0.8802806038739521, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [-0.0079, 0.0352, 0.0769, 0.0769, 0.1556, 0.1293, 0.1165, 0.0696, 0.0325, 0.0088, 0.0574, -0.0064, 0.0132, -0.0202, -0.0131, 0.0209, -0.034, 0.0059, -0.0577, -0.0364, 0.035, 0.0267, 0.0907, 0.1158, 0.1007, 0.1044, 0.1146, 0.025, 0.049, 0.0192, -0.003, -0.0135, -0.043, -0.0585, -0.0388, -0.04, -0.1945, -0.2987, -0.2592, -0.2898, -0.1562, -0.241, -0.2464, -0.2305, -0.2303, -0.2488, -0.2545, -0.2545, -0.2768, -0.2392, -0.2098, -0.1611, -0.1356, -0.1488, -0.167, -0.1662, -0.1573, -0.1254, -0.1286, -0.1276, -0.1048, -0.0774, -0.0697, 0.0, 0.0503, 0.0328, 0.0342, 0.0619, 0.0689, 0.063, 0.0385, 0.0275, 0.0117, 0.0117, 0.0443, 0.0421, 0.0489, 0.0235, 0.0638, 0.1079, 0.1188, 0.1517, 0.1763, 0.2022, 0.2368, 0.2572, 0.2589, 0.2526, 0.2985, 0.3039, 0.32, 0.32, 0.3393, 0.3228, 0.386, 0.3788, 0.3653, 0.3518, 0.3355, 0.3099, 0.3191, 0.3251, 0.3251, 0.3006, 0.3494, 0.3258, 0.3352, 0.3506, 0.2864, 0.3027, 0.2628, 0.2284, 0.2406, 0.228, 0.1846, 0.1912, 0.2118, 0.2067, 0.2066, 0.2143, 0.2444, 0.1837, 0.1375, 0.1688, 0.1828, 0.1798, 0.2133, 0.2894, 0.3418, 0.3855, 0.3478, 0.3588, 0.3109, 0.34, 0.3237, 0.2712, 0.2558, 0.2763, 0.2626, 0.2635, 0.2771, 0.3232, 0.2907, 0.2907, 0.285, 0.2876, 0.3471, 0.4248, 0.4258, 0.4668, 0.5184, 0.633, 0.7053, 0.7111, 0.7225, 0.7352, 0.8317, 0.7649, 0.7854, 0.8048, 0.7538, 0.7009, 0.7057, 0.7776, 0.8147, 0.9755, 0.9929, 1.0384, 1.0724, 1.0152, 1.1329, 1.0872, 0.9708, 1.092, 1.0301, 1.083, 1.1979, 1.1963, 1.244, 1.1953, 1.1539, 1.2433, 1.3769, 1.3886, 1.4083, 1.4595, 1.431, 1.4284, 1.5471, 1.3661]}
{"unit_id": "orig:1925348719459016980", "ticker_idx": 0, "ts": "2025-05-22T00:31:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's Growth Will Continue to Outpace ASICs", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Jensen Huang argues NVIDIA's AI chip growth will outpace ASICs due to fast architecture cadence, cost reduction, and NVLink collaboration potential.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA CEO Jensen Huang: NVIDIA’s Growth Will Continue to Outpace ASICs\n\nJensen Huang explained that the reason he believes NVIDIA’s AI chips have greater growth potential than ASICs is because, while there are many ASIC projects emerging in the world, about 90% of them fail. He compared this to how new startups constantly emerge, but most of them end up failing. Even if some projects manage to avoid failure, he emphasized that sustaining them over the long term is often difficult.\n\nIn contrast, he stated that continuing NVIDIA’s growth is not particularly difficult, and that this is precisely his responsibility. Since NVIDIA develops at a very fast pace, anyone trying to build an ASIC would have to make one that’s at least better than NVIDIA’s own ASICs.\n\nWhile acknowledging that market competition is indeed fierce, Huang emphasized that NVIDIA continues to optimize its architecture and is aggressively working to reduce costs. As a result, its technology is already widely adopted. He pointed out that ASIC vendors must compete not only with NVIDIA, but also with one another.\n\nHe also revealed that many companies are trying to develop semi-custom AI infrastructure and have approached NVIDIA to request access to its high-speed GPU interconnect technology, NVLink. NVIDIA welcomes this, he said, and sees potential benefit from such collaborations. \n\nhttps://t.co/IyaxQCdvD2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007754252050075072, "ret_signed_m1d": -0.007754252050075072, "alpha_spy_m1d": -0.007359972366447476, "alpha_c_m1d": -0.010495028149897223, "bench_spy_m1d": -0.00039427968362759636, "bench_c_m1d": 0.002740776099822151, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01159380523896325, "ret_signed_p1d": -0.01159380523896325, "alpha_spy_p1d": -0.004768129373284302, "alpha_c_p1d": 0.002400615323976396, "bench_spy_p1d": -0.0068256758656789485, "bench_c_p1d": -0.013994420562939647, "ret_p1w": 0.047880777139537534, "ret_signed_p1w": 0.047880777139537534, "alpha_spy_p1w": 0.03594439380651471, "alpha_c_p1w": 0.03430160449543229, "bench_spy_p1w": 0.011936383333022826, "bench_c_p1w": 0.013579172644105242, "ret_p1m": 0.08303838082447901, "ret_signed_p1m": 0.08303838082447901, "alpha_spy_p1m": 0.06083442091819147, "alpha_c_p1m": 0.00222781790389992, "bench_spy_p1m": 0.02220395990628754, "bench_c_p1m": 0.08081056292057909, "ret_p3m": 0.322383436882691, "ret_signed_p3m": 0.322383436882691, "alpha_spy_p3m": 0.22186466900034985, "alpha_c_p3m": 0.11450167162457059, "bench_spy_p3m": 0.10051876788234115, "bench_c_p3m": 0.2078817652581204, "ret_p6m": 0.4318597657661136, "ret_signed_p6m": 0.4318597657661136, "alpha_spy_p6m": 0.2728879760328009, "alpha_c_p6m": -0.0007220364212952024, "bench_spy_p6m": 0.15897178973331272, "bench_c_p6m": 0.4325818021874088, "price_path": [-0.0193, -0.0468, -0.0118, -0.0955, -0.0596, -0.1414, -0.1269, -0.117, -0.1677, -0.1517, -0.1947, -0.1813, -0.1287, -0.1299, -0.084, -0.1001, -0.131, -0.1153, -0.1077, -0.1139, -0.086, -0.0914, -0.1436, -0.1611, -0.1744, -0.1841, -0.1707, -0.1687, -0.2336, -0.29, -0.2649, -0.275, -0.1393, -0.1902, -0.1649, -0.1665, -0.1553, -0.2134, -0.2359, -0.2359, -0.2704, -0.2555, -0.2268, -0.1988, -0.1643, -0.1814, -0.1793, -0.18, -0.1598, -0.138, -0.1431, -0.1452, -0.1187, -0.1164, -0.1218, -0.074, -0.0218, 0.0189, 0.0151, 0.0193, 0.0206, 0.0117, -0.0078, 0.0, -0.0116, -0.0116, 0.0201, 0.0149, 0.0479, 0.0173, 0.0343, 0.0632, 0.0684, 0.0539, 0.0669, 0.0738, 0.0838, 0.0754, 0.0917, 0.0689, 0.0894, 0.0851, 0.0953, 0.0953, 0.083, 0.0854, 0.1135, 0.1618, 0.1671, 0.1877, 0.1895, 0.1542, 0.1839, 0.1997, 0.1997, 0.1914, 0.2046, 0.2263, 0.2355, 0.2417, 0.2353, 0.2852, 0.2902, 0.3025, 0.2981, 0.2903, 0.2576, 0.2858, 0.3081, 0.3063, 0.3307, 0.3214, 0.3497, 0.3392, 0.3079, 0.3552, 0.3421, 0.3508, 0.361, 0.3755, 0.3707, 0.379, 0.3672, 0.3704, 0.3586, 0.3703, 0.3224, 0.3206, 0.3174, 0.3401, 0.3538, 0.3685, 0.3673, 0.3565, 0.3114, 0.3114, 0.2858, 0.2846, 0.2924, 0.2575, 0.2672, 0.2856, 0.3351, 0.334, 0.3389, 0.3383, 0.3167, 0.2822, 0.327, 0.3302, 0.3825, 0.3435, 0.3325, 0.3379, 0.3417, 0.3692, 0.4048, 0.4098, 0.4222, 0.4127, 0.397, 0.3932, 0.4239, 0.4499, 0.3791, 0.4179, 0.3555, 0.354, 0.3689, 0.3795, 0.3752, 0.364, 0.3574, 0.3715, 0.4024, 0.4418, 0.5136, 0.5589, 0.5276, 0.5246, 0.5577, 0.496, 0.4698, 0.4161, 0.4167, 0.4987, 0.4544, 0.4592, 0.4069, 0.4319]}
{"unit_id": "orig:1940163911657169108", "ticker_idx": 0, "ts": "2025-07-01T21:41:37+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I expect strong performance from AVGO", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Expects strong performance from Broadcom going forward.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I expect strong performance from AVGO.\n\n$AVGO https://t.co/FKflGg9LYQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04121022606265656, "ret_signed_m1d": 0.04121022606265656, "alpha_spy_m1d": 0.0408864419522148, "alpha_c_m1d": 0.03018931278490089, "bench_spy_m1d": 0.0003237841104417605, "bench_c_m1d": 0.011020913277755673, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01949086253144583, "ret_signed_p1d": 0.01949086253144583, "alpha_spy_p1d": 0.014957586015629909, "alpha_c_p1d": -0.00012195944317161533, "bench_spy_p1d": 0.0045332765158159205, "bench_c_p1d": 0.019612821974617445, "ret_p1w": 0.02666769771454014, "ret_signed_p1w": 0.02666769771454014, "alpha_spy_p1w": 0.022312507442294516, "alpha_c_p1w": -0.00102963546327417, "bench_spy_p1w": 0.004355190272245624, "bench_c_p1w": 0.02769733317781431, "ret_p1m": 0.143083977805883, "ret_signed_p1m": 0.143083977805883, "alpha_spy_p1m": 0.11586797503782287, "alpha_c_p1m": 0.06905543813959558, "bench_spy_p1m": 0.027216002768060132, "bench_c_p1m": 0.07402853966628742, "ret_p3m": 0.26578230738079944, "ret_signed_p3m": 0.26578230738079944, "alpha_spy_p3m": 0.19129860800233733, "alpha_c_p3m": 0.0990913547375658, "bench_spy_p3m": 0.07448369937846211, "bench_c_p3m": 0.16669095264323364, "ret_p6m": 0.32768506140319054, "ret_signed_p6m": 0.32768506140319054, "alpha_spy_p6m": 0.20352147877436244, "alpha_c_p6m": 0.0033550741336294543, "bench_spy_p6m": 0.1241635826288281, "bench_c_p6m": 0.3243299872695611, "price_path": [-0.4196, -0.4487, -0.4191, -0.412, -0.3023, -0.3507, -0.3144, -0.3279, -0.3256, -0.342, -0.3556, -0.3556, -0.3737, -0.361, -0.3333, -0.291, -0.2753, -0.2747, -0.2796, -0.2747, -0.2564, -0.2326, -0.2436, -0.246, -0.2282, -0.217, -0.2154, -0.165, -0.1241, -0.1253, -0.1233, -0.1385, -0.1309, -0.1269, -0.1343, -0.1313, -0.1381, -0.1381, -0.112, -0.0977, -0.0882, -0.0878, -0.0628, -0.0321, -0.0161, -0.0205, -0.0695, -0.0795, -0.0781, -0.0469, -0.035, -0.0628, -0.05, -0.0603, -0.0531, -0.0531, -0.0557, -0.0414, -0.0037, -0.0003, 0.0205, 0.0174, 0.0412, 0.0, 0.0195, 0.0394, 0.0394, 0.0357, 0.0267, 0.0497, 0.0403, 0.0364, 0.041, 0.0612, 0.0607, 0.082, 0.0703, 0.0887, 0.0523, 0.0716, 0.0905, 0.0961, 0.1117, 0.1234, 0.1431, 0.1094, 0.0903, 0.1246, 0.1065, 0.1395, 0.1474, 0.152, 0.1479, 0.1817, 0.1675, 0.1756, 0.1571, 0.1549, 0.114, 0.0998, 0.0939, 0.1105, 0.1114, 0.1257, 0.1341, 0.1659, 0.1233, 0.1233, 0.1265, 0.1422, 0.1562, 0.265, 0.3056, 0.2717, 0.396, 0.3584, 0.3593, 0.3753, 0.3598, 0.3076, 0.3045, 0.3029, 0.2819, 0.2825, 0.2839, 0.2717, 0.2658, 0.2407, 0.2483, 0.2615, 0.2796, 0.2803, 0.2694, 0.2729, 0.3073, 0.3055, 0.2283, 0.3497, 0.3021, 0.3293, 0.34, 0.3218, 0.3214, 0.2965, 0.2876, 0.3027, 0.3399, 0.3699, 0.4112, 0.4605, 0.4245, 0.3986, 0.3718, 0.3317, 0.3583, 0.3455, 0.3222, 0.3561, 0.3317, 0.3441, 0.2864, 0.2958, 0.2965, 0.2884, 0.341, 0.3123, 0.2872, 0.4301, 0.4569, 0.5043, 0.5043, 0.5247, 0.4608, 0.4438, 0.4401, 0.4417, 0.4766, 0.5177, 0.5373, 0.5626, 0.5376, 0.3619, 0.2858, 0.2914, 0.2336, 0.2482, 0.2878, 0.2944, 0.3243, 0.3277]}
{"unit_id": "orig:1925327380933153174", "ticker_idx": 0, "ts": "2025-05-21T23:06:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain an Overweight rating on Samsung Electronics and SK Hynix, as valuation risks are low and fundamentals are expected to improve.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan Overweight Samsung, SK Hynix; prefers Advantest, Tokyo Electron, Hanmi on HBM/DDR5/3D NAND equipment cycle into 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Equipment / Foundry: Marginal Improvement in Industry Sentiment\n JP Morgan (J. Kwon, May 20, 2025)\n\n1QCY25 results and outlook indicate that as chip stacking and process node evolution continues, demand for testing high-performance memory such as HBM (High Bandwidth Memory) and DDR5 is rising sharply. This is driving stable to mid-single-digit growth in both front-end and back-end testing and wafer fabrication equipment (WFE) markets. While the WFE market is projected to reach approximately $100 billion for the full year 2025 with a broadly optimistic outlook for 2026, macroeconomic and geopolitical uncertainties—especially tariffs—in 2H25 remain key concerns. We prefer Samsung Electronics (SEC), SK Hynix, Tokyo Electron, Advantest, and Hanmi to capture long-term structural opportunities.\n\n• WFE Market: Solid Growth Momentum\n\n• Leading manufacturers forecast the 2025 WFE market at $99–110 billion, indicating stable to mid-single-digit growth.\n\n• Leading equipment vendors (LAM, KLA) are optimistic about continued adoption by leading foundries and logic fabs. Tokyo Electron expects double-digit growth in 2026.\n\n• DRAM and NAND together are projected to account for around 35% of WFE demand, with DRAM growing to 20% due to HBM and DDR5 transitions, and NAND reaching 10%.\n\n• HBM and Substrate Testing Ramping Up\n\n• Front-end equipment suppliers like KLA and TEL classify HBM as a testing-intensive technology, benefiting from increased stacking layers and stricter quality control.\n\n• Back-end testing equipment vendors (Advantest, Hanmi) report a significant increase in demand for HBM-related tools, with orders expected to continue releasing through 2H25.\n\n• Applied Materials forecasts over 40% YoY revenue growth from HBM and DDR5 customers in 2025, reflecting surging demand for high-value-added memory process equipment.\n\n• DRAM Technology Evolution Drives Equipment Refresh\n\n• ASML is promoting broader use of EUV to simplify DRAM multi-patterning to single exposure, improving efficiency and output.\n\n• LAM Research (LRCX) has made breakthroughs in etch technology, with its Akara system adopted for critical applications by major DRAM makers.\n\n• ASMI sees increased demand for ALD and epitaxy tools due to high-k metal gate processes and future 4F² or vertical channel transistor adoption.\n\n• NAND Demand Driven by Technology Transition\n\n• LRCX’s ALTUS Halo system advances atomic layer deposition (ALD) for 3D NAND inter-layer technology, enabling low-resistance interconnects for 256+ layers.\n\n• 3xx/4xx-layer NAND processes are expected to be adopted gradually over the next 3–5 years, driving a new round of equipment upgrades.\n\n• Tokyo Electron has secured new orders in NAND channel etch and low-temperature dry etch, expecting 2026 NAND investments to focus on brownfield upgrades and capacity enhancement.\n\n• Macro and Tariff Risk Mitigation\n\n• While direct tariff impact is limited, the industry is closely watching potential disruptions to end-customer purchasing patterns and CapEx decisions.\n\n• Major equipment vendors are strengthening supply chain diversification and regionalization to improve geopolitical risk resilience.\n\n• Test equipment vendors like Advantest and Teradyne state that HPC and AI markets are relatively insulated from tariffs but continue to monitor macro demand shifts closely.\n\n• China Market Developments\n• China contributed 27% of total sales in 1Q25, down from 38% in 2024. ASML remains most optimistic, forecasting over 25% of its sales from China.\n\n• LRCX, KLA, and AMAT all reported reduced spending in China on DRAM and mature logic nodes, though capacity build-out continues for 28nm and above.\n\n• ASMI reported strong margins in China, and Teradyne saw a slight increase in local customer share. Overall demand in the China market is expected to gradually stabilize.\n\n• Investment Recommendations and Catalysts\n\n• We maintain an Overweight rating on Samsung Electronics and SK Hynix, as valuation risks are low and fundamentals are expected to improve.\n\n• In the SPE (semiconductor production equipment) sector, we prefer Advantest and Tokyo Electron, which are set to benefit from equipment growth driven by HBM, DDR5, and 3D NAND technologies.\n\n• Medium-term catalysts include:\n1. Progress in 2026 equipment order negotiations among major manufacturers,\n2. Tariff resolution developments, and\n3. Expansion of AI accelerated computing demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0035906204532480945, "ret_signed_m1d": 0.0035906204532480945, "alpha_spy_m1d": -0.013548881236552957, "alpha_c_m1d": -0.015556192573253691, "bench_spy_m1d": 0.01713950168980105, "bench_c_m1d": 0.019146813026501786, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017953387734554127, "ret_signed_p1d": -0.017953387734554127, "alpha_spy_p1d": -0.01834782293596815, "alpha_c_p1d": -0.018215021869636416, "bench_spy_p1d": 0.0003944352014140229, "bench_c_p1d": 0.0002616341350822893, "ret_p1w": 0.0035906204532480945, "ret_signed_p1w": 0.0035906204532480945, "alpha_spy_p1w": -0.004764648307102659, "alpha_c_p1w": -0.004957834837801878, "bench_spy_p1w": 0.008355268760350754, "bench_c_p1w": 0.008548455291049972, "ret_p1m": 0.0628365002355471, "ret_signed_p1m": 0.0628365002355471, "alpha_spy_p1m": 0.03782195782062914, "alpha_c_p1m": 0.009059541428165119, "bench_spy_p1m": 0.025014542414917962, "bench_c_p1m": 0.05377695880738198, "ret_p3m": 0.26444078933987103, "ret_signed_p3m": 0.26444078933987103, "alpha_spy_p3m": 0.1574825456653448, "alpha_c_p3m": 0.10118503799042045, "bench_spy_p3m": 0.10695824367452622, "bench_c_p3m": 0.16325575134945058, "ret_p6m": 0.8652067150996965, "ret_signed_p6m": 0.8652067150996965, "alpha_spy_p6m": 0.7055880886069541, "alpha_c_p6m": 0.6115697228574932, "bench_spy_p6m": 0.1596186264927424, "bench_c_p6m": 0.25363699224220326, "price_path": [0.0387, 0.0226, 0.0209, 0.0102, 0.0048, -0.0273, -0.0273, -0.0273, -0.0362, -0.0309, -0.0416, -0.0416, -0.0434, -0.0202, -0.0238, -0.0238, 0.028, 0.028, 0.0441, 0.0744, 0.1012, 0.0798, 0.0673, 0.0958, 0.103, 0.0808, 0.0377, 0.0557, 0.0557, 0.0341, 0.0072, -0.0449, -0.0395, -0.0485, 0.0126, -0.009, 0.009, 0.0162, -0.018, -0.0108, -0.0072, -0.0054, -0.0126, 0.0, 0.0, 0.0, 0.0018, 0.0018, -0.0036, -0.0036, -0.0251, -0.0251, -0.0251, -0.0197, -0.0197, -0.0162, 0.0341, 0.0215, 0.0305, 0.0287, 0.0197, 0.0018, 0.0036, 0.0, -0.018, -0.0269, -0.018, -0.0323, 0.0036, 0.0072, 0.009, 0.0197, 0.0197, 0.0377, 0.061, 0.061, 0.0736, 0.0628, 0.0754, 0.0682, 0.0467, 0.0269, 0.0431, 0.0736, 0.0628, 0.0682, 0.0413, 0.0862, 0.1005, 0.0808, 0.0983, 0.0802, 0.0874, 0.0983, 0.1524, 0.1434, 0.1145, 0.1091, 0.091, 0.1019, 0.1308, 0.129, 0.1506, 0.1687, 0.2048, 0.2121, 0.2247, 0.1922, 0.1994, 0.1922, 0.1904, 0.2717, 0.2753, 0.3114, 0.2897, 0.2446, 0.259, 0.2626, 0.2428, 0.2735, 0.297, 0.2825, 0.2843, 0.2988, 0.2933, 0.2933, 0.2644, 0.2644, 0.2735, 0.2753, 0.2897, 0.2915, 0.2699, 0.2753, 0.2572, 0.259, 0.2211, 0.2482, 0.2608, 0.2662, 0.2554, 0.2662, 0.2915, 0.3114, 0.3259, 0.362, 0.3819, 0.4342, 0.4126, 0.4505, 0.4505, 0.5083, 0.53, 0.5426, 0.5553, 0.5047, 0.5277, 0.5223, 0.5604, 0.6284, 0.6284, 0.6284, 0.6284, 0.6284, 0.6284, 0.7128, 0.6928, 0.662, 0.7237, 0.7727, 0.7763, 0.7799, 0.769, 0.789, 0.7509, 0.7926, 0.8507, 0.8053, 0.8235, 0.8888, 0.9505, 1.0158, 0.9033, 0.8253, 0.7999, 0.7763, 0.8253, 0.8779, 0.8706, 0.8652]}
{"unit_id": "orig:1925327380933153174", "ticker_idx": 1, "ts": "2025-05-21T23:06:33+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain an Overweight rating on Samsung Electronics and SK Hynix, as valuation risks are low and fundamentals are expected to improve.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan Overweight Samsung, SK Hynix; prefers Advantest, Tokyo Electron, Hanmi on HBM/DDR5/3D NAND equipment cycle into 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Equipment / Foundry: Marginal Improvement in Industry Sentiment\n JP Morgan (J. Kwon, May 20, 2025)\n\n1QCY25 results and outlook indicate that as chip stacking and process node evolution continues, demand for testing high-performance memory such as HBM (High Bandwidth Memory) and DDR5 is rising sharply. This is driving stable to mid-single-digit growth in both front-end and back-end testing and wafer fabrication equipment (WFE) markets. While the WFE market is projected to reach approximately $100 billion for the full year 2025 with a broadly optimistic outlook for 2026, macroeconomic and geopolitical uncertainties—especially tariffs—in 2H25 remain key concerns. We prefer Samsung Electronics (SEC), SK Hynix, Tokyo Electron, Advantest, and Hanmi to capture long-term structural opportunities.\n\n• WFE Market: Solid Growth Momentum\n\n• Leading manufacturers forecast the 2025 WFE market at $99–110 billion, indicating stable to mid-single-digit growth.\n\n• Leading equipment vendors (LAM, KLA) are optimistic about continued adoption by leading foundries and logic fabs. Tokyo Electron expects double-digit growth in 2026.\n\n• DRAM and NAND together are projected to account for around 35% of WFE demand, with DRAM growing to 20% due to HBM and DDR5 transitions, and NAND reaching 10%.\n\n• HBM and Substrate Testing Ramping Up\n\n• Front-end equipment suppliers like KLA and TEL classify HBM as a testing-intensive technology, benefiting from increased stacking layers and stricter quality control.\n\n• Back-end testing equipment vendors (Advantest, Hanmi) report a significant increase in demand for HBM-related tools, with orders expected to continue releasing through 2H25.\n\n• Applied Materials forecasts over 40% YoY revenue growth from HBM and DDR5 customers in 2025, reflecting surging demand for high-value-added memory process equipment.\n\n• DRAM Technology Evolution Drives Equipment Refresh\n\n• ASML is promoting broader use of EUV to simplify DRAM multi-patterning to single exposure, improving efficiency and output.\n\n• LAM Research (LRCX) has made breakthroughs in etch technology, with its Akara system adopted for critical applications by major DRAM makers.\n\n• ASMI sees increased demand for ALD and epitaxy tools due to high-k metal gate processes and future 4F² or vertical channel transistor adoption.\n\n• NAND Demand Driven by Technology Transition\n\n• LRCX’s ALTUS Halo system advances atomic layer deposition (ALD) for 3D NAND inter-layer technology, enabling low-resistance interconnects for 256+ layers.\n\n• 3xx/4xx-layer NAND processes are expected to be adopted gradually over the next 3–5 years, driving a new round of equipment upgrades.\n\n• Tokyo Electron has secured new orders in NAND channel etch and low-temperature dry etch, expecting 2026 NAND investments to focus on brownfield upgrades and capacity enhancement.\n\n• Macro and Tariff Risk Mitigation\n\n• While direct tariff impact is limited, the industry is closely watching potential disruptions to end-customer purchasing patterns and CapEx decisions.\n\n• Major equipment vendors are strengthening supply chain diversification and regionalization to improve geopolitical risk resilience.\n\n• Test equipment vendors like Advantest and Teradyne state that HPC and AI markets are relatively insulated from tariffs but continue to monitor macro demand shifts closely.\n\n• China Market Developments\n• China contributed 27% of total sales in 1Q25, down from 38% in 2024. ASML remains most optimistic, forecasting over 25% of its sales from China.\n\n• LRCX, KLA, and AMAT all reported reduced spending in China on DRAM and mature logic nodes, though capacity build-out continues for 28nm and above.\n\n• ASMI reported strong margins in China, and Teradyne saw a slight increase in local customer share. Overall demand in the China market is expected to gradually stabilize.\n\n• Investment Recommendations and Catalysts\n\n• We maintain an Overweight rating on Samsung Electronics and SK Hynix, as valuation risks are low and fundamentals are expected to improve.\n\n• In the SPE (semiconductor production equipment) sector, we prefer Advantest and Tokyo Electron, which are set to benefit from equipment growth driven by HBM, DDR5, and 3D NAND technologies.\n\n• Medium-term catalysts include:\n1. Progress in 2026 equipment order negotiations among major manufacturers,\n2. Tariff resolution developments, and\n3. Expansion of AI accelerated computing demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007481442213208567, "ret_signed_m1d": 0.007481442213208567, "alpha_spy_m1d": -0.009658059476592484, "alpha_c_m1d": -0.00937358149702261, "bench_spy_m1d": 0.01713950168980105, "bench_c_m1d": 0.016855023710231176, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017954912531490175, "ret_signed_p1d": -0.017954912531490175, "alpha_spy_p1d": -0.018349347732904198, "alpha_c_p1d": -0.01522162775326208, "bench_spy_p1d": 0.0003944352014140229, "bench_c_p1d": -0.0027332847782280956, "ret_p1w": 0.037406583888659695, "ret_signed_p1w": 0.037406583888659695, "alpha_spy_p1w": 0.02905131512830894, "alpha_c_p1w": 0.03372083303720208, "bench_spy_p1w": 0.008355268760350754, "bench_c_p1w": 0.0036857508514576143, "ret_p1m": 0.22914882151017735, "ret_signed_p1m": 0.22914882151017735, "alpha_spy_p1m": 0.20413427909525939, "alpha_c_p1m": 0.14168459603700279, "bench_spy_p1m": 0.025014542414917962, "bench_c_p1m": 0.08746422547317456, "ret_p3m": 0.3365742724624756, "ret_signed_p3m": 0.3365742724624756, "alpha_spy_p3m": 0.2296160287879494, "alpha_c_p3m": 0.1071876578245694, "bench_spy_p3m": 0.10695824367452622, "bench_c_p3m": 0.22938661463790622, "ret_p6m": 2.062299254239053, "ret_signed_p6m": 2.062299254239053, "alpha_spy_p6m": 1.9026806277463104, "alpha_c_p6m": 1.6342957628353982, "bench_spy_p6m": 0.1596186264927424, "bench_c_p6m": 0.42800349140365457, "price_path": [0.0382, 0.0159, -0.0064, 0.006, -0.0065, -0.0514, -0.0514, -0.0718, -0.0369, -0.0404, -0.0404, -0.0628, -0.0633, -0.008, -0.004, 0.02, 0.0274, 0.0125, 0.0249, 0.0474, 0.0748, 0.0549, 0.0374, 0.0673, 0.0324, -0.006, -0.0489, -0.0175, -0.013, -0.0294, -0.0913, -0.1781, -0.1546, -0.1771, -0.0863, -0.0983, -0.1012, -0.0993, -0.1322, -0.1272, -0.1272, -0.1192, -0.1332, -0.0973, -0.1107, -0.0803, -0.0923, -0.0983, -0.1147, -0.1147, -0.0723, -0.0723, -0.0723, -0.0484, -0.0509, -0.0519, -0.0274, -0.01, 0.0274, 0.0, 0.02, -0.0055, 0.0075, 0.0, -0.018, -0.0025, 0.0125, 0.01, 0.0374, 0.0593, 0.0218, 0.0368, 0.0368, 0.0867, 0.1217, 0.1217, 0.1442, 0.1517, 0.1992, 0.1767, 0.1767, 0.2391, 0.2441, 0.2316, 0.2291, 0.2841, 0.2966, 0.3915, 0.429, 0.464, 0.419, 0.459, 0.4265, 0.394, 0.3915, 0.3516, 0.3541, 0.409, 0.404, 0.484, 0.4715, 0.499, 0.4915, 0.479, 0.3466, 0.3441, 0.3616, 0.3416, 0.3441, 0.3466, 0.3291, 0.3091, 0.3116, 0.3166, 0.3666, 0.2891, 0.2891, 0.3166, 0.2916, 0.3091, 0.2816, 0.3341, 0.3441, 0.389, 0.3815, 0.3815, 0.3366, 0.3141, 0.2766, 0.2242, 0.2541, 0.2966, 0.3066, 0.2991, 0.3435, 0.346, 0.281, 0.3035, 0.3135, 0.3285, 0.3685, 0.386, 0.4411, 0.5211, 0.5362, 0.6437, 0.6562, 0.7413, 0.6688, 0.7788, 0.7788, 0.7563, 0.8064, 0.7888, 0.7838, 0.6838, 0.7463, 0.7388, 0.8014, 0.979, 0.979, 0.979, 0.979, 0.979, 0.979, 1.1416, 1.0766, 1.059, 1.1141, 1.2642, 1.3292, 1.4293, 1.3968, 1.4093, 1.3943, 1.5519, 1.677, 1.607, 1.7921, 1.8421, 1.7971, 2.1023, 1.9322, 1.8972, 1.9672, 1.9022, 2.0323, 2.0973, 2.0873, 2.0623]}
{"unit_id": "orig:1925327380933153174", "ticker_idx": 2, "ts": "2025-05-21T23:06:33+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we prefer Advantest and Tokyo Electron, which are set to benefit from equipment growth driven by HBM, DDR5, and 3D NAND technologies.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan Overweight Samsung, SK Hynix; prefers Advantest, Tokyo Electron, Hanmi on HBM/DDR5/3D NAND equipment cycle into 2026.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Equipment / Foundry: Marginal Improvement in Industry Sentiment\n JP Morgan (J. Kwon, May 20, 2025)\n\n1QCY25 results and outlook indicate that as chip stacking and process node evolution continues, demand for testing high-performance memory such as HBM (High Bandwidth Memory) and DDR5 is rising sharply. This is driving stable to mid-single-digit growth in both front-end and back-end testing and wafer fabrication equipment (WFE) markets. While the WFE market is projected to reach approximately $100 billion for the full year 2025 with a broadly optimistic outlook for 2026, macroeconomic and geopolitical uncertainties—especially tariffs—in 2H25 remain key concerns. We prefer Samsung Electronics (SEC), SK Hynix, Tokyo Electron, Advantest, and Hanmi to capture long-term structural opportunities.\n\n• WFE Market: Solid Growth Momentum\n\n• Leading manufacturers forecast the 2025 WFE market at $99–110 billion, indicating stable to mid-single-digit growth.\n\n• Leading equipment vendors (LAM, KLA) are optimistic about continued adoption by leading foundries and logic fabs. Tokyo Electron expects double-digit growth in 2026.\n\n• DRAM and NAND together are projected to account for around 35% of WFE demand, with DRAM growing to 20% due to HBM and DDR5 transitions, and NAND reaching 10%.\n\n• HBM and Substrate Testing Ramping Up\n\n• Front-end equipment suppliers like KLA and TEL classify HBM as a testing-intensive technology, benefiting from increased stacking layers and stricter quality control.\n\n• Back-end testing equipment vendors (Advantest, Hanmi) report a significant increase in demand for HBM-related tools, with orders expected to continue releasing through 2H25.\n\n• Applied Materials forecasts over 40% YoY revenue growth from HBM and DDR5 customers in 2025, reflecting surging demand for high-value-added memory process equipment.\n\n• DRAM Technology Evolution Drives Equipment Refresh\n\n• ASML is promoting broader use of EUV to simplify DRAM multi-patterning to single exposure, improving efficiency and output.\n\n• LAM Research (LRCX) has made breakthroughs in etch technology, with its Akara system adopted for critical applications by major DRAM makers.\n\n• ASMI sees increased demand for ALD and epitaxy tools due to high-k metal gate processes and future 4F² or vertical channel transistor adoption.\n\n• NAND Demand Driven by Technology Transition\n\n• LRCX’s ALTUS Halo system advances atomic layer deposition (ALD) for 3D NAND inter-layer technology, enabling low-resistance interconnects for 256+ layers.\n\n• 3xx/4xx-layer NAND processes are expected to be adopted gradually over the next 3–5 years, driving a new round of equipment upgrades.\n\n• Tokyo Electron has secured new orders in NAND channel etch and low-temperature dry etch, expecting 2026 NAND investments to focus on brownfield upgrades and capacity enhancement.\n\n• Macro and Tariff Risk Mitigation\n\n• While direct tariff impact is limited, the industry is closely watching potential disruptions to end-customer purchasing patterns and CapEx decisions.\n\n• Major equipment vendors are strengthening supply chain diversification and regionalization to improve geopolitical risk resilience.\n\n• Test equipment vendors like Advantest and Teradyne state that HPC and AI markets are relatively insulated from tariffs but continue to monitor macro demand shifts closely.\n\n• China Market Developments\n• China contributed 27% of total sales in 1Q25, down from 38% in 2024. ASML remains most optimistic, forecasting over 25% of its sales from China.\n\n• LRCX, KLA, and AMAT all reported reduced spending in China on DRAM and mature logic nodes, though capacity build-out continues for 28nm and above.\n\n• ASMI reported strong margins in China, and Teradyne saw a slight increase in local customer share. Overall demand in the China market is expected to gradually stabilize.\n\n• Investment Recommendations and Catalysts\n\n• We maintain an Overweight rating on Samsung Electronics and SK Hynix, as valuation risks are low and fundamentals are expected to improve.\n\n• In the SPE (semiconductor production equipment) sector, we prefer Advantest and Tokyo Electron, which are set to benefit from equipment growth driven by HBM, DDR5, and 3D NAND technologies.\n\n• Medium-term catalysts include:\n1. Progress in 2026 equipment order negotiations among major manufacturers,\n2. Tariff resolution developments, and\n3. Expansion of AI accelerated computing demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03780306065745509, "ret_signed_m1d": 0.03780306065745509, "alpha_spy_m1d": 0.02066355896765404, "alpha_c_m1d": 0.020948036947223914, "bench_spy_m1d": 0.01713950168980105, "bench_c_m1d": 0.016855023710231176, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0322397601436879, "ret_signed_p1d": -0.0322397601436879, "alpha_spy_p1d": -0.032634195345101924, "alpha_c_p1d": -0.029506475365459806, "bench_spy_p1d": 0.0003944352014140229, "bench_c_p1d": -0.0027332847782280956, "ret_p1w": 0.0349500179372253, "ret_signed_p1w": 0.0349500179372253, "alpha_spy_p1w": 0.026594749176874544, "alpha_c_p1w": 0.03126426708576768, "bench_spy_p1w": 0.008355268760350754, "bench_c_p1w": 0.0036857508514576143, "ret_p1m": 0.3453636098880972, "ret_signed_p1m": 0.3453636098880972, "alpha_spy_p1m": 0.32034906747317926, "alpha_c_p1m": 0.25789938441492266, "bench_spy_p1m": 0.025014542414917962, "bench_c_p1m": 0.08746422547317456, "ret_p3m": 0.6348073168783315, "ret_signed_p3m": 0.6348073168783315, "alpha_spy_p3m": 0.5278490732038053, "alpha_c_p3m": 0.40542070224042526, "bench_spy_p3m": 0.10695824367452622, "bench_c_p3m": 0.22938661463790622, "ret_p6m": 1.9545660879299351, "ret_signed_p6m": 1.9545660879299351, "alpha_spy_p6m": 1.7949474614371927, "alpha_c_p6m": 1.5265625965262806, "bench_spy_p6m": 0.1596186264927424, "bench_c_p6m": 0.42800349140365457, "price_path": [0.3699, 0.3699, 0.2803, 0.2903, 0.2657, 0.1546, 0.1563, 0.1062, 0.1029, 0.0937, 0.0681, 0.108, 0.1268, 0.0745, 0.1139, 0.1729, 0.2049, 0.2382, 0.1783, 0.1783, 0.152, 0.1388, 0.0999, 0.1069, 0.0245, -0.0003, -0.0767, -0.1037, -0.0809, -0.1224, -0.1934, -0.2819, -0.197, -0.2596, -0.1579, -0.1966, -0.1571, -0.1722, -0.2264, -0.1994, -0.2175, -0.2245, -0.231, -0.2071, -0.182, -0.1441, -0.1854, -0.1854, -0.1636, -0.106, -0.096, -0.096, -0.096, -0.0942, -0.0603, -0.0449, 0.0414, 0.0197, 0.0693, 0.0575, 0.0375, 0.008, 0.0378, 0.0, -0.0322, -0.0278, 0.013, 0.0187, 0.035, 0.0903, 0.0509, 0.0114, 0.0381, 0.0581, 0.1064, 0.1311, 0.186, 0.1846, 0.1899, 0.1845, 0.2131, 0.33, 0.3623, 0.3785, 0.3454, 0.4024, 0.3852, 0.407, 0.4536, 0.5264, 0.51, 0.52, 0.5357, 0.4872, 0.5021, 0.5228, 0.5243, 0.5628, 0.5663, 0.6041, 0.6155, 0.6427, 0.6719, 0.6826, 0.6698, 0.5956, 0.5956, 0.6184, 0.6277, 0.617, 0.6398, 0.4929, 0.4772, 0.4615, 0.4765, 0.4572, 0.4337, 0.433, 0.4379, 0.4422, 0.4957, 0.4957, 0.5906, 0.6769, 0.5991, 0.6113, 0.6348, 0.6277, 0.5357, 0.5613, 0.5357, 0.5521, 0.5556, 0.6255, 0.6491, 0.6655, 0.5335, 0.5214, 0.5257, 0.5977, 0.632, 0.7033, 0.8131, 0.8723, 0.9544, 0.99, 0.99, 1.0157, 0.975, 1.0728, 1.1427, 1.2111, 1.2111, 1.1976, 1.1441, 1.0642, 1.1648, 1.0941, 1.0976, 1.1505, 1.2427, 1.5572, 1.5736, 1.5679, 1.5736, 1.5493, 1.5493, 1.43, 1.4821, 1.5, 1.4114, 1.5029, 1.4872, 1.4443, 1.3535, 1.4414, 1.6008, 1.5901, 2.1618, 2.1833, 2.3069, 2.3069, 2.1132, 1.9281, 2.0203, 1.8531, 1.9617, 1.8409, 1.8345, 1.9546]}
{"unit_id": "orig:1925327380933153174", "ticker_idx": 3, "ts": "2025-05-21T23:06:33+00:00", "symbol": "8035.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we prefer Advantest and Tokyo Electron, which are set to benefit from equipment growth driven by HBM, DDR5, and 3D NAND technologies.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JP Morgan Overweight Samsung, SK Hynix; prefers Advantest, Tokyo Electron, Hanmi on HBM/DDR5/3D NAND equipment cycle into 2026.", "resolved_tickers": ["8035.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Equipment / Foundry: Marginal Improvement in Industry Sentiment\n JP Morgan (J. Kwon, May 20, 2025)\n\n1QCY25 results and outlook indicate that as chip stacking and process node evolution continues, demand for testing high-performance memory such as HBM (High Bandwidth Memory) and DDR5 is rising sharply. This is driving stable to mid-single-digit growth in both front-end and back-end testing and wafer fabrication equipment (WFE) markets. While the WFE market is projected to reach approximately $100 billion for the full year 2025 with a broadly optimistic outlook for 2026, macroeconomic and geopolitical uncertainties—especially tariffs—in 2H25 remain key concerns. We prefer Samsung Electronics (SEC), SK Hynix, Tokyo Electron, Advantest, and Hanmi to capture long-term structural opportunities.\n\n• WFE Market: Solid Growth Momentum\n\n• Leading manufacturers forecast the 2025 WFE market at $99–110 billion, indicating stable to mid-single-digit growth.\n\n• Leading equipment vendors (LAM, KLA) are optimistic about continued adoption by leading foundries and logic fabs. Tokyo Electron expects double-digit growth in 2026.\n\n• DRAM and NAND together are projected to account for around 35% of WFE demand, with DRAM growing to 20% due to HBM and DDR5 transitions, and NAND reaching 10%.\n\n• HBM and Substrate Testing Ramping Up\n\n• Front-end equipment suppliers like KLA and TEL classify HBM as a testing-intensive technology, benefiting from increased stacking layers and stricter quality control.\n\n• Back-end testing equipment vendors (Advantest, Hanmi) report a significant increase in demand for HBM-related tools, with orders expected to continue releasing through 2H25.\n\n• Applied Materials forecasts over 40% YoY revenue growth from HBM and DDR5 customers in 2025, reflecting surging demand for high-value-added memory process equipment.\n\n• DRAM Technology Evolution Drives Equipment Refresh\n\n• ASML is promoting broader use of EUV to simplify DRAM multi-patterning to single exposure, improving efficiency and output.\n\n• LAM Research (LRCX) has made breakthroughs in etch technology, with its Akara system adopted for critical applications by major DRAM makers.\n\n• ASMI sees increased demand for ALD and epitaxy tools due to high-k metal gate processes and future 4F² or vertical channel transistor adoption.\n\n• NAND Demand Driven by Technology Transition\n\n• LRCX’s ALTUS Halo system advances atomic layer deposition (ALD) for 3D NAND inter-layer technology, enabling low-resistance interconnects for 256+ layers.\n\n• 3xx/4xx-layer NAND processes are expected to be adopted gradually over the next 3–5 years, driving a new round of equipment upgrades.\n\n• Tokyo Electron has secured new orders in NAND channel etch and low-temperature dry etch, expecting 2026 NAND investments to focus on brownfield upgrades and capacity enhancement.\n\n• Macro and Tariff Risk Mitigation\n\n• While direct tariff impact is limited, the industry is closely watching potential disruptions to end-customer purchasing patterns and CapEx decisions.\n\n• Major equipment vendors are strengthening supply chain diversification and regionalization to improve geopolitical risk resilience.\n\n• Test equipment vendors like Advantest and Teradyne state that HPC and AI markets are relatively insulated from tariffs but continue to monitor macro demand shifts closely.\n\n• China Market Developments\n• China contributed 27% of total sales in 1Q25, down from 38% in 2024. ASML remains most optimistic, forecasting over 25% of its sales from China.\n\n• LRCX, KLA, and AMAT all reported reduced spending in China on DRAM and mature logic nodes, though capacity build-out continues for 28nm and above.\n\n• ASMI reported strong margins in China, and Teradyne saw a slight increase in local customer share. Overall demand in the China market is expected to gradually stabilize.\n\n• Investment Recommendations and Catalysts\n\n• We maintain an Overweight rating on Samsung Electronics and SK Hynix, as valuation risks are low and fundamentals are expected to improve.\n\n• In the SPE (semiconductor production equipment) sector, we prefer Advantest and Tokyo Electron, which are set to benefit from equipment growth driven by HBM, DDR5, and 3D NAND technologies.\n\n• Medium-term catalysts include:\n1. Progress in 2026 equipment order negotiations among major manufacturers,\n2. Tariff resolution developments, and\n3. Expansion of AI accelerated computing demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00430207151656381, "ret_signed_m1d": 0.00430207151656381, "alpha_spy_m1d": -0.012837430173237241, "alpha_c_m1d": -0.012552952193667366, "bench_spy_m1d": 0.01713950168980105, "bench_c_m1d": 0.016855023710231176, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024306250064781154, "ret_signed_p1d": -0.024306250064781154, "alpha_spy_p1d": -0.024700685266195177, "alpha_c_p1d": -0.02157296528655306, "bench_spy_p1d": 0.0003944352014140229, "bench_c_p1d": -0.0027332847782280956, "ret_p1w": -0.0034415201554989228, "ret_signed_p1w": -0.0034415201554989228, "alpha_spy_p1w": -0.011796788915849676, "alpha_c_p1w": -0.007127271006956537, "bench_spy_p1w": 0.008355268760350754, "bench_c_p1w": 0.0036857508514576143, "ret_p1m": 0.03248011741737633, "ret_signed_p1m": 0.03248011741737633, "alpha_spy_p1m": 0.007465575002458369, "alpha_c_p1m": -0.05498410805579823, "bench_spy_p1m": 0.025014542414917962, "bench_c_p1m": 0.08746422547317456, "ret_p3m": -0.0946439458035957, "ret_signed_p3m": -0.0946439458035957, "alpha_spy_p3m": -0.20160218947812192, "alpha_c_p3m": -0.3240305604415019, "bench_spy_p3m": 0.10695824367452622, "bench_c_p3m": 0.22938661463790622, "ret_p6m": 0.4734626638835948, "ret_signed_p6m": 0.4734626638835948, "alpha_spy_p6m": 0.3138440373908524, "alpha_c_p6m": 0.045459172479940246, "bench_spy_p6m": 0.1596186264927424, "bench_c_p6m": 0.42800349140365457, "price_path": [0.079, 0.079, 0.0258, -0.0274, -0.0185, -0.0621, -0.0513, -0.0484, -0.066, -0.0768, -0.105, -0.0897, -0.0941, -0.0956, -0.0963, -0.0895, -0.0694, -0.0535, -0.059, -0.059, -0.052, -0.0594, -0.0543, -0.039, -0.0552, -0.074, -0.1349, -0.1288, -0.1121, -0.1448, -0.1843, -0.2661, -0.202, -0.2506, -0.154, -0.1551, -0.1437, -0.14, -0.1529, -0.1478, -0.1605, -0.1641, -0.1749, -0.1547, -0.123, -0.0858, -0.088, -0.088, -0.0869, -0.0875, -0.0906, -0.0906, -0.0906, -0.0576, -0.0383, -0.0297, -0.0153, 0.0325, 0.0422, 0.0381, 0.0095, -0.0077, 0.0043, 0.0, -0.0243, -0.0211, 0.0047, -0.0022, -0.0034, 0.0389, -0.0105, -0.0275, -0.0364, -0.032, 0.0045, 0.0174, 0.0262, 0.0213, 0.0703, 0.0755, 0.0239, 0.0331, 0.0635, 0.0641, 0.0325, 0.0318, 0.0198, 0.057, 0.0914, 0.1349, 0.1837, 0.1908, 0.1646, 0.1437, 0.1637, 0.1691, 0.1669, 0.1729, 0.1691, 0.1512, 0.1615, 0.1394, 0.1792, 0.1998, 0.1947, 0.1981, 0.1981, 0.1824, 0.2054, 0.2054, 0.2028, 0.1757, 0.1615, 0.1729, 0.1757, -0.0361, -0.0327, -0.0366, -0.0729, -0.0957, -0.0841, -0.0841, -0.0742, -0.0669, -0.0856, -0.0749, -0.0946, -0.0871, -0.0998, -0.1215, -0.1355, -0.1299, -0.1291, -0.1252, -0.1073, -0.111, -0.1273, -0.1252, -0.142, -0.1359, -0.1293, -0.1138, -0.0957, -0.0901, -0.0792, -0.0284, -0.0284, -0.0099, 0.0432, 0.0951, 0.1017, 0.1443, 0.1443, 0.1725, 0.1925, 0.142, 0.1517, 0.1454, 0.1215, 0.2099, 0.238, 0.3283, 0.3075, 0.2712, 0.2905, 0.2723, 0.2723, 0.2358, 0.2632, 0.3144, 0.307, 0.3648, 0.3344, 0.3353, 0.2918, 0.3153, 0.3305, 0.3666, 0.4118, 0.4335, 0.4852, 0.4852, 0.5117, 0.45, 0.4448, 0.4252, 0.4861, 0.4856, 0.4626, 0.4735]}
{"unit_id": "quote:1929328913349800427", "ticker_idx": 0, "ts": "2025-06-02T00:07:12+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC is optimistic about AI accelerators driving its revenue growth, forecasting a compound annual growth rate (CAGR) in the mid-40% range over the next five years starting in 2024", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares TSMC's own bullish forecast: mid-40% CAGR over 5 years driven by AI accelerators and broad 2nm adoption.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Taiwanese media reports that TSMC is optimistic about AI accelerators driving its revenue growth, forecasting a compound annual growth rate (CAGR) in the mid-40% range over the next five years starting in 2024. It is also expected that the following chips will be manufactured using TSMC’s 2nm process:\n\n1. Apple M6 Mac chip, A20 (2026)\n2. Qualcomm Snapdragon 8 Elite 3 (2026)\n3. MediaTek Dimensity 9600 (2026)\n4. AMD EPYC server processor (2026)\n5. Fujitsu TSMC 2nm AI CPU\n6. Microsoft MAIA 300 (2026)\n7. AWS Trainium 4 (2027)\n8. Google TPU v8p/e (2027)", "reply_text": null, "parent_text": null, "quoted_text": "Rumor: The price per wafer for TSMC’s 2nm process is $30,000, and the A14 process could reach as high as $45,000.\n\n$TSM\nhttps://t.co/nsPkXwU35h", "parent_author": null, "ret_m1d": -0.00780119143883562, "ret_signed_m1d": -0.00780119143883562, "alpha_spy_m1d": -0.00219984214880542, "alpha_c_m1d": 0.006789903156856569, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014216887472481288, "ret_signed_p1d": 0.014216887472481288, "alpha_spy_p1d": 0.008514297684293792, "alpha_c_p1d": -0.00834781013373287, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": 0.06241024507055015, "ret_signed_p1w": 0.06241024507055015, "alpha_spy_p1w": 0.05065082790751507, "alpha_c_p1w": 0.0062656844481985274, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.1573678584252829, "ret_signed_p1m": 0.1573678584252829, "alpha_spy_p1m": 0.11220928424563037, "alpha_c_p1m": 0.023623602443586522, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.2273724418581511, "ret_signed_p3m": 0.2273724418581511, "alpha_spy_p3m": 0.12930039632825752, "alpha_c_p3m": -0.0010697367805272417, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 0.47106502853406984, "ret_signed_p6m": 0.47106502853406984, "alpha_spy_p6m": 0.3256609669777639, "alpha_c_p6m": 0.07357219898266809, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [-0.0579, -0.101, -0.0946, -0.1276, -0.1259, -0.0942, -0.1227, -0.11, -0.099, -0.1114, -0.1082, -0.0891, -0.0929, -0.0702, -0.0715, -0.1095, -0.1365, -0.1519, -0.148, -0.1345, -0.1255, -0.1923, -0.2466, -0.2498, -0.2744, -0.1852, -0.2243, -0.1938, -0.2002, -0.1925, -0.2216, -0.2212, -0.2212, -0.2411, -0.223, -0.1901, -0.1574, -0.1526, -0.1614, -0.1558, -0.1445, -0.1135, -0.0799, -0.0946, -0.1158, -0.1042, -0.1007, -0.094, -0.0403, -0.0044, -0.0004, -0.0032, -0.0032, -0.0069, -0.0071, -0.0158, 0.0069, -0.0147, -0.0147, 0.0146, 0.0067, 0.0119, -0.0078, 0.0, 0.0142, 0.0388, 0.0436, 0.0531, 0.0624, 0.0904, 0.0989, 0.1097, 0.0874, 0.111, 0.1018, 0.0998, 0.0998, 0.0792, 0.0834, 0.1337, 0.1474, 0.1539, 0.1774, 0.1667, 0.1574, 0.2033, 0.2095, 0.2095, 0.1805, 0.1737, 0.1943, 0.1835, 0.1868, 0.1779, 0.2206, 0.2237, 0.2651, 0.2383, 0.2304, 0.2085, 0.238, 0.2445, 0.2651, 0.2504, 0.2431, 0.2513, 0.2446, 0.2116, 0.2311, 0.1975, 0.1918, 0.2498, 0.2457, 0.247, 0.2584, 0.2437, 0.2414, 0.2305, 0.2435, 0.1987, 0.1776, 0.171, 0.2002, 0.2136, 0.2297, 0.2326, 0.2274, 0.1893, 0.1893, 0.1765, 0.1919, 0.2116, 0.2538, 0.2733, 0.2925, 0.3416, 0.3337, 0.3359, 0.3464, 0.3542, 0.358, 0.3882, 0.3687, 0.4088, 0.4609, 0.4506, 0.4296, 0.4126, 0.4119, 0.4432, 0.4906, 0.4888, 0.5099, 0.5626, 0.5194, 0.5736, 0.5496, 0.4503, 0.5652, 0.5293, 0.5746, 0.5494, 0.5248, 0.5383, 0.5219, 0.4928, 0.5023, 0.5242, 0.5412, 0.5581, 0.5765, 0.5669, 0.5525, 0.5753, 0.5195, 0.5174, 0.4946, 0.4805, 0.5258, 0.5046, 0.5018, 0.4583, 0.4718, 0.4573, 0.4361, 0.4591, 0.434, 0.4214, 0.4709, 0.4711]}
{"unit_id": "thread:1939473288385962049", "ticker_idx": 0, "ts": "2025-06-29T23:57:20+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "institutional investors anticipate that TSMC will raise wafer prices again by 3% to 5% next year to reflect construction costs and strong AI demand, while US fab quotations will see adjustments exceeding 10%", "tweet_type": "analysis", "timeline": "months", "summary": "TSMC Arizona Fab 2 accelerating; wafer price hikes of 3-5% expected next year plus 10%+ on US fab quotes; supply chain benefits flow through.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: 3nm Process on Track! TSMC's US Fab 2 to Start Early, Specialty Gas and Chemical Firms See Upside\n\nTo meet growing customer demand for US manufacturing, TSMC is accelerating construction at its Arizona fabs! Supply chain sources reveal that Fab 2 (P2), planned for 3nm advanced process configurations, began construction in April 2025. The overall timeline currently shows signs of acceleration, with equipment installation anticipated in Q3 2026 and a push for online production in 2027. The fab construction speed is being compressed to approximately two years, operating at full throttle to meet customer demand.\n\nWith TSMC's Arizona Fab 2 commencing construction in April, supply chain analysis suggests that TSMC is actively cooperating to meet customer demand and address US government tariffs, leading to an accelerated overall timeline. Equipment move-in could happen as early as September next year, with the first wafer output expected in 2027. Equipment vendors admit that after fab construction is complete, internal facility operations still require about two years for adjustment and configuration. In terms of fab construction speed, TSMC's actions are already very fast.\n\nSupply chain sources indicate that TSMC's acceleration benefits Taiwanese facility engineering firms like United Integrated Services (UIS) and Marketech International Corp. (MIC), as they have accumulated learning experience from constructing Fab 1 (P1), which is expected to improve long-term profitability.\nSupply chain vendors also revealed that they will begin entering the fab after mid-year to cooperate with TSMC's US P2 project. Specialty gas and chemical segments are also expected to receive North American orders sequentially, with institutional investors optimistic about companies like Topco Scientific and Sunshine Chem.\nHowever, advanced packaging will still rely on Taiwanese capacity. Institutional investors judge that despite TSMC's declaration to invest in two advanced packaging fabs in the US, related evaluations, including human resources and factory permits, require time. Arizona AP1 (the first advanced packaging fab) is expected to break ground at the earliest in Q3 next year and will primarily focus on SoIC (System-on-Integrated-Chips). In other words, CoWoS will still need to be shipped back to Taiwan for production.\n\nFab construction cannot be completed overnight. Facility engineering firms estimate that due to the massive scale planned for the US fabs, only partial progress can be made to the perfected stage, with most fabs unable to fully come online and begin production before 2029. With international conditions and industry cycles continuously changing during the construction period, institutional investors anticipate that TSMC will raise wafer prices again by 3% to 5% next year to reflect construction costs and strong AI demand, while US fab quotations will see adjustments exceeding 10%.\nTaiwan's importance remains irreplaceable. From TSMC's strategic layout, nine new fabs and 11 production lines are simultaneously under construction this year, especially the upcoming 2nm production and the highly demanded advanced packaging capacity, all of which are progressing concurrently.\nRelevant industry sources reveal that TSMC's Kaohsiung 2nm Fab F22 Fab 2 will move in equipment in Q3 2025, followed by the completion of Fab 3 in Q1 2026. Advanced process research and development continue, and judging by the fab construction timelines, production will first take place in Taiwan.\n\n$TSM\n\n---\n\nhttps://t.co/LjOWDLZ7UR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00918365939265664, "ret_signed_m1d": 0.00918365939265664, "alpha_spy_m1d": 0.013942213457711072, "alpha_c_m1d": 0.010833056927868845, "bench_spy_m1d": -0.004758554065054432, "bench_c_m1d": -0.0016493975352122048, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007991564135397589, "ret_signed_p1d": -0.007991564135397589, "alpha_spy_p1d": -0.007667884827172733, "alpha_c_p1d": 0.0029092126269995022, "bench_spy_p1d": -0.0003236793082248557, "bench_c_p1d": -0.01090077676239709, "ret_p1w": 0.01183267690623624, "ret_signed_p1w": 0.01183267690623624, "alpha_spy_p1w": 0.00725235069049357, "alpha_c_p1w": 0.008677131122611703, "bench_spy_p1w": 0.004580326215742669, "bench_c_p1w": 0.003155545783624536, "ret_p1m": 0.06552167894744954, "ret_signed_p1m": 0.06552167894744954, "alpha_spy_p1m": 0.03734344751166341, "alpha_c_p1m": 0.015249192459260152, "bench_spy_p1m": 0.028178231435786127, "bench_c_p1m": 0.05027248648818938, "ret_p3m": 0.22536451428302318, "ret_signed_p3m": 0.22536451428302318, "alpha_spy_p3m": 0.15734742712382666, "alpha_c_p3m": 0.07411664834064546, "bench_spy_p3m": 0.06801708715919652, "bench_c_p3m": 0.1512478659423777, "ret_p6m": 0.3187823231477893, "ret_signed_p6m": 0.3187823231477893, "alpha_spy_p6m": 0.19892185293908748, "alpha_c_p6m": 0.012521495065923594, "bench_spy_p6m": 0.11986047020870183, "bench_c_p6m": 0.3062608280818657, "price_path": [-0.2504, -0.3077, -0.3542, -0.357, -0.3781, -0.3016, -0.3352, -0.309, -0.3144, -0.3079, -0.3328, -0.3325, -0.3325, -0.3495, -0.334, -0.3058, -0.2778, -0.2737, -0.2812, -0.2764, -0.2667, -0.2402, -0.2113, -0.224, -0.2421, -0.2322, -0.2292, -0.2235, -0.1775, -0.1466, -0.1432, -0.1456, -0.1456, -0.1488, -0.149, -0.1564, -0.1369, -0.1555, -0.1555, -0.1304, -0.1372, -0.1327, -0.1496, -0.1429, -0.1307, -0.1096, -0.1055, -0.0974, -0.0894, -0.0654, -0.0581, -0.0488, -0.0679, -0.0477, -0.0556, -0.0574, -0.0574, -0.075, -0.0714, -0.0283, -0.0166, -0.0109, 0.0092, 0.0, -0.008, 0.0314, 0.0367, 0.0367, 0.0118, 0.006, 0.0236, 0.0144, 0.0173, 0.0096, 0.0462, 0.0489, 0.0844, 0.0614, 0.0546, 0.0358, 0.0611, 0.0667, 0.0844, 0.0718, 0.0655, 0.0725, 0.0668, 0.0385, 0.0552, 0.0264, 0.0215, 0.0712, 0.0677, 0.0689, 0.0786, 0.066, 0.0641, 0.0547, 0.0659, 0.0274, 0.0093, 0.0037, 0.0287, 0.0402, 0.054, 0.0565, 0.052, 0.0193, 0.0193, 0.0084, 0.0216, 0.0385, 0.0747, 0.0914, 0.1079, 0.1499, 0.1431, 0.145, 0.154, 0.1607, 0.1639, 0.1898, 0.1731, 0.2075, 0.2522, 0.2433, 0.2254, 0.2107, 0.2102, 0.237, 0.2777, 0.2761, 0.2941, 0.3394, 0.3023, 0.3488, 0.3282, 0.2431, 0.3415, 0.3108, 0.3496, 0.328, 0.3069, 0.3186, 0.3044, 0.2795, 0.2877, 0.3064, 0.321, 0.3355, 0.3513, 0.343, 0.3306, 0.3503, 0.3024, 0.3006, 0.2811, 0.2689, 0.3078, 0.2896, 0.2872, 0.2499, 0.2615, 0.2491, 0.2309, 0.2507, 0.2291, 0.2183, 0.2607, 0.2609, 0.2843, 0.2843, 0.2911, 0.2742, 0.2937, 0.3086, 0.2974, 0.3054, 0.337, 0.3438, 0.3737, 0.3539, 0.297, 0.2779, 0.274, 0.23, 0.2643, 0.2833, 0.3025, 0.3188]}
{"unit_id": "orig:1921132141221486893", "ticker_idx": 0, "ts": "2025-05-10T09:16:10+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "He expressed confidence in ASML achieving 1000W source power from the current 600W", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish ASML on continued EUV intensity ramp and 1000W source power milestone; bullish ASMI as sole vertically integrated ALD leader.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Expert Call with Former Intel Lithography and Foundry Executive\n\nYesterday, we hosted an expert call with Klaus Shuegraf, currently VP at Photonic and formerly Director of Lithography Sourcing and VP of Foundry Strategy & Planning at Intel.\n\n1. EUV Source Power and Increasing Lithographic Intensity Expected to Continue\n\nShuegraf acknowledged that the transition from N3 to N2 nodes does not significantly shrink feature sizes due to the Gate-All-Around (GAA) structure. However, he projected that lithographic intensity will continue to increase for nodes beyond that. He expressed confidence in ASML achieving 1000W source power from the current 600W, but noted that reaching 1500W would approach technical limitations.\n\n2. 18A Node Outlook and High-NA Adoption Timing\n\nIn line with industry trends, Shuegraf stated that Intel must first prove the viability of its 18A node through its own products before attracting potential customers such as AVGO (Broadcom), NVDA (Nvidia), and AAPL (Apple M-series). He predicted that 2026 would be the mainstream ramp-up period for 18A, which would be about one year behind TSMC’s 2nm node production schedule. Regarding High-NA EUV, he noted that TSMC’s full adoption would likely be in the late 2020s, as it will take time for ASML to scale up production capacity.\n\n3. Potential Expansion of EUV Adoption in DRAM\n\nShuegraf suggested that if leading HBM suppliers adopt EUV, it could drive broader EUV adoption across the DRAM industry. This is due to trends such as AI PCs and advanced packaging making DRAM a more specialized, rather than purely commoditized, product segment. While some customers may be able to maintain current lithographic intensity levels through 4F2 architecture, others are expected to adopt additional litho layers alongside 4F2.\n\n4. Additional Notes on China and ALD\n\n(A) Regarding China’s development of EUV using DPP (Discharge Produced Plasma), Shuegraf acknowledged that for AI use cases, high wafer throughput is not always essential. However, he commented that DPP is economically inefficient and presents challenges in supply availability.\n\n(B) On the topic of ALD (Atomic Layer Deposition), he emphasized that ASMI remains the leading player due to its position as the only vertically integrated vendor with an in-house chemical supply chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06012902404463538, "ret_signed_m1d": -0.06012902404463538, "alpha_spy_m1d": -0.028138712873499117, "alpha_c_m1d": -0.001117141942292621, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025191146732793213, "ret_signed_p1d": 0.025191146732793213, "alpha_spy_p1d": 0.01858720641562317, "alpha_c_p1d": -0.00913761303010241, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": -0.006747701073078538, "ret_signed_p1w": -0.006747701073078538, "alpha_spy_p1w": -0.02709100134682374, "alpha_c_p1w": -0.042971934001199585, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.02639070318381087, "ret_signed_p1m": 0.02639070318381087, "alpha_spy_p1m": -0.008069467264435026, "alpha_c_p1m": -0.07748025919240065, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": -0.08184198290749678, "ret_signed_p3m": -0.08184198290749678, "alpha_spy_p3m": -0.1695433777256845, "alpha_c_p3m": -0.308075022049211, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.38260481704527005, "ret_signed_p6m": 0.38260481704527005, "alpha_spy_p6m": 0.2177241319822958, "alpha_c_p6m": -0.10326356013764282, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0739, 0.103, 0.0945, 0.0879, 0.0706, 0.0725, 0.0612, 0.0615, 0.0665, 0.0431, 0.068, 0.045, 0.0144, 0.0319, -0.0127, 0.0056, 0.0015, -0.0045, -0.0448, -0.0553, -0.0351, -0.0397, -0.0199, -0.0115, -0.0046, 0.0062, 0.0053, -0.0124, 0.0084, 0.0085, -0.0176, -0.0396, -0.0648, -0.0941, -0.0737, -0.0787, -0.135, -0.1568, -0.1779, -0.1423, -0.1714, -0.1366, -0.137, -0.1181, -0.095, -0.142, -0.1566, -0.1566, -0.1566, -0.1613, -0.1317, -0.1155, -0.117, -0.1236, -0.1228, -0.1266, -0.1266, -0.087, -0.0901, -0.0981, -0.0933, -0.0546, -0.0601, 0.0, 0.0252, 0.0295, 0.0229, 0.0022, -0.0067, -0.0103, -0.0033, -0.0151, -0.0364, -0.0193, -0.0012, -0.0091, -0.0021, -0.0195, -0.031, -0.0207, -0.0217, -0.0118, 0.001, 0.0198, 0.0264, 0.0349, 0.0145, -0.0034, 0.0048, -0.0013, -0.0028, -0.0207, -0.0247, 0.0027, 0.0367, 0.0414, 0.0157, 0.0234, 0.016, 0.0019, 0.016, 0.0153, -0.0112, 0.007, 0.018, 0.0193, 0.0346, 0.0255, 0.0306, 0.0588, -0.0616, -0.025, -0.0501, -0.0642, -0.0967, -0.0904, -0.0765, -0.093, -0.0554, -0.0633, -0.0566, -0.0782, -0.1054, -0.0943, -0.1023, -0.1086, -0.0818, -0.0725, -0.0573, -0.047, -0.0369, -0.0334, -0.0428, -0.0397, -0.0343, -0.0405, -0.0461, -0.0285, -0.0277, -0.0219, -0.0074, -0.0163, -0.0429, -0.0437, -0.0721, -0.0599, -0.0264, -0.0074, 0.0172, 0.0272, 0.0226, 0.0314, 0.0371, 0.0966, 0.1019, 0.108, 0.1937, 0.1933, 0.219, 0.2325, 0.2148, 0.2204, 0.222, 0.2411, 0.245, 0.265, 0.3194, 0.3232, 0.349, 0.3152, 0.2806, 0.2776, 0.2237, 0.2689, 0.2728, 0.3125, 0.3191, 0.3166, 0.3516, 0.3378, 0.3082, 0.3346, 0.3447, 0.3653, 0.3666, 0.3847, 0.412, 0.3828, 0.3954, 0.3826]}
{"unit_id": "orig:1921132141221486893", "ticker_idx": 1, "ts": "2025-05-10T09:16:10+00:00", "symbol": "ASM International", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASMI remains the leading player due to its position as the only vertically integrated vendor with an in-house chemical supply chain", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish ASML on continued EUV intensity ramp and 1000W source power milestone; bullish ASMI as sole vertically integrated ALD leader.", "resolved_tickers": ["ASMIY"], "resolution_method": "exchange_ticker", "resolution_reason": "ASM International Amsterdam; US OTC ADR ASMIY", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Expert Call with Former Intel Lithography and Foundry Executive\n\nYesterday, we hosted an expert call with Klaus Shuegraf, currently VP at Photonic and formerly Director of Lithography Sourcing and VP of Foundry Strategy & Planning at Intel.\n\n1. EUV Source Power and Increasing Lithographic Intensity Expected to Continue\n\nShuegraf acknowledged that the transition from N3 to N2 nodes does not significantly shrink feature sizes due to the Gate-All-Around (GAA) structure. However, he projected that lithographic intensity will continue to increase for nodes beyond that. He expressed confidence in ASML achieving 1000W source power from the current 600W, but noted that reaching 1500W would approach technical limitations.\n\n2. 18A Node Outlook and High-NA Adoption Timing\n\nIn line with industry trends, Shuegraf stated that Intel must first prove the viability of its 18A node through its own products before attracting potential customers such as AVGO (Broadcom), NVDA (Nvidia), and AAPL (Apple M-series). He predicted that 2026 would be the mainstream ramp-up period for 18A, which would be about one year behind TSMC’s 2nm node production schedule. Regarding High-NA EUV, he noted that TSMC’s full adoption would likely be in the late 2020s, as it will take time for ASML to scale up production capacity.\n\n3. Potential Expansion of EUV Adoption in DRAM\n\nShuegraf suggested that if leading HBM suppliers adopt EUV, it could drive broader EUV adoption across the DRAM industry. This is due to trends such as AI PCs and advanced packaging making DRAM a more specialized, rather than purely commoditized, product segment. While some customers may be able to maintain current lithographic intensity levels through 4F2 architecture, others are expected to adopt additional litho layers alongside 4F2.\n\n4. Additional Notes on China and ALD\n\n(A) Regarding China’s development of EUV using DPP (Discharge Produced Plasma), Shuegraf acknowledged that for AI use cases, high wafer throughput is not always essential. However, he commented that DPP is economically inefficient and presents challenges in supply availability.\n\n(B) On the topic of ALD (Atomic Layer Deposition), he emphasized that ASMI remains the leading player due to its position as the only vertically integrated vendor with an in-house chemical supply chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06790809825691857, "ret_signed_m1d": -0.06790809825691857, "alpha_spy_m1d": -0.035917787085782305, "alpha_c_m1d": -0.008896216154575809, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.031092876584287055, "ret_signed_p1d": 0.031092876584287055, "alpha_spy_p1d": 0.024488936267117012, "alpha_c_p1d": -0.0032358831786085673, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.002619418047691724, "ret_signed_p1w": 0.002619418047691724, "alpha_spy_p1w": -0.017723882226053478, "alpha_c_p1w": -0.03360481488042932, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.13492400932527793, "ret_signed_p1m": 0.13492400932527793, "alpha_spy_p1m": 0.10046383887703203, "alpha_c_p1m": 0.03105304694906641, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": -0.09880165280822972, "ret_signed_p3m": -0.09880165280822972, "alpha_spy_p3m": -0.18650304762641745, "alpha_c_p3m": -0.32503469194994394, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.17431006571731333, "ret_signed_p6m": 0.17431006571731333, "alpha_spy_p6m": 0.00942938065433907, "alpha_c_p6m": -0.31155831146559954, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0889, 0.1024, 0.1047, 0.1047, 0.0951, 0.0952, 0.0951, 0.0826, 0.0814, 0.045, 0.0616, -0.0232, -0.0233, -0.0484, -0.072, -0.0307, -0.1089, -0.0835, -0.16, -0.1341, -0.105, -0.1274, -0.0891, -0.0768, -0.0953, -0.0723, -0.0884, -0.1117, -0.0889, -0.0949, -0.1102, -0.1267, -0.1386, -0.1625, -0.1718, -0.1723, -0.2207, -0.2629, -0.2622, -0.3071, -0.2024, -0.2458, -0.1896, -0.1985, -0.1529, -0.1857, -0.1559, -0.1559, -0.1873, -0.1794, -0.1618, -0.1219, -0.0947, -0.0924, -0.0635, -0.1029, -0.1025, -0.072, -0.0869, -0.0991, -0.0652, -0.0716, -0.0679, 0.0, 0.0311, 0.0295, 0.0174, 0.0107, 0.0026, 0.0152, -0.0119, 0.0018, -0.0091, -0.0091, 0.026, 0.0121, 0.0434, 0.0072, 0.0108, 0.0274, 0.0381, 0.0376, 0.067, 0.0939, 0.1349, 0.1343, 0.1253, 0.106, 0.116, 0.1042, 0.0992, 0.0992, 0.0885, 0.1209, 0.1533, 0.1723, 0.1695, 0.1745, 0.1863, 0.1496, 0.1323, 0.1182, 0.1182, 0.1081, 0.1181, 0.1158, 0.1417, 0.1346, 0.1349, 0.1383, 0.1013, 0.1136, 0.1173, 0.1232, 0.0365, -0.0062, -0.0403, -0.073, -0.0384, -0.0518, -0.0486, -0.104, -0.1082, -0.1013, -0.1079, -0.1212, -0.0988, -0.091, -0.102, -0.0692, -0.0711, -0.0746, -0.1042, -0.1106, -0.1138, -0.1192, -0.1357, -0.1148, -0.1215, -0.1127, -0.1127, -0.0743, -0.1116, -0.1116, -0.1251, -0.1307, -0.1141, -0.0946, -0.0814, -0.0825, -0.0893, -0.0732, -0.076, -0.0179, 0.0039, 0.0124, 0.0869, 0.0739, 0.1013, 0.0857, 0.0889, 0.0778, 0.0695, 0.0906, 0.1124, 0.142, 0.2019, 0.1855, 0.2111, 0.1544, 0.1822, 0.1689, 0.1257, 0.1656, 0.1665, 0.1727, 0.1751, 0.1729, 0.1781, 0.1759, 0.1505, 0.1792, 0.1966, 0.2176, 0.1291, 0.1829, 0.2071, 0.2004, 0.2086, 0.1743]}
{"unit_id": "orig:1932966867226013745", "ticker_idx": 0, "ts": "2025-06-12T01:03:08+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note: GB200 shipments accelerating toward Q2 target of 5k-6k units; Giga-Byte top pick among AI server ODMs, followed by Hon Hai, Quanta, Wistron, Wiwynn.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach\nMorgan Stanley (S. Shih, 2025/06/09)\n\nMorgan Stanley estimates that GB200 rack shipments in May rose significantly month-over-month to 2,000–2,500 units. If this momentum continues in June, the total for Q2 is expected to reach 5,000–6,000 units.\n• Quanta: Stable Increment\nMorgan Stanley estimates Quanta’s May revenue at approximately NT$160 billion (MoM +4%, YoY +58%), driven by 3.8 million NB shipments (MoM +15%) and around 400 GB200 rack shipments. Rack shipments rose slightly from April’s 300–400 units.\n• Wistron: Significant Ramp-Up\nMorgan Stanley estimates Wistron’s May revenue at NT$208.4 billion (MoM +56%, YoY +162%), with cumulative April–May revenue accounting for 76%–79% of the estimated Q2 total. Strong shipments in NB (+6%), DT (+47%), and monitors (+7%), along with Wiwynn’s May revenue of approximately NT$71 billion (+11%), drove GB200-equivalent rack shipments in May to 900–1,000 units, a nearly 6-fold increase from April’s ~150 units.\n• Hon Hai: Slight Decline from High Base\nBased on supply chain checks, Morgan Stanley notes that Hon Hai’s cloud and networking business saw a slight decline in May due to a high base and unfavorable FX conditions. However, GB200 rack shipments still reached nearly 1,000 units. Looking ahead to Q2, Hon Hai is expected to ship a total of 3,000–4,000 units.\n• OEM/ODM Preference Ranking\nMorgan Stanley highlights that among downstream AI server OEM/ODMs, Giga-Byte is rated the “top pick” for its stable deliveries and strong price-performance ratio; followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014965570977212073, "ret_signed_m1d": -0.014965570977212073, "alpha_spy_m1d": -0.0110070768307996, "alpha_c_m1d": -0.010139457522481021, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.004826113454731051, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02089664809731795, "ret_signed_p1d": -0.02089664809731795, "alpha_spy_p1d": -0.00971658370085049, "alpha_c_p1d": 0.002512084485439159, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.023408732582757108, "ret_p1w": 0.003310234002421364, "ret_signed_p1w": 0.003310234002421364, "alpha_spy_p1w": 0.013761443846784127, "alpha_c_p1w": 0.0054382794497384035, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.0021280454473170396, "ret_p1m": 0.13737928949796663, "ret_signed_p1m": 0.13737928949796663, "alpha_spy_p1m": 0.10141469375098833, "alpha_c_p1m": 0.044884520618266865, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.09249476887969976, "ret_p3m": 0.17765508374204741, "ret_signed_p3m": 0.17765508374204741, "alpha_spy_p3m": 0.09731952434157742, "alpha_c_p3m": 0.04594311172108134, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.13171197202096607, "ret_p6m": 0.25814099590775474, "ret_signed_p6m": 0.25814099590775474, "alpha_spy_p6m": 0.11590699330911414, "alpha_c_p6m": -0.1266967064546165, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.38483770236237125, "price_path": [-0.1757, -0.204, -0.1896, -0.1826, -0.1883, -0.1627, -0.1677, -0.2155, -0.2316, -0.2437, -0.2526, -0.2404, -0.2385, -0.298, -0.3496, -0.3267, -0.3359, -0.2116, -0.2582, -0.235, -0.2365, -0.2263, -0.2794, -0.3001, -0.3001, -0.3317, -0.318, -0.2917, -0.2661, -0.2345, -0.2502, -0.2482, -0.2489, -0.2303, -0.2104, -0.2151, -0.217, -0.1927, -0.1906, -0.1956, -0.1518, -0.104, -0.0667, -0.0702, -0.0663, -0.0651, -0.0733, -0.0911, -0.084, -0.0946, -0.0946, -0.0656, -0.0703, -0.0401, -0.0681, -0.0526, -0.0261, -0.0213, -0.0346, -0.0227, -0.0164, -0.0072, -0.015, 0.0, -0.0209, -0.0021, -0.0061, 0.0033, 0.0033, -0.0079, -0.0057, 0.02, 0.0642, 0.0691, 0.0879, 0.0896, 0.0572, 0.0845, 0.0989, 0.0989, 0.0913, 0.1034, 0.1233, 0.1317, 0.1374, 0.1315, 0.1772, 0.1819, 0.1931, 0.189, 0.1819, 0.1519, 0.1778, 0.1982, 0.1966, 0.219, 0.2104, 0.2363, 0.2267, 0.1981, 0.2414, 0.2294, 0.2374, 0.2467, 0.26, 0.2556, 0.2632, 0.2523, 0.2553, 0.2445, 0.2552, 0.2113, 0.2097, 0.2068, 0.2275, 0.2401, 0.2536, 0.2524, 0.2426, 0.2012, 0.2012, 0.1778, 0.1767, 0.1839, 0.1519, 0.1608, 0.1777, 0.223, 0.2219, 0.2264, 0.2259, 0.2061, 0.1745, 0.2155, 0.2185, 0.2663, 0.2306, 0.2206, 0.2255, 0.229, 0.2542, 0.2868, 0.2914, 0.3028, 0.294, 0.2797, 0.2762, 0.3043, 0.3281, 0.2632, 0.2988, 0.2417, 0.2403, 0.2539, 0.2637, 0.2597, 0.2494, 0.2434, 0.2563, 0.2846, 0.3207, 0.3865, 0.4279, 0.3993, 0.3966, 0.4268, 0.3704, 0.3464, 0.2972, 0.2977, 0.3728, 0.3322, 0.3366, 0.2888, 0.3116, 0.287, 0.2508, 0.2864, 0.2459, 0.2337, 0.259, 0.2264, 0.2432, 0.2432, 0.2208, 0.2409, 0.2515, 0.2386, 0.2648, 0.2581]}
{"unit_id": "orig:1932966867226013745", "ticker_idx": 1, "ts": "2025-06-12T01:03:08+00:00", "symbol": "Giga-Byte Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Giga-Byte is rated the 'top pick' for its stable deliveries and strong price-performance ratio", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note: GB200 shipments accelerating toward Q2 target of 5k-6k units; Giga-Byte top pick among AI server ODMs, followed by Hon Hai, Quanta, Wistron, Wiwynn.", "resolved_tickers": ["2376.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Gigabyte Technology 2376.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach\nMorgan Stanley (S. Shih, 2025/06/09)\n\nMorgan Stanley estimates that GB200 rack shipments in May rose significantly month-over-month to 2,000–2,500 units. If this momentum continues in June, the total for Q2 is expected to reach 5,000–6,000 units.\n• Quanta: Stable Increment\nMorgan Stanley estimates Quanta’s May revenue at approximately NT$160 billion (MoM +4%, YoY +58%), driven by 3.8 million NB shipments (MoM +15%) and around 400 GB200 rack shipments. Rack shipments rose slightly from April’s 300–400 units.\n• Wistron: Significant Ramp-Up\nMorgan Stanley estimates Wistron’s May revenue at NT$208.4 billion (MoM +56%, YoY +162%), with cumulative April–May revenue accounting for 76%–79% of the estimated Q2 total. Strong shipments in NB (+6%), DT (+47%), and monitors (+7%), along with Wiwynn’s May revenue of approximately NT$71 billion (+11%), drove GB200-equivalent rack shipments in May to 900–1,000 units, a nearly 6-fold increase from April’s ~150 units.\n• Hon Hai: Slight Decline from High Base\nBased on supply chain checks, Morgan Stanley notes that Hon Hai’s cloud and networking business saw a slight decline in May due to a high base and unfavorable FX conditions. However, GB200 rack shipments still reached nearly 1,000 units. Looking ahead to Q2, Hon Hai is expected to ship a total of 3,000–4,000 units.\n• OEM/ODM Preference Ranking\nMorgan Stanley highlights that among downstream AI server OEM/ODMs, Giga-Byte is rated the “top pick” for its stable deliveries and strong price-performance ratio; followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003533561004010588, "ret_signed_m1d": 0.003533561004010588, "alpha_spy_m1d": 0.00749205515042306, "alpha_c_m1d": 0.012726857405829128, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014134244016042796, "ret_signed_p1d": 0.014134244016042796, "alpha_spy_p1d": 0.025314308412510256, "alpha_c_p1d": 0.02815080088246713, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.010600683012032097, "ret_signed_p1w": -0.010600683012032097, "alpha_spy_p1w": -0.00014947316766933394, "alpha_c_p1w": -0.006643002947173682, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": -0.008833902510026692, "ret_signed_p1m": -0.008833902510026692, "alpha_spy_p1m": -0.04479849825700499, "alpha_c_p1m": -0.06532906106111269, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": 0.042192113567631084, "ret_signed_p3m": 0.042192113567631084, "alpha_spy_p3m": -0.03814344583283891, "alpha_c_p3m": -0.053490581182423336, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": -0.1061691539173919, "ret_signed_p6m": -0.1061691539173919, "alpha_spy_p6m": -0.2484031565160325, "alpha_c_p6m": -0.31845163983241487, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [-0.0866, -0.0636, -0.0671, -0.053, -0.0565, -0.0583, -0.0512, -0.0636, -0.0777, -0.0919, -0.1449, -0.129, -0.1272, -0.1272, -0.1272, -0.2138, -0.2898, -0.3604, -0.2968, -0.2862, -0.2491, -0.2244, -0.2403, -0.2244, -0.2261, -0.2491, -0.2615, -0.2226, -0.2261, -0.1837, -0.1855, -0.1767, -0.1926, -0.1926, -0.1661, -0.1784, -0.1537, -0.1608, -0.1466, -0.136, -0.1201, -0.1219, -0.0707, -0.0618, -0.0795, -0.0936, -0.0919, -0.0813, -0.0883, -0.0954, -0.0866, -0.0936, -0.0459, -0.0247, -0.0247, -0.0601, -0.0018, 0.0141, 0.0159, 0.0177, 0.023, 0.0141, 0.0035, 0.0, 0.0141, 0.0, 0.0, 0.0071, -0.0106, -0.0106, -0.0212, 0.0035, 0.0053, 0.0035, -0.0053, 0.0, 0.0124, -0.0053, -0.023, -0.0353, -0.0371, -0.0318, -0.0159, -0.0177, -0.0088, -0.023, 0.0018, 0.0053, 0.0141, 0.023, 0.0265, -0.0106, 0.0177, 0.0265, 0.0159, 0.0212, -0.0018, -0.0164, 0.0074, 0.0074, 0.0019, 0.0184, 0.0184, 0.0422, 0.0477, 0.0605, 0.0459, 0.0514, 0.033, 0.022, 0.033, 0.0404, -0.0109, -0.0018, -0.0073, 0.0165, 0.0275, 0.0422, 0.0294, 0.0147, -0.0237, -0.0347, -0.0329, -0.0292, -0.0146, 0.0056, 0.0422, 0.0147, 0.0037, 0.0037, -0.0054, 0.0129, -0.0109, 0.0092, 0.033, 0.0294, 0.0367, 0.0184, 0.0733, 0.0733, 0.0733, 0.0953, 0.0733, 0.0953, 0.1411, 0.1411, 0.1393, 0.1228, 0.0953, 0.0953, 0.0587, 0.0257, 0.022, 0.0367, 0.0202, 0.044, 0.055, 0.033, 0.0239, 0.0239, 0.044, 0.0495, 0.0422, 0.0312, 0.0275, 0.0147, -0.0219, -0.0219, 0.0202, -0.0073, 0.0111, 0.0074, -0.0073, -0.0109, -0.0091, -0.0622, -0.0988, -0.1227, -0.0915, -0.1227, -0.108, -0.1117, -0.0988, -0.0952, -0.108, -0.1098, -0.1135, -0.1025, -0.1043, -0.1062]}
{"unit_id": "orig:1932966867226013745", "ticker_idx": 2, "ts": "2025-06-12T01:03:08+00:00", "symbol": "Hon Hai", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note: GB200 shipments accelerating toward Q2 target of 5k-6k units; Giga-Byte top pick among AI server ODMs, followed by Hon Hai, Quanta, Wistron, Wiwynn.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach\nMorgan Stanley (S. Shih, 2025/06/09)\n\nMorgan Stanley estimates that GB200 rack shipments in May rose significantly month-over-month to 2,000–2,500 units. If this momentum continues in June, the total for Q2 is expected to reach 5,000–6,000 units.\n• Quanta: Stable Increment\nMorgan Stanley estimates Quanta’s May revenue at approximately NT$160 billion (MoM +4%, YoY +58%), driven by 3.8 million NB shipments (MoM +15%) and around 400 GB200 rack shipments. Rack shipments rose slightly from April’s 300–400 units.\n• Wistron: Significant Ramp-Up\nMorgan Stanley estimates Wistron’s May revenue at NT$208.4 billion (MoM +56%, YoY +162%), with cumulative April–May revenue accounting for 76%–79% of the estimated Q2 total. Strong shipments in NB (+6%), DT (+47%), and monitors (+7%), along with Wiwynn’s May revenue of approximately NT$71 billion (+11%), drove GB200-equivalent rack shipments in May to 900–1,000 units, a nearly 6-fold increase from April’s ~150 units.\n• Hon Hai: Slight Decline from High Base\nBased on supply chain checks, Morgan Stanley notes that Hon Hai’s cloud and networking business saw a slight decline in May due to a high base and unfavorable FX conditions. However, GB200 rack shipments still reached nearly 1,000 units. Looking ahead to Q2, Hon Hai is expected to ship a total of 3,000–4,000 units.\n• OEM/ODM Preference Ranking\nMorgan Stanley highlights that among downstream AI server OEM/ODMs, Giga-Byte is rated the “top pick” for its stable deliveries and strong price-performance ratio; followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00952381812950076, "ret_signed_m1d": -0.00952381812950076, "alpha_spy_m1d": -0.005565323983088288, "alpha_c_m1d": -0.0003305217276822203, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006349178619756701, "ret_signed_p1d": -0.006349178619756701, "alpha_spy_p1d": 0.004830885776710758, "alpha_c_p1d": 0.007667378246667633, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.022222175369014163, "ret_signed_p1w": -0.022222175369014163, "alpha_spy_p1w": -0.0117709655246514, "alpha_c_p1w": -0.018264495304155748, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": 0.06263718368311588, "ret_signed_p1m": 0.06263718368311588, "alpha_spy_p1m": 0.02667258793613758, "alpha_c_p1m": 0.006142025132029882, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": 0.36530783662072164, "ret_signed_p3m": 0.36530783662072164, "alpha_spy_p3m": 0.28497227722025165, "alpha_c_p3m": 0.2696251418706672, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 0.5199330614910203, "ret_signed_p6m": 0.5199330614910203, "alpha_spy_p6m": 0.37769905889237965, "alpha_c_p6m": 0.3076505755759973, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [0.0667, 0.0635, 0.0413, 0.054, 0.0476, 0.0349, 0.0476, 0.054, 0.019, -0.0222, -0.073, -0.0349, -0.0254, -0.0254, -0.0254, -0.1206, -0.2063, -0.2857, -0.2159, -0.146, -0.1206, -0.1143, -0.1365, -0.146, -0.1397, -0.1333, -0.1619, -0.1175, -0.1333, -0.1175, -0.0952, -0.0921, -0.1016, -0.1016, -0.0635, -0.0952, -0.073, -0.0794, -0.073, -0.0667, -0.0286, 0.0032, 0.0349, 0.0127, 0.0032, -0.0222, -0.0222, -0.0159, -0.0286, -0.0222, -0.0286, -0.0349, -0.0349, -0.0095, -0.0095, -0.0317, -0.0381, -0.0095, -0.0127, -0.0286, -0.019, -0.0127, -0.0095, 0.0, -0.0063, -0.0159, -0.0159, -0.0095, -0.0222, -0.0127, -0.0095, 0.019, 0.0286, 0.0349, 0.0476, 0.0222, 0.0508, 0.0659, 0.0791, 0.0593, 0.0593, 0.0462, 0.0593, 0.0626, 0.0626, 0.0561, 0.0725, 0.0692, 0.0791, 0.089, 0.0857, 0.0659, 0.0922, 0.1449, 0.1482, 0.158, 0.1284, 0.1284, 0.1712, 0.1712, 0.1844, 0.214, 0.2238, 0.2798, 0.2798, 0.3028, 0.2962, 0.3061, 0.3127, 0.362, 0.3818, 0.3686, 0.3192, 0.3554, 0.3324, 0.3653, 0.3752, 0.3686, 0.3554, 0.339, 0.3061, 0.3061, 0.3192, 0.3357, 0.3489, 0.339, 0.3653, 0.3818, 0.4114, 0.4311, 0.4212, 0.4179, 0.3949, 0.4147, 0.4081, 0.4212, 0.4541, 0.4673, 0.5134, 0.4443, 0.4443, 0.4212, 0.441, 0.4837, 0.4903, 0.4903, 0.5035, 0.4805, 0.4574, 0.4574, 0.4015, 0.3554, 0.3587, 0.4673, 0.4903, 0.5693, 0.5726, 0.5956, 0.5726, 0.5726, 0.6219, 0.6449, 0.7107, 0.7239, 0.6943, 0.6548, 0.6055, 0.6252, 0.6318, 0.6055, 0.6417, 0.6219, 0.6515, 0.6581, 0.5857, 0.5528, 0.5134, 0.5068, 0.5561, 0.4805, 0.4476, 0.441, 0.4936, 0.5199, 0.4837, 0.4607, 0.4607, 0.4936, 0.5035, 0.5199]}
{"unit_id": "orig:1932966867226013745", "ticker_idx": 3, "ts": "2025-06-12T01:03:08+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note: GB200 shipments accelerating toward Q2 target of 5k-6k units; Giga-Byte top pick among AI server ODMs, followed by Hon Hai, Quanta, Wistron, Wiwynn.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach\nMorgan Stanley (S. Shih, 2025/06/09)\n\nMorgan Stanley estimates that GB200 rack shipments in May rose significantly month-over-month to 2,000–2,500 units. If this momentum continues in June, the total for Q2 is expected to reach 5,000–6,000 units.\n• Quanta: Stable Increment\nMorgan Stanley estimates Quanta’s May revenue at approximately NT$160 billion (MoM +4%, YoY +58%), driven by 3.8 million NB shipments (MoM +15%) and around 400 GB200 rack shipments. Rack shipments rose slightly from April’s 300–400 units.\n• Wistron: Significant Ramp-Up\nMorgan Stanley estimates Wistron’s May revenue at NT$208.4 billion (MoM +56%, YoY +162%), with cumulative April–May revenue accounting for 76%–79% of the estimated Q2 total. Strong shipments in NB (+6%), DT (+47%), and monitors (+7%), along with Wiwynn’s May revenue of approximately NT$71 billion (+11%), drove GB200-equivalent rack shipments in May to 900–1,000 units, a nearly 6-fold increase from April’s ~150 units.\n• Hon Hai: Slight Decline from High Base\nBased on supply chain checks, Morgan Stanley notes that Hon Hai’s cloud and networking business saw a slight decline in May due to a high base and unfavorable FX conditions. However, GB200 rack shipments still reached nearly 1,000 units. Looking ahead to Q2, Hon Hai is expected to ship a total of 3,000–4,000 units.\n• OEM/ODM Preference Ranking\nMorgan Stanley highlights that among downstream AI server OEM/ODMs, Giga-Byte is rated the “top pick” for its stable deliveries and strong price-performance ratio; followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012216429426923914, "ret_signed_m1d": -0.012216429426923914, "alpha_spy_m1d": -0.008257935280511441, "alpha_c_m1d": -0.0030231330251053734, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003490408407692547, "ret_signed_p1d": 0.003490408407692547, "alpha_spy_p1d": 0.014670472804160006, "alpha_c_p1d": 0.01750696527411688, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.026177951457476478, "ret_signed_p1w": -0.026177951457476478, "alpha_spy_p1w": -0.015726741613113715, "alpha_c_p1w": -0.022220271392618063, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": -0.0016619504414681074, "ret_signed_p1m": -0.0016619504414681074, "alpha_spy_p1m": -0.03762654618844641, "alpha_c_p1m": -0.05815710899255411, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": -0.04005956773218089, "ret_signed_p3m": -0.04005956773218089, "alpha_spy_p3m": -0.12039512713265088, "alpha_c_p3m": -0.1357422624822353, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 0.08793248990352831, "ret_signed_p6m": 0.08793248990352831, "alpha_spy_p6m": -0.0543015126951123, "alpha_c_p6m": -0.12434999601149466, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [-0.1099, -0.0942, -0.1152, -0.103, -0.1065, -0.1291, -0.1291, -0.1204, -0.1414, -0.1588, -0.2164, -0.1832, -0.1867, -0.1867, -0.1867, -0.267, -0.3316, -0.377, -0.3159, -0.2478, -0.2042, -0.2042, -0.2077, -0.1972, -0.1955, -0.1902, -0.2199, -0.1798, -0.1972, -0.1798, -0.1449, -0.1536, -0.1675, -0.1675, -0.1239, -0.1396, -0.1134, -0.1134, -0.1222, -0.1047, -0.0977, -0.0855, -0.0349, -0.0314, -0.0401, -0.0716, -0.0698, -0.0506, -0.0663, -0.0785, -0.0733, -0.0716, -0.0611, -0.0524, -0.0524, -0.0541, -0.0401, -0.0227, -0.0175, -0.0366, -0.0366, -0.0105, -0.0122, 0.0, 0.0035, -0.0017, -0.0122, -0.0122, -0.0262, -0.0244, -0.0157, 0.0105, 0.0227, 0.0122, -0.0035, 0.0038, 0.0276, 0.0313, 0.0239, 0.0166, 0.013, -0.0053, 0.0093, -0.0017, -0.0017, -0.0145, -0.0108, -0.0035, -0.009, 0.0002, 0.0002, -0.0401, -0.0327, -0.0163, -0.0181, -0.0053, -0.0218, -0.0126, 0.0294, 0.0294, 0.0276, 0.0532, 0.055, 0.0568, 0.0623, 0.0605, 0.0349, -0.0071, -0.0108, -0.0108, -0.0163, -0.0199, -0.051, -0.0529, -0.0547, -0.0199, -0.0309, -0.0163, -0.0273, -0.0401, -0.0638, -0.073, -0.073, -0.0602, -0.0547, -0.0492, -0.0401, -0.0017, -0.0126, 0.002, 0.0093, 0.0002, -0.0017, 0.0221, 0.0276, 0.0276, 0.0477, 0.0367, 0.0422, 0.0422, 0.0422, 0.0605, 0.0477, 0.0751, 0.0971, 0.0971, 0.1282, 0.119, 0.0898, 0.0898, 0.077, 0.0386, 0.0459, 0.066, 0.055, 0.0843, 0.0642, 0.0587, 0.0733, 0.0733, 0.1318, 0.1117, 0.1263, 0.119, 0.0989, 0.0879, 0.0514, 0.0587, 0.0715, 0.0495, 0.066, 0.066, 0.0532, 0.0422, 0.0221, 0.0093, -0.0126, -0.0236, 0.002, -0.0126, -0.0053, 0.0057, 0.0221, 0.0294, 0.0313, 0.0495, 0.055, 0.0696, 0.077, 0.0879]}
{"unit_id": "orig:1932966867226013745", "ticker_idx": 4, "ts": "2025-06-12T01:03:08+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note: GB200 shipments accelerating toward Q2 target of 5k-6k units; Giga-Byte top pick among AI server ODMs, followed by Hon Hai, Quanta, Wistron, Wiwynn.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach\nMorgan Stanley (S. Shih, 2025/06/09)\n\nMorgan Stanley estimates that GB200 rack shipments in May rose significantly month-over-month to 2,000–2,500 units. If this momentum continues in June, the total for Q2 is expected to reach 5,000–6,000 units.\n• Quanta: Stable Increment\nMorgan Stanley estimates Quanta’s May revenue at approximately NT$160 billion (MoM +4%, YoY +58%), driven by 3.8 million NB shipments (MoM +15%) and around 400 GB200 rack shipments. Rack shipments rose slightly from April’s 300–400 units.\n• Wistron: Significant Ramp-Up\nMorgan Stanley estimates Wistron’s May revenue at NT$208.4 billion (MoM +56%, YoY +162%), with cumulative April–May revenue accounting for 76%–79% of the estimated Q2 total. Strong shipments in NB (+6%), DT (+47%), and monitors (+7%), along with Wiwynn’s May revenue of approximately NT$71 billion (+11%), drove GB200-equivalent rack shipments in May to 900–1,000 units, a nearly 6-fold increase from April’s ~150 units.\n• Hon Hai: Slight Decline from High Base\nBased on supply chain checks, Morgan Stanley notes that Hon Hai’s cloud and networking business saw a slight decline in May due to a high base and unfavorable FX conditions. However, GB200 rack shipments still reached nearly 1,000 units. Looking ahead to Q2, Hon Hai is expected to ship a total of 3,000–4,000 units.\n• OEM/ODM Preference Ranking\nMorgan Stanley highlights that among downstream AI server OEM/ODMs, Giga-Byte is rated the “top pick” for its stable deliveries and strong price-performance ratio; followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012875536480686733, "ret_signed_m1d": -0.012875536480686733, "alpha_spy_m1d": -0.008917042334274261, "alpha_c_m1d": -0.003682240078868193, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01716738197424883, "ret_signed_p1d": 0.01716738197424883, "alpha_spy_p1d": 0.02834744637071629, "alpha_c_p1d": 0.031183938840673164, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.008583690987124415, "ret_signed_p1w": -0.008583690987124415, "alpha_spy_p1w": 0.0018675188572383483, "alpha_c_p1w": -0.0046260109222659995, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": 0.025751072961373467, "ret_signed_p1m": 0.025751072961373467, "alpha_spy_p1m": -0.010213522785604834, "alpha_c_p1m": -0.030744085589712533, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": -0.0042918454935622075, "ret_signed_p3m": -0.0042918454935622075, "alpha_spy_p3m": -0.0846274048940322, "alpha_c_p3m": -0.09997454024361663, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 0.296137339055794, "ret_signed_p6m": 0.296137339055794, "alpha_spy_p6m": 0.15390333645715337, "alpha_c_p6m": 0.08385485314077101, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [-0.1202, -0.1077, -0.1451, -0.1285, -0.1243, -0.1409, -0.1409, -0.1243, -0.1326, -0.1658, -0.209, -0.1575, -0.1617, -0.1617, -0.1617, -0.2455, -0.3202, -0.38, -0.3186, -0.2505, -0.1924, -0.1492, -0.1725, -0.1658, -0.1285, -0.1451, -0.1799, -0.1492, -0.1617, -0.1492, -0.1492, -0.1368, -0.1575, -0.1575, -0.1243, -0.1409, -0.0953, -0.116, -0.1119, -0.1119, -0.0994, -0.0994, -0.0496, -0.0745, -0.0787, -0.0953, -0.0911, -0.0621, -0.0621, -0.0621, -0.0372, -0.0455, -0.0413, -0.033, -0.033, -0.0455, -0.0601, -0.0215, -0.0129, -0.0558, -0.0258, -0.0172, -0.0129, 0.0, 0.0172, 0.0215, 0.0129, -0.0043, -0.0086, 0.0172, 0.0172, 0.03, 0.0773, 0.073, 0.0386, 0.0515, 0.03, 0.03, 0.0386, 0.0215, 0.0129, 0.0172, 0.0343, 0.03, 0.0258, 0.0043, 0.03, 0.0129, 0.0129, 0.0215, 0.0172, -0.0172, -0.0043, 0.0043, 0.0, 0.03, 0.0343, 0.0343, 0.0558, 0.0558, 0.0258, 0.0644, 0.0515, 0.0601, 0.0687, 0.0944, 0.0815, 0.0086, 0.0, 0.0, -0.0043, -0.0086, -0.0258, -0.0429, -0.0386, -0.0258, -0.03, -0.0129, -0.0129, -0.03, -0.0601, -0.0558, -0.0558, -0.03, -0.0086, -0.0172, -0.0043, 0.03, 0.0215, 0.03, 0.0258, 0.0129, 0.0215, 0.03, 0.0386, 0.0429, 0.0515, 0.0429, 0.0773, 0.0987, 0.0987, 0.206, 0.2532, 0.2833, 0.3176, 0.3176, 0.3391, 0.2918, 0.2961, 0.2961, 0.2489, 0.1845, 0.1845, 0.1974, 0.1717, 0.2189, 0.2232, 0.2318, 0.2275, 0.2275, 0.2575, 0.2704, 0.309, 0.3305, 0.2918, 0.2318, 0.1931, 0.1803, 0.2189, 0.1803, 0.1888, 0.2403, 0.1931, 0.2361, 0.206, 0.1845, 0.176, 0.176, 0.2275, 0.1931, 0.1931, 0.1974, 0.2232, 0.2489, 0.2403, 0.2103, 0.2275, 0.2446, 0.2532, 0.2961]}
{"unit_id": "orig:1932966867226013745", "ticker_idx": 5, "ts": "2025-06-12T01:03:08+00:00", "symbol": "Wiwynn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note: GB200 shipments accelerating toward Q2 target of 5k-6k units; Giga-Byte top pick among AI server ODMs, followed by Hon Hai, Quanta, Wistron, Wiwynn.", "resolved_tickers": ["6669.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wiwynn Corp listed on TWSE as 6669.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA GB200 NVL72 May Shipments: Accelerating Continuously, 2Q Target Within Reach\nMorgan Stanley (S. Shih, 2025/06/09)\n\nMorgan Stanley estimates that GB200 rack shipments in May rose significantly month-over-month to 2,000–2,500 units. If this momentum continues in June, the total for Q2 is expected to reach 5,000–6,000 units.\n• Quanta: Stable Increment\nMorgan Stanley estimates Quanta’s May revenue at approximately NT$160 billion (MoM +4%, YoY +58%), driven by 3.8 million NB shipments (MoM +15%) and around 400 GB200 rack shipments. Rack shipments rose slightly from April’s 300–400 units.\n• Wistron: Significant Ramp-Up\nMorgan Stanley estimates Wistron’s May revenue at NT$208.4 billion (MoM +56%, YoY +162%), with cumulative April–May revenue accounting for 76%–79% of the estimated Q2 total. Strong shipments in NB (+6%), DT (+47%), and monitors (+7%), along with Wiwynn’s May revenue of approximately NT$71 billion (+11%), drove GB200-equivalent rack shipments in May to 900–1,000 units, a nearly 6-fold increase from April’s ~150 units.\n• Hon Hai: Slight Decline from High Base\nBased on supply chain checks, Morgan Stanley notes that Hon Hai’s cloud and networking business saw a slight decline in May due to a high base and unfavorable FX conditions. However, GB200 rack shipments still reached nearly 1,000 units. Looking ahead to Q2, Hon Hai is expected to ship a total of 3,000–4,000 units.\n• OEM/ODM Preference Ranking\nMorgan Stanley highlights that among downstream AI server OEM/ODMs, Giga-Byte is rated the “top pick” for its stable deliveries and strong price-performance ratio; followed by Hon Hai, Quanta, Wistron, and Wiwynn, in that order.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0039447377658428096, "ret_signed_m1d": -0.0039447377658428096, "alpha_spy_m1d": 1.3756380569662774e-05, "alpha_c_m1d": 0.005248558635975731, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.00919329640181854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011834312483814058, "ret_signed_p1d": 0.011834312483814058, "alpha_spy_p1d": 0.023014376880281517, "alpha_c_p1d": 0.025850869350238392, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.014016556866424335, "ret_p1w": -0.0019724184760642194, "ret_signed_p1w": -0.0019724184760642194, "alpha_spy_p1w": 0.008478791368298544, "alpha_c_p1w": 0.001985261588794196, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.003957680064858415, "ret_p1m": 0.05832560258892072, "ret_signed_p1m": 0.05832560258892072, "alpha_spy_p1m": 0.02236100684194242, "alpha_c_p1m": 0.001830444037834722, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.056495158551086, "ret_p3m": 0.22286374809698706, "ret_signed_p3m": 0.22286374809698706, "alpha_spy_p3m": 0.14252818869651707, "alpha_c_p3m": 0.12718105334693264, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.09568269475005442, "ret_p6m": 0.903361016556274, "ret_signed_p6m": 0.903361016556274, "alpha_spy_p6m": 0.7611270139576334, "alpha_c_p6m": 0.691078530641251, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.21228248591502297, "price_path": [-0.2465, -0.217, -0.2327, -0.2051, -0.2308, -0.2525, -0.2446, -0.2426, -0.2722, -0.3077, -0.355, -0.3195, -0.3136, -0.3136, -0.3136, -0.3807, -0.3531, -0.3669, -0.3037, -0.2505, -0.2465, -0.2663, -0.2742, -0.2584, -0.2564, -0.2742, -0.3018, -0.2505, -0.2643, -0.2485, -0.2505, -0.2387, -0.2465, -0.2465, -0.1854, -0.1953, -0.1617, -0.1755, -0.1519, -0.1321, -0.1105, -0.1223, -0.075, -0.0611, -0.0611, -0.1085, -0.0927, -0.075, -0.0769, -0.1006, -0.0651, -0.0828, -0.0611, -0.0454, -0.0454, -0.0493, -0.0217, 0.0, -0.0197, -0.0276, -0.0276, -0.0197, -0.0039, 0.0, 0.0118, -0.0039, 0.0059, 0.0099, -0.002, 0.0, 0.0059, 0.0218, 0.101, 0.0644, 0.0279, 0.0279, -0.0046, 0.0116, 0.0035, -0.0026, -0.0026, -0.0168, 0.04, 0.0543, 0.0583, 0.0014, 0.0238, 0.0177, -0.0026, 0.0522, 0.0258, 0.0035, -0.0067, 0.0197, 0.0218, 0.0563, 0.0685, 0.0766, 0.1233, 0.1233, 0.0746, 0.1193, 0.1274, 0.1416, 0.1741, 0.2899, 0.3285, 0.3752, 0.3793, 0.359, 0.3915, 0.3427, 0.2127, 0.231, 0.2066, 0.2391, 0.2411, 0.2127, 0.2086, 0.2046, 0.1518, 0.1721, 0.1883, 0.1924, 0.229, 0.2696, 0.2229, 0.3447, 0.3447, 0.3163, 0.3082, 0.3001, 0.2838, 0.3001, 0.3122, 0.2696, 0.2757, 0.2513, 0.2574, 0.2391, 0.2391, 0.3468, 0.3407, 0.3468, 0.4077, 0.4077, 0.426, 0.3955, 0.3691, 0.3691, 0.3935, 0.426, 0.5418, 0.5682, 0.5377, 0.5763, 0.6515, 0.6515, 0.6677, 0.6677, 0.6474, 0.6596, 0.6698, 0.7469, 0.7734, 0.753, 0.7469, 0.6921, 0.7551, 0.6555, 0.7449, 0.9176, 0.8526, 0.812, 0.7388, 0.7673, 0.6779, 0.686, 0.7388, 0.6637, 0.6373, 0.6941, 0.8079, 0.8079, 0.8566, 0.7998, 0.8201, 0.8688, 0.8465, 0.9034]}
{"unit_id": "reply:1923940550266454366", "ticker_idx": 0, "ts": "2025-05-18T03:15:46+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it will only accelerate the decline of iPhone sales in China, likely leading Apple to follow the same path as Samsung", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bearish AAPL: US blocking Chinese AI on iPhone accelerates China sales decline, echoing Samsung's exit trajectory.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@kyleichan Absolutely. If the U.S. blocks China’s AI from being integrated into the iPhone, it will only accelerate the decline of iPhone sales in China, likely leading Apple to follow the same path as Samsung.", "reply_text": null, "parent_text": "The US wants to block Apple from running Alibaba’s AI models on iPhones in China, according to the New York Times.\n\nI think blocking iPhones from using Chinese AI models will eventually make the iPhone irrelevant in China where Chinese smartphone makers are expanding AI features. https://t.co/rX2zXQDXY8", "quoted_text": null, "parent_author": "kyleichan", "ret_m1d": 0.011878603559677403, "ret_signed_m1d": -0.011878603559677403, "alpha_spy_m1d": -0.012971326310520004, "alpha_c_m1d": -0.010429125656985727, "bench_spy_m1d": -0.0010927227508426007, "bench_c_m1d": 0.001449477902691676, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00919621752531341, "ret_signed_p1d": 0.00919621752531341, "alpha_spy_p1d": 0.00583408947551789, "alpha_c_p1d": 0.005444219700642572, "bench_spy_p1d": -0.003362128049795521, "bench_c_p1d": -0.0037519978246708385, "ret_p1w": -0.06470921532449037, "ret_signed_p1w": 0.06470921532449037, "alpha_spy_p1w": 0.038248802628357614, "alpha_c_p1w": 0.03170981019011698, "bench_spy_p1w": -0.02646041269613275, "bench_c_p1w": -0.03299940513437338, "ret_p1m": -0.06293697827309142, "ret_signed_p1m": 0.06293697827309142, "alpha_spy_p1m": 0.06744247478580212, "alpha_c_p1m": 0.09180071949895419, "bench_spy_p1m": 0.004505496512710705, "bench_c_p1m": 0.02886374122586277, "ret_p3m": 0.11621899575728523, "ret_signed_p3m": -0.11621899575728523, "alpha_spy_p3m": -0.028790797470189533, "alpha_c_p3m": 0.027135245545651054, "bench_spy_p3m": 0.0874281982870957, "bench_c_p3m": 0.14335424130293628, "ret_p6m": 0.32114919443303447, "ret_signed_p6m": -0.32114919443303447, "alpha_spy_p6m": -0.1663735113489928, "alpha_c_p6m": -0.06822740447634468, "bench_spy_p6m": 0.15477568308404166, "bench_c_p6m": 0.2529217899566898, "price_path": [0.1713, 0.1759, 0.1746, 0.182, 0.1817, 0.1498, 0.1351, 0.1568, 0.1386, 0.1286, 0.1277, 0.1257, 0.1436, 0.0881, 0.0564, 0.0379, 0.003, 0.0212, 0.0237, 0.0174, 0.0296, 0.0241, 0.0441, 0.0559, 0.0703, 0.0597, 0.0708, 0.0423, 0.0625, 0.0676, 0.071, -0.028, -0.0989, -0.132, -0.1752, -0.0488, -0.0891, -0.0522, -0.0313, -0.0331, -0.0707, -0.0578, -0.0578, -0.076, -0.0446, -0.0213, -0.0033, 0.0011, 0.0052, 0.0103, 0.0165, 0.0204, -0.0177, -0.0486, -0.0504, -0.0612, -0.0553, -0.0503, 0.0096, 0.0199, 0.017, 0.0128, 0.0119, 0.0, -0.0092, -0.032, -0.0355, -0.0647, -0.0647, -0.041, -0.04, -0.0423, -0.038, -0.0339, -0.0264, -0.0285, -0.039, -0.0233, -0.0351, -0.0293, -0.0479, -0.0459, -0.0591, -0.0496, -0.0629, -0.0584, -0.0584, -0.0373, -0.0349, -0.0406, -0.0346, -0.0373, -0.0369, -0.0173, -0.0046, 0.0175, 0.0228, 0.0228, 0.0056, 0.0059, 0.0113, 0.0174, 0.0114, -0.0008, 0.0016, 0.0066, 0.0059, 0.0115, 0.0177, 0.0269, 0.0257, 0.0239, 0.0244, 0.0252, 0.0119, 0.0013, -0.0058, -0.0307, -0.026, -0.0281, 0.0214, 0.0539, 0.0985, 0.0894, 0.1012, 0.1189, 0.1162, 0.1105, 0.1072, 0.1056, 0.0838, 0.0784, 0.0921, 0.0893, 0.0996, 0.1052, 0.1152, 0.1132, 0.1132, 0.1015, 0.1435, 0.1498, 0.1494, 0.1407, 0.1237, 0.0875, 0.103, 0.1224, 0.135, 0.142, 0.146, 0.1407, 0.1772, 0.2279, 0.22, 0.2099, 0.2317, 0.225, 0.22, 0.221, 0.2249, 0.233, 0.2372, 0.2309, 0.2299, 0.2374, 0.2182, 0.1761, 0.1876, 0.1881, 0.1956, 0.1866, 0.2098, 0.2575, 0.26, 0.2393, 0.2447, 0.2603, 0.289, 0.2899, 0.2933, 0.3014, 0.2965, 0.2901, 0.2949, 0.2954, 0.2936, 0.2874, 0.2932, 0.3211]}
{"unit_id": "orig:1925564651032719790", "ticker_idx": 0, "ts": "2025-05-22T14:49:22+00:00", "symbol": "ASPEED Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Positive for BMC supplier ASPEED: GB200 features higher BMC content compared to GB300", "tweet_type": "news_reaction", "timeline": "months", "summary": "JPM: GB300 Bianca switch is positive for ASPEED and Foxconn, negative for Wistron on OAM/UBB board content loss.", "resolved_tickers": ["6674.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "ASPEED Technology listed on TWSE as 6674.TW (BMC supplier)", "bench_c": "EWT", "bench_c_reason": "country_fallback (suffix='TW')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "JPM’s report on GB200/GB300\n\nNvidia Returns to Bianca Design for GB300\n\nWe previously pointed out that GB300 was delayed due to changes in board design. To accelerate GB300 growth, Nvidia plans to switch back from the Cordelia board to the Bianca board and delay Cordelia’s adoption until Vera-Rubin. However, this may require additional testing and could result in delays in board-level mass production. We expect board mass production to begin between July and August, with GB300 systems entering mass production in November.\n\nThe design change for GB300 impacts various components as follows:\n\n• Positive for BMC supplier ASPEED: GB200 features higher BMC content compared to GB300 (1–2 BMCs per system for GB200, vs. 1 for GB300). Although there are concerns that GB300 compute boards may integrate HMC and SCM modules, we believe customers will stick with the original design to accelerate the ramp-up.\n\n• Negative for OAM/UBB manufacturers (Foxconn, Wistron): The Bianca board integrates two boards, resulting in lower OAM/UBB board content compared to Cordelia’s separate OAM/UBB boards. Previously, we expected Wistron to gain some market share with Cordelia, but with the switch back to Bianca for GB300, Foxconn is likely to remain the primary compute board supplier this year.\n\n• Negative for socket suppliers: The adoption of LGA GPU and memory CAMM sockets has been canceled.\n\n• Negative for cold plate and quick-disconnect (QD) suppliers: Bianca compute boards feature significantly less QD and cold plate content compared to Cordelia (approximately threefold difference).\n\n• Neutral for power supply suppliers: GB300 consumes slightly more power than GB200 (1.4kW for the B300 GPU vs. 1.2kW for the B200 GPU). However, we haven’t seen changes in PSU specifications between GB300 and GB200. Steady growth in GB racks could benefit Delta’s power and thermal revenue in the coming quarters.\n\nMeanwhile, we expect GB200 shipments to continue increasing in 2H 2025 due to improved yields. Given strong customer demand, we anticipate customers will increase GB200 orders in the second half of 2025. Overall, we project the shipment ratio of GB200/300 systems this year to be approximately 80%/20%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007211612085621599, "ret_signed_m1d": 0.007211612085621599, "alpha_spy_m1d": 0.007605891769249196, "alpha_c_m1d": 0.00683914208420533, "bench_spy_m1d": -0.00039427968362759636, "bench_c_m1d": 0.0003724700014162696, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009615329948400753, "ret_signed_p1d": -0.009615329948400753, "alpha_spy_p1d": -0.002789654082721804, "alpha_c_p1d": -0.006263322838109708, "bench_spy_p1d": -0.0068256758656789485, "bench_c_p1d": -0.0033520071102910443, "ret_p1w": 0.03605769363028055, "ret_signed_p1w": 0.03605769363028055, "alpha_spy_p1w": 0.024121310297257725, "alpha_c_p1w": 0.05151397272555214, "bench_spy_p1w": 0.011936383333022826, "bench_c_p1w": -0.015456279095271586, "ret_p1m": 0.2980769706901025, "ret_signed_p1m": 0.2980769706901025, "alpha_spy_p1m": 0.275873010783815, "alpha_c_p1m": 0.27684796382901644, "bench_spy_p1m": 0.02220395990628754, "bench_c_p1m": 0.021229006861086086, "ret_p3m": 0.37019235795066385, "ret_signed_p3m": 0.37019235795066385, "alpha_spy_p3m": 0.2696735900683227, "alpha_c_p3m": 0.2511979630557839, "bench_spy_p3m": 0.10051876788234115, "bench_c_p3m": 0.11899439489487995, "ret_p6m": 0.014423132471786282, "ret_signed_p6m": 0.014423132471786282, "alpha_spy_p6m": -0.14454865726152644, "alpha_c_p6m": -0.19917084119689532, "bench_spy_p6m": 0.15897178973331272, "bench_c_p6m": 0.2135939736686816, "price_path": [0.1538, 0.1514, 0.1346, 0.1538, 0.1538, 0.1394, 0.1611, 0.1587, 0.1514, 0.1298, 0.1394, 0.1346, 0.1274, 0.125, 0.1274, 0.1442, 0.1466, 0.1394, 0.137, 0.1274, 0.1298, 0.1322, 0.1298, 0.1082, 0.0793, 0.0048, 0.0072, 0.0096, 0.0096, 0.0096, -0.0913, -0.1803, -0.262, -0.1923, -0.2067, -0.2163, -0.1851, -0.1899, -0.1851, -0.1587, -0.1659, -0.1731, -0.1875, -0.1779, -0.1635, -0.0817, 0.0024, 0.0192, 0.0192, 0.1154, 0.1226, 0.1274, 0.0962, 0.0986, 0.1514, 0.137, 0.0817, 0.0433, 0.0649, 0.0553, 0.0361, 0.0216, 0.0072, 0.0, -0.0096, -0.0385, -0.0505, -0.0577, 0.0361, 0.0361, 0.1394, 0.1587, 0.149, 0.262, 0.2788, 0.2692, 0.2668, 0.226, 0.3486, 0.3221, 0.3149, 0.2909, 0.274, 0.2236, 0.2981, 0.262, 0.3005, 0.2861, 0.2861, 0.2933, 0.262, 0.2668, 0.274, 0.274, 0.2428, 0.2332, 0.2188, 0.2043, 0.2067, 0.2548, 0.2716, 0.3125, 0.3389, 0.3582, 0.3702, 0.2909, 0.274, 0.3053, 0.2548, 0.2788, 0.2813, 0.2981, 0.3029, 0.2909, 0.2909, 0.3365, 0.3317, 0.3197, 0.2861, 0.3125, 0.3918, 0.3966, 0.4543, 0.4183, 0.3654, 0.3293, 0.3702, 0.3317, 0.3678, 0.3125, 0.2957, 0.2284, 0.2115, 0.1635, 0.1322, 0.0962, 0.0481, 0.0529, 0.0553, 0.0361, 0.0216, 0.0192, 0.0361, 0.1394, 0.0288, -0.0216, -0.0144, -0.0144, -0.0144, -0.0096, 0.0048, 0.0144, 0.0337, 0.0361, 0.0216, 0.0216, 0.0505, 0.0288, 0.0216, 0.0144, 0.0144, 0.0072, 0.0096, 0.0168, 0.0168, 0.024, -0.0048, 0.0048, 0.0096, 0.0, -0.0288, -0.012, 0.0, -0.0192, -0.0192, -0.0168, -0.0216, -0.0288, 0.0216, -0.0024, 0.0962, 0.0072, 0.0144, -0.0048, -0.0048, -0.012, 0.0072, 0.0216, 0.0192, 0.0144]}
{"unit_id": "orig:1925564651032719790", "ticker_idx": 1, "ts": "2025-05-22T14:49:22+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn is likely to remain the primary compute board supplier this year", "tweet_type": "news_reaction", "timeline": "months", "summary": "JPM: GB300 Bianca switch is positive for ASPEED and Foxconn, negative for Wistron on OAM/UBB board content loss.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "JPM’s report on GB200/GB300\n\nNvidia Returns to Bianca Design for GB300\n\nWe previously pointed out that GB300 was delayed due to changes in board design. To accelerate GB300 growth, Nvidia plans to switch back from the Cordelia board to the Bianca board and delay Cordelia’s adoption until Vera-Rubin. However, this may require additional testing and could result in delays in board-level mass production. We expect board mass production to begin between July and August, with GB300 systems entering mass production in November.\n\nThe design change for GB300 impacts various components as follows:\n\n• Positive for BMC supplier ASPEED: GB200 features higher BMC content compared to GB300 (1–2 BMCs per system for GB200, vs. 1 for GB300). Although there are concerns that GB300 compute boards may integrate HMC and SCM modules, we believe customers will stick with the original design to accelerate the ramp-up.\n\n• Negative for OAM/UBB manufacturers (Foxconn, Wistron): The Bianca board integrates two boards, resulting in lower OAM/UBB board content compared to Cordelia’s separate OAM/UBB boards. Previously, we expected Wistron to gain some market share with Cordelia, but with the switch back to Bianca for GB300, Foxconn is likely to remain the primary compute board supplier this year.\n\n• Negative for socket suppliers: The adoption of LGA GPU and memory CAMM sockets has been canceled.\n\n• Negative for cold plate and quick-disconnect (QD) suppliers: Bianca compute boards feature significantly less QD and cold plate content compared to Cordelia (approximately threefold difference).\n\n• Neutral for power supply suppliers: GB300 consumes slightly more power than GB200 (1.4kW for the B300 GPU vs. 1.2kW for the B200 GPU). However, we haven’t seen changes in PSU specifications between GB300 and GB200. Steady growth in GB racks could benefit Delta’s power and thermal revenue in the coming quarters.\n\nMeanwhile, we expect GB200 shipments to continue increasing in 2H 2025 due to improved yields. Given strong customer demand, we anticipate customers will increase GB200 orders in the second half of 2025. Overall, we project the shipment ratio of GB200/300 systems this year to be approximately 80%/20%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01307194203486306, "ret_signed_m1d": 0.01307194203486306, "alpha_spy_m1d": 0.013466221718490656, "alpha_c_m1d": 0.013333507735429517, "bench_spy_m1d": -0.00039427968362759636, "bench_c_m1d": -0.00026156570056645734, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006536022693765409, "ret_signed_p1d": 0.006536022693765409, "alpha_spy_p1d": 0.013361698559444357, "alpha_c_p1d": 0.017567618010340902, "bench_spy_p1d": -0.0068256758656789485, "bench_c_p1d": -0.011031595316575493, "ret_p1w": 0.019607861375960933, "ret_signed_p1w": 0.019607861375960933, "alpha_spy_p1w": 0.007671478042938107, "alpha_c_p1w": 0.009622662689075634, "bench_spy_p1w": 0.011936383333022826, "bench_c_p1w": 0.0099851986868853, "ret_p1m": 0.016339953381745875, "ret_signed_p1m": 0.016339953381745875, "alpha_spy_p1m": -0.005864006524541665, "alpha_c_p1m": -0.03249581494157039, "bench_spy_p1m": 0.02220395990628754, "bench_c_p1m": 0.048835768323316264, "ret_p3m": 0.4088506470238722, "ret_signed_p3m": 0.4088506470238722, "alpha_spy_p3m": 0.30833187914153104, "alpha_c_p3m": 0.2662082780643402, "bench_spy_p3m": 0.10051876788234115, "bench_c_p3m": 0.14264236895953197, "ret_p6m": 0.6323702208305442, "ret_signed_p6m": 0.6323702208305442, "alpha_spy_p6m": 0.47339843109723145, "alpha_c_p6m": 0.37223888259645066, "bench_spy_p6m": 0.15897178973331272, "bench_c_p6m": 0.2601313382340935, "price_path": [0.2026, 0.1601, 0.1797, 0.1373, 0.1373, 0.1111, 0.1144, 0.134, 0.134, 0.1242, 0.1242, 0.1013, 0.1144, 0.098, 0.1111, 0.098, 0.0948, 0.0719, 0.085, 0.0784, 0.0654, 0.0784, 0.085, 0.049, 0.0065, -0.0458, -0.0065, 0.0033, 0.0033, 0.0033, -0.0948, -0.183, -0.2647, -0.1928, -0.1209, -0.0948, -0.0882, -0.1111, -0.1209, -0.1144, -0.1078, -0.1373, -0.0915, -0.1078, -0.0915, -0.0686, -0.0654, -0.0752, -0.0752, -0.0359, -0.0686, -0.0458, -0.0523, -0.0458, -0.0392, 0.0, 0.0327, 0.0654, 0.0425, 0.0327, 0.0065, 0.0065, 0.0131, 0.0, 0.0065, 0.0, -0.0065, -0.0065, 0.0196, 0.0196, -0.0033, -0.0098, 0.0196, 0.0163, 0.0, 0.0098, 0.0163, 0.0196, 0.0294, 0.0229, 0.0131, 0.0131, 0.0196, 0.0065, 0.0163, 0.0196, 0.049, 0.0588, 0.0654, 0.0784, 0.0523, 0.0817, 0.0973, 0.1108, 0.0905, 0.0905, 0.077, 0.0905, 0.0939, 0.0939, 0.0871, 0.1041, 0.1007, 0.1108, 0.121, 0.1176, 0.0973, 0.1244, 0.1786, 0.1819, 0.1921, 0.1616, 0.1616, 0.2057, 0.2057, 0.2192, 0.2497, 0.2598, 0.3174, 0.3174, 0.3411, 0.3343, 0.3445, 0.3513, 0.4021, 0.4224, 0.4089, 0.3581, 0.3953, 0.3716, 0.4055, 0.4156, 0.4089, 0.3953, 0.3784, 0.3445, 0.3445, 0.3581, 0.375, 0.3885, 0.3784, 0.4055, 0.4224, 0.4529, 0.4732, 0.463, 0.4597, 0.4359, 0.4563, 0.4495, 0.463, 0.4969, 0.5105, 0.5579, 0.4867, 0.4867, 0.463, 0.4834, 0.5274, 0.5342, 0.5342, 0.5477, 0.524, 0.5003, 0.5003, 0.4427, 0.3953, 0.3987, 0.5105, 0.5342, 0.6154, 0.6188, 0.6425, 0.6188, 0.6188, 0.6696, 0.6933, 0.7611, 0.7746, 0.7441, 0.7035, 0.6527, 0.673, 0.6798, 0.6527, 0.6899, 0.6696, 0.7001, 0.7069, 0.6324]}
{"unit_id": "orig:1925564651032719790", "ticker_idx": 2, "ts": "2025-05-22T14:49:22+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Negative for OAM/UBB manufacturers (Foxconn, Wistron): The Bianca board integrates two boards, resulting in lower OAM/UBB board content", "tweet_type": "news_reaction", "timeline": "months", "summary": "JPM: GB300 Bianca switch is positive for ASPEED and Foxconn, negative for Wistron on OAM/UBB board content loss.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "JPM’s report on GB200/GB300\n\nNvidia Returns to Bianca Design for GB300\n\nWe previously pointed out that GB300 was delayed due to changes in board design. To accelerate GB300 growth, Nvidia plans to switch back from the Cordelia board to the Bianca board and delay Cordelia’s adoption until Vera-Rubin. However, this may require additional testing and could result in delays in board-level mass production. We expect board mass production to begin between July and August, with GB300 systems entering mass production in November.\n\nThe design change for GB300 impacts various components as follows:\n\n• Positive for BMC supplier ASPEED: GB200 features higher BMC content compared to GB300 (1–2 BMCs per system for GB200, vs. 1 for GB300). Although there are concerns that GB300 compute boards may integrate HMC and SCM modules, we believe customers will stick with the original design to accelerate the ramp-up.\n\n• Negative for OAM/UBB manufacturers (Foxconn, Wistron): The Bianca board integrates two boards, resulting in lower OAM/UBB board content compared to Cordelia’s separate OAM/UBB boards. Previously, we expected Wistron to gain some market share with Cordelia, but with the switch back to Bianca for GB300, Foxconn is likely to remain the primary compute board supplier this year.\n\n• Negative for socket suppliers: The adoption of LGA GPU and memory CAMM sockets has been canceled.\n\n• Negative for cold plate and quick-disconnect (QD) suppliers: Bianca compute boards feature significantly less QD and cold plate content compared to Cordelia (approximately threefold difference).\n\n• Neutral for power supply suppliers: GB300 consumes slightly more power than GB200 (1.4kW for the B300 GPU vs. 1.2kW for the B200 GPU). However, we haven’t seen changes in PSU specifications between GB300 and GB200. Steady growth in GB racks could benefit Delta’s power and thermal revenue in the coming quarters.\n\nMeanwhile, we expect GB200 shipments to continue increasing in 2H 2025 due to improved yields. Given strong customer demand, we anticipate customers will increase GB200 orders in the second half of 2025. Overall, we project the shipment ratio of GB200/300 systems this year to be approximately 80%/20%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.00039427968362759636, "alpha_c_m1d": -0.00026156570056645734, "bench_spy_m1d": -0.00039427968362759636, "bench_c_m1d": -0.00026156570056645734, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": -0.0068256758656789485, "alpha_c_p1d": -0.011031595316575493, "bench_spy_p1d": -0.0068256758656789485, "bench_c_p1d": -0.011031595316575493, "ret_p1w": 0.030973442058822354, "ret_signed_p1w": -0.030973442058822354, "alpha_spy_p1w": -0.019037058725799527, "alpha_c_p1w": -0.020988243371937054, "bench_spy_p1w": 0.011936383333022826, "bench_c_p1w": 0.0099851986868853, "ret_p1m": 0.08450142293387608, "ret_signed_p1m": -0.08450142293387608, "alpha_spy_p1m": -0.06229746302758854, "alpha_c_p1m": -0.035665654610559816, "bench_spy_p1m": 0.02220395990628754, "bench_c_p1m": 0.048835768323316264, "ret_p3m": 0.05704569070770216, "ret_signed_p3m": -0.05704569070770216, "alpha_spy_p3m": 0.043473077174639, "alpha_c_p3m": 0.08559667825182982, "bench_spy_p3m": 0.10051876788234115, "bench_c_p3m": 0.14264236895953197, "ret_p6m": 0.2858434592591528, "ret_signed_p6m": -0.2858434592591528, "alpha_spy_p6m": -0.12687166952584006, "alpha_c_p6m": -0.025712121025059265, "bench_spy_p6m": 0.15897178973331272, "bench_c_p6m": 0.2601313382340935, "price_path": [0.0177, -0.0088, 0.0177, -0.0088, -0.0088, -0.0531, -0.0398, -0.0221, -0.0531, -0.0531, -0.0619, -0.0752, -0.0752, -0.0796, -0.0752, -0.0619, -0.0487, -0.0885, -0.0708, -0.0664, -0.0841, -0.0841, -0.0664, -0.0752, -0.1106, -0.1566, -0.1018, -0.1062, -0.1062, -0.1062, -0.1956, -0.2752, -0.3389, -0.2735, -0.2009, -0.1389, -0.0929, -0.1177, -0.1106, -0.0708, -0.0885, -0.1257, -0.0929, -0.1062, -0.0929, -0.0929, -0.0796, -0.1018, -0.1018, -0.0664, -0.0841, -0.0354, -0.0575, -0.0531, -0.0531, -0.0398, -0.0398, 0.0133, -0.0133, -0.0177, -0.0354, -0.031, 0.0, 0.0, 0.0, 0.0265, 0.0177, 0.0221, 0.031, 0.031, 0.0177, 0.0021, 0.0433, 0.0525, 0.0067, 0.0387, 0.0479, 0.0525, 0.0662, 0.0845, 0.0891, 0.0799, 0.0616, 0.057, 0.0845, 0.0845, 0.0982, 0.1486, 0.144, 0.1074, 0.1211, 0.0982, 0.0982, 0.1074, 0.0891, 0.0799, 0.0845, 0.1028, 0.0982, 0.0937, 0.0708, 0.0982, 0.0799, 0.0799, 0.0891, 0.0845, 0.0479, 0.0616, 0.0708, 0.0662, 0.0982, 0.1028, 0.1028, 0.1257, 0.1257, 0.0937, 0.1348, 0.1211, 0.1303, 0.1394, 0.1669, 0.1531, 0.0753, 0.0662, 0.0662, 0.0616, 0.057, 0.0387, 0.0204, 0.025, 0.0387, 0.0342, 0.0525, 0.0525, 0.0342, 0.0021, 0.0067, 0.0067, 0.0342, 0.057, 0.0479, 0.0616, 0.0982, 0.0891, 0.0982, 0.0937, 0.0799, 0.0891, 0.0982, 0.1074, 0.112, 0.1211, 0.112, 0.1486, 0.1714, 0.1714, 0.2858, 0.3362, 0.3682, 0.4048, 0.4048, 0.4277, 0.3774, 0.3819, 0.3819, 0.3316, 0.263, 0.263, 0.2767, 0.2492, 0.2996, 0.3041, 0.3133, 0.3087, 0.3087, 0.3408, 0.3545, 0.3957, 0.4185, 0.3774, 0.3133, 0.2721, 0.2584, 0.2996, 0.2584, 0.2675, 0.3225, 0.2721, 0.3179, 0.2858]}
{"unit_id": "orig:1921691715607384144", "ticker_idx": 0, "ts": "2025-05-11T22:19:42+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA RAISES GPU PRICES BY 15–20%", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares NVIDIA price hike news; implied bullish on margin/revenue uplift.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "und", "tweet_text": "🚨NVIDIA RAISES GPU PRICES BY 15–20%", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.051626008542716484, "ret_signed_m1d": -0.051626008542716484, "alpha_spy_m1d": -0.01963569737158022, "alpha_c_m1d": 0.007385873559626277, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0563414029159095, "ret_signed_p1d": 0.0563414029159095, "alpha_spy_p1d": 0.04973746259873946, "alpha_c_p1d": 0.022012643153013878, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.10219506593164018, "ret_signed_p1w": 0.10219506593164018, "alpha_spy_p1w": 0.08185176565789498, "alpha_c_p1w": 0.06597083300351914, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.1704065092705338, "ret_signed_p1m": 0.1704065092705338, "alpha_spy_p1m": 0.1359463388222879, "alpha_c_p1m": 0.06653554689432228, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.46977681802609217, "ret_signed_p3m": 0.46977681802609217, "alpha_spy_p3m": 0.38207542320790444, "alpha_c_p3m": 0.24354377888437795, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.6155690716586328, "ret_signed_p6m": 0.6155690716586328, "alpha_spy_p6m": 0.4506883865956586, "alpha_c_p6m": 0.12970069447571997, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0661, 0.0998, 0.1288, 0.1288, 0.1332, 0.1318, 0.139, 0.0928, 0.0591, 0.0294, 0.0672, -0.0233, 0.0155, -0.0728, -0.0571, -0.0464, -0.1011, -0.0839, -0.1303, -0.1159, -0.059, -0.0603, -0.0108, -0.0282, -0.0615, -0.0446, -0.0363, -0.0431, -0.0129, -0.0188, -0.0751, -0.0941, -0.1084, -0.1189, -0.1045, -0.1023, -0.1724, -0.2333, -0.2062, -0.2171, -0.0705, -0.1254, -0.0981, -0.0999, -0.0878, -0.1505, -0.1749, -0.1749, -0.2121, -0.196, -0.165, -0.1347, -0.0975, -0.116, -0.1137, -0.1145, -0.0926, -0.0691, -0.0746, -0.0769, -0.0483, -0.0458, -0.0516, 0.0, 0.0563, 0.1003, 0.0962, 0.1008, 0.1022, 0.0925, 0.0715, 0.0799, 0.0674, 0.0674, 0.1016, 0.096, 0.1316, 0.0986, 0.1169, 0.1481, 0.1538, 0.1381, 0.1522, 0.1596, 0.1704, 0.1613, 0.1789, 0.1543, 0.1764, 0.1718, 0.1828, 0.1828, 0.1696, 0.1722, 0.2025, 0.2546, 0.2604, 0.2826, 0.2846, 0.2464, 0.2785, 0.2955, 0.2955, 0.2866, 0.3009, 0.3243, 0.3342, 0.3409, 0.334, 0.3879, 0.3933, 0.4066, 0.4018, 0.3934, 0.3581, 0.3886, 0.4126, 0.4107, 0.4371, 0.427, 0.4576, 0.4462, 0.4125, 0.4635, 0.4494, 0.4588, 0.4698, 0.4855, 0.4803, 0.4892, 0.4764, 0.4799, 0.4672, 0.4799, 0.4281, 0.4261, 0.4227, 0.4472, 0.462, 0.4779, 0.4765, 0.4649, 0.4162, 0.4162, 0.3886, 0.3873, 0.3957, 0.358, 0.3685, 0.3884, 0.4418, 0.4406, 0.4459, 0.4453, 0.422, 0.3846, 0.433, 0.4365, 0.493, 0.4508, 0.439, 0.4448, 0.4489, 0.4786, 0.5171, 0.5225, 0.5359, 0.5256, 0.5086, 0.5046, 0.5377, 0.5658, 0.4893, 0.5312, 0.4638, 0.4622, 0.4783, 0.4898, 0.4851, 0.473, 0.4659, 0.4812, 0.5145, 0.557, 0.6346, 0.6835, 0.6497, 0.6465, 0.6822, 0.6156]}
{"unit_id": "orig:1933471551657292068", "ticker_idx": 0, "ts": "2025-06-13T10:28:34+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the annual growth rate in 2026 alone could exceed 150%", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley analysis implying NVDA European revenue 8x+ by 2026 with 150%+ growth rate; bullish outlook adopted.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley analysts made the following assumptions in their review report of NVIDIA's European GTC:\n\nIf Hopper chips (approximately $22,000 each) are exclusively used in the European market in 2024, and then transitioned entirely to Blackwell chips (approximately $18,000 each) in 2026, this would imply an over 8-fold increase in Nvidia's European market revenue in 2026 compared to 2024. The analyst further projected that the annual growth rate in 2026 alone could exceed 150%.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02134263768652178, "ret_signed_m1d": 0.02134263768652178, "alpha_spy_m1d": 0.010036166210859676, "alpha_c_m1d": -0.0026271983787395925, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.023969836065261374, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019159054367494432, "ret_signed_p1d": 0.019159054367494432, "alpha_spy_p1d": 0.009644857888779113, "alpha_c_p1d": -0.005705846168853279, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.02486490053634771, "ret_p1w": 0.013242320775086691, "ret_signed_p1w": 0.013242320775086691, "alpha_spy_p1w": 0.014855595770896013, "alpha_c_p1w": 0.0004791350126391869, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.012763185762447504, "ret_p1m": 0.155666819563006, "ret_signed_p1m": 0.155666819563006, "alpha_spy_p1m": 0.1059900277972956, "alpha_c_p1m": 0.04542892109910057, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.11023789846390541, "ret_p3m": 0.2490668468027053, "ret_signed_p3m": 0.2490668468027053, "alpha_spy_p3m": 0.1533581003828346, "alpha_c_p3m": 0.07847654022942518, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.17059030657328011, "ret_p6m": 0.3071128831845811, "ret_signed_p6m": 0.3071128831845811, "alpha_spy_p6m": 0.15543463149858416, "alpha_c_p6m": -0.12698969258559245, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.4341025757701735, "price_path": [-0.187, -0.1723, -0.1652, -0.171, -0.1449, -0.1499, -0.1988, -0.2152, -0.2276, -0.2367, -0.2242, -0.2223, -0.283, -0.3358, -0.3123, -0.3217, -0.1947, -0.2424, -0.2187, -0.2202, -0.2097, -0.2641, -0.2852, -0.2852, -0.3174, -0.3035, -0.2766, -0.2504, -0.2181, -0.2342, -0.2321, -0.2328, -0.2139, -0.1935, -0.1983, -0.2003, -0.1755, -0.1733, -0.1784, -0.1337, -0.0849, -0.0468, -0.0504, -0.0463, -0.0451, -0.0535, -0.0717, -0.0644, -0.0753, -0.0753, -0.0456, -0.0505, -0.0196, -0.0482, -0.0324, -0.0054, -0.0004, -0.014, -0.0018, 0.0046, 0.0139, 0.0061, 0.0213, 0.0, 0.0192, 0.0151, 0.0247, 0.0247, 0.0132, 0.0155, 0.0418, 0.0869, 0.0919, 0.1112, 0.1128, 0.0798, 0.1076, 0.1223, 0.1223, 0.1146, 0.127, 0.1473, 0.1559, 0.1617, 0.1557, 0.2024, 0.2071, 0.2186, 0.2144, 0.2072, 0.1765, 0.2029, 0.2238, 0.2221, 0.245, 0.2362, 0.2627, 0.2529, 0.2236, 0.2679, 0.2556, 0.2638, 0.2733, 0.2869, 0.2824, 0.2901, 0.2791, 0.2821, 0.271, 0.282, 0.2372, 0.2355, 0.2325, 0.2537, 0.2665, 0.2803, 0.2791, 0.2691, 0.2269, 0.2269, 0.2029, 0.2018, 0.2091, 0.1764, 0.1855, 0.2028, 0.2491, 0.248, 0.2526, 0.2521, 0.2319, 0.1995, 0.2415, 0.2445, 0.2934, 0.2569, 0.2466, 0.2517, 0.2552, 0.281, 0.3143, 0.3189, 0.3306, 0.3216, 0.307, 0.3034, 0.3321, 0.3565, 0.2902, 0.3266, 0.2682, 0.2667, 0.2807, 0.2906, 0.2865, 0.2761, 0.2699, 0.2832, 0.312, 0.3489, 0.4161, 0.4584, 0.4292, 0.4264, 0.4573, 0.3996, 0.3751, 0.3249, 0.3254, 0.4021, 0.3606, 0.3652, 0.3163, 0.3396, 0.3144, 0.2775, 0.3139, 0.2725, 0.2601, 0.2859, 0.2526, 0.2698, 0.2698, 0.2468, 0.2674, 0.2782, 0.2651, 0.2918, 0.285, 0.3071]}
{"unit_id": "orig:1935165626890469724", "ticker_idx": 0, "ts": "2025-06-18T02:40:13+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the 2028 AI market size was revised upward by 26% compared to previous estimates", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish MRVL: AI market upwardly revised 26%, ASIC market ~$41B by 2028 at ~50% CAGR plus ecosystem at ~90% CAGR.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of the Marvell Event\n\nAt the Marvell event, the 2028 AI market size was revised upward by 26% compared to previous estimates:\n• 2028 data center CAPEX: $1 trillion\n• AI accelerators (GPU + ASIC): $349 billion\n• 2028 ASIC + related ecosystem market size: $55.4 billion\n• Of that, ASIC alone: $40.8 billion (2023–2028 CAGR of approximately 50%)\n• ASIC-related ecosystem: $14.6 billion (2023–2028 CAGR of approximately 90%)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06617742597775589, "ret_signed_m1d": -0.06617742597775589, "alpha_spy_m1d": -0.06632808442647598, "alpha_c_m1d": -0.06236924915785513, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.013075462479225308, "ret_signed_p1w": 0.013075462479225308, "alpha_spy_p1w": -0.0061311130904015965, "alpha_c_p1w": -0.034946199805228506, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": -0.038439644806319784, "ret_signed_p1m": -0.038439644806319784, "alpha_spy_p1m": -0.09276575876176651, "alpha_c_p1m": -0.14933492892037248, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": -0.09959714653714336, "ret_signed_p3m": -0.09959714653714336, "alpha_spy_p3m": -0.20910393231560576, "alpha_c_p3m": -0.26936761837288603, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.19496294868851893, "ret_signed_p6m": 0.19496294868851893, "alpha_spy_p6m": 0.03480692109508299, "alpha_c_p6m": -0.21746717081217692, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [-0.0619, -0.0303, -0.0448, -0.1103, -0.1356, -0.1732, -0.1795, -0.1648, -0.1573, -0.2585, -0.3412, -0.3203, -0.3332, -0.1876, -0.2954, -0.2877, -0.3027, -0.2886, -0.3071, -0.3102, -0.3102, -0.3412, -0.3246, -0.2826, -0.2351, -0.2139, -0.2168, -0.2171, -0.2212, -0.1868, -0.1684, -0.173, -0.1832, -0.2487, -0.2311, -0.2041, -0.1394, -0.1264, -0.1207, -0.1301, -0.1493, -0.1653, -0.1805, -0.198, -0.1748, -0.1903, -0.1903, -0.1485, -0.1382, -0.1497, -0.1969, -0.1799, -0.168, -0.1154, -0.1306, -0.0881, -0.0775, -0.0815, -0.0895, -0.0708, -0.1035, -0.0604, -0.0662, 0.0, 0.0, -0.0192, -0.0556, 0.0035, 0.0131, 0.067, 0.0295, 0.0327, 0.0172, -0.0093, 0.0031, 0.0031, -0.0454, -0.04, -0.0359, -0.0212, -0.0291, -0.0318, -0.0331, -0.0539, -0.0384, -0.0032, -0.0244, -0.0387, -0.0216, -0.0113, -0.0091, 0.0136, 0.0194, 0.0915, 0.0732, -0.0059, 0.0219, 0.0233, 0.0058, 0.0128, 0.0327, 0.0319, 0.039, 0.0592, 0.0554, 0.0174, 0.0247, -0.0376, -0.049, -0.0491, -0.0252, -0.0259, -0.0084, -0.0013, 0.0313, -0.1605, -0.1605, -0.1374, -0.168, -0.1441, -0.1543, -0.1187, -0.1075, -0.104, -0.1108, -0.1007, -0.0996, -0.0805, -0.0522, -0.0088, -0.0084, 0.0086, -0.0036, 0.0695, 0.1191, 0.1106, 0.1002, 0.1226, 0.1202, 0.151, 0.1513, 0.1874, 0.1613, 0.2352, 0.2109, 0.1439, 0.1944, 0.1521, 0.1877, 0.1789, 0.1752, 0.147, 0.1259, 0.083, 0.106, 0.1241, 0.1853, 0.182, 0.2046, 0.1835, 0.2526, 0.2075, 0.1704, 0.2413, 0.2471, 0.2149, 0.2457, 0.1936, 0.1936, 0.1694, 0.1551, 0.1151, 0.0513, 0.0866, 0.0246, 0.0349, 0.1196, 0.1148, 0.1721, 0.1721, 0.1946, 0.2173, 0.2412, 0.3389, 0.312, 0.3216, 0.2293, 0.1879, 0.2356, 0.195]}
{"unit_id": "orig:1934945075118960850", "ticker_idx": 0, "ts": "2025-06-17T12:03:50+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "valuation is expected to be revised upwards… Morgan Stanley… maintained its target price of NT$1,288… TSMC's 2026 USD revenue will grow by 20% year-over-year… Leading process technology, customer stickiness, and AI trends jointly support its long-term growth prospects", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Morgan Stanley Overweight on TSMC: three uncertainties resolved, 20% 2026 revenue growth, NT$1,288 PT maintained despite TWD FX headwind; valuation re-rate expected.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC: Three Major Uncertainties Resolved, AI-Driven Growth Continues, Valuation Expected to Re-rate Under Exchange Rate Influence\n\nMorgan Stanley (C. Chan, 2025/06/15)\n\nDriven by strong and continuous AI demand and price increases, despite the appreciation of the TWD compressing gross margins, TSMC's fundamentals remain robust, with sustainable growth projected for 2026, and valuation is expected to be revised upwards.\n\nThree major uncertainties gradually resolved, TSMC stock price rebounds significantly.\n\nTSMC's stock price has risen by 31% in the past three months, mainly due to the denial of Intel JV rumors, confirmed strong AI demand, and easing tariff concerns. Despite this, some investors still question why its performance lags behind NVIDIA (which rose 53% during the same period). Morgan Stanley believes the impact of TWD appreciation on gross margin is a key influencing factor.\n\nExchange rate appreciation impacts TSMC's gross margin, potentially leading to a 6% EPS cut for 2025.\n\nManagement noted at the shareholder meeting that every 1% TWD appreciation will compress gross margin by 40bps. The current 8-10% appreciation could result in a gross margin impact of over 3%, leading to a revision of 2H25 gross margin from the original forecast of 58-59% down to 55-56%. Morgan Stanley thus lowered its 2025/26 EPS forecasts by 6% and 12% respectively, but maintained its target price of NT$1,288, pointing out that TWD appreciation has historically led to PE expansion and increased QFII holdings.\n\n2026 revenue expected to grow by 20%, Capex flat at US$40 billion.\n\nMorgan Stanley forecasts that, benefiting from 1) 30-40% growth in cloud AI semiconductor revenue; 2) Intel outsourcing NovaLake CPU and GPU to TSMC 2nm; 3) potential orders for China-specific AI GPUs (e.g., B30); 4) a global average wafer price increase of 3-5%, with US fabs seeing up to 10% increases; and 5) 2nm capacity expansion to 90kwpm, TSMC's 2026 USD revenue will grow by 20% year-over-year, with capital expenditure maintained at US$40 billion.\n\nRobust AI growth momentum, active 2nm expansion pace supports future growth.\n\nMorgan Stanley emphasizes that strong 2nm demand in 2026 will drive TSMC to continue significant capacity expansion, with full-year Capex including 20kwpm for US 3nm capacity. Although some investors worry about capacity and capital expenditure scale, TSMC's global pricing power and customer demand structure support its expansion logic. Leading process technology, customer stickiness, and AI trends jointly support its long-term growth prospects.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019138735645889193, "ret_signed_m1d": -0.019138735645889193, "alpha_spy_m1d": -0.027757415078643266, "alpha_c_m1d": -0.025981530986239676, "bench_spy_m1d": 0.008618679432754073, "bench_c_m1d": 0.006842795340350483, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009569367822944708, "ret_signed_p1d": 0.009569367822944708, "alpha_spy_p1d": 0.009720003577115888, "alpha_c_p1d": 0.00574663335425174, "bench_spy_p1d": -0.00015063575417118003, "bench_c_p1d": 0.0038227344686929676, "ret_p1w": 0.004784683911472243, "ret_signed_p1w": 0.004784683911472243, "alpha_spy_p1w": -0.013697641235079994, "alpha_c_p1w": -0.03325191023410112, "bench_spy_p1w": 0.018482325146552236, "bench_c_p1w": 0.038036594145573366, "ret_p1m": 0.08133962649502924, "ret_signed_p1m": 0.08133962649502924, "alpha_spy_p1m": 0.03358426970474038, "alpha_c_p1m": -0.024512997998370034, "bench_spy_p1m": 0.04775535679028886, "bench_c_p1m": 0.10585262449339927, "ret_p3m": 0.2057417623607385, "ret_signed_p3m": 0.2057417623607385, "alpha_spy_p3m": 0.10227679906968001, "alpha_c_p3m": 0.042394467180010054, "bench_spy_p3m": 0.10346496329105848, "bench_c_p3m": 0.16334729518072844, "ret_p6m": 0.4459521614247919, "ret_signed_p6m": 0.4459521614247919, "alpha_spy_p6m": 0.288663844694625, "alpha_c_p6m": 0.01585158556880173, "bench_spy_p6m": 0.15728831673016685, "bench_c_p6m": 0.43010057585599015, "price_path": [-0.0614, -0.0738, -0.0738, -0.0566, -0.0662, -0.0871, -0.0928, -0.1329, -0.1005, -0.1024, -0.1024, -0.1024, -0.1919, -0.2224, -0.252, -0.1777, -0.1529, -0.1757, -0.1643, -0.1853, -0.1929, -0.19, -0.2043, -0.2224, -0.1681, -0.1767, -0.1538, -0.1453, -0.1405, -0.1348, -0.1348, -0.0948, -0.1062, -0.1233, -0.1157, -0.1252, -0.0957, -0.0881, -0.0766, -0.0481, -0.0538, -0.049, -0.0624, -0.0681, -0.0566, -0.0643, -0.0624, -0.07, -0.0805, -0.0786, -0.0786, -0.0786, -0.0986, -0.0948, -0.0566, -0.049, -0.0519, -0.0423, -0.0042, 0.0148, 0.0, -0.0144, -0.0191, 0.0, 0.0096, -0.0096, 0.0096, -0.0239, 0.0048, 0.0239, 0.0287, 0.0335, 0.0144, 0.0383, 0.0383, 0.0431, 0.0383, 0.0335, 0.0335, 0.0431, 0.0526, 0.0526, 0.0478, 0.0622, 0.0813, 0.0813, 0.1053, 0.1005, 0.0813, 0.0957, 0.0957, 0.0957, 0.0957, 0.0861, 0.1053, 0.11, 0.11, 0.0861, 0.1005, 0.0766, 0.1292, 0.1244, 0.1292, 0.1292, 0.1483, 0.1244, 0.1292, 0.1292, 0.134, 0.0861, 0.1005, 0.0861, 0.1196, 0.1244, 0.1388, 0.11, 0.11, 0.1148, 0.11, 0.11, 0.11, 0.1292, 0.1292, 0.1483, 0.1722, 0.1866, 0.2057, 0.201, 0.2298, 0.2154, 0.2346, 0.2154, 0.2442, 0.2874, 0.2874, 0.2682, 0.249, 0.249, 0.2538, 0.273, 0.3114, 0.3451, 0.3451, 0.3787, 0.3595, 0.3835, 0.3835, 0.3595, 0.3691, 0.4075, 0.4267, 0.3931, 0.4219, 0.4219, 0.4027, 0.3931, 0.3931, 0.4219, 0.4171, 0.446, 0.446, 0.4411, 0.4508, 0.446, 0.4027, 0.4075, 0.4027, 0.4171, 0.4075, 0.4171, 0.4027, 0.3739, 0.3883, 0.3499, 0.3403, 0.3979, 0.3307, 0.3211, 0.3595, 0.3835, 0.3787, 0.3835, 0.3547, 0.3739, 0.3931, 0.3883, 0.4027, 0.4363, 0.4219, 0.446]}
{"unit_id": "thread:1925371305916080451", "ticker_idx": 0, "ts": "2025-05-22T02:01:05+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still don't recommend Qualcomm stock. It's way overvalued.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Short QCOM: overvalued and facing 10-15% chip price erosion as Xiaomi's in-house Xring chip pressures Snapdragon pricing, similar to Apple's modem displacement.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "I don’t understand why my friends are so dismissive of Xiaomi’s in-house chip. They say things like, “Qualcomm signed a multi-year deal with Xiaomi, so what’s the point of Xiaomi’s own chip?”\n\nBut this is what I want to say. Qualcomm said something similar about Apple’s modem.\n\nIn an interview this past March, Cristiano Amon emphasized that Qualcomm would continue supplying modems to Apple. But unfortunately, starting with the iPhone 18, it’s highly likely that almost all models will feature Apple’s own modem.\n\nI believe what Amon says is just meant to reassure shareholders.\n\nTo say that Xiaomi’s in-house chip has no impact is obviously a lie. Why else would Xiaomi push so hard to reach flagship-level performance? It’s to lower the price they pay for Qualcomm’s flagship Snapdragon chips.\n\nTake Samsung as an example—by using Exynos, Samsung is estimated to have saved as much as $20–30 per chip compared to the prices Chinese OEMs pay for Qualcomm chips.\n\nThat’s clearly what Xiaomi wants too. I expect Xiaomi will be able to cut Qualcomm’s chip prices by about 10–15% through Xring. In other words, Qualcomm will earn that much less.\n\nI still don’t recommend Qualcomm stock. It’s way overvalued.\n\n---\n\nLink: \n1. https://t.co/ORB5hJWtTB\n\n2. \nhttps://t.co/Cp4tHh2AnV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026735398080328432, "ret_signed_m1d": -0.026735398080328432, "alpha_spy_m1d": -0.027129677763956028, "alpha_c_m1d": -0.02399462198050628, "bench_spy_m1d": -0.00039427968362759636, "bench_c_m1d": 0.002740776099822151, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013503336414491751, "ret_signed_p1d": 0.013503336414491751, "alpha_spy_p1d": 0.006677660548812803, "alpha_c_p1d": -0.0004910841484478956, "bench_spy_p1d": -0.0068256758656789485, "bench_c_p1d": -0.013994420562939647, "ret_p1w": 0.0065819095559391805, "ret_signed_p1w": -0.0065819095559391805, "alpha_spy_p1w": 0.005354473777083646, "alpha_c_p1w": 0.006997263088166061, "bench_spy_p1w": 0.011936383333022826, "bench_c_p1w": 0.013579172644105242, "ret_p1m": 0.0329714323317325, "ret_signed_p1m": -0.0329714323317325, "alpha_spy_p1m": -0.010767472425444957, "alpha_c_p1m": 0.04783913058884659, "bench_spy_p1m": 0.02220395990628754, "bench_c_p1m": 0.08081056292057909, "ret_p3m": 0.06662539616974894, "ret_signed_p3m": -0.06662539616974894, "alpha_spy_p3m": 0.03389337171259221, "alpha_c_p3m": 0.14125636908837147, "bench_spy_p3m": 0.10051876788234115, "bench_c_p3m": 0.2078817652581204, "ret_p6m": 0.19441636032577136, "ret_signed_p6m": -0.19441636032577136, "alpha_spy_p6m": -0.035444570592458646, "alpha_c_p6m": 0.23816544186163746, "bench_spy_p6m": 0.15897178973331272, "bench_c_p6m": 0.4325818021874088, "price_path": [0.0873, 0.0881, 0.0941, 0.0424, 0.0608, 0.0368, 0.0425, 0.0697, 0.0588, 0.094, 0.0516, 0.0389, 0.0368, 0.031, 0.0625, 0.0718, 0.0605, 0.0739, 0.0718, 0.0641, 0.0862, 0.0867, 0.0748, 0.0719, 0.0363, 0.0423, 0.0385, 0.0455, -0.0539, -0.1351, -0.1198, -0.1541, -0.0256, -0.088, -0.0551, -0.0555, -0.0595, -0.0789, -0.0727, -0.0727, -0.0762, -0.0586, -0.0469, -0.0011, 0.0081, 0.0017, -0.0033, 0.0074, -0.0825, -0.0513, -0.0538, -0.0507, -0.0208, -0.0154, -0.0149, 0.0322, 0.0269, 0.0381, 0.0356, 0.0348, 0.043, 0.0438, 0.0267, 0.0, -0.0135, -0.0135, 0.0086, 0.0016, 0.0066, -0.0147, -0.005, 0.0107, 0.0114, 0.0073, 0.0188, 0.0609, 0.0863, 0.0887, 0.0834, 0.0562, 0.0709, 0.0544, 0.0487, 0.0487, 0.033, 0.0454, 0.0629, 0.0644, 0.0799, 0.0823, 0.0872, 0.0881, 0.1081, 0.1073, 0.1073, 0.0792, 0.0885, 0.0878, 0.086, 0.0749, 0.0532, 0.0533, 0.0517, 0.0418, 0.0567, 0.0852, 0.0785, 0.0914, 0.0843, 0.0813, 0.0994, 0.1064, 0.0858, 0.0018, 0.0116, 0.007, 0.0015, -0.0044, -0.004, 0.0073, 0.0101, 0.0494, 0.0689, 0.0792, 0.0775, 0.0847, 0.0666, 0.0611, 0.0522, 0.0786, 0.0678, 0.0866, 0.0907, 0.0977, 0.0972, 0.0972, 0.0839, 0.0737, 0.0964, 0.0973, 0.1001, 0.0892, 0.0912, 0.1088, 0.111, 0.1068, 0.1269, 0.1346, 0.1543, 0.1455, 0.1652, 0.1639, 0.1915, 0.1649, 0.1616, 0.1348, 0.1421, 0.143, 0.1592, 0.1615, 0.1576, 0.1359, 0.1518, 0.1373, 0.0544, 0.1107, 0.1104, 0.1188, 0.1265, 0.1221, 0.1468, 0.1591, 0.1621, 0.1673, 0.1598, 0.2885, 0.2428, 0.2266, 0.2169, 0.2419, 0.2407, 0.1866, 0.2338, 0.1891, 0.1732, 0.1779, 0.1944, 0.2129, 0.198, 0.1944]}
{"unit_id": "thread:1932634155944480991", "ticker_idx": 0, "ts": "2025-06-11T03:01:04+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's server market share surged further to 39.4%, with expectations to easily surpass 40% for the full year… AMD is confident it will capture 50% of the server processor market in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "AMD gaining server share (39.4% → 50% by 2026) at Intel's expense as supply chains shift and TSMC partnership widens process advantage.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD nears 40% server market share in 1Q25, poised to match Intel in 2026\n\nIn 2017, AMD held nearly zero market share in the server processor segment. After CEO Lisa Su implemented a comprehensive platform strategy overhaul, AMD set ambitious goals for annual market share growth. This effort led to reaching 25% in 2023 and soaring to a new high of 33% in 2024. In the first quarter of 2025, AMD's server market share surged further to 39.4%, with expectations to easily surpass 40% for the full year.\n\nSupply chain sources reveal that AMD is confident it will capture 50% of the server processor market in 2026, establishing a competitive balance with Intel.\n\nRecord-breaking quarter across all segments\nAccording to Mercury Research's first quarter 2025 report, AMD has demonstrated strong growth in cloud and enterprise markets, achieving a record 39.4% revenue market share in servers. Additionally, driven by Ryzen channel sales and AI PC demand, its client-side market share also shows robust expansion.\n\nAMD's server revenue market share in the first quarter of 2025 increased by 6.5% year-over-year and 3.1% quarter-over-quarter, reaching 39.4%. Client product revenue market share grew 10.2% year-over-year and 2.7% quarter-over-quarter, hitting 26.5%.\n\nDesktop product revenue market share rose sharply by 15.2% year-over-year and 6.4% quarter-over-quarter. Notebook product revenue market share increased by 7.3% year-over-year and 0.5% quarter-over-quarter. Overall revenue market share expanded by 9% compared to the same period in 2024 and 2.9% quarter-over-quarter, reaching 31.6%.\n\nFinancial performance exceeds expectations\nAMD reported revenue of US$7.44 billion in the first quarter of 2025, up 36% year-over-year, with net income soaring to US$709 million from US$123 million in the same quarter of 2024. Gross margin stood at 50%. The data center business unit generated US$3.7 billion in revenue, growing 57% year-over-year, fueled by increased sales of Epyc CPUs and Instinct GPUs.\n\nGrowth across server, notebook, and desktop platform processors has been key to AMD's revenue increase.\n\nCautious optimism amid geopolitical headwinds\nLooking ahead to the second quarter and full year of 2025, AMD expects second-quarter revenue to remain roughly the same as the first quarter while maintaining an overall growth trajectory. However, AMD cautioned that US AI chip export restrictions on China could reduce its 2025 revenue by US$1.5 billion.\n\nSince 2019, with TSMC's support, AMD has significantly improved performance, yield, and time-to-market across server and other platforms. Meanwhile, Intel continues to face supply shortages due to process delays, prompting supply chains to increasingly favor AMD products.\n\nThe path to processor parity\nIndustry sources note that AMD's server platform emphasizes high cost-performance, gaining increasing market acceptance. Its market share has climbed from near zero to almost 40% and continues to rise. Expansion in notebook and desktop segments is breaking Intel's near-monopoly, with the potential to reach parity by 2026 and possibly surpass Intel within three years. Beyond AMD's technological and strategic enhancements, a critical factor is its partnership with TSMC for advanced 7-nanometer and below manufacturing processes.\n\nOn June 13, 2025, Taiwan time, AMD will hold its \"AMD Advancing AI 2025\" conference in San Jose, featuring participation from major customers and supply chain leaders.\n\n$AMD $INTC\n\n---\n\nhttps://t.co/j3tjKm3qLE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017335302003481212, "ret_signed_m1d": 0.017335302003481212, "alpha_spy_m1d": 0.014475230733708155, "alpha_c_m1d": 0.016609732971847535, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0007255690316336771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02179296188665847, "ret_signed_p1d": -0.02179296188665847, "alpha_spy_p1d": -0.025767187983812945, "alpha_c_p1d": -0.026642479664385488, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.004849517777727019, "ret_p1w": 0.046640263780302726, "ret_signed_p1w": 0.046640263780302726, "alpha_spy_p1w": 0.05315878299842125, "alpha_c_p1w": 0.04392911144412115, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.002711152336181577, "ret_p1m": 0.1900281029259112, "ret_signed_p1m": 0.1900281029259112, "alpha_spy_p1m": 0.14627712253536473, "alpha_c_p1m": 0.09212070386352833, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.09790739906238288, "ret_p3m": 0.2498762128526768, "ret_signed_p3m": 0.2498762128526768, "alpha_spy_p3m": 0.16774894809918495, "alpha_c_p3m": 0.11760195079799662, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.13227426205468018, "ret_p6m": 0.782895796728845, "ret_signed_p6m": 0.782895796728845, "alpha_spy_p6m": 0.6382965064700055, "alpha_c_p6m": 0.40211069292875457, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.38078510380009045, "price_path": [-0.1665, -0.1366, -0.1455, -0.1231, -0.1156, -0.1213, -0.0602, -0.0523, -0.0904, -0.1196, -0.1479, -0.1519, -0.1516, -0.1501, -0.2257, -0.2921, -0.3096, -0.3544, -0.2006, -0.2678, -0.229, -0.2199, -0.2134, -0.2712, -0.2777, -0.2777, -0.2937, -0.2879, -0.2538, -0.2202, -0.2022, -0.2043, -0.207, -0.1964, -0.2022, -0.1844, -0.1696, -0.1859, -0.1715, -0.1605, -0.1511, -0.1075, -0.0717, -0.0282, -0.0508, -0.0328, -0.0528, -0.063, -0.075, -0.0861, -0.0894, -0.0894, -0.0543, -0.0684, -0.0669, -0.0859, -0.0537, -0.0316, -0.0211, -0.045, -0.0409, 0.0049, 0.0173, 0.0, -0.0218, -0.0411, 0.0433, 0.0492, 0.0466, 0.0466, 0.0586, 0.0697, 0.1427, 0.1838, 0.1861, 0.1871, 0.1714, 0.1236, 0.1435, 0.1384, 0.1384, 0.1128, 0.1377, 0.1426, 0.19, 0.2087, 0.2072, 0.2845, 0.3214, 0.3242, 0.2959, 0.296, 0.2772, 0.3096, 0.3383, 0.3742, 0.4335, 0.4648, 0.4818, 0.4554, 0.4174, 0.4593, 0.4389, 0.3465, 0.4231, 0.4261, 0.4222, 0.4442, 0.5224, 0.4937, 0.4653, 0.454, 0.3749, 0.3637, 0.3514, 0.3848, 0.3485, 0.3754, 0.3796, 0.3916, 0.3425, 0.3425, 0.3399, 0.3384, 0.3356, 0.2476, 0.2499, 0.2863, 0.317, 0.285, 0.309, 0.3304, 0.3246, 0.3139, 0.3036, 0.2992, 0.3191, 0.3282, 0.3281, 0.3313, 0.3163, 0.332, 0.3356, 0.3539, 0.4011, 0.3593, 0.6816, 0.746, 0.9445, 0.9225, 0.774, 0.7865, 0.8003, 0.9696, 0.9363, 0.9241, 0.9858, 0.9649, 0.9005, 0.9398, 1.0878, 1.1436, 1.1298, 1.182, 1.1037, 1.1142, 1.1434, 1.0641, 1.116, 0.9622, 0.9279, 1.014, 0.9607, 1.1371, 1.0469, 1.0374, 0.9855, 0.901, 0.8454, 0.7007, 0.6822, 0.7752, 0.7016, 0.7685, 0.7685, 0.7957, 0.8141, 0.7768, 0.7963, 0.7829]}
{"unit_id": "thread:1932634155944480991", "ticker_idx": 1, "ts": "2025-06-11T03:01:04+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel continues to face supply shortages due to process delays, prompting supply chains to increasingly favor AMD products… breaking Intel's near-monopoly", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "AMD gaining server share (39.4% → 50% by 2026) at Intel's expense as supply chains shift and TSMC partnership widens process advantage.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD nears 40% server market share in 1Q25, poised to match Intel in 2026\n\nIn 2017, AMD held nearly zero market share in the server processor segment. After CEO Lisa Su implemented a comprehensive platform strategy overhaul, AMD set ambitious goals for annual market share growth. This effort led to reaching 25% in 2023 and soaring to a new high of 33% in 2024. In the first quarter of 2025, AMD's server market share surged further to 39.4%, with expectations to easily surpass 40% for the full year.\n\nSupply chain sources reveal that AMD is confident it will capture 50% of the server processor market in 2026, establishing a competitive balance with Intel.\n\nRecord-breaking quarter across all segments\nAccording to Mercury Research's first quarter 2025 report, AMD has demonstrated strong growth in cloud and enterprise markets, achieving a record 39.4% revenue market share in servers. Additionally, driven by Ryzen channel sales and AI PC demand, its client-side market share also shows robust expansion.\n\nAMD's server revenue market share in the first quarter of 2025 increased by 6.5% year-over-year and 3.1% quarter-over-quarter, reaching 39.4%. Client product revenue market share grew 10.2% year-over-year and 2.7% quarter-over-quarter, hitting 26.5%.\n\nDesktop product revenue market share rose sharply by 15.2% year-over-year and 6.4% quarter-over-quarter. Notebook product revenue market share increased by 7.3% year-over-year and 0.5% quarter-over-quarter. Overall revenue market share expanded by 9% compared to the same period in 2024 and 2.9% quarter-over-quarter, reaching 31.6%.\n\nFinancial performance exceeds expectations\nAMD reported revenue of US$7.44 billion in the first quarter of 2025, up 36% year-over-year, with net income soaring to US$709 million from US$123 million in the same quarter of 2024. Gross margin stood at 50%. The data center business unit generated US$3.7 billion in revenue, growing 57% year-over-year, fueled by increased sales of Epyc CPUs and Instinct GPUs.\n\nGrowth across server, notebook, and desktop platform processors has been key to AMD's revenue increase.\n\nCautious optimism amid geopolitical headwinds\nLooking ahead to the second quarter and full year of 2025, AMD expects second-quarter revenue to remain roughly the same as the first quarter while maintaining an overall growth trajectory. However, AMD cautioned that US AI chip export restrictions on China could reduce its 2025 revenue by US$1.5 billion.\n\nSince 2019, with TSMC's support, AMD has significantly improved performance, yield, and time-to-market across server and other platforms. Meanwhile, Intel continues to face supply shortages due to process delays, prompting supply chains to increasingly favor AMD products.\n\nThe path to processor parity\nIndustry sources note that AMD's server platform emphasizes high cost-performance, gaining increasing market acceptance. Its market share has climbed from near zero to almost 40% and continues to rise. Expansion in notebook and desktop segments is breaking Intel's near-monopoly, with the potential to reach parity by 2026 and possibly surpass Intel within three years. Beyond AMD's technological and strategic enhancements, a critical factor is its partnership with TSMC for advanced 7-nanometer and below manufacturing processes.\n\nOn June 13, 2025, Taiwan time, AMD will hold its \"AMD Advancing AI 2025\" conference in San Jose, featuring participation from major customers and supply chain leaders.\n\n$AMD $INTC\n\n---\n\nhttps://t.co/j3tjKm3qLE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06769823974228295, "ret_signed_m1d": -0.06769823974228295, "alpha_spy_m1d": -0.0648381684725099, "alpha_c_m1d": -0.06697267071064927, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0007255690316336771, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004352038262076974, "ret_signed_p1d": -0.004352038262076974, "alpha_spy_p1d": -0.0003778121649224975, "alpha_c_p1d": 0.0004974795156500456, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.004849517777727019, "ret_p1w": 0.03916825212713637, "ret_signed_p1w": -0.03916825212713637, "alpha_spy_p1w": -0.045686771345254895, "alpha_c_p1w": -0.036457099790954794, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.002711152336181577, "ret_p1m": 0.15183749242317757, "ret_signed_p1m": -0.15183749242317757, "alpha_spy_p1m": -0.10808651203263109, "alpha_c_p1m": -0.053930093360794684, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.09790739906238288, "ret_p3m": 0.18375237819070467, "ret_signed_p3m": -0.18375237819070467, "alpha_spy_p3m": -0.10162511343721281, "alpha_c_p3m": -0.05147811613602449, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.13227426205468018, "ret_p6m": 0.9584138975986223, "ret_signed_p6m": -0.9584138975986223, "alpha_spy_p6m": -0.8138146073397827, "alpha_c_p6m": -0.5776287937985318, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.38078510380009045, "price_path": [0.163, 0.2423, 0.2534, 0.1663, 0.1586, 0.1731, 0.1712, 0.1702, 0.1325, 0.1422, 0.0982, 0.0982, 0.0662, 0.0629, 0.0846, -0.0401, -0.0537, -0.1233, 0.0411, -0.0387, -0.0455, -0.0179, -0.0401, -0.0701, -0.0846, -0.0846, -0.089, -0.0566, -0.0044, 0.0392, -0.0305, -0.0082, -0.0164, -0.028, -0.0338, -0.0029, -0.0198, -0.0358, -0.0179, 0.0155, 0.0358, 0.0725, 0.0909, 0.0406, 0.0421, 0.0474, 0.0338, 0.0285, 0.0005, -0.0063, -0.0305, -0.0305, -0.0063, -0.015, -0.0208, -0.0546, -0.0455, -0.0189, -0.0208, -0.0334, -0.03, -0.0097, 0.0677, 0.0, 0.0044, -0.0261, 0.0029, 0.0058, 0.0392, 0.0392, 0.0193, 0.0247, 0.0904, 0.0735, 0.088, 0.0972, 0.0832, 0.1049, 0.058, 0.0875, 0.0875, 0.0638, 0.1407, 0.1335, 0.1518, 0.133, 0.1267, 0.1083, 0.0972, 0.1025, 0.117, 0.1248, 0.1238, 0.1359, 0.0943, 0.001, 0.0, -0.0131, -0.0164, -0.0426, -0.0662, -0.0571, -0.0237, -0.0131, -0.044, -0.0353, -0.0015, 0.0546, 0.0745, 0.1538, 0.1876, 0.1441, 0.2239, 0.1383, 0.1364, 0.1992, 0.1871, 0.1775, 0.2016, 0.2055, 0.1775, 0.1775, 0.1707, 0.1605, 0.19, 0.1842, 0.1838, 0.1818, 0.1978, 0.19, 0.1644, 0.1978, 0.222, 0.2041, 0.4782, 0.4304, 0.3907, 0.4188, 0.5097, 0.6436, 0.7166, 0.6673, 0.6223, 0.7379, 0.8037, 0.7809, 0.7693, 0.7974, 0.81, 0.8279, 0.7587, 0.7998, 0.7229, 0.7964, 0.7814, 0.7897, 0.8424, 0.8433, 0.7853, 0.8453, 0.8511, 0.912, 1.0082, 0.999, 0.942, 0.9338, 0.9101, 0.7906, 0.8559, 0.8008, 0.8438, 0.8593, 0.8317, 0.8322, 0.7365, 0.7176, 0.6784, 0.6601, 0.6978, 0.6257, 0.6683, 0.7307, 0.7326, 0.78, 0.78, 0.9613, 0.9347, 1.102, 1.1161, 0.9584]}
{"unit_id": "orig:1930965974037221876", "ticker_idx": 0, "ts": "2025-06-06T12:32:18+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our \"Buy\" ratings on both SK Hynix and Samsung Electronics (SEC).", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman reiterates Buy on SK Hynix and Samsung on accelerating DRAM ASP growth from HBM3E ramp and rising NAND prices in Q2-Q3 2025.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Korea Technology: May 2025 Memory Price Trends\n\nQ2 Price Outlook Largely Maintained; Commodity DRAM in Line with GSe Forecasts\n\nCompared to last month, TrendForce’s outlook for Q2 2025 prices (Commodity DRAM: +3–8% QoQ; NAND: +3–8%) remains largely unchanged. When compared with GSe (Goldman Sachs estimates), the commodity DRAM forecast is broadly in line, while the NAND outlook is more optimistic. Although the outlook for Q3 2025 remains fluid, we expect Korean companies to see an acceleration in DRAM ASP growth, driven by the full-scale ramp-up of 12-layer HBM3E production. Meanwhile, the price increase for commodity DRAM may moderate. We maintain our “Buy” ratings on both SK Hynix and Samsung Electronics (SEC).\n\nDRAM:\n• Q2: Despite a sharp price spike in legacy DDR4 (which has a smaller market share), PC and server DRAM prices are expected to rise 3–8% overall. This has narrowed the price gap with next-generation DDR5.\n• Q3: As high-value HBM3E enters full-scale production, overall DRAM ASP growth is expected to accelerate. However, the price increase for commodity DRAM could slightly decelerate.\n\nNAND Flash:\n• Q2: Prices are expected to rise 3–8% QoQ. Notably, outlooks for PC and mobile NAND have been revised upward.\n• Q3: Driven by solid demand for server SSDs, the projected price increase has been revised up from 0–5% to 5–10%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.010164599788460738, "alpha_c_m1d": 0.005658225860998067, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020044577207391834, "ret_signed_p1d": 0.020044577207391834, "alpha_spy_p1d": 0.01914329431924755, "alpha_c_p1d": 0.003307430524006394, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": 0.04899788645804648, "ret_signed_p1w": 0.04899788645804648, "alpha_spy_p1w": 0.052569633346652433, "alpha_c_p1w": 0.032142034888574145, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": 0.20712704154656958, "ret_signed_p1m": 0.20712704154656958, "alpha_spy_p1m": 0.16811294013240974, "alpha_c_p1m": 0.10017490820173713, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.170953960764761, "ret_signed_p3m": 0.170953960764761, "alpha_spy_p3m": 0.0933376051645094, "alpha_c_p3m": 0.037610097419763955, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 1.4016165824807283, "ret_signed_p6m": 1.4016165824807283, "alpha_spy_p6m": 1.2596919875144967, "alpha_c_p6m": 1.0050273130148228, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.165, -0.1156, -0.1121, -0.0907, -0.0841, -0.0974, -0.0863, -0.0663, -0.0418, -0.0596, -0.0752, -0.0485, -0.0796, -0.1138, -0.1521, -0.1241, -0.1201, -0.1347, -0.1899, -0.2672, -0.2463, -0.2664, -0.1854, -0.1961, -0.1988, -0.197, -0.2263, -0.2219, -0.2219, -0.2148, -0.2272, -0.1952, -0.2072, -0.1801, -0.1908, -0.1961, -0.2108, -0.2108, -0.173, -0.173, -0.173, -0.1516, -0.1539, -0.1548, -0.133, -0.1174, -0.0841, -0.1085, -0.0907, -0.1134, -0.1018, -0.1085, -0.1245, -0.1107, -0.0974, -0.0996, -0.0752, -0.0557, -0.0891, -0.0757, -0.0757, -0.0312, 0.0, 0.0, 0.02, 0.0267, 0.069, 0.049, 0.049, 0.1047, 0.1091, 0.098, 0.0958, 0.1448, 0.1559, 0.2405, 0.2739, 0.3051, 0.265, 0.3007, 0.2717, 0.2428, 0.2405, 0.2049, 0.2071, 0.2561, 0.2517, 0.3229, 0.3118, 0.3363, 0.3296, 0.3185, 0.2004, 0.1982, 0.2138, 0.196, 0.1982, 0.2004, 0.1849, 0.167, 0.1693, 0.1737, 0.2183, 0.1492, 0.1492, 0.1737, 0.1514, 0.167, 0.1425, 0.1893, 0.1982, 0.2383, 0.2316, 0.2316, 0.1915, 0.1715, 0.1381, 0.0913, 0.118, 0.1559, 0.1648, 0.1581, 0.1977, 0.1999, 0.142, 0.162, 0.171, 0.1843, 0.22, 0.2356, 0.2847, 0.3561, 0.3695, 0.4654, 0.4765, 0.5524, 0.4877, 0.5858, 0.5858, 0.5657, 0.6103, 0.5947, 0.5903, 0.5011, 0.5568, 0.5501, 0.6059, 0.7642, 0.7642, 0.7642, 0.7642, 0.7642, 0.7642, 0.9092, 0.8512, 0.8356, 0.8847, 1.0185, 1.0765, 1.1657, 1.1367, 1.1479, 1.1345, 1.275, 1.3865, 1.3241, 1.4891, 1.5337, 1.4936, 1.7657, 1.614, 1.5828, 1.6452, 1.5873, 1.7032, 1.7612, 1.7523, 1.73, 1.498, 1.7032, 1.5426, 1.507, 1.5471, 1.3241, 1.3196, 1.3151, 1.3374, 1.4284, 1.3659, 1.4016]}
{"unit_id": "orig:1930965974037221876", "ticker_idx": 1, "ts": "2025-06-06T12:32:18+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our \"Buy\" ratings on both SK Hynix and Samsung Electronics (SEC).", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman reiterates Buy on SK Hynix and Samsung on accelerating DRAM ASP growth from HBM3E ramp and rising NAND prices in Q2-Q3 2025.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Korea Technology: May 2025 Memory Price Trends\n\nQ2 Price Outlook Largely Maintained; Commodity DRAM in Line with GSe Forecasts\n\nCompared to last month, TrendForce’s outlook for Q2 2025 prices (Commodity DRAM: +3–8% QoQ; NAND: +3–8%) remains largely unchanged. When compared with GSe (Goldman Sachs estimates), the commodity DRAM forecast is broadly in line, while the NAND outlook is more optimistic. Although the outlook for Q3 2025 remains fluid, we expect Korean companies to see an acceleration in DRAM ASP growth, driven by the full-scale ramp-up of 12-layer HBM3E production. Meanwhile, the price increase for commodity DRAM may moderate. We maintain our “Buy” ratings on both SK Hynix and Samsung Electronics (SEC).\n\nDRAM:\n• Q2: Despite a sharp price spike in legacy DDR4 (which has a smaller market share), PC and server DRAM prices are expected to rise 3–8% overall. This has narrowed the price gap with next-generation DDR5.\n• Q3: As high-value HBM3E enters full-scale production, overall DRAM ASP growth is expected to accelerate. However, the price increase for commodity DRAM could slightly decelerate.\n\nNAND Flash:\n• Q2: Prices are expected to rise 3–8% QoQ. Notably, outlooks for PC and mobile NAND have been revised upward.\n• Q3: Driven by solid demand for server SSDs, the projected price increase has been revised up from 0–5% to 5–10%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.010164599788460738, "alpha_c_m1d": 0.009355607467128713, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.009355607467128713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011844390025516294, "ret_signed_p1d": 0.011844390025516294, "alpha_spy_p1d": 0.01094310713737201, "alpha_c_p1d": 0.006935874945164233, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.0049085150803520605, "ret_p1w": -0.013536349655783209, "ret_signed_p1w": -0.013536349655783209, "alpha_spy_p1w": -0.009964602767177255, "alpha_c_p1w": -0.01693460192800189, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.0033982522722186825, "ret_p1m": 0.050396855504293514, "ret_signed_p1m": 0.050396855504293514, "alpha_spy_p1m": 0.011382754090133673, "alpha_c_p1m": -0.020850426627838647, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.07124728213213216, "ret_p3m": 0.1882933477972173, "ret_signed_p3m": 0.1882933477972173, "alpha_spy_p3m": 0.1106769921969657, "alpha_c_p3m": 0.09002544373399357, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.09826790406322372, "ret_p6m": 0.7237018825648858, "ret_signed_p6m": 0.7237018825648858, "alpha_spy_p6m": 0.5817772875986542, "alpha_c_p6m": 0.5188302066174015, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.20487167594748423, "price_path": [-0.0984, -0.0766, -0.0799, -0.0799, -0.0311, -0.0311, -0.016, 0.0126, 0.0378, 0.0176, 0.0059, 0.0328, 0.0395, 0.0186, -0.022, -0.0051, -0.0051, -0.0254, -0.0508, -0.0998, -0.0948, -0.1032, -0.0457, -0.066, -0.0491, -0.0423, -0.0745, -0.0677, -0.0643, -0.0626, -0.0694, -0.0575, -0.0575, -0.0575, -0.0558, -0.0558, -0.0609, -0.0609, -0.0812, -0.0812, -0.0812, -0.0761, -0.0761, -0.0728, -0.0254, -0.0372, -0.0288, -0.0305, -0.0389, -0.0558, -0.0541, -0.0575, -0.0745, -0.0829, -0.0745, -0.088, -0.0541, -0.0508, -0.0491, -0.0389, -0.0389, -0.022, 0.0, 0.0, 0.0118, 0.0017, 0.0135, 0.0068, -0.0135, -0.0321, -0.0169, 0.0118, 0.0017, 0.0068, -0.0186, 0.0237, 0.0372, 0.0186, 0.0351, 0.0181, 0.0249, 0.0351, 0.0861, 0.0776, 0.0504, 0.0453, 0.0283, 0.0385, 0.0657, 0.064, 0.0844, 0.1015, 0.1355, 0.1423, 0.1542, 0.1236, 0.1304, 0.1236, 0.1219, 0.1985, 0.2019, 0.236, 0.2155, 0.173, 0.1866, 0.19, 0.1713, 0.2002, 0.2223, 0.2087, 0.2104, 0.224, 0.2189, 0.2189, 0.1917, 0.1917, 0.2002, 0.2019, 0.2155, 0.2172, 0.1968, 0.2019, 0.1849, 0.1866, 0.1508, 0.1764, 0.1883, 0.1934, 0.1832, 0.1934, 0.2172, 0.236, 0.2496, 0.2836, 0.3024, 0.3517, 0.3313, 0.367, 0.367, 0.4215, 0.442, 0.4539, 0.4658, 0.4181, 0.4398, 0.4347, 0.4706, 0.5347, 0.5347, 0.5347, 0.5347, 0.5347, 0.5347, 0.6143, 0.5955, 0.5664, 0.6245, 0.6707, 0.6741, 0.6775, 0.6673, 0.6861, 0.6502, 0.6895, 0.7442, 0.7015, 0.7186, 0.7801, 0.8383, 0.8998, 0.7938, 0.7203, 0.6963, 0.6741, 0.7203, 0.7699, 0.763, 0.7579, 0.6621, 0.7203, 0.6724, 0.6502, 0.7203, 0.6211, 0.6536, 0.6981, 0.7579, 0.7699, 0.7186, 0.7237]}
{"unit_id": "orig:1933341684521447846", "ticker_idx": 0, "ts": "2025-06-13T01:52:31+00:00", "symbol": "Foxconn Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley maintains its \"Underweight\" rating on Foxconn Tech", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/adopts MS Underweight on Foxconn Tech: revenue surge not translating to profits, front-loaded performance risk, low-value assembly constraints.", "resolved_tickers": ["2354.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Foxconn Technology (FIH) 2354.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronics & Computer Distribution'", "bench_c_industry": "Electronics & Computer Distribution", "unit_kind": "original", "lang": "en", "tweet_text": "Foxconn: Switch 2 Drives Revenue Surge, Limited Profit Elasticity, Front-End Sales Possibly Overly Concentrated\nMorgan Stanley (S. Shih, 25/06/12)\n\nSwitch 2's sales exceeded 3.5 million units within the first four days of its launch. Foxconn Tech's revenue for 1H25 tripled year-on-year due to inventory stocking, but profit only increased by 21%, indicating that the assembly business is more sensitive to revenue than profitability.\n\nSwitch 2 Assembly and Supply\n   \nFoxconn Tech is the exclusive contract manufacturer for Switch 2, responsible for producing metal parts and thermal modules; its inventory preparation began late last year, possibly to pre-empt rising tariffs.\n\nRevenue and Profit Divergence\n   \nConsolidated operating revenue in 1H25 increased by approximately 200% compared to the same period last year, but net profit only grew by 21% because BoM (Bill of Materials) costs increased while service fees remained unchanged, preventing Foxconn Tech from sharing in the benefits of economies of scale.\n \nEarly Performance Driven by High-End User Pre-orders\n   \nSwitch 2's initial launch was stronger than Switch 1's, primarily due to purchases by core gamers; if subsequent demand from casual users falls short of expectations, Foxconn Tech's performance this year may be significantly front-loaded.\n \nProfit Margins Under Pressure\n\nWhile assembly projects bring substantial revenue, Foxconn Tech finds it difficult to achieve a corresponding increase in profit due to low added value and cost structure constraints.\n \nInvestment Rating and Risk Warning\n   \nGiven the potentially highly front-loaded performance and insufficient profit conversion, Morgan Stanley maintains its \"Underweight\" rating on Foxconn Tech; future attention should be paid to the sustained sales and gross profit margin of Switch 2.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010886440158685007, "ret_signed_m1d": -0.010886440158685007, "alpha_spy_m1d": 0.0004200313169770986, "alpha_c_m1d": 0.003329373467837904, "bench_spy_m1d": 0.011306471475662105, "bench_c_m1d": 0.01421581362652291, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.001555223060520028, "ret_signed_p1d": -0.001555223060520028, "alpha_spy_p1d": 0.007958973418195292, "alpha_c_p1d": 0.014625662638643178, "bench_spy_p1d": 0.00951419647871532, "bench_c_p1d": 0.016180885699163206, "ret_p1w": -0.024883387070887997, "ret_signed_p1w": 0.024883387070887997, "alpha_spy_p1w": 0.023270112075078675, "alpha_c_p1w": 0.030611457309927204, "bench_spy_p1w": -0.0016132749958093218, "bench_c_p1w": 0.005728070239039207, "ret_p1m": 0.004528639781296251, "ret_signed_p1m": -0.004528639781296251, "alpha_spy_p1m": 0.04514815198441413, "alpha_c_p1m": 0.06610596762359777, "bench_spy_p1m": 0.04967679176571038, "bench_c_p1m": 0.07063460740489402, "ret_p3m": 0.14122086283532864, "ret_signed_p3m": -0.14122086283532864, "alpha_spy_p3m": -0.04551211641545794, "alpha_c_p3m": -0.00981762389915719, "bench_spy_p3m": 0.0957087464198707, "bench_c_p3m": 0.13140323893617145, "ret_p6m": 0.04426467992630245, "ret_signed_p6m": -0.04426467992630245, "alpha_spy_p6m": 0.10741357175969446, "alpha_c_p6m": 0.19388985137551962, "bench_spy_p6m": 0.1516782516859969, "bench_c_p6m": 0.23815453130182207, "price_path": [0.0995, 0.0855, 0.0918, 0.0824, 0.0669, 0.0653, 0.0747, 0.0435, 0.0156, -0.0389, -0.0047, -0.0047, -0.0047, -0.0047, -0.1042, -0.1928, -0.2566, -0.1835, -0.1524, -0.1244, -0.0933, -0.112, -0.1058, -0.1073, -0.1291, -0.14, -0.0995, -0.126, -0.1011, -0.0886, -0.0793, -0.0824, -0.0824, -0.0575, -0.084, -0.0653, -0.0731, -0.07, -0.0793, -0.0513, -0.0544, -0.0295, -0.0529, -0.0513, -0.0793, -0.0778, -0.0638, -0.0793, -0.0824, -0.07, -0.0653, -0.0747, -0.0575, -0.0575, -0.098, -0.1042, -0.0731, -0.0622, -0.0404, -0.0513, -0.0358, -0.0233, 0.0109, 0.0, 0.0016, -0.0109, -0.0124, -0.028, -0.0249, -0.0264, 0.0047, 0.0062, 0.0031, 0.0047, -0.0124, 0.0109, 0.0077, 0.0252, 0.0029, 0.0013, 0.0141, 0.0157, 0.0125, 0.022, 0.0045, 0.0157, 0.0109, 0.0268, 0.0395, 0.0236, -0.0082, 0.0077, 0.0268, 0.0268, 0.0268, 0.0141, 0.0284, 0.0315, 0.0315, 0.0506, 0.1142, 0.1063, 0.0999, 0.1031, 0.2127, 0.1396, 0.1158, 0.1428, 0.1412, 0.1571, 0.138, 0.0999, 0.1126, 0.1015, 0.1603, 0.1571, 0.1412, 0.1396, 0.1301, 0.0999, 0.0761, 0.1031, 0.1078, 0.0919, 0.0904, 0.1047, 0.1412, 0.1253, 0.1349, 0.1253, 0.1126, 0.0935, 0.1142, 0.111, 0.1221, 0.1237, 0.1078, 0.1428, 0.1142, 0.1142, 0.1174, 0.1142, 0.1285, 0.1221, 0.1221, 0.1301, 0.1142, 0.1047, 0.1047, 0.0745, 0.0379, 0.0284, 0.0633, 0.0617, 0.0919, 0.0951, 0.1047, 0.0983, 0.0983, 0.1078, 0.1253, 0.1635, 0.1476, 0.1365, 0.1269, 0.0967, 0.0872, 0.1142, 0.0999, 0.111, 0.1237, 0.1269, 0.0919, 0.0617, 0.0554, 0.0236, 0.0204, 0.057, 0.0188, 0.0013, 0.0029, 0.0236, 0.0315, 0.0363, 0.03, 0.0379, 0.049, 0.0538, 0.049, 0.0443]}
{"unit_id": "quote:1929696874761801937", "ticker_idx": 0, "ts": "2025-06-03T00:29:21+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The U.S.-based global tech team reiterates Nvidia, TSMC, and GIGABYTE as their current top picks.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares brokerage reiterating Nvidia, TSMC, and Gigabyte as top picks; NVL72 shipment forecast raised to 25–30K units, B300 demand growing.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report:\n\nThe U.S.-based global tech team reiterates Nvidia, TSMC, and GIGABYTE as their current top picks.\n\nCurrently, shipments are focused on HGX racks. Following Computex, the brokerage has revised up its shipment forecast for NVL72 to 25,000–30,000 units this year. However, the brokerage sees the ultimate target as “1,000 units per week,” implying over 10,000 units in the coming quarter. They suggest closely monitoring the monthly revenue trends of Taiwan’s supply chain companies as a result.\n\nThe brokerage maintains its forecast of 390,000 units for CoWoS-L this year.\n\nFinally, it sees the B30 shipped outside of China as equivalent to the RTX PRO 6000, with an estimated volume of 500,000 units, and notes that demand continues to grow.", "reply_text": null, "parent_text": null, "quoted_text": "6/3 外電整理\n\n6274台燿：亞系升目標,\n亞系券商上修25-27年EPS至$11.3/$14.6/$17.4，給予明年17xPE評價，同步升目標，現在僅交易在11xPE。", "parent_author": null, "ret_m1d": -0.02719151462371494, "ret_signed_m1d": -0.02719151462371494, "alpha_spy_m1d": -0.021521259971801765, "alpha_c_m1d": -0.005124746957045678, "bench_spy_m1d": -0.005670254651913176, "bench_c_m1d": -0.022066767666669262, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00495687095373043, "ret_signed_p1d": 0.00495687095373043, "alpha_spy_p1d": 0.005225418356696876, "alpha_c_p1d": -0.00694072427209691, "bench_spy_p1d": -0.00026854740296644586, "bench_c_p1d": 0.01189759522582734, "ret_p1w": 0.01940242388990243, "ret_signed_p1w": 0.01940242388990243, "alpha_spy_p1w": 0.0076761301371559565, "alpha_c_p1w": -0.0339759646730009, "bench_spy_p1w": 0.011726293752746475, "bench_c_p1w": 0.053378388562903334, "ret_p1m": 0.11358815962237645, "ret_signed_p1m": 0.11358815962237645, "alpha_spy_p1m": 0.06964477351614695, "alpha_c_p1m": -0.01688327455202554, "bench_spy_p1m": 0.0439433861062295, "bench_c_p1m": 0.130471434174402, "ret_p3m": 0.23348034808441498, "ret_signed_p3m": 0.23348034808441498, "alpha_spy_p3m": 0.1481460707788247, "alpha_c_p3m": 0.06667290689640892, "bench_spy_p3m": 0.08533427730559029, "bench_c_p3m": 0.16680744118800606, "ret_p6m": 0.2766087459024622, "ret_signed_p6m": 0.2766087459024622, "alpha_spy_p6m": 0.12983687106978659, "alpha_c_p6m": -0.12099561606719167, "bench_spy_p6m": 0.14677187483267562, "bench_c_p6m": 0.3976043619696539, "price_path": [-0.2171, -0.2021, -0.2425, -0.2299, -0.1804, -0.1816, -0.1384, -0.1536, -0.1826, -0.1678, -0.1607, -0.1665, -0.1403, -0.1454, -0.1944, -0.2109, -0.2234, -0.2325, -0.22, -0.2181, -0.2791, -0.3322, -0.3086, -0.3181, -0.1904, -0.2383, -0.2145, -0.216, -0.2055, -0.2601, -0.2813, -0.2813, -0.3138, -0.2997, -0.2727, -0.2464, -0.2139, -0.2301, -0.228, -0.2287, -0.2097, -0.1892, -0.194, -0.196, -0.1711, -0.1689, -0.174, -0.129, -0.0799, -0.0416, -0.0452, -0.0412, -0.04, -0.0484, -0.0667, -0.0594, -0.0703, -0.0703, -0.0405, -0.0454, -0.0144, -0.0431, -0.0272, 0.0, 0.005, -0.0087, 0.0035, 0.01, 0.0194, 0.0115, 0.0268, 0.0054, 0.0246, 0.0206, 0.0302, 0.0302, 0.0187, 0.021, 0.0474, 0.0928, 0.0978, 0.1171, 0.1188, 0.0856, 0.1136, 0.1284, 0.1284, 0.1206, 0.1331, 0.1535, 0.1621, 0.1679, 0.1619, 0.2088, 0.2136, 0.2251, 0.2209, 0.2137, 0.1828, 0.2094, 0.2304, 0.2287, 0.2517, 0.2429, 0.2695, 0.2596, 0.2302, 0.2747, 0.2624, 0.2706, 0.2801, 0.2938, 0.2893, 0.2971, 0.286, 0.289, 0.2779, 0.2889, 0.2438, 0.2421, 0.2391, 0.2605, 0.2734, 0.2872, 0.286, 0.2759, 0.2335, 0.2335, 0.2094, 0.2083, 0.2156, 0.1828, 0.1919, 0.2093, 0.2558, 0.2547, 0.2593, 0.2588, 0.2385, 0.206, 0.2481, 0.2512, 0.3003, 0.2636, 0.2533, 0.2584, 0.2619, 0.2879, 0.3214, 0.326, 0.3377, 0.3287, 0.314, 0.3105, 0.3393, 0.3638, 0.2971, 0.3337, 0.275, 0.2736, 0.2876, 0.2976, 0.2935, 0.283, 0.2768, 0.2901, 0.3191, 0.3561, 0.4237, 0.4663, 0.4369, 0.434, 0.4651, 0.4071, 0.3825, 0.332, 0.3325, 0.4097, 0.368, 0.3725, 0.3234, 0.3468, 0.3215, 0.2844, 0.3209, 0.2793, 0.2668, 0.2928, 0.2593, 0.2766]}
{"unit_id": "quote:1929696874761801937", "ticker_idx": 1, "ts": "2025-06-03T00:29:21+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The U.S.-based global tech team reiterates Nvidia, TSMC, and GIGABYTE as their current top picks.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares brokerage reiterating Nvidia, TSMC, and Gigabyte as top picks; NVL72 shipment forecast raised to 25–30K units, B300 demand growing.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report:\n\nThe U.S.-based global tech team reiterates Nvidia, TSMC, and GIGABYTE as their current top picks.\n\nCurrently, shipments are focused on HGX racks. Following Computex, the brokerage has revised up its shipment forecast for NVL72 to 25,000–30,000 units this year. However, the brokerage sees the ultimate target as “1,000 units per week,” implying over 10,000 units in the coming quarter. They suggest closely monitoring the monthly revenue trends of Taiwan’s supply chain companies as a result.\n\nThe brokerage maintains its forecast of 390,000 units for CoWoS-L this year.\n\nFinally, it sees the B30 shipped outside of China as equivalent to the RTX PRO 6000, with an estimated volume of 500,000 units, and notes that demand continues to grow.", "reply_text": null, "parent_text": null, "quoted_text": "6/3 外電整理\n\n6274台燿：亞系升目標,\n亞系券商上修25-27年EPS至$11.3/$14.6/$17.4，給予明年17xPE評價，同步升目標，現在僅交易在11xPE。", "parent_author": null, "ret_m1d": -0.004210495840462847, "ret_signed_m1d": -0.004210495840462847, "alpha_spy_m1d": 0.001459758811450329, "alpha_c_m1d": 0.017856271826206416, "bench_spy_m1d": -0.005670254651913176, "bench_c_m1d": -0.022066767666669262, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04210528443571282, "ret_signed_p1d": 0.04210528443571282, "alpha_spy_p1d": 0.042373831838679266, "alpha_c_p1d": 0.03020768920988548, "bench_spy_p1d": -0.00026854740296644586, "bench_c_p1d": 0.01189759522582734, "ret_p1w": 0.10000005868559514, "ret_signed_p1w": 0.10000005868559514, "alpha_spy_p1w": 0.08827376493284866, "alpha_c_p1w": 0.046621670122691805, "bench_spy_p1w": 0.011726293752746475, "bench_c_p1w": 0.053378388562903334, "ret_p1m": 0.14695153437075126, "ret_signed_p1m": 0.14695153437075126, "alpha_spy_p1m": 0.10300814826452176, "alpha_c_p1m": 0.016480100196349268, "bench_spy_p1m": 0.0439433861062295, "bench_c_p1m": 0.130471434174402, "ret_p3m": 0.22623394777803618, "ret_signed_p3m": 0.22623394777803618, "alpha_spy_p3m": 0.1408996704724459, "alpha_c_p3m": 0.059426506590030126, "bench_spy_p3m": 0.08533427730559029, "bench_c_p3m": 0.16680744118800606, "ret_p6m": 0.5283104854009475, "ret_signed_p6m": 0.5283104854009475, "alpha_spy_p6m": 0.3815386105682719, "alpha_c_p6m": 0.13070612343129362, "bench_spy_p6m": 0.14677187483267562, "bench_c_p6m": 0.3976043619696539, "price_path": [0.053, 0.053, 0.0457, 0.0174, 0.0352, 0.0111, 0.0048, 0.0163, 0.0221, 0.0021, 0.0368, 0.0232, 0.0232, 0.0421, 0.0316, 0.0084, 0.0021, -0.0421, -0.0063, -0.0084, -0.0084, -0.0084, -0.1074, -0.1411, -0.1737, -0.0916, -0.0642, -0.0895, -0.0768, -0.1, -0.1084, -0.1053, -0.1211, -0.1411, -0.0811, -0.0905, -0.0653, -0.0558, -0.0505, -0.0442, -0.0442, 0.0, -0.0126, -0.0316, -0.0232, -0.0337, -0.0011, 0.0074, 0.02, 0.0516, 0.0453, 0.0505, 0.0358, 0.0295, 0.0421, 0.0337, 0.0358, 0.0274, 0.0158, 0.0179, 0.0179, 0.0179, -0.0042, 0.0, 0.0421, 0.0505, 0.0474, 0.0579, 0.1, 0.1211, 0.1047, 0.0888, 0.0835, 0.1047, 0.1152, 0.0941, 0.1152, 0.0782, 0.11, 0.1311, 0.1364, 0.1417, 0.1205, 0.147, 0.147, 0.1522, 0.147, 0.1417, 0.1417, 0.1522, 0.1628, 0.1628, 0.1575, 0.1734, 0.1945, 0.1945, 0.2209, 0.2157, 0.1945, 0.2104, 0.2104, 0.2104, 0.2104, 0.1998, 0.2209, 0.2262, 0.2262, 0.1998, 0.2157, 0.1892, 0.2474, 0.2421, 0.2474, 0.2474, 0.2685, 0.2421, 0.2474, 0.2474, 0.2527, 0.1998, 0.2157, 0.1998, 0.2368, 0.2421, 0.2579, 0.2262, 0.2262, 0.2315, 0.2262, 0.2262, 0.2262, 0.2474, 0.2474, 0.2685, 0.2949, 0.3108, 0.3319, 0.3267, 0.3585, 0.3426, 0.3638, 0.3426, 0.3744, 0.4222, 0.4222, 0.401, 0.3797, 0.3797, 0.385, 0.4063, 0.4487, 0.4859, 0.4859, 0.523, 0.5018, 0.5283, 0.5283, 0.5018, 0.5124, 0.5548, 0.5761, 0.5389, 0.5708, 0.5708, 0.5495, 0.5389, 0.5389, 0.5708, 0.5655, 0.5973, 0.5973, 0.592, 0.6026, 0.5973, 0.5495, 0.5548, 0.5495, 0.5655, 0.5548, 0.5655, 0.5495, 0.5177, 0.5336, 0.4912, 0.4806, 0.5442, 0.4699, 0.4593, 0.5018, 0.5283]}
{"unit_id": "quote:1929696874761801937", "ticker_idx": 2, "ts": "2025-06-03T00:29:21+00:00", "symbol": "GIGABYTE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The U.S.-based global tech team reiterates Nvidia, TSMC, and GIGABYTE as their current top picks.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares brokerage reiterating Nvidia, TSMC, and Gigabyte as top picks; NVL72 shipment forecast raised to 25–30K units, B300 demand growing.", "resolved_tickers": ["2376.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Gigabyte Technology 2376.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report:\n\nThe U.S.-based global tech team reiterates Nvidia, TSMC, and GIGABYTE as their current top picks.\n\nCurrently, shipments are focused on HGX racks. Following Computex, the brokerage has revised up its shipment forecast for NVL72 to 25,000–30,000 units this year. However, the brokerage sees the ultimate target as “1,000 units per week,” implying over 10,000 units in the coming quarter. They suggest closely monitoring the monthly revenue trends of Taiwan’s supply chain companies as a result.\n\nThe brokerage maintains its forecast of 390,000 units for CoWoS-L this year.\n\nFinally, it sees the B30 shipped outside of China as equivalent to the RTX PRO 6000, with an estimated volume of 500,000 units, and notes that demand continues to grow.", "reply_text": null, "parent_text": null, "quoted_text": "6/3 外電整理\n\n6274台燿：亞系升目標,\n亞系券商上修25-27年EPS至$11.3/$14.6/$17.4，給予明年17xPE評價，同步升目標，現在僅交易在11xPE。", "parent_author": null, "ret_m1d": -0.05840706081565372, "ret_signed_m1d": -0.05840706081565372, "alpha_spy_m1d": -0.052736806163740546, "alpha_c_m1d": -0.04360481576435449, "bench_spy_m1d": -0.005670254651913176, "bench_c_m1d": -0.014802245051299234, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015929167861238547, "ret_signed_p1d": 0.015929167861238547, "alpha_spy_p1d": 0.016197715264204993, "alpha_c_p1d": 0.014448826533170811, "bench_spy_p1d": -0.00026854740296644586, "bench_c_p1d": 0.0014803413280677358, "ret_p1w": 0.015929167861238547, "ret_signed_p1w": 0.015929167861238547, "alpha_spy_p1w": 0.004202874108492072, "alpha_c_p1w": -0.0027639949400275476, "bench_spy_p1w": 0.011726293752746475, "bench_c_p1w": 0.018693162801266094, "ret_p1m": -0.0035398150802752326, "ret_signed_p1m": -0.0035398150802752326, "alpha_spy_p1m": -0.04748320118650473, "alpha_c_p1m": -0.07793331095022737, "bench_spy_p1m": 0.0439433861062295, "bench_c_p1m": 0.07439349586995214, "ret_p3m": 0.016513761467889854, "ret_signed_p3m": 0.016513761467889854, "alpha_spy_p3m": -0.06882051583770044, "alpha_c_p3m": -0.09528585500240494, "bench_spy_p3m": 0.08533427730559029, "bench_c_p3m": 0.1117996164702948, "ret_p6m": -0.0972477064220183, "ret_signed_p6m": -0.0972477064220183, "alpha_spy_p6m": -0.24401958125469392, "alpha_c_p6m": -0.300951019514047, "bench_spy_p6m": 0.14677187483267562, "bench_c_p6m": 0.20370331309202871, "price_path": [-0.0354, -0.0619, -0.0549, -0.0655, -0.0726, -0.0938, -0.0796, -0.085, -0.0619, -0.0655, -0.0513, -0.0549, -0.0566, -0.0496, -0.0619, -0.0761, -0.0903, -0.1434, -0.1274, -0.1257, -0.1257, -0.1257, -0.2124, -0.2885, -0.3593, -0.2956, -0.285, -0.2478, -0.223, -0.2389, -0.223, -0.2248, -0.2478, -0.2602, -0.2212, -0.2248, -0.1823, -0.1841, -0.1752, -0.1912, -0.1912, -0.1646, -0.177, -0.1522, -0.1593, -0.1451, -0.1345, -0.1186, -0.1204, -0.069, -0.0602, -0.0779, -0.092, -0.0903, -0.0796, -0.0867, -0.0938, -0.085, -0.092, -0.0442, -0.023, -0.023, -0.0584, 0.0, 0.0159, 0.0177, 0.0195, 0.0248, 0.0159, 0.0053, 0.0018, 0.0159, 0.0018, 0.0018, 0.0088, -0.0088, -0.0088, -0.0195, 0.0053, 0.0071, 0.0053, -0.0035, 0.0018, 0.0142, -0.0035, -0.0212, -0.0336, -0.0354, -0.0301, -0.0142, -0.0159, -0.0071, -0.0212, 0.0035, 0.0071, 0.0159, 0.0248, 0.0283, -0.0088, 0.0195, 0.0283, 0.0177, 0.023, 0.0, -0.0147, 0.0092, 0.0092, 0.0037, 0.0202, 0.0202, 0.044, 0.0495, 0.0624, 0.0477, 0.0532, 0.0349, 0.0239, 0.0349, 0.0422, -0.0092, 0.0, -0.0055, 0.0183, 0.0294, 0.044, 0.0312, 0.0165, -0.022, -0.033, -0.0312, -0.0275, -0.0128, 0.0073, 0.044, 0.0165, 0.0055, 0.0055, -0.0037, 0.0147, -0.0092, 0.011, 0.0349, 0.0312, 0.0385, 0.0202, 0.0752, 0.0752, 0.0752, 0.0972, 0.0752, 0.0972, 0.1431, 0.1431, 0.1413, 0.1248, 0.0972, 0.0972, 0.0606, 0.0275, 0.0239, 0.0385, 0.022, 0.0459, 0.0569, 0.0349, 0.0257, 0.0257, 0.0459, 0.0514, 0.044, 0.033, 0.0294, 0.0165, -0.0202, -0.0202, 0.022, -0.0055, 0.0128, 0.0092, -0.0055, -0.0092, -0.0073, -0.0606, -0.0972, -0.1211, -0.0899, -0.1211, -0.1064, -0.1101, -0.0972]}
{"unit_id": "thread:1935676301689713085", "ticker_idx": 0, "ts": "2025-06-19T12:29:28+00:00", "symbol": "LG Display", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "LG Display may have already reached a strategic agreement with Apple, including securing commitments for future iPhone panel supply volumes", "tweet_type": "analysis", "timeline": "quarters", "summary": "LG Display set to gain iPhone share as BOE fails LTPO qual tests; BOE faces structural exclusion from iPhone 17+ supply chain.", "resolved_tickers": ["034220.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "LG Display 034220.KS Korea", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: China's BOE's Approval for Apple's Successor Products Unclear… Will Korean Displays Benefit?\n\nLG Display's decision to concentrate its new investment capacity of 1.26 trillion Korean Won on 6th-generation small and medium-sized organic light-emitting diodes (OLEDs) is drawing attention. This is because small and medium-sized OLEDs, primarily smartphone panels, are considered a red ocean, making it difficult to undertake investment without a clear demand source.\n\nIndustry speculation suggests that LG Display may have already reached a strategic agreement with Apple, including securing commitments for future iPhone panel supply volumes.\n\nAccording to industry sources on the 18th, China's BOE has yet to pass the quality test for LTPO OLEDs for Apple's premium smartphone model, the 'iPhone 16 Pro'. The iPhone 16 Pro is equipped with LTPO (Low-Temperature Polycrystalline Oxide) OLED, a display capable of low-power operation.\n\nApple has so far applied LTPO OLED only to iPhone Pro models, but from the upcoming iPhone 17 series, scheduled for release in the second half of this year, this panel will be installed across all models. For this reason, BOE's entry into the iPhone 16 Pro supply chain is considered a preliminary hurdle to gauge its entry into subsequent models. If BOE fails the test, Apple's long-maintained 3-vendor (Samsung Display, LG Display, BOE) supply chain will narrow to a 2-vendor system.\n\nAn industry official explained, \"While Samsung Display and LG Display have already started mass production of panels for the iPhone 17, BOE is struggling, failing to even pass the qual test for the preceding iPhone 16 Pro.\" They added, \"LTPO is highly difficult in terms of low-power driver design and oxide TFT uniformity.\"\n\nAs BOE struggles with the qual test, the iPhone panel volumes it previously handled are expected to be transferred to LG Display. According to market research firm Omdia, BOE's total OLED share in the iPhone 16 series last year was in the late 10%. In the same survey, Samsung Display accounted for over 50%, and LG Display was in the early 30% range.\n\nThe industry anticipates that display volumes for this year's iPhone 17 series will be allocated to Samsung Display and LG Display at a 6:4 ratio, respectively. Furthermore, even if BOE succeeds in entering the iPhone 17 supply chain, it will likely be challenging for it to stably enter the technically more advanced iPhone 18 and 19 series.\n\nAccording to major US media, Apple is highly likely to introduce its first foldable model with advanced LTPO technology starting with the iPhone 18 series. Samsung Display and LG Display have been strengthening their 'LTPO+' panel capabilities since last year, which is expected to further widen the technological gap with BOE.\n\nAn industry official stated, \"Even if BOE secures some volume in the iPhone 17 series, it's hard to see them continuing to enter subsequent generations.\" They added, \"There's talk in the industry that LG Display has pre-coordinated a certain level of demand from Apple.\"\n\n---\n\nhttps://t.co/o5pNv0jCfl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0021762785636560977, "ret_signed_m1d": 0.0021762785636560977, "alpha_spy_m1d": 0.0021762785636560977, "alpha_c_m1d": 0.0021762785636560977, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0032644178454841466, "ret_signed_p1d": 0.0032644178454841466, "alpha_spy_p1d": 0.005613056995663057, "alpha_c_p1d": 0.00769303929708498, "bench_spy_p1d": -0.0023486391501789106, "bench_c_p1d": -0.004428621451600834, "ret_p1w": 0.01196953210010876, "ret_signed_p1w": 0.01196953210010876, "alpha_spy_p1w": -0.015211091363404927, "alpha_c_p1w": -0.0289067868081454, "bench_spy_p1w": 0.027180623463513687, "bench_c_p1w": 0.04087631890825416, "ret_p1m": 0.0021762785636560977, "ret_signed_p1m": 0.0021762785636560977, "alpha_spy_p1m": -0.05137763334289769, "alpha_c_p1m": -0.07941153641073084, "bench_spy_p1m": 0.053553911906553786, "bench_c_p1m": 0.08158781497438694, "ret_p3m": 0.44831338411316657, "ret_signed_p3m": 0.44831338411316657, "alpha_spy_p3m": 0.34033415927258504, "alpha_c_p3m": 0.31913235419969443, "bench_spy_p3m": 0.10797922484058153, "bench_c_p3m": 0.12918102991347213, "ret_p6m": 0.3623503808487487, "ret_signed_p6m": 0.3623503808487487, "alpha_spy_p6m": 0.21466841814735793, "alpha_c_p6m": 0.16941028623472087, "bench_spy_p6m": 0.14768196270139078, "bench_c_p6m": 0.19294009461402784, "price_path": [-0.0033, -0.0022, 0.0207, 0.0, -0.0087, -0.0413, -0.0413, -0.0675, -0.1034, -0.0947, -0.1545, -0.1708, -0.1708, -0.1262, -0.1284, -0.1132, -0.1099, -0.1251, -0.123, -0.1121, -0.1088, -0.1121, -0.0773, -0.0849, -0.0773, -0.0773, -0.0718, -0.0783, -0.0783, -0.0925, -0.0925, -0.0925, -0.0773, -0.0838, -0.0805, -0.0501, -0.0566, -0.0501, -0.0577, -0.0675, -0.0925, -0.1012, -0.0968, -0.1088, -0.1208, -0.111, -0.1197, -0.0871, -0.0751, -0.0631, -0.0958, -0.0958, -0.0838, -0.0751, -0.0751, -0.0479, -0.0446, -0.0337, -0.0152, -0.0381, -0.0435, -0.0272, 0.0022, 0.0, 0.0033, -0.0196, 0.0261, 0.0163, 0.012, -0.0076, -0.025, -0.0044, -0.0076, 0.0359, 0.012, -0.0033, -0.0033, 0.0098, 0.0141, 0.0054, 0.0076, 0.0054, -0.0011, 0.0022, 0.0022, -0.0022, -0.0087, 0.0022, 0.0033, 0.0936, 0.1317, 0.1262, 0.1447, 0.1795, 0.1491, 0.1415, 0.173, 0.1915, 0.197, 0.1893, 0.2024, 0.1806, 0.4461, 0.2927, 0.2927, 0.2992, 0.259, 0.2383, 0.2437, 0.2405, 0.2688, 0.2797, 0.2862, 0.3025, 0.3058, 0.2688, 0.2775, 0.3058, 0.3123, 0.3047, 0.296, 0.2894, 0.2666, 0.2786, 0.3166, 0.3112, 0.4483, 0.4015, 0.4276, 0.4276, 0.4516, 0.4494, 0.4853, 0.556, 0.5267, 0.5919, 0.568, 0.5702, 0.5952, 0.5952, 0.5952, 0.5952, 0.5952, 0.5952, 0.6997, 0.6529, 0.5267, 0.5767, 0.5647, 0.5637, 0.5811, 0.5332, 0.5441, 0.5397, 0.4984, 0.5495, 0.5571, 0.5778, 0.518, 0.5996, 0.58, 0.5854, 0.5245, 0.5277, 0.4592, 0.494, 0.4679, 0.4799, 0.4646, 0.4353, 0.4048, 0.3199, 0.3471, 0.3602, 0.3221, 0.3297, 0.3308, 0.3732, 0.3819, 0.3645, 0.3417, 0.3613, 0.3917, 0.3983, 0.444, 0.4081, 0.3972, 0.3798, 0.358, 0.3624]}
{"unit_id": "thread:1935676301689713085", "ticker_idx": 1, "ts": "2025-06-19T12:29:28+00:00", "symbol": "BOE Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Even if BOE secures some volume in the iPhone 17 series, it's hard to see them continuing to enter subsequent generations", "tweet_type": "analysis", "timeline": "quarters", "summary": "LG Display set to gain iPhone share as BOE fails LTPO qual tests; BOE faces structural exclusion from iPhone 17+ supply chain.", "resolved_tickers": ["200725.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "BOE Technology Group 200725.SZ Shenzhen B-share", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: China's BOE's Approval for Apple's Successor Products Unclear… Will Korean Displays Benefit?\n\nLG Display's decision to concentrate its new investment capacity of 1.26 trillion Korean Won on 6th-generation small and medium-sized organic light-emitting diodes (OLEDs) is drawing attention. This is because small and medium-sized OLEDs, primarily smartphone panels, are considered a red ocean, making it difficult to undertake investment without a clear demand source.\n\nIndustry speculation suggests that LG Display may have already reached a strategic agreement with Apple, including securing commitments for future iPhone panel supply volumes.\n\nAccording to industry sources on the 18th, China's BOE has yet to pass the quality test for LTPO OLEDs for Apple's premium smartphone model, the 'iPhone 16 Pro'. The iPhone 16 Pro is equipped with LTPO (Low-Temperature Polycrystalline Oxide) OLED, a display capable of low-power operation.\n\nApple has so far applied LTPO OLED only to iPhone Pro models, but from the upcoming iPhone 17 series, scheduled for release in the second half of this year, this panel will be installed across all models. For this reason, BOE's entry into the iPhone 16 Pro supply chain is considered a preliminary hurdle to gauge its entry into subsequent models. If BOE fails the test, Apple's long-maintained 3-vendor (Samsung Display, LG Display, BOE) supply chain will narrow to a 2-vendor system.\n\nAn industry official explained, \"While Samsung Display and LG Display have already started mass production of panels for the iPhone 17, BOE is struggling, failing to even pass the qual test for the preceding iPhone 16 Pro.\" They added, \"LTPO is highly difficult in terms of low-power driver design and oxide TFT uniformity.\"\n\nAs BOE struggles with the qual test, the iPhone panel volumes it previously handled are expected to be transferred to LG Display. According to market research firm Omdia, BOE's total OLED share in the iPhone 16 series last year was in the late 10%. In the same survey, Samsung Display accounted for over 50%, and LG Display was in the early 30% range.\n\nThe industry anticipates that display volumes for this year's iPhone 17 series will be allocated to Samsung Display and LG Display at a 6:4 ratio, respectively. Furthermore, even if BOE succeeds in entering the iPhone 17 supply chain, it will likely be challenging for it to stably enter the technically more advanced iPhone 18 and 19 series.\n\nAccording to major US media, Apple is highly likely to introduce its first foldable model with advanced LTPO technology starting with the iPhone 18 series. Samsung Display and LG Display have been strengthening their 'LTPO+' panel capabilities since last year, which is expected to further widen the technological gap with BOE.\n\nAn industry official stated, \"Even if BOE secures some volume in the iPhone 17 series, it's hard to see them continuing to enter subsequent generations.\" They added, \"There's talk in the industry that LG Display has pre-coordinated a certain level of demand from Apple.\"\n\n---\n\nhttps://t.co/o5pNv0jCfl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011627915757930873, "ret_signed_m1d": -0.011627915757930873, "alpha_spy_m1d": -0.011627915757930873, "alpha_c_m1d": -0.011627915757930873, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007751943838620656, "ret_signed_p1d": 0.007751943838620656, "alpha_spy_p1d": 0.005403304688441746, "alpha_c_p1d": 0.0033233223870198225, "bench_spy_p1d": -0.0023486391501789106, "bench_c_p1d": -0.004428621451600834, "ret_p1w": 0.0175475047274396, "ret_signed_p1w": -0.0175475047274396, "alpha_spy_p1w": 0.009633118736074087, "alpha_c_p1w": 0.02332881418081456, "bench_spy_p1w": 0.027180623463513687, "bench_c_p1w": 0.04087631890825416, "ret_p1m": 0.09277477676818324, "ret_signed_p1m": -0.09277477676818324, "alpha_spy_p1m": -0.039220864861629456, "alpha_c_p1m": -0.011186961793796302, "bench_spy_p1m": 0.053553911906553786, "bench_c_p1m": 0.08158781497438694, "ret_p3m": 0.17592069908713603, "ret_signed_p3m": -0.17592069908713603, "alpha_spy_p3m": -0.0679414742465545, "alpha_c_p3m": -0.0467396691736639, "bench_spy_p3m": 0.10797922484058153, "bench_c_p3m": 0.12918102991347213, "ret_p6m": 0.14028677283519508, "ret_signed_p6m": -0.14028677283519508, "alpha_spy_p6m": 0.007395189866195695, "alpha_c_p6m": 0.05265332177883275, "bench_spy_p6m": 0.14768196270139078, "bench_c_p6m": 0.19294009461402784, "price_path": [0.0853, 0.0891, 0.1008, 0.0969, 0.093, 0.0891, 0.093, 0.0891, 0.0775, 0.0775, -0.0271, 0.0116, 0.0116, -0.0039, -0.0039, 0.0155, 0.0116, 0.0194, 0.0271, 0.0465, 0.0659, 0.0349, 0.0233, 0.0194, 0.0271, 0.0116, 0.0039, 0.0, 0.0, 0.0, 0.0, 0.0039, 0.0078, 0.0039, 0.0, 0.0116, 0.0116, 0.0116, 0.0078, 0.0039, 0.0039, 0.0, -0.0078, -0.0039, -0.0039, -0.0039, -0.0039, -0.0039, 0.0078, -0.0039, -0.0039, 0.0078, 0.0155, 0.0194, 0.0194, 0.0194, 0.0116, 0.0155, 0.0155, 0.0, 0.0, 0.0039, 0.0116, 0.0, -0.0078, -0.0039, 0.0039, 0.0215, 0.0175, 0.0255, 0.0334, 0.0413, 0.0413, 0.0611, 0.0492, 0.0373, 0.0453, 0.0571, 0.069, 0.0849, 0.0769, 0.0769, 0.0809, 0.0967, 0.0928, 0.1126, 0.1165, 0.1205, 0.1205, 0.1126, 0.1205, 0.1165, 0.1086, 0.069, 0.0809, 0.0888, 0.0928, 0.0928, 0.0928, 0.0928, 0.0928, 0.0967, 0.0928, 0.0888, 0.0849, 0.0888, 0.0928, 0.1047, 0.1126, 0.1324, 0.1482, 0.1284, 0.1324, 0.1482, 0.172, 0.1561, 0.1363, 0.1403, 0.1363, 0.1561, 0.164, 0.168, 0.172, 0.1838, 0.1878, 0.1759, 0.1759, 0.1799, 0.1838, 0.1522, 0.168, 0.1561, 0.168, 0.172, 0.1759, 0.1838, 0.1838, 0.1838, 0.1838, 0.1838, 0.1838, 0.1838, 0.1838, 0.2116, 0.2036, 0.2076, 0.1878, 0.1918, 0.1878, 0.168, 0.164, 0.168, 0.164, 0.1601, 0.1561, 0.1561, 0.1601, 0.1522, 0.1442, 0.1363, 0.1442, 0.1244, 0.1324, 0.1363, 0.1363, 0.164, 0.168, 0.164, 0.164, 0.164, 0.1522, 0.1363, 0.1165, 0.1165, 0.0809, 0.0928, 0.1047, 0.1086, 0.1007, 0.1126, 0.1244, 0.1165, 0.1403, 0.1482, 0.168, 0.1601, 0.1442, 0.1442, 0.1363, 0.1403]}
{"unit_id": "orig:1942458646417531303", "ticker_idx": 0, "ts": "2025-07-08T05:40:05+00:00", "symbol": "optical modules", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs has revised its 2025-2027 net income estimates for covered optical module companies upward by up to 42%, and their target prices by 23-82%", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Goldman Sachs bullish thesis on optical module sector: raised estimates up to 42%, target prices 23-82%, 800G volumes up 10-58% vs prior forecasts; stocks still undervalued.", "resolved_tickers": ["COHR", "LITE", "CIEN", "AAOI"], "resolution_method": "basket", "resolution_reason": "Optical modules basket: Coherent, Lumentum, Ciena, AAOI", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: AI Technology's Rapid Advance Drives Increased Integration of AI Chips and Optical Modules, Fueling Industry Growth\n\nAs AI technology rapidly advances, the integration of AI chips and optical modules continues to increase, acting as a key growth driver for the optical module industry.\n\n * The trend of increasing optical modules used per AI chip is significantly boosting the industry's mid-to-long-term growth potential. Notably, the projected shipments for 800G optical modules in 2025 and 2026 have been raised to 19.9 million and 33.5 million units, respectively, representing increases of 10% and 58% from previous forecasts. Consequently, the market size is expected to grow by 60% and 52% during the same periods.\n\n * Reflecting this projected increase in demand, Goldman Sachs has revised its 2025-2027 net income estimates for covered optical module companies upward by up to 42%, and their target prices by 23-82%. This clearly indicates high confidence in the industry's outlook.\n\n * The core reason for this trend is the continuous increase in the number of optical modules required per GPU or ASIC chip in AI servers. For example, NVIDIA's H100 typically required 3 optical modules, but the next-generation B300 will require 4.5, and some ASIC chips are evolving to demand as many as 8 modules. This is rapidly accelerating the demand for high-speed optical modules.\n \n* Although optical module-related stocks have already rebounded from their April lows, their current stock prices are still considered undervalued. Furthermore, the valuation gap among industry-leading companies is likely to narrow significantly in the future.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00654886446061842, "ret_signed_m1d": -0.00654886446061842, "alpha_spy_m1d": -0.007096880835820052, "alpha_c_m1d": -0.007096880835820052, "bench_spy_m1d": 0.000548016375201632, "bench_c_m1d": 0.000548016375201632, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011770717055312657, "ret_signed_p1d": 0.011770717055312657, "alpha_spy_p1d": 0.005774090288101302, "alpha_c_p1d": 0.005774090288101302, "bench_spy_p1d": 0.005996626767211355, "bench_c_p1d": 0.005996626767211355, "ret_p1w": 0.06888377361718023, "ret_signed_p1w": 0.06888377361718023, "alpha_spy_p1w": 0.06598215441384492, "alpha_c_p1w": 0.06598215441384492, "bench_spy_p1w": 0.002901619203335315, "bench_c_p1w": 0.002901619203335315, "ret_p1m": 0.11331332010104786, "ret_signed_p1m": 0.11331332010104786, "alpha_spy_p1m": 0.0932598698998903, "alpha_c_p1m": 0.0932598698998903, "bench_spy_p1m": 0.020053450201157563, "bench_c_p1m": 0.020053450201157563, "ret_p3m": 0.5037129658460525, "ret_signed_p3m": 0.5037129658460525, "alpha_spy_p3m": 0.42194273106325564, "alpha_c_p3m": 0.42194273106325564, "bench_spy_p3m": 0.08177023478279688, "bench_c_p3m": 0.08177023478279688, "ret_p6m": 1.5871106036947877, "ret_signed_p6m": 1.5871106036947877, "alpha_spy_p6m": 1.4815376254366404, "alpha_c_p6m": 1.4815376254366404, "bench_spy_p6m": 0.10557297825814738, "bench_c_p6m": 0.10557297825814738, "price_path": [-0.4031, -0.4152, -0.4108, -0.4042, -0.4196, -0.4144, -0.4144, -0.4415, -0.4152, -0.3775, -0.3402, -0.331, -0.3234, -0.3232, -0.3281, -0.2764, -0.2591, -0.2648, -0.27, -0.263, -0.2501, -0.2532, -0.1678, -0.133, -0.132, -0.14, -0.1372, -0.1328, -0.1352, -0.1593, -0.1557, -0.1543, -0.1543, -0.1266, -0.1433, -0.157, -0.1951, -0.1758, -0.1443, -0.1287, -0.1728, -0.1688, -0.1635, -0.1656, -0.1678, -0.1491, -0.187, -0.1484, -0.1535, -0.1117, -0.1117, -0.075, -0.0832, -0.0559, -0.0155, 0.0252, 0.0055, 0.0044, -0.0195, -0.0049, 0.0305, 0.0305, -0.0065, 0.0, 0.0118, 0.0248, 0.0239, 0.047, 0.0689, 0.0802, 0.1015, 0.0917, 0.0947, 0.0543, 0.0722, 0.072, 0.097, 0.1042, 0.1093, 0.1072, 0.1067, 0.0583, 0.1004, 0.0781, 0.1133, 0.1292, 0.1405, 0.1199, 0.1683, 0.1594, 0.0532, 0.0723, 0.1112, 0.0596, 0.036, 0.0591, 0.0914, 0.103, 0.1184, 0.1138, 0.1826, 0.1376, 0.1376, 0.1204, 0.1273, 0.2415, 0.2738, 0.2799, 0.2988, 0.4082, 0.4085, 0.4002, 0.463, 0.4628, 0.4407, 0.4772, 0.4731, 0.4947, 0.4555, 0.401, 0.4113, 0.4268, 0.443, 0.4475, 0.5329, 0.5221, 0.5037, 0.5584, 0.5172, 0.5763, 0.5968, 0.465, 0.5523, 0.5204, 0.5767, 0.6, 0.6131, 0.641, 0.6302, 0.5534, 0.6379, 0.734, 0.8302, 0.8391, 0.9, 0.8417, 0.8511, 0.8212, 0.7272, 0.925, 1.0077, 0.9837, 1.0725, 1.0084, 1.0167, 0.8199, 0.8299, 0.8581, 0.8481, 0.9257, 0.7557, 0.8436, 1.0672, 1.0442, 1.1443, 1.1443, 1.2426, 1.2079, 1.1707, 1.1554, 1.2542, 1.2985, 1.3876, 1.4834, 1.5786, 1.6758, 1.3737, 1.375, 1.2864, 1.2486, 1.3507, 1.5542, 1.7117, 1.733, 1.7582, 1.7582, 1.7139, 1.6443, 1.6255, 1.5871]}
{"unit_id": "quote:1942712434311168355", "ticker_idx": 0, "ts": "2025-07-08T22:28:32+00:00", "symbol": "copper", "kind": "commodity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The PwC report, in essence, suggests that copper will also become a 'strategic material' like rare earth elements. That's how I interpret it.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish copper long-term: PwC report implies it becomes a strategic material as climate-driven supply disruptions worsen through 2035–2050.", "resolved_tickers": ["CPER"], "resolution_method": "etf", "resolution_reason": "Copper commodity proxied by CPER ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "The PwC report, in essence, suggests that copper will also become a 'strategic material' like rare earth elements. That's how I interpret it.\n\nTo summarize parts of the report:\n\nIncreased Copper Supply Risk: Most of the 17 countries supplying copper to the semiconductor industry are expected to face drought risks by 2035. Chile, the world's largest copper producer, is already grappling with water shortages that are slowing down production. Currently, 25% of Chile's copper output is at risk of supply disruptions, rising to 75% within a decade and to 90-100% by 2050.\n\nImportance of Copper: Copper is essential for making billions of tiny wires inside every chip's circuit, and currently, there is no match for its price and performance.\n\nLong-term Outlook: The report warns that the risk will only increase over time if innovation in materials does not adapt to climate change, and a more secure water supply is not developed in the affected countries. Regardless of global efforts to reduce carbon emissions, approximately half of each country's copper supply is projected to be at risk by 2050.\n\nMention of Past Cases: The report cited the previous global chip shortage (fueled by a pandemic-driven demand spike coinciding with factory shutdowns), which cost the U.S. economy a full percentage point in GDP growth and Germany 2.4%, emphasizing the potential economic impact of copper supply disruptions.", "reply_text": null, "parent_text": null, "quoted_text": "It's quite a striking coincidence that on the very day Trump announced a 50% tariff on copper, PwC released a report stating that global semiconductor production could be destabilized by copper supply disruptions due to climate change.\n\nGiven this, should one invest in copper https://t.co/X2qwlaGGob", "parent_author": null, "ret_m1d": -0.07758367793066057, "ret_signed_m1d": -0.07758367793066057, "alpha_spy_m1d": -0.0781316943058622, "alpha_c_m1d": -0.0781316943058622, "bench_spy_m1d": 0.000548016375201632, "bench_c_m1d": 0.000548016375201632, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01569436695625015, "ret_signed_p1d": 0.01569436695625015, "alpha_spy_p1d": 0.009697740189038795, "alpha_c_p1d": 0.009697740189038795, "bench_spy_p1d": 0.005996626767211355, "bench_c_p1d": 0.005996626767211355, "ret_p1w": 0.015990551027759814, "ret_signed_p1w": 0.015990551027759814, "alpha_spy_p1w": 0.013088931824424499, "alpha_c_p1w": 0.013088931824424499, "bench_spy_p1w": 0.002901619203335315, "bench_c_p1w": 0.002901619203335315, "ret_p1m": -0.1868522738453311, "ret_signed_p1m": -0.1868522738453311, "alpha_spy_p1m": -0.20690572404648866, "alpha_c_p1m": -0.20690572404648866, "bench_spy_p1m": 0.020053450201157563, "bench_c_p1m": 0.020053450201157563, "ret_p3m": -0.0725496218456877, "ret_signed_p3m": -0.0725496218456877, "alpha_spy_p3m": -0.15431985662848458, "alpha_c_p3m": -0.15431985662848458, "bench_spy_p3m": 0.08177023478279688, "bench_c_p3m": 0.08177023478279688, "ret_p6m": 0.03523833611424787, "ret_signed_p6m": 0.03523833611424787, "alpha_spy_p6m": -0.0703346421438995, "alpha_c_p6m": -0.0703346421438995, "bench_spy_p6m": 0.10557297825814738, "bench_c_p6m": 0.10557297825814738, "price_path": [-0.186, -0.1558, -0.1294, -0.141, -0.1318, -0.1273, -0.1273, -0.1179, -0.1004, -0.1007, -0.0936, -0.1004, -0.0986, -0.1063, -0.1549, -0.1466, -0.1383, -0.1341, -0.1217, -0.1489, -0.1537, -0.143, -0.1436, -0.1291, -0.1412, -0.1333, -0.1528, -0.1386, -0.1341, -0.135, -0.1344, -0.1001, -0.1001, -0.122, -0.135, -0.1365, -0.1306, -0.1007, -0.1001, -0.0965, -0.0891, -0.1048, -0.0915, -0.0909, -0.111, -0.1039, -0.1184, -0.1057, -0.1128, -0.1013, -0.1013, -0.1031, -0.0924, -0.0912, -0.0862, -0.0613, -0.0622, -0.0631, -0.0651, -0.0453, -0.0551, -0.0551, -0.0776, 0.0, 0.0157, 0.0305, 0.0219, 0.0151, 0.016, 0.0121, 0.0113, 0.0264, 0.0346, 0.0512, 0.069, 0.0711, 0.0613, 0.0302, 0.0364, -0.1638, -0.1871, -0.1836, -0.1792, -0.1943, -0.1869, -0.1863, -0.1777, -0.18, -0.1691, -0.1723, -0.1712, -0.1732, -0.1756, -0.1836, -0.1809, -0.1777, -0.1777, -0.175, -0.1729, -0.1809, -0.1729, -0.167, -0.167, -0.1564, -0.1602, -0.1685, -0.1726, -0.17, -0.167, -0.1611, -0.1472, -0.1531, -0.1404, -0.1427, -0.1584, -0.1608, -0.1537, -0.1525, -0.1525, -0.1205, -0.1261, -0.1303, -0.1031, -0.1116, -0.1072, -0.0951, -0.0725, -0.0782, -0.0731, -0.0708, -0.0622, -0.1036, -0.0646, -0.0865, -0.0897, -0.0868, -0.0874, -0.0752, -0.0956, -0.0829, -0.0699, -0.0646, -0.0554, -0.0527, -0.0465, -0.0651, -0.0663, -0.0737, -0.0959, -0.0856, -0.0871, -0.0918, -0.064, -0.072, -0.0672, -0.0743, -0.0725, -0.0847, -0.0879, -0.0782, -0.0888, -0.08, -0.0788, -0.0767, -0.0572, -0.0572, -0.0441, -0.0438, -0.0483, -0.0219, -0.0275, -0.0124, -0.0157, -0.0332, -0.0207, -0.0047, -0.0278, -0.0184, -0.0261, -0.016, -0.0151, -0.005, -0.0012, 0.0113, 0.0127, 0.0127, 0.0604, 0.0121, 0.0438, 0.0352]}
{"unit_id": "orig:1935312841336291754", "ticker_idx": 0, "ts": "2025-06-18T12:25:12+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell should still be considered a key beneficiary of AI infrastructure", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley Equal-weight on MRVL (key AI beneficiary but execution risk); prefers NVDA and AVGO over MRVL near-term.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell: AI Custom Chip Opportunity is Real, But Growth Execution is More Closely Watched Than Long-Term Targets\n\nMorgan Stanley (J. Moore, 06/18/25)\n\nMarvell's vision in the AI custom chip space is impressive. However, despite raised long-term expectations, the market is more focused on seeing near-term revenue materialization. Overall, Marvell should still be considered a key beneficiary of AI infrastructure.\n\nRaised Datacenter Market TAM and Introduced New \"XPU attach\" Concept, Targeting $19B in Revenue\n\nMarvell has raised its 2028 AI Datacenter Total Addressable Market (TAM) forecast to $94 billion, targeting a 20% market share. The newly proposed \"XPU attach\" concept covers custom networking, memory, co-processor, and other chips that support custom XPU chips, with a combined TAM of $55.4 billion. Marvell projects its revenue from AI datacenters to reach $19 billion by 2028, with $11 billion coming from XPU compute and attach.\n\nAdded 6 XPU-related Sockets, Bringing Total Designs to 18; Project Funnel Reaches $75B\n\nThe company disclosed two new XPU accelerator projects and four XPU attach sockets, with customers including two \"emerging hyperscalers.\" It has now won a cumulative 18 socket designs. Meanwhile, the custom chip project pipeline exceeds 50 projects across 10 customers, with a potential total lifetime revenue of $75 billion. XPU-type projects account for about one-third of this, with a scale of several billion dollars, while attach projects typically last 2 to 4 years and contribute several hundred million dollars in revenue.\n\nAI Chip Customization Platform is Well-Developed, But Differentiation from Competitors like AVGO is Limited\n\nThe institution believes that while Marvell's IP portfolio (e.g., SRAM, SERDES, advanced packaging) is strong, it is not exclusive to the industry. Competitors like Broadcom (AVGO) and AMD also possess world-class capabilities, and AVGO has superior experience in actual mass production, such as with TPUs. Nevertheless, the market for AI custom chips is vast, and AVGO's exclusivity with certain customers still provides Marvell with opportunities to enter the market.\n\nMarket Sentiment is Cautious; Investors Focus on Revenue Materialization, Not Another Long-Term Guidance Hike\n\nThe institution notes that several past analyst days have featured lofty targets that were not met (e.g., 5G, ARM server chips, the 2023 AI day). Although this announcement is positive, against the backdrop of weaker-than-expected performance in the first half of 2025, investors would rather see revenue materialize than another upward revision of the 2028 forecast. In comparison, the institution still prefers NVIDIA (NVDA) and Broadcom (AVGO), maintaining its \"Equal-weight\" rating on Marvell.\n\nGrowth Inflection Point May Appear in 2026; Progress on Maia and Trainium 3 Remains in Focus\n\nMicrosoft has confirmed that its Maia project is developing well, while it remains debated whether Marvell is leading the XPU for AWS's Trainium 3. Even if it is an XPU attach rather than the core XPU, the project could still drive growth in the partnership with AWS and potentially have higher profit margins. Overall, the majority of Marvell's projects are expected to enter the mass production cycle between 2026 and 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06617742597775589, "ret_signed_m1d": -0.06617742597775589, "alpha_spy_m1d": -0.06632808442647598, "alpha_c_m1d": -0.06236924915785513, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.013075462479225308, "ret_signed_p1w": 0.013075462479225308, "alpha_spy_p1w": -0.0061311130904015965, "alpha_c_p1w": -0.034946199805228506, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": -0.038439644806319784, "ret_signed_p1m": -0.038439644806319784, "alpha_spy_p1m": -0.09276575876176651, "alpha_c_p1m": -0.14933492892037248, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": -0.09959714653714336, "ret_signed_p3m": -0.09959714653714336, "alpha_spy_p3m": -0.20910393231560576, "alpha_c_p3m": -0.26936761837288603, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.19496294868851893, "ret_signed_p6m": 0.19496294868851893, "alpha_spy_p6m": 0.03480692109508299, "alpha_c_p6m": -0.21746717081217692, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [-0.0619, -0.0303, -0.0448, -0.1103, -0.1356, -0.1732, -0.1795, -0.1648, -0.1573, -0.2585, -0.3412, -0.3203, -0.3332, -0.1876, -0.2954, -0.2877, -0.3027, -0.2886, -0.3071, -0.3102, -0.3102, -0.3412, -0.3246, -0.2826, -0.2351, -0.2139, -0.2168, -0.2171, -0.2212, -0.1868, -0.1684, -0.173, -0.1832, -0.2487, -0.2311, -0.2041, -0.1394, -0.1264, -0.1207, -0.1301, -0.1493, -0.1653, -0.1805, -0.198, -0.1748, -0.1903, -0.1903, -0.1485, -0.1382, -0.1497, -0.1969, -0.1799, -0.168, -0.1154, -0.1306, -0.0881, -0.0775, -0.0815, -0.0895, -0.0708, -0.1035, -0.0604, -0.0662, 0.0, 0.0, -0.0192, -0.0556, 0.0035, 0.0131, 0.067, 0.0295, 0.0327, 0.0172, -0.0093, 0.0031, 0.0031, -0.0454, -0.04, -0.0359, -0.0212, -0.0291, -0.0318, -0.0331, -0.0539, -0.0384, -0.0032, -0.0244, -0.0387, -0.0216, -0.0113, -0.0091, 0.0136, 0.0194, 0.0915, 0.0732, -0.0059, 0.0219, 0.0233, 0.0058, 0.0128, 0.0327, 0.0319, 0.039, 0.0592, 0.0554, 0.0174, 0.0247, -0.0376, -0.049, -0.0491, -0.0252, -0.0259, -0.0084, -0.0013, 0.0313, -0.1605, -0.1605, -0.1374, -0.168, -0.1441, -0.1543, -0.1187, -0.1075, -0.104, -0.1108, -0.1007, -0.0996, -0.0805, -0.0522, -0.0088, -0.0084, 0.0086, -0.0036, 0.0695, 0.1191, 0.1106, 0.1002, 0.1226, 0.1202, 0.151, 0.1513, 0.1874, 0.1613, 0.2352, 0.2109, 0.1439, 0.1944, 0.1521, 0.1877, 0.1789, 0.1752, 0.147, 0.1259, 0.083, 0.106, 0.1241, 0.1853, 0.182, 0.2046, 0.1835, 0.2526, 0.2075, 0.1704, 0.2413, 0.2471, 0.2149, 0.2457, 0.1936, 0.1936, 0.1694, 0.1551, 0.1151, 0.0513, 0.0866, 0.0246, 0.0349, 0.1196, 0.1148, 0.1721, 0.1721, 0.1946, 0.2173, 0.2412, 0.3389, 0.312, 0.3216, 0.2293, 0.1879, 0.2356, 0.195]}
{"unit_id": "orig:1935312841336291754", "ticker_idx": 1, "ts": "2025-06-18T12:25:12+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the institution still prefers NVIDIA (NVDA) and Broadcom (AVGO)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley Equal-weight on MRVL (key AI beneficiary but execution risk); prefers NVDA and AVGO over MRVL near-term.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell: AI Custom Chip Opportunity is Real, But Growth Execution is More Closely Watched Than Long-Term Targets\n\nMorgan Stanley (J. Moore, 06/18/25)\n\nMarvell's vision in the AI custom chip space is impressive. However, despite raised long-term expectations, the market is more focused on seeing near-term revenue materialization. Overall, Marvell should still be considered a key beneficiary of AI infrastructure.\n\nRaised Datacenter Market TAM and Introduced New \"XPU attach\" Concept, Targeting $19B in Revenue\n\nMarvell has raised its 2028 AI Datacenter Total Addressable Market (TAM) forecast to $94 billion, targeting a 20% market share. The newly proposed \"XPU attach\" concept covers custom networking, memory, co-processor, and other chips that support custom XPU chips, with a combined TAM of $55.4 billion. Marvell projects its revenue from AI datacenters to reach $19 billion by 2028, with $11 billion coming from XPU compute and attach.\n\nAdded 6 XPU-related Sockets, Bringing Total Designs to 18; Project Funnel Reaches $75B\n\nThe company disclosed two new XPU accelerator projects and four XPU attach sockets, with customers including two \"emerging hyperscalers.\" It has now won a cumulative 18 socket designs. Meanwhile, the custom chip project pipeline exceeds 50 projects across 10 customers, with a potential total lifetime revenue of $75 billion. XPU-type projects account for about one-third of this, with a scale of several billion dollars, while attach projects typically last 2 to 4 years and contribute several hundred million dollars in revenue.\n\nAI Chip Customization Platform is Well-Developed, But Differentiation from Competitors like AVGO is Limited\n\nThe institution believes that while Marvell's IP portfolio (e.g., SRAM, SERDES, advanced packaging) is strong, it is not exclusive to the industry. Competitors like Broadcom (AVGO) and AMD also possess world-class capabilities, and AVGO has superior experience in actual mass production, such as with TPUs. Nevertheless, the market for AI custom chips is vast, and AVGO's exclusivity with certain customers still provides Marvell with opportunities to enter the market.\n\nMarket Sentiment is Cautious; Investors Focus on Revenue Materialization, Not Another Long-Term Guidance Hike\n\nThe institution notes that several past analyst days have featured lofty targets that were not met (e.g., 5G, ARM server chips, the 2023 AI day). Although this announcement is positive, against the backdrop of weaker-than-expected performance in the first half of 2025, investors would rather see revenue materialize than another upward revision of the 2028 forecast. In comparison, the institution still prefers NVIDIA (NVDA) and Broadcom (AVGO), maintaining its \"Equal-weight\" rating on Marvell.\n\nGrowth Inflection Point May Appear in 2026; Progress on Maia and Trainium 3 Remains in Focus\n\nMicrosoft has confirmed that its Maia project is developing well, while it remains debated whether Marvell is leading the XPU for AWS's Trainium 3. Even if it is an XPU attach rather than the core XPU, the project could still drive growth in the partnership with AWS and potentially have higher profit margins. Overall, the majority of Marvell's projects are expected to enter the mass production cycle between 2026 and 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009348349326185335, "ret_signed_m1d": -0.009348349326185335, "alpha_spy_m1d": -0.009499007774905421, "alpha_c_m1d": -0.00554017250628458, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.06069576878985283, "ret_signed_p1w": 0.06069576878985283, "alpha_spy_p1w": 0.04148919322022593, "alpha_c_p1w": 0.012674106505399019, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": 0.1891670177429181, "ret_signed_p1m": 0.1891670177429181, "alpha_spy_p1m": 0.13484090378747138, "alpha_c_p1m": 0.07827173362886541, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": 0.22188645019109177, "ret_signed_p3m": 0.22188645019109177, "alpha_spy_p3m": 0.11237966441262937, "alpha_c_p3m": 0.05211597835534909, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.24381553306802606, "ret_signed_p6m": 0.24381553306802606, "alpha_spy_p6m": 0.08365950547459011, "alpha_c_p6m": -0.1686145864326698, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [-0.191, -0.1655, -0.1705, -0.2181, -0.2341, -0.2462, -0.2551, -0.2429, -0.241, -0.3003, -0.3518, -0.3289, -0.3381, -0.2142, -0.2606, -0.2375, -0.2391, -0.2288, -0.2818, -0.3024, -0.3024, -0.3339, -0.3203, -0.294, -0.2685, -0.237, -0.2527, -0.2507, -0.2514, -0.2329, -0.213, -0.2177, -0.2196, -0.1954, -0.1933, -0.1982, -0.1546, -0.1069, -0.0698, -0.0733, -0.0694, -0.0682, -0.0764, -0.0941, -0.087, -0.0976, -0.0976, -0.0687, -0.0734, -0.0433, -0.0712, -0.0557, -0.0293, -0.0245, -0.0378, -0.0259, -0.0197, -0.0105, -0.0182, -0.0033, -0.0241, -0.0054, -0.0093, 0.0, 0.0, -0.0112, -0.009, 0.0166, 0.0607, 0.0656, 0.0843, 0.086, 0.0538, 0.0809, 0.0953, 0.0953, 0.0877, 0.0998, 0.1196, 0.128, 0.1336, 0.1278, 0.1734, 0.178, 0.1892, 0.1851, 0.178, 0.1481, 0.1739, 0.1943, 0.1926, 0.2149, 0.2064, 0.2323, 0.2226, 0.1941, 0.2373, 0.2253, 0.2333, 0.2426, 0.2558, 0.2514, 0.259, 0.2482, 0.2512, 0.2404, 0.2511, 0.2073, 0.2057, 0.2028, 0.2235, 0.236, 0.2495, 0.2483, 0.2385, 0.1973, 0.1973, 0.1739, 0.1728, 0.18, 0.1481, 0.1569, 0.1738, 0.2189, 0.2179, 0.2224, 0.2219, 0.2022, 0.1706, 0.2115, 0.2145, 0.2622, 0.2266, 0.2165, 0.2215, 0.2249, 0.2501, 0.2826, 0.2871, 0.2985, 0.2897, 0.2754, 0.272, 0.3, 0.3238, 0.2591, 0.2945, 0.2376, 0.2362, 0.2498, 0.2595, 0.2555, 0.2453, 0.2393, 0.2522, 0.2804, 0.3163, 0.3819, 0.4232, 0.3947, 0.392, 0.4221, 0.3658, 0.3419, 0.2929, 0.2934, 0.3683, 0.3278, 0.3322, 0.2845, 0.3073, 0.2827, 0.2467, 0.2822, 0.2418, 0.2297, 0.2549, 0.2224, 0.2391, 0.2391, 0.2167, 0.2368, 0.2474, 0.2345, 0.2607, 0.254, 0.2756, 0.2716, 0.2634, 0.2438]}
{"unit_id": "orig:1935312841336291754", "ticker_idx": 2, "ts": "2025-06-18T12:25:12+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the institution still prefers NVIDIA (NVDA) and Broadcom (AVGO)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley Equal-weight on MRVL (key AI beneficiary but execution risk); prefers NVDA and AVGO over MRVL near-term.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Marvell: AI Custom Chip Opportunity is Real, But Growth Execution is More Closely Watched Than Long-Term Targets\n\nMorgan Stanley (J. Moore, 06/18/25)\n\nMarvell's vision in the AI custom chip space is impressive. However, despite raised long-term expectations, the market is more focused on seeing near-term revenue materialization. Overall, Marvell should still be considered a key beneficiary of AI infrastructure.\n\nRaised Datacenter Market TAM and Introduced New \"XPU attach\" Concept, Targeting $19B in Revenue\n\nMarvell has raised its 2028 AI Datacenter Total Addressable Market (TAM) forecast to $94 billion, targeting a 20% market share. The newly proposed \"XPU attach\" concept covers custom networking, memory, co-processor, and other chips that support custom XPU chips, with a combined TAM of $55.4 billion. Marvell projects its revenue from AI datacenters to reach $19 billion by 2028, with $11 billion coming from XPU compute and attach.\n\nAdded 6 XPU-related Sockets, Bringing Total Designs to 18; Project Funnel Reaches $75B\n\nThe company disclosed two new XPU accelerator projects and four XPU attach sockets, with customers including two \"emerging hyperscalers.\" It has now won a cumulative 18 socket designs. Meanwhile, the custom chip project pipeline exceeds 50 projects across 10 customers, with a potential total lifetime revenue of $75 billion. XPU-type projects account for about one-third of this, with a scale of several billion dollars, while attach projects typically last 2 to 4 years and contribute several hundred million dollars in revenue.\n\nAI Chip Customization Platform is Well-Developed, But Differentiation from Competitors like AVGO is Limited\n\nThe institution believes that while Marvell's IP portfolio (e.g., SRAM, SERDES, advanced packaging) is strong, it is not exclusive to the industry. Competitors like Broadcom (AVGO) and AMD also possess world-class capabilities, and AVGO has superior experience in actual mass production, such as with TPUs. Nevertheless, the market for AI custom chips is vast, and AVGO's exclusivity with certain customers still provides Marvell with opportunities to enter the market.\n\nMarket Sentiment is Cautious; Investors Focus on Revenue Materialization, Not Another Long-Term Guidance Hike\n\nThe institution notes that several past analyst days have featured lofty targets that were not met (e.g., 5G, ARM server chips, the 2023 AI day). Although this announcement is positive, against the backdrop of weaker-than-expected performance in the first half of 2025, investors would rather see revenue materialize than another upward revision of the 2028 forecast. In comparison, the institution still prefers NVIDIA (NVDA) and Broadcom (AVGO), maintaining its \"Equal-weight\" rating on Marvell.\n\nGrowth Inflection Point May Appear in 2026; Progress on Maia and Trainium 3 Remains in Focus\n\nMicrosoft has confirmed that its Maia project is developing well, while it remains debated whether Marvell is leading the XPU for AWS's Trainium 3. Even if it is an XPU attach rather than the core XPU, the project could still drive growth in the partnership with AWS and potentially have higher profit margins. Overall, the majority of Marvell's projects are expected to enter the mass production cycle between 2026 and 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007522058284243127, "ret_signed_m1d": -0.007522058284243127, "alpha_spy_m1d": -0.007672716732963214, "alpha_c_m1d": -0.0037138814643423723, "bench_spy_m1d": 0.00015065844872008682, "bench_c_m1d": -0.003808176819900755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.05577062543398137, "ret_signed_p1w": 0.05577062543398137, "alpha_spy_p1w": 0.036564049864354464, "alpha_c_p1w": 0.007748963149527555, "bench_spy_p1w": 0.019206575569626905, "bench_c_p1w": 0.048021662284453814, "ret_p1m": 0.14273758670903658, "ret_signed_p1m": 0.14273758670903658, "alpha_spy_p1m": 0.08841147275358985, "alpha_c_p1m": 0.031842302594983884, "bench_spy_p1m": 0.05432611395544673, "bench_c_p1m": 0.1108952841140527, "ret_p3m": 0.45246735290671247, "ret_signed_p3m": 0.45246735290671247, "alpha_spy_p3m": 0.34296056712825007, "alpha_c_p3m": 0.2826968810709698, "bench_spy_p3m": 0.1095067857784624, "bench_c_p3m": 0.16977047183574268, "ret_p6m": 0.6239130032852491, "ret_signed_p6m": 0.6239130032852491, "alpha_spy_p6m": 0.4637569756918132, "alpha_c_p6m": 0.21148288378455327, "bench_spy_p6m": 0.16015602759343595, "bench_c_p6m": 0.41243011950069586, "price_path": [-0.2372, -0.2388, -0.2507, -0.2865, -0.3155, -0.3269, -0.3336, -0.3293, -0.3151, -0.387, -0.4178, -0.3865, -0.379, -0.2631, -0.3143, -0.2759, -0.2901, -0.2878, -0.3051, -0.3195, -0.3195, -0.3385, -0.3251, -0.2959, -0.2512, -0.2346, -0.234, -0.2392, -0.234, -0.2146, -0.1895, -0.2011, -0.2037, -0.1849, -0.1731, -0.1714, -0.1181, -0.075, -0.0762, -0.0741, -0.0901, -0.0821, -0.0779, -0.0857, -0.0825, -0.0897, -0.0897, -0.0621, -0.0471, -0.037, -0.0366, -0.0101, 0.0222, 0.0391, 0.0345, -0.0172, -0.0278, -0.0264, 0.0066, 0.0191, -0.0102, 0.0033, -0.0075, 0.0, 0.0, -0.0027, 0.0124, 0.0523, 0.0558, 0.0778, 0.0745, 0.0997, 0.0561, 0.0767, 0.0978, 0.0978, 0.0938, 0.0843, 0.1086, 0.0987, 0.0946, 0.0995, 0.1208, 0.1202, 0.1427, 0.1303, 0.1498, 0.1114, 0.1317, 0.1518, 0.1576, 0.1741, 0.1865, 0.2072, 0.1717, 0.1515, 0.1877, 0.1686, 0.2035, 0.2118, 0.2166, 0.2124, 0.248, 0.2331, 0.2416, 0.2221, 0.2198, 0.1765, 0.1616, 0.1553, 0.1729, 0.1738, 0.1889, 0.1978, 0.2313, 0.1864, 0.1864, 0.1898, 0.2063, 0.2211, 0.336, 0.3789, 0.3431, 0.4743, 0.4347, 0.4356, 0.4525, 0.4362, 0.381, 0.3777, 0.3761, 0.3539, 0.3545, 0.3559, 0.3431, 0.3368, 0.3103, 0.3184, 0.3323, 0.3514, 0.3522, 0.3407, 0.3443, 0.3807, 0.3787, 0.2973, 0.4254, 0.3752, 0.404, 0.4152, 0.396, 0.3956, 0.3693, 0.3599, 0.3758, 0.4152, 0.4468, 0.4904, 0.5424, 0.5044, 0.4771, 0.4488, 0.4064, 0.4345, 0.421, 0.3964, 0.4322, 0.4065, 0.4195, 0.3586, 0.3685, 0.3693, 0.3607, 0.4163, 0.3859, 0.3595, 0.5104, 0.5386, 0.5887, 0.5887, 0.6103, 0.5428, 0.5248, 0.521, 0.5227, 0.5595, 0.6029, 0.6236, 0.6503, 0.6239]}
{"unit_id": "orig:1930080557503000742", "ticker_idx": 0, "ts": "2025-06-04T01:53:58+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes demand is significantly exceeding expectations… leading to potential sharp earnings upside", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish thesis: NVDA demand ahead of consensus with sharp EPS upside potential; MRVL tech edge intact; AVGO AI/networking strength; sector recovery by Q4.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Weekly: AI & Networking Face Supply-Demand Bottlenecks — NVDA & AVGO Under Short-Term Pressure, But Long-Term Outlook Remains Strong\nMorgan Stanley (J. Moore, June 2, 2025)\n\nLeading players like NVIDIA (NVDA) and Broadcom (AVGO) maintain a strong competitive edge in AI and networking, and while short-term challenges persist due to supply-demand imbalances and policy risks, their long-term growth trajectories remain robust.\n\n⸻\n\n• NVIDIA $NVDA\n• Q1 results exceeded expectations, with strong revenue outside of China,\n• Networking business has rebounded,\n• And the rack server supply chain is steadily scaling.\n• While export controls to China have raised market concerns,\n→ The company has not shown meaningful signs of slowing,\n→ Although it did not issue clear guidance for 2H25,\n→ Morgan Stanley believes demand is significantly exceeding expectations.\n• As rack supply bottlenecks ease,\n→ Supply for the “Bianca” card (next-gen AI accelerator) will likely tighten,\n→ Leading to potential sharp earnings upside.\n\n⸻\n\n• Marvell $MRVL\n• Faces short-term pressure in optics due to the Trainium project reset and China export restrictions,\n→ But these impacts are expected to be temporary.\n• The Trainium program is on track to ship ~$2 billion worth in 2025, in line with expectations.\n• Morgan Stanley believes Marvell retains a technological edge in AI and optical interconnects,\n→ And valuation recovery could follow once market sentiment bottoms out.\n\n⸻\n\n• Broadcom $AVGO\n• Custom ASIC revenue is primarily driven by Google,\n• Products for Meta and ByteDance are in mass production but showing limited incremental upside.\n• Networking remains strong, supported by the increasing use of Ethernet in AI systems.\n• The software business continues its transition to a subscription model,\n→ Over 60% of customers have now migrated,\n→ FY2025 software revenue is expected to grow 24% YoY.\n• Short-term pressure exists in non-AI segments due to cyclical fluctuations,\n→ But the company’s long-term positioning in AI and networking is a clear strength.\n\n⸻\n\n• Sector Outlook:\n• Capital continues to flow rapidly into AI infrastructure and applications.\n• The market is closely watching the supply-demand dynamics of leading players such as NVDA and AVGO.\n• Once bottlenecks in rack servers and custom ASICs are resolved,\n→ The resulting supply shortage could act as a major earnings catalyst.\n• Non-AI semiconductor demand is expected to recover in Q3,\n→ With a notable rebound likely in Q4,\n→ Supporting a continued recovery in overall industry sentiment through the end of the year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004932421576486434, "ret_signed_m1d": -0.004932421576486434, "alpha_spy_m1d": -0.005201041116532745, "alpha_c_m1d": 0.006825285213173493, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013599137425771679, "ret_signed_p1d": -0.013599137425771679, "alpha_spy_p1d": -0.008766368369582445, "alpha_c_p1d": -0.011811632812611439, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.006481889145962372, "ret_signed_p1w": 0.006481889145962372, "alpha_spy_p1w": -0.0026300420161626814, "alpha_c_p1w": -0.03375642048266836, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.12282315830271795, "ret_signed_p1m": 0.12282315830271795, "alpha_spy_p1m": 0.07036941773677508, "alpha_c_p1m": -0.003770715177219053, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.22739630300135127, "ret_signed_p3m": 0.22739630300135127, "alpha_spy_p3m": 0.14177048370139467, "alpha_c_p3m": 0.07430784158682724, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.2703119733794417, "ret_signed_p6m": 0.2703119733794417, "alpha_spy_p6m": 0.12323205321321051, "alpha_c_p6m": -0.11085976629422323, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.206, -0.2463, -0.2337, -0.1845, -0.1856, -0.1427, -0.1578, -0.1867, -0.1719, -0.1648, -0.1707, -0.1445, -0.1496, -0.1984, -0.2148, -0.2272, -0.2363, -0.2239, -0.222, -0.2827, -0.3355, -0.312, -0.3214, -0.1944, -0.242, -0.2184, -0.2199, -0.2094, -0.2637, -0.2849, -0.2849, -0.3172, -0.3032, -0.2763, -0.2501, -0.2178, -0.2339, -0.2318, -0.2325, -0.2136, -0.1932, -0.198, -0.2, -0.1752, -0.173, -0.1781, -0.1333, -0.0845, -0.0464, -0.05, -0.0459, -0.0447, -0.0531, -0.0713, -0.0641, -0.0749, -0.0749, -0.0452, -0.0501, -0.0192, -0.0478, -0.032, -0.0049, 0.0, -0.0136, -0.0014, 0.005, 0.0144, 0.0065, 0.0218, 0.0004, 0.0196, 0.0156, 0.0252, 0.0252, 0.0137, 0.0159, 0.0422, 0.0874, 0.0924, 0.1116, 0.1133, 0.0803, 0.1081, 0.1228, 0.1228, 0.1151, 0.1275, 0.1478, 0.1564, 0.1621, 0.1562, 0.2029, 0.2076, 0.2191, 0.2149, 0.2077, 0.177, 0.2034, 0.2243, 0.2226, 0.2455, 0.2368, 0.2633, 0.2534, 0.2242, 0.2684, 0.2561, 0.2643, 0.2738, 0.2874, 0.2829, 0.2907, 0.2796, 0.2826, 0.2716, 0.2826, 0.2377, 0.236, 0.233, 0.2542, 0.2671, 0.2809, 0.2797, 0.2696, 0.2274, 0.2274, 0.2034, 0.2023, 0.2096, 0.1769, 0.186, 0.2033, 0.2496, 0.2485, 0.2531, 0.2526, 0.2324, 0.2001, 0.242, 0.245, 0.2939, 0.2574, 0.2471, 0.2522, 0.2557, 0.2815, 0.3148, 0.3195, 0.3311, 0.3222, 0.3075, 0.304, 0.3327, 0.3571, 0.2907, 0.3271, 0.2687, 0.2673, 0.2812, 0.2912, 0.2871, 0.2767, 0.2705, 0.2837, 0.3126, 0.3495, 0.4167, 0.459, 0.4298, 0.427, 0.4579, 0.4002, 0.3757, 0.3254, 0.3259, 0.4027, 0.3612, 0.3657, 0.3168, 0.3401, 0.315, 0.2781, 0.3144, 0.273, 0.2606, 0.2864, 0.2531, 0.2703, 0.2703]}
{"unit_id": "orig:1930080557503000742", "ticker_idx": 1, "ts": "2025-06-04T01:53:58+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes Marvell retains a technological edge in AI and optical interconnects, and valuation recovery could follow once market sentiment bottoms out", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish thesis: NVDA demand ahead of consensus with sharp EPS upside potential; MRVL tech edge intact; AVGO AI/networking strength; sector recovery by Q4.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Weekly: AI & Networking Face Supply-Demand Bottlenecks — NVDA & AVGO Under Short-Term Pressure, But Long-Term Outlook Remains Strong\nMorgan Stanley (J. Moore, June 2, 2025)\n\nLeading players like NVIDIA (NVDA) and Broadcom (AVGO) maintain a strong competitive edge in AI and networking, and while short-term challenges persist due to supply-demand imbalances and policy risks, their long-term growth trajectories remain robust.\n\n⸻\n\n• NVIDIA $NVDA\n• Q1 results exceeded expectations, with strong revenue outside of China,\n• Networking business has rebounded,\n• And the rack server supply chain is steadily scaling.\n• While export controls to China have raised market concerns,\n→ The company has not shown meaningful signs of slowing,\n→ Although it did not issue clear guidance for 2H25,\n→ Morgan Stanley believes demand is significantly exceeding expectations.\n• As rack supply bottlenecks ease,\n→ Supply for the “Bianca” card (next-gen AI accelerator) will likely tighten,\n→ Leading to potential sharp earnings upside.\n\n⸻\n\n• Marvell $MRVL\n• Faces short-term pressure in optics due to the Trainium project reset and China export restrictions,\n→ But these impacts are expected to be temporary.\n• The Trainium program is on track to ship ~$2 billion worth in 2025, in line with expectations.\n• Morgan Stanley believes Marvell retains a technological edge in AI and optical interconnects,\n→ And valuation recovery could follow once market sentiment bottoms out.\n\n⸻\n\n• Broadcom $AVGO\n• Custom ASIC revenue is primarily driven by Google,\n• Products for Meta and ByteDance are in mass production but showing limited incremental upside.\n• Networking remains strong, supported by the increasing use of Ethernet in AI systems.\n• The software business continues its transition to a subscription model,\n→ Over 60% of customers have now migrated,\n→ FY2025 software revenue is expected to grow 24% YoY.\n• Short-term pressure exists in non-AI segments due to cyclical fluctuations,\n→ But the company’s long-term positioning in AI and networking is a clear strength.\n\n⸻\n\n• Sector Outlook:\n• Capital continues to flow rapidly into AI infrastructure and applications.\n• The market is closely watching the supply-demand dynamics of leading players such as NVDA and AVGO.\n• Once bottlenecks in rack servers and custom ASICs are resolved,\n→ The resulting supply shortage could act as a major earnings catalyst.\n• Non-AI semiconductor demand is expected to recover in Q3,\n→ With a notable rebound likely in Q4,\n→ Supporting a continued recovery in overall industry sentiment through the end of the year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05942698234749233, "ret_signed_m1d": -0.05942698234749233, "alpha_spy_m1d": -0.05969560188753864, "alpha_c_m1d": -0.0476692755578324, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017194660355265512, "ret_signed_p1d": -0.017194660355265512, "alpha_spy_p1d": -0.012361891299076277, "alpha_c_p1d": -0.015407155742105272, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": 0.029260807297496028, "ret_signed_p1w": 0.029260807297496028, "alpha_spy_p1w": 0.020148876135370974, "alpha_c_p1w": -0.010977502331134703, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.13393661952654345, "ret_signed_p1m": 0.13393661952654345, "alpha_spy_p1m": 0.08148287896060058, "alpha_c_p1m": 0.007342746046606452, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": -0.05095836404495213, "ret_signed_p3m": -0.05095836404495213, "alpha_spy_p3m": -0.13658418334490874, "alpha_c_p3m": -0.20404682545947617, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.3250366577881818, "ret_signed_p6m": 0.3250366577881818, "alpha_spy_p6m": 0.17795673762195063, "alpha_c_p6m": -0.05613508188548311, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [0.0673, -0.0106, 0.0118, 0.0523, 0.0359, 0.0356, 0.0606, 0.0287, 0.0517, 0.062, 0.0605, 0.0962, 0.0798, 0.0058, -0.0228, -0.0653, -0.0724, -0.0558, -0.0474, -0.1617, -0.2553, -0.2316, -0.2463, -0.0816, -0.2035, -0.1947, -0.2118, -0.1958, -0.2167, -0.2202, -0.2202, -0.2552, -0.2365, -0.189, -0.1353, -0.1113, -0.1146, -0.1149, -0.1196, -0.0807, -0.0599, -0.0652, -0.0766, -0.1507, -0.1308, -0.1003, -0.0271, -0.0124, -0.006, -0.0166, -0.0383, -0.0564, -0.0736, -0.0934, -0.0671, -0.0846, -0.0846, -0.0374, -0.0258, -0.0388, -0.0922, -0.0729, -0.0594, 0.0, -0.0172, 0.0309, 0.0428, 0.0383, 0.0293, 0.0504, 0.0134, 0.0621, 0.0557, 0.1305, 0.1305, 0.1087, 0.0676, 0.1344, 0.1452, 0.2062, 0.1638, 0.1674, 0.1499, 0.1199, 0.1339, 0.1339, 0.0792, 0.0852, 0.0899, 0.1065, 0.0976, 0.0946, 0.0931, 0.0695, 0.087, 0.1269, 0.1029, 0.0867, 0.106, 0.1177, 0.1202, 0.1459, 0.1524, 0.2339, 0.2132, 0.1238, 0.1552, 0.1568, 0.137, 0.145, 0.1675, 0.1666, 0.1746, 0.1974, 0.1931, 0.1501, 0.1584, 0.0879, 0.0751, 0.0749, 0.102, 0.1012, 0.121, 0.129, 0.1658, -0.051, -0.051, -0.0248, -0.0594, -0.0324, -0.044, -0.0037, 0.009, 0.0129, 0.0052, 0.0167, 0.0179, 0.0395, 0.0715, 0.1205, 0.121, 0.1401, 0.1264, 0.209, 0.2651, 0.2555, 0.2437, 0.2691, 0.2663, 0.3012, 0.3015, 0.3423, 0.3128, 0.3963, 0.3688, 0.2932, 0.3503, 0.3024, 0.3427, 0.3327, 0.3285, 0.2966, 0.2728, 0.2243, 0.2503, 0.2708, 0.34, 0.3362, 0.3617, 0.3379, 0.416, 0.3651, 0.3231, 0.4033, 0.4098, 0.3734, 0.4083, 0.3494, 0.3494, 0.322, 0.3059, 0.2605, 0.1885, 0.2284, 0.1583, 0.1699, 0.2657, 0.2602, 0.325, 0.325]}
{"unit_id": "orig:1930080557503000742", "ticker_idx": 2, "ts": "2025-06-04T01:53:58+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the company's long-term positioning in AI and networking is a clear strength… once bottlenecks in rack servers and custom ASICs are resolved, the resulting supply shortage could act as a major earnings catalyst", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish thesis: NVDA demand ahead of consensus with sharp EPS upside potential; MRVL tech edge intact; AVGO AI/networking strength; sector recovery by Q4.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Weekly: AI & Networking Face Supply-Demand Bottlenecks — NVDA & AVGO Under Short-Term Pressure, But Long-Term Outlook Remains Strong\nMorgan Stanley (J. Moore, June 2, 2025)\n\nLeading players like NVIDIA (NVDA) and Broadcom (AVGO) maintain a strong competitive edge in AI and networking, and while short-term challenges persist due to supply-demand imbalances and policy risks, their long-term growth trajectories remain robust.\n\n⸻\n\n• NVIDIA $NVDA\n• Q1 results exceeded expectations, with strong revenue outside of China,\n• Networking business has rebounded,\n• And the rack server supply chain is steadily scaling.\n• While export controls to China have raised market concerns,\n→ The company has not shown meaningful signs of slowing,\n→ Although it did not issue clear guidance for 2H25,\n→ Morgan Stanley believes demand is significantly exceeding expectations.\n• As rack supply bottlenecks ease,\n→ Supply for the “Bianca” card (next-gen AI accelerator) will likely tighten,\n→ Leading to potential sharp earnings upside.\n\n⸻\n\n• Marvell $MRVL\n• Faces short-term pressure in optics due to the Trainium project reset and China export restrictions,\n→ But these impacts are expected to be temporary.\n• The Trainium program is on track to ship ~$2 billion worth in 2025, in line with expectations.\n• Morgan Stanley believes Marvell retains a technological edge in AI and optical interconnects,\n→ And valuation recovery could follow once market sentiment bottoms out.\n\n⸻\n\n• Broadcom $AVGO\n• Custom ASIC revenue is primarily driven by Google,\n• Products for Meta and ByteDance are in mass production but showing limited incremental upside.\n• Networking remains strong, supported by the increasing use of Ethernet in AI systems.\n• The software business continues its transition to a subscription model,\n→ Over 60% of customers have now migrated,\n→ FY2025 software revenue is expected to grow 24% YoY.\n• Short-term pressure exists in non-AI segments due to cyclical fluctuations,\n→ But the company’s long-term positioning in AI and networking is a clear strength.\n\n⸻\n\n• Sector Outlook:\n• Capital continues to flow rapidly into AI infrastructure and applications.\n• The market is closely watching the supply-demand dynamics of leading players such as NVDA and AVGO.\n• Once bottlenecks in rack servers and custom ASICs are resolved,\n→ The resulting supply shortage could act as a major earnings catalyst.\n• Non-AI semiconductor demand is expected to recover in Q3,\n→ With a notable rebound likely in Q4,\n→ Supporting a continued recovery in overall industry sentiment through the end of the year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016201804129663877, "ret_signed_m1d": -0.016201804129663877, "alpha_spy_m1d": -0.016470423669710188, "alpha_c_m1d": -0.00444409734000395, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": -0.011757706789659927, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004404742321395871, "ret_signed_p1d": -0.004404742321395871, "alpha_spy_p1d": 0.00042802673479336306, "alpha_c_p1d": -0.0026172377082356313, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0017875046131602401, "ret_p1w": -0.031292995729120365, "ret_signed_p1w": -0.031292995729120365, "alpha_spy_p1w": -0.04040492689124542, "alpha_c_p1w": -0.0715313053577511, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.04023830962863073, "ret_p1m": 0.05648734408414113, "ret_signed_p1m": 0.05648734408414113, "alpha_spy_p1m": 0.004033603518198259, "alpha_c_p1m": -0.07010652939579587, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.126593873479937, "ret_p3m": 0.14175740498695077, "ret_signed_p3m": 0.14175740498695077, "alpha_spy_p3m": 0.05613158568699417, "alpha_c_p3m": -0.01133105642757326, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.15308846141452404, "ret_p6m": 0.5289895205311899, "ret_signed_p6m": 0.5289895205311899, "alpha_spy_p6m": 0.3819096003649587, "alpha_c_p6m": 0.14781778085752495, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.3811717396736649, "price_path": [-0.2555, -0.2956, -0.2741, -0.2583, -0.2693, -0.2533, -0.2573, -0.2795, -0.2532, -0.2702, -0.2659, -0.2675, -0.2789, -0.3134, -0.3412, -0.3522, -0.3587, -0.3545, -0.3409, -0.4101, -0.4397, -0.4096, -0.4024, -0.2908, -0.34, -0.3031, -0.3168, -0.3146, -0.3312, -0.3451, -0.3451, -0.3634, -0.3505, -0.3224, -0.2793, -0.2634, -0.2628, -0.2678, -0.2628, -0.2442, -0.22, -0.2312, -0.2336, -0.2155, -0.2042, -0.2025, -0.1513, -0.1098, -0.1109, -0.1089, -0.1244, -0.1166, -0.1126, -0.1201, -0.117, -0.1239, -0.1239, -0.0974, -0.0829, -0.0732, -0.0728, -0.0474, -0.0162, 0.0, -0.0044, -0.0542, -0.0643, -0.063, -0.0313, -0.0192, -0.0474, -0.0344, -0.0449, -0.0376, -0.0376, -0.0402, -0.0257, 0.0127, 0.0161, 0.0373, 0.0341, 0.0583, 0.0164, 0.0362, 0.0565, 0.0565, 0.0526, 0.0435, 0.0669, 0.0573, 0.0534, 0.0581, 0.0786, 0.0781, 0.0998, 0.0878, 0.1065, 0.0696, 0.0892, 0.1084, 0.1141, 0.1299, 0.1419, 0.1618, 0.1276, 0.1082, 0.143, 0.1246, 0.1582, 0.1662, 0.1709, 0.1668, 0.201, 0.1867, 0.1949, 0.1761, 0.1739, 0.1322, 0.1179, 0.1118, 0.1287, 0.1296, 0.1441, 0.1527, 0.185, 0.1418, 0.1418, 0.145, 0.161, 0.1752, 0.2857, 0.327, 0.2926, 0.4189, 0.3807, 0.3816, 0.3978, 0.3821, 0.329, 0.3259, 0.3243, 0.3029, 0.3035, 0.3049, 0.2926, 0.2865, 0.2611, 0.2688, 0.2822, 0.3006, 0.3013, 0.2902, 0.2938, 0.3287, 0.3269, 0.2485, 0.3718, 0.3235, 0.3512, 0.362, 0.3435, 0.3431, 0.3178, 0.3087, 0.3241, 0.3619, 0.3924, 0.4344, 0.4844, 0.4478, 0.4215, 0.3943, 0.3535, 0.3806, 0.3675, 0.3439, 0.3783, 0.3536, 0.3661, 0.3075, 0.317, 0.3178, 0.3095, 0.363, 0.3338, 0.3084, 0.4536, 0.4808, 0.529, 0.529]}
{"unit_id": "orig:1930080557503000742", "ticker_idx": 3, "ts": "2025-06-04T01:53:58+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Non-AI semiconductor demand is expected to recover in Q3, with a notable rebound likely in Q4, supporting a continued recovery in overall industry sentiment through the end of the year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish thesis: NVDA demand ahead of consensus with sharp EPS upside potential; MRVL tech edge intact; AVGO AI/networking strength; sector recovery by Q4.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "US semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Weekly: AI & Networking Face Supply-Demand Bottlenecks — NVDA & AVGO Under Short-Term Pressure, But Long-Term Outlook Remains Strong\nMorgan Stanley (J. Moore, June 2, 2025)\n\nLeading players like NVIDIA (NVDA) and Broadcom (AVGO) maintain a strong competitive edge in AI and networking, and while short-term challenges persist due to supply-demand imbalances and policy risks, their long-term growth trajectories remain robust.\n\n⸻\n\n• NVIDIA $NVDA\n• Q1 results exceeded expectations, with strong revenue outside of China,\n• Networking business has rebounded,\n• And the rack server supply chain is steadily scaling.\n• While export controls to China have raised market concerns,\n→ The company has not shown meaningful signs of slowing,\n→ Although it did not issue clear guidance for 2H25,\n→ Morgan Stanley believes demand is significantly exceeding expectations.\n• As rack supply bottlenecks ease,\n→ Supply for the “Bianca” card (next-gen AI accelerator) will likely tighten,\n→ Leading to potential sharp earnings upside.\n\n⸻\n\n• Marvell $MRVL\n• Faces short-term pressure in optics due to the Trainium project reset and China export restrictions,\n→ But these impacts are expected to be temporary.\n• The Trainium program is on track to ship ~$2 billion worth in 2025, in line with expectations.\n• Morgan Stanley believes Marvell retains a technological edge in AI and optical interconnects,\n→ And valuation recovery could follow once market sentiment bottoms out.\n\n⸻\n\n• Broadcom $AVGO\n• Custom ASIC revenue is primarily driven by Google,\n• Products for Meta and ByteDance are in mass production but showing limited incremental upside.\n• Networking remains strong, supported by the increasing use of Ethernet in AI systems.\n• The software business continues its transition to a subscription model,\n→ Over 60% of customers have now migrated,\n→ FY2025 software revenue is expected to grow 24% YoY.\n• Short-term pressure exists in non-AI segments due to cyclical fluctuations,\n→ But the company’s long-term positioning in AI and networking is a clear strength.\n\n⸻\n\n• Sector Outlook:\n• Capital continues to flow rapidly into AI infrastructure and applications.\n• The market is closely watching the supply-demand dynamics of leading players such as NVDA and AVGO.\n• Once bottlenecks in rack servers and custom ASICs are resolved,\n→ The resulting supply shortage could act as a major earnings catalyst.\n• Non-AI semiconductor demand is expected to recover in Q3,\n→ With a notable rebound likely in Q4,\n→ Supporting a continued recovery in overall industry sentiment through the end of the year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011757706789659927, "ret_signed_m1d": -0.011757706789659927, "alpha_spy_m1d": -0.012026326329706238, "alpha_c_m1d": -0.012026326329706238, "bench_spy_m1d": 0.00026861954004631095, "bench_c_m1d": 0.00026861954004631095, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0017875046131602401, "ret_signed_p1d": -0.0017875046131602401, "alpha_spy_p1d": 0.0030452644430289944, "alpha_c_p1d": 0.0030452644430289944, "bench_spy_p1d": -0.0048327690561892345, "bench_c_p1d": -0.0048327690561892345, "ret_p1w": 0.04023830962863073, "ret_signed_p1w": 0.04023830962863073, "alpha_spy_p1w": 0.031126378466505678, "alpha_c_p1w": 0.031126378466505678, "bench_spy_p1w": 0.009111931162125053, "bench_c_p1w": 0.009111931162125053, "ret_p1m": 0.126593873479937, "ret_signed_p1m": 0.126593873479937, "alpha_spy_p1m": 0.07414013291399413, "alpha_c_p1m": 0.07414013291399413, "bench_spy_p1m": 0.052453740565942875, "bench_c_p1m": 0.052453740565942875, "ret_p3m": 0.15308846141452404, "ret_signed_p3m": 0.15308846141452404, "alpha_spy_p3m": 0.06746264211456743, "alpha_c_p3m": 0.06746264211456743, "bench_spy_p3m": 0.08562581929995661, "bench_c_p3m": 0.08562581929995661, "ret_p6m": 0.3811717396736649, "ret_signed_p6m": 0.3811717396736649, "alpha_spy_p6m": 0.23409181950743374, "alpha_c_p6m": 0.23409181950743374, "bench_spy_p6m": 0.14707992016623117, "bench_c_p6m": 0.14707992016623117, "price_path": [-0.1059, -0.1478, -0.1491, -0.1241, -0.1279, -0.1, -0.0905, -0.1046, -0.0967, -0.0981, -0.1072, -0.0839, -0.0878, -0.1181, -0.1351, -0.1572, -0.16, -0.1559, -0.1496, -0.2232, -0.2818, -0.2655, -0.2852, -0.1625, -0.2206, -0.2004, -0.1992, -0.194, -0.228, -0.2352, -0.2352, -0.2539, -0.2377, -0.2099, -0.1699, -0.158, -0.1633, -0.1647, -0.1608, -0.1567, -0.13, -0.1356, -0.1436, -0.126, -0.1184, -0.1126, -0.057, -0.0246, -0.0149, -0.0184, -0.0212, -0.0228, -0.0247, -0.0408, -0.0435, -0.0568, -0.0568, -0.0269, -0.0373, -0.0305, -0.0477, -0.0336, -0.0118, 0.0, -0.0018, 0.0039, 0.0207, 0.041, 0.0402, 0.0453, 0.0208, 0.0462, 0.0391, 0.0431, 0.0431, 0.0338, 0.0404, 0.0786, 0.0931, 0.1011, 0.1059, 0.1078, 0.0957, 0.1172, 0.1266, 0.1266, 0.1113, 0.126, 0.1337, 0.1421, 0.142, 0.1333, 0.1551, 0.1491, 0.1587, 0.1532, 0.154, 0.1335, 0.1385, 0.1432, 0.142, 0.1575, 0.1635, 0.1768, 0.1471, 0.1279, 0.1527, 0.1404, 0.1385, 0.1564, 0.166, 0.1654, 0.1921, 0.1967, 0.1993, 0.1744, 0.1792, 0.1554, 0.1478, 0.1418, 0.1664, 0.1675, 0.1787, 0.1824, 0.1872, 0.1531, 0.1531, 0.138, 0.1378, 0.1514, 0.165, 0.1778, 0.183, 0.195, 0.2073, 0.2088, 0.2201, 0.2203, 0.2125, 0.2591, 0.2541, 0.2797, 0.2779, 0.2755, 0.2753, 0.2783, 0.2817, 0.2964, 0.3255, 0.3436, 0.3374, 0.364, 0.3388, 0.3747, 0.3712, 0.2923, 0.3496, 0.3247, 0.3576, 0.3637, 0.3618, 0.3792, 0.3716, 0.345, 0.3696, 0.3948, 0.4298, 0.4423, 0.4642, 0.4448, 0.442, 0.4542, 0.4012, 0.428, 0.3946, 0.3828, 0.4252, 0.3944, 0.4122, 0.3697, 0.3703, 0.3517, 0.3239, 0.3483, 0.2914, 0.2955, 0.3471, 0.3506, 0.3812, 0.3812]}
{"unit_id": "quote:1923171003242574111", "ticker_idx": 1, "ts": "2025-05-16T00:17:52+00:00", "symbol": "3583.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Underperform for Chunghwa Precision Test Tech (Chun-Yuen)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Australian broker downgrade: Topco Scientific cut to Neutral, Chun-Yuen cut to Underperform on CoWoS order deferrals and SoIC/CoPoS delays.", "resolved_tickers": ["3583.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronics & Computer Distribution'", "bench_c_industry": "Electronics & Computer Distribution", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Another brokerage has downgraded the CoWoS supply chain. An Australian brokerage estimates CoWoS capacity for 2025–2027 at 70K, 90K, and 110K wafers per month (wpm), respectively. They also believe that some orders have been deferred from 4Q25 to 1Q26. SoIC and CoPoS still require more time to develop, leading the firm to downgrade its rating on Topco Scientific to Neutral, and Underperform for Chunghwa Precision Test Tech (Chun-Yuen).\n\nAI Supply Chain Tracking Update:\nA major U.S. investment bank reiterated that demand for CoWoS-L used in Blackwell remains unchanged, maintaining forecasts of 390K units for this year and 540K units for next year. They estimate that NVIDIA will consume around 160K units in the first half of this year, and over 100K more in 3Q. The brokerage believes demand is not an issue; if assembly yields improve, it will further accelerate chip inventory consumption.\"", "reply_text": null, "parent_text": null, "quoted_text": "早安！5/16 外電綜合整理\n\n- 3131弘塑、3583辛耘：澳系降評\n又一家券商降評CoWoS供應鏈，澳系券商預估CoWoS 25-27年產能為70、90、110wpm, 同時認為有部分訂單從4Q25遞延到1Q26。SoIC, CoPoS也尚待時間，是故降評覆蓋裡弘塑的評等到Neutral, 降評辛耘到Underperform。\n\n- AI 供應鏈追蹤報導：", "parent_author": null, "ret_m1d": -0.024013700300621887, "ret_signed_m1d": 0.024013700300621887, "alpha_spy_m1d": 0.01771959331815376, "alpha_c_m1d": 0.022140482004181727, "bench_spy_m1d": -0.006294106982468128, "bench_c_m1d": -0.00187321829644016, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03773584505511407, "ret_signed_p1d": 0.03773584505511407, "alpha_spy_p1d": 0.038829763155152075, "alpha_c_p1d": 0.03628846509768746, "bench_spy_p1d": 0.001093918100038005, "bench_c_p1d": -0.0014473799574266089, "ret_p1w": 0.03259006728204916, "ret_signed_p1w": -0.03259006728204916, "alpha_spy_p1w": -0.057985507402526615, "alpha_c_p1w": -0.06698908969625073, "bench_spy_p1w": -0.025395440120477453, "bench_c_p1w": -0.03439902241420156, "ret_p1m": 0.09090902342942275, "ret_signed_p1m": -0.09090902342942275, "alpha_spy_p1m": -0.0766376987017725, "alpha_c_p1m": -0.05621195337751339, "bench_spy_p1m": 0.014271324727650248, "bench_c_p1m": 0.03469707005190936, "ret_p3m": 0.22656967589033394, "ret_signed_p3m": -0.22656967589033394, "alpha_spy_p3m": -0.13805311465185532, "alpha_c_p3m": -0.0823971997338182, "bench_spy_p3m": 0.08851656123847862, "bench_c_p3m": 0.14417247615651574, "ret_p6m": 0.1379647842891909, "ret_signed_p6m": -0.1379647842891909, "alpha_spy_p6m": 0.015433826728846967, "alpha_c_p6m": 0.12390345043122908, "bench_spy_p6m": 0.15339861101803787, "bench_c_p6m": 0.26186823472042, "price_path": [0.2607, 0.2487, 0.2521, 0.271, 0.2539, 0.2144, 0.2213, 0.1715, 0.1715, 0.1063, 0.1544, 0.1664, 0.1818, 0.1424, 0.1801, 0.175, 0.1801, 0.1835, 0.1561, 0.1407, 0.1441, 0.1046, 0.1115, 0.0858, 0.0635, 0.0789, 0.0995, 0.0703, 0.0292, -0.0566, -0.0377, -0.0412, -0.0412, -0.0412, -0.1355, -0.2213, -0.2985, -0.2298, -0.1852, -0.1595, -0.1149, -0.1458, -0.1407, -0.1441, -0.1732, -0.1767, -0.0943, -0.0978, -0.0738, -0.0703, -0.0515, -0.072, -0.072, -0.0566, -0.0943, -0.0772, -0.0892, -0.0635, -0.0446, -0.0309, -0.0292, -0.0034, -0.024, 0.0, -0.0377, -0.0377, -0.0206, 0.0292, 0.0326, 0.0034, -0.0017, -0.0034, -0.0223, -0.0223, -0.0583, -0.0497, -0.0189, -0.0189, -0.0086, -0.0086, 0.0446, 0.06, 0.0806, 0.0738, 0.0909, 0.1321, 0.1149, 0.1286, 0.1184, 0.1389, 0.2075, 0.235, 0.2144, 0.2144, 0.1988, 0.2179, 0.2144, 0.204, 0.1849, 0.1675, 0.1744, 0.2214, 0.2022, 0.1883, 0.1606, 0.1918, 0.1849, 0.2266, 0.2335, 0.2231, 0.1744, 0.1953, 0.1953, 0.1779, 0.1918, 0.1762, 0.1658, 0.1779, 0.1779, 0.2161, 0.2005, 0.1797, 0.2961, 0.2839, 0.2613, 0.2526, 0.2266, 0.2196, 0.2248, 0.2283, 0.27, 0.2353, 0.2804, 0.2717, 0.2943, 0.3082, 0.3308, 0.3673, 0.3551, 0.3239, 0.2978, 0.3273, 0.2961, 0.3378, 0.3291, 0.3464, 0.3204, 0.2804, 0.2596, 0.2439, 0.237, 0.2422, 0.2978, 0.2717, 0.3013, 0.3482, 0.482, 0.4073, 0.3134, 0.3134, 0.3725, 0.3395, 0.3152, 0.3256, 0.3256, 0.3204, 0.3117, 0.3239, 0.3239, 0.31, 0.2683, 0.3169, 0.3412, 0.2856, 0.2874, 0.2735, 0.2474, 0.2422, 0.2422, 0.2439, 0.2214, 0.2248, 0.2214, 0.2405, 0.2683, 0.2161, 0.1953, 0.1779, 0.1692, 0.138]}
{"unit_id": "orig:1935848722027757881", "ticker_idx": 0, "ts": "2025-06-19T23:54:36+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Investment Opinion: Outperform (Buy)\nTarget Price: KRW 360,000 (Up 16% from previous)\n12-Month Expected Return: 47%", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Macquarie Outperform on SK Hynix, PT raised to KRW 360K (+16%), 47% upside on HBM leadership and DRAM upcycle.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix: Securing Leadership in HBM- Macquarie Research\n\n🔍 Executive Summary\n- Investment Opinion: Outperform (Buy)\n- Target Price: KRW 360,000 (Up 16% from previous)\n- Current Stock Price (as of 6/19): KRW 246,000\n- 12-Month Expected Return: 47%\n- 2025E PER: 5.0x, P/B: 1.7x\n\n📌 Investment Points Summary\n\n1. Securing HBM Leadership\n\n- HBM revenue proportion within DRAM expected to increase from 45% to 74% by 2027.\n- Effectively a leading monopoly in HBM, a high-value-added segment within DRAM.\n- Major Customers: Nvidia (65%), Broadcom (20%), AWS (15%).\n\n2. HBM Market Structure and Tight Supply\n\n- Three major structural constraints:Front-end wafer capacity shortage (die increase limited by geometry shrink alone).\n- Yield degradation due to increasing die size (TSV count increases from HBM4 to HBM5).\n- Increased production complexity due to rising heat generation.\n- HBM is a bottleneck for AI computing, leading to stronger demand than general DRAM.\n\n3. HBM Profitability: Margin Pressure vs. Economies of Scale\n\n- HBM4 may see some margin decline due to outsourcing the base die to TSMC.\n- However, HBM revenue is expected to reach KRW 56 trillion in 2025 (+330% YoY), with operating profit of KRW 40 trillion.\n- This would account for half of total DRAM revenue and 60% of total operating profit.\n\n[2] Memory ASP and Bit Growth Rate\n- DRAM ASP: $3.5 in 2024 → $7.6 in 2027.\n- HBM ASP: $12.7 → $18.2.\n- DRAM Bit Growth Rate: Maintained at 17% annually.\n- HBM Bit Growth Rate: Rapid increase at 60-100% annually.\n\n🔧 Strategic Implications\n- Entering DRAM Up-cycle: Increased server and mobile DRAM content.\nExamples: iPhone 17 → 12GB, Galaxy S26 → 16GB.\n- DDR4 discontinuation → DDR5 transition leads to server DRAM module increases to 64GB/96GB.\n- Limited entry of Chinese competitor (CXMT) into high-end DRAM → Maintaining technological gap.\n\n📉 Risk Factors\n- M15X / Yongin fab construction delays.\n- Difficulty in technology upgrades for Wuxi Fab due to lack of EUV.\n- Potential entry of CXMT into HBM/DDR5.\n- Risk of HBM4 yield degradation and production disruptions.\n\n🧠 Investor Perspective Summary\n- 2025-26 cumulative profit of KRW 87 trillion > total capital of KRW 74 trillion → ROE excess return period.\n- Increased likelihood of share buybacks: Excess cash generation expected after 2026.\n- Current P/B of 2.4x is not overvalued; expected to drop to 1.2x by 2026.\n\n🔄 Factors for Upward Revision vs. Consensus\n\n- Stronger KRW (1,426 → 1,381 KRW/USD), upward revision of DRAM prices.\n- 2026 operating profit estimate revised up by +6%.\n- HBM ASP lowered somewhat conservatively (-5% YoY), but bit growth rate maintained.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0020324669145308505, "ret_signed_m1d": 0.0020324669145308505, "alpha_spy_m1d": 0.0020324669145308505, "alpha_c_m1d": 0.0020324669145308505, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04471535640818325, "ret_signed_p1d": 0.04471535640818325, "alpha_spy_p1d": 0.04706399555836216, "alpha_c_p1d": 0.05355042922006703, "bench_spy_p1d": -0.0023486391501789106, "bench_c_p1d": -0.008835072811883782, "ret_p1w": 0.1910568649366864, "ret_signed_p1w": 0.1910568649366864, "alpha_spy_p1w": 0.1638762414731727, "alpha_c_p1w": 0.13541873243321056, "bench_spy_p1w": 0.027180623463513687, "bench_c_p1w": 0.05563813250347582, "ret_p1m": 0.09349577420878186, "ret_signed_p1m": 0.09349577420878186, "alpha_spy_p1m": 0.03994186230222807, "alpha_c_p1m": -0.012068108989098159, "bench_spy_p1m": 0.053553911906553786, "bench_c_p1m": 0.10556388319788002, "ret_p3m": 0.41667740324893665, "ret_signed_p3m": 0.41667740324893665, "alpha_spy_p3m": 0.3086981784083551, "alpha_c_p3m": 0.24675467895109904, "bench_spy_p3m": 0.10797922484058153, "bench_c_p3m": 0.1699227242978376, "ret_p6m": 1.3261555135050047, "ret_signed_p6m": 1.3261555135050047, "alpha_spy_p6m": 1.178473550803614, "alpha_c_p6m": 0.9775892397336272, "bench_spy_p6m": 0.14768196270139078, "bench_c_p6m": 0.3485662737713775, "price_path": [-0.1418, -0.156, -0.1316, -0.1601, -0.1913, -0.2262, -0.2006, -0.197, -0.2104, -0.2607, -0.3313, -0.3122, -0.3305, -0.2566, -0.2664, -0.2688, -0.2672, -0.294, -0.2899, -0.2899, -0.2834, -0.2948, -0.2656, -0.2765, -0.2518, -0.2615, -0.2664, -0.2798, -0.2798, -0.2453, -0.2453, -0.2453, -0.2258, -0.2278, -0.2286, -0.2087, -0.1945, -0.1641, -0.1864, -0.1702, -0.1909, -0.1803, -0.1864, -0.201, -0.1885, -0.1763, -0.1783, -0.156, -0.1382, -0.1687, -0.1565, -0.1565, -0.1159, -0.0874, -0.0874, -0.0691, -0.063, -0.0244, -0.0427, -0.0427, 0.0081, 0.0122, 0.002, 0.0, 0.0447, 0.0549, 0.1321, 0.1626, 0.1911, 0.1545, 0.187, 0.1606, 0.1341, 0.1321, 0.0996, 0.1016, 0.1463, 0.1423, 0.2073, 0.1972, 0.2195, 0.2134, 0.2033, 0.0955, 0.0935, 0.1077, 0.0915, 0.0935, 0.0955, 0.0813, 0.065, 0.0671, 0.0711, 0.1118, 0.0488, 0.0488, 0.0711, 0.0508, 0.065, 0.0427, 0.0854, 0.0935, 0.1301, 0.124, 0.124, 0.0874, 0.0691, 0.0386, -0.0041, 0.0203, 0.0549, 0.063, 0.0569, 0.093, 0.0951, 0.0422, 0.0605, 0.0686, 0.0808, 0.1134, 0.1276, 0.1724, 0.2376, 0.2498, 0.3373, 0.3475, 0.4167, 0.3576, 0.4472, 0.4472, 0.4289, 0.4696, 0.4554, 0.4513, 0.3699, 0.4207, 0.4146, 0.4655, 0.61, 0.61, 0.61, 0.61, 0.61, 0.61, 0.7424, 0.6894, 0.6752, 0.72, 0.8421, 0.895, 0.9764, 0.95, 0.9601, 0.9479, 1.0762, 1.1779, 1.1209, 1.2716, 1.3123, 1.2756, 1.524, 1.3856, 1.3571, 1.4141, 1.3611, 1.467, 1.5199, 1.5118, 1.4914, 1.2797, 1.467, 1.3204, 1.2879, 1.3245, 1.1209, 1.1169, 1.1128, 1.1332, 1.2162, 1.1591, 1.1917, 1.2732, 1.2488, 1.208, 1.2162, 1.3506, 1.3058, 1.3913, 1.3017, 1.3262]}
{"unit_id": "thread:1921692105275232447", "ticker_idx": 0, "ts": "2025-05-11T22:21:15+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI chip demand outside China remains robust and cloud service providers (CSPs) continue expanding capital expenditures, NVIDIA is expected to deliver strong quarterly earnings within guidance by the end of this month", "tweet_type": "analysis", "timeline": "weeks", "summary": "NVDA expected to deliver strong earnings within guidance; robust non-China AI demand and CSP capex growth support the print despite H20 ban headwinds.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Wafer Input Costs at TSMC’s U.S. Plant Soar, NVIDIA GPU Prices “Rise Across the Board” -Digitimes\n\nSources in the supply chain have revealed that NVIDIA is facing multiple crises. Following a reported $5.5 billion write-down in its quarterly earnings due to the ban on downgraded H20 chips being sold to China, CEO Jensen Huang has been shuttling between the U.S. and China in hopes of minimizing the impact and persuading the U.S. government to ease AI chip restrictions.\n\nAt the same time, market rumors indicate that in order to maintain stable profitability, NVIDIA has recently raised official prices for almost all its products, allowing its partners to increase prices accordingly.\n\nFor example, ASUS’s high-end RTX 5090 graphics card, launched earlier this year at a retail price of around NT$90,000, has recently surged to NT$100,000 through distribution channels—an increase of more than 10%. Prices for H200/B200 GPU modules and servers have also gone up.\n\nGiven that AI chip demand outside China remains robust and cloud service providers (CSPs) continue expanding capital expenditures, NVIDIA is expected to deliver strong quarterly earnings within guidance by the end of this month.\n\nSince April, U.S. President Donald Trump has imposed reciprocal tariffs and tightened AI chip export rules, effectively blocking NVIDIA’s downgraded H20 chips from entering China. As a result, NVIDIA has become a major target of these trade restrictions.\n\nJensen Huang is doing everything he can to defuse the situation and minimize the impact.\n\nIn response to Trump’s “Made in America” initiative, NVIDIA announced a $500 billion investment over four years to establish AI server manufacturing sites in the U.S. in partnership with TSMC, Foxconn, and others. Production of 4nm Blackwell chips has already begun at TSMC’s Arizona facility. Huang has also appeared alongside Trump, expressing that U.S. policies should focus on enhancing the competitiveness of American companies. He warned that restricting chip exports to China and other nations would undermine America’s technological leadership. He also visited Chinese government agencies during a sensitive time.\n\nAs Huang works both sides of the U.S.-China divide, he is seeking a resolution while repeatedly stressing to the U.S. government that China’s AI capabilities are catching up fast. He has warned that broader restrictions on AI chips could seriously harm U.S. chipmakers and accelerate China’s development efforts.\n\nIn a recent joint appearance with Trump, Huang even mentioned Huawei—blacklisted by the U.S.—noting that it has entered the AI chip field, achieving excellent performance in computing and networking technologies. He reiterated that U.S. policy should prioritize strengthening domestic competitiveness, and that limiting chip exports to China and others would threaten America's current technological lead.\n\nIt’s not just Huang raising concerns. AMD CEO Lisa Su also recently stated that China represents a \"huge opportunity\" for the global semiconductor and AI industries, urging a balanced approach between export controls and AI advancement. Rumors have emerged that the Trump administration is considering loosening some of the existing AI chip restrictions.\n\nAccording to the supply chain, NVIDIA’s Jensen Huang is actively participating in public events and issuing warnings to the Trump administration, while reiterating support for the “Made in America” initiative.\n\nOn the operational side, NVIDIA has historically maintained strong control. Despite uncertainty around tariffs and AI chip restrictions, the company is striving to post stable earnings. Numerous products have reportedly seen official price hikes, and partners have followed suit with their own increases.\n\nTake the high-demand RTX 50 series as an example: the RTX 5090, upon global launch earlier this year, was immediately snapped up at premium prices. Chinese AI chip buyers, in particular, paid without question—willing to purchase at any cost. The RTX 5090, originally priced around NT$80,000, quickly doubled in channel pricing.\n\nAfter the AI chip ban was enacted, the RTX 5090 became even harder to obtain, pushing prices higher. Due to the sharp decline in Chinese orders, companies have introduced bundled offerings to maintain profitability and clear inventory. ASUS and others have pushed packages that include power supplies or fully assembled systems. The price of the RTX 5090 surged from around NT$90,000 to NT$100,000 overnight. Other RTX 50 series cards also saw price hikes of 5–10%.\n\nAdditionally, NVIDIA has reportedly raised prices for H200 and B200 chips and modules in response to rising costs and platform transition strategies. The price increases have rippled out, with server vendors also adjusting their quotes to clients—by approximately 10–15%.\n\nThe supply chain notes that shifting Blackwell chip production to TSMC’s U.S. plant has led to noticeable cost increases in manufacturing, materials, and logistics. Much like TSMC itself, NVIDIA has the pricing power to pass these costs on to customers, making the price hikes expected.\n\nEven though AI chips like the H20 are banned from being sold to China, Chinese buyers still manage to acquire them through other channels outside the mainland. Meanwhile, demand from U.S. CSPs and other international customers remains strong. NVIDIA is expected to post impressive earnings within guidance in its upcoming report.\n\nNVIDIA’s fiscal Q4 earnings for the period ending January 26, 2025, beat market expectations.\n\nRevenue reached $39.33 billion, up 12% quarter-over-quarter and 78% year-over-year. Data center revenue came in at $35.6 billion, rising 16% QoQ and 93% YoY. The overall gross margin was 73%.\n\nLooking ahead to Q1 of fiscal 2026, NVIDIA estimates revenue of approximately $43 billion, a 65% increase YoY, although this growth is slower than the 262% surge in the same period a year earlier. Gross margins are expected to reach 70%.\n\nNVIDIA plans to release its earnings at the end of May. Before that, Jensen Huang will visit Taiwan on May 12 to meet with supply chain partners, dine with major manufacturers on the 17th, and deliver a keynote speech on the 19th—reviving what many call the “trillion-dollar banquet,” with full momentum behind the AI boom.\n\n---\n\nhttps://t.co/GthtAfqaWg", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.051626008542716484, "ret_signed_m1d": -0.051626008542716484, "alpha_spy_m1d": -0.01963569737158022, "alpha_c_m1d": 0.007385873559626277, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0563414029159095, "ret_signed_p1d": 0.0563414029159095, "alpha_spy_p1d": 0.04973746259873946, "alpha_c_p1d": 0.022012643153013878, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.10219506593164018, "ret_signed_p1w": 0.10219506593164018, "alpha_spy_p1w": 0.08185176565789498, "alpha_c_p1w": 0.06597083300351914, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.1704065092705338, "ret_signed_p1m": 0.1704065092705338, "alpha_spy_p1m": 0.1359463388222879, "alpha_c_p1m": 0.06653554689432228, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.46977681802609217, "ret_signed_p3m": 0.46977681802609217, "alpha_spy_p3m": 0.38207542320790444, "alpha_c_p3m": 0.24354377888437795, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 0.6155690716586328, "ret_signed_p6m": 0.6155690716586328, "alpha_spy_p6m": 0.4506883865956586, "alpha_c_p6m": 0.12970069447571997, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [0.0661, 0.0998, 0.1288, 0.1288, 0.1332, 0.1318, 0.139, 0.0928, 0.0591, 0.0294, 0.0672, -0.0233, 0.0155, -0.0728, -0.0571, -0.0464, -0.1011, -0.0839, -0.1303, -0.1159, -0.059, -0.0603, -0.0108, -0.0282, -0.0615, -0.0446, -0.0363, -0.0431, -0.0129, -0.0188, -0.0751, -0.0941, -0.1084, -0.1189, -0.1045, -0.1023, -0.1724, -0.2333, -0.2062, -0.2171, -0.0705, -0.1254, -0.0981, -0.0999, -0.0878, -0.1505, -0.1749, -0.1749, -0.2121, -0.196, -0.165, -0.1347, -0.0975, -0.116, -0.1137, -0.1145, -0.0926, -0.0691, -0.0746, -0.0769, -0.0483, -0.0458, -0.0516, 0.0, 0.0563, 0.1003, 0.0962, 0.1008, 0.1022, 0.0925, 0.0715, 0.0799, 0.0674, 0.0674, 0.1016, 0.096, 0.1316, 0.0986, 0.1169, 0.1481, 0.1538, 0.1381, 0.1522, 0.1596, 0.1704, 0.1613, 0.1789, 0.1543, 0.1764, 0.1718, 0.1828, 0.1828, 0.1696, 0.1722, 0.2025, 0.2546, 0.2604, 0.2826, 0.2846, 0.2464, 0.2785, 0.2955, 0.2955, 0.2866, 0.3009, 0.3243, 0.3342, 0.3409, 0.334, 0.3879, 0.3933, 0.4066, 0.4018, 0.3934, 0.3581, 0.3886, 0.4126, 0.4107, 0.4371, 0.427, 0.4576, 0.4462, 0.4125, 0.4635, 0.4494, 0.4588, 0.4698, 0.4855, 0.4803, 0.4892, 0.4764, 0.4799, 0.4672, 0.4799, 0.4281, 0.4261, 0.4227, 0.4472, 0.462, 0.4779, 0.4765, 0.4649, 0.4162, 0.4162, 0.3886, 0.3873, 0.3957, 0.358, 0.3685, 0.3884, 0.4418, 0.4406, 0.4459, 0.4453, 0.422, 0.3846, 0.433, 0.4365, 0.493, 0.4508, 0.439, 0.4448, 0.4489, 0.4786, 0.5171, 0.5225, 0.5359, 0.5256, 0.5086, 0.5046, 0.5377, 0.5658, 0.4893, 0.5312, 0.4638, 0.4622, 0.4783, 0.4898, 0.4851, 0.473, 0.4659, 0.4812, 0.5145, 0.557, 0.6346, 0.6835, 0.6497, 0.6465, 0.6822, 0.6156]}
{"unit_id": "thread:1931607868111028587", "ticker_idx": 0, "ts": "2025-06-08T07:02:57+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the Q2 earnings outlooks for Samsung Electronics and SK hynix are being sharply revised upwards", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "DRAM prices rising 8-13% QoQ in Q3 on Chinese stockpiling + US AI demand; Samsung and SK Hynix earnings outlooks sharply revised upward; memory boom may extend beyond 2025.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: DRAM Prices to Continue Rising in Q3… Driven by Chinese Stockpiling and Expanding US AI Demand\n\nAccording to market research firm TrendForce, DRAM contract prices are expected to rise by 8-13% quarter-on-quarter in the third quarter. In particular, prices for general-purpose PC DDR4 products are projected to increase by 13-18%. This is an upward revision from initial forecasts of 5-10% for server DRAM and 3-8% for PC DRAM. The industry considers it extremely unusual for prices of older memory products to show stronger growth than new products.\n\nPrices for the newer DDR5 are also on the rise. This is attributed to the expanding investment in AI server data centers by major US tech companies, leading to a comprehensive increase in demand for high-performance memory. Previously, demand was limited to NVIDIA's high-performance graphics processing units (GPUs), but recently, Google and Microsoft are also expanding their own AI infrastructure, broadening the demand for DDR5.\n\nNVIDIA's agreement last month to supply AI chips to the Middle East is also stimulating memory demand. The Middle East has not traditionally been a region for AI chip demand. This suggests the possibility of HBM (High Bandwidth Memory) demand diversifying from its current North America-centric focus to regions like the Middle East and Europe.\n\nAn industry source explained, \"While it was initially expected that pre-order demand from Chinese set manufacturers would be a temporary boost in Q2, there's a growing observation that additional orders will continue into the second half of the year. Simultaneously, with expanding AI investment demand from major US tech companies, the upward trend in memory prices is likely to continue for some time.\"\n\nDue to these factors, the Q2 earnings outlooks for Samsung Electronics and SK hynix are being sharply revised upwards. According to financial information provider FnGuide, as of June 5, the consensus operating profit forecasts from securities firms for Samsung Electronics and SK hynix were KRW 6.8692 trillion and KRW 8.7538 trillion, respectively, representing double-digit increases compared to a month ago. This is because the average selling price (ASP) increase for memory is larger than previously expected, and demand for high-spec AI memory, such as HBM, is steadily growing.\n\nNoh Geun-chang, a researcher at Hyundai Motor Securities, stated, \"Considering that the substantial demand for NVIDIA's Blackwell Ultra will likely materialize after 2026, memory demand is highly probable to continue into the second half of this year. Furthermore, with AI investments from Middle Eastern countries, the memory boom could extend longer than previously anticipated.\"\n\n---\n\nhttps://t.co/Y7oUQx8eHn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01170574264410118, "ret_signed_m1d": -0.01170574264410118, "alpha_spy_m1d": -0.010805271335338706, "alpha_c_m1d": -0.006821203398297726, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.004884539245803454, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010033522183809818, "ret_signed_p1d": -0.010033522183809818, "alpha_spy_p1d": -0.015703186852700557, "alpha_c_p1d": -0.01562779668708647, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005594274503276653, "ret_p1w": -0.043478263759740066, "ret_signed_p1w": -0.043478263759740066, "alpha_spy_p1w": -0.048480939334144435, "alpha_c_p1w": -0.058131945564243614, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.014653681804503549, "ret_p1m": 0.03305374138721051, "ret_signed_p1m": 0.03305374138721051, "alpha_spy_p1m": -0.004456185210016894, "alpha_c_p1m": -0.038689998953473914, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.07174374034068443, "ret_p3m": 0.17943112995664356, "ret_signed_p3m": 0.17943112995664356, "alpha_spy_p3m": 0.09378721833480896, "alpha_c_p3m": 0.08113318119917046, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.09829794875747311, "ret_p6m": 0.7474646642387253, "ret_signed_p6m": 0.7474646642387253, "alpha_spy_p6m": 0.6044551608537478, "alpha_c_p6m": 0.5361262238611317, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.21133844037759353, "price_path": [-0.0874, -0.0907, -0.0907, -0.0425, -0.0425, -0.0275, 0.0007, 0.0257, 0.0057, -0.0059, 0.0207, 0.0273, 0.0067, -0.0334, -0.0167, -0.0167, -0.0368, -0.0619, -0.1104, -0.1054, -0.1137, -0.0569, -0.0769, -0.0602, -0.0535, -0.0853, -0.0786, -0.0753, -0.0736, -0.0803, -0.0686, -0.0686, -0.0686, -0.0669, -0.0669, -0.0719, -0.0719, -0.092, -0.092, -0.092, -0.087, -0.087, -0.0836, -0.0368, -0.0485, -0.0401, -0.0418, -0.0502, -0.0669, -0.0652, -0.0686, -0.0853, -0.0936, -0.0853, -0.0987, -0.0652, -0.0619, -0.0602, -0.0502, -0.0502, -0.0334, -0.0117, -0.0117, 0.0, -0.01, 0.0017, -0.005, -0.0251, -0.0435, -0.0284, 0.0, -0.01, -0.005, -0.0301, 0.0117, 0.0251, 0.0067, 0.023, 0.0061, 0.0129, 0.023, 0.0734, 0.065, 0.0381, 0.0331, 0.0162, 0.0263, 0.0532, 0.0516, 0.0718, 0.0886, 0.1222, 0.129, 0.1407, 0.1104, 0.1172, 0.1104, 0.1088, 0.1845, 0.1878, 0.2215, 0.2013, 0.1592, 0.1727, 0.1761, 0.1576, 0.1862, 0.208, 0.1946, 0.1963, 0.2097, 0.2047, 0.2047, 0.1777, 0.1777, 0.1862, 0.1878, 0.2013, 0.203, 0.1828, 0.1878, 0.171, 0.1727, 0.1374, 0.1626, 0.1744, 0.1794, 0.1693, 0.1794, 0.203, 0.2215, 0.235, 0.2686, 0.2871, 0.3359, 0.3157, 0.351, 0.351, 0.4049, 0.4251, 0.4369, 0.4486, 0.4015, 0.423, 0.4179, 0.4534, 0.5168, 0.5168, 0.5168, 0.5168, 0.5168, 0.5168, 0.5954, 0.5768, 0.548, 0.6055, 0.6511, 0.6545, 0.6579, 0.6478, 0.6663, 0.6309, 0.6697, 0.7238, 0.6816, 0.6985, 0.7593, 0.8168, 0.8776, 0.7728, 0.7001, 0.6765, 0.6545, 0.7001, 0.7492, 0.7424, 0.7373, 0.6427, 0.7001, 0.6528, 0.6309, 0.7001, 0.6021, 0.6342, 0.6782, 0.7373, 0.7492, 0.6985, 0.7035, 0.7475]}
{"unit_id": "thread:1931607868111028587", "ticker_idx": 1, "ts": "2025-06-08T07:02:57+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the Q2 earnings outlooks for Samsung Electronics and SK hynix are being sharply revised upwards", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "DRAM prices rising 8-13% QoQ in Q3 on Chinese stockpiling + US AI demand; Samsung and SK Hynix earnings outlooks sharply revised upward; memory boom may extend beyond 2025.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: DRAM Prices to Continue Rising in Q3… Driven by Chinese Stockpiling and Expanding US AI Demand\n\nAccording to market research firm TrendForce, DRAM contract prices are expected to rise by 8-13% quarter-on-quarter in the third quarter. In particular, prices for general-purpose PC DDR4 products are projected to increase by 13-18%. This is an upward revision from initial forecasts of 5-10% for server DRAM and 3-8% for PC DRAM. The industry considers it extremely unusual for prices of older memory products to show stronger growth than new products.\n\nPrices for the newer DDR5 are also on the rise. This is attributed to the expanding investment in AI server data centers by major US tech companies, leading to a comprehensive increase in demand for high-performance memory. Previously, demand was limited to NVIDIA's high-performance graphics processing units (GPUs), but recently, Google and Microsoft are also expanding their own AI infrastructure, broadening the demand for DDR5.\n\nNVIDIA's agreement last month to supply AI chips to the Middle East is also stimulating memory demand. The Middle East has not traditionally been a region for AI chip demand. This suggests the possibility of HBM (High Bandwidth Memory) demand diversifying from its current North America-centric focus to regions like the Middle East and Europe.\n\nAn industry source explained, \"While it was initially expected that pre-order demand from Chinese set manufacturers would be a temporary boost in Q2, there's a growing observation that additional orders will continue into the second half of the year. Simultaneously, with expanding AI investment demand from major US tech companies, the upward trend in memory prices is likely to continue for some time.\"\n\nDue to these factors, the Q2 earnings outlooks for Samsung Electronics and SK hynix are being sharply revised upwards. According to financial information provider FnGuide, as of June 5, the consensus operating profit forecasts from securities firms for Samsung Electronics and SK hynix were KRW 6.8692 trillion and KRW 8.7538 trillion, respectively, representing double-digit increases compared to a month ago. This is because the average selling price (ASP) increase for memory is larger than previously expected, and demand for high-spec AI memory, such as HBM, is steadily growing.\n\nNoh Geun-chang, a researcher at Hyundai Motor Securities, stated, \"Considering that the substantial demand for NVIDIA's Blackwell Ultra will likely materialize after 2026, memory demand is highly probable to continue into the second half of this year. Furthermore, with AI investments from Middle Eastern countries, the memory boom could extend longer than previously anticipated.\"\n\n---\n\nhttps://t.co/Y7oUQx8eHn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019650687484922047, "ret_signed_m1d": -0.019650687484922047, "alpha_spy_m1d": -0.018750216176159573, "alpha_c_m1d": -0.00318906145111153, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006550252000495194, "ret_signed_p1d": 0.006550252000495194, "alpha_spy_p1d": 0.0008805873316044543, "alpha_c_p1d": -0.013336223323506147, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.08296944757355784, "ret_signed_p1w": 0.08296944757355784, "alpha_spy_p1w": 0.07796677199915347, "alpha_c_p1w": 0.05798489322927214, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.23144107256924995, "ret_signed_p1m": 0.23144107256924995, "alpha_spy_p1m": 0.19393114597202255, "alpha_c_p1m": 0.1282341898945223, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.1610633249950959, "ret_signed_p3m": 0.1610633249950959, "alpha_spy_p3m": 0.0754194133732613, "alpha_c_p3m": 0.033027833701728015, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 1.441948416482865, "ret_signed_p6m": 1.441948416482865, "alpha_spy_p6m": 1.2989389130978877, "alpha_c_p6m": 1.0431702971096883, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.133, -0.1295, -0.1086, -0.1021, -0.1151, -0.1042, -0.0846, -0.0606, -0.0781, -0.0933, -0.0672, -0.0977, -0.1313, -0.1688, -0.1413, -0.1374, -0.1518, -0.2058, -0.2816, -0.2612, -0.2808, -0.2014, -0.2119, -0.2145, -0.2128, -0.2415, -0.2372, -0.2372, -0.2302, -0.2424, -0.211, -0.2228, -0.1962, -0.2067, -0.2119, -0.2263, -0.2263, -0.1892, -0.1892, -0.1892, -0.1683, -0.1705, -0.1714, -0.15, -0.1348, -0.1021, -0.126, -0.1086, -0.1308, -0.1195, -0.126, -0.1417, -0.1282, -0.1151, -0.1173, -0.0933, -0.0742, -0.107, -0.0939, -0.0939, -0.0502, -0.0197, -0.0197, 0.0, 0.0066, 0.048, 0.0284, 0.0284, 0.083, 0.0873, 0.0764, 0.0742, 0.1223, 0.1332, 0.2162, 0.2489, 0.2795, 0.2402, 0.2751, 0.2467, 0.2183, 0.2162, 0.1812, 0.1834, 0.2314, 0.2271, 0.2969, 0.286, 0.31, 0.3035, 0.2926, 0.1769, 0.1747, 0.19, 0.1725, 0.1747, 0.1769, 0.1616, 0.1441, 0.1463, 0.1507, 0.1943, 0.1266, 0.1266, 0.1507, 0.1288, 0.1441, 0.1201, 0.1659, 0.1747, 0.214, 0.2074, 0.2074, 0.1681, 0.1485, 0.1157, 0.0699, 0.0961, 0.1332, 0.1419, 0.1354, 0.1742, 0.1764, 0.1195, 0.1392, 0.1479, 0.1611, 0.196, 0.2114, 0.2595, 0.3294, 0.3425, 0.4366, 0.4475, 0.5218, 0.4584, 0.5546, 0.5546, 0.535, 0.5787, 0.5634, 0.559, 0.4716, 0.5262, 0.5197, 0.5743, 0.7296, 0.7296, 0.7296, 0.7296, 0.7296, 0.7296, 0.8717, 0.8148, 0.7995, 0.8476, 0.9788, 1.0357, 1.1231, 1.0947, 1.1057, 1.0925, 1.2303, 1.3396, 1.2784, 1.4402, 1.4839, 1.4446, 1.7113, 1.5626, 1.532, 1.5933, 1.5364, 1.6501, 1.707, 1.6982, 1.6763, 1.4489, 1.6501, 1.4927, 1.4577, 1.4971, 1.2784, 1.274, 1.2696, 1.2915, 1.3807, 1.3194, 1.3544, 1.4419]}
{"unit_id": "thread:1931607868111028587", "ticker_idx": 2, "ts": "2025-06-08T07:02:57+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the upward trend in memory prices is likely to continue for some time", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "DRAM prices rising 8-13% QoQ in Q3 on Chinese stockpiling + US AI demand; Samsung and SK Hynix earnings outlooks sharply revised upward; memory boom may extend beyond 2025.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: DRAM Prices to Continue Rising in Q3… Driven by Chinese Stockpiling and Expanding US AI Demand\n\nAccording to market research firm TrendForce, DRAM contract prices are expected to rise by 8-13% quarter-on-quarter in the third quarter. In particular, prices for general-purpose PC DDR4 products are projected to increase by 13-18%. This is an upward revision from initial forecasts of 5-10% for server DRAM and 3-8% for PC DRAM. The industry considers it extremely unusual for prices of older memory products to show stronger growth than new products.\n\nPrices for the newer DDR5 are also on the rise. This is attributed to the expanding investment in AI server data centers by major US tech companies, leading to a comprehensive increase in demand for high-performance memory. Previously, demand was limited to NVIDIA's high-performance graphics processing units (GPUs), but recently, Google and Microsoft are also expanding their own AI infrastructure, broadening the demand for DDR5.\n\nNVIDIA's agreement last month to supply AI chips to the Middle East is also stimulating memory demand. The Middle East has not traditionally been a region for AI chip demand. This suggests the possibility of HBM (High Bandwidth Memory) demand diversifying from its current North America-centric focus to regions like the Middle East and Europe.\n\nAn industry source explained, \"While it was initially expected that pre-order demand from Chinese set manufacturers would be a temporary boost in Q2, there's a growing observation that additional orders will continue into the second half of the year. Simultaneously, with expanding AI investment demand from major US tech companies, the upward trend in memory prices is likely to continue for some time.\"\n\nDue to these factors, the Q2 earnings outlooks for Samsung Electronics and SK hynix are being sharply revised upwards. According to financial information provider FnGuide, as of June 5, the consensus operating profit forecasts from securities firms for Samsung Electronics and SK hynix were KRW 6.8692 trillion and KRW 8.7538 trillion, respectively, representing double-digit increases compared to a month ago. This is because the average selling price (ASP) increase for memory is larger than previously expected, and demand for high-spec AI memory, such as HBM, is steadily growing.\n\nNoh Geun-chang, a researcher at Hyundai Motor Securities, stated, \"Considering that the substantial demand for NVIDIA's Blackwell Ultra will likely materialize after 2026, memory demand is highly probable to continue into the second half of this year. Furthermore, with AI investments from Middle Eastern countries, the memory boom could extend longer than previously anticipated.\"\n\n---\n\nhttps://t.co/Y7oUQx8eHn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01763257724948845, "ret_signed_m1d": -0.01763257724948845, "alpha_spy_m1d": -0.016732105940725977, "alpha_c_m1d": -0.0011709512156779338, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008422801136441152, "ret_signed_p1d": 0.008422801136441152, "alpha_spy_p1d": 0.002753136467550413, "alpha_c_p1d": -0.011463674187560188, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.03987246510246708, "ret_signed_p1w": 0.03987246510246708, "alpha_spy_p1w": 0.03486978952806271, "alpha_c_p1w": 0.014887910758181376, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.12898545698485306, "ret_signed_p1m": 0.12898545698485306, "alpha_spy_p1m": 0.09147553038762565, "alpha_c_p1m": 0.025778574310125396, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.153687167442089, "ret_signed_p3m": 0.153687167442089, "alpha_spy_p3m": 0.06804325582025439, "alpha_c_p3m": 0.025651676148721103, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 1.1171126185425055, "ret_signed_p6m": 1.1171126185425055, "alpha_spy_p6m": 0.9741031151575281, "alpha_c_p6m": 0.7183344991693288, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1198, -0.1221, -0.0973, -0.0721, -0.0807, -0.071, -0.0522, -0.0608, -0.0666, -0.0838, -0.0724, -0.0833, -0.1095, -0.1397, -0.1195, -0.1185, -0.1728, -0.2281, -0.2586, -0.2586, -0.2309, -0.209, -0.2207, -0.2115, -0.2087, -0.234, -0.2319, -0.2308, -0.2341, -0.23, -0.2074, -0.1979, -0.1819, -0.1885, -0.1953, -0.2015, -0.1991, -0.1846, -0.1855, -0.1852, -0.1702, -0.1633, -0.1604, -0.1183, -0.1032, -0.0944, -0.1025, -0.0918, -0.1029, -0.1002, -0.1103, -0.1241, -0.1268, -0.1196, -0.1158, -0.0972, -0.0879, -0.1053, -0.0864, -0.0742, -0.051, -0.0245, -0.0176, 0.0, 0.0084, 0.0318, 0.0235, 0.0151, 0.0399, 0.0478, 0.0581, 0.0541, 0.0771, 0.0678, 0.1269, 0.1403, 0.1406, 0.1292, 0.1307, 0.1164, 0.1129, 0.1306, 0.1162, 0.1011, 0.129, 0.1154, 0.1446, 0.1542, 0.1439, 0.1529, 0.1438, 0.107, 0.1119, 0.1174, 0.0894, 0.0942, 0.0984, 0.0914, 0.1107, 0.1147, 0.1358, 0.1267, 0.0774, 0.0905, 0.1035, 0.0892, 0.1132, 0.1336, 0.1589, 0.1745, 0.1816, 0.1808, 0.1675, 0.1535, 0.1424, 0.1198, 0.1008, 0.1197, 0.1288, 0.1252, 0.1285, 0.1486, 0.1409, 0.1102, 0.1236, 0.1311, 0.1537, 0.1835, 0.1923, 0.2275, 0.2713, 0.312, 0.3745, 0.386, 0.4302, 0.4058, 0.4764, 0.4579, 0.475, 0.5017, 0.4864, 0.4742, 0.4306, 0.4759, 0.4824, 0.557, 0.6347, 0.6473, 0.6567, 0.6409, 0.6735, 0.6608, 0.7021, 0.7106, 0.6787, 0.7286, 0.8194, 0.839, 0.8825, 0.8562, 0.8545, 0.8631, 0.9591, 1.0168, 0.9877, 1.0615, 1.0885, 1.0938, 1.2359, 1.1012, 1.1254, 1.1404, 1.1129, 1.2123, 1.2109, 1.2171, 1.1842, 1.1066, 1.1781, 1.0694, 1.0427, 1.005, 0.9175, 0.9766, 0.9916, 1.0358, 1.0695, 1.0509, 1.0762, 1.1171]}
{"unit_id": "orig:1937649545124716754", "ticker_idx": 0, "ts": "2025-06-24T23:10:26+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rating Upgrade - Previous: Hold → Current: Buy - Target Price: USD 240 → USD 400 (approximately 58% upside from current price)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/endorses HSBC upgrade of AVGO to Buy, PT $400; ASIC explosive growth + multiple re-rating thesis with FY27e EPS 32% above consensus.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Broadcom (AVGO) - HSBC\n\n📌 Executive Summary (Investor Perspective)\n\nRating Upgrade \n- Previous: Hold → Current: Buy\n- Target Price: USD 240 → USD 400 (approximately 58% upside from current price)\n\n- Basis: Applying 32x P/E to FY27e EPS of USD 12.54 (10% premium over historical high of 29x)\n\n📈 Investment Points\n\n1. Explosive Growth Expectation and Visibility in ASIC Business\n- Expected expansion from 3 existing customers → 7 by FY27\n- Surging demand for Hyperscalers' in-house AI chips (ASIC) → Broadcom emerging as a key partner\n- FY26-27 Estimated Revenue: USD 28.4bn → 42.8bn (42-69% above consensus)\n- Industry-wide ASIC market CAGR: 96%, double the growth rate of the AI GPU market\n\n2. Explosive Rise in ASIC ASP (Average Selling Price)\n- Specification advancement from 5nm → 3nm, HBM3e → HBM4e\n\n- FY26 ASP expected to increase 92% y-y, FY27 ASP expected to increase 25%\n- Anticipated revenue and margin leverage effect due to price increase\n\n3. CoWoS (Advanced Packaging) Supply Advantage\n\n- Broadcom's CoWoS capacity growth rate of 48%, significantly higher than competitors Nvidia (11%) and AMD (15%)\n\n3. Mitigation of Concerns Regarding VMware and Wireless Businesses\n\n- Apple's wireless chip replacement speed slower than expected → Apple's share expected to increase to 88% by FY26\n- VMware's existing customer contract conversion is on a 3-year cycle → Low likelihood of revenue slowdown until FY26\n\n🔍 Performance and Valuation Outlook\n\nFor FY25e, revenue is expected to be $62.5 billion, with adjusted EPS at $6.82. The EV/EBITDA multiple is projected at 35.1x, and the free cash flow yield is 3.3%.\nIn FY26e, revenue is projected to grow to $82.6 billion, with adjusted EPS at $9.60, EV/EBITDA at 24.4x, and FCF yield rising to 4.3%.\nBy FY27e, revenue is forecasted to reach $102.7 billion, adjusted EPS at $12.54, EV/EBITDA at 19.0x, and FCF yield at 5.3%.\n\n- HSBC's estimated EPS for FY27 is 32% above consensus.\n- The company's valuation remains undervalued compared to its ASIC growth potential.\n⚠️ Risk Factors\n- Failure to meet expectations for new ASIC customer count\n- Lower-than-expected unit price premium\n\n- Accelerated decline in Apple wireless chip market share\n- Slowdown in VMware customer conversion speed\n📌 Conclusion\nBroadcom is judged to be capable of steep earnings growth + multiple re-rating by reflecting much stronger ASIC revenue growth and price premiums than previously anticipated. There is a high probability of a scenario similar to Nvidia's past AI GPU rally, involving an \"earnings revision → stock price rally,\" and the current stock price does not yet reflect this, making it a timely opportunity to buy.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.037911888243872705, "ret_signed_m1d": -0.037911888243872705, "alpha_spy_m1d": -0.026985386598029848, "alpha_c_m1d": -0.002484364558445251, "bench_spy_m1d": -0.010926501645842857, "bench_c_m1d": -0.03542752368542745, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003336193104735541, "ret_signed_p1d": 0.003336193104735541, "alpha_spy_p1d": 0.00277582847413127, "alpha_c_p1d": -0.010142499987725984, "bench_spy_p1d": 0.0005603646306042709, "bench_c_p1d": 0.013478693092461524, "ret_p1w": 0.003677247898428826, "ret_signed_p1w": 0.003677247898428826, "alpha_spy_p1w": -0.01423707372916061, "alpha_c_p1w": -0.012158319689394848, "bench_spy_p1w": 0.017914321627589436, "bench_c_p1w": 0.015835567587823673, "ret_p1m": 0.07552030637448981, "ret_signed_p1m": 0.07552030637448981, "alpha_spy_p1m": 0.030314380995297707, "alpha_c_p1m": 0.02002203361423649, "bench_spy_p1m": 0.0452059253791921, "bench_c_p1m": 0.05549827276025332, "ret_p3m": 0.3077300982426312, "ret_signed_p3m": 0.3077300982426312, "alpha_spy_p3m": 0.21089089357492763, "alpha_c_p3m": 0.14506528556939702, "bench_spy_p3m": 0.09683920466770357, "bench_c_p3m": 0.16266481267323418, "ret_p6m": 0.238118615622404, "ret_signed_p6m": 0.238118615622404, "alpha_spy_p6m": 0.12855434578593727, "alpha_c_p6m": -0.011200059829010955, "bench_spy_p6m": 0.10956426983646672, "bench_c_p6m": 0.24931867545141495, "price_path": [-0.3495, -0.3603, -0.3667, -0.3626, -0.3491, -0.4175, -0.4467, -0.417, -0.4099, -0.2997, -0.3483, -0.3119, -0.3254, -0.3232, -0.3396, -0.3533, -0.3533, -0.3713, -0.3586, -0.3309, -0.2884, -0.2726, -0.272, -0.2769, -0.272, -0.2536, -0.2298, -0.2408, -0.2432, -0.2254, -0.2142, -0.2125, -0.1619, -0.1209, -0.1221, -0.1201, -0.1353, -0.1277, -0.1237, -0.1311, -0.1281, -0.1349, -0.1349, -0.1087, -0.0944, -0.0848, -0.0844, -0.0593, -0.0285, -0.0125, -0.0169, -0.066, -0.0761, -0.0747, -0.0434, -0.0315, -0.0593, -0.0465, -0.0568, -0.0497, -0.0497, -0.0522, -0.0379, 0.0, 0.0033, 0.0243, 0.0212, 0.045, 0.0037, 0.0232, 0.0433, 0.0433, 0.0395, 0.0304, 0.0536, 0.0441, 0.0402, 0.0448, 0.0651, 0.0646, 0.086, 0.0742, 0.0927, 0.0562, 0.0755, 0.0946, 0.1001, 0.1157, 0.1276, 0.1473, 0.1135, 0.0943, 0.1287, 0.1106, 0.1437, 0.1516, 0.1562, 0.1521, 0.186, 0.1718, 0.1799, 0.1614, 0.1592, 0.1181, 0.1039, 0.0979, 0.1146, 0.1155, 0.1298, 0.1383, 0.1701, 0.1275, 0.1275, 0.1307, 0.1464, 0.1605, 0.2696, 0.3104, 0.2764, 0.4011, 0.3634, 0.3643, 0.3803, 0.3648, 0.3124, 0.3093, 0.3077, 0.2866, 0.2872, 0.2886, 0.2764, 0.2704, 0.2453, 0.2529, 0.2661, 0.2843, 0.285, 0.2741, 0.2776, 0.3121, 0.3103, 0.2328, 0.3546, 0.3069, 0.3342, 0.3449, 0.3266, 0.3263, 0.3013, 0.2923, 0.3075, 0.3449, 0.3749, 0.4164, 0.4658, 0.4297, 0.4037, 0.3768, 0.3366, 0.3633, 0.3504, 0.327, 0.361, 0.3366, 0.349, 0.2911, 0.3006, 0.3013, 0.2931, 0.346, 0.3171, 0.292, 0.4354, 0.4622, 0.5098, 0.5098, 0.5303, 0.4662, 0.4491, 0.4454, 0.447, 0.482, 0.5232, 0.543, 0.5683, 0.5433, 0.3669, 0.2905, 0.2961, 0.2381]}
{"unit_id": "orig:1933301215728959966", "ticker_idx": 0, "ts": "2025-06-12T23:11:43+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "earnings are likely to exceed market expectations. Guidance is also expected to meet or exceed market expectations. With HBM production capacity increasing and sales volume continuing to grow, positive guidance is anticipated", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish MU into FY3Q25 earnings (Jun 26): DDR4/DDR5 price tailwinds, solid server demand, and HBM ramp expected to drive beat-and-raise.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor: Micron’s US Investment Announcement\n\n[Daishin Securities – Semiconductors / Hyungkeun Ryu]\n\nMicron Details Its US Investment Plans\nYesterday, Micron announced its US investment plan. The core of this announcement is that Micron will establish six advanced fabs in the US (two in Idaho, up to four in New York) and expand its existing Virginia fab, with the mid- to long-term goal of producing 40% of its DRAM in the United States.\nExisting plants are reaching the limits of their fab space.\nAccording to the current design capacity, the remaining fab space is only about 50–60K wafers per month. As process technology advances, the length of the process line is increasing overall, making it difficult to respond to mid- to long-term production needs through conversion investments alone. Thus, new fab projects have been initiated. For the Idaho plant, production is expected to begin contributing from 1H27. The initial operation is estimated to be on a DRAM 1c line.\n\nMeanwhile, the Hiroshima Plant 2 is expected to be delayed compared to initial plans. Recruitment and infrastructure construction are overall being delayed, so it is deemed difficult to start operations within 2027. The New York plant is a long-term project set for after 2027. Thus, the short-term impact of this US investment decision on production will be limited.\n\nMicron’s Upcoming Earnings Announcement: Proof of Cycle Improvement\nMicron is scheduled to announce its FY3Q25 (March–May 2025 quarter) earnings on June 26th, Korea time. Through the quarterly results and guidance for the next quarter, we expect to get a sense of the short-term market environment. Due to the phase-out and steep production cuts of DDR4 in the supply industry, DDR4 prices are soaring (up over 30% quarter-on-quarter), and as customers normalize inventories amid tight supply-demand conditions, prices for DDR5 and LPDDR5 continue to rise. As such, earnings are likely to exceed market expectations.\nGuidance for the next quarter (June–August 2025) is also expected to be in line with market expectations. The rebound in demand for mobile and sets is continuing into CY2Q25, and purchasing orders for servers remain solid, so the short-term outlook remains favorable. With HBM production capacity increasing and sales volume continuing to grow, Micron is expected to provide positive guidance. We believe that this earnings release will strengthen confidence in the short-term profit growth of Korean semiconductors.\n\nMicron Technology Announces US Investment\nYesterday, Micron announced a new US investment plan. They plan to invest a total of $200 billion (with $150 billion in manufacturing and $50 billion in R&D), and the core of the new US investment plan is to establish six advanced fabs in the US (two in Idaho, up to four in New York) and expand the existing Virginia fab, with the goal of producing 40% of their DRAM in the US in the mid- to long-term.\n\nWhat Is the Background of This Fab Expansion?\nSo far, Micron has expanded its capacity mainly through M&A. However, as the trend of treating semiconductors as strategic assets strengthens and domestic protectionist policies for the semiconductor industry are reinforced, it is becoming increasingly difficult to expand capacity via M&A. With the DRAM industry structure consolidated into three major players (Samsung, SK hynix, Micron) plus China, there are fewer new fab sites to acquire via M&A.\nThe option of acquiring Nanya’s fab in Taiwan is under consideration, but since Nanya continues to focus on consumer products, the feasibility of that M&A is considered low.\n\nAs available space at existing fabs decreases, it seems inevitable that Micron would choose to build new fabs. The remaining space at current fabs is estimated to be just 50–60K wafers per month (with current capacity at 320–330K wafers/month and maximum capacity at 380K). With process miniaturization, process lines are getting longer overall, so increasing production just through conversion investments is not sufficient for mid- to long-term production needs. Therefore, Micron is actively pursuing new fab projects.\n\n- Currently, three new fab projects are underway (for front-end process): Hiroshima Plant 2 in Japan, a new plant in Idaho, and the New York project. Among these, the Idaho plant is expected to come online first, contributing to production starting in 1H27 (wafer input in early 2027, wafer output in 1H27), initially operating a DRAM 1c line.\n\n- The start-up of Hiroshima Plant 2 is expected to be delayed compared to the original plan. As the Trump administration’s onshoring policy for semiconductor supply chains strengthens, the burden of ramping up US fabs has likely increased. For Hiroshima Plant 2, overall hiring and infrastructure buildout are believed to be delayed, and given the current situation, it is judged unlikely to start up within 2027.\n\n- The New York project is a long-term one and is unlikely to contribute to production within the next three years (2025–2027).\n- As for the Taiwan site, while there is available land, due to geopolitical risk, fab investment has not yet been considered.\n\nWill Korean Companies Also Invest in US Front-End Fabs?\nThe US’s semiconductor strategy since the Biden administration has been to build a domestic semiconductor supply chain to enhance supply chain stability and stimulate the economy directly and indirectly. The use of tariffs despite an underdeveloped domestic semiconductor manufacturing supply chain is ultimately a strategic choice to accelerate the localization of fab construction in the US.\n\nHowever, there are practical concerns about mass production of memory semiconductors in the US. While government subsidies are expected via the CHIPS Act (potentially reducing building depreciation costs), the operation efficiency of US fabs is considered inferior to that of Asia. Because semiconductor fabs operate 24/7, managing fixed costs such as labor and electricity is critical, but the US faces significantly higher fixed costs compared to Asia. In the memory semiconductor industry, where costs remain critical, an increase in fixed costs is a burden for building new production lines.\n\nEven if Korean semiconductor companies decide to build new US front-end fabs, it is considered realistically impossible for the new fab to come online within the Trump administration’s term. Given the current US construction market, it is expected to take at least 40 months to complete a fab, and considering the time needed to make an investment decision, it is realistically unlikely for the fab to be operational within the administration’s term. Investment in back-end process lines or R&D facilities, which require shorter construction times than front-end fabs, seems possible.\n\nAre There Any Disadvantages?\nUltimately, I believe any disadvantages will be limited. There is still time before Micron’s US projects come online, and it will take at least until after 2028 for the US production lines to ramp up enough to fully meet local demand. This means the supply-side still holds the initiative, and the effectiveness of the Trump administration’s semiconductor tariff policy will be limited accordingly.\n\nMicron’s Upcoming Earnings Announcement: Impact on Korean Semiconductors?\nMicron is scheduled to announce its FY3Q25 (March–May 2025 quarter) earnings at 5:30 a.m. KST on June 26. Through the quarterly results and next-quarter guidance, we believe we can get a sense of the memory semiconductor market. As DDR4 has been phased out and production drastically reduced in the supply industry, DDR4 prices have surged, and with customer inventory normalization and tight supply-demand conditions, prices for DDR5/LPDDR5 continue to rise. This means that earnings are likely to exceed market expectations.\nGuidance is also expected to meet or exceed market expectations. The rebound in set demand (restocking demand from customer inventory normalization, preparations for new product launches in the second half of the year, etc.) continues into CY2Q25, and purchasing orders for servers remain solid, so the short-term outlook is expected to remain favorable. With HBM production capacity increasing and sales volume continuing to grow, positive guidance is anticipated, which we believe will lead to increased confidence in the short-term earnings of Korean semiconductor companies.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0012911097738207422, "ret_signed_m1d": -0.0012911097738207422, "alpha_spy_m1d": 0.00266738437259173, "alpha_c_m1d": 0.003535003680910309, "bench_spy_m1d": -0.003958494146412472, "bench_c_m1d": -0.004826113454731051, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004992304290626248, "ret_signed_p1d": -0.004992304290626248, "alpha_spy_p1d": 0.006187760105841211, "alpha_c_p1d": 0.01841642829213086, "bench_spy_p1d": -0.01118006439646746, "bench_c_p1d": -0.023408732582757108, "ret_p1w": 0.048545372035644574, "ret_signed_p1w": 0.048545372035644574, "alpha_spy_p1w": 0.05899658188000734, "alpha_c_p1w": 0.050673417482961614, "bench_spy_p1w": -0.010451209844362763, "bench_c_p1w": -0.0021280454473170396, "ret_p1m": 0.0728801648175943, "ret_signed_p1m": 0.0728801648175943, "alpha_spy_p1m": 0.036915569070615994, "alpha_c_p1m": -0.019614604062105467, "bench_spy_p1m": 0.0359645957469783, "bench_c_p1m": 0.09249476887969976, "ret_p3m": 0.16515159513999267, "ret_signed_p3m": 0.16515159513999267, "alpha_spy_p3m": 0.08481603573952268, "alpha_c_p3m": 0.033439623119026596, "bench_spy_p3m": 0.08033555940046999, "bench_c_p3m": 0.13171197202096607, "ret_p6m": 1.0450333612391387, "ret_signed_p6m": 1.0450333612391387, "alpha_spy_p6m": 0.9027993586404981, "alpha_c_p6m": 0.6601956588767675, "bench_spy_p6m": 0.1422340025986406, "bench_c_p6m": 0.38483770236237125, "price_path": [-0.1137, -0.1256, -0.1227, -0.1146, -0.1858, -0.1667, -0.1904, -0.208, -0.2164, -0.2398, -0.2521, -0.2364, -0.2374, -0.3601, -0.4429, -0.4115, -0.4359, -0.3297, -0.3971, -0.4014, -0.3887, -0.3885, -0.4033, -0.4078, -0.4078, -0.4255, -0.3957, -0.3723, -0.3336, -0.3133, -0.3238, -0.3383, -0.3377, -0.3306, -0.3052, -0.3078, -0.307, -0.2889, -0.2671, -0.261, -0.2056, -0.1657, -0.1795, -0.1784, -0.1565, -0.1509, -0.1556, -0.1751, -0.1838, -0.1963, -0.1963, -0.1704, -0.1721, -0.1668, -0.187, -0.1549, -0.1199, -0.1113, -0.0851, -0.0656, -0.045, -0.0176, -0.0013, 0.0, -0.005, 0.0315, 0.0358, 0.0485, 0.0485, 0.0639, 0.0508, 0.101, 0.0953, 0.0845, 0.0739, 0.0609, 0.0405, 0.0479, 0.0526, 0.0526, 0.0332, 0.0719, 0.0532, 0.0606, 0.0729, 0.0219, 0.0348, 0.0031, -0.0242, -0.0145, -0.0245, -0.059, -0.0538, -0.0374, -0.0414, -0.0415, -0.0354, -0.0115, -0.0597, -0.0964, -0.0715, -0.0604, -0.0628, -0.0362, 0.0243, 0.0659, 0.1006, 0.0706, 0.0794, 0.0413, 0.0644, 0.0515, 0.0098, -0.0024, 0.0139, 0.003, 0.0037, 0.0145, 0.0511, 0.0253, 0.0253, 0.0208, 0.0228, 0.0701, 0.1318, 0.1326, 0.1652, 0.2062, 0.2972, 0.3546, 0.3593, 0.3683, 0.3784, 0.4551, 0.402, 0.4183, 0.4337, 0.3932, 0.3512, 0.3549, 0.4121, 0.4415, 0.5693, 0.5831, 0.6193, 0.6462, 0.6008, 0.6943, 0.658, 0.5655, 0.6618, 0.6126, 0.6547, 0.746, 0.7447, 0.7825, 0.7439, 0.711, 0.782, 0.8881, 0.8974, 0.913, 0.9537, 0.9312, 0.9291, 1.0233, 0.8796, 1.0474, 1.0546, 1.0511, 1.1837, 1.0786, 1.1112, 1.0427, 1.1279, 1.0858, 0.9699, 0.9476, 0.736, 0.7877, 0.9305, 0.9356, 0.985, 0.985, 1.0387, 1.073, 1.0646, 1.0187, 0.9539, 1.045]}
{"unit_id": "orig:1929684382824255612", "ticker_idx": 0, "ts": "2025-06-02T23:39:43+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "These initiatives are set to significantly expand Broadcom's market opportunity, with projected revenues rising from approximately $75 billion in FY2027 to over $100 billion in FY2028", "tweet_type": "analysis", "timeline": "years", "summary": "Long AVGO: AI ASIC dominance, TPU/MTIA/OpenAI ramp, and revenue projected to grow from ~$75B in FY2027 to >$100B in FY2028.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI ASIC and Networking Chips\n\nBroadcom’s dominance in the AI ASIC market is striking, with over 100 chip designs completed across 7nm, 5nm, 3nm, and 2nm process nodes—firmly securing its No.1 position with over 80% market share. In the high-performance switching and routing chip sector, Broadcom also holds a commanding lead, boasting more than 80% market share with its Tomahawk 5 and Jericho 3 chipsets.\n\nBroadcom is set to begin mass production of its next-generation 3nm Tomahawk 6 switch chipset in the second half of 2025. The chip supports 102Tbps switching throughput (200Gbps SERDES) and is positioned as the highest-performing Ethernet networking chipset globally, placing Broadcom 1–2 generations ahead of competitors.\n\nMeanwhile, Broadcom’s AI ASIC programs are progressing rapidly.\nGoogle’s TPUv6 3nm ASIC has entered mass production, featuring\ntwo 3nm compute dies and eight HBM3e stacks per package,\nand is hailed as “the world’s most powerful custom AI XPU.”\nThe project is expected to bring Broadcom over $15 billion in lifetime revenue,\nwith revenue contributions starting in July 2025 and full-scale production in October and into the first half of 2026.\n\nBroadcom has completed design for the next-gen TPUv7 3nm,\nand has already started collaborating with Google on the TPUv8 2nm project,\ntargeting mass production in 2027–2028.\n\nWith Meta, Broadcom co-designed 7nm and 5nm MTIA AI chips (mainly for inference workloads),\nand has sampled its next-gen 3nm MTIA chip,\nwhich is powerful enough to support both AI training and inference,\nwith mass production expected in 2026.\n\nAdditionally, Broadcom’s AI ASIC projects with OpenAI and SoftBank/ARM are also steadily progressing,\nutilizing TSMC’s 3D SoIC chip stacking technology.\nThese projects are expected to tape out in the second half of 2025,\nwith mass production slated for mid to late 2026.\n\nThese initiatives are set to significantly expand Broadcom’s market opportunity,\nwith projected revenues rising from approximately $75 billion in FY2027\nto over $100 billion in FY2028.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02669779229960967, "ret_signed_m1d": -0.02669779229960967, "alpha_spy_m1d": -0.02109644300957947, "alpha_c_m1d": -0.01210669770391748, "bench_spy_m1d": -0.0056013492900302, "bench_c_m1d": -0.014591094595692189, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03272891970555247, "ret_signed_p1d": 0.03272891970555247, "alpha_spy_p1d": 0.027026329917364977, "alpha_c_p1d": 0.010164222099338316, "bench_spy_p1d": 0.005702589788187495, "bench_c_p1d": 0.022564697606214157, "ret_p1w": -0.01781197001345769, "ret_signed_p1w": -0.01781197001345769, "alpha_spy_p1w": -0.02957138717649277, "alpha_c_p1w": -0.07395653063580931, "bench_spy_p1w": 0.01175941716303508, "bench_c_p1w": 0.056144560622351625, "ret_p1m": 0.06695785923365705, "ret_signed_p1m": 0.06695785923365705, "alpha_spy_p1m": 0.021799285054004525, "alpha_c_p1m": -0.06678639674803932, "bench_spy_p1m": 0.04515857417965252, "bench_c_p1m": 0.13374425598169637, "ret_p3m": 0.24392444966750992, "ret_signed_p3m": 0.24392444966750992, "alpha_spy_p3m": 0.14585240413761635, "alpha_c_p3m": 0.015482271028831596, "bench_spy_p3m": 0.09807204552989357, "bench_c_p3m": 0.22844217863867833, "ret_p6m": 0.5544108023198013, "ret_signed_p6m": 0.5544108023198013, "alpha_spy_p6m": 0.4090067407634954, "alpha_c_p6m": 0.15691797276839958, "bench_spy_p6m": 0.14540406155630592, "bench_c_p6m": 0.39749282955140175, "price_path": [-0.232, -0.2807, -0.2185, -0.2606, -0.238, -0.2214, -0.2329, -0.2162, -0.2203, -0.2437, -0.216, -0.2339, -0.2294, -0.231, -0.2431, -0.2792, -0.3085, -0.32, -0.3268, -0.3224, -0.3081, -0.3808, -0.4118, -0.3802, -0.3726, -0.2556, -0.3072, -0.2685, -0.2829, -0.2805, -0.2979, -0.3125, -0.3125, -0.3317, -0.3182, -0.2887, -0.2435, -0.2268, -0.2261, -0.2314, -0.2261, -0.2066, -0.1812, -0.193, -0.1955, -0.1765, -0.1646, -0.1629, -0.1091, -0.0655, -0.0667, -0.0646, -0.0808, -0.0727, -0.0685, -0.0763, -0.0731, -0.0804, -0.0804, -0.0525, -0.0373, -0.0271, -0.0267, 0.0, 0.0327, 0.0497, 0.0451, -0.0072, -0.0178, -0.0164, 0.0169, 0.0296, -0.0, 0.0136, 0.0027, 0.0103, 0.0103, 0.0075, 0.0227, 0.063, 0.0666, 0.0888, 0.0855, 0.1109, 0.067, 0.0878, 0.109, 0.109, 0.105, 0.0954, 0.12, 0.1099, 0.1058, 0.1107, 0.1322, 0.1317, 0.1545, 0.1419, 0.1615, 0.1228, 0.1433, 0.1636, 0.1695, 0.1861, 0.1987, 0.2196, 0.1837, 0.1633, 0.1999, 0.1806, 0.2158, 0.2242, 0.2291, 0.2248, 0.2608, 0.2457, 0.2543, 0.2346, 0.2323, 0.1885, 0.1735, 0.1671, 0.1849, 0.1858, 0.201, 0.2101, 0.2439, 0.1985, 0.1985, 0.202, 0.2187, 0.2336, 0.3497, 0.393, 0.3569, 0.4894, 0.4494, 0.4504, 0.4674, 0.4509, 0.3951, 0.3918, 0.3902, 0.3677, 0.3683, 0.3698, 0.3569, 0.3505, 0.3238, 0.3319, 0.3459, 0.3653, 0.366, 0.3544, 0.3581, 0.3948, 0.3929, 0.3106, 0.44, 0.3893, 0.4184, 0.4297, 0.4103, 0.4099, 0.3834, 0.3738, 0.3899, 0.4297, 0.4616, 0.5057, 0.5582, 0.5199, 0.4922, 0.4637, 0.4208, 0.4492, 0.4356, 0.4107, 0.4469, 0.4209, 0.4341, 0.3725, 0.3826, 0.3833, 0.3746, 0.4308, 0.4002, 0.3734, 0.5259, 0.5544]}
{"unit_id": "orig:1934435254460436832", "ticker_idx": 0, "ts": "2025-06-16T02:17:59+00:00", "symbol": "Getty Images", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If its value drops, it will keep dropping; there's no way it's going up", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Getty Images: generative AI undercuts image-licensing revenue; stock won't recover if it falls.", "resolved_tickers": ["GETY"], "resolution_method": "exchange_ticker", "resolution_reason": "Getty Images Holdings GETY US", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "My opinion on Getty Images: If its value drops, it will keep dropping; there's no way it's going up.\n\nI recently watched a shocking video on YouTube Shorts. It was a viral video introducing a website that converts copyrighted images into royalty-free generative AI images.\n\nWatching that, I realized that businesses selling images for profit might no longer be sustainable.\n\nEspecially among the small Korean media outlets I know, most are already using generative AI to create images for their articles instead of buying them.\n\nCan Getty continue its business? I really don't know. While places like Bloomberg or NYT might keep buying images from Getty, I can definitely foresee Getty's revenue steadily declining.\n\nThis is not a buy or sell recommendation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.033149140705165725, "ret_signed_m1d": 0.033149140705165725, "alpha_spy_m1d": 0.023724611053277944, "alpha_c_m1d": 0.016281370435640707, "bench_spy_m1d": -0.00942452965188778, "bench_c_m1d": -0.016867770269525018, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04419885427355441, "ret_signed_p1d": 0.04419885427355441, "alpha_spy_p1d": 0.03565382173697551, "alpha_c_p1d": 0.035381619135760545, "bench_spy_p1d": -0.0085450325365789, "bench_c_p1d": -0.008817235137793866, "ret_p1w": 0.02209949299826519, "ret_signed_p1w": -0.02209949299826519, "alpha_spy_p1w": -0.02335348830290196, "alpha_c_p1w": -0.023995478375814838, "bench_spy_p1w": -0.001253995304636768, "bench_c_p1w": -0.0018959853775496471, "ret_p1m": -0.08287291762440252, "ret_signed_p1m": 0.08287291762440252, "alpha_spy_p1m": 0.11821382826345972, "alpha_c_p1m": 0.10115216506063596, "bench_spy_p1m": 0.0353409106390572, "bench_c_p1m": 0.018279247436233437, "ret_p3m": 0.08287298348589056, "ret_signed_p3m": -0.08287298348589056, "alpha_spy_p3m": 0.011529036837601758, "alpha_c_p3m": 0.0352262815495763, "bench_spy_p3m": 0.09440202032349232, "bench_c_p3m": 0.11809926503546686, "ret_p6m": -0.20994475538384738, "ret_signed_p6m": 0.20994475538384738, "alpha_spy_p6m": 0.34978428454866206, "alpha_c_p6m": 0.3320399361890607, "bench_spy_p6m": 0.13983952916481468, "bench_c_p6m": 0.12209518080521331, "price_path": [0.2431, 0.2099, 0.2099, 0.2155, 0.1381, 0.0884, 0.0663, -0.011, -0.0442, -0.0442, -0.0497, -0.1657, -0.2541, -0.1934, -0.221, -0.116, -0.1657, -0.1381, -0.1105, -0.0939, -0.0829, -0.0442, -0.0442, 0.0055, 0.0497, 0.0166, 0.0221, 0.0829, 0.105, 0.0994, 0.0552, 0.0608, 0.0939, 0.0829, 0.0442, 0.0221, 0.0608, 0.0387, 0.1105, 0.0, 0.0055, -0.0276, -0.0055, 0.0, 0.0, -0.0221, -0.0221, -0.0497, -0.0497, -0.0221, -0.0331, 0.0, -0.0166, -0.0387, -0.0276, -0.0442, -0.105, -0.0829, -0.0994, -0.0663, 0.0442, 0.0387, -0.0331, 0.0, -0.0442, -0.0055, -0.0055, 0.0166, 0.0221, -0.0221, -0.0552, -0.0331, -0.0829, -0.0829, -0.0331, -0.0442, 0.0, 0.0, -0.0608, -0.0276, -0.011, 0.0055, -0.0331, -0.0608, -0.0829, -0.0773, -0.0387, -0.0497, 0.0276, 0.0829, 0.1105, 0.0663, 0.0221, 0.0166, 0.0166, -0.011, -0.0221, -0.0829, -0.0331, -0.0387, 0.0221, -0.0166, -0.0663, -0.0552, -0.0773, 0.011, -0.011, 0.0, 0.0663, 0.0166, 0.011, 0.0055, 0.0994, 0.0718, 0.0552, 0.0387, 0.011, 0.0166, 0.0166, -0.0221, -0.0221, -0.0055, 0.0276, 0.0497, 0.0331, 0.0276, 0.0829, 0.0773, 0.1271, 0.1436, 0.105, 0.1271, 0.1436, 0.1823, 0.1105, 0.0939, 0.0829, 0.1105, 0.0939, 0.0939, 0.1436, 0.1547, 0.2707, 0.2155, 0.1547, 0.2762, 0.221, 0.1547, 0.1934, 0.2044, 0.3039, 0.1989, 0.1823, 0.1105, 0.105, 0.1326, 0.1934, 0.1547, 0.1878, 0.1492, 0.1381, 0.1105, 0.0387, 0.0829, -0.0166, -0.0442, -0.0055, -0.0221, -0.0497, -0.0608, -0.0829, -0.1436, -0.1768, -0.1989, -0.1934, -0.2099, -0.221, -0.1657, -0.1713, -0.1215, -0.1271, -0.1271, -0.1492, -0.1934, -0.1878, -0.1934, -0.1878, -0.1934, -0.1989, -0.2099]}
{"unit_id": "quote:1924249045973061967", "ticker_idx": 0, "ts": "2025-05-18T23:41:38+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the AI up-cycle is expected to continue over the medium to long term", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPMorgan note: NVDA mid-long term AI cycle intact despite short-term supply mismatch; ASIC growth (Broadcom) slows in 2026 before reaccelerating in 2027.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "250519_AI Servers Update – JPMorgan\n[Key Takeaways]\n\n(1) AI server demand remains solid, but in the short term, there is a supply imbalance between upstream and downstream, which presents a short-term downside risk.\n\n(2) In the mid-to-long term, 1) macroeconomic uncertainty and 2) a shift in CSPs’ resource allocation stance—from proactive computing infrastructure buildout to optimization/utilization—lead to downward revisions in AI GPU shipment growth rates for 2026 and 2027.\n■ '26 AI GPU Shipments YoY +28% [Previous +38%]\n■ '27 AI GPU Shipments YoY +10% [Previous +17%]\n\n(3) While the ASIC market is expected to experience slower growth in the short term, growth will reaccelerate from 2027 as projects from Meta and OpenAI are reflected.\n[Contents]\n(1) Final demand for AI remains solid, but the gap between upstream production and system manufacturing is widening.\n\n(2) JPMorgan maintains its forecast of 5.5 million GPU modules produced by NVIDIA, but adjusts the GPU mix.\n\n(3) Specifically, the production share of H200 increases, while the proportion of Blackwell HGX products is reduced.\n\n(4) Over the past 6 months, adoption of AI inference has increased, and demand for AI training continues, indicating a solid foundation for AI demand.\n\n(5) The recent upward trend in CAPEX by U.S. CSPs also supports this view.\n\n(6) However, a mismatch has emerged between GPU module production [Upstream] and GPU board/rack assembly [Downstream].\n\n(7) In other words, a growing mismatch between GPU supply and actual demand within the supply chain is forming, potentially signaling a period of downside risk.\n\n(8) Considering the preemptive computing infrastructure built this year and more cautious IT budget execution, the growth rate of CAPEX for AI servers may moderate.\n\n(9) Accordingly, AI GPU shipment growth rates for 2026 and 2027 are revised downward by -10% and -7%, respectively.\n■ '26 shipment growth rate: YoY +28%; '27: +10%, factoring in average selling price (ASP) increases\n\n(10) While there are short-term risks from inventory and policies (e.g., tariffs), the AI up-cycle is expected to continue over the medium to long term.\n\n(11) The most critical timing will be Q3, when GB200 GPU rack production normalizes.\n\n(12) In 1Q25, production was delayed due to GB200 supply issues, but in 2Q25, system assembly yields are improving, and ramp-up is accelerating.\n\n(13) The key focus for customers is whether they will adopt the GB200 instead of the delayed GB300.\n(14) Our base case is: 1) continued learning curve improvement for GB200, 2) solid AI demand, and 3) projected GB rack shipments at approximately 25K units.\n■ 1Q25: 1.0–1.5K\n■ 2Q25: 4–5K\n■ 3Q25: 8–10K\n■ 4Q25: 8–10K\n\n(15) Meanwhile, NVIDIA's growth will ultimately be led by the four major U.S. hyperscalers, while ASIC growth is expected to slow.\n\n(16) According to our analysis, the combined data center CAPEX of the four major U.S. big tech firms will reach approximately $280 billion this year, consuming about 2.0 million Blackwell GPUs.\n\n(17) This is expected to account for around 40–50% of NVIDIA’s topline GPU shipment growth.\n\n(18) In addition to demand from these big tech firms, secondary CSPs are projected to drive total NVL72 GB rack demand to approximately 38K units.\n\n(19) In contrast, ASIC server growth is expected to slow next year.\n(20) This is because ASIC-based platforms like Google TPU and AWS Inferentia are already large, and macro uncertainties may have an impact.\n(21) However, ASIC-related projects from Meta and OpenAI are structurally likely to grow in the coming years.\n(22) Overall, 2026 ASIC shipments are expected to grow by YoY +10%, with a reacceleration of growth in 2027 driven by new projects.\n\n$NVDA $AVGO $TSM", "reply_text": null, "parent_text": null, "quoted_text": "「 AI Servers: End demand intact, but rising gap between upstream build and system production (2025.5.18)」\n\n📌 Key Summary\n\n🧠 1. AI Server Demand Remains Solid, but Supply Chain Mismatch Deepens\n• NVIDIA’s 2025 high-end GPU shipment forecast remains at 5.5 million units\n\n• https://t.co/cJucmPA4GS", "parent_author": null, "ret_m1d": -0.0012539090372190032, "ret_signed_m1d": -0.0012539090372190032, "alpha_spy_m1d": -0.00016118628637640242, "alpha_c_m1d": -0.002920544594848584, "bench_spy_m1d": -0.0010927227508426007, "bench_c_m1d": 0.001666635557629581, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00877770099864661, "ret_signed_p1d": -0.00877770099864661, "alpha_spy_p1d": -0.005415572948851088, "alpha_c_p1d": -0.0068671705895627655, "bench_spy_p1d": -0.003362128049795521, "bench_c_p1d": -0.0019105304090838438, "ret_p1w": -0.03157040792684862, "ret_signed_p1w": -0.03157040792684862, "alpha_spy_p1w": -0.005109995230715869, "alpha_c_p1w": 0.0032655541201260796, "bench_spy_p1w": -0.02646041269613275, "bench_c_p1w": -0.0348359620469747, "ret_p1m": 0.06314081585369724, "ret_signed_p1m": 0.06314081585369724, "alpha_spy_p1m": 0.058635319340986536, "alpha_c_p1m": -0.00018996653858782686, "bench_spy_p1m": 0.004505496512710705, "bench_c_p1m": 0.06333078239228507, "ret_p3m": 0.3427206516274137, "ret_signed_p3m": 0.3427206516274137, "alpha_spy_p3m": 0.255292453340318, "alpha_c_p3m": 0.115453456107115, "bench_spy_p3m": 0.0874281982870957, "bench_c_p3m": 0.22726719552029873, "ret_p6m": 0.4249783453579359, "ret_signed_p6m": 0.4249783453579359, "alpha_spy_p6m": 0.2702026622738942, "alpha_c_p6m": -0.0019961763406175503, "bench_spy_p6m": 0.15477568308404166, "bench_c_p6m": 0.42697452169855343, "price_path": [0.0269, 0.0334, -0.0085, -0.0391, -0.066, -0.0317, -0.1138, -0.0786, -0.1587, -0.1445, -0.1348, -0.1845, -0.1688, -0.211, -0.1978, -0.1463, -0.1475, -0.1025, -0.1183, -0.1486, -0.1331, -0.1257, -0.1318, -0.1044, -0.1098, -0.1609, -0.1781, -0.191, -0.2006, -0.1875, -0.1855, -0.2491, -0.3043, -0.2798, -0.2897, -0.1567, -0.2065, -0.1818, -0.1834, -0.1724, -0.2293, -0.2514, -0.2514, -0.2852, -0.2706, -0.2424, -0.2149, -0.1812, -0.198, -0.1958, -0.1966, -0.1767, -0.1554, -0.1604, -0.1625, -0.1365, -0.1342, -0.1396, -0.0927, -0.0416, -0.0017, -0.0055, -0.0013, 0.0, -0.0088, -0.0278, -0.0202, -0.0316, -0.0316, -0.0005, -0.0056, 0.0267, -0.0032, 0.0134, 0.0417, 0.0468, 0.0326, 0.0454, 0.0521, 0.0619, 0.0536, 0.0696, 0.0473, 0.0673, 0.0631, 0.0732, 0.0732, 0.0611, 0.0635, 0.091, 0.1383, 0.1435, 0.1637, 0.1655, 0.1309, 0.16, 0.1754, 0.1754, 0.1673, 0.1803, 0.2015, 0.2105, 0.2166, 0.2103, 0.2592, 0.2642, 0.2762, 0.2718, 0.2642, 0.2321, 0.2598, 0.2816, 0.2799, 0.3038, 0.2947, 0.3224, 0.3121, 0.2815, 0.3278, 0.315, 0.3235, 0.3335, 0.3477, 0.343, 0.3511, 0.3395, 0.3427, 0.3311, 0.3426, 0.2957, 0.2939, 0.2908, 0.313, 0.3264, 0.3409, 0.3396, 0.3291, 0.2849, 0.2849, 0.2598, 0.2586, 0.2663, 0.2321, 0.2416, 0.2597, 0.3081, 0.307, 0.3118, 0.3113, 0.2901, 0.2563, 0.3002, 0.3033, 0.3545, 0.3163, 0.3055, 0.3109, 0.3145, 0.3415, 0.3764, 0.3813, 0.3935, 0.3841, 0.3688, 0.3651, 0.3951, 0.4206, 0.3512, 0.3893, 0.3281, 0.3266, 0.3412, 0.3516, 0.3474, 0.3365, 0.33, 0.3438, 0.3741, 0.4127, 0.483, 0.5274, 0.4968, 0.4938, 0.5262, 0.4658, 0.4401, 0.3875, 0.388, 0.4684, 0.425]}
{"unit_id": "quote:1924249045973061967", "ticker_idx": 1, "ts": "2025-05-18T23:41:38+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASIC server growth is expected to slow next year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPMorgan note: NVDA mid-long term AI cycle intact despite short-term supply mismatch; ASIC growth (Broadcom) slows in 2026 before reaccelerating in 2027.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "250519_AI Servers Update – JPMorgan\n[Key Takeaways]\n\n(1) AI server demand remains solid, but in the short term, there is a supply imbalance between upstream and downstream, which presents a short-term downside risk.\n\n(2) In the mid-to-long term, 1) macroeconomic uncertainty and 2) a shift in CSPs’ resource allocation stance—from proactive computing infrastructure buildout to optimization/utilization—lead to downward revisions in AI GPU shipment growth rates for 2026 and 2027.\n■ '26 AI GPU Shipments YoY +28% [Previous +38%]\n■ '27 AI GPU Shipments YoY +10% [Previous +17%]\n\n(3) While the ASIC market is expected to experience slower growth in the short term, growth will reaccelerate from 2027 as projects from Meta and OpenAI are reflected.\n[Contents]\n(1) Final demand for AI remains solid, but the gap between upstream production and system manufacturing is widening.\n\n(2) JPMorgan maintains its forecast of 5.5 million GPU modules produced by NVIDIA, but adjusts the GPU mix.\n\n(3) Specifically, the production share of H200 increases, while the proportion of Blackwell HGX products is reduced.\n\n(4) Over the past 6 months, adoption of AI inference has increased, and demand for AI training continues, indicating a solid foundation for AI demand.\n\n(5) The recent upward trend in CAPEX by U.S. CSPs also supports this view.\n\n(6) However, a mismatch has emerged between GPU module production [Upstream] and GPU board/rack assembly [Downstream].\n\n(7) In other words, a growing mismatch between GPU supply and actual demand within the supply chain is forming, potentially signaling a period of downside risk.\n\n(8) Considering the preemptive computing infrastructure built this year and more cautious IT budget execution, the growth rate of CAPEX for AI servers may moderate.\n\n(9) Accordingly, AI GPU shipment growth rates for 2026 and 2027 are revised downward by -10% and -7%, respectively.\n■ '26 shipment growth rate: YoY +28%; '27: +10%, factoring in average selling price (ASP) increases\n\n(10) While there are short-term risks from inventory and policies (e.g., tariffs), the AI up-cycle is expected to continue over the medium to long term.\n\n(11) The most critical timing will be Q3, when GB200 GPU rack production normalizes.\n\n(12) In 1Q25, production was delayed due to GB200 supply issues, but in 2Q25, system assembly yields are improving, and ramp-up is accelerating.\n\n(13) The key focus for customers is whether they will adopt the GB200 instead of the delayed GB300.\n(14) Our base case is: 1) continued learning curve improvement for GB200, 2) solid AI demand, and 3) projected GB rack shipments at approximately 25K units.\n■ 1Q25: 1.0–1.5K\n■ 2Q25: 4–5K\n■ 3Q25: 8–10K\n■ 4Q25: 8–10K\n\n(15) Meanwhile, NVIDIA's growth will ultimately be led by the four major U.S. hyperscalers, while ASIC growth is expected to slow.\n\n(16) According to our analysis, the combined data center CAPEX of the four major U.S. big tech firms will reach approximately $280 billion this year, consuming about 2.0 million Blackwell GPUs.\n\n(17) This is expected to account for around 40–50% of NVIDIA’s topline GPU shipment growth.\n\n(18) In addition to demand from these big tech firms, secondary CSPs are projected to drive total NVL72 GB rack demand to approximately 38K units.\n\n(19) In contrast, ASIC server growth is expected to slow next year.\n(20) This is because ASIC-based platforms like Google TPU and AWS Inferentia are already large, and macro uncertainties may have an impact.\n(21) However, ASIC-related projects from Meta and OpenAI are structurally likely to grow in the coming years.\n(22) Overall, 2026 ASIC shipments are expected to grow by YoY +10%, with a reacceleration of growth in 2027 driven by new projects.\n\n$NVDA $AVGO $TSM", "reply_text": null, "parent_text": null, "quoted_text": "「 AI Servers: End demand intact, but rising gap between upstream build and system production (2025.5.18)」\n\n📌 Key Summary\n\n🧠 1. AI Server Demand Remains Solid, but Supply Chain Mismatch Deepens\n• NVIDIA’s 2025 high-end GPU shipment forecast remains at 5.5 million units\n\n• https://t.co/cJucmPA4GS", "parent_author": null, "ret_m1d": -0.008758637147815995, "ret_signed_m1d": 0.008758637147815995, "alpha_spy_m1d": 0.0076659143969733945, "alpha_c_m1d": 0.010425272705445576, "bench_spy_m1d": -0.0010927227508426007, "bench_c_m1d": 0.001666635557629581, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004552765187900709, "ret_signed_p1d": -0.004552765187900709, "alpha_spy_p1d": -0.00791489323769623, "alpha_c_p1d": -0.006463295596984553, "bench_spy_p1d": -0.003362128049795521, "bench_c_p1d": -0.0019105304090838438, "ret_p1w": -0.008281750668274745, "ret_signed_p1w": 0.008281750668274745, "alpha_spy_p1w": -0.018178662027858006, "alpha_c_p1w": -0.026554211378699955, "bench_spy_p1w": -0.02646041269613275, "bench_c_p1w": -0.0348359620469747, "ret_p1m": 0.08125565343919194, "ret_signed_p1m": -0.08125565343919194, "alpha_spy_p1m": -0.07675015692648124, "alpha_c_p1m": -0.017924871046906876, "bench_spy_p1m": 0.004505496512710705, "bench_c_p1m": 0.06333078239228507, "ret_p3m": 0.35265374986843945, "ret_signed_p3m": -0.35265374986843945, "alpha_spy_p3m": -0.26522555158134375, "alpha_c_p3m": -0.12538655434814072, "bench_spy_p3m": 0.0874281982870957, "bench_c_p3m": 0.22726719552029873, "ret_p6m": 0.5322934851437522, "ret_signed_p6m": -0.5322934851437522, "alpha_spy_p6m": -0.3775178020597105, "alpha_c_p6m": -0.10531896344519875, "bench_spy_p6m": 0.15477568308404166, "bench_c_p6m": 0.42697452169855343, "price_path": [-0.0112, -0.0198, -0.0548, -0.1011, -0.1244, -0.0795, -0.1449, -0.1379, -0.19, -0.1895, -0.1718, -0.2243, -0.1572, -0.2026, -0.1783, -0.1604, -0.1728, -0.1547, -0.1592, -0.1844, -0.1546, -0.1738, -0.169, -0.1707, -0.1837, -0.2227, -0.2543, -0.2667, -0.274, -0.2693, -0.2538, -0.3322, -0.3657, -0.3317, -0.3235, -0.1972, -0.2529, -0.2111, -0.2266, -0.2241, -0.2429, -0.2586, -0.2586, -0.2793, -0.2647, -0.2329, -0.1842, -0.1662, -0.1655, -0.1711, -0.1655, -0.1444, -0.117, -0.1297, -0.1324, -0.112, -0.0991, -0.0973, -0.0392, 0.0078, 0.0065, 0.0087, -0.0088, 0.0, 0.0046, -0.0039, -0.0004, -0.0083, -0.0083, 0.0218, 0.0382, 0.0492, 0.0496, 0.0784, 0.1137, 0.132, 0.127, 0.0707, 0.0592, 0.0607, 0.0966, 0.1103, 0.0784, 0.0931, 0.0813, 0.0895, 0.0895, 0.0865, 0.1029, 0.1464, 0.1502, 0.1742, 0.1706, 0.198, 0.1506, 0.173, 0.196, 0.196, 0.1916, 0.1813, 0.2078, 0.1969, 0.1925, 0.1978, 0.221, 0.2204, 0.245, 0.2314, 0.2526, 0.2108, 0.233, 0.2548, 0.2612, 0.2791, 0.2926, 0.3152, 0.2765, 0.2545, 0.2939, 0.2731, 0.3111, 0.3202, 0.3254, 0.3208, 0.3596, 0.3434, 0.3527, 0.3314, 0.3289, 0.2817, 0.2655, 0.2586, 0.2778, 0.2788, 0.2952, 0.3049, 0.3414, 0.2925, 0.2925, 0.2962, 0.3142, 0.3304, 0.4555, 0.5022, 0.4632, 0.6062, 0.563, 0.5641, 0.5824, 0.5646, 0.5045, 0.5009, 0.4992, 0.475, 0.4756, 0.4772, 0.4632, 0.4564, 0.4275, 0.4363, 0.4514, 0.4723, 0.4731, 0.4606, 0.4646, 0.5042, 0.5021, 0.4133, 0.5529, 0.4982, 0.5296, 0.5418, 0.5208, 0.5205, 0.4918, 0.4815, 0.4989, 0.5417, 0.5762, 0.6238, 0.6804, 0.639, 0.6092, 0.5784, 0.5322, 0.5629, 0.5481, 0.5213, 0.5603, 0.5323]}
{"unit_id": "quote:1937893341880049925", "ticker_idx": 0, "ts": "2025-06-25T15:19:11+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rating Upgrade - Previous: Hold → Current: Buy - Target Price: USD 240 → USD 400", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares HSBC upgrade of AVGO from Hold to Buy, PT raised to $400 (~58% upside) on 32x FY27e EPS.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "HSBC's AVGO Outlook Chart 1 https://t.co/YIkt1ogaPi", "reply_text": null, "parent_text": null, "quoted_text": "Broadcom (AVGO) - HSBC\n\n📌 Executive Summary (Investor Perspective)\n\nRating Upgrade \n- Previous: Hold → Current: Buy\n- Target Price: USD 240 → USD 400 (approximately 58% upside from current price)\n\n- Basis: Applying 32x P/E to FY27e EPS of USD 12.54 (10% premium over historical https://t.co/dAo3tvBbQE", "parent_author": null, "ret_m1d": -0.003325099929278874, "ret_signed_m1d": -0.003325099929278874, "alpha_spy_m1d": -0.0027650491313332415, "alpha_c_m1d": 0.009974334172619725, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02085763478710234, "ret_signed_p1d": 0.02085763478710234, "alpha_spy_p1d": 0.013033854912614995, "alpha_c_p1d": 0.01359016075186159, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": 0.01983740863252592, "ret_signed_p1w": 0.01983740863252592, "alpha_spy_p1w": -0.0021187320091662354, "alpha_c_p1w": -0.002146552934502166, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": 0.09091239090585002, "ret_signed_p1m": 0.09091239090585002, "alpha_spy_p1m": 0.045946009668403454, "alpha_c_p1m": 0.04512751583492225, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": 0.28233694763457606, "ret_signed_p3m": 0.28233694763457606, "alpha_spy_p3m": 0.1809257798120696, "alpha_c_p3m": 0.11166104351633344, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.24861219321601213, "ret_signed_p6m": 0.24861219321601213, "alpha_spy_p6m": 0.13129533940782379, "alpha_c_p6m": -0.01319741037722344, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.3625, -0.3688, -0.3647, -0.3513, -0.4194, -0.4485, -0.4189, -0.4118, -0.302, -0.3505, -0.3141, -0.3276, -0.3254, -0.3418, -0.3554, -0.3554, -0.3734, -0.3607, -0.3331, -0.2907, -0.275, -0.2744, -0.2793, -0.2744, -0.2561, -0.2323, -0.2433, -0.2457, -0.2279, -0.2168, -0.2151, -0.1647, -0.1238, -0.125, -0.123, -0.1382, -0.1306, -0.1266, -0.134, -0.131, -0.1378, -0.1378, -0.1117, -0.0974, -0.0878, -0.0875, -0.0624, -0.0318, -0.0158, -0.0201, -0.0691, -0.0791, -0.0778, -0.0466, -0.0347, -0.0625, -0.0497, -0.0599, -0.0528, -0.0528, -0.0554, -0.0411, -0.0033, 0.0, 0.0209, 0.0178, 0.0416, 0.0003, 0.0198, 0.0398, 0.0398, 0.036, 0.027, 0.0501, 0.0406, 0.0368, 0.0414, 0.0616, 0.0611, 0.0824, 0.0706, 0.089, 0.0527, 0.0719, 0.0909, 0.0965, 0.112, 0.1238, 0.1435, 0.1098, 0.0906, 0.125, 0.1069, 0.1399, 0.1478, 0.1524, 0.1483, 0.1821, 0.1679, 0.176, 0.1575, 0.1553, 0.1143, 0.1002, 0.0943, 0.1109, 0.1118, 0.1261, 0.1345, 0.1663, 0.1237, 0.1237, 0.1269, 0.1426, 0.1566, 0.2654, 0.3061, 0.2721, 0.3964, 0.3589, 0.3598, 0.3757, 0.3603, 0.308, 0.3049, 0.3034, 0.2823, 0.2829, 0.2843, 0.2722, 0.2662, 0.2411, 0.2487, 0.2619, 0.28, 0.2807, 0.2698, 0.2733, 0.3077, 0.3059, 0.2287, 0.3501, 0.3025, 0.3298, 0.3405, 0.3222, 0.3219, 0.297, 0.2881, 0.3032, 0.3404, 0.3704, 0.4117, 0.461, 0.425, 0.3991, 0.3723, 0.3321, 0.3588, 0.3459, 0.3226, 0.3565, 0.3322, 0.3445, 0.2868, 0.2962, 0.2969, 0.2888, 0.3415, 0.3127, 0.2877, 0.4306, 0.4574, 0.5048, 0.5048, 0.5252, 0.4613, 0.4443, 0.4406, 0.4422, 0.4771, 0.5182, 0.5378, 0.5631, 0.5381, 0.3624, 0.2862, 0.2918, 0.234, 0.2486]}
{"unit_id": "thread:1942028012075246065", "ticker_idx": 0, "ts": "2025-07-07T01:08:53+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GPGPU growth in 2026 — primarily driven by NVIDIA's Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin delays weigh on NVDA 2026 GPGPU growth vs ASICs; bullish Wistron assembly quality, MediaTek TPU co-design, Alchip/Marvell Trainium3 backend, and Unimicron as rising high-end substrate winner.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Recent Trends and Future Outlook in the AI Server Sector By WT\n\nNVIDIA’s GB200 NVL72 shipments are ramping steadily. Shipments reached around 0.5K cabinets in Q1 2025. WT forecasts Q2–Q4 shipments of 4.0K, 8.1K, and 11.3K cabinets respectively — totaling about 24K units for 2025, aligning with market expectations of 21–26K. WT's research indicates ODM shipments are stabilizing, but CSP design revisions have hindered rapid volume increases. Currently, only Corewave, OpenAI, and Oracle are receiving shipments smoothly. Thanks to accumulated experience, 3231 Wistron has significantly better quality in sub-L10 assembly than peers. Also, some CSPs are still struggling with minor design revisions.\n\nRubin hasn’t taped out yet, but foundries expect mass production to begin in Q2 2026. Estimated wafer starts for 2026 are 340–350K. The CoWoS-L reticle area is increasing from 3.5X to 5.5X, and HBM count per package will rise from 8 to 12. HBM4 12hi will be adopted, with no hybrid bonding in the initial phase, pushing capacity to 576GB. Based on current validation, Samsung’s participation is unlikely. With these wafer volumes, only 1.2 million Rubin chips may be available in 2026 — equivalent to around 16–17K VR144 cabinets. However, ODMs likely won’t receive the chips until late 2026, and CSP customers may not receive VR144 servers until 2H 2027 or even year-end. Rubin chips require second-gen CoWoS-L packaging and 12 HBM4 stacks. WT expects foundries to delay output, pushing VR144 availability to late 2027.\n\nCloud service providers are also beginning volume production of in-house ASICs. Google’s TPU V6p is shipping this year, though some industry insiders believe it’s been rebranded as V7. Volume production will begin in 2H 2025 with about 2–3 million units expected for the year. For 2026, TPU V7 or V8 may be co-designed with 2454 MTK. However, due to unresolved issues with Google, no tape-out has occurred yet. If resolved, 2026 TPU shipments could exceed 3 million units, driven by Gemini-related applications. AWS's Trainium2 is expected to ship around 1.7 million units in 2025, with a small contribution from the minor Trainium2.5 update by year-end. Mass production of Trainium3 is expected by Q2 2026; it has recently taped out and will be produced by Annapurna, with backend handled by 3661 Alchip and Marvell. Total 2026 shipments are expected to reach 2.3–2.5 million units. Meta's MTIA 1 will begin shipping in Q4 2025, with 1.5 million units projected for 2026, and MTIA 1.5 launching mid-2026 using Andes RISC-V. Microsoft’s Maia200 recently taped out, with mass production expected in early 2026 — as this is Microsoft's first ASIC chip, market expectations remain limited.\n\nIn conclusion, GPGPU growth in 2026 — primarily driven by NVIDIA’s Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.\n\nIn NVIDIA’s substrate supply chain, Ibiden holds 65–75% market share, 3037 Unimicron holds 15–25%, and 3189 Kinsus has 5–10%. Japanese leader Ibiden has nearly all its high-end capacity allocated to NVIDIA. 3037 Unimicron has a strong track record with Blackwell substrates and high-end OAM HDI server boards. It has also demonstrated flexibility in substrate line adjustments and continuous upgrades to its HDI and PCB lines, significantly improving its operational foundation and optimizing its product mix. This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance. As we enter the ASIC era, Unimicron's competitive advantage will increasingly stand out.\n\n---\n\nhttps://t.co/CMrR2CZzJv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00695140583761078, "ret_signed_m1d": -0.00695140583761078, "alpha_spy_m1d": 0.0005565607408903883, "alpha_c_m1d": 0.006846119860680178, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011122326495506485, "ret_signed_p1d": -0.011122326495506485, "alpha_spy_p1d": -0.011670042713251783, "alpha_c_p1d": 0.0021748019926659534, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.03684273062740595, "ret_signed_p1w": -0.03684273062740595, "alpha_spy_p1w": -0.03018881656877115, "alpha_c_p1w": -0.016968546658937544, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": 0.1265166085526297, "ret_signed_p1m": -0.1265166085526297, "alpha_spy_p1m": -0.11477150217340681, "alpha_c_p1m": -0.10027984068309848, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.19376037136403101, "ret_signed_p3m": -0.19376037136403101, "alpha_spy_p3m": -0.11256667330372783, "alpha_c_p3m": 0.015275885275286782, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.18529450285638682, "ret_signed_p6m": -0.18529450285638682, "alpha_spy_p6m": -0.07207928965415489, "alpha_c_p6m": 0.11739514582603539, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.2775, -0.3203, -0.299, -0.3004, -0.291, -0.3397, -0.3587, -0.3587, -0.3876, -0.3751, -0.351, -0.3275, -0.2985, -0.3129, -0.3111, -0.3117, -0.2947, -0.2765, -0.2808, -0.2825, -0.2603, -0.2583, -0.2629, -0.2228, -0.179, -0.1448, -0.148, -0.1444, -0.1433, -0.1508, -0.1671, -0.1606, -0.1704, -0.1704, -0.1438, -0.1481, -0.1204, -0.1461, -0.1319, -0.1076, -0.1032, -0.1154, -0.1045, -0.0987, -0.0903, -0.0974, -0.0837, -0.1028, -0.0856, -0.0892, -0.0806, -0.0806, -0.0909, -0.0889, -0.0653, -0.0248, -0.0203, -0.0031, -0.0016, -0.0312, -0.0063, 0.007, 0.007, 0.0, 0.0111, 0.0293, 0.037, 0.0422, 0.0368, 0.0787, 0.083, 0.0933, 0.0895, 0.083, 0.0555, 0.0792, 0.098, 0.0964, 0.117, 0.1091, 0.1329, 0.1241, 0.0978, 0.1375, 0.1265, 0.1338, 0.1424, 0.1546, 0.1505, 0.1575, 0.1476, 0.1503, 0.1404, 0.1502, 0.11, 0.1084, 0.1058, 0.1248, 0.1363, 0.1487, 0.1476, 0.1386, 0.1007, 0.1007, 0.0792, 0.0782, 0.0848, 0.0555, 0.0636, 0.0791, 0.1206, 0.1197, 0.1238, 0.1234, 0.1052, 0.0762, 0.1138, 0.1165, 0.1604, 0.1277, 0.1184, 0.123, 0.1261, 0.1493, 0.1792, 0.1833, 0.1938, 0.1857, 0.1726, 0.1694, 0.1952, 0.217, 0.1575, 0.1902, 0.1378, 0.1365, 0.149, 0.1579, 0.1543, 0.1449, 0.1393, 0.1512, 0.1771, 0.2102, 0.2705, 0.3085, 0.2822, 0.2797, 0.3075, 0.2557, 0.2337, 0.1886, 0.1891, 0.258, 0.2207, 0.2248, 0.1809, 0.2018, 0.1793, 0.1462, 0.1788, 0.1416, 0.1305, 0.1537, 0.1238, 0.1392, 0.1392, 0.1186, 0.1371, 0.1468, 0.135, 0.159, 0.1529, 0.1727, 0.1691, 0.1615, 0.1435, 0.1062, 0.1142, 0.1232, 0.0804, 0.1006, 0.1439, 0.161, 0.1958, 0.1921, 0.1921, 0.2042, 0.1896, 0.1853]}
{"unit_id": "thread:1942028012075246065", "ticker_idx": 1, "ts": "2025-07-07T01:08:53+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "3231 Wistron has significantly better quality in sub-L10 assembly than peers", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin delays weigh on NVDA 2026 GPGPU growth vs ASICs; bullish Wistron assembly quality, MediaTek TPU co-design, Alchip/Marvell Trainium3 backend, and Unimicron as rising high-end substrate winner.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Recent Trends and Future Outlook in the AI Server Sector By WT\n\nNVIDIA’s GB200 NVL72 shipments are ramping steadily. Shipments reached around 0.5K cabinets in Q1 2025. WT forecasts Q2–Q4 shipments of 4.0K, 8.1K, and 11.3K cabinets respectively — totaling about 24K units for 2025, aligning with market expectations of 21–26K. WT's research indicates ODM shipments are stabilizing, but CSP design revisions have hindered rapid volume increases. Currently, only Corewave, OpenAI, and Oracle are receiving shipments smoothly. Thanks to accumulated experience, 3231 Wistron has significantly better quality in sub-L10 assembly than peers. Also, some CSPs are still struggling with minor design revisions.\n\nRubin hasn’t taped out yet, but foundries expect mass production to begin in Q2 2026. Estimated wafer starts for 2026 are 340–350K. The CoWoS-L reticle area is increasing from 3.5X to 5.5X, and HBM count per package will rise from 8 to 12. HBM4 12hi will be adopted, with no hybrid bonding in the initial phase, pushing capacity to 576GB. Based on current validation, Samsung’s participation is unlikely. With these wafer volumes, only 1.2 million Rubin chips may be available in 2026 — equivalent to around 16–17K VR144 cabinets. However, ODMs likely won’t receive the chips until late 2026, and CSP customers may not receive VR144 servers until 2H 2027 or even year-end. Rubin chips require second-gen CoWoS-L packaging and 12 HBM4 stacks. WT expects foundries to delay output, pushing VR144 availability to late 2027.\n\nCloud service providers are also beginning volume production of in-house ASICs. Google’s TPU V6p is shipping this year, though some industry insiders believe it’s been rebranded as V7. Volume production will begin in 2H 2025 with about 2–3 million units expected for the year. For 2026, TPU V7 or V8 may be co-designed with 2454 MTK. However, due to unresolved issues with Google, no tape-out has occurred yet. If resolved, 2026 TPU shipments could exceed 3 million units, driven by Gemini-related applications. AWS's Trainium2 is expected to ship around 1.7 million units in 2025, with a small contribution from the minor Trainium2.5 update by year-end. Mass production of Trainium3 is expected by Q2 2026; it has recently taped out and will be produced by Annapurna, with backend handled by 3661 Alchip and Marvell. Total 2026 shipments are expected to reach 2.3–2.5 million units. Meta's MTIA 1 will begin shipping in Q4 2025, with 1.5 million units projected for 2026, and MTIA 1.5 launching mid-2026 using Andes RISC-V. Microsoft’s Maia200 recently taped out, with mass production expected in early 2026 — as this is Microsoft's first ASIC chip, market expectations remain limited.\n\nIn conclusion, GPGPU growth in 2026 — primarily driven by NVIDIA’s Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.\n\nIn NVIDIA’s substrate supply chain, Ibiden holds 65–75% market share, 3037 Unimicron holds 15–25%, and 3189 Kinsus has 5–10%. Japanese leader Ibiden has nearly all its high-end capacity allocated to NVIDIA. 3037 Unimicron has a strong track record with Blackwell substrates and high-end OAM HDI server boards. It has also demonstrated flexibility in substrate line adjustments and continuous upgrades to its HDI and PCB lines, significantly improving its operational foundation and optimizing its product mix. This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance. As we enter the ASIC era, Unimicron's competitive advantage will increasingly stand out.\n\n---\n\nhttps://t.co/CMrR2CZzJv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008474576271186418, "ret_signed_m1d": 0.008474576271186418, "alpha_spy_m1d": 0.0009666096926852497, "alpha_c_m1d": 0.00047206987613490625, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.008002506395051512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004237288135593209, "ret_signed_p1d": 0.004237288135593209, "alpha_spy_p1d": 0.004785004353338507, "alpha_c_p1d": -0.0011369411542387908, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.005374229289832, "ret_p1w": -0.008474576271186418, "ret_signed_p1w": -0.008474576271186418, "alpha_spy_p1w": -0.01512849032982122, "alpha_c_p1w": -0.011298958475517207, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.0028243822043307887, "ret_p1m": 0.05084745762711873, "ret_signed_p1m": 0.05084745762711873, "alpha_spy_p1m": 0.03910235124789585, "alpha_c_p1m": 0.030448981441267398, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.020398476185851333, "ret_p3m": 0.26694915254237284, "ret_signed_p3m": 0.26694915254237284, "alpha_spy_p3m": 0.18575545448206965, "alpha_c_p3m": 0.1429582629253836, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.12399088961698923, "ret_p6m": 0.2457627118644068, "ret_signed_p6m": 0.2457627118644068, "alpha_spy_p6m": 0.13254749866217486, "alpha_c_p6m": 0.10173544634417264, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.14402726552023415, "price_path": [-0.3879, -0.3272, -0.26, -0.2027, -0.1601, -0.183, -0.1764, -0.1396, -0.156, -0.1904, -0.1601, -0.1723, -0.1601, -0.1601, -0.1478, -0.1682, -0.1682, -0.1355, -0.1519, -0.1068, -0.1273, -0.1232, -0.1232, -0.1109, -0.1109, -0.0617, -0.0863, -0.0904, -0.1068, -0.1027, -0.074, -0.074, -0.074, -0.0494, -0.0576, -0.0535, -0.0453, -0.0453, -0.0576, -0.072, -0.0339, -0.0254, -0.0678, -0.0381, -0.0297, -0.0254, -0.0127, 0.0042, 0.0085, 0.0, -0.0169, -0.0212, 0.0042, 0.0042, 0.0169, 0.0636, 0.0593, 0.0254, 0.0381, 0.0169, 0.0169, 0.0254, 0.0085, 0.0, 0.0042, 0.0212, 0.0169, 0.0127, -0.0085, 0.0169, 0.0, 0.0, 0.0085, 0.0042, -0.0297, -0.0169, -0.0085, -0.0127, 0.0169, 0.0212, 0.0212, 0.0424, 0.0424, 0.0127, 0.0508, 0.0381, 0.0466, 0.0551, 0.0805, 0.0678, -0.0042, -0.0127, -0.0127, -0.0169, -0.0212, -0.0381, -0.0551, -0.0508, -0.0381, -0.0424, -0.0254, -0.0254, -0.0424, -0.072, -0.0678, -0.0678, -0.0424, -0.0212, -0.0297, -0.0169, 0.0169, 0.0085, 0.0169, 0.0127, 0.0, 0.0085, 0.0169, 0.0254, 0.0297, 0.0381, 0.0297, 0.0636, 0.0847, 0.0847, 0.1907, 0.2373, 0.2669, 0.3008, 0.3008, 0.322, 0.2754, 0.2797, 0.2797, 0.2331, 0.1695, 0.1695, 0.1822, 0.1568, 0.2034, 0.2076, 0.2161, 0.2119, 0.2119, 0.2415, 0.2542, 0.2924, 0.3136, 0.2754, 0.2161, 0.178, 0.1653, 0.2034, 0.1653, 0.1737, 0.2246, 0.178, 0.2203, 0.1907, 0.1695, 0.161, 0.161, 0.2119, 0.178, 0.178, 0.1822, 0.2076, 0.2331, 0.2246, 0.1949, 0.2119, 0.2288, 0.2373, 0.2797, 0.2627, 0.2669, 0.25, 0.2203, 0.2161, 0.1992, 0.178, 0.1695, 0.1822, 0.2203, 0.2415, 0.2288, 0.2331, 0.2331, 0.2415, 0.2373, 0.2458]}
{"unit_id": "thread:1942028012075246065", "ticker_idx": 2, "ts": "2025-07-07T01:08:53+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TPU V7 or V8 may be co-designed with 2454 MTK", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin delays weigh on NVDA 2026 GPGPU growth vs ASICs; bullish Wistron assembly quality, MediaTek TPU co-design, Alchip/Marvell Trainium3 backend, and Unimicron as rising high-end substrate winner.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Recent Trends and Future Outlook in the AI Server Sector By WT\n\nNVIDIA’s GB200 NVL72 shipments are ramping steadily. Shipments reached around 0.5K cabinets in Q1 2025. WT forecasts Q2–Q4 shipments of 4.0K, 8.1K, and 11.3K cabinets respectively — totaling about 24K units for 2025, aligning with market expectations of 21–26K. WT's research indicates ODM shipments are stabilizing, but CSP design revisions have hindered rapid volume increases. Currently, only Corewave, OpenAI, and Oracle are receiving shipments smoothly. Thanks to accumulated experience, 3231 Wistron has significantly better quality in sub-L10 assembly than peers. Also, some CSPs are still struggling with minor design revisions.\n\nRubin hasn’t taped out yet, but foundries expect mass production to begin in Q2 2026. Estimated wafer starts for 2026 are 340–350K. The CoWoS-L reticle area is increasing from 3.5X to 5.5X, and HBM count per package will rise from 8 to 12. HBM4 12hi will be adopted, with no hybrid bonding in the initial phase, pushing capacity to 576GB. Based on current validation, Samsung’s participation is unlikely. With these wafer volumes, only 1.2 million Rubin chips may be available in 2026 — equivalent to around 16–17K VR144 cabinets. However, ODMs likely won’t receive the chips until late 2026, and CSP customers may not receive VR144 servers until 2H 2027 or even year-end. Rubin chips require second-gen CoWoS-L packaging and 12 HBM4 stacks. WT expects foundries to delay output, pushing VR144 availability to late 2027.\n\nCloud service providers are also beginning volume production of in-house ASICs. Google’s TPU V6p is shipping this year, though some industry insiders believe it’s been rebranded as V7. Volume production will begin in 2H 2025 with about 2–3 million units expected for the year. For 2026, TPU V7 or V8 may be co-designed with 2454 MTK. However, due to unresolved issues with Google, no tape-out has occurred yet. If resolved, 2026 TPU shipments could exceed 3 million units, driven by Gemini-related applications. AWS's Trainium2 is expected to ship around 1.7 million units in 2025, with a small contribution from the minor Trainium2.5 update by year-end. Mass production of Trainium3 is expected by Q2 2026; it has recently taped out and will be produced by Annapurna, with backend handled by 3661 Alchip and Marvell. Total 2026 shipments are expected to reach 2.3–2.5 million units. Meta's MTIA 1 will begin shipping in Q4 2025, with 1.5 million units projected for 2026, and MTIA 1.5 launching mid-2026 using Andes RISC-V. Microsoft’s Maia200 recently taped out, with mass production expected in early 2026 — as this is Microsoft's first ASIC chip, market expectations remain limited.\n\nIn conclusion, GPGPU growth in 2026 — primarily driven by NVIDIA’s Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.\n\nIn NVIDIA’s substrate supply chain, Ibiden holds 65–75% market share, 3037 Unimicron holds 15–25%, and 3189 Kinsus has 5–10%. Japanese leader Ibiden has nearly all its high-end capacity allocated to NVIDIA. 3037 Unimicron has a strong track record with Blackwell substrates and high-end OAM HDI server boards. It has also demonstrated flexibility in substrate line adjustments and continuous upgrades to its HDI and PCB lines, significantly improving its operational foundation and optimizing its product mix. This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance. As we enter the ASIC era, Unimicron's competitive advantage will increasingly stand out.\n\n---\n\nhttps://t.co/CMrR2CZzJv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0078124825314769275, "ret_signed_m1d": 0.0078124825314769275, "alpha_spy_m1d": 0.00030451595297575906, "alpha_c_m1d": -0.00598504316681403, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0078124825314769275, "ret_signed_p1d": -0.0078124825314769275, "alpha_spy_p1d": -0.00726476631373163, "alpha_c_p1d": -0.021109611019649366, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.08203126101276448, "ret_signed_p1w": 0.08203126101276448, "alpha_spy_p1w": 0.07537734695412968, "alpha_c_p1w": 0.06215707704429607, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": 0.046874992404990046, "ret_signed_p1m": 0.046874992404990046, "alpha_spy_p1m": 0.035129886025767165, "alpha_c_p1m": 0.020638224535458827, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.0, "ret_signed_p3m": 0.0, "alpha_spy_p3m": -0.08119369806030319, "alpha_c_p3m": -0.2090362566393178, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.10937494987293372, "ret_signed_p6m": 0.10937494987293372, "alpha_spy_p6m": -0.0038402633292982102, "alpha_c_p6m": -0.19331469880948848, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.0921, -0.004, 0.0611, 0.0726, 0.0611, 0.0458, 0.0267, 0.0458, 0.0305, -0.004, 0.0496, 0.019, 0.0573, 0.0343, 0.0496, 0.0343, 0.0343, -0.004, -0.0078, -0.0193, -0.0231, -0.0116, 0.0152, 0.0075, 0.0382, 0.042, 0.0496, 0.0458, 0.0037, 0.0075, 0.0267, 0.0152, 0.0075, -0.0193, -0.0193, -0.0193, -0.0346, -0.0346, -0.0346, -0.0231, -0.0231, -0.0461, -0.0193, -0.0155, 0.0228, 0.0075, -0.0155, -0.0423, -0.0385, -0.0231, -0.0155, -0.0308, -0.0423, -0.0385, -0.027, -0.0078, -0.0116, -0.0155, -0.0423, -0.0231, -0.0078, -0.0078, 0.0078, 0.0, -0.0078, 0.0586, 0.0937, 0.1094, 0.082, 0.0898, 0.1016, 0.0859, 0.1016, 0.1133, 0.1172, 0.1172, 0.1172, 0.1172, 0.0937, 0.0625, 0.0781, 0.0703, 0.0703, 0.0391, 0.0469, 0.0312, 0.0586, 0.0547, 0.0586, 0.0625, 0.0703, 0.0859, 0.0703, 0.1016, 0.0859, 0.0586, 0.0664, 0.0664, 0.1016, 0.0937, 0.0937, 0.082, 0.0703, 0.0625, 0.0625, 0.0898, 0.082, 0.1211, 0.1172, 0.1797, 0.1641, 0.1562, 0.1602, 0.1602, 0.1992, 0.1797, 0.1797, 0.125, 0.0977, 0.0742, 0.0391, 0.0508, 0.0234, 0.0234, 0.0273, 0.0039, 0.0, 0.0312, 0.0312, 0.0352, 0.043, 0.0508, 0.0508, 0.0273, 0.0195, 0.0273, 0.0391, 0.0391, 0.0469, 0.0469, 0.0391, 0.0117, 0.0117, 0.043, 0.0273, 0.0156, 0.0234, 0.0234, 0.0156, 0.0273, 0.0391, 0.0078, -0.0156, -0.0234, -0.0313, -0.0273, -0.0273, -0.0391, -0.0391, -0.0859, -0.0937, -0.0742, -0.1055, -0.1016, -0.0742, 0.0156, 0.0469, 0.0898, 0.1289, 0.1055, 0.0937, 0.0977, 0.1133, 0.125, 0.1094, 0.1406, 0.0898, 0.0977, 0.1094, 0.1094, 0.1133, 0.1094, 0.1016, 0.0937, 0.0859, 0.0781, 0.0781, 0.082, 0.1094, 0.1094]}
{"unit_id": "thread:1942028012075246065", "ticker_idx": 3, "ts": "2025-07-07T01:08:53+00:00", "symbol": "Alchip", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "backend handled by 3661 Alchip and Marvell", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin delays weigh on NVDA 2026 GPGPU growth vs ASICs; bullish Wistron assembly quality, MediaTek TPU co-design, Alchip/Marvell Trainium3 backend, and Unimicron as rising high-end substrate winner.", "resolved_tickers": ["3661.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Alchip Technologies 3661.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Recent Trends and Future Outlook in the AI Server Sector By WT\n\nNVIDIA’s GB200 NVL72 shipments are ramping steadily. Shipments reached around 0.5K cabinets in Q1 2025. WT forecasts Q2–Q4 shipments of 4.0K, 8.1K, and 11.3K cabinets respectively — totaling about 24K units for 2025, aligning with market expectations of 21–26K. WT's research indicates ODM shipments are stabilizing, but CSP design revisions have hindered rapid volume increases. Currently, only Corewave, OpenAI, and Oracle are receiving shipments smoothly. Thanks to accumulated experience, 3231 Wistron has significantly better quality in sub-L10 assembly than peers. Also, some CSPs are still struggling with minor design revisions.\n\nRubin hasn’t taped out yet, but foundries expect mass production to begin in Q2 2026. Estimated wafer starts for 2026 are 340–350K. The CoWoS-L reticle area is increasing from 3.5X to 5.5X, and HBM count per package will rise from 8 to 12. HBM4 12hi will be adopted, with no hybrid bonding in the initial phase, pushing capacity to 576GB. Based on current validation, Samsung’s participation is unlikely. With these wafer volumes, only 1.2 million Rubin chips may be available in 2026 — equivalent to around 16–17K VR144 cabinets. However, ODMs likely won’t receive the chips until late 2026, and CSP customers may not receive VR144 servers until 2H 2027 or even year-end. Rubin chips require second-gen CoWoS-L packaging and 12 HBM4 stacks. WT expects foundries to delay output, pushing VR144 availability to late 2027.\n\nCloud service providers are also beginning volume production of in-house ASICs. Google’s TPU V6p is shipping this year, though some industry insiders believe it’s been rebranded as V7. Volume production will begin in 2H 2025 with about 2–3 million units expected for the year. For 2026, TPU V7 or V8 may be co-designed with 2454 MTK. However, due to unresolved issues with Google, no tape-out has occurred yet. If resolved, 2026 TPU shipments could exceed 3 million units, driven by Gemini-related applications. AWS's Trainium2 is expected to ship around 1.7 million units in 2025, with a small contribution from the minor Trainium2.5 update by year-end. Mass production of Trainium3 is expected by Q2 2026; it has recently taped out and will be produced by Annapurna, with backend handled by 3661 Alchip and Marvell. Total 2026 shipments are expected to reach 2.3–2.5 million units. Meta's MTIA 1 will begin shipping in Q4 2025, with 1.5 million units projected for 2026, and MTIA 1.5 launching mid-2026 using Andes RISC-V. Microsoft’s Maia200 recently taped out, with mass production expected in early 2026 — as this is Microsoft's first ASIC chip, market expectations remain limited.\n\nIn conclusion, GPGPU growth in 2026 — primarily driven by NVIDIA’s Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.\n\nIn NVIDIA’s substrate supply chain, Ibiden holds 65–75% market share, 3037 Unimicron holds 15–25%, and 3189 Kinsus has 5–10%. Japanese leader Ibiden has nearly all its high-end capacity allocated to NVIDIA. 3037 Unimicron has a strong track record with Blackwell substrates and high-end OAM HDI server boards. It has also demonstrated flexibility in substrate line adjustments and continuous upgrades to its HDI and PCB lines, significantly improving its operational foundation and optimizing its product mix. This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance. As we enter the ASIC era, Unimicron's competitive advantage will increasingly stand out.\n\n---\n\nhttps://t.co/CMrR2CZzJv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009316774227175806, "ret_signed_m1d": -0.009316774227175806, "alpha_spy_m1d": -0.016824740805676974, "alpha_c_m1d": -0.023114299925466764, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001552808460907995, "ret_signed_p1d": -0.001552808460907995, "alpha_spy_p1d": -0.0010050922431626974, "alpha_c_p1d": -0.014849936949080433, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.02795032268152764, "ret_signed_p1w": 0.02795032268152764, "alpha_spy_p1w": 0.021296408622892837, "alpha_c_p1w": 0.008076138713059233, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": 0.1645963191673494, "ret_signed_p1m": 0.1645963191673494, "alpha_spy_p1m": 0.15285121278812652, "alpha_c_p1m": 0.13835955129781818, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.0298500025908206, "ret_signed_p3m": 0.0298500025908206, "alpha_spy_p3m": -0.051343695469482586, "alpha_c_p3m": -0.1791862540484972, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.09255015495555852, "ret_signed_p6m": 0.09255015495555852, "alpha_spy_p6m": -0.020665058246673418, "alpha_c_p6m": -0.2101394937268637, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.396, -0.337, -0.3401, -0.3075, -0.2671, -0.3261, -0.3121, -0.3152, -0.3463, -0.3804, -0.3525, -0.3447, -0.3292, -0.3199, -0.3245, -0.3401, -0.3401, -0.2857, -0.2935, -0.2935, -0.3075, -0.2717, -0.2857, -0.2143, -0.163, -0.1444, -0.1553, -0.1522, -0.191, -0.1102, -0.0745, -0.0994, -0.0776, -0.1134, -0.1118, -0.1134, -0.1273, -0.1273, -0.1553, -0.1553, -0.1289, -0.1382, -0.132, -0.1537, -0.1444, -0.118, -0.1366, -0.1817, -0.1817, -0.1258, -0.1289, -0.1087, -0.1304, -0.1211, -0.1398, -0.1056, -0.1056, -0.0683, -0.0388, -0.0109, -0.0093, -0.014, -0.0093, 0.0, -0.0016, 0.045, 0.0745, 0.0652, 0.028, 0.0466, 0.118, 0.1661, 0.2205, 0.1739, 0.1553, 0.1879, 0.2345, 0.2065, 0.205, 0.1848, 0.1646, 0.2065, 0.2065, 0.1661, 0.1646, 0.1522, 0.1693, 0.1925, 0.2143, 0.2655, 0.2795, 0.3199, 0.337, 0.3137, 0.2531, 0.1366, 0.1755, 0.146, 0.2096, 0.236, 0.2671, 0.2422, 0.2593, 0.2158, 0.2054, 0.2477, 0.207, 0.2446, 0.2462, 0.2274, 0.2352, 0.2305, 0.1882, 0.1678, 0.1537, 0.1788, 0.1662, 0.1756, 0.1788, 0.16, 0.069, 0.0565, 0.0267, 0.0267, 0.0863, 0.0643, 0.0299, 0.0377, 0.0377, 0.0847, 0.08, 0.102, 0.102, 0.0565, -0.036, -0.0344, -0.0297, -0.0329, 0.0142, 0.0518, 0.0095, -0.0125, -0.0125, -0.0391, -0.0391, 0.0189, 0.0518, 0.0957, 0.0816, 0.08, 0.0471, 0.1505, 0.1176, 0.1647, 0.1396, 0.1035, 0.1051, 0.1004, 0.1317, 0.0189, 0.0408, -0.036, -0.0626, -0.0689, -0.0752, -0.025, 0.0299, 0.0361, 0.0314, 0.0722, 0.0001, -0.0062, -0.0109, -0.0219, 0.022, 0.0565, 0.033, 0.0095, -0.0062, -0.0548, 0.0032, -0.0062, 0.0079, 0.011, 0.0095, -0.0031, -0.0031, 0.0079, 0.0628, 0.0926]}
{"unit_id": "thread:1942028012075246065", "ticker_idx": 4, "ts": "2025-07-07T01:08:53+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "backend handled by 3661 Alchip and Marvell", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin delays weigh on NVDA 2026 GPGPU growth vs ASICs; bullish Wistron assembly quality, MediaTek TPU co-design, Alchip/Marvell Trainium3 backend, and Unimicron as rising high-end substrate winner.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Recent Trends and Future Outlook in the AI Server Sector By WT\n\nNVIDIA’s GB200 NVL72 shipments are ramping steadily. Shipments reached around 0.5K cabinets in Q1 2025. WT forecasts Q2–Q4 shipments of 4.0K, 8.1K, and 11.3K cabinets respectively — totaling about 24K units for 2025, aligning with market expectations of 21–26K. WT's research indicates ODM shipments are stabilizing, but CSP design revisions have hindered rapid volume increases. Currently, only Corewave, OpenAI, and Oracle are receiving shipments smoothly. Thanks to accumulated experience, 3231 Wistron has significantly better quality in sub-L10 assembly than peers. Also, some CSPs are still struggling with minor design revisions.\n\nRubin hasn’t taped out yet, but foundries expect mass production to begin in Q2 2026. Estimated wafer starts for 2026 are 340–350K. The CoWoS-L reticle area is increasing from 3.5X to 5.5X, and HBM count per package will rise from 8 to 12. HBM4 12hi will be adopted, with no hybrid bonding in the initial phase, pushing capacity to 576GB. Based on current validation, Samsung’s participation is unlikely. With these wafer volumes, only 1.2 million Rubin chips may be available in 2026 — equivalent to around 16–17K VR144 cabinets. However, ODMs likely won’t receive the chips until late 2026, and CSP customers may not receive VR144 servers until 2H 2027 or even year-end. Rubin chips require second-gen CoWoS-L packaging and 12 HBM4 stacks. WT expects foundries to delay output, pushing VR144 availability to late 2027.\n\nCloud service providers are also beginning volume production of in-house ASICs. Google’s TPU V6p is shipping this year, though some industry insiders believe it’s been rebranded as V7. Volume production will begin in 2H 2025 with about 2–3 million units expected for the year. For 2026, TPU V7 or V8 may be co-designed with 2454 MTK. However, due to unresolved issues with Google, no tape-out has occurred yet. If resolved, 2026 TPU shipments could exceed 3 million units, driven by Gemini-related applications. AWS's Trainium2 is expected to ship around 1.7 million units in 2025, with a small contribution from the minor Trainium2.5 update by year-end. Mass production of Trainium3 is expected by Q2 2026; it has recently taped out and will be produced by Annapurna, with backend handled by 3661 Alchip and Marvell. Total 2026 shipments are expected to reach 2.3–2.5 million units. Meta's MTIA 1 will begin shipping in Q4 2025, with 1.5 million units projected for 2026, and MTIA 1.5 launching mid-2026 using Andes RISC-V. Microsoft’s Maia200 recently taped out, with mass production expected in early 2026 — as this is Microsoft's first ASIC chip, market expectations remain limited.\n\nIn conclusion, GPGPU growth in 2026 — primarily driven by NVIDIA’s Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.\n\nIn NVIDIA’s substrate supply chain, Ibiden holds 65–75% market share, 3037 Unimicron holds 15–25%, and 3189 Kinsus has 5–10%. Japanese leader Ibiden has nearly all its high-end capacity allocated to NVIDIA. 3037 Unimicron has a strong track record with Blackwell substrates and high-end OAM HDI server boards. It has also demonstrated flexibility in substrate line adjustments and continuous upgrades to its HDI and PCB lines, significantly improving its operational foundation and optimizing its product mix. This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance. As we enter the ASIC era, Unimicron's competitive advantage will increasingly stand out.\n\n---\n\nhttps://t.co/CMrR2CZzJv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0507336908902325, "ret_signed_m1d": 0.0507336908902325, "alpha_spy_m1d": 0.04322572431173133, "alpha_c_m1d": 0.03693616519194154, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.013797525698290958, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005590341494530637, "ret_signed_p1d": 0.005590341494530637, "alpha_spy_p1d": 0.006138057712275935, "alpha_c_p1d": -0.007706786993641801, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.013297128488172438, "ret_p1w": 0.014246571417238174, "ret_signed_p1w": 0.014246571417238174, "alpha_spy_p1w": 0.007592657358603372, "alpha_c_p1w": -0.005627612551230232, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.019874183968468406, "ret_p1m": 0.07187586503761345, "ret_signed_p1m": 0.07187586503761345, "alpha_spy_p1m": 0.06013075865839057, "alpha_c_p1m": 0.04563909716808223, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.02623676786953122, "ret_p3m": 0.2057379531968475, "ret_signed_p3m": 0.2057379531968475, "alpha_spy_p3m": 0.12454425513654432, "alpha_c_p3m": -0.003298303442470285, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.2090362566393178, "ret_p6m": 0.2143747239497582, "ret_signed_p6m": 0.2143747239497582, "alpha_spy_p6m": 0.10115951074752627, "alpha_c_p6m": -0.088314924732664, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.3026896486824222, "price_path": [-0.149, -0.2619, -0.2538, -0.2696, -0.2548, -0.2742, -0.2774, -0.2774, -0.3099, -0.2925, -0.2485, -0.1987, -0.1765, -0.1796, -0.1799, -0.1842, -0.1481, -0.1289, -0.1338, -0.1444, -0.213, -0.1945, -0.1663, -0.0985, -0.0848, -0.079, -0.0887, -0.1089, -0.1256, -0.1416, -0.1599, -0.1356, -0.1518, -0.1518, -0.108, -0.0973, -0.1093, -0.1588, -0.1409, -0.1284, -0.0734, -0.0893, -0.0447, -0.0337, -0.0379, -0.0463, -0.0267, -0.0609, -0.0158, -0.0218, 0.0475, 0.0475, 0.0274, -0.0108, 0.0512, 0.0612, 0.1177, 0.0784, 0.0818, 0.0655, 0.0377, 0.0507, 0.0507, 0.0, 0.0056, 0.0099, 0.0253, 0.017, 0.0142, 0.0128, -0.009, 0.0073, 0.0442, 0.0219, 0.007, 0.0249, 0.0356, 0.038, 0.0618, 0.0678, 0.1434, 0.1242, 0.0414, 0.0705, 0.0719, 0.0536, 0.061, 0.0818, 0.081, 0.0884, 0.1095, 0.1056, 0.0657, 0.0734, 0.0081, -0.0038, -0.0039, 0.0211, 0.0204, 0.0387, 0.0461, 0.0803, -0.1206, -0.1206, -0.0964, -0.1284, -0.1034, -0.1142, -0.0768, -0.0651, -0.0614, -0.0686, -0.0579, -0.0568, -0.0368, -0.0072, 0.0383, 0.0387, 0.0565, 0.0438, 0.1203, 0.1723, 0.1634, 0.1524, 0.1759, 0.1734, 0.2057, 0.206, 0.2438, 0.2165, 0.2939, 0.2684, 0.1983, 0.2512, 0.2068, 0.2442, 0.235, 0.231, 0.2015, 0.1794, 0.1345, 0.1585, 0.1776, 0.2417, 0.2382, 0.2618, 0.2397, 0.3121, 0.2649, 0.226, 0.3003, 0.3063, 0.2726, 0.3049, 0.2503, 0.2503, 0.225, 0.21, 0.168, 0.1013, 0.1382, 0.0733, 0.0841, 0.1728, 0.1678, 0.2278, 0.2278, 0.2513, 0.2751, 0.3002, 0.4025, 0.3744, 0.3844, 0.2877, 0.2443, 0.2943, 0.2517, 0.1818, 0.1794, 0.1767, 0.1435, 0.1823, 0.177, 0.1869, 0.2273, 0.2106, 0.2106, 0.2085, 0.2004, 0.2144]}
{"unit_id": "thread:1942028012075246065", "ticker_idx": 5, "ts": "2025-07-07T01:08:53+00:00", "symbol": "3037.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron's competitive advantage will increasingly stand out... This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin delays weigh on NVDA 2026 GPGPU growth vs ASICs; bullish Wistron assembly quality, MediaTek TPU co-design, Alchip/Marvell Trainium3 backend, and Unimicron as rising high-end substrate winner.", "resolved_tickers": ["3037.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Recent Trends and Future Outlook in the AI Server Sector By WT\n\nNVIDIA’s GB200 NVL72 shipments are ramping steadily. Shipments reached around 0.5K cabinets in Q1 2025. WT forecasts Q2–Q4 shipments of 4.0K, 8.1K, and 11.3K cabinets respectively — totaling about 24K units for 2025, aligning with market expectations of 21–26K. WT's research indicates ODM shipments are stabilizing, but CSP design revisions have hindered rapid volume increases. Currently, only Corewave, OpenAI, and Oracle are receiving shipments smoothly. Thanks to accumulated experience, 3231 Wistron has significantly better quality in sub-L10 assembly than peers. Also, some CSPs are still struggling with minor design revisions.\n\nRubin hasn’t taped out yet, but foundries expect mass production to begin in Q2 2026. Estimated wafer starts for 2026 are 340–350K. The CoWoS-L reticle area is increasing from 3.5X to 5.5X, and HBM count per package will rise from 8 to 12. HBM4 12hi will be adopted, with no hybrid bonding in the initial phase, pushing capacity to 576GB. Based on current validation, Samsung’s participation is unlikely. With these wafer volumes, only 1.2 million Rubin chips may be available in 2026 — equivalent to around 16–17K VR144 cabinets. However, ODMs likely won’t receive the chips until late 2026, and CSP customers may not receive VR144 servers until 2H 2027 or even year-end. Rubin chips require second-gen CoWoS-L packaging and 12 HBM4 stacks. WT expects foundries to delay output, pushing VR144 availability to late 2027.\n\nCloud service providers are also beginning volume production of in-house ASICs. Google’s TPU V6p is shipping this year, though some industry insiders believe it’s been rebranded as V7. Volume production will begin in 2H 2025 with about 2–3 million units expected for the year. For 2026, TPU V7 or V8 may be co-designed with 2454 MTK. However, due to unresolved issues with Google, no tape-out has occurred yet. If resolved, 2026 TPU shipments could exceed 3 million units, driven by Gemini-related applications. AWS's Trainium2 is expected to ship around 1.7 million units in 2025, with a small contribution from the minor Trainium2.5 update by year-end. Mass production of Trainium3 is expected by Q2 2026; it has recently taped out and will be produced by Annapurna, with backend handled by 3661 Alchip and Marvell. Total 2026 shipments are expected to reach 2.3–2.5 million units. Meta's MTIA 1 will begin shipping in Q4 2025, with 1.5 million units projected for 2026, and MTIA 1.5 launching mid-2026 using Andes RISC-V. Microsoft’s Maia200 recently taped out, with mass production expected in early 2026 — as this is Microsoft's first ASIC chip, market expectations remain limited.\n\nIn conclusion, GPGPU growth in 2026 — primarily driven by NVIDIA’s Rubin — will face delays and a high base effect. In contrast, ASICs will show much stronger growth.\n\nIn NVIDIA’s substrate supply chain, Ibiden holds 65–75% market share, 3037 Unimicron holds 15–25%, and 3189 Kinsus has 5–10%. Japanese leader Ibiden has nearly all its high-end capacity allocated to NVIDIA. 3037 Unimicron has a strong track record with Blackwell substrates and high-end OAM HDI server boards. It has also demonstrated flexibility in substrate line adjustments and continuous upgrades to its HDI and PCB lines, significantly improving its operational foundation and optimizing its product mix. This positions Unimicron as a true high-end substrate provider, which will be reflected in its future gross margin performance. As we enter the ASIC era, Unimicron's competitive advantage will increasingly stand out.\n\n---\n\nhttps://t.co/CMrR2CZzJv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026086946593948168, "ret_signed_m1d": 0.026086946593948168, "alpha_spy_m1d": 0.018578980015447, "alpha_c_m1d": 0.018084440198896656, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.008002506395051512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008695626478454344, "ret_signed_p1d": 0.008695626478454344, "alpha_spy_p1d": 0.009243342696199641, "alpha_c_p1d": 0.0033213971886223437, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.005374229289832, "ret_p1w": 0.05631321259876265, "ret_signed_p1w": 0.05631321259876265, "alpha_spy_p1w": 0.04965929854012785, "alpha_c_p1w": 0.05348883039443186, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.0028243822043307887, "ret_p1m": 0.1927536692261027, "ret_signed_p1m": 0.1927536692261027, "alpha_spy_p1m": 0.18100856284687983, "alpha_c_p1m": 0.17235519304025138, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.020398476185851333, "ret_p3m": 0.3467993460634222, "ret_signed_p3m": 0.3467993460634222, "alpha_spy_p3m": 0.26560564800311903, "alpha_c_p3m": 0.222808456446433, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.12399088961698923, "ret_p6m": 0.9321729180452365, "ret_signed_p6m": 0.9321729180452365, "alpha_spy_p6m": 0.8189577048430046, "alpha_c_p6m": 0.7881456525250023, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.14402726552023415, "price_path": [-0.3878, -0.327, -0.3191, -0.2635, -0.2278, -0.2591, -0.2635, -0.2591, -0.2939, -0.2957, -0.2278, -0.2478, -0.1948, -0.1713, -0.1652, -0.1835, -0.1835, -0.1774, -0.1957, -0.1974, -0.2096, -0.1948, -0.1835, -0.1217, -0.1322, -0.1087, -0.1217, -0.113, -0.1391, -0.1261, -0.1087, -0.113, -0.0783, -0.0957, -0.0391, -0.0478, -0.0783, -0.0783, -0.1217, -0.113, -0.1043, -0.1043, -0.1043, -0.1304, -0.0696, -0.0522, -0.0696, -0.1087, -0.1043, -0.1174, -0.1, -0.1261, -0.1304, -0.0435, -0.0304, -0.0087, -0.0087, -0.0174, -0.0087, -0.0261, -0.0261, 0.0174, 0.0261, 0.0, 0.0087, 0.0652, 0.1047, 0.0915, 0.0563, 0.1619, 0.1928, 0.1751, 0.1707, 0.1751, 0.1311, 0.1575, 0.1663, 0.1531, 0.1707, 0.2324, 0.25, 0.2104, 0.2104, 0.1884, 0.1928, 0.1795, 0.1972, 0.206, 0.2544, 0.2192, 0.2148, 0.2148, 0.1928, 0.25, 0.2236, 0.1751, 0.1751, 0.1884, 0.2324, 0.25, 0.338, 0.3204, 0.272, 0.184, 0.1531, 0.1663, 0.1751, 0.1795, 0.1751, 0.1884, 0.2456, 0.2544, 0.272, 0.25, 0.338, 0.2984, 0.3072, 0.3864, 0.3952, 0.3996, 0.382, 0.382, 0.3116, 0.3116, 0.3336, 0.3644, 0.3468, 0.3732, 0.3732, 0.3908, 0.4084, 0.4172, 0.4172, 0.3512, 0.2852, 0.3952, 0.4568, 0.4128, 0.4084, 0.4084, 0.3996, 0.3732, 0.3732, 0.4392, 0.4304, 0.4612, 0.4084, 0.4392, 0.5713, 0.4832, 0.4325, 0.4456, 0.4106, 0.4368, 0.4543, 0.463, 0.5591, 0.5096, 0.4744, 0.4964, 0.4128, 0.5096, 0.4084, 0.4436, 0.5493, 0.5713, 0.6461, 0.6417, 0.7341, 0.7253, 0.7297, 0.7473, 0.8926, 0.941, 0.9894, 1.0026, 0.9806, 0.9806, 0.9498, 0.8838, 0.8662, 0.8221, 0.8926, 0.941, 0.897, 0.9102, 0.9102, 0.941, 0.941, 0.9322]}
{"unit_id": "orig:1942727532094316856", "ticker_idx": 0, "ts": "2025-07-08T23:28:32+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price remains ₩74,000, with 'Buy' investment opinion maintained", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Sachs Samsung Electronics summary: Buy maintained, ₩74,000 PT despite 14% EPS cut; Q3 recovery expected on HBM3E ramp and memory price recovery.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electronics Q2 2025 Earnings Summary and Key Interpretations (Goldman Sachs)\n\n- Q2 2025 operating profit was ₩4.6 trillion, falling below GS's estimate of ₩5.5 trillion and the market consensus of ₩6.1 trillion.\n\n- Reasons for Underperformance: Weak profitability in memory, including HBM, and larger-than-expected losses in Foundry/LSI.\n\n- Memory Business: Incurred one-off costs (inventory valuation losses). HBM3E is currently undergoing customer qualification, with shipments in progress.\n\n- Foundry Business: Operating losses widened due to low utilization rates, primarily in mature nodes, and the impact of inventory losses.\n\n- Display Business: Met expectations.\n\n- Smartphone Business: Exceeded expectations due to product mix and foreign exchange effects.\n\n- CE Division: Underperformed expectations due to weak TV demand.\n\n1. Factors Expected to Drive Earnings Improvement from Q3 2025\n\n- Full-scale shipments of HBM3E to multiple customers, and expected progress in HBM4 qualification.\n\n- Increased memory (especially DRAM) shipments and price recovery.\n\n- Gradual recovery of Foundry utilization rates → Expected reduction in losses.\n\n- Quarterly earnings improvement across OLED, Smartphone, and Home Appliance divisions.\n\n- Potential for QoQ increase in profits across all divisions.\n\n2. Key Focus Points for Q2 2025 Earnings Conference Call\n\n- HBM3E 12Hi and HBM4 customer qualification and timeline.\n\n- 2025 HBM revenue target and initial 2026 outlook.\n\n- Rising memory spot prices and whether they are driving general memory demand.\n\n- Potential changes in Foundry and Memory Capital Expenditure (Capex) outlook.\n\n- Scale of inventory losses in Memory/Foundry.\n\n- H2 smartphone demand and OLED competitive environment.\n\n- Capital allocation strategy (M&A, shareholder returns, etc.).\n\n3. Profit Estimates and Target Price\n\n- 14% downward adjustment of 2025E~2027E EPS.\n\n- Slight downward revision of DRAM/NAND ASP, reflecting reduced HBM shipments.\n\n- Partially offset by improved smartphone profitability.\n\n- Target price remains ₩74,000, with 'Buy' investment opinion maintained.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004885940247018272, "ret_signed_m1d": 0.004885940247018272, "alpha_spy_m1d": 0.00433792387181664, "alpha_c_m1d": 0.010231441586980305, "bench_spy_m1d": 0.000548016375201632, "bench_c_m1d": -0.005345501339962033, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016286703429456373, "ret_signed_p1d": -0.016286703429456373, "alpha_spy_p1d": -0.022283330196667728, "alpha_c_p1d": -0.022295565333592227, "bench_spy_p1d": 0.005996626767211355, "bench_c_p1d": 0.006008861904135854, "ret_p1w": 0.037459282758823154, "ret_signed_p1w": 0.037459282758823154, "alpha_spy_p1w": 0.03455766355548784, "alpha_c_p1w": 0.031138258210178238, "bench_spy_p1w": 0.002901619203335315, "bench_c_p1w": 0.006321024548644916, "ret_p1m": 0.12052110347053557, "ret_signed_p1m": 0.12052110347053557, "alpha_spy_p1m": 0.10046765326937801, "alpha_c_p1m": 0.09414479469039039, "bench_spy_p1m": 0.020053450201157563, "bench_c_p1m": 0.026376308780145186, "ret_p3m": 0.4682479839729372, "ret_signed_p3m": 0.4682479839729372, "alpha_spy_p3m": 0.38647774919014033, "alpha_c_p3m": 0.3558912510545016, "bench_spy_p3m": 0.08177023478279688, "bench_c_p3m": 0.11235673291843562, "ret_p6m": 0.9710162609002393, "ret_signed_p6m": 0.9710162609002393, "alpha_spy_p6m": 0.8654432826420919, "alpha_c_p6m": 0.8443732150584198, "bench_spy_p6m": 0.10557297825814738, "bench_c_p6m": 0.1266430458418195, "price_path": [-0.087, -0.1065, -0.0903, -0.0838, -0.1146, -0.1081, -0.1048, -0.1032, -0.1097, -0.0984, -0.0984, -0.0984, -0.0967, -0.0967, -0.1016, -0.1016, -0.121, -0.121, -0.121, -0.1162, -0.1162, -0.1129, -0.0676, -0.0789, -0.0708, -0.0725, -0.0806, -0.0967, -0.0951, -0.0984, -0.1146, -0.1226, -0.1146, -0.1275, -0.0951, -0.0919, -0.0903, -0.0806, -0.0806, -0.0644, -0.0433, -0.0433, -0.032, -0.0417, -0.0304, -0.0369, -0.0563, -0.0741, -0.0595, -0.032, -0.0417, -0.0369, -0.0611, -0.0207, -0.0077, -0.0255, -0.0098, -0.0261, -0.0195, -0.0098, 0.0391, 0.0309, 0.0049, 0.0, -0.0163, -0.0065, 0.0195, 0.0179, 0.0375, 0.0537, 0.0863, 0.0928, 0.1042, 0.0749, 0.0814, 0.0749, 0.0733, 0.1466, 0.1498, 0.1824, 0.1629, 0.1221, 0.1352, 0.1384, 0.1205, 0.1482, 0.1694, 0.1564, 0.158, 0.171, 0.1661, 0.1661, 0.1401, 0.1401, 0.1482, 0.1498, 0.1629, 0.1645, 0.145, 0.1498, 0.1336, 0.1352, 0.101, 0.1254, 0.1368, 0.1417, 0.1319, 0.1417, 0.1645, 0.1824, 0.1954, 0.228, 0.2459, 0.2932, 0.2736, 0.3078, 0.3078, 0.3599, 0.3795, 0.3909, 0.4023, 0.3567, 0.3775, 0.3725, 0.4069, 0.4682, 0.4682, 0.4682, 0.4682, 0.4682, 0.4682, 0.5443, 0.5263, 0.4985, 0.5541, 0.5983, 0.6016, 0.6048, 0.595, 0.613, 0.5787, 0.6163, 0.6686, 0.6278, 0.6441, 0.703, 0.7586, 0.8175, 0.7161, 0.6457, 0.6228, 0.6016, 0.6457, 0.6932, 0.6866, 0.6817, 0.5901, 0.6457, 0.5999, 0.5787, 0.6457, 0.5509, 0.5819, 0.6245, 0.6817, 0.6932, 0.6441, 0.649, 0.6916, 0.7095, 0.7194, 0.7733, 0.7913, 0.7733, 0.7668, 0.7554, 0.7815, 0.7145, 0.6817, 0.7652, 0.7603, 0.739, 0.8077, 0.8241, 0.8175, 0.8175, 0.914, 0.9644, 0.971, 0.971]}
{"unit_id": "orig:1942351853926375502", "ticker_idx": 0, "ts": "2025-07-07T22:35:43+00:00", "symbol": "GDS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "anticipating that with deleveraging progress and improved chip availability, the performance and valuations of GDS and VNET are likely to recover in the second half of the year", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman Sachs bullish on GDS and VNET for H2 recovery driven by REITs-led deleveraging and improved AI chip supply in China.", "resolved_tickers": ["GDS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "original", "lang": "en", "tweet_text": "China Data Centers: Deleveraging and Improved Chip Supply Expected to Drive Stock Price Rebound\nGoldman Sachs (T. Zhao, 25/07/04)\n\nGoldman Sachs maintains a positive outlook on the Chinese data center industry, anticipating that with deleveraging progress and improved chip availability, the performance and valuations of GDS and VNET are likely to recover in the second half of the year.\n\nGDS and VNET Stock Prices Suppressed by Financing and Macro Disruptions, But With Significant Valuation Upside\nGDS and VNET's stock prices have consistently underperformed the market over the past three months. This is primarily due to dilution from convertible bonds and equity financing (GDS issued $550 million in bonds and an additional $142 million in equity; VNET issued $430 million in bonds), geopolitical tensions, and weaker capital expenditure from Chinese cloud vendors.\n\nREITs Issuance to Drive GDS/VNET Deleveraging, Expected to Improve Valuation and Cash Flow Structure\nGDS plans to issue C-REITs on the Shanghai Stock Exchange on July 14th. This issuance implies a 2026 EV/EBITDA multiple of 16.9x, which is higher than the current 13.5x valuation for GDS's China segment (excluding DayOne). This move is expected to generate approximately RMB 1.6 billion in net proceeds, aiding in deleveraging. VNET has also received approval to issue RMB 860 million in private REITs/ABS, and its future progress is worth monitoring.\n\nImproved Chip Supply Expected to Drive Order and Revenue Rebound in H2\nAlthough maintaining a cautious stance on cloud vendor capital expenditures for Q2 2025, Goldman Sachs believes that with improved supply of AI chips in China, spending is expected to rebound in the second half of 2025. Despite being affected by the H20 ban, NVIDIA is developing B30 chips for the Chinese market, with annual production potentially exceeding 1 million units. GDS and VNET have high requirements for customer chip allocation and deployment pace, so improved chip availability will drive an increase in order volumes and accelerate revenue and EBITDA growth in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05136138190242456, "ret_signed_m1d": 0.05136138190242456, "alpha_spy_m1d": 0.043853415323923395, "alpha_c_m1d": 0.04335887550737305, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.008002506395051512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0015470061117777156, "ret_signed_p1d": 0.0015470061117777156, "alpha_spy_p1d": 0.0020947223295230133, "alpha_c_p1d": -0.0038272231780542842, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.005374229289832, "ret_p1w": 0.08292080230449272, "ret_signed_p1w": 0.08292080230449272, "alpha_spy_p1w": 0.07626688824585792, "alpha_c_p1w": 0.08009642010016194, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.0028243822043307887, "ret_p1m": 0.09993810559204874, "ret_signed_p1m": 0.09993810559204874, "alpha_spy_p1m": 0.08819299921282586, "alpha_c_p1m": 0.07953962940619741, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.020398476185851333, "ret_p3m": 0.27939352922527894, "ret_signed_p3m": 0.27939352922527894, "alpha_spy_p3m": 0.19819983116497575, "alpha_c_p3m": 0.1554026396082897, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.12399088961698923, "ret_p6m": 0.07054451735226963, "ret_signed_p6m": 0.07054451735226963, "alpha_spy_p6m": -0.042670695849962303, "alpha_c_p6m": -0.07348274816796452, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.14402726552023415, "price_path": [-0.3988, -0.3691, -0.3608, -0.3688, -0.3617, -0.3868, -0.3357, -0.3357, -0.3438, -0.3069, -0.2757, -0.2862, -0.2837, -0.2766, -0.2624, -0.22, -0.203, -0.1306, -0.1402, -0.1318, -0.1897, -0.1816, -0.2126, -0.0987, -0.1541, -0.1556, -0.1819, -0.1575, -0.1621, -0.1467, -0.0826, -0.1287, -0.1182, -0.1182, -0.1603, -0.2336, -0.2237, -0.267, -0.2457, -0.2574, -0.2426, -0.2061, -0.2178, -0.1934, -0.19, -0.168, -0.1733, -0.2054, -0.13, -0.1578, -0.1569, -0.1569, -0.1631, -0.1479, -0.104, -0.1111, -0.0814, -0.0665, -0.0541, -0.0882, -0.0459, 0.0514, 0.0514, 0.0, 0.0015, -0.0028, 0.0059, 0.0492, 0.0829, 0.1671, 0.1795, 0.1742, 0.1627, 0.1058, 0.0845, 0.1361, 0.1145, 0.138, 0.13, 0.1247, 0.0913, 0.116, 0.0557, 0.0808, 0.0999, 0.1037, 0.1532, 0.1253, 0.1408, 0.1235, 0.0495, -0.0149, -0.0282, -0.0028, -0.0226, 0.0489, -0.0099, 0.0368, 0.0275, 0.0458, 0.0418, 0.0854, 0.0693, 0.0693, 0.0204, 0.0569, 0.0136, 0.0263, 0.0368, 0.06, 0.0365, 0.1912, 0.1822, 0.151, 0.1547, 0.242, 0.2067, 0.2054, 0.2302, 0.1643, 0.2584, 0.3017, 0.2401, 0.2367, 0.1974, 0.2447, 0.2794, 0.2262, 0.2271, 0.1603, 0.1708, 0.1887, 0.0306, 0.1058, 0.0424, 0.0476, 0.0297, 0.0127, 0.064, 0.056, 0.0316, 0.06, 0.0804, 0.1108, 0.1098, 0.1658, 0.0993, 0.1046, 0.0746, 0.0217, 0.0572, 0.0492, 0.0161, 0.034, 0.0313, -0.0084, -0.0532, -0.0823, -0.0919, -0.1021, -0.0795, -0.0628, -0.0433, 0.0368, 0.0541, 0.0551, 0.0551, 0.0511, 0.0712, 0.0458, 0.0439, 0.0452, 0.0894, 0.1027, 0.1145, 0.121, 0.1185, 0.1219, 0.1009, 0.0749, 0.0424, 0.0662, 0.1117, 0.1154, 0.0808, 0.0767, 0.0767, 0.0953, 0.0721, 0.0705]}
{"unit_id": "orig:1942351853926375502", "ticker_idx": 1, "ts": "2025-07-07T22:35:43+00:00", "symbol": "VNET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "anticipating that with deleveraging progress and improved chip availability, the performance and valuations of GDS and VNET are likely to recover in the second half of the year", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman Sachs bullish on GDS and VNET for H2 recovery driven by REITs-led deleveraging and improved AI chip supply in China.", "resolved_tickers": ["VNET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "original", "lang": "en", "tweet_text": "China Data Centers: Deleveraging and Improved Chip Supply Expected to Drive Stock Price Rebound\nGoldman Sachs (T. Zhao, 25/07/04)\n\nGoldman Sachs maintains a positive outlook on the Chinese data center industry, anticipating that with deleveraging progress and improved chip availability, the performance and valuations of GDS and VNET are likely to recover in the second half of the year.\n\nGDS and VNET Stock Prices Suppressed by Financing and Macro Disruptions, But With Significant Valuation Upside\nGDS and VNET's stock prices have consistently underperformed the market over the past three months. This is primarily due to dilution from convertible bonds and equity financing (GDS issued $550 million in bonds and an additional $142 million in equity; VNET issued $430 million in bonds), geopolitical tensions, and weaker capital expenditure from Chinese cloud vendors.\n\nREITs Issuance to Drive GDS/VNET Deleveraging, Expected to Improve Valuation and Cash Flow Structure\nGDS plans to issue C-REITs on the Shanghai Stock Exchange on July 14th. This issuance implies a 2026 EV/EBITDA multiple of 16.9x, which is higher than the current 13.5x valuation for GDS's China segment (excluding DayOne). This move is expected to generate approximately RMB 1.6 billion in net proceeds, aiding in deleveraging. VNET has also received approval to issue RMB 860 million in private REITs/ABS, and its future progress is worth monitoring.\n\nImproved Chip Supply Expected to Drive Order and Revenue Rebound in H2\nAlthough maintaining a cautious stance on cloud vendor capital expenditures for Q2 2025, Goldman Sachs believes that with improved supply of AI chips in China, spending is expected to rebound in the second half of 2025. Despite being affected by the H20 ban, NVIDIA is developing B30 chips for the Chinese market, with annual production potentially exceeding 1 million units. GDS and VNET have high requirements for customer chip allocation and deployment pace, so improved chip availability will drive an increase in order volumes and accelerate revenue and EBITDA growth in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.027851464086130262, "ret_signed_m1d": 0.027851464086130262, "alpha_spy_m1d": 0.020343497507629094, "alpha_c_m1d": 0.01984895769107875, "bench_spy_m1d": 0.0075079665785011684, "bench_c_m1d": 0.008002506395051512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00265251737767902, "ret_signed_p1d": 0.00265251737767902, "alpha_spy_p1d": 0.0032002335954243177, "alpha_c_p1d": -0.00272171191215298, "bench_spy_p1d": -0.0005477162177452977, "bench_c_p1d": 0.005374229289832, "ret_p1w": 0.050397893416902484, "ret_signed_p1w": 0.050397893416902484, "alpha_spy_p1w": 0.04374397935826768, "alpha_c_p1w": 0.047573511212571695, "bench_spy_p1w": 0.006653914058634802, "bench_c_p1w": 0.0028243822043307887, "ret_p1m": 0.0384615335968459, "ret_signed_p1m": 0.0384615335968459, "alpha_spy_p1m": 0.026716427217623018, "alpha_c_p1m": 0.018063057410994565, "bench_spy_p1m": 0.01174510637922288, "bench_c_p1m": 0.020398476185851333, "ret_p3m": 0.49071615755762243, "ret_signed_p3m": 0.49071615755762243, "alpha_spy_p3m": 0.40952245949731925, "alpha_c_p3m": 0.3667252679406332, "bench_spy_p3m": 0.08119369806030319, "bench_c_p3m": 0.12399088961698923, "ret_p6m": 0.12466837999191283, "ret_signed_p6m": 0.12466837999191283, "alpha_spy_p6m": 0.01145316678968089, "alpha_c_p6m": -0.019358885528321323, "bench_spy_p6m": 0.11321521320223193, "bench_c_p6m": 0.14402726552023415, "price_path": [-0.2175, -0.248, -0.2653, -0.2706, -0.2851, -0.3714, -0.3342, -0.3342, -0.3634, -0.3541, -0.2984, -0.2812, -0.2865, -0.2785, -0.2493, -0.1658, -0.1684, -0.061, -0.0769, -0.0942, -0.1645, -0.2042, -0.2095, -0.0411, -0.1525, -0.1737, -0.2056, -0.1472, -0.1645, -0.1751, -0.1406, -0.1976, -0.2056, -0.2056, -0.1989, -0.2851, -0.2759, -0.2918, -0.2679, -0.2653, -0.2281, -0.2003, -0.2162, -0.1844, -0.1963, -0.2016, -0.2334, -0.2785, -0.2467, -0.2639, -0.2785, -0.2785, -0.2732, -0.2772, -0.2255, -0.2042, -0.2162, -0.0743, -0.0849, -0.1233, -0.061, 0.0279, 0.0279, 0.0, 0.0027, -0.0477, -0.0756, 0.0544, 0.0504, 0.1989, 0.1658, 0.1618, 0.195, 0.118, 0.0729, 0.1592, 0.1048, 0.0955, 0.1061, 0.0981, 0.0637, 0.1021, -0.0066, 0.0371, 0.0385, 0.0623, 0.0809, 0.0584, 0.0637, 0.0345, 0.0623, 0.0544, 0.0305, 0.0279, 0.0398, 0.0584, -0.0265, 0.0995, 0.0398, 0.0358, 0.0345, 0.1313, 0.1485, 0.1485, 0.069, 0.061, 0.0133, 0.0477, 0.0186, 0.1048, 0.0517, 0.2082, 0.2692, 0.2719, 0.2719, 0.3064, 0.3196, 0.2772, 0.3329, 0.2401, 0.3753, 0.4761, 0.4324, 0.3806, 0.37, 0.4271, 0.4907, 0.4032, 0.4072, 0.317, 0.3117, 0.2706, 0.1286, 0.248, 0.2175, 0.2719, 0.1976, 0.1618, 0.2427, 0.2414, 0.2029, 0.2454, 0.2759, 0.3289, 0.3581, 0.4642, 0.3806, 0.3793, 0.3448, 0.2308, 0.3037, 0.3223, 0.2798, 0.2865, 0.2984, 0.1897, 0.1631, 0.1671, 0.1379, 0.1499, 0.1034, 0.0902, 0.1127, 0.179, 0.1857, 0.1817, 0.1817, 0.1857, 0.1897, 0.1631, 0.179, 0.1737, 0.1857, 0.2374, 0.2626, 0.2056, 0.2255, 0.2175, 0.1817, 0.1605, 0.1207, 0.1273, 0.1605, 0.1538, 0.1419, 0.1379, 0.1379, 0.1525, 0.1459, 0.1247]}
{"unit_id": "orig:1920979784760471994", "ticker_idx": 0, "ts": "2025-05-09T23:10:45+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%)... hardware ODMs are expected to maintain momentum in Q2", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS bullish on Taiwanese ODMs (Quanta, Wistron, Wiwynn, Inventec, Hon Hai) and Supermicro into Q2/2H25 on Blackwell rack ramp and HGX demand expansion.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "UBS APAC Technology (May 7)\n\n1. UBS AI Market Outlook and Implications for the Asian Supply Chain\n\nUnlike Taiwanese ODMs, Supermicro has shown mixed quarterly results in recent quarters. While Supermicro and HPE were hit earlier this year by a reduction in Hopper product volumes, Taiwanese ODMs continued to deliver solid growth.\n\nIn Q1, Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%). Hon Hai met expectations.\n\nQ2 also started strong, with positive outlooks from companies such as Wistron and Delta. As Blackwell rack deployment gradually improves and demand increases for HGX for B200, hardware ODMs are expected to maintain momentum in Q2.\n\nFactors contributing to this growth include Taiwan's limited exposure to Hopper downsizing, the ability to supply H20 products before Chinese export restrictions, Wiwynn’s expansion in ASIC servers, and tariff cuts on iPhone and notebook assembly.\n\n2. Supermicro’s Position and Relationship with Taiwanese ODMs\n\nSupermicro supports clients like xAI and Coreweave and focuses on second- and third-tier supply chains that serve mid-sized data centers and channel customers. This structure is shared with OEM partners such as Wistron and, to some extent, Gigabyte.\n\nUBS believes Wistron is producing AI server motherboards on an outsourced basis, while final assembly is handled by Hon Hai. Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025.\n\n3. Rack Deployment and Growth Outlook for 2H25\n\nGB200 rack shipments were around 1,000 units in Q1 2025 but are expected to rise to over 5,000 in Q2, with further expansion projected in the second half. GB300 is expected to enter pilot production in Q3 and move into full production in Q4.\n\nDemand for HGX systems is also increasing. Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected for Inventec and Wistron.\n\nThe next momentum driver is Computex in early June, followed by increased capital expenditure from hyperscalers in Q2, which will drive further supply expansion to meet demand.\n\n4. Tariff Impact and Supply Chain Volatility\n\nKey variables include the timing of the reinstatement of Section 232 semiconductor tariffs and resolution of the delayed mutual tariff adjustment (90-day delay).\n\nIn the AI supply chain, the USMCA exemption for Mexico remains a major protective factor. Although volumes are relatively small, the high-value nature of these products may lead to accelerated relocation of assembly to the U.S. if needed (comment by Randy Abrams).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019493153939566743, "ret_signed_m1d": -0.019493153939566743, "alpha_spy_m1d": -0.020769004322343743, "alpha_c_m1d": -0.020733880053138365, "bench_spy_m1d": 0.0012758503827769996, "bench_c_m1d": 0.0012407261135716219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007797286506401324, "ret_signed_p1d": 0.007797286506401324, "alpha_spy_p1d": -0.02525022483601025, "alpha_c_p1d": -0.038618056662658384, "bench_spy_p1d": 0.03304751134241157, "bench_c_p1d": 0.04641534316905971, "ret_p1w": 0.072124775531339, "ret_signed_p1w": 0.072124775531339, "alpha_spy_p1w": 0.019213467206554702, "alpha_c_p1w": -0.007332896053190874, "bench_spy_p1w": 0.0529113083247843, "bench_c_p1w": 0.07945767158452988, "ret_p1m": 0.07602341878453966, "ret_signed_p1m": 0.07602341878453966, "alpha_spy_p1m": 0.013401623369987181, "alpha_c_p1m": -0.02475777391483125, "bench_spy_p1m": 0.06262179541455248, "bench_c_p1m": 0.10078119269937091, "ret_p3m": 0.17841441047044238, "ret_signed_p3m": 0.17841441047044238, "alpha_spy_p3m": 0.05382520847695216, "alpha_c_p3m": -0.03245853375929619, "bench_spy_p3m": 0.12458920199349022, "bench_c_p3m": 0.21087294422973857, "ret_p6m": 0.21517603852671274, "ret_signed_p6m": 0.21517603852671274, "alpha_spy_p6m": -0.0026364430815917217, "alpha_c_p6m": -0.17636363819195355, "bench_spy_p6m": 0.21781248160830446, "bench_c_p6m": 0.3915396767186663, "price_path": [0.0058, 0.0058, 0.0136, 0.0136, 0.0156, 0.0312, 0.0507, 0.039, 0.0526, 0.0331, -0.0078, 0.0136, -0.0234, -0.0234, -0.0234, 0.0078, 0.0214, 0.0156, 0.0058, 0.0175, -0.0078, 0.0117, 0.0039, 0.0, -0.0058, 0.0117, -0.0117, 0.0019, -0.0019, -0.0273, -0.0273, -0.0175, -0.0409, -0.0604, -0.1248, -0.0877, -0.0916, -0.0916, -0.0916, -0.1813, -0.2534, -0.3041, -0.2359, -0.1598, -0.1111, -0.1111, -0.115, -0.1033, -0.1014, -0.0955, -0.1287, -0.0838, -0.1033, -0.0838, -0.0448, -0.0546, -0.0702, -0.0702, -0.0214, -0.039, -0.0097, -0.0097, -0.0195, 0.0, 0.0078, 0.0214, 0.078, 0.0819, 0.0721, 0.037, 0.039, 0.0604, 0.0429, 0.0292, 0.0351, 0.037, 0.0487, 0.0585, 0.0585, 0.0565, 0.0721, 0.0916, 0.0975, 0.076, 0.076, 0.1053, 0.1033, 0.117, 0.1209, 0.115, 0.1033, 0.1033, 0.0877, 0.0897, 0.0994, 0.1287, 0.1423, 0.1306, 0.1131, 0.1212, 0.1478, 0.1519, 0.1437, 0.1355, 0.1314, 0.111, 0.1274, 0.1151, 0.1151, 0.1008, 0.1049, 0.1131, 0.1069, 0.1171, 0.1171, 0.0722, 0.0804, 0.0988, 0.0967, 0.111, 0.0926, 0.1028, 0.1498, 0.1498, 0.1478, 0.1764, 0.1784, 0.1805, 0.1866, 0.1845, 0.1559, 0.109, 0.1049, 0.1049, 0.0988, 0.0947, 0.06, 0.0579, 0.0559, 0.0947, 0.0824, 0.0988, 0.0865, 0.0722, 0.0457, 0.0355, 0.0355, 0.0497, 0.0559, 0.062, 0.0722, 0.1151, 0.1028, 0.1192, 0.1274, 0.1171, 0.1151, 0.1417, 0.1478, 0.1478, 0.1702, 0.158, 0.1641, 0.1641, 0.1641, 0.1845, 0.1702, 0.2009, 0.2254, 0.2254, 0.2601, 0.2499, 0.2172, 0.2172, 0.2029, 0.16, 0.1682, 0.1907, 0.1784, 0.2111, 0.1886, 0.1825, 0.1988, 0.1988, 0.2642, 0.2417, 0.2581, 0.2499, 0.2274, 0.2152]}
{"unit_id": "orig:1920979784760471994", "ticker_idx": 1, "ts": "2025-05-09T23:10:45+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "hardware ODMs are expected to maintain momentum in Q2... positive outlooks from companies such as Wistron", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS bullish on Taiwanese ODMs (Quanta, Wistron, Wiwynn, Inventec, Hon Hai) and Supermicro into Q2/2H25 on Blackwell rack ramp and HGX demand expansion.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "UBS APAC Technology (May 7)\n\n1. UBS AI Market Outlook and Implications for the Asian Supply Chain\n\nUnlike Taiwanese ODMs, Supermicro has shown mixed quarterly results in recent quarters. While Supermicro and HPE were hit earlier this year by a reduction in Hopper product volumes, Taiwanese ODMs continued to deliver solid growth.\n\nIn Q1, Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%). Hon Hai met expectations.\n\nQ2 also started strong, with positive outlooks from companies such as Wistron and Delta. As Blackwell rack deployment gradually improves and demand increases for HGX for B200, hardware ODMs are expected to maintain momentum in Q2.\n\nFactors contributing to this growth include Taiwan's limited exposure to Hopper downsizing, the ability to supply H20 products before Chinese export restrictions, Wiwynn’s expansion in ASIC servers, and tariff cuts on iPhone and notebook assembly.\n\n2. Supermicro’s Position and Relationship with Taiwanese ODMs\n\nSupermicro supports clients like xAI and Coreweave and focuses on second- and third-tier supply chains that serve mid-sized data centers and channel customers. This structure is shared with OEM partners such as Wistron and, to some extent, Gigabyte.\n\nUBS believes Wistron is producing AI server motherboards on an outsourced basis, while final assembly is handled by Hon Hai. Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025.\n\n3. Rack Deployment and Growth Outlook for 2H25\n\nGB200 rack shipments were around 1,000 units in Q1 2025 but are expected to rise to over 5,000 in Q2, with further expansion projected in the second half. GB300 is expected to enter pilot production in Q3 and move into full production in Q4.\n\nDemand for HGX systems is also increasing. Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected for Inventec and Wistron.\n\nThe next momentum driver is Computex in early June, followed by increased capital expenditure from hyperscalers in Q2, which will drive further supply expansion to meet demand.\n\n4. Tariff Impact and Supply Chain Volatility\n\nKey variables include the timing of the reinstatement of Section 232 semiconductor tariffs and resolution of the delayed mutual tariff adjustment (90-day delay).\n\nIn the AI supply chain, the USMCA exemption for Mexico remains a major protective factor. Although volumes are relatively small, the high-value nature of these products may lead to accelerated relocation of assembly to the U.S. if needed (comment by Randy Abrams).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.0012758503827769996, "alpha_c_m1d": -0.0012407261135716219, "bench_spy_m1d": 0.0012758503827769996, "bench_c_m1d": 0.0012407261135716219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01401871949929534, "ret_signed_p1d": 0.01401871949929534, "alpha_spy_p1d": -0.019028791843116233, "alpha_c_p1d": -0.03239662366976437, "bench_spy_p1d": 0.03304751134241157, "bench_c_p1d": 0.04641534316905971, "ret_p1w": 0.037383178259241756, "ret_signed_p1w": 0.037383178259241756, "alpha_spy_p1w": -0.015528130065542545, "alpha_c_p1w": -0.04207449332528812, "bench_spy_p1w": 0.0529113083247843, "bench_c_p1w": 0.07945767158452988, "ret_p1m": 0.09698911621520168, "ret_signed_p1m": 0.09698911621520168, "alpha_spy_p1m": 0.03436732080064919, "alpha_c_p1m": -0.003792076484169238, "bench_spy_p1m": 0.06262179541455248, "bench_c_p1m": 0.10078119269937091, "ret_p3m": 0.18397503732477705, "ret_signed_p3m": 0.18397503732477705, "alpha_spy_p3m": 0.05938583533128683, "alpha_c_p3m": -0.02689790690496152, "bench_spy_p3m": 0.12458920199349022, "bench_c_p3m": 0.21087294422973857, "ret_p6m": 0.3869421865804532, "ret_signed_p6m": 0.3869421865804532, "alpha_spy_p6m": 0.16912970497214874, "alpha_c_p6m": -0.0045974901382130895, "bench_spy_p6m": 0.21781248160830446, "bench_c_p6m": 0.3915396767186663, "price_path": [0.0093, 0.014, 0.0467, 0.0421, 0.0561, 0.0841, 0.0841, 0.0888, 0.0981, 0.0748, 0.0467, 0.0748, 0.0467, 0.0467, 0.0, 0.014, 0.0327, 0.0, 0.0, -0.0093, -0.0234, -0.0234, -0.028, -0.0234, -0.0093, 0.0047, -0.0374, -0.0187, -0.014, -0.0327, -0.0327, -0.014, -0.0234, -0.0607, -0.1093, -0.0514, -0.0561, -0.0561, -0.0561, -0.1505, -0.2346, -0.3019, -0.2327, -0.1561, -0.0907, -0.0421, -0.0682, -0.0607, -0.0187, -0.0374, -0.0766, -0.0421, -0.0561, -0.0421, -0.0421, -0.028, -0.0514, -0.0514, -0.014, -0.0327, 0.0187, -0.0047, 0.0, 0.0, 0.014, 0.014, 0.0701, 0.0421, 0.0374, 0.0187, 0.0234, 0.0561, 0.0561, 0.0561, 0.0841, 0.0748, 0.0794, 0.0888, 0.0888, 0.0748, 0.0583, 0.1018, 0.1115, 0.0632, 0.097, 0.1067, 0.1115, 0.126, 0.1453, 0.1501, 0.1405, 0.1212, 0.1163, 0.1453, 0.1453, 0.1598, 0.213, 0.2081, 0.1695, 0.184, 0.1598, 0.1598, 0.1695, 0.1501, 0.1405, 0.1453, 0.1646, 0.1598, 0.155, 0.1308, 0.1598, 0.1405, 0.1405, 0.1501, 0.1453, 0.1067, 0.1212, 0.1308, 0.126, 0.1598, 0.1646, 0.1646, 0.1888, 0.1888, 0.155, 0.1985, 0.184, 0.1936, 0.2033, 0.2323, 0.2178, 0.1356, 0.126, 0.126, 0.1212, 0.1163, 0.097, 0.0777, 0.0825, 0.097, 0.0922, 0.1115, 0.1115, 0.0922, 0.0583, 0.0632, 0.0632, 0.0922, 0.1163, 0.1067, 0.1212, 0.1598, 0.1501, 0.1598, 0.155, 0.1405, 0.1501, 0.1598, 0.1695, 0.1743, 0.184, 0.1743, 0.213, 0.2371, 0.2371, 0.3579, 0.4111, 0.4449, 0.4836, 0.4836, 0.5078, 0.4546, 0.4594, 0.4594, 0.4063, 0.3338, 0.3338, 0.3483, 0.3193, 0.3724, 0.3773, 0.3869, 0.3821, 0.3821, 0.4159, 0.4304, 0.4739, 0.4981, 0.4546, 0.3869]}
{"unit_id": "orig:1920979784760471994", "ticker_idx": 2, "ts": "2025-05-09T23:10:45+00:00", "symbol": "Wiwynn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "hardware ODMs are expected to maintain momentum in Q2... Wiwynn's expansion in ASIC servers", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS bullish on Taiwanese ODMs (Quanta, Wistron, Wiwynn, Inventec, Hon Hai) and Supermicro into Q2/2H25 on Blackwell rack ramp and HGX demand expansion.", "resolved_tickers": ["6669.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wiwynn Corp listed on TWSE as 6669.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "UBS APAC Technology (May 7)\n\n1. UBS AI Market Outlook and Implications for the Asian Supply Chain\n\nUnlike Taiwanese ODMs, Supermicro has shown mixed quarterly results in recent quarters. While Supermicro and HPE were hit earlier this year by a reduction in Hopper product volumes, Taiwanese ODMs continued to deliver solid growth.\n\nIn Q1, Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%). Hon Hai met expectations.\n\nQ2 also started strong, with positive outlooks from companies such as Wistron and Delta. As Blackwell rack deployment gradually improves and demand increases for HGX for B200, hardware ODMs are expected to maintain momentum in Q2.\n\nFactors contributing to this growth include Taiwan's limited exposure to Hopper downsizing, the ability to supply H20 products before Chinese export restrictions, Wiwynn’s expansion in ASIC servers, and tariff cuts on iPhone and notebook assembly.\n\n2. Supermicro’s Position and Relationship with Taiwanese ODMs\n\nSupermicro supports clients like xAI and Coreweave and focuses on second- and third-tier supply chains that serve mid-sized data centers and channel customers. This structure is shared with OEM partners such as Wistron and, to some extent, Gigabyte.\n\nUBS believes Wistron is producing AI server motherboards on an outsourced basis, while final assembly is handled by Hon Hai. Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025.\n\n3. Rack Deployment and Growth Outlook for 2H25\n\nGB200 rack shipments were around 1,000 units in Q1 2025 but are expected to rise to over 5,000 in Q2, with further expansion projected in the second half. GB300 is expected to enter pilot production in Q3 and move into full production in Q4.\n\nDemand for HGX systems is also increasing. Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected for Inventec and Wistron.\n\nThe next momentum driver is Computex in early June, followed by increased capital expenditure from hyperscalers in Q2, which will drive further supply expansion to meet demand.\n\n4. Tariff Impact and Supply Chain Volatility\n\nKey variables include the timing of the reinstatement of Section 232 semiconductor tariffs and resolution of the delayed mutual tariff adjustment (90-day delay).\n\nIn the AI supply chain, the USMCA exemption for Mexico remains a major protective factor. Although volumes are relatively small, the high-value nature of these products may lead to accelerated relocation of assembly to the U.S. if needed (comment by Randy Abrams).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022727296104700256, "ret_signed_m1d": -0.022727296104700256, "alpha_spy_m1d": -0.024003146487477256, "alpha_c_m1d": -0.02396802221827188, "bench_spy_m1d": 0.0012758503827769996, "bench_c_m1d": 0.0012407261135716219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024999945712418148, "ret_signed_p1d": 0.024999945712418148, "alpha_spy_p1d": -0.008047565629993425, "alpha_c_p1d": -0.02141539745664156, "bench_spy_p1d": 0.03304751134241157, "bench_c_p1d": 0.04641534316905971, "ret_p1w": 0.08181812882934603, "ret_signed_p1w": 0.08181812882934603, "alpha_spy_p1w": 0.028906820504561725, "alpha_c_p1w": 0.00236045724481615, "bench_spy_p1w": 0.0529113083247843, "bench_c_p1w": 0.07945767158452988, "ret_p1m": 0.12045454363630115, "ret_signed_p1m": 0.12045454363630115, "alpha_spy_p1m": 0.057832748221748664, "alpha_c_p1m": 0.019673350936930234, "bench_spy_p1m": 0.06262179541455248, "bench_c_p1m": 0.10078119269937091, "ret_p3m": 0.29906183058411484, "ret_signed_p3m": 0.29906183058411484, "alpha_spy_p3m": 0.17447262859062462, "alpha_c_p3m": 0.08818888635437627, "bench_spy_p3m": 0.12458920199349022, "bench_c_p3m": 0.21087294422973857, "ret_p6m": 1.0199826302596238, "ret_signed_p6m": 1.0199826302596238, "alpha_spy_p6m": 0.8021701486513193, "alpha_c_p6m": 0.6284429535409575, "bench_spy_p6m": 0.21781248160830446, "bench_c_p6m": 0.3915396767186663, "price_path": [0.0114, 0.0091, -0.0364, -0.0705, -0.05, -0.0091, -0.0386, -0.0545, -0.0545, -0.0727, -0.1205, -0.0795, -0.1068, -0.1068, -0.1273, -0.1205, -0.0682, -0.0636, -0.0773, -0.0909, -0.1091, -0.1159, -0.1114, -0.1273, -0.1318, -0.0977, -0.1159, -0.0841, -0.1136, -0.1386, -0.1295, -0.1273, -0.1614, -0.2023, -0.2568, -0.2159, -0.2091, -0.2091, -0.2091, -0.2864, -0.2545, -0.2705, -0.1977, -0.1364, -0.1318, -0.1545, -0.1636, -0.1455, -0.1432, -0.1636, -0.1955, -0.1364, -0.1523, -0.1341, -0.1364, -0.1227, -0.1318, -0.1318, -0.0614, -0.0727, -0.0341, -0.05, -0.0227, 0.0, 0.025, 0.0114, 0.0659, 0.0818, 0.0818, 0.0273, 0.0455, 0.0659, 0.0636, 0.0364, 0.0773, 0.0568, 0.0818, 0.1, 0.1, 0.0955, 0.1273, 0.1523, 0.1295, 0.1205, 0.1205, 0.1295, 0.1477, 0.1523, 0.1659, 0.1477, 0.1591, 0.1636, 0.15, 0.1523, 0.1591, 0.1773, 0.2686, 0.2265, 0.1844, 0.1844, 0.1469, 0.1656, 0.1563, 0.1493, 0.1493, 0.1329, 0.1984, 0.2148, 0.2195, 0.1539, 0.1797, 0.1727, 0.1493, 0.2125, 0.182, 0.1563, 0.1446, 0.175, 0.1773, 0.2171, 0.2312, 0.2405, 0.2944, 0.2944, 0.2382, 0.2897, 0.2991, 0.3154, 0.3529, 0.4863, 0.5308, 0.5846, 0.5893, 0.5659, 0.6033, 0.5472, 0.3974, 0.4184, 0.3903, 0.4278, 0.4301, 0.3974, 0.3927, 0.388, 0.3271, 0.3506, 0.3693, 0.374, 0.4161, 0.4629, 0.4091, 0.5495, 0.5495, 0.5167, 0.5074, 0.498, 0.4793, 0.498, 0.5121, 0.4629, 0.4699, 0.4418, 0.4489, 0.4278, 0.4278, 0.5519, 0.5448, 0.5519, 0.6221, 0.6221, 0.6431, 0.608, 0.5776, 0.5776, 0.6057, 0.6431, 0.7766, 0.807, 0.7719, 0.8163, 0.903, 0.903, 0.9217, 0.9217, 0.8983, 0.9123, 0.924, 1.013, 1.0434, 1.02]}
{"unit_id": "orig:1920979784760471994", "ticker_idx": 3, "ts": "2025-05-09T23:10:45+00:00", "symbol": "Inventec", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "further volume expansion is expected for Inventec and Wistron", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS bullish on Taiwanese ODMs (Quanta, Wistron, Wiwynn, Inventec, Hon Hai) and Supermicro into Q2/2H25 on Blackwell rack ramp and HGX demand expansion.", "resolved_tickers": ["2356.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Inventec Corporation 2356.TW Taiwan server ODM", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "UBS APAC Technology (May 7)\n\n1. UBS AI Market Outlook and Implications for the Asian Supply Chain\n\nUnlike Taiwanese ODMs, Supermicro has shown mixed quarterly results in recent quarters. While Supermicro and HPE were hit earlier this year by a reduction in Hopper product volumes, Taiwanese ODMs continued to deliver solid growth.\n\nIn Q1, Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%). Hon Hai met expectations.\n\nQ2 also started strong, with positive outlooks from companies such as Wistron and Delta. As Blackwell rack deployment gradually improves and demand increases for HGX for B200, hardware ODMs are expected to maintain momentum in Q2.\n\nFactors contributing to this growth include Taiwan's limited exposure to Hopper downsizing, the ability to supply H20 products before Chinese export restrictions, Wiwynn’s expansion in ASIC servers, and tariff cuts on iPhone and notebook assembly.\n\n2. Supermicro’s Position and Relationship with Taiwanese ODMs\n\nSupermicro supports clients like xAI and Coreweave and focuses on second- and third-tier supply chains that serve mid-sized data centers and channel customers. This structure is shared with OEM partners such as Wistron and, to some extent, Gigabyte.\n\nUBS believes Wistron is producing AI server motherboards on an outsourced basis, while final assembly is handled by Hon Hai. Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025.\n\n3. Rack Deployment and Growth Outlook for 2H25\n\nGB200 rack shipments were around 1,000 units in Q1 2025 but are expected to rise to over 5,000 in Q2, with further expansion projected in the second half. GB300 is expected to enter pilot production in Q3 and move into full production in Q4.\n\nDemand for HGX systems is also increasing. Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected for Inventec and Wistron.\n\nThe next momentum driver is Computex in early June, followed by increased capital expenditure from hyperscalers in Q2, which will drive further supply expansion to meet demand.\n\n4. Tariff Impact and Supply Chain Volatility\n\nKey variables include the timing of the reinstatement of Section 232 semiconductor tariffs and resolution of the delayed mutual tariff adjustment (90-day delay).\n\nIn the AI supply chain, the USMCA exemption for Mexico remains a major protective factor. Although volumes are relatively small, the high-value nature of these products may lead to accelerated relocation of assembly to the U.S. if needed (comment by Randy Abrams).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0036231543736731098, "ret_signed_m1d": -0.0036231543736731098, "alpha_spy_m1d": -0.004899004756450109, "alpha_c_m1d": -0.004863880487244732, "bench_spy_m1d": 0.0012758503827769996, "bench_c_m1d": 0.0012407261135716219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028985522526105978, "ret_signed_p1d": 0.028985522526105978, "alpha_spy_p1d": -0.004061988816305595, "alpha_c_p1d": -0.01742982064295373, "bench_spy_p1d": 0.03304751134241157, "bench_c_p1d": 0.04641534316905971, "ret_p1w": -0.00241550014616454, "ret_signed_p1w": -0.00241550014616454, "alpha_spy_p1w": -0.05532680847094884, "alpha_c_p1w": -0.08187317173069442, "bench_spy_p1w": 0.0529113083247843, "bench_c_p1w": 0.07945767158452988, "ret_p1m": 0.0, "ret_signed_p1m": 0.0, "alpha_spy_p1m": -0.06262179541455248, "alpha_c_p1m": -0.10078119269937091, "bench_spy_p1m": 0.06262179541455248, "bench_c_p1m": 0.10078119269937091, "ret_p3m": 0.10425880221786654, "ret_signed_p3m": 0.10425880221786654, "alpha_spy_p3m": -0.020330399775623675, "alpha_c_p3m": -0.10661414201187203, "bench_spy_p3m": 0.12458920199349022, "bench_c_p3m": 0.21087294422973857, "ret_p6m": 0.14445931115786248, "ret_signed_p6m": 0.14445931115786248, "alpha_spy_p6m": -0.07335317045044198, "alpha_c_p6m": -0.24708036556080382, "bench_spy_p6m": 0.21781248160830446, "bench_c_p6m": 0.3915396767186663, "price_path": [0.099, 0.122, 0.1594, 0.1461, 0.1582, 0.1715, 0.1341, 0.1256, 0.1715, 0.1715, 0.1522, 0.1582, 0.1232, 0.1232, 0.1196, 0.1473, 0.151, 0.1341, 0.1824, 0.1618, 0.1546, 0.1268, 0.1111, 0.1232, 0.1268, 0.1498, 0.1377, 0.1413, 0.1473, 0.1111, 0.1002, 0.1075, 0.0797, 0.064, 0.0193, 0.0459, 0.035, 0.035, 0.035, -0.0676, -0.1135, -0.1812, -0.1002, -0.0338, 0.006, -0.0012, -0.0242, -0.0242, -0.0386, -0.0374, -0.0495, -0.0266, -0.0362, -0.029, -0.0169, 0.0024, -0.0181, -0.0181, 0.0072, -0.0145, -0.0024, -0.0266, -0.0036, 0.0, 0.029, 0.0362, 0.0193, 0.0085, -0.0024, 0.0012, -0.0024, 0.0254, 0.0169, 0.0024, 0.029, 0.0048, 0.0109, 0.0181, 0.0181, -0.0242, -0.0205, 0.0048, 0.0048, -0.0072, 0.0, 0.0109, 0.0169, 0.0399, 0.058, 0.0604, 0.0338, 0.0205, -0.0048, -0.0254, -0.029, -0.0085, -0.0024, 0.0169, 0.0266, 0.0217, 0.0338, 0.0628, 0.0713, 0.0628, 0.0543, 0.029, 0.0628, 0.0477, 0.0553, 0.0465, 0.1005, 0.0678, 0.0666, 0.0854, 0.0904, 0.0641, 0.0741, 0.0729, 0.0879, 0.0992, 0.0967, 0.1093, 0.1181, 0.1181, 0.0967, 0.1181, 0.1043, 0.1105, 0.0754, 0.0766, 0.0578, 0.0528, 0.049, 0.0339, 0.0477, 0.0276, 0.0352, 0.0276, 0.0276, 0.0301, 0.0402, 0.0402, 0.0389, 0.0339, 0.0276, 0.0188, 0.0188, 0.0377, 0.0477, 0.0515, 0.0603, 0.0553, 0.0477, 0.0904, 0.0955, 0.0867, 0.0741, 0.0879, 0.1193, 0.1218, 0.1306, 0.1218, 0.1834, 0.1281, 0.1281, 0.1495, 0.1319, 0.1432, 0.1495, 0.1495, 0.1683, 0.1784, 0.1646, 0.1646, 0.1369, 0.0816, 0.0904, 0.1005, 0.1017, 0.1332, 0.1407, 0.1558, 0.152, 0.152, 0.1746, 0.1809, 0.1947, 0.1734, 0.1533, 0.1445]}
{"unit_id": "orig:1920979784760471994", "ticker_idx": 4, "ts": "2025-05-09T23:10:45+00:00", "symbol": "Hon Hai", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS bullish on Taiwanese ODMs (Quanta, Wistron, Wiwynn, Inventec, Hon Hai) and Supermicro into Q2/2H25 on Blackwell rack ramp and HGX demand expansion.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "UBS APAC Technology (May 7)\n\n1. UBS AI Market Outlook and Implications for the Asian Supply Chain\n\nUnlike Taiwanese ODMs, Supermicro has shown mixed quarterly results in recent quarters. While Supermicro and HPE were hit earlier this year by a reduction in Hopper product volumes, Taiwanese ODMs continued to deliver solid growth.\n\nIn Q1, Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%). Hon Hai met expectations.\n\nQ2 also started strong, with positive outlooks from companies such as Wistron and Delta. As Blackwell rack deployment gradually improves and demand increases for HGX for B200, hardware ODMs are expected to maintain momentum in Q2.\n\nFactors contributing to this growth include Taiwan's limited exposure to Hopper downsizing, the ability to supply H20 products before Chinese export restrictions, Wiwynn’s expansion in ASIC servers, and tariff cuts on iPhone and notebook assembly.\n\n2. Supermicro’s Position and Relationship with Taiwanese ODMs\n\nSupermicro supports clients like xAI and Coreweave and focuses on second- and third-tier supply chains that serve mid-sized data centers and channel customers. This structure is shared with OEM partners such as Wistron and, to some extent, Gigabyte.\n\nUBS believes Wistron is producing AI server motherboards on an outsourced basis, while final assembly is handled by Hon Hai. Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025.\n\n3. Rack Deployment and Growth Outlook for 2H25\n\nGB200 rack shipments were around 1,000 units in Q1 2025 but are expected to rise to over 5,000 in Q2, with further expansion projected in the second half. GB300 is expected to enter pilot production in Q3 and move into full production in Q4.\n\nDemand for HGX systems is also increasing. Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected for Inventec and Wistron.\n\nThe next momentum driver is Computex in early June, followed by increased capital expenditure from hyperscalers in Q2, which will drive further supply expansion to meet demand.\n\n4. Tariff Impact and Supply Chain Volatility\n\nKey variables include the timing of the reinstatement of Section 232 semiconductor tariffs and resolution of the delayed mutual tariff adjustment (90-day delay).\n\nIn the AI supply chain, the USMCA exemption for Mexico remains a major protective factor. Although volumes are relatively small, the high-value nature of these products may lead to accelerated relocation of assembly to the U.S. if needed (comment by Randy Abrams).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0068027986567377186, "ret_signed_m1d": -0.0068027986567377186, "alpha_spy_m1d": -0.008078649039514718, "alpha_c_m1d": -0.00804352477030934, "bench_spy_m1d": 0.0012758503827769996, "bench_c_m1d": 0.0012407261135716219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.040816254084744985, "ret_signed_p1d": 0.040816254084744985, "alpha_spy_p1d": 0.007768742742333412, "alpha_c_p1d": -0.005599089084314723, "bench_spy_p1d": 0.03304751134241157, "bench_c_p1d": 0.04641534316905971, "ret_p1w": 0.0748299246550248, "ret_signed_p1w": 0.0748299246550248, "alpha_spy_p1w": 0.021918616330240503, "alpha_c_p1w": -0.004627746929505072, "bench_spy_p1w": 0.0529113083247843, "bench_c_p1w": 0.07945767158452988, "ret_p1m": 0.051020344498715176, "ret_signed_p1m": 0.051020344498715176, "alpha_spy_p1m": -0.011601450915837308, "alpha_c_p1m": -0.04976084820065574, "bench_spy_p1m": 0.06262179541455248, "bench_c_p1m": 0.10078119269937091, "ret_p3m": 0.31125944931753646, "ret_signed_p3m": 0.31125944931753646, "alpha_spy_p3m": 0.18667024732404625, "alpha_c_p3m": 0.1003865050877979, "bench_spy_p3m": 0.12458920199349022, "bench_c_p3m": 0.21087294422973857, "ret_p6m": 0.7730201693729055, "ret_signed_p6m": 0.7730201693729055, "alpha_spy_p6m": 0.555207687764601, "alpha_c_p6m": 0.3814804926542392, "bench_spy_p6m": 0.21781248160830446, "bench_c_p6m": 0.3915396767186663, "price_path": [0.2177, 0.2041, 0.2245, 0.2109, 0.2415, 0.2483, 0.2449, 0.2381, 0.2449, 0.2517, 0.2075, 0.2279, 0.1837, 0.1837, 0.1565, 0.1599, 0.1803, 0.1803, 0.1701, 0.1701, 0.1463, 0.1599, 0.1429, 0.1565, 0.1429, 0.1395, 0.1156, 0.1293, 0.1224, 0.1088, 0.1224, 0.1293, 0.0918, 0.0476, -0.0068, 0.034, 0.0442, 0.0442, 0.0442, -0.0578, -0.1497, -0.2347, -0.1599, -0.085, -0.0578, -0.051, -0.0748, -0.085, -0.0782, -0.0714, -0.102, -0.0544, -0.0714, -0.0544, -0.0306, -0.0272, -0.0374, -0.0374, 0.0034, -0.0306, -0.0068, -0.0136, -0.0068, 0.0, 0.0408, 0.0748, 0.1088, 0.085, 0.0748, 0.0476, 0.0476, 0.0544, 0.0408, 0.0476, 0.0408, 0.034, 0.034, 0.0612, 0.0612, 0.0374, 0.0306, 0.0612, 0.0578, 0.0408, 0.051, 0.0578, 0.0612, 0.0714, 0.0646, 0.0544, 0.0544, 0.0612, 0.0476, 0.0578, 0.0612, 0.0918, 0.102, 0.1088, 0.1224, 0.0952, 0.1259, 0.1421, 0.1562, 0.135, 0.135, 0.1209, 0.135, 0.1385, 0.1385, 0.1315, 0.1491, 0.1456, 0.1562, 0.1667, 0.1632, 0.1421, 0.1703, 0.2267, 0.2302, 0.2408, 0.209, 0.209, 0.2549, 0.2549, 0.269, 0.3007, 0.3113, 0.3712, 0.3712, 0.3959, 0.3888, 0.3994, 0.4064, 0.4593, 0.4805, 0.4664, 0.4135, 0.4523, 0.4276, 0.4628, 0.4734, 0.4664, 0.4523, 0.4346, 0.3994, 0.3994, 0.4135, 0.4311, 0.4452, 0.4346, 0.4628, 0.4805, 0.5122, 0.5333, 0.5228, 0.5192, 0.4946, 0.5157, 0.5087, 0.5228, 0.558, 0.5721, 0.6214, 0.5474, 0.5474, 0.5228, 0.5439, 0.5897, 0.5968, 0.5968, 0.6109, 0.5862, 0.5615, 0.5615, 0.5016, 0.4523, 0.4558, 0.5721, 0.5968, 0.6814, 0.6849, 0.7096, 0.6849, 0.6849, 0.7378, 0.7624, 0.8329, 0.847, 0.8153, 0.773]}
{"unit_id": "orig:1920979784760471994", "ticker_idx": 5, "ts": "2025-05-09T23:10:45+00:00", "symbol": "SMCI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected", "tweet_type": "analysis", "timeline": "quarters", "summary": "UBS bullish on Taiwanese ODMs (Quanta, Wistron, Wiwynn, Inventec, Hon Hai) and Supermicro into Q2/2H25 on Blackwell rack ramp and HGX demand expansion.", "resolved_tickers": ["SMCI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "UBS APAC Technology (May 7)\n\n1. UBS AI Market Outlook and Implications for the Asian Supply Chain\n\nUnlike Taiwanese ODMs, Supermicro has shown mixed quarterly results in recent quarters. While Supermicro and HPE were hit earlier this year by a reduction in Hopper product volumes, Taiwanese ODMs continued to deliver solid growth.\n\nIn Q1, Taiwanese ODMs exceeded market expectations: Quanta (+16%), Wistron (+20%), Wiwynn (+38%), and Inventec (+3%). Hon Hai met expectations.\n\nQ2 also started strong, with positive outlooks from companies such as Wistron and Delta. As Blackwell rack deployment gradually improves and demand increases for HGX for B200, hardware ODMs are expected to maintain momentum in Q2.\n\nFactors contributing to this growth include Taiwan's limited exposure to Hopper downsizing, the ability to supply H20 products before Chinese export restrictions, Wiwynn’s expansion in ASIC servers, and tariff cuts on iPhone and notebook assembly.\n\n2. Supermicro’s Position and Relationship with Taiwanese ODMs\n\nSupermicro supports clients like xAI and Coreweave and focuses on second- and third-tier supply chains that serve mid-sized data centers and channel customers. This structure is shared with OEM partners such as Wistron and, to some extent, Gigabyte.\n\nUBS believes Wistron is producing AI server motherboards on an outsourced basis, while final assembly is handled by Hon Hai. Sales of GB200 racks are expected to visibly ramp up starting in Q2 2025.\n\n3. Rack Deployment and Growth Outlook for 2H25\n\nGB200 rack shipments were around 1,000 units in Q1 2025 but are expected to rise to over 5,000 in Q2, with further expansion projected in the second half. GB300 is expected to enter pilot production in Q3 and move into full production in Q4.\n\nDemand for HGX systems is also increasing. Supermicro is supplying these in the form of B200 HGX racks, and further volume expansion is expected for Inventec and Wistron.\n\nThe next momentum driver is Computex in early June, followed by increased capital expenditure from hyperscalers in Q2, which will drive further supply expansion to meet demand.\n\n4. Tariff Impact and Supply Chain Volatility\n\nKey variables include the timing of the reinstatement of Section 232 semiconductor tariffs and resolution of the delayed mutual tariff adjustment (90-day delay).\n\nIn the AI supply chain, the USMCA exemption for Mexico remains a major protective factor. Although volumes are relatively small, the high-value nature of these products may lead to accelerated relocation of assembly to the U.S. if needed (comment by Randy Abrams).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00375119850240635, "ret_signed_m1d": 0.00375119850240635, "alpha_spy_m1d": 0.0024753481196293503, "alpha_c_m1d": 0.002510472388834728, "bench_spy_m1d": 0.0012758503827769996, "bench_c_m1d": 0.0012407261135716219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04782746788347447, "ret_signed_p1d": 0.04782746788347447, "alpha_spy_p1d": 0.0147799565410629, "alpha_c_p1d": 0.0014121247144147642, "bench_spy_p1d": 0.03304751134241157, "bench_c_p1d": 0.04641534316905971, "ret_p1w": 0.4426383824967999, "ret_signed_p1w": 0.4426383824967999, "alpha_spy_p1w": 0.3897270741720156, "alpha_c_p1w": 0.36318071091227, "bench_spy_p1w": 0.0529113083247843, "bench_c_p1w": 0.07945767158452988, "ret_p1m": 0.34792120163799445, "ret_signed_p1m": 0.34792120163799445, "alpha_spy_p1m": 0.28529940622344196, "alpha_c_p1m": 0.24714000893862353, "bench_spy_p1m": 0.06262179541455248, "bench_c_p1m": 0.10078119269937091, "ret_p3m": 0.4626446155142272, "ret_signed_p3m": 0.4626446155142272, "alpha_spy_p3m": 0.338055413520737, "alpha_c_p3m": 0.25177167128448863, "bench_spy_p3m": 0.12458920199349022, "bench_c_p3m": 0.21087294422973857, "ret_p6m": 0.5864332717444751, "ret_signed_p6m": 0.5864332717444751, "alpha_spy_p6m": 0.36862079013617066, "alpha_c_p6m": 0.19489359502580883, "bench_spy_p6m": 0.21781248160830446, "bench_c_p6m": 0.3915396767186663, "price_path": [0.2069, 0.2404, 0.3217, 0.4977, 0.4977, 0.7443, 0.8834, 0.8528, 0.7527, 0.6133, 0.4236, 0.5977, 0.3426, 0.296, 0.1275, 0.2235, 0.216, 0.1582, 0.1954, 0.1535, 0.2766, 0.3276, 0.2216, 0.3182, 0.311, 0.1847, 0.2535, 0.2223, 0.3176, 0.3042, 0.2704, 0.1579, 0.0853, 0.071, 0.0703, 0.0957, 0.0969, 0.0103, -0.0678, 0.0316, -0.0088, 0.1475, 0.0528, 0.0363, 0.0353, 0.0466, -0.0031, -0.015, -0.015, -0.0775, -0.0441, 0.0284, 0.1194, 0.14, 0.1651, 0.1254, -0.0041, 0.0231, 0.0538, 0.0056, 0.0297, 0.0153, 0.0038, 0.0, 0.0478, 0.2157, 0.4067, 0.3742, 0.4426, 0.4001, 0.337, 0.302, 0.2917, 0.2532, 0.2532, 0.2995, 0.3117, 0.2863, 0.251, 0.2879, 0.3501, 0.3795, 0.2745, 0.2988, 0.3479, 0.3414, 0.3507, 0.3554, 0.2992, 0.3657, 0.3476, 0.3901, 0.3901, 0.4167, 0.2782, 0.3392, 0.457, 0.5402, 0.4873, 0.532, 0.4755, 0.5236, 0.518, 0.518, 0.4726, 0.5352, 0.5636, 0.5742, 0.5392, 0.5545, 0.6621, 0.6636, 0.649, 0.6183, 0.6099, 0.5586, 0.6161, 0.6418, 0.7027, 0.8771, 0.8328, 0.8978, 0.8434, 0.7706, 0.8203, 0.7899, 0.4626, 0.4589, 0.3942, 0.4129, 0.4514, 0.4401, 0.422, 0.4183, 0.4336, 0.3517, 0.332, 0.3226, 0.3717, 0.3776, 0.3867, 0.3992, 0.3745, 0.2985, 0.2985, 0.2748, 0.2535, 0.2723, 0.2632, 0.2516, 0.3417, 0.3726, 0.3739, 0.4067, 0.4186, 0.4039, 0.4036, 0.4361, 0.432, 0.4651, 0.4689, 0.4442, 0.4455, 0.4323, 0.4492, 0.4986, 0.6377, 0.6411, 0.6243, 0.7074, 0.7215, 0.8343, 0.8124, 0.6524, 0.7115, 0.6602, 0.6862, 0.683, 0.6311, 0.7205, 0.7137, 0.6411, 0.498, 0.5095, 0.6121, 0.6368, 0.6471, 0.5746, 0.6243, 0.5864]}
{"unit_id": "orig:1921065866072125461", "ticker_idx": 0, "ts": "2025-05-10T04:52:48+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI demand is driving robust growth...price target has been slightly lowered from $112 to $98...stock price could potentially return to the high range of $112–$140", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's bullish MU view: EPS upgrades, AI revenue >1/3 of sales by Q4 2025, PT $98 with upside to $112–$140.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s Evaluation of Micron $MU\n\n1. AI Demand: AI demand is driving robust growth, with AI-related revenue (HBM and LPDDR) expected to increase by 2 to 2.5 times over the next three quarters. By Q4 2025, AI revenue could account for more than one-third of total sales, significantly boosting DRAM average selling prices (ASP).\n\n2. Earnings Forecast Upgrade: The EPS forecasts for the August and November quarters have been raised by 25% and 15%, respectively, compared to market expectations. However, the 2026 forecast has been revised downward to reflect macroeconomic risks. The price target has been slightly lowered from $112 to $98.\n\n3. Data Center Demand Surge: Strong demand for AI inference, combined with tight inventory levels, is driving substantial growth in data center demand. As a result, pricing is expected to remain supported even in the face of macroeconomic or trade-related disruptions.\n\n4. Current Valuation: Micron’s current valuation is considered to be at a mid-cycle level. If EPS exceeds $2.5, the stock price could potentially return to the high range of $112–$140.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06967173414031069, "ret_signed_m1d": -0.06967173414031069, "alpha_spy_m1d": -0.03768142296917443, "alpha_c_m1d": -0.010659852037967932, "bench_spy_m1d": -0.031990311171136265, "bench_c_m1d": -0.05901188210234276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05027631913789388, "ret_signed_p1d": 0.05027631913789388, "alpha_spy_p1d": 0.043672378820723834, "alpha_c_p1d": 0.015947559374998255, "bench_spy_p1d": 0.006603940317170043, "bench_c_p1d": 0.03432875976289562, "ret_p1w": 0.068913266260344, "ret_signed_p1w": 0.068913266260344, "alpha_spy_p1w": 0.048569965986598795, "alpha_c_p1w": 0.03268903333222295, "bench_spy_p1w": 0.020343300273745202, "bench_c_p1w": 0.03622423292812105, "ret_p1m": 0.23675368123748686, "ret_signed_p1m": 0.23675368123748686, "alpha_spy_p1m": 0.20229351078924096, "alpha_c_p1m": 0.13288271886127534, "bench_spy_p1m": 0.0344601704482459, "bench_c_p1m": 0.10387096237621152, "ret_p3m": 0.21329835415038878, "ret_signed_p3m": 0.21329835415038878, "alpha_spy_p3m": 0.12559695933220105, "alpha_c_p3m": -0.01293468499132544, "bench_spy_p3m": 0.08770139481818773, "bench_c_p3m": 0.22623303914171422, "ret_p6m": 1.366148981056865, "ret_signed_p6m": 1.366148981056865, "alpha_spy_p6m": 1.2012682959938907, "alpha_c_p6m": 0.8802806038739521, "bench_spy_p6m": 0.16488068506297426, "bench_c_p6m": 0.48586837718291287, "price_path": [-0.0079, 0.0352, 0.0769, 0.0769, 0.1556, 0.1293, 0.1165, 0.0696, 0.0325, 0.0088, 0.0574, -0.0064, 0.0132, -0.0202, -0.0131, 0.0209, -0.034, 0.0059, -0.0577, -0.0364, 0.035, 0.0267, 0.0907, 0.1158, 0.1007, 0.1044, 0.1146, 0.025, 0.049, 0.0192, -0.003, -0.0135, -0.043, -0.0585, -0.0388, -0.04, -0.1945, -0.2987, -0.2592, -0.2898, -0.1562, -0.241, -0.2464, -0.2305, -0.2303, -0.2488, -0.2545, -0.2545, -0.2768, -0.2392, -0.2098, -0.1611, -0.1356, -0.1488, -0.167, -0.1662, -0.1573, -0.1254, -0.1286, -0.1276, -0.1048, -0.0774, -0.0697, 0.0, 0.0503, 0.0328, 0.0342, 0.0619, 0.0689, 0.063, 0.0385, 0.0275, 0.0117, 0.0117, 0.0443, 0.0421, 0.0489, 0.0235, 0.0638, 0.1079, 0.1188, 0.1517, 0.1763, 0.2022, 0.2368, 0.2572, 0.2589, 0.2526, 0.2985, 0.3039, 0.32, 0.32, 0.3393, 0.3228, 0.386, 0.3788, 0.3653, 0.3518, 0.3355, 0.3099, 0.3191, 0.3251, 0.3251, 0.3006, 0.3494, 0.3258, 0.3352, 0.3506, 0.2864, 0.3027, 0.2628, 0.2284, 0.2406, 0.228, 0.1846, 0.1912, 0.2118, 0.2067, 0.2066, 0.2143, 0.2444, 0.1837, 0.1375, 0.1688, 0.1828, 0.1798, 0.2133, 0.2894, 0.3418, 0.3855, 0.3478, 0.3588, 0.3109, 0.34, 0.3237, 0.2712, 0.2558, 0.2763, 0.2626, 0.2635, 0.2771, 0.3232, 0.2907, 0.2907, 0.285, 0.2876, 0.3471, 0.4248, 0.4258, 0.4668, 0.5184, 0.633, 0.7053, 0.7111, 0.7225, 0.7352, 0.8317, 0.7649, 0.7854, 0.8048, 0.7538, 0.7009, 0.7057, 0.7776, 0.8147, 0.9755, 0.9929, 1.0384, 1.0724, 1.0152, 1.1329, 1.0872, 0.9708, 1.092, 1.0301, 1.083, 1.1979, 1.1963, 1.244, 1.1953, 1.1539, 1.2433, 1.3769, 1.3886, 1.4083, 1.4595, 1.431, 1.4284, 1.5471, 1.3661]}
{"unit_id": "orig:1932072542451495091", "ticker_idx": 0, "ts": "2025-06-09T13:49:24+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if AMZN joins MSFT and META as the third-largest customer, shipment volumes will expand further. More importantly, investors are watching the MI400 rack-scale solution's release timing and performance jump… if realized smoothly, it will deliver breakthrough market share and revenue growth for AMD", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS bullish thesis on AMD (MI400 ramp/new customers), MRVL (ASIC TAM upside to $8-9B), and MU (HBM scarcity supports above-consensus EPS in FY26/27).", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Brief: AMD / Marvell / Micron\nUBS (T. Arcuri, 25/06/08)\n\nAs AI compute demand accelerates, vendors such as AMD, Marvell, and Micron will face significant growth opportunities amid tight supply and demand for GPUs, custom ASICs, and HBM.\n\n• AMD: AI new-product push and MI400 on the horizon\n\nAMD will spotlight the MI355X at this week’s AI launch, slated for deployment in 3Q25, offering performance comparable to NVIDIA’s Blackwell B200/B300 at more competitive pricing. The key point is that if AMZN joins MSFT and META as the third-largest customer, shipment volumes will expand further. More importantly, investors are watching the MI400 rack-scale solution’s release timing and performance jump: a dual Active Interposer Die architecture, each package containing 288 GB of HBM4, with an in-house UA Link interconnect, expected to enter mass production as early as end-2026; if realized smoothly, it will deliver breakthrough market share and revenue growth for AMD.\n\n• Marvell: Custom ASIC outlook and TAM expansion\n\nMarvell had originally planned an AI Investor Day in June, later changed to an online seminar on June 17, and is expected to raise its 2028 TAM forecasts for accelerated-compute ASICs and the data-center market (2028 TAM: last MRVL forecast was $43 billion for ASICs and $75 billion for data centers). The company will reaffirm its target share above 20% and showcase progress on multiple SKUs in collaboration with AMZN on custom network accelerators. If Marvell can lift its 2028 custom-ASIC revenue from the Street’s $4.5 billion estimate to $8–9 billion, there is clear upside for the share price.\n\n• Micron: HBM supply–demand stress scenario simulation\n\nMicron simulated new AI compute deployment scenarios for C 2026 and found that if GPU shipments exceed the baseline by 20%, overall HBM supply–demand will be in shortage; if driven solely by XPU growth, a 75% increment is required to cause shortage; and under a 70% GPU / 30% XPU mix, roughly a 35% increment will tighten the HBM market. This outcome underscores HBM’s scarcity over the next two years, supporting Micron’s FY 2026 EPS above $4 and FY 2027 EPS above $5, and providing a rationale for valuation premium.\n\n• Industry consensus and key points\n\nUnder AI-driven acceleration of hardware refresh cycles for GPUs and custom ASICs, demand for high-bandwidth memory (HBM) is rising in tandem. Advanced nodes such as TSMC’s 2 nm (N2) will also scale up due to flagship SoC and server-chip migrations. Investors should monitor: the release timing and customer-base expansion for AMD’s MI400 and Marvell’s custom ASICs; HBM capacity ramps and price trends; and foundry N2 capacity expansion along with mature-node price resilience. These factors will determine semiconductor vendors’ market positions and profitability in the AI and high-performance computing wave.\n\n$AMD $MRVL $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04551056241478191, "ret_signed_m1d": -0.04551056241478191, "alpha_spy_m1d": -0.044610091106019434, "alpha_c_m1d": -0.02904893638097139, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012404456298324096, "ret_signed_p1d": 0.012404456298324096, "alpha_spy_p1d": 0.0067347916294333565, "alpha_c_p1d": -0.0074820190256772445, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.03828140889022169, "ret_signed_p1w": 0.03828140889022169, "alpha_spy_p1w": 0.03327873331581732, "alpha_c_p1w": 0.013296854545935988, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.1321777994214497, "ret_signed_p1m": 0.1321777994214497, "alpha_spy_p1m": 0.09466787282422229, "alpha_c_p1m": 0.02897091674672203, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.32908887558095556, "ret_signed_p3m": 0.32908887558095556, "alpha_spy_p3m": 0.24344496395912096, "alpha_c_p3m": 0.20105338428758768, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.7681754666681708, "ret_signed_p6m": 0.7681754666681708, "alpha_spy_p6m": 0.6251659632831934, "alpha_c_p6m": 0.36939734729499407, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.172, -0.194, -0.1705, -0.1408, -0.1497, -0.1273, -0.1199, -0.1256, -0.0647, -0.0568, -0.0948, -0.1239, -0.1521, -0.156, -0.1557, -0.1542, -0.2294, -0.2955, -0.3129, -0.3575, -0.2045, -0.2713, -0.2327, -0.2237, -0.2172, -0.2747, -0.2812, -0.2812, -0.2971, -0.2914, -0.2575, -0.2239, -0.206, -0.2082, -0.2109, -0.2003, -0.206, -0.1884, -0.1737, -0.1898, -0.1756, -0.1645, -0.1552, -0.1118, -0.0762, -0.0329, -0.0554, -0.0375, -0.0574, -0.0675, -0.0794, -0.0905, -0.0938, -0.0938, -0.0589, -0.0729, -0.0715, -0.0904, -0.0583, -0.0363, -0.0259, -0.0496, -0.0455, 0.0, 0.0124, -0.0048, -0.0265, -0.0458, 0.0383, 0.0441, 0.0416, 0.0416, 0.0535, 0.0645, 0.1372, 0.178, 0.1803, 0.1814, 0.1657, 0.1181, 0.1379, 0.1329, 0.1329, 0.1074, 0.1322, 0.137, 0.1843, 0.2028, 0.2013, 0.2783, 0.315, 0.3178, 0.2897, 0.2897, 0.271, 0.3033, 0.3318, 0.3675, 0.4266, 0.4577, 0.4747, 0.4484, 0.4105, 0.4522, 0.4319, 0.34, 0.4162, 0.4192, 0.4153, 0.4372, 0.515, 0.4865, 0.4582, 0.447, 0.3682, 0.3571, 0.3449, 0.3781, 0.342, 0.3688, 0.373, 0.3849, 0.336, 0.336, 0.3334, 0.3319, 0.3291, 0.2416, 0.2438, 0.28, 0.3106, 0.2788, 0.3026, 0.3239, 0.3182, 0.3075, 0.2973, 0.2929, 0.3127, 0.3218, 0.3216, 0.3248, 0.3099, 0.3256, 0.3291, 0.3473, 0.3943, 0.3527, 0.6735, 0.7375, 0.9351, 0.9132, 0.7654, 0.7779, 0.7916, 0.9601, 0.9269, 0.9147, 0.9762, 0.9554, 0.8913, 0.9304, 1.0777, 1.1332, 1.1195, 1.1714, 1.0935, 1.104, 1.133, 1.0541, 1.1057, 0.9527, 0.9185, 1.0043, 0.9512, 1.1268, 1.037, 1.0275, 0.9758, 0.8918, 0.8364, 0.6924, 0.674, 0.7666, 0.6933, 0.76, 0.76, 0.787, 0.8053, 0.7682]}
{"unit_id": "orig:1932072542451495091", "ticker_idx": 1, "ts": "2025-06-09T13:49:24+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If Marvell can lift its 2028 custom-ASIC revenue from the Street's $4.5 billion estimate to $8–9 billion, there is clear upside for the share price", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS bullish thesis on AMD (MI400 ramp/new customers), MRVL (ASIC TAM upside to $8-9B), and MU (HBM scarcity supports above-consensus EPS in FY26/27).", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Brief: AMD / Marvell / Micron\nUBS (T. Arcuri, 25/06/08)\n\nAs AI compute demand accelerates, vendors such as AMD, Marvell, and Micron will face significant growth opportunities amid tight supply and demand for GPUs, custom ASICs, and HBM.\n\n• AMD: AI new-product push and MI400 on the horizon\n\nAMD will spotlight the MI355X at this week’s AI launch, slated for deployment in 3Q25, offering performance comparable to NVIDIA’s Blackwell B200/B300 at more competitive pricing. The key point is that if AMZN joins MSFT and META as the third-largest customer, shipment volumes will expand further. More importantly, investors are watching the MI400 rack-scale solution’s release timing and performance jump: a dual Active Interposer Die architecture, each package containing 288 GB of HBM4, with an in-house UA Link interconnect, expected to enter mass production as early as end-2026; if realized smoothly, it will deliver breakthrough market share and revenue growth for AMD.\n\n• Marvell: Custom ASIC outlook and TAM expansion\n\nMarvell had originally planned an AI Investor Day in June, later changed to an online seminar on June 17, and is expected to raise its 2028 TAM forecasts for accelerated-compute ASICs and the data-center market (2028 TAM: last MRVL forecast was $43 billion for ASICs and $75 billion for data centers). The company will reaffirm its target share above 20% and showcase progress on multiple SKUs in collaboration with AMZN on custom network accelerators. If Marvell can lift its 2028 custom-ASIC revenue from the Street’s $4.5 billion estimate to $8–9 billion, there is clear upside for the share price.\n\n• Micron: HBM supply–demand stress scenario simulation\n\nMicron simulated new AI compute deployment scenarios for C 2026 and found that if GPU shipments exceed the baseline by 20%, overall HBM supply–demand will be in shortage; if driven solely by XPU growth, a 75% increment is required to cause shortage; and under a 70% GPU / 30% XPU mix, roughly a 35% increment will tighten the HBM market. This outcome underscores HBM’s scarcity over the next two years, supporting Micron’s FY 2026 EPS above $4 and FY 2027 EPS above $5, and providing a rationale for valuation premium.\n\n• Industry consensus and key points\n\nUnder AI-driven acceleration of hardware refresh cycles for GPUs and custom ASICs, demand for high-bandwidth memory (HBM) is rising in tandem. Advanced nodes such as TSMC’s 2 nm (N2) will also scale up due to flagship SoC and server-chip migrations. Investors should monitor: the release timing and customer-base expansion for AMD’s MI400 and Marvell’s custom ASICs; HBM capacity ramps and price trends; and foundry N2 capacity expansion along with mature-node price resilience. These factors will determine semiconductor vendors’ market positions and profitability in the AI and high-performance computing wave.\n\n$AMD $MRVL $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011426124585564024, "ret_signed_m1d": -0.011426124585564024, "alpha_spy_m1d": -0.01052565327680155, "alpha_c_m1d": 0.005035501448246493, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004339068266524748, "ret_signed_p1d": -0.004339068266524748, "alpha_spy_p1d": -0.010008732935415487, "alpha_c_p1d": -0.02422554359052609, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.018513070259247533, "ret_signed_p1w": 0.018513070259247533, "alpha_spy_p1w": 0.013510394684843163, "alpha_c_p1w": -0.0064714840850381705, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.040642030740599244, "ret_signed_p1m": 0.040642030740599244, "alpha_spy_p1m": 0.003132104143371839, "alpha_c_p1m": -0.06256485193412842, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": -0.07213678864026263, "ret_signed_p3m": -0.07213678864026263, "alpha_spy_p3m": -0.15778070026209723, "alpha_c_p3m": -0.20017227993363051, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 0.3454958765242089, "ret_signed_p6m": 0.3454958765242089, "alpha_spy_p6m": 0.20248637313923146, "alpha_c_p6m": -0.053282242848967876, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [0.0091, -0.0066, -0.0069, 0.0171, -0.0136, 0.0085, 0.0184, 0.0169, 0.0512, 0.0354, -0.0355, -0.063, -0.1037, -0.1105, -0.0946, -0.0865, -0.1962, -0.2859, -0.2632, -0.2772, -0.1193, -0.2362, -0.2278, -0.2441, -0.2288, -0.2489, -0.2522, -0.2522, -0.2858, -0.2679, -0.2223, -0.1708, -0.1478, -0.151, -0.1513, -0.1558, -0.1185, -0.0985, -0.1036, -0.1146, -0.1856, -0.1665, -0.1373, -0.0671, -0.0529, -0.0469, -0.057, -0.0778, -0.0952, -0.1117, -0.1306, -0.1054, -0.1222, -0.1222, -0.0769, -0.0658, -0.0782, -0.1294, -0.1109, -0.0981, -0.0411, -0.0576, -0.0114, 0.0, -0.0043, -0.013, 0.0072, -0.0282, 0.0185, 0.0123, 0.084, 0.084, 0.0632, 0.0237, 0.0878, 0.0982, 0.1566, 0.116, 0.1195, 0.1027, 0.0739, 0.0874, 0.0874, 0.0349, 0.0406, 0.0451, 0.061, 0.0525, 0.0496, 0.0482, 0.0256, 0.0424, 0.0806, 0.0576, 0.0421, 0.0606, 0.0717, 0.0742, 0.0988, 0.105, 0.1832, 0.1634, 0.0777, 0.1078, 0.1092, 0.0903, 0.0979, 0.1195, 0.1186, 0.1263, 0.1482, 0.1441, 0.1029, 0.1108, 0.0432, 0.0309, 0.0308, 0.0567, 0.056, 0.0749, 0.0826, 0.1179, -0.0899, -0.0899, -0.0649, -0.098, -0.0721, -0.0833, -0.0446, -0.0325, -0.0287, -0.0361, -0.0251, -0.0239, -0.0032, 0.0275, 0.0745, 0.0749, 0.0933, 0.0801, 0.1593, 0.2132, 0.2039, 0.1926, 0.2169, 0.2143, 0.2478, 0.2481, 0.2871, 0.2589, 0.339, 0.3126, 0.24, 0.2948, 0.2489, 0.2876, 0.278, 0.2739, 0.2434, 0.2205, 0.174, 0.1989, 0.2186, 0.2849, 0.2813, 0.3058, 0.2829, 0.3578, 0.309, 0.2687, 0.3456, 0.3519, 0.317, 0.3504, 0.2939, 0.2939, 0.2677, 0.2522, 0.2088, 0.1397, 0.1779, 0.1107, 0.1219, 0.2137, 0.2085, 0.2706, 0.2706, 0.2949, 0.3196, 0.3455]}
{"unit_id": "orig:1932072542451495091", "ticker_idx": 2, "ts": "2025-06-09T13:49:24+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "supporting Micron's FY 2026 EPS above $4 and FY 2027 EPS above $5, and providing a rationale for valuation premium", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS bullish thesis on AMD (MI400 ramp/new customers), MRVL (ASIC TAM upside to $8-9B), and MU (HBM scarcity supports above-consensus EPS in FY26/27).", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Brief: AMD / Marvell / Micron\nUBS (T. Arcuri, 25/06/08)\n\nAs AI compute demand accelerates, vendors such as AMD, Marvell, and Micron will face significant growth opportunities amid tight supply and demand for GPUs, custom ASICs, and HBM.\n\n• AMD: AI new-product push and MI400 on the horizon\n\nAMD will spotlight the MI355X at this week’s AI launch, slated for deployment in 3Q25, offering performance comparable to NVIDIA’s Blackwell B200/B300 at more competitive pricing. The key point is that if AMZN joins MSFT and META as the third-largest customer, shipment volumes will expand further. More importantly, investors are watching the MI400 rack-scale solution’s release timing and performance jump: a dual Active Interposer Die architecture, each package containing 288 GB of HBM4, with an in-house UA Link interconnect, expected to enter mass production as early as end-2026; if realized smoothly, it will deliver breakthrough market share and revenue growth for AMD.\n\n• Marvell: Custom ASIC outlook and TAM expansion\n\nMarvell had originally planned an AI Investor Day in June, later changed to an online seminar on June 17, and is expected to raise its 2028 TAM forecasts for accelerated-compute ASICs and the data-center market (2028 TAM: last MRVL forecast was $43 billion for ASICs and $75 billion for data centers). The company will reaffirm its target share above 20% and showcase progress on multiple SKUs in collaboration with AMZN on custom network accelerators. If Marvell can lift its 2028 custom-ASIC revenue from the Street’s $4.5 billion estimate to $8–9 billion, there is clear upside for the share price.\n\n• Micron: HBM supply–demand stress scenario simulation\n\nMicron simulated new AI compute deployment scenarios for C 2026 and found that if GPU shipments exceed the baseline by 20%, overall HBM supply–demand will be in shortage; if driven solely by XPU growth, a 75% increment is required to cause shortage; and under a 70% GPU / 30% XPU mix, roughly a 35% increment will tighten the HBM market. This outcome underscores HBM’s scarcity over the next two years, supporting Micron’s FY 2026 EPS above $4 and FY 2027 EPS above $5, and providing a rationale for valuation premium.\n\n• Industry consensus and key points\n\nUnder AI-driven acceleration of hardware refresh cycles for GPUs and custom ASICs, demand for high-bandwidth memory (HBM) is rising in tandem. Advanced nodes such as TSMC’s 2 nm (N2) will also scale up due to flagship SoC and server-chip migrations. Investors should monitor: the release timing and customer-base expansion for AMD’s MI400 and Marvell’s custom ASICs; HBM capacity ramps and price trends; and foundry N2 capacity expansion along with mature-node price resilience. These factors will determine semiconductor vendors’ market positions and profitability in the AI and high-performance computing wave.\n\n$AMD $MRVL $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02154130161944212, "ret_signed_m1d": -0.02154130161944212, "alpha_spy_m1d": -0.020640830310679648, "alpha_c_m1d": -0.0050796755856316045, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.016461626033810517, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028751673592638083, "ret_signed_p1d": 0.028751673592638083, "alpha_spy_p1d": 0.023082008923747344, "alpha_c_p1d": 0.008865198268636743, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01988647532400134, "ret_p1w": 0.08012621149358345, "ret_signed_p1w": 0.08012621149358345, "alpha_spy_p1w": 0.07512353591917909, "alpha_c_p1w": 0.05514165714929775, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.024984554344285703, "ret_p1m": 0.12246155699809869, "ret_signed_p1m": 0.12246155699809869, "alpha_spy_p1m": 0.08495163040087128, "alpha_c_p1m": 0.019254674323371024, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.10320688267472766, "ret_p3m": 0.1205670473745275, "ret_signed_p3m": 0.1205670473745275, "alpha_spy_p3m": 0.0349231357526929, "alpha_c_p3m": -0.007468443918840384, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.1280354912933679, "ret_p6m": 1.161924774905926, "ret_signed_p6m": 1.161924774905926, "alpha_spy_p6m": 1.0189152715209486, "alpha_c_p6m": 0.7631466555327493, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.39877811937317675, "price_path": [-0.1391, -0.146, -0.0928, -0.0719, -0.0844, -0.0813, -0.0729, -0.1474, -0.1274, -0.1523, -0.1707, -0.1794, -0.2039, -0.2169, -0.2005, -0.2014, -0.33, -0.4167, -0.3838, -0.4093, -0.2982, -0.3686, -0.3731, -0.3599, -0.3597, -0.3751, -0.3799, -0.3799, -0.3985, -0.3672, -0.3427, -0.3022, -0.2809, -0.2919, -0.3071, -0.3064, -0.2991, -0.2725, -0.2752, -0.2744, -0.2553, -0.2325, -0.2261, -0.1682, -0.1264, -0.1409, -0.1397, -0.1167, -0.1109, -0.1158, -0.1362, -0.1453, -0.1584, -0.1584, -0.1313, -0.1331, -0.1275, -0.1486, -0.1151, -0.0784, -0.0694, -0.042, -0.0215, 0.0, 0.0288, 0.0458, 0.0471, 0.0419, 0.0801, 0.0846, 0.098, 0.098, 0.114, 0.1003, 0.1529, 0.1469, 0.1356, 0.1245, 0.1109, 0.0896, 0.0973, 0.1022, 0.1022, 0.0819, 0.1225, 0.1028, 0.1106, 0.1235, 0.07, 0.0836, 0.0504, 0.0218, 0.032, 0.0215, -0.0147, -0.0092, 0.008, 0.0037, 0.0036, 0.0101, 0.0351, -0.0154, -0.0538, -0.0277, -0.0161, -0.0186, 0.0092, 0.0726, 0.1161, 0.1525, 0.1211, 0.1303, 0.0904, 0.1146, 0.1011, 0.0574, 0.0446, 0.0617, 0.0503, 0.051, 0.0623, 0.1006, 0.0737, 0.0737, 0.0689, 0.071, 0.1206, 0.1852, 0.186, 0.2201, 0.263, 0.3584, 0.4185, 0.4233, 0.4328, 0.4434, 0.5237, 0.4681, 0.4851, 0.5013, 0.4589, 0.4149, 0.4188, 0.4786, 0.5095, 0.6433, 0.6577, 0.6956, 0.7238, 0.6763, 0.7742, 0.7362, 0.6393, 0.7402, 0.6886, 0.7327, 0.8283, 0.8269, 0.8666, 0.8261, 0.7916, 0.866, 0.9771, 0.9869, 1.0032, 1.0458, 1.0222, 1.02, 1.1187, 0.9682, 1.144, 1.1515, 1.1478, 1.2866, 1.1765, 1.2108, 1.139, 1.2282, 1.1841, 1.0627, 1.0394, 0.8178, 0.872, 1.0215, 1.0269, 1.0786, 1.0786, 1.1348, 1.1707, 1.1619]}
{"unit_id": "orig:1940902779570241911", "ticker_idx": 0, "ts": "2025-07-03T22:37:37+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC and MediaTek are considered the best beneficiaries in the Taiwanese semiconductor sector", "tweet_type": "news_reaction", "timeline": "years", "summary": "Citi names TSMC and MediaTek top long-term beneficiaries of AI ASIC/system integration trend; downgrades GUC to Sell; keeps Alchip Neutral; Broadcom leads ASIC market.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GPU vs. AI ASIC: Intensifying Differentiation in AI Systems, TSMC and MediaTek Have the Greatest Long-Term Potential\n\nCitigroup (L. Chen, 25/07/02)\n\nCitigroup believes that GPUs will continue to dominate the AI chip market, but as AI systems become increasingly complex, ASICs show significant growth potential. Long-term value growth will be driven by performance improvements and higher Average Selling Prices (ASPs). Companies with strong system integration capabilities will be the ultimate winners.\n\nAI System Design Needs to Focus on Full-Stack Efficiency, ASIC Success Beyond the Chip Itself\n\nCitigroup points out that CSPs (Cloud Service Providers) are accelerating the development of in-house ASICs to achieve differentiation and cost advantages. However, the real challenge lies in system-level design, including efficient interconnects and packaging, a competitive CUDA software stack, and low-latency, low-power system integration. Manufacturers with end-to-end system capabilities will stand out, and TSMC and MediaTek are considered the best beneficiaries in the Taiwanese semiconductor sector.\n\nAI Chip Volume Growth Slows but Value Growth Accelerates, ASIC Still Has High CAGR Potential\n\nStarting from the second half of 2026, AI GPU/accelerators will migrate from N5/N4 processes to N3, driving significant increases in single-chip ASP and performance. Citigroup expects GPU shipment growth of approximately 10% in 2026, and AI ASIC shipments of 4.5 million units (slightly lower than market expectations of 5 million units). However, driven by ASP increases, the overall AI chip value growth CAGR (Compound Annual Growth Rate) can still exceed 40%.\n\nBroadcom Leads in ASIC Market Share, TPU v8 and Meta's In-House Chips are Future Catalysts\n\nBroadcom maintains its lead in the ASIC market thanks to its TPU project. TPU v7 shipments are expected to remain stable in 2026, with TPU v8 beginning to ramp up in late 2026 and continuing to grow in 2027. Meta's self-developed AI accelerator is expected to enter mass production from late 2025, with shipments potentially reaching 0.5 million to 1 million units in 2026, depending on its supply chain readiness.\n\nAWS Trainium Project Faces Potential Delay, Trainium 3 May Be Pushed to H2 2026\n\nDue to the design complexity of N3 and the difficulty of SerDes/IO implementation, Trainium 3 may be delayed longer than expected. Citigroup believes that Marvell and Alchip will be responsible for back-end services and collaboration with TSMC, respectively, while Trainium 2's lifecycle will be extended to 2026. Trainium 2+3 shipments are expected to be flat year-on-year, but due to the N3 migration, Trainium 3 will bring higher ASPs and larger die sizes.\n\nGUC Downgraded, Alchip Maintained Neutral, MediaTek More Advantageous Due to Advanced Process and SerDes Capability\n\nIn the design service segment, Citigroup has downgraded GUC's rating to \"Sell/Neutral,\" citing slow progress in customer mass production and system ecosystem development. Alchip maintains a neutral rating, awaiting clear progress on customer chip redesigns. MediaTek, with its design capabilities in advanced nodes and SerDes, is expected to enter a period of AI ASIC revenue realization in late 2026 and see further growth in 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004587151290653346, "ret_signed_m1d": -0.004587151290653346, "alpha_spy_m1d": 0.003232604153493579, "alpha_c_m1d": 0.0037691794015815017, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004587151290653346, "ret_signed_p1d": -0.004587151290653346, "alpha_spy_p1d": -0.004587151290653346, "alpha_c_p1d": -0.004587151290653346, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.009174415763257837, "ret_signed_p1w": 0.009174415763257837, "alpha_spy_p1w": 0.00840685052291934, "alpha_c_p1w": -0.0045762922905956405, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.06422023125109866, "ret_signed_p1m": 0.06422023125109866, "alpha_spy_p1m": 0.07000925593435492, "alpha_c_p1m": 0.06305669153445836, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.2020367997531729, "ret_signed_p3m": 0.2020367997531729, "alpha_spy_p3m": 0.13377487436527824, "alpha_c_p3m": 0.051342324127495065, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.3954987085373485, "ret_signed_p6m": 0.3954987085373485, "alpha_spy_p6m": 0.28527189094177663, "alpha_c_p6m": 0.1015194976892726, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.2253, -0.2545, -0.2829, -0.2116, -0.1878, -0.2098, -0.1988, -0.2189, -0.2262, -0.2235, -0.2372, -0.2545, -0.2025, -0.2107, -0.1888, -0.1805, -0.176, -0.1705, -0.1705, -0.1321, -0.1431, -0.1595, -0.1522, -0.1614, -0.133, -0.1257, -0.1148, -0.0874, -0.0928, -0.0883, -0.1011, -0.1065, -0.0956, -0.1029, -0.1011, -0.1084, -0.1184, -0.1166, -0.1166, -0.1166, -0.1358, -0.1321, -0.0956, -0.0883, -0.091, -0.0819, -0.0453, -0.0271, -0.0413, -0.055, -0.0596, -0.0413, -0.0321, -0.0505, -0.0321, -0.0642, -0.0367, -0.0183, -0.0138, -0.0092, -0.0275, -0.0046, -0.0046, 0.0, -0.0046, -0.0092, -0.0092, 0.0, 0.0092, 0.0092, 0.0046, 0.0183, 0.0367, 0.0367, 0.0596, 0.055, 0.0367, 0.0505, 0.0505, 0.0505, 0.0505, 0.0413, 0.0596, 0.0642, 0.0642, 0.0413, 0.055, 0.0321, 0.0826, 0.078, 0.0826, 0.0826, 0.1009, 0.078, 0.0826, 0.0826, 0.0872, 0.0413, 0.055, 0.0413, 0.0734, 0.078, 0.0917, 0.0642, 0.0642, 0.0688, 0.0642, 0.0642, 0.0642, 0.0826, 0.0826, 0.1009, 0.1239, 0.1376, 0.156, 0.1514, 0.179, 0.1652, 0.1836, 0.1652, 0.1928, 0.2343, 0.2343, 0.2159, 0.1974, 0.1974, 0.202, 0.2205, 0.2573, 0.2895, 0.2895, 0.3218, 0.3034, 0.3264, 0.3264, 0.3034, 0.3126, 0.3494, 0.3678, 0.3356, 0.3632, 0.3632, 0.3448, 0.3356, 0.3356, 0.3632, 0.3586, 0.3863, 0.3863, 0.3817, 0.3909, 0.3863, 0.3448, 0.3494, 0.3448, 0.3586, 0.3494, 0.3586, 0.3448, 0.3172, 0.331, 0.2941, 0.2849, 0.3402, 0.2757, 0.2665, 0.3034, 0.3264, 0.3218, 0.3264, 0.2988, 0.3172, 0.3356, 0.331, 0.3448, 0.377, 0.3632, 0.3863, 0.3585, 0.3678, 0.34, 0.3262, 0.3216, 0.3216, 0.3216, 0.3539, 0.377, 0.3816, 0.3816, 0.3955]}
{"unit_id": "orig:1940902779570241911", "ticker_idx": 1, "ts": "2025-07-03T22:37:37+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC and MediaTek are considered the best beneficiaries in the Taiwanese semiconductor sector", "tweet_type": "news_reaction", "timeline": "years", "summary": "Citi names TSMC and MediaTek top long-term beneficiaries of AI ASIC/system integration trend; downgrades GUC to Sell; keeps Alchip Neutral; Broadcom leads ASIC market.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GPU vs. AI ASIC: Intensifying Differentiation in AI Systems, TSMC and MediaTek Have the Greatest Long-Term Potential\n\nCitigroup (L. Chen, 25/07/02)\n\nCitigroup believes that GPUs will continue to dominate the AI chip market, but as AI systems become increasingly complex, ASICs show significant growth potential. Long-term value growth will be driven by performance improvements and higher Average Selling Prices (ASPs). Companies with strong system integration capabilities will be the ultimate winners.\n\nAI System Design Needs to Focus on Full-Stack Efficiency, ASIC Success Beyond the Chip Itself\n\nCitigroup points out that CSPs (Cloud Service Providers) are accelerating the development of in-house ASICs to achieve differentiation and cost advantages. However, the real challenge lies in system-level design, including efficient interconnects and packaging, a competitive CUDA software stack, and low-latency, low-power system integration. Manufacturers with end-to-end system capabilities will stand out, and TSMC and MediaTek are considered the best beneficiaries in the Taiwanese semiconductor sector.\n\nAI Chip Volume Growth Slows but Value Growth Accelerates, ASIC Still Has High CAGR Potential\n\nStarting from the second half of 2026, AI GPU/accelerators will migrate from N5/N4 processes to N3, driving significant increases in single-chip ASP and performance. Citigroup expects GPU shipment growth of approximately 10% in 2026, and AI ASIC shipments of 4.5 million units (slightly lower than market expectations of 5 million units). However, driven by ASP increases, the overall AI chip value growth CAGR (Compound Annual Growth Rate) can still exceed 40%.\n\nBroadcom Leads in ASIC Market Share, TPU v8 and Meta's In-House Chips are Future Catalysts\n\nBroadcom maintains its lead in the ASIC market thanks to its TPU project. TPU v7 shipments are expected to remain stable in 2026, with TPU v8 beginning to ramp up in late 2026 and continuing to grow in 2027. Meta's self-developed AI accelerator is expected to enter mass production from late 2025, with shipments potentially reaching 0.5 million to 1 million units in 2026, depending on its supply chain readiness.\n\nAWS Trainium Project Faces Potential Delay, Trainium 3 May Be Pushed to H2 2026\n\nDue to the design complexity of N3 and the difficulty of SerDes/IO implementation, Trainium 3 may be delayed longer than expected. Citigroup believes that Marvell and Alchip will be responsible for back-end services and collaboration with TSMC, respectively, while Trainium 2's lifecycle will be extended to 2026. Trainium 2+3 shipments are expected to be flat year-on-year, but due to the N3 migration, Trainium 3 will bring higher ASPs and larger die sizes.\n\nGUC Downgraded, Alchip Maintained Neutral, MediaTek More Advantageous Due to Advanced Process and SerDes Capability\n\nIn the design service segment, Citigroup has downgraded GUC's rating to \"Sell/Neutral,\" citing slow progress in customer mass production and system ecosystem development. Alchip maintains a neutral rating, awaiting clear progress on customer chip redesigns. MediaTek, with its design capabilities in advanced nodes and SerDes, is expected to enter a period of AI ASIC revenue realization in late 2026 and see further growth in 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -1.9596321654269389e-07, "ret_signed_m1d": -1.9596321654269389e-07, "alpha_spy_m1d": 0.007819559480930383, "alpha_c_m1d": 0.008356134729018305, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015747996006661547, "ret_signed_p1d": 0.015747996006661547, "alpha_spy_p1d": 0.015747996006661547, "alpha_c_p1d": 0.015747996006661547, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.1023621700065167, "ret_signed_p1w": 0.1023621700065167, "alpha_spy_p1w": 0.10159460476617821, "alpha_c_p1w": 0.08861146195266323, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.07874007801491612, "ret_signed_p1m": 0.07874007801491612, "alpha_spy_p1m": 0.08452910269817238, "alpha_c_p1m": 0.07757653829827582, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.035432991014988646, "ret_signed_p3m": 0.035432991014988646, "alpha_spy_p3m": -0.032828934372906016, "alpha_c_p3m": -0.11526148461068919, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.09055117300152049, "ret_signed_p6m": 0.09055117300152049, "alpha_spy_p6m": -0.019675644594051356, "alpha_c_p6m": -0.2034280378465554, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.0, -0.027, -0.0849, 0.0039, 0.0695, 0.0811, 0.0695, 0.0541, 0.0347, 0.0541, 0.0386, 0.0039, 0.0579, 0.027, 0.0656, 0.0425, 0.0579, 0.0425, 0.0425, 0.0039, -0.0, -0.0116, -0.0154, -0.0039, 0.0232, 0.0154, 0.0463, 0.0502, 0.0579, 0.0541, 0.0116, 0.0154, 0.0347, 0.0232, 0.0154, -0.0116, -0.0116, -0.0116, -0.027, -0.027, -0.027, -0.0154, -0.0154, -0.0386, -0.0116, -0.0077, 0.0309, 0.0154, -0.0077, -0.0347, -0.0309, -0.0154, -0.0077, -0.0232, -0.0347, -0.0309, -0.0193, -0.0, -0.0039, -0.0077, -0.0347, -0.0154, -0.0, 0.0, 0.0157, 0.0079, 0.0, 0.0669, 0.1024, 0.1181, 0.0906, 0.0984, 0.1102, 0.0945, 0.1102, 0.122, 0.126, 0.126, 0.126, 0.126, 0.1024, 0.0709, 0.0866, 0.0787, 0.0787, 0.0472, 0.0551, 0.0394, 0.0669, 0.063, 0.0669, 0.0709, 0.0787, 0.0945, 0.0787, 0.1102, 0.0945, 0.0669, 0.0748, 0.0748, 0.1102, 0.1024, 0.1024, 0.0906, 0.0787, 0.0709, 0.0709, 0.0984, 0.0906, 0.1299, 0.126, 0.189, 0.1732, 0.1654, 0.1693, 0.1693, 0.2087, 0.189, 0.189, 0.1339, 0.1063, 0.0827, 0.0472, 0.0591, 0.0315, 0.0315, 0.0354, 0.0118, 0.0079, 0.0394, 0.0394, 0.0433, 0.0512, 0.0591, 0.0591, 0.0354, 0.0276, 0.0354, 0.0472, 0.0472, 0.0551, 0.0551, 0.0472, 0.0197, 0.0197, 0.0512, 0.0354, 0.0236, 0.0315, 0.0315, 0.0236, 0.0354, 0.0472, 0.0157, -0.0079, -0.0157, -0.0236, -0.0197, -0.0197, -0.0315, -0.0315, -0.0787, -0.0866, -0.0669, -0.0984, -0.0945, -0.0669, 0.0236, 0.0551, 0.0984, 0.1378, 0.1142, 0.1024, 0.1063, 0.122, 0.1339, 0.1181, 0.1496, 0.0984, 0.1063, 0.1181, 0.1181, 0.122, 0.1181, 0.1102, 0.1024, 0.0945, 0.0866, 0.0866, 0.0906]}
{"unit_id": "orig:1940902779570241911", "ticker_idx": 2, "ts": "2025-07-03T22:37:37+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom maintains its lead in the ASIC market thanks to its TPU project", "tweet_type": "news_reaction", "timeline": "years", "summary": "Citi names TSMC and MediaTek top long-term beneficiaries of AI ASIC/system integration trend; downgrades GUC to Sell; keeps Alchip Neutral; Broadcom leads ASIC market.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GPU vs. AI ASIC: Intensifying Differentiation in AI Systems, TSMC and MediaTek Have the Greatest Long-Term Potential\n\nCitigroup (L. Chen, 25/07/02)\n\nCitigroup believes that GPUs will continue to dominate the AI chip market, but as AI systems become increasingly complex, ASICs show significant growth potential. Long-term value growth will be driven by performance improvements and higher Average Selling Prices (ASPs). Companies with strong system integration capabilities will be the ultimate winners.\n\nAI System Design Needs to Focus on Full-Stack Efficiency, ASIC Success Beyond the Chip Itself\n\nCitigroup points out that CSPs (Cloud Service Providers) are accelerating the development of in-house ASICs to achieve differentiation and cost advantages. However, the real challenge lies in system-level design, including efficient interconnects and packaging, a competitive CUDA software stack, and low-latency, low-power system integration. Manufacturers with end-to-end system capabilities will stand out, and TSMC and MediaTek are considered the best beneficiaries in the Taiwanese semiconductor sector.\n\nAI Chip Volume Growth Slows but Value Growth Accelerates, ASIC Still Has High CAGR Potential\n\nStarting from the second half of 2026, AI GPU/accelerators will migrate from N5/N4 processes to N3, driving significant increases in single-chip ASP and performance. Citigroup expects GPU shipment growth of approximately 10% in 2026, and AI ASIC shipments of 4.5 million units (slightly lower than market expectations of 5 million units). However, driven by ASP increases, the overall AI chip value growth CAGR (Compound Annual Growth Rate) can still exceed 40%.\n\nBroadcom Leads in ASIC Market Share, TPU v8 and Meta's In-House Chips are Future Catalysts\n\nBroadcom maintains its lead in the ASIC market thanks to its TPU project. TPU v7 shipments are expected to remain stable in 2026, with TPU v8 beginning to ramp up in late 2026 and continuing to grow in 2027. Meta's self-developed AI accelerator is expected to enter mass production from late 2025, with shipments potentially reaching 0.5 million to 1 million units in 2026, depending on its supply chain readiness.\n\nAWS Trainium Project Faces Potential Delay, Trainium 3 May Be Pushed to H2 2026\n\nDue to the design complexity of N3 and the difficulty of SerDes/IO implementation, Trainium 3 may be delayed longer than expected. Citigroup believes that Marvell and Alchip will be responsible for back-end services and collaboration with TSMC, respectively, while Trainium 2's lifecycle will be extended to 2026. Trainium 2+3 shipments are expected to be flat year-on-year, but due to the N3 migration, Trainium 3 will bring higher ASPs and larger die sizes.\n\nGUC Downgraded, Alchip Maintained Neutral, MediaTek More Advantageous Due to Advanced Process and SerDes Capability\n\nIn the design service segment, Citigroup has downgraded GUC's rating to \"Sell/Neutral,\" citing slow progress in customer mass production and system ecosystem development. Alchip maintains a neutral rating, awaiting clear progress on customer chip redesigns. MediaTek, with its design capabilities in advanced nodes and SerDes, is expected to enter a period of AI ASIC revenue realization in late 2026 and see further growth in 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019187508093365446, "ret_signed_m1d": -0.019187508093365446, "alpha_spy_m1d": -0.011367752649218521, "alpha_c_m1d": -0.010831177401130598, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.0007993819094753718, "ret_signed_p1w": 0.0007993819094753718, "alpha_spy_p1w": 3.1816669136874864e-05, "alpha_c_p1w": -0.012951326144378106, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.04891342840033164, "ret_signed_p1m": 0.04891342840033164, "alpha_spy_p1m": 0.0547024530835879, "alpha_c_p1m": 0.047749888683691344, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.20094210398166257, "ret_signed_p3m": 0.20094210398166257, "alpha_spy_p3m": 0.1326801785937679, "alpha_c_p3m": 0.05024762835598473, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.28428026126749817, "ret_signed_p6m": 0.28428026126749817, "alpha_spy_p6m": 0.17405344367192632, "alpha_c_p6m": -0.00969894958057771, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.4412, -0.4343, -0.3287, -0.3753, -0.3404, -0.3534, -0.3512, -0.367, -0.3801, -0.3801, -0.3974, -0.3852, -0.3586, -0.3179, -0.3028, -0.3022, -0.3069, -0.3022, -0.2846, -0.2617, -0.2723, -0.2746, -0.2575, -0.2467, -0.2452, -0.1967, -0.1574, -0.1585, -0.1566, -0.1712, -0.1639, -0.1601, -0.1671, -0.1642, -0.1708, -0.1708, -0.1457, -0.132, -0.1227, -0.1224, -0.0983, -0.0688, -0.0535, -0.0576, -0.1048, -0.1144, -0.1131, -0.0831, -0.0716, -0.0984, -0.086, -0.0959, -0.0891, -0.0891, -0.0915, -0.0778, -0.0415, -0.0383, -0.0182, -0.0212, 0.0017, -0.0379, -0.0192, 0.0, 0.0, -0.0036, -0.0123, 0.0099, 0.0008, -0.0029, 0.0015, 0.0209, 0.0205, 0.041, 0.0297, 0.0474, 0.0124, 0.0309, 0.0492, 0.0545, 0.0695, 0.0808, 0.0997, 0.0673, 0.0489, 0.0819, 0.0645, 0.0963, 0.1039, 0.1083, 0.1044, 0.1368, 0.1232, 0.131, 0.1132, 0.1111, 0.0717, 0.0581, 0.0524, 0.0684, 0.0692, 0.083, 0.0911, 0.1216, 0.0807, 0.0807, 0.0838, 0.0989, 0.1124, 0.217, 0.2561, 0.2235, 0.343, 0.3069, 0.3078, 0.3231, 0.3082, 0.258, 0.255, 0.2535, 0.2333, 0.2338, 0.2352, 0.2235, 0.2178, 0.1936, 0.2009, 0.2136, 0.231, 0.2317, 0.2213, 0.2246, 0.2577, 0.2559, 0.1817, 0.2985, 0.2527, 0.2789, 0.2892, 0.2716, 0.2713, 0.2474, 0.2388, 0.2533, 0.2891, 0.3179, 0.3577, 0.405, 0.3704, 0.3455, 0.3198, 0.2811, 0.3068, 0.2944, 0.272, 0.3046, 0.2812, 0.2931, 0.2376, 0.2466, 0.2473, 0.2395, 0.2902, 0.2625, 0.2384, 0.3759, 0.4016, 0.4472, 0.4472, 0.4669, 0.4054, 0.389, 0.3855, 0.387, 0.4206, 0.4601, 0.479, 0.5033, 0.4793, 0.3102, 0.237, 0.2424, 0.1868, 0.2008, 0.239, 0.2453, 0.274, 0.2773, 0.2773, 0.2843]}
{"unit_id": "orig:1940902779570241911", "ticker_idx": 4, "ts": "2025-07-03T22:37:37+00:00", "symbol": "Alchip", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Alchip maintains a neutral rating, awaiting clear progress on customer chip redesigns", "tweet_type": "news_reaction", "timeline": "years", "summary": "Citi names TSMC and MediaTek top long-term beneficiaries of AI ASIC/system integration trend; downgrades GUC to Sell; keeps Alchip Neutral; Broadcom leads ASIC market.", "resolved_tickers": ["3661.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Alchip Technologies 3661.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GPU vs. AI ASIC: Intensifying Differentiation in AI Systems, TSMC and MediaTek Have the Greatest Long-Term Potential\n\nCitigroup (L. Chen, 25/07/02)\n\nCitigroup believes that GPUs will continue to dominate the AI chip market, but as AI systems become increasingly complex, ASICs show significant growth potential. Long-term value growth will be driven by performance improvements and higher Average Selling Prices (ASPs). Companies with strong system integration capabilities will be the ultimate winners.\n\nAI System Design Needs to Focus on Full-Stack Efficiency, ASIC Success Beyond the Chip Itself\n\nCitigroup points out that CSPs (Cloud Service Providers) are accelerating the development of in-house ASICs to achieve differentiation and cost advantages. However, the real challenge lies in system-level design, including efficient interconnects and packaging, a competitive CUDA software stack, and low-latency, low-power system integration. Manufacturers with end-to-end system capabilities will stand out, and TSMC and MediaTek are considered the best beneficiaries in the Taiwanese semiconductor sector.\n\nAI Chip Volume Growth Slows but Value Growth Accelerates, ASIC Still Has High CAGR Potential\n\nStarting from the second half of 2026, AI GPU/accelerators will migrate from N5/N4 processes to N3, driving significant increases in single-chip ASP and performance. Citigroup expects GPU shipment growth of approximately 10% in 2026, and AI ASIC shipments of 4.5 million units (slightly lower than market expectations of 5 million units). However, driven by ASP increases, the overall AI chip value growth CAGR (Compound Annual Growth Rate) can still exceed 40%.\n\nBroadcom Leads in ASIC Market Share, TPU v8 and Meta's In-House Chips are Future Catalysts\n\nBroadcom maintains its lead in the ASIC market thanks to its TPU project. TPU v7 shipments are expected to remain stable in 2026, with TPU v8 beginning to ramp up in late 2026 and continuing to grow in 2027. Meta's self-developed AI accelerator is expected to enter mass production from late 2025, with shipments potentially reaching 0.5 million to 1 million units in 2026, depending on its supply chain readiness.\n\nAWS Trainium Project Faces Potential Delay, Trainium 3 May Be Pushed to H2 2026\n\nDue to the design complexity of N3 and the difficulty of SerDes/IO implementation, Trainium 3 may be delayed longer than expected. Citigroup believes that Marvell and Alchip will be responsible for back-end services and collaboration with TSMC, respectively, while Trainium 2's lifecycle will be extended to 2026. Trainium 2+3 shipments are expected to be flat year-on-year, but due to the N3 migration, Trainium 3 will bring higher ASPs and larger die sizes.\n\nGUC Downgraded, Alchip Maintained Neutral, MediaTek More Advantageous Due to Advanced Process and SerDes Capability\n\nIn the design service segment, Citigroup has downgraded GUC's rating to \"Sell/Neutral,\" citing slow progress in customer mass production and system ecosystem development. Alchip maintains a neutral rating, awaiting clear progress on customer chip redesigns. MediaTek, with its design capabilities in advanced nodes and SerDes, is expected to enter a period of AI ASIC revenue realization in late 2026 and see further growth in 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00472437253202318, "ret_signed_m1d": -0.00472437253202318, "alpha_spy_m1d": -0.012544127976170105, "alpha_c_m1d": -0.013080703224258028, "bench_spy_m1d": -0.007819755444146925, "bench_c_m1d": -0.008356330692234848, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00472437253202318, "ret_signed_p1d": -0.00472437253202318, "alpha_spy_p1d": -0.00472437253202318, "alpha_c_p1d": -0.00472437253202318, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.0897637767160151, "ret_signed_p1w": -0.0897637767160151, "alpha_spy_p1w": -0.0889962114756766, "alpha_c_p1w": -0.07601306866216162, "bench_spy_p1w": 0.0007675652403384969, "bench_c_p1w": 0.013750708053853478, "ret_p1m": 0.2236220334308563, "ret_signed_p1m": -0.2236220334308563, "alpha_spy_p1m": -0.22941105811411255, "alpha_c_p1m": -0.222458493714216, "bench_spy_p1m": -0.005789024683256261, "bench_c_p1m": 0.0011635397166402939, "ret_p3m": 0.10167620020199064, "ret_signed_p3m": -0.10167620020199064, "alpha_spy_p3m": -0.03341427481409598, "alpha_c_p3m": 0.049018275423687196, "bench_spy_p3m": 0.06826192538789466, "bench_c_p3m": 0.15069447562567784, "ret_p6m": 0.022190182871399733, "ret_signed_p6m": -0.022190182871399733, "alpha_spy_p6m": 0.08803663472417211, "alpha_c_p6m": 0.27178902797667615, "bench_spy_p6m": 0.11022681759557185, "bench_c_p6m": 0.2939792108480759, "price_path": [-0.2457, -0.3197, -0.3874, -0.3276, -0.3307, -0.2976, -0.2567, -0.3165, -0.3024, -0.3055, -0.337, -0.3717, -0.3433, -0.3354, -0.3197, -0.3102, -0.315, -0.3307, -0.3307, -0.2756, -0.2835, -0.2835, -0.2976, -0.2614, -0.2756, -0.2031, -0.1512, -0.1323, -0.1433, -0.1402, -0.1795, -0.0976, -0.0614, -0.0866, -0.0646, -0.1008, -0.0992, -0.1008, -0.115, -0.115, -0.1433, -0.1433, -0.1165, -0.126, -0.1197, -0.1417, -0.1323, -0.1055, -0.1244, -0.1701, -0.1701, -0.1134, -0.1165, -0.0961, -0.1181, -0.1087, -0.1276, -0.0929, -0.0929, -0.0551, -0.0252, 0.0031, 0.0047, 0.0, 0.0047, 0.0142, 0.0126, 0.0598, 0.0898, 0.0803, 0.0425, 0.0614, 0.1339, 0.1827, 0.2378, 0.1906, 0.1717, 0.2047, 0.252, 0.2236, 0.222, 0.2016, 0.1811, 0.2236, 0.2236, 0.1827, 0.1811, 0.1685, 0.1858, 0.2094, 0.2315, 0.2835, 0.2976, 0.3386, 0.3559, 0.3323, 0.2709, 0.1528, 0.1921, 0.1622, 0.2268, 0.2535, 0.285, 0.2598, 0.2772, 0.2331, 0.2225, 0.2654, 0.2241, 0.2622, 0.2638, 0.2448, 0.2527, 0.2479, 0.205, 0.1843, 0.17, 0.1955, 0.1828, 0.1923, 0.1955, 0.1764, 0.0842, 0.0715, 0.0413, 0.0413, 0.1017, 0.0794, 0.0444, 0.0524, 0.0524, 0.1001, 0.0953, 0.1176, 0.1176, 0.0715, -0.0223, -0.0207, -0.016, -0.0191, 0.0285, 0.0667, 0.0238, 0.0015, 0.0015, -0.0255, -0.0255, 0.0333, 0.0667, 0.1112, 0.0969, 0.0953, 0.0619, 0.1669, 0.1335, 0.1812, 0.1557, 0.1192, 0.1208, 0.116, 0.1478, 0.0333, 0.0556, -0.0223, -0.0493, -0.0557, -0.0621, -0.0112, 0.0444, 0.0508, 0.046, 0.0874, 0.0142, 0.0079, 0.0031, -0.008, 0.0365, 0.0715, 0.0476, 0.0238, 0.0079, -0.0414, 0.0174, 0.0079, 0.0222, 0.0254, 0.0238, 0.0111, 0.0111, 0.0222]}
{"unit_id": "orig:1930782373639328200", "ticker_idx": 0, "ts": "2025-06-06T00:22:44+00:00", "symbol": "ALAB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The company expects significant revenue growth in 2025/2026 driven by this initiative", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi note on Astera Labs; PCIe Gen 6 retimer leadership, NVLink Fusion chiplet opportunity, and UA Link positioning support significant 2025/2026 revenue growth.", "resolved_tickers": ["ALAB"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Astera Labs: Scorpio Series and Open Interconnect Strategy Position Company to Capitalize on AI Networking Growth\n\nCiti (A. Malik, 06/04/2025)\n\nDuring a recent visit to Silicon Valley, Astera Labs’ management detailed how its Scorpio product family, featuring the industry’s first PCIe Gen 6 retimer, is enabling large-scale AI network expansion. The company expects significant revenue growth in 2025/2026 driven by this initiative. Key themes included chiplet compatibility with NVLink Fusion and UA Link, multiple AI product roadmaps, customer diversification plans, and IP strategy flexibility.\n\n⸻\n\n• NVLink Fusion Integration & Chiplet Compatibility\n\nAstera explained that in NVLink Fusion deployments—where hyperscale data centers adopt NVIDIA’s cluster racks and switches—existing ASICs must connect to the NVLink protocol via an I/O chiplet.\nWhile customers may use NVIDIA’s in-house chiplets, Astera’s compatible chiplets offer an alternative when customers’ ASICs lack native NVLink integration.\n\nBeyond NVLink, Astera emphasized the monetization potential of its UA Link stack, which spans from retimers to PCIe switches, while also supporting PCIe and Ethernet for flexible, high-value system designs in large-scale deployments.\n\n⸻\n\n• Diverse AI Product Roadmaps & Market Strategy\n\nAstera was first to market with a PCIe Gen 6 retimer over a year ago and is now undergoing validation with NVIDIA’s Blackwell platform, with no competing products currently in sight.\nManagement expects AI hardware refresh cycles to accelerate, resulting in a faster Gen 6 to Gen 7 transition.\n\nThe retimer content is similar across NVIDIA’s B200 and B300, while the Taurus series is positioned to gain additional market share. Astera has partnered with cable manufacturers to ensure supply chain resilience at scale.\n\nFor Intel’s next-gen Leo CPU platform, CXL interconnects are expected to ramp in 2026. The base Scorpio SKU has already passed GB200 validation, and Scorpio X is on track for mass production in 2026.\n\nIn NVIDIA’s ecosystem, each Hopper GPU contributes ~$50 in chiplet content, but this could rise to hundreds of dollars in the Blackwell era, supporting the projection that Astera’s sales could see ~10% contribution from this segment by 2025.\n\n⸻\n\n• UA Link & Open Interconnect Value Proposition\n\nAstera believes UA Link matches NVLink in performance at 200G/s but offers the advantage of being open-standard, mitigating risks of vendor lock-in.\n\nCompared to a single Ethernet switch supporting both east-west and north-south scaling, Astera advocates dedicated low-latency interconnects like UA Link or NVLink for AI node-to-node communication.\n\nAstera plans to sample the first UA Link platforms shortly after large-scale Ethernet switches reach the market, with full-scale deployment expected between late 2026 and 2027, addressing the hyperscaler need to balance cost and performance in AI cluster expansion.\n\n⸻\n\n• Customer Concentration & Diversification Strategy\n\nCurrently, Astera primarily serves a single hyperscaler, but management aims to align its customer base with its AI market positioning, adding a second hyperscale cloud provider to its production pipeline to diversify risk and expand market reach.\n\nThe company is also targeting further penetration into traditional data center supply chains to diversify its revenue base.\n\n⸻\n\n• SerDes IP Strategy & Flexibility\n\nAstera has not yet developed in-house SerDes IP, but it shares high-quality IP from partner firms that is competitive with industry standards.\n\nThe company does not rule out developing its own link-layer IP if needed, but will weigh cost vs. performance when deciding whether to continue using third-party IP.\n\nThis multi-source IP approach gives Astera agility across projects, enabling it to meet diverse customer needs in bandwidth and latency performance.\n\n$ALAB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0004407814383178277, "ret_signed_m1d": -0.0004407814383178277, "alpha_spy_m1d": 0.00972381835014291, "alpha_c_m1d": 0.00521744442268024, "bench_spy_m1d": -0.010164599788460738, "bench_c_m1d": -0.005658225860998067, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017630837180397707, "ret_signed_p1d": 0.017630837180397707, "alpha_spy_p1d": 0.016729554292253423, "alpha_c_p1d": 0.0008936904970122672, "bench_spy_p1d": 0.0009012828881442836, "bench_c_p1d": 0.01673714668338544, "ret_p1w": -0.01123963243048387, "ret_signed_p1w": -0.01123963243048387, "alpha_spy_p1w": -0.007667885541877917, "alpha_c_p1w": -0.028095483999956206, "bench_spy_p1w": -0.0035717468886059534, "bench_c_p1w": 0.016855851569472335, "ret_p1m": -0.012451760368242026, "ret_signed_p1m": -0.012451760368242026, "alpha_spy_p1m": -0.05146586178240187, "alpha_c_p1m": -0.11940389371307447, "bench_spy_p1m": 0.03901410141415984, "bench_c_p1m": 0.10695213334483245, "ret_p3m": 0.9461157092049759, "ret_signed_p3m": 0.9461157092049759, "alpha_spy_p3m": 0.8684993536047243, "alpha_c_p3m": 0.8127718458599789, "bench_spy_p3m": 0.0776163556002516, "bench_c_p3m": 0.13334386334499704, "ret_p6m": 0.8202755089962122, "ret_signed_p6m": 0.8202755089962122, "alpha_spy_p6m": 0.6783509140299806, "alpha_c_p6m": 0.4236862395303067, "bench_spy_p6m": 0.14192459496623155, "bench_c_p6m": 0.39658926946590545, "price_path": [-0.2722, -0.2417, -0.2802, -0.2046, -0.2061, -0.2409, -0.2409, -0.24, -0.2165, -0.187, -0.2045, -0.2623, -0.3081, -0.3302, -0.3425, -0.3167, -0.2939, -0.3864, -0.4166, -0.3922, -0.4082, -0.2911, -0.3514, -0.3374, -0.3453, -0.338, -0.3522, -0.3503, -0.3503, -0.4003, -0.3905, -0.3372, -0.2887, -0.2715, -0.2718, -0.2766, -0.2803, -0.2372, -0.2147, -0.205, -0.2137, -0.2166, -0.2106, -0.2154, -0.1154, -0.0329, -0.001, 0.0159, -0.0021, -0.0051, 0.0409, 0.0156, 0.0391, 0.0426, 0.0426, 0.0748, 0.0629, 0.0646, -0.0003, 0.0239, 0.0478, 0.0493, -0.0004, 0.0, 0.0176, 0.0078, 0.0418, 0.0396, -0.0112, 0.0501, 0.0238, 0.0967, 0.0967, 0.0256, -0.0529, -0.0385, -0.0123, 0.0794, 0.003, -0.0036, -0.023, -0.024, 0.0006, 0.0006, -0.0125, 0.0171, 0.1004, 0.0691, 0.0567, -0.0047, 0.0177, 0.0131, 0.0793, 0.1254, 0.3431, 0.2883, 0.3166, 0.3408, 0.3469, 0.3669, 0.3048, 0.4201, 0.5067, 0.4446, 0.5199, 0.4936, 0.9217, 0.8831, 0.9755, 0.9772, 1.1157, 1.1338, 1.1013, 1.0479, 1.0543, 0.885, 0.9003, 0.9563, 0.9734, 0.919, 0.9676, 0.9747, 1.0843, 1.0077, 1.0077, 0.92, 0.9461, 1.0711, 1.1069, 1.3813, 1.3635, 1.5385, 1.5664, 1.5289, 1.5487, 1.6313, 1.748, 1.7755, 1.7019, 1.6149, 1.5356, 1.2619, 1.2109, 1.1794, 1.1906, 1.1576, 1.1433, 1.3096, 1.212, 1.4332, 1.3372, 1.4172, 1.4841, 1.2723, 1.1987, 0.7802, 0.7792, 0.8022, 0.7609, 0.7224, 0.7356, 0.7063, 0.8032, 0.8179, 0.8764, 0.854, 0.9132, 0.8683, 1.0571, 1.1109, 0.9759, 1.0048, 0.7943, 0.8236, 0.9145, 0.8463, 0.7387, 0.592, 0.5905, 0.558, 0.5374, 0.5648, 0.5349, 0.5625, 0.6281, 0.5954, 0.6994, 0.6994, 0.7363, 0.8203]}
{"unit_id": "orig:1937895180520313270", "ticker_idx": 0, "ts": "2025-06-25T15:26:30+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MRVL's data center business growth expectations are significantly underestimated by the market, with its long-term potential possibly far exceeding current forecasts… MRVL's 2028 EPS could reach USD 8, or even USD 10 under more optimistic assumptions, nearly double the current expectation of USD 4.25", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS note: MRVL data center EPS massively underestimated (2028E $8-$10 vs consensus $4.25); US semis face Section 232 tariff risk with final ruling ~Dec 2025.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "U.S. Semiconductors: Tariff and VEU Change Risks Persist, MRVL Data Center Growth Expectations Underestimated by Market\n\nUBS (T. Arcuri, 23/06/25)\n\nDespite waning investor attention, the Section 232 investigation is still progressing and may impose comprehensive tariffs on electronic products. Simultaneously, MRVL's data center business growth expectations are significantly underestimated by the market, with its long-term potential possibly far exceeding current forecasts.\n\nSection 232 Tariffs Still Pose Risks, Final Decision Expected by Year-End\n\nAlthough investor enthusiasm has waned, UBS notes that the Section 232 tariff investigation is ongoing. The final outcome could be a uniform tariff on all electronic products, initially without exemptions for U.S.-produced content due to high administrative costs. Once rules are established, U.S. companies will seek exemptions, but the outcome is uncertain. Given that the investigation began on April 1st, a final ruling is expected around December 2025, which may bring a series of negative news in the interim, potentially leading to a 25% tariff framework similar to the automotive industry.\n\nRevocation of VEU Exemptions Has Limited Impact on Korean and Taiwanese Manufacturers, Status Quo Largely Maintained\n\nRegarding rumors of VEU license revocations for Samsung, SK Hynix, and TSMC, UBS believes the impact will be limited. VEU licenses already require per-transaction approval with only slight simplification, and relevant factories are currently operating normally. Due to EUV restrictions, SK Hynix's Wuxi plant has remained at the 1a nm process with limited investment and will shift towards its Yongin plant in Korea in the future. Samsung's Xi'an plant also has limited investment, and both companies plan to transfer production capacity. Therefore, the institution believes the overall situation will maintain the status quo.\n\nMRVL Data Center Revenue Severely Underestimated by Market, Long-Term EPS Potential Up to 2x\n\nIn its latest event, MRVL raised its 2028 data center TAM to USD 94 billion (previously USD 75 billion), with custom accelerated computing accounting for USD 55 billion and a target market share of over 20%. If this target is achieved, data center revenue could reach USD 19 billion (of which custom computing is USD 11 billion), far exceeding current market expectations of USD 11 billion and USD 5 billion. Although UBS remains somewhat cautious, believing cloud vendors have limited willingness to deploy ASICs, if based on the company's guidance, MRVL's 2028 EPS could reach USD 8, or even USD 10 under more optimistic assumptions, nearly double the current expectation of USD 4.25.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009482377851554369, "ret_signed_m1d": -0.009482377851554369, "alpha_spy_m1d": -0.008922327053608736, "alpha_c_m1d": 0.0038170562503442307, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0132994341018986, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05320688282180064, "ret_signed_p1d": 0.05320688282180064, "alpha_spy_p1d": 0.04538310294731329, "alpha_c_p1d": 0.04593940878655989, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007267474035240751, "ret_p1w": -0.022125615487598393, "ret_signed_p1w": -0.022125615487598393, "alpha_spy_p1w": -0.04408175612929055, "alpha_c_p1w": -0.04410957705462648, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021983961567028087, "ret_p1m": -0.02409318093340518, "ret_signed_p1m": -0.02409318093340518, "alpha_spy_p1m": -0.06905956217085174, "alpha_c_p1m": -0.06987805600433294, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.045784875070927766, "ret_p3m": -0.004453696069020019, "ret_signed_p3m": -0.004453696069020019, "alpha_spy_p3m": -0.1058648638915265, "alpha_c_p3m": -0.17512960018726265, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.17067590411824263, "ret_p6m": 0.11411980336637151, "ret_signed_p6m": 0.11411980336637151, "alpha_spy_p6m": -0.00319705044181684, "alpha_c_p6m": -0.14768980022686407, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.2618096035932356, "price_path": [-0.1839, -0.19, -0.1756, -0.1682, -0.2681, -0.3497, -0.3291, -0.3419, -0.1981, -0.3045, -0.2969, -0.3117, -0.2978, -0.3161, -0.3191, -0.3191, -0.3497, -0.3333, -0.2918, -0.245, -0.224, -0.2269, -0.2272, -0.2313, -0.1973, -0.1791, -0.1837, -0.1937, -0.2584, -0.241, -0.2144, -0.1505, -0.1376, -0.1321, -0.1413, -0.1603, -0.1761, -0.1911, -0.2083, -0.1854, -0.2007, -0.2007, -0.1595, -0.1493, -0.1607, -0.2073, -0.1904, -0.1787, -0.1268, -0.1418, -0.0998, -0.0894, -0.0934, -0.1013, -0.0828, -0.1151, -0.0726, -0.0782, -0.0129, -0.0129, -0.0319, -0.0678, -0.0095, 0.0, 0.0532, 0.0162, 0.0194, 0.0041, -0.0221, -0.0099, -0.0099, -0.0577, -0.0524, -0.0483, -0.0338, -0.0416, -0.0443, -0.0456, -0.0661, -0.0509, -0.0161, -0.037, -0.0511, -0.0342, -0.0241, -0.0219, 0.0006, 0.0062, 0.0774, 0.0593, -0.0187, 0.0087, 0.01, -0.0072, -0.0002, 0.0194, 0.0186, 0.0256, 0.0455, 0.0418, 0.0042, 0.0115, -0.0501, -0.0613, -0.0614, -0.0378, -0.0385, -0.0212, -0.0142, 0.018, -0.1713, -0.1713, -0.1485, -0.1787, -0.1551, -0.1653, -0.1301, -0.119, -0.1156, -0.1223, -0.1123, -0.1112, -0.0924, -0.0644, -0.0216, -0.0212, -0.0045, -0.0164, 0.0557, 0.1047, 0.0962, 0.086, 0.1081, 0.1057, 0.1362, 0.1364, 0.172, 0.1463, 0.2192, 0.1952, 0.1292, 0.179, 0.1372, 0.1724, 0.1637, 0.16, 0.1322, 0.1113, 0.069, 0.0917, 0.1096, 0.17, 0.1667, 0.189, 0.1682, 0.2364, 0.1919, 0.1553, 0.2253, 0.231, 0.1992, 0.2297, 0.1782, 0.1782, 0.1543, 0.1402, 0.1007, 0.0378, 0.0726, 0.0114, 0.0215, 0.1052, 0.1004, 0.157, 0.157, 0.1791, 0.2016, 0.2252, 0.3216, 0.2951, 0.3046, 0.2134, 0.1725, 0.2196, 0.1795, 0.1136, 0.1113, 0.1088, 0.0776, 0.1141]}
{"unit_id": "orig:1937895180520313270", "ticker_idx": 1, "ts": "2025-06-25T15:26:30+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Section 232 tariff investigation is ongoing… may bring a series of negative news in the interim, potentially leading to a 25% tariff framework similar to the automotive industry", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS note: MRVL data center EPS massively underestimated (2028E $8-$10 vs consensus $4.25); US semis face Section 232 tariff risk with final ruling ~Dec 2025.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "US semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "U.S. Semiconductors: Tariff and VEU Change Risks Persist, MRVL Data Center Growth Expectations Underestimated by Market\n\nUBS (T. Arcuri, 23/06/25)\n\nDespite waning investor attention, the Section 232 investigation is still progressing and may impose comprehensive tariffs on electronic products. Simultaneously, MRVL's data center business growth expectations are significantly underestimated by the market, with its long-term potential possibly far exceeding current forecasts.\n\nSection 232 Tariffs Still Pose Risks, Final Decision Expected by Year-End\n\nAlthough investor enthusiasm has waned, UBS notes that the Section 232 tariff investigation is ongoing. The final outcome could be a uniform tariff on all electronic products, initially without exemptions for U.S.-produced content due to high administrative costs. Once rules are established, U.S. companies will seek exemptions, but the outcome is uncertain. Given that the investigation began on April 1st, a final ruling is expected around December 2025, which may bring a series of negative news in the interim, potentially leading to a 25% tariff framework similar to the automotive industry.\n\nRevocation of VEU Exemptions Has Limited Impact on Korean and Taiwanese Manufacturers, Status Quo Largely Maintained\n\nRegarding rumors of VEU license revocations for Samsung, SK Hynix, and TSMC, UBS believes the impact will be limited. VEU licenses already require per-transaction approval with only slight simplification, and relevant factories are currently operating normally. Due to EUV restrictions, SK Hynix's Wuxi plant has remained at the 1a nm process with limited investment and will shift towards its Yongin plant in Korea in the future. Samsung's Xi'an plant also has limited investment, and both companies plan to transfer production capacity. Therefore, the institution believes the overall situation will maintain the status quo.\n\nMRVL Data Center Revenue Severely Underestimated by Market, Long-Term EPS Potential Up to 2x\n\nIn its latest event, MRVL raised its 2028 data center TAM to USD 94 billion (previously USD 75 billion), with custom accelerated computing accounting for USD 55 billion and a target market share of over 20%. If this target is achieved, data center revenue could reach USD 19 billion (of which custom computing is USD 11 billion), far exceeding current market expectations of USD 11 billion and USD 5 billion. Although UBS remains somewhat cautious, believing cloud vendors have limited willingness to deploy ASICs, if based on the company's guidance, MRVL's 2028 EPS could reach USD 8, or even USD 10 under more optimistic assumptions, nearly double the current expectation of USD 4.25.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0132994341018986, "ret_signed_m1d": 0.0132994341018986, "alpha_spy_m1d": 0.012739383303952967, "alpha_c_m1d": 0.012739383303952967, "bench_spy_m1d": -0.0005600507979456326, "bench_c_m1d": -0.0005600507979456326, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007267474035240751, "ret_signed_p1d": -0.007267474035240751, "alpha_spy_p1d": 0.0005563058392465958, "alpha_c_p1d": 0.0005563058392465958, "bench_spy_p1d": 0.007823779874487347, "bench_c_p1d": 0.007823779874487347, "ret_p1w": 0.021983961567028087, "ret_signed_p1w": -0.021983961567028087, "alpha_spy_p1w": -2.7820925335930724e-05, "alpha_c_p1w": -2.7820925335930724e-05, "bench_spy_p1w": 0.021956140641692157, "bench_c_p1w": 0.021956140641692157, "ret_p1m": 0.045784875070927766, "ret_signed_p1m": -0.045784875070927766, "alpha_spy_p1m": -0.0008184938334812042, "alpha_c_p1m": -0.0008184938334812042, "bench_spy_p1m": 0.04496638123744656, "bench_c_p1m": 0.04496638123744656, "ret_p3m": 0.17067590411824263, "ret_signed_p3m": -0.17067590411824263, "alpha_spy_p3m": -0.06926473629573615, "alpha_c_p3m": -0.06926473629573615, "bench_spy_p3m": 0.10141116782250648, "bench_c_p3m": 0.10141116782250648, "ret_p6m": 0.2618096035932356, "ret_signed_p6m": -0.2618096035932356, "alpha_spy_p6m": -0.14449274978504723, "alpha_c_p6m": -0.14449274978504723, "bench_spy_p6m": 0.11731685380818835, "bench_c_p6m": 0.11731685380818835, "price_path": [-0.229, -0.2316, -0.2279, -0.2221, -0.2894, -0.343, -0.3281, -0.3461, -0.2339, -0.287, -0.2685, -0.2674, -0.2626, -0.2938, -0.3004, -0.3004, -0.3175, -0.3027, -0.2772, -0.2407, -0.2298, -0.2346, -0.2359, -0.2323, -0.2286, -0.2041, -0.2093, -0.2166, -0.2005, -0.1935, -0.1882, -0.1373, -0.1077, -0.0989, -0.102, -0.1046, -0.1061, -0.1078, -0.1226, -0.125, -0.1372, -0.1372, -0.1098, -0.1193, -0.1131, -0.1288, -0.1159, -0.096, -0.0852, -0.0868, -0.0816, -0.0663, -0.0477, -0.0484, -0.0438, -0.0662, -0.043, -0.0495, -0.0458, -0.0458, -0.0543, -0.0483, -0.0133, 0.0, 0.0073, 0.0117, 0.0134, 0.0023, 0.022, 0.0306, 0.0306, 0.0166, 0.0301, 0.0371, 0.0448, 0.0447, 0.0368, 0.0567, 0.0512, 0.06, 0.0549, 0.0556, 0.037, 0.0415, 0.0458, 0.0447, 0.0589, 0.0643, 0.0765, 0.0493, 0.0318, 0.0545, 0.0432, 0.0415, 0.0578, 0.0666, 0.0661, 0.0905, 0.0948, 0.0971, 0.0743, 0.0787, 0.0569, 0.05, 0.0445, 0.067, 0.068, 0.0783, 0.0816, 0.086, 0.0548, 0.0548, 0.041, 0.0408, 0.0533, 0.0657, 0.0775, 0.0822, 0.0931, 0.1044, 0.1058, 0.1162, 0.1163, 0.1092, 0.1518, 0.1472, 0.1707, 0.169, 0.1668, 0.1666, 0.1694, 0.1725, 0.1859, 0.2125, 0.2291, 0.2234, 0.2477, 0.2247, 0.2575, 0.2544, 0.1822, 0.2346, 0.2118, 0.2419, 0.2475, 0.2457, 0.2617, 0.2547, 0.2304, 0.2529, 0.2759, 0.3079, 0.3194, 0.3394, 0.3217, 0.3191, 0.3303, 0.2818, 0.3063, 0.2758, 0.265, 0.3037, 0.2756, 0.2919, 0.253, 0.2536, 0.2366, 0.2111, 0.2334, 0.1813, 0.1851, 0.2323, 0.2355, 0.2635, 0.2635, 0.2801, 0.2826, 0.3061, 0.324, 0.314, 0.3242, 0.3392, 0.3408, 0.3594, 0.3477, 0.2868, 0.2823, 0.2789, 0.2327, 0.2618]}
{"unit_id": "quote:1927703167489212832", "ticker_idx": 0, "ts": "2025-05-28T12:27:04+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we continue to prefer positioning for market upside around earnings", "tweet_type": "analysis", "timeline": "days", "summary": "JPM relayed: prefer long NVDA positioning into earnings; SMH put-call skew flags hedging pressure on semis ETF.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "J.P. Morgan – May 28: Equity & Macro Narrative\n\nIt appears that the macro environment is setting the stage for a potential melt-up.\n(i) The Bank of Japan (BOJ) stabilized the JGB market, leading to lower bond yields and triggering a global bond rally, which strengthened demand at the 2-year U.S. Treasury auction.\n(ii) The “short America” narrative has temporarily paused (or ended?), resulting in a stronger U.S. dollar.\n(iii) Expectations heading into NVIDIA’s earnings are relatively low.\n\nOn the micro level, however:\n(iv) Put option volume surged yesterday, with 121,000 puts traded on SMH (versus the 20-day average of 43,000), while call volume was only 19,000 (below the 20-day average of 26,000).\n(v) Retail investors were net sellers throughout the day, dumping around $1.2 billion in equities—equivalent to a -3.7z move—and particularly sold off NVIDIA stock.\n\nDespite this, we continue to prefer positioning for market upside around earnings.\n\n⸻\n\nBefore diving into the full preview, here are two perspectives worth considering:\n\n1. Page Hanson (Industrials)\n\nAI, TMT (Tech/Media/Telecom), and ex-TMT led today’s rally.\nHigh-beta names like VRT, which is still trading below levels seen before the bearish sentiment surged earlier this year (mostly among tech investors), rose +5.4%, leading the way.\nThis is an interesting move given the sharp AI rally over the past six weeks and NVIDIA’s upcoming earnings.\n\nAs the broader market strengthens, supportive data continues to emerge behind the AI trade.\nHistorically, NVIDIA and AI stocks often rallied sharply before earnings, only to weaken just ahead of the report as bulls took profits and de-risked.\nGiven that my colleague Josh Myers scores current market positioning at 7/10 (10 = max overbought), the continued rally right before earnings feels somewhat surprising.\n\n⸻\n\n2. Manish Sinha (Delta One)\n\nHe views NVIDIA’s earnings not just as company-specific, but as a warning for the broader market—one that could impact the ongoing growth narrative.\n\nHe specifically raised concerns about beta exposure, which he believes is beginning to show signs of exhaustion.\nHe notes that flows into NVIDIA from both retail and institutional investors have started to drop sharply.\nHistorically, when both flows and sentiment have turned down like this, NVIDIA’s stock has declined 15% to 20%.", "reply_text": null, "parent_text": null, "quoted_text": "Here are the earnings expectations for NVIDIA from the sell-side and buy-side:\n• Sell-side (consensus estimates):\n• Q1: $43.1 billion\n• Q2: $46.6 billion\n• Individual firm estimates:\n• Morgan Stanley: Q1 $42.2B / Q2 $43.5B\n• UBS: Q1 $44.0B / Q2 $44.6B\n• HSBC: Q1 $43.5B /", "parent_author": null, "ret_m1d": 0.005118180170402997, "ret_signed_m1d": 0.005118180170402997, "alpha_spy_m1d": -0.0007009000979054658, "alpha_c_m1d": -0.005650899803900433, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.01076907997430343, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03249016424752793, "ret_signed_p1d": 0.03249016424752793, "alpha_spy_p1d": 0.02854288690673168, "alpha_c_p1d": 0.025393300251513207, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.007096863996014724, "ret_p1w": 0.05274087220320278, "ret_signed_p1w": 0.05274087220320278, "alpha_spy_p1w": 0.03878900778980032, "alpha_c_p1w": 0.013996865305656359, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.038744006897546424, "ret_p1m": 0.14999444685150065, "ret_signed_p1m": 0.14999444685150065, "alpha_spy_p1m": 0.10584353163755589, "alpha_c_p1m": 0.006241331726588806, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.14375311512491185, "ret_p3m": 0.3338957051915541, "ret_signed_p3m": 0.3338957051915541, "alpha_spy_p3m": 0.2375262150750832, "alpha_c_p3m": 0.121154241170631, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.2127414640209231, "ret_p6m": 0.3401284775531095, "ret_signed_p6m": 0.3401284775531095, "alpha_spy_p6m": 0.22350433610615128, "alpha_c_p6m": -0.0012644559983867598, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.34139293355149625, "price_path": [-0.0734, -0.154, -0.1397, -0.13, -0.1799, -0.1642, -0.2065, -0.1933, -0.1415, -0.1426, -0.0975, -0.1133, -0.1438, -0.1283, -0.1208, -0.1269, -0.0994, -0.1047, -0.1561, -0.1734, -0.1865, -0.1961, -0.1829, -0.1809, -0.2449, -0.3004, -0.2757, -0.2857, -0.1519, -0.2021, -0.1771, -0.1788, -0.1677, -0.2249, -0.2472, -0.2472, -0.2811, -0.2664, -0.2381, -0.2105, -0.1765, -0.1935, -0.1913, -0.192, -0.1721, -0.1507, -0.1557, -0.1578, -0.1317, -0.1294, -0.1347, -0.0876, -0.0362, 0.0039, 0.0001, 0.0044, 0.0056, -0.0032, -0.0223, -0.0147, -0.0261, -0.0261, 0.0051, 0.0, 0.0325, 0.0024, 0.0191, 0.0475, 0.0527, 0.0384, 0.0513, 0.058, 0.0679, 0.0596, 0.0757, 0.0532, 0.0734, 0.0691, 0.0792, 0.0792, 0.0671, 0.0695, 0.0972, 0.1447, 0.15, 0.1702, 0.172, 0.1372, 0.1665, 0.182, 0.182, 0.1739, 0.1869, 0.2083, 0.2174, 0.2234, 0.2171, 0.2663, 0.2713, 0.2834, 0.279, 0.2714, 0.2391, 0.2669, 0.2889, 0.2871, 0.3112, 0.302, 0.3299, 0.3195, 0.2887, 0.3353, 0.3224, 0.331, 0.341, 0.3553, 0.3506, 0.3587, 0.3471, 0.3503, 0.3386, 0.3502, 0.303, 0.3012, 0.2981, 0.3204, 0.3339, 0.3484, 0.3472, 0.3366, 0.2921, 0.2921, 0.2669, 0.2657, 0.2734, 0.239, 0.2486, 0.2668, 0.3155, 0.3144, 0.3192, 0.3187, 0.2974, 0.2633, 0.3075, 0.3107, 0.3622, 0.3237, 0.3129, 0.3182, 0.322, 0.3491, 0.3842, 0.3891, 0.4013, 0.3919, 0.3765, 0.3728, 0.403, 0.4286, 0.3588, 0.3971, 0.3356, 0.3341, 0.3488, 0.3593, 0.355, 0.344, 0.3375, 0.3514, 0.3818, 0.4206, 0.4914, 0.536, 0.5052, 0.5022, 0.5348, 0.474, 0.4482, 0.3953, 0.3958, 0.4767, 0.433, 0.4378, 0.3863, 0.4108, 0.3843, 0.3455, 0.3838, 0.3401]}
{"unit_id": "quote:1927703167489212832", "ticker_idx": 1, "ts": "2025-05-28T12:27:04+00:00", "symbol": "SMH", "kind": "etf", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Put option volume surged yesterday, with 121,000 puts traded on SMH (versus the 20-day average of 43,000), while call volume was only 19,000", "tweet_type": "analysis", "timeline": "days", "summary": "JPM relayed: prefer long NVDA positioning into earnings; SMH put-call skew flags hedging pressure on semis ETF.", "resolved_tickers": ["SMH"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "J.P. Morgan – May 28: Equity & Macro Narrative\n\nIt appears that the macro environment is setting the stage for a potential melt-up.\n(i) The Bank of Japan (BOJ) stabilized the JGB market, leading to lower bond yields and triggering a global bond rally, which strengthened demand at the 2-year U.S. Treasury auction.\n(ii) The “short America” narrative has temporarily paused (or ended?), resulting in a stronger U.S. dollar.\n(iii) Expectations heading into NVIDIA’s earnings are relatively low.\n\nOn the micro level, however:\n(iv) Put option volume surged yesterday, with 121,000 puts traded on SMH (versus the 20-day average of 43,000), while call volume was only 19,000 (below the 20-day average of 26,000).\n(v) Retail investors were net sellers throughout the day, dumping around $1.2 billion in equities—equivalent to a -3.7z move—and particularly sold off NVIDIA stock.\n\nDespite this, we continue to prefer positioning for market upside around earnings.\n\n⸻\n\nBefore diving into the full preview, here are two perspectives worth considering:\n\n1. Page Hanson (Industrials)\n\nAI, TMT (Tech/Media/Telecom), and ex-TMT led today’s rally.\nHigh-beta names like VRT, which is still trading below levels seen before the bearish sentiment surged earlier this year (mostly among tech investors), rose +5.4%, leading the way.\nThis is an interesting move given the sharp AI rally over the past six weeks and NVIDIA’s upcoming earnings.\n\nAs the broader market strengthens, supportive data continues to emerge behind the AI trade.\nHistorically, NVIDIA and AI stocks often rallied sharply before earnings, only to weaken just ahead of the report as bulls took profits and de-risked.\nGiven that my colleague Josh Myers scores current market positioning at 7/10 (10 = max overbought), the continued rally right before earnings feels somewhat surprising.\n\n⸻\n\n2. Manish Sinha (Delta One)\n\nHe views NVIDIA’s earnings not just as company-specific, but as a warning for the broader market—one that could impact the ongoing growth narrative.\n\nHe specifically raised concerns about beta exposure, which he believes is beginning to show signs of exhaustion.\nHe notes that flows into NVIDIA from both retail and institutional investors have started to drop sharply.\nHistorically, when both flows and sentiment have turned down like this, NVIDIA’s stock has declined 15% to 20%.", "reply_text": null, "parent_text": null, "quoted_text": "Here are the earnings expectations for NVIDIA from the sell-side and buy-side:\n• Sell-side (consensus estimates):\n• Q1: $43.1 billion\n• Q2: $46.6 billion\n• Individual firm estimates:\n• Morgan Stanley: Q1 $42.2B / Q2 $43.5B\n• UBS: Q1 $44.0B / Q2 $44.6B\n• HSBC: Q1 $43.5B /", "parent_author": null, "ret_m1d": 0.01076907997430343, "ret_signed_m1d": -0.01076907997430343, "alpha_spy_m1d": -0.004949999705994967, "alpha_c_m1d": -0.004949999705994967, "bench_spy_m1d": 0.005819080268308463, "bench_c_m1d": 0.005819080268308463, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007096863996014724, "ret_signed_p1d": -0.007096863996014724, "alpha_spy_p1d": -0.0031495866552184726, "alpha_c_p1d": -0.0031495866552184726, "bench_spy_p1d": 0.003947277340796251, "bench_c_p1d": 0.003947277340796251, "ret_p1w": 0.038744006897546424, "ret_signed_p1w": -0.038744006897546424, "alpha_spy_p1w": -0.024792142484143964, "alpha_c_p1w": -0.024792142484143964, "bench_spy_p1w": 0.01395186441340246, "bench_c_p1w": 0.01395186441340246, "ret_p1m": 0.14375311512491185, "ret_signed_p1m": -0.14375311512491185, "alpha_spy_p1m": -0.09960219991096708, "alpha_c_p1m": -0.09960219991096708, "bench_spy_p1m": 0.044150915213944764, "bench_c_p1m": 0.044150915213944764, "ret_p3m": 0.2127414640209231, "ret_signed_p3m": -0.2127414640209231, "alpha_spy_p3m": -0.1163719739044522, "alpha_c_p3m": -0.1163719739044522, "bench_spy_p3m": 0.09636949011647089, "bench_c_p3m": 0.09636949011647089, "ret_p6m": 0.34139293355149625, "ret_signed_p6m": -0.34139293355149625, "alpha_spy_p6m": -0.22476879210453804, "alpha_c_p6m": -0.22476879210453804, "bench_spy_p6m": 0.11662414144695821, "bench_c_p6m": 0.11662414144695821, "price_path": [-0.0396, -0.0798, -0.0707, -0.0535, -0.0932, -0.0713, -0.1148, -0.1161, -0.0901, -0.0941, -0.0652, -0.0553, -0.0699, -0.0617, -0.0632, -0.0726, -0.0484, -0.0524, -0.084, -0.1015, -0.1246, -0.1275, -0.1232, -0.1166, -0.1931, -0.254, -0.2371, -0.2575, -0.1301, -0.1904, -0.1694, -0.1681, -0.1627, -0.1981, -0.2056, -0.2056, -0.225, -0.2082, -0.1793, -0.1378, -0.1254, -0.1308, -0.1323, -0.1282, -0.124, -0.0963, -0.1021, -0.1105, -0.0922, -0.0842, -0.0782, -0.0204, 0.0132, 0.0232, 0.0197, 0.0168, 0.0151, 0.0131, -0.0037, -0.0064, -0.0203, -0.0203, 0.0108, 0.0, 0.0071, -0.0108, 0.0039, 0.0265, 0.0387, 0.0369, 0.0428, 0.0602, 0.0813, 0.0805, 0.0858, 0.0604, 0.0867, 0.0793, 0.0835, 0.0835, 0.0739, 0.0807, 0.1204, 0.1355, 0.1438, 0.1488, 0.1507, 0.1381, 0.1605, 0.1702, 0.1702, 0.1543, 0.1697, 0.1777, 0.1863, 0.1862, 0.1773, 0.1999, 0.1936, 0.2036, 0.1978, 0.1987, 0.1775, 0.1826, 0.1875, 0.1862, 0.2024, 0.2085, 0.2224, 0.1915, 0.1716, 0.1974, 0.1846, 0.1826, 0.2012, 0.2111, 0.2106, 0.2382, 0.2431, 0.2458, 0.2199, 0.2249, 0.2002, 0.1923, 0.1861, 0.2116, 0.2127, 0.2244, 0.2282, 0.2332, 0.1978, 0.1978, 0.182, 0.1818, 0.196, 0.2101, 0.2235, 0.2288, 0.2413, 0.254, 0.2557, 0.2674, 0.2676, 0.2595, 0.3078, 0.3026, 0.3293, 0.3274, 0.3249, 0.3247, 0.3279, 0.3313, 0.3466, 0.3768, 0.3956, 0.3892, 0.4168, 0.3907, 0.4279, 0.4244, 0.3424, 0.4018, 0.3761, 0.4102, 0.4165, 0.4145, 0.4327, 0.4247, 0.3971, 0.4227, 0.4488, 0.4851, 0.4982, 0.5209, 0.5008, 0.4979, 0.5106, 0.4555, 0.4833, 0.4486, 0.4364, 0.4804, 0.4485, 0.4669, 0.4228, 0.4234, 0.4041, 0.3752, 0.4006, 0.3414]}
{"unit_id": "orig:1936209724271398986", "ticker_idx": 0, "ts": "2025-06-20T23:49:05+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the stock still has upside potential given the continued AI dividend", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish thesis on MU: HBM/LPDDR driving AI revenue share to 30%, margin convergence with SK Hynix, stock has upside elasticity.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron: Rapid Increase in AI Revenue Share Drives Performance Inflection Point, Valuation Has Upside Elasticity\n\nMorgan Stanley (J. Moore, 06/20/25)\n\nThe firm notes that Micron's AI business is growing significantly, with HBM and LPDDR demand driving ASP and gross margins higher. Although the valuation is not low, the stock still has upside potential given the continued AI dividend.\n\n- HBM and LPDDR drive AI-related revenue share from 14% to 30%, gross margin expected to converge with SK Hynix\n\nMicron's AI-related business is growing rapidly. HBM revenue in the August quarter is expected to increase by 54% quarter-over-quarter to $2 billion, with LPDDR and eSSD pricing also recovering. AI server demand is driving high-capacity DIMM growth, and the AI revenue share is expected to rise to 30% in the November quarter, leading to an overall ASP increase of about 30%. SK Hynix's AI revenue share has already reached 30%, with a gross margin close to 60%. The firm believes Micron is likely to converge in the coming quarters.\n\n- Consumer products and NAND drag factors weaken, gross margin improvement enters acceleration phase\n\nMarket concerns were raised in the May quarter guidance due to drag from consumer products and NAND. However, starting from the May quarter, eSSD volume ramp-up combined with rising NAND prices will significantly reduce the proportion of consumer products. NAND utilization rates are expected to increase, diluting fixed costs, and DDR4 spot prices have risen by 150% year-over-year. Although it accounts for a small proportion of revenue, it still contributes marginally to profitability.\n\n- Risks from changes in HBM supply-demand dynamics and smartphone inventory fluctuations\n\nSamsung still faces challenges in the high-end HBM market, which is beneficial for Micron in the short term. However, if Samsung's market share rebounds, it will compress market profit margins. Simultaneously, premature inventory build-up by end customers like iPhone could lead to a temporary demand decline in early 2026, requiring attention to inventory adjustment risks.\n\n- AI drives DRAM with clear logic, but free cash flow remains weaker than NVIDIA\n\nAlthough the AI trend will continue to drive HBM growth, Micron's capital expenditures still exceed 40% of its revenue, resulting in approximately $4 per share in free cash flow, significantly lower than NVIDIA. While both benefit from the AI wave, NVIDIA is more attractive in terms of free cash flow and valuation alignment.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014401291835186347, "ret_signed_m1d": -0.014401291835186347, "alpha_spy_m1d": -0.0167554600770643, "alpha_c_m1d": -0.023315118958739278, "bench_spy_m1d": 0.002354168241877952, "bench_c_m1d": 0.008913827123552931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012297753210128493, "ret_signed_p1d": -0.012297753210128493, "alpha_spy_p1d": -0.022175233714624154, "alpha_c_p1d": -0.018637384302408133, "bench_spy_p1d": 0.009877480504495662, "bench_c_p1d": 0.006339631092279641, "ret_p1w": 0.009385085114826941, "ret_signed_p1w": 0.009385085114826941, "alpha_spy_p1w": -0.025329082671318615, "alpha_c_p1w": -0.06035026880831085, "bench_spy_p1w": 0.034714167786145556, "bench_c_p1w": 0.06973535392313779, "ret_p1m": -0.08303739914343777, "ret_signed_p1m": -0.08303739914343777, "alpha_spy_p1m": -0.14107397994944193, "alpha_c_p1m": -0.19922448859645592, "bench_spy_p1m": 0.05803658080600416, "bench_c_p1m": 0.11618708945301814, "ret_p3m": 0.295635873547363, "ret_signed_p3m": 0.295635873547363, "alpha_spy_p3m": 0.18642811242568147, "alpha_c_p3m": 0.12281529522245593, "bench_spy_p3m": 0.10920776112168151, "bench_c_p3m": 0.17282057832490705, "ret_p6m": 0.9245342137955554, "ret_signed_p6m": 0.9245342137955554, "alpha_spy_p6m": 0.7758884120743372, "alpha_c_p6m": 0.5686344989966217, "bench_spy_p6m": 0.14864580172121822, "bench_c_p6m": 0.3558997147989338, "price_path": [-0.239, -0.2556, -0.2634, -0.2854, -0.297, -0.2823, -0.2832, -0.3985, -0.4764, -0.4468, -0.4697, -0.37, -0.4333, -0.4373, -0.4254, -0.4252, -0.4391, -0.4434, -0.4434, -0.46, -0.432, -0.41, -0.3736, -0.3545, -0.3644, -0.378, -0.3774, -0.3708, -0.3469, -0.3494, -0.3486, -0.3316, -0.3111, -0.3053, -0.2533, -0.2158, -0.2288, -0.2278, -0.2071, -0.2019, -0.2063, -0.2246, -0.2328, -0.2446, -0.2446, -0.2202, -0.2218, -0.2168, -0.2358, -0.2057, -0.1727, -0.1646, -0.14, -0.1217, -0.1023, -0.0765, -0.0612, -0.06, -0.0647, -0.0304, -0.0264, -0.0144, -0.0144, 0.0, -0.0123, 0.0349, 0.0295, 0.0194, 0.0094, -0.0028, -0.0219, -0.015, -0.0106, -0.0106, -0.0289, 0.0076, -0.0101, -0.003, 0.0085, -0.0395, -0.0273, -0.0571, -0.0828, -0.0736, -0.083, -0.1155, -0.1106, -0.0952, -0.099, -0.0991, -0.0933, -0.0708, -0.1162, -0.1507, -0.1273, -0.1168, -0.1191, -0.0941, -0.0372, 0.0019, 0.0345, 0.0064, 0.0146, -0.0212, 0.0005, -0.0116, -0.0508, -0.0623, -0.047, -0.0572, -0.0566, -0.0464, -0.012, -0.0362, -0.0362, -0.0405, -0.0386, 0.0059, 0.0639, 0.0646, 0.0952, 0.1338, 0.2194, 0.2733, 0.2777, 0.2862, 0.2956, 0.3677, 0.3178, 0.3331, 0.3476, 0.3096, 0.27, 0.2736, 0.3273, 0.355, 0.4751, 0.488, 0.522, 0.5474, 0.5047, 0.5926, 0.5585, 0.4716, 0.5621, 0.5158, 0.5553, 0.6412, 0.6399, 0.6755, 0.6392, 0.6083, 0.675, 0.7748, 0.7835, 0.7982, 0.8365, 0.8152, 0.8133, 0.9018, 0.7668, 0.9245, 0.9313, 0.9279, 1.0526, 0.9538, 0.9845, 0.9201, 1.0001, 0.9606, 0.8516, 0.8307, 0.6318, 0.6804, 0.8146, 0.8194, 0.8659, 0.8659, 0.9163, 0.9485, 0.9407, 0.8975, 0.8366, 0.9223, 1.0009, 1.0454, 1.1369, 1.0944, 0.954, 0.9245]}
{"unit_id": "reply:1932641061958234577", "ticker_idx": 0, "ts": "2025-06-11T03:28:30+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a decrease in Intel's server CPU market share from 70% to 50% is expected to result in a 7-11% reduction in Intel's revenue and a 10-20% decrease in its operating profit. I'm quite concerned about this", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish INTC: modeled server CPU share loss from 70% to 50% implies 7-11% revenue decline and 10-20% operating profit drop.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@Mojo_flyin According to my modeling, a decrease in Intel's server CPU market share from 70% to 50% is expected to result in a 7-11% reduction in Intel's revenue and a 10-20% decrease in its operating profit. I'm quite concerned about this.", "reply_text": null, "parent_text": "@Jukanlosreve Lisa Su has done an incredible transformation at $AMD. She also fully capitalized on the mistakes at $INTC.", "quoted_text": null, "parent_author": "Mojo_flyin", "ret_m1d": 0.06769823974228295, "ret_signed_m1d": -0.06769823974228295, "alpha_spy_m1d": -0.0648381684725099, "alpha_c_m1d": -0.06697267071064927, "bench_spy_m1d": 0.002860071269773057, "bench_c_m1d": 0.0007255690316336771, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004352038262076974, "ret_signed_p1d": -0.004352038262076974, "alpha_spy_p1d": -0.0003778121649224975, "alpha_c_p1d": 0.0004974795156500456, "bench_spy_p1d": 0.003974226097154476, "bench_c_p1d": 0.004849517777727019, "ret_p1w": 0.03916825212713637, "ret_signed_p1w": -0.03916825212713637, "alpha_spy_p1w": -0.045686771345254895, "alpha_c_p1w": -0.036457099790954794, "bench_spy_p1w": -0.006518519218118524, "bench_c_p1w": 0.002711152336181577, "ret_p1m": 0.15183749242317757, "ret_signed_p1m": -0.15183749242317757, "alpha_spy_p1m": -0.10808651203263109, "alpha_c_p1m": -0.053930093360794684, "bench_spy_p1m": 0.04375098039054648, "bench_c_p1m": 0.09790739906238288, "ret_p3m": 0.18375237819070467, "ret_signed_p3m": -0.18375237819070467, "alpha_spy_p3m": -0.10162511343721281, "alpha_c_p3m": -0.05147811613602449, "bench_spy_p3m": 0.08212726475349186, "bench_c_p3m": 0.13227426205468018, "ret_p6m": 0.9584138975986223, "ret_signed_p6m": -0.9584138975986223, "alpha_spy_p6m": -0.8138146073397827, "alpha_c_p6m": -0.5776287937985318, "bench_spy_p6m": 0.14459929025883955, "bench_c_p6m": 0.38078510380009045, "price_path": [0.163, 0.2423, 0.2534, 0.1663, 0.1586, 0.1731, 0.1712, 0.1702, 0.1325, 0.1422, 0.0982, 0.0982, 0.0662, 0.0629, 0.0846, -0.0401, -0.0537, -0.1233, 0.0411, -0.0387, -0.0455, -0.0179, -0.0401, -0.0701, -0.0846, -0.0846, -0.089, -0.0566, -0.0044, 0.0392, -0.0305, -0.0082, -0.0164, -0.028, -0.0338, -0.0029, -0.0198, -0.0358, -0.0179, 0.0155, 0.0358, 0.0725, 0.0909, 0.0406, 0.0421, 0.0474, 0.0338, 0.0285, 0.0005, -0.0063, -0.0305, -0.0305, -0.0063, -0.015, -0.0208, -0.0546, -0.0455, -0.0189, -0.0208, -0.0334, -0.03, -0.0097, 0.0677, 0.0, 0.0044, -0.0261, 0.0029, 0.0058, 0.0392, 0.0392, 0.0193, 0.0247, 0.0904, 0.0735, 0.088, 0.0972, 0.0832, 0.1049, 0.058, 0.0875, 0.0875, 0.0638, 0.1407, 0.1335, 0.1518, 0.133, 0.1267, 0.1083, 0.0972, 0.1025, 0.117, 0.1248, 0.1238, 0.1359, 0.0943, 0.001, 0.0, -0.0131, -0.0164, -0.0426, -0.0662, -0.0571, -0.0237, -0.0131, -0.044, -0.0353, -0.0015, 0.0546, 0.0745, 0.1538, 0.1876, 0.1441, 0.2239, 0.1383, 0.1364, 0.1992, 0.1871, 0.1775, 0.2016, 0.2055, 0.1775, 0.1775, 0.1707, 0.1605, 0.19, 0.1842, 0.1838, 0.1818, 0.1978, 0.19, 0.1644, 0.1978, 0.222, 0.2041, 0.4782, 0.4304, 0.3907, 0.4188, 0.5097, 0.6436, 0.7166, 0.6673, 0.6223, 0.7379, 0.8037, 0.7809, 0.7693, 0.7974, 0.81, 0.8279, 0.7587, 0.7998, 0.7229, 0.7964, 0.7814, 0.7897, 0.8424, 0.8433, 0.7853, 0.8453, 0.8511, 0.912, 1.0082, 0.999, 0.942, 0.9338, 0.9101, 0.7906, 0.8559, 0.8008, 0.8438, 0.8593, 0.8317, 0.8322, 0.7365, 0.7176, 0.6784, 0.6601, 0.6978, 0.6257, 0.6683, 0.7307, 0.7326, 0.78, 0.78, 0.9613, 0.9347, 1.102, 1.1161, 0.9584]}
{"unit_id": "orig:1925323152177209409", "ticker_idx": 0, "ts": "2025-05-21T22:49:44+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS maintains a Buy rating but lowers the target price from $180 to $175", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS Buy on NVDA with PT cut to $175; expects Q1 beat, stable Q2 guidance, and 2H reacceleration despite H20 ban and margin headwinds.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: 1Q25 Earnings Preview\nUBS (T. Arcuri, 2025/05/12)\n\nDue to the impact of the H20 export ban (estimated ~$800 million), Nvidia is still expected to slightly exceed its Q1 revenue guidance of $43 billion. For Q2, revenue guidance is expected to remain flat or increase slightly to around $44.6 billion. This shows unexpected stability amid widespread investor expectations of a downward revision. With GB300 rack shipments starting in late Q3 and China possibly resuming deliveries of modified Blackwell products, revenue growth in the second half is expected to reaccelerate. However, the continued use of the Bianca board design will weigh on gross margin, while the higher-margin Cordelia board won’t be adopted until next year’s Rubin platform.\n\n• FQ1 Revenue and Segment Breakdown:\n1Q (April) revenue is expected to reach approximately $44 billion, surpassing the $43 billion guidance. Data center revenue is estimated at $39.7 billion, including $36.3 billion from compute and $3.4 billion from networking. The H20 ban drags non-GAAP gross margin down to approximately 58.5%, well below the original guidance of 71%.\n\n• FQ2 Revenue Guidance Stability:\nGuidance for Q2 is expected to be flat or slightly up at around $44.6 billion. Despite the uncertainties caused by the export ban, this “stable” outlook should boost market confidence, especially as nearly half of investors anticipated a sequential decline.\n\n• 2H Acceleration Logic:\nWith GB300 racks expected to enter volume shipment in late Q3 and China possibly resuming shipments of a modified Blackwell SKU, revenue growth in the second half should notably rebound.\n\n• Gross Margin and Board Strategy:\nDue to continued use of the Bianca board, management is likely to remain cautious about gross margins in 2H. The rollout of the high-margin Cordelia board is delayed until the Rubin platform launches next year.\n\n• Valuation and Outlook:\nFY2026/CY2025e EPS is revised down to $4.22. UBS maintains a Buy rating but lowers the target price from $180 to $175, reflecting risks from the H20 export ban and gross margin pressure.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019575108185756473, "ret_signed_m1d": 0.019575108185756473, "alpha_spy_m1d": 0.0024356064959554224, "alpha_c_m1d": 0.0027200844755252973, "bench_spy_m1d": 0.01713950168980105, "bench_c_m1d": 0.016855023710231176, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0078148503695743, "ret_signed_p1d": 0.0078148503695743, "alpha_spy_p1d": 0.007420415168160277, "alpha_c_p1d": 0.010548135147802395, "bench_spy_p1d": 0.0003944352014140229, "bench_c_p1d": -0.0027332847782280956, "ret_p1w": 0.022837645516645688, "ret_signed_p1w": 0.022837645516645688, "alpha_spy_p1w": 0.014482376756294935, "alpha_c_p1w": 0.019151894665188074, "bench_spy_p1w": 0.008355268760350754, "bench_c_p1w": 0.0036857508514576143, "ret_p1m": 0.10387013772043896, "ret_signed_p1m": 0.10387013772043896, "alpha_spy_p1m": 0.078855595305521, "alpha_c_p1m": 0.016405912247264398, "bench_spy_p1m": 0.025014542414917962, "bench_c_p1m": 0.08746422547317456, "ret_p3m": 0.38105171493386103, "ret_signed_p3m": 0.38105171493386103, "alpha_spy_p3m": 0.2740934712593348, "alpha_c_p3m": 0.1516651002959548, "bench_spy_p3m": 0.10695824367452622, "bench_c_p3m": 0.22938661463790622, "ret_p6m": 0.41793249116159714, "ret_signed_p6m": 0.41793249116159714, "alpha_spy_p6m": 0.25831386466885475, "alpha_c_p6m": -0.010071000242057426, "bench_spy_p6m": 0.1596186264927424, "bench_c_p6m": 0.42800349140365457, "price_path": [0.0199, -0.0116, -0.0393, -0.004, -0.0885, -0.0523, -0.1347, -0.12, -0.1101, -0.1612, -0.1451, -0.1884, -0.1749, -0.1219, -0.1231, -0.0769, -0.0931, -0.1242, -0.1083, -0.1007, -0.107, -0.0788, -0.0843, -0.1369, -0.1546, -0.1679, -0.1777, -0.1643, -0.1622, -0.2276, -0.2844, -0.2592, -0.2693, -0.1325, -0.1838, -0.1583, -0.16, -0.1487, -0.2072, -0.23, -0.23, -0.2647, -0.2497, -0.2207, -0.1925, -0.1577, -0.175, -0.1728, -0.1736, -0.1532, -0.1313, -0.1364, -0.1385, -0.1118, -0.1095, -0.1149, -0.0668, -0.0142, 0.0269, 0.023, 0.0273, 0.0286, 0.0196, 0.0, 0.0078, -0.0039, -0.0039, 0.0281, 0.0228, 0.0561, 0.0253, 0.0423, 0.0715, 0.0768, 0.0621, 0.0753, 0.0822, 0.0923, 0.0838, 0.1002, 0.0772, 0.0979, 0.0936, 0.1039, 0.1039, 0.0915, 0.0939, 0.1222, 0.1709, 0.1763, 0.197, 0.1988, 0.1632, 0.1932, 0.209, 0.209, 0.2007, 0.214, 0.2359, 0.2452, 0.2514, 0.2449, 0.2952, 0.3003, 0.3127, 0.3082, 0.3004, 0.2674, 0.2958, 0.3183, 0.3165, 0.3411, 0.3317, 0.3603, 0.3496, 0.3181, 0.3658, 0.3526, 0.3614, 0.3716, 0.3863, 0.3814, 0.3898, 0.3779, 0.3811, 0.3692, 0.3811, 0.3327, 0.3309, 0.3277, 0.3505, 0.3644, 0.3792, 0.3779, 0.3671, 0.3216, 0.3216, 0.2958, 0.2946, 0.3025, 0.2673, 0.2771, 0.2957, 0.3455, 0.3444, 0.3493, 0.3488, 0.327, 0.2922, 0.3373, 0.3406, 0.3933, 0.354, 0.3429, 0.3483, 0.3521, 0.3799, 0.4158, 0.4208, 0.4333, 0.4237, 0.4079, 0.4041, 0.435, 0.4613, 0.3899, 0.429, 0.3661, 0.3646, 0.3796, 0.3903, 0.3859, 0.3747, 0.368, 0.3823, 0.4134, 0.4531, 0.5255, 0.5711, 0.5396, 0.5365, 0.5698, 0.5077, 0.4813, 0.4272, 0.4277, 0.5104, 0.4657, 0.4706, 0.4179]}
{"unit_id": "thread:1936933517830803963", "ticker_idx": 0, "ts": "2025-06-22T23:45:11+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Supply chain operators state that the order visibility from Apple, NVIDIA, AMD, plus the large demand from ASIC armies, extends to 2027. This forms the strongest foundation for TSMC's exceeding performance outlook.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish TSMC: advanced packaging (CoWoS, WMCM, SoIC) demand locked in through 2027 across Apple, NVIDIA, AMD and ASICs; AI accelerator revenue doubling in 2025.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: TSMC's WMCM and SoIC Dual Support Ensures Apple's Presence in Advanced Packaging\n\nAdvanced packaging continues to be a hot topic, and the industry is closely watching not only NVIDIA's large orders with TSMC, but also Apple's entry into the fray, with clear plans for collaboration with TSMC in advanced packaging.\n\nIt is understood that the monthly production capacity for Wafer-level Multi-chip Module (WMCM) packaging is estimated to reach 10,000 wafers by 2026, to be used for the A20 processor in the next-generation iPhone. Concurrently, Apple's proprietary AI server chips are, as expected, gradually beginning to adopt TSMC's 3D wafer stacking SoIC (System-on-Integrated-Chips) packaging technology.\n\nWMCM technology is understood to be an upgraded version of TSMC's InFO-PoP (Integrated Fan-Out Package-on-Package) technology, integrating advanced packaging techniques such as CoW (Chip-on-Wafer) and RDL (Redistribution Layer). By horizontally packaging logic chips and DRAM, it replaces the traditional vertical stacking method, achieving significant improvements in heat dissipation and performance. This is also an advanced packaging technology jointly developed by TSMC and Apple, specifically for Apple, becoming another highlight in addition to the A20 processor's adoption of the 2nm process.\n\nTo support its major client Apple, TSMC has even set up a dedicated production line at its P1 factory in Chiayi. The partnership between the two companies is rock solid, with mass production expected to begin in 2026. If Apple introduces its first foldable phone and enhances AI features, the market anticipates it will trigger another wave of device upgrades. Apple's proprietary AI server chips are gradually adopting SoIC packaging, with TSMC's AP6 factory in Zhunan serving as the mass production hub.\n\nTSMC's advanced packaging technology, along with its advanced process technology, continues to break through the Moore's Law bottleneck faced by next-generation chips. Supply chain operators state that the order visibility from Apple, NVIDIA, AMD, plus the large demand from ASIC (Application-Specific Integrated Circuit) armies, extends to 2027. This forms the strongest foundation for TSMC's exceeding performance outlook.\n\nTSMC's CoWoS (Chip-on-Wafer-on-Substrate) advanced packaging capacity remains in short supply, with major AI GPU and ASIC manufacturers having already reserved capacity for the next two years. Supply chain operators also confirmed that TSMC continues to expand CoWoS capacity, though the pace has slightly slowed due to a downward adjustment from previously ambitious targets. It is estimated that CoWoS monthly capacity will be around 75,000 wafers by the end of 2025, and between 105,000 and 115,000 wafers by the end of 2026.\n\nAfter resolving production bottlenecks, NVIDIA's GB200 is entering a stable large-scale shipment phase, and the GB300 chip platform has also begun sending samples to major CSP (Cloud Service Provider) manufacturers, with mass production expected to begin after Q2. The GB300 server rack solution is anticipated to ramp up quarter by quarter starting in Q4. NVIDIA is progressing according to its established roadmap, and TSMC Chairman C.C. Wei has also clearly stated that revenue from AI accelerators will double in 2025, and the company is working to double its full-year CoWoS-related capacity.\n\nPanel-level packaging is becoming TSMC's important weapon for the next generation. In fact, CoPoS (Chip on Panel on Substrate), which replaces wafers with panels, arranges chips on rectangular substrates, and then connects them to the underlying interposer through a packaging process, allows multiple chips to be packaged together. In fact, international packaging equipment manufacturers, material distributors, and others have reportedly received inquiries and plans from TSMC.\n\nThe \"turning circles into squares\" concept of CoPoS is estimated to be the future trend of advanced packaging. It is expected to enter full mass production at the AP7 factory in Chiayi by 2029, with major clients rumored to be NVIDIA and Broadcom.\n\nLooking ahead, the advancement of 2nm technology combined with advanced packaging will continue to drive the operational growth of TSMC's \"Grand Alliance\" partners. This includes companies like Topco, Giga Solution, Taiwan Speciality Chemicals, InnoLux, Kuniang, and Hermes Microvision, whose order visibility for related equipment and material supply chains is expected to extend without issues until 2027.\n\n$TSM $AAPL\n\n---\n\nhttps://t.co/XOJecq5w6s", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0038513868739630253, "ret_signed_m1d": -0.0038513868739630253, "alpha_spy_m1d": 0.00592948327630638, "alpha_c_m1d": 0.002448306486417251, "bench_spy_m1d": -0.009780870150269405, "bench_c_m1d": -0.006299693360380276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04645287181090407, "ret_signed_p1d": 0.04645287181090407, "alpha_spy_p1d": 0.03540566281781632, "alpha_c_p1d": 0.009724140113324786, "bench_spy_p1d": 0.011047208993087754, "bench_c_p1d": 0.03672873169757929, "ret_p1w": 0.07688276527060123, "ret_signed_p1w": 0.07688276527060123, "alpha_spy_p1w": 0.04739010592313453, "alpha_c_p1w": 0.012130215823267898, "bench_spy_p1w": 0.0294926593474667, "bench_c_p1w": 0.06475254944733333, "ret_p1m": 0.11544310415283543, "ret_signed_p1m": 0.11544310415283543, "alpha_spy_p1m": 0.06760509542374349, "alpha_c_p1m": 0.025911955617052396, "bench_spy_p1m": 0.04783800872909194, "bench_c_p1m": 0.08953114853578303, "ret_p3m": 0.2813211688404449, "ret_signed_p3m": 0.2813211688404449, "alpha_spy_p3m": 0.17783040382183746, "alpha_c_p3m": 0.07114247727857448, "bench_spy_p3m": 0.10349076501860743, "bench_c_p3m": 0.2101786915618704, "ret_p6m": 0.37196590218525305, "ret_signed_p6m": 0.37196590218525305, "alpha_spy_p6m": 0.23766260136593265, "alpha_c_p6m": 0.028273215953928288, "bench_spy_p6m": 0.1343033008193204, "bench_c_p6m": 0.34369268623132476, "price_path": [-0.1781, -0.203, -0.2172, -0.2136, -0.2011, -0.1928, -0.2544, -0.3046, -0.3075, -0.3303, -0.2479, -0.284, -0.2559, -0.2617, -0.2547, -0.2815, -0.2812, -0.2812, -0.2995, -0.2828, -0.2524, -0.2222, -0.2179, -0.2259, -0.2208, -0.2103, -0.1818, -0.1507, -0.1643, -0.1839, -0.1731, -0.1699, -0.1638, -0.1142, -0.081, -0.0774, -0.0799, -0.0799, -0.0833, -0.0836, -0.0916, -0.0706, -0.0905, -0.0905, -0.0635, -0.0708, -0.066, -0.0842, -0.077, -0.0639, -0.0412, -0.0367, -0.028, -0.0194, 0.0065, 0.0143, 0.0243, 0.0037, 0.0255, 0.017, 0.0151, 0.0151, -0.0039, 0.0, 0.0465, 0.0591, 0.0651, 0.0868, 0.0769, 0.0683, 0.1107, 0.1164, 0.1164, 0.0896, 0.0834, 0.1023, 0.0924, 0.0955, 0.0872, 0.1266, 0.1295, 0.1677, 0.143, 0.1357, 0.1154, 0.1427, 0.1487, 0.1677, 0.1542, 0.1474, 0.155, 0.1488, 0.1183, 0.1364, 0.1053, 0.1001, 0.1536, 0.1498, 0.1511, 0.1615, 0.148, 0.1459, 0.1358, 0.1478, 0.1064, 0.0869, 0.0809, 0.1078, 0.1202, 0.135, 0.1377, 0.1329, 0.0977, 0.0977, 0.0859, 0.1002, 0.1183, 0.1573, 0.1753, 0.193, 0.2383, 0.231, 0.233, 0.2428, 0.2499, 0.2534, 0.2813, 0.2633, 0.3004, 0.3484, 0.3389, 0.3196, 0.3038, 0.3032, 0.3321, 0.3759, 0.3742, 0.3936, 0.4423, 0.4024, 0.4525, 0.4303, 0.3387, 0.4447, 0.4115, 0.4534, 0.4301, 0.4074, 0.4199, 0.4047, 0.3779, 0.3867, 0.4069, 0.4226, 0.4382, 0.4552, 0.4463, 0.4329, 0.4541, 0.4025, 0.4006, 0.3796, 0.3665, 0.4083, 0.3888, 0.3862, 0.346, 0.3585, 0.3451, 0.3255, 0.3468, 0.3236, 0.3119, 0.3576, 0.3578, 0.383, 0.383, 0.3904, 0.3721, 0.3932, 0.4092, 0.3972, 0.4057, 0.4398, 0.4472, 0.4793, 0.458, 0.3967, 0.3761, 0.372]}
{"unit_id": "orig:1931265080513245279", "ticker_idx": 0, "ts": "2025-06-07T08:20:51+00:00", "symbol": "TSLA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if things like the approval for autonomous driving or subsidies get delayed or fizzle out due to Trump's pressure, it will inevitably have a direct impact on Tesla", "tweet_type": "analysis", "timeline": "weeks", "summary": "Expects TSLA to trade sideways/down if Trump-related benefits fizzle; AAPL pressured by tariff uncertainty overriding fundamentals.", "resolved_tickers": ["TSLA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Auto Manufacturers'", "bench_c_industry": "Auto Manufacturers", "unit_kind": "original", "lang": "en", "tweet_text": "A Quick Take: What Is Trump Afraid Of?\n\nYesterday, Trump and Musk had an incredibly fierce war of words. Who would have ever thought the relationship between these two ambitious, powerful figures, who once seemed so close, would come to this?\n\nBut amidst it all, what really stood out to me was how triggered Trump got by Musk's claim that he was on the Epstein list.\n\nAnd it got me thinking:\n\nWhat is Trump truly afraid of?\n\nThe one answer that immediately comes to mind is this: He's afraid of becoming a meme. Specifically, a meme that ridicules him.\n\nI'm not saying Trump dislikes all memes related to him. On the contrary, he seemed to enjoy his strongman image being consumed as a meme, and he even launched a memecoin.\n\nLet's think about this lightly. Why does Trump hate being ridiculed?\n\nOf course, who likes to be mocked? But I see it this way:\n\n“Trump wants to be feared by the people; he does not want to be a laughingstock.”\n\nFrom a political science perspective, leaders have always maintained power by being objects of fear. The moment they become objects of ridicule, they become a lame duck. It's just like the saying, \"Power grows out of the barrel of a gun.\"\n\nThat's why I think he got so furious when people started calling him \"TACO\" (Trump Always Chickens Out). The moment he becomes the subject of a meme, his power risks crumbling like a sandcastle.\n\nI believe Musk's claim about the Epstein list was another way to drag Trump into the realm of memes, ridicule, and scandal. And that's likely why Trump reacted so angrily and lashed out at Musk.\n\nSo, has he already become a lame duck?\n\nI believe he has, to some extent. What's the definition of a lame duck, after all? It means you give orders, but those below you don't carry them out properly.\n\nAlthough Trump still has a lot of his term left, his words are clearly losing their impact. As we can see from the \"TACO trade,\" the market doesn't seem to be taking him seriously anymore. The general attitude is, \"He'll throw a fit and then get over it.\"\n\nOf course, Trump will never admit he's a lame duck and will likely escalate the intensity of his statements to try and shake the market. That's why, in the short term, I expect both Tesla and Apple to trade sideways.\n\nSpecifically, if we want to \"predict\" the future impact on Tesla:\n\nThe reason Tesla was rallying since late last year was based on the logic that it would benefit from a Trump administration, right?\n\nBut if things like the approval for autonomous driving or subsidies get delayed or fizzle out due to Trump's pressure, it will inevitably have a direct impact on Tesla.\n\nOf course, if Musk manages to overpower Trump, the stock could rise.\n\nAs for Apple... hmm. It's directly exposed to tariffs, which are Trump's most powerful card. For that reason, I'm taking a wait-and-see approach with Apple.\n\nTariff policy is symbolic of Trump and his strongest weapon. The more cornered he feels, the more fiercely he will struggle. By constantly brandishing the sword of tariffs, Apple's stock price will likely be moved more by external variables like policy than by the company's own fundamentals.\n\nI'm not sure about semiconductor stocks. During the Trump-Musk conflict, the S&P 500 didn't show a major drop, aside from Tesla, so it seems we need to watch a bit longer for now.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.043554256622798326, "ret_signed_m1d": 0.043554256622798326, "alpha_spy_m1d": 0.04265378531403585, "alpha_c_m1d": 0.03560554053636511, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.007948716086433216, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05674382827003055, "ret_signed_p1d": -0.05674382827003055, "alpha_spy_p1d": -0.05107416360113981, "alpha_c_p1d": -0.04477397668293004, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.01196985158710051, "ret_p1w": 0.06659543458672856, "ret_signed_p1w": -0.06659543458672856, "alpha_spy_p1w": -0.061592759012324194, "alpha_c_p1w": -0.06589409631998544, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.0007013382667431234, "ret_p1m": -0.03490177419898777, "ret_signed_p1m": 0.03490177419898777, "alpha_spy_p1m": 0.07241170079621517, "alpha_c_p1m": 0.054786687990758476, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.01988491379177071, "ret_p3m": 0.0970575329259642, "ret_signed_p3m": -0.0970575329259642, "alpha_spy_p3m": -0.011413621304129595, "alpha_c_p3m": 0.004826789621755534, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.10188432254771973, "ret_p6m": 0.3910169450793346, "ret_signed_p6m": -0.3910169450793346, "alpha_spy_p6m": -0.24800744169435718, "alpha_c_p6m": -0.2811808862250236, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.10983605885431103, "price_path": [-0.196, -0.22, -0.1899, -0.2287, -0.2698, -0.2357, -0.2344, -0.194, -0.0978, -0.0662, -0.1183, -0.1149, -0.1459, -0.1602, -0.13, -0.0837, -0.1338, -0.2241, -0.244, -0.281, -0.1179, -0.1821, -0.1824, -0.1822, -0.1765, -0.2172, -0.2178, -0.2178, -0.2628, -0.2288, -0.1874, -0.159, -0.0766, -0.0736, -0.0536, -0.0856, -0.0909, -0.0693, -0.0918, -0.1077, -0.1049, -0.077, -0.0334, 0.0318, 0.0826, 0.1267, 0.111, 0.1342, 0.1086, 0.1142, 0.0844, 0.1052, 0.0997, 0.0997, 0.176, 0.1566, 0.1615, 0.1228, 0.1105, 0.1157, 0.0761, -0.0774, -0.0436, 0.0, 0.0567, 0.0578, 0.0341, 0.0542, 0.0666, 0.0252, 0.0437, 0.0437, 0.044, 0.13, 0.1033, 0.0615, 0.0557, 0.0488, 0.0294, -0.0255, 0.0229, 0.0219, 0.0219, -0.0474, -0.0349, -0.0412, 0.0042, 0.016, 0.027, 0.0071, 0.0424, 0.0351, 0.0683, 0.0645, 0.0763, 0.0777, -0.0106, 0.0242, 0.0551, 0.0409, 0.0339, -0.001, -0.0193, 0.0022, 0.0005, 0.0367, 0.0444, 0.0683, 0.0987, 0.1045, 0.0998, 0.0875, 0.0712, 0.0861, 0.0672, 0.0496, 0.0374, 0.1019, 0.1232, 0.1396, 0.1329, 0.1212, 0.082, 0.082, 0.0673, 0.0827, 0.0971, 0.1369, 0.1226, 0.1244, 0.1271, 0.1952, 0.2831, 0.3288, 0.3663, 0.3801, 0.3509, 0.3807, 0.4071, 0.38, 0.4349, 0.3721, 0.4272, 0.4363, 0.4412, 0.4889, 0.4129, 0.3929, 0.4688, 0.4035, 0.4216, 0.4114, 0.34, 0.4126, 0.391, 0.4102, 0.3894, 0.4237, 0.45, 0.4343, 0.4225, 0.455, 0.4055, 0.4661, 0.4925, 0.4956, 0.4262, 0.4796, 0.5178, 0.4397, 0.4974, 0.445, 0.3919, 0.4428, 0.4247, 0.3954, 0.3027, 0.3104, 0.3252, 0.3003, 0.3092, 0.2808, 0.2674, 0.3539, 0.3591, 0.3824, 0.3824, 0.394, 0.3939, 0.391]}
{"unit_id": "orig:1931265080513245279", "ticker_idx": 1, "ts": "2025-06-07T08:20:51+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Apple's stock price will likely be moved more by external variables like policy than by the company's own fundamentals", "tweet_type": "analysis", "timeline": "weeks", "summary": "Expects TSLA to trade sideways/down if Trump-related benefits fizzle; AAPL pressured by tariff uncertainty overriding fundamentals.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "A Quick Take: What Is Trump Afraid Of?\n\nYesterday, Trump and Musk had an incredibly fierce war of words. Who would have ever thought the relationship between these two ambitious, powerful figures, who once seemed so close, would come to this?\n\nBut amidst it all, what really stood out to me was how triggered Trump got by Musk's claim that he was on the Epstein list.\n\nAnd it got me thinking:\n\nWhat is Trump truly afraid of?\n\nThe one answer that immediately comes to mind is this: He's afraid of becoming a meme. Specifically, a meme that ridicules him.\n\nI'm not saying Trump dislikes all memes related to him. On the contrary, he seemed to enjoy his strongman image being consumed as a meme, and he even launched a memecoin.\n\nLet's think about this lightly. Why does Trump hate being ridiculed?\n\nOf course, who likes to be mocked? But I see it this way:\n\n“Trump wants to be feared by the people; he does not want to be a laughingstock.”\n\nFrom a political science perspective, leaders have always maintained power by being objects of fear. The moment they become objects of ridicule, they become a lame duck. It's just like the saying, \"Power grows out of the barrel of a gun.\"\n\nThat's why I think he got so furious when people started calling him \"TACO\" (Trump Always Chickens Out). The moment he becomes the subject of a meme, his power risks crumbling like a sandcastle.\n\nI believe Musk's claim about the Epstein list was another way to drag Trump into the realm of memes, ridicule, and scandal. And that's likely why Trump reacted so angrily and lashed out at Musk.\n\nSo, has he already become a lame duck?\n\nI believe he has, to some extent. What's the definition of a lame duck, after all? It means you give orders, but those below you don't carry them out properly.\n\nAlthough Trump still has a lot of his term left, his words are clearly losing their impact. As we can see from the \"TACO trade,\" the market doesn't seem to be taking him seriously anymore. The general attitude is, \"He'll throw a fit and then get over it.\"\n\nOf course, Trump will never admit he's a lame duck and will likely escalate the intensity of his statements to try and shake the market. That's why, in the short term, I expect both Tesla and Apple to trade sideways.\n\nSpecifically, if we want to \"predict\" the future impact on Tesla:\n\nThe reason Tesla was rallying since late last year was based on the logic that it would benefit from a Trump administration, right?\n\nBut if things like the approval for autonomous driving or subsidies get delayed or fizzle out due to Trump's pressure, it will inevitably have a direct impact on Tesla.\n\nOf course, if Musk manages to overpower Trump, the stock could rise.\n\nAs for Apple... hmm. It's directly exposed to tariffs, which are Trump's most powerful card. For that reason, I'm taking a wait-and-see approach with Apple.\n\nTariff policy is symbolic of Trump and his strongest weapon. The more cornered he feels, the more fiercely he will struggle. By constantly brandishing the sword of tariffs, Apple's stock price will likely be moved more by external variables like policy than by the company's own fundamentals.\n\nI'm not sure about semiconductor stocks. During the Trump-Musk conflict, the S&P 500 didn't show a major drop, aside from Tesla, so it seems we need to watch a bit longer for now.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01226100962740917, "ret_signed_m1d": -0.01226100962740917, "alpha_spy_m1d": -0.013161480936171643, "alpha_c_m1d": -0.017145548873212624, "bench_spy_m1d": -0.0009004713087624738, "bench_c_m1d": -0.004884539245803454, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00605594241060059, "ret_signed_p1d": -0.00605594241060059, "alpha_spy_p1d": -0.00038627774170985063, "alpha_c_p1d": -0.00046166790732393714, "bench_spy_p1d": 0.005669664668890739, "bench_c_p1d": 0.005594274503276653, "ret_p1w": -0.015040997154461389, "ret_signed_p1w": 0.015040997154461389, "alpha_spy_p1w": 0.020043672728865758, "alpha_c_p1w": 0.029694678958964937, "bench_spy_p1w": 0.005002675574404369, "bench_c_p1w": 0.014653681804503549, "ret_p1m": 0.04249182456139544, "ret_signed_p1m": -0.04249182456139544, "alpha_spy_p1m": -0.004981897964168036, "alpha_c_p1m": 0.029251915779288984, "bench_spy_p1m": 0.037509926597227405, "bench_c_p1m": 0.07174374034068443, "ret_p3m": 0.19162134557723287, "ret_signed_p3m": -0.19162134557723287, "alpha_spy_p3m": -0.10597743395539827, "alpha_c_p3m": -0.09332339681975976, "bench_spy_p3m": 0.0856439116218346, "bench_c_p3m": 0.09829794875747311, "ret_p6m": 0.42364121712305636, "ret_signed_p6m": -0.42364121712305636, "alpha_spy_p6m": -0.28063171373807894, "alpha_c_p6m": -0.21230277674546283, "bench_spy_p6m": 0.14300950338497742, "bench_c_p6m": 0.21133844037759353, "price_path": [0.0757, 0.0395, 0.0584, 0.0609, 0.0544, 0.0671, 0.0614, 0.0821, 0.0943, 0.1092, 0.0982, 0.1097, 0.0802, 0.1012, 0.1065, 0.1099, 0.0073, -0.0661, -0.1004, -0.1452, -0.0142, -0.056, -0.0177, 0.004, 0.0021, -0.0369, -0.0235, -0.0235, -0.0424, -0.0098, 0.0143, 0.033, 0.0375, 0.0418, 0.0471, 0.0535, 0.0575, 0.018, -0.014, -0.0159, -0.0271, -0.0209, -0.0158, 0.0464, 0.057, 0.054, 0.0496, 0.0487, 0.0364, 0.0269, 0.0032, -0.0004, -0.0307, -0.0307, -0.0062, -0.0051, -0.0074, -0.003, 0.0012, 0.009, 0.0068, -0.0041, 0.0123, 0.0, 0.0061, -0.0133, -0.0112, -0.0248, -0.015, -0.0288, -0.0242, -0.0242, -0.0022, 0.0002, -0.0057, 0.0005, -0.0022, -0.0018, 0.0185, 0.0316, 0.0546, 0.0601, 0.0601, 0.0422, 0.0425, 0.0481, 0.0544, 0.0482, 0.0356, 0.038, 0.0432, 0.0425, 0.0483, 0.0548, 0.0643, 0.063, 0.0611, 0.0617, 0.0625, 0.0487, 0.0377, 0.0304, 0.0046, 0.0094, 0.0073, 0.0586, 0.0922, 0.1385, 0.129, 0.1413, 0.1596, 0.1568, 0.1509, 0.1474, 0.1458, 0.1232, 0.1177, 0.1319, 0.1289, 0.1396, 0.1455, 0.1557, 0.1537, 0.1537, 0.1416, 0.1851, 0.1916, 0.1912, 0.1822, 0.1646, 0.1271, 0.1432, 0.1632, 0.1763, 0.1835, 0.1877, 0.1822, 0.22, 0.2726, 0.2644, 0.2539, 0.2766, 0.2695, 0.2644, 0.2654, 0.2695, 0.2778, 0.2823, 0.2757, 0.2746, 0.2825, 0.2625, 0.2189, 0.2308, 0.2313, 0.2391, 0.2297, 0.2538, 0.3032, 0.3059, 0.2844, 0.29, 0.3061, 0.3359, 0.3368, 0.3403, 0.3488, 0.3436, 0.3371, 0.342, 0.3425, 0.3407, 0.3342, 0.3403, 0.3692, 0.3604, 0.3578, 0.3551, 0.3305, 0.3304, 0.3359, 0.3245, 0.3505, 0.3726, 0.3778, 0.3807, 0.3807, 0.3871, 0.4083, 0.4236]}
{"unit_id": "orig:1944914612559990879", "ticker_idx": 0, "ts": "2025-07-15T00:19:13+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Companies like AVGO (Broadcom), which previously requested HBM supply from Samsung, are now urgently trying to secure HBM3E from MU (Micron) and SK Hynix, with MU stating it has no additional supply. Harlan still believes supply is very tight.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "JPM/Harlan sees HBM supply tight, bullish MU despite noise; NAND weakness is negative for SNDK.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron - JPM\n\nMicron's stock price has fallen due to concerns stemming from a competitor's (Edgewater) research findings, with the key points being:\n\nPotential HBM (High Bandwidth Memory) oversupply – Specifically, there's speculation that NVIDIA holds a significant amount of HBM3E inventory and may have overbooked its volume for next year. This could be one reason why they haven't entered into supply agreements.\n\nNAND does not appear to be in shortage – This is negative for SanDisk (SNDK), which indeed dropped 4%. These two issues are resonating with investors I've spoken with, and some believe they connect with related data points.\n\nThe first point differs somewhat from our outlook. We anticipate Ultra demand to be stronger than expected, but we've sufficiently highlighted the negative feedback we've heard in the market regarding MU (Micron). The second point aligns more with what we've previously mentioned: the NAND market has still not recovered and continues to pose a persistent risk.\n\nConversation with Harlan Regarding Memory-Related Noise:\nIt's difficult to believe the talk that NVIDIA won't purchase any HBM3E next year – Several individuals have also commented that this might just be noise generated during supply contract negotiations.\n\nCompanies like AVGO (Broadcom), which previously requested HBM supply from Samsung, are now urgently trying to secure HBM3E from MU (Micron) and SK Hynix, with MU stating it has no additional supply.\n\nHarlan still believes supply is very tight.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012488607049965128, "ret_signed_m1d": -0.012488607049965128, "alpha_spy_m1d": -0.01678011241993682, "alpha_c_m1d": 0.0063558938940968535, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.030638638779744487, "ret_signed_p1d": -0.030638638779744487, "alpha_spy_p1d": -0.0339818298982596, "alpha_c_p1d": -0.02541166387107685, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.09066689024244401, "ret_signed_p1w": -0.09066689024244401, "alpha_spy_p1w": -0.10146827366325939, "alpha_c_p1w": -0.07199429532673585, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.0346348225549622, "ret_signed_p1m": 0.0346348225549622, "alpha_spy_p1m": -0.0019325113681034622, "alpha_c_p1m": -0.001403755136877738, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.5128941986156961, "ret_signed_p3m": 0.5128941986156961, "alpha_spy_p3m": 0.460349080475132, "alpha_c_p3m": 0.3941183969209321, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 1.8299049735093238, "ret_signed_p6m": 1.8299049735093238, "alpha_spy_p6m": 1.7151476702281367, "alpha_c_p6m": 1.501782473673545, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.4277, -0.4277, -0.4449, -0.416, -0.3934, -0.356, -0.3364, -0.3465, -0.3605, -0.3599, -0.3531, -0.3286, -0.3311, -0.3303, -0.3128, -0.2917, -0.2858, -0.2323, -0.1937, -0.2071, -0.2061, -0.1848, -0.1794, -0.184, -0.2028, -0.2112, -0.2234, -0.2234, -0.1983, -0.2, -0.1948, -0.2143, -0.1834, -0.1495, -0.1412, -0.1159, -0.097, -0.0771, -0.0506, -0.0349, -0.0336, -0.0385, -0.0032, 0.001, 0.0133, 0.0133, 0.0281, 0.0154, 0.0639, 0.0584, 0.0481, 0.0377, 0.0252, 0.0055, 0.0126, 0.0172, 0.0172, -0.0016, 0.0359, 0.0177, 0.025, 0.0368, -0.0125, 0.0, -0.0306, -0.057, -0.0476, -0.0573, -0.0907, -0.0856, -0.0698, -0.0737, -0.0738, -0.0679, -0.0447, -0.0913, -0.1268, -0.1027, -0.092, -0.0943, -0.0686, -0.0102, 0.0301, 0.0636, 0.0346, 0.0431, 0.0063, 0.0286, 0.0162, -0.0241, -0.036, -0.0202, -0.0307, -0.0301, -0.0196, 0.0157, -0.0092, -0.0092, -0.0136, -0.0116, 0.0341, 0.0937, 0.0945, 0.126, 0.1656, 0.2536, 0.309, 0.3135, 0.3223, 0.332, 0.4061, 0.3548, 0.3706, 0.3855, 0.3463, 0.3057, 0.3094, 0.3646, 0.3931, 0.5165, 0.5298, 0.5648, 0.5909, 0.547, 0.6374, 0.6023, 0.5129, 0.606, 0.5584, 0.599, 0.6873, 0.686, 0.7226, 0.6853, 0.6534, 0.7221, 0.8246, 0.8336, 0.8487, 0.888, 0.8662, 0.8642, 0.9553, 0.8164, 0.9786, 0.9855, 0.9821, 1.1102, 1.0087, 1.0402, 0.974, 1.0563, 1.0157, 0.9036, 0.8821, 0.6776, 0.7276, 0.8655, 0.8705, 0.9183, 0.9183, 0.9701, 1.0033, 0.9952, 0.9508, 0.8882, 0.9763, 1.0571, 1.1029, 1.1969, 1.1532, 1.0089, 0.9786, 0.937, 0.8788, 1.0706, 1.2154, 1.3042, 1.3016, 1.3883, 1.3883, 1.3726, 1.4534, 1.4389, 1.3787, 1.3787, 1.6288, 1.6015, 1.8622, 1.8299]}
{"unit_id": "orig:1944914612559990879", "ticker_idx": 1, "ts": "2025-07-15T00:19:13+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND does not appear to be in shortage – This is negative for SanDisk (SNDK), which indeed dropped 4%", "tweet_type": "analysis", "timeline": "unspecified", "summary": "JPM/Harlan sees HBM supply tight, bullish MU despite noise; NAND weakness is negative for SNDK.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Micron - JPM\n\nMicron's stock price has fallen due to concerns stemming from a competitor's (Edgewater) research findings, with the key points being:\n\nPotential HBM (High Bandwidth Memory) oversupply – Specifically, there's speculation that NVIDIA holds a significant amount of HBM3E inventory and may have overbooked its volume for next year. This could be one reason why they haven't entered into supply agreements.\n\nNAND does not appear to be in shortage – This is negative for SanDisk (SNDK), which indeed dropped 4%. These two issues are resonating with investors I've spoken with, and some believe they connect with related data points.\n\nThe first point differs somewhat from our outlook. We anticipate Ultra demand to be stronger than expected, but we've sufficiently highlighted the negative feedback we've heard in the market regarding MU (Micron). The second point aligns more with what we've previously mentioned: the NAND market has still not recovered and continues to pose a persistent risk.\n\nConversation with Harlan Regarding Memory-Related Noise:\nIt's difficult to believe the talk that NVIDIA won't purchase any HBM3E next year – Several individuals have also commented that this might just be noise generated during supply contract negotiations.\n\nCompanies like AVGO (Broadcom), which previously requested HBM supply from Samsung, are now urgently trying to secure HBM3E from MU (Micron) and SK Hynix, with MU stating it has no additional supply.\n\nHarlan still believes supply is very tight.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005618016657510938, "ret_signed_m1d": 0.005618016657510938, "alpha_spy_m1d": 0.009909522027482631, "alpha_c_m1d": -0.0031835895964245076, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.008801606253935446, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03183521937008915, "ret_signed_p1d": 0.03183521937008915, "alpha_spy_p1d": 0.03517841048860426, "alpha_c_p1d": 0.03493700026849844, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00310178089840929, "ret_p1w": -0.03183521937008915, "ret_signed_p1w": 0.03183521937008915, "alpha_spy_p1w": 0.04263660279090453, "alpha_c_p1w": 0.03532478228368163, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.003489562913592481, "ret_p1m": 0.10042127758065345, "ret_signed_p1m": -0.10042127758065345, "alpha_spy_p1m": -0.06385394365758779, "alpha_c_p1m": -0.06009729426098276, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.040323983319670686, "ret_p3m": 1.736657310895675, "ret_signed_p3m": -1.736657310895675, "alpha_spy_p3m": -1.684112192755111, "alpha_c_p3m": -1.6558625003314664, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.08079481056420867, "ret_p6m": 7.276216934826541, "ret_signed_p6m": -7.276216934826541, "alpha_spy_p6m": -7.1614596315453545, "alpha_c_p6m": -7.136743125557316, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.1394738092692258, "price_path": [-0.2676, -0.2676, -0.3022, -0.3066, -0.2893, -0.2441, -0.231, -0.2446, -0.2289, -0.2484, -0.2371, -0.1948, -0.1898, -0.2081, -0.1814, -0.1419, -0.1166, -0.0339, -0.0169, -0.0185, -0.051, -0.0534, -0.0922, -0.0595, -0.0964, -0.1142, -0.1273, -0.1273, -0.1063, -0.0969, -0.0964, -0.1177, -0.1262, -0.0955, -0.0679, -0.0843, -0.0836, -0.0211, -0.0265, -0.0583, -0.0332, -0.0051, 0.0349, 0.0321, 0.0913, 0.0913, 0.0904, 0.099, 0.1081, 0.106, 0.1105, 0.1037, 0.0616, 0.0524, 0.0817, 0.0864, 0.0864, 0.0585, 0.0808, 0.0815, 0.099, 0.0789, -0.0056, 0.0, -0.0318, -0.0281, -0.0124, -0.026, -0.0318, 0.0066, -0.0154, -0.0056, -0.0194, 0.0049, 0.0157, 0.0047, -0.0325, -0.0049, -0.0185, -0.0145, -0.0475, 0.0379, 0.0152, 0.0962, 0.1004, 0.0927, 0.0426, 0.0655, 0.0435, 0.0393, 0.0651, 0.0854, 0.095, 0.1084, 0.1339, 0.1908, 0.2282, 0.2282, 0.1955, 0.2409, 0.463, 0.6046, 0.6502, 0.6505, 0.7303, 0.9733, 1.0162, 1.1088, 1.143, 1.1997, 1.3144, 1.3926, 1.4092, 1.4906, 1.3368, 1.2072, 1.2734, 1.6568, 1.6264, 1.8352, 1.9059, 2.0059, 1.8364, 1.8312, 2.0871, 2.0356, 1.7367, 2.151, 1.9796, 2.3778, 2.3771, 2.2809, 2.4654, 2.4946, 2.4398, 2.9103, 3.3577, 3.1313, 3.1088, 3.7837, 3.5838, 3.666, 3.8457, 3.5545, 4.0679, 3.8617, 4.6058, 5.2722, 5.3572, 5.6269, 4.7015, 4.9494, 5.2238, 4.7334, 4.7575, 3.5871, 3.688, 4.3127, 4.1615, 4.0337, 4.0337, 4.2266, 3.9197, 3.8069, 3.5501, 3.9932, 4.3481, 4.2779, 4.1372, 4.4508, 4.6557, 3.8263, 3.7254, 3.8996, 3.8415, 4.1372, 4.562, 4.6426, 4.7327, 4.8539, 4.8539, 4.8532, 4.7175, 4.6231, 4.5566, 4.5566, 5.4429, 5.4157, 7.1842, 7.2762]}
{"unit_id": "thread:1947071472318767388", "ticker_idx": 0, "ts": "2025-07-20T23:09:48+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC plans to significantly expand its 2nm production capacity… from approximately 40,000 wafers per month by the end of this year to 100,000 wafers per month next year… and then double it again to 200,000 wafers per month by 2027", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC 2nm capacity tripling by 2027 on strong demand from Apple/AMD/Intel/NVIDIA/Qualcomm/MediaTek; supply shortages signal durable revenue tailwind.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "• TSMC: The 2nm process is scheduled to enter mass production as planned in the second half of this year. According to industry sources, demand from primary customers like Apple, AMD, and Intel is very strong, and with Qualcomm, MediaTek, and NVIDIA also planning to successively adopt the 2nm process, TSMC's 2nm supply is facing severe shortages.\n\n• In response, TSMC plans to significantly expand its 2nm production capacity. Specifically, it plans to increase 2nm production capacity from approximately 40,000 wafers per month by the end of this year to 100,000 wafers per month next year (a 1.5x increase), and then double it again to 200,000 wafers per month by 2027.\n\n• It is anticipated that as early as 2027, the 2nm process will become the largest production volume node among TSMC's advanced processes below 7nm.\n\n---\n\nhttps://t.co/zUNoPnKQsZ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004347821422747922, "ret_signed_m1d": 0.004347821422747922, "alpha_spy_m1d": 0.006240407912912804, "alpha_c_m1d": 0.005036237310665825, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01739139296779124, "ret_signed_p1d": -0.01739139296779124, "alpha_spy_p1d": -0.01753451413101581, "alpha_c_p1d": 0.0003016224402693224, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.004347821422747811, "ret_signed_p1w": -0.004347821422747811, "alpha_spy_p1w": -0.01734150010751334, "alpha_c_p1w": -0.007445850981110946, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.030434749959234786, "ret_signed_p1m": 0.030434749959234786, "alpha_spy_p1m": 0.01287666131789611, "alpha_c_p1m": 0.029195538135889487, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.29646962779241504, "ret_signed_p3m": 0.29646962779241504, "alpha_spy_p3m": 0.2428703605884337, "alpha_c_p3m": 0.11475480942177407, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.4978807468181701, "ret_signed_p6m": 0.4978807468181701, "alpha_spy_p6m": 0.38817582996011235, "alpha_c_p6m": 0.14595386200174998, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.2441, -0.2519, -0.2311, -0.2233, -0.219, -0.2138, -0.2138, -0.1774, -0.1878, -0.2034, -0.1965, -0.2051, -0.1783, -0.1713, -0.161, -0.135, -0.1402, -0.1358, -0.148, -0.1532, -0.1428, -0.1497, -0.148, -0.1549, -0.1644, -0.1627, -0.1627, -0.1627, -0.1809, -0.1774, -0.1428, -0.1358, -0.1384, -0.1298, -0.0951, -0.0778, -0.0913, -0.1043, -0.1087, -0.0913, -0.0826, -0.1, -0.0826, -0.113, -0.087, -0.0696, -0.0652, -0.0609, -0.0783, -0.0565, -0.0565, -0.0522, -0.0565, -0.0609, -0.0609, -0.0522, -0.0435, -0.0435, -0.0478, -0.0348, -0.0174, -0.0174, 0.0043, 0.0, -0.0174, -0.0043, -0.0043, -0.0043, -0.0043, -0.013, 0.0043, 0.0087, 0.0087, -0.013, 0.0, -0.0217, 0.0261, 0.0217, 0.0261, 0.0261, 0.0435, 0.0217, 0.0261, 0.0261, 0.0304, -0.013, 0.0, -0.013, 0.0174, 0.0217, 0.0348, 0.0087, 0.0087, 0.013, 0.0087, 0.0087, 0.0087, 0.0261, 0.0261, 0.0435, 0.0652, 0.0783, 0.0957, 0.0913, 0.1175, 0.1044, 0.1219, 0.1044, 0.1306, 0.1699, 0.1699, 0.1524, 0.135, 0.135, 0.1393, 0.1568, 0.1917, 0.2223, 0.2223, 0.2528, 0.2354, 0.2572, 0.2572, 0.2354, 0.2441, 0.279, 0.2965, 0.2659, 0.2921, 0.2921, 0.2746, 0.2659, 0.2659, 0.2921, 0.2877, 0.3139, 0.3139, 0.3096, 0.3183, 0.3139, 0.2746, 0.279, 0.2746, 0.2877, 0.279, 0.2877, 0.2746, 0.2485, 0.2615, 0.2266, 0.2179, 0.2703, 0.2092, 0.2004, 0.2354, 0.2572, 0.2528, 0.2572, 0.231, 0.2485, 0.2659, 0.2615, 0.2746, 0.3052, 0.2921, 0.3139, 0.2877, 0.2964, 0.2701, 0.257, 0.2526, 0.2526, 0.2526, 0.2833, 0.3052, 0.3096, 0.3096, 0.3227, 0.3402, 0.3314, 0.3577, 0.3577, 0.3884, 0.4628, 0.4935, 0.4672, 0.476, 0.4716, 0.4804, 0.4979]}
{"unit_id": "orig:1945581656842408403", "ticker_idx": 0, "ts": "2025-07-16T20:29:48+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "upside of $0.30 and $0.45 to CY2026 non-GAAP EPS in Bull #1 and Bull #2 scenarios, respectively", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs analysis flagging $12–17.5B revenue upside and $0.30–0.45 EPS upside to NVDA from expected H20 China export ban lift.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) Implications of H20 China Export Resumption and Our View\n\nOur analysis suggests upside of $0.30 and $0.45 to CY2026 non-GAAP EPS in Bull #1 and Bull #2 scenarios, respectively.\n\nWhile we are not changing our base case estimates, which assume zero China contribution, we provide an updated bridge chart below that includes potential upside scenarios from the expected lift of the ban. In conclusion, all else being equal, we believe the expected lift of the ban this time could add $12.0 billion and $17.5 billion to 2026 revenue and $0.30 and $0.45 to non-GAAP EPS in our Bull #1 and Bull #2 scenarios, respectively.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003909684872381747, "ret_signed_m1d": -0.003909684872381747, "alpha_spy_m1d": -0.0005776334386263793, "alpha_c_m1d": -0.00916412460569449, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009511589505708296, "ret_signed_p1d": 0.009511589505708296, "alpha_spy_p1d": 0.003391899528190745, "alpha_c_p1d": 0.0011114490994763582, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0034428594862712103, "ret_signed_p1w": -0.0034428594862712103, "alpha_spy_p1w": -0.019446981286452614, "alpha_c_p1w": 0.005786934868547422, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.0621461963170844, "ret_signed_p1m": 0.0621461963170844, "alpha_spy_p1m": 0.028936714204513825, "alpha_c_p1m": 0.01845147637661637, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.09897081121285733, "ret_signed_p3m": 0.09897081121285733, "alpha_spy_p3m": 0.03383631036483137, "alpha_c_p3m": -0.07549684249508726, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.07988992467564904, "ret_signed_p6m": 0.07988992467564904, "alpha_spy_p6m": -0.031040142771990187, "alpha_c_p6m": -0.23426679716688548, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4078, -0.4345, -0.423, -0.4007, -0.379, -0.3523, -0.3656, -0.3639, -0.3645, -0.3488, -0.3319, -0.3359, -0.3375, -0.317, -0.3152, -0.3194, -0.2823, -0.2419, -0.2103, -0.2133, -0.21, -0.209, -0.2159, -0.231, -0.2249, -0.2339, -0.2339, -0.2094, -0.2134, -0.1878, -0.2115, -0.1984, -0.176, -0.1719, -0.1832, -0.1731, -0.1678, -0.16, -0.1665, -0.1539, -0.1716, -0.1557, -0.159, -0.1511, -0.1511, -0.1606, -0.1587, -0.137, -0.0996, -0.0954, -0.0795, -0.0781, -0.1054, -0.0824, -0.0702, -0.0702, -0.0766, -0.0663, -0.0495, -0.0424, -0.0376, -0.0426, -0.0039, 0.0, 0.0095, 0.0061, 0.0001, -0.0253, -0.0034, 0.0138, 0.0124, 0.0314, 0.0242, 0.0461, 0.0379, 0.0137, 0.0504, 0.0402, 0.047, 0.0549, 0.0661, 0.0624, 0.0688, 0.0596, 0.0621, 0.053, 0.0621, 0.0249, 0.0235, 0.0211, 0.0386, 0.0493, 0.0607, 0.0597, 0.0514, 0.0164, 0.0164, -0.0034, -0.0044, 0.0017, -0.0254, -0.0179, -0.0036, 0.0348, 0.0339, 0.0377, 0.0373, 0.0205, -0.0062, 0.0285, 0.031, 0.0715, 0.0413, 0.0327, 0.0369, 0.0399, 0.0612, 0.0888, 0.0927, 0.1023, 0.0949, 0.0827, 0.0798, 0.1036, 0.1238, 0.0689, 0.099, 0.0506, 0.0494, 0.061, 0.0692, 0.0658, 0.0572, 0.0521, 0.063, 0.0869, 0.1175, 0.1731, 0.2082, 0.184, 0.1817, 0.2073, 0.1595, 0.1392, 0.0976, 0.098, 0.1616, 0.1272, 0.131, 0.0905, 0.1098, 0.0889, 0.0584, 0.0885, 0.0542, 0.0439, 0.0653, 0.0377, 0.0519, 0.0519, 0.0329, 0.05, 0.0589, 0.048, 0.0702, 0.0645, 0.0829, 0.0795, 0.0725, 0.0559, 0.0214, 0.0288, 0.0372, -0.0024, 0.0163, 0.0563, 0.072, 0.1042, 0.1007, 0.1007, 0.1119, 0.0984, 0.0945, 0.0884, 0.0884, 0.1021, 0.0979, 0.0927, 0.1036, 0.0799]}
{"unit_id": "thread:1964988652649648570", "ticker_idx": 0, "ts": "2025-09-08T09:46:17+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Analysts say booming demand for cloud ASICs makes Broadcom one of the few credible challengers to Nvidia's dominance in AI chips", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Digitimes report of Broadcom's $10B ASIC win + pipeline from Apple, xAI, OpenAI, ByteDance; analysts see AVGO as credible Nvidia challenger.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Broadcom secures US$10B ASIC win, Apple and xAI next in line - Digitimes\n\nBroadcom delivered strong earnings guidance with its latest results, as CEO Hock Tan revealed a US$10 billion ASIC mass-production order from a major new customer outside the core hyperscale cloud service provider segment.\n\nDIGITIMES sources indicate that further orders from ByteDance, Apple, and Elon Musk's xAI are already in the pipeline, a development analysts see as reinforcing Broadcom's position as a credible challenger to Nvidia.\n\nOpenAI deal targets 2026 production\nIndustry sources say OpenAI's custom ASIC is set for mass production in 2026, positioning it as Broadcom's fourth confirmed large-scale ASIC customer. Although neither company has commented, insiders familiar with Broadcom's roadmap confirm the order is genuine, echoing long-standing reports of OpenAI's chip ambitions.\n\nOpenAI's chip project first surfaced in 2024, when the company began developing custom silicon to meet rising AI workloads, mirroring strategies of other hyperscalers. Sources said the process encountered few hurdles, enabling the design to pass specification tests quickly.\n\nWith OpenAI racing to expand its AI infrastructure, a 2026 production start aligns with market expectations, though ramp-up could begin in the second half of the year or slightly earlier.\n\nMajor players queue for custom chips\nBeyond OpenAI, Broadcom has confirmed ASIC orders from Google, Meta, and ByteDance, with Google's demand described as the largest.\n\nAdditional contracts are expected as projects advance, including ByteDance's second-generation ASIC, xAI's cloud-focused chip, and Apple's custom silicon. Production is slated for 2026 and 2027, with ByteDance likely leading in 2026, followed by xAI and Apple in 2027.\n\nGoogle and Meta are developing successive ASIC generations scheduled for mass production between 2027 and 2028. ByteDance and OpenAI are also advancing next-generation designs, with rollout expected after 2028.\n\nStrengthening market position\nThe influx of orders underlines Broadcom's strong ASIC pipeline, explaining Tan's optimism during the earnings call. His upbeat stance contrasted with Marvell's more cautious outlook. Successful execution would solidify Broadcom's leadership in cloud ASICs.\n\nMediaTek has secured AI inference ASIC orders from Google and Meta, while competitors, including Marvell and Alchip, supply AWS, Microsoft, and Intel. Broadcom, however, is viewed as holding a stronger and broader customer roster, anchored by contracts with OpenAI, Google, and xAI.\n\nAnalysts say booming demand for cloud ASICs makes Broadcom one of the few credible challengers to Nvidia's dominance in AI chips.\n\nIndustry insiders caution that while GPUs and ASICs overlap in some applications, the two markets will remain complementary. Nvidia and Broadcom are likely to dominate their respective domains, with fiercer rivalry expected in networking chips than in direct GPU–ASIC competition.\n\n$AVGO\n\n---\n\nhttps://t.co/Hvqw1yjul0", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031129663717725742, "ret_signed_m1d": -0.031129663717725742, "alpha_spy_m1d": -0.028679141619245563, "alpha_c_m1d": -0.020203014982361656, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025979983430521436, "ret_signed_p1d": -0.025979983430521436, "alpha_spy_p1d": -0.028291904128131318, "alpha_c_p1d": -0.0303304905991022, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.05334865965704916, "ret_signed_p1w": 0.05334865965704916, "alpha_spy_p1w": 0.03473053554589822, "alpha_c_p1w": 0.017431888348674818, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": -0.025064755049650667, "ret_signed_p1m": -0.025064755049650667, "alpha_spy_p1m": -0.05919561089954517, "alpha_c_p1m": -0.16175038093886418, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.1042466170647871, "ret_signed_p3m": 0.1042466170647871, "alpha_spy_p3m": 0.04651586067396751, "alpha_c_p3m": -0.11523275297268976, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": -0.08873355372693392, "ret_signed_p6m": -0.08873355372693392, "alpha_spy_p6m": -0.14329658980532434, "alpha_c_p6m": -0.41166909921848605, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.27, -0.2609, -0.2822, -0.2724, -0.2802, -0.2748, -0.2748, -0.2768, -0.2658, -0.2369, -0.2343, -0.2184, -0.2207, -0.2025, -0.2341, -0.2192, -0.2039, -0.2039, -0.2068, -0.2137, -0.196, -0.2032, -0.2062, -0.2027, -0.1872, -0.1876, -0.1713, -0.1803, -0.1662, -0.194, -0.1793, -0.1647, -0.1605, -0.1486, -0.1395, -0.1245, -0.1503, -0.1649, -0.1387, -0.1525, -0.1272, -0.1212, -0.1177, -0.1208, -0.095, -0.1058, -0.0996, -0.1137, -0.1154, -0.1468, -0.1576, -0.1622, -0.1494, -0.1488, -0.1378, -0.1313, -0.107, -0.1396, -0.1396, -0.1372, -0.1252, -0.1144, -0.0311, 0.0, -0.026, 0.0692, 0.0404, 0.0411, 0.0533, 0.0415, 0.0015, -0.0009, -0.0021, -0.0182, -0.0177, -0.0167, -0.026, -0.0305, -0.0497, -0.0439, -0.0338, -0.0199, -0.0194, -0.0277, -0.0251, 0.0013, -0.0001, -0.0592, 0.0337, -0.0027, 0.0182, 0.0263, 0.0124, 0.0121, -0.007, -0.0138, -0.0022, 0.0263, 0.0492, 0.0809, 0.1186, 0.091, 0.0712, 0.0507, 0.0199, 0.0403, 0.0305, 0.0127, 0.0386, 0.02, 0.0294, -0.0147, -0.0075, -0.007, -0.0132, 0.0271, 0.0051, -0.0141, 0.0953, 0.1158, 0.1522, 0.1522, 0.1678, 0.1189, 0.1058, 0.103, 0.1042, 0.1309, 0.1624, 0.1775, 0.1968, 0.1777, 0.0431, -0.0152, -0.0109, -0.0552, -0.044, -0.0136, -0.0086, 0.0143, 0.0169, 0.0169, 0.0224, 0.0145, 0.0158, 0.0049, 0.0049, 0.0094, -0.0028, -0.0018, -0.0026, -0.0346, 0.0017, 0.0227, 0.0296, -0.0131, -0.004, 0.0212, 0.0212, -0.0343, -0.0453, -0.0549, -0.0707, -0.0568, -0.0337, -0.0324, -0.0397, -0.038, -0.0386, -0.0699, -0.1055, -0.0984, -0.0333, -0.0013, -0.0115, -0.0048, -0.0384, -0.0558, -0.0558, -0.0344, -0.0316, -0.0302, -0.0341, -0.0408, -0.0549, -0.0351, -0.0659, -0.0722, -0.0743, -0.0887]}
{"unit_id": "thread:1971408271069376937", "ticker_idx": 0, "ts": "2025-09-26T02:55:33+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron has announced a price increase of over 20% across DRAM and NAND", "tweet_type": "analysis", "timeline": "quarters", "summary": "Micron announcing 20%+ DRAM/NAND price hikes; DDR5 rising fastest on CSP demand, NAND rebound driven by enterprise SSD orders — bullish for memory pricing into Q4.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Just in: Micron Notifies Memory Price Hike\n\nMicron has announced a price increase of over 20% across DRAM and NAND.\n\nIts strategy is to “focus on the big accounts (抓大放小),” meaning higher hikes are being applied to server RDIMM DRAM and enterprise SSDs, both high-value products.\n   \n •    DDR5 average price increase: 15–20%\n •    DDR4 average price increase: 10–15%\n\nIn other words, DDR5 is rising more sharply than DDR4, reflecting Micron’s confidence in CSP customers’ strong demand for DDR5 in 2026.\n\n⸻\n\nNAND Flash — Rebound Driven by CSP SSD Demand\n\nNAND price hikes have also begun.\n    \n• While consumer demand remains weak, CSPs have placed large short-term orders for high-capacity SSDs, triggering a sharp rebound in the market.\n    \n• Major NAND suppliers are allocating capacity toward high-margin SSDs, making it harder for module makers to secure wafer supply.\n\nCurrently, the price increase is around 10%, but cumulative 4Q gains could expand to as much as 20%.\n\n---\n\nhttps://t.co/EnBPf4dQNn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0027977303347074267, "ret_signed_m1d": -0.0027977303347074267, "alpha_spy_m1d": 0.0028987779141415038, "alpha_c_m1d": -0.0004361076591725688, "bench_spy_m1d": -0.0056965082488489305, "bench_c_m1d": -0.002361622675534858, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.042156590115308035, "ret_signed_p1d": 0.042156590115308035, "alpha_spy_p1d": 0.03934622367416818, "alpha_c_p1d": 0.03954650574772445, "bench_spy_p1d": 0.0028103664411398555, "bench_c_p1d": 0.002610084367583587, "ret_p1w": 0.19506332835661078, "ret_signed_p1w": 0.19506332835661078, "alpha_spy_p1w": 0.18389717421708696, "alpha_c_p1w": 0.1488885492539893, "bench_spy_p1w": 0.01116615413952382, "bench_c_p1w": 0.046174779102621466, "ret_p1m": 0.4003803374182877, "ret_signed_p1m": 0.4003803374182877, "alpha_spy_p1m": 0.3649931484821962, "alpha_c_p1m": 0.28192907942877476, "bench_spy_p1m": 0.035387188936091496, "bench_c_p1m": 0.11845125798951295, "ret_p3m": 0.8239938654753625, "ret_signed_p3m": 0.8239938654753625, "alpha_spy_p3m": 0.7777578143569477, "alpha_c_p3m": 0.688877505595153, "bench_spy_p3m": 0.04623605111841478, "bench_c_p3m": 0.13511635988020942, "ret_p6m": 1.5737053379964117, "ret_signed_p6m": 1.5737053379964117, "alpha_spy_p6m": 1.5777975787843714, "alpha_c_p6m": 1.3538379529976954, "bench_spy_p6m": -0.004092240787959889, "bench_c_p6m": 0.21986738499871628, "price_path": [-0.232, -0.2266, -0.2232, -0.2232, -0.2375, -0.2089, -0.2227, -0.2172, -0.2082, -0.2458, -0.2363, -0.2597, -0.2798, -0.2727, -0.28, -0.3055, -0.3016, -0.2896, -0.2926, -0.2926, -0.2881, -0.2704, -0.306, -0.3331, -0.3147, -0.3065, -0.3083, -0.2887, -0.244, -0.2133, -0.1877, -0.2098, -0.2033, -0.2314, -0.2144, -0.2239, -0.2547, -0.2638, -0.2517, -0.2597, -0.2592, -0.2513, -0.2243, -0.2433, -0.2433, -0.2466, -0.2451, -0.2102, -0.1647, -0.1641, -0.1401, -0.1098, -0.0426, -0.0003, 0.0032, 0.0099, 0.0173, 0.0739, 0.0347, 0.0467, 0.0581, 0.0282, -0.0028, 0.0, 0.0422, 0.0639, 0.1582, 0.1684, 0.1951, 0.215, 0.1814, 0.2505, 0.2237, 0.1554, 0.2265, 0.1902, 0.2212, 0.2886, 0.2876, 0.3156, 0.2871, 0.2628, 0.3152, 0.3935, 0.4004, 0.4119, 0.4419, 0.4253, 0.4237, 0.4933, 0.3872, 0.5111, 0.5164, 0.5138, 0.6116, 0.5341, 0.5582, 0.5076, 0.5704, 0.5394, 0.4538, 0.4374, 0.2812, 0.3194, 0.4247, 0.4286, 0.465, 0.465, 0.5046, 0.5299, 0.5237, 0.4898, 0.4421, 0.5093, 0.571, 0.606, 0.6778, 0.6444, 0.5342, 0.5111, 0.4793, 0.4349, 0.5814, 0.6919, 0.7598, 0.7578, 0.824, 0.824, 0.812, 0.8737, 0.8626, 0.8166, 0.8166, 1.0077, 0.9868, 1.1859, 1.1613, 1.0815, 1.1965, 1.2015, 1.1522, 1.1218, 1.1427, 1.3089, 1.3089, 1.3232, 1.4767, 1.5306, 1.5438, 1.4766, 1.6112, 1.7706, 1.7738, 1.6407, 1.7866, 1.6698, 1.4149, 1.4371, 1.5122, 1.441, 1.3758, 1.6118, 1.6349, 1.6202, 1.6202, 1.5446, 1.6794, 1.6565, 1.7253, 1.6795, 1.6607, 1.7306, 1.6451, 1.6248, 1.6267, 1.4167, 1.5509, 1.5272, 1.357, 1.478, 1.5658, 1.665, 1.5801, 1.7123, 1.8121, 1.9387, 1.9389, 1.8278, 1.6918, 1.5737]}
{"unit_id": "thread:1971408271069376937", "ticker_idx": 1, "ts": "2025-09-26T02:55:33+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DDR5 average price increase: 15–20%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Micron announcing 20%+ DRAM/NAND price hikes; DDR5 rising fastest on CSP demand, NAND rebound driven by enterprise SSD orders — bullish for memory pricing into Q4.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Just in: Micron Notifies Memory Price Hike\n\nMicron has announced a price increase of over 20% across DRAM and NAND.\n\nIts strategy is to “focus on the big accounts (抓大放小),” meaning higher hikes are being applied to server RDIMM DRAM and enterprise SSDs, both high-value products.\n   \n •    DDR5 average price increase: 15–20%\n •    DDR4 average price increase: 10–15%\n\nIn other words, DDR5 is rising more sharply than DDR4, reflecting Micron’s confidence in CSP customers’ strong demand for DDR5 in 2026.\n\n⸻\n\nNAND Flash — Rebound Driven by CSP SSD Demand\n\nNAND price hikes have also begun.\n    \n• While consumer demand remains weak, CSPs have placed large short-term orders for high-capacity SSDs, triggering a sharp rebound in the market.\n    \n• Major NAND suppliers are allocating capacity toward high-margin SSDs, making it harder for module makers to secure wafer supply.\n\nCurrently, the price increase is around 10%, but cumulative 4Q gains could expand to as much as 20%.\n\n---\n\nhttps://t.co/EnBPf4dQNn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030083699527090224, "ret_signed_m1d": 0.030083699527090224, "alpha_spy_m1d": 0.035780207775939155, "alpha_c_m1d": 0.03244532220262508, "bench_spy_m1d": -0.0056965082488489305, "bench_c_m1d": -0.002361622675534858, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03153926371720872, "ret_signed_p1d": 0.03153926371720872, "alpha_spy_p1d": 0.028728897276068865, "alpha_c_p1d": 0.028929179349625134, "bench_spy_p1d": 0.0028103664411398555, "bench_c_p1d": 0.002610084367583587, "ret_p1w": 0.15087860828307575, "ret_signed_p1w": 0.15087860828307575, "alpha_spy_p1w": 0.13971245414355193, "alpha_c_p1w": 0.10470382918045429, "bench_spy_p1w": 0.01116615413952382, "bench_c_p1w": 0.046174779102621466, "ret_p1m": 0.4067431142068836, "ret_signed_p1m": 0.4067431142068836, "alpha_spy_p1m": 0.3713559252707921, "alpha_c_p1m": 0.28829185621737063, "bench_spy_p1m": 0.035387188936091496, "bench_c_p1m": 0.11845125798951295, "ret_p3m": 0.6374430594229074, "ret_signed_p3m": 0.6374430594229074, "alpha_spy_p3m": 0.5912070083044926, "alpha_c_p3m": 0.5023266995426979, "bench_spy_p3m": 0.04623605111841478, "bench_c_p3m": 0.13511635988020942, "ret_p6m": 1.5369547263630805, "ret_signed_p6m": 1.5369547263630805, "alpha_spy_p6m": 1.5410469671510403, "alpha_c_p6m": 1.3170873413643642, "bench_spy_p6m": -0.004092240787959889, "bench_c_p6m": 0.21986738499871628, "price_path": [-0.2207, -0.2229, -0.2103, -0.2202, -0.2309, -0.2117, -0.2213, -0.2012, -0.1943, -0.2018, -0.1953, -0.2015, -0.2265, -0.223, -0.2192, -0.2388, -0.2354, -0.2325, -0.2374, -0.2233, -0.2205, -0.2056, -0.2124, -0.2468, -0.2375, -0.2285, -0.2384, -0.2216, -0.207, -0.1896, -0.1786, -0.1739, -0.1744, -0.1838, -0.1934, -0.2011, -0.2167, -0.2297, -0.2166, -0.2104, -0.2131, -0.2107, -0.1969, -0.2024, -0.2237, -0.2143, -0.209, -0.1932, -0.1725, -0.1665, -0.142, -0.1116, -0.083, -0.0396, -0.0316, -0.0009, -0.0176, 0.0314, 0.0184, 0.0307, 0.0492, 0.0386, 0.0301, 0.0, 0.0315, 0.0361, 0.0884, 0.142, 0.1509, 0.1575, 0.1463, 0.1694, 0.1604, 0.1886, 0.1949, 0.1725, 0.2074, 0.2705, 0.2838, 0.3138, 0.2954, 0.2942, 0.3003, 0.3668, 0.4067, 0.3867, 0.4373, 0.4562, 0.4604, 0.5585, 0.4645, 0.4816, 0.4916, 0.4726, 0.5419, 0.5405, 0.545, 0.522, 0.4689, 0.5178, 0.4423, 0.4237, 0.3971, 0.3369, 0.3787, 0.3894, 0.4206, 0.4436, 0.4309, 0.4485, 0.4767, 0.4638, 0.4404, 0.4781, 0.5358, 0.5321, 0.5753, 0.5395, 0.5152, 0.4741, 0.4317, 0.4582, 0.5068, 0.5335, 0.6057, 0.613, 0.6374, 0.6374, 0.6681, 0.7417, 0.7505, 0.7352, 0.7352, 0.8593, 0.91, 1.0093, 1.0255, 1.0039, 1.0311, 1.0369, 1.0047, 1.0095, 1.0379, 1.1205, 1.13, 1.0974, 1.163, 1.2071, 1.2226, 1.1695, 1.3077, 1.4131, 1.4272, 1.4296, 1.3591, 1.4656, 1.3801, 1.2905, 1.3097, 1.365, 1.33, 1.4009, 1.4799, 1.4776, 1.4776, 1.4524, 1.4973, 1.5391, 1.617, 1.6154, 1.6909, 1.7413, 1.8536, 1.8031, 1.8037, 1.5261, 1.389, 1.5508, 1.4634, 1.357, 1.5458, 1.6042, 1.5426, 1.549, 1.6668, 1.7261, 1.8705, 1.7585, 1.7027, 1.537]}
{"unit_id": "thread:1971408271069376937", "ticker_idx": 2, "ts": "2025-09-26T02:55:33+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND price hikes have also begun. Currently, the price increase is around 10%, but cumulative 4Q gains could expand to as much as 20%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Micron announcing 20%+ DRAM/NAND price hikes; DDR5 rising fastest on CSP demand, NAND rebound driven by enterprise SSD orders — bullish for memory pricing into Q4.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Just in: Micron Notifies Memory Price Hike\n\nMicron has announced a price increase of over 20% across DRAM and NAND.\n\nIts strategy is to “focus on the big accounts (抓大放小),” meaning higher hikes are being applied to server RDIMM DRAM and enterprise SSDs, both high-value products.\n   \n •    DDR5 average price increase: 15–20%\n •    DDR4 average price increase: 10–15%\n\nIn other words, DDR5 is rising more sharply than DDR4, reflecting Micron’s confidence in CSP customers’ strong demand for DDR5 in 2026.\n\n⸻\n\nNAND Flash — Rebound Driven by CSP SSD Demand\n\nNAND price hikes have also begun.\n    \n• While consumer demand remains weak, CSPs have placed large short-term orders for high-capacity SSDs, triggering a sharp rebound in the market.\n    \n• Major NAND suppliers are allocating capacity toward high-margin SSDs, making it harder for module makers to secure wafer supply.\n\nCurrently, the price increase is around 10%, but cumulative 4Q gains could expand to as much as 20%.\n\n---\n\nhttps://t.co/EnBPf4dQNn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03474797153484612, "ret_signed_m1d": 0.03474797153484612, "alpha_spy_m1d": 0.04044447978369505, "alpha_c_m1d": 0.03710959421038098, "bench_spy_m1d": -0.0056965082488489305, "bench_c_m1d": -0.002361622675534858, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06539676983857748, "ret_signed_p1d": 0.06539676983857748, "alpha_spy_p1d": 0.06258640339743762, "alpha_c_p1d": 0.06278668547099389, "bench_spy_p1d": 0.0028103664411398555, "bench_c_p1d": 0.002610084367583587, "ret_p1w": 0.23619158356305436, "ret_signed_p1w": 0.23619158356305436, "alpha_spy_p1w": 0.22502542942353054, "alpha_c_p1w": 0.1900168044604329, "bench_spy_p1w": 0.01116615413952382, "bench_c_p1w": 0.046174779102621466, "ret_p1m": 0.6539095335237557, "ret_signed_p1m": 0.6539095335237557, "alpha_spy_p1m": 0.6185223445876642, "alpha_c_p1m": 0.5354582755342427, "bench_spy_p1m": 0.035387188936091496, "bench_c_p1m": 0.11845125798951295, "ret_p3m": 0.9800514442833329, "ret_signed_p3m": 0.9800514442833329, "alpha_spy_p3m": 0.9338153931649181, "alpha_c_p3m": 0.8449350844031235, "bench_spy_p3m": 0.04623605111841478, "bench_c_p3m": 0.13511635988020942, "ret_p6m": 2.9456079115114884, "ret_signed_p6m": 2.9456079115114884, "alpha_spy_p6m": 2.949700152299448, "alpha_c_p6m": 2.725740526512772, "bench_spy_p6m": -0.004092240787959889, "bench_c_p6m": 0.21986738499871628, "price_path": [-0.3323, -0.3288, -0.3146, -0.3282, -0.3335, -0.3157, -0.3144, -0.3077, -0.3084, -0.3227, -0.3177, -0.3242, -0.3405, -0.3398, -0.3387, -0.349, -0.3373, -0.3361, -0.3431, -0.3361, -0.3334, -0.3225, -0.3263, -0.3527, -0.3465, -0.3429, -0.3502, -0.3445, -0.3253, -0.3168, -0.2916, -0.2902, -0.2923, -0.3021, -0.3112, -0.3163, -0.3313, -0.3392, -0.3251, -0.3211, -0.3196, -0.3168, -0.301, -0.2952, -0.3044, -0.3044, -0.2969, -0.2677, -0.2229, -0.2114, -0.1955, -0.1562, -0.094, -0.0444, -0.0314, 0.0021, -0.015, 0.0282, 0.0277, 0.0497, 0.068, 0.054, 0.0347, 0.0, 0.0654, 0.0746, 0.1165, 0.187, 0.2362, 0.2316, 0.2054, 0.2408, 0.2464, 0.2342, 0.2745, 0.2387, 0.3092, 0.3725, 0.3575, 0.4185, 0.405, 0.4118, 0.4596, 0.603, 0.6539, 0.6203, 0.742, 0.773, 0.7793, 0.854, 0.769, 0.8147, 0.846, 0.9246, 1.0831, 1.084, 1.1125, 1.0075, 0.8609, 0.9663, 0.8367, 0.8538, 0.7576, 0.671, 0.751, 0.7361, 0.6768, 0.7208, 0.7464, 0.7061, 0.7283, 0.6887, 0.7163, 0.7865, 0.8477, 0.8205, 0.8556, 0.8586, 0.7837, 0.7188, 0.6854, 0.7241, 0.7888, 0.8344, 0.9206, 0.925, 0.9801, 0.9887, 1.0352, 1.0338, 1.0198, 1.0048, 1.0048, 1.1573, 1.2269, 1.4535, 1.5208, 1.4828, 1.5733, 1.6013, 1.6283, 1.6086, 1.6858, 1.7953, 1.8219, 1.8835, 2.0809, 2.176, 2.0982, 2.0653, 2.2156, 2.3966, 2.4427, 2.6164, 2.6209, 2.8575, 2.5829, 2.4496, 2.4861, 2.4805, 2.3709, 2.5323, 2.7495, 2.8164, 2.7916, 2.7056, 2.7096, 2.7695, 2.8443, 2.8774, 2.9429, 2.9238, 3.0212, 2.9554, 2.9362, 2.5992, 2.5428, 2.619, 2.473, 2.4471, 2.6925, 2.8852, 2.7667, 2.8505, 3.0782, 3.1021, 3.3388, 3.2626, 3.1007, 2.9456]}
{"unit_id": "quote:1966653148358955109", "ticker_idx": 0, "ts": "2025-09-13T00:00:23+00:00", "symbol": "BESI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Why I'm Bullish on BESI", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Author signals bullish thesis on BESI in context of HBM hybrid bonding equipment dynamics.", "resolved_tickers": ["BESI.AS"], "resolution_method": "exchange_ticker", "resolution_reason": "BE Semiconductor Industries Amsterdam BESI.AS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Actually, I was thinking of turning this into a paid report… but I’ll just keep it light in note format and release it for free.\n\nI hope this will be helpful to everyone.", "reply_text": null, "parent_text": null, "quoted_text": "Why is LG developing CMP equipment? feat. Why I’m Bullish on BESI\n\nEveryone’s probably heard that LG Electronics is independently developing an HBM hybrid bonder. What’s interesting is that they’re also said to be developing CMP equipment for HBM hybrid bonding.\n\nWhy might that", "parent_author": null, "ret_m1d": -0.05332202492433302, "ret_signed_m1d": -0.05332202492433302, "alpha_spy_m1d": -0.04802636000084637, "alpha_c_m1d": -0.04404378691155164, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013118928428609844, "ret_signed_p1d": -0.013118928428609844, "alpha_spy_p1d": -0.011742135643328155, "alpha_c_p1d": -0.013249084271795053, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.03935678528582942, "ret_signed_p1w": 0.03935678528582942, "alpha_spy_p1w": 0.027586943911767303, "alpha_c_p1w": -0.009476072586680706, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.18916628078843667, "ret_signed_p1m": 0.18916628078843667, "alpha_spy_p1m": 0.18439104595925793, "alpha_c_p1m": 0.10344829342033024, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.15869647799636333, "ret_signed_p3m": 0.15869647799636333, "alpha_spy_p3m": 0.11304624632208249, "alpha_c_p3m": -0.048745625824348604, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 0.4777825762560579, "ret_signed_p6m": 0.4777825762560579, "alpha_spy_p6m": 0.4472882412419439, "alpha_c_p6m": 0.18023954186093127, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [0.0643, 0.0554, 0.0423, 0.0639, 0.091, 0.0935, 0.0986, 0.1041, 0.0753, 0.0339, 0.0432, 0.0322, 0.0245, 0.0275, 0.0343, 0.0322, 0.0783, 0.0762, 0.055, 0.069, 0.0152, 0.0791, 0.0829, 0.1016, 0.0647, 0.0466, 0.0749, -0.0089, 0.0394, 0.0157, 0.036, 0.008, -0.0085, 0.0021, -0.0241, -0.0284, 0.0152, 0.022, 0.0288, 0.0474, 0.0453, 0.0381, 0.0034, -0.003, 0.0106, -0.0089, -0.0229, 0.0072, 0.0085, 0.0216, 0.014, 0.0135, -0.0258, -0.0224, -0.0952, -0.1045, -0.0931, -0.08, -0.0618, -0.0567, -0.0491, -0.0381, -0.0533, 0.0, -0.0131, -0.0212, 0.0563, 0.0284, 0.0394, 0.0741, 0.088, 0.0724, 0.036, 0.0698, 0.0736, 0.0817, 0.1303, 0.1045, 0.2421, 0.2425, 0.2387, 0.2306, 0.1875, 0.2082, 0.1892, 0.2315, 0.2281, 0.1993, 0.2289, 0.2213, 0.1722, 0.2209, 0.2446, 0.2526, 0.2408, 0.2497, 0.2446, 0.2493, 0.234, 0.2065, 0.1976, 0.1659, 0.1354, 0.1549, 0.1663, 0.1456, 0.1325, 0.1121, 0.102, 0.0762, 0.0927, 0.1007, 0.0394, 0.0529, 0.0546, 0.0969, 0.0994, 0.0994, 0.0982, 0.1117, 0.1507, 0.1638, 0.1811, 0.2429, 0.2129, 0.1887, 0.1587, 0.1181, 0.121, 0.1308, 0.0944, 0.1126, 0.1066, 0.1105, 0.1164, 0.1134, 0.1134, 0.1134, 0.1181, 0.1265, 0.132, 0.132, 0.262, 0.3013, 0.3618, 0.3441, 0.2789, 0.2789, 0.3724, 0.4152, 0.3669, 0.4697, 0.4634, 0.4287, 0.4634, 0.4731, 0.4812, 0.4816, 0.4799, 0.4875, 0.4342, 0.3758, 0.3923, 0.3931, 0.3555, 0.3335, 0.3606, 0.4008, 0.4168, 0.4308, 0.4465, 0.4278, 0.4913, 0.5108, 0.5243, 0.5827, 0.4642, 0.5616, 0.5882, 0.6276, 0.6725, 0.5988, 0.603, 0.5954, 0.5349, 0.6166, 0.5967, 0.3229, 0.4075, 0.4778]}
{"unit_id": "orig:1960698305157194041", "ticker_idx": 0, "ts": "2025-08-27T13:37:58+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "July quarter results are expected to beat Wall Street expectations (revenue around $46–47bn), with October quarter guidance likely coming in at $53–54bn or higher, modestly above consensus", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares JPM NVDA preview: July rev $46-47B beats, Oct guide $53-54B+; names SK Hynix, TSMC, Hon Hai as AI supply-chain winners.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) NVIDIA (NVDA) Earnings Preview\n\nIndustry Commentary – We (Harlan and team) believe near-term AI fundamentals remain solid, supported by strong hyperscaler CAPEX spending. This trend was clearly demonstrated during the 2Q25 earnings season, where cloud/hyperscaler companies raised CAPEX guidance, and other AI beneficiaries (e.g., MTSI, ALAB, AMD) reported strong results and outlooks. In cyclical sectors such as telecom/service providers and enterprises, conditions are also showing gradual improvement.\n\nNVIDIA (NVDA) – Thanks to the ramp-up of GB200 production and the start of the GB300 program, July quarter results are expected to beat Wall Street expectations (revenue around $46–47bn), with October quarter guidance likely coming in at $53–54bn or higher, modestly above consensus. NVIDIA is scheduled to report July quarter results on Wednesday (Aug 27, local time), and revenue is likely to exceed guidance ($45bn) and land in the $46–47bn range. This upside is almost entirely driven by the strong ramp of GB200 rack shipments.\n\nFor perspective: April quarter revenue ($44bn) included ~$4.5bn from H20. Excluding that, the comparable base was $39.5bn. Thus, July quarter revenue adds $6.5–7.5bn on a like-for-like basis (versus a reported increase of only $2–3bn), again driven by GB200.\n\nFor October quarter, we expect NVIDIA to guide revenue to $53–54bn or above. This would modestly top consensus (consistent with recent guidance trends – see Figure 1) and sit above our own estimate of $51bn, as GB200 shipments continue to accelerate (estimated 8,000–9,000 racks). Total Blackwell rack shipments could reach ~12,000 in fiscal 4Q, implying ~28,000–30,000 racks for the year. Like AMD, NVIDIA is unlikely to include H20 (or B20/30) related revenue in October guidance, due to (1) delays and uncertainty around US government licensing approvals, (2) NVIDIA’s need to reassess Chinese customer orders, and (3) China’s government discouraging local adoption of NVIDIA GPUs due to “backdoor” concerns.\n\nThat said, potential H20 upside remains: NVIDIA’s existing $1.9bn inventory could convert into $5–6bn of revenue. Every additional 100k units shipped could yield at least $1bn in extra sales, with ~$3–4bn per quarter possible across 2–3 quarters. Gross margin for October is expected to improve slightly (J.P. Morgan estimate: 73%) on supply chain optimization, moving toward the mid-70% range by year-end.\n\nNVIDIA (NVDA) Buyside Dynamics (Buyside Bars) – Josh Meyers\n\nInvestor Sentiment: Positioning in NVIDIA remains relatively high, based on trading flows and client conversations. Even long-only investors who have trimmed positions remain optimistic. Our survey results mirror this elevated expectation. Investors anticipate ~60% YoY revenue growth this quarter, and ~55% YoY growth for next quarter guidance. Respondents expect both July revenue and October guidance to come in ~3.5% above consensus (roughly $2bn higher than current guidance, as usual).\nSurvey respondents see 2026 EPS >$6.5 (with $7 commonly cited; Harlan Sur views this as easily achievable). Optimism is driven by smooth GB200 shipments and the ramp of GB300 in 2H.\n\nRemaining uncertainties:\n\n1. Order volatility – hints of GB300 system order pull-ins (e.g., Quanta).\n\n2. H20 shipment guidance – balancing Chinese demand vs political risk.\n\n3. Networking supply – parts shortages.\n\nOptimists cite Hon Hai’s server strength as support against order volatility concerns (Gokul expects at least 6,000 servers in 3Q). The second and third issues are already well known, making any surprises easier to justify. Notably, margin expectations showed little dispersion in the survey.\n\nWhen shared with Asia tech specialist Duncan Wagner, he emphasized that Asian investors seem less optimistic than these survey results suggest. Another point of discussion: NVIDIA’s stock has shown a recent pattern of strong gains in 1H (2023, 2024, 2025) followed by flat trading in 2H (2023, 2024).\n\nPositioning score (1 = max sell/underweight, 10 = max buy/overweight): 8\nExpected move: 5.8%\n\nNVIDIA (NVDA) and Related Plays (Laterals) – Ashwin Panjabi\nNVIDIA could be the spark that unleashes “animal spirits” this week. Conditions are perfect. Last week’s Altman comments and the M.I.T. report reignited the recurring six-monthly concerns over AI CAPEX sustainability. Gokul Hariharan and his team produced an excellent chart showing the massive capacity for US CSPs to keep spending through 2027.\n\nIf EBITDA and OCF grow 10% and are fully reinvested, 2027 CAPEX could grow 76%… This is not our base case, but it illustrates the upside potential.\nSpending appetite is also enormous. The Big 4 already monetize AI based on FCF this year, with minimal inclusion of reasoning models (whose token usage is 10–30x higher than chatbots). Video-related GenAI, followed by agentic AI, represents the next wave of growth. Appetite extends to startups, sovereign wealth funds/governments, and enterprises, reinforcing confidence in the 2028 outlook.\n\nMuch of this demand remains latent due to severe supply constraints, driving price increases across the stack: advanced packaging, ABF substrates, power ICs, MLCCs, and commodity DRAM.\n\nWinners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0009361007650832143, "ret_signed_m1d": 0.0009361007650832143, "alpha_spy_m1d": 0.0032094342266061338, "alpha_c_m1d": 0.0039932407749171395, "bench_spy_m1d": -0.0022733334615229195, "bench_c_m1d": -0.003057140009833925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00787445711365986, "ret_signed_p1d": -0.00787445711365986, "alpha_spy_p1d": -0.011415819845802644, "alpha_c_p1d": -0.011973098061379361, "bench_spy_p1d": 0.003541362732142783, "bench_c_p1d": 0.004098640947719501, "ret_p1w": -0.060462495461940824, "ret_signed_p1w": -0.060462495461940824, "alpha_spy_p1w": -0.05599312301886983, "alpha_c_p1w": -0.02273490303014669, "bench_spy_p1w": -0.004469372443070996, "bench_c_p1w": -0.03772759243179413, "ret_p1m": -0.021475608772236532, "ret_signed_p1m": -0.021475608772236532, "alpha_spy_p1m": -0.041957789253324274, "alpha_c_p1m": -0.10008882358543025, "bench_spy_p1m": 0.02048218048108774, "bench_c_p1m": 0.07861321481319372, "ret_p3m": 0.0052880154899086484, "ret_signed_p3m": 0.0052880154899086484, "alpha_spy_p3m": -0.03175631594044481, "alpha_c_p3m": -0.1339984139287338, "bench_spy_p3m": 0.03704433143035346, "bench_c_p3m": 0.13928642941864244, "ret_p6m": 0.034807586290355186, "ret_signed_p6m": 0.034807586290355186, "alpha_spy_p6m": -0.029797843475520613, "alpha_c_p6m": -0.3475629200321324, "bench_spy_p6m": 0.0646054297658758, "bench_c_p6m": 0.3823705063224876, "price_path": [-0.2559, -0.2436, -0.2224, -0.2186, -0.2292, -0.2197, -0.2146, -0.2073, -0.2135, -0.2015, -0.2182, -0.2032, -0.2064, -0.1989, -0.1989, -0.2079, -0.2061, -0.1856, -0.1503, -0.1464, -0.1313, -0.13, -0.1558, -0.1341, -0.1226, -0.1226, -0.1286, -0.1189, -0.1031, -0.0964, -0.0919, -0.0965, -0.06, -0.0563, -0.0474, -0.0506, -0.0563, -0.0802, -0.0596, -0.0433, -0.0446, -0.0267, -0.0335, -0.0128, -0.0205, -0.0434, -0.0088, -0.0184, -0.012, -0.0046, 0.0061, 0.0025, 0.0086, -0.0001, 0.0023, -0.0063, 0.0023, -0.0328, -0.0341, -0.0365, -0.0199, -0.0099, 0.0009, 0.0, -0.0079, -0.0409, -0.0409, -0.0596, -0.0605, -0.0547, -0.0803, -0.0732, -0.0597, -0.0235, -0.0243, -0.0208, -0.0211, -0.037, -0.0622, -0.0295, -0.0271, 0.0111, -0.0174, -0.0254, -0.0215, -0.0187, 0.0014, 0.0275, 0.0311, 0.0402, 0.0332, 0.0218, 0.019, 0.0414, 0.0605, 0.0086, 0.0371, -0.0086, -0.0097, 0.0012, 0.009, 0.0058, -0.0024, -0.0072, 0.0031, 0.0257, 0.0545, 0.1071, 0.1402, 0.1173, 0.1151, 0.1393, 0.0942, 0.075, 0.0357, 0.0361, 0.0962, 0.0637, 0.0672, 0.029, 0.0473, 0.0276, -0.0013, 0.0272, -0.0052, -0.0149, 0.0053, -0.0208, -0.0073, -0.0073, -0.0253, -0.0092, -0.0007, -0.011, 0.0099, 0.0046, 0.0219, 0.0187, 0.0121, -0.0036, -0.0361, -0.0291, -0.0213, -0.0586, -0.041, -0.0032, 0.0116, 0.042, 0.0387, 0.0387, 0.0493, 0.0366, 0.0328, 0.0271, 0.0271, 0.04, 0.036, 0.0312, 0.0415, 0.0191, 0.0181, 0.0185, 0.0233, 0.0086, 0.0301, 0.0256, 0.0256, -0.0193, 0.0096, 0.018, 0.0335, 0.0269, 0.0382, 0.0547, 0.0602, 0.0526, 0.0222, -0.0068, -0.0407, -0.0534, 0.0211, 0.0466, 0.0383, 0.0466, 0.0295, 0.0068, 0.0068, 0.0187, 0.0352, 0.0348]}
{"unit_id": "orig:1960698305157194041", "ticker_idx": 1, "ts": "2025-08-27T13:37:58+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Winners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares JPM NVDA preview: July rev $46-47B beats, Oct guide $53-54B+; names SK Hynix, TSMC, Hon Hai as AI supply-chain winners.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) NVIDIA (NVDA) Earnings Preview\n\nIndustry Commentary – We (Harlan and team) believe near-term AI fundamentals remain solid, supported by strong hyperscaler CAPEX spending. This trend was clearly demonstrated during the 2Q25 earnings season, where cloud/hyperscaler companies raised CAPEX guidance, and other AI beneficiaries (e.g., MTSI, ALAB, AMD) reported strong results and outlooks. In cyclical sectors such as telecom/service providers and enterprises, conditions are also showing gradual improvement.\n\nNVIDIA (NVDA) – Thanks to the ramp-up of GB200 production and the start of the GB300 program, July quarter results are expected to beat Wall Street expectations (revenue around $46–47bn), with October quarter guidance likely coming in at $53–54bn or higher, modestly above consensus. NVIDIA is scheduled to report July quarter results on Wednesday (Aug 27, local time), and revenue is likely to exceed guidance ($45bn) and land in the $46–47bn range. This upside is almost entirely driven by the strong ramp of GB200 rack shipments.\n\nFor perspective: April quarter revenue ($44bn) included ~$4.5bn from H20. Excluding that, the comparable base was $39.5bn. Thus, July quarter revenue adds $6.5–7.5bn on a like-for-like basis (versus a reported increase of only $2–3bn), again driven by GB200.\n\nFor October quarter, we expect NVIDIA to guide revenue to $53–54bn or above. This would modestly top consensus (consistent with recent guidance trends – see Figure 1) and sit above our own estimate of $51bn, as GB200 shipments continue to accelerate (estimated 8,000–9,000 racks). Total Blackwell rack shipments could reach ~12,000 in fiscal 4Q, implying ~28,000–30,000 racks for the year. Like AMD, NVIDIA is unlikely to include H20 (or B20/30) related revenue in October guidance, due to (1) delays and uncertainty around US government licensing approvals, (2) NVIDIA’s need to reassess Chinese customer orders, and (3) China’s government discouraging local adoption of NVIDIA GPUs due to “backdoor” concerns.\n\nThat said, potential H20 upside remains: NVIDIA’s existing $1.9bn inventory could convert into $5–6bn of revenue. Every additional 100k units shipped could yield at least $1bn in extra sales, with ~$3–4bn per quarter possible across 2–3 quarters. Gross margin for October is expected to improve slightly (J.P. Morgan estimate: 73%) on supply chain optimization, moving toward the mid-70% range by year-end.\n\nNVIDIA (NVDA) Buyside Dynamics (Buyside Bars) – Josh Meyers\n\nInvestor Sentiment: Positioning in NVIDIA remains relatively high, based on trading flows and client conversations. Even long-only investors who have trimmed positions remain optimistic. Our survey results mirror this elevated expectation. Investors anticipate ~60% YoY revenue growth this quarter, and ~55% YoY growth for next quarter guidance. Respondents expect both July revenue and October guidance to come in ~3.5% above consensus (roughly $2bn higher than current guidance, as usual).\nSurvey respondents see 2026 EPS >$6.5 (with $7 commonly cited; Harlan Sur views this as easily achievable). Optimism is driven by smooth GB200 shipments and the ramp of GB300 in 2H.\n\nRemaining uncertainties:\n\n1. Order volatility – hints of GB300 system order pull-ins (e.g., Quanta).\n\n2. H20 shipment guidance – balancing Chinese demand vs political risk.\n\n3. Networking supply – parts shortages.\n\nOptimists cite Hon Hai’s server strength as support against order volatility concerns (Gokul expects at least 6,000 servers in 3Q). The second and third issues are already well known, making any surprises easier to justify. Notably, margin expectations showed little dispersion in the survey.\n\nWhen shared with Asia tech specialist Duncan Wagner, he emphasized that Asian investors seem less optimistic than these survey results suggest. Another point of discussion: NVIDIA’s stock has shown a recent pattern of strong gains in 1H (2023, 2024, 2025) followed by flat trading in 2H (2023, 2024).\n\nPositioning score (1 = max sell/underweight, 10 = max buy/overweight): 8\nExpected move: 5.8%\n\nNVIDIA (NVDA) and Related Plays (Laterals) – Ashwin Panjabi\nNVIDIA could be the spark that unleashes “animal spirits” this week. Conditions are perfect. Last week’s Altman comments and the M.I.T. report reignited the recurring six-monthly concerns over AI CAPEX sustainability. Gokul Hariharan and his team produced an excellent chart showing the massive capacity for US CSPs to keep spending through 2027.\n\nIf EBITDA and OCF grow 10% and are fully reinvested, 2027 CAPEX could grow 76%… This is not our base case, but it illustrates the upside potential.\nSpending appetite is also enormous. The Big 4 already monetize AI based on FCF this year, with minimal inclusion of reasoning models (whose token usage is 10–30x higher than chatbots). Video-related GenAI, followed by agentic AI, represents the next wave of growth. Appetite extends to startups, sovereign wealth funds/governments, and enterprises, reinforcing confidence in the 2028 outlook.\n\nMuch of this demand remains latent due to severe supply constraints, driving price increases across the stack: advanced packaging, ABF substrates, power ICs, MLCCs, and commodity DRAM.\n\nWinners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005769260362608719, "ret_signed_m1d": 0.005769260362608719, "alpha_spy_m1d": 0.008042593824131639, "alpha_c_m1d": 0.008826400372442644, "bench_spy_m1d": -0.0022733334615229195, "bench_c_m1d": -0.003057140009833925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.034183898078270936, "ret_signed_p1d": 0.034183898078270936, "alpha_spy_p1d": 0.030642535346128152, "alpha_c_p1d": 0.030085257130551435, "bench_spy_p1d": 0.003541362732142783, "bench_c_p1d": 0.004098640947719501, "ret_p1w": 0.011073665623563933, "ret_signed_p1w": 0.011073665623563933, "alpha_spy_p1w": 0.015543038066634929, "alpha_c_p1w": 0.048801258055358065, "bench_spy_p1w": -0.004469372443070996, "bench_c_p1w": -0.03772759243179413, "ret_p1m": 0.37313429593280456, "ret_signed_p1m": 0.37313429593280456, "alpha_spy_p1m": 0.3526521154517168, "alpha_c_p1m": 0.29452108111961084, "bench_spy_p1m": 0.02048218048108774, "bench_c_p1m": 0.07861321481319372, "ret_p3m": 1.0028888243172989, "ret_signed_p3m": 1.0028888243172989, "alpha_spy_p3m": 0.9658444928869454, "alpha_c_p3m": 0.8636023948986564, "bench_spy_p3m": 0.03704433143035346, "bench_c_p3m": 0.13928642941864244, "ret_p6m": 2.4458941671637775, "ret_signed_p6m": 2.4458941671637775, "alpha_spy_p6m": 2.3812887373979015, "alpha_c_p6m": 2.06352366084129, "bench_spy_p6m": 0.0646054297658758, "bench_c_p6m": 0.3823705063224876, "price_path": [-0.2135, -0.2019, -0.2019, -0.1635, -0.1365, -0.1365, -0.1192, -0.1135, -0.0769, -0.0942, -0.0942, -0.0462, -0.0423, -0.0519, -0.0538, -0.0115, -0.0019, 0.0712, 0.1, 0.1269, 0.0923, 0.1231, 0.0981, 0.0731, 0.0712, 0.0404, 0.0423, 0.0846, 0.0808, 0.1423, 0.1327, 0.1538, 0.1481, 0.1385, 0.0365, 0.0346, 0.0481, 0.0327, 0.0346, 0.0365, 0.0231, 0.0077, 0.0096, 0.0135, 0.0519, -0.0077, -0.0077, 0.0135, -0.0058, 0.0077, -0.0135, 0.0269, 0.0346, 0.0692, 0.0635, 0.0635, 0.0288, 0.0115, -0.0173, -0.0577, -0.0346, -0.0019, 0.0058, 0.0, 0.0342, 0.0361, -0.014, 0.0034, 0.0111, 0.0226, 0.0534, 0.0669, 0.1093, 0.1709, 0.1825, 0.2653, 0.2749, 0.3404, 0.2845, 0.3693, 0.3693, 0.3519, 0.3905, 0.377, 0.3731, 0.2961, 0.3442, 0.3385, 0.3866, 0.5234, 0.5234, 0.5234, 0.5234, 0.5234, 0.5234, 0.6485, 0.5985, 0.585, 0.6273, 0.7429, 0.793, 0.87, 0.845, 0.8546, 0.843, 0.9644, 1.0607, 1.0067, 1.1493, 1.1878, 1.1531, 1.3881, 1.2571, 1.2301, 1.2841, 1.234, 1.3341, 1.3842, 1.3765, 1.3572, 1.157, 1.3341, 1.1955, 1.1647, 1.1993, 1.0067, 1.0029, 0.999, 1.0183, 1.0968, 1.0429, 1.0737, 1.1508, 1.1277, 1.0891, 1.0968, 1.224, 1.1816, 1.2626, 1.1778, 1.2009, 1.1354, 1.0429, 1.1238, 1.1277, 1.1084, 1.2356, 1.251, 1.2664, 1.2664, 1.3088, 1.4669, 1.5093, 1.5093, 1.5093, 1.6095, 1.6827, 1.7983, 1.86, 1.914, 1.8677, 1.887, 1.8446, 1.86, 1.887, 1.914, 1.9448, 1.8639, 1.8523, 1.9101, 1.9564, 1.8369, 2.0836, 2.2416, 2.3187, 2.5037, 2.1992, 2.496, 2.469, 2.2455, 2.2339, 2.4189, 2.3765, 2.3148, 2.4228, 2.3919, 2.3919, 2.3919, 2.3919, 2.4459]}
{"unit_id": "orig:1960698305157194041", "ticker_idx": 2, "ts": "2025-08-27T13:37:58+00:00", "symbol": "TSMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Winners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares JPM NVDA preview: July rev $46-47B beats, Oct guide $53-54B+; names SK Hynix, TSMC, Hon Hai as AI supply-chain winners.", "resolved_tickers": ["TSM"], "resolution_method": "exchange_ticker", "resolution_reason": "TSMC listed on NYSE as TSM (ADR); also 2330.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) NVIDIA (NVDA) Earnings Preview\n\nIndustry Commentary – We (Harlan and team) believe near-term AI fundamentals remain solid, supported by strong hyperscaler CAPEX spending. This trend was clearly demonstrated during the 2Q25 earnings season, where cloud/hyperscaler companies raised CAPEX guidance, and other AI beneficiaries (e.g., MTSI, ALAB, AMD) reported strong results and outlooks. In cyclical sectors such as telecom/service providers and enterprises, conditions are also showing gradual improvement.\n\nNVIDIA (NVDA) – Thanks to the ramp-up of GB200 production and the start of the GB300 program, July quarter results are expected to beat Wall Street expectations (revenue around $46–47bn), with October quarter guidance likely coming in at $53–54bn or higher, modestly above consensus. NVIDIA is scheduled to report July quarter results on Wednesday (Aug 27, local time), and revenue is likely to exceed guidance ($45bn) and land in the $46–47bn range. This upside is almost entirely driven by the strong ramp of GB200 rack shipments.\n\nFor perspective: April quarter revenue ($44bn) included ~$4.5bn from H20. Excluding that, the comparable base was $39.5bn. Thus, July quarter revenue adds $6.5–7.5bn on a like-for-like basis (versus a reported increase of only $2–3bn), again driven by GB200.\n\nFor October quarter, we expect NVIDIA to guide revenue to $53–54bn or above. This would modestly top consensus (consistent with recent guidance trends – see Figure 1) and sit above our own estimate of $51bn, as GB200 shipments continue to accelerate (estimated 8,000–9,000 racks). Total Blackwell rack shipments could reach ~12,000 in fiscal 4Q, implying ~28,000–30,000 racks for the year. Like AMD, NVIDIA is unlikely to include H20 (or B20/30) related revenue in October guidance, due to (1) delays and uncertainty around US government licensing approvals, (2) NVIDIA’s need to reassess Chinese customer orders, and (3) China’s government discouraging local adoption of NVIDIA GPUs due to “backdoor” concerns.\n\nThat said, potential H20 upside remains: NVIDIA’s existing $1.9bn inventory could convert into $5–6bn of revenue. Every additional 100k units shipped could yield at least $1bn in extra sales, with ~$3–4bn per quarter possible across 2–3 quarters. Gross margin for October is expected to improve slightly (J.P. Morgan estimate: 73%) on supply chain optimization, moving toward the mid-70% range by year-end.\n\nNVIDIA (NVDA) Buyside Dynamics (Buyside Bars) – Josh Meyers\n\nInvestor Sentiment: Positioning in NVIDIA remains relatively high, based on trading flows and client conversations. Even long-only investors who have trimmed positions remain optimistic. Our survey results mirror this elevated expectation. Investors anticipate ~60% YoY revenue growth this quarter, and ~55% YoY growth for next quarter guidance. Respondents expect both July revenue and October guidance to come in ~3.5% above consensus (roughly $2bn higher than current guidance, as usual).\nSurvey respondents see 2026 EPS >$6.5 (with $7 commonly cited; Harlan Sur views this as easily achievable). Optimism is driven by smooth GB200 shipments and the ramp of GB300 in 2H.\n\nRemaining uncertainties:\n\n1. Order volatility – hints of GB300 system order pull-ins (e.g., Quanta).\n\n2. H20 shipment guidance – balancing Chinese demand vs political risk.\n\n3. Networking supply – parts shortages.\n\nOptimists cite Hon Hai’s server strength as support against order volatility concerns (Gokul expects at least 6,000 servers in 3Q). The second and third issues are already well known, making any surprises easier to justify. Notably, margin expectations showed little dispersion in the survey.\n\nWhen shared with Asia tech specialist Duncan Wagner, he emphasized that Asian investors seem less optimistic than these survey results suggest. Another point of discussion: NVIDIA’s stock has shown a recent pattern of strong gains in 1H (2023, 2024, 2025) followed by flat trading in 2H (2023, 2024).\n\nPositioning score (1 = max sell/underweight, 10 = max buy/overweight): 8\nExpected move: 5.8%\n\nNVIDIA (NVDA) and Related Plays (Laterals) – Ashwin Panjabi\nNVIDIA could be the spark that unleashes “animal spirits” this week. Conditions are perfect. Last week’s Altman comments and the M.I.T. report reignited the recurring six-monthly concerns over AI CAPEX sustainability. Gokul Hariharan and his team produced an excellent chart showing the massive capacity for US CSPs to keep spending through 2027.\n\nIf EBITDA and OCF grow 10% and are fully reinvested, 2027 CAPEX could grow 76%… This is not our base case, but it illustrates the upside potential.\nSpending appetite is also enormous. The Big 4 already monetize AI based on FCF this year, with minimal inclusion of reasoning models (whose token usage is 10–30x higher than chatbots). Video-related GenAI, followed by agentic AI, represents the next wave of growth. Appetite extends to startups, sovereign wealth funds/governments, and enterprises, reinforcing confidence in the 2028 outlook.\n\nMuch of this demand remains latent due to severe supply constraints, driving price increases across the stack: advanced packaging, ABF substrates, power ICs, MLCCs, and commodity DRAM.\n\nWinners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0023820835209913405, "ret_signed_m1d": -0.0023820835209913405, "alpha_spy_m1d": -0.00010875005946842098, "alpha_c_m1d": 0.0006750564888425847, "bench_spy_m1d": -0.0022733334615229195, "bench_c_m1d": -0.003057140009833925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004262587755470615, "ret_signed_p1d": -0.004262587755470615, "alpha_spy_p1d": -0.007803950487613398, "alpha_c_p1d": -0.008361228703190116, "bench_spy_p1d": 0.003541362732142783, "bench_c_p1d": 0.004098640947719501, "ret_p1w": -0.03301431141165556, "ret_signed_p1w": -0.03301431141165556, "alpha_spy_p1w": -0.028544938968584566, "alpha_c_p1w": 0.004713281020138571, "bench_spy_p1w": -0.004469372443070996, "bench_c_p1w": -0.03772759243179413, "ret_p1m": 0.15981790317731637, "ret_signed_p1m": 0.15981790317731637, "alpha_spy_p1m": 0.13933572269622863, "alpha_c_p1m": 0.08120468836412265, "bench_spy_p1m": 0.02048218048108774, "bench_c_p1m": 0.07861321481319372, "ret_p3m": 0.19327178792111144, "ret_signed_p3m": 0.19327178792111144, "alpha_spy_p3m": 0.15622745649075798, "alpha_c_p3m": 0.053985358502468994, "bench_spy_p3m": 0.03704433143035346, "bench_c_p3m": 0.13928642941864244, "ret_p6m": 0.5149105667022622, "ret_signed_p6m": 0.5149105667022622, "alpha_spy_p6m": 0.4503051369363864, "alpha_c_p6m": 0.13254006037977462, "bench_spy_p6m": 0.0646054297658758, "bench_c_p6m": 0.3823705063224876, "price_path": [-0.1951, -0.1887, -0.1772, -0.1572, -0.1533, -0.1457, -0.1381, -0.1154, -0.1085, -0.0997, -0.1178, -0.0987, -0.1061, -0.1078, -0.1078, -0.1245, -0.1211, -0.0802, -0.0692, -0.0639, -0.0448, -0.0535, -0.0611, -0.0238, -0.0188, -0.0188, -0.0423, -0.0478, -0.0311, -0.0398, -0.0372, -0.0444, -0.0098, -0.0072, 0.0264, 0.0046, -0.0018, -0.0196, 0.0043, 0.0097, 0.0264, 0.0145, 0.0085, 0.0151, 0.0097, -0.0171, -0.0012, -0.0285, -0.0331, 0.0139, 0.0106, 0.0117, 0.0209, 0.009, 0.0071, -0.0017, 0.0089, -0.0275, -0.0447, -0.05, -0.0263, -0.0155, -0.0024, 0.0, -0.0043, -0.0352, -0.0352, -0.0456, -0.033, -0.0171, 0.0172, 0.033, 0.0486, 0.0884, 0.082, 0.0837, 0.0923, 0.0986, 0.1017, 0.1262, 0.1104, 0.1429, 0.1852, 0.1768, 0.1598, 0.146, 0.1454, 0.1708, 0.2093, 0.2078, 0.2249, 0.2677, 0.2326, 0.2766, 0.2572, 0.1766, 0.2698, 0.2406, 0.2774, 0.257, 0.237, 0.248, 0.2346, 0.211, 0.2188, 0.2365, 0.2503, 0.2641, 0.279, 0.2712, 0.2595, 0.278, 0.2327, 0.231, 0.2126, 0.2011, 0.2378, 0.2206, 0.2183, 0.183, 0.194, 0.1822, 0.1651, 0.1838, 0.1633, 0.1531, 0.1933, 0.1934, 0.2156, 0.2156, 0.2221, 0.206, 0.2245, 0.2386, 0.228, 0.2355, 0.2655, 0.272, 0.3002, 0.2814, 0.2276, 0.2095, 0.2059, 0.1642, 0.1967, 0.2146, 0.2328, 0.2482, 0.256, 0.256, 0.273, 0.2649, 0.2593, 0.2774, 0.2774, 0.3435, 0.3546, 0.3764, 0.3396, 0.3368, 0.3604, 0.3946, 0.3923, 0.375, 0.4361, 0.4393, 0.4393, 0.3752, 0.3709, 0.3761, 0.4076, 0.3986, 0.4222, 0.4389, 0.4273, 0.3895, 0.4349, 0.4113, 0.3693, 0.3902, 0.4664, 0.494, 0.5213, 0.5725, 0.5473, 0.54, 0.54, 0.5309, 0.5228, 0.5149]}
{"unit_id": "orig:1960698305157194041", "ticker_idx": 3, "ts": "2025-08-27T13:37:58+00:00", "symbol": "Hon Hai", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Winners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares JPM NVDA preview: July rev $46-47B beats, Oct guide $53-54B+; names SK Hynix, TSMC, Hon Hai as AI supply-chain winners.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) NVIDIA (NVDA) Earnings Preview\n\nIndustry Commentary – We (Harlan and team) believe near-term AI fundamentals remain solid, supported by strong hyperscaler CAPEX spending. This trend was clearly demonstrated during the 2Q25 earnings season, where cloud/hyperscaler companies raised CAPEX guidance, and other AI beneficiaries (e.g., MTSI, ALAB, AMD) reported strong results and outlooks. In cyclical sectors such as telecom/service providers and enterprises, conditions are also showing gradual improvement.\n\nNVIDIA (NVDA) – Thanks to the ramp-up of GB200 production and the start of the GB300 program, July quarter results are expected to beat Wall Street expectations (revenue around $46–47bn), with October quarter guidance likely coming in at $53–54bn or higher, modestly above consensus. NVIDIA is scheduled to report July quarter results on Wednesday (Aug 27, local time), and revenue is likely to exceed guidance ($45bn) and land in the $46–47bn range. This upside is almost entirely driven by the strong ramp of GB200 rack shipments.\n\nFor perspective: April quarter revenue ($44bn) included ~$4.5bn from H20. Excluding that, the comparable base was $39.5bn. Thus, July quarter revenue adds $6.5–7.5bn on a like-for-like basis (versus a reported increase of only $2–3bn), again driven by GB200.\n\nFor October quarter, we expect NVIDIA to guide revenue to $53–54bn or above. This would modestly top consensus (consistent with recent guidance trends – see Figure 1) and sit above our own estimate of $51bn, as GB200 shipments continue to accelerate (estimated 8,000–9,000 racks). Total Blackwell rack shipments could reach ~12,000 in fiscal 4Q, implying ~28,000–30,000 racks for the year. Like AMD, NVIDIA is unlikely to include H20 (or B20/30) related revenue in October guidance, due to (1) delays and uncertainty around US government licensing approvals, (2) NVIDIA’s need to reassess Chinese customer orders, and (3) China’s government discouraging local adoption of NVIDIA GPUs due to “backdoor” concerns.\n\nThat said, potential H20 upside remains: NVIDIA’s existing $1.9bn inventory could convert into $5–6bn of revenue. Every additional 100k units shipped could yield at least $1bn in extra sales, with ~$3–4bn per quarter possible across 2–3 quarters. Gross margin for October is expected to improve slightly (J.P. Morgan estimate: 73%) on supply chain optimization, moving toward the mid-70% range by year-end.\n\nNVIDIA (NVDA) Buyside Dynamics (Buyside Bars) – Josh Meyers\n\nInvestor Sentiment: Positioning in NVIDIA remains relatively high, based on trading flows and client conversations. Even long-only investors who have trimmed positions remain optimistic. Our survey results mirror this elevated expectation. Investors anticipate ~60% YoY revenue growth this quarter, and ~55% YoY growth for next quarter guidance. Respondents expect both July revenue and October guidance to come in ~3.5% above consensus (roughly $2bn higher than current guidance, as usual).\nSurvey respondents see 2026 EPS >$6.5 (with $7 commonly cited; Harlan Sur views this as easily achievable). Optimism is driven by smooth GB200 shipments and the ramp of GB300 in 2H.\n\nRemaining uncertainties:\n\n1. Order volatility – hints of GB300 system order pull-ins (e.g., Quanta).\n\n2. H20 shipment guidance – balancing Chinese demand vs political risk.\n\n3. Networking supply – parts shortages.\n\nOptimists cite Hon Hai’s server strength as support against order volatility concerns (Gokul expects at least 6,000 servers in 3Q). The second and third issues are already well known, making any surprises easier to justify. Notably, margin expectations showed little dispersion in the survey.\n\nWhen shared with Asia tech specialist Duncan Wagner, he emphasized that Asian investors seem less optimistic than these survey results suggest. Another point of discussion: NVIDIA’s stock has shown a recent pattern of strong gains in 1H (2023, 2024, 2025) followed by flat trading in 2H (2023, 2024).\n\nPositioning score (1 = max sell/underweight, 10 = max buy/overweight): 8\nExpected move: 5.8%\n\nNVIDIA (NVDA) and Related Plays (Laterals) – Ashwin Panjabi\nNVIDIA could be the spark that unleashes “animal spirits” this week. Conditions are perfect. Last week’s Altman comments and the M.I.T. report reignited the recurring six-monthly concerns over AI CAPEX sustainability. Gokul Hariharan and his team produced an excellent chart showing the massive capacity for US CSPs to keep spending through 2027.\n\nIf EBITDA and OCF grow 10% and are fully reinvested, 2027 CAPEX could grow 76%… This is not our base case, but it illustrates the upside potential.\nSpending appetite is also enormous. The Big 4 already monetize AI based on FCF this year, with minimal inclusion of reasoning models (whose token usage is 10–30x higher than chatbots). Video-related GenAI, followed by agentic AI, represents the next wave of growth. Appetite extends to startups, sovereign wealth funds/governments, and enterprises, reinforcing confidence in the 2028 outlook.\n\nMuch of this demand remains latent due to severe supply constraints, driving price increases across the stack: advanced packaging, ABF substrates, power ICs, MLCCs, and commodity DRAM.\n\nWinners will keep winning: Hon Hai, ASE, TSMC, Delta, Elite Materials, SK Hynix, GDS (BUY-rated), Unimicron.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004807692307692291, "ret_signed_m1d": 0.004807692307692291, "alpha_spy_m1d": 0.00708102576921521, "alpha_c_m1d": 0.010177230726912345, "bench_spy_m1d": -0.0022733334615229195, "bench_c_m1d": -0.005369538419220055, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009615384615384581, "ret_signed_p1d": -0.009615384615384581, "alpha_spy_p1d": -0.013156747347527364, "alpha_c_p1d": -0.017518716702368264, "bench_spy_p1d": 0.003541362732142783, "bench_c_p1d": 0.007903332086983683, "ret_p1w": -0.03605769230769229, "ret_signed_p1w": -0.03605769230769229, "alpha_spy_p1w": -0.031588319864621295, "alpha_c_p1w": -0.024335246763457885, "bench_spy_p1w": -0.004469372443070996, "bench_c_p1w": -0.011722445544234406, "ret_p1m": 0.10576923076923084, "ret_signed_p1m": 0.10576923076923084, "alpha_spy_p1m": 0.0852870502881431, "alpha_c_p1m": 0.053179687383923424, "bench_spy_p1m": 0.02048218048108774, "bench_c_p1m": 0.05258954338530741, "ret_p3m": 0.05769230769230771, "ret_signed_p3m": 0.05769230769230771, "alpha_spy_p3m": 0.02064797626195425, "alpha_c_p3m": -0.0013716879779501756, "bench_spy_p3m": 0.03704433143035346, "bench_c_p3m": 0.059063995670257885, "ret_p6m": 0.09134615384615374, "ret_signed_p6m": 0.09134615384615374, "alpha_spy_p6m": 0.026740724080277944, "alpha_c_p6m": 0.02798335266703833, "bench_spy_p6m": 0.0646054297658758, "bench_c_p6m": 0.06336280117911541, "price_path": [-0.2763, -0.2925, -0.2972, -0.2763, -0.2786, -0.2902, -0.2832, -0.2786, -0.2763, -0.2693, -0.274, -0.2809, -0.2809, -0.2763, -0.2856, -0.2786, -0.2763, -0.2554, -0.2484, -0.2438, -0.2345, -0.2531, -0.2322, -0.2212, -0.2115, -0.226, -0.226, -0.2356, -0.226, -0.2236, -0.2236, -0.2284, -0.2163, -0.2188, -0.2115, -0.2043, -0.2067, -0.2212, -0.2019, -0.1635, -0.1611, -0.1538, -0.1755, -0.1755, -0.1442, -0.1442, -0.1346, -0.113, -0.1058, -0.0649, -0.0649, -0.0481, -0.0529, -0.0457, -0.0409, -0.0048, 0.0096, 0.0, -0.0361, -0.0096, -0.0264, -0.0024, 0.0048, 0.0, -0.0096, -0.0216, -0.0457, -0.0457, -0.0361, -0.024, -0.0144, -0.0216, -0.0024, 0.0096, 0.0312, 0.0457, 0.0385, 0.0361, 0.0192, 0.0337, 0.0288, 0.0385, 0.0625, 0.0721, 0.1058, 0.0553, 0.0553, 0.0385, 0.0529, 0.0841, 0.0889, 0.0889, 0.0986, 0.0817, 0.0649, 0.0649, 0.024, -0.0096, -0.0072, 0.0721, 0.0889, 0.1466, 0.149, 0.1659, 0.149, 0.149, 0.1851, 0.2019, 0.25, 0.2596, 0.238, 0.2091, 0.1731, 0.1875, 0.1923, 0.1731, 0.1995, 0.1851, 0.2067, 0.2115, 0.1587, 0.1346, 0.1058, 0.101, 0.137, 0.0817, 0.0577, 0.0529, 0.0913, 0.1106, 0.0841, 0.0673, 0.0673, 0.0913, 0.0986, 0.1106, 0.1154, 0.1298, 0.1226, 0.0865, 0.0913, 0.0649, 0.0481, 0.0409, 0.0385, 0.0649, 0.0769, 0.0769, 0.0769, 0.0769, 0.0841, 0.1106, 0.0962, 0.1082, 0.1082, 0.1154, 0.1274, 0.1346, 0.1466, 0.1034, 0.1082, 0.0986, 0.0889, 0.1274, 0.1226, 0.1274, 0.1034, 0.0745, 0.0529, 0.0745, 0.0649, 0.0769, 0.0841, 0.0841, 0.0769, 0.0601, 0.0312, 0.0409, 0.0529, 0.0361, 0.0337, 0.0505, 0.0625, 0.0913, 0.0913, 0.0913, 0.0913, 0.0913, 0.0913, 0.0913]}
{"unit_id": "orig:1947434273818218540", "ticker_idx": 0, "ts": "2025-07-21T23:11:27+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek has a high probability of securing the contract", "tweet_type": "analysis", "timeline": "months", "summary": "MediaTek likely to win Meta's Arke 2nm ASIC design contract; also expected to benefit from broader ASIC demand expansion.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250722_Taiwan design service - CLSA\n\n(1) CSP (Cloud Service Provider) firms are adopting more flexible chip development models, and automotive OEMs are increasingly moving towards in-house development of ASICs for ADAS.\n\n(2) This is opening up greater opportunities for Taiwanese ASIC companies.\n\n(3) Recent developments like Meta's new 2nm project and the design changes to its Olympus chip demonstrate the continued expansion of ASIC demand. Companies like MediaTek are expected to benefit.\n\n(4) Furthermore, CSP firms are increasingly separating design from the supply chain and tending to handle Front-End design in-house for cost efficiency. This presents a structural opportunity for design service providers.\n\n*Why is in-house Front-End design by CSPs beneficial for them? Specifically, complex areas within Front-End design such as RTL implementation, verification, DFT, and physical design require high levels of expertise, making them difficult to perform entirely in-house. Therefore, CSPs often design only the architecture or key blocks directly, while outsourcing the rest of the Front-End and Back-End (e.g., place & route, tape-out) to specialized design service firms.\n**Initially, ASIC development companies primarily outsourced turnkey projects to experienced fabless firms like Broadcom. However, as generations progress, optimizing cost structures becomes increasingly important.\n\n(5) Meanwhile, BEV (Battery Electric Vehicle) companies are directly developing ADAS ASICs to strengthen software and hardware integration, further broadening market opportunities for Taiwanese ASIC companies.\n\n(6) Below is an update on Meta's ASIC roadmap.\n\n(7) Meta recently announced 'Arke', a new ASIC project, in addition to its existing ASIC lineup.\n\n(8) It will use TSMC's 2nm process with a target for mass production in Q1 2027. Its purpose is to act as a bridge between Iris and Olympus.\n\n(9) The selection of the design house is expected to be completed within a few weeks, and our assessment is that MediaTek has a high probability of securing the contract.\n\n(10) On the other hand, the schedule for Olympus is being delayed. This is due to some design changes aimed at enhancing its competitiveness against NVIDIA's Vera Rubin.\n\n(11) Consequently, the mass production schedule has been pushed back from 4Q27 to 1H28.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010526292809482696, "ret_signed_m1d": -0.010526292809482696, "alpha_spy_m1d": -0.008633706319317813, "alpha_c_m1d": -0.009837876921564792, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0035087642698274912, "ret_signed_p1d": 0.0035087642698274912, "alpha_spy_p1d": 0.0033656431066029224, "alpha_c_p1d": 0.021201779677888055, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.01754382134913779, "ret_signed_p1w": -0.01754382134913779, "alpha_spy_p1w": -0.03053750003390332, "alpha_c_p1w": -0.020641850907500925, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": -0.024561349888792883, "ret_signed_p1m": -0.024561349888792883, "alpha_spy_p1m": -0.04211943853013156, "alpha_c_p1m": -0.02580056171213818, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": -0.0666666084506895, "ret_signed_p3m": -0.0666666084506895, "alpha_spy_p3m": -0.12026587565467084, "alpha_c_p3m": -0.24838142682133046, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.062306522838273315, "ret_signed_p6m": 0.062306522838273315, "alpha_spy_p6m": -0.04739839401978441, "alpha_c_p6m": -0.2896203619781468, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.0572, -0.0847, -0.0503, -0.0709, -0.0572, -0.0709, -0.0709, -0.1053, -0.1088, -0.1191, -0.1225, -0.1122, -0.0881, -0.095, -0.0675, -0.064, -0.0572, -0.0606, -0.0984, -0.095, -0.0778, -0.0881, -0.095, -0.1191, -0.1191, -0.1191, -0.1329, -0.1329, -0.1329, -0.1225, -0.1225, -0.1432, -0.1191, -0.1157, -0.0812, -0.095, -0.1157, -0.1397, -0.1363, -0.1225, -0.1157, -0.1294, -0.1397, -0.1363, -0.126, -0.1088, -0.1122, -0.1157, -0.1397, -0.1225, -0.1088, -0.1088, -0.0947, -0.1018, -0.1088, -0.0491, -0.0175, -0.0035, -0.0281, -0.0211, -0.0105, -0.0246, -0.0105, 0.0, 0.0035, 0.0035, 0.0035, 0.0035, -0.0175, -0.0456, -0.0316, -0.0386, -0.0386, -0.0667, -0.0596, -0.0737, -0.0491, -0.0526, -0.0491, -0.0456, -0.0386, -0.0246, -0.0386, -0.0105, -0.0246, -0.0491, -0.0421, -0.0421, -0.0105, -0.0175, -0.0175, -0.0281, -0.0386, -0.0456, -0.0456, -0.0211, -0.0281, 0.007, 0.0035, 0.0596, 0.0456, 0.0386, 0.0421, 0.0421, 0.0772, 0.0596, 0.0596, 0.0105, -0.014, -0.0351, -0.0667, -0.0561, -0.0807, -0.0807, -0.0772, -0.0982, -0.1018, -0.0737, -0.0737, -0.0702, -0.0632, -0.0561, -0.0561, -0.0772, -0.0842, -0.0772, -0.0667, -0.0667, -0.0596, -0.0596, -0.0667, -0.0912, -0.0912, -0.0632, -0.0772, -0.0877, -0.0807, -0.0807, -0.0877, -0.0772, -0.0667, -0.0947, -0.1158, -0.1228, -0.1298, -0.1263, -0.1263, -0.1368, -0.1368, -0.1789, -0.186, -0.1684, -0.1965, -0.193, -0.1684, -0.0877, -0.0596, -0.0211, 0.014, -0.007, -0.0175, -0.014, 0.0, 0.0105, -0.0035, 0.0246, -0.0211, -0.014, -0.0035, -0.0035, 0.0, -0.0035, -0.0105, -0.0175, -0.0246, -0.0316, -0.0316, -0.0281, -0.0035, -0.0035, 0.0035, 0.0035, 0.0316, 0.0702, 0.0623, 0.0659, 0.0337, 0.0158, 0.0337, 0.0623]}
{"unit_id": "orig:1957688306822705168", "ticker_idx": 0, "ts": "2025-08-19T06:17:19+00:00", "symbol": "software", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Goldman expects that as the pressure from enterprise software renewal cycles eases in 2026, AI's incremental contribution will provide stability for key indicators such as net revenue retention (NRR), paving a durable, multi-year growth path for the industry", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Goldman Sachs bullish view: AI is a force multiplier for SaaS/enterprise software, not a disruptor; multi-year growth path ahead as renewal headwinds ease in 2026.", "resolved_tickers": ["IGV"], "resolution_method": "etf", "resolution_reason": "Global software sector proxied by IGV ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Will software companies be doomed to die in the age of AI? Goldman Sachs says no.\n\nThe AI wave is reshaping the technology industry with unprecedented momentum, sparking an intense debate over the outlook for traditional software businesses.\n\nOver the past two weeks, sentiment in the software sector has clearly turned bearish. Last Tuesday, after OpenAI released its latest large model, GPT-5, worries that AI would replace traditional software triggered panic selling in the market. Shares of German software giant SAP plunged as much as 7.1%, wiping out nearly €22 billion in market value—the biggest one-day drop since late 2020—and dragged European software stocks sharply lower across the board.\n\nAccording to Goldman Sachs analyst Gabriela Borges and her team, their report on the 17th pointed out that investors’ concerns are focused on an existential risk: “Will AI erode existing pricing models, lower barriers to entry for new players, and ultimately pressure the profit margins of SaaS (Software-as-a-Service) giants?”\n\nIn Goldman’s view, the answer is no.\n\nThe report argues that the “software is dead” rhetoric is overly pessimistic. In some cases, AI has the potential to become a “force multiplier” for industry-leading suppliers. The current stage is similar to the historical period when the software industry shifted from on-premises deployment to cloud computing—a time that gave rise to a new cohort of leaders, but also prompted established companies such as Adobe, Intuit, and Autodesk to transform into larger, faster-growing, and more profitable enterprises, creating meaningful excess returns for investors.\n\nGoldman expects that as the pressure from enterprise software renewal cycles eases in 2026, AI’s incremental contribution will provide stability for key indicators such as net revenue retention (NRR), paving a durable, multi-year growth path for the industry. The future leaders of the software market will consist of two types of companies: today’s giants that continue to innovate, and emerging firms that successfully build differentiated, AI-native software.\n\nIs AI a Disruptor or a Force Multiplier?\nIn the debate over whether AI-native companies can replace traditional SaaS firms, the crux is whether their products can be “significantly better and cheaper.”\n\nGoldman points out that today’s SaaS giants are not comparable to the on-premises software companies they once disrupted. Back then, SaaS had a huge advantage in cost and experience thanks to cloud architecture, whereas SaaS enterprises today have set rather high competitive barriers.\n\nIn terms of pricing strategy, the biggest risk for AI-native companies is offering value-based pricing, which could cannibalize seat-based models. However, SaaS leaders have begun evolving their pricing toward outcome-based approaches, as Salesforce has done, or encouraging adoption by incorporating AI features into existing subscription plans like HubSpot.\n\nOn the innovation front, SaaS leaders are maintaining a rapid pace through organic growth and acquisitions. Examples abound: ServiceNow acquired Moveworks, Braze acquired Offerfit, and Salesforce acquired Bluebirds. In organic innovation, Salesforce launched Agentforce in September 2024, which to a large extent defined the next phase of AI adoption.\n\nTherefore, the report believes it is very difficult for AI-native companies to disrupt existing SaaS leaders on pricing and product functionality.\n\nA Goldman case study shows that in call-center scenarios, AI applications can reduce operating costs by 30%, yet total software budgets increase by 130%. This suggests that AI not only fails to compress the market, but may actually expand the total addressable market (TAM) for application software.\n\nThe Software Giants’ Hybrid AI Strategies and Moats\nTo meet challenges and seize opportunities, giants in enterprise software are broadly deploying a hybrid AI model strategy, which has become key to strengthening their moats.\n\nThis strategy combines domain-specific models trained on a company’s own proprietary data with cutting-edge external large language models (LLMs), preserving their data advantages while offering flexibility and high performance.\n\nFrom Snowflake’s Arctic model and Salesforce’s Agentforce to ServiceNow’s Now LLM, these industry leaders embed their own models within their platforms to handle core enterprise workflows, while also allowing customers to call external models such as OpenAI and Google via interfaces.\n\nThe report explains that this strategy greatly reduces the risk of being “undercut at the source” by AI-native upstarts, because it locks customers into ecosystems that are familiar, secure, and deeply integrated. Meanwhile, vertical software vendors possess stronger barriers thanks to their deep industry expertise.\n\nGoldman believes that in these specialized domains, AI is more likely to play an “augmenting” rather than a “replacing” role. Due to complex workflows and high data-migration costs, even technically superb AI-native products will struggle to earn customer trust and adoption in the short term.\n\nHigher Barriers for Enterprise Software Than for Consumer Software\nThe report also notes that the market often underestimates the fundamental differences between enterprise and consumer software—differences that constitute another important moat for existing giants. Goldman emphasizes that entry barriers for enterprise software are much higher than for consumer applications, with the core being “mission criticality.”\n\nThe report points out that “hallucinations” in AI models may be harmless in consumer settings, but in enterprise environments they can lead to reputational damage, customer churn, and other serious consequences. Therefore, enterprise software is about more than product features; it involves an entire complex system “beneath the iceberg.”\n\nIn addition, a current view holds that customers can now build their own SaaS-equivalent tech stacks at lower cost than before. Agentic AI significantly lowers development thresholds through code-generation and debugging tools, and lowers the barrier to full UIs through natural-language-driven agents. For example, Palantir’s value proposition combines forward-deployed engineers, an ontology-based operating system, and an AI toolkit.\n\nHowever, Goldman sees this as a classic insourcing vs. outsourcing issue. Even if the cost of maintaining software applications in-house falls, the costs of specialized suppliers fall as well; thus, the efficiency frontier of professional vendors typically remains ahead of internal efforts.\n\nGoldman argues that simply using AI to replicate a software stack is not only fraught with difficulty but also “somewhat meaningless.” The essence of technological change is to create new, differentiated applications, rather than to rebuild the past.\n\nThe report suggests that the industry may currently be at a local peak in the share of opportunities for customized versus packaged solutions: existing SaaS firms lack comprehensive AI features; AI-native firms often address only a small part of broader enterprise problems; and many developers and IT professionals are still early in the learning curve of successfully implementing AI projects.\n\nFuture Indicators to Watch\nWhile remaining optimistic about the long-term outlook for the software industry, Goldman also highlights several key factors that investors should monitor closely to determine whether the bullish or bearish arguments are being validated.\n\nFirst is the stability of net revenue retention (NRR). The report shows that over the past two years, SaaS leaders have faced particular renewal pressure. Pandemic-related demand peaked in 2021 and the first half of 2022, and those three-year contracts faced the greatest pressure upon renewal in 2024 and the first half of 2025.\n\nSecond is the degree of AI revenue contribution to growth. SaaS leaders need to demonstrate that their AI products generate incremental revenue. For example, Adobe expects its AI products to contribute $250 million in annualized recurring revenue (ARR) by the end of 2025. The report notes that such clear financial figures will be key validation signals.\n\nThird is customer feedback on SaaS leaders’ innovation. The market will watch the progress SaaS leaders make in addressing obstacles to AI adoption, such as model selection and tool integration.\n\nFinally, the momentum of AI-native companies will be tracked to gauge their staying power and the risk they pose to incumbents. Many AI-native firms founded over the past two years have the potential to grow into larger enterprises over time, expanding from single-product companies into suites. Tracking such companies over the long run will help assess their actual impact on existing profit pools.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02655284089355403, "ret_signed_m1d": 0.02655284089355403, "alpha_spy_m1d": 0.02109811824704133, "alpha_c_m1d": 0.02109811824704133, "bench_spy_m1d": 0.0054547226465127, "bench_c_m1d": 0.0054547226465127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0023456558050757303, "ret_signed_p1d": -0.0023456558050757303, "alpha_spy_p1d": 0.0003114278255709202, "alpha_c_p1d": 0.0003114278255709202, "bench_spy_p1d": -0.0026570836306466505, "bench_c_p1d": -0.0026570836306466505, "ret_p1w": -9.384627564867909e-05, "ret_signed_p1w": -9.384627564867909e-05, "alpha_spy_p1w": -0.00845568913472794, "alpha_c_p1w": -0.00845568913472794, "bench_spy_p1w": 0.00836184285907926, "bench_c_p1w": 0.00836184285907926, "ret_p1m": 0.062300612456113846, "ret_signed_p1m": 0.062300612456113846, "alpha_spy_p1m": 0.03202598301100168, "alpha_c_p1m": 0.03202598301100168, "bench_spy_p1m": 0.030274629445112167, "bench_c_p1m": 0.030274629445112167, "ret_p3m": 0.008069061696159396, "ret_signed_p3m": 0.008069061696159396, "alpha_spy_p3m": -0.04504500010530266, "alpha_c_p3m": -0.04504500010530266, "bench_spy_p3m": 0.053114061801462054, "bench_c_p3m": 0.053114061801462054, "ret_p6m": -0.21908423787733045, "ret_signed_p6m": -0.21908423787733045, "alpha_spy_p6m": -0.3067958971148931, "alpha_c_p6m": -0.3067958971148931, "bench_spy_p6m": 0.08771165923756263, "bench_c_p6m": 0.08771165923756263, "price_path": [-0.0363, -0.0436, -0.0436, -0.0261, -0.0328, -0.0397, -0.0275, -0.0248, -0.0149, -0.0167, -0.018, -0.0056, -0.0042, -0.0066, -0.0074, 0.0043, -0.0029, 0.0077, 0.0007, -0.0046, -0.0046, -0.0138, -0.002, 0.0138, 0.0112, 0.02, 0.0113, 0.0274, 0.0156, 0.0209, 0.044, 0.044, 0.0388, 0.0422, 0.0454, 0.0241, 0.0084, 0.0232, 0.0226, 0.0303, 0.0444, 0.042, 0.0434, 0.0405, 0.0491, 0.0517, 0.0573, 0.0591, 0.0603, 0.0576, 0.0477, 0.0192, 0.0413, 0.0372, 0.0491, 0.0283, 0.0286, 0.0168, 0.028, 0.0278, 0.0165, 0.0237, 0.0266, 0.0, -0.0023, -0.0058, 0.0083, 0.001, -0.0001, 0.0082, 0.0236, 0.0143, 0.0143, 0.0065, 0.0069, 0.0024, 0.0146, 0.0323, 0.037, 0.0531, 0.0546, 0.0478, 0.0585, 0.0576, 0.0623, 0.0825, 0.0995, 0.1052, 0.0921, 0.0811, 0.0687, 0.0787, 0.0882, 0.0791, 0.0797, 0.0858, 0.0774, 0.0848, 0.0731, 0.0855, 0.0853, 0.0517, 0.0756, 0.0658, 0.063, 0.0584, 0.0572, 0.0686, 0.0791, 0.065, 0.077, 0.0884, 0.0985, 0.0951, 0.0811, 0.0693, 0.0838, 0.0856, 0.0509, 0.0485, 0.0256, 0.0272, 0.0488, 0.0418, 0.0307, 0.0024, 0.0081, -0.0102, -0.0204, -0.0252, -0.0542, -0.0531, -0.0408, -0.0318, -0.0339, -0.0339, -0.0234, -0.0236, -0.0128, 0.0009, 0.0129, 0.027, 0.0265, 0.0316, 0.0328, 0.03, 0.013, -0.0112, -0.0025, -0.0176, -0.0038, 0.0111, 0.0222, 0.014, 0.0155, 0.0155, 0.0143, 0.0076, 0.0038, -0.0084, -0.0084, -0.0372, -0.0277, -0.0156, -0.0032, -0.0201, -0.0148, -0.0115, -0.0303, -0.0517, -0.0638, -0.0776, -0.0776, -0.0999, -0.1049, -0.0904, -0.0832, -0.0729, -0.0828, -0.0893, -0.1343, -0.1527, -0.1601, -0.1988, -0.2134, -0.2525, -0.2263, -0.2019, -0.1986, -0.2191]}
{"unit_id": "orig:1948531254594601151", "ticker_idx": 0, "ts": "2025-07-24T23:50:27+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": "search/cloud/AI", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "maintaining an Overweight rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley reiterates Overweight on Google; raises revenue/cloud estimates, lifts PT on accelerating Search/YouTube/Cloud growth and AI monetization progress.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Google Quarterly Report: Search, YouTube, and Cloud Businesses Accelerate Across the Board, GenAI Drives Product Suite Innovation, Increased Capital Expenditures Bolster Long-Term Growth Momentum\nMorgan Stanley (B. Nowak, 25/07/24)\n\nMorgan Stanley notes that Google's core businesses are accelerating across the board, with GenAI-driven product innovation permeating core product lines, cloud business growth momentum significantly strengthening, and capital expenditures substantially raised to support future development. Although increased short-term investment creates some pressure on EPS and free cash flow, the sustained growth momentum and valuation re-rating logic remain valid.\n\nCore Businesses Accelerate Across the Board, Revenue Growth Exhibits Durability\n\nIn Q2 2025, Google's Search, YouTube, and Cloud businesses grew year-over-year by 12%, 13%, and 32% respectively, all showing acceleration compared to the previous quarter. Morgan Stanley believes that Google's leading position in traffic entry points and the continuous commercialization progress of its AI products are key drivers for its long-term sustainable revenue.\n\nAI Product Penetration Accelerates, User Engagement and Data Disclosure Simultaneously Increase\n\nGoogle continues to advance its AI applications, not only integrating GenAI products into more product lines but also significantly increasing the frequency and breadth of related data disclosure. For example, enterprise users of Gemini have reached 85,000, with usage growing 35 times year-over-year. Morgan Stanley points out that these data strengthen investor confidence in the long-term value of its AI ecosystem.\n\nCapital Expenditures Substantially Raised, OBBBA Tax Benefits Provide Funding Assurance\n\nGoogle has raised its 2025 capital expenditure guidance from $75 billion to $85 billion, primarily for server and data center expansion to meet cloud demand. Morgan Stanley believes this increase could be partially offset by up to $25 billion in OBBBA tax benefits, which are expected to be a significant valuation driver in the coming quarters.\n\nGoogle Cloud Business Shows Strong Growth, Order Scale and Customer Quality Both Improve\n\nCloud business growth is primarily led by GCP, which is expected to account for approximately 80%. In Q2, cloud business backlog orders grew 18% quarter-over-quarter and 38% year-over-year; the number of large orders over $2.5 million doubled year-over-year, and orders over $1 million in 1H 2025 have already reached 2024 full-year levels. New GCP customers increased by 28% quarter-over-quarter, reflecting the rapidly increasing enterprise demand for AI tools.\n\nRevenue Expectations Raised but EPS Maintained, Target Price Increased\n\nMorgan Stanley has raised its 2026/2027 revenue expectations by 1% and 2% respectively, primarily due to a 4% and 5% upward revision of Google Cloud revenue expectations. However, due to an 8% simultaneous increase in capital expenditures, which offsets the impact on EPS and free cash flow, EPS remains unchanged, and FCF forecasts are lowered, maintaining an Overweight rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010095315600073818, "ret_signed_m1d": -0.010095315600073818, "alpha_spy_m1d": -0.009764372636749785, "alpha_c_m1d": -0.008438353675228827, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0016569619248449907, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005255644141298976, "ret_signed_p1d": 0.005255644141298976, "alpha_spy_p1d": 0.0010312945009798291, "alpha_c_p1d": 0.014553214996814057, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.00929757085551508, "ret_p1w": -0.0014050915788491602, "ret_signed_p1w": -0.0014050915788491602, "alpha_spy_p1w": 0.002283210019146442, "alpha_c_p1w": 0.009825648810044663, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": -0.011230740388893823, "ret_p1m": 0.07243574964890609, "ret_signed_p1m": 0.07243574964890609, "alpha_spy_p1m": 0.05527037557760317, "alpha_c_p1m": 0.04776489172920617, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.024670857919699918, "ret_p3m": 0.3044908098845678, "ret_signed_p3m": 0.3044908098845678, "alpha_spy_p3m": 0.2434412928226195, "alpha_c_p3m": 0.22691711568724693, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.07757369419732085, "ret_p6m": 0.7198895503794487, "ret_signed_p6m": 0.7198895503794487, "alpha_spy_p6m": 0.6234124484113714, "alpha_c_p6m": 0.65254341110428, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.06734613927516864, "price_path": [-0.1652, -0.1676, -0.1746, -0.1617, -0.1475, -0.1465, -0.1516, -0.2132, -0.1981, -0.2061, -0.1764, -0.1709, -0.1405, -0.1478, -0.1362, -0.1344, -0.1477, -0.1239, -0.1119, -0.1244, -0.1244, -0.1014, -0.1042, -0.1068, -0.1074, -0.1215, -0.1363, -0.1266, -0.1257, -0.0973, -0.0837, -0.0706, -0.0771, -0.0857, -0.0911, -0.0801, -0.0844, -0.0981, -0.0981, -0.1329, -0.1404, -0.1322, -0.1118, -0.0969, -0.071, -0.0829, -0.085, -0.0704, -0.0658, -0.0658, -0.08, -0.0927, -0.0809, -0.0757, -0.0623, -0.0552, -0.0529, -0.0479, -0.0447, -0.037, -0.0108, -0.0043, -0.0101, 0.0, 0.0053, 0.0021, 0.0186, 0.0227, -0.0014, -0.0158, 0.0149, 0.013, 0.0204, 0.0226, 0.0481, 0.0459, 0.0581, 0.0509, 0.056, 0.061, 0.059, 0.0489, 0.0372, 0.0394, 0.0724, 0.0849, 0.0779, 0.0797, 0.1013, 0.1079, 0.1079, 0.0998, 0.2003, 0.2088, 0.2229, 0.219, 0.2481, 0.2457, 0.2519, 0.2542, 0.3105, 0.3081, 0.2996, 0.3127, 0.3267, 0.3153, 0.3107, 0.2872, 0.2802, 0.2841, 0.2711, 0.2662, 0.2755, 0.2796, 0.2779, 0.3043, 0.28, 0.2741, 0.258, 0.2321, 0.2716, 0.2784, 0.3075, 0.3097, 0.3193, 0.3362, 0.3045, 0.3109, 0.3181, 0.3538, 0.4025, 0.3931, 0.4301, 0.4661, 0.4645, 0.4777, 0.4455, 0.4808, 0.4831, 0.4523, 0.511, 0.5173, 0.4933, 0.4509, 0.4396, 0.4845, 0.4806, 0.5251, 0.5076, 0.5607, 0.6593, 0.6846, 0.6664, 0.6664, 0.6676, 0.6401, 0.6449, 0.6648, 0.6543, 0.6733, 0.635, 0.6526, 0.6689, 0.6283, 0.612, 0.6064, 0.5978, 0.5464, 0.5764, 0.6009, 0.6145, 0.6383, 0.637, 0.637, 0.6339, 0.6342, 0.6357, 0.6313, 0.6313, 0.6425, 0.6497, 0.6383, 0.6781, 0.6961, 0.7124, 0.7296, 0.751, 0.7503, 0.7344, 0.7199]}
{"unit_id": "orig:1948200334624792919", "ticker_idx": 0, "ts": "2025-07-24T01:55:30+00:00", "symbol": "Delta Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Delta Electronics continues to strengthen its leading position in AI server power supplies and liquid cooling systems. Benefiting from the multi-fold expansion of the AI server market, the company has clear growth and valuation catalysts for the next two years", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM bullish on Delta Electronics: AI server power/liquid cooling share gains, 40% CAGR in cooling biz FY25-27, margins rising 11.3%→14% by 2027.", "resolved_tickers": ["2308.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Delta Electronics Taiwan 2308.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Delta Electronics: Increasing Market Share in AI Server Power and Liquid Cooling, Clear Upside for Performance and Valuation\nJPM (G. Hariharan, 2025/07/20)\n\nJPM states that Delta Electronics continues to strengthen its leading position in AI server power supplies and liquid cooling systems. Benefiting from the multi-fold expansion of the AI server market, the company has clear growth and valuation catalysts for the next two years.\n\nAI Servers Drive Revenue Mix Optimization, Multiple Catalysts in the Next Two Years\nAlthough 2Q revenue was lower than expected due to exchange rates and production line transfers, Delta Electronics' product portfolio optimization and improved AI server pricing led to gross margin improvement. The company anticipates increased GB300 cabinet shipments in the second half of the year, with power cabinet and HVDC architecture design changes driving a new round of performance in the next two years.\n\nHigh Market Share in Liquid Cooling Systems, Amazon IRHX a Key Catalyst\nDelta Electronics holds a higher-than-expected market share in Meta and Microsoft's L2A liquid cooling systems. Recently, Amazon announced its self-developed liquid cooling solution, reinforcing the trend of cloud service providers transitioning to an ODM model. JPM projects the liquid cooling business CAGR to reach 40% for FY25-27, with its revenue contribution potentially reaching 9% by 2027.\n\nHigh-Power Cabinets Drive Increased Power Value\nThe power value contributed by Delta Electronics in GB200 cabinets is $37,000, projected to rise to $53,000 for GB300. Future higher-power platforms like VR200 and Kyber are expected to push the value of a single power cabinet up to $150,000, and drive the full adoption of HVDC architecture.\n\nServer Business Structure Transformation Boosts Overall Profit Margins\nDelta Electronics' server-related revenue share is expected to increase from 25-30% in 2024 to 45-50% in 2027, boosting overall operating profit margins from 11.3% to 14%. High-value products such as liquid cooling and power cabinets are simultaneously driving up ASP and market share, strengthening the company's profitability", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0038610038610038533, "ret_signed_m1d": 0.0038610038610038533, "alpha_spy_m1d": 0.004191946824327886, "alpha_c_m1d": 0.008069347094884716, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.004208343233880862, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0019305019305019266, "ret_signed_p1d": -0.0019305019305019266, "alpha_spy_p1d": -0.006154851570821074, "alpha_c_p1d": -0.004225631517581374, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": 0.0022951295870794475, "ret_p1w": 0.09459459459459452, "ret_signed_p1w": 0.09459459459459452, "alpha_spy_p1w": 0.09828289619259012, "alpha_c_p1w": 0.08943005482147237, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.005164539773122145, "ret_p1m": 0.28185328185328196, "ret_signed_p1m": 0.28185328185328196, "alpha_spy_p1m": 0.26468790778197904, "alpha_c_p1m": 0.27791291278744157, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.003940369065840388, "ret_p3m": 0.9691119691119692, "ret_signed_p3m": 0.9691119691119692, "alpha_spy_p3m": 0.9080624520500209, "alpha_c_p3m": 0.86459184508424, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.10452012402772914, "ret_p6m": 1.1718146718146718, "ret_signed_p6m": 1.1718146718146718, "alpha_spy_p6m": 1.0753375698465946, "alpha_c_p6m": 1.0544935647929867, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.11732110702168508, "price_path": [-0.3542, -0.3551, -0.3674, -0.3674, -0.3285, -0.321, -0.3191, -0.3105, -0.2944, -0.2906, -0.2982, -0.302, -0.2944, -0.2973, -0.2963, -0.3115, -0.3067, -0.2821, -0.283, -0.2916, -0.2973, -0.3029, -0.2982, -0.2906, -0.2906, -0.3029, -0.2802, -0.2622, -0.2451, -0.2489, -0.246, -0.2375, -0.2375, -0.2422, -0.2451, -0.2403, -0.2375, -0.2268, -0.2201, -0.2278, -0.2278, -0.195, -0.1824, -0.1892, -0.1815, -0.2027, -0.1737, -0.1708, -0.1139, -0.1178, -0.1313, -0.1284, -0.1293, -0.0685, -0.0676, -0.0869, -0.0463, -0.0347, -0.0367, -0.0174, -0.0058, -0.0097, 0.0039, 0.0, -0.0019, 0.0077, 0.0135, 0.0232, 0.0946, 0.0946, 0.1448, 0.2201, 0.1988, 0.2027, 0.2297, 0.2317, 0.2394, 0.3031, 0.2934, 0.3243, 0.3282, 0.2761, 0.2066, 0.251, 0.2819, 0.305, 0.3224, 0.39, 0.3687, 0.3726, 0.3282, 0.3224, 0.3127, 0.334, 0.4015, 0.4015, 0.5135, 0.6158, 0.6042, 0.6158, 0.6139, 0.6236, 0.6139, 0.722, 0.7143, 0.751, 0.7239, 0.7201, 0.7066, 0.6371, 0.6371, 0.6486, 0.6815, 0.7259, 0.8185, 0.8185, 0.9054, 0.8938, 0.9286, 0.9286, 0.9884, 0.9151, 0.9595, 0.9788, 0.9286, 0.9595, 0.9691, 0.9691, 0.9498, 0.9498, 1.0656, 1.0753, 1.0463, 0.9498, 0.9208, 0.8977, 0.8822, 0.8668, 0.888, 0.8629, 0.8958, 0.8764, 0.8668, 0.8494, 0.7799, 0.7683, 0.7336, 0.7259, 0.8417, 0.7278, 0.7239, 0.7568, 0.8243, 0.8185, 0.7992, 0.8784, 0.9093, 0.9247, 0.9286, 0.89, 0.89, 0.8745, 0.8687, 0.8243, 0.8108, 0.7799, 0.7394, 0.7181, 0.7201, 0.7587, 0.8398, 0.8205, 0.8436, 0.8436, 0.8456, 0.8571, 0.8552, 0.8591, 0.8591, 0.9208, 0.9498, 1.027, 0.9884, 0.9498, 0.9595, 1.0367, 0.9884, 1.0463, 1.0174, 1.1718]}
{"unit_id": "orig:1973173076092395933", "ticker_idx": 0, "ts": "2025-09-30T23:48:15+00:00", "symbol": "memory", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "general DRAM ASP rose by +10–15% QoQ in 3Q25, and expects an additional +8–13% QoQ increase in 4Q25", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman/TrendForce: DRAM and NAND ASPs rising strongly into 4Q25; server demand upcycle with 2026 CSP bit-demand growth >25%; bullish memory pricing outlook.", "resolved_tickers": ["000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: SK Hynix, Samsung Electronics", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251001_South Korea Technology: September 2025 Memory Pricing – Goldman Sachs\n\n[Key Takeaways]\n    1. Solid server demand is driving the memory price upcycle.\n    2. TrendForce forecasts that in 2026, bit demand growth from U.S. CSPs will exceed +25%, while Chinese CSPs will exceed +20%.\n\n⸻\n\n[Contents]\n    1. TrendForce reports that general DRAM ASP rose by +10–15% QoQ in 3Q25, and expects an additional +8–13% QoQ increase in 4Q25.\n    2. For NAND ASP, prices rose +3–8% in 3Q25 and are projected to rise another +5–10% in 4Q25, supported by strong eSSD demand.\n    3. PC DRAM rose +8–13% QoQ in 3Q25. There were no new deals in September, while increased server DRAM supply kept PC DRAM supply tight.\n    4. PC OEMs, already holding sufficient inventories, may build up additional stock in anticipation of future price hikes. In 4Q25, PC DRAM prices are expected to rise +3–8% QoQ.\n    5. Server DRAM rose +3–8% QoQ in 3Q25. With Chinese CSPs requesting additional orders, server DRAM contract prices continued to rise.\n■ September: DDR5 contract prices MoM +1%, DDR4 contract prices MoM +2%.\n    6. In 2026, U.S. CSP bit demand growth is expected to exceed +25%, and Chinese CSPs over +20%.\n    7. TrendForce noted that with supply already tight, long-term contract discussions for server DRAM in 2026 have already begun.\n    8. In 4Q25, server DRAM prices are projected to rise +5–10% QoQ. However, there is a possibility that CSPs may have placed excessive orders to secure DRAM on time.\n■ DDR5: +5–10% / DDR4: +15–20%.\n    9. For LPDDR, demand is recovering due to rising smartphone content per device and the expanding applications of LPDDR.\n    10. In 3Q25, LPDDR4/5 prices rose +10–15% and +38–43% QoQ, respectively, and are forecast to rise another +8–13% and +15–20% in 4Q25.\n    11. In NAND, AI inference demand and HDD supply shortages are giving NAND manufacturers room to raise ASPs.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003946043277751832, "ret_signed_m1d": 0.003946043277751832, "alpha_spy_m1d": 0.007698774000249209, "alpha_c_m1d": 0.015283169441726696, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.030500512396686608, "ret_signed_p1d": 0.030500512396686608, "alpha_spy_p1d": 0.027092942970753597, "alpha_c_p1d": 0.008040592863877594, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.10392768042143374, "ret_signed_p1w": 0.10392768042143374, "alpha_spy_p1w": 0.09951448088586279, "alpha_c_p1w": 0.07117237652895314, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.40180491255419315, "ret_signed_p1m": 0.40180491255419315, "alpha_spy_p1m": 0.369966685396965, "alpha_c_p1m": 0.2723465910904568, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.5597464318755226, "ret_signed_p3m": 0.5597464318755226, "alpha_spy_p3m": 0.5204630958224853, "alpha_c_p3m": 0.4352262066919538, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 1.5671397216486262, "ret_signed_p6m": 1.5671397216486262, "alpha_spy_p6m": 1.575575994470528, "alpha_c_p6m": 1.3406977332818788, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.2213, -0.2358, -0.2446, -0.2305, -0.2379, -0.2114, -0.2055, -0.1982, -0.1932, -0.1908, -0.2171, -0.2154, -0.2062, -0.2226, -0.2196, -0.2212, -0.2268, -0.2059, -0.204, -0.1907, -0.1834, -0.2205, -0.2158, -0.2067, -0.2204, -0.2053, -0.2055, -0.1951, -0.1917, -0.174, -0.1779, -0.1779, -0.2004, -0.2068, -0.2146, -0.2291, -0.2158, -0.2029, -0.2072, -0.2076, -0.2007, -0.1994, -0.2306, -0.2152, -0.2082, -0.2021, -0.1941, -0.1855, -0.1614, -0.1319, -0.1228, -0.08, -0.0699, -0.0282, -0.0562, -0.0121, -0.0121, 0.0004, 0.0219, 0.0211, 0.0238, -0.0216, 0.0039, 0.0, 0.0305, 0.1039, 0.1039, 0.1039, 0.1039, 0.1039, 0.1039, 0.1784, 0.1531, 0.138, 0.1741, 0.2333, 0.2532, 0.2832, 0.2703, 0.2804, 0.2636, 0.3226, 0.3777, 0.3426, 0.4018, 0.4376, 0.445, 0.5542, 0.4683, 0.4326, 0.4444, 0.418, 0.4715, 0.5075, 0.5022, 0.4932, 0.385, 0.4715, 0.403, 0.3837, 0.4211, 0.3146, 0.3245, 0.3385, 0.3666, 0.4001, 0.3621, 0.3754, 0.4197, 0.4176, 0.4068, 0.4293, 0.4834, 0.461, 0.4888, 0.453, 0.4712, 0.4222, 0.3758, 0.4364, 0.4361, 0.4211, 0.4937, 0.5054, 0.5087, 0.5087, 0.5597, 0.6371, 0.6554, 0.6554, 0.6554, 0.7443, 0.8292, 0.8771, 0.9128, 0.9197, 0.9037, 0.9097, 0.8866, 0.9086, 0.9402, 0.9802, 0.9941, 0.9394, 0.9608, 0.9991, 1.0152, 0.9706, 1.1071, 1.1835, 1.2021, 1.27, 1.0958, 1.309, 1.3085, 1.1663, 1.1578, 1.2737, 1.2542, 1.2432, 1.3481, 1.3522, 1.3522, 1.3522, 1.3522, 1.4251, 1.5048, 1.5251, 1.6448, 1.6844, 1.8908, 1.827, 1.827, 1.5229, 1.2559, 1.5048, 1.4599, 1.245, 1.4783, 1.5154, 1.4668, 1.4116, 1.535, 1.5604, 1.7719, 1.662, 1.6467, 1.4615, 1.5583, 1.5671]}
{"unit_id": "orig:1973173076092395933", "ticker_idx": 1, "ts": "2025-09-30T23:48:15+00:00", "symbol": "memory", "kind": "sector", "geography": "Korea", "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND ASP, prices rose +3–8% in 3Q25 and are projected to rise another +5–10% in 4Q25, supported by strong eSSD demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman/TrendForce: DRAM and NAND ASPs rising strongly into 4Q25; server demand upcycle with 2026 CSP bit-demand growth >25%; bullish memory pricing outlook.", "resolved_tickers": ["000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Korea NAND sector: SK Hynix, Samsung Electronics", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251001_South Korea Technology: September 2025 Memory Pricing – Goldman Sachs\n\n[Key Takeaways]\n    1. Solid server demand is driving the memory price upcycle.\n    2. TrendForce forecasts that in 2026, bit demand growth from U.S. CSPs will exceed +25%, while Chinese CSPs will exceed +20%.\n\n⸻\n\n[Contents]\n    1. TrendForce reports that general DRAM ASP rose by +10–15% QoQ in 3Q25, and expects an additional +8–13% QoQ increase in 4Q25.\n    2. For NAND ASP, prices rose +3–8% in 3Q25 and are projected to rise another +5–10% in 4Q25, supported by strong eSSD demand.\n    3. PC DRAM rose +8–13% QoQ in 3Q25. There were no new deals in September, while increased server DRAM supply kept PC DRAM supply tight.\n    4. PC OEMs, already holding sufficient inventories, may build up additional stock in anticipation of future price hikes. In 4Q25, PC DRAM prices are expected to rise +3–8% QoQ.\n    5. Server DRAM rose +3–8% QoQ in 3Q25. With Chinese CSPs requesting additional orders, server DRAM contract prices continued to rise.\n■ September: DDR5 contract prices MoM +1%, DDR4 contract prices MoM +2%.\n    6. In 2026, U.S. CSP bit demand growth is expected to exceed +25%, and Chinese CSPs over +20%.\n    7. TrendForce noted that with supply already tight, long-term contract discussions for server DRAM in 2026 have already begun.\n    8. In 4Q25, server DRAM prices are projected to rise +5–10% QoQ. However, there is a possibility that CSPs may have placed excessive orders to secure DRAM on time.\n■ DDR5: +5–10% / DDR4: +15–20%.\n    9. For LPDDR, demand is recovering due to rising smartphone content per device and the expanding applications of LPDDR.\n    10. In 3Q25, LPDDR4/5 prices rose +10–15% and +38–43% QoQ, respectively, and are forecast to rise another +8–13% and +15–20% in 4Q25.\n    11. In NAND, AI inference demand and HDD supply shortages are giving NAND manufacturers room to raise ASPs.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003946043277751832, "ret_signed_m1d": 0.003946043277751832, "alpha_spy_m1d": 0.007698774000249209, "alpha_c_m1d": 0.015283169441726696, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.030500512396686608, "ret_signed_p1d": 0.030500512396686608, "alpha_spy_p1d": 0.027092942970753597, "alpha_c_p1d": 0.008040592863877594, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.10392768042143374, "ret_signed_p1w": 0.10392768042143374, "alpha_spy_p1w": 0.09951448088586279, "alpha_c_p1w": 0.07117237652895314, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.40180491255419315, "ret_signed_p1m": 0.40180491255419315, "alpha_spy_p1m": 0.369966685396965, "alpha_c_p1m": 0.2723465910904568, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.5597464318755226, "ret_signed_p3m": 0.5597464318755226, "alpha_spy_p3m": 0.5204630958224853, "alpha_c_p3m": 0.4352262066919538, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 1.5671397216486262, "ret_signed_p6m": 1.5671397216486262, "alpha_spy_p6m": 1.575575994470528, "alpha_c_p6m": 1.3406977332818788, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.2213, -0.2358, -0.2446, -0.2305, -0.2379, -0.2114, -0.2055, -0.1982, -0.1932, -0.1908, -0.2171, -0.2154, -0.2062, -0.2226, -0.2196, -0.2212, -0.2268, -0.2059, -0.204, -0.1907, -0.1834, -0.2205, -0.2158, -0.2067, -0.2204, -0.2053, -0.2055, -0.1951, -0.1917, -0.174, -0.1779, -0.1779, -0.2004, -0.2068, -0.2146, -0.2291, -0.2158, -0.2029, -0.2072, -0.2076, -0.2007, -0.1994, -0.2306, -0.2152, -0.2082, -0.2021, -0.1941, -0.1855, -0.1614, -0.1319, -0.1228, -0.08, -0.0699, -0.0282, -0.0562, -0.0121, -0.0121, 0.0004, 0.0219, 0.0211, 0.0238, -0.0216, 0.0039, 0.0, 0.0305, 0.1039, 0.1039, 0.1039, 0.1039, 0.1039, 0.1039, 0.1784, 0.1531, 0.138, 0.1741, 0.2333, 0.2532, 0.2832, 0.2703, 0.2804, 0.2636, 0.3226, 0.3777, 0.3426, 0.4018, 0.4376, 0.445, 0.5542, 0.4683, 0.4326, 0.4444, 0.418, 0.4715, 0.5075, 0.5022, 0.4932, 0.385, 0.4715, 0.403, 0.3837, 0.4211, 0.3146, 0.3245, 0.3385, 0.3666, 0.4001, 0.3621, 0.3754, 0.4197, 0.4176, 0.4068, 0.4293, 0.4834, 0.461, 0.4888, 0.453, 0.4712, 0.4222, 0.3758, 0.4364, 0.4361, 0.4211, 0.4937, 0.5054, 0.5087, 0.5087, 0.5597, 0.6371, 0.6554, 0.6554, 0.6554, 0.7443, 0.8292, 0.8771, 0.9128, 0.9197, 0.9037, 0.9097, 0.8866, 0.9086, 0.9402, 0.9802, 0.9941, 0.9394, 0.9608, 0.9991, 1.0152, 0.9706, 1.1071, 1.1835, 1.2021, 1.27, 1.0958, 1.309, 1.3085, 1.1663, 1.1578, 1.2737, 1.2542, 1.2432, 1.3481, 1.3522, 1.3522, 1.3522, 1.3522, 1.4251, 1.5048, 1.5251, 1.6448, 1.6844, 1.8908, 1.827, 1.827, 1.5229, 1.2559, 1.5048, 1.4599, 1.245, 1.4783, 1.5154, 1.4668, 1.4116, 1.535, 1.5604, 1.7719, 1.662, 1.6467, 1.4615, 1.5583, 1.5671]}
{"unit_id": "orig:1957647284159877214", "ticker_idx": 0, "ts": "2025-08-19T03:34:18+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Jeff Pu of Hong Kong's GF Securities downgraded AMD, citing that the company has implemented order cuts.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares GF Securities downgrade on AMD citing order cuts; adopted bearish stance.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Jeff Pu of Hong Kong’s GF Securities downgraded AMD, citing that the company has implemented order cuts.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05758028317123709, "ret_signed_m1d": -0.05758028317123709, "alpha_spy_m1d": -0.05212556052472439, "alpha_c_m1d": -0.0369869407216048, "bench_spy_m1d": 0.0054547226465127, "bench_c_m1d": 0.02059334244963229, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008105710470002747, "ret_signed_p1d": 0.008105710470002747, "alpha_spy_p1d": 0.005448626839356097, "alpha_c_p1d": 0.0015391757486598046, "bench_spy_p1d": -0.0026570836306466505, "bench_c_p1d": -0.006566534721342943, "ret_p1w": 0.0004202465574734138, "ret_signed_p1w": -0.0004202465574734138, "alpha_spy_p1w": 0.007941596301605847, "alpha_c_p1w": 0.019794951561003415, "bench_spy_p1w": 0.00836184285907926, "bench_c_p1w": 0.02021519811847683, "ret_p1m": -0.04437105526411722, "ret_signed_p1m": 0.04437105526411722, "alpha_spy_p1m": 0.07464568470922939, "alpha_c_p1m": 0.09380892735029844, "bench_spy_p1m": 0.030274629445112167, "bench_c_p1m": 0.04943787208618122, "ret_p3m": 0.4818972863176352, "ret_signed_p3m": -0.4818972863176352, "alpha_spy_p3m": -0.4287832245161731, "alpha_c_p3m": -0.29586911419660744, "bench_spy_p3m": 0.053114061801462054, "bench_c_p3m": 0.18602817212102774, "ret_p6m": 0.282377651861595, "ret_signed_p6m": -0.282377651861595, "alpha_spy_p6m": -0.1946659926240324, "alpha_c_p6m": 0.14805731515417886, "bench_spy_p6m": 0.08771165923756263, "bench_c_p6m": 0.4304349670157739, "price_path": [-0.3353, -0.3377, -0.3377, -0.3122, -0.3224, -0.3213, -0.3352, -0.3117, -0.2956, -0.288, -0.3054, -0.3024, -0.2691, -0.26, -0.2727, -0.2885, -0.3026, -0.2411, -0.2369, -0.2387, -0.2387, -0.23, -0.222, -0.1688, -0.139, -0.1373, -0.1365, -0.148, -0.1828, -0.1683, -0.172, -0.172, -0.1906, -0.1725, -0.169, -0.1344, -0.1209, -0.1219, -0.0657, -0.0388, -0.0369, -0.0574, -0.0573, -0.071, -0.0474, -0.0266, -0.0005, 0.0427, 0.0654, 0.0778, 0.0586, 0.0309, 0.0614, 0.0466, -0.0206, 0.0351, 0.0373, 0.0344, 0.0504, 0.1073, 0.0865, 0.0658, 0.0576, 0.0, -0.0081, -0.0171, 0.0073, -0.0192, 0.0004, 0.0035, 0.0122, -0.0235, -0.0235, -0.0254, -0.0265, -0.0286, -0.0925, -0.0909, -0.0644, -0.0421, -0.0653, -0.0479, -0.0324, -0.0366, -0.0444, -0.0518, -0.055, -0.0406, -0.0339, -0.034, -0.0317, -0.0426, -0.0312, -0.0286, -0.0153, 0.0191, -0.0113, 0.2231, 0.2699, 0.4144, 0.3983, 0.2903, 0.2994, 0.3095, 0.4326, 0.4083, 0.3995, 0.4444, 0.4292, 0.3823, 0.4109, 0.5186, 0.5591, 0.5491, 0.5871, 0.5301, 0.5378, 0.559, 0.5014, 0.5391, 0.4272, 0.4022, 0.4649, 0.4261, 0.5544, 0.4888, 0.4819, 0.4441, 0.3827, 0.3422, 0.237, 0.2235, 0.2912, 0.2376, 0.2863, 0.2863, 0.3061, 0.3195, 0.2923, 0.3065, 0.2968, 0.3087, 0.3276, 0.3307, 0.3295, 0.3295, 0.2656, 0.2464, 0.2559, 0.1895, 0.2072, 0.2815, 0.2906, 0.2903, 0.2911, 0.2911, 0.2908, 0.2946, 0.2929, 0.2859, 0.2859, 0.3418, 0.3274, 0.287, 0.261, 0.2289, 0.2199, 0.247, 0.3267, 0.3425, 0.3685, 0.392, 0.392, 0.3925, 0.4998, 0.5234, 0.5592, 0.5089, 0.5132, 0.5175, 0.5141, 0.4214, 0.4787, 0.4537, 0.202, 0.1558, 0.2515, 0.2969, 0.2823, 0.2824]}
{"unit_id": "thread:1951952435372765543", "ticker_idx": 0, "ts": "2025-08-03T10:25:00+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the market may be placing too much emphasis on its current growth figures, thereby underestimating the future potential signaled by its business backlog and its unique profit growth model", "tweet_type": "analysis", "timeline": "months", "summary": "Amazon undervalued as AWS backlog accelerates and margin mix improves; MSFT premium ignores cloud margin dilution risk; GOOG margin risk already priced in.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Why has Microsoft reached a $4 trillion valuation while Amazon seems to be lagging? Is it possible that the market is undervaluing Amazon?\n\nMicrosoft's market capitalization surging past $4 trillion seems to have dimmed Amazon's spotlight in the AI race. However, in the artificial intelligence-driven cloud competition, investors may need to shift their focus from growth speed to the deeper structure of profitability.\n\nThe AI cloud race is not just a battle of technology and growth; it is a reshaping of the profit models of tech giants. A recent analysis by The Information points out that in this competition, Microsoft and Google are facing margin pressures from their cloud business expansion. This very situation presents an opportunity for Amazon to enhance its overall profitability, a dynamic the market appears to misperceive.\n\nIn the latest earnings season, both Microsoft and Google announced accelerating growth in their cloud businesses. In comparison, Amazon Web Services' (AWS) 17% growth rate for the quarter seemed lackluster, but its business structure reveals a different story. AWS's profitability far surpasses that of its core e-commerce business, which means that every point of growth in its cloud division structurally optimizes Amazon's overall profit.\n\nThe Price of High Growth: Microsoft and Google's \"Sweet Sorrow\"\n\nFor Microsoft and Google, the robust growth of their cloud businesses is a double-edged sword.\n\nAccording to the latest quarterly data, Microsoft's \"Intelligent Cloud\" segment had a profit margin of 40.6%, whereas its \"Productivity and Business Processes\" segment, which includes commercial software, boasted a much higher margin of 57.4%. A similar pattern is evident at Google, where its cloud business reported a profit margin of 20.7%, significantly lower than the 40% margin of its \"Google Services\" segment, which is predominantly driven by advertising.\n\nThe issue is that for both companies, their lower-margin cloud businesses are growing at a much faster pace than their high-margin core operations. Data shows that Microsoft's cloud business grew by 26% in the quarter, far outpacing the 16% growth of its software business. Similarly, Google's cloud business expanded by 32%, well ahead of the 12% growth in its services segment.\n\nThe analysis suggests that as AI continues to drive enterprises toward the cloud, this \"growth imbalance\" will become more pronounced. In the long term, this implies that the overall profit margins of both companies could be \"diluted\" by their expanding cloud businesses, gradually converging toward the margin levels of their cloud operations. Furthermore, there is a potential risk that new AI products could cannibalize Microsoft's lucrative enterprise software and Google's search advertising businesses.\n\nIs Amazon's Potential Underestimated? The Cloud Business as a Profit Engine\n\nAmazon's situation is the complete opposite. Its cloud business, AWS, is the core engine of the company's profits. In the most recent quarter, AWS's operating margin stood at an impressive 33%, while its vast e-commerce business had a margin of just 6.6%.\n\nHistorical data corroborates this trend. From 2017 to 2024, as AWS's share of Amazon's total revenue increased from 9.8% to 17%, the company's overall operating margin leaped from 2.3% to 10.7%. The expansion of the cloud business has been the key driver of Amazon's leap in profitability.\n\nAlthough the market has expressed concern over AWS's 17% growth rate for the quarter, there are signs that its growth may be on the verge of accelerating. In a report, New Street Research analyst Dan Salmon noted that AWS's backlog (i.e., customer commitments for future orders) saw a substantial 25% increase during the quarter, a powerful indicator of future revenue.\n\nOf course, all cloud service providers face a common challenge: massive capital expenditures to support AI will increase depreciation expenses, thereby putting pressure on the margins of the cloud businesses themselves. However, the core issue lies in the long-term evolution of the business mix. For Google, the market seems to have already priced in the margin risk, with its stock performance lagging behind advertising peers like Meta. Yet, judging by the premium valuation of Microsoft's stock, investors appear to be selectively ignoring the same risk.\n\nAs for Amazon, the market may be placing too much emphasis on its current growth figures, thereby underestimating the future potential signaled by its business backlog and its unique profit growth model.\n\n$MSFT $AMZN $GOOG\n\n---\n\nhttps://t.co/wN7zuAoQz6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014646851844615671, "ret_signed_m1d": 0.014646851844615671, "alpha_spy_m1d": 0.02961909078371039, "alpha_c_m1d": 0.026264707857832525, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.011617856013216854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009922070229506863, "ret_signed_p1d": 0.009922070229506863, "alpha_spy_p1d": 0.014992036801938169, "alpha_c_p1d": 0.00864136003446414, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": 0.0012807101950427224, "ret_p1w": 0.04559418584246755, "ret_signed_p1w": 0.04559418584246755, "alpha_spy_p1w": 0.038068423735933, "alpha_c_p1w": 0.019385499363271386, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.026208686479196164, "ret_p1m": 0.06468227184595099, "ret_signed_p1m": 0.06468227184595099, "alpha_spy_p1m": 0.05026454221536669, "alpha_c_p1m": 0.012630896309151263, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.05205137553679973, "ret_p3m": 0.052964833626925945, "ret_signed_p3m": 0.052964833626925945, "alpha_spy_p3m": -0.027116316265054063, "alpha_c_p3m": -0.018078817777868128, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.07104365140479407, "ret_p6m": 0.1560595309794881, "ret_signed_p6m": 0.1560595309794881, "alpha_spy_p6m": 0.047833474132629394, "alpha_c_p6m": 0.029498970339387087, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.12656056064010102, "price_path": [-0.1084, -0.0925, -0.0878, -0.0142, -0.0013, -0.0066, -0.0306, -0.0286, -0.0259, -0.0358, -0.0498, -0.0404, -0.0504, -0.0504, -0.0266, -0.0327, -0.0281, -0.0314, -0.0236, -0.0281, -0.0209, -0.0177, 0.0091, 0.0252, 0.0282, 0.0073, 0.0075, 0.0021, 0.021, 0.015, 0.0041, 0.0041, -0.0093, -0.015, 0.0053, 0.0016, 0.0258, 0.055, 0.0366, 0.0416, 0.0391, 0.0556, 0.0556, 0.0558, 0.0364, 0.0515, 0.0501, 0.0632, 0.0663, 0.0695, 0.0545, 0.0578, 0.0684, 0.0834, 0.0747, 0.0786, 0.0972, 0.0935, 0.0999, 0.0915, 0.0876, 0.1061, 0.0146, 0.0, 0.0099, 0.0504, 0.0542, 0.0522, 0.0456, 0.0464, 0.061, 0.0913, 0.0916, 0.0937, 0.0773, 0.0575, 0.0487, 0.0812, 0.077, 0.0806, 0.0825, 0.0943, 0.082, 0.082, 0.0647, 0.0678, 0.1135, 0.0977, 0.1143, 0.1256, 0.0883, 0.0865, 0.078, 0.0935, 0.1058, 0.0944, 0.0925, 0.0937, 0.0755, 0.0428, 0.0404, 0.0307, 0.0384, 0.0497, 0.0374, 0.0424, 0.0508, 0.0371, 0.0437, 0.0479, 0.0641, 0.076, 0.0223, 0.0398, 0.0224, 0.0185, 0.0133, 0.0066, 0.0228, 0.049, 0.0298, 0.0446, 0.0593, 0.0724, 0.0832, 0.0881, 0.053, 0.1539, 0.2001, 0.178, 0.1821, 0.1483, 0.1548, 0.1736, 0.1769, 0.1538, 0.1225, 0.1089, 0.1003, 0.0515, 0.0522, 0.0259, 0.0427, 0.0691, 0.0851, 0.0827, 0.0827, 0.1019, 0.105, 0.1076, 0.0979, 0.0825, 0.0845, 0.072, 0.0769, 0.0951, 0.088, 0.0687, 0.0515, 0.0515, 0.0455, 0.0714, 0.0742, 0.0793, 0.0968, 0.0979, 0.0979, 0.0986, 0.0965, 0.0987, 0.0906, 0.0906, 0.0702, 0.1012, 0.1383, 0.1413, 0.1637, 0.1688, 0.1645, 0.1462, 0.1181, 0.1253, 0.1298, 0.1298, 0.0914, 0.0929, 0.1072, 0.13, 0.1265, 0.1561]}
{"unit_id": "thread:1951952435372765543", "ticker_idx": 1, "ts": "2025-08-03T10:25:00+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "judging by the premium valuation of Microsoft's stock, investors appear to be selectively ignoring the same risk", "tweet_type": "analysis", "timeline": "months", "summary": "Amazon undervalued as AWS backlog accelerates and margin mix improves; MSFT premium ignores cloud margin dilution risk; GOOG margin risk already priced in.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Why has Microsoft reached a $4 trillion valuation while Amazon seems to be lagging? Is it possible that the market is undervaluing Amazon?\n\nMicrosoft's market capitalization surging past $4 trillion seems to have dimmed Amazon's spotlight in the AI race. However, in the artificial intelligence-driven cloud competition, investors may need to shift their focus from growth speed to the deeper structure of profitability.\n\nThe AI cloud race is not just a battle of technology and growth; it is a reshaping of the profit models of tech giants. A recent analysis by The Information points out that in this competition, Microsoft and Google are facing margin pressures from their cloud business expansion. This very situation presents an opportunity for Amazon to enhance its overall profitability, a dynamic the market appears to misperceive.\n\nIn the latest earnings season, both Microsoft and Google announced accelerating growth in their cloud businesses. In comparison, Amazon Web Services' (AWS) 17% growth rate for the quarter seemed lackluster, but its business structure reveals a different story. AWS's profitability far surpasses that of its core e-commerce business, which means that every point of growth in its cloud division structurally optimizes Amazon's overall profit.\n\nThe Price of High Growth: Microsoft and Google's \"Sweet Sorrow\"\n\nFor Microsoft and Google, the robust growth of their cloud businesses is a double-edged sword.\n\nAccording to the latest quarterly data, Microsoft's \"Intelligent Cloud\" segment had a profit margin of 40.6%, whereas its \"Productivity and Business Processes\" segment, which includes commercial software, boasted a much higher margin of 57.4%. A similar pattern is evident at Google, where its cloud business reported a profit margin of 20.7%, significantly lower than the 40% margin of its \"Google Services\" segment, which is predominantly driven by advertising.\n\nThe issue is that for both companies, their lower-margin cloud businesses are growing at a much faster pace than their high-margin core operations. Data shows that Microsoft's cloud business grew by 26% in the quarter, far outpacing the 16% growth of its software business. Similarly, Google's cloud business expanded by 32%, well ahead of the 12% growth in its services segment.\n\nThe analysis suggests that as AI continues to drive enterprises toward the cloud, this \"growth imbalance\" will become more pronounced. In the long term, this implies that the overall profit margins of both companies could be \"diluted\" by their expanding cloud businesses, gradually converging toward the margin levels of their cloud operations. Furthermore, there is a potential risk that new AI products could cannibalize Microsoft's lucrative enterprise software and Google's search advertising businesses.\n\nIs Amazon's Potential Underestimated? The Cloud Business as a Profit Engine\n\nAmazon's situation is the complete opposite. Its cloud business, AWS, is the core engine of the company's profits. In the most recent quarter, AWS's operating margin stood at an impressive 33%, while its vast e-commerce business had a margin of just 6.6%.\n\nHistorical data corroborates this trend. From 2017 to 2024, as AWS's share of Amazon's total revenue increased from 9.8% to 17%, the company's overall operating margin leaped from 2.3% to 10.7%. The expansion of the cloud business has been the key driver of Amazon's leap in profitability.\n\nAlthough the market has expressed concern over AWS's 17% growth rate for the quarter, there are signs that its growth may be on the verge of accelerating. In a report, New Street Research analyst Dan Salmon noted that AWS's backlog (i.e., customer commitments for future orders) saw a substantial 25% increase during the quarter, a powerful indicator of future revenue.\n\nOf course, all cloud service providers face a common challenge: massive capital expenditures to support AI will increase depreciation expenses, thereby putting pressure on the margins of the cloud businesses themselves. However, the core issue lies in the long-term evolution of the business mix. For Google, the market seems to have already priced in the margin risk, with its stock performance lagging behind advertising peers like Meta. Yet, judging by the premium valuation of Microsoft's stock, investors appear to be selectively ignoring the same risk.\n\nAs for Amazon, the market may be placing too much emphasis on its current growth figures, thereby underestimating the future potential signaled by its business backlog and its unique profit growth model.\n\n$MSFT $AMZN $GOOG\n\n---\n\nhttps://t.co/wN7zuAoQz6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021525586457659895, "ret_signed_m1d": 0.021525586457659895, "alpha_spy_m1d": 0.006553347518565178, "alpha_c_m1d": 0.006553347518565178, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.014972238939094717, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014730112378469884, "ret_signed_p1d": 0.014730112378469884, "alpha_spy_p1d": 0.009660145806038578, "alpha_c_p1d": 0.009660145806038578, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.005069966572431306, "ret_p1w": -0.025894335225087106, "ret_signed_p1w": 0.025894335225087106, "alpha_spy_p1w": 0.033420097331621657, "alpha_c_p1w": 0.033420097331621657, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.00752576210653455, "ret_p1m": -0.0554283564139092, "ret_signed_p1m": 0.0554283564139092, "alpha_spy_p1m": 0.0698460860444935, "alpha_c_p1m": 0.0698460860444935, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.014417729630584297, "ret_p3m": -0.016831673956911875, "ret_signed_p3m": 0.016831673956911875, "alpha_spy_p3m": 0.09691282384889188, "alpha_c_p3m": 0.09691282384889188, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.08008114989198001, "ret_p6m": -0.09963597971740967, "ret_signed_p6m": 0.09963597971740967, "alpha_spy_p6m": 0.20786203656426838, "alpha_c_p6m": 0.20786203656426838, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.10822605684685871, "price_path": [-0.1925, -0.1835, -0.1824, -0.1628, -0.163, -0.1559, -0.154, -0.1519, -0.1433, -0.1446, -0.1551, -0.1508, -0.1595, -0.1595, -0.1399, -0.1461, -0.1437, -0.1405, -0.1375, -0.1357, -0.134, -0.1269, -0.1218, -0.1174, -0.1208, -0.1177, -0.106, -0.1133, -0.1055, -0.1075, -0.1034, -0.1034, -0.1087, -0.0927, -0.085, -0.081, -0.0713, -0.0741, -0.0714, -0.0814, -0.0832, -0.0687, -0.0687, -0.0708, -0.0728, -0.06, -0.0638, -0.0603, -0.0609, -0.0557, -0.056, -0.0447, -0.0478, -0.0478, -0.0567, -0.0556, -0.0462, -0.0409, -0.0432, -0.0431, -0.0418, -0.004, -0.0215, 0.0, -0.0147, -0.02, -0.0276, -0.0254, -0.0259, -0.0119, -0.0281, -0.0246, -0.0289, -0.0346, -0.0483, -0.0559, -0.0571, -0.0515, -0.057, -0.0612, -0.0524, -0.047, -0.0525, -0.0525, -0.0554, -0.055, -0.0501, -0.0744, -0.0684, -0.068, -0.0643, -0.0631, -0.0465, -0.0363, -0.0481, -0.0463, -0.0492, -0.0315, -0.038, -0.0477, -0.046, -0.0519, -0.0436, -0.0377, -0.0314, -0.0281, -0.0356, -0.0326, -0.0116, -0.0202, -0.0185, -0.0231, -0.0445, -0.0387, -0.0396, -0.0399, -0.0433, -0.0396, -0.0336, -0.032, -0.0266, -0.0266, -0.0209, -0.0061, 0.0137, 0.0127, -0.0168, -0.0317, -0.0332, -0.0382, -0.0516, -0.0704, -0.0709, -0.0538, -0.0488, -0.0442, -0.0589, -0.046, -0.051, -0.0766, -0.0891, -0.1037, -0.1155, -0.112, -0.1064, -0.0904, -0.0904, -0.0782, -0.0881, -0.082, -0.105, -0.0991, -0.0948, -0.0801, -0.0782, -0.1034, -0.0942, -0.1035, -0.1104, -0.1075, -0.108, -0.0933, -0.0896, -0.0915, -0.0879, -0.0857, -0.0857, -0.0863, -0.0874, -0.0867, -0.0939, -0.0939, -0.1139, -0.1141, -0.1035, -0.0942, -0.1043, -0.1021, -0.106, -0.1182, -0.1394, -0.1444, -0.1385, -0.1385, -0.1485, -0.168, -0.1548, -0.127, -0.1189, -0.0996]}
{"unit_id": "thread:1951952435372765543", "ticker_idx": 2, "ts": "2025-08-03T10:25:00+00:00", "symbol": "GOOG", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "For Google, the market seems to have already priced in the margin risk, with its stock performance lagging behind advertising peers like Meta", "tweet_type": "analysis", "timeline": "months", "summary": "Amazon undervalued as AWS backlog accelerates and margin mix improves; MSFT premium ignores cloud margin dilution risk; GOOG margin risk already priced in.", "resolved_tickers": ["GOOG"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Why has Microsoft reached a $4 trillion valuation while Amazon seems to be lagging? Is it possible that the market is undervaluing Amazon?\n\nMicrosoft's market capitalization surging past $4 trillion seems to have dimmed Amazon's spotlight in the AI race. However, in the artificial intelligence-driven cloud competition, investors may need to shift their focus from growth speed to the deeper structure of profitability.\n\nThe AI cloud race is not just a battle of technology and growth; it is a reshaping of the profit models of tech giants. A recent analysis by The Information points out that in this competition, Microsoft and Google are facing margin pressures from their cloud business expansion. This very situation presents an opportunity for Amazon to enhance its overall profitability, a dynamic the market appears to misperceive.\n\nIn the latest earnings season, both Microsoft and Google announced accelerating growth in their cloud businesses. In comparison, Amazon Web Services' (AWS) 17% growth rate for the quarter seemed lackluster, but its business structure reveals a different story. AWS's profitability far surpasses that of its core e-commerce business, which means that every point of growth in its cloud division structurally optimizes Amazon's overall profit.\n\nThe Price of High Growth: Microsoft and Google's \"Sweet Sorrow\"\n\nFor Microsoft and Google, the robust growth of their cloud businesses is a double-edged sword.\n\nAccording to the latest quarterly data, Microsoft's \"Intelligent Cloud\" segment had a profit margin of 40.6%, whereas its \"Productivity and Business Processes\" segment, which includes commercial software, boasted a much higher margin of 57.4%. A similar pattern is evident at Google, where its cloud business reported a profit margin of 20.7%, significantly lower than the 40% margin of its \"Google Services\" segment, which is predominantly driven by advertising.\n\nThe issue is that for both companies, their lower-margin cloud businesses are growing at a much faster pace than their high-margin core operations. Data shows that Microsoft's cloud business grew by 26% in the quarter, far outpacing the 16% growth of its software business. Similarly, Google's cloud business expanded by 32%, well ahead of the 12% growth in its services segment.\n\nThe analysis suggests that as AI continues to drive enterprises toward the cloud, this \"growth imbalance\" will become more pronounced. In the long term, this implies that the overall profit margins of both companies could be \"diluted\" by their expanding cloud businesses, gradually converging toward the margin levels of their cloud operations. Furthermore, there is a potential risk that new AI products could cannibalize Microsoft's lucrative enterprise software and Google's search advertising businesses.\n\nIs Amazon's Potential Underestimated? The Cloud Business as a Profit Engine\n\nAmazon's situation is the complete opposite. Its cloud business, AWS, is the core engine of the company's profits. In the most recent quarter, AWS's operating margin stood at an impressive 33%, while its vast e-commerce business had a margin of just 6.6%.\n\nHistorical data corroborates this trend. From 2017 to 2024, as AWS's share of Amazon's total revenue increased from 9.8% to 17%, the company's overall operating margin leaped from 2.3% to 10.7%. The expansion of the cloud business has been the key driver of Amazon's leap in profitability.\n\nAlthough the market has expressed concern over AWS's 17% growth rate for the quarter, there are signs that its growth may be on the verge of accelerating. In a report, New Street Research analyst Dan Salmon noted that AWS's backlog (i.e., customer commitments for future orders) saw a substantial 25% increase during the quarter, a powerful indicator of future revenue.\n\nOf course, all cloud service providers face a common challenge: massive capital expenditures to support AI will increase depreciation expenses, thereby putting pressure on the margins of the cloud businesses themselves. However, the core issue lies in the long-term evolution of the business mix. For Google, the market seems to have already priced in the margin risk, with its stock performance lagging behind advertising peers like Meta. Yet, judging by the premium valuation of Microsoft's stock, investors appear to be selectively ignoring the same risk.\n\nAs for Amazon, the market may be placing too much emphasis on its current growth figures, thereby underestimating the future potential signaled by its business backlog and its unique profit growth model.\n\n$MSFT $AMZN $GOOG\n\n---\n\nhttps://t.co/wN7zuAoQz6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029629738427260355, "ret_signed_m1d": -0.029629738427260355, "alpha_spy_m1d": -0.014657499488165637, "alpha_c_m1d": -0.010768714751138164, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.01886102367612219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0021967456944589348, "ret_signed_p1d": -0.0021967456944589348, "alpha_spy_p1d": 0.002873220877972371, "alpha_c_p1d": 0.005569604268524819, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.007766349962983754, "ret_p1w": 0.030038259063021266, "ret_signed_p1w": 0.030038259063021266, "alpha_spy_p1w": 0.022512496956486716, "alpha_c_p1w": 0.02985346300548919, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.00018479605753207728, "ret_p1m": 0.0829628925917294, "ret_signed_p1m": 0.0829628925917294, "alpha_spy_p1m": 0.0685451629611451, "alpha_c_p1m": 0.05485650933410979, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.028106383257619605, "ret_p3m": 0.4413890857880054, "ret_signed_p3m": 0.4413890857880054, "alpha_spy_p3m": 0.3613079358960254, "alpha_c_p3m": 0.3807503325872008, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.060638753200804585, "ret_p6m": 0.7140135234471234, "ret_signed_p6m": 0.7140135234471234, "alpha_spy_p6m": 0.6057874666002647, "alpha_c_p6m": 0.6259268565182943, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.08808666692882916, "price_path": [-0.2203, -0.2053, -0.2123, -0.1858, -0.1791, -0.1489, -0.1561, -0.1457, -0.1435, -0.1565, -0.1323, -0.1225, -0.1347, -0.1347, -0.1123, -0.1153, -0.1175, -0.118, -0.1307, -0.1443, -0.1357, -0.1336, -0.1075, -0.0926, -0.0804, -0.0866, -0.0959, -0.1015, -0.091, -0.0946, -0.1112, -0.1112, -0.1431, -0.1519, -0.1431, -0.1239, -0.1089, -0.0893, -0.0938, -0.0962, -0.0817, -0.0777, -0.0777, -0.0929, -0.1052, -0.0924, -0.0871, -0.0738, -0.0661, -0.0646, -0.0612, -0.0564, -0.0501, -0.0235, -0.0186, -0.0217, -0.013, -0.0085, -0.0119, 0.0035, 0.0086, -0.0148, -0.0296, 0.0, -0.0022, 0.006, 0.0078, 0.0324, 0.03, 0.043, 0.0372, 0.0412, 0.0468, 0.0436, 0.0344, 0.0227, 0.0249, 0.056, 0.0685, 0.0623, 0.0637, 0.0849, 0.0908, 0.0908, 0.083, 0.1806, 0.1886, 0.2014, 0.1973, 0.2268, 0.2249, 0.2311, 0.2342, 0.2873, 0.2855, 0.2775, 0.2902, 0.3051, 0.293, 0.2902, 0.2672, 0.2607, 0.2639, 0.2494, 0.2453, 0.2555, 0.26, 0.2601, 0.286, 0.2636, 0.2551, 0.2384, 0.2143, 0.2509, 0.2588, 0.287, 0.2879, 0.2977, 0.3142, 0.2851, 0.2912, 0.2974, 0.332, 0.3802, 0.3725, 0.407, 0.4414, 0.441, 0.4527, 0.4218, 0.456, 0.459, 0.4301, 0.4858, 0.4917, 0.4697, 0.4272, 0.4162, 0.4603, 0.457, 0.4981, 0.4827, 0.5321, 0.6284, 0.6548, 0.6376, 0.6376, 0.6368, 0.6112, 0.6158, 0.6394, 0.628, 0.6469, 0.6089, 0.6258, 0.6424, 0.605, 0.5888, 0.5826, 0.5745, 0.525, 0.5541, 0.579, 0.5929, 0.6152, 0.6151, 0.6151, 0.6115, 0.6086, 0.6094, 0.6055, 0.6055, 0.6133, 0.6236, 0.6094, 0.6497, 0.668, 0.684, 0.7024, 0.7213, 0.7207, 0.7046, 0.6902, 0.6902, 0.6483, 0.6801, 0.6927, 0.6804, 0.7068, 0.714]}
{"unit_id": "quote:1970326732373926228", "ticker_idx": 0, "ts": "2025-09-23T03:17:54+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS says that the OAI–AVGO deal has a volume target even larger than the one between Google and AVGO", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS view: OpenAI–NVDA deal = ~$400B revenue; OAI–AVGO deal volume exceeds Google–AVGO, bullish both.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting. UBS says that the OAI–AVGO deal has a volume target even larger than the one between Google and AVGO.", "reply_text": null, "parent_text": null, "quoted_text": "(Must read) UBS Commentary on the OpenAI–NVDA Deal:\n\nWhile timing will, to a large degree, depend on OpenAI's ramp plan, we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period. Given the sheer size of the numbers, it does seem likely", "parent_author": null, "ret_m1d": -0.00044250703604264974, "ret_signed_m1d": -0.00044250703604264974, "alpha_spy_m1d": -0.005915775474888529, "alpha_c_m1d": -0.0019034776713225954, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.001091668021910408, "ret_signed_p1d": 0.001091668021910408, "alpha_spy_p1d": 0.004273301474524094, "alpha_c_p1d": 0.0029257556936610474, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.026641906763270273, "ret_signed_p1w": -0.026641906763270273, "alpha_spy_p1w": -0.031120035485962316, "alpha_c_p1w": -0.04112733592922291, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.004012485659201204, "ret_signed_p1m": 0.004012485659201204, "alpha_spy_p1m": -0.0029083260423534707, "alpha_c_p1m": -0.04852088956931033, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.004189488473618219, "ret_signed_p3m": 0.004189488473618219, "alpha_spy_p3m": -0.025048664139091015, "alpha_c_p3m": -0.10314651012640774, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": -0.06610467519757557, "ret_signed_p6m": -0.06610467519757557, "alpha_spy_p6m": -0.06636762087592318, "alpha_c_p6m": -0.29363013084808154, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.2043, -0.2067, -0.1881, -0.2203, -0.2051, -0.1895, -0.1895, -0.1924, -0.1995, -0.1815, -0.1889, -0.1919, -0.1883, -0.1725, -0.1729, -0.1563, -0.1655, -0.1511, -0.1795, -0.1644, -0.1497, -0.1453, -0.1332, -0.124, -0.1087, -0.135, -0.1499, -0.1231, -0.1372, -0.1115, -0.1053, -0.1018, -0.1049, -0.0786, -0.0896, -0.0833, -0.0977, -0.0994, -0.1314, -0.1424, -0.147, -0.1341, -0.1334, -0.1223, -0.1157, -0.0909, -0.1241, -0.1241, -0.1216, -0.1094, -0.0984, -0.0136, 0.0181, -0.0084, 0.0885, 0.0592, 0.0599, 0.0724, 0.0603, 0.0196, 0.0172, 0.016, -0.0004, 0.0, 0.0011, -0.0084, -0.013, -0.0326, -0.0266, -0.0164, -0.0022, -0.0017, -0.0102, -0.0075, 0.0194, 0.0179, -0.0422, 0.0524, 0.0153, 0.0366, 0.0449, 0.0307, 0.0304, 0.011, 0.004, 0.0158, 0.0448, 0.0682, 0.1004, 0.1388, 0.1107, 0.0905, 0.0697, 0.0384, 0.0591, 0.0491, 0.0309, 0.0574, 0.0384, 0.048, 0.0031, 0.0104, 0.0109, 0.0046, 0.0457, 0.0232, 0.0037, 0.1151, 0.136, 0.173, 0.173, 0.1889, 0.1391, 0.1258, 0.1229, 0.1242, 0.1514, 0.1834, 0.1987, 0.2184, 0.1989, 0.0619, 0.0026, 0.007, -0.0381, -0.0267, 0.0042, 0.0093, 0.0326, 0.0353, 0.0353, 0.0409, 0.0328, 0.0342, 0.0231, 0.0231, 0.0276, 0.0152, 0.0162, 0.0154, -0.0172, 0.0197, 0.0411, 0.0482, 0.0047, 0.014, 0.0397, 0.0397, -0.0168, -0.0281, -0.0378, -0.0539, -0.0397, -0.0163, -0.0149, -0.0224, -0.0207, -0.0212, -0.0531, -0.0894, -0.0821, -0.0159, 0.0167, 0.0063, 0.0132, -0.0211, -0.0388, -0.0388, -0.017, -0.0141, -0.0127, -0.0167, -0.0235, -0.0378, -0.0177, -0.049, -0.0554, -0.0576, -0.0723, -0.0614, -0.0163, -0.0231, 0.022, 0.0127, 0.0097, -0.0069, -0.0477, -0.0395, -0.0502, -0.0661]}
{"unit_id": "quote:1970326732373926228", "ticker_idx": 1, "ts": "2025-09-23T03:17:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS view: OpenAI–NVDA deal = ~$400B revenue; OAI–AVGO deal volume exceeds Google–AVGO, bullish both.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting. UBS says that the OAI–AVGO deal has a volume target even larger than the one between Google and AVGO.", "reply_text": null, "parent_text": null, "quoted_text": "(Must read) UBS Commentary on the OpenAI–NVDA Deal:\n\nWhile timing will, to a large degree, depend on OpenAI's ramp plan, we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period. Given the sheer size of the numbers, it does seem likely", "parent_author": null, "ret_m1d": 0.02903106179928261, "ret_signed_m1d": 0.02903106179928261, "alpha_spy_m1d": 0.02355779336043673, "alpha_c_m1d": 0.027570091164002664, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008182476268150385, "ret_signed_p1d": -0.008182476268150385, "alpha_spy_p1d": -0.005000842815536699, "alpha_c_p1d": -0.0063483885963997455, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.04567619863458905, "ret_signed_p1w": 0.04567619863458905, "alpha_spy_p1w": 0.041198069911897006, "alpha_c_p1w": 0.031190769468636415, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.01036827301103349, "ret_signed_p1m": 0.01036827301103349, "alpha_spy_p1m": 0.0034474613094788165, "alpha_c_p1m": -0.042165102217478045, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.014403919022290212, "ret_signed_p3m": 0.014403919022290212, "alpha_spy_p3m": -0.014834233590419021, "alpha_c_p3m": -0.09293207957773575, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 0.011151779837404119, "ret_signed_p6m": 0.011151779837404119, "alpha_spy_p6m": 0.010888834159056504, "alpha_c_p6m": -0.21637367581310185, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.1312, -0.1159, -0.1146, -0.1409, -0.1188, -0.107, -0.107, -0.1132, -0.1033, -0.0872, -0.0804, -0.0758, -0.0805, -0.0434, -0.0396, -0.0305, -0.0338, -0.0396, -0.0639, -0.0429, -0.0263, -0.0277, -0.0095, -0.0164, 0.0047, -0.0032, -0.0265, 0.0087, -0.001, 0.0055, 0.0131, 0.0239, 0.0203, 0.0265, 0.0177, 0.0201, 0.0113, 0.02, -0.0157, -0.017, -0.0194, -0.0025, 0.0077, 0.0187, 0.0177, 0.0097, -0.0239, -0.0239, -0.0429, -0.0438, -0.038, -0.064, -0.0568, -0.043, -0.0062, -0.0071, -0.0034, -0.0038, -0.0199, -0.0456, -0.0123, -0.0099, 0.029, 0.0, -0.0082, -0.0041, -0.0013, 0.0192, 0.0457, 0.0494, 0.0586, 0.0515, 0.0398, 0.037, 0.0599, 0.0792, 0.0265, 0.0554, 0.009, 0.0078, 0.0189, 0.0268, 0.0236, 0.0153, 0.0104, 0.0209, 0.0439, 0.0732, 0.1267, 0.1603, 0.1371, 0.1348, 0.1594, 0.1135, 0.094, 0.0541, 0.0545, 0.1156, 0.0826, 0.0861, 0.0472, 0.0658, 0.0458, 0.0164, 0.0453, 0.0124, 0.0025, 0.0231, -0.0034, 0.0103, 0.0103, -0.008, 0.0084, 0.017, 0.0065, 0.0278, 0.0224, 0.04, 0.0367, 0.03, 0.0141, -0.0191, -0.0119, -0.0039, -0.0419, -0.024, 0.0144, 0.0295, 0.0605, 0.0571, 0.0571, 0.0679, 0.0549, 0.0511, 0.0453, 0.0453, 0.0585, 0.0544, 0.0494, 0.0599, 0.0371, 0.0361, 0.0365, 0.0414, 0.0265, 0.0484, 0.0438, 0.0438, -0.002, 0.0275, 0.036, 0.0518, 0.0451, 0.0566, 0.0734, 0.079, 0.0712, 0.0403, 0.0108, -0.0237, -0.0367, 0.0392, 0.0651, 0.0567, 0.0652, 0.0478, 0.0246, 0.0246, 0.0367, 0.0536, 0.0531, 0.0639, 0.0736, 0.0809, 0.0961, 0.0363, -0.0069, 0.0228, 0.0091, 0.0259, 0.0276, -0.0034, 0.0237, 0.0356, 0.0427, 0.0265, 0.0103, 0.027, 0.0197, 0.0112]}
{"unit_id": "quote:1944305774559474106", "ticker_idx": 0, "ts": "2025-07-13T07:59:54+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We believe HBM-related strength should also be accretive to MU's margins as HBM commands a much higher premium compared to traditional DRAM... The net result is a significantly stronger GM profile.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses Deutsche Buy initiation on MU at $150 PT; HBM margins ~60% vs non-HBM ~35% drive strong gross margin expansion.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "We believe HBM-related strength should also be accretive to MU's margins as HBM  commands a much higher premium compared to traditional DRAM (DBe HBM  ASPs are ≥5x greater than conventional DRAM).  HBM offers significant  performance improvements vs. standard DRAM, which drives pricing substantially  higher, more-than-offsetting the higher manufacturing costs associated with a  more complex and silicon-intensive technology.  The net result is a significantly  stronger GM profile.  As per our estimates, MU's HBM GMs currently stand in the  ~60% range, while non-HBM GMs stand at ~35%.", "reply_text": null, "parent_text": null, "quoted_text": "Deutsche initiation on Micron $MU\n\nWe are launching coverage of Micron (Ticker: MU) with Buy rating and $150 price  target.  We are constructive on both the cyclical and secular setup facing Micron.  On  the secular side, we believe High Bandwidth Memory (HBM) is an underrated", "parent_author": null, "ret_m1d": 0.04991154298359102, "ret_signed_m1d": 0.04991154298359102, "alpha_spy_m1d": 0.051816026232916035, "alpha_c_m1d": 0.04230615413884076, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012646544778240454, "ret_signed_p1d": 0.012646544778240454, "alpha_spy_p1d": 0.016919711828867534, "alpha_c_p1d": -0.006559891879551305, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": -0.045358699256708546, "ret_signed_p1w": -0.045358699256708546, "alpha_spy_p1w": -0.05169673124142693, "alpha_c_p1w": -0.06354875940993221, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.07705931824968992, "ret_signed_p1m": 0.07705931824968992, "alpha_spy_p1m": 0.04844241817742945, "alpha_c_p1m": 0.025257913476379068, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.6225482680796428, "ret_signed_p3m": 0.6225482680796428, "alpha_spy_p3m": 0.545387536175657, "alpha_c_p3m": 0.4126437203500206, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 1.8984393378890543, "ret_signed_p6m": 1.8984393378890543, "alpha_spy_p6m": 1.7848561027008398, "alpha_c_p6m": 1.535667386361405, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.416, -0.4205, -0.4205, -0.4378, -0.4086, -0.3857, -0.3479, -0.328, -0.3383, -0.3524, -0.3518, -0.3449, -0.3201, -0.3226, -0.3219, -0.3041, -0.2828, -0.2768, -0.2226, -0.1836, -0.1971, -0.196, -0.1745, -0.1691, -0.1737, -0.1927, -0.2012, -0.2135, -0.2135, -0.1882, -0.1899, -0.1846, -0.2044, -0.173, -0.1387, -0.1303, -0.1047, -0.0856, -0.0655, -0.0386, -0.0227, -0.0214, -0.0263, 0.0094, 0.0136, 0.0261, 0.0261, 0.0411, 0.0283, 0.0774, 0.0718, 0.0613, 0.0509, 0.0381, 0.0183, 0.0254, 0.0301, 0.0301, 0.011, 0.049, 0.0306, 0.0379, 0.0499, 0.0, 0.0126, -0.0184, -0.0451, -0.0356, -0.0454, -0.0792, -0.074, -0.058, -0.062, -0.0621, -0.0561, -0.0326, -0.0798, -0.1158, -0.0914, -0.0805, -0.0829, -0.0568, 0.0024, 0.0431, 0.0771, 0.0477, 0.0563, 0.0191, 0.0416, 0.029, -0.0118, -0.0238, -0.0078, -0.0185, -0.0178, -0.0073, 0.0286, 0.0034, 0.0034, -0.0011, 0.0009, 0.0472, 0.1076, 0.1083, 0.1402, 0.1803, 0.2695, 0.3256, 0.3302, 0.339, 0.3489, 0.4239, 0.372, 0.3879, 0.403, 0.3634, 0.3222, 0.3259, 0.3818, 0.4107, 0.5357, 0.5492, 0.5846, 0.611, 0.5665, 0.6581, 0.6225, 0.532, 0.6263, 0.5781, 0.6193, 0.7086, 0.7073, 0.7444, 0.7066, 0.6743, 0.7439, 0.8477, 0.8568, 0.8721, 0.9119, 0.8898, 0.8878, 0.98, 0.8394, 1.0036, 1.0106, 1.0072, 1.1369, 1.0341, 1.066, 0.999, 1.0823, 1.0412, 0.9277, 0.9059, 0.6988, 0.7494, 0.8891, 0.8942, 0.9425, 0.9425, 0.995, 1.0286, 1.0204, 0.9754, 0.9121, 1.0013, 1.0831, 1.1295, 1.2247, 1.1804, 1.0343, 1.0036, 0.9615, 0.9025, 1.0968, 1.2434, 1.3334, 1.3307, 1.4185, 1.4185, 1.4026, 1.4844, 1.4697, 1.4088, 1.4088, 1.662, 1.6344, 1.8984]}
{"unit_id": "thread:1963506113376399405", "ticker_idx": 0, "ts": "2025-09-04T07:35:12+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia is bound to succeed in China. China's competing chips are still inferior to Nvidia's, and now Nvidia is about to launch a chip in China that is six times more powerful than the H20.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish NVDA on China dominance: local chips inferior and a new chip 6x more powerful than H20 incoming.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia is bound to succeed in China.\n\nChina’s competing chips are still inferior to Nvidia’s, and now Nvidia is about to launch a chip in China that is six times more powerful than the H20.\n\n$NVDA https://t.co/DiWOmLlsNw\n\n---\n\nhttps://t.co/Rimiqmox8J", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006058477205052815, "ret_signed_m1d": -0.006058477205052815, "alpha_spy_m1d": 0.0022296166490474167, "alpha_c_m1d": 0.005774759335373303, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.011833236540426118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027030129068697395, "ret_signed_p1d": -0.027030129068697395, "alpha_spy_p1d": -0.02413398058016225, "alpha_c_p1d": -0.03882904188315939, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.011798912814461993, "ret_p1w": 0.03215653285758835, "ret_signed_p1w": 0.03215653285758835, "alpha_spy_p1w": 0.019046297719764738, "alpha_c_p1w": -0.016384186012191604, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.04854071886977995, "ret_p1m": 0.0930360228950089, "ret_signed_p1m": 0.0930360228950089, "alpha_spy_p1m": 0.05922802667284133, "alpha_c_p1m": -0.06849026976555406, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.16152629266056295, "ret_p3m": 0.057149151572260415, "ret_signed_p3m": 0.057149151572260415, "alpha_spy_p3m": 0.004308997310772611, "alpha_c_p3m": -0.18286335704959544, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.24001250862185586, "ret_p6m": 0.03233051870988013, "ret_signed_p6m": 0.03233051870988013, "alpha_spy_p6m": -0.0305308479285602, "alpha_c_p6m": -0.3739845610547625, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.40631507976464265, "price_path": [-0.1692, -0.1614, -0.1679, -0.1553, -0.173, -0.1571, -0.1604, -0.1525, -0.1525, -0.162, -0.1601, -0.1384, -0.1011, -0.0969, -0.081, -0.0796, -0.107, -0.0839, -0.0718, -0.0718, -0.0782, -0.0679, -0.0511, -0.044, -0.0393, -0.0442, -0.0056, -0.0017, 0.0078, 0.0044, -0.0016, -0.027, -0.0051, 0.0121, 0.0107, 0.0297, 0.0224, 0.0443, 0.0362, 0.012, 0.0486, 0.0384, 0.0452, 0.0531, 0.0643, 0.0606, 0.067, 0.0578, 0.0604, 0.0512, 0.0603, 0.0232, 0.0218, 0.0193, 0.0369, 0.0475, 0.0589, 0.0579, 0.0496, 0.0147, 0.0147, -0.0051, -0.0061, 0.0, -0.027, -0.0195, -0.0052, 0.033, 0.0322, 0.0359, 0.0355, 0.0188, -0.0079, 0.0267, 0.0292, 0.0697, 0.0395, 0.031, 0.0352, 0.0381, 0.0594, 0.087, 0.0908, 0.1004, 0.093, 0.0809, 0.078, 0.1017, 0.1219, 0.0671, 0.0971, 0.0488, 0.0477, 0.0592, 0.0674, 0.064, 0.0554, 0.0503, 0.0612, 0.0851, 0.1156, 0.1712, 0.2062, 0.182, 0.1797, 0.2052, 0.1575, 0.1373, 0.0957, 0.0961, 0.1596, 0.1253, 0.129, 0.0886, 0.1079, 0.0871, 0.0566, 0.0866, 0.0524, 0.0421, 0.0635, 0.0359, 0.0502, 0.0502, 0.0312, 0.0482, 0.0571, 0.0463, 0.0684, 0.0627, 0.081, 0.0777, 0.0707, 0.0541, 0.0197, 0.0271, 0.0354, -0.0041, 0.0146, 0.0545, 0.0702, 0.1024, 0.0989, 0.0989, 0.1101, 0.0966, 0.0926, 0.0866, 0.0866, 0.1003, 0.096, 0.0909, 0.1018, 0.0781, 0.077, 0.0775, 0.0826, 0.067, 0.0898, 0.085, 0.085, 0.0375, 0.068, 0.0769, 0.0934, 0.0864, 0.0983, 0.1158, 0.1216, 0.1135, 0.0814, 0.0507, 0.0149, 0.0014, 0.0802, 0.1072, 0.0985, 0.1073, 0.0891, 0.0651, 0.0651, 0.0777, 0.0952, 0.0947, 0.1059, 0.116, 0.1236, 0.1394, 0.0772, 0.0323]}
{"unit_id": "orig:1963408855054082560", "ticker_idx": 0, "ts": "2025-09-04T01:08:44+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA has clearly emerged as the true 'big winner'", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA wins AI value chain with 70%+ margins; server OEMs (SMCI, DELL) structurally squeezed with razor-thin margins and no pricing power.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI Server Business Booming, but All the Money Goes to NVIDIA\n\nAmid surging demand for artificial intelligence, global major AI server manufacturers are facing a common challenge. Revenues are growing significantly, but profit margins continue to shrink.\n\nWith companies such as HPE, Dell, and Supermicro releasing earnings one after another, the harsh reality has emerged: while revenue growth is evident, soaring NVIDIA chip costs and fierce market competition are making hardware makers’ margins extremely thin.\n\nHPE, in its third-quarter results released Wednesday, reported revenue of $9.14 billion, up 18% year-on-year, and earnings per share of $0.44, both exceeding market expectations. However, its server division’s operating margin fell to 6.4% from 10.8% a year earlier.\n\nConscious of investor concerns, HPE CEO Antonio Neri pledged on an analyst conference call that margins would recover to around 10% by the end of the current fiscal quarter. This statement temporarily reassured the market, and HPE’s stock rose 1.5% in after-hours trading.\n\nThis trend is spreading across the entire industry, and the root cause lies in the fact that NVIDIA virtually monopolizes the core of AI servers—high-performance GPU chips. These very expensive chips account for most of the server’s cost, leaving the overwhelming majority of value at the top of the supply chain with NVIDIA.\n\n⸻\n\nThe Dilemma of “Higher Revenue, Lower Profits”\n\nFor server OEMs, the current AI market presents an awkward situation of “增收不增利 (increased revenue without increased profit).”\n\nSupermicro, in addition to HPE, is facing the same issue. While its Q4 2025 revenue soared 46.59% year-on-year, its gross margin declined to 9.7%.\n\nDell shows a similar picture. Its Q2 2026 gross margin fell to 18.7% from 22% in the same period last year, which the company attributed to pricing pressure brought on by AI servers.\n\nIn other words, the phenomenon of “high revenue but low profit” has become a common dilemma for AI hardware manufacturers. In stark contrast to the razor-thin margins of server makers is NVIDIA’s overwhelming profitability.\n\nWith a 98% share of the data center GPU market, NVIDIA holds absolute pricing power. Its Q2 2026 results showed a non-GAAP gross margin of as high as 72.7%, several times that of server makers. Data also shows that its latest Blackwell GPU platform delivers margins of up to 77.6% in AI inference workloads.\n\n⸻\n\nThe Current State of the AI Value Chain\n\n- NVIDIA: Gross margin over 70%\n\n- Dell: Gross margin about 18.7%\n\n- HPE: Server division operating margin between 6.4%–10%\n\n- Supermicro: Gross margin 9.7%\n\n⸻\n\nBackground to Margin Pressure for Server Makers\n\nThere are three structural reasons why server manufacturers’ profits continue to be under pressure.\n\n1. High component costs\nThe core of AI servers, NVIDIA GPUs, are expensive and in short supply, leaving OEMs with virtually no bargaining power. One report even pointed out that in AI hardware for cloud service providers, OEMs may lose $1 for every $7.9 of revenue, highlighting the asymmetry in cost structure.\n\n2. Fierce market competition\nTo secure market share, server makers are engaged in aggressive price wars. In order to win large customer orders, they are offering steep discounts, further eroding already thin margins. Dell’s Infrastructure Solutions Group operating margin has already dropped to 8.8%.\n\n3. Complex supply chain management\nTo meet urgent delivery demands for AI components, manufacturers are incurring additional logistics costs, while challenges in inventory management are further adding to operating costs and profitability pressures.\n⸻\n\nInvestor Perspective\n\nFrom an investor’s standpoint, the growth story of AI server manufacturers is becoming increasingly complicated. Revenue forecasts are dazzling, but sustainability faces severe tests. Analysts warn that maintaining margins at the 10–11% level, as in the case of Supermicro, will be insufficient to support long-term technological innovation or deliver substantial returns to shareholders.\n\nIn the grand game of AI, server assemblers are little more than “freight carriers,” while NVIDIA has clearly emerged as the true “big winner.”\n\n$NVDA $SMCI $DELL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006058477205052815, "ret_signed_m1d": -0.006058477205052815, "alpha_spy_m1d": 0.0022296166490474167, "alpha_c_m1d": 0.005774759335373303, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.011833236540426118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027030129068697395, "ret_signed_p1d": -0.027030129068697395, "alpha_spy_p1d": -0.02413398058016225, "alpha_c_p1d": -0.03882904188315939, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.011798912814461993, "ret_p1w": 0.03215653285758835, "ret_signed_p1w": 0.03215653285758835, "alpha_spy_p1w": 0.019046297719764738, "alpha_c_p1w": -0.016384186012191604, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.04854071886977995, "ret_p1m": 0.0930360228950089, "ret_signed_p1m": 0.0930360228950089, "alpha_spy_p1m": 0.05922802667284133, "alpha_c_p1m": -0.06849026976555406, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.16152629266056295, "ret_p3m": 0.057149151572260415, "ret_signed_p3m": 0.057149151572260415, "alpha_spy_p3m": 0.004308997310772611, "alpha_c_p3m": -0.18286335704959544, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.24001250862185586, "ret_p6m": 0.03233051870988013, "ret_signed_p6m": 0.03233051870988013, "alpha_spy_p6m": -0.0305308479285602, "alpha_c_p6m": -0.3739845610547625, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.40631507976464265, "price_path": [-0.1692, -0.1614, -0.1679, -0.1553, -0.173, -0.1571, -0.1604, -0.1525, -0.1525, -0.162, -0.1601, -0.1384, -0.1011, -0.0969, -0.081, -0.0796, -0.107, -0.0839, -0.0718, -0.0718, -0.0782, -0.0679, -0.0511, -0.044, -0.0393, -0.0442, -0.0056, -0.0017, 0.0078, 0.0044, -0.0016, -0.027, -0.0051, 0.0121, 0.0107, 0.0297, 0.0224, 0.0443, 0.0362, 0.012, 0.0486, 0.0384, 0.0452, 0.0531, 0.0643, 0.0606, 0.067, 0.0578, 0.0604, 0.0512, 0.0603, 0.0232, 0.0218, 0.0193, 0.0369, 0.0475, 0.0589, 0.0579, 0.0496, 0.0147, 0.0147, -0.0051, -0.0061, 0.0, -0.027, -0.0195, -0.0052, 0.033, 0.0322, 0.0359, 0.0355, 0.0188, -0.0079, 0.0267, 0.0292, 0.0697, 0.0395, 0.031, 0.0352, 0.0381, 0.0594, 0.087, 0.0908, 0.1004, 0.093, 0.0809, 0.078, 0.1017, 0.1219, 0.0671, 0.0971, 0.0488, 0.0477, 0.0592, 0.0674, 0.064, 0.0554, 0.0503, 0.0612, 0.0851, 0.1156, 0.1712, 0.2062, 0.182, 0.1797, 0.2052, 0.1575, 0.1373, 0.0957, 0.0961, 0.1596, 0.1253, 0.129, 0.0886, 0.1079, 0.0871, 0.0566, 0.0866, 0.0524, 0.0421, 0.0635, 0.0359, 0.0502, 0.0502, 0.0312, 0.0482, 0.0571, 0.0463, 0.0684, 0.0627, 0.081, 0.0777, 0.0707, 0.0541, 0.0197, 0.0271, 0.0354, -0.0041, 0.0146, 0.0545, 0.0702, 0.1024, 0.0989, 0.0989, 0.1101, 0.0966, 0.0926, 0.0866, 0.0866, 0.1003, 0.096, 0.0909, 0.1018, 0.0781, 0.077, 0.0775, 0.0826, 0.067, 0.0898, 0.085, 0.085, 0.0375, 0.068, 0.0769, 0.0934, 0.0864, 0.0983, 0.1158, 0.1216, 0.1135, 0.0814, 0.0507, 0.0149, 0.0014, 0.0802, 0.1072, 0.0985, 0.1073, 0.0891, 0.0651, 0.0651, 0.0777, 0.0952, 0.0947, 0.1059, 0.116, 0.1236, 0.1394, 0.0772, 0.0323]}
{"unit_id": "orig:1963408855054082560", "ticker_idx": 1, "ts": "2025-09-04T01:08:44+00:00", "symbol": "SMCI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "server assemblers are little more than 'freight carriers'", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA wins AI value chain with 70%+ margins; server OEMs (SMCI, DELL) structurally squeezed with razor-thin margins and no pricing power.", "resolved_tickers": ["SMCI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "AI Server Business Booming, but All the Money Goes to NVIDIA\n\nAmid surging demand for artificial intelligence, global major AI server manufacturers are facing a common challenge. Revenues are growing significantly, but profit margins continue to shrink.\n\nWith companies such as HPE, Dell, and Supermicro releasing earnings one after another, the harsh reality has emerged: while revenue growth is evident, soaring NVIDIA chip costs and fierce market competition are making hardware makers’ margins extremely thin.\n\nHPE, in its third-quarter results released Wednesday, reported revenue of $9.14 billion, up 18% year-on-year, and earnings per share of $0.44, both exceeding market expectations. However, its server division’s operating margin fell to 6.4% from 10.8% a year earlier.\n\nConscious of investor concerns, HPE CEO Antonio Neri pledged on an analyst conference call that margins would recover to around 10% by the end of the current fiscal quarter. This statement temporarily reassured the market, and HPE’s stock rose 1.5% in after-hours trading.\n\nThis trend is spreading across the entire industry, and the root cause lies in the fact that NVIDIA virtually monopolizes the core of AI servers—high-performance GPU chips. These very expensive chips account for most of the server’s cost, leaving the overwhelming majority of value at the top of the supply chain with NVIDIA.\n\n⸻\n\nThe Dilemma of “Higher Revenue, Lower Profits”\n\nFor server OEMs, the current AI market presents an awkward situation of “增收不增利 (increased revenue without increased profit).”\n\nSupermicro, in addition to HPE, is facing the same issue. While its Q4 2025 revenue soared 46.59% year-on-year, its gross margin declined to 9.7%.\n\nDell shows a similar picture. Its Q2 2026 gross margin fell to 18.7% from 22% in the same period last year, which the company attributed to pricing pressure brought on by AI servers.\n\nIn other words, the phenomenon of “high revenue but low profit” has become a common dilemma for AI hardware manufacturers. In stark contrast to the razor-thin margins of server makers is NVIDIA’s overwhelming profitability.\n\nWith a 98% share of the data center GPU market, NVIDIA holds absolute pricing power. Its Q2 2026 results showed a non-GAAP gross margin of as high as 72.7%, several times that of server makers. Data also shows that its latest Blackwell GPU platform delivers margins of up to 77.6% in AI inference workloads.\n\n⸻\n\nThe Current State of the AI Value Chain\n\n- NVIDIA: Gross margin over 70%\n\n- Dell: Gross margin about 18.7%\n\n- HPE: Server division operating margin between 6.4%–10%\n\n- Supermicro: Gross margin 9.7%\n\n⸻\n\nBackground to Margin Pressure for Server Makers\n\nThere are three structural reasons why server manufacturers’ profits continue to be under pressure.\n\n1. High component costs\nThe core of AI servers, NVIDIA GPUs, are expensive and in short supply, leaving OEMs with virtually no bargaining power. One report even pointed out that in AI hardware for cloud service providers, OEMs may lose $1 for every $7.9 of revenue, highlighting the asymmetry in cost structure.\n\n2. Fierce market competition\nTo secure market share, server makers are engaged in aggressive price wars. In order to win large customer orders, they are offering steep discounts, further eroding already thin margins. Dell’s Infrastructure Solutions Group operating margin has already dropped to 8.8%.\n\n3. Complex supply chain management\nTo meet urgent delivery demands for AI components, manufacturers are incurring additional logistics costs, while challenges in inventory management are further adding to operating costs and profitability pressures.\n⸻\n\nInvestor Perspective\n\nFrom an investor’s standpoint, the growth story of AI server manufacturers is becoming increasingly complicated. Revenue forecasts are dazzling, but sustainability faces severe tests. Analysts warn that maintaining margins at the 10–11% level, as in the case of Supermicro, will be insufficient to support long-term technological innovation or deliver substantial returns to shareholders.\n\nIn the grand game of AI, server assemblers are little more than “freight carriers,” while NVIDIA has clearly emerged as the true “big winner.”\n\n$NVDA $SMCI $DELL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014742070701991405, "ret_signed_m1d": 0.014742070701991405, "alpha_spy_m1d": 0.006453976847891174, "alpha_c_m1d": 0.009830309389721292, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.004911761312270113, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007125329486268983, "ret_signed_p1d": 0.007125329486268983, "alpha_spy_p1d": 0.004229180997733839, "alpha_c_p1d": 0.00796287573126997, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.0008375462450009863, "ret_p1w": 0.07985257835570803, "ret_signed_p1w": -0.07985257835570803, "alpha_spy_p1w": -0.06674234321788441, "alpha_c_p1w": -0.0477935387537356, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.03205903960197243, "ret_p1m": 0.276658430232422, "ret_signed_p1m": -0.276658430232422, "alpha_spy_p1m": -0.2428504340102544, "alpha_c_p1m": -0.19119581338982083, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.08546261684260115, "ret_p3m": -0.19115485130600895, "ret_signed_p3m": 0.19115485130600895, "alpha_spy_p3m": 0.24399500556749676, "alpha_c_p3m": 0.29407818986637435, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.1029233385603654, "ret_p6m": -0.20417693409131144, "ret_signed_p6m": 0.20417693409131144, "alpha_spy_p6m": 0.26703830072975177, "alpha_c_p6m": 0.26379508346994174, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.0596181493786303, "price_path": [0.0595, 0.0543, 0.0617, 0.0654, 0.0211, 0.0735, 0.0592, 0.0926, 0.0926, 0.1135, 0.0047, 0.0526, 0.1452, 0.2106, 0.169, 0.2042, 0.1597, 0.1975, 0.1931, 0.1931, 0.1575, 0.2066, 0.229, 0.2373, 0.2098, 0.2219, 0.3064, 0.3076, 0.2961, 0.272, 0.2654, 0.2251, 0.2703, 0.2904, 0.3383, 0.4754, 0.4405, 0.4916, 0.4489, 0.3916, 0.4307, 0.4069, 0.1496, 0.1467, 0.0958, 0.1106, 0.1408, 0.1319, 0.1177, 0.1147, 0.1268, 0.0624, 0.0469, 0.0396, 0.0781, 0.0828, 0.0899, 0.0998, 0.0803, 0.0206, 0.0206, 0.002, -0.0147, 0.0, -0.0071, -0.0162, 0.0545, 0.0789, 0.0799, 0.1057, 0.115, 0.1034, 0.1032, 0.1287, 0.1256, 0.1516, 0.1545, 0.1351, 0.1361, 0.1258, 0.1391, 0.1779, 0.2872, 0.2899, 0.2767, 0.342, 0.3531, 0.4418, 0.4246, 0.2988, 0.3452, 0.3049, 0.3253, 0.3229, 0.2821, 0.3523, 0.3469, 0.2899, 0.1774, 0.1865, 0.2671, 0.2865, 0.2946, 0.2376, 0.2767, 0.2469, 0.1646, 0.0327, -0.0088, -0.0231, -0.0125, -0.0462, -0.0686, -0.1378, -0.1052, -0.1622, -0.1425, -0.1713, -0.2246, -0.2091, -0.1813, -0.202, -0.1934, -0.1934, -0.1683, -0.1791, -0.1912, -0.1725, -0.159, -0.1477, -0.131, -0.1396, -0.1425, -0.1641, -0.2057, -0.2292, -0.2221, -0.2668, -0.2784, -0.2356, -0.2366, -0.2442, -0.2494, -0.2494, -0.2472, -0.2609, -0.2715, -0.2808, -0.2808, -0.2393, -0.2612, -0.2496, -0.2627, -0.2654, -0.259, -0.26, -0.2973, -0.3054, -0.2771, -0.198, -0.198, -0.2283, -0.2079, -0.2027, -0.2211, -0.2432, -0.2334, -0.2332, -0.26, -0.2848, -0.27, -0.271, -0.1705, -0.242, -0.1553, -0.1762, -0.1811, -0.2128, -0.2523, -0.2496, -0.2496, -0.2602, -0.27, -0.2098, -0.2034, -0.2455, -0.2351, -0.1744, -0.2069, -0.2042]}
{"unit_id": "orig:1963408855054082560", "ticker_idx": 2, "ts": "2025-09-04T01:08:44+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "high revenue but low profit has become a common dilemma for AI hardware manufacturers", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA wins AI value chain with 70%+ margins; server OEMs (SMCI, DELL) structurally squeezed with razor-thin margins and no pricing power.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "AI Server Business Booming, but All the Money Goes to NVIDIA\n\nAmid surging demand for artificial intelligence, global major AI server manufacturers are facing a common challenge. Revenues are growing significantly, but profit margins continue to shrink.\n\nWith companies such as HPE, Dell, and Supermicro releasing earnings one after another, the harsh reality has emerged: while revenue growth is evident, soaring NVIDIA chip costs and fierce market competition are making hardware makers’ margins extremely thin.\n\nHPE, in its third-quarter results released Wednesday, reported revenue of $9.14 billion, up 18% year-on-year, and earnings per share of $0.44, both exceeding market expectations. However, its server division’s operating margin fell to 6.4% from 10.8% a year earlier.\n\nConscious of investor concerns, HPE CEO Antonio Neri pledged on an analyst conference call that margins would recover to around 10% by the end of the current fiscal quarter. This statement temporarily reassured the market, and HPE’s stock rose 1.5% in after-hours trading.\n\nThis trend is spreading across the entire industry, and the root cause lies in the fact that NVIDIA virtually monopolizes the core of AI servers—high-performance GPU chips. These very expensive chips account for most of the server’s cost, leaving the overwhelming majority of value at the top of the supply chain with NVIDIA.\n\n⸻\n\nThe Dilemma of “Higher Revenue, Lower Profits”\n\nFor server OEMs, the current AI market presents an awkward situation of “增收不增利 (increased revenue without increased profit).”\n\nSupermicro, in addition to HPE, is facing the same issue. While its Q4 2025 revenue soared 46.59% year-on-year, its gross margin declined to 9.7%.\n\nDell shows a similar picture. Its Q2 2026 gross margin fell to 18.7% from 22% in the same period last year, which the company attributed to pricing pressure brought on by AI servers.\n\nIn other words, the phenomenon of “high revenue but low profit” has become a common dilemma for AI hardware manufacturers. In stark contrast to the razor-thin margins of server makers is NVIDIA’s overwhelming profitability.\n\nWith a 98% share of the data center GPU market, NVIDIA holds absolute pricing power. Its Q2 2026 results showed a non-GAAP gross margin of as high as 72.7%, several times that of server makers. Data also shows that its latest Blackwell GPU platform delivers margins of up to 77.6% in AI inference workloads.\n\n⸻\n\nThe Current State of the AI Value Chain\n\n- NVIDIA: Gross margin over 70%\n\n- Dell: Gross margin about 18.7%\n\n- HPE: Server division operating margin between 6.4%–10%\n\n- Supermicro: Gross margin 9.7%\n\n⸻\n\nBackground to Margin Pressure for Server Makers\n\nThere are three structural reasons why server manufacturers’ profits continue to be under pressure.\n\n1. High component costs\nThe core of AI servers, NVIDIA GPUs, are expensive and in short supply, leaving OEMs with virtually no bargaining power. One report even pointed out that in AI hardware for cloud service providers, OEMs may lose $1 for every $7.9 of revenue, highlighting the asymmetry in cost structure.\n\n2. Fierce market competition\nTo secure market share, server makers are engaged in aggressive price wars. In order to win large customer orders, they are offering steep discounts, further eroding already thin margins. Dell’s Infrastructure Solutions Group operating margin has already dropped to 8.8%.\n\n3. Complex supply chain management\nTo meet urgent delivery demands for AI components, manufacturers are incurring additional logistics costs, while challenges in inventory management are further adding to operating costs and profitability pressures.\n⸻\n\nInvestor Perspective\n\nFrom an investor’s standpoint, the growth story of AI server manufacturers is becoming increasingly complicated. Revenue forecasts are dazzling, but sustainability faces severe tests. Analysts warn that maintaining margins at the 10–11% level, as in the case of Supermicro, will be insufficient to support long-term technological innovation or deliver substantial returns to shareholders.\n\nIn the grand game of AI, server assemblers are little more than “freight carriers,” while NVIDIA has clearly emerged as the true “big winner.”\n\n$NVDA $SMCI $DELL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020920497760438583, "ret_signed_m1d": 0.020920497760438583, "alpha_spy_m1d": 0.012632403906338352, "alpha_c_m1d": 0.01600873644816847, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.004911761312270113, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014525848251563556, "ret_signed_p1d": 0.014525848251563556, "alpha_spy_p1d": 0.011629699763028412, "alpha_c_p1d": 0.015363394496564542, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.0008375462450009863, "ret_p1w": -0.0102629109719512, "ret_signed_p1w": 0.0102629109719512, "alpha_spy_p1w": 0.02337314610977481, "alpha_c_p1w": 0.04232195057392363, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.03205903960197243, "ret_p1m": 0.111076061294902, "ret_signed_p1m": -0.111076061294902, "alpha_spy_p1m": -0.07726806507273443, "alpha_c_p1m": -0.02561344445230085, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.08546261684260115, "ret_p3m": 0.07708522318153155, "ret_signed_p3m": -0.07708522318153155, "alpha_spy_p3m": -0.024245068920043744, "alpha_c_p3m": 0.025838115378833848, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.1029233385603654, "ret_p6m": 0.17831966146911715, "ret_signed_p6m": -0.17831966146911715, "alpha_spy_p6m": -0.11545829483067682, "alpha_c_p6m": -0.11870151209048685, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.0596181493786303, "price_path": [-0.102, -0.1096, -0.1254, -0.1084, -0.1386, -0.1057, -0.0885, -0.0836, -0.0836, -0.0615, -0.0721, -0.0519, -0.0529, -0.0095, -0.0251, -0.0361, -0.0429, -0.0292, -0.0155, -0.0155, -0.019, -0.022, -0.0044, 0.0057, -0.0028, -0.0119, -0.0118, -0.0284, -0.026, 0.0319, 0.0139, -0.0185, 0.0043, 0.0133, 0.0359, 0.0568, 0.054, 0.0542, 0.0475, 0.0051, 0.0281, 0.0301, 0.0116, 0.0573, 0.0864, 0.092, 0.1182, 0.0984, 0.0962, 0.0917, 0.0905, 0.0673, 0.0143, 0.0092, 0.0329, 0.0343, 0.0341, 0.046, 0.0583, -0.0357, -0.0357, -0.0451, -0.0209, 0.0, -0.0145, -0.029, -0.0425, -0.0175, -0.0103, -0.0129, 0.001, 0.008, 0.029, 0.0429, 0.0416, 0.0712, 0.0606, 0.0428, 0.0339, 0.0323, 0.0571, 0.1192, 0.1817, 0.1634, 0.1111, 0.1507, 0.191, 0.2989, 0.2312, 0.1887, 0.211, 0.1745, 0.2134, 0.1945, 0.1809, 0.1674, 0.1839, 0.1894, 0.2219, 0.2569, 0.285, 0.3063, 0.2961, 0.2756, 0.2835, 0.2685, 0.2252, 0.2075, 0.1819, 0.1623, 0.1305, 0.0993, 0.1148, 0.0612, 0.0597, -0.0296, -0.028, -0.0542, -0.0699, -0.0294, 0.0079, -0.0024, 0.0558, 0.0558, 0.0565, 0.0465, 0.0771, 0.0587, 0.1012, 0.1005, 0.1124, 0.0951, 0.1142, 0.0981, 0.0298, 0.034, 0.0597, 0.0132, -0.026, 0.0016, 0.0031, 0.0111, 0.0171, 0.0171, 0.0239, 0.0098, 0.0135, -0.0027, -0.0027, 0.0125, -0.0175, -0.0181, -0.0487, -0.0612, -0.0444, -0.0456, -0.052, -0.0597, -0.052, -0.0451, -0.0451, -0.1162, -0.0988, -0.0676, -0.0815, -0.0775, -0.0876, -0.0664, -0.0571, -0.0894, -0.0518, -0.0678, -0.0289, -0.0818, -0.0368, -0.0379, 0.0027, -0.012, -0.1023, -0.0651, -0.0651, -0.0762, -0.0707, -0.0526, -0.0271, -0.052, -0.0469, -0.0174, -0.0336, 0.1783]}
{"unit_id": "quote:1955807892030022131", "ticker_idx": 0, "ts": "2025-08-14T01:45:13+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "That's bad news for $MRVL.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Amazon ASIC work with Alchip seen as bad news for Marvell; implied bearish.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "That’s bad news for $MRVL.", "reply_text": null, "parent_text": null, "quoted_text": "Amazon’s chip division is working with Taiwan’s Alchip on Trainium3 and Trainium4 to bring the chips to mass production on TSMC 3nm and 2nm processes, respectively, media report, citing unnamed industry sources. Trainium3 will be in mass production in the 1st quarter of 2026.", "parent_author": null, "ret_m1d": 0.00354259038479543, "ret_signed_m1d": -0.00354259038479543, "alpha_spy_m1d": -0.0036355471978617437, "alpha_c_m1d": -0.005662309818544653, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.036057790874986506, "ret_signed_p1d": 0.036057790874986506, "alpha_spy_p1d": 0.03371659623296375, "alpha_c_p1d": 0.015290705111012493, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.09906381304654266, "ret_signed_p1w": 0.09906381304654266, "alpha_spy_p1w": 0.08448906278924129, "alpha_c_p1w": 0.0511702804315981, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": -0.14789986050905302, "ret_signed_p1m": 0.14789986050905302, "alpha_spy_p1m": 0.16721922576099468, "alpha_c_p1m": 0.15584906186779868, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.1309355216056236, "ret_signed_p3m": -0.1309355216056236, "alpha_spy_p3m": -0.0690027108664586, "alpha_c_p3m": 0.031789699851623165, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.017091992848284443, "ret_signed_p6m": -0.017091992848284443, "alpha_spy_p6m": 0.059861360010433406, "alpha_c_p6m": 0.3173666999988336, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2091, -0.2236, -0.2401, -0.2181, -0.2328, -0.2328, -0.1932, -0.1835, -0.1944, -0.2391, -0.2229, -0.2117, -0.1619, -0.1763, -0.136, -0.126, -0.1298, -0.1373, -0.1196, -0.1506, -0.1098, -0.1152, -0.0525, -0.0525, -0.0707, -0.1052, -0.0492, -0.0401, 0.0109, -0.0246, -0.0216, -0.0362, -0.0614, -0.0496, -0.0496, -0.0955, -0.0904, -0.0865, -0.0726, -0.0801, -0.0826, -0.0839, -0.1036, -0.0889, -0.0555, -0.0757, -0.0892, -0.073, -0.0633, -0.0611, -0.0396, -0.0342, 0.0342, 0.0168, -0.0581, -0.0318, -0.0305, -0.0471, -0.0404, -0.0215, -0.0223, -0.0156, 0.0035, 0.0, -0.0361, -0.0291, -0.0882, -0.0989, -0.0991, -0.0764, -0.077, -0.0605, -0.0538, -0.0229, -0.2046, -0.2046, -0.1827, -0.2117, -0.189, -0.1988, -0.165, -0.1544, -0.1511, -0.1575, -0.1479, -0.1469, -0.1288, -0.102, -0.0609, -0.0605, -0.0444, -0.0559, 0.0133, 0.0603, 0.0523, 0.0424, 0.0636, 0.0614, 0.0906, 0.0908, 0.125, 0.1003, 0.1703, 0.1473, 0.0838, 0.1317, 0.0916, 0.1254, 0.117, 0.1135, 0.0868, 0.0667, 0.0261, 0.0479, 0.0651, 0.1231, 0.1199, 0.1413, 0.1213, 0.1868, 0.1441, 0.1089, 0.1761, 0.1816, 0.1511, 0.1803, 0.1309, 0.1309, 0.108, 0.0945, 0.0565, -0.0039, 0.0295, -0.0292, -0.0195, 0.0608, 0.0562, 0.1106, 0.1106, 0.1318, 0.1533, 0.176, 0.2686, 0.2431, 0.2522, 0.1647, 0.1255, 0.1707, 0.1322, 0.0689, 0.0667, 0.0643, 0.0343, 0.0694, 0.0646, 0.0736, 0.11, 0.095, 0.095, 0.0931, 0.0857, 0.0984, 0.0759, 0.0759, 0.1317, 0.1423, 0.117, 0.0716, 0.0565, 0.0543, 0.0502, 0.0522, 0.0289, 0.0184, 0.0194, 0.0194, 0.011, 0.0459, 0.0528, 0.0165, 0.036, 0.0507, 0.0594, 0.0305, -0.0001, -0.0034, -0.043, -0.0659, -0.0598, 0.0171]}
{"unit_id": "orig:1960976697718976938", "ticker_idx": 0, "ts": "2025-08-28T08:04:12+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We believe that AVGO has revised up 2026 CoWoS guidance in TSMC to 180k for now, mainly driven by better demand f TPU", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish AVGO: revised up 2026 CoWoS guidance at TSMC to 180k on stronger TPU demand, with external sales to XAI/Meta.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Hong Kong GF Securities, Jeff Pu: We believe that AVGO has revised up 2026 CoWoS guidance in TSMC to 180k for now, mainly driven by better demand f TPU. We believe that, beyond internal use, we expect TPU to be sold extern for customers such as XAI, Meta, etc\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02721524016058996, "ret_signed_m1d": -0.02721524016058996, "alpha_spy_m1d": -0.02368637442206034, "alpha_c_m1d": -0.02313332949905267, "bench_spy_m1d": -0.0035288657385296185, "bench_c_m1d": -0.004081910661537291, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03648129550752288, "ret_signed_p1d": -0.03648129550752288, "alpha_spy_p1d": -0.030517613903543683, "alpha_c_p1d": -0.007740764622537033, "bench_spy_p1d": -0.0059636816039791984, "bench_c_p1d": -0.02874053088498585, "ret_p1w": -0.00826159708763785, "ret_signed_p1w": -0.00826159708763785, "alpha_spy_p1w": -0.008569778814500717, "alpha_c_p1w": 0.021917788613127875, "bench_spy_p1w": 0.00030818172686286616, "bench_c_p1w": -0.030179385700765726, "ret_p1m": 0.08570611446687382, "ret_signed_p1m": 0.08570611446687382, "alpha_spy_p1m": 0.06299922033066441, "alpha_c_p1m": 0.008952817432668292, "bench_spy_p1m": 0.02270689413620941, "bench_c_p1m": 0.07675329703420553, "ret_p3m": 0.24960225899192, "ret_signed_p3m": 0.24960225899192, "alpha_spy_p3m": 0.20649760459415534, "alpha_c_p3m": 0.11198841452098174, "bench_spy_p3m": 0.043104654397764675, "bench_c_p3m": 0.13761384447093827, "ret_p6m": 0.08167028252414665, "ret_signed_p6m": 0.08167028252414665, "alpha_spy_p6m": 0.013149789103290566, "alpha_c_p6m": -0.3112680827040748, "bench_spy_p6m": 0.06852049342085609, "bench_c_p6m": 0.39293836522822145, "price_path": [-0.1961, -0.1698, -0.1561, -0.1598, -0.2018, -0.2104, -0.2093, -0.1825, -0.1723, -0.1961, -0.1851, -0.194, -0.1879, -0.1879, -0.1901, -0.1778, -0.1454, -0.1426, -0.1247, -0.1273, -0.1069, -0.1423, -0.1255, -0.1084, -0.1084, -0.1117, -0.1194, -0.0996, -0.1077, -0.111, -0.1071, -0.0898, -0.0902, -0.0719, -0.082, -0.0662, -0.0974, -0.0809, -0.0646, -0.0598, -0.0465, -0.0364, -0.0195, -0.0484, -0.0648, -0.0354, -0.0509, -0.0226, -0.0158, -0.0119, -0.0154, 0.0135, 0.0014, 0.0084, -0.0075, -0.0094, -0.0445, -0.0566, -0.0617, -0.0475, -0.0467, -0.0345, -0.0272, 0.0, -0.0365, -0.0365, -0.0337, -0.0203, -0.0083, 0.085, 0.1199, 0.0908, 0.1974, 0.1652, 0.1659, 0.1796, 0.1664, 0.1216, 0.1189, 0.1176, 0.0995, 0.1, 0.1012, 0.0908, 0.0857, 0.0642, 0.0707, 0.082, 0.0976, 0.0982, 0.0888, 0.0918, 0.1213, 0.1198, 0.0536, 0.1577, 0.1169, 0.1402, 0.1494, 0.1337, 0.1334, 0.1121, 0.1044, 0.1174, 0.1493, 0.175, 0.2105, 0.2527, 0.2218, 0.1996, 0.1766, 0.1422, 0.1651, 0.1541, 0.1341, 0.1631, 0.1423, 0.1529, 0.1034, 0.1114, 0.1121, 0.1051, 0.1503, 0.1256, 0.1041, 0.2267, 0.2496, 0.2903, 0.2903, 0.3078, 0.253, 0.2384, 0.2353, 0.2366, 0.2665, 0.3018, 0.3186, 0.3403, 0.3189, 0.1681, 0.1028, 0.1077, 0.0581, 0.0706, 0.1046, 0.1103, 0.1359, 0.1388, 0.1388, 0.145, 0.1361, 0.1376, 0.1254, 0.1254, 0.1303, 0.1167, 0.1178, 0.117, 0.0811, 0.1217, 0.1453, 0.1531, 0.1052, 0.1154, 0.1436, 0.1436, 0.0815, 0.0692, 0.0584, 0.0407, 0.0563, 0.0821, 0.0836, 0.0754, 0.0773, 0.0767, 0.0416, 0.0017, 0.0097, 0.0825, 0.1184, 0.107, 0.1145, 0.0769, 0.0573, 0.0573, 0.0813, 0.0845, 0.086, 0.0817]}
{"unit_id": "quote:1943992296900358523", "ticker_idx": 0, "ts": "2025-07-12T11:14:15+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This design precisely addresses two major pain points: first, becoming the preferred solution for inference in small and medium-sized models, fully aligning with the arrival of the inference era; and second, serving as a low-cost computing power pool for cloud services.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish NVDA: B30 positions as preferred inference solution for SMB/cloud, leveraging CUDA ecosystem stickiness and cost advantages over domestic alternatives.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "NVIDIA B30 Industry Notes 2\n\nThis design precisely addresses two major pain points: first, becoming the preferred solution for inference in small and medium-sized models, fully aligning with the arrival of the inference era; and second, serving as a low-cost computing power pool for cloud services.\n\nFirstly, regarding inference for small and medium-sized models, the B30's disadvantage in single-card energy efficiency is mitigated in scenarios with low memory bandwidth requirements, such as intelligent customer service, text generation, and image recognition. For instance, when processing a 4096-long text, the H20's throughput reaches 961 tokens/s, while the B30 is only 60% of that; however, if expanded to an 8-card cluster, the B30 can increase its effective bandwidth to 1.2TB/s through dynamic compression technology, meeting medium concurrency demands. More importantly, its deep compatibility with the CUDA-X ecosystem allows enterprises to seamlessly migrate frameworks like PyTorch, saving technical reconstruction costs.\n\nSecondly, regarding its role as a low-cost computing power pool for cloud services, the B30's cluster solution is highly cost-effective for small and medium-sized enterprises and academic institutions. Tests by a major company show that a computing power pool constructed with 100 B30 cards can support lightweight training of billion-parameter models, reducing procurement costs by 40% and unit power consumption by nearly 30% compared to the H20. Although a certain domestic card significantly surpasses the B30 in single-card FP16 computing power (approximately 200 TFLOPS), the latter maintains an advantage in mainstream model deployment efficiency due to the stickiness of the CUDA ecosystem.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": "NVIDIA B30 Industry Notes: \n\n• China's major internet company A estimates that the performance of NVIDIA's B30 is approximately 75% that of the H20. They placed an order for hundreds of thousands of units (USD 1 billion) in late June, with delivery expected in August.\n \n•", "parent_author": null, "ret_m1d": 0.005180683499361294, "ret_signed_m1d": 0.005180683499361294, "alpha_spy_m1d": 0.007085166748686311, "alpha_c_m1d": -0.0024247053453889666, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.040409535932759866, "ret_signed_p1d": 0.040409535932759866, "alpha_spy_p1d": 0.044682702983386946, "alpha_c_p1d": 0.021203099274968107, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.04455419435283536, "ret_signed_p1w": 0.04455419435283536, "alpha_spy_p1w": 0.03821616236811698, "alpha_c_p1w": 0.0263641341996117, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.11635274151788177, "ret_signed_p1m": 0.11635274151788177, "alpha_spy_p1m": 0.0877358414456213, "alpha_c_p1m": 0.06455133674457092, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.17377251073301703, "ret_signed_p3m": 0.17377251073301703, "alpha_spy_p3m": 0.09661177882903127, "alpha_c_p3m": -0.036132036996605166, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.1413481255415343, "ret_signed_p6m": 0.1413481255415343, "alpha_spy_p6m": 0.02776489035331986, "alpha_c_p6m": -0.22142382598611499, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.3632, -0.3815, -0.3815, -0.4094, -0.3973, -0.374, -0.3514, -0.3234, -0.3373, -0.3356, -0.3362, -0.3198, -0.3022, -0.3063, -0.308, -0.2866, -0.2847, -0.2891, -0.2504, -0.2081, -0.1752, -0.1783, -0.1748, -0.1738, -0.181, -0.1967, -0.1905, -0.1998, -0.1998, -0.1742, -0.1784, -0.1517, -0.1764, -0.1627, -0.1393, -0.1351, -0.1468, -0.1363, -0.1307, -0.1226, -0.1295, -0.1162, -0.1347, -0.1181, -0.1216, -0.1133, -0.1133, -0.1232, -0.1213, -0.0986, -0.0595, -0.0552, -0.0385, -0.0371, -0.0656, -0.0416, -0.0288, -0.0288, -0.0355, -0.0248, -0.0073, 0.0002, 0.0052, 0.0, 0.0404, 0.0445, 0.0544, 0.0508, 0.0446, 0.018, 0.0409, 0.0589, 0.0575, 0.0773, 0.0697, 0.0926, 0.0841, 0.0588, 0.0971, 0.0865, 0.0936, 0.1018, 0.1135, 0.1096, 0.1164, 0.1068, 0.1094, 0.0998, 0.1093, 0.0705, 0.0691, 0.0665, 0.0848, 0.0959, 0.1079, 0.1068, 0.0981, 0.0616, 0.0616, 0.0409, 0.0399, 0.0463, 0.018, 0.0258, 0.0408, 0.0808, 0.0799, 0.0839, 0.0834, 0.0659, 0.038, 0.0742, 0.0769, 0.1192, 0.0876, 0.0787, 0.0831, 0.0861, 0.1084, 0.1373, 0.1413, 0.1513, 0.1436, 0.1309, 0.1279, 0.1527, 0.1738, 0.1164, 0.1479, 0.0973, 0.0961, 0.1082, 0.1168, 0.1132, 0.1042, 0.0989, 0.1103, 0.1353, 0.1672, 0.2253, 0.262, 0.2367, 0.2342, 0.261, 0.2111, 0.1899, 0.1464, 0.1468, 0.2133, 0.1774, 0.1813, 0.139, 0.1591, 0.1374, 0.1054, 0.1369, 0.1011, 0.0903, 0.1127, 0.0839, 0.0987, 0.0987, 0.0789, 0.0967, 0.1061, 0.0947, 0.1178, 0.1119, 0.131, 0.1275, 0.1203, 0.1029, 0.0669, 0.0746, 0.0833, 0.042, 0.0615, 0.1033, 0.1197, 0.1534, 0.1497, 0.1497, 0.1614, 0.1473, 0.1432, 0.1368, 0.1368, 0.1512, 0.1467, 0.1413]}
{"unit_id": "orig:1952960642547396973", "ticker_idx": 0, "ts": "2025-08-06T05:11:16+00:00", "symbol": "S&P500", "kind": "index", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "maintaining stock allocation is a necessary consideration", "tweet_type": "analysis", "timeline": "months", "summary": "Merrill endorses maintaining equity allocation at all-time highs; S&P 500 historically outperforms after new highs, supported by earnings, consumption, and tech innovation.", "resolved_tickers": ["SPY"], "resolution_method": "etf", "resolution_reason": "S&P 500 best proxied by SPY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Merrill: Why Stock Allocation Should Be Maintained Even at All-Time Highs\n\nThe S&P 500 index recently surpassed the 6,400 mark for the first time, setting new all-time highs on a closing basis 15 times since the beginning of the year. The fact that 12 of these instances have occurred within the last 30 trading days suggests that the market's upward momentum remains strong. This environment may foster a perception among investors, particularly those who reduced risk during the high volatility period in the spring, that they have not fully participated in the rally.\n\nIt is natural to feel a psychological burden about deploying new capital when the index is near its all-time high. However, this can lead to investment decisions based on market timing. We have consistently cautioned against strategies centered on market timing, emphasizing that reducing stock allocation based solely on valuation concerns can result in missing out on medium- to long-term return opportunities.\n\nBased on data since 1990, investing in the S&P 500 on a day it reached an all-time high has, on average, yielded superior returns over 1, 2, 3, and 5-year periods compared to investing on any given day (Chart 3A). This indicates that all-time highs frequently occur during momentum phases within a bull market, suggesting a high probability of subsequent return improvements even after these peaks (Chart 3B).\n\nFurthermore, historical cases show that a correction of -10% or more within one year of an all-time high has occurred in less than 5% of instances. Out of the 24 years since 1990 that saw a market peak, only four coincided with a recession. This implies that market peaks have generally occurred during economic expansion phases.\n\nThe current market rally is supported by structural factors such as solid corporate earnings, a recovery in consumption, technological innovation, and the proliferation of asset-light business models. While there may be some noise in future economic indicators or the possibility of short-term corrections, we view these as temporary fluctuations within a medium- to long-term upward trend. In this environment, excessively increasing cash holdings could lead to opportunity costs. From a strategic standpoint, maintaining stock allocation is a necessary consideration.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007601455526256684, "ret_signed_m1d": -0.007601455526256684, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": -0.007601455526256684, "bench_c_m1d": -0.007601455526256684, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0008376240775045396, "ret_signed_p1d": -0.0008376240775045396, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": -0.0008376240775045396, "bench_c_p1d": -0.0008376240775045396, "ret_p1w": 0.019137827924330386, "ret_signed_p1w": 0.019137827924330386, "alpha_spy_p1w": 0.0, "alpha_c_p1w": 0.0, "bench_spy_p1w": 0.019137827924330386, "bench_c_p1w": 0.019137827924330386, "ret_p1m": 0.025822478557951944, "ret_signed_p1m": 0.025822478557951944, "alpha_spy_p1m": 0.0, "alpha_c_p1m": 0.0, "bench_spy_p1m": 0.025822478557951944, "bench_c_p1m": 0.025822478557951944, "ret_p3m": 0.08289540700689901, "ret_signed_p3m": 0.08289540700689901, "alpha_spy_p3m": 0.0, "alpha_c_p3m": 0.0, "bench_spy_p3m": 0.08289540700689901, "bench_c_p3m": 0.08289540700689901, "ret_p6m": 0.10310168956684795, "ret_signed_p6m": 0.10310168956684795, "alpha_spy_p6m": 0.0, "alpha_c_p6m": 0.0, "bench_spy_p6m": 0.10310168956684795, "bench_c_p6m": 0.10310168956684795, "price_path": [-0.1108, -0.0814, -0.0753, -0.0742, -0.0696, -0.0637, -0.0627, -0.0659, -0.0816, -0.0812, -0.0875, -0.0875, -0.0685, -0.0739, -0.0703, -0.0713, -0.0661, -0.0608, -0.061, -0.0655, -0.056, -0.0551, -0.0497, -0.0525, -0.0487, -0.0593, -0.0504, -0.0585, -0.0586, -0.0586, -0.0608, -0.0516, -0.0411, -0.0406, -0.033, -0.0282, -0.0236, -0.0239, -0.0195, -0.0118, -0.0118, -0.0191, -0.0197, -0.0138, -0.011, -0.0145, -0.0126, -0.0168, -0.0135, -0.0075, -0.0082, -0.0063, -0.0062, 0.0023, 0.0026, 0.0068, 0.0066, 0.0039, 0.0027, -0.0011, -0.0175, -0.0025, -0.0076, 0.0, -0.0008, 0.007, 0.005, 0.0157, 0.0191, 0.0192, 0.0168, 0.0166, 0.0111, 0.0084, 0.0044, 0.0198, 0.0153, 0.0196, 0.0219, 0.0255, 0.0194, 0.0194, 0.0118, 0.0173, 0.0258, 0.0229, 0.0254, 0.0277, 0.0307, 0.0393, 0.0389, 0.0445, 0.043, 0.0417, 0.0466, 0.0518, 0.0567, 0.051, 0.0477, 0.0428, 0.0488, 0.0517, 0.0557, 0.0593, 0.0605, 0.0605, 0.0643, 0.0604, 0.0667, 0.0636, 0.0348, 0.0507, 0.0494, 0.0541, 0.0469, 0.0529, 0.0638, 0.0638, 0.0583, 0.0645, 0.0732, 0.0859, 0.0888, 0.0893, 0.0773, 0.0809, 0.0829, 0.0701, 0.0738, 0.0622, 0.0633, 0.0799, 0.0824, 0.083, 0.065, 0.0648, 0.0549, 0.046, 0.0501, 0.0341, 0.0444, 0.0597, 0.0697, 0.0771, 0.0771, 0.083, 0.078, 0.08, 0.0838, 0.0846, 0.0866, 0.0834, 0.0824, 0.0896, 0.0921, 0.0804, 0.0788, 0.0758, 0.064, 0.072, 0.0817, 0.0885, 0.0934, 0.0973, 0.0973, 0.0972, 0.0933, 0.0919, 0.0838, 0.0838, 0.0858, 0.0931, 0.0996, 0.096, 0.0959, 0.1031, 0.1049, 0.1027, 0.0973, 0.1002, 0.0993, 0.0993, 0.0769, 0.0894, 0.0951, 0.0955, 0.101, 0.1054, 0.1053, 0.1031]}
{"unit_id": "quote:1946797685887172727", "ticker_idx": 0, "ts": "2025-07-20T05:01:52+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It seems that VMware's customers are increasingly leaving. This is obviously bad news for Broadcom.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish AVGO as VMware customer churn accelerates, including major Samsung affiliates.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "It seems that VMware's customers are increasingly leaving. This is obviously bad news for Broadcom. \n\nI also know that many companies I'm familiar with are reviewing VMware alternatives.\n\n$AVGO https://t.co/TIvh4PsKZP", "reply_text": null, "parent_text": null, "quoted_text": "Report: Samsung Affiliates Actively Seeking VMware Alternatives; Reviewing Costs and Efficiency from Multiple Angles\n\nMajor Samsung Group affiliates are embarking on a quest for alternatives to the virtualization software 'VMware.'\n\nThis move comes as VMware's prices have surged https://t.co/I40dOH777F", "parent_author": null, "ret_m1d": -0.01689728278649416, "ret_signed_m1d": 0.01689728278649416, "alpha_spy_m1d": 0.015004696296329278, "alpha_c_m1d": 0.016208866898576257, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0333783873584057, "ret_signed_p1d": 0.0333783873584057, "alpha_spy_p1d": 0.03352150852163027, "alpha_c_p1d": 0.01568537195034514, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.021130415854473172, "ret_signed_p1w": -0.021130415854473172, "alpha_spy_p1w": -0.008136737169707642, "alpha_c_p1w": -0.018032386296110037, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.023247088882376143, "ret_signed_p1m": -0.023247088882376143, "alpha_spy_p1m": -0.005689000241037467, "alpha_c_p1m": -0.022007877059030845, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.23089701527378304, "ret_signed_p3m": -0.23089701527378304, "alpha_spy_p3m": -0.1772977480698017, "alpha_c_p3m": -0.04918219690314207, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.234854009623642, "ret_signed_p6m": -0.234854009623642, "alpha_spy_p6m": -0.12514909276558428, "alpha_c_p6m": 0.11707287519277809, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.3876, -0.3487, -0.3343, -0.3338, -0.3383, -0.3338, -0.3169, -0.2951, -0.3052, -0.3074, -0.291, -0.2808, -0.2793, -0.233, -0.1955, -0.1965, -0.1947, -0.2087, -0.2017, -0.198, -0.2048, -0.202, -0.2083, -0.2083, -0.1843, -0.1712, -0.1624, -0.1621, -0.1391, -0.1109, -0.0963, -0.1002, -0.1452, -0.1544, -0.1532, -0.1245, -0.1136, -0.1391, -0.1273, -0.1368, -0.1303, -0.1303, -0.1326, -0.1195, -0.0848, -0.0817, -0.0626, -0.0654, -0.0436, -0.0814, -0.0635, -0.0452, -0.0452, -0.0487, -0.0569, -0.0358, -0.0444, -0.048, -0.0438, -0.0252, -0.0257, -0.0061, -0.0169, 0.0, -0.0334, -0.0157, 0.0017, 0.0068, 0.0211, 0.032, 0.05, 0.019, 0.0015, 0.033, 0.0164, 0.0467, 0.054, 0.0582, 0.0544, 0.0854, 0.0724, 0.0799, 0.0629, 0.0609, 0.0232, 0.0103, 0.0048, 0.0201, 0.0209, 0.034, 0.0418, 0.0709, 0.0319, 0.0319, 0.0348, 0.0492, 0.0621, 0.162, 0.1993, 0.1681, 0.2823, 0.2478, 0.2486, 0.2633, 0.2491, 0.2011, 0.1983, 0.1968, 0.1775, 0.178, 0.1793, 0.1682, 0.1627, 0.1397, 0.1466, 0.1587, 0.1754, 0.1761, 0.166, 0.1692, 0.2008, 0.1992, 0.1283, 0.2398, 0.1961, 0.2211, 0.2309, 0.2141, 0.2138, 0.191, 0.1828, 0.1966, 0.2308, 0.2584, 0.2963, 0.3415, 0.3085, 0.2847, 0.2601, 0.2232, 0.2477, 0.2359, 0.2145, 0.2456, 0.2233, 0.2346, 0.1816, 0.1903, 0.1909, 0.1835, 0.2318, 0.2054, 0.1824, 0.3137, 0.3382, 0.3818, 0.3818, 0.4005, 0.3419, 0.3262, 0.3229, 0.3243, 0.3563, 0.3941, 0.4121, 0.4353, 0.4124, 0.251, 0.1811, 0.1862, 0.1331, 0.1465, 0.183, 0.189, 0.2164, 0.2196, 0.2196, 0.2262, 0.2167, 0.2183, 0.2052, 0.2052, 0.2105, 0.1959, 0.1971, 0.1962, 0.1578, 0.2013, 0.2265, 0.2349]}
{"unit_id": "thread:1950339218695434251", "ticker_idx": 0, "ts": "2025-07-29T23:34:39+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics is showing signs of passing the low point in the sub-3nm foundry competition. Considering that its competitor's CAPA has virtually reached its limit, Samsung is rising again to a position where it can expect reflective benefits as a second choice", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Shinhan Securities bullish view: Samsung foundry passing its low point, set to benefit as TSMC Plan B with re-evaluation of its foundry ecosystem.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Samsung Electronics' 'Changed Strategy,' a 'Volume Offensive' Despite Low-Margin Tesla Order\n\nAn analysis suggests that although Samsung Electronics has signed a 22 trillion KRW foundry (semiconductor contract manufacturing) supply deal with Tesla, it is a structure where profits are difficult to expect in the short term, considering the yield and cost of the 2-nanometer (nm) process. Nevertheless, the reason Samsung pushed for this order appears to be a strategic shift. Unlike in the past, when it secured orders based on technological prowess from introducing advanced processes, the new strategy seems to be securing orders first and then increasing stability by pouring in massive funds and volume, even if it incurs some cost losses due to low yields.\n\nRecently, Samsung Electronics has been experiencing multi-trillion won foundry deficits as its 3nm process failed to achieve expected yields, causing it to lose major clients to Taiwan's TSMC. Consequently, it seems to have decided that it is more beneficial to produce some finished products and deliver them to clients, even while accepting losses, rather than waiting for yields to be secured.\n\nA semiconductor industry official stated, \"I understand that the 2nm process, AI6, accounts for the majority of this order volume, with some 4nm process, AI4, mixed in.\" Samsung plans to mass-produce the AI6 using its 2nm Gate-All-Around (GAA) process at its Taylor fab in Texas. The mass production timing is anticipated to be after 2027.\n\nIn the case of the Dojo series, it appears a separate order was not received this time, as Tesla recently revealed plans to integrate the A16 and Dojo series. However, as Tesla CEO Elon Musk has hinted that Samsung's actual production scale could become several times larger in the future, the possibility of additional negotiations expanding to the Dojo series is considered open.\n\nOf course, from Samsung Electronics' perspective, this order is not a \"profitable business.\" It is highly likely that Samsung offered Tesla a lower price than TSMC to win this contract, and with existing yield issues, it is at a disadvantage compared to TSMC in terms of profit margins. While there are talks that the 2nm yield has recently improved significantly to over 40%, yield improvement is still in progress as the target is set for over 60% by the end of this year.\n\nAnother industry official commented, \"Samsung's 2nm process yield seems to have risen from the 20% range at the beginning of this year to the current 30% range.\" TSMC's 2nm yield is estimated to be between 60% and 70%.\n\nAmidst this, Samsung must also bear massive costs to enhance the completeness of its 2nm process. It is actually accelerating the commercialization of 'High-NA,' a next-generation Extreme Ultraviolet (EUV) technology. A High-NA machine is an EUV equipment that costs 500 billion KRW per unit, a greater investment burden than existing equipment (200-300 billion KRW).\n\nThe first official mentioned, \"Until now, High-NA technology was considered a big risk for full-scale adoption due to its low efficiency relative to its price. However, with recent successive news of performance improvements, the interest from major companies has rapidly increased.\" He added, \"Currently, Samsung Foundry is conducting trial runs of High-NA and checking its performance.\"\n\nThe possibility of the Dojo2 volume, currently produced by TSMC, being allocated to Samsung is analyzed to be low. The reasoning is that there is no need to dual-source Dojo2 before the mass production of AI6. A semiconductor industry official said, \"As of now, the possibility of Dojo2 volume coming over to Samsung is almost non-existent. This is because there have been no cases of dual-sourcing when Tesla clearly defines the roles of each supplier.\" Tesla CEO Elon Musk recently drew a line on social media, stating, 'AI4/AI6 will be produced by Samsung, AI5 by TSMC.'\n\nWithin Samsung's foundry business division, the atmosphere is more focused on securing clients rather than immediate profitability. An official familiar with the company's situation said, \"The strategy for responding to orders has recently changed. In the past, the basic approach was to deliver after raising the yield. Now, the approach is to somehow meet the volume (number of good chips) presented by the client through mass production, even if the yield is low.\"\n\nHe further explained, \"Currently, we are in a situation where we cannot afford to wait for the yield to increase, and this Tesla order was also a decision made considering that it's difficult for it to contribute to profitability right away. Above all, we are viewing it positively because foundry yield is an aspect that naturally improves as the utilization rate increases.\"\n\nIt was reported that not only in foundry but also in the memory business, the company has emphasized a stance of \"injecting massive volume this year.\" This is a similar trend to the moves of its competitor, Micron. Micron is also in a situation where it is absorbing the cost burden to maximize production and meet the High-Bandwidth Memory (HBM) volume demanded by clients like NVIDIA, even if it means many chips are discarded due to low yields.\n\nMeanwhile, there is also an interpretation that this order could allow Samsung's foundry business to rapidly emerge as a 'Plan B' for TSMC's customers. This analysis suggests a green light has been lit for securing future clients.\n\nShinhan Securities analyzed, \"Samsung Electronics is showing signs of passing the low point in the sub-3nm foundry competition. Considering that its competitor's CAPA (production capacity) has virtually reached its limit, Samsung is rising again to a position where it can expect reflective benefits as a second choice.\" They explained, \"This change is expected to act as an opportunity for a re-evaluation of Samsung's entire foundry ecosystem.\"\n\n---\n\nhttps://t.co/DGAyZtadak", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0028327933379326264, "ret_signed_m1d": -0.0028327933379326264, "alpha_spy_m1d": -0.005477514679550044, "alpha_c_m1d": -0.0027949689941849654, "bench_spy_m1d": 0.0026447213416174176, "bench_c_m1d": -3.782434374766108e-05, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028328716846468538, "ret_signed_p1d": 0.028328716846468538, "alpha_spy_p1d": 0.02958795108554746, "alpha_c_p1d": 0.02609452800228307, "bench_spy_p1d": -0.0012592342390789213, "bench_c_p1d": 0.0022341888441854696, "ret_p1w": -0.009915000530519191, "ret_signed_p1w": -0.009915000530519191, "alpha_spy_p1w": 0.0015606386742602618, "alpha_c_p1w": 0.005117739571685487, "bench_spy_p1w": -0.011475639204779453, "bench_c_p1w": -0.015032740102204678, "ret_p1m": 0.0, "ret_signed_p1m": 0.0, "alpha_spy_p1m": -0.017898257051605082, "alpha_c_p1m": -0.0013631847076409631, "bench_spy_p1m": 0.017898257051605082, "bench_c_p1m": 0.0013631847076409631, "ret_p3m": 0.40567727152866984, "ret_signed_p3m": 0.40567727152866984, "alpha_spy_p3m": 0.33662257506749005, "alpha_c_p3m": 0.29258201962086106, "bench_spy_p3m": 0.06905469646117979, "bench_c_p3m": 0.11309525190780878, "ret_p6m": 1.1373514154565085, "ret_signed_p6m": 1.1373514154565085, "alpha_spy_p6m": 1.0522347967620542, "alpha_c_p6m": 1.0451992474254583, "bench_spy_p6m": 0.0851166186944543, "bench_c_p6m": 0.0921521680310502, "price_path": [-0.2187, -0.2356, -0.2356, -0.2356, -0.2313, -0.2313, -0.2285, -0.1891, -0.199, -0.1919, -0.1933, -0.2004, -0.2144, -0.213, -0.2159, -0.2299, -0.237, -0.2299, -0.2412, -0.213, -0.2102, -0.2088, -0.2004, -0.2004, -0.1863, -0.168, -0.168, -0.1581, -0.1666, -0.1567, -0.1624, -0.1793, -0.1947, -0.1821, -0.1581, -0.1666, -0.1624, -0.1835, -0.1483, -0.137, -0.1525, -0.1388, -0.153, -0.1473, -0.1388, -0.0963, -0.1034, -0.1261, -0.1303, -0.1445, -0.136, -0.1133, -0.1147, -0.0977, -0.0836, -0.0552, -0.0496, -0.0397, -0.0652, -0.0595, -0.0652, -0.0666, -0.0028, 0.0, 0.0283, 0.0113, -0.0241, -0.0127, -0.0099, -0.0255, -0.0014, 0.017, 0.0057, 0.0071, 0.0184, 0.0142, 0.0142, -0.0085, -0.0085, -0.0014, 0.0, 0.0113, 0.0127, -0.0042, 0.0, -0.0142, -0.0127, -0.0425, -0.0212, -0.0113, -0.0071, -0.0156, -0.0071, 0.0127, 0.0283, 0.0397, 0.068, 0.0836, 0.1246, 0.1076, 0.1374, 0.1374, 0.1827, 0.1997, 0.2096, 0.2195, 0.1799, 0.198, 0.1937, 0.2236, 0.2769, 0.2769, 0.2769, 0.2769, 0.2769, 0.2769, 0.3431, 0.3274, 0.3032, 0.3516, 0.39, 0.3929, 0.3957, 0.3872, 0.4028, 0.373, 0.4057, 0.4512, 0.4156, 0.4299, 0.4811, 0.5295, 0.5807, 0.4925, 0.4313, 0.4114, 0.3929, 0.4313, 0.4725, 0.4669, 0.4626, 0.3829, 0.4313, 0.3914, 0.373, 0.4313, 0.3488, 0.3758, 0.4128, 0.4626, 0.4725, 0.4299, 0.4341, 0.4711, 0.4868, 0.4953, 0.5423, 0.5579, 0.5423, 0.5366, 0.5266, 0.5494, 0.491, 0.4626, 0.5351, 0.5309, 0.5124, 0.5721, 0.5864, 0.5807, 0.5807, 0.6646, 0.7085, 0.7142, 0.7142, 0.7142, 0.8371, 0.9744, 0.9858, 1.0158, 0.9844, 0.9872, 0.9844, 0.9672, 1.0058, 1.0573, 1.1288, 1.1345, 1.0759, 1.1374]}
{"unit_id": "orig:1948162014024798613", "ticker_idx": 0, "ts": "2025-07-23T23:23:13+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Recovered to +32% after a slight slowdown to +28% last quarter. Annual revenue run rate exceeds $50B. Google Cloud's backlog increased by 18% QoQ and 38% YoY, reaching $106B at quarter-end.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish GOOGL: GCP reaccelerating to +32%, $106B backlog, token usage doubling, subscription momentum, and improving Cloud margins all point to durable strength.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Alphabet (GOOGL US EQUITY) FY2Q25 Conference Call\n\n▶️ Key Comments\n\nAt I/O on May 20th, disclosed 480 trillion tokens used in April (50x increase YoY).\n\nThis number has since doubled, now processing 980 trillion tokens monthly.\n\nSearch\n\nRecorded double-digit revenue growth (+7% QoQ, +12% YoY).\n\nNew features like AI Overviews, Circle to Search, and Lens are preventing user churn.\n\nSubscription\n\nYouTube and subscription services also showed strong momentum, reflecting robust performance.\n\nGemini MAU (Monthly Active Users) reached 450 million, with daily requests increasing by over 50% compared to Q1.\n\nGCP\n\nRecovered to +32% after a slight slowdown to +28% last quarter.\n\nAnnual revenue run rate exceeds $50B.\nGoogle Cloud's backlog increased by 18% QoQ and 38% YoY, reaching $106B at quarter-end.\nSupply is still expected to remain tight relative to demand until 2026.\n\nCapex\n\nCapex forecast raised from the previous $75B to approximately $85B.\n\n▶️Q&A\n\nQ: Do you anticipate changes in the traditional long-standing methods related to clicks and click monetization over the next 3-5 years?\n\nA: All indicators we are seeing suggest that people are very positive about AI. AI is being well-received across all products. For example, people are already actively using the image-based Circle to Search feature in multimodality.\n\nQ: What are your thoughts on the mix between advertising and subscriptions?\nA: We have a very deep understanding of our monetization strategies for both advertising and subscriptions. We plan to continue aggressively pursuing this dual strategy.\n\nQ: How should we understand the contribution of cost per click and click volume growth to the 12% growth?\nA: Paid clicks increased by 4% YoY. Some policy changes are leading to a decrease in click volume but are favorable for monetization.\n\nQ: You mentioned a $10B increase in Capex and supply shortages simultaneously. How should this be understood?\nA: While additional investments are being executed, it takes time for them to translate into actual capacity expansion.\n\nQ: In conversations with engineering teams developing agentic capabilities, what are the biggest technical hurdles?\nA: Currently, during the experimental process, reducing latency, optimizing costs, and maintaining user trust are challenging. We expect agentic-based experiences to be more widely used by 2026.\n\nQ: What are the efficiency gains from using internal Gen AI tools? Are they better than before?\nA: Many agents are being utilized in actual workflows, and previously friction-filled experiences are gradually improving.\n\nQ: What is the difference in user behavior between AI mode and the standalone Gemini app?\nA: AI mode and the standalone Gemini app are optimized for different use cases. AI mode is very powerful when focusing on information retrieval and seeking accurate, information-based responses, whereas Gemini's queries vary from psychological counseling to conversational.\n\nQ: How is Google currently responding to attracting and retaining AI talent?\nA: We have invested deeply in AI talent over a long period and currently boast a very competitive position in the AI talent pool, possessing both breadth and depth. The factors top talent values most are: 1) conducting meaningful research on the front lines, 2) access to abundant computing resources, and 3) an environment for collaboration with industry-leading colleagues.\n\nQ: Given that Click-Through Rate (CTR) has been a key driver of monetization in Search, can CTR be maintained in an environment where ad impressions per SERP (Search Engine Results Page) decrease due to the introduction of AI Overviews and new AI formats?\nA: User satisfaction continues to increase, and the number of search users is also growing. From a monetization perspective, we see AI Overviews yielding nearly the same level of monetization performance as the traditional search experience.\n\nQ: There have been reports about collaboration between OpenAI and Google Cloud. In what direction can this partnership expand in the future?\nA: The collaboration with OpenAI is a very encouraging partnership from Google Cloud's perspective. We hope to further strengthen and grow it.\n\nQ: Is the advertising environment in the second half of the year more uncertain or more stable compared to last year?\nA: It's early in Q3, so it's premature to give specific comments for the entire second half, but currently, there are no particular concerns.\n\nQ: Why are you pursuing both AI Mode and Gemini instead of moving to a unified, integrated search experience?\nA: Because the two platforms can broadly and deeply cover the entire spectrum of human information consumption behavior. Search is information-centric, while Gemini is focused on being a personal assistant. It's closer to an everyday helper. Over time, we will allow users to integrate functionalities from both interfaces without complex choices. In the past, text, image, and video search were separate. We have experience integrating them through Universal Search.\n\nQ: What is the outlook for the Cloud segment in the second half? Last quarter you mentioned resolving supply constraints by end of 2025, but today you are talking about supply constraints until 2026. Can you explain the change in outlook?\nA: Current growth rates are a result of past Capex execution, and cloud growth is not linear. Growth can continue to vary quarter by quarter.\n\nQ: How are subscription services related to Gemini growing, and what is the growth strategy?\nA: The past quarter was a period of very healthy growth for AI-based subscription services, and the future outlook is very encouraging.\n\nQ: How do you assess the ROI on CapEx investment in the Cloud segment?\nA: As the Cloud customer base expands, operational efficiency gradually increases. Churn rates also decrease. This structure allows for greater revenue generation with the same cost. The Cloud segment has shown a strong trend of improving operating margins over the past few years.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005835094076843239, "ret_signed_m1d": 0.005835094076843239, "alpha_spy_m1d": 0.014270867509582286, "alpha_c_m1d": 0.011275411195515006, "bench_spy_m1d": -0.008435773432739047, "bench_c_m1d": -0.005440317118671767, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010198270357911765, "ret_signed_p1d": 0.010198270357911765, "alpha_spy_p1d": 0.009867217835084707, "alpha_c_p1d": 0.008538558353469172, "bench_spy_p1d": 0.00033105252282705777, "bench_c_p1d": 0.0016597120044425928, "ret_p1w": 0.033117869028981284, "ret_signed_p1w": 0.033117869028981284, "alpha_spy_p1w": 0.032723731152933144, "alpha_c_p1w": 0.05155957139615264, "bench_spy_p1w": 0.00039413787604813955, "bench_c_p1w": -0.018441702367171353, "ret_p1m": 0.05004464230444006, "ret_signed_p1m": 0.05004464230444006, "alpha_spy_p1m": 0.047931865297867304, "alpha_c_p1m": 0.038241930053669604, "bench_spy_p1m": 0.0021127770065727525, "bench_c_p1m": 0.011802712250770453, "ret_p3m": 0.3498368099877851, "ret_signed_p3m": 0.3498368099877851, "alpha_spy_p3m": 0.2884201128864794, "alpha_c_p3m": 0.27426840303208566, "bench_spy_p3m": 0.061416697101305706, "bench_c_p3m": 0.07556840695569944, "ret_p6m": 0.7520659828212746, "ret_signed_p6m": 0.7520659828212746, "alpha_spy_p6m": 0.6543061048924037, "alpha_c_p6m": 0.6732940332898767, "bench_spy_p6m": 0.09775987792887086, "bench_c_p6m": 0.07877194953139788, "price_path": [-0.1496, -0.1567, -0.1591, -0.1662, -0.1531, -0.1388, -0.1378, -0.143, -0.2052, -0.19, -0.198, -0.168, -0.1624, -0.1317, -0.1391, -0.1274, -0.1256, -0.139, -0.115, -0.1029, -0.1155, -0.1155, -0.0922, -0.095, -0.0977, -0.0983, -0.1125, -0.1275, -0.1177, -0.1168, -0.0881, -0.0743, -0.0611, -0.0677, -0.0764, -0.0818, -0.0708, -0.0751, -0.0889, -0.0889, -0.124, -0.1316, -0.1233, -0.1028, -0.0877, -0.0615, -0.0736, -0.0756, -0.0609, -0.0562, -0.0562, -0.0707, -0.0834, -0.0715, -0.0663, -0.0528, -0.0456, -0.0433, -0.0382, -0.035, -0.0272, -0.0007, 0.0058, 0.0, 0.0102, 0.0155, 0.0124, 0.029, 0.0331, 0.0088, -0.0058, 0.0253, 0.0233, 0.0308, 0.0331, 0.0588, 0.0566, 0.0689, 0.0617, 0.0668, 0.0719, 0.0698, 0.0596, 0.0478, 0.05, 0.0834, 0.096, 0.0889, 0.0907, 0.1125, 0.1192, 0.1192, 0.111, 0.2125, 0.2212, 0.2353, 0.2314, 0.2608, 0.2584, 0.2647, 0.267, 0.3238, 0.3215, 0.3129, 0.3261, 0.3402, 0.3287, 0.3241, 0.3003, 0.2932, 0.2972, 0.2841, 0.2791, 0.2885, 0.2927, 0.2909, 0.3176, 0.2931, 0.2871, 0.2708, 0.2447, 0.2846, 0.2914, 0.3208, 0.3231, 0.3327, 0.3498, 0.3178, 0.3243, 0.3316, 0.3676, 0.4168, 0.4073, 0.4446, 0.481, 0.4795, 0.4928, 0.4603, 0.4959, 0.4982, 0.4671, 0.5264, 0.5327, 0.5085, 0.4657, 0.4543, 0.4996, 0.4957, 0.5406, 0.5229, 0.5767, 0.6762, 0.7018, 0.6834, 0.6834, 0.6846, 0.6568, 0.6616, 0.6817, 0.6712, 0.6904, 0.6517, 0.6694, 0.6859, 0.6449, 0.6284, 0.6228, 0.6141, 0.5622, 0.5924, 0.6172, 0.631, 0.655, 0.6537, 0.6537, 0.6506, 0.6509, 0.6524, 0.6479, 0.6479, 0.6592, 0.6666, 0.655, 0.6952, 0.7134, 0.7299, 0.7472, 0.7689, 0.7682, 0.7521]}
{"unit_id": "thread:1968962935730196959", "ticker_idx": 0, "ts": "2025-09-19T08:58:40+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung is competing with a more advanced process node… Samsung holds the edge… Samsung's advantage, as its implementation (11 Gbps) is faster than SK hynix's", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung leads HBM4 race with 1c DRAM + 4nm logic die + 11Gbps speed advantage over rivals ahead of Rubin ramp.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“A Matter of Pride, Not Money”… How Samsung Broke Through NVIDIA’s Wall\n\nSamsung Electronics has passed NVIDIA’s quality test for its fifth-generation high-bandwidth memory (HBM3E) 12-high product. It took a year and six months since the company completed development of the HBM3E 12-high in February last year. Analysts say this was the result of Samsung revamping its existing HBM3E to improve performance, combined with NVIDIA’s strategy of diversifying its suppliers. With Samsung having regained a significant portion of its HBM competitiveness, many expect the company to perform well in the race to develop and supply sixth-generation HBM (HBM4).\n\n◇ Tackling Heat Issues Through DRAM Redesign\n\nAccording to industry sources on the 19th, Samsung recently passed NVIDIA’s quality test for HBM3E 12-high. Stacking 12 DRAM layers to maximize data throughput, the HBM3E 12-high is used in the latest AI accelerators such as NVIDIA’s B300 and AMD’s MI350. Samsung had broken into AMD’s supply chain but had repeatedly failed to clear NVIDIA’s hurdles.\n\nSamsung went through multiple setbacks in its efforts to supply HBM3E 8-high and 12-high to NVIDIA. Around October last year, the HBM3E 8-high test was close to passing when orders were suddenly postponed. The issue was that Samsung had used fourth-generation 10-nanometer (1a DRAM) chips as the core die, which failed to meet NVIDIA’s strict heat dissipation requirements.\n\nThe reason Samsung managed to break through with the more advanced 12-high was its redesign of the 1a DRAM, solving the overheating problem. The decisive move came from Jun Young-hyun, who took over as head of the Device Solutions (DS) division last May. After meeting NVIDIA’s executives early this year, he ordered a full redesign of the 1a DRAM for HBM3E. The HBM development team, effectively betting everything, fixed the thermal issues.\n\nIt was only natural that NVIDIA, after verification, placed an order with Samsung. Adding Samsung to the supplier list helps lower costs. However, shipment volumes are expected to be small, as Samsung is the third supplier after SK hynix and Micron.\n\nAn industry insider commented, “For Samsung, HBM3E 12-high was a matter of pride. Even if it doesn’t bring big sales, it carries great significance in proving their technical capability.” Samsung officially stated regarding its HBM3E 12-high supply to NVIDIA, “The process is proceeding in line with customer schedules.”\n\n◇ Advantage in HBM4 Speed\n\nThe next battleground for the memory “big three”—Samsung Electronics, SK hynix, and Micron—is HBM4. HBM4 will be used in NVIDIA’s next-generation AI accelerator, Rubin, set for release next year. Unlike the HBM3E round, the HBM4 race is expected to be different—Samsung is competing with a more advanced process node.\n\nSamsung plans to use sixth-generation 10-nanometer (1c DRAM) chips in HBM4, while SK hynix and Micron will continue with the 1b DRAM they used for HBM3E. For the logic die, Samsung will use a 4-nanometer foundry process, compared to SK hynix and Micron’s reliance on TSMC’s 12-nanometer process. On paper, Samsung holds the edge.\n\nNVIDIA recently asked HBM suppliers to raise data processing speeds to at least 10 gigabits per second (Gbps) to boost Rubin’s performance. This plays to Samsung’s advantage, as its implementation (11 Gbps) is faster than SK hynix’s (around 10 Gbps). Micron is reportedly struggling to meet NVIDIA’s requirements. Samsung plans to deliver HBM4 samples this month that meet NVIDIA’s requested specifications.\n\n$NVDA\n\n---\n\nhttps://t.co/zgLE52GZXb", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.004928464706804525, "alpha_c_m1d": 0.010038694785118873, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": -0.010038694785118873, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.039850587603541765, "ret_signed_p1d": 0.039850587603541765, "alpha_spy_p1d": 0.03511957963544554, "alpha_c_p1d": 0.025139252809774915, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.01471133479376685, "ret_p1w": 0.037359888976894684, "ret_signed_p1w": 0.037359888976894684, "alpha_spy_p1w": 0.04019244520004206, "alpha_c_p1w": 0.036361841897415426, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.0009980470794792584, "ret_p1m": 0.22711941862246676, "ret_signed_p1m": 0.22711941862246676, "alpha_spy_p1m": 0.21566850003244653, "alpha_c_p1m": 0.1925561199719561, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.034563298650510665, "ret_p3m": 0.34970615148733897, "ret_signed_p3m": 0.34970615148733897, "alpha_spy_p3m": 0.3381045738681123, "alpha_c_p3m": 0.34892353317207103, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.0007826183152679356, "ret_p6m": 1.3718967008058978, "ret_signed_p6m": 1.3718967008058978, "alpha_spy_p6m": 1.3608873084316122, "alpha_c_p6m": 1.3739823280820576, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": -0.002085627276159796, "price_path": [-0.2512, -0.2413, -0.2549, -0.2428, -0.2553, -0.2503, -0.2428, -0.2055, -0.2117, -0.2316, -0.2354, -0.2478, -0.2403, -0.2204, -0.2217, -0.2067, -0.1943, -0.1694, -0.1644, -0.1557, -0.1781, -0.1731, -0.1781, -0.1793, -0.1233, -0.1208, -0.0959, -0.1108, -0.142, -0.132, -0.1295, -0.1432, -0.122, -0.1059, -0.1158, -0.1146, -0.1046, -0.1083, -0.1083, -0.1283, -0.1283, -0.122, -0.1208, -0.1108, -0.1096, -0.1245, -0.1208, -0.1333, -0.132, -0.1582, -0.1395, -0.1308, -0.127, -0.1345, -0.127, -0.1096, -0.0959, -0.0859, -0.061, -0.0473, -0.0112, -0.0262, 0.0, 0.0, 0.0399, 0.0548, 0.0635, 0.0722, 0.0374, 0.0532, 0.0495, 0.0758, 0.1227, 0.1227, 0.1227, 0.1227, 0.1227, 0.1227, 0.1808, 0.1671, 0.1458, 0.1883, 0.2221, 0.2246, 0.2271, 0.2196, 0.2334, 0.2071, 0.2359, 0.2759, 0.2446, 0.2571, 0.3022, 0.3447, 0.3897, 0.3122, 0.2584, 0.2409, 0.2246, 0.2584, 0.2947, 0.2897, 0.2859, 0.2159, 0.2584, 0.2234, 0.2071, 0.2584, 0.1858, 0.2096, 0.2421, 0.2859, 0.2947, 0.2571, 0.2609, 0.2934, 0.3072, 0.3147, 0.356, 0.3697, 0.356, 0.351, 0.3422, 0.3622, 0.3109, 0.2859, 0.3497, 0.346, 0.3297, 0.3822, 0.3947, 0.3897, 0.3897, 0.4635, 0.5021, 0.5071, 0.5071, 0.5071, 0.6152, 0.7359, 0.7459, 0.7723, 0.7447, 0.7472, 0.7447, 0.7296, 0.7635, 0.8088, 0.8716, 0.8767, 0.8251, 0.8792, 0.9144, 0.9118, 0.9118, 1.0049, 1.0413, 1.0199, 1.0174, 0.8905, 1.1054, 1.1255, 1.0023, 0.9935, 1.0916, 1.0841, 1.1092, 1.2449, 1.2776, 1.2776, 1.2776, 1.2776, 1.3882, 1.3895, 1.4259, 1.5139, 1.5579, 1.7402, 1.7213, 1.7213, 1.4523, 1.1645, 1.4083, 1.3656, 1.1808, 1.3618, 1.3882, 1.3618, 1.3065, 1.3719]}
{"unit_id": "thread:1968176574072754463", "ticker_idx": 0, "ts": "2025-09-17T04:53:56+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron announced NAND price increases, signaling that manufacturers anticipate tight capacity supply in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND/DRAM contract prices rising 15-20% per quarter through 1Q26 on AI-driven CSP demand; Samsung V9 sold out, Micron signaling hikes, Phison posting record profits — entire memory supply chain bullish.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory prices surged in 4Q25, defying seasonal trends as AI drives demand-Digitimes\n\nDriven by strong AI demand, memory contract prices for NAND and DRAM rose sharply by an estimated 15-20% in the fourth quarter of 2025, breaking traditional year-end price decline patterns. Supply shortages led to aggressive procurement from cloud service providers, with high-stack 3D NAND products nearly sold out.\n\nIndustry insiders reveal that large cloud service providers (CSPs) have expanded their big data storage needs from HDD shortages to high-capacity SSDs.\n\nSamsung's V9 NAND nearly sold out\nRecent reports indicate that Samsung Electronics' most advanced V9 NAND (estimated around 290-layer stacking) for 2026 is nearly sold out. Due to three key advantages—better cost structure from higher stacking layers, faster read speeds, and larger die capacity per chip—this 3D NAND product has attracted strong priority purchasing interest from CSP customers.\n\nThe scramble for supplies has now spread to the main V8 NAND products, reflecting the booming development of AI servers in recent years. As computing power infrastructure becomes more complete, \"data storage capability\" is becoming a critical requirement for the next phase of AI advancement.\n\nSanDisk and Micron signal price hikes\nRecently, SanDisk and Micron announced NAND price increases, signaling that manufacturers anticipate tight capacity supply in 2026. Micron's fiscal year-end report concluded at the end of August, likely clearing internal inventory levels just as major CSPs ramp up orders for 2026. This has prompted several manufacturers to preemptively signal upcoming price hikes, reminding downstream customers of another impending price cycle.\n\nAlthough fourth-quarter contract prices are not yet finalized, supply chains estimate that this round of price increases will definitely exceed single-digit percentages. Contract prices could rise by 15-20% per quarter starting from the fourth quarter of 2025 through the first quarter of 2026, clearly reflecting industry-wide supply constraints.\n\nDRAM and NAND both facing steep gains\nAnother source revealed that manufacturers recently proposed a 10% price increase for October contracts. Based on this, quarterly price growth may surpass 20%, with contract price rises for NAND and DRAM being comparable. In terms of spot market performance, DRAM price increases might even be more pronounced than those of NAND.\n\nThe NAND market slowed down in the third quarter of 2024 but began showing signs of recovery starting in the second quarter of 2025, boosting confidence across related supply chains.\n\nThe market initially expected prices to decline during the traditional slow season from the fourth quarter through the following first quarter, but the 2025 price trend is proving to be an exception.\n\nDespite consumer purchasing power not significantly improving, the clear trend of increased orders from major cloud players indicates that the 2025 price rally will extend from the third quarter into the fourth quarter and potentially all the way through the first quarter of 2026. Price forecasts for the entire year of 2026 remain optimistic, with expectations for quarterly price increases smoothly connecting to the traditional peak season in the third quarter.\n\nAs the NAND market warms up, manufacturers have successively initiated price hikes, benefiting the entire supply chain. Many memory module makers currently hold low-cost inventories.\n\nPhison reports record-breaking profits\nNAND controller giant Phison(https://t.co/dzoVOYHImM) reported a significant profit leap in August 2025, with monthly revenue reaching approximately NT$5.934 billion (approx. US$197.6 million), a 23.48% year-over-year increase, and net income after tax hitting NT$650 million, soaring 49x compared to the previous year. This reflects positive contributions from its strategic focus on non-consumer markets and R&D investments.\n\nPhison stated it will continue strengthening cooperation with NAND manufacturers to ensure supply meets diverse global customer applications and seize growth opportunities brought by market recovery.\n\n---\n\nhttps://t.co/q4YaPbacmK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007312954749480682, "ret_signed_m1d": -0.007312954749480682, "alpha_spy_m1d": -0.008556935607750948, "alpha_c_m1d": -0.01373391548670877, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.055628446960352074, "ret_signed_p1d": 0.055628446960352074, "alpha_spy_p1d": 0.050956001818021734, "alpha_c_p1d": 0.017233657058628005, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.010750710153943022, "ret_signed_p1w": 0.010750710153943022, "alpha_spy_p1w": 0.005057550797399113, "alpha_c_p1w": -0.04120666195159273, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.26668435544784375, "ret_signed_p1m": 0.26668435544784375, "alpha_spy_p1m": 0.2616908769628672, "alpha_c_p1m": 0.14203251350581048, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.4853974431306163, "ret_signed_p3m": 0.4853974431306163, "alpha_spy_p3m": 0.4498423000980296, "alpha_c_p3m": 0.32929587068420574, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 1.5362065568262122, "ret_signed_p6m": 1.5362065568262122, "alpha_spy_p6m": 1.5199740166326137, "alpha_c_p6m": 1.2607349457543522, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.2282, -0.2377, -0.2013, -0.2054, -0.2132, -0.2209, -0.2304, -0.2451, -0.2398, -0.2364, -0.2364, -0.2505, -0.2223, -0.236, -0.2305, -0.2216, -0.2586, -0.2493, -0.2723, -0.2921, -0.285, -0.2923, -0.3173, -0.3135, -0.3016, -0.3046, -0.3046, -0.3002, -0.2828, -0.3178, -0.3445, -0.3264, -0.3183, -0.3201, -0.3008, -0.2569, -0.2267, -0.2015, -0.2233, -0.2169, -0.2445, -0.2278, -0.2371, -0.2674, -0.2763, -0.2645, -0.2723, -0.2718, -0.264, -0.2375, -0.2561, -0.2561, -0.2595, -0.258, -0.2236, -0.1789, -0.1783, -0.1547, -0.1249, -0.0589, -0.0173, -0.0139, -0.0073, 0.0, 0.0556, 0.0171, 0.0289, 0.0401, 0.0108, -0.0198, -0.017, 0.0244, 0.0458, 0.1385, 0.1485, 0.1747, 0.1943, 0.1614, 0.2292, 0.2029, 0.1358, 0.2056, 0.1699, 0.2005, 0.2667, 0.2657, 0.2932, 0.2652, 0.2413, 0.2928, 0.3698, 0.3766, 0.3879, 0.4174, 0.401, 0.3995, 0.4679, 0.3636, 0.4854, 0.4906, 0.488, 0.5842, 0.508, 0.5317, 0.482, 0.5438, 0.5132, 0.4291, 0.413, 0.2594, 0.297, 0.4005, 0.4043, 0.4401, 0.4401, 0.479, 0.5039, 0.4978, 0.4645, 0.4175, 0.4836, 0.5443, 0.5787, 0.6493, 0.6165, 0.5082, 0.4854, 0.4542, 0.4105, 0.5545, 0.6631, 0.7299, 0.7279, 0.793, 0.793, 0.7812, 0.8418, 0.8309, 0.7858, 0.7858, 0.9735, 0.9531, 1.1488, 1.1245, 1.0461, 1.1592, 1.1641, 1.1156, 1.0857, 1.1062, 1.2697, 1.2697, 1.2837, 1.4346, 1.4876, 1.5005, 1.4345, 1.5668, 1.7235, 1.7267, 1.5958, 1.7392, 1.6244, 1.3738, 1.3957, 1.4695, 1.3995, 1.3354, 1.5674, 1.5901, 1.5757, 1.5757, 1.5014, 1.6338, 1.6113, 1.679, 1.6339, 1.6154, 1.6842, 1.6001, 1.5801, 1.582, 1.3756, 1.5076, 1.4843, 1.3169, 1.4359, 1.5222, 1.6197, 1.5362]}
{"unit_id": "thread:1968176574072754463", "ticker_idx": 1, "ts": "2025-09-17T04:53:56+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' most advanced V9 NAND (estimated around 290-layer stacking) for 2026 is nearly sold out", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND/DRAM contract prices rising 15-20% per quarter through 1Q26 on AI-driven CSP demand; Samsung V9 sold out, Micron signaling hikes, Phison posting record profits — entire memory supply chain bullish.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory prices surged in 4Q25, defying seasonal trends as AI drives demand-Digitimes\n\nDriven by strong AI demand, memory contract prices for NAND and DRAM rose sharply by an estimated 15-20% in the fourth quarter of 2025, breaking traditional year-end price decline patterns. Supply shortages led to aggressive procurement from cloud service providers, with high-stack 3D NAND products nearly sold out.\n\nIndustry insiders reveal that large cloud service providers (CSPs) have expanded their big data storage needs from HDD shortages to high-capacity SSDs.\n\nSamsung's V9 NAND nearly sold out\nRecent reports indicate that Samsung Electronics' most advanced V9 NAND (estimated around 290-layer stacking) for 2026 is nearly sold out. Due to three key advantages—better cost structure from higher stacking layers, faster read speeds, and larger die capacity per chip—this 3D NAND product has attracted strong priority purchasing interest from CSP customers.\n\nThe scramble for supplies has now spread to the main V8 NAND products, reflecting the booming development of AI servers in recent years. As computing power infrastructure becomes more complete, \"data storage capability\" is becoming a critical requirement for the next phase of AI advancement.\n\nSanDisk and Micron signal price hikes\nRecently, SanDisk and Micron announced NAND price increases, signaling that manufacturers anticipate tight capacity supply in 2026. Micron's fiscal year-end report concluded at the end of August, likely clearing internal inventory levels just as major CSPs ramp up orders for 2026. This has prompted several manufacturers to preemptively signal upcoming price hikes, reminding downstream customers of another impending price cycle.\n\nAlthough fourth-quarter contract prices are not yet finalized, supply chains estimate that this round of price increases will definitely exceed single-digit percentages. Contract prices could rise by 15-20% per quarter starting from the fourth quarter of 2025 through the first quarter of 2026, clearly reflecting industry-wide supply constraints.\n\nDRAM and NAND both facing steep gains\nAnother source revealed that manufacturers recently proposed a 10% price increase for October contracts. Based on this, quarterly price growth may surpass 20%, with contract price rises for NAND and DRAM being comparable. In terms of spot market performance, DRAM price increases might even be more pronounced than those of NAND.\n\nThe NAND market slowed down in the third quarter of 2024 but began showing signs of recovery starting in the second quarter of 2025, boosting confidence across related supply chains.\n\nThe market initially expected prices to decline during the traditional slow season from the fourth quarter through the following first quarter, but the 2025 price trend is proving to be an exception.\n\nDespite consumer purchasing power not significantly improving, the clear trend of increased orders from major cloud players indicates that the 2025 price rally will extend from the third quarter into the fourth quarter and potentially all the way through the first quarter of 2026. Price forecasts for the entire year of 2026 remain optimistic, with expectations for quarterly price increases smoothly connecting to the traditional peak season in the third quarter.\n\nAs the NAND market warms up, manufacturers have successively initiated price hikes, benefiting the entire supply chain. Many memory module makers currently hold low-cost inventories.\n\nPhison reports record-breaking profits\nNAND controller giant Phison(https://t.co/dzoVOYHImM) reported a significant profit leap in August 2025, with monthly revenue reaching approximately NT$5.934 billion (approx. US$197.6 million), a 23.48% year-over-year increase, and net income after tax hitting NT$650 million, soaring 49x compared to the previous year. This reflects positive contributions from its strategic focus on non-consumer markets and R&D investments.\n\nPhison stated it will continue strengthening cooperation with NAND manufacturers to ensure supply meets diverse global customer applications and seize growth opportunities brought by market recovery.\n\n---\n\nhttps://t.co/q4YaPbacmK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01534540628571035, "ret_signed_m1d": 0.01534540628571035, "alpha_spy_m1d": 0.014101425427440084, "alpha_c_m1d": 0.011882113978149977, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0034632923075603728, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026854284168853715, "ret_signed_p1d": 0.026854284168853715, "alpha_spy_p1d": 0.022181839026523376, "alpha_c_p1d": 0.009575301461721697, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.017278982707132018, "ret_p1w": 0.09207173038970451, "ret_signed_p1w": 0.09207173038970451, "alpha_spy_p1w": 0.0863785710331606, "alpha_c_p1w": 0.06474255088484493, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.02732917950485958, "ret_p1m": 0.25493492765635994, "ret_signed_p1m": 0.25493492765635994, "alpha_spy_p1m": 0.2499414491713834, "alpha_c_p1m": 0.20539841209937149, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.04953651555698846, "ret_p3m": 0.3461328090778437, "ret_signed_p3m": 0.3461328090778437, "alpha_spy_p3m": 0.31057766604525705, "alpha_c_p3m": 0.2962642897829064, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.0498685192949373, "ret_p6m": 1.425266562838158, "ret_signed_p6m": 1.425266562838158, "alpha_spy_p6m": 1.4090340226445595, "alpha_c_p6m": 1.4067608257548645, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.018505737083293505, "price_path": [-0.2438, -0.2628, -0.2311, -0.2209, -0.2349, -0.2225, -0.2353, -0.2302, -0.2225, -0.1841, -0.1905, -0.211, -0.2148, -0.2276, -0.2199, -0.1995, -0.2008, -0.1854, -0.1726, -0.1471, -0.1419, -0.133, -0.156, -0.1509, -0.156, -0.1573, -0.0997, -0.0972, -0.0716, -0.087, -0.1189, -0.1087, -0.1061, -0.1202, -0.0985, -0.0818, -0.0921, -0.0908, -0.0806, -0.0844, -0.0844, -0.1049, -0.1049, -0.0985, -0.0972, -0.087, -0.0857, -0.101, -0.0972, -0.11, -0.1087, -0.1355, -0.1164, -0.1074, -0.1036, -0.1113, -0.1036, -0.0857, -0.0716, -0.0614, -0.0358, -0.0217, 0.0153, 0.0, 0.0269, 0.0269, 0.0678, 0.0831, 0.0921, 0.101, 0.0652, 0.0815, 0.0777, 0.1047, 0.1528, 0.1528, 0.1528, 0.1528, 0.1528, 0.1528, 0.2125, 0.1984, 0.1766, 0.2203, 0.2549, 0.2575, 0.2601, 0.2524, 0.2665, 0.2395, 0.2691, 0.3102, 0.2781, 0.2909, 0.3371, 0.3808, 0.4271, 0.3474, 0.2922, 0.2742, 0.2575, 0.2922, 0.3294, 0.3243, 0.3204, 0.2485, 0.2922, 0.2562, 0.2395, 0.2922, 0.2177, 0.2421, 0.2755, 0.3204, 0.3294, 0.2909, 0.2948, 0.3282, 0.3423, 0.35, 0.3924, 0.4065, 0.3924, 0.3872, 0.3782, 0.3988, 0.3461, 0.3204, 0.386, 0.3821, 0.3654, 0.4193, 0.4322, 0.4271, 0.4271, 0.5028, 0.5424, 0.5476, 0.5476, 0.5476, 0.6586, 0.7825, 0.7928, 0.8199, 0.7915, 0.7941, 0.7915, 0.776, 0.8109, 0.8573, 0.9219, 0.927, 0.8741, 0.9296, 0.9658, 0.9632, 0.9632, 1.0587, 1.0961, 1.0742, 1.0716, 0.9412, 1.162, 1.1826, 1.0561, 1.0471, 1.1478, 1.14, 1.1658, 1.3052, 1.3388, 1.3388, 1.3388, 1.3388, 1.4524, 1.4537, 1.4911, 1.5814, 1.6266, 1.8138, 1.7944, 1.7944, 1.5182, 1.2226, 1.473, 1.4291, 1.2394, 1.4253, 1.4524, 1.4253]}
{"unit_id": "thread:1968176574072754463", "ticker_idx": 3, "ts": "2025-09-17T04:53:56+00:00", "symbol": "NAND", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "contract prices could rise by 15-20% per quarter starting from the fourth quarter of 2025 through the first quarter of 2026, clearly reflecting industry-wide supply constraints", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND/DRAM contract prices rising 15-20% per quarter through 1Q26 on AI-driven CSP demand; Samsung V9 sold out, Micron signaling hikes, Phison posting record profits — entire memory supply chain bullish.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "Global NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory prices surged in 4Q25, defying seasonal trends as AI drives demand-Digitimes\n\nDriven by strong AI demand, memory contract prices for NAND and DRAM rose sharply by an estimated 15-20% in the fourth quarter of 2025, breaking traditional year-end price decline patterns. Supply shortages led to aggressive procurement from cloud service providers, with high-stack 3D NAND products nearly sold out.\n\nIndustry insiders reveal that large cloud service providers (CSPs) have expanded their big data storage needs from HDD shortages to high-capacity SSDs.\n\nSamsung's V9 NAND nearly sold out\nRecent reports indicate that Samsung Electronics' most advanced V9 NAND (estimated around 290-layer stacking) for 2026 is nearly sold out. Due to three key advantages—better cost structure from higher stacking layers, faster read speeds, and larger die capacity per chip—this 3D NAND product has attracted strong priority purchasing interest from CSP customers.\n\nThe scramble for supplies has now spread to the main V8 NAND products, reflecting the booming development of AI servers in recent years. As computing power infrastructure becomes more complete, \"data storage capability\" is becoming a critical requirement for the next phase of AI advancement.\n\nSanDisk and Micron signal price hikes\nRecently, SanDisk and Micron announced NAND price increases, signaling that manufacturers anticipate tight capacity supply in 2026. Micron's fiscal year-end report concluded at the end of August, likely clearing internal inventory levels just as major CSPs ramp up orders for 2026. This has prompted several manufacturers to preemptively signal upcoming price hikes, reminding downstream customers of another impending price cycle.\n\nAlthough fourth-quarter contract prices are not yet finalized, supply chains estimate that this round of price increases will definitely exceed single-digit percentages. Contract prices could rise by 15-20% per quarter starting from the fourth quarter of 2025 through the first quarter of 2026, clearly reflecting industry-wide supply constraints.\n\nDRAM and NAND both facing steep gains\nAnother source revealed that manufacturers recently proposed a 10% price increase for October contracts. Based on this, quarterly price growth may surpass 20%, with contract price rises for NAND and DRAM being comparable. In terms of spot market performance, DRAM price increases might even be more pronounced than those of NAND.\n\nThe NAND market slowed down in the third quarter of 2024 but began showing signs of recovery starting in the second quarter of 2025, boosting confidence across related supply chains.\n\nThe market initially expected prices to decline during the traditional slow season from the fourth quarter through the following first quarter, but the 2025 price trend is proving to be an exception.\n\nDespite consumer purchasing power not significantly improving, the clear trend of increased orders from major cloud players indicates that the 2025 price rally will extend from the third quarter into the fourth quarter and potentially all the way through the first quarter of 2026. Price forecasts for the entire year of 2026 remain optimistic, with expectations for quarterly price increases smoothly connecting to the traditional peak season in the third quarter.\n\nAs the NAND market warms up, manufacturers have successively initiated price hikes, benefiting the entire supply chain. Many memory module makers currently hold low-cost inventories.\n\nPhison reports record-breaking profits\nNAND controller giant Phison(https://t.co/dzoVOYHImM) reported a significant profit leap in August 2025, with monthly revenue reaching approximately NT$5.934 billion (approx. US$197.6 million), a 23.48% year-over-year increase, and net income after tax hitting NT$650 million, soaring 49x compared to the previous year. This reflects positive contributions from its strategic focus on non-consumer markets and R&D investments.\n\nPhison stated it will continue strengthening cooperation with NAND manufacturers to ensure supply meets diverse global customer applications and seize growth opportunities brought by market recovery.\n\n---\n\nhttps://t.co/q4YaPbacmK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017088492788325672, "ret_signed_m1d": 0.017088492788325672, "alpha_spy_m1d": 0.015844511930055406, "alpha_c_m1d": 0.010667532051097585, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043722415302998786, "ret_signed_p1d": 0.043722415302998786, "alpha_spy_p1d": 0.03904997016066845, "alpha_c_p1d": 0.005327625401274717, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.07040231638139702, "ret_signed_p1w": 0.07040231638139702, "alpha_spy_p1w": 0.06470915702485311, "alpha_c_p1w": 0.01844494427586127, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.3926535304208163, "ret_signed_p1m": 0.3926535304208163, "alpha_spy_p1m": 0.38766005193583974, "alpha_c_p1m": 0.26800168847878303, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.7428400321487472, "ret_signed_p3m": 0.7428400321487472, "alpha_spy_p3m": 0.7072848891161605, "alpha_c_p3m": 0.5867384597023366, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 2.8254856478969606, "ret_signed_p6m": 2.8254856478969606, "alpha_spy_p6m": 2.809253107703362, "alpha_c_p6m": 2.5500140368251003, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.3385, -0.3297, -0.3041, -0.2995, -0.2992, -0.3046, -0.309, -0.322, -0.3184, -0.3038, -0.3174, -0.323, -0.3053, -0.3041, -0.2971, -0.2977, -0.312, -0.3069, -0.3132, -0.3293, -0.3285, -0.3273, -0.3378, -0.3261, -0.325, -0.332, -0.3243, -0.3215, -0.3103, -0.3142, -0.341, -0.3347, -0.3311, -0.3386, -0.3327, -0.3133, -0.3051, -0.2798, -0.2781, -0.2802, -0.2901, -0.2995, -0.3046, -0.3195, -0.3273, -0.313, -0.3089, -0.3076, -0.3047, -0.2889, -0.283, -0.2926, -0.2924, -0.2847, -0.2549, -0.21, -0.1983, -0.1823, -0.1427, -0.0801, -0.0303, -0.0169, 0.0171, 0.0, 0.0437, 0.0436, 0.066, 0.0847, 0.0704, 0.051, 0.0161, 0.0826, 0.0915, 0.1342, 0.2054, 0.2544, 0.2492, 0.2232, 0.2591, 0.2645, 0.2526, 0.2937, 0.2573, 0.3293, 0.3927, 0.3776, 0.439, 0.4256, 0.4324, 0.4809, 0.6253, 0.6757, 0.6417, 0.7653, 0.796, 0.8032, 0.879, 0.7924, 0.8385, 0.8686, 0.949, 1.109, 1.1111, 1.1401, 1.0326, 0.8868, 0.9933, 0.8624, 0.8791, 0.7804, 0.6934, 0.7752, 0.7603, 0.7018, 0.7457, 0.771, 0.73, 0.7523, 0.7122, 0.7413, 0.8127, 0.8735, 0.8458, 0.8818, 0.8851, 0.8084, 0.7428, 0.7098, 0.7491, 0.8141, 0.8604, 0.947, 0.952, 1.0068, 1.0154, 1.0627, 1.0616, 1.0477, 1.0326, 1.0326, 1.1886, 1.2594, 1.4907, 1.5583, 1.5188, 1.612, 1.6407, 1.6673, 1.6482, 1.7272, 1.8367, 1.8633, 1.9268, 2.1271, 2.222, 2.1424, 2.1093, 2.2609, 2.445, 2.4916, 2.6676, 2.6771, 2.9181, 2.637, 2.5017, 2.5396, 2.5348, 2.4225, 2.5872, 2.8073, 2.8736, 2.849, 2.7608, 2.7654, 2.8283, 2.9055, 2.9404, 3.0046, 2.9859, 3.0878, 3.0205, 3.0, 2.6566, 2.5984, 2.6754, 2.5269, 2.5029, 2.7522, 2.9466, 2.8255]}
{"unit_id": "quote:1965031634086019245", "ticker_idx": 0, "ts": "2025-09-08T12:37:04+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our Buy rating. The new $200 target price is derived by applying the same 30x P/E multiple to our revised 2026 EPS estimate.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi note: Buy NVDA (PT cut to $200) + Buy AVGO (PT raised to $350, 4th AI customer est. OpenAI with $10B order).", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Citi) NVIDIA – GPU vs. XPU: Target Price Lowered to $200 Amid Intensifying TPU Competition\n\nFollowing Broadcom’s comments last week that year-over-year growth in XPUs is accelerating due to both new and existing customers, we are lowering our 2026 GPU revenue forecast by about 4% and cutting our target price (TP) to $200. We had previously expected AI XPU chip sales to surpass GPU sales in 2026. Broadcom’s remarks on the rapid pace of XPU adoption appear to stem from Google’s strategy shift toward providing compute power to competitors such as Meta, OpenAI, and Oracle, thereby indirectly competing with NVIDIA. This has also been a risk we recently flagged.\n\nWe estimate that these transactions will impact NVIDIA’s 2026 revenue by reducing GPU sales by roughly $12 billion. Supported by growth in spending from new cloud players (“neo cloud”) and sovereign AI initiatives worldwide, our revised 2025/2026 forecasts remain 2% and 5% above consensus, respectively. Importantly, our estimates exclude China-related revenue. Should NVIDIA resume GPU exports to China, that would represent an additional upside to earnings.\n\nThe new $200 target price is derived by applying the same 30x P/E multiple to our revised 2026 EPS estimate. We maintain our Buy rating.\n\nBroadcom (AVGO.O) – Fourth AI Customer Announced, Estimated to be OpenAI with $10B Order for 2H 2026. Estimates Raised, Buy Rating Maintained.\n\nBroadcom reported solid results and guidance. Strength in the AI segment (33% of Q3 FY25 revenue) continued, though partially offset by higher stock-based compensation expenses. The company announced a fourth AI customer, which we believe to be OpenAI. We estimate this will add $10 billion of incremental revenue in Q3 2026, although we remain cautious on the extreme volatility of the AI business.\n\nWe are raising our estimates for Broadcom (AVGO), maintaining our Buy rating, and lifting our target price from $315 to $350.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": "CITI: NVIDIA Price Target Cut, Broadcom XPU and GPU Pose a Real Threat\n\n• CITI lowered its NVIDIA price target to $200 due to intensifying TPU competition. As a result, it forecasts NVIDIA’s GPU revenue in 2026 will decline by about $12 billion. Specifically, the 2026 XPU market", "parent_author": null, "ret_m1d": -0.007664302707223625, "ret_signed_m1d": -0.007664302707223625, "alpha_spy_m1d": -0.005213780608743446, "alpha_c_m1d": 0.0032623460281404615, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014556426417988044, "ret_signed_p1d": 0.014556426417988044, "alpha_spy_p1d": 0.012244505720378163, "alpha_c_p1d": 0.01020591924940728, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.0561465529724674, "ret_signed_p1w": 0.0561465529724674, "alpha_spy_p1w": 0.037528428861316465, "alpha_c_p1w": 0.02022978166409306, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": 0.09946193196537245, "ret_signed_p1m": 0.09946193196537245, "alpha_spy_p1m": 0.06533107611547795, "alpha_c_p1m": -0.03722369392384106, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.08965935717095963, "ret_signed_p3m": 0.08965935717095963, "alpha_spy_p3m": 0.03192860078014004, "alpha_c_p3m": -0.12982001286651723, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": 0.06987213437582929, "ret_signed_p6m": 0.06987213437582929, "alpha_spy_p6m": 0.015309098297438872, "alpha_c_p6m": -0.25306341111572284, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.1514, -0.1385, -0.1565, -0.1403, -0.1437, -0.1356, -0.1356, -0.1453, -0.1434, -0.1213, -0.0832, -0.079, -0.0627, -0.0613, -0.0892, -0.0657, -0.0533, -0.0533, -0.0598, -0.0494, -0.0323, -0.025, -0.0201, -0.0252, 0.0142, 0.0182, 0.0279, 0.0244, 0.0182, -0.0076, 0.0147, 0.0323, 0.0308, 0.0501, 0.0428, 0.0651, 0.0568, 0.0321, 0.0695, 0.0591, 0.066, 0.074, 0.0855, 0.0817, 0.0882, 0.0789, 0.0815, 0.0721, 0.0814, 0.0436, 0.0421, 0.0396, 0.0575, 0.0683, 0.08, 0.079, 0.0705, 0.0349, 0.0349, 0.0147, 0.0137, 0.0199, -0.0077, 0.0, 0.0146, 0.0536, 0.0527, 0.0566, 0.0561, 0.0391, 0.0118, 0.0472, 0.0497, 0.091, 0.0602, 0.0515, 0.0558, 0.0588, 0.0805, 0.1086, 0.1125, 0.1223, 0.1148, 0.1024, 0.0995, 0.1236, 0.1442, 0.0883, 0.119, 0.0697, 0.0685, 0.0803, 0.0886, 0.0852, 0.0764, 0.0712, 0.0823, 0.1067, 0.1378, 0.1945, 0.2302, 0.2055, 0.2031, 0.2292, 0.1806, 0.1599, 0.1175, 0.1179, 0.1827, 0.1477, 0.1515, 0.1103, 0.1299, 0.1087, 0.0776, 0.1083, 0.0733, 0.0629, 0.0847, 0.0566, 0.0711, 0.0711, 0.0517, 0.069, 0.0782, 0.0671, 0.0897, 0.0839, 0.1026, 0.0991, 0.092, 0.0751, 0.04, 0.0475, 0.056, 0.0157, 0.0348, 0.0755, 0.0915, 0.1243, 0.1207, 0.1207, 0.1321, 0.1184, 0.1144, 0.1082, 0.1082, 0.1222, 0.1178, 0.1126, 0.1237, 0.0995, 0.0985, 0.0989, 0.1041, 0.0882, 0.1115, 0.1066, 0.1066, 0.0581, 0.0893, 0.0983, 0.1152, 0.108, 0.1202, 0.138, 0.1439, 0.1357, 0.1029, 0.0716, 0.0351, 0.0213, 0.1017, 0.1292, 0.1203, 0.1293, 0.1108, 0.0863, 0.0863, 0.0991, 0.117, 0.1165, 0.1279, 0.1382, 0.1459, 0.162, 0.0986, 0.0529, 0.0843, 0.0699]}
{"unit_id": "quote:1965031634086019245", "ticker_idx": 1, "ts": "2025-09-08T12:37:04+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We are raising our estimates for Broadcom (AVGO), maintaining our Buy rating, and lifting our target price from $315 to $350.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi note: Buy NVDA (PT cut to $200) + Buy AVGO (PT raised to $350, 4th AI customer est. OpenAI with $10B order).", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Citi) NVIDIA – GPU vs. XPU: Target Price Lowered to $200 Amid Intensifying TPU Competition\n\nFollowing Broadcom’s comments last week that year-over-year growth in XPUs is accelerating due to both new and existing customers, we are lowering our 2026 GPU revenue forecast by about 4% and cutting our target price (TP) to $200. We had previously expected AI XPU chip sales to surpass GPU sales in 2026. Broadcom’s remarks on the rapid pace of XPU adoption appear to stem from Google’s strategy shift toward providing compute power to competitors such as Meta, OpenAI, and Oracle, thereby indirectly competing with NVIDIA. This has also been a risk we recently flagged.\n\nWe estimate that these transactions will impact NVIDIA’s 2026 revenue by reducing GPU sales by roughly $12 billion. Supported by growth in spending from new cloud players (“neo cloud”) and sovereign AI initiatives worldwide, our revised 2025/2026 forecasts remain 2% and 5% above consensus, respectively. Importantly, our estimates exclude China-related revenue. Should NVIDIA resume GPU exports to China, that would represent an additional upside to earnings.\n\nThe new $200 target price is derived by applying the same 30x P/E multiple to our revised 2026 EPS estimate. We maintain our Buy rating.\n\nBroadcom (AVGO.O) – Fourth AI Customer Announced, Estimated to be OpenAI with $10B Order for 2H 2026. Estimates Raised, Buy Rating Maintained.\n\nBroadcom reported solid results and guidance. Strength in the AI segment (33% of Q3 FY25 revenue) continued, though partially offset by higher stock-based compensation expenses. The company announced a fourth AI customer, which we believe to be OpenAI. We estimate this will add $10 billion of incremental revenue in Q3 2026, although we remain cautious on the extreme volatility of the AI business.\n\nWe are raising our estimates for Broadcom (AVGO), maintaining our Buy rating, and lifting our target price from $315 to $350.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": "CITI: NVIDIA Price Target Cut, Broadcom XPU and GPU Pose a Real Threat\n\n• CITI lowered its NVIDIA price target to $200 due to intensifying TPU competition. As a result, it forecasts NVIDIA’s GPU revenue in 2026 will decline by about $12 billion. Specifically, the 2026 XPU market", "parent_author": null, "ret_m1d": -0.031129663717725742, "ret_signed_m1d": -0.031129663717725742, "alpha_spy_m1d": -0.028679141619245563, "alpha_c_m1d": -0.020203014982361656, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025979983430521436, "ret_signed_p1d": -0.025979983430521436, "alpha_spy_p1d": -0.028291904128131318, "alpha_c_p1d": -0.0303304905991022, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.05334865965704916, "ret_signed_p1w": 0.05334865965704916, "alpha_spy_p1w": 0.03473053554589822, "alpha_c_p1w": 0.017431888348674818, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": -0.025064755049650667, "ret_signed_p1m": -0.025064755049650667, "alpha_spy_p1m": -0.05919561089954517, "alpha_c_p1m": -0.16175038093886418, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.1042466170647871, "ret_signed_p3m": 0.1042466170647871, "alpha_spy_p3m": 0.04651586067396751, "alpha_c_p3m": -0.11523275297268976, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": -0.08873355372693392, "ret_signed_p6m": -0.08873355372693392, "alpha_spy_p6m": -0.14329658980532434, "alpha_c_p6m": -0.41166909921848605, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.27, -0.2609, -0.2822, -0.2724, -0.2802, -0.2748, -0.2748, -0.2768, -0.2658, -0.2369, -0.2343, -0.2184, -0.2207, -0.2025, -0.2341, -0.2192, -0.2039, -0.2039, -0.2068, -0.2137, -0.196, -0.2032, -0.2062, -0.2027, -0.1872, -0.1876, -0.1713, -0.1803, -0.1662, -0.194, -0.1793, -0.1647, -0.1605, -0.1486, -0.1395, -0.1245, -0.1503, -0.1649, -0.1387, -0.1525, -0.1272, -0.1212, -0.1177, -0.1208, -0.095, -0.1058, -0.0996, -0.1137, -0.1154, -0.1468, -0.1576, -0.1622, -0.1494, -0.1488, -0.1378, -0.1313, -0.107, -0.1396, -0.1396, -0.1372, -0.1252, -0.1144, -0.0311, 0.0, -0.026, 0.0692, 0.0404, 0.0411, 0.0533, 0.0415, 0.0015, -0.0009, -0.0021, -0.0182, -0.0177, -0.0167, -0.026, -0.0305, -0.0497, -0.0439, -0.0338, -0.0199, -0.0194, -0.0277, -0.0251, 0.0013, -0.0001, -0.0592, 0.0337, -0.0027, 0.0182, 0.0263, 0.0124, 0.0121, -0.007, -0.0138, -0.0022, 0.0263, 0.0492, 0.0809, 0.1186, 0.091, 0.0712, 0.0507, 0.0199, 0.0403, 0.0305, 0.0127, 0.0386, 0.02, 0.0294, -0.0147, -0.0075, -0.007, -0.0132, 0.0271, 0.0051, -0.0141, 0.0953, 0.1158, 0.1522, 0.1522, 0.1678, 0.1189, 0.1058, 0.103, 0.1042, 0.1309, 0.1624, 0.1775, 0.1968, 0.1777, 0.0431, -0.0152, -0.0109, -0.0552, -0.044, -0.0136, -0.0086, 0.0143, 0.0169, 0.0169, 0.0224, 0.0145, 0.0158, 0.0049, 0.0049, 0.0094, -0.0028, -0.0018, -0.0026, -0.0346, 0.0017, 0.0227, 0.0296, -0.0131, -0.004, 0.0212, 0.0212, -0.0343, -0.0453, -0.0549, -0.0707, -0.0568, -0.0337, -0.0324, -0.0397, -0.038, -0.0386, -0.0699, -0.1055, -0.0984, -0.0333, -0.0013, -0.0115, -0.0048, -0.0384, -0.0558, -0.0558, -0.0344, -0.0316, -0.0302, -0.0341, -0.0408, -0.0549, -0.0351, -0.0659, -0.0722, -0.0743, -0.0887]}
{"unit_id": "orig:1951089623234650248", "ticker_idx": 0, "ts": "2025-08-01T01:16:30+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's earnings bottomed out in Q2 and project a bullish outlook for Samsung", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung: earnings bottomed in Q2, inventory losses shrink in Q3, HBM ramp underway — projects recovery from here.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "What We Can Glean from Samsung Electronics' Earnings Release\n\nThe increase in DRAM's ASP (Average Selling Price) was lower than expected. The company announced a 30% increase in bit growth. Since Q1's bit growth was 600 million Gb, Q2's was nearly 800 million Gb. This suggests that the ASP movement was not significant compared to the volume increase. Another interpretation is that the company lowered HBM prices significantly, resulting in no mix improvement.\n\nRegarding the Q3 outlook, it's estimated that there will be considerable noise in the semiconductor sector.\nFirst, DRAM bit growth is projected to be around 7%. The key issue is pricing, and it seems that an overall price increase is a plausible scenario. Considering that Samsung tried to maximize shipments in Q2 to clear out DDR4 inventory before prices rose in Q3, a potential price increase for DDR4 in Q3 might be suppressed.\n\nAs for HBM, the company aims for a 100% quarter-on-quarter increase in Q3, which amounts to roughly 1.5 billion Gb. Not all suppliers have to meet all demand, but this year, SK Hynix and Micron combined are at around 16 billion Gb, while market demand is 20-22 billion Gb. This indicates a significant supply shortage compared to demand.\nIn the conference call, Samsung mentioned something that could be interpreted as a potential HBM oversupply. This statement seems to be somewhat disconnected from the current market sentiment. If we are to interpret it, it might be related to the recent price cuts for 12-stack HBM in the second half of the year. It could also be interpreted that with Micron's entry, prices are falling, and the supply situation is not as tight as before, and will fall further when Samsung enters. Conversely, it could also mean that we will enter the market at a lower price than before.\n\nIn order for Samsung to ship 1.5 billion Gb, they will need to supply NVIDIA. There are rumors that the 12-stack HBM qualification will proceed in August. There are also rumors that HBM3E still has thermal issues. From NVIDIA's perspective, which continues to push for lower prices, they need to bring in as many suppliers as possible, so there is a possibility that NVIDIA might lower their standards.\n\nFinally, Samsung stated that Q2 was not the last quarter for inventory valuation losses. While they did not disclose the exact scale of the Q2 losses, they did mention that the losses would be significantly reduced in Q3. Based on this, I conclude that Samsung's earnings bottomed out in Q2 and project a bullish outlook for Samsung.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03628451801845611, "ret_signed_m1d": 0.03628451801845611, "alpha_spy_m1d": 0.019620828077359542, "alpha_c_m1d": 0.014228266800442535, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.02205625121801358, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01161109622749179, "ret_signed_p1d": 0.01161109622749179, "alpha_spy_p1d": -0.0035887179614699782, "alpha_c_p1d": -0.008694791299870719, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02030588752736251, "ret_p1w": 0.042090123474913144, "ret_signed_p1w": 0.042090123474913144, "alpha_spy_p1w": 0.017223541503019302, "alpha_c_p1w": 0.007663562763754239, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.034426560711158904, "ret_p1m": -0.01886781633450707, "ret_signed_p1m": -0.01886781633450707, "alpha_spy_p1m": -0.05639287429155182, "alpha_c_p1m": -0.039796134858980814, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.020928318524473744, "ret_p3m": 0.4651436595476073, "ret_signed_p3m": 0.4651436595476073, "alpha_spy_p3m": 0.3564519244954558, "alpha_c_p3m": 0.28056009055783004, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.18458356898977724, "ret_p6m": 1.2281755934741243, "ret_signed_p6m": 1.2281755934741243, "alpha_spy_p6m": 1.1075694978265855, "alpha_c_p6m": 1.0884890617390235, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.13968653173510082, "price_path": [-0.2167, -0.2124, -0.2124, -0.2095, -0.1691, -0.1792, -0.172, -0.1734, -0.1806, -0.1951, -0.1936, -0.1965, -0.2109, -0.2181, -0.2109, -0.2225, -0.1936, -0.1907, -0.1893, -0.1806, -0.1806, -0.1662, -0.1475, -0.1475, -0.1374, -0.146, -0.1359, -0.1417, -0.159, -0.1749, -0.1619, -0.1374, -0.146, -0.1417, -0.1633, -0.1273, -0.1157, -0.1316, -0.1176, -0.1321, -0.1263, -0.1176, -0.074, -0.0813, -0.1045, -0.1089, -0.1234, -0.1147, -0.0914, -0.0929, -0.0755, -0.061, -0.0319, -0.0261, -0.016, -0.0421, -0.0363, -0.0421, -0.0435, 0.0218, 0.0247, 0.0537, 0.0363, 0.0, 0.0116, 0.0145, -0.0015, 0.0232, 0.0421, 0.0305, 0.0319, 0.0435, 0.0392, 0.0392, 0.016, 0.016, 0.0232, 0.0247, 0.0363, 0.0377, 0.0203, 0.0247, 0.0102, 0.0116, -0.0189, 0.0029, 0.0131, 0.0174, 0.0087, 0.0174, 0.0377, 0.0537, 0.0653, 0.0943, 0.1103, 0.1524, 0.135, 0.1655, 0.1655, 0.2119, 0.2293, 0.2395, 0.2496, 0.209, 0.2275, 0.2231, 0.2538, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3762, 0.3602, 0.3354, 0.385, 0.4243, 0.4272, 0.4302, 0.4214, 0.4374, 0.4068, 0.4404, 0.487, 0.4506, 0.4651, 0.5176, 0.5672, 0.6197, 0.5293, 0.4666, 0.4462, 0.4272, 0.4666, 0.5089, 0.503, 0.4987, 0.417, 0.4666, 0.4258, 0.4068, 0.4666, 0.382, 0.4097, 0.4476, 0.4987, 0.5089, 0.4651, 0.4695, 0.5074, 0.5235, 0.5322, 0.5803, 0.5964, 0.5803, 0.5745, 0.5643, 0.5876, 0.5278, 0.4987, 0.573, 0.5687, 0.5497, 0.6109, 0.6255, 0.6197, 0.6197, 0.7057, 0.7506, 0.7565, 0.7565, 0.7565, 0.8824, 1.0231, 1.0348, 1.0656, 1.0333, 1.0363, 1.0333, 1.0158, 1.0553, 1.1081, 1.1813, 1.1872, 1.1271, 1.1901, 1.2311, 1.2282, 1.2282]}
{"unit_id": "quote:1970330663015547159", "ticker_idx": 0, "ts": "2025-09-23T03:33:31+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "That's the reason I'm bullish on BESI.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish BESI on increasing adoption of hybrid bonding in AI accelerator packaging.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "The increasing adoption of hybrid bonding is also worth noting.\n\nThat’s the reason I’m bullish on BESI. https://t.co/MGfR4z0XaZ", "reply_text": null, "parent_text": null, "quoted_text": "AI Accelerator Roadmap Summary\n\nFeynman’s specs stand out.\n\nIt uses the A16 node,\n\nand notably, its HBM4E capacity is smaller than Rubin Ultra’s.\n\nThe speculation that Jaguar Shore might be built on Intel 18A is also interesting. https://t.co/XBDC4mbeAf", "parent_author": null, "ret_m1d": -0.03230887504278135, "ret_signed_m1d": -0.03230887504278135, "alpha_spy_m1d": -0.03778214348162723, "alpha_c_m1d": -0.033769845678061294, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013002369858779561, "ret_signed_p1d": 0.013002369858779561, "alpha_spy_p1d": 0.016184003311393247, "alpha_c_p1d": 0.0148364575305302, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.0003940387079959251, "ret_signed_p1w": -0.0003940387079959251, "alpha_spy_p1w": -0.004872167430687968, "alpha_c_p1w": -0.01487946787394856, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.09141050671877449, "ret_signed_p1m": 0.09141050671877449, "alpha_spy_p1m": 0.08448969501721981, "alpha_c_p1m": 0.03887713149026295, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.030338802503479867, "ret_signed_p3m": 0.030338802503479867, "alpha_spy_p3m": 0.001100649890770633, "alpha_c_p3m": -0.0769971960965461, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 0.5161544677715602, "ret_signed_p6m": 0.5161544677715602, "alpha_spy_p6m": 0.5158915220932125, "alpha_c_p6m": 0.2886290121210542, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [0.0229, 0.028, 0.0012, -0.0374, -0.0288, -0.039, -0.0461, -0.0433, -0.037, -0.039, 0.0039, 0.002, -0.0177, -0.0047, -0.0548, 0.0047, 0.0083, 0.0256, -0.0087, -0.0256, 0.0008, -0.0772, -0.0323, -0.0544, -0.0355, -0.0615, -0.0768, -0.067, -0.0914, -0.0954, -0.0548, -0.0485, -0.0422, -0.0248, -0.0268, -0.0335, -0.0658, -0.0717, -0.0591, -0.0772, -0.0902, -0.0623, -0.0611, -0.0489, -0.0559, -0.0563, -0.093, -0.0898, -0.1576, -0.1663, -0.1556, -0.1434, -0.1265, -0.1217, -0.1147, -0.1044, -0.1186, -0.069, -0.0812, -0.0887, -0.0165, -0.0426, -0.0323, 0.0, 0.013, -0.0016, -0.0355, -0.0039, -0.0004, 0.0071, 0.0524, 0.0284, 0.1564, 0.1568, 0.1533, 0.1458, 0.1056, 0.1249, 0.1072, 0.1466, 0.1434, 0.1166, 0.1442, 0.1371, 0.0914, 0.1367, 0.1588, 0.1663, 0.1552, 0.1635, 0.1588, 0.1631, 0.1489, 0.1233, 0.1151, 0.0855, 0.0571, 0.0753, 0.0859, 0.0666, 0.0544, 0.0355, 0.026, 0.002, 0.0173, 0.0248, -0.0323, -0.0197, -0.0181, 0.0213, 0.0236, 0.0236, 0.0225, 0.0351, 0.0713, 0.0835, 0.0997, 0.1572, 0.1292, 0.1068, 0.0788, 0.041, 0.0437, 0.0528, 0.0189, 0.0359, 0.0303, 0.0339, 0.0394, 0.0366, 0.0366, 0.0366, 0.041, 0.0489, 0.054, 0.054, 0.1749, 0.2116, 0.2679, 0.2514, 0.1907, 0.1907, 0.2778, 0.3176, 0.2727, 0.3684, 0.3625, 0.3302, 0.3625, 0.3716, 0.379, 0.3794, 0.3779, 0.3849, 0.3353, 0.2809, 0.2963, 0.2971, 0.262, 0.2415, 0.2667, 0.3042, 0.3191, 0.3322, 0.3467, 0.3294, 0.3885, 0.4066, 0.4192, 0.4736, 0.3633, 0.4539, 0.4787, 0.5154, 0.5571, 0.4886, 0.4925, 0.4854, 0.4291, 0.5051, 0.4866, 0.2317, 0.3105, 0.3759, 0.3786, 0.381, 0.4586, 0.4421, 0.485, 0.5162]}
{"unit_id": "quote:1963033676972994591", "ticker_idx": 0, "ts": "2025-09-03T00:17:54+00:00", "symbol": "GIS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bullish on Apple foldable value chain", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish GIS as Apple foldable iPhone/iPad supply chain beneficiary on raised shipment forecasts for 2026–2027.", "resolved_tickers": ["6456.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "GIS = General Interface Solution 6456.TW; Apple foldable context", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "quote", "lang": "en", "tweet_text": "Bullish on Apple foldable value chain\n\n\"Apple recently revised its shipment forecasts for the foldable iPhone upward to 8–10 million units in 2026 and 20–25 million in 2027 (vs. previous estimate of 6–8 million and 10–15 million, respectively), which will significantly boost demand for relatively low-yield UTG.\"", "reply_text": null, "parent_text": null, "quoted_text": "GIS: Worst Over, Exclusive Apple Foldable iPhone/iPad and Vision Air Supply, Enhanced Long-Term Profitability Visibility from Next-Gen Orders\n\nFull story\nhttps://t.co/xg3iixB75m https://t.co/hlPXfHuaZc", "parent_author": null, "ret_m1d": -0.06470590479233684, "ret_signed_m1d": -0.06470590479233684, "alpha_spy_m1d": -0.0593155583831696, "alpha_c_m1d": -0.05877510055671131, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": -0.005930804235625531, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04117651546702661, "ret_signed_p1d": -0.04117651546702661, "alpha_spy_p1d": -0.049533875908841174, "alpha_c_p1d": -0.04611252126159571, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.004936005794569098, "ret_p1w": -0.08235291873707495, "ret_signed_p1w": -0.08235291873707495, "alpha_spy_p1w": -0.0955104737798228, "alpha_c_p1w": -0.11602423733214995, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.033671318595074995, "ret_p1m": -0.09117648180793314, "ret_signed_p1m": -0.09117648180793314, "alpha_spy_p1m": -0.13363977851526465, "alpha_c_p1m": -0.18751388836312144, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.0963374065551883, "ret_p3m": -0.3801470363841337, "ret_signed_p3m": -0.3801470363841337, "alpha_spy_p3m": -0.43982341289027826, "alpha_c_p3m": -0.4772123985693184, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.09706536218518469, "ret_p6m": -0.2977941176470589, "ret_signed_p6m": -0.2977941176470589, "alpha_spy_p6m": -0.3747094863207836, "alpha_c_p6m": -0.37990913065037013, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.08211501300331125, "price_path": [-0.3103, -0.3279, -0.3206, -0.3412, -0.3456, -0.3721, -0.3787, -0.3868, -0.3868, -0.4074, -0.414, -0.4206, -0.4037, -0.3971, -0.3978, -0.3926, -0.4088, -0.4037, -0.4051, -0.3787, -0.3904, -0.3993, -0.4066, -0.4059, -0.4066, -0.3985, -0.4103, -0.3515, -0.3662, -0.3485, -0.3478, -0.339, -0.3603, -0.3449, -0.3382, -0.3243, -0.325, -0.3221, -0.3191, -0.3243, -0.3243, -0.3257, -0.3184, -0.3221, -0.3228, -0.325, -0.3162, -0.3103, -0.3176, -0.3088, -0.2765, -0.2721, -0.2426, -0.2588, -0.2015, -0.2, -0.2, -0.1632, -0.1221, -0.1191, -0.0735, -0.0721, -0.0647, 0.0, -0.0412, -0.0309, -0.0735, -0.1059, -0.0824, -0.0824, -0.0721, -0.1, -0.0706, -0.0794, -0.0706, -0.0574, -0.0632, -0.0676, -0.0853, -0.0662, -0.1, -0.1, -0.0779, -0.0926, -0.0912, -0.1088, -0.1088, -0.0735, -0.1412, -0.1662, -0.1662, -0.1809, -0.2235, -0.2338, -0.2338, -0.2382, -0.2353, -0.2324, -0.2309, -0.2397, -0.2397, -0.2324, -0.2382, -0.2471, -0.2618, -0.2721, -0.2794, -0.2971, -0.2978, -0.2882, -0.3015, -0.3125, -0.3375, -0.3162, -0.3279, -0.3456, -0.3551, -0.3787, -0.3912, -0.3794, -0.3868, -0.3978, -0.3882, -0.3772, -0.3713, -0.3721, -0.3801, -0.3801, -0.3728, -0.3779, -0.3794, -0.3735, -0.3647, -0.3015, -0.3419, -0.3294, -0.3471, -0.3654, -0.3588, -0.3662, -0.3559, -0.3449, -0.3574, -0.3603, -0.3603, -0.375, -0.3699, -0.3699, -0.3713, -0.3713, -0.375, -0.3912, -0.3757, -0.3676, -0.3846, -0.3846, -0.3691, -0.3757, -0.3662, -0.361, -0.364, -0.3544, -0.3618, -0.3801, -0.3728, -0.3779, -0.3743, -0.3868, -0.3941, -0.4118, -0.4235, -0.4324, -0.4096, -0.3779, -0.3978, -0.3838, -0.3801, -0.3574, -0.3676, -0.3676, -0.3676, -0.3676, -0.3676, -0.3676, -0.3676, -0.3676, -0.3654, -0.3485, -0.361, -0.2978]}
{"unit_id": "thread:1973932507918987331", "ticker_idx": 0, "ts": "2025-10-03T02:05:58+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia is positioning itself as the new king of AI networking by completing the last piece of the AI infrastructure puzzle with Spectrum-X", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA dominates AI networking via Spectrum-X (647% YoY growth); Arista benefits from data-center focus; Cisco loses ground with weak DC growth and legacy exposure.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Nvidia’s Spectrum-X dominates Ethernet switch market, surpassing Cisco\n\n    • Building on the success of its Spectrum-X platform, Nvidia is replicating its computing dominance in the networking domain.\n    • Spectrum-X is a dedicated Ethernet solution designed for hyperscale GenAI clusters.\n    • It delivers InfiniBand-level performance on Ethernet while breaking the barriers of closed ecosystems and high costs.\n    • According to IDC, the data center Ethernet market revenue in Q2 2025 surged 71.6% year-on-year.\n    • Over the same period, Nvidia’s switch business revenue skyrocketed 647% YoY to $2.3 billion, capturing 25.9% market share and overtaking Cisco and Arista.\n    • In an era where GPUs are scaling into the millions, networking is no longer a background element but a core factor determining AI training efficiency.\n    • Nvidia is positioning itself as the new king of AI networking by completing the last piece of the AI infrastructure puzzle with Spectrum-X.\n    • IDC data shows the global Ethernet switch market reached $14.5 billion in Q2 2025, up 42.1% YoY, driven by data center demand.\n    • With model parameters and training datasets growing exponentially, requirements for compute density, memory bandwidth, and network communication are intensifying.\n    • Hyperscalers and cloud service providers are accelerating the buildout of AI-dedicated data centers with bandwidth of 800GbE and above, driving growth in the high-speed switch market.\n    • Nvidia grew 647% YoY centered on data centers, while Arista rose 33.5%, with 90.7% of its revenue from data centers.\n    • Cisco still held the top spot with $4 billion in revenue, but its data center growth was just 9.1%, with more than two-thirds of its sales coming from the slower-growing non–data center segment.\n    • As legacy leaders lose ground at the gateway to the AI era, Nvidia’s new weapon, Spectrum-X, underscores its competitive edge.\n    • Nvidia CFO Colette Kress announced plans to release new Spectrum-X chips and switches annually to support GPU clusters ranging from tens of thousands to millions.\n    • InfiniBand still retains a performance edge thanks to ultra-low latency, native RDMA architecture, and in-network computing.\n    • Ethernet, however, is preferred by many vendors for its lower cost and flexibility. Nvidia’s Ethernet platform continues to improve with technologies such as lossless networking, RDMA, and congestion control.\n    • IDC forecasts that Ethernet and InfiniBand will continue to maintain distinct roles in the future.\n\n$NVDA\n\n---\n\nhttps://t.co/wxCUaCmOYJ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006769035359998732, "ret_signed_m1d": 0.006769035359998732, "alpha_spy_m1d": 0.006754267779698786, "alpha_c_m1d": 0.002135511121102329, "bench_spy_m1d": 1.4767580299945848e-05, "bench_c_m1d": 0.004633524238896403, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011086245458746924, "ret_signed_p1d": -0.011086245458746924, "alpha_spy_p1d": -0.014672474369083033, "alpha_c_p1d": -0.03098648648202118, "bench_spy_p1d": 0.003586228910336109, "bench_c_p1d": 0.019900241023274257, "ret_p1w": -0.023771415445946076, "ret_signed_p1w": -0.023771415445946076, "alpha_spy_p1w": 0.000421366601079054, "alpha_c_p1w": 0.009910377822496041, "bench_spy_p1w": -0.02419278204702513, "bench_c_p1w": -0.03368179326844212, "ret_p1m": 0.1026543220821905, "ret_signed_p1m": 0.1026543220821905, "alpha_spy_p1m": 0.08153989259692773, "alpha_c_p1m": 0.015271564975240848, "bench_spy_p1m": 0.021114429485262765, "bench_c_p1m": 0.08738275710694965, "ret_p3m": -0.005914101199933519, "ret_signed_p3m": -0.005914101199933519, "alpha_spy_p3m": -0.027917612398528968, "alpha_c_p3m": -0.07889319905931069, "bench_spy_p3m": 0.02200351119859545, "bench_c_p3m": 0.07297909785937717, "ret_p6m": -0.11956007370441812, "ret_signed_p6m": -0.11956007370441812, "alpha_spy_p6m": -0.06928939588165473, "alpha_c_p6m": -0.19968976510224767, "bench_spy_p6m": -0.0502706778227634, "bench_c_p6m": 0.08012969139782955, "price_path": [-0.1473, -0.1319, -0.1254, -0.121, -0.1256, -0.0902, -0.0867, -0.078, -0.0811, -0.0866, -0.1098, -0.0898, -0.074, -0.0753, -0.058, -0.0646, -0.0446, -0.052, -0.0741, -0.0407, -0.0499, -0.0438, -0.0366, -0.0263, -0.0297, -0.0238, -0.0322, -0.0299, -0.0383, -0.03, -0.0639, -0.0652, -0.0674, -0.0514, -0.0417, -0.0312, -0.0321, -0.0398, -0.0717, -0.0717, -0.0898, -0.0907, -0.0851, -0.1098, -0.103, -0.0899, -0.0549, -0.0557, -0.0522, -0.0526, -0.0679, -0.0924, -0.0607, -0.0584, -0.0214, -0.049, -0.0568, -0.0529, -0.0503, -0.0308, -0.0055, -0.002, 0.0068, 0.0, -0.0111, -0.0138, 0.0079, 0.0264, -0.0238, 0.0037, -0.0405, -0.0415, -0.031, -0.0235, -0.0265, -0.0344, -0.0391, -0.0291, -0.0072, 0.0206, 0.0715, 0.1035, 0.0814, 0.0793, 0.1027, 0.059, 0.0405, 0.0025, 0.0028, 0.0609, 0.0295, 0.0329, -0.0041, 0.0136, -0.0054, -0.0334, -0.0059, -0.0372, -0.0466, -0.027, -0.0522, -0.0392, -0.0392, -0.0566, -0.041, -0.0328, -0.0428, -0.0225, -0.0277, -0.011, -0.0141, -0.0204, -0.0356, -0.0671, -0.0603, -0.0527, -0.0889, -0.0718, -0.0353, -0.0209, 0.0085, 0.0053, 0.0053, 0.0156, 0.0033, -0.0004, -0.0059, -0.0059, 0.0066, 0.0027, -0.002, 0.008, -0.0137, -0.0147, -0.0142, -0.0096, -0.0238, -0.003, -0.0074, -0.0074, -0.0508, -0.0229, -0.0148, 0.0003, -0.0061, 0.0049, 0.0208, 0.0261, 0.0188, -0.0107, -0.0387, -0.0715, -0.0838, -0.0117, 0.013, 0.005, 0.013, -0.0036, -0.0256, -0.0256, -0.0141, 0.002, 0.0015, 0.0118, 0.021, 0.0279, 0.0424, -0.0145, -0.0555, -0.0273, -0.0403, -0.0244, -0.0228, -0.0522, -0.0264, -0.0151, -0.0084, -0.0238, -0.0392, -0.0233, -0.0302, -0.0384, -0.0482, -0.0794, -0.0637, -0.0661, -0.0475, -0.0872, -0.107, -0.1196]}
{"unit_id": "thread:1973932507918987331", "ticker_idx": 1, "ts": "2025-10-03T02:05:58+00:00", "symbol": "Arista Networks", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Arista rose 33.5%, with 90.7% of its revenue from data centers", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA dominates AI networking via Spectrum-X (647% YoY growth); Arista benefits from data-center focus; Cisco loses ground with weak DC growth and legacy exposure.", "resolved_tickers": ["ANET"], "resolution_method": "exchange_ticker", "resolution_reason": "Arista Networks ANET US", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Nvidia’s Spectrum-X dominates Ethernet switch market, surpassing Cisco\n\n    • Building on the success of its Spectrum-X platform, Nvidia is replicating its computing dominance in the networking domain.\n    • Spectrum-X is a dedicated Ethernet solution designed for hyperscale GenAI clusters.\n    • It delivers InfiniBand-level performance on Ethernet while breaking the barriers of closed ecosystems and high costs.\n    • According to IDC, the data center Ethernet market revenue in Q2 2025 surged 71.6% year-on-year.\n    • Over the same period, Nvidia’s switch business revenue skyrocketed 647% YoY to $2.3 billion, capturing 25.9% market share and overtaking Cisco and Arista.\n    • In an era where GPUs are scaling into the millions, networking is no longer a background element but a core factor determining AI training efficiency.\n    • Nvidia is positioning itself as the new king of AI networking by completing the last piece of the AI infrastructure puzzle with Spectrum-X.\n    • IDC data shows the global Ethernet switch market reached $14.5 billion in Q2 2025, up 42.1% YoY, driven by data center demand.\n    • With model parameters and training datasets growing exponentially, requirements for compute density, memory bandwidth, and network communication are intensifying.\n    • Hyperscalers and cloud service providers are accelerating the buildout of AI-dedicated data centers with bandwidth of 800GbE and above, driving growth in the high-speed switch market.\n    • Nvidia grew 647% YoY centered on data centers, while Arista rose 33.5%, with 90.7% of its revenue from data centers.\n    • Cisco still held the top spot with $4 billion in revenue, but its data center growth was just 9.1%, with more than two-thirds of its sales coming from the slower-growing non–data center segment.\n    • As legacy leaders lose ground at the gateway to the AI era, Nvidia’s new weapon, Spectrum-X, underscores its competitive edge.\n    • Nvidia CFO Colette Kress announced plans to release new Spectrum-X chips and switches annually to support GPU clusters ranging from tens of thousands to millions.\n    • InfiniBand still retains a performance edge thanks to ultra-low latency, native RDMA architecture, and in-network computing.\n    • Ethernet, however, is preferred by many vendors for its lower cost and flexibility. Nvidia’s Ethernet platform continues to improve with technologies such as lossless networking, RDMA, and congestion control.\n    • IDC forecasts that Ethernet and InfiniBand will continue to maintain distinct roles in the future.\n\n$NVDA\n\n---\n\nhttps://t.co/wxCUaCmOYJ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0071477201796069645, "ret_signed_m1d": -0.0071477201796069645, "alpha_spy_m1d": -0.00716248775990691, "alpha_c_m1d": -0.012205326004710026, "bench_spy_m1d": 1.4767580299945848e-05, "bench_c_m1d": 0.005057605825103062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027491408934707806, "ret_signed_p1d": 0.027491408934707806, "alpha_spy_p1d": 0.023905180024371697, "alpha_c_p1d": 0.017376197284501682, "bench_spy_p1d": 0.003586228910336109, "bench_c_p1d": 0.010115211650206124, "ret_p1w": 0.059106571158183074, "ret_signed_p1w": 0.059106571158183074, "alpha_spy_p1w": 0.0832993532052082, "alpha_c_p1w": 0.08133882012360194, "bench_spy_p1w": -0.02419278204702513, "bench_c_p1w": -0.022232248965418866, "ret_p1m": 0.08309275833601815, "ret_signed_p1m": 0.08309275833601815, "alpha_spy_p1m": 0.06197832885075538, "alpha_c_p1m": 0.022717588798849997, "bench_spy_p1m": 0.021114429485262765, "bench_c_p1m": 0.06037516953716815, "ret_p3m": -0.09945018021101804, "ret_signed_p3m": -0.09945018021101804, "alpha_spy_p3m": -0.12145369140961348, "alpha_c_p3m": -0.11229346375540927, "bench_spy_p3m": 0.02200351119859545, "bench_c_p3m": 0.012843283544391237, "ret_p6m": -0.2018556889799452, "ret_signed_p6m": -0.2018556889799452, "alpha_spy_p6m": -0.15158501115718181, "alpha_c_p6m": -0.09994615688264563, "bench_spy_p6m": -0.0502706778227634, "bench_c_p6m": -0.10190953209729958, "price_path": [-0.2894, -0.2696, -0.2695, -0.2538, -0.2552, -0.2621, -0.2557, -0.2304, -0.2318, -0.2329, -0.2455, -0.2231, -0.2162, -0.2146, -0.1921, -0.1847, -0.1609, -0.1531, -0.192, -0.1729, -0.1882, -0.0462, -0.0427, -0.0434, -0.054, -0.0292, -0.0515, -0.062, -0.0564, -0.0513, -0.0874, -0.0964, -0.0926, -0.0842, -0.0856, -0.0772, -0.0841, -0.0637, -0.0615, -0.0615, -0.0662, -0.0558, -0.0298, -0.0182, -0.0377, -0.0247, 0.0359, 0.0518, -0.042, -0.0005, -0.023, -0.0183, 0.008, 0.0282, -0.0007, -0.0097, -0.0197, -0.0168, -0.0206, -0.0146, 0.0014, 0.0259, -0.0071, 0.0, 0.0275, -0.0014, 0.0815, 0.0875, 0.0591, 0.0134, -0.0461, -0.0146, 0.0035, -0.0165, 0.0067, 0.003, 0.0075, 0.0499, 0.0572, 0.0777, 0.0775, 0.1136, 0.0889, 0.0838, 0.0831, 0.0553, -0.0349, -0.0789, -0.0746, -0.0566, -0.0726, -0.0723, -0.1045, -0.0971, -0.1254, -0.1515, -0.1422, -0.1781, -0.1929, -0.1603, -0.1406, -0.1227, -0.1227, -0.1019, -0.1195, -0.1256, -0.1216, -0.1165, -0.1162, -0.1126, -0.1063, -0.0903, -0.0764, -0.1425, -0.1348, -0.1331, -0.159, -0.1435, -0.0988, -0.1015, -0.0975, -0.1012, -0.1012, -0.0939, -0.078, -0.0898, -0.0995, -0.0995, -0.0818, -0.0571, -0.0888, -0.106, -0.1497, -0.1554, -0.1518, -0.107, -0.1403, -0.1025, -0.1077, -0.1077, -0.1236, -0.1252, -0.0487, -0.063, -0.0122, 0.0082, 0.032, 0.0182, -0.0258, -0.049, -0.042, -0.1046, -0.1157, -0.0551, -0.0258, -0.0141, -0.0333, -0.0713, -0.0269, -0.0269, -0.0201, -0.041, -0.0568, -0.0874, -0.1242, -0.115, -0.0867, -0.1048, -0.0825, -0.1113, -0.1436, -0.0733, -0.0419, -0.0867, -0.0573, -0.0404, -0.05, -0.0788, -0.082, -0.0698, -0.0854, -0.0648, -0.0635, -0.0981, -0.0661, -0.101, -0.0721, -0.1577, -0.17, -0.2019]}
{"unit_id": "thread:1973932507918987331", "ticker_idx": 2, "ts": "2025-10-03T02:05:58+00:00", "symbol": "Cisco", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Cisco still held the top spot with $4 billion in revenue, but its data center growth was just 9.1%, with more than two-thirds of its sales coming from the slower-growing non–data center segment", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA dominates AI networking via Spectrum-X (647% YoY growth); Arista benefits from data-center focus; Cisco loses ground with weak DC growth and legacy exposure.", "resolved_tickers": ["CSCO"], "resolution_method": "exchange_ticker", "resolution_reason": "Cisco Systems CSCO US", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Nvidia’s Spectrum-X dominates Ethernet switch market, surpassing Cisco\n\n    • Building on the success of its Spectrum-X platform, Nvidia is replicating its computing dominance in the networking domain.\n    • Spectrum-X is a dedicated Ethernet solution designed for hyperscale GenAI clusters.\n    • It delivers InfiniBand-level performance on Ethernet while breaking the barriers of closed ecosystems and high costs.\n    • According to IDC, the data center Ethernet market revenue in Q2 2025 surged 71.6% year-on-year.\n    • Over the same period, Nvidia’s switch business revenue skyrocketed 647% YoY to $2.3 billion, capturing 25.9% market share and overtaking Cisco and Arista.\n    • In an era where GPUs are scaling into the millions, networking is no longer a background element but a core factor determining AI training efficiency.\n    • Nvidia is positioning itself as the new king of AI networking by completing the last piece of the AI infrastructure puzzle with Spectrum-X.\n    • IDC data shows the global Ethernet switch market reached $14.5 billion in Q2 2025, up 42.1% YoY, driven by data center demand.\n    • With model parameters and training datasets growing exponentially, requirements for compute density, memory bandwidth, and network communication are intensifying.\n    • Hyperscalers and cloud service providers are accelerating the buildout of AI-dedicated data centers with bandwidth of 800GbE and above, driving growth in the high-speed switch market.\n    • Nvidia grew 647% YoY centered on data centers, while Arista rose 33.5%, with 90.7% of its revenue from data centers.\n    • Cisco still held the top spot with $4 billion in revenue, but its data center growth was just 9.1%, with more than two-thirds of its sales coming from the slower-growing non–data center segment.\n    • As legacy leaders lose ground at the gateway to the AI era, Nvidia’s new weapon, Spectrum-X, underscores its competitive edge.\n    • Nvidia CFO Colette Kress announced plans to release new Spectrum-X chips and switches annually to support GPU clusters ranging from tens of thousands to millions.\n    • InfiniBand still retains a performance edge thanks to ultra-low latency, native RDMA architecture, and in-network computing.\n    • Ethernet, however, is preferred by many vendors for its lower cost and flexibility. Nvidia’s Ethernet platform continues to improve with technologies such as lossless networking, RDMA, and congestion control.\n    • IDC forecasts that Ethernet and InfiniBand will continue to maintain distinct roles in the future.\n\n$NVDA\n\n---\n\nhttps://t.co/wxCUaCmOYJ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0002945283356923234, "ret_signed_m1d": 0.0002945283356923234, "alpha_spy_m1d": 0.00030929591599226924, "alpha_c_m1d": 0.00030929591599226924, "bench_spy_m1d": 1.4767580299945848e-05, "bench_c_m1d": 1.4767580299945848e-05, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01457620052088604, "ret_signed_p1d": -0.01457620052088604, "alpha_spy_p1d": -0.010989971610549931, "alpha_c_p1d": -0.010989971610549931, "bench_spy_p1d": 0.003586228910336109, "bench_c_p1d": 0.003586228910336109, "ret_p1w": 0.0002945283356923234, "ret_signed_p1w": -0.0002945283356923234, "alpha_spy_p1w": -0.024487310382717453, "alpha_c_p1w": -0.024487310382717453, "bench_spy_p1w": -0.02419278204702513, "bench_c_p1w": -0.02419278204702513, "ret_p1m": 0.09614260984805556, "ret_signed_p1m": -0.09614260984805556, "alpha_spy_p1m": -0.0750281803627928, "alpha_c_p1m": -0.0750281803627928, "bench_spy_p1m": 0.021114429485262765, "bench_c_p1m": 0.021114429485262765, "ret_p3m": 0.1341284765754598, "ret_signed_p3m": -0.1341284765754598, "alpha_spy_p3m": -0.11212496537686434, "alpha_c_p3m": -0.11212496537686434, "bench_spy_p3m": 0.02200351119859545, "bench_c_p3m": 0.02200351119859545, "ret_p6m": 0.140345249884698, "ret_signed_p6m": -0.140345249884698, "alpha_spy_p6m": -0.19061592770746139, "alpha_c_p6m": -0.19061592770746139, "bench_spy_p6m": -0.0502706778227634, "bench_c_p6m": -0.0502706778227634, "price_path": [0.0038, 0.0138, 0.0063, -0.0056, -0.0075, -0.0168, -0.0141, -0.0004, -0.0041, 0.0001, -0.0025, 0.0042, -0.0006, 0.0053, -0.006, -0.0054, -0.0007, -0.0037, -0.0179, 0.0023, -0.0119, 0.0129, 0.023, 0.0506, 0.0342, 0.0446, 0.0303, 0.0142, -0.0312, -0.0202, -0.023, -0.0179, -0.0192, -0.0148, -0.0174, 0.0009, 0.0016, 0.0161, 0.0111, 0.0111, -0.0078, -0.0097, -0.005, -0.0209, -0.0212, -0.0145, -0.0029, -0.0094, -0.0263, -0.0192, -0.0205, -0.0089, 0.0051, -0.0018, -0.0085, -0.0116, -0.0148, -0.007, -0.0162, -0.0089, 0.0013, 0.0061, -0.0003, 0.0, 0.0146, 0.0158, 0.0355, 0.03, 0.0003, -0.0068, 0.0109, 0.0236, 0.0172, 0.0325, 0.0402, 0.0412, 0.0403, 0.0346, 0.0399, 0.0511, 0.0692, 0.0502, 0.0735, 0.0764, 0.0961, 0.0648, 0.0615, 0.0459, 0.0464, 0.0614, 0.0558, 0.0889, 0.1393, 0.1484, 0.1452, 0.1391, 0.1542, 0.1107, 0.1204, 0.1225, 0.1237, 0.12, 0.12, 0.1328, 0.1196, 0.1318, 0.1447, 0.1449, 0.148, 0.1611, 0.1706, 0.1815, 0.1671, 0.1455, 0.1521, 0.1418, 0.119, 0.133, 0.1546, 0.15, 0.1487, 0.1487, 0.1487, 0.1508, 0.1453, 0.1397, 0.1341, 0.1341, 0.1255, 0.1187, 0.1136, 0.106, 0.0948, 0.0936, 0.0955, 0.1171, 0.1014, 0.1138, 0.113, 0.113, 0.0857, 0.0908, 0.1002, 0.1041, 0.1399, 0.1646, 0.1688, 0.1609, 0.1593, 0.1936, 0.2302, 0.2013, 0.2191, 0.2555, 0.2845, 0.2773, 0.2662, 0.1101, 0.1375, 0.1375, 0.1375, 0.1572, 0.1628, 0.1723, 0.1507, 0.1566, 0.1711, 0.156, 0.1762, 0.1756, 0.1688, 0.197, 0.1843, 0.164, 0.1281, 0.1501, 0.156, 0.1507, 0.1594, 0.1679, 0.1734, 0.1486, 0.1621, 0.1494, 0.1667, 0.1969, 0.2112, 0.2161, 0.183, 0.1403]}
{"unit_id": "quote:1947950504170197183", "ticker_idx": 0, "ts": "2025-07-23T09:22:46+00:00", "symbol": "AMEC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMEC is an extraordinarily high-margin business — with gross margins projected to stay in the 44% range", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish AMEC on durable ~44% gross margins; vouches for quality of the Chinese semicap name.", "resolved_tickers": ["688012.SS"], "resolution_method": "exchange_ticker", "resolution_reason": "AMEC = Advanced Micro-Fabrication Equipment, Shanghai STAR Market", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Even though it’s a Chinese company, AMEC is an extraordinarily high-margin business — with gross margins projected to stay in the 44% range. https://t.co/8nD2E5y66T", "reply_text": null, "parent_text": null, "quoted_text": "Goldman modeling on Chinese Semicap name AMEC - heavy growth in Etch https://t.co/MpWqwEPb2g", "parent_author": null, "ret_m1d": -0.05546311152614103, "ret_signed_m1d": -0.05546311152614103, "alpha_spy_m1d": -0.04702733809340198, "alpha_c_m1d": -0.05113670482981281, "bench_spy_m1d": -0.008435773432739047, "bench_c_m1d": -0.004326406696328222, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002569143923314199, "ret_signed_p1d": 0.002569143923314199, "alpha_spy_p1d": 0.002238091400487141, "alpha_c_p1d": -0.001582843621898089, "bench_spy_p1d": 0.00033105252282705777, "bench_c_p1d": 0.004151987545212288, "ret_p1w": 0.026598195911958333, "ret_signed_p1w": 0.026598195911958333, "alpha_spy_p1w": 0.026204058035910194, "alpha_c_p1w": -0.007071322413481074, "bench_spy_p1w": 0.00039413787604813955, "bench_c_p1w": 0.03366951832543941, "ret_p1m": 0.019646509370908394, "ret_signed_p1m": 0.019646509370908394, "alpha_spy_p1m": 0.01753373236433564, "alpha_c_p1m": 0.016680849756770977, "bench_spy_p1m": 0.0021127770065727525, "bench_c_p1m": 0.002965659614137417, "ret_p3m": 0.32401392791100947, "ret_signed_p3m": 0.32401392791100947, "alpha_spy_p3m": 0.26259723080970376, "alpha_c_p3m": 0.11254186872684246, "bench_spy_p3m": 0.061416697101305706, "bench_c_p3m": 0.211472059184167, "ret_p6m": 0.8689235520874266, "ret_signed_p6m": 0.8689235520874266, "alpha_spy_p6m": 0.7711636741585557, "alpha_c_p6m": 0.48152088715932617, "bench_spy_p6m": 0.09775987792887086, "bench_c_p6m": 0.3874026649281004, "price_path": [-0.0572, -0.064, -0.0569, -0.0532, -0.0532, -0.0532, -0.0532, -0.0501, -0.0478, -0.0441, -0.0862, -0.0776, -0.0799, -0.056, -0.0811, -0.0931, -0.1007, -0.1038, -0.1126, -0.1169, -0.1222, -0.1224, -0.1293, -0.1404, -0.1113, -0.1198, -0.1198, -0.1003, -0.0977, -0.0947, -0.0947, -0.1008, -0.1385, -0.1456, -0.1652, -0.136, -0.1291, -0.1402, -0.1357, -0.1304, -0.1369, -0.1016, -0.099, -0.0907, -0.0999, -0.0958, -0.0832, -0.0807, -0.1009, -0.1047, -0.09, -0.092, -0.0721, -0.0829, -0.0903, -0.0712, -0.0983, -0.0933, -0.1019, -0.1024, -0.0819, -0.0746, -0.0555, 0.0, 0.0026, 0.0069, 0.0078, 0.0397, 0.0266, -0.0045, -0.0302, -0.0137, -0.0167, -0.0123, -0.0115, -0.013, -0.0257, -0.0295, -0.0303, -0.0326, -0.0217, -0.0204, -0.0348, -0.0025, 0.0196, 0.0871, 0.0629, 0.078, 0.0429, 0.136, 0.0789, 0.1244, 0.0705, 0.068, -0.0157, 0.0076, 0.0715, 0.0296, 0.0357, 0.0669, 0.0655, 0.0836, 0.078, 0.1445, 0.2754, 0.2795, 0.2926, 0.4105, 0.4634, 0.4508, 0.4609, 0.5364, 0.5062, 0.5062, 0.5062, 0.5062, 0.5062, 0.5062, 0.5062, 0.6075, 0.476, 0.4866, 0.3778, 0.3803, 0.3823, 0.3194, 0.324, 0.3585, 0.344, 0.346, 0.4174, 0.4879, 0.4886, 0.4944, 0.5115, 0.4073, 0.3999, 0.4752, 0.4785, 0.5415, 0.5573, 0.594, 0.542, 0.5251, 0.529, 0.4811, 0.4569, 0.5107, 0.5037, 0.4284, 0.345, 0.3657, 0.3573, 0.3614, 0.3385, 0.3506, 0.3199, 0.3118, 0.3071, 0.3624, 0.3794, 0.3917, 0.3929, 0.3793, 0.3502, 0.4082, 0.3897, 0.3728, 0.4063, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.3738, 0.5683, 0.6607, 0.7749, 0.7424, 0.696, 0.7767, 0.743, 0.7751, 0.8689]}
{"unit_id": "orig:1943990459736510814", "ticker_idx": 0, "ts": "2025-07-12T11:06:57+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The demand for B30 is significant.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish NVDA on strong B30 demand from Chinese internet majors; hundreds of thousands of units ordered.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA B30 Industry Notes: \n\n• China's major internet company A estimates that the performance of NVIDIA's B30 is approximately 75% that of the H20. They placed an order for hundreds of thousands of units (USD 1 billion) in late June, with delivery expected in August.\n \n• Internet company B plans to increase its Q3 capital expenditure and intends to order 300,000 units of B30, with delivery scheduled for September.\n \n• Internet company C's detailed information remains confidential.\n\nThe demand for B30 is significant.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005180683499361294, "ret_signed_m1d": 0.005180683499361294, "alpha_spy_m1d": 0.007085166748686311, "alpha_c_m1d": -0.0024247053453889666, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.040409535932759866, "ret_signed_p1d": 0.040409535932759866, "alpha_spy_p1d": 0.044682702983386946, "alpha_c_p1d": 0.021203099274968107, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.04455419435283536, "ret_signed_p1w": 0.04455419435283536, "alpha_spy_p1w": 0.03821616236811698, "alpha_c_p1w": 0.0263641341996117, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.11635274151788177, "ret_signed_p1m": 0.11635274151788177, "alpha_spy_p1m": 0.0877358414456213, "alpha_c_p1m": 0.06455133674457092, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.17377251073301703, "ret_signed_p3m": 0.17377251073301703, "alpha_spy_p3m": 0.09661177882903127, "alpha_c_p3m": -0.036132036996605166, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.1413481255415343, "ret_signed_p6m": 0.1413481255415343, "alpha_spy_p6m": 0.02776489035331986, "alpha_c_p6m": -0.22142382598611499, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.3632, -0.3815, -0.3815, -0.4094, -0.3973, -0.374, -0.3514, -0.3234, -0.3373, -0.3356, -0.3362, -0.3198, -0.3022, -0.3063, -0.308, -0.2866, -0.2847, -0.2891, -0.2504, -0.2081, -0.1752, -0.1783, -0.1748, -0.1738, -0.181, -0.1967, -0.1905, -0.1998, -0.1998, -0.1742, -0.1784, -0.1517, -0.1764, -0.1627, -0.1393, -0.1351, -0.1468, -0.1363, -0.1307, -0.1226, -0.1295, -0.1162, -0.1347, -0.1181, -0.1216, -0.1133, -0.1133, -0.1232, -0.1213, -0.0986, -0.0595, -0.0552, -0.0385, -0.0371, -0.0656, -0.0416, -0.0288, -0.0288, -0.0355, -0.0248, -0.0073, 0.0002, 0.0052, 0.0, 0.0404, 0.0445, 0.0544, 0.0508, 0.0446, 0.018, 0.0409, 0.0589, 0.0575, 0.0773, 0.0697, 0.0926, 0.0841, 0.0588, 0.0971, 0.0865, 0.0936, 0.1018, 0.1135, 0.1096, 0.1164, 0.1068, 0.1094, 0.0998, 0.1093, 0.0705, 0.0691, 0.0665, 0.0848, 0.0959, 0.1079, 0.1068, 0.0981, 0.0616, 0.0616, 0.0409, 0.0399, 0.0463, 0.018, 0.0258, 0.0408, 0.0808, 0.0799, 0.0839, 0.0834, 0.0659, 0.038, 0.0742, 0.0769, 0.1192, 0.0876, 0.0787, 0.0831, 0.0861, 0.1084, 0.1373, 0.1413, 0.1513, 0.1436, 0.1309, 0.1279, 0.1527, 0.1738, 0.1164, 0.1479, 0.0973, 0.0961, 0.1082, 0.1168, 0.1132, 0.1042, 0.0989, 0.1103, 0.1353, 0.1672, 0.2253, 0.262, 0.2367, 0.2342, 0.261, 0.2111, 0.1899, 0.1464, 0.1468, 0.2133, 0.1774, 0.1813, 0.139, 0.1591, 0.1374, 0.1054, 0.1369, 0.1011, 0.0903, 0.1127, 0.0839, 0.0987, 0.0987, 0.0789, 0.0967, 0.1061, 0.0947, 0.1178, 0.1119, 0.131, 0.1275, 0.1203, 0.1029, 0.0669, 0.0746, 0.0833, 0.042, 0.0615, 0.1033, 0.1197, 0.1534, 0.1497, 0.1497, 0.1614, 0.1473, 0.1432, 0.1368, 0.1368, 0.1512, 0.1467, 0.1413]}
{"unit_id": "orig:1963883771910070654", "ticker_idx": 0, "ts": "2025-09-05T08:35:53+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I think the outlook is good for Micron", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Micron outlook; Samsung expected to underperform expectations.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "@efekarhanli I think the outlook is good for Micron.\n\nSamsung doesn’t seem like it will do as well as expected.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.054502454289519275, "ret_signed_m1d": -0.054502454289519275, "alpha_spy_m1d": -0.057407014816634905, "alpha_c_m1d": -0.04284113239538745, "bench_spy_m1d": 0.0029045605271156294, "bench_c_m1d": -0.011661321894131826, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0006853562334441232, "ret_signed_p1d": 0.0006853562334441232, "alpha_spy_p1d": -0.0017711856752693489, "alpha_c_p1d": -0.010362003116995533, "bench_spy_p1d": 0.002456541908713472, "bench_c_p1d": 0.011047359350439656, "ret_p1w": 0.19684868737182293, "ret_signed_p1w": 0.19684868737182293, "alpha_spy_p1w": 0.18113579664652169, "alpha_c_p1w": 0.15920543509887253, "bench_spy_p1w": 0.015712890725301243, "bench_c_p1w": 0.0376432522729504, "ret_p1m": 0.45451483071573073, "ret_signed_p1m": 0.45451483071573073, "alpha_spy_p1m": 0.41398583589414506, "alpha_c_p1m": 0.2836883642747543, "bench_spy_p1m": 0.04052899482158567, "bench_c_p1m": 0.17082646644097643, "ret_p3m": 0.7835630591129057, "ret_signed_p3m": 0.7835630591129057, "alpha_spy_p3m": 0.724008536645024, "alpha_c_p3m": 0.5411667430551756, "bench_spy_p3m": 0.05955452246788173, "bench_c_p3m": 0.24239631605773004, "ret_p6m": 2.1445179236590857, "ret_signed_p6m": 2.1445179236590857, "alpha_spy_p6m": 2.0779633361424383, "alpha_c_p6m": 1.7545338635128849, "bench_spy_p6m": 0.06655458751664733, "bench_c_p6m": 0.3899840601462008, "price_path": [-0.132, -0.1176, -0.1165, -0.1209, -0.0886, -0.0848, -0.0736, -0.0736, -0.06, -0.0716, -0.0273, -0.0323, -0.0418, -0.0512, -0.0627, -0.0806, -0.0742, -0.07, -0.07, -0.0872, -0.0529, -0.0695, -0.0629, -0.0521, -0.0971, -0.0857, -0.1137, -0.1379, -0.1293, -0.1381, -0.1686, -0.164, -0.1495, -0.1531, -0.1532, -0.1478, -0.1266, -0.1692, -0.2016, -0.1796, -0.1698, -0.172, -0.1484, -0.095, -0.0582, -0.0276, -0.054, -0.0463, -0.0799, -0.0595, -0.0709, -0.1078, -0.1186, -0.1042, -0.1138, -0.1132, -0.1037, -0.0713, -0.0941, -0.0941, -0.0981, -0.0963, -0.0545, 0.0, 0.0007, 0.0295, 0.0657, 0.1462, 0.1968, 0.201, 0.209, 0.2179, 0.2856, 0.2387, 0.2531, 0.2667, 0.231, 0.1938, 0.1972, 0.2476, 0.2737, 0.3865, 0.3987, 0.4307, 0.4545, 0.4144, 0.497, 0.4649, 0.3832, 0.4683, 0.4248, 0.462, 0.5426, 0.5415, 0.5749, 0.5408, 0.5117, 0.5745, 0.6682, 0.6765, 0.6903, 0.7262, 0.7063, 0.7044, 0.7877, 0.6607, 0.809, 0.8153, 0.8122, 0.9293, 0.8365, 0.8654, 0.8048, 0.8801, 0.8429, 0.7405, 0.7208, 0.5338, 0.5795, 0.7056, 0.7102, 0.7539, 0.7539, 0.8012, 0.8315, 0.8242, 0.7836, 0.7264, 0.8069, 0.8808, 0.9226, 1.0086, 0.9687, 0.8367, 0.809, 0.771, 0.7178, 0.8932, 1.0255, 1.1067, 1.1043, 1.1836, 1.1836, 1.1692, 1.2431, 1.2298, 1.1748, 1.1748, 1.4035, 1.3786, 1.6169, 1.5873, 1.4919, 1.6296, 1.6355, 1.5765, 1.5401, 1.5651, 1.7641, 1.7641, 1.7813, 1.965, 2.0295, 2.0453, 1.9648, 2.126, 2.3168, 2.3207, 2.1614, 2.336, 2.1961, 1.891, 1.9176, 2.0075, 1.9222, 1.8441, 2.1268, 2.1544, 2.1368, 2.1368, 2.0463, 2.2076, 2.1802, 2.2626, 2.2078, 2.1852, 2.269, 2.1665, 2.1422, 2.1445]}
{"unit_id": "orig:1963883771910070654", "ticker_idx": 1, "ts": "2025-09-05T08:35:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung doesn't seem like it will do as well as expected", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Micron outlook; Samsung expected to underperform expectations.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "@efekarhanli I think the outlook is good for Micron.\n\nSamsung doesn’t seem like it will do as well as expected.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008633112828845269, "ret_signed_m1d": -0.008633112828845269, "alpha_spy_m1d": -0.005728552301729639, "alpha_c_m1d": -0.009469958177167115, "bench_spy_m1d": 0.0029045605271156294, "bench_c_m1d": -0.0008368453483218463, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008633112828845269, "ret_signed_p1d": -0.008633112828845269, "alpha_spy_p1d": -0.006176570920131796, "alpha_c_p1d": -0.0011385734807942693, "bench_spy_p1d": 0.002456541908713472, "bench_c_p1d": 0.007494539348050999, "ret_p1w": 0.08489197296279882, "ret_signed_p1w": -0.08489197296279882, "alpha_spy_p1w": -0.06917908223749758, "alpha_c_p1w": -0.054381214038838, "bench_spy_p1w": 0.015712890725301243, "bench_c_p1w": 0.030510758923960823, "ret_p1m": 0.2971283860028786, "ret_signed_p1m": -0.2971283860028786, "alpha_spy_p1m": -0.25659939118129294, "alpha_c_p1m": -0.20160363769178158, "bench_spy_p1m": 0.04052899482158567, "bench_c_p1m": 0.09552474831109703, "ret_p3m": 0.5103055140978214, "ret_signed_p3m": -0.5103055140978214, "alpha_spy_p3m": -0.4507509916299397, "alpha_c_p3m": -0.4056767649921933, "bench_spy_p3m": 0.05955452246788173, "bench_c_p3m": 0.10462874910562814, "ret_p6m": 2.144216711460727, "ret_signed_p6m": -2.144216711460727, "alpha_spy_p6m": -2.0776621239440796, "alpha_c_p6m": -2.079533893216449, "bench_spy_p6m": 0.06655458751664733, "bench_c_p6m": 0.0646828182442778, "price_path": [-0.1534, -0.1434, -0.1491, -0.1663, -0.182, -0.1691, -0.1448, -0.1534, -0.1491, -0.1706, -0.1348, -0.1234, -0.1391, -0.1252, -0.1396, -0.1338, -0.1252, -0.082, -0.0892, -0.1122, -0.1165, -0.1309, -0.1223, -0.0993, -0.1007, -0.0835, -0.0691, -0.0403, -0.0345, -0.0245, -0.0504, -0.0446, -0.0504, -0.0518, 0.0129, 0.0158, 0.0446, 0.0273, -0.0086, 0.0029, 0.0058, -0.0101, 0.0144, 0.0331, 0.0216, 0.023, 0.0345, 0.0302, 0.0302, 0.0072, 0.0072, 0.0144, 0.0158, 0.0273, 0.0288, 0.0115, 0.0158, 0.0014, 0.0029, -0.0273, -0.0058, 0.0043, 0.0086, 0.0, 0.0086, 0.0288, 0.0446, 0.0561, 0.0849, 0.1007, 0.1424, 0.1252, 0.1554, 0.1554, 0.2014, 0.2187, 0.2288, 0.2388, 0.1986, 0.2169, 0.2126, 0.2429, 0.2971, 0.2971, 0.2971, 0.2971, 0.2971, 0.2971, 0.3643, 0.3484, 0.3239, 0.373, 0.412, 0.4149, 0.4178, 0.4091, 0.425, 0.3947, 0.4279, 0.4742, 0.438, 0.4525, 0.5045, 0.5537, 0.6057, 0.5161, 0.4539, 0.4337, 0.4149, 0.4539, 0.4959, 0.4901, 0.4857, 0.4048, 0.4539, 0.4135, 0.3947, 0.4539, 0.3701, 0.3976, 0.4352, 0.4857, 0.4959, 0.4525, 0.4568, 0.4944, 0.5103, 0.519, 0.5667, 0.5826, 0.5667, 0.5609, 0.5508, 0.5739, 0.5146, 0.4857, 0.5594, 0.5551, 0.5363, 0.597, 0.6115, 0.6057, 0.6057, 0.691, 0.7355, 0.7413, 0.7413, 0.7413, 0.8662, 1.0056, 1.0172, 1.0477, 1.0158, 1.0187, 1.0158, 0.9984, 1.0376, 1.0899, 1.1625, 1.1683, 1.1087, 1.1712, 1.2118, 1.2089, 1.2089, 1.3164, 1.3585, 1.3338, 1.3309, 1.1843, 1.4326, 1.4558, 1.3135, 1.3033, 1.4166, 1.4079, 1.4369, 1.5938, 1.6316, 1.6316, 1.6316, 1.6316, 1.7594, 1.7608, 1.8029, 1.9046, 1.9554, 2.166, 2.1442, 2.1442]}
{"unit_id": "orig:1952203720030400833", "ticker_idx": 0, "ts": "2025-08-04T03:03:31+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Apple's profit margins will decline due to tariffs... it appears they will raise prices by about $50 to defend profit margins", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish AAPL: tariff pressure forces ~$50 iPhone 17 price hike, defending margins despite rising BoM costs.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "A Brief Note on iPhone 17 Series Price Increase\n\nA few days ago, Jefferies analysts revealed in their note that the iPhone 17 series could see a $50 price increase.\n\nAdditionally, WSJ reported last May that Apple was considering raising prices for the iPhone 17 series.\n\nApple's iPhone price increase speculation has been one of the consistently raised issues. Since the iPhone 14 series.\n\nHowever, it's now becoming a realistic necessity rather than mere speculation. As confirmed with the iPhone 16 series, Apple's profit margins decreased compared to the previous generation due to rising BoM costs.\n\nTherefore, many industry analysts/stakeholders expected a price increase for the iPhone 16 series. Because Apple is not a company that tolerates profit margin decline. However, contrary to everyone's expectations, Apple did not raise prices.\n\nBut now, I think Apple really has no choice but to raise prices.\n\nBecause of tariffs.\n\nAs seen in the headlines that dominated the media a few days ago, the US now imports the largest share of iPhones from India.\n\nAnd DJT imposed a 25% tariff on India.\n\nBased on media reports and industry supply chain news, the iPhone 17's BoM is expected to actually decrease compared to the previous generation. This is the result of Apple's tremendous pressure on component suppliers for cost reduction.\n\nThis news isn't exactly a secret. That's why the stock performance of companies in Apple's value chain was quite poor in the first half of this year.\n\nHowever, despite the decrease in iPhone 17 series BoM, Apple's profit margins will decline due to tariffs.\n\nAccording to my modeling, the weak dollar effect can reduce tariff burden by 5%, and component cost reduction can alleviate tariff burden by about 7-10%.\n\nStill, a 10% tariff remains burdensome for Apple. Therefore, it appears they will raise prices by about $50 to defend profit margins.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004770046296976416, "ret_signed_m1d": -0.004770046296976416, "alpha_spy_m1d": 0.010202192642118302, "alpha_c_m1d": 0.015131718238139968, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.019901764535116384, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0021146466969286237, "ret_signed_p1d": -0.0021146466969286237, "alpha_spy_p1d": 0.002955319875502682, "alpha_c_p1d": 0.006158791117942708, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.008273437814871332, "ret_p1w": 0.1184550196868559, "ret_signed_p1w": 0.1184550196868559, "alpha_spy_p1w": 0.11092925758032135, "alpha_c_p1w": 0.11147807038343527, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.006976949303420632, "ret_p1m": 0.1309599870725553, "ret_signed_p1m": 0.1309599870725553, "alpha_spy_p1m": 0.116542257441971, "alpha_c_p1m": 0.14045351193136157, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": -0.00949352485880628, "ret_p3m": 0.33615929082870877, "ret_signed_p3m": 0.33615929082870877, "alpha_spy_p3m": 0.25607814093672876, "alpha_c_p3m": 0.1894284033158511, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.14673088751285768, "ret_p6m": 0.27275002382461766, "ret_signed_p6m": 0.27275002382461766, "alpha_spy_p6m": 0.16452396697775895, "alpha_c_p6m": 0.14068167605308135, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.1320683477715363, "price_path": [-0.0362, -0.0301, -0.025, 0.0366, 0.0471, 0.0442, 0.0398, 0.0389, 0.0267, 0.0173, -0.0062, -0.0098, -0.0397, -0.0397, -0.0154, -0.0144, -0.0167, -0.0123, -0.0081, -0.0004, -0.0026, -0.0134, 0.0028, -0.0093, -0.0033, -0.0225, -0.0204, -0.0339, -0.0242, -0.0379, -0.0333, -0.0333, -0.0116, -0.0091, -0.015, -0.0088, -0.0116, -0.0112, 0.0089, 0.022, 0.0447, 0.0502, 0.0502, 0.0325, 0.0328, 0.0383, 0.0446, 0.0384, 0.0259, 0.0283, 0.0335, 0.0328, 0.0385, 0.0449, 0.0543, 0.0531, 0.0512, 0.0518, 0.0526, 0.0389, 0.028, 0.0208, -0.0048, 0.0, -0.0021, 0.0487, 0.082, 0.1279, 0.1185, 0.1306, 0.1487, 0.146, 0.1402, 0.1367, 0.1351, 0.1127, 0.1072, 0.1213, 0.1184, 0.1289, 0.1348, 0.1449, 0.1429, 0.1429, 0.131, 0.174, 0.1805, 0.18, 0.1711, 0.1538, 0.1165, 0.1325, 0.1524, 0.1653, 0.1725, 0.1766, 0.1711, 0.2086, 0.2607, 0.2526, 0.2422, 0.2646, 0.2577, 0.2526, 0.2536, 0.2576, 0.2659, 0.2703, 0.2637, 0.2627, 0.2705, 0.2507, 0.2075, 0.2193, 0.2198, 0.2276, 0.2182, 0.2421, 0.2911, 0.2937, 0.2724, 0.278, 0.2939, 0.3234, 0.3243, 0.3278, 0.3362, 0.3311, 0.3246, 0.3295, 0.33, 0.3281, 0.3217, 0.3277, 0.3564, 0.3477, 0.3451, 0.3424, 0.318, 0.3179, 0.3235, 0.3121, 0.3379, 0.3597, 0.3649, 0.3678, 0.3678, 0.3742, 0.3951, 0.4103, 0.4003, 0.3833, 0.3738, 0.3694, 0.3659, 0.3738, 0.3701, 0.3714, 0.3508, 0.3533, 0.3396, 0.3413, 0.3486, 0.3353, 0.3422, 0.3493, 0.3493, 0.3473, 0.3491, 0.3457, 0.3397, 0.3397, 0.3355, 0.3171, 0.2929, 0.2829, 0.2765, 0.2782, 0.2825, 0.2864, 0.2811, 0.2725, 0.2592, 0.2592, 0.2157, 0.2204, 0.2239, 0.2223, 0.2587, 0.2728]}
{"unit_id": "quote:1956636229107364305", "ticker_idx": 0, "ts": "2025-08-16T08:36:44+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "The profit margin of the GB200 is tremendous.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Endorses Morgan Stanley's finding that GB200 margins are huge; implied bullish on Nvidia.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "The profit margin of the GB200 is tremendous. https://t.co/9nCOtTcDM6", "reply_text": null, "parent_text": null, "quoted_text": "Morgan Stanley: AI inference is an incredibly profitable business. In its newly released report, the firm used a detailed financial model to calculate, for the first time from an economic perspective, the profitability of global AI compute competition. The conclusion is that https://t.co/j8GpKzYYRD", "parent_author": null, "ret_m1d": -0.008570860315888718, "ret_signed_m1d": -0.008570860315888718, "alpha_spy_m1d": -0.008788539126832307, "alpha_c_m1d": -0.004494894694822116, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.004075965621066602, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03499800104375028, "ret_signed_p1d": -0.03499800104375028, "alpha_spy_p1d": -0.02957287097707262, "alpha_c_p1d": -0.014820187223993253, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.02017781381975703, "ret_p1w": -0.01208714510141684, "ret_signed_p1w": -0.01208714510141684, "alpha_spy_p1w": -0.01079685992973567, "alpha_c_p1w": -0.0021835984152863697, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.009903546686130471, "ret_p1m": -0.039119339029687894, "ret_signed_p1m": -0.039119339029687894, "alpha_spy_p1m": -0.06507928344935043, "alpha_c_p1m": -0.07398428232234966, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.034864943292661765, "ret_p3m": 0.026704848072617082, "ret_signed_p3m": 0.026704848072617082, "alpha_spy_p3m": -0.02086730121045899, "alpha_c_p3m": -0.13485285428145266, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.16155770235406974, "ret_p6m": 0.03599328672679625, "ret_signed_p6m": 0.03599328672679625, "alpha_spy_p6m": -0.046067414947216, "alpha_c_p6m": -0.3317203549813268, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.36771364170812304, "price_path": [-0.2759, -0.2703, -0.2787, -0.2787, -0.2556, -0.2594, -0.2353, -0.2576, -0.2453, -0.2242, -0.2203, -0.2309, -0.2214, -0.2164, -0.2091, -0.2153, -0.2033, -0.22, -0.205, -0.2082, -0.2007, -0.2007, -0.2097, -0.2079, -0.1874, -0.1522, -0.1483, -0.1333, -0.132, -0.1577, -0.136, -0.1246, -0.1246, -0.1306, -0.1209, -0.1051, -0.0984, -0.0939, -0.0986, -0.0621, -0.0585, -0.0495, -0.0527, -0.0584, -0.0823, -0.0617, -0.0454, -0.0468, -0.0289, -0.0357, -0.0151, -0.0227, -0.0455, -0.011, -0.0206, -0.0142, -0.0068, 0.0038, 0.0003, 0.0063, -0.0023, 0.0001, -0.0086, 0.0, -0.035, -0.0363, -0.0386, -0.0221, -0.0121, -0.0013, -0.0023, -0.0101, -0.043, -0.043, -0.0617, -0.0626, -0.0569, -0.0824, -0.0753, -0.0618, -0.0257, -0.0265, -0.023, -0.0234, -0.0391, -0.0643, -0.0316, -0.0293, 0.0088, -0.0196, -0.0276, -0.0237, -0.0209, -0.0008, 0.0252, 0.0288, 0.0379, 0.0309, 0.0195, 0.0167, 0.0391, 0.0581, 0.0064, 0.0347, -0.0108, -0.0119, -0.001, 0.0067, 0.0035, -0.0046, -0.0094, 0.0009, 0.0234, 0.0521, 0.1046, 0.1376, 0.1148, 0.1126, 0.1367, 0.0917, 0.0726, 0.0334, 0.0338, 0.0937, 0.0613, 0.0648, 0.0267, 0.0449, 0.0253, -0.0035, 0.0248, -0.0075, -0.0171, 0.003, -0.023, -0.0096, -0.0096, -0.0275, -0.0114, -0.003, -0.0132, 0.0076, 0.0023, 0.0196, 0.0164, 0.0098, -0.0058, -0.0383, -0.0313, -0.0235, -0.0607, -0.0431, -0.0055, 0.0093, 0.0397, 0.0364, 0.0364, 0.0469, 0.0342, 0.0305, 0.0248, 0.0248, 0.0377, 0.0337, 0.0289, 0.0391, 0.0168, 0.0158, 0.0162, 0.021, 0.0063, 0.0278, 0.0233, 0.0233, -0.0215, 0.0073, 0.0157, 0.0312, 0.0246, 0.0359, 0.0524, 0.0578, 0.0502, 0.0199, -0.0091, -0.0429, -0.0556, 0.0188, 0.0442, 0.036]}
{"unit_id": "orig:1945589675885994252", "ticker_idx": 0, "ts": "2025-07-16T21:01:40+00:00", "symbol": "GDS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM states that Nvidia's resumption of H20 shipments to China significantly alleviates the AI server supply bottleneck in China, helping to drive an upward revision of GDS Holdings' China business outlook. Simultaneously, its international subsidiary DayOne continues its strong growth momentum, opening up room for valuation re-rating.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM bullish on GDS: H20 resumption drives China AI data center rebound, DayOne IPO and C-REIT support valuation re-rating.", "resolved_tickers": ["GDS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "original", "lang": "en", "tweet_text": "GDS Holdings: H20 Resumption Drives China AI Data Center Rebound, DayOne Shows Strong Growth\n\nJPM (G. Hariharan, 25/07/16)\n\nJPM states that Nvidia's resumption of H20 shipments to China significantly alleviates the AI server supply bottleneck in China, helping to drive an upward revision of GDS Holdings' China business outlook. Simultaneously, its international subsidiary DayOne continues its strong growth momentum, opening up room for valuation re-rating.\n\nH20 Resumption Alleviates Chip Bottleneck, Drives China AI Inference Demand & Valuation Uplift\n\nNvidia's resumption of H20 shipments to China is expected to drive a rebound in domestic AI inference server demand. GDS Holdings stands to benefit from the initial build-out phase of inference compute power in tier-one cities and its collaborations with clients like Alibaba. JPM expects an upward revision of the FY26/27 China AI data center pipeline, noting GDS's lower exposure to training compute power.\n\nDayOne International Business Expands Steadily, Large Social Media Clients Drive Mid-to-Long-Term Revenue Potential\n\nDayOne, a subsidiary of GDS, has already signed up for over 500MW of client capacity and secured 1.2GW of underlying resources. It is expected to launch its IPO in the second half of 2026, at which point GDS will hold approximately a 30% stake. If valued at 20x FY27E EV/EBITDA, GDS could achieve significant value realization.\n\nC-REIT Mechanism Aids Capital Recycling and Supports a New AI Capex Cycle\n\nGDS raised capital through its C-REIT listing (oversubscribed by 166 times) and plans to continuously inject mature assets. This is expected to stabilize dividend payout ratios, expand the Free Cash Flow (FCF) pool, and support a new AI-driven capital expenditure cycle domestically, alleviating liquidity pressure before 2026.\n\nCore Pricing Stabilizes with Upside Potential, Alibaba Capex Expected to Drive Revenue Growth from 2H25\n\nAfter two years of消化, pricing in the China IDC market has largely stabilized. High-power density/low-PUE demand related to AI inference will enhance pricing power. Alibaba's Capex plans for the next three years are clear, and with chip availability, orders are expected to materialize, potentially marking a growth inflection point for GDS's core business.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010493119684929164, "ret_signed_m1d": -0.010493119684929164, "alpha_spy_m1d": -0.007161068251173797, "alpha_c_m1d": -0.007400930081167156, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.003092189603762008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004459553350173895, "ret_signed_p1d": -0.004459553350173895, "alpha_spy_p1d": -0.010579243327691445, "alpha_c_p1d": -0.013542949140020522, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.009083395789846627, "ret_p1w": -0.03672606900339226, "ret_signed_p1w": -0.03672606900339226, "alpha_spy_p1w": -0.05273019080357366, "alpha_c_p1w": -0.0428333448160928, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.006107275812700541, "ret_p1m": -0.164742889697305, "ret_signed_p1m": -0.164742889697305, "alpha_spy_p1m": -0.19795237180987557, "alpha_c_p1m": -0.19960831051839578, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.03486542082109079, "ret_p3m": -0.06243434732699493, "ret_signed_p3m": -0.06243434732699493, "alpha_spy_p3m": -0.1275688481750209, "alpha_c_p3m": -0.16612763547055354, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1036932881435586, "ret_p6m": 0.09522563404583373, "ret_signed_p6m": 0.09522563404583373, "alpha_spy_p6m": -0.015704433401805495, "alpha_c_p6m": -0.02297185676571556, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11819749081154929, "price_path": [-0.4368, -0.4436, -0.4124, -0.3859, -0.3948, -0.3927, -0.3867, -0.3746, -0.3387, -0.3242, -0.2629, -0.271, -0.2639, -0.313, -0.3061, -0.3324, -0.2358, -0.2828, -0.2841, -0.3064, -0.2857, -0.2896, -0.2765, -0.2222, -0.2613, -0.2524, -0.2524, -0.288, -0.3502, -0.3418, -0.3785, -0.3604, -0.3704, -0.3578, -0.3269, -0.3368, -0.3161, -0.3132, -0.2946, -0.2991, -0.3263, -0.2623, -0.2859, -0.2852, -0.2852, -0.2904, -0.2775, -0.2403, -0.2463, -0.2211, -0.2086, -0.1981, -0.2269, -0.1911, -0.1086, -0.1086, -0.1522, -0.1508, -0.1545, -0.1472, -0.1104, -0.0818, -0.0105, 0.0, -0.0045, -0.0142, -0.0624, -0.0805, -0.0367, -0.0551, -0.0352, -0.042, -0.0464, -0.0748, -0.0538, -0.1049, -0.0837, -0.0674, -0.0643, -0.0223, -0.0459, -0.0328, -0.0475, -0.1102, -0.1647, -0.176, -0.1545, -0.1713, -0.1107, -0.1605, -0.1209, -0.1288, -0.1133, -0.1167, -0.0797, -0.0934, -0.0934, -0.1348, -0.1039, -0.1406, -0.1299, -0.1209, -0.1013, -0.1212, 0.01, 0.0024, -0.0241, -0.021, 0.053, 0.0231, 0.022, 0.043, -0.0129, 0.0669, 0.1036, 0.0514, 0.0485, 0.0152, 0.0554, 0.0847, 0.0396, 0.0404, -0.0163, -0.0073, 0.0079, -0.1262, -0.0624, -0.1162, -0.1118, -0.127, -0.1414, -0.0978, -0.1047, -0.1254, -0.1013, -0.0839, -0.0582, -0.059, -0.0115, -0.0679, -0.0635, -0.0889, -0.1338, -0.1036, -0.1104, -0.1385, -0.1233, -0.1257, -0.1592, -0.1973, -0.2219, -0.2301, -0.2387, -0.2196, -0.2054, -0.1889, -0.1209, -0.1062, -0.1055, -0.1055, -0.1089, -0.0918, -0.1133, -0.1149, -0.1139, -0.0763, -0.0651, -0.0551, -0.0496, -0.0517, -0.0488, -0.0666, -0.0887, -0.1162, -0.096, -0.0575, -0.0543, -0.0837, -0.0871, -0.0871, -0.0714, -0.091, -0.0923, -0.0845, -0.0845, 0.0058, -0.0155, -0.0118, 0.0084, 0.0952]}
{"unit_id": "thread:1965599893827022877", "ticker_idx": 0, "ts": "2025-09-10T02:15:08+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Negative for ASML", "tweet_type": "analysis", "timeline": "unspecified", "summary": "TSMC in-house EUV pellicle development reduces ASML supply chain dependence and may delay High-NA EUV adoption — bearish for ASML.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "TSMC Begins In-House EUV Pellicle Development, May Delay High-NA EUV Adoption…Negative for ASML\n\n- TSMC’s 6-inch wafer fab (Fab 2) in Hsinchu Science Park will cease production on related lines.\n\n- According to semiconductor equipment industry sources, the 6-inch fab will be converted into a panel-level packaging (CoPoS) facility, while the 8-inch fab will officially begin in-house EUV pellicle production, reducing reliance on ASML and other suppliers.\n\n- This means TSMC can leverage its own R&D and manufacturing capabilities to significantly improve yields in EUV-based advanced processes, achieving minimum cost and maximum efficiency.\n\nCurrently, EUV equipment procurement costs are extremely high, prompting TSMC to delay High-NA EUV adoption while exploring alternatives to improve EUV process yields and cut costs.\n\n- One such alternative is the EUV pellicle, which TSMC has been developing for years. Industry sources say Fab 3 in Hsinchu will soon begin full-scale in-house EUV pellicle production, reducing dependence on ASML’s supply chain and enabling TSMC to enhance EUV yields while building processes and products with lower costs and higher efficiency.\n\n- With the advent of the EUV era, conventional organic pellicles have become unsuitable as they fail to meet both transmittance and stability requirements.\n\n- Currently, most EUV processes use pellicle-free photomasks, which necessitates frequent pattern inspections.\n\n- When defects are detected, masks must be repaired or remade, leading to higher production costs and slower throughput. For this reason, ASML and other semiconductor companies have been striving to develop EUV pellicles, but the high technical difficulty has so far prevented mass production.\n\n- TSMC has determined that EUV pellicles are critically important for improving production efficiency and reducing costs in sub-7nm processes, and has recently decided to accelerate its in-house development plans.\n\n$ASML\n\n---\n\nhttps://t.co/LJa38NBtQT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00441050973628232, "ret_signed_m1d": -0.00441050973628232, "alpha_spy_m1d": -0.0072930310661663, "alpha_c_m1d": -0.014449362820823985, "bench_spy_m1d": -0.0028825213298839802, "bench_c_m1d": -0.010038853084541666, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008526937378555921, "ret_signed_p1d": -0.008526937378555921, "alpha_spy_p1d": -0.0002166726467702773, "alpha_c_p1d": 0.0017779170233740427, "bench_spy_p1d": 0.008310264731785644, "bench_c_p1d": 0.010304854401929964, "ret_p1w": 0.08350479101227393, "ret_signed_p1w": -0.08350479101227393, "alpha_spy_p1w": -0.072818048213245, "alpha_c_p1w": -0.06881203423628723, "bench_spy_p1w": 0.010686742799028925, "bench_c_p1w": 0.014692756775986693, "ret_p1m": 0.24933840549771147, "ret_signed_p1m": -0.24933840549771147, "alpha_spy_p1m": -0.21743040533311997, "alpha_c_p1m": -0.10181319973592351, "bench_spy_p1m": 0.0319080001645915, "bench_c_p1m": 0.14752520576178796, "ret_p3m": 0.418553422808863, "ret_signed_p3m": -0.418553422808863, "alpha_spy_p3m": -0.3674727816722456, "alpha_c_p3m": -0.19344283739125068, "bench_spy_p3m": 0.051080641136617366, "bench_c_p3m": 0.2251105854176123, "ret_p6m": 0.7490265521851383, "ret_signed_p6m": -0.7490265521851383, "alpha_spy_p6m": -0.6984176015048205, "alpha_c_p6m": -0.4307397960488166, "bench_spy_p6m": 0.05060895068031779, "bench_c_p6m": 0.3182867561363216, "price_path": [-0.0255, -0.0175, -0.0235, -0.0249, -0.0424, -0.0463, -0.0195, 0.0138, 0.0183, -0.0068, 0.0007, -0.0065, -0.0202, -0.0065, -0.0072, -0.0331, -0.0153, -0.0046, -0.0032, 0.0117, 0.0028, 0.0078, 0.0353, -0.0824, -0.0466, -0.0711, -0.0849, -0.1167, -0.1106, -0.0969, -0.1131, -0.0763, -0.0841, -0.0775, -0.0986, -0.1253, -0.1144, -0.1222, -0.1283, -0.1022, -0.0931, -0.0782, -0.0681, -0.0582, -0.0548, -0.064, -0.061, -0.0557, -0.0617, -0.0672, -0.05, -0.0493, -0.0435, -0.0294, -0.0381, -0.0641, -0.0648, -0.0926, -0.0807, -0.0479, -0.0294, -0.0053, 0.0044, 0.0, 0.0085, 0.0141, 0.0723, 0.0775, 0.0835, 0.1673, 0.1669, 0.192, 0.2052, 0.1879, 0.1933, 0.1949, 0.2136, 0.2174, 0.237, 0.2902, 0.2939, 0.3192, 0.2861, 0.2523, 0.2493, 0.1966, 0.2408, 0.2446, 0.2834, 0.2899, 0.2874, 0.3217, 0.3081, 0.2792, 0.3051, 0.3149, 0.335, 0.3364, 0.354, 0.3807, 0.3521, 0.3645, 0.352, 0.3352, 0.3186, 0.2825, 0.3059, 0.3062, 0.3134, 0.2982, 0.2873, 0.2907, 0.2748, 0.3057, 0.3105, 0.2283, 0.2648, 0.267, 0.3392, 0.3214, 0.3305, 0.3648, 0.3831, 0.4188, 0.4099, 0.4015, 0.4186, 0.4034, 0.3932, 0.3857, 0.3627, 0.3711, 0.3381, 0.2872, 0.3147, 0.3278, 0.3196, 0.3302, 0.324, 0.324, 0.324, 0.3355, 0.3526, 0.357, 0.357, 0.4526, 0.5511, 0.5626, 0.549, 0.4919, 0.5929, 0.6, 0.6239, 0.5968, 0.6928, 0.719, 0.6501, 0.6789, 0.7007, 0.7325, 0.7352, 0.7349, 0.7932, 0.7591, 0.7555, 0.7903, 0.8038, 0.7532, 0.6798, 0.6931, 0.7582, 0.7768, 0.7593, 0.7812, 0.7399, 0.7555, 0.7623, 0.7685, 0.8357, 0.826, 0.8517, 0.8422, 0.8632, 0.9, 0.8175, 0.8189, 0.785, 0.7133, 0.7694, 0.749]}
{"unit_id": "thread:1945428075006812261", "ticker_idx": 0, "ts": "2025-07-16T10:19:32+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TrendForce expects that the US government's lifting of export restrictions on NVIDIA's H20 GPU to China will stimulate a recovery in demand from local Chinese AI companies and CSPs", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares TrendForce bullish thesis: H20 export approval revives NVIDIA demand in China and drives HBM consumption higher.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶️ H20 Export Approval Reignites Demand Momentum, China’s Procurement Ratio for Foreign AI Chips Expected to Rise to 49%, Says TrendForce\n\n- TrendForce expects that the US government's lifting of export restrictions on NVIDIA's H20 GPU to China will stimulate a recovery in demand from local Chinese AI companies and CSPs (Cloud Service Providers).\n\n- The H20 is anticipated to once again establish itself as a primary high-performance AI chip in the market and is expected to contribute to an increase in HBM demand.\n\n- Following the strengthening of US AI chip export regulations in April, the annual shipment forecast for the H20, which was developed specifically for the Chinese market, was drastically lowered.\n\n- However, considering recent trends and the likelihood that NVIDIA will strive to meet its initial shipment targets, TrendForce has revised China's procurement ratio for foreign AI chips (primarily NVIDIA and AMD) upwards from the previously estimated 42% to 49%.\n\n- While Chinese companies like Huawei are developing their own AI chips, their ecosystems are not yet mature enough to fully replace the advantages of NVIDIA's system integration, computational efficiency, and supply chain scale.\n\n- The resumption of H20 shipments is expected to significantly restore potential demand for AI infrastructure construction, especially among major CSPs in China who have been prioritizing H20 for their domestic data center builds.\n\n- Furthermore, NVIDIA plans to launch a custom version of the RTX PRO 6000 targeting the Chinese market to support a wider range of edge AI inference applications.\n\n- H20 chips to be shipped in 2024 will primarily use HBM3 8hi, with an upgrade to HBM3e 8hi and an increase in total capacity expected in early 2025.\n\n- As Chinese domestic ASIC manufacturers face HBM export restrictions, many of their products still rely on previously supplied HBM2e.\n\n- Therefore, the H20 is expected to be preferred, increasing its proportion of overall HBM consumption.\n\n- Additionally, with China's AI market still vulnerable to geopolitical uncertainties and persistent risks for NVIDIA, an aggressive inventory build is anticipated if H20 export restrictions are indeed lifted.\n\n---\n\nhttps://t.co/8ZPQYl48dY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003909684872381747, "ret_signed_m1d": -0.003909684872381747, "alpha_spy_m1d": -0.0005776334386263793, "alpha_c_m1d": -0.00916412460569449, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009511589505708296, "ret_signed_p1d": 0.009511589505708296, "alpha_spy_p1d": 0.003391899528190745, "alpha_c_p1d": 0.0011114490994763582, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0034428594862712103, "ret_signed_p1w": -0.0034428594862712103, "alpha_spy_p1w": -0.019446981286452614, "alpha_c_p1w": 0.005786934868547422, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.0621461963170844, "ret_signed_p1m": 0.0621461963170844, "alpha_spy_p1m": 0.028936714204513825, "alpha_c_p1m": 0.01845147637661637, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.09897081121285733, "ret_signed_p3m": 0.09897081121285733, "alpha_spy_p3m": 0.03383631036483137, "alpha_c_p3m": -0.07549684249508726, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.07988992467564904, "ret_signed_p6m": 0.07988992467564904, "alpha_spy_p6m": -0.031040142771990187, "alpha_c_p6m": -0.23426679716688548, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4078, -0.4345, -0.423, -0.4007, -0.379, -0.3523, -0.3656, -0.3639, -0.3645, -0.3488, -0.3319, -0.3359, -0.3375, -0.317, -0.3152, -0.3194, -0.2823, -0.2419, -0.2103, -0.2133, -0.21, -0.209, -0.2159, -0.231, -0.2249, -0.2339, -0.2339, -0.2094, -0.2134, -0.1878, -0.2115, -0.1984, -0.176, -0.1719, -0.1832, -0.1731, -0.1678, -0.16, -0.1665, -0.1539, -0.1716, -0.1557, -0.159, -0.1511, -0.1511, -0.1606, -0.1587, -0.137, -0.0996, -0.0954, -0.0795, -0.0781, -0.1054, -0.0824, -0.0702, -0.0702, -0.0766, -0.0663, -0.0495, -0.0424, -0.0376, -0.0426, -0.0039, 0.0, 0.0095, 0.0061, 0.0001, -0.0253, -0.0034, 0.0138, 0.0124, 0.0314, 0.0242, 0.0461, 0.0379, 0.0137, 0.0504, 0.0402, 0.047, 0.0549, 0.0661, 0.0624, 0.0688, 0.0596, 0.0621, 0.053, 0.0621, 0.0249, 0.0235, 0.0211, 0.0386, 0.0493, 0.0607, 0.0597, 0.0514, 0.0164, 0.0164, -0.0034, -0.0044, 0.0017, -0.0254, -0.0179, -0.0036, 0.0348, 0.0339, 0.0377, 0.0373, 0.0205, -0.0062, 0.0285, 0.031, 0.0715, 0.0413, 0.0327, 0.0369, 0.0399, 0.0612, 0.0888, 0.0927, 0.1023, 0.0949, 0.0827, 0.0798, 0.1036, 0.1238, 0.0689, 0.099, 0.0506, 0.0494, 0.061, 0.0692, 0.0658, 0.0572, 0.0521, 0.063, 0.0869, 0.1175, 0.1731, 0.2082, 0.184, 0.1817, 0.2073, 0.1595, 0.1392, 0.0976, 0.098, 0.1616, 0.1272, 0.131, 0.0905, 0.1098, 0.0889, 0.0584, 0.0885, 0.0542, 0.0439, 0.0653, 0.0377, 0.0519, 0.0519, 0.0329, 0.05, 0.0589, 0.048, 0.0702, 0.0645, 0.0829, 0.0795, 0.0725, 0.0559, 0.0214, 0.0288, 0.0372, -0.0024, 0.0163, 0.0563, 0.072, 0.1042, 0.1007, 0.1007, 0.1119, 0.0984, 0.0945, 0.0884, 0.0884, 0.1021, 0.0979, 0.0927, 0.1036, 0.0799]}
{"unit_id": "thread:1945428075006812261", "ticker_idx": 1, "ts": "2025-07-16T10:19:32+00:00", "symbol": "HBM", "kind": "sector", "geography": "global", "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "H20 is anticipated to once again establish itself as a primary high-performance AI chip in the market and is expected to contribute to an increase in HBM demand", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares TrendForce bullish thesis: H20 export approval revives NVIDIA demand in China and drives HBM consumption higher.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "Global HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶️ H20 Export Approval Reignites Demand Momentum, China’s Procurement Ratio for Foreign AI Chips Expected to Rise to 49%, Says TrendForce\n\n- TrendForce expects that the US government's lifting of export restrictions on NVIDIA's H20 GPU to China will stimulate a recovery in demand from local Chinese AI companies and CSPs (Cloud Service Providers).\n\n- The H20 is anticipated to once again establish itself as a primary high-performance AI chip in the market and is expected to contribute to an increase in HBM demand.\n\n- Following the strengthening of US AI chip export regulations in April, the annual shipment forecast for the H20, which was developed specifically for the Chinese market, was drastically lowered.\n\n- However, considering recent trends and the likelihood that NVIDIA will strive to meet its initial shipment targets, TrendForce has revised China's procurement ratio for foreign AI chips (primarily NVIDIA and AMD) upwards from the previously estimated 42% to 49%.\n\n- While Chinese companies like Huawei are developing their own AI chips, their ecosystems are not yet mature enough to fully replace the advantages of NVIDIA's system integration, computational efficiency, and supply chain scale.\n\n- The resumption of H20 shipments is expected to significantly restore potential demand for AI infrastructure construction, especially among major CSPs in China who have been prioritizing H20 for their domestic data center builds.\n\n- Furthermore, NVIDIA plans to launch a custom version of the RTX PRO 6000 targeting the Chinese market to support a wider range of edge AI inference applications.\n\n- H20 chips to be shipped in 2024 will primarily use HBM3 8hi, with an upgrade to HBM3e 8hi and an increase in total capacity expected in early 2025.\n\n- As Chinese domestic ASIC manufacturers face HBM export restrictions, many of their products still rely on previously supplied HBM2e.\n\n- Therefore, the H20 is expected to be preferred, increasing its proportion of overall HBM consumption.\n\n- Additionally, with China's AI market still vulnerable to geopolitical uncertainties and persistent risks for NVIDIA, an aggressive inventory build is anticipated if H20 export restrictions are indeed lifted.\n\n---\n\nhttps://t.co/8ZPQYl48dY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008199044061293015, "ret_signed_m1d": 0.008199044061293015, "alpha_spy_m1d": 0.011531095495048382, "alpha_c_m1d": 0.0029446043279802714, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02861389333227719, "ret_signed_p1d": -0.02861389333227719, "alpha_spy_p1d": -0.034733583309794744, "alpha_c_p1d": -0.03701403373850913, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.04054252466373239, "ret_signed_p1w": -0.04054252466373239, "alpha_spy_p1w": -0.056546646463913795, "alpha_c_p1w": -0.03131273030891376, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.03895492576354095, "ret_signed_p1m": 0.03895492576354095, "alpha_spy_p1m": 0.005745443650970371, "alpha_c_p1m": -0.004739794176927083, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.5030786269450117, "ret_signed_p3m": 0.5030786269450117, "alpha_spy_p3m": 0.43794412609698574, "alpha_c_p3m": 0.3286109732370671, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 1.512175882780585, "ret_signed_p6m": 1.512175882780585, "alpha_spy_p6m": 1.4012458153329457, "alpha_c_p6m": 1.1980191609380504, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.3233, -0.3269, -0.3222, -0.3027, -0.2929, -0.2793, -0.285, -0.2911, -0.2962, -0.2938, -0.282, -0.2828, -0.2826, -0.2696, -0.263, -0.2601, -0.2219, -0.2083, -0.2019, -0.2082, -0.199, -0.208, -0.2061, -0.2153, -0.2273, -0.2306, -0.2247, -0.2207, -0.2048, -0.1971, -0.2118, -0.1947, -0.183, -0.1638, -0.1405, -0.134, -0.1186, -0.1108, -0.0911, -0.0978, -0.1056, -0.0851, -0.0779, -0.0678, -0.0714, -0.0524, -0.0616, -0.0107, -0.0001, -0.0014, -0.0101, -0.0106, -0.0226, -0.0244, -0.0079, -0.0195, -0.0336, -0.0099, -0.0224, 0.0012, 0.0107, -0.0006, 0.0082, 0.0, -0.0286, -0.0239, -0.0197, -0.0449, -0.0405, -0.0366, -0.0424, -0.0238, -0.0201, -0.0007, -0.0117, -0.0542, -0.0418, -0.0309, -0.043, -0.0215, -0.0009, 0.0207, 0.035, 0.0393, 0.039, 0.0263, 0.0156, 0.0062, -0.0135, -0.0289, -0.0126, -0.0061, -0.0098, -0.0064, 0.0107, 0.0032, -0.0223, -0.011, -0.0045, 0.0162, 0.0426, 0.0499, 0.0803, 0.1177, 0.1555, 0.2091, 0.2191, 0.2562, 0.237, 0.2981, 0.2805, 0.2973, 0.3199, 0.3061, 0.2946, 0.2589, 0.2986, 0.3051, 0.3725, 0.4365, 0.4486, 0.4575, 0.4424, 0.4735, 0.4615, 0.4914, 0.5031, 0.474, 0.518, 0.5961, 0.6114, 0.6475, 0.6243, 0.6218, 0.6312, 0.7139, 0.7617, 0.7382, 0.7986, 0.821, 0.8278, 0.9465, 0.8283, 0.8539, 0.8649, 0.8423, 0.9297, 0.9244, 0.9309, 0.901, 0.8417, 0.8971, 0.8035, 0.7804, 0.7414, 0.6722, 0.7284, 0.7424, 0.7826, 0.8092, 0.7957, 0.8177, 0.8509, 0.8346, 0.8049, 0.8545, 0.9252, 0.9229, 0.9768, 0.9333, 0.8988, 0.8479, 0.7962, 0.8263, 0.8918, 0.9292, 1.0188, 1.0275, 1.0598, 1.0598, 1.0973, 1.1873, 1.1968, 1.1761, 1.1761, 1.3362, 1.3982, 1.5259, 1.5437, 1.5122]}
{"unit_id": "quote:1968833380797268390", "ticker_idx": 0, "ts": "2025-09-19T00:23:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "SK hynix is likely to handle the entire HBM4 volume in the first half of next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "SK Hynix to capture ~70–100% of HBM4 volume in H1 2025; Micron lagging in HBM and likely retreating to commodity memory.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Summary:\n\n- Rubin’s tape-out has been completed.\n\n- The industry expects about 70% of Nvidia’s HBM4 volume to be allocated to SK hynix.\n\n- In fact, SK hynix is likely to handle the entire HBM4 volume in the first half of next year.\n\n- Samsung’s 1c DRAM yield remains at only around 60%.\n\n- Considering that yields decline in the latter stages of processing, a stable mass production ramp requires reaching 80%.\n\n- In Micron’s case, CS samples may be shipped at the same time as Samsung’s or may not be released at all within this year.\n\n- Micron is somewhat lagging in HBM competitiveness, and there is a possibility it will expand into the commodity memory space within its limited capacity.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics, HBM4 ‘CS Sample’ Shipment in November… Yield Issues Remain a Drag\n\nSamsung Electronics’ HBM4 12-high, developed based on 10nm-class 6th-generation (1c) DRAM, is expected to face difficulties securing Nvidia’s initial orders. The CS sample shipment schedule", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.004928464706804525, "alpha_c_m1d": -0.003991042456217064, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": 0.003991042456217064, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012658236774113485, "ret_signed_p1d": -0.012658236774113485, "alpha_spy_p1d": -0.01738924474220971, "alpha_c_p1d": -0.033120086490420864, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.02046184971630738, "ret_p1w": -0.05344589839563141, "ret_signed_p1w": -0.05344589839563141, "alpha_spy_p1w": -0.05061334217248403, "alpha_c_p1w": -0.0727991575288014, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.019353259133169987, "ret_p1m": 0.3656820822944049, "ret_signed_p1m": 0.3656820822944049, "alpha_spy_p1m": 0.3542311637043847, "alpha_c_p1m": 0.2658752832685063, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.09980679902589862, "ret_p3m": 0.551039645222434, "ret_signed_p3m": 0.551039645222434, "alpha_spy_p3m": 0.5394380676032073, "alpha_c_p3m": 0.47650925449103854, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.07453039073139545, "ret_p6m": 1.7468243346157752, "ret_signed_p6m": 1.7468243346157752, "alpha_spy_p6m": 1.7358149422414897, "alpha_c_p6m": 1.495214512346127, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": 0.25160982226964834, "price_path": [-0.2177, -0.1967, -0.177, -0.2023, -0.1798, -0.1981, -0.2163, -0.2177, -0.2402, -0.2388, -0.2079, -0.2107, -0.1658, -0.1728, -0.1573, -0.1615, -0.1686, -0.243, -0.2444, -0.2346, -0.2458, -0.2444, -0.243, -0.2528, -0.2641, -0.2627, -0.2599, -0.2318, -0.2753, -0.2753, -0.2599, -0.2739, -0.2641, -0.2795, -0.25, -0.2444, -0.2191, -0.2233, -0.2233, -0.2486, -0.2613, -0.2823, -0.3118, -0.295, -0.2711, -0.2655, -0.2697, -0.2447, -0.2433, -0.2799, -0.2672, -0.2616, -0.2532, -0.2307, -0.2208, -0.1899, -0.1449, -0.1364, -0.0759, -0.0689, -0.0211, -0.0619, 0.0, 0.0, -0.0127, 0.0155, 0.0056, 0.0028, -0.0534, -0.0183, -0.0225, 0.0127, 0.1125, 0.1125, 0.1125, 0.1125, 0.1125, 0.1125, 0.2039, 0.1674, 0.1575, 0.1885, 0.2729, 0.3094, 0.3657, 0.3474, 0.3544, 0.346, 0.4346, 0.5049, 0.4655, 0.5696, 0.5977, 0.5724, 0.744, 0.6484, 0.6287, 0.6681, 0.6315, 0.7046, 0.7412, 0.7356, 0.7215, 0.5752, 0.7046, 0.6034, 0.5809, 0.6062, 0.4655, 0.4627, 0.4599, 0.474, 0.5313, 0.4919, 0.5144, 0.5707, 0.5539, 0.5257, 0.5313, 0.6242, 0.5933, 0.6524, 0.5904, 0.6073, 0.5595, 0.4919, 0.551, 0.5539, 0.5398, 0.6327, 0.6439, 0.6552, 0.6552, 0.6862, 0.8016, 0.8325, 0.8325, 0.8325, 0.9057, 0.9592, 1.0437, 1.0887, 1.1281, 1.0943, 1.1084, 1.0774, 1.0887, 1.1084, 1.1281, 1.1506, 1.0915, 1.0831, 1.1253, 1.1591, 1.0718, 1.252, 1.3674, 1.4237, 1.5588, 1.3364, 1.5532, 1.5335, 1.3702, 1.3617, 1.4969, 1.4659, 1.4209, 1.4997, 1.4772, 1.4772, 1.4772, 1.4772, 1.5166, 1.6714, 1.677, 1.829, 1.8656, 2.0993, 1.9922, 1.9922, 1.6481, 1.3943, 1.6538, 1.6058, 1.3576, 1.6453, 1.6932, 1.6227, 1.5663, 1.7468]}
{"unit_id": "quote:1968833380797268390", "ticker_idx": 1, "ts": "2025-09-19T00:23:51+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is somewhat lagging in HBM competitiveness, and there is a possibility it will expand into the commodity memory space within its limited capacity", "tweet_type": "analysis", "timeline": "quarters", "summary": "SK Hynix to capture ~70–100% of HBM4 volume in H1 2025; Micron lagging in HBM and likely retreating to commodity memory.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Summary:\n\n- Rubin’s tape-out has been completed.\n\n- The industry expects about 70% of Nvidia’s HBM4 volume to be allocated to SK hynix.\n\n- In fact, SK hynix is likely to handle the entire HBM4 volume in the first half of next year.\n\n- Samsung’s 1c DRAM yield remains at only around 60%.\n\n- Considering that yields decline in the latter stages of processing, a stable mass production ramp requires reaching 80%.\n\n- In Micron’s case, CS samples may be shipped at the same time as Samsung’s or may not be released at all within this year.\n\n- Micron is somewhat lagging in HBM competitiveness, and there is a possibility it will expand into the commodity memory space within its limited capacity.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics, HBM4 ‘CS Sample’ Shipment in November… Yield Issues Remain a Drag\n\nSamsung Electronics’ HBM4 12-high, developed based on 10nm-class 6th-generation (1c) DRAM, is expected to face difficulties securing Nvidia’s initial orders. The CS sample shipment schedule", "parent_author": null, "ret_m1d": 0.03785413036315055, "ret_signed_m1d": -0.03785413036315055, "alpha_spy_m1d": -0.042782595069955076, "alpha_c_m1d": -0.03386308790693349, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": 0.003991042456217064, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011614370666173546, "ret_signed_p1d": -0.011614370666173546, "alpha_spy_p1d": -0.006883362698077322, "alpha_c_p1d": 0.008847479050133833, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.02046184971630738, "ret_p1w": -0.03355240726058739, "ret_signed_p1w": 0.03355240726058739, "alpha_spy_p1w": 0.030719851037440016, "alpha_c_p1w": 0.05290566639375738, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.019353259133169987, "ret_p1m": 0.27142811666693745, "ret_signed_p1m": -0.27142811666693745, "alpha_spy_p1m": -0.2599771980769172, "alpha_c_p1m": -0.17162131764103883, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.09980679902589862, "ret_p3m": 0.3867218081022077, "ret_signed_p3m": -0.3867218081022077, "alpha_spy_p3m": -0.37512023048298104, "alpha_c_p3m": -0.31219141737081224, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.07453039073139545, "ret_p6m": 1.717724235415905, "ret_signed_p6m": -1.717724235415905, "alpha_spy_p6m": -1.7067148430416195, "alpha_c_p6m": -1.4661144131462567, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": 0.25160982226964834, "price_path": [-0.2147, -0.2188, -0.2264, -0.2341, -0.2433, -0.2578, -0.2526, -0.2492, -0.2492, -0.2631, -0.2354, -0.2488, -0.2435, -0.2347, -0.2711, -0.2619, -0.2845, -0.304, -0.2971, -0.3042, -0.3288, -0.3251, -0.3134, -0.3163, -0.3164, -0.312, -0.2949, -0.3293, -0.3555, -0.3377, -0.3298, -0.3315, -0.3125, -0.2694, -0.2397, -0.215, -0.2363, -0.2301, -0.2572, -0.2408, -0.25, -0.2797, -0.2885, -0.2768, -0.2846, -0.2841, -0.2764, -0.2503, -0.2687, -0.2687, -0.2719, -0.2704, -0.2367, -0.1927, -0.1922, -0.1689, -0.1397, -0.0747, -0.0338, -0.0305, -0.024, -0.0168, 0.0379, 0.0, 0.0116, 0.0226, -0.0063, -0.0363, -0.0336, 0.0072, 0.0282, 0.1193, 0.1292, 0.155, 0.1742, 0.1418, 0.2085, 0.1826, 0.1167, 0.1853, 0.1502, 0.1802, 0.2454, 0.2444, 0.2714, 0.2439, 0.2204, 0.2711, 0.3468, 0.3534, 0.3645, 0.3935, 0.3774, 0.376, 0.4432, 0.3407, 0.4604, 0.4655, 0.463, 0.5575, 0.4826, 0.5059, 0.457, 0.5178, 0.4877, 0.405, 0.3892, 0.2382, 0.2751, 0.3769, 0.3806, 0.4159, 0.4159, 0.4541, 0.4786, 0.4726, 0.4398, 0.3937, 0.4587, 0.5183, 0.5521, 0.6216, 0.5893, 0.4828, 0.4604, 0.4297, 0.3867, 0.5283, 0.6351, 0.7008, 0.6988, 0.7628, 0.7628, 0.7512, 0.8108, 0.8001, 0.7557, 0.7557, 0.9403, 0.9202, 1.1126, 1.0887, 1.0117, 1.1228, 1.1276, 1.08, 1.0506, 1.0708, 1.2314, 1.2314, 1.2453, 1.3936, 1.4457, 1.4584, 1.3935, 1.5236, 1.6776, 1.6808, 1.5521, 1.6931, 1.5802, 1.3339, 1.3553, 1.4279, 1.3591, 1.296, 1.5242, 1.5465, 1.5323, 1.5323, 1.4592, 1.5895, 1.5673, 1.6339, 1.5896, 1.5714, 1.639, 1.5563, 1.5367, 1.5385, 1.3356, 1.4653, 1.4424, 1.2779, 1.3949, 1.4797, 1.5756, 1.4935, 1.6213, 1.7177]}
{"unit_id": "orig:1948197651914346913", "ticker_idx": 0, "ts": "2025-07-24T01:44:50+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom. Although the official export license is still pending and supply chain availability remains uncertain, this event has positive implications for the AI industry starting from 2026.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: H20 export resumption boosts NVDA, AMD, AVGO revenue; TSMC 45% AI semi CAGR; Samsung reverses impairment loss; China AI capex RMB 380B on track.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "China AI: H20 Export Resumption Sends Positive Signal to China-US AI Industrial Chain\nMorgan Stanley (C. Chan, 2025/07/20)\n\nMorgan Stanley stated that the US's proposed resumption of exports of medium-level AI chips (such as H20) to China is beneficial for China's AI infrastructure investment, US semiconductor companies, and the Asian supply chain, while also creating temporary pressure on domestic Chinese GPUs.\n\nH20 Export Approval to Facilitate China's AI Capital Expenditure Realization\nMorgan Stanley pointed out that if H20 chip shipments resume, China's CSPs' AI capital expenditure, which could reach RMB 380 billion in 2025, is expected to proceed as planned, and may even see upside. This further validates China's strategic path of promoting AI popularization and enhancing daily user experience through application and large model deployment.\n\nUS AI Semiconductor Companies to Benefit Significantly, NVDA Outlook Boosted\nAccording to Nvidia's estimation of the AI accelerator market, China's AI accelerator expenditure is approximately $50 billion. Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom. Although the official export license is still pending and supply chain availability remains uncertain, this event has positive implications for the AI industry starting from 2026.\n\nAsian Supply Chain to Benefit Simultaneously, TSMC and Samsung Show Strong Performance\nTSMC stated that the resumption of H20 shipments is beneficial for the company. Morgan Stanley expects its AI semiconductor revenue to achieve a compound annual growth rate of 45% over the next five years. Samsung is also expected to reverse its previous KRW 1.5 trillion impairment loss due to this event. In addition, server ODM manufacturers such as Wistron, Foxconn Industrial Internet, and Lenovo are expected to benefit from China's CSPs' H20 server projects.\n\nDomestic GPU Manufacturers Under Pressure, Self-Sufficiency Process to Continue\nMorgan Stanley noted that this event may slow down the penetration rate of domestic GPUs, cautiously estimating the self-sufficiency rate of domestic GPUs to 39% by 2027, while emphasizing that China's will to promote AI chip self-sufficiency will not weaken amidst geopolitical risks.\n\nChina to Achieve Leadership and Accelerated Development in ToC AI Applications\nChina is expected to take the lead in ToC AI applications, benefiting from breakthroughs in AI-native applications, the integration capabilities of super apps, the open-source large model ecosystem, declining inference costs, and accelerated capital expenditures. Starting from H2 2024, AI-native applications have significantly improved user retention, product differentiation, and functional diversity, overcoming the challenges of early AI chatbots. At the same time, AI functionalities are gradually being integrated into super apps, leveraging user activity and proprietary data resources to enhance usage scenarios and penetration.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017037010424765353, "ret_signed_m1d": -0.017037010424765353, "alpha_spy_m1d": -0.01670606746144132, "alpha_c_m1d": -0.012902190599967334, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0013813721014661606, "ret_signed_p1d": -0.0013813721014661606, "alpha_spy_p1d": -0.005605721741785308, "alpha_c_p1d": -0.0003041894717301652, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.023771089619391894, "ret_signed_p1w": 0.023771089619391894, "alpha_spy_p1w": 0.027459391217387497, "alpha_c_p1w": 0.020365950138228728, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": 0.024461863510141457, "ret_signed_p1m": 0.024461863510141457, "alpha_spy_p1m": 0.007296489438838538, "alpha_c_p1m": 0.00416994244368496, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.042766229694791935, "ret_signed_p3m": 0.042766229694791935, "alpha_spy_p3m": -0.018283287367156342, "alpha_c_p3m": -0.15702524766006798, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 0.07200904962987287, "ret_signed_p6m": 0.07200904962987287, "alpha_spy_p6m": -0.02446805233820437, "alpha_c_p6m": -0.32352907207293735, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3742, -0.3726, -0.3731, -0.3576, -0.341, -0.3449, -0.3465, -0.3263, -0.3245, -0.3286, -0.2921, -0.2522, -0.2211, -0.224, -0.2207, -0.2198, -0.2266, -0.2414, -0.2355, -0.2444, -0.2444, -0.2202, -0.2241, -0.1989, -0.2223, -0.2093, -0.1872, -0.1832, -0.1943, -0.1844, -0.1791, -0.1715, -0.1779, -0.1654, -0.1829, -0.1672, -0.1705, -0.1627, -0.1627, -0.172, -0.1702, -0.1487, -0.1118, -0.1077, -0.092, -0.0907, -0.1176, -0.0949, -0.0829, -0.0829, -0.0892, -0.0791, -0.0625, -0.0555, -0.0508, -0.0557, -0.0175, -0.0136, -0.0043, -0.0077, -0.0136, -0.0386, -0.017, 0.0, -0.0014, 0.0173, 0.0102, 0.0318, 0.0238, -0.0001, 0.036, 0.026, 0.0327, 0.0405, 0.0516, 0.0479, 0.0542, 0.0452, 0.0477, 0.0386, 0.0476, 0.0109, 0.0096, 0.0071, 0.0245, 0.0349, 0.0462, 0.0452, 0.037, 0.0025, 0.0025, -0.017, -0.018, -0.012, -0.0387, -0.0313, -0.0172, 0.0207, 0.0198, 0.0235, 0.0231, 0.0066, -0.0198, 0.0144, 0.0169, 0.0569, 0.0271, 0.0186, 0.0228, 0.0257, 0.0467, 0.074, 0.0778, 0.0873, 0.08, 0.068, 0.0651, 0.0885, 0.1084, 0.0543, 0.084, 0.0363, 0.0351, 0.0465, 0.0546, 0.0513, 0.0428, 0.0377, 0.0485, 0.0721, 0.1022, 0.1571, 0.1917, 0.1678, 0.1655, 0.1908, 0.1437, 0.1236, 0.0826, 0.083, 0.1457, 0.1118, 0.1155, 0.0756, 0.0946, 0.0741, 0.0439, 0.0736, 0.0398, 0.0296, 0.0508, 0.0235, 0.0376, 0.0376, 0.0188, 0.0356, 0.0445, 0.0337, 0.0556, 0.05, 0.0681, 0.0648, 0.0579, 0.0415, 0.0075, 0.0148, 0.023, -0.016, 0.0024, 0.0418, 0.0574, 0.0892, 0.0857, 0.0857, 0.0968, 0.0835, 0.0795, 0.0736, 0.0736, 0.0871, 0.0829, 0.0778, 0.0886, 0.0652, 0.0641, 0.0646, 0.0696, 0.0542, 0.0767, 0.072]}
{"unit_id": "orig:1948197651914346913", "ticker_idx": 1, "ts": "2025-07-24T01:44:50+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: H20 export resumption boosts NVDA, AMD, AVGO revenue; TSMC 45% AI semi CAGR; Samsung reverses impairment loss; China AI capex RMB 380B on track.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "China AI: H20 Export Resumption Sends Positive Signal to China-US AI Industrial Chain\nMorgan Stanley (C. Chan, 2025/07/20)\n\nMorgan Stanley stated that the US's proposed resumption of exports of medium-level AI chips (such as H20) to China is beneficial for China's AI infrastructure investment, US semiconductor companies, and the Asian supply chain, while also creating temporary pressure on domestic Chinese GPUs.\n\nH20 Export Approval to Facilitate China's AI Capital Expenditure Realization\nMorgan Stanley pointed out that if H20 chip shipments resume, China's CSPs' AI capital expenditure, which could reach RMB 380 billion in 2025, is expected to proceed as planned, and may even see upside. This further validates China's strategic path of promoting AI popularization and enhancing daily user experience through application and large model deployment.\n\nUS AI Semiconductor Companies to Benefit Significantly, NVDA Outlook Boosted\nAccording to Nvidia's estimation of the AI accelerator market, China's AI accelerator expenditure is approximately $50 billion. Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom. Although the official export license is still pending and supply chain availability remains uncertain, this event has positive implications for the AI industry starting from 2026.\n\nAsian Supply Chain to Benefit Simultaneously, TSMC and Samsung Show Strong Performance\nTSMC stated that the resumption of H20 shipments is beneficial for the company. Morgan Stanley expects its AI semiconductor revenue to achieve a compound annual growth rate of 45% over the next five years. Samsung is also expected to reverse its previous KRW 1.5 trillion impairment loss due to this event. In addition, server ODM manufacturers such as Wistron, Foxconn Industrial Internet, and Lenovo are expected to benefit from China's CSPs' H20 server projects.\n\nDomestic GPU Manufacturers Under Pressure, Self-Sufficiency Process to Continue\nMorgan Stanley noted that this event may slow down the penetration rate of domestic GPUs, cautiously estimating the self-sufficiency rate of domestic GPUs to 39% by 2027, while emphasizing that China's will to promote AI chip self-sufficiency will not weaken amidst geopolitical risks.\n\nChina to Achieve Leadership and Accelerated Development in ToC AI Applications\nChina is expected to take the lead in ToC AI applications, benefiting from breakthroughs in AI-native applications, the integration capabilities of super apps, the open-source large model ecosystem, declining inference costs, and accelerated capital expenditures. Starting from H2 2024, AI-native applications have significantly improved user retention, product differentiation, and functional diversity, overcoming the challenges of early AI chatbots. At the same time, AI functionalities are gradually being integrated into super apps, leveraging user activity and proprietary data resources to enhance usage scenarios and penetration.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02140390652118418, "ret_signed_m1d": -0.02140390652118418, "alpha_spy_m1d": -0.021072963557860147, "alpha_c_m1d": -0.01726908669638616, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026832014770116652, "ret_signed_p1d": 0.026832014770116652, "alpha_spy_p1d": 0.022607665129797505, "alpha_c_p1d": 0.027909197399852648, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.08752777491233621, "ret_signed_p1w": 0.08752777491233621, "alpha_spy_p1w": 0.09121607651033181, "alpha_c_p1w": 0.08412263543117304, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": 0.03478904243471992, "ret_signed_p1m": 0.03478904243471992, "alpha_spy_p1m": 0.017623668363417, "alpha_c_p1m": 0.014497121368263421, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.46823344404395195, "ret_signed_p3m": 0.46823344404395195, "alpha_spy_p3m": 0.4071839269820037, "alpha_c_p3m": 0.26844196668909204, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 0.429990185130944, "ret_signed_p6m": 0.429990185130944, "alpha_spy_p6m": 0.33351308316286676, "alpha_c_p6m": 0.03445206342813378, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.4054, -0.4075, -0.3995, -0.4038, -0.3906, -0.3795, -0.3917, -0.381, -0.3727, -0.3657, -0.3331, -0.3063, -0.2739, -0.2907, -0.2773, -0.2923, -0.2998, -0.3088, -0.3171, -0.3196, -0.3196, -0.2934, -0.3038, -0.3028, -0.317, -0.2929, -0.2764, -0.2686, -0.2864, -0.2833, -0.2491, -0.2398, -0.2528, -0.2691, -0.2835, -0.2204, -0.216, -0.2179, -0.2179, -0.209, -0.2007, -0.1461, -0.1155, -0.1137, -0.1129, -0.1247, -0.1604, -0.1456, -0.1493, -0.1493, -0.1685, -0.1499, -0.1462, -0.1108, -0.0968, -0.098, -0.0402, -0.0126, -0.0105, -0.0316, -0.0316, -0.0456, -0.0214, 0.0, 0.0268, 0.0712, 0.0945, 0.1073, 0.0875, 0.0591, 0.0904, 0.0752, 0.0062, 0.0634, 0.0656, 0.0627, 0.0791, 0.1376, 0.1161, 0.0949, 0.0865, 0.0273, 0.019, 0.0098, 0.0348, 0.0076, 0.0278, 0.0309, 0.0398, 0.0031, 0.0031, 0.0012, 0.0001, -0.002, -0.0677, -0.0661, -0.0389, -0.0159, -0.0398, -0.0219, -0.0059, -0.0102, -0.0183, -0.0259, -0.0292, -0.0144, -0.0075, -0.0076, -0.0052, -0.0164, -0.0047, -0.002, 0.0117, 0.0469, 0.0157, 0.2565, 0.3047, 0.453, 0.4365, 0.3256, 0.3349, 0.3452, 0.4717, 0.4468, 0.4377, 0.4838, 0.4682, 0.4201, 0.4495, 0.5601, 0.6017, 0.5915, 0.6305, 0.5719, 0.5798, 0.6016, 0.5424, 0.5811, 0.4662, 0.4405, 0.5049, 0.4651, 0.5969, 0.5295, 0.5224, 0.4836, 0.4205, 0.3789, 0.2708, 0.257, 0.3265, 0.2715, 0.3215, 0.3215, 0.3418, 0.3555, 0.3277, 0.3422, 0.3322, 0.3445, 0.3639, 0.367, 0.3658, 0.3658, 0.3001, 0.2804, 0.2902, 0.222, 0.2402, 0.3165, 0.3259, 0.3256, 0.3264, 0.3264, 0.3261, 0.3299, 0.3283, 0.321, 0.321, 0.3784, 0.3637, 0.3222, 0.2955, 0.2625, 0.2532, 0.2811, 0.363, 0.3792, 0.4059, 0.43]}
{"unit_id": "orig:1948197651914346913", "ticker_idx": 2, "ts": "2025-07-24T01:44:50+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: H20 export resumption boosts NVDA, AMD, AVGO revenue; TSMC 45% AI semi CAGR; Samsung reverses impairment loss; China AI capex RMB 380B on track.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "China AI: H20 Export Resumption Sends Positive Signal to China-US AI Industrial Chain\nMorgan Stanley (C. Chan, 2025/07/20)\n\nMorgan Stanley stated that the US's proposed resumption of exports of medium-level AI chips (such as H20) to China is beneficial for China's AI infrastructure investment, US semiconductor companies, and the Asian supply chain, while also creating temporary pressure on domestic Chinese GPUs.\n\nH20 Export Approval to Facilitate China's AI Capital Expenditure Realization\nMorgan Stanley pointed out that if H20 chip shipments resume, China's CSPs' AI capital expenditure, which could reach RMB 380 billion in 2025, is expected to proceed as planned, and may even see upside. This further validates China's strategic path of promoting AI popularization and enhancing daily user experience through application and large model deployment.\n\nUS AI Semiconductor Companies to Benefit Significantly, NVDA Outlook Boosted\nAccording to Nvidia's estimation of the AI accelerator market, China's AI accelerator expenditure is approximately $50 billion. Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom. Although the official export license is still pending and supply chain availability remains uncertain, this event has positive implications for the AI industry starting from 2026.\n\nAsian Supply Chain to Benefit Simultaneously, TSMC and Samsung Show Strong Performance\nTSMC stated that the resumption of H20 shipments is beneficial for the company. Morgan Stanley expects its AI semiconductor revenue to achieve a compound annual growth rate of 45% over the next five years. Samsung is also expected to reverse its previous KRW 1.5 trillion impairment loss due to this event. In addition, server ODM manufacturers such as Wistron, Foxconn Industrial Internet, and Lenovo are expected to benefit from China's CSPs' H20 server projects.\n\nDomestic GPU Manufacturers Under Pressure, Self-Sufficiency Process to Continue\nMorgan Stanley noted that this event may slow down the penetration rate of domestic GPUs, cautiously estimating the self-sufficiency rate of domestic GPUs to 39% by 2027, while emphasizing that China's will to promote AI chip self-sufficiency will not weaken amidst geopolitical risks.\n\nChina to Achieve Leadership and Accelerated Development in ToC AI Applications\nChina is expected to take the lead in ToC AI applications, benefiting from breakthroughs in AI-native applications, the integration capabilities of super apps, the open-source large model ecosystem, declining inference costs, and accelerated capital expenditures. Starting from H2 2024, AI-native applications have significantly improved user retention, product differentiation, and functional diversity, overcoming the challenges of early AI chatbots. At the same time, AI functionalities are gradually being integrated into super apps, leveraging user activity and proprietary data resources to enhance usage scenarios and penetration.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017387553009892365, "ret_signed_m1d": -0.017387553009892365, "alpha_spy_m1d": -0.017056610046568332, "alpha_c_m1d": -0.013252733185094345, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005091586146982019, "ret_signed_p1d": 0.005091586146982019, "alpha_spy_p1d": 0.0008672365066628718, "alpha_c_p1d": 0.006168768776718014, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.01728390126868673, "ret_signed_p1w": 0.01728390126868673, "alpha_spy_p1w": 0.020972202866682332, "alpha_c_p1w": 0.013878761787523564, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": 0.018322970108219927, "ret_signed_p1m": 0.018322970108219927, "alpha_spy_p1m": 0.001157596036917008, "alpha_c_p1m": -0.0019689509582365705, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.1888993981714131, "ret_signed_p3m": 0.1888993981714131, "alpha_spy_p3m": 0.12784988110946482, "alpha_c_p3m": -0.010892079183446812, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 0.22263426594740499, "ret_signed_p6m": 0.22263426594740499, "alpha_spy_p6m": 0.12615716397932775, "alpha_c_p6m": -0.17290385575540523, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3349, -0.3394, -0.3349, -0.3181, -0.2963, -0.3064, -0.3086, -0.2923, -0.282, -0.2806, -0.2343, -0.1969, -0.1979, -0.1961, -0.21, -0.203, -0.1994, -0.2062, -0.2034, -0.2096, -0.2096, -0.1857, -0.1726, -0.1639, -0.1635, -0.1406, -0.1124, -0.0978, -0.1018, -0.1467, -0.1559, -0.1547, -0.1261, -0.1151, -0.1406, -0.1289, -0.1383, -0.1318, -0.1318, -0.1341, -0.121, -0.0864, -0.0833, -0.0642, -0.0671, -0.0452, -0.083, -0.0652, -0.0469, -0.0469, -0.0503, -0.0586, -0.0374, -0.0461, -0.0496, -0.0454, -0.0269, -0.0274, -0.0078, -0.0186, -0.0017, -0.0351, -0.0174, 0.0, 0.0051, 0.0194, 0.0302, 0.0482, 0.0173, -0.0002, 0.0312, 0.0146, 0.0449, 0.0521, 0.0563, 0.0526, 0.0835, 0.0706, 0.078, 0.0611, 0.0591, 0.0215, 0.0085, 0.0031, 0.0183, 0.0191, 0.0322, 0.04, 0.0691, 0.0301, 0.0301, 0.033, 0.0474, 0.0602, 0.16, 0.1972, 0.1661, 0.2801, 0.2456, 0.2465, 0.2611, 0.2469, 0.199, 0.1962, 0.1948, 0.1755, 0.176, 0.1773, 0.1661, 0.1607, 0.1377, 0.1447, 0.1567, 0.1734, 0.174, 0.164, 0.1672, 0.1988, 0.1971, 0.1263, 0.2376, 0.194, 0.219, 0.2288, 0.212, 0.2117, 0.1889, 0.1807, 0.1946, 0.2287, 0.2562, 0.2941, 0.3392, 0.3062, 0.2825, 0.2579, 0.2211, 0.2455, 0.2338, 0.2124, 0.2435, 0.2212, 0.2325, 0.1796, 0.1882, 0.1889, 0.1814, 0.2297, 0.2033, 0.1804, 0.3114, 0.3359, 0.3794, 0.3794, 0.3981, 0.3396, 0.3239, 0.3206, 0.322, 0.354, 0.3917, 0.4097, 0.4328, 0.4099, 0.2488, 0.179, 0.1842, 0.1312, 0.1446, 0.1809, 0.187, 0.2143, 0.2175, 0.2175, 0.2241, 0.2146, 0.2162, 0.2031, 0.2031, 0.2084, 0.1938, 0.195, 0.1941, 0.1558, 0.1992, 0.2244, 0.2327, 0.1815, 0.1924, 0.2226]}
{"unit_id": "orig:1948197651914346913", "ticker_idx": 3, "ts": "2025-07-24T01:44:50+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC stated that the resumption of H20 shipments is beneficial for the company. Morgan Stanley expects its AI semiconductor revenue to achieve a compound annual growth rate of 45% over the next five years.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: H20 export resumption boosts NVDA, AMD, AVGO revenue; TSMC 45% AI semi CAGR; Samsung reverses impairment loss; China AI capex RMB 380B on track.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "China AI: H20 Export Resumption Sends Positive Signal to China-US AI Industrial Chain\nMorgan Stanley (C. Chan, 2025/07/20)\n\nMorgan Stanley stated that the US's proposed resumption of exports of medium-level AI chips (such as H20) to China is beneficial for China's AI infrastructure investment, US semiconductor companies, and the Asian supply chain, while also creating temporary pressure on domestic Chinese GPUs.\n\nH20 Export Approval to Facilitate China's AI Capital Expenditure Realization\nMorgan Stanley pointed out that if H20 chip shipments resume, China's CSPs' AI capital expenditure, which could reach RMB 380 billion in 2025, is expected to proceed as planned, and may even see upside. This further validates China's strategic path of promoting AI popularization and enhancing daily user experience through application and large model deployment.\n\nUS AI Semiconductor Companies to Benefit Significantly, NVDA Outlook Boosted\nAccording to Nvidia's estimation of the AI accelerator market, China's AI accelerator expenditure is approximately $50 billion. Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom. Although the official export license is still pending and supply chain availability remains uncertain, this event has positive implications for the AI industry starting from 2026.\n\nAsian Supply Chain to Benefit Simultaneously, TSMC and Samsung Show Strong Performance\nTSMC stated that the resumption of H20 shipments is beneficial for the company. Morgan Stanley expects its AI semiconductor revenue to achieve a compound annual growth rate of 45% over the next five years. Samsung is also expected to reverse its previous KRW 1.5 trillion impairment loss due to this event. In addition, server ODM manufacturers such as Wistron, Foxconn Industrial Internet, and Lenovo are expected to benefit from China's CSPs' H20 server projects.\n\nDomestic GPU Manufacturers Under Pressure, Self-Sufficiency Process to Continue\nMorgan Stanley noted that this event may slow down the penetration rate of domestic GPUs, cautiously estimating the self-sufficiency rate of domestic GPUs to 39% by 2027, while emphasizing that China's will to promote AI chip self-sufficiency will not weaken amidst geopolitical risks.\n\nChina to Achieve Leadership and Accelerated Development in ToC AI Applications\nChina is expected to take the lead in ToC AI applications, benefiting from breakthroughs in AI-native applications, the integration capabilities of super apps, the open-source large model ecosystem, declining inference costs, and accelerated capital expenditures. Starting from H2 2024, AI-native applications have significantly improved user retention, product differentiation, and functional diversity, overcoming the challenges of early AI chatbots. At the same time, AI functionalities are gradually being integrated into super apps, leveraging user activity and proprietary data resources to enhance usage scenarios and penetration.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0003309429633240324, "alpha_c_m1d": 0.0041348198247980195, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.004224349640319147, "alpha_c_p1d": 0.0010771826297359954, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.013100422566123537, "ret_signed_p1w": 0.013100422566123537, "alpha_spy_p1w": 0.01678872416411914, "alpha_c_p1w": 0.009695283084960371, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.008733722789339371, "ret_signed_p1m": -0.008733722789339371, "alpha_spy_p1m": -0.02589909686064229, "alpha_c_p1m": -0.02902564385579587, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.2977467988614775, "ret_signed_p3m": 0.2977467988614775, "alpha_spy_p3m": 0.2366972817995292, "alpha_c_p3m": 0.09795532150661757, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 0.5308149818917933, "ret_signed_p6m": 0.5308149818917933, "alpha_spy_p6m": 0.4343378799237161, "alpha_c_p6m": 0.1352768601889831, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.2199, -0.2156, -0.2103, -0.2103, -0.1738, -0.1842, -0.1999, -0.1929, -0.2016, -0.1747, -0.1677, -0.1573, -0.1312, -0.1364, -0.1321, -0.1442, -0.1495, -0.139, -0.146, -0.1442, -0.1512, -0.1608, -0.159, -0.159, -0.159, -0.1773, -0.1738, -0.139, -0.1321, -0.1347, -0.126, -0.0912, -0.0738, -0.0873, -0.1004, -0.1048, -0.0873, -0.0786, -0.0961, -0.0786, -0.1092, -0.083, -0.0655, -0.0611, -0.0568, -0.0742, -0.0524, -0.0524, -0.048, -0.0524, -0.0568, -0.0568, -0.048, -0.0393, -0.0393, -0.0437, -0.0306, -0.0131, -0.0131, 0.0087, 0.0044, -0.0131, 0.0, 0.0, 0.0, 0.0, -0.0087, 0.0087, 0.0131, 0.0131, -0.0087, 0.0044, -0.0175, 0.0306, 0.0262, 0.0306, 0.0306, 0.048, 0.0262, 0.0306, 0.0306, 0.0349, -0.0087, 0.0044, -0.0087, 0.0218, 0.0262, 0.0393, 0.0131, 0.0131, 0.0175, 0.0131, 0.0131, 0.0131, 0.0306, 0.0306, 0.048, 0.0699, 0.083, 0.1004, 0.0961, 0.1224, 0.1092, 0.1268, 0.1092, 0.1355, 0.175, 0.175, 0.1574, 0.1399, 0.1399, 0.1443, 0.1618, 0.1969, 0.2276, 0.2276, 0.2583, 0.2408, 0.2627, 0.2627, 0.2408, 0.2495, 0.2846, 0.3021, 0.2714, 0.2977, 0.2977, 0.2802, 0.2714, 0.2714, 0.2977, 0.2934, 0.3197, 0.3197, 0.3153, 0.3241, 0.3197, 0.2802, 0.2846, 0.2802, 0.2934, 0.2846, 0.2934, 0.2802, 0.2539, 0.2671, 0.232, 0.2232, 0.2758, 0.2144, 0.2057, 0.2408, 0.2627, 0.2583, 0.2627, 0.2364, 0.2539, 0.2714, 0.2671, 0.2802, 0.3109, 0.2977, 0.3197, 0.2933, 0.3021, 0.2757, 0.2625, 0.2581, 0.2581, 0.2581, 0.2889, 0.3109, 0.3153, 0.3153, 0.3285, 0.3461, 0.3373, 0.3637, 0.3637, 0.3944, 0.4692, 0.5, 0.4736, 0.4824, 0.478, 0.4868, 0.5044, 0.5044, 0.4868, 0.5308]}
{"unit_id": "orig:1948197651914346913", "ticker_idx": 4, "ts": "2025-07-24T01:44:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung is also expected to reverse its previous KRW 1.5 trillion impairment loss due to this event.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: H20 export resumption boosts NVDA, AMD, AVGO revenue; TSMC 45% AI semi CAGR; Samsung reverses impairment loss; China AI capex RMB 380B on track.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "China AI: H20 Export Resumption Sends Positive Signal to China-US AI Industrial Chain\nMorgan Stanley (C. Chan, 2025/07/20)\n\nMorgan Stanley stated that the US's proposed resumption of exports of medium-level AI chips (such as H20) to China is beneficial for China's AI infrastructure investment, US semiconductor companies, and the Asian supply chain, while also creating temporary pressure on domestic Chinese GPUs.\n\nH20 Export Approval to Facilitate China's AI Capital Expenditure Realization\nMorgan Stanley pointed out that if H20 chip shipments resume, China's CSPs' AI capital expenditure, which could reach RMB 380 billion in 2025, is expected to proceed as planned, and may even see upside. This further validates China's strategic path of promoting AI popularization and enhancing daily user experience through application and large model deployment.\n\nUS AI Semiconductor Companies to Benefit Significantly, NVDA Outlook Boosted\nAccording to Nvidia's estimation of the AI accelerator market, China's AI accelerator expenditure is approximately $50 billion. Resuming shipments to China will provide revenue support for Nvidia, AMD, and Broadcom. Although the official export license is still pending and supply chain availability remains uncertain, this event has positive implications for the AI industry starting from 2026.\n\nAsian Supply Chain to Benefit Simultaneously, TSMC and Samsung Show Strong Performance\nTSMC stated that the resumption of H20 shipments is beneficial for the company. Morgan Stanley expects its AI semiconductor revenue to achieve a compound annual growth rate of 45% over the next five years. Samsung is also expected to reverse its previous KRW 1.5 trillion impairment loss due to this event. In addition, server ODM manufacturers such as Wistron, Foxconn Industrial Internet, and Lenovo are expected to benefit from China's CSPs' H20 server projects.\n\nDomestic GPU Manufacturers Under Pressure, Self-Sufficiency Process to Continue\nMorgan Stanley noted that this event may slow down the penetration rate of domestic GPUs, cautiously estimating the self-sufficiency rate of domestic GPUs to 39% by 2027, while emphasizing that China's will to promote AI chip self-sufficiency will not weaken amidst geopolitical risks.\n\nChina to Achieve Leadership and Accelerated Development in ToC AI Applications\nChina is expected to take the lead in ToC AI applications, benefiting from breakthroughs in AI-native applications, the integration capabilities of super apps, the open-source large model ecosystem, declining inference costs, and accelerated capital expenditures. Starting from H2 2024, AI-native applications have significantly improved user retention, product differentiation, and functional diversity, overcoming the challenges of early AI chatbots. At the same time, AI functionalities are gradually being integrated into super apps, leveraging user activity and proprietary data resources to enhance usage scenarios and penetration.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006060579938871102, "ret_signed_m1d": 0.006060579938871102, "alpha_spy_m1d": 0.006391522902195135, "alpha_c_m1d": 0.010268923172751965, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.004208343233880862, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0015151150535629432, "ret_signed_p1d": -0.0015151150535629432, "alpha_spy_p1d": -0.00573946469388209, "alpha_c_p1d": -0.0038102446406423907, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": 0.0022951295870794475, "ret_p1w": 0.08181812848630599, "ret_signed_p1w": 0.08181812848630599, "alpha_spy_p1w": 0.08550643008430159, "alpha_c_p1w": 0.07665358871318384, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.005164539773122145, "ret_p1m": 0.08181812848630599, "ret_signed_p1m": 0.08181812848630599, "alpha_spy_p1m": 0.06465275441500307, "alpha_c_p1m": 0.0778777594204656, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.003940369065840388, "ret_p3m": 0.4838636954368518, "ret_signed_p3m": 0.4838636954368518, "alpha_spy_p3m": 0.42281417837490354, "alpha_c_p3m": 0.37934357140912267, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.10452012402772914, "ret_p6m": 1.277142139821354, "ret_signed_p6m": 1.277142139821354, "alpha_spy_p6m": 1.1806650378532768, "alpha_c_p6m": 1.159821032799669, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.11732110702168508, "price_path": [-0.1597, -0.1597, -0.1642, -0.1642, -0.1823, -0.1823, -0.1823, -0.1778, -0.1778, -0.1748, -0.1326, -0.1431, -0.1356, -0.1371, -0.1446, -0.1597, -0.1582, -0.1612, -0.1763, -0.1838, -0.1763, -0.1883, -0.1582, -0.1552, -0.1537, -0.1446, -0.1446, -0.1296, -0.11, -0.11, -0.0995, -0.1085, -0.098, -0.104, -0.1221, -0.1386, -0.1251, -0.0995, -0.1085, -0.104, -0.1266, -0.0889, -0.0769, -0.0934, -0.0788, -0.0939, -0.0879, -0.0788, -0.0333, -0.0409, -0.0652, -0.0697, -0.0848, -0.0758, -0.0515, -0.053, -0.0348, -0.0197, 0.0106, 0.0167, 0.0273, 0.0, 0.0061, 0.0, -0.0015, 0.0667, 0.0697, 0.1, 0.0818, 0.0439, 0.0561, 0.0591, 0.0424, 0.0682, 0.0879, 0.0758, 0.0773, 0.0894, 0.0848, 0.0848, 0.0606, 0.0606, 0.0682, 0.0697, 0.0818, 0.0833, 0.0652, 0.0697, 0.0545, 0.0561, 0.0242, 0.047, 0.0576, 0.0621, 0.053, 0.0621, 0.0833, 0.1, 0.1121, 0.1424, 0.1591, 0.203, 0.1848, 0.2167, 0.2167, 0.2652, 0.2833, 0.2939, 0.3045, 0.2621, 0.2814, 0.2769, 0.3088, 0.3659, 0.3659, 0.3659, 0.3659, 0.3659, 0.3659, 0.4367, 0.4199, 0.3941, 0.4458, 0.4869, 0.49, 0.493, 0.4839, 0.5006, 0.4686, 0.5036, 0.5523, 0.5143, 0.5295, 0.5843, 0.6361, 0.6908, 0.5965, 0.531, 0.5097, 0.49, 0.531, 0.5752, 0.5691, 0.5645, 0.4793, 0.531, 0.4884, 0.4686, 0.531, 0.4428, 0.4717, 0.5113, 0.5645, 0.5752, 0.5295, 0.5341, 0.5737, 0.5904, 0.5995, 0.6498, 0.6665, 0.6498, 0.6437, 0.633, 0.6574, 0.595, 0.5645, 0.6421, 0.6376, 0.6178, 0.6817, 0.6969, 0.6908, 0.6908, 0.7806, 0.8275, 0.8336, 0.8336, 0.8336, 0.9652, 1.112, 1.1242, 1.1563, 1.1227, 1.1257, 1.1227, 1.1043, 1.1456, 1.2007, 1.2771]}
{"unit_id": "quote:1964104588493246943", "ticker_idx": 0, "ts": "2025-09-05T23:13:19+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "could not only offset cost pressures but also serve as a positive catalyst to break the 'post-event share price drop' jinx and drive Apple's stock higher into year-end", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Morgan Stanley view: iPhone 17 price hike could drive AAPL ASP +5% and push stock higher into year-end.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Morgan Stanley: First Moderate Price Hike in 7 Years? All Eyes on iPhone 17 Pricing Ahead of Next Week’s Apple Launch Event\n\n• Morgan Stanley projects that Apple may implement its first “moderate price hike” in seven years by eliminating the 128GB entry-level configuration of the Pro model and launching a new ultra-thin model at a higher starting price. This move could drive iPhone average selling prices (ASP) up by +5%, well above the market consensus of +1%. If this strategy materializes, it could not only offset cost pressures but also serve as a positive catalyst to break the “post-event share price drop” jinx and drive Apple’s stock higher into year-end.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": "iPhone 17 Series Supply Chain Analysis\n\nSource: Citi https://t.co/hQYhHTFAVT", "parent_author": null, "ret_m1d": 0.0003755743740683215, "ret_signed_m1d": 0.0003755743740683215, "alpha_spy_m1d": -0.002528986153047308, "alpha_c_m1d": 0.0012124197223901678, "bench_spy_m1d": 0.0029045605271156294, "bench_c_m1d": -0.0008368453483218463, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0075513878695669945, "ret_signed_p1d": -0.0075513878695669945, "alpha_spy_p1d": -0.010007929778280467, "alpha_c_p1d": -0.015045927217617994, "bench_spy_p1d": 0.002456541908713472, "bench_c_p1d": 0.007494539348050999, "ret_p1w": -0.023446808760567484, "ret_signed_p1w": -0.023446808760567484, "alpha_spy_p1w": -0.039159699485868726, "alpha_c_p1w": -0.05395756768452831, "bench_spy_p1w": 0.015712890725301243, "bench_c_p1w": 0.030510758923960823, "ret_p1m": 0.07092507868994957, "ret_signed_p1m": 0.07092507868994957, "alpha_spy_p1m": 0.0303960838683639, "alpha_c_p1m": -0.024599669621147457, "bench_spy_p1m": 0.04052899482158567, "bench_c_p1m": 0.09552474831109703, "ret_p3m": 0.1866388219413575, "ret_signed_p3m": 0.1866388219413575, "alpha_spy_p3m": 0.12708429947347577, "alpha_c_p3m": 0.08201007283572936, "bench_spy_p3m": 0.05955452246788173, "bench_c_p3m": 0.10462874910562814, "ret_p6m": 0.10653161055586802, "ret_signed_p6m": 0.10653161055586802, "alpha_spy_p6m": 0.03997702303922068, "alpha_c_p6m": 0.04184879231159022, "bench_spy_p6m": 0.06655458751664733, "bench_c_p6m": 0.0646828182442778, "price_path": [-0.1554, -0.1716, -0.1699, -0.1813, -0.1731, -0.1847, -0.1808, -0.1808, -0.1624, -0.1603, -0.1653, -0.16, -0.1624, -0.162, -0.145, -0.1339, -0.1147, -0.1101, -0.1101, -0.1251, -0.1248, -0.1201, -0.1148, -0.12, -0.1306, -0.1286, -0.1242, -0.1248, -0.1199, -0.1145, -0.1065, -0.1076, -0.1092, -0.1087, -0.108, -0.1196, -0.1288, -0.135, -0.1566, -0.1526, -0.1544, -0.1113, -0.0831, -0.0442, -0.0522, -0.0419, -0.0265, -0.0288, -0.0338, -0.0367, -0.0381, -0.0571, -0.0617, -0.0498, -0.0523, -0.0433, -0.0384, -0.0297, -0.0315, -0.0315, -0.0416, -0.0051, 0.0004, 0.0, -0.0076, -0.0223, -0.0538, -0.0403, -0.0234, -0.0125, -0.0064, -0.0029, -0.0076, 0.0242, 0.0684, 0.0615, 0.0527, 0.0717, 0.0658, 0.0615, 0.0623, 0.0658, 0.0728, 0.0765, 0.0709, 0.07, 0.0766, 0.0599, 0.0233, 0.0333, 0.0337, 0.0403, 0.0324, 0.0526, 0.0941, 0.0963, 0.0783, 0.083, 0.0965, 0.1215, 0.1223, 0.1252, 0.1323, 0.128, 0.1225, 0.1266, 0.127, 0.1255, 0.1201, 0.1252, 0.1495, 0.142, 0.1399, 0.1376, 0.1169, 0.1169, 0.1215, 0.1119, 0.1338, 0.1523, 0.1567, 0.1591, 0.1591, 0.1645, 0.1823, 0.1952, 0.1866, 0.1722, 0.1642, 0.1605, 0.1575, 0.1642, 0.1611, 0.1621, 0.1447, 0.1468, 0.1352, 0.1367, 0.1429, 0.1316, 0.1374, 0.1435, 0.1435, 0.1417, 0.1432, 0.1404, 0.1353, 0.1353, 0.1318, 0.1161, 0.0956, 0.0872, 0.0818, 0.0832, 0.0868, 0.0902, 0.0856, 0.0783, 0.0671, 0.0671, 0.0302, 0.0342, 0.0371, 0.0358, 0.0666, 0.0786, 0.0709, 0.0786, 0.0836, 0.1276, 0.1254, 0.1546, 0.1522, 0.1615, 0.1479, 0.144, 0.1516, 0.094, 0.0692, 0.0692, 0.103, 0.105, 0.0892, 0.1059, 0.1126, 0.1375, 0.1463, 0.1409, 0.1043, 0.1065]}
{"unit_id": "quote:1958143503072108570", "ticker_idx": 0, "ts": "2025-08-20T12:26:06+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MU -5%", "tweet_type": "news_reaction", "timeline": "intraday", "summary": "Samsung HBM4 passing Nvidia testing is bearish for Micron; author flags MU down 5%.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "und", "tweet_text": "MU -5% https://t.co/PkxgJ18ttG", "reply_text": null, "parent_text": null, "quoted_text": "[Exclusive] Samsung’s HBM4 Sample Sent to Nvidia ‘Passes’… Mass Production Imminent\n\nSamsung Electronics’ 6th generation High Bandwidth Memory (HBM4), sent last month to Nvidia, the unrivaled No.1 in the AI semiconductor sector, has passed reliability testing and entered the https://t.co/15vEQCOVbA", "parent_author": null, "ret_m1d": 0.041293451682645976, "ret_signed_m1d": -0.041293451682645976, "alpha_spy_m1d": -0.038629289149342716, "alpha_c_m1d": -0.03468351256659141, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006609939116054564, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012114970846485495, "ret_signed_p1d": 0.012114970846485495, "alpha_spy_p1d": 0.00810314855892691, "alpha_c_p1d": 0.006923891451574682, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.005191079394910814, "ret_p1w": 0.004607224616834271, "ret_signed_p1w": -0.004607224616834271, "alpha_spy_p1w": 0.008744745024889422, "alpha_c_p1w": 0.025500718052461124, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.030107942669295396, "ret_p1m": 0.4409180708897755, "ret_signed_p1m": -0.4409180708897755, "alpha_spy_p1m": -0.4030718956737507, "alpha_c_p1m": -0.343984197825439, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.09693387306433654, "ret_p3m": 1.0655364305996948, "ret_signed_p3m": -1.0655364305996948, "alpha_spy_p3m": -1.0194542031942895, "alpha_c_p3m": -0.8878648521831436, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.17767157841655123, "ret_p6m": 2.535505608646391, "ret_signed_p6m": -2.535505608646391, "alpha_spy_p6m": -2.4617447977547613, "alpha_c_p6m": -2.125713114099673, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.4097924945467182, "price_path": [-0.2041, -0.2041, -0.1785, -0.1802, -0.1749, -0.1949, -0.1631, -0.1285, -0.1199, -0.094, -0.0747, -0.0543, -0.0271, -0.011, -0.0097, -0.0147, 0.0215, 0.0257, 0.0384, 0.0384, 0.0535, 0.0406, 0.0903, 0.0846, 0.074, 0.0634, 0.0505, 0.0304, 0.0377, 0.0424, 0.0424, 0.0231, 0.0615, 0.0429, 0.0503, 0.0625, 0.0119, 0.0247, -0.0067, -0.0337, -0.0241, -0.034, -0.0682, -0.063, -0.0468, -0.0508, -0.0508, -0.0448, -0.0211, -0.0689, -0.1052, -0.0805, -0.0695, -0.0719, -0.0456, 0.0143, 0.0555, 0.0899, 0.0602, 0.0689, 0.0312, 0.0541, 0.0413, 0.0, -0.0121, 0.004, -0.0067, -0.0061, 0.0046, 0.0409, 0.0154, 0.0154, 0.0108, 0.0129, 0.0597, 0.1208, 0.1216, 0.1538, 0.1944, 0.2846, 0.3414, 0.346, 0.355, 0.365, 0.4409, 0.3884, 0.4045, 0.4198, 0.3797, 0.338, 0.3418, 0.3983, 0.4275, 0.554, 0.5677, 0.6035, 0.6302, 0.5852, 0.6779, 0.6419, 0.5503, 0.6457, 0.5969, 0.6386, 0.729, 0.7277, 0.7652, 0.727, 0.6943, 0.7647, 0.8698, 0.879, 0.8945, 0.9347, 0.9124, 0.9103, 1.0036, 0.8613, 1.0275, 1.0346, 1.0311, 1.1624, 1.0584, 1.0907, 1.0229, 1.1072, 1.0655, 0.9507, 0.9287, 0.7191, 0.7703, 0.9117, 0.9168, 0.9657, 0.9657, 1.0188, 1.0528, 1.0445, 0.999, 0.9349, 1.0252, 1.108, 1.1549, 1.2513, 1.2065, 1.0586, 1.0275, 0.9849, 0.9253, 1.1219, 1.2702, 1.3613, 1.3585, 1.4474, 1.4474, 1.4313, 1.5141, 1.4992, 1.4375, 1.4375, 1.6938, 1.6659, 1.9331, 1.8999, 1.7929, 1.9472, 1.9539, 1.8878, 1.847, 1.875, 2.0981, 2.0981, 2.1173, 2.3232, 2.3955, 2.4132, 2.323, 2.5036, 2.7175, 2.7219, 2.5433, 2.739, 2.5822, 2.2403, 2.2701, 2.3708, 2.2753, 2.1877, 2.5045, 2.5355]}
{"unit_id": "orig:1947797264870871356", "ticker_idx": 0, "ts": "2025-07-22T23:13:50+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix has been preferred for the long term", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM Asia Tech marketing review: long-term preference for SK Hynix, growing interest in Samsung on 1cnm yield; TSMC confirms AI demand; HBM supply tight; NAND price stabilization expected; high interest in Hanmi and Leeno.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Tech Marketing Review (JPM)\n\n• Recently held meetings with over 45 institutional investors during Hong Kong marketing.\n\n• Overall interest was concentrated in the order of Memory > Holding Companies > EV Batteries & Materials = Hardware.\n\n• Investor confidence in the trend of AI hardware spending has increased compared to three months ago.\n\n• Confirmation of increased server build-up due to easing supply chain bottlenecks.\n\n• Rising demand confidence as AI use cases and adoption increase, particularly in the U.S. market.\n\n• TSMC's performance and outlook reconfirmed strong AI semiconductor demand.\n\n• MSFT's CAPEX announcement on July 31st mentioned as a future market catalyst.\n\n• In HBM4 price discussions, investors did not sympathize with the bearish view of making excessive price concessions to GPU customers.\n\n• Emphasized that the HBM market supply-demand is tight, and logic die costs have significantly increased.\n\n• Pointed out that HBM4 supplier certification is still ongoing, leading to high uncertainty.\n\n• HBM4 premium expectations have fluctuated, but the market still anticipates a reasonable price premium.\n\n1. Regarding SK Hynix\n\n• Some investors are concerned that HBM4 margins are lower than HBM3E.\n\n• They consider the risk of further margin decline depending on negotiation outcomes.\n\n• Mentioned that technological leadership can maintain long-term competitiveness.\n\n2. Regarding Samsung Electronics\n\n• Interest in HBM4 certification and customer sample plans due to 1cnm DRAM technology improvements.\n\n• Noted the possibility of certification results coming out by 1Q26.\n\n• Maintained a cautious view due to past failures even after reaching the final certification stage.\n\n3. Comparison of Samsung Electronics and SK Hynix Positions\n\n• While SK Hynix has been preferred for the long term, recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment.\n\n• Opinions were also raised regarding the possibility of SK Hynix's premium narrowing in the short term.\n\n4. General Memory Prices\n\n• Feedback that the rising trend in DDR4 prices is likely to be reflected in contract prices starting from 3Q25.\n\n• Shared that PC DRAM prices have high volatility.\n\n• NAND is expected to stabilize in price due to supply discipline and increased seasonal demand.\n\n5. HBM Post-processing\n\n• Concerns regarding Hanmi Semiconductor have eased; the opinion is to closely monitor the extension of TCB lifespan and the potential introduction of HCB.\n\n6. Hardware and Components\n\n• Interest in traditional hardware and consumer cycles was low.\n\n• Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01489765662390452, "ret_signed_m1d": 0.01489765662390452, "alpha_spy_m1d": 0.015040757306392982, "alpha_c_m1d": -0.003114040010228969, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.01801169663413349, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.001862148641138761, "ret_signed_p1d": 0.001862148641138761, "alpha_spy_p1d": -0.0066453924809761755, "alpha_c_p1d": -0.002483057182762982, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.004345205823901743, "ret_p1w": -0.022346309625308813, "ret_signed_p1w": -0.022346309625308813, "alpha_spy_p1w": -0.03252339133648363, "alpha_c_p1w": -0.048733034874552605, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.026386725249243792, "ret_p1m": -0.048417033406593646, "ret_signed_p1m": -0.048417033406593646, "alpha_spy_p1m": -0.06312615875701666, "alpha_c_p1m": -0.06099716912427988, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.012580135717686236, "ret_p3m": 0.736210101510651, "ret_signed_p3m": 0.736210101510651, "alpha_spy_p3m": 0.6767819150025005, "alpha_c_p3m": 0.5348575588475868, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.20135254266306424, "ret_p6m": 1.7694749273776056, "ret_signed_p6m": 1.7694749273776056, "alpha_spy_p6m": 1.6653824925586451, "alpha_c_p6m": 1.4043756214133691, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.36509930596423645, "price_path": [-0.3371, -0.3145, -0.3234, -0.3278, -0.3401, -0.3401, -0.3085, -0.3085, -0.3085, -0.2907, -0.2925, -0.2933, -0.2751, -0.262, -0.2342, -0.2546, -0.2397, -0.2587, -0.249, -0.2546, -0.268, -0.2565, -0.2453, -0.2472, -0.2267, -0.2104, -0.2384, -0.2272, -0.2272, -0.1899, -0.1639, -0.1639, -0.1471, -0.1415, -0.1061, -0.1229, -0.1229, -0.0764, -0.0726, -0.0819, -0.0838, -0.0428, -0.0335, 0.0372, 0.0652, 0.0912, 0.0577, 0.0875, 0.0633, 0.0391, 0.0372, 0.0074, 0.0093, 0.0503, 0.0466, 0.1061, 0.0968, 0.1173, 0.1117, 0.1024, 0.0037, 0.0019, 0.0149, 0.0, 0.0019, 0.0037, -0.0093, -0.0242, -0.0223, -0.0186, 0.0186, -0.0391, -0.0391, -0.0186, -0.0372, -0.0242, -0.0447, -0.0056, 0.0019, 0.0354, 0.0298, 0.0298, -0.0037, -0.0205, -0.0484, -0.0875, -0.0652, -0.0335, -0.0261, -0.0317, 0.0014, 0.0033, -0.0452, -0.0284, -0.0209, -0.0097, 0.0201, 0.0331, 0.0742, 0.1339, 0.145, 0.2252, 0.2346, 0.298, 0.2439, 0.3259, 0.3259, 0.3092, 0.3464, 0.3334, 0.3297, 0.2551, 0.3017, 0.2961, 0.3427, 0.4751, 0.4751, 0.4751, 0.4751, 0.4751, 0.4751, 0.5963, 0.5479, 0.5348, 0.5758, 0.6877, 0.7362, 0.8108, 0.7866, 0.7959, 0.7847, 0.9022, 0.9954, 0.9432, 1.0812, 1.1185, 1.0849, 1.3125, 1.1856, 1.1595, 1.2118, 1.1633, 1.2602, 1.3087, 1.3013, 1.2826, 1.0887, 1.2602, 1.126, 1.0961, 1.1297, 0.9432, 0.9395, 0.9358, 0.9544, 1.0305, 0.9782, 1.0081, 1.0827, 1.0603, 1.023, 1.0305, 1.1536, 1.1126, 1.1909, 1.1088, 1.1312, 1.0678, 0.9782, 1.0566, 1.0603, 1.0416, 1.1648, 1.1797, 1.1947, 1.1947, 1.2357, 1.3888, 1.4298, 1.4298, 1.4298, 1.5269, 1.5978, 1.7098, 1.7695, 1.8217, 1.7769, 1.7956, 1.7545, 1.7695]}
{"unit_id": "orig:1947797264870871356", "ticker_idx": 1, "ts": "2025-07-22T23:13:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM Asia Tech marketing review: long-term preference for SK Hynix, growing interest in Samsung on 1cnm yield; TSMC confirms AI demand; HBM supply tight; NAND price stabilization expected; high interest in Hanmi and Leeno.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Tech Marketing Review (JPM)\n\n• Recently held meetings with over 45 institutional investors during Hong Kong marketing.\n\n• Overall interest was concentrated in the order of Memory > Holding Companies > EV Batteries & Materials = Hardware.\n\n• Investor confidence in the trend of AI hardware spending has increased compared to three months ago.\n\n• Confirmation of increased server build-up due to easing supply chain bottlenecks.\n\n• Rising demand confidence as AI use cases and adoption increase, particularly in the U.S. market.\n\n• TSMC's performance and outlook reconfirmed strong AI semiconductor demand.\n\n• MSFT's CAPEX announcement on July 31st mentioned as a future market catalyst.\n\n• In HBM4 price discussions, investors did not sympathize with the bearish view of making excessive price concessions to GPU customers.\n\n• Emphasized that the HBM market supply-demand is tight, and logic die costs have significantly increased.\n\n• Pointed out that HBM4 supplier certification is still ongoing, leading to high uncertainty.\n\n• HBM4 premium expectations have fluctuated, but the market still anticipates a reasonable price premium.\n\n1. Regarding SK Hynix\n\n• Some investors are concerned that HBM4 margins are lower than HBM3E.\n\n• They consider the risk of further margin decline depending on negotiation outcomes.\n\n• Mentioned that technological leadership can maintain long-term competitiveness.\n\n2. Regarding Samsung Electronics\n\n• Interest in HBM4 certification and customer sample plans due to 1cnm DRAM technology improvements.\n\n• Noted the possibility of certification results coming out by 1Q26.\n\n• Maintained a cautious view due to past failures even after reaching the final certification stage.\n\n3. Comparison of Samsung Electronics and SK Hynix Positions\n\n• While SK Hynix has been preferred for the long term, recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment.\n\n• Opinions were also raised regarding the possibility of SK Hynix's premium narrowing in the short term.\n\n4. General Memory Prices\n\n• Feedback that the rising trend in DDR4 prices is likely to be reflected in contract prices starting from 3Q25.\n\n• Shared that PC DRAM prices have high volatility.\n\n• NAND is expected to stabilize in price due to supply discipline and increased seasonal demand.\n\n5. HBM Post-processing\n\n• Concerns regarding Hanmi Semiconductor have eased; the opinion is to closely monitor the extension of TCB lifespan and the potential introduction of HCB.\n\n6. Hardware and Components\n\n• Interest in traditional hardware and consumer cycles was low.\n\n• Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02727266958722896, "ret_signed_m1d": 0.02727266958722896, "alpha_spy_m1d": 0.02741577026971742, "alpha_c_m1d": 0.017922256775820822, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.009350412811408138, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006060579938871102, "ret_signed_p1d": 0.006060579938871102, "alpha_spy_p1d": -0.0024469611832438343, "alpha_c_p1d": 0.00034209771339277495, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.005718482225478327, "ret_p1w": 0.06969684888394512, "ret_signed_p1w": 0.06969684888394512, "alpha_spy_p1w": 0.0595197671727703, "alpha_c_p1w": 0.04929590515264515, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.02040094373129997, "ret_p1m": 0.06818173383038206, "ret_signed_p1m": 0.06818173383038206, "alpha_spy_p1m": 0.05347260847995905, "alpha_c_p1m": 0.06416348810601336, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.0040182457243687075, "ret_p3m": 0.4899514528642379, "ret_signed_p3m": 0.4899514528642379, "alpha_spy_p3m": 0.4305232663560874, "alpha_c_p3m": 0.38730362546363795, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.10264782740059997, "ret_p6m": 1.1456215329995172, "ret_signed_p6m": 1.1456215329995172, "alpha_spy_p6m": 1.0415290981805567, "alpha_c_p6m": 1.0242914676183734, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.12133006538114377, "price_path": [-0.1612, -0.1612, -0.1597, -0.1597, -0.1642, -0.1642, -0.1823, -0.1823, -0.1823, -0.1778, -0.1778, -0.1748, -0.1326, -0.1431, -0.1356, -0.1371, -0.1446, -0.1597, -0.1582, -0.1612, -0.1763, -0.1838, -0.1763, -0.1883, -0.1582, -0.1552, -0.1537, -0.1446, -0.1446, -0.1296, -0.11, -0.11, -0.0995, -0.1085, -0.098, -0.104, -0.1221, -0.1386, -0.1251, -0.0995, -0.1085, -0.104, -0.1266, -0.0889, -0.0769, -0.0934, -0.0788, -0.0939, -0.0879, -0.0788, -0.0333, -0.0409, -0.0652, -0.0697, -0.0848, -0.0758, -0.0515, -0.053, -0.0348, -0.0197, 0.0106, 0.0167, 0.0273, 0.0, 0.0061, 0.0, -0.0015, 0.0667, 0.0697, 0.1, 0.0818, 0.0439, 0.0561, 0.0591, 0.0424, 0.0682, 0.0879, 0.0758, 0.0773, 0.0894, 0.0848, 0.0848, 0.0606, 0.0606, 0.0682, 0.0697, 0.0818, 0.0833, 0.0652, 0.0697, 0.0545, 0.0561, 0.0242, 0.047, 0.0576, 0.0621, 0.053, 0.0621, 0.0833, 0.1, 0.1121, 0.1424, 0.1591, 0.203, 0.1848, 0.2167, 0.2167, 0.2652, 0.2833, 0.2939, 0.3045, 0.2621, 0.2814, 0.2769, 0.3088, 0.3659, 0.3659, 0.3659, 0.3659, 0.3659, 0.3659, 0.4367, 0.4199, 0.3941, 0.4458, 0.4869, 0.49, 0.493, 0.4839, 0.5006, 0.4686, 0.5036, 0.5523, 0.5143, 0.5295, 0.5843, 0.6361, 0.6908, 0.5965, 0.531, 0.5097, 0.49, 0.531, 0.5752, 0.5691, 0.5645, 0.4793, 0.531, 0.4884, 0.4686, 0.531, 0.4428, 0.4717, 0.5113, 0.5645, 0.5752, 0.5295, 0.5341, 0.5737, 0.5904, 0.5995, 0.6498, 0.6665, 0.6498, 0.6437, 0.633, 0.6574, 0.595, 0.5645, 0.6421, 0.6376, 0.6178, 0.6817, 0.6969, 0.6908, 0.6908, 0.7806, 0.8275, 0.8336, 0.8336, 0.8336, 0.9652, 1.112, 1.1242, 1.1563, 1.1227, 1.1257, 1.1227, 1.1043, 1.1456]}
{"unit_id": "orig:1947797264870871356", "ticker_idx": 2, "ts": "2025-07-22T23:13:50+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's performance and outlook reconfirmed strong AI semiconductor demand", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM Asia Tech marketing review: long-term preference for SK Hynix, growing interest in Samsung on 1cnm yield; TSMC confirms AI demand; HBM supply tight; NAND price stabilization expected; high interest in Hanmi and Leeno.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Tech Marketing Review (JPM)\n\n• Recently held meetings with over 45 institutional investors during Hong Kong marketing.\n\n• Overall interest was concentrated in the order of Memory > Holding Companies > EV Batteries & Materials = Hardware.\n\n• Investor confidence in the trend of AI hardware spending has increased compared to three months ago.\n\n• Confirmation of increased server build-up due to easing supply chain bottlenecks.\n\n• Rising demand confidence as AI use cases and adoption increase, particularly in the U.S. market.\n\n• TSMC's performance and outlook reconfirmed strong AI semiconductor demand.\n\n• MSFT's CAPEX announcement on July 31st mentioned as a future market catalyst.\n\n• In HBM4 price discussions, investors did not sympathize with the bearish view of making excessive price concessions to GPU customers.\n\n• Emphasized that the HBM market supply-demand is tight, and logic die costs have significantly increased.\n\n• Pointed out that HBM4 supplier certification is still ongoing, leading to high uncertainty.\n\n• HBM4 premium expectations have fluctuated, but the market still anticipates a reasonable price premium.\n\n1. Regarding SK Hynix\n\n• Some investors are concerned that HBM4 margins are lower than HBM3E.\n\n• They consider the risk of further margin decline depending on negotiation outcomes.\n\n• Mentioned that technological leadership can maintain long-term competitiveness.\n\n2. Regarding Samsung Electronics\n\n• Interest in HBM4 certification and customer sample plans due to 1cnm DRAM technology improvements.\n\n• Noted the possibility of certification results coming out by 1Q26.\n\n• Maintained a cautious view due to past failures even after reaching the final certification stage.\n\n3. Comparison of Samsung Electronics and SK Hynix Positions\n\n• While SK Hynix has been preferred for the long term, recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment.\n\n• Opinions were also raised regarding the possibility of SK Hynix's premium narrowing in the short term.\n\n4. General Memory Prices\n\n• Feedback that the rising trend in DDR4 prices is likely to be reflected in contract prices starting from 3Q25.\n\n• Shared that PC DRAM prices have high volatility.\n\n• NAND is expected to stabilize in price due to supply discipline and increased seasonal demand.\n\n5. HBM Post-processing\n\n• Concerns regarding Hanmi Semiconductor have eased; the opinion is to closely monitor the extension of TCB lifespan and the potential introduction of HCB.\n\n6. Hardware and Components\n\n• Interest in traditional hardware and consumer cycles was low.\n\n• Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01769920682897208, "ret_signed_m1d": 0.01769920682897208, "alpha_spy_m1d": 0.017842307511460542, "alpha_c_m1d": -0.0003124898051614089, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.01801169663413349, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013274432415607551, "ret_signed_p1d": 0.013274432415607551, "alpha_spy_p1d": 0.004766891293492614, "alpha_c_p1d": 0.008929226591705808, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.004345205823901743, "ret_p1w": 0.0044247744133645295, "ret_signed_p1w": 0.0044247744133645295, "alpha_spy_p1w": -0.0057523072978102885, "alpha_c_p1w": -0.021961950835879263, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.026386725249243792, "ret_p1m": 0.0044247744133645295, "ret_signed_p1m": 0.0044247744133645295, "alpha_spy_p1m": -0.010284350937058484, "alpha_c_p1m": -0.008155361304321707, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.012580135717686236, "ret_p3m": 0.28831866760455704, "ret_signed_p3m": 0.28831866760455704, "alpha_spy_p3m": 0.22889048109640653, "alpha_c_p3m": 0.0869661249414928, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.20135254266306424, "ret_p6m": 0.5243920479612401, "ret_signed_p6m": 0.5243920479612401, "alpha_spy_p6m": 0.4202996131422796, "alpha_c_p6m": 0.1592927419970036, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.36509930596423645, "price_path": [-0.2386, -0.2175, -0.2095, -0.2051, -0.1999, -0.1999, -0.1628, -0.1734, -0.1893, -0.1822, -0.191, -0.1637, -0.1567, -0.1461, -0.1197, -0.125, -0.1205, -0.1329, -0.1382, -0.1276, -0.1346, -0.1329, -0.1399, -0.1496, -0.1479, -0.1479, -0.1479, -0.1664, -0.1628, -0.1276, -0.1205, -0.1232, -0.1144, -0.0791, -0.0615, -0.0752, -0.0885, -0.0929, -0.0752, -0.0664, -0.0841, -0.0664, -0.0973, -0.0708, -0.0531, -0.0487, -0.0442, -0.0619, -0.0398, -0.0398, -0.0354, -0.0398, -0.0442, -0.0442, -0.0354, -0.0265, -0.0265, -0.031, -0.0177, 0.0, 0.0, 0.0221, 0.0177, 0.0, 0.0133, 0.0133, 0.0133, 0.0133, 0.0044, 0.0221, 0.0265, 0.0265, 0.0044, 0.0177, -0.0044, 0.0442, 0.0398, 0.0442, 0.0442, 0.0619, 0.0398, 0.0442, 0.0442, 0.0487, 0.0044, 0.0177, 0.0044, 0.0354, 0.0398, 0.0531, 0.0265, 0.0265, 0.031, 0.0265, 0.0265, 0.0265, 0.0442, 0.0442, 0.0619, 0.0841, 0.0973, 0.115, 0.1106, 0.1373, 0.1239, 0.1417, 0.1239, 0.1506, 0.1906, 0.1906, 0.1728, 0.155, 0.155, 0.1595, 0.1773, 0.2128, 0.2439, 0.2439, 0.275, 0.2572, 0.2794, 0.2794, 0.2572, 0.2661, 0.3016, 0.3194, 0.2883, 0.315, 0.315, 0.2972, 0.2883, 0.2883, 0.315, 0.3105, 0.3372, 0.3372, 0.3327, 0.3416, 0.3372, 0.2972, 0.3016, 0.2972, 0.3105, 0.3016, 0.3105, 0.2972, 0.2705, 0.2839, 0.2483, 0.2395, 0.2928, 0.2306, 0.2217, 0.2572, 0.2794, 0.275, 0.2794, 0.2528, 0.2705, 0.2883, 0.2839, 0.2972, 0.3283, 0.315, 0.3372, 0.3104, 0.3194, 0.2926, 0.2792, 0.2748, 0.2748, 0.2748, 0.306, 0.3283, 0.3327, 0.3327, 0.3461, 0.3639, 0.355, 0.3818, 0.3818, 0.413, 0.4887, 0.5199, 0.4932, 0.5021, 0.4976, 0.5066, 0.5244, 0.5244]}
{"unit_id": "orig:1947797264870871356", "ticker_idx": 3, "ts": "2025-07-22T23:13:50+00:00", "symbol": "HBM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "HBM market supply-demand is tight, and logic die costs have significantly increased", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM Asia Tech marketing review: long-term preference for SK Hynix, growing interest in Samsung on 1cnm yield; TSMC confirms AI demand; HBM supply tight; NAND price stabilization expected; high interest in Hanmi and Leeno.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Tech Marketing Review (JPM)\n\n• Recently held meetings with over 45 institutional investors during Hong Kong marketing.\n\n• Overall interest was concentrated in the order of Memory > Holding Companies > EV Batteries & Materials = Hardware.\n\n• Investor confidence in the trend of AI hardware spending has increased compared to three months ago.\n\n• Confirmation of increased server build-up due to easing supply chain bottlenecks.\n\n• Rising demand confidence as AI use cases and adoption increase, particularly in the U.S. market.\n\n• TSMC's performance and outlook reconfirmed strong AI semiconductor demand.\n\n• MSFT's CAPEX announcement on July 31st mentioned as a future market catalyst.\n\n• In HBM4 price discussions, investors did not sympathize with the bearish view of making excessive price concessions to GPU customers.\n\n• Emphasized that the HBM market supply-demand is tight, and logic die costs have significantly increased.\n\n• Pointed out that HBM4 supplier certification is still ongoing, leading to high uncertainty.\n\n• HBM4 premium expectations have fluctuated, but the market still anticipates a reasonable price premium.\n\n1. Regarding SK Hynix\n\n• Some investors are concerned that HBM4 margins are lower than HBM3E.\n\n• They consider the risk of further margin decline depending on negotiation outcomes.\n\n• Mentioned that technological leadership can maintain long-term competitiveness.\n\n2. Regarding Samsung Electronics\n\n• Interest in HBM4 certification and customer sample plans due to 1cnm DRAM technology improvements.\n\n• Noted the possibility of certification results coming out by 1Q26.\n\n• Maintained a cautious view due to past failures even after reaching the final certification stage.\n\n3. Comparison of Samsung Electronics and SK Hynix Positions\n\n• While SK Hynix has been preferred for the long term, recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment.\n\n• Opinions were also raised regarding the possibility of SK Hynix's premium narrowing in the short term.\n\n4. General Memory Prices\n\n• Feedback that the rising trend in DDR4 prices is likely to be reflected in contract prices starting from 3Q25.\n\n• Shared that PC DRAM prices have high volatility.\n\n• NAND is expected to stabilize in price due to supply discipline and increased seasonal demand.\n\n5. HBM Post-processing\n\n• Concerns regarding Hanmi Semiconductor have eased; the opinion is to closely monitor the extension of TCB lifespan and the potential introduction of HCB.\n\n6. Hardware and Components\n\n• Interest in traditional hardware and consumer cycles was low.\n\n• Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026295066868389155, "ret_signed_m1d": 0.026295066868389155, "alpha_spy_m1d": 0.026438167550877616, "alpha_c_m1d": 0.008283370234255665, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.01801169663413349, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004502616975360511, "ret_signed_p1d": 0.004502616975360511, "alpha_spy_p1d": -0.0040049241467544254, "alpha_c_p1d": 0.00015741115145876793, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.004345205823901743, "ret_p1w": 0.024145818205509983, "ret_signed_p1w": 0.024145818205509983, "alpha_spy_p1w": 0.013968736494335166, "alpha_c_p1w": -0.0022409070437338086, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.026386725249243792, "ret_p1m": 0.030973245553227374, "ret_signed_p1m": 0.030973245553227374, "alpha_spy_p1m": 0.01626412020280436, "alpha_c_p1m": 0.018393109835541138, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.012580135717686236, "ret_p3m": 0.6934264418472603, "ret_signed_p3m": 0.6934264418472603, "alpha_spy_p3m": 0.6339982553391098, "alpha_c_p3m": 0.4920738991841961, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.20135254266306424, "ret_p6m": 1.656779412501202, "ret_signed_p6m": 1.656779412501202, "alpha_spy_p6m": 1.5526869776822416, "alpha_c_p6m": 1.2916801065369656, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.36509930596423645, "price_path": [-0.2634, -0.2486, -0.2548, -0.2614, -0.2668, -0.2643, -0.2508, -0.2517, -0.2515, -0.2376, -0.2305, -0.2275, -0.1878, -0.1728, -0.1659, -0.1729, -0.1626, -0.172, -0.17, -0.1797, -0.1923, -0.1954, -0.1892, -0.1846, -0.1684, -0.1601, -0.176, -0.1579, -0.1455, -0.125, -0.1005, -0.0936, -0.0772, -0.0687, -0.0476, -0.0547, -0.0625, -0.0396, -0.0323, -0.0224, -0.026, -0.0054, -0.0145, 0.0394, 0.0508, 0.0501, 0.04, 0.0403, 0.0271, 0.0246, 0.0408, 0.0284, 0.014, 0.0399, 0.027, 0.0525, 0.0618, 0.0501, 0.0589, 0.0496, 0.0171, 0.022, 0.0263, 0.0, 0.0045, 0.0089, 0.0026, 0.0203, 0.0241, 0.044, 0.0332, -0.0116, 0.0012, 0.013, 0.0004, 0.0227, 0.0439, 0.0676, 0.0829, 0.0875, 0.0873, 0.0738, 0.0627, 0.0525, 0.031, 0.0141, 0.0314, 0.0386, 0.0352, 0.0387, 0.0577, 0.0497, 0.0229, 0.0345, 0.0412, 0.0632, 0.092, 0.0996, 0.1319, 0.1719, 0.2119, 0.2691, 0.2794, 0.3184, 0.2979, 0.363, 0.3442, 0.3605, 0.3845, 0.3693, 0.3567, 0.319, 0.3613, 0.3683, 0.4398, 0.5078, 0.5206, 0.5302, 0.5141, 0.5472, 0.5344, 0.5656, 0.578, 0.5475, 0.5934, 0.6767, 0.6934, 0.7327, 0.7079, 0.7049, 0.7157, 0.8041, 0.8547, 0.8302, 0.8957, 0.9184, 0.9237, 1.0512, 0.9265, 0.9555, 0.9683, 0.9443, 1.0373, 1.031, 1.038, 1.006, 0.9431, 1.0026, 0.9026, 0.8782, 0.8352, 0.7619, 0.8209, 0.8347, 0.8762, 0.9051, 0.8914, 0.915, 0.9502, 0.932, 0.8997, 0.9512, 1.0274, 1.025, 1.0835, 1.0366, 0.9993, 0.9462, 0.891, 0.9216, 0.9917, 1.0319, 1.1268, 1.1359, 1.1707, 1.1707, 1.2085, 1.3048, 1.3152, 1.2931, 1.2931, 1.461, 1.5236, 1.6605, 1.6793, 1.6472, 1.6885, 1.6961, 1.6526, 1.6568]}
{"unit_id": "orig:1947797264870871356", "ticker_idx": 5, "ts": "2025-07-22T23:13:50+00:00", "symbol": "Leeno Industrial", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial)", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM Asia Tech marketing review: long-term preference for SK Hynix, growing interest in Samsung on 1cnm yield; TSMC confirms AI demand; HBM supply tight; NAND price stabilization expected; high interest in Hanmi and Leeno.", "resolved_tickers": ["058610.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "Leeno Industrial 058610.KQ Korea KOSDAQ test sockets", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Asia Tech Marketing Review (JPM)\n\n• Recently held meetings with over 45 institutional investors during Hong Kong marketing.\n\n• Overall interest was concentrated in the order of Memory > Holding Companies > EV Batteries & Materials = Hardware.\n\n• Investor confidence in the trend of AI hardware spending has increased compared to three months ago.\n\n• Confirmation of increased server build-up due to easing supply chain bottlenecks.\n\n• Rising demand confidence as AI use cases and adoption increase, particularly in the U.S. market.\n\n• TSMC's performance and outlook reconfirmed strong AI semiconductor demand.\n\n• MSFT's CAPEX announcement on July 31st mentioned as a future market catalyst.\n\n• In HBM4 price discussions, investors did not sympathize with the bearish view of making excessive price concessions to GPU customers.\n\n• Emphasized that the HBM market supply-demand is tight, and logic die costs have significantly increased.\n\n• Pointed out that HBM4 supplier certification is still ongoing, leading to high uncertainty.\n\n• HBM4 premium expectations have fluctuated, but the market still anticipates a reasonable price premium.\n\n1. Regarding SK Hynix\n\n• Some investors are concerned that HBM4 margins are lower than HBM3E.\n\n• They consider the risk of further margin decline depending on negotiation outcomes.\n\n• Mentioned that technological leadership can maintain long-term competitiveness.\n\n2. Regarding Samsung Electronics\n\n• Interest in HBM4 certification and customer sample plans due to 1cnm DRAM technology improvements.\n\n• Noted the possibility of certification results coming out by 1Q26.\n\n• Maintained a cautious view due to past failures even after reaching the final certification stage.\n\n3. Comparison of Samsung Electronics and SK Hynix Positions\n\n• While SK Hynix has been preferred for the long term, recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment.\n\n• Opinions were also raised regarding the possibility of SK Hynix's premium narrowing in the short term.\n\n4. General Memory Prices\n\n• Feedback that the rising trend in DDR4 prices is likely to be reflected in contract prices starting from 3Q25.\n\n• Shared that PC DRAM prices have high volatility.\n\n• NAND is expected to stabilize in price due to supply discipline and increased seasonal demand.\n\n5. HBM Post-processing\n\n• Concerns regarding Hanmi Semiconductor have eased; the opinion is to closely monitor the extension of TCB lifespan and the potential introduction of HCB.\n\n6. Hardware and Components\n\n• Interest in traditional hardware and consumer cycles was low.\n\n• Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020912542276398716, "ret_signed_m1d": 0.020912542276398716, "alpha_spy_m1d": 0.021055642958887177, "alpha_c_m1d": 0.007467613253196781, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.013444929023201935, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011406834453921877, "ret_signed_p1d": -0.011406834453921877, "alpha_spy_p1d": -0.019914375576036814, "alpha_c_p1d": -0.024577361881827575, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.013170527427905698, "ret_p1w": -0.04372621118424236, "ret_signed_p1w": -0.04372621118424236, "alpha_spy_p1w": -0.053903292895417176, "alpha_c_p1w": -0.05195772393861897, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.008231512754376613, "ret_p1m": -0.11977187376405674, "ret_signed_p1m": -0.11977187376405674, "alpha_spy_p1m": -0.13448099911447975, "alpha_c_p1m": -0.09878132895381608, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": -0.020990544810240652, "ret_p3m": 0.1772615040672776, "ret_signed_p3m": 0.1772615040672776, "alpha_spy_p3m": 0.11783331755912707, "alpha_c_p3m": -0.03717114227390783, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.21443264634118542, "ret_p6m": 2.3297118197556936, "ret_signed_p6m": 2.3297118197556936, "alpha_spy_p6m": 2.225619384936733, "alpha_c_p6m": 1.7917428902782948, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.5379689294773988, "price_path": [-0.019, -0.0209, -0.0456, -0.0209, -0.0494, -0.0494, -0.0437, -0.0437, -0.0437, -0.0133, -0.0228, -0.0494, -0.0342, -0.038, -0.0019, 0.0266, -0.0019, -0.0551, -0.0722, -0.0779, -0.0913, -0.1084, -0.0817, -0.097, -0.097, -0.0932, -0.0875, -0.0494, -0.0494, 0.0057, 0.019, 0.019, 0.0057, -0.0114, -0.0038, -0.0057, -0.0285, -0.0266, -0.0228, -0.0171, -0.0095, 0.0038, 0.0494, 0.0951, 0.0627, 0.0285, 0.0475, 0.0494, 0.0494, 0.038, 0.0475, 0.0114, -0.0038, 0.0, 0.019, 0.0209, 0.0057, -0.0038, 0.019, 0.0266, 0.0114, 0.0076, 0.0209, 0.0, -0.0114, -0.0266, -0.0342, -0.0361, -0.0437, -0.0323, -0.0475, -0.0856, -0.0779, -0.0494, -0.0342, -0.0323, -0.038, -0.0323, -0.0418, -0.0646, -0.0627, -0.0627, -0.0875, -0.1027, -0.1198, -0.1141, -0.1027, -0.0209, -0.0057, -0.0171, -0.0266, -0.0361, -0.0304, 0.0418, 0.019, 0.0266, 0.0266, 0.0456, 0.0475, 0.0475, 0.0798, 0.0798, 0.0875, 0.0475, 0.0266, 0.0646, 0.0646, 0.0856, 0.0646, 0.0228, 0.0342, 0.0, 0.0076, 0.0513, 0.0418, 0.0303, 0.0303, 0.0303, 0.0303, 0.0303, 0.0303, 0.1639, 0.1696, 0.1238, 0.2192, 0.2421, 0.1773, 0.3089, 0.2669, 0.2345, 0.2097, 0.1982, 0.1982, 0.204, 0.2021, 0.1868, 0.2898, 0.3623, 0.3604, 0.2688, 0.2307, 0.183, 0.204, 0.2078, 0.1906, 0.2021, 0.1334, 0.1792, 0.1544, 0.4635, 0.6218, 0.7592, 0.7592, 0.8031, 0.7058, 0.9061, 1.4041, 1.3927, 1.4652, 1.3927, 1.4728, 1.5415, 1.8621, 1.8659, 1.7819, 1.7896, 1.8621, 2.0071, 2.1254, 1.7323, 1.6713, 2.4268, 2.4841, 2.2895, 2.2055, 2.2055, 2.2284, 2.2838, 2.108, 2.108, 2.108, 2.1615, 2.0583, 1.8748, 2.0583, 1.9054, 2.0201, 2.1577, 2.4788, 2.3297]}
{"unit_id": "orig:1947797264870871356", "ticker_idx": 6, "ts": "2025-07-22T23:13:50+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND is expected to stabilize in price due to supply discipline and increased seasonal demand", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM Asia Tech marketing review: long-term preference for SK Hynix, growing interest in Samsung on 1cnm yield; TSMC confirms AI demand; HBM supply tight; NAND price stabilization expected; high interest in Hanmi and Leeno.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Tech Marketing Review (JPM)\n\n• Recently held meetings with over 45 institutional investors during Hong Kong marketing.\n\n• Overall interest was concentrated in the order of Memory > Holding Companies > EV Batteries & Materials = Hardware.\n\n• Investor confidence in the trend of AI hardware spending has increased compared to three months ago.\n\n• Confirmation of increased server build-up due to easing supply chain bottlenecks.\n\n• Rising demand confidence as AI use cases and adoption increase, particularly in the U.S. market.\n\n• TSMC's performance and outlook reconfirmed strong AI semiconductor demand.\n\n• MSFT's CAPEX announcement on July 31st mentioned as a future market catalyst.\n\n• In HBM4 price discussions, investors did not sympathize with the bearish view of making excessive price concessions to GPU customers.\n\n• Emphasized that the HBM market supply-demand is tight, and logic die costs have significantly increased.\n\n• Pointed out that HBM4 supplier certification is still ongoing, leading to high uncertainty.\n\n• HBM4 premium expectations have fluctuated, but the market still anticipates a reasonable price premium.\n\n1. Regarding SK Hynix\n\n• Some investors are concerned that HBM4 margins are lower than HBM3E.\n\n• They consider the risk of further margin decline depending on negotiation outcomes.\n\n• Mentioned that technological leadership can maintain long-term competitiveness.\n\n2. Regarding Samsung Electronics\n\n• Interest in HBM4 certification and customer sample plans due to 1cnm DRAM technology improvements.\n\n• Noted the possibility of certification results coming out by 1Q26.\n\n• Maintained a cautious view due to past failures even after reaching the final certification stage.\n\n3. Comparison of Samsung Electronics and SK Hynix Positions\n\n• While SK Hynix has been preferred for the long term, recent improvements in Samsung Electronics' 1cnm yield have led to considerations for portfolio adjustment.\n\n• Opinions were also raised regarding the possibility of SK Hynix's premium narrowing in the short term.\n\n4. General Memory Prices\n\n• Feedback that the rising trend in DDR4 prices is likely to be reflected in contract prices starting from 3Q25.\n\n• Shared that PC DRAM prices have high volatility.\n\n• NAND is expected to stabilize in price due to supply discipline and increased seasonal demand.\n\n5. HBM Post-processing\n\n• Concerns regarding Hanmi Semiconductor have eased; the opinion is to closely monitor the extension of TCB lifespan and the potential introduction of HCB.\n\n6. Hardware and Components\n\n• Interest in traditional hardware and consumer cycles was low.\n\n• Relatively high interest in AI-centric supply chains (Hanmi Semiconductor, Leeno Industrial).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013232809982348371, "ret_signed_m1d": 0.013232809982348371, "alpha_spy_m1d": 0.013375910664836833, "alpha_c_m1d": -0.0047788866517851185, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.01801169663413349, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02205812311202191, "ret_signed_p1d": 0.02205812311202191, "alpha_spy_p1d": 0.013550581989906975, "alpha_c_p1d": 0.017712917288120168, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.004345205823901743, "ret_p1w": 0.02458145061535728, "ret_signed_p1w": 0.02458145061535728, "alpha_spy_p1w": 0.014404368904182462, "alpha_c_p1w": -0.0018052746338865118, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.026386725249243792, "ret_p1m": 0.029948031537860832, "ret_signed_p1m": 0.029948031537860832, "alpha_spy_p1m": 0.015238906187437819, "alpha_c_p1m": 0.017367895820174596, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.012580135717686236, "ret_p3m": 1.2409347612505308, "ret_signed_p3m": 1.2409347612505308, "alpha_spy_p3m": 1.1815065747423803, "alpha_c_p3m": 1.0395822185874666, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.20135254266306424, "ret_p6m": 3.5769474014662253, "ret_signed_p6m": 3.5769474014662253, "alpha_spy_p6m": 3.472854966647265, "alpha_c_p6m": 3.211848095501989, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.36509930596423645, "price_path": [-0.2507, -0.2344, -0.2396, -0.2403, -0.2501, -0.2481, -0.2319, -0.2315, -0.2351, -0.2134, -0.1998, -0.1834, -0.1279, -0.1165, -0.1075, -0.123, -0.1168, -0.1327, -0.1254, -0.1399, -0.1593, -0.1653, -0.16, -0.1533, -0.1395, -0.1276, -0.1483, -0.1478, -0.1352, -0.1107, -0.0947, -0.0863, -0.068, -0.0623, -0.0513, -0.059, -0.0644, -0.0424, -0.0324, -0.0173, -0.0148, 0.0125, 0.0314, 0.0652, 0.0699, 0.0705, 0.0625, 0.0522, 0.0307, 0.0394, 0.0614, 0.0399, 0.0323, 0.0602, 0.0654, 0.0718, 0.0675, 0.0387, 0.0462, 0.0343, 0.0124, 0.0134, 0.0132, 0.0, 0.0221, 0.0227, 0.0119, 0.0194, 0.0246, 0.0407, 0.0342, -0.005, 0.005, 0.0082, -0.0016, 0.0022, 0.0379, 0.0475, 0.0946, 0.0948, 0.0912, 0.0732, 0.0614, 0.0533, 0.0299, 0.0207, 0.0435, 0.0489, 0.053, 0.0589, 0.0867, 0.1, 0.0906, 0.0858, 0.0998, 0.1592, 0.2423, 0.2634, 0.2849, 0.3512, 0.4688, 0.5482, 0.5736, 0.6261, 0.6034, 0.6729, 0.6786, 0.716, 0.7472, 0.7172, 0.6778, 0.6276, 0.7555, 0.7701, 0.842, 0.9562, 1.0487, 1.0311, 0.9854, 1.0565, 1.0665, 1.0184, 1.1115, 1.0445, 1.1809, 1.2775, 1.2409, 1.3518, 1.3338, 1.3432, 1.4519, 1.715, 1.7845, 1.7292, 1.9663, 2.007, 2.0209, 2.1346, 1.9942, 2.0997, 2.1444, 2.3288, 2.629, 2.6323, 2.696, 2.4677, 2.231, 2.4185, 2.1817, 2.2199, 1.9941, 1.8621, 2.0266, 1.9888, 1.865, 1.9376, 1.999, 1.9061, 1.9318, 1.8507, 1.9277, 2.062, 2.1533, 2.0997, 2.166, 2.1918, 2.0214, 1.9112, 1.8713, 1.9288, 2.0494, 2.1471, 2.2862, 2.2958, 2.3962, 2.412, 2.4863, 2.4543, 2.421, 2.394, 2.394, 2.6778, 2.7861, 3.2545, 3.3756, 3.2903, 3.4965, 3.5584, 3.6179, 3.5769]}
{"unit_id": "thread:1959877363006861399", "ticker_idx": 0, "ts": "2025-08-25T07:15:50+00:00", "symbol": "AMEC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This will be a major blow to AMEC", "tweet_type": "analysis", "timeline": "unspecified", "summary": "TSMC's decision to exclude Chinese equipment from Taiwan fabs (to qualify for US subsidies) is a major blow to AMEC.", "resolved_tickers": ["688012.SS"], "resolution_method": "exchange_ticker", "resolution_reason": "AMEC = Advanced Micro-Fabrication Equipment, Shanghai STAR Market", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Interesting.\n\nNikkei reported that in order to receive U.S. subsidies, TSMC decided not to use Chinese equipment at its Taiwan fabs, rather than removing Chinese equipment from its U.S. fabs.\n\nThis will be a major blow to AMEC. https://t.co/OO9hpqRD4E\n\n---\n\nhttps://t.co/Dh9jheYkpk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.022748847886497447, "ret_signed_m1d": -0.022748847886497447, "alpha_spy_m1d": -0.018328235017351036, "alpha_c_m1d": -0.023701531774172557, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014170672454562716, "ret_signed_p1d": -0.014170672454562716, "alpha_spy_p1d": -0.0099836361046719, "alpha_c_p1d": -0.004542284786897888, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": 0.05781991646098339, "ret_signed_p1w": -0.05781991646098339, "alpha_spy_p1w": -0.053804086526149986, "alpha_c_p1w": -0.07017022833554698, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.3270142568652157, "ret_signed_p1m": -0.3270142568652157, "alpha_spy_p1m": -0.29187047557178225, "alpha_c_p1m": -0.23249878728694218, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.3438388896560056, "ret_signed_p3m": -0.3438388896560056, "alpha_spy_p3m": -0.32536459653395267, "alpha_c_p3m": -0.23775571088513958, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.63270146883684, "ret_signed_p6m": -0.63270146883684, "alpha_spy_p6m": -0.5637542407472611, "alpha_c_p6m": -0.24192172410237922, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1912, -0.1639, -0.1719, -0.1719, -0.1535, -0.1511, -0.1483, -0.1483, -0.154, -0.1895, -0.1962, -0.2147, -0.1872, -0.1806, -0.191, -0.1869, -0.1819, -0.188, -0.1548, -0.1523, -0.1445, -0.1531, -0.1493, -0.1374, -0.1351, -0.1541, -0.1577, -0.1438, -0.1457, -0.127, -0.1372, -0.1442, -0.1262, -0.1517, -0.147, -0.155, -0.1555, -0.1363, -0.1294, -0.1114, -0.0592, -0.0568, -0.0527, -0.0518, -0.0218, -0.0342, -0.0634, -0.0876, -0.072, -0.0749, -0.0708, -0.07, -0.0714, -0.0834, -0.0869, -0.0877, -0.0899, -0.0796, -0.0784, -0.0919, -0.0616, -0.0407, 0.0227, 0.0, 0.0142, -0.0189, 0.0687, 0.015, 0.0578, 0.0072, 0.0047, -0.0739, -0.0521, 0.0081, -0.0314, -0.0256, 0.0038, 0.0024, 0.0194, 0.0142, 0.0768, 0.1999, 0.2038, 0.2161, 0.327, 0.3768, 0.3649, 0.3744, 0.4455, 0.417, 0.417, 0.417, 0.417, 0.417, 0.417, 0.417, 0.5124, 0.3886, 0.3986, 0.2962, 0.2986, 0.3005, 0.2413, 0.2456, 0.2781, 0.2644, 0.2664, 0.3335, 0.3999, 0.4005, 0.406, 0.422, 0.324, 0.3171, 0.3879, 0.391, 0.4502, 0.4651, 0.4997, 0.4507, 0.4348, 0.4385, 0.3935, 0.3706, 0.4212, 0.4147, 0.3438, 0.2654, 0.2848, 0.2769, 0.2808, 0.2592, 0.2706, 0.2418, 0.2341, 0.2298, 0.2818, 0.2978, 0.3093, 0.3104, 0.2976, 0.2703, 0.3248, 0.3074, 0.2915, 0.3231, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.2925, 0.4755, 0.5624, 0.6699, 0.6392, 0.5956, 0.6715, 0.6398, 0.67, 0.7583, 0.7874, 0.7962, 0.7266, 0.7299, 0.6752, 0.7419, 0.6735, 0.6923, 0.7469, 0.6464, 0.6491, 0.5767, 0.6415, 0.655, 0.6914, 0.655, 0.6687, 0.6611, 0.6439, 0.6351, 0.6327, 0.6327, 0.6327]}
{"unit_id": "orig:1970351707323596814", "ticker_idx": 0, "ts": "2025-09-23T04:57:09+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung up 5% on expectation for NVDA HBM entry", "tweet_type": "analysis", "timeline": "months", "summary": "Relays Citi checks: Hynix dominant HBM4 share in 2026 (long); buyside turning cautious on MU into print given high EPS bar and HBM supply concerns.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Citi’s note on Samsung, SK Hynix, and Micron:\n\nSamsung denied the press report and our checks (Peter & me) indicate there's no firm orders yet from NVDA. However Samsung up 5% on expectation for NVDA HBM entry. Looking back over the last 3 years, this is probably 10-11th of qualification false alarm but buyside continues to view HBM oversupply as a base case on the back of Samsung's entry simply because of huge new supply addition. Our checks confirms 1) NVDA pulling in Rubin schedule so asking for early shipment of HBM4 from Hynix and 2) Hynix preparing HBM4 mass production in 2 weeks and I reckon there could be potential announcement or event for successful HBM4 production given numerous noise from the market (mgmt is well aware of this) and confidence for dominant m/s in 2026 amid higher HBM4 spec. Also don't forget below local feedback on 28b GB Socamm orders for 2026E vs. CITI 25b / and iPhone probably total 16-18b.\n\nMU buyside bar is high since a lot of you said MU Nov Q eps is close to 3.6-3.7 and they won't be able to talk about HBM sold out in 2026. I can sense that buyside turning more cautious (long Samsung vs. short MU, or long Semicap vs. short MU) into the print.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014167682693315031, "ret_signed_m1d": -0.014167682693315031, "alpha_spy_m1d": -0.01964095113216091, "alpha_c_m1d": -0.02287516051591476, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.00870747782259973, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00826445047379698, "ret_signed_p1d": 0.00826445047379698, "alpha_spy_p1d": 0.011446083926410666, "alpha_c_p1d": 0.014438117118731553, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.006173666644934572, "ret_p1w": -0.00502563521762156, "ret_signed_p1w": -0.00502563521762156, "alpha_spy_p1w": -0.009503763940313603, "alpha_c_p1w": -0.010878234503607032, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.005852599285985471, "ret_p1m": 0.1693023341729294, "ret_signed_p1m": 0.1693023341729294, "alpha_spy_p1m": 0.16238152247137472, "alpha_c_p1m": 0.150495649812612, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.018806684360317405, "ret_p3m": 0.2606170159033787, "ret_signed_p3m": 0.2606170159033787, "alpha_spy_p3m": 0.23137886329066948, "alpha_c_p3m": 0.22857067124980013, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.032046344653578585, "ret_p6m": 1.4846316970647475, "ret_signed_p6m": 1.4846316970647475, "alpha_spy_p6m": 1.4843687513864, "alpha_c_p6m": 1.4984832418519514, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": -0.013851544787203851, "price_path": [-0.2936, -0.2822, -0.294, -0.2893, -0.2822, -0.2468, -0.2527, -0.2715, -0.2751, -0.2869, -0.2798, -0.2609, -0.2621, -0.2479, -0.2361, -0.2125, -0.2078, -0.1995, -0.2208, -0.2161, -0.2208, -0.222, -0.1688, -0.1665, -0.1429, -0.157, -0.1865, -0.1771, -0.1747, -0.1877, -0.1677, -0.1523, -0.1617, -0.1606, -0.1511, -0.1547, -0.1547, -0.1736, -0.1736, -0.1677, -0.1665, -0.157, -0.1558, -0.17, -0.1665, -0.1783, -0.1771, -0.2019, -0.1842, -0.1759, -0.1724, -0.1795, -0.1724, -0.1558, -0.1429, -0.1334, -0.1098, -0.0968, -0.0626, -0.0767, -0.0519, -0.0519, -0.0142, 0.0, 0.0083, 0.0165, -0.0165, -0.0015, -0.005, 0.0199, 0.0643, 0.0643, 0.0643, 0.0643, 0.0643, 0.0643, 0.1195, 0.1064, 0.0863, 0.1266, 0.1586, 0.161, 0.1634, 0.1563, 0.1693, 0.1444, 0.1717, 0.2096, 0.18, 0.1918, 0.2345, 0.2748, 0.3175, 0.244, 0.193, 0.1764, 0.161, 0.193, 0.2274, 0.2227, 0.2191, 0.1527, 0.193, 0.1598, 0.1444, 0.193, 0.1242, 0.1468, 0.1776, 0.2191, 0.2274, 0.1918, 0.1954, 0.2262, 0.2393, 0.2464, 0.2855, 0.2986, 0.2855, 0.2808, 0.2725, 0.2915, 0.2428, 0.2191, 0.2796, 0.276, 0.2606, 0.3104, 0.3223, 0.3175, 0.3175, 0.3875, 0.424, 0.4288, 0.4288, 0.4288, 0.5313, 0.6457, 0.6552, 0.6803, 0.654, 0.6564, 0.654, 0.6397, 0.6719, 0.7148, 0.7744, 0.7792, 0.7303, 0.7815, 0.8149, 0.8125, 0.8125, 0.9007, 0.9353, 0.915, 0.9126, 0.7923, 0.996, 1.0151, 0.8983, 0.89, 0.9829, 0.9758, 0.9996, 1.1283, 1.1593, 1.1593, 1.1593, 1.1593, 1.2642, 1.2654, 1.2999, 1.3833, 1.425, 1.5978, 1.58, 1.58, 1.3249, 1.0521, 1.2832, 1.2427, 1.0675, 1.2391, 1.2642, 1.2391, 1.1867, 1.2487, 1.3106, 1.4846]}
{"unit_id": "orig:1970351707323596814", "ticker_idx": 1, "ts": "2025-09-23T04:57:09+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVDA pulling in Rubin schedule so asking for early shipment of HBM4 from Hynix and 2) Hynix preparing HBM4 mass production in 2 weeks… confidence for dominant m/s in 2026 amid higher HBM4 spec", "tweet_type": "analysis", "timeline": "months", "summary": "Relays Citi checks: Hynix dominant HBM4 share in 2026 (long); buyside turning cautious on MU into print given high EPS bar and HBM supply concerns.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi’s note on Samsung, SK Hynix, and Micron:\n\nSamsung denied the press report and our checks (Peter & me) indicate there's no firm orders yet from NVDA. However Samsung up 5% on expectation for NVDA HBM entry. Looking back over the last 3 years, this is probably 10-11th of qualification false alarm but buyside continues to view HBM oversupply as a base case on the back of Samsung's entry simply because of huge new supply addition. Our checks confirms 1) NVDA pulling in Rubin schedule so asking for early shipment of HBM4 from Hynix and 2) Hynix preparing HBM4 mass production in 2 weeks and I reckon there could be potential announcement or event for successful HBM4 production given numerous noise from the market (mgmt is well aware of this) and confidence for dominant m/s in 2026 amid higher HBM4 spec. Also don't forget below local feedback on 28b GB Socamm orders for 2026E vs. CITI 25b / and iPhone probably total 16-18b.\n\nMU buyside bar is high since a lot of you said MU Nov Q eps is close to 3.6-3.7 and they won't be able to talk about HBM sold out in 2026. I can sense that buyside turning more cautious (long Samsung vs. short MU, or long Semicap vs. short MU) into the print.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027700775241169207, "ret_signed_m1d": -0.027700775241169207, "alpha_spy_m1d": -0.033174043680015086, "alpha_c_m1d": -0.029161745876449152, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009695288694700488, "ret_signed_p1d": -0.009695288694700488, "alpha_spy_p1d": -0.006513655242086802, "alpha_c_p1d": -0.007861201022949849, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.037396063935869694, "ret_signed_p1w": -0.037396063935869694, "alpha_spy_p1w": -0.04187419265856174, "alpha_c_p1w": -0.05188149310182233, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.33379508814861514, "ret_signed_p1m": 0.33379508814861514, "alpha_spy_p1m": 0.32687427644706046, "alpha_c_p1m": 0.2812617129201036, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.5163205834724471, "ret_signed_p3m": 0.5163205834724471, "alpha_spy_p3m": 0.4870824308597379, "alpha_c_p3m": 0.4089845848724212, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 1.9327045667287934, "ret_signed_p6m": 1.9327045667287934, "alpha_spy_p6m": 1.9324416210504458, "alpha_c_p6m": 1.7051791110782875, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.1895, -0.2144, -0.1923, -0.2103, -0.2283, -0.2296, -0.2518, -0.2504, -0.22, -0.2227, -0.1785, -0.1854, -0.1702, -0.1743, -0.1812, -0.2545, -0.2559, -0.2462, -0.2573, -0.2559, -0.2545, -0.2642, -0.2753, -0.2739, -0.2711, -0.2435, -0.2863, -0.2863, -0.2711, -0.285, -0.2753, -0.2905, -0.2615, -0.2559, -0.231, -0.2352, -0.2352, -0.2601, -0.2725, -0.2933, -0.3223, -0.3057, -0.2822, -0.2767, -0.2808, -0.2562, -0.2548, -0.2909, -0.2784, -0.2729, -0.2645, -0.2424, -0.2327, -0.2022, -0.1579, -0.1496, -0.09, -0.0831, -0.036, -0.0762, -0.0152, -0.0152, -0.0277, 0.0, -0.0097, -0.0125, -0.0679, -0.0332, -0.0374, -0.0028, 0.0956, 0.0956, 0.0956, 0.0956, 0.0956, 0.0956, 0.1856, 0.1496, 0.1399, 0.1704, 0.2535, 0.2895, 0.3449, 0.3269, 0.3338, 0.3255, 0.4127, 0.482, 0.4432, 0.5457, 0.5734, 0.5485, 0.7175, 0.6233, 0.6039, 0.6427, 0.6066, 0.6787, 0.7147, 0.7091, 0.6953, 0.5512, 0.6787, 0.5789, 0.5568, 0.5817, 0.4432, 0.4404, 0.4377, 0.4515, 0.508, 0.4692, 0.4914, 0.5468, 0.5302, 0.5025, 0.508, 0.5995, 0.569, 0.6272, 0.5662, 0.5829, 0.5357, 0.4692, 0.5274, 0.5302, 0.5163, 0.6078, 0.6189, 0.63, 0.63, 0.6605, 0.7741, 0.8046, 0.8046, 0.8046, 0.8767, 0.9294, 1.0125, 1.0569, 1.0957, 1.0624, 1.0763, 1.0458, 1.0569, 1.0763, 1.0957, 1.1179, 1.0596, 1.0513, 1.0929, 1.1262, 1.0402, 1.2177, 1.3313, 1.3867, 1.5198, 1.3008, 1.5143, 1.4949, 1.3341, 1.3258, 1.4588, 1.4283, 1.384, 1.4616, 1.4394, 1.4394, 1.4394, 1.4394, 1.4782, 1.6307, 1.6362, 1.7859, 1.822, 2.0521, 1.9466, 1.9466, 1.6078, 1.3578, 1.6133, 1.5661, 1.3217, 1.605, 1.6522, 1.5828, 1.5272, 1.705, 1.6939, 1.9327]}
{"unit_id": "orig:1970351707323596814", "ticker_idx": 2, "ts": "2025-09-23T04:57:09+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "buyside turning more cautious (long Samsung vs. short MU, or long Semicap vs. short MU) into the print", "tweet_type": "analysis", "timeline": "months", "summary": "Relays Citi checks: Hynix dominant HBM4 share in 2026 (long); buyside turning cautious on MU into print given high EPS bar and HBM supply concerns.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi’s note on Samsung, SK Hynix, and Micron:\n\nSamsung denied the press report and our checks (Peter & me) indicate there's no firm orders yet from NVDA. However Samsung up 5% on expectation for NVDA HBM entry. Looking back over the last 3 years, this is probably 10-11th of qualification false alarm but buyside continues to view HBM oversupply as a base case on the back of Samsung's entry simply because of huge new supply addition. Our checks confirms 1) NVDA pulling in Rubin schedule so asking for early shipment of HBM4 from Hynix and 2) Hynix preparing HBM4 mass production in 2 weeks and I reckon there could be potential announcement or event for successful HBM4 production given numerous noise from the market (mgmt is well aware of this) and confidence for dominant m/s in 2026 amid higher HBM4 spec. Also don't forget below local feedback on 28b GB Socamm orders for 2026E vs. CITI 25b / and iPhone probably total 16-18b.\n\nMU buyside bar is high since a lot of you said MU Nov Q eps is close to 3.6-3.7 and they won't be able to talk about HBM sold out in 2026. I can sense that buyside turning more cautious (long Samsung vs. short MU, or long Semicap vs. short MU) into the print.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010756522195869866, "ret_signed_m1d": 0.010756522195869866, "alpha_spy_m1d": 0.016229790634715746, "alpha_c_m1d": 0.012217492831149812, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02824341205600267, "ret_signed_p1d": 0.02824341205600267, "alpha_spy_p1d": 0.025061778603388984, "alpha_c_p1d": 0.02640932438425203, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.0054684811190846006, "ret_signed_p1w": -0.0054684811190846006, "alpha_spy_p1w": -0.0009903523963925576, "alpha_c_p1w": 0.009016948046868034, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.19340362628470564, "ret_signed_p1m": -0.19340362628470564, "alpha_spy_p1m": -0.18648281458315097, "alpha_c_p1m": -0.1408702510561941, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.5989816019108829, "ret_signed_p3m": -0.5989816019108829, "alpha_spy_p3m": -0.5697434492981737, "alpha_c_p3m": -0.49164560331085694, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 1.7775123773604586, "ret_signed_p6m": -1.7775123773604586, "alpha_spy_p6m": -1.777249431682111, "alpha_c_p6m": -1.5499869217099527, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.2435, -0.251, -0.2601, -0.2742, -0.2691, -0.2658, -0.2658, -0.2794, -0.2523, -0.2654, -0.2602, -0.2517, -0.2872, -0.2782, -0.3003, -0.3194, -0.3126, -0.3196, -0.3437, -0.34, -0.3286, -0.3314, -0.3315, -0.3272, -0.3105, -0.3442, -0.3697, -0.3524, -0.3446, -0.3463, -0.3277, -0.2856, -0.2565, -0.2323, -0.2532, -0.2471, -0.2737, -0.2576, -0.2666, -0.2957, -0.3042, -0.2928, -0.3004, -0.2999, -0.2924, -0.2669, -0.2848, -0.2848, -0.288, -0.2866, -0.2536, -0.2106, -0.21, -0.1873, -0.1587, -0.0952, -0.0552, -0.0519, -0.0456, -0.0386, 0.0149, -0.0221, -0.0108, 0.0, -0.0282, -0.0576, -0.0549, -0.0151, 0.0055, 0.0946, 0.1042, 0.1294, 0.1482, 0.1166, 0.1818, 0.1565, 0.092, 0.1591, 0.1248, 0.1541, 0.2178, 0.2169, 0.2433, 0.2164, 0.1934, 0.243, 0.317, 0.3235, 0.3343, 0.3627, 0.347, 0.3455, 0.4113, 0.311, 0.4281, 0.4331, 0.4306, 0.5231, 0.4498, 0.4726, 0.4248, 0.4842, 0.4548, 0.374, 0.3585, 0.2108, 0.2469, 0.3465, 0.3501, 0.3846, 0.3846, 0.422, 0.4459, 0.4401, 0.408, 0.3629, 0.4264, 0.4847, 0.5178, 0.5857, 0.5541, 0.45, 0.4281, 0.3981, 0.3561, 0.4945, 0.599, 0.6631, 0.6612, 0.7238, 0.7238, 0.7124, 0.7708, 0.7603, 0.7169, 0.7169, 0.8974, 0.8777, 1.0659, 1.0425, 0.9672, 1.0759, 1.0806, 1.034, 1.0052, 1.025, 1.1821, 1.1821, 1.1956, 1.3407, 1.3916, 1.4041, 1.3406, 1.4678, 1.6184, 1.6215, 1.4957, 1.6336, 1.5231, 1.2823, 1.3033, 1.3742, 1.3069, 1.2453, 1.4684, 1.4902, 1.4763, 1.4763, 1.4049, 1.5322, 1.5105, 1.5756, 1.5323, 1.5145, 1.5806, 1.4998, 1.4806, 1.4824, 1.2839, 1.4108, 1.3884, 1.2275, 1.3419, 1.4249, 1.5186, 1.4384, 1.5634, 1.6576, 1.7773, 1.7775]}
{"unit_id": "orig:1948513061595480549", "ticker_idx": 0, "ts": "2025-07-24T22:38:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPMorgan: Overweight / TP: ₩360,000 [Maintained]", "tweet_type": "news_reaction", "timeline": "months", "summary": "Compilation of 2Q25 broker views on SK Hynix: JPM/Nomura/Citi/CLSA all Overweight (TPs ₩360-380K); Goldman downgrades to Neutral, cuts TP to ₩300K on 2026 HBM ASP/competition concerns.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix 2Q25 Foreign Broker Comments (July 25, 2025)\n\n(1) JPMorgan: Overweight / TP: ₩360,000 [Maintained]\n\n■ We expect HBM negotiations in 2H25 to proceed smoothly, based on an estimated HBM4 price formation with a +35% premium over existing HBM.\n■ While past capital expenditure expansions by memory companies were interpreted as a sign of oversupply, we believe the current market perceives 1) the increase in HBM-related CAPEX as a positive sign of expanding HBM revenue, and 2) furthermore, a positive indication that the HBM business model is transitioning to a ‘customized’ model.\n■ We are relieved that the trend is expected to be better than pessimists' concerns of a double-digit (DD%) decline in HBM prices in 2026.\n■ However, reflecting stronger price trends for supply-driven general DRAM and some assumption adjustments in the HBM market, we slightly lower our 2026 and 2027 EPS forecasts by -2% and -3% respectively, but maintain our target price at ₩360,000.\n■ Management's comment that eSSD could be utilized as cache storage if AI inference demand increases in the next 2-3 years is still at a cautious expectation level.\n\n(2) Goldman Sachs: Neutral / TP: ₩300,000 [Downgraded]\n\n■ DRAM strength offsets NAND sluggishness. Higher-than-expected legacy shipments improved top-line.\n\n■ HBM revenue is estimated at around $5.3bn. The HBM mix within DRAM appears to have slightly decreased from 46% last quarter to 43% this quarter.\n■ HBM demand and performance are both solid, but they met market expectations. We maintain our outlook for a lack of guidance for 2026 ['26 full booking confirm X] and ASP decline due to intensifying competition.\n■ We foresee HBM3E prices themselves declining significantly in 2026, even with an expected +40-50% premium price for HBM3E 12-hi. We expect OPM to fall from the 60-70% range to the 50-60% range in 2026.\n■ We upgrade our 2025 EPS estimate by +2% but downgrade 2026 and 2027 EPS estimates by -2%. We lower our target price from ₩310,000 to ₩300,000.\n\n(3) Nomura: Overweight / TP: ₩360,000 [Maintained]\n■ The implications of this earnings report are: 1) HBM cost reduction is progressing faster than expected, 2) HBM profitability significantly exceeds market expectations, and 3) NAND inventory reduction is rapid, reducing future burden.\n\n■ DRAM shipments exceeded expectations, attributed to tariff Pull-in Demand.\n\n■ We forecast operating profit of ₩11 trillion for 3Q25(F) and maintain ₩40 trillion for the full year 2025 operating profit.\n■ For 2026, we forecast revenue of ₩110 trillion and operating profit of ₩50 trillion vs. market consensus of ₩96 trillion and ₩41 trillion respectively.\n■ HBM profitability in 2026 is expected to only slightly decrease compared to initial expectations, despite a 7% ASP decline, due to faster-than-expected cost reductions. HBM4 costs are +25% higher than HBM3E, but price increases are limited [+0-10%], suggesting some decline in profitability.\n■ Samsung Electronics' HBM4 costs are expected to be +30% higher than SK Hynix, allowing SK Hynix to maintain an HBM operating profit margin of around 59% in 2026.\n\n(4) Citi: Overweight / TP: ₩380,000 [Downgraded]\n■ This earnings report is positive due to increased HBM3e 12-Hi sales and higher server/PC DRAM sales.\n■ As 2026 HBM supply contracts are not yet finalized and due to intensifying competition in the HBM market, the 2026 HBM ASP growth rate is projected to be revised down from the previous expectation of +7% YoY to -6%.\n\n■ Currently, HBM supply shortages persist, but fierce competition among suppliers is expected from 2026. The HBM market is projected to enter an oversupply state (HBM S/D ratio +3%) in 2026.\n\n■ Accordingly, we lower our target price from ₩430,000 to ₩380,000.\n\n(5) CLSA: Overweight / TP: ₩375,000 [Upgraded]\n■ Despite significant demand pull-ins in the first half of the year, SK Hynix's customer inventories are low, and AI demand remains very strong, making a sharp decline in second-half demand unlikely.\n■ Negotiations with HBM customers for 2026 supply are ongoing, but a certain level of demand visibility has already been secured, and CAPEX is accelerating to proactively address anticipated demand.\n■ For HBM4, production costs have increased due to the integration of advanced features, so Hynix is negotiating ASPs with customers to maintain HBM profitability at current levels.\n\n■ Accordingly, we raise our 2026 and 2027 operating profit estimates by +10% and +18% respectively, and upgrade our target price from ₩350,000 to ₩375,000.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018553554029104857, "ret_signed_m1d": -0.0018553554029104857, "alpha_spy_m1d": -0.0015244124395864533, "alpha_c_m1d": 0.0022794644218875337, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012987138500289652, "ret_signed_p1d": -0.012987138500289652, "alpha_spy_p1d": -0.0172114881406088, "alpha_c_p1d": -0.011909955870553657, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.014842261023144454, "ret_signed_p1w": 0.014842261023144454, "alpha_spy_p1w": 0.018530562621140056, "alpha_c_p1w": 0.011437121541981288, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.06864553000706097, "ret_signed_p1m": -0.06864553000706097, "alpha_spy_p1m": -0.08581090407836389, "alpha_c_p1m": -0.08893745107351747, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.7799327721855513, "ret_signed_p3m": 0.7799327721855513, "alpha_spy_p3m": 0.7188832551236031, "alpha_c_p3m": 0.5801412948306914, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.8112586328636606, "ret_signed_p6m": 1.8112586328636606, "alpha_spy_p6m": 1.7147815308955834, "alpha_c_p6m": 1.4157205111608504, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666, 1.7852, 1.7443, 1.7592, 1.7852, 1.8113]}
{"unit_id": "orig:1948513061595480549", "ticker_idx": 1, "ts": "2025-07-24T22:38:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs: Neutral / TP: ₩300,000 [Downgraded]", "tweet_type": "news_reaction", "timeline": "months", "summary": "Compilation of 2Q25 broker views on SK Hynix: JPM/Nomura/Citi/CLSA all Overweight (TPs ₩360-380K); Goldman downgrades to Neutral, cuts TP to ₩300K on 2026 HBM ASP/competition concerns.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix 2Q25 Foreign Broker Comments (July 25, 2025)\n\n(1) JPMorgan: Overweight / TP: ₩360,000 [Maintained]\n\n■ We expect HBM negotiations in 2H25 to proceed smoothly, based on an estimated HBM4 price formation with a +35% premium over existing HBM.\n■ While past capital expenditure expansions by memory companies were interpreted as a sign of oversupply, we believe the current market perceives 1) the increase in HBM-related CAPEX as a positive sign of expanding HBM revenue, and 2) furthermore, a positive indication that the HBM business model is transitioning to a ‘customized’ model.\n■ We are relieved that the trend is expected to be better than pessimists' concerns of a double-digit (DD%) decline in HBM prices in 2026.\n■ However, reflecting stronger price trends for supply-driven general DRAM and some assumption adjustments in the HBM market, we slightly lower our 2026 and 2027 EPS forecasts by -2% and -3% respectively, but maintain our target price at ₩360,000.\n■ Management's comment that eSSD could be utilized as cache storage if AI inference demand increases in the next 2-3 years is still at a cautious expectation level.\n\n(2) Goldman Sachs: Neutral / TP: ₩300,000 [Downgraded]\n\n■ DRAM strength offsets NAND sluggishness. Higher-than-expected legacy shipments improved top-line.\n\n■ HBM revenue is estimated at around $5.3bn. The HBM mix within DRAM appears to have slightly decreased from 46% last quarter to 43% this quarter.\n■ HBM demand and performance are both solid, but they met market expectations. We maintain our outlook for a lack of guidance for 2026 ['26 full booking confirm X] and ASP decline due to intensifying competition.\n■ We foresee HBM3E prices themselves declining significantly in 2026, even with an expected +40-50% premium price for HBM3E 12-hi. We expect OPM to fall from the 60-70% range to the 50-60% range in 2026.\n■ We upgrade our 2025 EPS estimate by +2% but downgrade 2026 and 2027 EPS estimates by -2%. We lower our target price from ₩310,000 to ₩300,000.\n\n(3) Nomura: Overweight / TP: ₩360,000 [Maintained]\n■ The implications of this earnings report are: 1) HBM cost reduction is progressing faster than expected, 2) HBM profitability significantly exceeds market expectations, and 3) NAND inventory reduction is rapid, reducing future burden.\n\n■ DRAM shipments exceeded expectations, attributed to tariff Pull-in Demand.\n\n■ We forecast operating profit of ₩11 trillion for 3Q25(F) and maintain ₩40 trillion for the full year 2025 operating profit.\n■ For 2026, we forecast revenue of ₩110 trillion and operating profit of ₩50 trillion vs. market consensus of ₩96 trillion and ₩41 trillion respectively.\n■ HBM profitability in 2026 is expected to only slightly decrease compared to initial expectations, despite a 7% ASP decline, due to faster-than-expected cost reductions. HBM4 costs are +25% higher than HBM3E, but price increases are limited [+0-10%], suggesting some decline in profitability.\n■ Samsung Electronics' HBM4 costs are expected to be +30% higher than SK Hynix, allowing SK Hynix to maintain an HBM operating profit margin of around 59% in 2026.\n\n(4) Citi: Overweight / TP: ₩380,000 [Downgraded]\n■ This earnings report is positive due to increased HBM3e 12-Hi sales and higher server/PC DRAM sales.\n■ As 2026 HBM supply contracts are not yet finalized and due to intensifying competition in the HBM market, the 2026 HBM ASP growth rate is projected to be revised down from the previous expectation of +7% YoY to -6%.\n\n■ Currently, HBM supply shortages persist, but fierce competition among suppliers is expected from 2026. The HBM market is projected to enter an oversupply state (HBM S/D ratio +3%) in 2026.\n\n■ Accordingly, we lower our target price from ₩430,000 to ₩380,000.\n\n(5) CLSA: Overweight / TP: ₩375,000 [Upgraded]\n■ Despite significant demand pull-ins in the first half of the year, SK Hynix's customer inventories are low, and AI demand remains very strong, making a sharp decline in second-half demand unlikely.\n■ Negotiations with HBM customers for 2026 supply are ongoing, but a certain level of demand visibility has already been secured, and CAPEX is accelerating to proactively address anticipated demand.\n■ For HBM4, production costs have increased due to the integration of advanced features, so Hynix is negotiating ASPs with customers to maintain HBM profitability at current levels.\n\n■ Accordingly, we raise our 2026 and 2027 operating profit estimates by +10% and +18% respectively, and upgrade our target price from ₩350,000 to ₩375,000.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018553554029104857, "ret_signed_m1d": 0.0018553554029104857, "alpha_spy_m1d": 0.0015244124395864533, "alpha_c_m1d": -0.0022794644218875337, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012987138500289652, "ret_signed_p1d": 0.012987138500289652, "alpha_spy_p1d": 0.0172114881406088, "alpha_c_p1d": 0.011909955870553657, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.014842261023144454, "ret_signed_p1w": -0.014842261023144454, "alpha_spy_p1w": -0.018530562621140056, "alpha_c_p1w": -0.011437121541981288, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.06864553000706097, "ret_signed_p1m": 0.06864553000706097, "alpha_spy_p1m": 0.08581090407836389, "alpha_c_p1m": 0.08893745107351747, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.7799327721855513, "ret_signed_p3m": -0.7799327721855513, "alpha_spy_p3m": -0.7188832551236031, "alpha_c_p3m": -0.5801412948306914, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.8112586328636606, "ret_signed_p6m": -1.8112586328636606, "alpha_spy_p6m": -1.7147815308955834, "alpha_c_p6m": -1.4157205111608504, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666, 1.7852, 1.7443, 1.7592, 1.7852, 1.8113]}
{"unit_id": "orig:1948513061595480549", "ticker_idx": 2, "ts": "2025-07-24T22:38:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura: Overweight / TP: ₩360,000 [Maintained]", "tweet_type": "news_reaction", "timeline": "months", "summary": "Compilation of 2Q25 broker views on SK Hynix: JPM/Nomura/Citi/CLSA all Overweight (TPs ₩360-380K); Goldman downgrades to Neutral, cuts TP to ₩300K on 2026 HBM ASP/competition concerns.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix 2Q25 Foreign Broker Comments (July 25, 2025)\n\n(1) JPMorgan: Overweight / TP: ₩360,000 [Maintained]\n\n■ We expect HBM negotiations in 2H25 to proceed smoothly, based on an estimated HBM4 price formation with a +35% premium over existing HBM.\n■ While past capital expenditure expansions by memory companies were interpreted as a sign of oversupply, we believe the current market perceives 1) the increase in HBM-related CAPEX as a positive sign of expanding HBM revenue, and 2) furthermore, a positive indication that the HBM business model is transitioning to a ‘customized’ model.\n■ We are relieved that the trend is expected to be better than pessimists' concerns of a double-digit (DD%) decline in HBM prices in 2026.\n■ However, reflecting stronger price trends for supply-driven general DRAM and some assumption adjustments in the HBM market, we slightly lower our 2026 and 2027 EPS forecasts by -2% and -3% respectively, but maintain our target price at ₩360,000.\n■ Management's comment that eSSD could be utilized as cache storage if AI inference demand increases in the next 2-3 years is still at a cautious expectation level.\n\n(2) Goldman Sachs: Neutral / TP: ₩300,000 [Downgraded]\n\n■ DRAM strength offsets NAND sluggishness. Higher-than-expected legacy shipments improved top-line.\n\n■ HBM revenue is estimated at around $5.3bn. The HBM mix within DRAM appears to have slightly decreased from 46% last quarter to 43% this quarter.\n■ HBM demand and performance are both solid, but they met market expectations. We maintain our outlook for a lack of guidance for 2026 ['26 full booking confirm X] and ASP decline due to intensifying competition.\n■ We foresee HBM3E prices themselves declining significantly in 2026, even with an expected +40-50% premium price for HBM3E 12-hi. We expect OPM to fall from the 60-70% range to the 50-60% range in 2026.\n■ We upgrade our 2025 EPS estimate by +2% but downgrade 2026 and 2027 EPS estimates by -2%. We lower our target price from ₩310,000 to ₩300,000.\n\n(3) Nomura: Overweight / TP: ₩360,000 [Maintained]\n■ The implications of this earnings report are: 1) HBM cost reduction is progressing faster than expected, 2) HBM profitability significantly exceeds market expectations, and 3) NAND inventory reduction is rapid, reducing future burden.\n\n■ DRAM shipments exceeded expectations, attributed to tariff Pull-in Demand.\n\n■ We forecast operating profit of ₩11 trillion for 3Q25(F) and maintain ₩40 trillion for the full year 2025 operating profit.\n■ For 2026, we forecast revenue of ₩110 trillion and operating profit of ₩50 trillion vs. market consensus of ₩96 trillion and ₩41 trillion respectively.\n■ HBM profitability in 2026 is expected to only slightly decrease compared to initial expectations, despite a 7% ASP decline, due to faster-than-expected cost reductions. HBM4 costs are +25% higher than HBM3E, but price increases are limited [+0-10%], suggesting some decline in profitability.\n■ Samsung Electronics' HBM4 costs are expected to be +30% higher than SK Hynix, allowing SK Hynix to maintain an HBM operating profit margin of around 59% in 2026.\n\n(4) Citi: Overweight / TP: ₩380,000 [Downgraded]\n■ This earnings report is positive due to increased HBM3e 12-Hi sales and higher server/PC DRAM sales.\n■ As 2026 HBM supply contracts are not yet finalized and due to intensifying competition in the HBM market, the 2026 HBM ASP growth rate is projected to be revised down from the previous expectation of +7% YoY to -6%.\n\n■ Currently, HBM supply shortages persist, but fierce competition among suppliers is expected from 2026. The HBM market is projected to enter an oversupply state (HBM S/D ratio +3%) in 2026.\n\n■ Accordingly, we lower our target price from ₩430,000 to ₩380,000.\n\n(5) CLSA: Overweight / TP: ₩375,000 [Upgraded]\n■ Despite significant demand pull-ins in the first half of the year, SK Hynix's customer inventories are low, and AI demand remains very strong, making a sharp decline in second-half demand unlikely.\n■ Negotiations with HBM customers for 2026 supply are ongoing, but a certain level of demand visibility has already been secured, and CAPEX is accelerating to proactively address anticipated demand.\n■ For HBM4, production costs have increased due to the integration of advanced features, so Hynix is negotiating ASPs with customers to maintain HBM profitability at current levels.\n\n■ Accordingly, we raise our 2026 and 2027 operating profit estimates by +10% and +18% respectively, and upgrade our target price from ₩350,000 to ₩375,000.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018553554029104857, "ret_signed_m1d": -0.0018553554029104857, "alpha_spy_m1d": -0.0015244124395864533, "alpha_c_m1d": 0.0022794644218875337, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012987138500289652, "ret_signed_p1d": -0.012987138500289652, "alpha_spy_p1d": -0.0172114881406088, "alpha_c_p1d": -0.011909955870553657, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.014842261023144454, "ret_signed_p1w": 0.014842261023144454, "alpha_spy_p1w": 0.018530562621140056, "alpha_c_p1w": 0.011437121541981288, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.06864553000706097, "ret_signed_p1m": -0.06864553000706097, "alpha_spy_p1m": -0.08581090407836389, "alpha_c_p1m": -0.08893745107351747, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.7799327721855513, "ret_signed_p3m": 0.7799327721855513, "alpha_spy_p3m": 0.7188832551236031, "alpha_c_p3m": 0.5801412948306914, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.8112586328636606, "ret_signed_p6m": 1.8112586328636606, "alpha_spy_p6m": 1.7147815308955834, "alpha_c_p6m": 1.4157205111608504, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666, 1.7852, 1.7443, 1.7592, 1.7852, 1.8113]}
{"unit_id": "orig:1948513061595480549", "ticker_idx": 3, "ts": "2025-07-24T22:38:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi: Overweight / TP: ₩380,000 [Downgraded]", "tweet_type": "news_reaction", "timeline": "months", "summary": "Compilation of 2Q25 broker views on SK Hynix: JPM/Nomura/Citi/CLSA all Overweight (TPs ₩360-380K); Goldman downgrades to Neutral, cuts TP to ₩300K on 2026 HBM ASP/competition concerns.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix 2Q25 Foreign Broker Comments (July 25, 2025)\n\n(1) JPMorgan: Overweight / TP: ₩360,000 [Maintained]\n\n■ We expect HBM negotiations in 2H25 to proceed smoothly, based on an estimated HBM4 price formation with a +35% premium over existing HBM.\n■ While past capital expenditure expansions by memory companies were interpreted as a sign of oversupply, we believe the current market perceives 1) the increase in HBM-related CAPEX as a positive sign of expanding HBM revenue, and 2) furthermore, a positive indication that the HBM business model is transitioning to a ‘customized’ model.\n■ We are relieved that the trend is expected to be better than pessimists' concerns of a double-digit (DD%) decline in HBM prices in 2026.\n■ However, reflecting stronger price trends for supply-driven general DRAM and some assumption adjustments in the HBM market, we slightly lower our 2026 and 2027 EPS forecasts by -2% and -3% respectively, but maintain our target price at ₩360,000.\n■ Management's comment that eSSD could be utilized as cache storage if AI inference demand increases in the next 2-3 years is still at a cautious expectation level.\n\n(2) Goldman Sachs: Neutral / TP: ₩300,000 [Downgraded]\n\n■ DRAM strength offsets NAND sluggishness. Higher-than-expected legacy shipments improved top-line.\n\n■ HBM revenue is estimated at around $5.3bn. The HBM mix within DRAM appears to have slightly decreased from 46% last quarter to 43% this quarter.\n■ HBM demand and performance are both solid, but they met market expectations. We maintain our outlook for a lack of guidance for 2026 ['26 full booking confirm X] and ASP decline due to intensifying competition.\n■ We foresee HBM3E prices themselves declining significantly in 2026, even with an expected +40-50% premium price for HBM3E 12-hi. We expect OPM to fall from the 60-70% range to the 50-60% range in 2026.\n■ We upgrade our 2025 EPS estimate by +2% but downgrade 2026 and 2027 EPS estimates by -2%. We lower our target price from ₩310,000 to ₩300,000.\n\n(3) Nomura: Overweight / TP: ₩360,000 [Maintained]\n■ The implications of this earnings report are: 1) HBM cost reduction is progressing faster than expected, 2) HBM profitability significantly exceeds market expectations, and 3) NAND inventory reduction is rapid, reducing future burden.\n\n■ DRAM shipments exceeded expectations, attributed to tariff Pull-in Demand.\n\n■ We forecast operating profit of ₩11 trillion for 3Q25(F) and maintain ₩40 trillion for the full year 2025 operating profit.\n■ For 2026, we forecast revenue of ₩110 trillion and operating profit of ₩50 trillion vs. market consensus of ₩96 trillion and ₩41 trillion respectively.\n■ HBM profitability in 2026 is expected to only slightly decrease compared to initial expectations, despite a 7% ASP decline, due to faster-than-expected cost reductions. HBM4 costs are +25% higher than HBM3E, but price increases are limited [+0-10%], suggesting some decline in profitability.\n■ Samsung Electronics' HBM4 costs are expected to be +30% higher than SK Hynix, allowing SK Hynix to maintain an HBM operating profit margin of around 59% in 2026.\n\n(4) Citi: Overweight / TP: ₩380,000 [Downgraded]\n■ This earnings report is positive due to increased HBM3e 12-Hi sales and higher server/PC DRAM sales.\n■ As 2026 HBM supply contracts are not yet finalized and due to intensifying competition in the HBM market, the 2026 HBM ASP growth rate is projected to be revised down from the previous expectation of +7% YoY to -6%.\n\n■ Currently, HBM supply shortages persist, but fierce competition among suppliers is expected from 2026. The HBM market is projected to enter an oversupply state (HBM S/D ratio +3%) in 2026.\n\n■ Accordingly, we lower our target price from ₩430,000 to ₩380,000.\n\n(5) CLSA: Overweight / TP: ₩375,000 [Upgraded]\n■ Despite significant demand pull-ins in the first half of the year, SK Hynix's customer inventories are low, and AI demand remains very strong, making a sharp decline in second-half demand unlikely.\n■ Negotiations with HBM customers for 2026 supply are ongoing, but a certain level of demand visibility has already been secured, and CAPEX is accelerating to proactively address anticipated demand.\n■ For HBM4, production costs have increased due to the integration of advanced features, so Hynix is negotiating ASPs with customers to maintain HBM profitability at current levels.\n\n■ Accordingly, we raise our 2026 and 2027 operating profit estimates by +10% and +18% respectively, and upgrade our target price from ₩350,000 to ₩375,000.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018553554029104857, "ret_signed_m1d": -0.0018553554029104857, "alpha_spy_m1d": -0.0015244124395864533, "alpha_c_m1d": 0.0022794644218875337, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012987138500289652, "ret_signed_p1d": -0.012987138500289652, "alpha_spy_p1d": -0.0172114881406088, "alpha_c_p1d": -0.011909955870553657, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.014842261023144454, "ret_signed_p1w": 0.014842261023144454, "alpha_spy_p1w": 0.018530562621140056, "alpha_c_p1w": 0.011437121541981288, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.06864553000706097, "ret_signed_p1m": -0.06864553000706097, "alpha_spy_p1m": -0.08581090407836389, "alpha_c_p1m": -0.08893745107351747, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.7799327721855513, "ret_signed_p3m": 0.7799327721855513, "alpha_spy_p3m": 0.7188832551236031, "alpha_c_p3m": 0.5801412948306914, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.8112586328636606, "ret_signed_p6m": 1.8112586328636606, "alpha_spy_p6m": 1.7147815308955834, "alpha_c_p6m": 1.4157205111608504, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666, 1.7852, 1.7443, 1.7592, 1.7852, 1.8113]}
{"unit_id": "orig:1948513061595480549", "ticker_idx": 4, "ts": "2025-07-24T22:38:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CLSA: Overweight / TP: ₩375,000 [Upgraded]", "tweet_type": "news_reaction", "timeline": "months", "summary": "Compilation of 2Q25 broker views on SK Hynix: JPM/Nomura/Citi/CLSA all Overweight (TPs ₩360-380K); Goldman downgrades to Neutral, cuts TP to ₩300K on 2026 HBM ASP/competition concerns.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix 2Q25 Foreign Broker Comments (July 25, 2025)\n\n(1) JPMorgan: Overweight / TP: ₩360,000 [Maintained]\n\n■ We expect HBM negotiations in 2H25 to proceed smoothly, based on an estimated HBM4 price formation with a +35% premium over existing HBM.\n■ While past capital expenditure expansions by memory companies were interpreted as a sign of oversupply, we believe the current market perceives 1) the increase in HBM-related CAPEX as a positive sign of expanding HBM revenue, and 2) furthermore, a positive indication that the HBM business model is transitioning to a ‘customized’ model.\n■ We are relieved that the trend is expected to be better than pessimists' concerns of a double-digit (DD%) decline in HBM prices in 2026.\n■ However, reflecting stronger price trends for supply-driven general DRAM and some assumption adjustments in the HBM market, we slightly lower our 2026 and 2027 EPS forecasts by -2% and -3% respectively, but maintain our target price at ₩360,000.\n■ Management's comment that eSSD could be utilized as cache storage if AI inference demand increases in the next 2-3 years is still at a cautious expectation level.\n\n(2) Goldman Sachs: Neutral / TP: ₩300,000 [Downgraded]\n\n■ DRAM strength offsets NAND sluggishness. Higher-than-expected legacy shipments improved top-line.\n\n■ HBM revenue is estimated at around $5.3bn. The HBM mix within DRAM appears to have slightly decreased from 46% last quarter to 43% this quarter.\n■ HBM demand and performance are both solid, but they met market expectations. We maintain our outlook for a lack of guidance for 2026 ['26 full booking confirm X] and ASP decline due to intensifying competition.\n■ We foresee HBM3E prices themselves declining significantly in 2026, even with an expected +40-50% premium price for HBM3E 12-hi. We expect OPM to fall from the 60-70% range to the 50-60% range in 2026.\n■ We upgrade our 2025 EPS estimate by +2% but downgrade 2026 and 2027 EPS estimates by -2%. We lower our target price from ₩310,000 to ₩300,000.\n\n(3) Nomura: Overweight / TP: ₩360,000 [Maintained]\n■ The implications of this earnings report are: 1) HBM cost reduction is progressing faster than expected, 2) HBM profitability significantly exceeds market expectations, and 3) NAND inventory reduction is rapid, reducing future burden.\n\n■ DRAM shipments exceeded expectations, attributed to tariff Pull-in Demand.\n\n■ We forecast operating profit of ₩11 trillion for 3Q25(F) and maintain ₩40 trillion for the full year 2025 operating profit.\n■ For 2026, we forecast revenue of ₩110 trillion and operating profit of ₩50 trillion vs. market consensus of ₩96 trillion and ₩41 trillion respectively.\n■ HBM profitability in 2026 is expected to only slightly decrease compared to initial expectations, despite a 7% ASP decline, due to faster-than-expected cost reductions. HBM4 costs are +25% higher than HBM3E, but price increases are limited [+0-10%], suggesting some decline in profitability.\n■ Samsung Electronics' HBM4 costs are expected to be +30% higher than SK Hynix, allowing SK Hynix to maintain an HBM operating profit margin of around 59% in 2026.\n\n(4) Citi: Overweight / TP: ₩380,000 [Downgraded]\n■ This earnings report is positive due to increased HBM3e 12-Hi sales and higher server/PC DRAM sales.\n■ As 2026 HBM supply contracts are not yet finalized and due to intensifying competition in the HBM market, the 2026 HBM ASP growth rate is projected to be revised down from the previous expectation of +7% YoY to -6%.\n\n■ Currently, HBM supply shortages persist, but fierce competition among suppliers is expected from 2026. The HBM market is projected to enter an oversupply state (HBM S/D ratio +3%) in 2026.\n\n■ Accordingly, we lower our target price from ₩430,000 to ₩380,000.\n\n(5) CLSA: Overweight / TP: ₩375,000 [Upgraded]\n■ Despite significant demand pull-ins in the first half of the year, SK Hynix's customer inventories are low, and AI demand remains very strong, making a sharp decline in second-half demand unlikely.\n■ Negotiations with HBM customers for 2026 supply are ongoing, but a certain level of demand visibility has already been secured, and CAPEX is accelerating to proactively address anticipated demand.\n■ For HBM4, production costs have increased due to the integration of advanced features, so Hynix is negotiating ASPs with customers to maintain HBM profitability at current levels.\n\n■ Accordingly, we raise our 2026 and 2027 operating profit estimates by +10% and +18% respectively, and upgrade our target price from ₩350,000 to ₩375,000.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018553554029104857, "ret_signed_m1d": -0.0018553554029104857, "alpha_spy_m1d": -0.0015244124395864533, "alpha_c_m1d": 0.0022794644218875337, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012987138500289652, "ret_signed_p1d": -0.012987138500289652, "alpha_spy_p1d": -0.0172114881406088, "alpha_c_p1d": -0.011909955870553657, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.014842261023144454, "ret_signed_p1w": 0.014842261023144454, "alpha_spy_p1w": 0.018530562621140056, "alpha_c_p1w": 0.011437121541981288, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.06864553000706097, "ret_signed_p1m": -0.06864553000706097, "alpha_spy_p1m": -0.08581090407836389, "alpha_c_p1m": -0.08893745107351747, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.7799327721855513, "ret_signed_p3m": 0.7799327721855513, "alpha_spy_p3m": 0.7188832551236031, "alpha_c_p3m": 0.5801412948306914, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.8112586328636606, "ret_signed_p6m": 1.8112586328636606, "alpha_spy_p6m": 1.7147815308955834, "alpha_c_p6m": 1.4157205111608504, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666, 1.7852, 1.7443, 1.7592, 1.7852, 1.8113]}
{"unit_id": "orig:1946553572730265647", "ticker_idx": 0, "ts": "2025-07-19T12:51:51+00:00", "symbol": "TEAM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "re-affirming our bullish stance on top picks TEAM and GTLB", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Piper Sandler reaffirms bullish stance on TEAM and GTLB; AI will not hurt developer hiring, supports DevOps names.", "resolved_tickers": ["TEAM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Application", "unit_kind": "original", "lang": "en", "tweet_text": "Piper Sandler) AI: Foe or Friend to Developer Hiring?\n\nAs the adoption of AI coding tools proliferates across organizations, the debate around its long-term impact on developer headcount has become a paramount topic of conversation for investors. While the bearish view that AI will lead to developer job cuts in the future has gained traction in several feedback loops, we believe a more optimistic view is warranted for several reasons at this juncture. In this report, we review survey feedback, hiring data, and recent tax law changes to maintain our positive view on how AI will positively impact developer hiring, ultimately re-affirming our bullish stance on top picks TEAM and GTLB.\n\nBackground on the Bear Case. Code writing has become one of the most widely adopted use cases for Generative AI (GenAI). Large Language Models’ (LLMs) ability to rapidly generate code based on natural language input, or through autocomplete, has led to significant improvements in developer productivity within enterprises. This has naturally led to a debate around the ultimate impact on a developer’s job function, or if the job will even exist in the long-term. Conversations across our DevOps coverage have been focused on this risk for some time, with many investors hesitant to invest in potentially impacted names due to this long-term risk to the business model. While we believe in the power of AI, we do not believe it is a structural risk to these stocks and outline our four key reasons why below.\n\nFour Reasons Why AI Will Not Be Detrimental to Developer Hiring\n\n1. Coding Only Accounts for ~17% of a Typical Developer’s Day. While often thought of as a core part of a developer’s day, a MSFT survey of 6,000 developers indicated a typical day includes ~90 minutes of coding time. In our view, efficiency gains in the code writing process alone are not enough to fully displace a developer’s overall responsibilities.\n\n2. CIOs Believe AI Can Accelerate Developer Hiring. Contrary to current conventional wisdom, our 2025 CIO survey data indicates that engineering/DevOps hiring could accelerate as a result of AI adoption, with this cohort representing the largest net beneficiary of any job function we surveyed.\n\n3. Hiring Data Not Indicating AI Induced Pressure. Software developer employment growth slowed in 2024 after two years of strong growth, primarily due to slowing growth within software companies focused on profitability. With recent data indicating hiring in the sector is starting to pick up again, we do not see widespread evidence of AI induced developer cuts at this time.\n\n4. Recent Tax Law Changes Make Developer Hiring More Tax Efficient. Passed earlier this month, the “One Big Beautiful Bill” (OBBB) reversed tax law changes related to domestic R&D expenditures. We believe the prior changes, originally changed for the 2022 tax year, contributed to the slowdown in developer hiring, which could be reversed as a result of the recently passed changes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002572545783236402, "ret_signed_m1d": 0.002572545783236402, "alpha_spy_m1d": 0.004465132273401284, "alpha_c_m1d": 0.004465132273401284, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0018925864901648826, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02099198301215366, "ret_signed_p1d": 0.02099198301215366, "alpha_spy_p1d": 0.02084886184892909, "alpha_c_p1d": 0.02084886184892909, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": 0.0001431211632245688, "ret_p1w": 0.038485262935014974, "ret_signed_p1w": 0.038485262935014974, "alpha_spy_p1w": 0.025491584250249444, "alpha_c_p1w": 0.025491584250249444, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.01299367868476553, "ret_p1m": -0.1448342961930672, "ret_signed_p1m": -0.1448342961930672, "alpha_spy_p1m": -0.16239238483440588, "alpha_c_p1m": -0.16239238483440588, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.017558088641338676, "ret_p3m": -0.2275673780028965, "ret_signed_p3m": -0.2275673780028965, "alpha_spy_p3m": -0.28116664520687784, "alpha_c_p3m": -0.28116664520687784, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.05359926720398134, "ret_p6m": -0.2905433270393576, "ret_signed_p6m": -0.2905433270393576, "alpha_spy_p6m": -0.4002482438974153, "alpha_c_p6m": -0.4002482438974153, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.10970491685805772, "price_path": [0.0716, 0.1581, 0.1766, 0.1825, 0.1756, 0.1747, 0.1786, 0.0726, 0.069, 0.0696, 0.0584, 0.0713, 0.0677, 0.1512, 0.1809, 0.157, 0.1364, 0.136, 0.1035, 0.1036, 0.0762, 0.0721, 0.0641, 0.0641, 0.086, 0.0961, 0.0492, 0.0683, 0.058, 0.085, 0.0852, 0.1247, 0.1037, 0.1061, 0.0962, 0.0367, 0.0255, 0.0127, 0.0307, 0.0198, 0.0092, 0.0092, -0.0249, -0.0083, 0.0325, 0.0206, 0.0213, 0.041, 0.0449, 0.0687, 0.0639, 0.0986, 0.0986, 0.1116, 0.1286, 0.1365, 0.0293, -0.0378, -0.0141, -0.0343, -0.0201, -0.0192, 0.0026, 0.0, 0.021, 0.0422, 0.0281, 0.0445, 0.0385, 0.0293, 0.0146, -0.0133, -0.0587, -0.0446, -0.0459, -0.0456, -0.1202, -0.1353, -0.1805, -0.1871, -0.1563, -0.1542, -0.1408, -0.1296, -0.1448, -0.1431, -0.1532, -0.1213, -0.1322, -0.1549, -0.1157, -0.0905, -0.0853, -0.0853, -0.1115, -0.1218, -0.1352, -0.1128, -0.1114, -0.0617, -0.1031, -0.0908, -0.1003, -0.1054, -0.1148, -0.1138, -0.1272, -0.1302, -0.1221, -0.1476, -0.1571, -0.1643, -0.1426, -0.1495, -0.1783, -0.2304, -0.2228, -0.2257, -0.2273, -0.2276, -0.2307, -0.2399, -0.2562, -0.2314, -0.1955, -0.2291, -0.2276, -0.2182, -0.183, -0.132, -0.1362, -0.151, -0.1632, -0.1549, -0.1564, -0.1805, -0.1733, -0.1283, -0.0941, -0.1511, -0.1693, -0.1967, -0.188, -0.1813, -0.1894, -0.1999, -0.2177, -0.219, -0.2487, -0.2347, -0.2462, -0.2589, -0.2474, -0.2465, -0.2332, -0.2384, -0.2384, -0.2307, -0.2201, -0.2081, -0.1935, -0.1946, -0.1802, -0.173, -0.1695, -0.1679, -0.16, -0.1663, -0.1811, -0.179, -0.1666, -0.1511, -0.1713, -0.1539, -0.1645, -0.171, -0.171, -0.1606, -0.1563, -0.1617, -0.1658, -0.1658, -0.2037, -0.2083, -0.1852, -0.1723, -0.2312, -0.2467, -0.2466, -0.2905]}
{"unit_id": "orig:1946553572730265647", "ticker_idx": 1, "ts": "2025-07-19T12:51:51+00:00", "symbol": "GTLB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "re-affirming our bullish stance on top picks TEAM and GTLB", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Piper Sandler reaffirms bullish stance on TEAM and GTLB; AI will not hurt developer hiring, supports DevOps names.", "resolved_tickers": ["GTLB"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Piper Sandler) AI: Foe or Friend to Developer Hiring?\n\nAs the adoption of AI coding tools proliferates across organizations, the debate around its long-term impact on developer headcount has become a paramount topic of conversation for investors. While the bearish view that AI will lead to developer job cuts in the future has gained traction in several feedback loops, we believe a more optimistic view is warranted for several reasons at this juncture. In this report, we review survey feedback, hiring data, and recent tax law changes to maintain our positive view on how AI will positively impact developer hiring, ultimately re-affirming our bullish stance on top picks TEAM and GTLB.\n\nBackground on the Bear Case. Code writing has become one of the most widely adopted use cases for Generative AI (GenAI). Large Language Models’ (LLMs) ability to rapidly generate code based on natural language input, or through autocomplete, has led to significant improvements in developer productivity within enterprises. This has naturally led to a debate around the ultimate impact on a developer’s job function, or if the job will even exist in the long-term. Conversations across our DevOps coverage have been focused on this risk for some time, with many investors hesitant to invest in potentially impacted names due to this long-term risk to the business model. While we believe in the power of AI, we do not believe it is a structural risk to these stocks and outline our four key reasons why below.\n\nFour Reasons Why AI Will Not Be Detrimental to Developer Hiring\n\n1. Coding Only Accounts for ~17% of a Typical Developer’s Day. While often thought of as a core part of a developer’s day, a MSFT survey of 6,000 developers indicated a typical day includes ~90 minutes of coding time. In our view, efficiency gains in the code writing process alone are not enough to fully displace a developer’s overall responsibilities.\n\n2. CIOs Believe AI Can Accelerate Developer Hiring. Contrary to current conventional wisdom, our 2025 CIO survey data indicates that engineering/DevOps hiring could accelerate as a result of AI adoption, with this cohort representing the largest net beneficiary of any job function we surveyed.\n\n3. Hiring Data Not Indicating AI Induced Pressure. Software developer employment growth slowed in 2024 after two years of strong growth, primarily due to slowing growth within software companies focused on profitability. With recent data indicating hiring in the sector is starting to pick up again, we do not see widespread evidence of AI induced developer cuts at this time.\n\n4. Recent Tax Law Changes Make Developer Hiring More Tax Efficient. Passed earlier this month, the “One Big Beautiful Bill” (OBBB) reversed tax law changes related to domestic R&D expenditures. We believe the prior changes, originally changed for the 2022 tax year, contributed to the slowdown in developer hiring, which could be reversed as a result of the recently passed changes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010017826831759669, "ret_signed_m1d": 0.010017826831759669, "alpha_spy_m1d": 0.011910413321924551, "alpha_c_m1d": 0.011910413321924551, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0018925864901648826, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024710611210998268, "ret_signed_p1d": 0.024710611210998268, "alpha_spy_p1d": 0.0245674900477737, "alpha_c_p1d": 0.0245674900477737, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": 0.0001431211632245688, "ret_p1w": 0.0770258913670645, "ret_signed_p1w": 0.0770258913670645, "alpha_spy_p1w": 0.06403221268229897, "alpha_c_p1w": 0.06403221268229897, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.01299367868476553, "ret_p1m": -0.014470118826296119, "ret_signed_p1m": -0.014470118826296119, "alpha_spy_p1m": -0.032028207467634795, "alpha_c_p1m": -0.032028207467634795, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.017558088641338676, "ret_p3m": 0.07457707981685369, "ret_signed_p3m": 0.07457707981685369, "alpha_spy_p3m": 0.02097781261287235, "alpha_c_p3m": 0.02097781261287235, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.05359926720398134, "ret_p6m": -0.2019145161295801, "ret_signed_p6m": -0.2019145161295801, "alpha_spy_p6m": -0.31161943298763783, "alpha_c_p6m": -0.31161943298763783, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.10970491685805772, "price_path": [-0.0445, 0.0167, 0.0499, 0.0454, 0.0496, 0.039, 0.0454, 0.0363, 0.0125, 0.035, 0.0481, 0.1026, 0.0944, 0.1589, 0.1879, 0.1894, 0.1534, 0.1463, 0.0897, 0.1167, 0.0597, 0.0634, 0.0574, 0.0574, 0.0686, 0.0296, 0.0169, 0.0131, 0.014, 0.039, 0.055, 0.0857, 0.0857, 0.0828, 0.0799, -0.0345, -0.0269, -0.0821, -0.0632, -0.0679, -0.0641, -0.0641, -0.0926, -0.0937, -0.0465, -0.0519, -0.0154, 0.0049, 0.0042, 0.0131, 0.0049, 0.0323, 0.0323, 0.0363, 0.0374, 0.0436, 0.0038, -0.0643, -0.0523, -0.045, -0.0523, -0.0223, 0.01, 0.0, 0.0247, 0.0485, 0.0416, 0.0661, 0.077, 0.0505, 0.0278, -0.0247, -0.069, -0.0171, -0.0205, -0.0258, -0.0739, -0.1002, -0.1358, -0.1262, -0.0755, -0.0884, -0.0142, 0.0096, -0.0145, -0.0272, -0.0165, 0.0058, 0.0069, -0.0256, 0.0521, 0.0914, 0.069, 0.069, 0.0617, 0.0443, -0.0325, 0.018, 0.0661, 0.0724, 0.0911, 0.1073, 0.1198, 0.1151, 0.12, 0.1227, 0.114, 0.1175, 0.1015, 0.0873, 0.0514, 0.0401, 0.0568, 0.043, 0.0036, -0.018, 0.035, 0.0443, 0.0492, 0.0274, 0.0539, 0.0432, -0.0069, -0.0163, -0.0036, -0.0284, 0.0746, 0.0087, 0.0485, 0.1022, 0.0706, 0.0712, 0.0804, 0.0799, 0.0732, 0.0692, 0.057, 0.0853, 0.1362, 0.0352, 0.0203, 0.0142, 0.014, 0.0283, 0.0358, 0.0341, -0.0211, -0.0131, -0.0508, -0.0145, -0.0338, -0.0537, -0.0764, -0.0772, -0.0819, -0.0968, -0.0968, -0.0859, -0.0839, -0.0345, -0.1578, -0.165, -0.1685, -0.1302, -0.1313, -0.0984, -0.0939, -0.1202, -0.1409, -0.1518, -0.1569, -0.1525, -0.1541, -0.1456, -0.1636, -0.1656, -0.1656, -0.1576, -0.1492, -0.1505, -0.1645, -0.1645, -0.1946, -0.165, -0.1821, -0.1817, -0.2086, -0.2122, -0.2061, -0.2019]}
{"unit_id": "orig:1963050551987622392", "ticker_idx": 0, "ts": "2025-09-03T01:24:57+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA, AMD, AWS/Amazon, Microsoft, Google Cloud on sustained demand from rising token consumption despite falling unit inference costs.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI Inference Costs (Token Pricing) and Their Investment/Industry Implications\n\nAI Inference Costs (Token Pricing)\n\n- Between 2022 and 2025, costs have been dropping by roughly 10x annually.\n\n- The most advanced models (e.g., GPT-4o level) have seen nearly a 900x annual reduction.\n\n- Less sophisticated models only saw about a 9x annual reduction.\n\nDual Cost Structure\n\n- Training: Still extremely expensive → heavy burden on companies.\n\n- Inference: Rapidly becoming cheaper → enabling broader enterprise adoption.\n\nNew Challenge: Token Usage Growth\n\n- More advanced models (with stronger “reasoning/thinking” capabilities) consume more tokens.\n\n- As a result, even if unit costs fall, total usage may drive overall costs higher.\n\nCorporate Dilemma\n\n- Absorb higher costs to maintain high-performance models.\n\n- Cut costs by choosing cheaper, simpler models.\n\n- Ultimately, companies face the hard decision of passing costs to customers vs absorbing them internally.\n\nInvestment & Industry Implications\n\n1. Opportunities for AI Infrastructure Companies\n\n- Falling token prices may drive wider adoption.\n\n- Yet, increased token consumption → continued demand for data centers and GPUs.\n\n- Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong.\n\n2. Risks for SaaS Companies\n\nSaaS firms embedding AI features face margin pressure.\n\nPassing costs to customers risks price resistance, while absorbing them erodes profitability.\n\n3. Shifts in Consumer Choice\n\n- Some consumers may opt for low-cost, lightweight models over premium ones.\n\n- This encourages market diversification and segmentation of AI models.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0009376653358550424, "ret_signed_m1d": 0.0009376653358550424, "alpha_spy_m1d": 0.006328011745022288, "alpha_c_m1d": 0.0007630296980438711, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": 0.0001746356378111713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006095406083867605, "ret_signed_p1d": 0.006095406083867605, "alpha_spy_p1d": -0.0022619543579469603, "alpha_c_p1d": -0.005879532740221016, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.011974938824088621, "ret_p1w": 0.03932711353729168, "ret_signed_p1w": 0.03932711353729168, "alpha_spy_p1w": 0.026169558494543832, "alpha_c_p1w": -0.010946895549746483, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.05027400908703816, "ret_p1m": 0.10714241949499526, "ret_signed_p1m": 0.10714241949499526, "alpha_spy_p1m": 0.06467912278766375, "alpha_c_p1m": -0.07373948833842836, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.18088190783342362, "ret_p3m": 0.05456652320974387, "ret_signed_p3m": 0.05456652320974387, "alpha_spy_p3m": -0.005109853296400679, "alpha_c_p3m": -0.17770667703855092, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.2322732002482948, "ret_p6m": 0.0837574950390858, "ret_signed_p6m": 0.0837574950390858, "alpha_spy_p6m": 0.006842126365361079, "alpha_c_p6m": -0.35915011779220896, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.44290761283129476, "price_path": [-0.1694, -0.1641, -0.1563, -0.1629, -0.1502, -0.1679, -0.152, -0.1553, -0.1473, -0.1473, -0.1569, -0.155, -0.1332, -0.0956, -0.0914, -0.0754, -0.074, -0.1015, -0.0784, -0.0661, -0.0661, -0.0726, -0.0622, -0.0454, -0.0382, -0.0334, -0.0384, 0.0005, 0.0044, 0.0139, 0.0105, 0.0045, -0.021, 0.0009, 0.0183, 0.0169, 0.0359, 0.0287, 0.0507, 0.0425, 0.0182, 0.055, 0.0448, 0.0516, 0.0595, 0.0708, 0.067, 0.0735, 0.0643, 0.0668, 0.0576, 0.0668, 0.0294, 0.028, 0.0256, 0.0432, 0.0539, 0.0653, 0.0644, 0.056, 0.0209, 0.0209, 0.0009, 0.0, 0.0061, -0.0211, -0.0135, 0.0008, 0.0393, 0.0384, 0.0423, 0.0418, 0.025, -0.0019, 0.033, 0.0355, 0.0762, 0.0458, 0.0373, 0.0415, 0.0444, 0.0659, 0.0936, 0.0975, 0.1071, 0.0997, 0.0875, 0.0846, 0.1084, 0.1287, 0.0736, 0.1038, 0.0552, 0.054, 0.0656, 0.0739, 0.0705, 0.0618, 0.0567, 0.0677, 0.0917, 0.1224, 0.1783, 0.2135, 0.1892, 0.1869, 0.2126, 0.1646, 0.1442, 0.1024, 0.1028, 0.1667, 0.1322, 0.1359, 0.0952, 0.1146, 0.0937, 0.063, 0.0933, 0.0588, 0.0485, 0.07, 0.0423, 0.0566, 0.0566, 0.0375, 0.0546, 0.0636, 0.0526, 0.0749, 0.0692, 0.0876, 0.0842, 0.0773, 0.0605, 0.0259, 0.0333, 0.0417, 0.002, 0.0207, 0.0609, 0.0767, 0.1091, 0.1056, 0.1056, 0.1168, 0.1033, 0.0993, 0.0932, 0.0932, 0.107, 0.1027, 0.0975, 0.1085, 0.0846, 0.0836, 0.0841, 0.0892, 0.0735, 0.0964, 0.0916, 0.0916, 0.0438, 0.0746, 0.0835, 0.1001, 0.093, 0.105, 0.1226, 0.1284, 0.1203, 0.088, 0.0571, 0.021, 0.0075, 0.0868, 0.1139, 0.1052, 0.114, 0.0958, 0.0716, 0.0716, 0.0842, 0.1019, 0.1014, 0.1127, 0.1228, 0.1304, 0.1463, 0.0838]}
{"unit_id": "orig:1963050551987622392", "ticker_idx": 1, "ts": "2025-09-03T01:24:57+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA, AMD, AWS/Amazon, Microsoft, Google Cloud on sustained demand from rising token consumption despite falling unit inference costs.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI Inference Costs (Token Pricing) and Their Investment/Industry Implications\n\nAI Inference Costs (Token Pricing)\n\n- Between 2022 and 2025, costs have been dropping by roughly 10x annually.\n\n- The most advanced models (e.g., GPT-4o level) have seen nearly a 900x annual reduction.\n\n- Less sophisticated models only saw about a 9x annual reduction.\n\nDual Cost Structure\n\n- Training: Still extremely expensive → heavy burden on companies.\n\n- Inference: Rapidly becoming cheaper → enabling broader enterprise adoption.\n\nNew Challenge: Token Usage Growth\n\n- More advanced models (with stronger “reasoning/thinking” capabilities) consume more tokens.\n\n- As a result, even if unit costs fall, total usage may drive overall costs higher.\n\nCorporate Dilemma\n\n- Absorb higher costs to maintain high-performance models.\n\n- Cut costs by choosing cheaper, simpler models.\n\n- Ultimately, companies face the hard decision of passing costs to customers vs absorbing them internally.\n\nInvestment & Industry Implications\n\n1. Opportunities for AI Infrastructure Companies\n\n- Falling token prices may drive wider adoption.\n\n- Yet, increased token consumption → continued demand for data centers and GPUs.\n\n- Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong.\n\n2. Risks for SaaS Companies\n\nSaaS firms embedding AI features face margin pressure.\n\nPassing costs to customers risks price resistance, while absorbing them erodes profitability.\n\n3. Shifts in Consumer Choice\n\n- Some consumers may opt for low-cost, lightweight models over premium ones.\n\n- This encourages market diversification and segmentation of AI models.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0011719141163510916, "ret_signed_m1d": 0.0011719141163510916, "alpha_spy_m1d": 0.0065622605255183375, "alpha_c_m1d": 0.0009972784785399202, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": 0.0001746356378111713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002097154051931671, "ret_signed_p1d": -0.002097154051931671, "alpha_spy_p1d": -0.010454514493746236, "alpha_c_p1d": -0.014072092876020292, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.011974938824088621, "ret_p1w": -0.01597490605487706, "ret_signed_p1w": -0.01597490605487706, "alpha_spy_p1w": -0.029132461097624907, "alpha_c_p1w": -0.06624891514191522, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.05027400908703816, "ret_p1m": 0.04687590585234269, "ret_signed_p1m": 0.04687590585234269, "alpha_spy_p1m": 0.0044126091450111815, "alpha_c_p1m": -0.13400600198108092, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.18088190783342362, "ret_p3m": 0.35545542397089536, "ret_signed_p3m": 0.35545542397089536, "alpha_spy_p3m": 0.2957790474647508, "alpha_c_p3m": 0.12318222372260057, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.2322732002482948, "ret_p6m": 0.2562757450294353, "ret_signed_p6m": 0.2562757450294353, "alpha_spy_p6m": 0.17936037635571056, "alpha_c_p6m": -0.18663186780185947, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.44290761283129476, "price_path": [-0.2834, -0.2492, -0.2399, -0.2528, -0.2691, -0.2835, -0.2204, -0.2161, -0.218, -0.218, -0.209, -0.2008, -0.1462, -0.1155, -0.1138, -0.113, -0.1248, -0.1605, -0.1456, -0.1494, -0.1494, -0.1686, -0.1499, -0.1463, -0.1108, -0.0969, -0.098, -0.0402, -0.0126, -0.0106, -0.0317, -0.0316, -0.0457, -0.0215, -0.0001, 0.0268, 0.0711, 0.0944, 0.1072, 0.0875, 0.059, 0.0904, 0.0751, 0.0061, 0.0633, 0.0656, 0.0626, 0.0791, 0.1375, 0.1161, 0.0949, 0.0864, 0.0273, 0.0189, 0.0097, 0.0347, 0.0076, 0.0277, 0.0308, 0.0398, 0.0031, 0.0031, 0.0012, 0.0, -0.0021, -0.0678, -0.0661, -0.0389, -0.016, -0.0398, -0.022, -0.006, -0.0103, -0.0183, -0.026, -0.0292, -0.0144, -0.0076, -0.0077, -0.0053, -0.0165, -0.0047, -0.0021, 0.0116, 0.0469, 0.0157, 0.2565, 0.3046, 0.4529, 0.4364, 0.3255, 0.3349, 0.3452, 0.4717, 0.4467, 0.4376, 0.4837, 0.4681, 0.42, 0.4494, 0.56, 0.6016, 0.5914, 0.6304, 0.5718, 0.5797, 0.6015, 0.5423, 0.581, 0.4661, 0.4404, 0.5048, 0.465, 0.5968, 0.5294, 0.5223, 0.4835, 0.4204, 0.3788, 0.2707, 0.2569, 0.3264, 0.2714, 0.3214, 0.3214, 0.3417, 0.3555, 0.3276, 0.3421, 0.3321, 0.3444, 0.3638, 0.3669, 0.3657, 0.3658, 0.3001, 0.2803, 0.2901, 0.2219, 0.2401, 0.3164, 0.3258, 0.3255, 0.3263, 0.3263, 0.326, 0.3299, 0.3282, 0.3209, 0.3209, 0.3783, 0.3636, 0.3221, 0.2954, 0.2624, 0.2531, 0.281, 0.3629, 0.3791, 0.4058, 0.4299, 0.4299, 0.4305, 0.5407, 0.565, 0.6017, 0.5501, 0.5545, 0.5589, 0.5554, 0.4601, 0.519, 0.4933, 0.2347, 0.1873, 0.2856, 0.3323, 0.3173, 0.3173, 0.2702, 0.2787, 0.2787, 0.2526, 0.2343, 0.2544, 0.2345, 0.2126, 0.3189, 0.3006, 0.2563]}
{"unit_id": "orig:1963050551987622392", "ticker_idx": 2, "ts": "2025-09-03T01:24:57+00:00", "symbol": "Amazon Web Services", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA, AMD, AWS/Amazon, Microsoft, Google Cloud on sustained demand from rising token consumption despite falling unit inference costs.", "resolved_tickers": ["AMZN"], "resolution_method": "exchange_ticker", "resolution_reason": "AWS is a division of Amazon AMZN; no separate listing", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "AI Inference Costs (Token Pricing) and Their Investment/Industry Implications\n\nAI Inference Costs (Token Pricing)\n\n- Between 2022 and 2025, costs have been dropping by roughly 10x annually.\n\n- The most advanced models (e.g., GPT-4o level) have seen nearly a 900x annual reduction.\n\n- Less sophisticated models only saw about a 9x annual reduction.\n\nDual Cost Structure\n\n- Training: Still extremely expensive → heavy burden on companies.\n\n- Inference: Rapidly becoming cheaper → enabling broader enterprise adoption.\n\nNew Challenge: Token Usage Growth\n\n- More advanced models (with stronger “reasoning/thinking” capabilities) consume more tokens.\n\n- As a result, even if unit costs fall, total usage may drive overall costs higher.\n\nCorporate Dilemma\n\n- Absorb higher costs to maintain high-performance models.\n\n- Cut costs by choosing cheaper, simpler models.\n\n- Ultimately, companies face the hard decision of passing costs to customers vs absorbing them internally.\n\nInvestment & Industry Implications\n\n1. Opportunities for AI Infrastructure Companies\n\n- Falling token prices may drive wider adoption.\n\n- Yet, increased token consumption → continued demand for data centers and GPUs.\n\n- Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong.\n\n2. Risks for SaaS Companies\n\nSaaS firms embedding AI features face margin pressure.\n\nPassing costs to customers risks price resistance, while absorbing them erodes profitability.\n\n3. Shifts in Consumer Choice\n\n- Some consumers may opt for low-cost, lightweight models over premium ones.\n\n- This encourages market diversification and segmentation of AI models.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002876273903595683, "ret_signed_m1d": -0.002876273903595683, "alpha_spy_m1d": 0.002514072505571563, "alpha_c_m1d": 0.0011508056144785561, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": -0.004027079518074239, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.042877945692648245, "ret_signed_p1d": 0.042877945692648245, "alpha_spy_p1d": 0.03452058525083368, "alpha_c_p1d": 0.02460483979228556, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.018273105900362685, "ret_p1w": 0.019204372903216305, "ret_signed_p1w": 0.019204372903216305, "alpha_spy_p1w": 0.006046817860468456, "alpha_c_p1w": 0.010847252554808895, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.00835712034840741, "ret_p1m": -0.015841416629210125, "ret_signed_p1m": -0.015841416629210125, "alpha_spy_p1m": -0.058304713336541636, "alpha_c_p1m": -0.05430261771165712, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.038461201082446994, "ret_p3m": 0.03491304570054132, "ret_signed_p3m": 0.03491304570054132, "alpha_spy_p3m": -0.024763330805603223, "alpha_c_p3m": 0.00859754302598903, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.026315502674552294, "ret_p6m": -0.07995932069998268, "ret_signed_p6m": -0.07995932069998268, "alpha_spy_p6m": -0.1568746893737074, "alpha_c_p6m": -0.09744282661959014, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.017483505919607456, "price_path": [-0.055, -0.0399, -0.0371, -0.0566, -0.0564, -0.0615, -0.0438, -0.0494, -0.0596, -0.0596, -0.0721, -0.0775, -0.0585, -0.0619, -0.0392, -0.0119, -0.0292, -0.0245, -0.0269, -0.0114, -0.0114, -0.0112, -0.0293, -0.0153, -0.0165, -0.0043, -0.0013, 0.0016, -0.0124, -0.0093, 0.0006, 0.0146, 0.0065, 0.0102, 0.0276, 0.0241, 0.0301, 0.0222, 0.0186, 0.0359, -0.0497, -0.0635, -0.0542, -0.0163, -0.0127, -0.0146, -0.0208, -0.02, -0.0063, 0.0221, 0.0223, 0.0243, 0.0089, -0.0096, -0.0179, 0.0126, 0.0086, 0.012, 0.0139, 0.0248, 0.0133, 0.0133, -0.0029, 0.0, 0.0429, 0.0281, 0.0436, 0.0542, 0.0192, 0.0175, 0.0096, 0.0241, 0.0357, 0.0249, 0.0232, 0.0243, 0.0073, -0.0234, -0.0256, -0.0347, -0.0275, -0.0169, -0.0284, -0.0237, -0.0158, -0.0287, -0.0225, -0.0186, -0.0034, 0.0077, -0.0426, -0.0262, -0.0425, -0.0461, -0.051, -0.0573, -0.0421, -0.0175, -0.0356, -0.0217, -0.0079, 0.0043, 0.0144, 0.0191, -0.0139, 0.0807, 0.1239, 0.1032, 0.1071, 0.0754, 0.0815, 0.0992, 0.1023, 0.0806, 0.0513, 0.0385, 0.0304, -0.0152, -0.0146, -0.0392, -0.0235, 0.0013, 0.0163, 0.014, 0.014, 0.032, 0.0349, 0.0373, 0.0283, 0.0138, 0.0157, 0.004, 0.0085, 0.0256, 0.019, 0.0009, -0.0153, -0.0152, -0.0209, 0.0034, 0.006, 0.0108, 0.0272, 0.0283, 0.0283, 0.0289, 0.0269, 0.0289, 0.0214, 0.0214, 0.0023, 0.0313, 0.0661, 0.0689, 0.0898, 0.0947, 0.0906, 0.0735, 0.0472, 0.0539, 0.0581, 0.0581, 0.0222, 0.0235, 0.0369, 0.0583, 0.055, 0.0827, 0.0753, 0.0696, 0.0589, 0.0751, 0.0559, 0.031, -0.0146, -0.0693, -0.0764, -0.0842, -0.097, -0.1168, -0.1204, -0.1204, -0.1099, -0.0938, -0.0935, -0.0703, -0.0917, -0.0771, -0.0679, -0.08]}
{"unit_id": "orig:1963050551987622392", "ticker_idx": 3, "ts": "2025-09-03T01:24:57+00:00", "symbol": "Microsoft", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA, AMD, AWS/Amazon, Microsoft, Google Cloud on sustained demand from rising token consumption despite falling unit inference costs.", "resolved_tickers": ["MSFT"], "resolution_method": "exchange_ticker", "resolution_reason": "Microsoft Corporation MSFT US", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "AI Inference Costs (Token Pricing) and Their Investment/Industry Implications\n\nAI Inference Costs (Token Pricing)\n\n- Between 2022 and 2025, costs have been dropping by roughly 10x annually.\n\n- The most advanced models (e.g., GPT-4o level) have seen nearly a 900x annual reduction.\n\n- Less sophisticated models only saw about a 9x annual reduction.\n\nDual Cost Structure\n\n- Training: Still extremely expensive → heavy burden on companies.\n\n- Inference: Rapidly becoming cheaper → enabling broader enterprise adoption.\n\nNew Challenge: Token Usage Growth\n\n- More advanced models (with stronger “reasoning/thinking” capabilities) consume more tokens.\n\n- As a result, even if unit costs fall, total usage may drive overall costs higher.\n\nCorporate Dilemma\n\n- Absorb higher costs to maintain high-performance models.\n\n- Cut costs by choosing cheaper, simpler models.\n\n- Ultimately, companies face the hard decision of passing costs to customers vs absorbing them internally.\n\nInvestment & Industry Implications\n\n1. Opportunities for AI Infrastructure Companies\n\n- Falling token prices may drive wider adoption.\n\n- Yet, increased token consumption → continued demand for data centers and GPUs.\n\n- Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong.\n\n2. Risks for SaaS Companies\n\nSaaS firms embedding AI features face margin pressure.\n\nPassing costs to customers risks price resistance, while absorbing them erodes profitability.\n\n3. Shifts in Consumer Choice\n\n- Some consumers may opt for low-cost, lightweight models over premium ones.\n\n- This encourages market diversification and segmentation of AI models.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0004551781682208267, "ret_signed_m1d": -0.0004551781682208267, "alpha_spy_m1d": 0.004935168240946419, "alpha_c_m1d": 0.004935168240946419, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": -0.005390346409167246, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005184534030341359, "ret_signed_p1d": 0.005184534030341359, "alpha_spy_p1d": -0.0031728264114732063, "alpha_c_p1d": -0.0031728264114732063, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.008357360441814565, "ret_p1w": -0.0098545769562558, "ret_signed_p1w": -0.0098545769562558, "alpha_spy_p1w": -0.02301213199900365, "alpha_c_p1w": -0.02301213199900365, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.013157555042747848, "ret_p1m": 0.02055989345549203, "ret_signed_p1m": 0.02055989345549203, "alpha_spy_p1m": -0.02190340325183948, "alpha_c_p1m": -0.02190340325183948, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.04246329670733151, "ret_p3m": -0.035023316486474276, "ret_signed_p3m": -0.035023316486474276, "alpha_spy_p3m": -0.09469969299261882, "alpha_c_p3m": -0.09469969299261882, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.059676376506144546, "ret_p6m": -0.20176018430765663, "ret_signed_p6m": -0.20176018430765663, "alpha_spy_p6m": -0.27867555298138136, "alpha_c_p6m": -0.27867555298138136, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.07691536867372473, "price_path": [-0.0707, -0.066, -0.0697, -0.0663, -0.054, -0.0617, -0.0534, -0.0556, -0.0512, -0.0512, -0.0569, -0.0399, -0.0317, -0.0275, -0.0172, -0.0202, -0.0173, -0.0279, -0.0298, -0.0145, -0.0145, -0.0167, -0.0189, -0.0053, -0.0093, -0.0057, -0.0062, -0.0007, -0.0011, 0.0109, 0.0076, 0.0077, -0.0018, -0.0006, 0.0093, 0.0149, 0.0125, 0.0126, 0.0139, 0.054, 0.0354, 0.0582, 0.0426, 0.0371, 0.029, 0.0313, 0.0308, 0.0456, 0.0284, 0.0322, 0.0276, 0.0216, 0.0071, -0.0009, -0.0022, 0.0037, -0.0022, -0.0065, 0.0028, 0.0085, 0.0027, 0.0027, -0.0005, 0.0, 0.0052, -0.0205, -0.0141, -0.0137, -0.0099, -0.0086, 0.009, 0.0198, 0.0073, 0.0092, 0.0061, 0.0249, 0.018, 0.0077, 0.0095, 0.0033, 0.0121, 0.0183, 0.0249, 0.0284, 0.0206, 0.0237, 0.0459, 0.0369, 0.0386, 0.0337, 0.0111, 0.0172, 0.0163, 0.016, 0.0124, 0.0163, 0.0226, 0.0244, 0.0301, 0.0301, 0.0361, 0.0518, 0.0727, 0.0716, 0.0404, 0.0247, 0.0231, 0.0178, 0.0036, -0.0163, -0.0169, 0.0013, 0.0066, 0.0115, -0.0041, 0.0096, 0.0042, -0.0229, -0.0361, -0.0515, -0.064, -0.0603, -0.0544, -0.0375, -0.0375, -0.0246, -0.035, -0.0286, -0.0529, -0.0467, -0.0421, -0.0265, -0.0246, -0.0512, -0.0415, -0.0513, -0.0587, -0.0555, -0.0561, -0.0405, -0.0366, -0.0386, -0.0348, -0.0325, -0.0325, -0.0331, -0.0343, -0.0336, -0.0412, -0.0412, -0.0624, -0.0626, -0.0513, -0.0415, -0.0521, -0.0498, -0.054, -0.0669, -0.0893, -0.0947, -0.0883, -0.0883, -0.0989, -0.1195, -0.1056, -0.0762, -0.0677, -0.0472, -0.0452, -0.1406, -0.1469, -0.1607, -0.1848, -0.1789, -0.2195, -0.2047, -0.18, -0.1807, -0.1983, -0.2033, -0.2044, -0.2044, -0.2132, -0.2078, -0.2082, -0.2107, -0.236, -0.227, -0.204, -0.2018]}
{"unit_id": "orig:1963050551987622392", "ticker_idx": 4, "ts": "2025-09-03T01:24:57+00:00", "symbol": "Alphabet", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA, AMD, AWS/Amazon, Microsoft, Google Cloud on sustained demand from rising token consumption despite falling unit inference costs.", "resolved_tickers": ["GOOGL"], "resolution_method": "exchange_ticker", "resolution_reason": "Alphabet Inc. GOOGL US", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "AI Inference Costs (Token Pricing) and Their Investment/Industry Implications\n\nAI Inference Costs (Token Pricing)\n\n- Between 2022 and 2025, costs have been dropping by roughly 10x annually.\n\n- The most advanced models (e.g., GPT-4o level) have seen nearly a 900x annual reduction.\n\n- Less sophisticated models only saw about a 9x annual reduction.\n\nDual Cost Structure\n\n- Training: Still extremely expensive → heavy burden on companies.\n\n- Inference: Rapidly becoming cheaper → enabling broader enterprise adoption.\n\nNew Challenge: Token Usage Growth\n\n- More advanced models (with stronger “reasoning/thinking” capabilities) consume more tokens.\n\n- As a result, even if unit costs fall, total usage may drive overall costs higher.\n\nCorporate Dilemma\n\n- Absorb higher costs to maintain high-performance models.\n\n- Cut costs by choosing cheaper, simpler models.\n\n- Ultimately, companies face the hard decision of passing costs to customers vs absorbing them internally.\n\nInvestment & Industry Implications\n\n1. Opportunities for AI Infrastructure Companies\n\n- Falling token prices may drive wider adoption.\n\n- Yet, increased token consumption → continued demand for data centers and GPUs.\n\n- Long-term growth momentum for semiconductors (NVIDIA, AMD) and cloud infrastructure providers (AWS, Microsoft Azure, Google Cloud) remains strong.\n\n2. Risks for SaaS Companies\n\nSaaS firms embedding AI features face margin pressure.\n\nPassing costs to customers risks price resistance, while absorbing them erodes profitability.\n\n3. Shifts in Consumer Choice\n\n- Some consumers may opt for low-cost, lightweight models over premium ones.\n\n- This encourages market diversification and segmentation of AI models.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08371623205250311, "ret_signed_m1d": -0.08371623205250311, "alpha_spy_m1d": -0.07832588564333587, "alpha_c_m1d": -0.06726481305355125, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": -0.01645141899895186, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0071101442827747885, "ret_signed_p1d": 0.0071101442827747885, "alpha_spy_p1d": -0.0012472161590397768, "alpha_c_p1d": -0.0019117203367786662, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.009021864619553455, "ret_p1w": 0.03782155983725555, "ret_signed_p1w": 0.03782155983725555, "alpha_spy_p1w": 0.0246640047945077, "alpha_c_p1w": 0.024023582324787185, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.013797977512468362, "ret_p1m": 0.06611355451852097, "ret_signed_p1m": 0.06611355451852097, "alpha_spy_p1m": 0.023650257811189457, "alpha_c_p1m": 0.03253737183765493, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.03357618268086604, "ret_p3m": 0.36639072940987716, "ret_signed_p3m": 0.36639072940987716, "alpha_spy_p3m": 0.3067143529037326, "alpha_c_p3m": 0.347015645071411, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.01937508433846613, "ret_p6m": 0.3346752386120768, "ret_signed_p6m": 0.3346752386120768, "alpha_spy_p6m": 0.2577598699383521, "alpha_c_p6m": 0.29534870596917173, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.03932653264290509, "price_path": [-0.2479, -0.2366, -0.2257, -0.2311, -0.2383, -0.2427, -0.2336, -0.2372, -0.2486, -0.2486, -0.2776, -0.2838, -0.277, -0.26, -0.2476, -0.226, -0.236, -0.2377, -0.2255, -0.2217, -0.2217, -0.2335, -0.2441, -0.2343, -0.2299, -0.2188, -0.2129, -0.211, -0.2068, -0.2041, -0.1977, -0.1758, -0.1705, -0.1753, -0.1669, -0.1625, -0.1651, -0.1513, -0.148, -0.168, -0.18, -0.1544, -0.156, -0.1499, -0.148, -0.1268, -0.1286, -0.1184, -0.1244, -0.1202, -0.116, -0.1177, -0.1261, -0.1359, -0.134, -0.1065, -0.0961, -0.102, -0.1005, -0.0825, -0.077, -0.077, -0.0837, 0.0, 0.0071, 0.0188, 0.0156, 0.0398, 0.0378, 0.043, 0.0449, 0.0918, 0.0898, 0.0828, 0.0936, 0.1053, 0.0958, 0.092, 0.0724, 0.0665, 0.0698, 0.059, 0.0549, 0.0627, 0.0661, 0.0646, 0.0867, 0.0664, 0.0615, 0.0481, 0.0265, 0.0594, 0.0651, 0.0893, 0.0912, 0.0991, 0.1132, 0.0868, 0.0921, 0.0982, 0.1279, 0.1684, 0.1606, 0.1914, 0.2214, 0.2202, 0.2311, 0.2043, 0.2337, 0.2356, 0.2099, 0.2588, 0.2641, 0.2441, 0.2088, 0.1994, 0.2368, 0.2336, 0.2706, 0.256, 0.3003, 0.3824, 0.4035, 0.3883, 0.3883, 0.3893, 0.3664, 0.3704, 0.387, 0.3782, 0.3941, 0.3622, 0.3768, 0.3904, 0.3566, 0.343, 0.3383, 0.3312, 0.2884, 0.3133, 0.3337, 0.3451, 0.3649, 0.3638, 0.3638, 0.3613, 0.3615, 0.3628, 0.3591, 0.3591, 0.3684, 0.3744, 0.3649, 0.3981, 0.4131, 0.4267, 0.441, 0.4588, 0.4583, 0.445, 0.4329, 0.4329, 0.3982, 0.4259, 0.4352, 0.4239, 0.447, 0.4526, 0.459, 0.4687, 0.4676, 0.4923, 0.4751, 0.4461, 0.4383, 0.4019, 0.4082, 0.3833, 0.3502, 0.3417, 0.3275, 0.3275, 0.3114, 0.3171, 0.315, 0.3677, 0.3525, 0.35, 0.3586, 0.3347]}
{"unit_id": "orig:1945589120488951954", "ticker_idx": 0, "ts": "2025-07-16T20:59:28+00:00", "symbol": "data center", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "related sectors and regional assets are expected to benefit significantly", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley DC deep-dive: bullish global DC/AI infrastructure, nuclear energy, and NVDA on 23% CAGR capacity growth and surging hyperscaler capex.", "resolved_tickers": ["DTCR"], "resolution_method": "etf", "resolution_reason": "Global data center sector proxied by DTCR ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Global Data Center Deep Dive: AI Drives Soaring DC Investment, Energy & Approvals Become Key Expansion Barriers\n\nMorgan Stanley (E. Kelly, 25/07/15)\n\nMorgan Stanley states that the AI investment wave is driving a 23% compound annual growth rate in global data centers. However, factors such as energy supply, approval delays, and building material shortages pose bottlenecks to expansion. Nuclear power and the conversion of Bitcoin mining farms are emerging as important energy solutions, and related sectors and regional assets are expected to benefit significantly.\n\nEnergy Remains the Primary Constraint, But Approvals and Building Material Supply Are Also Increasingly Important\n\nAlphaWise surveys indicate that energy will remain the top obstacle for DC expansion over the next two years, followed by planning permit acquisition, low-carbon building material supply, and cooling technologies. Operators are increasingly focused on data sovereignty and the ability to match green energy sources. DC construction costs are projected to increase by 11% annually over the next two years.\n\nAI Drives High Growth in Global DC Construction, U.S. Remains Core, China and Middle East Accelerate Catch-Up\n\nGlobal DC capacity is expected to grow by 23% annually until 2030, with the U.S. contributing over 60%, and projects like Stargate potentially adding 15GW. China's growth rate has increased to 16%, while Middle Eastern regions like the UAE and KSA are projected to grow 5 to 6 times. Europe, due to difficulties in obtaining approvals and green energy, is expected to lag but still holds potential.\n\nNuclear Power, Natural Gas, and Bitcoin Mining Farm Conversions Key to Future Energy Supply\n\nAs AI drives a multiple increase in energy consumption, nuclear power is becoming a focal point for energy supply. The U.S. has eased nuclear energy approval processes, with Three Mile Island restarting to power Microsoft, and Meta signing a 1.1GW nuclear power purchase agreement. New supercomputing centers in France also rely on the country's nuclear energy advantage.\n\nAI Wave Drives Hyperscale Capital Expenditures, Some Companies Adjust Guidance Upward\n\nMorgan Stanley expects capital expenditures by the four major cloud providers to grow by 43% year-over-year in 2025. Meta has raised its full-year Capex guidance to $64-72 billion, and Nvidia also noted continuous upward revisions in GB200 orders, with AI training and inference demand exceeding expectations. Billions of dollars in AI investments are continually being added in the Middle East.\n\nCross-Industry Beneficiaries Emerge, Multiple Regions and Sectors Poised for Structural Tailwinds\n\nMorgan Stanley, through its global DC stock screening model, has identified relevant beneficiaries in 8 regions and 8 sectors, including real estate, infrastructure, industrial goods, utilities, energy, technology, and telecommunications. These will benefit from the long-term prosperity of DC investment and AI computing infrastructure construction.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009419205576346457, "ret_signed_m1d": -0.009419205576346457, "alpha_spy_m1d": -0.006087154142591089, "alpha_c_m1d": -0.006087154142591089, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.0033320514337553675, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007325926009262407, "ret_signed_p1d": 0.007325926009262407, "alpha_spy_p1d": 0.0012062360317448562, "alpha_c_p1d": 0.0012062360317448562, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.006119689977517551, "ret_p1w": 0.02197797892798059, "ret_signed_p1w": 0.02197797892798059, "alpha_spy_p1w": 0.005973857127799187, "alpha_c_p1w": 0.005973857127799187, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.016004121800181403, "ret_p1m": -0.018838310702596228, "ret_signed_p1m": -0.018838310702596228, "alpha_spy_p1m": -0.052047792815166805, "alpha_c_p1m": -0.052047792815166805, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.033209482112570576, "ret_p3m": 0.13082148490958279, "ret_signed_p3m": 0.13082148490958279, "alpha_spy_p3m": 0.06568698406155682, "alpha_c_p3m": 0.06568698406155682, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.06513450084802597, "ret_p6m": 0.19127737566487912, "ret_signed_p6m": 0.19127737566487912, "alpha_spy_p6m": 0.0803473082172399, "alpha_c_p6m": 0.0803473082172399, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11093006744763922, "price_path": [-0.1885, -0.202, -0.1854, -0.1781, -0.1698, -0.1609, -0.1562, -0.1442, -0.1338, -0.1276, -0.1073, -0.1057, -0.1052, -0.1187, -0.1203, -0.1177, -0.1047, -0.1083, -0.1057, -0.1021, -0.0854, -0.0833, -0.0854, -0.0912, -0.0979, -0.099, -0.099, -0.0886, -0.099, -0.0943, -0.1036, -0.087, -0.0823, -0.063, -0.062, -0.0604, -0.0578, -0.05, -0.0521, -0.0474, -0.0677, -0.0479, -0.0563, -0.0532, -0.0532, -0.0558, -0.0464, -0.0235, -0.037, -0.0412, -0.0319, -0.0246, -0.034, -0.0225, -0.0162, -0.0162, -0.0366, -0.0345, -0.0309, -0.0314, -0.0304, -0.023, -0.0094, 0.0, 0.0073, 0.0107, 0.0094, 0.0126, 0.022, 0.022, 0.0204, 0.0131, 0.0073, -0.0016, -0.0026, -0.0194, -0.0031, -0.0068, -0.0152, -0.0099, -0.0167, -0.0152, -0.0058, -0.0084, -0.0188, -0.0199, -0.0173, -0.0215, -0.0199, -0.0272, -0.0042, -0.011, -0.0094, -0.0089, -0.0052, -0.0105, -0.0105, -0.0304, -0.0408, -0.0413, -0.0293, -0.0309, -0.0157, 0.0031, 0.0277, 0.0319, 0.0382, 0.0387, 0.0471, 0.0586, 0.056, 0.0801, 0.0754, 0.0717, 0.0722, 0.0659, 0.0628, 0.0717, 0.0895, 0.1052, 0.1062, 0.1177, 0.1125, 0.1261, 0.1298, 0.1073, 0.1308, 0.1245, 0.1507, 0.1528, 0.1392, 0.1638, 0.1418, 0.1261, 0.1486, 0.1659, 0.1852, 0.1758, 0.1722, 0.1669, 0.1701, 0.1675, 0.1361, 0.1554, 0.1371, 0.1376, 0.1612, 0.1512, 0.135, 0.0874, 0.0911, 0.0806, 0.0691, 0.0638, 0.034, 0.0324, 0.0628, 0.0633, 0.0748, 0.0748, 0.0942, 0.0863, 0.0874, 0.0973, 0.1026, 0.1172, 0.1371, 0.1397, 0.1329, 0.1334, 0.1047, 0.0822, 0.0733, 0.0529, 0.0654, 0.0863, 0.0968, 0.0955, 0.1041, 0.1041, 0.1036, 0.1068, 0.117, 0.1112, 0.1112, 0.156, 0.1702, 0.1902, 0.1823, 0.1913]}
{"unit_id": "orig:1945589120488951954", "ticker_idx": 1, "ts": "2025-07-16T20:59:28+00:00", "symbol": "nuclear power", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Nuclear power and the conversion of Bitcoin mining farms are emerging as important energy solutions, and related sectors and regional assets are expected to benefit significantly", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley DC deep-dive: bullish global DC/AI infrastructure, nuclear energy, and NVDA on 23% CAGR capacity growth and surging hyperscaler capex.", "resolved_tickers": ["NLR"], "resolution_method": "etf", "resolution_reason": "US nuclear power sector proxied by NLR ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Global Data Center Deep Dive: AI Drives Soaring DC Investment, Energy & Approvals Become Key Expansion Barriers\n\nMorgan Stanley (E. Kelly, 25/07/15)\n\nMorgan Stanley states that the AI investment wave is driving a 23% compound annual growth rate in global data centers. However, factors such as energy supply, approval delays, and building material shortages pose bottlenecks to expansion. Nuclear power and the conversion of Bitcoin mining farms are emerging as important energy solutions, and related sectors and regional assets are expected to benefit significantly.\n\nEnergy Remains the Primary Constraint, But Approvals and Building Material Supply Are Also Increasingly Important\n\nAlphaWise surveys indicate that energy will remain the top obstacle for DC expansion over the next two years, followed by planning permit acquisition, low-carbon building material supply, and cooling technologies. Operators are increasingly focused on data sovereignty and the ability to match green energy sources. DC construction costs are projected to increase by 11% annually over the next two years.\n\nAI Drives High Growth in Global DC Construction, U.S. Remains Core, China and Middle East Accelerate Catch-Up\n\nGlobal DC capacity is expected to grow by 23% annually until 2030, with the U.S. contributing over 60%, and projects like Stargate potentially adding 15GW. China's growth rate has increased to 16%, while Middle Eastern regions like the UAE and KSA are projected to grow 5 to 6 times. Europe, due to difficulties in obtaining approvals and green energy, is expected to lag but still holds potential.\n\nNuclear Power, Natural Gas, and Bitcoin Mining Farm Conversions Key to Future Energy Supply\n\nAs AI drives a multiple increase in energy consumption, nuclear power is becoming a focal point for energy supply. The U.S. has eased nuclear energy approval processes, with Three Mile Island restarting to power Microsoft, and Meta signing a 1.1GW nuclear power purchase agreement. New supercomputing centers in France also rely on the country's nuclear energy advantage.\n\nAI Wave Drives Hyperscale Capital Expenditures, Some Companies Adjust Guidance Upward\n\nMorgan Stanley expects capital expenditures by the four major cloud providers to grow by 43% year-over-year in 2025. Meta has raised its full-year Capex guidance to $64-72 billion, and Nvidia also noted continuous upward revisions in GB200 orders, with AI training and inference demand exceeding expectations. Billions of dollars in AI investments are continually being added in the Middle East.\n\nCross-Industry Beneficiaries Emerge, Multiple Regions and Sectors Poised for Structural Tailwinds\n\nMorgan Stanley, through its global DC stock screening model, has identified relevant beneficiaries in 8 regions and 8 sectors, including real estate, infrastructure, industrial goods, utilities, energy, technology, and telecommunications. These will benefit from the long-term prosperity of DC investment and AI computing infrastructure construction.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02779235144985992, "ret_signed_m1d": -0.02779235144985992, "alpha_spy_m1d": -0.024460300016104553, "alpha_c_m1d": -0.024460300016104553, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.0033320514337553675, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009701042559154605, "ret_signed_p1d": 0.009701042559154605, "alpha_spy_p1d": 0.003581352581637054, "alpha_c_p1d": 0.003581352581637054, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.006119689977517551, "ret_p1w": 0.029977274639261475, "ret_signed_p1w": 0.029977274639261475, "alpha_spy_p1w": 0.013973152839080072, "alpha_c_p1w": 0.013973152839080072, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.016004121800181403, "ret_p1m": 0.013459170584814606, "ret_signed_p1m": 0.013459170584814606, "alpha_spy_p1m": -0.01975031152775597, "alpha_c_p1m": -0.01975031152775597, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.033209482112570576, "ret_p3m": 0.35719274103333043, "ret_signed_p3m": 0.35719274103333043, "alpha_spy_p3m": 0.29205824018530446, "alpha_c_p3m": 0.29205824018530446, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.06513450084802597, "ret_p6m": 0.23802486661416733, "ret_signed_p6m": 0.23802486661416733, "alpha_spy_p6m": 0.1270947991665281, "alpha_c_p6m": 0.1270947991665281, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11093006744763922, "price_path": [-0.3461, -0.3687, -0.3552, -0.3376, -0.315, -0.3185, -0.3062, -0.3035, -0.3101, -0.301, -0.2866, -0.293, -0.2717, -0.268, -0.2723, -0.2617, -0.2582, -0.2428, -0.2328, -0.2304, -0.2305, -0.227, -0.2297, -0.2338, -0.2292, -0.1526, -0.1526, -0.1258, -0.1292, -0.15, -0.1506, -0.1523, -0.137, -0.1387, -0.1469, -0.129, -0.109, -0.1318, -0.081, -0.0799, -0.0821, -0.0407, -0.0561, -0.0484, -0.0484, -0.0644, -0.0547, -0.0298, -0.0511, -0.0321, -0.0437, -0.0283, -0.0594, -0.0524, -0.0517, -0.0517, -0.0462, -0.0714, -0.0809, -0.068, -0.0546, -0.0248, -0.0278, 0.0, 0.0097, 0.0237, 0.0028, 0.0073, 0.03, 0.0535, 0.0571, 0.0357, 0.0163, 0.0131, 0.0016, -0.0207, 0.0173, 0.0442, 0.0535, 0.0426, 0.0315, 0.0258, 0.0417, 0.017, 0.0135, 0.0117, 0.0108, -0.0388, -0.0501, -0.0346, 0.0065, 0.007, 0.0412, 0.0207, 0.0401, 0.0377, 0.0377, 0.0351, 0.0444, 0.0334, 0.0381, 0.055, 0.0674, 0.0753, 0.0796, 0.0696, 0.1425, 0.1103, 0.1131, 0.139, 0.2005, 0.2238, 0.2186, 0.1866, 0.1983, 0.1906, 0.1952, 0.1855, 0.2018, 0.2401, 0.2357, 0.2715, 0.2735, 0.272, 0.2861, 0.2812, 0.3572, 0.3989, 0.4254, 0.3762, 0.3236, 0.3216, 0.2558, 0.2295, 0.2504, 0.293, 0.2771, 0.34, 0.377, 0.3797, 0.3555, 0.3018, 0.2416, 0.2415, 0.1833, 0.1959, 0.2205, 0.1921, 0.1949, 0.1359, 0.1271, 0.1038, 0.1051, 0.1327, 0.0717, 0.0577, 0.0954, 0.1052, 0.1217, 0.1217, 0.136, 0.112, 0.137, 0.1633, 0.2154, 0.1813, 0.1751, 0.178, 0.1709, 0.2073, 0.1343, 0.1056, 0.1024, 0.0635, 0.0936, 0.1378, 0.1473, 0.151, 0.1523, 0.1523, 0.1351, 0.1279, 0.1096, 0.1125, 0.1125, 0.1917, 0.2434, 0.2426, 0.249, 0.238]}
{"unit_id": "orig:1945589120488951954", "ticker_idx": 2, "ts": "2025-07-16T20:59:28+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia also noted continuous upward revisions in GB200 orders, with AI training and inference demand exceeding expectations", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley DC deep-dive: bullish global DC/AI infrastructure, nuclear energy, and NVDA on 23% CAGR capacity growth and surging hyperscaler capex.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Data Center Deep Dive: AI Drives Soaring DC Investment, Energy & Approvals Become Key Expansion Barriers\n\nMorgan Stanley (E. Kelly, 25/07/15)\n\nMorgan Stanley states that the AI investment wave is driving a 23% compound annual growth rate in global data centers. However, factors such as energy supply, approval delays, and building material shortages pose bottlenecks to expansion. Nuclear power and the conversion of Bitcoin mining farms are emerging as important energy solutions, and related sectors and regional assets are expected to benefit significantly.\n\nEnergy Remains the Primary Constraint, But Approvals and Building Material Supply Are Also Increasingly Important\n\nAlphaWise surveys indicate that energy will remain the top obstacle for DC expansion over the next two years, followed by planning permit acquisition, low-carbon building material supply, and cooling technologies. Operators are increasingly focused on data sovereignty and the ability to match green energy sources. DC construction costs are projected to increase by 11% annually over the next two years.\n\nAI Drives High Growth in Global DC Construction, U.S. Remains Core, China and Middle East Accelerate Catch-Up\n\nGlobal DC capacity is expected to grow by 23% annually until 2030, with the U.S. contributing over 60%, and projects like Stargate potentially adding 15GW. China's growth rate has increased to 16%, while Middle Eastern regions like the UAE and KSA are projected to grow 5 to 6 times. Europe, due to difficulties in obtaining approvals and green energy, is expected to lag but still holds potential.\n\nNuclear Power, Natural Gas, and Bitcoin Mining Farm Conversions Key to Future Energy Supply\n\nAs AI drives a multiple increase in energy consumption, nuclear power is becoming a focal point for energy supply. The U.S. has eased nuclear energy approval processes, with Three Mile Island restarting to power Microsoft, and Meta signing a 1.1GW nuclear power purchase agreement. New supercomputing centers in France also rely on the country's nuclear energy advantage.\n\nAI Wave Drives Hyperscale Capital Expenditures, Some Companies Adjust Guidance Upward\n\nMorgan Stanley expects capital expenditures by the four major cloud providers to grow by 43% year-over-year in 2025. Meta has raised its full-year Capex guidance to $64-72 billion, and Nvidia also noted continuous upward revisions in GB200 orders, with AI training and inference demand exceeding expectations. Billions of dollars in AI investments are continually being added in the Middle East.\n\nCross-Industry Beneficiaries Emerge, Multiple Regions and Sectors Poised for Structural Tailwinds\n\nMorgan Stanley, through its global DC stock screening model, has identified relevant beneficiaries in 8 regions and 8 sectors, including real estate, infrastructure, industrial goods, utilities, energy, technology, and telecommunications. These will benefit from the long-term prosperity of DC investment and AI computing infrastructure construction.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003909684872381747, "ret_signed_m1d": -0.003909684872381747, "alpha_spy_m1d": -0.0005776334386263793, "alpha_c_m1d": -0.00916412460569449, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009511589505708296, "ret_signed_p1d": 0.009511589505708296, "alpha_spy_p1d": 0.003391899528190745, "alpha_c_p1d": 0.0011114490994763582, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0034428594862712103, "ret_signed_p1w": -0.0034428594862712103, "alpha_spy_p1w": -0.019446981286452614, "alpha_c_p1w": 0.005786934868547422, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.0621461963170844, "ret_signed_p1m": 0.0621461963170844, "alpha_spy_p1m": 0.028936714204513825, "alpha_c_p1m": 0.01845147637661637, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.09897081121285733, "ret_signed_p3m": 0.09897081121285733, "alpha_spy_p3m": 0.03383631036483137, "alpha_c_p3m": -0.07549684249508726, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.07988992467564904, "ret_signed_p6m": 0.07988992467564904, "alpha_spy_p6m": -0.031040142771990187, "alpha_c_p6m": -0.23426679716688548, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4078, -0.4345, -0.423, -0.4007, -0.379, -0.3523, -0.3656, -0.3639, -0.3645, -0.3488, -0.3319, -0.3359, -0.3375, -0.317, -0.3152, -0.3194, -0.2823, -0.2419, -0.2103, -0.2133, -0.21, -0.209, -0.2159, -0.231, -0.2249, -0.2339, -0.2339, -0.2094, -0.2134, -0.1878, -0.2115, -0.1984, -0.176, -0.1719, -0.1832, -0.1731, -0.1678, -0.16, -0.1665, -0.1539, -0.1716, -0.1557, -0.159, -0.1511, -0.1511, -0.1606, -0.1587, -0.137, -0.0996, -0.0954, -0.0795, -0.0781, -0.1054, -0.0824, -0.0702, -0.0702, -0.0766, -0.0663, -0.0495, -0.0424, -0.0376, -0.0426, -0.0039, 0.0, 0.0095, 0.0061, 0.0001, -0.0253, -0.0034, 0.0138, 0.0124, 0.0314, 0.0242, 0.0461, 0.0379, 0.0137, 0.0504, 0.0402, 0.047, 0.0549, 0.0661, 0.0624, 0.0688, 0.0596, 0.0621, 0.053, 0.0621, 0.0249, 0.0235, 0.0211, 0.0386, 0.0493, 0.0607, 0.0597, 0.0514, 0.0164, 0.0164, -0.0034, -0.0044, 0.0017, -0.0254, -0.0179, -0.0036, 0.0348, 0.0339, 0.0377, 0.0373, 0.0205, -0.0062, 0.0285, 0.031, 0.0715, 0.0413, 0.0327, 0.0369, 0.0399, 0.0612, 0.0888, 0.0927, 0.1023, 0.0949, 0.0827, 0.0798, 0.1036, 0.1238, 0.0689, 0.099, 0.0506, 0.0494, 0.061, 0.0692, 0.0658, 0.0572, 0.0521, 0.063, 0.0869, 0.1175, 0.1731, 0.2082, 0.184, 0.1817, 0.2073, 0.1595, 0.1392, 0.0976, 0.098, 0.1616, 0.1272, 0.131, 0.0905, 0.1098, 0.0889, 0.0584, 0.0885, 0.0542, 0.0439, 0.0653, 0.0377, 0.0519, 0.0519, 0.0329, 0.05, 0.0589, 0.048, 0.0702, 0.0645, 0.0829, 0.0795, 0.0725, 0.0559, 0.0214, 0.0288, 0.0372, -0.0024, 0.0163, 0.0563, 0.072, 0.1042, 0.1007, 0.1007, 0.1119, 0.0984, 0.0945, 0.0884, 0.0884, 0.1021, 0.0979, 0.0927, 0.1036, 0.0799]}
{"unit_id": "orig:1945084400833831013", "ticker_idx": 0, "ts": "2025-07-15T11:33:53+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVDA is the top pick", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (adopted): NVDA top pick, ALAB and MRVL upside potential on AI supply chain strength; hyperscaler Blackwell demand drives 2026 capex growth.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "North American Quarterly Earnings Outlook: AI Momentum Remains Strong, But Inventory and Durability Spark Debate\n\nMorgan Stanley (J. Moore, 25/07/14)\n\nMorgan Stanley believes that AI supply and demand remain strong, but other sectors are affected by inventory replenishment and trade uncertainty. Some company managements are adopting a more conservative stance, and while market sentiment is strengthening, it still shows skepticism.\n\nAI Remains Core Growth Driver, NVDA, ALAB, MRVL Expected to Show Strong Performance\n\nMorgan Stanley emphasizes that AI investment remains the primary driving factor, optimistic about the performance across the entire chain from processors to networking, storage, and memory. NVDA is the top pick, while ALAB and MRVL have upside potential, especially ALAB, which maintains strong fundamentals amidst weak market sentiment. Hyperscale cloud service providers are expected to continue revising up their Blackwell procurement plans, driving sustained expenditure growth in 2026.\n\nPartial Inventory Replenishment Driven by Tariffs, Durability Remains Major Market Debate Point\n\nRecent inventory growth is partly in response to tariff uncertainties. After inventory destocking in 1Q, 2Q/3Q entered a replenishment phase. Although demand is not strong, customers need to maintain higher buffer inventories. In contrast, conservative statements from management have led to stock price pressure; Morgan Stanley tends to favor more conservative players as more noteworthy.\n\n2026 May Face Mismatch Between Chip Shipments and Server Sales\n\nSome Asian companies anticipate that 2026 might experience a phase where \"chip shipments lead server installations,\" fueling skepticism about the market's short-term strength but long-term sustainability. Despite this, Morgan Stanley believes that strong demand for non-server forms and accelerated rack shipments will alleviate this pressure, maintaining an optimistic stance.\n\nAnalog Chips and Consumer Electronics Show Signs of Recovery, But Sustainability of Inventory Replenishment Unclear\n\nAnalog chip companies like ADI and NXPI are expected to show recovery, but there is debate in the market about the durability of their recovery. The PC and smartphone supply chains are also in an inventory replenishment period, but PCs are already showing signs of slowing down. Smartphones have not yet seen clear destocking signals, but investors are generally cautious about this sector.\n\nOverall Outlook Positive But Uncertainties Remain\n\nMorgan Stanley believes that this quarter's earnings reports will generally be positive, with AI remaining the strongest theme. It also favors representative companies in the analog and storage sectors. However, confidence in the outlook still diverges across different sectors, emphasizing that investors should adopt a more cautious attitude toward the recent strong market rally.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03884002840913159, "ret_signed_m1d": -0.03884002840913159, "alpha_spy_m1d": -0.04313153377910328, "alpha_c_m1d": -0.019995527465069607, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00392503050477, "ret_signed_p1d": 0.00392503050477, "alpha_spy_p1d": 0.0005818393862548898, "alpha_c_p1d": 0.009152005413437636, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.021499604475081924, "ret_signed_p1w": -0.021499604475081924, "alpha_spy_p1w": -0.032300987895897304, "alpha_c_p1w": -0.002827009559373761, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.06379609510335871, "ret_signed_p1m": 0.06379609510335871, "alpha_spy_p1m": 0.02722876118029305, "alpha_c_p1m": 0.027757517411518773, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.07305409607893898, "ret_signed_p3m": 0.07305409607893898, "alpha_spy_p3m": 0.020508977938374917, "alpha_c_p3m": -0.045721705615825003, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.10797419003219977, "ret_signed_p6m": 0.10797419003219977, "alpha_spy_p6m": -0.0067831132489872825, "alpha_c_p6m": -0.22014830980357902, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.4055, -0.4055, -0.4323, -0.4207, -0.3983, -0.3766, -0.3497, -0.3631, -0.3614, -0.362, -0.3462, -0.3293, -0.3333, -0.3349, -0.3143, -0.3125, -0.3167, -0.2795, -0.2389, -0.2072, -0.2102, -0.2069, -0.2059, -0.2128, -0.2279, -0.2219, -0.2309, -0.2309, -0.2063, -0.2103, -0.1846, -0.2084, -0.1953, -0.1728, -0.1687, -0.18, -0.1698, -0.1645, -0.1567, -0.1633, -0.1506, -0.1683, -0.1524, -0.1557, -0.1477, -0.1477, -0.1573, -0.1554, -0.1336, -0.096, -0.0919, -0.0759, -0.0745, -0.1019, -0.0788, -0.0665, -0.0665, -0.073, -0.0627, -0.0458, -0.0387, -0.0339, -0.0388, 0.0, 0.0039, 0.0135, 0.01, 0.004, -0.0215, 0.0005, 0.0178, 0.0164, 0.0354, 0.0282, 0.0502, 0.042, 0.0177, 0.0545, 0.0443, 0.0511, 0.059, 0.0703, 0.0665, 0.073, 0.0638, 0.0663, 0.0571, 0.0663, 0.0289, 0.0275, 0.0251, 0.0427, 0.0534, 0.0649, 0.0639, 0.0555, 0.0204, 0.0204, 0.0005, -0.0005, 0.0056, -0.0216, -0.014, 0.0004, 0.0388, 0.038, 0.0418, 0.0414, 0.0245, -0.0023, 0.0325, 0.035, 0.0757, 0.0453, 0.0368, 0.041, 0.0439, 0.0654, 0.0931, 0.097, 0.1066, 0.0992, 0.087, 0.0841, 0.1079, 0.1282, 0.0731, 0.1033, 0.0547, 0.0535, 0.0651, 0.0734, 0.07, 0.0613, 0.0562, 0.0672, 0.0912, 0.1219, 0.1777, 0.213, 0.1886, 0.1863, 0.212, 0.164, 0.1436, 0.1019, 0.1023, 0.1661, 0.1316, 0.1354, 0.0947, 0.1141, 0.0932, 0.0625, 0.0927, 0.0583, 0.048, 0.0695, 0.0418, 0.0561, 0.0561, 0.037, 0.0541, 0.0631, 0.0521, 0.0744, 0.0687, 0.0871, 0.0837, 0.0767, 0.06, 0.0254, 0.0329, 0.0412, 0.0015, 0.0203, 0.0604, 0.0762, 0.1086, 0.105, 0.105, 0.1163, 0.1028, 0.0988, 0.0927, 0.0927, 0.1065, 0.1022, 0.097, 0.108]}
{"unit_id": "orig:1945084400833831013", "ticker_idx": 1, "ts": "2025-07-15T11:33:53+00:00", "symbol": "ALAB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ALAB and MRVL have upside potential, especially ALAB, which maintains strong fundamentals amidst weak market sentiment", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (adopted): NVDA top pick, ALAB and MRVL upside potential on AI supply chain strength; hyperscaler Blackwell demand drives 2026 capex growth.", "resolved_tickers": ["ALAB"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "North American Quarterly Earnings Outlook: AI Momentum Remains Strong, But Inventory and Durability Spark Debate\n\nMorgan Stanley (J. Moore, 25/07/14)\n\nMorgan Stanley believes that AI supply and demand remain strong, but other sectors are affected by inventory replenishment and trade uncertainty. Some company managements are adopting a more conservative stance, and while market sentiment is strengthening, it still shows skepticism.\n\nAI Remains Core Growth Driver, NVDA, ALAB, MRVL Expected to Show Strong Performance\n\nMorgan Stanley emphasizes that AI investment remains the primary driving factor, optimistic about the performance across the entire chain from processors to networking, storage, and memory. NVDA is the top pick, while ALAB and MRVL have upside potential, especially ALAB, which maintains strong fundamentals amidst weak market sentiment. Hyperscale cloud service providers are expected to continue revising up their Blackwell procurement plans, driving sustained expenditure growth in 2026.\n\nPartial Inventory Replenishment Driven by Tariffs, Durability Remains Major Market Debate Point\n\nRecent inventory growth is partly in response to tariff uncertainties. After inventory destocking in 1Q, 2Q/3Q entered a replenishment phase. Although demand is not strong, customers need to maintain higher buffer inventories. In contrast, conservative statements from management have led to stock price pressure; Morgan Stanley tends to favor more conservative players as more noteworthy.\n\n2026 May Face Mismatch Between Chip Shipments and Server Sales\n\nSome Asian companies anticipate that 2026 might experience a phase where \"chip shipments lead server installations,\" fueling skepticism about the market's short-term strength but long-term sustainability. Despite this, Morgan Stanley believes that strong demand for non-server forms and accelerated rack shipments will alleviate this pressure, maintaining an optimistic stance.\n\nAnalog Chips and Consumer Electronics Show Signs of Recovery, But Sustainability of Inventory Replenishment Unclear\n\nAnalog chip companies like ADI and NXPI are expected to show recovery, but there is debate in the market about the durability of their recovery. The PC and smartphone supply chains are also in an inventory replenishment period, but PCs are already showing signs of slowing down. Smartphones have not yet seen clear destocking signals, but investors are generally cautious about this sector.\n\nOverall Outlook Positive But Uncertainties Remain\n\nMorgan Stanley believes that this quarter's earnings reports will generally be positive, with AI remaining the strongest theme. It also favors representative companies in the analog and storage sectors. However, confidence in the outlook still diverges across different sectors, emphasizing that investors should adopt a more cautious attitude toward the recent strong market rally.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022087493525835944, "ret_signed_m1d": -0.022087493525835944, "alpha_spy_m1d": -0.026378998895807637, "alpha_c_m1d": -0.0032429925817739624, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004547403271652262, "ret_signed_p1d": -0.004547403271652262, "alpha_spy_p1d": -0.007890594390167371, "alpha_c_p1d": 0.0006795716370153748, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.2658077402503427, "ret_signed_p1w": 0.2658077402503427, "alpha_spy_p1w": 0.2550063568295273, "alpha_c_p1w": 0.28448033516605087, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 1.0965785849934866, "ret_signed_p1m": 1.0965785849934866, "alpha_spy_p1m": 1.060011251070421, "alpha_c_p1m": 1.0605400073016467, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 1.2326765412640706, "ret_signed_p3m": 1.2326765412640706, "alpha_spy_p3m": 1.1801314231235065, "alpha_c_p3m": 1.1139007395693066, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.794824633791543, "ret_signed_p6m": 0.794824633791543, "alpha_spy_p6m": 0.680067330510356, "alpha_c_p6m": 0.4667021339557642, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.3616, -0.3616, -0.4108, -0.4011, -0.3487, -0.3011, -0.2842, -0.2845, -0.2892, -0.2929, -0.2505, -0.2283, -0.2188, -0.2274, -0.2303, -0.2243, -0.2291, -0.1308, -0.0498, -0.0184, -0.0018, -0.0195, -0.0225, 0.0227, -0.0021, 0.021, 0.0245, 0.0245, 0.0561, 0.0444, 0.046, -0.0178, 0.0061, 0.0296, 0.031, -0.0179, -0.0174, -0.0001, -0.0097, 0.0236, 0.0214, -0.0285, 0.0318, 0.006, 0.0776, 0.0776, 0.0077, -0.0694, -0.0552, -0.0296, 0.0606, -0.0145, -0.021, -0.0401, -0.041, -0.0169, -0.0169, -0.0297, -0.0006, 0.0812, 0.0505, 0.0383, -0.0221, 0.0, -0.0045, 0.0605, 0.1058, 0.3197, 0.2658, 0.2936, 0.3175, 0.3234, 0.3431, 0.282, 0.3953, 0.4804, 0.4194, 0.4934, 0.4675, 0.8882, 0.8503, 0.9411, 0.9427, 1.0788, 1.0966, 1.0646, 1.0122, 1.0185, 0.8521, 0.8672, 0.9222, 0.939, 0.8856, 0.9333, 0.9402, 1.048, 0.9727, 0.9727, 0.8865, 0.9122, 1.035, 1.0702, 1.3398, 1.3223, 1.4943, 1.5217, 1.4848, 1.5042, 1.5854, 1.7001, 1.7272, 1.6548, 1.5693, 1.4914, 1.2225, 1.1724, 1.1414, 1.1524, 1.12, 1.1059, 1.2694, 1.1735, 1.3908, 1.2964, 1.3751, 1.4408, 1.2327, 1.1604, 0.7491, 0.7482, 0.7708, 0.7302, 0.6924, 0.7054, 0.6766, 0.7718, 0.7862, 0.8437, 0.8217, 0.8798, 0.8358, 1.0212, 1.0741, 0.9414, 0.9699, 0.763, 0.7918, 0.8811, 0.8141, 0.7084, 0.5642, 0.5628, 0.5309, 0.5106, 0.5376, 0.5081, 0.5353, 0.5997, 0.5676, 0.6698, 0.6698, 0.706, 0.7885, 0.5476, 0.6511, 0.6513, 0.7457, 0.9028, 0.809, 0.7791, 0.8807, 0.6116, 0.5554, 0.5693, 0.5184, 0.5795, 0.78, 0.869, 0.828, 0.8403, 0.8403, 0.811, 0.8418, 0.8497, 0.8012, 0.8012, 0.9441, 0.8093, 0.7433, 0.7948]}
{"unit_id": "orig:1945084400833831013", "ticker_idx": 2, "ts": "2025-07-15T11:33:53+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ALAB and MRVL have upside potential", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (adopted): NVDA top pick, ALAB and MRVL upside potential on AI supply chain strength; hyperscaler Blackwell demand drives 2026 capex growth.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "North American Quarterly Earnings Outlook: AI Momentum Remains Strong, But Inventory and Durability Spark Debate\n\nMorgan Stanley (J. Moore, 25/07/14)\n\nMorgan Stanley believes that AI supply and demand remain strong, but other sectors are affected by inventory replenishment and trade uncertainty. Some company managements are adopting a more conservative stance, and while market sentiment is strengthening, it still shows skepticism.\n\nAI Remains Core Growth Driver, NVDA, ALAB, MRVL Expected to Show Strong Performance\n\nMorgan Stanley emphasizes that AI investment remains the primary driving factor, optimistic about the performance across the entire chain from processors to networking, storage, and memory. NVDA is the top pick, while ALAB and MRVL have upside potential, especially ALAB, which maintains strong fundamentals amidst weak market sentiment. Hyperscale cloud service providers are expected to continue revising up their Blackwell procurement plans, driving sustained expenditure growth in 2026.\n\nPartial Inventory Replenishment Driven by Tariffs, Durability Remains Major Market Debate Point\n\nRecent inventory growth is partly in response to tariff uncertainties. After inventory destocking in 1Q, 2Q/3Q entered a replenishment phase. Although demand is not strong, customers need to maintain higher buffer inventories. In contrast, conservative statements from management have led to stock price pressure; Morgan Stanley tends to favor more conservative players as more noteworthy.\n\n2026 May Face Mismatch Between Chip Shipments and Server Sales\n\nSome Asian companies anticipate that 2026 might experience a phase where \"chip shipments lead server installations,\" fueling skepticism about the market's short-term strength but long-term sustainability. Despite this, Morgan Stanley believes that strong demand for non-server forms and accelerated rack shipments will alleviate this pressure, maintaining an optimistic stance.\n\nAnalog Chips and Consumer Electronics Show Signs of Recovery, But Sustainability of Inventory Replenishment Unclear\n\nAnalog chip companies like ADI and NXPI are expected to show recovery, but there is debate in the market about the durability of their recovery. The PC and smartphone supply chains are also in an inventory replenishment period, but PCs are already showing signs of slowing down. Smartphones have not yet seen clear destocking signals, but investors are generally cautious about this sector.\n\nOverall Outlook Positive But Uncertainties Remain\n\nMorgan Stanley believes that this quarter's earnings reports will generally be positive, with AI remaining the strongest theme. It also favors representative companies in the analog and storage sectors. However, confidence in the outlook still diverges across different sectors, emphasizing that investors should adopt a more cautious attitude toward the recent strong market rally.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0013809655611551896, "ret_signed_m1d": 0.0013809655611551896, "alpha_spy_m1d": -0.0029105398088165035, "alpha_c_m1d": 0.02022546650521717, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021544097264227235, "ret_signed_p1d": -0.021544097264227235, "alpha_spy_p1d": -0.024887288382742345, "alpha_c_p1d": -0.0163171223555596, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.005800224256477415, "ret_signed_p1w": -0.005800224256477415, "alpha_spy_p1w": -0.016601607677292796, "alpha_c_p1w": 0.012872370659230747, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.09542892165400874, "ret_signed_p1m": 0.09542892165400874, "alpha_spy_p1m": 0.05886158773094308, "alpha_c_p1m": 0.0593903439621688, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.1830781154433161, "ret_signed_p3m": 0.1830781154433161, "alpha_spy_p3m": 0.13053299730275203, "alpha_c_p3m": 0.06430231374855211, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.1696731855588287, "ret_signed_p6m": 0.1696731855588287, "alpha_spy_p6m": 0.05491588227764166, "alpha_c_p6m": -0.15844931427695008, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2866, -0.2866, -0.3186, -0.3015, -0.258, -0.2089, -0.187, -0.19, -0.1903, -0.1946, -0.159, -0.1399, -0.1447, -0.1552, -0.223, -0.2048, -0.1769, -0.11, -0.0964, -0.0906, -0.1003, -0.1202, -0.1367, -0.1525, -0.1705, -0.1465, -0.1625, -0.1625, -0.1194, -0.1087, -0.1206, -0.1694, -0.1518, -0.1395, -0.0851, -0.1009, -0.0568, -0.0459, -0.0501, -0.0584, -0.039, -0.0728, -0.0283, -0.0342, 0.0342, 0.0342, 0.0144, -0.0233, 0.0378, 0.0478, 0.1035, 0.0647, 0.068, 0.052, 0.0246, 0.0374, 0.0374, -0.0127, -0.0072, -0.0029, 0.0123, 0.0041, 0.0014, 0.0, -0.0215, -0.0055, 0.0309, 0.009, -0.0058, 0.0119, 0.0225, 0.0249, 0.0483, 0.0543, 0.1288, 0.1099, 0.0282, 0.0569, 0.0583, 0.0402, 0.0475, 0.0681, 0.0673, 0.0746, 0.0954, 0.0916, 0.0522, 0.0598, -0.0047, -0.0164, -0.0166, 0.0081, 0.0075, 0.0255, 0.0329, 0.0666, -0.1317, -0.1317, -0.1079, -0.1395, -0.1148, -0.1254, -0.0885, -0.0769, -0.0733, -0.0804, -0.0699, -0.0688, -0.049, -0.0197, 0.0251, 0.0255, 0.0431, 0.0305, 0.1061, 0.1574, 0.1486, 0.1378, 0.161, 0.1585, 0.1904, 0.1907, 0.228, 0.2011, 0.2774, 0.2523, 0.1831, 0.2353, 0.1915, 0.2284, 0.2193, 0.2154, 0.1863, 0.1644, 0.1201, 0.1438, 0.1626, 0.2259, 0.2225, 0.2458, 0.224, 0.2954, 0.2489, 0.2104, 0.2838, 0.2898, 0.2565, 0.2884, 0.2345, 0.2345, 0.2095, 0.1947, 0.1532, 0.0873, 0.1238, 0.0597, 0.0703, 0.1579, 0.153, 0.2122, 0.2122, 0.2355, 0.2589, 0.2837, 0.3847, 0.3569, 0.3669, 0.2714, 0.2285, 0.2779, 0.2359, 0.1668, 0.1644, 0.1618, 0.129, 0.1673, 0.1621, 0.1719, 0.2117, 0.1952, 0.1952, 0.1932, 0.1852, 0.199, 0.1744, 0.1744, 0.2353, 0.2469, 0.2193, 0.1697]}
{"unit_id": "orig:1947901094417563963", "ticker_idx": 0, "ts": "2025-07-23T06:06:25+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's HBM competitiveness is expected to continue into 2026 due to its technological leadership and competitors' delays. We expect SK Hynix to maintain over 60% HBM4 market share in FY26.", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM bullish SK Hynix: >60% HBM4 share in FY26, HBM4 price premium +35% at median, DRAM ASP +2% QoQ — author adopts the stance by sharing the note.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM Tech Sales: SK Hynix 2Q25 Earnings Preview\n\n2Q25 Earnings Estimates\n\nOperating Profit (OP): JP Morgan estimates KRW 9 trillion.\n\nCurrent sell-side consensus is around KRW 9.2-9.3 trillion, and the results are expected to converge within this range.\n2Q JPMe Gross Margin (GM): 57%\n\n3Q25 Guidance Estimates\n\n3Q DRAM Bit Growth, Average Selling Price (ASP), HBM Proportion:\n\nJP Morgan anticipates DRAM bit growth of +8% QoQ and +25% YoY.\n\nDRAM ASP is projected to increase by +2% QoQ and +12% YoY.\nHBM revenue proportion is forecasted to be 44%.\n\n2025 Annual Capital Expenditure (CAPEX) Estimates\n\nCAPEX: JP Morgan estimates 2025 CAPEX at KRW 21.284 trillion.\n\nSK Hynix is expected to begin operation of the M15X fab in 4Q, with a possibility of an earlier schedule. Consequently, sell-side and street (market consensus) expect CAPEX to exceed KRW 23 trillion, and the company's guidance is likely to be around this level.\n\nYongin Cluster will commence full-scale operation in 2H 2027. Even if the schedule is accelerated, CAPEX reflection will begin from 2H 2026.\n\nHBM3E / HBM4 Contract Price Outlook\nPrice Forecast:\n\nThe most bearish market scenario suggests a -30% price decline for HBM3E 12Hi and a +20% premium for HBM4 relative to HBM3E.\n\nHowever, we believe there is upside risk. Our analyst's model projects an HBM4 premium of +35% at the median, with HBM3E price decline of <10%.\n\nCurrent consensus estimates an HBM4 premium of 20-30% and an HBM3E 12Hi price decline of 10-20%.\n\nNote: SK Hynix is currently in price negotiations, and as this is a sensitive matter, it is highly probable that specific prices will not be disclosed. Therefore, the 2Q earnings themselves are unlikely to be a catalyst; rather, the results of the price negotiations with Nvidia (NVDA) expected at the end of 3Q will be the key event.\n\n2026 HBM Market Share and Product Mix Outlook\n\nHBM Product Mix and Market Share:\nSK Hynix's HBM competitiveness is expected to continue into 2026 due to its technological leadership and competitors' delays. In particular, the proportion of HBM4 is expected to increase.\n\nNVDA needs to make FY26 volume procurement decisions by the end of 3Q to match Rubin product yields. We expect SK Hynix to maintain over 60% HBM4 market share in FY26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018586874887572824, "ret_signed_m1d": -0.0018586874887572824, "alpha_spy_m1d": 0.006577085943981764, "alpha_c_m1d": 0.0024677192075709398, "bench_spy_m1d": -0.008435773432739047, "bench_c_m1d": -0.004326406696328222, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0018588041452243242, "ret_signed_p1d": 0.0018588041452243242, "alpha_spy_p1d": 0.0015277516223972665, "alpha_c_p1d": -0.0022931833999879636, "bench_spy_p1d": 0.00033105252282705777, "bench_c_p1d": 0.004151987545212288, "ret_p1w": -0.020445795689263524, "ret_signed_p1w": -0.020445795689263524, "alpha_spy_p1w": -0.020839933565311664, "alpha_c_p1w": -0.05411531401470293, "bench_spy_p1w": 0.00039413787604813955, "bench_c_p1w": 0.03366951832543941, "ret_p1m": -0.08921921593322024, "ret_signed_p1m": -0.08921921593322024, "alpha_spy_p1m": -0.091331992939793, "alpha_c_p1m": -0.09218487554735766, "bench_spy_p1m": 0.0021127770065727525, "bench_c_p1m": 0.002965659614137417, "ret_p3m": 0.8074398361769903, "ret_signed_p3m": 0.8074398361769903, "alpha_spy_p3m": 0.7460231390756846, "alpha_c_p3m": 0.5959677769928233, "bench_spy_p3m": 0.061416697101305706, "bench_c_p3m": 0.211472059184167, "ret_p6m": 1.7904056588652266, "ret_signed_p6m": 1.7904056588652266, "alpha_spy_p6m": 1.6926457809363558, "alpha_c_p6m": 1.4030029939371262, "bench_spy_p6m": 0.09775987792887086, "bench_c_p6m": 0.3874026649281004, "price_path": [-0.3157, -0.3246, -0.3291, -0.3413, -0.3413, -0.3098, -0.3098, -0.3098, -0.292, -0.2938, -0.2946, -0.2764, -0.2634, -0.2356, -0.256, -0.2411, -0.2601, -0.2504, -0.256, -0.2693, -0.2578, -0.2467, -0.2486, -0.2282, -0.2119, -0.2398, -0.2286, -0.2286, -0.1914, -0.1654, -0.1654, -0.1487, -0.1431, -0.1078, -0.1245, -0.1245, -0.0781, -0.0743, -0.0836, -0.0855, -0.0446, -0.0353, 0.0353, 0.0632, 0.0892, 0.0558, 0.0855, 0.0613, 0.0372, 0.0353, 0.0056, 0.0074, 0.0483, 0.0446, 0.1041, 0.0948, 0.1152, 0.1097, 0.1004, 0.0019, 0.0, 0.013, -0.0019, 0.0, 0.0019, -0.0112, -0.026, -0.0242, -0.0204, 0.0167, -0.0409, -0.0409, -0.0204, -0.039, -0.026, -0.0465, -0.0074, 0.0, 0.0335, 0.0279, 0.0279, -0.0056, -0.0223, -0.0502, -0.0892, -0.0669, -0.0353, -0.0279, -0.0335, -0.0004, 0.0014, -0.047, -0.0302, -0.0228, -0.0116, 0.0182, 0.0312, 0.0722, 0.1317, 0.1429, 0.223, 0.2323, 0.2955, 0.2416, 0.3235, 0.3235, 0.3067, 0.3439, 0.3309, 0.3272, 0.2527, 0.2993, 0.2937, 0.3402, 0.4724, 0.4724, 0.4724, 0.4724, 0.4724, 0.4724, 0.5934, 0.545, 0.5319, 0.5729, 0.6846, 0.733, 0.8074, 0.7832, 0.7925, 0.7814, 0.8986, 0.9917, 0.9396, 1.0773, 1.1146, 1.0811, 1.3082, 1.1816, 1.1555, 1.2076, 1.1592, 1.256, 1.3044, 1.297, 1.2784, 1.0848, 1.256, 1.122, 1.0922, 1.1257, 0.9396, 0.9359, 0.9322, 0.9508, 1.0267, 0.9745, 1.0043, 1.0788, 1.0565, 1.0192, 1.0267, 1.1496, 1.1086, 1.1869, 1.1049, 1.1273, 1.0639, 0.9745, 1.0528, 1.0565, 1.0379, 1.1608, 1.1757, 1.1906, 1.1906, 1.2316, 1.3843, 1.4253, 1.4253, 1.4253, 1.5222, 1.593, 1.7047, 1.7643, 1.8165, 1.7718, 1.7904, 1.7494, 1.7643, 1.7904]}
{"unit_id": "thread:1967788434522771601", "ticker_idx": 0, "ts": "2025-09-16T03:11:37+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citigroup recently projected Micron's fiscal 2026 revenue at $56 billion... This represents a 48% increase from the company's estimated fiscal 2025 revenue", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's bullish Micron thesis: 48% revenue growth to $56B in FY2026 driven by HBM/AI demand and data center tailwinds.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Citigroup: “Micron to See 48% Revenue Growth Next Year”… Chasing SK\n\nMicron is expected to see a sharp jump in revenue for fiscal 2026 (September 2025–August 2026). With demand for AI memory—particularly high-bandwidth memory (HBM)—surging, the company is forecast to be one of the major beneficiaries.\n\nCitigroup recently projected Micron’s fiscal 2026 revenue at $56 billion (approx. ₩78 trillion). This represents a 48% increase from the company’s estimated fiscal 2025 revenue of $37.7–37.9 billion (approx. ₩53 trillion).\n\nChristopher Danely, an analyst at Citigroup, said: “Micron’s Q4 earnings will be in line with market expectations, but its revenue guidance will far exceed forecasts thanks to rising DRAM/NAND sales and prices. Data center demand and expanding AI investments will translate into significant revenue growth for Micron.”\n\nCitigroup also highlighted that data center demand accounts for 55% of Micron’s revenue, noting that growing investment by cloud service providers (CSPs) will act as a tailwind.\n\nNext year, Micron’s advanced packaging plant in Singapore is set to begin operations, expanding its HBM production capacity. The following year, advanced DRAM production will also begin at its Idaho fab in the U.S.\n\nImportantly, Micron is expected to post around 50% revenue growth for two consecutive years, far outpacing the HBM market growth rate of 33%. This has raised concerns that Micron could emerge as a serious threat to current market leader SK Hynix.\n\nAccording to market research firm Yole, the HBM market is projected to grow from $17 billion in 2024 to $98 billion in 2030, representing a CAGR of 33%. This year, the market size is estimated at around $34 billion.\n\nMeanwhile, Micron is scheduled to announce its Q4 FY2025 earnings on September 23 (local time).\n\n$MU\n\n---\n\nhttps://t.co/qFm5DFi17Q", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006611248522732871, "ret_signed_m1d": -0.006611248522732871, "alpha_spy_m1d": -0.007989939479777441, "alpha_c_m1d": -0.006481109617886616, "bench_spy_m1d": 0.0013786909570445705, "bench_c_m1d": -0.00013013890484625534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0073668280295078326, "ret_signed_p1d": 0.0073668280295078326, "alpha_spy_p1d": 0.00860926332205647, "alpha_c_p1d": 0.013746823069083147, "bench_spy_p1d": -0.0012424352925486382, "bench_c_p1d": -0.006379995039575315, "ret_p1w": 0.04778989867267969, "ret_signed_p1w": 0.04778989867267969, "alpha_spy_p1w": 0.04014027624356098, "alpha_c_p1w": 0.0006234142436600809, "bench_spy_p1w": 0.007649622429118708, "bench_c_p1w": 0.047166484429019606, "ret_p1m": 0.20929475365346706, "ret_signed_p1m": 0.20929475365346706, "alpha_spy_p1m": 0.1986672781323895, "alpha_c_p1m": 0.09676585479501254, "bench_spy_p1m": 0.010627475521077567, "bench_c_p1m": 0.11252889885845452, "ret_p3m": 0.5192734352081192, "ret_signed_p3m": 0.5192734352081192, "alpha_spy_p3m": 0.4834399674221852, "alpha_c_p3m": 0.3665765555225331, "bench_spy_p3m": 0.03583346778593399, "bench_c_p3m": 0.1526968796855861, "ret_p6m": 1.6389714497168235, "ret_signed_p6m": 1.6389714497168235, "alpha_spy_p6m": 1.608351603020767, "alpha_c_p6m": 1.3295158817397232, "bench_spy_p6m": 0.030619846696056552, "bench_c_p6m": 0.3094555679771003, "price_path": [-0.2337, -0.2225, -0.2321, -0.1954, -0.1995, -0.2074, -0.2152, -0.2247, -0.2395, -0.2342, -0.2307, -0.2307, -0.2449, -0.2166, -0.2303, -0.2248, -0.2159, -0.2532, -0.2437, -0.2669, -0.2869, -0.2798, -0.2871, -0.3123, -0.3085, -0.2965, -0.2995, -0.2995, -0.2951, -0.2775, -0.3128, -0.3396, -0.3214, -0.3133, -0.3151, -0.2956, -0.2514, -0.221, -0.1956, -0.2175, -0.2111, -0.2389, -0.2221, -0.2315, -0.262, -0.2709, -0.259, -0.267, -0.2665, -0.2586, -0.2318, -0.2507, -0.2507, -0.254, -0.2525, -0.2179, -0.1728, -0.1723, -0.1485, -0.1185, -0.0519, -0.01, -0.0066, 0.0, 0.0074, 0.0634, 0.0246, 0.0365, 0.0478, 0.0182, -0.0125, -0.0098, 0.032, 0.0535, 0.1469, 0.157, 0.1834, 0.2031, 0.1699, 0.2383, 0.2118, 0.1441, 0.2145, 0.1785, 0.2093, 0.276, 0.2751, 0.3027, 0.2745, 0.2504, 0.3024, 0.3799, 0.3867, 0.3981, 0.4279, 0.4113, 0.4098, 0.4787, 0.3737, 0.4963, 0.5016, 0.499, 0.5959, 0.5191, 0.543, 0.4929, 0.5551, 0.5244, 0.4396, 0.4234, 0.2687, 0.3065, 0.4108, 0.4146, 0.4507, 0.4507, 0.4899, 0.515, 0.5089, 0.4753, 0.428, 0.4946, 0.5557, 0.5903, 0.6615, 0.6284, 0.5193, 0.4963, 0.4649, 0.4209, 0.566, 0.6754, 0.7426, 0.7406, 0.8062, 0.8062, 0.7943, 0.8554, 0.8444, 0.7989, 0.7989, 0.9881, 0.9675, 1.1646, 1.1402, 1.0612, 1.1751, 1.18, 1.1312, 1.1011, 1.1218, 1.2864, 1.2864, 1.3006, 1.4525, 1.5059, 1.519, 1.4524, 1.5857, 1.7435, 1.7468, 1.615, 1.7594, 1.6437, 1.3913, 1.4133, 1.4877, 1.4172, 1.3526, 1.5863, 1.6092, 1.5947, 1.5947, 1.5198, 1.6532, 1.6305, 1.6987, 1.6533, 1.6347, 1.704, 1.6192, 1.5991, 1.601, 1.3931, 1.526, 1.5026, 1.334, 1.4539, 1.5408, 1.639]}
{"unit_id": "orig:1966652328179871881", "ticker_idx": 0, "ts": "2025-09-12T23:57:08+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "That's also why I'm extremely bullish on BESI.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Extremely bullish BESI as dominant hybrid bonder paired with AMAT's CMP monopoly; both set to lead if hybrid bonding enters HBM.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Why is LG developing CMP equipment? feat. Why I’m Bullish on BESI\n\nEveryone’s probably heard that LG Electronics is independently developing an HBM hybrid bonder. What’s interesting is that they’re also said to be developing CMP equipment for HBM hybrid bonding.\n\nWhy might that be?\n\nFirst, let’s look at CMP equipment.\n\nCMP is a technology used to planarize surfaces in both front-end and back-end semiconductor processes, and in hybrid bonding it’s performed right before two chips are mated.\n\nOriginally, CMP wasn’t needed for TC bonding.\n\nThat’s because TC bonding uses bumps as an intermediary to join semiconductor chips, and the spaces between bumps are filled with a compliant molding material, so flatness requirements aren’t stringent.\n\nHybrid bonding is different. Because chips are directly joined without separate bumps, surface planarity basically dictates bonding yield.\n\nThe problem is that AMAT literally has a 100% monopoly on CMP equipment for hybrid bonding.\n\nAMAT also holds a 60% share of the overall CMP market, and in hybrid-bonding CMP, AMAT is, quite literally, the god.\n\nIn April this year, Applied acquired a 9% stake in the Netherlands’ Besi, becoming its largest shareholder. The two companies had already been collaborating in hybrid bonding, and the equity investment made their relationship even tighter.\n\nEven now, in the advanced packaging market for logic, the “Applied (CMP) + Besi (hybrid bonder)” alliance has tremendous influence.\n\nIndustry insiders unanimously say that if hybrid bonding technology is introduced into the HBM (high-bandwidth memory) industry, the pair is likewise highly likely to dominate the market.\n\nJust as Japan’s Lasertec mask inspection tools are bundled alongside ASML’s EUV lithography systems, the two companies’ equipment could be supplied in tandem.\n\n(That’s also why I’m extremely bullish on BESI.)\n\nIt’s reasonable to interpret LG’s decision to pursue CMP development in parallel from the outset—while kicking off hybrid bonder development—as a response to this market structure.\n\nThere are alternatives in the CMP equipment market—Japan’s Ebara and Korea’s KC Tech—but neither has any track record of supplying CMP for hybrid bonding.\n\nSince AMAT has no reason at all to cooperate with LG, LG has no choice but to develop it in-house.\n\nI think any company developing a hybrid bonder will ultimately have to find a solution for CMP equipment one way or another—either by developing it internally or by forming a partnership.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016092944819933397, "ret_signed_m1d": 0.016092944819933397, "alpha_spy_m1d": 0.015758219904827175, "alpha_c_m1d": 0.017374603272491518, "bench_spy_m1d": 0.0003347249151062215, "bench_c_m1d": -0.001281658452558121, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.056325409831226914, "ret_signed_p1d": 0.056325409831226914, "alpha_spy_p1d": 0.051001551538119916, "alpha_c_p1d": 0.04696027991405338, "bench_spy_p1d": 0.005323858293106998, "bench_c_p1d": 0.009365129917173531, "ret_p1w": 0.0862762121000582, "ret_signed_p1w": 0.0862762121000582, "alpha_spy_p1w": 0.0739093670757276, "alpha_c_p1w": 0.04884858642037493, "bench_spy_p1w": 0.012366845024330608, "bench_c_p1w": 0.037427625679683274, "ret_p1m": 0.27626289437842777, "ret_signed_p1m": 0.27626289437842777, "alpha_spy_p1m": 0.2649028457211289, "alpha_c_p1m": 0.15983949904679906, "bench_spy_p1m": 0.011360048657298893, "bench_c_p1m": 0.11642339533162871, "ret_p3m": 0.2556995688935033, "ret_signed_p3m": 0.2556995688935033, "alpha_spy_p3m": 0.2069228924881361, "alpha_c_p3m": 0.026401506500346006, "bench_spy_p3m": 0.0487766764053672, "bench_c_p3m": 0.2292980623931573, "ret_p6m": 0.4868127648597753, "ret_signed_p6m": 0.4868127648597753, "alpha_spy_p6m": 0.44916459173686585, "alpha_c_p6m": 0.18687491136256074, "bench_spy_p6m": 0.03764817312290947, "bench_c_p6m": 0.2999378534972146, "price_path": [0.1341, 0.1243, 0.1149, 0.101, 0.1238, 0.1524, 0.1551, 0.1605, 0.1663, 0.1359, 0.0921, 0.1019, 0.0903, 0.0823, 0.0854, 0.0925, 0.0903, 0.139, 0.1368, 0.1144, 0.1292, 0.0724, 0.1399, 0.1439, 0.1636, 0.1247, 0.1055, 0.1354, 0.0469, 0.0979, 0.0729, 0.0943, 0.0648, 0.0474, 0.0586, 0.0308, 0.0264, 0.0724, 0.0796, 0.0867, 0.1064, 0.1042, 0.0966, 0.0599, 0.0532, 0.0675, 0.0469, 0.0322, 0.0639, 0.0653, 0.0791, 0.0711, 0.0706, 0.0291, 0.0326, -0.0443, -0.0541, -0.042, -0.0282, -0.0089, -0.0036, 0.0045, 0.0161, 0.0, 0.0563, 0.0425, 0.034, 0.1158, 0.0863, 0.0979, 0.1346, 0.1493, 0.1328, 0.0943, 0.1301, 0.1341, 0.1426, 0.194, 0.1667, 0.312, 0.3125, 0.3084, 0.3, 0.2544, 0.2763, 0.2561, 0.3008, 0.2973, 0.2669, 0.2982, 0.2901, 0.2383, 0.2897, 0.3147, 0.3232, 0.3107, 0.3201, 0.3147, 0.3196, 0.3035, 0.2745, 0.2651, 0.2316, 0.1994, 0.2199, 0.232, 0.2101, 0.1962, 0.1748, 0.1641, 0.1368, 0.1542, 0.1627, 0.0979, 0.1122, 0.114, 0.1587, 0.1614, 0.1614, 0.16, 0.1743, 0.2155, 0.2293, 0.2477, 0.3129, 0.2812, 0.2557, 0.224, 0.181, 0.1842, 0.1945, 0.156, 0.1752, 0.169, 0.173, 0.1793, 0.1761, 0.1761, 0.1761, 0.181, 0.19, 0.1958, 0.1958, 0.333, 0.3746, 0.4385, 0.4198, 0.3509, 0.3509, 0.4497, 0.4949, 0.4439, 0.5525, 0.5458, 0.5092, 0.5458, 0.5561, 0.5646, 0.565, 0.5633, 0.5713, 0.515, 0.4533, 0.4707, 0.4716, 0.4318, 0.4086, 0.4372, 0.4797, 0.4966, 0.5114, 0.5279, 0.5083, 0.5753, 0.5959, 0.6102, 0.6719, 0.5467, 0.6495, 0.6777, 0.7193, 0.7667, 0.6889, 0.6933, 0.6853, 0.6214, 0.7076, 0.6866, 0.3974, 0.4868]}
{"unit_id": "orig:1966652328179871881", "ticker_idx": 1, "ts": "2025-09-12T23:57:08+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the 'Applied (CMP) + Besi (hybrid bonder)' alliance has tremendous influence... highly likely to dominate the market", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Extremely bullish BESI as dominant hybrid bonder paired with AMAT's CMP monopoly; both set to lead if hybrid bonding enters HBM.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Why is LG developing CMP equipment? feat. Why I’m Bullish on BESI\n\nEveryone’s probably heard that LG Electronics is independently developing an HBM hybrid bonder. What’s interesting is that they’re also said to be developing CMP equipment for HBM hybrid bonding.\n\nWhy might that be?\n\nFirst, let’s look at CMP equipment.\n\nCMP is a technology used to planarize surfaces in both front-end and back-end semiconductor processes, and in hybrid bonding it’s performed right before two chips are mated.\n\nOriginally, CMP wasn’t needed for TC bonding.\n\nThat’s because TC bonding uses bumps as an intermediary to join semiconductor chips, and the spaces between bumps are filled with a compliant molding material, so flatness requirements aren’t stringent.\n\nHybrid bonding is different. Because chips are directly joined without separate bumps, surface planarity basically dictates bonding yield.\n\nThe problem is that AMAT literally has a 100% monopoly on CMP equipment for hybrid bonding.\n\nAMAT also holds a 60% share of the overall CMP market, and in hybrid-bonding CMP, AMAT is, quite literally, the god.\n\nIn April this year, Applied acquired a 9% stake in the Netherlands’ Besi, becoming its largest shareholder. The two companies had already been collaborating in hybrid bonding, and the equity investment made their relationship even tighter.\n\nEven now, in the advanced packaging market for logic, the “Applied (CMP) + Besi (hybrid bonder)” alliance has tremendous influence.\n\nIndustry insiders unanimously say that if hybrid bonding technology is introduced into the HBM (high-bandwidth memory) industry, the pair is likewise highly likely to dominate the market.\n\nJust as Japan’s Lasertec mask inspection tools are bundled alongside ASML’s EUV lithography systems, the two companies’ equipment could be supplied in tandem.\n\n(That’s also why I’m extremely bullish on BESI.)\n\nIt’s reasonable to interpret LG’s decision to pursue CMP development in parallel from the outset—while kicking off hybrid bonder development—as a response to this market structure.\n\nThere are alternatives in the CMP equipment market—Japan’s Ebara and Korea’s KC Tech—but neither has any track record of supplying CMP for hybrid bonding.\n\nSince AMAT has no reason at all to cooperate with LG, LG has no choice but to develop it in-house.\n\nI think any company developing a hybrid bonder will ultimately have to find a solution for CMP equipment one way or another—either by developing it internally or by forming a partnership.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014004654304025532, "ret_signed_m1d": 0.014004654304025532, "alpha_spy_m1d": 0.01366992938891931, "alpha_c_m1d": 0.015286312756583653, "bench_spy_m1d": 0.0003347249151062215, "bench_c_m1d": -0.001281658452558121, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018653112590935494, "ret_signed_p1d": 0.018653112590935494, "alpha_spy_p1d": 0.013329254297828497, "alpha_c_p1d": 0.009287982673761963, "bench_spy_p1d": 0.005323858293106998, "bench_c_p1d": 0.009365129917173531, "ret_p1w": 0.1328962912990017, "ret_signed_p1w": 0.1328962912990017, "alpha_spy_p1w": 0.1205294462746711, "alpha_c_p1w": 0.09546866561931844, "bench_spy_p1w": 0.012366845024330608, "bench_c_p1w": 0.037427625679683274, "ret_p1m": 0.3079855712556263, "ret_signed_p1m": 0.3079855712556263, "alpha_spy_p1m": 0.2966255225983274, "alpha_c_p1m": 0.1915621759239976, "bench_spy_p1m": 0.011360048657298893, "bench_c_p1m": 0.11642339533162871, "ret_p3m": 0.6429636579563334, "ret_signed_p3m": 0.6429636579563334, "alpha_spy_p3m": 0.5941869815509662, "alpha_c_p3m": 0.41366559556317606, "bench_spy_p3m": 0.0487766764053672, "bench_c_p3m": 0.2292980623931573, "ret_p6m": 1.0263880295348637, "ret_signed_p6m": 1.0263880295348637, "alpha_spy_p6m": 0.9887398564119543, "alpha_c_p6m": 0.7264501760376492, "bench_spy_p6m": 0.03764817312290947, "bench_c_p6m": 0.2999378534972146, "price_path": [0.0345, 0.0271, 0.0271, 0.007, 0.0219, 0.0707, 0.0879, 0.0906, 0.0887, 0.0879, 0.092, 0.1291, 0.1353, 0.1353, 0.1337, 0.1587, 0.1611, 0.1768, 0.1762, 0.1713, 0.1843, 0.1576, 0.144, 0.1317, 0.1446, 0.1121, 0.1113, 0.1179, 0.1035, 0.1307, 0.1196, 0.1254, 0.07, 0.0696, 0.0864, 0.0646, 0.0586, 0.0884, 0.0986, 0.0957, 0.1199, 0.1292, 0.1186, -0.0388, -0.0282, -0.036, -0.0435, -0.0474, -0.0316, -0.0346, -0.0196, -0.0203, -0.0151, -0.042, -0.042, -0.061, -0.0688, -0.057, -0.0301, -0.0343, -0.0256, -0.0261, 0.014, 0.0, 0.0187, 0.0342, 0.0616, 0.1309, 0.1329, 0.195, 0.1971, 0.2005, 0.1895, 0.2153, 0.2214, 0.2201, 0.2976, 0.3325, 0.2964, 0.3344, 0.2608, 0.2962, 0.3129, 0.2512, 0.308, 0.3003, 0.3563, 0.3571, 0.3408, 0.3595, 0.3468, 0.3144, 0.3616, 0.3632, 0.3786, 0.3566, 0.4049, 0.3859, 0.3892, 0.4166, 0.3718, 0.4356, 0.3917, 0.3711, 0.401, 0.3628, 0.375, 0.3303, 0.3469, 0.363, 0.3416, 0.4013, 0.315, 0.3376, 0.3788, 0.4478, 0.4926, 0.4926, 0.5062, 0.5212, 0.5843, 0.604, 0.6089, 0.6003, 0.6012, 0.5951, 0.643, 0.6129, 0.5478, 0.5601, 0.5456, 0.4825, 0.5137, 0.5311, 0.5466, 0.5539, 0.5572, 0.5572, 0.5638, 0.5707, 0.5523, 0.5345, 0.5345, 0.6055, 0.6977, 0.7675, 0.7448, 0.6817, 0.7984, 0.8346, 0.8204, 0.8026, 0.9053, 0.9526, 0.9526, 0.9002, 0.9421, 0.9035, 0.925, 0.9075, 0.9867, 1.0108, 1.0382, 0.9246, 0.9609, 0.9028, 0.777, 0.8152, 0.9258, 0.9739, 0.9649, 1.0295, 0.9609, 1.1192, 1.1192, 1.1444, 1.2051, 1.2111, 1.2442, 1.2333, 1.2595, 1.3612, 1.2463, 1.2258, 1.2251, 1.1004, 1.1389, 1.0718, 0.9415, 1.0264]}
{"unit_id": "orig:1945086781025870033", "ticker_idx": 0, "ts": "2025-07-15T11:43:21+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS expects results to meet guidance with a small upside... 3Q25 guidance may exceed market expectations", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares UBS Neutral on INTC with short-term upside bias into Q2 earnings; 3Q25 guidance may beat on server/PC pull-ins despite structural foundry headwinds.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Intel Outlook: Fundamentals Under Pressure, Strategic Adjustment Focuses on Product Business\n\nUBS (T. Arcuri, 25/07/13)\n\nUBS points out that while INTC may slightly exceed expectations in the short term, its fundamental and strategic uncertainties remain significant, maintaining a \"Neutral\" stance.\n\nShort-Term Upside Bias, But Fundamentals Still Face Structural Challenges\n\nINTC is set to release its earnings after market close on July 24. UBS expects results to meet guidance with a small upside. While 2Q25 originally assumed tariff-induced demand destruction, actual pull-ins were evident, especially in server and PC segments, suggesting 3Q25 guidance may exceed market expectations. More importantly, a significant strategic shift at the company level may be underway.\n\nPotential Foundry Strategy Adjustment, Shifting to Product-Focused Model and Capex Cuts\n\nMedia reports indicate that INTC may abandon external foundry services for its 18A/18AP processes, instead focusing on the 14A process, signaling a potential pivot from the foundry business to an in-house product-centric route. UBS believes this move might be strategically more reasonable but challenging to execute. The company is expected to announce significant capex reductions (possibly up to $5 billion) either this quarter or when communicating its 2026 capital expenditure guidance later this year, aiming to alleviate its current net debt pressure of up to $30 billion.\n\nValuation Adjustment But Maintain Neutral Rating\n\nDespite short-term earnings and order support, the delay in foundry profitability timeline beyond 2027, coupled with increased difficulties in managing capital expenditure and deleveraging, keeps UBS cautious on INTC's fundamentals. Nevertheless, based on a valuation cycle update, UBS maintains a neutral stance.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016579369960807, "ret_signed_m1d": 0.016579369960807, "alpha_spy_m1d": 0.012287864590835307, "alpha_c_m1d": 0.03542387090486898, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01003488400832142, "ret_signed_p1d": -0.01003488400832142, "alpha_spy_p1d": -0.01337807512683653, "alpha_c_p1d": -0.004807909099653784, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.013961592223343589, "ret_signed_p1w": 0.013961592223343589, "alpha_spy_p1w": 0.0031602088025282082, "alpha_c_p1w": 0.03263418713905175, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.030541045401804356, "ret_signed_p1m": -0.030541045401804356, "alpha_spy_p1m": -0.06710837932487002, "alpha_c_p1m": -0.06657962309364429, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.5868236828455378, "ret_signed_p3m": 0.5868236828455378, "alpha_spy_p3m": 0.5342785647049737, "alpha_c_p3m": 0.4680478811507738, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.8599476843897247, "ret_signed_p6m": 0.8599476843897247, "alpha_spy_p6m": 0.7451903811085376, "alpha_c_p6m": 0.5318251845539459, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.1741, -0.1741, -0.178, -0.1488, -0.1017, -0.0624, -0.1252, -0.1051, -0.1126, -0.123, -0.1283, -0.1003, -0.1156, -0.13, -0.1139, -0.0838, -0.0654, -0.0323, -0.0157, -0.0611, -0.0598, -0.055, -0.0672, -0.072, -0.0973, -0.1034, -0.1252, -0.1252, -0.1034, -0.1113, -0.1165, -0.147, -0.1387, -0.1147, -0.1165, -0.1278, -0.1248, -0.1065, -0.0366, -0.0977, -0.0938, -0.1213, -0.0951, -0.0925, -0.0624, -0.0624, -0.0803, -0.0755, -0.0161, -0.0314, -0.0183, -0.01, -0.0227, -0.0031, -0.0454, -0.0188, -0.0188, -0.0401, 0.0292, 0.0227, 0.0393, 0.0223, 0.0166, 0.0, -0.01, -0.0052, 0.0079, 0.0148, 0.014, 0.0249, -0.0127, -0.0969, -0.0977, -0.1095, -0.1126, -0.1361, -0.1575, -0.1492, -0.1191, -0.1095, -0.1374, -0.1296, -0.099, -0.0484, -0.0305, 0.041, 0.0716, 0.0323, 0.1043, 0.0271, 0.0253, 0.082, 0.0711, 0.0624, 0.0842, 0.0877, 0.0624, 0.0624, 0.0563, 0.0471, 0.0737, 0.0685, 0.0681, 0.0663, 0.0807, 0.0737, 0.0506, 0.0807, 0.1025, 0.0864, 0.3338, 0.2906, 0.2548, 0.2801, 0.3621, 0.483, 0.5489, 0.5044, 0.4638, 0.5681, 0.6274, 0.6069, 0.5964, 0.6217, 0.6331, 0.6492, 0.5868, 0.6239, 0.5545, 0.6209, 0.6073, 0.6147, 0.6623, 0.6632, 0.6108, 0.6649, 0.6702, 0.7251, 0.812, 0.8037, 0.7522, 0.7448, 0.7234, 0.6156, 0.6745, 0.6248, 0.6636, 0.6776, 0.6527, 0.6531, 0.5668, 0.5497, 0.5144, 0.4978, 0.5318, 0.4668, 0.5052, 0.5615, 0.5633, 0.606, 0.606, 0.7696, 0.7456, 0.8966, 0.9092, 0.767, 0.8067, 0.7583, 0.767, 0.7792, 0.7238, 0.6497, 0.6366, 0.6278, 0.5729, 0.5829, 0.6065, 0.5868, 0.586, 0.5777, 0.5777, 0.5794, 0.6003, 0.6274, 0.6099, 0.6099, 0.7182, 0.7177, 0.7469, 0.8599]}
{"unit_id": "orig:1950806783343341849", "ticker_idx": 0, "ts": "2025-07-31T06:32:36+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia is expected to be the biggest beneficiary of this investment boom… Nvidia will capture the majority of those orders, further solidifying its dominant position in the AI chip market", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long NVDA: OBBBA tax changes boost Big Tech free cash flow, accelerating AI capex with Nvidia as the dominant chip beneficiary capturing ~80% share through 2030.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: The Biggest Beneficiary of OBBBA?\n\nNvidia, which became the world's first company to surpass a $4 trillion market capitalization earlier this month, could see an unexpected boost from the One Big Beautiful Bill Act (OBBBA), recently signed by President Trump.\n\nAccording to reports, the 'Great America Bill,' signed earlier this month, allows companies to immediately expense capital investments instead of depreciating them over several years. This tax policy change could induce tech giants to significantly increase their spending on Nvidia chips and other AI infrastructure.\n\nMorgan Stanley predicts that the bill and related depreciation policy changes will increase the combined free cash flow of Amazon, Alphabet, and Meta by nearly $49.5 billion in 2025 and about $29.2 billion in 2026. Analysts expect these tech giants to invest the saved funds into infrastructure to meet the surging demand for artificial intelligence and cloud computing.\n\nLess than a month after the bill's passage, signs of accelerating AI investment are already emerging. Alphabet last week raised its full-year capital expenditure forecast by $10 billion (about 13%), citing the need for increased AI investment to meet rapidly growing demand. As the undisputed leader in the AI chip market, Nvidia is expected to be the biggest beneficiary of this investment boom.\n\nTax Reform Benefits Unleash Big Tech's Cash Flow\n\nThe tax changes from the 'Great America Bill,' signed by President Trump on July 4, could become a significant catalyst for tech giants to increase spending on Nvidia chips and other AI infrastructure. Brian Nowak, an analyst at Morgan Stanley, stated in a recent report:\n\n\"We believe these tax incentives may be intended to encourage/enable leading U.S. tech companies to invest more aggressively in the field of generative AI to ensure the U.S. maintains its leading position in the global generative AI competition.\"\n\nAnalysts point out that the impact of this change will be significant. The ability to immediately write off AI investments not only incentivizes tech giants to increase investment to enhance computing power but also helps them save on tax costs.\n\nMorgan Stanley analyst Joseph Moore viewed this as a sign that \"GPU demand has not yet peaked in 2025.\"\n\nMeanwhile, according to previous reports, Microsoft's total capital expenditure for the fourth quarter of its 2025 fiscal year reached a record high of $24.2 billion for a single quarter.\n\nMeta also raised the lower end of its 2025 capital expenditure forecast from $64 billion to $66 billion.\n\nAnalysts expect this pattern of AI spending growth to be replicated among other tech giants. Being able to fully deduct AI investment costs in the year of investment provides a stronger incentive for companies, allowing them to achieve the dual benefits of increasing computing power and saving on taxes.\n\nDriven by the new tax policy, tech giants will use their increased free cash flow to continue expanding the scale of data center construction, thereby alleviating the current supply shortage for AI services.\n\nNvidia's Market Dominance Expected to Solidify\n\nAs the biggest beneficiary of the AI investment boom, Nvidia has seen its quarterly revenue grow more than fourfold over the past two years, reaching a record $44 billion in the first quarter of the current fiscal year.\n\nNvidia chips are one of the largest expenditure items for tech giants building data centers, and the company's share in the fast-growing market is expected to expand further.\n\nBank of America (BoA) predicts that AI accelerators will account for over 65% of all data center spending by 2030, up from about 38% in 2024. During this period, Nvidia is expected to maintain a market share of about 80%.\n\nThis means that as the cash flow unleashed by 'OBBBA' drives tech giants to increase AI investment, Nvidia will capture the majority of those orders, further solidifying its dominant position in the AI chip market.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007870989762204905, "ret_signed_m1d": 0.007870989762204905, "alpha_spy_m1d": 0.004105736298273932, "alpha_c_m1d": -0.01803115146168177, "bench_spy_m1d": 0.003765253463930973, "bench_c_m1d": 0.025902141223886677, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02333162962246693, "ret_signed_p1d": -0.02333162962246693, "alpha_spy_p1d": -0.006941066935938878, "alpha_c_p1d": -0.006606147054694422, "bench_spy_p1d": -0.016390562686528054, "bench_c_p1d": -0.01672548256777251, "ret_p1w": 0.016304058249579922, "ret_signed_p1w": 0.016304058249579922, "alpha_spy_p1w": 0.016035285469017335, "alpha_c_p1w": 0.008201032720051238, "bench_spy_p1w": 0.00026877278056258724, "bench_c_p1w": 0.008103025529528685, "ret_p1m": -0.020745518363826654, "ret_signed_p1m": -0.020745518363826654, "alpha_spy_p1m": -0.041264956819582976, "alpha_c_p1m": -0.02597449755370429, "bench_spy_p1m": 0.020519438455756323, "bench_c_p1m": 0.005228979189877636, "ret_p3m": 0.1302712469576235, "ret_signed_p3m": 0.1302712469576235, "alpha_spy_p3m": 0.04027511592232336, "alpha_c_p3m": -0.12712190651106603, "bench_spy_p3m": 0.08999613103530013, "bench_c_p3m": 0.2573931534686895, "ret_p6m": 0.05521464654806363, "ret_signed_p6m": 0.05521464654806363, "alpha_spy_p6m": -0.04145503472344969, "alpha_c_p6m": -0.3345454502994112, "bench_spy_p6m": 0.09666968127151332, "bench_c_p6m": 0.3897600968474748, "price_path": [-0.3601, -0.3617, -0.3419, -0.3402, -0.3442, -0.3085, -0.2696, -0.2392, -0.242, -0.2388, -0.2379, -0.2446, -0.2591, -0.2533, -0.2619, -0.2619, -0.2383, -0.2421, -0.2175, -0.2403, -0.2277, -0.2061, -0.2022, -0.213, -0.2033, -0.1982, -0.1907, -0.197, -0.1848, -0.2018, -0.1865, -0.1897, -0.1821, -0.1821, -0.1913, -0.1895, -0.1685, -0.1325, -0.1285, -0.1131, -0.1118, -0.1381, -0.1159, -0.1042, -0.1042, -0.1104, -0.1005, -0.0843, -0.0774, -0.0728, -0.0776, -0.0403, -0.0365, -0.0274, -0.0307, -0.0365, -0.0609, -0.0399, -0.0232, -0.0246, -0.0063, -0.0133, 0.0079, 0.0, -0.0233, 0.012, 0.0022, 0.0087, 0.0163, 0.0272, 0.0236, 0.0297, 0.0209, 0.0233, 0.0145, 0.0233, -0.0125, -0.0139, -0.0162, 0.0007, 0.0109, 0.0219, 0.021, 0.0129, -0.0207, -0.0207, -0.0399, -0.0408, -0.0349, -0.061, -0.0537, -0.04, -0.003, -0.0039, -0.0002, -0.0006, -0.0168, -0.0426, -0.0091, -0.0067, 0.0323, 0.0032, -0.005, -0.001, 0.0019, 0.0224, 0.049, 0.0527, 0.062, 0.0549, 0.0432, 0.0404, 0.0633, 0.0827, 0.0298, 0.0588, 0.0122, 0.0111, 0.0222, 0.0301, 0.0269, 0.0186, 0.0136, 0.0242, 0.0472, 0.0766, 0.1303, 0.1641, 0.1407, 0.1385, 0.1632, 0.1171, 0.0975, 0.0575, 0.0579, 0.1191, 0.086, 0.0896, 0.0506, 0.0692, 0.0491, 0.0197, 0.0487, 0.0156, 0.0057, 0.0264, -0.0002, 0.0135, 0.0135, -0.0048, 0.0116, 0.0202, 0.0097, 0.0311, 0.0256, 0.0433, 0.04, 0.0333, 0.0173, -0.0159, -0.0088, -0.0007, -0.0389, -0.0209, 0.0177, 0.0328, 0.0639, 0.0605, 0.0605, 0.0713, 0.0583, 0.0545, 0.0486, 0.0486, 0.0618, 0.0577, 0.0528, 0.0633, 0.0404, 0.0394, 0.0399, 0.0448, 0.0297, 0.0517, 0.0471, 0.0471, 0.0012, 0.0308, 0.0393, 0.0552]}
{"unit_id": "orig:1943951861247356958", "ticker_idx": 0, "ts": "2025-07-12T08:33:35+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raised Amazon's 2026/2027 EPS forecasts to $8/$9 and increased its price target to $300, maintaining an \"Overweight\" rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Morgan Stanley Overweight on AMZN, PT $300, on macro improvement, AWS acceleration, and Anthropic contribution.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Amazon: Macro Improvement and Accelerated AWS Growth Drive Upward Revision in Earnings Forecast\nMorgan Stanley (B. Nowak, 25/07/10) \n\nMorgan Stanley raised Amazon's 2026/2027 EPS forecasts to $8/$9 and increased its price target to $300, maintaining an \"Overweight\" rating, based on an improving macro environment and a more positive outlook for AWS growth. Macro Improvement Drives 2026/2027 EPS Revisions to $8/$9.\n\nBenefiting from easing US-China trade tensions, Morgan Stanley has revised upwards its EPS expectations, which were previously lowered due to a 145% tariff scenario. It now expects 2026 EPS to be $8, consistent with early 2025 forecasts, and 2027 EPS to be $9, representing increases of 9% and 6% respectively. Macroeconomic recovery, improved advertising prospects, and a rebound in e-commerce business margins are the main drivers. \n\nAnthropic's Contribution Fuels AWS Acceleration.\nBased on the latest reports and modeling, Morgan Stanley predicts that Anthropic's annualized contribution to AWS revenue will rise from the current approximately 60 basis points to over 150 basis points in 2026/2027. Anthropic's revenue is projected to reach $10 billion and $19 billion in 2026 and 2027, respectively, becoming a significant engine for AWS's mid-term accelerated growth. Non-Anthropic Business Performance Remains Robust, AWS Fundamentals Resilient.\nEven excluding Anthropic's contribution, AWS has maintained 16% to 19% year-over-year growth over the past five quarters, demonstrating the high resilience of its business fundamentals. With improved GPU supply, GenAI and non-GenAI workloads are expected to accelerate deployment from 2H25, adding further momentum to overall growth. Azure Comparison and CIO Survey Indicate AWS's Potential for Tactical Market Share Expansion.\n\nMicrosoft Azure's experience suggests that initial investments in non-GenAI workloads can drive subsequent GenAI acceleration, a path AWS may replicate. Furthermore, the latest quarterly CIO survey indicates that AWS is tactically projected to surpass MSFT, GOOGL, and ORCL in IT budgets, achieving further market share gains.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0029686656994000193, "ret_signed_m1d": -0.0029686656994000193, "alpha_spy_m1d": -0.001064182450075002, "alpha_c_m1d": 0.0007207721574640225, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": -0.003689437856864042, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0029243814744550622, "ret_signed_p1d": 0.0029243814744550622, "alpha_spy_p1d": 0.0071975485250821425, "alpha_c_p1d": 0.01723246755802388, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": -0.014308086083568816, "ret_p1w": 0.015995394440605715, "ret_signed_p1w": 0.015995394440605715, "alpha_spy_p1w": 0.00965736245588733, "alpha_c_p1w": 0.011630837101680003, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.004364557338925712, "ret_p1m": -0.018698219571328667, "ret_signed_p1m": -0.018698219571328667, "alpha_spy_p1m": -0.04731511964358914, "alpha_c_p1m": -0.03890077475045206, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.020202555179123394, "ret_p3m": 0.009083269216987322, "ret_signed_p3m": 0.009083269216987322, "alpha_spy_p3m": -0.06807746268699844, "alpha_c_p3m": -0.05114122786319175, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.060224497080179074, "ret_p6m": 0.06752620882412197, "ret_signed_p6m": 0.06752620882412197, "alpha_spy_p6m": -0.04605702636409248, "alpha_c_p6m": -0.026052242759619038, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.093578451583741, "price_path": [-0.2276, -0.2352, -0.2352, -0.2586, -0.2327, -0.1998, -0.1735, -0.1626, -0.1683, -0.1697, -0.1829, -0.1573, -0.1582, -0.1743, -0.1802, -0.1639, -0.1489, -0.1446, -0.0755, -0.0634, -0.0684, -0.0909, -0.0891, -0.0865, -0.0958, -0.1089, -0.1001, -0.1094, -0.1094, -0.0872, -0.0929, -0.0886, -0.0916, -0.0844, -0.0885, -0.0818, -0.0788, -0.0537, -0.0386, -0.0358, -0.0553, -0.0552, -0.0602, -0.0425, -0.0482, -0.0584, -0.0584, -0.0709, -0.0763, -0.0572, -0.0607, -0.038, -0.0106, -0.0279, -0.0232, -0.0256, -0.0101, -0.0101, -0.0098, -0.028, -0.014, -0.0152, -0.003, 0.0, 0.0029, -0.0111, -0.008, 0.0019, 0.016, 0.0079, 0.0115, 0.029, 0.0255, 0.0315, 0.0236, 0.0199, 0.0373, -0.0485, -0.0622, -0.0529, -0.015, -0.0113, -0.0133, -0.0195, -0.0187, -0.005, 0.0234, 0.0237, 0.0257, 0.0103, -0.0083, -0.0166, 0.014, 0.01, 0.0134, 0.0152, 0.0262, 0.0147, 0.0147, -0.0016, 0.0013, 0.0443, 0.0294, 0.045, 0.0556, 0.0206, 0.0189, 0.0109, 0.0254, 0.037, 0.0263, 0.0245, 0.0257, 0.0086, -0.0221, -0.0243, -0.0334, -0.0262, -0.0156, -0.0271, -0.0224, -0.0145, -0.0274, -0.0212, -0.0173, -0.0021, 0.0091, -0.0413, -0.0249, -0.0412, -0.0448, -0.0497, -0.0561, -0.0408, -0.0162, -0.0343, -0.0204, -0.0066, 0.0057, 0.0158, 0.0204, -0.0125, 0.0821, 0.1254, 0.1047, 0.1086, 0.0769, 0.0829, 0.1006, 0.1037, 0.082, 0.0527, 0.0399, 0.0318, -0.0139, -0.0133, -0.0379, -0.0222, 0.0026, 0.0176, 0.0154, 0.0154, 0.0334, 0.0363, 0.0387, 0.0296, 0.0152, 0.017, 0.0053, 0.0099, 0.027, 0.0203, 0.0022, -0.014, -0.0139, -0.0196, 0.0047, 0.0074, 0.0121, 0.0286, 0.0296, 0.0296, 0.0303, 0.0283, 0.0303, 0.0227, 0.0227, 0.0036, 0.0327, 0.0675]}
{"unit_id": "orig:1968838703058669977", "ticker_idx": 0, "ts": "2025-09-19T00:45:00+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "despite the noise around Micron's HBM4, I believe one should actually take a long position on the company", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long MU: even if HBM4 misses NVIDIA qualification, strong ASIC/CSP HBM demand, soaring commodity memory prices, and diversifying customer base make shorting extremely risky.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Recently, after I covered some negative aspects about Micron, I received inquiries from several hedge funds and research firms.\n\nThey asked me whether they should short Micron.\n\nMeanwhile, some passionate Micron fanboys began labeling me as a Micron short-seller.\n\nBut in fact, despite the noise around Micron’s HBM4, I believe one should actually take a long position on the company.\n\nLet me explain why by quoting the perspective of Mr. Peter Lee, Managing Director at Citigroup, whom I personally respect.\n\nLast year, NVIDIA accounted for 80% of total HBM demand, this year it will be 70%, by 2026 it will fall to 60%, and by 2027 to 50%. In other words, ASICs will take up an increasingly larger share.\n\nThat’s why, even if Micron fails to qualify its HBM4 at NVIDIA, I don’t think the financial impact will be nearly as severe as what Samsung faced in the past.\n\nWhen Samsung failed its HBM3E qualification last year and this year, commodity memory prices were also weak, and HBM demand from ASIC players (whose qualification standards are far looser than NVIDIA’s) was not very strong either. Today, however, the situation is the complete opposite. Demand for HBM from CSP-targeted ASICs is extremely strong, and commodity memory prices are soaring.\n\nThis is why Micron’s stock has remained resilient despite rumors about its HBM4 performance.\n\nFurthermore, the HBM market going forward will become increasingly segmented—on one end, ultra-high-performance, high-priced products for NVIDIA-like customers, and on the other, lower-priced, lower-performance products. The rise of custom HBM will accelerate this segmentation trend even further.\n\nFinally, the fact that NVIDIA’s share of HBM demand is steadily declining means that suppliers will gain more bargaining power in HBM pricing.\n\nIn other words, even if you can’t sell to NVIDIA, there are plenty of other customers to sell HBM to.\n\nThat’s why I think shorting Micron is extremely risky.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03785413036315055, "ret_signed_m1d": 0.03785413036315055, "alpha_spy_m1d": 0.042782595069955076, "alpha_c_m1d": 0.03386308790693349, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": 0.003991042456217064, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011614370666173546, "ret_signed_p1d": 0.011614370666173546, "alpha_spy_p1d": 0.006883362698077322, "alpha_c_p1d": -0.008847479050133833, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.02046184971630738, "ret_p1w": -0.03355240726058739, "ret_signed_p1w": -0.03355240726058739, "alpha_spy_p1w": -0.030719851037440016, "alpha_c_p1w": -0.05290566639375738, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.019353259133169987, "ret_p1m": 0.27142811666693745, "ret_signed_p1m": 0.27142811666693745, "alpha_spy_p1m": 0.2599771980769172, "alpha_c_p1m": 0.17162131764103883, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.09980679902589862, "ret_p3m": 0.3867218081022077, "ret_signed_p3m": 0.3867218081022077, "alpha_spy_p3m": 0.37512023048298104, "alpha_c_p3m": 0.31219141737081224, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.07453039073139545, "ret_p6m": 1.717724235415905, "ret_signed_p6m": 1.717724235415905, "alpha_spy_p6m": 1.7067148430416195, "alpha_c_p6m": 1.4661144131462567, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": 0.25160982226964834, "price_path": [-0.2147, -0.2188, -0.2264, -0.2341, -0.2433, -0.2578, -0.2526, -0.2492, -0.2492, -0.2631, -0.2354, -0.2488, -0.2435, -0.2347, -0.2711, -0.2619, -0.2845, -0.304, -0.2971, -0.3042, -0.3288, -0.3251, -0.3134, -0.3163, -0.3164, -0.312, -0.2949, -0.3293, -0.3555, -0.3377, -0.3298, -0.3315, -0.3125, -0.2694, -0.2397, -0.215, -0.2363, -0.2301, -0.2572, -0.2408, -0.25, -0.2797, -0.2885, -0.2768, -0.2846, -0.2841, -0.2764, -0.2503, -0.2687, -0.2687, -0.2719, -0.2704, -0.2367, -0.1927, -0.1922, -0.1689, -0.1397, -0.0747, -0.0338, -0.0305, -0.024, -0.0168, 0.0379, 0.0, 0.0116, 0.0226, -0.0063, -0.0363, -0.0336, 0.0072, 0.0282, 0.1193, 0.1292, 0.155, 0.1742, 0.1418, 0.2085, 0.1826, 0.1167, 0.1853, 0.1502, 0.1802, 0.2454, 0.2444, 0.2714, 0.2439, 0.2204, 0.2711, 0.3468, 0.3534, 0.3645, 0.3935, 0.3774, 0.376, 0.4432, 0.3407, 0.4604, 0.4655, 0.463, 0.5575, 0.4826, 0.5059, 0.457, 0.5178, 0.4877, 0.405, 0.3892, 0.2382, 0.2751, 0.3769, 0.3806, 0.4159, 0.4159, 0.4541, 0.4786, 0.4726, 0.4398, 0.3937, 0.4587, 0.5183, 0.5521, 0.6216, 0.5893, 0.4828, 0.4604, 0.4297, 0.3867, 0.5283, 0.6351, 0.7008, 0.6988, 0.7628, 0.7628, 0.7512, 0.8108, 0.8001, 0.7557, 0.7557, 0.9403, 0.9202, 1.1126, 1.0887, 1.0117, 1.1228, 1.1276, 1.08, 1.0506, 1.0708, 1.2314, 1.2314, 1.2453, 1.3936, 1.4457, 1.4584, 1.3935, 1.5236, 1.6776, 1.6808, 1.5521, 1.6931, 1.5802, 1.3339, 1.3553, 1.4279, 1.3591, 1.296, 1.5242, 1.5465, 1.5323, 1.5323, 1.4592, 1.5895, 1.5673, 1.6339, 1.5896, 1.5714, 1.639, 1.5563, 1.5367, 1.5385, 1.3356, 1.4653, 1.4424, 1.2779, 1.3949, 1.4797, 1.5756, 1.4935, 1.6213, 1.7177]}
{"unit_id": "orig:1968922002779160883", "ticker_idx": 0, "ts": "2025-09-19T06:16:00+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": "foundry", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "raised its price target from $25 to $35", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS note: raises INTC PT $25→$35 on NVIDIA partnership/SOTP; KLAC benefits from improved CapEx outlook; AMD faces medium-to-long-term competitive pressure.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Intel: NVIDIA and Intel Reach $5 Billion Investment Partnership, Potentially Driving Intel Foundry Spin-Off and Long-Term Valuation Re-Rating\n\nUBS (T. Arcuri, 250918)\n\nUBS stated that NVIDIA announced a $5 billion joint investment with Intel to develop AI infrastructure and PC products. This follows the U.S. government’s conversion of CHIPS Act commitments into an equity stake in Intel.\n\nInvestment Background and Comparisons\n\nUBS pointed out that this partnership extends the logic of government injections being followed by customer investments. Intel’s manufacturing roadmap focus has already shifted from 18A to the 14A process node, which is expected to go into mass production in 2029. Similar to how Intel, Samsung, and TSMC once invested in ASML to accelerate EUV adoption, this investment increases the likelihood that the market will evaluate Intel using a sum-of-the-parts (SOTP) approach, with a reasonable valuation range of $35–$40 per share.\n\nDual Significance: Products and Foundry\n\nUBS believes this investment is more significant for Intel’s foundry business, though it also carries potential implications for Intel’s product side. While NVIDIA remains committed to the Grace CPU roadmap, it may in the future launch Rubin or Rubin Ultra systems with x86 CPUs, which would be a mild positive for Intel’s product business. More importantly, the funding will allow Intel to continue investing in the 14A node and may pave the way for a split between its foundry and product businesses.\n\nIndustry Chain Impacts and Competitive Landscape\n\nUBS emphasized that this move first benefits semiconductor equipment makers, especially KLA (KLAC), given the improved long-term CapEx outlook for Intel. However, it will not significantly alter the short-term expectation of lower CapEx in 2026. The market may interpret this as a negative for AMD, but UBS views the impact as more medium-to-long term. If Intel does spin off businesses and opens up to TSMC capacity, it could put competitive pressure on AMD, though this would take 3–4 years to materialize.\n\nValuation and Target Price\n\nBased on SOTP analysis of Intel’s five business segments (Intel Products, Intel Foundry, Altera, Mobileye, and Others), UBS applied differentiated valuation methods and ultimately raised its price target from $25 to $35.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03346855218632894, "ret_signed_m1d": 0.03346855218632894, "alpha_spy_m1d": 0.03839701689313346, "alpha_c_m1d": 0.029477509730111873, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": 0.003991042456217064, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027721423155483316, "ret_signed_p1d": -0.027721423155483316, "alpha_spy_p1d": -0.03245243112357954, "alpha_c_p1d": -0.048183272871790694, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.02046184971630738, "ret_p1w": 0.20013522959983265, "ret_signed_p1w": 0.20013522959983265, "alpha_spy_p1w": 0.20296778582298003, "alpha_c_p1w": 0.18078197046666267, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.019353259133169987, "ret_p1m": 0.288032406098383, "ret_signed_p1m": 0.288032406098383, "alpha_spy_p1m": 0.27658148750836276, "alpha_c_p1m": 0.18822560707248437, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.09980679902589862, "ret_p3m": 0.21872884820967475, "ret_signed_p3m": 0.21872884820967475, "alpha_spy_p3m": 0.2071272705904481, "alpha_c_p3m": 0.1441984574782793, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.07453039073139545, "ret_p6m": 0.5469911575239776, "ret_signed_p6m": 0.5469911575239776, "alpha_spy_p6m": 0.5359817651496921, "alpha_c_p6m": 0.2953813352543293, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": 0.25160982226964834, "price_path": [-0.2377, -0.2495, -0.2394, -0.2329, -0.2427, -0.2275, -0.2603, -0.2397, -0.2397, -0.2563, -0.2025, -0.2076, -0.1947, -0.2079, -0.2123, -0.2252, -0.2329, -0.2292, -0.2191, -0.2137, -0.2143, -0.2059, -0.235, -0.3002, -0.3009, -0.31, -0.3124, -0.3306, -0.3472, -0.3408, -0.3174, -0.31, -0.3316, -0.3256, -0.3019, -0.2627, -0.2488, -0.1934, -0.1697, -0.2001, -0.1444, -0.2042, -0.2055, -0.1616, -0.17, -0.1768, -0.1599, -0.1572, -0.1768, -0.1768, -0.1815, -0.1886, -0.168, -0.1721, -0.1724, -0.1738, -0.1626, -0.168, -0.1859, -0.1626, -0.1457, -0.1582, 0.0335, 0.0, -0.0277, -0.0081, 0.0554, 0.1491, 0.2001, 0.1657, 0.1342, 0.215, 0.261, 0.2451, 0.237, 0.2566, 0.2654, 0.2779, 0.2295, 0.2583, 0.2045, 0.2559, 0.2454, 0.2512, 0.288, 0.2887, 0.2481, 0.2901, 0.2941, 0.3367, 0.404, 0.3976, 0.3577, 0.3519, 0.3354, 0.2519, 0.2975, 0.259, 0.289, 0.2999, 0.2806, 0.2809, 0.214, 0.2008, 0.1734, 0.1606, 0.187, 0.1366, 0.1663, 0.2099, 0.2113, 0.2444, 0.2444, 0.3712, 0.3526, 0.4696, 0.4794, 0.3692, 0.3999, 0.3624, 0.3692, 0.3786, 0.3357, 0.2782, 0.2681, 0.2613, 0.2187, 0.2265, 0.2448, 0.2295, 0.2289, 0.2224, 0.2224, 0.2238, 0.24, 0.261, 0.2475, 0.2475, 0.3313, 0.331, 0.3536, 0.4412, 0.3898, 0.5399, 0.4895, 0.5987, 0.6471, 0.6335, 0.5876, 0.5876, 0.6416, 0.834, 0.8364, 0.5237, 0.4364, 0.4851, 0.6491, 0.645, 0.571, 0.6501, 0.665, 0.643, 0.6308, 0.7103, 0.6984, 0.5933, 0.6325, 0.5713, 0.5818, 0.5818, 0.5612, 0.5368, 0.5085, 0.4912, 0.475, 0.5592, 0.5849, 0.5368, 0.5419, 0.5382, 0.4571, 0.5409, 0.5534, 0.4679, 0.5409, 0.5815, 0.622, 0.5297, 0.5473, 0.547]}
{"unit_id": "orig:1968922002779160883", "ticker_idx": 1, "ts": "2025-09-19T06:16:00+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this move first benefits semiconductor equipment makers, especially KLA (KLAC), given the improved long-term CapEx outlook for Intel", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS note: raises INTC PT $25→$35 on NVIDIA partnership/SOTP; KLAC benefits from improved CapEx outlook; AMD faces medium-to-long-term competitive pressure.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Intel: NVIDIA and Intel Reach $5 Billion Investment Partnership, Potentially Driving Intel Foundry Spin-Off and Long-Term Valuation Re-Rating\n\nUBS (T. Arcuri, 250918)\n\nUBS stated that NVIDIA announced a $5 billion joint investment with Intel to develop AI infrastructure and PC products. This follows the U.S. government’s conversion of CHIPS Act commitments into an equity stake in Intel.\n\nInvestment Background and Comparisons\n\nUBS pointed out that this partnership extends the logic of government injections being followed by customer investments. Intel’s manufacturing roadmap focus has already shifted from 18A to the 14A process node, which is expected to go into mass production in 2029. Similar to how Intel, Samsung, and TSMC once invested in ASML to accelerate EUV adoption, this investment increases the likelihood that the market will evaluate Intel using a sum-of-the-parts (SOTP) approach, with a reasonable valuation range of $35–$40 per share.\n\nDual Significance: Products and Foundry\n\nUBS believes this investment is more significant for Intel’s foundry business, though it also carries potential implications for Intel’s product side. While NVIDIA remains committed to the Grace CPU roadmap, it may in the future launch Rubin or Rubin Ultra systems with x86 CPUs, which would be a mild positive for Intel’s product business. More importantly, the funding will allow Intel to continue investing in the 14A node and may pave the way for a split between its foundry and product businesses.\n\nIndustry Chain Impacts and Competitive Landscape\n\nUBS emphasized that this move first benefits semiconductor equipment makers, especially KLA (KLAC), given the improved long-term CapEx outlook for Intel. However, it will not significantly alter the short-term expectation of lower CapEx in 2026. The market may interpret this as a negative for AMD, but UBS views the impact as more medium-to-long term. If Intel does spin off businesses and opens up to TSMC capacity, it could put competitive pressure on AMD, though this would take 3–4 years to materialize.\n\nValuation and Target Price\n\nBased on SOTP analysis of Intel’s five business segments (Intel Products, Intel Foundry, Altera, Mobileye, and Others), UBS applied differentiated valuation methods and ultimately raised its price target from $25 to $35.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0017992953239089893, "ret_signed_m1d": 0.0017992953239089893, "alpha_spy_m1d": 0.006727760030713514, "alpha_c_m1d": -0.002191747132308075, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": 0.003991042456217064, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025305941804676158, "ret_signed_p1d": 0.025305941804676158, "alpha_spy_p1d": 0.020574933836579934, "alpha_c_p1d": 0.004844092088368779, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.02046184971630738, "ret_p1w": 0.01864450113173155, "ret_signed_p1w": 0.01864450113173155, "alpha_spy_p1w": 0.021477057354878926, "alpha_c_p1w": -0.0007087580014384365, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.019353259133169987, "ret_p1m": 0.10344452121907155, "ret_signed_p1m": 0.10344452121907155, "alpha_spy_p1m": 0.09199360262905132, "alpha_c_p1m": 0.003637722193172932, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.09980679902589862, "ret_p3m": 0.12363616993091808, "ret_signed_p3m": 0.12363616993091808, "alpha_spy_p3m": 0.11203459231169144, "alpha_c_p3m": 0.04910577919952264, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.07453039073139545, "ret_p6m": 0.38065767500068337, "ret_signed_p6m": 0.38065767500068337, "alpha_spy_p6m": 0.3696482826263978, "alpha_c_p6m": 0.12904785273103503, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": 0.25160982226964834, "price_path": [-0.1509, -0.1467, -0.1377, -0.1501, -0.1445, -0.1416, -0.1203, -0.117, -0.117, -0.1284, -0.1221, -0.1183, -0.1131, -0.117, -0.1196, -0.1056, -0.1085, -0.1051, -0.1107, -0.1044, -0.1479, -0.1432, -0.1365, -0.1385, -0.1184, -0.1251, -0.1166, -0.1605, -0.1532, -0.1256, -0.1563, -0.1517, -0.129, -0.1263, -0.1307, -0.1065, -0.0932, -0.0876, -0.1644, -0.1545, -0.1615, -0.1592, -0.165, -0.167, -0.1582, -0.15, -0.1492, -0.1443, -0.1654, -0.1654, -0.1899, -0.1923, -0.1642, -0.1337, -0.13, -0.1216, -0.1074, -0.0819, -0.0773, -0.0535, -0.0519, -0.0526, 0.0018, 0.0, 0.0253, 0.0253, 0.0228, 0.0137, 0.0186, 0.0185, 0.0323, 0.0805, 0.0904, 0.0543, 0.0908, 0.0382, 0.017, 0.0083, -0.0594, -0.019, -0.0183, 0.0404, 0.0517, 0.0592, 0.1034, 0.0982, 0.0665, 0.1093, 0.1321, 0.163, 0.1543, 0.1823, 0.1623, 0.1569, 0.1669, 0.1423, 0.1745, 0.1547, 0.1422, 0.1657, 0.1398, 0.1475, 0.1119, 0.0857, 0.0869, 0.0767, 0.1193, 0.0569, 0.0518, 0.0898, 0.0986, 0.1112, 0.1112, 0.1269, 0.1094, 0.1407, 0.1617, 0.1582, 0.1643, 0.174, 0.175, 0.1878, 0.1947, 0.1446, 0.1745, 0.1729, 0.1236, 0.1719, 0.1942, 0.2134, 0.2164, 0.2243, 0.2243, 0.2268, 0.2084, 0.1923, 0.1649, 0.1649, 0.2219, 0.2966, 0.3374, 0.3036, 0.2699, 0.3422, 0.3692, 0.3823, 0.3753, 0.4812, 0.5031, 0.5031, 0.4248, 0.4573, 0.4381, 0.4503, 0.4793, 0.5496, 0.56, 0.6152, 0.369, 0.3522, 0.2996, 0.2533, 0.2761, 0.3834, 0.3807, 0.3718, 0.4184, 0.391, 0.4037, 0.4037, 0.4113, 0.421, 0.4111, 0.4361, 0.4281, 0.4463, 0.4848, 0.4633, 0.4635, 0.4735, 0.3836, 0.4168, 0.3721, 0.2907, 0.3719, 0.3948, 0.4063, 0.3531, 0.3618, 0.3807]}
{"unit_id": "orig:1968922002779160883", "ticker_idx": 2, "ts": "2025-09-19T06:16:00+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The market may interpret this as a negative for AMD", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS note: raises INTC PT $25→$35 on NVIDIA partnership/SOTP; KLAC benefits from improved CapEx outlook; AMD faces medium-to-long-term competitive pressure.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Intel: NVIDIA and Intel Reach $5 Billion Investment Partnership, Potentially Driving Intel Foundry Spin-Off and Long-Term Valuation Re-Rating\n\nUBS (T. Arcuri, 250918)\n\nUBS stated that NVIDIA announced a $5 billion joint investment with Intel to develop AI infrastructure and PC products. This follows the U.S. government’s conversion of CHIPS Act commitments into an equity stake in Intel.\n\nInvestment Background and Comparisons\n\nUBS pointed out that this partnership extends the logic of government injections being followed by customer investments. Intel’s manufacturing roadmap focus has already shifted from 18A to the 14A process node, which is expected to go into mass production in 2029. Similar to how Intel, Samsung, and TSMC once invested in ASML to accelerate EUV adoption, this investment increases the likelihood that the market will evaluate Intel using a sum-of-the-parts (SOTP) approach, with a reasonable valuation range of $35–$40 per share.\n\nDual Significance: Products and Foundry\n\nUBS believes this investment is more significant for Intel’s foundry business, though it also carries potential implications for Intel’s product side. While NVIDIA remains committed to the Grace CPU roadmap, it may in the future launch Rubin or Rubin Ultra systems with x86 CPUs, which would be a mild positive for Intel’s product business. More importantly, the funding will allow Intel to continue investing in the 14A node and may pave the way for a split between its foundry and product businesses.\n\nIndustry Chain Impacts and Competitive Landscape\n\nUBS emphasized that this move first benefits semiconductor equipment makers, especially KLA (KLAC), given the improved long-term CapEx outlook for Intel. However, it will not significantly alter the short-term expectation of lower CapEx in 2026. The market may interpret this as a negative for AMD, but UBS views the impact as more medium-to-long term. If Intel does spin off businesses and opens up to TSMC capacity, it could put competitive pressure on AMD, though this would take 3–4 years to materialize.\n\nValuation and Target Price\n\nBased on SOTP analysis of Intel’s five business segments (Intel Products, Intel Foundry, Altera, Mobileye, and Others), UBS applied differentiated valuation methods and ultimately raised its price target from $25 to $35.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003367423478951581, "ret_signed_m1d": -0.003367423478951581, "alpha_spy_m1d": -0.008295888185756106, "alpha_c_m1d": 0.0006236189772654832, "bench_spy_m1d": -0.004928464706804525, "bench_c_m1d": 0.003991042456217064, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015248706434916048, "ret_signed_p1d": -0.015248706434916048, "alpha_spy_p1d": -0.010517698466819825, "alpha_c_p1d": 0.00521314328139133, "bench_spy_p1d": 0.004731007968096224, "bench_c_p1d": 0.02046184971630738, "ret_p1w": 0.01315208928296685, "ret_signed_p1w": -0.01315208928296685, "alpha_spy_p1w": -0.015984645506114226, "alpha_c_p1w": 0.0062011698502031365, "bench_spy_p1w": -0.0028325562231473755, "bench_c_p1w": 0.019353259133169987, "ret_p1m": 0.5284325464862758, "ret_signed_p1m": -0.5284325464862758, "alpha_spy_p1m": -0.5169816278962556, "alpha_c_p1m": -0.4286257474603772, "bench_spy_p1m": 0.011450918590020231, "bench_c_p1m": 0.09980679902589862, "ret_p3m": 0.2587203848949331, "ret_signed_p3m": -0.2587203848949331, "alpha_spy_p3m": -0.24711880727570645, "alpha_c_p3m": -0.18418999416353765, "bench_spy_p3m": 0.011601577619226644, "bench_c_p3m": 0.07453039073139545, "ret_p6m": 0.24899931757661453, "ret_signed_p6m": -0.24899931757661453, "alpha_spy_p6m": -0.23798992520232898, "alpha_c_p6m": 0.0026105046930338105, "bench_spy_p6m": 0.011009392374285554, "bench_c_p6m": 0.25160982226964834, "price_path": [-0.1205, -0.0889, -0.0871, -0.0863, -0.0984, -0.1352, -0.1199, -0.1238, -0.1238, -0.1435, -0.1243, -0.1206, -0.0841, -0.0697, -0.0708, -0.0113, 0.0171, 0.0192, -0.0025, -0.0025, -0.017, 0.008, 0.0301, 0.0577, 0.1034, 0.1274, 0.1405, 0.1202, 0.0909, 0.1232, 0.1075, 0.0364, 0.0954, 0.0977, 0.0946, 0.1116, 0.1717, 0.1497, 0.1278, 0.1191, 0.0582, 0.0496, 0.0402, 0.0659, 0.0379, 0.0586, 0.0619, 0.0711, 0.0333, 0.0333, 0.0313, 0.0301, 0.028, -0.0397, -0.038, -0.01, 0.0137, -0.0109, 0.0075, 0.024, 0.0195, 0.0112, 0.0034, 0.0, 0.0152, 0.0223, 0.0222, 0.0247, 0.0132, 0.0252, 0.028, 0.0421, 0.0784, 0.0463, 0.2943, 0.3439, 0.4967, 0.4797, 0.3654, 0.3751, 0.3857, 0.516, 0.4903, 0.4809, 0.5284, 0.5124, 0.4628, 0.493, 0.607, 0.6499, 0.6393, 0.6795, 0.6192, 0.6273, 0.6497, 0.5887, 0.6286, 0.5103, 0.4838, 0.5502, 0.5091, 0.6449, 0.5754, 0.5681, 0.5282, 0.4632, 0.4204, 0.309, 0.2947, 0.3664, 0.3097, 0.3612, 0.3612, 0.3821, 0.3963, 0.3676, 0.3826, 0.3723, 0.3849, 0.4049, 0.4081, 0.4068, 0.4069, 0.3392, 0.3189, 0.329, 0.2587, 0.2775, 0.3561, 0.3657, 0.3654, 0.3663, 0.3663, 0.366, 0.3699, 0.3682, 0.3607, 0.3607, 0.4198, 0.4047, 0.3619, 0.3344, 0.3005, 0.2909, 0.3196, 0.404, 0.4207, 0.4481, 0.473, 0.473, 0.4735, 0.5871, 0.6121, 0.6499, 0.5967, 0.6013, 0.6058, 0.6023, 0.5041, 0.5647, 0.5383, 0.2719, 0.2231, 0.3244, 0.3724, 0.3569, 0.357, 0.3085, 0.3172, 0.3172, 0.2903, 0.2715, 0.2921, 0.2717, 0.2491, 0.3587, 0.3397, 0.2941, 0.2721, 0.262, 0.2132, 0.2839, 0.2672, 0.2226, 0.2878, 0.2913, 0.3014, 0.2564, 0.2287, 0.249]}
{"unit_id": "orig:1945588717160452370", "ticker_idx": 0, "ts": "2025-07-16T20:57:52+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Resumption of Shipments to China Drives Model Revisions Upward, H20 Inventory Release to Boost Gross Margins", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: H20 shipment resumption drives NVDA estimates up; SK Hynix and Micron preferred in HBM; Samsung's near-term benefit limited.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Drives Model Revisions Upward, H20 Inventory Release to Boost Gross Margins\n\nUBS (N. Gaudois, T. Arcuri, 25/07/15)\n\nUBS stated that the U.S. government is expected to approve Nvidia's resumption of H20 and RTX PRO GPU shipments to China. This development is anticipated to provide upside for Nvidia's revenue and gross margins, especially given the H20 inventory has already been impaired.\n\nResumption of Shipments to China Leads to Upward Revision of Revenue and Profit Models; H20 Impaired Inventory Becomes Source of Short-Term Gross Margin Flexibility\n\nUBS's model previously assumed that products like H20 would resume shipments in Q3, contributing approximately $2 billion to $3 billion in quarterly revenue, significantly lower than the pre-ban peak of $7 billion to $8 billion, though that peak was influenced by pull-ins. If H20, due to its prior impairment, is resold, it will bring upside to recent gross margins.\n\nH20 Accounts for 13% of GPU Shipments but Has Limited Impact on HBM, Approximately 5% of HBM Bit Demand\n\nPrior to the ban, H20 GPUs accounted for about 13% of Nvidia's GPU shipments. However, because they are configured with only 96GB of HBM, their overall share of HBM bit demand is only about 5%. UBS believes that if shipments resume, the Chinese market might see an initial H20 inventory replenishment rush due to a shortage of alternatives, despite ongoing efforts for domestic substitution.\n\nDifferentiated Impact on HBM Manufacturers; Samsung's Short-Term Benefit Limited, SK Hynix and Micron Remain Preferred\n\nSamsung had previously recorded an impairment of approximately 1.3 trillion Korean Won related to HBM in 2Q25, partly due to H20 and AMD MI308 products. While the resumption of H20 shipments might partially reverse this impairment, Samsung's packaging business among Chinese AI customers and the adoption of HBM3E 8-Hi are still restricted. UBS considers SK Hynix to have a leadership advantage in the HBM sector, while Micron is expected to continue increasing its market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003909684872381747, "ret_signed_m1d": -0.003909684872381747, "alpha_spy_m1d": -0.0005776334386263793, "alpha_c_m1d": -0.00916412460569449, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009511589505708296, "ret_signed_p1d": 0.009511589505708296, "alpha_spy_p1d": 0.003391899528190745, "alpha_c_p1d": 0.0011114490994763582, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0034428594862712103, "ret_signed_p1w": -0.0034428594862712103, "alpha_spy_p1w": -0.019446981286452614, "alpha_c_p1w": 0.005786934868547422, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.0621461963170844, "ret_signed_p1m": 0.0621461963170844, "alpha_spy_p1m": 0.028936714204513825, "alpha_c_p1m": 0.01845147637661637, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.09897081121285733, "ret_signed_p3m": 0.09897081121285733, "alpha_spy_p3m": 0.03383631036483137, "alpha_c_p3m": -0.07549684249508726, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.07988992467564904, "ret_signed_p6m": 0.07988992467564904, "alpha_spy_p6m": -0.031040142771990187, "alpha_c_p6m": -0.23426679716688548, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4078, -0.4345, -0.423, -0.4007, -0.379, -0.3523, -0.3656, -0.3639, -0.3645, -0.3488, -0.3319, -0.3359, -0.3375, -0.317, -0.3152, -0.3194, -0.2823, -0.2419, -0.2103, -0.2133, -0.21, -0.209, -0.2159, -0.231, -0.2249, -0.2339, -0.2339, -0.2094, -0.2134, -0.1878, -0.2115, -0.1984, -0.176, -0.1719, -0.1832, -0.1731, -0.1678, -0.16, -0.1665, -0.1539, -0.1716, -0.1557, -0.159, -0.1511, -0.1511, -0.1606, -0.1587, -0.137, -0.0996, -0.0954, -0.0795, -0.0781, -0.1054, -0.0824, -0.0702, -0.0702, -0.0766, -0.0663, -0.0495, -0.0424, -0.0376, -0.0426, -0.0039, 0.0, 0.0095, 0.0061, 0.0001, -0.0253, -0.0034, 0.0138, 0.0124, 0.0314, 0.0242, 0.0461, 0.0379, 0.0137, 0.0504, 0.0402, 0.047, 0.0549, 0.0661, 0.0624, 0.0688, 0.0596, 0.0621, 0.053, 0.0621, 0.0249, 0.0235, 0.0211, 0.0386, 0.0493, 0.0607, 0.0597, 0.0514, 0.0164, 0.0164, -0.0034, -0.0044, 0.0017, -0.0254, -0.0179, -0.0036, 0.0348, 0.0339, 0.0377, 0.0373, 0.0205, -0.0062, 0.0285, 0.031, 0.0715, 0.0413, 0.0327, 0.0369, 0.0399, 0.0612, 0.0888, 0.0927, 0.1023, 0.0949, 0.0827, 0.0798, 0.1036, 0.1238, 0.0689, 0.099, 0.0506, 0.0494, 0.061, 0.0692, 0.0658, 0.0572, 0.0521, 0.063, 0.0869, 0.1175, 0.1731, 0.2082, 0.184, 0.1817, 0.2073, 0.1595, 0.1392, 0.0976, 0.098, 0.1616, 0.1272, 0.131, 0.0905, 0.1098, 0.0889, 0.0584, 0.0885, 0.0542, 0.0439, 0.0653, 0.0377, 0.0519, 0.0519, 0.0329, 0.05, 0.0589, 0.048, 0.0702, 0.0645, 0.0829, 0.0795, 0.0725, 0.0559, 0.0214, 0.0288, 0.0372, -0.0024, 0.0163, 0.0563, 0.072, 0.1042, 0.1007, 0.1007, 0.1119, 0.0984, 0.0945, 0.0884, 0.0884, 0.1021, 0.0979, 0.0927, 0.1036, 0.0799]}
{"unit_id": "orig:1945588717160452370", "ticker_idx": 1, "ts": "2025-07-16T20:57:52+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS considers SK Hynix to have a leadership advantage in the HBM sector", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: H20 shipment resumption drives NVDA estimates up; SK Hynix and Micron preferred in HBM; Samsung's near-term benefit limited.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Drives Model Revisions Upward, H20 Inventory Release to Boost Gross Margins\n\nUBS (N. Gaudois, T. Arcuri, 25/07/15)\n\nUBS stated that the U.S. government is expected to approve Nvidia's resumption of H20 and RTX PRO GPU shipments to China. This development is anticipated to provide upside for Nvidia's revenue and gross margins, especially given the H20 inventory has already been impaired.\n\nResumption of Shipments to China Leads to Upward Revision of Revenue and Profit Models; H20 Impaired Inventory Becomes Source of Short-Term Gross Margin Flexibility\n\nUBS's model previously assumed that products like H20 would resume shipments in Q3, contributing approximately $2 billion to $3 billion in quarterly revenue, significantly lower than the pre-ban peak of $7 billion to $8 billion, though that peak was influenced by pull-ins. If H20, due to its prior impairment, is resold, it will bring upside to recent gross margins.\n\nH20 Accounts for 13% of GPU Shipments but Has Limited Impact on HBM, Approximately 5% of HBM Bit Demand\n\nPrior to the ban, H20 GPUs accounted for about 13% of Nvidia's GPU shipments. However, because they are configured with only 96GB of HBM, their overall share of HBM bit demand is only about 5%. UBS believes that if shipments resume, the Chinese market might see an initial H20 inventory replenishment rush due to a shortage of alternatives, despite ongoing efforts for domestic substitution.\n\nDifferentiated Impact on HBM Manufacturers; Samsung's Short-Term Benefit Limited, SK Hynix and Micron Remain Preferred\n\nSamsung had previously recorded an impairment of approximately 1.3 trillion Korean Won related to HBM in 2Q25, partly due to H20 and AMD MI308 products. While the resumption of H20 shipments might partially reverse this impairment, Samsung's packaging business among Chinese AI customers and the adoption of HBM3E 8-Hi are still restricted. UBS considers SK Hynix to have a leadership advantage in the HBM sector, while Micron is expected to continue increasing its market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008446041932942183, "ret_signed_m1d": 0.008446041932942183, "alpha_spy_m1d": 0.01177809336669755, "alpha_c_m1d": 0.0031916021996294397, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0895271115528894, "ret_signed_p1d": -0.0895271115528894, "alpha_spy_p1d": -0.09564680153040694, "alpha_c_p1d": -0.09792725195912133, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.09121636234567321, "ret_signed_p1w": -0.09121636234567321, "alpha_spy_p1w": -0.10722048414585461, "alpha_c_p1w": -0.08198656799085458, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": -0.06587855459331182, "ret_signed_p1m": -0.06587855459331182, "alpha_spy_p1m": -0.0990880367058824, "alpha_c_p1m": -0.10957327453377985, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.4040520603898148, "ret_signed_p3m": 0.4040520603898148, "alpha_spy_p3m": 0.3389175595417888, "alpha_c_p3m": 0.2295844066818702, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 1.5595747676352523, "ret_signed_p6m": 1.5595747676352523, "alpha_spy_p6m": 1.448644700187613, "alpha_c_p6m": 1.2454180457927178, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4098, -0.4045, -0.4139, -0.3896, -0.3987, -0.3782, -0.3862, -0.3903, -0.4014, -0.4014, -0.3728, -0.3728, -0.3728, -0.3566, -0.3583, -0.3589, -0.3424, -0.3306, -0.3053, -0.3239, -0.3104, -0.3276, -0.3188, -0.3239, -0.336, -0.3255, -0.3154, -0.3171, -0.2986, -0.2838, -0.3091, -0.299, -0.299, -0.2652, -0.2416, -0.2416, -0.2264, -0.2213, -0.1892, -0.2044, -0.2044, -0.1622, -0.1588, -0.1672, -0.1689, -0.1318, -0.1233, -0.0591, -0.0338, -0.0101, -0.0405, -0.0135, -0.0355, -0.0574, -0.0591, -0.0861, -0.0845, -0.0473, -0.0507, 0.0034, -0.0051, 0.0135, 0.0084, 0.0, -0.0895, -0.0912, -0.0794, -0.0929, -0.0912, -0.0895, -0.1014, -0.1149, -0.1132, -0.1098, -0.076, -0.1284, -0.1284, -0.1098, -0.1267, -0.1149, -0.1334, -0.098, -0.0912, -0.0608, -0.0659, -0.0659, -0.0963, -0.1115, -0.1368, -0.1723, -0.152, -0.1233, -0.1166, -0.1216, -0.0916, -0.0899, -0.1339, -0.1187, -0.1119, -0.1017, -0.0747, -0.0628, -0.0256, 0.0285, 0.0387, 0.1114, 0.1199, 0.1774, 0.1283, 0.2027, 0.2027, 0.1875, 0.2214, 0.2095, 0.2061, 0.1385, 0.1808, 0.1757, 0.218, 0.3381, 0.3381, 0.3381, 0.3381, 0.3381, 0.3381, 0.448, 0.4041, 0.3922, 0.4294, 0.5309, 0.5749, 0.6426, 0.6206, 0.629, 0.6189, 0.7255, 0.81, 0.7627, 0.8879, 0.9217, 0.8912, 1.0976, 0.9826, 0.9589, 1.0063, 0.9623, 1.0503, 1.0942, 1.0875, 1.0706, 0.8946, 1.0503, 0.9285, 0.9014, 0.9318, 0.7627, 0.7593, 0.7559, 0.7728, 0.8418, 0.7944, 0.8215, 0.8892, 0.8689, 0.835, 0.8418, 0.9535, 0.9163, 0.9874, 0.9129, 0.9332, 0.8757, 0.7944, 0.8655, 0.8689, 0.852, 0.9637, 0.9772, 0.9908, 0.9908, 1.028, 1.1668, 1.2041, 1.2041, 1.2041, 1.2921, 1.3564, 1.458, 1.5122, 1.5596]}
{"unit_id": "orig:1945588717160452370", "ticker_idx": 2, "ts": "2025-07-16T20:57:52+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is expected to continue increasing its market share", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: H20 shipment resumption drives NVDA estimates up; SK Hynix and Micron preferred in HBM; Samsung's near-term benefit limited.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Drives Model Revisions Upward, H20 Inventory Release to Boost Gross Margins\n\nUBS (N. Gaudois, T. Arcuri, 25/07/15)\n\nUBS stated that the U.S. government is expected to approve Nvidia's resumption of H20 and RTX PRO GPU shipments to China. This development is anticipated to provide upside for Nvidia's revenue and gross margins, especially given the H20 inventory has already been impaired.\n\nResumption of Shipments to China Leads to Upward Revision of Revenue and Profit Models; H20 Impaired Inventory Becomes Source of Short-Term Gross Margin Flexibility\n\nUBS's model previously assumed that products like H20 would resume shipments in Q3, contributing approximately $2 billion to $3 billion in quarterly revenue, significantly lower than the pre-ban peak of $7 billion to $8 billion, though that peak was influenced by pull-ins. If H20, due to its prior impairment, is resold, it will bring upside to recent gross margins.\n\nH20 Accounts for 13% of GPU Shipments but Has Limited Impact on HBM, Approximately 5% of HBM Bit Demand\n\nPrior to the ban, H20 GPUs accounted for about 13% of Nvidia's GPU shipments. However, because they are configured with only 96GB of HBM, their overall share of HBM bit demand is only about 5%. UBS believes that if shipments resume, the Chinese market might see an initial H20 inventory replenishment rush due to a shortage of alternatives, despite ongoing efforts for domestic substitution.\n\nDifferentiated Impact on HBM Manufacturers; Samsung's Short-Term Benefit Limited, SK Hynix and Micron Remain Preferred\n\nSamsung had previously recorded an impairment of approximately 1.3 trillion Korean Won related to HBM in 2Q25, partly due to H20 and AMD MI308 products. While the resumption of H20 shipments might partially reverse this impairment, Samsung's packaging business among Chinese AI customers and the adoption of HBM3E 8-Hi are still restricted. UBS considers SK Hynix to have a leadership advantage in the HBM sector, while Micron is expected to continue increasing its market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03160703531774334, "ret_signed_m1d": 0.03160703531774334, "alpha_spy_m1d": 0.034939086751498705, "alpha_c_m1d": 0.026352595584430594, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027226580707769732, "ret_signed_p1d": -0.027226580707769732, "alpha_spy_p1d": -0.03334627068528728, "alpha_c_p1d": -0.03562672111400167, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.05668630604607339, "ret_signed_p1w": -0.05668630604607339, "alpha_spy_p1w": -0.07269042784625479, "alpha_c_p1w": -0.047456511691254755, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.07609727428390567, "ret_signed_p1m": 0.07609727428390567, "alpha_spy_p1m": 0.04288779217133509, "alpha_c_p1m": 0.03240255434343764, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.6567098252654557, "ret_signed_p3m": 0.6567098252654557, "alpha_spy_p3m": 0.5915753244174298, "alpha_c_p3m": 0.48224217155751115, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 1.8116205680560835, "ret_signed_p6m": 1.8116205680560835, "alpha_spy_p6m": 1.7006905006084443, "alpha_c_p6m": 1.497463846213549, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4096, -0.4273, -0.3975, -0.3742, -0.3357, -0.3154, -0.3259, -0.3403, -0.3397, -0.3327, -0.3074, -0.3099, -0.3092, -0.2911, -0.2693, -0.2633, -0.2081, -0.1683, -0.1821, -0.181, -0.1591, -0.1535, -0.1582, -0.1776, -0.1863, -0.1988, -0.1988, -0.173, -0.1747, -0.1694, -0.1895, -0.1575, -0.1226, -0.114, -0.0879, -0.0685, -0.048, -0.0206, -0.0044, -0.0031, -0.0081, 0.0283, 0.0326, 0.0453, 0.0453, 0.0606, 0.0475, 0.0976, 0.0919, 0.0812, 0.0705, 0.0576, 0.0373, 0.0446, 0.0493, 0.0493, 0.03, 0.0686, 0.0499, 0.0574, 0.0696, 0.0187, 0.0316, 0.0, -0.0272, -0.0175, -0.0275, -0.0619, -0.0567, -0.0404, -0.0444, -0.0445, -0.0384, -0.0145, -0.0626, -0.0992, -0.0744, -0.0633, -0.0657, -0.0392, 0.0211, 0.0626, 0.0972, 0.0673, 0.0761, 0.0381, 0.0612, 0.0483, 0.0067, -0.0055, 0.0107, -0.0001, 0.0006, 0.0113, 0.0478, 0.0222, 0.0222, 0.0176, 0.0197, 0.0668, 0.1283, 0.1291, 0.1616, 0.2024, 0.2932, 0.3504, 0.3551, 0.3641, 0.3741, 0.4506, 0.3977, 0.4139, 0.4293, 0.3889, 0.347, 0.3508, 0.4077, 0.4371, 0.5645, 0.5782, 0.6143, 0.6412, 0.5959, 0.6891, 0.6529, 0.5607, 0.6567, 0.6076, 0.6496, 0.7406, 0.7393, 0.777, 0.7385, 0.7057, 0.7765, 0.8823, 0.8916, 0.9071, 0.9477, 0.9252, 0.9231, 1.0171, 0.8738, 1.0411, 1.0483, 1.0447, 1.1769, 1.0722, 1.1047, 1.0364, 1.1213, 1.0794, 0.9638, 0.9416, 0.7306, 0.7822, 0.9245, 0.9297, 0.9789, 0.9789, 1.0324, 1.0666, 1.0582, 1.0124, 0.9479, 1.0387, 1.1221, 1.1694, 1.2664, 1.2213, 1.0724, 1.0411, 0.9982, 0.9382, 1.1361, 1.2854, 1.3771, 1.3743, 1.4638, 1.4638, 1.4476, 1.5309, 1.5159, 1.4539, 1.4539, 1.7119, 1.6838, 1.9527, 1.9193, 1.8116]}
{"unit_id": "orig:1945588717160452370", "ticker_idx": 3, "ts": "2025-07-16T20:57:52+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's short-term benefit limited", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: H20 shipment resumption drives NVDA estimates up; SK Hynix and Micron preferred in HBM; Samsung's near-term benefit limited.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Drives Model Revisions Upward, H20 Inventory Release to Boost Gross Margins\n\nUBS (N. Gaudois, T. Arcuri, 25/07/15)\n\nUBS stated that the U.S. government is expected to approve Nvidia's resumption of H20 and RTX PRO GPU shipments to China. This development is anticipated to provide upside for Nvidia's revenue and gross margins, especially given the H20 inventory has already been impaired.\n\nResumption of Shipments to China Leads to Upward Revision of Revenue and Profit Models; H20 Impaired Inventory Becomes Source of Short-Term Gross Margin Flexibility\n\nUBS's model previously assumed that products like H20 would resume shipments in Q3, contributing approximately $2 billion to $3 billion in quarterly revenue, significantly lower than the pre-ban peak of $7 billion to $8 billion, though that peak was influenced by pull-ins. If H20, due to its prior impairment, is resold, it will bring upside to recent gross margins.\n\nH20 Accounts for 13% of GPU Shipments but Has Limited Impact on HBM, Approximately 5% of HBM Bit Demand\n\nPrior to the ban, H20 GPUs accounted for about 13% of Nvidia's GPU shipments. However, because they are configured with only 96GB of HBM, their overall share of HBM bit demand is only about 5%. UBS believes that if shipments resume, the Chinese market might see an initial H20 inventory replenishment rush due to a shortage of alternatives, despite ongoing efforts for domestic substitution.\n\nDifferentiated Impact on HBM Manufacturers; Samsung's Short-Term Benefit Limited, SK Hynix and Micron Remain Preferred\n\nSamsung had previously recorded an impairment of approximately 1.3 trillion Korean Won related to HBM in 2Q25, partly due to H20 and AMD MI308 products. While the resumption of H20 shipments might partially reverse this impairment, Samsung's packaging business among Chinese AI customers and the adoption of HBM3E 8-Hi are still restricted. UBS considers SK Hynix to have a leadership advantage in the HBM sector, while Micron is expected to continue increasing its market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015455945066806476, "ret_signed_m1d": 0.015455945066806476, "alpha_spy_m1d": 0.012123893633051108, "alpha_c_m1d": 0.012363755463044468, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.003092189603762008, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03091201226382756, "ret_signed_p1d": -0.03091201226382756, "alpha_spy_p1d": -0.02479232228631001, "alpha_c_p1d": -0.021828616473980933, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.009083395789846627, "ret_p1w": 0.026275094400549426, "ret_signed_p1w": -0.026275094400549426, "alpha_spy_p1w": -0.010270972600368022, "alpha_c_p1w": -0.020167818587848885, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.006107275812700541, "ret_p1m": 0.10664605760002899, "ret_signed_p1m": -0.10664605760002899, "alpha_spy_p1m": -0.07343657548745841, "alpha_c_p1m": -0.0717806367789382, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.03486542082109079, "ret_p3m": 0.4484739951797647, "ret_signed_p3m": -0.4484739951797647, "alpha_spy_p3m": -0.38333949433173875, "alpha_c_p3m": -0.3447807070362061, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1036932881435586, "ret_p6m": 1.1653323126504187, "ret_signed_p6m": -1.1653323126504187, "alpha_spy_p6m": -1.0544022452027795, "alpha_c_p6m": -1.0471348218388694, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11819749081154929, "price_path": [-0.1505, -0.149, -0.1551, -0.1444, -0.1444, -0.1444, -0.1428, -0.1428, -0.1474, -0.1474, -0.1659, -0.1659, -0.1659, -0.1612, -0.1612, -0.1582, -0.1152, -0.1259, -0.1182, -0.1198, -0.1275, -0.1428, -0.1413, -0.1444, -0.1597, -0.1674, -0.1597, -0.172, -0.1413, -0.1382, -0.1367, -0.1275, -0.1275, -0.1121, -0.0921, -0.0921, -0.0814, -0.0906, -0.0798, -0.086, -0.1044, -0.1213, -0.1075, -0.0814, -0.0906, -0.086, -0.109, -0.0706, -0.0583, -0.0752, -0.0603, -0.0757, -0.0696, -0.0603, -0.0139, -0.0216, -0.0464, -0.051, -0.0665, -0.0572, -0.0325, -0.034, -0.0155, 0.0, 0.0309, 0.0371, 0.0479, 0.0201, 0.0263, 0.0201, 0.0185, 0.0881, 0.0912, 0.1221, 0.1036, 0.0649, 0.0773, 0.0804, 0.0634, 0.0896, 0.1097, 0.0974, 0.0989, 0.1113, 0.1066, 0.1066, 0.0819, 0.0819, 0.0896, 0.0912, 0.1036, 0.1051, 0.0866, 0.0912, 0.0757, 0.0773, 0.0448, 0.068, 0.0788, 0.0835, 0.0742, 0.0835, 0.1051, 0.1221, 0.1345, 0.1654, 0.1824, 0.2272, 0.2087, 0.2411, 0.2411, 0.2906, 0.3091, 0.3199, 0.3308, 0.2875, 0.3072, 0.3025, 0.3351, 0.3934, 0.3934, 0.3934, 0.3934, 0.3934, 0.3934, 0.4656, 0.4485, 0.4221, 0.4749, 0.5168, 0.5199, 0.523, 0.5137, 0.5308, 0.4982, 0.5339, 0.5835, 0.5447, 0.5603, 0.6161, 0.6689, 0.7248, 0.6286, 0.5618, 0.5401, 0.5199, 0.5618, 0.6068, 0.6006, 0.596, 0.509, 0.5618, 0.5183, 0.4982, 0.5618, 0.4718, 0.5013, 0.5416, 0.596, 0.6068, 0.5603, 0.5649, 0.6053, 0.6224, 0.6317, 0.6829, 0.7, 0.6829, 0.6767, 0.6658, 0.6907, 0.627, 0.596, 0.6751, 0.6705, 0.6503, 0.7155, 0.731, 0.7248, 0.7248, 0.8164, 0.8642, 0.8705, 0.8705, 0.8705, 1.0046, 1.1544, 1.1669, 1.1997, 1.1653]}
{"unit_id": "thread:1972604384531128489", "ticker_idx": 0, "ts": "2025-09-29T10:08:29+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics is cited as the biggest beneficiary of this supercycle. Based on its industry-leading memory semiconductor production capacity, it is predicted to benefit the most from rising legacy memory prices.", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supercycle underway with record-low inventories; Samsung named biggest beneficiary, Hynix and Micron also positioned to gain from supply constraints and surging prices.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Supercycle Returns After 7 Years\n\nDRAM semiconductor inventories have fallen to record lows and prices are soaring, signaling the full onset of a “semiconductor supercycle” after seven years.\n\nAccording to market research firm TrendForce on the 29th, as of the end of the third quarter the average inventory of global DRAM manufacturers stood at 3.3 weeks, the lowest level ever recorded. Even compared with the 2018 semiconductor supercycle, when inventories averaged 3 to 4 weeks, this is a lower level. SK Hynix (000660) and Micron reportedly hold only two weeks of inventory each, while Samsung Electronics (005930) has six weeks. Ahn Ki-hyun, executive director of the Korea Semiconductor Industry Association, pointed out, “The average inventory level of DRAM memory is about eight weeks. Two weeks is an emergency situation, and competition among buyers to secure supply will intensify.”\n\nIn reality, the rapid growth of the artificial intelligence (AI) market has driven up demand for high-bandwidth memory (HBM), which is essential for AI accelerators. As semiconductor companies converted existing DRAM production lines to HBM, output of general-purpose DRAM fell sharply. On top of this, the server replacement cycle in data centers coincided, further boosting demand for standard DRAM.\n\nThe shortage of supply has led to surging prices. On this day, TrendForce reported that the price of the standard PC DRAM product “DDR4 8Gb” hit $6.3, setting a new annual high. The latest standard “DDR5 16Gb” was priced at $7.5, which, depending on the institution, represents an increase of nearly twofold. NAND memory prices have also been rising daily, and Samsung Electronics and SK Hynix are reportedly planning additional price hikes for both DRAM and NAND.\n\nThe semiconductor boom is expected to continue for the time being. Since Samsung Electronics, SK Hynix, and Micron are all focusing on HBM4, which will be mounted on NVIDIA’s next-generation AI accelerator Rubin, general-purpose memory production will inevitably remain constrained. An industry insider said, “At present, there are no plans to expand general-purpose memory production lines.”\n\nSamsung Electronics is cited as the biggest beneficiary of this supercycle. Based on its industry-leading memory semiconductor production capacity, it is predicted to benefit the most from rising legacy memory prices. Samsung Electronics’ operating profit forecast for the third quarter is on average in the 9 trillion won range, but some securities firms suggest it could exceed 10 trillion won.\n\n---\n\nhttps://t.co/FZXiMrReNi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01508318736599934, "ret_signed_m1d": -0.01508318736599934, "alpha_spy_m1d": -0.012280696949876546, "alpha_c_m1d": -0.009768114069644174, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.005315073296355166, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00356291718410795, "ret_signed_p1d": -0.00356291718410795, "alpha_spy_p1d": -0.007329783943219925, "alpha_c_p1d": -0.00902088581503413, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.00545796863092618, "ret_p1w": 0.06591448176582415, "ret_signed_p1w": 0.06591448176582415, "alpha_spy_p1w": 0.05396600038341148, "alpha_c_p1w": 0.03998067057913901, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.02593381118668514, "ret_p1m": 0.18171027125445693, "ret_signed_p1m": 0.18171027125445693, "alpha_spy_p1m": 0.14648246451188762, "alpha_c_p1m": 0.10455115713566654, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.0771591141187904, "ret_p3m": 0.3194773989499322, "ret_signed_p3m": 0.3194773989499322, "alpha_spy_p3m": 0.276173414337779, "alpha_c_p3m": 0.2741242318751569, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.045353167074775325, "ret_p6m": 1.263921210152973, "ret_signed_p6m": 1.263921210152973, "alpha_spy_p6m": 1.274138216564915, "alpha_c_p6m": 1.2898830764206246, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": -0.0259618662676514, "price_path": [-0.2811, -0.2456, -0.2516, -0.2705, -0.274, -0.2858, -0.2788, -0.2598, -0.261, -0.2468, -0.235, -0.2114, -0.2066, -0.1984, -0.2196, -0.2149, -0.2196, -0.2208, -0.1676, -0.1652, -0.1416, -0.1558, -0.1853, -0.1759, -0.1735, -0.1865, -0.1664, -0.1511, -0.1605, -0.1593, -0.1499, -0.1534, -0.1534, -0.1723, -0.1723, -0.1664, -0.1652, -0.1558, -0.1546, -0.1688, -0.1652, -0.1771, -0.1759, -0.2007, -0.183, -0.1747, -0.1712, -0.1783, -0.1712, -0.1546, -0.1416, -0.1321, -0.1085, -0.0955, -0.0612, -0.0754, -0.0506, -0.0506, -0.0127, 0.0015, 0.0097, 0.018, -0.0151, 0.0, -0.0036, 0.0214, 0.0659, 0.0659, 0.0659, 0.0659, 0.0659, 0.0659, 0.1211, 0.1081, 0.0879, 0.1283, 0.1603, 0.1627, 0.1651, 0.158, 0.171, 0.1461, 0.1734, 0.2114, 0.1817, 0.1936, 0.2363, 0.2767, 0.3195, 0.2458, 0.1948, 0.1781, 0.1627, 0.1948, 0.2292, 0.2245, 0.2209, 0.1544, 0.1948, 0.1615, 0.1461, 0.1948, 0.1259, 0.1485, 0.1793, 0.2209, 0.2292, 0.1936, 0.1971, 0.228, 0.2411, 0.2482, 0.2874, 0.3005, 0.2874, 0.2827, 0.2743, 0.2933, 0.2447, 0.2209, 0.2815, 0.2779, 0.2625, 0.3124, 0.3242, 0.3195, 0.3195, 0.3895, 0.4261, 0.4309, 0.4309, 0.4309, 0.5335, 0.6481, 0.6577, 0.6827, 0.6565, 0.6589, 0.6565, 0.6421, 0.6744, 0.7173, 0.777, 0.7818, 0.7328, 0.7842, 0.8176, 0.8152, 0.8152, 0.9035, 0.9381, 0.9178, 0.9154, 0.7949, 0.999, 1.0181, 0.9011, 0.8928, 0.9859, 0.9787, 1.0026, 1.1315, 1.1625, 1.1625, 1.1625, 1.1625, 1.2675, 1.2687, 1.3033, 1.3868, 1.4286, 1.6017, 1.5838, 1.5838, 1.3284, 1.0551, 1.2866, 1.246, 1.0706, 1.2424, 1.2675, 1.2424, 1.1899, 1.252, 1.314, 1.4883, 1.3928, 1.3797, 1.2233, 1.2639]}
{"unit_id": "thread:1972604384531128489", "ticker_idx": 1, "ts": "2025-09-29T10:08:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix (000660) and Micron reportedly hold only two weeks of inventory each", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supercycle underway with record-low inventories; Samsung named biggest beneficiary, Hynix and Micron also positioned to gain from supply constraints and surging prices.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Supercycle Returns After 7 Years\n\nDRAM semiconductor inventories have fallen to record lows and prices are soaring, signaling the full onset of a “semiconductor supercycle” after seven years.\n\nAccording to market research firm TrendForce on the 29th, as of the end of the third quarter the average inventory of global DRAM manufacturers stood at 3.3 weeks, the lowest level ever recorded. Even compared with the 2018 semiconductor supercycle, when inventories averaged 3 to 4 weeks, this is a lower level. SK Hynix (000660) and Micron reportedly hold only two weeks of inventory each, while Samsung Electronics (005930) has six weeks. Ahn Ki-hyun, executive director of the Korea Semiconductor Industry Association, pointed out, “The average inventory level of DRAM memory is about eight weeks. Two weeks is an emergency situation, and competition among buyers to secure supply will intensify.”\n\nIn reality, the rapid growth of the artificial intelligence (AI) market has driven up demand for high-bandwidth memory (HBM), which is essential for AI accelerators. As semiconductor companies converted existing DRAM production lines to HBM, output of general-purpose DRAM fell sharply. On top of this, the server replacement cycle in data centers coincided, further boosting demand for standard DRAM.\n\nThe shortage of supply has led to surging prices. On this day, TrendForce reported that the price of the standard PC DRAM product “DDR4 8Gb” hit $6.3, setting a new annual high. The latest standard “DDR5 16Gb” was priced at $7.5, which, depending on the institution, represents an increase of nearly twofold. NAND memory prices have also been rising daily, and Samsung Electronics and SK Hynix are reportedly planning additional price hikes for both DRAM and NAND.\n\nThe semiconductor boom is expected to continue for the time being. Since Samsung Electronics, SK Hynix, and Micron are all focusing on HBM4, which will be mounted on NVIDIA’s next-generation AI accelerator Rubin, general-purpose memory production will inevitably remain constrained. An industry insider said, “At present, there are no plans to expand general-purpose memory production lines.”\n\nSamsung Electronics is cited as the biggest beneficiary of this supercycle. Based on its industry-leading memory semiconductor production capacity, it is predicted to benefit the most from rising legacy memory prices. Samsung Electronics’ operating profit forecast for the third quarter is on average in the 9 trillion won range, but some securities firms suggest it could exceed 10 trillion won.\n\n---\n\nhttps://t.co/FZXiMrReNi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03581654880602969, "ret_signed_m1d": -0.03581654880602969, "alpha_spy_m1d": -0.0330140583899069, "alpha_c_m1d": -0.03321325924383656, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.002603289562193134, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004297878113476816, "ret_signed_p1d": -0.004297878113476816, "alpha_spy_p1d": -0.008064744872588792, "alpha_c_p1d": -0.015765008582315554, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.011467130468838738, "ret_p1w": 0.13323799253141777, "ret_signed_p1w": 0.13323799253141777, "alpha_spy_p1w": 0.1212895111490051, "alpha_c_p1w": 0.069021777279342, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.06421621525207577, "ret_p1m": 0.4928368249178525, "ret_signed_p1m": 0.4928368249178525, "alpha_spy_p1m": 0.4576090181752832, "alpha_c_p1m": 0.3674726689295844, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.1253641559882681, "ret_p3m": 0.6860205561544783, "ret_signed_p3m": 0.6860205561544783, "alpha_spy_p3m": 0.6427165715423251, "alpha_c_p3m": 0.5538592328458196, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.1321613233086587, "ret_p6m": 1.8324553904799323, "ret_signed_p6m": 1.8324553904799323, "alpha_spy_p6m": 1.8426723968918743, "alpha_c_p6m": 1.6057219529479712, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": 0.22673343753196118, "price_path": [-0.2017, -0.2032, -0.226, -0.2246, -0.1931, -0.196, -0.1502, -0.1574, -0.1416, -0.1459, -0.1531, -0.2289, -0.2303, -0.2203, -0.2318, -0.2303, -0.2289, -0.2389, -0.2504, -0.2489, -0.2461, -0.2175, -0.2618, -0.2618, -0.2461, -0.2604, -0.2504, -0.2661, -0.2361, -0.2303, -0.2046, -0.2089, -0.2089, -0.2346, -0.2475, -0.269, -0.299, -0.2818, -0.2575, -0.2518, -0.2561, -0.2307, -0.2292, -0.2665, -0.2536, -0.2479, -0.2393, -0.2163, -0.2063, -0.1748, -0.1289, -0.1203, -0.0587, -0.0516, -0.0029, -0.0444, 0.0186, 0.0186, 0.0057, 0.0344, 0.0244, 0.0215, -0.0358, 0.0, -0.0043, 0.0315, 0.1332, 0.1332, 0.1332, 0.1332, 0.1332, 0.1332, 0.2264, 0.1891, 0.1791, 0.2106, 0.2966, 0.3338, 0.3911, 0.3725, 0.3797, 0.3711, 0.4613, 0.533, 0.4928, 0.5989, 0.6275, 0.6017, 0.7765, 0.6791, 0.659, 0.6991, 0.6619, 0.7364, 0.7736, 0.7679, 0.7536, 0.6046, 0.7364, 0.6332, 0.6103, 0.6361, 0.4928, 0.49, 0.4871, 0.5014, 0.5599, 0.5197, 0.5427, 0.6, 0.5828, 0.5541, 0.5599, 0.6545, 0.6229, 0.6832, 0.6201, 0.6373, 0.5885, 0.5197, 0.5799, 0.5828, 0.5685, 0.6631, 0.6746, 0.686, 0.686, 0.7176, 0.8351, 0.8667, 0.8667, 0.8667, 0.9412, 0.9957, 1.0817, 1.1276, 1.1677, 1.1333, 1.1477, 1.1161, 1.1276, 1.1477, 1.1677, 1.1907, 1.1305, 1.1219, 1.1649, 1.1993, 1.1104, 1.2939, 1.4115, 1.4688, 1.6065, 1.3799, 1.6007, 1.5806, 1.4143, 1.4057, 1.5434, 1.5118, 1.4659, 1.5462, 1.5233, 1.5233, 1.5233, 1.5233, 1.5634, 1.7211, 1.7269, 1.8817, 1.919, 2.1571, 2.0479, 2.0479, 1.6974, 1.4389, 1.7032, 1.6543, 1.4016, 1.6946, 1.7434, 1.6716, 1.6141, 1.798, 1.7865, 2.0335, 1.91, 1.8928, 1.6802, 1.8325]}
{"unit_id": "thread:1972604384531128489", "ticker_idx": 2, "ts": "2025-09-29T10:08:29+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix (000660) and Micron reportedly hold only two weeks of inventory each", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supercycle underway with record-low inventories; Samsung named biggest beneficiary, Hynix and Micron also positioned to gain from supply constraints and surging prices.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Supercycle Returns After 7 Years\n\nDRAM semiconductor inventories have fallen to record lows and prices are soaring, signaling the full onset of a “semiconductor supercycle” after seven years.\n\nAccording to market research firm TrendForce on the 29th, as of the end of the third quarter the average inventory of global DRAM manufacturers stood at 3.3 weeks, the lowest level ever recorded. Even compared with the 2018 semiconductor supercycle, when inventories averaged 3 to 4 weeks, this is a lower level. SK Hynix (000660) and Micron reportedly hold only two weeks of inventory each, while Samsung Electronics (005930) has six weeks. Ahn Ki-hyun, executive director of the Korea Semiconductor Industry Association, pointed out, “The average inventory level of DRAM memory is about eight weeks. Two weeks is an emergency situation, and competition among buyers to secure supply will intensify.”\n\nIn reality, the rapid growth of the artificial intelligence (AI) market has driven up demand for high-bandwidth memory (HBM), which is essential for AI accelerators. As semiconductor companies converted existing DRAM production lines to HBM, output of general-purpose DRAM fell sharply. On top of this, the server replacement cycle in data centers coincided, further boosting demand for standard DRAM.\n\nThe shortage of supply has led to surging prices. On this day, TrendForce reported that the price of the standard PC DRAM product “DDR4 8Gb” hit $6.3, setting a new annual high. The latest standard “DDR5 16Gb” was priced at $7.5, which, depending on the institution, represents an increase of nearly twofold. NAND memory prices have also been rising daily, and Samsung Electronics and SK Hynix are reportedly planning additional price hikes for both DRAM and NAND.\n\nThe semiconductor boom is expected to continue for the time being. Since Samsung Electronics, SK Hynix, and Micron are all focusing on HBM4, which will be mounted on NVIDIA’s next-generation AI accelerator Rubin, general-purpose memory production will inevitably remain constrained. An industry insider said, “At present, there are no plans to expand general-purpose memory production lines.”\n\nSamsung Electronics is cited as the biggest beneficiary of this supercycle. Based on its industry-leading memory semiconductor production capacity, it is predicted to benefit the most from rising legacy memory prices. Samsung Electronics’ operating profit forecast for the third quarter is on average in the 9 trillion won range, but some securities firms suggest it could exceed 10 trillion won.\n\n---\n\nhttps://t.co/FZXiMrReNi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.040451301191353206, "ret_signed_m1d": -0.040451301191353206, "alpha_spy_m1d": -0.03764881077523041, "alpha_c_m1d": -0.03784801162916007, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.002603289562193134, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020866536028222082, "ret_signed_p1d": 0.020866536028222082, "alpha_spy_p1d": 0.017099669269110107, "alpha_c_p1d": 0.009399405559383345, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.011467130468838738, "ret_p1w": 0.16583045755341375, "ret_signed_p1w": 0.16583045755341375, "alpha_spy_p1w": 0.15388197617100108, "alpha_c_p1w": 0.10161424230133798, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.06421621525207577, "ret_p1m": 0.3547834163389052, "ret_signed_p1m": 0.3547834163389052, "alpha_spy_p1m": 0.3195556095963359, "alpha_c_p1m": 0.2294192603506371, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.1253641559882681, "ret_p3m": 0.750210940251838, "ret_signed_p3m": 0.750210940251838, "alpha_spy_p3m": 0.7069069556396848, "alpha_c_p3m": 0.6180496169431793, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.1321613233086587, "ret_p6m": 1.4157269220176496, "ret_signed_p6m": 1.4157269220176496, "alpha_spy_p6m": 1.4259439284295916, "alpha_c_p6m": 1.1889934844856884, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": 0.22673343753196118, "price_path": [-0.2579, -0.2546, -0.2546, -0.2683, -0.2409, -0.2542, -0.2489, -0.2402, -0.2763, -0.2672, -0.2896, -0.309, -0.3021, -0.3092, -0.3336, -0.3299, -0.3183, -0.3212, -0.3212, -0.3169, -0.2999, -0.3341, -0.3601, -0.3425, -0.3346, -0.3363, -0.3174, -0.2746, -0.2451, -0.2206, -0.2418, -0.2356, -0.2625, -0.2462, -0.2553, -0.2849, -0.2935, -0.282, -0.2897, -0.2892, -0.2816, -0.2556, -0.2739, -0.2739, -0.2771, -0.2757, -0.2422, -0.1985, -0.1979, -0.1749, -0.1458, -0.0813, -0.0407, -0.0374, -0.031, -0.0239, 0.0304, -0.0071, 0.0044, 0.0153, -0.0134, -0.0431, -0.0405, 0.0, 0.0209, 0.1113, 0.1211, 0.1467, 0.1658, 0.1337, 0.1999, 0.1742, 0.1087, 0.1769, 0.142, 0.1718, 0.2365, 0.2356, 0.2624, 0.235, 0.2117, 0.262, 0.3371, 0.3437, 0.3548, 0.3836, 0.3676, 0.3661, 0.4329, 0.3311, 0.45, 0.455, 0.4525, 0.5464, 0.472, 0.4951, 0.4466, 0.5069, 0.4771, 0.395, 0.3793, 0.2294, 0.266, 0.3671, 0.3708, 0.4058, 0.4058, 0.4437, 0.468, 0.4621, 0.4296, 0.3837, 0.4483, 0.5075, 0.5411, 0.61, 0.5779, 0.4722, 0.45, 0.4195, 0.3768, 0.5174, 0.6235, 0.6886, 0.6867, 0.7502, 0.7502, 0.7387, 0.7979, 0.7873, 0.7432, 0.7432, 0.9264, 0.9065, 1.0975, 1.0738, 0.9973, 1.1077, 1.1124, 1.0652, 1.036, 1.056, 1.2155, 1.2155, 1.2293, 1.3765, 1.4282, 1.4409, 1.3764, 1.5056, 1.6585, 1.6616, 1.5339, 1.6739, 1.5618, 1.3172, 1.3385, 1.4106, 1.3423, 1.2797, 1.5062, 1.5284, 1.5142, 1.5142, 1.4417, 1.571, 1.549, 1.6151, 1.5711, 1.553, 1.6201, 1.5381, 1.5186, 1.5204, 1.3189, 1.4477, 1.425, 1.2616, 1.3778, 1.462, 1.5572, 1.4757, 1.6026, 1.6983, 1.8198, 1.82, 1.7134, 1.5829, 1.4696, 1.4157]}
{"unit_id": "thread:1947805691034669273", "ticker_idx": 0, "ts": "2025-07-22T23:47:19+00:00", "symbol": "Elite Material Co.", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "capacity constraints for EMC and TUC's high-end production lines, with extended lead times and rising average selling prices (ASP) anticipated", "tweet_type": "analysis", "timeline": "quarters", "summary": "EMC and TUC are beneficiaries of AI server upgrade cycle; M8/M9-grade CCL penetration rising 40%→70% by 2026, driving capacity tightness and ASP gains.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Elite Material Co. 2383.TW Taiwan CCL", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: US copper tariffs send prices soaring, but AI server demand boosts Taiwan's CCL makers\n\nAs the US prepares to impose steep tariffs on imported copper, global copper prices have rebounded, raising concerns about rising costs across copper-intensive industries. Among the most vulnerable is the copper-clad laminate (CCL) industry, a foundational pillar of printed circuit board (PCB) manufacturing, where higher raw material prices could compress margins and trigger price hikes downstream.\n\nDespite the focus on tariff-driven price changes, industry analysts indicate that attention is now gravitating toward the growing demand for advanced materials, fueled by a surge in AI server upgrades. Suppliers based in Taiwan, including Elite Material Co. (EMC) and Taiwan Union Technology Corp. (TUC), are emerging as likely beneficiaries amid this evolving market dynamic.\n\nTrump tariffs fuel copper price volatility\nCopper prices have swung sharply in recent months, partly in response to US President Donald Trump's proposed tariff policies. Trump has announced plans to implement a 50% import duty on copper, categorizing it as a strategic raw material, effective August 1, 2025. Although copper prices have not surpassed the May 2024 peak of US$10,116 per metric ton, futures on the London Metal Exchange (LME) rose to US$9,481 on July 21, 2025, marking a two-week high.\n\nManufacturers of CCL in Taiwan report that the rising copper costs are placing pressure on cost-control efforts. However, much of the increased expense is typically transferred to downstream PCB customers through pricing mechanisms linked to average copper prices. Industry insiders suggest that moderate fluctuations in copper prices could potentially stimulate customer orders and contribute to a more stable demand cycle.\n\nAI demand outshines copper volatility\nAlthough copper accounts for a significant portion of overall costs, some CCL firms believe that carefully managed price increases could create profit opportunities. Industry sources point out that AI remains the primary driver of demand in the global technology market, leading to strong orders for premium CCL materials. Meanwhile, Chinese and mid-tier competitors are still facing challenges related to overcapacity and intense price competition in the low- to mid-range product segments, limiting the possibility of widespread price increases.\n\nFor suppliers based in Taiwan, the emphasis has shifted beyond copper tariffs. Renewed optimism has emerged following a positive outlook from a leading global foundry regarding AI-related demand in the second half of 2025. As a result, industry participants are increasingly focusing on securing high-margin contracts for next-generation AI server components.\n\nRubin GPUs and new ASIC servers to boost CCL penetration\nMarket analysts anticipate that Nvidia's upcoming Rubin-series AI GPUs will upgrade from M8-grade to M9-grade CCLs, featuring higher layer counts and more stringent technical specifications. Concurrently, US-based cloud service providers are expected to roll out a new generation of AI ASIC servers in 2026, a move projected to significantly boost demand for both upstream CCL materials and downstream PCB manufacturing.\n\nForecasts suggest that as demand for AI servers and ASIC chips grows from late 2024 onward, penetration of M8/M9-grade CCLs will jump from 40% in 2025 to 70% by 2026. This will likely lead to capacity constraints for EMC and TUC's high-end production lines, with extended lead times and rising average selling prices (ASP) anticipated.\n\n---\n\nhttps://t.co/hZjFuFwHMT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025252471624815254, "ret_signed_m1d": 0.025252471624815254, "alpha_spy_m1d": 0.025395572307303715, "alpha_c_m1d": 0.015902058813407116, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.009350412811408138, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0020202227141890017, "ret_signed_p1d": -0.0020202227141890017, "alpha_spy_p1d": -0.010527763836303938, "alpha_c_p1d": -0.007738704939667329, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.005718482225478327, "ret_p1w": 0.025252471624815254, "ret_signed_p1w": 0.025252471624815254, "alpha_spy_p1w": 0.015075389913640436, "alpha_c_p1w": 0.004851527893515284, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.02040094373129997, "ret_p1m": 0.16161619316187337, "ret_signed_p1m": 0.16161619316187337, "alpha_spy_p1m": 0.14690706781145035, "alpha_c_p1m": 0.15759794743750466, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.0040182457243687075, "ret_p3m": 0.13080504850808072, "ret_signed_p3m": 0.13080504850808072, "alpha_spy_p3m": 0.07137686199993021, "alpha_c_p3m": 0.028157221107480757, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.10264782740059997, "ret_p6m": 0.708999485075561, "ret_signed_p6m": 0.708999485075561, "alpha_spy_p6m": 0.6049070502566005, "alpha_c_p6m": 0.5876694196944172, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.12133006538114377, "price_path": [-0.4798, -0.4283, -0.4374, -0.4293, -0.4384, -0.4384, -0.3828, -0.3707, -0.3697, -0.3818, -0.3313, -0.3141, -0.3182, -0.3192, -0.3061, -0.3081, -0.3091, -0.3263, -0.3172, -0.2727, -0.2788, -0.2828, -0.2859, -0.2869, -0.2717, -0.2434, -0.2434, -0.2343, -0.2081, -0.2121, -0.203, -0.204, -0.1717, -0.101, -0.1182, -0.1333, -0.1313, -0.1121, -0.1242, -0.1465, -0.1465, -0.1535, -0.1576, -0.1313, -0.1333, -0.1364, -0.1253, -0.1091, -0.0909, -0.0879, -0.0626, -0.0677, -0.0899, -0.0374, -0.0333, -0.0333, -0.0495, -0.0626, -0.0202, -0.0071, -0.0172, 0.0354, 0.0253, 0.0, -0.002, -0.001, 0.004, 0.0354, 0.0253, 0.0152, 0.1162, 0.1162, 0.1212, 0.1364, 0.1212, 0.1566, 0.1616, 0.1667, 0.1869, 0.2121, 0.2222, 0.2929, 0.3182, 0.2424, 0.1616, 0.2475, 0.2172, 0.2626, 0.2626, 0.2929, 0.2792, 0.2485, 0.182, 0.1769, 0.1922, 0.141, 0.1922, 0.1717, 0.2536, 0.2843, 0.3559, 0.3764, 0.3457, 0.315, 0.3457, 0.3662, 0.3201, 0.3304, 0.3764, 0.3252, 0.2843, 0.2536, 0.2536, 0.2536, 0.2587, 0.3252, 0.3355, 0.3355, 0.2638, 0.2229, 0.2638, 0.2638, 0.228, 0.1973, 0.1666, 0.1666, 0.1308, 0.1513, 0.1564, 0.1462, 0.141, 0.141, 0.1922, 0.2843, 0.2792, 0.3048, 0.3918, 0.4429, 0.5043, 0.4736, 0.4839, 0.402, 0.3918, 0.3508, 0.3764, 0.4071, 0.4122, 0.4839, 0.3815, 0.3611, 0.4634, 0.3406, 0.3508, 0.4839, 0.5043, 0.5299, 0.5606, 0.5094, 0.4736, 0.4634, 0.4583, 0.5299, 0.5197, 0.535, 0.6374, 0.6322, 0.6425, 0.6425, 0.5299, 0.5299, 0.535, 0.5862, 0.6374, 0.6527, 0.7295, 0.7295, 0.6783, 0.7141, 0.6834, 0.6834, 0.6834, 0.6271, 0.6578, 0.7346, 0.6169, 0.5811, 0.6015, 0.6476, 0.6732, 0.709]}
{"unit_id": "orig:1946366333672985081", "ticker_idx": 0, "ts": "2025-07-19T00:27:50+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintained its \"Buy\" rating on Dell and raised its 12-month price target from $130 to $140", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman reiterates Buy on DELL, raises 12-month PT from $130 to $140 after HQ visit; confident in AI server tech advantages and $14.4B backlog.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: After Visiting Dell Headquarters, More Confident in its AI Server Technology Advantages\n\nEarlier this year, Dell reached a $5 billion agreement with Elon Musk's xAI to provide advanced AI servers to support the startup's next-generation AI computing workloads. Dell also recently signed an agreement with AI hyperscaler CoreWeave to deliver high-performance server racks, supporting its rapidly expanding cloud services platform. \n\nOn Thursday, a team of Goldman Sachs analysts led by Michael Ng visited Dell's headquarters in Round Rock, Texas. They met with several Dell executives, including Arthur Lewis, President of Infrastructure Solutions Group, and Paul Frantz, Head of Investor Relations, to gain a deeper understanding of Dell's rapid innovation and execution capabilities in the AI infrastructure space. \n\nNg told clients: \"This site visit gave us a deeper appreciation for Dell's engineering and design capabilities for AI servers, which we believe is its core competitive advantage in the market.\" The analysts' main conclusion is that Dell has successfully established its dominant position in the AI infrastructure domain, thanks to its world-class engineering strength, deep relationships with customers and suppliers, and the growing demand for AI servers—with a current backlog of advanced server orders reaching $14.4 billion. \n\nAnalysts stated that Dell's engineering and design capabilities are outstanding. Following the first shipments of NVIDIA GB300 NVL72 servers on July 3, just seven months after the earlier first shipments of GB200 NVL72, the analysts' on-site visit to the AI lab further deepened their understanding of Dell's leading capabilities. Dell assigns dedicated teams (\"pods\") comprising thermal, networking, storage, and system architects to AI clients, working closely with them to meet evolving AI demands. Dell can also design and deliver multiple custom components to meet specific scenario requirements, strongly supported by its top-tier supply chain management. Furthermore, Dell collaborates closely with key suppliers to pre-position new product deliveries, which will lead to operational leverage as revenue expands. \n\nMeanwhile, Dell's AI labs and first fiscal quarter (F1Q26) results show strong AI server demand: last quarter, Dell recorded $12.1 billion in AI server orders, and the backlog increased to $14.4 billion. Despite converting $12.1 billion in orders into sales, Dell's orders continue to expand, reflecting sustained strong demand for AI systems. While sovereign government customers contributed to the growth, the primary demand still comes from Tier 2 Cloud Service Providers (CSPs). Analysts interacted with several AI team members in the lab, gaining a deeper understanding of the complex elements involved in server execution, including cooling, rack design, power, cabling, networking, and software. \n\nAdditionally, Dell's traditional server refreshes continue, and the storage business is also progressing: 70% of Dell's traditional installed base is still 14th generation or older products, presenting a clear refresh opportunity. Enterprises are also attempting to consolidate data center equipment to optimize power and cooling efficiency. Although traditional servers have achieved year-over-year growth for six consecutive quarters, the growth primarily comes from increased average selling prices and configurations rather than unit shipments, indicating that most equipment is still relatively old. While the storage business was impacted by the gradual phasing out of VXRail products, Dell's self-developed technology has driven three consecutive quarters of storage revenue growth, with positive progress particularly in private cloud, AI, and security domains. \n\nGoldman Sachs maintained its \"Buy\" rating on Dell and raised its 12-month price target from $130 to $140.\n\n$DELL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017679946180996664, "ret_signed_m1d": 0.017679946180996664, "alpha_spy_m1d": 0.019572532671161547, "alpha_c_m1d": 0.0189813437611831, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0013013975801864364, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.031961669876199084, "ret_signed_p1d": -0.031961669876199084, "alpha_spy_p1d": -0.03210479103942365, "alpha_c_p1d": -0.022697877349118434, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.00926379252708065, "ret_p1w": 0.042239280869334817, "ret_signed_p1w": 0.042239280869334817, "alpha_spy_p1w": 0.029245602184569286, "alpha_c_p1w": 0.031329358225955284, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.010909922643379533, "ret_p1m": 0.0526725085903017, "ret_signed_p1m": 0.0526725085903017, "alpha_spy_p1m": 0.03511441994896303, "alpha_c_p1m": 0.05117956686875025, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.001492941721551455, "ret_p3m": 0.17810559727123576, "ret_signed_p3m": 0.17810559727123576, "alpha_spy_p3m": 0.12450633006725442, "alpha_c_p3m": 0.08758896765051194, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.09051662962072382, "ret_p6m": -0.06500286698745772, "ret_signed_p6m": -0.06500286698745772, "alpha_spy_p6m": -0.17470778384551544, "alpha_c_p6m": -0.18961118678289368, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.12460831979543596, "price_path": [-0.3086, -0.268, -0.2642, -0.2674, -0.2728, -0.2885, -0.287, -0.2665, -0.2683, -0.2715, -0.2572, -0.2539, -0.2563, -0.198, -0.1637, -0.1391, -0.1403, -0.1145, -0.1138, -0.1142, -0.1321, -0.1319, -0.1307, -0.1307, -0.1161, -0.1178, -0.1189, -0.1372, -0.1619, -0.1323, -0.1261, -0.1329, -0.1179, -0.1143, -0.1218, -0.1374, -0.1207, -0.1504, -0.118, -0.101, -0.0962, -0.0962, -0.0744, -0.0848, -0.0649, -0.0659, -0.0231, -0.0385, -0.0493, -0.0561, -0.0425, -0.029, -0.029, -0.0325, -0.0354, -0.0181, -0.0081, -0.0165, -0.0255, -0.0254, -0.0418, -0.0394, 0.0177, 0.0, -0.032, -0.0095, -0.0007, 0.0217, 0.0422, 0.0395, 0.0397, 0.0331, -0.0087, 0.014, 0.0159, -0.0023, 0.0428, 0.0714, 0.077, 0.1028, 0.0833, 0.0812, 0.0767, 0.0755, 0.0527, 0.0004, -0.0047, 0.0187, 0.02, 0.0199, 0.0317, 0.0437, -0.0489, -0.0489, -0.0582, -0.0344, -0.0137, -0.0281, -0.0423, -0.0556, -0.031, -0.0239, -0.0264, -0.0127, -0.0059, 0.0148, 0.0286, 0.0273, 0.0565, 0.046, 0.0285, 0.0197, 0.0181, 0.0426, 0.1038, 0.1654, 0.1474, 0.0958, 0.1349, 0.1747, 0.281, 0.2142, 0.1723, 0.1944, 0.1583, 0.1967, 0.1781, 0.1647, 0.1513, 0.1676, 0.1731, 0.2051, 0.2396, 0.2673, 0.2883, 0.2783, 0.2581, 0.2659, 0.2511, 0.2083, 0.1909, 0.1657, 0.1463, 0.1149, 0.0842, 0.0995, 0.0466, 0.0452, -0.043, -0.0413, -0.0672, -0.0827, -0.0427, -0.0059, -0.0161, 0.0413, 0.0413, 0.042, 0.0321, 0.0623, 0.0442, 0.086, 0.0854, 0.0971, 0.08, 0.0989, 0.083, 0.0156, 0.0198, 0.0451, -0.0007, -0.0394, -0.0122, -0.0107, -0.0028, 0.0031, 0.0031, 0.0099, -0.0041, -0.0005, -0.0164, -0.0164, -0.0014, -0.031, -0.0316, -0.0618, -0.0741, -0.0575, -0.0587, -0.065]}
{"unit_id": "orig:1948900335541780852", "ticker_idx": 1, "ts": "2025-07-26T00:17:03+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain an Overweight rating on Hanmi Semiconductor and BESI", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM maintains Overweight on Hanmi Semi and BESI, Neutral on ASMPT; selectively bullish on ASE and Unimicron as advanced packaging beneficiaries, cautious on mainstream semi recovery and hybrid bonding adoption pace.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Asian Tech- JPM\n\nKey highlights from BESI 2Q25 results\n\nSemiconductor Equipment Industry Outlook: Hopes and Concerns for H2 2025 📊\n\n🌅 Mainstream Recovery, but Double-Dip Concerns...\n\nBESI's Q2 results were still impacted by weak demand for mobile and automotive semiconductors, but management emphasized early signs of recovery in mainstream semiconductors. They stated that demand for HPC in high-end mobile and consumer markets is driving growth, and they anticipate a semiconductor upturn in H2 2025. This aligns with ASMPT's recent outlook.\n\nHowever, we are more cautious about the mainstream semiconductor recovery. Recent guidance from fabless companies and IDMs suggests a potential double-dip in the second half of this year. This is due to the fading impact of Chinese subsidies and the strong tariff avoidance demand seen in the first half of the year also diminishing.\n\n🔗 Hybrid Bonding, Driving Force in HBM4 and Logic... But Cautiously\n\nBESI reported that hybrid bonding revenue more than doubled year-over-year in H1 2025 and expects significant order increases in H2. Management expressed confidence that hybrid bonding is progressing smoothly in HBM4 (estimated to be Samsung's 16-Hi HBM4) and logic (primarily TSMC's SoIC-X), preparing for customer qualification in 2026 and mass production in 2027.\n\nHowever, we remain cautious about the rapid adoption of hybrid bonding, especially in DRAM applications. We believe that TCB will remain the preferred option for the next two years due to better TCO and improved throughput. The transition to hybrid bonding is likely to start with 20Hi/24Hi HBM4e or HBM5 stacks.\n\nIn logic, SoIC adoption also remains slow (only AMD among AI customers by 2027, and for high-end Mac processors like Apple's M5 Pro/Max. Nvidia seems more cautious, and CPO appears delayed). NAND's 3D stacking could emerge as a new application for hybrid bonding in the mid-term, but it will require much lower specifications and precision.\n\n🎯 Focus on BESI's New TCB Platform, Memory TCB Market Intensifies\n\nBESI's new TCB platform is attracting decent interest (estimated to be from Micron). The improved cycle time and accuracy of the next-generation TCB platform look highly competitive. BESI announced that it plans to ship 5 system orders (revenue scale over $20 million) in late Q3 to Q4 this year, which appears to be for pilot testing given the low volume.\n\nWe believe the memory TCB market is becoming increasingly competitive. Multiple vendors (ASMPT, Hanmi Semiconductor, Hanwha Systems, Shinkawa, SEMES, BESI, etc.) are competing for market share. This competition will put pressure on margins and necessitate further innovation to maintain a competitive edge.\n\n🚀 Advanced Packaging Market Growing Rapidly - Selective Beneficiaries Expected\n\nThe advanced packaging market is expected to show rapid compound annual growth (Advanced Packaging ~12% CAGR to '29, High-Performance Packaging ~37% CAGR to '29). This will benefit major OSATs like ASE and substrate manufacturers like Unimicron. Meanwhile, increasing complexity and precision requirements are likely to positively impact test vendors (BESI noted the assembly equipment market needs to grow 63% from 2024-29).\n\nNevertheless, we maintain a more selective view on equipment vendors due to the challenge of fierce competition among multiple vendors in some areas. We maintain an Overweight rating on Hanmi Semiconductor and BESI, and a Neutral rating on ASMPT.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04641694600215507, "ret_signed_m1d": -0.04641694600215507, "alpha_spy_m1d": -0.04666810921446585, "alpha_c_m1d": -0.032965015750539006, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02280134081034435, "ret_signed_p1d": -0.02280134081034435, "alpha_spy_p1d": -0.02016359556985825, "alpha_c_p1d": -0.02791432019549467, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.03583063375889861, "ret_signed_p1w": -0.03583063375889861, "alpha_spy_p1w": -0.026771556479776026, "alpha_c_p1w": -0.03164408774812455, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": -0.017101052497982994, "ret_signed_p1m": -0.017101052497982994, "alpha_spy_p1m": -0.030006434899530476, "alpha_c_p1m": -0.035425713452676866, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.17467427705433103, "ret_signed_p3m": 0.17467427705433103, "alpha_spy_p3m": 0.11708271675675053, "alpha_c_p3m": -0.008538296387232913, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 0.40798040363730603, "ret_signed_p6m": 0.40798040363730603, "alpha_spy_p6m": 0.33807392158735805, "alpha_c_p6m": 0.06415302467899364, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.2326, -0.2326, -0.1881, -0.1849, -0.1951, -0.1531, -0.136, -0.1213, -0.0635, -0.0334, -0.0216, -0.0594, -0.0774, -0.0973, -0.1164, -0.1136, -0.1197, -0.1425, -0.1246, -0.1087, -0.1132, -0.1002, -0.1323, -0.1535, -0.1307, -0.1209, -0.0993, -0.0749, -0.0647, -0.0204, -0.0354, -0.0004, 0.0252, 0.0342, 0.033, 0.024, 0.0155, 0.0029, 0.0236, 0.0497, 0.0521, 0.057, 0.0623, 0.0346, -0.0053, 0.0037, -0.0069, -0.0143, -0.0114, -0.0049, -0.0069, 0.0375, 0.0354, 0.0151, 0.0285, -0.0232, 0.0383, 0.0419, 0.0599, 0.0244, 0.0069, 0.0342, -0.0464, 0.0, -0.0228, -0.0033, -0.0301, -0.046, -0.0358, -0.0611, -0.0651, -0.0232, -0.0167, -0.0102, 0.0077, 0.0057, -0.0012, -0.0346, -0.0407, -0.0277, -0.0464, -0.0599, -0.0309, -0.0297, -0.0171, -0.0244, -0.0248, -0.0627, -0.0594, -0.1295, -0.1384, -0.1274, -0.1148, -0.0973, -0.0924, -0.0851, -0.0745, -0.0892, -0.0379, -0.0505, -0.0582, 0.0163, -0.0106, 0.0, 0.0334, 0.0468, 0.0318, -0.0033, 0.0293, 0.033, 0.0407, 0.0875, 0.0627, 0.195, 0.1954, 0.1918, 0.184, 0.1425, 0.1625, 0.1441, 0.1849, 0.1816, 0.1539, 0.1824, 0.1751, 0.1279, 0.1747, 0.1975, 0.2052, 0.1938, 0.2024, 0.1975, 0.202, 0.1873, 0.1608, 0.1523, 0.1217, 0.0924, 0.1112, 0.1221, 0.1022, 0.0896, 0.07, 0.0603, 0.0354, 0.0513, 0.059, 0.0, 0.013, 0.0147, 0.0554, 0.0578, 0.0578, 0.0566, 0.0696, 0.1071, 0.1197, 0.1364, 0.1958, 0.1669, 0.1437, 0.1148, 0.0757, 0.0786, 0.0879, 0.0529, 0.0704, 0.0647, 0.0684, 0.0741, 0.0713, 0.0713, 0.0713, 0.0757, 0.0839, 0.0892, 0.0892, 0.2142, 0.252, 0.3103, 0.2932, 0.2305, 0.2305, 0.3204, 0.3616, 0.3151, 0.4141, 0.408, 0.3746, 0.408]}
{"unit_id": "orig:1948900335541780852", "ticker_idx": 2, "ts": "2025-07-26T00:17:03+00:00", "symbol": "ASMPT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a Neutral rating on ASMPT", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM maintains Overweight on Hanmi Semi and BESI, Neutral on ASMPT; selectively bullish on ASE and Unimicron as advanced packaging beneficiaries, cautious on mainstream semi recovery and hybrid bonding adoption pace.", "resolved_tickers": ["0522.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "ASMPT listed on Hong Kong Exchange 0522.HK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Asian Tech- JPM\n\nKey highlights from BESI 2Q25 results\n\nSemiconductor Equipment Industry Outlook: Hopes and Concerns for H2 2025 📊\n\n🌅 Mainstream Recovery, but Double-Dip Concerns...\n\nBESI's Q2 results were still impacted by weak demand for mobile and automotive semiconductors, but management emphasized early signs of recovery in mainstream semiconductors. They stated that demand for HPC in high-end mobile and consumer markets is driving growth, and they anticipate a semiconductor upturn in H2 2025. This aligns with ASMPT's recent outlook.\n\nHowever, we are more cautious about the mainstream semiconductor recovery. Recent guidance from fabless companies and IDMs suggests a potential double-dip in the second half of this year. This is due to the fading impact of Chinese subsidies and the strong tariff avoidance demand seen in the first half of the year also diminishing.\n\n🔗 Hybrid Bonding, Driving Force in HBM4 and Logic... But Cautiously\n\nBESI reported that hybrid bonding revenue more than doubled year-over-year in H1 2025 and expects significant order increases in H2. Management expressed confidence that hybrid bonding is progressing smoothly in HBM4 (estimated to be Samsung's 16-Hi HBM4) and logic (primarily TSMC's SoIC-X), preparing for customer qualification in 2026 and mass production in 2027.\n\nHowever, we remain cautious about the rapid adoption of hybrid bonding, especially in DRAM applications. We believe that TCB will remain the preferred option for the next two years due to better TCO and improved throughput. The transition to hybrid bonding is likely to start with 20Hi/24Hi HBM4e or HBM5 stacks.\n\nIn logic, SoIC adoption also remains slow (only AMD among AI customers by 2027, and for high-end Mac processors like Apple's M5 Pro/Max. Nvidia seems more cautious, and CPO appears delayed). NAND's 3D stacking could emerge as a new application for hybrid bonding in the mid-term, but it will require much lower specifications and precision.\n\n🎯 Focus on BESI's New TCB Platform, Memory TCB Market Intensifies\n\nBESI's new TCB platform is attracting decent interest (estimated to be from Micron). The improved cycle time and accuracy of the next-generation TCB platform look highly competitive. BESI announced that it plans to ship 5 system orders (revenue scale over $20 million) in late Q3 to Q4 this year, which appears to be for pilot testing given the low volume.\n\nWe believe the memory TCB market is becoming increasingly competitive. Multiple vendors (ASMPT, Hanmi Semiconductor, Hanwha Systems, Shinkawa, SEMES, BESI, etc.) are competing for market share. This competition will put pressure on margins and necessitate further innovation to maintain a competitive edge.\n\n🚀 Advanced Packaging Market Growing Rapidly - Selective Beneficiaries Expected\n\nThe advanced packaging market is expected to show rapid compound annual growth (Advanced Packaging ~12% CAGR to '29, High-Performance Packaging ~37% CAGR to '29). This will benefit major OSATs like ASE and substrate manufacturers like Unimicron. Meanwhile, increasing complexity and precision requirements are likely to positively impact test vendors (BESI noted the assembly equipment market needs to grow 63% from 2024-29).\n\nNevertheless, we maintain a more selective view on equipment vendors due to the challenge of fierce competition among multiple vendors in some areas. We maintain an Overweight rating on Hanmi Semiconductor and BESI, and a Neutral rating on ASMPT.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025417738922480115, "ret_signed_m1d": -0.025417738922480115, "alpha_spy_m1d": -0.025166575710169337, "alpha_c_m1d": -0.03886966917409618, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007262211120708573, "ret_signed_p1d": -0.007262211120708573, "alpha_spy_p1d": -0.009899956361194673, "alpha_c_p1d": -0.0021492317355582546, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.018881480208448265, "ret_signed_p1w": 0.018881480208448265, "alpha_spy_p1w": 0.00982240292932568, "alpha_c_p1w": 0.014694934197674203, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": 0.05991256998236061, "ret_signed_p1m": -0.05991256998236061, "alpha_spy_p1m": -0.047007187580813126, "alpha_c_p1m": -0.041587909027666736, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.2137237921272679, "ret_signed_p3m": -0.2137237921272679, "alpha_spy_p3m": -0.15613223182968738, "alpha_c_p3m": -0.030511218685703945, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 0.42293624460699886, "ret_signed_p6m": -0.42293624460699886, "alpha_spy_p6m": -0.3530297625570509, "alpha_c_p6m": -0.07910886564868647, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.2462, -0.2462, -0.2289, -0.2289, -0.231, -0.239, -0.2303, -0.226, -0.1956, -0.191, -0.1786, -0.1786, -0.1888, -0.2244, -0.2186, -0.2157, -0.2259, -0.2389, -0.2295, -0.2397, -0.2309, -0.2179, -0.2317, -0.2222, -0.2092, -0.2106, -0.1924, -0.1975, -0.1721, -0.1663, -0.191, -0.2106, -0.2106, -0.2077, -0.2128, -0.2295, -0.2397, -0.2295, -0.2033, -0.1939, -0.1816, -0.1859, -0.1598, -0.1641, -0.1641, -0.1685, -0.1649, -0.146, -0.1576, -0.1416, -0.1474, -0.1467, -0.1235, -0.1198, -0.1009, -0.1097, -0.0908, -0.0959, -0.0821, -0.1111, -0.0821, 0.0196, 0.0254, 0.0, 0.0073, -0.029, -0.0283, -0.053, -0.0189, -0.0182, -0.0174, 0.0087, 0.0022, 0.0235, 0.0257, 0.0468, 0.0191, 0.0461, 0.022, -0.0276, -0.0297, -0.037, 0.0162, 0.0934, 0.0599, 0.0205, 0.0599, 0.0278, 0.0351, -0.0079, -0.0159, -0.0225, -0.0093, -0.0042, -0.0013, 0.0038, 0.0351, 0.0344, 0.0242, 0.0453, 0.0388, 0.1146, 0.1219, 0.1554, 0.1518, 0.2232, 0.1875, 0.1496, 0.1598, 0.1962, 0.1962, 0.2392, 0.2232, 0.3019, 0.3019, 0.3136, 0.3901, 0.353, 0.3099, 0.2239, 0.2801, 0.2873, 0.2276, 0.2691, 0.2516, 0.2181, 0.2137, 0.3121, 0.3668, 0.272, 0.272, 0.2225, 0.1955, 0.2042, 0.1904, 0.1889, 0.1999, 0.1831, 0.2006, 0.1795, 0.1561, 0.1678, 0.1357, 0.1365, 0.108, 0.0942, 0.1058, 0.038, 0.0329, 0.0687, 0.057, 0.0796, 0.0978, 0.1153, 0.1146, 0.1277, 0.1379, 0.1459, 0.135, 0.1532, 0.1241, 0.1496, 0.1481, 0.0964, 0.0898, 0.1073, 0.1, 0.084, 0.1277, 0.1292, 0.1284, 0.1284, 0.1284, 0.1102, 0.1262, 0.1328, 0.1328, 0.1838, 0.1984, 0.2429, 0.3194, 0.2976, 0.3114, 0.3129, 0.3602, 0.3756, 0.3894, 0.4266, 0.42, 0.4229]}
{"unit_id": "orig:1948900335541780852", "ticker_idx": 3, "ts": "2025-07-26T00:17:03+00:00", "symbol": "ASE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This will benefit major OSATs like ASE", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM maintains Overweight on Hanmi Semi and BESI, Neutral on ASMPT; selectively bullish on ASE and Unimicron as advanced packaging beneficiaries, cautious on mainstream semi recovery and hybrid bonding adoption pace.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asian Tech- JPM\n\nKey highlights from BESI 2Q25 results\n\nSemiconductor Equipment Industry Outlook: Hopes and Concerns for H2 2025 📊\n\n🌅 Mainstream Recovery, but Double-Dip Concerns...\n\nBESI's Q2 results were still impacted by weak demand for mobile and automotive semiconductors, but management emphasized early signs of recovery in mainstream semiconductors. They stated that demand for HPC in high-end mobile and consumer markets is driving growth, and they anticipate a semiconductor upturn in H2 2025. This aligns with ASMPT's recent outlook.\n\nHowever, we are more cautious about the mainstream semiconductor recovery. Recent guidance from fabless companies and IDMs suggests a potential double-dip in the second half of this year. This is due to the fading impact of Chinese subsidies and the strong tariff avoidance demand seen in the first half of the year also diminishing.\n\n🔗 Hybrid Bonding, Driving Force in HBM4 and Logic... But Cautiously\n\nBESI reported that hybrid bonding revenue more than doubled year-over-year in H1 2025 and expects significant order increases in H2. Management expressed confidence that hybrid bonding is progressing smoothly in HBM4 (estimated to be Samsung's 16-Hi HBM4) and logic (primarily TSMC's SoIC-X), preparing for customer qualification in 2026 and mass production in 2027.\n\nHowever, we remain cautious about the rapid adoption of hybrid bonding, especially in DRAM applications. We believe that TCB will remain the preferred option for the next two years due to better TCO and improved throughput. The transition to hybrid bonding is likely to start with 20Hi/24Hi HBM4e or HBM5 stacks.\n\nIn logic, SoIC adoption also remains slow (only AMD among AI customers by 2027, and for high-end Mac processors like Apple's M5 Pro/Max. Nvidia seems more cautious, and CPO appears delayed). NAND's 3D stacking could emerge as a new application for hybrid bonding in the mid-term, but it will require much lower specifications and precision.\n\n🎯 Focus on BESI's New TCB Platform, Memory TCB Market Intensifies\n\nBESI's new TCB platform is attracting decent interest (estimated to be from Micron). The improved cycle time and accuracy of the next-generation TCB platform look highly competitive. BESI announced that it plans to ship 5 system orders (revenue scale over $20 million) in late Q3 to Q4 this year, which appears to be for pilot testing given the low volume.\n\nWe believe the memory TCB market is becoming increasingly competitive. Multiple vendors (ASMPT, Hanmi Semiconductor, Hanwha Systems, Shinkawa, SEMES, BESI, etc.) are competing for market share. This competition will put pressure on margins and necessitate further innovation to maintain a competitive edge.\n\n🚀 Advanced Packaging Market Growing Rapidly - Selective Beneficiaries Expected\n\nThe advanced packaging market is expected to show rapid compound annual growth (Advanced Packaging ~12% CAGR to '29, High-Performance Packaging ~37% CAGR to '29). This will benefit major OSATs like ASE and substrate manufacturers like Unimicron. Meanwhile, increasing complexity and precision requirements are likely to positively impact test vendors (BESI noted the assembly equipment market needs to grow 63% from 2024-29).\n\nNevertheless, we maintain a more selective view on equipment vendors due to the challenge of fierce competition among multiple vendors in some areas. We maintain an Overweight rating on Hanmi Semiconductor and BESI, and a Neutral rating on ASMPT.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006622487875614147, "ret_signed_m1d": -0.006622487875614147, "alpha_spy_m1d": -0.006873651087924926, "alpha_c_m1d": 0.006829442376001915, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008514678825578437, "ret_signed_p1d": -0.008514678825578437, "alpha_spy_p1d": -0.0058769335850923365, "alpha_c_p1d": -0.013627658210728755, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.0558183698791207, "ret_signed_p1w": -0.0558183698791207, "alpha_spy_p1w": -0.046759292599998115, "alpha_c_p1w": -0.05163182386834664, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": -0.05014188725385815, "ret_signed_p1m": -0.05014188725385815, "alpha_spy_p1m": -0.06304726965540564, "alpha_c_p1m": -0.06846654820855202, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.21286660974093974, "ret_signed_p3m": 0.21286660974093974, "alpha_spy_p3m": 0.15527504944335924, "alpha_c_p3m": 0.029654036299375797, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 0.7880794857340423, "ret_signed_p6m": 0.7880794857340423, "alpha_spy_p6m": 0.7181730036840943, "alpha_c_p6m": 0.4442521067757299, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.2027, -0.2055, -0.1689, -0.1717, -0.1781, -0.1689, -0.1616, -0.1543, -0.0968, -0.0794, -0.0694, -0.0721, -0.0895, -0.0977, -0.095, -0.1041, -0.0977, -0.1196, -0.1196, -0.0995, -0.1196, -0.1306, -0.1434, -0.1388, -0.1342, -0.1169, -0.1205, -0.1142, -0.0959, -0.074, -0.0694, -0.0657, -0.0822, -0.0539, -0.0575, -0.0511, -0.0511, -0.0749, -0.0612, -0.032, -0.031, -0.0274, -0.0329, -0.0566, -0.063, -0.0255, -0.0189, -0.0189, -0.0416, -0.0322, -0.0199, -0.0199, -0.0151, -0.035, -0.0274, -0.0142, 0.0009, -0.0047, -0.0085, -0.0265, 0.0009, 0.0, -0.0066, 0.0, -0.0085, -0.0227, -0.1012, -0.053, -0.0558, -0.0795, -0.0804, -0.0624, -0.0568, -0.053, -0.0416, -0.0511, -0.0596, -0.07, -0.0445, -0.07, -0.0833, -0.0889, -0.0672, -0.0653, -0.0501, -0.052, -0.0568, -0.0624, -0.0624, -0.0274, -0.0227, -0.0246, -0.0095, 0.0889, 0.0587, 0.0691, 0.0511, 0.0454, 0.0596, 0.0691, 0.0691, 0.0851, 0.0842, 0.1079, 0.1069, 0.0918, 0.0842, 0.0568, 0.053, 0.0492, 0.053, 0.0549, 0.0482, 0.0776, 0.0766, 0.122, 0.1079, 0.0482, 0.1097, 0.071, 0.1287, 0.1902, 0.2034, 0.2289, 0.1996, 0.1788, 0.2129, 0.228, 0.2517, 0.264, 0.3652, 0.4229, 0.5147, 0.5317, 0.4541, 0.4976, 0.4541, 0.4399, 0.4551, 0.4172, 0.4182, 0.3841, 0.3822, 0.3529, 0.3406, 0.3311, 0.3009, 0.3075, 0.3396, 0.3415, 0.3623, 0.3623, 0.4144, 0.403, 0.4447, 0.457, 0.4342, 0.439, 0.4891, 0.5109, 0.5544, 0.5516, 0.491, 0.474, 0.4475, 0.3955, 0.4068, 0.4342, 0.4522, 0.4693, 0.4711, 0.4711, 0.4872, 0.5156, 0.5166, 0.5232, 0.5232, 0.5951, 0.6045, 0.6556, 0.667, 0.6395, 0.6689, 0.7285, 0.7692, 0.7862, 0.8155, 0.8363, 0.8363, 0.7881]}
{"unit_id": "orig:1948900335541780852", "ticker_idx": 4, "ts": "2025-07-26T00:17:03+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "substrate manufacturers like Unimicron", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPM maintains Overweight on Hanmi Semi and BESI, Neutral on ASMPT; selectively bullish on ASE and Unimicron as advanced packaging beneficiaries, cautious on mainstream semi recovery and hybrid bonding adoption pace.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Asian Tech- JPM\n\nKey highlights from BESI 2Q25 results\n\nSemiconductor Equipment Industry Outlook: Hopes and Concerns for H2 2025 📊\n\n🌅 Mainstream Recovery, but Double-Dip Concerns...\n\nBESI's Q2 results were still impacted by weak demand for mobile and automotive semiconductors, but management emphasized early signs of recovery in mainstream semiconductors. They stated that demand for HPC in high-end mobile and consumer markets is driving growth, and they anticipate a semiconductor upturn in H2 2025. This aligns with ASMPT's recent outlook.\n\nHowever, we are more cautious about the mainstream semiconductor recovery. Recent guidance from fabless companies and IDMs suggests a potential double-dip in the second half of this year. This is due to the fading impact of Chinese subsidies and the strong tariff avoidance demand seen in the first half of the year also diminishing.\n\n🔗 Hybrid Bonding, Driving Force in HBM4 and Logic... But Cautiously\n\nBESI reported that hybrid bonding revenue more than doubled year-over-year in H1 2025 and expects significant order increases in H2. Management expressed confidence that hybrid bonding is progressing smoothly in HBM4 (estimated to be Samsung's 16-Hi HBM4) and logic (primarily TSMC's SoIC-X), preparing for customer qualification in 2026 and mass production in 2027.\n\nHowever, we remain cautious about the rapid adoption of hybrid bonding, especially in DRAM applications. We believe that TCB will remain the preferred option for the next two years due to better TCO and improved throughput. The transition to hybrid bonding is likely to start with 20Hi/24Hi HBM4e or HBM5 stacks.\n\nIn logic, SoIC adoption also remains slow (only AMD among AI customers by 2027, and for high-end Mac processors like Apple's M5 Pro/Max. Nvidia seems more cautious, and CPO appears delayed). NAND's 3D stacking could emerge as a new application for hybrid bonding in the mid-term, but it will require much lower specifications and precision.\n\n🎯 Focus on BESI's New TCB Platform, Memory TCB Market Intensifies\n\nBESI's new TCB platform is attracting decent interest (estimated to be from Micron). The improved cycle time and accuracy of the next-generation TCB platform look highly competitive. BESI announced that it plans to ship 5 system orders (revenue scale over $20 million) in late Q3 to Q4 this year, which appears to be for pilot testing given the low volume.\n\nWe believe the memory TCB market is becoming increasingly competitive. Multiple vendors (ASMPT, Hanmi Semiconductor, Hanwha Systems, Shinkawa, SEMES, BESI, etc.) are competing for market share. This competition will put pressure on margins and necessitate further innovation to maintain a competitive edge.\n\n🚀 Advanced Packaging Market Growing Rapidly - Selective Beneficiaries Expected\n\nThe advanced packaging market is expected to show rapid compound annual growth (Advanced Packaging ~12% CAGR to '29, High-Performance Packaging ~37% CAGR to '29). This will benefit major OSATs like ASE and substrate manufacturers like Unimicron. Meanwhile, increasing complexity and precision requirements are likely to positively impact test vendors (BESI noted the assembly equipment market needs to grow 63% from 2024-29).\n\nNevertheless, we maintain a more selective view on equipment vendors due to the challenge of fierce competition among multiple vendors in some areas. We maintain an Overweight rating on Hanmi Semiconductor and BESI, and a Neutral rating on ASMPT.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015037593984962405, "ret_signed_m1d": -0.015037593984962405, "alpha_spy_m1d": -0.015288757197273184, "alpha_c_m1d": -0.007123188839357586, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.00791440514560482, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05263157894736836, "ret_signed_p1d": 0.05263157894736836, "alpha_spy_p1d": 0.05526932418785446, "alpha_c_p1d": 0.05259375317288573, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 3.782577448263602e-05, "ret_p1w": 0.015037593984962516, "ret_signed_p1w": 0.015037593984962516, "alpha_spy_p1w": 0.0240966712640851, "alpha_c_p1w": 0.0218157175063054, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0067781235213428825, "ret_p1m": 0.06766917293233088, "ret_signed_p1m": 0.06766917293233088, "alpha_spy_p1m": 0.054763790530783396, "alpha_c_p1m": 0.07164517236239165, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": -0.003975999430060773, "ret_p3m": 0.1729323308270676, "ret_signed_p3m": 0.1729323308270676, "alpha_spy_p3m": 0.1153407705294871, "alpha_c_p3m": 0.0768192424361227, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.0961130883909449, "ret_p6m": 1.3157894736842106, "ret_signed_p6m": 1.3157894736842106, "alpha_spy_p6m": 1.2458829916342626, "alpha_c_p6m": 1.2385574803192692, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.07723199336494146, "price_path": [-0.3026, -0.3026, -0.2974, -0.313, -0.3144, -0.3248, -0.3122, -0.3026, -0.2498, -0.2587, -0.2387, -0.2498, -0.2424, -0.2647, -0.2535, -0.2387, -0.2424, -0.2127, -0.2275, -0.1793, -0.1867, -0.2127, -0.2127, -0.2498, -0.2424, -0.235, -0.235, -0.235, -0.2573, -0.2053, -0.1904, -0.2053, -0.2387, -0.235, -0.2461, -0.2313, -0.2535, -0.2573, -0.183, -0.1718, -0.1533, -0.1533, -0.1607, -0.1533, -0.1681, -0.1681, -0.131, -0.1236, -0.1458, -0.1384, -0.0901, -0.0564, -0.0677, -0.0977, -0.0075, 0.0188, 0.0038, 0.0, 0.0038, -0.0338, -0.0113, -0.0038, -0.015, 0.0, 0.0526, 0.0677, 0.0338, 0.0338, 0.015, 0.0188, 0.0075, 0.0226, 0.0301, 0.0714, 0.0414, 0.0376, 0.0376, 0.0188, 0.0677, 0.0451, 0.0038, 0.0038, 0.015, 0.0526, 0.0677, 0.1429, 0.1278, 0.0865, 0.0113, -0.015, -0.0038, 0.0038, 0.0075, 0.0038, 0.015, 0.0639, 0.0714, 0.0865, 0.0677, 0.1429, 0.109, 0.1165, 0.1842, 0.1917, 0.1955, 0.1805, 0.1805, 0.1203, 0.1203, 0.1391, 0.1654, 0.1504, 0.1729, 0.1729, 0.188, 0.203, 0.2105, 0.2105, 0.1541, 0.0977, 0.1917, 0.2444, 0.2068, 0.203, 0.203, 0.1955, 0.1729, 0.1729, 0.2293, 0.2218, 0.2481, 0.203, 0.2293, 0.3421, 0.2669, 0.2236, 0.2348, 0.2049, 0.2273, 0.2422, 0.2497, 0.3317, 0.2895, 0.2594, 0.2782, 0.2068, 0.2895, 0.203, 0.2331, 0.3233, 0.3421, 0.406, 0.4023, 0.4812, 0.4737, 0.4774, 0.4925, 0.6165, 0.6579, 0.6992, 0.7105, 0.6917, 0.6917, 0.6654, 0.609, 0.594, 0.5564, 0.6165, 0.6579, 0.6203, 0.6316, 0.6316, 0.6579, 0.6579, 0.6504, 0.6541, 0.6541, 0.6466, 0.7293, 0.7331, 0.6992, 0.7105, 0.6955, 0.8308, 0.9098, 0.8797, 0.8759, 1.0602, 1.2632, 1.3158]}
{"unit_id": "reply:1965982096964596139", "ticker_idx": 0, "ts": "2025-09-11T03:33:52+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I ultimately think BESI will win", "tweet_type": "analysis", "timeline": "years", "summary": "Long BESI as winner in hybrid bonding race; long AMAT as best-in-class CMP equipment provider and BESI's largest shareholder.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@foggycap I ultimately think BESI will win. Hybrid bonders require the CMP process, and AMAT is the best at CMP equipment — plus, AMAT is BESI’s largest shareholder, right?", "reply_text": null, "parent_text": "@Jukanlosreve Gotcha, so no bumps when stacking NAND. \n\nWho do you think will win in the memory hybrid bonding race? \n\nBESI clear leader today but hasn’t seemed to figure out memory yet… ASMPT similar. Then Hanmi/Hanwha/LG etc all talk about HBonder launches by ‘27-28", "quoted_text": null, "parent_author": "foggycap", "ret_m1d": -0.011438620350767925, "ret_signed_m1d": -0.011438620350767925, "alpha_spy_m1d": -0.0031968469379024533, "alpha_c_m1d": -0.001238872861649698, "bench_spy_m1d": -0.008241773412865472, "bench_c_m1d": -0.010199747489118227, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015838063734204266, "ret_signed_p1d": -0.015838063734204266, "alpha_spy_p1d": -0.015503450822376474, "alpha_c_p1d": -0.017121366943167238, "bench_spy_p1d": -0.00033461291182779185, "bench_c_p1d": 0.0012833032089629715, "ret_p1w": 0.0981082967658804, "ret_signed_p1w": 0.0981082967658804, "alpha_spy_p1w": 0.09106794750300695, "alpha_c_p1w": 0.05520360577443517, "bench_spy_p1w": 0.0070403492628734465, "bench_c_p1w": 0.04290469099144523, "ret_p1m": 0.23449185229840763, "ret_signed_p1m": 0.23449185229840763, "alpha_spy_p1m": 0.23874897091901248, "alpha_c_p1m": 0.1640479639531396, "bench_spy_p1m": -0.0042571186206048495, "bench_c_p1m": 0.07044388834526805, "ret_p3m": 0.2608887828143589, "ret_signed_p3m": 0.2608887828143589, "alpha_spy_p3m": 0.21937066790089066, "alpha_c_p3m": 0.046793397760444666, "bench_spy_p3m": 0.04151811491346824, "bench_c_p3m": 0.21409538505391423, "ret_p6m": 0.3752749863220377, "ret_signed_p6m": 0.3752749863220377, "alpha_spy_p6m": 0.3469818968602103, "alpha_c_p6m": 0.11924839159797407, "bench_spy_p6m": 0.028293089461827403, "bench_c_p6m": 0.25602659472406364, "price_path": [0.1175, 0.1161, 0.1065, 0.0972, 0.0836, 0.106, 0.1342, 0.1368, 0.1421, 0.1478, 0.1179, 0.0748, 0.0845, 0.073, 0.0651, 0.0682, 0.0752, 0.073, 0.121, 0.1188, 0.0968, 0.1113, 0.0554, 0.1219, 0.1258, 0.1452, 0.1069, 0.088, 0.1175, 0.0304, 0.0805, 0.0559, 0.077, 0.048, 0.0308, 0.0418, 0.0145, 0.0101, 0.0554, 0.0625, 0.0695, 0.0889, 0.0867, 0.0792, 0.0431, 0.0365, 0.0506, 0.0304, 0.0158, 0.0471, 0.0484, 0.062, 0.0541, 0.0537, 0.0128, 0.0163, -0.0594, -0.0691, -0.0572, -0.0436, -0.0246, -0.0194, -0.0114, 0.0, -0.0158, 0.0396, 0.026, 0.0176, 0.0981, 0.0691, 0.0805, 0.1166, 0.1311, 0.1148, 0.077, 0.1122, 0.1161, 0.1245, 0.1751, 0.1483, 0.2912, 0.2917, 0.2877, 0.2794, 0.2345, 0.256, 0.2363, 0.2802, 0.2767, 0.2468, 0.2776, 0.2697, 0.2187, 0.2692, 0.2939, 0.3022, 0.2899, 0.2992, 0.2939, 0.2987, 0.2829, 0.2543, 0.2451, 0.2121, 0.1804, 0.2006, 0.2125, 0.1909, 0.1773, 0.1562, 0.1456, 0.1188, 0.1359, 0.1443, 0.0805, 0.0946, 0.0963, 0.1403, 0.143, 0.143, 0.1417, 0.1557, 0.1962, 0.2099, 0.2279, 0.2921, 0.2609, 0.2358, 0.2046, 0.1623, 0.1654, 0.1755, 0.1377, 0.1566, 0.1505, 0.1544, 0.1606, 0.1575, 0.1575, 0.1575, 0.1623, 0.1711, 0.1769, 0.1769, 0.3119, 0.3528, 0.4158, 0.3973, 0.3295, 0.3295, 0.4267, 0.4712, 0.421, 0.5279, 0.5213, 0.4853, 0.5213, 0.5315, 0.5398, 0.5403, 0.5385, 0.5464, 0.491, 0.4303, 0.4474, 0.4483, 0.4092, 0.3863, 0.4144, 0.4562, 0.4729, 0.4875, 0.5037, 0.4844, 0.5504, 0.5706, 0.5847, 0.6454, 0.5222, 0.6234, 0.6511, 0.692, 0.7387, 0.6621, 0.6665, 0.6586, 0.5957, 0.6806, 0.6599, 0.3753]}
{"unit_id": "reply:1965982096964596139", "ticker_idx": 1, "ts": "2025-09-11T03:33:52+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMAT is the best at CMP equipment — plus, AMAT is BESI's largest shareholder, right?", "tweet_type": "analysis", "timeline": "years", "summary": "Long BESI as winner in hybrid bonding race; long AMAT as best-in-class CMP equipment provider and BESI's largest shareholder.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@foggycap I ultimately think BESI will win. Hybrid bonders require the CMP process, and AMAT is the best at CMP equipment — plus, AMAT is BESI’s largest shareholder, right?", "reply_text": null, "parent_text": "@Jukanlosreve Gotcha, so no bumps when stacking NAND. \n\nWho do you think will win in the memory hybrid bonding race? \n\nBESI clear leader today but hasn’t seemed to figure out memory yet… ASMPT similar. Then Hanmi/Hanwha/LG etc all talk about HBonder launches by ‘27-28", "quoted_text": null, "parent_author": "foggycap", "ret_m1d": -0.03955322677508233, "ret_signed_m1d": -0.03955322677508233, "alpha_spy_m1d": -0.03131145336221686, "alpha_c_m1d": -0.0293534792859641, "bench_spy_m1d": -0.008241773412865472, "bench_c_m1d": -0.010199747489118227, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013811232763658032, "ret_signed_p1d": -0.013811232763658032, "alpha_spy_p1d": -0.01347661985183024, "alpha_c_p1d": -0.015094535972621004, "bench_spy_p1d": -0.00033461291182779185, "bench_c_p1d": 0.0012833032089629715, "ret_p1w": 0.11525121310500208, "ret_signed_p1w": 0.11525121310500208, "alpha_spy_p1w": 0.10821086384212864, "alpha_c_p1w": 0.07234652211355685, "bench_spy_p1w": 0.0070403492628734465, "bench_c_p1w": 0.04290469099144523, "ret_p1m": 0.23391134382101253, "ret_signed_p1m": 0.23391134382101253, "alpha_spy_p1m": 0.23816846244161738, "alpha_c_p1m": 0.16346745547574448, "bench_spy_p1m": -0.0042571186206048495, "bench_c_p1m": 0.07044388834526805, "ret_p3m": 0.5731041408836721, "ret_signed_p3m": 0.5731041408836721, "alpha_spy_p3m": 0.5315860259702039, "alpha_c_p3m": 0.3590087558297579, "bench_spy_p3m": 0.04151811491346824, "bench_c_p3m": 0.21409538505391423, "ret_p6m": 0.9146774332878949, "ret_signed_p6m": 0.9146774332878949, "alpha_spy_p6m": 0.8863843438260675, "alpha_c_p6m": 0.6586508385638312, "bench_spy_p6m": 0.028293089461827403, "bench_c_p6m": 0.25602659472406364, "price_path": [0.0346, 0.0202, 0.0129, 0.0129, -0.0069, 0.0077, 0.0559, 0.0729, 0.0755, 0.0737, 0.0729, 0.0769, 0.1135, 0.1196, 0.1196, 0.118, 0.1427, 0.1451, 0.1605, 0.1599, 0.1551, 0.1679, 0.1417, 0.1282, 0.116, 0.1288, 0.0967, 0.0959, 0.1025, 0.0882, 0.1151, 0.1042, 0.1099, 0.0552, 0.0548, 0.0714, 0.0499, 0.044, 0.0733, 0.0834, 0.0805, 0.1044, 0.1136, 0.1032, -0.052, -0.0417, -0.0493, -0.0567, -0.0606, -0.045, -0.048, -0.0331, -0.0339, -0.0287, -0.0552, -0.0552, -0.0739, -0.0817, -0.07, -0.0435, -0.0476, -0.0391, -0.0396, 0.0, -0.0138, 0.0046, 0.0199, 0.0469, 0.1153, 0.1172, 0.1785, 0.1805, 0.1839, 0.1731, 0.1985, 0.2045, 0.2033, 0.2797, 0.3141, 0.2785, 0.316, 0.2434, 0.2783, 0.2947, 0.2339, 0.2899, 0.2823, 0.3375, 0.3383, 0.3223, 0.3408, 0.3282, 0.2963, 0.3428, 0.3444, 0.3596, 0.3379, 0.3855, 0.3667, 0.37, 0.3971, 0.3529, 0.4158, 0.3725, 0.3522, 0.3816, 0.3439, 0.356, 0.312, 0.3283, 0.3442, 0.3231, 0.3819, 0.2969, 0.3191, 0.3598, 0.4278, 0.472, 0.472, 0.4854, 0.5001, 0.5624, 0.5819, 0.5866, 0.5782, 0.5791, 0.5731, 0.6203, 0.5906, 0.5264, 0.5385, 0.5242, 0.462, 0.4928, 0.5099, 0.5252, 0.5324, 0.5357, 0.5357, 0.5422, 0.549, 0.5309, 0.5133, 0.5133, 0.5833, 0.6743, 0.7431, 0.7207, 0.6585, 0.7736, 0.8092, 0.7953, 0.7777, 0.879, 0.9257, 0.9257, 0.874, 0.9152, 0.8773, 0.8984, 0.8812, 0.9592, 0.983, 1.01, 0.898, 0.9338, 0.8765, 0.7525, 0.7901, 0.8992, 0.9466, 0.9378, 1.0014, 0.9338, 1.09, 1.09, 1.1148, 1.1747, 1.1805, 1.2133, 1.2025, 1.2283, 1.3286, 1.2153, 1.1951, 1.1944, 1.0714, 1.1094, 1.0432, 0.9147]}
{"unit_id": "orig:1959783335489892427", "ticker_idx": 0, "ts": "2025-08-25T01:02:12+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raised its July quarter revenue estimate to $46.6 billion... maintained its \"Overweight\" rating and lifted its price target from $200 to $206", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley reiterates Overweight on NVDA, raises PT $200→$206 on strong demand, supply improvements, and ~85% market share through 2026.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA 2Q Earnings Preview: Strong Demand and Supply Improvements Support Continued Growth\nMorgan Stanley (J. Moore, 25/08/18)\n\nEarnings and Target Price Update\n\nMorgan Stanley raised its July quarter revenue estimate to $46.6 billion (previously $45.2 billion) and October quarter revenue estimate to $52.5 billion (previously $51.3 billion). The firm maintained its “Overweight” rating and lifted its price target from $200 to $206, reflecting an approximate 4% increase in FY27 EPS.\n\nDemand Remains Strong\n\nHyperscale customers described demand using terms like “remarkable” and “insatiable” in earnings calls. The focus has shifted from last year’s supply constraints toward a surge in inference demand, underscoring the durability of NVIDIA’s growth. All four major hyperscale customers are maintaining strong growth, while second-tier cloud providers such as CoreWeave and sovereign customers are also adding new momentum.\n\nSupply Bottlenecks Improving\n\n- Cabinet assembly is accelerating: Hon Hai expects Q3 cabinet shipments to be three times higher than in Q2.\n\n- China hardware teams estimate that the top four ODMs will double cabinet output in Q3, reaching about 34,000 cabinets for the year, equivalent to roughly 2.4 million GPUs, or about $90 billion in revenue.\n\n- Testing bottlenecks are easing: KYEC expects Q3 test volumes to rise to 1.5 million units.\n\n- Overall, every stage of the Blackwell supply chain is showing improvements.\n\nProduct Mix and China Uncertainty\n\n- Hopper production has ceased; in Q2 it still contributed around $6 billion, but is expected to decline to $2.8 billion in July.\n\n- B200 server shipments are compensating for cabinet shortages, but with lower ASPs and less additional content.\n\n- The company may exclude China contributions from October guidance; however, if license approvals accelerate, there could be upside.\n\n- Previous U.S. Commerce Department restrictions on China created about $8 billion in headwinds, and recovery for the H20 remains uncertain.\n\nCompetitive Landscape Outlook\n\n- Morgan Stanley expects NVIDIA to further expand market share in 2025 and maintain close to 85% share in 2026.\n\n- Even with Google ramping up ASIC development, its spending on NVIDIA is still expected to grow over 3x YoY.\n\n- AMD’s UALink is projected to scale in 2026, but NVIDIA will already be ahead with the Rubin platform’s next-gen NVLink Fusion.\n\n- Overall, competitors are unlikely to surpass NVIDIA in mainstream workload performance.\n\nConclusion\nMorgan Stanley believes that in the short term, supply remains the main constraint, while demand determines the outlook through 2026. Even excluding China, NVIDIA’s revenue remains below actual demand, and growth momentum will continue into next year.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010121735824010547, "ret_signed_m1d": -0.010121735824010547, "alpha_spy_m1d": -0.014542348693156959, "alpha_c_m1d": -0.009169051936335437, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01090037658917331, "ret_signed_p1d": 0.01090037658917331, "alpha_spy_p1d": 0.006713340239282495, "alpha_c_p1d": 0.0012719889215084823, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.03131088170414875, "ret_signed_p1w": -0.03131088170414875, "alpha_spy_p1w": -0.035326711638982156, "alpha_c_p1w": -0.01896056982958516, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": -0.007618864936297265, "ret_signed_p1m": -0.007618864936297265, "alpha_spy_p1m": -0.042762646229730694, "alpha_c_p1m": -0.10213433451457077, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.00467260846353823, "ret_signed_p3m": 0.00467260846353823, "alpha_spy_p3m": -0.013801684658514723, "alpha_c_p3m": -0.10141057030732781, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.028812254552892602, "ret_signed_p6m": 0.028812254552892602, "alpha_spy_p6m": -0.04013497353668627, "alpha_c_p6m": -0.36196749018156815, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.2503, -0.226, -0.2485, -0.236, -0.2147, -0.2108, -0.2215, -0.2119, -0.2068, -0.1994, -0.2057, -0.1936, -0.2104, -0.1953, -0.1985, -0.1909, -0.1909, -0.2, -0.1982, -0.1775, -0.1418, -0.1379, -0.1227, -0.1214, -0.1474, -0.1255, -0.1138, -0.1138, -0.12, -0.1102, -0.0942, -0.0874, -0.0828, -0.0875, -0.0507, -0.0469, -0.0379, -0.0412, -0.0469, -0.0711, -0.0502, -0.0338, -0.0351, -0.017, -0.0239, -0.003, -0.0108, -0.0339, 0.0011, -0.0086, -0.0022, 0.0053, 0.0161, 0.0125, 0.0186, 0.0099, 0.0123, 0.0036, 0.0122, -0.0232, -0.0245, -0.0269, -0.0101, 0.0, 0.0109, 0.01, 0.002, -0.0313, -0.0313, -0.0502, -0.0511, -0.0453, -0.0711, -0.064, -0.0503, -0.0138, -0.0146, -0.011, -0.0114, -0.0274, -0.0529, -0.0198, -0.0174, 0.0212, -0.0076, -0.0157, -0.0117, -0.009, 0.0114, 0.0377, 0.0414, 0.0506, 0.0435, 0.0319, 0.0291, 0.0518, 0.071, 0.0187, 0.0474, 0.0013, 0.0002, 0.0112, 0.019, 0.0158, 0.0076, 0.0027, 0.0131, 0.0359, 0.065, 0.1181, 0.1515, 0.1284, 0.1262, 0.1506, 0.1051, 0.0857, 0.0461, 0.0464, 0.1071, 0.0743, 0.0779, 0.0393, 0.0577, 0.0378, 0.0087, 0.0374, 0.0047, -0.0051, 0.0153, -0.011, 0.0026, 0.0026, -0.0156, 0.0007, 0.0092, -0.0012, 0.02, 0.0146, 0.032, 0.0288, 0.0222, 0.0063, -0.0265, -0.0195, -0.0115, -0.0492, -0.0314, 0.0067, 0.0217, 0.0524, 0.0491, 0.0491, 0.0597, 0.0469, 0.0431, 0.0373, 0.0373, 0.0504, 0.0463, 0.0414, 0.0518, 0.0292, 0.0282, 0.0286, 0.0335, 0.0186, 0.0404, 0.0358, 0.0358, -0.0096, 0.0196, 0.0281, 0.0438, 0.0372, 0.0486, 0.0652, 0.0708, 0.0631, 0.0324, 0.0031, -0.0311, -0.044, 0.0313, 0.057, 0.0487, 0.0571, 0.0398, 0.0168, 0.0168, 0.0288]}
{"unit_id": "thread:1948176460541558891", "ticker_idx": 0, "ts": "2025-07-24T00:20:38+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM plays a critical role in performance improvement within the AI market. Considering its continued importance, there is no doubt about its growth potential. High growth will continue. Given the secured visibility for HBM supply next year, we decided to make proactive investments to respond in a timely manner. We expect to be able to respond quickly with additional supply. SK Hynix will strive to take leadership in new memory areas like PIM", "tweet_type": "analysis", "timeline": "quarters", "summary": "SK Hynix 2Q25 earnings call: strong HBM demand visibility into 2026, proactive capex, NAND normalizing, and resumed NVDA China exports as upside catalyst — durably bullish.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "■ SK Hynix 2Q25 Earnings Call Key Q&A (1)\n\n1. What is the mid-to-long-term demand outlook and growth momentum for HBM beyond 2025?\n\n- The AI market is showing rapid growth driven by expanding big tech CAPEX, increasing token usage, and the growth rate of AI startups.\n\n- In particular, AI demand is expanding from training to inference and becoming more sophisticated, leading to a steep increase in workloads. Bandwidth bottlenecks are also intensifying.\n\nHBM plays a critical role in performance improvement within the AI market.Considering its continued importance, there is no doubt about its growth potential. The AI market is expanding into areas such as AI agents, which will increase computational demands. This will drive market demand growth.\n\nMoving forward, while the HBM market may not experience the initial explosive growth rates, it is seeing rapid expansion and diversification of its customer base.Considering the launch of their new products and services, high growth will continue.\n\n2. What is the status of supply negotiations with key customers for HBM 2026?\n\n- It is difficult to discuss customer-related specifics in detail. Things are proceeding as planned. We are closely discussing product mix and pricing due to the increasing diversity of customer projects.\n\nAs a collaborative partner in AI semiconductors with our key customers, we are negotiating for optimal supply conditions. We are preparing to stably supply the highest quality products on time, grow with our customers, and respond to increased supply needs in a timely manner.\n\n---\n\n■ SK Hynix 2Q25 Earnings Call Key Q&A (2)\n\n3. What was the intensity of the pull-in demand impact in Q2, current inventory levels, and the demand outlook for the second half?\n\n- Our shipments in Q2 significantly exceeded the initial guidance, indicating higher-than-expected purchasing demand.\n\n- Customers had initially aimed for conservative inventory levels in the first half of this year but shifted towards securing adequate inventory due to tariff uncertainties.\n\n- Purchasing demand significantly increased from customers who had been holding low inventory.\n\n- Additionally, for some legacy products, there was competition to secure volume among small and medium-sized companies, including module makers.\n\nWe achieved much higher shipments in the first half than initially planned. Demand slowdown persists in the second half.\nWe believe there's a low probability of abrupt market supply-demand fluctuations.\nCustomers' inventory levels in Q2 did not rise to concerning levels.\nMemory suppliers' inventories have also substantially decreased, so future supply increases will only be possible due to production increases.\nWhile purchasing demand may be influenced by future tariff policies,we will focus on stable business operations, primarily for products with secured demand visibility.\n\n4. Concerns about reduced utilization of Chinese factories. What are the future operational plans for Chinese fabs?\n\n- Currently, there are no issues with the Validated End-User (VEU) status that SK Hynix secured in October 2023. We plan to continue operating Chinese fabs while complying with regulations and considering customer needs.\n\nRecent circumstances show that the supply and demand for legacy DRAM are creating concerns for Chinese fab operations. For instance, the industry-wide expansion of HBM mass production is acting as a constraint on general DRAM. A shortage of legacy supply is emerging during the transition process to D5/LPD5 products.\nWe addressed the short-term increase in legacy demand with inventory from the last quarter. After reviewing the IT ecosystem, we've determined that demand for legacy products will remain at a certain level for a considerable period, so we plan to actively utilize our Chinese fabs to address legacy products.\nChinese fabs are also important production facilities for global memory semiconductor supply and demand. We will continue to monitor US regulations and communicate with various countries in the future.\n\n---\n\n■ SK Hynix 2Q25 Earnings Call Key Q&A (3)\n\n5. NAND Q2 shipment growth significantly exceeded guidance. What's the background?\n\n- In the last earnings announcement, we provided a guidance of 20%.\n- At that time, there were signs of rising spot prices and decreasing customer inventory, but demand visibility for some products wasn't high.\n\nHowever, in Q2, hyperscaler customers increased their AI investments, leading to increased eSSD demand.\nThere was also pull-in demand, primarily for component products, due to tariff impacts.\nFurthermore, mobile demand increased thanks to promotional events in China.\nWith the increase in Q2 sales volume, our inventory levels have dropped to a normal level.\nWe'll continue to appropriately manage production and inventory according to future market conditions.\n\n6. As we move to HBM4, there are concerns about HBM profitability in the market due to increasing cost rates. What's the company's opinion on this?\n\n- HBM4 is a product with many technical changes, such as an increased number of I/O due to bandwidth and design modifications.\n\nWe aim to reflect the increased costs of HBM4 in its pricing as much as possible. We want to establish optimal pricing with customers to revitalize the HBM market.\n\n---\n\n■ SK Hynix 2Q25 Earnings Call Key Q&A (4)\n\n7. What are your operational plans for DDR4 products going forward?\n\n- With some suppliers deciding to exit, short-term supply shortage concerns have emerged.\n- As this is a period of generational transition from D4 to D5, we view this more as a demand concentration phenomenon rather than a significant change in D4 itself.\n\nOur D4 revenue proportion has been continuously decreasing since 2023 due to expanding D5 and HBM demand. From a double-digit percentage last year, it has fallen to a single-digit percentage this year. While we do have plans to end production of D4 for the mass market in the future, we will continue to support customers with long-term contracts.\n\n8. Capex. How much will this year's investment increase compared to the original plan? Will the increase be in equipment investment or infrastructure?\n\n- We are simultaneously pursuing investments for future growth and financial soundness. This year's investments are primarily focused on meeting the volumes agreed upon with customers. Long-term growth infrastructure investments, such as M15X and the Yongin fab construction, are also underway.\n\nGiven the secured visibility for HBM supply next year, we decided to make proactive investments to respond in a timely manner. Accordingly, our investment scale this year will increase compared to the original plan, with most of the increase going towards HBM equipment investment. The exact investment amount will be finalized once investment discussions with key customers are complete.Going forward, we will adhere to capex discipline while enhancing investment efficiency.\n\n---\n\n■ SK Hynix 2Q25 Earnings Call Key Q&A (5)\n\n9. Concerns about limited capacity expansion due to no available space in M16 fab. Considering M15X investment, what is the ramp-up plan by the end of next year? What is the timeline for 1c DRAM to ramp up to a meaningful proportion of 15-20%?\n\n- As the HBM market grows and demand increases, SK Hynix has been converting a lot of its general DRAM capacity to HBM.\n- Compared to the past, the required cleanroom size for producing the same volume has also significantly increased.\n- Currently, SK Hynix's existing fabs are being utilized for product mix optimization, technology transitions, and securing TSV.\n- For the mid-to-long term, M15X, Yongin, and advanced packaging US fabs are being prepared simultaneously.\n\nM15X is scheduled to open in Q4 this year. Mass production will begin next year.We plan to mass produce next-generation HBM products primarily. Since customer volume visibility for next year has been secured, we plan to gradually increase capacity.\nWhile it's difficult to disclose the M15X capacity plan for the end of next year, inferring the intent of the question, we can comment that there will be no instances where we cannot meet customer demand due to fab space constraints.\n1c transition will begin in the second half of this year. Next year will be the full-scale period. We will share concrete plans once they are finalized.\n\n10. Growth has slowed compared to the past, and suppliers remain conservative. NAND profitability improvement is limited. How long is this situation expected to continue?\n\n- The current slowdown in NAND market growth is due to stagnation in customer IT set demand and AI investment not yet leading to storage device purchases.\n\nHowever, the need for storage investment will gradually emerge. Increased AI utilization and a significant rise in token generation will make it difficult to cope with existing data processing methods.\nTherefore, we plan to offload AI inference data caching to eSSD. Initial demand will emerge in 2-3 years. If this trend materializes, eSSD will become part of the computing cache rather than just a storage device, increasing its proportion in AI systems. From this perspective, we expect significant growth to begin in the NAND market.\n\n---\n\n■ SK Hynix 2Q25 Earnings Call Key Q&A (6)\n\n11. Will next-generation DRAM technology also be efficient for HBM DRAM Die production?\n\n- Traditionally, the technologies pursued by memory companies have been reaching their limits in terms of performance and capacity. SK Hynix is primarily considering a transition to a Vertical Gate Platform and 3D technology based on groundbreaking changes in structure and materials below 10nm.\n\n- The Vertical Platform is a next-generation technology that significantly improves area efficiency and power efficiency by vertically stacking gates. 3D DRAM is a technology that achieves high capacity without further miniaturization. We are striving to secure technological capabilities so that these technologies can overcome existing weaknesses and become standardized.\n\n- However, technological attempts require time to secure stability and mass production feasibility. We will continue to maintain technological leadership by actively pursuing miniaturization with existing methods.\n\n- One of the foundations of SK Hynix's HBM competitiveness is its core DRAM technology. We expect that preparing for next-generation memory DRAM will allow us to maintain our competitive edge in the next generation as well.\n\n12. Regarding the issue of a major customer resuming AI chip exports to China. What is the upside demand effect?\n\n- It hasn't been long since the major customer resumed exports to China, so we are also currently verifying the information on the HBM products for those specific products.\n\n- However, we have been a key vendor supplying HBM products for those specific products until the export restrictions were imposed. We expect to be able to respond quickly with additional supply.\n\n- Based on close collaboration, we will work to ensure that this situation also brings a positive outcome for our company.\n\n---\n\nSK Hynix 2Q25 Earnings Call Key Q&A (7)\n\n13. There are concerns about intensifying HBM competition. As a leader in HBM, what is the company's future strategy in response to this?\n\n- Concerns about intensifying competition are natural. Competition is an inherent part of operating a memory business. It's crucial to deliver strong performance and bring value to investors.\n- As HBM has emerged as a critical product, the memory market has transformed from a pure commodity market. Now, leading players hold significant bargaining power. HBM greatly influences the performance of AI chips and AI systems. As a result, the advantage of proactively negotiating with customers is much stronger than in the past.\n- SK Hynix's corporate culture also contributed to its rise as a leader in the HBM market, where it was once weaker. Our customer-centric mindset and strong organizational teamwork are not easily copied by other companies. Ensuring our continued presence with HBM customers will be a key competitive advantage in the future HBM market.\n- To overcome the current constraints in AI Computing, we must respond with HBM and new memory products. SK Hynix will strive to take leadership in new memory areas like PIM, aligning with customer needs.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0018553554029104857, "ret_signed_m1d": -0.0018553554029104857, "alpha_spy_m1d": -0.0015244124395864533, "alpha_c_m1d": 0.0022794644218875337, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012987138500289652, "ret_signed_p1d": -0.012987138500289652, "alpha_spy_p1d": -0.0172114881406088, "alpha_c_p1d": -0.011909955870553657, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": 0.014842261023144454, "ret_signed_p1w": 0.014842261023144454, "alpha_spy_p1w": 0.018530562621140056, "alpha_c_p1w": 0.011437121541981288, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.06864553000706097, "ret_signed_p1m": -0.06864553000706097, "alpha_spy_p1m": -0.08581090407836389, "alpha_c_p1m": -0.08893745107351747, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.7799327721855513, "ret_signed_p3m": 0.7799327721855513, "alpha_spy_p3m": 0.7188832551236031, "alpha_c_p3m": 0.5801412948306914, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.8112586328636606, "ret_signed_p6m": 1.8112586328636606, "alpha_spy_p6m": 1.7147815308955834, "alpha_c_p6m": 1.4157205111608504, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666, 1.7852, 1.7443, 1.7592, 1.7852, 1.8113]}
{"unit_id": "reply:1945433237595324658", "ticker_idx": 0, "ts": "2025-07-16T10:40:02+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I believe Samsung has no choice but to pull it off — I've heard too much from people on the inside.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Samsung: insider signals suggest the company will overcome HBM yield/stack challenges through sustained R&D push.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "Semiconductors aren’t an industry that runs because of one genius. It’s a factory powered by hundreds of reasonably smart people working together.\n\nSamsung has recently been pushing countless engineers into nonstop R&D.\n\nI believe Samsung has no choice but to pull it off — I’ve heard too much from people on the inside.", "reply_text": null, "parent_text": "@Jukanlosreve I think you understimated the leaps that need to be done. 1a/1b DRAMs aren’t sufficient to bridge the gap. 1c has to yield a good HVM’able yields for 12Hi/16Hi outs to make sense. Coupled if this culture story is true, its a challenge to tackle both issue. https://t.co/JOrSOJSvbV", "quoted_text": null, "parent_author": "bobnorman_ch", "ret_m1d": -0.015455945066806476, "ret_signed_m1d": -0.015455945066806476, "alpha_spy_m1d": -0.012123893633051108, "alpha_c_m1d": -0.012363755463044468, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.003092189603762008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03091201226382756, "ret_signed_p1d": 0.03091201226382756, "alpha_spy_p1d": 0.02479232228631001, "alpha_c_p1d": 0.021828616473980933, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.009083395789846627, "ret_p1w": 0.026275094400549426, "ret_signed_p1w": 0.026275094400549426, "alpha_spy_p1w": 0.010270972600368022, "alpha_c_p1w": 0.020167818587848885, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.006107275812700541, "ret_p1m": 0.10664605760002899, "ret_signed_p1m": 0.10664605760002899, "alpha_spy_p1m": 0.07343657548745841, "alpha_c_p1m": 0.0717806367789382, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.03486542082109079, "ret_p3m": 0.4484739951797647, "ret_signed_p3m": 0.4484739951797647, "alpha_spy_p3m": 0.38333949433173875, "alpha_c_p3m": 0.3447807070362061, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1036932881435586, "ret_p6m": 1.1653323126504187, "ret_signed_p6m": 1.1653323126504187, "alpha_spy_p6m": 1.0544022452027795, "alpha_c_p6m": 1.0471348218388694, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11819749081154929, "price_path": [-0.1505, -0.149, -0.1551, -0.1444, -0.1444, -0.1444, -0.1428, -0.1428, -0.1474, -0.1474, -0.1659, -0.1659, -0.1659, -0.1612, -0.1612, -0.1582, -0.1152, -0.1259, -0.1182, -0.1198, -0.1275, -0.1428, -0.1413, -0.1444, -0.1597, -0.1674, -0.1597, -0.172, -0.1413, -0.1382, -0.1367, -0.1275, -0.1275, -0.1121, -0.0921, -0.0921, -0.0814, -0.0906, -0.0798, -0.086, -0.1044, -0.1213, -0.1075, -0.0814, -0.0906, -0.086, -0.109, -0.0706, -0.0583, -0.0752, -0.0603, -0.0757, -0.0696, -0.0603, -0.0139, -0.0216, -0.0464, -0.051, -0.0665, -0.0572, -0.0325, -0.034, -0.0155, 0.0, 0.0309, 0.0371, 0.0479, 0.0201, 0.0263, 0.0201, 0.0185, 0.0881, 0.0912, 0.1221, 0.1036, 0.0649, 0.0773, 0.0804, 0.0634, 0.0896, 0.1097, 0.0974, 0.0989, 0.1113, 0.1066, 0.1066, 0.0819, 0.0819, 0.0896, 0.0912, 0.1036, 0.1051, 0.0866, 0.0912, 0.0757, 0.0773, 0.0448, 0.068, 0.0788, 0.0835, 0.0742, 0.0835, 0.1051, 0.1221, 0.1345, 0.1654, 0.1824, 0.2272, 0.2087, 0.2411, 0.2411, 0.2906, 0.3091, 0.3199, 0.3308, 0.2875, 0.3072, 0.3025, 0.3351, 0.3934, 0.3934, 0.3934, 0.3934, 0.3934, 0.3934, 0.4656, 0.4485, 0.4221, 0.4749, 0.5168, 0.5199, 0.523, 0.5137, 0.5308, 0.4982, 0.5339, 0.5835, 0.5447, 0.5603, 0.6161, 0.6689, 0.7248, 0.6286, 0.5618, 0.5401, 0.5199, 0.5618, 0.6068, 0.6006, 0.596, 0.509, 0.5618, 0.5183, 0.4982, 0.5618, 0.4718, 0.5013, 0.5416, 0.596, 0.6068, 0.5603, 0.5649, 0.6053, 0.6224, 0.6317, 0.6829, 0.7, 0.6829, 0.6767, 0.6658, 0.6907, 0.627, 0.596, 0.6751, 0.6705, 0.6503, 0.7155, 0.731, 0.7248, 0.7248, 0.8164, 0.8642, 0.8705, 0.8705, 0.8705, 1.0046, 1.1544, 1.1669, 1.1997, 1.1653]}
{"unit_id": "orig:1970052496610758732", "ticker_idx": 0, "ts": "2025-09-22T09:08:11+00:00", "symbol": "DRAM", "kind": "sector", "geography": "US", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JP Morgan: Positive Long-Term DRAM Growth Outlook After the Memory Executive Summit", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPMorgan adopts positive long-term DRAM/HBM growth outlook: cHBM customization, power efficiency, Edge AI TAM 21% CAGR to $50B by 2030, hybrid bonding ahead of schedule.", "resolved_tickers": ["MU"], "resolution_method": "exchange_ticker", "resolution_reason": "US DRAM sector: Micron Technology MU", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Positive Long-Term DRAM Growth Outlook After the Memory Executive Summit\n    \n1. From Passive to Active: Customized HBM Redefines the Role of Memory\n    \n• In XPU-to-XPU environments connecting hundreds of chips, memory is expected to account for a larger share of total cost of ownership (TCO) in advanced AI infrastructure design. Memory vendors anticipate HBM evolving into an active component of system intelligence. Customized HBM (cHBM) will integrate diverse features, processing capabilities, and logic die designs (e.g., LPDDR working alongside HBM and compute logic integrated within HBM stacks), making customization a key differentiator of performance. Vendors disclosed that multiple customer projects are already underway, stressing the importance of collaboration with foundries and supply chain partners.\n    \n• Investor concerns over GPU vendors taking control of next-generation logic die design were deemed excessive. Memory companies are now providing full-stack solutions that include HBM dies, logic dies, LPDDR, PIM (processing-in-memory), and storage solutions across multiple AI infrastructure layers.\n    \n2. AI-Driven Explosive Power Consumption: HBM as the Key to Reducing AI Infrastructure TCO\n    \n• Compared to HBM3/3E, vendors are placing greater emphasis on HBM’s power-saving value proposition. With training compute demand increasing ~4x per year and power consumption surging accordingly (data center power demand rising from 1.5% of global electricity in 2025 to 3.0% in 2030; AI servers expected to account for 65% of data center consumption by 2030, representing a 56x increase vs. 2023), vendors shared design challenges and solutions for energy efficiency. Every 10% improvement in HBM power efficiency reduces rack-level power consumption by ~2%, enabling large-scale system-level gains.\n    \n3. Expansion of Emerging Memory Solutions Targeting Customized AI Markets\n    \n• Similar to creative approaches in packaging back-end processes (CoWoS, CoPoS, etc.), memory vendors are focusing on commercializing new memory solutions optimized for specific applications.\n    \n4. Edge AI as the Next Growth Phase: Expanding AI Inference Market\n    \n• As the focus shifts from training to inference, the Edge AI hardware TAM is projected to grow at a 21% CAGR between 2023–2030, reaching $50B by 2030. Demand is expanding from cloud toward edge devices, with smartphones, automobiles, and consumer electronics requiring low-latency, low-power, high-bandwidth memory. Winbond highlighted its CUBE (4-Hi HBM with TSV and TCB stacking) technology, while other vendors are expected to prepare VFO (Vertical Fan-Out) solutions for Edge AI. Suppliers anticipate advanced Edge AI solutions achieving meaningful market adoption within 2–3 years.\n    \n5. Accelerated Adoption of HBM Hybrid Bonding Solutions\n    \n• Vendors are moving faster than the base case scenario of 2H 2027 for adoption of hybrid bonding in HBM. Testing is already underway not only for 20-Hi HBM but also for 16-Hi HBM4E.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011481025777169696, "ret_signed_m1d": -0.011481025777169696, "alpha_spy_m1d": -0.006772294852597338, "alpha_c_m1d": 0.008570531977778018, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01087348305772684, "ret_signed_p1d": 0.01087348305772684, "alpha_spy_p1d": 0.0163169578975344, "alpha_c_p1d": 0.012332322371607751, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.004373843452624926, "ret_signed_p1w": -0.004373843452624926, "alpha_spy_p1w": 0.00036489580524323717, "alpha_c_p1w": -0.005894730709706697, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.22959944416106892, "ret_signed_p1m": 0.22959944416106892, "alpha_spy_p1m": 0.22292627200685078, "alpha_c_p1m": 0.15780512786668988, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5107863663515086, "ret_signed_p3m": 0.5107863663515086, "alpha_spy_p3m": 0.49634517812828816, "alpha_c_p3m": 0.43293928745085, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.8074702227486457, "ret_signed_p6m": 1.8074702227486457, "alpha_spy_p6m": 1.7985744088282953, "alpha_c_p6m": 1.5717408696055801, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2277, -0.2353, -0.2428, -0.252, -0.2663, -0.2612, -0.2578, -0.2578, -0.2715, -0.2442, -0.2574, -0.2522, -0.2435, -0.2795, -0.2704, -0.2927, -0.312, -0.3051, -0.3122, -0.3365, -0.3328, -0.3213, -0.3241, -0.3242, -0.3199, -0.303, -0.337, -0.3629, -0.3453, -0.3375, -0.3392, -0.3204, -0.2778, -0.2485, -0.224, -0.2451, -0.2389, -0.2658, -0.2495, -0.2586, -0.288, -0.2966, -0.2851, -0.2928, -0.2923, -0.2847, -0.2589, -0.2771, -0.2771, -0.2803, -0.2788, -0.2455, -0.202, -0.2014, -0.1785, -0.1496, -0.0853, -0.0449, -0.0416, -0.0352, -0.0281, 0.0259, -0.0115, 0.0, 0.0109, -0.0177, -0.0473, -0.0446, -0.0044, 0.0164, 0.1065, 0.1162, 0.1417, 0.1607, 0.1287, 0.1946, 0.1691, 0.1038, 0.1717, 0.137, 0.1667, 0.2311, 0.2301, 0.2568, 0.2296, 0.2064, 0.2565, 0.3313, 0.3379, 0.3489, 0.3775, 0.3616, 0.3602, 0.4266, 0.3253, 0.4436, 0.4487, 0.4462, 0.5397, 0.4656, 0.4886, 0.4403, 0.5003, 0.4707, 0.3889, 0.3732, 0.224, 0.2605, 0.3611, 0.3648, 0.3996, 0.3996, 0.4374, 0.4616, 0.4557, 0.4233, 0.3777, 0.4419, 0.5009, 0.5343, 0.6029, 0.571, 0.4657, 0.4436, 0.4133, 0.3708, 0.5108, 0.6164, 0.6812, 0.6793, 0.7426, 0.7426, 0.7311, 0.79, 0.7794, 0.7355, 0.7355, 0.918, 0.8981, 1.0883, 1.0648, 0.9886, 1.0984, 1.1032, 1.0561, 1.0271, 1.047, 1.2058, 1.2058, 1.2195, 1.3661, 1.4176, 1.4302, 1.366, 1.4946, 1.6469, 1.65, 1.5228, 1.6622, 1.5506, 1.3071, 1.3283, 1.4001, 1.332, 1.2697, 1.4952, 1.5173, 1.5032, 1.5032, 1.431, 1.5597, 1.5378, 1.6036, 1.5599, 1.5419, 1.6087, 1.527, 1.5076, 1.5094, 1.3088, 1.437, 1.4144, 1.2517, 1.3674, 1.4513, 1.546, 1.4649, 1.5912, 1.6865, 1.8075]}
{"unit_id": "orig:1962300280823067129", "ticker_idx": 0, "ts": "2025-08-31T23:43:39+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Maintain a positive view on SEC/SKH/MU", "tweet_type": "analysis", "timeline": "months", "summary": "CLSA reiterates positive view on Samsung, SK Hynix, and Micron; US curbs on China ops seen tightening memory supply mid-term.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250901_US curbs SEC/Hynix China operations - CLSA\n\n[Key Takeaways]\n\n(1) This U.S. Department of Commerce action will not cause production disruptions in the short term.\n\n(2) If disruptions do occur, memory supply will become even tighter.\n\n(3) In the mid to long term, domestic companies are expected to reduce their production share in China and accelerate fab construction in Korea.\n\n(4) Maintain a positive view on SEC/SKH/MU.\n\n[Contents]\n\n(1) Status of Korean memory companies’ China fabs as of 1H25\n\nSKH DRAM: 185K [36% of SKH DRAM / 10% of global DRAM]\nSKH NAND: 85K [37% of SKH NAND / 6% of global NAND]\nSEC NAND: 120K [30% of SEC NAND / 9% of global NAND]\n(2) For SKH’s DRAM fab, process upgrades to 1-a are underway, but still 50% remains on 1-z and 1-y processes. For NAND, mainly 144-layer, with review underway for upgrade to 192-layer.\n\n(3) For SEC’s China NAND fab, mainly based on 128-layer.\n\n(4) This U.S. action aims to restrict Korean companies from expanding production capacity and upgrading processes in China.\n\n(5) In the short term, there will be no production suspension, but the burden of securing export licenses each time U.S. equipment is purchased will increase, leading to greater uncertainty.\n\n(6) However, in the long term, fabs in China will face upgrade restrictions, falling behind fabs in Korea in terms of competitiveness, which could result in cost structure deterioration for domestic memory companies.\n\n(7) Accordingly, the need for Korean companies to accelerate construction of new domestic fabs will increase.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03106504207020322, "ret_signed_m1d": 0.03106504207020322, "alpha_spy_m1d": 0.03106504207020322, "alpha_c_m1d": 0.03106504207020322, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022189282367031327, "ret_signed_p1d": 0.022189282367031327, "alpha_spy_p1d": 0.029599577922905573, "alpha_c_p1d": 0.032286690889147485, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.010097408522116158, "ret_p1w": 0.0369822931329169, "ret_signed_p1w": 0.0369822931329169, "alpha_spy_p1w": 0.031122341678621268, "alpha_c_p1w": 0.027914041086600117, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.009068252046316783, "ret_p1m": 0.24666165438494003, "ret_signed_p1m": 0.24666165438494003, "alpha_spy_p1m": 0.21104124144777803, "alpha_c_p1m": 0.17132459834220382, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.07533705604273622, "ret_p3m": 0.5378960292863133, "ret_signed_p3m": 0.5378960292863133, "alpha_spy_p3m": 0.48128895549186157, "alpha_c_p3m": 0.4552339242988097, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.08266210498750359, "ret_p6m": 1.9862260440870387, "ret_signed_p6m": 1.9862260440870387, "alpha_spy_p6m": 1.9145381595485984, "alpha_c_p6m": 1.9139194831823991, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.0723065609046396, "price_path": [-0.1502, -0.1311, -0.1311, -0.1208, -0.1296, -0.1193, -0.1252, -0.1428, -0.159, -0.1458, -0.1208, -0.1296, -0.1252, -0.1472, -0.1105, -0.0987, -0.1149, -0.1006, -0.1154, -0.1095, -0.1006, -0.0562, -0.0636, -0.0873, -0.0917, -0.1065, -0.0976, -0.074, -0.0754, -0.0577, -0.0429, -0.0133, -0.0074, 0.003, -0.0237, -0.0178, -0.0237, -0.0251, 0.0414, 0.0444, 0.074, 0.0562, 0.0192, 0.0311, 0.034, 0.0178, 0.0429, 0.0621, 0.0503, 0.0518, 0.0636, 0.0592, 0.0592, 0.0355, 0.0355, 0.0429, 0.0444, 0.0562, 0.0577, 0.0399, 0.0444, 0.0296, 0.0311, 0.0, 0.0222, 0.0325, 0.037, 0.0281, 0.037, 0.0577, 0.074, 0.0858, 0.1154, 0.1317, 0.1746, 0.1568, 0.1879, 0.1879, 0.2352, 0.253, 0.2633, 0.2737, 0.2322, 0.2511, 0.2467, 0.2779, 0.3336, 0.3336, 0.3336, 0.3336, 0.3336, 0.3336, 0.4027, 0.3863, 0.3611, 0.4116, 0.4517, 0.4547, 0.4577, 0.4487, 0.4651, 0.4339, 0.4681, 0.5156, 0.4785, 0.4933, 0.5468, 0.5973, 0.6508, 0.5587, 0.4948, 0.474, 0.4547, 0.4948, 0.5379, 0.532, 0.5275, 0.4443, 0.4948, 0.4532, 0.4339, 0.4948, 0.4086, 0.4369, 0.4755, 0.5275, 0.5379, 0.4933, 0.4978, 0.5364, 0.5528, 0.5617, 0.6107, 0.627, 0.6107, 0.6048, 0.5944, 0.6181, 0.5572, 0.5275, 0.6033, 0.5988, 0.5795, 0.6419, 0.6568, 0.6508, 0.6508, 0.7385, 0.7843, 0.7902, 0.7902, 0.7902, 0.9187, 1.062, 1.0739, 1.1053, 1.0724, 1.0754, 1.0724, 1.0545, 1.0948, 1.1486, 1.2232, 1.2292, 1.168, 1.2322, 1.274, 1.271, 1.271, 1.3815, 1.4248, 1.3994, 1.3964, 1.2456, 1.501, 1.5249, 1.3785, 1.3681, 1.4845, 1.4756, 1.5054, 1.6667, 1.7055, 1.7055, 1.7055, 1.7055, 1.8369, 1.8384, 1.8817, 1.9862]}
{"unit_id": "orig:1962300280823067129", "ticker_idx": 1, "ts": "2025-08-31T23:43:39+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain a positive view on SEC/SKH/MU", "tweet_type": "analysis", "timeline": "months", "summary": "CLSA reiterates positive view on Samsung, SK Hynix, and Micron; US curbs on China ops seen tightening memory supply mid-term.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250901_US curbs SEC/Hynix China operations - CLSA\n\n[Key Takeaways]\n\n(1) This U.S. Department of Commerce action will not cause production disruptions in the short term.\n\n(2) If disruptions do occur, memory supply will become even tighter.\n\n(3) In the mid to long term, domestic companies are expected to reduce their production share in China and accelerate fab construction in Korea.\n\n(4) Maintain a positive view on SEC/SKH/MU.\n\n[Contents]\n\n(1) Status of Korean memory companies’ China fabs as of 1H25\n\nSKH DRAM: 185K [36% of SKH DRAM / 10% of global DRAM]\nSKH NAND: 85K [37% of SKH NAND / 6% of global NAND]\nSEC NAND: 120K [30% of SEC NAND / 9% of global NAND]\n(2) For SKH’s DRAM fab, process upgrades to 1-a are underway, but still 50% remains on 1-z and 1-y processes. For NAND, mainly 144-layer, with review underway for upgrade to 192-layer.\n\n(3) For SEC’s China NAND fab, mainly based on 128-layer.\n\n(4) This U.S. action aims to restrict Korean companies from expanding production capacity and upgrading processes in China.\n\n(5) In the short term, there will be no production suspension, but the burden of securing export licenses each time U.S. equipment is purchased will increase, leading to greater uncertainty.\n\n(6) However, in the long term, fabs in China will face upgrade restrictions, falling behind fabs in Korea in terms of competitiveness, which could result in cost structure deterioration for domestic memory companies.\n\n(7) Accordingly, the need for Korean companies to accelerate construction of new domestic fabs will increase.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05078115843636, "ret_signed_m1d": 0.05078115843636, "alpha_spy_m1d": 0.05078115843636, "alpha_c_m1d": 0.05078115843636, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017578138029290757, "ret_signed_p1d": 0.017578138029290757, "alpha_spy_p1d": 0.024988433585165004, "alpha_c_p1d": 0.03070296290806762, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.013124824878776864, "ret_p1w": 0.08203129040275492, "ret_signed_p1w": 0.08203129040275492, "alpha_spy_p1w": 0.07617133894845929, "alpha_c_p1w": 0.06057014166236763, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.021461148740387292, "ret_p1m": 0.35742191552228975, "ret_signed_p1m": 0.35742191552228975, "alpha_spy_p1m": 0.32180150258512774, "alpha_c_p1m": 0.23316702106535292, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.12425489445693683, "ret_p3m": 1.1265218823692713, "ret_signed_p3m": 1.1265218823692713, "alpha_spy_p3m": 1.0699148085748196, "alpha_c_p3m": 0.9287198178343916, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.1978020645348797, "ret_p6m": 2.9285929147770355, "ret_signed_p6m": 2.9285929147770355, "alpha_spy_p6m": 2.856905030238595, "alpha_c_p6m": 2.480164684766586, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.4484282300104496, "price_path": [-0.1516, -0.1243, -0.1243, -0.1068, -0.1009, -0.0639, -0.0814, -0.0814, -0.0326, -0.0287, -0.0385, -0.0404, 0.0025, 0.0122, 0.0863, 0.1156, 0.1429, 0.1078, 0.139, 0.1136, 0.0883, 0.0863, 0.0551, 0.0571, 0.1, 0.0961, 0.1585, 0.1487, 0.1702, 0.1643, 0.1546, 0.0512, 0.0493, 0.0629, 0.0473, 0.0493, 0.0512, 0.0376, 0.022, 0.0239, 0.0278, 0.0668, 0.0064, 0.0064, 0.0278, 0.0083, 0.022, 0.0005, 0.0415, 0.0493, 0.0844, 0.0785, 0.0785, 0.0434, 0.0259, -0.0034, -0.0443, -0.0209, 0.0122, 0.02, 0.0142, 0.0488, 0.0508, 0.0, 0.0176, 0.0254, 0.0371, 0.0684, 0.082, 0.125, 0.1875, 0.1992, 0.2832, 0.293, 0.3594, 0.3027, 0.3887, 0.3887, 0.3711, 0.4102, 0.3965, 0.3926, 0.3145, 0.3633, 0.3574, 0.4063, 0.5449, 0.5449, 0.5449, 0.5449, 0.5449, 0.5449, 0.6719, 0.6211, 0.6074, 0.6504, 0.7676, 0.8184, 0.8965, 0.8711, 0.8809, 0.8691, 0.9922, 1.0898, 1.0352, 1.1797, 1.2187, 1.1836, 1.4219, 1.2891, 1.2617, 1.3164, 1.2656, 1.3672, 1.418, 1.4102, 1.3906, 1.1875, 1.3672, 1.2266, 1.1953, 1.2305, 1.0352, 1.0313, 1.0273, 1.0469, 1.1265, 1.0718, 1.1031, 1.1812, 1.1578, 1.1187, 1.1265, 1.2555, 1.2125, 1.2946, 1.2086, 1.2321, 1.1656, 1.0718, 1.1539, 1.1578, 1.1382, 1.2672, 1.2829, 1.2985, 1.2985, 1.3415, 1.5018, 1.5448, 1.5448, 1.5448, 1.6464, 1.7207, 1.838, 1.9005, 1.9552, 1.9083, 1.9279, 1.8849, 1.9005, 1.9279, 1.9552, 1.9865, 1.9044, 1.8927, 1.9513, 1.9982, 1.8771, 2.1272, 2.2875, 2.3657, 2.5533, 2.2445, 2.5455, 2.5181, 2.2914, 2.2797, 2.4673, 2.4243, 2.3618, 2.4712, 2.44, 2.44, 2.44, 2.44, 2.4947, 2.7097, 2.7175, 2.9286]}
{"unit_id": "orig:1962300280823067129", "ticker_idx": 2, "ts": "2025-08-31T23:43:39+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain a positive view on SEC/SKH/MU", "tweet_type": "analysis", "timeline": "months", "summary": "CLSA reiterates positive view on Samsung, SK Hynix, and Micron; US curbs on China ops seen tightening memory supply mid-term.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250901_US curbs SEC/Hynix China operations - CLSA\n\n[Key Takeaways]\n\n(1) This U.S. Department of Commerce action will not cause production disruptions in the short term.\n\n(2) If disruptions do occur, memory supply will become even tighter.\n\n(3) In the mid to long term, domestic companies are expected to reduce their production share in China and accelerate fab construction in Korea.\n\n(4) Maintain a positive view on SEC/SKH/MU.\n\n[Contents]\n\n(1) Status of Korean memory companies’ China fabs as of 1H25\n\nSKH DRAM: 185K [36% of SKH DRAM / 10% of global DRAM]\nSKH NAND: 85K [37% of SKH NAND / 6% of global NAND]\nSEC NAND: 120K [30% of SEC NAND / 9% of global NAND]\n(2) For SKH’s DRAM fab, process upgrades to 1-a are underway, but still 50% remains on 1-z and 1-y processes. For NAND, mainly 144-layer, with review underway for upgrade to 192-layer.\n\n(3) For SEC’s China NAND fab, mainly based on 128-layer.\n\n(4) This U.S. action aims to restrict Korean companies from expanding production capacity and upgrading processes in China.\n\n(5) In the short term, there will be no production suspension, but the burden of securing export licenses each time U.S. equipment is purchased will increase, leading to greater uncertainty.\n\n(6) However, in the long term, fabs in China will face upgrade restrictions, falling behind fabs in Korea in terms of competitiveness, which could result in cost structure deterioration for domestic memory companies.\n\n(7) Accordingly, the need for Korean companies to accelerate construction of new domestic fabs will increase.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004453374548142985, "ret_signed_p1d": -0.004453374548142985, "alpha_spy_p1d": 0.0029569210077312613, "alpha_c_p1d": 0.008671450330633879, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.013124824878776864, "ret_p1w": 0.10461318165995759, "ret_signed_p1w": 0.10461318165995759, "alpha_spy_p1w": 0.09875323020566196, "alpha_c_p1w": 0.0831520329195703, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.021461148740387292, "ret_p1m": 0.40593227727445536, "ret_signed_p1m": 0.40593227727445536, "alpha_spy_p1m": 0.37031186433729335, "alpha_c_p1m": 0.28167738281751853, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.12425489445693683, "ret_p3m": 0.9360069202668313, "ret_signed_p3m": 0.9360069202668313, "alpha_spy_p3m": 0.8793998464723796, "alpha_c_p3m": 0.7382048557319516, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.1978020645348797, "ret_p6m": 2.5160132606775507, "ret_signed_p6m": 2.5160132606775507, "alpha_spy_p6m": 2.44432537613911, "alpha_c_p6m": 2.067585030667101, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.4484282300104496, "price_path": [-0.1332, -0.1077, -0.0887, -0.0686, -0.0418, -0.026, -0.0247, -0.0296, 0.006, 0.0102, 0.0226, 0.0226, 0.0376, 0.0248, 0.0738, 0.0682, 0.0577, 0.0473, 0.0347, 0.0148, 0.022, 0.0266, 0.0266, 0.0076, 0.0455, 0.0271, 0.0345, 0.0464, -0.0034, 0.0092, -0.0217, -0.0483, -0.0388, -0.0486, -0.0823, -0.0771, -0.0612, -0.0651, -0.0652, -0.0592, -0.0359, -0.0829, -0.1187, -0.0944, -0.0836, -0.086, -0.06, -0.001, 0.0396, 0.0734, 0.0442, 0.0528, 0.0156, 0.0381, 0.0255, -0.0151, -0.0271, -0.0112, -0.0218, -0.0211, -0.0106, 0.0251, 0.0, 0.0, -0.0045, -0.0024, 0.0437, 0.1039, 0.1046, 0.1364, 0.1764, 0.2652, 0.3211, 0.3257, 0.3345, 0.3443, 0.4191, 0.3674, 0.3832, 0.3983, 0.3588, 0.3178, 0.3215, 0.3772, 0.4059, 0.5305, 0.544, 0.5793, 0.6056, 0.5613, 0.6525, 0.6171, 0.5269, 0.6208, 0.5728, 0.6138, 0.7029, 0.7016, 0.7385, 0.7008, 0.6687, 0.738, 0.8415, 0.8506, 0.8658, 0.9055, 0.8835, 0.8814, 0.9733, 0.8332, 0.9969, 1.0039, 1.0004, 1.1297, 1.0272, 1.0591, 0.9923, 1.0753, 1.0343, 0.9212, 0.8995, 0.6931, 0.7435, 0.8828, 0.8878, 0.936, 0.936, 0.9883, 1.0218, 1.0136, 0.9688, 0.9057, 0.9945, 1.0761, 1.1223, 1.2173, 1.1731, 1.0275, 0.9969, 0.9549, 0.8962, 1.0898, 1.2358, 1.3255, 1.3229, 1.4104, 1.4104, 1.3945, 1.476, 1.4614, 1.4007, 1.4007, 1.6531, 1.6256, 1.8887, 1.8561, 1.7507, 1.9027, 1.9092, 1.8441, 1.8039, 1.8315, 2.0512, 2.0512, 2.0701, 2.2729, 2.3442, 2.3616, 2.2728, 2.4507, 2.6613, 2.6656, 2.4897, 2.6825, 2.528, 2.1913, 2.2206, 2.3199, 2.2257, 2.1395, 2.4515, 2.482, 2.4626, 2.4626, 2.3627, 2.5407, 2.5105, 2.6015, 2.5409, 2.516]}
{"unit_id": "quote:1951090591468757485", "ticker_idx": 0, "ts": "2025-08-01T01:20:21+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I remain bullish on Samsung's stock. I believe it has at least 20% upside from the current level.", "tweet_type": "prediction", "timeline": "quarters", "summary": "Long Samsung 20%+ upside; bearish on Micron and SK Hynix as Samsung's HBM3E pricing war pressures peers and HBM3E ASP may fall as early as Q4.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "I remain bullish on Samsung’s stock. I believe it has at least 20% upside from the current level.\n\nMicron’s sharp drop yesterday and SK Hynix’s plunge today are both driven by Samsung effectively declaring a game of chicken in the HBM3E market.\n\nHBM customers like Nvidia are using Samsung as leverage to push down HBM3E pricing — and that’s unavoidable.\n\nI’m starting to think that the ASP of HBM3E might begin to decline as early as Q4 this year.", "reply_text": null, "parent_text": null, "quoted_text": "What We Can Glean from Samsung Electronics' Earnings Release\n\nThe increase in DRAM's ASP (Average Selling Price) was lower than expected. The company announced a 30% increase in bit growth. Since Q1's bit growth was 600 million Gb, Q2's was nearly 800 million Gb. This suggests", "parent_author": null, "ret_m1d": 0.03628451801845611, "ret_signed_m1d": 0.03628451801845611, "alpha_spy_m1d": 0.019620828077359542, "alpha_c_m1d": 0.014228266800442535, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.02205625121801358, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01161109622749179, "ret_signed_p1d": 0.01161109622749179, "alpha_spy_p1d": -0.0035887179614699782, "alpha_c_p1d": -0.008694791299870719, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02030588752736251, "ret_p1w": 0.042090123474913144, "ret_signed_p1w": 0.042090123474913144, "alpha_spy_p1w": 0.017223541503019302, "alpha_c_p1w": 0.007663562763754239, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.034426560711158904, "ret_p1m": -0.01886781633450707, "ret_signed_p1m": -0.01886781633450707, "alpha_spy_p1m": -0.05639287429155182, "alpha_c_p1m": -0.039796134858980814, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.020928318524473744, "ret_p3m": 0.4651436595476073, "ret_signed_p3m": 0.4651436595476073, "alpha_spy_p3m": 0.3564519244954558, "alpha_c_p3m": 0.28056009055783004, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.18458356898977724, "ret_p6m": 1.2281755934741243, "ret_signed_p6m": 1.2281755934741243, "alpha_spy_p6m": 1.1075694978265855, "alpha_c_p6m": 1.0884890617390235, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.13968653173510082, "price_path": [-0.2167, -0.2124, -0.2124, -0.2095, -0.1691, -0.1792, -0.172, -0.1734, -0.1806, -0.1951, -0.1936, -0.1965, -0.2109, -0.2181, -0.2109, -0.2225, -0.1936, -0.1907, -0.1893, -0.1806, -0.1806, -0.1662, -0.1475, -0.1475, -0.1374, -0.146, -0.1359, -0.1417, -0.159, -0.1749, -0.1619, -0.1374, -0.146, -0.1417, -0.1633, -0.1273, -0.1157, -0.1316, -0.1176, -0.1321, -0.1263, -0.1176, -0.074, -0.0813, -0.1045, -0.1089, -0.1234, -0.1147, -0.0914, -0.0929, -0.0755, -0.061, -0.0319, -0.0261, -0.016, -0.0421, -0.0363, -0.0421, -0.0435, 0.0218, 0.0247, 0.0537, 0.0363, 0.0, 0.0116, 0.0145, -0.0015, 0.0232, 0.0421, 0.0305, 0.0319, 0.0435, 0.0392, 0.0392, 0.016, 0.016, 0.0232, 0.0247, 0.0363, 0.0377, 0.0203, 0.0247, 0.0102, 0.0116, -0.0189, 0.0029, 0.0131, 0.0174, 0.0087, 0.0174, 0.0377, 0.0537, 0.0653, 0.0943, 0.1103, 0.1524, 0.135, 0.1655, 0.1655, 0.2119, 0.2293, 0.2395, 0.2496, 0.209, 0.2275, 0.2231, 0.2538, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3762, 0.3602, 0.3354, 0.385, 0.4243, 0.4272, 0.4302, 0.4214, 0.4374, 0.4068, 0.4404, 0.487, 0.4506, 0.4651, 0.5176, 0.5672, 0.6197, 0.5293, 0.4666, 0.4462, 0.4272, 0.4666, 0.5089, 0.503, 0.4987, 0.417, 0.4666, 0.4258, 0.4068, 0.4666, 0.382, 0.4097, 0.4476, 0.4987, 0.5089, 0.4651, 0.4695, 0.5074, 0.5235, 0.5322, 0.5803, 0.5964, 0.5803, 0.5745, 0.5643, 0.5876, 0.5278, 0.4987, 0.573, 0.5687, 0.5497, 0.6109, 0.6255, 0.6197, 0.6197, 0.7057, 0.7506, 0.7565, 0.7565, 0.7565, 0.8824, 1.0231, 1.0348, 1.0656, 1.0333, 1.0363, 1.0333, 1.0158, 1.0553, 1.1081, 1.1813, 1.1872, 1.1271, 1.1901, 1.2311, 1.2282, 1.2282]}
{"unit_id": "quote:1951090591468757485", "ticker_idx": 1, "ts": "2025-08-01T01:20:21+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's sharp drop yesterday and SK Hynix's plunge today are both driven by Samsung effectively declaring a game of chicken in the HBM3E market.", "tweet_type": "prediction", "timeline": "quarters", "summary": "Long Samsung 20%+ upside; bearish on Micron and SK Hynix as Samsung's HBM3E pricing war pressures peers and HBM3E ASP may fall as early as Q4.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I remain bullish on Samsung’s stock. I believe it has at least 20% upside from the current level.\n\nMicron’s sharp drop yesterday and SK Hynix’s plunge today are both driven by Samsung effectively declaring a game of chicken in the HBM3E market.\n\nHBM customers like Nvidia are using Samsung as leverage to push down HBM3E pricing — and that’s unavoidable.\n\nI’m starting to think that the ASP of HBM3E might begin to decline as early as Q4 this year.", "reply_text": null, "parent_text": null, "quoted_text": "What We Can Glean from Samsung Electronics' Earnings Release\n\nThe increase in DRAM's ASP (Average Selling Price) was lower than expected. The company announced a 30% increase in bit growth. Since Q1's bit growth was 600 million Gb, Q2's was nearly 800 million Gb. This suggests", "parent_author": null, "ret_m1d": 0.040617723692302654, "ret_signed_m1d": -0.040617723692302654, "alpha_spy_m1d": -0.023954033751206083, "alpha_c_m1d": -0.02360774095477547, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.017009982737527185, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027555283000971764, "ret_signed_p1d": -0.027555283000971764, "alpha_spy_p1d": -0.012355468812009995, "alpha_c_p1d": -0.005579590184152883, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02197569281681888, "ret_p1w": 0.13358115305450302, "ret_signed_p1w": -0.13358115305450302, "alpha_spy_p1w": -0.10871457108260918, "alpha_c_p1w": -0.09984289405876234, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.03373825899574068, "ret_p1m": 0.1347254014995649, "ret_signed_p1m": -0.1347254014995649, "alpha_spy_p1m": -0.09720034354252016, "alpha_c_p1m": -0.11239749472640526, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.022327906773159656, "ret_p3m": 1.1622036824882853, "ret_signed_p3m": -1.1622036824882853, "alpha_spy_p3m": -1.0535119474361339, "alpha_c_p3m": -0.8640526817780356, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.2981510007102497, "ret_p6m": 2.71368160389145, "ret_signed_p6m": -2.71368160389145, "alpha_spy_p6m": -2.593075508243911, "alpha_c_p6m": -2.30476820820392, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.40891339568752993, "price_path": [-0.2331, -0.213, -0.1889, -0.1821, -0.1209, -0.0767, -0.092, -0.0908, -0.0665, -0.0603, -0.0655, -0.0871, -0.0967, -0.1106, -0.1106, -0.0819, -0.0838, -0.0779, -0.1002, -0.0648, -0.026, -0.0165, 0.0125, 0.0341, 0.0569, 0.0873, 0.1053, 0.1067, 0.1012, 0.1416, 0.1463, 0.1604, 0.1604, 0.1774, 0.1629, 0.2184, 0.2122, 0.2002, 0.1884, 0.174, 0.1516, 0.1597, 0.1649, 0.1649, 0.1434, 0.1863, 0.1655, 0.1738, 0.1874, 0.1309, 0.1452, 0.1101, 0.0799, 0.0907, 0.0796, 0.0414, 0.0472, 0.0653, 0.0608, 0.0607, 0.0675, 0.094, 0.0406, 0.0, 0.0276, 0.0399, 0.0372, 0.0666, 0.1336, 0.1796, 0.2181, 0.1849, 0.1946, 0.1525, 0.178, 0.1637, 0.1176, 0.104, 0.122, 0.11, 0.1108, 0.1227, 0.1632, 0.1347, 0.1347, 0.1297, 0.132, 0.1843, 0.2526, 0.2534, 0.2895, 0.3349, 0.4356, 0.4991, 0.5043, 0.5143, 0.5255, 0.6103, 0.5516, 0.5696, 0.5867, 0.5419, 0.4953, 0.4995, 0.5627, 0.5953, 0.7367, 0.752, 0.792, 0.8219, 0.7716, 0.8751, 0.835, 0.7326, 0.8392, 0.7847, 0.8312, 0.9323, 0.9308, 0.9727, 0.93, 0.8935, 0.9722, 1.0896, 1.0999, 1.1172, 1.1622, 1.1372, 1.1349, 1.2392, 1.0802, 1.2659, 1.2738, 1.2699, 1.4167, 1.3004, 1.3365, 1.2607, 1.3549, 1.3084, 1.18, 1.1554, 0.9212, 0.9785, 1.1364, 1.1422, 1.1968, 1.1968, 1.2562, 1.2942, 1.2849, 1.234, 1.1624, 1.2632, 1.3558, 1.4083, 1.516, 1.4659, 1.3006, 1.2659, 1.2183, 1.1516, 1.3713, 1.5371, 1.6389, 1.6358, 1.7351, 1.7351, 1.7171, 1.8096, 1.793, 1.7241, 1.7241, 2.0105, 1.9793, 2.2779, 2.2408, 2.1213, 2.2937, 2.3012, 2.2273, 2.1817, 2.213, 2.4623, 2.4623, 2.4838, 2.7139, 2.7947, 2.8145, 2.7137]}
{"unit_id": "quote:1951090591468757485", "ticker_idx": 2, "ts": "2025-08-01T01:20:21+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's plunge today are both driven by Samsung effectively declaring a game of chicken in the HBM3E market.", "tweet_type": "prediction", "timeline": "quarters", "summary": "Long Samsung 20%+ upside; bearish on Micron and SK Hynix as Samsung's HBM3E pricing war pressures peers and HBM3E ASP may fall as early as Q4.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I remain bullish on Samsung’s stock. I believe it has at least 20% upside from the current level.\n\nMicron’s sharp drop yesterday and SK Hynix’s plunge today are both driven by Samsung effectively declaring a game of chicken in the HBM3E market.\n\nHBM customers like Nvidia are using Samsung as leverage to push down HBM3E pricing — and that’s unavoidable.\n\nI’m starting to think that the ASP of HBM3E might begin to decline as early as Q4 this year.", "reply_text": null, "parent_text": null, "quoted_text": "What We Can Glean from Samsung Electronics' Earnings Release\n\nThe increase in DRAM's ASP (Average Selling Price) was lower than expected. The company announced a 30% increase in bit growth. Since Q1's bit growth was 600 million Gb, Q2's was nearly 800 million Gb. This suggests", "parent_author": null, "ret_m1d": 0.060077465046982415, "ret_signed_m1d": -0.060077465046982415, "alpha_spy_m1d": -0.04341377510588584, "alpha_c_m1d": -0.04306748230945523, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.017009982737527185, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.015199814188961769, "alpha_c_p1d": 0.02197569281681888, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02197569281681888, "ret_p1w": -0.005813983542341639, "ret_signed_p1w": 0.005813983542341639, "alpha_spy_p1w": 0.03068056551423548, "alpha_c_p1w": 0.039552242538082316, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.03373825899574068, "ret_p1m": -0.006318687984950944, "ret_signed_p1m": 0.006318687984950944, "alpha_spy_p1m": 0.04384374594199569, "alpha_c_p1m": 0.0286465947581106, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.022327906773159656, "ret_p3m": 1.1659146373836324, "ret_signed_p3m": -1.1659146373836324, "alpha_spy_p3m": -1.057222902331481, "alpha_c_p3m": -0.8677636366733827, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.2981510007102497, "ret_p6m": 1.8588797361500777, "ret_signed_p6m": -1.8588797361500777, "alpha_spy_p6m": -1.738273640502539, "alpha_c_p6m": -1.4499663404625478, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.40891339568752993, "price_path": [-0.2804, -0.2618, -0.2637, -0.2645, -0.2455, -0.232, -0.203, -0.2243, -0.2088, -0.2285, -0.2185, -0.2243, -0.2382, -0.2262, -0.2146, -0.2165, -0.1953, -0.1783, -0.2074, -0.1957, -0.1957, -0.157, -0.1298, -0.1298, -0.1124, -0.1066, -0.0698, -0.0872, -0.0872, -0.0388, -0.0349, -0.0446, -0.0465, -0.0039, 0.0058, 0.0795, 0.1085, 0.1357, 0.1008, 0.1318, 0.1066, 0.0814, 0.0795, 0.0484, 0.0504, 0.093, 0.0891, 0.1512, 0.1415, 0.1628, 0.157, 0.1473, 0.0446, 0.0426, 0.0562, 0.0407, 0.0426, 0.0446, 0.031, 0.0155, 0.0174, 0.0213, 0.0601, 0.0, 0.0, 0.0213, 0.0019, 0.0155, -0.0058, 0.0349, 0.0426, 0.0775, 0.0717, 0.0717, 0.0368, 0.0194, -0.0097, -0.0504, -0.0271, 0.0058, 0.0136, 0.0078, 0.0422, 0.0441, -0.0063, 0.0111, 0.0189, 0.0306, 0.0616, 0.0752, 0.1179, 0.18, 0.1916, 0.2751, 0.2848, 0.3508, 0.2945, 0.3799, 0.3799, 0.3624, 0.4012, 0.3877, 0.3838, 0.3061, 0.3547, 0.3488, 0.3974, 0.5352, 0.5352, 0.5352, 0.5352, 0.5352, 0.5352, 0.6613, 0.6109, 0.5973, 0.64, 0.7564, 0.8069, 0.8845, 0.8593, 0.869, 0.8573, 0.9796, 1.0766, 1.0223, 1.1659, 1.2047, 1.1698, 1.4066, 1.2746, 1.2474, 1.3018, 1.2513, 1.3522, 1.4027, 1.3949, 1.3755, 1.1737, 1.3522, 1.2125, 1.1814, 1.2164, 1.0223, 1.0184, 1.0145, 1.0339, 1.1131, 1.0587, 1.0898, 1.1675, 1.1442, 1.1053, 1.1131, 1.2413, 1.1985, 1.2801, 1.1947, 1.218, 1.1519, 1.0587, 1.1403, 1.1442, 1.1247, 1.2529, 1.2685, 1.284, 1.284, 1.3267, 1.486, 1.5287, 1.5287, 1.5287, 1.6297, 1.7035, 1.82, 1.8822, 1.9366, 1.89, 1.9094, 1.8666, 1.8822, 1.9094, 1.9366, 1.9676, 1.8861, 1.8744, 1.9327, 1.9793, 1.8589]}
{"unit_id": "reply:1967037558833197286", "ticker_idx": 0, "ts": "2025-09-14T01:27:54+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buying more AMD doesn't seem like a bad idea either", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Endorses buying more AMD shares.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@treasureh8nter I’m using an iPhone 15 Pro, haha. Buying more AMD doesn’t seem like a bad idea either.", "reply_text": null, "parent_text": "@Jukanlosreve what are you using now?\n\nis the iphone Air an actual upgrade?! 😅\n\nhow about buying $AMD shares instead 🤣", "quoted_text": null, "parent_author": "treasureh8nter", "ret_m1d": -0.016070962267541566, "ret_signed_m1d": -0.016070962267541566, "alpha_spy_m1d": -0.010775297344054913, "alpha_c_m1d": -0.006792724254760185, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004343490520823057, "ret_signed_p1d": -0.004343490520823057, "alpha_spy_p1d": -0.002966697735541368, "alpha_c_p1d": -0.004473646364008266, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": -0.00850093289181697, "ret_signed_p1w": -0.00850093289181697, "alpha_spy_p1w": -0.020270774265879088, "alpha_c_p1w": -0.0573337907643271, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.3532513736791765, "ret_signed_p1m": 0.3532513736791765, "alpha_spy_p1m": 0.34847613884999773, "alpha_c_p1m": 0.26753338631107004, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.3739760960792411, "ret_signed_p3m": 0.3739760960792411, "alpha_spy_p3m": 0.32832586440496025, "alpha_c_p3m": 0.16653399225852916, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 0.2610448691328793, "ret_signed_p6m": 0.2610448691328793, "alpha_spy_p6m": 0.23055053411876525, "alpha_c_p6m": -0.03649816526224736, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.2133, -0.2133, -0.2043, -0.196, -0.141, -0.1102, -0.1085, -0.1077, -0.1195, -0.1554, -0.1405, -0.1443, -0.1443, -0.1636, -0.1448, -0.1412, -0.1055, -0.0915, -0.0926, -0.0344, -0.0067, -0.0047, -0.0259, -0.0258, -0.04, -0.0156, 0.006, 0.0329, 0.0776, 0.101, 0.1139, 0.094, 0.0654, 0.0969, 0.0816, 0.0122, 0.0697, 0.072, 0.069, 0.0856, 0.1443, 0.1228, 0.1015, 0.093, 0.0334, 0.0251, 0.0158, 0.041, 0.0137, 0.0339, 0.037, 0.046, 0.0091, 0.0091, 0.0072, 0.006, 0.0039, -0.0622, -0.0605, -0.0331, -0.0101, -0.0341, -0.0161, 0.0, -0.0043, -0.0124, -0.0201, -0.0234, -0.0085, -0.0016, -0.0017, 0.0007, -0.0105, 0.0012, 0.0039, 0.0177, 0.0532, 0.0218, 0.264, 0.3124, 0.4617, 0.4451, 0.3335, 0.3429, 0.3533, 0.4805, 0.4554, 0.4463, 0.4927, 0.477, 0.4286, 0.4581, 0.5694, 0.6113, 0.601, 0.6402, 0.5813, 0.5892, 0.6111, 0.5516, 0.5905, 0.4749, 0.4491, 0.5139, 0.4738, 0.6064, 0.5386, 0.5315, 0.4924, 0.429, 0.3871, 0.2784, 0.2645, 0.3344, 0.279, 0.3294, 0.3294, 0.3498, 0.3636, 0.3356, 0.3502, 0.3402, 0.3525, 0.372, 0.3752, 0.3739, 0.374, 0.3079, 0.288, 0.2979, 0.2293, 0.2476, 0.3243, 0.3338, 0.3335, 0.3343, 0.3343, 0.334, 0.3379, 0.3362, 0.3289, 0.3289, 0.3866, 0.3718, 0.33, 0.3032, 0.27, 0.2607, 0.2887, 0.3711, 0.3874, 0.4142, 0.4385, 0.4385, 0.4391, 0.55, 0.5744, 0.6113, 0.5594, 0.5638, 0.5683, 0.5648, 0.4689, 0.5281, 0.5023, 0.2422, 0.1945, 0.2934, 0.3403, 0.3252, 0.3253, 0.2779, 0.2864, 0.2864, 0.2601, 0.2417, 0.2619, 0.2419, 0.2199, 0.3269, 0.3084, 0.2638, 0.2423, 0.2324, 0.1848, 0.2538, 0.2376, 0.194, 0.2576, 0.261]}
{"unit_id": "thread:1949729428533309684", "ticker_idx": 0, "ts": "2025-07-28T07:11:34+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's AI chip revenue next year will be $15.1 billion, a significant increase from the previous estimate of $9.6 billion", "tweet_type": "news_reaction", "timeline": "months", "summary": "AMD MI350 price hike drives higher AI chip revenue ($15.1B est.); Samsung benefits via HBM pricing power and expanding AMD partnership.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD Raises Price of New AI Chips; Samsung Anticipates HBM Benefits\n\nAMD, a U.S. fabless semiconductor company, is significantly raising the price of its new artificial intelligence (AI) chip. Samsung Electronics, which supplies High Bandwidth Memory (HBM) for this product, is also expected to benefit directly and indirectly.\n\nAs AMD's AI chip profitability increases, its room for component unit price increases will grow, allowing Samsung Electronics to sell HBM at a higher price.\n\nAccording to industry sources on the 28th, analysis suggests that the average selling price of AMD's recently launched new AI chip, the 'MI350,' has increased by $10,000 (KRW 14 million) to $25,000 (KRW 35 million) from previous estimates.\n\nExperts believe that the MI350's performance rivals that of Nvidia's latest AI chip, the 'Blackwell B200,' leading to a higher price for the MI350 than initially anticipated by the market. Although still 30% cheaper than the Blackwell B200, the price has risen significantly.\n\nThe MI350 is primarily used in AI data centers, with AI computing performance improved by up to 4 times and inference performance by up to 35 times. As this chip incorporates eight 12-layer 5th-generation HBM3E units, it could be good news for memory companies.\n\nAs AMD's AI chip prices rise, the demand for AI chips in the AI market is expected to grow further.\n\nGlobal investment bank HSBC commented, \"AMD's MI350 is now capable of competing effectively with Nvidia's Blackwell B200 in the market,\" adding, \"AMD's AI chip revenue next year will be $15.1 billion, a significant increase from the previous estimate of $9.6 billion.\"\n\nAs the price of AMD's MI350 increases, Samsung Electronics, which supplies HBM for this chip, can sequentially increase its profitability.\n\nThis is because as the profit margin from AMD's MI350 expands, the acceptable range for increasing component supply prices, including HBM, can widen. Conversely, if AI chip prices decrease, pressure for component supply price reductions intensifies.\n\nFurthermore, with the expansion of demand for AMD's AI chips in the AI market, Samsung Electronics' HBM price negotiation power is also likely to increase. As HBM supply becomes tighter, memory suppliers gain greater negotiation leverage.\n\nPreviously, AMD officially confirmed the adoption of Samsung Electronics' HBM3E in the MI350 series in June. As Samsung Electronics has not yet passed Nvidia's HBM3E quality verification (qual-test), expanding HBM cooperation with AMD is considered more important than ever.\n\nIn addition, attention is focused on whether Samsung Electronics will succeed in supplying 6th-generation HBM4 for AMD's next-generation chip 'MI400,' scheduled for release next year. Unlike its competitors, there has been no news yet of Samsung Electronics' HBM4 sample supply. Samsung Electronics aims for HBM4 mass production in the second half of this year.\n\nAn industry official stated, \"Samsung's negotiation power can increase at a time when HBM demand is rapidly growing,\" adding, \"We need to observe more closely whether they will cooperate with AMD for HBM4 as well.\"\n\n---\n\nhttps://t.co/e2DPQCZiLF", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04140275417358541, "ret_signed_m1d": -0.04140275417358541, "alpha_spy_m1d": -0.04165391738589619, "alpha_c_m1d": -0.027950823921969348, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021766663017303678, "ret_signed_p1d": 0.021766663017303678, "alpha_spy_p1d": 0.024404408257789778, "alpha_c_p1d": 0.01665368363215336, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": 0.01796611223882083, "ret_signed_p1w": 0.01796611223882083, "alpha_spy_p1w": 0.027025189517943415, "alpha_c_p1w": 0.022152658249594892, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": -0.04053903257205749, "ret_signed_p1m": -0.04053903257205749, "alpha_spy_p1m": -0.05344441497360497, "alpha_c_p1m": -0.05886369352675136, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.3531613528604123, "ret_signed_p3m": 0.3531613528604123, "alpha_spy_p3m": 0.2955697925628318, "alpha_c_p3m": 0.16994877941884834, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 0.33548308924484727, "ret_signed_p6m": 0.33548308924484727, "alpha_spy_p6m": 0.2655766071948993, "alpha_c_p6m": -0.008344289713465125, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.4394, -0.4435, -0.4311, -0.4208, -0.4321, -0.4221, -0.4144, -0.4078, -0.3774, -0.3524, -0.3221, -0.3378, -0.3253, -0.3393, -0.3464, -0.3547, -0.3625, -0.3648, -0.3648, -0.3403, -0.3501, -0.3491, -0.3624, -0.3399, -0.3245, -0.3172, -0.3338, -0.3309, -0.299, -0.2903, -0.3024, -0.3176, -0.3311, -0.2722, -0.2681, -0.2699, -0.2699, -0.2615, -0.2538, -0.2029, -0.1742, -0.1726, -0.1719, -0.1829, -0.2162, -0.2023, -0.2059, -0.2059, -0.2238, -0.2064, -0.203, -0.1699, -0.1569, -0.1579, -0.1039, -0.0782, -0.0763, -0.096, -0.0959, -0.1091, -0.0864, -0.0665, -0.0414, 0.0, 0.0218, 0.0337, 0.0153, -0.0113, 0.018, 0.0037, -0.0607, -0.0073, -0.0052, -0.0079, 0.0074, 0.062, 0.042, 0.0222, 0.0143, -0.0409, -0.0487, -0.0573, -0.034, -0.0593, -0.0405, -0.0376, -0.0293, -0.0635, -0.0635, -0.0653, -0.0664, -0.0684, -0.1297, -0.1281, -0.1027, -0.0813, -0.1036, -0.0869, -0.072, -0.076, -0.0835, -0.0906, -0.0937, -0.0799, -0.0735, -0.0736, -0.0713, -0.0818, -0.0708, -0.0684, -0.0556, -0.0226, -0.0518, 0.173, 0.218, 0.3564, 0.3411, 0.2375, 0.2462, 0.2558, 0.3739, 0.3507, 0.3422, 0.3852, 0.3707, 0.3258, 0.3532, 0.4564, 0.4953, 0.4857, 0.5221, 0.4675, 0.4748, 0.4952, 0.4399, 0.476, 0.3688, 0.3448, 0.4049, 0.3677, 0.4908, 0.4278, 0.4212, 0.385, 0.3261, 0.2873, 0.1863, 0.1734, 0.2383, 0.187, 0.2337, 0.2337, 0.2526, 0.2655, 0.2394, 0.253, 0.2437, 0.2552, 0.2732, 0.2762, 0.275, 0.2751, 0.2138, 0.1953, 0.2045, 0.1408, 0.1578, 0.229, 0.2378, 0.2375, 0.2383, 0.2383, 0.238, 0.2416, 0.24, 0.2332, 0.2332, 0.2868, 0.2731, 0.2343, 0.2094, 0.1786, 0.1699, 0.196, 0.2724, 0.2876, 0.3124, 0.335, 0.335, 0.3355]}
{"unit_id": "thread:1949729428533309684", "ticker_idx": 1, "ts": "2025-07-28T07:11:34+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "As the price of AMD's MI350 increases, Samsung Electronics, which supplies HBM for this chip, can sequentially increase its profitability", "tweet_type": "news_reaction", "timeline": "months", "summary": "AMD MI350 price hike drives higher AI chip revenue ($15.1B est.); Samsung benefits via HBM pricing power and expanding AMD partnership.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: AMD Raises Price of New AI Chips; Samsung Anticipates HBM Benefits\n\nAMD, a U.S. fabless semiconductor company, is significantly raising the price of its new artificial intelligence (AI) chip. Samsung Electronics, which supplies High Bandwidth Memory (HBM) for this product, is also expected to benefit directly and indirectly.\n\nAs AMD's AI chip profitability increases, its room for component unit price increases will grow, allowing Samsung Electronics to sell HBM at a higher price.\n\nAccording to industry sources on the 28th, analysis suggests that the average selling price of AMD's recently launched new AI chip, the 'MI350,' has increased by $10,000 (KRW 14 million) to $25,000 (KRW 35 million) from previous estimates.\n\nExperts believe that the MI350's performance rivals that of Nvidia's latest AI chip, the 'Blackwell B200,' leading to a higher price for the MI350 than initially anticipated by the market. Although still 30% cheaper than the Blackwell B200, the price has risen significantly.\n\nThe MI350 is primarily used in AI data centers, with AI computing performance improved by up to 4 times and inference performance by up to 35 times. As this chip incorporates eight 12-layer 5th-generation HBM3E units, it could be good news for memory companies.\n\nAs AMD's AI chip prices rise, the demand for AI chips in the AI market is expected to grow further.\n\nGlobal investment bank HSBC commented, \"AMD's MI350 is now capable of competing effectively with Nvidia's Blackwell B200 in the market,\" adding, \"AMD's AI chip revenue next year will be $15.1 billion, a significant increase from the previous estimate of $9.6 billion.\"\n\nAs the price of AMD's MI350 increases, Samsung Electronics, which supplies HBM for this chip, can sequentially increase its profitability.\n\nThis is because as the profit margin from AMD's MI350 expands, the acceptable range for increasing component supply prices, including HBM, can widen. Conversely, if AI chip prices decrease, pressure for component supply price reductions intensifies.\n\nFurthermore, with the expansion of demand for AMD's AI chips in the AI market, Samsung Electronics' HBM price negotiation power is also likely to increase. As HBM supply becomes tighter, memory suppliers gain greater negotiation leverage.\n\nPreviously, AMD officially confirmed the adoption of Samsung Electronics' HBM3E in the MI350 series in June. As Samsung Electronics has not yet passed Nvidia's HBM3E quality verification (qual-test), expanding HBM cooperation with AMD is considered more important than ever.\n\nIn addition, attention is focused on whether Samsung Electronics will succeed in supplying 6th-generation HBM4 for AMD's next-generation chip 'MI400,' scheduled for release next year. Unlike its competitors, there has been no news yet of Samsung Electronics' HBM4 sample supply. Samsung Electronics aims for HBM4 mass production in the second half of this year.\n\nAn industry official stated, \"Samsung's negotiation power can increase at a time when HBM demand is rapidly growing,\" adding, \"We need to observe more closely whether they will cooperate with AMD for HBM4 as well.\"\n\n---\n\nhttps://t.co/e2DPQCZiLF", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06392037833579933, "ret_signed_m1d": -0.06392037833579933, "alpha_spy_m1d": -0.06417154154811011, "alpha_c_m1d": -0.056005973190194513, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.00791440514560482, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0028408408529752016, "ret_signed_p1d": 0.0028408408529752016, "alpha_spy_p1d": 0.005478586093461302, "alpha_c_p1d": 0.0028030150784925656, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 3.782577448263602e-05, "ret_p1w": -0.009943167469083525, "ret_signed_p1w": -0.009943167469083525, "alpha_spy_p1w": -0.0008840901899609399, "alpha_c_p1w": -0.0031650439477406422, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0067781235213428825, "ret_p1m": -0.0014204204264876008, "ret_signed_p1m": -0.0014204204264876008, "alpha_spy_p1m": -0.014325802828035084, "alpha_c_p1m": 0.002555579003573172, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": -0.003975999430060773, "ret_p3m": 0.37685443154336795, "ret_signed_p3m": 0.37685443154336795, "alpha_spy_p3m": 0.31926287124578745, "alpha_c_p3m": 0.28074134315242305, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.0961130883909449, "ret_p6m": 1.0817728912172333, "ret_signed_p6m": 1.0817728912172333, "alpha_spy_p6m": 1.0118664091672853, "alpha_c_p6m": 1.0045408978522918, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.07723199336494146, "price_path": [-0.2165, -0.2165, -0.2334, -0.2334, -0.2334, -0.2292, -0.2292, -0.2263, -0.1868, -0.1967, -0.1896, -0.191, -0.1981, -0.2122, -0.2108, -0.2136, -0.2277, -0.2348, -0.2277, -0.239, -0.2108, -0.208, -0.2066, -0.1981, -0.1981, -0.184, -0.1656, -0.1656, -0.1557, -0.1642, -0.1543, -0.16, -0.1769, -0.1925, -0.1797, -0.1557, -0.1642, -0.16, -0.1812, -0.1459, -0.1346, -0.1501, -0.1364, -0.1506, -0.1449, -0.1364, -0.0937, -0.1009, -0.1236, -0.1278, -0.142, -0.1335, -0.1108, -0.1122, -0.0952, -0.081, -0.0526, -0.0469, -0.0369, -0.0625, -0.0568, -0.0625, -0.0639, 0.0, 0.0028, 0.0313, 0.0142, -0.0213, -0.0099, -0.0071, -0.0227, 0.0014, 0.0199, 0.0085, 0.0099, 0.0213, 0.017, 0.017, -0.0057, -0.0057, 0.0014, 0.0028, 0.0142, 0.0156, -0.0014, 0.0028, -0.0114, -0.0099, -0.0398, -0.0185, -0.0085, -0.0043, -0.0128, -0.0043, 0.0156, 0.0313, 0.0426, 0.071, 0.0866, 0.1278, 0.1108, 0.1406, 0.1406, 0.1861, 0.2031, 0.2131, 0.223, 0.1832, 0.2014, 0.1971, 0.227, 0.2805, 0.2805, 0.2805, 0.2805, 0.2805, 0.2805, 0.3469, 0.3312, 0.3069, 0.3555, 0.394, 0.3968, 0.3997, 0.3911, 0.4068, 0.3769, 0.4097, 0.4553, 0.4197, 0.4339, 0.4853, 0.5338, 0.5852, 0.4967, 0.4354, 0.4154, 0.3968, 0.4354, 0.4767, 0.471, 0.4667, 0.3868, 0.4354, 0.3954, 0.3769, 0.4354, 0.3526, 0.3797, 0.4168, 0.4667, 0.4767, 0.4339, 0.4382, 0.4753, 0.491, 0.4996, 0.5466, 0.5623, 0.5466, 0.5409, 0.5309, 0.5538, 0.4953, 0.4667, 0.5395, 0.5352, 0.5167, 0.5766, 0.5909, 0.5852, 0.5852, 0.6693, 0.7133, 0.719, 0.719, 0.719, 0.8423, 0.98, 0.9914, 1.0216, 0.99, 0.9929, 0.99, 0.9728, 1.0115, 1.0631, 1.1348, 1.1406, 1.0818]}
{"unit_id": "quote:1973943159647560139", "ticker_idx": 0, "ts": "2025-10-03T02:48:18+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Upgrade to Buy, Target Price USD 175: Dell's stock has risen 16% over the past month, driven by accelerating AI infrastructure trends.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses GF Securities upgrade of DELL to Buy with $175 PT on GB300 demand acceleration and storage recovery.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Telecom Report\n\n☄️ Dell (DELL US): Significant Growth in GB300 Demand by 2026, Plus Strengthening Storage Business\n\n☀️ Upgrade to Buy, Target Price USD 175: Dell’s stock has risen 16% over the past month, driven by accelerating AI infrastructure trends. On the AI server side, we expect GB300 momentum to pick up speed. We estimate FY26/FY27 NVL72/144 rack shipments will reach 5.5k/10k units, supported by demand from xAI and CoreWeave. Meanwhile, we expect strong HDD+SSD demand to underpin Dell’s storage business and support margin performance. In the near term, we expect Dell’s stock to react positively ahead of the October 7 analyst meeting, as long-term upside remains. Considering increased NVL72/144 racks, reduced HGX demand, and improved storage demand, we raise our FY26/FY27 EPS estimates by 3%/10% respectively, and adjust our target price to USD 175. This is based on 15x FY27E P/E, reflecting improved visibility in AI servers and easing margin concerns. Given attractive valuation and better risk/reward, we upgrade Dell to Buy.\n\n☀️ Notable Increase in GB300 Demand by 2026: We now estimate Dell’s FY26 GB200/300 NVL72 shipments at 5.5k units, above our previous forecast of 4.5k, while AI HGX server shipments are expected to decline to 26k units as the market shifts toward NVL72/144 configurations. For FY27, we forecast NVL72/144 rack shipments of 10k units, supported by xAI’s Colossus 2 and CoreWeave’s agreements with OpenAI/Meta. Dell’s strong positioning at the L11/L12 tier will also help. We project FY26/FY27 AI server revenues at USD 24.5bn/39.0bn, both above consensus. With increased order visibility for GB300 among second-tier cloud service providers, Dell’s AI server pipeline is expected to continue growing, becoming the primary growth driver of the ISG business.\n\n☀️ Storage Business Recovery Supporting Margins: In our report “A Stronger NAND Cycle in 2026”, we highlighted that eSSD demand for 2026 was revised upward, driven by AI data centers and HDD replacement. Dell previously guided for flat storage business in FY26, but we now expect slight upside and a stronger rebound in FY27, benefiting from rising storage demand tied to AI infrastructure. With Dell’s IP and higher proportion of all-flash product mix boosting profitability, we expect momentum in storage improvement to support overall margin recovery, alleviating key investor concerns. We forecast storage revenue growth of +2%/+6% YoY in FY26/FY27.", "reply_text": null, "parent_text": null, "quoted_text": "《广发海外电子通信》\n☄️ 戴尔 (DELL US) ：2026年GB300需求显著增长，加上存储业务增強\n\n☀️升评至买入，目标价175美元：DELL股价过去一个月上涨16%，受AI基础设施趋势加速推动。在AI服务器方面，我们预计GB300动能开始加速动，预估FY26/FY27", "parent_author": null, "ret_m1d": 0.04710809391692883, "ret_signed_m1d": 0.04710809391692883, "alpha_spy_m1d": 0.04709332633662888, "alpha_c_m1d": 0.042050488091825766, "bench_spy_m1d": 1.4767580299945848e-05, "bench_c_m1d": 0.005057605825103062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03566850424169621, "ret_signed_p1d": 0.03566850424169621, "alpha_spy_p1d": 0.0320822753313601, "alpha_c_p1d": 0.025553292591490084, "bench_spy_p1d": 0.003586228910336109, "bench_c_p1d": 0.010115211650206124, "ret_p1w": 0.06984520741504818, "ret_signed_p1w": 0.06984520741504818, "alpha_spy_p1w": 0.09403798946207331, "alpha_c_p1w": 0.09207745638046705, "bench_spy_p1w": -0.02419278204702513, "bench_c_p1w": -0.022232248965418866, "ret_p1m": 0.14168306379918727, "ret_signed_p1m": 0.14168306379918727, "alpha_spy_p1m": 0.1205686343139245, "alpha_c_p1m": 0.08130789426201912, "bench_spy_p1m": 0.021114429485262765, "bench_c_p1m": 0.06037516953716815, "ret_p3m": -0.1023979641612579, "ret_signed_p3m": -0.1023979641612579, "alpha_spy_p3m": -0.12440147535985335, "alpha_c_p3m": -0.11524124770564914, "bench_spy_p3m": 0.02200351119859545, "bench_c_p3m": 0.012843283544391237, "ret_p6m": 0.17926408573060315, "ret_signed_p6m": 0.17926408573060315, "alpha_spy_p6m": 0.22953476355336655, "alpha_c_p6m": 0.28117361782790273, "bench_spy_p6m": -0.0502706778227634, "bench_c_p6m": -0.10190953209729958, "price_path": [-0.1198, -0.1039, -0.0949, -0.1025, -0.1107, -0.1106, -0.1256, -0.1234, -0.0713, -0.0874, -0.1166, -0.0961, -0.088, -0.0676, -0.0489, -0.0514, -0.0512, -0.0572, -0.0954, -0.0747, -0.0729, -0.0895, -0.0484, -0.0222, -0.0172, 0.0064, -0.0114, -0.0134, -0.0175, -0.0185, -0.0394, -0.0871, -0.0917, -0.0703, -0.0691, -0.0693, -0.0585, -0.0475, -0.1321, -0.1321, -0.1405, -0.1188, -0.1, -0.113, -0.126, -0.1382, -0.1157, -0.1092, -0.1116, -0.099, -0.0928, -0.0739, -0.0613, -0.0625, -0.0359, -0.0455, -0.0615, -0.0695, -0.0709, -0.0486, 0.0073, 0.0635, 0.0471, 0.0, 0.0357, 0.072, 0.169, 0.1081, 0.0698, 0.09, 0.0571, 0.0921, 0.0751, 0.0629, 0.0507, 0.0655, 0.0705, 0.0998, 0.1312, 0.1565, 0.1757, 0.1666, 0.1481, 0.1552, 0.1417, 0.1027, 0.0868, 0.0637, 0.0461, 0.0175, -0.0106, 0.0033, -0.0449, -0.0462, -0.1266, -0.1251, -0.1487, -0.1629, -0.1264, -0.0928, -0.1021, -0.0498, -0.0498, -0.0491, -0.0581, -0.0306, -0.0471, -0.0089, -0.0095, 0.0012, -0.0144, 0.0028, -0.0117, -0.0732, -0.0694, -0.0463, -0.0881, -0.1234, -0.0985, -0.0972, -0.09, -0.0846, -0.0846, -0.0784, -0.0911, -0.0879, -0.1024, -0.1024, -0.0887, -0.1157, -0.1163, -0.1438, -0.155, -0.1399, -0.141, -0.1468, -0.1537, -0.1468, -0.1405, -0.1405, -0.2045, -0.1889, -0.1609, -0.1733, -0.1697, -0.1788, -0.1598, -0.1514, -0.1804, -0.1466, -0.161, -0.126, -0.1736, -0.1331, -0.1341, -0.0975, -0.1108, -0.192, -0.1586, -0.1586, -0.1686, -0.1636, -0.1473, -0.1243, -0.1467, -0.1422, -0.1157, -0.1302, 0.0605, 0.0997, 0.0398, 0.0535, 0.0493, 0.0491, 0.0493, 0.0299, 0.0554, 0.0736, 0.0859, 0.1211, 0.0958, 0.0686, 0.1227, 0.1292, 0.1788, 0.267, 0.3178, 0.2592, 0.2305, 0.1793]}
{"unit_id": "orig:1945755188402487741", "ticker_idx": 0, "ts": "2025-07-17T07:59:21+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "N2 is more profitable than N3. Structurally… all nodes below N7 are very tight. Demand for 5nm, 3nm, and future 2nm is very strong. This kind of strong demand has not been seen for a very long time.", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC 2Q25 call summary: all nodes below N7 tight, N2 more profitable than N3, AI/HPC demand explosive, CoWoS expanding — structurally bullish.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC 2Q25 Conference Call Key Comments & Q&A Summary\n\n▶️ Key Q&A\n\nQ. Is the previous outlook for CoWoS supply-demand balance by 2026 still valid?\n\n- AI demand is growing stronger, and we are working hard to narrow the CoWoS supply-demand gap.\n\n- The momentum remains strong and very healthy.\n\nQ. Any changes in the outlook for on-device AI?\n- It takes 1-2 years for customers to complete new product designs, and development is still ongoing.\n\n- While the increase in the quantity of edge devices is moderate, die sizes are increasing (by about 5-10%).\n\n- We expect full-scale growth in 6 months to 1 year.\n\nQ. According to the guidance, revenue seems to decrease in Q4. Is this due to the consumer segment?\n\n- We are being conservative, considering the potential impact and uncertainty of tariffs.\n\nQ. Considering the resumption of H20 shipments to China, is it possible to revise upward the 5-year CAGR for AI accelerator growth to the 40% range?\n\n- It's too early to revise the outlook upward due to the resumption of H20 shipments to China.\n\n- It would be more appropriate to answer this in the next quarter.\n\nQ. What is the expected revenue contribution of N2 ramp-up in 2026? Will it be similar to N3's 10-11% (based on the second year)?\n\n- New nodes have historically ramped up through smartphones, but now HPC products are also participating in the ramp-up.\n\n- Since N2 prices are higher than N3, the revenue contribution will definitely be larger.\n\nQ. With the N5→N3 transition mostly complete, is there a possibility of N3 supply becoming tight in the coming years? Conversely, will N5 utilization rates decline?\n\n- The tight supply situation will continue for the next few years.\n\n- N5 is also very tight.\n\n- The high demand is because many AI products are still on the N4 node.\n\n- These products will transition to N3 within the next two years.\n\n- Currently, all nodes below N7 are very tight.\n\n- One of our strengths is having GIGAFAB clusters.\n\n- 85-90% of equipment from N7, N5, N3 to N2 is common.\n\n- This allows for easy CAPA adjustment and transition between nodes.\n\n- We are currently utilizing N7 CAPA to support N5.\n\nQ. Can wafer prices be adjusted upward to offset increased costs like exchange rates?\n\n- Pricing is just one of several factors; other factors can also be utilized.\n\nQ. Please provide examples and quantify the benefits of integrating AI across the fab, contributing to cost structure improvements and profits.\n\n- Even a 1% improvement in productivity due to AI represents a value of approximately $1 billion.\n\nQ. Large-scale fab investment plans and ability to meet GW-scale demand?\n\n- Recently, AI data center demand has exploded globally.\n\n- Demand for 5nm, 3nm, and future 2nm is very strong.\n\n- This kind of strong demand has not been seen for a very long time.\n\n- We are working very hard to close the gap between supply and demand.\n\nQ. N2 ROI level compared to N3 and future N2 development direction?\n\n- Structurally, N2 is more profitable than N3.\n\n- N2 development is proceeding as planned, with ramp-up starting in the second half of this year, and revenue contribution expected in the first half of next year.\n\nQ. Possibility of CapEx increase in 2026-2027?\n\n- CapEx plans for 2025 were made considering macroeconomic uncertainties.\n\n- It's too early to discuss future CapEx, but there won't be a drastic reduction.\n\nQ. Any change in the 2025 AI growth rate from the previous 100% guidance? Possibility of further CoWoS Capa expansion in 2025?\n\n- We are working very hard to narrow the gap between supply and demand.\n\n- The demand momentum is very strong and healthy.\n\n- We are continuously expanding CoWoS capacity to support customer demand.\n\nQ. Prioritization of advanced packaging technology development?\n\n- We are simultaneously developing various advanced backend packaging technologies.\n\n- These are closely linked to cutting-edge node processes.\n\n- Some transferability between technologies is also secured, allowing for efficient response.\n\nQ. Scale and opportunities in the humanoid robot market?\n\n- It's difficult for humanoid robots to play a significant role within this year, and 2026 is also too early, we believe.\n\n- The medical industry will likely be the first area where humanoids are utilized.\n\n- Humanoid robots require a wide variety of sensors, as well as extreme stability and precision.\n\nQ. Are customers pulling demand forward into 2H25 before the 2026 price increase?\n\n- No pull-in demand from customers detected.\n\n- Considering the 3nm process cycle time of approximately 4 months, even orders placed now would be difficult to produce within this year.\n\n- Current process capacity is also very tight, so there is no room to pull demand forward.\n\nQ. Have AI and HPC entered a phase where they, along with smartphones, are driving A16 demand?\n\n- Yes. Until now, HPC customers have used N+1, N+2 nodes, but moving forward, AI demand will rapidly shift to A16.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032736137765441486, "ret_signed_m1d": -0.032736137765441486, "alpha_spy_m1d": -0.02665367060126722, "alpha_c_m1d": -0.024405971921902414, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021172647928687227, "ret_signed_p1d": -0.021172647928687227, "alpha_spy_p1d": -0.02044023502657677, "alpha_c_p1d": -0.016373454888251704, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": -0.016286599225248666, "ret_signed_p1w": -0.016286599225248666, "alpha_spy_p1w": -0.02644521420138235, "alpha_c_p1w": -0.002882922782284747, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": -0.02736158446200665, "ret_signed_p1m": -0.02736158446200665, "alpha_spy_p1m": -0.051882372472820015, "alpha_c_p1m": -0.040868208695853725, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.20876894271373292, "ret_signed_p3m": 0.20876894271373292, "alpha_spy_p3m": 0.15140639950328283, "alpha_c_p3m": 0.06551011307794985, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 0.3254373776897599, "ret_signed_p6m": 0.3254373776897599, "alpha_spy_p6m": 0.2139622326356443, "alpha_c_p6m": -0.012984752748134376, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.4002, -0.3858, -0.3598, -0.334, -0.3302, -0.3371, -0.3327, -0.3238, -0.2993, -0.2727, -0.2844, -0.3011, -0.2919, -0.2892, -0.2839, -0.2415, -0.213, -0.2099, -0.2121, -0.2121, -0.215, -0.2152, -0.2221, -0.2041, -0.2212, -0.2212, -0.198, -0.2043, -0.2002, -0.2157, -0.2096, -0.1983, -0.1789, -0.1751, -0.1676, -0.1602, -0.1381, -0.1314, -0.1228, -0.1405, -0.1218, -0.1291, -0.1307, -0.1307, -0.1469, -0.1436, -0.1039, -0.0931, -0.0879, -0.0693, -0.0778, -0.0852, -0.0489, -0.044, -0.044, -0.0669, -0.0722, -0.056, -0.0645, -0.0619, -0.0689, -0.0352, -0.0327, 0.0, -0.0212, -0.0275, -0.0448, -0.0215, -0.0163, 0.0, -0.0116, -0.0174, -0.011, -0.0162, -0.0423, -0.0269, -0.0535, -0.0579, -0.0121, -0.0154, -0.0143, -0.0053, -0.0169, -0.0187, -0.0274, -0.0171, -0.0525, -0.0692, -0.0744, -0.0513, -0.0408, -0.028, -0.0257, -0.0298, -0.06, -0.06, -0.0701, -0.0579, -0.0423, -0.0089, 0.0065, 0.0217, 0.0604, 0.0542, 0.0559, 0.0643, 0.0704, 0.0734, 0.0973, 0.0819, 0.1136, 0.1547, 0.1466, 0.13, 0.1165, 0.116, 0.1408, 0.1783, 0.1768, 0.1935, 0.2352, 0.201, 0.2438, 0.2249, 0.1464, 0.2372, 0.2088, 0.2446, 0.2247, 0.2053, 0.216, 0.2029, 0.1799, 0.1875, 0.2048, 0.2182, 0.2316, 0.2461, 0.2385, 0.2271, 0.2452, 0.201, 0.1994, 0.1814, 0.1702, 0.206, 0.1893, 0.187, 0.1526, 0.1633, 0.1519, 0.1351, 0.1533, 0.1335, 0.1235, 0.1626, 0.1628, 0.1843, 0.1843, 0.1907, 0.175, 0.193, 0.2068, 0.1965, 0.2038, 0.233, 0.2393, 0.2668, 0.2485, 0.1961, 0.1784, 0.1749, 0.1343, 0.1659, 0.1834, 0.2011, 0.2162, 0.2237, 0.2237, 0.2403, 0.2324, 0.2269, 0.2446, 0.2446, 0.309, 0.3198, 0.341, 0.3052, 0.3024, 0.3254]}
{"unit_id": "orig:1966772742013022711", "ticker_idx": 0, "ts": "2025-09-13T07:55:37+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Why did I turn bullish on NAND?", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Author turned bullish on NAND, references HBF as part of the thesis.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I wrote this piece on HBF and NAND.\nI hope you find it helpful.\n\nWhy did I turn bullish on NAND? feat. HBF\n\nhttps://t.co/hwvIBaqsWA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01386212617592295, "ret_signed_m1d": -0.01386212617592295, "alpha_spy_m1d": -0.008566461252436297, "alpha_c_m1d": -0.004583888163141569, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03436283827846291, "ret_signed_p1d": 0.03436283827846291, "alpha_spy_p1d": 0.0357396310637446, "alpha_c_p1d": 0.0342326824352777, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.08466854941825827, "ret_signed_p1w": 0.08466854941825827, "alpha_spy_p1w": 0.07289870804419615, "alpha_c_p1w": 0.035835691545748144, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.2793189320400511, "ret_signed_p1m": 0.2793189320400511, "alpha_spy_p1m": 0.27454369721087235, "alpha_c_p1m": 0.19360094467194466, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.9205130151148138, "ret_signed_p3m": 0.9205130151148138, "alpha_spy_p3m": 0.8748627834405329, "alpha_c_p3m": 0.7130709112941018, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 2.8310571784434377, "ret_signed_p6m": 2.8310571784434377, "alpha_spy_p6m": 2.800562843429324, "alpha_c_p6m": 2.5335141440483113, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.3459, -0.3452, -0.3275, -0.3188, -0.2928, -0.2881, -0.2879, -0.2934, -0.298, -0.311, -0.3073, -0.2925, -0.3062, -0.312, -0.2941, -0.293, -0.2859, -0.2864, -0.3012, -0.2959, -0.3024, -0.3185, -0.3176, -0.3164, -0.3272, -0.3152, -0.3143, -0.3212, -0.3133, -0.3104, -0.299, -0.3031, -0.3303, -0.3238, -0.3202, -0.3277, -0.3219, -0.3019, -0.2937, -0.268, -0.2663, -0.2684, -0.2786, -0.2879, -0.2932, -0.3082, -0.316, -0.3015, -0.2973, -0.296, -0.293, -0.2769, -0.271, -0.2807, -0.2805, -0.2726, -0.2418, -0.1963, -0.1843, -0.1681, -0.128, -0.0642, -0.0139, 0.0, 0.0344, 0.0174, 0.0619, 0.062, 0.0847, 0.1038, 0.0889, 0.0689, 0.034, 0.1023, 0.1111, 0.1551, 0.2269, 0.2764, 0.2706, 0.2445, 0.2816, 0.2867, 0.2739, 0.3166, 0.2793, 0.3533, 0.4171, 0.4017, 0.464, 0.4506, 0.4571, 0.5076, 0.6542, 0.704, 0.6698, 0.7963, 0.8264, 0.834, 0.9113, 0.8227, 0.8709, 0.9002, 0.9833, 1.1465, 1.1489, 1.1791, 1.0677, 0.9225, 1.0304, 0.897, 0.9137, 0.8102, 0.7232, 0.8078, 0.7925, 0.7338, 0.7778, 0.8037, 0.7617, 0.7838, 0.7426, 0.7733, 0.8465, 0.9074, 0.8791, 0.9166, 0.9205, 0.8403, 0.774, 0.7414, 0.7807, 0.8472, 0.8954, 0.9829, 0.9883, 1.0438, 1.0524, 1.1, 1.099, 1.0847, 1.0692, 1.0692, 1.2298, 1.3011, 1.54, 1.6081, 1.5666, 1.664, 1.6939, 1.7202, 1.701, 1.7823, 1.893, 1.9196, 1.9864, 2.1918, 2.2873, 2.205, 2.1714, 2.3251, 2.5141, 2.5618, 2.7411, 2.7582, 3.003, 2.711, 2.5743, 2.6143, 2.6086, 2.4922, 2.6629, 2.8866, 2.9527, 2.9281, 2.8363, 2.8422, 2.9076, 2.9879, 3.0243, 3.0866, 3.0677, 3.1723, 3.1032, 3.0812, 2.7306, 2.6741, 2.7495, 2.5967, 2.5774, 2.8311]}
{"unit_id": "orig:1969891421084860675", "ticker_idx": 0, "ts": "2025-09-21T22:28:08+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the recent surge in commodity memory demand is expected to offset much of this risk", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi: Samsung HBM3E qualification nearing completion, HBM oversupply fears easing on commodity demand surge and HBM4 transition; bullish read for Samsung and MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250922_Samsung’s HBM3e Progressing; HBM Oversupply Risks Easing on HBM4 Transition and Strong Commodity Memory Demand - CITI\n\n(1) According to the Korean media outlet Chosun Media, Samsung has reportedly passed NVIDIA’s qualification test for HBM3E 12hi.\n\n(2) If this is true, Samsung will have joined NVIDIA’s HBM3E 12hi supply chain approximately 19 months after first beginning development of the product.\n\n(3) We believe the qualification test is now in the final stages, and the results are expected to come out between late September and early October.\n\n(4) If Samsung passes the qualification test, it could potentially be adopted in the B300, which for the first time will be equipped with liquid cooling.\n\n(5) However, we expect Samsung’s HBM3E 12hi supply volume to be more limited than the level anticipated by the market.\n\n(6) Meanwhile, although concerns remain about HBM oversupply due to Samsung’s entry, the recent surge in commodity memory demand is expected to offset much of this risk.\n\n(7) In addition, Samsung is prioritizing HBM4 development over HBM3E 12hi in order to regain memory leadership, and the specification upgrade [10Gbps] is also expected to slow down the overall pace of HBM supply growth.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011481025777169696, "ret_signed_m1d": -0.011481025777169696, "alpha_spy_m1d": -0.006772294852597338, "alpha_c_m1d": 0.008570531977778018, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01087348305772684, "ret_signed_p1d": 0.01087348305772684, "alpha_spy_p1d": 0.0163169578975344, "alpha_c_p1d": 0.012332322371607751, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.004373843452624926, "ret_signed_p1w": -0.004373843452624926, "alpha_spy_p1w": 0.00036489580524323717, "alpha_c_p1w": -0.005894730709706697, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.22959944416106892, "ret_signed_p1m": 0.22959944416106892, "alpha_spy_p1m": 0.22292627200685078, "alpha_c_p1m": 0.15780512786668988, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5107863663515086, "ret_signed_p3m": 0.5107863663515086, "alpha_spy_p3m": 0.49634517812828816, "alpha_c_p3m": 0.43293928745085, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.8074702227486457, "ret_signed_p6m": 1.8074702227486457, "alpha_spy_p6m": 1.7985744088282953, "alpha_c_p6m": 1.5717408696055801, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2277, -0.2353, -0.2428, -0.252, -0.2663, -0.2612, -0.2578, -0.2578, -0.2715, -0.2442, -0.2574, -0.2522, -0.2435, -0.2795, -0.2704, -0.2927, -0.312, -0.3051, -0.3122, -0.3365, -0.3328, -0.3213, -0.3241, -0.3242, -0.3199, -0.303, -0.337, -0.3629, -0.3453, -0.3375, -0.3392, -0.3204, -0.2778, -0.2485, -0.224, -0.2451, -0.2389, -0.2658, -0.2495, -0.2586, -0.288, -0.2966, -0.2851, -0.2928, -0.2923, -0.2847, -0.2589, -0.2771, -0.2771, -0.2803, -0.2788, -0.2455, -0.202, -0.2014, -0.1785, -0.1496, -0.0853, -0.0449, -0.0416, -0.0352, -0.0281, 0.0259, -0.0115, 0.0, 0.0109, -0.0177, -0.0473, -0.0446, -0.0044, 0.0164, 0.1065, 0.1162, 0.1417, 0.1607, 0.1287, 0.1946, 0.1691, 0.1038, 0.1717, 0.137, 0.1667, 0.2311, 0.2301, 0.2568, 0.2296, 0.2064, 0.2565, 0.3313, 0.3379, 0.3489, 0.3775, 0.3616, 0.3602, 0.4266, 0.3253, 0.4436, 0.4487, 0.4462, 0.5397, 0.4656, 0.4886, 0.4403, 0.5003, 0.4707, 0.3889, 0.3732, 0.224, 0.2605, 0.3611, 0.3648, 0.3996, 0.3996, 0.4374, 0.4616, 0.4557, 0.4233, 0.3777, 0.4419, 0.5009, 0.5343, 0.6029, 0.571, 0.4657, 0.4436, 0.4133, 0.3708, 0.5108, 0.6164, 0.6812, 0.6793, 0.7426, 0.7426, 0.7311, 0.79, 0.7794, 0.7355, 0.7355, 0.918, 0.8981, 1.0883, 1.0648, 0.9886, 1.0984, 1.1032, 1.0561, 1.0271, 1.047, 1.2058, 1.2058, 1.2195, 1.3661, 1.4176, 1.4302, 1.366, 1.4946, 1.6469, 1.65, 1.5228, 1.6622, 1.5506, 1.3071, 1.3283, 1.4001, 1.332, 1.2697, 1.4952, 1.5173, 1.5032, 1.5032, 1.431, 1.5597, 1.5378, 1.6036, 1.5599, 1.5419, 1.6087, 1.527, 1.5076, 1.5094, 1.3088, 1.437, 1.4144, 1.2517, 1.3674, 1.4513, 1.546, 1.4649, 1.5912, 1.6865, 1.8075]}
{"unit_id": "orig:1969891421084860675", "ticker_idx": 1, "ts": "2025-09-21T22:28:08+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has reportedly passed NVIDIA's qualification test for HBM3E 12hi", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi: Samsung HBM3E qualification nearing completion, HBM oversupply fears easing on commodity demand surge and HBM4 transition; bullish read for Samsung and MU.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250922_Samsung’s HBM3e Progressing; HBM Oversupply Risks Easing on HBM4 Transition and Strong Commodity Memory Demand - CITI\n\n(1) According to the Korean media outlet Chosun Media, Samsung has reportedly passed NVIDIA’s qualification test for HBM3E 12hi.\n\n(2) If this is true, Samsung will have joined NVIDIA’s HBM3E 12hi supply chain approximately 19 months after first beginning development of the product.\n\n(3) We believe the qualification test is now in the final stages, and the results are expected to come out between late September and early October.\n\n(4) If Samsung passes the qualification test, it could potentially be adopted in the B300, which for the first time will be equipped with liquid cooling.\n\n(5) However, we expect Samsung’s HBM3E 12hi supply volume to be more limited than the level anticipated by the market.\n\n(6) Meanwhile, although concerns remain about HBM oversupply due to Samsung’s entry, the recent surge in commodity memory demand is expected to offset much of this risk.\n\n(7) In addition, Samsung is prioritizing HBM4 development over HBM3E 12hi in order to regain memory leadership, and the specification upgrade [10Gbps] is also expected to slow down the overall pace of HBM supply growth.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "thread:1943850009528873425", "ticker_idx": 0, "ts": "2025-07-12T01:48:51+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Suppliers are taking comfort from the momentum driven by both GB200 and GB300, viewing it as a positive sign for business performance in the second half of 2025", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GB300 on track per Dell/CoreWeave delivery; supply chain expects continued AI server revenue growth in H2 2025, bullish for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: With GB300 on track and no Cordelia detours, Nvidia eyes a strong AI finish in 2025\n\nWith Nvidia's GB200 series products entering volume shipments starting in the second quarter of 2025, the GB300 has also begun its delivery recently, as Dell chairman and CEO Michael Dell has unveiled that his company's GB300 NVL72 products have officially been delivered to AI cloud computing provider CoreWeave for use. This indirectly confirms that the GB300, originally expected to start small-volume shipments in the third quarter of 2025, has not been affected by the volume ramp-up of GB200 and is shipping according to plan.\n\nSuppliers are taking comfort from the momentum driven by both GB200 and GB300, viewing it as a positive sign for business performance in the second half of 2025.\n\nAt Nvidia GTC in March 2025, the company first revealed the architecture and products of GB300, announcing plans to begin engineering sample shipments and limited test orders in the second quarter of 2025, with mass production and initial shipments slated for the third quarter of 2025. However, since the GB200 series had yet to reach volume in the market at that time, there was concern whether this would cause cannibalization or delay GB300 shipments.\n\nAccording to supply chain sources, key challenges related to manufacturing process technology, liquid cooling systems, and thermal management pushed back the mass production timeline for GB200 products from September 2024 to the second quarter of 2025. Many suppliers then expected that GB300 might also face delays as a result.\n\nHowever, in April 2025, reports emerged that the Cordelia architecture initially planned for GB300 would be replaced by the existing Bianca architecture used in GB200. This change boosted confidence among industry players regarding GB300's planned volume shipments in the fourth quarter of 2025. Dell's delivery to CoreWeave further confirmed that GB300's shipment schedule remains on track.\n\nIn addition, many designs and components originally tailored for GB300 were reportedly deferred or scaled down, favoring mature designs based on GB200.\n\nSupply chain insiders noted that the overall design and manufacturing processes have become more refined, making the launch of GB300 less challenging. The critical factor now lies in real-world testing. Previously, GB200 encountered problems requiring fixes only after being shipped for field testing.\n\nDemand for AI servers remains strong on the client side, with manufacturers working overtime to support client deliveries. Major GB200 shipment leaders such as Foxconn, Wistron, Quanta Computer, and their associated supply chains are expected to continue benefiting from the growing market demand for AI computing power, anticipating revenue growth in the second half of 2025.\n\nSupply chain experts believe that because architectural changes are limited, the overall learning curve should shorten for GB300. Provided field tests meet targets smoothly and upstream semiconductor supplies remain stable, downstream assembly and supply chain partners can fully cooperate.\n\nAs for clients potentially shifting investment focus from GB200 to GB300, supply chain insiders said this depends entirely on considerations of return on investment. Regardless of whether it is GB200 or GB300, supply chain stakeholders expect to continue enjoying the benefits brought by the AI wave throughout the latter half of 2025.\n\n$NVDA\n\n---\n\nhttps://t.co/NmTHEVIaXI", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005180683499361294, "ret_signed_m1d": 0.005180683499361294, "alpha_spy_m1d": 0.007085166748686311, "alpha_c_m1d": -0.0024247053453889666, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.040409535932759866, "ret_signed_p1d": 0.040409535932759866, "alpha_spy_p1d": 0.044682702983386946, "alpha_c_p1d": 0.021203099274968107, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.04455419435283536, "ret_signed_p1w": 0.04455419435283536, "alpha_spy_p1w": 0.03821616236811698, "alpha_c_p1w": 0.0263641341996117, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.11635274151788177, "ret_signed_p1m": 0.11635274151788177, "alpha_spy_p1m": 0.0877358414456213, "alpha_c_p1m": 0.06455133674457092, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.17377251073301703, "ret_signed_p3m": 0.17377251073301703, "alpha_spy_p3m": 0.09661177882903127, "alpha_c_p3m": -0.036132036996605166, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.1413481255415343, "ret_signed_p6m": 0.1413481255415343, "alpha_spy_p6m": 0.02776489035331986, "alpha_c_p6m": -0.22142382598611499, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.3632, -0.3815, -0.3815, -0.4094, -0.3973, -0.374, -0.3514, -0.3234, -0.3373, -0.3356, -0.3362, -0.3198, -0.3022, -0.3063, -0.308, -0.2866, -0.2847, -0.2891, -0.2504, -0.2081, -0.1752, -0.1783, -0.1748, -0.1738, -0.181, -0.1967, -0.1905, -0.1998, -0.1998, -0.1742, -0.1784, -0.1517, -0.1764, -0.1627, -0.1393, -0.1351, -0.1468, -0.1363, -0.1307, -0.1226, -0.1295, -0.1162, -0.1347, -0.1181, -0.1216, -0.1133, -0.1133, -0.1232, -0.1213, -0.0986, -0.0595, -0.0552, -0.0385, -0.0371, -0.0656, -0.0416, -0.0288, -0.0288, -0.0355, -0.0248, -0.0073, 0.0002, 0.0052, 0.0, 0.0404, 0.0445, 0.0544, 0.0508, 0.0446, 0.018, 0.0409, 0.0589, 0.0575, 0.0773, 0.0697, 0.0926, 0.0841, 0.0588, 0.0971, 0.0865, 0.0936, 0.1018, 0.1135, 0.1096, 0.1164, 0.1068, 0.1094, 0.0998, 0.1093, 0.0705, 0.0691, 0.0665, 0.0848, 0.0959, 0.1079, 0.1068, 0.0981, 0.0616, 0.0616, 0.0409, 0.0399, 0.0463, 0.018, 0.0258, 0.0408, 0.0808, 0.0799, 0.0839, 0.0834, 0.0659, 0.038, 0.0742, 0.0769, 0.1192, 0.0876, 0.0787, 0.0831, 0.0861, 0.1084, 0.1373, 0.1413, 0.1513, 0.1436, 0.1309, 0.1279, 0.1527, 0.1738, 0.1164, 0.1479, 0.0973, 0.0961, 0.1082, 0.1168, 0.1132, 0.1042, 0.0989, 0.1103, 0.1353, 0.1672, 0.2253, 0.262, 0.2367, 0.2342, 0.261, 0.2111, 0.1899, 0.1464, 0.1468, 0.2133, 0.1774, 0.1813, 0.139, 0.1591, 0.1374, 0.1054, 0.1369, 0.1011, 0.0903, 0.1127, 0.0839, 0.0987, 0.0987, 0.0789, 0.0967, 0.1061, 0.0947, 0.1178, 0.1119, 0.131, 0.1275, 0.1203, 0.1029, 0.0669, 0.0746, 0.0833, 0.042, 0.0615, 0.1033, 0.1197, 0.1534, 0.1497, 0.1497, 0.1614, 0.1473, 0.1432, 0.1368, 0.1368, 0.1512, 0.1467, 0.1413]}
{"unit_id": "quote:1965978991447388338", "ticker_idx": 0, "ts": "2025-09-11T03:21:32+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think NAND could actually end up in shortage", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Shifting view on NAND; expects shortage conditions ahead, turning bullish on the segment.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket: dominant producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "To be honest, I used to overlook NAND too.\n\nBut lately I’ve been starting to change my mind. I think NAND could actually end up in shortage. https://t.co/GL6GOFH5vy", "reply_text": null, "parent_text": null, "quoted_text": "Morgan Stanley: Why We Forecast Strength in the NAND Market in the AI Era https://t.co/EH5trLt4RT", "parent_author": null, "ret_m1d": -0.06876806263566948, "ret_signed_m1d": -0.06876806263566948, "alpha_spy_m1d": -0.06052628922280401, "alpha_c_m1d": -0.05856831514655125, "bench_spy_m1d": -0.008241773412865472, "bench_c_m1d": -0.010199747489118227, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.054366957986921746, "ret_signed_p1d": 0.054366957986921746, "alpha_spy_p1d": 0.05470157089874954, "alpha_c_p1d": 0.053083654777958775, "bench_spy_p1d": -0.00033461291182779185, "bench_c_p1d": 0.0012833032089629715, "ret_p1w": 0.1350162529046902, "ret_signed_p1w": 0.1350162529046902, "alpha_spy_p1w": 0.12797590364181674, "alpha_c_p1w": 0.09211156191324496, "bench_spy_p1w": 0.0070403492628734465, "bench_c_p1w": 0.04290469099144523, "ret_p1m": 0.36354547239914675, "ret_signed_p1m": 0.36354547239914675, "alpha_spy_p1m": 0.3678025910197516, "alpha_c_p1m": 0.2931015840538787, "bench_spy_p1m": -0.0042571186206048495, "bench_c_p1m": 0.07044388834526805, "ret_p3m": 1.0149232074003147, "ret_signed_p3m": 1.0149232074003147, "alpha_spy_p3m": 0.9734050924868465, "alpha_c_p3m": 0.8008278223464005, "bench_spy_p3m": 0.04151811491346824, "bench_c_p3m": 0.21409538505391423, "ret_p6m": 2.86204917344888, "ret_signed_p6m": 2.86204917344888, "alpha_spy_p6m": 2.8337560839870526, "alpha_c_p6m": 2.6060225787248164, "bench_spy_p6m": 0.028293089461827403, "bench_c_p6m": 0.25602659472406364, "price_path": [-0.3195, -0.3123, -0.3033, -0.3024, -0.2831, -0.2732, -0.2456, -0.2406, -0.2403, -0.2462, -0.251, -0.2651, -0.2612, -0.2454, -0.2603, -0.2662, -0.2469, -0.2454, -0.238, -0.2388, -0.2545, -0.249, -0.2559, -0.2734, -0.2726, -0.2714, -0.2825, -0.2696, -0.2685, -0.2761, -0.268, -0.2651, -0.253, -0.257, -0.286, -0.2793, -0.2754, -0.2835, -0.2775, -0.2563, -0.2476, -0.2197, -0.2179, -0.2203, -0.231, -0.2412, -0.2468, -0.2629, -0.2714, -0.2558, -0.2513, -0.2498, -0.2466, -0.2293, -0.2226, -0.2328, -0.2328, -0.2244, -0.1916, -0.1423, -0.1294, -0.1121, -0.0688, 0.0, 0.0544, 0.0691, 0.1062, 0.0875, 0.135, 0.1353, 0.1597, 0.1801, 0.1644, 0.1431, 0.105, 0.1783, 0.1879, 0.2342, 0.3121, 0.3662, 0.3602, 0.3317, 0.371, 0.3772, 0.3635, 0.4089, 0.369, 0.4483, 0.5172, 0.5003, 0.5679, 0.5535, 0.561, 0.6149, 0.7739, 0.8288, 0.7915, 0.928, 0.9616, 0.9693, 1.0516, 0.9576, 1.0084, 1.0416, 1.1314, 1.3077, 1.3102, 1.3425, 1.2233, 1.0625, 1.1801, 1.0362, 1.0551, 0.9459, 0.8505, 0.9404, 0.9235, 0.8577, 0.906, 0.9343, 0.8884, 0.9126, 0.8683, 0.9013, 0.9796, 1.0458, 1.0149, 1.0541, 1.0587, 0.9739, 0.902, 0.8664, 0.9094, 0.9803, 1.0311, 1.1257, 1.1311, 1.1913, 1.2008, 1.2525, 1.2499, 1.2343, 1.2179, 1.2179, 1.3883, 1.4656, 1.7207, 1.7955, 1.7523, 1.8553, 1.8872, 1.9174, 1.8961, 1.9832, 2.1026, 2.1319, 2.2034, 2.4242, 2.5285, 2.4396, 2.4039, 2.5692, 2.7715, 2.8233, 3.0195, 3.032, 3.2977, 2.9868, 2.8385, 2.8801, 2.8738, 2.7495, 2.9308, 3.1733, 3.2467, 3.2195, 3.1219, 3.1259, 3.195, 3.2796, 3.3185, 3.388, 3.366, 3.4772, 3.4033, 3.3803, 3.0049, 2.9433, 3.0251, 2.862]}
{"unit_id": "orig:1945590322270278113", "ticker_idx": 0, "ts": "2025-07-16T21:04:14+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "with the expected resumption of NVIDIA H20 chip supply, China's AI compute bottleneck is likely to ease", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish thesis on China AI: H20 supply resumption eases compute bottleneck, 44% QoQ capex surge in 3Q25 drives sustained cloud revenue growth.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "China AI Application Tracking: Improved External Compute Supply + Local Model Evolution + Accelerated Enterprise Adoption Jointly Drive China's AI Development into a New Phase\n\nGoldman Sachs (R. Keung, 25/07/15)\n\nGoldman Sachs believes that with the expected resumption of NVIDIA H20 chip supply, China's AI compute bottleneck is likely to ease. Coupled with the continuous evolution of local models, accelerated enterprise adoption, and increasingly clear commercialization paths, China's AI ecosystem is poised for multiple catalysts in the second half of the year. Goldman Sachs expects Chinese cloud vendors to initiate a new capital expenditure cycle in 3Q25, with overall Capex seeing a 44% quarter-over-quarter growth, driving sustained growth in AI cloud revenue.\n\nRapid Evolution of Local Chinese Large Models, Gradually Narrowing the Technology Gap\n\nWhile a gap still exists between top-tier Chinese and U.S. models, recent launches of models like Kimi, Tencent Hunyuan, and ByteDance Doubao, which feature capabilities such as code generation and MoE inference optimization, demonstrate rapid progress in model capabilities by Chinese internet platforms and AI startups.\n\nSignificant Acceleration in Generative AI Adoption by Enterprises\n\nAccording to Gartner data, over 40% of Chinese enterprises are piloting generative AI tools, a significant increase from 8% last year. Meituan's recently launched code agent tool, NoCode, is an example, indicating that enterprises are accelerating AI deployment for internal tools and efficiency improvements.\n\nClear Commercialization Paths for To C, Accelerated Adoption in To B Sector Led by Healthcare\n\nIn the To C (to consumer) segment, subscription-based AI tools like Kuaishou Keling are achieving steady Annual Recurring Revenue (ARR) growth. Ad-driven search and assistant AI (e.g., Quark, WeChat Qbot, Baidu) are enhancing their commercial closed-loop capabilities through improved advertising technology. In the To B (to business) segment, Alibaba DAMO Academy and Ant Group have launched a model that can identify gastric cancer up to six months in advance and an AI health application, respectively, reflecting substantial breakthroughs for AI in the healthcare vertical.\n\nIntensifying Competition in Video Generation Models, All Leading Platforms Have Deployments\n\nKuaishou Keling's AI revenue is projected to grow from $142 million in 2025 to $400 million in 2027, representing a 2-year compound annual growth rate of 68%. Additionally, ByteDance's Seedance, Alibaba's Tongyi Qianxiang (integrating Qwen3 multimodal capabilities), and Baidu's Steamer-I2V and MuseSteamer have also entered the video generation market, indicating a fiercely competitive landscape.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003909684872381747, "ret_signed_m1d": -0.003909684872381747, "alpha_spy_m1d": -0.0005776334386263793, "alpha_c_m1d": -0.00916412460569449, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009511589505708296, "ret_signed_p1d": 0.009511589505708296, "alpha_spy_p1d": 0.003391899528190745, "alpha_c_p1d": 0.0011114490994763582, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0034428594862712103, "ret_signed_p1w": -0.0034428594862712103, "alpha_spy_p1w": -0.019446981286452614, "alpha_c_p1w": 0.005786934868547422, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.0621461963170844, "ret_signed_p1m": 0.0621461963170844, "alpha_spy_p1m": 0.028936714204513825, "alpha_c_p1m": 0.01845147637661637, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.09897081121285733, "ret_signed_p3m": 0.09897081121285733, "alpha_spy_p3m": 0.03383631036483137, "alpha_c_p3m": -0.07549684249508726, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.07988992467564904, "ret_signed_p6m": 0.07988992467564904, "alpha_spy_p6m": -0.031040142771990187, "alpha_c_p6m": -0.23426679716688548, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4078, -0.4345, -0.423, -0.4007, -0.379, -0.3523, -0.3656, -0.3639, -0.3645, -0.3488, -0.3319, -0.3359, -0.3375, -0.317, -0.3152, -0.3194, -0.2823, -0.2419, -0.2103, -0.2133, -0.21, -0.209, -0.2159, -0.231, -0.2249, -0.2339, -0.2339, -0.2094, -0.2134, -0.1878, -0.2115, -0.1984, -0.176, -0.1719, -0.1832, -0.1731, -0.1678, -0.16, -0.1665, -0.1539, -0.1716, -0.1557, -0.159, -0.1511, -0.1511, -0.1606, -0.1587, -0.137, -0.0996, -0.0954, -0.0795, -0.0781, -0.1054, -0.0824, -0.0702, -0.0702, -0.0766, -0.0663, -0.0495, -0.0424, -0.0376, -0.0426, -0.0039, 0.0, 0.0095, 0.0061, 0.0001, -0.0253, -0.0034, 0.0138, 0.0124, 0.0314, 0.0242, 0.0461, 0.0379, 0.0137, 0.0504, 0.0402, 0.047, 0.0549, 0.0661, 0.0624, 0.0688, 0.0596, 0.0621, 0.053, 0.0621, 0.0249, 0.0235, 0.0211, 0.0386, 0.0493, 0.0607, 0.0597, 0.0514, 0.0164, 0.0164, -0.0034, -0.0044, 0.0017, -0.0254, -0.0179, -0.0036, 0.0348, 0.0339, 0.0377, 0.0373, 0.0205, -0.0062, 0.0285, 0.031, 0.0715, 0.0413, 0.0327, 0.0369, 0.0399, 0.0612, 0.0888, 0.0927, 0.1023, 0.0949, 0.0827, 0.0798, 0.1036, 0.1238, 0.0689, 0.099, 0.0506, 0.0494, 0.061, 0.0692, 0.0658, 0.0572, 0.0521, 0.063, 0.0869, 0.1175, 0.1731, 0.2082, 0.184, 0.1817, 0.2073, 0.1595, 0.1392, 0.0976, 0.098, 0.1616, 0.1272, 0.131, 0.0905, 0.1098, 0.0889, 0.0584, 0.0885, 0.0542, 0.0439, 0.0653, 0.0377, 0.0519, 0.0519, 0.0329, 0.05, 0.0589, 0.048, 0.0702, 0.0645, 0.0829, 0.0795, 0.0725, 0.0559, 0.0214, 0.0288, 0.0372, -0.0024, 0.0163, 0.0563, 0.072, 0.1042, 0.1007, 0.1007, 0.1119, 0.0984, 0.0945, 0.0884, 0.0884, 0.1021, 0.0979, 0.0927, 0.1036, 0.0799]}
{"unit_id": "orig:1945590322270278113", "ticker_idx": 1, "ts": "2025-07-16T21:04:14+00:00", "symbol": "China internet", "kind": "sector", "geography": "China", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Goldman Sachs expects Chinese cloud vendors to initiate a new capital expenditure cycle in 3Q25, with overall Capex seeing a 44% quarter-over-quarter growth, driving sustained growth in AI cloud revenue", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish thesis on China AI: H20 supply resumption eases compute bottleneck, 44% QoQ capex surge in 3Q25 drives sustained cloud revenue growth.", "resolved_tickers": ["KWEB"], "resolution_method": "etf", "resolution_reason": "China internet sector; KWEB ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "China AI Application Tracking: Improved External Compute Supply + Local Model Evolution + Accelerated Enterprise Adoption Jointly Drive China's AI Development into a New Phase\n\nGoldman Sachs (R. Keung, 25/07/15)\n\nGoldman Sachs believes that with the expected resumption of NVIDIA H20 chip supply, China's AI compute bottleneck is likely to ease. Coupled with the continuous evolution of local models, accelerated enterprise adoption, and increasingly clear commercialization paths, China's AI ecosystem is poised for multiple catalysts in the second half of the year. Goldman Sachs expects Chinese cloud vendors to initiate a new capital expenditure cycle in 3Q25, with overall Capex seeing a 44% quarter-over-quarter growth, driving sustained growth in AI cloud revenue.\n\nRapid Evolution of Local Chinese Large Models, Gradually Narrowing the Technology Gap\n\nWhile a gap still exists between top-tier Chinese and U.S. models, recent launches of models like Kimi, Tencent Hunyuan, and ByteDance Doubao, which feature capabilities such as code generation and MoE inference optimization, demonstrate rapid progress in model capabilities by Chinese internet platforms and AI startups.\n\nSignificant Acceleration in Generative AI Adoption by Enterprises\n\nAccording to Gartner data, over 40% of Chinese enterprises are piloting generative AI tools, a significant increase from 8% last year. Meituan's recently launched code agent tool, NoCode, is an example, indicating that enterprises are accelerating AI deployment for internal tools and efficiency improvements.\n\nClear Commercialization Paths for To C, Accelerated Adoption in To B Sector Led by Healthcare\n\nIn the To C (to consumer) segment, subscription-based AI tools like Kuaishou Keling are achieving steady Annual Recurring Revenue (ARR) growth. Ad-driven search and assistant AI (e.g., Quark, WeChat Qbot, Baidu) are enhancing their commercial closed-loop capabilities through improved advertising technology. In the To B (to business) segment, Alibaba DAMO Academy and Ant Group have launched a model that can identify gastric cancer up to six months in advance and an AI health application, respectively, reflecting substantial breakthroughs for AI in the healthcare vertical.\n\nIntensifying Competition in Video Generation Models, All Leading Platforms Have Deployments\n\nKuaishou Keling's AI revenue is projected to grow from $142 million in 2025 to $400 million in 2027, representing a 2-year compound annual growth rate of 68%. Additionally, ByteDance's Seedance, Alibaba's Tongyi Qianxiang (integrating Qwen3 multimodal capabilities), and Baidu's Steamer-I2V and MuseSteamer have also entered the video generation market, indicating a fiercely competitive landscape.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008774292195602884, "ret_signed_m1d": 0.008774292195602884, "alpha_spy_m1d": 0.012106343629358252, "alpha_c_m1d": 0.012106343629358252, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.0033320514337553675, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010755572896089083, "ret_signed_p1d": 0.010755572896089083, "alpha_spy_p1d": 0.004635882918571532, "alpha_c_p1d": 0.004635882918571532, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.006119689977517551, "ret_p1w": 0.04896682099410676, "ret_signed_p1w": 0.04896682099410676, "alpha_spy_p1w": 0.03296269919392536, "alpha_c_p1w": 0.03296269919392536, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.016004121800181403, "ret_p1m": 0.03651279475932001, "ret_signed_p1m": 0.03651279475932001, "alpha_spy_p1m": 0.0033033126467494345, "alpha_c_p1m": 0.0033033126467494345, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.033209482112570576, "ret_p3m": 0.1282195381815221, "ret_signed_p3m": 0.1282195381815221, "alpha_spy_p3m": 0.06308503733349613, "alpha_c_p3m": 0.06308503733349613, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.06513450084802597, "ret_p6m": 0.07864009639727887, "ret_signed_p6m": 0.07864009639727887, "alpha_spy_p6m": -0.032289971050360355, "alpha_c_p6m": -0.032289971050360355, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11093006744763922, "price_path": [-0.1449, -0.1432, -0.1149, -0.094, -0.0923, -0.094, -0.0937, -0.0892, -0.0914, -0.0903, -0.049, -0.0478, -0.0382, -0.0654, -0.0552, -0.0679, -0.0161, -0.0184, -0.0031, -0.0294, -0.0292, -0.028, -0.0289, -0.0408, -0.0459, -0.0425, -0.0425, -0.0509, -0.0597, -0.0461, -0.0733, -0.0614, -0.0628, -0.0396, -0.0314, -0.0309, -0.0091, -0.0127, -0.013, -0.0136, -0.0376, -0.0142, -0.0269, -0.0391, -0.0391, -0.0575, -0.0476, -0.0187, -0.0255, -0.028, -0.0241, -0.0283, -0.028, -0.0402, -0.0481, -0.0481, -0.0413, -0.0238, -0.0371, -0.0359, -0.0416, -0.0297, 0.0088, 0.0, 0.0108, 0.0252, 0.0215, 0.0306, 0.049, 0.0357, 0.0277, 0.019, 0.0122, -0.0057, 0.0031, -0.0235, -0.002, 0.0048, 0.0178, 0.015, 0.0142, 0.0045, 0.028, 0.0651, 0.0365, 0.0521, 0.0617, 0.0512, 0.0515, 0.0586, 0.0863, 0.09, 0.0903, 0.0606, 0.0623, 0.0807, 0.0807, 0.0818, 0.0829, 0.0739, 0.088, 0.105, 0.1313, 0.1268, 0.1531, 0.15, 0.1588, 0.1727, 0.2055, 0.1877, 0.1809, 0.1729, 0.1497, 0.1763, 0.1797, 0.1625, 0.1874, 0.1891, 0.2012, 0.2154, 0.2072, 0.2134, 0.1922, 0.197, 0.1701, 0.0875, 0.1282, 0.1081, 0.1226, 0.1189, 0.1177, 0.1424, 0.1308, 0.1169, 0.1398, 0.1483, 0.1704, 0.1622, 0.165, 0.1384, 0.1308, 0.1308, 0.1053, 0.1138, 0.118, 0.103, 0.1274, 0.1206, 0.1129, 0.0985, 0.0824, 0.0736, 0.071, 0.0558, 0.0255, 0.032, 0.062, 0.0682, 0.0614, 0.0614, 0.0634, 0.0753, 0.0643, 0.0512, 0.0507, 0.064, 0.0626, 0.049, 0.0526, 0.0476, 0.0442, 0.0263, 0.0232, 0.0147, 0.0243, 0.0331, 0.0438, 0.0384, 0.0363, 0.0363, 0.0459, 0.0351, 0.0327, 0.0225, 0.0225, 0.0699, 0.0946, 0.0913, 0.0666, 0.0786]}
{"unit_id": "orig:1945088732211896830", "ticker_idx": 0, "ts": "2025-07-15T11:51:06+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley's base case is Scenario 2, where TSMC upgrades its full-year revenue growth to a high double-digit percentage... This scenario has a 60% probability, with an expected stock price increase of 2%", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares Morgan Stanley Buy-thesis on TSMC: Q2 beat, full-year guidance upgrade likely, earnings call July 17 as catalyst with base-case 2% stock upside.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC: Q2 Revenue Exceeds Expectations, Increasing Likelihood of Full-Year Guidance Upgrade; Focus on Information from the Earnings Call\n\nMorgan Stanley (C. Chan, 14/07/2025)\n\nMorgan Stanley notes that TSMC's Q2'25 revenue in USD surpassed the high end of its guidance, significantly increasing the probability of an upward revision to high double-digit growth for the full year. Strong AI demand persists, with customers pulling in orders ahead of schedule. The earnings call on July 17th is expected to be a catalyst for the stock price.\n\nQ2'25 USD Revenue Reaches $29.6 Billion, Exceeding the Upper Limit of Company Guidance and Morgan Stanley's Expectations\n\nTSMC announced preliminary Q2'25 revenue of NT$933 billion. Based on an average exchange rate of 31.5, this corresponds to a USD revenue of approximately $29.6 billion, a 16% sequential increase. This figure exceeds the company's guidance ceiling of $28.4-29.2 billion and is also higher than Morgan Stanley's previous forecast of +15% QoQ growth. Morgan Stanley believes that about 3-4 percentage points of this quarter's revenue growth came from customers pulling in orders early in response to tariffs.\n\nFull-Year Revenue Growth Guidance Likely to Be Upgraded to a High Double-Digit Percentage\n\nPreviously, Morgan Stanley had forecast a 27% year-over-year growth in full-year USD revenue, which was higher than the company's original guidance of mid-teens growth. Assuming a 1% QoQ growth in 3Q and flat performance in 4Q, this 27% growth target is now achievable. Morgan Stanley's industry research indicates that even with moderate demand for mobile phones and PCs, TSMC's capacity utilization will remain high in the second half of the year. Furthermore, adjustments in some PC CPU orders have been offset by demand from cryptocurrency mining, reflecting customers' intent to place orders early before the price hikes in 2026.\n\nSemiconductor Tariffs Remain an Unresolved Uncertainty\n\nOne of the three major uncertainties facing TSMC is the issue of tariffs. The market is currently focused on whether the Section 232 investigation will grant the semiconductor industry a grace period of 1-1.5 years, similar to the pharmaceutical industry. Morgan Stanley believes that due to its expanding investments in the U.S., TSMC is likely to receive an exemption or a grace period, and could also benefit from new U.S. tax credit policies.\n\nEarnings Call is a Key Catalyst; Market Focus on 2025 Revenue Outlook and 2026 Wafer Price Hike Plans\n\nAt the July 17th earnings call, the key focus will be whether the company revises its full-year revenue guidance upward. Morgan Stanley's base case is Scenario 2, where TSMC upgrades its full-year revenue growth to a high double-digit percentage, provides guidance for slight sequential growth in 3Q, maintains its $40 billion capital expenditure plan, and hints that the 2026 price hike strategy will continue. This scenario has a 60% probability, with an expected stock price increase of 2%.\n\nTwo Other Scenarios Correspond to Stronger/Weaker Stock Price Reactions\n\nIn Scenario 1, TSMC would upgrade its full-year growth forecast to over 30% and explicitly state its price hike plans, potentially leading to a 5% increase in the stock price. In Scenario 3, TSMC would maintain its original growth guidance and project a sequential revenue decline in 3Q, which could cause the stock price to fall by 5%. The outcome of the earnings call will significantly impact market sentiment and stock price trends.\n\nKey Questions for the July 17th Earnings Call:\n\n- What factors will TSMC consider for its 2026 wafer pricing, and will this include foreign exchange impacts?\n- What is the sustainability of demand for cloud AI semiconductors, and what will TSMC's CoWoS capacity be in 2026?\n- How will TSMC capitalize on the strong AI demand opportunities in China?\n- How will semiconductor tariffs affect customer ordering behavior? Does the company anticipate a recovery in other semiconductor segments besides AI?", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013513498494239662, "ret_signed_m1d": -0.013513498494239662, "alpha_spy_m1d": -0.017805003864211355, "alpha_c_m1d": 0.00533100244982232, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01801788684969341, "ret_signed_p1d": 0.01801788684969341, "alpha_spy_p1d": 0.0146746957311783, "alpha_c_p1d": 0.023244861758361046, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.01801788684969341, "ret_signed_p1w": 0.01801788684969341, "alpha_spy_p1w": 0.007216503428878029, "alpha_c_p1w": 0.03669048176540157, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.0810811021080644, "ret_signed_p1m": 0.0810811021080644, "alpha_spy_p1m": 0.04451376818499875, "alpha_c_p1m": 0.04504252441622447, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.3024865495193749, "ret_signed_p3m": 0.3024865495193749, "alpha_spy_p3m": 0.24994143137881086, "alpha_c_p3m": 0.18371074782461094, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.5200951425337497, "ret_signed_p6m": 0.5200951425337497, "alpha_spy_p6m": 0.4053378392525626, "alpha_c_p6m": 0.19197264269797087, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2402, -0.2375, -0.2509, -0.268, -0.2168, -0.2249, -0.2034, -0.1953, -0.1908, -0.1854, -0.1854, -0.1478, -0.1585, -0.1747, -0.1675, -0.1765, -0.1487, -0.1415, -0.1307, -0.1038, -0.1092, -0.1047, -0.1173, -0.1226, -0.1119, -0.1191, -0.1173, -0.1244, -0.1343, -0.1325, -0.1325, -0.1325, -0.1513, -0.1478, -0.1119, -0.1047, -0.1074, -0.0984, -0.0625, -0.0446, -0.0586, -0.0721, -0.0766, -0.0586, -0.0495, -0.0676, -0.0495, -0.0811, -0.0541, -0.036, -0.0315, -0.027, -0.045, -0.0225, -0.0225, -0.018, -0.0225, -0.027, -0.027, -0.018, -0.009, -0.009, -0.0135, 0.0, 0.018, 0.018, 0.0405, 0.036, 0.018, 0.0315, 0.0315, 0.0315, 0.0315, 0.0225, 0.0405, 0.045, 0.045, 0.0225, 0.036, 0.0135, 0.0631, 0.0586, 0.0631, 0.0631, 0.0811, 0.0586, 0.0631, 0.0631, 0.0676, 0.0225, 0.036, 0.0225, 0.0541, 0.0586, 0.0721, 0.045, 0.045, 0.0495, 0.045, 0.045, 0.045, 0.0631, 0.0631, 0.0811, 0.1036, 0.1171, 0.1351, 0.1306, 0.1578, 0.1442, 0.1623, 0.1442, 0.1713, 0.212, 0.212, 0.1939, 0.1759, 0.1759, 0.1804, 0.1985, 0.2346, 0.2663, 0.2663, 0.298, 0.2799, 0.3025, 0.3025, 0.2799, 0.2889, 0.3251, 0.3432, 0.3115, 0.3387, 0.3387, 0.3206, 0.3115, 0.3115, 0.3387, 0.3341, 0.3613, 0.3613, 0.3568, 0.3658, 0.3613, 0.3206, 0.3251, 0.3206, 0.3341, 0.3251, 0.3341, 0.3206, 0.2934, 0.307, 0.2708, 0.2618, 0.3161, 0.2527, 0.2437, 0.2799, 0.3025, 0.298, 0.3025, 0.2754, 0.2934, 0.3115, 0.307, 0.3206, 0.3522, 0.3387, 0.3613, 0.3341, 0.3431, 0.3159, 0.3023, 0.2978, 0.2978, 0.2978, 0.3295, 0.3522, 0.3567, 0.3567, 0.3704, 0.3885, 0.3794, 0.4067, 0.4067, 0.4384, 0.5156, 0.5473, 0.5201]}
{"unit_id": "quote:1969672983867506764", "ticker_idx": 0, "ts": "2025-09-21T08:00:08+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC is an incredible company.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish TSMC on strong N2 customer adoption; quality endorsement.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Also, it’s shocking that TSMC has already gathered 15 customers for N2.\n\nTSMC is an incredible company.", "reply_text": null, "parent_text": null, "quoted_text": "Reasons for a bullish outlook on ASML:\n\nAhmad Khan, President of KLA’s Semiconductor Products &amp; Solutions Division, recently stated at a Goldman Sachs event that TSMC has secured about 15 customers for its first-generation 2nm (N2) process node, of which around 10 are https://t.co/pysGHzPcb5", "parent_author": null, "ret_m1d": -0.023165933043046705, "ret_signed_m1d": -0.023165933043046705, "alpha_spy_m1d": -0.018457202118474347, "alpha_c_m1d": -0.0031143752880989917, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03474908933571563, "ret_signed_p1d": 0.03474908933571563, "alpha_spy_p1d": 0.04019256417552319, "alpha_c_p1d": 0.036207928649596544, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.003860988840507895, "ret_signed_p1w": 0.003860988840507895, "alpha_spy_p1w": 0.008599728098376058, "alpha_c_p1w": 0.002340101583426124, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.14285725129779747, "ret_signed_p1m": 0.14285725129779747, "alpha_spy_p1m": 0.13618407914357933, "alpha_c_p1m": 0.07106293500341843, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.1079279742594419, "ret_signed_p3m": 0.1079279742594419, "alpha_spy_p3m": 0.0934867860362214, "alpha_c_p3m": 0.030080895358783222, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 0.45355588428135096, "ret_signed_p6m": 0.45355588428135096, "alpha_spy_p6m": 0.44466007036100064, "alpha_c_p6m": 0.21782653113828543, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.177, -0.1732, -0.1693, -0.1847, -0.1655, -0.1655, -0.1617, -0.1655, -0.1693, -0.1693, -0.1617, -0.154, -0.154, -0.1578, -0.1463, -0.1309, -0.1309, -0.1117, -0.1155, -0.1309, -0.1194, -0.1194, -0.1194, -0.1194, -0.127, -0.1117, -0.1078, -0.1078, -0.127, -0.1155, -0.1347, -0.0924, -0.0963, -0.0924, -0.0924, -0.0771, -0.0963, -0.0924, -0.0924, -0.0886, -0.127, -0.1155, -0.127, -0.1001, -0.0963, -0.0847, -0.1078, -0.1078, -0.104, -0.1078, -0.1078, -0.1078, -0.0924, -0.0924, -0.0771, -0.0578, -0.0463, -0.0309, -0.0347, -0.0116, -0.0232, -0.0077, -0.0232, 0.0, 0.0347, 0.0347, 0.0193, 0.0039, 0.0039, 0.0077, 0.0232, 0.0541, 0.0811, 0.0811, 0.1081, 0.0927, 0.112, 0.112, 0.0927, 0.1004, 0.1313, 0.1467, 0.1197, 0.1429, 0.1429, 0.1274, 0.1197, 0.1197, 0.1429, 0.139, 0.1622, 0.1622, 0.1583, 0.166, 0.1622, 0.1274, 0.1313, 0.1274, 0.139, 0.1313, 0.139, 0.1274, 0.1042, 0.1158, 0.0849, 0.0772, 0.1236, 0.0695, 0.0618, 0.0927, 0.112, 0.1081, 0.112, 0.0888, 0.1042, 0.1197, 0.1158, 0.1274, 0.1544, 0.1429, 0.1622, 0.1389, 0.1467, 0.1234, 0.1118, 0.1079, 0.1079, 0.1079, 0.135, 0.1544, 0.1583, 0.1583, 0.1699, 0.1854, 0.1777, 0.2009, 0.2009, 0.228, 0.2939, 0.321, 0.2977, 0.3055, 0.3016, 0.3094, 0.3249, 0.3249, 0.3094, 0.3481, 0.3636, 0.3752, 0.3481, 0.3636, 0.3714, 0.3597, 0.3791, 0.4101, 0.3985, 0.3752, 0.3675, 0.3946, 0.383, 0.3675, 0.3791, 0.4062, 0.4566, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4721, 0.5224, 0.5612, 0.5457, 0.5457, 0.5302, 0.4992, 0.445, 0.4721, 0.4643, 0.4023, 0.4333, 0.5031, 0.4605, 0.445, 0.4295, 0.4536]}
{"unit_id": "quote:1955766698742833457", "ticker_idx": 0, "ts": "2025-08-13T23:01:32+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Pricing power now favors AMD. AMD's revenue share sits above its unit share, while Intel's revenue share trails its unit share. This signals AMD is winning higher-value sockets and holding better ASPs", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses thesis that AMD has pricing power advantage over INTC in servers, winning higher-value sockets while Intel discounts.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Pricing power now favors AMD. AMD’s revenue share sits above its unit share, while Intel’s revenue share trails its unit share. This signals AMD is winning higher-value sockets and holding better ASPs, as Intel discounts during its platform transition.\"\n\n👀\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": "The big takes from the latest Mercury Microprocessor Report on @Intel @AMD  ( $INTC $AMD ) market share and pricing dynamics in servers: \n\nPricing power now favors AMD. AMD’s revenue share sits above its unit share, while Intel’s revenue share trails its unit share. This signals", "parent_author": null, "ret_m1d": -0.05135018606836639, "ret_signed_m1d": -0.05135018606836639, "alpha_spy_m1d": -0.04793881202176342, "alpha_c_m1d": -0.04743349264485908, "bench_spy_m1d": -0.0034113740466029663, "bench_c_m1d": -0.003916693423507311, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01881575347118425, "ret_signed_p1d": -0.01881575347118425, "alpha_spy_p1d": -0.01890871892602297, "alpha_c_p1d": -0.02093997565998873, "bench_spy_p1d": 9.296545483872265e-05, "bench_c_p1d": 0.002124222188804481, "ret_p1w": -0.10421863903878725, "ret_signed_p1w": -0.10421863903878725, "alpha_spy_p1w": -0.09370514329906765, "alpha_c_p1w": -0.06332639662587569, "bench_spy_p1w": -0.010513495739719603, "bench_c_p1w": -0.04089224241291156, "ret_p1m": -0.15589415619483094, "ret_signed_p1m": -0.15589415619483094, "alpha_spy_p1m": -0.17564950624222042, "alpha_c_p1m": -0.16468987482937847, "bench_spy_p1m": 0.019755350047389486, "bench_c_p1m": 0.008795718634547534, "ret_p3m": 0.3229584543430666, "ret_signed_p3m": 0.3229584543430666, "alpha_spy_p3m": 0.26335251715297137, "alpha_c_p3m": 0.13207291807501598, "bench_spy_p3m": 0.059605937190095215, "bench_c_p3m": 0.1908855362680506, "ret_p6m": 0.043813045826259556, "ret_signed_p6m": 0.043813045826259556, "alpha_spy_p6m": -0.012966417016669896, "alpha_c_p6m": -0.22495928258787945, "bench_spy_p6m": 0.05677946284292945, "bench_c_p6m": 0.268772328414139, "price_path": [-0.3647, -0.3778, -0.3845, -0.3924, -0.3997, -0.4019, -0.4019, -0.3788, -0.388, -0.3871, -0.3996, -0.3784, -0.3639, -0.357, -0.3727, -0.37, -0.3399, -0.3317, -0.3431, -0.3574, -0.3701, -0.3147, -0.3108, -0.3125, -0.3125, -0.3046, -0.2974, -0.2494, -0.2224, -0.2209, -0.2202, -0.2306, -0.262, -0.2489, -0.2522, -0.2522, -0.2691, -0.2527, -0.2495, -0.2183, -0.2061, -0.207, -0.1562, -0.132, -0.1302, -0.1487, -0.1487, -0.161, -0.1397, -0.1209, -0.0973, -0.0583, -0.0378, -0.0266, -0.044, -0.069, -0.0414, -0.0548, -0.1155, -0.0652, -0.0632, -0.0658, -0.0514, 0.0, -0.0188, -0.0375, -0.0449, -0.0969, -0.1042, -0.1123, -0.0903, -0.1142, -0.0965, -0.0938, -0.0859, -0.1182, -0.1182, -0.1198, -0.1209, -0.1227, -0.1805, -0.179, -0.1551, -0.1349, -0.1559, -0.1402, -0.1261, -0.1299, -0.137, -0.1437, -0.1466, -0.1336, -0.1275, -0.1276, -0.1255, -0.1353, -0.125, -0.1227, -0.1107, -0.0797, -0.1071, 0.1046, 0.1469, 0.2773, 0.2628, 0.1653, 0.1735, 0.1826, 0.2938, 0.2719, 0.2639, 0.3044, 0.2907, 0.2484, 0.2742, 0.3714, 0.408, 0.399, 0.4333, 0.3818, 0.3888, 0.4079, 0.3559, 0.3899, 0.2889, 0.2663, 0.323, 0.2879, 0.4038, 0.3445, 0.3383, 0.3042, 0.2487, 0.2122, 0.1171, 0.105, 0.1661, 0.1177, 0.1617, 0.1617, 0.1795, 0.1916, 0.1671, 0.1799, 0.1711, 0.1819, 0.1989, 0.2017, 0.2006, 0.2007, 0.1429, 0.1256, 0.1342, 0.0742, 0.0902, 0.1573, 0.1655, 0.1653, 0.166, 0.166, 0.1658, 0.1691, 0.1677, 0.1613, 0.1613, 0.2117, 0.1988, 0.1623, 0.1388, 0.1099, 0.1017, 0.1262, 0.1982, 0.2124, 0.2359, 0.2571, 0.2571, 0.2576, 0.3545, 0.3758, 0.4081, 0.3627, 0.3666, 0.3705, 0.3674, 0.2836, 0.3354, 0.3128, 0.0855, 0.0438]}
{"unit_id": "quote:1955766698742833457", "ticker_idx": 1, "ts": "2025-08-13T23:01:32+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel's revenue share trails its unit share... Intel discounts during its platform transition", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses thesis that AMD has pricing power advantage over INTC in servers, winning higher-value sockets while Intel discounts.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Pricing power now favors AMD. AMD’s revenue share sits above its unit share, while Intel’s revenue share trails its unit share. This signals AMD is winning higher-value sockets and holding better ASPs, as Intel discounts during its platform transition.\"\n\n👀\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": "The big takes from the latest Mercury Microprocessor Report on @Intel @AMD  ( $INTC $AMD ) market share and pricing dynamics in servers: \n\nPricing power now favors AMD. AMD’s revenue share sits above its unit share, while Intel’s revenue share trails its unit share. This signals", "parent_author": null, "ret_m1d": -0.018451838887577954, "ret_signed_m1d": 0.018451838887577954, "alpha_spy_m1d": 0.015040464840974987, "alpha_c_m1d": 0.014535145464070642, "bench_spy_m1d": -0.0034113740466029663, "bench_c_m1d": -0.003916693423507311, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07380744138958684, "ret_signed_p1d": -0.07380744138958684, "alpha_spy_p1d": -0.07371447593474811, "alpha_c_p1d": -0.07168321920078236, "bench_spy_p1d": 9.296545483872265e-05, "bench_c_p1d": 0.002124222188804481, "ret_p1w": 0.059406014534821194, "ret_signed_p1w": -0.059406014534821194, "alpha_spy_p1w": -0.0699195102745408, "alpha_c_p1w": -0.10029825694773276, "bench_spy_p1w": -0.010513495739719603, "bench_c_p1w": -0.04089224241291156, "ret_p1m": 0.10756081777017212, "ret_signed_p1m": -0.10756081777017212, "alpha_spy_p1m": -0.08780546772278264, "alpha_c_p1m": -0.09876509913562459, "bench_spy_p1m": 0.019755350047389486, "bench_c_p1m": 0.008795718634547534, "ret_p3m": 0.7304231301137138, "ret_signed_p3m": -0.7304231301137138, "alpha_spy_p3m": -0.6708171929236186, "alpha_c_p3m": -0.5395375938456632, "bench_spy_p3m": 0.059605937190095215, "bench_c_p3m": 0.1908855362680506, "ret_p6m": 1.1710172443378055, "ret_signed_p6m": -1.1710172443378055, "alpha_spy_p6m": -1.114237781494876, "alpha_c_p6m": -0.9022449159236665, "bench_spy_p6m": 0.05677946284292945, "bench_c_p6m": 0.268772328414139, "price_path": [-0.0252, -0.0378, -0.0428, -0.0689, -0.0752, -0.0977, -0.0977, -0.0752, -0.0833, -0.0887, -0.1202, -0.1116, -0.0869, -0.0887, -0.1004, -0.0972, -0.0783, -0.0063, -0.0693, -0.0653, -0.0936, -0.0666, -0.0639, -0.0329, -0.0329, -0.0513, -0.0464, 0.0149, -0.0009, 0.0126, 0.0212, 0.0081, 0.0284, -0.0153, 0.0122, 0.0122, -0.0099, 0.0617, 0.0549, 0.072, 0.0545, 0.0486, 0.0315, 0.0212, 0.0261, 0.0396, 0.0468, 0.0459, 0.0572, 0.0185, -0.0684, -0.0693, -0.0815, -0.0846, -0.1089, -0.131, -0.1224, -0.0914, -0.0815, -0.1103, -0.1022, -0.0707, -0.0185, 0.0, 0.0738, 0.1053, 0.0648, 0.1391, 0.0594, 0.0576, 0.1161, 0.1049, 0.0959, 0.1184, 0.122, 0.0959, 0.0959, 0.0896, 0.0801, 0.1076, 0.1022, 0.1017, 0.0999, 0.1148, 0.1076, 0.0837, 0.1148, 0.1373, 0.1206, 0.3758, 0.3312, 0.2943, 0.3204, 0.405, 0.5297, 0.5977, 0.5518, 0.5099, 0.6175, 0.6787, 0.6575, 0.6467, 0.6728, 0.6845, 0.7012, 0.6368, 0.6751, 0.6035, 0.6719, 0.658, 0.6656, 0.7147, 0.7156, 0.6616, 0.7174, 0.7228, 0.7795, 0.869, 0.8605, 0.8074, 0.7997, 0.7777, 0.6665, 0.7273, 0.676, 0.716, 0.7304, 0.7048, 0.7052, 0.6161, 0.5986, 0.5621, 0.545, 0.5801, 0.5131, 0.5527, 0.6107, 0.6125, 0.6566, 0.6566, 0.8254, 0.8006, 0.9563, 0.9694, 0.8227, 0.8636, 0.8137, 0.8227, 0.8353, 0.7781, 0.7016, 0.6881, 0.6791, 0.6224, 0.6328, 0.6571, 0.6368, 0.6359, 0.6274, 0.6274, 0.6292, 0.6508, 0.6787, 0.6607, 0.6607, 0.7723, 0.7718, 0.802, 0.9185, 0.8501, 1.05, 0.9829, 1.1283, 1.1926, 1.1746, 1.1134, 1.1134, 1.1854, 1.4415, 1.4446, 1.0284, 0.9122, 0.977, 1.1953, 1.1899, 1.0914, 1.1967, 1.2165, 1.1872, 1.171]}
{"unit_id": "orig:1953800307802354099", "ticker_idx": 0, "ts": "2025-08-08T12:47:47+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our Buy rating and $200 price objective", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA reiterates Buy on AMD, $200 PT, raises 2025-27 EPS estimates; implies bearish on Intel losing CPU share.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: AMD - Reiterate Buy, Raising Estimates, China AI Not Essential for Growth\n\nDespite a Q2 revenue beat of $300M (4%) and Q3 revenue guidance that was $450M (5%) above consensus, AMD shares are trading down in the after-hours on what we believe is ignorable earnings-related noise: 1) No China GPU license yet. While this did not impact guidance, it may have disappointed investors who were modeling a ~$500M/quarter China contribution in Q3/Q4. 2) Much of the Q2 beat was driven by gaming consoles. The Data Center segment was largely in-line with estimates, with MI GPU sales slightly below $1B in Q2. 3) The stock has surged over 80% during the quarter, including a 15% rise since July 15th on news of a potential China license.\n\nIn our view, most of the above is noise that overlooks AMD's increasing participation in the ~$1T AI market and its opportunity to continue gaining PC/server CPU share against a struggling Intel (INTC). We are removing the China MI contribution from our estimates and raising our annual revenue forecasts for 2025/26/27 by about 3% each, leading to adjusted Earnings Per Share (EPS) estimates of $3.85/$6.00/$7.50, up 1%/7%/10% respectively. Longer-term, we see $9-$10 in EPS power for a company with 15-20% top-line growth. We are adjusting our 2026 target Price-to-Earnings (P/E) multiple to a slightly lower 33x (from 36x) to reflect near-term China uncertainty, but we maintain our Buy rating and $200 price objective.\n\nPositives: Improving Pipeline in a Large TAM, Strong FCF\nAMD continues to execute successfully on its AI GPU pipeline. The MI355X (Blackwell competitor) is on track for a production ramp with multiple customers in the second half of the year, and the MI400 series (Rubin competitor) and Helios rack system are also on track for a 2026 launch. AMD now expects Q3 GPU revenue of ~$1.7B without the ~$800M China loss. This implies a ~$500M raise to the rest-of-world outlook compared to the pre-sanction consensus of $2.0B. Server CPUs also continue to perform well, taking share from incumbent Intel (INTC) in both cloud and enterprise. Overall, we expect AMD to be a company that consistently generates 20%+ Free Cash Flow (FCF) margins long-term.\n\nRisks: GPU Competition, High Opex, China Uncertainty\n(1) Meanwhile, NVIDIA (NVDA) is now ramping its Blackwell Ultra systems and will launch its next-gen Rubin platform in 2026. While AMD has made significant progress, we see NVIDIA's hardware as generally 0.5-1.0 year ahead, especially in its system/rack approach, where AMD still lacks experience in large-scale deployments. (2) AMD will need to maintain a high level of operating expenses (opex) to support this effort. (3) There is currently no guarantee on when, or if, AMD will receive a license to resume shipments to China, and even if a license is granted, it could take another 8-9 months for the supply chain to fully normalize. By that point, AMD's current MI308 product may no longer be competitive in the Chinese market.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0020838192972812974, "ret_signed_m1d": -0.0020838192972812974, "alpha_spy_m1d": 0.00565342245051581, "alpha_c_m1d": 0.006126594054887025, "bench_spy_m1d": -0.007737241747797108, "bench_c_m1d": -0.008210413352168322, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0027783962884999625, "ret_signed_p1d": -0.0027783962884999625, "alpha_spy_p1d": -0.0008009530520851627, "alpha_c_p1d": -0.002335472406851502, "bench_spy_p1d": -0.0019774432364148, "bench_c_p1d": -0.0004429238816484604, "ret_p1w": 0.027494791334992996, "ret_signed_p1w": 0.027494791334992996, "alpha_spy_p1w": 0.017670173258302446, "alpha_c_p1w": 0.020272326636193272, "bench_spy_p1w": 0.00982461807669055, "bench_c_p1w": 0.007222464698799724, "ret_p1m": -0.12358179858521445, "ret_signed_p1m": -0.12358179858521445, "alpha_spy_p1m": -0.14186546866189997, "alpha_c_p1m": -0.13376810924411975, "bench_spy_p1m": 0.01828367007668552, "bench_c_p1m": 0.010186310658905295, "ret_p3m": 0.48373463025389785, "ret_signed_p3m": 0.48373463025389785, "alpha_spy_p3m": 0.4173819909757699, "alpha_c_p3m": 0.258987522895177, "bench_spy_p3m": 0.06635263927812796, "bench_c_p3m": 0.22474710735872083, "ret_p6m": 0.4255036588503496, "ret_signed_p6m": 0.4255036588503496, "alpha_spy_p6m": 0.32785689358329484, "alpha_c_p6m": 0.0313039071672232, "bench_spy_p6m": 0.09764676526705474, "bench_c_p6m": 0.3941997516831264, "price_path": [-0.349, -0.3186, -0.3344, -0.3218, -0.3358, -0.343, -0.3514, -0.3592, -0.3615, -0.3615, -0.3369, -0.3467, -0.3457, -0.3591, -0.3365, -0.321, -0.3136, -0.3303, -0.3274, -0.2954, -0.2866, -0.2988, -0.3141, -0.3276, -0.2684, -0.2643, -0.2661, -0.2661, -0.2577, -0.2499, -0.1987, -0.1699, -0.1683, -0.1676, -0.1786, -0.2121, -0.1982, -0.2017, -0.2017, -0.2197, -0.2022, -0.1988, -0.1655, -0.1525, -0.1535, -0.0993, -0.0734, -0.0715, -0.0913, -0.0912, -0.1044, -0.0817, -0.0616, -0.0364, 0.0052, 0.0271, 0.0391, 0.0205, -0.0061, 0.0233, 0.009, -0.0558, -0.0021, 0.0, -0.0028, 0.0127, 0.0675, 0.0474, 0.0275, 0.0196, -0.0359, -0.0438, -0.0524, -0.0289, -0.0544, -0.0355, -0.0326, -0.0242, -0.0586, -0.0586, -0.0604, -0.0615, -0.0635, -0.1251, -0.1236, -0.0981, -0.0765, -0.0989, -0.0821, -0.0671, -0.0712, -0.0787, -0.0859, -0.089, -0.0751, -0.0687, -0.0688, -0.0665, -0.077, -0.066, -0.0635, -0.0506, -0.0175, -0.0468, 0.1792, 0.2243, 0.3635, 0.3481, 0.2439, 0.2527, 0.2624, 0.3811, 0.3577, 0.3492, 0.3925, 0.3778, 0.3327, 0.3602, 0.464, 0.5031, 0.4935, 0.53, 0.4751, 0.4825, 0.503, 0.4474, 0.4837, 0.3759, 0.3518, 0.4122, 0.3749, 0.4986, 0.4353, 0.4286, 0.3922, 0.333, 0.294, 0.1925, 0.1796, 0.2448, 0.1932, 0.2401, 0.2401, 0.2591, 0.2721, 0.2459, 0.2596, 0.2502, 0.2617, 0.2799, 0.2828, 0.2817, 0.2817, 0.2201, 0.2016, 0.2108, 0.1467, 0.1638, 0.2354, 0.2442, 0.2439, 0.2447, 0.2447, 0.2444, 0.248, 0.2465, 0.2396, 0.2396, 0.2935, 0.2797, 0.2407, 0.2157, 0.1848, 0.176, 0.2022, 0.2791, 0.2943, 0.3193, 0.3419, 0.3419, 0.3424, 0.4459, 0.4687, 0.5031, 0.4547, 0.4588, 0.463, 0.4597, 0.3703, 0.4255]}
{"unit_id": "orig:1953800307802354099", "ticker_idx": 1, "ts": "2025-08-08T12:47:47+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "opportunity to continue gaining PC/server CPU share against a struggling Intel (INTC)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA reiterates Buy on AMD, $200 PT, raises 2025-27 EPS estimates; implies bearish on Intel losing CPU share.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: AMD - Reiterate Buy, Raising Estimates, China AI Not Essential for Growth\n\nDespite a Q2 revenue beat of $300M (4%) and Q3 revenue guidance that was $450M (5%) above consensus, AMD shares are trading down in the after-hours on what we believe is ignorable earnings-related noise: 1) No China GPU license yet. While this did not impact guidance, it may have disappointed investors who were modeling a ~$500M/quarter China contribution in Q3/Q4. 2) Much of the Q2 beat was driven by gaming consoles. The Data Center segment was largely in-line with estimates, with MI GPU sales slightly below $1B in Q2. 3) The stock has surged over 80% during the quarter, including a 15% rise since July 15th on news of a potential China license.\n\nIn our view, most of the above is noise that overlooks AMD's increasing participation in the ~$1T AI market and its opportunity to continue gaining PC/server CPU share against a struggling Intel (INTC). We are removing the China MI contribution from our estimates and raising our annual revenue forecasts for 2025/26/27 by about 3% each, leading to adjusted Earnings Per Share (EPS) estimates of $3.85/$6.00/$7.50, up 1%/7%/10% respectively. Longer-term, we see $9-$10 in EPS power for a company with 15-20% top-line growth. We are adjusting our 2026 target Price-to-Earnings (P/E) multiple to a slightly lower 33x (from 36x) to reflect near-term China uncertainty, but we maintain our Buy rating and $200 price objective.\n\nPositives: Improving Pipeline in a Large TAM, Strong FCF\nAMD continues to execute successfully on its AI GPU pipeline. The MI355X (Blackwell competitor) is on track for a production ramp with multiple customers in the second half of the year, and the MI400 series (Rubin competitor) and Helios rack system are also on track for a 2026 launch. AMD now expects Q3 GPU revenue of ~$1.7B without the ~$800M China loss. This implies a ~$500M raise to the rest-of-world outlook compared to the pre-sanction consensus of $2.0B. Server CPUs also continue to perform well, taking share from incumbent Intel (INTC) in both cloud and enterprise. Overall, we expect AMD to be a company that consistently generates 20%+ Free Cash Flow (FCF) margins long-term.\n\nRisks: GPU Competition, High Opex, China Uncertainty\n(1) Meanwhile, NVIDIA (NVDA) is now ramping its Blackwell Ultra systems and will launch its next-gen Rubin platform in 2026. While AMD has made significant progress, we see NVIDIA's hardware as generally 0.5-1.0 year ahead, especially in its system/rack approach, where AMD still lacks experience in large-scale deployments. (2) AMD will need to maintain a high level of operating expenses (opex) to support this effort. (3) There is currently no guarantee on when, or if, AMD will receive a license to resume shipments to China, and even if a license is granted, it could take another 8-9 months for the supply chain to fully normalize. By that point, AMD's current MI308 product may no longer be competitive in the Chinese market.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009022571342962693, "ret_signed_m1d": 0.009022571342962693, "alpha_spy_m1d": 0.0012853295951655852, "alpha_c_m1d": 0.000812157990794371, "bench_spy_m1d": -0.007737241747797108, "bench_c_m1d": -0.008210413352168322, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03508766059253432, "ret_signed_p1d": -0.03508766059253432, "alpha_spy_p1d": -0.03706510382894912, "alpha_c_p1d": -0.03553058447418278, "bench_spy_p1d": -0.0019774432364148, "bench_c_p1d": -0.0004429238816484604, "ret_p1w": 0.2310776203862004, "ret_signed_p1w": -0.2310776203862004, "alpha_spy_p1w": -0.22125300230950984, "alpha_c_p1w": -0.22385515568740066, "bench_spy_p1w": 0.00982461807669055, "bench_c_p1w": 0.007222464698799724, "ret_p1m": 0.22706759930115505, "ret_signed_p1m": -0.22706759930115505, "alpha_spy_p1m": -0.20878392922446953, "alpha_c_p1m": -0.21688128864224976, "bench_spy_p1m": 0.01828367007668552, "bench_c_p1m": 0.010186310658905295, "ret_p3m": 0.9238095037776974, "ret_signed_p3m": -0.9238095037776974, "alpha_spy_p3m": -0.8574568644995695, "alpha_c_p3m": -0.6990623964189766, "bench_spy_p3m": 0.06635263927812796, "bench_c_p3m": 0.22474710735872083, "ret_p6m": 1.4466165166251002, "ret_signed_p6m": -1.4466165166251002, "alpha_spy_p6m": -1.3489697513580454, "alpha_c_p6m": -1.0524167649419738, "bench_spy_p6m": 0.09764676526705474, "bench_c_p6m": 0.3941997516831264, "price_path": [0.1308, 0.0787, 0.0802, 0.0857, 0.0717, 0.0662, 0.0371, 0.0301, 0.005, 0.005, 0.0301, 0.0211, 0.015, -0.0201, -0.0105, 0.017, 0.015, 0.002, 0.0055, 0.0266, 0.1068, 0.0366, 0.0411, 0.0095, 0.0396, 0.0426, 0.0772, 0.0772, 0.0566, 0.0622, 0.1303, 0.1128, 0.1278, 0.1373, 0.1228, 0.1454, 0.0967, 0.1273, 0.1273, 0.1028, 0.1825, 0.1749, 0.194, 0.1744, 0.1679, 0.1489, 0.1373, 0.1429, 0.1579, 0.1659, 0.1649, 0.1774, 0.1343, 0.0376, 0.0366, 0.0231, 0.0195, -0.0075, -0.0321, -0.0226, 0.012, 0.0231, -0.009, 0.0, 0.0351, 0.0932, 0.1138, 0.196, 0.2311, 0.186, 0.2687, 0.1799, 0.1779, 0.2431, 0.2306, 0.2206, 0.2456, 0.2496, 0.2206, 0.2206, 0.2135, 0.203, 0.2336, 0.2276, 0.2271, 0.2251, 0.2416, 0.2336, 0.207, 0.2416, 0.2667, 0.2481, 0.5323, 0.4827, 0.4416, 0.4707, 0.5649, 0.7038, 0.7794, 0.7283, 0.6817, 0.8015, 0.8697, 0.8461, 0.8341, 0.8632, 0.8762, 0.8947, 0.8231, 0.8657, 0.786, 0.8622, 0.8466, 0.8551, 0.9098, 0.9108, 0.8506, 0.9128, 0.9188, 0.982, 1.0817, 1.0722, 1.013, 1.0045, 0.9799, 0.8561, 0.9238, 0.8667, 0.9113, 0.9273, 0.8987, 0.8992, 0.8, 0.7805, 0.7398, 0.7208, 0.7599, 0.6852, 0.7293, 0.794, 0.796, 0.8451, 0.8451, 1.0331, 1.0055, 1.1789, 1.1935, 1.0301, 1.0757, 1.0201, 1.0301, 1.0441, 0.9805, 0.8952, 0.8802, 0.8702, 0.807, 0.8185, 0.8456, 0.8231, 0.8221, 0.8125, 0.8125, 0.8145, 0.8386, 0.8697, 0.8496, 0.8496, 0.9739, 0.9734, 1.007, 1.1368, 1.0607, 1.2832, 1.2085, 1.3704, 1.4421, 1.4221, 1.3539, 1.3539, 1.4341, 1.7193, 1.7228, 1.2591, 1.1298, 1.202, 1.4451, 1.4391, 1.3293, 1.4466]}
{"unit_id": "orig:1946462358559449109", "ticker_idx": 0, "ts": "2025-07-19T06:49:24+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML expects the Chinese market to account for over 25% of its total sales, up from a previous expectation in the low 20% range", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises China WFE TAM to $4.03B for 2025 on AI-driven logic foundry demand; bullish on ASML, KLA, AMAT, Naura, AMEC.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "China Semiconductor Equipment: Strong China WFE Demand Expected in H2 2025, Anticipating Growth\nMorgan Stanley (C. Chan, 25/07/17)\n\nMorgan Stanley believes that China's semiconductor equipment (WFE) market will continue to grow, driven by strong logic foundry demand and increasing localization needs, with demand expected to rise significantly in the second half of 2025. Based on surveys in Taiwan and Beijing, China's WFE demand is anticipated to strengthen remarkably in H2 2025, with AI-driven logic foundry being the primary growth engine. Morgan Stanley has also adjusted its 2025/2026 outlook for the Chinese WFE market, further raising its Total Addressable Market (TAM) and forecasts.\n\nRising Demand in China's WFE Market: According to Morgan Stanley's survey, China's logic foundry spending in H2 2025 could significantly exceed that of H1. Morgan Stanley expects China's lithography equipment demand to be stronger than anticipated in Q2. China's AI demand far exceeds the current AI chip supply, which is expected to drive more logic foundry spending.\n\nGrowth in China's Local Equipment Market Share: Morgan Stanley points out that approximately 25% of China's foundry CapEx in 2025 is expected to be allocated to local WFE equipment, an increase from 20% in 2024. Chinese WFE manufacturers are expanding their product portfolios; for example, Naura has launched ion implantation equipment in Q1 2025. AMEC expects its product pipeline to cover 60% of wafer processes in the next 5-10 years.\n\nMaturity Node Oversupply: Although semiconductor equipment import data from January to May indicates a decline in demand for China's mature nodes, Morgan Stanley's industry survey suggests that the government's willingness to promote localization may override economic rationality in H2 2025. Despite some government-supported projects facing limited demand and low capacity utilization rates, government policies to promote localization of the semiconductor industry may continue to support WFE demand.\n\n2025 China WFE Market Outlook: Morgan Stanley has revised its 2025 China WFE market forecast upwards from $3.66 billion (down 12% YoY) to $4.03 billion (down 3% YoY). Although the increase in 2025 primarily comes from Taiwan and South Korea, China remains the largest region for global WFE demand. The Chinese WFE market is expected to grow to $3.62 billion by 2026, despite a 10% YoY decline.\n\nForeign Equipment Suppliers Optimistic on China Market: According to Morgan Stanley's analysis, foreign equipment suppliers have improved their outlook on the Chinese market. For instance, ASML expects the Chinese market to account for over 25% of its total sales, up from a previous expectation in the low 20% range. KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015061663527752067, "ret_signed_m1d": 0.015061663527752067, "alpha_spy_m1d": 0.01695425001791695, "alpha_c_m1d": 0.01575007941566997, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.034769939171927944, "ret_signed_p1d": -0.034769939171927944, "alpha_spy_p1d": -0.03491306033515251, "alpha_c_p1d": -0.01707692376386738, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.009399590799242796, "ret_signed_p1w": 0.009399590799242796, "alpha_spy_p1w": -0.0035940878855227343, "alpha_c_p1w": 0.006301561240879661, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.031891338518251544, "ret_signed_p1m": 0.031891338518251544, "alpha_spy_p1m": 0.014333249876912868, "alpha_c_p1m": 0.030652126694906245, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.4095930024797507, "ret_signed_p3m": 0.4095930024797507, "alpha_spy_p3m": 0.3559937352757694, "alpha_c_p3m": 0.22787818410910976, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.7745160888057321, "ret_signed_p6m": 0.7745160888057321, "alpha_spy_p6m": 0.6648111719476744, "alpha_c_p6m": 0.422589203989312, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.0721, -0.0549, -0.0564, -0.0635, -0.0627, -0.0667, -0.0667, -0.0244, -0.0277, -0.0362, -0.0311, 0.0103, 0.0043, 0.0686, 0.0955, 0.1001, 0.0931, 0.071, 0.0614, 0.0575, 0.0651, 0.0524, 0.0296, 0.0479, 0.0673, 0.0588, 0.0663, 0.0477, 0.0354, 0.0465, 0.0453, 0.0559, 0.0697, 0.0897, 0.0968, 0.1059, 0.0841, 0.0649, 0.0737, 0.0671, 0.0655, 0.0465, 0.0421, 0.0715, 0.1078, 0.1128, 0.0854, 0.0936, 0.0857, 0.0707, 0.0857, 0.0849, 0.0566, 0.0761, 0.0878, 0.0892, 0.1056, 0.0958, 0.1013, 0.1314, 0.0027, 0.0418, 0.0151, 0.0, -0.0348, -0.028, -0.0131, -0.0308, 0.0094, 0.0009, 0.0081, -0.015, -0.0441, -0.0322, -0.0407, -0.0475, -0.0189, -0.0089, 0.0073, 0.0184, 0.0292, 0.0329, 0.0229, 0.0261, 0.0319, 0.0253, 0.0194, 0.0382, 0.039, 0.0452, 0.0606, 0.0512, 0.0227, 0.0219, -0.0084, 0.0046, 0.0404, 0.0606, 0.087, 0.0976, 0.0928, 0.1021, 0.1082, 0.1718, 0.1774, 0.184, 0.2756, 0.2751, 0.3026, 0.3171, 0.2981, 0.304, 0.3058, 0.3262, 0.3304, 0.3518, 0.4099, 0.4139, 0.4416, 0.4054, 0.3685, 0.3653, 0.3076, 0.3559, 0.3601, 0.4025, 0.4096, 0.4069, 0.4443, 0.4295, 0.3979, 0.4261, 0.4369, 0.4589, 0.4604, 0.4797, 0.5088, 0.4776, 0.4911, 0.4774, 0.4591, 0.4409, 0.4015, 0.427, 0.4274, 0.4353, 0.4187, 0.4068, 0.4105, 0.3931, 0.4269, 0.432, 0.3422, 0.3821, 0.3846, 0.4634, 0.4439, 0.4539, 0.4914, 0.5114, 0.5505, 0.5407, 0.5315, 0.5502, 0.5336, 0.5225, 0.5143, 0.4892, 0.4983, 0.4623, 0.4066, 0.4367, 0.451, 0.442, 0.4536, 0.4468, 0.4468, 0.4468, 0.4594, 0.4781, 0.4829, 0.4829, 0.5873, 0.695, 0.7076, 0.6928, 0.6303, 0.7407, 0.7484, 0.7745]}
{"unit_id": "orig:1946462358559449109", "ticker_idx": 1, "ts": "2025-07-19T06:49:24+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises China WFE TAM to $4.03B for 2025 on AI-driven logic foundry demand; bullish on ASML, KLA, AMAT, Naura, AMEC.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "China Semiconductor Equipment: Strong China WFE Demand Expected in H2 2025, Anticipating Growth\nMorgan Stanley (C. Chan, 25/07/17)\n\nMorgan Stanley believes that China's semiconductor equipment (WFE) market will continue to grow, driven by strong logic foundry demand and increasing localization needs, with demand expected to rise significantly in the second half of 2025. Based on surveys in Taiwan and Beijing, China's WFE demand is anticipated to strengthen remarkably in H2 2025, with AI-driven logic foundry being the primary growth engine. Morgan Stanley has also adjusted its 2025/2026 outlook for the Chinese WFE market, further raising its Total Addressable Market (TAM) and forecasts.\n\nRising Demand in China's WFE Market: According to Morgan Stanley's survey, China's logic foundry spending in H2 2025 could significantly exceed that of H1. Morgan Stanley expects China's lithography equipment demand to be stronger than anticipated in Q2. China's AI demand far exceeds the current AI chip supply, which is expected to drive more logic foundry spending.\n\nGrowth in China's Local Equipment Market Share: Morgan Stanley points out that approximately 25% of China's foundry CapEx in 2025 is expected to be allocated to local WFE equipment, an increase from 20% in 2024. Chinese WFE manufacturers are expanding their product portfolios; for example, Naura has launched ion implantation equipment in Q1 2025. AMEC expects its product pipeline to cover 60% of wafer processes in the next 5-10 years.\n\nMaturity Node Oversupply: Although semiconductor equipment import data from January to May indicates a decline in demand for China's mature nodes, Morgan Stanley's industry survey suggests that the government's willingness to promote localization may override economic rationality in H2 2025. Despite some government-supported projects facing limited demand and low capacity utilization rates, government policies to promote localization of the semiconductor industry may continue to support WFE demand.\n\n2025 China WFE Market Outlook: Morgan Stanley has revised its 2025 China WFE market forecast upwards from $3.66 billion (down 12% YoY) to $4.03 billion (down 3% YoY). Although the increase in 2025 primarily comes from Taiwan and South Korea, China remains the largest region for global WFE demand. The Chinese WFE market is expected to grow to $3.62 billion by 2026, despite a 10% YoY decline.\n\nForeign Equipment Suppliers Optimistic on China Market: According to Morgan Stanley's analysis, foreign equipment suppliers have improved their outlook on the Chinese market. For instance, ASML expects the Chinese market to account for over 25% of its total sales, up from a previous expectation in the low 20% range. KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007080697005804493, "ret_signed_m1d": -0.007080697005804493, "alpha_spy_m1d": -0.00518811051563961, "alpha_c_m1d": -0.006392281117886589, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04856264333444471, "ret_signed_p1d": -0.04856264333444471, "alpha_spy_p1d": -0.04870576449766928, "alpha_c_p1d": -0.030869627926384147, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.015579655794265235, "ret_signed_p1w": -0.015579655794265235, "alpha_spy_p1w": -0.028573334479030765, "alpha_c_p1w": -0.01867768535262837, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": -0.06374061877901493, "ret_signed_p1m": -0.06374061877901493, "alpha_spy_p1m": -0.08129870742035361, "alpha_c_p1m": -0.06497983060236023, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.17427843942074706, "ret_signed_p3m": 0.17427843942074706, "alpha_spy_p3m": 0.12067917221676572, "alpha_c_p3m": -0.007436378949893907, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.5434462628071639, "ret_signed_p6m": 0.5434462628071639, "alpha_spy_p6m": 0.4337413459491062, "alpha_c_p6m": 0.1915193779907438, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.301, -0.2666, -0.2611, -0.262, -0.2682, -0.2525, -0.2805, -0.2573, -0.2625, -0.2772, -0.2613, -0.2518, -0.2541, -0.191, -0.1533, -0.1436, -0.1426, -0.1583, -0.1608, -0.1573, -0.1694, -0.1815, -0.1926, -0.1926, -0.1586, -0.1709, -0.1747, -0.1929, -0.187, -0.1727, -0.166, -0.1556, -0.1384, -0.1157, -0.0866, -0.0701, -0.0669, -0.0747, -0.0484, -0.0477, -0.071, -0.071, -0.0936, -0.0869, -0.052, -0.0472, -0.0371, -0.0511, -0.0448, -0.0415, -0.0178, -0.0141, -0.0141, -0.0268, -0.0198, -0.0155, -0.0097, -0.0141, -0.017, -0.0013, -0.0046, -0.0007, -0.0071, 0.0, -0.0486, -0.0434, -0.0358, -0.038, -0.0156, -0.0231, -0.0136, -0.0626, -0.0545, -0.0236, -0.058, -0.0528, -0.0274, -0.0245, -0.0294, -0.0024, 0.0125, 0.0188, -0.067, -0.0559, -0.0637, -0.0612, -0.0677, -0.0699, -0.06, -0.0509, -0.0501, -0.0446, -0.0681, -0.0681, -0.0955, -0.0981, -0.0667, -0.0327, -0.0286, -0.0192, -0.0033, 0.0252, 0.0302, 0.0568, 0.0586, 0.0579, 0.1186, 0.1166, 0.1448, 0.1448, 0.1421, 0.1319, 0.1374, 0.1372, 0.1527, 0.2064, 0.2175, 0.1772, 0.218, 0.1593, 0.1356, 0.1258, 0.0503, 0.0954, 0.0962, 0.1617, 0.1743, 0.1827, 0.2321, 0.2262, 0.1909, 0.2386, 0.2641, 0.2986, 0.2889, 0.3201, 0.2978, 0.2918, 0.3029, 0.2755, 0.3114, 0.2893, 0.2753, 0.3016, 0.2727, 0.2813, 0.2415, 0.2122, 0.2136, 0.2023, 0.2497, 0.1802, 0.1745, 0.2169, 0.2267, 0.2408, 0.2408, 0.2583, 0.2387, 0.2737, 0.2972, 0.2932, 0.3001, 0.3109, 0.312, 0.3262, 0.334, 0.2781, 0.3115, 0.3096, 0.2546, 0.3085, 0.3335, 0.3549, 0.3582, 0.367, 0.367, 0.3698, 0.3492, 0.3313, 0.3007, 0.3007, 0.3643, 0.4478, 0.4933, 0.4555, 0.418, 0.4987, 0.5288, 0.5434]}
{"unit_id": "orig:1946462358559449109", "ticker_idx": 2, "ts": "2025-07-19T06:49:24+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises China WFE TAM to $4.03B for 2025 on AI-driven logic foundry demand; bullish on ASML, KLA, AMAT, Naura, AMEC.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "China Semiconductor Equipment: Strong China WFE Demand Expected in H2 2025, Anticipating Growth\nMorgan Stanley (C. Chan, 25/07/17)\n\nMorgan Stanley believes that China's semiconductor equipment (WFE) market will continue to grow, driven by strong logic foundry demand and increasing localization needs, with demand expected to rise significantly in the second half of 2025. Based on surveys in Taiwan and Beijing, China's WFE demand is anticipated to strengthen remarkably in H2 2025, with AI-driven logic foundry being the primary growth engine. Morgan Stanley has also adjusted its 2025/2026 outlook for the Chinese WFE market, further raising its Total Addressable Market (TAM) and forecasts.\n\nRising Demand in China's WFE Market: According to Morgan Stanley's survey, China's logic foundry spending in H2 2025 could significantly exceed that of H1. Morgan Stanley expects China's lithography equipment demand to be stronger than anticipated in Q2. China's AI demand far exceeds the current AI chip supply, which is expected to drive more logic foundry spending.\n\nGrowth in China's Local Equipment Market Share: Morgan Stanley points out that approximately 25% of China's foundry CapEx in 2025 is expected to be allocated to local WFE equipment, an increase from 20% in 2024. Chinese WFE manufacturers are expanding their product portfolios; for example, Naura has launched ion implantation equipment in Q1 2025. AMEC expects its product pipeline to cover 60% of wafer processes in the next 5-10 years.\n\nMaturity Node Oversupply: Although semiconductor equipment import data from January to May indicates a decline in demand for China's mature nodes, Morgan Stanley's industry survey suggests that the government's willingness to promote localization may override economic rationality in H2 2025. Despite some government-supported projects facing limited demand and low capacity utilization rates, government policies to promote localization of the semiconductor industry may continue to support WFE demand.\n\n2025 China WFE Market Outlook: Morgan Stanley has revised its 2025 China WFE market forecast upwards from $3.66 billion (down 12% YoY) to $4.03 billion (down 3% YoY). Although the increase in 2025 primarily comes from Taiwan and South Korea, China remains the largest region for global WFE demand. The Chinese WFE market is expected to grow to $3.62 billion by 2026, despite a 10% YoY decline.\n\nForeign Equipment Suppliers Optimistic on China Market: According to Morgan Stanley's analysis, foreign equipment suppliers have improved their outlook on the Chinese market. For instance, ASML expects the Chinese market to account for over 25% of its total sales, up from a previous expectation in the low 20% range. KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011266409947964129, "ret_signed_m1d": -0.011266409947964129, "alpha_spy_m1d": -0.009373823457799246, "alpha_c_m1d": -0.010577994060046225, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028399486520033657, "ret_signed_p1d": -0.028399486520033657, "alpha_spy_p1d": -0.028542607683258225, "alpha_c_p1d": -0.010706471111973093, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.012149023467145303, "ret_signed_p1w": -0.012149023467145303, "alpha_spy_p1w": -0.025142702151910834, "alpha_c_p1w": -0.015247053025508439, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": -0.15778009452742914, "ret_signed_p1m": -0.15778009452742914, "alpha_spy_p1m": -0.1753381831687678, "alpha_c_p1m": -0.15901930635077444, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.18567379402354134, "ret_signed_p3m": 0.18567379402354134, "alpha_spy_p3m": 0.13207452681956, "alpha_c_p3m": 0.003958975652900376, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.5904836067706192, "ret_signed_p6m": 0.5904836067706192, "alpha_spy_p6m": 0.48077868991256145, "alpha_c_p6m": 0.23855672195419908, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.2567, -0.2226, -0.2154, -0.2193, -0.2258, -0.2198, -0.2288, -0.197, -0.1996, -0.2077, -0.1934, -0.1924, -0.1944, -0.1303, -0.1042, -0.0985, -0.0953, -0.1428, -0.1409, -0.142, -0.1601, -0.1666, -0.1822, -0.1822, -0.1598, -0.1611, -0.172, -0.1862, -0.1835, -0.1603, -0.1593, -0.1476, -0.1343, -0.1185, -0.0978, -0.1029, -0.0914, -0.1143, -0.0834, -0.0962, -0.1026, -0.1026, -0.1202, -0.1072, -0.0645, -0.0495, -0.0472, -0.0488, -0.0495, -0.0459, -0.0135, -0.0081, -0.0081, -0.0095, 0.0124, 0.0144, 0.0281, 0.0276, 0.0233, 0.0347, 0.0114, -0.0005, -0.0113, 0.0, -0.0284, -0.0291, -0.0233, -0.0359, -0.0121, -0.0218, -0.0167, -0.0652, -0.0655, -0.0508, -0.0699, -0.0751, -0.0491, -0.0402, -0.0427, -0.0216, -0.0134, -0.0227, -0.1602, -0.151, -0.1578, -0.1643, -0.1678, -0.154, -0.1566, -0.1434, -0.1441, -0.1395, -0.163, -0.163, -0.1796, -0.1865, -0.1761, -0.1526, -0.1563, -0.1487, -0.1491, -0.1141, -0.1263, -0.11, -0.0964, -0.0725, -0.012, -0.0102, 0.0441, 0.0459, 0.0488, 0.0393, 0.0618, 0.0671, 0.066, 0.1337, 0.1642, 0.1326, 0.1658, 0.1015, 0.1325, 0.147, 0.0932, 0.1428, 0.1361, 0.1849, 0.1857, 0.1715, 0.1878, 0.1767, 0.1484, 0.1896, 0.191, 0.2045, 0.1853, 0.2275, 0.2108, 0.2137, 0.2377, 0.1985, 0.2542, 0.2159, 0.1979, 0.224, 0.1906, 0.2013, 0.1623, 0.1768, 0.1908, 0.1721, 0.2243, 0.1489, 0.1686, 0.2046, 0.2649, 0.3041, 0.3041, 0.316, 0.329, 0.3842, 0.4014, 0.4056, 0.3981, 0.399, 0.3936, 0.4354, 0.4091, 0.3523, 0.363, 0.3503, 0.2952, 0.3225, 0.3377, 0.3512, 0.3576, 0.3605, 0.3605, 0.3663, 0.3723, 0.3562, 0.3407, 0.3407, 0.4027, 0.4833, 0.5443, 0.5244, 0.4693, 0.5712, 0.6028, 0.5905]}
{"unit_id": "orig:1946462358559449109", "ticker_idx": 3, "ts": "2025-07-19T06:49:24+00:00", "symbol": "Naura Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Naura has launched ion implantation equipment in Q1 2025", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises China WFE TAM to $4.03B for 2025 on AI-driven logic foundry demand; bullish on ASML, KLA, AMAT, Naura, AMEC.", "resolved_tickers": ["002371.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Naura Technology Group listed on Shenzhen exchange", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "China Semiconductor Equipment: Strong China WFE Demand Expected in H2 2025, Anticipating Growth\nMorgan Stanley (C. Chan, 25/07/17)\n\nMorgan Stanley believes that China's semiconductor equipment (WFE) market will continue to grow, driven by strong logic foundry demand and increasing localization needs, with demand expected to rise significantly in the second half of 2025. Based on surveys in Taiwan and Beijing, China's WFE demand is anticipated to strengthen remarkably in H2 2025, with AI-driven logic foundry being the primary growth engine. Morgan Stanley has also adjusted its 2025/2026 outlook for the Chinese WFE market, further raising its Total Addressable Market (TAM) and forecasts.\n\nRising Demand in China's WFE Market: According to Morgan Stanley's survey, China's logic foundry spending in H2 2025 could significantly exceed that of H1. Morgan Stanley expects China's lithography equipment demand to be stronger than anticipated in Q2. China's AI demand far exceeds the current AI chip supply, which is expected to drive more logic foundry spending.\n\nGrowth in China's Local Equipment Market Share: Morgan Stanley points out that approximately 25% of China's foundry CapEx in 2025 is expected to be allocated to local WFE equipment, an increase from 20% in 2024. Chinese WFE manufacturers are expanding their product portfolios; for example, Naura has launched ion implantation equipment in Q1 2025. AMEC expects its product pipeline to cover 60% of wafer processes in the next 5-10 years.\n\nMaturity Node Oversupply: Although semiconductor equipment import data from January to May indicates a decline in demand for China's mature nodes, Morgan Stanley's industry survey suggests that the government's willingness to promote localization may override economic rationality in H2 2025. Despite some government-supported projects facing limited demand and low capacity utilization rates, government policies to promote localization of the semiconductor industry may continue to support WFE demand.\n\n2025 China WFE Market Outlook: Morgan Stanley has revised its 2025 China WFE market forecast upwards from $3.66 billion (down 12% YoY) to $4.03 billion (down 3% YoY). Although the increase in 2025 primarily comes from Taiwan and South Korea, China remains the largest region for global WFE demand. The Chinese WFE market is expected to grow to $3.62 billion by 2026, despite a 10% YoY decline.\n\nForeign Equipment Suppliers Optimistic on China Market: According to Morgan Stanley's analysis, foreign equipment suppliers have improved their outlook on the Chinese market. For instance, ASML expects the Chinese market to account for over 25% of its total sales, up from a previous expectation in the low 20% range. KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016060037008226447, "ret_signed_m1d": 0.016060037008226447, "alpha_spy_m1d": 0.01795262349839133, "alpha_c_m1d": 0.01674845289614435, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006530313810918509, "ret_signed_p1d": 0.006530313810918509, "alpha_spy_p1d": 0.00638719264769394, "alpha_c_p1d": 0.024223329218979073, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.07373850024572359, "ret_signed_p1w": 0.07373850024572359, "alpha_spy_p1w": 0.06074482156095806, "alpha_c_p1w": 0.07064047068736046, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.06220904365410762, "ret_signed_p1m": 0.06220904365410762, "alpha_spy_p1m": 0.044650955012768945, "alpha_c_p1m": 0.06096983183076232, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.28320580567267495, "ret_signed_p3m": 0.28320580567267495, "alpha_spy_p3m": 0.2296065384686936, "alpha_c_p3m": 0.10149098730203399, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.5316357784838528, "ret_signed_p6m": 0.5316357784838528, "alpha_spy_p6m": 0.4219308616257951, "alpha_c_p6m": 0.1797088936674327, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [0.0523, 0.0532, 0.0466, 0.0492, 0.0492, 0.0412, 0.0412, 0.0412, 0.0412, 0.0565, 0.0351, 0.0364, 0.0179, 0.0082, 0.0043, 0.0088, -0.0097, -0.0104, 0.0043, 0.0013, -0.0141, -0.0237, -0.0191, -0.0036, -0.009, -0.0364, -0.0182, -0.0336, -0.0336, -0.0193, -0.018, -0.0081, -0.0061, -0.0076, -0.0372, -0.0378, -0.0667, -0.0521, -0.0479, -0.0398, -0.0396, -0.0221, -0.0147, 0.0126, 0.0062, 0.0195, 0.0111, 0.0032, 0.021, 0.0235, 0.0092, 0.0254, 0.0472, 0.0395, 0.0435, 0.0347, 0.0306, 0.0337, 0.0129, 0.0102, 0.0025, -0.0045, 0.0161, 0.0, 0.0065, 0.0358, 0.0581, 0.0685, 0.0737, 0.0773, 0.0911, 0.0473, 0.0305, 0.0257, 0.0264, 0.036, 0.0401, 0.0282, 0.0208, 0.0381, 0.0352, 0.0675, 0.0687, 0.0849, 0.0622, 0.0951, 0.1279, 0.1979, 0.1876, 0.1967, 0.167, 0.2402, 0.1639, 0.1806, 0.1827, 0.1668, 0.0965, 0.1163, 0.1573, 0.1364, 0.1504, 0.1793, 0.1781, 0.1873, 0.1806, 0.2167, 0.2739, 0.2561, 0.2895, 0.3085, 0.4394, 0.4174, 0.4148, 0.4342, 0.4134, 0.4134, 0.4134, 0.4134, 0.4134, 0.4134, 0.4134, 0.4312, 0.3682, 0.3617, 0.2561, 0.2885, 0.2832, 0.248, 0.2503, 0.2763, 0.2635, 0.261, 0.296, 0.3407, 0.331, 0.3236, 0.2982, 0.2718, 0.2529, 0.2668, 0.2399, 0.2842, 0.2928, 0.2988, 0.2982, 0.2909, 0.2925, 0.2732, 0.2519, 0.3229, 0.3101, 0.2654, 0.2087, 0.3005, 0.306, 0.3176, 0.3185, 0.337, 0.395, 0.3538, 0.3556, 0.4054, 0.4076, 0.4466, 0.4293, 0.4327, 0.4157, 0.422, 0.4029, 0.3838, 0.3896, 0.3845, 0.4009, 0.4538, 0.5042, 0.5032, 0.4785, 0.4342, 0.4148, 0.4342, 0.4344, 0.4344, 0.4344, 0.4826, 0.5184, 0.6104, 0.5831, 0.5544, 0.568, 0.5316]}
{"unit_id": "orig:1946462358559449109", "ticker_idx": 4, "ts": "2025-07-19T06:49:24+00:00", "symbol": "AMEC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMEC expects its product pipeline to cover 60% of wafer processes in the next 5-10 years", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises China WFE TAM to $4.03B for 2025 on AI-driven logic foundry demand; bullish on ASML, KLA, AMAT, Naura, AMEC.", "resolved_tickers": ["688012.SS"], "resolution_method": "exchange_ticker", "resolution_reason": "AMEC = Advanced Micro-Fabrication Equipment, Shanghai STAR Market", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "China Semiconductor Equipment: Strong China WFE Demand Expected in H2 2025, Anticipating Growth\nMorgan Stanley (C. Chan, 25/07/17)\n\nMorgan Stanley believes that China's semiconductor equipment (WFE) market will continue to grow, driven by strong logic foundry demand and increasing localization needs, with demand expected to rise significantly in the second half of 2025. Based on surveys in Taiwan and Beijing, China's WFE demand is anticipated to strengthen remarkably in H2 2025, with AI-driven logic foundry being the primary growth engine. Morgan Stanley has also adjusted its 2025/2026 outlook for the Chinese WFE market, further raising its Total Addressable Market (TAM) and forecasts.\n\nRising Demand in China's WFE Market: According to Morgan Stanley's survey, China's logic foundry spending in H2 2025 could significantly exceed that of H1. Morgan Stanley expects China's lithography equipment demand to be stronger than anticipated in Q2. China's AI demand far exceeds the current AI chip supply, which is expected to drive more logic foundry spending.\n\nGrowth in China's Local Equipment Market Share: Morgan Stanley points out that approximately 25% of China's foundry CapEx in 2025 is expected to be allocated to local WFE equipment, an increase from 20% in 2024. Chinese WFE manufacturers are expanding their product portfolios; for example, Naura has launched ion implantation equipment in Q1 2025. AMEC expects its product pipeline to cover 60% of wafer processes in the next 5-10 years.\n\nMaturity Node Oversupply: Although semiconductor equipment import data from January to May indicates a decline in demand for China's mature nodes, Morgan Stanley's industry survey suggests that the government's willingness to promote localization may override economic rationality in H2 2025. Despite some government-supported projects facing limited demand and low capacity utilization rates, government policies to promote localization of the semiconductor industry may continue to support WFE demand.\n\n2025 China WFE Market Outlook: Morgan Stanley has revised its 2025 China WFE market forecast upwards from $3.66 billion (down 12% YoY) to $4.03 billion (down 3% YoY). Although the increase in 2025 primarily comes from Taiwan and South Korea, China remains the largest region for global WFE demand. The Chinese WFE market is expected to grow to $3.62 billion by 2026, despite a 10% YoY decline.\n\nForeign Equipment Suppliers Optimistic on China Market: According to Morgan Stanley's analysis, foreign equipment suppliers have improved their outlook on the Chinese market. For instance, ASML expects the Chinese market to account for over 25% of its total sales, up from a previous expectation in the low 20% range. KLA and AMAT have also raised their outlook for their China businesses, reflecting strong confidence in the 2025 China WFE market.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007893274082459434, "ret_signed_m1d": -0.007893274082459434, "alpha_spy_m1d": -0.006000687592294551, "alpha_c_m1d": -0.00720485819454153, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020685909323066465, "ret_signed_p1d": 0.020685909323066465, "alpha_spy_p1d": 0.020542788159841896, "alpha_c_p1d": 0.03837892473112703, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.08905833090268289, "ret_signed_p1w": 0.08905833090268289, "alpha_spy_p1w": 0.07606465221791736, "alpha_c_p1w": 0.08596030134431976, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.043059313736301785, "ret_signed_p1m": 0.043059313736301785, "alpha_spy_p1m": 0.02550122509496311, "alpha_c_p1m": 0.04182010191295649, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.49373982565235397, "ret_signed_p3m": 0.49373982565235397, "alpha_spy_p3m": 0.44014055844837263, "alpha_c_p3m": 0.312025007281713, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.8835057174598682, "ret_signed_p6m": 0.8835057174598682, "alpha_spy_p6m": 0.7738008006018104, "alpha_c_p6m": 0.5315788326434481, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [0.0219, 0.0182, 0.0188, 0.0115, 0.0191, 0.0231, 0.0231, 0.0231, 0.0231, 0.0265, 0.029, 0.033, -0.0125, -0.0033, -0.0057, 0.0201, -0.007, -0.02, -0.0282, -0.0315, -0.041, -0.0457, -0.0515, -0.0517, -0.0591, -0.071, -0.0396, -0.0489, -0.0489, -0.0277, -0.0249, -0.0218, -0.0217, -0.0283, -0.069, -0.0768, -0.0979, -0.0664, -0.0589, -0.0708, -0.066, -0.0603, -0.0674, -0.0292, -0.0263, -0.0174, -0.0273, -0.0229, -0.0093, -0.0066, -0.0284, -0.0325, -0.0166, -0.0188, 0.0027, -0.0089, -0.017, 0.0036, -0.0256, -0.0203, -0.0295, -0.03, -0.0079, 0.0, 0.0207, 0.0806, 0.0834, 0.0881, 0.0891, 0.1235, 0.1094, 0.0758, 0.048, 0.0659, 0.0625, 0.0673, 0.0682, 0.0666, 0.0528, 0.0488, 0.0479, 0.0454, 0.0572, 0.0586, 0.0431, 0.0779, 0.1019, 0.1747, 0.1486, 0.1649, 0.1269, 0.2275, 0.1659, 0.215, 0.1568, 0.1541, 0.0637, 0.0888, 0.1579, 0.1126, 0.1192, 0.153, 0.1514, 0.1709, 0.1649, 0.2368, 0.3782, 0.3827, 0.3968, 0.5242, 0.5814, 0.5678, 0.5787, 0.6603, 0.6276, 0.6276, 0.6276, 0.6276, 0.6276, 0.6276, 0.6276, 0.7371, 0.595, 0.6064, 0.4888, 0.4916, 0.4937, 0.4257, 0.4308, 0.468, 0.4523, 0.4545, 0.5316, 0.6079, 0.6086, 0.6149, 0.6333, 0.5207, 0.5128, 0.5942, 0.5977, 0.6658, 0.6829, 0.7225, 0.6663, 0.6481, 0.6523, 0.6005, 0.5743, 0.6324, 0.6249, 0.5435, 0.4535, 0.4758, 0.4667, 0.4711, 0.4464, 0.4594, 0.4263, 0.4175, 0.4125, 0.4722, 0.4906, 0.5039, 0.5052, 0.4905, 0.4591, 0.5217, 0.5017, 0.4835, 0.5197, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.4846, 0.6948, 0.7946, 0.918, 0.8829, 0.8328, 0.9199, 0.8835]}
{"unit_id": "orig:1955844621705339196", "ticker_idx": 0, "ts": "2025-08-14T04:11:10+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "plans to take back most of the wafer orders for the TR2.5 and TR3 projects that were previously commissioned through ASIC vendors Marvell and Alchip", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Amazon pulling TR2.5/TR3 wafer orders from Marvell/Alchip to place directly with TSMC; bearish Marvell, bullish TSMC on $5B direct order flow in 2026.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "According to a discussion with a business representative from a certain company, among CSP operators, Amazon—driven by the goals of strengthening cooperation with TSMC and reducing costs—plans to take back most of the wafer orders for the TR2.5 and TR3 projects that were previously commissioned through ASIC vendors Marvell and Alchip, and place them directly with TSMC via its own subsidiary, Annapurna. As a result, in 2026, the value of its self-placed orders is expected to reach USD 5 billion.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00354259038479543, "ret_signed_m1d": -0.00354259038479543, "alpha_spy_m1d": -0.0036355471978617437, "alpha_c_m1d": -0.005662309818544653, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.036057790874986506, "ret_signed_p1d": 0.036057790874986506, "alpha_spy_p1d": 0.03371659623296375, "alpha_c_p1d": 0.015290705111012493, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.09906381304654266, "ret_signed_p1w": 0.09906381304654266, "alpha_spy_p1w": 0.08448906278924129, "alpha_c_p1w": 0.0511702804315981, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": -0.14789986050905302, "ret_signed_p1m": 0.14789986050905302, "alpha_spy_p1m": 0.16721922576099468, "alpha_c_p1m": 0.15584906186779868, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.1309355216056236, "ret_signed_p3m": -0.1309355216056236, "alpha_spy_p3m": -0.0690027108664586, "alpha_c_p3m": 0.031789699851623165, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.017091992848284443, "ret_signed_p6m": -0.017091992848284443, "alpha_spy_p6m": 0.059861360010433406, "alpha_c_p6m": 0.3173666999988336, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2091, -0.2236, -0.2401, -0.2181, -0.2328, -0.2328, -0.1932, -0.1835, -0.1944, -0.2391, -0.2229, -0.2117, -0.1619, -0.1763, -0.136, -0.126, -0.1298, -0.1373, -0.1196, -0.1506, -0.1098, -0.1152, -0.0525, -0.0525, -0.0707, -0.1052, -0.0492, -0.0401, 0.0109, -0.0246, -0.0216, -0.0362, -0.0614, -0.0496, -0.0496, -0.0955, -0.0904, -0.0865, -0.0726, -0.0801, -0.0826, -0.0839, -0.1036, -0.0889, -0.0555, -0.0757, -0.0892, -0.073, -0.0633, -0.0611, -0.0396, -0.0342, 0.0342, 0.0168, -0.0581, -0.0318, -0.0305, -0.0471, -0.0404, -0.0215, -0.0223, -0.0156, 0.0035, 0.0, -0.0361, -0.0291, -0.0882, -0.0989, -0.0991, -0.0764, -0.077, -0.0605, -0.0538, -0.0229, -0.2046, -0.2046, -0.1827, -0.2117, -0.189, -0.1988, -0.165, -0.1544, -0.1511, -0.1575, -0.1479, -0.1469, -0.1288, -0.102, -0.0609, -0.0605, -0.0444, -0.0559, 0.0133, 0.0603, 0.0523, 0.0424, 0.0636, 0.0614, 0.0906, 0.0908, 0.125, 0.1003, 0.1703, 0.1473, 0.0838, 0.1317, 0.0916, 0.1254, 0.117, 0.1135, 0.0868, 0.0667, 0.0261, 0.0479, 0.0651, 0.1231, 0.1199, 0.1413, 0.1213, 0.1868, 0.1441, 0.1089, 0.1761, 0.1816, 0.1511, 0.1803, 0.1309, 0.1309, 0.108, 0.0945, 0.0565, -0.0039, 0.0295, -0.0292, -0.0195, 0.0608, 0.0562, 0.1106, 0.1106, 0.1318, 0.1533, 0.176, 0.2686, 0.2431, 0.2522, 0.1647, 0.1255, 0.1707, 0.1322, 0.0689, 0.0667, 0.0643, 0.0343, 0.0694, 0.0646, 0.0736, 0.11, 0.095, 0.095, 0.0931, 0.0857, 0.0984, 0.0759, 0.0759, 0.1317, 0.1423, 0.117, 0.0716, 0.0565, 0.0543, 0.0502, 0.0522, 0.0289, 0.0184, 0.0194, 0.0194, 0.011, 0.0459, 0.0528, 0.0165, 0.036, 0.0507, 0.0594, 0.0305, -0.0001, -0.0034, -0.043, -0.0659, -0.0598, 0.0171]}
{"unit_id": "orig:1955844621705339196", "ticker_idx": 1, "ts": "2025-08-14T04:11:10+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "strengthening cooperation with TSMC and reducing costs—plans to take back most of the wafer orders… and place them directly with TSMC via its own subsidiary", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Amazon pulling TR2.5/TR3 wafer orders from Marvell/Alchip to place directly with TSMC; bearish Marvell, bullish TSMC on $5B direct order flow in 2026.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "According to a discussion with a business representative from a certain company, among CSP operators, Amazon—driven by the goals of strengthening cooperation with TSMC and reducing costs—plans to take back most of the wafer orders for the TR2.5 and TR3 projects that were previously commissioned through ASIC vendors Marvell and Alchip, and place them directly with TSMC via its own subsidiary, Annapurna. As a result, in 2026, the value of its self-placed orders is expected to reach USD 5 billion.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.021276678399781535, "ret_signed_m1d": 0.021276678399781535, "alpha_spy_m1d": 0.02136963521284785, "alpha_c_m1d": 0.023396397833530758, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004255314681092859, "ret_signed_p1d": 0.004255314681092859, "alpha_spy_p1d": 0.006596509323115618, "alpha_c_p1d": 0.025022400445066872, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.021276573405464405, "ret_signed_p1w": -0.021276573405464405, "alpha_spy_p1w": -0.006701823148163033, "alpha_c_p1w": 0.02661695920948015, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.07234055956721352, "ret_signed_p1m": 0.07234055956721352, "alpha_spy_p1m": 0.05302119431527186, "alpha_c_p1m": 0.06439135820846786, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.2517956915581945, "ret_signed_p3m": 0.2517956915581945, "alpha_spy_p3m": 0.18986288081902947, "alpha_c_p3m": 0.08907047010094771, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.526023103999496, "ret_signed_p6m": 0.526023103999496, "alpha_spy_p6m": 0.4490697511407782, "alpha_c_p6m": 0.191564411152378, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.1661, -0.1712, -0.161, -0.1678, -0.1661, -0.1729, -0.1822, -0.1805, -0.1805, -0.1805, -0.1983, -0.1949, -0.161, -0.1542, -0.1568, -0.1483, -0.1144, -0.0974, -0.1106, -0.1234, -0.1277, -0.1106, -0.1021, -0.1191, -0.1021, -0.1319, -0.1064, -0.0894, -0.0851, -0.0809, -0.0979, -0.0766, -0.0766, -0.0723, -0.0766, -0.0809, -0.0809, -0.0723, -0.0638, -0.0638, -0.0681, -0.0553, -0.0383, -0.0383, -0.017, -0.0213, -0.0383, -0.0255, -0.0255, -0.0255, -0.0255, -0.034, -0.017, -0.0128, -0.0128, -0.034, -0.0213, -0.0426, 0.0043, 0.0, 0.0043, 0.0043, 0.0213, 0.0, 0.0043, 0.0043, 0.0085, -0.034, -0.0213, -0.034, -0.0043, 0.0, 0.0128, -0.0128, -0.0128, -0.0085, -0.0128, -0.0128, -0.0128, 0.0043, 0.0043, 0.0213, 0.0426, 0.0553, 0.0723, 0.0681, 0.0937, 0.0809, 0.098, 0.0809, 0.1065, 0.145, 0.145, 0.1279, 0.1108, 0.1108, 0.1151, 0.1322, 0.1663, 0.1963, 0.1963, 0.2262, 0.2091, 0.2304, 0.2304, 0.2091, 0.2176, 0.2518, 0.2689, 0.239, 0.2646, 0.2646, 0.2475, 0.239, 0.239, 0.2646, 0.2603, 0.286, 0.286, 0.2817, 0.2902, 0.286, 0.2475, 0.2518, 0.2475, 0.2603, 0.2518, 0.2603, 0.2475, 0.2219, 0.2347, 0.2005, 0.192, 0.2433, 0.1834, 0.1749, 0.2091, 0.2304, 0.2262, 0.2304, 0.2048, 0.2219, 0.239, 0.2347, 0.2475, 0.2774, 0.2646, 0.286, 0.2603, 0.2688, 0.2431, 0.2302, 0.226, 0.226, 0.226, 0.256, 0.2774, 0.2817, 0.2817, 0.2945, 0.3117, 0.3031, 0.3288, 0.3288, 0.3588, 0.4317, 0.4617, 0.436, 0.4446, 0.4403, 0.4489, 0.466, 0.466, 0.4489, 0.4917, 0.5089, 0.5217, 0.4917, 0.5089, 0.5174, 0.5046, 0.526, 0.5603, 0.5475, 0.5217, 0.5132, 0.5432, 0.5303, 0.5132, 0.526]}
{"unit_id": "quote:1966382807665500291", "ticker_idx": 0, "ts": "2025-09-12T06:06:09+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron to Raise Prices by 20%–30%, Suspends Quotations", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares bullish news: Micron raising DDR4/DDR5/LPDDR prices 20–30%, suspending quotes amid supply shortage vs customer demand — implied long MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Micron to Raise Prices by 20%–30%, Suspends Quotations\n\nXincun News, 2025-09-12 11:28\n\nEvolution Beyond Single-Processor Embedded Systems\n\nFollowing SanDisk’s announcement last week of raising storage product prices by more than 10%, today Micron has informed channels that it will increase storage product prices by 20%–30%.\n\nStarting today, all quotations for DDR4, DDR5, LPDDR4, and LPDDR5 products are suspended. Prices for contract customers are all canceled, and quotations are paused. It is expected that all product quotations will be suspended for one week.\n\nIt is reported that this adjustment involves not only consumer and industrial storage products, but also automotive electronic products, which are expected to see a 70% price increase.\n\nAccording to supply chain sources, Micron executives have seen a significant supply shortage relative to customers’ FCST (forecasted) demand. Therefore, they urgently suspended all product quotations and will readjust subsequent pricing.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": "美光即將漲價20%-30%,暫停報價\n\n新村社 2025-09-12 11:28\n\n超越單處理器嵌入式系統的演進\n\n繼上週閃迪宣布將存儲產品價格上調10%以上之後, 今日美光向渠道通知存儲產品即將上漲20%-30%。 https://t.co/9AIzBDDUnO", "parent_author": null, "ret_m1d": -0.042358295684981284, "ret_signed_m1d": -0.042358295684981284, "alpha_spy_m1d": -0.042693020600087506, "alpha_c_m1d": -0.04107663723242316, "bench_spy_m1d": 0.0003347249151062215, "bench_c_m1d": -0.001281658452558121, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003434540461712876, "ret_signed_p1d": 0.003434540461712876, "alpha_spy_p1d": -0.0018893178313941217, "alpha_c_p1d": -0.005930589455460655, "bench_spy_p1d": 0.005323858293106998, "bench_c_p1d": 0.009365129917173531, "ret_p1w": 0.03498062792445933, "ret_signed_p1w": 0.03498062792445933, "alpha_spy_p1w": 0.02261378290012872, "alpha_c_p1w": -0.0024469977552239452, "bench_spy_p1w": 0.012366845024330608, "bench_c_p1w": 0.037427625679683274, "ret_p1m": 0.2268061125665073, "ret_signed_p1m": 0.2268061125665073, "alpha_spy_p1m": 0.2154460639092084, "alpha_c_p1m": 0.11038271723487858, "bench_spy_p1m": 0.011360048657298893, "bench_c_p1m": 0.11642339533162871, "ret_p3m": 0.6782747506024465, "ret_signed_p3m": 0.6782747506024465, "alpha_spy_p3m": 0.6294980741970793, "alpha_c_p3m": 0.44897668820928915, "bench_spy_p3m": 0.0487766764053672, "bench_c_p3m": 0.2292980623931573, "ret_p6m": 1.4786695298365924, "ret_signed_p6m": 1.4786695298365924, "alpha_spy_p6m": 1.441021356713683, "alpha_c_p6m": 1.1787316763393778, "bench_spy_p6m": 0.03764817312290947, "bench_c_p6m": 0.2999378534972146, "price_path": [-0.2353, -0.2259, -0.2259, -0.2146, -0.2243, -0.1872, -0.1914, -0.1994, -0.2073, -0.2169, -0.2318, -0.2264, -0.223, -0.223, -0.2373, -0.2087, -0.2225, -0.217, -0.208, -0.2456, -0.2361, -0.2595, -0.2797, -0.2725, -0.2798, -0.3053, -0.3015, -0.2894, -0.2924, -0.2924, -0.2879, -0.2702, -0.3059, -0.333, -0.3146, -0.3064, -0.3081, -0.2885, -0.2438, -0.2131, -0.1875, -0.2096, -0.2031, -0.2313, -0.2142, -0.2237, -0.2545, -0.2636, -0.2515, -0.2596, -0.259, -0.2511, -0.2241, -0.2431, -0.2431, -0.2465, -0.2449, -0.21, -0.1645, -0.1639, -0.1399, -0.1096, -0.0424, 0.0, 0.0034, 0.0101, 0.0176, 0.0742, 0.035, 0.047, 0.0584, 0.0285, -0.0025, 0.0003, 0.0424, 0.0642, 0.1585, 0.1687, 0.1954, 0.2153, 0.1817, 0.2508, 0.224, 0.1557, 0.2268, 0.1905, 0.2215, 0.2889, 0.288, 0.3159, 0.2874, 0.2631, 0.3155, 0.3939, 0.4007, 0.4123, 0.4423, 0.4256, 0.4241, 0.4937, 0.3876, 0.5115, 0.5168, 0.5141, 0.612, 0.5344, 0.5586, 0.508, 0.5708, 0.5398, 0.4542, 0.4378, 0.2815, 0.3197, 0.4251, 0.4289, 0.4654, 0.4654, 0.505, 0.5303, 0.5241, 0.4902, 0.4424, 0.5097, 0.5714, 0.6064, 0.6783, 0.6449, 0.5346, 0.5115, 0.4797, 0.4352, 0.5818, 0.6923, 0.7602, 0.7582, 0.8245, 0.8245, 0.8124, 0.8742, 0.8631, 0.8171, 0.8171, 1.0082, 0.9874, 1.1865, 1.1618, 1.082, 1.1971, 1.202, 1.1528, 1.1223, 1.1432, 1.3095, 1.3095, 1.3238, 1.4773, 1.5313, 1.5444, 1.4772, 1.6119, 1.7713, 1.7745, 1.6414, 1.7873, 1.6704, 1.4155, 1.4377, 1.5129, 1.4416, 1.3764, 1.6125, 1.6356, 1.6209, 1.6209, 1.5453, 1.68, 1.6571, 1.726, 1.6802, 1.6613, 1.7313, 1.6457, 1.6254, 1.6273, 1.4173, 1.5516, 1.5279, 1.3576, 1.4787]}
{"unit_id": "reply:1968819460363424034", "ticker_idx": 0, "ts": "2025-09-18T23:28:32+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC에게는 단기적 불확실성, 노이즈, 주가에 하방압력을 가져올 것이라고 하던데", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Relays Samsung Securities view that NVDA-Intel tie-up brings short-term uncertainty and downward price pressure on TSMC.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@dons_korea 삼성증권에서는 인텔에 엔비디아가 투자한 것 자체가 TSMC에게는 단기적 불확실성, 노이즈, 주가에 하방압력을 가져올 것이라고 하던데...", "reply_text": null, "parent_text": "궈밍치 분석: \n엔비디아–인텔 협력의 세 가지 핵심 포인트\n\n애널리스트 궈밍치는 이번 엔비디아와 인텔의 협력이 PC 생태계, 서버 시장, 그리고 TSMC의 지위 세 가지 측면에서 의미가 있다고 짚었음.\n\n1. AI PC 생태계의 재편 가능성\n\n엔비디아가 자체 ARM 기반 프로세서를 개발하는 건 불확실성이", "quoted_text": null, "parent_author": "dons_korea", "ret_m1d": -0.015564141313881175, "ret_signed_m1d": 0.015564141313881175, "alpha_spy_m1d": 0.010913426381942926, "alpha_c_m1d": -0.021410995960601586, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.03697513727448276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015564141313881175, "ret_signed_p1d": 0.015564141313881175, "alpha_spy_p1d": 0.02051701608921319, "alpha_c_p1d": 0.011588963959257348, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": -0.003975177354623827, "ret_p1w": 0.027237342923270846, "ret_signed_p1w": -0.027237342923270846, "alpha_spy_p1w": -0.03083956041032332, "alpha_c_p1w": -0.01433395199847709, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": 0.012903390924793756, "ret_p1m": 0.1284046439594142, "ret_signed_p1m": -0.1284046439594142, "alpha_spy_p1m": -0.12240698948565476, "alpha_c_p1m": -0.0468197303176312, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": 0.081584913641783, "ret_p3m": 0.12045398822754061, "ret_signed_p3m": -0.12045398822754061, "alpha_spy_p3m": -0.09253120284728111, "alpha_c_p3m": -0.010128136681485378, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": 0.11032585154605523, "ret_p6m": 0.4561997233024544, "ret_signed_p6m": -0.4561997233024544, "alpha_spy_p6m": -0.4504186275598876, "alpha_c_p6m": -0.23042065952625812, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": 0.22577906377619628, "price_path": [-0.2094, -0.1861, -0.1706, -0.1668, -0.1629, -0.1784, -0.159, -0.159, -0.1551, -0.159, -0.1629, -0.1629, -0.1551, -0.1474, -0.1474, -0.1513, -0.1396, -0.1241, -0.1241, -0.1047, -0.1086, -0.1241, -0.1125, -0.1125, -0.1125, -0.1125, -0.1203, -0.1047, -0.1009, -0.1009, -0.1203, -0.1086, -0.128, -0.0854, -0.0892, -0.0854, -0.0854, -0.0699, -0.0892, -0.0854, -0.0854, -0.0815, -0.1203, -0.1086, -0.1203, -0.0931, -0.0892, -0.0776, -0.1009, -0.1009, -0.097, -0.1009, -0.1009, -0.1009, -0.0854, -0.0854, -0.0699, -0.0505, -0.0389, -0.0234, -0.0272, -0.0039, -0.0156, 0.0, -0.0156, 0.0078, 0.0428, 0.0428, 0.0272, 0.0117, 0.0117, 0.0156, 0.0311, 0.0623, 0.0895, 0.0895, 0.1167, 0.1012, 0.1206, 0.1206, 0.1012, 0.1089, 0.1401, 0.1556, 0.1284, 0.1518, 0.1518, 0.1362, 0.1284, 0.1284, 0.1518, 0.1479, 0.1712, 0.1712, 0.1673, 0.1751, 0.1712, 0.1362, 0.1401, 0.1362, 0.1479, 0.1401, 0.1479, 0.1362, 0.1128, 0.1245, 0.0934, 0.0856, 0.1323, 0.0778, 0.07, 0.1012, 0.1206, 0.1167, 0.1206, 0.0973, 0.1128, 0.1284, 0.1245, 0.1362, 0.1634, 0.1518, 0.1712, 0.1478, 0.1556, 0.1322, 0.1205, 0.1165, 0.1165, 0.1165, 0.1439, 0.1634, 0.1673, 0.1673, 0.179, 0.1946, 0.1868, 0.2102, 0.2102, 0.2376, 0.3039, 0.3313, 0.3078, 0.3157, 0.3118, 0.3196, 0.3352, 0.3352, 0.3196, 0.3586, 0.3742, 0.3859, 0.3586, 0.3742, 0.382, 0.3703, 0.3898, 0.4211, 0.4094, 0.3859, 0.3781, 0.4054, 0.3937, 0.3781, 0.3898, 0.4172, 0.4679, 0.4952, 0.4952, 0.4952, 0.4952, 0.4952, 0.4952, 0.4952, 0.4952, 0.4835, 0.5343, 0.5733, 0.5577, 0.5577, 0.5421, 0.5109, 0.4562, 0.4835, 0.4757, 0.4133, 0.4445, 0.5148, 0.4718, 0.4562]}
{"unit_id": "orig:1963874307240861729", "ticker_idx": 0, "ts": "2025-09-05T07:58:16+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Our investment rating is Overweight, and Broadcom is presented as a top pick in the semiconductor sector.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan reiterates Overweight on AVGO as top semi pick; raises PT on AI revenue now projected $45B in FY2026 (+125% YoY) and >60% in 2027.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Broadcom – JPMorgan\n\nBroadcom reported quarterly results in July that exceeded market expectations and also provided a solid revenue outlook for the October quarter. AI revenue came in at $5.2B, up 18% QoQ, and is expected to increase to $6.2B in the October quarter (+19% QoQ, +66% YoY). This is above the consensus of $5.7B, and the company is targeting approximately $20B in AI revenue for FY2025. These results were driven by the ramp of Google’s next-generation 3nm TPU (V6, Ironwood) and demand for the Tomahawk 5 switching platform.\n\nOne of its major customers (believed to be OpenAI) has been secured as an official client, with a $10B order for AI inference. Ramp-up for this customer is expected to begin in the second half of FY2026, leading to AI revenue growth far above prior guidance (+60% YoY), now expected to exceed 100%. Accordingly, we have raised our forecast for Broadcom’s AI revenue in FY2026 to $45B (+125% YoY), and we expect growth of more than 60% in 2027 as well. In addition, we project that the next-generation AI ASIC project with SoftBank and Arm could emerge as a key customer driver going forward.\n\nOutside of AI, semiconductor orders have increased by more than 20%, showing gradual recovery. In the infrastructure software segment, the transition to VCF full-stack solutions and upselling are proceeding steadily. This structural growth is supported by strong capex from cloud and hyperscalers, rising AI training and inference workloads, Google TPU v6/v7 3nm ASICs, new projects with Meta and OpenAI, and demand for AI networking. Broadcom also highlighted that it secured a technological lead by being the first in the industry to complete tape-out of a 2nm process and 3.5D packaging in the second half of 2025.\n\nWe have raised our price target and earnings estimates in line with the strengthened fundamentals. Our investment rating is Overweight, and Broadcom is presented as a top pick in the semiconductor sector.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08596852489089546, "ret_signed_m1d": -0.08596852489089546, "alpha_spy_m1d": -0.08887308541801109, "alpha_c_m1d": -0.07430720299676363, "bench_spy_m1d": 0.0029045605271156294, "bench_c_m1d": -0.011661321894131826, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032129855308787425, "ret_signed_p1d": 0.032129855308787425, "alpha_spy_p1d": 0.029673313400073953, "alpha_c_p1d": 0.02108249595834777, "bench_spy_p1d": 0.002456541908713472, "bench_c_p1d": 0.011047359350439656, "ret_p1w": 0.07459148474333199, "ret_signed_p1w": 0.07459148474333199, "alpha_spy_p1w": 0.05887859401803075, "alpha_c_p1w": 0.03694823247038159, "bench_spy_p1w": 0.015712890725301243, "bench_c_p1w": 0.0376432522729504, "ret_p1m": 0.003507896272217481, "ret_signed_p1m": 0.003507896272217481, "alpha_spy_p1m": -0.03702109854936819, "alpha_c_p1m": -0.16731857016875895, "bench_spy_p1m": 0.04052899482158567, "bench_c_p1m": 0.17082646644097643, "ret_p3m": 0.1384694739246224, "ret_signed_p3m": 0.1384694739246224, "alpha_spy_p3m": 0.07891495145674066, "alpha_c_p3m": -0.10392684213310766, "bench_spy_p3m": 0.05955452246788173, "bench_c_p3m": 0.24239631605773004, "ret_p6m": -0.044530165525046184, "ret_signed_p6m": -0.044530165525046184, "alpha_spy_p6m": -0.11108475304169352, "alpha_c_p6m": -0.434514225671247, "bench_spy_p6m": 0.06655458751664733, "bench_c_p6m": 0.3899840601462008, "price_path": [-0.2712, -0.2466, -0.2372, -0.2591, -0.249, -0.2571, -0.2515, -0.2515, -0.2535, -0.2422, -0.2124, -0.2097, -0.1933, -0.1957, -0.1769, -0.2095, -0.1941, -0.1783, -0.1783, -0.1813, -0.1884, -0.1702, -0.1776, -0.1807, -0.177, -0.1611, -0.1615, -0.1446, -0.1539, -0.1394, -0.1681, -0.1529, -0.1379, -0.1335, -0.1212, -0.1119, -0.0964, -0.123, -0.1381, -0.111, -0.1253, -0.0992, -0.093, -0.0893, -0.0925, -0.0659, -0.077, -0.0707, -0.0853, -0.087, -0.1194, -0.1306, -0.1352, -0.1221, -0.1214, -0.1101, -0.1034, -0.0784, -0.112, -0.112, -0.1094, -0.097, -0.086, 0.0, 0.0321, 0.0053, 0.1036, 0.0739, 0.0746, 0.0872, 0.075, 0.0337, 0.0312, 0.03, 0.0134, 0.0138, 0.0149, 0.0053, 0.0006, -0.0192, -0.0132, -0.0028, 0.0116, 0.0121, 0.0035, 0.0063, 0.0334, 0.032, -0.029, 0.067, 0.0294, 0.0509, 0.0593, 0.0449, 0.0446, 0.025, 0.0179, 0.0298, 0.0593, 0.083, 0.1156, 0.1545, 0.1261, 0.1056, 0.0844, 0.0527, 0.0738, 0.0636, 0.0452, 0.072, 0.0528, 0.0625, 0.0169, 0.0244, 0.0249, 0.0185, 0.0601, 0.0374, 0.0176, 0.1305, 0.1517, 0.1892, 0.1892, 0.2053, 0.1548, 0.1413, 0.1385, 0.1397, 0.1673, 0.1998, 0.2153, 0.2353, 0.2155, 0.0766, 0.0164, 0.0209, -0.0248, -0.0133, 0.0181, 0.0233, 0.0469, 0.0496, 0.0496, 0.0553, 0.0471, 0.0485, 0.0372, 0.0372, 0.0418, 0.0292, 0.0302, 0.0294, -0.0036, 0.0338, 0.0555, 0.0627, 0.0186, 0.028, 0.054, 0.054, -0.0032, -0.0146, -0.0245, -0.0408, -0.0265, -0.0027, -0.0013, -0.0088, -0.0071, -0.0077, -0.04, -0.0768, -0.0694, -0.0023, 0.0308, 0.0203, 0.0272, -0.0075, -0.0255, -0.0255, -0.0034, -0.0005, 0.0009, -0.0031, -0.01, -0.0245, -0.0041, -0.0359, -0.0423, -0.0445]}
{"unit_id": "orig:1945084879701704742", "ticker_idx": 0, "ts": "2025-07-15T11:35:47+00:00", "symbol": "data center", "kind": "sector", "geography": "US", "sub_sector": "AI infrastructure", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "AI infrastructure spending remains resilient", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi bullish on AI infra, Ethernet/optical, and storage; bearish on consumer hardware/PC/smartphone on tariff uncertainty; 2026 cloud capex +15% vs 9% consensus.", "resolved_tickers": ["DTCR"], "resolution_method": "etf", "resolution_reason": "US data center/AI infra sector proxied by DTCR ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "North American Hardware & Communications Outlook: AI-Driven Growth Sustains Infrastructure Spending, Consumer Electronics Face Uncertainty\n\nCiti (A. Malik, 25/07/11)\n\nCiti notes that while AI infrastructure spending remains resilient, the consumer side faces uncertainty due to tariff impacts. Citi is optimistic about AI-related communications and storage sectors and advises avoiding consumer-oriented hardware companies.\n\nAI Infrastructure Spending Continues to Strengthen, US Top 5 Cloud Vendors' 2026 Capex Expected to Grow 15%\n\nCiti raised its forecast for the year-over-year growth rate of data center capital expenditure for the top five US cloud vendors in 2026 to 15% (market consensus is 9%), primarily driven by AI. Although enterprise AI applications are still in their early stages, sovereign AI demand is growing rapidly. Due to limitations in power, connectivity, and computing power, there are some imbalances in AI infrastructure construction, but long-term growth potential remains significant.\n\nEthernet Demand Accelerating, 2025 Data Center Switch Shipment Forecast Revised Up from 14% to 23%\n\nIn AI backend networks, the trend of Ethernet replacing InfiniBand is clear, driving accelerated 800G shipments. Data center switch growth is expected to reach 23% in 2025. As GPUs evolve to next-generation products, demand for optical modules will also increase, benefiting communications infrastructure vendors.\n\nAI Evolution Towards Inference Era Drives Demand for High-Performance Storage\n\nAI inference applications place higher demands on data collection and processing. Hard drive manufacturers are focusing on high-capacity and high-performance storage products while addressing supply tightness. The overall storage industry fundamentals are benefiting.\n\nIncreased Tariff Uncertainty Puts Pressure on Consumer Electronics\n\nThe current average tariff rate is approximately 15%. If a new round of tariffs is implemented as scheduled in August, it could rise to 20%. Key raw materials such as copper and semiconductors may face higher tariffs, increasing consumer-side uncertainty and impacting demand for PCs and smartphones. Citi advises avoiding consumer-related stocks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013734834590792278, "ret_signed_m1d": -0.013734834590792278, "alpha_spy_m1d": -0.01802633996076397, "alpha_c_m1d": -0.01802633996076397, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": 0.004291505369971693, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009508770641800002, "ret_signed_p1d": 0.009508770641800002, "alpha_spy_p1d": 0.006165579523284892, "alpha_c_p1d": 0.006165579523284892, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.0033431911185151097, "ret_p1w": 0.02218706389403069, "ret_signed_p1w": 0.02218706389403069, "alpha_spy_p1w": 0.01138568047321531, "alpha_c_p1w": 0.01138568047321531, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.01080138342081538, "ret_p1m": 0.0010565413385614786, "ret_signed_p1m": 0.0010565413385614786, "alpha_spy_p1m": -0.03551079258450418, "alpha_c_p1m": -0.03551079258450418, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.03656733392306566, "ret_p3m": 0.11780238184716163, "ret_signed_p3m": 0.11780238184716163, "alpha_spy_p3m": 0.06525726370659757, "alpha_c_p3m": 0.06525726370659757, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.05254511814056406, "ret_p6m": 0.19356681014157995, "ret_signed_p6m": 0.19356681014157995, "alpha_spy_p6m": 0.0788095068603929, "alpha_c_p6m": 0.0788095068603929, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.11475730328118705, "price_path": [-0.1808, -0.1808, -0.1944, -0.1776, -0.1703, -0.1619, -0.1529, -0.1482, -0.1361, -0.1256, -0.1193, -0.0988, -0.0972, -0.0967, -0.1104, -0.1119, -0.1093, -0.0962, -0.0999, -0.0972, -0.0935, -0.0767, -0.0746, -0.0767, -0.0825, -0.0893, -0.0904, -0.0904, -0.0799, -0.0904, -0.0857, -0.0951, -0.0783, -0.0736, -0.0541, -0.0531, -0.0515, -0.0489, -0.041, -0.0431, -0.0384, -0.0589, -0.0389, -0.0473, -0.0442, -0.0442, -0.0468, -0.0373, -0.0142, -0.0279, -0.0321, -0.0227, -0.0153, -0.0248, -0.0132, -0.0069, -0.0069, -0.0275, -0.0254, -0.0217, -0.0222, -0.0211, -0.0137, 0.0, 0.0095, 0.0169, 0.0203, 0.019, 0.0222, 0.0317, 0.0317, 0.0301, 0.0227, 0.0169, 0.0079, 0.0069, -0.01, 0.0063, 0.0026, -0.0058, -0.0005, -0.0074, -0.0058, 0.0037, 0.0011, -0.0095, -0.0106, -0.0079, -0.0121, -0.0106, -0.018, 0.0053, -0.0016, 0.0, 0.0005, 0.0042, -0.0011, -0.0011, -0.0211, -0.0317, -0.0322, -0.0201, -0.0217, -0.0063, 0.0127, 0.0375, 0.0417, 0.0481, 0.0486, 0.0571, 0.0687, 0.066, 0.0903, 0.0856, 0.0819, 0.0824, 0.0761, 0.0729, 0.0819, 0.0998, 0.1157, 0.1167, 0.1284, 0.1231, 0.1368, 0.1405, 0.1178, 0.1416, 0.1352, 0.1616, 0.1638, 0.15, 0.1749, 0.1527, 0.1368, 0.1595, 0.177, 0.1965, 0.187, 0.1833, 0.178, 0.1812, 0.1786, 0.1469, 0.1664, 0.1479, 0.1484, 0.1722, 0.1622, 0.1458, 0.0977, 0.1014, 0.0909, 0.0792, 0.074, 0.0438, 0.0423, 0.0729, 0.0734, 0.0851, 0.0851, 0.1046, 0.0967, 0.0977, 0.1078, 0.113, 0.1278, 0.1479, 0.1506, 0.1437, 0.1442, 0.1152, 0.0924, 0.0835, 0.0629, 0.0755, 0.0967, 0.1072, 0.1059, 0.1146, 0.1146, 0.1141, 0.1173, 0.1276, 0.1218, 0.1218, 0.167, 0.1813, 0.2015, 0.1936]}
{"unit_id": "orig:1945084879701704742", "ticker_idx": 1, "ts": "2025-07-15T11:35:47+00:00", "symbol": "communications infrastructure", "kind": "sector", "geography": "US", "sub_sector": "Ethernet/optical", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Citi is optimistic about AI-related communications and storage sectors", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi bullish on AI infra, Ethernet/optical, and storage; bearish on consumer hardware/PC/smartphone on tariff uncertainty; 2026 cloud capex +15% vs 9% consensus.", "resolved_tickers": ["COHR", "CIEN", "LITE"], "resolution_method": "basket", "resolution_reason": "US optical/comms infra basket: Coherent, Ciena, Lumentum; IIVI merged→COHR", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "North American Hardware & Communications Outlook: AI-Driven Growth Sustains Infrastructure Spending, Consumer Electronics Face Uncertainty\n\nCiti (A. Malik, 25/07/11)\n\nCiti notes that while AI infrastructure spending remains resilient, the consumer side faces uncertainty due to tariff impacts. Citi is optimistic about AI-related communications and storage sectors and advises avoiding consumer-oriented hardware companies.\n\nAI Infrastructure Spending Continues to Strengthen, US Top 5 Cloud Vendors' 2026 Capex Expected to Grow 15%\n\nCiti raised its forecast for the year-over-year growth rate of data center capital expenditure for the top five US cloud vendors in 2026 to 15% (market consensus is 9%), primarily driven by AI. Although enterprise AI applications are still in their early stages, sovereign AI demand is growing rapidly. Due to limitations in power, connectivity, and computing power, there are some imbalances in AI infrastructure construction, but long-term growth potential remains significant.\n\nEthernet Demand Accelerating, 2025 Data Center Switch Shipment Forecast Revised Up from 14% to 23%\n\nIn AI backend networks, the trend of Ethernet replacing InfiniBand is clear, driving accelerated 800G shipments. Data center switch growth is expected to reach 23% in 2025. As GPUs evolve to next-generation products, demand for optical modules will also increase, benefiting communications infrastructure vendors.\n\nAI Evolution Towards Inference Era Drives Demand for High-Performance Storage\n\nAI inference applications place higher demands on data collection and processing. Hard drive manufacturers are focusing on high-capacity and high-performance storage products while addressing supply tightness. The overall storage industry fundamentals are benefiting.\n\nIncreased Tariff Uncertainty Puts Pressure on Consumer Electronics\n\nThe current average tariff rate is approximately 15%. If a new round of tariffs is implemented as scheduled in August, it could rise to 20%. Key raw materials such as copper and semiconductors may face higher tariffs, increasing consumer-side uncertainty and impacting demand for PCs and smartphones. Citi advises avoiding consumer-related stocks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030219126765356836, "ret_signed_m1d": -0.030219126765356836, "alpha_spy_m1d": -0.034510632135328526, "alpha_c_m1d": -0.034510632135328526, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": 0.004291505369971693, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01722564653962766, "ret_signed_p1d": 0.01722564653962766, "alpha_spy_p1d": 0.01388245542111255, "alpha_c_p1d": 0.01388245542111255, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.0033431911185151097, "ret_p1w": 0.015418279286629163, "ret_signed_p1w": 0.015418279286629163, "alpha_spy_p1w": 0.004616895865813783, "alpha_c_p1w": 0.004616895865813783, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.01080138342081538, "ret_p1m": 0.18589722294239341, "ret_signed_p1m": 0.18589722294239341, "alpha_spy_p1m": 0.14932988901932776, "alpha_c_p1m": 0.14932988901932776, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.03656733392306566, "ret_p3m": 0.522856078044016, "ret_signed_p3m": 0.522856078044016, "alpha_spy_p3m": 0.47031095990345195, "alpha_c_p3m": 0.47031095990345195, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.05254511814056406, "ret_p6m": 2.0454432753708756, "ret_signed_p6m": 2.0454432753708756, "alpha_spy_p6m": 1.9306859720896885, "alpha_c_p6m": 1.9306859720896885, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.11475730328118705, "price_path": [-0.3889, -0.3889, -0.4157, -0.3932, -0.3542, -0.3206, -0.3064, -0.3028, -0.2996, -0.3053, -0.2628, -0.244, -0.2451, -0.2459, -0.2375, -0.2303, -0.2332, -0.1677, -0.148, -0.1332, -0.1404, -0.1347, -0.1333, -0.1354, -0.1435, -0.1453, -0.1467, -0.1467, -0.1198, -0.1302, -0.1389, -0.169, -0.1461, -0.121, -0.1056, -0.1516, -0.1532, -0.147, -0.1447, -0.146, -0.1296, -0.1603, -0.1308, -0.1336, -0.1162, -0.1162, -0.113, -0.113, -0.0871, -0.0707, -0.0376, -0.0442, -0.0394, -0.0654, -0.0627, -0.0412, -0.0412, -0.0618, -0.0588, -0.0561, -0.0437, -0.0467, -0.0302, 0.0, 0.0172, 0.0388, 0.0357, 0.0441, 0.0154, 0.0384, 0.0437, 0.06, 0.0913, 0.1105, 0.1202, 0.1218, 0.0766, 0.1217, 0.1029, 0.126, 0.1564, 0.179, 0.1616, 0.1998, 0.1859, 0.0762, 0.0843, 0.089, 0.0633, 0.0435, 0.0518, 0.0797, 0.0993, 0.1102, 0.1164, 0.1857, 0.1442, 0.1442, 0.1325, 0.1451, 0.2855, 0.318, 0.33, 0.3527, 0.4431, 0.4538, 0.4408, 0.4906, 0.4977, 0.4639, 0.5144, 0.51, 0.5192, 0.499, 0.4469, 0.4637, 0.4899, 0.5084, 0.5142, 0.5922, 0.5839, 0.5616, 0.5621, 0.539, 0.603, 0.6273, 0.5229, 0.6097, 0.577, 0.5956, 0.6459, 0.6531, 0.6624, 0.6573, 0.6008, 0.6903, 0.7802, 0.8632, 0.8782, 0.9705, 0.9027, 0.9088, 0.9008, 0.8242, 1.0478, 1.1797, 1.1557, 1.3035, 1.2427, 1.2511, 1.0398, 1.0455, 1.0786, 1.0743, 1.1711, 0.9747, 1.0715, 1.3291, 1.2976, 1.3896, 1.3896, 1.4983, 1.4579, 1.4177, 1.4024, 1.5187, 1.5703, 1.6676, 1.7575, 1.8234, 1.9305, 1.6019, 1.6282, 1.5284, 1.5018, 1.6049, 1.8358, 1.9422, 1.9511, 1.9783, 1.9783, 1.9676, 1.8858, 1.8707, 1.836, 1.836, 1.9785, 1.791, 2.049, 2.0454]}
{"unit_id": "orig:1945084879701704742", "ticker_idx": 2, "ts": "2025-07-15T11:35:47+00:00", "symbol": "storage", "kind": "sector", "geography": "US", "sub_sector": "high-capacity HDD", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The overall storage industry fundamentals are benefiting", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi bullish on AI infra, Ethernet/optical, and storage; bearish on consumer hardware/PC/smartphone on tariff uncertainty; 2026 cloud capex +15% vs 9% consensus.", "resolved_tickers": ["STX", "WDC"], "resolution_method": "basket", "resolution_reason": "High-capacity HDD storage basket: Seagate (STX), Western Digital (WDC)", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "North American Hardware & Communications Outlook: AI-Driven Growth Sustains Infrastructure Spending, Consumer Electronics Face Uncertainty\n\nCiti (A. Malik, 25/07/11)\n\nCiti notes that while AI infrastructure spending remains resilient, the consumer side faces uncertainty due to tariff impacts. Citi is optimistic about AI-related communications and storage sectors and advises avoiding consumer-oriented hardware companies.\n\nAI Infrastructure Spending Continues to Strengthen, US Top 5 Cloud Vendors' 2026 Capex Expected to Grow 15%\n\nCiti raised its forecast for the year-over-year growth rate of data center capital expenditure for the top five US cloud vendors in 2026 to 15% (market consensus is 9%), primarily driven by AI. Although enterprise AI applications are still in their early stages, sovereign AI demand is growing rapidly. Due to limitations in power, connectivity, and computing power, there are some imbalances in AI infrastructure construction, but long-term growth potential remains significant.\n\nEthernet Demand Accelerating, 2025 Data Center Switch Shipment Forecast Revised Up from 14% to 23%\n\nIn AI backend networks, the trend of Ethernet replacing InfiniBand is clear, driving accelerated 800G shipments. Data center switch growth is expected to reach 23% in 2025. As GPUs evolve to next-generation products, demand for optical modules will also increase, benefiting communications infrastructure vendors.\n\nAI Evolution Towards Inference Era Drives Demand for High-Performance Storage\n\nAI inference applications place higher demands on data collection and processing. Hard drive manufacturers are focusing on high-capacity and high-performance storage products while addressing supply tightness. The overall storage industry fundamentals are benefiting.\n\nIncreased Tariff Uncertainty Puts Pressure on Consumer Electronics\n\nThe current average tariff rate is approximately 15%. If a new round of tariffs is implemented as scheduled in August, it could rise to 20%. Key raw materials such as copper and semiconductors may face higher tariffs, increasing consumer-side uncertainty and impacting demand for PCs and smartphones. Citi advises avoiding consumer-related stocks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004341794626474882, "ret_signed_m1d": -0.004341794626474882, "alpha_spy_m1d": -0.008633299996446575, "alpha_c_m1d": 0.004459811627460564, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.008801606253935446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013878406311816027, "ret_signed_p1d": -0.013878406311816027, "alpha_spy_p1d": -0.017221597430331137, "alpha_c_p1d": -0.016980187210225317, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00310178089840929, "ret_p1w": -0.01173216266598176, "ret_signed_p1w": -0.01173216266598176, "alpha_spy_p1w": -0.02253354608679714, "alpha_c_p1w": -0.015221725579574241, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.003489562913592481, "ret_p1m": 0.08963169882702937, "ret_signed_p1m": 0.08963169882702937, "alpha_spy_p1m": 0.05306436490396371, "alpha_c_p1m": 0.049307715507358685, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.040323983319670686, "ret_p3m": 0.576999566890543, "ret_signed_p3m": 0.576999566890543, "alpha_spy_p3m": 0.5244544487499789, "alpha_c_p3m": 0.4962047563263343, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.08079481056420867, "ret_p6m": 1.5229127444222932, "ret_signed_p6m": 1.5229127444222932, "alpha_spy_p6m": 1.4081554411411061, "alpha_c_p6m": 1.3834389351530674, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.1394738092692258, "price_path": [-0.4773, -0.4773, -0.4875, -0.4775, -0.4608, -0.426, -0.4227, -0.4235, -0.4275, -0.3721, -0.3748, -0.3592, -0.3549, -0.3594, -0.3541, -0.3513, -0.3547, -0.3156, -0.2859, -0.2854, -0.2739, -0.2709, -0.2613, -0.2689, -0.2858, -0.2684, -0.253, -0.253, -0.2265, -0.2216, -0.2174, -0.2255, -0.2167, -0.1914, -0.1711, -0.1663, -0.1658, -0.1435, -0.1587, -0.1657, -0.1663, -0.1629, -0.1377, -0.1296, -0.1236, -0.1236, -0.124, -0.1089, -0.0856, -0.0721, -0.0578, -0.0569, -0.042, -0.0408, -0.0033, -0.0094, -0.0094, -0.0193, -0.0414, -0.045, -0.0336, -0.0166, -0.0043, 0.0, -0.0139, -0.0116, 0.0035, 0.0109, -0.0117, 0.0257, 0.0234, 0.0157, 0.0155, 0.035, 0.0234, 0.1093, 0.0861, 0.0916, 0.0706, 0.0403, 0.048, 0.0598, 0.0615, 0.084, 0.0896, 0.0869, 0.0738, 0.0972, 0.0915, 0.0914, 0.0714, 0.104, 0.1367, 0.1445, 0.1583, 0.1857, 0.1564, 0.1564, 0.1784, 0.2282, 0.288, 0.3135, 0.3264, 0.3435, 0.3519, 0.3729, 0.3814, 0.4672, 0.4724, 0.464, 0.5062, 0.5326, 0.6026, 0.5825, 0.5635, 0.5322, 0.5219, 0.634, 0.6844, 0.8323, 0.8302, 0.824, 0.7458, 0.6462, 0.6532, 0.6334, 0.577, 0.6198, 0.5508, 0.631, 0.694, 0.694, 0.6224, 0.622, 0.6167, 0.6938, 0.7473, 0.714, 0.6764, 0.9419, 0.9269, 0.9745, 1.065, 0.9706, 1.1148, 1.1498, 1.148, 1.2808, 1.2293, 1.1845, 1.0485, 1.0389, 1.0838, 0.9874, 1.0134, 0.8488, 0.831, 0.9715, 1.0333, 1.0855, 1.0855, 1.1418, 1.1212, 1.084, 1.0238, 1.0883, 1.1911, 1.22, 1.2096, 1.3557, 1.4247, 1.2762, 1.2373, 1.2648, 1.1678, 1.281, 1.3407, 1.2632, 1.274, 1.2946, 1.2946, 1.3125, 1.2821, 1.2511, 1.2072, 1.2072, 1.3626, 1.3717, 1.7423, 1.5229]}
{"unit_id": "orig:1945084879701704742", "ticker_idx": 3, "ts": "2025-07-15T11:35:47+00:00", "symbol": "consumer electronics", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Citi advises avoiding consumer-related stocks", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi bullish on AI infra, Ethernet/optical, and storage; bearish on consumer hardware/PC/smartphone on tariff uncertainty; 2026 cloud capex +15% vs 9% consensus.", "resolved_tickers": ["XLY"], "resolution_method": "etf", "resolution_reason": "Consumer electronics proxied by XLY consumer discretionary ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "North American Hardware & Communications Outlook: AI-Driven Growth Sustains Infrastructure Spending, Consumer Electronics Face Uncertainty\n\nCiti (A. Malik, 25/07/11)\n\nCiti notes that while AI infrastructure spending remains resilient, the consumer side faces uncertainty due to tariff impacts. Citi is optimistic about AI-related communications and storage sectors and advises avoiding consumer-oriented hardware companies.\n\nAI Infrastructure Spending Continues to Strengthen, US Top 5 Cloud Vendors' 2026 Capex Expected to Grow 15%\n\nCiti raised its forecast for the year-over-year growth rate of data center capital expenditure for the top five US cloud vendors in 2026 to 15% (market consensus is 9%), primarily driven by AI. Although enterprise AI applications are still in their early stages, sovereign AI demand is growing rapidly. Due to limitations in power, connectivity, and computing power, there are some imbalances in AI infrastructure construction, but long-term growth potential remains significant.\n\nEthernet Demand Accelerating, 2025 Data Center Switch Shipment Forecast Revised Up from 14% to 23%\n\nIn AI backend networks, the trend of Ethernet replacing InfiniBand is clear, driving accelerated 800G shipments. Data center switch growth is expected to reach 23% in 2025. As GPUs evolve to next-generation products, demand for optical modules will also increase, benefiting communications infrastructure vendors.\n\nAI Evolution Towards Inference Era Drives Demand for High-Performance Storage\n\nAI inference applications place higher demands on data collection and processing. Hard drive manufacturers are focusing on high-capacity and high-performance storage products while addressing supply tightness. The overall storage industry fundamentals are benefiting.\n\nIncreased Tariff Uncertainty Puts Pressure on Consumer Electronics\n\nThe current average tariff rate is approximately 15%. If a new round of tariffs is implemented as scheduled in August, it could rise to 20%. Key raw materials such as copper and semiconductors may face higher tariffs, increasing consumer-side uncertainty and impacting demand for PCs and smartphones. Citi advises avoiding consumer-related stocks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014515779100509052, "ret_signed_m1d": -0.014515779100509052, "alpha_spy_m1d": -0.010224273730537359, "alpha_c_m1d": -0.010224273730537359, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": 0.004291505369971693, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0018258912130597604, "ret_signed_p1d": -0.0018258912130597604, "alpha_spy_p1d": 0.0015172999054553493, "alpha_c_p1d": 0.0015172999054553493, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.0033431911185151097, "ret_p1w": 0.029260004765082837, "ret_signed_p1w": -0.029260004765082837, "alpha_spy_p1w": -0.018458621344267456, "alpha_c_p1w": -0.018458621344267456, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.01080138342081538, "ret_p1m": 0.04920799236521911, "ret_signed_p1m": -0.04920799236521911, "alpha_spy_p1m": -0.012640658442153452, "alpha_c_p1m": -0.012640658442153452, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.03656733392306566, "ret_p3m": 0.04602844492479785, "ret_signed_p3m": -0.04602844492479785, "alpha_spy_p3m": 0.006516673215766211, "alpha_c_p3m": 0.006516673215766211, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.05254511814056406, "ret_p6m": 0.10789480081259994, "ret_signed_p6m": -0.10789480081259994, "alpha_spy_p6m": 0.006862502468587106, "alpha_c_p6m": 0.006862502468587106, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.11475730328118705, "price_path": [-0.1555, -0.1555, -0.178, -0.152, -0.1331, -0.1158, -0.0999, -0.0996, -0.0932, -0.1015, -0.0965, -0.0825, -0.0922, -0.1001, -0.0931, -0.0824, -0.078, -0.0322, -0.0194, -0.0156, -0.0178, -0.0085, -0.0108, -0.016, -0.0359, -0.0318, -0.0405, -0.0405, -0.0121, -0.0212, -0.0208, -0.0261, -0.0238, -0.0187, -0.0213, -0.0459, -0.0336, -0.0258, -0.0141, -0.0219, -0.0265, -0.0355, -0.0251, -0.0419, -0.0415, -0.0415, -0.0403, -0.0213, -0.0159, -0.0275, -0.0185, -0.0022, -0.0079, -0.0037, 0.0043, 0.0098, 0.0098, -0.0031, -0.0064, -0.0005, 0.0105, 0.0108, 0.0145, 0.0, 0.0018, 0.0052, 0.0142, 0.0189, 0.0293, 0.0341, 0.017, 0.0265, 0.0325, 0.0253, 0.0189, 0.0108, -0.0136, -0.002, -0.0007, 0.02, 0.0197, 0.0215, 0.0241, 0.035, 0.0492, 0.048, 0.0455, 0.0497, 0.0492, 0.0388, 0.0327, 0.064, 0.0619, 0.0668, 0.067, 0.0688, 0.0578, 0.0578, 0.0499, 0.0542, 0.0734, 0.0732, 0.0774, 0.0759, 0.063, 0.0837, 0.0893, 0.0987, 0.1055, 0.1027, 0.0977, 0.1019, 0.0983, 0.088, 0.0956, 0.0802, 0.0961, 0.1009, 0.0958, 0.1027, 0.0947, 0.0877, 0.097, 0.0778, 0.081, 0.0756, 0.046, 0.0694, 0.0716, 0.0732, 0.0628, 0.072, 0.0805, 0.0937, 0.0851, 0.09, 0.0868, 0.1024, 0.1027, 0.0936, 0.0689, 0.0972, 0.1074, 0.0891, 0.1023, 0.0768, 0.0785, 0.0927, 0.0954, 0.0881, 0.0609, 0.0558, 0.0459, 0.0266, 0.0266, 0.0113, 0.0312, 0.0449, 0.0673, 0.0746, 0.0746, 0.0813, 0.0819, 0.0812, 0.0929, 0.089, 0.095, 0.0791, 0.079, 0.0948, 0.0991, 0.1039, 0.1133, 0.1151, 0.1022, 0.1189, 0.1144, 0.1211, 0.1195, 0.1234, 0.1234, 0.1184, 0.1074, 0.1029, 0.0942, 0.0942, 0.0845, 0.1024, 0.1095, 0.1079]}
{"unit_id": "thread:1955135584567169100", "ticker_idx": 0, "ts": "2025-08-12T05:13:42+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this can be interpreted as confidence that they would choose SK hynix or Micron over Samsung, making it difficult for Samsung to break in", "tweet_type": "analysis", "timeline": "quarters", "summary": "HBM supplier consolidation post-HBM4 favors SK Hynix and Micron over Samsung, which faces difficulty breaking into key customers like AMD and ASIC makers.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Here are some insights from Micron at the KeyBanc Forum:\n\n“With HBM becoming more customized, it’s highly likely that each customer’s HBM suppliers will be narrowed down to one or two.”\n\nWhat does this mean?\n\nI’ll explain in this thread.🧵 https://t.co/uCkpIdnVLo\n\n---\n\nMicron stated that as major customers’ demand shifts to post-HBM4 standards, the trend will likely become even more pronounced for them to work with fewer suppliers rather than diversifying their memory sources.\n\nThis is a highly provocative comment — and, indirectly, a strong criticism of Samsung.\n\nIt stands in stark contrast to the previous narrative that customers would expand their pool of HBM suppliers, leading to falling HBM prices and even talk of oversupply.\n\n---\n\nOf course, it could also be an implicit suggestion that Samsung will enter the largest demand source — Nvidia. * Most companies outside of one or two will most likely work with only one or two HBM suppliers.\n\nHowever, for companies like ASIC makers or AMD, which receive comparatively smaller volumes of HBM than Nvidia, this can be interpreted as confidence that they would choose SK hynix or Micron over Samsung, making it difficult for Samsung to break in.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0014064359978266383, "ret_signed_m1d": 0.0014064359978266383, "alpha_spy_m1d": -0.009127532622271173, "alpha_c_m1d": -0.013842760643237306, "bench_spy_m1d": -0.010533968620097811, "bench_c_m1d": -0.015249196641063945, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011251710256180214, "ret_signed_p1d": -0.011251710256180214, "alpha_spy_p1d": -0.007828658901024133, "alpha_c_p1d": -0.010878934344884295, "bench_spy_p1d": 0.0034230513551560815, "bench_c_p1d": 0.00037277591129591947, "ret_p1w": -0.01547124052322757, "ret_signed_p1w": 0.01547124052322757, "alpha_spy_p1w": 0.010989984461344693, "alpha_c_p1w": -0.009099067549844242, "bench_spy_p1w": -0.004481256061882877, "bench_c_p1w": -0.02457030807307181, "ret_p1m": 0.021097095651317677, "ret_signed_p1m": -0.021097095651317677, "alpha_spy_p1m": -0.006284432554523489, "alpha_c_p1m": -0.013864009144457512, "bench_spy_p1m": 0.014812663096794187, "bench_c_p1m": 0.007233086506860165, "ret_p3m": 0.3830773797636142, "ret_signed_p3m": -0.3830773797636142, "alpha_spy_p3m": -0.3361805897278436, "alpha_c_p3m": -0.3073146457476239, "bench_spy_p3m": 0.04689679003577063, "bench_c_p3m": 0.07576273401599032, "ret_p6m": 1.4005647526800913, "ret_signed_p6m": -1.4005647526800913, "alpha_spy_p6m": -1.3267568659184343, "alpha_c_p6m": -1.3677377380679154, "bench_spy_p6m": 0.07380788676165695, "bench_c_p6m": 0.03282701461217585, "price_path": [-0.199, -0.206, -0.22, -0.2186, -0.2214, -0.2354, -0.2423, -0.2354, -0.2465, -0.2186, -0.2158, -0.2144, -0.206, -0.206, -0.192, -0.1738, -0.1738, -0.1641, -0.1724, -0.1627, -0.1683, -0.185, -0.2004, -0.1878, -0.1641, -0.1724, -0.1683, -0.1892, -0.1543, -0.1431, -0.1585, -0.1449, -0.1589, -0.1533, -0.1449, -0.1027, -0.1097, -0.1322, -0.1364, -0.1505, -0.1421, -0.1195, -0.121, -0.1041, -0.09, -0.0619, -0.0563, -0.0464, -0.0717, -0.0661, -0.0717, -0.0731, -0.0098, -0.007, 0.0211, 0.0042, -0.0309, -0.0197, -0.0169, -0.0323, -0.0084, 0.0098, -0.0014, 0.0, 0.0113, 0.007, 0.007, -0.0155, -0.0155, -0.0084, -0.007, 0.0042, 0.0056, -0.0113, -0.007, -0.0211, -0.0197, -0.0492, -0.0281, -0.0183, -0.0141, -0.0225, -0.0141, 0.0056, 0.0211, 0.0323, 0.0605, 0.0759, 0.1167, 0.0999, 0.1294, 0.1294, 0.1744, 0.1913, 0.2011, 0.211, 0.1716, 0.1895, 0.1853, 0.215, 0.2679, 0.2679, 0.2679, 0.2679, 0.2679, 0.2679, 0.3336, 0.3181, 0.2941, 0.3421, 0.3803, 0.3831, 0.3859, 0.3774, 0.393, 0.3633, 0.3958, 0.441, 0.4057, 0.4198, 0.4707, 0.5187, 0.5696, 0.482, 0.4212, 0.4014, 0.3831, 0.4212, 0.4622, 0.4565, 0.4523, 0.3732, 0.4212, 0.3817, 0.3633, 0.4212, 0.3393, 0.3661, 0.4029, 0.4523, 0.4622, 0.4198, 0.424, 0.4608, 0.4763, 0.4848, 0.5314, 0.547, 0.5314, 0.5258, 0.5159, 0.5385, 0.4806, 0.4523, 0.5244, 0.5201, 0.5017, 0.5611, 0.5752, 0.5696, 0.5696, 0.6529, 0.6964, 0.7021, 0.7021, 0.7021, 0.8242, 0.9605, 0.9718, 1.0017, 0.9704, 0.9733, 0.9704, 0.9534, 0.9917, 1.0428, 1.1138, 1.1195, 1.0613, 1.1223, 1.1621, 1.1592, 1.1592, 1.2643, 1.3055, 1.2813, 1.2785, 1.1351, 1.3779, 1.4006]}
{"unit_id": "thread:1955135584567169100", "ticker_idx": 1, "ts": "2025-08-12T05:13:42+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "they would choose SK hynix or Micron over Samsung", "tweet_type": "analysis", "timeline": "quarters", "summary": "HBM supplier consolidation post-HBM4 favors SK Hynix and Micron over Samsung, which faces difficulty breaking into key customers like AMD and ASIC makers.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Here are some insights from Micron at the KeyBanc Forum:\n\n“With HBM becoming more customized, it’s highly likely that each customer’s HBM suppliers will be narrowed down to one or two.”\n\nWhat does this mean?\n\nI’ll explain in this thread.🧵 https://t.co/uCkpIdnVLo\n\n---\n\nMicron stated that as major customers’ demand shifts to post-HBM4 standards, the trend will likely become even more pronounced for them to work with fewer suppliers rather than diversifying their memory sources.\n\nThis is a highly provocative comment — and, indirectly, a strong criticism of Samsung.\n\nIt stands in stark contrast to the previous narrative that customers would expand their pool of HBM suppliers, leading to falling HBM prices and even talk of oversupply.\n\n---\n\nOf course, it could also be an implicit suggestion that Samsung will enter the largest demand source — Nvidia. * Most companies outside of one or two will most likely work with only one or two HBM suppliers.\n\nHowever, for companies like ASIC makers or AMD, which receive comparatively smaller volumes of HBM than Nvidia, this can be interpreted as confidence that they would choose SK hynix or Micron over Samsung, making it difficult for Samsung to break in.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007434866611495949, "ret_signed_m1d": -0.007434866611495949, "alpha_spy_m1d": 0.0030991020086018617, "alpha_c_m1d": 0.014891068062920176, "bench_spy_m1d": -0.010533968620097811, "bench_c_m1d": -0.022325934674416126, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03345730804936631, "ret_signed_p1d": 0.03345730804936631, "alpha_spy_p1d": 0.030034256694210226, "alpha_c_p1d": 0.029525213818243357, "bench_spy_p1d": 0.0034230513551560815, "bench_c_p1d": 0.003932094231122951, "ret_p1w": -0.022304658162721314, "ret_signed_p1w": -0.022304658162721314, "alpha_spy_p1w": -0.017823402100838437, "alpha_c_p1w": 0.00845171020902935, "bench_spy_p1w": -0.004481256061882877, "bench_c_p1w": -0.030756368371750664, "ret_p1m": 0.131744114506255, "ret_signed_p1m": 0.131744114506255, "alpha_spy_p1m": 0.11693145140946082, "alpha_c_p1m": 0.12931163677685054, "bench_spy_p1m": 0.014812663096794187, "bench_c_p1m": 0.0024324777294044697, "ret_p3m": 1.1592484430444632, "ret_signed_p3m": 1.1592484430444632, "alpha_spy_p3m": 1.1123516530086925, "alpha_c_p3m": 0.9992350473816507, "bench_spy_p3m": 0.04689679003577063, "bench_c_p3m": 0.16001339566281247, "ret_p6m": 2.3529574397545177, "ret_signed_p6m": 2.3529574397545177, "alpha_spy_p6m": 2.2791495529928607, "alpha_c_p6m": 2.076020778532687, "bench_spy_p6m": 0.07380788676165695, "bench_c_p6m": 0.2769366612218309, "price_path": [-0.256, -0.2411, -0.2601, -0.2504, -0.256, -0.2693, -0.2578, -0.2467, -0.2486, -0.2282, -0.2119, -0.2398, -0.2286, -0.2286, -0.1914, -0.1654, -0.1654, -0.1487, -0.1431, -0.1078, -0.1245, -0.1245, -0.0781, -0.0743, -0.0836, -0.0855, -0.0446, -0.0353, 0.0353, 0.0632, 0.0892, 0.0558, 0.0855, 0.0613, 0.0372, 0.0353, 0.0056, 0.0074, 0.0483, 0.0446, 0.1041, 0.0948, 0.1152, 0.1097, 0.1004, 0.0019, 0.0, 0.013, -0.0019, 0.0, 0.0019, -0.0112, -0.026, -0.0242, -0.0204, 0.0167, -0.0409, -0.0409, -0.0204, -0.039, -0.026, -0.0465, -0.0074, 0.0, 0.0335, 0.0279, 0.0279, -0.0056, -0.0223, -0.0502, -0.0892, -0.0669, -0.0353, -0.0279, -0.0335, -0.0004, 0.0014, -0.047, -0.0302, -0.0228, -0.0116, 0.0182, 0.0312, 0.0722, 0.1317, 0.1429, 0.223, 0.2323, 0.2955, 0.2416, 0.3235, 0.3235, 0.3067, 0.3439, 0.3309, 0.3272, 0.2527, 0.2993, 0.2937, 0.3402, 0.4724, 0.4724, 0.4724, 0.4724, 0.4724, 0.4724, 0.5934, 0.545, 0.5319, 0.5729, 0.6846, 0.733, 0.8074, 0.7832, 0.7925, 0.7814, 0.8986, 0.9917, 0.9396, 1.0773, 1.1146, 1.0811, 1.3082, 1.1816, 1.1555, 1.2076, 1.1592, 1.256, 1.3044, 1.297, 1.2784, 1.0848, 1.256, 1.122, 1.0922, 1.1257, 0.9396, 0.9359, 0.9322, 0.9508, 1.0267, 0.9745, 1.0043, 1.0788, 1.0565, 1.0192, 1.0267, 1.1496, 1.1086, 1.1869, 1.1049, 1.1273, 1.0639, 0.9745, 1.0528, 1.0565, 1.0379, 1.1608, 1.1757, 1.1906, 1.1906, 1.2316, 1.3843, 1.4253, 1.4253, 1.4253, 1.5222, 1.593, 1.7047, 1.7643, 1.8165, 1.7718, 1.7904, 1.7494, 1.7643, 1.7904, 1.8165, 1.8463, 1.7681, 1.7569, 1.8128, 1.8575, 1.742, 1.9804, 2.1332, 2.2077, 2.3865, 2.0922, 2.379, 2.353]}
{"unit_id": "thread:1955135584567169100", "ticker_idx": 2, "ts": "2025-08-12T05:13:42+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "they would choose SK hynix or Micron over Samsung", "tweet_type": "analysis", "timeline": "quarters", "summary": "HBM supplier consolidation post-HBM4 favors SK Hynix and Micron over Samsung, which faces difficulty breaking into key customers like AMD and ASIC makers.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Here are some insights from Micron at the KeyBanc Forum:\n\n“With HBM becoming more customized, it’s highly likely that each customer’s HBM suppliers will be narrowed down to one or two.”\n\nWhat does this mean?\n\nI’ll explain in this thread.🧵 https://t.co/uCkpIdnVLo\n\n---\n\nMicron stated that as major customers’ demand shifts to post-HBM4 standards, the trend will likely become even more pronounced for them to work with fewer suppliers rather than diversifying their memory sources.\n\nThis is a highly provocative comment — and, indirectly, a strong criticism of Samsung.\n\nIt stands in stark contrast to the previous narrative that customers would expand their pool of HBM suppliers, leading to falling HBM prices and even talk of oversupply.\n\n---\n\nOf course, it could also be an implicit suggestion that Samsung will enter the largest demand source — Nvidia. * Most companies outside of one or two will most likely work with only one or two HBM suppliers.\n\nHowever, for companies like ASIC makers or AMD, which receive comparatively smaller volumes of HBM than Nvidia, this can be interpreted as confidence that they would choose SK hynix or Micron over Samsung, making it difficult for Samsung to break in.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03154599885754861, "ret_signed_m1d": -0.03154599885754861, "alpha_spy_m1d": -0.021012030237450796, "alpha_c_m1d": -0.009220064183132481, "bench_spy_m1d": -0.010533968620097811, "bench_c_m1d": -0.022325934674416126, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027240783850084882, "ret_signed_p1d": -0.027240783850084882, "alpha_spy_p1d": -0.030663835205240964, "alpha_c_p1d": -0.031172878081207833, "bench_spy_p1d": 0.0034230513551560815, "bench_c_p1d": 0.003932094231122951, "ret_p1w": -0.04461840049994659, "ret_signed_p1w": -0.04461840049994659, "alpha_spy_p1w": -0.04013714443806371, "alpha_c_p1w": -0.013862032128195922, "bench_spy_p1w": -0.004481256061882877, "bench_c_p1w": -0.030756368371750664, "ret_p1m": 0.0958904945261676, "ret_signed_p1m": 0.0958904945261676, "alpha_spy_p1m": 0.08107783142937341, "alpha_c_p1m": 0.09345801679676313, "bench_spy_p1m": 0.014812663096794187, "bench_c_p1m": 0.0024324777294044697, "ret_p3m": 0.8635536791544258, "ret_signed_p3m": 0.8635536791544258, "alpha_spy_p3m": 0.8166568891186552, "alpha_c_p3m": 0.7035402834916133, "bench_spy_p3m": 0.04689679003577063, "bench_c_p3m": 0.16001339566281247, "ret_p6m": 1.9729234487628635, "ret_signed_p6m": 1.9729234487628635, "alpha_spy_p6m": 1.8991155620012066, "alpha_c_p6m": 1.6959867875410326, "bench_spy_p6m": 0.07380788676165695, "bench_c_p6m": 0.2769366612218309, "price_path": [-0.2535, -0.2336, -0.2285, -0.2328, -0.2505, -0.2584, -0.2698, -0.2698, -0.2463, -0.2478, -0.243, -0.2613, -0.2322, -0.2004, -0.1925, -0.1688, -0.151, -0.1323, -0.1074, -0.0926, -0.0914, -0.096, -0.0628, -0.0589, -0.0473, -0.0473, -0.0334, -0.0453, 0.0003, -0.0049, -0.0146, -0.0243, -0.0361, -0.0546, -0.0479, -0.0436, -0.0436, -0.0613, -0.0261, -0.0431, -0.0363, -0.0252, -0.0715, -0.0598, -0.0886, -0.1134, -0.1046, -0.1137, -0.145, -0.1403, -0.1254, -0.1291, -0.1292, -0.1236, -0.1018, -0.1457, -0.179, -0.1564, -0.1463, -0.1485, -0.1243, -0.0694, -0.0315, 0.0, -0.0272, -0.0193, -0.0539, -0.0329, -0.0446, -0.0825, -0.0936, -0.0788, -0.0887, -0.0881, -0.0783, -0.045, -0.0684, -0.0684, -0.0726, -0.0707, -0.0277, 0.0283, 0.029, 0.0586, 0.0959, 0.1786, 0.2308, 0.235, 0.2432, 0.2524, 0.322, 0.2738, 0.2886, 0.3026, 0.2658, 0.2276, 0.2311, 0.283, 0.3097, 0.4258, 0.4384, 0.4712, 0.4957, 0.4545, 0.5394, 0.5065, 0.4224, 0.5099, 0.4652, 0.5034, 0.5864, 0.5852, 0.6196, 0.5845, 0.5546, 0.6191, 0.7155, 0.724, 0.7382, 0.7751, 0.7546, 0.7527, 0.8383, 0.7078, 0.8603, 0.8668, 0.8636, 0.984, 0.8885, 0.9182, 0.856, 0.9333, 0.8951, 0.7898, 0.7696, 0.5773, 0.6243, 0.754, 0.7587, 0.8036, 0.8036, 0.8523, 0.8834, 0.8759, 0.8341, 0.7753, 0.8581, 0.934, 0.9771, 1.0656, 1.0244, 0.8888, 0.8603, 0.8212, 0.7664, 0.9468, 1.0829, 1.1664, 1.1639, 1.2455, 1.2455, 1.2307, 1.3066, 1.293, 1.2364, 1.2364, 1.4716, 1.446, 1.6911, 1.6607, 1.5625, 1.7041, 1.7102, 1.6495, 1.6121, 1.6378, 1.8425, 1.8425, 1.8601, 2.049, 2.1154, 2.1316, 2.0489, 2.2146, 2.4108, 2.4148, 2.2509, 2.4305, 2.2867, 1.9729]}
{"unit_id": "orig:1952501933706821732", "ticker_idx": 0, "ts": "2025-08-04T22:48:31+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MSFT provided guidance for F1Q26 Capex at $30B (an increase from the consensus of $23B)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley sharply revises up 2025/2026 global cloud capex forecasts; all four hyperscalers signal surging AI infrastructure investment above consensus.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Global Cloud Capex Outlook: Sharp Upward Revision (Morgan Stanley)\n\n- Global cloud Capex for 2025 is estimated at $445B, a +56% Y/Y increase. This figure represents a +12 percentage point (approx. $35B) upward revision from two weeks ago.\n- The majority of this upward revision is driven by Amazon, Microsoft, Alphabet, and Meta, which account for 77% of the total Capex growth.\n\n2026 Outlook: Significantly Higher Upside Than Consensus\n\n- While the 2026 consensus forecast is for +16% Y/Y growth ($518B), Morgan Stanley's estimate is for +31% Y/Y growth ($582B), projecting a Capex-to-revenue ratio of 21.6%—an all-time high.\n- Morgan Stanley believes there is a high probability that Capex estimates will be revised further upward within the next 12 months.\n\nSummary of Messages from the Big 4 Hyperscalers\n\n- All four—MSFT, META, AMZN, and GOOGL—emphasized clear expectations for a return on investment from their Capex expansion and stressed the necessity of responding to the surging demand for AI infrastructure.\n- MSFT provided guidance for F1Q26 Capex at $30B (an increase from the consensus of $23B).\n- GOOGL raised its CY25 Capex guidance to $85B.\n- META raised the lower end of its Capex guidance to $66B.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021525586457659895, "ret_signed_m1d": -0.021525586457659895, "alpha_spy_m1d": -0.006553347518565178, "alpha_c_m1d": -0.006553347518565178, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.014972238939094717, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014730112378469884, "ret_signed_p1d": -0.014730112378469884, "alpha_spy_p1d": -0.009660145806038578, "alpha_c_p1d": -0.009660145806038578, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.005069966572431306, "ret_p1w": -0.025894335225087106, "ret_signed_p1w": -0.025894335225087106, "alpha_spy_p1w": -0.033420097331621657, "alpha_c_p1w": -0.033420097331621657, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.00752576210653455, "ret_p1m": -0.0554283564139092, "ret_signed_p1m": -0.0554283564139092, "alpha_spy_p1m": -0.0698460860444935, "alpha_c_p1m": -0.0698460860444935, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.014417729630584297, "ret_p3m": -0.016831673956911875, "ret_signed_p3m": -0.016831673956911875, "alpha_spy_p3m": -0.09691282384889188, "alpha_c_p3m": -0.09691282384889188, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.08008114989198001, "ret_p6m": -0.09963597971740967, "ret_signed_p6m": -0.09963597971740967, "alpha_spy_p6m": -0.20786203656426838, "alpha_c_p6m": -0.20786203656426838, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.10822605684685871, "price_path": [-0.1925, -0.1835, -0.1824, -0.1628, -0.163, -0.1559, -0.154, -0.1519, -0.1433, -0.1446, -0.1551, -0.1508, -0.1595, -0.1595, -0.1399, -0.1461, -0.1437, -0.1405, -0.1375, -0.1357, -0.134, -0.1269, -0.1218, -0.1174, -0.1208, -0.1177, -0.106, -0.1133, -0.1055, -0.1075, -0.1034, -0.1034, -0.1087, -0.0927, -0.085, -0.081, -0.0713, -0.0741, -0.0714, -0.0814, -0.0832, -0.0687, -0.0687, -0.0708, -0.0728, -0.06, -0.0638, -0.0603, -0.0609, -0.0557, -0.056, -0.0447, -0.0478, -0.0478, -0.0567, -0.0556, -0.0462, -0.0409, -0.0432, -0.0431, -0.0418, -0.004, -0.0215, 0.0, -0.0147, -0.02, -0.0276, -0.0254, -0.0259, -0.0119, -0.0281, -0.0246, -0.0289, -0.0346, -0.0483, -0.0559, -0.0571, -0.0515, -0.057, -0.0612, -0.0524, -0.047, -0.0525, -0.0525, -0.0554, -0.055, -0.0501, -0.0744, -0.0684, -0.068, -0.0643, -0.0631, -0.0465, -0.0363, -0.0481, -0.0463, -0.0492, -0.0315, -0.038, -0.0477, -0.046, -0.0519, -0.0436, -0.0377, -0.0314, -0.0281, -0.0356, -0.0326, -0.0116, -0.0202, -0.0185, -0.0231, -0.0445, -0.0387, -0.0396, -0.0399, -0.0433, -0.0396, -0.0336, -0.032, -0.0266, -0.0266, -0.0209, -0.0061, 0.0137, 0.0127, -0.0168, -0.0317, -0.0332, -0.0382, -0.0516, -0.0704, -0.0709, -0.0538, -0.0488, -0.0442, -0.0589, -0.046, -0.051, -0.0766, -0.0891, -0.1037, -0.1155, -0.112, -0.1064, -0.0904, -0.0904, -0.0782, -0.0881, -0.082, -0.105, -0.0991, -0.0948, -0.0801, -0.0782, -0.1034, -0.0942, -0.1035, -0.1104, -0.1075, -0.108, -0.0933, -0.0896, -0.0915, -0.0879, -0.0857, -0.0857, -0.0863, -0.0874, -0.0867, -0.0939, -0.0939, -0.1139, -0.1141, -0.1035, -0.0942, -0.1043, -0.1021, -0.106, -0.1182, -0.1394, -0.1444, -0.1385, -0.1385, -0.1485, -0.168, -0.1548, -0.127, -0.1189, -0.0996]}
{"unit_id": "orig:1952501933706821732", "ticker_idx": 1, "ts": "2025-08-04T22:48:31+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "META raised the lower end of its Capex guidance to $66B", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley sharply revises up 2025/2026 global cloud capex forecasts; all four hyperscalers signal surging AI infrastructure investment above consensus.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Global Cloud Capex Outlook: Sharp Upward Revision (Morgan Stanley)\n\n- Global cloud Capex for 2025 is estimated at $445B, a +56% Y/Y increase. This figure represents a +12 percentage point (approx. $35B) upward revision from two weeks ago.\n- The majority of this upward revision is driven by Amazon, Microsoft, Alphabet, and Meta, which account for 77% of the total Capex growth.\n\n2026 Outlook: Significantly Higher Upside Than Consensus\n\n- While the 2026 consensus forecast is for +16% Y/Y growth ($518B), Morgan Stanley's estimate is for +31% Y/Y growth ($582B), projecting a Capex-to-revenue ratio of 21.6%—an all-time high.\n- Morgan Stanley believes there is a high probability that Capex estimates will be revised further upward within the next 12 months.\n\nSummary of Messages from the Big 4 Hyperscalers\n\n- All four—MSFT, META, AMZN, and GOOGL—emphasized clear expectations for a return on investment from their Capex expansion and stressed the necessity of responding to the surging demand for AI infrastructure.\n- MSFT provided guidance for F1Q26 Capex at $30B (an increase from the consensus of $23B).\n- GOOGL raised its CY25 Capex guidance to $85B.\n- META raised the lower end of its Capex guidance to $66B.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03395285653710489, "ret_signed_m1d": -0.03395285653710489, "alpha_spy_m1d": -0.018980617598010174, "alpha_c_m1d": -0.015091832860982701, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.01886102367612219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016628680790430894, "ret_signed_p1d": -0.016628680790430894, "alpha_spy_p1d": -0.011558714217999588, "alpha_c_p1d": -0.00886233082744714, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.007766349962983754, "ret_p1w": -0.013524484913330692, "ret_signed_p1w": -0.013524484913330692, "alpha_spy_p1w": -0.021050247019865242, "alpha_c_p1w": -0.01370928097086277, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.00018479605753207728, "ret_p1m": -0.05314484521691187, "ret_signed_p1m": -0.05314484521691187, "alpha_spy_p1m": -0.06756257484749617, "alpha_c_p1m": -0.08125122847453148, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.028106383257619605, "ret_p3m": -0.14097689067104546, "ret_signed_p3m": -0.14097689067104546, "alpha_spy_p3m": -0.22105804056302547, "alpha_c_p3m": -0.20161564387185005, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.060638753200804585, "ret_p6m": -0.13189148545067264, "ret_signed_p6m": -0.13189148545067264, "alpha_spy_p6m": -0.24011754229753135, "alpha_c_p6m": -0.2199781523795018, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.08808666692882916, "price_path": [-0.2319, -0.2303, -0.2374, -0.177, -0.1557, -0.1514, -0.1713, -0.1758, -0.1757, -0.18, -0.1821, -0.1807, -0.1929, -0.1929, -0.1733, -0.1717, -0.1698, -0.1666, -0.1365, -0.1417, -0.1146, -0.1189, -0.102, -0.1067, -0.096, -0.1066, -0.1076, -0.1211, -0.0956, -0.1019, -0.1038, -0.1038, -0.1211, -0.1003, -0.0827, -0.0872, -0.0648, -0.0551, -0.0493, -0.0736, -0.0809, -0.0739, -0.0739, -0.0747, -0.0717, -0.0561, -0.0633, -0.0758, -0.0714, -0.085, -0.0946, -0.0966, -0.0929, -0.0817, -0.0922, -0.0809, -0.0793, -0.082, -0.0757, -0.0984, -0.1045, -0.0038, -0.034, 0.0, -0.0166, -0.0056, -0.0187, -0.0091, -0.0135, 0.0176, 0.0048, 0.0074, 0.0114, -0.0116, -0.0321, -0.0369, -0.048, -0.0278, -0.0297, -0.0287, -0.0373, -0.0325, -0.0485, -0.0485, -0.0531, -0.0506, -0.0357, -0.0308, -0.031, -0.0137, -0.0314, -0.0328, -0.0268, -0.015, 0.0034, -0.0008, 0.005, 0.0026, -0.0138, -0.0264, -0.0196, -0.0347, -0.0414, -0.0418, -0.0534, -0.0754, -0.0629, -0.0841, -0.0776, -0.0809, -0.0748, -0.0546, -0.0909, -0.0775, -0.0866, -0.0751, -0.0822, -0.076, -0.0563, -0.0549, -0.0547, -0.0539, -0.0483, -0.0323, -0.0315, -0.0312, -0.141, -0.1643, -0.178, -0.1914, -0.1803, -0.2022, -0.1987, -0.1857, -0.1917, -0.215, -0.2139, -0.2145, -0.2241, -0.2296, -0.2391, -0.2406, -0.2341, -0.2098, -0.18, -0.1833, -0.1833, -0.1648, -0.174, -0.1659, -0.1756, -0.1473, -0.132, -0.1406, -0.1532, -0.162, -0.1587, -0.1696, -0.1647, -0.1523, -0.1622, -0.1429, -0.1502, -0.1467, -0.1422, -0.1389, -0.1389, -0.1444, -0.1503, -0.1409, -0.1485, -0.1485, -0.161, -0.1502, -0.1478, -0.1632, -0.1666, -0.1576, -0.1719, -0.1859, -0.206, -0.1992, -0.1999, -0.1999, -0.2207, -0.2093, -0.1646, -0.1502, -0.1327, -0.1319]}
{"unit_id": "orig:1952501933706821732", "ticker_idx": 2, "ts": "2025-08-04T22:48:31+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Amazon, Microsoft, Alphabet, and Meta, which account for 77% of the total Capex growth", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley sharply revises up 2025/2026 global cloud capex forecasts; all four hyperscalers signal surging AI infrastructure investment above consensus.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Global Cloud Capex Outlook: Sharp Upward Revision (Morgan Stanley)\n\n- Global cloud Capex for 2025 is estimated at $445B, a +56% Y/Y increase. This figure represents a +12 percentage point (approx. $35B) upward revision from two weeks ago.\n- The majority of this upward revision is driven by Amazon, Microsoft, Alphabet, and Meta, which account for 77% of the total Capex growth.\n\n2026 Outlook: Significantly Higher Upside Than Consensus\n\n- While the 2026 consensus forecast is for +16% Y/Y growth ($518B), Morgan Stanley's estimate is for +31% Y/Y growth ($582B), projecting a Capex-to-revenue ratio of 21.6%—an all-time high.\n- Morgan Stanley believes there is a high probability that Capex estimates will be revised further upward within the next 12 months.\n\nSummary of Messages from the Big 4 Hyperscalers\n\n- All four—MSFT, META, AMZN, and GOOGL—emphasized clear expectations for a return on investment from their Capex expansion and stressed the necessity of responding to the surging demand for AI infrastructure.\n- MSFT provided guidance for F1Q26 Capex at $30B (an increase from the consensus of $23B).\n- GOOGL raised its CY25 Capex guidance to $85B.\n- META raised the lower end of its Capex guidance to $66B.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014646851844615671, "ret_signed_m1d": 0.014646851844615671, "alpha_spy_m1d": 0.02961909078371039, "alpha_c_m1d": 0.026264707857832525, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.011617856013216854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009922070229506863, "ret_signed_p1d": 0.009922070229506863, "alpha_spy_p1d": 0.014992036801938169, "alpha_c_p1d": 0.00864136003446414, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": 0.0012807101950427224, "ret_p1w": 0.04559418584246755, "ret_signed_p1w": 0.04559418584246755, "alpha_spy_p1w": 0.038068423735933, "alpha_c_p1w": 0.019385499363271386, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.026208686479196164, "ret_p1m": 0.06468227184595099, "ret_signed_p1m": 0.06468227184595099, "alpha_spy_p1m": 0.05026454221536669, "alpha_c_p1m": 0.012630896309151263, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.05205137553679973, "ret_p3m": 0.052964833626925945, "ret_signed_p3m": 0.052964833626925945, "alpha_spy_p3m": -0.027116316265054063, "alpha_c_p3m": -0.018078817777868128, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.07104365140479407, "ret_p6m": 0.1560595309794881, "ret_signed_p6m": 0.1560595309794881, "alpha_spy_p6m": 0.047833474132629394, "alpha_c_p6m": 0.029498970339387087, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.12656056064010102, "price_path": [-0.1084, -0.0925, -0.0878, -0.0142, -0.0013, -0.0066, -0.0306, -0.0286, -0.0259, -0.0358, -0.0498, -0.0404, -0.0504, -0.0504, -0.0266, -0.0327, -0.0281, -0.0314, -0.0236, -0.0281, -0.0209, -0.0177, 0.0091, 0.0252, 0.0282, 0.0073, 0.0075, 0.0021, 0.021, 0.015, 0.0041, 0.0041, -0.0093, -0.015, 0.0053, 0.0016, 0.0258, 0.055, 0.0366, 0.0416, 0.0391, 0.0556, 0.0556, 0.0558, 0.0364, 0.0515, 0.0501, 0.0632, 0.0663, 0.0695, 0.0545, 0.0578, 0.0684, 0.0834, 0.0747, 0.0786, 0.0972, 0.0935, 0.0999, 0.0915, 0.0876, 0.1061, 0.0146, 0.0, 0.0099, 0.0504, 0.0542, 0.0522, 0.0456, 0.0464, 0.061, 0.0913, 0.0916, 0.0937, 0.0773, 0.0575, 0.0487, 0.0812, 0.077, 0.0806, 0.0825, 0.0943, 0.082, 0.082, 0.0647, 0.0678, 0.1135, 0.0977, 0.1143, 0.1256, 0.0883, 0.0865, 0.078, 0.0935, 0.1058, 0.0944, 0.0925, 0.0937, 0.0755, 0.0428, 0.0404, 0.0307, 0.0384, 0.0497, 0.0374, 0.0424, 0.0508, 0.0371, 0.0437, 0.0479, 0.0641, 0.076, 0.0223, 0.0398, 0.0224, 0.0185, 0.0133, 0.0066, 0.0228, 0.049, 0.0298, 0.0446, 0.0593, 0.0724, 0.0832, 0.0881, 0.053, 0.1539, 0.2001, 0.178, 0.1821, 0.1483, 0.1548, 0.1736, 0.1769, 0.1538, 0.1225, 0.1089, 0.1003, 0.0515, 0.0522, 0.0259, 0.0427, 0.0691, 0.0851, 0.0827, 0.0827, 0.1019, 0.105, 0.1076, 0.0979, 0.0825, 0.0845, 0.072, 0.0769, 0.0951, 0.088, 0.0687, 0.0515, 0.0515, 0.0455, 0.0714, 0.0742, 0.0793, 0.0968, 0.0979, 0.0979, 0.0986, 0.0965, 0.0987, 0.0906, 0.0906, 0.0702, 0.1012, 0.1383, 0.1413, 0.1637, 0.1688, 0.1645, 0.1462, 0.1181, 0.1253, 0.1298, 0.1298, 0.0914, 0.0929, 0.1072, 0.13, 0.1265, 0.1561]}
{"unit_id": "orig:1952501933706821732", "ticker_idx": 3, "ts": "2025-08-04T22:48:31+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GOOGL raised its CY25 Capex guidance to $85B", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley sharply revises up 2025/2026 global cloud capex forecasts; all four hyperscalers signal surging AI infrastructure investment above consensus.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Global Cloud Capex Outlook: Sharp Upward Revision (Morgan Stanley)\n\n- Global cloud Capex for 2025 is estimated at $445B, a +56% Y/Y increase. This figure represents a +12 percentage point (approx. $35B) upward revision from two weeks ago.\n- The majority of this upward revision is driven by Amazon, Microsoft, Alphabet, and Meta, which account for 77% of the total Capex growth.\n\n2026 Outlook: Significantly Higher Upside Than Consensus\n\n- While the 2026 consensus forecast is for +16% Y/Y growth ($518B), Morgan Stanley's estimate is for +31% Y/Y growth ($582B), projecting a Capex-to-revenue ratio of 21.6%—an all-time high.\n- Morgan Stanley believes there is a high probability that Capex estimates will be revised further upward within the next 12 months.\n\nSummary of Messages from the Big 4 Hyperscalers\n\n- All four—MSFT, META, AMZN, and GOOGL—emphasized clear expectations for a return on investment from their Capex expansion and stressed the necessity of responding to the surging demand for AI infrastructure.\n- MSFT provided guidance for F1Q26 Capex at $30B (an increase from the consensus of $23B).\n- GOOGL raised its CY25 Capex guidance to $85B.\n- META raised the lower end of its Capex guidance to $66B.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030301348041262388, "ret_signed_m1d": -0.030301348041262388, "alpha_spy_m1d": -0.01532910910216767, "alpha_c_m1d": -0.011440324365140198, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.01886102367612219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0018969853747939913, "ret_signed_p1d": -0.0018969853747939913, "alpha_spy_p1d": 0.0031729811976373146, "alpha_c_p1d": 0.005869364588189763, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.007766349962983754, "ret_p1w": 0.030557907217318547, "ret_signed_p1w": 0.030557907217318547, "alpha_spy_p1w": 0.023032145110783997, "alpha_c_p1w": 0.03037311115978647, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.00018479605753207728, "ret_p1m": 0.08362394227167624, "ret_signed_p1m": 0.08362394227167624, "alpha_spy_p1m": 0.06920621264109195, "alpha_c_p1m": 0.05551755901405664, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.028106383257619605, "ret_p3m": 0.444482107624222, "ret_signed_p3m": 0.444482107624222, "alpha_spy_p3m": 0.364400957732242, "alpha_c_p3m": 0.38384335442341744, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.060638753200804585, "ret_p6m": 0.7179462751598731, "ret_signed_p6m": 0.7179462751598731, "alpha_spy_p6m": 0.6097202183130144, "alpha_c_p6m": 0.6298596082310439, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.08808666692882916, "price_path": [-0.2248, -0.2099, -0.2178, -0.1885, -0.1831, -0.1531, -0.1604, -0.1489, -0.1472, -0.1603, -0.1368, -0.125, -0.1373, -0.1373, -0.1146, -0.1174, -0.1199, -0.1205, -0.1344, -0.149, -0.1394, -0.1386, -0.1106, -0.0972, -0.0843, -0.0907, -0.0992, -0.1044, -0.0937, -0.0979, -0.1114, -0.1114, -0.1456, -0.153, -0.1449, -0.1249, -0.1102, -0.0846, -0.0964, -0.0984, -0.0841, -0.0795, -0.0795, -0.0936, -0.106, -0.0944, -0.0893, -0.0761, -0.0691, -0.0669, -0.0619, -0.0588, -0.0512, -0.0253, -0.019, -0.0247, -0.0147, -0.0095, -0.0126, 0.0036, 0.0076, -0.0161, -0.0303, 0.0, -0.0019, 0.0054, 0.0076, 0.0327, 0.0306, 0.0426, 0.0355, 0.0405, 0.0454, 0.0434, 0.0335, 0.0219, 0.0241, 0.0567, 0.069, 0.062, 0.0638, 0.0851, 0.0916, 0.0916, 0.0836, 0.1826, 0.191, 0.2049, 0.201, 0.2297, 0.2274, 0.2335, 0.2357, 0.2912, 0.2889, 0.2805, 0.2934, 0.3072, 0.2959, 0.2915, 0.2683, 0.2613, 0.2652, 0.2524, 0.2475, 0.2568, 0.2608, 0.2591, 0.2851, 0.2612, 0.2553, 0.2395, 0.214, 0.2529, 0.2596, 0.2882, 0.2904, 0.2999, 0.3165, 0.2853, 0.2916, 0.2987, 0.3338, 0.3818, 0.3726, 0.409, 0.4445, 0.443, 0.456, 0.4243, 0.459, 0.4613, 0.4309, 0.4887, 0.4949, 0.4713, 0.4295, 0.4185, 0.4626, 0.4589, 0.5026, 0.4854, 0.5378, 0.6349, 0.6598, 0.6419, 0.6419, 0.6431, 0.6159, 0.6207, 0.6403, 0.6299, 0.6487, 0.611, 0.6282, 0.6443, 0.6044, 0.5882, 0.5827, 0.5743, 0.5237, 0.5532, 0.5773, 0.5908, 0.6142, 0.6129, 0.6129, 0.6099, 0.6102, 0.6117, 0.6073, 0.6073, 0.6183, 0.6255, 0.6142, 0.6534, 0.6712, 0.6872, 0.7041, 0.7252, 0.7246, 0.7089, 0.6946, 0.6946, 0.6535, 0.6863, 0.6974, 0.684, 0.7113, 0.7179]}
{"unit_id": "orig:1958315188643598527", "ticker_idx": 0, "ts": "2025-08-20T23:48:19+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "we maintain our preference ranking of Asian memory makers as SKH > SEC > Nanya. We continue to support SKH's technology leadership in HBM and believe concerns about 2026 OP declines are excessive", "tweet_type": "analysis", "timeline": "months", "summary": "JPMorgan maintains positive view on memory: SKH top pick on HBM tech lead, SEC OW on legacy upcycle underpricing + HBM4 upside, MU/DRAM ASP strength into 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250821_Memory Market Update: HBM FAQs – JPMorgan\n\n(1) In this report, we aim to address recent investor FAQs related to the HBM industry.\n\n(2) The most frequently asked questions are:\n\n2026 pricing outlook,\nchanges in logic die specifications and design ownership,\nthe impact of CSP CAPEX expansion on HBM supply/demand.\n(3) In conclusion, we maintain our preference ranking of Asian memory makers as SKH > SEC > Nanya.\n\n[HBM Base Die]\n\n(4) Who is responsible for the HBM4E base die design? Why does NVIDIA want to directly control base die design?\n\n(5) Up to HBM3E, since the design involved I/O circuit structure, memory makers traditionally handled the die design.\n\n(6) However, starting with HBM4, the role and significance of this design has become much more important in controlling HBM performance.\n\n(7) Recently, as reported by TrendForce [NVIDIA direct entry], discussions regarding customers [NVIDIA = memory makers’ customer] entering base die design are being actively debated among investors.\n\n(8) In other words, NVIDIA may directly design the logic die for HBM4E at the 3nm process starting from late 2027.\n\n(9) We believe NVIDIA’s insourcing signals 1) cost control [one of the main drivers of high BOM cost in HBM4], and 2) supply chain diversification.\n\n[Are Memory Makers Being Excluded?]\n\n(10) We believe it is an over-interpretation to say that memory makers are being excluded from HBM.\n\n(11) In HBM4E, customers will have two choices:\n\n(12) First, custom HBM dies embedding the GPU/ASIC customer’s core IP to pursue “performance differentiation.”\n\n(13) Second, generic EKDLs where memory makers continue to handle design as they do today.\n\n(14) Embedding IP into custom logic dies is important, but at the same time, integrating and stacking them into HBM is also critical.\n\n(15) With HBM4E expanding to 16-high stacks, the stable yield and mass production capability of memory makers will become even more essential to supply chain stability.\n\n(16) Therefore, it is too early to conclude that the value of memory makers will disappear or that HBM will become commoditized.\n\n[Rising Importance of Thermal Management in HBM4]\n\n(17) As HBM specification requirements increase, thermal characteristics are becoming even more important.\n\n(18) Recently, customers have requested that memory makers raise HBM4 reading speed from ~8Gbps to ~10Gbps.\n\n(19) This is because, with double the I/Os versus HBM3E, both performance and thermal control become far more critical.\n\n(20) SKH’s advanced MR-MUF and Micron’s strong base die performance are currently viewed as superior to SEC up to the 5th generation HBM in terms of thermal management.\n\n(21) NVIDIA’s chips consume ~700W per chip, far higher than other GPU/ASICs [300–400W].\n\n(22) Therefore, the leading suppliers who can meet NVIDIA’s upgrade requests and thermal requirements will likely remain in a favorable position.\n\n[HBM4 Qualification Tests: SKH Advantage Continues]\n\n(23) SKH is about to begin customer sampling, with results expected around November this year.\n\n(24) Meanwhile, Micron and SEC will start customer qualification in November, with results expected by February 2026 at the earliest.\n\n(25) If the JEDEC standard [HBM4: ~8Gbps] does not change, this spec change will be optional rather than mandatory.\n\n(26) However, if stricter spec requirements are added, the qualification process could lengthen, and the 2Q25 GPU launch schedule [with HBM4 onboard] could become challenging.\n\n(27) If the launch of the new Rubin GPU slips to late 2026, SKH’s early qualification advantage may be diluted in its competition with MU/SEC.\n\n[2026 HBM Pricing Outlook]\n\n(28) We maintain our HBM4 ASP outlook at $19/GB, $2.4/Gb.\n\n(29) The rationale is: 1) wafer die penalty from large die sizes, 2) yield loss from increased I/Os, 3) additional new logic die costs compared to prior generations.\n\n(30) SKH management has also stated that “profitability maintenance” is the top priority for next-gen HBM, making a +30–40% premium vs. HBM3E reasonable.\n\n(31) By contrast, the HBM3E market shows signs of easing.\n\n(32) SEC is expected to start shipping HBM3E 12-high to Broadcom/AMD in 3Q25.\n\n(33) We expect HBM3E bit mix to peak by 1H26, then gradually decline from 3Q26.\n\n(34) We estimate 2026 HBM3E 12-high ASP to fall -11% YoY.\n\nSKH: 4Q25 -7% → 1Q26 -7% → 2Q26 -7% → 4Q26 -10%\n\n(35) For 2026, bull/bear case HBM operating profit is $18.2bn / $12.5bn respectively.\n\n[CSP CAPEX Update]\n\n(36) CSP CAPEX expansion is positively impacting HBM bit demand, with ASIC demand growth and B30A as key variables.\n\n(37) Major firms reported 2Q25 cloud and AI revenue above expectations and raised CAPEX.\n\n(38) While noise around AI ROIC persists, CAPEX trends remain upward.\n\n(39) Our channel checks indicate HBM suppliers also see stronger procurement needs across customers versus 3 months ago, and that 2026 volumes will largely be decided in current CAPEX negotiations.\n\n(40) In 2026–27, if ASIC demand rises +10%, HBM bit demand would increase by +3% and +2% respectively.\n\n(41) Memory makers have recently detected production codes for B30A, which is reportedly targeting 2026 mass production.\n\n(42) These products could use HBM3E 96GB or 144GB.\n\n(43) If 1,000K chips are produced in 2026, this could add +3–4% to HBM bit demand.\n\n[DRAM Outlook]\n\n(44) The DRAM market is expected to remain tight into early 2026, driven by strong GPU server demand.\n\n(45) Beyond HBM, ASP uptrends are also appearing in legacy DRAM/NAND.\n\n(46) Micron recently raised guidance, which we attribute to D4/D5 price strength.\n\n(47) ASP strength is expected to continue through 4Q25, with server customers likely to further raise GP server purchases over the next 6 months.\n\n(48) Rather than the mild year-end decline initially expected, demand recovery looks stronger, and visibility into early 2026 is improving.\n\n[Positive View on Memory Sector Maintained]\n\n(49) We maintain a medium-term positive view on memory stocks.\n\n(50) The main market debate centers on 2026 HBM ASP and supply-demand, with 2026 volume and price negotiations to be announced within 1–2 months.\n\n(51) We continue to support SKH’s technology leadership in HBM and believe concerns about 2026 OP declines are excessive.\n\n(52) The market is pricing in an annual OP growth scenario even in 2026, with gradual share price recovery expected into year-end.\n\n(53) Over the past month, SEC has outperformed SKH, driven by 1) expectations of new foundry orders, and 2) better-than-expected 2H25 HBM outlook.\n\n(54) HBM4 could provide SEC with a rebound catalyst, and we estimate SEC has a 30–40% chance of securing HBM4 orders in 2026.\n\n(55) We also believe the recovery effect of the legacy memory upcycle has not been fully priced into SEC, and we maintain an OW rating on SEC as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02935423423573913, "ret_signed_m1d": 0.02935423423573913, "alpha_spy_m1d": 0.02669007170243587, "alpha_c_m1d": 0.022744295119684566, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006609939116054564, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04109591564800541, "ret_signed_p1d": -0.04109591564800541, "alpha_spy_p1d": -0.037084093360446824, "alpha_c_p1d": -0.035904836253094596, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.005191079394910814, "ret_p1w": 0.017612491413326214, "ret_signed_p1w": 0.017612491413326214, "alpha_spy_p1w": 0.004260521771602521, "alpha_c_p1w": -0.012495451255969181, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.030107942669295396, "ret_p1m": 0.3933991707419464, "ret_signed_p1m": 0.3933991707419464, "alpha_spy_p1m": 0.35555299552592157, "alpha_c_p1m": 0.2964652976776099, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.09693387306433654, "ret_p3m": 1.3752457173485015, "ret_signed_p3m": 1.3752457173485015, "alpha_spy_p3m": 1.3291634899430962, "alpha_c_p3m": 1.1975741389319503, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.17767157841655123, "ret_p6m": 2.4830508609886763, "ret_signed_p6m": 2.4830508609886763, "alpha_spy_p6m": 2.409290050097047, "alpha_c_p6m": 2.073258366441958, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.4097924945467182, "price_path": [-0.2186, -0.2069, -0.2089, -0.1874, -0.1703, -0.1996, -0.1879, -0.1879, -0.1487, -0.1213, -0.1213, -0.1037, -0.0978, -0.0607, -0.0783, -0.0783, -0.0294, -0.0254, -0.0352, -0.0372, 0.0059, 0.0157, 0.09, 0.1194, 0.1468, 0.1115, 0.1429, 0.1174, 0.092, 0.09, 0.0587, 0.0607, 0.1037, 0.0998, 0.1624, 0.1526, 0.1742, 0.1683, 0.1585, 0.0548, 0.0528, 0.0665, 0.0509, 0.0528, 0.0548, 0.0411, 0.0254, 0.0274, 0.0313, 0.0704, 0.0098, 0.0098, 0.0313, 0.0117, 0.0254, 0.0039, 0.045, 0.0528, 0.0881, 0.0822, 0.0822, 0.047, 0.0294, 0.0, -0.0411, -0.0176, 0.0157, 0.0235, 0.0176, 0.0524, 0.0544, 0.0034, 0.021, 0.0289, 0.0406, 0.072, 0.0857, 0.1288, 0.1915, 0.2033, 0.2876, 0.2974, 0.364, 0.3072, 0.3934, 0.3934, 0.3758, 0.415, 0.4012, 0.3973, 0.3189, 0.3679, 0.362, 0.411, 0.5502, 0.5502, 0.5502, 0.5502, 0.5502, 0.5502, 0.6776, 0.6266, 0.6129, 0.656, 0.7736, 0.8245, 0.9029, 0.8775, 0.8873, 0.8755, 0.999, 1.097, 1.0421, 1.1871, 1.2263, 1.191, 1.4301, 1.2969, 1.2694, 1.3243, 1.2733, 1.3752, 1.4262, 1.4184, 1.3988, 1.1949, 1.3752, 1.2341, 1.2028, 1.2381, 1.0421, 1.0382, 1.0342, 1.0538, 1.1338, 1.0788, 1.1102, 1.1887, 1.1651, 1.1259, 1.1338, 1.2632, 1.2201, 1.3024, 1.2161, 1.2397, 1.173, 1.0788, 1.1612, 1.1651, 1.1455, 1.275, 1.2907, 1.3063, 1.3063, 1.3495, 1.5103, 1.5535, 1.5535, 1.5535, 1.6554, 1.73, 1.8476, 1.9104, 1.9653, 1.9182, 1.9378, 1.8947, 1.9104, 1.9378, 1.9653, 1.9967, 1.9143, 1.9025, 1.9614, 2.0084, 1.8869, 2.1379, 2.2987, 2.3771, 2.5654, 2.2556, 2.5576, 2.5301, 2.3026, 2.2909, 2.4791, 2.436, 2.3732, 2.4831]}
{"unit_id": "orig:1958315188643598527", "ticker_idx": 1, "ts": "2025-08-20T23:48:19+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "we maintain an OW rating on SEC as well. We believe the recovery effect of the legacy memory upcycle has not been fully priced into SEC", "tweet_type": "analysis", "timeline": "months", "summary": "JPMorgan maintains positive view on memory: SKH top pick on HBM tech lead, SEC OW on legacy upcycle underpricing + HBM4 upside, MU/DRAM ASP strength into 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250821_Memory Market Update: HBM FAQs – JPMorgan\n\n(1) In this report, we aim to address recent investor FAQs related to the HBM industry.\n\n(2) The most frequently asked questions are:\n\n2026 pricing outlook,\nchanges in logic die specifications and design ownership,\nthe impact of CSP CAPEX expansion on HBM supply/demand.\n(3) In conclusion, we maintain our preference ranking of Asian memory makers as SKH > SEC > Nanya.\n\n[HBM Base Die]\n\n(4) Who is responsible for the HBM4E base die design? Why does NVIDIA want to directly control base die design?\n\n(5) Up to HBM3E, since the design involved I/O circuit structure, memory makers traditionally handled the die design.\n\n(6) However, starting with HBM4, the role and significance of this design has become much more important in controlling HBM performance.\n\n(7) Recently, as reported by TrendForce [NVIDIA direct entry], discussions regarding customers [NVIDIA = memory makers’ customer] entering base die design are being actively debated among investors.\n\n(8) In other words, NVIDIA may directly design the logic die for HBM4E at the 3nm process starting from late 2027.\n\n(9) We believe NVIDIA’s insourcing signals 1) cost control [one of the main drivers of high BOM cost in HBM4], and 2) supply chain diversification.\n\n[Are Memory Makers Being Excluded?]\n\n(10) We believe it is an over-interpretation to say that memory makers are being excluded from HBM.\n\n(11) In HBM4E, customers will have two choices:\n\n(12) First, custom HBM dies embedding the GPU/ASIC customer’s core IP to pursue “performance differentiation.”\n\n(13) Second, generic EKDLs where memory makers continue to handle design as they do today.\n\n(14) Embedding IP into custom logic dies is important, but at the same time, integrating and stacking them into HBM is also critical.\n\n(15) With HBM4E expanding to 16-high stacks, the stable yield and mass production capability of memory makers will become even more essential to supply chain stability.\n\n(16) Therefore, it is too early to conclude that the value of memory makers will disappear or that HBM will become commoditized.\n\n[Rising Importance of Thermal Management in HBM4]\n\n(17) As HBM specification requirements increase, thermal characteristics are becoming even more important.\n\n(18) Recently, customers have requested that memory makers raise HBM4 reading speed from ~8Gbps to ~10Gbps.\n\n(19) This is because, with double the I/Os versus HBM3E, both performance and thermal control become far more critical.\n\n(20) SKH’s advanced MR-MUF and Micron’s strong base die performance are currently viewed as superior to SEC up to the 5th generation HBM in terms of thermal management.\n\n(21) NVIDIA’s chips consume ~700W per chip, far higher than other GPU/ASICs [300–400W].\n\n(22) Therefore, the leading suppliers who can meet NVIDIA’s upgrade requests and thermal requirements will likely remain in a favorable position.\n\n[HBM4 Qualification Tests: SKH Advantage Continues]\n\n(23) SKH is about to begin customer sampling, with results expected around November this year.\n\n(24) Meanwhile, Micron and SEC will start customer qualification in November, with results expected by February 2026 at the earliest.\n\n(25) If the JEDEC standard [HBM4: ~8Gbps] does not change, this spec change will be optional rather than mandatory.\n\n(26) However, if stricter spec requirements are added, the qualification process could lengthen, and the 2Q25 GPU launch schedule [with HBM4 onboard] could become challenging.\n\n(27) If the launch of the new Rubin GPU slips to late 2026, SKH’s early qualification advantage may be diluted in its competition with MU/SEC.\n\n[2026 HBM Pricing Outlook]\n\n(28) We maintain our HBM4 ASP outlook at $19/GB, $2.4/Gb.\n\n(29) The rationale is: 1) wafer die penalty from large die sizes, 2) yield loss from increased I/Os, 3) additional new logic die costs compared to prior generations.\n\n(30) SKH management has also stated that “profitability maintenance” is the top priority for next-gen HBM, making a +30–40% premium vs. HBM3E reasonable.\n\n(31) By contrast, the HBM3E market shows signs of easing.\n\n(32) SEC is expected to start shipping HBM3E 12-high to Broadcom/AMD in 3Q25.\n\n(33) We expect HBM3E bit mix to peak by 1H26, then gradually decline from 3Q26.\n\n(34) We estimate 2026 HBM3E 12-high ASP to fall -11% YoY.\n\nSKH: 4Q25 -7% → 1Q26 -7% → 2Q26 -7% → 4Q26 -10%\n\n(35) For 2026, bull/bear case HBM operating profit is $18.2bn / $12.5bn respectively.\n\n[CSP CAPEX Update]\n\n(36) CSP CAPEX expansion is positively impacting HBM bit demand, with ASIC demand growth and B30A as key variables.\n\n(37) Major firms reported 2Q25 cloud and AI revenue above expectations and raised CAPEX.\n\n(38) While noise around AI ROIC persists, CAPEX trends remain upward.\n\n(39) Our channel checks indicate HBM suppliers also see stronger procurement needs across customers versus 3 months ago, and that 2026 volumes will largely be decided in current CAPEX negotiations.\n\n(40) In 2026–27, if ASIC demand rises +10%, HBM bit demand would increase by +3% and +2% respectively.\n\n(41) Memory makers have recently detected production codes for B30A, which is reportedly targeting 2026 mass production.\n\n(42) These products could use HBM3E 96GB or 144GB.\n\n(43) If 1,000K chips are produced in 2026, this could add +3–4% to HBM bit demand.\n\n[DRAM Outlook]\n\n(44) The DRAM market is expected to remain tight into early 2026, driven by strong GPU server demand.\n\n(45) Beyond HBM, ASP uptrends are also appearing in legacy DRAM/NAND.\n\n(46) Micron recently raised guidance, which we attribute to D4/D5 price strength.\n\n(47) ASP strength is expected to continue through 4Q25, with server customers likely to further raise GP server purchases over the next 6 months.\n\n(48) Rather than the mild year-end decline initially expected, demand recovery looks stronger, and visibility into early 2026 is improving.\n\n[Positive View on Memory Sector Maintained]\n\n(49) We maintain a medium-term positive view on memory stocks.\n\n(50) The main market debate centers on 2026 HBM ASP and supply-demand, with 2026 volume and price negotiations to be announced within 1–2 months.\n\n(51) We continue to support SKH’s technology leadership in HBM and believe concerns about 2026 OP declines are excessive.\n\n(52) The market is pricing in an annual OP growth scenario even in 2026, with gradual share price recovery expected into year-end.\n\n(53) Over the past month, SEC has outperformed SKH, driven by 1) expectations of new foundry orders, and 2) better-than-expected 2H25 HBM outlook.\n\n(54) HBM4 could provide SEC with a rebound catalyst, and we estimate SEC has a 30–40% chance of securing HBM4 orders in 2026.\n\n(55) We also believe the recovery effect of the legacy memory upcycle has not been fully priced into SEC, and we maintain an OW rating on SEC as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007092252635595098, "ret_signed_m1d": -0.007092252635595098, "alpha_spy_m1d": -0.009756415168898358, "alpha_c_m1d": -0.01390394999180844, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006811697356213342, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0014184056940667045, "ret_signed_p1d": 0.0014184056940667045, "alpha_spy_p1d": 0.005430227981625291, "alpha_c_p1d": 0.005035682779189421, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.0036172770851227165, "ret_p1w": 0.0014184056940667045, "ret_signed_p1w": 0.0014184056940667045, "alpha_spy_p1w": -0.011933563947656989, "alpha_c_p1w": -0.01628415404005934, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.017702559734126044, "ret_p1m": 0.13900712049745856, "ret_signed_p1m": 0.13900712049745856, "alpha_spy_p1m": 0.10116094528143371, "alpha_c_p1m": 0.07639400688929898, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.06261311360815958, "ret_p3m": 0.43331694311660773, "ret_signed_p3m": 0.43331694311660773, "alpha_spy_p3m": 0.3872347157112024, "alpha_c_p3m": 0.3403611999286589, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.09295574318794886, "ret_p6m": 1.5570060666965628, "ret_signed_p6m": 1.5570060666965628, "alpha_spy_p6m": 1.4832452558049332, "alpha_c_p6m": 1.4825372790952058, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.074468787601357, "price_path": [-0.2359, -0.2288, -0.2401, -0.2119, -0.2091, -0.2077, -0.1992, -0.1992, -0.1851, -0.1668, -0.1668, -0.1569, -0.1654, -0.1555, -0.1612, -0.1781, -0.1936, -0.1809, -0.1569, -0.1654, -0.1612, -0.1823, -0.1471, -0.1358, -0.1513, -0.1376, -0.1518, -0.1461, -0.1376, -0.095, -0.1021, -0.1248, -0.1291, -0.1433, -0.1348, -0.1121, -0.1135, -0.0965, -0.0823, -0.0539, -0.0482, -0.0383, -0.0638, -0.0582, -0.0638, -0.0652, -0.0014, 0.0014, 0.0298, 0.0128, -0.0227, -0.0113, -0.0085, -0.0241, 0.0, 0.0184, 0.0071, 0.0085, 0.0199, 0.0156, 0.0156, -0.0071, -0.0071, 0.0, 0.0014, 0.0128, 0.0142, -0.0028, 0.0014, -0.0128, -0.0113, -0.0411, -0.0199, -0.0099, -0.0057, -0.0142, -0.0057, 0.0142, 0.0298, 0.0411, 0.0695, 0.0851, 0.1262, 0.1092, 0.139, 0.139, 0.1844, 0.2014, 0.2113, 0.2213, 0.1816, 0.1997, 0.1954, 0.2253, 0.2787, 0.2787, 0.2787, 0.2787, 0.2787, 0.2787, 0.345, 0.3293, 0.3051, 0.3535, 0.392, 0.3948, 0.3977, 0.3891, 0.4048, 0.3749, 0.4077, 0.4533, 0.4176, 0.4319, 0.4832, 0.5316, 0.5829, 0.4946, 0.4333, 0.4134, 0.3948, 0.4333, 0.4746, 0.4689, 0.4647, 0.3849, 0.4333, 0.3934, 0.3749, 0.4333, 0.3507, 0.3778, 0.4148, 0.4647, 0.4746, 0.4319, 0.4362, 0.4732, 0.4889, 0.4974, 0.5444, 0.5601, 0.5444, 0.5387, 0.5288, 0.5516, 0.4932, 0.4647, 0.5373, 0.5331, 0.5145, 0.5744, 0.5886, 0.5829, 0.5829, 0.667, 0.7109, 0.7166, 0.7166, 0.7166, 0.8397, 0.9772, 0.9886, 1.0187, 0.9872, 0.9901, 0.9872, 0.97, 1.0087, 1.0602, 1.1318, 1.1375, 1.0788, 1.1404, 1.1805, 1.1776, 1.1776, 1.2836, 1.3251, 1.3007, 1.2979, 1.1533, 1.3981, 1.421, 1.2807, 1.2707, 1.3823, 1.3737, 1.4024, 1.557]}
{"unit_id": "orig:1958315188643598527", "ticker_idx": 2, "ts": "2025-08-20T23:48:19+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The DRAM market is expected to remain tight into early 2026, driven by strong GPU server demand. ASP strength is expected to continue through 4Q25", "tweet_type": "analysis", "timeline": "months", "summary": "JPMorgan maintains positive view on memory: SKH top pick on HBM tech lead, SEC OW on legacy upcycle underpricing + HBM4 upside, MU/DRAM ASP strength into 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250821_Memory Market Update: HBM FAQs – JPMorgan\n\n(1) In this report, we aim to address recent investor FAQs related to the HBM industry.\n\n(2) The most frequently asked questions are:\n\n2026 pricing outlook,\nchanges in logic die specifications and design ownership,\nthe impact of CSP CAPEX expansion on HBM supply/demand.\n(3) In conclusion, we maintain our preference ranking of Asian memory makers as SKH > SEC > Nanya.\n\n[HBM Base Die]\n\n(4) Who is responsible for the HBM4E base die design? Why does NVIDIA want to directly control base die design?\n\n(5) Up to HBM3E, since the design involved I/O circuit structure, memory makers traditionally handled the die design.\n\n(6) However, starting with HBM4, the role and significance of this design has become much more important in controlling HBM performance.\n\n(7) Recently, as reported by TrendForce [NVIDIA direct entry], discussions regarding customers [NVIDIA = memory makers’ customer] entering base die design are being actively debated among investors.\n\n(8) In other words, NVIDIA may directly design the logic die for HBM4E at the 3nm process starting from late 2027.\n\n(9) We believe NVIDIA’s insourcing signals 1) cost control [one of the main drivers of high BOM cost in HBM4], and 2) supply chain diversification.\n\n[Are Memory Makers Being Excluded?]\n\n(10) We believe it is an over-interpretation to say that memory makers are being excluded from HBM.\n\n(11) In HBM4E, customers will have two choices:\n\n(12) First, custom HBM dies embedding the GPU/ASIC customer’s core IP to pursue “performance differentiation.”\n\n(13) Second, generic EKDLs where memory makers continue to handle design as they do today.\n\n(14) Embedding IP into custom logic dies is important, but at the same time, integrating and stacking them into HBM is also critical.\n\n(15) With HBM4E expanding to 16-high stacks, the stable yield and mass production capability of memory makers will become even more essential to supply chain stability.\n\n(16) Therefore, it is too early to conclude that the value of memory makers will disappear or that HBM will become commoditized.\n\n[Rising Importance of Thermal Management in HBM4]\n\n(17) As HBM specification requirements increase, thermal characteristics are becoming even more important.\n\n(18) Recently, customers have requested that memory makers raise HBM4 reading speed from ~8Gbps to ~10Gbps.\n\n(19) This is because, with double the I/Os versus HBM3E, both performance and thermal control become far more critical.\n\n(20) SKH’s advanced MR-MUF and Micron’s strong base die performance are currently viewed as superior to SEC up to the 5th generation HBM in terms of thermal management.\n\n(21) NVIDIA’s chips consume ~700W per chip, far higher than other GPU/ASICs [300–400W].\n\n(22) Therefore, the leading suppliers who can meet NVIDIA’s upgrade requests and thermal requirements will likely remain in a favorable position.\n\n[HBM4 Qualification Tests: SKH Advantage Continues]\n\n(23) SKH is about to begin customer sampling, with results expected around November this year.\n\n(24) Meanwhile, Micron and SEC will start customer qualification in November, with results expected by February 2026 at the earliest.\n\n(25) If the JEDEC standard [HBM4: ~8Gbps] does not change, this spec change will be optional rather than mandatory.\n\n(26) However, if stricter spec requirements are added, the qualification process could lengthen, and the 2Q25 GPU launch schedule [with HBM4 onboard] could become challenging.\n\n(27) If the launch of the new Rubin GPU slips to late 2026, SKH’s early qualification advantage may be diluted in its competition with MU/SEC.\n\n[2026 HBM Pricing Outlook]\n\n(28) We maintain our HBM4 ASP outlook at $19/GB, $2.4/Gb.\n\n(29) The rationale is: 1) wafer die penalty from large die sizes, 2) yield loss from increased I/Os, 3) additional new logic die costs compared to prior generations.\n\n(30) SKH management has also stated that “profitability maintenance” is the top priority for next-gen HBM, making a +30–40% premium vs. HBM3E reasonable.\n\n(31) By contrast, the HBM3E market shows signs of easing.\n\n(32) SEC is expected to start shipping HBM3E 12-high to Broadcom/AMD in 3Q25.\n\n(33) We expect HBM3E bit mix to peak by 1H26, then gradually decline from 3Q26.\n\n(34) We estimate 2026 HBM3E 12-high ASP to fall -11% YoY.\n\nSKH: 4Q25 -7% → 1Q26 -7% → 2Q26 -7% → 4Q26 -10%\n\n(35) For 2026, bull/bear case HBM operating profit is $18.2bn / $12.5bn respectively.\n\n[CSP CAPEX Update]\n\n(36) CSP CAPEX expansion is positively impacting HBM bit demand, with ASIC demand growth and B30A as key variables.\n\n(37) Major firms reported 2Q25 cloud and AI revenue above expectations and raised CAPEX.\n\n(38) While noise around AI ROIC persists, CAPEX trends remain upward.\n\n(39) Our channel checks indicate HBM suppliers also see stronger procurement needs across customers versus 3 months ago, and that 2026 volumes will largely be decided in current CAPEX negotiations.\n\n(40) In 2026–27, if ASIC demand rises +10%, HBM bit demand would increase by +3% and +2% respectively.\n\n(41) Memory makers have recently detected production codes for B30A, which is reportedly targeting 2026 mass production.\n\n(42) These products could use HBM3E 96GB or 144GB.\n\n(43) If 1,000K chips are produced in 2026, this could add +3–4% to HBM bit demand.\n\n[DRAM Outlook]\n\n(44) The DRAM market is expected to remain tight into early 2026, driven by strong GPU server demand.\n\n(45) Beyond HBM, ASP uptrends are also appearing in legacy DRAM/NAND.\n\n(46) Micron recently raised guidance, which we attribute to D4/D5 price strength.\n\n(47) ASP strength is expected to continue through 4Q25, with server customers likely to further raise GP server purchases over the next 6 months.\n\n(48) Rather than the mild year-end decline initially expected, demand recovery looks stronger, and visibility into early 2026 is improving.\n\n[Positive View on Memory Sector Maintained]\n\n(49) We maintain a medium-term positive view on memory stocks.\n\n(50) The main market debate centers on 2026 HBM ASP and supply-demand, with 2026 volume and price negotiations to be announced within 1–2 months.\n\n(51) We continue to support SKH’s technology leadership in HBM and believe concerns about 2026 OP declines are excessive.\n\n(52) The market is pricing in an annual OP growth scenario even in 2026, with gradual share price recovery expected into year-end.\n\n(53) Over the past month, SEC has outperformed SKH, driven by 1) expectations of new foundry orders, and 2) better-than-expected 2H25 HBM outlook.\n\n(54) HBM4 could provide SEC with a rebound catalyst, and we estimate SEC has a 30–40% chance of securing HBM4 orders in 2026.\n\n(55) We also believe the recovery effect of the legacy memory upcycle has not been fully priced into SEC, and we maintain an OW rating on SEC as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.041293451682645976, "ret_signed_m1d": 0.041293451682645976, "alpha_spy_m1d": 0.038629289149342716, "alpha_c_m1d": 0.03468351256659141, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006609939116054564, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012114970846485495, "ret_signed_p1d": -0.012114970846485495, "alpha_spy_p1d": -0.00810314855892691, "alpha_c_p1d": -0.006923891451574682, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.005191079394910814, "ret_p1w": 0.004607224616834271, "ret_signed_p1w": 0.004607224616834271, "alpha_spy_p1w": -0.008744745024889422, "alpha_c_p1w": -0.025500718052461124, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.030107942669295396, "ret_p1m": 0.4409180708897755, "ret_signed_p1m": 0.4409180708897755, "alpha_spy_p1m": 0.4030718956737507, "alpha_c_p1m": 0.343984197825439, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.09693387306433654, "ret_p3m": 1.0655364305996948, "ret_signed_p3m": 1.0655364305996948, "alpha_spy_p3m": 1.0194542031942895, "alpha_c_p3m": 0.8878648521831436, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.17767157841655123, "ret_p6m": 2.535505608646391, "ret_signed_p6m": 2.535505608646391, "alpha_spy_p6m": 2.4617447977547613, "alpha_c_p6m": 2.125713114099673, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.4097924945467182, "price_path": [-0.2041, -0.2041, -0.1785, -0.1802, -0.1749, -0.1949, -0.1631, -0.1285, -0.1199, -0.094, -0.0747, -0.0543, -0.0271, -0.011, -0.0097, -0.0147, 0.0215, 0.0257, 0.0384, 0.0384, 0.0535, 0.0406, 0.0903, 0.0846, 0.074, 0.0634, 0.0505, 0.0304, 0.0377, 0.0424, 0.0424, 0.0231, 0.0615, 0.0429, 0.0503, 0.0625, 0.0119, 0.0247, -0.0067, -0.0337, -0.0241, -0.034, -0.0682, -0.063, -0.0468, -0.0508, -0.0508, -0.0448, -0.0211, -0.0689, -0.1052, -0.0805, -0.0695, -0.0719, -0.0456, 0.0143, 0.0555, 0.0899, 0.0602, 0.0689, 0.0312, 0.0541, 0.0413, 0.0, -0.0121, 0.004, -0.0067, -0.0061, 0.0046, 0.0409, 0.0154, 0.0154, 0.0108, 0.0129, 0.0597, 0.1208, 0.1216, 0.1538, 0.1944, 0.2846, 0.3414, 0.346, 0.355, 0.365, 0.4409, 0.3884, 0.4045, 0.4198, 0.3797, 0.338, 0.3418, 0.3983, 0.4275, 0.554, 0.5677, 0.6035, 0.6302, 0.5852, 0.6779, 0.6419, 0.5503, 0.6457, 0.5969, 0.6386, 0.729, 0.7277, 0.7652, 0.727, 0.6943, 0.7647, 0.8698, 0.879, 0.8945, 0.9347, 0.9124, 0.9103, 1.0036, 0.8613, 1.0275, 1.0346, 1.0311, 1.1624, 1.0584, 1.0907, 1.0229, 1.1072, 1.0655, 0.9507, 0.9287, 0.7191, 0.7703, 0.9117, 0.9168, 0.9657, 0.9657, 1.0188, 1.0528, 1.0445, 0.999, 0.9349, 1.0252, 1.108, 1.1549, 1.2513, 1.2065, 1.0586, 1.0275, 0.9849, 0.9253, 1.1219, 1.2702, 1.3613, 1.3585, 1.4474, 1.4474, 1.4313, 1.5141, 1.4992, 1.4375, 1.4375, 1.6938, 1.6659, 1.9331, 1.8999, 1.7929, 1.9472, 1.9539, 1.8878, 1.847, 1.875, 2.0981, 2.0981, 2.1173, 2.3232, 2.3955, 2.4132, 2.323, 2.5036, 2.7175, 2.7219, 2.5433, 2.739, 2.5822, 2.2403, 2.2701, 2.3708, 2.2753, 2.1877, 2.5045, 2.5355]}
{"unit_id": "orig:1958315188643598527", "ticker_idx": 3, "ts": "2025-08-20T23:48:19+00:00", "symbol": "Nanya Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our preference ranking of Asian memory makers as SKH > SEC > Nanya", "tweet_type": "analysis", "timeline": "months", "summary": "JPMorgan maintains positive view on memory: SKH top pick on HBM tech lead, SEC OW on legacy upcycle underpricing + HBM4 upside, MU/DRAM ASP strength into 2026.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya Technology Corp listed on TWSE as 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250821_Memory Market Update: HBM FAQs – JPMorgan\n\n(1) In this report, we aim to address recent investor FAQs related to the HBM industry.\n\n(2) The most frequently asked questions are:\n\n2026 pricing outlook,\nchanges in logic die specifications and design ownership,\nthe impact of CSP CAPEX expansion on HBM supply/demand.\n(3) In conclusion, we maintain our preference ranking of Asian memory makers as SKH > SEC > Nanya.\n\n[HBM Base Die]\n\n(4) Who is responsible for the HBM4E base die design? Why does NVIDIA want to directly control base die design?\n\n(5) Up to HBM3E, since the design involved I/O circuit structure, memory makers traditionally handled the die design.\n\n(6) However, starting with HBM4, the role and significance of this design has become much more important in controlling HBM performance.\n\n(7) Recently, as reported by TrendForce [NVIDIA direct entry], discussions regarding customers [NVIDIA = memory makers’ customer] entering base die design are being actively debated among investors.\n\n(8) In other words, NVIDIA may directly design the logic die for HBM4E at the 3nm process starting from late 2027.\n\n(9) We believe NVIDIA’s insourcing signals 1) cost control [one of the main drivers of high BOM cost in HBM4], and 2) supply chain diversification.\n\n[Are Memory Makers Being Excluded?]\n\n(10) We believe it is an over-interpretation to say that memory makers are being excluded from HBM.\n\n(11) In HBM4E, customers will have two choices:\n\n(12) First, custom HBM dies embedding the GPU/ASIC customer’s core IP to pursue “performance differentiation.”\n\n(13) Second, generic EKDLs where memory makers continue to handle design as they do today.\n\n(14) Embedding IP into custom logic dies is important, but at the same time, integrating and stacking them into HBM is also critical.\n\n(15) With HBM4E expanding to 16-high stacks, the stable yield and mass production capability of memory makers will become even more essential to supply chain stability.\n\n(16) Therefore, it is too early to conclude that the value of memory makers will disappear or that HBM will become commoditized.\n\n[Rising Importance of Thermal Management in HBM4]\n\n(17) As HBM specification requirements increase, thermal characteristics are becoming even more important.\n\n(18) Recently, customers have requested that memory makers raise HBM4 reading speed from ~8Gbps to ~10Gbps.\n\n(19) This is because, with double the I/Os versus HBM3E, both performance and thermal control become far more critical.\n\n(20) SKH’s advanced MR-MUF and Micron’s strong base die performance are currently viewed as superior to SEC up to the 5th generation HBM in terms of thermal management.\n\n(21) NVIDIA’s chips consume ~700W per chip, far higher than other GPU/ASICs [300–400W].\n\n(22) Therefore, the leading suppliers who can meet NVIDIA’s upgrade requests and thermal requirements will likely remain in a favorable position.\n\n[HBM4 Qualification Tests: SKH Advantage Continues]\n\n(23) SKH is about to begin customer sampling, with results expected around November this year.\n\n(24) Meanwhile, Micron and SEC will start customer qualification in November, with results expected by February 2026 at the earliest.\n\n(25) If the JEDEC standard [HBM4: ~8Gbps] does not change, this spec change will be optional rather than mandatory.\n\n(26) However, if stricter spec requirements are added, the qualification process could lengthen, and the 2Q25 GPU launch schedule [with HBM4 onboard] could become challenging.\n\n(27) If the launch of the new Rubin GPU slips to late 2026, SKH’s early qualification advantage may be diluted in its competition with MU/SEC.\n\n[2026 HBM Pricing Outlook]\n\n(28) We maintain our HBM4 ASP outlook at $19/GB, $2.4/Gb.\n\n(29) The rationale is: 1) wafer die penalty from large die sizes, 2) yield loss from increased I/Os, 3) additional new logic die costs compared to prior generations.\n\n(30) SKH management has also stated that “profitability maintenance” is the top priority for next-gen HBM, making a +30–40% premium vs. HBM3E reasonable.\n\n(31) By contrast, the HBM3E market shows signs of easing.\n\n(32) SEC is expected to start shipping HBM3E 12-high to Broadcom/AMD in 3Q25.\n\n(33) We expect HBM3E bit mix to peak by 1H26, then gradually decline from 3Q26.\n\n(34) We estimate 2026 HBM3E 12-high ASP to fall -11% YoY.\n\nSKH: 4Q25 -7% → 1Q26 -7% → 2Q26 -7% → 4Q26 -10%\n\n(35) For 2026, bull/bear case HBM operating profit is $18.2bn / $12.5bn respectively.\n\n[CSP CAPEX Update]\n\n(36) CSP CAPEX expansion is positively impacting HBM bit demand, with ASIC demand growth and B30A as key variables.\n\n(37) Major firms reported 2Q25 cloud and AI revenue above expectations and raised CAPEX.\n\n(38) While noise around AI ROIC persists, CAPEX trends remain upward.\n\n(39) Our channel checks indicate HBM suppliers also see stronger procurement needs across customers versus 3 months ago, and that 2026 volumes will largely be decided in current CAPEX negotiations.\n\n(40) In 2026–27, if ASIC demand rises +10%, HBM bit demand would increase by +3% and +2% respectively.\n\n(41) Memory makers have recently detected production codes for B30A, which is reportedly targeting 2026 mass production.\n\n(42) These products could use HBM3E 96GB or 144GB.\n\n(43) If 1,000K chips are produced in 2026, this could add +3–4% to HBM bit demand.\n\n[DRAM Outlook]\n\n(44) The DRAM market is expected to remain tight into early 2026, driven by strong GPU server demand.\n\n(45) Beyond HBM, ASP uptrends are also appearing in legacy DRAM/NAND.\n\n(46) Micron recently raised guidance, which we attribute to D4/D5 price strength.\n\n(47) ASP strength is expected to continue through 4Q25, with server customers likely to further raise GP server purchases over the next 6 months.\n\n(48) Rather than the mild year-end decline initially expected, demand recovery looks stronger, and visibility into early 2026 is improving.\n\n[Positive View on Memory Sector Maintained]\n\n(49) We maintain a medium-term positive view on memory stocks.\n\n(50) The main market debate centers on 2026 HBM ASP and supply-demand, with 2026 volume and price negotiations to be announced within 1–2 months.\n\n(51) We continue to support SKH’s technology leadership in HBM and believe concerns about 2026 OP declines are excessive.\n\n(52) The market is pricing in an annual OP growth scenario even in 2026, with gradual share price recovery expected into year-end.\n\n(53) Over the past month, SEC has outperformed SKH, driven by 1) expectations of new foundry orders, and 2) better-than-expected 2H25 HBM outlook.\n\n(54) HBM4 could provide SEC with a rebound catalyst, and we estimate SEC has a 30–40% chance of securing HBM4 orders in 2026.\n\n(55) We also believe the recovery effect of the legacy memory upcycle has not been fully priced into SEC, and we maintain an OW rating on SEC as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026936044522783975, "ret_signed_m1d": 0.026936044522783975, "alpha_spy_m1d": 0.024271881989480715, "alpha_c_m1d": 0.02032610540672941, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006609939116054564, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04264874429668497, "ret_signed_p1d": 0.04264874429668497, "alpha_spy_p1d": 0.04666056658424356, "alpha_c_p1d": 0.047839823691595784, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.005191079394910814, "ret_p1w": 0.0729517408677387, "ret_signed_p1w": 0.0729517408677387, "alpha_spy_p1w": 0.05959977122601501, "alpha_c_p1w": 0.042843798198443306, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.030107942669295396, "ret_p1m": 0.7957351598212614, "ret_signed_p1m": 0.7957351598212614, "alpha_spy_p1m": 0.7578889846052366, "alpha_c_p1m": 0.6988012867569249, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.09693387306433654, "ret_p3m": 2.7373738013780007, "ret_signed_p3m": 2.7373738013780007, "alpha_spy_p3m": 2.6912915739725953, "alpha_c_p3m": 2.5597022229614494, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.17767157841655123, "ret_p6m": 5.240179680378883, "ret_signed_p6m": 5.240179680378883, "alpha_spy_p6m": 5.166418869487254, "alpha_c_p6m": 4.830387185832166, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.4097924945467182, "price_path": [-0.0449, -0.0393, -0.0584, -0.064, 0.0224, 0.0224, 0.0281, 0.0292, 0.1055, 0.1785, 0.1672, 0.1605, 0.2189, 0.1829, 0.2144, 0.2009, 0.1874, 0.2076, 0.3266, 0.3558, 0.3221, 0.3109, 0.2727, 0.2346, 0.2099, 0.1919, 0.1493, 0.1111, 0.11, 0.1358, 0.0875, 0.0976, 0.0887, 0.0864, 0.0584, -0.0359, -0.0258, -0.0629, -0.0707, -0.0325, -0.0516, -0.0584, -0.0898, -0.0337, -0.0348, -0.0292, -0.0281, -0.0527, 0.0079, 0.0, 0.0, -0.018, -0.0034, -0.0269, -0.0527, -0.0135, -0.0168, 0.0651, 0.0337, 0.0404, 0.0774, 0.0629, 0.0269, 0.0, 0.0426, 0.018, 0.0539, 0.0651, 0.073, 0.0471, 0.055, 0.0494, 0.0247, 0.0404, 0.0819, 0.1897, 0.2009, 0.2099, 0.2211, 0.2682, 0.2884, 0.4164, 0.5533, 0.6453, 0.7957, 0.7778, 0.8137, 0.8182, 0.7059, 0.7194, 0.5937, 0.5937, 0.6386, 0.7104, 0.789, 0.8631, 0.8631, 1.0471, 1.0382, 1.211, 1.211, 1.1459, 1.0, 0.9371, 1.1302, 1.3345, 1.4018, 1.3681, 1.4691, 1.4579, 1.4579, 1.6936, 1.9181, 1.9854, 2.0191, 1.9742, 2.0752, 1.9517, 2.0976, 2.3221, 2.2997, 2.6251, 2.6925, 2.6588, 2.67, 2.5578, 2.7374, 2.6027, 2.5915, 2.4905, 2.1425, 2.2099, 2.2211, 2.064, 2.2772, 2.2772, 2.3558, 2.3558, 2.3895, 2.3895, 2.4343, 2.67, 2.6364, 2.5017, 2.5017, 2.6364, 2.6476, 2.5466, 2.6925, 2.8159, 2.9169, 3.0404, 2.9618, 3.2424, 3.2424, 3.2424, 3.2312, 3.3659, 3.3322, 3.3322, 3.6465, 3.6689, 4.1291, 4.4097, 4.4209, 3.8822, 4.3648, 4.1627, 4.3086, 4.3311, 4.6117, 5.1728, 5.1055, 4.6341, 4.982, 5.0943, 5.4198, 5.2177, 5.6667, 5.6667, 6.3288, 5.5993, 5.229, 5.6218, 5.1728, 4.9259, 5.1616, 5.0831, 5.2402, 5.2402]}
{"unit_id": "orig:1965028624073199634", "ticker_idx": 0, "ts": "2025-09-08T12:25:07+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CITI lowered its NVIDIA price target to $200 due to intensifying TPU competition. forecasts NVIDIA's GPU revenue in 2026 will decline by about $12 billion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's PT cut on NVDA to $200 on XPU competition; implied AVGO beneficiary of GPU share shift.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CITI: NVIDIA Price Target Cut, Broadcom XPU and GPU Pose a Real Threat\n\n• CITI lowered its NVIDIA price target to $200 due to intensifying TPU competition. As a result, it forecasts NVIDIA’s GPU revenue in 2026 will decline by about $12 billion. Specifically, the 2026 XPU market is expected to grow +53% YoY, significantly outpacing the AI GPU revenue growth rate of +34% YoY. This growth will be mainly driven by production capacity expansions at Google, Meta, and Amazon.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007664302707223625, "ret_signed_m1d": 0.007664302707223625, "alpha_spy_m1d": 0.005213780608743446, "alpha_c_m1d": -0.0032623460281404615, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014556426417988044, "ret_signed_p1d": -0.014556426417988044, "alpha_spy_p1d": -0.012244505720378163, "alpha_c_p1d": -0.01020591924940728, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.0561465529724674, "ret_signed_p1w": -0.0561465529724674, "alpha_spy_p1w": -0.037528428861316465, "alpha_c_p1w": -0.02022978166409306, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": 0.09946193196537245, "ret_signed_p1m": -0.09946193196537245, "alpha_spy_p1m": -0.06533107611547795, "alpha_c_p1m": 0.03722369392384106, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.08965935717095963, "ret_signed_p3m": -0.08965935717095963, "alpha_spy_p3m": -0.03192860078014004, "alpha_c_p3m": 0.12982001286651723, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": 0.06987213437582929, "ret_signed_p6m": -0.06987213437582929, "alpha_spy_p6m": -0.015309098297438872, "alpha_c_p6m": 0.25306341111572284, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.1514, -0.1385, -0.1565, -0.1403, -0.1437, -0.1356, -0.1356, -0.1453, -0.1434, -0.1213, -0.0832, -0.079, -0.0627, -0.0613, -0.0892, -0.0657, -0.0533, -0.0533, -0.0598, -0.0494, -0.0323, -0.025, -0.0201, -0.0252, 0.0142, 0.0182, 0.0279, 0.0244, 0.0182, -0.0076, 0.0147, 0.0323, 0.0308, 0.0501, 0.0428, 0.0651, 0.0568, 0.0321, 0.0695, 0.0591, 0.066, 0.074, 0.0855, 0.0817, 0.0882, 0.0789, 0.0815, 0.0721, 0.0814, 0.0436, 0.0421, 0.0396, 0.0575, 0.0683, 0.08, 0.079, 0.0705, 0.0349, 0.0349, 0.0147, 0.0137, 0.0199, -0.0077, 0.0, 0.0146, 0.0536, 0.0527, 0.0566, 0.0561, 0.0391, 0.0118, 0.0472, 0.0497, 0.091, 0.0602, 0.0515, 0.0558, 0.0588, 0.0805, 0.1086, 0.1125, 0.1223, 0.1148, 0.1024, 0.0995, 0.1236, 0.1442, 0.0883, 0.119, 0.0697, 0.0685, 0.0803, 0.0886, 0.0852, 0.0764, 0.0712, 0.0823, 0.1067, 0.1378, 0.1945, 0.2302, 0.2055, 0.2031, 0.2292, 0.1806, 0.1599, 0.1175, 0.1179, 0.1827, 0.1477, 0.1515, 0.1103, 0.1299, 0.1087, 0.0776, 0.1083, 0.0733, 0.0629, 0.0847, 0.0566, 0.0711, 0.0711, 0.0517, 0.069, 0.0782, 0.0671, 0.0897, 0.0839, 0.1026, 0.0991, 0.092, 0.0751, 0.04, 0.0475, 0.056, 0.0157, 0.0348, 0.0755, 0.0915, 0.1243, 0.1207, 0.1207, 0.1321, 0.1184, 0.1144, 0.1082, 0.1082, 0.1222, 0.1178, 0.1126, 0.1237, 0.0995, 0.0985, 0.0989, 0.1041, 0.0882, 0.1115, 0.1066, 0.1066, 0.0581, 0.0893, 0.0983, 0.1152, 0.108, 0.1202, 0.138, 0.1439, 0.1357, 0.1029, 0.0716, 0.0351, 0.0213, 0.1017, 0.1292, 0.1203, 0.1293, 0.1108, 0.0863, 0.0863, 0.0991, 0.117, 0.1165, 0.1279, 0.1382, 0.1459, 0.162, 0.0986, 0.0529, 0.0843, 0.0699]}
{"unit_id": "orig:1965028624073199634", "ticker_idx": 1, "ts": "2025-09-08T12:25:07+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom XPU and GPU Pose a Real Threat", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's PT cut on NVDA to $200 on XPU competition; implied AVGO beneficiary of GPU share shift.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CITI: NVIDIA Price Target Cut, Broadcom XPU and GPU Pose a Real Threat\n\n• CITI lowered its NVIDIA price target to $200 due to intensifying TPU competition. As a result, it forecasts NVIDIA’s GPU revenue in 2026 will decline by about $12 billion. Specifically, the 2026 XPU market is expected to grow +53% YoY, significantly outpacing the AI GPU revenue growth rate of +34% YoY. This growth will be mainly driven by production capacity expansions at Google, Meta, and Amazon.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031129663717725742, "ret_signed_m1d": -0.031129663717725742, "alpha_spy_m1d": -0.028679141619245563, "alpha_c_m1d": -0.020203014982361656, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025979983430521436, "ret_signed_p1d": -0.025979983430521436, "alpha_spy_p1d": -0.028291904128131318, "alpha_c_p1d": -0.0303304905991022, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.05334865965704916, "ret_signed_p1w": 0.05334865965704916, "alpha_spy_p1w": 0.03473053554589822, "alpha_c_p1w": 0.017431888348674818, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": -0.025064755049650667, "ret_signed_p1m": -0.025064755049650667, "alpha_spy_p1m": -0.05919561089954517, "alpha_c_p1m": -0.16175038093886418, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.1042466170647871, "ret_signed_p3m": 0.1042466170647871, "alpha_spy_p3m": 0.04651586067396751, "alpha_c_p3m": -0.11523275297268976, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": -0.08873355372693392, "ret_signed_p6m": -0.08873355372693392, "alpha_spy_p6m": -0.14329658980532434, "alpha_c_p6m": -0.41166909921848605, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.27, -0.2609, -0.2822, -0.2724, -0.2802, -0.2748, -0.2748, -0.2768, -0.2658, -0.2369, -0.2343, -0.2184, -0.2207, -0.2025, -0.2341, -0.2192, -0.2039, -0.2039, -0.2068, -0.2137, -0.196, -0.2032, -0.2062, -0.2027, -0.1872, -0.1876, -0.1713, -0.1803, -0.1662, -0.194, -0.1793, -0.1647, -0.1605, -0.1486, -0.1395, -0.1245, -0.1503, -0.1649, -0.1387, -0.1525, -0.1272, -0.1212, -0.1177, -0.1208, -0.095, -0.1058, -0.0996, -0.1137, -0.1154, -0.1468, -0.1576, -0.1622, -0.1494, -0.1488, -0.1378, -0.1313, -0.107, -0.1396, -0.1396, -0.1372, -0.1252, -0.1144, -0.0311, 0.0, -0.026, 0.0692, 0.0404, 0.0411, 0.0533, 0.0415, 0.0015, -0.0009, -0.0021, -0.0182, -0.0177, -0.0167, -0.026, -0.0305, -0.0497, -0.0439, -0.0338, -0.0199, -0.0194, -0.0277, -0.0251, 0.0013, -0.0001, -0.0592, 0.0337, -0.0027, 0.0182, 0.0263, 0.0124, 0.0121, -0.007, -0.0138, -0.0022, 0.0263, 0.0492, 0.0809, 0.1186, 0.091, 0.0712, 0.0507, 0.0199, 0.0403, 0.0305, 0.0127, 0.0386, 0.02, 0.0294, -0.0147, -0.0075, -0.007, -0.0132, 0.0271, 0.0051, -0.0141, 0.0953, 0.1158, 0.1522, 0.1522, 0.1678, 0.1189, 0.1058, 0.103, 0.1042, 0.1309, 0.1624, 0.1775, 0.1968, 0.1777, 0.0431, -0.0152, -0.0109, -0.0552, -0.044, -0.0136, -0.0086, 0.0143, 0.0169, 0.0169, 0.0224, 0.0145, 0.0158, 0.0049, 0.0049, 0.0094, -0.0028, -0.0018, -0.0026, -0.0346, 0.0017, 0.0227, 0.0296, -0.0131, -0.004, 0.0212, 0.0212, -0.0343, -0.0453, -0.0549, -0.0707, -0.0568, -0.0337, -0.0324, -0.0397, -0.038, -0.0386, -0.0699, -0.1055, -0.0984, -0.0333, -0.0013, -0.0115, -0.0048, -0.0384, -0.0558, -0.0558, -0.0344, -0.0316, -0.0302, -0.0341, -0.0408, -0.0549, -0.0351, -0.0659, -0.0722, -0.0743, -0.0887]}
{"unit_id": "orig:1945974655917122025", "ticker_idx": 0, "ts": "2025-07-17T22:31:27+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TP was raised from NT1,210 to NT1,370... TP was raised to NT$1,310... TP was raised to NT$1,275... N2 revenue contribution is expected to be much larger than N3 during its initial ramp-up phase", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares multiple analyst upgrades (GS/Nomura/JPM/Citi) all raising TPs and EPS on TSMC; relayed bullish thesis adopted by author.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250718_2Q25 TSMC Comment\n\n(1) Goldman Sachs\n\n- TSMC stated that there were no signs of softening demand from AI customers, delivering a positive earnings announcement.\n\n- Consequently, the N2 revenue contribution is expected to be much larger than N3 during its initial ramp-up phase, especially in the first two years, indicating future strength in the HPC segment.\n\n- N5/N3 wafer revenue estimates were upgraded, and the 2026 5nm/CoWoS ASP growth rates were raised from +3%/+5% to +7%/+8% respectively.\n\n- For 2026 GPM, the forecast was raised by +380BP from the previous 54.1% to 57.9%. Ultimately, TP was raised from NT1,210toNT1,370.\n\n(2) Nomura\n\n- Leading-edge process supply remains tight, and customers are still focusing on securing 2026 leading-edge process and CoWoS CAPA rather than tariff uncertainties.\n\n- CoWoS CAPA is judged not to exceed 90-100K per month by the end of 2026.\n\n- It is highly likely that Apple will adopt N2 across all applications in 2026.\n\n- EPS Growth for 2025-2027 was raised by +3-5%, and TP was raised to NT$1,310.\n\n(3) JPMorgan\n\n- Encouragingly robust margin despite the unprecedented strength of the TWD and margin dilution from overseas fabs.\n\n- Q4 2025 is expected to be the bottom for GPM, and 2026 GPM is projected to improve due to 1) ASP increases, 2) N3 process yield improvements, and 3) accelerated cost reduction.\n\n- CoWoS CAPA is expected to remain tight until 2026, and AI demand remains robust from CSPs, Sovereign AI, and new demand from China.\n\n- EPS Growth for 2025-2026 was raised by +5% and +4% respectively, and TP was raised to NT$1,275.\n\n(4) CITI\n\n- Strong margins were recorded due to robust demand in leading-edge processes and high production efficiency, despite negative FX effects.\n\n- N2 process will begin mass production from 2H25 and expand to a larger scale in 2H26; N2P/A16 processes are also expected to enter mass production from 2H26.\n\n- NVIDIA's H20 has not yet received concrete orders, but products based on the license for the recent resumption of China exports will be able to meet China's AI demand.\n\n- This is identified as a factor that can positively impact global AI growth and will be a favorable factor for the overall AI Value-Chain supply chain.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032736137765441486, "ret_signed_m1d": -0.032736137765441486, "alpha_spy_m1d": -0.02665367060126722, "alpha_c_m1d": -0.024405971921902414, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021172647928687227, "ret_signed_p1d": -0.021172647928687227, "alpha_spy_p1d": -0.02044023502657677, "alpha_c_p1d": -0.016373454888251704, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": -0.016286599225248666, "ret_signed_p1w": -0.016286599225248666, "alpha_spy_p1w": -0.02644521420138235, "alpha_c_p1w": -0.002882922782284747, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": -0.02736158446200665, "ret_signed_p1m": -0.02736158446200665, "alpha_spy_p1m": -0.051882372472820015, "alpha_c_p1m": -0.040868208695853725, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.20876894271373292, "ret_signed_p3m": 0.20876894271373292, "alpha_spy_p3m": 0.15140639950328283, "alpha_c_p3m": 0.06551011307794985, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 0.3254373776897599, "ret_signed_p6m": 0.3254373776897599, "alpha_spy_p6m": 0.2139622326356443, "alpha_c_p6m": -0.012984752748134376, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.4002, -0.3858, -0.3598, -0.334, -0.3302, -0.3371, -0.3327, -0.3238, -0.2993, -0.2727, -0.2844, -0.3011, -0.2919, -0.2892, -0.2839, -0.2415, -0.213, -0.2099, -0.2121, -0.2121, -0.215, -0.2152, -0.2221, -0.2041, -0.2212, -0.2212, -0.198, -0.2043, -0.2002, -0.2157, -0.2096, -0.1983, -0.1789, -0.1751, -0.1676, -0.1602, -0.1381, -0.1314, -0.1228, -0.1405, -0.1218, -0.1291, -0.1307, -0.1307, -0.1469, -0.1436, -0.1039, -0.0931, -0.0879, -0.0693, -0.0778, -0.0852, -0.0489, -0.044, -0.044, -0.0669, -0.0722, -0.056, -0.0645, -0.0619, -0.0689, -0.0352, -0.0327, 0.0, -0.0212, -0.0275, -0.0448, -0.0215, -0.0163, 0.0, -0.0116, -0.0174, -0.011, -0.0162, -0.0423, -0.0269, -0.0535, -0.0579, -0.0121, -0.0154, -0.0143, -0.0053, -0.0169, -0.0187, -0.0274, -0.0171, -0.0525, -0.0692, -0.0744, -0.0513, -0.0408, -0.028, -0.0257, -0.0298, -0.06, -0.06, -0.0701, -0.0579, -0.0423, -0.0089, 0.0065, 0.0217, 0.0604, 0.0542, 0.0559, 0.0643, 0.0704, 0.0734, 0.0973, 0.0819, 0.1136, 0.1547, 0.1466, 0.13, 0.1165, 0.116, 0.1408, 0.1783, 0.1768, 0.1935, 0.2352, 0.201, 0.2438, 0.2249, 0.1464, 0.2372, 0.2088, 0.2446, 0.2247, 0.2053, 0.216, 0.2029, 0.1799, 0.1875, 0.2048, 0.2182, 0.2316, 0.2461, 0.2385, 0.2271, 0.2452, 0.201, 0.1994, 0.1814, 0.1702, 0.206, 0.1893, 0.187, 0.1526, 0.1633, 0.1519, 0.1351, 0.1533, 0.1335, 0.1235, 0.1626, 0.1628, 0.1843, 0.1843, 0.1907, 0.175, 0.193, 0.2068, 0.1965, 0.2038, 0.233, 0.2393, 0.2668, 0.2485, 0.1961, 0.1784, 0.1749, 0.1343, 0.1659, 0.1834, 0.2011, 0.2162, 0.2237, 0.2237, 0.2403, 0.2324, 0.2269, 0.2446, 0.2446, 0.309, 0.3198, 0.341, 0.3052, 0.3024, 0.3254]}
{"unit_id": "orig:1944891592143503374", "ticker_idx": 0, "ts": "2025-07-14T22:47:44+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Potential for Gradual Upside to Nvidia's Revenue Estimates... resumption of even partial revenue from the Chinese AI market... would have a substantial impact on Nvidia's overall revenue performance. This represents an upside not currently reflected in market consensus estimates.", "tweet_type": "analysis", "timeline": "years", "summary": "Stifel analysis: NVDA China chip resumption = gradual revenue upside above consensus + long-term US-China trade thaw benefiting US AI ecosystem.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Stifel) Our Analysis on Nvidia Resuming AI Chip Production Compliant with China Export Controls\n\nReports last week indicated that Nvidia is set to release another scaled-down version of its AI chip for the Chinese market. Currently, U.S. export controls restrict China's access to cutting-edge AI infrastructure, leading Nvidia to sell a series of modified chips compliant with these export control regulations in the Chinese region.\n\nThe upcoming AI chip platform is a derivative of the server-oriented 'RTX Pro 6000D' platform, expected to be priced between approximately $6,500 and $8,000. This chip will not include Nvidia's proprietary networking technology, 'NVLink'. Reports related to this processor emerged ahead of the International Supply Chain Expo in Beijing this week.\n\nOur analysis of this news is as follows:\n\n1. Potential for Gradual Upside to Nvidia's Revenue Estimates: Following the most recent U.S. export control measures, Nvidia has not included China-based revenue in its revenue guidance. Although this new chip is not expected to generate as much revenue as the previously restricted H20 chips, the resumption of even partial revenue from the Chinese AI market (estimated by CEO Jensen Huang to reach $50 billion in the coming years) would have a substantial impact on Nvidia's overall revenue performance. This represents an upside not currently reflected in market consensus estimates.\n\n2. Potential Thawing of U.S.-China Trade Relations: Beyond a short-term perspective, and considering CEO Jensen Huang's direct communication with the White House, we view Nvidia's resumption of AI chip sales for the Chinese market as indicative of improving prospects for U.S.-China trade relations. If this holds true, it is expected to act as an upside factor that positively impacts the entire U.S. AI ecosystem in the long run, much like the proverb, \"a rising tide lifts all boats.\"\n\nHowever, these potential tailwinds are still several years away, and inherent obstacles in U.S.-China relations, such as national security concerns, still persist. Therefore, we will continue to pay close attention to further news related to this and continuously monitor the evolving AI supply chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005180683499361294, "ret_signed_m1d": 0.005180683499361294, "alpha_spy_m1d": 0.007085166748686311, "alpha_c_m1d": -0.0024247053453889666, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.040409535932759866, "ret_signed_p1d": 0.040409535932759866, "alpha_spy_p1d": 0.044682702983386946, "alpha_c_p1d": 0.021203099274968107, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.04455419435283536, "ret_signed_p1w": 0.04455419435283536, "alpha_spy_p1w": 0.03821616236811698, "alpha_c_p1w": 0.0263641341996117, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.11635274151788177, "ret_signed_p1m": 0.11635274151788177, "alpha_spy_p1m": 0.0877358414456213, "alpha_c_p1m": 0.06455133674457092, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.17377251073301703, "ret_signed_p3m": 0.17377251073301703, "alpha_spy_p3m": 0.09661177882903127, "alpha_c_p3m": -0.036132036996605166, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.1413481255415343, "ret_signed_p6m": 0.1413481255415343, "alpha_spy_p6m": 0.02776489035331986, "alpha_c_p6m": -0.22142382598611499, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.3632, -0.3815, -0.3815, -0.4094, -0.3973, -0.374, -0.3514, -0.3234, -0.3373, -0.3356, -0.3362, -0.3198, -0.3022, -0.3063, -0.308, -0.2866, -0.2847, -0.2891, -0.2504, -0.2081, -0.1752, -0.1783, -0.1748, -0.1738, -0.181, -0.1967, -0.1905, -0.1998, -0.1998, -0.1742, -0.1784, -0.1517, -0.1764, -0.1627, -0.1393, -0.1351, -0.1468, -0.1363, -0.1307, -0.1226, -0.1295, -0.1162, -0.1347, -0.1181, -0.1216, -0.1133, -0.1133, -0.1232, -0.1213, -0.0986, -0.0595, -0.0552, -0.0385, -0.0371, -0.0656, -0.0416, -0.0288, -0.0288, -0.0355, -0.0248, -0.0073, 0.0002, 0.0052, 0.0, 0.0404, 0.0445, 0.0544, 0.0508, 0.0446, 0.018, 0.0409, 0.0589, 0.0575, 0.0773, 0.0697, 0.0926, 0.0841, 0.0588, 0.0971, 0.0865, 0.0936, 0.1018, 0.1135, 0.1096, 0.1164, 0.1068, 0.1094, 0.0998, 0.1093, 0.0705, 0.0691, 0.0665, 0.0848, 0.0959, 0.1079, 0.1068, 0.0981, 0.0616, 0.0616, 0.0409, 0.0399, 0.0463, 0.018, 0.0258, 0.0408, 0.0808, 0.0799, 0.0839, 0.0834, 0.0659, 0.038, 0.0742, 0.0769, 0.1192, 0.0876, 0.0787, 0.0831, 0.0861, 0.1084, 0.1373, 0.1413, 0.1513, 0.1436, 0.1309, 0.1279, 0.1527, 0.1738, 0.1164, 0.1479, 0.0973, 0.0961, 0.1082, 0.1168, 0.1132, 0.1042, 0.0989, 0.1103, 0.1353, 0.1672, 0.2253, 0.262, 0.2367, 0.2342, 0.261, 0.2111, 0.1899, 0.1464, 0.1468, 0.2133, 0.1774, 0.1813, 0.139, 0.1591, 0.1374, 0.1054, 0.1369, 0.1011, 0.0903, 0.1127, 0.0839, 0.0987, 0.0987, 0.0789, 0.0967, 0.1061, 0.0947, 0.1178, 0.1119, 0.131, 0.1275, 0.1203, 0.1029, 0.0669, 0.0746, 0.0833, 0.042, 0.0615, 0.1033, 0.1197, 0.1534, 0.1497, 0.1497, 0.1614, 0.1473, 0.1432, 0.1368, 0.1368, 0.1512, 0.1467, 0.1413]}
{"unit_id": "orig:1945645307859276068", "ticker_idx": 0, "ts": "2025-07-17T00:42:44+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Investment opinion has been downgraded from Buy to Neutral, with a target price of KRW 310,000, representing approximately 5% upside potential from the current stock price.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs downgrades SK Hynix to Neutral from Buy (PT KRW 310K) on 2026 HBM oversupply, ASP decline, and 19%-below-consensus operating profit forecast.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix, Downgraded to Neutral (Goldman Sachs)\n\n- Investment opinion has been downgraded from Buy to Neutral, with a target price of KRW 310,000, representing approximately 5% upside potential from the current stock price.\n\n- The long-term growth potential of the HBM market itself is positive, but from 2026 onwards, oversupply and intensified competition are expected to exert downward pressure on prices.\n\n- It is highly probable that HBM prices will record their first decline in 2026. Customer bargaining power is strengthening, and SK Hynix, with its high proportion of GPU-bound HBM, is vulnerable to price defense.\n\n- In 2025, strong demand for HBM and general-purpose memory will lead to upward revisions in earnings. However, the 2026 earnings outlook has been lowered under the premise that HBM average selling prices (ASPs) will fall by a double-digit percentage.\n\n- The 2026 operating profit forecast is calculated to be approximately 19% lower than the consensus, predicting negative growth unlike market expectations (17% annual growth).\n\n- SK Hynix's stock price has risen 73% year-to-date, outperforming KOSPI (which rose 33%). It is assessed that the short-term valuation burden has increased.\n\n- Currently trading at a 12-month forward P/B of 1.8x (historical average 1.3x), it is judged that the risk and reward are nearly balanced.\n\n- HBM demand is still predominantly driven by GPUs, but ASICs (application-specific AI chips) are rapidly increasing, leading the overall growth trend.\n2026 HBM demand growth is expected to be 23% for GPUs and 82% for ASICs.\nWithin GPUs, the transition from B300 to Rubin shows limited increase in HBM content, leading to a slowdown in GPU contribution growth.\nASIC families like TPU and Trainium are seeing rapid increases in HBM content, leading to strong demand growth momentum.\n\n- On the HBM supply side, total industry production capacity is expected to expand to 485K WPM (wafer per month) level by the end of 2026, with supply likely to exceed demand.\nSupply growth rate is revised up to 48%, while demand growth is expected at 38%.\n2026 HBM ASP is projected to decrease by approximately 10% year-on-year.\n\nScenario-based Outlook:\nBase Scenario: HBM3E ASP drop of 30%, HBM4 premium of 45%, HBM margin of 59%, operating profit of KRW 37 trillion (19% lower than consensus).\nBull Case Scenario: Samsung Electronics' delayed entry to key customers, HBM3E ASP drop of 10%, HBM4 premium of 55%, margin of 65%, operating profit of KRW 45 trillion (at consensus level).\nBear Case Scenario: Samsung Electronics' early entry, HBM3E ASP drop of 35%, HBM4 premium of 35%, margin of 50%, operating profit of KRW 32 trillion (29% lower than consensus).\n\nQ2 2025 earnings are revised upward to around KRW 9 trillion in operating profit, expected to meet expectations due to stronger demand compared to the previous quarter.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09833034315341949, "ret_signed_m1d": -0.09833034315341949, "alpha_spy_m1d": -0.10441281031759375, "alpha_c_m1d": -0.10666050899695856, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0018553554029104857, "ret_signed_p1d": 0.0018553554029104857, "alpha_spy_p1d": 0.0011229425008000282, "alpha_c_p1d": -0.0029438376375250375, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": 0.0, "ret_signed_p1w": -0.0, "alpha_spy_p1w": 0.010158614976133684, "alpha_c_p1w": -0.013403676442963919, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": 0.025973927680496, "ret_signed_p1m": -0.025973927680496, "alpha_spy_p1m": -0.0014531396696826349, "alpha_c_p1m": -0.012467303446648925, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.5291072123884737, "ret_signed_p3m": -0.5291072123884737, "alpha_spy_p3m": -0.4717446691780236, "alpha_c_p3m": -0.3858483827526906, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 1.7666355538912395, "ret_signed_p6m": -1.7666355538912395, "alpha_spy_p6m": -1.655160408837124, "alpha_c_p6m": -1.4282134234533452, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.3459, -0.3563, -0.3296, -0.3396, -0.317, -0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666]}
{"unit_id": "orig:1955808354946965760", "ticker_idx": 0, "ts": "2025-08-14T01:47:03+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD is still expected to continue expanding its market share in both the client and server segments, directly eroding Intel's share", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS data shows AMD taking share from Intel in client and server CPUs; AMD expected to keep expanding share through 2025 at Intel's expense.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CPU Industry Update: Server CPUs Continue to Outperform Seasonality, AMD's Momentum Persists\nUBS (T. Arcuri, 13/08/25)\n\nUBS data shows that server CPU shipments in the second calendar quarter (CQ2) once again surpassed seasonal trends, while PC MPU (Micro-Processor Unit) shipments fell below seasonality. AMD continues to take share from Intel in both the client and server segments, particularly expanding its market share in dollar terms.\n\nPC MPU Below Seasonality\nAccording to Mercury Research data, global PC (desktop + notebook) MPU shipments in CQ2:25 grew approximately 2% quarter-over-quarter (QoQ). This is below the 3-year seasonal average and showed no clear signs of tariff-related pull-in. Desktop and notebook CPU shipments increased by 1.6% and 2.0% QoQ, respectively. Intel's notebook share gained 190 basis points (bps) QoQ, but its desktop share fell by 420 bps QoQ, maintaining an overall market share of about 76%. In dollar terms, AMD's market share increased by 120 bps QoQ, benefiting from a $10 increase in notebook Average Selling Price (ASP) (a $30 increase on a trailing-twelve-month basis) and a $3 increase in desktop ASP.\n\nServer CPU Shipments Again Exceed Seasonality\nServer CPU shipments were flat QoQ, significantly outperforming the 3-year seasonal trend of a ~4% decline. Shipments were up 18% year-over-year (YoY), marking the fourth consecutive quarter of positive growth and reflecting strong cloud demand. Intel's unit shipments were flat QoQ (+13% YoY), driven by a 12% QoQ increase in Emerald Rapids and a small-scale ramp-up of Granite Rapids/Sierra Forest; Sapphire Rapids, while gradually declining, remains the main product. AMD's overall shipments grew 1% QoQ (+34% YoY), driven by an 88% QoQ increase in Turin to 435,000 units, which was partially offset by declines in Bergamo and Genoa shipments. ARM-based server CPUs grew 15% QoQ (+147% YoY), but their absolute scale remains small. AMD's server CPU unit market share held steady at about 27%, while its dollar market share increased by 150 bps QoQ.\n\n2025 Outlook\nUBS forecasts global PC shipments to grow by approximately 2% YoY in 2025, a downward revision from the previous forecast of 5%. This is mainly because the replacement cycle benefits from the end-of-life of Windows 10 are being offset by uncertainties in the tariff environment, declining commercial procurement budgets, and macroeconomic pressures on the consumer side. Global PC shipments in 2025 are projected to be around 249 million units, with desktop and notebook shipments up 1.1% and 2.2% YoY, respectively. In this environment, AMD is still expected to continue expanding its market share in both the client and server segments, directly eroding Intel's share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019176575182234812, "ret_signed_m1d": 0.019176575182234812, "alpha_spy_m1d": 0.019269531995301126, "alpha_c_m1d": 0.021296294615984035, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01901079027036512, "ret_signed_p1d": -0.01901079027036512, "alpha_spy_p1d": -0.01666959562834236, "alpha_c_p1d": 0.0017562954936088948, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.09527488546632146, "ret_signed_p1w": -0.09527488546632146, "alpha_spy_p1w": -0.08070013520902009, "alpha_c_p1w": -0.0473813528513769, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": -0.12368051948862357, "ret_signed_p1m": -0.12368051948862357, "alpha_spy_p1m": -0.14299988474056524, "alpha_c_p1m": -0.13162972084736924, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.31262784348318995, "ret_signed_p3m": 0.31262784348318995, "alpha_spy_p3m": 0.25069503274402494, "alpha_c_p3m": 0.14990262202594318, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.15192045292505374, "ret_signed_p6m": 0.15192045292505374, "alpha_spy_p6m": 0.07496710006633589, "alpha_c_p6m": -0.1825382399220643, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.3659, -0.3727, -0.3807, -0.3882, -0.3904, -0.3904, -0.3669, -0.3763, -0.3754, -0.3881, -0.3665, -0.3517, -0.3447, -0.3607, -0.3579, -0.3273, -0.3189, -0.3305, -0.3451, -0.3581, -0.3015, -0.2976, -0.2993, -0.2993, -0.2913, -0.2839, -0.235, -0.2075, -0.206, -0.2053, -0.2158, -0.2478, -0.2345, -0.2379, -0.2379, -0.255, -0.2384, -0.2351, -0.2033, -0.1908, -0.1918, -0.14, -0.1153, -0.1135, -0.1324, -0.1324, -0.145, -0.1232, -0.1041, -0.08, -0.0403, -0.0194, -0.008, -0.0256, -0.0511, -0.023, -0.0367, -0.0985, -0.0473, -0.0453, -0.0479, -0.0332, 0.0192, 0.0, -0.019, -0.0266, -0.0796, -0.087, -0.0953, -0.0729, -0.0972, -0.0792, -0.0764, -0.0684, -0.1012, -0.1012, -0.103, -0.104, -0.1059, -0.1647, -0.1632, -0.1389, -0.1183, -0.1397, -0.1237, -0.1094, -0.1132, -0.1204, -0.1273, -0.1302, -0.1169, -0.1108, -0.1109, -0.1088, -0.1188, -0.1083, -0.1059, -0.0936, -0.062, -0.09, 0.1258, 0.1689, 0.3018, 0.287, 0.1876, 0.196, 0.2053, 0.3186, 0.2963, 0.2881, 0.3294, 0.3154, 0.2723, 0.2986, 0.3977, 0.435, 0.4259, 0.4608, 0.4083, 0.4154, 0.4349, 0.3819, 0.4166, 0.3136, 0.2906, 0.3483, 0.3126, 0.4307, 0.3703, 0.364, 0.3292, 0.2727, 0.2354, 0.1385, 0.1262, 0.1884, 0.1392, 0.184, 0.184, 0.2022, 0.2145, 0.1895, 0.2025, 0.1936, 0.2046, 0.2219, 0.2248, 0.2237, 0.2237, 0.1649, 0.1472, 0.156, 0.0948, 0.1111, 0.1795, 0.1879, 0.1876, 0.1884, 0.1884, 0.1881, 0.1915, 0.1901, 0.1835, 0.1835, 0.235, 0.2218, 0.1846, 0.1607, 0.1311, 0.1228, 0.1478, 0.2212, 0.2357, 0.2596, 0.2812, 0.2812, 0.2817, 0.3805, 0.4022, 0.4351, 0.3888, 0.3928, 0.3967, 0.3936, 0.3083, 0.361, 0.338, 0.1063, 0.0638, 0.1519]}
{"unit_id": "orig:1955808354946965760", "ticker_idx": 1, "ts": "2025-08-14T01:47:03+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "directly eroding Intel's share", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS data shows AMD taking share from Intel in client and server CPUs; AMD expected to keep expanding share through 2025 at Intel's expense.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CPU Industry Update: Server CPUs Continue to Outperform Seasonality, AMD's Momentum Persists\nUBS (T. Arcuri, 13/08/25)\n\nUBS data shows that server CPU shipments in the second calendar quarter (CQ2) once again surpassed seasonal trends, while PC MPU (Micro-Processor Unit) shipments fell below seasonality. AMD continues to take share from Intel in both the client and server segments, particularly expanding its market share in dollar terms.\n\nPC MPU Below Seasonality\nAccording to Mercury Research data, global PC (desktop + notebook) MPU shipments in CQ2:25 grew approximately 2% quarter-over-quarter (QoQ). This is below the 3-year seasonal average and showed no clear signs of tariff-related pull-in. Desktop and notebook CPU shipments increased by 1.6% and 2.0% QoQ, respectively. Intel's notebook share gained 190 basis points (bps) QoQ, but its desktop share fell by 420 bps QoQ, maintaining an overall market share of about 76%. In dollar terms, AMD's market share increased by 120 bps QoQ, benefiting from a $10 increase in notebook Average Selling Price (ASP) (a $30 increase on a trailing-twelve-month basis) and a $3 increase in desktop ASP.\n\nServer CPU Shipments Again Exceed Seasonality\nServer CPU shipments were flat QoQ, significantly outperforming the 3-year seasonal trend of a ~4% decline. Shipments were up 18% year-over-year (YoY), marking the fourth consecutive quarter of positive growth and reflecting strong cloud demand. Intel's unit shipments were flat QoQ (+13% YoY), driven by a 12% QoQ increase in Emerald Rapids and a small-scale ramp-up of Granite Rapids/Sierra Forest; Sapphire Rapids, while gradually declining, remains the main product. AMD's overall shipments grew 1% QoQ (+34% YoY), driven by an 88% QoQ increase in Turin to 435,000 units, which was partially offset by declines in Bergamo and Genoa shipments. ARM-based server CPUs grew 15% QoQ (+147% YoY), but their absolute scale remains small. AMD's server CPU unit market share held steady at about 27%, while its dollar market share increased by 150 bps QoQ.\n\n2025 Outlook\nUBS forecasts global PC shipments to grow by approximately 2% YoY in 2025, a downward revision from the previous forecast of 5%. This is mainly because the replacement cycle benefits from the end-of-life of Windows 10 are being offset by uncertainties in the tariff environment, declining commercial procurement budgets, and macroeconomic pressures on the consumer side. Global PC shipments in 2025 are projected to be around 249 million units, with desktop and notebook shipments up 1.1% and 2.2% YoY, respectively. In this environment, AMD is still expected to continue expanding its market share in both the client and server segments, directly eroding Intel's share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06873433591973854, "ret_signed_m1d": 0.06873433591973854, "alpha_spy_m1d": 0.06864137910667223, "alpha_c_m1d": 0.06661461648598932, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02933775514184722, "ret_signed_p1d": -0.02933775514184722, "alpha_spy_p1d": -0.03167894978386998, "alpha_c_p1d": -0.050104840905821235, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.015088038606142273, "ret_signed_p1w": 0.015088038606142273, "alpha_spy_p1w": 0.0005132883488409012, "alpha_c_p1w": -0.03280549400880228, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.009220423626437269, "ret_signed_p1m": -0.009220423626437269, "alpha_spy_p1m": 0.010098941625504398, "alpha_c_p1m": -0.0012712222676916074, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.5875943042382468, "ret_signed_p3m": -0.5875943042382468, "alpha_spy_p3m": -0.5256614934990818, "alpha_c_p3m": -0.424869082781, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 1.1202849479659958, "ret_signed_p6m": -1.1202849479659958, "alpha_spy_p6m": -1.043331595107278, "alpha_c_p6m": -0.7858262551188777, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.1039, -0.1085, -0.1329, -0.1387, -0.1597, -0.1597, -0.1387, -0.1463, -0.1513, -0.1806, -0.1727, -0.1496, -0.1513, -0.1622, -0.1593, -0.1417, -0.0746, -0.1333, -0.1295, -0.1559, -0.1308, -0.1282, -0.0993, -0.0993, -0.1165, -0.1119, -0.0549, -0.0696, -0.057, -0.049, -0.0612, -0.0423, -0.083, -0.0574, -0.0574, -0.078, -0.0113, -0.0176, -0.0017, -0.018, -0.0235, -0.0394, -0.049, -0.0444, -0.0319, -0.0251, -0.026, -0.0155, -0.0516, -0.1324, -0.1333, -0.1446, -0.1475, -0.1702, -0.1907, -0.1827, -0.1538, -0.1446, -0.1714, -0.1639, -0.1345, -0.0859, -0.0687, 0.0, 0.0293, -0.0084, 0.0608, -0.0134, -0.0151, 0.0394, 0.0289, 0.0205, 0.0415, 0.0448, 0.0205, 0.0205, 0.0147, 0.0059, 0.0314, 0.0264, 0.026, 0.0243, 0.0381, 0.0314, 0.0092, 0.0381, 0.0591, 0.0436, 0.2812, 0.2397, 0.2054, 0.2297, 0.3085, 0.4246, 0.4878, 0.4451, 0.4061, 0.5063, 0.5633, 0.5436, 0.5335, 0.5578, 0.5687, 0.5842, 0.5243, 0.5599, 0.4933, 0.557, 0.544, 0.5511, 0.5968, 0.5977, 0.5474, 0.5993, 0.6044, 0.6572, 0.7406, 0.7326, 0.6832, 0.676, 0.6555, 0.552, 0.6085, 0.5608, 0.5981, 0.6115, 0.5876, 0.588, 0.505, 0.4887, 0.4547, 0.4388, 0.4715, 0.4091, 0.4459, 0.5, 0.5017, 0.5427, 0.5427, 0.6999, 0.6769, 0.8219, 0.834, 0.6974, 0.7355, 0.689, 0.6974, 0.7091, 0.6559, 0.5847, 0.5721, 0.5637, 0.5109, 0.5205, 0.5432, 0.5243, 0.5235, 0.5155, 0.5155, 0.5172, 0.5373, 0.5633, 0.5465, 0.5465, 0.6505, 0.65, 0.6781, 0.7867, 0.723, 0.9091, 0.8466, 0.982, 1.0419, 1.0251, 0.9681, 0.9681, 1.0352, 1.2737, 1.2766, 0.8889, 0.7808, 0.8412, 1.0444, 1.0394, 0.9476, 1.0457, 1.0641, 1.0369, 1.0218, 1.1203]}
{"unit_id": "orig:1972905654890471697", "ticker_idx": 0, "ts": "2025-09-30T06:05:37+00:00", "symbol": "SanDisk", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's not too late to buy SanDisk and Kioxia yet, right? I should buy them now", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author signals intent to buy both SanDisk and Kioxia, implying bullish forward stance on both.", "resolved_tickers": ["SNDK"], "resolution_method": "exchange_ticker", "resolution_reason": "SanDisk (spun off from WD) listed on Nasdaq as SNDK", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Fuck, it’s not too late to buy SanDisk and Kioxia yet, right? I should buy them now. https://t.co/7rHeZIrHG8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011586480277334665, "ret_signed_m1d": 0.011586480277334665, "alpha_spy_m1d": 0.015339210999832043, "alpha_c_m1d": 0.017014821193820806, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.005428340916486141, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0795009451065729, "ret_signed_p1d": 0.0795009451065729, "alpha_spy_p1d": 0.0760933756806399, "alpha_c_p1d": 0.06970891208009422, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.009792033026478686, "ret_p1w": 0.07798574187160079, "ret_signed_p1w": 0.07798574187160079, "alpha_spy_p1w": 0.07357254233602983, "alpha_c_p1w": 0.06443302096430203, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.013552720907298754, "ret_p1m": 0.8213904293118899, "ret_signed_p1m": 0.8213904293118899, "alpha_spy_p1m": 0.7895522021546617, "alpha_c_p1m": 0.7423795039045118, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.07901092540737809, "ret_p3m": 1.2286097134842686, "ret_signed_p3m": 1.2286097134842686, "alpha_spy_p3m": 1.1893263774312313, "alpha_c_p3m": 1.1872966495181576, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.04131306396611101, "ret_p6m": 5.041533010595883, "ret_signed_p6m": 5.041533010595883, "alpha_spy_p6m": 5.049969283417784, "alpha_c_p6m": 5.068441938608752, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": -0.026908928012868705, "price_path": [-0.5864, -0.5864, -0.597, -0.5885, -0.5882, -0.5816, -0.5892, -0.6214, -0.6193, -0.6314, -0.6299, -0.624, -0.6291, -0.6314, -0.6168, -0.6251, -0.6214, -0.6266, -0.6174, -0.6133, -0.6175, -0.6316, -0.6211, -0.6263, -0.6248, -0.6373, -0.6048, -0.6135, -0.5826, -0.581, -0.584, -0.603, -0.5943, -0.6027, -0.6043, -0.5945, -0.5867, -0.5831, -0.578, -0.5683, -0.5466, -0.5324, -0.5324, -0.5448, -0.5275, -0.443, -0.389, -0.3717, -0.3716, -0.3412, -0.2487, -0.2324, -0.1971, -0.184, -0.1625, -0.1188, -0.089, -0.0827, -0.0517, -0.1102, -0.1596, -0.1344, 0.0116, 0.0, 0.0795, 0.1064, 0.1445, 0.0799, 0.078, 0.1754, 0.1558, 0.042, 0.1997, 0.1345, 0.2861, 0.2858, 0.2492, 0.3194, 0.3306, 0.3097, 0.4889, 0.6592, 0.573, 0.5644, 0.8214, 0.7453, 0.7766, 0.845, 0.7341, 0.9296, 0.8511, 1.1344, 1.3881, 1.4205, 1.5232, 1.1709, 1.2652, 1.3697, 1.183, 1.1922, 0.7465, 0.7849, 1.0228, 0.9652, 0.9166, 0.9166, 0.99, 0.8732, 0.8302, 0.7324, 0.9012, 1.0363, 1.0095, 0.956, 1.0754, 1.1534, 0.8376, 0.7992, 0.8655, 0.8434, 0.956, 1.1177, 1.1484, 1.1827, 1.2289, 1.2289, 1.2286, 1.1769, 1.141, 1.1157, 1.1157, 1.4531, 1.4428, 2.1161, 2.1512, 1.9816, 2.3637, 2.4694, 2.4742, 2.4564, 2.6474, 2.6865, 2.6865, 3.0385, 3.4678, 3.487, 3.2231, 3.1961, 3.2908, 3.7026, 3.8066, 4.1359, 4.9291, 5.1988, 4.2099, 4.1355, 4.3293, 4.1996, 3.8275, 4.3417, 4.6176, 4.5843, 4.5843, 4.2637, 4.3512, 4.5356, 4.793, 4.9402, 4.6909, 4.6362, 4.8102, 4.6627, 4.5176, 4.0393, 4.3392, 4.0409, 3.6999, 4.2471, 4.516, 4.8416, 4.5153, 4.8968, 5.2712, 5.4186, 5.7174, 5.8814, 5.3254, 5.2611, 5.261, 5.0415]}
{"unit_id": "orig:1972905654890471697", "ticker_idx": 1, "ts": "2025-09-30T06:05:37+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's not too late to buy SanDisk and Kioxia yet, right? I should buy them now", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author signals intent to buy both SanDisk and Kioxia, implying bullish forward stance on both.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Fuck, it’s not too late to buy SanDisk and Kioxia yet, right? I should buy them now. https://t.co/7rHeZIrHG8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04102564102564099, "ret_signed_m1d": -0.04102564102564099, "alpha_spy_m1d": -0.037272910303143614, "alpha_c_m1d": -0.029688514861666127, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.034871794871794926, "ret_signed_p1d": -0.034871794871794926, "alpha_spy_p1d": -0.038279364297727936, "alpha_c_p1d": -0.05733171440460394, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.21025641025641018, "ret_signed_p1w": 0.21025641025641018, "alpha_spy_p1w": 0.20584321072083922, "alpha_c_p1w": 0.17750110636392957, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 1.0676923076923077, "ret_signed_p1m": 1.0676923076923077, "alpha_spy_p1m": 1.0358540805350795, "alpha_c_p1m": 0.9382339862285713, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 1.3415384615384616, "ret_signed_p3m": 1.3415384615384616, "alpha_spy_p3m": 1.3022551254854242, "alpha_c_p3m": 1.2170182363548927, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 3.604102564102564, "ret_signed_p6m": 3.604102564102564, "alpha_spy_p6m": 3.6125388369244655, "alpha_c_p6m": 3.3776605757358165, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.4773, -0.5118, -0.4954, -0.4761, -0.4445, -0.4755, -0.4896, -0.5001, -0.4974, -0.497, -0.5048, -0.5173, -0.5173, -0.5081, -0.48, -0.4736, -0.496, -0.4966, -0.5019, -0.497, -0.4917, -0.5028, -0.5114, -0.5138, -0.5214, -0.5282, -0.5151, -0.5151, -0.463, -0.4712, -0.4759, -0.4749, -0.4991, -0.4927, -0.5163, -0.5272, -0.5071, -0.5093, -0.5009, -0.5046, -0.4931, -0.4675, -0.4511, -0.4634, -0.4619, -0.4613, -0.3713, -0.3538, -0.3487, -0.2851, -0.1785, -0.0892, -0.0892, -0.0349, -0.0892, -0.0728, -0.0708, -0.0113, -0.0113, 0.0154, 0.0031, -0.0985, -0.041, 0.0, -0.0349, 0.1097, 0.2677, 0.2964, 0.2103, 0.2062, 0.2759, 0.2636, 0.2636, 0.2308, 0.2964, 0.4113, 0.3456, 0.4667, 0.4441, 0.4995, 0.5118, 0.801, 1.0123, 0.9241, 1.0677, 1.2359, 1.2205, 1.2205, 1.2072, 1.1631, 1.359, 1.4697, 1.7323, 1.7067, 1.7159, 1.6718, 1.0564, 1.2903, 1.1046, 1.2226, 1.3251, 1.0574, 1.0574, 1.0211, 0.7202, 0.8562, 0.9294, 0.8219, 0.8907, 0.8484, 0.8605, 0.9344, 1.0821, 1.0254, 0.9426, 0.9713, 1.0287, 0.8872, 0.7822, 0.9077, 0.9508, 0.9159, 1.0769, 1.0414, 1.1754, 1.2144, 1.3415, 1.1897, 1.1405, 1.1405, 1.1405, 1.1405, 1.3282, 1.3795, 1.6031, 1.6667, 1.6031, 1.6031, 1.8072, 1.7241, 1.7959, 2.0256, 2.12, 2.119, 2.3846, 2.6738, 2.5559, 2.5795, 2.7846, 2.8862, 2.9477, 3.3815, 2.7662, 3.2636, 3.2872, 3.0062, 2.9169, 2.8769, 2.8656, 2.8656, 3.3436, 3.6862, 3.5744, 3.5887, 3.3938, 3.359, 3.2174, 3.2174, 3.5682, 3.4031, 3.3569, 3.3508, 3.4133, 3.1436, 2.9477, 3.1641, 3.0974, 2.6985, 3.0144, 3.3856, 3.3579, 3.321, 3.639, 3.4328, 3.7979, 3.5867, 3.5867, 3.4031, 3.3282, 3.6041]}
{"unit_id": "thread:1956967786225504737", "ticker_idx": 0, "ts": "2025-08-17T06:34:13+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Quanta expects strong growth in 2026, with Meta and Google as major customers", "tweet_type": "analysis", "timeline": "quarters", "summary": "Quanta and Foxconn both set for ASIC server shipment surge in 2026 from major CSP customers, with margin improvement vs GPU servers.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Quanta and Foxconn ramp up ASIC server efforts as 2026 market competition intensifies\n\n- Both Quanta and Foxconn expect a surge in ASIC server shipments in the 2026 AI server market, anticipating a margin relief effect.\n\n- While the two companies won in the Nvidia GB series AI server market in 2024–2025, the ASIC server market for CSPs is projected to expand in earnest in 2026.\n\n- Quanta’s ASIC shipments in 2025 were minimal, but the company expects strong growth in 2026, with Meta and Google as major customers.\n\n- Foxconn had 20% of its 2024 AI server revenue from ASICs, and it anticipates growth in 2026 based on confirmed orders from the three major CSPs—AWS, Google, and Microsoft.\n\n- Currently, the AI server market is led by GPU products, and according to Digitimes Research, GPU shipments in 2025 will reach 62.73 million units, while ASIC shipments will stand at 50.62 million units.\n\n- However, more and more institutions expect ASIC server growth to outpace GPU growth. Notably, Goldman Sachs projects a 40% growth rate for ASIC servers in 2026 (vs. 25.6% for GPUs).\n\n- This is because CSPs are accelerating their own ASIC development for cost efficiency, workload customization, and to avoid supply chain dependency.\n\n- Expansion of Nvidia GB200 and GB300 shipments exerts margin pressure on ODMs due to high cabinet costs.\n\n- In contrast, ASIC servers, while more complex in design, are more profitable than GPU servers, making them advantageous for ODMs to defend their margins.\n\n- As a result, the expansion of ASIC shipments in 2026 is expected to be the key breakthrough for ODMs in improving profitability and driving strategic transformation.\n\n---\n\nhttps://t.co/TnziTtbKJy", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005576208178438735, "ret_signed_m1d": 0.005576208178438735, "alpha_spy_m1d": 0.005358529367495146, "alpha_c_m1d": 0.00771684584035881, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.002140637661920075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0037174721189591198, "ret_signed_p1d": -0.0037174721189591198, "alpha_spy_p1d": 0.0017076579477185438, "alpha_c_p1d": 0.013745951653879707, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.017463423772838826, "ret_p1w": -0.0037174721189591198, "ret_signed_p1w": -0.0037174721189591198, "alpha_spy_p1w": -0.0024271869472779484, "alpha_c_p1w": 0.012957338069472746, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.016674810188431866, "ret_p1m": 0.016728624535315983, "ret_signed_p1m": 0.016728624535315983, "alpha_spy_p1m": -0.009231319884346556, "alpha_c_p1m": -0.006180525292519334, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.022909149827835318, "ret_p3m": 0.059479553903345694, "ret_signed_p3m": 0.059479553903345694, "alpha_spy_p3m": 0.011907404620269624, "alpha_c_p3m": -0.01821723873052239, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.07769679263386808, "ret_p6m": 0.059479553903345694, "ret_signed_p6m": 0.059479553903345694, "alpha_spy_p6m": -0.022581147770666554, "alpha_c_p6m": -0.014240453636412465, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.07372000753975816, "price_path": [-0.0349, -0.0509, -0.0633, -0.058, -0.0562, -0.0455, -0.0367, -0.0367, -0.0384, -0.0242, -0.0065, -0.0012, -0.0207, -0.0207, 0.0059, 0.0041, 0.0166, 0.0201, 0.0148, 0.0041, 0.0041, -0.0101, -0.0083, 0.0006, 0.0272, 0.0396, 0.029, 0.013, 0.0204, 0.0446, 0.0483, 0.0409, 0.0335, 0.0297, 0.0112, 0.026, 0.0149, 0.0149, 0.0019, 0.0056, 0.013, 0.0074, 0.0167, 0.0167, -0.0242, -0.0167, 0.0, -0.0019, 0.0112, -0.0056, 0.0037, 0.0465, 0.0465, 0.0446, 0.0706, 0.0725, 0.0743, 0.0799, 0.0781, 0.052, 0.0093, 0.0056, 0.0056, 0.0, -0.0037, -0.0353, -0.0372, -0.039, -0.0037, -0.0149, 0.0, -0.0112, -0.0242, -0.0483, -0.0576, -0.0576, -0.0446, -0.039, -0.0335, -0.0242, 0.0149, 0.0037, 0.0186, 0.026, 0.0167, 0.0149, 0.039, 0.0446, 0.0446, 0.0651, 0.0539, 0.0595, 0.0595, 0.0595, 0.0781, 0.0651, 0.0929, 0.1152, 0.1152, 0.1468, 0.1375, 0.1078, 0.1078, 0.0948, 0.0558, 0.0632, 0.0836, 0.0725, 0.1022, 0.0818, 0.0762, 0.0911, 0.0911, 0.1506, 0.1301, 0.145, 0.1375, 0.1171, 0.1059, 0.0688, 0.0762, 0.0892, 0.0669, 0.0836, 0.0836, 0.0706, 0.0595, 0.039, 0.026, 0.0037, -0.0074, 0.0186, 0.0037, 0.0112, 0.0223, 0.039, 0.0465, 0.0483, 0.0669, 0.0725, 0.0874, 0.0948, 0.1059, 0.1059, 0.1004, 0.0874, 0.0706, 0.0651, 0.0595, 0.0149, -0.0149, -0.0335, -0.0149, -0.0074, -0.0167, -0.0186, -0.0186, -0.0223, -0.0223, -0.0223, 0.0112, 0.0112, 0.0297, 0.0335, 0.0483, 0.0483, 0.0149, 0.052, 0.0428, 0.0372, 0.0632, 0.052, 0.0762, 0.0483, 0.0316, 0.0223, 0.052, 0.0409, 0.0669, 0.0688, 0.0669, 0.052, 0.0409, 0.0149, 0.0279, 0.0335, 0.0149, 0.039, 0.0558, 0.0595]}
{"unit_id": "thread:1956967786225504737", "ticker_idx": 1, "ts": "2025-08-17T06:34:13+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn anticipates growth in 2026 based on confirmed orders from the three major CSPs—AWS, Google, and Microsoft", "tweet_type": "analysis", "timeline": "quarters", "summary": "Quanta and Foxconn both set for ASIC server shipment surge in 2026 from major CSP customers, with margin improvement vs GPU servers.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Quanta and Foxconn ramp up ASIC server efforts as 2026 market competition intensifies\n\n- Both Quanta and Foxconn expect a surge in ASIC server shipments in the 2026 AI server market, anticipating a margin relief effect.\n\n- While the two companies won in the Nvidia GB series AI server market in 2024–2025, the ASIC server market for CSPs is projected to expand in earnest in 2026.\n\n- Quanta’s ASIC shipments in 2025 were minimal, but the company expects strong growth in 2026, with Meta and Google as major customers.\n\n- Foxconn had 20% of its 2024 AI server revenue from ASICs, and it anticipates growth in 2026 based on confirmed orders from the three major CSPs—AWS, Google, and Microsoft.\n\n- Currently, the AI server market is led by GPU products, and according to Digitimes Research, GPU shipments in 2025 will reach 62.73 million units, while ASIC shipments will stand at 50.62 million units.\n\n- However, more and more institutions expect ASIC server growth to outpace GPU growth. Notably, Goldman Sachs projects a 40% growth rate for ASIC servers in 2026 (vs. 25.6% for GPUs).\n\n- This is because CSPs are accelerating their own ASIC development for cost efficiency, workload customization, and to avoid supply chain dependency.\n\n- Expansion of Nvidia GB200 and GB300 shipments exerts margin pressure on ODMs due to high cabinet costs.\n\n- In contrast, ASIC servers, while more complex in design, are more profitable than GPU servers, making them advantageous for ODMs to defend their margins.\n\n- As a result, the expansion of ASIC shipments in 2026 is expected to be the key breakthrough for ODMs in improving profitability and driving strategic transformation.\n\n---\n\nhttps://t.co/TnziTtbKJy", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014285714285714235, "ret_signed_m1d": -0.014285714285714235, "alpha_spy_m1d": -0.014503393096657824, "alpha_c_m1d": -0.01214507662379416, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.002140637661920075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00952380952380949, "ret_signed_p1d": -0.00952380952380949, "alpha_spy_p1d": -0.004098679457131826, "alpha_c_p1d": 0.007939614249029336, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.017463423772838826, "ret_p1w": -0.011904761904761862, "ret_signed_p1w": -0.011904761904761862, "alpha_spy_p1w": -0.010614476733080691, "alpha_c_p1w": 0.004770048283670003, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.016674810188431866, "ret_p1m": 0.026190476190476097, "ret_signed_p1m": 0.026190476190476097, "alpha_spy_p1m": 0.0002305317708135579, "alpha_c_p1m": 0.0032813263626407796, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.022909149827835318, "ret_p3m": 0.19999999999999996, "ret_signed_p3m": 0.19999999999999996, "alpha_spy_p3m": 0.15242785071692388, "alpha_c_p3m": 0.12230320736613187, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.07769679263386808, "ret_p6m": 0.05238095238095242, "ret_signed_p6m": 0.05238095238095242, "alpha_spy_p6m": -0.02967974929305983, "alpha_c_p6m": -0.021339055158805742, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.07372000753975816, "price_path": [-0.2878, -0.297, -0.2924, -0.297, -0.3016, -0.3016, -0.2832, -0.2832, -0.2993, -0.3039, -0.2832, -0.2855, -0.297, -0.2901, -0.2855, -0.2832, -0.2763, -0.2809, -0.2878, -0.2878, -0.2832, -0.2924, -0.2855, -0.2832, -0.2625, -0.2556, -0.251, -0.2418, -0.2602, -0.2395, -0.2286, -0.219, -0.2333, -0.2333, -0.2429, -0.2333, -0.231, -0.231, -0.2357, -0.2238, -0.2262, -0.219, -0.2119, -0.2143, -0.2286, -0.2095, -0.1714, -0.169, -0.1619, -0.1833, -0.1833, -0.1524, -0.1524, -0.1429, -0.1214, -0.1143, -0.0738, -0.0738, -0.0571, -0.0619, -0.0548, -0.05, -0.0143, 0.0, -0.0095, -0.0452, -0.019, -0.0357, -0.0119, -0.0048, -0.0095, -0.019, -0.031, -0.0548, -0.0548, -0.0452, -0.0333, -0.0238, -0.031, -0.0119, 0.0, 0.0214, 0.0357, 0.0286, 0.0262, 0.0095, 0.0238, 0.019, 0.0286, 0.0524, 0.0619, 0.0952, 0.0452, 0.0452, 0.0286, 0.0429, 0.0738, 0.0786, 0.0786, 0.0881, 0.0714, 0.0548, 0.0548, 0.0143, -0.019, -0.0167, 0.0619, 0.0786, 0.1357, 0.1381, 0.1548, 0.1381, 0.1381, 0.1738, 0.1905, 0.2381, 0.2476, 0.2262, 0.1976, 0.1619, 0.1762, 0.181, 0.1619, 0.1881, 0.1738, 0.1952, 0.2, 0.1476, 0.1238, 0.0952, 0.0905, 0.1262, 0.0714, 0.0476, 0.0429, 0.081, 0.1, 0.0738, 0.0571, 0.0571, 0.081, 0.0881, 0.1, 0.1048, 0.119, 0.1119, 0.0762, 0.081, 0.0548, 0.0381, 0.031, 0.0286, 0.0548, 0.0667, 0.0667, 0.0667, 0.0667, 0.0738, 0.1, 0.0857, 0.0976, 0.0976, 0.1048, 0.1167, 0.1238, 0.1357, 0.0929, 0.0976, 0.0881, 0.0786, 0.1167, 0.1119, 0.1167, 0.0929, 0.0643, 0.0429, 0.0643, 0.0548, 0.0667, 0.0738, 0.0738, 0.0667, 0.05, 0.0214, 0.031, 0.0429, 0.0262, 0.0238, 0.0405, 0.0524]}
{"unit_id": "orig:1943954147264672207", "ticker_idx": 0, "ts": "2025-07-12T08:42:40+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI continues to be the primary growth engine in the mid-to-long term", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares JPM Buy-thesis on TSMC: short-term consolidation on FX/tariff risks, but AI drives mid-to-long-term upside; 2026 revenue +14-19%.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC: AI and Advanced Processes Drive Performance Growth, Exchange Rates and Tariffs Remain Short-Term Risks\nJPM (G. Hariharan, 25/07/10)\n\nJPM states that TSMC's 2Q25 revenue slightly exceeded the upper end of its guidance, primarily driven by strong demand for N3/N5 and rush orders for mature nodes. Looking ahead to 2026, AI and N2 will continue to drive growth, though TWD appreciation and tariff issues still present uncertainties.\n\n2Q Revenue Meets Expectations and Exceeds Company Guidance; N3/N5 and Rush Orders Are Key Drivers\n2Q25 revenue was NT$934 billion, a 39% year-over-year increase, slightly above the company's dollar-denominated guidance. Growth was primarily driven by strong ramp-up of N3, AI-driven growth in N4/N5, and rush order shipments for older nodes. Even with an 11% appreciation of the New Taiwan Dollar, the gross margin is still expected to fall within the company's guidance range of 57-59%.\n\n3Q Revenue Outlook for Moderate Growth, Full-Year Guidance Potentially Revised Up to Nearly 30%\nJPM forecasts 3Q revenue to increase by 3-6% quarter-over-quarter, benefiting from continued N3/N5 demand. Although 4Q revenue is expected to decline by 6% due to diminishing rush order effects, the full-year revenue growth guidance may be revised upwards from mid-20% to high-20%, due to sustained strong AI orders and Apple not making additional cuts for tariff risks.\n\n2026 Growth Driven by Price Increases, N3 Continuation, and N2 Introduction; Gross Margin Under Pressure\nJPM estimates 2026 revenue to grow by 14% (19% in USD terms), with key drivers being 6-10% price increases for advanced processes, continued strong N3 performance, and N2 customer expansion. Gross margin may remain flat due to a stronger exchange rate and initial depreciation costs from N2 ramp-up. Capital expenditure is expected to reach $45 billion for expanding N2 capacity in Hsinchu and Kaohsiung, and the second phase of the N3 fab in Arizona.\n\nFive Key Topics for the 7/17 Earnings Call, Including CoWoS Supply/Demand, 4Q Outlook, and N2 Prospects\nInvestor focus areas include: 1) The sustainability of AI demand and whether CoWoS capacity bottlenecks will ease; 2) Whether non-AI customer demand will be weak in the second half of the year; 3) Whether exchange rate pressures can be passed on to customers; 4) Initial demand and ramp-up speed for N2; 5) Whether TSMC's expansion in the US will help mitigate tariff impacts.\n\nShort-Term Stock Price May Consolidate, AI Remains Core Mid-to-Long-Term Driver\nJPM notes that TSMC's stock price has seen significant gains recently, and in the short term, it is pressured by exchange rate and tariff investigation (Section 232) uncertainties. However, if consumer demand in the second half of the year exceeds expectations, it could provide upside for the stock price. AI continues to be the primary growth engine in the mid-to-long term.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007565506754671603, "ret_signed_m1d": 0.007565506754671603, "alpha_spy_m1d": 0.00946999000399662, "alpha_c_m1d": -3.988209007865784e-05, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.036209363418472806, "ret_signed_p1d": 0.036209363418472806, "alpha_spy_p1d": 0.04048253046909989, "alpha_c_p1d": 0.017002926760681047, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.04451836294549416, "ret_signed_p1w": 0.04451836294549416, "alpha_spy_p1w": 0.03818033096077578, "alpha_c_p1w": 0.026328302792270497, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.06830797777975661, "ret_signed_p1m": 0.06830797777975661, "alpha_spy_p1m": 0.039691077707496136, "alpha_c_p1m": 0.016506573006445757, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.3155467192590542, "ret_signed_p3m": 0.3155467192590542, "alpha_spy_p3m": 0.23838598735506844, "alpha_c_p3m": 0.105642171529432, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.4402837522732699, "ret_signed_p6m": 0.4402837522732699, "alpha_spy_p6m": 0.32670051708505543, "alpha_c_p6m": 0.07751180074562058, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.3391, -0.3388, -0.3388, -0.3557, -0.3403, -0.3124, -0.2846, -0.2806, -0.288, -0.2833, -0.2737, -0.2474, -0.2188, -0.2314, -0.2493, -0.2395, -0.2365, -0.2309, -0.1853, -0.1548, -0.1514, -0.1537, -0.1537, -0.1569, -0.1571, -0.1645, -0.1452, -0.1635, -0.1635, -0.1387, -0.1454, -0.141, -0.1577, -0.151, -0.139, -0.1181, -0.114, -0.106, -0.0981, -0.0743, -0.0671, -0.0579, -0.0768, -0.0568, -0.0646, -0.0663, -0.0663, -0.0838, -0.0802, -0.0375, -0.0259, -0.0204, -0.0004, -0.0095, -0.0174, 0.0216, 0.0268, 0.0268, 0.0022, -0.0035, 0.0139, 0.0048, 0.0076, 0.0, 0.0362, 0.0389, 0.074, 0.0513, 0.0445, 0.0259, 0.051, 0.0565, 0.074, 0.0616, 0.0554, 0.0623, 0.0566, 0.0286, 0.0452, 0.0166, 0.0118, 0.061, 0.0576, 0.0587, 0.0683, 0.0558, 0.0539, 0.0446, 0.0557, 0.0176, -0.0003, -0.0059, 0.0189, 0.0303, 0.0439, 0.0464, 0.042, 0.0096, 0.0096, -0.0012, 0.0119, 0.0286, 0.0645, 0.081, 0.0973, 0.1389, 0.1322, 0.1341, 0.143, 0.1496, 0.1528, 0.1785, 0.162, 0.196, 0.2402, 0.2314, 0.2137, 0.1992, 0.1986, 0.2252, 0.2655, 0.2639, 0.2818, 0.3266, 0.2899, 0.3359, 0.3155, 0.2312, 0.3288, 0.2983, 0.3367, 0.3154, 0.2945, 0.306, 0.292, 0.2673, 0.2754, 0.294, 0.3084, 0.3228, 0.3384, 0.3302, 0.318, 0.3374, 0.29, 0.2882, 0.2689, 0.2568, 0.2953, 0.2773, 0.2749, 0.238, 0.2495, 0.2372, 0.2192, 0.2387, 0.2174, 0.2067, 0.2487, 0.2489, 0.272, 0.272, 0.2788, 0.262, 0.2814, 0.2961, 0.2851, 0.2929, 0.3243, 0.331, 0.3606, 0.341, 0.2846, 0.2657, 0.2619, 0.2183, 0.2522, 0.271, 0.2901, 0.3062, 0.3143, 0.3143, 0.3321, 0.3237, 0.3178, 0.3367, 0.3367, 0.4059, 0.4175, 0.4403]}
{"unit_id": "quote:1945399671125668056", "ticker_idx": 0, "ts": "2025-07-16T08:26:40+00:00", "symbol": "HBM memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I anticipate a decline in HBM ASP next year. This is because Samsung will enter the supply chain, and all HBM makers will increase their supply.", "tweet_type": "prediction", "timeline": "months", "summary": "Bearish HBM ASP: Samsung entry + all-maker supply increases will drive average selling prices down next year.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM memory basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Unlike Micron's directors, I anticipate a decline in HBM ASP next year. This is because Samsung will enter the supply chain, and all HBM makers will increase their supply.\n\nPlease refer to the table below. (Source: Samsung Securities)\n\nSamsung will supply more than double the amount of HBM next year compared to this year.", "reply_text": null, "parent_text": null, "quoted_text": "Director at Micron - HBM packaging yields have matured, capacity will catch up (via Tegus): https://t.co/1Lb80QfhNQ", "parent_author": null, "ret_m1d": 0.008199044061293015, "ret_signed_m1d": -0.008199044061293015, "alpha_spy_m1d": -0.011531095495048382, "alpha_c_m1d": -0.0029446043279802714, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02861389333227719, "ret_signed_p1d": 0.02861389333227719, "alpha_spy_p1d": 0.034733583309794744, "alpha_c_p1d": 0.03701403373850913, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.04054252466373239, "ret_signed_p1w": 0.04054252466373239, "alpha_spy_p1w": 0.056546646463913795, "alpha_c_p1w": 0.03131273030891376, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.03895492576354095, "ret_signed_p1m": -0.03895492576354095, "alpha_spy_p1m": -0.005745443650970371, "alpha_c_p1m": 0.004739794176927083, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.5030786269450117, "ret_signed_p3m": -0.5030786269450117, "alpha_spy_p3m": -0.43794412609698574, "alpha_c_p3m": -0.3286109732370671, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 1.512175882780585, "ret_signed_p6m": -1.512175882780585, "alpha_spy_p6m": -1.4012458153329457, "alpha_c_p6m": -1.1980191609380504, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.3233, -0.3269, -0.3222, -0.3027, -0.2929, -0.2793, -0.285, -0.2911, -0.2962, -0.2938, -0.282, -0.2828, -0.2826, -0.2696, -0.263, -0.2601, -0.2219, -0.2083, -0.2019, -0.2082, -0.199, -0.208, -0.2061, -0.2153, -0.2273, -0.2306, -0.2247, -0.2207, -0.2048, -0.1971, -0.2118, -0.1947, -0.183, -0.1638, -0.1405, -0.134, -0.1186, -0.1108, -0.0911, -0.0978, -0.1056, -0.0851, -0.0779, -0.0678, -0.0714, -0.0524, -0.0616, -0.0107, -0.0001, -0.0014, -0.0101, -0.0106, -0.0226, -0.0244, -0.0079, -0.0195, -0.0336, -0.0099, -0.0224, 0.0012, 0.0107, -0.0006, 0.0082, 0.0, -0.0286, -0.0239, -0.0197, -0.0449, -0.0405, -0.0366, -0.0424, -0.0238, -0.0201, -0.0007, -0.0117, -0.0542, -0.0418, -0.0309, -0.043, -0.0215, -0.0009, 0.0207, 0.035, 0.0393, 0.039, 0.0263, 0.0156, 0.0062, -0.0135, -0.0289, -0.0126, -0.0061, -0.0098, -0.0064, 0.0107, 0.0032, -0.0223, -0.011, -0.0045, 0.0162, 0.0426, 0.0499, 0.0803, 0.1177, 0.1555, 0.2091, 0.2191, 0.2562, 0.237, 0.2981, 0.2805, 0.2973, 0.3199, 0.3061, 0.2946, 0.2589, 0.2986, 0.3051, 0.3725, 0.4365, 0.4486, 0.4575, 0.4424, 0.4735, 0.4615, 0.4914, 0.5031, 0.474, 0.518, 0.5961, 0.6114, 0.6475, 0.6243, 0.6218, 0.6312, 0.7139, 0.7617, 0.7382, 0.7986, 0.821, 0.8278, 0.9465, 0.8283, 0.8539, 0.8649, 0.8423, 0.9297, 0.9244, 0.9309, 0.901, 0.8417, 0.8971, 0.8035, 0.7804, 0.7414, 0.6722, 0.7284, 0.7424, 0.7826, 0.8092, 0.7957, 0.8177, 0.8509, 0.8346, 0.8049, 0.8545, 0.9252, 0.9229, 0.9768, 0.9333, 0.8988, 0.8479, 0.7962, 0.8263, 0.8918, 0.9292, 1.0188, 1.0275, 1.0598, 1.0598, 1.0973, 1.1873, 1.1968, 1.1761, 1.1761, 1.3362, 1.3982, 1.5259, 1.5437, 1.5122]}
{"unit_id": "orig:1967810540035768448", "ticker_idx": 0, "ts": "2025-09-16T04:39:27+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$NVDA", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish NVDA on Hon Hai's 2026 AI rack outlook: base 50–60K, bull case 100K, 90% GPU racks.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Hon Hai’s 2026 AI rack outlook:\n\nThe base case is 50–60K, with a bull case up to 100K(What?)…\n\nAnd 90% of these are said to be GPU racks.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01641119410161429, "ret_signed_m1d": 0.01641119410161429, "alpha_spy_m1d": 0.01503250314456972, "alpha_c_m1d": 0.016541333006460546, "bench_spy_m1d": 0.0013786909570445705, "bench_c_m1d": -0.00013013890484625534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0262466362492908, "ret_signed_p1d": -0.0262466362492908, "alpha_spy_p1d": -0.025004200956742162, "alpha_c_p1d": -0.019866641209715485, "bench_spy_p1d": -0.0012424352925486382, "bench_c_p1d": -0.006379995039575315, "ret_p1w": 0.020299523250129736, "ret_signed_p1w": 0.020299523250129736, "alpha_spy_p1w": 0.012649900821011029, "alpha_c_p1w": -0.02686696117888987, "bench_spy_p1w": 0.007649622429118708, "bench_c_p1w": 0.047166484429019606, "ret_p1m": 0.028305071049989472, "ret_signed_p1m": 0.028305071049989472, "alpha_spy_p1m": 0.017677595528911905, "alpha_c_p1m": -0.08422382780846505, "bench_spy_p1m": 0.010627475521077567, "bench_c_p1m": 0.11252889885845452, "ret_p3m": 0.0008562195203041245, "ret_signed_p3m": 0.0008562195203041245, "alpha_spy_p3m": -0.03497724826562987, "alpha_c_p3m": -0.15184066016528197, "bench_spy_p3m": 0.03583346778593399, "bench_c_p3m": 0.1526968796855861, "ret_p6m": 0.06387474412457306, "ret_signed_p6m": 0.06387474412457306, "alpha_spy_p6m": 0.033254897428516506, "alpha_c_p6m": -0.24558082385252722, "bench_spy_p6m": 0.030619846696056552, "bench_c_p6m": 0.3094555679771003, "price_path": [-0.1682, -0.1775, -0.1757, -0.1543, -0.1177, -0.1136, -0.098, -0.0966, -0.1234, -0.1009, -0.0889, -0.0889, -0.0952, -0.0851, -0.0687, -0.0617, -0.057, -0.0619, -0.024, -0.0201, -0.0108, -0.0142, -0.0201, -0.0449, -0.0235, -0.0066, -0.0079, 0.0106, 0.0035, 0.025, 0.017, -0.0067, 0.0292, 0.0193, 0.0259, 0.0336, 0.0447, 0.041, 0.0473, 0.0383, 0.0408, 0.0318, 0.0407, 0.0043, 0.0029, 0.0005, 0.0177, 0.0281, 0.0393, 0.0384, 0.0302, -0.0041, -0.0041, -0.0235, -0.0244, -0.0185, -0.045, -0.0376, -0.0236, 0.014, 0.0131, 0.0168, 0.0164, 0.0, -0.0262, 0.0078, 0.0102, 0.0499, 0.0203, 0.012, 0.0161, 0.0189, 0.0399, 0.0669, 0.0707, 0.0801, 0.0728, 0.061, 0.0581, 0.0814, 0.1012, 0.0473, 0.0769, 0.0294, 0.0283, 0.0396, 0.0477, 0.0444, 0.0359, 0.0309, 0.0416, 0.0651, 0.095, 0.1495, 0.1839, 0.1602, 0.1579, 0.183, 0.1362, 0.1163, 0.0755, 0.0759, 0.1382, 0.1045, 0.1082, 0.0685, 0.0874, 0.067, 0.0371, 0.0666, 0.0329, 0.0229, 0.0439, 0.0168, 0.0308, 0.0308, 0.0121, 0.0288, 0.0376, 0.0269, 0.0487, 0.0431, 0.0611, 0.0578, 0.051, 0.0347, 0.0009, 0.0081, 0.0163, -0.0225, -0.0042, 0.035, 0.0504, 0.082, 0.0786, 0.0786, 0.0896, 0.0763, 0.0725, 0.0665, 0.0665, 0.0799, 0.0758, 0.0707, 0.0814, 0.0582, 0.0571, 0.0576, 0.0626, 0.0473, 0.0697, 0.065, 0.065, 0.0183, 0.0483, 0.057, 0.0732, 0.0663, 0.0781, 0.0952, 0.1009, 0.093, 0.0614, 0.0313, -0.0039, -0.0171, 0.0603, 0.0867, 0.0782, 0.0868, 0.069, 0.0454, 0.0454, 0.0578, 0.075, 0.0745, 0.0855, 0.0954, 0.1028, 0.1183, 0.0573, 0.0133, 0.0435, 0.0296, 0.0467, 0.0484, 0.0169, 0.0445, 0.0566, 0.0639]}
{"unit_id": "orig:1956635675241128198", "ticker_idx": 0, "ts": "2025-08-16T08:34:32+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's GB200 ranked first with an astonishing profit margin of about 78%", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Morgan Stanley model: NVDA GB200 leads with ~78% margin; AMD suffering significant losses in AI inference — author shares this directional finding.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: AI inference is an incredibly profitable business. In its newly released report, the firm used a detailed financial model to calculate, for the first time from an economic perspective, the profitability of global AI compute competition. The conclusion is that standardized “AI inference factories,” regardless of which big tech company’s chips they use, generally enjoy an average profit margin exceeding 50%.\n\nAmong them, NVIDIA’s GB200 ranked first with an astonishing profit margin of about 78%. Google and Huawei’s chips were also assessed as “guaranteed profit businesses.” However, AMD, which had been highly anticipated by the market, was found to be suffering significant losses in the AI platform’s inference sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008570860315888718, "ret_signed_m1d": -0.008570860315888718, "alpha_spy_m1d": -0.008788539126832307, "alpha_c_m1d": -0.004494894694822116, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.004075965621066602, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03499800104375028, "ret_signed_p1d": -0.03499800104375028, "alpha_spy_p1d": -0.02957287097707262, "alpha_c_p1d": -0.014820187223993253, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.02017781381975703, "ret_p1w": -0.01208714510141684, "ret_signed_p1w": -0.01208714510141684, "alpha_spy_p1w": -0.01079685992973567, "alpha_c_p1w": -0.0021835984152863697, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.009903546686130471, "ret_p1m": -0.039119339029687894, "ret_signed_p1m": -0.039119339029687894, "alpha_spy_p1m": -0.06507928344935043, "alpha_c_p1m": -0.07398428232234966, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.034864943292661765, "ret_p3m": 0.026704848072617082, "ret_signed_p3m": 0.026704848072617082, "alpha_spy_p3m": -0.02086730121045899, "alpha_c_p3m": -0.13485285428145266, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.16155770235406974, "ret_p6m": 0.03599328672679625, "ret_signed_p6m": 0.03599328672679625, "alpha_spy_p6m": -0.046067414947216, "alpha_c_p6m": -0.3317203549813268, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.36771364170812304, "price_path": [-0.2759, -0.2703, -0.2787, -0.2787, -0.2556, -0.2594, -0.2353, -0.2576, -0.2453, -0.2242, -0.2203, -0.2309, -0.2214, -0.2164, -0.2091, -0.2153, -0.2033, -0.22, -0.205, -0.2082, -0.2007, -0.2007, -0.2097, -0.2079, -0.1874, -0.1522, -0.1483, -0.1333, -0.132, -0.1577, -0.136, -0.1246, -0.1246, -0.1306, -0.1209, -0.1051, -0.0984, -0.0939, -0.0986, -0.0621, -0.0585, -0.0495, -0.0527, -0.0584, -0.0823, -0.0617, -0.0454, -0.0468, -0.0289, -0.0357, -0.0151, -0.0227, -0.0455, -0.011, -0.0206, -0.0142, -0.0068, 0.0038, 0.0003, 0.0063, -0.0023, 0.0001, -0.0086, 0.0, -0.035, -0.0363, -0.0386, -0.0221, -0.0121, -0.0013, -0.0023, -0.0101, -0.043, -0.043, -0.0617, -0.0626, -0.0569, -0.0824, -0.0753, -0.0618, -0.0257, -0.0265, -0.023, -0.0234, -0.0391, -0.0643, -0.0316, -0.0293, 0.0088, -0.0196, -0.0276, -0.0237, -0.0209, -0.0008, 0.0252, 0.0288, 0.0379, 0.0309, 0.0195, 0.0167, 0.0391, 0.0581, 0.0064, 0.0347, -0.0108, -0.0119, -0.001, 0.0067, 0.0035, -0.0046, -0.0094, 0.0009, 0.0234, 0.0521, 0.1046, 0.1376, 0.1148, 0.1126, 0.1367, 0.0917, 0.0726, 0.0334, 0.0338, 0.0937, 0.0613, 0.0648, 0.0267, 0.0449, 0.0253, -0.0035, 0.0248, -0.0075, -0.0171, 0.003, -0.023, -0.0096, -0.0096, -0.0275, -0.0114, -0.003, -0.0132, 0.0076, 0.0023, 0.0196, 0.0164, 0.0098, -0.0058, -0.0383, -0.0313, -0.0235, -0.0607, -0.0431, -0.0055, 0.0093, 0.0397, 0.0364, 0.0364, 0.0469, 0.0342, 0.0305, 0.0248, 0.0248, 0.0377, 0.0337, 0.0289, 0.0391, 0.0168, 0.0158, 0.0162, 0.021, 0.0063, 0.0278, 0.0233, 0.0233, -0.0215, 0.0073, 0.0157, 0.0312, 0.0246, 0.0359, 0.0524, 0.0578, 0.0502, 0.0199, -0.0091, -0.0429, -0.0556, 0.0188, 0.0442, 0.036]}
{"unit_id": "orig:1956635675241128198", "ticker_idx": 1, "ts": "2025-08-16T08:34:32+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD, which had been highly anticipated by the market, was found to be suffering significant losses in the AI platform's inference sector", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Morgan Stanley model: NVDA GB200 leads with ~78% margin; AMD suffering significant losses in AI inference — author shares this directional finding.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: AI inference is an incredibly profitable business. In its newly released report, the firm used a detailed financial model to calculate, for the first time from an economic perspective, the profitability of global AI compute competition. The conclusion is that standardized “AI inference factories,” regardless of which big tech company’s chips they use, generally enjoy an average profit margin exceeding 50%.\n\nAmong them, NVIDIA’s GB200 ranked first with an astonishing profit margin of about 78%. Google and Huawei’s chips were also assessed as “guaranteed profit businesses.” However, AMD, which had been highly anticipated by the market, was found to be suffering significant losses in the AI platform’s inference sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007777876245797399, "ret_signed_m1d": -0.007777876245797399, "alpha_spy_m1d": -0.00756019743485381, "alpha_c_m1d": -0.011853841866864001, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.004075965621066602, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05444530697809358, "ret_signed_p1d": 0.05444530697809358, "alpha_spy_p1d": 0.049020176911415914, "alpha_c_p1d": 0.03426749315833655, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.02017781381975703, "ret_p1w": -0.07255591474725498, "ret_signed_p1w": 0.07255591474725498, "alpha_spy_p1w": 0.07126562957557381, "alpha_c_p1w": 0.06265236806112451, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.009903546686130471, "ret_p1m": -0.08902005637626198, "ret_signed_p1m": 0.08902005637626198, "alpha_spy_p1m": 0.11498000079592452, "alpha_c_p1m": 0.12388499966892375, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.034864943292661765, "ret_p3m": 0.40774388312186827, "ret_signed_p3m": -0.40774388312186827, "alpha_spy_p3m": -0.3601717338387922, "alpha_c_p3m": -0.24618618076779852, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.16155770235406974, "ret_p6m": 0.21250146510883905, "ret_signed_p6m": -0.21250146510883905, "alpha_spy_p6m": -0.1304407634348268, "alpha_c_p6m": 0.155212176599284, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.36771364170812304, "price_path": [-0.3638, -0.3715, -0.3737, -0.3737, -0.3496, -0.3593, -0.3583, -0.3714, -0.3492, -0.334, -0.3268, -0.3432, -0.3404, -0.3089, -0.3003, -0.3123, -0.3272, -0.3405, -0.2824, -0.2784, -0.2802, -0.2802, -0.2719, -0.2643, -0.2141, -0.1859, -0.1843, -0.1835, -0.1944, -0.2273, -0.2136, -0.217, -0.217, -0.2347, -0.2176, -0.2142, -0.1816, -0.1687, -0.1698, -0.1166, -0.0912, -0.0893, -0.1087, -0.1087, -0.1216, -0.0993, -0.0796, -0.0549, -0.0141, 0.0074, 0.0191, 0.001, -0.0252, 0.0036, -0.0104, -0.0739, -0.0212, -0.0192, -0.0219, -0.0068, 0.047, 0.0273, 0.0078, 0.0, -0.0544, -0.0621, -0.0706, -0.0476, -0.0726, -0.054, -0.0512, -0.0429, -0.0767, -0.0767, -0.0785, -0.0795, -0.0815, -0.1419, -0.1404, -0.1154, -0.0942, -0.1162, -0.0998, -0.085, -0.089, -0.0964, -0.1034, -0.1064, -0.0928, -0.0865, -0.0866, -0.0844, -0.0947, -0.0839, -0.0815, -0.0689, -0.0364, -0.0651, 0.1565, 0.2008, 0.3373, 0.3222, 0.2201, 0.2287, 0.2382, 0.3546, 0.3317, 0.3233, 0.3657, 0.3514, 0.3071, 0.3341, 0.4359, 0.4742, 0.4648, 0.5007, 0.4468, 0.4541, 0.4741, 0.4196, 0.4553, 0.3495, 0.3259, 0.3851, 0.3485, 0.4698, 0.4077, 0.4012, 0.3655, 0.3074, 0.2692, 0.1696, 0.1569, 0.2209, 0.1703, 0.2163, 0.2163, 0.235, 0.2476, 0.222, 0.2354, 0.2262, 0.2375, 0.2553, 0.2582, 0.2571, 0.2571, 0.1967, 0.1785, 0.1875, 0.1247, 0.1415, 0.2117, 0.2203, 0.2201, 0.2208, 0.2208, 0.2206, 0.2241, 0.2226, 0.2159, 0.2159, 0.2687, 0.2551, 0.2169, 0.1923, 0.162, 0.1535, 0.1791, 0.2545, 0.2694, 0.294, 0.3162, 0.3162, 0.3167, 0.4182, 0.4405, 0.4743, 0.4268, 0.4309, 0.4349, 0.4317, 0.344, 0.3981, 0.3745, 0.1365, 0.0929, 0.1834, 0.2263, 0.2125]}
{"unit_id": "orig:1973679502770524636", "ticker_idx": 0, "ts": "2025-10-02T09:20:37+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We raise our target price to KRW 120,000… We recommend Buy.", "tweet_type": "news_reaction", "timeline": "months", "summary": "KIS raises Samsung Electronics TP to KRW 120,000 (+26%), Buy; 3Q beat expected, HBM3e certified, 2026 OP estimate lifted 36% to KRW 73T.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "KIS: Samsung Electronics (005930) – Target Price KRW 120,000\n\n3Q25 Preview: Normalization of the Abnormal\n\n3Q Earnings Expected to Beat Consensus\nWe project 3Q revenue of KRW 82.4 trillion (+11% QoQ, +4% YoY) and operating profit of KRW 10.5 trillion (+125% QoQ, +15% YoY), which would surpass the consensus estimate of KRW 9.8 trillion by 7%. Conventional DRAM demand growth is expected to exceed guidance, driving stronger bit growth. HBM revenue is estimated to have surged 98% QoQ, boosting ASPs along with conventional DRAM. In foundry, new customer wins in mature processes (7nm and above) are gradually raising utilization. In addition, one-off costs recognized since 2H24 are coming to an end, significantly narrowing foundry losses compared to the previous quarter.\n\nSigns of Rebound Led by HBM\nSamsung’s HBM3e 12-high is believed to have essentially completed NVIDIA certification during 3Q. HBM4 samples of 11Gbps and above have also reportedly been shipped to NVIDIA. In 2H25, sales will mainly focus on HBM3e for customers other than NVIDIA, but starting in 2026, sales are expected to expand to NVIDIA and a broader customer base, with HBM bit growth outpacing the market average. Reflecting recent supply-demand conditions, we revise up our 2026 conventional DRAM ASP forecast from +10% to +25%, and HBM ASP from –2% to +9%. Accordingly, we raise our 2026 operating profit estimate by 36% to KRW 73 trillion (from the previous KRW 54 trillion).\n\nTarget Price Raised, Maintain “Second Preferred Pick” in the Sector\nWe raise our target price to KRW 120,000 (12-month forward BPS of KRW 71,381, target PBR 1.7x), up 26% from the previous estimate, while maintaining our stance as the second preferred pick within the sector. We believe much of the HBM risk has been alleviated with the certification of HBM3e 12-high for NVIDIA, and thus raise the target PBR from 1.3x to 1.7x. As HBM revenue grows, Samsung will gradually narrow the profitability gap with competitors. Investors can expect to benefit from both the conventional DRAM upcycle and the normalization of the HBM business. We recommend Buy.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04178270636615855, "ret_signed_m1d": -0.04178270636615855, "alpha_spy_m1d": -0.0406323196768168, "alpha_c_m1d": -0.03640107221397715, "bench_spy_m1d": -0.0011503866893417536, "bench_c_m1d": -0.005381634152181403, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 1.4767362221612679e-05, "alpha_c_p1d": 0.0050321551678135235, "bench_spy_p1d": -1.4767362221612679e-05, "bench_c_p1d": -0.0050321551678135235, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.0028989891327806383, "alpha_c_p1w": -0.014117967123095987, "bench_spy_p1w": 0.0028989891327806383, "bench_c_p1w": 0.014117967123095987, "ret_p1m": 0.19777161704230184, "ret_signed_p1m": 0.19777161704230184, "alpha_spy_p1m": 0.1785849611610073, "alpha_c_p1m": 0.1470307547467531, "bench_spy_p1m": 0.019186655881294534, "bench_c_p1m": 0.05074086229554875, "ret_p3m": 0.3424273572416967, "ret_signed_p3m": 0.3424273572416967, "alpha_spy_p3m": 0.31281054094006966, "alpha_c_p3m": 0.32460125554609265, "bench_spy_p3m": 0.029616816301627047, "bench_c_p3m": 0.017826101695604057, "ret_p6m": 1.0164401005503554, "ret_signed_p6m": 1.0164401005503554, "alpha_spy_p6m": 1.0635388137627695, "alpha_c_p6m": 1.1059086454633775, "bench_spy_p6m": -0.04709871321241399, "bench_c_p6m": -0.08946854491302214, "price_path": [-0.3156, -0.3189, -0.33, -0.3234, -0.3056, -0.3067, -0.2934, -0.2823, -0.2601, -0.2557, -0.2479, -0.2679, -0.2635, -0.2679, -0.269, -0.2191, -0.2169, -0.1947, -0.208, -0.2357, -0.2268, -0.2246, -0.2368, -0.218, -0.2036, -0.2124, -0.2113, -0.2024, -0.2058, -0.2058, -0.2235, -0.2235, -0.218, -0.2169, -0.208, -0.2069, -0.2202, -0.2169, -0.228, -0.2268, -0.2501, -0.2335, -0.2257, -0.2224, -0.2291, -0.2224, -0.2069, -0.1947, -0.1858, -0.1636, -0.1514, -0.1192, -0.1326, -0.1093, -0.1093, -0.0738, -0.0605, -0.0527, -0.0449, -0.076, -0.0618, -0.0652, -0.0418, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0518, 0.0396, 0.0206, 0.0585, 0.0886, 0.0908, 0.093, 0.0864, 0.0986, 0.0752, 0.1008, 0.1365, 0.1086, 0.1198, 0.1599, 0.1978, 0.2379, 0.1688, 0.1209, 0.1053, 0.0908, 0.1209, 0.1532, 0.1487, 0.1454, 0.083, 0.1209, 0.0897, 0.0752, 0.1209, 0.0563, 0.0774, 0.1064, 0.1454, 0.1532, 0.1198, 0.1231, 0.1521, 0.1643, 0.171, 0.2078, 0.2201, 0.2078, 0.2033, 0.1955, 0.2134, 0.1677, 0.1454, 0.2022, 0.1989, 0.1844, 0.2312, 0.2423, 0.2379, 0.2379, 0.3036, 0.3379, 0.3424, 0.3424, 0.3424, 0.4387, 0.5462, 0.5552, 0.5787, 0.554, 0.5563, 0.554, 0.5406, 0.5708, 0.6111, 0.6671, 0.6716, 0.6257, 0.6738, 0.7052, 0.7029, 0.7029, 0.7858, 0.8183, 0.7992, 0.797, 0.6839, 0.8754, 0.8933, 0.7836, 0.7757, 0.8631, 0.8563, 0.8787, 0.9996, 1.0288, 1.0288, 1.0288, 1.0288, 1.1273, 1.1284, 1.1609, 1.2392, 1.2784, 1.4408, 1.424, 1.424, 1.1844, 0.928, 1.1452, 1.1071, 0.9425, 1.1038, 1.1273, 1.1038, 1.0545, 1.1127, 1.1709, 1.3344, 1.2448, 1.2325, 1.0859, 1.1239, 1.1161, 1.0164, 1.0164]}
{"unit_id": "thread:1952187102768308540", "ticker_idx": 0, "ts": "2025-08-04T01:57:29+00:00", "symbol": "Doosan Enerbility", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If Doosan successfully acquires Westinghouse, it could become the ultimate all-in-one provider for the entire nuclear power value chain — from A to Z", "tweet_type": "prediction", "timeline": "years", "summary": "Long-term bullish on nuclear stocks; especially Doosan Enerbility on potential Westinghouse acquisition; also likes Hitachi and GE Vernova.", "resolved_tickers": ["034020.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Enerbility 034020.KS Korea nuclear/power", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "self_thread", "lang": "en", "tweet_text": "I have a bullish long-term outlook on nuclear-related stocks.\n\nIn particular, I heard that Korea’s Doosan Enerbility is considering acquiring Westinghouse in the U.S. If Doosan successfully acquires Westinghouse, it could become the ultimate all-in-one provider for the entire nuclear power value chain — from A to Z.\n\nHitachi and GE Vernova also look promising.\n\n---\n\nhttps://t.co/UCnwMvCoFK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04510108864696738, "ret_signed_m1d": -0.04510108864696738, "alpha_spy_m1d": -0.030128849707872662, "alpha_c_m1d": -0.03596943860931634, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.00913165003765104, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007776049766718529, "ret_signed_p1d": 0.007776049766718529, "alpha_spy_p1d": 0.012846016339149835, "alpha_c_p1d": 0.00982729665977733, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.0020512468930588, "ret_p1w": 0.07931570762052886, "ret_signed_p1w": 0.07931570762052886, "alpha_spy_p1w": 0.07178994551399431, "alpha_c_p1w": 0.08447708390456032, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": -0.00516137628403146, "ret_p1m": -0.06065318818040433, "ret_signed_p1m": -0.06065318818040433, "alpha_spy_p1m": -0.07507091781098862, "alpha_c_p1m": -0.05694755853504074, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": -0.003705629645363584, "ret_p3m": 0.3950233281493001, "ret_signed_p3m": 0.3950233281493001, "alpha_spy_p3m": 0.31494217825732007, "alpha_c_p3m": 0.3671705049608447, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.02785282318845539, "ret_p6m": 0.4556765163297045, "ret_signed_p6m": 0.4556765163297045, "alpha_spy_p6m": 0.3474504594828458, "alpha_c_p6m": 0.3551628680290737, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.10051364830063081, "price_path": [-0.5754, -0.5684, -0.5669, -0.5521, -0.5389, -0.5132, -0.5031, -0.4666, -0.4619, -0.4215, -0.3919, -0.4051, -0.3655, -0.3546, -0.3414, -0.381, -0.3437, -0.3725, -0.3631, -0.3631, -0.3367, -0.2862, -0.2862, -0.2613, -0.2535, -0.2053, -0.1509, -0.1509, -0.0731, -0.0762, -0.0513, -0.0513, -0.0638, 0.0669, 0.0715, 0.0233, 0.0062, 0.0233, 0.0638, -0.028, -0.042, -0.0373, -0.0669, -0.0218, 0.0358, 0.0016, -0.0187, -0.0591, -0.0638, 0.0156, -0.0109, -0.0156, 0.0078, 0.0638, -0.0264, -0.014, 0.0171, 0.0249, -0.0124, 0.0124, -0.0093, 0.0202, -0.0451, 0.0, 0.0078, 0.0311, 0.0171, 0.0327, 0.0793, 0.0187, 0.0342, 0.0187, 0.0187, 0.0124, -0.0747, -0.1073, -0.0435, -0.0327, 0.0249, -0.0156, -0.0187, -0.0218, -0.0404, -0.0731, -0.0607, -0.0295, -0.0233, -0.0389, -0.0529, -0.0327, -0.0264, -0.0513, -0.0513, -0.0855, -0.0156, -0.0467, -0.0544, -0.0544, -0.0295, -0.0187, 0.0249, -0.0047, -0.0187, -0.0218, -0.0249, 0.0047, 0.014, 0.014, 0.014, 0.014, 0.014, 0.014, 0.1586, 0.2068, 0.1788, 0.2893, 0.3017, 0.2473, 0.2488, 0.2442, 0.2286, 0.1866, 0.2582, 0.2737, 0.3437, 0.4992, 0.395, 0.3795, 0.3919, 0.3919, 0.3002, 0.2333, 0.2115, 0.2379, 0.2162, 0.2193, 0.2924, 0.2193, 0.2255, 0.1726, 0.1571, 0.2084, 0.1369, 0.1275, 0.1431, 0.2084, 0.2068, 0.1882, 0.1649, 0.1664, 0.2193, 0.2317, 0.2504, 0.1944, 0.1928, 0.1944, 0.2022, 0.2395, 0.1991, 0.2022, 0.1742, 0.1586, 0.2037, 0.2131, 0.2068, 0.1804, 0.1804, 0.1446, 0.1897, 0.1711, 0.1711, 0.1711, 0.1695, 0.2939, 0.3359, 0.3064, 0.3017, 0.3095, 0.3701, 0.353, 0.3935, 0.3919, 0.4821, 0.4837, 0.4821, 0.4199, 0.3997, 0.451, 0.4277, 0.4557]}
{"unit_id": "thread:1952187102768308540", "ticker_idx": 1, "ts": "2025-08-04T01:57:29+00:00", "symbol": "Hitachi", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hitachi and GE Vernova also look promising", "tweet_type": "prediction", "timeline": "years", "summary": "Long-term bullish on nuclear stocks; especially Doosan Enerbility on potential Westinghouse acquisition; also likes Hitachi and GE Vernova.", "resolved_tickers": ["6501.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Hitachi Ltd 6501.T Japan", "bench_c": "XLI", "bench_c_reason": "industry='Conglomerates'", "bench_c_industry": "Conglomerates", "unit_kind": "self_thread", "lang": "en", "tweet_text": "I have a bullish long-term outlook on nuclear-related stocks.\n\nIn particular, I heard that Korea’s Doosan Enerbility is considering acquiring Westinghouse in the U.S. If Doosan successfully acquires Westinghouse, it could become the ultimate all-in-one provider for the entire nuclear power value chain — from A to Z.\n\nHitachi and GE Vernova also look promising.\n\n---\n\nhttps://t.co/UCnwMvCoFK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03329323120100813, "ret_signed_m1d": 0.03329323120100813, "alpha_spy_m1d": 0.04826547014010285, "alpha_c_m1d": 0.04242488123865917, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.00913165003765104, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004583798857714871, "ret_signed_p1d": -0.004583798857714871, "alpha_spy_p1d": 0.0004861677147164345, "alpha_c_p1d": -0.0025325519646560712, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.0020512468930588, "ret_p1w": 0.020024132783940107, "ret_signed_p1w": 0.020024132783940107, "alpha_spy_p1w": 0.012498370677405557, "alpha_c_p1w": 0.025185509067971568, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": -0.00516137628403146, "ret_p1m": -0.042943305833877066, "ret_signed_p1m": -0.042943305833877066, "alpha_spy_p1m": -0.05736103546446136, "alpha_c_p1m": -0.03923767618851348, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": -0.003705629645363584, "ret_p3m": 0.20437869998778457, "ret_signed_p3m": 0.20437869998778457, "alpha_spy_p3m": 0.12429755009580457, "alpha_c_p3m": 0.17652587679932918, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.02785282318845539, "ret_p6m": 0.2361686673042489, "ret_signed_p6m": 0.2361686673042489, "alpha_spy_p6m": 0.1279426104573902, "alpha_c_p6m": 0.1356550190036181, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.10051364830063081, "price_path": [-0.0828, -0.0789, -0.0615, -0.0743, -0.0673, -0.0661, -0.0666, -0.0794, -0.0823, -0.0712, -0.0796, -0.0996, -0.0832, -0.0697, -0.0519, -0.0425, -0.0152, -0.0239, -0.0403, -0.0519, -0.0372, -0.0314, -0.0302, -0.008, -0.0135, -0.0046, -0.0176, -0.0381, -0.0316, -0.0304, -0.0171, -0.0193, -0.0374, -0.0531, -0.0343, -0.021, -0.0121, 0.0326, 0.0145, 0.0109, 0.0007, -0.0162, -0.0287, -0.0405, -0.0063, -0.006, -0.0072, -0.0253, -0.0135, 0.0089, 0.014, 0.022, 0.0478, 0.0478, 0.0639, 0.0736, 0.1153, 0.1131, 0.0922, 0.0753, 0.1076, 0.1332, 0.0333, 0.0, -0.0046, 0.0014, 0.015, 0.02, 0.02, 0.0357, 0.04, 0.013, 0.0311, 0.0294, 0.0294, 0.0, -0.0094, -0.0051, -0.0075, -0.0236, -0.0248, -0.0075, -0.0232, -0.0393, -0.0429, -0.0762, -0.0753, -0.0601, -0.0516, -0.0408, -0.0063, -0.007, -0.013, -0.013, -0.0224, -0.0446, -0.0456, -0.068, -0.0719, -0.0719, -0.0277, -0.0277, -0.0497, -0.0492, -0.0463, -0.0596, -0.0536, 0.0435, 0.0862, 0.0855, 0.0864, 0.1085, 0.0712, 0.0712, 0.0343, 0.0932, 0.1168, 0.0896, 0.1105, 0.0915, 0.0847, 0.0748, 0.1088, 0.1498, 0.1352, 0.177, 0.2044, 0.2905, 0.2905, 0.267, 0.2129, 0.2621, 0.2468, 0.241, 0.2425, 0.2711, 0.298, 0.2335, 0.243, 0.1575, 0.1406, 0.183, 0.143, 0.143, 0.143, 0.1777, 0.2032, 0.2056, 0.1804, 0.1874, 0.1835, 0.2029, 0.185, 0.1918, 0.1915, 0.1859, 0.1663, 0.2158, 0.1925, 0.1791, 0.2034, 0.1677, 0.1918, 0.2143, 0.216, 0.2104, 0.2037, 0.2066, 0.1961, 0.1896, 0.1896, 0.1896, 0.1896, 0.2299, 0.3213, 0.2772, 0.2347, 0.2461, 0.2461, 0.2934, 0.3099, 0.3002, 0.2629, 0.2665, 0.2612, 0.2573, 0.2774, 0.2859, 0.2393, 0.2362]}
{"unit_id": "thread:1952187102768308540", "ticker_idx": 2, "ts": "2025-08-04T01:57:29+00:00", "symbol": "GEV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hitachi and GE Vernova also look promising", "tweet_type": "prediction", "timeline": "years", "summary": "Long-term bullish on nuclear stocks; especially Doosan Enerbility on potential Westinghouse acquisition; also likes Hitachi and GE Vernova.", "resolved_tickers": ["GEV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "self_thread", "lang": "en", "tweet_text": "I have a bullish long-term outlook on nuclear-related stocks.\n\nIn particular, I heard that Korea’s Doosan Enerbility is considering acquiring Westinghouse in the U.S. If Doosan successfully acquires Westinghouse, it could become the ultimate all-in-one provider for the entire nuclear power value chain — from A to Z.\n\nHitachi and GE Vernova also look promising.\n\n---\n\nhttps://t.co/UCnwMvCoFK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009460266548386054, "ret_signed_m1d": -0.009460266548386054, "alpha_spy_m1d": 0.005511972390708664, "alpha_c_m1d": -0.00032861651073501363, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.00913165003765104, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019690107472879026, "ret_signed_p1d": -0.019690107472879026, "alpha_spy_p1d": -0.01462014090044772, "alpha_c_p1d": -0.017638860579820226, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.0020512468930588, "ret_p1w": -0.018120884248522873, "ret_signed_p1w": -0.018120884248522873, "alpha_spy_p1w": -0.025646646355057423, "alpha_c_p1w": -0.012959507964491412, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": -0.00516137628403146, "ret_p1m": -0.12536783570260512, "ret_signed_p1m": -0.12536783570260512, "alpha_spy_p1m": -0.13978556533318942, "alpha_c_p1m": -0.12166220605724154, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": -0.003705629645363584, "ret_p3m": -0.1334711531382896, "ret_signed_p3m": -0.1334711531382896, "alpha_spy_p3m": -0.21355230303026962, "alpha_c_p3m": -0.161323976326745, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.02785282318845539, "ret_p6m": 0.04636450370940026, "ret_signed_p6m": 0.04636450370940026, "alpha_spy_p6m": -0.06186155313745845, "alpha_c_p6m": -0.05414914459123055, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.10051364830063081, "price_path": [-0.3865, -0.4058, -0.3979, -0.3717, -0.3465, -0.3472, -0.3561, -0.3544, -0.3265, -0.3213, -0.3168, -0.308, -0.2996, -0.2996, -0.2889, -0.2685, -0.2894, -0.2867, -0.2683, -0.2565, -0.2638, -0.2696, -0.2685, -0.2761, -0.2982, -0.2708, -0.2645, -0.2784, -0.2642, -0.263, -0.2607, -0.2607, -0.2656, -0.2461, -0.2296, -0.2419, -0.2356, -0.2163, -0.202, -0.2369, -0.2383, -0.2202, -0.2202, -0.2003, -0.2007, -0.192, -0.1866, -0.1869, -0.1629, -0.156, -0.1537, -0.1401, -0.1334, -0.1461, -0.1717, -0.0509, -0.0585, -0.0274, -0.0228, -0.0454, -0.0117, -0.0037, -0.0095, 0.0, -0.0197, 0.0027, -0.0255, -0.0206, -0.0181, -0.008, -0.0429, -0.0566, -0.0617, -0.057, -0.09, -0.0878, -0.0857, -0.084, -0.0912, -0.0556, -0.0609, -0.0439, -0.0751, -0.0751, -0.1254, -0.1294, -0.0965, -0.1217, -0.0944, -0.0861, -0.029, -0.0432, -0.0562, -0.0514, -0.0677, -0.0724, -0.0781, -0.0582, -0.0278, -0.0443, -0.051, -0.0834, -0.0869, -0.091, -0.0722, -0.0854, -0.0853, -0.1023, -0.0899, -0.0855, -0.0563, -0.043, -0.0878, -0.0219, -0.0277, -0.0706, -0.0917, -0.0947, -0.1033, -0.1165, -0.1306, -0.1017, -0.1179, -0.1179, -0.1381, -0.1276, -0.1335, -0.1168, -0.1226, -0.1729, -0.1552, -0.1695, -0.1319, -0.1248, -0.1304, -0.1315, -0.1575, -0.1271, -0.129, -0.1624, -0.1013, -0.1577, -0.161, -0.1238, -0.1358, -0.1098, -0.1098, -0.0947, -0.1292, -0.0919, -0.0914, -0.0504, -0.0471, -0.0613, -0.0561, 0.0913, 0.063, 0.0139, 0.0285, 0.0358, -0.0729, -0.0348, -0.0064, -0.001, -0.0016, 0.0073, 0.0073, 0.0015, 0.0015, -0.0043, -0.0135, -0.0135, 0.0257, 0.0285, 0.0367, 0.0005, -0.0508, -0.0597, -0.0336, -0.015, -0.0269, -0.0299, 0.0295, 0.0295, 0.0345, 0.0089, -0.0005, -0.0064, 0.006, 0.0464]}
{"unit_id": "quote:1968166465909166258", "ticker_idx": 0, "ts": "2025-09-17T04:13:46+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I bought a bit of LRCX yesterday. It looks promising going forward.", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Bought LRCX; sees it as promising going forward on stronger memory prices lifting equipment bull case.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "I bought a bit of LRCX yesterday. It looks promising going forward. https://t.co/97G8huBwSk", "reply_text": null, "parent_text": null, "quoted_text": "Memory Semiconductor Equipment: Stronger Memory Prices Push 2026 Bull Case Expectations Higher\n\nMorgan Stanley says that strengthening memory prices are pushing up the 2026 semiconductor equipment bull case scenario to USD 50 billion (vs. previous estimate of USD 43.8 billion), https://t.co/7R9nVPx7IK", "parent_author": null, "ret_m1d": -0.01173095054577944, "ret_signed_m1d": -0.01173095054577944, "alpha_spy_m1d": -0.012974931404049705, "alpha_c_m1d": -0.018151911283007527, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03625926738363949, "ret_signed_p1d": 0.03625926738363949, "alpha_spy_p1d": 0.031586822241309154, "alpha_c_p1d": -0.002135522518084576, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.054744855143045834, "ret_signed_p1w": 0.054744855143045834, "alpha_spy_p1w": 0.049051695786501925, "alpha_c_p1w": 0.0027874830375100856, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.1702308192326889, "ret_signed_p1m": 0.1702308192326889, "alpha_spy_p1m": 0.16523734074771235, "alpha_c_p1m": 0.04557897729065563, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.35271098642796184, "ret_signed_p3m": 0.35271098642796184, "alpha_spy_p3m": 0.31715584339537517, "alpha_c_p3m": 0.1966094139815513, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.7268346853718224, "ret_signed_p6m": 0.7268346853718224, "alpha_spy_p6m": 0.7106021451782238, "alpha_c_p6m": 0.45136307429996236, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.2577, -0.2485, -0.2155, -0.2123, -0.2056, -0.2026, -0.2015, -0.2058, -0.1893, -0.1894, -0.1894, -0.1949, -0.181, -0.1812, -0.171, -0.1655, -0.1828, -0.1709, -0.1766, -0.1732, -0.1742, -0.1654, -0.1986, -0.2034, -0.1979, -0.2046, -0.191, -0.1884, -0.1871, -0.222, -0.2094, -0.1927, -0.2069, -0.213, -0.1866, -0.1653, -0.1632, -0.1363, -0.1244, -0.1191, -0.1837, -0.1888, -0.1769, -0.1866, -0.1927, -0.179, -0.1692, -0.1499, -0.1495, -0.1461, -0.1784, -0.1784, -0.204, -0.1982, -0.1762, -0.1555, -0.1381, -0.134, -0.1193, -0.0518, -0.0405, -0.0221, -0.0117, 0.0, 0.0363, 0.0412, 0.0845, 0.0823, 0.0547, 0.0532, 0.0548, 0.0775, 0.1006, 0.1737, 0.2082, 0.1985, 0.226, 0.1536, 0.1716, 0.159, 0.0798, 0.1327, 0.1369, 0.19, 0.1702, 0.1632, 0.184, 0.1922, 0.161, 0.2127, 0.2468, 0.2897, 0.2791, 0.3207, 0.3234, 0.2943, 0.3253, 0.2805, 0.3567, 0.3331, 0.3098, 0.3675, 0.3084, 0.3268, 0.2602, 0.2186, 0.2121, 0.1774, 0.2231, 0.1474, 0.1725, 0.2361, 0.2488, 0.2752, 0.2752, 0.2823, 0.2723, 0.3003, 0.3152, 0.2933, 0.3066, 0.3399, 0.3651, 0.3853, 0.389, 0.3216, 0.3527, 0.3441, 0.276, 0.356, 0.4183, 0.4429, 0.4421, 0.46, 0.46, 0.4661, 0.448, 0.4308, 0.4094, 0.4094, 0.5236, 0.6035, 0.7039, 0.672, 0.6545, 0.7978, 0.8146, 0.765, 0.719, 0.7905, 0.8357, 0.8357, 0.8311, 0.8804, 0.8171, 0.7943, 0.8349, 0.9633, 0.9725, 1.0432, 0.9221, 0.9554, 0.8945, 0.7272, 0.7562, 0.9019, 0.8877, 0.8657, 0.9358, 0.9043, 0.9392, 0.9392, 0.9396, 0.9767, 0.9545, 1.0165, 0.9947, 1.011, 1.054, 0.9683, 0.9257, 0.9019, 0.7888, 0.8381, 0.7696, 0.6431, 0.7405, 0.7741, 0.8042, 0.7268]}
{"unit_id": "orig:1960130067440508968", "ticker_idx": 0, "ts": "2025-08-26T00:00:00+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We continue to like NVIDIA at this share price range, because the ramp-up over the next 12 months will be very strong", "tweet_type": "analysis", "timeline": "months", "summary": "Morgan Stanley (shared/adopted): most bullish on NVDA into 12-month ramp; sees near-term sentiment recovery potential in MRVL and MU despite EW ratings.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: Latest Commentary on Investor Feedback Regarding “2Q Outlook”\n\nMorgan Stanley (J. Moore, 25/08/25)\n\nRegarding our feedback on NVIDIA’s outlook: Although for most of this year we have enjoyed significant optimism around NVIDIA’s performance, earlier on market sentiment was still relatively cautious, with a long list of concerns—DeepSeek, Blackwell delays, cabinet challenges, some customers’ data center construction slowdowns, and macroeconomic worries.\n\nFrom our investor interactions, it appears that these pessimistic concerns have largely passed, and we share in the current optimism. We had previously forecast October-quarter revenue of $52.5 billion and believed there could be upside, but in some investor conversations the optimism was noticeably higher. Some sell-side analysts have even suggested figures as high as $55 billion, and the market has shown some emotional buy-in to these numbers, though they are not yet locked in. We continue to like NVIDIA at this share price range, because the ramp-up over the next 12 months will be very strong, although we expect the company will remain cautious regarding supply-side issues and the China variable. It is worth noting that the upper bound of this expectation is based on several billion dollars of contribution from China.\n\nFrom this perspective, the China factor has actually clarified market expectations about the outlook. One major variable in guidance is how management assesses the China opportunity; while a small number of licenses have been approved, there is no clarity on future approvals. Media reports last week intensified market concerns, one claiming that the Chinese government might block domestic customers from purchasing U.S. chips (link), and another suggesting the company would slow its build-out pace (link). NVIDIA’s statement in response was only that “we constantly manage the supply chain to respond to market conditions”—a vague expression, but it did not deny the reports.\n\nThis remains an important issue requiring a long-term solution, and we believe ultimately both sides will have incentives to ensure China can obtain U.S. hardware. But how the company guides this part of the business will be one of the key factors affecting the specific guidance number. Against this backdrop, they will clearly take a conservative stance on this portion, which will in some sense lower the market’s hurdle.\n\nWe want to make clear that we remain very optimistic about the next 12 months, with many strong supports. It is only necessary to understand the factors that may affect near-term guidance.\n\nMeanwhile, are there any other AI stocks where market sentiment has not turned into “one-sided” optimism?\n\nYes, we can name a few. We rate them all Equal Weight (EW), but we believe both companies could see a positive sentiment shift.\n\nMarvell (outlook as follows): Market sentiment has been weighed down by Amazon’s Trainium performing only in line with expectations (rather than clearly above) and by concerns about growth next year. We think that, in the long term, expectations for ASIC may be too high, since commercial ASICs usually take longer to reach mass production. However, we still expect Amazon-related business to grow in line with market expectations next year and for the optical business to remain strong. Although reservations about ASIC keep us at EW, there is potential for a near-term positive sentiment shift.\n\nMicron: High Bandwidth Memory (HBM) pricing will face a hard reset next year, with at least one customer—NVIDIA—already committed to full-year 2025 pricing. But current market sentiment is quite negative. We expect HBM will maintain a significant but gradually narrowing premium relative to DDR5. We maintain EW on Micron because the stock is expensive relative to its own history, and even in a strong cycle its free cash flow is quite limited. Still, we believe there is short-term potential for sentiment recovery.\n\nWe continue to prefer NVDA the most. Despite higher expectations, in a very strong AI environment, its positive sentiment still has the chance to spread to these peers.\n\n$NVDA $MRVL $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01078283957708237, "ret_signed_m1d": -0.01078283957708237, "alpha_spy_m1d": -0.006613261402572013, "alpha_c_m1d": -0.0012462736630559634, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0009352252999643706, "ret_signed_p1d": -0.0009352252999643706, "alpha_spy_p1d": -0.003213738581973713, "alpha_c_p1d": -0.004001740074807136, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.06046102686377797, "ret_signed_p1w": -0.06046102686377797, "alpha_spy_p1w": -0.0528815021516118, "alpha_c_p1w": -0.02585281936163819, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": -0.026352128487647786, "ret_signed_p1m": -0.026352128487647786, "alpha_spy_p1m": -0.0539000810695367, "alpha_c_p1m": -0.1084413858754818, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": -0.01584366868256315, "ret_signed_p3m": -0.01584366868256315, "alpha_spy_p3m": -0.040174301739457574, "alpha_c_p3m": -0.1148496148905338, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.034279923956166636, "ret_signed_p6m": 0.034279923956166636, "alpha_spy_p6m": -0.03557288954310378, "alpha_c_p6m": -0.3602733243718008, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.2343, -0.2566, -0.2443, -0.2231, -0.2193, -0.2299, -0.2204, -0.2154, -0.2081, -0.2142, -0.2023, -0.219, -0.204, -0.2071, -0.1996, -0.1996, -0.2086, -0.2069, -0.1863, -0.1511, -0.1472, -0.1321, -0.1308, -0.1566, -0.1349, -0.1234, -0.1234, -0.1294, -0.1198, -0.1039, -0.0972, -0.0927, -0.0974, -0.0609, -0.0572, -0.0482, -0.0515, -0.0572, -0.0811, -0.0605, -0.0442, -0.0455, -0.0276, -0.0344, -0.0138, -0.0215, -0.0443, -0.0097, -0.0193, -0.0129, -0.0055, 0.0051, 0.0016, 0.0076, -0.001, 0.0014, -0.0073, 0.0013, -0.0337, -0.035, -0.0374, -0.0208, -0.0108, 0.0, -0.0009, -0.0088, -0.0418, -0.0418, -0.0605, -0.0613, -0.0556, -0.0811, -0.074, -0.0606, -0.0244, -0.0253, -0.0217, -0.0221, -0.0379, -0.0631, -0.0304, -0.028, 0.0102, -0.0183, -0.0264, -0.0224, -0.0196, 0.0005, 0.0265, 0.0302, 0.0392, 0.0322, 0.0208, 0.018, 0.0404, 0.0595, 0.0077, 0.0361, -0.0095, -0.0106, 0.0003, 0.008, 0.0048, -0.0033, -0.0081, 0.0022, 0.0248, 0.0535, 0.106, 0.1391, 0.1163, 0.1141, 0.1382, 0.0931, 0.074, 0.0348, 0.0352, 0.0951, 0.0627, 0.0662, 0.0281, 0.0463, 0.0266, -0.0022, 0.0262, -0.0062, -0.0158, 0.0043, -0.0217, -0.0083, -0.0083, -0.0262, -0.0101, -0.0016, -0.0119, 0.009, 0.0036, 0.0209, 0.0177, 0.0112, -0.0045, -0.037, -0.03, -0.0222, -0.0595, -0.0419, -0.0042, 0.0107, 0.041, 0.0377, 0.0377, 0.0483, 0.0356, 0.0319, 0.0261, 0.0261, 0.0391, 0.0351, 0.0302, 0.0405, 0.0181, 0.0171, 0.0176, 0.0223, 0.0076, 0.0292, 0.0247, 0.0247, -0.0202, 0.0086, 0.017, 0.0326, 0.026, 0.0373, 0.0538, 0.0592, 0.0516, 0.0212, -0.0078, -0.0416, -0.0543, 0.0201, 0.0456, 0.0374, 0.0457, 0.0286, 0.0058, 0.0058, 0.0177, 0.0343]}
{"unit_id": "orig:1960130067440508968", "ticker_idx": 1, "ts": "2025-08-26T00:00:00+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "there is potential for a near-term positive sentiment shift", "tweet_type": "analysis", "timeline": "months", "summary": "Morgan Stanley (shared/adopted): most bullish on NVDA into 12-month ramp; sees near-term sentiment recovery potential in MRVL and MU despite EW ratings.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: Latest Commentary on Investor Feedback Regarding “2Q Outlook”\n\nMorgan Stanley (J. Moore, 25/08/25)\n\nRegarding our feedback on NVIDIA’s outlook: Although for most of this year we have enjoyed significant optimism around NVIDIA’s performance, earlier on market sentiment was still relatively cautious, with a long list of concerns—DeepSeek, Blackwell delays, cabinet challenges, some customers’ data center construction slowdowns, and macroeconomic worries.\n\nFrom our investor interactions, it appears that these pessimistic concerns have largely passed, and we share in the current optimism. We had previously forecast October-quarter revenue of $52.5 billion and believed there could be upside, but in some investor conversations the optimism was noticeably higher. Some sell-side analysts have even suggested figures as high as $55 billion, and the market has shown some emotional buy-in to these numbers, though they are not yet locked in. We continue to like NVIDIA at this share price range, because the ramp-up over the next 12 months will be very strong, although we expect the company will remain cautious regarding supply-side issues and the China variable. It is worth noting that the upper bound of this expectation is based on several billion dollars of contribution from China.\n\nFrom this perspective, the China factor has actually clarified market expectations about the outlook. One major variable in guidance is how management assesses the China opportunity; while a small number of licenses have been approved, there is no clarity on future approvals. Media reports last week intensified market concerns, one claiming that the Chinese government might block domestic customers from purchasing U.S. chips (link), and another suggesting the company would slow its build-out pace (link). NVIDIA’s statement in response was only that “we constantly manage the supply chain to respond to market conditions”—a vague expression, but it did not deny the reports.\n\nThis remains an important issue requiring a long-term solution, and we believe ultimately both sides will have incentives to ensure China can obtain U.S. hardware. But how the company guides this part of the business will be one of the key factors affecting the specific guidance number. Against this backdrop, they will clearly take a conservative stance on this portion, which will in some sense lower the market’s hurdle.\n\nWe want to make clear that we remain very optimistic about the next 12 months, with many strong supports. It is only necessary to understand the factors that may affect near-term guidance.\n\nMeanwhile, are there any other AI stocks where market sentiment has not turned into “one-sided” optimism?\n\nYes, we can name a few. We rate them all Equal Weight (EW), but we believe both companies could see a positive sentiment shift.\n\nMarvell (outlook as follows): Market sentiment has been weighed down by Amazon’s Trainium performing only in line with expectations (rather than clearly above) and by concerns about growth next year. We think that, in the long term, expectations for ASIC may be too high, since commercial ASICs usually take longer to reach mass production. However, we still expect Amazon-related business to grow in line with market expectations next year and for the optical business to remain strong. Although reservations about ASIC keep us at EW, there is potential for a near-term positive sentiment shift.\n\nMicron: High Bandwidth Memory (HBM) pricing will face a hard reset next year, with at least one customer—NVIDIA—already committed to full-year 2025 pricing. But current market sentiment is quite negative. We expect HBM will maintain a significant but gradually narrowing premium relative to DDR5. We maintain EW on Micron because the stock is expensive relative to its own history, and even in a strong cycle its free cash flow is quite limited. Still, we believe there is short-term potential for sentiment recovery.\n\nWe continue to prefer NVDA the most. Despite higher expectations, in a very strong AI environment, its positive sentiment still has the chance to spread to these peers.\n\n$NVDA $MRVL $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017640767834846915, "ret_signed_m1d": -0.017640767834846915, "alpha_spy_m1d": -0.013471189660336558, "alpha_c_m1d": -0.008104201920820508, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007137026755229758, "ret_signed_p1d": 0.007137026755229758, "alpha_spy_p1d": 0.004858513473220416, "alpha_c_p1d": 0.004070511980386993, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.13008355026345564, "ret_signed_p1w": -0.13008355026345564, "alpha_spy_p1w": -0.12250402555128947, "alpha_c_p1w": -0.09547534276131586, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": 0.07850780896973886, "ret_signed_p1m": 0.07850780896973886, "alpha_spy_p1m": 0.050959856387849944, "alpha_c_p1m": -0.0035814484180951567, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.043647678719026306, "ret_signed_p3m": 0.043647678719026306, "alpha_spy_p3m": 0.019317045662131882, "alpha_c_p3m": -0.055358267488944346, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.0665136061248921, "ret_signed_p6m": 0.0665136061248921, "alpha_spy_p6m": -0.003339207374378317, "alpha_c_p6m": -0.32803964220307535, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.1425, -0.1901, -0.1729, -0.1609, -0.1079, -0.1233, -0.0803, -0.0697, -0.0737, -0.0818, -0.063, -0.0959, -0.0525, -0.0583, 0.0085, 0.0085, -0.0109, -0.0476, 0.012, 0.0217, 0.076, 0.0382, 0.0414, 0.0258, -0.001, 0.0116, 0.0116, -0.0373, -0.0319, -0.0277, -0.0129, -0.0209, -0.0236, -0.0249, -0.0459, -0.0303, 0.0053, -0.0162, -0.0306, -0.0133, -0.003, -0.0007, 0.0222, 0.028, 0.1007, 0.0823, 0.0026, 0.0306, 0.0319, 0.0143, 0.0214, 0.0415, 0.0407, 0.0478, 0.0681, 0.0644, 0.026, 0.0334, -0.0295, -0.0409, -0.0411, -0.017, -0.0176, 0.0, 0.0071, 0.04, -0.1534, -0.1534, -0.1301, -0.1609, -0.1368, -0.1472, -0.1112, -0.0999, -0.0964, -0.1033, -0.0931, -0.092, -0.0727, -0.0442, -0.0004, 0.0, 0.0171, 0.0048, 0.0785, 0.1286, 0.12, 0.1095, 0.1321, 0.1297, 0.1608, 0.1611, 0.1974, 0.1712, 0.2456, 0.2211, 0.1536, 0.2045, 0.1618, 0.1978, 0.1889, 0.1851, 0.1567, 0.1354, 0.0922, 0.1153, 0.1337, 0.1954, 0.192, 0.2148, 0.1935, 0.2632, 0.2177, 0.1803, 0.2518, 0.2576, 0.2252, 0.2563, 0.2037, 0.2037, 0.1793, 0.1649, 0.1245, 0.0602, 0.0958, 0.0333, 0.0436, 0.1291, 0.1242, 0.182, 0.182, 0.2047, 0.2276, 0.2517, 0.3502, 0.3231, 0.3328, 0.2397, 0.1979, 0.246, 0.2051, 0.1377, 0.1354, 0.1329, 0.1009, 0.1382, 0.1331, 0.1427, 0.1815, 0.1655, 0.1655, 0.1634, 0.1556, 0.1691, 0.1451, 0.1451, 0.2045, 0.2159, 0.1889, 0.1405, 0.1245, 0.1222, 0.1178, 0.1199, 0.0951, 0.0839, 0.085, 0.085, 0.0761, 0.1132, 0.1206, 0.0819, 0.1027, 0.1183, 0.1276, 0.0969, 0.0642, 0.0607, 0.0186, -0.0058, 0.0007, 0.0826, 0.1105, 0.1059, 0.0969, 0.0549, 0.06, 0.06, 0.0654, 0.0665]}
{"unit_id": "orig:1960130067440508968", "ticker_idx": 2, "ts": "2025-08-26T00:00:00+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe there is short-term potential for sentiment recovery", "tweet_type": "analysis", "timeline": "months", "summary": "Morgan Stanley (shared/adopted): most bullish on NVDA into 12-month ramp; sees near-term sentiment recovery potential in MRVL and MU despite EW ratings.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: Latest Commentary on Investor Feedback Regarding “2Q Outlook”\n\nMorgan Stanley (J. Moore, 25/08/25)\n\nRegarding our feedback on NVIDIA’s outlook: Although for most of this year we have enjoyed significant optimism around NVIDIA’s performance, earlier on market sentiment was still relatively cautious, with a long list of concerns—DeepSeek, Blackwell delays, cabinet challenges, some customers’ data center construction slowdowns, and macroeconomic worries.\n\nFrom our investor interactions, it appears that these pessimistic concerns have largely passed, and we share in the current optimism. We had previously forecast October-quarter revenue of $52.5 billion and believed there could be upside, but in some investor conversations the optimism was noticeably higher. Some sell-side analysts have even suggested figures as high as $55 billion, and the market has shown some emotional buy-in to these numbers, though they are not yet locked in. We continue to like NVIDIA at this share price range, because the ramp-up over the next 12 months will be very strong, although we expect the company will remain cautious regarding supply-side issues and the China variable. It is worth noting that the upper bound of this expectation is based on several billion dollars of contribution from China.\n\nFrom this perspective, the China factor has actually clarified market expectations about the outlook. One major variable in guidance is how management assesses the China opportunity; while a small number of licenses have been approved, there is no clarity on future approvals. Media reports last week intensified market concerns, one claiming that the Chinese government might block domestic customers from purchasing U.S. chips (link), and another suggesting the company would slow its build-out pace (link). NVIDIA’s statement in response was only that “we constantly manage the supply chain to respond to market conditions”—a vague expression, but it did not deny the reports.\n\nThis remains an important issue requiring a long-term solution, and we believe ultimately both sides will have incentives to ensure China can obtain U.S. hardware. But how the company guides this part of the business will be one of the key factors affecting the specific guidance number. Against this backdrop, they will clearly take a conservative stance on this portion, which will in some sense lower the market’s hurdle.\n\nWe want to make clear that we remain very optimistic about the next 12 months, with many strong supports. It is only necessary to understand the factors that may affect near-term guidance.\n\nMeanwhile, are there any other AI stocks where market sentiment has not turned into “one-sided” optimism?\n\nYes, we can name a few. We rate them all Equal Weight (EW), but we believe both companies could see a positive sentiment shift.\n\nMarvell (outlook as follows): Market sentiment has been weighed down by Amazon’s Trainium performing only in line with expectations (rather than clearly above) and by concerns about growth next year. We think that, in the long term, expectations for ASIC may be too high, since commercial ASICs usually take longer to reach mass production. However, we still expect Amazon-related business to grow in line with market expectations next year and for the optical business to remain strong. Although reservations about ASIC keep us at EW, there is potential for a near-term positive sentiment shift.\n\nMicron: High Bandwidth Memory (HBM) pricing will face a hard reset next year, with at least one customer—NVIDIA—already committed to full-year 2025 pricing. But current market sentiment is quite negative. We expect HBM will maintain a significant but gradually narrowing premium relative to DDR5. We maintain EW on Micron because the stock is expensive relative to its own history, and even in a strong cycle its free cash flow is quite limited. Still, we believe there is short-term potential for sentiment recovery.\n\nWe continue to prefer NVDA the most. Despite higher expectations, in a very strong AI environment, its positive sentiment still has the chance to spread to these peers.\n\n$NVDA $MRVL $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0006867236402662291, "ret_signed_m1d": -0.0006867236402662291, "alpha_spy_m1d": 0.0034828545342441286, "alpha_c_m1d": 0.008849842273760178, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010729638785070383, "ret_signed_p1d": 0.010729638785070383, "alpha_spy_p1d": 0.008451125503061041, "alpha_c_p1d": 0.007663124010227618, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": 0.016995737866876803, "ret_signed_p1w": 0.016995737866876803, "alpha_spy_p1w": 0.024575262579042967, "alpha_c_p1w": 0.05160394536901658, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": 0.38806870305705155, "ret_signed_p1m": 0.38806870305705155, "alpha_spy_p1m": 0.36052075047516263, "alpha_c_p1m": 0.30597944566921753, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.7811146604839896, "ret_signed_p3m": 0.7811146604839896, "alpha_spy_p3m": 0.7567840274270952, "alpha_c_p3m": 0.6821087142760189, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 2.61702817540985, "ret_signed_p6m": 2.61702817540985, "alpha_spy_p6m": 2.54717536191058, "alpha_c_p6m": 2.2224749270818824, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.1699, -0.1899, -0.158, -0.1231, -0.1146, -0.0885, -0.069, -0.0485, -0.0212, -0.005, -0.0037, -0.0087, 0.0277, 0.032, 0.0447, 0.0447, 0.0599, 0.0469, 0.0969, 0.0912, 0.0805, 0.0699, 0.0569, 0.0367, 0.044, 0.0487, 0.0487, 0.0294, 0.068, 0.0493, 0.0567, 0.0689, 0.0181, 0.031, -0.0006, -0.0278, -0.0181, -0.0281, -0.0625, -0.0573, -0.0409, -0.045, -0.0451, -0.039, -0.0151, -0.0632, -0.0997, -0.0749, -0.0639, -0.0663, -0.0397, 0.0205, 0.062, 0.0966, 0.0667, 0.0755, 0.0375, 0.0605, 0.0476, 0.0061, -0.0061, 0.0101, -0.0007, 0.0, 0.0107, 0.0472, 0.0215, 0.0215, 0.017, 0.0191, 0.0662, 0.1276, 0.1284, 0.1609, 0.2017, 0.2924, 0.3496, 0.3542, 0.3633, 0.3733, 0.4497, 0.3968, 0.413, 0.4284, 0.3881, 0.3462, 0.35, 0.4069, 0.4362, 0.5635, 0.5773, 0.6133, 0.6402, 0.5949, 0.6881, 0.6519, 0.5598, 0.6557, 0.6067, 0.6486, 0.7395, 0.7383, 0.776, 0.7375, 0.7047, 0.7754, 0.8812, 0.8905, 0.906, 0.9465, 0.924, 0.922, 1.0159, 0.8727, 1.0399, 1.047, 1.0435, 1.1756, 1.0709, 1.1035, 1.0352, 1.12, 1.0781, 0.9626, 0.9404, 0.7296, 0.7811, 0.9233, 0.9285, 0.9777, 0.9777, 1.0311, 1.0653, 1.057, 1.0112, 0.9467, 1.0375, 1.1208, 1.1681, 1.265, 1.2199, 1.0712, 1.0399, 0.997, 0.937, 1.1348, 1.284, 1.3756, 1.3729, 1.4623, 1.4623, 1.4461, 1.5294, 1.5144, 1.4524, 1.4524, 1.7103, 1.6822, 1.9509, 1.9176, 1.8099, 1.9652, 1.9719, 1.9054, 1.8643, 1.8925, 2.1169, 2.1169, 2.1363, 2.3434, 2.4162, 2.434, 2.3433, 2.525, 2.7402, 2.7445, 2.5649, 2.7618, 2.6041, 2.26, 2.29, 2.3914, 2.2952, 2.2072, 2.5259, 2.5571, 2.5372, 2.5372, 2.4351, 2.617]}
{"unit_id": "quote:1972634218619621807", "ticker_idx": 0, "ts": "2025-09-29T12:07:02+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$NVDA", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses Barclays Buy on NVDA; $2T+ planned AI spend makes Jensen's $3-4T TAM forecast look credible.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "With the wave of announcements that have come over the last 6-9mo, we now estimate over $2T of planned spend at ~40GW of power in total.  Within that, we attribute ~65-70% to compute & networking with more deals likely in the pipeline, which starts to make the updated guidance of $3-4T look much more real.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": "Barclays buy thesis on $NVDA:\n\n\"When tracking AI capacity additions over the LTM, AI TAMs don't seem so outlandish anymore and NVDA looks like the most interesting name in our group.  When Jensen first forecast a $1T industry by the end of the decade, we admittedly balked.  With", "parent_author": null, "ret_m1d": -0.02012652464071829, "ret_signed_m1d": -0.02012652464071829, "alpha_spy_m1d": -0.017324034224595497, "alpha_c_m1d": -0.017523235078525157, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.002603289562193134, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026010341701790507, "ret_signed_p1d": 0.026010341701790507, "alpha_spy_p1d": 0.02224347494267853, "alpha_c_p1d": 0.01454321123295177, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.011467130468838738, "ret_p1w": 0.020291339324061797, "ret_signed_p1w": 0.020291339324061797, "alpha_spy_p1w": 0.008342857941649129, "alpha_c_p1w": -0.04392487592801397, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.06421621525207577, "ret_p1m": 0.10547149503197861, "ret_signed_p1m": 0.10547149503197861, "alpha_spy_p1m": 0.0702436882894093, "alpha_c_p1m": -0.0198926609562895, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.1253641559882681, "ret_p3m": 0.03723116702726181, "ret_signed_p3m": 0.03723116702726181, "alpha_spy_p3m": -0.006072817584891421, "alpha_c_p3m": -0.0949301562813969, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.1321613233086587, "ret_p6m": -0.03646289898957855, "ret_signed_p6m": -0.03646289898957855, "alpha_spy_p6m": -0.026245892577636654, "alpha_c_p6m": -0.26319633652153973, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": 0.22673343753196118, "price_path": [-0.1353, -0.1238, -0.1238, -0.1299, -0.1202, -0.1044, -0.0977, -0.0931, -0.0978, -0.0614, -0.0577, -0.0487, -0.052, -0.0576, -0.0815, -0.0609, -0.0447, -0.046, -0.0281, -0.0349, -0.0142, -0.0219, -0.0448, -0.0102, -0.0198, -0.0134, -0.006, 0.0046, 0.0011, 0.0071, -0.0015, 0.0009, -0.0078, 0.0008, -0.0342, -0.0355, -0.0378, -0.0213, -0.0113, -0.0005, -0.0014, -0.0093, -0.0422, -0.0422, -0.0609, -0.0618, -0.0561, -0.0816, -0.0745, -0.061, -0.0249, -0.0257, -0.0222, -0.0225, -0.0383, -0.0636, -0.0308, -0.0285, 0.0097, -0.0188, -0.0268, -0.0229, -0.0201, 0.0, 0.026, 0.0296, 0.0387, 0.0317, 0.0203, 0.0175, 0.0399, 0.0589, 0.0072, 0.0356, -0.01, -0.0111, -0.0002, 0.0075, 0.0043, -0.0038, -0.0086, 0.0017, 0.0243, 0.053, 0.1055, 0.1385, 0.1157, 0.1135, 0.1376, 0.0926, 0.0735, 0.0343, 0.0346, 0.0946, 0.0622, 0.0657, 0.0276, 0.0458, 0.0261, -0.0027, 0.0257, -0.0067, -0.0163, 0.0038, -0.0222, -0.0087, -0.0087, -0.0267, -0.0106, -0.0021, -0.0124, 0.0085, 0.0031, 0.0204, 0.0172, 0.0107, -0.005, -0.0375, -0.0305, -0.0227, -0.0599, -0.0423, -0.0047, 0.0102, 0.0405, 0.0372, 0.0372, 0.0478, 0.0351, 0.0313, 0.0256, 0.0256, 0.0386, 0.0345, 0.0297, 0.04, 0.0176, 0.0166, 0.017, 0.0218, 0.0071, 0.0287, 0.0241, 0.0241, -0.0207, 0.0081, 0.0165, 0.0321, 0.0255, 0.0367, 0.0532, 0.0587, 0.0511, 0.0207, -0.0082, -0.0421, -0.0548, 0.0196, 0.0451, 0.0368, 0.0452, 0.028, 0.0053, 0.0053, 0.0172, 0.0338, 0.0333, 0.0439, 0.0534, 0.0605, 0.0755, 0.0168, -0.0256, 0.0035, -0.0098, 0.0066, 0.0082, -0.0221, 0.0045, 0.0161, 0.0231, 0.0072, -0.0087, 0.0076, 0.0005, -0.0079, -0.018, -0.0502, -0.034, -0.0365]}
{"unit_id": "thread:1970275538670702949", "ticker_idx": 0, "ts": "2025-09-22T23:54:29+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory prices are likely to keep climbing, significantly boosting corporate profitability for the foreseeable future", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory supercycle thesis: Samsung, Hynix, Micron all benefit as HBM-driven supply tightness + AI/server demand push DRAM/NAND prices higher through at least 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung, SK Hynix Raise DRAM and NAND Prices – Growing Expectations for a “Memory Supercycle” [biz-plus]\n\nThe tectonic plates of the memory semiconductor market are shifting. Samsung Electronics (005930) and SK Hynix (000660) have simultaneously raised prices for their flagship memory products such as DRAM and NAND flash, signaling the end of a prolonged downturn. With the expansion of the artificial intelligence (AI) market sparking supply chain restructuring and surging demand, expectations are mounting that the industry is entering a new “supercycle.”\n\nAccording to industry sources on the 22nd, Samsung Electronics is currently negotiating fourth-quarter DRAM and NAND flash prices with major clients. Foreign media reported that Samsung has notified some customers that it will raise DRAM prices by up to 30% and NAND flash prices by up to 10% in Q4. SK Hynix has yet to formally announce price increases but is reportedly negotiating with clients under the principle of charging higher prices in line with market conditions. With the two memory giants effectively moving together on price hikes, analysts say the market has decisively shifted to a supplier-led dynamic.\n\nThe root cause of this price surge lies in the structural changes to the memory market brought on by the AI era. The centerpiece is, without question, high-bandwidth memory (HBM). Serving as the “heart” of AI semiconductors, HBM boasts far higher profitability than standard DRAM. Global memory leaders—Samsung Electronics, SK Hynix, and Micron—are allocating their limited wafer capacity to HBM production as a top priority. Because HBM chips are two to three times larger than conventional DRAM, far fewer chips can be produced from a single wafer. As a result, the more capacity is devoted to HBM, the less supply is available for PC, mobile, and general-purpose server DRAM.\n\nWhile supply has tightened, demand is exploding at a pace far exceeding forecasts. The strongest demand driver, by far, is AI data centers. Training and inference for AI models require processing enormous amounts of data at high speed, making high-performance DRAM and large-capacity enterprise solid-state drives (eSSD) essential. Market research firm TechInsights projects that DRAM capacity per AI server will grow from an average of 1TB in 2024 to 2.5TB by 2030—more than a 2.5-fold increase.\n\nOn top of this, replacement demand from the general server market is further fueling growth. Data center servers worldwide, which were massively deployed during the last supercycle in 2017–2018, are now hitting their 4–5-year replacement cycle. As enterprises resume general server investments that had briefly slowed due to the AI boom, demand for commodity DRAM is soaring. Global investment bank Morgan Stanley noted in a report that “AI-related demand is stimulating demand for general-purpose server and mobile DRAM as well, driving a rebound in prices.”\n\nThe NAND flash market is following a similar pattern. Exploding demand to replace or supplement traditional hard disk drives (HDDs) with eSSDs for AI data processing is already creating visible shortages. Morgan Stanley predicts that the 2026 order volume for large-capacity eSSDs from U.S. hyperscale data center operators alone could surpass the total size of this year’s entire eSSD market. It forecasts the NAND market will flip into shortage by 2026.\n\nThese market shifts are raising expectations for upcoming corporate earnings. All eyes are on Micron’s earnings release scheduled for the 23rd (local time) in the U.S. The outlook and pricing policies Micron presents for the HBM market will serve as an important benchmark for the strategies of Samsung Electronics and SK Hynix. Wall Street expects Micron’s Q4 revenue to surge 43% year-on-year.\n\nSamsung Electronics and SK Hynix are set to announce their Q3 earnings in the weeks of October 13 and October 20, respectively. With HBM revenue beginning to flow in and the decline in commodity memory prices halted, both firms are projected to show clear earnings improvements. An industry insider commented, “With supply structurally constrained and demand rising from both AI and general servers, memory prices are likely to keep climbing, significantly boosting corporate profitability for the foreseeable future.”\n\n---\n\nhttps://t.co/sNy0byQgbp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "thread:1970275538670702949", "ticker_idx": 1, "ts": "2025-09-22T23:54:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory prices are likely to keep climbing, significantly boosting corporate profitability for the foreseeable future", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory supercycle thesis: Samsung, Hynix, Micron all benefit as HBM-driven supply tightness + AI/server demand push DRAM/NAND prices higher through at least 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung, SK Hynix Raise DRAM and NAND Prices – Growing Expectations for a “Memory Supercycle” [biz-plus]\n\nThe tectonic plates of the memory semiconductor market are shifting. Samsung Electronics (005930) and SK Hynix (000660) have simultaneously raised prices for their flagship memory products such as DRAM and NAND flash, signaling the end of a prolonged downturn. With the expansion of the artificial intelligence (AI) market sparking supply chain restructuring and surging demand, expectations are mounting that the industry is entering a new “supercycle.”\n\nAccording to industry sources on the 22nd, Samsung Electronics is currently negotiating fourth-quarter DRAM and NAND flash prices with major clients. Foreign media reported that Samsung has notified some customers that it will raise DRAM prices by up to 30% and NAND flash prices by up to 10% in Q4. SK Hynix has yet to formally announce price increases but is reportedly negotiating with clients under the principle of charging higher prices in line with market conditions. With the two memory giants effectively moving together on price hikes, analysts say the market has decisively shifted to a supplier-led dynamic.\n\nThe root cause of this price surge lies in the structural changes to the memory market brought on by the AI era. The centerpiece is, without question, high-bandwidth memory (HBM). Serving as the “heart” of AI semiconductors, HBM boasts far higher profitability than standard DRAM. Global memory leaders—Samsung Electronics, SK Hynix, and Micron—are allocating their limited wafer capacity to HBM production as a top priority. Because HBM chips are two to three times larger than conventional DRAM, far fewer chips can be produced from a single wafer. As a result, the more capacity is devoted to HBM, the less supply is available for PC, mobile, and general-purpose server DRAM.\n\nWhile supply has tightened, demand is exploding at a pace far exceeding forecasts. The strongest demand driver, by far, is AI data centers. Training and inference for AI models require processing enormous amounts of data at high speed, making high-performance DRAM and large-capacity enterprise solid-state drives (eSSD) essential. Market research firm TechInsights projects that DRAM capacity per AI server will grow from an average of 1TB in 2024 to 2.5TB by 2030—more than a 2.5-fold increase.\n\nOn top of this, replacement demand from the general server market is further fueling growth. Data center servers worldwide, which were massively deployed during the last supercycle in 2017–2018, are now hitting their 4–5-year replacement cycle. As enterprises resume general server investments that had briefly slowed due to the AI boom, demand for commodity DRAM is soaring. Global investment bank Morgan Stanley noted in a report that “AI-related demand is stimulating demand for general-purpose server and mobile DRAM as well, driving a rebound in prices.”\n\nThe NAND flash market is following a similar pattern. Exploding demand to replace or supplement traditional hard disk drives (HDDs) with eSSDs for AI data processing is already creating visible shortages. Morgan Stanley predicts that the 2026 order volume for large-capacity eSSDs from U.S. hyperscale data center operators alone could surpass the total size of this year’s entire eSSD market. It forecasts the NAND market will flip into shortage by 2026.\n\nThese market shifts are raising expectations for upcoming corporate earnings. All eyes are on Micron’s earnings release scheduled for the 23rd (local time) in the U.S. The outlook and pricing policies Micron presents for the HBM market will serve as an important benchmark for the strategies of Samsung Electronics and SK Hynix. Wall Street expects Micron’s Q4 revenue to surge 43% year-on-year.\n\nSamsung Electronics and SK Hynix are set to announce their Q3 earnings in the weeks of October 13 and October 20, respectively. With HBM revenue beginning to flow in and the decline in commodity memory prices halted, both firms are projected to show clear earnings improvements. An industry insider commented, “With supply structurally constrained and demand rising from both AI and general servers, memory prices are likely to keep climbing, significantly boosting corporate profitability for the foreseeable future.”\n\n---\n\nhttps://t.co/sNy0byQgbp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012820521976864452, "ret_signed_m1d": 0.012820521976864452, "alpha_spy_m1d": 0.01752925290143681, "alpha_c_m1d": 0.032872079731812165, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028489969482429656, "ret_signed_p1d": 0.028489969482429656, "alpha_spy_p1d": 0.033933444322237216, "alpha_c_p1d": 0.029948808796310566, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.0056981188806843885, "ret_signed_p1w": -0.0056981188806843885, "alpha_spy_p1w": -0.0009593796228162255, "alpha_c_p1w": -0.00721900613776616, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.3646722878607491, "ret_signed_p1m": 0.3646722878607491, "alpha_spy_p1m": 0.357999115706531, "alpha_c_p1m": 0.2928779715663701, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5737758511959559, "ret_signed_p3m": 0.5737758511959559, "alpha_spy_p3m": 0.5593346629727354, "alpha_c_p3m": 0.4959287722952972, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.7706148936884407, "ret_signed_p6m": 1.7706148936884407, "alpha_spy_p6m": 1.7617190797680904, "alpha_c_p6m": 1.5348855405453752, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.1864, -0.1664, -0.1921, -0.1693, -0.1878, -0.2063, -0.2077, -0.2305, -0.229, -0.1977, -0.2006, -0.1551, -0.1622, -0.1465, -0.1508, -0.1579, -0.2333, -0.2347, -0.2248, -0.2361, -0.2347, -0.2333, -0.2433, -0.2546, -0.2532, -0.2504, -0.2219, -0.266, -0.266, -0.2504, -0.2646, -0.2546, -0.2703, -0.2404, -0.2347, -0.2091, -0.2134, -0.2134, -0.239, -0.2518, -0.2731, -0.303, -0.2859, -0.2618, -0.2561, -0.2603, -0.235, -0.2336, -0.2707, -0.2578, -0.2521, -0.2436, -0.2208, -0.2108, -0.1795, -0.1339, -0.1254, -0.0641, -0.057, -0.0085, -0.0499, 0.0128, 0.0128, 0.0, 0.0285, 0.0185, 0.0157, -0.0413, -0.0057, -0.01, 0.0256, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.2194, 0.1823, 0.1724, 0.2037, 0.2892, 0.3262, 0.3832, 0.3647, 0.3718, 0.3632, 0.453, 0.5242, 0.4843, 0.5897, 0.6182, 0.5926, 0.7664, 0.6695, 0.6496, 0.6895, 0.6524, 0.7265, 0.7635, 0.7578, 0.7436, 0.5954, 0.7265, 0.6239, 0.6011, 0.6268, 0.4843, 0.4815, 0.4786, 0.4929, 0.551, 0.5111, 0.5339, 0.5909, 0.5738, 0.5453, 0.551, 0.6451, 0.6137, 0.6736, 0.6108, 0.6279, 0.5795, 0.5111, 0.5709, 0.5738, 0.5595, 0.6536, 0.665, 0.6764, 0.6764, 0.7078, 0.8247, 0.856, 0.856, 0.856, 0.9302, 0.9843, 1.0699, 1.1155, 1.1554, 1.1212, 1.1354, 1.1041, 1.1155, 1.1354, 1.1554, 1.1782, 1.1183, 1.1098, 1.1525, 1.1868, 1.0984, 1.2808, 1.3977, 1.4547, 1.5916, 1.3664, 1.5859, 1.5659, 1.4006, 1.392, 1.5289, 1.4975, 1.4519, 1.5317, 1.5089, 1.5089, 1.5089, 1.5089, 1.5488, 1.7056, 1.7113, 1.8653, 1.9024, 2.1391, 2.0305, 2.0305, 1.6821, 1.425, 1.6878, 1.6392, 1.3879, 1.6792, 1.7278, 1.6564, 1.5992, 1.782, 1.7706]}
{"unit_id": "thread:1970275538670702949", "ticker_idx": 2, "ts": "2025-09-22T23:54:29+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Wall Street expects Micron's Q4 revenue to surge 43% year-on-year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory supercycle thesis: Samsung, Hynix, Micron all benefit as HBM-driven supply tightness + AI/server demand push DRAM/NAND prices higher through at least 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung, SK Hynix Raise DRAM and NAND Prices – Growing Expectations for a “Memory Supercycle” [biz-plus]\n\nThe tectonic plates of the memory semiconductor market are shifting. Samsung Electronics (005930) and SK Hynix (000660) have simultaneously raised prices for their flagship memory products such as DRAM and NAND flash, signaling the end of a prolonged downturn. With the expansion of the artificial intelligence (AI) market sparking supply chain restructuring and surging demand, expectations are mounting that the industry is entering a new “supercycle.”\n\nAccording to industry sources on the 22nd, Samsung Electronics is currently negotiating fourth-quarter DRAM and NAND flash prices with major clients. Foreign media reported that Samsung has notified some customers that it will raise DRAM prices by up to 30% and NAND flash prices by up to 10% in Q4. SK Hynix has yet to formally announce price increases but is reportedly negotiating with clients under the principle of charging higher prices in line with market conditions. With the two memory giants effectively moving together on price hikes, analysts say the market has decisively shifted to a supplier-led dynamic.\n\nThe root cause of this price surge lies in the structural changes to the memory market brought on by the AI era. The centerpiece is, without question, high-bandwidth memory (HBM). Serving as the “heart” of AI semiconductors, HBM boasts far higher profitability than standard DRAM. Global memory leaders—Samsung Electronics, SK Hynix, and Micron—are allocating their limited wafer capacity to HBM production as a top priority. Because HBM chips are two to three times larger than conventional DRAM, far fewer chips can be produced from a single wafer. As a result, the more capacity is devoted to HBM, the less supply is available for PC, mobile, and general-purpose server DRAM.\n\nWhile supply has tightened, demand is exploding at a pace far exceeding forecasts. The strongest demand driver, by far, is AI data centers. Training and inference for AI models require processing enormous amounts of data at high speed, making high-performance DRAM and large-capacity enterprise solid-state drives (eSSD) essential. Market research firm TechInsights projects that DRAM capacity per AI server will grow from an average of 1TB in 2024 to 2.5TB by 2030—more than a 2.5-fold increase.\n\nOn top of this, replacement demand from the general server market is further fueling growth. Data center servers worldwide, which were massively deployed during the last supercycle in 2017–2018, are now hitting their 4–5-year replacement cycle. As enterprises resume general server investments that had briefly slowed due to the AI boom, demand for commodity DRAM is soaring. Global investment bank Morgan Stanley noted in a report that “AI-related demand is stimulating demand for general-purpose server and mobile DRAM as well, driving a rebound in prices.”\n\nThe NAND flash market is following a similar pattern. Exploding demand to replace or supplement traditional hard disk drives (HDDs) with eSSDs for AI data processing is already creating visible shortages. Morgan Stanley predicts that the 2026 order volume for large-capacity eSSDs from U.S. hyperscale data center operators alone could surpass the total size of this year’s entire eSSD market. It forecasts the NAND market will flip into shortage by 2026.\n\nThese market shifts are raising expectations for upcoming corporate earnings. All eyes are on Micron’s earnings release scheduled for the 23rd (local time) in the U.S. The outlook and pricing policies Micron presents for the HBM market will serve as an important benchmark for the strategies of Samsung Electronics and SK Hynix. Wall Street expects Micron’s Q4 revenue to surge 43% year-on-year.\n\nSamsung Electronics and SK Hynix are set to announce their Q3 earnings in the weeks of October 13 and October 20, respectively. With HBM revenue beginning to flow in and the decline in commodity memory prices halted, both firms are projected to show clear earnings improvements. An industry insider commented, “With supply structurally constrained and demand rising from both AI and general servers, memory prices are likely to keep climbing, significantly boosting corporate profitability for the foreseeable future.”\n\n---\n\nhttps://t.co/sNy0byQgbp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011481025777169696, "ret_signed_m1d": -0.011481025777169696, "alpha_spy_m1d": -0.006772294852597338, "alpha_c_m1d": 0.008570531977778018, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01087348305772684, "ret_signed_p1d": 0.01087348305772684, "alpha_spy_p1d": 0.0163169578975344, "alpha_c_p1d": 0.012332322371607751, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.004373843452624926, "ret_signed_p1w": -0.004373843452624926, "alpha_spy_p1w": 0.00036489580524323717, "alpha_c_p1w": -0.005894730709706697, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.22959944416106892, "ret_signed_p1m": 0.22959944416106892, "alpha_spy_p1m": 0.22292627200685078, "alpha_c_p1m": 0.15780512786668988, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5107863663515086, "ret_signed_p3m": 0.5107863663515086, "alpha_spy_p3m": 0.49634517812828816, "alpha_c_p3m": 0.43293928745085, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.8074702227486457, "ret_signed_p6m": 1.8074702227486457, "alpha_spy_p6m": 1.7985744088282953, "alpha_c_p6m": 1.5717408696055801, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2277, -0.2353, -0.2428, -0.252, -0.2663, -0.2612, -0.2578, -0.2578, -0.2715, -0.2442, -0.2574, -0.2522, -0.2435, -0.2795, -0.2704, -0.2927, -0.312, -0.3051, -0.3122, -0.3365, -0.3328, -0.3213, -0.3241, -0.3242, -0.3199, -0.303, -0.337, -0.3629, -0.3453, -0.3375, -0.3392, -0.3204, -0.2778, -0.2485, -0.224, -0.2451, -0.2389, -0.2658, -0.2495, -0.2586, -0.288, -0.2966, -0.2851, -0.2928, -0.2923, -0.2847, -0.2589, -0.2771, -0.2771, -0.2803, -0.2788, -0.2455, -0.202, -0.2014, -0.1785, -0.1496, -0.0853, -0.0449, -0.0416, -0.0352, -0.0281, 0.0259, -0.0115, 0.0, 0.0109, -0.0177, -0.0473, -0.0446, -0.0044, 0.0164, 0.1065, 0.1162, 0.1417, 0.1607, 0.1287, 0.1946, 0.1691, 0.1038, 0.1717, 0.137, 0.1667, 0.2311, 0.2301, 0.2568, 0.2296, 0.2064, 0.2565, 0.3313, 0.3379, 0.3489, 0.3775, 0.3616, 0.3602, 0.4266, 0.3253, 0.4436, 0.4487, 0.4462, 0.5397, 0.4656, 0.4886, 0.4403, 0.5003, 0.4707, 0.3889, 0.3732, 0.224, 0.2605, 0.3611, 0.3648, 0.3996, 0.3996, 0.4374, 0.4616, 0.4557, 0.4233, 0.3777, 0.4419, 0.5009, 0.5343, 0.6029, 0.571, 0.4657, 0.4436, 0.4133, 0.3708, 0.5108, 0.6164, 0.6812, 0.6793, 0.7426, 0.7426, 0.7311, 0.79, 0.7794, 0.7355, 0.7355, 0.918, 0.8981, 1.0883, 1.0648, 0.9886, 1.0984, 1.1032, 1.0561, 1.0271, 1.047, 1.2058, 1.2058, 1.2195, 1.3661, 1.4176, 1.4302, 1.366, 1.4946, 1.6469, 1.65, 1.5228, 1.6622, 1.5506, 1.3071, 1.3283, 1.4001, 1.332, 1.2697, 1.4952, 1.5173, 1.5032, 1.5032, 1.431, 1.5597, 1.5378, 1.6036, 1.5599, 1.5419, 1.6087, 1.527, 1.5076, 1.5094, 1.3088, 1.437, 1.4144, 1.2517, 1.3674, 1.4513, 1.546, 1.4649, 1.5912, 1.6865, 1.8075]}
{"unit_id": "orig:1944385898784891123", "ticker_idx": 0, "ts": "2025-07-13T13:18:17+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if the MTIA v4 project materializes, it will bring annual revenues of $3-4 billion to MediaTek from 2028 to 2029, which accounts for over 10% of its total revenue growth", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish MediaTek on cloud ASIC wins: Meta MTIA v4 ($3-4B/yr), AI glasses ($2B), Google TPU v7e ($1.5-2B); TSMC 2nm as process node beneficiary.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "MediaTek Industry Note:\n\nEuropean institutional investors point out that MediaTek views cloud ASICs as a mid-to-long-term growth driver and is actively expanding its collaborations with key clients. According to their report, MediaTek is highly likely to cooperate with Meta on the development of the next-generation MTIA v4 ASIC, with mass production expected in 2028 and adoption of TSMC (2330) 2nm process. Previous versions, MTIA v1/v2/v3, were all designed by Broadcom. Compared to Google's TPU, the MTIA project requires more front-end design services and relies on SerDes IP technology support. If MediaTek successfully secures this project and achieves mass production, it will help establish its technological and market position in the cloud ASIC field. This report estimates that if the MTIA v4 project materializes, it will bring annual revenues of $3-4 billion to MediaTek from 2028 to 2029, which accounts for over 10% of its total revenue growth.\n\nPotential Collaboration on AI Glasses ASIC\nFurthermore, European institutional investors suggest that Meta may also collaborate with MediaTek on the development of an AI glasses-specific ASIC, potentially replacing the current Qualcomm solution. Meta is a leader in the AI glasses market, with shipments of approximately 2 million units in 2024, expected to reach 10 million units in 2025. This project aims to enable large language model execution on devices by adopting the N3 advanced process and a four-layer stack VHM (Very High Bandwidth Mobile Memory). This ASIC is scheduled for mass production in mid-2027, and is expected to contribute approximately $2 billion in revenue in 2028. (Calculated at $100 per chip, this would imply shipments of 20 million units.) European institutional investors also anticipate that this project could generate $200 million in NRE (Non-Recurring Engineering) revenue between 2026 and 2027.\n\nGoogle TPU Project and Short-Term Outlook\nRegarding the collaboration with Google, European institutional investors expect that the TPU v7e will undergo Tape Out (design completion) in September 2025. Although the mass production timeline remains uncertain, it is likely to fall between late 2026 and early 2027, and Google's commitment to its self-developed computing architecture remains unchanged. If TPU v7e successfully reaches mass production, it is estimated to contribute $1.5-2 billion in revenue in 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025270701360860626, "ret_signed_m1d": 0.025270701360860626, "alpha_spy_m1d": 0.027175184610185643, "alpha_c_m1d": 0.017665312516110365, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007220200388817322, "ret_signed_p1d": 0.007220200388817322, "alpha_spy_p1d": 0.011493367439444402, "alpha_c_p1d": -0.011986236268974437, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.028880801555269286, "ret_signed_p1w": 0.028880801555269286, "alpha_spy_p1w": 0.022542769570550902, "alpha_c_p1w": 0.010690741402045623, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": -0.018050590817995316, "ret_signed_p1m": -0.018050590817995316, "alpha_spy_p1m": -0.04666749089025579, "alpha_c_p1m": -0.06985199559130617, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": -0.028880891401221298, "ret_signed_p3m": -0.028880891401221298, "alpha_spy_p3m": -0.10604162330520706, "alpha_c_p3m": -0.2387854391308435, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.09298678671523386, "ret_signed_p6m": 0.09298678671523386, "alpha_spy_p6m": -0.02059644847298059, "alpha_c_p6m": -0.26978516481241543, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.0335, -0.0512, -0.0335, -0.0476, -0.0795, -0.0299, -0.0582, -0.0228, -0.0441, -0.0299, -0.0441, -0.0441, -0.0795, -0.083, -0.0937, -0.0972, -0.0866, -0.0618, -0.0689, -0.0405, -0.037, -0.0299, -0.0335, -0.0724, -0.0689, -0.0512, -0.0618, -0.0689, -0.0937, -0.0937, -0.0937, -0.1078, -0.1078, -0.1078, -0.0972, -0.0972, -0.1184, -0.0937, -0.0901, -0.0547, -0.0689, -0.0901, -0.1149, -0.1114, -0.0972, -0.0901, -0.1043, -0.1149, -0.1114, -0.1007, -0.083, -0.0866, -0.0901, -0.1149, -0.0972, -0.083, -0.083, -0.0686, -0.0758, -0.083, -0.0217, 0.0108, 0.0253, 0.0, 0.0072, 0.0181, 0.0036, 0.0181, 0.0289, 0.0325, 0.0325, 0.0325, 0.0325, 0.0108, -0.0181, -0.0036, -0.0108, -0.0108, -0.0397, -0.0325, -0.0469, -0.0217, -0.0253, -0.0217, -0.0181, -0.0108, 0.0036, -0.0108, 0.0181, 0.0036, -0.0217, -0.0144, -0.0144, 0.0181, 0.0108, 0.0108, 0.0, -0.0108, -0.0181, -0.0181, 0.0072, 0.0, 0.0361, 0.0325, 0.0903, 0.0758, 0.0686, 0.0722, 0.0722, 0.1083, 0.0903, 0.0903, 0.0397, 0.0144, -0.0072, -0.0397, -0.0289, -0.0542, -0.0542, -0.0505, -0.0722, -0.0758, -0.0469, -0.0469, -0.0433, -0.0361, -0.0289, -0.0289, -0.0505, -0.0578, -0.0505, -0.0397, -0.0397, -0.0325, -0.0325, -0.0397, -0.065, -0.065, -0.0361, -0.0505, -0.0614, -0.0542, -0.0542, -0.0614, -0.0505, -0.0397, -0.0686, -0.0903, -0.0975, -0.1047, -0.1011, -0.1011, -0.1119, -0.1119, -0.1552, -0.1625, -0.1444, -0.1733, -0.1697, -0.1444, -0.0614, -0.0325, 0.0072, 0.0433, 0.0217, 0.0108, 0.0144, 0.0289, 0.0397, 0.0253, 0.0542, 0.0072, 0.0144, 0.0253, 0.0253, 0.0289, 0.0253, 0.0181, 0.0108, 0.0036, -0.0036, -0.0036, 0.0, 0.0253, 0.0253, 0.0325, 0.0325, 0.0614, 0.1011, 0.093]}
{"unit_id": "orig:1944385898784891123", "ticker_idx": 1, "ts": "2025-07-13T13:18:17+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "adoption of TSMC (2330) 2nm process", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish MediaTek on cloud ASIC wins: Meta MTIA v4 ($3-4B/yr), AI glasses ($2B), Google TPU v7e ($1.5-2B); TSMC 2nm as process node beneficiary.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "MediaTek Industry Note:\n\nEuropean institutional investors point out that MediaTek views cloud ASICs as a mid-to-long-term growth driver and is actively expanding its collaborations with key clients. According to their report, MediaTek is highly likely to cooperate with Meta on the development of the next-generation MTIA v4 ASIC, with mass production expected in 2028 and adoption of TSMC (2330) 2nm process. Previous versions, MTIA v1/v2/v3, were all designed by Broadcom. Compared to Google's TPU, the MTIA project requires more front-end design services and relies on SerDes IP technology support. If MediaTek successfully secures this project and achieves mass production, it will help establish its technological and market position in the cloud ASIC field. This report estimates that if the MTIA v4 project materializes, it will bring annual revenues of $3-4 billion to MediaTek from 2028 to 2029, which accounts for over 10% of its total revenue growth.\n\nPotential Collaboration on AI Glasses ASIC\nFurthermore, European institutional investors suggest that Meta may also collaborate with MediaTek on the development of an AI glasses-specific ASIC, potentially replacing the current Qualcomm solution. Meta is a leader in the AI glasses market, with shipments of approximately 2 million units in 2024, expected to reach 10 million units in 2025. This project aims to enable large language model execution on devices by adopting the N3 advanced process and a four-layer stack VHM (Very High Bandwidth Mobile Memory). This ASIC is scheduled for mass production in mid-2027, and is expected to contribute approximately $2 billion in revenue in 2028. (Calculated at $100 per chip, this would imply shipments of 20 million units.) European institutional investors also anticipate that this project could generate $200 million in NRE (Non-Recurring Engineering) revenue between 2026 and 2027.\n\nGoogle TPU Project and Short-Term Outlook\nRegarding the collaboration with Google, European institutional investors expect that the TPU v7e will undergo Tape Out (design completion) in September 2025. Although the mass production timeline remains uncertain, it is likely to fall between late 2026 and early 2027, and Google's commitment to its self-developed computing architecture remains unchanged. If TPU v7e successfully reaches mass production, it is estimated to contribute $1.5-2 billion in revenue in 2027.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004566204901135684, "ret_signed_m1d": 0.004566204901135684, "alpha_spy_m1d": 0.006470688150460702, "alpha_c_m1d": -0.0030391839436145762, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013698614703407275, "ret_signed_p1d": 0.013698614703407275, "alpha_spy_p1d": 0.017971781754034355, "alpha_c_p1d": -0.005507821954384484, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.05022825391249364, "ret_signed_p1w": 0.05022825391249364, "alpha_spy_p1w": 0.04389022192777525, "alpha_c_p1w": 0.03203819375926997, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.07762548331930841, "ret_signed_p1m": 0.07762548331930841, "alpha_spy_p1m": 0.04900858324704793, "alpha_c_p1m": 0.025824078545997553, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.32032881091761123, "ret_signed_p3m": 0.32032881091761123, "alpha_spy_p3m": 0.24316807901362547, "alpha_c_p3m": 0.11042426318798904, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.5685169021912353, "ret_signed_p6m": 0.5685169021912353, "alpha_spy_p6m": 0.45493366700302085, "alpha_c_p6m": 0.205744950663586, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.2225, -0.2298, -0.227, -0.2407, -0.2579, -0.2061, -0.2143, -0.1925, -0.1843, -0.1797, -0.1743, -0.1743, -0.1361, -0.147, -0.1634, -0.1561, -0.1652, -0.137, -0.1297, -0.1188, -0.0915, -0.097, -0.0924, -0.1052, -0.1106, -0.0997, -0.107, -0.1052, -0.1124, -0.1224, -0.1206, -0.1206, -0.1206, -0.1397, -0.1361, -0.0997, -0.0924, -0.0952, -0.0861, -0.0497, -0.0315, -0.0457, -0.0594, -0.0639, -0.0457, -0.0365, -0.0548, -0.0365, -0.0685, -0.0411, -0.0228, -0.0183, -0.0137, -0.032, -0.0091, -0.0091, -0.0046, -0.0091, -0.0137, -0.0137, -0.0046, 0.0046, 0.0046, 0.0, 0.0137, 0.032, 0.032, 0.0548, 0.0502, 0.032, 0.0457, 0.0457, 0.0457, 0.0457, 0.0365, 0.0548, 0.0594, 0.0594, 0.0365, 0.0502, 0.0274, 0.0776, 0.0731, 0.0776, 0.0776, 0.0959, 0.0731, 0.0776, 0.0776, 0.0822, 0.0365, 0.0502, 0.0365, 0.0685, 0.0731, 0.0868, 0.0594, 0.0594, 0.0639, 0.0594, 0.0594, 0.0594, 0.0776, 0.0776, 0.0959, 0.1187, 0.1324, 0.1507, 0.1461, 0.1736, 0.1599, 0.1782, 0.1599, 0.1874, 0.2286, 0.2286, 0.2103, 0.192, 0.192, 0.1965, 0.2149, 0.2516, 0.2837, 0.2837, 0.3157, 0.2974, 0.3203, 0.3203, 0.2974, 0.3066, 0.3433, 0.3616, 0.3295, 0.357, 0.357, 0.3387, 0.3295, 0.3295, 0.357, 0.3524, 0.3799, 0.3799, 0.3753, 0.3845, 0.3799, 0.3387, 0.3433, 0.3387, 0.3524, 0.3433, 0.3524, 0.3387, 0.3112, 0.3249, 0.2882, 0.2791, 0.3341, 0.2699, 0.2607, 0.2974, 0.3203, 0.3157, 0.3203, 0.2928, 0.3112, 0.3295, 0.3249, 0.3387, 0.3708, 0.357, 0.3799, 0.3523, 0.3615, 0.3339, 0.3201, 0.3155, 0.3155, 0.3155, 0.3477, 0.3707, 0.3753, 0.3753, 0.3891, 0.4075, 0.3983, 0.4259, 0.4259, 0.4581, 0.5363, 0.5685]}
{"unit_id": "thread:1946073863693169003", "ticker_idx": 0, "ts": "2025-07-18T05:05:40+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Competitors can't even see their taillights... the AI server market continues to boom, providing a critical growth engine for ODMs and component suppliers in the second half of the year", "tweet_type": "analysis", "timeline": "quarters", "summary": "GB300 production underway with smooth ramp, robust GB200 demand, and booming AI server market into H2 2025; bullish NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: GB300 production underway: Nvidia's strategic shift brings relief to ODMs\n\nNvidia has begun limited production of its next-generation AI server platform, the GB300, with volume shipments expected to ramp up starting in September 2025, according to sources in the supply chain. Industry insiders expect a smooth production trajectory into the second half of the year, thanks to a strategic shift that's easing the burden on manufacturers.\n\nIn a key move, Nvidia has chosen to reuse the motherboard design from its current GB200 platform—known as the Bianca board—for the GB300. This reuse has significantly shortened the learning curve for suppliers, many of whom had struggled to keep pace with Nvidia's blistering product update cycle in the past.\n\n\"There are no major issues with the GB300 at this stage. Shipments should proceed smoothly in the second half,\" noted one ODM representative.\n\nCurrently, demand for the GB200 remains robust, though its reliance on liquid cooling has led to widespread reports of coolant leaks. Despite the unresolved issue, enterprise customers remain eager to deploy the systems, opting for temporary solutions such as leak testing and localized shutdowns. The urgency of staying competitive in AI development has pushed data center operators to prioritize speed over technical perfection.\n\n\"Nvidia's upgrade pace in AI servers is like a military blitz,\" said one industry executive. \"Competitors can't even see their taillights, but the supply chain is feeling the strain.\"\n\nNvidia's rapid generational shifts—from Hopper to the current Blackwell architecture—have consistently introduced sweeping changes. The GB200, for example, dramatically compressed server layout, leading to repeated delays in mass production. However, with the GB300 reusing existing infrastructure, suppliers no longer face the same steep ramp-up challenges.\n\nODM firms are now actively testing the GB300, and early production results are promising. The transition is expected to occur seamlessly, with steady shipments projected through the third quarter, and volume is expected to pick up further in the fourth quarter. Compute board supplier Wistron has confirmed that revenue this quarter will not grow sequentially due to the generational overlap—an indirect confirmation that GB300 production is underway.\n\nLooking beyond: The Rubin series will bring new challenges\nWhile the supply chain is momentarily relieved, preparations are already underway for the next architectural leap: Nvidia's Rubin platform. Unlike the GB300, Rubin will introduce chiplet-based GPUs—designed in partnership with TSMC—and feature a completely redesigned Kyber rack (VR300 NVL 576), replacing the current Oberon rack.\n\nThe new design will increase GPU density, thereby demanding more robust thermal solutions. Liquid cooling, which has become more prominent with the GB200, will become essential with Rubin.\n\nCurrently, the most vulnerable component in the GB200's liquid cooling systems is the quick-connect fittings, which are prone to leaks, even after stress testing at the factory. ODMs note that real-world deployment scenarios vary widely in water pressure and plumbing design, making leaks difficult to fully eliminate. Post-shipment servicing has consumed considerable manpower.\n\nDespite these challenges, ODMs remain confident about the GB300 rollout. Even as the US tariffs on Chinese imports threaten consumer electronics, the AI server market continues to boom, providing a critical growth engine for ODMs and component suppliers in the second half of the year.\n\n$NVDA\n\n---\n\nhttps://t.co/iTGxxTYcM9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0034220029148854803, "ret_signed_m1d": 0.0034220029148854803, "alpha_spy_m1d": 0.0026890531909407045, "alpha_c_m1d": -0.0014003334486645258, "bench_spy_m1d": 0.0007329497239447758, "bench_c_m1d": 0.004822336363550006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0059741431161833836, "ret_signed_p1d": -0.0059741431161833836, "alpha_spy_p1d": -0.00787031828184992, "alpha_c_p1d": -0.006663033247012451, "bench_spy_p1d": 0.001896175165666536, "bench_c_p1d": 0.0006888901308290674, "ret_p1w": 0.0063221059051437045, "ret_signed_p1w": 0.0063221059051437045, "alpha_spy_p1w": -0.008847295420187917, "alpha_c_p1w": 0.0160359523542436, "bench_spy_p1w": 0.015169401325331622, "bench_c_p1w": -0.009713846449099894, "ret_p1m": 0.05568103912611866, "ret_signed_p1m": 0.05568103912611866, "alpha_spy_p1m": 0.030632460242367765, "alpha_c_p1m": 0.03311901739755596, "bench_spy_p1m": 0.025048578883750894, "bench_c_p1m": 0.022562021728562698, "ret_p3m": 0.04309569153731663, "ret_signed_p3m": 0.04309569153731663, "alpha_spy_p3m": -0.01973956945477151, "alpha_c_p3m": -0.1341975069775574, "bench_spy_p3m": 0.06283526099208814, "bench_c_p3m": 0.17729319851487402, "ret_p6m": 0.07279574513473941, "ret_signed_p6m": 0.07279574513473941, "alpha_spy_p6m": -0.04124085507703423, "alpha_c_p6m": -0.27705636440016224, "bench_spy_p6m": 0.11403660021177364, "bench_c_p6m": 0.34985210953490165, "price_path": [-0.4265, -0.4043, -0.3827, -0.3562, -0.3694, -0.3677, -0.3683, -0.3527, -0.3359, -0.3399, -0.3415, -0.3211, -0.3193, -0.3235, -0.2866, -0.2464, -0.2151, -0.218, -0.2147, -0.2137, -0.2206, -0.2356, -0.2296, -0.2386, -0.2386, -0.2141, -0.2181, -0.1927, -0.2163, -0.2032, -0.181, -0.1769, -0.1881, -0.1781, -0.1728, -0.1651, -0.1716, -0.159, -0.1766, -0.1608, -0.1641, -0.1562, -0.1562, -0.1657, -0.1638, -0.1422, -0.105, -0.1009, -0.085, -0.0836, -0.1108, -0.0879, -0.0758, -0.0758, -0.0822, -0.072, -0.0553, -0.0482, -0.0434, -0.0484, -0.0099, -0.006, 0.0034, 0.0, -0.006, -0.0312, -0.0095, 0.0077, 0.0063, 0.0252, 0.018, 0.0398, 0.0317, 0.0076, 0.044, 0.0339, 0.0407, 0.0485, 0.0597, 0.056, 0.0624, 0.0532, 0.0557, 0.0466, 0.0557, 0.0187, 0.0173, 0.0149, 0.0324, 0.0429, 0.0543, 0.0533, 0.045, 0.0103, 0.0103, -0.0095, -0.0104, -0.0044, -0.0313, -0.0238, -0.0096, 0.0285, 0.0277, 0.0314, 0.031, 0.0144, -0.0122, 0.0223, 0.0248, 0.065, 0.035, 0.0265, 0.0307, 0.0336, 0.0548, 0.0822, 0.0861, 0.0956, 0.0883, 0.0762, 0.0733, 0.0969, 0.117, 0.0624, 0.0923, 0.0443, 0.0431, 0.0546, 0.0628, 0.0594, 0.0508, 0.0457, 0.0566, 0.0804, 0.1107, 0.1661, 0.2009, 0.1769, 0.1745, 0.2, 0.1525, 0.1323, 0.0909, 0.0914, 0.1546, 0.1204, 0.1241, 0.0839, 0.1031, 0.0824, 0.052, 0.0819, 0.0478, 0.0376, 0.0589, 0.0314, 0.0456, 0.0456, 0.0267, 0.0436, 0.0526, 0.0417, 0.0637, 0.0581, 0.0763, 0.073, 0.0661, 0.0495, 0.0153, 0.0226, 0.0309, -0.0084, 0.0101, 0.0499, 0.0655, 0.0976, 0.0941, 0.0941, 0.1052, 0.0918, 0.0879, 0.0818, 0.0818, 0.0955, 0.0912, 0.0861, 0.097, 0.0734, 0.0723, 0.0728]}
{"unit_id": "thread:1963165479981502733", "ticker_idx": 0, "ts": "2025-09-03T09:01:38+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I maintain my view of Samsung's strength. Samsung has a higher chance of seizing leadership in HBM compared to other memory makers. I believe this is Samsung's true strength.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Samsung: in-house design capabilities give it a structural edge in HBM leadership over Micron and Hynix.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Myron’s view is truly fascinating.\n\nNVIDIA designing its own HBM base die could be an example of this.\n\nThe more parts of HBM that get outsourced, the more memory companies risk ending up producing just the “shell” of HBM.\n\nI expect memory makers and fabless companies will wrestle for control over HBM.\n\nAlso, I maintain my view of Samsung’s strength.\n\nUnlike Micron and SK Hynix, Samsung possesses its own design capabilities.\n\nThis means Samsung has a higher chance of seizing leadership in HBM compared to other memory makers.\n\nI believe this is Samsung’s true strength.\n\n---\n\nhttps://t.co/jaaPZKkAmH\n\nI really enjoyed reading this.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010028639022655161, "ret_signed_m1d": -0.010028639022655161, "alpha_spy_m1d": -0.0046382926134879154, "alpha_c_m1d": -0.0040978347870296306, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": -0.005930804235625531, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004298117531614576, "ret_signed_p1d": 0.004298117531614576, "alpha_spy_p1d": -0.004059242910199989, "alpha_c_p1d": -0.0006378882629545224, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.004936005794569098, "ret_p1w": 0.0401146692972878, "ret_signed_p1w": 0.0401146692972878, "alpha_spy_p1w": 0.02695711425453995, "alpha_c_p1w": 0.006443350702212802, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.033671318595074995, "ret_p1m": 0.29155346943340743, "ret_signed_p1m": 0.29155346943340743, "alpha_spy_p1m": 0.24909017272607592, "alpha_c_p1m": 0.19521606287821913, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.0963374065551883, "ret_p3m": 0.4505693276497549, "ret_signed_p3m": 0.4505693276497549, "alpha_spy_p3m": 0.39089295114361033, "alpha_c_p3m": 0.3535039654645702, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.09706536218518469, "ret_p6m": 2.1523940606348493, "ret_signed_p6m": 2.1523940606348493, "alpha_spy_p6m": 2.0754786919611243, "alpha_c_p6m": 2.070279047631538, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.08211501300331125, "price_path": [-0.1585, -0.1485, -0.157, -0.1471, -0.1528, -0.1698, -0.1855, -0.1727, -0.1485, -0.157, -0.1528, -0.1741, -0.1385, -0.1271, -0.1428, -0.1289, -0.1433, -0.1375, -0.1289, -0.086, -0.0931, -0.116, -0.1203, -0.1347, -0.1261, -0.1032, -0.1046, -0.0874, -0.0731, -0.0444, -0.0387, -0.0287, -0.0544, -0.0487, -0.0544, -0.0559, 0.0086, 0.0115, 0.0401, 0.0229, -0.0129, -0.0014, 0.0014, -0.0143, 0.01, 0.0287, 0.0172, 0.0186, 0.0301, 0.0258, 0.0258, 0.0029, 0.0029, 0.01, 0.0115, 0.0229, 0.0244, 0.0072, 0.0115, -0.0029, -0.0014, -0.0315, -0.01, 0.0, 0.0043, -0.0043, 0.0043, 0.0244, 0.0401, 0.0516, 0.0802, 0.096, 0.1375, 0.1203, 0.1504, 0.1504, 0.1963, 0.2135, 0.2235, 0.2335, 0.1934, 0.2117, 0.2074, 0.2376, 0.2916, 0.2916, 0.2916, 0.2916, 0.2916, 0.2916, 0.3585, 0.3426, 0.3182, 0.3671, 0.406, 0.4088, 0.4117, 0.4031, 0.4189, 0.3887, 0.4218, 0.4678, 0.4319, 0.4463, 0.4981, 0.547, 0.5988, 0.5096, 0.4477, 0.4275, 0.4088, 0.4477, 0.4894, 0.4837, 0.4794, 0.3988, 0.4477, 0.4074, 0.3887, 0.4477, 0.3642, 0.3916, 0.429, 0.4794, 0.4894, 0.4463, 0.4506, 0.488, 0.5038, 0.5124, 0.5599, 0.5758, 0.5599, 0.5542, 0.5441, 0.5671, 0.5081, 0.4794, 0.5527, 0.5484, 0.5297, 0.5902, 0.6045, 0.5988, 0.5988, 0.6837, 0.728, 0.7338, 0.7338, 0.7338, 0.8582, 0.997, 1.0086, 1.0389, 1.0071, 1.01, 1.0071, 0.9898, 1.0288, 1.0809, 1.1532, 1.159, 1.0997, 1.1618, 1.2023, 1.1994, 1.1994, 1.3065, 1.3484, 1.3238, 1.3209, 1.1749, 1.4221, 1.4453, 1.3036, 1.2934, 1.4062, 1.3976, 1.4265, 1.5826, 1.6202, 1.6202, 1.6202, 1.6202, 1.7475, 1.7489, 1.7909, 1.8921, 1.9427, 2.1524]}
{"unit_id": "orig:1960853513078235519", "ticker_idx": 0, "ts": "2025-08-27T23:54:43+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reasons for showing strength in ASML are gradually emerging", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long ASML as foundry monopoly breaks down and Intel/Samsung challenge TSMC, restoring ASML's pricing power on EUV.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "It seems that reasons for showing strength in ASML are gradually emerging.\n\nUntil now, the foundry sector has been virtually monopolized by TSMC, so Intel and Samsung had no reason to expand their fabs.\n\nThe reason why semiconductor equipment companies’ stock prices have been relatively weak during the first half of this year is that investors were focused on the back-end process.\n\nSince TSMC monopolized the volume from Big Tech, naturally, market attention concentrated on CoWoS, and CoWoS was mistakenly perceived as an unrivaled technology unique to TSMC that no one else could challenge. In reality, the front-end process is the most important, and companies like Samsung, lacking customers, simply did not have the capacity to invest in the back-end, which inevitably put them at a disadvantage in packaging.\n\nAlthough ASML is a company with strengths in the front-end process, its performance has been sluggish because the foundry industry has transformed into a TSMC monopoly structure. In fact, TSMC has tried to rein in ASML by delaying the introduction of High-NA EUV and even succeeded in making the supply conditions of EUV machines favorable to itself.\n\nHowever, now with Intel receiving support from the U.S. government and Samsung joining forces with Elon Musk, I’m beginning to think that competitors in the foundry space are finally emerging.\n\nIn other words, TSMC can no longer strong-arm ASML when purchasing EUV lithography machines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014541150822964366, "ret_signed_m1d": -0.014541150822964366, "alpha_spy_m1d": -0.012267817361441447, "alpha_c_m1d": -0.01148401081313044, "bench_spy_m1d": -0.0022733334615229195, "bench_c_m1d": -0.003057140009833925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008936654064967908, "ret_signed_p1d": -0.008936654064967908, "alpha_spy_p1d": -0.012478016797110691, "alpha_c_p1d": -0.013035295012687409, "bench_spy_p1d": 0.003541362732142783, "bench_c_p1d": 0.004098640947719501, "ret_p1w": -0.052862725751519846, "ret_signed_p1w": -0.052862725751519846, "alpha_spy_p1w": -0.04839335330844885, "alpha_c_p1w": -0.015135133319725713, "bench_spy_p1w": -0.004469372443070996, "bench_c_p1w": -0.03772759243179413, "ret_p1m": 0.22947598545494552, "ret_signed_p1m": 0.22947598545494552, "alpha_spy_p1m": 0.20899380497385778, "alpha_c_p1m": 0.1508627706417518, "bench_spy_p1m": 0.02048218048108774, "bench_c_p1m": 0.07861321481319372, "ret_p3m": 0.30311394803202707, "ret_signed_p3m": 0.30311394803202707, "alpha_spy_p3m": 0.2660696166016736, "alpha_c_p3m": 0.16382751861338463, "bench_spy_p3m": 0.03704433143035346, "bench_c_p3m": 0.13928642941864244, "ret_p6m": 0.8813240074071071, "ret_signed_p6m": 0.8813240074071071, "alpha_spy_p6m": 0.8167185776412313, "alpha_c_p6m": 0.49895350108461956, "bench_spy_p6m": 0.0646054297658758, "bench_c_p6m": 0.3823705063224876, "price_path": [-0.0122, -0.0238, -0.0134, -0.0144, -0.0045, 0.0085, 0.0274, 0.0341, 0.0427, 0.0221, 0.004, 0.0123, 0.0061, 0.0046, -0.0134, -0.0174, 0.0102, 0.0445, 0.0492, 0.0233, 0.031, 0.0236, 0.0094, 0.0236, 0.0229, -0.0039, 0.0146, 0.0256, 0.027, 0.0424, 0.0332, 0.0383, 0.0667, -0.0546, -0.0178, -0.043, -0.0572, -0.09, -0.0836, -0.0696, -0.0862, -0.0483, -0.0563, -0.0495, -0.0713, -0.0988, -0.0875, -0.0956, -0.1019, -0.075, -0.0656, -0.0503, -0.0398, -0.0297, -0.0262, -0.0356, -0.0326, -0.0271, -0.0333, -0.0389, -0.0212, -0.0204, -0.0145, 0.0, -0.0089, -0.0357, -0.0365, -0.0651, -0.0529, -0.0191, 0.0, 0.0248, 0.0348, 0.0303, 0.0391, 0.0448, 0.1048, 0.1101, 0.1163, 0.2027, 0.2022, 0.2281, 0.2417, 0.2239, 0.2295, 0.2311, 0.2504, 0.2543, 0.2745, 0.3293, 0.3331, 0.3591, 0.3251, 0.2902, 0.2872, 0.2328, 0.2784, 0.2823, 0.3223, 0.329, 0.3264, 0.3617, 0.3478, 0.3179, 0.3446, 0.3547, 0.3755, 0.3769, 0.3951, 0.4225, 0.3931, 0.4058, 0.3929, 0.3756, 0.3585, 0.3213, 0.3454, 0.3458, 0.3532, 0.3376, 0.3263, 0.3298, 0.3134, 0.3453, 0.3502, 0.2655, 0.3031, 0.3054, 0.3797, 0.3614, 0.3708, 0.4061, 0.425, 0.4618, 0.4526, 0.4439, 0.4615, 0.4459, 0.4354, 0.4277, 0.404, 0.4127, 0.3787, 0.3262, 0.3546, 0.3681, 0.3596, 0.3705, 0.3641, 0.3641, 0.3641, 0.3759, 0.3935, 0.3981, 0.3981, 0.4966, 0.5981, 0.6099, 0.596, 0.5371, 0.6412, 0.6485, 0.673, 0.6451, 0.7441, 0.7711, 0.7001, 0.7298, 0.7523, 0.785, 0.7878, 0.7875, 0.8475, 0.8123, 0.8087, 0.8445, 0.8585, 0.8063, 0.7307, 0.7444, 0.8114, 0.8306, 0.8126, 0.8351, 0.7926, 0.8087, 0.8157, 0.8221, 0.8914, 0.8813]}
{"unit_id": "orig:1960853513078235519", "ticker_idx": 1, "ts": "2025-08-27T23:54:43+00:00", "symbol": "TSMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC can no longer strong-arm ASML when purchasing EUV lithography machines", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long ASML as foundry monopoly breaks down and Intel/Samsung challenge TSMC, restoring ASML's pricing power on EUV.", "resolved_tickers": ["TSM"], "resolution_method": "exchange_ticker", "resolution_reason": "TSMC listed on NYSE as TSM (ADR); also 2330.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "It seems that reasons for showing strength in ASML are gradually emerging.\n\nUntil now, the foundry sector has been virtually monopolized by TSMC, so Intel and Samsung had no reason to expand their fabs.\n\nThe reason why semiconductor equipment companies’ stock prices have been relatively weak during the first half of this year is that investors were focused on the back-end process.\n\nSince TSMC monopolized the volume from Big Tech, naturally, market attention concentrated on CoWoS, and CoWoS was mistakenly perceived as an unrivaled technology unique to TSMC that no one else could challenge. In reality, the front-end process is the most important, and companies like Samsung, lacking customers, simply did not have the capacity to invest in the back-end, which inevitably put them at a disadvantage in packaging.\n\nAlthough ASML is a company with strengths in the front-end process, its performance has been sluggish because the foundry industry has transformed into a TSMC monopoly structure. In fact, TSMC has tried to rein in ASML by delaying the introduction of High-NA EUV and even succeeded in making the supply conditions of EUV machines favorable to itself.\n\nHowever, now with Intel receiving support from the U.S. government and Samsung joining forces with Elon Musk, I’m beginning to think that competitors in the foundry space are finally emerging.\n\nIn other words, TSMC can no longer strong-arm ASML when purchasing EUV lithography machines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0023820835209913405, "ret_signed_m1d": 0.0023820835209913405, "alpha_spy_m1d": 0.00010875005946842098, "alpha_c_m1d": -0.0006750564888425847, "bench_spy_m1d": -0.0022733334615229195, "bench_c_m1d": -0.003057140009833925, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004262587755470615, "ret_signed_p1d": 0.004262587755470615, "alpha_spy_p1d": 0.007803950487613398, "alpha_c_p1d": 0.008361228703190116, "bench_spy_p1d": 0.003541362732142783, "bench_c_p1d": 0.004098640947719501, "ret_p1w": -0.03301431141165556, "ret_signed_p1w": 0.03301431141165556, "alpha_spy_p1w": 0.028544938968584566, "alpha_c_p1w": -0.004713281020138571, "bench_spy_p1w": -0.004469372443070996, "bench_c_p1w": -0.03772759243179413, "ret_p1m": 0.15981790317731637, "ret_signed_p1m": -0.15981790317731637, "alpha_spy_p1m": -0.13933572269622863, "alpha_c_p1m": -0.08120468836412265, "bench_spy_p1m": 0.02048218048108774, "bench_c_p1m": 0.07861321481319372, "ret_p3m": 0.19327178792111144, "ret_signed_p3m": -0.19327178792111144, "alpha_spy_p3m": -0.15622745649075798, "alpha_c_p3m": -0.053985358502468994, "bench_spy_p3m": 0.03704433143035346, "bench_c_p3m": 0.13928642941864244, "ret_p6m": 0.5149105667022622, "ret_signed_p6m": -0.5149105667022622, "alpha_spy_p6m": -0.4503051369363864, "alpha_c_p6m": -0.13254006037977462, "bench_spy_p6m": 0.0646054297658758, "bench_c_p6m": 0.3823705063224876, "price_path": [-0.1951, -0.1887, -0.1772, -0.1572, -0.1533, -0.1457, -0.1381, -0.1154, -0.1085, -0.0997, -0.1178, -0.0987, -0.1061, -0.1078, -0.1078, -0.1245, -0.1211, -0.0802, -0.0692, -0.0639, -0.0448, -0.0535, -0.0611, -0.0238, -0.0188, -0.0188, -0.0423, -0.0478, -0.0311, -0.0398, -0.0372, -0.0444, -0.0098, -0.0072, 0.0264, 0.0046, -0.0018, -0.0196, 0.0043, 0.0097, 0.0264, 0.0145, 0.0085, 0.0151, 0.0097, -0.0171, -0.0012, -0.0285, -0.0331, 0.0139, 0.0106, 0.0117, 0.0209, 0.009, 0.0071, -0.0017, 0.0089, -0.0275, -0.0447, -0.05, -0.0263, -0.0155, -0.0024, 0.0, -0.0043, -0.0352, -0.0352, -0.0456, -0.033, -0.0171, 0.0172, 0.033, 0.0486, 0.0884, 0.082, 0.0837, 0.0923, 0.0986, 0.1017, 0.1262, 0.1104, 0.1429, 0.1852, 0.1768, 0.1598, 0.146, 0.1454, 0.1708, 0.2093, 0.2078, 0.2249, 0.2677, 0.2326, 0.2766, 0.2572, 0.1766, 0.2698, 0.2406, 0.2774, 0.257, 0.237, 0.248, 0.2346, 0.211, 0.2188, 0.2365, 0.2503, 0.2641, 0.279, 0.2712, 0.2595, 0.278, 0.2327, 0.231, 0.2126, 0.2011, 0.2378, 0.2206, 0.2183, 0.183, 0.194, 0.1822, 0.1651, 0.1838, 0.1633, 0.1531, 0.1933, 0.1934, 0.2156, 0.2156, 0.2221, 0.206, 0.2245, 0.2386, 0.228, 0.2355, 0.2655, 0.272, 0.3002, 0.2814, 0.2276, 0.2095, 0.2059, 0.1642, 0.1967, 0.2146, 0.2328, 0.2482, 0.256, 0.256, 0.273, 0.2649, 0.2593, 0.2774, 0.2774, 0.3435, 0.3546, 0.3764, 0.3396, 0.3368, 0.3604, 0.3946, 0.3923, 0.375, 0.4361, 0.4393, 0.4393, 0.3752, 0.3709, 0.3761, 0.4076, 0.3986, 0.4222, 0.4389, 0.4273, 0.3895, 0.4349, 0.4113, 0.3693, 0.3902, 0.4664, 0.494, 0.5213, 0.5725, 0.5473, 0.54, 0.54, 0.5309, 0.5228, 0.5149]}
{"unit_id": "orig:1943955440619884726", "ticker_idx": 0, "ts": "2025-07-12T08:47:48+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes that GOOGL and META are leaders in Diffusion model-driven video and image generation tools...bringing upside potential to mid-to-long-term valuations", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Buy-aligned thesis: GOOGL and META diffusion model video tools to boost engagement and ad monetization, driving mid-to-long-term valuation upside.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Diffusion Models - Google and Meta: Video Generation Tools Drive Growth in Both Engagement and Monetization\nMorgan Stanley (B. Nowak, 25/07/09)\n\nMorgan Stanley believes that GOOGL and META are leaders in Diffusion model-driven video and image generation tools, which can enhance content creation efficiency, improve ad personalization and relevance, and thus significantly boost user engagement time and ad monetization capabilities, bringing upside potential to mid-to-long-term valuations.\n\nGoogle Strengthens YouTube Creation and Ad Tools with Veo 3 and Imagen 4\nGOOGL's YouTube has embedded Veo 3 and Imagen 4 models into its creator and advertiser tools, including features like Dream Screen and Flow, supporting text-to-video/image generation. \nApproximately 57% of advertisers have adopted these tools, which is expected to enhance ad content diversity and \npersonalization.\n\nMeta Launches Emu and MovieGen, Accelerating Innovation in Video Creation Experience\nMETA has introduced multimodal Diffusion models like Emu and MovieGen, which support text-to-video generation with sound effects and music. New features such as Video Generation and Video Highlights have been added to improve automatic ad generation efficiency and viewability. Surveys show that 56% of advertisers have already adopted these tools.\n\nAI Content Generation Will Drive Higher Personalization and Engagement\nBased on their leading data infrastructure and algorithmic capabilities, GOOGL and META can enhance recommendation accuracy and user experience through AI-generated content. For example, Meta's AI recommendations have already increased user time spent by 6-7%, and AI creation tools are expected to further drive content diversity and ad suitability in the future.\n\nAd Monetization Has Quantifiable Growth Potential\nOn the YouTube platform, every 1% increase in engagement or monetization capability can bring approximately $1 billion in incremental revenue. Although this accounts for only 15% of GOOGL's advertising revenue, its positive impact on valuation is significant. For META, due to higher average ad value per person, every 1% increase can boost ad revenue by approximately $5 billion and has about 2% upside elasticity on the stock price target.\n\nSmall and Medium-Sized Advertisers Benefit from AI Automation Empowerment\nThrough Diffusion model-driven video generation tools, small and medium-sized advertisers, traditionally limited by production resources, can efficiently achieve customized and automated video advertising, improving ad quality and return on investment. This is an important driver for the democratization of advertising technology in the future.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00754568224020391, "ret_signed_m1d": -0.00754568224020391, "alpha_spy_m1d": -0.005641198990878893, "alpha_c_m1d": 0.0021693995151328416, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": -0.009715081755336752, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0024235267064269816, "ret_signed_p1d": 0.0024235267064269816, "alpha_spy_p1d": 0.006696693757054062, "alpha_c_p1d": 0.012325536619593569, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": -0.009902009913166587, "ret_p1w": 0.04703681912540936, "ret_signed_p1w": 0.04703681912540936, "alpha_spy_p1w": 0.04069878714069097, "alpha_c_p1w": 0.04180563786757352, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.005231181257835837, "ret_p1m": 0.11996036036187796, "ret_signed_p1m": 0.11996036036187796, "alpha_spy_p1m": 0.09134346028961748, "alpha_c_p1m": 0.09006776235703562, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.02989259800484234, "ret_p3m": 0.3314938441057096, "ret_signed_p3m": 0.3314938441057096, "alpha_spy_p3m": 0.25433311220172383, "alpha_c_p3m": 0.25423291672611126, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.07726092737959833, "ret_p6m": 0.7340104473104894, "ret_signed_p6m": 0.7340104473104894, "alpha_spy_p6m": 0.6204272121222749, "alpha_c_p6m": 0.6305974285020761, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.10341301880841325, "price_path": [-0.1565, -0.1684, -0.1684, -0.1876, -0.1667, -0.1454, -0.1238, -0.109, -0.1165, -0.1189, -0.1264, -0.1127, -0.0976, -0.0967, -0.102, -0.1672, -0.1513, -0.1597, -0.1283, -0.1224, -0.0903, -0.098, -0.0858, -0.0838, -0.0979, -0.0727, -0.06, -0.0732, -0.0732, -0.0488, -0.0518, -0.0546, -0.0552, -0.0701, -0.0858, -0.0755, -0.0746, -0.0446, -0.0301, -0.0163, -0.0232, -0.0323, -0.0379, -0.0264, -0.0309, -0.0454, -0.0454, -0.0822, -0.0902, -0.0815, -0.0599, -0.0442, -0.0167, -0.0294, -0.0315, -0.0161, -0.0112, -0.0112, -0.0263, -0.0397, -0.0272, -0.0217, -0.0075, 0.0, 0.0024, 0.0078, 0.0111, 0.0193, 0.047, 0.0539, 0.0478, 0.0584, 0.064, 0.0607, 0.0782, 0.0825, 0.057, 0.0417, 0.0742, 0.0722, 0.08, 0.0824, 0.1094, 0.1071, 0.12, 0.1124, 0.1178, 0.123, 0.1208, 0.1102, 0.0978, 0.1002, 0.1351, 0.1483, 0.1409, 0.1428, 0.1657, 0.1727, 0.1727, 0.1641, 0.2704, 0.2795, 0.2943, 0.2902, 0.321, 0.3185, 0.3251, 0.3275, 0.3871, 0.3846, 0.3756, 0.3894, 0.4042, 0.3921, 0.3873, 0.3624, 0.355, 0.3591, 0.3454, 0.3401, 0.3501, 0.3544, 0.3526, 0.3806, 0.3548, 0.3485, 0.3315, 0.3042, 0.3459, 0.3531, 0.3839, 0.3862, 0.3964, 0.4143, 0.3807, 0.3875, 0.3952, 0.4329, 0.4844, 0.4745, 0.5136, 0.5517, 0.5501, 0.5641, 0.53, 0.5673, 0.5698, 0.5371, 0.5992, 0.6059, 0.5806, 0.5357, 0.5238, 0.5712, 0.5672, 0.6142, 0.5957, 0.6519, 0.7563, 0.783, 0.7638, 0.7638, 0.7651, 0.7359, 0.741, 0.762, 0.751, 0.7711, 0.7306, 0.7491, 0.7664, 0.7235, 0.7062, 0.7003, 0.6911, 0.6368, 0.6685, 0.6944, 0.7089, 0.7341, 0.7326, 0.7326, 0.7294, 0.7297, 0.7313, 0.7266, 0.7266, 0.7385, 0.7461, 0.734]}
{"unit_id": "orig:1943955440619884726", "ticker_idx": 1, "ts": "2025-07-12T08:47:48+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes that GOOGL and META are leaders in Diffusion model-driven video and image generation tools...bringing upside potential to mid-to-long-term valuations", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Buy-aligned thesis: GOOGL and META diffusion model video tools to boost engagement and ad monetization, driving mid-to-long-term valuation upside.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Diffusion Models - Google and Meta: Video Generation Tools Drive Growth in Both Engagement and Monetization\nMorgan Stanley (B. Nowak, 25/07/09)\n\nMorgan Stanley believes that GOOGL and META are leaders in Diffusion model-driven video and image generation tools, which can enhance content creation efficiency, improve ad personalization and relevance, and thus significantly boost user engagement time and ad monetization capabilities, bringing upside potential to mid-to-long-term valuations.\n\nGoogle Strengthens YouTube Creation and Ad Tools with Veo 3 and Imagen 4\nGOOGL's YouTube has embedded Veo 3 and Imagen 4 models into its creator and advertiser tools, including features like Dream Screen and Flow, supporting text-to-video/image generation. \nApproximately 57% of advertisers have adopted these tools, which is expected to enhance ad content diversity and \npersonalization.\n\nMeta Launches Emu and MovieGen, Accelerating Innovation in Video Creation Experience\nMETA has introduced multimodal Diffusion models like Emu and MovieGen, which support text-to-video generation with sound effects and music. New features such as Video Generation and Video Highlights have been added to improve automatic ad generation efficiency and viewability. Surveys show that 56% of advertisers have already adopted these tools.\n\nAI Content Generation Will Drive Higher Personalization and Engagement\nBased on their leading data infrastructure and algorithmic capabilities, GOOGL and META can enhance recommendation accuracy and user experience through AI-generated content. For example, Meta's AI recommendations have already increased user time spent by 6-7%, and AI creation tools are expected to further drive content diversity and ad suitability in the future.\n\nAd Monetization Has Quantifiable Growth Potential\nOn the YouTube platform, every 1% increase in engagement or monetization capability can bring approximately $1 billion in incremental revenue. Although this accounts for only 15% of GOOGL's advertising revenue, its positive impact on valuation is significant. For META, due to higher average ad value per person, every 1% increase can boost ad revenue by approximately $5 billion and has about 2% upside elasticity on the stock price target.\n\nSmall and Medium-Sized Advertisers Benefit from AI Automation Empowerment\nThrough Diffusion model-driven video generation tools, small and medium-sized advertisers, traditionally limited by production resources, can efficiently achieve customized and automated video advertising, improving ad quality and return on investment. This is an important driver for the democratization of advertising technology in the future.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0047299369528435165, "ret_signed_m1d": -0.0047299369528435165, "alpha_spy_m1d": -0.002825453703518499, "alpha_c_m1d": 0.0049851448024932354, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": -0.009715081755336752, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01460631211558705, "ret_signed_p1d": -0.01460631211558705, "alpha_spy_p1d": -0.01033314506495997, "alpha_c_p1d": -0.004704302202420463, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": -0.009902009913166587, "ret_p1w": -0.011027609077419176, "ret_signed_p1w": -0.011027609077419176, "alpha_spy_p1w": -0.01736564106213756, "alpha_c_p1w": -0.016258790335255013, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.005231181257835837, "ret_p1m": 0.09582202056063371, "ret_signed_p1m": 0.09582202056063371, "alpha_spy_p1m": 0.06720512048837324, "alpha_c_p1m": 0.06592942255579137, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.02989259800484234, "ret_p3m": 0.018150561300787293, "ret_signed_p3m": 0.018150561300787293, "alpha_spy_p3m": -0.059010170603198464, "alpha_c_p3m": -0.05911036607881104, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.07726092737959833, "ret_p6m": -0.08227673436537619, "ret_signed_p6m": -0.08227673436537619, "alpha_spy_p6m": -0.19585996955359064, "alpha_c_p6m": -0.18568975317378944, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.10341301880841325, "price_path": [-0.3038, -0.3049, -0.3049, -0.3282, -0.3066, -0.2789, -0.261, -0.2415, -0.238, -0.2315, -0.2391, -0.2069, -0.1725, -0.1694, -0.186, -0.1728, -0.1711, -0.1788, -0.1137, -0.0907, -0.0861, -0.1076, -0.1125, -0.1123, -0.1169, -0.1192, -0.1177, -0.1309, -0.1309, -0.1097, -0.108, -0.1059, -0.1025, -0.0701, -0.0757, -0.0465, -0.0511, -0.0329, -0.038, -0.0264, -0.0379, -0.039, -0.0535, -0.0261, -0.0329, -0.0349, -0.0349, -0.0535, -0.0311, -0.0121, -0.017, 0.0072, 0.0176, 0.0238, -0.0024, -0.0102, -0.0026, -0.0026, -0.0036, -0.0003, 0.0165, 0.0088, -0.0047, 0.0, -0.0146, -0.025, -0.0271, -0.0231, -0.011, -0.0223, -0.0102, -0.0085, -0.0114, -0.0046, -0.029, -0.0357, 0.0729, 0.0404, 0.0769, 0.059, 0.0708, 0.0567, 0.0671, 0.0624, 0.0958, 0.0821, 0.0849, 0.0892, 0.0644, 0.0424, 0.0372, 0.0252, 0.047, 0.0449, 0.046, 0.0367, 0.0419, 0.0247, 0.0247, 0.0197, 0.0224, 0.0385, 0.0437, 0.0435, 0.0621, 0.0431, 0.0416, 0.0481, 0.0607, 0.0806, 0.076, 0.0823, 0.0797, 0.0621, 0.0485, 0.0558, 0.0395, 0.0324, 0.0319, 0.0194, -0.0043, 0.0092, -0.0137, -0.0066, -0.0102, -0.0036, 0.0182, -0.021, -0.0066, -0.0164, -0.004, -0.0116, -0.0049, 0.0163, 0.0178, 0.018, 0.0188, 0.0249, 0.0422, 0.043, 0.0434, -0.0749, -0.1001, -0.1148, -0.1292, -0.1173, -0.1409, -0.137, -0.1231, -0.1296, -0.1547, -0.1534, -0.154, -0.1644, -0.1704, -0.1806, -0.1822, -0.1751, -0.1491, -0.1169, -0.1205, -0.1205, -0.1006, -0.1104, -0.1018, -0.1122, -0.0818, -0.0653, -0.0744, -0.0881, -0.0976, -0.094, -0.1058, -0.1005, -0.0871, -0.0977, -0.077, -0.0848, -0.0811, -0.0763, -0.0726, -0.0726, -0.0786, -0.085, -0.0749, -0.083, -0.083, -0.0965, -0.0848, -0.0823]}
{"unit_id": "thread:1944960532412555629", "ticker_idx": 0, "ts": "2025-07-15T03:21:41+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If the RTX 6000D can be supplied smoothly and maintain Chinese clients' purchasing dependence, Nvidia's revenue from China is expected to rebound significantly in the second half of 2025, driving robust overall growth", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVDA China revenue expected to rebound significantly H2 2025 via RTX 6000D; TSMC benefits from 4nm utilization surge.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Jensen Huang's third visit to China in 2025; RTX 6000D aims for two million shipments\n\nNvidia CEO Jensen Huang made his third visit to China in 2025, demonstrating significant sincerity to revive the Chinese market. According to supply chain sources, Nvidia has confirmed it will supply the RTX 6000D in the third quarter of 2025, manufactured using TSMC's 4nm process. The shipment target is set at around two million units by year-end, aiming to fill a revenue gap exceeding US$10 billion.\n\nThe RTX 6000D and the Blackwell series have also driven TSMC's 4nm production capacity utilization to unprecedented levels, contributing significantly to its revenue.\n\nReportedly, to counteract the decline in revenue from China—from 13% down to single-digit percentages—Huang visited China again on July 14 under close scrutiny from US authorities. This marks his third appearance in China in 2025.\n\nSupply chain insiders revealed that besides showing his commitment to the Chinese market, Huang also presented multiple clients with Nvidia's solutions amid ongoing US-China tensions.\n\nImpact of US export bans on Nvidia\nHuang's active push to ship downgraded AI chips to China reflects the critical situation. Nvidia's fiscal year 2025 financial report showed that revenue from China (including Hong Kong) reached US$17.108 billion, accounting for 13.11% of total annual revenue.\n\nFollowing the US ban on H20 chips, Nvidia immediately recognized inventory losses up to US$5.5 billion. Its May financial report disclosed impairments of US$4.5 billion related to inventory and procurement obligations, along with approximately US$2.5 billion in lost potential revenue recognition.\n\nAs Nvidia nearly monopolizes the AI GPU market, it has been one of the hardest-hit companies under these restrictions. The US Department of Commerce first imposed export controls targeting AI chips to China in 2022, banning products like Nvidia's A100 and H100 from being exported to China. In response, Nvidia urgently launched downgraded versions, the A800 and H800, to continue supplying the Chinese market.\n\nA year later, the US government further blocked exports of the A800 and H800. Nvidia then shifted focus to developing downgraded H20 and L20 products, but even flagship gaming GPUs such as the RTX 4090 faced restrictions.\n\nIn 2025, the supply of the special edition RTX 5090D was also hindered, and the ban on H20 sales to China almost pushed Nvidia's revenue share in the Chinese market below 5%.\n\nThe new RTX 6000D: A tactical response\nDetermined not to abandon the Chinese market, Nvidia plans to launch the \"RTX 6000D\" in the third quarter of 2025, complying with US regulations. This modified version of the RTX 5090 uses TSMC's 4nm process, offers up to 1,100GB/s bidirectional bandwidth, and features next-generation GDDR7 graphics memory. It aims to deliver memory performance comparable to high-bandwidth memory (HBM).\n\nIndustry experts caution that supply constraints on GDDR7 could pose risks in the second half of 2025. The current shipment target for the RTX 6000D starts at a minimum of one million units, potentially reaching up to two million if obstacles are overcome.\n\nAnalysts note that while Chinese firms like Huawei and Cambricon accelerate AI chip development, they face limitations due to SMIC's manufacturing capacity, Huawei's process capabilities, high production costs, memory bandwidth challenges, system integration difficulties, and immature software ecosystems. Replacing Nvidia's CUDA ecosystem remains highly challenging.\n\nIf the RTX 6000D can be supplied smoothly and maintain Chinese clients' purchasing dependence, Nvidia's revenue from China is expected to rebound significantly in the second half of 2025, driving robust overall growth.\n\n---\n\nhttps://t.co/BHM5yrkHG7", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03884002840913159, "ret_signed_m1d": -0.03884002840913159, "alpha_spy_m1d": -0.04313153377910328, "alpha_c_m1d": -0.019995527465069607, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00392503050477, "ret_signed_p1d": 0.00392503050477, "alpha_spy_p1d": 0.0005818393862548898, "alpha_c_p1d": 0.009152005413437636, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.021499604475081924, "ret_signed_p1w": -0.021499604475081924, "alpha_spy_p1w": -0.032300987895897304, "alpha_c_p1w": -0.002827009559373761, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.06379609510335871, "ret_signed_p1m": 0.06379609510335871, "alpha_spy_p1m": 0.02722876118029305, "alpha_c_p1m": 0.027757517411518773, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.07305409607893898, "ret_signed_p3m": 0.07305409607893898, "alpha_spy_p3m": 0.020508977938374917, "alpha_c_p3m": -0.045721705615825003, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.10797419003219977, "ret_signed_p6m": 0.10797419003219977, "alpha_spy_p6m": -0.0067831132489872825, "alpha_c_p6m": -0.22014830980357902, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.4055, -0.4055, -0.4323, -0.4207, -0.3983, -0.3766, -0.3497, -0.3631, -0.3614, -0.362, -0.3462, -0.3293, -0.3333, -0.3349, -0.3143, -0.3125, -0.3167, -0.2795, -0.2389, -0.2072, -0.2102, -0.2069, -0.2059, -0.2128, -0.2279, -0.2219, -0.2309, -0.2309, -0.2063, -0.2103, -0.1846, -0.2084, -0.1953, -0.1728, -0.1687, -0.18, -0.1698, -0.1645, -0.1567, -0.1633, -0.1506, -0.1683, -0.1524, -0.1557, -0.1477, -0.1477, -0.1573, -0.1554, -0.1336, -0.096, -0.0919, -0.0759, -0.0745, -0.1019, -0.0788, -0.0665, -0.0665, -0.073, -0.0627, -0.0458, -0.0387, -0.0339, -0.0388, 0.0, 0.0039, 0.0135, 0.01, 0.004, -0.0215, 0.0005, 0.0178, 0.0164, 0.0354, 0.0282, 0.0502, 0.042, 0.0177, 0.0545, 0.0443, 0.0511, 0.059, 0.0703, 0.0665, 0.073, 0.0638, 0.0663, 0.0571, 0.0663, 0.0289, 0.0275, 0.0251, 0.0427, 0.0534, 0.0649, 0.0639, 0.0555, 0.0204, 0.0204, 0.0005, -0.0005, 0.0056, -0.0216, -0.014, 0.0004, 0.0388, 0.038, 0.0418, 0.0414, 0.0245, -0.0023, 0.0325, 0.035, 0.0757, 0.0453, 0.0368, 0.041, 0.0439, 0.0654, 0.0931, 0.097, 0.1066, 0.0992, 0.087, 0.0841, 0.1079, 0.1282, 0.0731, 0.1033, 0.0547, 0.0535, 0.0651, 0.0734, 0.07, 0.0613, 0.0562, 0.0672, 0.0912, 0.1219, 0.1777, 0.213, 0.1886, 0.1863, 0.212, 0.164, 0.1436, 0.1019, 0.1023, 0.1661, 0.1316, 0.1354, 0.0947, 0.1141, 0.0932, 0.0625, 0.0927, 0.0583, 0.048, 0.0695, 0.0418, 0.0561, 0.0561, 0.037, 0.0541, 0.0631, 0.0521, 0.0744, 0.0687, 0.0871, 0.0837, 0.0767, 0.06, 0.0254, 0.0329, 0.0412, 0.0015, 0.0203, 0.0604, 0.0762, 0.1086, 0.105, 0.105, 0.1163, 0.1028, 0.0988, 0.0927, 0.0927, 0.1065, 0.1022, 0.097, 0.108]}
{"unit_id": "thread:1944960532412555629", "ticker_idx": 1, "ts": "2025-07-15T03:21:41+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The RTX 6000D and the Blackwell series have also driven TSMC's 4nm production capacity utilization to unprecedented levels, contributing significantly to its revenue", "tweet_type": "analysis", "timeline": "quarters", "summary": "NVDA China revenue expected to rebound significantly H2 2025 via RTX 6000D; TSMC benefits from 4nm utilization surge.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Jensen Huang's third visit to China in 2025; RTX 6000D aims for two million shipments\n\nNvidia CEO Jensen Huang made his third visit to China in 2025, demonstrating significant sincerity to revive the Chinese market. According to supply chain sources, Nvidia has confirmed it will supply the RTX 6000D in the third quarter of 2025, manufactured using TSMC's 4nm process. The shipment target is set at around two million units by year-end, aiming to fill a revenue gap exceeding US$10 billion.\n\nThe RTX 6000D and the Blackwell series have also driven TSMC's 4nm production capacity utilization to unprecedented levels, contributing significantly to its revenue.\n\nReportedly, to counteract the decline in revenue from China—from 13% down to single-digit percentages—Huang visited China again on July 14 under close scrutiny from US authorities. This marks his third appearance in China in 2025.\n\nSupply chain insiders revealed that besides showing his commitment to the Chinese market, Huang also presented multiple clients with Nvidia's solutions amid ongoing US-China tensions.\n\nImpact of US export bans on Nvidia\nHuang's active push to ship downgraded AI chips to China reflects the critical situation. Nvidia's fiscal year 2025 financial report showed that revenue from China (including Hong Kong) reached US$17.108 billion, accounting for 13.11% of total annual revenue.\n\nFollowing the US ban on H20 chips, Nvidia immediately recognized inventory losses up to US$5.5 billion. Its May financial report disclosed impairments of US$4.5 billion related to inventory and procurement obligations, along with approximately US$2.5 billion in lost potential revenue recognition.\n\nAs Nvidia nearly monopolizes the AI GPU market, it has been one of the hardest-hit companies under these restrictions. The US Department of Commerce first imposed export controls targeting AI chips to China in 2022, banning products like Nvidia's A100 and H100 from being exported to China. In response, Nvidia urgently launched downgraded versions, the A800 and H800, to continue supplying the Chinese market.\n\nA year later, the US government further blocked exports of the A800 and H800. Nvidia then shifted focus to developing downgraded H20 and L20 products, but even flagship gaming GPUs such as the RTX 4090 faced restrictions.\n\nIn 2025, the supply of the special edition RTX 5090D was also hindered, and the ban on H20 sales to China almost pushed Nvidia's revenue share in the Chinese market below 5%.\n\nThe new RTX 6000D: A tactical response\nDetermined not to abandon the Chinese market, Nvidia plans to launch the \"RTX 6000D\" in the third quarter of 2025, complying with US regulations. This modified version of the RTX 5090 uses TSMC's 4nm process, offers up to 1,100GB/s bidirectional bandwidth, and features next-generation GDDR7 graphics memory. It aims to deliver memory performance comparable to high-bandwidth memory (HBM).\n\nIndustry experts caution that supply constraints on GDDR7 could pose risks in the second half of 2025. The current shipment target for the RTX 6000D starts at a minimum of one million units, potentially reaching up to two million if obstacles are overcome.\n\nAnalysts note that while Chinese firms like Huawei and Cambricon accelerate AI chip development, they face limitations due to SMIC's manufacturing capacity, Huawei's process capabilities, high production costs, memory bandwidth challenges, system integration difficulties, and immature software ecosystems. Replacing Nvidia's CUDA ecosystem remains highly challenging.\n\nIf the RTX 6000D can be supplied smoothly and maintain Chinese clients' purchasing dependence, Nvidia's revenue from China is expected to rebound significantly in the second half of 2025, driving robust overall growth.\n\n---\n\nhttps://t.co/BHM5yrkHG7", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013513498494239662, "ret_signed_m1d": -0.013513498494239662, "alpha_spy_m1d": -0.017805003864211355, "alpha_c_m1d": 0.00533100244982232, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01801788684969341, "ret_signed_p1d": 0.01801788684969341, "alpha_spy_p1d": 0.0146746957311783, "alpha_c_p1d": 0.023244861758361046, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.01801788684969341, "ret_signed_p1w": 0.01801788684969341, "alpha_spy_p1w": 0.007216503428878029, "alpha_c_p1w": 0.03669048176540157, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.0810811021080644, "ret_signed_p1m": 0.0810811021080644, "alpha_spy_p1m": 0.04451376818499875, "alpha_c_p1m": 0.04504252441622447, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.3024865495193749, "ret_signed_p3m": 0.3024865495193749, "alpha_spy_p3m": 0.24994143137881086, "alpha_c_p3m": 0.18371074782461094, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.5200951425337497, "ret_signed_p6m": 0.5200951425337497, "alpha_spy_p6m": 0.4053378392525626, "alpha_c_p6m": 0.19197264269797087, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2402, -0.2375, -0.2509, -0.268, -0.2168, -0.2249, -0.2034, -0.1953, -0.1908, -0.1854, -0.1854, -0.1478, -0.1585, -0.1747, -0.1675, -0.1765, -0.1487, -0.1415, -0.1307, -0.1038, -0.1092, -0.1047, -0.1173, -0.1226, -0.1119, -0.1191, -0.1173, -0.1244, -0.1343, -0.1325, -0.1325, -0.1325, -0.1513, -0.1478, -0.1119, -0.1047, -0.1074, -0.0984, -0.0625, -0.0446, -0.0586, -0.0721, -0.0766, -0.0586, -0.0495, -0.0676, -0.0495, -0.0811, -0.0541, -0.036, -0.0315, -0.027, -0.045, -0.0225, -0.0225, -0.018, -0.0225, -0.027, -0.027, -0.018, -0.009, -0.009, -0.0135, 0.0, 0.018, 0.018, 0.0405, 0.036, 0.018, 0.0315, 0.0315, 0.0315, 0.0315, 0.0225, 0.0405, 0.045, 0.045, 0.0225, 0.036, 0.0135, 0.0631, 0.0586, 0.0631, 0.0631, 0.0811, 0.0586, 0.0631, 0.0631, 0.0676, 0.0225, 0.036, 0.0225, 0.0541, 0.0586, 0.0721, 0.045, 0.045, 0.0495, 0.045, 0.045, 0.045, 0.0631, 0.0631, 0.0811, 0.1036, 0.1171, 0.1351, 0.1306, 0.1578, 0.1442, 0.1623, 0.1442, 0.1713, 0.212, 0.212, 0.1939, 0.1759, 0.1759, 0.1804, 0.1985, 0.2346, 0.2663, 0.2663, 0.298, 0.2799, 0.3025, 0.3025, 0.2799, 0.2889, 0.3251, 0.3432, 0.3115, 0.3387, 0.3387, 0.3206, 0.3115, 0.3115, 0.3387, 0.3341, 0.3613, 0.3613, 0.3568, 0.3658, 0.3613, 0.3206, 0.3251, 0.3206, 0.3341, 0.3251, 0.3341, 0.3206, 0.2934, 0.307, 0.2708, 0.2618, 0.3161, 0.2527, 0.2437, 0.2799, 0.3025, 0.298, 0.3025, 0.2754, 0.2934, 0.3115, 0.307, 0.3206, 0.3522, 0.3387, 0.3613, 0.3341, 0.3431, 0.3159, 0.3023, 0.2978, 0.2978, 0.2978, 0.3295, 0.3522, 0.3567, 0.3567, 0.3704, 0.3885, 0.3794, 0.4067, 0.4067, 0.4384, 0.5156, 0.5473, 0.5201]}
{"unit_id": "orig:1946477919657947301", "ticker_idx": 0, "ts": "2025-07-19T07:51:14+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron announced strong demand-driven results in its May 2024 quarterly earnings and provided a positive outlook for the August quarter. HBM demand surpassed $1.5 billion, a 50% increase quarter-over-quarter. HBM revenue could approach $8 billion on an annualized basis, and Micron is expanding its customer base by mass-producing additional HBM4 products by 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM bullish thesis: DRAM supply balanced, no ASP downcycle; MU HBM ramping to ~$8B annualized; NAND faces declines in 2025-2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) Regarding Memory Semiconductors:\n\nThe HBM market continues to face tight overall supply due to a surge in memory content, driven by increasing demand for AI and high-performance servers. This, coupled with rising shipments and improved average selling prices (ASPs) for DDR4 and DDR5, and the stable trend in NAND prices, indicates a recovering demand in consumer applications as well. In particular, the rapid inventory clearance by customers and distribution channels is creating a favorable short-term supply-demand structure, accompanied by some pre-emptive demand related to tariffs. However, it's also assessed that these tariff factors could act as a demand headwind in the second half of the year, potentially limiting the quarter-over-quarter improvement in ASPs and shipments.\n\nDRAM revenue surged by 96% in 2024, and we anticipate growth of 31% in 2025 and 15% in 2026. In contrast, NAND is projected to increase by 91% in 2024, followed by a 9% decrease in 2025 and a 5% decrease in 2026. For DRAM, supply and demand are expected to balance out in 2025, and a double-digit ASP decline, typical in a downcycle, is not foreseen. This is because the increasing proportion of cleanroom investment in infrastructure and the additional cleanroom space occupied by HBM expansion limit the overall capacity for bit production growth. In the NAND market, companies with portfolios focused on data centers and eSSD are expected to perform relatively well, while slowing consumer sentiment and macro uncertainties could weigh on consumer-facing demand.\n\nMicron announced strong demand-driven results in its May 2024 quarterly earnings and provided a positive outlook for the August quarter. Notably, HBM demand surpassed $1.5 billion, a 50% increase quarter-over-quarter, and improvements were also observed across the datacenter, mobile, and industrial/embedded markets. The ramp-up of HBM3E 12-high is proceeding smoothly, and HBM market share is expected to reach a level similar to Micron's overall DRAM share by the end of the year. This implies that HBM revenue could approach $8 billion on an annualized basis, and Micron is expanding its customer base by mass-producing additional HBM4 products by 2026. Currently, collaboration with a total of four customers is underway.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010244619707217018, "ret_signed_m1d": 0.010244619707217018, "alpha_spy_m1d": 0.0121372061973819, "alpha_c_m1d": 0.010933035595134921, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03541463067701622, "ret_signed_p1d": -0.03541463067701622, "alpha_spy_p1d": -0.03555775184024079, "alpha_c_p1d": -0.01772161526895566, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.017486495581344408, "ret_signed_p1w": -0.017486495581344408, "alpha_spy_p1w": -0.03048017426610994, "alpha_c_p1w": -0.020584525139707543, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.07789454889981817, "ret_signed_p1m": 0.07789454889981817, "alpha_spy_p1m": 0.060336460258479496, "alpha_c_p1m": 0.07665533707647287, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.7897804009088125, "ret_signed_p3m": 0.7897804009088125, "alpha_spy_p3m": 0.7361811337048312, "alpha_c_p3m": 0.6080655825381716, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 1.9892999079136064, "ret_signed_p6m": 1.9892999079136064, "alpha_spy_p6m": 1.8795949910555487, "alpha_c_p6m": 1.6373730230971864, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.3565, -0.3169, -0.2961, -0.3068, -0.3217, -0.321, -0.3138, -0.2878, -0.2904, -0.2896, -0.271, -0.2487, -0.2424, -0.1857, -0.1448, -0.159, -0.1578, -0.1353, -0.1296, -0.1344, -0.1544, -0.1633, -0.1762, -0.1762, -0.1496, -0.1514, -0.1459, -0.1666, -0.1337, -0.0978, -0.089, -0.0622, -0.0421, -0.0211, 0.0071, 0.0238, 0.0251, 0.02, 0.0574, 0.0618, 0.0749, 0.0749, 0.0906, 0.0771, 0.1286, 0.1228, 0.1117, 0.1008, 0.0875, 0.0666, 0.0741, 0.079, 0.079, 0.0591, 0.0988, 0.0796, 0.0873, 0.0998, 0.0475, 0.0608, 0.0283, 0.0003, 0.0102, 0.0, -0.0354, -0.03, -0.0132, -0.0174, -0.0175, -0.0112, 0.0133, -0.0361, -0.0737, -0.0482, -0.0368, -0.0393, -0.012, 0.05, 0.0926, 0.1282, 0.0975, 0.1065, 0.0675, 0.0911, 0.0779, 0.0351, 0.0226, 0.0393, 0.0282, 0.0289, 0.0399, 0.0775, 0.051, 0.051, 0.0464, 0.0485, 0.097, 0.1602, 0.161, 0.1944, 0.2364, 0.3298, 0.3886, 0.3934, 0.4026, 0.413, 0.4916, 0.4372, 0.4539, 0.4697, 0.4282, 0.3851, 0.3889, 0.4475, 0.4777, 0.6087, 0.6228, 0.6599, 0.6875, 0.641, 0.7368, 0.6996, 0.6048, 0.7035, 0.6531, 0.6962, 0.7898, 0.7885, 0.8272, 0.7877, 0.7539, 0.8267, 0.9355, 0.945, 0.961, 1.0028, 0.9796, 0.9775, 1.0741, 0.9268, 1.0988, 1.1061, 1.1025, 1.2384, 1.1307, 1.1642, 1.094, 1.1813, 1.1381, 1.0193, 0.9965, 0.7795, 0.8326, 0.9789, 0.9842, 1.0348, 1.0348, 1.0898, 1.125, 1.1164, 1.0693, 1.0029, 1.0963, 1.1821, 1.2307, 1.3304, 1.284, 1.131, 1.0988, 1.0547, 0.9929, 1.1965, 1.35, 1.4443, 1.4414, 1.5334, 1.5334, 1.5167, 1.6024, 1.587, 1.5232, 1.5232, 1.7885, 1.7596, 2.0362, 2.0019, 1.8911, 2.0508, 2.0577, 1.9893]}
{"unit_id": "orig:1946477919657947301", "ticker_idx": 1, "ts": "2025-07-19T07:51:14+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM revenue surged by 96% in 2024, and we anticipate growth of 31% in 2025 and 15% in 2026. For DRAM, supply and demand are expected to balance out in 2025, and a double-digit ASP decline, typical in a downcycle, is not foreseen", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM bullish thesis: DRAM supply balanced, no ASP downcycle; MU HBM ramping to ~$8B annualized; NAND faces declines in 2025-2026.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) Regarding Memory Semiconductors:\n\nThe HBM market continues to face tight overall supply due to a surge in memory content, driven by increasing demand for AI and high-performance servers. This, coupled with rising shipments and improved average selling prices (ASPs) for DDR4 and DDR5, and the stable trend in NAND prices, indicates a recovering demand in consumer applications as well. In particular, the rapid inventory clearance by customers and distribution channels is creating a favorable short-term supply-demand structure, accompanied by some pre-emptive demand related to tariffs. However, it's also assessed that these tariff factors could act as a demand headwind in the second half of the year, potentially limiting the quarter-over-quarter improvement in ASPs and shipments.\n\nDRAM revenue surged by 96% in 2024, and we anticipate growth of 31% in 2025 and 15% in 2026. In contrast, NAND is projected to increase by 91% in 2024, followed by a 9% decrease in 2025 and a 5% decrease in 2026. For DRAM, supply and demand are expected to balance out in 2025, and a double-digit ASP decline, typical in a downcycle, is not foreseen. This is because the increasing proportion of cleanroom investment in infrastructure and the additional cleanroom space occupied by HBM expansion limit the overall capacity for bit production growth. In the NAND market, companies with portfolios focused on data centers and eSSD are expected to perform relatively well, while slowing consumer sentiment and macro uncertainties could weigh on consumer-facing demand.\n\nMicron announced strong demand-driven results in its May 2024 quarterly earnings and provided a positive outlook for the August quarter. Notably, HBM demand surpassed $1.5 billion, a 50% increase quarter-over-quarter, and improvements were also observed across the datacenter, mobile, and industrial/embedded markets. The ramp-up of HBM3E 12-high is proceeding smoothly, and HBM market share is expected to reach a level similar to Micron's overall DRAM share by the end of the year. This implies that HBM revenue could approach $8 billion on an annualized basis, and Micron is expanding its customer base by mass-producing additional HBM4 products by 2026. Currently, collaboration with a total of four customers is underway.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00430800312716458, "ret_signed_m1d": -0.00430800312716458, "alpha_spy_m1d": -0.002415416636999697, "alpha_c_m1d": -0.003619587239246676, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02554740762843018, "ret_signed_p1d": -0.02554740762843018, "alpha_spy_p1d": -0.02569052879165475, "alpha_c_p1d": -0.007854392220369617, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.005890159510396324, "ret_signed_p1w": -0.005890159510396324, "alpha_spy_p1w": -0.018883838195161855, "alpha_c_p1w": -0.00898818906875946, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.025160171020784827, "ret_signed_p1m": 0.025160171020784827, "alpha_spy_p1m": 0.007602082379446151, "alpha_c_p1m": 0.02392095919743953, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.6333872171104805, "ret_signed_p3m": 0.6333872171104805, "alpha_spy_p3m": 0.5797879499064992, "alpha_c_p3m": 0.45167239873983955, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 1.583958051997555, "ret_signed_p6m": 1.583958051997555, "alpha_spy_p6m": 1.4742531351394972, "alpha_c_p6m": 1.2320311671811348, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.2923, -0.2824, -0.268, -0.2741, -0.2805, -0.2858, -0.2833, -0.2701, -0.271, -0.2708, -0.2572, -0.2504, -0.2476, -0.209, -0.1945, -0.1876, -0.1945, -0.1845, -0.1937, -0.1917, -0.2011, -0.2134, -0.2163, -0.2102, -0.2059, -0.19, -0.1818, -0.1974, -0.1799, -0.1679, -0.1478, -0.124, -0.1173, -0.1014, -0.0931, -0.0725, -0.0795, -0.0871, -0.0647, -0.0576, -0.048, -0.0515, -0.0314, -0.0401, 0.0125, 0.0236, 0.0232, 0.0132, 0.0137, 0.0008, -0.0017, 0.014, 0.0018, -0.0121, 0.0131, 0.0005, 0.0256, 0.0346, 0.0234, 0.0319, 0.0229, -0.009, -0.0043, 0.0, -0.0255, -0.0212, -0.0169, -0.0231, -0.0059, -0.0022, 0.017, 0.0069, -0.0369, -0.0245, -0.013, -0.0253, -0.0036, 0.0168, 0.0399, 0.0547, 0.0594, 0.0591, 0.0461, 0.0351, 0.0252, 0.0042, -0.0123, 0.0045, 0.0117, 0.0085, 0.0118, 0.0302, 0.0226, -0.0037, 0.0076, 0.0142, 0.0355, 0.0635, 0.071, 0.1025, 0.1415, 0.1802, 0.236, 0.246, 0.2842, 0.264, 0.3275, 0.3093, 0.3251, 0.3485, 0.3339, 0.3217, 0.2847, 0.3258, 0.3326, 0.4019, 0.4686, 0.481, 0.4902, 0.4747, 0.5067, 0.4943, 0.5254, 0.537, 0.5075, 0.5521, 0.6334, 0.6499, 0.6883, 0.6642, 0.6614, 0.6716, 0.7578, 0.8074, 0.7833, 0.8474, 0.8698, 0.8748, 0.9995, 0.8781, 0.9057, 0.9184, 0.8948, 0.9853, 0.9796, 0.9864, 0.9554, 0.8931, 0.9519, 0.8543, 0.8305, 0.7895, 0.7172, 0.7742, 0.7876, 0.8278, 0.8563, 0.8426, 0.8656, 0.9001, 0.8825, 0.8511, 0.901, 0.9754, 0.9727, 1.0297, 0.9839, 0.9481, 0.8963, 0.8423, 0.8726, 0.9402, 0.9788, 1.0715, 1.0804, 1.1139, 1.1139, 1.151, 1.245, 1.2554, 1.2341, 1.2341, 1.3971, 1.4584, 1.5913, 1.6099, 1.5792, 1.6188, 1.6262, 1.584]}
{"unit_id": "orig:1946477919657947301", "ticker_idx": 2, "ts": "2025-07-19T07:51:14+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": "NAND", "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND is projected to increase by 91% in 2024, followed by a 9% decrease in 2025 and a 5% decrease in 2026. slowing consumer sentiment and macro uncertainties could weigh on consumer-facing demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM bullish thesis: DRAM supply balanced, no ASP downcycle; MU HBM ramping to ~$8B annualized; NAND faces declines in 2025-2026.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket: dominant producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan) Regarding Memory Semiconductors:\n\nThe HBM market continues to face tight overall supply due to a surge in memory content, driven by increasing demand for AI and high-performance servers. This, coupled with rising shipments and improved average selling prices (ASPs) for DDR4 and DDR5, and the stable trend in NAND prices, indicates a recovering demand in consumer applications as well. In particular, the rapid inventory clearance by customers and distribution channels is creating a favorable short-term supply-demand structure, accompanied by some pre-emptive demand related to tariffs. However, it's also assessed that these tariff factors could act as a demand headwind in the second half of the year, potentially limiting the quarter-over-quarter improvement in ASPs and shipments.\n\nDRAM revenue surged by 96% in 2024, and we anticipate growth of 31% in 2025 and 15% in 2026. In contrast, NAND is projected to increase by 91% in 2024, followed by a 9% decrease in 2025 and a 5% decrease in 2026. For DRAM, supply and demand are expected to balance out in 2025, and a double-digit ASP decline, typical in a downcycle, is not foreseen. This is because the increasing proportion of cleanroom investment in infrastructure and the additional cleanroom space occupied by HBM expansion limit the overall capacity for bit production growth. In the NAND market, companies with portfolios focused on data centers and eSSD are expected to perform relatively well, while slowing consumer sentiment and macro uncertainties could weigh on consumer-facing demand.\n\nMicron announced strong demand-driven results in its May 2024 quarterly earnings and provided a positive outlook for the August quarter. Notably, HBM demand surpassed $1.5 billion, a 50% increase quarter-over-quarter, and improvements were also observed across the datacenter, mobile, and industrial/embedded markets. The ramp-up of HBM3E 12-high is proceeding smoothly, and HBM market share is expected to reach a level similar to Micron's overall DRAM share by the end of the year. This implies that HBM revenue could approach $8 billion on an annualized basis, and Micron is expanding its customer base by mass-producing additional HBM4 products by 2026. Currently, collaboration with a total of four customers is underway.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00020297994465674486, "ret_signed_m1d": -0.00020297994465674486, "alpha_spy_m1d": -0.0020955664348216276, "alpha_c_m1d": -0.0008913958325746485, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01270517440458414, "ret_signed_p1d": 0.01270517440458414, "alpha_spy_p1d": 0.012848295567808709, "alpha_c_p1d": -0.004987841003476424, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.006396514116279839, "ret_signed_p1w": -0.006396514116279839, "alpha_spy_p1w": 0.006597164568485692, "alpha_c_p1w": -0.0032984845579167033, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.039571267840609735, "ret_signed_p1m": -0.039571267840609735, "alpha_spy_p1m": -0.02201317919927106, "alpha_c_p1m": -0.038332056017264436, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 1.2582567981408048, "ret_signed_p3m": -1.2582567981408048, "alpha_spy_p3m": -1.2046575309368235, "alpha_c_p3m": -1.0765419797701639, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 3.587206857240017, "ret_signed_p6m": -3.587206857240017, "alpha_spy_p6m": -3.477501940381959, "alpha_c_p6m": -3.235279972423597, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.2731, -0.2601, -0.244, -0.249, -0.2496, -0.2594, -0.2576, -0.2416, -0.2411, -0.2447, -0.2231, -0.2097, -0.1932, -0.1383, -0.1272, -0.1181, -0.1335, -0.1276, -0.1434, -0.1361, -0.1504, -0.1698, -0.1757, -0.1704, -0.1639, -0.1502, -0.1383, -0.1589, -0.1589, -0.1465, -0.1222, -0.1065, -0.0982, -0.0803, -0.0747, -0.0639, -0.0717, -0.0771, -0.0555, -0.0455, -0.0308, -0.0281, -0.0008, 0.0187, 0.0517, 0.0563, 0.057, 0.049, 0.0389, 0.0174, 0.0261, 0.048, 0.0264, 0.0192, 0.0469, 0.0526, 0.0584, 0.0538, 0.0254, 0.0327, 0.021, -0.0006, 0.0002, 0.0, -0.0127, 0.0094, 0.0101, -0.0008, 0.0064, 0.0114, 0.0272, 0.0211, -0.0175, -0.0079, -0.0048, -0.0145, -0.0111, 0.0241, 0.0333, 0.0804, 0.0807, 0.077, 0.0593, 0.0474, 0.0396, 0.0164, 0.0072, 0.0298, 0.0352, 0.0395, 0.0452, 0.0727, 0.0864, 0.0774, 0.0724, 0.0863, 0.1449, 0.2281, 0.2492, 0.2702, 0.3364, 0.4537, 0.533, 0.558, 0.6105, 0.5875, 0.6559, 0.6619, 0.6994, 0.7302, 0.7009, 0.6619, 0.6112, 0.7384, 0.7532, 0.8232, 0.9377, 1.0311, 1.0137, 0.9677, 1.0377, 1.0486, 1.0008, 1.0928, 1.0263, 1.162, 1.2583, 1.2212, 1.3323, 1.3144, 1.3245, 1.4323, 1.6958, 1.7666, 1.7109, 1.9475, 1.9895, 2.0031, 2.1148, 1.9767, 2.0803, 2.1268, 2.3113, 2.6113, 2.6147, 2.6779, 2.451, 2.2096, 2.3981, 2.1619, 2.2015, 1.979, 1.8455, 2.0079, 1.9699, 1.8431, 1.9165, 1.9783, 1.8845, 1.9105, 1.8297, 1.9069, 2.0403, 2.1317, 2.0777, 2.142, 2.1685, 2.0005, 1.89, 1.8499, 1.9082, 2.0273, 2.1233, 2.2621, 2.2712, 2.3718, 2.3879, 2.4627, 2.4284, 2.3948, 2.3684, 2.3684, 2.6482, 2.7571, 3.2213, 3.344, 3.2605, 3.4639, 3.5254, 3.5872]}
{"unit_id": "orig:1948531643209449671", "ticker_idx": 0, "ts": "2025-07-24T23:52:00+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Azure Growth Expected to Meet or Exceed Upper Guidance… UBS has therefore raised its FY26 operating margin forecast by approximately 50 basis points", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS bullish outlook on MSFT: Azure to meet/beat guidance, FY26 margins revised up ~50bps, enterprise spending improving.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Microsoft Outlook: Azure Growth and Profitability Expectations Positive, OpenAI Partnership Draws Attention\nUBS (K. Keirstead, 25/07/22)\n\nMicrosoft's upcoming Q4 earnings report is expected to bring positive signals regarding Azure growth and profitability revisions, but the market is also closely watching changes in its relationship with OpenAI and long-term capital expenditure trends.\n\nAzure Growth Expected to Meet or Exceed Upper Guidance, Enterprise Spending Environment Marginally Improving\n\nUBS states that market expectations for Azure's year-over-year growth this quarter are 36%, with guidance for a slight slowdown to 34% next quarter. Although this requires historical levels of sequential revenue growth, feedback from enterprise customers and partners indicates that spending sentiment has improved compared to the February-April period. AI demand remains strong, ChatGPT inference load is increasing, and supply bottlenecks have not worsened.\n\nFY26 Margin Outlook Improves, Cost Control Measures Show Initial Results\n\nMicrosoft continues to implement cost control measures on operating expenses to offset gross margin pressure. UBS has therefore raised its FY26 operating margin forecast by approximately 50 basis points, now expecting margins to remain stable rather than decline.\n\nMicrosoft-OpenAI Partnership Creates Uncertainty, Compute Demand Structure May Change\n\nThe market continues to focus on reports of adjustments to the terms of the Microsoft-OpenAI partnership, especially the risk of compute workload potentially shifting from Microsoft to Oracle and CoreWeave. UBS believes the market currently has no clear conclusion on this but generally tends to trust Microsoft's execution capabilities.\n\nOther Financial Variables Also Under Scrutiny, Including Capex and Tax Policy Impact\n\nIn addition to core growth metrics, UBS points out that the scale of FY26 capital expenditures, changes in tax legislation, and adjustments to asset depreciation rules could also impact financial performance, warranting continuous monitoring.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009806681896403324, "ret_signed_m1d": -0.009806681896403324, "alpha_spy_m1d": -0.009475738933079292, "alpha_c_m1d": -0.009475738933079292, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0003309429633240324, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005539485520711374, "ret_signed_p1d": 0.005539485520711374, "alpha_spy_p1d": 0.0013151358803922264, "alpha_c_p1d": 0.0013151358803922264, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": 0.004224349640319147, "ret_p1w": 0.04427650522102189, "ret_signed_p1w": 0.04427650522102189, "alpha_spy_p1w": 0.04796480681901749, "alpha_c_p1w": 0.04796480681901749, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": -0.0036883015979956024, "ret_p1m": -0.005512329704128072, "ret_signed_p1m": -0.005512329704128072, "alpha_spy_p1m": -0.02267770377543099, "alpha_c_p1m": -0.02267770377543099, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.01716537407130292, "ret_p3m": 0.014936903369850096, "ret_signed_p3m": 0.014936903369850096, "alpha_spy_p3m": -0.04611261369209818, "alpha_c_p3m": -0.04611261369209818, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.06104951706194828, "ret_p6m": -0.09669969520455979, "ret_signed_p6m": -0.09669969520455979, "alpha_spy_p6m": -0.19317679717263703, "alpha_c_p6m": -0.19317679717263703, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.09647710196807724, "price_path": [-0.2357, -0.2301, -0.2277, -0.1688, -0.1495, -0.1478, -0.1534, -0.1533, -0.1439, -0.1428, -0.1222, -0.1225, -0.115, -0.113, -0.1108, -0.1018, -0.1032, -0.1141, -0.1097, -0.1188, -0.1188, -0.0982, -0.1048, -0.1022, -0.0989, -0.0957, -0.0938, -0.092, -0.0846, -0.0793, -0.0746, -0.0782, -0.0749, -0.0627, -0.0703, -0.0621, -0.0643, -0.06, -0.06, -0.0655, -0.0487, -0.0407, -0.0364, -0.0263, -0.0292, -0.0264, -0.0369, -0.0387, -0.0236, -0.0236, -0.0258, -0.0279, -0.0144, -0.0184, -0.0148, -0.0154, -0.0099, -0.0103, 0.0016, -0.0016, -0.0016, -0.011, -0.0098, 0.0, 0.0055, 0.0032, 0.0033, 0.0046, 0.0443, 0.0259, 0.0485, 0.033, 0.0275, 0.0195, 0.0218, 0.0213, 0.0359, 0.019, 0.0227, 0.0182, 0.0122, -0.0022, -0.0101, -0.0114, -0.0055, -0.0113, -0.0157, -0.0065, -0.0008, -0.0066, -0.0066, -0.0096, -0.0092, -0.0041, -0.0295, -0.0232, -0.0228, -0.019, -0.0177, -0.0003, 0.0104, -0.002, -0.0, -0.0031, 0.0155, 0.0086, -0.0016, 0.0002, -0.0059, 0.0028, 0.0089, 0.0155, 0.019, 0.0112, 0.0143, 0.0363, 0.0273, 0.029, 0.0242, 0.0018, 0.0079, 0.0069, 0.0066, 0.0031, 0.0069, 0.0132, 0.0149, 0.0206, 0.0206, 0.0266, 0.0421, 0.0628, 0.0618, 0.0308, 0.0152, 0.0137, 0.0084, -0.0056, -0.0254, -0.0259, -0.0079, -0.0027, 0.0022, -0.0132, 0.0003, -0.005, -0.0319, -0.0449, -0.0602, -0.0726, -0.0689, -0.0631, -0.0463, -0.0463, -0.0335, -0.0439, -0.0375, -0.0616, -0.0555, -0.0509, -0.0355, -0.0335, -0.06, -0.0503, -0.06, -0.0673, -0.0642, -0.0648, -0.0493, -0.0455, -0.0475, -0.0437, -0.0414, -0.0414, -0.042, -0.0432, -0.0424, -0.05, -0.05, -0.071, -0.0712, -0.0601, -0.0503, -0.0609, -0.0586, -0.0627, -0.0755, -0.0976, -0.103, -0.0967]}
{"unit_id": "orig:1946461164361470439", "ticker_idx": 0, "ts": "2025-07-19T06:44:39+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley believes that TSMC's 2025 revenue growth guidance has been raised to a 30% year-on-year growth rate, and strong growth is expected to continue in 2026, with AI demand, especially AI GPU demand from the Chinese market, being the main driving force.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises TSMC outlook: 30% revenue growth in 2025, 33% in 2026, driven by AI/GPU demand, 2nm ramp, and potential tariff exemptions; bullish on gross margin staying above 53%.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC: 2025 Revenue Growth Guidance Raised, Strong AI Demand Drives Growth\nMorgan Stanley (C. Chan, 25/07/17)\n\nMorgan Stanley believes that TSMC's 2025 revenue growth guidance has been raised to a 30% year-on-year growth rate, and strong growth is expected to continue in 2026, with AI demand, especially AI GPU demand from the Chinese market, being the main driving force. TSMC expects revenue to grow by 33% in 2026 and plans to continue expanding 2nm capacity in the coming years. TSMC also anticipates AI semiconductors to achieve a compound annual growth rate (CAGR) of over 45% in the next five years, further boosting the company's revenue growth.\n\nRevenue Growth Guidance Raised: TSMC has raised its 2025 revenue growth guidance to 30% year-on-year, better than the originally anticipated mid-20%. This upward revision reflects strong expectations for High-Performance Computing (HPC) and AI demand. TSMC expects to achieve 33% USD revenue growth in 2025, with 2026 revenue also projected to grow by 33%. Although a 10% sequential decline is expected in the fourth quarter, Morgan Stanley considers this too conservative and expects the fourth quarter to be flat quarter-on-quarter.\n\nAI Demand Drives Growth: TSMC's AI business, especially GPU demand from the Chinese market, is expected to contribute 2-3% to the company's revenue in 2025. The company expects revenue from AI semiconductors in cloud computing to achieve a mid-40% CAGR over the next five years. TSMC also stated that with the resumption of H20 shipments, AI demand will further drive the company's growth in the coming years.\n\n2026 Gross Margin Outlook: TSMC expects its gross margin in 2026 to remain above 53%, despite potential foreign exchange impacts. TSMC plans to offset possible cost pressures by raising wafer prices. Morgan Stanley believes that due to tight 3nm and 4nm capacity, TSMC may raise prices by 3-5% in 2026. Prices for 4nm capacity at US fabs are expected to increase by more than 10%.\n\nSemiconductor Tariffs: Semiconductor tariffs remain a potential burden for TSMC. TSMC mentioned in its Q1 2025 earnings call that tariffs could threaten electronics demand and reduce company revenue. However, Morgan Stanley believes that TSMC's $165 billion investment in the US increases the likelihood of obtaining tariff exemptions. If tariff exemptions apply to equipment and chemical imports, TSMC's long-term gross margin will be able to stay above 53%.\n\n2nm Capacity Build-out: TSMC plans to begin mass production of 2nm in the second half of 2025, with 2nm capacity expected to reach 90k wafers/month by the end of 2026. Major customers include Apple, AMD, and Intel. The application of 2nm technology will drive TSMC's continued growth in smartphones and High-Performance Computing (HPC). 2nm technology is expected to become a key pillar for the company's future growth, with AI semiconductor revenue contribution projected to grow from 13% in 2024 to 34% by 2027.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006489367268648127, "ret_signed_m1d": 0.006489367268648127, "alpha_spy_m1d": 0.00838195375881301, "alpha_c_m1d": 0.007177783156566031, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01779356357664763, "ret_signed_p1d": -0.01779356357664763, "alpha_spy_p1d": -0.0179366847398722, "alpha_c_p1d": -0.00010054816858706683, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.01632826207599436, "ret_signed_p1w": 0.01632826207599436, "alpha_spy_p1w": 0.003334583391228829, "alpha_c_p1w": 0.013230232517631224, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": -0.025748490106008992, "ret_signed_p1m": -0.025748490106008992, "alpha_spy_p1m": -0.04330657874734767, "alpha_c_p1m": -0.02698770192935429, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.2593088112441124, "ret_signed_p3m": 0.2593088112441124, "alpha_spy_p3m": 0.2057095440401311, "alpha_c_p3m": 0.07759399287347146, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.39481611010684925, "ret_signed_p6m": 0.39481611010684925, "alpha_spy_p6m": 0.28511119324879153, "alpha_c_p6m": 0.042889225290429156, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.3417, -0.3151, -0.3113, -0.3184, -0.3139, -0.3047, -0.2795, -0.2521, -0.2642, -0.2813, -0.2719, -0.2691, -0.2637, -0.22, -0.1908, -0.1876, -0.1898, -0.1898, -0.1928, -0.193, -0.2001, -0.1816, -0.1992, -0.1992, -0.1754, -0.1818, -0.1776, -0.1936, -0.1872, -0.1757, -0.1557, -0.1518, -0.1441, -0.1365, -0.1137, -0.1069, -0.0981, -0.1162, -0.097, -0.1045, -0.1061, -0.1061, -0.1228, -0.1194, -0.0785, -0.0674, -0.0621, -0.043, -0.0517, -0.0593, -0.022, -0.017, -0.017, -0.0405, -0.046, -0.0293, -0.0381, -0.0354, -0.0426, -0.008, -0.0054, 0.0283, 0.0065, 0.0, -0.0178, 0.0062, 0.0115, 0.0283, 0.0163, 0.0104, 0.017, 0.0116, -0.0152, 0.0006, -0.0267, -0.0313, 0.0158, 0.0125, 0.0136, 0.0228, 0.0108, 0.009, 0.0001, 0.0107, -0.0257, -0.0429, -0.0482, -0.0245, -0.0136, -0.0005, 0.0018, -0.0024, -0.0334, -0.0334, -0.0438, -0.0312, -0.0152, 0.0191, 0.0349, 0.0505, 0.0904, 0.084, 0.0857, 0.0943, 0.1006, 0.1037, 0.1283, 0.1124, 0.145, 0.1874, 0.179, 0.162, 0.1481, 0.1475, 0.173, 0.2116, 0.21, 0.2272, 0.2701, 0.2349, 0.279, 0.2595, 0.1788, 0.2721, 0.2429, 0.2798, 0.2593, 0.2393, 0.2503, 0.2369, 0.2133, 0.221, 0.2388, 0.2526, 0.2664, 0.2814, 0.2735, 0.2618, 0.2804, 0.235, 0.2333, 0.2148, 0.2033, 0.2401, 0.2229, 0.2206, 0.1852, 0.1962, 0.1844, 0.1672, 0.1859, 0.1655, 0.1552, 0.1955, 0.1956, 0.2178, 0.2178, 0.2243, 0.2082, 0.2268, 0.2409, 0.2303, 0.2378, 0.2678, 0.2743, 0.3026, 0.2838, 0.2299, 0.2118, 0.2081, 0.1664, 0.1989, 0.2168, 0.2351, 0.2505, 0.2583, 0.2583, 0.2753, 0.2673, 0.2616, 0.2798, 0.2798, 0.346, 0.3571, 0.3789, 0.342, 0.3392, 0.3629, 0.3972, 0.3948]}
{"unit_id": "orig:1963422625394561506", "ticker_idx": 0, "ts": "2025-09-04T02:03:27+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's HBM price negotiations with Nvidia are likely to turn out better than the market had feared", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares UBS view that SK Hynix HBM pricing with Nvidia will beat market fears; implied bullish.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: SK Hynix’s HBM price negotiations with Nvidia are likely to turn out better than the market had feared. https://t.co/6PdDViWxWr", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011299482371136316, "ret_signed_m1d": -0.011299482371136316, "alpha_spy_m1d": -0.0030113885170360843, "alpha_c_m1d": 0.0005337541692898018, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.011833236540426118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.030131716942059406, "ret_signed_p1d": 0.030131716942059406, "alpha_spy_p1d": 0.03302786543059455, "alpha_c_p1d": 0.018332804127597413, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.011798912814461993, "ret_p1w": 0.15630885700304953, "ret_signed_p1w": 0.15630885700304953, "alpha_spy_p1w": 0.14319862186522592, "alpha_c_p1w": 0.10776813813326958, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.04854071886977995, "ret_p1m": 0.4896422296776557, "ret_signed_p1m": 0.4896422296776557, "alpha_spy_p1m": 0.4558342334554881, "alpha_c_p1m": 0.32811593701709274, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.16152629266056295, "ret_p3m": 1.1032000270038496, "ret_signed_p3m": 1.1032000270038496, "alpha_spy_p3m": 1.0503598727423618, "alpha_c_p3m": 0.8631875183819937, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.24001250862185586, "ret_p6m": 3.006474786689651, "ret_signed_p6m": 3.006474786689651, "alpha_spy_p6m": 2.9436134200512107, "alpha_c_p6m": 2.6001597069250084, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.40631507976464265, "price_path": [-0.1387, -0.1331, -0.0973, -0.1143, -0.1143, -0.0673, -0.0635, -0.0729, -0.0748, -0.0334, -0.024, 0.0475, 0.0757, 0.102, 0.0681, 0.0982, 0.0738, 0.0493, 0.0475, 0.0174, 0.0192, 0.0606, 0.0569, 0.117, 0.1076, 0.1283, 0.1227, 0.1133, 0.0136, 0.0117, 0.0249, 0.0098, 0.0117, 0.0136, 0.0004, -0.0146, -0.0127, -0.009, 0.0286, -0.0297, -0.0297, -0.009, -0.0278, -0.0146, -0.0353, 0.0042, 0.0117, 0.0456, 0.0399, 0.0399, 0.0061, -0.0108, -0.0391, -0.0785, -0.056, -0.024, -0.0165, -0.0221, 0.0113, 0.0132, -0.0358, -0.0188, -0.0113, 0.0, 0.0301, 0.0433, 0.0847, 0.145, 0.1563, 0.2373, 0.2467, 0.3107, 0.2561, 0.339, 0.339, 0.322, 0.3597, 0.3465, 0.3427, 0.2674, 0.3145, 0.3089, 0.3559, 0.4896, 0.4896, 0.4896, 0.4896, 0.4896, 0.4896, 0.6121, 0.5631, 0.5499, 0.5913, 0.7043, 0.7533, 0.8286, 0.8041, 0.8136, 0.8023, 0.9209, 1.0151, 0.9623, 1.1017, 1.1394, 1.1055, 1.3352, 1.2072, 1.1808, 1.2335, 1.1846, 1.2825, 1.3314, 1.3239, 1.3051, 1.1092, 1.2825, 1.1469, 1.1168, 1.1507, 0.9623, 0.9586, 0.9548, 0.9736, 1.0504, 0.9977, 1.0278, 1.1032, 1.0806, 1.0429, 1.0504, 1.1748, 1.1334, 1.2125, 1.1296, 1.1522, 1.0881, 0.9977, 1.0768, 1.0806, 1.0617, 1.1861, 1.2012, 1.2163, 1.2163, 1.2577, 1.4123, 1.4537, 1.4537, 1.4537, 1.5517, 1.6233, 1.7364, 1.7967, 1.8495, 1.8043, 1.8231, 1.7817, 1.7967, 1.8231, 1.8495, 1.8797, 1.8005, 1.7892, 1.8457, 1.891, 1.7741, 2.0153, 2.1699, 2.2453, 2.4262, 2.1284, 2.4186, 2.3923, 2.1736, 2.1623, 2.3433, 2.3018, 2.2415, 2.347, 2.3169, 2.3169, 2.3169, 2.3169, 2.3696, 2.5769, 2.5845, 2.788, 2.837, 3.15, 3.0065]}
{"unit_id": "orig:1963829050432250263", "ticker_idx": 0, "ts": "2025-09-05T04:58:26+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Market Undervalues Amazon AWS's 'AI Potential'", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Barclays call that market undervalues AMZN AWS AI potential; Anthropic/Claude 5 ramp could add 4pp to AWS quarterly revenue growth.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Barclays: Market Undervalues Amazon AWS’s ‘AI Potential’; “Deep Integration” with Anthropic, API Business Already Surpassing OpenAI\n\n• Barclays analyzed that Anthropic’s API business has already outpaced OpenAI in both scale and growth rate, contributing significantly to AWS’s revenue. According to the data, Anthropic’s API business is projected to surge to $3.907B in 2025 (YoY +662%), while OpenAI is expected to see only +80% growth over the same period.\n\n• Furthermore, Barclays noted that with Anthropic beginning pre-training of Claude 5 in Q4 this year and adding inference revenue on top, Anthropic alone could lift AWS’s quarterly revenue growth rate by 4 percentage points.\n\n$AMZN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014419105661448661, "ret_signed_m1d": 0.014419105661448661, "alpha_spy_m1d": 0.011514545134333032, "alpha_c_m1d": 0.014206440310548185, "bench_spy_m1d": 0.0029045605271156294, "bench_c_m1d": 0.0002126653509004761, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015107797009308976, "ret_signed_p1d": 0.015107797009308976, "alpha_spy_p1d": 0.012651255100595504, "alpha_c_p1d": 0.011237261512994534, "bench_spy_p1d": 0.002456541908713472, "bench_c_p1d": 0.0038705354963144423, "ret_p1w": -0.017991683818820436, "ret_signed_p1w": -0.017991683818820436, "alpha_spy_p1w": -0.03370457454412168, "alpha_c_p1w": -0.032963442016880395, "bench_spy_p1w": 0.015712890725301243, "bench_c_p1w": 0.014971758198059959, "ret_p1m": -0.04919729627894531, "ret_signed_p1m": -0.04919729627894531, "alpha_spy_p1m": -0.08972629110053099, "alpha_c_p1m": -0.07137036801518493, "bench_spy_p1m": 0.04052899482158567, "bench_c_p1m": 0.022173071736239613, "ret_p3m": 0.00021522425585907, "ret_signed_p3m": 0.00021522425585907, "alpha_spy_p3m": -0.05933929821202266, "alpha_c_p3m": -0.018080588678006215, "bench_spy_p3m": 0.05955452246788173, "bench_c_p3m": 0.018295812933865285, "ret_p6m": -0.10304309496289199, "ret_signed_p6m": -0.10304309496289199, "alpha_spy_p6m": -0.16959768247953932, "alpha_c_p6m": -0.08856233393164392, "bench_spy_p6m": 0.06655458751664733, "bench_c_p6m": -0.014480761031248068, "price_path": [-0.0634, -0.0823, -0.0822, -0.0871, -0.0699, -0.0754, -0.0853, -0.0853, -0.0974, -0.1027, -0.0842, -0.0875, -0.0655, -0.0389, -0.0557, -0.0511, -0.0534, -0.0384, -0.0384, -0.0381, -0.0558, -0.0421, -0.0433, -0.0315, -0.0286, -0.0257, -0.0393, -0.0364, -0.0267, -0.013, -0.0209, -0.0174, -0.0004, -0.0038, 0.002, -0.0057, -0.0092, 0.0077, -0.0757, -0.089, -0.08, -0.0431, -0.0396, -0.0415, -0.0475, -0.0467, -0.0334, -0.0058, -0.0056, -0.0036, -0.0186, -0.0367, -0.0447, -0.015, -0.0189, -0.0156, -0.0138, -0.0031, -0.0143, -0.0143, -0.0301, -0.0273, 0.0144, 0.0, 0.0151, 0.0254, -0.0086, -0.0102, -0.018, -0.0039, 0.0074, -0.0031, -0.0047, -0.0037, -0.0202, -0.05, -0.0522, -0.061, -0.054, -0.0437, -0.0549, -0.0504, -0.0427, -0.0552, -0.0492, -0.0454, -0.0306, -0.0198, -0.0687, -0.0528, -0.0686, -0.0721, -0.0769, -0.083, -0.0682, -0.0443, -0.0619, -0.0484, -0.035, -0.0231, -0.0133, -0.0087, -0.0408, 0.0512, 0.0933, 0.0731, 0.0769, 0.0461, 0.052, 0.0692, 0.0722, 0.0511, 0.0226, 0.0102, 0.0023, -0.0421, -0.0415, -0.0654, -0.0501, -0.026, -0.0114, -0.0136, -0.0136, 0.0038, 0.0067, 0.009, 0.0002, -0.0139, -0.0121, -0.0234, -0.019, -0.0024, -0.0088, -0.0264, -0.0421, -0.0421, -0.0476, -0.024, -0.0214, -0.0168, -0.0008, 0.0002, 0.0002, 0.0008, -0.0011, 0.0009, -0.0065, -0.0065, -0.0251, 0.0031, 0.037, 0.0397, 0.0601, 0.0648, 0.0609, 0.0442, 0.0186, 0.0252, 0.0292, 0.0292, -0.0057, -0.0044, 0.0087, 0.0294, 0.0262, 0.0532, 0.046, 0.0405, 0.03, 0.0458, 0.0271, 0.0028, -0.0415, -0.0947, -0.1016, -0.1092, -0.1216, -0.1409, -0.1444, -0.1444, -0.1342, -0.1185, -0.1182, -0.0956, -0.1165, -0.1023, -0.0934, -0.1051, -0.0961, -0.103]}
{"unit_id": "thread:1967523158040429044", "ticker_idx": 0, "ts": "2025-09-15T09:37:30+00:00", "symbol": "LG Display", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "LG Display will accelerate its OLED-focused business transition and achieve improved earnings not only this year but also next year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "LGD gains iPhone OLED share from BOE, expanding AP4 line for iPhone 18; equipment suppliers Shinsung ENG, DMS, and Jusung Engineering seen as beneficiaries.", "resolved_tickers": ["034220.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "LG Display 034220.KS Korea", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] LGD to Supply iPhone Panels Instead of China’s BOE- Fnnews\n\nLG Display has begun expanding its Paju-based AP4 plant line to respond to increased orders for Apple iPhone OLED (organic light-emitting diode) panels. This comes as part of the iPhone panel volume previously handled by China’s BOE has shifted to LG Display. Analysts say this once again proves big tech’s trust in Korean display products. There is growing consensus that Korea should leverage its technology and quality edge to further expand its strategic gains during the global supply chain restructuring triggered by U.S.-China tensions.\n\nAccording to Financial News’ coverage on the 15th, LG Display last month formed a task force (TF) for the expansion of the Paju AP4 line.\n\nAP4 is LG Display’s 6th-generation mobile OLED production line, where TFT (thin-film transistor) substrates used in iPhone panels are processed. Back in late 2023, LG Display increased the production capacity of the AP4 plant from 30,000 sheets per month to about 45,000. Now, nearly two years later, it has decided on another expansion. The TF is expected to complete equipment selection this month, after which sequential purchase orders will be placed with equipment partners starting in October. Industry watchers expect that, with LG Display’s expansion of this line for the first time in years, the company will accelerate its OLED-focused business transition and achieve improved earnings not only this year but also next year.\n\nApple has been reducing BOE’s share of iPhone panel supply, and LG Display has absorbed part of that volume—hence the push for expansion. According to industry estimates, Samsung Display currently supplies 50% of iPhone OLED panels, LG Display 30%, and BOE had raised its share to about 20%. However, with U.S.-China trade conflicts dragging on, restrictions on Chinese firms are intensifying, and Apple, as a U.S. strategic partner, appears to be restructuring its supply chain to reduce reliance on China.\n\nThrough this expansion, LG Display is expected to secure an additional capacity of 15,000 sheets per month, bringing AP4’s total capacity to 60,000. The bulk of the 1.26 trillion won OLED investment plan approved by the board in June is also interpreted as being allocated to the AP4 line expansion. Mass production at the expanded line is expected to begin next year, with the new volume likely to be adopted in Apple’s next-generation iPhone (iPhone 18 series), anticipated for release in the second half of next year.\n\nThese developments have also heightened expectations for a rebound in earnings. LG Display posted an operating loss of 116 billion won in Q2 this year, reverting to a quarterly deficit after achieving consecutive profits in Q4 2023 and Q1 2024.\n\nStill, the company has pledged to drive toward “annual profit” from the second half onward, based on an upgraded focus on OLED panels, raising expectations. Industry insiders believe that not only in the second half of this year but also with expanded iPhone OLED orders next year, LG Display will stabilize earnings and return to annual profitability.\n\nThe expansion is also expected to bring a boost to equipment makers. As LG Display pushes ahead with OLED investments, suppliers are expected to benefit from new contracts. Shinsung ENG is seen as a strong candidate to supply cleanroom systems for OLED production spaces. DMS has the capability to supply LG Display with wet process equipment such as cleaning, etching, stripping, and developing in an integrated package. Jusung Engineering is focusing on oxide deposition equipment as well as encapsulation deposition equipment.\n\nAn industry official said, “As LG Display accelerates its expansion of iPhone OLED lines, a trickle-down effect is expected across the domestic display equipment ecosystem.”\n\n---\n\nhttps://t.co/bc8XJoayU3", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004149377593360981, "ret_signed_m1d": 0.004149377593360981, "alpha_spy_m1d": 0.009445042516847635, "alpha_c_m1d": 0.013076490624647774, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.008927113031286793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10456431535269717, "ret_signed_p1d": 0.10456431535269717, "alpha_spy_p1d": 0.10594110813797886, "alpha_c_p1d": 0.10804009348181953, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": -0.0034757781291223644, "ret_p1w": 0.10705394190871376, "ret_signed_p1w": 0.10705394190871376, "alpha_spy_p1w": 0.09528410053465164, "alpha_c_p1w": 0.07155244006180017, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.035501501846913586, "ret_p1m": 0.16431535269709552, "ret_signed_p1m": 0.16431535269709552, "alpha_spy_p1m": 0.15954011786791678, "alpha_c_p1m": 0.13302673519782404, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.031288617499271476, "ret_p3m": 0.03568464730290466, "ret_signed_p3m": 0.03568464730290466, "alpha_spy_p3m": -0.00996558437137618, "alpha_c_p3m": -0.04846709814780459, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.08415174545070925, "ret_p6m": -0.0058091286307053736, "ret_signed_p6m": -0.0058091286307053736, "alpha_spy_p6m": -0.0363034636448194, "alpha_c_p6m": -0.03136064944058714, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.025551520809881767, "price_path": [-0.2357, -0.2373, -0.2349, -0.2523, -0.2174, -0.2249, -0.2282, -0.2432, -0.2564, -0.2407, -0.2432, -0.21, -0.2282, -0.2398, -0.2398, -0.2299, -0.2266, -0.2332, -0.2315, -0.2332, -0.2382, -0.2357, -0.2357, -0.239, -0.244, -0.2357, -0.2349, -0.166, -0.1369, -0.1411, -0.127, -0.1004, -0.1237, -0.1295, -0.1054, -0.0913, -0.0871, -0.0929, -0.083, -0.0996, 0.1029, -0.0141, -0.0141, -0.0091, -0.0398, -0.0556, -0.0515, -0.0539, -0.0324, -0.0241, -0.0191, -0.0066, -0.0041, -0.0324, -0.0257, -0.0041, 0.0008, -0.005, -0.0116, -0.0166, -0.034, -0.0249, 0.0041, 0.0, 0.1046, 0.0689, 0.0888, 0.0888, 0.1071, 0.1054, 0.1328, 0.1867, 0.1643, 0.2141, 0.1959, 0.1975, 0.2166, 0.2166, 0.2166, 0.2166, 0.2166, 0.2166, 0.2963, 0.2606, 0.1643, 0.2025, 0.1934, 0.1925, 0.2058, 0.1693, 0.1776, 0.1743, 0.1427, 0.1817, 0.1876, 0.2033, 0.1577, 0.2199, 0.205, 0.2091, 0.1627, 0.1651, 0.1129, 0.1394, 0.1195, 0.1286, 0.117, 0.0946, 0.0714, 0.0066, 0.0274, 0.0373, 0.0083, 0.0141, 0.0149, 0.0473, 0.0539, 0.0407, 0.0232, 0.0382, 0.0614, 0.0664, 0.1012, 0.0739, 0.0656, 0.0523, 0.0357, 0.039, 0.0207, 0.0, 0.0091, -0.0133, -0.0066, -0.0232, -0.0108, 0.0, 0.0, -0.0232, -0.0041, -0.0199, -0.0199, -0.0199, -0.0183, -0.0025, 0.0216, 0.0407, 0.0149, -0.0224, -0.0058, 0.0066, 0.0, 0.0158, 0.039, 0.034, 0.0349, 0.0116, 0.0506, 0.0398, 0.0324, 0.0199, 0.0066, -0.0199, -0.0315, -0.0838, -0.0564, -0.0415, -0.0498, -0.0705, -0.0481, -0.0315, -0.0116, -0.0116, -0.0373, -0.0373, -0.0373, -0.0373, 0.0191, 0.1046, 0.1585, 0.1668, 0.2299, 0.3154, 0.19, 0.19, 0.0672, -0.0871, -0.0349, -0.0075, -0.0598, -0.0058]}
{"unit_id": "thread:1967523158040429044", "ticker_idx": 1, "ts": "2025-09-15T09:37:30+00:00", "symbol": "Shinsung ENG", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Shinsung ENG is seen as a strong candidate to supply cleanroom systems for OLED production spaces", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "LGD gains iPhone OLED share from BOE, expanding AP4 line for iPhone 18; equipment suppliers Shinsung ENG, DMS, and Jusung Engineering seen as beneficiaries.", "resolved_tickers": ["011930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Shinsung ENG (Korea cleanroom/display equipment) listed as 011930.KS", "bench_c": "XLI", "bench_c_reason": "industry='Building Products & Equipment'", "bench_c_industry": "Building Products & Equipment", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] LGD to Supply iPhone Panels Instead of China’s BOE- Fnnews\n\nLG Display has begun expanding its Paju-based AP4 plant line to respond to increased orders for Apple iPhone OLED (organic light-emitting diode) panels. This comes as part of the iPhone panel volume previously handled by China’s BOE has shifted to LG Display. Analysts say this once again proves big tech’s trust in Korean display products. There is growing consensus that Korea should leverage its technology and quality edge to further expand its strategic gains during the global supply chain restructuring triggered by U.S.-China tensions.\n\nAccording to Financial News’ coverage on the 15th, LG Display last month formed a task force (TF) for the expansion of the Paju AP4 line.\n\nAP4 is LG Display’s 6th-generation mobile OLED production line, where TFT (thin-film transistor) substrates used in iPhone panels are processed. Back in late 2023, LG Display increased the production capacity of the AP4 plant from 30,000 sheets per month to about 45,000. Now, nearly two years later, it has decided on another expansion. The TF is expected to complete equipment selection this month, after which sequential purchase orders will be placed with equipment partners starting in October. Industry watchers expect that, with LG Display’s expansion of this line for the first time in years, the company will accelerate its OLED-focused business transition and achieve improved earnings not only this year but also next year.\n\nApple has been reducing BOE’s share of iPhone panel supply, and LG Display has absorbed part of that volume—hence the push for expansion. According to industry estimates, Samsung Display currently supplies 50% of iPhone OLED panels, LG Display 30%, and BOE had raised its share to about 20%. However, with U.S.-China trade conflicts dragging on, restrictions on Chinese firms are intensifying, and Apple, as a U.S. strategic partner, appears to be restructuring its supply chain to reduce reliance on China.\n\nThrough this expansion, LG Display is expected to secure an additional capacity of 15,000 sheets per month, bringing AP4’s total capacity to 60,000. The bulk of the 1.26 trillion won OLED investment plan approved by the board in June is also interpreted as being allocated to the AP4 line expansion. Mass production at the expanded line is expected to begin next year, with the new volume likely to be adopted in Apple’s next-generation iPhone (iPhone 18 series), anticipated for release in the second half of next year.\n\nThese developments have also heightened expectations for a rebound in earnings. LG Display posted an operating loss of 116 billion won in Q2 this year, reverting to a quarterly deficit after achieving consecutive profits in Q4 2023 and Q1 2024.\n\nStill, the company has pledged to drive toward “annual profit” from the second half onward, based on an upgraded focus on OLED panels, raising expectations. Industry insiders believe that not only in the second half of this year but also with expanded iPhone OLED orders next year, LG Display will stabilize earnings and return to annual profitability.\n\nThe expansion is also expected to bring a boost to equipment makers. As LG Display pushes ahead with OLED investments, suppliers are expected to benefit from new contracts. Shinsung ENG is seen as a strong candidate to supply cleanroom systems for OLED production spaces. DMS has the capability to supply LG Display with wet process equipment such as cleaning, etching, stripping, and developing in an integrated package. Jusung Engineering is focusing on oxide deposition equipment as well as encapsulation deposition equipment.\n\nAn industry official said, “As LG Display accelerates its expansion of iPhone OLED lines, a trickle-down effect is expected across the domestic display equipment ecosystem.”\n\n---\n\nhttps://t.co/bc8XJoayU3", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03737780333525009, "ret_signed_m1d": -0.03737780333525009, "alpha_spy_m1d": -0.03208213841176344, "alpha_c_m1d": -0.03278100081774571, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.004596802517504384, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.022426682001150056, "ret_signed_p1d": -0.022426682001150056, "alpha_spy_p1d": -0.021049889215868367, "alpha_c_p1d": -0.01947155850973059, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": -0.0029551234914194646, "ret_p1w": 0.03967797584818866, "ret_signed_p1w": 0.03967797584818866, "alpha_spy_p1w": 0.027908134474126545, "alpha_c_p1w": 0.030798836126471274, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.008879139721717388, "ret_p1m": -0.0011500862564692849, "ret_signed_p1m": -0.0011500862564692849, "alpha_spy_p1m": -0.005925321085648028, "alpha_c_p1m": -0.011611778235771197, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.010461691979301913, "ret_p3m": -0.07245543415756184, "ret_signed_p3m": -0.07245543415756184, "alpha_spy_p3m": -0.11810566583184268, "alpha_c_p3m": -0.11252404550769435, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.04006861135013251, "ret_p6m": 0.2507188039102932, "ret_signed_p6m": 0.2507188039102932, "alpha_spy_p6m": 0.2202244688961792, "alpha_c_p6m": 0.12577027406600627, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.12494852984428695, "price_path": [-0.1553, -0.1489, -0.1271, -0.1374, -0.1173, -0.13, -0.1403, -0.1415, -0.1581, -0.1317, -0.1323, -0.1121, -0.1323, -0.1489, -0.1432, -0.1277, -0.0224, -0.0311, -0.0385, -0.0477, -0.0707, -0.0822, -0.0822, -0.073, -0.0955, -0.0851, -0.1001, -0.1012, -0.0978, -0.1196, -0.1202, -0.1196, -0.157, -0.1673, -0.1547, -0.1317, -0.1156, -0.1248, -0.1133, -0.0449, -0.0621, -0.0742, -0.0742, -0.0799, -0.0932, -0.1162, -0.1248, -0.1213, -0.0909, -0.0897, -0.0799, -0.0909, -0.1024, -0.1208, -0.1058, -0.1081, -0.0857, -0.0857, -0.0512, -0.0466, -0.0345, -0.0489, -0.0374, 0.0, -0.0224, 0.0224, 0.0334, 0.0334, 0.0397, 0.065, 0.0564, 0.0311, 0.0029, 0.0017, -0.0121, -0.0052, 0.0311, 0.0311, 0.0311, 0.0311, 0.0311, 0.0311, 0.061, 0.0265, -0.0012, 0.0408, 0.0552, 0.0276, 0.069, 0.0017, -0.0063, 0.0006, 0.088, 0.1219, 0.1673, 0.1282, 0.0983, 0.0627, 0.1587, 0.1409, 0.0817, 0.0598, 0.0213, 0.0322, 0.042, 0.0155, -0.0109, -0.0656, -0.065, -0.1144, -0.1288, -0.0926, -0.1334, -0.1369, -0.1317, -0.1047, -0.0776, -0.0696, -0.0759, -0.0759, -0.0673, -0.0782, -0.065, -0.0667, -0.0656, -0.073, -0.0725, -0.0667, -0.0748, -0.111, -0.0886, -0.1116, -0.0949, -0.0707, -0.0845, -0.0989, -0.0989, -0.0978, -0.0725, -0.0863, -0.0863, -0.0863, -0.0679, -0.042, -0.0437, -0.069, -0.1041, -0.1185, -0.1029, -0.0932, -0.0443, -0.0702, -0.0972, -0.0817, -0.0523, -0.0719, -0.0512, -0.0339, -0.0173, -0.0207, -0.0167, 0.0408, 0.0236, 0.0075, 0.0575, 0.1265, 0.2105, 0.268, 0.3916, 0.3974, 0.4117, 0.4117, 0.3341, 0.3341, 0.3341, 0.3341, 0.3542, 0.3341, 0.2018, 0.2335, 0.199, 0.2277, 0.4089, 0.4089, 0.2421, 0.0661, 0.245, 0.3283, 0.2076, 0.2507]}
{"unit_id": "thread:1967523158040429044", "ticker_idx": 3, "ts": "2025-09-15T09:37:30+00:00", "symbol": "Jusung Engineering", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Jusung Engineering is focusing on oxide deposition equipment as well as encapsulation deposition equipment", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "LGD gains iPhone OLED share from BOE, expanding AP4 line for iPhone 18; equipment suppliers Shinsung ENG, DMS, and Jusung Engineering seen as beneficiaries.", "resolved_tickers": ["036930.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "Jusung Engineering 036930.KQ Korea KOSDAQ deposition equip", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] LGD to Supply iPhone Panels Instead of China’s BOE- Fnnews\n\nLG Display has begun expanding its Paju-based AP4 plant line to respond to increased orders for Apple iPhone OLED (organic light-emitting diode) panels. This comes as part of the iPhone panel volume previously handled by China’s BOE has shifted to LG Display. Analysts say this once again proves big tech’s trust in Korean display products. There is growing consensus that Korea should leverage its technology and quality edge to further expand its strategic gains during the global supply chain restructuring triggered by U.S.-China tensions.\n\nAccording to Financial News’ coverage on the 15th, LG Display last month formed a task force (TF) for the expansion of the Paju AP4 line.\n\nAP4 is LG Display’s 6th-generation mobile OLED production line, where TFT (thin-film transistor) substrates used in iPhone panels are processed. Back in late 2023, LG Display increased the production capacity of the AP4 plant from 30,000 sheets per month to about 45,000. Now, nearly two years later, it has decided on another expansion. The TF is expected to complete equipment selection this month, after which sequential purchase orders will be placed with equipment partners starting in October. Industry watchers expect that, with LG Display’s expansion of this line for the first time in years, the company will accelerate its OLED-focused business transition and achieve improved earnings not only this year but also next year.\n\nApple has been reducing BOE’s share of iPhone panel supply, and LG Display has absorbed part of that volume—hence the push for expansion. According to industry estimates, Samsung Display currently supplies 50% of iPhone OLED panels, LG Display 30%, and BOE had raised its share to about 20%. However, with U.S.-China trade conflicts dragging on, restrictions on Chinese firms are intensifying, and Apple, as a U.S. strategic partner, appears to be restructuring its supply chain to reduce reliance on China.\n\nThrough this expansion, LG Display is expected to secure an additional capacity of 15,000 sheets per month, bringing AP4’s total capacity to 60,000. The bulk of the 1.26 trillion won OLED investment plan approved by the board in June is also interpreted as being allocated to the AP4 line expansion. Mass production at the expanded line is expected to begin next year, with the new volume likely to be adopted in Apple’s next-generation iPhone (iPhone 18 series), anticipated for release in the second half of next year.\n\nThese developments have also heightened expectations for a rebound in earnings. LG Display posted an operating loss of 116 billion won in Q2 this year, reverting to a quarterly deficit after achieving consecutive profits in Q4 2023 and Q1 2024.\n\nStill, the company has pledged to drive toward “annual profit” from the second half onward, based on an upgraded focus on OLED panels, raising expectations. Industry insiders believe that not only in the second half of this year but also with expanded iPhone OLED orders next year, LG Display will stabilize earnings and return to annual profitability.\n\nThe expansion is also expected to bring a boost to equipment makers. As LG Display pushes ahead with OLED investments, suppliers are expected to benefit from new contracts. Shinsung ENG is seen as a strong candidate to supply cleanroom systems for OLED production spaces. DMS has the capability to supply LG Display with wet process equipment such as cleaning, etching, stripping, and developing in an integrated package. Jusung Engineering is focusing on oxide deposition equipment as well as encapsulation deposition equipment.\n\nAn industry official said, “As LG Display accelerates its expansion of iPhone OLED lines, a trickle-down effect is expected across the domestic display equipment ecosystem.”\n\n---\n\nhttps://t.co/bc8XJoayU3", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03697478858939185, "ret_signed_m1d": -0.03697478858939185, "alpha_spy_m1d": -0.031679123665905196, "alpha_c_m1d": -0.030282039950361717, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.006692748639030133, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01680675198426429, "ret_signed_p1d": 0.01680675198426429, "alpha_spy_p1d": 0.01818354476954598, "alpha_c_p1d": -0.002135021043472074, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.018941773027736364, "ret_p1w": 0.08739497876619495, "ret_signed_p1w": 0.08739497876619495, "alpha_spy_p1w": 0.07562513739213284, "alpha_c_p1w": 0.057467038456584074, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.02992794030961088, "ret_p1m": 0.035294146278960215, "ret_signed_p1m": 0.035294146278960215, "alpha_spy_p1m": 0.030518911449781472, "alpha_c_p1m": -0.008903291283492987, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.0441974375624532, "ret_p3m": -0.03529408050297056, "ret_signed_p3m": -0.03529408050297056, "alpha_spy_p3m": -0.0809443121772514, "alpha_c_p3m": -0.2344353179689661, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.19914123746599555, "ret_p6m": 1.3574114674912696, "ret_signed_p6m": 1.3574114674912696, "alpha_spy_p6m": 1.3269171324771556, "alpha_c_p6m": 0.6750452475988864, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.6823662198923832, "price_path": [0.0672, 0.0639, 0.0588, 0.0168, 0.0605, 0.079, 0.0992, 0.0689, 0.0437, 0.0319, 0.037, 0.0605, 0.0269, 0.0151, 0.0303, 0.0118, 0.0168, 0.0118, 0.0151, 0.0218, 0.0218, 0.0202, 0.0134, 0.0168, 0.0017, -0.0084, -0.0202, -0.0286, 0.0, -0.0202, -0.0067, -0.084, -0.1143, -0.0908, -0.0891, -0.0706, -0.0773, -0.0605, -0.0487, -0.0605, -0.0454, -0.0437, -0.0437, -0.0723, -0.0824, -0.0992, -0.1126, -0.1092, -0.0958, -0.0975, -0.1076, -0.079, -0.079, -0.116, -0.1025, -0.1025, -0.0655, -0.0706, -0.0655, -0.0571, -0.0487, -0.0336, -0.037, 0.0, 0.0168, 0.005, 0.0437, 0.0437, 0.0874, 0.0756, 0.0924, 0.079, 0.0319, 0.0521, 0.0387, 0.0504, 0.0437, 0.0437, 0.0437, 0.0437, 0.0437, 0.0437, 0.0941, 0.084, 0.0353, 0.0689, 0.0706, 0.042, 0.0571, 0.0353, 0.0303, 0.0151, 0.042, 0.0487, 0.0218, 0.0756, 0.042, 0.0319, 0.0504, 0.0706, 0.0218, 0.037, -0.0168, 0.0084, 0.0151, 0.0235, 0.0218, -0.0151, 0.005, -0.0218, -0.0555, -0.0487, -0.1025, -0.1076, -0.1109, -0.1008, -0.0891, -0.0807, -0.0134, 0.0202, 0.0286, -0.0084, -0.0168, -0.0168, -0.0235, -0.0336, -0.0353, -0.0538, -0.0689, -0.0975, -0.0807, -0.0908, -0.0958, -0.0622, -0.0622, -0.0824, -0.0824, -0.0605, -0.0537, -0.0671, -0.0671, -0.0671, 0.0255, 0.0507, 0.0928, 0.044, 0.177, 0.1248, 0.1012, 0.1147, 0.1147, 0.0928, 0.076, 0.0659, 0.0507, 0.0406, 0.0558, 0.0659, 0.1282, 0.2309, 0.2915, 0.3067, 0.3926, 0.3757, 0.5121, 0.6165, 0.4869, 0.5845, 0.7479, 0.6552, 0.6468, 0.667, 0.7175, 0.7175, 0.7175, 0.7175, 0.8523, 0.8792, 0.7849, 0.7815, 0.7681, 0.8893, 0.8893, 0.8893, 0.987, 0.8118, 1.2496, 1.2597, 1.0173, 1.3574]}
{"unit_id": "orig:1943452344508912076", "ticker_idx": 0, "ts": "2025-07-10T23:28:41+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's 2Q25 revenue met its guidance and slightly exceeded market expectations, despite negative exchange rate effects [strong TWD]", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish TSMC on AI-driven node demand and FX-adjusted beat; domestic memory IDMs to benefit from DRAM ASP rebound; focus on AI-linked CCL/PCB names.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250711_TSMC: June 2025 Performance\n\n[Implication]\n\n(1) TSMC's 2Q25 revenue met its guidance and slightly exceeded market expectations, despite negative exchange rate effects [strong TWD].\n\n(2) TSMC is currently experiencing 1) continued demand for advanced nodes, centered on N3, for AI GPUs and ASICs, 2) anticipation for N2 for Apple in the second half of the year, and 3) even urgent orders for older nodes recently.\n\n(3) Meanwhile, it is also notable that Nanya mentioned 1) a QoQ increase of +70% in bit shipments and 2) the possibility of a DRAM ASP rebound in Q3. This is because a rebound in D3 and D4 ASPs is highly likely to promote buyers' transition to D5, thereby positively impacting domestic memory IDMs.\n\n(4) Furthermore, starting next week with TSMC's 2Q25 conference call, earnings announcements from major US tech companies are scheduled from mid to late July. Accordingly, there is a high probability that additional momentum will flow into the domestic memory industry.\n\n(5) Meanwhile, material, component, and equipment companies are in a somewhat conservative situation due to persistent uncertainty in IDM CAPEX and continued sluggish demand for consumer products. Therefore, attention should be paid to CCL/PCB companies directly related to AI products.\n\n[Contents]\n\n(1) TSMC June revenue increased YoY +27%. 2Q25 revenue increased YoY QoQ +17%.\n\n(2) In Taiwan Dollar terms, 2Q25 top line was NT933,792mnvs.NT927,996mn.\n\n(3) Considering that TWD strengthened by +13.7% against the USD from April 1st to June 30th, it is essentially a huge surprise.\n\n(4) By node, the continued ramp-up of N3, sustained strong performance of N5, and even urgent orders for some older generation processes are expected to result in a GPM that at least meets guidance, similar to the top line.\n\n(5) Below are the opinions from each firm.\n\n(6) Bernstein ■ Top line is consistent with the midpoint of guidance and slightly exceeds consensus by +0.5%. Considering the negative FX effect of approximately 5%, 2Q25 performance in USD terms exceeded the upper end by +3%. ■ Future price increases are expected to offset future FX headwinds.\n(7) Wedbush ■ Our estimate is $30.2bn, which is approximately +5% higher than TSMC's guidance midpoint. ■ This strong performance is evidence of continued momentum in AI semiconductors.\n(8) JPMorgan ■ '25 and '26 estimates adjusted down by -1% and -2% respectively, reflecting FX. ■ Q3 guidance is expected to show +3~6% growth QoQ in USD terms, and GPM is expected to slightly decline to around 56% in the second half due to FX impact. ■ N3 demand is expected to remain strong throughout the year, with additional upside expected in the second half due to the N3 transition of AI accelerators.\n(9) Mizuho ■ Annual revenue growth guidance [+25%] could easily be raised, and it is possible to exceed the market consensus estimate of 27%. ■ However, +27% growth alone may not be enough for the stock price to maintain momentum until the end of '25.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00909101104819332, "ret_signed_m1d": -0.00909101104819332, "alpha_spy_m1d": -0.006278632609499124, "alpha_c_m1d": -0.0017873142690471022, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0035154238705955576, "alpha_c_p1d": 0.00010434156744454359, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": 0.027272584532529454, "ret_signed_p1w": 0.027272584532529454, "alpha_spy_p1w": 0.023725293465524322, "alpha_c_p1w": 0.012699687574086127, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": 0.06818174171385505, "ret_signed_p1m": 0.06818174171385505, "alpha_spy_p1m": 0.050029635847881115, "alpha_c_p1m": 0.04727882515410298, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.3097635926647444, "ret_signed_p3m": 0.3097635926647444, "alpha_spy_p3m": 0.2376101619495865, "alpha_c_p3m": 0.13749726187855194, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 0.45149494358142706, "ret_signed_p6m": 0.45149494358142706, "alpha_spy_p6m": 0.35359409678936693, "alpha_c_p6m": 0.14911052582334405, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.217, -0.2061, -0.226, -0.2333, -0.2305, -0.2441, -0.2613, -0.2097, -0.2179, -0.1961, -0.188, -0.1835, -0.178, -0.178, -0.14, -0.1509, -0.1672, -0.1599, -0.169, -0.1409, -0.1337, -0.1228, -0.0957, -0.1011, -0.0966, -0.1092, -0.1147, -0.1038, -0.111, -0.1092, -0.1165, -0.1264, -0.1246, -0.1246, -0.1246, -0.1436, -0.14, -0.1038, -0.0966, -0.0993, -0.0902, -0.054, -0.0359, -0.05, -0.0636, -0.0682, -0.05, -0.0409, -0.0591, -0.0409, -0.0727, -0.0455, -0.0273, -0.0227, -0.0182, -0.0364, -0.0136, -0.0136, -0.0091, -0.0136, -0.0182, -0.0182, -0.0091, 0.0, 0.0, -0.0045, 0.0091, 0.0273, 0.0273, 0.05, 0.0455, 0.0273, 0.0409, 0.0409, 0.0409, 0.0409, 0.0318, 0.05, 0.0545, 0.0545, 0.0318, 0.0455, 0.0227, 0.0727, 0.0682, 0.0727, 0.0727, 0.0909, 0.0682, 0.0727, 0.0727, 0.0773, 0.0318, 0.0455, 0.0318, 0.0636, 0.0682, 0.0818, 0.0545, 0.0545, 0.0591, 0.0545, 0.0545, 0.0545, 0.0727, 0.0727, 0.0909, 0.1136, 0.1273, 0.1455, 0.1409, 0.1683, 0.1546, 0.1729, 0.1546, 0.182, 0.2231, 0.2231, 0.2048, 0.1865, 0.1865, 0.1911, 0.2094, 0.2459, 0.2778, 0.2778, 0.3098, 0.2915, 0.3143, 0.3143, 0.2915, 0.3006, 0.3371, 0.3554, 0.3235, 0.3508, 0.3508, 0.3326, 0.3235, 0.3235, 0.3508, 0.3463, 0.3737, 0.3737, 0.3691, 0.3782, 0.3737, 0.3326, 0.3371, 0.3326, 0.3463, 0.3371, 0.3463, 0.3326, 0.3052, 0.3189, 0.2824, 0.2733, 0.328, 0.2641, 0.255, 0.2915, 0.3143, 0.3098, 0.3143, 0.2869, 0.3052, 0.3235, 0.3189, 0.3326, 0.3645, 0.3508, 0.3737, 0.3462, 0.3553, 0.3279, 0.3141, 0.3096, 0.3096, 0.3096, 0.3416, 0.3645, 0.3691, 0.3691, 0.3828, 0.4011, 0.392, 0.4194, 0.4194, 0.4515]}
{"unit_id": "orig:1943452344508912076", "ticker_idx": 1, "ts": "2025-07-10T23:28:41+00:00", "symbol": "Nanya Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a rebound in D3 and D4 ASPs is highly likely to promote buyers' transition to D5, thereby positively impacting domestic memory IDMs", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish TSMC on AI-driven node demand and FX-adjusted beat; domestic memory IDMs to benefit from DRAM ASP rebound; focus on AI-linked CCL/PCB names.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya Technology Corp listed on TWSE as 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250711_TSMC: June 2025 Performance\n\n[Implication]\n\n(1) TSMC's 2Q25 revenue met its guidance and slightly exceeded market expectations, despite negative exchange rate effects [strong TWD].\n\n(2) TSMC is currently experiencing 1) continued demand for advanced nodes, centered on N3, for AI GPUs and ASICs, 2) anticipation for N2 for Apple in the second half of the year, and 3) even urgent orders for older nodes recently.\n\n(3) Meanwhile, it is also notable that Nanya mentioned 1) a QoQ increase of +70% in bit shipments and 2) the possibility of a DRAM ASP rebound in Q3. This is because a rebound in D3 and D4 ASPs is highly likely to promote buyers' transition to D5, thereby positively impacting domestic memory IDMs.\n\n(4) Furthermore, starting next week with TSMC's 2Q25 conference call, earnings announcements from major US tech companies are scheduled from mid to late July. Accordingly, there is a high probability that additional momentum will flow into the domestic memory industry.\n\n(5) Meanwhile, material, component, and equipment companies are in a somewhat conservative situation due to persistent uncertainty in IDM CAPEX and continued sluggish demand for consumer products. Therefore, attention should be paid to CCL/PCB companies directly related to AI products.\n\n[Contents]\n\n(1) TSMC June revenue increased YoY +27%. 2Q25 revenue increased YoY QoQ +17%.\n\n(2) In Taiwan Dollar terms, 2Q25 top line was NT933,792mnvs.NT927,996mn.\n\n(3) Considering that TWD strengthened by +13.7% against the USD from April 1st to June 30th, it is essentially a huge surprise.\n\n(4) By node, the continued ramp-up of N3, sustained strong performance of N5, and even urgent orders for some older generation processes are expected to result in a GPM that at least meets guidance, similar to the top line.\n\n(5) Below are the opinions from each firm.\n\n(6) Bernstein ■ Top line is consistent with the midpoint of guidance and slightly exceeds consensus by +0.5%. Considering the negative FX effect of approximately 5%, 2Q25 performance in USD terms exceeded the upper end by +3%. ■ Future price increases are expected to offset future FX headwinds.\n(7) Wedbush ■ Our estimate is $30.2bn, which is approximately +5% higher than TSMC's guidance midpoint. ■ This strong performance is evidence of continued momentum in AI semiconductors.\n(8) JPMorgan ■ '25 and '26 estimates adjusted down by -1% and -2% respectively, reflecting FX. ■ Q3 guidance is expected to show +3~6% growth QoQ in USD terms, and GPM is expected to slightly decline to around 56% in the second half due to FX impact. ■ N3 demand is expected to remain strong throughout the year, with additional upside expected in the second half due to the N3 transition of AI accelerators.\n(9) Mizuho ■ Annual revenue growth guidance [+25%] could easily be raised, and it is possible to exceed the market consensus estimate of 27%. ■ However, +27% growth alone may not be enough for the stock price to maintain momentum until the end of '25.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026511133818604815, "ret_signed_m1d": 0.026511133818604815, "alpha_spy_m1d": 0.02932351225729901, "alpha_c_m1d": 0.03381483059775103, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.007303696779146218, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08907742581162437, "ret_signed_p1d": -0.08907742581162437, "alpha_spy_p1d": -0.08556200194102881, "alpha_c_p1d": -0.08897308424417982, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.00010434156744454359, "ret_p1w": -0.0858961382967276, "ret_signed_p1w": -0.0858961382967276, "alpha_spy_p1w": -0.08944342936373273, "alpha_c_p1w": -0.10046903525517092, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": 0.014572896958443327, "ret_p1m": -0.06786851875674049, "ret_signed_p1m": -0.06786851875674049, "alpha_spy_p1m": -0.08602062462271443, "alpha_c_p1m": -0.08877143531649256, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.02090291655975207, "ret_p3m": 0.9342522586809643, "ret_signed_p3m": 0.9342522586809643, "alpha_spy_p3m": 0.8620988279658064, "alpha_c_p3m": 0.7619859278947718, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.17226633078619247, "ret_p6m": 3.390243760360968, "ret_signed_p6m": 3.390243760360968, "alpha_spy_p6m": 3.2923429135689077, "alpha_c_p6m": 3.0878593426028846, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.302384417758083, "price_path": [-0.2534, -0.2672, -0.3054, -0.3192, -0.2513, -0.1771, -0.1654, -0.1813, -0.2068, -0.2153, -0.2131, -0.211, -0.2365, -0.2365, -0.2439, -0.2736, -0.2492, -0.2651, -0.2587, -0.2715, -0.211, -0.1326, -0.0467, -0.0392, -0.0318, -0.0679, -0.0403, -0.0742, -0.0912, -0.0976, -0.0923, -0.1103, -0.1156, -0.0339, -0.0339, -0.0286, -0.0276, 0.0445, 0.1135, 0.1029, 0.0965, 0.1516, 0.1177, 0.1474, 0.1347, 0.122, 0.141, 0.2534, 0.281, 0.2492, 0.2386, 0.2025, 0.1665, 0.1432, 0.1262, 0.0859, 0.0498, 0.0488, 0.0732, 0.0276, 0.0371, 0.0286, 0.0265, 0.0, -0.0891, -0.0795, -0.1145, -0.122, -0.0859, -0.1039, -0.1103, -0.14, -0.087, -0.088, -0.0827, -0.0817, -0.105, -0.0477, -0.0551, -0.0551, -0.0721, -0.0583, -0.0806, -0.105, -0.0679, -0.071, 0.0064, -0.0233, -0.017, 0.018, 0.0042, -0.0297, -0.0551, -0.0148, -0.0382, -0.0042, 0.0064, 0.0138, -0.0106, -0.0032, -0.0085, -0.0318, -0.017, 0.0223, 0.1241, 0.1347, 0.1432, 0.1538, 0.1983, 0.2174, 0.3383, 0.4677, 0.5546, 0.6967, 0.6797, 0.7137, 0.7179, 0.6119, 0.6246, 0.5058, 0.5058, 0.5483, 0.6161, 0.6903, 0.7603, 0.7603, 0.9343, 0.9258, 1.0891, 1.0891, 1.0276, 0.8897, 0.8303, 1.0127, 1.2057, 1.2694, 1.2375, 1.333, 1.3224, 1.3224, 1.5451, 1.7572, 1.8208, 1.8526, 1.8102, 1.9056, 1.789, 1.9268, 2.1389, 2.1177, 2.4252, 2.4889, 2.4571, 2.4677, 2.3616, 2.5313, 2.404, 2.3934, 2.298, 1.9692, 2.0329, 2.0435, 1.895, 2.0965, 2.0965, 2.1707, 2.1707, 2.2025, 2.2025, 2.245, 2.4677, 2.4358, 2.3086, 2.3086, 2.4358, 2.4464, 2.351, 2.4889, 2.6055, 2.701, 2.8176, 2.7434, 3.0085, 3.0085, 3.0085, 2.9979, 3.1251, 3.0933, 3.0933, 3.3902]}
{"unit_id": "orig:1943452344508912076", "ticker_idx": 2, "ts": "2025-07-10T23:28:41+00:00", "symbol": "CCL/PCB", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "attention should be paid to CCL/PCB companies directly related to AI products", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish TSMC on AI-driven node demand and FX-adjusted beat; domestic memory IDMs to benefit from DRAM ASP rebound; focus on AI-linked CCL/PCB names.", "resolved_tickers": ["006280.KS", "008060.KS"], "resolution_method": "basket", "resolution_reason": "Korea CCL/PCB: Doosan+Iljin Materials basket", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "original", "lang": "en", "tweet_text": "250711_TSMC: June 2025 Performance\n\n[Implication]\n\n(1) TSMC's 2Q25 revenue met its guidance and slightly exceeded market expectations, despite negative exchange rate effects [strong TWD].\n\n(2) TSMC is currently experiencing 1) continued demand for advanced nodes, centered on N3, for AI GPUs and ASICs, 2) anticipation for N2 for Apple in the second half of the year, and 3) even urgent orders for older nodes recently.\n\n(3) Meanwhile, it is also notable that Nanya mentioned 1) a QoQ increase of +70% in bit shipments and 2) the possibility of a DRAM ASP rebound in Q3. This is because a rebound in D3 and D4 ASPs is highly likely to promote buyers' transition to D5, thereby positively impacting domestic memory IDMs.\n\n(4) Furthermore, starting next week with TSMC's 2Q25 conference call, earnings announcements from major US tech companies are scheduled from mid to late July. Accordingly, there is a high probability that additional momentum will flow into the domestic memory industry.\n\n(5) Meanwhile, material, component, and equipment companies are in a somewhat conservative situation due to persistent uncertainty in IDM CAPEX and continued sluggish demand for consumer products. Therefore, attention should be paid to CCL/PCB companies directly related to AI products.\n\n[Contents]\n\n(1) TSMC June revenue increased YoY +27%. 2Q25 revenue increased YoY QoQ +17%.\n\n(2) In Taiwan Dollar terms, 2Q25 top line was NT933,792mnvs.NT927,996mn.\n\n(3) Considering that TWD strengthened by +13.7% against the USD from April 1st to June 30th, it is essentially a huge surprise.\n\n(4) By node, the continued ramp-up of N3, sustained strong performance of N5, and even urgent orders for some older generation processes are expected to result in a GPM that at least meets guidance, similar to the top line.\n\n(5) Below are the opinions from each firm.\n\n(6) Bernstein ■ Top line is consistent with the midpoint of guidance and slightly exceeds consensus by +0.5%. Considering the negative FX effect of approximately 5%, 2Q25 performance in USD terms exceeded the upper end by +3%. ■ Future price increases are expected to offset future FX headwinds.\n(7) Wedbush ■ Our estimate is $30.2bn, which is approximately +5% higher than TSMC's guidance midpoint. ■ This strong performance is evidence of continued momentum in AI semiconductors.\n(8) JPMorgan ■ '25 and '26 estimates adjusted down by -1% and -2% respectively, reflecting FX. ■ Q3 guidance is expected to show +3~6% growth QoQ in USD terms, and GPM is expected to slightly decline to around 56% in the second half due to FX impact. ■ N3 demand is expected to remain strong throughout the year, with additional upside expected in the second half due to the N3 transition of AI accelerators.\n(9) Mizuho ■ Annual revenue growth guidance [+25%] could easily be raised, and it is possible to exceed the market consensus estimate of 27%. ■ However, +27% growth alone may not be enough for the stock price to maintain momentum until the end of '25.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0038583658035736823, "ret_signed_m1d": -0.0038583658035736823, "alpha_spy_m1d": -0.001045987364879486, "alpha_c_m1d": 0.010634355359367853, "bench_spy_m1d": -0.0028123784386941963, "bench_c_m1d": -0.014492721162941535, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0037017703912553546, "ret_signed_p1d": -0.0037017703912553546, "alpha_spy_p1d": -0.00018634652065979695, "alpha_c_p1d": 0.004228249360201997, "bench_spy_p1d": -0.0035154238705955576, "bench_c_p1d": -0.007930019751457351, "ret_p1w": 0.04794523203933376, "ret_signed_p1w": 0.04794523203933376, "alpha_spy_p1w": 0.04439794097232863, "alpha_c_p1w": 0.049039086711312296, "bench_spy_p1w": 0.0035472910670051316, "bench_c_p1w": -0.0010938546719785336, "ret_p1m": -0.006330638756704476, "ret_signed_p1m": -0.006330638756704476, "alpha_spy_p1m": -0.02448274462267841, "alpha_c_p1m": -0.011252771470414369, "bench_spy_p1m": 0.018152105865973933, "bench_c_p1m": 0.0049221327137098925, "ret_p3m": 0.022080610097158504, "ret_signed_p3m": 0.022080610097158504, "alpha_spy_p3m": -0.05007282061799939, "alpha_c_p3m": -0.11751469993694225, "bench_spy_p3m": 0.0721534307151579, "bench_c_p3m": 0.13959531003410075, "ret_p6m": 0.186440891391486, "ret_signed_p6m": 0.186440891391486, "alpha_spy_p6m": 0.08854004459942588, "alpha_c_p6m": -0.24254329240425576, "bench_spy_p6m": 0.09790084679206013, "bench_c_p6m": 0.42898418379574177, "price_path": [-0.1166, -0.0997, -0.1103, -0.0967, -0.0894, -0.0842, -0.0826, -0.0789, -0.0762, -0.0691, -0.0757, -0.063, -0.0676, -0.0676, -0.0688, -0.0688, -0.0688, -0.0928, -0.0901, -0.0917, -0.0878, -0.0806, -0.0736, -0.0708, -0.0939, -0.0876, -0.0865, -0.0845, -0.0777, -0.0754, -0.063, -0.077, -0.0695, -0.0653, -0.0624, -0.0684, -0.0684, -0.0528, -0.0313, -0.0313, -0.0239, -0.0273, -0.0272, -0.0237, -0.0419, -0.0441, -0.048, -0.0423, -0.0482, -0.0433, -0.0569, -0.0481, -0.0478, -0.0547, -0.0675, -0.064, -0.0461, -0.0417, -0.0195, -0.0355, -0.0382, -0.0164, -0.0039, 0.0, -0.0037, -0.0057, 0.0011, 0.003, 0.0479, 0.0704, 0.0724, 0.0735, 0.0582, 0.061, 0.0555, 0.018, 0.0194, 0.0379, 0.0393, -0.0073, -0.0102, 0.0052, 0.0024, 0.0022, -0.0063, -0.0134, -0.0182, -0.0093, -0.0224, -0.0224, -0.036, -0.0471, -0.0489, -0.0527, -0.051, -0.034, -0.0369, -0.0369, -0.0382, -0.0466, -0.0585, -0.0507, -0.0396, -0.0318, -0.0346, -0.0212, -0.0227, -0.0117, -0.01, -0.0056, 0.003, -0.002, -0.0064, -0.0067, -0.0067, 0.0339, 0.0366, 0.0274, 0.026, -0.0105, -0.0055, -0.0139, -0.0022, 0.0221, 0.0221, 0.0221, 0.0221, 0.0221, 0.0221, 0.0363, 0.0254, 0.0111, 0.0271, 0.025, 0.0152, 0.0328, 0.0227, 0.0306, 0.024, 0.0228, 0.0397, 0.0353, 0.0289, 0.0223, 0.0249, 0.0269, 0.0273, 0.0463, 0.0677, 0.0398, 0.0993, 0.0906, 0.1232, 0.1249, 0.1271, 0.1261, 0.1047, 0.1011, 0.1061, 0.0903, 0.0887, 0.103, 0.1135, 0.1111, 0.1196, 0.1219, 0.1427, 0.1761, 0.1499, 0.1504, 0.1538, 0.1632, 0.1773, 0.248, 0.2879, 0.2816, 0.2607, 0.2632, 0.264, 0.2754, 0.251, 0.2126, 0.1941, 0.1941, 0.1853, 0.1633, 0.2142, 0.2142, 0.2142, 0.1864]}
{"unit_id": "orig:1958389641692017015", "ticker_idx": 0, "ts": "2025-08-21T04:44:10+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC will continue to maintain over 90% market share at advanced process nodes, and it maintains an 'Overweight' rating on TSMC with a target price of NT$1,275", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses JPMorgan Overweight on TSMC with NT$1,275 PT; Intel foundry 'illusory competition' is net positive for TSMC's dominance.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Intel’s entry into foundry is actually a benefit for TSMC – JPMorgan\n\nThe “revival” of Intel’s foundry business as a potential threat to TSMC is likely overstated by the market.\n\nOn August 21, JPMorgan’s Technology and Telecoms team said in its latest report that Intel Foundry’s “illusory competition” is actually more beneficial to TSMC.\n\nThe analysts believe that the existence of Intel’s foundry business can in fact help TSMC avoid regulatory pressure stemming from a monopolistic position. JPMorgan also said that customer participation in Intel Foundry’s “revival plan” is not purely negative, and that the fundamental challenges facing Intel Foundry go far beyond capital.\n\nThe bank said TSMC will continue to maintain over 90% market share at advanced process nodes, and it maintains an “Overweight” rating on TSMC with a target price of NT$1,275.\n\nThe illusion of “having a choice” is more valuable than an absolute monopoly\nJPMorgan’s analysts noted that aside from the initial frenzy (in the second half of 2020, when the market expected Intel to massively outsource manufacturing, TSMC’s P/E once reached 25–30x), talk of TSMC potentially becoming a monopolist in advanced-node foundry has not led to a significant uplift in its valuation multiple.\n\nJPMorgan believes that such a potential monopoly would instead invite more scrutiny (from government agencies) and amplify geopolitical risks—factors that typically depress valuation multiples.\n\nThe bank argues that if there is a somewhat weaker competitor in advanced nodes that creates for customers an illusion of “having a choice,” it could actually be a more favorable scenario for TSMC. This is because it can ease ongoing scrutiny from governments and reduce pressure from policy demands such as “manufacturing reshoring to the United States.”\n\nCustomer participation in Intel Foundry’s revival is not purely negative\nThe market may view participation by TSMC’s major customers—such as Apple or Nvidia—in Intel Foundry’s revival plan as a direct loss of TSMC’s market share. But JPMorgan does not see this as purely negative.\n\nThe analysts point out that Intel Foundry would need to execute flawlessly on at least two to three advanced process nodes in order to gain credibility with fabless customers and achieve meaningful scale at the leading edge.\n\nMeanwhile, Intel’s product business may continue to face competitive pressure from AMD, ARM, Nvidia, and in-house chips, forcing Intel to keep outsourcing more products to the most competitive foundry—TSMC.\n\nThe analysts believe that, regardless of how many customers support Intel Foundry, it is unlikely to resolve the inherent conflict of interest between the product business and the foundry business. Therefore, TSMC will continue to maintain over 90% market share at advanced nodes for the foreseeable future.\n\nIntel Foundry’s issues go far beyond funding\nWhile most observers attribute Intel’s foundry problems to funding, JPMorgan believes the fundamental challenges lie elsewhere. The report states:\n\nDue to a severe deterioration in cash flow in its product business, Intel may currently be unable to fund foundry-related expansion, but even when Intel was quite dominant in CPUs and held a process technology lead, its foundry strategy never truly worked.\n\nThe analysts note that the foundry business requires a very different corporate culture, a service mindset, a focus on cost efficiency, and customer-centric innovation—capabilities that are difficult to cultivate quickly for a company that has focused on products for almost its entire existence.\n\nIntel Foundry’s best chance of success is to adopt an N-1 approach (considering that even at N-2 or N-3 nodes there is a lack of a second supplier in the foundry space), which would reduce risk for potential customers and may make capacity ramps easier, rather than competing head-on with TSMC at the cutting edge.\n\nHowever, JPMorgan warns that as the leading domestic foundry recipient of U.S. government investment, such a cautious approach may not be a viable strategy.\n\n$INTC $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0055865865051230035, "ret_signed_m1d": 0.0055865865051230035, "alpha_spy_m1d": 0.0015586046702653622, "alpha_c_m1d": 0.00036841918938024243, "bench_spy_m1d": 0.004027981834857641, "bench_c_m1d": 0.005218167315742761, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024897693556911404, "ret_signed_p1d": 0.024897693556911404, "alpha_spy_p1d": 0.009540807005760765, "alpha_c_p1d": 0.00339895070926155, "bench_spy_p1d": 0.01535688655115064, "bench_c_p1d": 0.021498742847649854, "ret_p1w": 0.04812387295679943, "ret_signed_p1w": 0.04812387295679943, "alpha_spy_p1w": 0.0270870380846846, "alpha_c_p1w": 0.0083965807703128, "bench_spy_p1w": 0.02103683487211483, "bench_c_p1w": 0.03972729218648663, "ret_p1m": 0.1688100658084224, "ret_signed_p1m": 0.1688100658084224, "alpha_spy_p1m": 0.12162243778513582, "alpha_c_p1m": 0.07053546880535433, "bench_spy_p1m": 0.047187628023286576, "bench_c_p1m": 0.09827459700306806, "ret_p3m": 0.22635272604811885, "ret_signed_p3m": 0.22635272604811885, "alpha_spy_p3m": 0.1848767162198044, "alpha_c_p3m": 0.06692174057059797, "bench_spy_p3m": 0.04147600982831445, "bench_c_p3m": 0.15943098547752088, "ret_p6m": 0.6210264131464642, "ret_signed_p6m": 0.6210264131464642, "alpha_spy_p6m": 0.5421810173359916, "alpha_c_p6m": 0.19825912317503636, "bench_spy_p6m": 0.07884539581047267, "bench_c_p6m": 0.4227672899714279, "price_path": [-0.1586, -0.1336, -0.1403, -0.1359, -0.1527, -0.146, -0.1339, -0.1129, -0.1088, -0.1007, -0.0927, -0.0688, -0.0616, -0.0523, -0.0714, -0.0512, -0.0591, -0.0608, -0.0608, -0.0784, -0.0748, -0.0318, -0.0202, -0.0146, 0.0055, -0.0037, -0.0117, 0.0276, 0.0329, 0.0329, 0.0081, 0.0023, 0.0198, 0.0107, 0.0135, 0.0059, 0.0423, 0.045, 0.0804, 0.0575, 0.0507, 0.032, 0.0572, 0.0628, 0.0804, 0.0678, 0.0616, 0.0685, 0.0629, 0.0347, 0.0513, 0.0226, 0.0178, 0.0673, 0.0638, 0.0649, 0.0746, 0.0621, 0.0601, 0.0508, 0.0619, 0.0236, 0.0056, 0.0, 0.0249, 0.0363, 0.0501, 0.0526, 0.0481, 0.0156, 0.0156, 0.0047, 0.0179, 0.0347, 0.0707, 0.0874, 0.1038, 0.1456, 0.1389, 0.1408, 0.1498, 0.1564, 0.1596, 0.1854, 0.1688, 0.2031, 0.2475, 0.2387, 0.2208, 0.2063, 0.2057, 0.2324, 0.273, 0.2714, 0.2894, 0.3344, 0.2975, 0.3438, 0.3233, 0.2385, 0.3366, 0.3059, 0.3446, 0.3231, 0.3021, 0.3137, 0.2996, 0.2748, 0.2829, 0.3016, 0.3161, 0.3306, 0.3463, 0.338, 0.3257, 0.3453, 0.2976, 0.2958, 0.2763, 0.2643, 0.303, 0.2849, 0.2824, 0.2453, 0.2568, 0.2444, 0.2264, 0.246, 0.2245, 0.2138, 0.2561, 0.2562, 0.2795, 0.2795, 0.2864, 0.2695, 0.2889, 0.3038, 0.2926, 0.3005, 0.3321, 0.3389, 0.3686, 0.3489, 0.2922, 0.2732, 0.2693, 0.2255, 0.2596, 0.2785, 0.2977, 0.3139, 0.3221, 0.3221, 0.34, 0.3315, 0.3255, 0.3446, 0.3446, 0.4142, 0.4259, 0.4488, 0.4101, 0.4071, 0.432, 0.468, 0.4655, 0.4474, 0.5116, 0.515, 0.515, 0.4476, 0.443, 0.4485, 0.4817, 0.4721, 0.497, 0.5146, 0.5024, 0.4626, 0.5104, 0.4856, 0.4413, 0.4634, 0.5436, 0.5726, 0.6013, 0.6552, 0.6287, 0.621]}
{"unit_id": "orig:1954705260125343896", "ticker_idx": 0, "ts": "2025-08-11T00:43:45+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA, AMD, Marvell, Broadcom: Core beneficiaries of AI datacenter and accelerator market growth", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA cloud capex review names NVDA, AMD, MRVL, AVGO as core beneficiaries of AI datacenter/accelerator buildout; upward-revised 2025/26 capex forecasts support the theme.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "2Q Cloud Capex Review (BofA)\n\n1. Industry Trends\nThe 2025/26 global cloud AI Capex forecast has been revised upward to +54%/+16% YoY, with total spending expected to reach $432B/$500B, respectively.\nQ2 global hyperscale Capex came in at $112B, up +27% QoQ and +74% YoY, exceeding expectations.\nAll major cloud players are increasing the proportion of spend on servers (especially CPUs/GPUs). Capex intensity remains at 23–25%, well above the historical average of 10–12%.\nGPU shipments to China could add about 10% to AI TAM, and there is potential for easing of U.S.–China trade tariff risks.\nAI infrastructure demand is expanding beyond hyperscalers to neo-cloud providers, enterprises, and national-level projects.\n\n2. Company-Level AI Investment Trends\nGoogle: 2025 Capex outlook raised to $85B, with servers accounting for over two-thirds of spending.\nMicrosoft: Higher proportion of short-term assets such as servers; demand exceeds supply; backlog of $368B.\nMeta: 2026 Capex expected to approach $100B, server-centric; hinted at the possibility of raising external capital.\nAmazon: Q2 Capex +30% QoQ; 2025 Capex guidance of +40% YoY.\nOracle: Sustained high growth driven by cloud expansion.\n\n3. AI TAM Outlook\nAI accelerator market projected to grow from $126B in 2024 to over $650B in 2030, with a 2025–2030 CAGR of +26%.\nAI servers to account for 85% of AI DC systems, reaching ~$700B by 2030.\nHBM market projected to grow from $17B in 2024 to $87B in 2030.\nMarket share expectations: NVIDIA 80%+, AMD 5%, ASIC ecosystem 10–15%.\n\n4. Key Beneficiaries\nNVIDIA, AMD, Marvell, Broadcom: Core beneficiaries of AI datacenter and accelerator market growth.\nSemiconductor equipment, EDA, and optical component companies also expected to benefit.\n\n5. Large-Scale AI Projects by Country\nU.S.: Stargate ($500B, ramping in 2028), Pennsylvania AI Energy Corridor ($90B).\nSaudi Arabia: HUMAIN AI Factories ($30B, 2025–2030).\nUAE: 5GW AI Campus (multi-phase expansion through 2030).\nFrance, Malaysia, Brazil, Italy, and several other countries have AI projects exceeding $1B in scale.\n\n6. Individual Stock Valuation & Risks\nAMD: 33x 2026E EPS; AI growth and CPU market share gains expected, but timing/scale of Middle East projects remain uncertain.\nMarvell (MRVL): 25x 2026E EPS; growth in custom silicon and HBM, but faces integration and rising debt risks.\nNVIDIA (NVDA): 37x 2026E PE; maintaining dominance in the AI accelerator market, but subject to potential headwinds from gaming, China export restrictions, and power supply constraints.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0035153188720971063, "ret_signed_m1d": 0.0035153188720971063, "alpha_spy_m1d": 0.001533957606248304, "alpha_c_m1d": 0.0030721987219517466, "bench_spy_m1d": 0.0019813612658488022, "bench_c_m1d": 0.00044312015014535966, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0060419962243181136, "ret_signed_p1d": 0.0060419962243181136, "alpha_spy_p1d": -0.0046041182313820705, "alpha_c_p1d": -0.016793768235672024, "bench_spy_p1d": 0.010646114455700184, "bench_c_p1d": 0.022835764459990138, "ret_p1w": -0.0002746971562347378, "ret_signed_p1w": -0.0002746971562347378, "alpha_spy_p1w": -0.01187993759060102, "alpha_c_p1w": -0.01206751516667437, "bench_spy_p1w": 0.011605240434366282, "bench_c_p1w": 0.011792818010439632, "ret_p1m": -0.06206747457416706, "ret_signed_p1m": -0.06206747457416706, "alpha_spy_p1m": -0.08472758806832126, "alpha_c_p1m": -0.07709818936338297, "bench_spy_p1m": 0.0226601134941542, "bench_c_p1m": 0.015030714789215915, "ret_p3m": 0.03312435281241011, "ret_signed_p3m": 0.03312435281241011, "alpha_spy_p3m": -0.023877194236789245, "alpha_c_p3m": -0.16350152774151194, "bench_spy_p3m": 0.05700154704919935, "bench_c_p3m": 0.19662588055392205, "ret_p6m": -0.009336434105667979, "ret_signed_p6m": -0.009336434105667979, "alpha_spy_p6m": -0.0998586065255257, "alpha_c_p6m": -0.36897324878232485, "bench_spy_p6m": 0.09052217241985772, "bench_c_p6m": 0.35963681467665687, "price_path": [-0.2567, -0.2595, -0.2563, -0.2554, -0.2619, -0.2761, -0.2705, -0.2789, -0.2789, -0.2558, -0.2596, -0.2355, -0.2578, -0.2455, -0.2244, -0.2205, -0.2311, -0.2216, -0.2166, -0.2093, -0.2155, -0.2036, -0.2202, -0.2053, -0.2084, -0.2009, -0.2009, -0.2099, -0.2081, -0.1876, -0.1524, -0.1485, -0.1335, -0.1322, -0.158, -0.1363, -0.1248, -0.1248, -0.1308, -0.1212, -0.1053, -0.0986, -0.0941, -0.0988, -0.0624, -0.0587, -0.0498, -0.053, -0.0587, -0.0826, -0.062, -0.0457, -0.047, -0.0292, -0.036, -0.0153, -0.023, -0.0458, -0.0113, -0.0209, -0.0145, -0.0071, 0.0035, 0.0, 0.006, -0.0026, -0.0002, -0.0088, -0.0003, -0.0353, -0.0366, -0.0389, -0.0224, -0.0124, -0.0016, -0.0025, -0.0104, -0.0433, -0.0433, -0.062, -0.0628, -0.0571, -0.0826, -0.0755, -0.0621, -0.026, -0.0268, -0.0232, -0.0236, -0.0394, -0.0646, -0.0319, -0.0296, 0.0086, -0.0199, -0.0279, -0.0239, -0.0212, -0.0011, 0.0249, 0.0285, 0.0376, 0.0306, 0.0192, 0.0164, 0.0388, 0.0578, 0.0061, 0.0344, -0.0111, -0.0122, -0.0013, 0.0064, 0.0032, -0.0049, -0.0097, 0.0006, 0.0231, 0.0519, 0.1043, 0.1373, 0.1145, 0.1123, 0.1364, 0.0914, 0.0723, 0.0331, 0.0335, 0.0934, 0.061, 0.0645, 0.0264, 0.0446, 0.025, -0.0038, 0.0246, -0.0077, -0.0174, 0.0027, -0.0232, -0.0098, -0.0098, -0.0277, -0.0117, -0.0032, -0.0135, 0.0074, 0.002, 0.0193, 0.0161, 0.0096, -0.0061, -0.0386, -0.0316, -0.0237, -0.061, -0.0434, -0.0058, 0.0091, 0.0394, 0.0361, 0.0361, 0.0466, 0.034, 0.0302, 0.0245, 0.0245, 0.0374, 0.0334, 0.0286, 0.0388, 0.0165, 0.0155, 0.0159, 0.0207, 0.006, 0.0275, 0.023, 0.023, -0.0218, 0.007, 0.0154, 0.0309, 0.0243, 0.0356, 0.0521, 0.0575, 0.0499, 0.0196, -0.0093]}
{"unit_id": "orig:1954705260125343896", "ticker_idx": 1, "ts": "2025-08-11T00:43:45+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD: 33x 2026E EPS; AI growth and CPU market share gains expected", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA cloud capex review names NVDA, AMD, MRVL, AVGO as core beneficiaries of AI datacenter/accelerator buildout; upward-revised 2025/26 capex forecasts support the theme.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "2Q Cloud Capex Review (BofA)\n\n1. Industry Trends\nThe 2025/26 global cloud AI Capex forecast has been revised upward to +54%/+16% YoY, with total spending expected to reach $432B/$500B, respectively.\nQ2 global hyperscale Capex came in at $112B, up +27% QoQ and +74% YoY, exceeding expectations.\nAll major cloud players are increasing the proportion of spend on servers (especially CPUs/GPUs). Capex intensity remains at 23–25%, well above the historical average of 10–12%.\nGPU shipments to China could add about 10% to AI TAM, and there is potential for easing of U.S.–China trade tariff risks.\nAI infrastructure demand is expanding beyond hyperscalers to neo-cloud providers, enterprises, and national-level projects.\n\n2. Company-Level AI Investment Trends\nGoogle: 2025 Capex outlook raised to $85B, with servers accounting for over two-thirds of spending.\nMicrosoft: Higher proportion of short-term assets such as servers; demand exceeds supply; backlog of $368B.\nMeta: 2026 Capex expected to approach $100B, server-centric; hinted at the possibility of raising external capital.\nAmazon: Q2 Capex +30% QoQ; 2025 Capex guidance of +40% YoY.\nOracle: Sustained high growth driven by cloud expansion.\n\n3. AI TAM Outlook\nAI accelerator market projected to grow from $126B in 2024 to over $650B in 2030, with a 2025–2030 CAGR of +26%.\nAI servers to account for 85% of AI DC systems, reaching ~$700B by 2030.\nHBM market projected to grow from $17B in 2024 to $87B in 2030.\nMarket share expectations: NVIDIA 80%+, AMD 5%, ASIC ecosystem 10–15%.\n\n4. Key Beneficiaries\nNVIDIA, AMD, Marvell, Broadcom: Core beneficiaries of AI datacenter and accelerator market growth.\nSemiconductor equipment, EDA, and optical component companies also expected to benefit.\n\n5. Large-Scale AI Projects by Country\nU.S.: Stargate ($500B, ramping in 2028), Pennsylvania AI Energy Corridor ($90B).\nSaudi Arabia: HUMAIN AI Factories ($30B, 2025–2030).\nUAE: 5GW AI Campus (multi-phase expansion through 2030).\nFrance, Malaysia, Brazil, Italy, and several other countries have AI projects exceeding $1B in scale.\n\n6. Individual Stock Valuation & Risks\nAMD: 33x 2026E EPS; AI growth and CPU market share gains expected, but timing/scale of Middle East projects remain uncertain.\nMarvell (MRVL): 25x 2026E EPS; growth in custom silicon and HBM, but faces integration and rising debt risks.\nNVIDIA (NVDA): 37x 2026E PE; maintaining dominance in the AI accelerator market, but subject to potential headwinds from gaming, China export restrictions, and power supply constraints.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002786137281983514, "ret_signed_m1d": 0.002786137281983514, "alpha_spy_m1d": 0.0008047760161347117, "alpha_c_m1d": 0.0023430171318381543, "bench_spy_m1d": 0.0019813612658488022, "bench_c_m1d": 0.00044312015014535966, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01549801594998712, "ret_signed_p1d": 0.01549801594998712, "alpha_spy_p1d": 0.0048519014942869365, "alpha_c_p1d": -0.007337748510003017, "bench_spy_p1d": 0.010646114455700184, "bench_c_p1d": 0.022835764459990138, "ret_p1w": 0.022405390281529547, "ret_signed_p1w": 0.022405390281529547, "alpha_spy_p1w": 0.010800149847163265, "alpha_c_p1w": 0.010612572271089915, "bench_spy_p1w": 0.011605240434366282, "bench_c_p1w": 0.011792818010439632, "ret_p1m": -0.09554209177911921, "ret_signed_p1m": -0.09554209177911921, "alpha_spy_p1m": -0.11820220527327341, "alpha_c_p1m": -0.11057280656833512, "bench_spy_p1m": 0.0226601134941542, "bench_c_p1m": 0.015030714789215915, "ret_p3m": 0.37973066306294245, "ret_signed_p3m": 0.37973066306294245, "alpha_spy_p3m": 0.3227291160137431, "alpha_c_p3m": 0.1831047825090204, "bench_spy_p3m": 0.05700154704919935, "bench_c_p3m": 0.19662588055392205, "ret_p6m": 0.4053285484434672, "ret_signed_p6m": 0.4053285484434672, "alpha_spy_p6m": 0.3148063760236095, "alpha_c_p6m": 0.04569173376681035, "bench_spy_p6m": 0.09052217241985772, "bench_c_p6m": 0.35963681467665687, "price_path": [-0.3167, -0.3325, -0.3199, -0.334, -0.3411, -0.3495, -0.3574, -0.3597, -0.3597, -0.335, -0.3449, -0.3439, -0.3573, -0.3346, -0.3191, -0.3117, -0.3285, -0.3256, -0.2934, -0.2847, -0.2968, -0.3122, -0.3257, -0.2664, -0.2622, -0.264, -0.264, -0.2556, -0.2479, -0.1965, -0.1676, -0.166, -0.1653, -0.1763, -0.2099, -0.196, -0.1995, -0.1995, -0.2176, -0.2, -0.1966, -0.1632, -0.1501, -0.1511, -0.0968, -0.0708, -0.0689, -0.0888, -0.0887, -0.1019, -0.0791, -0.059, -0.0337, 0.008, 0.03, 0.042, 0.0234, -0.0034, 0.0261, 0.0118, -0.0532, 0.0007, 0.0028, 0.0, 0.0155, 0.0705, 0.0503, 0.0304, 0.0224, -0.0333, -0.0411, -0.0497, -0.0262, -0.0518, -0.0329, -0.0299, -0.0215, -0.056, -0.056, -0.0578, -0.0589, -0.0609, -0.1227, -0.1211, -0.0955, -0.0739, -0.0964, -0.0796, -0.0645, -0.0686, -0.0762, -0.0834, -0.0864, -0.0725, -0.0661, -0.0662, -0.0639, -0.0744, -0.0634, -0.0609, -0.048, -0.0148, -0.0442, 0.1824, 0.2277, 0.3673, 0.3518, 0.2474, 0.2562, 0.2659, 0.385, 0.3615, 0.3529, 0.3963, 0.3816, 0.3364, 0.364, 0.4681, 0.5073, 0.4976, 0.5343, 0.4792, 0.4866, 0.5071, 0.4514, 0.4879, 0.3797, 0.3556, 0.4162, 0.3787, 0.5027, 0.4393, 0.4326, 0.3961, 0.3367, 0.2976, 0.1958, 0.1828, 0.2483, 0.1965, 0.2436, 0.2436, 0.2627, 0.2756, 0.2494, 0.2631, 0.2537, 0.2652, 0.2834, 0.2864, 0.2852, 0.2853, 0.2235, 0.2049, 0.2141, 0.1499, 0.1671, 0.2389, 0.2477, 0.2474, 0.2482, 0.2482, 0.2479, 0.2515, 0.2499, 0.2431, 0.2431, 0.2971, 0.2833, 0.2442, 0.2191, 0.1881, 0.1793, 0.2055, 0.2826, 0.2979, 0.323, 0.3457, 0.3457, 0.3462, 0.45, 0.4728, 0.5073, 0.4587, 0.4629, 0.467, 0.4638, 0.3741, 0.4295, 0.4053]}
{"unit_id": "orig:1954705260125343896", "ticker_idx": 2, "ts": "2025-08-11T00:43:45+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell (MRVL): 25x 2026E EPS; growth in custom silicon and HBM", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA cloud capex review names NVDA, AMD, MRVL, AVGO as core beneficiaries of AI datacenter/accelerator buildout; upward-revised 2025/26 capex forecasts support the theme.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "2Q Cloud Capex Review (BofA)\n\n1. Industry Trends\nThe 2025/26 global cloud AI Capex forecast has been revised upward to +54%/+16% YoY, with total spending expected to reach $432B/$500B, respectively.\nQ2 global hyperscale Capex came in at $112B, up +27% QoQ and +74% YoY, exceeding expectations.\nAll major cloud players are increasing the proportion of spend on servers (especially CPUs/GPUs). Capex intensity remains at 23–25%, well above the historical average of 10–12%.\nGPU shipments to China could add about 10% to AI TAM, and there is potential for easing of U.S.–China trade tariff risks.\nAI infrastructure demand is expanding beyond hyperscalers to neo-cloud providers, enterprises, and national-level projects.\n\n2. Company-Level AI Investment Trends\nGoogle: 2025 Capex outlook raised to $85B, with servers accounting for over two-thirds of spending.\nMicrosoft: Higher proportion of short-term assets such as servers; demand exceeds supply; backlog of $368B.\nMeta: 2026 Capex expected to approach $100B, server-centric; hinted at the possibility of raising external capital.\nAmazon: Q2 Capex +30% QoQ; 2025 Capex guidance of +40% YoY.\nOracle: Sustained high growth driven by cloud expansion.\n\n3. AI TAM Outlook\nAI accelerator market projected to grow from $126B in 2024 to over $650B in 2030, with a 2025–2030 CAGR of +26%.\nAI servers to account for 85% of AI DC systems, reaching ~$700B by 2030.\nHBM market projected to grow from $17B in 2024 to $87B in 2030.\nMarket share expectations: NVIDIA 80%+, AMD 5%, ASIC ecosystem 10–15%.\n\n4. Key Beneficiaries\nNVIDIA, AMD, Marvell, Broadcom: Core beneficiaries of AI datacenter and accelerator market growth.\nSemiconductor equipment, EDA, and optical component companies also expected to benefit.\n\n5. Large-Scale AI Projects by Country\nU.S.: Stargate ($500B, ramping in 2028), Pennsylvania AI Energy Corridor ($90B).\nSaudi Arabia: HUMAIN AI Factories ($30B, 2025–2030).\nUAE: 5GW AI Campus (multi-phase expansion through 2030).\nFrance, Malaysia, Brazil, Italy, and several other countries have AI projects exceeding $1B in scale.\n\n6. Individual Stock Valuation & Risks\nAMD: 33x 2026E EPS; AI growth and CPU market share gains expected, but timing/scale of Middle East projects remain uncertain.\nMarvell (MRVL): 25x 2026E EPS; growth in custom silicon and HBM, but faces integration and rising debt risks.\nNVIDIA (NVDA): 37x 2026E PE; maintaining dominance in the AI accelerator market, but subject to potential headwinds from gaming, China export restrictions, and power supply constraints.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00077634445124386, "ret_signed_m1d": 0.00077634445124386, "alpha_spy_m1d": -0.0012050168146049423, "alpha_c_m1d": 0.0003332243010985003, "bench_spy_m1d": 0.0019813612658488022, "bench_c_m1d": 0.00044312015014535966, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006858121444248022, "ret_signed_p1d": 0.006858121444248022, "alpha_spy_p1d": -0.003787993011452162, "alpha_c_p1d": -0.015977643015742116, "bench_spy_p1d": 0.010646114455700184, "bench_c_p1d": 0.022835764459990138, "ret_p1w": -0.006987594611107206, "ret_signed_p1w": -0.006987594611107206, "alpha_spy_p1w": -0.018592835045473488, "alpha_c_p1w": -0.018780412621546838, "bench_spy_p1w": 0.011605240434366282, "bench_c_p1w": 0.011792818010439632, "ret_p1m": -0.13509323205478663, "ret_signed_p1m": -0.13509323205478663, "alpha_spy_p1m": -0.15775334554894083, "alpha_c_p1m": -0.15012394684400254, "bench_spy_p1m": 0.0226601134941542, "bench_c_p1m": 0.015030714789215915, "ret_p3m": 0.20848592258731102, "ret_signed_p3m": 0.20848592258731102, "alpha_spy_p3m": 0.15148437553811167, "alpha_c_p3m": 0.011860042033388973, "bench_spy_p3m": 0.05700154704919935, "bench_c_p3m": 0.19662588055392205, "ret_p6m": -0.02116456019725821, "ret_signed_p6m": -0.02116456019725821, "alpha_spy_p6m": -0.11168673261711592, "alpha_c_p6m": -0.3808013748739151, "bench_spy_p6m": 0.09052217241985772, "bench_c_p6m": 0.35963681467665687, "price_path": [-0.148, -0.157, -0.1756, -0.1911, -0.2059, -0.2228, -0.2003, -0.2153, -0.2153, -0.1748, -0.1649, -0.176, -0.2218, -0.2052, -0.1937, -0.1428, -0.1575, -0.1163, -0.1061, -0.1099, -0.1177, -0.0996, -0.1313, -0.0895, -0.0951, -0.0309, -0.0309, -0.0496, -0.0849, -0.0276, -0.0183, 0.034, -0.0024, 0.0007, -0.0143, -0.04, -0.028, -0.028, -0.0749, -0.0697, -0.0657, -0.0515, -0.0591, -0.0617, -0.063, -0.0832, -0.0682, -0.034, -0.0546, -0.0685, -0.0519, -0.0419, -0.0397, -0.0177, -0.0122, 0.0577, 0.04, -0.0366, -0.0097, -0.0084, -0.0254, -0.0185, 0.0008, 0.0, 0.0069, 0.0264, 0.0228, -0.0141, -0.007, -0.0674, -0.0784, -0.0785, -0.0554, -0.056, -0.0391, -0.0322, -0.0006, -0.1865, -0.1865, -0.1641, -0.1937, -0.1705, -0.1805, -0.146, -0.1351, -0.1317, -0.1383, -0.1285, -0.1275, -0.109, -0.0815, -0.0395, -0.0391, -0.0226, -0.0344, 0.0364, 0.0845, 0.0762, 0.0661, 0.0879, 0.0855, 0.1154, 0.1157, 0.1506, 0.1254, 0.1969, 0.1734, 0.1085, 0.1575, 0.1164, 0.151, 0.1424, 0.1388, 0.1115, 0.091, 0.0495, 0.0717, 0.0894, 0.1487, 0.1454, 0.1673, 0.1469, 0.2138, 0.1702, 0.1342, 0.2029, 0.2085, 0.1773, 0.2072, 0.1567, 0.1567, 0.1333, 0.1194, 0.0806, 0.0188, 0.053, -0.0071, 0.0029, 0.085, 0.0803, 0.1358, 0.1358, 0.1576, 0.1796, 0.2028, 0.2974, 0.2714, 0.2807, 0.1913, 0.1511, 0.1974, 0.158, 0.0932, 0.091, 0.0886, 0.0579, 0.0938, 0.0888, 0.098, 0.1353, 0.1199, 0.1199, 0.118, 0.1105, 0.1234, 0.1004, 0.1004, 0.1575, 0.1683, 0.1424, 0.096, 0.0806, 0.0784, 0.0741, 0.0761, 0.0523, 0.0416, 0.0426, 0.0426, 0.034, 0.0697, 0.0768, 0.0396, 0.0596, 0.0746, 0.0835, 0.054, 0.0226, 0.0193, -0.0212]}
{"unit_id": "orig:1954705260125343896", "ticker_idx": 3, "ts": "2025-08-11T00:43:45+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA, AMD, Marvell, Broadcom: Core beneficiaries of AI datacenter and accelerator market growth", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA cloud capex review names NVDA, AMD, MRVL, AVGO as core beneficiaries of AI datacenter/accelerator buildout; upward-revised 2025/26 capex forecasts support the theme.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "2Q Cloud Capex Review (BofA)\n\n1. Industry Trends\nThe 2025/26 global cloud AI Capex forecast has been revised upward to +54%/+16% YoY, with total spending expected to reach $432B/$500B, respectively.\nQ2 global hyperscale Capex came in at $112B, up +27% QoQ and +74% YoY, exceeding expectations.\nAll major cloud players are increasing the proportion of spend on servers (especially CPUs/GPUs). Capex intensity remains at 23–25%, well above the historical average of 10–12%.\nGPU shipments to China could add about 10% to AI TAM, and there is potential for easing of U.S.–China trade tariff risks.\nAI infrastructure demand is expanding beyond hyperscalers to neo-cloud providers, enterprises, and national-level projects.\n\n2. Company-Level AI Investment Trends\nGoogle: 2025 Capex outlook raised to $85B, with servers accounting for over two-thirds of spending.\nMicrosoft: Higher proportion of short-term assets such as servers; demand exceeds supply; backlog of $368B.\nMeta: 2026 Capex expected to approach $100B, server-centric; hinted at the possibility of raising external capital.\nAmazon: Q2 Capex +30% QoQ; 2025 Capex guidance of +40% YoY.\nOracle: Sustained high growth driven by cloud expansion.\n\n3. AI TAM Outlook\nAI accelerator market projected to grow from $126B in 2024 to over $650B in 2030, with a 2025–2030 CAGR of +26%.\nAI servers to account for 85% of AI DC systems, reaching ~$700B by 2030.\nHBM market projected to grow from $17B in 2024 to $87B in 2030.\nMarket share expectations: NVIDIA 80%+, AMD 5%, ASIC ecosystem 10–15%.\n\n4. Key Beneficiaries\nNVIDIA, AMD, Marvell, Broadcom: Core beneficiaries of AI datacenter and accelerator market growth.\nSemiconductor equipment, EDA, and optical component companies also expected to benefit.\n\n5. Large-Scale AI Projects by Country\nU.S.: Stargate ($500B, ramping in 2028), Pennsylvania AI Energy Corridor ($90B).\nSaudi Arabia: HUMAIN AI Factories ($30B, 2025–2030).\nUAE: 5GW AI Campus (multi-phase expansion through 2030).\nFrance, Malaysia, Brazil, Italy, and several other countries have AI projects exceeding $1B in scale.\n\n6. Individual Stock Valuation & Risks\nAMD: 33x 2026E EPS; AI growth and CPU market share gains expected, but timing/scale of Middle East projects remain uncertain.\nMarvell (MRVL): 25x 2026E EPS; growth in custom silicon and HBM, but faces integration and rising debt risks.\nNVIDIA (NVDA): 37x 2026E PE; maintaining dominance in the AI accelerator market, but subject to potential headwinds from gaming, China export restrictions, and power supply constraints.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0035210147406323777, "ret_signed_m1d": 0.0035210147406323777, "alpha_spy_m1d": 0.0015396534747835755, "alpha_c_m1d": 0.003077894590487018, "bench_spy_m1d": 0.0019813612658488022, "bench_c_m1d": 0.00044312015014535966, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029384712697902327, "ret_signed_p1d": 0.029384712697902327, "alpha_spy_p1d": 0.018738598242202142, "alpha_c_p1d": 0.006548948237912189, "bench_spy_p1d": 0.010646114455700184, "bench_c_p1d": 0.022835764459990138, "ret_p1w": 0.0061205442822418465, "ret_signed_p1w": 0.0061205442822418465, "alpha_spy_p1w": -0.005484696152124435, "alpha_c_p1w": -0.005672273728197785, "bench_spy_p1w": 0.011605240434366282, "bench_c_p1w": 0.011792818010439632, "ret_p1m": 0.1078316445329437, "ret_signed_p1m": 0.1078316445329437, "alpha_spy_p1m": 0.0851715310387895, "alpha_c_p1m": 0.09280092974372778, "bench_spy_p1m": 0.0226601134941542, "bench_c_p1m": 0.015030714789215915, "ret_p3m": 0.172093622653831, "ret_signed_p3m": 0.172093622653831, "alpha_spy_p3m": 0.11509207560463164, "alpha_c_p3m": -0.02453225790009106, "bench_spy_p3m": 0.05700154704919935, "bench_c_p3m": 0.19662588055392205, "ret_p6m": 0.05789005542343628, "ret_signed_p6m": 0.05789005542343628, "alpha_spy_p6m": -0.032632116996421434, "alpha_c_p6m": -0.3017467592532206, "bench_spy_p6m": 0.09052217241985772, "bench_c_p6m": 0.35963681467665687, "price_path": [-0.238, -0.2363, -0.2495, -0.2429, -0.2394, -0.2458, -0.2432, -0.2492, -0.2492, -0.2264, -0.214, -0.2057, -0.2053, -0.1835, -0.1568, -0.1429, -0.1467, -0.1894, -0.1981, -0.1969, -0.1697, -0.1594, -0.1836, -0.1724, -0.1814, -0.1752, -0.1752, -0.1774, -0.165, -0.132, -0.1292, -0.111, -0.1137, -0.093, -0.1289, -0.1119, -0.0945, -0.0945, -0.0978, -0.1056, -0.0856, -0.0938, -0.0971, -0.0931, -0.0756, -0.076, -0.0574, -0.0677, -0.0516, -0.0833, -0.0665, -0.05, -0.0451, -0.0316, -0.0213, -0.0042, -0.0336, -0.0502, -0.0203, -0.0361, -0.0073, -0.0005, 0.0035, 0.0, 0.0294, 0.0171, 0.0241, 0.008, 0.0061, -0.0296, -0.0419, -0.0471, -0.0326, -0.0318, -0.0194, -0.012, 0.0156, -0.0214, -0.0214, -0.0186, -0.005, 0.0072, 0.102, 0.1374, 0.1078, 0.2161, 0.1834, 0.1842, 0.1981, 0.1846, 0.1391, 0.1364, 0.135, 0.1167, 0.1172, 0.1184, 0.1079, 0.1027, 0.0808, 0.0874, 0.0989, 0.1147, 0.1153, 0.1058, 0.1089, 0.1388, 0.1373, 0.07, 0.1758, 0.1343, 0.1581, 0.1673, 0.1515, 0.1512, 0.1295, 0.1217, 0.1348, 0.1673, 0.1934, 0.2294, 0.2723, 0.2409, 0.2184, 0.195, 0.1601, 0.1833, 0.1721, 0.1518, 0.1813, 0.1601, 0.1709, 0.1206, 0.1288, 0.1294, 0.1224, 0.1682, 0.1432, 0.1214, 0.2458, 0.2691, 0.3105, 0.3105, 0.3282, 0.2726, 0.2577, 0.2546, 0.2559, 0.2863, 0.3221, 0.3392, 0.3612, 0.3395, 0.1864, 0.1201, 0.125, 0.0746, 0.0873, 0.1219, 0.1276, 0.1536, 0.1566, 0.1566, 0.1629, 0.1539, 0.1554, 0.143, 0.143, 0.148, 0.1341, 0.1353, 0.1344, 0.098, 0.1393, 0.1632, 0.1711, 0.1225, 0.1328, 0.1615, 0.1615, 0.0984, 0.0859, 0.0749, 0.057, 0.0728, 0.099, 0.1005, 0.0922, 0.0941, 0.0935, 0.0579]}
{"unit_id": "quote:1944924593971716250", "ticker_idx": 0, "ts": "2025-07-15T00:58:52+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell might face a market void in the second half of 2025", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish MRVL: Trainium 2 MAX to be designed in-house by Annapurna, leaving Marvell with a revenue gap in H2 2025.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AWS's ASIC Strategy: Strong Momentum, but Risks Emerge\nAWS's momentum in the ASIC (Application-Specific Integrated Circuit) space is currently very strong, and there's no questioning their commitment to actively deploying custom chips. However, we've identified several downstream risks, including the delay of Trainium 3, a mismatch in upstream and downstream forecasts, a reduction in Trainium 3 capacity, and cost reduction/pull-in demand pressures.\n\nTrainium 3 and Trainium 2 Delays\n\nAccording to our channel checks, AWS is postponing the launch of Trainium 3 due to insufficient preparation for its custom liquid cooling system. Consequently, the lifecycle of Trainium 2 may be extended. We anticipate Trainium 2 will be phased out around September 2025. Before Trainium 3's release, AWS might introduce a \"Trainium 2 MAX\" version, which will feature higher HBM capacity but still utilize an air-cooled design.\nIt's worth noting that while Trainium 3 is extended, Marvell's involvement in Trainium 2 is gradually decreasing. Trainium 2 MAX will be designed and manufactured independently by Annapurna, which implies that Marvell might face a market void in the second half of 2025.", "reply_text": null, "parent_text": null, "quoted_text": "轉發LumenAlpha substack\nJul 11, 2025\n\nAWS 的 ASIC 勢頭近期非常強勁，我們並不質疑他們積極部署客製化晶片的決心。然而，我們發現下游存在一些風險，包括 Trainium 3 的延期、上下游預測不匹配、Trainium 3 的容量減少以及成本下降/拉動壓力。\n\nTrainium 3 延期及 Trainium 2 延期", "parent_author": null, "ret_m1d": 0.0013809655611551896, "ret_signed_m1d": -0.0013809655611551896, "alpha_spy_m1d": 0.0029105398088165035, "alpha_c_m1d": -0.02022546650521717, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021544097264227235, "ret_signed_p1d": 0.021544097264227235, "alpha_spy_p1d": 0.024887288382742345, "alpha_c_p1d": 0.0163171223555596, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.005800224256477415, "ret_signed_p1w": 0.005800224256477415, "alpha_spy_p1w": 0.016601607677292796, "alpha_c_p1w": -0.012872370659230747, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.09542892165400874, "ret_signed_p1m": -0.09542892165400874, "alpha_spy_p1m": -0.05886158773094308, "alpha_c_p1m": -0.0593903439621688, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.1830781154433161, "ret_signed_p3m": -0.1830781154433161, "alpha_spy_p3m": -0.13053299730275203, "alpha_c_p3m": -0.06430231374855211, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.1696731855588287, "ret_signed_p6m": -0.1696731855588287, "alpha_spy_p6m": -0.05491588227764166, "alpha_c_p6m": 0.15844931427695008, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2866, -0.2866, -0.3186, -0.3015, -0.258, -0.2089, -0.187, -0.19, -0.1903, -0.1946, -0.159, -0.1399, -0.1447, -0.1552, -0.223, -0.2048, -0.1769, -0.11, -0.0964, -0.0906, -0.1003, -0.1202, -0.1367, -0.1525, -0.1705, -0.1465, -0.1625, -0.1625, -0.1194, -0.1087, -0.1206, -0.1694, -0.1518, -0.1395, -0.0851, -0.1009, -0.0568, -0.0459, -0.0501, -0.0584, -0.039, -0.0728, -0.0283, -0.0342, 0.0342, 0.0342, 0.0144, -0.0233, 0.0378, 0.0478, 0.1035, 0.0647, 0.068, 0.052, 0.0246, 0.0374, 0.0374, -0.0127, -0.0072, -0.0029, 0.0123, 0.0041, 0.0014, 0.0, -0.0215, -0.0055, 0.0309, 0.009, -0.0058, 0.0119, 0.0225, 0.0249, 0.0483, 0.0543, 0.1288, 0.1099, 0.0282, 0.0569, 0.0583, 0.0402, 0.0475, 0.0681, 0.0673, 0.0746, 0.0954, 0.0916, 0.0522, 0.0598, -0.0047, -0.0164, -0.0166, 0.0081, 0.0075, 0.0255, 0.0329, 0.0666, -0.1317, -0.1317, -0.1079, -0.1395, -0.1148, -0.1254, -0.0885, -0.0769, -0.0733, -0.0804, -0.0699, -0.0688, -0.049, -0.0197, 0.0251, 0.0255, 0.0431, 0.0305, 0.1061, 0.1574, 0.1486, 0.1378, 0.161, 0.1585, 0.1904, 0.1907, 0.228, 0.2011, 0.2774, 0.2523, 0.1831, 0.2353, 0.1915, 0.2284, 0.2193, 0.2154, 0.1863, 0.1644, 0.1201, 0.1438, 0.1626, 0.2259, 0.2225, 0.2458, 0.224, 0.2954, 0.2489, 0.2104, 0.2838, 0.2898, 0.2565, 0.2884, 0.2345, 0.2345, 0.2095, 0.1947, 0.1532, 0.0873, 0.1238, 0.0597, 0.0703, 0.1579, 0.153, 0.2122, 0.2122, 0.2355, 0.2589, 0.2837, 0.3847, 0.3569, 0.3669, 0.2714, 0.2285, 0.2779, 0.2359, 0.1668, 0.1644, 0.1618, 0.129, 0.1673, 0.1621, 0.1719, 0.2117, 0.1952, 0.1952, 0.1932, 0.1852, 0.199, 0.1744, 0.1744, 0.2353, 0.2469, 0.2193, 0.1697]}
{"unit_id": "thread:1969315945694052655", "ticker_idx": 0, "ts": "2025-09-20T08:21:24+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ARM will end up being the biggest loser in this NVIDIA–Intel deal... one of ARM's biggest future growth expectations—the push into notebooks—has basically evaporated... ARM's penetration in server CPUs will also be delayed... that, too, would undermine ARM's valuation", "tweet_type": "analysis", "timeline": "months", "summary": "Short ARM: notebook growth story evaporates on NVIDIA-Intel partnership, server CPU share defense by x86 delays ARM penetration, and NVIDIA's own CPU architecture may cut royalty rates.", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "I’ve learned a lot from @insane_analyst as always.\n\nI can’t shake the thought that ARM will end up being the biggest loser in this NVIDIA–Intel deal.\n\n1. Windows on ARM essentially collapsing\n\nAfter Qualcomm entered the WoA space, ARM quietly asked AMD, NVIDIA, and Samsung to join. But the strongest candidate, the NVIDIA + MTK partnership, has basically fallen apart.\n\nNow, NVIDIA has climbed onto the back of a giant in the notebook SoC market: Intel.\n\nSo what happens to N1/N1X? NVIDIA won’t really need MediaTek anymore.\n\nFor ARM, the only hope for breaking into the notebook chipset market lies with AMD. But honestly… even if AMD is developing “Sound Wave,” it won’t have the bandwidth to bother with ARM while it’s busy fending off the Intel–NVIDIA alliance.\n\nIn other words, one of ARM’s biggest future growth expectations—the push into notebooks—has basically evaporated.\n\n2. ARM’s penetration in server CPUs will also be delayed\n\nAs my friend IA pointed out, x86 had been steadily losing share in the server CPU market. It’s already fallen to around the 80% range.\n\nBut if Intel supplies x86 CPUs for DGX systems, x86’s market share will naturally be defended.\n\nThis directly threatens ARM’s valuation, because NVIDIA—the biggest player in the server CPU market—choosing x86 carries huge significance.\n\nOne more thing I recently learned: NVIDIA is actually designing its own server CPU architecture.\n\nWould that lower ARM’s royalty rates? Hard to say for sure, but right now it looks more likely than not.\n\nAnd that, too, would undermine ARM’s valuation.\n\n---\n\nLink: https://t.co/zMZ3e6fKmv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010866501327909672, "ret_signed_m1d": 0.010866501327909672, "alpha_spy_m1d": 0.006157770403337315, "alpha_c_m1d": -0.00918505642703804, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024155525592321037, "ret_signed_p1d": 0.024155525592321037, "alpha_spy_p1d": 0.018712050752513476, "alpha_c_p1d": 0.022696686278440126, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.032391976842294445, "ret_signed_p1w": 0.032391976842294445, "alpha_spy_p1w": 0.027653237584426282, "alpha_c_p1w": 0.033912864099376216, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.17234226115447826, "ret_signed_p1m": -0.17234226115447826, "alpha_spy_p1m": -0.16566908900026012, "alpha_c_p1m": -0.10054794486009921, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": -0.2143548899995258, "ret_signed_p3m": 0.2143548899995258, "alpha_spy_p3m": 0.22879607822274628, "alpha_c_p3m": 0.29220196890018446, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": -0.11883996869244484, "ret_signed_p6m": 0.11883996869244484, "alpha_spy_p6m": 0.12773578261279517, "alpha_c_p6m": 0.3545693218355104, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [0.0888, 0.0946, 0.1452, 0.1195, 0.082, 0.0703, 0.0734, 0.0734, 0.0166, 0.0229, 0.0245, 0.0282, 0.0101, 0.0004, 0.0182, 0.0652, 0.0879, 0.0849, 0.1207, 0.0832, 0.1024, 0.1074, 0.1294, 0.1377, 0.1314, 0.1304, -0.0215, -0.0478, -0.0307, -0.0502, -0.0579, -0.0617, -0.0414, -0.0237, -0.0145, -0.0199, -0.0272, -0.0386, -0.0237, -0.0725, -0.0922, -0.0775, -0.0454, -0.0464, -0.0292, -0.0264, -0.0134, -0.0427, -0.0427, -0.084, -0.0904, -0.0623, -0.0437, -0.037, -0.0255, 0.0669, 0.0707, 0.0426, 0.0649, 0.0649, 0.0615, 0.0143, -0.0109, 0.0, -0.0242, -0.0012, -0.0265, -0.0336, -0.0324, -0.0207, 0.0408, 0.0531, 0.0565, 0.0813, 0.1029, 0.1543, 0.1812, 0.0715, 0.1901, 0.1639, 0.1813, 0.1849, 0.1462, 0.187, 0.1723, 0.1469, 0.1531, 0.1813, 0.2363, 0.198, 0.1793, 0.1451, 0.1754, 0.1675, 0.1125, 0.1087, 0.0953, 0.0547, 0.0717, 0.0364, 0.0296, -0.0289, -0.0326, -0.0292, -0.0584, -0.0518, -0.0827, -0.0894, -0.0676, -0.0903, -0.0822, -0.0822, -0.0617, -0.0655, -0.0554, -0.0366, -0.0276, -0.0219, -0.0325, -0.0176, -0.0205, -0.0577, -0.0941, -0.1392, -0.1618, -0.2069, -0.2144, -0.2108, -0.2159, -0.2247, -0.2279, -0.2279, -0.2368, -0.2351, -0.2327, -0.2434, -0.2434, -0.2059, -0.1964, -0.2004, -0.1993, -0.2173, -0.2263, -0.2308, -0.2536, -0.2733, -0.2725, -0.2679, -0.2679, -0.2582, -0.2115, -0.175, -0.1966, -0.2059, -0.2049, -0.2389, -0.2495, -0.2708, -0.2599, -0.2764, -0.2739, -0.2326, -0.1438, -0.1375, -0.1283, -0.1329, -0.1543, -0.1329, -0.1329, -0.1217, -0.1193, -0.1215, -0.1308, -0.1433, -0.1131, -0.0882, -0.1053, -0.1179, -0.1392, -0.1575, -0.141, -0.1651, -0.2083, -0.1858, -0.1656, -0.1687, -0.2032, -0.1989, -0.1577, -0.1188]}
{"unit_id": "orig:1955808046862754094", "ticker_idx": 0, "ts": "2025-08-14T01:45:50+00:00", "symbol": "Lumentum", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM generally prefers Lumentum", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM adopts optical sector note: prefers Lumentum (beat expected on rev/EPS), followed by Coherent; cautious on Fabrinet due to high valuation and 1.6T ramp scrutiny.", "resolved_tickers": ["LITE"], "resolution_method": "exchange_ticker", "resolution_reason": "Lumentum Holdings LITE US optical components", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Communications F4Q25 Outlook: Fundamentals Remain Solid, but High Valuations Set a High Bar for Earnings\nJPM (S. Chatterjee, 11/08/25)\n\nStock prices for optical component companies have risen by an average of about 50% since early June, with valuations exceeding 27 times NTM (Next Twelve Months) earnings, setting a high execution bar for F4Q25 financial reports. Although the capital expenditure outlook for US-based cloud service providers and second-tier CSPs is robust, and there is a recovery in demand for optical components both inside and outside the data center as well as in telecom investment, the market is concerned that the magnitude of earnings revisions will be insufficient to justify the high valuations. Within the optical components sector, JPM generally prefers Lumentum, followed by Coherent, and remains cautious on Fabrinet.\n\nIndustry Background and Overall Judgment\nThe optical components sector has significantly outperformed the broader market in recent months. The capex outlook for US-based cloud providers and second-tier CSPs is favorable, and industry checks indicate strong demand from data center and telecom investments. However, the high valuations mean that earnings reports must significantly beat expectations to drive stock prices higher. The overall strategy is to remain cautious on earnings and to favor stocks with greater potential for upside.\n\nLumentum: Expected to Beat on Both Revenue and Profit\nDriven by comprehensive growth in telecom, datacom lasers, and optical modules, improvements in production capacity and yield are contributing to upward revisions for both revenue and gross margin. F4Q25 revenue is projected to be $484 million (+57% YoY, higher than consensus and guidance), with a gross margin of 38.0%, an operating margin of 15.5%, and an EPS of $0.89, all of which are above market expectations. For F1Q26, revenue is projected at $515 million (+53% YoY) with a gross margin of 39.4% and an EPS of $0.97, both also above consensus. Its valuation is in line with the industry average, and it is positioned to benefit the most.\n\nCoherent: Solid Execution but Lacks Short-Term Catalysts\nAlthough its fundamentals are solid, investor focus has recently decreased due to a lack of new catalysts and macroeconomic uncertainty constraining industrial growth. F4Q25 revenue is projected at $1.55 billion (+18% YoY, higher than consensus and guidance), with a gross margin of 38.3% and an EPS of $0.98, slightly exceeding expectations. F1Q26 revenue is projected at $1.552 billion (+15% YoY) with an EPS of $0.99, which is largely in line with market expectations. In the long term, continuously improving revenue and profits, along with portfolio optimization, should help bolster confidence among long-only investors.\n\nFabrinet: 1.6T Ramp-Up Pace Under Scrutiny Amid High Valuation\nThe stock price has risen significantly, driven by ramp-ups at Amazon and Ciena, and is expected to ramp in sync with Nvidia's 1.6T optical modules. However, its current CY27E P/E ratio of over 21x is significantly higher than its peers, and the pace of the 1.6T volume ramp-up in F1Q26 will be under intense scrutiny. F4Q25 revenue is projected at $890 million (+18% YoY, higher than consensus and guidance), with an EPS of $2.67. For F1Q26, revenue is projected at $927 million (+15% YoY) with an EPS of $2.78, both slightly above expectations, but the high valuation magnifies the impact of any potential delay risks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.048944341964071825, "ret_signed_m1d": 0.048944341964071825, "alpha_spy_m1d": 0.04903729877713814, "alpha_c_m1d": 0.04903729877713814, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -9.295681306631387e-05, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010818337411063306, "ret_signed_p1d": 0.010818337411063306, "alpha_spy_p1d": 0.013159532053086065, "alpha_c_p1d": 0.013159532053086065, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.002341194642022759, "ret_p1w": 0.024515769422955636, "ret_signed_p1w": 0.024515769422955636, "alpha_spy_p1w": 0.03909051968025701, "alpha_c_p1w": 0.03909051968025701, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.014574750257301372, "ret_p1m": 0.42226487843390137, "ret_signed_p1m": 0.42226487843390137, "alpha_spy_p1m": 0.4029455131819597, "alpha_c_p1m": 0.4029455131819597, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.019319365251941667, "ret_p3m": 1.2026696490219004, "ret_signed_p3m": 1.2026696490219004, "alpha_spy_p3m": 1.1407368382827354, "alpha_c_p3m": 1.1407368382827354, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.06193281073916501, "ret_p6m": 3.815826007741353, "ret_signed_p6m": 3.815826007741353, "alpha_spy_p6m": 3.738872654882635, "alpha_c_p6m": 3.738872654882635, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.07695335285871785, "price_path": [-0.3206, -0.3273, -0.3374, -0.338, -0.3389, -0.3389, -0.3163, -0.3204, -0.3421, -0.3694, -0.3382, -0.3225, -0.2996, -0.2877, -0.2893, -0.2836, -0.2849, -0.2815, -0.2541, -0.2805, -0.2516, -0.2467, -0.2282, -0.2282, -0.2178, -0.2221, -0.199, -0.1994, -0.1737, -0.1734, -0.1707, -0.2018, -0.204, -0.1908, -0.1908, -0.2054, -0.2034, -0.211, -0.1919, -0.1887, -0.1953, -0.1438, -0.1308, -0.1045, -0.1082, -0.094, -0.1308, -0.109, -0.1027, -0.0881, -0.065, -0.0448, -0.0416, -0.0396, -0.0693, -0.0304, -0.0564, -0.0402, -0.0304, 0.0144, 0.0036, 0.044, 0.0489, 0.0, 0.0108, 0.038, 0.0291, 0.0111, 0.0245, 0.0412, 0.0768, 0.0872, 0.0979, 0.1826, 0.1587, 0.1587, 0.1565, 0.1701, 0.2381, 0.304, 0.3034, 0.3238, 0.4385, 0.4385, 0.4223, 0.4724, 0.491, 0.4251, 0.4962, 0.4721, 0.437, 0.4307, 0.3327, 0.3882, 0.4025, 0.4184, 0.4196, 0.4954, 0.481, 0.4292, 0.4012, 0.3711, 0.4229, 0.3938, 0.3053, 0.4008, 0.3678, 0.366, 0.4241, 0.4375, 0.4046, 0.4217, 0.379, 0.4701, 0.5643, 0.6908, 0.7308, 0.8695, 0.746, 0.7585, 0.7412, 0.6433, 1.0306, 1.0911, 1.0948, 1.2674, 1.2027, 1.2144, 0.9792, 1.0254, 1.1119, 1.1587, 1.3462, 1.0349, 1.2299, 1.6118, 1.5412, 1.6896, 1.6896, 1.8369, 1.7738, 1.6419, 1.6433, 1.8603, 1.8914, 1.9887, 2.1437, 2.1932, 2.2463, 1.8298, 1.92, 1.7582, 1.794, 1.9413, 2.2405, 2.4015, 2.38, 2.4542, 2.4542, 2.4093, 2.2508, 2.2384, 2.2158, 2.2158, 2.3686, 2.1151, 2.4673, 2.4277, 2.0384, 2.066, 1.9652, 2.1524, 1.8932, 1.9949, 1.8289, 1.8289, 2.1132, 2.1621, 2.0927, 1.9593, 1.9005, 2.2338, 2.3589, 2.3279, 2.4186, 2.6941, 2.796, 3.0616, 3.4008, 3.8158]}
{"unit_id": "orig:1955808046862754094", "ticker_idx": 1, "ts": "2025-08-14T01:45:50+00:00", "symbol": "Coherent", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "followed by Coherent", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM adopts optical sector note: prefers Lumentum (beat expected on rev/EPS), followed by Coherent; cautious on Fabrinet due to high valuation and 1.6T ramp scrutiny.", "resolved_tickers": ["COHR"], "resolution_method": "exchange_ticker", "resolution_reason": "Coherent Corp. COHR US", "bench_c": "XLK", "bench_c_reason": "industry='Scientific & Technical Instruments'", "bench_c_industry": "Scientific & Technical Instruments", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Communications F4Q25 Outlook: Fundamentals Remain Solid, but High Valuations Set a High Bar for Earnings\nJPM (S. Chatterjee, 11/08/25)\n\nStock prices for optical component companies have risen by an average of about 50% since early June, with valuations exceeding 27 times NTM (Next Twelve Months) earnings, setting a high execution bar for F4Q25 financial reports. Although the capital expenditure outlook for US-based cloud service providers and second-tier CSPs is robust, and there is a recovery in demand for optical components both inside and outside the data center as well as in telecom investment, the market is concerned that the magnitude of earnings revisions will be insufficient to justify the high valuations. Within the optical components sector, JPM generally prefers Lumentum, followed by Coherent, and remains cautious on Fabrinet.\n\nIndustry Background and Overall Judgment\nThe optical components sector has significantly outperformed the broader market in recent months. The capex outlook for US-based cloud providers and second-tier CSPs is favorable, and industry checks indicate strong demand from data center and telecom investments. However, the high valuations mean that earnings reports must significantly beat expectations to drive stock prices higher. The overall strategy is to remain cautious on earnings and to favor stocks with greater potential for upside.\n\nLumentum: Expected to Beat on Both Revenue and Profit\nDriven by comprehensive growth in telecom, datacom lasers, and optical modules, improvements in production capacity and yield are contributing to upward revisions for both revenue and gross margin. F4Q25 revenue is projected to be $484 million (+57% YoY, higher than consensus and guidance), with a gross margin of 38.0%, an operating margin of 15.5%, and an EPS of $0.89, all of which are above market expectations. For F1Q26, revenue is projected at $515 million (+53% YoY) with a gross margin of 39.4% and an EPS of $0.97, both also above consensus. Its valuation is in line with the industry average, and it is positioned to benefit the most.\n\nCoherent: Solid Execution but Lacks Short-Term Catalysts\nAlthough its fundamentals are solid, investor focus has recently decreased due to a lack of new catalysts and macroeconomic uncertainty constraining industrial growth. F4Q25 revenue is projected at $1.55 billion (+18% YoY, higher than consensus and guidance), with a gross margin of 38.3% and an EPS of $0.98, slightly exceeding expectations. F1Q26 revenue is projected at $1.552 billion (+15% YoY) with an EPS of $0.99, which is largely in line with market expectations. In the long term, continuously improving revenue and profits, along with portfolio optimization, should help bolster confidence among long-only investors.\n\nFabrinet: 1.6T Ramp-Up Pace Under Scrutiny Amid High Valuation\nThe stock price has risen significantly, driven by ramp-ups at Amazon and Ciena, and is expected to ramp in sync with Nvidia's 1.6T optical modules. However, its current CY27E P/E ratio of over 21x is significantly higher than its peers, and the pace of the 1.6T volume ramp-up in F1Q26 will be under intense scrutiny. F4Q25 revenue is projected at $890 million (+18% YoY, higher than consensus and guidance), with an EPS of $2.67. For F1Q26, revenue is projected at $927 million (+15% YoY) with an EPS of $2.78, both slightly above expectations, but the high valuation magnifies the impact of any potential delay risks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.24397163380338682, "ret_signed_m1d": 0.24397163380338682, "alpha_spy_m1d": 0.24406459061645314, "alpha_c_m1d": 0.2418054743707565, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": 0.0021661594326303213, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019094380478606565, "ret_signed_p1d": 0.019094380478606565, "alpha_spy_p1d": 0.021435575120629324, "alpha_c_p1d": 0.026676796853659424, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.007582416375052858, "ret_p1w": -0.05510095982193586, "ret_signed_p1w": -0.05510095982193586, "alpha_spy_p1w": -0.040526209564634486, "alpha_c_p1w": -0.02215729942336342, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.03294366039857244, "ret_p1m": 0.12373154554382171, "ret_signed_p1m": 0.12373154554382171, "alpha_spy_p1m": 0.10441218029188004, "alpha_c_p1m": 0.11196593619460105, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.011765609349220663, "ret_p3m": 0.7240588311634581, "ret_signed_p3m": 0.7240588311634581, "alpha_spy_p3m": 0.6621260204242931, "alpha_c_p3m": 0.6282289083838888, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.09582992277956937, "ret_p6m": 1.4842333757244073, "ret_signed_p6m": 1.4842333757244073, "alpha_spy_p6m": 1.4072800228656894, "alpha_c_p6m": 1.4270063436294942, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.05722703209491309, "price_path": [-0.1339, -0.1331, -0.1319, -0.1393, -0.1439, -0.1439, -0.1141, -0.1462, -0.1324, -0.1748, -0.1622, -0.1217, -0.1093, -0.1309, -0.13, -0.1141, -0.1051, -0.1106, -0.1167, -0.1558, -0.1188, -0.1296, -0.1021, -0.1021, -0.114, -0.1201, -0.1166, -0.0657, -0.0193, -0.0482, -0.0266, -0.0547, -0.0359, -0.0082, -0.0082, -0.0393, -0.0136, -0.0044, 0.0226, 0.018, 0.0313, 0.0482, 0.0673, 0.0942, 0.0867, 0.0898, 0.0586, 0.074, 0.0771, 0.0926, 0.138, 0.1673, 0.17, 0.174, 0.1215, 0.1646, 0.1522, 0.1691, 0.2419, 0.2596, 0.2395, 0.2718, 0.244, 0.0, 0.0191, -0.0127, -0.0424, -0.0556, -0.0551, -0.0191, -0.0125, -0.0008, -0.0103, 0.0387, -0.0129, -0.0129, -0.042, -0.0347, 0.0433, 0.0675, 0.0766, 0.0826, 0.1292, 0.1294, 0.1237, 0.1603, 0.1399, 0.1283, 0.1789, 0.1905, 0.2504, 0.1925, 0.1622, 0.1628, 0.1674, 0.1781, 0.1753, 0.251, 0.2307, 0.2393, 0.2523, 0.2391, 0.273, 0.335, 0.2122, 0.2562, 0.1933, 0.2047, 0.2652, 0.2695, 0.3115, 0.3179, 0.2588, 0.3259, 0.4112, 0.4729, 0.4647, 0.5064, 0.448, 0.4398, 0.4403, 0.4043, 0.469, 0.7381, 0.6859, 0.8191, 0.7241, 0.7094, 0.5272, 0.5202, 0.5174, 0.5074, 0.5596, 0.4797, 0.5222, 0.6564, 0.6241, 0.6803, 0.6803, 0.7923, 0.784, 0.7991, 0.8654, 0.9351, 0.9835, 1.0279, 1.1029, 1.1544, 1.1658, 0.9459, 0.9471, 0.9116, 0.8597, 0.9172, 1.0276, 1.0838, 1.0935, 1.0881, 1.0881, 1.0919, 1.0624, 1.0383, 1.0139, 1.0139, 1.1203, 1.0334, 1.1179, 1.0908, 0.8893, 0.9428, 1.0205, 1.0734, 1.0088, 1.1381, 1.0845, 1.0845, 1.1109, 1.1981, 1.2119, 1.1488, 1.1578, 1.335, 1.4129, 1.3553, 1.3151, 1.4271, 1.5006, 1.3022, 1.283, 1.4842]}
{"unit_id": "orig:1955808046862754094", "ticker_idx": 2, "ts": "2025-08-14T01:45:50+00:00", "symbol": "Fabrinet", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "remains cautious on Fabrinet", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM adopts optical sector note: prefers Lumentum (beat expected on rev/EPS), followed by Coherent; cautious on Fabrinet due to high valuation and 1.6T ramp scrutiny.", "resolved_tickers": ["FN"], "resolution_method": "exchange_ticker", "resolution_reason": "Fabrinet FN US optical manufacturing", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Communications F4Q25 Outlook: Fundamentals Remain Solid, but High Valuations Set a High Bar for Earnings\nJPM (S. Chatterjee, 11/08/25)\n\nStock prices for optical component companies have risen by an average of about 50% since early June, with valuations exceeding 27 times NTM (Next Twelve Months) earnings, setting a high execution bar for F4Q25 financial reports. Although the capital expenditure outlook for US-based cloud service providers and second-tier CSPs is robust, and there is a recovery in demand for optical components both inside and outside the data center as well as in telecom investment, the market is concerned that the magnitude of earnings revisions will be insufficient to justify the high valuations. Within the optical components sector, JPM generally prefers Lumentum, followed by Coherent, and remains cautious on Fabrinet.\n\nIndustry Background and Overall Judgment\nThe optical components sector has significantly outperformed the broader market in recent months. The capex outlook for US-based cloud providers and second-tier CSPs is favorable, and industry checks indicate strong demand from data center and telecom investments. However, the high valuations mean that earnings reports must significantly beat expectations to drive stock prices higher. The overall strategy is to remain cautious on earnings and to favor stocks with greater potential for upside.\n\nLumentum: Expected to Beat on Both Revenue and Profit\nDriven by comprehensive growth in telecom, datacom lasers, and optical modules, improvements in production capacity and yield are contributing to upward revisions for both revenue and gross margin. F4Q25 revenue is projected to be $484 million (+57% YoY, higher than consensus and guidance), with a gross margin of 38.0%, an operating margin of 15.5%, and an EPS of $0.89, all of which are above market expectations. For F1Q26, revenue is projected at $515 million (+53% YoY) with a gross margin of 39.4% and an EPS of $0.97, both also above consensus. Its valuation is in line with the industry average, and it is positioned to benefit the most.\n\nCoherent: Solid Execution but Lacks Short-Term Catalysts\nAlthough its fundamentals are solid, investor focus has recently decreased due to a lack of new catalysts and macroeconomic uncertainty constraining industrial growth. F4Q25 revenue is projected at $1.55 billion (+18% YoY, higher than consensus and guidance), with a gross margin of 38.3% and an EPS of $0.98, slightly exceeding expectations. F1Q26 revenue is projected at $1.552 billion (+15% YoY) with an EPS of $0.99, which is largely in line with market expectations. In the long term, continuously improving revenue and profits, along with portfolio optimization, should help bolster confidence among long-only investors.\n\nFabrinet: 1.6T Ramp-Up Pace Under Scrutiny Amid High Valuation\nThe stock price has risen significantly, driven by ramp-ups at Amazon and Ciena, and is expected to ramp in sync with Nvidia's 1.6T optical modules. However, its current CY27E P/E ratio of over 21x is significantly higher than its peers, and the pace of the 1.6T volume ramp-up in F1Q26 will be under intense scrutiny. F4Q25 revenue is projected at $890 million (+18% YoY, higher than consensus and guidance), with an EPS of $2.67. For F1Q26, revenue is projected at $927 million (+15% YoY) with an EPS of $2.78, both slightly above expectations, but the high valuation magnifies the impact of any potential delay risks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04728642455687315, "ret_signed_m1d": -0.04728642455687315, "alpha_spy_m1d": -0.047379381369939466, "alpha_c_m1d": -0.04512026512424283, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": 0.0021661594326303213, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023969204988062653, "ret_signed_p1d": -0.023969204988062653, "alpha_spy_p1d": -0.026310399630085413, "alpha_c_p1d": -0.03155162136311551, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.007582416375052858, "ret_p1w": -0.13984097089260794, "ret_signed_p1w": 0.13984097089260794, "alpha_spy_p1w": 0.12526622063530657, "alpha_c_p1w": 0.1068973104940355, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.03294366039857244, "ret_p1m": 0.11152513327516211, "ret_signed_p1m": -0.11152513327516211, "alpha_spy_p1m": -0.09220576802322045, "alpha_c_p1m": -0.09975952392594145, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.011765609349220663, "ret_p3m": 0.39496406879076273, "ret_signed_p3m": -0.39496406879076273, "alpha_spy_p3m": -0.3330312580515977, "alpha_c_p3m": -0.29913414601119337, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.09582992277956937, "ret_p6m": 0.564797600739138, "ret_signed_p6m": -0.564797600739138, "alpha_spy_p6m": -0.48784424788042013, "alpha_c_p6m": -0.5075705686442249, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.05722703209491309, "price_path": [-0.2928, -0.2912, -0.2953, -0.2913, -0.285, -0.285, -0.2713, -0.2803, -0.2708, -0.277, -0.2754, -0.2552, -0.2571, -0.2635, -0.2705, -0.2771, -0.2619, -0.2449, -0.2022, -0.239, -0.1976, -0.1965, -0.1755, -0.1755, -0.1735, -0.1621, -0.1415, -0.1124, -0.0719, -0.0822, -0.0851, -0.1024, -0.1032, -0.0851, -0.0851, -0.11, -0.1109, -0.1145, -0.1019, -0.0951, -0.0795, -0.064, -0.0765, -0.0426, -0.0346, -0.0391, -0.0661, -0.0524, -0.0535, -0.041, -0.0169, -0.004, -0.0044, 0.0051, -0.0214, 0.0233, 0.0187, 0.0322, 0.0405, 0.068, 0.0321, 0.0815, 0.0473, 0.0, 0.024, 0.0156, -0.1144, -0.1296, -0.1398, -0.0861, -0.0304, 0.0266, 0.0319, 0.1024, 0.0286, 0.0286, 0.0423, 0.048, 0.1244, 0.1489, 0.1509, 0.146, 0.1568, 0.1244, 0.1115, 0.1146, 0.1039, 0.1122, 0.1737, 0.1895, 0.2092, 0.1805, 0.1384, 0.1207, 0.1143, 0.1296, 0.1321, 0.1529, 0.1474, 0.1364, 0.1497, 0.1493, 0.1905, 0.1868, 0.1238, 0.1879, 0.1651, 0.2368, 0.2806, 0.2727, 0.2654, 0.2915, 0.2427, 0.2885, 0.3071, 0.3152, 0.3585, 0.3783, 0.3599, 0.3679, 0.3754, 0.427, 0.5118, 0.4321, 0.3945, 0.4507, 0.395, 0.3894, 0.2496, 0.2833, 0.2936, 0.2706, 0.2937, 0.2022, 0.2123, 0.2927, 0.3341, 0.3942, 0.3942, 0.4264, 0.3959, 0.371, 0.3888, 0.4648, 0.4855, 0.5246, 0.5476, 0.604, 0.6399, 0.462, 0.4593, 0.4008, 0.3243, 0.4048, 0.4762, 0.5013, 0.5118, 0.4957, 0.4957, 0.4851, 0.4657, 0.4342, 0.4136, 0.4136, 0.4885, 0.4186, 0.4877, 0.4608, 0.3752, 0.3886, 0.4373, 0.5183, 0.4812, 0.5449, 0.5352, 0.5352, 0.4983, 0.5163, 0.4605, 0.4497, 0.4721, 0.548, 0.5555, 0.5408, 0.5196, 0.5512, 0.3926, 0.3134, 0.3745, 0.5648]}
{"unit_id": "quote:1961966021683945889", "ticker_idx": 0, "ts": "2025-08-31T01:35:25+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "A single Stargate project alone would add almost as much NVIDIA chip demand over several years as the entire total volume projected for 2026", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish NVDA: Stargate alone could add ~70,000 GB200 racks of demand, nearly matching all projected 2026 volume — massive incremental demand signal.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "The particularly interesting part is this:\n\n“JPMorgan estimated that if the 10GW Stargate project is implemented, it could generate demand equivalent to about 70,000 GB200 racks over the coming years.”\n\nThis is insane. A single Stargate project alone would add almost as much NVIDIA chip demand over several years as the entire total volume projected for 2026.", "reply_text": null, "parent_text": null, "quoted_text": "Hon Hai: Sovereign AI Investment May Reach USD 1 Trillion in the Next Five Years, Becoming a New Growth Driver for the Compute Market\n\nHon Hai believes that sovereign AI investment projects worth more than USD 1 trillion will emerge in the next five years, becoming a new and", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019520005102879878, "ret_signed_p1d": -0.019520005102879878, "alpha_spy_p1d": -0.012109709547005632, "alpha_c_p1d": -0.006395180224103014, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.013124824878776864, "ret_p1w": -0.03370073659685746, "ret_signed_p1w": -0.03370073659685746, "alpha_spy_p1w": -0.03956068805115309, "alpha_c_p1w": -0.055161885337244754, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.021461148740387292, "ret_p1m": 0.07125115097366086, "ret_signed_p1m": 0.07125115097366086, "alpha_spy_p1m": 0.035630738036498855, "alpha_c_p1m": -0.05300374348327597, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.12425489445693683, "ret_p3m": 0.03496480335906882, "ret_signed_p3m": 0.03496480335906882, "alpha_spy_p3m": -0.021642270435382915, "alpha_c_p3m": -0.16283726117581088, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.1978020645348797, "ret_p6m": 0.10731214475776785, "ret_signed_p6m": 0.10731214475776785, "alpha_spy_p6m": 0.03562426021932752, "alpha_c_p6m": -0.3411160852526818, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.4484282300104496, "price_path": [-0.1853, -0.1963, -0.1864, -0.1812, -0.1736, -0.18, -0.1675, -0.1849, -0.1693, -0.1726, -0.1648, -0.1648, -0.1741, -0.1723, -0.1509, -0.1141, -0.11, -0.0943, -0.0929, -0.1199, -0.0972, -0.0852, -0.0852, -0.0915, -0.0814, -0.0649, -0.0579, -0.0532, -0.058, -0.02, -0.0161, -0.0068, -0.0102, -0.0161, -0.041, -0.0195, -0.0025, -0.0039, 0.0148, 0.0076, 0.0292, 0.0212, -0.0026, 0.0334, 0.0234, 0.0301, 0.0378, 0.0489, 0.0452, 0.0516, 0.0425, 0.045, 0.036, 0.045, 0.0084, 0.007, 0.0046, 0.0219, 0.0323, 0.0436, 0.0426, 0.0344, 0.0, 0.0, -0.0195, -0.0204, -0.0145, -0.0411, -0.0337, -0.0196, 0.0181, 0.0172, 0.021, 0.0206, 0.0041, -0.0223, 0.0119, 0.0144, 0.0542, 0.0245, 0.0161, 0.0202, 0.0231, 0.0441, 0.0713, 0.075, 0.0845, 0.0772, 0.0653, 0.0624, 0.0858, 0.1056, 0.0516, 0.0812, 0.0336, 0.0325, 0.0439, 0.052, 0.0486, 0.0401, 0.0351, 0.0459, 0.0694, 0.0994, 0.1542, 0.1887, 0.1649, 0.1626, 0.1878, 0.1408, 0.1208, 0.0799, 0.0803, 0.1428, 0.109, 0.1127, 0.0729, 0.0919, 0.0714, 0.0413, 0.0709, 0.0371, 0.027, 0.0481, 0.021, 0.035, 0.035, 0.0162, 0.033, 0.0419, 0.0311, 0.0529, 0.0474, 0.0654, 0.0621, 0.0552, 0.0389, 0.0049, 0.0122, 0.0204, -0.0185, -0.0001, 0.0392, 0.0547, 0.0864, 0.083, 0.083, 0.094, 0.0807, 0.0768, 0.0709, 0.0709, 0.0843, 0.0802, 0.0751, 0.0858, 0.0625, 0.0614, 0.0619, 0.0669, 0.0516, 0.074, 0.0693, 0.0693, 0.0224, 0.0526, 0.0613, 0.0776, 0.0707, 0.0824, 0.0997, 0.1054, 0.0974, 0.0657, 0.0355, 0.0002, -0.0131, 0.0646, 0.0912, 0.0826, 0.0912, 0.0734, 0.0497, 0.0497, 0.0621, 0.0793, 0.0789, 0.0899, 0.0998, 0.1073]}
{"unit_id": "reply:1964954681303875600", "ticker_idx": 0, "ts": "2025-09-08T07:31:17+00:00", "symbol": "005380.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "현대차도 미워도 다시한번....이 아닌가 해서", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Soft bullish lean on Hyundai Motor — 'hate it but give it another chance' sentiment implies a long.", "resolved_tickers": ["005380.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Auto Manufacturers'", "bench_c_industry": "Auto Manufacturers", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@dons_korea 현대차도 미워도 다시한번....이 아닌가 해서 ㅎㅎㅎ", "reply_text": null, "parent_text": "@Jukanlosreve 그럼 현대차를 사야하는데요,,,", "quoted_text": null, "parent_author": "dons_korea", "ret_m1d": 0.00686497923125251, "ret_signed_m1d": 0.00686497923125251, "alpha_spy_m1d": 0.00931550132973269, "alpha_c_m1d": 0.010720591443638905, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.0038556122123863945, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0022882778872903575, "ret_signed_p1d": 0.0022882778872903575, "alpha_spy_p1d": -2.3642810319524088e-05, "alpha_c_p1d": 0.0036864705040426937, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": -0.0013981926167523362, "ret_p1w": -0.016018381919176483, "ret_signed_p1w": -0.016018381919176483, "alpha_spy_p1w": -0.03463650603032742, "alpha_c_p1w": -0.03584718899161354, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.01982880707243706, "ret_p1m": 0.00686497923125251, "ret_signed_p1m": 0.00686497923125251, "alpha_spy_p1m": -0.027265876618641993, "alpha_c_p1m": 0.006459429896244151, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.00040554933500835944, "ret_p3m": 0.3100066648941169, "ret_signed_p3m": 0.3100066648941169, "alpha_spy_p3m": 0.2522759085032973, "alpha_c_p3m": 0.29924474972850934, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.010761915165607538, "ret_p6m": 1.7614662999968047, "ret_signed_p6m": 1.7614662999968047, "alpha_spy_p6m": 1.7069032639184143, "alpha_c_p6m": 1.7887628198282126, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": -0.027296519831408017, "price_path": [-0.0905, -0.0883, -0.0996, -0.0883, -0.0724, -0.0679, -0.0634, -0.0498, -0.0883, -0.0679, -0.0182, -0.0521, -0.0724, -0.0792, -0.0498, -0.034, -0.0295, -0.0476, -0.0566, -0.0453, -0.0521, -0.0589, -0.0589, -0.0182, -0.0453, -0.0611, -0.0498, -0.0476, -0.0543, -0.0657, 0.0045, -0.0159, -0.0204, -0.0114, -0.0136, 0.009, -0.0362, -0.0498, -0.0453, -0.0476, -0.0476, -0.0385, -0.0385, -0.0385, -0.0362, -0.0227, -0.0159, -0.0159, -0.0204, -0.0091, -0.0023, 0.0022, -0.0046, 0.0045, -0.0114, -0.0046, 0.016, 0.0069, 0.0092, 0.0069, 0.0137, 0.0137, 0.0069, 0.0, 0.0023, 0.0092, 0.0206, 0.0229, -0.016, -0.016, -0.0114, -0.0092, -0.0092, -0.0023, 0.0023, -0.0046, -0.0069, -0.0183, -0.016, -0.016, -0.0137, 0.0069, 0.0069, 0.0069, 0.0069, 0.0069, 0.0069, -0.0069, 0.0, 0.0206, 0.0229, 0.1076, 0.1121, 0.135, 0.1739, 0.1945, 0.1533, 0.1556, 0.1648, 0.1465, 0.1808, 0.2128, 0.3272, 0.3341, 0.2632, 0.2288, 0.2311, 0.2082, 0.238, 0.2311, 0.2609, 0.2746, 0.2471, 0.2426, 0.2105, 0.2082, 0.1991, 0.1876, 0.1785, 0.1785, 0.1968, 0.2083, 0.2083, 0.176, 0.2291, 0.2315, 0.31, 0.4556, 0.4579, 0.4186, 0.3978, 0.3655, 0.3932, 0.3562, 0.3216, 0.3216, 0.3054, 0.3331, 0.3354, 0.3262, 0.3354, 0.3354, 0.3216, 0.3562, 0.3701, 0.3701, 0.3701, 0.3793, 0.407, 0.4232, 0.6196, 0.5734, 0.6912, 0.6958, 0.8761, 0.9015, 0.95, 0.9084, 1.218, 1.2134, 1.5368, 1.4444, 1.3566, 1.2758, 1.2573, 1.2758, 1.4398, 1.3104, 1.2088, 1.2711, 1.3289, 1.2573, 1.1602, 1.2088, 1.2203, 1.352, 1.3381, 1.3058, 1.3058, 1.3058, 1.3058, 1.3705, 1.352, 1.4167, 1.4213, 1.6431, 1.8264, 2.1281, 2.1281, 1.7615]}
{"unit_id": "orig:1972892318245228853", "ticker_idx": 0, "ts": "2025-09-30T05:12:37+00:00", "symbol": "EMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains its Buy rating on EMC and TUC, raising their 12-month target prices to NT$1,530 and NT$410, respectively", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman Sachs reiterates Buy on Taiwan CCL makers EMC (PT NT$1,530) and TUC (PT NT$410) on M9 CCL adoption driving 50-60% growth in 2026 and 100%+ in 2027.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "CCL context: EMC = Elite Material Co. 2383.TW not US EMC", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "High-End CCL/PCB Demand Accelerates: M9 Adoption Brings New Opportunities\n\nGoldman Sachs (A. Chang, 250929)\n\nGoldman Sachs stated that the adoption of M9-grade CCL (Q-glass version) will accelerate significantly from the second half of 2026 to 2027, with application scenarios including Vera Rubin CPX midboards, major ASIC CSP projects (TPU, Trainium, META ASIC, etc.), Vera Rubin Ultra’s switchboards and compute boards, as well as backplanes in Kyber cabinet cross-back applications. EMC’s M9-grade CCL has passed customer certification and holds a leading advantage, with its ASP expected to exceed $400 per sheet, far higher than M7’s $100–110 and M8’s $160–180, significantly boosting the company’s profitability.\n\nAlthough NVIDIA’s backplane PCB solution failed in its first sample test in September, raising investor concerns that the project might be delayed until 2028, Goldman Sachs believes that the first failure is part of normal R&D and does not affect the mass production schedule for the second half of 2027. It is expected that NVIDIA’s backplane PCB will then adopt 76–98 layers of M9 CCL, with the related market size reaching over $5 billion, while the CCL market space is expected to reach $2 billion from the second half of 2027 to the first half of 2028. Compared to the overall M7+ CCL market of about $8 billion in 2027, the contribution is significant.\n\nIn the short term, AWS’s PCB/CCL orders have been delayed to early 2026. Coupled with Trainium 2 demand gradually tapering off after the third quarter of 2025, some suppliers may face revenue pressure during 4Q25–1Q26. However, as TUC holds a relatively high share in Trainium 2.5/3 projects, the negative impact will be limited. Leading companies such as EMC can fill capacity gaps through other high-end AI server and 800G projects, even though a 5–10% production loss may occur during the capacity switching period.\n\nOverall, Goldman Sachs expects that the CCL/PCB market driven by high-end AI projects will still maintain 50–60% year-on-year growth in 2026 and accelerate to over 100% in 2027. Any short-term pullback will become a good opportunity for medium- to long-term investment positioning. Goldman Sachs maintains its Buy rating on EMC and TUC, raising their 12-month target prices to NT$1,530 and NT$410, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.003752730722497377, "alpha_c_m1d": 0.005428340916486141, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.005428340916486141, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004081632653061273, "ret_signed_p1d": 0.004081632653061273, "alpha_spy_p1d": 0.0006740632271282632, "alpha_c_p1d": -0.0057104003734174125, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.009792033026478686, "ret_p1w": 0.008163265306122547, "ret_signed_p1w": 0.008163265306122547, "alpha_spy_p1w": 0.0037500657705515916, "alpha_c_p1w": -0.005389455601176207, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.013552720907298754, "ret_p1m": 0.020408163265306145, "ret_signed_p1m": 0.020408163265306145, "alpha_spy_p1m": -0.011430063891922027, "alpha_c_p1m": -0.05860276214207194, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.07901092540737809, "ret_p3m": 0.3387755102040817, "ret_signed_p3m": 0.3387755102040817, "alpha_spy_p3m": 0.29949217415104434, "alpha_c_p3m": 0.2974624462379707, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.04131306396611101, "ret_p6m": 1.3224489795918366, "ret_signed_p6m": 1.3224489795918366, "alpha_spy_p6m": 1.3308852524137385, "alpha_c_p6m": 1.3493579076047053, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": -0.026908928012868705, "price_path": [-0.2523, -0.2563, -0.274, -0.2321, -0.2289, -0.2289, -0.2418, -0.2523, -0.2184, -0.2079, -0.216, -0.1741, -0.1822, -0.2023, -0.2039, -0.2031, -0.1991, -0.1741, -0.1822, -0.1902, -0.1096, -0.1096, -0.1056, -0.0935, -0.1056, -0.0774, -0.0734, -0.0694, -0.0532, -0.0331, -0.025, 0.0314, 0.0515, -0.0089, -0.0734, -0.0049, -0.0291, 0.0072, 0.0072, 0.0314, 0.0204, -0.0041, -0.0571, -0.0612, -0.049, -0.0898, -0.049, -0.0653, 0.0, 0.0245, 0.0816, 0.098, 0.0735, 0.049, 0.0735, 0.0898, 0.0531, 0.0612, 0.098, 0.0571, 0.0245, 0.0, 0.0, 0.0, 0.0041, 0.0571, 0.0653, 0.0653, 0.0082, -0.0245, 0.0082, 0.0082, -0.0204, -0.0449, -0.0694, -0.0694, -0.098, -0.0816, -0.0776, -0.0857, -0.0898, -0.0898, -0.049, 0.0245, 0.0204, 0.0408, 0.1102, 0.151, 0.2, 0.1755, 0.1837, 0.1184, 0.1102, 0.0776, 0.098, 0.1224, 0.1265, 0.1837, 0.102, 0.0857, 0.1673, 0.0694, 0.0776, 0.1837, 0.2, 0.2204, 0.2449, 0.2041, 0.1755, 0.1673, 0.1633, 0.2204, 0.2122, 0.2245, 0.3061, 0.302, 0.3102, 0.3102, 0.2204, 0.2204, 0.2245, 0.2653, 0.3061, 0.3184, 0.3796, 0.3796, 0.3388, 0.3673, 0.3429, 0.3429, 0.3429, 0.298, 0.3224, 0.3837, 0.2898, 0.2612, 0.2776, 0.3143, 0.3347, 0.3633, 0.2816, 0.3265, 0.2898, 0.3184, 0.2735, 0.3551, 0.4694, 0.4204, 0.4857, 0.551, 0.4612, 0.4245, 0.4571, 0.5143, 0.6653, 0.5918, 0.5714, 0.698, 0.7184, 0.7918, 0.7918, 0.7918, 0.7918, 0.7918, 0.7918, 0.7918, 0.7918, 0.7755, 0.8857, 0.9755, 0.9918, 0.9918, 0.9429, 0.8939, 0.7878, 0.9306, 0.9265, 0.7755, 0.9388, 1.0, 1.1224, 1.2082, 1.1102, 1.2041, 1.2041, 1.3592, 1.2939, 1.2816, 1.1918, 1.3224]}
{"unit_id": "quote:1949654885882007569", "ticker_idx": 0, "ts": "2025-07-28T02:15:22+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "every SDE working at an internet company should buy Nvidia stock as a hedge", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish NVDA: AI-era GPU capex grows annually unlike one-time internet infra, making Nvidia the structural beneficiary as internet companies shift spend from people to GPUs.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "What are the differences in the operating logic and models between the AI era and the Internet era?\n\nI've been thinking about this question recently, and I'm trying to look at it from a long-term perspective.\n\nAt the GTC conference two years ago, Jensen Huang unveiled a slide that announced the curtain opening on a new era: the PC Era -> the Internet Era -> the AI Era. The great hardware infrastructure build-out for each new era has begun, and it might be a repeat of the Cisco story, with its rising and falling tides.\n\nIn the Cisco era, once the infrastructure was built, the subsequent demand for it decreased significantly. It was like laying down the pipelines; internet companies built various applications on top of these pipelines. Even if the infra needed expansion, the construction costs were largely a one-time expense with long depreciation cycles thereafter. Consequently, after a brief explosion in the early internet era, Cisco quickly became an outcast. If you had bought its stock in 2000, you wouldn't have broken even until 2021, massively underperforming the S&P 500.\n\nThe mobile internet era was the same. Once the mobile SoC infrastructure was established, the number of phones that could be sold each year was basically fixed. This is why Qualcomm became the most illusory and poor investment in the semiconductor field over the past decade. The prosperity of the mobile internet economy didn't bring much incremental growth to Qualcomm; the underlying infrastructure only saw routine replacement and upgrade cycles for phones.\n\nBut looking back now, two years later, the AI era might not allow for a direct application of past experiences. The logic is different. Training infrastructure is not a one-time expense. The daily operating costs far exceed those of the Internet era. The heavy workload on GPUs means they have a lifespan of only two to three years and need replacement, and they are extremely expensive.\n\nWhy is Meta spending what amounts to hundreds of millions in signing bonuses per person to poach top AI talent for foundational models? Is this a foolish expenditure?\n\nIt makes sense from the perspective of Meta's high infrastructure costs. Since Meta is investing $70 billion a year in GPU infrastructure—a figure that is still growing—and is even getting directly involved in building gigawatt-scale data centers, it seems absolutely necessary to hire a group of the world's foremost experts to make the best use of this $70B-a-year luxury GPU data center. In comparison, a few billion dollars in recruitment fees even looks like a bargain.\n\nAt Google/OpenAI, \"Compute-per-Researcher\" is already a key management KPI. The idea that after Meta spends hundreds of millions on people, the corresponding compute power must be fully matched, helps to understand Meta's motivation for poaching talent at high prices when viewed in reverse.\n\nThe people working on foundation model training at these large companies (especially pre-training) are different from SDEs, and even different from MLEs, a new role that only emerged a decade ago. This is a completely new profession born of this era: managers of compute expenditure efficiency. Without experience training on thousand-GPU clusters, you don't even meet the entry threshold for the industry.\n\nFor the first time in human history, a new profession has emerged where a few dozen to a hundred people operate a massive army of machines with annual expenses equivalent to the GDP of a country (in the $100B range = Croatia/Costa Rica).\n\nIn the Internet era, companies like Google, Amazon, and Facebook didn't need to bear much pressure on infrastructure. They just needed to wait for the internet infrastructure to mature, and they could build a business model for information flow on top of this network. The network and compute costs for each request had an extremely low marginal cost, which made scaling incredibly effective. The distribution marginal cost was nearly zero, and the returns from rapid expansion were astonishing, growing exponentially with scale.\n\nThis characteristic also led to a phenomenon: the largest OPEX for internet companies was SDE labor costs. This is a typical feature of the tertiary (service) industry and a key reason why SDEs enjoyed a golden age over the past fifteen years, with their salaries continuously rising along with business expansion.\n\nIn the Internet era, the scarce resource was the work time of software engineers. The focus of financial reports was on OPEX salaries, and the moat was the network effect and the near-zero cost of infinite replication and distribution. While Google/FB also built their own data centers, CAPEX was around 10%.\n\nHowever, in the LLM era, at least for the next few years, the scarce resources for internet companies have become GPUs and power supply capacity (at the gigawatt level). The heavyweight on financial reports is now CAPEX (Microsoft's CAPEX ratio is 33%, and Meta's is already approaching 40%). GPUs have undoubtedly become heavy, high-consumption assets.\n\nFor the first time in their history, internet companies are carrying balance sheets with high depreciation costs, similar to semiconductor foundries. The business model is almost shifting from \"traffic × conversion rate\" to \"gross profit per token.\"\n\nA key difference between the tertiary and secondary industries is the need to manage heavy assets and continuous operating costs. The nature of internet companies will transform from the tertiary sector to becoming more \"industrialized\" (like the secondary sector). The value of employees as assets will not be as precious. The golden age of unlimited SDE salary premiums may be difficult to sustain in the AI era. They will have to get a taste of the standard treatment for employees in the semiconductor industry—for example, asking for too high a salary could lead to the offer being rescinded.\n\nAs internet companies continuously shift resources from personnel to GPU purchases, squeezing personnel costs, cutting benefits, and constantly laying off and replacing staff, I think every SDE working at an internet company should buy Nvidia stock as a hedge (to compensate for the risk of being squeezed out of the value chain by GPUs).\n\nIn the AI era, the Internet era's characteristic of near-zero marginal cost, which favored scalability, faces a fundamental challenge. Not only are training costs no longer a one-time expense but they grow annually, but as for customer AI inference requests, due to the consensus on inference scaling and the need for larger models to achieve better results, the cost of inference may not decrease with the deflation in hardware compute prices.\n\nIt's like the mobile phone chip industry back then. As chip performance improved each year, battery life logically should have increased significantly. But the optimal business logic was to improve the user experience as much as possible within the power consumption limits people could tolerate (a battery that lasts a day), rather than maintaining the same experience and reducing power consumption. Otherwise, phones could have achieved standby times of several months.\n\nAI inference in the AI era is the same. Once the cost of GPT-4o drops, people will just use the better-experiencing GPT-4.5. The cost of GPT-4.5 might be higher than GPT-4 was two years ago, and usage might be limited to just a few queries a day. An agent might become ten times cheaper to run for the same task in half a year, but a new, more effective agent will pull the price right back up.\n\nThis AI inference cost might behave just like mobile chip power consumption back then: within the cost limits that people (and businesses) can tolerate, the experience will be improved as much as possible. Therefore, AI inference costs won't necessarily decrease, meaning the distribution marginal cost, a key metric in the Internet era, will become much higher.\n\nThis also leads internet companies to push their investment in this area to the limit of what their financial reports can tolerate, to maximize their compute capacity. This might lead to an unprecedented phenomenon: until scaling laws fail, the growth in compute spending will not be lower than the company's revenue growth.\n\nStarting from this change in the macroeconomic background—that is, starting from the differences in the underlying compute characteristics of AI and the internet—we can perhaps try to deduce what strategic and business model changes this will bring.\n\n(To be continued)", "reply_text": null, "parent_text": null, "quoted_text": "AI时代和互联网时代的运行逻辑和模式有什么不一样？\n\n最近一直在思考这个问题，也在尝试从时间线拉长的角度去看一看\n\n两年前的GTX大会，老黄po出来的这张图宣布了时代大幕的拉开，PC时代-&gt;互联网时代-&gt;AI时代，每一个时代前期的硬件大基建时代开始了，可能又是一次cisco时代涨潮退潮的故事重现 https://t.co/PvqPCJgRqi", "parent_author": null, "ret_m1d": -0.018387500892582453, "ret_signed_m1d": -0.018387500892582453, "alpha_spy_m1d": -0.018638664104893232, "alpha_c_m1d": -0.004935570640966391, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0070154606938631625, "ret_signed_p1d": -0.0070154606938631625, "alpha_spy_p1d": -0.004377715453377062, "alpha_c_p1d": -0.01212844007901348, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": 0.018387587236714076, "ret_signed_p1w": 0.018387587236714076, "alpha_spy_p1w": 0.02744666451583666, "alpha_c_p1w": 0.022574133247488137, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": 0.02840169327749642, "ret_signed_p1m": 0.02840169327749642, "alpha_spy_p1m": 0.015496310875948938, "alpha_c_p1m": 0.010077032322802548, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.03066632716154105, "ret_signed_p3m": 0.03066632716154105, "alpha_spy_p3m": -0.02692523313603945, "alpha_c_p3m": -0.1525462462800229, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 0.007581101100084586, "ret_signed_p6m": 0.007581101100084586, "alpha_spy_p6m": -0.062325380949863396, "alpha_c_p6m": -0.3362462778582278, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.3838, -0.3686, -0.3522, -0.3561, -0.3577, -0.3378, -0.336, -0.3401, -0.3042, -0.2649, -0.2343, -0.2372, -0.234, -0.233, -0.2398, -0.2544, -0.2485, -0.2573, -0.2573, -0.2334, -0.2373, -0.2126, -0.2355, -0.2228, -0.2011, -0.1971, -0.208, -0.1982, -0.1931, -0.1856, -0.1919, -0.1796, -0.1968, -0.1814, -0.1846, -0.1769, -0.1769, -0.1861, -0.1843, -0.1632, -0.127, -0.1229, -0.1075, -0.1061, -0.1327, -0.1103, -0.0985, -0.0985, -0.1047, -0.0948, -0.0785, -0.0716, -0.0669, -0.0717, -0.0342, -0.0304, -0.0212, -0.0246, -0.0304, -0.055, -0.0338, -0.017, -0.0184, 0.0, -0.007, 0.0143, 0.0063, -0.0171, 0.0184, 0.0085, 0.0151, 0.0227, 0.0337, 0.03, 0.0363, 0.0274, 0.0298, 0.0209, 0.0298, -0.0063, -0.0076, -0.01, 0.007, 0.0173, 0.0284, 0.0274, 0.0193, -0.0145, -0.0145, -0.0338, -0.0347, -0.0288, -0.055, -0.0478, -0.0339, 0.0033, 0.0024, 0.0061, 0.0057, -0.0105, -0.0365, -0.0028, -0.0004, 0.0389, 0.0096, 0.0013, 0.0054, 0.0082, 0.0289, 0.0557, 0.0594, 0.0687, 0.0616, 0.0498, 0.047, 0.07, 0.0896, 0.0363, 0.0655, 0.0186, 0.0175, 0.0287, 0.0367, 0.0334, 0.025, 0.02, 0.0307, 0.0539, 0.0835, 0.1374, 0.1714, 0.148, 0.1457, 0.1705, 0.1242, 0.1045, 0.0642, 0.0646, 0.1262, 0.0929, 0.0965, 0.0573, 0.076, 0.0558, 0.0261, 0.0553, 0.0221, 0.0121, 0.0329, 0.0061, 0.0199, 0.0199, 0.0015, 0.018, 0.0267, 0.0161, 0.0376, 0.0321, 0.0499, 0.0466, 0.0399, 0.0238, -0.0097, -0.0025, 0.0056, -0.0328, -0.0147, 0.0241, 0.0394, 0.0706, 0.0672, 0.0672, 0.0781, 0.065, 0.0612, 0.0553, 0.0553, 0.0686, 0.0644, 0.0595, 0.07, 0.047, 0.046, 0.0465, 0.0514, 0.0363, 0.0584, 0.0538, 0.0538, 0.0076]}
{"unit_id": "orig:1945464661815816540", "ticker_idx": 0, "ts": "2025-07-16T12:44:55+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": "EUV", "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EW, Target Price EUR 660", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Morgan Stanley reiterates Equal Weight on ASML with EUR 660 target; 2026 uncertainty offsets strong orders, medium-term growth story less clear.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley) ASML Earnings Quick Analysis: EW, Target Price EUR 660\n\nKey Summary\n\n1. Comments on potential stagnation in 2025 equipment unit shipments and uncertainty about 2026 growth may raise some concerns.\n\n2. Strong new orders are offset by 2026 uncertainty.\n\n3. 2025 guidance was narrowed to the midpoint of previous forecasts (approximately 15% growth year-over-year from 2024).\n\n4. New orders reached EUR 5.5 billion, exceeding estimates (Morgan Stanley's estimate of EUR 4.6 billion), with EUR 2.3 billion coming from EUV equipment (estimated at approximately 10 units).\n\n5. Margin was better than expected due to strong IBM (Installed Base Management) revenue and one-off factors, but the outlook for a slowdown in the upgrade business in the second half and a consequent weaker gross margin (GM%) forecast was maintained.\n\n6. In conclusion, this announcement somewhat increased the uncertainty surrounding the previously presented medium-term growth story.\n\nAnnouncement Details\nQ2 new orders (Order Book) recorded EUR 5.5 billion, surpassing the consensus (EUR 4.5 billion), with EUR 2.3 billion in EUV equipment orders. The company adjusted its full-year 2025 guidance to \"approximately 15% growth compared to 2024,\" which aligns with the midpoint of the previous guidance, EUR 32.5 billion.\n\nWithin this growth outlook, ASML projected that the EUV business unit would grow by 30%, the IBM (Installed Base Management) business unit by 20%, and the DUV business unit would remain at a similar level to 2024. Additionally, based on the current order backlog, the company maintained its previous expectation that China's revenue share would exceed 25% in 2025.\n\nRegarding the 2026 outlook, considering macroeconomic and geopolitical situations (including tariffs), the company took a more cautious stance than before, stating that they are \"preparing for growth but cannot confirm it at this stage.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.12831577350590062, "ret_signed_m1d": -0.12831577350590062, "alpha_spy_m1d": -0.131647824939656, "alpha_c_m1d": -0.12306133377258788, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03899006242456471, "ret_signed_p1d": -0.03899006242456471, "alpha_spy_p1d": -0.03287037244704716, "alpha_c_p1d": -0.03058992201833277, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.030680714528612762, "ret_signed_p1w": 0.030680714528612762, "alpha_spy_p1w": 0.046684836328794166, "alpha_c_p1w": 0.02145092017379413, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.03004965592608655, "ret_signed_p1m": -0.03004965592608655, "alpha_spy_p1m": 0.003159826186484027, "alpha_c_p1m": 0.01364506401438148, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.3522505413743511, "ret_signed_p3m": -0.3522505413743511, "alpha_spy_p3m": -0.2871160405263251, "alpha_c_p3m": -0.17778288766640649, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.625885770233126, "ret_signed_p6m": -0.625885770233126, "alpha_spy_p6m": -0.5149557027854867, "alpha_c_p6m": -0.31172904839059146, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.1012, -0.1012, -0.1062, -0.0746, -0.0574, -0.059, -0.066, -0.0652, -0.0692, -0.0692, -0.027, -0.0304, -0.0388, -0.0337, 0.0075, 0.0016, 0.0657, 0.0925, 0.0972, 0.0901, 0.0681, 0.0585, 0.0547, 0.0622, 0.0495, 0.0268, 0.0451, 0.0644, 0.0559, 0.0634, 0.0449, 0.0326, 0.0436, 0.0425, 0.0531, 0.0668, 0.0868, 0.0938, 0.1029, 0.0812, 0.062, 0.0708, 0.0642, 0.0626, 0.0436, 0.0393, 0.0686, 0.1048, 0.1098, 0.0825, 0.0906, 0.0828, 0.0678, 0.0828, 0.082, 0.0537, 0.0732, 0.0849, 0.0863, 0.1026, 0.0928, 0.0983, 0.1283, 0.0, 0.039, 0.0123, -0.0027, -0.0374, -0.0307, -0.0158, -0.0334, 0.0067, -0.0018, 0.0054, -0.0177, -0.0467, -0.0348, -0.0433, -0.0501, -0.0215, -0.0116, 0.0046, 0.0156, 0.0264, 0.03, 0.0201, 0.0233, 0.0291, 0.0225, 0.0166, 0.0353, 0.0361, 0.0424, 0.0578, 0.0483, 0.02, 0.0192, -0.0111, 0.0019, 0.0376, 0.0578, 0.084, 0.0946, 0.0898, 0.0991, 0.1052, 0.1686, 0.1742, 0.1808, 0.2721, 0.2717, 0.2991, 0.3135, 0.2946, 0.3005, 0.3023, 0.3226, 0.3268, 0.3481, 0.4061, 0.4101, 0.4376, 0.4016, 0.3647, 0.3615, 0.304, 0.3523, 0.3564, 0.3987, 0.4058, 0.403, 0.4404, 0.4256, 0.3941, 0.4223, 0.433, 0.455, 0.4564, 0.4757, 0.5047, 0.4736, 0.4871, 0.4734, 0.4551, 0.437, 0.3977, 0.4232, 0.4235, 0.4314, 0.4148, 0.4029, 0.4066, 0.3893, 0.423, 0.4281, 0.3386, 0.3784, 0.3808, 0.4594, 0.44, 0.45, 0.4874, 0.5073, 0.5463, 0.5365, 0.5273, 0.546, 0.5294, 0.5183, 0.5102, 0.4851, 0.4943, 0.4583, 0.4028, 0.4328, 0.4471, 0.4381, 0.4497, 0.4429, 0.4429, 0.4429, 0.4554, 0.4741, 0.4789, 0.4789, 0.583, 0.6904, 0.7029, 0.6882, 0.6259]}
{"unit_id": "orig:1970272780785754612", "ticker_idx": 0, "ts": "2025-09-22T23:43:31+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "12M target price raised from KRW 84,000 → KRW 96,000; Buy rating maintained", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman Sachs reiterates Buy on Samsung, raises 12M PT to KRW 96,000 on memory upcycle, HBM ramp, and EPS upgrades of 8–24%.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electronics: Profit Outlook Improvement in 2026, Target Price Raised to 96,000 KRW (Goldman Sachs)\n\nEarnings Outlook & Key Revisions\n    • 3Q25 operating profit forecast raised from KRW 8.9tn → KRW 10.2tn\n(reflecting strong traditional memory shipment & pricing, reduced losses in Foundry/System LSI, and higher KRW/USD FX assumptions).\n    • 2025–2027 EPS forecasts lifted by 8–24% (factoring in stronger DRAM/NAND pricing assumptions and HBM shipment growth).\n    • 12M target price raised from KRW 84,000 → KRW 96,000; Buy rating maintained.\n\n⸻\n\n2. Business Unit Highlights\n\n1) Memory\n    • DRAM ASP expected to rebound from 2Q25 and rise for six consecutive quarters through 3Q26, driven by server demand.\n    • NAND ASP projected to increase for four consecutive quarters from 3Q25 to 2Q26.\n    • Memory upcycle supported by hyperscaler AI CapEx expansion, server replacement demand, synchronized procurement patterns among customers, and constrained supply capacity.\n    • Rising memory content per device (iPhone 17 adopting 12GB DRAM, higher base NAND capacity) further supports demand.\n    • DRAM ASP outlook revised upward vs prior forecasts: ’25E +2%, ’26E +20%, ’27E +11%.\n    • NAND ASP outlook revised upward: ’25E +2%, ’26E +18%, ’27E +7%.\n\n2) HBM\n    • 3Q25 shipments expected to grow over 80%, generating revenue of about USD 2.3bn (doubling YoY).\n    • Full-scale ramp of HBM3E 12-high shipments, with higher chances of Nvidia qualification → positive spillover to HBM4 qualification prospects.\n    • HBM revenue forecasts raised by 7% to USD 14bn in ’26E and USD 19bn in ’27E.\n\n3) Foundry / System LSI\n    • 3Q25 operating loss forecast narrowed from -KRW 1.7tn → -KRW 1.1tn.\n    • Utilization improvement driven by demand for Nintendo Switch 2 SoC (8nm) and Galaxy Z Flip7 (Exynos 2500 on 3nm).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "orig:1944939778585321827", "ticker_idx": 0, "ts": "2025-07-15T01:59:13+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This could have a positive impact on Samsung Electronics' Q3 earnings", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM's Wagner sees Nvidia H20 write-back of $10B+ positively impacting Samsung Electronics Q3 earnings as primary HBM3 supplier.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "JPM: Samsung Electronics (SEC) and Nvidia's H20 Model\n\nFor reference, Samsung Electronics (SEC) is the primary supplier of HBM3 chips used in Nvidia's H20 model.\n\nWagner, JPM's Asia Tech Special Sales Lead, believes that Nvidia could record over $10 billion in cost write-backs related to its H20 chips. This could have a positive impact on Samsung Electronics' Q3 earnings.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01883834739473611, "ret_signed_m1d": -0.01883834739473611, "alpha_spy_m1d": -0.023129852764707803, "alpha_c_m1d": -0.010036741140800665, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.008801606253935446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015698581479784757, "ret_signed_p1d": 0.015698581479784757, "alpha_spy_p1d": 0.012355390361269647, "alpha_c_p1d": 0.012596800581375467, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00310178089840929, "ret_p1w": 0.03610674980825368, "ret_signed_p1w": 0.03610674980825368, "alpha_spy_p1w": 0.0253053663874383, "alpha_c_p1w": 0.0326171868946612, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.003489562913592481, "ret_p1m": 0.12872829370574368, "ret_signed_p1m": 0.12872829370574368, "alpha_spy_p1m": 0.09216095978267802, "alpha_c_p1m": 0.08840431038607299, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.040323983319670686, "ret_p3m": 0.48855841811668865, "ret_signed_p3m": 0.48855841811668865, "alpha_spy_p3m": 0.4360132999761246, "alpha_c_p3m": 0.40776360755248, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.08079481056420867, "ret_p6m": 1.2341847828716226, "ret_signed_p6m": 1.2341847828716226, "alpha_spy_p6m": 1.1194274795904355, "alpha_c_p6m": 1.0947109736023968, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.1394738092692258, "price_path": [-0.1403, -0.1372, -0.1356, -0.1418, -0.1309, -0.1309, -0.1309, -0.1294, -0.1294, -0.134, -0.134, -0.1528, -0.1528, -0.1528, -0.1481, -0.1481, -0.145, -0.1013, -0.1122, -0.1044, -0.106, -0.1138, -0.1294, -0.1278, -0.1309, -0.1465, -0.1543, -0.1465, -0.159, -0.1278, -0.1247, -0.1231, -0.1138, -0.1138, -0.0982, -0.0779, -0.0779, -0.0669, -0.0763, -0.0654, -0.0716, -0.0904, -0.1075, -0.0935, -0.0669, -0.0763, -0.0716, -0.095, -0.056, -0.0435, -0.0607, -0.0455, -0.0612, -0.0549, -0.0455, 0.0016, -0.0063, -0.0314, -0.0361, -0.0518, -0.0424, -0.0173, -0.0188, 0.0, 0.0157, 0.0471, 0.0534, 0.0644, 0.0361, 0.0424, 0.0361, 0.0345, 0.1052, 0.1083, 0.1397, 0.1209, 0.0816, 0.0942, 0.0973, 0.0801, 0.1068, 0.1272, 0.1146, 0.1162, 0.1287, 0.124, 0.124, 0.0989, 0.0989, 0.1068, 0.1083, 0.1209, 0.1224, 0.1036, 0.1083, 0.0926, 0.0942, 0.0612, 0.0848, 0.0958, 0.1005, 0.0911, 0.1005, 0.1224, 0.1397, 0.1523, 0.1837, 0.2009, 0.2465, 0.2276, 0.2606, 0.2606, 0.3108, 0.3297, 0.3407, 0.3516, 0.3077, 0.3277, 0.323, 0.3561, 0.4152, 0.4152, 0.4152, 0.4152, 0.4152, 0.4152, 0.4886, 0.4712, 0.4444, 0.498, 0.5406, 0.5437, 0.5469, 0.5374, 0.5548, 0.5217, 0.5579, 0.6084, 0.569, 0.5847, 0.6415, 0.6951, 0.7519, 0.6541, 0.5863, 0.5642, 0.5437, 0.5863, 0.6321, 0.6257, 0.621, 0.5327, 0.5863, 0.5422, 0.5217, 0.5863, 0.4949, 0.5248, 0.5658, 0.621, 0.6321, 0.5847, 0.5895, 0.6305, 0.6478, 0.6573, 0.7093, 0.7267, 0.7093, 0.703, 0.692, 0.7172, 0.6526, 0.621, 0.7014, 0.6967, 0.6762, 0.7424, 0.7582, 0.7519, 0.7519, 0.8449, 0.8935, 0.8998, 0.8998, 0.8998, 1.0361, 1.1882, 1.2009, 1.2342]}
{"unit_id": "orig:1973256688372425182", "ticker_idx": 0, "ts": "2025-10-01T05:20:30+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BESI of the Netherlands already has development and application experience in logic semiconductors and is considered the strongest candidate supplier", "tweet_type": "analysis", "timeline": "years", "summary": "BESI seen as strongest hybrid bonder supplier candidate; Hanwha Vision a latecomer with upside, as hybrid bonding reshapes HBM packaging equipment market by 2027–2028.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Hybrid Bonder – Who Will Do Better?\n\nShinYoung Securities\n\nTransition from TC (Thermal Compression) Bonder to Hybrid Bonder from 2027\n\nHybrid bonding is a technology that connects chips without solder bumps. In contrast to current HBM that uses TC bonding, where the micro-bump pitch remains around 40–55㎛, hybrid bonding can be reduced to below 10㎛, improving density by about 4–5x on the same area.\n\nMoreover, TC bonding passes through bumps and TSVs, resulting in relatively high parasitic resistance and higher power consumption per transmission. Hybrid bonding minimizes resistance paths, enabling ~30–40% power savings at the same performance level. In terms of density, bandwidth, and power efficiency, hybrid bonding is expected to outperform TC bonding.\n\nHybrid bonding has already been applied in some logic semiconductors and NAND, and is expected to be adopted in HBM starting in 2027.\n\nMarket Size Expected to Reach Trillions of KRW by 2028\n\nHybrid bonding is projected not only to provide technological superiority but also explosive growth opportunities in terms of market size. Currently, DRAM makers’ combined HBM production capacity is about 400K/M, and is expected to expand to 500K/M next year with additional capacity. Assuming conservative annual expansions of 50K/M during 2027–2028, HBM capacity could reach 600K/M by 2028.\n\nIf HBF adds another 100K/M capacity by 2028, total TSV capacity is estimated at 700K/M. Assuming hybrid bonding is applied to HBM4E and HBF from the second half of 2027 and penetration reaches 50% in 2028, about 350K/M would be hybrid-bonded.\n\nFor TC bonders, about 1.5–2 units are needed per 1K/M, whereas hybrid bonders, which require greater precision, will likely require 2 units. Thus, applying hybrid bonding to 350K/M TSV capacity would require roughly 700 new bonders.\n\nWhile TC bonders are currently priced at just over KRW 3 billion per unit, hybrid bonders are estimated to be around KRW 4 billion due to their technical complexity. Based on replacement demand alone, this implies a market of about KRW 2.8 trillion.\n\nWe believe that, rather than focusing on short-term TC bonder unit growth, it is more critical in the mid-to-long term which companies secure hybrid bonding capabilities and their positioning in the ecosystem.\n\nWho Will Do Better?\n\nCurrently, SK hynix (000660.KS) is known to be co-developing and evaluating hybrid bonders with two suppliers. BESI of the Netherlands already has development and application experience in logic semiconductors and is considered the strongest candidate supplier.\n\nAt the same time, Hanwha Vision (489790.KS) is said to have achieved positive results in recent qualification tests, suggesting that despite being a latecomer, it has the potential to secure meaningful opportunities.\n\nUltimately, hybrid bonding is expected to not only drive technological evolution but also reshape the bonder equipment market into a multi-trillion won scale, establishing itself as a key element in transforming the next-generation memory packaging supply chain landscape.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007433531264066762, "ret_signed_m1d": -0.007433531264066762, "alpha_spy_m1d": -0.00403753393480355, "alpha_c_m1d": 0.014533021266492585, "bench_spy_m1d": -0.0033959973292632117, "bench_c_m1d": -0.021966552530559347, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04499221106929885, "ret_signed_p1d": 0.04499221106929885, "alpha_spy_p1d": 0.0438404994662589, "alpha_c_p1d": 0.0313567931832619, "bench_spy_p1d": 0.001151711603039951, "bench_c_p1d": 0.013635417886036949, "ret_p1w": 0.14514865050039782, "ret_signed_p1w": 0.14514865050039782, "alpha_spy_p1w": 0.1381774068346051, "alpha_c_p1w": 0.10804829965204643, "bench_spy_p1w": 0.006971243665792715, "bench_c_p1w": 0.03710035084835139, "ret_p1m": 0.15062598300294372, "ret_signed_p1m": 0.15062598300294372, "alpha_spy_p1m": 0.1336016432467353, "alpha_c_p1m": 0.06060224450294238, "bench_spy_p1m": 0.01702433975620843, "bench_c_p1m": 0.09002373850000134, "ret_p3m": 0.03364634235647013, "ret_signed_p3m": 0.03364634235647013, "alpha_spy_p3m": 0.0015834516668673526, "alpha_c_p3m": -0.06121197340264262, "bench_spy_p3m": 0.032062890689602774, "bench_c_p3m": 0.09485831575911274, "ret_p6m": 0.45070421773706326, "ret_signed_p6m": 0.45070421773706326, "alpha_spy_p6m": 0.4801557846461171, "alpha_c_p6m": 0.3058541401805326, "bench_spy_p6m": -0.029451566909053817, "bench_c_p6m": 0.14485007755653068, "price_path": [-0.0528, -0.0501, -0.0438, -0.0458, -0.0031, -0.0051, -0.0246, -0.0117, -0.0614, -0.0023, 0.0012, 0.0184, -0.0156, -0.0325, -0.0063, -0.0837, -0.0391, -0.061, -0.0423, -0.0681, -0.0833, -0.0736, -0.0978, -0.1017, -0.0614, -0.0552, -0.0489, -0.0317, -0.0336, -0.0403, -0.0724, -0.0782, -0.0657, -0.0837, -0.0966, -0.0689, -0.0677, -0.0556, -0.0626, -0.063, -0.0994, -0.0962, -0.1635, -0.1721, -0.1616, -0.1495, -0.1326, -0.1279, -0.1209, -0.1107, -0.1248, -0.0755, -0.0876, -0.0951, -0.0235, -0.0493, -0.0391, -0.007, 0.0059, -0.0086, -0.0423, -0.011, -0.0074, 0.0, 0.045, 0.0211, 0.1483, 0.1487, 0.1451, 0.1377, 0.0978, 0.117, 0.0994, 0.1385, 0.1354, 0.1088, 0.1362, 0.1291, 0.0837, 0.1287, 0.1506, 0.1581, 0.1471, 0.1553, 0.1506, 0.1549, 0.1408, 0.1154, 0.1072, 0.0779, 0.0497, 0.0677, 0.0782, 0.0591, 0.0469, 0.0282, 0.0188, -0.0051, 0.0102, 0.0176, -0.0391, -0.0266, -0.025, 0.0141, 0.0164, 0.0164, 0.0153, 0.0278, 0.0638, 0.0759, 0.0919, 0.1491, 0.1213, 0.099, 0.0712, 0.0336, 0.0364, 0.0454, 0.0117, 0.0286, 0.0231, 0.0266, 0.0321, 0.0293, 0.0293, 0.0293, 0.0336, 0.0415, 0.0466, 0.0466, 0.1667, 0.2031, 0.259, 0.2426, 0.1823, 0.1823, 0.2688, 0.3083, 0.2637, 0.3588, 0.3529, 0.3208, 0.3529, 0.3619, 0.3693, 0.3697, 0.3682, 0.3752, 0.3259, 0.2719, 0.2872, 0.2879, 0.2531, 0.2328, 0.2578, 0.295, 0.3099, 0.3228, 0.3372, 0.32, 0.3787, 0.3967, 0.4092, 0.4632, 0.3537, 0.4437, 0.4683, 0.5047, 0.5462, 0.4781, 0.482, 0.475, 0.419, 0.4945, 0.4761, 0.223, 0.3013, 0.3662, 0.3689, 0.3713, 0.4484, 0.4319, 0.4746, 0.5055, 0.4386, 0.3967, 0.4284, 0.4405, 0.4515, 0.4507]}
{"unit_id": "orig:1973256688372425182", "ticker_idx": 1, "ts": "2025-10-01T05:20:30+00:00", "symbol": "Hanwha Vision", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hanwha Vision (489790.KS) is said to have achieved positive results in recent qualification tests, suggesting that despite being a latecomer, it has the potential to secure meaningful opportunities", "tweet_type": "analysis", "timeline": "years", "summary": "BESI seen as strongest hybrid bonder supplier candidate; Hanwha Vision a latecomer with upside, as hybrid bonding reshapes HBM packaging equipment market by 2027–2028.", "resolved_tickers": ["489790.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Hanwha Vision 489790.KS Korea as cited in context", "bench_c": "XLI", "bench_c_reason": "industry='Security & Protection Services'", "bench_c_industry": "Security & Protection Services", "unit_kind": "original", "lang": "en", "tweet_text": "Hybrid Bonder – Who Will Do Better?\n\nShinYoung Securities\n\nTransition from TC (Thermal Compression) Bonder to Hybrid Bonder from 2027\n\nHybrid bonding is a technology that connects chips without solder bumps. In contrast to current HBM that uses TC bonding, where the micro-bump pitch remains around 40–55㎛, hybrid bonding can be reduced to below 10㎛, improving density by about 4–5x on the same area.\n\nMoreover, TC bonding passes through bumps and TSVs, resulting in relatively high parasitic resistance and higher power consumption per transmission. Hybrid bonding minimizes resistance paths, enabling ~30–40% power savings at the same performance level. In terms of density, bandwidth, and power efficiency, hybrid bonding is expected to outperform TC bonding.\n\nHybrid bonding has already been applied in some logic semiconductors and NAND, and is expected to be adopted in HBM starting in 2027.\n\nMarket Size Expected to Reach Trillions of KRW by 2028\n\nHybrid bonding is projected not only to provide technological superiority but also explosive growth opportunities in terms of market size. Currently, DRAM makers’ combined HBM production capacity is about 400K/M, and is expected to expand to 500K/M next year with additional capacity. Assuming conservative annual expansions of 50K/M during 2027–2028, HBM capacity could reach 600K/M by 2028.\n\nIf HBF adds another 100K/M capacity by 2028, total TSV capacity is estimated at 700K/M. Assuming hybrid bonding is applied to HBM4E and HBF from the second half of 2027 and penetration reaches 50% in 2028, about 350K/M would be hybrid-bonded.\n\nFor TC bonders, about 1.5–2 units are needed per 1K/M, whereas hybrid bonders, which require greater precision, will likely require 2 units. Thus, applying hybrid bonding to 350K/M TSV capacity would require roughly 700 new bonders.\n\nWhile TC bonders are currently priced at just over KRW 3 billion per unit, hybrid bonders are estimated to be around KRW 4 billion due to their technical complexity. Based on replacement demand alone, this implies a market of about KRW 2.8 trillion.\n\nWe believe that, rather than focusing on short-term TC bonder unit growth, it is more critical in the mid-to-long term which companies secure hybrid bonding capabilities and their positioning in the ecosystem.\n\nWho Will Do Better?\n\nCurrently, SK hynix (000660.KS) is known to be co-developing and evaluating hybrid bonders with two suppliers. BESI of the Netherlands already has development and application experience in logic semiconductors and is considered the strongest candidate supplier.\n\nAt the same time, Hanwha Vision (489790.KS) is said to have achieved positive results in recent qualification tests, suggesting that despite being a latecomer, it has the potential to secure meaningful opportunities.\n\nUltimately, hybrid bonding is expected to not only drive technological evolution but also reshape the bonder equipment market into a multi-trillion won scale, establishing itself as a key element in transforming the next-generation memory packaging supply chain landscape.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015296367112810794, "ret_signed_m1d": 0.015296367112810794, "alpha_spy_m1d": 0.018692364442074005, "alpha_c_m1d": 0.012500688175831609, "bench_spy_m1d": -0.0033959973292632117, "bench_c_m1d": 0.0027956789369791846, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0248565965583174, "ret_signed_p1d": 0.0248565965583174, "alpha_spy_p1d": 0.02370488495527745, "alpha_c_p1d": 0.022255988205642563, "bench_spy_p1d": 0.001151711603039951, "bench_c_p1d": 0.002600608352674838, "ret_p1w": 0.0248565965583174, "ret_signed_p1w": 0.0248565965583174, "alpha_spy_p1w": 0.017885352892524686, "alpha_c_p1w": 0.013218250233755668, "bench_spy_p1w": 0.006971243665792715, "bench_c_p1w": 0.011638346324561732, "ret_p1m": -0.0248565965583174, "ret_signed_p1m": -0.0248565965583174, "alpha_spy_p1m": -0.04188093631452583, "alpha_c_p1m": -0.030643324510823966, "bench_spy_p1m": 0.01702433975620843, "bench_c_p1m": 0.005786727952506565, "ret_p3m": -0.18355640535372852, "ret_signed_p3m": -0.18355640535372852, "alpha_spy_p3m": -0.2156192960433313, "alpha_c_p3m": -0.2070106980895855, "bench_spy_p3m": 0.032062890689602774, "bench_c_p3m": 0.023454292735856974, "ret_p6m": 0.592734225621415, "ret_signed_p6m": 0.592734225621415, "alpha_spy_p6m": 0.6221857925304688, "alpha_c_p6m": 0.5376169022454, "bench_spy_p6m": -0.029451566909053817, "bench_c_p6m": 0.05511732337601494, "price_path": [-0.0402, -0.0287, 0.0268, 0.0153, 0.0478, 0.0612, 0.0229, 0.0325, 0.065, 0.0268, 0.0229, 0.0574, 0.065, 0.0459, 0.1415, 0.1434, 0.1625, 0.325, 0.2906, 0.1816, 0.1319, 0.1128, 0.1205, 0.0899, 0.0746, 0.0841, 0.0841, 0.0574, 0.1052, 0.1281, 0.1281, 0.0707, 0.0574, 0.0153, 0.0363, 0.0344, 0.0574, 0.0344, 0.0306, 0.0287, 0.0191, -0.0249, -0.0153, 0.065, 0.0765, 0.1033, 0.1033, 0.1166, 0.1377, 0.1205, 0.1453, 0.1663, 0.1721, 0.1759, 0.1759, 0.1759, 0.13, 0.1128, 0.1033, 0.0593, 0.0134, 0.0191, 0.0153, 0.0, 0.0249, 0.0249, 0.0249, 0.0249, 0.0249, 0.0249, 0.0688, 0.0497, -0.0191, 0.0115, 0.0019, -0.0229, 0.021, 0.0096, -0.0076, -0.0153, -0.0057, 0.0325, 0.0229, -0.0076, -0.0249, -0.0363, -0.0507, -0.0507, -0.1109, -0.1224, -0.1491, -0.1396, -0.0507, -0.0631, -0.0822, -0.1224, -0.0899, -0.1243, -0.0851, -0.0755, -0.1252, -0.1272, -0.1511, -0.152, -0.1597, -0.1711, -0.1558, -0.1692, -0.1673, -0.2151, -0.2055, -0.1912, -0.2046, -0.2256, -0.2667, -0.2667, -0.2811, -0.3011, -0.2916, -0.3126, -0.2897, -0.2505, -0.2457, -0.1683, -0.1683, -0.1826, -0.1836, -0.1415, -0.1415, -0.1415, -0.1195, -0.1071, -0.0516, -0.0717, -0.0841, -0.0631, -0.065, -0.0478, -0.0019, 0.0421, 0.0421, 0.0382, 0.0096, -0.0229, 0.0402, 0.0344, -0.0057, 0.0344, 0.065, 0.0593, 0.1816, 0.088, 0.2218, 0.2256, 0.1874, 0.2772, 0.2964, 0.1319, 0.13, 0.174, 0.1721, 0.1721, 0.1721, 0.1721, 0.1702, 0.1663, 0.1836, 0.5373, 0.4493, 0.6444, 0.6329, 0.6329, 0.4493, 0.262, 0.4971, 0.6405, 0.5201, 0.5201, 0.4895, 0.5086, 0.4665, 0.5029, 0.4704, 0.6042, 0.6673, 0.6711, 0.5335, 0.587, 0.7725, 0.5927]}
{"unit_id": "orig:1972892551846981925", "ticker_idx": 0, "ts": "2025-09-30T05:13:33+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS believes MRVL remains one of the core recommended stocks for the rest of 2025", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS reiterates MRVL as core pick; LRCX and AMAT benefit from silicon-inside cooling; ASMI lowers 2H25 guidance with Intel as key drag.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Latest Developments in the U.S. Semiconductor Industry: Tariffs, MRVL Growth Vectors, ASMI Outlook, and New Applications of Silicon-Inside Cooling\n\nUBS (T. Arcuri, 250929)\n\nThe market is closely watching the Trump administration’s proposed 1:1 chip tariff rule. According to the plan, tariffs will be calculated based on the value of the chip, and credits or transitional exemptions may be granted to newly built fabs that commit to investing in the U.S. UBS pointed out that the expansion plans of TSMC and Intel by 2030 may be sufficient to meet domestic U.S. advanced logic demand, but the localization levels of DRAM and NAND are significantly insufficient. Therefore, the policy is expected to include a longer transitional period. Similar tariff cases in the pharmaceutical industry also show that 100% tariffs are often accompanied by clear exemption arrangements.\n\nAt a recent event, MRVL reiterated its diversified growth path in the data center business. The company expects global DC capital expenditure to expand at an 18% compound annual growth rate and to outperform over the long term, mainly driven by optical modules and XPU attach businesses. Management confirmed that Custom ASIC business will follow DC capex growth in 2026, and Microsoft’s XPU in 2027 is expected to become the next major catalyst. XPU attach projects cover NICs, memory expansion, etc., with individual project lifecycle revenues of over $200 million, and are expected to contribute about $1.9 billion in annual incremental revenue by 2028. On optical modules, after the acquisition of IPHI, the DSP business has scaled up 5x and is expected to continue to outpace DC capex growth. Networking products such as switches and Retimer/AECs also show strong growth momentum. In addition, the company announced a new $5 billion share buyback program. UBS believes MRVL remains one of the core recommended stocks for the rest of 2025.\n\nASM International, at its Investor Day, lowered its 2H25 forecast, expecting a sequential decline of 5–10%, mainly due to differentiated customer demand, with Intel likely being the biggest drag. UBS believes the impact on U.S. equipment makers is limited, as LAM, KLA, and AMAT have largely digested the risks from Intel’s WFE cuts. ASMI also pointed out that declining power/analog demand has a negative impact on LAM and AMAT. On the other hand, the penetration of atomic layer deposition (ALD) at advanced nodes continues to rise, with the number of GAA layers increasing by more than 20% from 2nm to 1.4nm, and the share of front-end ALD is expected to rise from 50% to 60%. This will intensify competition with LAM’s single-wafer ALD business.\n\nHowever, UBS believes that LAM can offset this through new technology opportunities such as dry resist, backside power delivery, and moly ALD. In advanced packaging, ASMI plans to increase its market share from 15% to over 30%, with part of the share coming from AMAT’s handover. Microsoft recently showcased a new breakthrough in silicon-inside cooling based on microfluidic technology, achieving 3x heat dissipation efficiency and a 65% reduction in peak temperature, which is expected to significantly improve AI chip energy efficiency. The related technology relies on Bosch DRIE etching to build micro-fin heat sinks on the backside of the chip, and TSMC has also proposed a similar IMC-Si solution incorporated into the CoWoS-R process. UBS believes LRCX (with a leading share in the TSV etching market) and parts of AMAT will be the main beneficiaries. Overall, this development further reinforces the leadership of deposition/etching processes in future WFE growth.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019983255910710818, "ret_signed_m1d": -0.019983255910710818, "alpha_spy_m1d": -0.01623052518821344, "alpha_c_m1d": -0.008646129746735953, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0021410176725789043, "ret_signed_p1d": -0.0021410176725789043, "alpha_spy_p1d": -0.0055485870985119146, "alpha_c_p1d": -0.02460093720538792, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.03449513333990395, "ret_signed_p1w": 0.03449513333990395, "alpha_spy_p1w": 0.030081933804332994, "alpha_c_p1w": 0.0017398294474233467, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.07303072380362874, "ret_signed_p1m": 0.07303072380362874, "alpha_spy_p1m": 0.04119249664640057, "alpha_c_p1m": -0.05642759766010763, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.02768130004542435, "ret_signed_p3m": 0.02768130004542435, "alpha_spy_p3m": -0.011602036007613004, "alpha_c_p3m": -0.09683892513814452, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 0.17266651076673312, "ret_signed_p6m": 0.17266651076673312, "alpha_spy_p6m": 0.18110278358863496, "alpha_c_p6m": -0.05377547760001433, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.1065, -0.1065, -0.1496, -0.1449, -0.1412, -0.1281, -0.1351, -0.1375, -0.1387, -0.1572, -0.1435, -0.112, -0.131, -0.1437, -0.1285, -0.1193, -0.1173, -0.0971, -0.0919, -0.0277, -0.044, -0.1144, -0.0897, -0.0885, -0.1041, -0.0978, -0.0801, -0.0808, -0.0745, -0.0565, -0.0598, -0.0937, -0.0872, -0.1427, -0.1528, -0.153, -0.1317, -0.1323, -0.1167, -0.1104, -0.0814, -0.2522, -0.2522, -0.2316, -0.2588, -0.2375, -0.2467, -0.2149, -0.2049, -0.2019, -0.2079, -0.1989, -0.1979, -0.1809, -0.1557, -0.117, -0.1167, -0.1016, -0.1124, -0.0473, -0.0031, -0.0107, -0.02, 0.0, -0.0021, 0.0253, 0.0256, 0.0577, 0.0345, 0.1003, 0.0786, 0.019, 0.064, 0.0263, 0.058, 0.0502, 0.0468, 0.0217, 0.0029, -0.0353, -0.0148, 0.0014, 0.0559, 0.0529, 0.073, 0.0542, 0.1158, 0.0756, 0.0426, 0.1058, 0.1109, 0.0822, 0.1097, 0.0633, 0.0633, 0.0417, 0.029, -0.0067, -0.0635, -0.0321, -0.0873, -0.0781, -0.0027, -0.007, 0.0441, 0.0441, 0.0641, 0.0843, 0.1056, 0.1927, 0.1687, 0.1773, 0.0951, 0.0582, 0.1006, 0.0645, 0.0049, 0.0029, 0.0007, -0.0275, 0.0054, 0.0009, 0.0094, 0.0436, 0.0295, 0.0295, 0.0277, 0.0208, 0.0327, 0.0115, 0.0115, 0.064, 0.074, 0.0502, 0.0074, -0.0067, -0.0087, -0.0127, -0.0108, -0.0327, -0.0426, -0.0416, -0.0416, -0.0495, -0.0167, -0.0102, -0.0444, -0.026, -0.0122, -0.004, -0.0311, -0.06, -0.0631, -0.1002, -0.1218, -0.1161, -0.0438, -0.0191, -0.0232, -0.0311, -0.0682, -0.0637, -0.0637, -0.0589, -0.0579, -0.0517, -0.0533, -0.0734, -0.0663, -0.0361, -0.0556, -0.027, -0.0369, -0.0768, -0.0698, -0.0986, 0.0669, 0.1036, 0.1113, 0.0773, 0.0443, 0.0465, 0.0908, 0.0814, 0.0437, 0.0664, 0.0471, 0.0739, 0.1001, 0.1727]}
{"unit_id": "orig:1972892551846981925", "ticker_idx": 1, "ts": "2025-09-30T05:13:33+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS believes LRCX (with a leading share in the TSV etching market) and parts of AMAT will be the main beneficiaries", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS reiterates MRVL as core pick; LRCX and AMAT benefit from silicon-inside cooling; ASMI lowers 2H25 guidance with Intel as key drag.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Latest Developments in the U.S. Semiconductor Industry: Tariffs, MRVL Growth Vectors, ASMI Outlook, and New Applications of Silicon-Inside Cooling\n\nUBS (T. Arcuri, 250929)\n\nThe market is closely watching the Trump administration’s proposed 1:1 chip tariff rule. According to the plan, tariffs will be calculated based on the value of the chip, and credits or transitional exemptions may be granted to newly built fabs that commit to investing in the U.S. UBS pointed out that the expansion plans of TSMC and Intel by 2030 may be sufficient to meet domestic U.S. advanced logic demand, but the localization levels of DRAM and NAND are significantly insufficient. Therefore, the policy is expected to include a longer transitional period. Similar tariff cases in the pharmaceutical industry also show that 100% tariffs are often accompanied by clear exemption arrangements.\n\nAt a recent event, MRVL reiterated its diversified growth path in the data center business. The company expects global DC capital expenditure to expand at an 18% compound annual growth rate and to outperform over the long term, mainly driven by optical modules and XPU attach businesses. Management confirmed that Custom ASIC business will follow DC capex growth in 2026, and Microsoft’s XPU in 2027 is expected to become the next major catalyst. XPU attach projects cover NICs, memory expansion, etc., with individual project lifecycle revenues of over $200 million, and are expected to contribute about $1.9 billion in annual incremental revenue by 2028. On optical modules, after the acquisition of IPHI, the DSP business has scaled up 5x and is expected to continue to outpace DC capex growth. Networking products such as switches and Retimer/AECs also show strong growth momentum. In addition, the company announced a new $5 billion share buyback program. UBS believes MRVL remains one of the core recommended stocks for the rest of 2025.\n\nASM International, at its Investor Day, lowered its 2H25 forecast, expecting a sequential decline of 5–10%, mainly due to differentiated customer demand, with Intel likely being the biggest drag. UBS believes the impact on U.S. equipment makers is limited, as LAM, KLA, and AMAT have largely digested the risks from Intel’s WFE cuts. ASMI also pointed out that declining power/analog demand has a negative impact on LAM and AMAT. On the other hand, the penetration of atomic layer deposition (ALD) at advanced nodes continues to rise, with the number of GAA layers increasing by more than 20% from 2nm to 1.4nm, and the share of front-end ALD is expected to rise from 50% to 60%. This will intensify competition with LAM’s single-wafer ALD business.\n\nHowever, UBS believes that LAM can offset this through new technology opportunities such as dry resist, backside power delivery, and moly ALD. In advanced packaging, ASMI plans to increase its market share from 15% to over 30%, with part of the share coming from AMAT’s handover. Microsoft recently showcased a new breakthrough in silicon-inside cooling based on microfluidic technology, achieving 3x heat dissipation efficiency and a 65% reduction in peak temperature, which is expected to significantly improve AI chip energy efficiency. The related technology relies on Bosch DRIE etching to build micro-fin heat sinks on the backside of the chip, and TSMC has also proposed a similar IMC-Si solution incorporated into the CoWoS-R process. UBS believes LRCX (with a leading share in the TSV etching market) and parts of AMAT will be the main beneficiaries. Overall, this development further reinforces the leadership of deposition/etching processes in future WFE growth.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020985757962269647, "ret_signed_m1d": -0.020985757962269647, "alpha_spy_m1d": -0.01723302723977227, "alpha_c_m1d": -0.009648631798294782, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06639278850488828, "ret_signed_p1d": 0.06639278850488828, "alpha_spy_p1d": 0.06298521907895527, "alpha_c_p1d": 0.043932868972079264, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.048170344475994886, "ret_signed_p1w": 0.048170344475994886, "alpha_spy_p1w": 0.04375714494042393, "alpha_c_p1w": 0.015415040583514283, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.19992530595587854, "ret_signed_p1m": 0.19992530595587854, "alpha_spy_p1m": 0.16808707879865037, "alpha_c_p1m": 0.07046698449214217, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.3320625670043631, "ret_signed_p3m": 0.3320625670043631, "alpha_spy_p3m": 0.29277923095132574, "alpha_c_p3m": 0.20754234182079423, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 0.7484279223616208, "ret_signed_p6m": 0.7484279223616208, "alpha_spy_p6m": 0.7568641951835227, "alpha_c_p6m": 0.5219859339948734, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.2635, -0.2635, -0.2685, -0.2559, -0.2561, -0.2467, -0.2418, -0.2575, -0.2467, -0.2519, -0.2488, -0.2497, -0.2417, -0.2719, -0.2763, -0.2712, -0.2773, -0.2649, -0.2625, -0.2614, -0.2931, -0.2817, -0.2665, -0.2794, -0.2849, -0.261, -0.2416, -0.2397, -0.2153, -0.2044, -0.1996, -0.2583, -0.263, -0.2522, -0.261, -0.2665, -0.254, -0.2451, -0.2276, -0.2273, -0.2242, -0.2535, -0.2535, -0.2768, -0.2715, -0.2515, -0.2327, -0.2169, -0.2131, -0.1998, -0.1385, -0.1282, -0.1115, -0.1021, -0.0914, -0.0585, -0.054, -0.0146, -0.0167, -0.0417, -0.0431, -0.0416, -0.021, 0.0, 0.0664, 0.0978, 0.0889, 0.1139, 0.0482, 0.0645, 0.053, -0.0189, 0.0292, 0.0329, 0.0813, 0.0633, 0.0568, 0.0758, 0.0832, 0.0549, 0.1019, 0.1328, 0.1718, 0.1622, 0.1999, 0.2025, 0.176, 0.2042, 0.1634, 0.2326, 0.2113, 0.1901, 0.2425, 0.1888, 0.2055, 0.145, 0.1072, 0.1013, 0.0698, 0.1113, 0.0425, 0.0653, 0.1231, 0.1347, 0.1586, 0.1586, 0.165, 0.156, 0.1814, 0.195, 0.1751, 0.1872, 0.2174, 0.2404, 0.2587, 0.262, 0.2008, 0.2291, 0.2213, 0.1593, 0.232, 0.2887, 0.311, 0.3103, 0.3265, 0.3265, 0.3321, 0.3156, 0.3, 0.2805, 0.2805, 0.3844, 0.4569, 0.5482, 0.5192, 0.5033, 0.6335, 0.6487, 0.6037, 0.5619, 0.6268, 0.6679, 0.6679, 0.6638, 0.7085, 0.651, 0.6303, 0.6672, 0.7838, 0.7922, 0.8564, 0.7464, 0.7766, 0.7213, 0.5693, 0.5957, 0.7281, 0.7151, 0.6952, 0.7588, 0.7302, 0.7619, 0.7619, 0.7623, 0.796, 0.7758, 0.8321, 0.8123, 0.8271, 0.8662, 0.7884, 0.7496, 0.728, 0.6253, 0.6701, 0.6078, 0.4929, 0.5814, 0.612, 0.6392, 0.569, 0.5893, 0.6432, 0.6962, 0.683, 0.7525, 0.7103, 0.7474, 0.7888, 0.7484]}
{"unit_id": "orig:1972892551846981925", "ticker_idx": 2, "ts": "2025-09-30T05:13:33+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "parts of AMAT will be the main beneficiaries", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS reiterates MRVL as core pick; LRCX and AMAT benefit from silicon-inside cooling; ASMI lowers 2H25 guidance with Intel as key drag.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Latest Developments in the U.S. Semiconductor Industry: Tariffs, MRVL Growth Vectors, ASMI Outlook, and New Applications of Silicon-Inside Cooling\n\nUBS (T. Arcuri, 250929)\n\nThe market is closely watching the Trump administration’s proposed 1:1 chip tariff rule. According to the plan, tariffs will be calculated based on the value of the chip, and credits or transitional exemptions may be granted to newly built fabs that commit to investing in the U.S. UBS pointed out that the expansion plans of TSMC and Intel by 2030 may be sufficient to meet domestic U.S. advanced logic demand, but the localization levels of DRAM and NAND are significantly insufficient. Therefore, the policy is expected to include a longer transitional period. Similar tariff cases in the pharmaceutical industry also show that 100% tariffs are often accompanied by clear exemption arrangements.\n\nAt a recent event, MRVL reiterated its diversified growth path in the data center business. The company expects global DC capital expenditure to expand at an 18% compound annual growth rate and to outperform over the long term, mainly driven by optical modules and XPU attach businesses. Management confirmed that Custom ASIC business will follow DC capex growth in 2026, and Microsoft’s XPU in 2027 is expected to become the next major catalyst. XPU attach projects cover NICs, memory expansion, etc., with individual project lifecycle revenues of over $200 million, and are expected to contribute about $1.9 billion in annual incremental revenue by 2028. On optical modules, after the acquisition of IPHI, the DSP business has scaled up 5x and is expected to continue to outpace DC capex growth. Networking products such as switches and Retimer/AECs also show strong growth momentum. In addition, the company announced a new $5 billion share buyback program. UBS believes MRVL remains one of the core recommended stocks for the rest of 2025.\n\nASM International, at its Investor Day, lowered its 2H25 forecast, expecting a sequential decline of 5–10%, mainly due to differentiated customer demand, with Intel likely being the biggest drag. UBS believes the impact on U.S. equipment makers is limited, as LAM, KLA, and AMAT have largely digested the risks from Intel’s WFE cuts. ASMI also pointed out that declining power/analog demand has a negative impact on LAM and AMAT. On the other hand, the penetration of atomic layer deposition (ALD) at advanced nodes continues to rise, with the number of GAA layers increasing by more than 20% from 2nm to 1.4nm, and the share of front-end ALD is expected to rise from 50% to 60%. This will intensify competition with LAM’s single-wafer ALD business.\n\nHowever, UBS believes that LAM can offset this through new technology opportunities such as dry resist, backside power delivery, and moly ALD. In advanced packaging, ASMI plans to increase its market share from 15% to over 30%, with part of the share coming from AMAT’s handover. Microsoft recently showcased a new breakthrough in silicon-inside cooling based on microfluidic technology, achieving 3x heat dissipation efficiency and a 65% reduction in peak temperature, which is expected to significantly improve AI chip energy efficiency. The related technology relies on Bosch DRIE etching to build micro-fin heat sinks on the backside of the chip, and TSMC has also proposed a similar IMC-Si solution incorporated into the CoWoS-R process. UBS believes LRCX (with a leading share in the TSV etching market) and parts of AMAT will be the main beneficiaries. Overall, this development further reinforces the leadership of deposition/etching processes in future WFE growth.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0010256542316788853, "ret_signed_m1d": 0.0010256542316788853, "alpha_spy_m1d": 0.004778384954176262, "alpha_c_m1d": 0.01236278039565375, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06349519096849354, "ret_signed_p1d": 0.06349519096849354, "alpha_spy_p1d": 0.06008762154256053, "alpha_c_p1d": 0.04103527143568453, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.033310518559131364, "ret_signed_p1w": 0.033310518559131364, "alpha_spy_p1w": 0.02889731902356041, "alpha_c_p1w": 0.0005552146666507607, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.151460406310592, "ret_signed_p1m": 0.151460406310592, "alpha_spy_p1m": 0.11962217915336382, "alpha_c_p1m": 0.02200208484685562, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.2816907956048009, "ret_signed_p3m": 0.2816907956048009, "alpha_spy_p3m": 0.24240745955176357, "alpha_c_p3m": 0.15717057042123206, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 0.8097367607247716, "ret_signed_p6m": 0.8097367607247716, "alpha_spy_p6m": 0.8181730335466735, "alpha_c_p6m": 0.5832947723580242, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.0695, -0.0695, -0.0708, -0.0503, -0.0484, -0.0355, -0.036, -0.0401, -0.0294, -0.0512, -0.0624, -0.0725, -0.0619, -0.0886, -0.0892, -0.0838, -0.0956, -0.0733, -0.0824, -0.0776, -0.1231, -0.1234, -0.1096, -0.1275, -0.1324, -0.108, -0.0996, -0.102, -0.0822, -0.0745, -0.0832, -0.2122, -0.2036, -0.2099, -0.2161, -0.2193, -0.2064, -0.2088, -0.1965, -0.1971, -0.1928, -0.2148, -0.2148, -0.2304, -0.2368, -0.2271, -0.2051, -0.2085, -0.2014, -0.2018, -0.1689, -0.1804, -0.1651, -0.1524, -0.13, -0.0732, -0.0715, -0.0206, -0.0189, -0.0161, -0.0251, -0.004, 0.001, 0.0, 0.0635, 0.0921, 0.0625, 0.0936, 0.0333, 0.0624, 0.076, 0.0254, 0.072, 0.0657, 0.1116, 0.1122, 0.0989, 0.1142, 0.1038, 0.0773, 0.1159, 0.1173, 0.1299, 0.1118, 0.1515, 0.1358, 0.1385, 0.161, 0.1243, 0.1766, 0.1406, 0.1237, 0.1482, 0.1169, 0.1269, 0.0903, 0.1039, 0.1171, 0.0995, 0.1484, 0.0778, 0.0963, 0.13, 0.1866, 0.2233, 0.2233, 0.2345, 0.2467, 0.2985, 0.3146, 0.3186, 0.3115, 0.3123, 0.3073, 0.3465, 0.3219, 0.2685, 0.2786, 0.2667, 0.215, 0.2406, 0.2548, 0.2675, 0.2735, 0.2762, 0.2762, 0.2817, 0.2873, 0.2722, 0.2577, 0.2577, 0.3158, 0.3914, 0.4486, 0.43, 0.3783, 0.4739, 0.5036, 0.492, 0.4774, 0.5615, 0.6003, 0.6003, 0.5574, 0.5917, 0.5601, 0.5777, 0.5634, 0.6282, 0.648, 0.6705, 0.5774, 0.6071, 0.5595, 0.4564, 0.4877, 0.5783, 0.6177, 0.6104, 0.6633, 0.6071, 0.7369, 0.7369, 0.7575, 0.8073, 0.8121, 0.8393, 0.8304, 0.8518, 0.9352, 0.841, 0.8242, 0.8237, 0.7214, 0.753, 0.698, 0.5912, 0.6608, 0.6948, 0.7202, 0.6526, 0.6735, 0.6963, 0.727, 0.7124, 0.7503, 0.7496, 0.7727, 0.8325, 0.8097]}
{"unit_id": "quote:1961722335398088905", "ticker_idx": 0, "ts": "2025-08-30T09:27:06+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC itself will become more attractive", "tweet_type": "analysis", "timeline": "unspecified", "summary": "TSMC purging China-exposed suppliers boosts its own attractiveness while creating downside pressure on Taiwan local equipment/materials suppliers with high China revenue exposure.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Summary:\n\n- Full exclusion of Chinese equipment from TSMC’s 2nm line → aimed at avoiding U.S. regulations; Chinese equipment firms (AMEC, Naura, Xinkailai, etc.) are excluded from the list.\n\n- Companies with profit margins above 40% and those highly dependent on China sales are under close scrutiny and may be removed from TSMC’s supply chain.\n\n- High-margin companies are expected to face price reduction requests of up to 15%.\n\n- For Taiwanese local companies that have recently expanded their sales to Chinese customers, China orders are attractive due to high margins and large scale, which puts them in conflict with TSMC’s interests.\n\nInvestor Insights\n- Expanded risks for Taiwanese local equipment and materials suppliers, with potential downside pressure on share prices → a valuation discount factor.\n\n- Companies with high revenue exposure to China → heightened conflict with TSMC, leading to greater long-term risks.\n\n- In contrast, TSMC itself will become more attractive.\n\n- Supply chain cost reductions (strengthened bargaining power) → a factor stabilizing gross margins.\n\n- Minimization of U.S. regulatory risks → continued maintenance of strategic advantages.", "reply_text": null, "parent_text": null, "quoted_text": "TSMC Reportedly “Purges” Local Supply Chain: High-Margin, China-Dependent Firms May Be Eliminated\n\nNews that TSMC’s 2nm production lines will completely phase out Chinese equipment has sparked concern, with Taiwan’s semiconductor industry currently on edge. TSMC has also issued a https://t.co/dj3y16PX9y", "parent_author": null, "ret_m1d": -0.004291840948688619, "ret_signed_m1d": -0.004291840948688619, "alpha_spy_m1d": -0.004291840948688619, "alpha_c_m1d": -0.004291840948688619, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004291840948688619, "ret_signed_p1d": -0.004291840948688619, "alpha_spy_p1d": 0.003118454607185628, "alpha_c_p1d": 0.008832983930088245, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.013124824878776864, "ret_p1w": 0.012875522846065968, "ret_signed_p1w": 0.012875522846065968, "alpha_spy_p1w": 0.007015571391770337, "alpha_c_p1w": -0.008585625894321325, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.021461148740387292, "ret_p1m": 0.12465241107764924, "ret_signed_p1m": 0.12465241107764924, "alpha_spy_p1m": 0.08903199814048723, "alpha_c_p1m": 0.00039751662071241256, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.12425489445693683, "ret_p3m": 0.23668673140344332, "ret_signed_p3m": 0.23668673140344332, "alpha_spy_p3m": 0.18007965760899158, "alpha_c_p3m": 0.03888466686856362, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.1978020645348797, "ret_p6m": 0.6990868732187705, "ret_signed_p6m": 0.6990868732187705, "alpha_spy_p6m": 0.6273989886803302, "alpha_c_p6m": 0.2506586432083209, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.4484282300104496, "price_path": [-0.1538, -0.147, -0.1495, -0.141, -0.1068, -0.0897, -0.103, -0.1159, -0.1202, -0.103, -0.0944, -0.1116, -0.0944, -0.1245, -0.0987, -0.0815, -0.0773, -0.073, -0.0901, -0.0687, -0.0687, -0.0644, -0.0687, -0.073, -0.073, -0.0644, -0.0558, -0.0558, -0.0601, -0.0472, -0.03, -0.03, -0.0086, -0.0129, -0.03, -0.0172, -0.0172, -0.0172, -0.0172, -0.0258, -0.0086, -0.0043, -0.0043, -0.0258, -0.0129, -0.0343, 0.0129, 0.0086, 0.0129, 0.0129, 0.03, 0.0086, 0.0129, 0.0129, 0.0172, -0.0258, -0.0129, -0.0258, 0.0043, 0.0086, 0.0215, -0.0043, -0.0043, 0.0, -0.0043, -0.0043, -0.0043, 0.0129, 0.0129, 0.03, 0.0515, 0.0644, 0.0815, 0.0773, 0.1031, 0.0902, 0.1074, 0.0902, 0.116, 0.1548, 0.1548, 0.1376, 0.1203, 0.1203, 0.1247, 0.1419, 0.1764, 0.2065, 0.2065, 0.2367, 0.2195, 0.241, 0.241, 0.2195, 0.2281, 0.2625, 0.2798, 0.2496, 0.2755, 0.2755, 0.2582, 0.2496, 0.2496, 0.2755, 0.2712, 0.297, 0.297, 0.2927, 0.3013, 0.297, 0.2582, 0.2625, 0.2582, 0.2712, 0.2625, 0.2712, 0.2582, 0.2324, 0.2453, 0.2108, 0.2022, 0.2539, 0.1936, 0.185, 0.2195, 0.241, 0.2367, 0.241, 0.2151, 0.2324, 0.2496, 0.2453, 0.2582, 0.2884, 0.2755, 0.297, 0.2711, 0.2797, 0.2538, 0.2408, 0.2365, 0.2365, 0.2365, 0.2667, 0.2884, 0.2927, 0.2927, 0.3057, 0.323, 0.3143, 0.3402, 0.3402, 0.3705, 0.444, 0.4743, 0.4483, 0.457, 0.4527, 0.4613, 0.4786, 0.4786, 0.4613, 0.5045, 0.5218, 0.5348, 0.5045, 0.5218, 0.5305, 0.5175, 0.5391, 0.5737, 0.5607, 0.5348, 0.5262, 0.5564, 0.5434, 0.5262, 0.5391, 0.5694, 0.6256, 0.6559, 0.6559, 0.6559, 0.6559, 0.6559, 0.6559, 0.6559, 0.6559, 0.6429, 0.6991]}
{"unit_id": "quote:1961722335398088905", "ticker_idx": 1, "ts": "2025-08-30T09:27:06+00:00", "symbol": "Taiwan semiconductors", "kind": "sector", "geography": "Taiwan", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "Expanded risks for Taiwanese local equipment and materials suppliers, with potential downside pressure on share prices → a valuation discount factor", "tweet_type": "analysis", "timeline": "unspecified", "summary": "TSMC purging China-exposed suppliers boosts its own attractiveness while creating downside pressure on Taiwan local equipment/materials suppliers with high China revenue exposure.", "resolved_tickers": ["EWT"], "resolution_method": "etf", "resolution_reason": "Taiwan semiconductor sector proxied by EWT ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "Summary:\n\n- Full exclusion of Chinese equipment from TSMC’s 2nm line → aimed at avoiding U.S. regulations; Chinese equipment firms (AMEC, Naura, Xinkailai, etc.) are excluded from the list.\n\n- Companies with profit margins above 40% and those highly dependent on China sales are under close scrutiny and may be removed from TSMC’s supply chain.\n\n- High-margin companies are expected to face price reduction requests of up to 15%.\n\n- For Taiwanese local companies that have recently expanded their sales to Chinese customers, China orders are attractive due to high margins and large scale, which puts them in conflict with TSMC’s interests.\n\nInvestor Insights\n- Expanded risks for Taiwanese local equipment and materials suppliers, with potential downside pressure on share prices → a valuation discount factor.\n\n- Companies with high revenue exposure to China → heightened conflict with TSMC, leading to greater long-term risks.\n\n- In contrast, TSMC itself will become more attractive.\n\n- Supply chain cost reductions (strengthened bargaining power) → a factor stabilizing gross margins.\n\n- Minimization of U.S. regulatory risks → continued maintenance of strategic advantages.", "reply_text": null, "parent_text": null, "quoted_text": "TSMC Reportedly “Purges” Local Supply Chain: High-Margin, China-Dependent Firms May Be Eliminated\n\nNews that TSMC’s 2nm production lines will completely phase out Chinese equipment has sparked concern, with Taiwan’s semiconductor industry currently on edge. TSMC has also issued a https://t.co/dj3y16PX9y", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013090739779015426, "ret_signed_p1d": 0.013090739779015426, "alpha_spy_p1d": 0.005680444223141179, "alpha_c_p1d": 0.005680444223141179, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.007410295555874247, "ret_p1w": 0.03672223027177268, "ret_signed_p1w": -0.03672223027177268, "alpha_spy_p1w": -0.03086227881747705, "alpha_c_p1w": -0.03086227881747705, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.005859951454295631, "ret_p1m": 0.08160494108543026, "ret_signed_p1m": -0.08160494108543026, "alpha_spy_p1m": -0.045984528148268256, "alpha_c_p1m": -0.045984528148268256, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.035620412937162005, "ret_p3m": 0.08228496982503897, "ret_signed_p3m": -0.08228496982503897, "alpha_spy_p3m": -0.025677896030587233, "alpha_c_p3m": -0.025677896030587233, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.056607073794451734, "ret_p6m": 0.3466377479781608, "ret_signed_p6m": -0.3466377479781608, "alpha_spy_p6m": -0.2749498634397205, "alpha_c_p6m": -0.2749498634397205, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.07168788453844033, "price_path": [-0.0826, -0.0864, -0.0801, -0.0699, -0.0513, -0.0508, -0.0354, -0.0505, -0.0401, -0.0539, -0.0474, -0.0474, -0.0677, -0.0619, -0.034, -0.0257, -0.0165, -0.0165, -0.0241, -0.0211, -0.0012, 0.0105, 0.0105, -0.0202, -0.0197, -0.0134, -0.0075, -0.0088, -0.0172, -0.0094, 0.0027, 0.0112, 0.0071, 0.0092, -0.0017, 0.0167, 0.0097, 0.0104, 0.0041, -0.0034, -0.008, -0.0078, -0.0073, 0.0003, 0.0017, 0.0054, 0.0262, 0.025, 0.0248, 0.0386, 0.0376, 0.0311, 0.0272, 0.0377, 0.0216, -0.0007, -0.0043, 0.0143, 0.0075, 0.0116, 0.0162, 0.0184, 0.0, 0.0, -0.0131, -0.0015, 0.0082, 0.0236, 0.0367, 0.0476, 0.0636, 0.0666, 0.0682, 0.0714, 0.0765, 0.0746, 0.0842, 0.0775, 0.0884, 0.093, 0.0808, 0.0721, 0.0666, 0.0724, 0.0816, 0.0904, 0.0881, 0.0989, 0.1056, 0.1018, 0.1161, 0.1022, 0.0476, 0.0976, 0.083, 0.1049, 0.108, 0.1066, 0.125, 0.1183, 0.1056, 0.1171, 0.1302, 0.1387, 0.1406, 0.1413, 0.1312, 0.1331, 0.1374, 0.1124, 0.1226, 0.104, 0.0984, 0.1178, 0.1117, 0.112, 0.0879, 0.108, 0.0738, 0.0655, 0.0626, 0.0486, 0.0507, 0.0627, 0.0677, 0.0823, 0.0823, 0.0967, 0.0903, 0.0988, 0.1025, 0.0967, 0.111, 0.1148, 0.1173, 0.1357, 0.1207, 0.0898, 0.0899, 0.0868, 0.0675, 0.0824, 0.0961, 0.1021, 0.1114, 0.1119, 0.1119, 0.1206, 0.1261, 0.1329, 0.1297, 0.1297, 0.1517, 0.1612, 0.1706, 0.1663, 0.1637, 0.1722, 0.1846, 0.1754, 0.1827, 0.1966, 0.1974, 0.1974, 0.1832, 0.1994, 0.2136, 0.2396, 0.2437, 0.2682, 0.2659, 0.2417, 0.2102, 0.2259, 0.2239, 0.2118, 0.2067, 0.2456, 0.2583, 0.2704, 0.3043, 0.2903, 0.299, 0.299, 0.2961, 0.2977, 0.2922, 0.3226, 0.3013, 0.3466]}
{"unit_id": "orig:1943636215351349448", "ticker_idx": 0, "ts": "2025-07-11T11:39:19+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the market has partially priced in the strong demand expectations for its next-generation GPU (Blackwell) but has not fully valued the expansion of its customer base and the explosive growth of its AI inference business", "tweet_type": "news_reaction", "timeline": "months", "summary": "GS: NVDA not fully valued on inference/customer expansion; AVGO/NVDA rises not a bubble; neutral on AMD, ARM, MRVL as AI share still limited.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Although the AI investment cycle is shifting from \"investment\" to \"returns,\" this does not mean that a \"slowdown\" equates to \"peaking.\"\n\nGoldman Sachs pointed out in its latest AI report that despite a moderating growth rate, AI infrastructure investment will remain sustainable for the next 2-3 years. The market's excessive focus on \"slow returns\" may overlook the fact that cost-saving benefits have already begun to be realized, and stock prices have yet to reflect this structural change.\n\nThe report emphasizes that while the commercial monetization of AI is still in its early stages, the first phase of returns, driven by cost reduction, is already evident. It is estimated that by 2030, AI automation could save Fortune 500 companies approximately $935 billion in costs. Analysts believe this early gain is sufficient to support the current level of AI infrastructure investment, even though the growth rate may slow.\n\nAnalysts also noted that hyperscalers, as the primary investors in AI infrastructure, base their investment decisions more on long-term revenue growth opportunities rather than short-term cost savings.\n\nThe AI Investment Return Debate: Cost Savings vs. Commercial Monetization\n\nSince the generative AI boom began in late 2023, cumulative capital expenditures have surpassed $350 billion. Is the investment worthwhile? Are the returns sustainable? These questions are becoming the focal point of the market.\n\nThe Goldman Sachs research team divides AI value creation into three stages: the first stage achieves cost reduction through automation (currently underway), the second stage involves reinvestment and rebuilding, and the third stage realizes monetization through incremental revenue.\n\nThe report's analysis shows that AI applications in functions such as customer service, sales and marketing, and IT are already generating tangible benefits. In customer service, for example, 43% of call centers have adopted AI tools, reducing average operating costs by 30%.\n\nSpecifically, JPMorgan Chase plans to cut 10% of its back-office staff over the next five years, AT&T has reduced call center traffic by 30% with the help of AI, and T-Mobile expects to cut customer service interactions by 75% by 2027.\n\nThe core of the current debate is the sustainability of AI infrastructure investment. Goldman Sachs projects that by 2030, the world's Fortune 500 companies could save up to $935 billion in costs, representing about 14% of their total costs. The net present value (NPV) of AI investment returns is estimated at approximately $780 billion, which still shows a positive return relative to the cumulative investment of $350 billion.\n\nAnalysts point out that hyperscalers, the main investors in AI infrastructure, make investment decisions based more on long-term revenue growth opportunities than on short-term cost savings. This time lag between investment and returns complicates ROI calculations.\n\nInfrastructure Spending Secure in the Short Term, Inference Demand Becomes New Driver\n\nAnother key concern for the market regarding AI stocks is whether infrastructure spending has already peaked, especially with rising expectations of inventory buildup and weakening demand for training chips.\n\nGoldman Sachs believes these concerns are premature, stating in the report:\n\n- For the next 2-3 years, especially until 2026, major tech companies (like Microsoft, Amazon, Google, Meta) have the financial capacity to continue investing in AI infrastructure without significantly compressing their profit margins.\n\n- Key suppliers such as TSMC, Broadcom, and SK Hynix continue to raise their AI-related revenue forecasts. TSMC has been increasing its near-term and long-term AI revenue targets for six consecutive quarters, and SK Hynix expects its HBM revenue to double in 2025. These figures support the judgment of a fundamental balance between supply and demand.\n\n- Demand for \"inference\" computing power from enterprise customers and governments (sovereign AI) will become a new driver of spending, especially as small and medium-sized enterprises rapidly expand their deployment of custom models or edge AI applications.\n\nTaking NVIDIA as an example, Goldman Sachs responded by conducting an in-depth analysis of the supply-demand balance for the GB200 NVL72 rack.\n\nBased on NVIDIA's public statements, hyperscalers are currently deploying about 1,000 GB200 racks per week and plan to further accelerate deployment in the second quarter. The Goldman Sachs model shows that to avoid excess inventory at ODM manufacturers, hyperscalers will need to deploy 36,000 and 58,000 GB200-equivalent racks in 2025 and 2026, respectively.\n\nThe report notes that the rollout of NVIDIA's Blackwell architecture is on track, with the product already accounting for 70% of data center revenue in the first quarter, largely completing the transition from Hopper to Blackwell. This progress helps alleviate market concerns about inventory buildup.\n\nHowever, analysts also caution that the visibility of hyperscalers' capital expenditures for 2027 remains limited, which is a key uncertainty affecting the long-term supply-demand balance. In contrast, NVIDIA and Broadcom have better business visibility due to their longer product delivery cycles (about 12 months) and relatively concentrated customer bases.\n\nHow Are Stock Prices Pricing in AI Expectations?\nUsing NVIDIA as an example, Goldman Sachs points out that the market has partially priced in the strong demand expectations for its next-generation GPU (Blackwell) but has not fully valued the expansion of its customer base and the explosive growth of its AI inference business.\n\nFor Broadcom, the rise in its stock price is more based on the clear guidance provided by the company—that AI revenue will grow by 60% year-over-year in FY25 and FY26. This expectation also leads Goldman Sachs to believe that the rise in Broadcom's and NVIDIA's stock prices is not a \"bubble\" but rather a reflection of a clearer path to medium-term fundamental improvement.\n\nIn contrast, Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0049720531377652755, "ret_signed_m1d": -0.0049720531377652755, "alpha_spy_m1d": -0.008499878810943606, "alpha_c_m1d": -0.005076405593508637, "bench_spy_m1d": 0.003527825673178331, "bench_c_m1d": 0.00010435245574336172, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005153982348054642, "ret_signed_p1d": -0.005153982348054642, "alpha_spy_p1d": -0.007062099574675518, "alpha_c_p1d": 0.002394001147220215, "bench_spy_p1d": 0.0019081172266208757, "bench_c_p1d": -0.007547983495274857, "ret_p1w": 0.045416045545598305, "ret_signed_p1w": 0.045416045545598305, "alpha_spy_p1w": 0.0390660185553926, "alpha_c_p1w": 0.03560691470572075, "bench_spy_p1w": 0.006350026990205704, "bench_c_p1w": 0.009809130839877556, "ret_p1m": 0.10392914347709281, "ret_signed_p1m": 0.10392914347709281, "alpha_spy_p1m": 0.0842056151736994, "alpha_c_p1m": 0.08337192265984261, "bench_spy_p1m": 0.01972352830339341, "bench_c_p1m": 0.0205572208172502, "ret_p3m": 0.1467418245908143, "ret_signed_p3m": 0.1467418245908143, "alpha_spy_p3m": 0.06439016190235303, "alpha_c_p3m": -0.057021891634395416, "bench_spy_p3m": 0.08235166268846128, "bench_c_p3m": 0.20376371622520972, "ret_p6m": 0.14080210423433304, "ret_signed_p6m": 0.14080210423433304, "alpha_spy_p6m": 0.031690114078528175, "alpha_c_p6m": -0.1766869414198664, "bench_spy_p6m": 0.10911199015580486, "bench_c_p6m": 0.31748904565419944, "price_path": [-0.3197, -0.3665, -0.3847, -0.3847, -0.4124, -0.4004, -0.3773, -0.3547, -0.3269, -0.3408, -0.339, -0.3396, -0.3233, -0.3058, -0.3099, -0.3116, -0.2903, -0.2884, -0.2927, -0.2542, -0.2122, -0.1794, -0.1825, -0.1791, -0.178, -0.1852, -0.2009, -0.1946, -0.204, -0.204, -0.1784, -0.1826, -0.1561, -0.1807, -0.167, -0.1438, -0.1395, -0.1512, -0.1407, -0.1352, -0.1272, -0.1339, -0.1208, -0.1392, -0.1227, -0.1261, -0.1179, -0.1179, -0.1278, -0.1258, -0.1032, -0.0643, -0.06, -0.0435, -0.042, -0.0705, -0.0465, -0.0338, -0.0338, -0.0405, -0.0298, -0.0124, -0.005, 0.0, -0.0052, 0.035, 0.0391, 0.049, 0.0454, 0.0392, 0.0128, 0.0355, 0.0535, 0.052, 0.0717, 0.0642, 0.087, 0.0785, 0.0534, 0.0914, 0.0809, 0.0879, 0.0961, 0.1078, 0.1039, 0.1106, 0.1011, 0.1037, 0.0942, 0.1036, 0.065, 0.0635, 0.061, 0.0793, 0.0903, 0.1022, 0.1011, 0.0925, 0.0561, 0.0561, 0.0355, 0.0346, 0.0409, 0.0127, 0.0206, 0.0354, 0.0752, 0.0743, 0.0783, 0.0779, 0.0605, 0.0326, 0.0687, 0.0713, 0.1134, 0.082, 0.0731, 0.0775, 0.0805, 0.1027, 0.1314, 0.1354, 0.1454, 0.1377, 0.1251, 0.1221, 0.1467, 0.1677, 0.1107, 0.142, 0.0917, 0.0905, 0.1025, 0.111, 0.1075, 0.0985, 0.0932, 0.1046, 0.1295, 0.1612, 0.219, 0.2555, 0.2303, 0.2279, 0.2545, 0.2048, 0.1837, 0.1405, 0.1409, 0.207, 0.1713, 0.1752, 0.1331, 0.1532, 0.1315, 0.0997, 0.131, 0.0954, 0.0847, 0.107, 0.0783, 0.0931, 0.0931, 0.0733, 0.091, 0.1004, 0.089, 0.1121, 0.1062, 0.1252, 0.1217, 0.1145, 0.0972, 0.0614, 0.0691, 0.0777, 0.0366, 0.056, 0.0976, 0.1139, 0.1474, 0.1438, 0.1438, 0.1554, 0.1414, 0.1373, 0.131, 0.131, 0.1452, 0.1408]}
{"unit_id": "orig:1943636215351349448", "ticker_idx": 1, "ts": "2025-07-11T11:39:19+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the rise in Broadcom's and NVIDIA's stock prices is not a 'bubble' but rather a reflection of a clearer path to medium-term fundamental improvement", "tweet_type": "news_reaction", "timeline": "months", "summary": "GS: NVDA not fully valued on inference/customer expansion; AVGO/NVDA rises not a bubble; neutral on AMD, ARM, MRVL as AI share still limited.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Although the AI investment cycle is shifting from \"investment\" to \"returns,\" this does not mean that a \"slowdown\" equates to \"peaking.\"\n\nGoldman Sachs pointed out in its latest AI report that despite a moderating growth rate, AI infrastructure investment will remain sustainable for the next 2-3 years. The market's excessive focus on \"slow returns\" may overlook the fact that cost-saving benefits have already begun to be realized, and stock prices have yet to reflect this structural change.\n\nThe report emphasizes that while the commercial monetization of AI is still in its early stages, the first phase of returns, driven by cost reduction, is already evident. It is estimated that by 2030, AI automation could save Fortune 500 companies approximately $935 billion in costs. Analysts believe this early gain is sufficient to support the current level of AI infrastructure investment, even though the growth rate may slow.\n\nAnalysts also noted that hyperscalers, as the primary investors in AI infrastructure, base their investment decisions more on long-term revenue growth opportunities rather than short-term cost savings.\n\nThe AI Investment Return Debate: Cost Savings vs. Commercial Monetization\n\nSince the generative AI boom began in late 2023, cumulative capital expenditures have surpassed $350 billion. Is the investment worthwhile? Are the returns sustainable? These questions are becoming the focal point of the market.\n\nThe Goldman Sachs research team divides AI value creation into three stages: the first stage achieves cost reduction through automation (currently underway), the second stage involves reinvestment and rebuilding, and the third stage realizes monetization through incremental revenue.\n\nThe report's analysis shows that AI applications in functions such as customer service, sales and marketing, and IT are already generating tangible benefits. In customer service, for example, 43% of call centers have adopted AI tools, reducing average operating costs by 30%.\n\nSpecifically, JPMorgan Chase plans to cut 10% of its back-office staff over the next five years, AT&T has reduced call center traffic by 30% with the help of AI, and T-Mobile expects to cut customer service interactions by 75% by 2027.\n\nThe core of the current debate is the sustainability of AI infrastructure investment. Goldman Sachs projects that by 2030, the world's Fortune 500 companies could save up to $935 billion in costs, representing about 14% of their total costs. The net present value (NPV) of AI investment returns is estimated at approximately $780 billion, which still shows a positive return relative to the cumulative investment of $350 billion.\n\nAnalysts point out that hyperscalers, the main investors in AI infrastructure, make investment decisions based more on long-term revenue growth opportunities than on short-term cost savings. This time lag between investment and returns complicates ROI calculations.\n\nInfrastructure Spending Secure in the Short Term, Inference Demand Becomes New Driver\n\nAnother key concern for the market regarding AI stocks is whether infrastructure spending has already peaked, especially with rising expectations of inventory buildup and weakening demand for training chips.\n\nGoldman Sachs believes these concerns are premature, stating in the report:\n\n- For the next 2-3 years, especially until 2026, major tech companies (like Microsoft, Amazon, Google, Meta) have the financial capacity to continue investing in AI infrastructure without significantly compressing their profit margins.\n\n- Key suppliers such as TSMC, Broadcom, and SK Hynix continue to raise their AI-related revenue forecasts. TSMC has been increasing its near-term and long-term AI revenue targets for six consecutive quarters, and SK Hynix expects its HBM revenue to double in 2025. These figures support the judgment of a fundamental balance between supply and demand.\n\n- Demand for \"inference\" computing power from enterprise customers and governments (sovereign AI) will become a new driver of spending, especially as small and medium-sized enterprises rapidly expand their deployment of custom models or edge AI applications.\n\nTaking NVIDIA as an example, Goldman Sachs responded by conducting an in-depth analysis of the supply-demand balance for the GB200 NVL72 rack.\n\nBased on NVIDIA's public statements, hyperscalers are currently deploying about 1,000 GB200 racks per week and plan to further accelerate deployment in the second quarter. The Goldman Sachs model shows that to avoid excess inventory at ODM manufacturers, hyperscalers will need to deploy 36,000 and 58,000 GB200-equivalent racks in 2025 and 2026, respectively.\n\nThe report notes that the rollout of NVIDIA's Blackwell architecture is on track, with the product already accounting for 70% of data center revenue in the first quarter, largely completing the transition from Hopper to Blackwell. This progress helps alleviate market concerns about inventory buildup.\n\nHowever, analysts also caution that the visibility of hyperscalers' capital expenditures for 2027 remains limited, which is a key uncertainty affecting the long-term supply-demand balance. In contrast, NVIDIA and Broadcom have better business visibility due to their longer product delivery cycles (about 12 months) and relatively concentrated customer bases.\n\nHow Are Stock Prices Pricing in AI Expectations?\nUsing NVIDIA as an example, Goldman Sachs points out that the market has partially priced in the strong demand expectations for its next-generation GPU (Blackwell) but has not fully valued the expansion of its customer base and the explosive growth of its AI inference business.\n\nFor Broadcom, the rise in its stock price is more based on the clear guidance provided by the company—that AI revenue will grow by 60% year-over-year in FY25 and FY26. This expectation also leads Goldman Sachs to believe that the rise in Broadcom's and NVIDIA's stock prices is not a \"bubble\" but rather a reflection of a clearer path to medium-term fundamental improvement.\n\nIn contrast, Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0037173872870546187, "ret_signed_m1d": 0.0037173872870546187, "alpha_spy_m1d": 0.0001895616138762879, "alpha_c_m1d": 0.003613034831311257, "bench_spy_m1d": 0.003527825673178331, "bench_c_m1d": 0.00010435245574336172, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0044463898433175775, "ret_signed_p1d": 0.0044463898433175775, "alpha_spy_p1d": 0.002538272616696702, "alpha_c_p1d": 0.011994373338592434, "bench_spy_p1d": 0.0019081172266208757, "bench_c_p1d": -0.007547983495274857, "ret_p1w": 0.03265544410652277, "ret_signed_p1w": 0.03265544410652277, "alpha_spy_p1w": 0.026305417116317065, "alpha_c_p1w": 0.022846313266645213, "bench_spy_p1w": 0.006350026990205704, "bench_c_p1w": 0.009809130839877556, "ret_p1m": 0.10758800270078828, "ret_signed_p1m": 0.10758800270078828, "alpha_spy_p1m": 0.08786447439739486, "alpha_c_p1m": 0.08703078188353808, "bench_spy_p1m": 0.01972352830339341, "bench_c_p1m": 0.0205572208172502, "ret_p3m": 0.2613600008590977, "ret_signed_p3m": 0.2613600008590977, "alpha_spy_p3m": 0.17900833817063644, "alpha_c_p3m": 0.057596284633887995, "bench_spy_p3m": 0.08235166268846128, "bench_c_p3m": 0.20376371622520972, "ret_p6m": 0.25616536825088154, "ret_signed_p6m": 0.25616536825088154, "alpha_spy_p6m": 0.14705337809507668, "alpha_c_p6m": -0.0613236774033179, "bench_spy_p6m": 0.10911199015580486, "bench_c_p6m": 0.31748904565419944, "price_path": [-0.3493, -0.3651, -0.3783, -0.3783, -0.3957, -0.3834, -0.3568, -0.3159, -0.3008, -0.3002, -0.3049, -0.3002, -0.2825, -0.2596, -0.2702, -0.2725, -0.2553, -0.2445, -0.243, -0.1943, -0.1549, -0.156, -0.1541, -0.1688, -0.1614, -0.1576, -0.1647, -0.1618, -0.1684, -0.1684, -0.1432, -0.1294, -0.1202, -0.1198, -0.0957, -0.0661, -0.0507, -0.0549, -0.1022, -0.1118, -0.1105, -0.0804, -0.0689, -0.0957, -0.0834, -0.0933, -0.0864, -0.0864, -0.0889, -0.0751, -0.0387, -0.0355, -0.0153, -0.0183, 0.0046, -0.0351, -0.0163, 0.0029, 0.0029, -0.0007, -0.0094, 0.0128, 0.0037, 0.0, 0.0044, 0.0239, 0.0234, 0.044, 0.0327, 0.0504, 0.0153, 0.0339, 0.0522, 0.0576, 0.0726, 0.084, 0.1029, 0.0704, 0.052, 0.0851, 0.0676, 0.0995, 0.1071, 0.1115, 0.1076, 0.1401, 0.1265, 0.1343, 0.1165, 0.1144, 0.0748, 0.0612, 0.0555, 0.0715, 0.0723, 0.0861, 0.0943, 0.1249, 0.0839, 0.0839, 0.087, 0.1021, 0.1156, 0.2205, 0.2597, 0.227, 0.3469, 0.3107, 0.3116, 0.327, 0.312, 0.2616, 0.2587, 0.2572, 0.2369, 0.2374, 0.2388, 0.227, 0.2213, 0.1971, 0.2044, 0.2171, 0.2346, 0.2353, 0.2248, 0.2282, 0.2614, 0.2596, 0.1852, 0.3022, 0.2564, 0.2826, 0.2929, 0.2753, 0.275, 0.251, 0.2424, 0.2569, 0.2929, 0.3218, 0.3616, 0.4091, 0.3744, 0.3495, 0.3236, 0.2849, 0.3106, 0.2982, 0.2757, 0.3084, 0.2849, 0.2968, 0.2412, 0.2503, 0.251, 0.2431, 0.2939, 0.2662, 0.242, 0.3799, 0.4057, 0.4515, 0.4515, 0.4711, 0.4095, 0.393, 0.3895, 0.3911, 0.4247, 0.4643, 0.4833, 0.5077, 0.4836, 0.314, 0.2406, 0.246, 0.1902, 0.2043, 0.2426, 0.249, 0.2777, 0.281, 0.281, 0.288, 0.278, 0.2797, 0.266, 0.266, 0.2715, 0.2562]}
{"unit_id": "orig:1943636215351349448", "ticker_idx": 2, "ts": "2025-07-11T11:39:19+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share", "tweet_type": "news_reaction", "timeline": "months", "summary": "GS: NVDA not fully valued on inference/customer expansion; AVGO/NVDA rises not a bubble; neutral on AMD, ARM, MRVL as AI share still limited.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Although the AI investment cycle is shifting from \"investment\" to \"returns,\" this does not mean that a \"slowdown\" equates to \"peaking.\"\n\nGoldman Sachs pointed out in its latest AI report that despite a moderating growth rate, AI infrastructure investment will remain sustainable for the next 2-3 years. The market's excessive focus on \"slow returns\" may overlook the fact that cost-saving benefits have already begun to be realized, and stock prices have yet to reflect this structural change.\n\nThe report emphasizes that while the commercial monetization of AI is still in its early stages, the first phase of returns, driven by cost reduction, is already evident. It is estimated that by 2030, AI automation could save Fortune 500 companies approximately $935 billion in costs. Analysts believe this early gain is sufficient to support the current level of AI infrastructure investment, even though the growth rate may slow.\n\nAnalysts also noted that hyperscalers, as the primary investors in AI infrastructure, base their investment decisions more on long-term revenue growth opportunities rather than short-term cost savings.\n\nThe AI Investment Return Debate: Cost Savings vs. Commercial Monetization\n\nSince the generative AI boom began in late 2023, cumulative capital expenditures have surpassed $350 billion. Is the investment worthwhile? Are the returns sustainable? These questions are becoming the focal point of the market.\n\nThe Goldman Sachs research team divides AI value creation into three stages: the first stage achieves cost reduction through automation (currently underway), the second stage involves reinvestment and rebuilding, and the third stage realizes monetization through incremental revenue.\n\nThe report's analysis shows that AI applications in functions such as customer service, sales and marketing, and IT are already generating tangible benefits. In customer service, for example, 43% of call centers have adopted AI tools, reducing average operating costs by 30%.\n\nSpecifically, JPMorgan Chase plans to cut 10% of its back-office staff over the next five years, AT&T has reduced call center traffic by 30% with the help of AI, and T-Mobile expects to cut customer service interactions by 75% by 2027.\n\nThe core of the current debate is the sustainability of AI infrastructure investment. Goldman Sachs projects that by 2030, the world's Fortune 500 companies could save up to $935 billion in costs, representing about 14% of their total costs. The net present value (NPV) of AI investment returns is estimated at approximately $780 billion, which still shows a positive return relative to the cumulative investment of $350 billion.\n\nAnalysts point out that hyperscalers, the main investors in AI infrastructure, make investment decisions based more on long-term revenue growth opportunities than on short-term cost savings. This time lag between investment and returns complicates ROI calculations.\n\nInfrastructure Spending Secure in the Short Term, Inference Demand Becomes New Driver\n\nAnother key concern for the market regarding AI stocks is whether infrastructure spending has already peaked, especially with rising expectations of inventory buildup and weakening demand for training chips.\n\nGoldman Sachs believes these concerns are premature, stating in the report:\n\n- For the next 2-3 years, especially until 2026, major tech companies (like Microsoft, Amazon, Google, Meta) have the financial capacity to continue investing in AI infrastructure without significantly compressing their profit margins.\n\n- Key suppliers such as TSMC, Broadcom, and SK Hynix continue to raise their AI-related revenue forecasts. TSMC has been increasing its near-term and long-term AI revenue targets for six consecutive quarters, and SK Hynix expects its HBM revenue to double in 2025. These figures support the judgment of a fundamental balance between supply and demand.\n\n- Demand for \"inference\" computing power from enterprise customers and governments (sovereign AI) will become a new driver of spending, especially as small and medium-sized enterprises rapidly expand their deployment of custom models or edge AI applications.\n\nTaking NVIDIA as an example, Goldman Sachs responded by conducting an in-depth analysis of the supply-demand balance for the GB200 NVL72 rack.\n\nBased on NVIDIA's public statements, hyperscalers are currently deploying about 1,000 GB200 racks per week and plan to further accelerate deployment in the second quarter. The Goldman Sachs model shows that to avoid excess inventory at ODM manufacturers, hyperscalers will need to deploy 36,000 and 58,000 GB200-equivalent racks in 2025 and 2026, respectively.\n\nThe report notes that the rollout of NVIDIA's Blackwell architecture is on track, with the product already accounting for 70% of data center revenue in the first quarter, largely completing the transition from Hopper to Blackwell. This progress helps alleviate market concerns about inventory buildup.\n\nHowever, analysts also caution that the visibility of hyperscalers' capital expenditures for 2027 remains limited, which is a key uncertainty affecting the long-term supply-demand balance. In contrast, NVIDIA and Broadcom have better business visibility due to their longer product delivery cycles (about 12 months) and relatively concentrated customer bases.\n\nHow Are Stock Prices Pricing in AI Expectations?\nUsing NVIDIA as an example, Goldman Sachs points out that the market has partially priced in the strong demand expectations for its next-generation GPU (Blackwell) but has not fully valued the expansion of its customer base and the explosive growth of its AI inference business.\n\nFor Broadcom, the rise in its stock price is more based on the clear guidance provided by the company—that AI revenue will grow by 60% year-over-year in FY25 and FY26. This expectation also leads Goldman Sachs to believe that the rise in Broadcom's and NVIDIA's stock prices is not a \"bubble\" but rather a reflection of a clearer path to medium-term fundamental improvement.\n\nIn contrast, Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015435012533112191, "ret_signed_m1d": 0.015435012533112191, "alpha_spy_m1d": 0.018962838206290522, "alpha_c_m1d": 0.015539364988855553, "bench_spy_m1d": 0.003527825673178331, "bench_c_m1d": 0.00010435245574336172, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0012292902474535516, "ret_signed_p1d": 0.0012292902474535516, "alpha_spy_p1d": 0.0031374074740744273, "alpha_c_p1d": -0.006318693247821305, "bench_spy_p1d": 0.0019081172266208757, "bench_c_p1d": -0.007547983495274857, "ret_p1w": 0.07218964251059923, "ret_signed_p1w": -0.07218964251059923, "alpha_spy_p1w": -0.06583961552039352, "alpha_c_p1w": -0.06238051167072167, "bench_spy_p1w": 0.006350026990205704, "bench_c_p1w": 0.009809130839877556, "ret_p1m": 0.17661522287763765, "ret_signed_p1m": -0.17661522287763765, "alpha_spy_p1m": -0.15689169457424423, "alpha_c_p1m": -0.15605800206038745, "bench_spy_p1m": 0.01972352830339341, "bench_c_p1m": 0.0205572208172502, "ret_p3m": 0.6087966159294382, "ret_signed_p3m": -0.6087966159294382, "alpha_spy_p3m": -0.526444953240977, "alpha_c_p3m": -0.4050328997042285, "bench_spy_p3m": 0.08235166268846128, "bench_c_p3m": 0.20376371622520972, "ret_p6m": 0.5099030501008726, "ret_signed_p6m": -0.5099030501008726, "alpha_spy_p6m": -0.4007910599450677, "alpha_c_p6m": -0.19241400444667311, "bench_spy_p6m": 0.10911199015580486, "bench_c_p6m": 0.31748904565419944, "price_path": [-0.3492, -0.397, -0.4024, -0.4024, -0.4157, -0.4109, -0.3827, -0.3548, -0.3399, -0.3417, -0.3439, -0.3351, -0.3399, -0.3252, -0.313, -0.3265, -0.3146, -0.3054, -0.2976, -0.2616, -0.2319, -0.196, -0.2147, -0.1998, -0.2164, -0.2248, -0.2347, -0.2439, -0.2466, -0.2466, -0.2176, -0.2292, -0.228, -0.2438, -0.2171, -0.1988, -0.1901, -0.2099, -0.2065, -0.1686, -0.1583, -0.1727, -0.1907, -0.2067, -0.1368, -0.1319, -0.1341, -0.1341, -0.1242, -0.115, -0.0546, -0.0206, -0.0187, -0.0178, -0.0309, -0.0704, -0.054, -0.0581, -0.0581, -0.0794, -0.0587, -0.0547, -0.0154, 0.0, -0.0012, 0.0628, 0.0933, 0.0955, 0.0722, 0.0723, 0.0567, 0.0835, 0.1072, 0.1369, 0.186, 0.2119, 0.226, 0.2041, 0.1727, 0.2073, 0.1905, 0.1141, 0.1774, 0.1799, 0.1766, 0.1949, 0.2595, 0.2358, 0.2123, 0.203, 0.1375, 0.1283, 0.1181, 0.1457, 0.1157, 0.138, 0.1414, 0.1513, 0.1107, 0.1107, 0.1086, 0.1073, 0.105, 0.0322, 0.0341, 0.0642, 0.0896, 0.0632, 0.083, 0.1007, 0.0959, 0.087, 0.0785, 0.0749, 0.0913, 0.0989, 0.0988, 0.1014, 0.0891, 0.102, 0.105, 0.1201, 0.1592, 0.1246, 0.3913, 0.4445, 0.6088, 0.5906, 0.4677, 0.4781, 0.4895, 0.6296, 0.602, 0.5919, 0.6429, 0.6257, 0.5724, 0.6049, 0.7274, 0.7735, 0.7621, 0.8053, 0.7405, 0.7492, 0.7733, 0.7078, 0.7506, 0.6234, 0.595, 0.6663, 0.6222, 0.7681, 0.6935, 0.6856, 0.6427, 0.5728, 0.5268, 0.407, 0.3917, 0.4687, 0.4078, 0.4632, 0.4632, 0.4857, 0.5009, 0.47, 0.4861, 0.4751, 0.4887, 0.5101, 0.5136, 0.5122, 0.5123, 0.4396, 0.4177, 0.4286, 0.353, 0.3732, 0.4577, 0.468, 0.4677, 0.4687, 0.4687, 0.4683, 0.4725, 0.4707, 0.4626, 0.4626, 0.5262, 0.5099]}
{"unit_id": "orig:1943636215351349448", "ticker_idx": 3, "ts": "2025-07-11T11:39:19+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share", "tweet_type": "news_reaction", "timeline": "months", "summary": "GS: NVDA not fully valued on inference/customer expansion; AVGO/NVDA rises not a bubble; neutral on AMD, ARM, MRVL as AI share still limited.", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Although the AI investment cycle is shifting from \"investment\" to \"returns,\" this does not mean that a \"slowdown\" equates to \"peaking.\"\n\nGoldman Sachs pointed out in its latest AI report that despite a moderating growth rate, AI infrastructure investment will remain sustainable for the next 2-3 years. The market's excessive focus on \"slow returns\" may overlook the fact that cost-saving benefits have already begun to be realized, and stock prices have yet to reflect this structural change.\n\nThe report emphasizes that while the commercial monetization of AI is still in its early stages, the first phase of returns, driven by cost reduction, is already evident. It is estimated that by 2030, AI automation could save Fortune 500 companies approximately $935 billion in costs. Analysts believe this early gain is sufficient to support the current level of AI infrastructure investment, even though the growth rate may slow.\n\nAnalysts also noted that hyperscalers, as the primary investors in AI infrastructure, base their investment decisions more on long-term revenue growth opportunities rather than short-term cost savings.\n\nThe AI Investment Return Debate: Cost Savings vs. Commercial Monetization\n\nSince the generative AI boom began in late 2023, cumulative capital expenditures have surpassed $350 billion. Is the investment worthwhile? Are the returns sustainable? These questions are becoming the focal point of the market.\n\nThe Goldman Sachs research team divides AI value creation into three stages: the first stage achieves cost reduction through automation (currently underway), the second stage involves reinvestment and rebuilding, and the third stage realizes monetization through incremental revenue.\n\nThe report's analysis shows that AI applications in functions such as customer service, sales and marketing, and IT are already generating tangible benefits. In customer service, for example, 43% of call centers have adopted AI tools, reducing average operating costs by 30%.\n\nSpecifically, JPMorgan Chase plans to cut 10% of its back-office staff over the next five years, AT&T has reduced call center traffic by 30% with the help of AI, and T-Mobile expects to cut customer service interactions by 75% by 2027.\n\nThe core of the current debate is the sustainability of AI infrastructure investment. Goldman Sachs projects that by 2030, the world's Fortune 500 companies could save up to $935 billion in costs, representing about 14% of their total costs. The net present value (NPV) of AI investment returns is estimated at approximately $780 billion, which still shows a positive return relative to the cumulative investment of $350 billion.\n\nAnalysts point out that hyperscalers, the main investors in AI infrastructure, make investment decisions based more on long-term revenue growth opportunities than on short-term cost savings. This time lag between investment and returns complicates ROI calculations.\n\nInfrastructure Spending Secure in the Short Term, Inference Demand Becomes New Driver\n\nAnother key concern for the market regarding AI stocks is whether infrastructure spending has already peaked, especially with rising expectations of inventory buildup and weakening demand for training chips.\n\nGoldman Sachs believes these concerns are premature, stating in the report:\n\n- For the next 2-3 years, especially until 2026, major tech companies (like Microsoft, Amazon, Google, Meta) have the financial capacity to continue investing in AI infrastructure without significantly compressing their profit margins.\n\n- Key suppliers such as TSMC, Broadcom, and SK Hynix continue to raise their AI-related revenue forecasts. TSMC has been increasing its near-term and long-term AI revenue targets for six consecutive quarters, and SK Hynix expects its HBM revenue to double in 2025. These figures support the judgment of a fundamental balance between supply and demand.\n\n- Demand for \"inference\" computing power from enterprise customers and governments (sovereign AI) will become a new driver of spending, especially as small and medium-sized enterprises rapidly expand their deployment of custom models or edge AI applications.\n\nTaking NVIDIA as an example, Goldman Sachs responded by conducting an in-depth analysis of the supply-demand balance for the GB200 NVL72 rack.\n\nBased on NVIDIA's public statements, hyperscalers are currently deploying about 1,000 GB200 racks per week and plan to further accelerate deployment in the second quarter. The Goldman Sachs model shows that to avoid excess inventory at ODM manufacturers, hyperscalers will need to deploy 36,000 and 58,000 GB200-equivalent racks in 2025 and 2026, respectively.\n\nThe report notes that the rollout of NVIDIA's Blackwell architecture is on track, with the product already accounting for 70% of data center revenue in the first quarter, largely completing the transition from Hopper to Blackwell. This progress helps alleviate market concerns about inventory buildup.\n\nHowever, analysts also caution that the visibility of hyperscalers' capital expenditures for 2027 remains limited, which is a key uncertainty affecting the long-term supply-demand balance. In contrast, NVIDIA and Broadcom have better business visibility due to their longer product delivery cycles (about 12 months) and relatively concentrated customer bases.\n\nHow Are Stock Prices Pricing in AI Expectations?\nUsing NVIDIA as an example, Goldman Sachs points out that the market has partially priced in the strong demand expectations for its next-generation GPU (Blackwell) but has not fully valued the expansion of its customer base and the explosive growth of its AI inference business.\n\nFor Broadcom, the rise in its stock price is more based on the clear guidance provided by the company—that AI revenue will grow by 60% year-over-year in FY25 and FY26. This expectation also leads Goldman Sachs to believe that the rise in Broadcom's and NVIDIA's stock prices is not a \"bubble\" but rather a reflection of a clearer path to medium-term fundamental improvement.\n\nIn contrast, Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017884065826293538, "ret_signed_m1d": -0.017884065826293538, "alpha_spy_m1d": -0.014356240153115207, "alpha_c_m1d": -0.017779713370550176, "bench_spy_m1d": 0.003527825673178331, "bench_c_m1d": 0.00010435245574336172, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009593045990495486, "ret_signed_p1d": 0.009593045990495486, "alpha_spy_p1d": 0.011501163217116361, "alpha_c_p1d": 0.0020450624952206287, "bench_spy_p1d": 0.0019081172266208757, "bench_c_p1d": -0.007547983495274857, "ret_p1w": 0.07400303461070545, "ret_signed_p1w": -0.07400303461070545, "alpha_spy_p1w": -0.06765300762049975, "alpha_c_p1w": -0.0641939037708279, "bench_spy_p1w": 0.006350026990205704, "bench_c_p1w": 0.009809130839877556, "ret_p1m": -0.033506915909582324, "ret_signed_p1m": 0.033506915909582324, "alpha_spy_p1m": 0.053230444212975736, "alpha_c_p1m": 0.054064136726832523, "bench_spy_p1m": 0.01972352830339341, "bench_c_p1m": 0.0205572208172502, "ret_p3m": 0.14272989915439926, "ret_signed_p3m": -0.14272989915439926, "alpha_spy_p3m": -0.060378236465937984, "alpha_c_p3m": 0.06103381707081046, "bench_spy_p3m": 0.08235166268846128, "bench_c_p3m": 0.20376371622520972, "ret_p6m": -0.20439907723745032, "ret_signed_p6m": 0.20439907723745032, "alpha_spy_p6m": 0.3135110673932552, "alpha_c_p6m": 0.5218881228916498, "bench_spy_p6m": 0.10911199015580486, "bench_c_p6m": 0.31748904565419944, "price_path": [-0.2887, -0.3079, -0.3098, -0.3098, -0.3365, -0.311, -0.2794, -0.2325, -0.2234, -0.232, -0.235, -0.2185, -0.2093, -0.1553, -0.1644, -0.161, -0.149, -0.2016, -0.2065, -0.1448, -0.132, -0.0856, -0.0882, -0.0684, -0.0952, -0.1021, -0.1061, -0.1142, -0.1285, -0.1285, -0.0821, -0.0713, -0.1222, -0.1466, -0.1363, -0.1176, -0.1068, -0.1123, -0.0879, -0.0502, -0.0364, -0.0381, -0.0502, -0.0712, -0.0267, -0.0084, 0.0008, 0.0008, -0.0062, 0.0232, 0.0717, 0.0779, 0.0837, 0.1338, 0.1083, 0.0712, 0.0595, 0.0627, 0.0627, 0.0064, 0.0127, 0.0143, 0.0179, 0.0, -0.0096, 0.008, 0.0545, 0.077, 0.074, 0.1095, 0.0724, 0.0914, 0.0963, 0.1181, 0.1263, 0.1201, 0.1191, -0.0313, -0.0573, -0.0404, -0.0597, -0.0673, -0.0711, -0.051, -0.0335, -0.0243, -0.0297, -0.0369, -0.0482, -0.0334, -0.0817, -0.1013, -0.0867, -0.055, -0.0559, -0.0389, -0.0362, -0.0232, -0.0523, -0.0523, -0.0932, -0.0995, -0.0717, -0.0532, -0.0466, -0.0352, 0.0562, 0.06, 0.0322, 0.0543, 0.0542, 0.0509, 0.0041, -0.0208, -0.01, -0.0339, -0.0112, -0.0362, -0.0433, -0.0421, -0.0305, 0.0304, 0.0425, 0.0459, 0.0704, 0.0919, 0.1427, 0.1694, 0.0608, 0.1782, 0.1523, 0.1695, 0.173, 0.1348, 0.1751, 0.1606, 0.1355, 0.1416, 0.1695, 0.2239, 0.186, 0.1675, 0.1337, 0.1636, 0.1558, 0.1013, 0.0976, 0.0843, 0.0441, 0.061, 0.026, 0.0193, -0.0386, -0.0423, -0.0389, -0.0678, -0.0613, -0.0919, -0.0985, -0.0769, -0.0994, -0.0913, -0.0913, -0.0711, -0.0749, -0.0648, -0.0463, -0.0373, -0.0317, -0.0422, -0.0275, -0.0303, -0.0672, -0.1031, -0.1478, -0.1702, -0.2149, -0.2222, -0.2187, -0.2237, -0.2324, -0.2356, -0.2356, -0.2444, -0.2428, -0.2404, -0.251, -0.251, -0.2139, -0.2044]}
{"unit_id": "orig:1943636215351349448", "ticker_idx": 4, "ts": "2025-07-11T11:39:19+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share", "tweet_type": "news_reaction", "timeline": "months", "summary": "GS: NVDA not fully valued on inference/customer expansion; AVGO/NVDA rises not a bubble; neutral on AMD, ARM, MRVL as AI share still limited.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Although the AI investment cycle is shifting from \"investment\" to \"returns,\" this does not mean that a \"slowdown\" equates to \"peaking.\"\n\nGoldman Sachs pointed out in its latest AI report that despite a moderating growth rate, AI infrastructure investment will remain sustainable for the next 2-3 years. The market's excessive focus on \"slow returns\" may overlook the fact that cost-saving benefits have already begun to be realized, and stock prices have yet to reflect this structural change.\n\nThe report emphasizes that while the commercial monetization of AI is still in its early stages, the first phase of returns, driven by cost reduction, is already evident. It is estimated that by 2030, AI automation could save Fortune 500 companies approximately $935 billion in costs. Analysts believe this early gain is sufficient to support the current level of AI infrastructure investment, even though the growth rate may slow.\n\nAnalysts also noted that hyperscalers, as the primary investors in AI infrastructure, base their investment decisions more on long-term revenue growth opportunities rather than short-term cost savings.\n\nThe AI Investment Return Debate: Cost Savings vs. Commercial Monetization\n\nSince the generative AI boom began in late 2023, cumulative capital expenditures have surpassed $350 billion. Is the investment worthwhile? Are the returns sustainable? These questions are becoming the focal point of the market.\n\nThe Goldman Sachs research team divides AI value creation into three stages: the first stage achieves cost reduction through automation (currently underway), the second stage involves reinvestment and rebuilding, and the third stage realizes monetization through incremental revenue.\n\nThe report's analysis shows that AI applications in functions such as customer service, sales and marketing, and IT are already generating tangible benefits. In customer service, for example, 43% of call centers have adopted AI tools, reducing average operating costs by 30%.\n\nSpecifically, JPMorgan Chase plans to cut 10% of its back-office staff over the next five years, AT&T has reduced call center traffic by 30% with the help of AI, and T-Mobile expects to cut customer service interactions by 75% by 2027.\n\nThe core of the current debate is the sustainability of AI infrastructure investment. Goldman Sachs projects that by 2030, the world's Fortune 500 companies could save up to $935 billion in costs, representing about 14% of their total costs. The net present value (NPV) of AI investment returns is estimated at approximately $780 billion, which still shows a positive return relative to the cumulative investment of $350 billion.\n\nAnalysts point out that hyperscalers, the main investors in AI infrastructure, make investment decisions based more on long-term revenue growth opportunities than on short-term cost savings. This time lag between investment and returns complicates ROI calculations.\n\nInfrastructure Spending Secure in the Short Term, Inference Demand Becomes New Driver\n\nAnother key concern for the market regarding AI stocks is whether infrastructure spending has already peaked, especially with rising expectations of inventory buildup and weakening demand for training chips.\n\nGoldman Sachs believes these concerns are premature, stating in the report:\n\n- For the next 2-3 years, especially until 2026, major tech companies (like Microsoft, Amazon, Google, Meta) have the financial capacity to continue investing in AI infrastructure without significantly compressing their profit margins.\n\n- Key suppliers such as TSMC, Broadcom, and SK Hynix continue to raise their AI-related revenue forecasts. TSMC has been increasing its near-term and long-term AI revenue targets for six consecutive quarters, and SK Hynix expects its HBM revenue to double in 2025. These figures support the judgment of a fundamental balance between supply and demand.\n\n- Demand for \"inference\" computing power from enterprise customers and governments (sovereign AI) will become a new driver of spending, especially as small and medium-sized enterprises rapidly expand their deployment of custom models or edge AI applications.\n\nTaking NVIDIA as an example, Goldman Sachs responded by conducting an in-depth analysis of the supply-demand balance for the GB200 NVL72 rack.\n\nBased on NVIDIA's public statements, hyperscalers are currently deploying about 1,000 GB200 racks per week and plan to further accelerate deployment in the second quarter. The Goldman Sachs model shows that to avoid excess inventory at ODM manufacturers, hyperscalers will need to deploy 36,000 and 58,000 GB200-equivalent racks in 2025 and 2026, respectively.\n\nThe report notes that the rollout of NVIDIA's Blackwell architecture is on track, with the product already accounting for 70% of data center revenue in the first quarter, largely completing the transition from Hopper to Blackwell. This progress helps alleviate market concerns about inventory buildup.\n\nHowever, analysts also caution that the visibility of hyperscalers' capital expenditures for 2027 remains limited, which is a key uncertainty affecting the long-term supply-demand balance. In contrast, NVIDIA and Broadcom have better business visibility due to their longer product delivery cycles (about 12 months) and relatively concentrated customer bases.\n\nHow Are Stock Prices Pricing in AI Expectations?\nUsing NVIDIA as an example, Goldman Sachs points out that the market has partially priced in the strong demand expectations for its next-generation GPU (Blackwell) but has not fully valued the expansion of its customer base and the explosive growth of its AI inference business.\n\nFor Broadcom, the rise in its stock price is more based on the clear guidance provided by the company—that AI revenue will grow by 60% year-over-year in FY25 and FY26. This expectation also leads Goldman Sachs to believe that the rise in Broadcom's and NVIDIA's stock prices is not a \"bubble\" but rather a reflection of a clearer path to medium-term fundamental improvement.\n\nIn contrast, Goldman Sachs maintains a neutral rating on companies like AMD, Arm, and Marvell because their AI-related businesses are still in the early stages with limited market share.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008114626817470327, "ret_signed_m1d": -0.008114626817470327, "alpha_spy_m1d": -0.0045868011442919965, "alpha_c_m1d": -0.008010274361726966, "bench_spy_m1d": 0.003527825673178331, "bench_c_m1d": 0.00010435245574336172, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002750640773673818, "ret_signed_p1d": 0.002750640773673818, "alpha_spy_p1d": 0.004658758000294694, "alpha_c_p1d": -0.004797342721601039, "bench_spy_p1d": 0.0019081172266208757, "bench_c_p1d": -0.007547983495274857, "ret_p1w": 0.02668137319472441, "ret_signed_p1w": -0.02668137319472441, "alpha_spy_p1w": -0.020331346204518708, "alpha_c_p1w": -0.016872242354846856, "bench_spy_p1w": 0.006350026990205704, "bench_c_p1w": 0.009809130839877556, "ret_p1m": 0.0628525359036678, "ret_signed_p1m": -0.0628525359036678, "alpha_spy_p1m": -0.04312900760027438, "alpha_c_p1m": -0.042295315086417595, "bench_spy_p1m": 0.01972352830339341, "bench_c_p1m": 0.0205572208172502, "ret_p3m": 0.2721771923574132, "ret_signed_p3m": -0.2721771923574132, "alpha_spy_p3m": -0.18982552966895194, "alpha_c_p3m": -0.0684134761322035, "bench_spy_p3m": 0.08235166268846128, "bench_c_p3m": 0.20376371622520972, "ret_p6m": 0.24177906378133063, "ret_signed_p6m": -0.24177906378133063, "alpha_spy_p6m": -0.13266707362552577, "alpha_c_p6m": 0.07570998187286881, "bench_spy_p6m": 0.10911199015580486, "bench_c_p6m": 0.31748904565419944, "price_path": [-0.2673, -0.2864, -0.2895, -0.2895, -0.3214, -0.3044, -0.2611, -0.2122, -0.1903, -0.1933, -0.1936, -0.1979, -0.1624, -0.1435, -0.1483, -0.1587, -0.2262, -0.208, -0.1803, -0.1136, -0.1002, -0.0944, -0.104, -0.1238, -0.1403, -0.156, -0.174, -0.1501, -0.166, -0.166, -0.123, -0.1124, -0.1242, -0.1729, -0.1553, -0.143, -0.0889, -0.1046, -0.0607, -0.0499, -0.054, -0.0622, -0.043, -0.0767, -0.0323, -0.0382, 0.03, 0.03, 0.0102, -0.0273, 0.0335, 0.0434, 0.0989, 0.0603, 0.0636, 0.0477, 0.0203, 0.0331, 0.0331, -0.0168, -0.0113, -0.007, 0.0081, 0.0, -0.0028, -0.0041, -0.0256, -0.0096, 0.0267, 0.0048, -0.0099, 0.0077, 0.0183, 0.0206, 0.044, 0.0499, 0.1242, 0.1054, 0.0239, 0.0525, 0.0539, 0.0359, 0.0432, 0.0637, 0.0629, 0.0701, 0.0909, 0.0871, 0.0479, 0.0554, -0.0088, -0.0205, -0.0206, 0.004, 0.0033, 0.0213, 0.0286, 0.0622, -0.1353, -0.1353, -0.1115, -0.143, -0.1184, -0.129, -0.0923, -0.0807, -0.0772, -0.0842, -0.0737, -0.0726, -0.0529, -0.0238, 0.0209, 0.0213, 0.0388, 0.0263, 0.1015, 0.1527, 0.1439, 0.1331, 0.1562, 0.1538, 0.1855, 0.1858, 0.2229, 0.1961, 0.2722, 0.2471, 0.1782, 0.2302, 0.1866, 0.2233, 0.2143, 0.2104, 0.1814, 0.1596, 0.1154, 0.1391, 0.1578, 0.2209, 0.2174, 0.2407, 0.2189, 0.2901, 0.2437, 0.2054, 0.2785, 0.2844, 0.2513, 0.2831, 0.2294, 0.2294, 0.2045, 0.1898, 0.1485, 0.0828, 0.1192, 0.0553, 0.0659, 0.1531, 0.1482, 0.2072, 0.2072, 0.2304, 0.2538, 0.2784, 0.379, 0.3513, 0.3612, 0.2661, 0.2235, 0.2726, 0.2308, 0.162, 0.1596, 0.157, 0.1244, 0.1625, 0.1573, 0.167, 0.2067, 0.1903, 0.1903, 0.1882, 0.1803, 0.194, 0.1695, 0.1695, 0.2302, 0.2418]}
{"unit_id": "orig:1958391813926244698", "ticker_idx": 0, "ts": "2025-08-21T04:52:48+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KeyBanc analyst John Vinh, while issuing a cautionary note, still raised Nvidia's price target from $190 to $215 and maintained an 'Overweight' rating. Susquehanna analyst Christopher Rolland also saw continued momentum in Nvidia's data center business, raising his price target from $180 to $210 while maintaining a 'Positive' rating.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses KeyBanc and Susquehanna Buy/Overweight/Positive calls on NVDA with raised PTs of $215 and $210 respectively.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "The highly anticipated Nvidia earnings report: Will the Q3 guidance fall short of expectations?\n\nNvidia will release its latest earnings report next Wednesday, and the Q3 guidance will be the main focus of the market.\n\nAccording to the latest report from KeyBanc Capital Markets, Nvidia may temporarily exclude direct revenue from the Chinese market in its guidance for the next quarter, since under U.S. export restrictions, the approval timing for semiconductor export licenses remains uncertain.\n\nKeyBanc analysts said that if the China business based on chips such as the H20 and RTX6000D (B40) were included, it could have brought Nvidia an incremental revenue of $2–3 billion.\n\nCurrently, the market consensus expects Nvidia’s Q3 revenue to be $45.92 billion, with earnings per share of $1.01.\n\nCapacity Expansion Supports Fundamentals\nDespite short-term uncertainty in the Chinese market, Nvidia’s business fundamentals remain strong, providing solid support for its long-term growth.\n\nKeyBanc emphasized in its report that Nvidia’s GPU supply and production capacity are improving significantly, which is the core driver of its continued performance momentum.\n\nData shows that in the quarter ending in July, Nvidia’s GPU supply increased by 40%, and with the ramp-up of the B200, supply in the quarter ending in October is expected to grow another 20%. Meanwhile, the newer and more powerful B300 chips will begin shipping in the October quarter and are expected to account for half of the Blackwell series shipments.\n\nIn addition, server rack production efficiency is also improving.\n\nThe report noted that the manufacturing yield of GB200 racks by server ODMs has approached 85%, and rack shipments by the end of the year are expected to reach 15,000 to 17,000 units. Therefore, the bank raised its full-year GB200 rack shipment forecast from 25,000 to 30,000 units.\n\nWall Street Maintains Optimistic Outlook\nKeyBanc analyst John Vinh, while issuing a cautionary note, still raised Nvidia’s price target from $190 to $215 and maintained an “Overweight” rating.\n\nSusquehanna analyst Christopher Rolland also saw continued momentum in Nvidia’s data center business, raising his price target from $180 to $210 while maintaining a “Positive” rating.\n\nDespite the two Wall Street institutions raising their price targets, market reaction remained relatively cautious. On Wednesday morning, Nvidia’s stock price still fell by about 2.5%.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002400314052935615, "ret_signed_m1d": 0.002400314052935615, "alpha_spy_m1d": -0.0016276677819220264, "alpha_c_m1d": -0.002817853262807146, "bench_spy_m1d": 0.004027981834857641, "bench_c_m1d": 0.005218167315742761, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017202090813873028, "ret_signed_p1d": 0.017202090813873028, "alpha_spy_p1d": 0.0018452042627223886, "alpha_c_p1d": -0.004296652033776827, "bench_spy_p1d": 0.01535688655115064, "bench_c_p1d": 0.021498742847649854, "ret_p1w": 0.029660593796233536, "ret_signed_p1w": 0.029660593796233536, "alpha_spy_p1w": 0.008623758924118707, "alpha_c_p1w": -0.010066698390253093, "bench_spy_p1w": 0.02103683487211483, "bench_c_p1w": 0.03972729218648663, "ret_p1m": 0.00971516232536973, "ret_signed_p1m": 0.00971516232536973, "alpha_spy_p1m": -0.037472465697916846, "alpha_c_p1m": -0.08855943467769833, "bench_spy_p1m": 0.047187628023286576, "bench_c_p1m": 0.09827459700306806, "ret_p3m": 0.03651981761415879, "ret_signed_p3m": 0.03651981761415879, "alpha_spy_p3m": -0.004956192214155664, "alpha_c_p3m": -0.12291116786336209, "bench_spy_p3m": 0.04147600982831445, "bench_c_p3m": 0.15943098547752088, "ret_p6m": 0.04486514117947604, "ret_signed_p6m": 0.04486514117947604, "alpha_spy_p6m": -0.03398025463099663, "alpha_c_p6m": -0.37790214879195183, "bench_spy_p6m": 0.07884539581047267, "bench_c_p6m": 0.4227672899714279, "price_path": [-0.2497, -0.2257, -0.2296, -0.2046, -0.2278, -0.2149, -0.193, -0.189, -0.2, -0.1901, -0.1849, -0.1773, -0.1837, -0.1713, -0.1887, -0.1731, -0.1764, -0.1686, -0.1686, -0.1779, -0.1761, -0.1548, -0.1181, -0.1141, -0.0985, -0.0971, -0.1239, -0.1013, -0.0894, -0.0894, -0.0957, -0.0856, -0.0692, -0.0622, -0.0575, -0.0623, -0.0245, -0.0206, -0.0113, -0.0147, -0.0206, -0.0454, -0.024, -0.0071, -0.0085, 0.0101, 0.003, 0.0245, 0.0165, -0.0072, 0.0287, 0.0187, 0.0254, 0.0331, 0.0441, 0.0405, 0.0467, 0.0378, 0.0402, 0.0313, 0.0402, 0.0038, 0.0024, 0.0, 0.0172, 0.0276, 0.0388, 0.0378, 0.0297, -0.0046, -0.0046, -0.024, -0.0249, -0.019, -0.0455, -0.0381, -0.0241, 0.0134, 0.0126, 0.0163, 0.0159, -0.0005, -0.0267, 0.0073, 0.0097, 0.0494, 0.0198, 0.0114, 0.0155, 0.0184, 0.0393, 0.0664, 0.0701, 0.0796, 0.0723, 0.0604, 0.0576, 0.0808, 0.1006, 0.0468, 0.0763, 0.0289, 0.0278, 0.0391, 0.0472, 0.0438, 0.0354, 0.0303, 0.0411, 0.0645, 0.0944, 0.1489, 0.1833, 0.1596, 0.1573, 0.1824, 0.1356, 0.1157, 0.0749, 0.0753, 0.1376, 0.104, 0.1076, 0.068, 0.0869, 0.0665, 0.0365, 0.066, 0.0324, 0.0223, 0.0433, 0.0163, 0.0302, 0.0302, 0.0116, 0.0283, 0.0371, 0.0264, 0.0481, 0.0426, 0.0605, 0.0572, 0.0504, 0.0341, 0.0003, 0.0076, 0.0158, -0.023, -0.0047, 0.0345, 0.0499, 0.0814, 0.078, 0.078, 0.089, 0.0758, 0.0719, 0.066, 0.066, 0.0794, 0.0752, 0.0702, 0.0809, 0.0576, 0.0566, 0.057, 0.062, 0.0468, 0.0691, 0.0644, 0.0644, 0.0178, 0.0478, 0.0565, 0.0726, 0.0658, 0.0775, 0.0946, 0.1003, 0.0924, 0.0609, 0.0307, -0.0044, -0.0176, 0.0597, 0.0862, 0.0776, 0.0862, 0.0685, 0.0449]}
{"unit_id": "orig:1970253062209176029", "ticker_idx": 0, "ts": "2025-09-22T22:25:10+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "For Samsung Electronics and SK Hynix, valuations remain conservative (around 1.0–1.2x P/B), implying greater upside potential relative to risk", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish global memory sector; base-case double-digit upside, bull-case 40%+; Samsung and Hynix at conservative 1.0–1.2x P/B favor upside over downside risk.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Global Memory Risk-Reward\n\n    • Distribution of returns under bullish/bearish scenarios for the memory sector\n    • Even under the base-case scenario, double-digit upside potential is evident, while in the bull case, more than 40% additional returns are possible\n    • Conversely, in the bear case, downside risk is around 20%, but compared with the past, the risk-reward ratio remains favorable\n    • For Samsung Electronics and SK Hynix, valuations remain conservative (around 1.0–1.2x P/B), implying greater upside potential relative to risk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "orig:1970253062209176029", "ticker_idx": 1, "ts": "2025-09-22T22:25:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "For Samsung Electronics and SK Hynix, valuations remain conservative (around 1.0–1.2x P/B), implying greater upside potential relative to risk", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish global memory sector; base-case double-digit upside, bull-case 40%+; Samsung and Hynix at conservative 1.0–1.2x P/B favor upside over downside risk.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Memory Risk-Reward\n\n    • Distribution of returns under bullish/bearish scenarios for the memory sector\n    • Even under the base-case scenario, double-digit upside potential is evident, while in the bull case, more than 40% additional returns are possible\n    • Conversely, in the bear case, downside risk is around 20%, but compared with the past, the risk-reward ratio remains favorable\n    • For Samsung Electronics and SK Hynix, valuations remain conservative (around 1.0–1.2x P/B), implying greater upside potential relative to risk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012820521976864452, "ret_signed_m1d": 0.012820521976864452, "alpha_spy_m1d": 0.01752925290143681, "alpha_c_m1d": 0.032872079731812165, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028489969482429656, "ret_signed_p1d": 0.028489969482429656, "alpha_spy_p1d": 0.033933444322237216, "alpha_c_p1d": 0.029948808796310566, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.0056981188806843885, "ret_signed_p1w": -0.0056981188806843885, "alpha_spy_p1w": -0.0009593796228162255, "alpha_c_p1w": -0.00721900613776616, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.3646722878607491, "ret_signed_p1m": 0.3646722878607491, "alpha_spy_p1m": 0.357999115706531, "alpha_c_p1m": 0.2928779715663701, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5737758511959559, "ret_signed_p3m": 0.5737758511959559, "alpha_spy_p3m": 0.5593346629727354, "alpha_c_p3m": 0.4959287722952972, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.7706148936884407, "ret_signed_p6m": 1.7706148936884407, "alpha_spy_p6m": 1.7617190797680904, "alpha_c_p6m": 1.5348855405453752, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.1864, -0.1664, -0.1921, -0.1693, -0.1878, -0.2063, -0.2077, -0.2305, -0.229, -0.1977, -0.2006, -0.1551, -0.1622, -0.1465, -0.1508, -0.1579, -0.2333, -0.2347, -0.2248, -0.2361, -0.2347, -0.2333, -0.2433, -0.2546, -0.2532, -0.2504, -0.2219, -0.266, -0.266, -0.2504, -0.2646, -0.2546, -0.2703, -0.2404, -0.2347, -0.2091, -0.2134, -0.2134, -0.239, -0.2518, -0.2731, -0.303, -0.2859, -0.2618, -0.2561, -0.2603, -0.235, -0.2336, -0.2707, -0.2578, -0.2521, -0.2436, -0.2208, -0.2108, -0.1795, -0.1339, -0.1254, -0.0641, -0.057, -0.0085, -0.0499, 0.0128, 0.0128, 0.0, 0.0285, 0.0185, 0.0157, -0.0413, -0.0057, -0.01, 0.0256, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.2194, 0.1823, 0.1724, 0.2037, 0.2892, 0.3262, 0.3832, 0.3647, 0.3718, 0.3632, 0.453, 0.5242, 0.4843, 0.5897, 0.6182, 0.5926, 0.7664, 0.6695, 0.6496, 0.6895, 0.6524, 0.7265, 0.7635, 0.7578, 0.7436, 0.5954, 0.7265, 0.6239, 0.6011, 0.6268, 0.4843, 0.4815, 0.4786, 0.4929, 0.551, 0.5111, 0.5339, 0.5909, 0.5738, 0.5453, 0.551, 0.6451, 0.6137, 0.6736, 0.6108, 0.6279, 0.5795, 0.5111, 0.5709, 0.5738, 0.5595, 0.6536, 0.665, 0.6764, 0.6764, 0.7078, 0.8247, 0.856, 0.856, 0.856, 0.9302, 0.9843, 1.0699, 1.1155, 1.1554, 1.1212, 1.1354, 1.1041, 1.1155, 1.1354, 1.1554, 1.1782, 1.1183, 1.1098, 1.1525, 1.1868, 1.0984, 1.2808, 1.3977, 1.4547, 1.5916, 1.3664, 1.5859, 1.5659, 1.4006, 1.392, 1.5289, 1.4975, 1.4519, 1.5317, 1.5089, 1.5089, 1.5089, 1.5089, 1.5488, 1.7056, 1.7113, 1.8653, 1.9024, 2.1391, 2.0305, 2.0305, 1.6821, 1.425, 1.6878, 1.6392, 1.3879, 1.6792, 1.7278, 1.6564, 1.5992, 1.782, 1.7706]}
{"unit_id": "orig:1970253062209176029", "ticker_idx": 2, "ts": "2025-09-22T22:25:10+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Even under the base-case scenario, double-digit upside potential is evident, while in the bull case, more than 40% additional returns are possible", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish global memory sector; base-case double-digit upside, bull-case 40%+; Samsung and Hynix at conservative 1.0–1.2x P/B favor upside over downside risk.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Memory Risk-Reward\n\n    • Distribution of returns under bullish/bearish scenarios for the memory sector\n    • Even under the base-case scenario, double-digit upside potential is evident, while in the bull case, more than 40% additional returns are possible\n    • Conversely, in the bear case, downside risk is around 20%, but compared with the past, the risk-reward ratio remains favorable\n    • For Samsung Electronics and SK Hynix, valuations remain conservative (around 1.0–1.2x P/B), implying greater upside potential relative to risk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012327960751177628, "ret_signed_m1d": -0.012327960751177628, "alpha_spy_m1d": -0.00761922982660527, "alpha_c_m1d": 0.007723597003770085, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01791158103942525, "ret_signed_p1d": 0.01791158103942525, "alpha_spy_p1d": 0.02335505587923281, "alpha_c_p1d": 0.01937042035330616, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.0009367612410294823, "ret_signed_p1w": 0.0009367612410294823, "alpha_spy_p1w": 0.005675500498897645, "alpha_c_p1w": -0.0005841260160522888, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.25571535433485715, "ret_signed_p1m": 0.25571535433485715, "alpha_spy_p1m": 0.249042182180639, "alpha_c_p1m": 0.1839210380404781, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.4596447512769175, "ret_signed_p3m": 0.4596447512769175, "alpha_spy_p3m": 0.445203563053697, "alpha_c_p3m": 0.38179767237625883, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.6406466633660866, "ret_signed_p6m": 1.6406466633660866, "alpha_spy_p6m": 1.6317508494457362, "alpha_c_p6m": 1.404917310223021, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2281, -0.2284, -0.2356, -0.235, -0.2444, -0.2464, -0.2338, -0.2434, -0.2539, -0.2355, -0.2449, -0.2256, -0.2187, -0.2258, -0.2194, -0.2253, -0.2488, -0.2454, -0.2417, -0.2608, -0.2574, -0.2547, -0.2594, -0.2452, -0.2425, -0.228, -0.2346, -0.2679, -0.2589, -0.2502, -0.2599, -0.2436, -0.2294, -0.2129, -0.2024, -0.1977, -0.1983, -0.2072, -0.2167, -0.224, -0.2389, -0.2514, -0.2387, -0.2328, -0.2355, -0.2332, -0.2201, -0.2253, -0.246, -0.2369, -0.2317, -0.2165, -0.1968, -0.1909, -0.1672, -0.138, -0.1106, -0.0687, -0.0608, -0.031, -0.0472, 0.0001, -0.0123, 0.0, 0.0179, 0.0079, -0.0002, -0.0295, 0.0009, 0.0052, 0.0556, 0.1075, 0.116, 0.1224, 0.1117, 0.1337, 0.1252, 0.1529, 0.1588, 0.1371, 0.1711, 0.2318, 0.2447, 0.2734, 0.2557, 0.2548, 0.2602, 0.3243, 0.363, 0.3434, 0.3921, 0.4107, 0.4153, 0.5098, 0.4189, 0.4345, 0.4438, 0.4254, 0.4921, 0.4914, 0.4956, 0.4735, 0.4217, 0.4691, 0.3964, 0.3784, 0.3537, 0.2951, 0.3353, 0.346, 0.3764, 0.3985, 0.3858, 0.4027, 0.4301, 0.4181, 0.3957, 0.4323, 0.4877, 0.484, 0.5252, 0.4909, 0.4679, 0.4279, 0.387, 0.4132, 0.4596, 0.4849, 0.5547, 0.5619, 0.5851, 0.5851, 0.6154, 0.6864, 0.6949, 0.6803, 0.6803, 0.8005, 0.8506, 0.9457, 0.9615, 0.9406, 0.9666, 0.9721, 0.9412, 0.9462, 0.974, 1.0537, 1.0629, 1.031, 1.0943, 1.1371, 1.1518, 1.101, 1.2345, 1.3359, 1.3491, 1.3515, 1.2822, 1.3871, 1.3057, 1.2182, 1.2364, 1.2908, 1.2571, 1.3252, 1.4026, 1.4008, 1.4008, 1.3768, 1.4197, 1.4611, 1.5357, 1.5347, 1.6082, 1.657, 1.7671, 1.7184, 1.719, 1.4497, 1.3145, 1.4727, 1.3886, 1.2842, 1.4673, 1.5235, 1.4642, 1.4695, 1.5832, 1.6406]}
{"unit_id": "orig:1945589433857868097", "ticker_idx": 0, "ts": "2025-07-16T21:00:42+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi Remains Bullish on KLAC and AMAT, Memory Cycle Recovery and Reduced China Risk as Catalysts", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates bullish stance on KLAC and adds AMAT to key recommendations on memory recovery, reduced China risk, and attractive valuation.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Equipment Capex Outlook: 2025 WFE Spending Flat, Memory Recovery & AI Cloud Capex Drive Optimistic Revisions\n\nCiti (A. Malik, 25/07/15)\n\nCiti states that global Wafer Fab Equipment (WFE) spending for 2025 is expected to remain at the $100 billion level, largely flat compared to last year. Among this, the three major manufacturers—TSMC, Samsung, and Intel—will account for approximately 60%. Driven by improvements in memory (especially NAND) and an increase in AI cloud capital expenditure, Citi has revised its 2025/2026 WFE outlook upwards and maintains a positive view on equipment stocks.\n\nTSMC Maintains $40 Billion Capex Estimate, Still Potential for Upward Revision in September\n\nTSMC is scheduled to release its financial report on July 17. In its April conference call, the company provided a 2025 capital expenditure range of $38 billion to $42 billion, with 70% allocated to advanced processes, 10-20% to specialty processes, and another 10-20% for advanced packaging, testing, and mask manufacturing. Citi anticipates that TSMC may still raise its Capex in September, thus maintaining a $40 billion assumption in its WFE model, slightly below the market consensus of $43 billion.\n\nIntel Continues to Cut Capital Expenditures, Citi's Estimate Significantly Below Market Consensus\n\nIntel will release its financial report on July 24. In its April conference call, the company lowered its 2025 capital expenditure guidance to $18 billion, reflecting improved utilization of construction assets and enhanced operational efficiency. Citi believes there is still room for further reduction, thus assuming only $14 billion in spending for 2025 in its model, significantly lower than the market consensus of $19 billion.\n\nSamsung Capital Expenditure Expected to Decline 10% Year-on-Year, Below Market Expectations\n\nSamsung will release its financial report on July 30. Earlier this year, management stated that 2025 memory investments would be flat compared to 2024. Citi's model estimates Samsung's 2025 capital expenditure at $31 billion, a 10% year-on-year decrease, while market expectations are $38 billion (a 2% year-on-year increase when measured in Korean Won).\n\nCiti Remains Bullish on KLAC and AMAT, Memory Cycle Recovery and Reduced China Risk as Catalysts\n\nCiti notes that KLAC has a higher proportion of advanced process exposure, benefiting from increased spending by leading foundries. The firm has also added AMAT to its list of key U.S. recommendations due to a lower base in the second half of the year, a recovery in DRAM, reduced China business risk, and relatively attractive valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0032566355701806327, "ret_signed_m1d": 0.0032566355701806327, "alpha_spy_m1d": 0.006588687003936, "alpha_c_m1d": -0.001997804163132111, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0038457240873108756, "ret_signed_p1d": 0.0038457240873108756, "alpha_spy_p1d": -0.002273965890206675, "alpha_c_p1d": -0.004554416318921062, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0389934877817435, "ret_signed_p1w": -0.0389934877817435, "alpha_spy_p1w": -0.0549976095819249, "alpha_c_p1w": -0.029763693426924864, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.023481744770687873, "ret_signed_p1m": 0.023481744770687873, "alpha_spy_p1m": -0.009727737341882703, "alpha_c_p1m": -0.020212975169780156, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.1004196531826902, "ret_signed_p3m": 0.1004196531826902, "alpha_spy_p3m": 0.03528515233466423, "alpha_c_p3m": -0.0740480005252544, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.4244501215356995, "ret_signed_p6m": 0.4244501215356995, "alpha_spy_p6m": 0.3135200540880603, "alpha_c_p6m": 0.11029339969316498, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.322, -0.3366, -0.3199, -0.2978, -0.2632, -0.2577, -0.2586, -0.2648, -0.2491, -0.2772, -0.2539, -0.2591, -0.2739, -0.258, -0.2484, -0.2506, -0.1873, -0.1494, -0.1397, -0.1386, -0.1545, -0.1569, -0.1534, -0.1656, -0.1778, -0.1889, -0.1889, -0.1547, -0.1671, -0.1709, -0.1892, -0.1832, -0.1689, -0.1622, -0.1517, -0.1344, -0.1116, -0.0824, -0.0659, -0.0627, -0.0705, -0.044, -0.0434, -0.0668, -0.0668, -0.0894, -0.0827, -0.0476, -0.0429, -0.0327, -0.0467, -0.0404, -0.0371, -0.0133, -0.0095, -0.0095, -0.0224, -0.0153, -0.011, -0.0052, -0.0095, -0.0125, 0.0033, 0.0, 0.0038, -0.0025, 0.0046, -0.0442, -0.039, -0.0314, -0.0336, -0.0111, -0.0186, -0.0091, -0.0583, -0.0502, -0.0191, -0.0536, -0.0484, -0.023, -0.02, -0.025, 0.0022, 0.0171, 0.0235, -0.0627, -0.0516, -0.0595, -0.0569, -0.0634, -0.0657, -0.0557, -0.0466, -0.0457, -0.0402, -0.0638, -0.0638, -0.0914, -0.094, -0.0625, -0.0283, -0.0241, -0.0147, 0.0013, 0.0299, 0.035, 0.0617, 0.0635, 0.0627, 0.1237, 0.1217, 0.1501, 0.15, 0.1473, 0.137, 0.1426, 0.1424, 0.158, 0.2119, 0.2231, 0.1826, 0.2236, 0.1646, 0.1408, 0.131, 0.0551, 0.1004, 0.1012, 0.167, 0.1796, 0.1881, 0.2377, 0.2319, 0.1963, 0.2443, 0.2699, 0.3045, 0.2948, 0.3262, 0.3038, 0.2977, 0.3088, 0.2813, 0.3174, 0.2952, 0.2812, 0.3076, 0.2785, 0.2872, 0.2472, 0.2178, 0.2192, 0.2078, 0.2555, 0.1856, 0.1798, 0.2224, 0.2323, 0.2464, 0.2464, 0.2641, 0.2444, 0.2796, 0.3031, 0.2991, 0.306, 0.3169, 0.318, 0.3323, 0.3401, 0.2839, 0.3175, 0.3156, 0.2604, 0.3145, 0.3396, 0.3611, 0.3644, 0.3733, 0.3733, 0.3761, 0.3554, 0.3374, 0.3067, 0.3067, 0.3705, 0.4544, 0.5002, 0.4622, 0.4245]}
{"unit_id": "orig:1945589433857868097", "ticker_idx": 1, "ts": "2025-07-16T21:00:42+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "added AMAT to its list of key U.S. recommendations due to a lower base in the second half of the year, a recovery in DRAM, reduced China business risk, and relatively attractive valuation", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates bullish stance on KLAC and adds AMAT to key recommendations on memory recovery, reduced China risk, and attractive valuation.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Equipment Capex Outlook: 2025 WFE Spending Flat, Memory Recovery & AI Cloud Capex Drive Optimistic Revisions\n\nCiti (A. Malik, 25/07/15)\n\nCiti states that global Wafer Fab Equipment (WFE) spending for 2025 is expected to remain at the $100 billion level, largely flat compared to last year. Among this, the three major manufacturers—TSMC, Samsung, and Intel—will account for approximately 60%. Driven by improvements in memory (especially NAND) and an increase in AI cloud capital expenditure, Citi has revised its 2025/2026 WFE outlook upwards and maintains a positive view on equipment stocks.\n\nTSMC Maintains $40 Billion Capex Estimate, Still Potential for Upward Revision in September\n\nTSMC is scheduled to release its financial report on July 17. In its April conference call, the company provided a 2025 capital expenditure range of $38 billion to $42 billion, with 70% allocated to advanced processes, 10-20% to specialty processes, and another 10-20% for advanced packaging, testing, and mask manufacturing. Citi anticipates that TSMC may still raise its Capex in September, thus maintaining a $40 billion assumption in its WFE model, slightly below the market consensus of $43 billion.\n\nIntel Continues to Cut Capital Expenditures, Citi's Estimate Significantly Below Market Consensus\n\nIntel will release its financial report on July 24. In its April conference call, the company lowered its 2025 capital expenditure guidance to $18 billion, reflecting improved utilization of construction assets and enhanced operational efficiency. Citi believes there is still room for further reduction, thus assuming only $14 billion in spending for 2025 in its model, significantly lower than the market consensus of $19 billion.\n\nSamsung Capital Expenditure Expected to Decline 10% Year-on-Year, Below Market Expectations\n\nSamsung will release its financial report on July 30. Earlier this year, management stated that 2025 memory investments would be flat compared to 2024. Citi's model estimates Samsung's 2025 capital expenditure at $31 billion, a 10% year-on-year decrease, while market expectations are $38 billion (a 2% year-on-year increase when measured in Korean Won).\n\nCiti Remains Bullish on KLAC and AMAT, Memory Cycle Recovery and Reduced China Risk as Catalysts\n\nCiti notes that KLAC has a higher proportion of advanced process exposure, benefiting from increased spending by leading foundries. The firm has also added AMAT to its list of key U.S. recommendations due to a lower base in the second half of the year, a recovery in DRAM, reduced China business risk, and relatively attractive valuation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.022996752272785637, "ret_signed_m1d": 0.022996752272785637, "alpha_spy_m1d": 0.026328803706541004, "alpha_c_m1d": 0.017742312539472893, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01175508087599797, "ret_signed_p1d": -0.01175508087599797, "alpha_spy_p1d": -0.01787477085351552, "alpha_c_p1d": -0.020155221282229907, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.04003907498240111, "ret_signed_p1w": -0.04003907498240111, "alpha_spy_p1w": -0.056043196782582516, "alpha_c_p1w": -0.03080928062758248, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": -0.03372517199257008, "ret_signed_p1m": -0.03372517199257008, "alpha_spy_p1m": -0.06693465410514066, "alpha_c_p1m": -0.07741989193303811, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.1298651406562079, "ret_signed_p3m": 0.1298651406562079, "alpha_spy_p3m": 0.06473063980818194, "alpha_c_p3m": -0.04460251305173668, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.4527018657112689, "ret_signed_p6m": 0.4527018657112689, "alpha_spy_p6m": 0.3417717982636297, "alpha_c_p6m": 0.1385451438687344, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.2964, -0.3061, -0.2921, -0.2651, -0.2313, -0.2243, -0.2282, -0.2345, -0.2286, -0.2375, -0.2061, -0.2086, -0.2167, -0.2025, -0.2015, -0.2035, -0.1401, -0.1143, -0.1086, -0.1055, -0.1525, -0.1506, -0.1517, -0.1696, -0.176, -0.1915, -0.1915, -0.1693, -0.1705, -0.1814, -0.1954, -0.1927, -0.1698, -0.1688, -0.1572, -0.1441, -0.1284, -0.108, -0.113, -0.1017, -0.1243, -0.0937, -0.1064, -0.1128, -0.1128, -0.1301, -0.1173, -0.0751, -0.0603, -0.058, -0.0595, -0.0603, -0.0567, -0.0246, -0.0193, -0.0193, -0.0207, 0.0009, 0.003, 0.0165, 0.016, 0.0118, 0.023, 0.0, -0.0118, -0.0224, -0.0113, -0.0394, -0.04, -0.0343, -0.0468, -0.0233, -0.0329, -0.0278, -0.0757, -0.0761, -0.0615, -0.0804, -0.0856, -0.0599, -0.051, -0.0535, -0.0326, -0.0245, -0.0337, -0.1697, -0.1606, -0.1673, -0.1738, -0.1772, -0.1635, -0.1661, -0.1531, -0.1537, -0.1492, -0.1724, -0.1724, -0.1888, -0.1956, -0.1854, -0.1622, -0.1658, -0.1583, -0.1587, -0.1241, -0.1362, -0.1201, -0.1066, -0.083, -0.0231, -0.0214, 0.0323, 0.0341, 0.037, 0.0275, 0.0498, 0.0551, 0.054, 0.1209, 0.151, 0.1198, 0.1527, 0.0891, 0.1197, 0.1341, 0.0808, 0.1299, 0.1232, 0.1716, 0.1723, 0.1582, 0.1744, 0.1634, 0.1354, 0.1761, 0.1776, 0.1909, 0.1719, 0.2136, 0.1971, 0.2, 0.2237, 0.185, 0.2401, 0.2022, 0.1844, 0.2102, 0.1772, 0.1878, 0.1492, 0.1635, 0.1774, 0.1589, 0.2104, 0.1359, 0.1554, 0.191, 0.2506, 0.2893, 0.2893, 0.3011, 0.314, 0.3686, 0.3856, 0.3898, 0.3823, 0.3832, 0.3779, 0.4192, 0.3932, 0.337, 0.3476, 0.3351, 0.2806, 0.3076, 0.3226, 0.336, 0.3423, 0.3451, 0.3451, 0.3509, 0.3568, 0.3409, 0.3256, 0.3256, 0.3868, 0.4665, 0.5268, 0.5072, 0.4527]}
{"unit_id": "orig:1959892931613229204", "ticker_idx": 0, "ts": "2025-08-25T08:17:42+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe near-term upside catalysts remain limited", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman post-D/G: limited near-term upside for Hynix, low expectations for Samsung HBM; some rotation toward NVDA/TSMC as cleaner AI plays.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_South Korea Technology: Investor feedback post our Hynix D/G; HBM pricing the main pushback, while Hynix vs. Samsung now a debate - Goldman Sachs\n\n[Key Takeaways]\n\n1. The biggest difference between investors and our view was on SKH’s ’26 HBM blended ASP.\n\n2. Unlike in the past, there is no longer an overwhelming preference for SKH over SEC; instead, the two are now being compared side by side.\n\n3. However, the attention lost from SKH has not fully shifted to SEC, with some investors instead turning their eyes to TSMC/NVIDIA as alternatives.\n\n[Contents]\n\n1. After our downgrade report on SKH, we held more than 150 investor meetings in the U.S. and Asia over several weeks.\n\n2. This report summarizes the feedback we received in the process.\n\n3. The area where investors most pushed back on our view was that SKH’s ’26 HBM blended ASP would remain positive, contrary to our stance.\n\n4. In other words, most investors expect SKH’s earnings growth to be strong next year.\n\n5. This is based on: 1) low confidence that SEC’s HBM3E 12hi and HBM4 will be supplied on time, and 2) a positive outlook on the legacy market.\n\n6. Compared with several months ago, investors still showed a preference for SKH over SEC, but instead of a clear dominance, the debate has shifted to which of the two to choose.\n\n7. A majority of investors believe SKH’s HBM pricing could remain flat or even rise slightly.\n\n8. This is a stark contrast to our projection of a double-digit decline.\n\n9. Most investors agreed that HBM3E 12hi pricing would decline next year due to new competitors entering the market.\n\n10. However, they saw the decline as only about -10%, not the -30% level we modeled.\n\n11. Few investors believed that SEC/MU would supply HBM4 to NVIDIA in a meaningful volume next year.\n\n12. As a result, investors held a positive view that strong pricing momentum in HBM4 would support SKH’s HBM ASP.\n\n13. At the same time, there was awareness across investors that costs would rise significantly in the transition from HBM3E to HBM4\n\n* due to increased I/O counts, lower yields, outsourcing costs for base die, etc.\n\n14. However, they believed the price premium would offset those costs, taking a much more optimistic stance than our scenario of HBM OPM falling more than -10ppt.\n\n15. In short, only a minority of investors agreed with our base scenario that SKH’s operating profit would decline YoY next year.\n\n16. As for SEC, investor perception of its HBM qualification progress improved compared with 1–2 quarters ago, but confidence still remained low.\n\n17. Many investors now see HBM3E 12hi as less important, with the key factor being the timing of HBM4 qualification.\n\n18. Overall, SEC was evaluated slightly more favorably than 1–2 quarters ago, but expectations remain low.\n\n19. Lastly, with memory spot prices continuing to show strength and major players delivering better-than-expected fundamentals, investors were positive on legacy semiconductors for 2H25.\n\n20. While most investors still preferred SKH, the preference was no longer overwhelming, and the discussion has now shifted to a debate with SEC.\n\n21. Some investors saw valuation attractiveness in SKH due to the recent share price correction, but we believe near-term upside catalysts remain limited.\n\n22. Regarding SEC, the number of questions in this round of meetings rose significantly, but the interest lost from SKH did not flow directly into SEC.\n\n23. Instead, some investors looked toward cleaner AI beneficiaries such as NVIDIA and TSMC.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032755225139682875, "ret_signed_m1d": 0.032755225139682875, "alpha_spy_m1d": 0.03717583800882929, "alpha_c_m1d": 0.031802541252007765, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007707168704449874, "ret_signed_p1d": -0.007707168704449874, "alpha_spy_p1d": -0.003520132354559058, "alpha_c_p1d": 0.0019212189632149546, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.01206254011757013, "ret_signed_p1w": 0.01206254011757013, "alpha_spy_p1w": 0.016078370052403534, "alpha_c_p1w": -0.00028777175699346014, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.39314614138507986, "ret_signed_p1m": -0.39314614138507986, "alpha_spy_p1m": -0.35800236009164643, "alpha_c_p1m": -0.29863067180680636, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 1.2035636253174942, "ret_signed_p3m": -1.2035636253174942, "alpha_spy_p3m": -1.1850893321954412, "alpha_c_p3m": -1.0974804465466281, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 2.39846717626965, "ret_signed_p6m": -2.39846717626965, "alpha_spy_p6m": -2.3295199481800712, "alpha_c_p6m": -2.0076874315351896, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1999, -0.183, -0.2119, -0.2004, -0.2004, -0.1618, -0.1349, -0.1349, -0.1175, -0.1118, -0.0751, -0.0925, -0.0925, -0.0443, -0.0405, -0.0501, -0.052, -0.0096, 0.0, 0.0732, 0.1021, 0.1291, 0.0944, 0.1252, 0.1002, 0.0751, 0.0732, 0.0424, 0.0443, 0.0867, 0.0829, 0.1445, 0.1349, 0.1561, 0.1503, 0.1407, 0.0385, 0.0366, 0.0501, 0.0347, 0.0366, 0.0385, 0.025, 0.0096, 0.0116, 0.0154, 0.0539, -0.0058, -0.0058, 0.0154, -0.0039, 0.0096, -0.0116, 0.0289, 0.0366, 0.0713, 0.0655, 0.0655, 0.0308, 0.0135, -0.0154, -0.0559, -0.0328, 0.0, 0.0077, 0.0019, 0.0362, 0.0381, -0.0121, 0.0053, 0.013, 0.0246, 0.0555, 0.069, 0.1114, 0.1732, 0.1848, 0.2677, 0.2774, 0.343, 0.287, 0.3719, 0.3719, 0.3546, 0.3931, 0.3796, 0.3758, 0.2986, 0.3468, 0.341, 0.3893, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.6517, 0.6015, 0.588, 0.6305, 0.7463, 0.7964, 0.8736, 0.8485, 0.8582, 0.8466, 0.9682, 1.0646, 1.0106, 1.1534, 1.192, 1.1573, 1.3927, 1.2615, 1.2344, 1.2885, 1.2383, 1.3386, 1.3888, 1.3811, 1.3618, 1.1611, 1.3386, 1.1997, 1.1688, 1.2036, 1.0106, 1.0067, 1.0029, 1.0222, 1.1009, 1.0468, 1.0777, 1.1549, 1.1318, 1.0931, 1.1009, 1.2283, 1.1858, 1.2669, 1.182, 1.2051, 1.1395, 1.0468, 1.1279, 1.1318, 1.1125, 1.2399, 1.2553, 1.2708, 1.2708, 1.3133, 1.4716, 1.5141, 1.5141, 1.5141, 1.6145, 1.6879, 1.8037, 1.8655, 1.9196, 1.8732, 1.8926, 1.8501, 1.8655, 1.8926, 1.9196, 1.9505, 1.8694, 1.8578, 1.9157, 1.9621, 1.8424, 2.0895, 2.2479, 2.3251, 2.5105, 2.2054, 2.5027, 2.4757, 2.2517, 2.2401, 2.4255, 2.383, 2.3212, 2.4294, 2.3985, 2.3985, 2.3985]}
{"unit_id": "orig:1959892931613229204", "ticker_idx": 1, "ts": "2025-08-25T08:17:42+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "overall, SEC was evaluated slightly more favorably than 1–2 quarters ago, but expectations remain low", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman post-D/G: limited near-term upside for Hynix, low expectations for Samsung HBM; some rotation toward NVDA/TSMC as cleaner AI plays.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250825_South Korea Technology: Investor feedback post our Hynix D/G; HBM pricing the main pushback, while Hynix vs. Samsung now a debate - Goldman Sachs\n\n[Key Takeaways]\n\n1. The biggest difference between investors and our view was on SKH’s ’26 HBM blended ASP.\n\n2. Unlike in the past, there is no longer an overwhelming preference for SKH over SEC; instead, the two are now being compared side by side.\n\n3. However, the attention lost from SKH has not fully shifted to SEC, with some investors instead turning their eyes to TSMC/NVIDIA as alternatives.\n\n[Contents]\n\n1. After our downgrade report on SKH, we held more than 150 investor meetings in the U.S. and Asia over several weeks.\n\n2. This report summarizes the feedback we received in the process.\n\n3. The area where investors most pushed back on our view was that SKH’s ’26 HBM blended ASP would remain positive, contrary to our stance.\n\n4. In other words, most investors expect SKH’s earnings growth to be strong next year.\n\n5. This is based on: 1) low confidence that SEC’s HBM3E 12hi and HBM4 will be supplied on time, and 2) a positive outlook on the legacy market.\n\n6. Compared with several months ago, investors still showed a preference for SKH over SEC, but instead of a clear dominance, the debate has shifted to which of the two to choose.\n\n7. A majority of investors believe SKH’s HBM pricing could remain flat or even rise slightly.\n\n8. This is a stark contrast to our projection of a double-digit decline.\n\n9. Most investors agreed that HBM3E 12hi pricing would decline next year due to new competitors entering the market.\n\n10. However, they saw the decline as only about -10%, not the -30% level we modeled.\n\n11. Few investors believed that SEC/MU would supply HBM4 to NVIDIA in a meaningful volume next year.\n\n12. As a result, investors held a positive view that strong pricing momentum in HBM4 would support SKH’s HBM ASP.\n\n13. At the same time, there was awareness across investors that costs would rise significantly in the transition from HBM3E to HBM4\n\n* due to increased I/O counts, lower yields, outsourcing costs for base die, etc.\n\n14. However, they believed the price premium would offset those costs, taking a much more optimistic stance than our scenario of HBM OPM falling more than -10ppt.\n\n15. In short, only a minority of investors agreed with our base scenario that SKH’s operating profit would decline YoY next year.\n\n16. As for SEC, investor perception of its HBM qualification progress improved compared with 1–2 quarters ago, but confidence still remained low.\n\n17. Many investors now see HBM3E 12hi as less important, with the key factor being the timing of HBM4 qualification.\n\n18. Overall, SEC was evaluated slightly more favorably than 1–2 quarters ago, but expectations remain low.\n\n19. Lastly, with memory spot prices continuing to show strength and major players delivering better-than-expected fundamentals, investors were positive on legacy semiconductors for 2H25.\n\n20. While most investors still preferred SKH, the preference was no longer overwhelming, and the discussion has now shifted to a debate with SEC.\n\n21. Some investors saw valuation attractiveness in SKH due to the recent share price correction, but we believe near-term upside catalysts remain limited.\n\n22. Regarding SEC, the number of questions in this round of meetings rose significantly, but the interest lost from SKH did not flow directly into SEC.\n\n23. Instead, some investors looked toward cleaner AI beneficiaries such as NVIDIA and TSMC.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0013986782181396595, "ret_signed_m1d": 0.0013986782181396595, "alpha_spy_m1d": 0.0058192910872860715, "alpha_c_m1d": 0.003652151611039045, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": 0.0022534733928993855, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016783254497436784, "ret_signed_p1d": 0.016783254497436784, "alpha_spy_p1d": 0.0209702908473276, "alpha_c_p1d": 0.021366502937378073, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.00458324843994129, "ret_p1w": -0.054545466601639614, "ret_signed_p1w": 0.054545466601639614, "alpha_spy_p1w": 0.05856129653647302, "alpha_c_p1w": 0.056913626529906614, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": 0.0023681599282669996, "ret_p1m": 0.1846153574116849, "ret_signed_p1m": -0.1846153574116849, "alpha_spy_p1m": -0.14947157611825146, "alpha_c_p1m": -0.11300344634898662, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.07161191106269826, "ret_p3m": 0.4132704016547195, "ret_signed_p3m": -0.4132704016547195, "alpha_spy_p3m": -0.39479610853266656, "alpha_c_p3m": -0.37251958401579, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.040750817638929515, "ret_p6m": 1.5579468429337187, "ret_signed_p6m": -1.5579468429337187, "alpha_spy_p6m": -1.4889996148441398, "alpha_c_p6m": -1.489535271343165, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.06841157159055355, "price_path": [-0.2229, -0.2202, -0.2188, -0.2104, -0.2104, -0.1965, -0.1785, -0.1785, -0.1687, -0.1771, -0.1673, -0.1729, -0.1896, -0.2049, -0.1924, -0.1687, -0.1771, -0.1729, -0.1938, -0.159, -0.1479, -0.1632, -0.1497, -0.1636, -0.158, -0.1497, -0.1077, -0.1147, -0.1371, -0.1413, -0.1552, -0.1469, -0.1245, -0.1259, -0.1091, -0.0951, -0.0671, -0.0615, -0.0517, -0.0769, -0.0713, -0.0769, -0.0783, -0.0154, -0.0126, 0.0154, -0.0014, -0.0364, -0.0252, -0.0224, -0.0378, -0.014, 0.0042, -0.007, -0.0056, 0.0056, 0.0014, 0.0014, -0.021, -0.021, -0.014, -0.0126, -0.0014, 0.0, -0.0168, -0.0126, -0.0266, -0.0252, -0.0545, -0.0336, -0.0238, -0.0196, -0.028, -0.0196, 0.0, 0.0154, 0.0266, 0.0545, 0.0699, 0.1105, 0.0937, 0.1231, 0.1231, 0.1678, 0.1846, 0.1944, 0.2042, 0.165, 0.1829, 0.1787, 0.2082, 0.2608, 0.2608, 0.2608, 0.2608, 0.2608, 0.2608, 0.3262, 0.3107, 0.2868, 0.3346, 0.3725, 0.3753, 0.3781, 0.3697, 0.3852, 0.3557, 0.388, 0.4329, 0.3978, 0.4119, 0.4624, 0.5102, 0.5608, 0.4737, 0.4133, 0.3936, 0.3753, 0.4133, 0.454, 0.4484, 0.4442, 0.3655, 0.4133, 0.3739, 0.3557, 0.4133, 0.3318, 0.3585, 0.395, 0.4442, 0.454, 0.4119, 0.4161, 0.4526, 0.4681, 0.4765, 0.5228, 0.5383, 0.5228, 0.5172, 0.5074, 0.5299, 0.4723, 0.4442, 0.5158, 0.5116, 0.4933, 0.5523, 0.5664, 0.5608, 0.5608, 0.6437, 0.6869, 0.6926, 0.6926, 0.6926, 0.814, 0.9495, 0.9608, 0.9905, 0.9594, 0.9622, 0.9594, 0.9425, 0.9806, 1.0314, 1.102, 1.1076, 1.0497, 1.1104, 1.15, 1.1472, 1.1472, 1.2516, 1.2926, 1.2686, 1.2657, 1.1232, 1.3645, 1.3871, 1.2488, 1.2389, 1.349, 1.3405, 1.3688, 1.5212, 1.5579, 1.5579, 1.5579]}
{"unit_id": "orig:1959892931613229204", "ticker_idx": 2, "ts": "2025-08-25T08:17:42+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "some investors looked toward cleaner AI beneficiaries such as NVIDIA and TSMC", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman post-D/G: limited near-term upside for Hynix, low expectations for Samsung HBM; some rotation toward NVDA/TSMC as cleaner AI plays.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_South Korea Technology: Investor feedback post our Hynix D/G; HBM pricing the main pushback, while Hynix vs. Samsung now a debate - Goldman Sachs\n\n[Key Takeaways]\n\n1. The biggest difference between investors and our view was on SKH’s ’26 HBM blended ASP.\n\n2. Unlike in the past, there is no longer an overwhelming preference for SKH over SEC; instead, the two are now being compared side by side.\n\n3. However, the attention lost from SKH has not fully shifted to SEC, with some investors instead turning their eyes to TSMC/NVIDIA as alternatives.\n\n[Contents]\n\n1. After our downgrade report on SKH, we held more than 150 investor meetings in the U.S. and Asia over several weeks.\n\n2. This report summarizes the feedback we received in the process.\n\n3. The area where investors most pushed back on our view was that SKH’s ’26 HBM blended ASP would remain positive, contrary to our stance.\n\n4. In other words, most investors expect SKH’s earnings growth to be strong next year.\n\n5. This is based on: 1) low confidence that SEC’s HBM3E 12hi and HBM4 will be supplied on time, and 2) a positive outlook on the legacy market.\n\n6. Compared with several months ago, investors still showed a preference for SKH over SEC, but instead of a clear dominance, the debate has shifted to which of the two to choose.\n\n7. A majority of investors believe SKH’s HBM pricing could remain flat or even rise slightly.\n\n8. This is a stark contrast to our projection of a double-digit decline.\n\n9. Most investors agreed that HBM3E 12hi pricing would decline next year due to new competitors entering the market.\n\n10. However, they saw the decline as only about -10%, not the -30% level we modeled.\n\n11. Few investors believed that SEC/MU would supply HBM4 to NVIDIA in a meaningful volume next year.\n\n12. As a result, investors held a positive view that strong pricing momentum in HBM4 would support SKH’s HBM ASP.\n\n13. At the same time, there was awareness across investors that costs would rise significantly in the transition from HBM3E to HBM4\n\n* due to increased I/O counts, lower yields, outsourcing costs for base die, etc.\n\n14. However, they believed the price premium would offset those costs, taking a much more optimistic stance than our scenario of HBM OPM falling more than -10ppt.\n\n15. In short, only a minority of investors agreed with our base scenario that SKH’s operating profit would decline YoY next year.\n\n16. As for SEC, investor perception of its HBM qualification progress improved compared with 1–2 quarters ago, but confidence still remained low.\n\n17. Many investors now see HBM3E 12hi as less important, with the key factor being the timing of HBM4 qualification.\n\n18. Overall, SEC was evaluated slightly more favorably than 1–2 quarters ago, but expectations remain low.\n\n19. Lastly, with memory spot prices continuing to show strength and major players delivering better-than-expected fundamentals, investors were positive on legacy semiconductors for 2H25.\n\n20. While most investors still preferred SKH, the preference was no longer overwhelming, and the discussion has now shifted to a debate with SEC.\n\n21. Some investors saw valuation attractiveness in SKH due to the recent share price correction, but we believe near-term upside catalysts remain limited.\n\n22. Regarding SEC, the number of questions in this round of meetings rose significantly, but the interest lost from SKH did not flow directly into SEC.\n\n23. Instead, some investors looked toward cleaner AI beneficiaries such as NVIDIA and TSMC.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010121735824010547, "ret_signed_m1d": -0.010121735824010547, "alpha_spy_m1d": -0.014542348693156959, "alpha_c_m1d": -0.009169051936335437, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01090037658917331, "ret_signed_p1d": 0.01090037658917331, "alpha_spy_p1d": 0.006713340239282495, "alpha_c_p1d": 0.0012719889215084823, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.03131088170414875, "ret_signed_p1w": -0.03131088170414875, "alpha_spy_p1w": -0.035326711638982156, "alpha_c_p1w": -0.01896056982958516, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": -0.007618864936297265, "ret_signed_p1m": -0.007618864936297265, "alpha_spy_p1m": -0.042762646229730694, "alpha_c_p1m": -0.10213433451457077, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.00467260846353823, "ret_signed_p3m": 0.00467260846353823, "alpha_spy_p3m": -0.013801684658514723, "alpha_c_p3m": -0.10141057030732781, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.028812254552892602, "ret_signed_p6m": 0.028812254552892602, "alpha_spy_p6m": -0.04013497353668627, "alpha_c_p6m": -0.36196749018156815, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.2503, -0.226, -0.2485, -0.236, -0.2147, -0.2108, -0.2215, -0.2119, -0.2068, -0.1994, -0.2057, -0.1936, -0.2104, -0.1953, -0.1985, -0.1909, -0.1909, -0.2, -0.1982, -0.1775, -0.1418, -0.1379, -0.1227, -0.1214, -0.1474, -0.1255, -0.1138, -0.1138, -0.12, -0.1102, -0.0942, -0.0874, -0.0828, -0.0875, -0.0507, -0.0469, -0.0379, -0.0412, -0.0469, -0.0711, -0.0502, -0.0338, -0.0351, -0.017, -0.0239, -0.003, -0.0108, -0.0339, 0.0011, -0.0086, -0.0022, 0.0053, 0.0161, 0.0125, 0.0186, 0.0099, 0.0123, 0.0036, 0.0122, -0.0232, -0.0245, -0.0269, -0.0101, 0.0, 0.0109, 0.01, 0.002, -0.0313, -0.0313, -0.0502, -0.0511, -0.0453, -0.0711, -0.064, -0.0503, -0.0138, -0.0146, -0.011, -0.0114, -0.0274, -0.0529, -0.0198, -0.0174, 0.0212, -0.0076, -0.0157, -0.0117, -0.009, 0.0114, 0.0377, 0.0414, 0.0506, 0.0435, 0.0319, 0.0291, 0.0518, 0.071, 0.0187, 0.0474, 0.0013, 0.0002, 0.0112, 0.019, 0.0158, 0.0076, 0.0027, 0.0131, 0.0359, 0.065, 0.1181, 0.1515, 0.1284, 0.1262, 0.1506, 0.1051, 0.0857, 0.0461, 0.0464, 0.1071, 0.0743, 0.0779, 0.0393, 0.0577, 0.0378, 0.0087, 0.0374, 0.0047, -0.0051, 0.0153, -0.011, 0.0026, 0.0026, -0.0156, 0.0007, 0.0092, -0.0012, 0.02, 0.0146, 0.032, 0.0288, 0.0222, 0.0063, -0.0265, -0.0195, -0.0115, -0.0492, -0.0314, 0.0067, 0.0217, 0.0524, 0.0491, 0.0491, 0.0597, 0.0469, 0.0431, 0.0373, 0.0373, 0.0504, 0.0463, 0.0414, 0.0518, 0.0292, 0.0282, 0.0286, 0.0335, 0.0186, 0.0404, 0.0358, 0.0358, -0.0096, 0.0196, 0.0281, 0.0438, 0.0372, 0.0486, 0.0652, 0.0708, 0.0631, 0.0324, 0.0031, -0.0311, -0.044, 0.0313, 0.057, 0.0487, 0.0571, 0.0398, 0.0168, 0.0168, 0.0288]}
{"unit_id": "orig:1959892931613229204", "ticker_idx": 3, "ts": "2025-08-25T08:17:42+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "some investors looked toward cleaner AI beneficiaries such as NVIDIA and TSMC", "tweet_type": "analysis", "timeline": "quarters", "summary": "Goldman post-D/G: limited near-term upside for Hynix, low expectations for Samsung HBM; some rotation toward NVDA/TSMC as cleaner AI plays.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_South Korea Technology: Investor feedback post our Hynix D/G; HBM pricing the main pushback, while Hynix vs. Samsung now a debate - Goldman Sachs\n\n[Key Takeaways]\n\n1. The biggest difference between investors and our view was on SKH’s ’26 HBM blended ASP.\n\n2. Unlike in the past, there is no longer an overwhelming preference for SKH over SEC; instead, the two are now being compared side by side.\n\n3. However, the attention lost from SKH has not fully shifted to SEC, with some investors instead turning their eyes to TSMC/NVIDIA as alternatives.\n\n[Contents]\n\n1. After our downgrade report on SKH, we held more than 150 investor meetings in the U.S. and Asia over several weeks.\n\n2. This report summarizes the feedback we received in the process.\n\n3. The area where investors most pushed back on our view was that SKH’s ’26 HBM blended ASP would remain positive, contrary to our stance.\n\n4. In other words, most investors expect SKH’s earnings growth to be strong next year.\n\n5. This is based on: 1) low confidence that SEC’s HBM3E 12hi and HBM4 will be supplied on time, and 2) a positive outlook on the legacy market.\n\n6. Compared with several months ago, investors still showed a preference for SKH over SEC, but instead of a clear dominance, the debate has shifted to which of the two to choose.\n\n7. A majority of investors believe SKH’s HBM pricing could remain flat or even rise slightly.\n\n8. This is a stark contrast to our projection of a double-digit decline.\n\n9. Most investors agreed that HBM3E 12hi pricing would decline next year due to new competitors entering the market.\n\n10. However, they saw the decline as only about -10%, not the -30% level we modeled.\n\n11. Few investors believed that SEC/MU would supply HBM4 to NVIDIA in a meaningful volume next year.\n\n12. As a result, investors held a positive view that strong pricing momentum in HBM4 would support SKH’s HBM ASP.\n\n13. At the same time, there was awareness across investors that costs would rise significantly in the transition from HBM3E to HBM4\n\n* due to increased I/O counts, lower yields, outsourcing costs for base die, etc.\n\n14. However, they believed the price premium would offset those costs, taking a much more optimistic stance than our scenario of HBM OPM falling more than -10ppt.\n\n15. In short, only a minority of investors agreed with our base scenario that SKH’s operating profit would decline YoY next year.\n\n16. As for SEC, investor perception of its HBM qualification progress improved compared with 1–2 quarters ago, but confidence still remained low.\n\n17. Many investors now see HBM3E 12hi as less important, with the key factor being the timing of HBM4 qualification.\n\n18. Overall, SEC was evaluated slightly more favorably than 1–2 quarters ago, but expectations remain low.\n\n19. Lastly, with memory spot prices continuing to show strength and major players delivering better-than-expected fundamentals, investors were positive on legacy semiconductors for 2H25.\n\n20. While most investors still preferred SKH, the preference was no longer overwhelming, and the discussion has now shifted to a debate with SEC.\n\n21. Some investors saw valuation attractiveness in SKH due to the recent share price correction, but we believe near-term upside catalysts remain limited.\n\n22. Regarding SEC, the number of questions in this round of meetings rose significantly, but the interest lost from SKH did not flow directly into SEC.\n\n23. Instead, some investors looked toward cleaner AI beneficiaries such as NVIDIA and TSMC.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02991460381475941, "ret_signed_m1d": -0.02991460381475941, "alpha_spy_m1d": -0.03433521668390582, "alpha_c_m1d": -0.0289619199270843, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004273499767392774, "ret_signed_p1d": 0.004273499767392774, "alpha_spy_p1d": 8.646341750195852e-05, "alpha_c_p1d": -0.005354887900272054, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.004273499767392774, "ret_signed_p1w": -0.004273499767392774, "alpha_spy_p1w": -0.008289329702226178, "alpha_c_p1w": 0.008076812107170817, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.14988038544903337, "ret_signed_p1m": 0.14988038544903337, "alpha_spy_p1m": 0.11473660415559994, "alpha_c_p1m": 0.05536491587075987, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.2485641825277558, "ret_signed_p3m": 0.2485641825277558, "alpha_spy_p3m": 0.23008988940570285, "alpha_c_p3m": 0.14248100375688977, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.6487768189024732, "ret_signed_p6m": 0.6487768189024732, "alpha_spy_p6m": 0.5798295908128943, "alpha_c_p6m": 0.25799707416801243, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.177, -0.177, -0.177, -0.1949, -0.1915, -0.1574, -0.1506, -0.1532, -0.1447, -0.1106, -0.0936, -0.1068, -0.1197, -0.1239, -0.1068, -0.0983, -0.1154, -0.0983, -0.1282, -0.1026, -0.0855, -0.0812, -0.0769, -0.094, -0.0726, -0.0726, -0.0684, -0.0726, -0.0769, -0.0769, -0.0684, -0.0598, -0.0598, -0.0641, -0.0513, -0.0342, -0.0342, -0.0128, -0.0171, -0.0342, -0.0214, -0.0214, -0.0214, -0.0214, -0.0299, -0.0128, -0.0085, -0.0085, -0.0299, -0.0171, -0.0385, 0.0085, 0.0043, 0.0085, 0.0085, 0.0256, 0.0043, 0.0085, 0.0085, 0.0128, -0.0299, -0.0171, -0.0299, 0.0, 0.0043, 0.0171, -0.0085, -0.0085, -0.0043, -0.0085, -0.0085, -0.0085, 0.0085, 0.0085, 0.0256, 0.047, 0.0598, 0.0769, 0.0726, 0.0984, 0.0855, 0.1027, 0.0855, 0.1113, 0.1499, 0.1499, 0.1327, 0.1156, 0.1156, 0.1198, 0.137, 0.1713, 0.2014, 0.2014, 0.2314, 0.2142, 0.2357, 0.2357, 0.2142, 0.2228, 0.2571, 0.2743, 0.2443, 0.27, 0.27, 0.2529, 0.2443, 0.2443, 0.27, 0.2657, 0.2915, 0.2915, 0.2872, 0.2958, 0.2915, 0.2529, 0.2571, 0.2529, 0.2657, 0.2571, 0.2657, 0.2529, 0.2271, 0.24, 0.2057, 0.1971, 0.2486, 0.1885, 0.1799, 0.2142, 0.2357, 0.2314, 0.2357, 0.2099, 0.2271, 0.2443, 0.24, 0.2529, 0.2829, 0.27, 0.2915, 0.2656, 0.2743, 0.2484, 0.2355, 0.2312, 0.2312, 0.2312, 0.2613, 0.2829, 0.2872, 0.2872, 0.3001, 0.3173, 0.3087, 0.3345, 0.3345, 0.3647, 0.4378, 0.468, 0.4421, 0.4508, 0.4464, 0.4551, 0.4723, 0.4723, 0.4551, 0.4981, 0.5153, 0.5282, 0.4981, 0.5153, 0.5239, 0.511, 0.5325, 0.567, 0.5541, 0.5282, 0.5196, 0.5498, 0.5368, 0.5196, 0.5325, 0.5627, 0.6186, 0.6488, 0.6488, 0.6488, 0.6488, 0.6488]}
{"unit_id": "thread:1969945634062193071", "ticker_idx": 0, "ts": "2025-09-22T02:03:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DS operating profit in Q4 could exceed 6 trillion KRW, more than double last year's Q4 figure of 2.9 trillion KRW", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung DRAM up 30%/NAND up 10% in Q4; supply tightening drives memory price cycle recovery benefiting Samsung and Micron.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Electronics Raises DRAM Prices by Up to 30%… NAND Up 10%\n\nSamsung Electronics has implemented contract price hikes for DRAM and NAND flash in the fourth quarter. With global competitors also moving to adjust prices, the tightening supply situation—driven by the reduction of legacy product production and the expansion of demand from major cloud providers—has been cited as the main background.\n\nAccording to industry sources on the 22nd, Samsung recently notified major customers that it will increase contract prices for key DRAM and NAND flash-based products in the fourth quarter. Specifically, DRAM products such as Low Power Double Data Rate (LPDDR) 4X, 5, and 5X will rise by 15–30%, while NAND-based mobile storage such as embedded MultiMediaCard (eMMC) and Universal Flash Storage (UFS) will rise by 5–10%.\n\nThis move aligns with recent price hikes by global memory manufacturers including Micron and SanDisk. Earlier this month, SanDisk informed customers it would raise NAND flash product prices by 10%, while Micron announced it would increase major product prices by 20–30% and temporarily suspend new quotations.\n\nAn industry official commented, “With fourth-quarter demand from smartphones and AI PCs coinciding, and legacy product production continuing to decline, price increases became inevitable. Samsung’s decision reflects this supply-demand environment.”\n\nBoth supply and demand factors appear to be driving the global memory manufacturers’ price adjustments. The fourth quarter is traditionally a peak season for new smartphone and PC launches. This year in particular, the arrival of flagship smartphones and AI PCs equipped with artificial intelligence features has boosted demand for low-power, high-performance memory.\n\nAdditionally, as production of legacy products decreases, the justification for price hikes has grown. Global memory firms are gradually reducing output of older DRAM products such as DDR4 and LPDDR4X, and shifting focus to next-generation DDR5 and HBM. In this environment, rising demand has pushed prices higher. In fact, in July, the monthly average price of DDR4 spiked by 50% in just one month, even surpassing DDR5 prices—an unprecedented phenomenon.\n\nOriginally, Samsung had planned to end production of DDR4 and LPDDR4 by June this year, but unexpected demand extended the timeline first to the end of the year, and more recently by another year to late next year. It was also reported that Samsung raised prices slightly for LPDDR5X and other products in the third quarter compared to the previous quarter.\n\nThe NAND flash situation is similar. Recently, large cloud service providers have placed massive orders for server solid-state drives (SSDs) to expand data centers, exacerbating supply shortages. Traditionally, data centers have relied heavily on hard disk drives (HDDs), but the explosive increase in storage capacity required for AI training and inference has driven NAND flash demand to surge at an unprecedented pace.\n\nUltimately, tight supply across both DRAM and NAND, coupled with rising demand, has strengthened the rationale for price hikes. Citigroup has forecast DRAM and NAND supply in 2025 to fall short of demand by 1.8% and 4%, respectively. Global investment bank Morgan Stanley has also predicted NAND supply could fall up to 8% short of demand next year.\n\nIndustry observers expect the price increases to directly improve Samsung’s semiconductor division (DS—Device Solutions) profitability in the fourth quarter. As of Q2 this year, Samsung held 32.7% of the global DRAM market and 32.9% of the NAND flash market, ranking second and first, respectively. In particular, Samsung has technological leadership in LPDDR, having developed the world’s first LPDDR5 in 2018 and LPDDR5X in 2021.\n\nGiven that revenue contributions from Nvidia-related HBM remain limited for now, the latest price hikes are seen as critical for defending near-term earnings. Some predict DS operating profit in Q4 could exceed 6 trillion KRW, more than double last year’s Q4 figure of 2.9 trillion KRW.\n\nAn industry official added, “Samsung still holds leadership in mobile memory and UFS/eMMC. This round of price hikes, in line with global competitors’ moves, will bolster the rebound of semiconductor profitability in the short term.”\n\n---\n\nhttps://t.co/wQXgcvstH7", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "thread:1969945634062193071", "ticker_idx": 1, "ts": "2025-09-22T02:03:33+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This move aligns with recent price hikes by global memory manufacturers including Micron and SanDisk", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung DRAM up 30%/NAND up 10% in Q4; supply tightening drives memory price cycle recovery benefiting Samsung and Micron.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Electronics Raises DRAM Prices by Up to 30%… NAND Up 10%\n\nSamsung Electronics has implemented contract price hikes for DRAM and NAND flash in the fourth quarter. With global competitors also moving to adjust prices, the tightening supply situation—driven by the reduction of legacy product production and the expansion of demand from major cloud providers—has been cited as the main background.\n\nAccording to industry sources on the 22nd, Samsung recently notified major customers that it will increase contract prices for key DRAM and NAND flash-based products in the fourth quarter. Specifically, DRAM products such as Low Power Double Data Rate (LPDDR) 4X, 5, and 5X will rise by 15–30%, while NAND-based mobile storage such as embedded MultiMediaCard (eMMC) and Universal Flash Storage (UFS) will rise by 5–10%.\n\nThis move aligns with recent price hikes by global memory manufacturers including Micron and SanDisk. Earlier this month, SanDisk informed customers it would raise NAND flash product prices by 10%, while Micron announced it would increase major product prices by 20–30% and temporarily suspend new quotations.\n\nAn industry official commented, “With fourth-quarter demand from smartphones and AI PCs coinciding, and legacy product production continuing to decline, price increases became inevitable. Samsung’s decision reflects this supply-demand environment.”\n\nBoth supply and demand factors appear to be driving the global memory manufacturers’ price adjustments. The fourth quarter is traditionally a peak season for new smartphone and PC launches. This year in particular, the arrival of flagship smartphones and AI PCs equipped with artificial intelligence features has boosted demand for low-power, high-performance memory.\n\nAdditionally, as production of legacy products decreases, the justification for price hikes has grown. Global memory firms are gradually reducing output of older DRAM products such as DDR4 and LPDDR4X, and shifting focus to next-generation DDR5 and HBM. In this environment, rising demand has pushed prices higher. In fact, in July, the monthly average price of DDR4 spiked by 50% in just one month, even surpassing DDR5 prices—an unprecedented phenomenon.\n\nOriginally, Samsung had planned to end production of DDR4 and LPDDR4 by June this year, but unexpected demand extended the timeline first to the end of the year, and more recently by another year to late next year. It was also reported that Samsung raised prices slightly for LPDDR5X and other products in the third quarter compared to the previous quarter.\n\nThe NAND flash situation is similar. Recently, large cloud service providers have placed massive orders for server solid-state drives (SSDs) to expand data centers, exacerbating supply shortages. Traditionally, data centers have relied heavily on hard disk drives (HDDs), but the explosive increase in storage capacity required for AI training and inference has driven NAND flash demand to surge at an unprecedented pace.\n\nUltimately, tight supply across both DRAM and NAND, coupled with rising demand, has strengthened the rationale for price hikes. Citigroup has forecast DRAM and NAND supply in 2025 to fall short of demand by 1.8% and 4%, respectively. Global investment bank Morgan Stanley has also predicted NAND supply could fall up to 8% short of demand next year.\n\nIndustry observers expect the price increases to directly improve Samsung’s semiconductor division (DS—Device Solutions) profitability in the fourth quarter. As of Q2 this year, Samsung held 32.7% of the global DRAM market and 32.9% of the NAND flash market, ranking second and first, respectively. In particular, Samsung has technological leadership in LPDDR, having developed the world’s first LPDDR5 in 2018 and LPDDR5X in 2021.\n\nGiven that revenue contributions from Nvidia-related HBM remain limited for now, the latest price hikes are seen as critical for defending near-term earnings. Some predict DS operating profit in Q4 could exceed 6 trillion KRW, more than double last year’s Q4 figure of 2.9 trillion KRW.\n\nAn industry official added, “Samsung still holds leadership in mobile memory and UFS/eMMC. This round of price hikes, in line with global competitors’ moves, will bolster the rebound of semiconductor profitability in the short term.”\n\n---\n\nhttps://t.co/wQXgcvstH7", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011481025777169696, "ret_signed_m1d": -0.011481025777169696, "alpha_spy_m1d": -0.006772294852597338, "alpha_c_m1d": 0.008570531977778018, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01087348305772684, "ret_signed_p1d": 0.01087348305772684, "alpha_spy_p1d": 0.0163169578975344, "alpha_c_p1d": 0.012332322371607751, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.004373843452624926, "ret_signed_p1w": -0.004373843452624926, "alpha_spy_p1w": 0.00036489580524323717, "alpha_c_p1w": -0.005894730709706697, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.22959944416106892, "ret_signed_p1m": 0.22959944416106892, "alpha_spy_p1m": 0.22292627200685078, "alpha_c_p1m": 0.15780512786668988, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5107863663515086, "ret_signed_p3m": 0.5107863663515086, "alpha_spy_p3m": 0.49634517812828816, "alpha_c_p3m": 0.43293928745085, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.8074702227486457, "ret_signed_p6m": 1.8074702227486457, "alpha_spy_p6m": 1.7985744088282953, "alpha_c_p6m": 1.5717408696055801, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2277, -0.2353, -0.2428, -0.252, -0.2663, -0.2612, -0.2578, -0.2578, -0.2715, -0.2442, -0.2574, -0.2522, -0.2435, -0.2795, -0.2704, -0.2927, -0.312, -0.3051, -0.3122, -0.3365, -0.3328, -0.3213, -0.3241, -0.3242, -0.3199, -0.303, -0.337, -0.3629, -0.3453, -0.3375, -0.3392, -0.3204, -0.2778, -0.2485, -0.224, -0.2451, -0.2389, -0.2658, -0.2495, -0.2586, -0.288, -0.2966, -0.2851, -0.2928, -0.2923, -0.2847, -0.2589, -0.2771, -0.2771, -0.2803, -0.2788, -0.2455, -0.202, -0.2014, -0.1785, -0.1496, -0.0853, -0.0449, -0.0416, -0.0352, -0.0281, 0.0259, -0.0115, 0.0, 0.0109, -0.0177, -0.0473, -0.0446, -0.0044, 0.0164, 0.1065, 0.1162, 0.1417, 0.1607, 0.1287, 0.1946, 0.1691, 0.1038, 0.1717, 0.137, 0.1667, 0.2311, 0.2301, 0.2568, 0.2296, 0.2064, 0.2565, 0.3313, 0.3379, 0.3489, 0.3775, 0.3616, 0.3602, 0.4266, 0.3253, 0.4436, 0.4487, 0.4462, 0.5397, 0.4656, 0.4886, 0.4403, 0.5003, 0.4707, 0.3889, 0.3732, 0.224, 0.2605, 0.3611, 0.3648, 0.3996, 0.3996, 0.4374, 0.4616, 0.4557, 0.4233, 0.3777, 0.4419, 0.5009, 0.5343, 0.6029, 0.571, 0.4657, 0.4436, 0.4133, 0.3708, 0.5108, 0.6164, 0.6812, 0.6793, 0.7426, 0.7426, 0.7311, 0.79, 0.7794, 0.7355, 0.7355, 0.918, 0.8981, 1.0883, 1.0648, 0.9886, 1.0984, 1.1032, 1.0561, 1.0271, 1.047, 1.2058, 1.2058, 1.2195, 1.3661, 1.4176, 1.4302, 1.366, 1.4946, 1.6469, 1.65, 1.5228, 1.6622, 1.5506, 1.3071, 1.3283, 1.4001, 1.332, 1.2697, 1.4952, 1.5173, 1.5032, 1.5032, 1.431, 1.5597, 1.5378, 1.6036, 1.5599, 1.5419, 1.6087, 1.527, 1.5076, 1.5094, 1.3088, 1.437, 1.4144, 1.2517, 1.3674, 1.4513, 1.546, 1.4649, 1.5912, 1.6865, 1.8075]}
{"unit_id": "orig:1970596623710625810", "ticker_idx": 0, "ts": "2025-09-23T21:10:21+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs expects Micron's stock price to rise in the short term, as its earnings and guidance significantly exceeded market expectations", "tweet_type": "news_reaction", "timeline": "days", "summary": "Shares Goldman view: MU stock to rise short-term on blowout earnings/guidance; SNDK seen benefiting from same catalyst.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs First Take\n\n12M TP: $130 (maintained)\n\nGoldman Sachs expects Micron’s stock price to rise in the short term, as its earnings and guidance significantly exceeded market expectations. However, despite this strong earnings release, the official investment rating on Micron remains Neutral.\n\nInvestors are expected to focus on the following three key points from Micron’s conference call:\n    \n1. HBM roadmap progress: Whether Micron can successfully ramp up HBM production and achieve its market share target by the end of 2025.\n    \n2. Sustainability of DRAM pricing: Whether the current strong upward momentum in DRAM prices can continue going forward.\n    \n3. Profitability maintenance: Whether Micron can defend high gross margins in a stable manner during the transition of production mix toward advanced products such as HBM4.\n\n• Micron’s strong earnings are also expected to trigger a positive initial reaction in the stock price of SanDisk (SNDK), which is exposed to similar end markets.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010756522195869866, "ret_signed_m1d": -0.010756522195869866, "alpha_spy_m1d": -0.016229790634715746, "alpha_c_m1d": -0.012217492831149812, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02824341205600267, "ret_signed_p1d": -0.02824341205600267, "alpha_spy_p1d": -0.025061778603388984, "alpha_c_p1d": -0.02640932438425203, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.0054684811190846006, "ret_signed_p1w": 0.0054684811190846006, "alpha_spy_p1w": 0.0009903523963925576, "alpha_c_p1w": -0.009016948046868034, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.19340362628470564, "ret_signed_p1m": 0.19340362628470564, "alpha_spy_p1m": 0.18648281458315097, "alpha_c_p1m": 0.1408702510561941, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.5989816019108829, "ret_signed_p3m": 0.5989816019108829, "alpha_spy_p3m": 0.5697434492981737, "alpha_c_p3m": 0.49164560331085694, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 1.7775123773604586, "ret_signed_p6m": 1.7775123773604586, "alpha_spy_p6m": 1.777249431682111, "alpha_c_p6m": 1.5499869217099527, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.2435, -0.251, -0.2601, -0.2742, -0.2691, -0.2658, -0.2658, -0.2794, -0.2523, -0.2654, -0.2602, -0.2517, -0.2872, -0.2782, -0.3003, -0.3194, -0.3126, -0.3196, -0.3437, -0.34, -0.3286, -0.3314, -0.3315, -0.3272, -0.3105, -0.3442, -0.3697, -0.3524, -0.3446, -0.3463, -0.3277, -0.2856, -0.2565, -0.2323, -0.2532, -0.2471, -0.2737, -0.2576, -0.2666, -0.2957, -0.3042, -0.2928, -0.3004, -0.2999, -0.2924, -0.2669, -0.2848, -0.2848, -0.288, -0.2866, -0.2536, -0.2106, -0.21, -0.1873, -0.1587, -0.0952, -0.0552, -0.0519, -0.0456, -0.0386, 0.0149, -0.0221, -0.0108, 0.0, -0.0282, -0.0576, -0.0549, -0.0151, 0.0055, 0.0946, 0.1042, 0.1294, 0.1482, 0.1166, 0.1818, 0.1565, 0.092, 0.1591, 0.1248, 0.1541, 0.2178, 0.2169, 0.2433, 0.2164, 0.1934, 0.243, 0.317, 0.3235, 0.3343, 0.3627, 0.347, 0.3455, 0.4113, 0.311, 0.4281, 0.4331, 0.4306, 0.5231, 0.4498, 0.4726, 0.4248, 0.4842, 0.4548, 0.374, 0.3585, 0.2108, 0.2469, 0.3465, 0.3501, 0.3846, 0.3846, 0.422, 0.4459, 0.4401, 0.408, 0.3629, 0.4264, 0.4847, 0.5178, 0.5857, 0.5541, 0.45, 0.4281, 0.3981, 0.3561, 0.4945, 0.599, 0.6631, 0.6612, 0.7238, 0.7238, 0.7124, 0.7708, 0.7603, 0.7169, 0.7169, 0.8974, 0.8777, 1.0659, 1.0425, 0.9672, 1.0759, 1.0806, 1.034, 1.0052, 1.025, 1.1821, 1.1821, 1.1956, 1.3407, 1.3916, 1.4041, 1.3406, 1.4678, 1.6184, 1.6215, 1.4957, 1.6336, 1.5231, 1.2823, 1.3033, 1.3742, 1.3069, 1.2453, 1.4684, 1.4902, 1.4763, 1.4763, 1.4049, 1.5322, 1.5105, 1.5756, 1.5323, 1.5145, 1.5806, 1.4998, 1.4806, 1.4824, 1.2839, 1.4108, 1.3884, 1.2275, 1.3419, 1.4249, 1.5186, 1.4384, 1.5634, 1.6576, 1.7773, 1.7775]}
{"unit_id": "orig:1970596623710625810", "ticker_idx": 1, "ts": "2025-09-23T21:10:21+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's strong earnings are also expected to trigger a positive initial reaction in the stock price of SanDisk (SNDK)", "tweet_type": "news_reaction", "timeline": "days", "summary": "Shares Goldman view: MU stock to rise short-term on blowout earnings/guidance; SNDK seen benefiting from same catalyst.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs First Take\n\n12M TP: $130 (maintained)\n\nGoldman Sachs expects Micron’s stock price to rise in the short term, as its earnings and guidance significantly exceeded market expectations. However, despite this strong earnings release, the official investment rating on Micron remains Neutral.\n\nInvestors are expected to focus on the following three key points from Micron’s conference call:\n    \n1. HBM roadmap progress: Whether Micron can successfully ramp up HBM production and achieve its market share target by the end of 2025.\n    \n2. Sustainability of DRAM pricing: Whether the current strong upward momentum in DRAM prices can continue going forward.\n    \n3. Profitability maintenance: Whether Micron can defend high gross margins in a stable manner during the transition of production mix toward advanced products such as HBM4.\n\n• Micron’s strong earnings are also expected to trigger a positive initial reaction in the stock price of SanDisk (SNDK), which is exposed to similar end markets.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03270679799837384, "ret_signed_m1d": -0.03270679799837384, "alpha_spy_m1d": -0.03818006643721972, "alpha_c_m1d": -0.04141427582097357, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.00870747782259973, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06174811654703072, "ret_signed_p1d": -0.06174811654703072, "alpha_spy_p1d": -0.05856648309441703, "alpha_c_p1d": -0.055574449902096146, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.006173666644934572, "ret_p1w": 0.05451123439084338, "ret_signed_p1w": 0.05451123439084338, "alpha_spy_p1w": 0.050033105668151334, "alpha_c_p1w": 0.048658635104857906, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.005852599285985471, "ret_p1m": 0.38110897406801825, "ret_signed_p1m": 0.38110897406801825, "alpha_spy_p1m": 0.3741881623664636, "alpha_c_p1m": 0.36230228970770084, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.018806684360317405, "ret_p3m": 1.2331766654397969, "ret_signed_p3m": 1.2331766654397969, "alpha_spy_p3m": 1.2039385128270876, "alpha_c_p3m": 1.2011303207862183, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.032046344653578585, "ret_p6m": 6.0835525529395005, "ret_signed_p6m": 6.0835525529395005, "alpha_spy_p6m": 6.083289607261153, "alpha_c_p6m": 6.097404097726704, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": -0.013851544787203851, "price_path": [-0.5541, -0.5569, -0.5738, -0.5774, -0.5657, -0.5638, -0.5638, -0.575, -0.5661, -0.5658, -0.5587, -0.5668, -0.6008, -0.5985, -0.6113, -0.6098, -0.6035, -0.6089, -0.6113, -0.5959, -0.6047, -0.6008, -0.6063, -0.5965, -0.5922, -0.5966, -0.6116, -0.6005, -0.6059, -0.6043, -0.6176, -0.5833, -0.5924, -0.5599, -0.5582, -0.5613, -0.5814, -0.5722, -0.581, -0.5827, -0.5724, -0.5642, -0.5603, -0.555, -0.5447, -0.5219, -0.5069, -0.5069, -0.52, -0.5018, -0.4126, -0.3557, -0.3375, -0.3373, -0.3053, -0.2077, -0.1905, -0.1533, -0.1396, -0.1168, -0.0708, -0.0394, -0.0327, 0.0, -0.0617, -0.1138, -0.0872, 0.0667, 0.0545, 0.1383, 0.1667, 0.2069, 0.1388, 0.1367, 0.2395, 0.2188, 0.0988, 0.2651, 0.1963, 0.3562, 0.3559, 0.3173, 0.3914, 0.4031, 0.3811, 0.57, 0.7496, 0.6587, 0.6497, 0.9207, 0.8404, 0.8734, 0.9456, 0.8287, 1.0348, 0.952, 1.2508, 1.5183, 1.5524, 1.6607, 1.2892, 1.3887, 1.4989, 1.302, 1.3117, 0.8417, 0.8822, 1.1331, 1.0724, 1.0211, 1.0211, 1.0985, 0.9753, 0.93, 0.8269, 1.0048, 1.1473, 1.1191, 1.0626, 1.1885, 1.2708, 0.9378, 0.8973, 0.9672, 0.9439, 1.0626, 1.2332, 1.2655, 1.3017, 1.3504, 1.3504, 1.3501, 1.2956, 1.2577, 1.231, 1.231, 1.5868, 1.5759, 2.286, 2.3229, 2.1442, 2.5471, 2.6586, 2.6636, 2.6448, 2.8462, 2.8874, 2.8874, 3.2586, 3.7114, 3.7316, 3.4533, 3.4248, 3.5247, 3.9589, 4.0686, 4.4159, 5.2523, 5.5367, 4.4939, 4.4154, 4.6198, 4.4831, 4.0906, 4.6329, 4.9238, 4.8887, 4.8887, 4.5507, 4.6429, 4.8373, 5.1087, 5.264, 5.0011, 4.9434, 5.1269, 4.9714, 4.8184, 4.314, 4.6303, 4.3157, 3.9561, 4.5332, 4.8166, 5.1601, 4.816, 5.2182, 5.6131, 5.7685, 6.0836]}
{"unit_id": "orig:1969927807708094927", "ticker_idx": 0, "ts": "2025-09-22T00:52:43+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix, which received the largest volume allocation by a narrow margin, will need to make additional 1c nm investments next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "2026 SOCAMM orders shift to SK Hynix (11B Gb) and Samsung (10B Gb) at Micron's expense; SK Hynix leads, Samsung expanding; Micron growth limited.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[2026 SOCAMM: Supplier Composition Centered on SK hynix and Samsung Electronics]\n\nSince last year, NVIDIA has continued to use Micron as the supplier of LPDDR for its Grace CPU (Figure 1). This year’s SOCAMM orders have also been placed almost exclusively with Micron. However, recently, NVIDIA has been restructuring its allocation for next year to focus mainly on domestic suppliers. While Samsung Electronics and SK hynix received only minimal volumes this year, they have each been allocated explosive volumes of 10 billion Gb and 11 billion Gb, respectively, for next year. Meanwhile, Micron, which received 5 billion Gb orders this year, is expected to see only limited growth to 7 billion Gb next year.\n\nThis move is interpreted as NVIDIA’s intention to utilize all three suppliers for SOCAMM. Unlike HBM, product competitiveness differences between suppliers are not significant for SOCAMM. From the supplier’s perspective, technological capability translates into cost differences rather than performance, making the adoption of the 1c nm node the most important factor. SK hynix, which received the largest volume allocation by a narrow margin, will need to make additional 1c nm investments next year. Samsung Electronics, meanwhile, plans to allocate its entire 1c nm capacity to HBM, but is highly likely to gradually expand its application to SOCAMM and other products.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012820521976864452, "ret_signed_m1d": 0.012820521976864452, "alpha_spy_m1d": 0.01752925290143681, "alpha_c_m1d": 0.032872079731812165, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028489969482429656, "ret_signed_p1d": 0.028489969482429656, "alpha_spy_p1d": 0.033933444322237216, "alpha_c_p1d": 0.029948808796310566, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.0056981188806843885, "ret_signed_p1w": -0.0056981188806843885, "alpha_spy_p1w": -0.0009593796228162255, "alpha_c_p1w": -0.00721900613776616, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.3646722878607491, "ret_signed_p1m": 0.3646722878607491, "alpha_spy_p1m": 0.357999115706531, "alpha_c_p1m": 0.2928779715663701, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5737758511959559, "ret_signed_p3m": 0.5737758511959559, "alpha_spy_p3m": 0.5593346629727354, "alpha_c_p3m": 0.4959287722952972, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.7706148936884407, "ret_signed_p6m": 1.7706148936884407, "alpha_spy_p6m": 1.7617190797680904, "alpha_c_p6m": 1.5348855405453752, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.1864, -0.1664, -0.1921, -0.1693, -0.1878, -0.2063, -0.2077, -0.2305, -0.229, -0.1977, -0.2006, -0.1551, -0.1622, -0.1465, -0.1508, -0.1579, -0.2333, -0.2347, -0.2248, -0.2361, -0.2347, -0.2333, -0.2433, -0.2546, -0.2532, -0.2504, -0.2219, -0.266, -0.266, -0.2504, -0.2646, -0.2546, -0.2703, -0.2404, -0.2347, -0.2091, -0.2134, -0.2134, -0.239, -0.2518, -0.2731, -0.303, -0.2859, -0.2618, -0.2561, -0.2603, -0.235, -0.2336, -0.2707, -0.2578, -0.2521, -0.2436, -0.2208, -0.2108, -0.1795, -0.1339, -0.1254, -0.0641, -0.057, -0.0085, -0.0499, 0.0128, 0.0128, 0.0, 0.0285, 0.0185, 0.0157, -0.0413, -0.0057, -0.01, 0.0256, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.2194, 0.1823, 0.1724, 0.2037, 0.2892, 0.3262, 0.3832, 0.3647, 0.3718, 0.3632, 0.453, 0.5242, 0.4843, 0.5897, 0.6182, 0.5926, 0.7664, 0.6695, 0.6496, 0.6895, 0.6524, 0.7265, 0.7635, 0.7578, 0.7436, 0.5954, 0.7265, 0.6239, 0.6011, 0.6268, 0.4843, 0.4815, 0.4786, 0.4929, 0.551, 0.5111, 0.5339, 0.5909, 0.5738, 0.5453, 0.551, 0.6451, 0.6137, 0.6736, 0.6108, 0.6279, 0.5795, 0.5111, 0.5709, 0.5738, 0.5595, 0.6536, 0.665, 0.6764, 0.6764, 0.7078, 0.8247, 0.856, 0.856, 0.856, 0.9302, 0.9843, 1.0699, 1.1155, 1.1554, 1.1212, 1.1354, 1.1041, 1.1155, 1.1354, 1.1554, 1.1782, 1.1183, 1.1098, 1.1525, 1.1868, 1.0984, 1.2808, 1.3977, 1.4547, 1.5916, 1.3664, 1.5859, 1.5659, 1.4006, 1.392, 1.5289, 1.4975, 1.4519, 1.5317, 1.5089, 1.5089, 1.5089, 1.5089, 1.5488, 1.7056, 1.7113, 1.8653, 1.9024, 2.1391, 2.0305, 2.0305, 1.6821, 1.425, 1.6878, 1.6392, 1.3879, 1.6792, 1.7278, 1.6564, 1.5992, 1.782, 1.7706]}
{"unit_id": "orig:1969927807708094927", "ticker_idx": 1, "ts": "2025-09-22T00:52:43+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics, meanwhile, plans to allocate its entire 1c nm capacity to HBM, but is highly likely to gradually expand its application to SOCAMM and other products", "tweet_type": "analysis", "timeline": "quarters", "summary": "2026 SOCAMM orders shift to SK Hynix (11B Gb) and Samsung (10B Gb) at Micron's expense; SK Hynix leads, Samsung expanding; Micron growth limited.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "[2026 SOCAMM: Supplier Composition Centered on SK hynix and Samsung Electronics]\n\nSince last year, NVIDIA has continued to use Micron as the supplier of LPDDR for its Grace CPU (Figure 1). This year’s SOCAMM orders have also been placed almost exclusively with Micron. However, recently, NVIDIA has been restructuring its allocation for next year to focus mainly on domestic suppliers. While Samsung Electronics and SK hynix received only minimal volumes this year, they have each been allocated explosive volumes of 10 billion Gb and 11 billion Gb, respectively, for next year. Meanwhile, Micron, which received 5 billion Gb orders this year, is expected to see only limited growth to 7 billion Gb next year.\n\nThis move is interpreted as NVIDIA’s intention to utilize all three suppliers for SOCAMM. Unlike HBM, product competitiveness differences between suppliers are not significant for SOCAMM. From the supplier’s perspective, technological capability translates into cost differences rather than performance, making the adoption of the 1c nm node the most important factor. SK hynix, which received the largest volume allocation by a narrow margin, will need to make additional 1c nm investments next year. Samsung Electronics, meanwhile, plans to allocate its entire 1c nm capacity to HBM, but is highly likely to gradually expand its application to SOCAMM and other products.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "orig:1969927807708094927", "ticker_idx": 2, "ts": "2025-09-22T00:52:43+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron, which received 5 billion Gb orders this year, is expected to see only limited growth to 7 billion Gb next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "2026 SOCAMM orders shift to SK Hynix (11B Gb) and Samsung (10B Gb) at Micron's expense; SK Hynix leads, Samsung expanding; Micron growth limited.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[2026 SOCAMM: Supplier Composition Centered on SK hynix and Samsung Electronics]\n\nSince last year, NVIDIA has continued to use Micron as the supplier of LPDDR for its Grace CPU (Figure 1). This year’s SOCAMM orders have also been placed almost exclusively with Micron. However, recently, NVIDIA has been restructuring its allocation for next year to focus mainly on domestic suppliers. While Samsung Electronics and SK hynix received only minimal volumes this year, they have each been allocated explosive volumes of 10 billion Gb and 11 billion Gb, respectively, for next year. Meanwhile, Micron, which received 5 billion Gb orders this year, is expected to see only limited growth to 7 billion Gb next year.\n\nThis move is interpreted as NVIDIA’s intention to utilize all three suppliers for SOCAMM. Unlike HBM, product competitiveness differences between suppliers are not significant for SOCAMM. From the supplier’s perspective, technological capability translates into cost differences rather than performance, making the adoption of the 1c nm node the most important factor. SK hynix, which received the largest volume allocation by a narrow margin, will need to make additional 1c nm investments next year. Samsung Electronics, meanwhile, plans to allocate its entire 1c nm capacity to HBM, but is highly likely to gradually expand its application to SOCAMM and other products.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011481025777169696, "ret_signed_m1d": 0.011481025777169696, "alpha_spy_m1d": 0.006772294852597338, "alpha_c_m1d": -0.008570531977778018, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01087348305772684, "ret_signed_p1d": -0.01087348305772684, "alpha_spy_p1d": -0.0163169578975344, "alpha_c_p1d": -0.012332322371607751, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.004373843452624926, "ret_signed_p1w": 0.004373843452624926, "alpha_spy_p1w": -0.00036489580524323717, "alpha_c_p1w": 0.005894730709706697, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.22959944416106892, "ret_signed_p1m": -0.22959944416106892, "alpha_spy_p1m": -0.22292627200685078, "alpha_c_p1m": -0.15780512786668988, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5107863663515086, "ret_signed_p3m": -0.5107863663515086, "alpha_spy_p3m": -0.49634517812828816, "alpha_c_p3m": -0.43293928745085, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.8074702227486457, "ret_signed_p6m": -1.8074702227486457, "alpha_spy_p6m": -1.7985744088282953, "alpha_c_p6m": -1.5717408696055801, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2277, -0.2353, -0.2428, -0.252, -0.2663, -0.2612, -0.2578, -0.2578, -0.2715, -0.2442, -0.2574, -0.2522, -0.2435, -0.2795, -0.2704, -0.2927, -0.312, -0.3051, -0.3122, -0.3365, -0.3328, -0.3213, -0.3241, -0.3242, -0.3199, -0.303, -0.337, -0.3629, -0.3453, -0.3375, -0.3392, -0.3204, -0.2778, -0.2485, -0.224, -0.2451, -0.2389, -0.2658, -0.2495, -0.2586, -0.288, -0.2966, -0.2851, -0.2928, -0.2923, -0.2847, -0.2589, -0.2771, -0.2771, -0.2803, -0.2788, -0.2455, -0.202, -0.2014, -0.1785, -0.1496, -0.0853, -0.0449, -0.0416, -0.0352, -0.0281, 0.0259, -0.0115, 0.0, 0.0109, -0.0177, -0.0473, -0.0446, -0.0044, 0.0164, 0.1065, 0.1162, 0.1417, 0.1607, 0.1287, 0.1946, 0.1691, 0.1038, 0.1717, 0.137, 0.1667, 0.2311, 0.2301, 0.2568, 0.2296, 0.2064, 0.2565, 0.3313, 0.3379, 0.3489, 0.3775, 0.3616, 0.3602, 0.4266, 0.3253, 0.4436, 0.4487, 0.4462, 0.5397, 0.4656, 0.4886, 0.4403, 0.5003, 0.4707, 0.3889, 0.3732, 0.224, 0.2605, 0.3611, 0.3648, 0.3996, 0.3996, 0.4374, 0.4616, 0.4557, 0.4233, 0.3777, 0.4419, 0.5009, 0.5343, 0.6029, 0.571, 0.4657, 0.4436, 0.4133, 0.3708, 0.5108, 0.6164, 0.6812, 0.6793, 0.7426, 0.7426, 0.7311, 0.79, 0.7794, 0.7355, 0.7355, 0.918, 0.8981, 1.0883, 1.0648, 0.9886, 1.0984, 1.1032, 1.0561, 1.0271, 1.047, 1.2058, 1.2058, 1.2195, 1.3661, 1.4176, 1.4302, 1.366, 1.4946, 1.6469, 1.65, 1.5228, 1.6622, 1.5506, 1.3071, 1.3283, 1.4001, 1.332, 1.2697, 1.4952, 1.5173, 1.5032, 1.5032, 1.431, 1.5597, 1.5378, 1.6036, 1.5599, 1.5419, 1.6087, 1.527, 1.5076, 1.5094, 1.3088, 1.437, 1.4144, 1.2517, 1.3674, 1.4513, 1.546, 1.4649, 1.5912, 1.6865, 1.8075]}
{"unit_id": "orig:1959784312372629573", "ticker_idx": 0, "ts": "2025-08-25T01:06:05+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HSBC raises its price target from $125 to $200, based on FY27 EPS of $6.43 and applying a 31x target PE", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC note raising NVDA PT from $125 to $200 on FY27 EPS $6.43; rating maintained at Hold with supply chain and China headwinds.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA 2Q Earnings Preview: China and Supply Chain Uncertainties Persist, GPU TAM Expansion Supports Long-Term Potential\nHSBC (F. Lee, 25/08/20)\n\n2Q/3QFY26 Outlook\n\nHSBC forecasts 2QFY26 revenue of $46.7 billion, slightly above management guidance of $45.0 billion and in line with consensus of $46.3 billion; 3QFY26 revenue of $53.9 billion, also in line with consensus of $53.3 billion. Gross margin is expected to recover to 72% (2Q) and 73% (3Q), with EPS of $1.03 and $1.21, respectively, both consistent with market expectations. Overall performance is solid, but China market uncertainty remains the main constraint.\n\nChina Market Uncertainty\n\nAlthough H20 export licenses have resumed, HSBC believes risks remain in China’s AI GPU market:\n\n1. The U.S. side requires a 15% revenue share, depressing ASPs;\n\n2. Chinese regulators are discouraging domestic companies from purchasing U.S. products;\n\n3. NVIDIA itself may launch the B30 GPU as an H20 replacement.\n\nHSBC expects no H20 shipments in 3QFY26, recovery of 390,000 units in 4Q generating $3.3 billion revenue, and full-year revenue of $7.9 billion, below the previous estimate of $10 billion, and far below the potential $15 billion level absent restrictions.\n\nSupply Chain Bottlenecks and NVL Cabinet Shipments\n\nHSBC expects FY26 NVL cabinet shipments of about 26,000 units. Even if yields improve to 60–70%, this remains below NVIDIA’s target of 50,000 units and HSBC’s prior forecast of 40,000 units. GPU shipments are expected to reach 6.8 million units (equivalent to 284,000 CoWoS wafers). Supply chain bottlenecks imply some shipments will be deferred into FY27. If Rubin platform mass production is delayed, uncertainty will be further exacerbated.\n\nGPU TAM and FY27 Growth Drivers\n\nDriven by revised CSP capital expenditure, GPU TAM continues to expand. HSBC raises its FY27 data center revenue forecast to $253 billion (consensus: $240 billion), and under a bull case assumes CSP Capex of $545 billion (vs current $446 billion), corresponding to data center revenue of $310 billion and EPS of $7.74. Under the base case, FY27 EPS is estimated at $6.43, higher than the consensus of $5.94.\n\nValuation and Rating\n\nHSBC raises its price target from $125 to $200, based on FY27 EPS of $6.43 and applying a 31x target PE (a 0.5 standard deviation discount to the past five-year average). The rating is maintained at Hold, with short-term supply chain and China market uncertainties still seen as headwinds. Only further upward revisions of CSP Capex in 2026 could drive higher re-rating.\n\nRisk Factors\n\nUpside risks: AI GPU and NVL cabinet shipments exceeding expectations, ASPs higher than expected, stronger-than-expected China H20 demand.\n\nDownside risks: High-end GPU product mix below expectations, CSP investment in generative AI weakening, overall data center demand softening, geopolitical uncertainties.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010121735824010547, "ret_signed_m1d": -0.010121735824010547, "alpha_spy_m1d": -0.014542348693156959, "alpha_c_m1d": -0.009169051936335437, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01090037658917331, "ret_signed_p1d": 0.01090037658917331, "alpha_spy_p1d": 0.006713340239282495, "alpha_c_p1d": 0.0012719889215084823, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.03131088170414875, "ret_signed_p1w": -0.03131088170414875, "alpha_spy_p1w": -0.035326711638982156, "alpha_c_p1w": -0.01896056982958516, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": -0.007618864936297265, "ret_signed_p1m": -0.007618864936297265, "alpha_spy_p1m": -0.042762646229730694, "alpha_c_p1m": -0.10213433451457077, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.00467260846353823, "ret_signed_p3m": 0.00467260846353823, "alpha_spy_p3m": -0.013801684658514723, "alpha_c_p3m": -0.10141057030732781, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.028812254552892602, "ret_signed_p6m": 0.028812254552892602, "alpha_spy_p6m": -0.04013497353668627, "alpha_c_p6m": -0.36196749018156815, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.2503, -0.226, -0.2485, -0.236, -0.2147, -0.2108, -0.2215, -0.2119, -0.2068, -0.1994, -0.2057, -0.1936, -0.2104, -0.1953, -0.1985, -0.1909, -0.1909, -0.2, -0.1982, -0.1775, -0.1418, -0.1379, -0.1227, -0.1214, -0.1474, -0.1255, -0.1138, -0.1138, -0.12, -0.1102, -0.0942, -0.0874, -0.0828, -0.0875, -0.0507, -0.0469, -0.0379, -0.0412, -0.0469, -0.0711, -0.0502, -0.0338, -0.0351, -0.017, -0.0239, -0.003, -0.0108, -0.0339, 0.0011, -0.0086, -0.0022, 0.0053, 0.0161, 0.0125, 0.0186, 0.0099, 0.0123, 0.0036, 0.0122, -0.0232, -0.0245, -0.0269, -0.0101, 0.0, 0.0109, 0.01, 0.002, -0.0313, -0.0313, -0.0502, -0.0511, -0.0453, -0.0711, -0.064, -0.0503, -0.0138, -0.0146, -0.011, -0.0114, -0.0274, -0.0529, -0.0198, -0.0174, 0.0212, -0.0076, -0.0157, -0.0117, -0.009, 0.0114, 0.0377, 0.0414, 0.0506, 0.0435, 0.0319, 0.0291, 0.0518, 0.071, 0.0187, 0.0474, 0.0013, 0.0002, 0.0112, 0.019, 0.0158, 0.0076, 0.0027, 0.0131, 0.0359, 0.065, 0.1181, 0.1515, 0.1284, 0.1262, 0.1506, 0.1051, 0.0857, 0.0461, 0.0464, 0.1071, 0.0743, 0.0779, 0.0393, 0.0577, 0.0378, 0.0087, 0.0374, 0.0047, -0.0051, 0.0153, -0.011, 0.0026, 0.0026, -0.0156, 0.0007, 0.0092, -0.0012, 0.02, 0.0146, 0.032, 0.0288, 0.0222, 0.0063, -0.0265, -0.0195, -0.0115, -0.0492, -0.0314, 0.0067, 0.0217, 0.0524, 0.0491, 0.0491, 0.0597, 0.0469, 0.0431, 0.0373, 0.0373, 0.0504, 0.0463, 0.0414, 0.0518, 0.0292, 0.0282, 0.0286, 0.0335, 0.0186, 0.0404, 0.0358, 0.0358, -0.0096, 0.0196, 0.0281, 0.0438, 0.0372, 0.0486, 0.0652, 0.0708, 0.0631, 0.0324, 0.0031, -0.0311, -0.044, 0.0313, 0.057, 0.0487, 0.0571, 0.0398, 0.0168, 0.0168, 0.0288]}
{"unit_id": "quote:1959193240705601544", "ticker_idx": 0, "ts": "2025-08-23T09:57:23+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Even if this may look like a bad deal for Intel in the short term, in the long run it's actually a very good deal.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish INTC long-term: US gov equity stake acts as co-signer, Trump becomes Intel's salesman, accelerates process dev and could pressure NVDA/Apple/Qualcomm to use Intel Foundry.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Even if this may look like a bad deal for Intel in the short term, in the long run it’s actually a very good deal.\n\nIt’s essentially the U.S. government co-signing Intel’s guarantees. And it also means Trump himself is effectively becoming Intel’s salesman.\n\nWith this capital injection, Intel can accelerate process development, and we can even consider a scenario where the U.S. government pressures big fabless companies like NVIDIA, Apple, and Qualcomm to give orders to Intel Foundry.\n\nWhy is this deal being viewed so negatively in the first place? I honestly don’t understand.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": "@Jukanlosreve perhaps you could step through your logic for me. I'd love to know how coercing Intel to convert CHIPs Act funding etc into a 10% equity stake helps to save Intel? Even Mr. Lutnick clearly stated that this is a worse deal for Intel compared to before. Why would he say that?", "parent_author": null, "ret_m1d": 0.010183299705468052, "ret_signed_m1d": 0.010183299705468052, "alpha_spy_m1d": 0.00576268683632164, "alpha_c_m1d": 0.011135983593143162, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00814659314892785, "ret_signed_p1d": -0.00814659314892785, "alpha_spy_p1d": -0.012333629498818666, "alpha_c_p1d": -0.017774980816592678, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.00814659314892785, "ret_signed_p1w": -0.00814659314892785, "alpha_spy_p1w": -0.012162423083761253, "alpha_c_p1w": 0.004203718725635741, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.1951120596491256, "ret_signed_p1m": 0.1951120596491256, "alpha_spy_p1m": 0.15996827835569216, "alpha_c_p1m": 0.10059659007085209, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.36945010088359576, "ret_signed_p3m": 0.36945010088359576, "alpha_spy_p3m": 0.3509758077615428, "alpha_c_p3m": 0.2633669221127297, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.8810591340248475, "ret_signed_p6m": 0.8810591340248475, "alpha_spy_p6m": 0.8121119059352686, "alpha_c_p6m": 0.4902793892903867, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1703, -0.1752, -0.2037, -0.1959, -0.1735, -0.1752, -0.1857, -0.1829, -0.1658, -0.1006, -0.1576, -0.154, -0.1796, -0.1552, -0.1527, -0.1246, -0.1246, -0.1413, -0.1369, -0.0815, -0.0957, -0.0835, -0.0758, -0.0876, -0.0692, -0.1088, -0.0839, -0.0839, -0.1039, -0.0391, -0.0452, -0.0297, -0.0456, -0.0509, -0.0664, -0.0758, -0.0713, -0.0591, -0.0525, -0.0534, -0.0432, -0.0782, -0.1568, -0.1576, -0.1686, -0.1715, -0.1935, -0.2134, -0.2057, -0.1776, -0.1686, -0.1947, -0.1874, -0.1589, -0.1116, -0.0949, -0.0281, 0.0004, -0.0363, 0.031, -0.0411, -0.0428, 0.0102, 0.0, -0.0081, 0.0122, 0.0155, -0.0081, -0.0081, -0.0138, -0.0224, 0.0024, -0.0024, -0.0029, -0.0045, 0.009, 0.0024, -0.0191, 0.009, 0.0293, 0.0143, 0.2452, 0.2049, 0.1715, 0.1951, 0.2717, 0.3845, 0.446, 0.4045, 0.3666, 0.464, 0.5193, 0.5002, 0.4904, 0.5141, 0.5246, 0.5397, 0.4815, 0.5161, 0.4513, 0.5132, 0.5006, 0.5075, 0.5519, 0.5527, 0.5039, 0.5544, 0.5593, 0.6106, 0.6916, 0.6839, 0.6358, 0.6289, 0.609, 0.5084, 0.5633, 0.5169, 0.5532, 0.5662, 0.543, 0.5434, 0.4627, 0.4468, 0.4138, 0.3984, 0.4301, 0.3695, 0.4053, 0.4578, 0.4595, 0.4994, 0.4994, 0.6521, 0.6297, 0.7707, 0.7825, 0.6497, 0.6868, 0.6415, 0.6497, 0.6611, 0.6094, 0.5401, 0.5279, 0.5198, 0.4684, 0.4778, 0.4998, 0.4815, 0.4807, 0.4729, 0.4729, 0.4745, 0.4941, 0.5193, 0.5031, 0.5031, 0.6041, 0.6037, 0.631, 0.7365, 0.6745, 0.8554, 0.7947, 0.9263, 0.9845, 0.9682, 0.9128, 0.9128, 0.978, 1.2098, 1.2126, 0.8358, 0.7308, 0.7894, 0.987, 0.9821, 0.8929, 0.9882, 1.0061, 0.9796, 0.965, 1.0607, 1.0464, 0.9198, 0.967, 0.8933, 0.9059, 0.9059, 0.8811]}
{"unit_id": "reply:1951563761141285014", "ticker_idx": 0, "ts": "2025-08-02T08:40:33+00:00", "symbol": "Meta Platforms", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think Zuckerberg is moving with impressive agility", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Meta on Zuckerberg's agility in AI/social competition vs OpenAI.", "resolved_tickers": ["META"], "resolution_method": "exchange_ticker", "resolution_reason": "Meta Platforms META US", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@RonnieBateman5 You’re right. There have already been multiple media reports that OpenAI is considering building a social media platform like X.\n\nI’m not sure how bureaucratic Meta has become, but I think Zuckerberg is moving with impressive agility.", "reply_text": null, "parent_text": "@Jukanlosreve I was listening to bg2 podcast the other day. Hosts were saying the economic value of AI will accrue at the consumer layer. Whom owns the consumers shall win the war. Meta has a shot here to create another few trillion in market cap. If Meta had failed in SI, they will be in", "quoted_text": null, "parent_author": "RonnieBateman5", "ret_m1d": -0.03395285653710489, "ret_signed_m1d": -0.03395285653710489, "alpha_spy_m1d": -0.018980617598010174, "alpha_c_m1d": -0.015091832860982701, "bench_spy_m1d": -0.014972238939094717, "bench_c_m1d": -0.01886102367612219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016628680790430894, "ret_signed_p1d": -0.016628680790430894, "alpha_spy_p1d": -0.011558714217999588, "alpha_c_p1d": -0.00886233082744714, "bench_spy_p1d": -0.005069966572431306, "bench_c_p1d": -0.007766349962983754, "ret_p1w": -0.013524484913330692, "ret_signed_p1w": -0.013524484913330692, "alpha_spy_p1w": -0.021050247019865242, "alpha_c_p1w": -0.01370928097086277, "bench_spy_p1w": 0.00752576210653455, "bench_c_p1w": 0.00018479605753207728, "ret_p1m": -0.05314484521691187, "ret_signed_p1m": -0.05314484521691187, "alpha_spy_p1m": -0.06756257484749617, "alpha_c_p1m": -0.08125122847453148, "bench_spy_p1m": 0.014417729630584297, "bench_c_p1m": 0.028106383257619605, "ret_p3m": -0.14097689067104546, "ret_signed_p3m": -0.14097689067104546, "alpha_spy_p3m": -0.22105804056302547, "alpha_c_p3m": -0.20161564387185005, "bench_spy_p3m": 0.08008114989198001, "bench_c_p3m": 0.060638753200804585, "ret_p6m": -0.13189148545067264, "ret_signed_p6m": -0.13189148545067264, "alpha_spy_p6m": -0.24011754229753135, "alpha_c_p6m": -0.2199781523795018, "bench_spy_p6m": 0.10822605684685871, "bench_c_p6m": 0.08808666692882916, "price_path": [-0.2319, -0.2303, -0.2374, -0.177, -0.1557, -0.1514, -0.1713, -0.1758, -0.1757, -0.18, -0.1821, -0.1807, -0.1929, -0.1929, -0.1733, -0.1717, -0.1698, -0.1666, -0.1365, -0.1417, -0.1146, -0.1189, -0.102, -0.1067, -0.096, -0.1066, -0.1076, -0.1211, -0.0956, -0.1019, -0.1038, -0.1038, -0.1211, -0.1003, -0.0827, -0.0872, -0.0648, -0.0551, -0.0493, -0.0736, -0.0809, -0.0739, -0.0739, -0.0747, -0.0717, -0.0561, -0.0633, -0.0758, -0.0714, -0.085, -0.0946, -0.0966, -0.0929, -0.0817, -0.0922, -0.0809, -0.0793, -0.082, -0.0757, -0.0984, -0.1045, -0.0038, -0.034, 0.0, -0.0166, -0.0056, -0.0187, -0.0091, -0.0135, 0.0176, 0.0048, 0.0074, 0.0114, -0.0116, -0.0321, -0.0369, -0.048, -0.0278, -0.0297, -0.0287, -0.0373, -0.0325, -0.0485, -0.0485, -0.0531, -0.0506, -0.0357, -0.0308, -0.031, -0.0137, -0.0314, -0.0328, -0.0268, -0.015, 0.0034, -0.0008, 0.005, 0.0026, -0.0138, -0.0264, -0.0196, -0.0347, -0.0414, -0.0418, -0.0534, -0.0754, -0.0629, -0.0841, -0.0776, -0.0809, -0.0748, -0.0546, -0.0909, -0.0775, -0.0866, -0.0751, -0.0822, -0.076, -0.0563, -0.0549, -0.0547, -0.0539, -0.0483, -0.0323, -0.0315, -0.0312, -0.141, -0.1643, -0.178, -0.1914, -0.1803, -0.2022, -0.1987, -0.1857, -0.1917, -0.215, -0.2139, -0.2145, -0.2241, -0.2296, -0.2391, -0.2406, -0.2341, -0.2098, -0.18, -0.1833, -0.1833, -0.1648, -0.174, -0.1659, -0.1756, -0.1473, -0.132, -0.1406, -0.1532, -0.162, -0.1587, -0.1696, -0.1647, -0.1523, -0.1622, -0.1429, -0.1502, -0.1467, -0.1422, -0.1389, -0.1389, -0.1444, -0.1503, -0.1409, -0.1485, -0.1485, -0.161, -0.1502, -0.1478, -0.1632, -0.1666, -0.1576, -0.1719, -0.1859, -0.206, -0.1992, -0.1999, -0.1999, -0.2207, -0.2093, -0.1646, -0.1502, -0.1327, -0.1319]}
{"unit_id": "orig:1948534120545747444", "ticker_idx": 0, "ts": "2025-07-25T00:01:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Investment Rating: Maintained at 'Buy', expecting continued AI demand and memory demand recovery", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates Buy on SK Hynix despite PT cut to ₩380K from ₩430K on 2026 HBM ASP/oversupply risk; author shares/adopts the call.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Peter Lee, a guru and top analyst in the IT industry from Citi, released a note today regarding SK Hynix: \n\nQ2 Review: Operating Profit Meets Market Expectations; 2026 HBM ASP Growth Potentially Slowing Due to Intensifying Competition\n\nSK Hynix announced Q2 operating profit of KRW 9.2 trillion, aligning with the market consensus of KRW 9.0 trillion. The company recorded solid performance with strong increases in DRAM and NAND bit shipments, up +mid-20% and +70% QoQ respectively. However, with 2026 HBM contracts yet to be finalized and intensifying competition anticipated, we project SK Hynix's 2026 HBM ASP growth to turn negative, at -6% YoY (previously +7% YoY). We have consequently lowered our target price from KRW 430,000 to KRW 380,000.\n\nTarget Price Downgrade (₩430,000 → ₩380,000)\n\nHBM Supply Increase Risk: Potential for oversupply starting in 2026.\n\nASP Scenario:\nHBM5 Introduction Based ASP: Lowered from +3% to -3%\nHBM ASP Growth Rate: Adjusted from +7% to -6% YoYEPS Impact: 2026E/2027E EPS projected to be lowered by -12%/-14% respectively.\n\nTarget Price Calculation Basis: Applying a P/E multiple of 2.4x (down from 2.6x).\nInvestment Rating: Maintained at 'Buy', expecting continued AI demand and memory demand recovery.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0131580235748463, "ret_signed_m1d": 0.0131580235748463, "alpha_spy_m1d": 0.017364603152240643, "alpha_c_m1d": 0.01207967937146548, "bench_spy_m1d": -0.004206579577394343, "bench_c_m1d": 0.001078344203380821, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015037378885368491, "ret_signed_p1d": -0.015037378885368491, "alpha_spy_p1d": -0.014786278740176795, "alpha_c_p1d": -0.028672730941803892, "bench_spy_p1d": -0.0002511001451916961, "bench_c_p1d": 0.013635352056435401, "ret_p1w": -0.030075052701107996, "ret_signed_p1w": -0.030075052701107996, "alpha_spy_p1w": -0.005934271216396003, "alpha_c_p1w": -0.017761675887584816, "bench_spy_p1w": -0.024140781484711993, "bench_c_p1w": -0.01231337681352318, "ret_p1m": -0.02443592502020564, "ret_signed_p1m": -0.02443592502020564, "alpha_spy_p1m": -0.03286463905124515, "alpha_c_p1m": -0.046802063713138, "bench_spy_p1m": 0.008428714031039508, "bench_c_p1m": 0.022366138692932358, "ret_p3m": 0.812765341481591, "ret_signed_p3m": 0.812765341481591, "alpha_spy_p3m": 0.7616722104762976, "alpha_c_p3m": 0.6349851737722871, "bench_spy_p3m": 0.051093131005293335, "bench_c_p3m": 0.17778016770930383, "ret_p6m": 1.878389472542985, "ret_signed_p6m": 1.878389472542985, "alpha_spy_p6m": 1.7865247887591271, "alpha_c_p6m": 1.4813464803960397, "bench_spy_p6m": 0.09186468378385793, "bench_c_p6m": 0.3970429921469454, "price_path": [-0.3215, -0.3339, -0.3339, -0.302, -0.302, -0.302, -0.284, -0.2859, -0.2866, -0.2682, -0.2551, -0.227, -0.2476, -0.2326, -0.2517, -0.242, -0.2476, -0.2611, -0.2495, -0.2382, -0.2401, -0.2195, -0.203, -0.2312, -0.2199, -0.2199, -0.1823, -0.156, -0.156, -0.1391, -0.1335, -0.0977, -0.1147, -0.1147, -0.0677, -0.0639, -0.0733, -0.0752, -0.0338, -0.0244, 0.047, 0.0752, 0.1015, 0.0677, 0.0977, 0.0733, 0.0489, 0.047, 0.0169, 0.0188, 0.0602, 0.0564, 0.1165, 0.1071, 0.1278, 0.1222, 0.1128, 0.0132, 0.0113, 0.0244, 0.0094, 0.0113, 0.0132, 0.0, -0.015, -0.0132, -0.0094, 0.0282, -0.0301, -0.0301, -0.0094, -0.0282, -0.015, -0.0357, 0.0038, 0.0113, 0.0451, 0.0395, 0.0395, 0.0056, -0.0113, -0.0395, -0.0789, -0.0564, -0.0244, -0.0169, -0.0226, 0.0109, 0.0127, -0.0362, -0.0193, -0.0117, -0.0004, 0.0297, 0.0429, 0.0843, 0.1445, 0.1558, 0.2367, 0.2462, 0.3102, 0.2556, 0.3384, 0.3384, 0.3215, 0.3591, 0.3459, 0.3422, 0.2669, 0.3139, 0.3083, 0.3553, 0.489, 0.489, 0.489, 0.489, 0.489, 0.489, 0.6113, 0.5624, 0.5492, 0.5906, 0.7036, 0.7525, 0.8278, 0.8034, 0.8128, 0.8015, 0.9201, 1.0142, 0.9615, 1.1008, 1.1384, 1.1045, 1.3342, 1.2062, 1.1798, 1.2325, 1.1836, 1.2815, 1.3304, 1.3229, 1.3041, 1.1083, 1.2815, 1.146, 1.1158, 1.1497, 0.9615, 0.9577, 0.9539, 0.9728, 1.0495, 0.9968, 1.0269, 1.1023, 1.0797, 1.042, 1.0495, 1.1739, 1.1324, 1.2115, 1.1287, 1.1513, 1.0872, 0.9968, 1.0759, 1.0797, 1.0608, 1.1852, 1.2002, 1.2153, 1.2153, 1.2567, 1.4112, 1.4527, 1.4527, 1.4527, 1.5506, 1.6222, 1.7352, 1.7955, 1.8482, 1.803, 1.8219, 1.7804, 1.7955, 1.8219, 1.8482, 1.8784]}
{"unit_id": "orig:1973174172034670801", "ticker_idx": 0, "ts": "2025-09-30T23:52:37+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price raised to NT$1,600 (from NT$1,370), implying 23% upside from current levels. ADR target price raised to USD 320 (vs. current price of USD 273, 17% upside).", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Goldman Sachs Buy on TSMC with PT raised to NT$1,600 / ADR $320, implying 17-23% upside on AI-driven N2 ramp and revenue CAGR of ~19% to 2030.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC 2027 Growth Re-acceleration and AI-driven Long-term Growth Momentum, Target Price Raised (Goldman Sachs)\n\n1. Growth Outlook\n    • 2026 revenue is expected to grow +18.9% YoY in USD terms, and +20.4% in 2027, revised upward from previous forecasts.\n    • N2 process demand expansion: after initial demand from smartphone and HPC clients, AI will drive the second wave of demand in 2027. N2 capacity/shipments are revised up to 1,230k/990k (from 990k/780k).\n    • N2 revenue contribution is now projected at 9.6% in 2026 and 21.4% in 2027, higher than prior forecasts.\n    • 2026E Capex raised to USD 44bn (from 42bn); 2027E Capex maintained at USD 50bn.\n\n2. AI Demand Momentum\n    • According to Nvidia, the AI infrastructure market is expected to grow from USD 600bn in 2025 to USD 3–4tn by 2030, a CAGR of about 42%.\n    • TSMC’s AI revenue is also expected to expand at a similar growth rate (mid-40s%). Even under conservative assumptions, total revenue CAGR of 17.2% from 2026–2030 is achievable.\n    • Assuming 5% growth in non-AI revenue, total revenue could reach USD 292bn by 2030, a CAGR of 19.3%.\n\n3. FX and Profitability\n    • USD/TWD assumption revised up to 30 (from 29), with positive implications for revenue and margin.\n    • Since most revenue is USD-based and ~75% of costs are TWD-based, NT dollar weakness benefits gross margin (a 1% NT appreciation reduces GM by ~40bps).\n    • 2025–2027E GM revised up to 58.5% / 58.2% / 58.2% (from 57.9% / 57.9% / 57.4%).\n\n4. Earnings and Valuation\n    • EPS estimates revised up by +2.1%, +3.6%, and +9.4% for 2025–2027E.\n    • 2027E EPS projected at NT$84.8, implying P/E of 15.3x — still undervalued compared with long-term averages.\n    • Target price raised to NT$1,600 (from NT$1,370), implying 23% upside from current levels.\n    • ADR target price raised to USD 320 (vs. current price of USD 273, 17% upside).\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02169787584230709, "ret_signed_m1d": -0.02169787584230709, "alpha_spy_m1d": -0.017945145119809713, "alpha_c_m1d": -0.010360749678332226, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03286890388277075, "ret_signed_p1d": 0.03286890388277075, "alpha_spy_p1d": 0.02946133445683774, "alpha_c_p1d": 0.010408984349961736, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.052776523404409525, "ret_signed_p1w": 0.052776523404409525, "alpha_spy_p1w": 0.04836332386883857, "alpha_c_p1w": 0.020021219511928923, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.09237696245340055, "ret_signed_p1m": 0.09237696245340055, "alpha_spy_p1m": 0.060538735296172375, "alpha_c_p1m": -0.037081359010335824, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.08724802202716253, "ret_signed_p3m": 0.08724802202716253, "alpha_spy_p3m": 0.047964685974125176, "alpha_c_p3m": -0.037272203156406336, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 0.25200071636622146, "ret_signed_p6m": 0.25200071636622146, "alpha_spy_p6m": 0.2604369891881233, "alpha_c_p6m": 0.025558727999474007, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.1619, -0.1619, -0.182, -0.1867, -0.1725, -0.1799, -0.1776, -0.1838, -0.1543, -0.1521, -0.1234, -0.142, -0.1475, -0.1627, -0.1422, -0.1377, -0.1234, -0.1336, -0.1386, -0.133, -0.1376, -0.1605, -0.1469, -0.1703, -0.1742, -0.134, -0.1368, -0.1359, -0.1281, -0.1382, -0.1398, -0.1474, -0.1383, -0.1694, -0.1841, -0.1886, -0.1684, -0.1591, -0.1479, -0.1459, -0.1496, -0.176, -0.176, -0.1848, -0.1741, -0.1605, -0.1312, -0.1177, -0.1044, -0.0704, -0.0759, -0.0744, -0.0671, -0.0617, -0.0591, -0.0381, -0.0516, -0.0238, 0.0122, 0.0051, -0.0094, -0.0212, -0.0217, 0.0, 0.0329, 0.0316, 0.0462, 0.0827, 0.0528, 0.0903, 0.0737, 0.0049, 0.0845, 0.0596, 0.091, 0.0736, 0.0565, 0.0659, 0.0545, 0.0343, 0.041, 0.0561, 0.0679, 0.0796, 0.0924, 0.0857, 0.0757, 0.0916, 0.0528, 0.0514, 0.0356, 0.0258, 0.0572, 0.0425, 0.0406, 0.0104, 0.0198, 0.0097, -0.0049, 0.011, -0.0064, -0.0151, 0.0192, 0.0193, 0.0382, 0.0382, 0.0438, 0.03, 0.0458, 0.0579, 0.0488, 0.0552, 0.0808, 0.0864, 0.1105, 0.0945, 0.0485, 0.033, 0.0299, -0.0057, 0.0221, 0.0374, 0.0529, 0.0661, 0.0727, 0.0727, 0.0872, 0.0804, 0.0755, 0.091, 0.091, 0.1475, 0.1569, 0.1755, 0.1441, 0.1417, 0.1619, 0.1911, 0.1891, 0.1744, 0.2265, 0.2293, 0.2293, 0.1746, 0.1708, 0.1753, 0.2022, 0.1945, 0.2147, 0.2289, 0.219, 0.1868, 0.2255, 0.2054, 0.1695, 0.1874, 0.2524, 0.276, 0.2993, 0.343, 0.3215, 0.3153, 0.3153, 0.3075, 0.3006, 0.2939, 0.3303, 0.3285, 0.3849, 0.392, 0.3528, 0.3448, 0.3252, 0.2678, 0.2833, 0.2704, 0.2167, 0.2519, 0.2461, 0.2729, 0.2088, 0.2146, 0.2215, 0.2456, 0.2226, 0.2197, 0.1854, 0.2185, 0.2358, 0.252]}
{"unit_id": "thread:1968470173767139747", "ticker_idx": 0, "ts": "2025-09-18T00:20:36+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung reportedly failed 12-HI HBM3E qualification, implying MU benefits while Samsung loses share.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SAMSUNG SEEMS TO HAVE ULTIMATELY FAILED THE QUALIFICATION TEST FOR ITS 12-HI HBM3E — ACCORDING TO A LOCAL KOREAN SECURITIES REPORT.\n\n$MU\n\n---\n\nhttps://t.co/u4jHzGD8ir", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05269699497065694, "ret_signed_m1d": -0.05269699497065694, "alpha_spy_m1d": -0.04804628003871869, "alpha_c_m1d": -0.01572185769617418, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.03697513727448276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03647345928074219, "ret_signed_p1d": -0.03647345928074219, "alpha_spy_p1d": -0.04142633405607421, "alpha_c_p1d": -0.032498281926118366, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": -0.003975177354623827, "ret_p1w": -0.07140733481006467, "ret_signed_p1w": -0.07140733481006467, "alpha_spy_p1w": -0.0678051173230122, "alpha_c_p1w": -0.08431072573485843, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": 0.012903390924793756, "ret_p1m": 0.19904521559148836, "ret_signed_p1m": 0.19904521559148836, "alpha_spy_p1m": 0.19304756111772892, "alpha_c_p1m": 0.11746030194970536, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": 0.081584913641783, "ret_p3m": 0.3775570381578992, "ret_signed_p3m": 0.3775570381578992, "alpha_spy_p3m": 0.3496342527776397, "alpha_c_p3m": 0.267231186611844, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": 0.11032585154605523, "ret_p6m": 1.5257217070270563, "ret_signed_p6m": 1.5257217070270563, "alpha_spy_p6m": 1.5199406112844895, "alpha_c_p6m": 1.29994264325086, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": 0.22577906377619628, "price_path": [-0.2778, -0.2434, -0.2473, -0.2547, -0.262, -0.2709, -0.2849, -0.2799, -0.2766, -0.2766, -0.29, -0.2633, -0.2762, -0.2711, -0.2627, -0.2977, -0.2888, -0.3106, -0.3294, -0.3227, -0.3296, -0.3533, -0.3497, -0.3384, -0.3412, -0.3413, -0.3371, -0.3206, -0.3538, -0.379, -0.3619, -0.3543, -0.3559, -0.3376, -0.2961, -0.2675, -0.2436, -0.2642, -0.2582, -0.2843, -0.2685, -0.2773, -0.306, -0.3144, -0.3032, -0.3107, -0.3102, -0.3028, -0.2776, -0.2953, -0.2953, -0.2985, -0.2971, -0.2646, -0.2222, -0.2216, -0.1992, -0.1711, -0.1085, -0.069, -0.0658, -0.0596, -0.0527, 0.0, -0.0365, -0.0253, -0.0147, -0.0425, -0.0714, -0.0688, -0.0295, -0.0093, 0.0785, 0.088, 0.1128, 0.1314, 0.1002, 0.1644, 0.1395, 0.0759, 0.1421, 0.1083, 0.1372, 0.1999, 0.199, 0.2251, 0.1985, 0.1759, 0.2247, 0.2976, 0.304, 0.3148, 0.3427, 0.3272, 0.3258, 0.3905, 0.2918, 0.4071, 0.412, 0.4096, 0.5007, 0.4285, 0.451, 0.4039, 0.4624, 0.4335, 0.3538, 0.3385, 0.1931, 0.2286, 0.3267, 0.3303, 0.3642, 0.3642, 0.4011, 0.4247, 0.4189, 0.3873, 0.3428, 0.4055, 0.4629, 0.4955, 0.5624, 0.5313, 0.4287, 0.4071, 0.3776, 0.3361, 0.4726, 0.5755, 0.6387, 0.6368, 0.6985, 0.6985, 0.6873, 0.7448, 0.7345, 0.6917, 0.6917, 0.8695, 0.8501, 1.0355, 1.0126, 0.9383, 1.0454, 1.05, 1.0041, 0.9758, 0.9952, 1.1501, 1.1501, 1.1634, 1.3063, 1.3565, 1.3688, 1.3062, 1.4315, 1.58, 1.583, 1.459, 1.5949, 1.4861, 1.2487, 1.2694, 1.3394, 1.273, 1.2123, 1.4321, 1.4536, 1.44, 1.44, 1.3695, 1.495, 1.4737, 1.5378, 1.4951, 1.4776, 1.5427, 1.4631, 1.4442, 1.4459, 1.2504, 1.3754, 1.3534, 1.1948, 1.3075, 1.3893, 1.4816, 1.4026, 1.5257]}
{"unit_id": "thread:1968470173767139747", "ticker_idx": 1, "ts": "2025-09-18T00:20:36+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SAMSUNG SEEMS TO HAVE ULTIMATELY FAILED THE QUALIFICATION TEST FOR ITS 12-HI HBM3E", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung reportedly failed 12-HI HBM3E qualification, implying MU benefits while Samsung loses share.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SAMSUNG SEEMS TO HAVE ULTIMATELY FAILED THE QUALIFICATION TEST FOR ITS 12-HI HBM3E — ACCORDING TO A LOCAL KOREAN SECURITIES REPORT.\n\n$MU\n\n---\n\nhttps://t.co/u4jHzGD8ir", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026151991166487476, "ret_signed_m1d": 0.026151991166487476, "alpha_spy_m1d": 0.021501276234549227, "alpha_c_m1d": 0.009166500459552696, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.01698549070693478, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.004952874775332017, "alpha_c_p1d": 0.010140492090183084, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": 0.010140492090183084, "ret_p1w": 0.07222917773094406, "ret_signed_p1w": -0.07222917773094406, "alpha_spy_p1w": -0.07583139521799653, "alpha_c_p1w": -0.06412654947030916, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": 0.008102628260634903, "ret_p1m": 0.22461769877002147, "ret_signed_p1m": -0.22461769877002147, "alpha_spy_p1m": -0.21862004429626203, "alpha_c_p1m": -0.19109492755042035, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": 0.03352277121960112, "ret_p3m": 0.28591106520245746, "ret_signed_p3m": -0.28591106520245746, "alpha_spy_p3m": -0.25798827982219796, "alpha_c_p3m": -0.2519894656990731, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": 0.03392159950338436, "ret_p6m": 1.3065343371701958, "ret_signed_p6m": -1.3065343371701958, "alpha_spy_p6m": -1.300753241427629, "alpha_c_p6m": -1.3128824095057061, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": -0.006348072335510269, "price_path": [-0.2821, -0.2512, -0.2413, -0.2549, -0.2428, -0.2553, -0.2503, -0.2428, -0.2055, -0.2117, -0.2316, -0.2354, -0.2478, -0.2403, -0.2204, -0.2217, -0.2067, -0.1943, -0.1694, -0.1644, -0.1557, -0.1781, -0.1731, -0.1781, -0.1793, -0.1233, -0.1208, -0.0959, -0.1108, -0.142, -0.132, -0.1295, -0.1432, -0.122, -0.1059, -0.1158, -0.1146, -0.1046, -0.1083, -0.1083, -0.1283, -0.1283, -0.122, -0.1208, -0.1108, -0.1096, -0.1245, -0.1208, -0.1333, -0.132, -0.1582, -0.1395, -0.1308, -0.127, -0.1345, -0.127, -0.1096, -0.0959, -0.0859, -0.061, -0.0473, -0.0112, -0.0262, 0.0, 0.0, 0.0399, 0.0548, 0.0635, 0.0722, 0.0374, 0.0532, 0.0495, 0.0758, 0.1227, 0.1227, 0.1227, 0.1227, 0.1227, 0.1227, 0.1808, 0.1671, 0.1458, 0.1883, 0.2221, 0.2246, 0.2271, 0.2196, 0.2334, 0.2071, 0.2359, 0.2759, 0.2446, 0.2571, 0.3022, 0.3447, 0.3897, 0.3122, 0.2584, 0.2409, 0.2246, 0.2584, 0.2947, 0.2897, 0.2859, 0.2159, 0.2584, 0.2234, 0.2071, 0.2584, 0.1858, 0.2096, 0.2421, 0.2859, 0.2947, 0.2571, 0.2609, 0.2934, 0.3072, 0.3147, 0.356, 0.3697, 0.356, 0.351, 0.3422, 0.3622, 0.3109, 0.2859, 0.3497, 0.346, 0.3297, 0.3822, 0.3947, 0.3897, 0.3897, 0.4635, 0.5021, 0.5071, 0.5071, 0.5071, 0.6152, 0.7359, 0.7459, 0.7723, 0.7447, 0.7472, 0.7447, 0.7296, 0.7635, 0.8088, 0.8716, 0.8767, 0.8251, 0.8792, 0.9144, 0.9118, 0.9118, 1.0049, 1.0413, 1.0199, 1.0174, 0.8905, 1.1054, 1.1255, 1.0023, 0.9935, 1.0916, 1.0841, 1.1092, 1.2449, 1.2776, 1.2776, 1.2776, 1.2776, 1.3882, 1.3895, 1.4259, 1.5139, 1.5579, 1.7402, 1.7213, 1.7213, 1.4523, 1.1645, 1.4083, 1.3656, 1.1808, 1.3618, 1.3882, 1.3618, 1.3065]}
{"unit_id": "orig:1973659133661638765", "ticker_idx": 0, "ts": "2025-10-02T07:59:40+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we raise our target price to 560,000 KRW… We maintain SK hynix as our top pick in the sector", "tweet_type": "news_reaction", "timeline": "months", "summary": "KIS raises SK Hynix TP to 560,000 KRW (+36%), top pick on memory upcycle through 2027, 2026 OP raised to 70T KRW.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS: SK hynix (000660) – Target Price 560,000 KRW\n\n3Q25 Preview: Venturing into Uncharted Territory\n\n3Q25 Earnings Expected to Beat Consensus\nWe project Q3 revenue of 25.1 trillion KRW (+13% QoQ, +43% YoY) and operating profit of 12 trillion KRW (+31% QoQ, +71% YoY), exceeding the consensus estimate of 10.8 trillion KRW by 11%. Server DRAM demand is surging, and DRAM bit growth in Q3 is expected to surpass guidance. NAND will also exceed expectations, driven by soaring demand for QLC eSSDs used in AI storage, boosting both revenue and profit at subsidiary Solidigm. HBM sales once again surpassed plan in Q3, accounting for 43% of DRAM revenue and 48% of total operating profit, driving the strong performance.\n\nUpward Revision of 2026 ASP and Profit Estimates\nConventional DRAM demand is rapidly rising, led by OpenAI and others supporting AI workloads, which is pushing server DRAM ASP higher. While hyperscalers are aggressively expanding data centers and generating simultaneous server demand, memory suppliers have maintained conservative investment over the past two years, making it difficult to increase supply in the short term. We raise our 2026 ASP forecast for conventional DRAM from +15% to +30%, and revise HBM ASP from –2% decline to +8% increase. Reflecting this, we lift our 2026 annual operating profit forecast from 54 trillion KRW to 70 trillion KRW. Although annual HBM contracts are still under negotiation, the narrowing margin gap with conventional DRAM, combined with supply shortages requiring vendor allocation, suggests HBM prices will likely rise beyond prior expectations.\n\nTarget Price Raised, Maintains Top Pick in the Sector\nBased on the upward earnings revision, we raise our target price to 560,000 KRW (12MF BPS 245,997 KRW, target PBR 2.5x), a 36% increase from the previous target. We expect the memory upcycle to be the longest in history, lasting until 2027. Given the transformed nature of the memory cycle and the elevated profit levels, we set the target PBR at 2.5x. Considering supply conditions unlikely to be resolved in the near term, both ASPs and earnings still carry upside risk. We maintain SK hynix as our top pick in the sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08975981455193383, "ret_signed_m1d": -0.08975981455193383, "alpha_spy_m1d": -0.08860942786259207, "alpha_c_m1d": -0.07630782022987881, "bench_spy_m1d": -0.0011503866893417536, "bench_c_m1d": -0.013451994322055016, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 1.4767362221612679e-05, "alpha_c_p1d": 0.004612153712874267, "bench_spy_p1d": -1.4767362221612679e-05, "bench_c_p1d": -0.004612153712874267, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.0028989891327806383, "alpha_c_p1w": -0.020606624920762595, "bench_spy_p1w": 0.0028989891327806383, "bench_c_p1w": 0.020606624920762595, "ret_p1m": 0.41340069543024827, "ret_signed_p1m": 0.41340069543024827, "alpha_spy_p1m": 0.39421403954895373, "alpha_c_p1m": 0.34013913164948417, "bench_spy_p1m": 0.019186655881294534, "bench_c_p1m": 0.0732615637807641, "ret_p3m": 0.6471964314335994, "ret_signed_p3m": 0.6471964314335994, "alpha_spy_p3m": 0.6175796151319723, "alpha_c_p3m": 0.5697352374215796, "bench_spy_p3m": 0.029616816301627047, "bench_c_p3m": 0.07746119401201979, "ret_p6m": 1.3650846382521182, "ret_signed_p6m": 1.3650846382521182, "alpha_spy_p6m": 1.4121833514645323, "alpha_c_p6m": 1.2551788977860445, "bench_spy_p6m": -0.04709871321241399, "bench_c_p6m": 0.10990574046607371, "price_path": [-0.3158, -0.288, -0.2905, -0.2501, -0.2564, -0.2426, -0.2463, -0.2527, -0.3196, -0.3208, -0.312, -0.3221, -0.3208, -0.3196, -0.3284, -0.3385, -0.3372, -0.3347, -0.3095, -0.3486, -0.3486, -0.3347, -0.3473, -0.3385, -0.3524, -0.3259, -0.3208, -0.2981, -0.3019, -0.3019, -0.3246, -0.336, -0.3549, -0.3814, -0.3663, -0.3448, -0.3398, -0.3436, -0.3211, -0.3198, -0.3527, -0.3413, -0.3363, -0.3287, -0.3085, -0.2996, -0.2718, -0.2314, -0.2238, -0.1694, -0.1631, -0.1201, -0.1568, -0.1011, -0.1011, -0.1125, -0.0872, -0.0961, -0.0986, -0.1492, -0.1176, -0.1214, -0.0898, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0822, 0.0493, 0.0405, 0.0683, 0.1441, 0.177, 0.2276, 0.2111, 0.2174, 0.2099, 0.2895, 0.3527, 0.3173, 0.4109, 0.4362, 0.4134, 0.5676, 0.4817, 0.464, 0.4994, 0.4665, 0.5322, 0.5651, 0.5601, 0.5474, 0.4159, 0.5322, 0.4412, 0.421, 0.4437, 0.3173, 0.3148, 0.3123, 0.3249, 0.3765, 0.341, 0.3613, 0.4119, 0.3967, 0.3714, 0.3765, 0.46, 0.4321, 0.4853, 0.4296, 0.4448, 0.4018, 0.341, 0.3942, 0.3967, 0.384, 0.4675, 0.4777, 0.4878, 0.4878, 0.5156, 0.6194, 0.6472, 0.6472, 0.6472, 0.713, 0.7611, 0.837, 0.8774, 0.9129, 0.8825, 0.8952, 0.8673, 0.8774, 0.8952, 0.9129, 0.9331, 0.88, 0.8724, 0.9103, 0.9407, 0.8623, 1.0242, 1.1279, 1.1786, 1.3, 1.1001, 1.2949, 1.2772, 1.1305, 1.1229, 1.2443, 1.2165, 1.176, 1.2469, 1.2266, 1.2266, 1.2266, 1.2266, 1.262, 1.4012, 1.4063, 1.5429, 1.5758, 1.7859, 1.6896, 1.6896, 1.3803, 1.1522, 1.3854, 1.3423, 1.1192, 1.3778, 1.4209, 1.3575, 1.3068, 1.469, 1.4589, 1.6769, 1.5679, 1.5527, 1.3651, 1.4994, 1.5222, 1.3651, 1.3651]}
{"unit_id": "orig:1968663926591340946", "ticker_idx": 0, "ts": "2025-09-18T13:10:30+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek will drop tomorrow", "tweet_type": "prediction", "timeline": "intraday", "summary": "MediaTek to drop tomorrow as NVIDIA acquires Intel x86 capability, likely canceling NVIDIA's ARM-based N1/N1X; Wintel play favors Intel.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Investment Idea: MediaTek will drop tomorrow.\n\nNVIDIA has now gained Intel’s x86 CPU capabilities.\n\nI wouldn’t be surprised if NVIDIA’s N1/N1X ends up being canceled.\n\nFrom NVIDIA’s perspective, leveraging Wintel is far better than trying to build Windows on ARM from scratch.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.004650714931938249, "alpha_c_m1d": -0.03697513727448276, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.03697513727448276, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0463575951555083, "ret_signed_p1d": 0.0463575951555083, "alpha_spy_p1d": 0.05131046993084032, "alpha_c_p1d": 0.042382417800884475, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": -0.003975177354623827, "ret_p1w": -0.10927152727193912, "ret_signed_p1w": 0.10927152727193912, "alpha_spy_p1w": 0.10566930978488664, "alpha_c_p1w": 0.12217491819673287, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": 0.012903390924793756, "ret_p1m": -0.11920528000346775, "ret_signed_p1m": 0.11920528000346775, "alpha_spy_p1m": 0.1252029344772272, "alpha_c_p1m": 0.20079019364525075, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": 0.081584913641783, "ret_p3m": -0.05960268120592127, "ret_signed_p3m": 0.05960268120592127, "alpha_spy_p3m": 0.08752546658618077, "alpha_c_p3m": 0.1699285327519765, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": 0.11032585154605523, "ret_p6m": 0.16115377987849544, "ret_signed_p6m": -0.16115377987849544, "alpha_spy_p6m": -0.15537268413592864, "alpha_c_p6m": 0.06462528389770084, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": 0.22577906377619628, "price_path": [-0.1849, -0.1752, -0.1589, -0.1622, -0.1654, -0.1882, -0.1719, -0.1589, -0.1589, -0.1457, -0.1523, -0.1589, -0.1026, -0.0728, -0.0596, -0.0828, -0.0762, -0.0662, -0.0795, -0.0662, -0.0563, -0.053, -0.053, -0.053, -0.053, -0.0728, -0.0993, -0.0861, -0.0927, -0.0927, -0.1192, -0.1126, -0.1258, -0.1026, -0.106, -0.1026, -0.0993, -0.0927, -0.0795, -0.0927, -0.0662, -0.0795, -0.1026, -0.096, -0.096, -0.0662, -0.0728, -0.0728, -0.0828, -0.0927, -0.0993, -0.0993, -0.0762, -0.0828, -0.0497, -0.053, 0.0, -0.0132, -0.0199, -0.0166, -0.0166, 0.0166, 0.0, 0.0, -0.0464, -0.0695, -0.0894, -0.1192, -0.1093, -0.1325, -0.1325, -0.1291, -0.149, -0.1523, -0.1258, -0.1258, -0.1225, -0.1159, -0.1093, -0.1093, -0.1291, -0.1358, -0.1291, -0.1192, -0.1192, -0.1126, -0.1126, -0.1192, -0.1424, -0.1424, -0.1159, -0.1291, -0.1391, -0.1325, -0.1325, -0.1391, -0.1291, -0.1192, -0.1457, -0.1656, -0.1722, -0.1788, -0.1755, -0.1755, -0.1854, -0.1854, -0.2252, -0.2318, -0.2152, -0.2417, -0.2384, -0.2152, -0.1391, -0.1126, -0.0762, -0.043, -0.0629, -0.0728, -0.0695, -0.0563, -0.0464, -0.0596, -0.0331, -0.0762, -0.0695, -0.0596, -0.0596, -0.0563, -0.0596, -0.0662, -0.0728, -0.0795, -0.0861, -0.0861, -0.0828, -0.0596, -0.0596, -0.053, -0.053, -0.0265, 0.0099, 0.0025, 0.0059, -0.0245, -0.0414, -0.0245, 0.0025, 0.0126, 0.0059, 0.016, 0.0025, 0.0025, -0.011, 0.0025, 0.1004, 0.1949, 0.1983, 0.2017, 0.2017, 0.1882, 0.151, 0.2118, 0.2152, 0.1949, 0.1544, 0.2354, 0.2455, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2118, 0.232, 0.2624, 0.313, 0.313, 0.2827, 0.2253, 0.1612, 0.1983, 0.1915, 0.124, 0.151, 0.1915, 0.205, 0.1612]}
{"unit_id": "orig:1968663926591340946", "ticker_idx": 1, "ts": "2025-09-18T13:10:30+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA has now gained Intel's x86 CPU capabilities. From NVIDIA's perspective, leveraging Wintel is far better than trying to build Windows on ARM from scratch", "tweet_type": "prediction", "timeline": "intraday", "summary": "MediaTek to drop tomorrow as NVIDIA acquires Intel x86 capability, likely canceling NVIDIA's ARM-based N1/N1X; Wintel play favors Intel.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Investment Idea: MediaTek will drop tomorrow.\n\nNVIDIA has now gained Intel’s x86 CPU capabilities.\n\nI wouldn’t be surprised if NVIDIA’s N1/N1X ends up being canceled.\n\nFrom NVIDIA’s perspective, leveraging Wintel is far better than trying to build Windows on ARM from scratch.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.18547596116770415, "ret_signed_m1d": -0.18547596116770415, "alpha_spy_m1d": -0.1808252462357659, "alpha_c_m1d": -0.1485008238932214, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.03697513727448276, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03238468370955794, "ret_signed_p1d": -0.03238468370955794, "alpha_spy_p1d": -0.037337558484889954, "alpha_c_p1d": -0.02840950635493411, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": -0.003975177354623827, "ret_p1w": 0.11187445265894502, "ret_signed_p1w": 0.11187445265894502, "alpha_spy_p1w": 0.1154766701459975, "alpha_c_p1w": 0.09897106173415127, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": 0.012903390924793756, "ret_p1m": 0.21066400690214393, "ret_signed_p1m": 0.21066400690214393, "alpha_spy_p1m": 0.2046663524283845, "alpha_c_p1m": 0.12907909326036093, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": 0.081584913641783, "ret_p3m": 0.22047764951754134, "ret_signed_p3m": 0.22047764951754134, "alpha_spy_p3m": 0.19255486413728184, "alpha_c_p3m": 0.11015179797148611, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": 0.11032585154605523, "ret_p6m": 0.4972195261589405, "ret_signed_p6m": 0.4972195261589405, "alpha_spy_p6m": 0.4914384304163737, "alpha_c_p6m": 0.2714404623827442, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": 0.22577906377619628, "price_path": [-0.3068, -0.2623, -0.2738, -0.264, -0.2578, -0.2673, -0.2525, -0.2843, -0.2643, -0.2643, -0.2803, -0.2283, -0.2332, -0.2208, -0.2336, -0.2378, -0.2502, -0.2578, -0.2542, -0.2444, -0.2391, -0.2398, -0.2316, -0.2597, -0.3229, -0.3235, -0.3324, -0.3346, -0.3523, -0.3683, -0.3621, -0.3395, -0.3324, -0.3533, -0.3474, -0.3245, -0.2866, -0.2731, -0.2195, -0.1966, -0.226, -0.1721, -0.23, -0.2313, -0.1887, -0.1969, -0.2035, -0.1871, -0.1845, -0.2035, -0.2035, -0.208, -0.2149, -0.195, -0.1989, -0.1992, -0.2005, -0.1897, -0.195, -0.2123, -0.1897, -0.1734, -0.1855, 0.0, -0.0324, -0.0592, -0.0402, 0.0213, 0.1119, 0.1613, 0.1279, 0.0975, 0.1757, 0.2202, 0.2048, 0.1969, 0.2159, 0.2244, 0.2365, 0.1897, 0.2175, 0.1655, 0.2152, 0.2051, 0.2107, 0.2463, 0.247, 0.2077, 0.2483, 0.2522, 0.2934, 0.3585, 0.3523, 0.3137, 0.3081, 0.2921, 0.2113, 0.2555, 0.2182, 0.2473, 0.2578, 0.2391, 0.2395, 0.1747, 0.1619, 0.1354, 0.123, 0.1485, 0.0998, 0.1286, 0.1708, 0.1721, 0.2041, 0.2041, 0.3268, 0.3088, 0.422, 0.4315, 0.3248, 0.3546, 0.3183, 0.3248, 0.334, 0.2924, 0.2368, 0.227, 0.2205, 0.1793, 0.1868, 0.2044, 0.1897, 0.1891, 0.1829, 0.1829, 0.1842, 0.1999, 0.2202, 0.2071, 0.2071, 0.2882, 0.2879, 0.3098, 0.3945, 0.3448, 0.49, 0.4413, 0.5469, 0.5937, 0.5806, 0.5361, 0.5361, 0.5885, 0.7746, 0.7769, 0.4743, 0.3899, 0.437, 0.5957, 0.5918, 0.5201, 0.5967, 0.6111, 0.5898, 0.578, 0.6549, 0.6434, 0.5417, 0.5797, 0.5204, 0.5306, 0.5306, 0.5106, 0.4871, 0.4596, 0.4429, 0.4272, 0.5087, 0.5335, 0.4871, 0.492, 0.4884, 0.4099, 0.491, 0.5031, 0.4203, 0.491, 0.5303, 0.5695, 0.4802, 0.4972]}
{"unit_id": "orig:1973007877066666188", "ticker_idx": 0, "ts": "2025-09-30T12:51:49+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "picked up some Nvidia. I think we'll be hearing good news in the next few weeks", "tweet_type": "position_disclosure", "timeline": "weeks", "summary": "Bought Nvidia expecting good news in coming weeks.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I took profits on AMD and picked up some Nvidia.\nI think we’ll be hearing good news in the next few weeks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0253509547073848, "ret_signed_m1d": -0.0253509547073848, "alpha_spy_m1d": -0.02159822398488742, "alpha_c_m1d": -0.014013828543409934, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0035374376225016935, "ret_signed_p1d": 0.0035374376225016935, "alpha_spy_p1d": 0.0001298681965686832, "alpha_c_p1d": -0.01892248191030732, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": -0.008253803011263217, "ret_signed_p1w": -0.008253803011263217, "alpha_spy_p1w": -0.012667002546834172, "alpha_c_p1w": -0.04100910690374382, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.10965811258359404, "ret_signed_p1m": 0.10965811258359404, "alpha_spy_p1m": 0.07781988542636586, "alpha_c_p1m": -0.019800208880142334, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.02122740661082889, "ret_signed_p3m": 0.02122740661082889, "alpha_spy_p3m": -0.018055929442208463, "alpha_c_p3m": -0.10329281857273998, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": -0.0422359419366215, "ret_signed_p6m": -0.0422359419366215, "alpha_spy_p6m": -0.033799669114719655, "alpha_c_p6m": -0.26867793030336895, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.146, -0.146, -0.1519, -0.1425, -0.1271, -0.1205, -0.1161, -0.1207, -0.0852, -0.0816, -0.0728, -0.076, -0.0815, -0.1048, -0.0847, -0.0689, -0.0702, -0.0527, -0.0594, -0.0392, -0.0467, -0.069, -0.0353, -0.0446, -0.0384, -0.0312, -0.0209, -0.0243, -0.0184, -0.0268, -0.0245, -0.0329, -0.0245, -0.0587, -0.06, -0.0622, -0.0461, -0.0363, -0.0258, -0.0267, -0.0344, -0.0665, -0.0665, -0.0847, -0.0856, -0.08, -0.1049, -0.098, -0.0848, -0.0496, -0.0504, -0.047, -0.0473, -0.0627, -0.0873, -0.0554, -0.0531, -0.0159, -0.0437, -0.0515, -0.0476, -0.045, -0.0254, 0.0, 0.0035, 0.0124, 0.0056, -0.0056, -0.0083, 0.0136, 0.0321, -0.0183, 0.0093, -0.0351, -0.0362, -0.0256, -0.018, -0.0211, -0.029, -0.0338, -0.0237, -0.0017, 0.0263, 0.0774, 0.1097, 0.0874, 0.0853, 0.1088, 0.0649, 0.0463, 0.008, 0.0084, 0.0668, 0.0353, 0.0387, 0.0015, 0.0192, 0.0001, -0.028, -0.0003, -0.0318, -0.0413, -0.0216, -0.047, -0.0339, -0.0339, -0.0513, -0.0357, -0.0274, -0.0375, -0.0171, -0.0223, -0.0055, -0.0086, -0.015, -0.0302, -0.0619, -0.0551, -0.0474, -0.0838, -0.0666, -0.0299, -0.0154, 0.0142, 0.0109, 0.0109, 0.0212, 0.0088, 0.0052, -0.0004, -0.0004, 0.0122, 0.0083, 0.0036, 0.0136, -0.0082, -0.0092, -0.0087, -0.0041, -0.0184, 0.0026, -0.0018, -0.0018, -0.0456, -0.0174, -0.0093, 0.0059, -0.0005, 0.0105, 0.0265, 0.0318, 0.0244, -0.0051, -0.0334, -0.0664, -0.0787, -0.0062, 0.0186, 0.0106, 0.0187, 0.002, -0.0202, -0.0202, -0.0086, 0.0076, 0.0071, 0.0174, 0.0267, 0.0337, 0.0482, -0.009, -0.0503, -0.0219, -0.0349, -0.0189, -0.0173, -0.0469, -0.021, -0.0096, -0.0028, -0.0183, -0.0338, -0.0179, -0.0248, -0.033, -0.0429, -0.0743, -0.0585, -0.0609, -0.0422]}
{"unit_id": "orig:1961024466907230360", "ticker_idx": 0, "ts": "2025-08-28T11:14:01+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our Buy rating on NVIDIA (NVDA), our top pick in the sector... We raise our PO from $220 to $235", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA, raises PO to $235; top sector pick on 80%+ AI infra market share, $10+ EPS power by CY27, Blackwell ramp, and compelling PEG vs. peers.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) NVIDIA: Solid Position Amid the Most Attractive Visible Growth Opportunity, Price Objective (PO) Raised to $235\n\n1. Ignore China-related noise, focus on ~ $10 EPS potential in CY27\n\nWe maintain our Buy rating on NVIDIA (NVDA), our top pick in the sector. NVIDIA holds a solid position with over 80% market share in the fastest-growing and most attractive global AI infrastructure build-out market, backed by an industry-leading free cash flow (FCF) margin of over 45% and a valuation at roughly 1x earnings growth (compared to the S&P 500 and Mag-7 peers trading at more than 2x PEG). While China-related uncertainty remains, the geopolitical hurdles around exports to China—which we fully exclude from our model—are expected to be more than offset by performance in the rest of the world (FY26E sales of ~$200bn). Overall, we raise our FY26/27/28 pro-forma EPS estimates by 6%/7%/11% to $4.45/$6.26/$8.03, respectively. Our forecasts assume annual revenue growth of $30–35% over 2025–27, and considering NVIDIA’s cited “industry 50% growth,” we estimate EPS power of over $10 by 2027 (with potential to reach $15 longer term). In the near term, given shares are up over 35% within the quarter and expectations are elevated, volatility is possible. However, we expect an upward trend to resume as investors revisit NVIDIA’s strong fundamental outlook. We raise our PO from $220 to $235, continuing to apply a 37x FY26E PE multiple, which aligns with the mid-range of NVIDIA’s historical PE band (25x–56x).\n\n2. Positives: Blackwell ramp, GM, Networking, Rubin\n\nBlackwell Ultra is now in full production mode, producing ~1,000 racks per week (ASP ~$3m, equating to ~$39bn revenue) with further acceleration expected through 3Q. As system-level racks mature and economies of scale kick in, gross margins (GM) could recover faster toward the mid-70% range. Networking (NVLink “switch-up”; Spectrum-X “switch-out”) will also continue to benefit. Spectrum-X currently runs at an annualized revenue rate of $10bn, up from $8bn in April. Finally, next-gen Rubin remains on track for CY26 (FY27) launch, sustaining NVIDIA’s hardware lead of 0.5–1.0 years versus AMD GPUs.\n\n3. Risks: Slower near-term EPS revisions, China uncertainty\n\nDue to China-related uncertainties and policy resets, near-term EPS estimate revisions could be more muted than in recent quarters. However, NVIDIA’s October-quarter revenue guidance of $54bn should be compared not to the current $53.5bn consensus, but to the pre-H20 licensing approval estimate of $51–52bn. Once H20 shipments begin, some previously written-off inventory could be reused, potentially boosting GM by 200–300bps.\n\n4. FY2Q26 Earnings Highlights (Prepared Remarks)\n\n1. Blackwell: FY2Q Blackwell grew +17% QoQ, reaching record levels; GB300 production started in the quarter.\n\n2. Blackwell Ultra: The platform drove a very strong quarter, generating tens of billions in revenue.\n\n3. Late July/early August fab conversions successfully supported GB300 ramp; currently at ~1,000 racks/week with further acceleration expected in 3Q as more capacity comes online.\n\n4. AI TAM: NVIDIA projects AI infrastructure spend of $3–4tn by the late 2020s.\n\n5. Cloud Capex: $600bn this year in data center infrastructure and compute.\n\n6. Rubin: Chips are in-fab and on track for next year’s volume production.\n\n7. China H20: Despite some licenses granted, NVIDIA shipped zero H20 chips to China. The U.S. government is expected to take 15% of licensed H20 sales.\n\n8. H20 was excluded from 3Q guidance; if geopolitics ease, $2–5bn H20 revenue could be realized with additional supply.\n\n9. NVIDIA sold $650m of H20 to non-China customers in 2Q.\n\n10. RTX PRO: Fully ramped, air-cooled PCIe system for enterprise IT apps; ~90 firms adopted; expected multi-billion product line.\n\n11. Sovereign AI: >$20bn sales expected this year, over 2x last year.\n\n12. Networking: 2Q revenue hit a record $7.3bn (+46% QoQ, +98% YoY), with strong demand across SpectrumX Ethernet, InfiniBand, NVLink.\n\n13. SpectrumX: Double-digit QoQ and YoY growth; annualized revenue >$10bn.\n\n14. InfiniBand: Nearly doubled QoQ, driven by XDR adoption.\n\n15. Regional Sales: China fell to low-single-digit % of DC revenue. Singapore represented 22%, with 99%+ billed for U.S.-based customers.\n\n5. FY2Q26 Results vs. Estimates\n\n1. Revenue: $46.7bn (+6% QoQ, +56% YoY), 1% above consensus $46.2bn and 4% above BofA’s $45.1bn.\n\n2. Data Center: $41.1bn, modest miss vs. consensus $41.4bn, but above BofA $39.7bn (+5% QoQ, +56% YoY) despite $4bn H20 headwind.\n\n3. Gaming: $4.3bn (+14% QoQ, +49% YoY), beating consensus $3.9bn by 11% and BofA $4.1bn, driven by Blackwell GeForce ramp.\n\n4. Auto in line; Pro Visualization beat consensus by 13%; OEM beat by 56% (but small segment).\n\n5. EPS: $1.05 (+30% QoQ, +54% YoY), beating consensus $1.01 by 4% and BofA $0.98 by 7%.\n\n6. GM: 72.7% vs. consensus 72.1% (+54bps) and BofA 72.0%. Up 1,170bps QoQ (lapping prior H20 write-offs), but down 301bps YoY. Included $180m benefit (40bps) from reversing H20 reserves.\n\n7. Op. Margin: 64.5% vs. consensus 63.5% (+102bps) and BofA 63.2%. Up 1,171bps QoQ, but down 183bps YoY. Opex rose 6% QoQ (infra + comp).\n\n8. Inventory: Rose sequentially from $11bn to $15bn to support Blackwell/Ultra ramps.\n\n9. Capital Return: $10bn returned via buybacks/dividends; board approved $60bn new authorization, in addition to $14.7bn remaining.\n\n6. FY3Q26 Guidance (Above Consensus)\n\n1. Revenue: $54bn ±2% (+16% QoQ, +54% YoY); midpoint above BofA $50.6bn and consensus $53.5bn. Does not include China H20 shipments.\n\n2. GM: 73.5% ±50bps, above BofA 73.0%, in line with consensus 73.4%. Still guiding mid-70s non-GAAP GM through year-end.\n\n3. Op. Margin: 65.7% implied, above BofA 64.6% and in line with consensus 65.6%.\n\n4. EPS: $1.23 implied, +17% QoQ, +52% YoY, above BofA $1.13 and consensus $1.21.\n\n5. Other: Other income expected +$500m; tax rate 16.5%.\n\n7. Q&A Highlights\n\n1. 2026 Growth Outlook; Networking vs. DC mix: Growth driven by inference/agent AI, which cuts hallucinations and enables tool use, requiring 100–1,000x more compute per task. NVLink72 delivers order-of-magnitude throughput gains, lowering token cost/energy. With cloud capex doubling to $600bn, supports $3–4tn AI infra expansion in 5 years.\n\n2. H20 License; $2–5bn 3Q Potential; ASIC: Licenses being issued; supply ready. If resolved, $2–5bn shipments possible in 3Q. ASICs rarely reach volume and cannot match NVIDIA’s full-stack platform.\n\n3. $3–4tn AI TAM; Share; Power Constraints: At $600bn hyperscaler capex, $3–4tn TAM is “sensible.” A 1GW AI factory ($50–60bn) yields NVIDIA content share of 35%. Power is the bottleneck: tokens per 100MW and perf/watt drive customer P&L.\n\n4. China LT Outlook; Potential Blackwell License: China could be a $50bn opportunity this year, growing ~50% annually. ~50% of AI researchers are in China, with many open-source models originating there. NVIDIA pushing to bring Blackwell into China.\n\n5. Spectrum XGS Opportunity: Tackles “scale-across” challenge by linking AI factories into “super-factories.” Raises utilization from 65% to 85–90%, unlocking $10–20bn profit from $50bn GW-scale factories.\n\n6. + $7bn DC Sequential Growth; Hopper Strength: Blackwell is main driver of 2Q→3Q DC revenue growth, with NVLink boosting networking. H100/H200 (HDX) still shipping, but Blackwell is the key growth engine.\n\n7. Rubin vs. Blackwell: Need annual cadence to improve perf/watt (tokens/energy) for cost savings/revenue gains. Blackwell inference perf/watt is order-of-magnitude over Hopper. Rubin to bring many new ideas (details later).\n\n8. 50% CAGR TAM Visibility: “Very meaningful” visibility from major customer forecasts, continued new logos, and AI-native startup funding surge ($100bn last year → $180bn this year; could 10x next year). H100/H200 sold out, hyperscalers leasing capacity elsewhere. Demand remains robust, underpinning strong growth outlook.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00793695633594016, "ret_signed_m1d": 0.00793695633594016, "alpha_spy_m1d": 0.011465822074469778, "alpha_c_m1d": 0.012018866997477451, "bench_spy_m1d": -0.0035288657385296185, "bench_c_m1d": -0.004081910661537291, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03324642868305783, "ret_signed_p1d": -0.03324642868305783, "alpha_spy_p1d": -0.02728274707907863, "alpha_c_p1d": -0.004505897798071978, "bench_spy_p1d": -0.0059636816039791984, "bench_c_p1d": -0.02874053088498585, "ret_p1w": -0.04723311083269477, "ret_signed_p1w": -0.04723311083269477, "alpha_spy_p1w": -0.047541292559557635, "alpha_c_p1w": -0.017053725131929043, "bench_spy_p1w": 0.00030818172686286616, "bench_c_p1w": -0.030179385700765726, "ret_p1m": -0.010933824193973374, "ret_signed_p1m": -0.010933824193973374, "alpha_spy_p1m": -0.033640718330182784, "alpha_c_p1m": -0.0876871212281789, "bench_spy_p1m": 0.02270689413620941, "bench_c_p1m": 0.07675329703420553, "ret_p3m": -0.01298758502172559, "ret_signed_p3m": -0.01298758502172559, "alpha_spy_p3m": -0.056092239419490264, "alpha_c_p3m": -0.15060142949266386, "bench_spy_p3m": 0.043104654397764675, "bench_c_p3m": 0.13761384447093827, "ret_p6m": 0.053678663766249635, "ret_signed_p6m": 0.053678663766249635, "alpha_spy_p6m": -0.014841829654606453, "alpha_c_p6m": -0.3392597014619718, "bench_spy_p6m": 0.06852049342085609, "bench_c_p6m": 0.39293836522822145, "price_path": [-0.2376, -0.2162, -0.2124, -0.2231, -0.2135, -0.2084, -0.201, -0.2072, -0.1952, -0.212, -0.1969, -0.2001, -0.1925, -0.1925, -0.2016, -0.1998, -0.1791, -0.1435, -0.1396, -0.1244, -0.1231, -0.1491, -0.1272, -0.1156, -0.1156, -0.1217, -0.1119, -0.096, -0.0892, -0.0846, -0.0894, -0.0526, -0.0488, -0.0398, -0.0431, -0.0488, -0.0729, -0.0521, -0.0357, -0.037, -0.019, -0.0259, -0.005, -0.0128, -0.0358, -0.0009, -0.0106, -0.0042, 0.0033, 0.014, 0.0105, 0.0166, 0.0079, 0.0103, 0.0016, 0.0102, -0.0251, -0.0265, -0.0288, -0.0121, -0.002, 0.0089, 0.0079, 0.0, -0.0332, -0.0332, -0.0521, -0.053, -0.0472, -0.073, -0.0658, -0.0522, -0.0158, -0.0166, -0.013, -0.0134, -0.0293, -0.0548, -0.0218, -0.0194, 0.0192, -0.0096, -0.0177, -0.0137, -0.0109, 0.0094, 0.0356, 0.0393, 0.0485, 0.0414, 0.0299, 0.0271, 0.0497, 0.0689, 0.0167, 0.0453, -0.0007, -0.0018, 0.0092, 0.017, 0.0138, 0.0056, 0.0007, 0.0111, 0.0339, 0.0629, 0.1158, 0.1492, 0.1262, 0.1239, 0.1483, 0.1029, 0.0835, 0.044, 0.0444, 0.1049, 0.0722, 0.0757, 0.0372, 0.0556, 0.0357, 0.0067, 0.0353, 0.0027, -0.0071, 0.0133, -0.013, 0.0006, 0.0006, -0.0175, -0.0013, 0.0072, -0.0032, 0.0179, 0.0125, 0.03, 0.0268, 0.0202, 0.0043, -0.0285, -0.0214, -0.0135, -0.0511, -0.0334, 0.0047, 0.0197, 0.0503, 0.047, 0.047, 0.0576, 0.0448, 0.041, 0.0352, 0.0352, 0.0483, 0.0442, 0.0394, 0.0497, 0.0271, 0.0261, 0.0266, 0.0314, 0.0166, 0.0383, 0.0338, 0.0338, -0.0115, 0.0176, 0.026, 0.0417, 0.0351, 0.0465, 0.0631, 0.0686, 0.061, 0.0303, 0.0011, -0.0331, -0.0459, 0.0292, 0.0549, 0.0466, 0.055, 0.0377, 0.0148, 0.0148, 0.0268, 0.0435, 0.043, 0.0537]}
{"unit_id": "orig:1969214391792124049", "ticker_idx": 0, "ts": "2025-09-20T01:37:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix benefits from both its HBM market leadership and surging DDR4/DDR5 prices, emerging as the core beneficiary of the upcycle", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA adopts bullish memory upcycle thesis: SK Hynix as core beneficiary, Micron on HBM+DRAM recovery, Samsung undervalued; DDR5/HBM4/China demand driving the cycle.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "DDR5 Surge, Next-Gen Rubin Memory Architecture, and China’s Demand Recovery Driving the Memory Upcycle (BofA)\n\n1. Industry and Price Cycle\n    • DDR5 spot prices rose 10% this week, hitting record highs and shifting from a gradual recovery into a sharp rally, confirming the upside of the cycle.\n    • DDR4 also climbed 6%, showing that even legacy products are rebounding strongly, signaling tight supply-demand conditions across the board.\n    • Key drivers include PC/server replacement demand, production shifts from DDR5 to HBM4 and GDDR7, and distributors building inventories ahead of anticipated Q4 shortages.\n    • DRAM contract prices: DDR4 surged 150% QoQ in Q3, while DDR5 continued its gradual uptrend. For Q4, DDR5 contract prices are expected to rise by over 10%, expanding earnings momentum.\n\n2. Product and Technology Innovation\n    • NVIDIA’s Rubin platform introduces a hybrid memory architecture of HBM4, GDDR7, and SOCAMM, expanding total memory capacity by 2.5x compared to Blackwell.\n    • This design allocates tailored memory resources: HBM4 for AI training, GDDR7 for inference, and SOCAMM for CPU computing—supporting both qualitative and quantitative growth in memory demand.\n    • SOCAMM does not require TSV processing, lowering costs, and partially substituting DDR5 demand, which could prolong memory shortages.\n\n3. Positioning of Korean Companies\n    • Samsung Electronics has passed NVIDIA’s quality test for its 12-high HBM3e, securing the foundation for full-scale mass production. Although SK hynix’s dominance is unlikely to be challenged in the short term, Samsung is expected to ship about $1 billion worth in Q4, rapidly contributing to revenue.\n    • SK hynix benefits from both its HBM market leadership and surging DDR4/DDR5 prices, emerging as the core beneficiary of the upcycle.\n    • Micron is also set to benefit from both DRAM price recovery and its entry into HBM, strengthening its cyclical rebound.\n\n4. China Demand Recovery\n    • China’s semiconductor equipment imports rose 12% YoY in August, far above market expectations (around 0%), partly influenced by Korean memory makers’ fab expansions in Xi’an and Wuxi.\n    • Investments in equipment and infrastructure are being front-loaded ahead of new U.S. export controls in December, creating a short-term demand boost.\n    • Huawei’s next-gen GPU roadmap (Ascend series) remains at the HBM2 level but includes plans to adopt HBM4 by 2028, strengthening long-term demand momentum.\n\n5. Valuation and Stock Trends\n    • SK hynix and Micron are trading above 2x P/B, a level historically associated with cycle peaks, reflecting strong confidence in the uptrend.\n    • Samsung Electronics and Nanya remain relatively undervalued, leaving room for valuation re-rating.\n    • Overall, AI-driven investment and HBM4 expectations are powering memory stocks higher, with SK hynix and Micron maintaining YTD strength.\n\n6. Conclusion\n    • The combination of a DDR5/HBM-driven price rally, Rubin’s architectural innovation, and China’s unexpected demand recovery is powering a strong global memory upcycle.\n    • Korean memory makers enjoy a triple tailwind: technological leadership in HBM, cyclical price gains in DRAM/NAND, and Chinese investment demand.\n    • With both qualitative and quantitative improvements confirmed, maintaining a positive outlook on the memory sector remains warranted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012820521976864452, "ret_signed_m1d": 0.012820521976864452, "alpha_spy_m1d": 0.01752925290143681, "alpha_c_m1d": 0.032872079731812165, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028489969482429656, "ret_signed_p1d": 0.028489969482429656, "alpha_spy_p1d": 0.033933444322237216, "alpha_c_p1d": 0.029948808796310566, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.0056981188806843885, "ret_signed_p1w": -0.0056981188806843885, "alpha_spy_p1w": -0.0009593796228162255, "alpha_c_p1w": -0.00721900613776616, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.3646722878607491, "ret_signed_p1m": 0.3646722878607491, "alpha_spy_p1m": 0.357999115706531, "alpha_c_p1m": 0.2928779715663701, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5737758511959559, "ret_signed_p3m": 0.5737758511959559, "alpha_spy_p3m": 0.5593346629727354, "alpha_c_p3m": 0.4959287722952972, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.7706148936884407, "ret_signed_p6m": 1.7706148936884407, "alpha_spy_p6m": 1.7617190797680904, "alpha_c_p6m": 1.5348855405453752, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.1864, -0.1664, -0.1921, -0.1693, -0.1878, -0.2063, -0.2077, -0.2305, -0.229, -0.1977, -0.2006, -0.1551, -0.1622, -0.1465, -0.1508, -0.1579, -0.2333, -0.2347, -0.2248, -0.2361, -0.2347, -0.2333, -0.2433, -0.2546, -0.2532, -0.2504, -0.2219, -0.266, -0.266, -0.2504, -0.2646, -0.2546, -0.2703, -0.2404, -0.2347, -0.2091, -0.2134, -0.2134, -0.239, -0.2518, -0.2731, -0.303, -0.2859, -0.2618, -0.2561, -0.2603, -0.235, -0.2336, -0.2707, -0.2578, -0.2521, -0.2436, -0.2208, -0.2108, -0.1795, -0.1339, -0.1254, -0.0641, -0.057, -0.0085, -0.0499, 0.0128, 0.0128, 0.0, 0.0285, 0.0185, 0.0157, -0.0413, -0.0057, -0.01, 0.0256, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.1268, 0.2194, 0.1823, 0.1724, 0.2037, 0.2892, 0.3262, 0.3832, 0.3647, 0.3718, 0.3632, 0.453, 0.5242, 0.4843, 0.5897, 0.6182, 0.5926, 0.7664, 0.6695, 0.6496, 0.6895, 0.6524, 0.7265, 0.7635, 0.7578, 0.7436, 0.5954, 0.7265, 0.6239, 0.6011, 0.6268, 0.4843, 0.4815, 0.4786, 0.4929, 0.551, 0.5111, 0.5339, 0.5909, 0.5738, 0.5453, 0.551, 0.6451, 0.6137, 0.6736, 0.6108, 0.6279, 0.5795, 0.5111, 0.5709, 0.5738, 0.5595, 0.6536, 0.665, 0.6764, 0.6764, 0.7078, 0.8247, 0.856, 0.856, 0.856, 0.9302, 0.9843, 1.0699, 1.1155, 1.1554, 1.1212, 1.1354, 1.1041, 1.1155, 1.1354, 1.1554, 1.1782, 1.1183, 1.1098, 1.1525, 1.1868, 1.0984, 1.2808, 1.3977, 1.4547, 1.5916, 1.3664, 1.5859, 1.5659, 1.4006, 1.392, 1.5289, 1.4975, 1.4519, 1.5317, 1.5089, 1.5089, 1.5089, 1.5089, 1.5488, 1.7056, 1.7113, 1.8653, 1.9024, 2.1391, 2.0305, 2.0305, 1.6821, 1.425, 1.6878, 1.6392, 1.3879, 1.6792, 1.7278, 1.6564, 1.5992, 1.782, 1.7706]}
{"unit_id": "orig:1969214391792124049", "ticker_idx": 1, "ts": "2025-09-20T01:37:51+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is also set to benefit from both DRAM price recovery and its entry into HBM, strengthening its cyclical rebound", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA adopts bullish memory upcycle thesis: SK Hynix as core beneficiary, Micron on HBM+DRAM recovery, Samsung undervalued; DDR5/HBM4/China demand driving the cycle.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "DDR5 Surge, Next-Gen Rubin Memory Architecture, and China’s Demand Recovery Driving the Memory Upcycle (BofA)\n\n1. Industry and Price Cycle\n    • DDR5 spot prices rose 10% this week, hitting record highs and shifting from a gradual recovery into a sharp rally, confirming the upside of the cycle.\n    • DDR4 also climbed 6%, showing that even legacy products are rebounding strongly, signaling tight supply-demand conditions across the board.\n    • Key drivers include PC/server replacement demand, production shifts from DDR5 to HBM4 and GDDR7, and distributors building inventories ahead of anticipated Q4 shortages.\n    • DRAM contract prices: DDR4 surged 150% QoQ in Q3, while DDR5 continued its gradual uptrend. For Q4, DDR5 contract prices are expected to rise by over 10%, expanding earnings momentum.\n\n2. Product and Technology Innovation\n    • NVIDIA’s Rubin platform introduces a hybrid memory architecture of HBM4, GDDR7, and SOCAMM, expanding total memory capacity by 2.5x compared to Blackwell.\n    • This design allocates tailored memory resources: HBM4 for AI training, GDDR7 for inference, and SOCAMM for CPU computing—supporting both qualitative and quantitative growth in memory demand.\n    • SOCAMM does not require TSV processing, lowering costs, and partially substituting DDR5 demand, which could prolong memory shortages.\n\n3. Positioning of Korean Companies\n    • Samsung Electronics has passed NVIDIA’s quality test for its 12-high HBM3e, securing the foundation for full-scale mass production. Although SK hynix’s dominance is unlikely to be challenged in the short term, Samsung is expected to ship about $1 billion worth in Q4, rapidly contributing to revenue.\n    • SK hynix benefits from both its HBM market leadership and surging DDR4/DDR5 prices, emerging as the core beneficiary of the upcycle.\n    • Micron is also set to benefit from both DRAM price recovery and its entry into HBM, strengthening its cyclical rebound.\n\n4. China Demand Recovery\n    • China’s semiconductor equipment imports rose 12% YoY in August, far above market expectations (around 0%), partly influenced by Korean memory makers’ fab expansions in Xi’an and Wuxi.\n    • Investments in equipment and infrastructure are being front-loaded ahead of new U.S. export controls in December, creating a short-term demand boost.\n    • Huawei’s next-gen GPU roadmap (Ascend series) remains at the HBM2 level but includes plans to adopt HBM4 by 2028, strengthening long-term demand momentum.\n\n5. Valuation and Stock Trends\n    • SK hynix and Micron are trading above 2x P/B, a level historically associated with cycle peaks, reflecting strong confidence in the uptrend.\n    • Samsung Electronics and Nanya remain relatively undervalued, leaving room for valuation re-rating.\n    • Overall, AI-driven investment and HBM4 expectations are powering memory stocks higher, with SK hynix and Micron maintaining YTD strength.\n\n6. Conclusion\n    • The combination of a DDR5/HBM-driven price rally, Rubin’s architectural innovation, and China’s unexpected demand recovery is powering a strong global memory upcycle.\n    • Korean memory makers enjoy a triple tailwind: technological leadership in HBM, cyclical price gains in DRAM/NAND, and Chinese investment demand.\n    • With both qualitative and quantitative improvements confirmed, maintaining a positive outlook on the memory sector remains warranted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011481025777169696, "ret_signed_m1d": -0.011481025777169696, "alpha_spy_m1d": -0.006772294852597338, "alpha_c_m1d": 0.008570531977778018, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01087348305772684, "ret_signed_p1d": 0.01087348305772684, "alpha_spy_p1d": 0.0163169578975344, "alpha_c_p1d": 0.012332322371607751, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.004373843452624926, "ret_signed_p1w": -0.004373843452624926, "alpha_spy_p1w": 0.00036489580524323717, "alpha_c_p1w": -0.005894730709706697, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.22959944416106892, "ret_signed_p1m": 0.22959944416106892, "alpha_spy_p1m": 0.22292627200685078, "alpha_c_p1m": 0.15780512786668988, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.5107863663515086, "ret_signed_p3m": 0.5107863663515086, "alpha_spy_p3m": 0.49634517812828816, "alpha_c_p3m": 0.43293928745085, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.8074702227486457, "ret_signed_p6m": 1.8074702227486457, "alpha_spy_p6m": 1.7985744088282953, "alpha_c_p6m": 1.5717408696055801, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2277, -0.2353, -0.2428, -0.252, -0.2663, -0.2612, -0.2578, -0.2578, -0.2715, -0.2442, -0.2574, -0.2522, -0.2435, -0.2795, -0.2704, -0.2927, -0.312, -0.3051, -0.3122, -0.3365, -0.3328, -0.3213, -0.3241, -0.3242, -0.3199, -0.303, -0.337, -0.3629, -0.3453, -0.3375, -0.3392, -0.3204, -0.2778, -0.2485, -0.224, -0.2451, -0.2389, -0.2658, -0.2495, -0.2586, -0.288, -0.2966, -0.2851, -0.2928, -0.2923, -0.2847, -0.2589, -0.2771, -0.2771, -0.2803, -0.2788, -0.2455, -0.202, -0.2014, -0.1785, -0.1496, -0.0853, -0.0449, -0.0416, -0.0352, -0.0281, 0.0259, -0.0115, 0.0, 0.0109, -0.0177, -0.0473, -0.0446, -0.0044, 0.0164, 0.1065, 0.1162, 0.1417, 0.1607, 0.1287, 0.1946, 0.1691, 0.1038, 0.1717, 0.137, 0.1667, 0.2311, 0.2301, 0.2568, 0.2296, 0.2064, 0.2565, 0.3313, 0.3379, 0.3489, 0.3775, 0.3616, 0.3602, 0.4266, 0.3253, 0.4436, 0.4487, 0.4462, 0.5397, 0.4656, 0.4886, 0.4403, 0.5003, 0.4707, 0.3889, 0.3732, 0.224, 0.2605, 0.3611, 0.3648, 0.3996, 0.3996, 0.4374, 0.4616, 0.4557, 0.4233, 0.3777, 0.4419, 0.5009, 0.5343, 0.6029, 0.571, 0.4657, 0.4436, 0.4133, 0.3708, 0.5108, 0.6164, 0.6812, 0.6793, 0.7426, 0.7426, 0.7311, 0.79, 0.7794, 0.7355, 0.7355, 0.918, 0.8981, 1.0883, 1.0648, 0.9886, 1.0984, 1.1032, 1.0561, 1.0271, 1.047, 1.2058, 1.2058, 1.2195, 1.3661, 1.4176, 1.4302, 1.366, 1.4946, 1.6469, 1.65, 1.5228, 1.6622, 1.5506, 1.3071, 1.3283, 1.4001, 1.332, 1.2697, 1.4952, 1.5173, 1.5032, 1.5032, 1.431, 1.5597, 1.5378, 1.6036, 1.5599, 1.5419, 1.6087, 1.527, 1.5076, 1.5094, 1.3088, 1.437, 1.4144, 1.2517, 1.3674, 1.4513, 1.546, 1.4649, 1.5912, 1.6865, 1.8075]}
{"unit_id": "orig:1969214391792124049", "ticker_idx": 2, "ts": "2025-09-20T01:37:51+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and Nanya remain relatively undervalued, leaving room for valuation re-rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA adopts bullish memory upcycle thesis: SK Hynix as core beneficiary, Micron on HBM+DRAM recovery, Samsung undervalued; DDR5/HBM4/China demand driving the cycle.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "DDR5 Surge, Next-Gen Rubin Memory Architecture, and China’s Demand Recovery Driving the Memory Upcycle (BofA)\n\n1. Industry and Price Cycle\n    • DDR5 spot prices rose 10% this week, hitting record highs and shifting from a gradual recovery into a sharp rally, confirming the upside of the cycle.\n    • DDR4 also climbed 6%, showing that even legacy products are rebounding strongly, signaling tight supply-demand conditions across the board.\n    • Key drivers include PC/server replacement demand, production shifts from DDR5 to HBM4 and GDDR7, and distributors building inventories ahead of anticipated Q4 shortages.\n    • DRAM contract prices: DDR4 surged 150% QoQ in Q3, while DDR5 continued its gradual uptrend. For Q4, DDR5 contract prices are expected to rise by over 10%, expanding earnings momentum.\n\n2. Product and Technology Innovation\n    • NVIDIA’s Rubin platform introduces a hybrid memory architecture of HBM4, GDDR7, and SOCAMM, expanding total memory capacity by 2.5x compared to Blackwell.\n    • This design allocates tailored memory resources: HBM4 for AI training, GDDR7 for inference, and SOCAMM for CPU computing—supporting both qualitative and quantitative growth in memory demand.\n    • SOCAMM does not require TSV processing, lowering costs, and partially substituting DDR5 demand, which could prolong memory shortages.\n\n3. Positioning of Korean Companies\n    • Samsung Electronics has passed NVIDIA’s quality test for its 12-high HBM3e, securing the foundation for full-scale mass production. Although SK hynix’s dominance is unlikely to be challenged in the short term, Samsung is expected to ship about $1 billion worth in Q4, rapidly contributing to revenue.\n    • SK hynix benefits from both its HBM market leadership and surging DDR4/DDR5 prices, emerging as the core beneficiary of the upcycle.\n    • Micron is also set to benefit from both DRAM price recovery and its entry into HBM, strengthening its cyclical rebound.\n\n4. China Demand Recovery\n    • China’s semiconductor equipment imports rose 12% YoY in August, far above market expectations (around 0%), partly influenced by Korean memory makers’ fab expansions in Xi’an and Wuxi.\n    • Investments in equipment and infrastructure are being front-loaded ahead of new U.S. export controls in December, creating a short-term demand boost.\n    • Huawei’s next-gen GPU roadmap (Ascend series) remains at the HBM2 level but includes plans to adopt HBM4 by 2028, strengthening long-term demand momentum.\n\n5. Valuation and Stock Trends\n    • SK hynix and Micron are trading above 2x P/B, a level historically associated with cycle peaks, reflecting strong confidence in the uptrend.\n    • Samsung Electronics and Nanya remain relatively undervalued, leaving room for valuation re-rating.\n    • Overall, AI-driven investment and HBM4 expectations are powering memory stocks higher, with SK hynix and Micron maintaining YTD strength.\n\n6. Conclusion\n    • The combination of a DDR5/HBM-driven price rally, Rubin’s architectural innovation, and China’s unexpected demand recovery is powering a strong global memory upcycle.\n    • Korean memory makers enjoy a triple tailwind: technological leadership in HBM, cyclical price gains in DRAM/NAND, and Chinese investment demand.\n    • With both qualitative and quantitative improvements confirmed, maintaining a positive outlook on the memory sector remains warranted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "orig:1969214391792124049", "ticker_idx": 3, "ts": "2025-09-20T01:37:51+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "maintaining a positive outlook on the memory sector remains warranted", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA adopts bullish memory upcycle thesis: SK Hynix as core beneficiary, Micron on HBM+DRAM recovery, Samsung undervalued; DDR5/HBM4/China demand driving the cycle.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "DDR5 Surge, Next-Gen Rubin Memory Architecture, and China’s Demand Recovery Driving the Memory Upcycle (BofA)\n\n1. Industry and Price Cycle\n    • DDR5 spot prices rose 10% this week, hitting record highs and shifting from a gradual recovery into a sharp rally, confirming the upside of the cycle.\n    • DDR4 also climbed 6%, showing that even legacy products are rebounding strongly, signaling tight supply-demand conditions across the board.\n    • Key drivers include PC/server replacement demand, production shifts from DDR5 to HBM4 and GDDR7, and distributors building inventories ahead of anticipated Q4 shortages.\n    • DRAM contract prices: DDR4 surged 150% QoQ in Q3, while DDR5 continued its gradual uptrend. For Q4, DDR5 contract prices are expected to rise by over 10%, expanding earnings momentum.\n\n2. Product and Technology Innovation\n    • NVIDIA’s Rubin platform introduces a hybrid memory architecture of HBM4, GDDR7, and SOCAMM, expanding total memory capacity by 2.5x compared to Blackwell.\n    • This design allocates tailored memory resources: HBM4 for AI training, GDDR7 for inference, and SOCAMM for CPU computing—supporting both qualitative and quantitative growth in memory demand.\n    • SOCAMM does not require TSV processing, lowering costs, and partially substituting DDR5 demand, which could prolong memory shortages.\n\n3. Positioning of Korean Companies\n    • Samsung Electronics has passed NVIDIA’s quality test for its 12-high HBM3e, securing the foundation for full-scale mass production. Although SK hynix’s dominance is unlikely to be challenged in the short term, Samsung is expected to ship about $1 billion worth in Q4, rapidly contributing to revenue.\n    • SK hynix benefits from both its HBM market leadership and surging DDR4/DDR5 prices, emerging as the core beneficiary of the upcycle.\n    • Micron is also set to benefit from both DRAM price recovery and its entry into HBM, strengthening its cyclical rebound.\n\n4. China Demand Recovery\n    • China’s semiconductor equipment imports rose 12% YoY in August, far above market expectations (around 0%), partly influenced by Korean memory makers’ fab expansions in Xi’an and Wuxi.\n    • Investments in equipment and infrastructure are being front-loaded ahead of new U.S. export controls in December, creating a short-term demand boost.\n    • Huawei’s next-gen GPU roadmap (Ascend series) remains at the HBM2 level but includes plans to adopt HBM4 by 2028, strengthening long-term demand momentum.\n\n5. Valuation and Stock Trends\n    • SK hynix and Micron are trading above 2x P/B, a level historically associated with cycle peaks, reflecting strong confidence in the uptrend.\n    • Samsung Electronics and Nanya remain relatively undervalued, leaving room for valuation re-rating.\n    • Overall, AI-driven investment and HBM4 expectations are powering memory stocks higher, with SK hynix and Micron maintaining YTD strength.\n\n6. Conclusion\n    • The combination of a DDR5/HBM-driven price rally, Rubin’s architectural innovation, and China’s unexpected demand recovery is powering a strong global memory upcycle.\n    • Korean memory makers enjoy a triple tailwind: technological leadership in HBM, cyclical price gains in DRAM/NAND, and Chinese investment demand.\n    • With both qualitative and quantitative improvements confirmed, maintaining a positive outlook on the memory sector remains warranted.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012327960751177628, "ret_signed_m1d": -0.012327960751177628, "alpha_spy_m1d": -0.00761922982660527, "alpha_c_m1d": 0.007723597003770085, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01791158103942525, "ret_signed_p1d": 0.01791158103942525, "alpha_spy_p1d": 0.02335505587923281, "alpha_c_p1d": 0.01937042035330616, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.0009367612410294823, "ret_signed_p1w": 0.0009367612410294823, "alpha_spy_p1w": 0.005675500498897645, "alpha_c_p1w": -0.0005841260160522888, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.25571535433485715, "ret_signed_p1m": 0.25571535433485715, "alpha_spy_p1m": 0.249042182180639, "alpha_c_p1m": 0.1839210380404781, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.4596447512769175, "ret_signed_p3m": 0.4596447512769175, "alpha_spy_p3m": 0.445203563053697, "alpha_c_p3m": 0.38179767237625883, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.6406466633660866, "ret_signed_p6m": 1.6406466633660866, "alpha_spy_p6m": 1.6317508494457362, "alpha_c_p6m": 1.404917310223021, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2281, -0.2284, -0.2356, -0.235, -0.2444, -0.2464, -0.2338, -0.2434, -0.2539, -0.2355, -0.2449, -0.2256, -0.2187, -0.2258, -0.2194, -0.2253, -0.2488, -0.2454, -0.2417, -0.2608, -0.2574, -0.2547, -0.2594, -0.2452, -0.2425, -0.228, -0.2346, -0.2679, -0.2589, -0.2502, -0.2599, -0.2436, -0.2294, -0.2129, -0.2024, -0.1977, -0.1983, -0.2072, -0.2167, -0.224, -0.2389, -0.2514, -0.2387, -0.2328, -0.2355, -0.2332, -0.2201, -0.2253, -0.246, -0.2369, -0.2317, -0.2165, -0.1968, -0.1909, -0.1672, -0.138, -0.1106, -0.0687, -0.0608, -0.031, -0.0472, 0.0001, -0.0123, 0.0, 0.0179, 0.0079, -0.0002, -0.0295, 0.0009, 0.0052, 0.0556, 0.1075, 0.116, 0.1224, 0.1117, 0.1337, 0.1252, 0.1529, 0.1588, 0.1371, 0.1711, 0.2318, 0.2447, 0.2734, 0.2557, 0.2548, 0.2602, 0.3243, 0.363, 0.3434, 0.3921, 0.4107, 0.4153, 0.5098, 0.4189, 0.4345, 0.4438, 0.4254, 0.4921, 0.4914, 0.4956, 0.4735, 0.4217, 0.4691, 0.3964, 0.3784, 0.3537, 0.2951, 0.3353, 0.346, 0.3764, 0.3985, 0.3858, 0.4027, 0.4301, 0.4181, 0.3957, 0.4323, 0.4877, 0.484, 0.5252, 0.4909, 0.4679, 0.4279, 0.387, 0.4132, 0.4596, 0.4849, 0.5547, 0.5619, 0.5851, 0.5851, 0.6154, 0.6864, 0.6949, 0.6803, 0.6803, 0.8005, 0.8506, 0.9457, 0.9615, 0.9406, 0.9666, 0.9721, 0.9412, 0.9462, 0.974, 1.0537, 1.0629, 1.031, 1.0943, 1.1371, 1.1518, 1.101, 1.2345, 1.3359, 1.3491, 1.3515, 1.2822, 1.3871, 1.3057, 1.2182, 1.2364, 1.2908, 1.2571, 1.3252, 1.4026, 1.4008, 1.4008, 1.3768, 1.4197, 1.4611, 1.5357, 1.5347, 1.6082, 1.657, 1.7671, 1.7184, 1.719, 1.4497, 1.3145, 1.4727, 1.3886, 1.2842, 1.4673, 1.5235, 1.4642, 1.4695, 1.5832, 1.6406]}
{"unit_id": "orig:1970749965430665574", "ticker_idx": 0, "ts": "2025-09-24T07:19:41+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "generic DRAM ASPs are expected to rise 8–13% QoQ, and including HBM, the increase could expand to 13–18%", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM ASPs forecast +8–13% QoQ in 4Q25 (up to +18% incl. HBM) on server demand surge and capacity shift toward advanced/HBM nodes.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "▶ DRAM Prices to Continue Rising in 4Q25, Server Demand Surges Ahead While Legacy Process Products See Steeper Increases\n\n    • The three major memory makers are prioritizing advanced process capacity for high-end server DRAM and HBM, which naturally reduces capacity allocation for PC, mobile, and consumer DRAM product groups.\n    • n 4Q25, generic DRAM ASPs are expected to rise 8–13% QoQ, and including HBM, the increase could expand to 13–18%.\n\n⸻\n\n[PC DRAM: Modest Price Increase, U.S. CSPs Considering Early Procurement of Server DRAM]\n    • PC shipments in 4Q25 are expected to decline due to weaker momentum from promotions and slower inventory restocking. Accordingly, major OEMs are reducing their DRAM bit procurement.\n    • Meanwhile, as memory makers aggressively shift capacity toward server DDR5 products, supply constraints for PC DDR5/DDR4 persist.\n    • As a result, PC DRAM ASPs are expected to show a mild but steady upward trend throughout the quarter.\n    • Global CSPs are projected to significantly increase memory procurement in 2026, and some U.S. CSPs have already shown moves to start securing supply from 4Q25 to ensure sufficient inventory.\n    • Although suppliers are actively adjusting capacity, technical issues at some vendors and the prioritization of HBM4 capacity allocations create continued uncertainty in DDR5 supply.\n\n⸻\n\n[Mobile DRAM Prices Rising, Graphics DRAM Demand Sustained, Consumer DRAM Growth Slows]\n\n    • Suppliers have reduced shipments of LPDDR4X used in mid- to low-end smartphones. Brands, seeking to avoid supply disruptions, increased procurement, which has worsened the imbalance. Consequently, LPDDR4X prices are expected to rise +10% QoQ in 4Q25.\n    • LPDDR5X products are expanding beyond premium smartphones into other categories, and given supply dynamics and pricing strategies, LPDDR5X prices are also likely to continue rising.\n    • The PC inventory restocking cycle and expectations around NVIDIA’s RTX 6000 series launch are supporting strong demand inflows for graphics DRAM, especially GDDR7.\n    • Suppliers are anticipating shortages and have begun raising prices, with GDDR7 prices expected to increase more steeply than in the prior quarter.\n   • Suppliers are maintaining firm DDR4 pricing in consumer DRAM, but supply conditions remain limited. Following the sharp spike in 3Q25, buyers have scaled back inventory procurement.\n    • Accordingly, DDR4 price increases are expected to moderate in 4Q25. Meanwhile, DDR3 still benefits from early procurement demand, and with capacity conversions and rapid inventory depletion, its price increases are expected to continue.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010219261695896894, "ret_signed_m1d": 0.010219261695896894, "alpha_spy_m1d": 0.007027473142046511, "alpha_c_m1d": 0.008381803965575604, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008259317884715492, "ret_signed_p1d": -0.008259317884715492, "alpha_spy_p1d": -0.0036458203962434297, "alpha_c_p1d": -0.008103638291812413, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.04830357054726327, "ret_signed_p1w": 0.04830357054726327, "alpha_spy_p1w": 0.03718560612396956, "alpha_c_p1w": 0.009126939226759966, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.25085205846157504, "ret_signed_p1m": 0.25085205846157504, "alpha_spy_p1m": 0.2347273796390122, "alpha_c_p1m": 0.1770767101201443, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 0.5449006934175141, "ret_signed_p3m": 0.5449006934175141, "alpha_spy_p3m": 0.5059449215457171, "alpha_c_p3m": 0.42123818656422773, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 1.6535661009895588, "ret_signed_p6m": 1.6535661009895588, "alpha_spy_p6m": 1.65258338368522, "alpha_c_p6m": 1.4198802593573656, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.2413, -0.2409, -0.2503, -0.2522, -0.2398, -0.2492, -0.2597, -0.2413, -0.2506, -0.2316, -0.2248, -0.2322, -0.2259, -0.2319, -0.2553, -0.2518, -0.2482, -0.2673, -0.264, -0.2612, -0.2658, -0.252, -0.2492, -0.2348, -0.2417, -0.2747, -0.2656, -0.257, -0.2666, -0.2503, -0.2359, -0.2193, -0.2087, -0.2044, -0.2048, -0.2139, -0.223, -0.2303, -0.2453, -0.2576, -0.245, -0.2393, -0.242, -0.2396, -0.2265, -0.2318, -0.2521, -0.2432, -0.2381, -0.2228, -0.2029, -0.1971, -0.1736, -0.1446, -0.1169, -0.0753, -0.0676, -0.0382, -0.054, -0.007, -0.0197, -0.0075, 0.0102, 0.0, -0.0083, -0.0369, -0.0066, -0.0022, 0.0483, 0.0994, 0.1081, 0.1145, 0.1036, 0.126, 0.1173, 0.1437, 0.1503, 0.1286, 0.1623, 0.2227, 0.2353, 0.2638, 0.2461, 0.2449, 0.2509, 0.3146, 0.3527, 0.3336, 0.3817, 0.3998, 0.4042, 0.4978, 0.4074, 0.4241, 0.4334, 0.4154, 0.4819, 0.4803, 0.4846, 0.4624, 0.4123, 0.4585, 0.3862, 0.3683, 0.3422, 0.2852, 0.3258, 0.3363, 0.3666, 0.3883, 0.3763, 0.3932, 0.42, 0.4077, 0.3853, 0.4219, 0.477, 0.4738, 0.5151, 0.481, 0.4571, 0.4177, 0.3771, 0.4023, 0.4496, 0.4756, 0.5449, 0.5519, 0.5755, 0.5755, 0.605, 0.6754, 0.6836, 0.6687, 0.6687, 0.7888, 0.8376, 0.9333, 0.9485, 0.927, 0.9539, 0.9594, 0.9284, 0.9329, 0.9604, 1.0405, 1.0496, 1.0185, 1.0824, 1.1249, 1.1395, 1.0888, 1.2213, 1.3227, 1.3357, 1.3366, 1.2703, 1.3717, 1.2888, 1.2033, 1.2221, 1.2745, 1.2407, 1.3102, 1.3864, 1.3844, 1.3844, 1.3599, 1.4036, 1.4439, 1.5179, 1.5163, 1.5882, 1.6368, 1.7437, 1.6956, 1.6963, 1.4298, 1.299, 1.4538, 1.3693, 1.2684, 1.4489, 1.5052, 1.446, 1.4529, 1.5655, 1.6233, 1.7613, 1.6536]}
{"unit_id": "orig:1949760813067383114", "ticker_idx": 0, "ts": "2025-07-28T09:16:17+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CLSA assessed that this contract will strengthen the justification for investment in the Taylor plant, forecasting an operating profit turnaround from 2H26 and the possibility of valuation multiple recovery", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares CLSA bullish view: Samsung foundry turns profitable from 2H26, valuation multiple recovery likely on Tesla 22.8T KRW order.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Hong Kong CLSA Expects Samsung Foundry to Turn Profitable from 2H 2026\n\nSamsung Electronics announced a foundry order worth approximately 22.8 trillion KRW. Although the contract party remains undisclosed, Bloomberg reported it to be Tesla. The contract is said to involve the production of 2-nanometer process-based chips for Tesla's autonomous driving technology and humanoid robots at Samsung's Taylor, Texas plant. This is interpreted as a sign of the market's regained confidence in Samsung's advanced foundry capabilities, and it also suggests the potential for expanding major client relationships.\n\nThis order value accounts for approximately 7.6% of Samsung Electronics' total revenue in 2024, and about 86% of the Foundry and LSI business unit's revenue. If recognized uniformly over the project period, it is estimated to be approximately 2.7 trillion KRW annually. Samsung Electronics' external foundry customers include Qualcomm, Google, Tesla, NVIDIA, and Nintendo. As of 2024, this division recorded an operating loss of approximately 5.6 trillion KRW, and a loss of 6.6 trillion KRW was projected for 2025. However, this order is expected to reduce losses and bring about a structural turnaround starting from the second half of the year.\n\nSamsung Electronics' foundry business has been losing market share to TSMC over the past four years, with its share dropping to 7.7% as of Q1 2025, a significant decrease from 19% five years ago. Customer attrition also continued due to yield degradation and delays in process transitions. However, this order appears to be the first visible sign of the results from Samsung's restructuring and technology investments, providing momentum that could lead to new orders from other clients in the future.\n\nCLSA assessed that this contract will strengthen the justification for investment in the Taylor plant, forecasting an operating profit turnaround from 2H26 and the possibility of valuation multiple recovery. Accordingly, doubts about the sustainability of profitability are expected to ease, and the market is highly likely to positively reflect the possibility of a turnaround in the foundry business.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06392037833579933, "ret_signed_m1d": -0.06392037833579933, "alpha_spy_m1d": -0.06417154154811011, "alpha_c_m1d": -0.056005973190194513, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.00791440514560482, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0028408408529752016, "ret_signed_p1d": 0.0028408408529752016, "alpha_spy_p1d": 0.005478586093461302, "alpha_c_p1d": 0.0028030150784925656, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 3.782577448263602e-05, "ret_p1w": -0.009943167469083525, "ret_signed_p1w": -0.009943167469083525, "alpha_spy_p1w": -0.0008840901899609399, "alpha_c_p1w": -0.0031650439477406422, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0067781235213428825, "ret_p1m": -0.0014204204264876008, "ret_signed_p1m": -0.0014204204264876008, "alpha_spy_p1m": -0.014325802828035084, "alpha_c_p1m": 0.002555579003573172, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": -0.003975999430060773, "ret_p3m": 0.37685443154336795, "ret_signed_p3m": 0.37685443154336795, "alpha_spy_p3m": 0.31926287124578745, "alpha_c_p3m": 0.28074134315242305, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.0961130883909449, "ret_p6m": 1.0817728912172333, "ret_signed_p6m": 1.0817728912172333, "alpha_spy_p6m": 1.0118664091672853, "alpha_c_p6m": 1.0045408978522918, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.07723199336494146, "price_path": [-0.2165, -0.2165, -0.2334, -0.2334, -0.2334, -0.2292, -0.2292, -0.2263, -0.1868, -0.1967, -0.1896, -0.191, -0.1981, -0.2122, -0.2108, -0.2136, -0.2277, -0.2348, -0.2277, -0.239, -0.2108, -0.208, -0.2066, -0.1981, -0.1981, -0.184, -0.1656, -0.1656, -0.1557, -0.1642, -0.1543, -0.16, -0.1769, -0.1925, -0.1797, -0.1557, -0.1642, -0.16, -0.1812, -0.1459, -0.1346, -0.1501, -0.1364, -0.1506, -0.1449, -0.1364, -0.0937, -0.1009, -0.1236, -0.1278, -0.142, -0.1335, -0.1108, -0.1122, -0.0952, -0.081, -0.0526, -0.0469, -0.0369, -0.0625, -0.0568, -0.0625, -0.0639, 0.0, 0.0028, 0.0313, 0.0142, -0.0213, -0.0099, -0.0071, -0.0227, 0.0014, 0.0199, 0.0085, 0.0099, 0.0213, 0.017, 0.017, -0.0057, -0.0057, 0.0014, 0.0028, 0.0142, 0.0156, -0.0014, 0.0028, -0.0114, -0.0099, -0.0398, -0.0185, -0.0085, -0.0043, -0.0128, -0.0043, 0.0156, 0.0313, 0.0426, 0.071, 0.0866, 0.1278, 0.1108, 0.1406, 0.1406, 0.1861, 0.2031, 0.2131, 0.223, 0.1832, 0.2014, 0.1971, 0.227, 0.2805, 0.2805, 0.2805, 0.2805, 0.2805, 0.2805, 0.3469, 0.3312, 0.3069, 0.3555, 0.394, 0.3968, 0.3997, 0.3911, 0.4068, 0.3769, 0.4097, 0.4553, 0.4197, 0.4339, 0.4853, 0.5338, 0.5852, 0.4967, 0.4354, 0.4154, 0.3968, 0.4354, 0.4767, 0.471, 0.4667, 0.3868, 0.4354, 0.3954, 0.3769, 0.4354, 0.3526, 0.3797, 0.4168, 0.4667, 0.4767, 0.4339, 0.4382, 0.4753, 0.491, 0.4996, 0.5466, 0.5623, 0.5466, 0.5409, 0.5309, 0.5538, 0.4953, 0.4667, 0.5395, 0.5352, 0.5167, 0.5766, 0.5909, 0.5852, 0.5852, 0.6693, 0.7133, 0.719, 0.719, 0.719, 0.8423, 0.98, 0.9914, 1.0216, 0.99, 0.9929, 0.99, 0.9728, 1.0115, 1.0631, 1.1348, 1.1406, 1.0818]}
{"unit_id": "orig:1945993027073528163", "ticker_idx": 0, "ts": "2025-07-17T23:44:27+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs Raises TSMC Price Target from NT1,210 to NT1,370", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Goldman Sachs Buy on TSMC with PT raised to NT1,370; on APAC Conviction List on AI/HPC/EV structural growth.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs Raises TSMC Price Target from NT1,210 to NT1,370\n\nAPAC Conviction List\n\n- Applies a 20x P/E to 2026 EPS.\n\n- Maintains technology leadership with over 60% global foundry market share.\n\n- Expected to benefit from long-term structural growth in AI, HPC, 5G, and EV sectors.\n\n- Potential to achieve around 20% revenue CAGR in the future.\n\n- Valuation remains attractive, positioned at the lower end of its 10-year band.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032736137765441486, "ret_signed_m1d": -0.032736137765441486, "alpha_spy_m1d": -0.02665367060126722, "alpha_c_m1d": -0.024405971921902414, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021172647928687227, "ret_signed_p1d": -0.021172647928687227, "alpha_spy_p1d": -0.02044023502657677, "alpha_c_p1d": -0.016373454888251704, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": -0.016286599225248666, "ret_signed_p1w": -0.016286599225248666, "alpha_spy_p1w": -0.02644521420138235, "alpha_c_p1w": -0.002882922782284747, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": -0.02736158446200665, "ret_signed_p1m": -0.02736158446200665, "alpha_spy_p1m": -0.051882372472820015, "alpha_c_p1m": -0.040868208695853725, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.20876894271373292, "ret_signed_p3m": 0.20876894271373292, "alpha_spy_p3m": 0.15140639950328283, "alpha_c_p3m": 0.06551011307794985, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 0.3254373776897599, "ret_signed_p6m": 0.3254373776897599, "alpha_spy_p6m": 0.2139622326356443, "alpha_c_p6m": -0.012984752748134376, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.4002, -0.3858, -0.3598, -0.334, -0.3302, -0.3371, -0.3327, -0.3238, -0.2993, -0.2727, -0.2844, -0.3011, -0.2919, -0.2892, -0.2839, -0.2415, -0.213, -0.2099, -0.2121, -0.2121, -0.215, -0.2152, -0.2221, -0.2041, -0.2212, -0.2212, -0.198, -0.2043, -0.2002, -0.2157, -0.2096, -0.1983, -0.1789, -0.1751, -0.1676, -0.1602, -0.1381, -0.1314, -0.1228, -0.1405, -0.1218, -0.1291, -0.1307, -0.1307, -0.1469, -0.1436, -0.1039, -0.0931, -0.0879, -0.0693, -0.0778, -0.0852, -0.0489, -0.044, -0.044, -0.0669, -0.0722, -0.056, -0.0645, -0.0619, -0.0689, -0.0352, -0.0327, 0.0, -0.0212, -0.0275, -0.0448, -0.0215, -0.0163, 0.0, -0.0116, -0.0174, -0.011, -0.0162, -0.0423, -0.0269, -0.0535, -0.0579, -0.0121, -0.0154, -0.0143, -0.0053, -0.0169, -0.0187, -0.0274, -0.0171, -0.0525, -0.0692, -0.0744, -0.0513, -0.0408, -0.028, -0.0257, -0.0298, -0.06, -0.06, -0.0701, -0.0579, -0.0423, -0.0089, 0.0065, 0.0217, 0.0604, 0.0542, 0.0559, 0.0643, 0.0704, 0.0734, 0.0973, 0.0819, 0.1136, 0.1547, 0.1466, 0.13, 0.1165, 0.116, 0.1408, 0.1783, 0.1768, 0.1935, 0.2352, 0.201, 0.2438, 0.2249, 0.1464, 0.2372, 0.2088, 0.2446, 0.2247, 0.2053, 0.216, 0.2029, 0.1799, 0.1875, 0.2048, 0.2182, 0.2316, 0.2461, 0.2385, 0.2271, 0.2452, 0.201, 0.1994, 0.1814, 0.1702, 0.206, 0.1893, 0.187, 0.1526, 0.1633, 0.1519, 0.1351, 0.1533, 0.1335, 0.1235, 0.1626, 0.1628, 0.1843, 0.1843, 0.1907, 0.175, 0.193, 0.2068, 0.1965, 0.2038, 0.233, 0.2393, 0.2668, 0.2485, 0.1961, 0.1784, 0.1749, 0.1343, 0.1659, 0.1834, 0.2011, 0.2162, 0.2237, 0.2237, 0.2403, 0.2324, 0.2269, 0.2446, 0.2446, 0.309, 0.3198, 0.341, 0.3052, 0.3024, 0.3254]}
{"unit_id": "orig:1973358174741664061", "ticker_idx": 0, "ts": "2025-10-01T12:03:46+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "DRAM operating margin is projected to climb from the current 40–50% range to nearly 70% in 2026, approaching the peak levels seen during the 2017 supercycle", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Nomura's Memory Supercycle thesis: DRAM margins to ~70%, NAND breakout, HBM ASP +15%/yr; SK Hynix key beneficiary.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura Securities: On the Memory Supercycle\n\nAI and Data Center Demand Explosion\n    • U.S. Big Tech companies are expected to continue expanding their AI server investments, with traditional server capital expenditures projected to grow by about 20–30% in 2026.\n    • Accordingly, DDR4/DDR5 memory demand is expected to increase by around 50%, and enterprise SSD demand is projected to nearly double.\n    • Due to HDD supply shortages and surging AI storage demand, the NAND market is also set to boom, with bit shipments in 2026 forecast to rise more than 50% YoY.\n\n⸻\n\nPrice and Profitability Surge\n    • DRAM operating margin is projected to climb from the current 40–50% range to nearly 70% in 2026, approaching the peak levels seen during the 2017 supercycle.\n    • NAND margins are expected to break out from the current breakeven point to 30–40%.\n    • HBM competition is likely to intensify (with Samsung expected to rejoin NVIDIA’s supply chain in 2026). However, blended ASP is still forecast to maintain 15%+ annual growth through 2025–2026, contributing to around 75% of SK Hynix’s operating profit.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04685237086366715, "ret_signed_m1d": -0.04685237086366715, "alpha_spy_m1d": -0.04345637353440394, "alpha_c_m1d": -0.024885818333107802, "bench_spy_m1d": -0.0033959973292632117, "bench_c_m1d": -0.021966552530559347, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.050333279571154264, "ret_signed_p1d": 0.050333279571154264, "alpha_spy_p1d": 0.04918156796811431, "alpha_c_p1d": 0.036697861685117315, "bench_spy_p1d": 0.001151711603039951, "bench_c_p1d": 0.013635417886036949, "ret_p1w": 0.07396412336687101, "ret_signed_p1w": 0.07396412336687101, "alpha_spy_p1w": 0.0669928797010783, "alpha_c_p1w": 0.03686377251851962, "bench_spy_p1w": 0.006971243665792715, "bench_c_p1w": 0.03710035084835139, "ret_p1m": 0.33960787045402246, "ret_signed_p1m": 0.33960787045402246, "alpha_spy_p1m": 0.32258353069781404, "alpha_c_p1m": 0.24958413195402113, "bench_spy_p1m": 0.01702433975620843, "bench_c_p1m": 0.09002373850000134, "ret_p3m": 0.5976972811035476, "ret_signed_p3m": 0.5976972811035476, "alpha_spy_p3m": 0.5656343904139448, "alpha_c_p3m": 0.5028389653444348, "bench_spy_p3m": 0.032062890689602774, "bench_c_p3m": 0.09485831575911274, "ret_p6m": 1.2187180735779972, "ret_signed_p6m": 1.2187180735779972, "alpha_spy_p6m": 1.248169640487051, "alpha_c_p6m": 1.0738679960214665, "bench_spy_p6m": -0.029451566909053817, "bench_c_p6m": 0.14485007755653068, "price_path": [-0.2821, -0.2919, -0.2747, -0.2834, -0.2647, -0.2583, -0.2644, -0.2584, -0.2636, -0.2862, -0.283, -0.2792, -0.2972, -0.2941, -0.2917, -0.2962, -0.2825, -0.28, -0.2663, -0.2719, -0.3037, -0.2953, -0.2871, -0.2964, -0.281, -0.2682, -0.2528, -0.2431, -0.2381, -0.2388, -0.2469, -0.2565, -0.2634, -0.2772, -0.2892, -0.2771, -0.2711, -0.2738, -0.2717, -0.2596, -0.2642, -0.2843, -0.2753, -0.2703, -0.2564, -0.2382, -0.2324, -0.2099, -0.1822, -0.157, -0.1172, -0.1096, -0.0808, -0.0967, -0.0519, -0.0632, -0.0515, -0.0344, -0.0435, -0.0507, -0.0792, -0.0506, -0.0469, 0.0, 0.0503, 0.058, 0.0637, 0.0541, 0.074, 0.0663, 0.0947, 0.0989, 0.0786, 0.1109, 0.1685, 0.1811, 0.2084, 0.1918, 0.1914, 0.1956, 0.2562, 0.2938, 0.2744, 0.3212, 0.3396, 0.344, 0.4345, 0.3484, 0.3609, 0.37, 0.3522, 0.4149, 0.4158, 0.4194, 0.399, 0.3472, 0.3941, 0.3253, 0.3081, 0.2874, 0.2296, 0.2663, 0.2766, 0.3053, 0.3269, 0.3137, 0.3295, 0.3564, 0.3453, 0.3246, 0.3586, 0.4112, 0.4068, 0.4454, 0.4127, 0.3927, 0.3544, 0.3153, 0.3417, 0.3837, 0.4058, 0.4722, 0.4792, 0.5004, 0.5004, 0.53, 0.5977, 0.6063, 0.593, 0.593, 0.7056, 0.7546, 0.8428, 0.8587, 0.8402, 0.8629, 0.8682, 0.8392, 0.8446, 0.8711, 0.945, 0.9539, 0.9226, 0.9808, 1.0211, 1.0352, 0.9871, 1.114, 1.2092, 1.222, 1.2274, 1.1568, 1.2612, 1.1876, 1.102, 1.1181, 1.1725, 1.1412, 1.2021, 1.2768, 1.2752, 1.2752, 1.2535, 1.2923, 1.3329, 1.4041, 1.4041, 1.4759, 1.5217, 1.6305, 1.5836, 1.5841, 1.3271, 1.193, 1.3471, 1.2691, 1.165, 1.341, 1.3935, 1.3377, 1.3401, 1.4484, 1.5014, 1.6382, 1.5351, 1.4861, 1.3324, 1.3787, 1.3597, 1.2187]}
{"unit_id": "orig:1973358174741664061", "ticker_idx": 1, "ts": "2025-10-01T12:03:46+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "blended ASP is still forecast to maintain 15%+ annual growth through 2025–2026, contributing to around 75% of SK Hynix's operating profit", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Nomura's Memory Supercycle thesis: DRAM margins to ~70%, NAND breakout, HBM ASP +15%/yr; SK Hynix key beneficiary.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura Securities: On the Memory Supercycle\n\nAI and Data Center Demand Explosion\n    • U.S. Big Tech companies are expected to continue expanding their AI server investments, with traditional server capital expenditures projected to grow by about 20–30% in 2026.\n    • Accordingly, DDR4/DDR5 memory demand is expected to increase by around 50%, and enterprise SSD demand is projected to nearly double.\n    • Due to HDD supply shortages and surging AI storage demand, the NAND market is also set to boom, with bit shipments in 2026 forecast to rise more than 50% YoY.\n\n⸻\n\nPrice and Profitability Surge\n    • DRAM operating margin is projected to climb from the current 40–50% range to nearly 70% in 2026, approaching the peak levels seen during the 2017 supercycle.\n    • NAND margins are expected to break out from the current breakeven point to 30–40%.\n    • HBM competition is likely to intensify (with Samsung expected to rejoin NVIDIA’s supply chain in 2026). However, blended ASP is still forecast to maintain 15%+ annual growth through 2025–2026, contributing to around 75% of SK Hynix’s operating profit.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0347222367293164, "ret_signed_m1d": -0.0347222367293164, "alpha_spy_m1d": -0.03132623940005319, "alpha_c_m1d": -0.012755684198757056, "bench_spy_m1d": -0.0033959973292632117, "bench_c_m1d": -0.021966552530559347, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09861113142104294, "ret_signed_p1d": 0.09861113142104294, "alpha_spy_p1d": 0.09745941981800299, "alpha_c_p1d": 0.08497571353500599, "bench_spy_p1d": 0.001151711603039951, "bench_c_p1d": 0.013635417886036949, "ret_p1w": 0.09861113142104294, "ret_signed_p1w": 0.09861113142104294, "alpha_spy_p1w": 0.09163988775525023, "alpha_c_p1w": 0.06151078057269155, "bench_spy_p1w": 0.006971243665792715, "bench_c_p1w": 0.03710035084835139, "ret_p1m": 0.5777775804812966, "ret_signed_p1m": 0.5777775804812966, "alpha_spy_p1m": 0.5607532407250881, "alpha_c_p1m": 0.48775384198129523, "bench_spy_p1m": 0.01702433975620843, "bench_c_p1m": 0.09002373850000134, "ret_p3m": 0.77905097839325, "ret_signed_p3m": 0.77905097839325, "alpha_spy_p3m": 0.7469880877036472, "alpha_c_p3m": 0.6841926626341372, "bench_spy_p3m": 0.032062890689602774, "bench_c_p3m": 0.09485831575911274, "ret_p6m": 1.5983083103366877, "ret_signed_p6m": 1.5983083103366877, "alpha_spy_p6m": 1.6277598772457416, "alpha_c_p6m": 1.453458232780157, "bench_spy_p6m": -0.029451566909053817, "bench_c_p6m": 0.14485007755653068, "price_path": [-0.2497, -0.2483, -0.2178, -0.2206, -0.1762, -0.1831, -0.1679, -0.172, -0.179, -0.2525, -0.2539, -0.2441, -0.2552, -0.2539, -0.2525, -0.2622, -0.2733, -0.2719, -0.2691, -0.2414, -0.2844, -0.2844, -0.2691, -0.283, -0.2733, -0.2885, -0.2594, -0.2539, -0.2289, -0.2331, -0.2331, -0.258, -0.2705, -0.2913, -0.3204, -0.3038, -0.2802, -0.2747, -0.2788, -0.2542, -0.2528, -0.2889, -0.2764, -0.2708, -0.2625, -0.2403, -0.2306, -0.2, -0.1556, -0.1472, -0.0875, -0.0806, -0.0333, -0.0736, -0.0125, -0.0125, -0.025, 0.0028, -0.0069, -0.0097, -0.0653, -0.0306, -0.0347, 0.0, 0.0986, 0.0986, 0.0986, 0.0986, 0.0986, 0.0986, 0.1889, 0.1528, 0.1431, 0.1736, 0.2569, 0.2931, 0.3486, 0.3306, 0.3375, 0.3292, 0.4167, 0.4861, 0.4472, 0.55, 0.5778, 0.5528, 0.7222, 0.6278, 0.6083, 0.6472, 0.6111, 0.6833, 0.7194, 0.7139, 0.7, 0.5556, 0.6833, 0.5833, 0.5611, 0.5861, 0.4472, 0.4444, 0.4417, 0.4556, 0.5122, 0.4733, 0.4955, 0.5511, 0.5344, 0.5066, 0.5122, 0.6039, 0.5733, 0.6317, 0.5706, 0.5872, 0.54, 0.4733, 0.5317, 0.5344, 0.5205, 0.6123, 0.6234, 0.6345, 0.6345, 0.6651, 0.7791, 0.8096, 0.8096, 0.8096, 0.8819, 0.9347, 1.0181, 1.0626, 1.1015, 1.0681, 1.082, 1.0515, 1.0626, 1.082, 1.1015, 1.1237, 1.0654, 1.057, 1.0987, 1.1321, 1.0459, 1.2238, 1.3378, 1.3934, 1.5268, 1.3072, 1.5212, 1.5018, 1.3406, 1.3322, 1.4657, 1.4351, 1.3906, 1.4684, 1.4462, 1.4462, 1.4462, 1.4462, 1.4851, 1.638, 1.6436, 1.7937, 1.8298, 2.0606, 1.9548, 1.9548, 1.615, 1.3644, 1.6206, 1.5732, 1.3282, 1.6122, 1.6596, 1.59, 1.5343, 1.7125, 1.7013, 1.9409, 1.8211, 1.8044, 1.5983, 1.7459, 1.771, 1.5983]}
{"unit_id": "quote:1959774440537161961", "ticker_idx": 0, "ts": "2025-08-25T00:26:52+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The Trump administration's actions this time will raise the valuation floor for Intel.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish INTC: US government equity involvement raises the valuation floor, preventing downside.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I’m glad that Kuo and I seem to share similar views!\n\nThat’s right. Intel is like a “too big to fail” entity—it’s simply too large to collapse.\n\nThe Trump administration’s actions this time will raise the valuation floor for Intel. https://t.co/dwrTa5X39k", "reply_text": null, "parent_text": null, "quoted_text": "Analysis of the U.S. Government’s Equity Investment in Intel\n\nThis deal doesn't guarantee the technology gain, but it prevents the valuation pain\n\nFull story\nhttps://t.co/OtvnRW6mXu https://t.co/DPXZBEeNqP", "parent_author": null, "ret_m1d": 0.010183299705468052, "ret_signed_m1d": 0.010183299705468052, "alpha_spy_m1d": 0.00576268683632164, "alpha_c_m1d": 0.011135983593143162, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00814659314892785, "ret_signed_p1d": -0.00814659314892785, "alpha_spy_p1d": -0.012333629498818666, "alpha_c_p1d": -0.017774980816592678, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.00814659314892785, "ret_signed_p1w": -0.00814659314892785, "alpha_spy_p1w": -0.012162423083761253, "alpha_c_p1w": 0.004203718725635741, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.1951120596491256, "ret_signed_p1m": 0.1951120596491256, "alpha_spy_p1m": 0.15996827835569216, "alpha_c_p1m": 0.10059659007085209, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.36945010088359576, "ret_signed_p3m": 0.36945010088359576, "alpha_spy_p3m": 0.3509758077615428, "alpha_c_p3m": 0.2633669221127297, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.8810591340248475, "ret_signed_p6m": 0.8810591340248475, "alpha_spy_p6m": 0.8121119059352686, "alpha_c_p6m": 0.4902793892903867, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1703, -0.1752, -0.2037, -0.1959, -0.1735, -0.1752, -0.1857, -0.1829, -0.1658, -0.1006, -0.1576, -0.154, -0.1796, -0.1552, -0.1527, -0.1246, -0.1246, -0.1413, -0.1369, -0.0815, -0.0957, -0.0835, -0.0758, -0.0876, -0.0692, -0.1088, -0.0839, -0.0839, -0.1039, -0.0391, -0.0452, -0.0297, -0.0456, -0.0509, -0.0664, -0.0758, -0.0713, -0.0591, -0.0525, -0.0534, -0.0432, -0.0782, -0.1568, -0.1576, -0.1686, -0.1715, -0.1935, -0.2134, -0.2057, -0.1776, -0.1686, -0.1947, -0.1874, -0.1589, -0.1116, -0.0949, -0.0281, 0.0004, -0.0363, 0.031, -0.0411, -0.0428, 0.0102, 0.0, -0.0081, 0.0122, 0.0155, -0.0081, -0.0081, -0.0138, -0.0224, 0.0024, -0.0024, -0.0029, -0.0045, 0.009, 0.0024, -0.0191, 0.009, 0.0293, 0.0143, 0.2452, 0.2049, 0.1715, 0.1951, 0.2717, 0.3845, 0.446, 0.4045, 0.3666, 0.464, 0.5193, 0.5002, 0.4904, 0.5141, 0.5246, 0.5397, 0.4815, 0.5161, 0.4513, 0.5132, 0.5006, 0.5075, 0.5519, 0.5527, 0.5039, 0.5544, 0.5593, 0.6106, 0.6916, 0.6839, 0.6358, 0.6289, 0.609, 0.5084, 0.5633, 0.5169, 0.5532, 0.5662, 0.543, 0.5434, 0.4627, 0.4468, 0.4138, 0.3984, 0.4301, 0.3695, 0.4053, 0.4578, 0.4595, 0.4994, 0.4994, 0.6521, 0.6297, 0.7707, 0.7825, 0.6497, 0.6868, 0.6415, 0.6497, 0.6611, 0.6094, 0.5401, 0.5279, 0.5198, 0.4684, 0.4778, 0.4998, 0.4815, 0.4807, 0.4729, 0.4729, 0.4745, 0.4941, 0.5193, 0.5031, 0.5031, 0.6041, 0.6037, 0.631, 0.7365, 0.6745, 0.8554, 0.7947, 0.9263, 0.9845, 0.9682, 0.9128, 0.9128, 0.978, 1.2098, 1.2126, 0.8358, 0.7308, 0.7894, 0.987, 0.9821, 0.8929, 0.9882, 1.0061, 0.9796, 0.965, 1.0607, 1.0464, 0.9198, 0.967, 0.8933, 0.9059, 0.9059, 0.8811]}
{"unit_id": "quote:1948315075749798314", "ticker_idx": 0, "ts": "2025-07-24T09:31:26+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still believe it'll take a significant amount of time for Intel's stock to rise substantially. For their foundry business to normalize... it's going to take at least two years.", "tweet_type": "analysis", "timeline": "years", "summary": "Intel stock recovery is years away; foundry normalization (external customers + tape-outs) needs at least two years before meaningful upside.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I've thought about this before, but I still believe it'll take a significant amount of time for Intel's stock to rise substantially.\n\nFor their foundry business to normalize, meaning securing external customers and having those products come to fruition, it's going to take at least two years.", "reply_text": null, "parent_text": null, "quoted_text": "Intel’s CEO Needs a Path to Profits to Win Over Stock Skeptics- BBG\n\nBack in April, Intel Corp.’s newly appointed Chief Executive Officer Lip-Bu Tan told investors on his inaugural earnings call that turning around the troubled chipmaker would take time. That was three months ago https://t.co/tA2PJnMvOH", "parent_author": null, "ret_m1d": 0.038002679728001754, "ret_signed_m1d": -0.038002679728001754, "alpha_spy_m1d": -0.038333622691325786, "alpha_c_m1d": -0.04213749955279977, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08528495224927657, "ret_signed_p1d": 0.08528495224927657, "alpha_spy_p1d": 0.08950930188959572, "alpha_c_p1d": 0.08420776961954057, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": -0.1250552376781523, "ret_signed_p1w": 0.1250552376781523, "alpha_spy_p1w": 0.1213669360801567, "alpha_c_p1w": 0.12846037715931546, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": 0.09589041788636266, "ret_signed_p1m": -0.09589041788636266, "alpha_spy_p1m": -0.07872504381505974, "alpha_c_p1m": -0.07559849681990616, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.6844896308247779, "ret_signed_p3m": -0.6844896308247779, "alpha_spy_p3m": -0.6234401137628296, "alpha_c_p3m": -0.48469815346991796, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.0751215566055667, "ret_signed_p6m": -1.0751215566055667, "alpha_spy_p6m": -0.9786444546374895, "alpha_c_p6m": -0.6795834349027565, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.0937, -0.1012, -0.1118, -0.1171, -0.0888, -0.1043, -0.1189, -0.1025, -0.072, -0.0535, -0.0199, -0.0031, -0.049, -0.0477, -0.0429, -0.0552, -0.0601, -0.0857, -0.0919, -0.114, -0.114, -0.0919, -0.0999, -0.1052, -0.1361, -0.1277, -0.1034, -0.1052, -0.1167, -0.1136, -0.095, -0.0243, -0.0862, -0.0822, -0.11, -0.0835, -0.0809, -0.0504, -0.0504, -0.0685, -0.0636, -0.0035, -0.019, -0.0057, 0.0027, -0.0102, 0.0097, -0.0331, -0.0062, -0.0062, -0.0278, 0.0424, 0.0358, 0.0526, 0.0354, 0.0296, 0.0128, 0.0027, 0.0075, 0.0208, 0.0278, 0.027, 0.038, 0.0, -0.0853, -0.0862, -0.0981, -0.1012, -0.1251, -0.1467, -0.1383, -0.1078, -0.0981, -0.1264, -0.1184, -0.0875, -0.0362, -0.0181, 0.0544, 0.0853, 0.0455, 0.1184, 0.0402, 0.0384, 0.0959, 0.0848, 0.076, 0.0981, 0.1016, 0.076, 0.076, 0.0698, 0.0605, 0.0875, 0.0822, 0.0817, 0.08, 0.0946, 0.0875, 0.0641, 0.0946, 0.1167, 0.1003, 0.3509, 0.3071, 0.2709, 0.2965, 0.3796, 0.502, 0.5687, 0.5236, 0.4825, 0.5882, 0.6483, 0.6275, 0.6169, 0.6425, 0.654, 0.6703, 0.6072, 0.6447, 0.5745, 0.6416, 0.6279, 0.6354, 0.6836, 0.6845, 0.6315, 0.6863, 0.6916, 0.7472, 0.8352, 0.8268, 0.7746, 0.7671, 0.7455, 0.6363, 0.696, 0.6456, 0.6849, 0.6991, 0.6739, 0.6743, 0.5868, 0.5696, 0.5338, 0.517, 0.5515, 0.4856, 0.5245, 0.5815, 0.5833, 0.6266, 0.6266, 0.7923, 0.768, 0.9209, 0.9337, 0.7897, 0.8299, 0.7808, 0.7897, 0.802, 0.7459, 0.6708, 0.6575, 0.6487, 0.593, 0.6032, 0.627, 0.6072, 0.6063, 0.5979, 0.5979, 0.5996, 0.6209, 0.6483, 0.6306, 0.6306, 0.7402, 0.7397, 0.7693, 0.8838, 0.8166, 1.0128, 0.947, 1.0897, 1.1529, 1.1352, 1.0751]}
{"unit_id": "orig:1955533727616110816", "ticker_idx": 1, "ts": "2025-08-13T07:35:47+00:00", "symbol": "LG Display", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SDC and LGD are expected to benefit in the long run", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays Meritz thesis: Apple to cut BOE share mid-to-long term, benefiting SDC and LGD at BOE's expense.", "resolved_tickers": ["034220.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "LG Display 034220.KS Korea", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Despite the recent ITC ruling, is Apple still expected to use BOE panels for the iPhone 17?\n\nAmong the four iPhone 17 models to be released this year, the iPhone 17 and iPhone 17 Air are expected to adopt panels from BOE. Even if the recent ITC ruling results in a ban on the use of BOE panels, Apple is likely to first allocate the models using BOE panels to other regions. However, in the mid to long term, considering the business uncertainty associated with using BOE, Apple is likely to significantly reduce their share. In such a case, SDC and LGD are expected to benefit in the long run.\n\nSource: Meritz Securities Korea", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.18359668924003014, "ret_signed_m1d": -0.18359668924003014, "alpha_spy_m1d": -0.18018531519342718, "alpha_c_m1d": -0.18322405223883187, "bench_spy_m1d": -0.0034113740466029663, "bench_c_m1d": -0.0003726370011982727, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.10609480812641081, "ret_signed_p1d": -0.10609480812641081, "alpha_spy_p1d": -0.10618777358124953, "alpha_c_p1d": -0.10393333079828315, "bench_spy_p1d": 9.296545483872265e-05, "bench_c_p1d": -0.002161477328127659, "ret_p1w": -0.14371708051166288, "ret_signed_p1w": -0.14371708051166288, "alpha_spy_p1w": -0.13320358477194327, "alpha_c_p1w": -0.1121863715340804, "bench_spy_p1w": -0.010513495739719603, "bench_c_p1w": -0.03153070897758248, "ret_p1m": -0.11587659894657643, "ret_signed_p1m": -0.11587659894657643, "alpha_spy_p1m": -0.1356319489939659, "alpha_c_p1m": -0.12612600079167202, "bench_spy_p1m": 0.019755350047389486, "bench_c_p1m": 0.010249401845095596, "ret_p3m": 0.03310759969902177, "ret_signed_p3m": 0.03310759969902177, "alpha_spy_p3m": -0.026498337491073443, "alpha_c_p3m": -0.0697577995450378, "bench_spy_p3m": 0.059605937190095215, "bench_c_p3m": 0.10286539924405957, "ret_p6m": -0.13844996237772766, "ret_signed_p6m": -0.13844996237772766, "alpha_spy_p6m": -0.19522942522065712, "alpha_c_p6m": -0.15227955729287135, "bench_spy_p6m": 0.05677946284292945, "bench_c_p6m": 0.013829594915143684, "price_path": [-0.3552, -0.3725, -0.3785, -0.3755, -0.3837, -0.392, -0.3853, -0.3913, -0.3687, -0.3604, -0.3521, -0.3747, -0.3747, -0.3664, -0.3604, -0.3604, -0.3416, -0.3394, -0.3318, -0.319, -0.3348, -0.3386, -0.3273, -0.307, -0.3085, -0.3062, -0.322, -0.2904, -0.2972, -0.3002, -0.3138, -0.3258, -0.3115, -0.3138, -0.2837, -0.3002, -0.3108, -0.3108, -0.3017, -0.2987, -0.3047, -0.3032, -0.3047, -0.3093, -0.307, -0.307, -0.31, -0.3145, -0.307, -0.3062, -0.2438, -0.2175, -0.2212, -0.2084, -0.1843, -0.2054, -0.2107, -0.1889, -0.1761, -0.1723, -0.1776, -0.1685, -0.1836, 0.0, -0.1061, -0.1061, -0.1016, -0.1294, -0.1437, -0.14, -0.1422, -0.1226, -0.1151, -0.1106, -0.0993, -0.0971, -0.1226, -0.1166, -0.0971, -0.0926, -0.0978, -0.1038, -0.1084, -0.1242, -0.1159, -0.0895, -0.0933, 0.0015, -0.0309, -0.0128, -0.0128, 0.0038, 0.0023, 0.0271, 0.076, 0.0557, 0.1008, 0.0843, 0.0858, 0.1031, 0.1031, 0.1031, 0.1031, 0.1031, 0.1031, 0.1753, 0.143, 0.0557, 0.0903, 0.082, 0.0813, 0.0933, 0.0602, 0.0677, 0.0647, 0.0361, 0.0715, 0.0767, 0.091, 0.0497, 0.1061, 0.0926, 0.0963, 0.0542, 0.0564, 0.009, 0.0331, 0.015, 0.0233, 0.0128, -0.0075, -0.0286, -0.0873, -0.0685, -0.0594, -0.0858, -0.0805, -0.0798, -0.0504, -0.0444, -0.0564, -0.0722, -0.0587, -0.0376, -0.0331, -0.0015, -0.0263, -0.0339, -0.0459, -0.0609, -0.0579, -0.0745, -0.0933, -0.085, -0.1053, -0.0993, -0.1144, -0.1031, -0.0933, -0.0933, -0.1144, -0.0971, -0.1114, -0.1114, -0.1114, -0.1099, -0.0956, -0.0737, -0.0564, -0.0798, -0.1136, -0.0986, -0.0873, -0.0933, -0.079, -0.0579, -0.0625, -0.0617, -0.0828, -0.0474, -0.0572, -0.064, -0.0752, -0.0873, -0.1114, -0.1219, -0.1693, -0.1445, -0.1309, -0.1384]}
{"unit_id": "orig:1955533727616110816", "ticker_idx": 2, "ts": "2025-08-13T07:35:47+00:00", "symbol": "BOE Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Apple is likely to significantly reduce their share", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays Meritz thesis: Apple to cut BOE share mid-to-long term, benefiting SDC and LGD at BOE's expense.", "resolved_tickers": ["200725.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "BOE Technology Group 200725.SZ Shenzhen B-share", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Despite the recent ITC ruling, is Apple still expected to use BOE panels for the iPhone 17?\n\nAmong the four iPhone 17 models to be released this year, the iPhone 17 and iPhone 17 Air are expected to adopt panels from BOE. Even if the recent ITC ruling results in a ban on the use of BOE panels, Apple is likely to first allocate the models using BOE panels to other regions. However, in the mid to long term, considering the business uncertainty associated with using BOE, Apple is likely to significantly reduce their share. In such a case, SDC and LGD are expected to benefit in the long run.\n\nSource: Meritz Securities Korea", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003623184962974646, "ret_signed_m1d": -0.003623184962974646, "alpha_spy_m1d": -0.007034559009577612, "alpha_c_m1d": -0.003995821964172919, "bench_spy_m1d": -0.0034113740466029663, "bench_c_m1d": -0.0003726370011982727, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003623184962974757, "ret_signed_p1d": 0.003623184962974757, "alpha_spy_p1d": 0.0037161504178134797, "alpha_c_p1d": 0.001461707634847098, "bench_spy_p1d": 9.296545483872265e-05, "bench_c_p1d": -0.002161477328127659, "ret_p1w": 0.01086955488892416, "ret_signed_p1w": -0.01086955488892416, "alpha_spy_p1w": -0.021383050628643763, "alpha_c_p1w": -0.04240026386650664, "bench_spy_p1w": -0.010513495739719603, "bench_c_p1w": -0.03153070897758248, "ret_p1m": 0.08333334053196206, "ret_signed_p1m": -0.08333334053196206, "alpha_spy_p1m": -0.06357799048457258, "alpha_c_p1m": -0.07308393868686647, "bench_spy_p1m": 0.019755350047389486, "bench_c_p1m": 0.010249401845095596, "ret_p3m": 0.06521741571708839, "ret_signed_p3m": -0.06521741571708839, "alpha_spy_p3m": -0.005611478526993174, "alpha_c_p3m": 0.037647983526971185, "bench_spy_p3m": 0.059605937190095215, "bench_c_p3m": 0.10286539924405957, "ret_p6m": 0.14492757128607558, "ret_signed_p6m": -0.14492757128607558, "alpha_spy_p6m": -0.08814810844314613, "alpha_c_p6m": -0.1310979763709319, "bench_spy_p6m": 0.05677946284292945, "bench_c_p6m": 0.013829594915143684, "price_path": [-0.0814, -0.0814, -0.0849, -0.092, -0.0884, -0.0884, -0.0884, -0.0884, -0.0884, -0.0778, -0.0884, -0.0884, -0.0778, -0.0707, -0.0672, -0.0672, -0.0672, -0.0743, -0.0707, -0.0707, -0.0849, -0.0849, -0.0814, -0.0743, -0.0849, -0.092, -0.0884, -0.0814, -0.0652, -0.0688, -0.0616, -0.0543, -0.0471, -0.0471, -0.029, -0.0399, -0.0507, -0.0435, -0.0326, -0.0217, -0.0072, -0.0145, -0.0145, -0.0109, 0.0036, 0.0, 0.0181, 0.0217, 0.0254, 0.0254, 0.0181, 0.0254, 0.0217, 0.0145, -0.0217, -0.0109, -0.0036, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0036, 0.0, -0.0036, -0.0072, -0.0036, 0.0, 0.0109, 0.0181, 0.0362, 0.0507, 0.0326, 0.0362, 0.0507, 0.0725, 0.058, 0.0399, 0.0435, 0.0399, 0.058, 0.0652, 0.0688, 0.0725, 0.0833, 0.087, 0.0761, 0.0761, 0.0797, 0.0833, 0.0543, 0.0688, 0.058, 0.0688, 0.0725, 0.0761, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.1087, 0.1014, 0.1051, 0.087, 0.0906, 0.087, 0.0688, 0.0652, 0.0688, 0.0652, 0.0616, 0.058, 0.058, 0.0616, 0.0543, 0.0471, 0.0399, 0.0471, 0.029, 0.0362, 0.0399, 0.0399, 0.0652, 0.0688, 0.0652, 0.0652, 0.0652, 0.0543, 0.0399, 0.0217, 0.0217, -0.0109, 0.0, 0.0109, 0.0145, 0.0072, 0.0181, 0.029, 0.0217, 0.0435, 0.0507, 0.0688, 0.0616, 0.0471, 0.0471, 0.0399, 0.0435, 0.029, 0.0181, 0.0217, 0.0181, 0.0181, 0.029, 0.0254, 0.029, 0.0326, 0.029, 0.0435, 0.0399, 0.0471, 0.0471, 0.0471, 0.058, 0.087, 0.0652, 0.0833, 0.0906, 0.1159, 0.1051, 0.1486, 0.1558, 0.1522, 0.1884, 0.1703, 0.1522, 0.1268, 0.1558, 0.1667, 0.1812, 0.1993, 0.1739, 0.1558, 0.1232, 0.1667, 0.1667, 0.1449]}
{"unit_id": "orig:1945462652224082057", "ticker_idx": 0, "ts": "2025-07-16T12:36:55+00:00", "symbol": "VNET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our 'Buy' rating on VNET/GDS", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates Buy on VNET and GDS; H20 sales resumption and RTX6000D shipments drive upside in orders/revenue/EBITDA for Chinese IDCs.", "resolved_tickers": ["VNET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "original", "lang": "en", "tweet_text": "Citi) China Data Centers - Thoughts on H20 Sales and Latest RTX6000D Shipments; 'Buy' on VNET/GDS\n\nAccording to a Reuters report on July 15, NVIDIA will resume sales of H20 chips to China and has launched new models that meet the regulatory requirements of the Chinese market. CEO Jensen Huang expects licenses to be granted this time. We believe this is positive for Chinese Internet Data Centers (IDCs) such as VNET/GDS, as more orders could materialize if NVIDIA is ultimately able to sell its accumulated H20 inventory to Chinese Cloud Service Providers (CSPs), leading to an upside in VNET/GDS's EBITDA and revenue.\n\nDigitimes reported on July 15 that shipments of RTX6000D (a downgraded version) could reach 1-2 million units in 2025. While shipments may not be very high this year, we believe both pieces of news are positive for VNET/GDS. The previous sector correction was mainly attributable to the H20 sales ban and weakened investment sentiment towards AI accelerator development in China.\n\nWe expect VNET/GDS to benefit from potential H20/RTX6000D sales, securing more upside potential in terms of orders/revenue/EBITDA. At the same time, if the H20 license approval and RTX6000D shipments materialize, the beta of AI-related stocks across China could also be re-rated. We maintain our 'Buy' rating on VNET/GDS.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.028441410817400925, "ret_signed_m1d": 0.028441410817400925, "alpha_spy_m1d": 0.03177346225115629, "alpha_c_m1d": 0.03153360042116293, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.003092189603762008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003412938919383879, "ret_signed_p1d": -0.003412938919383879, "alpha_spy_p1d": -0.00953262889690143, "alpha_c_p1d": -0.012496334709230505, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.009083395789846627, "ret_p1w": -0.005688303862554589, "ret_signed_p1w": -0.005688303862554589, "alpha_spy_p1w": -0.021692425662735992, "alpha_c_p1w": -0.01179557967525513, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.006107275812700541, "ret_p1m": -0.09556315770572643, "ret_signed_p1m": -0.09556315770572643, "alpha_spy_p1m": -0.128772639818297, "alpha_c_p1m": -0.13042857852681722, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.03486542082109079, "ret_p3m": 0.07053468580770961, "ret_signed_p3m": 0.07053468580770961, "alpha_spy_p3m": 0.00540018495968364, "alpha_c_p3m": -0.033158602335849, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1036932881435586, "ret_p6m": 0.1740614038237891, "ret_signed_p6m": 0.1740614038237891, "alpha_spy_p6m": 0.06313133637614987, "alpha_c_p6m": 0.05586391301223981, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11819749081154929, "price_path": [-0.4289, -0.4539, -0.446, -0.3982, -0.3834, -0.3879, -0.3811, -0.3561, -0.2844, -0.2867, -0.1945, -0.2082, -0.223, -0.2833, -0.3174, -0.322, -0.1775, -0.273, -0.2912, -0.3185, -0.2685, -0.2833, -0.2924, -0.2628, -0.3117, -0.3185, -0.3185, -0.3129, -0.3868, -0.3788, -0.3925, -0.372, -0.3697, -0.3379, -0.314, -0.3276, -0.3003, -0.3106, -0.3151, -0.3424, -0.3811, -0.3538, -0.3686, -0.3811, -0.3811, -0.3766, -0.38, -0.3356, -0.3174, -0.3276, -0.2059, -0.215, -0.248, -0.1945, -0.1183, -0.1183, -0.1422, -0.1399, -0.1832, -0.2071, -0.0956, -0.099, 0.0284, 0.0, -0.0034, 0.025, -0.041, -0.0796, -0.0057, -0.0523, -0.0603, -0.0512, -0.058, -0.0876, -0.0546, -0.1479, -0.1104, -0.1092, -0.0887, -0.0728, -0.0922, -0.0876, -0.1126, -0.0887, -0.0956, -0.116, -0.1183, -0.1081, -0.0922, -0.165, -0.0569, -0.1081, -0.1115, -0.1126, -0.0296, -0.0148, -0.0148, -0.083, -0.0899, -0.1308, -0.1013, -0.1263, -0.0523, -0.0978, 0.0364, 0.0887, 0.091, 0.091, 0.1206, 0.132, 0.0956, 0.1433, 0.0637, 0.1797, 0.2662, 0.2287, 0.1843, 0.1752, 0.2241, 0.2787, 0.2036, 0.2071, 0.1297, 0.1251, 0.0899, -0.0319, 0.0705, 0.0444, 0.091, 0.0273, -0.0034, 0.066, 0.0648, 0.0319, 0.0683, 0.0944, 0.1399, 0.165, 0.256, 0.1843, 0.1832, 0.1536, 0.0557, 0.1183, 0.1342, 0.0978, 0.1035, 0.1138, 0.0205, -0.0023, 0.0011, -0.0239, -0.0137, -0.0535, -0.0648, -0.0455, 0.0114, 0.0171, 0.0137, 0.0137, 0.0171, 0.0205, -0.0023, 0.0114, 0.0068, 0.0171, 0.0614, 0.083, 0.0341, 0.0512, 0.0444, 0.0137, -0.0046, -0.0387, -0.033, -0.0046, -0.0102, -0.0205, -0.0239, -0.0239, -0.0114, -0.0171, -0.0353, -0.0375, -0.0375, 0.0364, 0.0626, 0.0569, 0.0592, 0.1741]}
{"unit_id": "orig:1945462652224082057", "ticker_idx": 1, "ts": "2025-07-16T12:36:55+00:00", "symbol": "GDS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our 'Buy' rating on VNET/GDS", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates Buy on VNET and GDS; H20 sales resumption and RTX6000D shipments drive upside in orders/revenue/EBITDA for Chinese IDCs.", "resolved_tickers": ["GDS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "original", "lang": "en", "tweet_text": "Citi) China Data Centers - Thoughts on H20 Sales and Latest RTX6000D Shipments; 'Buy' on VNET/GDS\n\nAccording to a Reuters report on July 15, NVIDIA will resume sales of H20 chips to China and has launched new models that meet the regulatory requirements of the Chinese market. CEO Jensen Huang expects licenses to be granted this time. We believe this is positive for Chinese Internet Data Centers (IDCs) such as VNET/GDS, as more orders could materialize if NVIDIA is ultimately able to sell its accumulated H20 inventory to Chinese Cloud Service Providers (CSPs), leading to an upside in VNET/GDS's EBITDA and revenue.\n\nDigitimes reported on July 15 that shipments of RTX6000D (a downgraded version) could reach 1-2 million units in 2025. While shipments may not be very high this year, we believe both pieces of news are positive for VNET/GDS. The previous sector correction was mainly attributable to the H20 sales ban and weakened investment sentiment towards AI accelerator development in China.\n\nWe expect VNET/GDS to benefit from potential H20/RTX6000D sales, securing more upside potential in terms of orders/revenue/EBITDA. At the same time, if the H20 license approval and RTX6000D shipments materialize, the beta of AI-related stocks across China could also be re-rated. We maintain our 'Buy' rating on VNET/GDS.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010493119684929164, "ret_signed_m1d": -0.010493119684929164, "alpha_spy_m1d": -0.007161068251173797, "alpha_c_m1d": -0.007400930081167156, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.003092189603762008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004459553350173895, "ret_signed_p1d": -0.004459553350173895, "alpha_spy_p1d": -0.010579243327691445, "alpha_c_p1d": -0.013542949140020522, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.009083395789846627, "ret_p1w": -0.03672606900339226, "ret_signed_p1w": -0.03672606900339226, "alpha_spy_p1w": -0.05273019080357366, "alpha_c_p1w": -0.0428333448160928, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.006107275812700541, "ret_p1m": -0.164742889697305, "ret_signed_p1m": -0.164742889697305, "alpha_spy_p1m": -0.19795237180987557, "alpha_c_p1m": -0.19960831051839578, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.03486542082109079, "ret_p3m": -0.06243434732699493, "ret_signed_p3m": -0.06243434732699493, "alpha_spy_p3m": -0.1275688481750209, "alpha_c_p3m": -0.16612763547055354, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1036932881435586, "ret_p6m": 0.09522563404583373, "ret_signed_p6m": 0.09522563404583373, "alpha_spy_p6m": -0.015704433401805495, "alpha_c_p6m": -0.02297185676571556, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11819749081154929, "price_path": [-0.4368, -0.4436, -0.4124, -0.3859, -0.3948, -0.3927, -0.3867, -0.3746, -0.3387, -0.3242, -0.2629, -0.271, -0.2639, -0.313, -0.3061, -0.3324, -0.2358, -0.2828, -0.2841, -0.3064, -0.2857, -0.2896, -0.2765, -0.2222, -0.2613, -0.2524, -0.2524, -0.288, -0.3502, -0.3418, -0.3785, -0.3604, -0.3704, -0.3578, -0.3269, -0.3368, -0.3161, -0.3132, -0.2946, -0.2991, -0.3263, -0.2623, -0.2859, -0.2852, -0.2852, -0.2904, -0.2775, -0.2403, -0.2463, -0.2211, -0.2086, -0.1981, -0.2269, -0.1911, -0.1086, -0.1086, -0.1522, -0.1508, -0.1545, -0.1472, -0.1104, -0.0818, -0.0105, 0.0, -0.0045, -0.0142, -0.0624, -0.0805, -0.0367, -0.0551, -0.0352, -0.042, -0.0464, -0.0748, -0.0538, -0.1049, -0.0837, -0.0674, -0.0643, -0.0223, -0.0459, -0.0328, -0.0475, -0.1102, -0.1647, -0.176, -0.1545, -0.1713, -0.1107, -0.1605, -0.1209, -0.1288, -0.1133, -0.1167, -0.0797, -0.0934, -0.0934, -0.1348, -0.1039, -0.1406, -0.1299, -0.1209, -0.1013, -0.1212, 0.01, 0.0024, -0.0241, -0.021, 0.053, 0.0231, 0.022, 0.043, -0.0129, 0.0669, 0.1036, 0.0514, 0.0485, 0.0152, 0.0554, 0.0847, 0.0396, 0.0404, -0.0163, -0.0073, 0.0079, -0.1262, -0.0624, -0.1162, -0.1118, -0.127, -0.1414, -0.0978, -0.1047, -0.1254, -0.1013, -0.0839, -0.0582, -0.059, -0.0115, -0.0679, -0.0635, -0.0889, -0.1338, -0.1036, -0.1104, -0.1385, -0.1233, -0.1257, -0.1592, -0.1973, -0.2219, -0.2301, -0.2387, -0.2196, -0.2054, -0.1889, -0.1209, -0.1062, -0.1055, -0.1055, -0.1089, -0.0918, -0.1133, -0.1149, -0.1139, -0.0763, -0.0651, -0.0551, -0.0496, -0.0517, -0.0488, -0.0666, -0.0887, -0.1162, -0.096, -0.0575, -0.0543, -0.0837, -0.0871, -0.0871, -0.0714, -0.091, -0.0923, -0.0845, -0.0845, 0.0058, -0.0155, -0.0118, 0.0084, 0.0952]}
{"unit_id": "quote:1973225787672371290", "ticker_idx": 0, "ts": "2025-10-01T03:17:43+00:00", "symbol": "GOOG", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi highlighted that this has driven a concurrent increase in Google Cloud demand, with new customers up 28% QoQ, and the number of large contracts over $250 million doubling YoY", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Citi's bullish GOOG thesis: Gemini token usage surging, Cloud demand accelerating, 60% of gen-AI startups on Google Cloud.", "resolved_tickers": ["GOOG"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "quote", "lang": "en", "tweet_text": "As of June 2025, Gemini’s token usage had increased to 980 trillion per month, a sharp rise from 480 trillion in April. Citi highlighted that this has driven a concurrent increase in Google Cloud demand, with new customers up 28% QoQ, and the number of large contracts over $250 million doubling YoY.\n\nIn addition, Gemini’s monthly active users (MAU) reached 450 million, and Citi assessed that the figure may have surpassed 500 million by September.\n\nAt the same time, Gemini’s daily request volume rose 50% QoQ in 2Q, with new features such as AI-Overviews and AI-Mode being rolled out rapidly. Citi also pointed out that about 60% of generative AI startups have chosen Google Cloud, and 9 out of the top 10 AI labs are using Google Cloud.\n\n$GOOG", "reply_text": null, "parent_text": null, "quoted_text": "Citi commented that Google’s 2026 capital expenditures (Capex) are expected to reach $111 billion, representing a 29% YoY increase, or about an additional $25 billion. The 2024–2029 CAGR is projected at 26%, with roughly two-thirds allocated to servers and the remaining one-third https://t.co/7wHrMKaA6D", "parent_author": null, "ret_m1d": -0.00810453284825241, "ret_signed_m1d": -0.00810453284825241, "alpha_spy_m1d": -0.004708535518989199, "alpha_c_m1d": -0.0223278533489647, "bench_spy_m1d": -0.0033959973292632117, "bench_c_m1d": 0.014223320500712289, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0036246656481162542, "ret_signed_p1d": 0.0036246656481162542, "alpha_spy_p1d": 0.0024729540450763032, "alpha_c_p1d": 0.005852429702895567, "bench_spy_p1d": 0.001151711603039951, "bench_c_p1d": -0.002227764054779313, "ret_p1w": -0.00032576439822640246, "ret_signed_p1w": -0.00032576439822640246, "alpha_spy_p1w": -0.007297008064019117, "alpha_c_p1w": 0.009356301742569295, "bench_spy_p1w": 0.006971243665792715, "bench_c_p1w": -0.009682066140795698, "ret_p1m": 0.14808172996265823, "ret_signed_p1m": 0.14808172996265823, "alpha_spy_p1m": 0.1310573902064498, "alpha_c_p1m": 0.1685597771673607, "bench_spy_p1m": 0.01702433975620843, "bench_c_p1m": -0.020478047204702476, "ret_p3m": 0.281237912180639, "ret_signed_p3m": 0.281237912180639, "alpha_spy_p3m": 0.2491750214910362, "alpha_c_p3m": 0.2678572484831401, "bench_spy_p3m": 0.032062890689602774, "bench_c_p3m": 0.013380663697498907, "ret_p6m": 0.14490946301083718, "ret_signed_p6m": 0.14490946301083718, "alpha_spy_p6m": 0.174361029919891, "alpha_c_p6m": 0.20659558585201432, "bench_spy_p6m": -0.029451566909053817, "bench_c_p6m": -0.06168612284117714, "price_path": [-0.2653, -0.2775, -0.2873, -0.2771, -0.2729, -0.2622, -0.2561, -0.255, -0.2522, -0.2485, -0.2434, -0.2222, -0.2183, -0.2207, -0.2139, -0.2103, -0.213, -0.2007, -0.1966, -0.2152, -0.2271, -0.2035, -0.2052, -0.1987, -0.1973, -0.1777, -0.1796, -0.1693, -0.1739, -0.1707, -0.1662, -0.1687, -0.1761, -0.1854, -0.1837, -0.1589, -0.1489, -0.1538, -0.1528, -0.1359, -0.1311, -0.1311, -0.1374, -0.0596, -0.0533, -0.0431, -0.0463, -0.0228, -0.0244, -0.0194, -0.0169, 0.0253, 0.0239, 0.0176, 0.0277, 0.0395, 0.0299, 0.0277, 0.0093, 0.0042, 0.0067, -0.0048, -0.0081, 0.0, 0.0036, 0.0037, 0.0243, 0.0065, -0.0003, -0.0136, -0.0328, -0.0037, 0.0026, 0.0251, 0.0258, 0.0336, 0.0468, 0.0236, 0.0285, 0.0334, 0.061, 0.0993, 0.0932, 0.1207, 0.1481, 0.1478, 0.1571, 0.1324, 0.1597, 0.1621, 0.1391, 0.1835, 0.1882, 0.1706, 0.1368, 0.128, 0.1632, 0.1605, 0.1932, 0.181, 0.2204, 0.297, 0.3181, 0.3044, 0.3044, 0.3037, 0.2834, 0.287, 0.3058, 0.2967, 0.3118, 0.2815, 0.2949, 0.3082, 0.2784, 0.2655, 0.2606, 0.2541, 0.2147, 0.2379, 0.2577, 0.2688, 0.2865, 0.2865, 0.2865, 0.2836, 0.2812, 0.2819, 0.2788, 0.2788, 0.285, 0.2932, 0.2819, 0.314, 0.3286, 0.3413, 0.356, 0.3711, 0.3706, 0.3577, 0.3462, 0.3462, 0.3129, 0.3383, 0.3483, 0.3385, 0.3595, 0.3652, 0.3704, 0.3801, 0.3796, 0.4056, 0.3885, 0.3585, 0.3503, 0.3167, 0.322, 0.2985, 0.2688, 0.2608, 0.2471, 0.2471, 0.2341, 0.2387, 0.2371, 0.2833, 0.2702, 0.2671, 0.2757, 0.2517, 0.2692, 0.2485, 0.2371, 0.2367, 0.2263, 0.2157, 0.248, 0.2517, 0.2578, 0.2365, 0.2294, 0.2415, 0.2618, 0.2491, 0.2468, 0.2185, 0.2195, 0.1794, 0.181, 0.1449]}
{"unit_id": "quote:1944949363194900925", "ticker_idx": 0, "ts": "2025-07-15T02:37:18+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Street estimates should rise by about 10% on the news.", "tweet_type": "analysis", "timeline": "weeks", "summary": "H20 China resumption means consensus estimates rise ~10%; bullish for NVDA near-term.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"In reality a handful of the 40 analysts have already factored in the return of the H20 before tonight’s announcement.  \n\nThat means that Street estimates should rise by about 10% on the news.\"", "reply_text": null, "parent_text": null, "quoted_text": "What does Nvidia resuming H20 to China mean to numbers.  \n\nIn theory that should add back about 15% to Street estimates. Last quarter the company outlined that impact. \n\nIn reality a handful of the 40 analysts have already factored in the return of the H20 before tonight’s", "parent_author": null, "ret_m1d": -0.03884002840913159, "ret_signed_m1d": -0.03884002840913159, "alpha_spy_m1d": -0.04313153377910328, "alpha_c_m1d": -0.019995527465069607, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00392503050477, "ret_signed_p1d": 0.00392503050477, "alpha_spy_p1d": 0.0005818393862548898, "alpha_c_p1d": 0.009152005413437636, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.021499604475081924, "ret_signed_p1w": -0.021499604475081924, "alpha_spy_p1w": -0.032300987895897304, "alpha_c_p1w": -0.002827009559373761, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.06379609510335871, "ret_signed_p1m": 0.06379609510335871, "alpha_spy_p1m": 0.02722876118029305, "alpha_c_p1m": 0.027757517411518773, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.07305409607893898, "ret_signed_p3m": 0.07305409607893898, "alpha_spy_p3m": 0.020508977938374917, "alpha_c_p3m": -0.045721705615825003, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.10797419003219977, "ret_signed_p6m": 0.10797419003219977, "alpha_spy_p6m": -0.0067831132489872825, "alpha_c_p6m": -0.22014830980357902, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.4055, -0.4055, -0.4323, -0.4207, -0.3983, -0.3766, -0.3497, -0.3631, -0.3614, -0.362, -0.3462, -0.3293, -0.3333, -0.3349, -0.3143, -0.3125, -0.3167, -0.2795, -0.2389, -0.2072, -0.2102, -0.2069, -0.2059, -0.2128, -0.2279, -0.2219, -0.2309, -0.2309, -0.2063, -0.2103, -0.1846, -0.2084, -0.1953, -0.1728, -0.1687, -0.18, -0.1698, -0.1645, -0.1567, -0.1633, -0.1506, -0.1683, -0.1524, -0.1557, -0.1477, -0.1477, -0.1573, -0.1554, -0.1336, -0.096, -0.0919, -0.0759, -0.0745, -0.1019, -0.0788, -0.0665, -0.0665, -0.073, -0.0627, -0.0458, -0.0387, -0.0339, -0.0388, 0.0, 0.0039, 0.0135, 0.01, 0.004, -0.0215, 0.0005, 0.0178, 0.0164, 0.0354, 0.0282, 0.0502, 0.042, 0.0177, 0.0545, 0.0443, 0.0511, 0.059, 0.0703, 0.0665, 0.073, 0.0638, 0.0663, 0.0571, 0.0663, 0.0289, 0.0275, 0.0251, 0.0427, 0.0534, 0.0649, 0.0639, 0.0555, 0.0204, 0.0204, 0.0005, -0.0005, 0.0056, -0.0216, -0.014, 0.0004, 0.0388, 0.038, 0.0418, 0.0414, 0.0245, -0.0023, 0.0325, 0.035, 0.0757, 0.0453, 0.0368, 0.041, 0.0439, 0.0654, 0.0931, 0.097, 0.1066, 0.0992, 0.087, 0.0841, 0.1079, 0.1282, 0.0731, 0.1033, 0.0547, 0.0535, 0.0651, 0.0734, 0.07, 0.0613, 0.0562, 0.0672, 0.0912, 0.1219, 0.1777, 0.213, 0.1886, 0.1863, 0.212, 0.164, 0.1436, 0.1019, 0.1023, 0.1661, 0.1316, 0.1354, 0.0947, 0.1141, 0.0932, 0.0625, 0.0927, 0.0583, 0.048, 0.0695, 0.0418, 0.0561, 0.0561, 0.037, 0.0541, 0.0631, 0.0521, 0.0744, 0.0687, 0.0871, 0.0837, 0.0767, 0.06, 0.0254, 0.0329, 0.0412, 0.0015, 0.0203, 0.0604, 0.0762, 0.1086, 0.105, 0.105, 0.1163, 0.1028, 0.0988, 0.0927, 0.0927, 0.1065, 0.1022, 0.097, 0.108]}
{"unit_id": "orig:1945646078088679591", "ticker_idx": 0, "ts": "2025-07-17T00:45:48+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi has added Samsung Electronics to its 30-day Upside Catalyst Watch list… Citi has raised its target price from KRW 83,000 to KRW 90,000", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi's 30-day Upside Catalyst Watch add on Samsung with PT raised to KRW 90,000 on foundry recovery and GDDR7/HBM demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250717_SEC: Potential Tailwinds from Foundry and Memory Businesses - CITI\n\n(1) Citi has added Samsung Electronics to its 30-day Upside Catalyst Watch list.\n\n(2) This is due to Citi's assessment that, despite potentially disappointing 2Q25 earnings, investor sentiment for 2H25 is likely to improve.\n\n(3) This improvement is attributed to 1) the gradual recovery of foundry utilization rates, and 2) increased demand for GDDR7 and eSSD driven by the proliferation of AI inference models.\n\n(4) Samsung Electronics holds a strong market position in the GDDR7 segment, and the possibility of passing HBM3E 12-Hi qualification tests is a positive factor. Consequently, Citi has raised its target price from KRW 83,000 to KRW 90,000.\n\n(5) Foundry utilization rates are expected to recover: improving from 60% in 1H25 to 77% in 2H25.\n\n(6) The improvement is attributed to increased Exynos chip production, and the foundry operating loss is projected to shrink to approximately KRW 1.2 trillion in 3Q25.\n\n(7) Furthermore, in the mid-to-long term, Samsung is expected to produce Exynos on a 2nm process, and there is a possibility of securing additional 2nm orders from external clients.\n\n(8) Meanwhile, given the limited HBM supply, GDDR7 and similar products are expected to be more actively utilized as alternatives.\n\n(9) This is due to expanding memory demand for AI inference. If Samsung passes NVIDIA's qualification tests in 3Q25, there is a possibility of an upward revision in earnings.\n\n(10) HBM4 qualification is expected around 1Q26.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029985112110534562, "ret_signed_m1d": -0.029985112110534562, "alpha_spy_m1d": -0.023902644946360296, "alpha_c_m1d": -0.020983481692528505, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.009001630418006057, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005996856566729658, "ret_signed_p1d": 0.005996856566729658, "alpha_spy_p1d": 0.0067292694688401156, "alpha_c_p1d": 0.006647803097681093, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.0006509465309514351, "ret_p1w": -0.010494854396156716, "ret_signed_p1w": -0.010494854396156716, "alpha_spy_p1w": -0.0206534693722904, "alpha_c_p1w": -0.011759188336187165, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": 0.0012643339400304487, "ret_p1m": 0.07346315149621097, "ret_signed_p1m": 0.07346315149621097, "alpha_spy_p1m": 0.0489423634853976, "alpha_c_p1m": 0.055689353420388166, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.0177737980758228, "ret_p3m": 0.3794403778943689, "ret_signed_p3m": 0.3794403778943689, "alpha_spy_p3m": 0.3220778346839188, "alpha_c_p3m": 0.29971985062533135, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.07972052726903756, "ret_p6m": 1.1034312041965202, "ret_signed_p6m": 1.1034312041965202, "alpha_spy_p6m": 0.9919560591424046, "alpha_c_p6m": 0.9806256692495765, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.12280553494694368, "price_path": [-0.1745, -0.1804, -0.17, -0.17, -0.17, -0.1685, -0.1685, -0.173, -0.173, -0.1909, -0.1909, -0.1909, -0.1864, -0.1864, -0.1834, -0.1417, -0.1521, -0.1447, -0.1462, -0.1536, -0.1685, -0.167, -0.17, -0.1849, -0.1924, -0.1849, -0.1968, -0.167, -0.164, -0.1626, -0.1536, -0.1536, -0.1387, -0.1193, -0.1193, -0.1089, -0.1179, -0.1074, -0.1134, -0.1313, -0.1477, -0.1342, -0.1089, -0.1179, -0.1134, -0.1357, -0.0985, -0.0866, -0.103, -0.0885, -0.1034, -0.0975, -0.0885, -0.0435, -0.051, -0.075, -0.0795, -0.0945, -0.0855, -0.0615, -0.063, -0.045, -0.03, 0.0, 0.006, 0.0165, -0.0105, -0.0045, -0.0105, -0.012, 0.0555, 0.0585, 0.0885, 0.0705, 0.033, 0.045, 0.048, 0.0315, 0.057, 0.0765, 0.0645, 0.066, 0.078, 0.0735, 0.0735, 0.0495, 0.0495, 0.057, 0.0585, 0.0705, 0.072, 0.054, 0.0585, 0.0435, 0.045, 0.0135, 0.036, 0.0465, 0.051, 0.042, 0.051, 0.072, 0.0885, 0.1004, 0.1304, 0.1469, 0.1904, 0.1724, 0.2039, 0.2039, 0.2519, 0.2699, 0.2804, 0.2909, 0.2489, 0.268, 0.2635, 0.2951, 0.3516, 0.3516, 0.3516, 0.3516, 0.3516, 0.3516, 0.4216, 0.405, 0.3794, 0.4306, 0.4713, 0.4743, 0.4773, 0.4683, 0.4849, 0.4532, 0.4879, 0.5361, 0.4984, 0.5135, 0.5677, 0.6189, 0.6731, 0.5797, 0.515, 0.4939, 0.4743, 0.515, 0.5586, 0.5526, 0.5481, 0.4638, 0.515, 0.4728, 0.4532, 0.515, 0.4276, 0.4562, 0.4954, 0.5481, 0.5586, 0.5135, 0.518, 0.5571, 0.5737, 0.5827, 0.6324, 0.649, 0.6324, 0.6264, 0.6159, 0.64, 0.5782, 0.5481, 0.6249, 0.6204, 0.6008, 0.6641, 0.6791, 0.6731, 0.6731, 0.7619, 0.8083, 0.8144, 0.8144, 0.8144, 0.9445, 1.0898, 1.1019, 1.1337, 1.1004, 1.1034]}
{"unit_id": "quote:1948520271704633387", "ticker_idx": 0, "ts": "2025-07-24T23:06:49+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Isn't 14A Too Late? The 2028-2029 timeline seems too slow in my view... Intel's intention to build an AI platform based on x86... is a long-term endeavor with low customer appeal... their ability to succeed with ASICs is full of uncertainty", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish INTC: 14A timeline too slow vs TSMC, IFS lacks external orders, ASIC strategy faces structural headwinds from custom silicon adoption.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Some Light Thoughts on the Intel Call:\n\n1. Regarding IFS (Intel Foundry Services): Isn't 14A Too Late?\n\nThe 2028-2029 timeline seems too slow in my view. Especially by then, TSMC's 1.4nm process will likely be ramping up. I question whether TSMC's A14 and Intel's 14A can truly compete. I understand that even within Intel, 14A is seen as competitive with TSMC's 2nm, not A14.\n\nMoreover, Intel emphasizing that 18A can generate profit from internal products alone can only be interpreted as a sign that external customer orders are still insufficient.\n\n2. ASIC: Still Closer to Wishful Thinking\nIntel's intention to build an AI platform based on x86, aiming to capture the ecosystem like NVIDIA, is a long-term endeavor with low customer appeal.\n\nAdjusting their target to focus on inference and edge rather than training is a pragmatic shift. However, the problem is that customers are increasingly adopting their own ASICs for inference rather than using Intel's chips.\n\nI question whether Intel can beat a strong competitor like Broadcom in the inference space. Especially with the recent significant reduction in Gaudi 3 production and the cancellation of Falcon Share, their ability to succeed with ASICs is full of uncertainty.", "reply_text": null, "parent_text": null, "quoted_text": "Intel FY2Q25 Conference Call (After Hours -4.2%)\n\n▶️ FY2Q25 Key Comments\n\nComplete Review of CEO-Direct Organizations: Workforce, costs, and restructuring completed.\n\n- Workforce reduction targeting 75,000 employees by year-end, with management layers reduced by 50%.\n- Expansion", "parent_author": null, "ret_m1d": 0.038002679728001754, "ret_signed_m1d": -0.038002679728001754, "alpha_spy_m1d": -0.038333622691325786, "alpha_c_m1d": -0.04213749955279977, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08528495224927657, "ret_signed_p1d": 0.08528495224927657, "alpha_spy_p1d": 0.08950930188959572, "alpha_c_p1d": 0.08420776961954057, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": -0.1250552376781523, "ret_signed_p1w": 0.1250552376781523, "alpha_spy_p1w": 0.1213669360801567, "alpha_c_p1w": 0.12846037715931546, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": 0.09589041788636266, "ret_signed_p1m": -0.09589041788636266, "alpha_spy_p1m": -0.07872504381505974, "alpha_c_p1m": -0.07559849681990616, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.6844896308247779, "ret_signed_p3m": -0.6844896308247779, "alpha_spy_p3m": -0.6234401137628296, "alpha_c_p3m": -0.48469815346991796, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 1.0751215566055667, "ret_signed_p6m": -1.0751215566055667, "alpha_spy_p6m": -0.9786444546374895, "alpha_c_p6m": -0.6795834349027565, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.0937, -0.1012, -0.1118, -0.1171, -0.0888, -0.1043, -0.1189, -0.1025, -0.072, -0.0535, -0.0199, -0.0031, -0.049, -0.0477, -0.0429, -0.0552, -0.0601, -0.0857, -0.0919, -0.114, -0.114, -0.0919, -0.0999, -0.1052, -0.1361, -0.1277, -0.1034, -0.1052, -0.1167, -0.1136, -0.095, -0.0243, -0.0862, -0.0822, -0.11, -0.0835, -0.0809, -0.0504, -0.0504, -0.0685, -0.0636, -0.0035, -0.019, -0.0057, 0.0027, -0.0102, 0.0097, -0.0331, -0.0062, -0.0062, -0.0278, 0.0424, 0.0358, 0.0526, 0.0354, 0.0296, 0.0128, 0.0027, 0.0075, 0.0208, 0.0278, 0.027, 0.038, 0.0, -0.0853, -0.0862, -0.0981, -0.1012, -0.1251, -0.1467, -0.1383, -0.1078, -0.0981, -0.1264, -0.1184, -0.0875, -0.0362, -0.0181, 0.0544, 0.0853, 0.0455, 0.1184, 0.0402, 0.0384, 0.0959, 0.0848, 0.076, 0.0981, 0.1016, 0.076, 0.076, 0.0698, 0.0605, 0.0875, 0.0822, 0.0817, 0.08, 0.0946, 0.0875, 0.0641, 0.0946, 0.1167, 0.1003, 0.3509, 0.3071, 0.2709, 0.2965, 0.3796, 0.502, 0.5687, 0.5236, 0.4825, 0.5882, 0.6483, 0.6275, 0.6169, 0.6425, 0.654, 0.6703, 0.6072, 0.6447, 0.5745, 0.6416, 0.6279, 0.6354, 0.6836, 0.6845, 0.6315, 0.6863, 0.6916, 0.7472, 0.8352, 0.8268, 0.7746, 0.7671, 0.7455, 0.6363, 0.696, 0.6456, 0.6849, 0.6991, 0.6739, 0.6743, 0.5868, 0.5696, 0.5338, 0.517, 0.5515, 0.4856, 0.5245, 0.5815, 0.5833, 0.6266, 0.6266, 0.7923, 0.768, 0.9209, 0.9337, 0.7897, 0.8299, 0.7808, 0.7897, 0.802, 0.7459, 0.6708, 0.6575, 0.6487, 0.593, 0.6032, 0.627, 0.6072, 0.6063, 0.5979, 0.5979, 0.5996, 0.6209, 0.6483, 0.6306, 0.6306, 0.7402, 0.7397, 0.7693, 0.8838, 0.8166, 1.0128, 0.947, 1.0897, 1.1529, 1.1352, 1.0751]}
{"unit_id": "orig:1956932544076878155", "ticker_idx": 1, "ts": "2025-08-17T04:14:11+00:00", "symbol": "Eoptolink", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Currently, Innolight and Eoptolink are trading at around 15x FY26 PE. Citi expects that as the ASIC super-cycle and SiPh dividends materialize, the industry's valuation could rise to 20x", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi bullish on optical module super-cycle driven by ASIC/SiPh; top pick Innolight, sector re-rating from 15x to 20x FY26 PE expected.", "resolved_tickers": ["300502.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Eoptolink Technology 300502.SZ China", "bench_c": "MCHI", "bench_c_reason": "country_fallback (suffix='SZ')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Module Industry: ASIC & SiPh Drive a Super-CycleCiti (L. Tsang, 08/14/25)\n\nIndustry Demand and Growth Expectations\n\nCiti raises its 2026 demand forecast for optical modules, expecting demand for 800G modules to reach 45 million units and 1.6T to reach 8 million units. The main drivers are strengthening demand for ASICs, cloud infrastructure upgrades, and new AI infrastructure customers such as Oracle and Coreweave. The analyst believes the market underestimates the profitability improvement brought by the migration to Silicon Photonics (SiPh), projecting that gross margins for 800G/1.6T SiPh optical modules could reach over 45–50%.\n\nASIC & SiPh Drive an Industry Super-Cycle\n\nCiti believes the ASIC up-cycle starting in 2026 will further solidify the growth logic for optical module manufacturers. As the AI accelerator attach rate increases, demand for optical modules could be amplified by more than 20 times if combined with LPO technology. Simultaneously, the replacement of front-end networks is creating upgrade demand for 800G optical modules. The large-scale adoption of SiPh will also drive gross margin improvement and upward EPS revisions, thereby creating a \"super-cycle\" for the optical module industry.\n\nValuation Re-rating Logic\n\nCurrently, investors generally assign a forward P/E multiple in the mid-to-high teens to optical module companies due to concerns about cyclicality, a potential peak in AI capital expenditure, and the risk of substitution by CPO (Co-Packaged Optics). However, Citi argues that a 20x valuation is more reasonable for the following reasons:\n\n- ASICs will bring sustained growth for the next 2–3 years.\n\n- Inference-related AI capital expenditure is more sustainable.\n\n- Large-scale CPO adoption is not expected until 2029–2030.\n\n- Tier-1 vendors (like Innolight, Eoptolink) have a stable market share.\n\n- Increasing SiPh penetration will lead to improved profitability.\n\nValuation and Stock Selection\n\nCiti has updated its models for Innolight, Eoptolink, and TFC, expecting industry growth in 2026 to be driven by ASIC opportunities, AI investments from US cloud vendors, and SiPh penetration exceeding 50%. Among them, Innolight is Citi's top pick due to its leading market share, stronger customer stickiness, and advantages in 1.6T products. Meanwhile, technology upgrades and BOM optimization are expected to gradually materialize into gross margin improvements in the second half of 2025. Currently, Innolight and Eoptolink are trading at around 15x FY26 PE. Citi expects that as the ASIC super-cycle and SiPh dividends materialize, the industry's valuation could rise to 20x. Any pullback in the stock price would present an opportunity to increase positions.\n\n*Note: Citi's forecasts have always been quite bullish and high; the actual expectations are likely not this high.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0453968506393182, "ret_signed_m1d": -0.0453968506393182, "alpha_spy_m1d": -0.04561452945026179, "alpha_c_m1d": -0.03940080379975641, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.005996046839561786, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05116875681497124, "ret_signed_p1d": 0.05116875681497124, "alpha_spy_p1d": 0.0565938868816489, "alpha_c_p1d": 0.05866379928629972, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.0074950424713284836, "ret_p1w": 0.2121658817614649, "ret_signed_p1w": 0.2121658817614649, "alpha_spy_p1w": 0.21345616693314606, "alpha_c_p1w": 0.18718253548202624, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": 0.024983346279438656, "ret_p1m": 0.4153260858091259, "ret_signed_p1m": 0.4153260858091259, "alpha_spy_p1m": 0.38936614138946335, "alpha_c_p1m": 0.33504634189119065, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.08027974391793524, "ret_p3m": 0.33386554938356783, "ret_signed_p3m": 0.33386554938356783, "alpha_spy_p3m": 0.28629340010049176, "alpha_c_p3m": 0.260248001127652, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.07361754825591582, "ret_p6m": 0.6091939750111635, "ret_signed_p6m": 0.6091939750111635, "alpha_spy_p6m": 0.5271332733371512, "alpha_c_p6m": 0.5640961885057514, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.045097786505412074, "price_path": [-0.66, -0.6572, -0.6624, -0.6629, -0.6735, -0.6536, -0.6432, -0.6387, -0.6387, -0.6465, -0.6211, -0.6033, -0.5918, -0.5871, -0.5923, -0.5974, -0.5873, -0.5904, -0.5632, -0.5615, -0.5406, -0.5374, -0.5497, -0.5608, -0.5632, -0.5481, -0.5219, -0.4958, -0.48, -0.4793, -0.4884, -0.4692, -0.4746, -0.4822, -0.4475, -0.4451, -0.4554, -0.4631, -0.4642, -0.357, -0.3049, -0.2488, -0.2537, -0.2703, -0.2766, -0.2869, -0.2926, -0.2919, -0.2722, -0.2084, -0.2208, -0.2255, -0.2504, -0.2448, -0.2531, -0.237, -0.2484, -0.2538, -0.2104, -0.1612, -0.0316, -0.0542, -0.0454, 0.0, 0.0512, 0.0444, 0.0555, 0.1208, 0.2122, 0.1545, 0.2621, 0.453, 0.4581, 0.5903, 0.4663, 0.5134, 0.2776, 0.4297, 0.2874, 0.2908, 0.3711, 0.5551, 0.4654, 0.434, 0.4153, 0.4231, 0.4143, 0.4327, 0.4041, 0.445, 0.4517, 0.5392, 0.531, 0.5894, 0.4973, 0.4973, 0.4973, 0.4973, 0.4973, 0.4973, 0.4973, 0.4374, 0.4172, 0.427, 0.2956, 0.3205, 0.3294, 0.2948, 0.3477, 0.4958, 0.4852, 0.4236, 0.5268, 0.6536, 0.6809, 0.6624, 0.531, 0.4094, 0.4195, 0.4247, 0.4168, 0.4602, 0.4319, 0.3754, 0.3345, 0.3472, 0.3339, 0.2714, 0.2759, 0.2971, 0.3197, 0.3416, 0.228, 0.1994, 0.2474, 0.3554, 0.3951, 0.4237, 0.4557, 0.4941, 0.5274, 0.532, 0.5453, 0.6559, 0.7157, 0.7695, 0.705, 0.752, 0.7471, 0.6669, 0.8261, 0.7418, 0.7778, 0.8952, 0.8748, 0.8564, 0.8338, 0.8228, 0.8196, 0.8257, 0.7638, 0.7638, 0.7638, 0.7464, 0.7092, 0.7205, 0.6797, 0.6301, 0.602, 0.5686, 0.5718, 0.6374, 0.6259, 0.6504, 0.566, 0.6161, 0.6771, 0.5679, 0.5909, 0.6343, 0.6089, 0.6087, 0.7172, 0.797, 0.7565, 0.6341, 0.5678, 0.4903, 0.5971, 0.6092]}
{"unit_id": "orig:1956932544076878155", "ticker_idx": 2, "ts": "2025-08-17T04:14:11+00:00", "symbol": "optical modules", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Any pullback in the stock price would present an opportunity to increase positions", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi bullish on optical module super-cycle driven by ASIC/SiPh; top pick Innolight, sector re-rating from 15x to 20x FY26 PE expected.", "resolved_tickers": ["COHR", "LITE", "CIEN", "AAOI"], "resolution_method": "basket", "resolution_reason": "Optical modules basket: Coherent, Lumentum, Ciena, AAOI", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Optical Module Industry: ASIC & SiPh Drive a Super-CycleCiti (L. Tsang, 08/14/25)\n\nIndustry Demand and Growth Expectations\n\nCiti raises its 2026 demand forecast for optical modules, expecting demand for 800G modules to reach 45 million units and 1.6T to reach 8 million units. The main drivers are strengthening demand for ASICs, cloud infrastructure upgrades, and new AI infrastructure customers such as Oracle and Coreweave. The analyst believes the market underestimates the profitability improvement brought by the migration to Silicon Photonics (SiPh), projecting that gross margins for 800G/1.6T SiPh optical modules could reach over 45–50%.\n\nASIC & SiPh Drive an Industry Super-Cycle\n\nCiti believes the ASIC up-cycle starting in 2026 will further solidify the growth logic for optical module manufacturers. As the AI accelerator attach rate increases, demand for optical modules could be amplified by more than 20 times if combined with LPO technology. Simultaneously, the replacement of front-end networks is creating upgrade demand for 800G optical modules. The large-scale adoption of SiPh will also drive gross margin improvement and upward EPS revisions, thereby creating a \"super-cycle\" for the optical module industry.\n\nValuation Re-rating Logic\n\nCurrently, investors generally assign a forward P/E multiple in the mid-to-high teens to optical module companies due to concerns about cyclicality, a potential peak in AI capital expenditure, and the risk of substitution by CPO (Co-Packaged Optics). However, Citi argues that a 20x valuation is more reasonable for the following reasons:\n\n- ASICs will bring sustained growth for the next 2–3 years.\n\n- Inference-related AI capital expenditure is more sustainable.\n\n- Large-scale CPO adoption is not expected until 2029–2030.\n\n- Tier-1 vendors (like Innolight, Eoptolink) have a stable market share.\n\n- Increasing SiPh penetration will lead to improved profitability.\n\nValuation and Stock Selection\n\nCiti has updated its models for Innolight, Eoptolink, and TFC, expecting industry growth in 2026 to be driven by ASIC opportunities, AI investments from US cloud vendors, and SiPh penetration exceeding 50%. Among them, Innolight is Citi's top pick due to its leading market share, stronger customer stickiness, and advantages in 1.6T products. Meanwhile, technology upgrades and BOM optimization are expected to gradually materialize into gross margin improvements in the second half of 2025. Currently, Innolight and Eoptolink are trading at around 15x FY26 PE. Citi expects that as the ASIC super-cycle and SiPh dividends materialize, the industry's valuation could rise to 20x. Any pullback in the stock price would present an opportunity to increase positions.\n\n*Note: Citi's forecasts have always been quite bullish and high; the actual expectations are likely not this high.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03742058730931669, "ret_signed_m1d": -0.03742058730931669, "alpha_spy_m1d": -0.03763826612026028, "alpha_c_m1d": -0.03763826612026028, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": 0.00021767881094358899, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0504259356272056, "ret_signed_p1d": -0.0504259356272056, "alpha_spy_p1d": -0.045000805560527934, "alpha_c_p1d": -0.045000805560527934, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.005425130066677664, "ret_p1w": -0.011005143410225737, "ret_signed_p1w": -0.011005143410225737, "alpha_spy_p1w": -0.009714858238544566, "alpha_c_p1w": -0.009714858238544566, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.0012902851716811714, "ret_p1m": 0.29837755899909074, "ret_signed_p1m": 0.29837755899909074, "alpha_spy_p1m": 0.2724176145794282, "alpha_c_p1m": 0.2724176145794282, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.02595994441966254, "ret_p3m": 0.5915807830989679, "ret_signed_p3m": 0.5915807830989679, "alpha_spy_p3m": 0.5440086338158918, "alpha_c_p3m": 0.5440086338158918, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.04757214928307607, "ret_p6m": 2.086201992663422, "ret_signed_p6m": 2.086201992663422, "alpha_spy_p6m": 2.00414129098941, "alpha_c_p6m": 2.00414129098941, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.08206070167401225, "price_path": [-0.24, -0.2363, -0.2348, -0.2348, -0.2095, -0.2264, -0.2381, -0.2739, -0.2577, -0.2277, -0.214, -0.2542, -0.2498, -0.2447, -0.2469, -0.2492, -0.2332, -0.2687, -0.2325, -0.2379, -0.1969, -0.1969, -0.1602, -0.1688, -0.1448, -0.1036, -0.0653, -0.0857, -0.0868, -0.108, -0.0926, -0.0585, -0.0585, -0.0945, -0.0873, -0.075, -0.0628, -0.0637, -0.0414, -0.024, -0.0143, 0.0048, -0.005, -0.0033, -0.0411, -0.0259, -0.0268, -0.0032, 0.0009, 0.0039, 0.0005, -0.0001, -0.0448, -0.0073, -0.0273, 0.0059, 0.0204, 0.0283, 0.0091, 0.0537, 0.0453, -0.0563, -0.0374, 0.0, -0.0504, -0.0719, -0.0497, -0.0195, -0.011, 0.0035, -0.0027, 0.0579, 0.0161, 0.0161, -0.001, 0.0043, 0.1034, 0.1312, 0.1362, 0.1522, 0.2507, 0.2496, 0.2431, 0.3008, 0.2984, 0.2817, 0.3121, 0.3096, 0.3337, 0.2943, 0.247, 0.2538, 0.2659, 0.28, 0.2837, 0.3614, 0.3511, 0.3371, 0.3946, 0.3557, 0.4075, 0.4283, 0.3051, 0.382, 0.3523, 0.4081, 0.4257, 0.4379, 0.4678, 0.4572, 0.3849, 0.4595, 0.5464, 0.6327, 0.6383, 0.6858, 0.6366, 0.6448, 0.6157, 0.5309, 0.695, 0.77, 0.7466, 0.8175, 0.7584, 0.765, 0.5916, 0.5994, 0.6213, 0.61, 0.673, 0.5311, 0.6041, 0.7929, 0.7744, 0.8625, 0.8625, 0.9486, 0.92, 0.8914, 0.8802, 0.9622, 1.0015, 1.0787, 1.1632, 1.2525, 1.3368, 1.075, 1.0702, 0.9961, 0.9579, 1.0467, 1.2199, 1.3647, 1.3864, 1.4063, 1.4063, 1.3636, 1.3067, 1.2889, 1.2539, 1.2539, 1.3962, 1.2278, 1.432, 1.4241, 1.1384, 1.183, 1.1854, 1.2814, 1.1835, 1.2796, 1.2255, 1.2255, 1.3158, 1.316, 1.2955, 1.2217, 1.2103, 1.4009, 1.5524, 1.4659, 1.5116, 1.6569, 1.7406, 1.6323, 1.6907, 1.9505, 2.1287, 2.0862]}
{"unit_id": "thread:1958118406064071082", "ticker_idx": 0, "ts": "2025-08-20T10:46:22+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BESI will maintain a long-term technological lead in hybrid bonding, making it difficult for competitors to catch up", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares thesis: BESI leads hybrid bonding with durable tech moat; Hanmi faces structural long-term threat from hybrid bonder transition starting 2027.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Next-Generation HBM ‘Hybrid Bonder’ Technology Led by Overseas Firms, Korean Equipment Makers Could Be Squeezed Out\n\nExperts say that equipment for “hybrid bonding” used in next-generation high bandwidth memory (HBM) has already been fully developed by overseas companies, with only the hurdle of price remaining.\n\nAs a result, concerns are rising that domestic HBM bonding equipment developers such as Hanmi Semiconductor and Hanwha Semitek may fall behind in technology development and lose the market.\n\nSamsung Electronics, which has been accelerating its catch-up in the HBM market, is reportedly preparing to build a test production line for hybrid bonding equipment.\n\nOn the 20th, the Institute of Industrial Education and Research held the “Seminar on Hybrid Bonding and Packaging Material Trends Leading HBM-AI Semiconductor Innovation” at the Lifelong Education Center in Gasan-dong, Geumcheon-gu, Seoul.\n\nAt the seminar, Joo Seung-hwan, professor at Inha University Graduate School of Manufacturing Innovation, who is researching hybrid bonding equipment for HBM with LG Electronics, said, “HBM hybrid bonding equipment can already be considered technically complete,” and added, “The delay in adoption is due to the fact that costs are three times higher.”\n\nUnlike the existing TC bonder that uses thermo-compression, hybrid bonding connects chips directly without bumps. It can reduce semiconductor thickness and dissipates heat about 100 times faster than TC bonders. It is already being used in the production of logic semiconductors, image sensors (CIS), and NAND flash.\n\nProfessor Joo said, “Korea is late in developing hybrid bonders, and from a national perspective, the situation is not favorable for competitiveness in hybrid bonding equipment.”\n\nAccording to U.S. brokerage Bernstein, Dutch semiconductor equipment manufacturer BESI, in partnership with U.S.-based Applied Materials, is leading in hybrid bonding technology. Bernstein predicts that Samsung Electronics, SK Hynix, and Micron are all highly likely to adopt BESI’s hybrid bonder for future next-generation HBM production.\n\nThe full-scale adoption of hybrid bonders is expected to begin with the production of 16-layer and 20-layer HBM4 and HBM5 products. As the number of stacked DRAM layers increases, the need to reduce thickness grows, and heat dissipation becomes more critical.\n\nDavid Dai, a Bernstein researcher, forecasted that after sample testing in 2026, hybrid bonders will be applied in earnest starting with the 7th-generation HBM4E in 2027, with Samsung Electronics and Micron expected to adopt the technology first.\n\nSamsung Electronics, which has been accelerating its catch-up in HBM, is moving quickly. A semiconductor equipment industry insider said, “Samsung is preparing to set up a hybrid bonding test line in some HBM production plants.”\n\nAs overheating issues were cited as one of the reasons Samsung had difficulty obtaining Nvidia’s certification for its HBM3E 12-layer product, the adoption of hybrid bonding, which can increase heat dissipation capacity by up to 100 times, is expected to serve as a stepping stone for catching up.\n\nSK Hynix, which currently leads with HBM manufactured using TC bonders, is expected to be relatively slow in adopting hybrid bonding. This is because numerous TC bonders are already deployed in its production facilities, making it less necessary to replace them with expensive hybrid bonders.\n\nKorean semiconductor equipment makers such as Hanmi Semiconductor and Hanwha Semitek are increasingly concerned about falling behind in the hybrid bonding equipment race. Although they hold an overwhelming market share in TC bonders, they could be completely pushed out in the future hybrid bonder market.\n\nU.S. brokerage Goldman Sachs stated, “Hanmi Semiconductor’s long-term issue is the transition from TC bonders to hybrid bonders,” adding, “From 2027 onwards, six competitors including BESI will enter the hybrid bonding equipment market, which will pose a major challenge for Hanmi Semiconductor.”\n\nAn industry insider said, “Hanwha Semitek is preparing to launch its second-generation hybrid bonder in the second half of this year, but it is likely to contain significant overseas technology,” adding, “Korean semiconductor manufacturing lacks infrastructure and basic technology, making it highly dependent on foreign solutions.”\n\nDutch company BESI has expressed strong confidence in the hybrid bonder market.\n\nBesi CEO Richard Blickman said, “Constructive exchanges of opinions have taken place with major memory semiconductor companies,” adding, “Although the adoption of hybrid bonders has been delayed as HBM4 standards were eased, considering the heat limitations in data centers, the low power consumption and heat dissipation of hybrid bonding equipment will be a major strength.”\n\nGoldman Sachs commented, “The technological complexity and ecosystem integration underpin BESI’s competitiveness in hybrid bonding,” predicting, “BESI will maintain a long-term technological lead in hybrid bonding, making it difficult for competitors to catch up.”\n\nIt added, “BESI has enjoyed multiple advantages, such as fine-tuning its process range with continuous feedback from the three major memory companies. Although alternative solutions are emerging in Korea, BESI has the strength of providing stable performance in large-scale environments.”\n\n---\n\nhttps://t.co/K2INI02l8z", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01964133454625916, "ret_signed_m1d": 0.01964133454625916, "alpha_spy_m1d": 0.0169771720129559, "alpha_c_m1d": 0.013031395430204595, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006609939116054564, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014090592448717021, "ret_signed_p1d": -0.014090592448717021, "alpha_spy_p1d": -0.010078770161158435, "alpha_c_p1d": -0.008899513053806207, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.005191079394910814, "ret_p1w": 0.023057196010457304, "ret_signed_p1w": 0.023057196010457304, "alpha_spy_p1w": 0.00970522636873361, "alpha_c_p1w": -0.007050746658838092, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.030107942669295396, "ret_p1m": 0.06575579213073368, "ret_signed_p1m": 0.06575579213073368, "alpha_spy_p1m": 0.027909616914708835, "alpha_c_p1m": -0.03117808093360286, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.09693387306433654, "ret_p3m": 0.11187011858808882, "ret_signed_p3m": 0.11187011858808882, "alpha_spy_p3m": 0.06578789118268347, "alpha_c_p3m": -0.06580145982846242, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.17767157841655123, "ret_p6m": 0.44064894811463784, "ret_signed_p6m": 0.44064894811463784, "alpha_spy_p6m": 0.36688813722300817, "alpha_c_p6m": 0.03085645356791966, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.4097924945467182, "price_path": [-0.1008, -0.082, -0.0653, -0.07, -0.0564, -0.0901, -0.1123, -0.0884, -0.0781, -0.0555, -0.0299, -0.0192, 0.0273, 0.0115, 0.0482, 0.0751, 0.0845, 0.0833, 0.0739, 0.0649, 0.0517, 0.0734, 0.1008, 0.1033, 0.1085, 0.114, 0.085, 0.0431, 0.0525, 0.0414, 0.0337, 0.0367, 0.0436, 0.0414, 0.088, 0.0858, 0.0645, 0.0786, 0.0243, 0.0888, 0.0927, 0.1114, 0.0743, 0.0559, 0.0845, 0.0, 0.0487, 0.0248, 0.0453, 0.0171, 0.0004, 0.0111, -0.0154, -0.0196, 0.0243, 0.0312, 0.038, 0.0568, 0.0547, 0.0474, 0.0124, 0.006, 0.0196, 0.0, -0.0141, 0.0162, 0.0175, 0.0307, 0.0231, 0.0226, -0.0171, -0.0137, -0.0871, -0.0965, -0.085, -0.0717, -0.0534, -0.0482, -0.0406, -0.0295, -0.0448, 0.009, -0.0043, -0.0124, 0.0658, 0.0376, 0.0487, 0.0837, 0.0978, 0.082, 0.0453, 0.0794, 0.0833, 0.0914, 0.1405, 0.1144, 0.2532, 0.2536, 0.2498, 0.2417, 0.1981, 0.219, 0.1998, 0.2425, 0.2391, 0.2101, 0.24, 0.2323, 0.1827, 0.2319, 0.2558, 0.2639, 0.2519, 0.2609, 0.2558, 0.2605, 0.2451, 0.2173, 0.2084, 0.1763, 0.1456, 0.1652, 0.1768, 0.1558, 0.1426, 0.1221, 0.1119, 0.0858, 0.1025, 0.1106, 0.0487, 0.0623, 0.064, 0.1067, 0.1093, 0.1093, 0.108, 0.1217, 0.161, 0.1742, 0.1917, 0.2541, 0.2237, 0.1994, 0.1691, 0.1281, 0.1311, 0.1409, 0.1042, 0.1225, 0.1166, 0.1204, 0.1264, 0.1234, 0.1234, 0.1234, 0.1281, 0.1366, 0.1422, 0.1422, 0.2733, 0.313, 0.374, 0.3561, 0.2904, 0.2904, 0.3847, 0.4278, 0.3792, 0.4829, 0.4765, 0.4415, 0.4765, 0.4863, 0.4944, 0.4949, 0.4932, 0.5009, 0.4471, 0.3881, 0.4048, 0.4056, 0.3676, 0.3454, 0.3728, 0.4133, 0.4295, 0.4436, 0.4594, 0.4406]}
{"unit_id": "orig:1955041158645874925", "ticker_idx": 0, "ts": "2025-08-11T22:58:29+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price raised to $185 (from $165), applying 2H CY26 annualized EPS of $17 × P/E 10–12x", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM Buy on MU with PT raised to $185; raised guidance, HBM ramp, DRAM pricing strength through 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron (JPM)\n\n1. Results and Guidance\n\n• Raises August-quarter revenue outlook to $11.2bn (vs. prior guidance midpoint of $10.7bn), non-GAAP gross margin to 44.5% (vs. 42% prior), and expects non-GAAP EPS of $2.85 (vs. $2.50 prior).\n\n• The main driver is an improved pricing environment, with strong demand across all DRAM product lines for AI/datacenter, smartphones, and PCs.\n\n• The 12-high HBM3E yield ramp is progressing faster than the 8-high, contributing to ASP/mix improvement.\n\n• DDR5 and LPDDR5 price uptrends continue; DDR4 is seeing some tightness due to EOL.\n\n2. Demand and Market Environment\n\n• DRAM price strength is expected to continue through 2H25–2026.\n\n• HBM demand has strong visibility through CY26, and the HBM4 ramp is expected to constrain non-HBM DRAM supply, maintaining upward price pressure.\n\n• With expanding AI infrastructure investment, HBM revenue is expected to reach about $8bn annually (assuming annualized from the quarterly run rate).\n\n• End demand across smartphones, PCs, and datacenters remains stable.\n\n3. Competitiveness and Positioning\n\n• Confident of fully absorbing/selling its entire HBM supply in FY26/CY26, with tighter roadmap alignment with key customers.\n\n• Increasing the share of high-margin products in the portfolio, creating opportunities to gain market share.\n\n• Continuing profitability improvement and cost-reduction efforts in the NAND business.\n\n• Expecting to mitigate earnings volatility through product and market diversification.\n\n4. Financial Outlook and Valuation\n\n• FY25E revenue $37.28bn, EPS $8.11 → FY26E revenue $51.47bn, EPS $14.57.\n\n• FY26E gross margin above 49%, with EBITDA margin expected at 54.7%.\n\n• Target price raised to $185 (from $165), applying 2H CY26 annualized EPS of $17 × P/E 10–12x.\n\n• EV/EBITDA multiple expected to decline from 7.5× in FY25E to 4.5× in FY26E.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03903979884308595, "ret_signed_m1d": -0.03903979884308595, "alpha_spy_m1d": -0.04102116010893475, "alpha_c_m1d": -0.03948291899323131, "bench_spy_m1d": 0.0019813612658488022, "bench_c_m1d": 0.00044312015014535966, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03257356448559756, "ret_signed_p1d": 0.03257356448559756, "alpha_spy_p1d": 0.021927450029897377, "alpha_c_p1d": 0.009737800025607424, "bench_spy_p1d": 0.010646114455700184, "bench_c_p1d": 0.022835764459990138, "ret_p1w": -0.00137401150096228, "ret_signed_p1w": -0.00137401150096228, "alpha_spy_p1w": -0.012979251935328562, "alpha_c_p1w": -0.013166829511401912, "bench_spy_p1w": 0.011605240434366282, "bench_c_p1w": 0.011792818010439632, "ret_p1m": 0.09311362653318778, "ret_signed_p1m": 0.09311362653318778, "alpha_spy_p1m": 0.07045351303903358, "alpha_c_p1m": 0.07808291174397186, "bench_spy_p1m": 0.0226601134941542, "bench_c_p1m": 0.015030714789215915, "ret_p3m": 0.9275724442299176, "ret_signed_p3m": 0.9275724442299176, "alpha_spy_p3m": 0.8705708971807182, "alpha_c_p3m": 0.7309465636759955, "bench_spy_p3m": 0.05700154704919935, "bench_c_p3m": 0.19662588055392205, "ret_p6m": 2.3937297301036047, "ret_signed_p6m": 2.3937297301036047, "alpha_spy_p6m": 2.3032075576837467, "alpha_c_p6m": 2.034092915426948, "bench_spy_p6m": 0.09052217241985772, "bench_c_p6m": 0.35963681467665687, "price_path": [-0.2303, -0.2292, -0.2086, -0.2034, -0.2078, -0.2261, -0.2342, -0.246, -0.246, -0.2217, -0.2233, -0.2183, -0.2372, -0.2072, -0.1743, -0.1662, -0.1417, -0.1234, -0.1041, -0.0783, -0.063, -0.0618, -0.0665, -0.0323, -0.0282, -0.0163, -0.0163, -0.0019, -0.0142, 0.0329, 0.0276, 0.0175, 0.0075, -0.0047, -0.0238, -0.0169, -0.0125, -0.0125, -0.0307, 0.0057, -0.012, -0.0049, 0.0065, -0.0413, -0.0292, -0.0589, -0.0845, -0.0754, -0.0848, -0.1172, -0.1123, -0.0969, -0.1007, -0.1008, -0.0951, -0.0726, -0.1178, -0.1523, -0.1289, -0.1185, -0.1208, -0.0958, -0.039, 0.0, 0.0326, 0.0044, 0.0127, -0.023, -0.0014, -0.0135, -0.0526, -0.0641, -0.0488, -0.059, -0.0584, -0.0483, -0.0139, -0.0381, -0.0381, -0.0424, -0.0404, 0.004, 0.0618, 0.0626, 0.0931, 0.1316, 0.217, 0.2709, 0.2752, 0.2837, 0.2932, 0.3651, 0.3153, 0.3306, 0.3451, 0.3071, 0.2676, 0.2712, 0.3248, 0.3524, 0.4723, 0.4852, 0.5191, 0.5445, 0.5018, 0.5896, 0.5555, 0.4687, 0.5591, 0.5129, 0.5524, 0.638, 0.6368, 0.6723, 0.6361, 0.6052, 0.6718, 0.7714, 0.7801, 0.7948, 0.8329, 0.8118, 0.8098, 0.8982, 0.7634, 0.9209, 0.9276, 0.9243, 1.0486, 0.9501, 0.9807, 0.9164, 0.9963, 0.9569, 0.8481, 0.8272, 0.6286, 0.6772, 0.8111, 0.816, 0.8623, 0.8623, 0.9126, 0.9448, 0.937, 0.8938, 0.8331, 0.9186, 0.997, 1.0415, 1.1328, 1.0904, 0.9503, 0.9209, 0.8805, 0.824, 1.0102, 1.1507, 1.237, 1.2344, 1.3186, 1.3186, 1.3033, 1.3818, 1.3677, 1.3093, 1.3093, 1.5521, 1.5256, 1.7787, 1.7473, 1.646, 1.7922, 1.7985, 1.7358, 1.6972, 1.7237, 1.935, 1.935, 1.9533, 2.1483, 2.2169, 2.2336, 2.1482, 2.3193, 2.5219, 2.526, 2.3568, 2.5423, 2.3937]}
{"unit_id": "orig:1945268751550800199", "ticker_idx": 0, "ts": "2025-07-15T23:46:26+00:00", "symbol": "8035.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is time to buy into beneficiary companies such as Tokyo Electron … we maintain our Overweight rating on our covered companies", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA channel checks point to imminent Samsung 1-c DRAM capex; buy TEL/Eugene Technology/EO Technics as beneficiaries; cautious on Kokusai Electric/Wonik IPS/HPSP on share-loss risk.", "resolved_tickers": ["8035.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "250716_Semicap equipment: SEC's Capex for 1c DRAM coming soon - CLSA\n\n(1) Our channel checks indicate that Samsung Electronics is soon expected to place large-scale capital expenditure orders for 1-c DRAM mass production.\n\n(2) This is because, even though they haven't completely resolved all complex technical issues of 1-c DRAM, they have made progress in yield improvement sufficient to obtain internal Production Readiness Approval.\n\n(3) Samsung Electronics is believed to be ready to execute investments to secure sufficient capacity for HBM4 and LPDDR5 demand in 2026.\n\n(4) We believe Samsung Electronics has achieved a partial breakthrough, sacrificing some of the cost structure to begin 1-c DRAM mass production by the end of 2025.\n\n(5) A year ago, 1-c yields were significantly low, but they have now reached approximately 50% at the lab level. This progress is mainly attributed to 1) core circuit redesign and 2) increasing die size to widen capacitor pitch.\n\n(6) Further improvements in chip design and process will still be necessary. Key challenges include 1) addressing excessive EUV usage, 2) aggressive clocking, and 3) resolving low power efficiency issues. Solving all technical problems may require more time. * Risk of overlay misalignment between multiple EUV layers, affecting transistor variability. \n* Increased process-induced parasitics, leading to leakage current and non-uniform cell characteristics. \n* Greater manufacturing process variations, necessitating guard-banding (e.g., over-supplying voltage) at the design stage.\n\n(7) Nevertheless, we anticipate a bold investment execution, primarily centered on P4.\n(8) Our channel checks suggest that P4 line will begin orders in 2H25 to build 50K (monthly) 1-c DRAM CAPA. For 1H26, we forecast 40K/10K orders for P4/P3 respectively, and 30K for P3 in 2H26.\n(9) To secure sufficient 1-c DRAM production volume for important customers like NVIDIA, Samsung is proactively starting with ample capacity to offset potential mass production yield risks that may arise in the future.\n\n(10) Accordingly, we believe that regardless of the actual progress of 1-c DRAM, it is time to buy into beneficiary companies such as Tokyo Electron, Eugene Technology, and EO Technics, and we maintain our Overweight rating on our covered companies.\n\n(11) Tokyo Electron: Expected to benefit from Samsung Electronics' DRAM investment due to its relatively high exposure to batch-type CVD equipment for titanium nitride (TiN) processes.\n\n(12) EO Technics: Expected to supply laser annealing and marking equipment, and potentially femtosecond laser-based cutting equipment, for the P4/P3 lines dedicated to 1-c DRAM.\n\n(13) Eugene Technology: Expected to benefit from Samsung Electronics' 1-c DRAM CAPEX execution due to its dominant market share in the LPCVD market, capable of controlling thickness and uniformity.\n\n(14) However, we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.033746747241298114, "ret_signed_m1d": -0.033746747241298114, "alpha_spy_m1d": -0.03803825261126981, "alpha_c_m1d": -0.014902246297236132, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01751188449661889, "ret_signed_p1d": 0.01751188449661889, "alpha_spy_p1d": 0.01416869337810378, "alpha_c_p1d": 0.022738859405286527, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.002736236492892452, "ret_signed_p1w": 0.002736236492892452, "alpha_spy_p1w": -0.008065146927922928, "alpha_c_p1w": 0.021408831408600615, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.20868296335007863, "ret_signed_p1m": -0.20868296335007863, "alpha_spy_p1m": -0.2452502972731443, "alpha_c_p1m": -0.24472154104191857, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.0789539274757125, "ret_signed_p3m": 0.0789539274757125, "alpha_spy_p3m": 0.026408809335148442, "alpha_c_p3m": -0.03982187421905148, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.4065461476026184, "ret_signed_p6m": 0.4065461476026184, "alpha_spy_p6m": 0.2917888443214314, "alpha_c_p6m": 0.07842364776683963, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2773, -0.288, -0.2911, -0.3003, -0.2831, -0.2563, -0.2247, -0.2266, -0.2266, -0.2256, -0.2262, -0.2287, -0.2287, -0.2287, -0.2008, -0.1844, -0.1771, -0.1649, -0.1244, -0.1162, -0.1197, -0.1439, -0.1585, -0.1483, -0.152, -0.1726, -0.1698, -0.1479, -0.1538, -0.1549, -0.1189, -0.1609, -0.1753, -0.1828, -0.1791, -0.1481, -0.1372, -0.1297, -0.1339, -0.0923, -0.0879, -0.1317, -0.1239, -0.0981, -0.0976, -0.1244, -0.125, -0.1352, -0.1036, -0.0744, -0.0376, 0.0038, 0.0099, -0.0124, -0.0301, -0.0131, -0.0086, -0.0104, -0.0053, -0.0086, -0.0237, -0.015, -0.0337, 0.0, 0.0175, 0.0131, 0.0161, 0.0161, 0.0027, 0.0223, 0.0223, 0.0201, -0.0029, -0.015, -0.0053, -0.0029, -0.1826, -0.1797, -0.183, -0.2138, -0.2331, -0.2233, -0.2233, -0.2149, -0.2087, -0.2246, -0.2154, -0.2322, -0.2258, -0.2366, -0.255, -0.2669, -0.2621, -0.2614, -0.2581, -0.243, -0.2461, -0.2599, -0.2581, -0.2723, -0.2672, -0.2616, -0.2484, -0.2331, -0.2284, -0.2191, -0.176, -0.176, -0.1603, -0.1153, -0.0713, -0.0657, -0.0296, -0.0296, -0.0057, 0.0113, -0.0316, -0.0233, -0.0286, -0.0489, 0.0261, 0.0498, 0.1265, 0.1088, 0.078, 0.0944, 0.079, 0.079, 0.048, 0.0712, 0.1147, 0.1084, 0.1574, 0.1316, 0.1324, 0.0955, 0.1154, 0.1283, 0.1589, 0.1972, 0.2157, 0.2595, 0.2595, 0.282, 0.2297, 0.2252, 0.2087, 0.2603, 0.2599, 0.2404, 0.2496, 0.1737, 0.2271, 0.16, 0.1372, 0.1976, 0.1121, 0.1121, 0.146, 0.1486, 0.1858, 0.1718, 0.1656, 0.1534, 0.2079, 0.2463, 0.2212, 0.2219, 0.2378, 0.2205, 0.2013, 0.1608, 0.1475, 0.1412, 0.1541, 0.1169, 0.1497, 0.2223, 0.216, 0.2241, 0.2451, 0.265, 0.2614, 0.2647, 0.2647, 0.2647, 0.2647, 0.3609, 0.3763, 0.4065]}
{"unit_id": "orig:1945268751550800199", "ticker_idx": 1, "ts": "2025-07-15T23:46:26+00:00", "symbol": "EO Technics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EO Technics: Expected to supply laser annealing and marking equipment … for the P4/P3 lines dedicated to 1-c DRAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA channel checks point to imminent Samsung 1-c DRAM capex; buy TEL/Eugene Technology/EO Technics as beneficiaries; cautious on Kokusai Electric/Wonik IPS/HPSP on share-loss risk.", "resolved_tickers": ["039030.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "EO Technics 039030.KQ Korea KOSDAQ laser equipment", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "250716_Semicap equipment: SEC's Capex for 1c DRAM coming soon - CLSA\n\n(1) Our channel checks indicate that Samsung Electronics is soon expected to place large-scale capital expenditure orders for 1-c DRAM mass production.\n\n(2) This is because, even though they haven't completely resolved all complex technical issues of 1-c DRAM, they have made progress in yield improvement sufficient to obtain internal Production Readiness Approval.\n\n(3) Samsung Electronics is believed to be ready to execute investments to secure sufficient capacity for HBM4 and LPDDR5 demand in 2026.\n\n(4) We believe Samsung Electronics has achieved a partial breakthrough, sacrificing some of the cost structure to begin 1-c DRAM mass production by the end of 2025.\n\n(5) A year ago, 1-c yields were significantly low, but they have now reached approximately 50% at the lab level. This progress is mainly attributed to 1) core circuit redesign and 2) increasing die size to widen capacitor pitch.\n\n(6) Further improvements in chip design and process will still be necessary. Key challenges include 1) addressing excessive EUV usage, 2) aggressive clocking, and 3) resolving low power efficiency issues. Solving all technical problems may require more time. * Risk of overlay misalignment between multiple EUV layers, affecting transistor variability. \n* Increased process-induced parasitics, leading to leakage current and non-uniform cell characteristics. \n* Greater manufacturing process variations, necessitating guard-banding (e.g., over-supplying voltage) at the design stage.\n\n(7) Nevertheless, we anticipate a bold investment execution, primarily centered on P4.\n(8) Our channel checks suggest that P4 line will begin orders in 2H25 to build 50K (monthly) 1-c DRAM CAPA. For 1H26, we forecast 40K/10K orders for P4/P3 respectively, and 30K for P3 in 2H26.\n(9) To secure sufficient 1-c DRAM production volume for important customers like NVIDIA, Samsung is proactively starting with ample capacity to offset potential mass production yield risks that may arise in the future.\n\n(10) Accordingly, we believe that regardless of the actual progress of 1-c DRAM, it is time to buy into beneficiary companies such as Tokyo Electron, Eugene Technology, and EO Technics, and we maintain our Overweight rating on our covered companies.\n\n(11) Tokyo Electron: Expected to benefit from Samsung Electronics' DRAM investment due to its relatively high exposure to batch-type CVD equipment for titanium nitride (TiN) processes.\n\n(12) EO Technics: Expected to supply laser annealing and marking equipment, and potentially femtosecond laser-based cutting equipment, for the P4/P3 lines dedicated to 1-c DRAM.\n\n(13) Eugene Technology: Expected to benefit from Samsung Electronics' 1-c DRAM CAPEX execution due to its dominant market share in the LPCVD market, capable of controlling thickness and uniformity.\n\n(14) However, we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.004291505369971693, "alpha_c_m1d": 0.01884450094406198, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03051643209788013, "ret_signed_p1d": 0.03051643209788013, "alpha_spy_p1d": 0.02717324097936502, "alpha_c_p1d": 0.03574340700654777, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.004694864052346093, "ret_signed_p1w": 0.004694864052346093, "alpha_spy_p1w": -0.0061065193684692876, "alpha_c_p1w": 0.023367458968054255, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.12206572839152074, "ret_signed_p1m": 0.12206572839152074, "alpha_spy_p1m": 0.08549839446845509, "alpha_c_p1m": 0.08602715069968081, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.20657276545511594, "ret_signed_p3m": 0.20657276545511594, "alpha_spy_p3m": 0.15402764731455187, "alpha_c_p3m": 0.08779696376035195, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.4503559304633924, "ret_signed_p6m": 0.4503559304633924, "alpha_spy_p6m": 0.33559862718220534, "alpha_c_p6m": 0.1222334306276136, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.461, -0.4563, -0.4563, -0.4601, -0.4338, -0.4385, -0.4338, -0.4526, -0.4488, -0.4587, -0.4587, -0.4437, -0.4437, -0.4437, -0.4516, -0.4408, -0.4512, -0.4047, -0.3869, -0.369, -0.3662, -0.3521, -0.3765, -0.3826, -0.3836, -0.3991, -0.3714, -0.3338, -0.3272, -0.3089, -0.3169, -0.3244, -0.3192, -0.3192, -0.3061, -0.293, -0.293, -0.3085, -0.3089, -0.3047, -0.3033, -0.3174, -0.3089, -0.2803, -0.2671, -0.2789, -0.2498, -0.2831, -0.2596, -0.207, -0.1953, -0.2066, -0.2103, -0.1836, -0.1817, -0.1131, -0.1455, -0.131, -0.1174, -0.131, -0.0638, 0.0047, 0.0, 0.0, 0.0305, 0.054, 0.0681, 0.0446, 0.0047, -0.007, -0.007, 0.0047, 0.0094, -0.0094, 0.0023, 0.0258, -0.0446, -0.0141, 0.0047, 0.0, 0.0892, 0.1033, 0.0798, 0.0798, 0.1221, 0.1033, 0.1033, 0.0469, -0.0047, -0.0047, -0.0094, -0.0188, -0.007, -0.0188, -0.0023, -0.0188, -0.0305, -0.077, -0.0869, -0.0756, -0.0587, -0.062, -0.0211, -0.007, -0.0047, -0.0094, 0.0188, 0.0282, 0.0681, 0.0376, 0.1784, 0.1784, 0.1549, 0.1408, 0.1244, 0.1033, 0.0845, 0.108, 0.0939, 0.0962, 0.1174, 0.1174, 0.1174, 0.1174, 0.1174, 0.1174, 0.2066, 0.2183, 0.1737, 0.2042, 0.223, 0.1667, 0.1315, 0.0751, 0.0728, 0.0869, 0.1009, 0.108, 0.0798, 0.1033, 0.0986, 0.108, 0.1432, 0.3779, 0.2817, 0.2911, 0.23, 0.3122, 0.3685, 0.3404, 0.4155, 0.3005, 0.3192, 0.2911, 0.2887, 0.2911, 0.1643, 0.1526, 0.1596, 0.2042, 0.2066, 0.2066, 0.2817, 0.2958, 0.3122, 0.3052, 0.338, 0.3028, 0.277, 0.2723, 0.2723, 0.2676, 0.1831, 0.1127, 0.1831, 0.1878, 0.1643, 0.2066, 0.2113, 0.1972, 0.1972, 0.2207, 0.2593, 0.2829, 0.2829, 0.2829, 0.3631, 0.4952, 0.5117, 0.4504]}
{"unit_id": "orig:1945268751550800199", "ticker_idx": 2, "ts": "2025-07-15T23:46:26+00:00", "symbol": "Eugene Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Eugene Technology: Expected to benefit from Samsung Electronics' 1-c DRAM CAPEX execution due to its dominant market share in the LPCVD market", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA channel checks point to imminent Samsung 1-c DRAM capex; buy TEL/Eugene Technology/EO Technics as beneficiaries; cautious on Kokusai Electric/Wonik IPS/HPSP on share-loss risk.", "resolved_tickers": ["084370.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "Eugene Technology 084370.KQ Korea KOSDAQ LPCVD", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "250716_Semicap equipment: SEC's Capex for 1c DRAM coming soon - CLSA\n\n(1) Our channel checks indicate that Samsung Electronics is soon expected to place large-scale capital expenditure orders for 1-c DRAM mass production.\n\n(2) This is because, even though they haven't completely resolved all complex technical issues of 1-c DRAM, they have made progress in yield improvement sufficient to obtain internal Production Readiness Approval.\n\n(3) Samsung Electronics is believed to be ready to execute investments to secure sufficient capacity for HBM4 and LPDDR5 demand in 2026.\n\n(4) We believe Samsung Electronics has achieved a partial breakthrough, sacrificing some of the cost structure to begin 1-c DRAM mass production by the end of 2025.\n\n(5) A year ago, 1-c yields were significantly low, but they have now reached approximately 50% at the lab level. This progress is mainly attributed to 1) core circuit redesign and 2) increasing die size to widen capacitor pitch.\n\n(6) Further improvements in chip design and process will still be necessary. Key challenges include 1) addressing excessive EUV usage, 2) aggressive clocking, and 3) resolving low power efficiency issues. Solving all technical problems may require more time. * Risk of overlay misalignment between multiple EUV layers, affecting transistor variability. \n* Increased process-induced parasitics, leading to leakage current and non-uniform cell characteristics. \n* Greater manufacturing process variations, necessitating guard-banding (e.g., over-supplying voltage) at the design stage.\n\n(7) Nevertheless, we anticipate a bold investment execution, primarily centered on P4.\n(8) Our channel checks suggest that P4 line will begin orders in 2H25 to build 50K (monthly) 1-c DRAM CAPA. For 1H26, we forecast 40K/10K orders for P4/P3 respectively, and 30K for P3 in 2H26.\n(9) To secure sufficient 1-c DRAM production volume for important customers like NVIDIA, Samsung is proactively starting with ample capacity to offset potential mass production yield risks that may arise in the future.\n\n(10) Accordingly, we believe that regardless of the actual progress of 1-c DRAM, it is time to buy into beneficiary companies such as Tokyo Electron, Eugene Technology, and EO Technics, and we maintain our Overweight rating on our covered companies.\n\n(11) Tokyo Electron: Expected to benefit from Samsung Electronics' DRAM investment due to its relatively high exposure to batch-type CVD equipment for titanium nitride (TiN) processes.\n\n(12) EO Technics: Expected to supply laser annealing and marking equipment, and potentially femtosecond laser-based cutting equipment, for the P4/P3 lines dedicated to 1-c DRAM.\n\n(13) Eugene Technology: Expected to benefit from Samsung Electronics' 1-c DRAM CAPEX execution due to its dominant market share in the LPCVD market, capable of controlling thickness and uniformity.\n\n(14) However, we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03154213692364327, "ret_signed_m1d": -0.03154213692364327, "alpha_spy_m1d": -0.035833642293614965, "alpha_c_m1d": -0.03332113999013897, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": 0.001779003066495699, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06308409100279877, "ret_signed_p1d": 0.06308409100279877, "alpha_spy_p1d": 0.05974089988428366, "alpha_c_p1d": 0.06294722818887366, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00013686281392510935, "ret_p1w": 0.03037376064577746, "ret_signed_p1w": 0.03037376064577746, "alpha_spy_p1w": 0.01957237722496208, "alpha_c_p1w": 0.03283718453916362, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.002463423893386163, "ret_p1m": 0.16355137751837434, "ret_signed_p1m": 0.16355137751837434, "alpha_spy_p1m": 0.12698404359530868, "alpha_c_p1m": 0.14097029430791852, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.02258108321045582, "ret_p3m": 0.8598129482604038, "ret_signed_p3m": 0.8598129482604038, "alpha_spy_p3m": 0.8072678301198397, "alpha_c_p3m": 0.7551188740145691, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.10469407424583466, "ret_p6m": 1.0771027778923097, "ret_signed_p6m": 1.0771027778923097, "alpha_spy_p6m": 0.9623454746111226, "alpha_c_p6m": 0.5794439647170966, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.4976588131752131, "price_path": [-0.2196, -0.2173, -0.2185, -0.2442, -0.1916, -0.1951, -0.1799, -0.2068, -0.2009, -0.2056, -0.2056, -0.2009, -0.2009, -0.2009, -0.1986, -0.2056, -0.2126, -0.16, -0.153, -0.132, -0.2208, -0.215, -0.2523, -0.2535, -0.25, -0.2629, -0.2722, -0.25, -0.2523, -0.2465, -0.2208, -0.2418, -0.2371, -0.2371, -0.2301, -0.2056, -0.2056, -0.1986, -0.1928, -0.1308, -0.1414, -0.1647, -0.1706, -0.1507, -0.1238, -0.1262, -0.0981, -0.1227, -0.0864, -0.0724, -0.0736, -0.0864, -0.0864, -0.0958, -0.1168, 0.0058, 0.0058, -0.0058, 0.0, -0.0315, -0.0537, -0.0187, -0.0315, 0.0, 0.0631, 0.0783, 0.0736, 0.0806, 0.0304, 0.0129, -0.0082, -0.0175, 0.014, 0.0269, 0.0783, 0.0456, -0.0199, -0.0245, -0.0152, -0.0269, -0.0117, 0.0269, 0.0806, 0.1215, 0.1636, 0.1869, 0.1869, 0.1542, 0.1507, 0.1507, 0.1542, 0.1215, 0.1449, 0.132, 0.1285, 0.1157, 0.1203, 0.0666, 0.0619, 0.0818, 0.0736, 0.0619, 0.0864, 0.1472, 0.1636, 0.1939, 0.278, 0.3505, 0.3551, 0.4439, 0.5023, 0.5023, 0.5818, 0.6542, 0.6262, 0.6098, 0.5584, 0.6215, 0.6332, 0.6939, 0.8178, 0.8178, 0.8178, 0.8178, 0.8178, 0.8178, 0.8598, 0.8341, 0.7944, 0.8528, 0.8598, 0.8972, 1.0561, 1.0935, 1.0841, 1.1262, 1.1472, 1.1612, 1.0794, 1.2944, 1.285, 1.2173, 1.1986, 1.3785, 1.4393, 1.4696, 1.3528, 1.3481, 1.2944, 1.278, 1.285, 1.0771, 1.1005, 0.986, 1.0304, 0.9439, 0.8061, 0.8014, 0.8037, 0.8481, 0.8575, 0.8645, 0.8762, 0.8738, 0.8855, 0.7944, 0.7991, 0.8201, 0.7757, 0.8131, 0.8014, 0.7523, 0.7009, 0.6776, 0.729, 0.6776, 0.715, 0.7897, 0.7897, 0.7266, 0.7266, 0.7664, 0.7523, 0.7523, 0.7523, 0.7523, 1.0, 1.0911, 1.1402, 1.0771]}
{"unit_id": "orig:1945268751550800199", "ticker_idx": 3, "ts": "2025-07-15T23:46:26+00:00", "symbol": "Kokusai Electric", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA channel checks point to imminent Samsung 1-c DRAM capex; buy TEL/Eugene Technology/EO Technics as beneficiaries; cautious on Kokusai Electric/Wonik IPS/HPSP on share-loss risk.", "resolved_tickers": ["6525.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kokusai Electric 6525.T Japan batch CVD equipment", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "250716_Semicap equipment: SEC's Capex for 1c DRAM coming soon - CLSA\n\n(1) Our channel checks indicate that Samsung Electronics is soon expected to place large-scale capital expenditure orders for 1-c DRAM mass production.\n\n(2) This is because, even though they haven't completely resolved all complex technical issues of 1-c DRAM, they have made progress in yield improvement sufficient to obtain internal Production Readiness Approval.\n\n(3) Samsung Electronics is believed to be ready to execute investments to secure sufficient capacity for HBM4 and LPDDR5 demand in 2026.\n\n(4) We believe Samsung Electronics has achieved a partial breakthrough, sacrificing some of the cost structure to begin 1-c DRAM mass production by the end of 2025.\n\n(5) A year ago, 1-c yields were significantly low, but they have now reached approximately 50% at the lab level. This progress is mainly attributed to 1) core circuit redesign and 2) increasing die size to widen capacitor pitch.\n\n(6) Further improvements in chip design and process will still be necessary. Key challenges include 1) addressing excessive EUV usage, 2) aggressive clocking, and 3) resolving low power efficiency issues. Solving all technical problems may require more time. * Risk of overlay misalignment between multiple EUV layers, affecting transistor variability. \n* Increased process-induced parasitics, leading to leakage current and non-uniform cell characteristics. \n* Greater manufacturing process variations, necessitating guard-banding (e.g., over-supplying voltage) at the design stage.\n\n(7) Nevertheless, we anticipate a bold investment execution, primarily centered on P4.\n(8) Our channel checks suggest that P4 line will begin orders in 2H25 to build 50K (monthly) 1-c DRAM CAPA. For 1H26, we forecast 40K/10K orders for P4/P3 respectively, and 30K for P3 in 2H26.\n(9) To secure sufficient 1-c DRAM production volume for important customers like NVIDIA, Samsung is proactively starting with ample capacity to offset potential mass production yield risks that may arise in the future.\n\n(10) Accordingly, we believe that regardless of the actual progress of 1-c DRAM, it is time to buy into beneficiary companies such as Tokyo Electron, Eugene Technology, and EO Technics, and we maintain our Overweight rating on our covered companies.\n\n(11) Tokyo Electron: Expected to benefit from Samsung Electronics' DRAM investment due to its relatively high exposure to batch-type CVD equipment for titanium nitride (TiN) processes.\n\n(12) EO Technics: Expected to supply laser annealing and marking equipment, and potentially femtosecond laser-based cutting equipment, for the P4/P3 lines dedicated to 1-c DRAM.\n\n(13) Eugene Technology: Expected to benefit from Samsung Electronics' 1-c DRAM CAPEX execution due to its dominant market share in the LPCVD market, capable of controlling thickness and uniformity.\n\n(14) However, we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04328085533006665, "ret_signed_m1d": 0.04328085533006665, "alpha_spy_m1d": 0.04757236070003834, "alpha_c_m1d": 0.024436354386004666, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021489145474934856, "ret_signed_p1d": -0.021489145474934856, "alpha_spy_p1d": -0.018145954356419747, "alpha_c_p1d": -0.026716120383602493, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": 0.01967316344798875, "ret_signed_p1w": -0.01967316344798875, "alpha_spy_p1w": -0.00887178002717337, "alpha_c_p1w": -0.038345758363696913, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.020883793309879328, "ret_signed_p1m": 0.020883793309879328, "alpha_spy_p1m": 0.05745112723294499, "alpha_c_p1m": 0.05692237100171926, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.35102747899679687, "ret_signed_p3m": -0.35102747899679687, "alpha_spy_p3m": -0.2984823608562328, "alpha_c_p3m": -0.2322516773020329, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.8829282125608973, "ret_signed_p6m": -0.8829282125608973, "alpha_spy_p6m": -0.7681709092797102, "alpha_c_p6m": -0.5548057127251185, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.3045, -0.3095, -0.3072, -0.3008, -0.2665, -0.2418, -0.2125, -0.2167, -0.2167, -0.2025, -0.1525, -0.1448, -0.1448, -0.1448, -0.192, -0.1265, -0.1311, -0.089, -0.1211, -0.0581, -0.0469, -0.0702, -0.0899, -0.0878, -0.0869, -0.0926, -0.1035, -0.0436, -0.0357, -0.0584, -0.0623, -0.0914, -0.1199, -0.1233, -0.1091, -0.0938, -0.0781, -0.0381, -0.0533, -0.0024, -0.0218, -0.0614, -0.0451, 0.0139, 0.0033, -0.0133, 0.0051, -0.0254, 0.0058, 0.0463, 0.0451, 0.0545, 0.0493, 0.0169, -0.0015, 0.0327, 0.0242, 0.0176, 0.0061, -0.0142, -0.0142, -0.0303, -0.0433, 0.0, 0.0215, 0.0303, 0.0112, 0.0112, 0.0197, 0.0094, 0.0318, 0.0321, -0.0012, -0.0115, 0.0015, 0.0318, -0.0539, -0.0157, -0.0357, -0.0708, -0.0448, -0.0702, -0.0702, -0.0524, -0.0209, -0.0481, -0.072, -0.089, -0.0672, -0.1091, -0.1392, -0.143, -0.1359, -0.1317, -0.136, -0.1289, -0.1303, -0.1836, -0.1922, -0.1979, -0.2011, -0.196, -0.169, -0.1628, -0.1056, -0.0645, 0.0336, 0.0336, 0.0763, 0.1365, 0.2276, 0.2721, 0.3559, 0.3559, 0.3469, 0.3593, 0.2939, 0.2784, 0.2763, 0.2887, 0.4343, 0.4203, 0.4799, 0.451, 0.4358, 0.3769, 0.351, 0.351, 0.2778, 0.3626, 0.4462, 0.4349, 0.4936, 0.465, 0.4869, 0.4328, 0.4823, 0.5419, 0.5225, 0.6759, 0.655, 0.724, 0.724, 0.7261, 0.6717, 0.7772, 0.7632, 0.799, 0.6541, 0.3501, 0.3294, 0.2498, 0.3541, 0.2258, 0.2197, 0.2829, 0.1899, 0.1899, 0.2696, 0.2617, 0.3285, 0.3057, 0.2969, 0.3437, 0.3963, 0.4228, 0.3957, 0.3981, 0.4817, 0.4452, 0.4489, 0.4644, 0.3726, 0.3337, 0.3899, 0.372, 0.3878, 0.5547, 0.5398, 0.5729, 0.6413, 0.734, 0.699, 0.6702, 0.6702, 0.6702, 0.6702, 0.8082, 0.7665, 0.8829]}
{"unit_id": "orig:1945268751550800199", "ticker_idx": 5, "ts": "2025-07-15T23:46:26+00:00", "symbol": "HPSP", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA channel checks point to imminent Samsung 1-c DRAM capex; buy TEL/Eugene Technology/EO Technics as beneficiaries; cautious on Kokusai Electric/Wonik IPS/HPSP on share-loss risk.", "resolved_tickers": ["403870.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "HPSP Co. 403870.KQ Korea KOSDAQ HP annealing equipment", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "original", "lang": "en", "tweet_text": "250716_Semicap equipment: SEC's Capex for 1c DRAM coming soon - CLSA\n\n(1) Our channel checks indicate that Samsung Electronics is soon expected to place large-scale capital expenditure orders for 1-c DRAM mass production.\n\n(2) This is because, even though they haven't completely resolved all complex technical issues of 1-c DRAM, they have made progress in yield improvement sufficient to obtain internal Production Readiness Approval.\n\n(3) Samsung Electronics is believed to be ready to execute investments to secure sufficient capacity for HBM4 and LPDDR5 demand in 2026.\n\n(4) We believe Samsung Electronics has achieved a partial breakthrough, sacrificing some of the cost structure to begin 1-c DRAM mass production by the end of 2025.\n\n(5) A year ago, 1-c yields were significantly low, but they have now reached approximately 50% at the lab level. This progress is mainly attributed to 1) core circuit redesign and 2) increasing die size to widen capacitor pitch.\n\n(6) Further improvements in chip design and process will still be necessary. Key challenges include 1) addressing excessive EUV usage, 2) aggressive clocking, and 3) resolving low power efficiency issues. Solving all technical problems may require more time. * Risk of overlay misalignment between multiple EUV layers, affecting transistor variability. \n* Increased process-induced parasitics, leading to leakage current and non-uniform cell characteristics. \n* Greater manufacturing process variations, necessitating guard-banding (e.g., over-supplying voltage) at the design stage.\n\n(7) Nevertheless, we anticipate a bold investment execution, primarily centered on P4.\n(8) Our channel checks suggest that P4 line will begin orders in 2H25 to build 50K (monthly) 1-c DRAM CAPA. For 1H26, we forecast 40K/10K orders for P4/P3 respectively, and 30K for P3 in 2H26.\n(9) To secure sufficient 1-c DRAM production volume for important customers like NVIDIA, Samsung is proactively starting with ample capacity to offset potential mass production yield risks that may arise in the future.\n\n(10) Accordingly, we believe that regardless of the actual progress of 1-c DRAM, it is time to buy into beneficiary companies such as Tokyo Electron, Eugene Technology, and EO Technics, and we maintain our Overweight rating on our covered companies.\n\n(11) Tokyo Electron: Expected to benefit from Samsung Electronics' DRAM investment due to its relatively high exposure to batch-type CVD equipment for titanium nitride (TiN) processes.\n\n(12) EO Technics: Expected to supply laser annealing and marking equipment, and potentially femtosecond laser-based cutting equipment, for the P4/P3 lines dedicated to 1-c DRAM.\n\n(13) Eugene Technology: Expected to benefit from Samsung Electronics' 1-c DRAM CAPEX execution due to its dominant market share in the LPCVD market, capable of controlling thickness and uniformity.\n\n(14) However, we maintain a cautious stance on other equipment stocks such as Kokusai Electric/Wonik IPS/HPSP due to potential market share declines.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.004291505369971693, "alpha_c_m1d": 0.007539333776067192, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": 0.007539333776067192, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.001919429239003012, "ret_signed_p1d": -0.001919429239003012, "alpha_spy_p1d": 0.0014237618795120977, "alpha_c_p1d": 0.0016166811383329094, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.0035361103773359215, "ret_p1w": -0.026871399688922493, "ret_signed_p1w": 0.026871399688922493, "alpha_spy_p1w": 0.03767278310973787, "alpha_c_p1w": 0.034477493286809935, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.007606093597887442, "ret_p1m": 0.019193911354330462, "ret_signed_p1m": -0.019193911354330462, "alpha_spy_p1m": 0.017373422568735197, "alpha_c_p1m": -0.0018467471140592995, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.017347164240271162, "ret_p3m": 0.4030709953338367, "ret_signed_p3m": -0.4030709953338367, "alpha_spy_p3m": -0.35052587719327266, "alpha_c_p3m": -0.3988056702001319, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.0042653251337048115, "ret_p6m": 0.2953994511561766, "ret_signed_p6m": -0.2953994511561766, "alpha_spy_p6m": -0.18064214787498956, "alpha_c_p6m": -0.2267211845332895, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.06867826662288712, "price_path": [-0.0768, -0.0672, -0.0864, -0.0979, -0.0633, -0.048, -0.0768, -0.1459, -0.1267, -0.1382, -0.1382, -0.1382, -0.1382, -0.1382, -0.1248, -0.1248, -0.1344, -0.0998, -0.0845, -0.0633, -0.0806, -0.1036, -0.1497, -0.1459, -0.1459, -0.1804, -0.1823, -0.1612, -0.1612, -0.142, -0.1228, -0.1459, -0.1555, -0.1555, -0.1459, -0.1228, -0.1228, -0.0787, -0.0729, 0.0633, 0.094, 0.0614, 0.0825, 0.0825, 0.094, 0.0825, 0.0806, 0.0595, 0.0672, 0.1036, 0.1228, 0.0998, 0.0653, 0.0633, 0.0345, 0.071, 0.0403, 0.0288, 0.023, 0.0096, -0.0038, 0.0077, 0.0, 0.0, 0.0019, -0.0058, 0.0, -0.0115, -0.0269, -0.0384, -0.0461, -0.0595, -0.0192, -0.0192, 0.0403, 0.0307, -0.048, -0.0576, -0.0345, 0.0326, 0.0038, 0.0288, 0.0154, -0.0019, 0.0192, 0.0403, 0.0403, 0.0269, 0.0096, -0.0115, -0.0211, -0.0403, -0.023, -0.0077, -0.0096, -0.0269, 0.025, -0.0192, -0.0115, 0.0019, 0.0326, 0.0326, 0.0365, 0.0518, 0.0672, 0.0595, 0.1036, 0.1689, 0.1843, 0.1497, 0.1958, 0.1958, 0.3148, 0.3225, 0.3608, 0.3186, 0.3052, 0.309, 0.2956, 0.3186, 0.3417, 0.3417, 0.3417, 0.3417, 0.3417, 0.3417, 0.4031, 0.4146, 0.357, 0.3896, 0.3992, 0.334, 0.3916, 0.38, 0.3666, 0.3378, 0.3992, 0.4223, 0.3743, 0.3589, 0.3493, 0.3282, 0.3378, 0.3071, 0.2514, 0.2783, 0.2131, 0.2495, 0.2361, 0.2534, 0.2668, 0.1939, 0.2131, 0.1478, 0.1152, 0.1574, 0.0883, 0.0825, 0.0845, 0.1056, 0.1056, 0.1612, 0.2015, 0.1996, 0.19, 0.1727, 0.1631, 0.167, 0.1555, 0.1382, 0.1305, 0.1305, 0.1075, 0.0806, 0.1267, 0.0979, 0.1017, 0.1497, 0.167, 0.1536, 0.1536, 0.1881, 0.2954, 0.3071, 0.3071, 0.3071, 0.3442, 0.3656, 0.5276, 0.2954]}
{"unit_id": "thread:1945698437837869119", "ticker_idx": 0, "ts": "2025-07-17T04:13:51+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura has raised TSMC's target price from NT1,170toNT1,280, forecasting that 2nm process demand will be a key growth driver in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Nomura's raised TSMC PT to NT$1,280 on strong 2nm demand in 2026, driven by full Apple iPhone lineup adoption.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nomura: TSMC Target Price Raised to NT$1,280 on Strong 2nm Demand in 2026\n\nNomura has raised TSMC's target price from NT1,170toNT1,280, forecasting that 2nm process demand will be a key growth driver in 2026. Notably, Apple is expected to fully adopt the 2nm process across its entire iPhone lineup starting in 2026, which is an expansion compared to its 2023 application solely in the high-end Pro series. This is projected to contribute approximately 4% of TSMC's wafer revenue.\n\n$TSM\n\n---\n\nhttps://t.co/cYCYmTZVkE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032736137765441486, "ret_signed_m1d": -0.032736137765441486, "alpha_spy_m1d": -0.02665367060126722, "alpha_c_m1d": -0.024405971921902414, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021172647928687227, "ret_signed_p1d": -0.021172647928687227, "alpha_spy_p1d": -0.02044023502657677, "alpha_c_p1d": -0.016373454888251704, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": -0.016286599225248666, "ret_signed_p1w": -0.016286599225248666, "alpha_spy_p1w": -0.02644521420138235, "alpha_c_p1w": -0.002882922782284747, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": -0.02736158446200665, "ret_signed_p1m": -0.02736158446200665, "alpha_spy_p1m": -0.051882372472820015, "alpha_c_p1m": -0.040868208695853725, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.20876894271373292, "ret_signed_p3m": 0.20876894271373292, "alpha_spy_p3m": 0.15140639950328283, "alpha_c_p3m": 0.06551011307794985, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 0.3254373776897599, "ret_signed_p6m": 0.3254373776897599, "alpha_spy_p6m": 0.2139622326356443, "alpha_c_p6m": -0.012984752748134376, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.4002, -0.3858, -0.3598, -0.334, -0.3302, -0.3371, -0.3327, -0.3238, -0.2993, -0.2727, -0.2844, -0.3011, -0.2919, -0.2892, -0.2839, -0.2415, -0.213, -0.2099, -0.2121, -0.2121, -0.215, -0.2152, -0.2221, -0.2041, -0.2212, -0.2212, -0.198, -0.2043, -0.2002, -0.2157, -0.2096, -0.1983, -0.1789, -0.1751, -0.1676, -0.1602, -0.1381, -0.1314, -0.1228, -0.1405, -0.1218, -0.1291, -0.1307, -0.1307, -0.1469, -0.1436, -0.1039, -0.0931, -0.0879, -0.0693, -0.0778, -0.0852, -0.0489, -0.044, -0.044, -0.0669, -0.0722, -0.056, -0.0645, -0.0619, -0.0689, -0.0352, -0.0327, 0.0, -0.0212, -0.0275, -0.0448, -0.0215, -0.0163, 0.0, -0.0116, -0.0174, -0.011, -0.0162, -0.0423, -0.0269, -0.0535, -0.0579, -0.0121, -0.0154, -0.0143, -0.0053, -0.0169, -0.0187, -0.0274, -0.0171, -0.0525, -0.0692, -0.0744, -0.0513, -0.0408, -0.028, -0.0257, -0.0298, -0.06, -0.06, -0.0701, -0.0579, -0.0423, -0.0089, 0.0065, 0.0217, 0.0604, 0.0542, 0.0559, 0.0643, 0.0704, 0.0734, 0.0973, 0.0819, 0.1136, 0.1547, 0.1466, 0.13, 0.1165, 0.116, 0.1408, 0.1783, 0.1768, 0.1935, 0.2352, 0.201, 0.2438, 0.2249, 0.1464, 0.2372, 0.2088, 0.2446, 0.2247, 0.2053, 0.216, 0.2029, 0.1799, 0.1875, 0.2048, 0.2182, 0.2316, 0.2461, 0.2385, 0.2271, 0.2452, 0.201, 0.1994, 0.1814, 0.1702, 0.206, 0.1893, 0.187, 0.1526, 0.1633, 0.1519, 0.1351, 0.1533, 0.1335, 0.1235, 0.1626, 0.1628, 0.1843, 0.1843, 0.1907, 0.175, 0.193, 0.2068, 0.1965, 0.2038, 0.233, 0.2393, 0.2668, 0.2485, 0.1961, 0.1784, 0.1749, 0.1343, 0.1659, 0.1834, 0.2011, 0.2162, 0.2237, 0.2237, 0.2403, 0.2324, 0.2269, 0.2446, 0.2446, 0.309, 0.3198, 0.341, 0.3052, 0.3024, 0.3254]}
{"unit_id": "orig:1968650264141037649", "ticker_idx": 0, "ts": "2025-09-18T12:16:13+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Buy ratings on Alphabet, Amazon, and Meta, reflecting the significant AI-driven revenue upside potential.", "tweet_type": "analysis", "timeline": "years", "summary": "BofA reiterates Buy on Alphabet, Amazon, and Meta on AI-driven revenue upside despite rising D&A and margin timing risk.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Deep Dive into Internet Companies’ Capital Expenditures (CAPEX): Surging Capex and Depreciation & Amortization (D&A)\nRising capital expenditures by large internet companies are leading to higher depreciation and amortization (D&A).\n\n1. Rising CAPEX = Rising D&A\n\nAs D&A is expected to rise, expectations for new AI capabilities and future revenue generation through AI infrastructure remain a key driver of investment sentiment and performance for large-cap internet stocks. In this report, we review the positive aspects and risks related to capex in the sector, present segment-specific capex estimates for major internet companies (Google, Meta, Amazon), and update depreciation estimates (for Meta). Our current capex forecasts are slightly above Street consensus (with room for further upside), and our depreciation estimates are even higher. The key points are as follows:\n\n2. Positives: AI Capex Drives Growth and Efficiency\n\nAI is emerging as a multi-dimensional growth driver for large internet companies, accelerating core businesses, creating new revenue streams, and improving operational efficiency.\n\n1. Google: AI is boosting search ad profitability and accelerating cloud adoption, with rising AI compute demand leading to larger-scale contracts.\n\n2. Meta: AI integration is improving content and ad targeting, driving higher engagement and improving ad conversion.\n\n3. Amazon: Expanding capacity is translating into AWS revenue growth, while also improving efficiency in the retail distribution segment.\n\nBeyond core businesses, AI is also opening up new opportunities. Google is expanding subscription services through Gemini Advanced, Workspace, and Vertex AI. Meta is offering automated ad campaigns and could launch premium AI services. Amazon is building shopping agents. In addition, AI can enhance productivity, curb headcount growth, and streamline support functions, thereby driving operating margin leverage. Looking forward, the ratio of capex to expected revenue is lower than in the past, suggesting potential for further revenue upside.\n\n3. Risks: Margin Pressure, Short Asset Life, and Overcapacity\n\nThe key risks are as follows:\n\n1. Margin pressure: AI-driven revenue streams may only begin to scale meaningfully in 2026–2027, whereas associated depreciation could rise earlier. This timing mismatch may pressure margins before new revenue is realized.\n\n2. Short asset lives: AI infrastructure faces high workloads and rapid innovation cycles, which could shorten asset life spans. AI-related assets may require replacement every 3–5 years, pulling forward cost recognition and impacting profitability.\n\n3. Overcapacity and pricing risk: The surge in AI infrastructure could outpace demand, while LLMs become increasingly commoditized as performance converges. If supply exceeds consumption (which we view as unlikely before 2027), hyperscalers may rely on aggressive pricing strategies to maintain utilization, hurting profitability.\n\n4. Market Consensus Tends to Underestimate Depreciation\n\nWith sector capex growth at 56% in 2024 and an estimated 63% in 2025, we believe Street consensus still underestimates rising depreciation. Based on our updated analysis (detailed in the depreciation breakdown table), the market may be underestimating Google, Meta, and Amazon’s depreciation for 2026–28, and this gap could widen over time. In 2027, the difference between Street consensus and BofA estimates for depreciation is expected to be largest at Alphabet ($7B), followed by Amazon ($5.9B) and Meta (~$3.5B).\n\nWe already forecast higher depreciation for Alphabet and Amazon compared with consensus, and we are now revising up Meta’s depreciation forecast. Accordingly, we lower our 2027 EPS estimate for Meta by 3% to $34.96, though this remains above the Street’s $34.26 estimate thanks to stronger revenue forecasts. We maintain Buy ratings on Alphabet, Amazon, and Meta, reflecting the significant AI-driven revenue upside potential.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009919470761090121, "ret_signed_m1d": -0.009919470761090121, "alpha_spy_m1d": -0.005268755829151872, "alpha_c_m1d": -0.00916274203735501, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.0007567287237351117, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010673352478958398, "ret_signed_p1d": 0.010673352478958398, "alpha_spy_p1d": 0.005720477703626381, "alpha_c_p1d": 0.006552845033354293, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": 0.004120507445604105, "ret_p1w": -0.024758967009263055, "ret_signed_p1w": -0.024758967009263055, "alpha_spy_p1w": -0.021156749522210583, "alpha_c_p1w": -0.012298050729286802, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": -0.012460916279976253, "ret_p1m": 0.005039033915886559, "ret_signed_p1m": 0.005039033915886559, "alpha_spy_p1m": -0.0009586205578728801, "alpha_c_p1m": 0.03513602529880744, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": -0.03009699138292088, "ret_p3m": 0.21719844579720804, "ret_signed_p3m": 0.21719844579720804, "alpha_spy_p3m": 0.18927566041694854, "alpha_c_p3m": 0.23286588584464762, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": -0.015667440047439585, "ret_p6m": 0.20101038944189709, "ret_signed_p6m": 0.20101038944189709, "alpha_spy_p6m": 0.1952292936993303, "alpha_c_p6m": 0.23211563972284377, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": -0.031105250280946684, "price_path": [-0.3451, -0.3389, -0.3234, -0.312, -0.2923, -0.3014, -0.3029, -0.2918, -0.2883, -0.2883, -0.2992, -0.3088, -0.2998, -0.2959, -0.2857, -0.2803, -0.2785, -0.2747, -0.2722, -0.2664, -0.2464, -0.2415, -0.2459, -0.2382, -0.2342, -0.2366, -0.224, -0.2209, -0.2393, -0.2502, -0.2268, -0.2283, -0.2227, -0.2209, -0.2015, -0.2032, -0.1939, -0.1994, -0.1955, -0.1917, -0.1933, -0.2009, -0.2098, -0.2081, -0.183, -0.1735, -0.1788, -0.1775, -0.161, -0.156, -0.156, -0.1622, -0.0856, -0.0791, -0.0684, -0.0714, -0.0492, -0.051, -0.0463, -0.0446, -0.0017, -0.0035, -0.0099, 0.0, 0.0107, 0.002, -0.0015, -0.0194, -0.0248, -0.0218, -0.0317, -0.0354, -0.0283, -0.0252, -0.0265, -0.0063, -0.0249, -0.0294, -0.0417, -0.0613, -0.0313, -0.0261, -0.004, -0.0023, 0.005, 0.0179, -0.0062, -0.0013, 0.0042, 0.0313, 0.0684, 0.0613, 0.0894, 0.1169, 0.1157, 0.1257, 0.1012, 0.1281, 0.1298, 0.1063, 0.1511, 0.1559, 0.1376, 0.1053, 0.0967, 0.1309, 0.128, 0.1618, 0.1485, 0.189, 0.2641, 0.2833, 0.2695, 0.2695, 0.2704, 0.2494, 0.2531, 0.2682, 0.2602, 0.2747, 0.2456, 0.2589, 0.2714, 0.2405, 0.228, 0.2237, 0.2172, 0.1781, 0.2009, 0.2195, 0.2299, 0.2481, 0.2471, 0.2471, 0.2448, 0.245, 0.2461, 0.2427, 0.2427, 0.2513, 0.2568, 0.248, 0.2784, 0.2921, 0.3045, 0.3176, 0.3339, 0.3334, 0.3213, 0.3102, 0.3102, 0.2785, 0.3038, 0.3124, 0.302, 0.3232, 0.3283, 0.3341, 0.343, 0.342, 0.3646, 0.3488, 0.3223, 0.3152, 0.2819, 0.2877, 0.2649, 0.2346, 0.2268, 0.2138, 0.2138, 0.1991, 0.2043, 0.2024, 0.2506, 0.2367, 0.2344, 0.2423, 0.2204, 0.2378, 0.217, 0.2053, 0.2035, 0.1946, 0.1852, 0.2172, 0.2199, 0.2265, 0.2061, 0.201]}
{"unit_id": "orig:1968650264141037649", "ticker_idx": 1, "ts": "2025-09-18T12:16:13+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Buy ratings on Alphabet, Amazon, and Meta, reflecting the significant AI-driven revenue upside potential.", "tweet_type": "analysis", "timeline": "years", "summary": "BofA reiterates Buy on Alphabet, Amazon, and Meta on AI-driven revenue upside despite rising D&A and margin timing risk.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Deep Dive into Internet Companies’ Capital Expenditures (CAPEX): Surging Capex and Depreciation & Amortization (D&A)\nRising capital expenditures by large internet companies are leading to higher depreciation and amortization (D&A).\n\n1. Rising CAPEX = Rising D&A\n\nAs D&A is expected to rise, expectations for new AI capabilities and future revenue generation through AI infrastructure remain a key driver of investment sentiment and performance for large-cap internet stocks. In this report, we review the positive aspects and risks related to capex in the sector, present segment-specific capex estimates for major internet companies (Google, Meta, Amazon), and update depreciation estimates (for Meta). Our current capex forecasts are slightly above Street consensus (with room for further upside), and our depreciation estimates are even higher. The key points are as follows:\n\n2. Positives: AI Capex Drives Growth and Efficiency\n\nAI is emerging as a multi-dimensional growth driver for large internet companies, accelerating core businesses, creating new revenue streams, and improving operational efficiency.\n\n1. Google: AI is boosting search ad profitability and accelerating cloud adoption, with rising AI compute demand leading to larger-scale contracts.\n\n2. Meta: AI integration is improving content and ad targeting, driving higher engagement and improving ad conversion.\n\n3. Amazon: Expanding capacity is translating into AWS revenue growth, while also improving efficiency in the retail distribution segment.\n\nBeyond core businesses, AI is also opening up new opportunities. Google is expanding subscription services through Gemini Advanced, Workspace, and Vertex AI. Meta is offering automated ad campaigns and could launch premium AI services. Amazon is building shopping agents. In addition, AI can enhance productivity, curb headcount growth, and streamline support functions, thereby driving operating margin leverage. Looking forward, the ratio of capex to expected revenue is lower than in the past, suggesting potential for further revenue upside.\n\n3. Risks: Margin Pressure, Short Asset Life, and Overcapacity\n\nThe key risks are as follows:\n\n1. Margin pressure: AI-driven revenue streams may only begin to scale meaningfully in 2026–2027, whereas associated depreciation could rise earlier. This timing mismatch may pressure margins before new revenue is realized.\n\n2. Short asset lives: AI infrastructure faces high workloads and rapid innovation cycles, which could shorten asset life spans. AI-related assets may require replacement every 3–5 years, pulling forward cost recognition and impacting profitability.\n\n3. Overcapacity and pricing risk: The surge in AI infrastructure could outpace demand, while LLMs become increasingly commoditized as performance converges. If supply exceeds consumption (which we view as unlikely before 2027), hyperscalers may rely on aggressive pricing strategies to maintain utilization, hurting profitability.\n\n4. Market Consensus Tends to Underestimate Depreciation\n\nWith sector capex growth at 56% in 2024 and an estimated 63% in 2025, we believe Street consensus still underestimates rising depreciation. Based on our updated analysis (detailed in the depreciation breakdown table), the market may be underestimating Google, Meta, and Amazon’s depreciation for 2026–28, and this gap could widen over time. In 2027, the difference between Street consensus and BofA estimates for depreciation is expected to be largest at Alphabet ($7B), followed by Amazon ($5.9B) and Meta (~$3.5B).\n\nWe already forecast higher depreciation for Alphabet and Amazon compared with consensus, and we are now revising up Meta’s depreciation forecast. Accordingly, we lower our 2027 EPS estimate for Meta by 3% to $34.96, though this remains above the Street’s $34.26 estimate thanks to stronger revenue forecasts. We maintain Buy ratings on Alphabet, Amazon, and Meta, reflecting the significant AI-driven revenue upside potential.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0016866297489706028, "ret_signed_m1d": 0.0016866297489706028, "alpha_spy_m1d": 0.006337344680908852, "alpha_c_m1d": -0.002846051491590007, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": 0.00453268124056061, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0010811746080494622, "ret_signed_p1d": 0.0010811746080494622, "alpha_spy_p1d": -0.0038717001672825546, "alpha_c_p1d": -0.00270299456269818, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": 0.003784169170747642, "ret_p1w": -0.05656706341190698, "ret_signed_p1w": -0.05656706341190698, "alpha_spy_p1w": -0.05296484592485451, "alpha_c_p1w": -0.04058506394975847, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": -0.01598199946214851, "ret_p1m": -0.07866627504002432, "ret_signed_p1m": -0.07866627504002432, "alpha_spy_p1m": -0.08466392951378376, "alpha_c_p1m": -0.05518600853424516, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": -0.023480266505779168, "ret_p3m": -0.037495127488395896, "ret_signed_p3m": -0.037495127488395896, "alpha_spy_p3m": -0.06541791286865539, "alpha_c_p3m": -0.05333896268836136, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": 0.01584383519996546, "ret_p6m": -0.1018898845042352, "ret_signed_p6m": -0.1018898845042352, "alpha_spy_p6m": -0.107670980246802, "alpha_c_p6m": -0.02733587427535733, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": -0.07455401022887786, "price_path": [-0.0984, -0.0798, -0.0832, -0.061, -0.0343, -0.0512, -0.0466, -0.0489, -0.0338, -0.0338, -0.0336, -0.0513, -0.0376, -0.0388, -0.0269, -0.024, -0.0211, -0.0348, -0.0318, -0.0221, -0.0083, -0.0163, -0.0127, 0.0043, 0.0009, 0.0067, -0.001, -0.0045, 0.0125, -0.0713, -0.0847, -0.0756, -0.0386, -0.035, -0.0369, -0.0429, -0.0422, -0.0288, -0.0011, -0.0009, 0.0011, -0.0139, -0.0321, -0.0401, -0.0103, -0.0142, -0.0109, -0.0091, 0.0016, -0.0096, -0.0096, -0.0255, -0.0227, 0.0192, 0.0048, 0.0199, 0.0303, -0.0039, -0.0055, -0.0133, 0.0009, 0.0122, 0.0017, 0.0, 0.0011, -0.0156, -0.0455, -0.0477, -0.0566, -0.0495, -0.0392, -0.0504, -0.0458, -0.0381, -0.0507, -0.0447, -0.0409, -0.026, -0.0151, -0.0643, -0.0483, -0.0642, -0.0677, -0.0725, -0.0787, -0.0638, -0.0398, -0.0574, -0.0439, -0.0304, -0.0184, -0.0086, -0.004, -0.0362, 0.0562, 0.0985, 0.0782, 0.082, 0.0511, 0.057, 0.0743, 0.0773, 0.0561, 0.0275, 0.015, 0.0071, -0.0375, -0.0369, -0.0609, -0.0456, -0.0214, -0.0067, -0.009, -0.009, 0.0086, 0.0115, 0.0138, 0.005, -0.0092, -0.0074, -0.0188, -0.0143, 0.0024, -0.0041, -0.0218, -0.0376, -0.0375, -0.0431, -0.0193, -0.0168, -0.0121, 0.0039, 0.005, 0.005, 0.0056, 0.0036, 0.0056, -0.0018, -0.0018, -0.0205, 0.0079, 0.0419, 0.0447, 0.0651, 0.0698, 0.0659, 0.0492, 0.0234, 0.0301, 0.0341, 0.0341, -0.001, 0.0003, 0.0134, 0.0343, 0.0311, 0.0582, 0.0509, 0.0454, 0.0349, 0.0507, 0.032, 0.0076, -0.0369, -0.0904, -0.0973, -0.105, -0.1174, -0.1368, -0.1403, -0.1403, -0.1301, -0.1143, -0.114, -0.0913, -0.1123, -0.098, -0.089, -0.1008, -0.0918, -0.0988, -0.0973, -0.0623, -0.0532, -0.0779, -0.0767, -0.0731, -0.0804, -0.0938, -0.1019]}
{"unit_id": "orig:1968650264141037649", "ticker_idx": 2, "ts": "2025-09-18T12:16:13+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Buy ratings on Alphabet, Amazon, and Meta, reflecting the significant AI-driven revenue upside potential.", "tweet_type": "analysis", "timeline": "years", "summary": "BofA reiterates Buy on Alphabet, Amazon, and Meta on AI-driven revenue upside despite rising D&A and margin timing risk.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Deep Dive into Internet Companies’ Capital Expenditures (CAPEX): Surging Capex and Depreciation & Amortization (D&A)\nRising capital expenditures by large internet companies are leading to higher depreciation and amortization (D&A).\n\n1. Rising CAPEX = Rising D&A\n\nAs D&A is expected to rise, expectations for new AI capabilities and future revenue generation through AI infrastructure remain a key driver of investment sentiment and performance for large-cap internet stocks. In this report, we review the positive aspects and risks related to capex in the sector, present segment-specific capex estimates for major internet companies (Google, Meta, Amazon), and update depreciation estimates (for Meta). Our current capex forecasts are slightly above Street consensus (with room for further upside), and our depreciation estimates are even higher. The key points are as follows:\n\n2. Positives: AI Capex Drives Growth and Efficiency\n\nAI is emerging as a multi-dimensional growth driver for large internet companies, accelerating core businesses, creating new revenue streams, and improving operational efficiency.\n\n1. Google: AI is boosting search ad profitability and accelerating cloud adoption, with rising AI compute demand leading to larger-scale contracts.\n\n2. Meta: AI integration is improving content and ad targeting, driving higher engagement and improving ad conversion.\n\n3. Amazon: Expanding capacity is translating into AWS revenue growth, while also improving efficiency in the retail distribution segment.\n\nBeyond core businesses, AI is also opening up new opportunities. Google is expanding subscription services through Gemini Advanced, Workspace, and Vertex AI. Meta is offering automated ad campaigns and could launch premium AI services. Amazon is building shopping agents. In addition, AI can enhance productivity, curb headcount growth, and streamline support functions, thereby driving operating margin leverage. Looking forward, the ratio of capex to expected revenue is lower than in the past, suggesting potential for further revenue upside.\n\n3. Risks: Margin Pressure, Short Asset Life, and Overcapacity\n\nThe key risks are as follows:\n\n1. Margin pressure: AI-driven revenue streams may only begin to scale meaningfully in 2026–2027, whereas associated depreciation could rise earlier. This timing mismatch may pressure margins before new revenue is realized.\n\n2. Short asset lives: AI infrastructure faces high workloads and rapid innovation cycles, which could shorten asset life spans. AI-related assets may require replacement every 3–5 years, pulling forward cost recognition and impacting profitability.\n\n3. Overcapacity and pricing risk: The surge in AI infrastructure could outpace demand, while LLMs become increasingly commoditized as performance converges. If supply exceeds consumption (which we view as unlikely before 2027), hyperscalers may rely on aggressive pricing strategies to maintain utilization, hurting profitability.\n\n4. Market Consensus Tends to Underestimate Depreciation\n\nWith sector capex growth at 56% in 2024 and an estimated 63% in 2025, we believe Street consensus still underestimates rising depreciation. Based on our updated analysis (detailed in the depreciation breakdown table), the market may be underestimating Google, Meta, and Amazon’s depreciation for 2026–28, and this gap could widen over time. In 2027, the difference between Street consensus and BofA estimates for depreciation is expected to be largest at Alphabet ($7B), followed by Amazon ($5.9B) and Meta (~$3.5B).\n\nWe already forecast higher depreciation for Alphabet and Amazon compared with consensus, and we are now revising up Meta’s depreciation forecast. Accordingly, we lower our 2027 EPS estimate for Meta by 3% to $34.96, though this remains above the Street’s $34.26 estimate thanks to stronger revenue forecasts. We maintain Buy ratings on Alphabet, Amazon, and Meta, reflecting the significant AI-driven revenue upside potential.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005805783013084209, "ret_signed_m1d": -0.005805783013084209, "alpha_spy_m1d": -0.0011550680811459602, "alpha_c_m1d": -0.0050490542893490975, "bench_spy_m1d": -0.004650714931938249, "bench_c_m1d": -0.0007567287237351117, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0023965663825364913, "ret_signed_p1d": -0.0023965663825364913, "alpha_spy_p1d": -0.007349441157868508, "alpha_c_p1d": -0.006517073828140596, "bench_spy_p1d": 0.004952874775332017, "bench_c_p1d": 0.004120507445604105, "ret_p1w": -0.03951876413858513, "ret_signed_p1w": -0.03951876413858513, "alpha_spy_p1w": -0.035916546651532655, "alpha_c_p1w": -0.027057847858608874, "bench_spy_p1w": -0.0036022174870524726, "bench_c_p1w": -0.012460916279976253, "ret_p1m": -0.08054611447707216, "ret_signed_p1m": -0.08054611447707216, "alpha_spy_p1m": -0.0865437689508316, "alpha_c_p1m": -0.05044912309415128, "bench_spy_p1m": 0.0059976544737594395, "bench_c_p1m": -0.03009699138292088, "ret_p3m": -0.15651401483636507, "ret_signed_p3m": -0.15651401483636507, "alpha_spy_p3m": -0.18443680021662456, "alpha_c_p3m": -0.14084657478892548, "bench_spy_p3m": 0.027922785380259496, "bench_c_p3m": -0.015667440047439585, "ret_p6m": -0.21227156464774077, "ret_signed_p6m": -0.21227156464774077, "alpha_spy_p6m": -0.21805266039030757, "alpha_c_p6m": -0.1811663143667941, "bench_spy_p6m": 0.0057810957425668, "bench_c_p6m": -0.031105250280946684, "price_path": [-0.1047, -0.0872, -0.0917, -0.0694, -0.0597, -0.054, -0.0782, -0.0855, -0.0785, -0.0785, -0.0793, -0.0764, -0.0608, -0.0679, -0.0804, -0.076, -0.0895, -0.0991, -0.101, -0.0974, -0.0862, -0.0967, -0.0854, -0.0839, -0.0866, -0.0803, -0.1029, -0.109, -0.0087, -0.0388, -0.005, -0.0215, -0.0106, -0.0236, -0.014, -0.0184, 0.0125, -0.0002, 0.0024, 0.0064, -0.0165, -0.0369, -0.0417, -0.0527, -0.0326, -0.0345, -0.0335, -0.0421, -0.0373, -0.0533, -0.0533, -0.0579, -0.0554, -0.0405, -0.0356, -0.0358, -0.0186, -0.0362, -0.0376, -0.0316, -0.0199, -0.0016, -0.0058, 0.0, -0.0024, -0.0187, -0.0312, -0.0244, -0.0395, -0.0461, -0.0466, -0.0582, -0.08, -0.0676, -0.0887, -0.0822, -0.0855, -0.0794, -0.0593, -0.0954, -0.0821, -0.0912, -0.0797, -0.0868, -0.0805, -0.061, -0.0596, -0.0594, -0.0586, -0.053, -0.0371, -0.0363, -0.036, -0.1452, -0.1685, -0.1821, -0.1955, -0.1844, -0.2062, -0.2027, -0.1898, -0.1958, -0.2189, -0.2178, -0.2184, -0.2279, -0.2335, -0.2429, -0.2444, -0.2379, -0.2138, -0.184, -0.1874, -0.1874, -0.169, -0.1781, -0.1701, -0.1797, -0.1516, -0.1363, -0.1448, -0.1574, -0.1662, -0.1629, -0.1738, -0.1689, -0.1565, -0.1663, -0.1471, -0.1544, -0.1509, -0.1465, -0.1432, -0.1432, -0.1486, -0.1545, -0.1452, -0.1527, -0.1527, -0.1652, -0.1544, -0.1521, -0.1674, -0.1707, -0.1618, -0.176, -0.19, -0.2099, -0.2032, -0.2039, -0.2039, -0.2246, -0.2132, -0.1687, -0.1544, -0.137, -0.1362, -0.1417, -0.0523, -0.0803, -0.0933, -0.1122, -0.1413, -0.1398, -0.151, -0.1308, -0.1391, -0.1417, -0.1659, -0.1788, -0.1788, -0.1794, -0.1744, -0.1724, -0.1584, -0.1821, -0.1794, -0.161, -0.1567, -0.168, -0.1611, -0.1592, -0.1429, -0.1521, -0.1723, -0.169, -0.1605, -0.1595, -0.1809, -0.2123]}
{"unit_id": "orig:1945086380583084259", "ticker_idx": 1, "ts": "2025-07-15T11:41:45+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Quanta being the primary ODM partner for AWS GB200 NVL72", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley sees AVC, Quanta, and Delta as key beneficiaries of AWS IRHX liquid cooling adoption for high-density AI racks.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Liquid Cooling: AWS Launches Proprietary IRHX Liquid Cooling Solution, Benefiting Upstream Component Supply Chain\n\nMorgan Stanley (S. Shih, 25/07/11)\n\nMorgan Stanley points out that AWS has launched its self-developed IRHX liquid cooling solution to support high-density AI server racks, indicating an accelerating adoption of liquid cooling architecture during a transition period. Liquid cooling component manufacturers and system integrators in the supply chain are expected to be the first to benefit.\n\nAWS Launches IRHX Liquid Cooling Solution, Suitable for High-Density AI Racks Like GB200 NVL72\n\nAWS has introduced its proprietary In-Row Heat Exchanger (IRHX) liquid cooling solution, which can be inserted into existing or new data centers to support high-power AI racks. The IRHX consists of a manifold, pump units, and dry coolers, featuring modular scalability to dynamically increase or decrease dry cooler units based on thermal load.\n\nIRHX Seen as Transitional Side-Car CDU Solution, Balancing Space and Efficiency But Requiring Attention to Water Leak Risk\n\nIRHX is positioned as a transitional side-car CDU design before the widespread adoption of liquid-to-liquid solutions, capable of supporting 4 to 8 racks and saving data center space. However, its highly integrated nature may pose concerns such as water leakage risks or decreased cooling efficiency, requiring further observation of actual deployment effects.\n\nLiquid Cooling Component Manufacturers to Benefit First, AVC Expected to Stand Out\n\nAs AWS does not have liquid cooling component manufacturing capabilities, relevant parts will depend on external suppliers. Morgan Stanley notes that AVC, with its expertise in cold plates, pump units, and dry coolers, and its close relationship with AWS, is expected to be in a favorable position during this adoption phase.\n\nODMs and System Integrators Play Key Roles, Quanta and Delta are Major Beneficiaries\n\nIRHX is expected to be integrated by server ODM manufacturers, with Quanta being the primary ODM partner for AWS GB200 NVL72. Additionally, if AWS prefers partners with accumulated liquid cooling technology and firmware capabilities for integration, liquid cooling system providers like Delta are also expected to participate in IRHX deployment.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003696857670979714, "ret_signed_m1d": -0.003696857670979714, "alpha_spy_m1d": -0.007988363040951407, "alpha_c_m1d": 0.0051047485829557315, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.008801606253935446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007393715341959428, "ret_signed_p1d": 0.007393715341959428, "alpha_spy_p1d": 0.004050524223444318, "alpha_c_p1d": 0.004291934443550138, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00310178089840929, "ret_p1w": -0.02957486136783738, "ret_signed_p1w": -0.02957486136783738, "alpha_spy_p1w": -0.04037624478865276, "alpha_c_p1w": -0.03306442428142986, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.003489562913592481, "ret_p1m": 0.003696857670979714, "ret_signed_p1m": 0.003696857670979714, "alpha_spy_p1m": -0.032870476252085945, "alpha_c_p1m": -0.03662712564869097, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.040323983319670686, "ret_p3m": 0.1016635859519408, "ret_signed_p3m": 0.1016635859519408, "alpha_spy_p3m": 0.04911846781137674, "alpha_c_p3m": 0.02086877538773213, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.08079481056420867, "ret_p6m": 0.04251386321626627, "ret_signed_p6m": 0.04251386321626627, "alpha_spy_p6m": -0.07224344006492078, "alpha_c_p6m": -0.09695994605295954, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.1394738092692258, "price_path": [-0.1884, -0.1867, -0.1814, -0.2114, -0.1708, -0.1884, -0.1708, -0.1355, -0.1443, -0.1584, -0.1584, -0.1143, -0.1302, -0.1038, -0.1038, -0.1126, -0.0949, -0.0879, -0.0755, -0.0244, -0.0208, -0.0297, -0.0614, -0.0596, -0.0402, -0.0561, -0.0685, -0.0632, -0.0614, -0.0508, -0.042, -0.042, -0.0438, -0.0297, -0.012, -0.0067, -0.0261, -0.0261, 0.0003, -0.0014, 0.0109, 0.0145, 0.0092, -0.0014, -0.0014, -0.0155, -0.0138, -0.005, 0.0215, 0.0339, 0.0233, 0.0074, 0.0148, 0.0388, 0.0425, 0.0351, 0.0277, 0.024, 0.0055, 0.0203, 0.0092, 0.0092, -0.0037, 0.0, 0.0074, 0.0018, 0.0111, 0.0111, -0.0296, -0.0222, -0.0055, -0.0074, 0.0055, -0.0111, -0.0018, 0.0407, 0.0407, 0.0388, 0.0647, 0.0665, 0.0684, 0.0739, 0.0721, 0.0462, 0.0037, 0.0, 0.0, -0.0055, -0.0092, -0.0407, -0.0425, -0.0444, -0.0092, -0.0203, -0.0055, -0.0166, -0.0296, -0.0536, -0.0628, -0.0628, -0.0499, -0.0444, -0.0388, -0.0296, 0.0092, -0.0018, 0.0129, 0.0203, 0.0111, 0.0092, 0.0333, 0.0388, 0.0388, 0.0591, 0.0481, 0.0536, 0.0536, 0.0536, 0.0721, 0.0591, 0.0869, 0.1091, 0.1091, 0.1405, 0.1312, 0.1017, 0.1017, 0.0887, 0.0499, 0.0573, 0.0776, 0.0665, 0.0961, 0.0758, 0.0702, 0.085, 0.085, 0.1442, 0.1238, 0.1386, 0.1312, 0.1109, 0.0998, 0.0628, 0.0702, 0.0832, 0.061, 0.0776, 0.0776, 0.0647, 0.0536, 0.0333, 0.0203, -0.0018, -0.0129, 0.0129, -0.0018, 0.0055, 0.0166, 0.0333, 0.0407, 0.0425, 0.061, 0.0665, 0.0813, 0.0887, 0.0998, 0.0998, 0.0943, 0.0813, 0.0647, 0.0591, 0.0536, 0.0092, -0.0203, -0.0388, -0.0203, -0.0129, -0.0222, -0.024, -0.024, -0.0277, -0.0277, -0.0277, 0.0055, 0.0055, 0.024, 0.0277, 0.0425, 0.0425]}
{"unit_id": "orig:1945086380583084259", "ticker_idx": 2, "ts": "2025-07-15T11:41:45+00:00", "symbol": "Delta Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "liquid cooling system providers like Delta are also expected to participate in IRHX deployment", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley sees AVC, Quanta, and Delta as key beneficiaries of AWS IRHX liquid cooling adoption for high-density AI racks.", "resolved_tickers": ["2308.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Delta Electronics Taiwan 2308.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Liquid Cooling: AWS Launches Proprietary IRHX Liquid Cooling Solution, Benefiting Upstream Component Supply Chain\n\nMorgan Stanley (S. Shih, 25/07/11)\n\nMorgan Stanley points out that AWS has launched its self-developed IRHX liquid cooling solution to support high-density AI server racks, indicating an accelerating adoption of liquid cooling architecture during a transition period. Liquid cooling component manufacturers and system integrators in the supply chain are expected to be the first to benefit.\n\nAWS Launches IRHX Liquid Cooling Solution, Suitable for High-Density AI Racks Like GB200 NVL72\n\nAWS has introduced its proprietary In-Row Heat Exchanger (IRHX) liquid cooling solution, which can be inserted into existing or new data centers to support high-power AI racks. The IRHX consists of a manifold, pump units, and dry coolers, featuring modular scalability to dynamically increase or decrease dry cooler units based on thermal load.\n\nIRHX Seen as Transitional Side-Car CDU Solution, Balancing Space and Efficiency But Requiring Attention to Water Leak Risk\n\nIRHX is positioned as a transitional side-car CDU design before the widespread adoption of liquid-to-liquid solutions, capable of supporting 4 to 8 racks and saving data center space. However, its highly integrated nature may pose concerns such as water leakage risks or decreased cooling efficiency, requiring further observation of actual deployment effects.\n\nLiquid Cooling Component Manufacturers to Benefit First, AVC Expected to Stand Out\n\nAs AWS does not have liquid cooling component manufacturing capabilities, relevant parts will depend on external suppliers. Morgan Stanley notes that AVC, with its expertise in cold plates, pump units, and dry coolers, and its close relationship with AWS, is expected to be in a favorable position during this adoption phase.\n\nODMs and System Integrators Play Key Roles, Quanta and Delta are Major Beneficiaries\n\nIRHX is expected to be integrated by server ODM manufacturers, with Quanta being the primary ODM partner for AWS GB200 NVL72. Additionally, if AWS prefers partners with accumulated liquid cooling technology and firmware capabilities for integration, liquid cooling system providers like Delta are also expected to participate in IRHX deployment.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04251012145748989, "ret_signed_m1d": -0.04251012145748989, "alpha_spy_m1d": -0.046801626827461584, "alpha_c_m1d": -0.033708515203554446, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.008801606253935446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012145748987854255, "ret_signed_p1d": 0.012145748987854255, "alpha_spy_p1d": 0.008802557869339145, "alpha_c_p1d": 0.009043968089444965, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00310178089840929, "ret_p1w": 0.03846153846153855, "ret_signed_p1w": 0.03846153846153855, "alpha_spy_p1w": 0.027660155040723167, "alpha_c_p1w": 0.034971975547946066, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.003489562913592481, "ret_p1m": 0.3663967611336032, "ret_signed_p1m": 0.3663967611336032, "alpha_spy_p1m": 0.32982942721053754, "alpha_c_p1m": 0.3260727778139325, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.040323983319670686, "ret_p3m": 1.022267206477733, "ret_signed_p3m": 1.022267206477733, "alpha_spy_p3m": 0.9697220883371689, "alpha_c_p3m": 0.9414723959135243, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.08079481056420867, "ret_p6m": 1.0850202429149798, "ret_signed_p6m": 1.0850202429149798, "alpha_spy_p6m": 0.9702629396337927, "alpha_c_p6m": 0.945546433645754, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.1394738092692258, "price_path": [-0.3725, -0.3645, -0.3795, -0.3834, -0.3447, -0.3506, -0.3427, -0.3228, -0.3238, -0.3367, -0.3367, -0.2959, -0.288, -0.286, -0.277, -0.2601, -0.2561, -0.2641, -0.2681, -0.2601, -0.2631, -0.2621, -0.278, -0.2731, -0.2472, -0.2482, -0.2571, -0.2631, -0.2691, -0.2641, -0.2561, -0.2561, -0.2691, -0.2452, -0.2263, -0.2084, -0.2124, -0.2094, -0.2005, -0.2005, -0.2054, -0.2084, -0.2034, -0.2004, -0.1893, -0.1822, -0.1903, -0.1903, -0.1559, -0.1427, -0.1498, -0.1417, -0.164, -0.1336, -0.1306, -0.0709, -0.0749, -0.0891, -0.086, -0.087, -0.0233, -0.0223, -0.0425, 0.0, 0.0121, 0.0101, 0.0304, 0.0425, 0.0385, 0.0526, 0.0486, 0.0466, 0.0567, 0.0628, 0.0729, 0.1478, 0.1478, 0.2004, 0.2794, 0.2571, 0.2611, 0.2895, 0.2915, 0.2996, 0.3664, 0.3563, 0.3887, 0.3927, 0.3381, 0.2652, 0.3117, 0.3441, 0.3684, 0.3866, 0.4575, 0.4352, 0.4393, 0.3927, 0.3866, 0.3765, 0.3988, 0.4696, 0.4696, 0.587, 0.6943, 0.6822, 0.6943, 0.6923, 0.7024, 0.6923, 0.8057, 0.7976, 0.836, 0.8077, 0.8036, 0.7895, 0.7166, 0.7166, 0.7287, 0.7632, 0.8097, 0.9069, 0.9069, 0.998, 0.9858, 1.0223, 1.0223, 1.085, 1.0081, 1.0547, 1.0749, 1.0223, 1.0547, 1.0648, 1.0648, 1.0445, 1.0445, 1.166, 1.1761, 1.1457, 1.0445, 1.0142, 0.9899, 0.9737, 0.9575, 0.9798, 0.9534, 0.9879, 0.9676, 0.9575, 0.9393, 0.8664, 0.8543, 0.8178, 0.8097, 0.9312, 0.8117, 0.8077, 0.8421, 0.913, 0.9069, 0.8866, 0.9696, 1.002, 1.0182, 1.0223, 0.9818, 0.9818, 0.9656, 0.9595, 0.913, 0.8988, 0.8664, 0.8239, 0.8016, 0.8036, 0.8441, 0.9291, 0.9089, 0.9332, 0.9332, 0.9352, 0.9474, 0.9453, 0.9494, 0.9494, 1.0142, 1.0445, 1.1255, 1.085]}
{"unit_id": "orig:1973710149941612905", "ticker_idx": 0, "ts": "2025-10-02T11:22:24+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we raise our target price (TP) from $200 to $210 and see the October 27–29 GTC Washington event as a potential positive catalyst for the stock. We reiterate our Buy rating.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates Buy on NVDA, raises PT $200→$210 on upgraded AI infrastructure outlook; Oct GTC event seen as positive catalyst.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi) NVIDIA: Mid-Quarter Update – Target Price Raised to $210 on Upgraded AI Infrastructure Outlook\n\nCiti’s View\nWe spoke with NVIDIA and provide below a brief note on key investor focus points. While awaiting further details on the recent $100B OpenAI investment, we have become incrementally more constructive on NVIDIA’s roadmap and competitive positioning following the Rubin CPX GPU announcement. We forecast October/January quarter revenues of $54B/$62B, and we revise our 2026/2027 estimates by +1%/+10% in line with Citi’s updated AI infrastructure spending outlook and funding flow assumptions (see related report). Applying a consistent 30x P/E multiple to our revised 2026 EPS estimate of about $7, we raise our target price (TP) from $200 to $210 and see the October 27–29 GTC Washington event as a potential positive catalyst for the stock. We reiterate our Buy rating.\n\nGPU vs. XPU — Management reiterated that its view on the debate between commercial GPUs and ASICs (application-specific chips) at Google and elsewhere has not changed, highlighting that the outlook remains solid and execution and roadmap confidence remain intact.\n\nRubin CPX Announcement — Management clarified that the one-year product cadence is unchanged and the CPX launch does not alter this. It will follow Vera Rubin, which is expected toward the end of 2026. CPX will be available as a standalone GPU as well as in configurations alongside Vera Rubin devices, designed to handle large-context inference. It is positioned as a product that augments the existing roadmap.\n\nOpenAI Investment — This is separate from product shipments and is milestone-based, tied to NVIDIA acquiring equity if OpenAI proceeds with building gigawatt-scale capacity. As the company has noted previously, historically 1 GW of capacity requires about $50B in total capex. We believe NVIDIA’s highly competitive products, combined with rising user counts and compute consumption per user, are why OpenAI turned to NVIDIA for support. According to NVIDIA, OpenAI is facing difficulties securing data centers, land, and power and thus requires assistance. NVIDIA emphasized there is no preferential treatment or pricing; allocations are strictly determined by purchase orders (POs) and data center readiness. While each product has a single set price, if NVIDIA were to transact directly with OpenAI (which currently procures through partners), the economic terms for OpenAI could differ.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008735200039813429, "ret_signed_m1d": -0.008735200039813429, "alpha_spy_m1d": -0.007584813350471675, "alpha_c_m1d": 0.004716794282241588, "bench_spy_m1d": -0.0011503866893417536, "bench_c_m1d": -0.013451994322055016, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006723523591067093, "ret_signed_p1d": -0.006723523591067093, "alpha_spy_p1d": -0.0067087562288454805, "alpha_c_p1d": -0.002111369878192826, "bench_spy_p1d": -1.4767362221612679e-05, "bench_c_p1d": -0.004612153712874267, "ret_p1w": 0.01948223810881089, "ret_signed_p1w": 0.01948223810881089, "alpha_spy_p1w": 0.01658324897603025, "alpha_c_p1w": -0.0011243868119517053, "bench_spy_p1w": 0.0028989891327806383, "bench_c_p1w": 0.020606624920762595, "ret_p1m": 0.0719995966952276, "ret_signed_p1m": 0.0719995966952276, "alpha_spy_p1m": 0.05281294081393306, "alpha_c_p1m": -0.0012619670855364973, "bench_spy_p1m": 0.019186655881294534, "bench_c_p1m": 0.0732615637807641, "ret_p3m": -0.007091765408472117, "ret_signed_p3m": -0.007091765408472117, "alpha_spy_p3m": -0.036708581710099164, "alpha_c_p3m": -0.0845529594204919, "bench_spy_p3m": 0.029616816301627047, "bench_c_p3m": 0.07746119401201979, "ret_p6m": -0.11303723071812166, "ret_signed_p6m": -0.11303723071812166, "alpha_spy_p6m": -0.06593851750570767, "alpha_c_p6m": -0.22294297118419537, "bench_spy_p6m": -0.04709871321241399, "bench_c_p6m": 0.10990574046607371, "price_path": [-0.1623, -0.153, -0.1377, -0.1313, -0.1269, -0.1314, -0.0964, -0.0928, -0.0842, -0.0873, -0.0928, -0.1158, -0.0959, -0.0803, -0.0815, -0.0643, -0.0709, -0.051, -0.0584, -0.0804, -0.0471, -0.0563, -0.0502, -0.043, -0.0328, -0.0362, -0.0304, -0.0387, -0.0364, -0.0447, -0.0365, -0.0702, -0.0715, -0.0737, -0.0578, -0.0481, -0.0377, -0.0386, -0.0462, -0.0779, -0.0779, -0.0959, -0.0968, -0.0913, -0.1158, -0.109, -0.096, -0.0613, -0.062, -0.0586, -0.059, -0.0742, -0.0985, -0.067, -0.0647, -0.028, -0.0554, -0.0631, -0.0593, -0.0566, -0.0373, -0.0122, -0.0087, 0.0, -0.0067, -0.0177, -0.0204, 0.0012, 0.0195, -0.0303, -0.003, -0.0469, -0.048, -0.0375, -0.03, -0.0331, -0.0409, -0.0456, -0.0356, -0.0139, 0.0138, 0.0643, 0.0961, 0.0741, 0.072, 0.0952, 0.0519, 0.0335, -0.0043, -0.0039, 0.0538, 0.0226, 0.026, -0.0107, 0.0068, -0.0121, -0.0399, -0.0125, -0.0437, -0.053, -0.0336, -0.0586, -0.0457, -0.0457, -0.0629, -0.0475, -0.0393, -0.0492, -0.0291, -0.0343, -0.0176, -0.0207, -0.027, -0.0421, -0.0734, -0.0667, -0.0591, -0.095, -0.078, -0.0418, -0.0275, 0.0017, -0.0014, -0.0014, 0.0087, -0.0035, -0.0071, -0.0126, -0.0126, -0.0002, -0.004, -0.0087, 0.0012, -0.0203, -0.0213, -0.0209, -0.0163, -0.0304, -0.0097, -0.014, -0.014, -0.0572, -0.0294, -0.0214, -0.0064, -0.0128, -0.0019, 0.014, 0.0192, 0.0119, -0.0173, -0.0452, -0.0778, -0.09, -0.0184, 0.0061, -0.0018, 0.0062, -0.0103, -0.0321, -0.0321, -0.0207, -0.0048, -0.0052, 0.005, 0.0141, 0.021, 0.0354, -0.0211, -0.0619, -0.0339, -0.0467, -0.0309, -0.0293, -0.0586, -0.033, -0.0218, -0.015, -0.0303, -0.0456, -0.0299, -0.0367, -0.0448, -0.0546, -0.0856, -0.07, -0.0724, -0.0539, -0.0933, -0.113]}
{"unit_id": "quote:1957607562179932358", "ticker_idx": 0, "ts": "2025-08-19T00:56:28+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has achieved its target yield earlier than planned and is preparing for initial mass production", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung ahead of SK Hynix by up to 6 months on HBM4 mass production ramp, implying share gain.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting.\n\n1. Samsung has achieved its target yield earlier than planned and is preparing for initial mass production.\n\n2. The original expectation was that Samsung would begin HBM4 mass production at the same time as SK hynix or about 1–2 months earlier, but at the current pace, the large-scale mass production timing could be 3 to as much as 6 months ahead of SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": "Samsung, Why the Confidence in HBM4 Next Year? Higher-than-expected Yield and Ramp-up Timing\n\nWhile Samsung Electronics is solidifying its foundation for the revival of its semiconductor business, Chairman Lee Jae-yong recently returned from visiting big tech companies in San https://t.co/mVL7CpLXdQ", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.0054547226465127, "alpha_c_m1d": -0.017773815443997476, "bench_spy_m1d": 0.0054547226465127, "bench_c_m1d": 0.017773815443997476, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00714291197185335, "ret_signed_p1d": 0.00714291197185335, "alpha_spy_p1d": 0.0097999956025, "alpha_c_p1d": 0.013908524026322144, "bench_spy_p1d": -0.0026570836306466505, "bench_c_p1d": -0.006765612054468795, "ret_p1w": 0.004285837489693689, "ret_signed_p1w": 0.004285837489693689, "alpha_spy_p1w": -0.004076005369385571, "alpha_c_p1w": -0.0011037198608323529, "bench_spy_p1w": 0.00836184285907926, "bench_c_p1w": 0.005389557350526042, "ret_p1m": 0.11714287584922056, "ret_signed_p1m": 0.11714287584922056, "alpha_spy_p1m": 0.0868682464041084, "alpha_c_p1m": 0.07964588293105934, "bench_spy_p1m": 0.030274629445112167, "bench_c_p1m": 0.03749699291816122, "ret_p3m": 0.3947668690437165, "ret_signed_p3m": 0.3947668690437165, "alpha_spy_p3m": 0.34165280724225444, "alpha_c_p3m": 0.29194469784981547, "bench_spy_p3m": 0.053114061801462054, "bench_c_p3m": 0.10282217119390102, "ret_p6m": 1.419543157063916, "ret_signed_p6m": 1.419543157063916, "alpha_spy_p6m": 1.3318314978263535, "alpha_c_p6m": 1.3235192746852633, "bench_spy_p6m": 0.08771165923756263, "bench_c_p6m": 0.09602388237865278, "price_path": [-0.2233, -0.2304, -0.2233, -0.2347, -0.2063, -0.2035, -0.202, -0.1935, -0.1935, -0.1793, -0.1609, -0.1609, -0.1509, -0.1594, -0.1495, -0.1552, -0.1722, -0.1878, -0.1751, -0.1509, -0.1594, -0.1552, -0.1765, -0.141, -0.1296, -0.1452, -0.1314, -0.1457, -0.14, -0.1314, -0.0886, -0.0957, -0.1186, -0.1229, -0.1371, -0.1286, -0.1057, -0.1071, -0.09, -0.0757, -0.0471, -0.0414, -0.0314, -0.0571, -0.0514, -0.0571, -0.0586, 0.0057, 0.0086, 0.0371, 0.02, -0.0157, -0.0043, -0.0014, -0.0171, 0.0071, 0.0257, 0.0143, 0.0157, 0.0271, 0.0229, 0.0229, 0.0, 0.0, 0.0071, 0.0086, 0.02, 0.0214, 0.0043, 0.0086, -0.0057, -0.0043, -0.0343, -0.0129, -0.0029, 0.0014, -0.0071, 0.0014, 0.0214, 0.0371, 0.0486, 0.0771, 0.0929, 0.1343, 0.1171, 0.1471, 0.1471, 0.1929, 0.21, 0.22, 0.23, 0.19, 0.2082, 0.2039, 0.2341, 0.2879, 0.2879, 0.2879, 0.2879, 0.2879, 0.2879, 0.3546, 0.3388, 0.3144, 0.3632, 0.4019, 0.4048, 0.4077, 0.3991, 0.4149, 0.3847, 0.4177, 0.4636, 0.4278, 0.4421, 0.4938, 0.5426, 0.5942, 0.5053, 0.4436, 0.4235, 0.4048, 0.4436, 0.4852, 0.4794, 0.4751, 0.3948, 0.4436, 0.4034, 0.3847, 0.4436, 0.3603, 0.3876, 0.4249, 0.4751, 0.4852, 0.4421, 0.4464, 0.4837, 0.4995, 0.5081, 0.5555, 0.5713, 0.5555, 0.5497, 0.5397, 0.5627, 0.5038, 0.4751, 0.5483, 0.544, 0.5253, 0.5856, 0.6, 0.5942, 0.5942, 0.6789, 0.7231, 0.7289, 0.7289, 0.7289, 0.8529, 0.9913, 1.0028, 1.0331, 1.0014, 1.0043, 1.0014, 0.9841, 1.023, 1.0749, 1.147, 1.1528, 1.0937, 1.1557, 1.196, 1.1932, 1.1932, 1.2999, 1.3417, 1.3172, 1.3143, 1.1686, 1.4152, 1.4383, 1.297, 1.2869, 1.3994, 1.3907, 1.4195]}
{"unit_id": "quote:1957607562179932358", "ticker_idx": 1, "ts": "2025-08-19T00:56:28+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the large-scale mass production timing could be 3 to as much as 6 months ahead of SK hynix", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung ahead of SK Hynix by up to 6 months on HBM4 mass production ramp, implying share gain.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting.\n\n1. Samsung has achieved its target yield earlier than planned and is preparing for initial mass production.\n\n2. The original expectation was that Samsung would begin HBM4 mass production at the same time as SK hynix or about 1–2 months earlier, but at the current pace, the large-scale mass production timing could be 3 to as much as 6 months ahead of SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": "Samsung, Why the Confidence in HBM4 Next Year? Higher-than-expected Yield and Ramp-up Timing\n\nWhile Samsung Electronics is solidifying its foundation for the revival of its semiconductor business, Chairman Lee Jae-yong recently returned from visiting big tech companies in San https://t.co/mVL7CpLXdQ", "parent_author": null, "ret_m1d": 0.017110114290354028, "ret_signed_m1d": -0.017110114290354028, "alpha_spy_m1d": -0.011655391643841329, "alpha_c_m1d": 0.003483228159278262, "bench_spy_m1d": 0.0054547226465127, "bench_c_m1d": 0.02059334244963229, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0285171355588133, "ret_signed_p1d": 0.0285171355588133, "alpha_spy_p1d": 0.02586005192816665, "alpha_c_p1d": 0.021950600837470358, "bench_spy_p1d": -0.0026570836306466505, "bench_c_p1d": -0.006566534721342943, "ret_p1w": -0.005703450975324653, "ret_signed_p1w": 0.005703450975324653, "alpha_spy_p1w": 0.014065293834403914, "alpha_c_p1w": 0.025918649093801482, "bench_spy_p1w": 0.00836184285907926, "bench_c_p1w": 0.02021519811847683, "ret_p1m": 0.2698924117273418, "ret_signed_p1m": -0.2698924117273418, "alpha_spy_p1m": -0.23961778228222963, "alpha_c_p1m": -0.22045453964116057, "bench_spy_p1m": 0.030274629445112167, "bench_c_p1m": 0.04943787208618122, "ret_p3m": 1.1323529227523972, "ret_signed_p3m": -1.1323529227523972, "alpha_spy_p3m": -1.0792388609509351, "alpha_c_p3m": -0.9463247506313694, "bench_spy_p3m": 0.053114061801462054, "bench_c_p3m": 0.18602817212102774, "ret_p6m": 2.277030241755184, "ret_signed_p6m": -2.277030241755184, "alpha_spy_p6m": -2.189318582517622, "alpha_c_p6m": -1.8465952747394103, "bench_spy_p6m": 0.08771165923756263, "bench_c_p6m": 0.4304349670157739, "price_path": [-0.2527, -0.2409, -0.2295, -0.2314, -0.2106, -0.1939, -0.2224, -0.211, -0.211, -0.173, -0.1464, -0.1464, -0.1293, -0.1236, -0.0875, -0.1046, -0.1046, -0.057, -0.0532, -0.0627, -0.0646, -0.0228, -0.0133, 0.0589, 0.0875, 0.1141, 0.0798, 0.1103, 0.0856, 0.0608, 0.0589, 0.0285, 0.0304, 0.0722, 0.0684, 0.1293, 0.1198, 0.1407, 0.135, 0.1255, 0.0247, 0.0228, 0.0361, 0.0209, 0.0228, 0.0247, 0.0114, -0.0038, -0.0019, 0.0019, 0.0399, -0.019, -0.019, 0.0019, -0.0171, -0.0038, -0.0247, 0.0152, 0.0228, 0.057, 0.0513, 0.0513, 0.0171, 0.0, -0.0285, -0.0684, -0.0456, -0.0133, -0.0057, -0.0114, 0.0224, 0.0243, -0.0252, -0.0081, -0.0005, 0.011, 0.0414, 0.0548, 0.0966, 0.1576, 0.169, 0.2509, 0.2604, 0.3251, 0.2699, 0.3537, 0.3537, 0.3365, 0.3746, 0.3613, 0.3575, 0.2813, 0.3289, 0.3232, 0.3708, 0.506, 0.506, 0.506, 0.506, 0.506, 0.506, 0.6297, 0.5802, 0.5669, 0.6088, 0.723, 0.7725, 0.8487, 0.8239, 0.8334, 0.822, 0.942, 1.0372, 0.9838, 1.1247, 1.1628, 1.1285, 1.3608, 1.2314, 1.2047, 1.258, 1.2085, 1.3075, 1.357, 1.3494, 1.3304, 1.1324, 1.3075, 1.1704, 1.14, 1.1742, 0.9838, 0.98, 0.9762, 0.9953, 1.0729, 1.0196, 1.05, 1.1263, 1.1034, 1.0653, 1.0729, 1.1987, 1.1567, 1.2368, 1.1529, 1.1758, 1.111, 1.0196, 1.0996, 1.1034, 1.0843, 1.2101, 1.2253, 1.2406, 1.2406, 1.2825, 1.4387, 1.4806, 1.4806, 1.4806, 1.5797, 1.6521, 1.7664, 1.8274, 1.8807, 1.835, 1.8541, 1.8121, 1.8274, 1.8541, 1.8807, 1.9112, 1.8312, 1.8198, 1.8769, 1.9227, 1.8045, 2.0484, 2.2046, 2.2808, 2.4637, 2.1627, 2.4561, 2.4295, 2.2084, 2.197, 2.3799, 2.338, 2.277]}
{"unit_id": "thread:1943832893073371630", "ticker_idx": 0, "ts": "2025-07-12T00:40:51+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$4 trillion. This number is not a myth, nor a bubble.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Argues NVDA's $4T market cap is justified by CUDA moat, upward integration, and Sovereign AI tailwind; bullish on continued growth.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Is a $4 Trillion Market Cap Justified for NVIDIA? — WallstCN\n\nIn July 2025, history was rewritten.\n\nNVIDIA, the company founded by a man with a penchant for leather jackets, saw its market value skyrocket past the $4 trillion ceiling, leaving a host of traditional giants in its dust to become the absolute core of the global capital markets.\n\nIn an instant, applause, gasps, bubble theories, and doubts intertwined. Media headlines were dominated by Jensen Huang's quotes, the staggering wealth effect, and the grand narrative of AI devouring everything. But for every decision-maker in the midst of this industrial wave—be they investors, corporate strategists, or tech leaders—the real questions are far more important than the rise and fall of stock prices:\n\nWhat exactly underpins this vast empire? Is it the GPUs that are being frantically snapped up? When AMD, Intel, and even the major cloud vendors all claim to have their own AI chips, why does NVIDIA's \"throne\" seem so unshakable? After $4 trillion, how will its growth story continue?\n\nTo find the answer, we decided to employ an \"old-school\" but most effective method—deep conversations with the people who are truly shaping this industry. Leveraging our extensive expert network in Silicon Valley, we spoke with several anonymous experts on the front lines of the AI battlefield. Among them were a former head of AI infrastructure at a top cloud vendor, a chief architect leading large-scale model training, and a top VC partner identifying the next tech trends on the streets of Silicon Valley.\n\nNow, allow us to present these precious first-hand insights to you. This is not just a deconstruction of a company, but a deep analysis of the core driving force of an era.\n\n01. The Deepest Moat Is Hidden in Invisible Code\nWhen we asked almost every expert the same question—\"What is NVIDIA's most critical barrier to entry?\"—not a single one answered \"chip performance.\" Instead, they all pointed to a creation born nearly two decades ago: CUDA.\n\nA senior technical director who once built AI platforms at a FAANG company began our conversation with a vivid analogy:\n\n\"The biggest misconception from the outside world is still viewing NVIDIA as a hardware company. That's like thinking Coca-Cola's success lies only in its bottle. Since officially launching CUDA in 2006, Jensen Huang hasn't been selling chips; he's been 'preaching a gospel.' He built the 'Church of NVIDIA,' and CUDA is its bible.\"\n\n\"Today, when any customer buys an H100 or B200, they are not just paying the price of silicon; they are buying a 'ticket' into this church's ecosystem. It's an invisible, yet almost universally mandatory, 'ecosystem tax'.\"\n\nCUDA (Compute Unified Device Architecture), a name that sounds rather technical, is the starting point of the entire NVIDIA mythos. When GPUs were still just \"treasures\" for gamers, Jensen Huang had the foresight to invest heavily in opening up the heart of the GPU—its thousands of parallel computing cores—to general scientific and commercial computing.\n\nThis grand strategic move has been playing out for nearly 20 years.\n\nIt is not a product, but an ecosystem. CUDA is not just a programming interface; it includes a rich set of highly optimized math libraries (like cuDNN for deep neural networks, cuBLAS for linear algebra), a powerful compiler, intuitive debugging tools (like NVIDIA Nsight), and a massive developer community.\n\nIt created a perfect model of network effects. The more developers use CUDA, the more CUDA-based applications and frameworks (like TensorFlow, PyTorch) emerge. These killer apps, in turn, attract more users and developers to the CUDA ecosystem. Once this flywheel starts spinning, the gravitational pull it generates is immense.\n\nToday, there are over 4 million developers worldwide using CUDA. For any AI PhD student, their first line of model code is almost certainly run on CUDA. This has formed a powerful \"muscle memory\" that has spread from academia to industry, becoming the de facto industry standard.\n\n02. Invisible Costs, Visible Barriers\n\"If CUDA is so powerful, why can't competitors like AMD's ROCm or Intel's oneAPI just create something better to replace it?\" This was the question we posed to a chief AI architect responsible for large model training. He smiled and asked us in return:\n\n\"Do you know what the real cost of migrating a top-tier company's core AI business from the NVIDIA platform to another is? It's not the hardware cost of purchasing tens of thousands of new chips. It's a 'technical bill' so long it's despairing, with an amount that could be several, or even ten times, the hardware cost.\"\n\nWith this expert's help, we got a glimpse of the tip of this \"technical bill\" iceberg:\n\n- Code Refactoring and Migration: This is by no means a simple \"find-and-replace.\" Countless kernels, painstakingly handwritten and optimized for the low-level architecture of NVIDIA GPUs, must be almost completely rewritten for AMD or Intel chips. The underlying hardware architecture differences involved are hard for outsiders to imagine.\n\n- Performance Optimization Hell: Even if the code is successfully migrated, the new hardware won't achieve the performance of the NVIDIA platform \"out of the box.\" Engineers need to spend months, or even years, on tedious performance tuning, fixing all sorts of unexpected bugs, just to slowly \"approach\" the original efficiency. For the AI race where every second counts, this time cost is fatal.\n\n- The Toolchain Chasm: NVIDIA provides extremely mature performance analysis and debugging tools like Nsight and NVProf, which help engineers quickly pinpoint bottlenecks. Competitors' toolchains are still years behind in terms of stability, ease of use, and feature richness. The architect confessed, \"A problem that can be solved in an afternoon on NVIDIA might take a week on other platforms, and you still wouldn't know where the problem is.\"\n\n- The Talent Pool Discontinuity: The harsh reality is that the number of engineers proficient in CUDA on the market is likely hundreds or thousands of times greater than those proficient in ROCm. For companies, this means higher recruitment costs, longer training cycles, and a huge risk of project delays.\n\n- Ecosystem Inertia: Model communities like Hugging Face have the vast majority of their open-source models pre-trained and optimized for NVIDIA GPUs. When a team wants to quickly validate a new idea, the fastest path is always to \"download the model and run it on an NVIDIA GPU.\"\n\n\"To sum it up,\" the architect concluded, \"NVIDIA's moat wasn't dug by itself. It was built for them over the past fifteen years by millions of developers, with every line of code, every debugging session, and every project. To fill this moat, you don't need money; you need time, and an equally massive and loyal army of developers. As of now, no one can do it.\"\n\n03. Upward Integration: From Selling Shovels to \"Gold-Mining Factories\"\nIf CUDA is NVIDIA's \"software soul,\" then its \"hardware\" evolution strategy is equally filled with wisdom. A top VC partner with 20 years in Silicon Valley gave us a unique business perspective:\n\n\"To understand NVIDIA's business model, you can't just look at the GPU. you have to look at how its 'average transaction value' has increased step by step. This is a textbook case of 'Upward Integration.' It's essentially not selling a product, but constantly solving bigger and more valuable problems for its customers.\"\n\nThis top VC partner depicted NVIDIA's strategy as a four-stage rocket:\n\nStage 1: Selling \"Parts\" - GPU Chips. This is the starting point. From the G80 to Fermi, and to today's Blackwell architecture, NVIDIA has consistently maintained leadership in single-card performance. This is the foundation of all its business.\n\nStage 2: Selling \"Equipment\" - DGX/HGX Servers. NVIDIA quickly realized that customers didn't need 8 separate GPUs, but a \"monster\" that could make these 8 GPUs work together efficiently. Thus, it used high-speed interconnect technology NVLink and NVSwitch to tightly couple the GPUs and launched the DGX server. It was no longer selling parts, but an \"out-of-the-box AI supercomputer.\" The average transaction value leaped from thousands to hundreds of thousands of dollars.\n\nStage 3: Selling \"Production Lines\" - SuperPOD Clusters.\nWhen customers needed to train models with hundreds of billions or trillions of parameters, one DGX was not enough. NVIDIA, through the InfiniBand high-speed networking technology acquired from Mellanox, connected hundreds or thousands of DGX servers into a massive cluster and provided a full suite of software to manage it. This is the SuperPOD. It was no longer selling equipment, but a complete blueprint for an \"AI model production line.\" The average transaction value soared to tens or even hundreds of millions of dollars.\n\nStage 4: Selling \"Factories\" - Data Center-Level Solutions. \nToday, NVIDIA is moving towards its ultimate form. It partners with cloud service providers to launch DGX Cloud, allowing customers to rent a complete \"AI factory\" on demand. It even gets directly involved in the design of customer data centers. What it sells is \"AI capability\" itself.\n\nThrough this tiered strategy, NVIDIA transformed itself from a chip supplier into an indispensable \"general contractor\" providing full-stack solutions within its customers' AI strategies. Each integration solved a deeper customer pain point, while also bringing higher profit margins and stronger customer stickiness.\n\nConclusion\nThe story up to this point seems legendary enough. But for a $4 trillion empire, its ambitions are far from over. Take NVIDIA AI Enterprise (NVAIE) for example; it's like the \"Windows Operating System\" of the AI era. After purchasing NVIDIA's hardware, enterprises can subscribe to the NVAIE service in exchange for the stability, security, technical support, and performance guarantees necessary to run mission-critical operations.\n\nThis not only opens up a new, high-margin software subscription market for NVIDIA but, more importantly, transforms its relationship with customers from a one-time transaction to a long-term service partnership.\n\nAnd when this \"hardware + software + service\" full-stack capability is polished to perfection, it perfectly aligns with one of the most important new trends of the 21st century: Sovereign AI.\n\nAn expert specializing in geo-technology revealed the final chapter of the NVIDIA story for us:\n\n\"We are entering an era of 'Sovereign AI.' Every nation will realize that having its own independent AI infrastructure, its own foundational large models, and AI trained on its own national data is as much a part of 21st-century national sovereignty as having its own currency and military. And who can provide these nations with the complete toolkit to build 'Sovereign AI'? Today, there is only one answer—NVIDIA.\"\n\nThis elevates NVIDIA beyond the scope of a commercial company; its products become strategic resources in 21st-century geopolitics. This not only opens up a brand-new blue ocean market with \"nations\" as the unit of sale but also raises the certainty and irreplaceability of its business to an unprecedented level.\n\n$4 trillion. This number is not a myth, nor a bubble.\n\n---\n\nhttps://t.co/8slUiaRi2F", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005180683499361294, "ret_signed_m1d": 0.005180683499361294, "alpha_spy_m1d": 0.007085166748686311, "alpha_c_m1d": -0.0024247053453889666, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0076053888447502604, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.040409535932759866, "ret_signed_p1d": 0.040409535932759866, "alpha_spy_p1d": 0.044682702983386946, "alpha_c_p1d": 0.021203099274968107, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.01920643665779176, "ret_p1w": 0.04455419435283536, "ret_signed_p1w": 0.04455419435283536, "alpha_spy_p1w": 0.03821616236811698, "alpha_c_p1w": 0.0263641341996117, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.018190060153223664, "ret_p1m": 0.11635274151788177, "ret_signed_p1m": 0.11635274151788177, "alpha_spy_p1m": 0.0877358414456213, "alpha_c_p1m": 0.06455133674457092, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.051801404773310855, "ret_p3m": 0.17377251073301703, "ret_signed_p3m": 0.17377251073301703, "alpha_spy_p3m": 0.09661177882903127, "alpha_c_p3m": -0.036132036996605166, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.2099045477296222, "ret_p6m": 0.1413481255415343, "ret_signed_p6m": 0.1413481255415343, "alpha_spy_p6m": 0.02776489035331986, "alpha_c_p6m": -0.22142382598611499, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.3627719515276493, "price_path": [-0.3632, -0.3815, -0.3815, -0.4094, -0.3973, -0.374, -0.3514, -0.3234, -0.3373, -0.3356, -0.3362, -0.3198, -0.3022, -0.3063, -0.308, -0.2866, -0.2847, -0.2891, -0.2504, -0.2081, -0.1752, -0.1783, -0.1748, -0.1738, -0.181, -0.1967, -0.1905, -0.1998, -0.1998, -0.1742, -0.1784, -0.1517, -0.1764, -0.1627, -0.1393, -0.1351, -0.1468, -0.1363, -0.1307, -0.1226, -0.1295, -0.1162, -0.1347, -0.1181, -0.1216, -0.1133, -0.1133, -0.1232, -0.1213, -0.0986, -0.0595, -0.0552, -0.0385, -0.0371, -0.0656, -0.0416, -0.0288, -0.0288, -0.0355, -0.0248, -0.0073, 0.0002, 0.0052, 0.0, 0.0404, 0.0445, 0.0544, 0.0508, 0.0446, 0.018, 0.0409, 0.0589, 0.0575, 0.0773, 0.0697, 0.0926, 0.0841, 0.0588, 0.0971, 0.0865, 0.0936, 0.1018, 0.1135, 0.1096, 0.1164, 0.1068, 0.1094, 0.0998, 0.1093, 0.0705, 0.0691, 0.0665, 0.0848, 0.0959, 0.1079, 0.1068, 0.0981, 0.0616, 0.0616, 0.0409, 0.0399, 0.0463, 0.018, 0.0258, 0.0408, 0.0808, 0.0799, 0.0839, 0.0834, 0.0659, 0.038, 0.0742, 0.0769, 0.1192, 0.0876, 0.0787, 0.0831, 0.0861, 0.1084, 0.1373, 0.1413, 0.1513, 0.1436, 0.1309, 0.1279, 0.1527, 0.1738, 0.1164, 0.1479, 0.0973, 0.0961, 0.1082, 0.1168, 0.1132, 0.1042, 0.0989, 0.1103, 0.1353, 0.1672, 0.2253, 0.262, 0.2367, 0.2342, 0.261, 0.2111, 0.1899, 0.1464, 0.1468, 0.2133, 0.1774, 0.1813, 0.139, 0.1591, 0.1374, 0.1054, 0.1369, 0.1011, 0.0903, 0.1127, 0.0839, 0.0987, 0.0987, 0.0789, 0.0967, 0.1061, 0.0947, 0.1178, 0.1119, 0.131, 0.1275, 0.1203, 0.1029, 0.0669, 0.0746, 0.0833, 0.042, 0.0615, 0.1033, 0.1197, 0.1534, 0.1497, 0.1497, 0.1614, 0.1473, 0.1432, 0.1368, 0.1368, 0.1512, 0.1467, 0.1413]}
{"unit_id": "orig:1955807027101622533", "ticker_idx": 0, "ts": "2025-08-14T01:41:47+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS has raised its 2025-26 CoWoS demand forecast for AI GPUs and ASICs, driven by TSMC's capacity expansion and increased capital expenditure from US cloud service providers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for TSMC, Nvidia (+10/8%), AMD (+23%), and Broadcom (+51%) on cloud AI capex growth and TSMC capacity expansion to 100k wpm by end-2026.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS Demand Forecast Revised Upward: TSMC's Expansion and Cloud Provider Investment Drive AI GPU and ASIC Growth\nUBS (S. Lin, 12/08/25)\n\nUBS has raised its 2025-26 CoWoS demand forecast for AI GPUs and ASICs, driven by TSMC's capacity expansion and increased capital expenditure from US cloud service providers. It also anticipates that advanced packaging technology will accelerate its development in multiple directions.\n\nTSMC's CoWoS Capacity Expansion and Rising Nvidia Demand\nUBS expects TSMC to expand its monthly CoWoS capacity from approximately 70k wafers per month (wpm) to 100k wpm between the end of 2025 and the end of 2026. Investment in cloud AI by US cloud service providers is projected to grow by 62% in 2025 and 19% in 2026. Nvidia's CoWoS demand forecast for 2025/26 is revised upward by 10% and 8%, respectively, mainly due to higher production volumes of Blackwell GPUs and a modest incremental increase from the H20. The total number of GPUs using CoWoS packaging in 2026 is expected to increase from 6.1 million units to 6.7 million units.\n\nStrong Demand for AMD and Major ASIC Products\nAMD's 2026 CoWoS demand forecast is raised by 23%, reflecting the potential from the larger package designs of the MI400 and Venice server CPUs, with related wafer demand expected to grow by 83% year-over-year. On the ASIC front, Amazon's Trainium is projected to double its CoWoS wafer demand in 2026, driven by upgrades to Trainium 3 and 2.5; Broadcom's TPU demand is expected to increase by 51%, and Meta's MTIA project will also see a significant rise in CoWoS demand.\n\nWMCM Mass Production Plans and Advantages\nTSMC plans to mass-produce WMCM (Wafer-on-Wafer Multi-Chip Module) packaging for Apple's iPhone 18 Pro in Q4 2025, with an initial monthly capacity of about 10k wpm, potentially ramping up to 30-50k wpm. Compared to InFO, WMCM offers higher interconnect density, better thermal performance, and a thinner form factor. By packaging the application processor and DRAM side-by-side, it supports devices like foldable screens and has the potential to support future chiplet architectures.\n\nExpansion Outlook for Multiple Advanced Packaging Technologies\nIn addition to CoWoS and WMCM, UBS expects TSMC and its supply chain to accelerate the capacity expansion of other advanced packaging technologies such as SoIC (System on Integrated Chips), Panel-Level Packaging (PLP), and Co-Packaged Optics (CPO). SoIC monthly capacity is projected to double to 8-10k wpm in 2025 and increase to 20-30k wpm in 2027-28. Trial production of 310x310mm² specifications for panel-level packaging is expected in the second half of 2026. CPO is anticipated to be commercialized in 2027-28, with a scaled-up upgrade for xPU integration to be completed between 2028-30.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0018256486973173836, "ret_signed_m1d": 0.0018256486973173836, "alpha_spy_m1d": 0.0019186055103836974, "alpha_c_m1d": 0.003945368131066607, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008796783445339962, "ret_signed_p1d": -0.008796783445339962, "alpha_spy_p1d": -0.006455588803317203, "alpha_c_p1d": 0.011970302318634052, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.0567220735016154, "ret_signed_p1w": -0.0567220735016154, "alpha_spy_p1w": -0.04214732324431403, "alpha_c_p1w": -0.008828540886670844, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.07605788122566892, "ret_signed_p1m": 0.07605788122566892, "alpha_spy_p1m": 0.056738515973727255, "alpha_c_p1m": 0.06810867986692326, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.2119857483677987, "ret_signed_p3m": 0.2119857483677987, "alpha_spy_p3m": 0.1500529376286337, "alpha_c_p3m": 0.04926052691055194, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.4559969232333436, "ret_signed_p6m": 0.4559969232333436, "alpha_spy_p6m": 0.37904357037462577, "alpha_c_p6m": 0.12153823038622558, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2, -0.2002, -0.2072, -0.1889, -0.2063, -0.2063, -0.1827, -0.1891, -0.1849, -0.2008, -0.1945, -0.183, -0.1632, -0.1593, -0.1517, -0.1442, -0.1216, -0.1149, -0.1061, -0.1241, -0.1051, -0.1124, -0.1141, -0.1141, -0.1307, -0.1273, -0.0868, -0.0758, -0.0705, -0.0516, -0.0602, -0.0677, -0.0307, -0.0257, -0.0257, -0.0491, -0.0545, -0.038, -0.0466, -0.044, -0.0512, -0.0168, -0.0143, 0.0191, -0.0025, -0.0089, -0.0266, -0.0028, 0.0025, 0.0191, 0.0073, 0.0014, 0.0079, 0.0026, -0.024, -0.0083, -0.0354, -0.04, 0.0067, 0.0034, 0.0045, 0.0137, 0.0018, 0.0, -0.0088, 0.0017, -0.0344, -0.0515, -0.0567, -0.0332, -0.0224, -0.0095, -0.0071, -0.0113, -0.042, -0.042, -0.0523, -0.0399, -0.024, 0.01, 0.0257, 0.0412, 0.0807, 0.0743, 0.0761, 0.0846, 0.0908, 0.0939, 0.1182, 0.1025, 0.1348, 0.1768, 0.1684, 0.1516, 0.1379, 0.1373, 0.1625, 0.2007, 0.1992, 0.2162, 0.2587, 0.2239, 0.2676, 0.2482, 0.1682, 0.2608, 0.2318, 0.2683, 0.2481, 0.2283, 0.2392, 0.2259, 0.2025, 0.2102, 0.2278, 0.2415, 0.2551, 0.2699, 0.2621, 0.2505, 0.269, 0.224, 0.2223, 0.204, 0.1925, 0.2291, 0.212, 0.2097, 0.1746, 0.1856, 0.1739, 0.1568, 0.1754, 0.1551, 0.1449, 0.1848, 0.185, 0.2069, 0.2069, 0.2134, 0.1975, 0.2158, 0.2298, 0.2193, 0.2268, 0.2565, 0.2629, 0.2909, 0.2724, 0.2189, 0.2009, 0.1973, 0.1559, 0.1882, 0.206, 0.2241, 0.2394, 0.2471, 0.2471, 0.264, 0.256, 0.2504, 0.2683, 0.2683, 0.334, 0.345, 0.3666, 0.3301, 0.3273, 0.3507, 0.3847, 0.3824, 0.3653, 0.4259, 0.4291, 0.4291, 0.3655, 0.3611, 0.3663, 0.3976, 0.3886, 0.4121, 0.4287, 0.4172, 0.3797, 0.4247, 0.4013, 0.3595, 0.3804, 0.456]}
{"unit_id": "orig:1955807027101622533", "ticker_idx": 1, "ts": "2025-08-14T01:41:47+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia's CoWoS demand forecast for 2025/26 is revised upward by 10% and 8%, respectively, mainly due to higher production volumes of Blackwell GPUs", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for TSMC, Nvidia (+10/8%), AMD (+23%), and Broadcom (+51%) on cloud AI capex growth and TSMC capacity expansion to 100k wpm by end-2026.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS Demand Forecast Revised Upward: TSMC's Expansion and Cloud Provider Investment Drive AI GPU and ASIC Growth\nUBS (S. Lin, 12/08/25)\n\nUBS has raised its 2025-26 CoWoS demand forecast for AI GPUs and ASICs, driven by TSMC's capacity expansion and increased capital expenditure from US cloud service providers. It also anticipates that advanced packaging technology will accelerate its development in multiple directions.\n\nTSMC's CoWoS Capacity Expansion and Rising Nvidia Demand\nUBS expects TSMC to expand its monthly CoWoS capacity from approximately 70k wafers per month (wpm) to 100k wpm between the end of 2025 and the end of 2026. Investment in cloud AI by US cloud service providers is projected to grow by 62% in 2025 and 19% in 2026. Nvidia's CoWoS demand forecast for 2025/26 is revised upward by 10% and 8%, respectively, mainly due to higher production volumes of Blackwell GPUs and a modest incremental increase from the H20. The total number of GPUs using CoWoS packaging in 2026 is expected to increase from 6.1 million units to 6.7 million units.\n\nStrong Demand for AMD and Major ASIC Products\nAMD's 2026 CoWoS demand forecast is raised by 23%, reflecting the potential from the larger package designs of the MI400 and Venice server CPUs, with related wafer demand expected to grow by 83% year-over-year. On the ASIC front, Amazon's Trainium is projected to double its CoWoS wafer demand in 2026, driven by upgrades to Trainium 3 and 2.5; Broadcom's TPU demand is expected to increase by 51%, and Meta's MTIA project will also see a significant rise in CoWoS demand.\n\nWMCM Mass Production Plans and Advantages\nTSMC plans to mass-produce WMCM (Wafer-on-Wafer Multi-Chip Module) packaging for Apple's iPhone 18 Pro in Q4 2025, with an initial monthly capacity of about 10k wpm, potentially ramping up to 30-50k wpm. Compared to InFO, WMCM offers higher interconnect density, better thermal performance, and a thinner form factor. By packaging the application processor and DRAM side-by-side, it supports devices like foldable screens and has the potential to support future chiplet architectures.\n\nExpansion Outlook for Multiple Advanced Packaging Technologies\nIn addition to CoWoS and WMCM, UBS expects TSMC and its supply chain to accelerate the capacity expansion of other advanced packaging technologies such as SoIC (System on Integrated Chips), Panel-Level Packaging (PLP), and Co-Packaged Optics (CPO). SoIC monthly capacity is projected to double to 8-10k wpm in 2025 and increase to 20-30k wpm in 2027-28. Trial production of 310x310mm² specifications for panel-level packaging is expected in the second half of 2026. CPO is anticipated to be commercialized in 2027-28, with a scaled-up upgrade for xPU integration to be completed between 2028-30.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0023623948593429756, "ret_signed_m1d": -0.0023623948593429756, "alpha_spy_m1d": -0.0022694380462766617, "alpha_c_m1d": -0.00024267542559375244, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008625390714084524, "ret_signed_p1d": -0.008625390714084524, "alpha_spy_p1d": -0.006284196072061765, "alpha_c_p1d": 0.01214169504988949, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.03867708906268863, "ret_signed_p1w": -0.03867708906268863, "alpha_spy_p1w": -0.024102338805387258, "alpha_c_p1w": 0.009216443552255926, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": -0.02301926404893817, "ret_signed_p1m": -0.02301926404893817, "alpha_spy_p1m": -0.04233862930087984, "alpha_c_p1m": -0.03096846540768383, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.06126195618636232, "ret_signed_p3m": 0.06126195618636232, "alpha_spy_p3m": -0.000670854552802691, "alpha_c_p3m": -0.10146326527088445, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.01873851334129295, "ret_signed_p6m": 0.01873851334129295, "alpha_spy_p6m": -0.0582148395174249, "alpha_c_p6m": -0.3157201795058251, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2552, -0.2618, -0.276, -0.2703, -0.2788, -0.2788, -0.2556, -0.2594, -0.2354, -0.2577, -0.2453, -0.2242, -0.2204, -0.231, -0.2215, -0.2165, -0.2092, -0.2153, -0.2034, -0.22, -0.2051, -0.2082, -0.2007, -0.2007, -0.2097, -0.2079, -0.1875, -0.1522, -0.1483, -0.1333, -0.132, -0.1578, -0.1361, -0.1246, -0.1246, -0.1306, -0.121, -0.1052, -0.0985, -0.0939, -0.0986, -0.0622, -0.0585, -0.0496, -0.0528, -0.0585, -0.0824, -0.0618, -0.0455, -0.0468, -0.029, -0.0358, -0.0151, -0.0228, -0.0456, -0.0111, -0.0207, -0.0143, -0.0069, 0.0037, 0.0002, 0.0063, -0.0024, 0.0, -0.0086, -0.0001, -0.0351, -0.0364, -0.0387, -0.0221, -0.0121, -0.0014, -0.0023, -0.0102, -0.0431, -0.0431, -0.0618, -0.0626, -0.0569, -0.0824, -0.0753, -0.0619, -0.0258, -0.0266, -0.023, -0.0234, -0.0392, -0.0644, -0.0317, -0.0293, 0.0088, -0.0197, -0.0277, -0.0237, -0.021, -0.0009, 0.0251, 0.0287, 0.0378, 0.0308, 0.0194, 0.0166, 0.039, 0.058, 0.0063, 0.0347, -0.0109, -0.012, -0.0011, 0.0066, 0.0035, -0.0047, -0.0095, 0.0008, 0.0234, 0.0521, 0.1045, 0.1375, 0.1147, 0.1125, 0.1366, 0.0916, 0.0725, 0.0334, 0.0337, 0.0936, 0.0613, 0.0648, 0.0266, 0.0448, 0.0252, -0.0036, 0.0248, -0.0075, -0.0172, 0.003, -0.023, -0.0096, -0.0096, -0.0275, -0.0115, -0.003, -0.0133, 0.0076, 0.0023, 0.0195, 0.0163, 0.0098, -0.0059, -0.0383, -0.0314, -0.0235, -0.0608, -0.0432, -0.0055, 0.0093, 0.0396, 0.0363, 0.0363, 0.0469, 0.0342, 0.0304, 0.0247, 0.0247, 0.0376, 0.0336, 0.0288, 0.0391, 0.0167, 0.0157, 0.0162, 0.0209, 0.0063, 0.0277, 0.0232, 0.0232, -0.0216, 0.0073, 0.0156, 0.0312, 0.0246, 0.0358, 0.0523, 0.0577, 0.0502, 0.0198, -0.0091, -0.0429, -0.0556, 0.0187]}
{"unit_id": "orig:1955807027101622533", "ticker_idx": 2, "ts": "2025-08-14T01:41:47+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's 2026 CoWoS demand forecast is raised by 23%, reflecting the potential from the larger package designs of the MI400 and Venice server CPUs", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for TSMC, Nvidia (+10/8%), AMD (+23%), and Broadcom (+51%) on cloud AI capex growth and TSMC capacity expansion to 100k wpm by end-2026.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS Demand Forecast Revised Upward: TSMC's Expansion and Cloud Provider Investment Drive AI GPU and ASIC Growth\nUBS (S. Lin, 12/08/25)\n\nUBS has raised its 2025-26 CoWoS demand forecast for AI GPUs and ASICs, driven by TSMC's capacity expansion and increased capital expenditure from US cloud service providers. It also anticipates that advanced packaging technology will accelerate its development in multiple directions.\n\nTSMC's CoWoS Capacity Expansion and Rising Nvidia Demand\nUBS expects TSMC to expand its monthly CoWoS capacity from approximately 70k wafers per month (wpm) to 100k wpm between the end of 2025 and the end of 2026. Investment in cloud AI by US cloud service providers is projected to grow by 62% in 2025 and 19% in 2026. Nvidia's CoWoS demand forecast for 2025/26 is revised upward by 10% and 8%, respectively, mainly due to higher production volumes of Blackwell GPUs and a modest incremental increase from the H20. The total number of GPUs using CoWoS packaging in 2026 is expected to increase from 6.1 million units to 6.7 million units.\n\nStrong Demand for AMD and Major ASIC Products\nAMD's 2026 CoWoS demand forecast is raised by 23%, reflecting the potential from the larger package designs of the MI400 and Venice server CPUs, with related wafer demand expected to grow by 83% year-over-year. On the ASIC front, Amazon's Trainium is projected to double its CoWoS wafer demand in 2026, driven by upgrades to Trainium 3 and 2.5; Broadcom's TPU demand is expected to increase by 51%, and Meta's MTIA project will also see a significant rise in CoWoS demand.\n\nWMCM Mass Production Plans and Advantages\nTSMC plans to mass-produce WMCM (Wafer-on-Wafer Multi-Chip Module) packaging for Apple's iPhone 18 Pro in Q4 2025, with an initial monthly capacity of about 10k wpm, potentially ramping up to 30-50k wpm. Compared to InFO, WMCM offers higher interconnect density, better thermal performance, and a thinner form factor. By packaging the application processor and DRAM side-by-side, it supports devices like foldable screens and has the potential to support future chiplet architectures.\n\nExpansion Outlook for Multiple Advanced Packaging Technologies\nIn addition to CoWoS and WMCM, UBS expects TSMC and its supply chain to accelerate the capacity expansion of other advanced packaging technologies such as SoIC (System on Integrated Chips), Panel-Level Packaging (PLP), and Co-Packaged Optics (CPO). SoIC monthly capacity is projected to double to 8-10k wpm in 2025 and increase to 20-30k wpm in 2027-28. Trial production of 310x310mm² specifications for panel-level packaging is expected in the second half of 2026. CPO is anticipated to be commercialized in 2027-28, with a scaled-up upgrade for xPU integration to be completed between 2028-30.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019176575182234812, "ret_signed_m1d": 0.019176575182234812, "alpha_spy_m1d": 0.019269531995301126, "alpha_c_m1d": 0.021296294615984035, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01901079027036512, "ret_signed_p1d": -0.01901079027036512, "alpha_spy_p1d": -0.01666959562834236, "alpha_c_p1d": 0.0017562954936088948, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.09527488546632146, "ret_signed_p1w": -0.09527488546632146, "alpha_spy_p1w": -0.08070013520902009, "alpha_c_p1w": -0.0473813528513769, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": -0.12368051948862357, "ret_signed_p1m": -0.12368051948862357, "alpha_spy_p1m": -0.14299988474056524, "alpha_c_p1m": -0.13162972084736924, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.31262784348318995, "ret_signed_p3m": 0.31262784348318995, "alpha_spy_p3m": 0.25069503274402494, "alpha_c_p3m": 0.14990262202594318, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.15192045292505374, "ret_signed_p6m": 0.15192045292505374, "alpha_spy_p6m": 0.07496710006633589, "alpha_c_p6m": -0.1825382399220643, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.3659, -0.3727, -0.3807, -0.3882, -0.3904, -0.3904, -0.3669, -0.3763, -0.3754, -0.3881, -0.3665, -0.3517, -0.3447, -0.3607, -0.3579, -0.3273, -0.3189, -0.3305, -0.3451, -0.3581, -0.3015, -0.2976, -0.2993, -0.2993, -0.2913, -0.2839, -0.235, -0.2075, -0.206, -0.2053, -0.2158, -0.2478, -0.2345, -0.2379, -0.2379, -0.255, -0.2384, -0.2351, -0.2033, -0.1908, -0.1918, -0.14, -0.1153, -0.1135, -0.1324, -0.1324, -0.145, -0.1232, -0.1041, -0.08, -0.0403, -0.0194, -0.008, -0.0256, -0.0511, -0.023, -0.0367, -0.0985, -0.0473, -0.0453, -0.0479, -0.0332, 0.0192, 0.0, -0.019, -0.0266, -0.0796, -0.087, -0.0953, -0.0729, -0.0972, -0.0792, -0.0764, -0.0684, -0.1012, -0.1012, -0.103, -0.104, -0.1059, -0.1647, -0.1632, -0.1389, -0.1183, -0.1397, -0.1237, -0.1094, -0.1132, -0.1204, -0.1273, -0.1302, -0.1169, -0.1108, -0.1109, -0.1088, -0.1188, -0.1083, -0.1059, -0.0936, -0.062, -0.09, 0.1258, 0.1689, 0.3018, 0.287, 0.1876, 0.196, 0.2053, 0.3186, 0.2963, 0.2881, 0.3294, 0.3154, 0.2723, 0.2986, 0.3977, 0.435, 0.4259, 0.4608, 0.4083, 0.4154, 0.4349, 0.3819, 0.4166, 0.3136, 0.2906, 0.3483, 0.3126, 0.4307, 0.3703, 0.364, 0.3292, 0.2727, 0.2354, 0.1385, 0.1262, 0.1884, 0.1392, 0.184, 0.184, 0.2022, 0.2145, 0.1895, 0.2025, 0.1936, 0.2046, 0.2219, 0.2248, 0.2237, 0.2237, 0.1649, 0.1472, 0.156, 0.0948, 0.1111, 0.1795, 0.1879, 0.1876, 0.1884, 0.1884, 0.1881, 0.1915, 0.1901, 0.1835, 0.1835, 0.235, 0.2218, 0.1846, 0.1607, 0.1311, 0.1228, 0.1478, 0.2212, 0.2357, 0.2596, 0.2812, 0.2812, 0.2817, 0.3805, 0.4022, 0.4351, 0.3888, 0.3928, 0.3967, 0.3936, 0.3083, 0.361, 0.338, 0.1063, 0.0638, 0.1519]}
{"unit_id": "orig:1955807027101622533", "ticker_idx": 3, "ts": "2025-08-14T01:41:47+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom's TPU demand is expected to increase by 51%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for TSMC, Nvidia (+10/8%), AMD (+23%), and Broadcom (+51%) on cloud AI capex growth and TSMC capacity expansion to 100k wpm by end-2026.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS Demand Forecast Revised Upward: TSMC's Expansion and Cloud Provider Investment Drive AI GPU and ASIC Growth\nUBS (S. Lin, 12/08/25)\n\nUBS has raised its 2025-26 CoWoS demand forecast for AI GPUs and ASICs, driven by TSMC's capacity expansion and increased capital expenditure from US cloud service providers. It also anticipates that advanced packaging technology will accelerate its development in multiple directions.\n\nTSMC's CoWoS Capacity Expansion and Rising Nvidia Demand\nUBS expects TSMC to expand its monthly CoWoS capacity from approximately 70k wafers per month (wpm) to 100k wpm between the end of 2025 and the end of 2026. Investment in cloud AI by US cloud service providers is projected to grow by 62% in 2025 and 19% in 2026. Nvidia's CoWoS demand forecast for 2025/26 is revised upward by 10% and 8%, respectively, mainly due to higher production volumes of Blackwell GPUs and a modest incremental increase from the H20. The total number of GPUs using CoWoS packaging in 2026 is expected to increase from 6.1 million units to 6.7 million units.\n\nStrong Demand for AMD and Major ASIC Products\nAMD's 2026 CoWoS demand forecast is raised by 23%, reflecting the potential from the larger package designs of the MI400 and Venice server CPUs, with related wafer demand expected to grow by 83% year-over-year. On the ASIC front, Amazon's Trainium is projected to double its CoWoS wafer demand in 2026, driven by upgrades to Trainium 3 and 2.5; Broadcom's TPU demand is expected to increase by 51%, and Meta's MTIA project will also see a significant rise in CoWoS demand.\n\nWMCM Mass Production Plans and Advantages\nTSMC plans to mass-produce WMCM (Wafer-on-Wafer Multi-Chip Module) packaging for Apple's iPhone 18 Pro in Q4 2025, with an initial monthly capacity of about 10k wpm, potentially ramping up to 30-50k wpm. Compared to InFO, WMCM offers higher interconnect density, better thermal performance, and a thinner form factor. By packaging the application processor and DRAM side-by-side, it supports devices like foldable screens and has the potential to support future chiplet architectures.\n\nExpansion Outlook for Multiple Advanced Packaging Technologies\nIn addition to CoWoS and WMCM, UBS expects TSMC and its supply chain to accelerate the capacity expansion of other advanced packaging technologies such as SoIC (System on Integrated Chips), Panel-Level Packaging (PLP), and Co-Packaged Optics (CPO). SoIC monthly capacity is projected to double to 8-10k wpm in 2025 and increase to 20-30k wpm in 2027-28. Trial production of 310x310mm² specifications for panel-level packaging is expected in the second half of 2026. CPO is anticipated to be commercialized in 2027-28, with a scaled-up upgrade for xPU integration to be completed between 2028-30.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0068758815754321745, "ret_signed_m1d": -0.0068758815754321745, "alpha_spy_m1d": -0.006782924762365861, "alpha_c_m1d": -0.004756162141682951, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015711877061529322, "ret_signed_p1d": -0.015711877061529322, "alpha_spy_p1d": -0.013370682419506563, "alpha_c_p1d": 0.005055208702444691, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.0694984441669797, "ret_signed_p1w": -0.0694984441669797, "alpha_spy_p1w": -0.05492369390967833, "alpha_c_p1w": -0.021604911552035144, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.15628308912611133, "ret_signed_p1m": 0.15628308912611133, "alpha_spy_p1m": 0.13696372387416966, "alpha_c_p1m": 0.14833388776736567, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.13280540958304, "ret_signed_p3m": 0.13280540958304, "alpha_spy_p3m": 0.07087259884387498, "alpha_c_p3m": -0.02991981187420678, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 0.07357428980633274, "ret_signed_p6m": 0.07357428980633274, "alpha_spy_p6m": -0.0033790630523851117, "alpha_c_p6m": -0.2608844030407853, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2607, -0.2573, -0.2636, -0.261, -0.2668, -0.2668, -0.2446, -0.2325, -0.2244, -0.224, -0.2028, -0.1767, -0.1631, -0.1668, -0.2085, -0.217, -0.2158, -0.1893, -0.1792, -0.2028, -0.1919, -0.2006, -0.1946, -0.1946, -0.1968, -0.1846, -0.1525, -0.1497, -0.1319, -0.1346, -0.1143, -0.1494, -0.1328, -0.1158, -0.1158, -0.119, -0.1267, -0.1071, -0.1151, -0.1184, -0.1145, -0.0973, -0.0977, -0.0796, -0.0896, -0.074, -0.1049, -0.0885, -0.0724, -0.0676, -0.0544, -0.0444, -0.0277, -0.0563, -0.0726, -0.0434, -0.0588, -0.0307, -0.024, -0.0201, -0.0236, 0.0051, -0.0069, 0.0, -0.0157, -0.0176, -0.0524, -0.0645, -0.0695, -0.0554, -0.0546, -0.0425, -0.0353, -0.0083, -0.0445, -0.0445, -0.0417, -0.0284, -0.0165, 0.076, 0.1106, 0.0817, 0.1874, 0.1555, 0.1563, 0.1698, 0.1567, 0.1123, 0.1096, 0.1083, 0.0904, 0.0909, 0.0921, 0.0818, 0.0767, 0.0554, 0.0618, 0.073, 0.0885, 0.0891, 0.0798, 0.0828, 0.112, 0.1105, 0.0448, 0.1481, 0.1076, 0.1308, 0.1399, 0.1243, 0.1241, 0.1029, 0.0953, 0.1081, 0.1398, 0.1653, 0.2004, 0.2423, 0.2117, 0.1897, 0.1669, 0.1327, 0.1554, 0.1445, 0.1247, 0.1535, 0.1328, 0.1433, 0.0942, 0.1022, 0.1028, 0.0959, 0.1407, 0.1163, 0.095, 0.2165, 0.2392, 0.2796, 0.2796, 0.297, 0.2426, 0.2281, 0.225, 0.2264, 0.256, 0.291, 0.3077, 0.3292, 0.3079, 0.1585, 0.0937, 0.0985, 0.0493, 0.0617, 0.0955, 0.1011, 0.1265, 0.1294, 0.1294, 0.1355, 0.1267, 0.1282, 0.1161, 0.1161, 0.121, 0.1074, 0.1086, 0.1077, 0.0722, 0.1124, 0.1358, 0.1435, 0.0961, 0.1061, 0.1342, 0.1342, 0.0725, 0.0603, 0.0496, 0.0321, 0.0476, 0.0732, 0.0746, 0.0665, 0.0684, 0.0677, 0.033, -0.0066, 0.0013, 0.0736]}
{"unit_id": "orig:1945273179980882374", "ticker_idx": 0, "ts": "2025-07-16T00:04:02+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "our target prices of 74,000 KRW for Samsung Electronics", "tweet_type": "news_reaction", "timeline": "months", "summary": "Samsung Securities sets 74,000 KRW target for Samsung Elec and 340,000 KRW for SK Hynix; HBM cycle extending into 2026 with earnings re-rating.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "The HBM Party Extends Next Year.\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\n1. HBM ASP (Average Selling Price) will likely decline in 2026 (annual).\n\nNow, with new products commanding a price premium and older products starting to be discounted, the market will shift to one where prices (P) gradually decrease and quantities (Q) increase. We estimate next year's ASP to decline by 4%. This phenomenon is more natural when looking at other semiconductors. I don't think there's too much cause for concern. \n\nThe key is whether new products command a price premium. For HBM, new products quickly replace old ones, so the importance of a first-mover advantage for new products will not change.\n\n2. Is HBM3E 12-layer product overbooked?\n\nI don't think so. Production is better than expected, and while demand has been slightly pushed back, leading to the term \"overbooked,\" cloud customers' investment intentions have not diminished. There might be concerns about inventory accumulation somewhere before the GB300 launch, but I believe this will be resolved upon its release.\n\n3. Will cloud customer investment falter?\n\nIt's true that it's slowing down. We should acknowledge that we are in the latter half of the cycle. However, the critical point is when the CSP investment inflection point will occur, and I believe it's still time to enjoy the party a bit longer.\n\n4. Cycle Extension to Next Year - Stock Prices Reflect Next Year's Earnings\n\nWe do not expect a re-rating of stock valuations in the latter half of the cycle. However, as the visibility of increased AI investment and earnings growth next year improves, the market is beginning to reflect next year's earnings in stock prices. Q3 holds many events that could enhance next year's visibility. Against this backdrop, our target prices of 74,000 KRW for Samsung Electronics and 340,000 KRW for SK Hynix apply the upper end of this year's valuation to next year's earnings.\n\nThank you.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015455945066806476, "ret_signed_m1d": -0.015455945066806476, "alpha_spy_m1d": -0.012123893633051108, "alpha_c_m1d": -0.012363755463044468, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": -0.003092189603762008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03091201226382756, "ret_signed_p1d": 0.03091201226382756, "alpha_spy_p1d": 0.02479232228631001, "alpha_c_p1d": 0.021828616473980933, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.009083395789846627, "ret_p1w": 0.026275094400549426, "ret_signed_p1w": 0.026275094400549426, "alpha_spy_p1w": 0.010270972600368022, "alpha_c_p1w": 0.020167818587848885, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": 0.006107275812700541, "ret_p1m": 0.10664605760002899, "ret_signed_p1m": 0.10664605760002899, "alpha_spy_p1m": 0.07343657548745841, "alpha_c_p1m": 0.0717806367789382, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.03486542082109079, "ret_p3m": 0.4484739951797647, "ret_signed_p3m": 0.4484739951797647, "alpha_spy_p3m": 0.38333949433173875, "alpha_c_p3m": 0.3447807070362061, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1036932881435586, "ret_p6m": 1.1653323126504187, "ret_signed_p6m": 1.1653323126504187, "alpha_spy_p6m": 1.0544022452027795, "alpha_c_p6m": 1.0471348218388694, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.11819749081154929, "price_path": [-0.1505, -0.149, -0.1551, -0.1444, -0.1444, -0.1444, -0.1428, -0.1428, -0.1474, -0.1474, -0.1659, -0.1659, -0.1659, -0.1612, -0.1612, -0.1582, -0.1152, -0.1259, -0.1182, -0.1198, -0.1275, -0.1428, -0.1413, -0.1444, -0.1597, -0.1674, -0.1597, -0.172, -0.1413, -0.1382, -0.1367, -0.1275, -0.1275, -0.1121, -0.0921, -0.0921, -0.0814, -0.0906, -0.0798, -0.086, -0.1044, -0.1213, -0.1075, -0.0814, -0.0906, -0.086, -0.109, -0.0706, -0.0583, -0.0752, -0.0603, -0.0757, -0.0696, -0.0603, -0.0139, -0.0216, -0.0464, -0.051, -0.0665, -0.0572, -0.0325, -0.034, -0.0155, 0.0, 0.0309, 0.0371, 0.0479, 0.0201, 0.0263, 0.0201, 0.0185, 0.0881, 0.0912, 0.1221, 0.1036, 0.0649, 0.0773, 0.0804, 0.0634, 0.0896, 0.1097, 0.0974, 0.0989, 0.1113, 0.1066, 0.1066, 0.0819, 0.0819, 0.0896, 0.0912, 0.1036, 0.1051, 0.0866, 0.0912, 0.0757, 0.0773, 0.0448, 0.068, 0.0788, 0.0835, 0.0742, 0.0835, 0.1051, 0.1221, 0.1345, 0.1654, 0.1824, 0.2272, 0.2087, 0.2411, 0.2411, 0.2906, 0.3091, 0.3199, 0.3308, 0.2875, 0.3072, 0.3025, 0.3351, 0.3934, 0.3934, 0.3934, 0.3934, 0.3934, 0.3934, 0.4656, 0.4485, 0.4221, 0.4749, 0.5168, 0.5199, 0.523, 0.5137, 0.5308, 0.4982, 0.5339, 0.5835, 0.5447, 0.5603, 0.6161, 0.6689, 0.7248, 0.6286, 0.5618, 0.5401, 0.5199, 0.5618, 0.6068, 0.6006, 0.596, 0.509, 0.5618, 0.5183, 0.4982, 0.5618, 0.4718, 0.5013, 0.5416, 0.596, 0.6068, 0.5603, 0.5649, 0.6053, 0.6224, 0.6317, 0.6829, 0.7, 0.6829, 0.6767, 0.6658, 0.6907, 0.627, 0.596, 0.6751, 0.6705, 0.6503, 0.7155, 0.731, 0.7248, 0.7248, 0.8164, 0.8642, 0.8705, 0.8705, 0.8705, 1.0046, 1.1544, 1.1669, 1.1997, 1.1653]}
{"unit_id": "orig:1945273179980882374", "ticker_idx": 1, "ts": "2025-07-16T00:04:02+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "340,000 KRW for SK Hynix", "tweet_type": "news_reaction", "timeline": "months", "summary": "Samsung Securities sets 74,000 KRW target for Samsung Elec and 340,000 KRW for SK Hynix; HBM cycle extending into 2026 with earnings re-rating.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "The HBM Party Extends Next Year.\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\n1. HBM ASP (Average Selling Price) will likely decline in 2026 (annual).\n\nNow, with new products commanding a price premium and older products starting to be discounted, the market will shift to one where prices (P) gradually decrease and quantities (Q) increase. We estimate next year's ASP to decline by 4%. This phenomenon is more natural when looking at other semiconductors. I don't think there's too much cause for concern. \n\nThe key is whether new products command a price premium. For HBM, new products quickly replace old ones, so the importance of a first-mover advantage for new products will not change.\n\n2. Is HBM3E 12-layer product overbooked?\n\nI don't think so. Production is better than expected, and while demand has been slightly pushed back, leading to the term \"overbooked,\" cloud customers' investment intentions have not diminished. There might be concerns about inventory accumulation somewhere before the GB300 launch, but I believe this will be resolved upon its release.\n\n3. Will cloud customer investment falter?\n\nIt's true that it's slowing down. We should acknowledge that we are in the latter half of the cycle. However, the critical point is when the CSP investment inflection point will occur, and I believe it's still time to enjoy the party a bit longer.\n\n4. Cycle Extension to Next Year - Stock Prices Reflect Next Year's Earnings\n\nWe do not expect a re-rating of stock valuations in the latter half of the cycle. However, as the visibility of increased AI investment and earnings growth next year improves, the market is beginning to reflect next year's earnings in stock prices. Q3 holds many events that could enhance next year's visibility. Against this backdrop, our target prices of 74,000 KRW for Samsung Electronics and 340,000 KRW for SK Hynix apply the upper end of this year's valuation to next year's earnings.\n\nThank you.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008446041932942183, "ret_signed_m1d": 0.008446041932942183, "alpha_spy_m1d": 0.01177809336669755, "alpha_c_m1d": 0.0031916021996294397, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0895271115528894, "ret_signed_p1d": -0.0895271115528894, "alpha_spy_p1d": -0.09564680153040694, "alpha_c_p1d": -0.09792725195912133, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.09121636234567321, "ret_signed_p1w": -0.09121636234567321, "alpha_spy_p1w": -0.10722048414585461, "alpha_c_p1w": -0.08198656799085458, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": -0.06587855459331182, "ret_signed_p1m": -0.06587855459331182, "alpha_spy_p1m": -0.0990880367058824, "alpha_c_p1m": -0.10957327453377985, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.4040520603898148, "ret_signed_p3m": 0.4040520603898148, "alpha_spy_p3m": 0.3389175595417888, "alpha_c_p3m": 0.2295844066818702, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 1.5595747676352523, "ret_signed_p6m": 1.5595747676352523, "alpha_spy_p6m": 1.448644700187613, "alpha_c_p6m": 1.2454180457927178, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4098, -0.4045, -0.4139, -0.3896, -0.3987, -0.3782, -0.3862, -0.3903, -0.4014, -0.4014, -0.3728, -0.3728, -0.3728, -0.3566, -0.3583, -0.3589, -0.3424, -0.3306, -0.3053, -0.3239, -0.3104, -0.3276, -0.3188, -0.3239, -0.336, -0.3255, -0.3154, -0.3171, -0.2986, -0.2838, -0.3091, -0.299, -0.299, -0.2652, -0.2416, -0.2416, -0.2264, -0.2213, -0.1892, -0.2044, -0.2044, -0.1622, -0.1588, -0.1672, -0.1689, -0.1318, -0.1233, -0.0591, -0.0338, -0.0101, -0.0405, -0.0135, -0.0355, -0.0574, -0.0591, -0.0861, -0.0845, -0.0473, -0.0507, 0.0034, -0.0051, 0.0135, 0.0084, 0.0, -0.0895, -0.0912, -0.0794, -0.0929, -0.0912, -0.0895, -0.1014, -0.1149, -0.1132, -0.1098, -0.076, -0.1284, -0.1284, -0.1098, -0.1267, -0.1149, -0.1334, -0.098, -0.0912, -0.0608, -0.0659, -0.0659, -0.0963, -0.1115, -0.1368, -0.1723, -0.152, -0.1233, -0.1166, -0.1216, -0.0916, -0.0899, -0.1339, -0.1187, -0.1119, -0.1017, -0.0747, -0.0628, -0.0256, 0.0285, 0.0387, 0.1114, 0.1199, 0.1774, 0.1283, 0.2027, 0.2027, 0.1875, 0.2214, 0.2095, 0.2061, 0.1385, 0.1808, 0.1757, 0.218, 0.3381, 0.3381, 0.3381, 0.3381, 0.3381, 0.3381, 0.448, 0.4041, 0.3922, 0.4294, 0.5309, 0.5749, 0.6426, 0.6206, 0.629, 0.6189, 0.7255, 0.81, 0.7627, 0.8879, 0.9217, 0.8912, 1.0976, 0.9826, 0.9589, 1.0063, 0.9623, 1.0503, 1.0942, 1.0875, 1.0706, 0.8946, 1.0503, 0.9285, 0.9014, 0.9318, 0.7627, 0.7593, 0.7559, 0.7728, 0.8418, 0.7944, 0.8215, 0.8892, 0.8689, 0.835, 0.8418, 0.9535, 0.9163, 0.9874, 0.9129, 0.9332, 0.8757, 0.7944, 0.8655, 0.8689, 0.852, 0.9637, 0.9772, 0.9908, 0.9908, 1.028, 1.1668, 1.2041, 1.2041, 1.2041, 1.2921, 1.3564, 1.458, 1.5122, 1.5596]}
{"unit_id": "thread:1950699951518015553", "ticker_idx": 1, "ts": "2025-07-30T23:28:05+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the global semiconductor company Applied Materials is developing its hybrid bonder in collaboration with the Dutch equipment company Besi", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Hanmi Semiconductor and Besi seen as beneficiaries as Samsung's SEMES struggles with hybrid bonder development, opening doors for alternative suppliers.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Samsung Subsidiary SEMES Faces Difficulties in Hybrid Bonder Development… Hanmi Semiconductor to Benefit Indirectly?\n\nAs Samsung Electronics faces difficulties in developing hybrid bonders, industry attention is focused on whether it will reach out to Hanmi Semiconductor. The observation is that with Samsung's semiconductor equipment subsidiary, SEMES, making little progress in its hybrid bonder development, an opportunity may arise for Hanmi Semiconductor.\n\nThe hybrid bonder is a piece of equipment expected to be used in next-generation High Bandwidth Memory (HBM). HBM is made by vertically stacking multiple DRAMs, and existing HBM used TC bonders for this process. However, as the number of layers increases, it is projected that future HBM will eliminate the bumps between DRAMs and connect them using hybrid bonders.\n\nAccording to the semiconductor industry on the 31st, SEMES is experiencing difficulties in developing its hybrid bonder. A semiconductor industry official stated, \"I understand that SEMES is having trouble finding companies to collaborate with, and the performance itself is not very good.\"\n\nHybrid bonders require technology from various front-end and back-end process equipment, such as CMP (Chemical Mechanical Polishing), plasma activation, cleaning, bonding, and laser dicing equipment. For this reason, the global semiconductor company Applied Materials is developing its hybrid bonder in collaboration with the Dutch equipment company Besi, and Hanmi Semiconductor is conducting research with TES. Hanwha Semitech absorbed the front-end process equipment division of its affiliate, Hanwha Momentum, in 2024.\n\nIt appears that although SEMES is a comprehensive equipment company with both front-end and back-end process technologies, the performance of its hybrid bonder is not reaching a satisfactory level, and it is also struggling to find partners.\n\nBesides SEMES, Samsung Electronics is currently collaborating with the Applied Materials-Besi alliance in the hybrid bonder field, but 'high price' is pointed out as the biggest obstacle.\n\nAn official said, \"It's true that the Applied-Besi equipment is better than competitors', but the price is too high.\" They added, \"I understand that because of this, Samsung has asked various equipment companies to form clusters and present their solutions.\"\n\nSamsung Electronics' difficulties are expected to act as an opportunity for Hanmi Semiconductor. This is because Hanmi Semiconductor has been actively expanding its customer base this year in response to its main client SK hynix's vendor diversification policy.\n\nWhen SK hynix awarded a TC bonder order to Hanwha Semitech in the first half of the year, Hanmi Semiconductor increased its proportion of business with the American memory company, Micron.\n\nHanmi Semiconductor is developing a hybrid bonder in collaboration with the domestic front-end process equipment specialist TES, and plans to invest 100 billion KRW to build a hybrid bonder factory. A Hanmi Semiconductor official stated, \"Rather than targeting a specific client, we are building (the hybrid bonder) from various angles and are developing it so that all three major memory companies can use it.\"\n\nAnother semiconductor industry official commented, \"Samsung Electronics gave its subsidiary SEMES the task of developing a hybrid bonder, but it's not going well,\" adding, \"It seems there will be an opportunity for Hanmi Semiconductor.\"\n\n---\n\nhttps://t.co/KSEP3TZ6FC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019607840677466015, "ret_signed_m1d": -0.019607840677466015, "alpha_spy_m1d": -0.020868662586662312, "alpha_c_m1d": -0.008266343932618159, "bench_spy_m1d": 0.0012608219091962969, "bench_c_m1d": -0.011341496744847857, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026960796612669924, "ret_signed_p1d": -0.026960796612669924, "alpha_spy_p1d": -0.023209667102122244, "alpha_c_p1d": -0.0017126367901494355, "bench_spy_p1d": -0.0037511295105476794, "bench_c_p1d": -0.02524815982252049, "ret_p1w": -0.06209147457043673, "ret_signed_p1w": -0.06209147457043673, "alpha_spy_p1w": -0.05944351572235995, "alpha_c_p1w": -0.02955245588476285, "bench_spy_p1w": -0.0026479588480767813, "bench_c_p1w": -0.03253901868567388, "ret_p1m": -0.02165034237604868, "ret_signed_p1m": -0.02165034237604868, "alpha_spy_p1m": -0.0444412797480479, "alpha_c_p1m": -0.03049383726405719, "bench_spy_p1m": 0.02279093737199922, "bench_c_p1m": 0.008843494888008507, "ret_p3m": 0.20915034237604857, "ret_signed_p3m": 0.20915034237604857, "alpha_spy_p3m": 0.12611945689604864, "alpha_c_p3m": -0.005795923459914087, "bench_spy_p3m": 0.08303088547999993, "bench_c_p3m": 0.21494626583596266, "ret_p6m": 0.4297386440844664, "ret_signed_p6m": 0.4297386440844664, "alpha_spy_p6m": 0.3375789267817715, "alpha_c_p6m": 0.06582384377220318, "bench_spy_p6m": 0.0921597173026949, "bench_c_p6m": 0.3639148003122632, "price_path": [-0.1855, -0.1822, -0.1925, -0.1503, -0.1332, -0.1185, -0.0605, -0.0302, -0.0184, -0.0564, -0.0743, -0.0944, -0.1136, -0.1107, -0.1168, -0.1397, -0.1217, -0.1058, -0.1103, -0.0972, -0.1295, -0.1507, -0.1279, -0.1181, -0.0964, -0.0719, -0.0617, -0.0172, -0.0323, 0.0029, 0.0286, 0.0376, 0.0364, 0.0274, 0.0188, 0.0061, 0.027, 0.0531, 0.0556, 0.0605, 0.0658, 0.038, -0.002, 0.0069, -0.0037, -0.011, -0.0082, -0.0016, -0.0037, 0.0408, 0.0388, 0.0184, 0.0319, -0.02, 0.0417, 0.0453, 0.0633, 0.0278, 0.0102, 0.0376, -0.0433, 0.0033, -0.0196, 0.0, -0.027, -0.0429, -0.0327, -0.058, -0.0621, -0.02, -0.0135, -0.0069, 0.011, 0.009, 0.002, -0.0315, -0.0376, -0.0245, -0.0433, -0.0568, -0.0278, -0.0266, -0.0139, -0.0212, -0.0217, -0.0596, -0.0564, -0.1266, -0.1356, -0.1246, -0.1119, -0.0944, -0.0895, -0.0821, -0.0715, -0.0862, -0.0347, -0.0474, -0.0551, 0.0196, -0.0074, 0.0033, 0.0368, 0.0502, 0.0351, 0.0, 0.0327, 0.0364, 0.0441, 0.0911, 0.0662, 0.1989, 0.1993, 0.1957, 0.1879, 0.1462, 0.1663, 0.1479, 0.1887, 0.1855, 0.1577, 0.1863, 0.1789, 0.1315, 0.1785, 0.2014, 0.2092, 0.1977, 0.2063, 0.2014, 0.2059, 0.1912, 0.1646, 0.156, 0.1254, 0.096, 0.1148, 0.1258, 0.1058, 0.0931, 0.0735, 0.0637, 0.0388, 0.0547, 0.0625, 0.0033, 0.0163, 0.018, 0.0588, 0.0613, 0.0613, 0.06, 0.0731, 0.1107, 0.1234, 0.1401, 0.1998, 0.1708, 0.1475, 0.1185, 0.0792, 0.0821, 0.0915, 0.0564, 0.0739, 0.0682, 0.0719, 0.0776, 0.0748, 0.0748, 0.0748, 0.0792, 0.0874, 0.0927, 0.0927, 0.2181, 0.2561, 0.3145, 0.2974, 0.2345, 0.2345, 0.3248, 0.366, 0.3194, 0.4187, 0.4126, 0.3791, 0.4126, 0.422, 0.4297]}
{"unit_id": "orig:1967374008619515929", "ticker_idx": 0, "ts": "2025-09-14T23:44:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "SEC leads the GDDR7 market. It is the sole supplier for RTX5090 and expressed confidence in leading supply for Rubin CPX as well", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung leads GDDR7 supply; SK Hynix holds HBM4 first-mover pricing leverage; GP DRAM shortage likely in 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250915_SEC: CLSA Investors’ Forum 2025 – CLSA\n\n[Key Takeaways]\n\n1. Demand for general-purpose DRAM, centered on GP servers, is increasing, with expectations of GDDR benefiting from rising inference demand.\n\n2. For HBM4, the first supplier will secure a pricing advantage, but in-house base-die production enables fast quality response, strengthening competitiveness.\n\n[Contents]\n\n(1) General-purpose servers and On-Device AI driving DRAM demand\n\n- Strong demand for high-density DIMMs [96/128/256GB, 1-b node] is emerging in general servers.\n- Adoption of LPDDR5X in server platforms is accelerating.\n- DRAM demand from PCs and smartphones remains robust in both 1H25 and 2H25. This strength is driven not only by seasonality but also by the spread of On-Device AI features.\n\n(2) Expansion of GDDR7 adoption in AI inference chips\n\n- The large-context AI inference GPU Rubin CPX [equipped with 128GB GDDR7] is scheduled for launch in 2H26.\n- SEC leads the GDDR7 market. It is the sole supplier for RTX5090 and expressed confidence in leading supply for Rubin CPX as well.\n- There is also potential supply to China-bound low-cost GPU B40, raising the likelihood of a general-purpose DRAM shortage in 2026.\n\n(3) Awaiting NVIDIA HBM4 certification\n\n- SEC is confident in HBM4 for NVIDIA but acknowledged that the first certified supplier [= SK Hynix] will have pricing negotiation leverage.\n\n- No aggressive price cuts are planned for HBM3E, but with full-scale HBM4 adoption in 2026, competition among suppliers is inevitable.\n\n- Since HBM4 production costs have risen significantly, ASPs are also expected to surge to defend margins.\n\n- SEC emphasized that because it produces the HBM4 base die in-house, it can respond quickly to any quality/reliability issues.\n\n(4) Update on China NAND Fab and shareholder returns\n\n- Around 50% of China capacity has been converted to 236L.\n- Transition to 290L may face delays due to U.S. regulations; discussions are ongoing with the U.S. and Korean governments.\n- Share buyback program will conclude on October 8. Additional shareholder return policies will be discussed afterward.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014379050266663418, "ret_signed_m1d": -0.014379050266663418, "alpha_spy_m1d": -0.009083385343176764, "alpha_c_m1d": -0.005451937235376625, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.008927113031286793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03790862122232053, "ret_signed_p1d": 0.03790862122232053, "alpha_spy_p1d": 0.039285414007602215, "alpha_c_p1d": 0.04138439935144289, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": -0.0034757781291223644, "ret_p1w": 0.09150334763684276, "ret_signed_p1w": 0.09150334763684276, "alpha_spy_p1w": 0.07973350626278064, "alpha_c_p1w": 0.05600184578992917, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.035501501846913586, "ret_p1m": 0.20272799943230813, "ret_signed_p1m": 0.20272799943230813, "alpha_spy_p1m": 0.19795276460312938, "alpha_c_p1m": 0.17143938193303665, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.031288617499271476, "ret_p3m": 0.40887237642603425, "ret_signed_p3m": 0.40887237642603425, "alpha_spy_p3m": 0.3632221447517534, "alpha_c_p3m": 0.324720630975325, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.08415174545070925, "ret_p6m": 1.4791613363774125, "ret_signed_p6m": 1.4791613363774125, "alpha_spy_p6m": 1.4486670013632985, "alpha_c_p6m": 1.4536098155675308, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.025551520809881767, "price_path": [-0.2231, -0.2309, -0.227, -0.2465, -0.214, -0.2036, -0.2179, -0.2052, -0.2183, -0.2131, -0.2052, -0.166, -0.1725, -0.1935, -0.1974, -0.2105, -0.2026, -0.1817, -0.183, -0.1673, -0.1542, -0.1281, -0.1229, -0.1137, -0.1373, -0.132, -0.1373, -0.1386, -0.0797, -0.0771, -0.051, -0.0667, -0.0993, -0.0889, -0.0863, -0.1007, -0.0784, -0.0614, -0.0719, -0.0706, -0.0601, -0.0641, -0.0641, -0.085, -0.085, -0.0784, -0.0771, -0.0667, -0.0654, -0.081, -0.0771, -0.0902, -0.0889, -0.1163, -0.0967, -0.0876, -0.0837, -0.0915, -0.0837, -0.0654, -0.051, -0.0405, -0.0144, 0.0, 0.0379, 0.0222, 0.0497, 0.0497, 0.0915, 0.1072, 0.1163, 0.1255, 0.0889, 0.1056, 0.1016, 0.1292, 0.1784, 0.1784, 0.1784, 0.1784, 0.1784, 0.1784, 0.2395, 0.225, 0.2027, 0.2474, 0.2828, 0.2854, 0.2881, 0.2802, 0.2946, 0.2671, 0.2973, 0.3393, 0.3065, 0.3196, 0.3669, 0.4115, 0.4588, 0.3774, 0.3209, 0.3025, 0.2854, 0.3209, 0.359, 0.3537, 0.3498, 0.2763, 0.3209, 0.2841, 0.2671, 0.3209, 0.2447, 0.2697, 0.3038, 0.3498, 0.359, 0.3196, 0.3235, 0.3577, 0.3721, 0.38, 0.4233, 0.4378, 0.4233, 0.4181, 0.4089, 0.4299, 0.376, 0.3498, 0.4168, 0.4128, 0.3957, 0.4509, 0.464, 0.4588, 0.4588, 0.5362, 0.5767, 0.582, 0.582, 0.582, 0.6954, 0.8221, 0.8327, 0.8604, 0.8313, 0.834, 0.8313, 0.8155, 0.8511, 0.8986, 0.9646, 0.9699, 0.9158, 0.9725, 1.0095, 1.0068, 1.0068, 1.1045, 1.1427, 1.1203, 1.1176, 0.9844, 1.21, 1.2311, 1.1018, 1.0926, 1.1955, 1.1876, 1.214, 1.3565, 1.3908, 1.3908, 1.3908, 1.3908, 1.5069, 1.5082, 1.5465, 1.6388, 1.685, 1.8763, 1.8565, 1.8565, 1.5742, 1.272, 1.528, 1.4831, 1.2892, 1.4792]}
{"unit_id": "orig:1967374008619515929", "ticker_idx": 1, "ts": "2025-09-14T23:44:50+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": "GDDR", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "raising the likelihood of a general-purpose DRAM shortage in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung leads GDDR7 supply; SK Hynix holds HBM4 first-mover pricing leverage; GP DRAM shortage likely in 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250915_SEC: CLSA Investors’ Forum 2025 – CLSA\n\n[Key Takeaways]\n\n1. Demand for general-purpose DRAM, centered on GP servers, is increasing, with expectations of GDDR benefiting from rising inference demand.\n\n2. For HBM4, the first supplier will secure a pricing advantage, but in-house base-die production enables fast quality response, strengthening competitiveness.\n\n[Contents]\n\n(1) General-purpose servers and On-Device AI driving DRAM demand\n\n- Strong demand for high-density DIMMs [96/128/256GB, 1-b node] is emerging in general servers.\n- Adoption of LPDDR5X in server platforms is accelerating.\n- DRAM demand from PCs and smartphones remains robust in both 1H25 and 2H25. This strength is driven not only by seasonality but also by the spread of On-Device AI features.\n\n(2) Expansion of GDDR7 adoption in AI inference chips\n\n- The large-context AI inference GPU Rubin CPX [equipped with 128GB GDDR7] is scheduled for launch in 2H26.\n- SEC leads the GDDR7 market. It is the sole supplier for RTX5090 and expressed confidence in leading supply for Rubin CPX as well.\n- There is also potential supply to China-bound low-cost GPU B40, raising the likelihood of a general-purpose DRAM shortage in 2026.\n\n(3) Awaiting NVIDIA HBM4 certification\n\n- SEC is confident in HBM4 for NVIDIA but acknowledged that the first certified supplier [= SK Hynix] will have pricing negotiation leverage.\n\n- No aggressive price cuts are planned for HBM3E, but with full-scale HBM4 adoption in 2026, competition among suppliers is inevitable.\n\n- Since HBM4 production costs have risen significantly, ASPs are also expected to surge to defend margins.\n\n- SEC emphasized that because it produces the HBM4 base die in-house, it can respond quickly to any quality/reliability issues.\n\n(4) Update on China NAND Fab and shareholder returns\n\n- Around 50% of China capacity has been converted to 236L.\n- Transition to 290L may face delays due to U.S. regulations; discussions are ongoing with the U.S. and Korean governments.\n- Share buyback program will conclude on October 8. Additional shareholder return policies will be discussed afterward.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014379050266663418, "ret_signed_m1d": -0.014379050266663418, "alpha_spy_m1d": -0.009083385343176764, "alpha_c_m1d": -0.005451937235376625, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.008927113031286793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03790862122232053, "ret_signed_p1d": 0.03790862122232053, "alpha_spy_p1d": 0.039285414007602215, "alpha_c_p1d": 0.04138439935144289, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": -0.0034757781291223644, "ret_p1w": 0.09150334763684276, "ret_signed_p1w": 0.09150334763684276, "alpha_spy_p1w": 0.07973350626278064, "alpha_c_p1w": 0.05600184578992917, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.035501501846913586, "ret_p1m": 0.20272799943230813, "ret_signed_p1m": 0.20272799943230813, "alpha_spy_p1m": 0.19795276460312938, "alpha_c_p1m": 0.17143938193303665, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.031288617499271476, "ret_p3m": 0.40887237642603425, "ret_signed_p3m": 0.40887237642603425, "alpha_spy_p3m": 0.3632221447517534, "alpha_c_p3m": 0.324720630975325, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.08415174545070925, "ret_p6m": 1.4791613363774125, "ret_signed_p6m": 1.4791613363774125, "alpha_spy_p6m": 1.4486670013632985, "alpha_c_p6m": 1.4536098155675308, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.025551520809881767, "price_path": [-0.2231, -0.2309, -0.227, -0.2465, -0.214, -0.2036, -0.2179, -0.2052, -0.2183, -0.2131, -0.2052, -0.166, -0.1725, -0.1935, -0.1974, -0.2105, -0.2026, -0.1817, -0.183, -0.1673, -0.1542, -0.1281, -0.1229, -0.1137, -0.1373, -0.132, -0.1373, -0.1386, -0.0797, -0.0771, -0.051, -0.0667, -0.0993, -0.0889, -0.0863, -0.1007, -0.0784, -0.0614, -0.0719, -0.0706, -0.0601, -0.0641, -0.0641, -0.085, -0.085, -0.0784, -0.0771, -0.0667, -0.0654, -0.081, -0.0771, -0.0902, -0.0889, -0.1163, -0.0967, -0.0876, -0.0837, -0.0915, -0.0837, -0.0654, -0.051, -0.0405, -0.0144, 0.0, 0.0379, 0.0222, 0.0497, 0.0497, 0.0915, 0.1072, 0.1163, 0.1255, 0.0889, 0.1056, 0.1016, 0.1292, 0.1784, 0.1784, 0.1784, 0.1784, 0.1784, 0.1784, 0.2395, 0.225, 0.2027, 0.2474, 0.2828, 0.2854, 0.2881, 0.2802, 0.2946, 0.2671, 0.2973, 0.3393, 0.3065, 0.3196, 0.3669, 0.4115, 0.4588, 0.3774, 0.3209, 0.3025, 0.2854, 0.3209, 0.359, 0.3537, 0.3498, 0.2763, 0.3209, 0.2841, 0.2671, 0.3209, 0.2447, 0.2697, 0.3038, 0.3498, 0.359, 0.3196, 0.3235, 0.3577, 0.3721, 0.38, 0.4233, 0.4378, 0.4233, 0.4181, 0.4089, 0.4299, 0.376, 0.3498, 0.4168, 0.4128, 0.3957, 0.4509, 0.464, 0.4588, 0.4588, 0.5362, 0.5767, 0.582, 0.582, 0.582, 0.6954, 0.8221, 0.8327, 0.8604, 0.8313, 0.834, 0.8313, 0.8155, 0.8511, 0.8986, 0.9646, 0.9699, 0.9158, 0.9725, 1.0095, 1.0068, 1.0068, 1.1045, 1.1427, 1.1203, 1.1176, 0.9844, 1.21, 1.2311, 1.1018, 1.0926, 1.1955, 1.1876, 1.214, 1.3565, 1.3908, 1.3908, 1.3908, 1.3908, 1.5069, 1.5082, 1.5465, 1.6388, 1.685, 1.8763, 1.8565, 1.8565, 1.5742, 1.272, 1.528, 1.4831, 1.2892, 1.4792]}
{"unit_id": "orig:1967374008619515929", "ticker_idx": 2, "ts": "2025-09-14T23:44:50+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the first certified supplier [= SK Hynix] will have pricing negotiation leverage", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung leads GDDR7 supply; SK Hynix holds HBM4 first-mover pricing leverage; GP DRAM shortage likely in 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250915_SEC: CLSA Investors’ Forum 2025 – CLSA\n\n[Key Takeaways]\n\n1. Demand for general-purpose DRAM, centered on GP servers, is increasing, with expectations of GDDR benefiting from rising inference demand.\n\n2. For HBM4, the first supplier will secure a pricing advantage, but in-house base-die production enables fast quality response, strengthening competitiveness.\n\n[Contents]\n\n(1) General-purpose servers and On-Device AI driving DRAM demand\n\n- Strong demand for high-density DIMMs [96/128/256GB, 1-b node] is emerging in general servers.\n- Adoption of LPDDR5X in server platforms is accelerating.\n- DRAM demand from PCs and smartphones remains robust in both 1H25 and 2H25. This strength is driven not only by seasonality but also by the spread of On-Device AI features.\n\n(2) Expansion of GDDR7 adoption in AI inference chips\n\n- The large-context AI inference GPU Rubin CPX [equipped with 128GB GDDR7] is scheduled for launch in 2H26.\n- SEC leads the GDDR7 market. It is the sole supplier for RTX5090 and expressed confidence in leading supply for Rubin CPX as well.\n- There is also potential supply to China-bound low-cost GPU B40, raising the likelihood of a general-purpose DRAM shortage in 2026.\n\n(3) Awaiting NVIDIA HBM4 certification\n\n- SEC is confident in HBM4 for NVIDIA but acknowledged that the first certified supplier [= SK Hynix] will have pricing negotiation leverage.\n\n- No aggressive price cuts are planned for HBM3E, but with full-scale HBM4 adoption in 2026, competition among suppliers is inevitable.\n\n- Since HBM4 production costs have risen significantly, ASPs are also expected to surge to defend margins.\n\n- SEC emphasized that because it produces the HBM4 base die in-house, it can respond quickly to any quality/reliability issues.\n\n(4) Update on China NAND Fab and shareholder returns\n\n- Around 50% of China capacity has been converted to 236L.\n- Transition to 290L may face delays due to U.S. regulations; discussions are ongoing with the U.S. and Korean governments.\n- Share buyback program will conclude on October 8. Additional shareholder return policies will be discussed afterward.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007552760263550007, "ret_signed_m1d": -0.007552760263550007, "alpha_spy_m1d": -0.002257095340063353, "alpha_c_m1d": 0.0017254777492313744, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05135954607514148, "ret_signed_p1d": 0.05135954607514148, "alpha_spy_p1d": 0.05273633886042317, "alpha_c_p1d": 0.05122939023195627, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.06042302879498718, "ret_signed_p1w": 0.06042302879498718, "alpha_spy_p1w": 0.04865318742092506, "alpha_c_p1w": 0.01159017092247705, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.24320245896160975, "ret_signed_p1m": 0.24320245896160975, "alpha_spy_p1m": 0.238427224132431, "alpha_c_p1m": 0.15748447159350332, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.7081711738099534, "ret_signed_p3m": 0.7081711738099534, "alpha_spy_p3m": 0.6625209421356726, "alpha_c_p3m": 0.5007290699892415, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 1.8410994136412615, "ret_signed_p6m": 1.8410994136412615, "alpha_spy_p6m": 1.8106050786271475, "alpha_c_p6m": 1.5435563792461349, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.2564, -0.2579, -0.2247, -0.2171, -0.1598, -0.1372, -0.1161, -0.1432, -0.1191, -0.1387, -0.1583, -0.1598, -0.184, -0.1824, -0.1493, -0.1523, -0.104, -0.1116, -0.095, -0.0995, -0.107, -0.187, -0.1885, -0.1779, -0.19, -0.1885, -0.187, -0.1975, -0.2096, -0.2081, -0.2051, -0.1749, -0.2217, -0.2217, -0.2051, -0.2202, -0.2096, -0.2262, -0.1945, -0.1885, -0.1613, -0.1659, -0.1659, -0.193, -0.2066, -0.2292, -0.2609, -0.2428, -0.2171, -0.2111, -0.2156, -0.1888, -0.1873, -0.2266, -0.213, -0.2069, -0.1979, -0.1737, -0.1631, -0.1299, -0.0816, -0.0725, -0.0076, 0.0, 0.0514, 0.0076, 0.074, 0.074, 0.0604, 0.0906, 0.0801, 0.077, 0.0166, 0.0544, 0.0498, 0.0876, 0.1949, 0.1949, 0.1949, 0.1949, 0.1949, 0.1949, 0.2931, 0.2538, 0.2432, 0.2764, 0.3671, 0.4063, 0.4668, 0.4471, 0.4547, 0.4456, 0.5408, 0.6163, 0.574, 0.6858, 0.716, 0.6888, 0.8731, 0.7704, 0.7492, 0.7915, 0.7523, 0.8308, 0.8701, 0.864, 0.8489, 0.6918, 0.8308, 0.7221, 0.6979, 0.7251, 0.574, 0.571, 0.568, 0.5831, 0.6447, 0.6024, 0.6265, 0.687, 0.6689, 0.6386, 0.6447, 0.7445, 0.7112, 0.7747, 0.7082, 0.7263, 0.6749, 0.6024, 0.6658, 0.6689, 0.6538, 0.7535, 0.7656, 0.7777, 0.7777, 0.811, 0.9349, 0.9682, 0.9682, 0.9682, 1.0468, 1.1042, 1.1949, 1.2433, 1.2856, 1.2493, 1.2645, 1.2312, 1.2433, 1.2645, 1.2856, 1.3098, 1.2463, 1.2373, 1.2826, 1.3189, 1.2252, 1.4186, 1.5426, 1.6031, 1.7482, 1.5093, 1.7421, 1.721, 1.5456, 1.5366, 1.6817, 1.6484, 1.6, 1.6847, 1.6605, 1.6605, 1.6605, 1.6605, 1.7028, 1.8691, 1.8752, 2.0384, 2.0777, 2.3288, 2.2137, 2.2137, 1.8441, 1.5715, 1.8502, 1.7987, 1.5322, 1.8411]}
{"unit_id": "orig:1972818494657101931", "ticker_idx": 0, "ts": "2025-09-30T00:19:16+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "redesigning metal layers and wiring is likely to increase HBM4 costs, further narrowing the profitability gap between HBM and conventional DRAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish MU: HBM4 redesign raises costs, compresses margins, and may force resource shift away from HBM; profitability at risk.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS Comment on Micron HBM4 Issues (KIS was among the first to raise concerns about Micron’s HBM4 speed)\n\nQ. What is your view on Micron’s 11Gbps HBM4 comment?\n\nA. We see two possible scenarios regarding Micron’s statement that it has already submitted improved samples. Considering that the speed issue emerged only within the past month and a half, there would not have been sufficient time to prepare actual improved samples.\n\n    1. The first possibility is that Micron sent estimated data based on a redesigned version of the product.\n    2. The second is that Micron selectively supplied sample quantities by picking only the “good dies” running at 11Gbps or higher from its current wafers.\n\nEither way, the current wafers cannot support full-scale mass production at the higher speed, and a redesign appears unavoidable. Channel checks suggest that Micron has decided to respond by redesigning the metal layers using its in-house CMOS process.\n\nSince Micron itself mentioned achieving 11Gbps, the issue is no longer about HBM4’s overall market share but rather about profit margin differences among the three DRAM makers. In other words, profitability will hinge on how many dies per wafer can reach 11Gbps (or NVIDIA’s hurdle of 10Gbps), which will directly impact both production supply capability and gross margin.\n\nFor Micron, however, redesigning metal layers and wiring is likely to increase HBM4 costs, further narrowing the profitability gap between HBM and conventional DRAM. Given Micron’s traditional strategy of prioritizing margin improvement and FCF maximization over market share, it is possible the company may eventually shift some resources back from HBM toward conventional DRAM.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0204400235405946, "ret_signed_m1d": 0.0204400235405946, "alpha_spy_m1d": 0.01668729281809722, "alpha_c_m1d": 0.009102897376619734, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08863240251055848, "ret_signed_p1d": -0.08863240251055848, "alpha_spy_p1d": -0.08522483308462547, "alpha_c_p1d": -0.06617248297774947, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.11048462427289563, "ret_signed_p1w": -0.11048462427289563, "alpha_spy_p1w": -0.10607142473732467, "alpha_c_p1w": -0.07772932038041502, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.3553186802508512, "ret_signed_p1m": -0.3553186802508512, "alpha_spy_p1m": -0.32348045309362305, "alpha_c_p1m": -0.22586035878711486, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": 0.7031338873141986, "ret_signed_p3m": -0.7031338873141986, "alpha_spy_p3m": -0.6638505512611612, "alpha_c_p3m": -0.5786136621306297, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 1.285941508709545, "ret_signed_p6m": -1.285941508709545, "alpha_spy_p6m": -1.2943777815314468, "alpha_c_p6m": -1.0594995203427975, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.2698, -0.2698, -0.2833, -0.2564, -0.2694, -0.2642, -0.2557, -0.2911, -0.2822, -0.3041, -0.3231, -0.3163, -0.3233, -0.3472, -0.3436, -0.3322, -0.335, -0.3351, -0.3309, -0.3142, -0.3477, -0.3732, -0.3559, -0.3482, -0.3499, -0.3314, -0.2894, -0.2606, -0.2365, -0.2573, -0.2512, -0.2776, -0.2616, -0.2706, -0.2995, -0.308, -0.2967, -0.3042, -0.3037, -0.2963, -0.2709, -0.2887, -0.2887, -0.2919, -0.2905, -0.2577, -0.2149, -0.2143, -0.1917, -0.1633, -0.1001, -0.0603, -0.0571, -0.0508, -0.0438, 0.0094, -0.0274, -0.0161, -0.0054, -0.0335, -0.0627, -0.0601, -0.0204, 0.0, 0.0886, 0.0982, 0.1233, 0.142, 0.1105, 0.1754, 0.1502, 0.086, 0.1528, 0.1187, 0.1479, 0.2112, 0.2103, 0.2365, 0.2098, 0.1869, 0.2362, 0.3098, 0.3163, 0.3271, 0.3553, 0.3397, 0.3382, 0.4036, 0.3039, 0.4203, 0.4253, 0.4228, 0.5148, 0.4419, 0.4646, 0.417, 0.4761, 0.4469, 0.3665, 0.3511, 0.2043, 0.2401, 0.3392, 0.3428, 0.377, 0.377, 0.4142, 0.438, 0.4322, 0.4004, 0.3554, 0.4187, 0.4767, 0.5096, 0.5771, 0.5457, 0.4421, 0.4203, 0.3905, 0.3487, 0.4864, 0.5903, 0.6541, 0.6522, 0.7144, 0.7144, 0.7031, 0.7611, 0.7507, 0.7075, 0.7075, 0.8871, 0.8675, 1.0546, 1.0314, 0.9565, 1.0646, 1.0692, 1.0229, 0.9943, 1.014, 1.1702, 1.1702, 1.1837, 1.3279, 1.3786, 1.391, 1.3278, 1.4544, 1.6042, 1.6072, 1.4821, 1.6192, 1.5094, 1.2698, 1.2907, 1.3613, 1.2944, 1.2331, 1.455, 1.4767, 1.4629, 1.4629, 1.3918, 1.5184, 1.4969, 1.5616, 1.5186, 1.5008, 1.5666, 1.4862, 1.4671, 1.4689, 1.2715, 1.3977, 1.3754, 1.2154, 1.3292, 1.4117, 1.5049, 1.4251, 1.5494, 1.6432, 1.7622, 1.7624, 1.6579, 1.5301, 1.4191, 1.3663, 1.2859]}
{"unit_id": "orig:1965585165423665215", "ticker_idx": 0, "ts": "2025-09-10T01:16:36+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MORGAN STANLEY CUTS GB200/300 SHIPMENT FORECAST", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley's cut to GB200/300 shipment forecast; implied bearish on NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "MORGAN STANLEY CUTS GB200/300 SHIPMENT FORECAST\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.037049656235068285, "ret_signed_m1d": 0.037049656235068285, "alpha_spy_m1d": 0.034167134905184304, "alpha_c_m1d": 0.02701080315052662, "bench_spy_m1d": -0.0028825213298839802, "bench_c_m1d": -0.010038853084541666, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0008459006393353175, "ret_signed_p1d": 0.0008459006393353175, "alpha_spy_p1d": 0.009156165371120961, "alpha_c_p1d": 0.011150755041265281, "bench_spy_p1d": 0.008310264731785644, "bench_c_p1d": 0.010304854401929964, "ret_p1w": -0.039645880122221366, "ret_signed_p1w": 0.039645880122221366, "alpha_spy_p1w": 0.05033262292125029, "alpha_c_p1w": 0.05433863689820806, "bench_spy_p1w": 0.010686742799028925, "bench_c_p1w": 0.014692756775986693, "ret_p1m": 0.08600268099468078, "ret_signed_p1m": -0.08600268099468078, "alpha_spy_p1m": -0.05409468083008928, "alpha_c_p1m": 0.06152252476710718, "bench_spy_p1m": 0.0319080001645915, "bench_c_p1m": 0.14752520576178796, "ret_p3m": 0.04647152114433162, "ret_signed_p3m": -0.04647152114433162, "alpha_spy_p3m": 0.004609119992285748, "alpha_c_p3m": 0.1786390642732807, "bench_spy_p3m": 0.051080641136617366, "bench_c_p3m": 0.2251105854176123, "ret_p6m": 0.034007460518325594, "ret_signed_p6m": -0.034007460518325594, "alpha_spy_p6m": 0.016601490161992194, "alpha_c_p6m": 0.28427929561799603, "bench_spy_p6m": 0.05060895068031779, "bench_c_p6m": 0.3182867561363216, "price_path": [-0.1994, -0.1841, -0.1873, -0.1796, -0.1796, -0.1888, -0.187, -0.166, -0.1298, -0.1258, -0.1104, -0.1091, -0.1355, -0.1132, -0.1014, -0.1014, -0.1077, -0.0977, -0.0815, -0.0746, -0.07, -0.0748, -0.0374, -0.0336, -0.0244, -0.0277, -0.0336, -0.0581, -0.0369, -0.0202, -0.0216, -0.0033, -0.0103, 0.0109, 0.003, -0.0204, 0.0151, 0.0052, 0.0118, 0.0194, 0.0303, 0.0267, 0.0329, 0.024, 0.0264, 0.0176, 0.0264, -0.0095, -0.0109, -0.0133, 0.0037, 0.014, 0.025, 0.0241, 0.016, -0.0178, -0.0178, -0.0369, -0.0378, -0.032, -0.0581, -0.0509, -0.037, 0.0, -0.0008, 0.0028, 0.0024, -0.0138, -0.0396, -0.0061, -0.0037, 0.0355, 0.0063, -0.002, 0.0021, 0.0049, 0.0255, 0.0522, 0.0559, 0.0652, 0.0581, 0.0464, 0.0435, 0.0665, 0.086, 0.0329, 0.062, 0.0153, 0.0142, 0.0253, 0.0333, 0.03, 0.0217, 0.0167, 0.0273, 0.0504, 0.0799, 0.1337, 0.1676, 0.1442, 0.1419, 0.1667, 0.1205, 0.1009, 0.0607, 0.0611, 0.1225, 0.0893, 0.0929, 0.0538, 0.0725, 0.0523, 0.0228, 0.0519, 0.0187, 0.0088, 0.0295, 0.0028, 0.0166, 0.0166, -0.0018, 0.0147, 0.0233, 0.0128, 0.0342, 0.0288, 0.0465, 0.0432, 0.0365, 0.0204, -0.0129, -0.0058, 0.0023, -0.0359, -0.0179, 0.0208, 0.036, 0.0671, 0.0637, 0.0637, 0.0746, 0.0615, 0.0577, 0.0518, 0.0518, 0.0651, 0.061, 0.056, 0.0665, 0.0436, 0.0426, 0.043, 0.0479, 0.0329, 0.0549, 0.0503, 0.0503, 0.0043, 0.0339, 0.0425, 0.0584, 0.0517, 0.0632, 0.0801, 0.0857, 0.0779, 0.0468, 0.0171, -0.0176, -0.0306, 0.0457, 0.0718, 0.0633, 0.0719, 0.0543, 0.031, 0.031, 0.0432, 0.0602, 0.0597, 0.0706, 0.0803, 0.0876, 0.1029, 0.0427, -0.0007, 0.0292, 0.0155, 0.0323, 0.034]}
{"unit_id": "orig:1969672513706999866", "ticker_idx": 0, "ts": "2025-09-21T07:58:16+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reasons for a bullish outlook on ASML... I believe this benefits not only TSMC, but also equipment makers such as ASML", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish ASML (and TSMC): HPC shift to 2nm nodes refutes the bear case that HPC demand won't drive cutting-edge equipment spend.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Reasons for a bullish outlook on ASML:\n\nAhmad Khan, President of KLA’s Semiconductor Products & Solutions Division, recently stated at a Goldman Sachs event that TSMC has secured about 15 customers for its first-generation 2nm (N2) process node, of which around 10 are high-performance computing (HPC) customers. Khan noted that inspection and metrology processes are becoming increasingly important at advanced nodes when making this point.\n\nWhat stands out is that the majority of the 2nm customers are in HPC. Traditionally, TSMC’s new advanced nodes had been adopted mainly by mobile application processor (AP) customers such as Apple and Qualcomm. However, with the advent of the AI era, demand for implementing high-performance computing has surged.\n\nIn my view, this serves as a rebuttal to the long-standing argument that has weighed on equipment makers like ASML: namely, that because HPC players don’t use the most advanced nodes, even if TSMC earns money, it won’t translate into demand for cutting-edge equipment.\n\nIn other words, if HPC makers in the past felt little need to adopt the most advanced nodes, they are now starting to recognize that necessity and are acting on it.\n\nI believe this benefits not only TSMC, but also equipment makers such as ASML.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02109022638613367, "ret_signed_m1d": -0.02109022638613367, "alpha_spy_m1d": -0.016381495461561313, "alpha_c_m1d": -0.0010386686311859572, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011100103218230739, "ret_signed_p1d": 0.011100103218230739, "alpha_spy_p1d": 0.0165435780580383, "alpha_c_p1d": 0.01255894253211165, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.018130279585259368, "ret_signed_p1w": 0.018130279585259368, "alpha_spy_p1w": 0.02286901884312753, "alpha_c_p1w": 0.016609392328177597, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.09743463457121293, "ret_signed_p1m": 0.09743463457121293, "alpha_spy_p1m": 0.09076146241699479, "alpha_c_p1m": 0.025640318276833884, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.10295464172134006, "ret_signed_p3m": 0.10295464172134006, "alpha_spy_p3m": 0.08851345349811957, "alpha_c_p3m": 0.025107562820681384, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 0.476951397071127, "ret_signed_p6m": 0.476951397071127, "alpha_spy_p6m": 0.46805558315077667, "alpha_c_p6m": 0.24122204392806146, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.1457, -0.1667, -0.1605, -0.1665, -0.1781, -0.1665, -0.1671, -0.1889, -0.1739, -0.1649, -0.1638, -0.1512, -0.1587, -0.1546, -0.1314, -0.2302, -0.2002, -0.2207, -0.2323, -0.259, -0.2538, -0.2424, -0.2559, -0.2251, -0.2316, -0.2261, -0.2438, -0.2662, -0.257, -0.2636, -0.2687, -0.2468, -0.2391, -0.2267, -0.2182, -0.2099, -0.2071, -0.2147, -0.2123, -0.2078, -0.2129, -0.2174, -0.203, -0.2024, -0.1976, -0.1857, -0.193, -0.2148, -0.2155, -0.2388, -0.2288, -0.2013, -0.1857, -0.1655, -0.1574, -0.1611, -0.1539, -0.1492, -0.1004, -0.0961, -0.091, -0.0207, -0.0211, 0.0, 0.0111, -0.0035, 0.0011, 0.0025, 0.0181, 0.0213, 0.0377, 0.0824, 0.0855, 0.1067, 0.0789, 0.0506, 0.0481, 0.0038, 0.0409, 0.0442, 0.0767, 0.0821, 0.08, 0.1088, 0.0974, 0.0731, 0.0948, 0.1031, 0.12, 0.1211, 0.1359, 0.1583, 0.1343, 0.1447, 0.1342, 0.1201, 0.1062, 0.0759, 0.0955, 0.0958, 0.1018, 0.0891, 0.08, 0.0828, 0.0695, 0.0954, 0.0994, 0.0304, 0.0611, 0.0629, 0.1235, 0.1085, 0.1162, 0.145, 0.1603, 0.1903, 0.1828, 0.1757, 0.1901, 0.1773, 0.1688, 0.1625, 0.1432, 0.1503, 0.1226, 0.0799, 0.103, 0.114, 0.107, 0.1159, 0.1107, 0.1107, 0.1107, 0.1204, 0.1347, 0.1384, 0.1384, 0.2186, 0.3013, 0.3109, 0.2995, 0.2516, 0.3363, 0.3423, 0.3623, 0.3396, 0.4201, 0.4421, 0.3843, 0.4085, 0.4268, 0.4535, 0.4557, 0.4555, 0.5044, 0.4757, 0.4727, 0.5019, 0.5133, 0.4708, 0.4092, 0.4204, 0.475, 0.4906, 0.476, 0.4943, 0.4596, 0.4727, 0.4784, 0.4836, 0.54, 0.5319, 0.5534, 0.5455, 0.5631, 0.594, 0.5247, 0.5259, 0.4975, 0.4374, 0.4844, 0.4673, 0.4191, 0.4198, 0.4846, 0.4831, 0.4604, 0.4594, 0.4814, 0.477]}
{"unit_id": "orig:1969672513706999866", "ticker_idx": 1, "ts": "2025-09-21T07:58:16+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I believe this benefits not only TSMC", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish ASML (and TSMC): HPC shift to 2nm nodes refutes the bear case that HPC demand won't drive cutting-edge equipment spend.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Reasons for a bullish outlook on ASML:\n\nAhmad Khan, President of KLA’s Semiconductor Products & Solutions Division, recently stated at a Goldman Sachs event that TSMC has secured about 15 customers for its first-generation 2nm (N2) process node, of which around 10 are high-performance computing (HPC) customers. Khan noted that inspection and metrology processes are becoming increasingly important at advanced nodes when making this point.\n\nWhat stands out is that the majority of the 2nm customers are in HPC. Traditionally, TSMC’s new advanced nodes had been adopted mainly by mobile application processor (AP) customers such as Apple and Qualcomm. However, with the advent of the AI era, demand for implementing high-performance computing has surged.\n\nIn my view, this serves as a rebuttal to the long-standing argument that has weighed on equipment makers like ASML: namely, that because HPC players don’t use the most advanced nodes, even if TSMC earns money, it won’t translate into demand for cutting-edge equipment.\n\nIn other words, if HPC makers in the past felt little need to adopt the most advanced nodes, they are now starting to recognize that necessity and are acting on it.\n\nI believe this benefits not only TSMC, but also equipment makers such as ASML.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023165933043046705, "ret_signed_m1d": -0.023165933043046705, "alpha_spy_m1d": -0.018457202118474347, "alpha_c_m1d": -0.0031143752880989917, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03474908933571563, "ret_signed_p1d": 0.03474908933571563, "alpha_spy_p1d": 0.04019256417552319, "alpha_c_p1d": 0.036207928649596544, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.003860988840507895, "ret_signed_p1w": 0.003860988840507895, "alpha_spy_p1w": 0.008599728098376058, "alpha_c_p1w": 0.002340101583426124, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.14285725129779747, "ret_signed_p1m": 0.14285725129779747, "alpha_spy_p1m": 0.13618407914357933, "alpha_c_p1m": 0.07106293500341843, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.1079279742594419, "ret_signed_p3m": 0.1079279742594419, "alpha_spy_p3m": 0.0934867860362214, "alpha_c_p3m": 0.030080895358783222, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 0.45355588428135096, "ret_signed_p6m": 0.45355588428135096, "alpha_spy_p6m": 0.44466007036100064, "alpha_c_p6m": 0.21782653113828543, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.177, -0.1732, -0.1693, -0.1847, -0.1655, -0.1655, -0.1617, -0.1655, -0.1693, -0.1693, -0.1617, -0.154, -0.154, -0.1578, -0.1463, -0.1309, -0.1309, -0.1117, -0.1155, -0.1309, -0.1194, -0.1194, -0.1194, -0.1194, -0.127, -0.1117, -0.1078, -0.1078, -0.127, -0.1155, -0.1347, -0.0924, -0.0963, -0.0924, -0.0924, -0.0771, -0.0963, -0.0924, -0.0924, -0.0886, -0.127, -0.1155, -0.127, -0.1001, -0.0963, -0.0847, -0.1078, -0.1078, -0.104, -0.1078, -0.1078, -0.1078, -0.0924, -0.0924, -0.0771, -0.0578, -0.0463, -0.0309, -0.0347, -0.0116, -0.0232, -0.0077, -0.0232, 0.0, 0.0347, 0.0347, 0.0193, 0.0039, 0.0039, 0.0077, 0.0232, 0.0541, 0.0811, 0.0811, 0.1081, 0.0927, 0.112, 0.112, 0.0927, 0.1004, 0.1313, 0.1467, 0.1197, 0.1429, 0.1429, 0.1274, 0.1197, 0.1197, 0.1429, 0.139, 0.1622, 0.1622, 0.1583, 0.166, 0.1622, 0.1274, 0.1313, 0.1274, 0.139, 0.1313, 0.139, 0.1274, 0.1042, 0.1158, 0.0849, 0.0772, 0.1236, 0.0695, 0.0618, 0.0927, 0.112, 0.1081, 0.112, 0.0888, 0.1042, 0.1197, 0.1158, 0.1274, 0.1544, 0.1429, 0.1622, 0.1389, 0.1467, 0.1234, 0.1118, 0.1079, 0.1079, 0.1079, 0.135, 0.1544, 0.1583, 0.1583, 0.1699, 0.1854, 0.1777, 0.2009, 0.2009, 0.228, 0.2939, 0.321, 0.2977, 0.3055, 0.3016, 0.3094, 0.3249, 0.3249, 0.3094, 0.3481, 0.3636, 0.3752, 0.3481, 0.3636, 0.3714, 0.3597, 0.3791, 0.4101, 0.3985, 0.3752, 0.3675, 0.3946, 0.383, 0.3675, 0.3791, 0.4062, 0.4566, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4837, 0.4721, 0.5224, 0.5612, 0.5457, 0.5457, 0.5302, 0.4992, 0.445, 0.4721, 0.4643, 0.4023, 0.4333, 0.5031, 0.4605, 0.445, 0.4295, 0.4536]}
{"unit_id": "orig:1965934136394396106", "ticker_idx": 0, "ts": "2025-09-11T00:23:18+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Estimation on 2026 ASIC volume and revenue contribution", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Fubon's 2026 ASIC volume and revenue contribution estimates for Broadcom; implied bullish.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Estimation on 2026 ASIC volume and revenue contribution\n\nSource: Fubon\n\n$AVGO https://t.co/mtUBrSGxip", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.027639561620509623, "ret_signed_m1d": 0.027639561620509623, "alpha_spy_m1d": 0.035881335033375095, "alpha_c_m1d": 0.03783930910962785, "bench_spy_m1d": -0.008241773412865472, "bench_c_m1d": -0.010199747489118227, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0006673122159186473, "ret_signed_p1d": 0.0006673122159186473, "alpha_spy_p1d": 0.0010019251277464392, "alpha_c_p1d": -0.0006159909930443241, "bench_spy_p1d": -0.00033461291182779185, "bench_c_p1d": 0.0012833032089629715, "ret_p1w": -0.039707423833466704, "ret_signed_p1w": -0.039707423833466704, "alpha_spy_p1w": -0.04674777309634015, "alpha_c_p1w": -0.08261211482491193, "bench_spy_p1w": 0.0070403492628734465, "bench_c_p1w": 0.04290469099144523, "ret_p1m": -0.09577564654352078, "ret_signed_p1m": -0.09577564654352078, "alpha_spy_p1m": -0.09151852792291593, "alpha_c_p1m": -0.16621953488878882, "bench_spy_p1m": -0.0042571186206048495, "bench_c_p1m": 0.07044388834526805, "ret_p3m": 0.13168009910939116, "ret_signed_p3m": 0.13168009910939116, "alpha_spy_p3m": 0.09016198419592292, "alpha_c_p3m": -0.08241528594452308, "bench_spy_p3m": 0.04151811491346824, "bench_c_p3m": 0.21409538505391423, "ret_p6m": -0.0777197261382756, "ret_signed_p6m": -0.0777197261382756, "alpha_spy_p6m": -0.106012815600103, "alpha_c_p6m": -0.33374632086233924, "bench_spy_p6m": 0.028293089461827403, "bench_c_p6m": 0.25602659472406364, "price_path": [-0.3006, -0.3082, -0.303, -0.303, -0.3049, -0.2944, -0.2666, -0.2641, -0.2488, -0.251, -0.2335, -0.2639, -0.2495, -0.2348, -0.2348, -0.2376, -0.2442, -0.2273, -0.2342, -0.237, -0.2337, -0.2188, -0.2192, -0.2035, -0.2121, -0.1986, -0.2253, -0.2112, -0.1972, -0.1931, -0.1817, -0.173, -0.1585, -0.1833, -0.1974, -0.1721, -0.1855, -0.1612, -0.1554, -0.152, -0.155, -0.1301, -0.1405, -0.1346, -0.1482, -0.1498, -0.18, -0.1904, -0.1947, -0.1825, -0.1819, -0.1713, -0.1651, -0.1418, -0.1731, -0.1731, -0.1707, -0.1592, -0.1488, -0.0688, -0.0389, -0.0638, 0.0276, 0.0, 0.0007, 0.0124, 0.001, -0.0374, -0.0397, -0.0408, -0.0563, -0.0559, -0.0549, -0.0638, -0.0682, -0.0867, -0.0811, -0.0714, -0.058, -0.0575, -0.0655, -0.063, -0.0376, -0.039, -0.0958, -0.0064, -0.0415, -0.0214, -0.0136, -0.027, -0.0272, -0.0456, -0.0521, -0.041, -0.0136, 0.0085, 0.0389, 0.0751, 0.0486, 0.0296, 0.0098, -0.0197, -0.0001, -0.0095, -0.0267, -0.0017, -0.0197, -0.0106, -0.053, -0.0461, -0.0456, -0.0516, -0.0128, -0.034, -0.0524, 0.0528, 0.0725, 0.1074, 0.1074, 0.1224, 0.0754, 0.0628, 0.0602, 0.0613, 0.087, 0.1172, 0.1317, 0.1503, 0.1319, 0.0025, -0.0535, -0.0493, -0.0919, -0.0812, -0.052, -0.0471, -0.0251, -0.0226, -0.0226, -0.0173, -0.0249, -0.0237, -0.0341, -0.0341, -0.0299, -0.0416, -0.0406, -0.0414, -0.0721, -0.0373, -0.0171, -0.0104, -0.0515, -0.0427, -0.0185, -0.0185, -0.0718, -0.0824, -0.0916, -0.1068, -0.0934, -0.0713, -0.07, -0.077, -0.0754, -0.076, -0.106, -0.1403, -0.1335, -0.0709, -0.0402, -0.0499, -0.0434, -0.0758, -0.0925, -0.0925, -0.072, -0.0693, -0.0679, -0.0717, -0.0781, -0.0916, -0.0726, -0.1022, -0.1082, -0.1103, -0.1242, -0.1139, -0.0713, -0.0777]}
{"unit_id": "quote:1968303370684518529", "ticker_idx": 0, "ts": "2025-09-17T13:17:47+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Outperform rating and raise target price to USD 132", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF Securities Outperform on SanDisk with TP raised to $132; bullish on NAND upcycle through 2026 driven by CSP eSSD demand surge.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Communication\n☄ SanDisk: Is NAND approaching its “HBM moment”?\n\n☀ Maintain Outperform rating and raise target price to USD 132: Following our September 9 upgrade of SanDisk, based on supply chain research with memory manufacturers and NAND procurement teams in the smartphone/PC sectors, we have become more optimistic on NAND. We believe NAND manufacturers remain disciplined in their 2026 expansion guidance, with limited new capacity additions. We estimate overall NAND supply growth in 2026 at ~15%, including YMTC’s planned expansion of 50KPM (about 50,000 wafers/month). By contrast, NAND bit demand growth is likely to exceed 20%, mainly driven by stronger adoption of eSSDs by CSPs. We believe this demand-driven upcycle could run through 2026. Accordingly, we raise SanDisk’s target price (TP) to USD 132, based on 10x CY2026E P/E, equivalent to 1.8x FY2026E P/B.\n\n☀ NAND’s “HBM moment”: We observe that incremental eSSD demand from CSPs equates to around 10% of global NAND demand, driven by the surge in tokens generated by AI inference workloads. During inference, KV caches are stored in HBM, but once a response is generated and the session ends, this data is offloaded from HBM to HDDs or SSDs. As AI usage and token generation accelerate, the volume of data requiring storage is rising rapidly. However, HDDs remain in shortage, with lead times exceeding 52 weeks, creating a bottleneck and forcing CSPs to turn to the more expensive eSSDs. Our supply chain research indicates that eSSD demand has already been booked out through 2026, with CSPs even requesting fulfillment with lower-capacity SSDs to meet demand.\n\n☀ NAND pricing may be stronger: In recent visits to NAND procurement teams at smartphone/PC OEMs, we noted a significant change in market sentiment following the surge in CSP eSSD orders. Previously, the industry broadly anchored to a down-cycle narrative, but the sudden ramp in eSSD volumes is shifting the story. Procurement teams are increasingly worried about supply shortages in 2026, which could prompt them to adopt more aggressive pricing strategies in order to secure short-term NAND supply. Our current model assumes NAND ASPs rise 6–8% QoQ in Q4 2025, and we believe there is still upside to this assumption.", "reply_text": null, "parent_text": null, "quoted_text": "轉\n\n《廣發海外電子通信》\n☄ 閃迪：NAND 是否迎來 HBM 時刻？\n\n☀ 維持跑贏大盤評級並上調目標價至 132 美元：繼我們在 9 月 9 日上調閃迪後，基於與存儲廠商及智能手機/PC 的 NAND 採購團隊的供應鏈調研，我們對 NAND 更加樂觀。我們認為各 NAND 廠商在 2026", "parent_author": null, "ret_m1d": -0.02575288004159515, "ret_signed_m1d": -0.02575288004159515, "alpha_spy_m1d": -0.026996860899865416, "alpha_c_m1d": -0.029216172349155523, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0034632923075603728, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.052144316933342205, "ret_signed_p1d": 0.052144316933342205, "alpha_spy_p1d": 0.047471871791011866, "alpha_c_p1d": 0.03486533422621019, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.017278982707132018, "ret_p1w": 0.062360333449272165, "ret_signed_p1w": 0.062360333449272165, "alpha_spy_p1w": 0.056667174092728256, "alpha_c_p1w": 0.035031153944412585, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.02732917950485958, "ret_p1m": 0.5352772416552434, "ret_signed_p1m": 0.5352772416552434, "alpha_spy_p1m": 0.5302837631702668, "alpha_c_p1m": 0.4857407260982549, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.04953651555698846, "ret_p3m": 1.1482387197491302, "ret_signed_p3m": 1.1482387197491302, "alpha_spy_p3m": 1.1126835767165435, "alpha_c_p3m": 1.0983702004541929, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.0498685192949373, "ret_p6m": 5.585293171070882, "ret_signed_p6m": 5.585293171070882, "alpha_spy_p6m": 5.569060630877283, "alpha_c_p6m": 5.566787433987589, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.018505737083293505, "price_path": [-0.5043, -0.5004, -0.4962, -0.4972, -0.4952, -0.4982, -0.5174, -0.5215, -0.5082, -0.5061, -0.5061, -0.5188, -0.5087, -0.5084, -0.5004, -0.5095, -0.5479, -0.5454, -0.5599, -0.5582, -0.551, -0.5572, -0.5599, -0.5424, -0.5524, -0.5479, -0.5542, -0.5432, -0.5383, -0.5433, -0.5602, -0.5476, -0.5538, -0.552, -0.567, -0.5281, -0.5385, -0.5016, -0.4997, -0.5032, -0.526, -0.5156, -0.5256, -0.5275, -0.5158, -0.5065, -0.5022, -0.4961, -0.4845, -0.4587, -0.4416, -0.4416, -0.4565, -0.4359, -0.3349, -0.2705, -0.2498, -0.2497, -0.2134, -0.1029, -0.0834, -0.0413, -0.0258, 0.0, 0.0521, 0.0877, 0.0952, 0.1323, 0.0624, 0.0034, 0.0335, 0.2078, 0.194, 0.2889, 0.3211, 0.3665, 0.2895, 0.2871, 0.4034, 0.38, 0.2441, 0.4325, 0.3546, 0.5356, 0.5353, 0.4915, 0.5754, 0.5887, 0.5638, 0.7777, 0.9811, 0.8782, 0.8679, 1.1747, 1.0839, 1.1212, 1.2029, 1.0706, 1.3039, 1.2102, 1.5485, 1.8514, 1.8901, 2.0127, 1.592, 1.7047, 1.8294, 1.6065, 1.6174, 1.0853, 1.1312, 1.4152, 1.3465, 1.2884, 1.2884, 1.3761, 1.2366, 1.1853, 1.0685, 1.27, 1.4313, 1.3994, 1.3354, 1.478, 1.5711, 1.1941, 1.1482, 1.2274, 1.201, 1.3354, 1.5286, 1.5652, 1.6062, 1.6613, 1.6613, 1.661, 1.5992, 1.5563, 1.5261, 1.5261, 1.929, 1.9167, 2.7207, 2.7625, 2.5601, 3.0163, 3.1425, 3.1482, 3.127, 3.355, 3.4016, 3.4016, 3.822, 4.3346, 4.3575, 4.0424, 4.0101, 4.1232, 4.6149, 4.7391, 5.1323, 6.0793, 6.4014, 5.2206, 5.1317, 5.3632, 5.2084, 4.764, 5.378, 5.7074, 5.6677, 5.6677, 5.2849, 5.3893, 5.6095, 5.9168, 6.0926, 5.7949, 5.7296, 5.9373, 5.7613, 5.5881, 5.0169, 5.375, 5.0188, 4.6117, 5.2651, 5.586, 5.9749, 5.5853]}
{"unit_id": "quote:1968303370684518529", "ticker_idx": 1, "ts": "2025-09-17T13:17:47+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We believe this demand-driven upcycle could run through 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF Securities Outperform on SanDisk with TP raised to $132; bullish on NAND upcycle through 2026 driven by CSP eSSD demand surge.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Communication\n☄ SanDisk: Is NAND approaching its “HBM moment”?\n\n☀ Maintain Outperform rating and raise target price to USD 132: Following our September 9 upgrade of SanDisk, based on supply chain research with memory manufacturers and NAND procurement teams in the smartphone/PC sectors, we have become more optimistic on NAND. We believe NAND manufacturers remain disciplined in their 2026 expansion guidance, with limited new capacity additions. We estimate overall NAND supply growth in 2026 at ~15%, including YMTC’s planned expansion of 50KPM (about 50,000 wafers/month). By contrast, NAND bit demand growth is likely to exceed 20%, mainly driven by stronger adoption of eSSDs by CSPs. We believe this demand-driven upcycle could run through 2026. Accordingly, we raise SanDisk’s target price (TP) to USD 132, based on 10x CY2026E P/E, equivalent to 1.8x FY2026E P/B.\n\n☀ NAND’s “HBM moment”: We observe that incremental eSSD demand from CSPs equates to around 10% of global NAND demand, driven by the surge in tokens generated by AI inference workloads. During inference, KV caches are stored in HBM, but once a response is generated and the session ends, this data is offloaded from HBM to HDDs or SSDs. As AI usage and token generation accelerate, the volume of data requiring storage is rising rapidly. However, HDDs remain in shortage, with lead times exceeding 52 weeks, creating a bottleneck and forcing CSPs to turn to the more expensive eSSDs. Our supply chain research indicates that eSSD demand has already been booked out through 2026, with CSPs even requesting fulfillment with lower-capacity SSDs to meet demand.\n\n☀ NAND pricing may be stronger: In recent visits to NAND procurement teams at smartphone/PC OEMs, we noted a significant change in market sentiment following the surge in CSP eSSD orders. Previously, the industry broadly anchored to a down-cycle narrative, but the sudden ramp in eSSD volumes is shifting the story. Procurement teams are increasingly worried about supply shortages in 2026, which could prompt them to adopt more aggressive pricing strategies in order to secure short-term NAND supply. Our current model assumes NAND ASPs rise 6–8% QoQ in Q4 2025, and we believe there is still upside to this assumption.", "reply_text": null, "parent_text": null, "quoted_text": "轉\n\n《廣發海外電子通信》\n☄ 閃迪：NAND 是否迎來 HBM 時刻？\n\n☀ 維持跑贏大盤評級並上調目標價至 132 美元：繼我們在 9 月 9 日上調閃迪後，基於與存儲廠商及智能手機/PC 的 NAND 採購團隊的供應鏈調研，我們對 NAND 更加樂觀。我們認為各 NAND 廠商在 2026", "parent_author": null, "ret_m1d": 0.017088492788325672, "ret_signed_m1d": 0.017088492788325672, "alpha_spy_m1d": 0.015844511930055406, "alpha_c_m1d": 0.010667532051097585, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043722415302998786, "ret_signed_p1d": 0.043722415302998786, "alpha_spy_p1d": 0.03904997016066845, "alpha_c_p1d": 0.005327625401274717, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.07040231638139702, "ret_signed_p1w": 0.07040231638139702, "alpha_spy_p1w": 0.06470915702485311, "alpha_c_p1w": 0.01844494427586127, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.3926535304208163, "ret_signed_p1m": 0.3926535304208163, "alpha_spy_p1m": 0.38766005193583974, "alpha_c_p1m": 0.26800168847878303, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.7428400321487472, "ret_signed_p3m": 0.7428400321487472, "alpha_spy_p3m": 0.7072848891161605, "alpha_c_p3m": 0.5867384597023366, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 2.8254856478969606, "ret_signed_p6m": 2.8254856478969606, "alpha_spy_p6m": 2.809253107703362, "alpha_c_p6m": 2.5500140368251003, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.3385, -0.3297, -0.3041, -0.2995, -0.2992, -0.3046, -0.309, -0.322, -0.3184, -0.3038, -0.3174, -0.323, -0.3053, -0.3041, -0.2971, -0.2977, -0.312, -0.3069, -0.3132, -0.3293, -0.3285, -0.3273, -0.3378, -0.3261, -0.325, -0.332, -0.3243, -0.3215, -0.3103, -0.3142, -0.341, -0.3347, -0.3311, -0.3386, -0.3327, -0.3133, -0.3051, -0.2798, -0.2781, -0.2802, -0.2901, -0.2995, -0.3046, -0.3195, -0.3273, -0.313, -0.3089, -0.3076, -0.3047, -0.2889, -0.283, -0.2926, -0.2924, -0.2847, -0.2549, -0.21, -0.1983, -0.1823, -0.1427, -0.0801, -0.0303, -0.0169, 0.0171, 0.0, 0.0437, 0.0436, 0.066, 0.0847, 0.0704, 0.051, 0.0161, 0.0826, 0.0915, 0.1342, 0.2054, 0.2544, 0.2492, 0.2232, 0.2591, 0.2645, 0.2526, 0.2937, 0.2573, 0.3293, 0.3927, 0.3776, 0.439, 0.4256, 0.4324, 0.4809, 0.6253, 0.6757, 0.6417, 0.7653, 0.796, 0.8032, 0.879, 0.7924, 0.8385, 0.8686, 0.949, 1.109, 1.1111, 1.1401, 1.0326, 0.8868, 0.9933, 0.8624, 0.8791, 0.7804, 0.6934, 0.7752, 0.7603, 0.7018, 0.7457, 0.771, 0.73, 0.7523, 0.7122, 0.7413, 0.8127, 0.8735, 0.8458, 0.8818, 0.8851, 0.8084, 0.7428, 0.7098, 0.7491, 0.8141, 0.8604, 0.947, 0.952, 1.0068, 1.0154, 1.0627, 1.0616, 1.0477, 1.0326, 1.0326, 1.1886, 1.2594, 1.4907, 1.5583, 1.5188, 1.612, 1.6407, 1.6673, 1.6482, 1.7272, 1.8367, 1.8633, 1.9268, 2.1271, 2.222, 2.1424, 2.1093, 2.2609, 2.445, 2.4916, 2.6676, 2.6771, 2.9181, 2.637, 2.5017, 2.5396, 2.5348, 2.4225, 2.5872, 2.8073, 2.8736, 2.849, 2.7608, 2.7654, 2.8283, 2.9055, 2.9404, 3.0046, 2.9859, 3.0878, 3.0205, 3.0, 2.6566, 2.5984, 2.6754, 2.5269, 2.5029, 2.7522, 2.9466, 2.8255]}
{"unit_id": "thread:1971731186415444406", "ticker_idx": 0, "ts": "2025-09-27T00:18:42+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Prices are expected to rise further in Q4 and throughout 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM and NAND prices set to rise further in Q4 and 2026 driven by AI-led structural demand surge; Micron flagged as beneficiary.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM, NAND Flash Prices to Rise Again in Q4\n\nThe new wave of computing power demand driven by generative AI and large language models is pushing the global memory market upward. Recently, both DRAM and NAND Flash have experienced shortages, with supply gaps proving more severe than earlier forecasts. Prices are expected to rise further in Q4 and throughout 2026. Industry insiders note that global cloud service providers (CSPs) have significantly revised upward their 2026 orders, leaving the current inventories of the three major memory suppliers insufficient. This shortage could mark the beginning of a global memory supply crunch in 2026.\n\nAt the same time, traditional hard disk drive (HDD) makers are cutting production, which could result in at least half a year of undersupply and force some orders to shift toward SSDs, further tightening the NAND supply chain. Most industry players believe that Q4 memory price hikes are already inevitable. While October contract prices have yet to be finalized, upstream suppliers are taking a more aggressive stance, with price increases likely to exceed earlier projections.\n\nSpecifically, DDR5 contract prices were previously expected to rise by 10–15%, with spot prices up 15–25%. DDR4 contract prices were forecast to increase by more than 10%, with spot prices climbing over 15%. All these numbers now have room for upward revision. The market is reacting quickly: leading memory module maker ADATA announced it would suspend DDR4 quotations starting from the 29th, prioritizing DDR5 and NAND supply for major customers.\n\nNAND controller chip leader Phison recently resumed some quotations and directly raised prices by around 10%, a signal that the NAND Flash market is entering an uptrend. Meanwhile, memory giant Micron’s latest earnings report highlighted that AI-driven demand is spreading across data centers, PCs, smartphones, and automotive applications, laying the foundation for growth in 2026.\n\nGlobal server shipments in 2025 are now projected to grow 10% year-on-year, a significant upward revision from earlier estimates. AI server demand is the strongest driver, but it is also fueling a rebound in traditional servers. The integration of AI technology into product design and manufacturing is boosting productivity by 30–40%, increasing wafer image analysis workloads fivefold year-on-year, shortening R&D timelines, and improving yields.\n\nPC demand is strengthening due to the phase-out of Windows 10 and the rise of AI PCs. In smartphones, the high-capacity trend is accelerating, with one-third of flagship models already equipped with 12GB or more of DRAM.\n\nIn automotive, advanced driver-assistance systems (ADAS) and AI-powered in-cabin experiences require more memory, boosting demand for high-end DDR5 and LPDDR5. Real-world AI applications such as robots, drones, and AR/VR devices are also emerging as new growth engines in the embedded market.\n\nIndustry experts point out that AI is extending from data centers to personal devices and vehicles, comprehensively driving up memory demand. With AI servers ramping up, PCs and smartphones moving toward higher capacity, and rapid expansion in automotive and embedded markets, the global memory industry is entering a new phase of structural growth.\n\n---\n\nhttps://t.co/R5Cfvsczge", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03045034578779408, "ret_signed_m1d": -0.03045034578779408, "alpha_spy_m1d": -0.027647855371671286, "alpha_c_m1d": -0.027847056225600946, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.002603289562193134, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0043352469102124385, "ret_signed_p1d": 0.0043352469102124385, "alpha_spy_p1d": 0.0005683801511004631, "alpha_c_p1d": -0.007131883558626299, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.011467130468838738, "ret_p1w": 0.1216609772835519, "ret_signed_p1w": 0.1216609772835519, "alpha_spy_p1w": 0.10971249590113923, "alpha_c_p1w": 0.057444762031476124, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.06421621525207577, "ret_p1m": 0.34311017083707157, "ret_signed_p1m": 0.34311017083707157, "alpha_spy_p1m": 0.30788236409450226, "alpha_c_p1m": 0.21774601484880346, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.1253641559882681, "ret_p3m": 0.5852362984520828, "ret_signed_p3m": 0.5852362984520828, "alpha_spy_p3m": 0.5419323138399296, "alpha_c_p3m": 0.4530749751434241, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.1321613233086587, "ret_p6m": 1.504034507550185, "ret_signed_p6m": 1.504034507550185, "alpha_spy_p6m": 1.5142515139621269, "alpha_c_p6m": 1.277301070018224, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": 0.22673343753196118, "price_path": [-0.2469, -0.2345, -0.2441, -0.2545, -0.236, -0.2453, -0.2259, -0.2191, -0.2263, -0.22, -0.2259, -0.2497, -0.2463, -0.2426, -0.2617, -0.2584, -0.2556, -0.2603, -0.2464, -0.2437, -0.2292, -0.2358, -0.2691, -0.2601, -0.2514, -0.2611, -0.2447, -0.2306, -0.2139, -0.2034, -0.1988, -0.1993, -0.2083, -0.2177, -0.2251, -0.2401, -0.2526, -0.2399, -0.2339, -0.2366, -0.2343, -0.2211, -0.2263, -0.247, -0.2379, -0.2327, -0.2175, -0.1977, -0.1918, -0.1681, -0.1388, -0.1113, -0.0693, -0.0615, -0.0317, -0.0479, -0.0005, -0.013, -0.0009, 0.0171, 0.0069, -0.0012, -0.0305, 0.0, 0.0043, 0.0547, 0.1068, 0.1153, 0.1217, 0.1109, 0.133, 0.1244, 0.1521, 0.158, 0.1363, 0.1702, 0.2311, 0.244, 0.2729, 0.2552, 0.2541, 0.2597, 0.324, 0.3627, 0.3431, 0.392, 0.4105, 0.4149, 0.5096, 0.4187, 0.4346, 0.4441, 0.4257, 0.4925, 0.4916, 0.4958, 0.4737, 0.422, 0.4694, 0.3966, 0.3786, 0.3534, 0.2949, 0.3352, 0.3457, 0.376, 0.3983, 0.3857, 0.4026, 0.43, 0.4178, 0.3954, 0.4318, 0.4875, 0.4838, 0.5253, 0.4908, 0.4676, 0.4277, 0.3867, 0.4127, 0.4594, 0.4848, 0.5547, 0.5618, 0.5852, 0.5852, 0.6153, 0.6864, 0.6949, 0.6802, 0.6802, 0.8004, 0.8501, 0.9456, 0.9614, 0.9405, 0.9666, 0.9722, 0.9411, 0.946, 0.9737, 1.0534, 1.0627, 1.0309, 1.0942, 1.1369, 1.1518, 1.1007, 1.2343, 1.336, 1.3494, 1.3519, 1.2829, 1.3872, 1.3053, 1.218, 1.2364, 1.2905, 1.2567, 1.3249, 1.402, 1.4, 1.4, 1.3758, 1.4189, 1.46, 1.535, 1.5338, 1.6072, 1.6559, 1.7656, 1.7167, 1.7174, 1.4482, 1.3139, 1.4716, 1.3873, 1.2833, 1.4663, 1.5227, 1.4632, 1.4689, 1.5828, 1.6401, 1.7806, 1.6721, 1.6185, 1.4577, 1.504]}
{"unit_id": "thread:1971731186415444406", "ticker_idx": 1, "ts": "2025-09-27T00:18:42+00:00", "symbol": "NAND Flash", "kind": "sector", "geography": "global", "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the NAND Flash market is entering an uptrend", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM and NAND prices set to rise further in Q4 and 2026 driven by AI-led structural demand surge; Micron flagged as beneficiary.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "Global NAND sector basket: Micron, SK Hynix, Samsung, Kioxia, SanDisk", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM, NAND Flash Prices to Rise Again in Q4\n\nThe new wave of computing power demand driven by generative AI and large language models is pushing the global memory market upward. Recently, both DRAM and NAND Flash have experienced shortages, with supply gaps proving more severe than earlier forecasts. Prices are expected to rise further in Q4 and throughout 2026. Industry insiders note that global cloud service providers (CSPs) have significantly revised upward their 2026 orders, leaving the current inventories of the three major memory suppliers insufficient. This shortage could mark the beginning of a global memory supply crunch in 2026.\n\nAt the same time, traditional hard disk drive (HDD) makers are cutting production, which could result in at least half a year of undersupply and force some orders to shift toward SSDs, further tightening the NAND supply chain. Most industry players believe that Q4 memory price hikes are already inevitable. While October contract prices have yet to be finalized, upstream suppliers are taking a more aggressive stance, with price increases likely to exceed earlier projections.\n\nSpecifically, DDR5 contract prices were previously expected to rise by 10–15%, with spot prices up 15–25%. DDR4 contract prices were forecast to increase by more than 10%, with spot prices climbing over 15%. All these numbers now have room for upward revision. The market is reacting quickly: leading memory module maker ADATA announced it would suspend DDR4 quotations starting from the 29th, prioritizing DDR5 and NAND supply for major customers.\n\nNAND controller chip leader Phison recently resumed some quotations and directly raised prices by around 10%, a signal that the NAND Flash market is entering an uptrend. Meanwhile, memory giant Micron’s latest earnings report highlighted that AI-driven demand is spreading across data centers, PCs, smartphones, and automotive applications, laying the foundation for growth in 2026.\n\nGlobal server shipments in 2025 are now projected to grow 10% year-on-year, a significant upward revision from earlier estimates. AI server demand is the strongest driver, but it is also fueling a rebound in traditional servers. The integration of AI technology into product design and manufacturing is boosting productivity by 30–40%, increasing wafer image analysis workloads fivefold year-on-year, shortening R&D timelines, and improving yields.\n\nPC demand is strengthening due to the phase-out of Windows 10 and the rise of AI PCs. In smartphones, the high-capacity trend is accelerating, with one-third of flagship models already equipped with 12GB or more of DRAM.\n\nIn automotive, advanced driver-assistance systems (ADAS) and AI-powered in-cabin experiences require more memory, boosting demand for high-end DDR5 and LPDDR5. Real-world AI applications such as robots, drones, and AR/VR devices are also emerging as new growth engines in the embedded market.\n\nIndustry experts point out that AI is extending from data centers to personal devices and vehicles, comprehensively driving up memory demand. With AI servers ramping up, PCs and smartphones moving toward higher capacity, and rapid expansion in automotive and embedded markets, the global memory industry is entering a new phase of structural growth.\n\n---\n\nhttps://t.co/R5Cfvsczge", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05911224838190079, "ret_signed_m1d": -0.05911224838190079, "alpha_spy_m1d": -0.056309757965777994, "alpha_c_m1d": -0.056508958819707654, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.002603289562193134, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008866543602522547, "ret_signed_p1d": 0.008866543602522547, "alpha_spy_p1d": 0.0050996768434105715, "alpha_c_p1d": -0.002600586866316191, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.011467130468838738, "ret_p1w": 0.1568863331966078, "ret_signed_p1w": 0.1568863331966078, "alpha_spy_p1w": 0.14493785181419513, "alpha_c_p1w": 0.09267011794453203, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.06421621525207577, "ret_p1m": 0.5164534875688265, "ret_signed_p1m": 0.5164534875688265, "alpha_spy_p1m": 0.48122568082625716, "alpha_c_p1m": 0.39108933158055836, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.1253641559882681, "ret_p3m": 0.8536295676401886, "ret_signed_p3m": 0.8536295676401886, "alpha_spy_p3m": 0.8103255830280354, "alpha_c_p3m": 0.7214682443315299, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.1321613233086587, "ret_p6m": 2.6429446871695723, "ret_signed_p6m": 2.6429446871695723, "alpha_spy_p6m": 2.6531616935815143, "alpha_c_p6m": 2.416211249637611, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": 0.22673343753196118, "price_path": [-0.3635, -0.3499, -0.3628, -0.3678, -0.351, -0.3499, -0.3434, -0.3438, -0.3567, -0.3519, -0.3578, -0.3734, -0.3728, -0.3716, -0.3815, -0.3708, -0.3695, -0.3762, -0.369, -0.3667, -0.3562, -0.3598, -0.3849, -0.3792, -0.3756, -0.3826, -0.3768, -0.3591, -0.3508, -0.3275, -0.3261, -0.328, -0.337, -0.3459, -0.3507, -0.3649, -0.3728, -0.3594, -0.3556, -0.3544, -0.3519, -0.3373, -0.3323, -0.3413, -0.3408, -0.334, -0.308, -0.2667, -0.2561, -0.2408, -0.2039, -0.1469, -0.0999, -0.0882, -0.0564, -0.0732, -0.0327, -0.0339, -0.013, 0.0039, -0.0082, -0.0254, -0.0591, 0.0, 0.0089, 0.0476, 0.1142, 0.1598, 0.1569, 0.1321, 0.1637, 0.1693, 0.1608, 0.1955, 0.1628, 0.2268, 0.2872, 0.274, 0.3305, 0.3173, 0.3241, 0.3655, 0.498, 0.5483, 0.5165, 0.6265, 0.6577, 0.6633, 0.7336, 0.6544, 0.6934, 0.7244, 0.7925, 0.9375, 0.938, 0.9628, 0.8706, 0.7299, 0.8278, 0.7085, 0.7241, 0.6423, 0.5589, 0.6301, 0.6175, 0.5633, 0.605, 0.6272, 0.5919, 0.6142, 0.5787, 0.6011, 0.6651, 0.724, 0.6994, 0.7306, 0.7313, 0.667, 0.6059, 0.5725, 0.61, 0.6692, 0.7091, 0.7907, 0.7944, 0.8455, 0.8536, 0.8981, 0.8989, 0.8867, 0.8729, 0.8729, 1.0117, 1.0786, 1.2797, 1.3427, 1.3099, 1.3879, 1.4121, 1.437, 1.4191, 1.4884, 1.5919, 1.6171, 1.6674, 1.8457, 1.9355, 1.8676, 1.8365, 1.9782, 2.1418, 2.1833, 2.3404, 2.3274, 2.5471, 2.3073, 2.1816, 2.2124, 2.2109, 2.1147, 2.2572, 2.4577, 2.5214, 2.4981, 2.4232, 2.4257, 2.4795, 2.5459, 2.5743, 2.6422, 2.6262, 2.7167, 2.657, 2.6417, 2.3294, 2.2673, 2.348, 2.2162, 2.1787, 2.4076, 2.5832, 2.4773, 2.5484, 2.757, 2.7776, 2.9971, 2.9203, 2.7782, 2.6308, 2.6429]}
{"unit_id": "thread:1971731186415444406", "ticker_idx": 2, "ts": "2025-09-27T00:18:42+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory giant Micron's latest earnings report highlighted that AI-driven demand is spreading across data centers, PCs, smartphones, and automotive applications, laying the foundation for growth in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM and NAND prices set to rise further in Q4 and 2026 driven by AI-led structural demand surge; Micron flagged as beneficiary.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM, NAND Flash Prices to Rise Again in Q4\n\nThe new wave of computing power demand driven by generative AI and large language models is pushing the global memory market upward. Recently, both DRAM and NAND Flash have experienced shortages, with supply gaps proving more severe than earlier forecasts. Prices are expected to rise further in Q4 and throughout 2026. Industry insiders note that global cloud service providers (CSPs) have significantly revised upward their 2026 orders, leaving the current inventories of the three major memory suppliers insufficient. This shortage could mark the beginning of a global memory supply crunch in 2026.\n\nAt the same time, traditional hard disk drive (HDD) makers are cutting production, which could result in at least half a year of undersupply and force some orders to shift toward SSDs, further tightening the NAND supply chain. Most industry players believe that Q4 memory price hikes are already inevitable. While October contract prices have yet to be finalized, upstream suppliers are taking a more aggressive stance, with price increases likely to exceed earlier projections.\n\nSpecifically, DDR5 contract prices were previously expected to rise by 10–15%, with spot prices up 15–25%. DDR4 contract prices were forecast to increase by more than 10%, with spot prices climbing over 15%. All these numbers now have room for upward revision. The market is reacting quickly: leading memory module maker ADATA announced it would suspend DDR4 quotations starting from the 29th, prioritizing DDR5 and NAND supply for major customers.\n\nNAND controller chip leader Phison recently resumed some quotations and directly raised prices by around 10%, a signal that the NAND Flash market is entering an uptrend. Meanwhile, memory giant Micron’s latest earnings report highlighted that AI-driven demand is spreading across data centers, PCs, smartphones, and automotive applications, laying the foundation for growth in 2026.\n\nGlobal server shipments in 2025 are now projected to grow 10% year-on-year, a significant upward revision from earlier estimates. AI server demand is the strongest driver, but it is also fueling a rebound in traditional servers. The integration of AI technology into product design and manufacturing is boosting productivity by 30–40%, increasing wafer image analysis workloads fivefold year-on-year, shortening R&D timelines, and improving yields.\n\nPC demand is strengthening due to the phase-out of Windows 10 and the rise of AI PCs. In smartphones, the high-capacity trend is accelerating, with one-third of flagship models already equipped with 12GB or more of DRAM.\n\nIn automotive, advanced driver-assistance systems (ADAS) and AI-powered in-cabin experiences require more memory, boosting demand for high-end DDR5 and LPDDR5. Real-world AI applications such as robots, drones, and AR/VR devices are also emerging as new growth engines in the embedded market.\n\nIndustry experts point out that AI is extending from data centers to personal devices and vehicles, comprehensively driving up memory demand. With AI servers ramping up, PCs and smartphones moving toward higher capacity, and rapid expansion in automotive and embedded markets, the global memory industry is entering a new phase of structural growth.\n\n---\n\nhttps://t.co/R5Cfvsczge", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.040451301191353206, "ret_signed_m1d": -0.040451301191353206, "alpha_spy_m1d": -0.03764881077523041, "alpha_c_m1d": -0.03784801162916007, "bench_spy_m1d": -0.0028024904161227937, "bench_c_m1d": -0.002603289562193134, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020866536028222082, "ret_signed_p1d": 0.020866536028222082, "alpha_spy_p1d": 0.017099669269110107, "alpha_c_p1d": 0.009399405559383345, "bench_spy_p1d": 0.0037668667591119753, "bench_c_p1d": 0.011467130468838738, "ret_p1w": 0.16583045755341375, "ret_signed_p1w": 0.16583045755341375, "alpha_spy_p1w": 0.15388197617100108, "alpha_c_p1w": 0.10161424230133798, "bench_spy_p1w": 0.011948481382412668, "bench_c_p1w": 0.06421621525207577, "ret_p1m": 0.3547834163389052, "ret_signed_p1m": 0.3547834163389052, "alpha_spy_p1m": 0.3195556095963359, "alpha_c_p1m": 0.2294192603506371, "bench_spy_p1m": 0.03522780674256931, "bench_c_p1m": 0.1253641559882681, "ret_p3m": 0.750210940251838, "ret_signed_p3m": 0.750210940251838, "alpha_spy_p3m": 0.7069069556396848, "alpha_c_p3m": 0.6180496169431793, "bench_spy_p3m": 0.043303984612153235, "bench_c_p3m": 0.1321613233086587, "ret_p6m": 1.4157269220176496, "ret_signed_p6m": 1.4157269220176496, "alpha_spy_p6m": 1.4259439284295916, "alpha_c_p6m": 1.1889934844856884, "bench_spy_p6m": -0.010217006411941898, "bench_c_p6m": 0.22673343753196118, "price_path": [-0.2579, -0.2546, -0.2546, -0.2683, -0.2409, -0.2542, -0.2489, -0.2402, -0.2763, -0.2672, -0.2896, -0.309, -0.3021, -0.3092, -0.3336, -0.3299, -0.3183, -0.3212, -0.3212, -0.3169, -0.2999, -0.3341, -0.3601, -0.3425, -0.3346, -0.3363, -0.3174, -0.2746, -0.2451, -0.2206, -0.2418, -0.2356, -0.2625, -0.2462, -0.2553, -0.2849, -0.2935, -0.282, -0.2897, -0.2892, -0.2816, -0.2556, -0.2739, -0.2739, -0.2771, -0.2757, -0.2422, -0.1985, -0.1979, -0.1749, -0.1458, -0.0813, -0.0407, -0.0374, -0.031, -0.0239, 0.0304, -0.0071, 0.0044, 0.0153, -0.0134, -0.0431, -0.0405, 0.0, 0.0209, 0.1113, 0.1211, 0.1467, 0.1658, 0.1337, 0.1999, 0.1742, 0.1087, 0.1769, 0.142, 0.1718, 0.2365, 0.2356, 0.2624, 0.235, 0.2117, 0.262, 0.3371, 0.3437, 0.3548, 0.3836, 0.3676, 0.3661, 0.4329, 0.3311, 0.45, 0.455, 0.4525, 0.5464, 0.472, 0.4951, 0.4466, 0.5069, 0.4771, 0.395, 0.3793, 0.2294, 0.266, 0.3671, 0.3708, 0.4058, 0.4058, 0.4437, 0.468, 0.4621, 0.4296, 0.3837, 0.4483, 0.5075, 0.5411, 0.61, 0.5779, 0.4722, 0.45, 0.4195, 0.3768, 0.5174, 0.6235, 0.6886, 0.6867, 0.7502, 0.7502, 0.7387, 0.7979, 0.7873, 0.7432, 0.7432, 0.9264, 0.9065, 1.0975, 1.0738, 0.9973, 1.1077, 1.1124, 1.0652, 1.036, 1.056, 1.2155, 1.2155, 1.2293, 1.3765, 1.4282, 1.4409, 1.3764, 1.5056, 1.6585, 1.6616, 1.5339, 1.6739, 1.5618, 1.3172, 1.3385, 1.4106, 1.3423, 1.2797, 1.5062, 1.5284, 1.5142, 1.5142, 1.4417, 1.571, 1.549, 1.6151, 1.5711, 1.553, 1.6201, 1.5381, 1.5186, 1.5204, 1.3189, 1.4477, 1.425, 1.2616, 1.3778, 1.462, 1.5572, 1.4757, 1.6026, 1.6983, 1.8198, 1.82, 1.7134, 1.5829, 1.4696, 1.4157]}
{"unit_id": "quote:1950103391025103163", "ticker_idx": 0, "ts": "2025-07-29T07:57:34+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$TSM is one of the safest choices in the AI era", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Endorses TSM as one of the safest AI-era plays, citing CoWoS capacity dominance.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "$TSM is one of the safest choices in the AI era. https://t.co/khmeQuQHJt", "reply_text": null, "parent_text": null, "quoted_text": "Morgan Stanley's Detailed Analysis of the TSMC CoWoS Capacity Battle: NVIDIA Secures 60%, Cloud AI Chip Market to Surge 40-50% by 2026\n\nAccording to a recent research report from Morgan Stanley, a detailed forecast has been released regarding the allocation of CoWoS capacity https://t.co/unpdIBsicl", "parent_author": null, "ret_m1d": 0.005884076442545405, "ret_signed_m1d": 0.005884076442545405, "alpha_spy_m1d": 0.0032393551009279875, "alpha_c_m1d": 0.010971046255907169, "bench_spy_m1d": 0.0026447213416174176, "bench_c_m1d": -0.005086969813361764, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006547046717751748, "ret_signed_p1d": 0.006547046717751748, "alpha_spy_p1d": 0.007806280956830669, "alpha_c_p1d": -0.004924555162479027, "bench_spy_p1d": -0.0012592342390789213, "bench_c_p1d": 0.011471601880230775, "ret_p1w": -0.03671318279560565, "ret_signed_p1w": -0.03671318279560565, "alpha_spy_p1w": -0.025237543590826195, "alpha_c_p1w": -0.01691129159083138, "bench_spy_p1w": -0.011475639204779453, "bench_c_p1w": -0.01980189120477427, "ret_p1m": -0.00845323773125517, "ret_signed_p1m": -0.00845323773125517, "alpha_spy_p1m": -0.026351494782860252, "alpha_c_p1m": -0.024704534374484277, "bench_spy_p1m": 0.017898257051605082, "bench_c_p1m": 0.016251296643229107, "ret_p3m": 0.2260826218494083, "ret_signed_p3m": 0.2260826218494083, "alpha_spy_p3m": 0.15702792538822852, "alpha_c_p3m": 0.02724326451413095, "bench_spy_p3m": 0.06905469646117979, "bench_c_p3m": 0.19883935733527736, "ret_p6m": 0.359267386913662, "ret_signed_p6m": 0.359267386913662, "alpha_spy_p6m": 0.2741507682192077, "alpha_c_p6m": -0.017245616098992222, "bench_spy_p6m": 0.0851166186944543, "bench_c_p6m": 0.3765130030126542, "price_path": [-0.2869, -0.2598, -0.2717, -0.2887, -0.2794, -0.2766, -0.2712, -0.228, -0.1991, -0.1959, -0.1981, -0.1981, -0.2011, -0.2013, -0.2083, -0.19, -0.2074, -0.2074, -0.1839, -0.1902, -0.186, -0.2019, -0.1956, -0.1841, -0.1644, -0.1605, -0.1529, -0.1454, -0.1228, -0.1161, -0.1073, -0.1253, -0.1063, -0.1137, -0.1153, -0.1153, -0.1319, -0.1285, -0.088, -0.077, -0.0718, -0.0529, -0.0615, -0.069, -0.032, -0.0271, -0.0271, -0.0504, -0.0558, -0.0393, -0.0479, -0.0453, -0.0525, -0.0181, -0.0156, 0.0177, -0.0039, -0.0103, -0.0279, -0.0041, 0.0011, 0.0177, 0.0059, 0.0, 0.0065, 0.0012, -0.0254, -0.0097, -0.0367, -0.0413, 0.0053, 0.0021, 0.0031, 0.0123, 0.0005, -0.0014, -0.0102, 0.0003, -0.0358, -0.0527, -0.058, -0.0346, -0.0238, -0.0108, -0.0085, -0.0127, -0.0433, -0.0433, -0.0536, -0.0412, -0.0254, 0.0086, 0.0243, 0.0397, 0.0792, 0.0728, 0.0746, 0.0831, 0.0893, 0.0924, 0.1167, 0.101, 0.1333, 0.1752, 0.1668, 0.15, 0.1363, 0.1358, 0.1609, 0.1991, 0.1976, 0.2146, 0.257, 0.2222, 0.2658, 0.2465, 0.1666, 0.259, 0.2302, 0.2666, 0.2464, 0.2266, 0.2375, 0.2242, 0.2008, 0.2085, 0.2261, 0.2398, 0.2534, 0.2682, 0.2604, 0.2488, 0.2672, 0.2223, 0.2206, 0.2023, 0.1909, 0.2274, 0.2103, 0.208, 0.173, 0.1839, 0.1723, 0.1552, 0.1737, 0.1535, 0.1434, 0.1832, 0.1834, 0.2053, 0.2053, 0.2117, 0.1958, 0.2142, 0.2281, 0.2176, 0.2251, 0.2548, 0.2612, 0.2892, 0.2706, 0.2172, 0.1993, 0.1957, 0.1544, 0.1865, 0.2043, 0.2224, 0.2377, 0.2454, 0.2454, 0.2622, 0.2542, 0.2486, 0.2666, 0.2666, 0.3321, 0.3431, 0.3647, 0.3283, 0.3255, 0.3489, 0.3828, 0.3805, 0.3634, 0.424, 0.4271, 0.4271, 0.3636, 0.3593]}
{"unit_id": "orig:1943953549425357115", "ticker_idx": 0, "ts": "2025-07-12T08:40:17+00:00", "symbol": "HPE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley notes that HPE has raised its Juniper acquisition synergy target to $600 million annually and emphasized Juniper's strong growth potential in the data center and AI networking fields, expecting the acquisition to significantly improve HPE's overall earnings structure", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note on HPE: raised Juniper synergy target to $600M, accretive EPS in Y1, Juniper to drive >50% of operating profit.", "resolved_tickers": ["HPE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "HPE: Synergies Target Raised, Juniper Drives Network Business Growth\nMorgan Stanley (M. Marshall, 25/07/10)\n\nMorgan Stanley notes that HPE has raised its Juniper acquisition synergy target to $600 million annually and emphasized Juniper's strong growth potential in the data center and AI networking fields, expecting the acquisition to significantly improve HPE's overall earnings structure.\n\nJuniper 2Q25 Orders and Revenue Significantly Exceed Expectations\nJuniper's second-quarter orders grew over 40% year-over-year, and revenue increased over 20% year-over-year, significantly exceeding previous expectations of approximately 10%. This was primarily driven by data center clients, followed by campus and branch scenarios. AI clients showed growth in both front-end and back-end, including deployments connecting over 1 million GPUs in new cloud services and sovereign cloud clusters.\n\nJuniper Empowers HPE to Build a Complete AI and IT Platform\nManagement believes that Juniper's network stack will enhance HPE's overall AI solutions, building a more complete IT platform. Concurrently, HPE's scale advantage in global sales channels will help Juniper expand its market share, especially in international markets.\n\nSynergy Target Raised from $450 Million to $600 Million Annually\nHPE has raised its synergy target for the next three years to $600 million annually, stating that one-third will be achieved in the first year, with the remainder completed in the subsequent two years. The focus will be on G&A cost savings (e.g., real estate) and COGS improvements from supply chain optimization.\n\nJuniper to Become a Key Component of HPE's Profit Structure\nUpon completion of the acquisition, HPE's networking business (including Juniper) is expected to contribute over 50% of operating profit. Juniper is projected to be accretive to non-GAAP EPS in the first full year post-acquisition and drive free cash flow growth in the second and third years. HPE plans to release its three-year financial guidance at the SAM conference on October 15.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00241890506647624, "ret_signed_m1d": 0.00241890506647624, "alpha_spy_m1d": 0.004323388315801258, "alpha_c_m1d": 0.004323388315801258, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": -0.0019044832493250174, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01886781642770341, "ret_signed_p1d": -0.01886781642770341, "alpha_spy_p1d": -0.01459464937707633, "alpha_c_p1d": -0.01459464937707633, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": -0.00427316705062708, "ret_p1w": -0.0004837058745186251, "ret_signed_p1w": -0.0004837058745186251, "alpha_spy_p1w": -0.006821737859237009, "alpha_c_p1w": -0.006821737859237009, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.006338031984718384, "ret_p1m": 0.03144661117542791, "ret_signed_p1m": 0.03144661117542791, "alpha_spy_p1m": 0.0028297111031674316, "alpha_c_p1m": 0.0028297111031674316, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.028616900072260476, "ret_p3m": 0.27666764409691935, "ret_signed_p3m": 0.27666764409691935, "alpha_spy_p3m": 0.1995069121929336, "alpha_c_p3m": 0.1995069121929336, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.07716073190398576, "ret_p6m": 0.1639842749568281, "ret_signed_p6m": 0.1639842749568281, "alpha_spy_p6m": 0.050401039768613654, "alpha_c_p6m": 0.050401039768613654, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.11358323518821445, "price_path": [-0.2853, -0.2729, -0.2729, -0.2882, -0.2724, -0.2628, -0.2229, -0.22, -0.208, -0.2167, -0.221, -0.2119, -0.1902, -0.1994, -0.208, -0.2047, -0.1898, -0.1888, -0.148, -0.1393, -0.1427, -0.1537, -0.1475, -0.1614, -0.1489, -0.1576, -0.1571, -0.1634, -0.1634, -0.1384, -0.1504, -0.1523, -0.1701, -0.1672, -0.1504, -0.1432, -0.147, -0.1302, -0.1206, -0.1192, -0.1249, -0.1293, -0.1523, -0.124, -0.1403, -0.1398, -0.1398, -0.1311, -0.1384, -0.1219, -0.1055, -0.1113, -0.1093, -0.0106, -0.0092, 0.0281, 0.0324, 0.0324, 0.0116, 0.0073, 0.0063, 0.03, 0.0024, 0.0, -0.0189, -0.0155, 0.0077, 0.0213, -0.0005, -0.0126, 0.0044, -0.0077, 0.0029, 0.0102, 0.0087, 0.0106, 0.001, -0.0416, -0.0314, -0.0387, -0.0261, -0.0131, 0.0068, -0.0039, 0.0314, 0.0431, 0.0314, 0.0237, 0.0464, 0.0324, 0.0179, 0.0556, 0.0861, 0.0914, 0.0885, 0.0997, 0.1205, 0.0919, 0.0919, 0.0972, 0.104, 0.1205, 0.1379, 0.1369, 0.1553, 0.1824, 0.2027, 0.1964, 0.2017, 0.1819, 0.1964, 0.2275, 0.2154, 0.2237, 0.2013, 0.1896, 0.1775, 0.1672, 0.174, 0.1945, 0.2125, 0.1906, 0.1882, 0.2057, 0.2115, 0.2684, 0.2767, 0.1867, 0.2061, 0.2159, 0.2178, 0.0943, 0.1167, 0.1123, 0.1317, 0.123, 0.1278, 0.1429, 0.176, 0.1794, 0.1804, 0.1872, 0.1877, 0.1916, 0.1458, 0.1531, 0.1381, 0.1405, 0.1614, 0.1133, 0.1419, 0.1152, 0.1103, 0.0325, 0.0165, 0.0019, -0.0263, -0.002, 0.0257, 0.0388, 0.0471, 0.0471, 0.0636, 0.0671, 0.0661, 0.0826, 0.1137, 0.1347, 0.1604, 0.2047, 0.2285, 0.1935, 0.1609, 0.1697, 0.1818, 0.1682, 0.1633, 0.1953, 0.2085, 0.2026, 0.1958, 0.1958, 0.1982, 0.1904, 0.1777, 0.1752, 0.1752, 0.1826, 0.1806, 0.164]}
{"unit_id": "quote:1965001356542128598", "ticker_idx": 0, "ts": "2025-09-08T10:36:46+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I don't want to buy ASML stock.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish on ASML; short-term growth limited and High-NA EUV demand won't scale until early 2030s.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "This analysis can be summarized as meaning that ASML’s short-term growth momentum is limited, and large-scale demand for High-NA EUV is unlikely to materialize until the early 2030s.\n\nI don’t want to buy ASML stock.", "reply_text": null, "parent_text": null, "quoted_text": "Barclays' bottom-up $ASML EUV model - muted outlook:\n\n\"Though we model large leading-edge capacity additions, we do not see any increase in EUV exposures at N2 vs N3, and expect a 25% increase at A14 from N2 (only 20% without backside power), but that could be optimistic given", "parent_author": null, "ret_m1d": -0.02423886812609788, "ret_signed_m1d": 0.02423886812609788, "alpha_spy_m1d": 0.0217883460276177, "alpha_c_m1d": 0.013312219390733793, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009754712891562045, "ret_signed_p1d": -0.009754712891562045, "alpha_spy_p1d": -0.0074427921939521635, "alpha_c_p1d": -0.005404205722981281, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.0780372496864421, "ret_signed_p1w": -0.0780372496864421, "alpha_spy_p1w": -0.059419125575291165, "alpha_c_p1w": -0.04212047837806776, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": 0.2929352197898367, "ret_signed_p1m": -0.2929352197898367, "alpha_spy_p1m": -0.2588043639399422, "alpha_c_p1m": -0.1562495939006232, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.4173657142183451, "ret_signed_p3m": -0.4173657142183451, "alpha_spy_p3m": -0.3596349578275255, "alpha_c_p3m": -0.19788634418086826, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": 0.7224543765252791, "ret_signed_p6m": -0.7224543765252791, "alpha_spy_p6m": -0.6678913404468887, "alpha_c_p6m": -0.399518831033727, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [0.0174, -0.0026, -0.0203, -0.0122, -0.0183, -0.0197, -0.0373, -0.0413, -0.0143, 0.0192, 0.0237, -0.0015, 0.0061, -0.0012, -0.015, -0.0012, -0.0019, -0.028, -0.01, 0.0007, 0.0021, 0.0171, 0.0081, 0.0131, 0.0408, -0.0775, -0.0416, -0.0662, -0.08, -0.112, -0.1058, -0.0921, -0.1083, -0.0714, -0.0792, -0.0726, -0.0939, -0.1206, -0.1097, -0.1175, -0.1237, -0.0974, -0.0882, -0.0733, -0.0631, -0.0532, -0.0498, -0.059, -0.056, -0.0507, -0.0568, -0.0622, -0.0449, -0.0442, -0.0384, -0.0242, -0.033, -0.0591, -0.0599, -0.0878, -0.0758, -0.0429, -0.0242, 0.0, 0.0098, 0.0053, 0.0139, 0.0195, 0.078, 0.0832, 0.0893, 0.1735, 0.1731, 0.1983, 0.2116, 0.1942, 0.1997, 0.2013, 0.2201, 0.2239, 0.2436, 0.2971, 0.3008, 0.3262, 0.2929, 0.2589, 0.256, 0.2029, 0.2474, 0.2513, 0.2903, 0.2968, 0.2943, 0.3287, 0.3151, 0.286, 0.312, 0.3219, 0.3422, 0.3435, 0.3613, 0.3881, 0.3593, 0.3718, 0.3592, 0.3423, 0.3256, 0.2893, 0.3128, 0.3131, 0.3204, 0.3051, 0.2942, 0.2976, 0.2816, 0.3127, 0.3174, 0.2348, 0.2715, 0.2737, 0.3463, 0.3284, 0.3376, 0.3721, 0.3904, 0.4264, 0.4174, 0.4089, 0.4261, 0.4109, 0.4006, 0.3931, 0.37, 0.3784, 0.3453, 0.294, 0.3217, 0.3349, 0.3266, 0.3373, 0.331, 0.331, 0.331, 0.3426, 0.3598, 0.3642, 0.3642, 0.4603, 0.5594, 0.5709, 0.5573, 0.4998, 0.6014, 0.6085, 0.6325, 0.6053, 0.7018, 0.7281, 0.6588, 0.6879, 0.7098, 0.7418, 0.7444, 0.7441, 0.8028, 0.7684, 0.7649, 0.7998, 0.8134, 0.7625, 0.6888, 0.7021, 0.7675, 0.7862, 0.7687, 0.7907, 0.7491, 0.7649, 0.7717, 0.7779, 0.8455, 0.8357, 0.8615, 0.852, 0.8731, 0.9101, 0.8271, 0.8286, 0.7945, 0.7225]}
{"unit_id": "orig:1949606756541403417", "ticker_idx": 0, "ts": "2025-07-27T23:04:07+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics (OW) preferred. 1x P/B level (undervalued) → stock defensibility despite short-term performance weakness. HBM3E/4 supply chain re-entry expands DRAM revenue leverage in 2026.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Asia Memory note: OW Samsung on valuation/HBM re-entry, EW Hynix on margin pressure, Neutral/cautious Micron on competition, positive Winbond on DDR4/NOR.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Technology: Memory Outlook – Bad Is Good, Good Is Not (Morgan Stanley)\n\n📉 Memory Market Outlook: Bad is Good, Good is Not\n\n(Key Analysis on HBM and Memory Cycle in 2026)\n\n🔑 Key Summary\n\n- 2026 will be a year to verify the sustainability of HBM (High Bandwidth Memory) profitability.\n- HBM and commodity memory (DRAM/NAND) prices are highly likely to decline over the next six months, and stock prices are expected to reflect this.\n\nHBM competition intensifies, leading to price pressure.\n\n- HBM3E (12-hi) 2026 unit price negotiation: Expected to be $1.56-$1.74/Gb next year, down from $1.88/Gb this year.\n- Samsung's potential re-entry into the Nvidia supply chain increases the risk of further declines due to aggressive pricing.\n- HBM4 initial negotiations also face continued price pressure due to intensifying competition.\n\nDRAM double-dip cycle is confirmed.\n- Inventory adjustment in 2H25-1H26 after early inventory stockpiling in 1H25.\n- PC/server DDR5 price negotiations delayed, customers refuse price increases.\n\nNAND remains range-bound.\n- Rebound of 5-10% in 3Q25, then decline again in 4Q25-1H26.\n- AI server (especially eSSD) demand is the only growth driver.\n\n🏭 HBM Detailed Outlook (Samsung Focus)\nSamsung Electronics' return to Nvidia supply chain becomes more visible.\n\n- HBM3E 12-hi: Customer sampling by end of August, mass production possible in Q4.\n- HBM4: Samples shipped in August-September, qualification expected in 4Q25.\n- 1c node ramp-up may secure process advantage over SK Hynix and Micron.\n\nOver-supply concerns expand (2026).\n- HBM bit supply +50% YoY (Samsung utilization 60% → 100%).\n- Demand is expected to be +67% (16bn Gb → 27bn Gb), but supply will exceed demand at 34bn Gb.\n- Inevitable price pressure, potentially falling to HBM3 levels ($1.7/Gb).\n\n🌀 DRAM Cycle (W-shaped trend)\n\n2H25 price weakness expected.\n- PC/server DDR5 attempting 3-8% price increases, but subject to adjustment due to customer resistance.\n- Legacy nodes like DDR4 and GDDR6 show sharp increases of 25-40% → supply reduction effect.\n\n1H26 inventory digestion period.\n- Early demand in 2Q25 due to tariff concerns → consumption contraction in 1H26.\n- AI demand sustained, but commodity DRAM continues a gradual downturn.\n\n💾 NAND Outlook\n\n2H25-2026 demand slowdown.\n- Smartphone demand slowdown (end of China subsidies) leads to eMMC/UFS price weakness.\n- PC demand also shows early purchases due to tariff risks, followed by contraction.\n- AI server eSSD and tray (4TB→8TB) increase are the only growth engines.\n\nYMTC (China) expands production to 200k WPM in 2026.\n- Full-scale supply to CSPs (Cloud Service Providers) with eSSD, bit growth +35-40% YoY.\n\n📊 Investment Implications\n\nSamsung Electronics (OW) preferred.\n- 1x P/B level (undervalued) → stock defensibility despite short-term performance weakness.\n- HBM3E/4 supply chain re-entry expands DRAM revenue leverage in 2026.\n\nSK Hynix (EW).\n- Maintains HBM leadership, but intensifying price competition leads to margin pressure.\n\nMicron (Neutral).\n- Valuation burden due to intensified competition in both HBM and DRAM.\nWinbond (NOR Flash/DDR4).\n- Strong DDR4 and NOR Flash prices expected to improve profitability in 2025.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06392037833579933, "ret_signed_m1d": -0.06392037833579933, "alpha_spy_m1d": -0.06417154154811011, "alpha_c_m1d": -0.056005973190194513, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.00791440514560482, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0028408408529752016, "ret_signed_p1d": 0.0028408408529752016, "alpha_spy_p1d": 0.005478586093461302, "alpha_c_p1d": 0.0028030150784925656, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 3.782577448263602e-05, "ret_p1w": -0.009943167469083525, "ret_signed_p1w": -0.009943167469083525, "alpha_spy_p1w": -0.0008840901899609399, "alpha_c_p1w": -0.0031650439477406422, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0067781235213428825, "ret_p1m": -0.0014204204264876008, "ret_signed_p1m": -0.0014204204264876008, "alpha_spy_p1m": -0.014325802828035084, "alpha_c_p1m": 0.002555579003573172, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": -0.003975999430060773, "ret_p3m": 0.37685443154336795, "ret_signed_p3m": 0.37685443154336795, "alpha_spy_p3m": 0.31926287124578745, "alpha_c_p3m": 0.28074134315242305, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.0961130883909449, "ret_p6m": 1.0817728912172333, "ret_signed_p6m": 1.0817728912172333, "alpha_spy_p6m": 1.0118664091672853, "alpha_c_p6m": 1.0045408978522918, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.07723199336494146, "price_path": [-0.2165, -0.2165, -0.2334, -0.2334, -0.2334, -0.2292, -0.2292, -0.2263, -0.1868, -0.1967, -0.1896, -0.191, -0.1981, -0.2122, -0.2108, -0.2136, -0.2277, -0.2348, -0.2277, -0.239, -0.2108, -0.208, -0.2066, -0.1981, -0.1981, -0.184, -0.1656, -0.1656, -0.1557, -0.1642, -0.1543, -0.16, -0.1769, -0.1925, -0.1797, -0.1557, -0.1642, -0.16, -0.1812, -0.1459, -0.1346, -0.1501, -0.1364, -0.1506, -0.1449, -0.1364, -0.0937, -0.1009, -0.1236, -0.1278, -0.142, -0.1335, -0.1108, -0.1122, -0.0952, -0.081, -0.0526, -0.0469, -0.0369, -0.0625, -0.0568, -0.0625, -0.0639, 0.0, 0.0028, 0.0313, 0.0142, -0.0213, -0.0099, -0.0071, -0.0227, 0.0014, 0.0199, 0.0085, 0.0099, 0.0213, 0.017, 0.017, -0.0057, -0.0057, 0.0014, 0.0028, 0.0142, 0.0156, -0.0014, 0.0028, -0.0114, -0.0099, -0.0398, -0.0185, -0.0085, -0.0043, -0.0128, -0.0043, 0.0156, 0.0313, 0.0426, 0.071, 0.0866, 0.1278, 0.1108, 0.1406, 0.1406, 0.1861, 0.2031, 0.2131, 0.223, 0.1832, 0.2014, 0.1971, 0.227, 0.2805, 0.2805, 0.2805, 0.2805, 0.2805, 0.2805, 0.3469, 0.3312, 0.3069, 0.3555, 0.394, 0.3968, 0.3997, 0.3911, 0.4068, 0.3769, 0.4097, 0.4553, 0.4197, 0.4339, 0.4853, 0.5338, 0.5852, 0.4967, 0.4354, 0.4154, 0.3968, 0.4354, 0.4767, 0.471, 0.4667, 0.3868, 0.4354, 0.3954, 0.3769, 0.4354, 0.3526, 0.3797, 0.4168, 0.4667, 0.4767, 0.4339, 0.4382, 0.4753, 0.491, 0.4996, 0.5466, 0.5623, 0.5466, 0.5409, 0.5309, 0.5538, 0.4953, 0.4667, 0.5395, 0.5352, 0.5167, 0.5766, 0.5909, 0.5852, 0.5852, 0.6693, 0.7133, 0.719, 0.719, 0.719, 0.8423, 0.98, 0.9914, 1.0216, 0.99, 0.9929, 0.99, 0.9728, 1.0115, 1.0631, 1.1348, 1.1406, 1.0818]}
{"unit_id": "orig:1949606756541403417", "ticker_idx": 1, "ts": "2025-07-27T23:04:07+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix (EW). Maintains HBM leadership, but intensifying price competition leads to margin pressure.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Asia Memory note: OW Samsung on valuation/HBM re-entry, EW Hynix on margin pressure, Neutral/cautious Micron on competition, positive Winbond on DDR4/NOR.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Technology: Memory Outlook – Bad Is Good, Good Is Not (Morgan Stanley)\n\n📉 Memory Market Outlook: Bad is Good, Good is Not\n\n(Key Analysis on HBM and Memory Cycle in 2026)\n\n🔑 Key Summary\n\n- 2026 will be a year to verify the sustainability of HBM (High Bandwidth Memory) profitability.\n- HBM and commodity memory (DRAM/NAND) prices are highly likely to decline over the next six months, and stock prices are expected to reflect this.\n\nHBM competition intensifies, leading to price pressure.\n\n- HBM3E (12-hi) 2026 unit price negotiation: Expected to be $1.56-$1.74/Gb next year, down from $1.88/Gb this year.\n- Samsung's potential re-entry into the Nvidia supply chain increases the risk of further declines due to aggressive pricing.\n- HBM4 initial negotiations also face continued price pressure due to intensifying competition.\n\nDRAM double-dip cycle is confirmed.\n- Inventory adjustment in 2H25-1H26 after early inventory stockpiling in 1H25.\n- PC/server DDR5 price negotiations delayed, customers refuse price increases.\n\nNAND remains range-bound.\n- Rebound of 5-10% in 3Q25, then decline again in 4Q25-1H26.\n- AI server (especially eSSD) demand is the only growth driver.\n\n🏭 HBM Detailed Outlook (Samsung Focus)\nSamsung Electronics' return to Nvidia supply chain becomes more visible.\n\n- HBM3E 12-hi: Customer sampling by end of August, mass production possible in Q4.\n- HBM4: Samples shipped in August-September, qualification expected in 4Q25.\n- 1c node ramp-up may secure process advantage over SK Hynix and Micron.\n\nOver-supply concerns expand (2026).\n- HBM bit supply +50% YoY (Samsung utilization 60% → 100%).\n- Demand is expected to be +67% (16bn Gb → 27bn Gb), but supply will exceed demand at 34bn Gb.\n- Inevitable price pressure, potentially falling to HBM3 levels ($1.7/Gb).\n\n🌀 DRAM Cycle (W-shaped trend)\n\n2H25 price weakness expected.\n- PC/server DDR5 attempting 3-8% price increases, but subject to adjustment due to customer resistance.\n- Legacy nodes like DDR4 and GDDR6 show sharp increases of 25-40% → supply reduction effect.\n\n1H26 inventory digestion period.\n- Early demand in 2Q25 due to tariff concerns → consumption contraction in 1H26.\n- AI demand sustained, but commodity DRAM continues a gradual downturn.\n\n💾 NAND Outlook\n\n2H25-2026 demand slowdown.\n- Smartphone demand slowdown (end of China subsidies) leads to eMMC/UFS price weakness.\n- PC demand also shows early purchases due to tariff risks, followed by contraction.\n- AI server eSSD and tray (4TB→8TB) increase are the only growth engines.\n\nYMTC (China) expands production to 200k WPM in 2026.\n- Full-scale supply to CSPs (Cloud Service Providers) with eSSD, bit growth +35-40% YoY.\n\n📊 Investment Implications\n\nSamsung Electronics (OW) preferred.\n- 1x P/B level (undervalued) → stock defensibility despite short-term performance weakness.\n- HBM3E/4 supply chain re-entry expands DRAM revenue leverage in 2026.\n\nSK Hynix (EW).\n- Maintains HBM leadership, but intensifying price competition leads to margin pressure.\n\nMicron (Neutral).\n- Valuation burden due to intensified competition in both HBM and DRAM.\nWinbond (NOR Flash/DDR4).\n- Strong DDR4 and NOR Flash prices expected to improve profitability in 2025.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01526695385491017, "ret_signed_m1d": 0.01526695385491017, "alpha_spy_m1d": 0.015015790642599391, "alpha_c_m1d": 0.028718884106526232, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0019083467743843396, "ret_signed_p1d": 0.0019083467743843396, "alpha_spy_p1d": 0.00454609201487044, "alpha_c_p1d": -0.0032046326107659784, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.015267253287969629, "ret_signed_p1w": -0.015267253287969629, "alpha_spy_p1w": -0.006208176008847044, "alpha_c_p1w": -0.011080707277195567, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": -0.0019084066609962314, "ret_signed_p1m": -0.0019084066609962314, "alpha_spy_p1m": -0.014813789062543714, "alpha_c_p1m": -0.020233067615690103, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.8289738974021368, "ret_signed_p3m": 0.8289738974021368, "alpha_spy_p3m": 0.7713823371045563, "alpha_c_p3m": 0.6457613239605728, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 1.8420077992730124, "ret_signed_p6m": 1.8420077992730124, "alpha_spy_p6m": 1.7721013172230644, "alpha_c_p6m": 1.4981804203147, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.3237, -0.3237, -0.2914, -0.2914, -0.2914, -0.2731, -0.275, -0.2757, -0.2571, -0.2437, -0.2152, -0.2361, -0.2209, -0.2403, -0.2304, -0.2361, -0.2498, -0.238, -0.2266, -0.2285, -0.2075, -0.1908, -0.2195, -0.208, -0.208, -0.1698, -0.1431, -0.1431, -0.126, -0.1202, -0.084, -0.1011, -0.1011, -0.0534, -0.0496, -0.0592, -0.0611, -0.0191, -0.0095, 0.063, 0.0916, 0.1183, 0.084, 0.1145, 0.0897, 0.0649, 0.063, 0.0324, 0.0344, 0.0763, 0.0725, 0.1336, 0.124, 0.145, 0.1393, 0.1298, 0.0286, 0.0267, 0.0401, 0.0248, 0.0267, 0.0286, 0.0153, 0.0, 0.0019, 0.0057, 0.0439, -0.0153, -0.0153, 0.0057, -0.0134, 0.0, -0.021, 0.0191, 0.0267, 0.0611, 0.0553, 0.0553, 0.021, 0.0038, -0.0248, -0.0649, -0.042, -0.0095, -0.0019, -0.0076, 0.0263, 0.0282, -0.0215, -0.0043, 0.0034, 0.0148, 0.0454, 0.0588, 0.1008, 0.162, 0.1734, 0.2556, 0.2652, 0.3302, 0.2747, 0.3588, 0.3588, 0.3416, 0.3799, 0.3665, 0.3627, 0.2862, 0.334, 0.3283, 0.376, 0.5117, 0.5117, 0.5117, 0.5117, 0.5117, 0.5117, 0.6359, 0.5863, 0.5729, 0.6149, 0.7296, 0.7793, 0.8557, 0.8309, 0.8404, 0.829, 0.9494, 1.0449, 0.9914, 1.1328, 1.1711, 1.1367, 1.3698, 1.2399, 1.2131, 1.2666, 1.2169, 1.3163, 1.366, 1.3584, 1.3393, 1.1405, 1.3163, 1.1787, 1.1481, 1.1825, 0.9914, 0.9876, 0.9838, 1.0029, 1.0808, 1.0273, 1.0579, 1.1344, 1.1114, 1.0732, 1.0808, 1.2071, 1.165, 1.2453, 1.1611, 1.1841, 1.1191, 1.0273, 1.1076, 1.1114, 1.0923, 1.2185, 1.2338, 1.2491, 1.2491, 1.2912, 1.448, 1.4901, 1.4901, 1.4901, 1.5896, 1.6622, 1.777, 1.8382, 1.8917, 1.8458, 1.865, 1.8229, 1.8382, 1.865, 1.8917, 1.9223, 1.842]}
{"unit_id": "orig:1949606756541403417", "ticker_idx": 2, "ts": "2025-07-27T23:04:07+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron (Neutral). Valuation burden due to intensified competition in both HBM and DRAM.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Asia Memory note: OW Samsung on valuation/HBM re-entry, EW Hynix on margin pressure, Neutral/cautious Micron on competition, positive Winbond on DDR4/NOR.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Technology: Memory Outlook – Bad Is Good, Good Is Not (Morgan Stanley)\n\n📉 Memory Market Outlook: Bad is Good, Good is Not\n\n(Key Analysis on HBM and Memory Cycle in 2026)\n\n🔑 Key Summary\n\n- 2026 will be a year to verify the sustainability of HBM (High Bandwidth Memory) profitability.\n- HBM and commodity memory (DRAM/NAND) prices are highly likely to decline over the next six months, and stock prices are expected to reflect this.\n\nHBM competition intensifies, leading to price pressure.\n\n- HBM3E (12-hi) 2026 unit price negotiation: Expected to be $1.56-$1.74/Gb next year, down from $1.88/Gb this year.\n- Samsung's potential re-entry into the Nvidia supply chain increases the risk of further declines due to aggressive pricing.\n- HBM4 initial negotiations also face continued price pressure due to intensifying competition.\n\nDRAM double-dip cycle is confirmed.\n- Inventory adjustment in 2H25-1H26 after early inventory stockpiling in 1H25.\n- PC/server DDR5 price negotiations delayed, customers refuse price increases.\n\nNAND remains range-bound.\n- Rebound of 5-10% in 3Q25, then decline again in 4Q25-1H26.\n- AI server (especially eSSD) demand is the only growth driver.\n\n🏭 HBM Detailed Outlook (Samsung Focus)\nSamsung Electronics' return to Nvidia supply chain becomes more visible.\n\n- HBM3E 12-hi: Customer sampling by end of August, mass production possible in Q4.\n- HBM4: Samples shipped in August-September, qualification expected in 4Q25.\n- 1c node ramp-up may secure process advantage over SK Hynix and Micron.\n\nOver-supply concerns expand (2026).\n- HBM bit supply +50% YoY (Samsung utilization 60% → 100%).\n- Demand is expected to be +67% (16bn Gb → 27bn Gb), but supply will exceed demand at 34bn Gb.\n- Inevitable price pressure, potentially falling to HBM3 levels ($1.7/Gb).\n\n🌀 DRAM Cycle (W-shaped trend)\n\n2H25 price weakness expected.\n- PC/server DDR5 attempting 3-8% price increases, but subject to adjustment due to customer resistance.\n- Legacy nodes like DDR4 and GDDR6 show sharp increases of 25-40% → supply reduction effect.\n\n1H26 inventory digestion period.\n- Early demand in 2Q25 due to tariff concerns → consumption contraction in 1H26.\n- AI demand sustained, but commodity DRAM continues a gradual downturn.\n\n💾 NAND Outlook\n\n2H25-2026 demand slowdown.\n- Smartphone demand slowdown (end of China subsidies) leads to eMMC/UFS price weakness.\n- PC demand also shows early purchases due to tariff risks, followed by contraction.\n- AI server eSSD and tray (4TB→8TB) increase are the only growth engines.\n\nYMTC (China) expands production to 200k WPM in 2026.\n- Full-scale supply to CSPs (Cloud Service Providers) with eSSD, bit growth +35-40% YoY.\n\n📊 Investment Implications\n\nSamsung Electronics (OW) preferred.\n- 1x P/B level (undervalued) → stock defensibility despite short-term performance weakness.\n- HBM3E/4 supply chain re-entry expands DRAM revenue leverage in 2026.\n\nSK Hynix (EW).\n- Maintains HBM leadership, but intensifying price competition leads to margin pressure.\n\nMicron (Neutral).\n- Valuation burden due to intensified competition in both HBM and DRAM.\nWinbond (NOR Flash/DDR4).\n- Strong DDR4 and NOR Flash prices expected to improve profitability in 2025.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 8.989993772945404e-05, "ret_signed_m1d": -8.989993772945404e-05, "alpha_spy_m1d": 0.00016126327458132472, "alpha_c_m1d": -0.013541830189345516, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006381934082124063, "ret_signed_p1d": -0.006381934082124063, "alpha_spy_p1d": -0.009019679322610163, "alpha_c_p1d": -0.0012689546969737453, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.03128092027317764, "ret_signed_p1w": 0.03128092027317764, "alpha_spy_p1w": 0.022221842994055052, "alpha_c_p1w": 0.027094374262403575, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": 0.047191011544621686, "ret_signed_p1m": -0.047191011544621686, "alpha_spy_p1m": -0.0342856291430742, "alpha_c_p1m": -0.028866350589927814, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.8592309826090665, "ret_signed_p3m": -0.8592309826090665, "alpha_spy_p3m": -0.801639422311486, "alpha_c_p3m": -0.6760184091675026, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 2.2842797425084473, "ret_signed_p6m": -2.2842797425084473, "alpha_spy_p6m": -2.2143732604584994, "alpha_c_p6m": -1.940452363550135, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.309, -0.3016, -0.2751, -0.2778, -0.277, -0.258, -0.2353, -0.229, -0.1712, -0.1295, -0.144, -0.1428, -0.1199, -0.1141, -0.119, -0.1393, -0.1484, -0.1615, -0.1615, -0.1345, -0.1363, -0.1307, -0.1517, -0.1183, -0.0818, -0.0728, -0.0455, -0.0251, -0.0036, 0.025, 0.042, 0.0433, 0.0381, 0.0762, 0.0807, 0.094, 0.094, 0.11, 0.0963, 0.1487, 0.1427, 0.1315, 0.1204, 0.1068, 0.0856, 0.0933, 0.0982, 0.0982, 0.0779, 0.1184, 0.0988, 0.1066, 0.1194, 0.0662, 0.0796, 0.0466, 0.0181, 0.0282, 0.0178, -0.0182, -0.0128, 0.0043, 0.0001, 0.0, 0.0064, 0.0314, -0.019, -0.0573, -0.0313, -0.0197, -0.0222, 0.0056, 0.0687, 0.1121, 0.1483, 0.117, 0.1262, 0.0865, 0.1106, 0.0971, 0.0536, 0.0408, 0.0578, 0.0465, 0.0472, 0.0584, 0.0966, 0.0698, 0.0698, 0.065, 0.0671, 0.1165, 0.1809, 0.1817, 0.2156, 0.2584, 0.3534, 0.4133, 0.4182, 0.4276, 0.4381, 0.5181, 0.4627, 0.4797, 0.4958, 0.4536, 0.4097, 0.4137, 0.4733, 0.504, 0.6373, 0.6517, 0.6894, 0.7176, 0.6702, 0.7678, 0.7299, 0.6334, 0.7338, 0.6825, 0.7264, 0.8216, 0.8203, 0.8598, 0.8195, 0.7851, 0.8592, 0.97, 0.9797, 0.9959, 1.0384, 1.0148, 1.0127, 1.111, 0.961, 1.1362, 1.1436, 1.1399, 1.2783, 1.1686, 1.2027, 1.1312, 1.2201, 1.1762, 1.0552, 1.032, 0.8112, 0.8652, 1.0141, 1.0195, 1.071, 1.071, 1.127, 1.1628, 1.1541, 1.1061, 1.0386, 1.1336, 1.2209, 1.2704, 1.3719, 1.3247, 1.1689, 1.1362, 1.0913, 1.0284, 1.2356, 1.3918, 1.4878, 1.4849, 1.5785, 1.5785, 1.5615, 1.6487, 1.6331, 1.5681, 1.5681, 1.8382, 1.8087, 2.0902, 2.0553, 1.9425, 2.1051, 2.1121, 2.0425, 1.9995, 2.029, 2.264, 2.264, 2.2843]}
{"unit_id": "orig:1949606756541403417", "ticker_idx": 3, "ts": "2025-07-27T23:04:07+00:00", "symbol": "Winbond", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Winbond (NOR Flash/DDR4). Strong DDR4 and NOR Flash prices expected to improve profitability in 2025.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Asia Memory note: OW Samsung on valuation/HBM re-entry, EW Hynix on margin pressure, Neutral/cautious Micron on competition, positive Winbond on DDR4/NOR.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Asia Technology: Memory Outlook – Bad Is Good, Good Is Not (Morgan Stanley)\n\n📉 Memory Market Outlook: Bad is Good, Good is Not\n\n(Key Analysis on HBM and Memory Cycle in 2026)\n\n🔑 Key Summary\n\n- 2026 will be a year to verify the sustainability of HBM (High Bandwidth Memory) profitability.\n- HBM and commodity memory (DRAM/NAND) prices are highly likely to decline over the next six months, and stock prices are expected to reflect this.\n\nHBM competition intensifies, leading to price pressure.\n\n- HBM3E (12-hi) 2026 unit price negotiation: Expected to be $1.56-$1.74/Gb next year, down from $1.88/Gb this year.\n- Samsung's potential re-entry into the Nvidia supply chain increases the risk of further declines due to aggressive pricing.\n- HBM4 initial negotiations also face continued price pressure due to intensifying competition.\n\nDRAM double-dip cycle is confirmed.\n- Inventory adjustment in 2H25-1H26 after early inventory stockpiling in 1H25.\n- PC/server DDR5 price negotiations delayed, customers refuse price increases.\n\nNAND remains range-bound.\n- Rebound of 5-10% in 3Q25, then decline again in 4Q25-1H26.\n- AI server (especially eSSD) demand is the only growth driver.\n\n🏭 HBM Detailed Outlook (Samsung Focus)\nSamsung Electronics' return to Nvidia supply chain becomes more visible.\n\n- HBM3E 12-hi: Customer sampling by end of August, mass production possible in Q4.\n- HBM4: Samples shipped in August-September, qualification expected in 4Q25.\n- 1c node ramp-up may secure process advantage over SK Hynix and Micron.\n\nOver-supply concerns expand (2026).\n- HBM bit supply +50% YoY (Samsung utilization 60% → 100%).\n- Demand is expected to be +67% (16bn Gb → 27bn Gb), but supply will exceed demand at 34bn Gb.\n- Inevitable price pressure, potentially falling to HBM3 levels ($1.7/Gb).\n\n🌀 DRAM Cycle (W-shaped trend)\n\n2H25 price weakness expected.\n- PC/server DDR5 attempting 3-8% price increases, but subject to adjustment due to customer resistance.\n- Legacy nodes like DDR4 and GDDR6 show sharp increases of 25-40% → supply reduction effect.\n\n1H26 inventory digestion period.\n- Early demand in 2Q25 due to tariff concerns → consumption contraction in 1H26.\n- AI demand sustained, but commodity DRAM continues a gradual downturn.\n\n💾 NAND Outlook\n\n2H25-2026 demand slowdown.\n- Smartphone demand slowdown (end of China subsidies) leads to eMMC/UFS price weakness.\n- PC demand also shows early purchases due to tariff risks, followed by contraction.\n- AI server eSSD and tray (4TB→8TB) increase are the only growth engines.\n\nYMTC (China) expands production to 200k WPM in 2026.\n- Full-scale supply to CSPs (Cloud Service Providers) with eSSD, bit growth +35-40% YoY.\n\n📊 Investment Implications\n\nSamsung Electronics (OW) preferred.\n- 1x P/B level (undervalued) → stock defensibility despite short-term performance weakness.\n- HBM3E/4 supply chain re-entry expands DRAM revenue leverage in 2026.\n\nSK Hynix (EW).\n- Maintains HBM leadership, but intensifying price competition leads to margin pressure.\n\nMicron (Neutral).\n- Valuation burden due to intensified competition in both HBM and DRAM.\nWinbond (NOR Flash/DDR4).\n- Strong DDR4 and NOR Flash prices expected to improve profitability in 2025.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002857157256717091, "ret_signed_m1d": -0.002857157256717091, "alpha_spy_m1d": -0.00310832046902787, "alpha_c_m1d": 0.010594772994898971, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005714314513434182, "ret_signed_p1d": -0.005714314513434182, "alpha_spy_p1d": -0.003076569272948082, "alpha_c_p1d": -0.0108272938985845, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.008571471770151273, "ret_signed_p1w": -0.008571471770151273, "alpha_spy_p1w": 0.0004876055089713116, "alpha_c_p1w": -0.004384925759377212, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": 0.07142849317001576, "ret_signed_p1m": 0.07142849317001576, "alpha_spy_p1m": 0.05852311076846828, "alpha_c_p1m": 0.05310383221532189, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 1.645714034034771, "ret_signed_p3m": 1.645714034034771, "alpha_spy_p3m": 1.5881224737371904, "alpha_c_p3m": 1.462501460593207, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 5.457142531587266, "ret_signed_p6m": 5.457142531587266, "alpha_spy_p6m": 5.387236049537318, "alpha_c_p6m": 5.113315152628953, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.1, -0.1, -0.0886, -0.1171, -0.1029, -0.1114, -0.1086, -0.1114, -0.0543, 0.0229, 0.0371, 0.0343, 0.0257, 0.0143, 0.06, 0.0371, 0.02, 0.0143, -0.0029, -0.0086, -0.0057, 0.0114, 0.0114, -0.0343, -0.0457, 0.0086, 0.0343, 0.0486, 0.0514, 0.0743, 0.0743, 0.0743, 0.0543, 0.0514, 0.0714, 0.1229, 0.12, 0.0829, 0.12, 0.1457, 0.1571, 0.1914, 0.2086, 0.1514, 0.1343, 0.1171, 0.1286, 0.0943, 0.0914, 0.08, 0.0829, 0.0571, 0.0543, 0.0429, 0.0229, 0.0086, 0.0229, 0.0086, -0.0057, -0.02, 0.0057, -0.0086, -0.0029, 0.0, -0.0057, 0.02, -0.0086, -0.0086, -0.0086, 0.0057, -0.0257, -0.0571, -0.0029, -0.0057, 0.0543, 0.0343, 0.0286, 0.0514, 0.0914, 0.0629, 0.0286, 0.0514, 0.04, 0.0514, 0.0714, 0.1314, 0.12, 0.1257, 0.16, 0.14, 0.1486, 0.1714, 0.2886, 0.4057, 0.4029, 0.4257, 0.4286, 0.4543, 0.5314, 0.5971, 0.68, 0.8457, 0.9143, 0.9571, 1.0829, 1.0114, 0.9086, 0.7943, 0.7943, 0.9457, 0.9314, 1.0571, 1.2171, 1.2171, 1.4371, 1.3571, 1.4829, 1.4829, 1.4571, 1.3257, 1.2857, 1.5143, 1.5114, 1.6486, 1.5486, 1.5971, 1.6457, 1.6457, 1.9086, 2.0857, 2.0857, 2.1543, 2.0971, 2.2343, 2.0457, 2.2, 2.3714, 2.32, 2.6514, 2.6857, 2.7029, 2.6686, 2.4457, 2.7829, 2.5086, 2.3943, 2.3143, 1.9829, 2.12, 2.2857, 2.0629, 2.2514, 2.3143, 2.2914, 2.2343, 2.1886, 2.3029, 2.5314, 2.88, 3.0743, 2.8571, 2.9086, 3.2629, 3.0629, 3.0343, 3.2229, 3.2571, 3.0743, 3.1771, 3.1714, 3.3943, 3.3943, 3.3714, 3.3943, 3.7829, 3.72, 3.72, 4.1886, 4.4571, 4.9429, 5.0857, 5.2, 4.5886, 4.8286, 4.7029, 4.7029, 4.6686, 5.0857, 5.6571, 5.4571]}
{"unit_id": "orig:1945263024098697624", "ticker_idx": 0, "ts": "2025-07-15T23:23:40+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix will benefit from the resumption of H20 exports. The absorption of HBM volume supplied for H20 in 1Q25 could accelerate with this export resumption, and given its proven quality stability, SK Hynix products are expected to be supplied predominantly.", "tweet_type": "analysis", "timeline": "months", "summary": "Long SK Hynix as primary H20 HBM beneficiary on export resumption; long Samsung on H20 HBM spillover, GDDR7 for B30, and MI308 supply wins.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA and AMD are expected to resume GPU exports to China, correct? Currently, SK Hynix is the primary supplier of HBM for H20.\n\n1. Regarding H20: I believe SK Hynix will benefit from the resumption of H20 exports. The absorption of HBM volume supplied for H20 in 1Q25 could accelerate with this export resumption, and given its proven quality stability, SK Hynix products are expected to be supplied predominantly. \n\nThe next likely beneficiary appears to be Samsung Electronics. Micron, from what I understand, is not currently addressing NVIDIA's H20 demand. \n\nFor Samsung, a pre-production and then sales strategy might even help resolve the accumulated HBM inventory issues. I understand Samsung currently holds approximately 1 trillion KRW worth of HBM3E inventory.\n\n2. Regarding B30 (RTX Pro 6000D): This product will feature GDDR7, and Samsung is estimated to supply most of it this year.\n\n3. Regarding MI308: Samsung Electronics is also estimated to largely handle this.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005024887902465247, "ret_signed_m1d": 0.005024887902465247, "alpha_spy_m1d": 0.0007333825324935539, "alpha_c_m1d": 0.02386938884652723, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00837530376613238, "ret_signed_p1d": -0.00837530376613238, "alpha_spy_p1d": -0.01171849488464749, "alpha_c_p1d": -0.003148328857464744, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.10050269904539955, "ret_signed_p1w": -0.10050269904539955, "alpha_spy_p1w": -0.11130408246621493, "alpha_c_p1w": -0.08183010412969138, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.06867690216785682, "ret_signed_p1m": -0.06867690216785682, "alpha_spy_p1m": -0.10524423609092248, "alpha_c_p1m": -0.10471547985969676, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.43590666016838076, "ret_signed_p3m": 0.43590666016838076, "alpha_spy_p3m": 0.3833615420278167, "alpha_c_p3m": 0.3171308584736168, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 1.4911352222375944, "ret_signed_p6m": 1.4911352222375944, "alpha_spy_p6m": 1.3763779189564074, "alpha_c_p6m": 1.1630127224018156, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.4148, -0.4148, -0.4094, -0.4188, -0.3947, -0.4038, -0.3834, -0.3914, -0.3954, -0.4064, -0.4064, -0.378, -0.378, -0.378, -0.362, -0.3636, -0.3643, -0.3479, -0.3362, -0.3111, -0.3295, -0.3161, -0.3332, -0.3245, -0.3295, -0.3416, -0.3312, -0.3212, -0.3228, -0.3044, -0.2898, -0.3149, -0.3049, -0.3049, -0.2714, -0.2479, -0.2479, -0.2328, -0.2278, -0.196, -0.2111, -0.2111, -0.1692, -0.1658, -0.1742, -0.1759, -0.139, -0.1307, -0.067, -0.0419, -0.0184, -0.0486, -0.0218, -0.0436, -0.0653, -0.067, -0.0938, -0.0921, -0.0553, -0.0586, -0.005, -0.0134, 0.005, 0.0, -0.0084, -0.0972, -0.0988, -0.0871, -0.1005, -0.0988, -0.0972, -0.1089, -0.1223, -0.1206, -0.1173, -0.0838, -0.1357, -0.1357, -0.1173, -0.134, -0.1223, -0.1407, -0.1055, -0.0988, -0.0687, -0.0737, -0.0737, -0.1039, -0.1189, -0.1441, -0.1792, -0.1591, -0.1307, -0.124, -0.129, -0.0992, -0.0975, -0.1411, -0.126, -0.1193, -0.1093, -0.0824, -0.0707, -0.0338, 0.0199, 0.03, 0.1021, 0.1105, 0.1675, 0.1189, 0.1927, 0.1927, 0.1776, 0.2111, 0.1994, 0.196, 0.1289, 0.1709, 0.1658, 0.2078, 0.3269, 0.3269, 0.3269, 0.3269, 0.3269, 0.3269, 0.4359, 0.3923, 0.3806, 0.4175, 0.5181, 0.5617, 0.6288, 0.607, 0.6154, 0.6053, 0.711, 0.7949, 0.7479, 0.872, 0.9056, 0.8754, 1.0801, 0.966, 0.9425, 0.9895, 0.9459, 1.0331, 1.0767, 1.07, 1.0532, 0.8788, 1.0331, 0.9123, 0.8855, 0.9157, 0.7479, 0.7446, 0.7412, 0.758, 0.8264, 0.7794, 0.8062, 0.8734, 0.8532, 0.8197, 0.8264, 0.9372, 0.9002, 0.9707, 0.8969, 0.917, 0.86, 0.7794, 0.8499, 0.8532, 0.8365, 0.9472, 0.9607, 0.9741, 0.9741, 1.011, 1.1487, 1.1856, 1.1856, 1.1856, 1.2729, 1.3367, 1.4374, 1.4911]}
{"unit_id": "orig:1945263024098697624", "ticker_idx": 1, "ts": "2025-07-15T23:23:40+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The next likely beneficiary appears to be Samsung Electronics... a pre-production and then sales strategy might even help resolve the accumulated HBM inventory issues... Samsung is estimated to supply most of it this year... Samsung Electronics is also estimated to largely handle this.", "tweet_type": "analysis", "timeline": "months", "summary": "Long SK Hynix as primary H20 HBM beneficiary on export resumption; long Samsung on H20 HBM spillover, GDDR7 for B30, and MI308 supply wins.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA and AMD are expected to resume GPU exports to China, correct? Currently, SK Hynix is the primary supplier of HBM for H20.\n\n1. Regarding H20: I believe SK Hynix will benefit from the resumption of H20 exports. The absorption of HBM volume supplied for H20 in 1Q25 could accelerate with this export resumption, and given its proven quality stability, SK Hynix products are expected to be supplied predominantly. \n\nThe next likely beneficiary appears to be Samsung Electronics. Micron, from what I understand, is not currently addressing NVIDIA's H20 demand. \n\nFor Samsung, a pre-production and then sales strategy might even help resolve the accumulated HBM inventory issues. I understand Samsung currently holds approximately 1 trillion KRW worth of HBM3E inventory.\n\n2. Regarding B30 (RTX Pro 6000D): This product will feature GDDR7, and Samsung is estimated to supply most of it this year.\n\n3. Regarding MI308: Samsung Electronics is also estimated to largely handle this.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01883834739473611, "ret_signed_m1d": -0.01883834739473611, "alpha_spy_m1d": -0.023129852764707803, "alpha_c_m1d": -0.010036741140800665, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.008801606253935446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015698581479784757, "ret_signed_p1d": 0.015698581479784757, "alpha_spy_p1d": 0.012355390361269647, "alpha_c_p1d": 0.012596800581375467, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.00310178089840929, "ret_p1w": 0.03610674980825368, "ret_signed_p1w": 0.03610674980825368, "alpha_spy_p1w": 0.0253053663874383, "alpha_c_p1w": 0.0326171868946612, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.003489562913592481, "ret_p1m": 0.12872829370574368, "ret_signed_p1m": 0.12872829370574368, "alpha_spy_p1m": 0.09216095978267802, "alpha_c_p1m": 0.08840431038607299, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.040323983319670686, "ret_p3m": 0.48855841811668865, "ret_signed_p3m": 0.48855841811668865, "alpha_spy_p3m": 0.4360132999761246, "alpha_c_p3m": 0.40776360755248, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.08079481056420867, "ret_p6m": 1.2341847828716226, "ret_signed_p6m": 1.2341847828716226, "alpha_spy_p6m": 1.1194274795904355, "alpha_c_p6m": 1.0947109736023968, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.1394738092692258, "price_path": [-0.1403, -0.1372, -0.1356, -0.1418, -0.1309, -0.1309, -0.1309, -0.1294, -0.1294, -0.134, -0.134, -0.1528, -0.1528, -0.1528, -0.1481, -0.1481, -0.145, -0.1013, -0.1122, -0.1044, -0.106, -0.1138, -0.1294, -0.1278, -0.1309, -0.1465, -0.1543, -0.1465, -0.159, -0.1278, -0.1247, -0.1231, -0.1138, -0.1138, -0.0982, -0.0779, -0.0779, -0.0669, -0.0763, -0.0654, -0.0716, -0.0904, -0.1075, -0.0935, -0.0669, -0.0763, -0.0716, -0.095, -0.056, -0.0435, -0.0607, -0.0455, -0.0612, -0.0549, -0.0455, 0.0016, -0.0063, -0.0314, -0.0361, -0.0518, -0.0424, -0.0173, -0.0188, 0.0, 0.0157, 0.0471, 0.0534, 0.0644, 0.0361, 0.0424, 0.0361, 0.0345, 0.1052, 0.1083, 0.1397, 0.1209, 0.0816, 0.0942, 0.0973, 0.0801, 0.1068, 0.1272, 0.1146, 0.1162, 0.1287, 0.124, 0.124, 0.0989, 0.0989, 0.1068, 0.1083, 0.1209, 0.1224, 0.1036, 0.1083, 0.0926, 0.0942, 0.0612, 0.0848, 0.0958, 0.1005, 0.0911, 0.1005, 0.1224, 0.1397, 0.1523, 0.1837, 0.2009, 0.2465, 0.2276, 0.2606, 0.2606, 0.3108, 0.3297, 0.3407, 0.3516, 0.3077, 0.3277, 0.323, 0.3561, 0.4152, 0.4152, 0.4152, 0.4152, 0.4152, 0.4152, 0.4886, 0.4712, 0.4444, 0.498, 0.5406, 0.5437, 0.5469, 0.5374, 0.5548, 0.5217, 0.5579, 0.6084, 0.569, 0.5847, 0.6415, 0.6951, 0.7519, 0.6541, 0.5863, 0.5642, 0.5437, 0.5863, 0.6321, 0.6257, 0.621, 0.5327, 0.5863, 0.5422, 0.5217, 0.5863, 0.4949, 0.5248, 0.5658, 0.621, 0.6321, 0.5847, 0.5895, 0.6305, 0.6478, 0.6573, 0.7093, 0.7267, 0.7093, 0.703, 0.692, 0.7172, 0.6526, 0.621, 0.7014, 0.6967, 0.6762, 0.7424, 0.7582, 0.7519, 0.7519, 0.8449, 0.8935, 0.8998, 0.8998, 0.8998, 1.0361, 1.1882, 1.2009, 1.2342]}
{"unit_id": "thread:1951163169117966496", "ticker_idx": 0, "ts": "2025-08-01T06:08:45+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "starting from the second half, the company's overall quarterly operating profit is expected to increase due to improved foundry utilization", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung foundry utilization at Pyeongtaek hitting maximum; legacy process recovery expected to shrink annual deficit and lift operating profit H2.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: [Exclusive] Samsung Electronics' Pyeongtaek Foundry Utilization Rate 'Hits Maximum'... Secures Series of Additional Orders in Mature Processes\n\nIt has been determined that the utilization rate of the legacy (mature) process production lines, which had been holding back Samsung Electronics' foundry (semiconductor contract manufacturing) business, has entered a recovery phase. As mass production of newly won orders in legacy processes—previously cited as the cause of the foundry division's large operating losses—begins in earnest, forecasts suggest that the annual deficit will shrink.\n\nAccording to industry sources on the 1st, the utilization rate of the production lines at Samsung's Pyeongtaek Campus P2 and P3, which had fallen below 50% last year, has reportedly reached its maximum level recently. These production lines are responsible for legacy processes such as 4nm (nanometer, 1 billionth of a meter), 5nm, and 7nm. It is also understood that the utilization rate of the legacy production line at Samsung's Giheung Campus is recovering rapidly as the input of mass production volume for the 8nm process increases.\n\nSamsung's legacy production lines were identified as the main culprit for the large operating losses after suffering from low utilization in the first half of the year. During its 2Q earnings conference call on the 31st, Samsung Electronics stated, \"We posted sluggish results as the low utilization of mature process lines continued.\" A mature process refers to a legacy process that has been in mass production for a long time, with stabilized yield and performance. Recently, the 4nm process has also begun to be classified as a legacy process. Samsung's foundry division is known to have recorded an operating loss in the 2 trillion KRW range in the second quarter of this year.\n\nIt is analyzed that profitability was hit after shipments of products for Baidu, a major 4nm process client for Samsung, were halted due to the impact of U.S. sanctions against China. Samsung had been mass-producing Baidu's artificial intelligence (AI) chips using its 4nm process. In its earnings call, Samsung stated, \"An inventory provision was established due to the impact of sanctions against China on AI chips.\"\n\nHowever, it is reported that since June this year, mass production volume has been continuously fed into the legacy production lines, bringing the utilization rate to its maximum. The analysis suggests that the utilization rate has gradually recovered through the mass production of the 'base die' for the 6th generation High Bandwidth Memory (HBM4) developed by Samsung's Memory Division, along with cryptocurrency miners for a Chinese client, Google's mobile application processors (AP), and previous-generation APs from the System LSI Division.\n\nThe utilization rate of the legacy production line at Samsung's Giheung Campus is also understood to be increasing. In particular, mass production of orders from domestic and international clients for the 8nm process is reportedly beginning in earnest. A semiconductor industry official stated, \"In addition to advanced processes of 3nm and below, Samsung has been competing for orders by advancing its legacy processes where it judges it has a competitive edge over TSMC to improve performance,\" adding, \"It's understood that mass production input is increasing, especially in the 4nm, 5nm, and 8nm processes.\"\n\nWith the rising utilization of legacy processes, which had been dragging down the performance of Samsung's Device Solutions (DS) division, analysis suggests that the deficit can be reduced in the second half of the year. Seo Seung-yeon, an analyst at DB Financial Investment, said, \"In the second quarter, the operating deficit widened compared to the previous quarter due to a slower-than-expected improvement in foundry utilization and inventory provisions from sanctions on AI semiconductors for China. However, starting from the second half, the company's overall quarterly operating profit is expected to increase due to improved foundry utilization.\"\n\nA Samsung Electronics official explained, \"We cannot confirm specific figures regarding the foundry division's utilization status,\" but added, \"it is true that the foundry business is also expected to follow a 'sangjeohago' (weak in the first half, strong in the second) trend of gradual recovery in the second half.\"\n\n---\n\nhttps://t.co/OEEvsafy31", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03628451801845611, "ret_signed_m1d": 0.03628451801845611, "alpha_spy_m1d": 0.019620828077359542, "alpha_c_m1d": 0.014228266800442535, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.02205625121801358, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01161109622749179, "ret_signed_p1d": 0.01161109622749179, "alpha_spy_p1d": -0.0035887179614699782, "alpha_c_p1d": -0.008694791299870719, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02030588752736251, "ret_p1w": 0.042090123474913144, "ret_signed_p1w": 0.042090123474913144, "alpha_spy_p1w": 0.017223541503019302, "alpha_c_p1w": 0.007663562763754239, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.034426560711158904, "ret_p1m": -0.01886781633450707, "ret_signed_p1m": -0.01886781633450707, "alpha_spy_p1m": -0.05639287429155182, "alpha_c_p1m": -0.039796134858980814, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.020928318524473744, "ret_p3m": 0.4651436595476073, "ret_signed_p3m": 0.4651436595476073, "alpha_spy_p3m": 0.3564519244954558, "alpha_c_p3m": 0.28056009055783004, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.18458356898977724, "ret_p6m": 1.2281755934741243, "ret_signed_p6m": 1.2281755934741243, "alpha_spy_p6m": 1.1075694978265855, "alpha_c_p6m": 1.0884890617390235, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.13968653173510082, "price_path": [-0.2167, -0.2124, -0.2124, -0.2095, -0.1691, -0.1792, -0.172, -0.1734, -0.1806, -0.1951, -0.1936, -0.1965, -0.2109, -0.2181, -0.2109, -0.2225, -0.1936, -0.1907, -0.1893, -0.1806, -0.1806, -0.1662, -0.1475, -0.1475, -0.1374, -0.146, -0.1359, -0.1417, -0.159, -0.1749, -0.1619, -0.1374, -0.146, -0.1417, -0.1633, -0.1273, -0.1157, -0.1316, -0.1176, -0.1321, -0.1263, -0.1176, -0.074, -0.0813, -0.1045, -0.1089, -0.1234, -0.1147, -0.0914, -0.0929, -0.0755, -0.061, -0.0319, -0.0261, -0.016, -0.0421, -0.0363, -0.0421, -0.0435, 0.0218, 0.0247, 0.0537, 0.0363, 0.0, 0.0116, 0.0145, -0.0015, 0.0232, 0.0421, 0.0305, 0.0319, 0.0435, 0.0392, 0.0392, 0.016, 0.016, 0.0232, 0.0247, 0.0363, 0.0377, 0.0203, 0.0247, 0.0102, 0.0116, -0.0189, 0.0029, 0.0131, 0.0174, 0.0087, 0.0174, 0.0377, 0.0537, 0.0653, 0.0943, 0.1103, 0.1524, 0.135, 0.1655, 0.1655, 0.2119, 0.2293, 0.2395, 0.2496, 0.209, 0.2275, 0.2231, 0.2538, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3762, 0.3602, 0.3354, 0.385, 0.4243, 0.4272, 0.4302, 0.4214, 0.4374, 0.4068, 0.4404, 0.487, 0.4506, 0.4651, 0.5176, 0.5672, 0.6197, 0.5293, 0.4666, 0.4462, 0.4272, 0.4666, 0.5089, 0.503, 0.4987, 0.417, 0.4666, 0.4258, 0.4068, 0.4666, 0.382, 0.4097, 0.4476, 0.4987, 0.5089, 0.4651, 0.4695, 0.5074, 0.5235, 0.5322, 0.5803, 0.5964, 0.5803, 0.5745, 0.5643, 0.5876, 0.5278, 0.4987, 0.573, 0.5687, 0.5497, 0.6109, 0.6255, 0.6197, 0.6197, 0.7057, 0.7506, 0.7565, 0.7565, 0.7565, 0.8824, 1.0231, 1.0348, 1.0656, 1.0333, 1.0363, 1.0333, 1.0158, 1.0553, 1.1081, 1.1813, 1.1872, 1.1271, 1.1901, 1.2311, 1.2282, 1.2282]}
{"unit_id": "orig:1945267218767602162", "ticker_idx": 0, "ts": "2025-07-15T23:40:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EPS is expected to increase by $0.40-$0.50, which represents more than a 10% upside to current market consensus estimates", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Bernstein estimate: NVDA China revenue recovery of $15-20B in FY2026 could add $0.40-$0.50 EPS, 10%+ above consensus.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein Estimates:\n\nFor every $10 billion increase in NVIDIA's revenue from the Chinese market, there could be an EPS uplift of approximately $0.25. If NVIDIA can recover $15-20 billion in revenue in FY2026 (ending January 2026), EPS is expected to increase by $0.40-$0.50, which represents more than a 10% upside to current market consensus estimates.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03884002840913159, "ret_signed_m1d": -0.03884002840913159, "alpha_spy_m1d": -0.04313153377910328, "alpha_c_m1d": -0.019995527465069607, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00392503050477, "ret_signed_p1d": 0.00392503050477, "alpha_spy_p1d": 0.0005818393862548898, "alpha_c_p1d": 0.009152005413437636, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.021499604475081924, "ret_signed_p1w": -0.021499604475081924, "alpha_spy_p1w": -0.032300987895897304, "alpha_c_p1w": -0.002827009559373761, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": 0.06379609510335871, "ret_signed_p1m": 0.06379609510335871, "alpha_spy_p1m": 0.02722876118029305, "alpha_c_p1m": 0.027757517411518773, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.07305409607893898, "ret_signed_p3m": 0.07305409607893898, "alpha_spy_p3m": 0.020508977938374917, "alpha_c_p3m": -0.045721705615825003, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.10797419003219977, "ret_signed_p6m": 0.10797419003219977, "alpha_spy_p6m": -0.0067831132489872825, "alpha_c_p6m": -0.22014830980357902, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.4055, -0.4055, -0.4323, -0.4207, -0.3983, -0.3766, -0.3497, -0.3631, -0.3614, -0.362, -0.3462, -0.3293, -0.3333, -0.3349, -0.3143, -0.3125, -0.3167, -0.2795, -0.2389, -0.2072, -0.2102, -0.2069, -0.2059, -0.2128, -0.2279, -0.2219, -0.2309, -0.2309, -0.2063, -0.2103, -0.1846, -0.2084, -0.1953, -0.1728, -0.1687, -0.18, -0.1698, -0.1645, -0.1567, -0.1633, -0.1506, -0.1683, -0.1524, -0.1557, -0.1477, -0.1477, -0.1573, -0.1554, -0.1336, -0.096, -0.0919, -0.0759, -0.0745, -0.1019, -0.0788, -0.0665, -0.0665, -0.073, -0.0627, -0.0458, -0.0387, -0.0339, -0.0388, 0.0, 0.0039, 0.0135, 0.01, 0.004, -0.0215, 0.0005, 0.0178, 0.0164, 0.0354, 0.0282, 0.0502, 0.042, 0.0177, 0.0545, 0.0443, 0.0511, 0.059, 0.0703, 0.0665, 0.073, 0.0638, 0.0663, 0.0571, 0.0663, 0.0289, 0.0275, 0.0251, 0.0427, 0.0534, 0.0649, 0.0639, 0.0555, 0.0204, 0.0204, 0.0005, -0.0005, 0.0056, -0.0216, -0.014, 0.0004, 0.0388, 0.038, 0.0418, 0.0414, 0.0245, -0.0023, 0.0325, 0.035, 0.0757, 0.0453, 0.0368, 0.041, 0.0439, 0.0654, 0.0931, 0.097, 0.1066, 0.0992, 0.087, 0.0841, 0.1079, 0.1282, 0.0731, 0.1033, 0.0547, 0.0535, 0.0651, 0.0734, 0.07, 0.0613, 0.0562, 0.0672, 0.0912, 0.1219, 0.1777, 0.213, 0.1886, 0.1863, 0.212, 0.164, 0.1436, 0.1019, 0.1023, 0.1661, 0.1316, 0.1354, 0.0947, 0.1141, 0.0932, 0.0625, 0.0927, 0.0583, 0.048, 0.0695, 0.0418, 0.0561, 0.0561, 0.037, 0.0541, 0.0631, 0.0521, 0.0744, 0.0687, 0.0871, 0.0837, 0.0767, 0.06, 0.0254, 0.0329, 0.0412, 0.0015, 0.0203, 0.0604, 0.0762, 0.1086, 0.105, 0.105, 0.1163, 0.1028, 0.0988, 0.0927, 0.0927, 0.1065, 0.1022, 0.097, 0.108]}
{"unit_id": "orig:1953627407237493123", "ticker_idx": 0, "ts": "2025-08-08T01:20:45+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "12-month target price raised to $200 (from $185), P/E 35x", "tweet_type": "news_reaction", "timeline": "months", "summary": "GS raises NVDA TP from $185 to $200 on higher hyperscaler CapEx, Blackwell ramp, and raised FY27 EPS to $6.75.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA Preview, TP Raised from $185 → $200 (GS)\n\n1. Earnings Outlook and Guidance\n\n• FY2Q26 and FY3Q26 revenue forecasts are $41.9bn and $51.5bn, respectively, exceeding consensus by 2% and 8%\n\n• Sales and EPS estimates for FY26–28 raised by an average of 7%\n\n• Reflecting expected hyperscaler CapEx increases in FY27/28 (+$60bn and +$90bn, respectively) → Nvidia Datacenter revenue estimates raised by +$24bn and +$36bn\n\n• AI startup contribution is reflected more conservatively (FY27/28 lowered by -$4bn/-$6bn)\n\n• Contribution from China has not yet been factored into estimates, but if export licenses are approved, there is potential for FY27 revenue/EPS to be boosted by +$20bn / +$0.40\n\n2. Key Earnings Events and Stock Reaction Points\n\n• In FY3Q guidance, the ramp-up speed and scale of the Blackwell product lineup will have the greatest short-term stock impact\n\n• If there is mention of the timing and scale of resuming H20 shipments to China, the stock could react\n\n• Need to watch gross margin: if previously accrued H20 inventory (~$2.5bn) is recognized, there is room for profitability improvement\n\n3. Mid- to Long-Term Growth and Valuation\n\n• Investor focus is shifting from upside in 2026 to sustainability of growth in 2027\n\n• FY27 EPS estimate raised to $6.75 (from $6.29)\n\n• 12-month target price raised to $200 (from $185), P/E 35x\n\n4. Investor Sentiment and Market Expectations\n\n• Most investors in the market are holding long positions ahead of this earnings release, and expectations are already high\n\n• As a result, market expectations for 2H25 guidance are very high, and a sensitive reaction is possible if guidance falls short", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010563727646062837, "ret_signed_m1d": -0.010563727646062837, "alpha_spy_m1d": -0.0028264858982657293, "alpha_c_m1d": -0.002353314293894515, "bench_spy_m1d": -0.007737241747797108, "bench_c_m1d": -0.008210413352168322, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0035030046935885517, "ret_signed_p1d": -0.0035030046935885517, "alpha_spy_p1d": -0.001525561457173752, "alpha_c_p1d": -0.0030600808119400913, "bench_spy_p1d": -0.0019774432364148, "bench_c_p1d": -0.0004429238816484604, "ret_p1w": -0.012315229992857168, "ret_signed_p1w": -0.012315229992857168, "alpha_spy_p1w": -0.02213984806954772, "alpha_c_p1w": -0.019537694691656893, "bench_spy_p1w": 0.00982461807669055, "bench_c_p1w": 0.007222464698799724, "ret_p1m": -0.0787629755726067, "ret_signed_p1m": -0.0787629755726067, "alpha_spy_p1m": -0.09704664564929222, "alpha_c_p1m": -0.08894928623151199, "bench_spy_p1m": 0.01828367007668552, "bench_c_p1m": 0.010186310658905295, "ret_p3m": 0.06853329303741895, "ret_signed_p3m": 0.06853329303741895, "alpha_spy_p3m": 0.002180653759290996, "alpha_c_p3m": -0.15621381432130188, "bench_spy_p3m": 0.06635263927812796, "bench_c_p3m": 0.22474710735872083, "ret_p6m": 0.016041683216476033, "ret_signed_p6m": 0.016041683216476033, "alpha_spy_p6m": -0.0816050820505787, "alpha_c_p6m": -0.37815806846665034, "bench_spy_p6m": 0.09764676526705474, "bench_c_p6m": 0.3941997516831264, "price_path": [-0.2889, -0.2593, -0.2621, -0.2589, -0.258, -0.2645, -0.2786, -0.273, -0.2814, -0.2814, -0.2584, -0.2622, -0.2382, -0.2604, -0.2481, -0.2271, -0.2233, -0.2338, -0.2244, -0.2194, -0.2121, -0.2182, -0.2063, -0.2229, -0.208, -0.2112, -0.2037, -0.2037, -0.2126, -0.2109, -0.1905, -0.1554, -0.1515, -0.1366, -0.1352, -0.1609, -0.1393, -0.1279, -0.1279, -0.1339, -0.1242, -0.1085, -0.1018, -0.0973, -0.102, -0.0657, -0.062, -0.0531, -0.0563, -0.062, -0.0858, -0.0652, -0.049, -0.0504, -0.0326, -0.0394, -0.0188, -0.0264, -0.0492, -0.0148, -0.0243, -0.018, -0.0106, 0.0, -0.0035, 0.0025, -0.0061, -0.0037, -0.0123, -0.0038, -0.0386, -0.04, -0.0423, -0.0258, -0.0158, -0.0051, -0.006, -0.0138, -0.0466, -0.0466, -0.0652, -0.0661, -0.0604, -0.0858, -0.0788, -0.0654, -0.0294, -0.0302, -0.0267, -0.027, -0.0427, -0.0679, -0.0353, -0.033, 0.005, -0.0233, -0.0313, -0.0274, -0.0246, -0.0046, 0.0213, 0.0249, 0.0339, 0.027, 0.0156, 0.0129, 0.0351, 0.0541, 0.0026, 0.0308, -0.0146, -0.0157, -0.0048, 0.0029, -0.0003, -0.0084, -0.0132, -0.0029, 0.0195, 0.0482, 0.1004, 0.1333, 0.1106, 0.1084, 0.1324, 0.0876, 0.0685, 0.0295, 0.0299, 0.0896, 0.0573, 0.0608, 0.0228, 0.0409, 0.0214, -0.0073, 0.021, -0.0112, -0.0209, -0.0008, -0.0267, -0.0133, -0.0133, -0.0311, -0.0152, -0.0067, -0.017, 0.0038, -0.0015, 0.0157, 0.0125, 0.006, -0.0096, -0.0419, -0.035, -0.0271, -0.0643, -0.0467, -0.0092, 0.0055, 0.0357, 0.0325, 0.0325, 0.043, 0.0303, 0.0266, 0.0209, 0.0209, 0.0338, 0.0298, 0.025, 0.0352, 0.0129, 0.0119, 0.0124, 0.0171, 0.0025, 0.0239, 0.0194, 0.0194, -0.0252, 0.0035, 0.0118, 0.0273, 0.0207, 0.032, 0.0484, 0.0538, 0.0463, 0.016]}
{"unit_id": "orig:1945588486666588437", "ticker_idx": 0, "ts": "2025-07-16T20:56:57+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is generally a significant positive for Nvidia's stock, especially considering that China AI revenue was set at a zero baseline in previous financial reports", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley (adopted): H20/RTX PRO China license resumption is structurally positive for NVDA; AMD, AVGO, ALAB, MRVL also benefit from licensing loosening.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Expected to Bring Structural Revenue Upside, but Short-term Supply and Licensing Processes Remain Uncertain\nMorgan Stanley (J. Moore, 25/07/16)\n\nMorgan Stanley noted that the U.S. government is expected to issue licenses to Nvidia to resume exports of H20 and RTX PRO GPU products to the Chinese market. Although this change still faces short-term supply and process obstacles, it is generally a significant positive for Nvidia's stock, especially considering that China AI revenue was set at a zero baseline in previous financial reports.\n\nU.S. Government Plans to Issue H20 Export Licenses, Potentially Restarting China Market Revenue Contribution\nNvidia announced it is awaiting U.S. government approval to resume sales of H20 and RTX Pro products in China. This license is expected to receive official support, and if implemented, it will reverse the stagnation in the Chinese market caused by previous export restrictions. Before 2023, China accounted for approximately 25% of Nvidia's data center revenue, which rebounded to 11% to 14% after the introduction of H20.\n\nLimited H20 Spot Supply, Short-term Supply Constraints May Restrict Revenue Recovery Pace\nAlthough Nvidia had previously made a $4.5 billion impairment charge for export restrictions, a significant portion of this included costs for canceled purchase commitments and some scrapped semi-finished products. Additionally, the company has shifted its production focus to the Blackwell architecture, and H20 spot resources are limited, facing certain cycle pressures to restart production.\n\nExport Licensing Process Still Requires Individual Customer Applications, With Delays and Additional Restrictions\nWhile the U.S. government has expressed support for issuing export licenses, the specific process requires each Chinese customer to submit an application to the Department of Commerce. Even under a \"presumed approval\" framework, the relevant process may still be lengthy, and it is not ruled out that individual licenses may come with further performance or usage restrictions.\n\nB30/RTX6000 Positioning Ambiguous, Still Disagreement on Whether It Competes with H20\nThere is still debate on whether RTX6000 (formerly B30) will be affected by demand diversion from the H20 restart. This product does not contain high-bandwidth memory, has simple packaging, low production difficulty, and can be manufactured on the same line as gaming chips, but its market role is not yet clear, and some customers still view it as an intermediate transitional solution.\n\nLong-term Chinese Demand Has Structural Upside, Especially Benefiting AMD and ASIC Vendors\nAlthough the short-term impact is mainly concentrated on Nvidia, AMD may have a higher market share in the Chinese AI market, and if licenses are restored, it will bring greater marginal benefits. In addition, as ByteDance is Broadcom's third-largest ASIC customer, the loosening of licenses will also benefit its deployment in the $600-900 billion TAM by 2027. At the same time, H20 connector suppliers such as ALAB and MRVL are also expected to benefit, although the short-term numerical impact is limited.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003909684872381747, "ret_signed_m1d": -0.003909684872381747, "alpha_spy_m1d": -0.0005776334386263793, "alpha_c_m1d": -0.00916412460569449, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009511589505708296, "ret_signed_p1d": 0.009511589505708296, "alpha_spy_p1d": 0.003391899528190745, "alpha_c_p1d": 0.0011114490994763582, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.0034428594862712103, "ret_signed_p1w": -0.0034428594862712103, "alpha_spy_p1w": -0.019446981286452614, "alpha_c_p1w": 0.005786934868547422, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.0621461963170844, "ret_signed_p1m": 0.0621461963170844, "alpha_spy_p1m": 0.028936714204513825, "alpha_c_p1m": 0.01845147637661637, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.09897081121285733, "ret_signed_p3m": 0.09897081121285733, "alpha_spy_p3m": 0.03383631036483137, "alpha_c_p3m": -0.07549684249508726, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.07988992467564904, "ret_signed_p6m": 0.07988992467564904, "alpha_spy_p6m": -0.031040142771990187, "alpha_c_p6m": -0.23426679716688548, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4078, -0.4345, -0.423, -0.4007, -0.379, -0.3523, -0.3656, -0.3639, -0.3645, -0.3488, -0.3319, -0.3359, -0.3375, -0.317, -0.3152, -0.3194, -0.2823, -0.2419, -0.2103, -0.2133, -0.21, -0.209, -0.2159, -0.231, -0.2249, -0.2339, -0.2339, -0.2094, -0.2134, -0.1878, -0.2115, -0.1984, -0.176, -0.1719, -0.1832, -0.1731, -0.1678, -0.16, -0.1665, -0.1539, -0.1716, -0.1557, -0.159, -0.1511, -0.1511, -0.1606, -0.1587, -0.137, -0.0996, -0.0954, -0.0795, -0.0781, -0.1054, -0.0824, -0.0702, -0.0702, -0.0766, -0.0663, -0.0495, -0.0424, -0.0376, -0.0426, -0.0039, 0.0, 0.0095, 0.0061, 0.0001, -0.0253, -0.0034, 0.0138, 0.0124, 0.0314, 0.0242, 0.0461, 0.0379, 0.0137, 0.0504, 0.0402, 0.047, 0.0549, 0.0661, 0.0624, 0.0688, 0.0596, 0.0621, 0.053, 0.0621, 0.0249, 0.0235, 0.0211, 0.0386, 0.0493, 0.0607, 0.0597, 0.0514, 0.0164, 0.0164, -0.0034, -0.0044, 0.0017, -0.0254, -0.0179, -0.0036, 0.0348, 0.0339, 0.0377, 0.0373, 0.0205, -0.0062, 0.0285, 0.031, 0.0715, 0.0413, 0.0327, 0.0369, 0.0399, 0.0612, 0.0888, 0.0927, 0.1023, 0.0949, 0.0827, 0.0798, 0.1036, 0.1238, 0.0689, 0.099, 0.0506, 0.0494, 0.061, 0.0692, 0.0658, 0.0572, 0.0521, 0.063, 0.0869, 0.1175, 0.1731, 0.2082, 0.184, 0.1817, 0.2073, 0.1595, 0.1392, 0.0976, 0.098, 0.1616, 0.1272, 0.131, 0.0905, 0.1098, 0.0889, 0.0584, 0.0885, 0.0542, 0.0439, 0.0653, 0.0377, 0.0519, 0.0519, 0.0329, 0.05, 0.0589, 0.048, 0.0702, 0.0645, 0.0829, 0.0795, 0.0725, 0.0559, 0.0214, 0.0288, 0.0372, -0.0024, 0.0163, 0.0563, 0.072, 0.1042, 0.1007, 0.1007, 0.1119, 0.0984, 0.0945, 0.0884, 0.0884, 0.1021, 0.0979, 0.0927, 0.1036, 0.0799]}
{"unit_id": "orig:1945588486666588437", "ticker_idx": 1, "ts": "2025-07-16T20:56:57+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD may have a higher market share in the Chinese AI market, and if licenses are restored, it will bring greater marginal benefits", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley (adopted): H20/RTX PRO China license resumption is structurally positive for NVDA; AMD, AVGO, ALAB, MRVL also benefit from licensing loosening.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Expected to Bring Structural Revenue Upside, but Short-term Supply and Licensing Processes Remain Uncertain\nMorgan Stanley (J. Moore, 25/07/16)\n\nMorgan Stanley noted that the U.S. government is expected to issue licenses to Nvidia to resume exports of H20 and RTX PRO GPU products to the Chinese market. Although this change still faces short-term supply and process obstacles, it is generally a significant positive for Nvidia's stock, especially considering that China AI revenue was set at a zero baseline in previous financial reports.\n\nU.S. Government Plans to Issue H20 Export Licenses, Potentially Restarting China Market Revenue Contribution\nNvidia announced it is awaiting U.S. government approval to resume sales of H20 and RTX Pro products in China. This license is expected to receive official support, and if implemented, it will reverse the stagnation in the Chinese market caused by previous export restrictions. Before 2023, China accounted for approximately 25% of Nvidia's data center revenue, which rebounded to 11% to 14% after the introduction of H20.\n\nLimited H20 Spot Supply, Short-term Supply Constraints May Restrict Revenue Recovery Pace\nAlthough Nvidia had previously made a $4.5 billion impairment charge for export restrictions, a significant portion of this included costs for canceled purchase commitments and some scrapped semi-finished products. Additionally, the company has shifted its production focus to the Blackwell architecture, and H20 spot resources are limited, facing certain cycle pressures to restart production.\n\nExport Licensing Process Still Requires Individual Customer Applications, With Delays and Additional Restrictions\nWhile the U.S. government has expressed support for issuing export licenses, the specific process requires each Chinese customer to submit an application to the Department of Commerce. Even under a \"presumed approval\" framework, the relevant process may still be lengthy, and it is not ruled out that individual licenses may come with further performance or usage restrictions.\n\nB30/RTX6000 Positioning Ambiguous, Still Disagreement on Whether It Competes with H20\nThere is still debate on whether RTX6000 (formerly B30) will be affected by demand diversion from the H20 restart. This product does not contain high-bandwidth memory, has simple packaging, low production difficulty, and can be manufactured on the same line as gaming chips, but its market role is not yet clear, and some customers still view it as an intermediate transitional solution.\n\nLong-term Chinese Demand Has Structural Upside, Especially Benefiting AMD and ASIC Vendors\nAlthough the short-term impact is mainly concentrated on Nvidia, AMD may have a higher market share in the Chinese AI market, and if licenses are restored, it will bring greater marginal benefits. In addition, as ByteDance is Broadcom's third-largest ASIC customer, the loosening of licenses will also benefit its deployment in the $600-900 billion TAM by 2027. At the same time, H20 connector suppliers such as ALAB and MRVL are also expected to benefit, although the short-term numerical impact is limited.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02792354553706633, "ret_signed_m1d": -0.02792354553706633, "alpha_spy_m1d": -0.024591494103310962, "alpha_c_m1d": -0.03317798527037907, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0020614806801599883, "ret_signed_p1d": 0.0020614806801599883, "alpha_spy_p1d": -0.004058209297357562, "alpha_c_p1d": -0.006338659726071949, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": -0.008933082947359727, "ret_signed_p1w": -0.008933082947359727, "alpha_spy_p1w": -0.02493720474754113, "alpha_c_p1w": 0.000296711407458905, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.13037228184950478, "ret_signed_p1m": 0.13037228184950478, "alpha_spy_p1m": 0.0971627997369342, "alpha_c_p1m": 0.08667756190903675, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.35194899858478745, "ret_signed_p3m": 0.35194899858478745, "alpha_spy_p3m": 0.2868144977367615, "alpha_c_p3m": 0.17748134487684286, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.27861063427395827, "ret_signed_p6m": 0.27861063427395827, "alpha_spy_p6m": 0.16768056682631904, "alpha_c_p6m": -0.035546087568576246, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.4534, -0.4655, -0.4611, -0.4353, -0.4099, -0.3962, -0.3979, -0.3999, -0.3919, -0.3962, -0.3828, -0.3716, -0.3839, -0.3731, -0.3647, -0.3576, -0.3246, -0.2975, -0.2646, -0.2817, -0.2681, -0.2832, -0.2909, -0.3, -0.3084, -0.3109, -0.3109, -0.2844, -0.295, -0.2939, -0.3083, -0.2839, -0.2672, -0.2592, -0.2773, -0.2742, -0.2396, -0.2301, -0.2433, -0.2597, -0.2744, -0.2105, -0.206, -0.208, -0.208, -0.1989, -0.1905, -0.1352, -0.1042, -0.1024, -0.1016, -0.1136, -0.1497, -0.1347, -0.1385, -0.1385, -0.1579, -0.1391, -0.1354, -0.0995, -0.0853, -0.0865, -0.0279, 0.0, 0.0021, -0.0193, -0.0192, -0.0335, -0.0089, 0.0127, 0.0399, 0.0848, 0.1084, 0.1214, 0.1014, 0.0726, 0.1043, 0.0889, 0.019, 0.077, 0.0792, 0.0762, 0.0929, 0.152, 0.1304, 0.1089, 0.1003, 0.0404, 0.032, 0.0227, 0.048, 0.0205, 0.0409, 0.044, 0.0531, 0.0159, 0.0159, 0.014, 0.0128, 0.0107, -0.0558, -0.0542, -0.0266, -0.0034, -0.0275, -0.0094, 0.0067, 0.0024, -0.0057, -0.0135, -0.0168, -0.0018, 0.0051, 0.005, 0.0074, -0.0039, 0.008, 0.0107, 0.0246, 0.0603, 0.0287, 0.2726, 0.3213, 0.4715, 0.4548, 0.3425, 0.3519, 0.3624, 0.4905, 0.4653, 0.456, 0.5027, 0.4869, 0.4382, 0.468, 0.58, 0.6221, 0.6118, 0.6512, 0.592, 0.5999, 0.622, 0.562, 0.6013, 0.4849, 0.4589, 0.5241, 0.4838, 0.6173, 0.549, 0.5418, 0.5025, 0.4386, 0.3965, 0.287, 0.273, 0.3434, 0.2877, 0.3383, 0.3383, 0.3589, 0.3728, 0.3446, 0.3593, 0.3492, 0.3616, 0.3812, 0.3844, 0.3832, 0.3832, 0.3167, 0.2967, 0.3067, 0.2376, 0.256, 0.3333, 0.3428, 0.3425, 0.3433, 0.3433, 0.343, 0.3469, 0.3452, 0.3378, 0.3378, 0.396, 0.3811, 0.339, 0.312, 0.2786]}
{"unit_id": "orig:1945588486666588437", "ticker_idx": 2, "ts": "2025-07-16T20:56:57+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "as ByteDance is Broadcom's third-largest ASIC customer, the loosening of licenses will also benefit its deployment in the $600-900 billion TAM by 2027", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley (adopted): H20/RTX PRO China license resumption is structurally positive for NVDA; AMD, AVGO, ALAB, MRVL also benefit from licensing loosening.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Expected to Bring Structural Revenue Upside, but Short-term Supply and Licensing Processes Remain Uncertain\nMorgan Stanley (J. Moore, 25/07/16)\n\nMorgan Stanley noted that the U.S. government is expected to issue licenses to Nvidia to resume exports of H20 and RTX PRO GPU products to the Chinese market. Although this change still faces short-term supply and process obstacles, it is generally a significant positive for Nvidia's stock, especially considering that China AI revenue was set at a zero baseline in previous financial reports.\n\nU.S. Government Plans to Issue H20 Export Licenses, Potentially Restarting China Market Revenue Contribution\nNvidia announced it is awaiting U.S. government approval to resume sales of H20 and RTX Pro products in China. This license is expected to receive official support, and if implemented, it will reverse the stagnation in the Chinese market caused by previous export restrictions. Before 2023, China accounted for approximately 25% of Nvidia's data center revenue, which rebounded to 11% to 14% after the introduction of H20.\n\nLimited H20 Spot Supply, Short-term Supply Constraints May Restrict Revenue Recovery Pace\nAlthough Nvidia had previously made a $4.5 billion impairment charge for export restrictions, a significant portion of this included costs for canceled purchase commitments and some scrapped semi-finished products. Additionally, the company has shifted its production focus to the Blackwell architecture, and H20 spot resources are limited, facing certain cycle pressures to restart production.\n\nExport Licensing Process Still Requires Individual Customer Applications, With Delays and Additional Restrictions\nWhile the U.S. government has expressed support for issuing export licenses, the specific process requires each Chinese customer to submit an application to the Department of Commerce. Even under a \"presumed approval\" framework, the relevant process may still be lengthy, and it is not ruled out that individual licenses may come with further performance or usage restrictions.\n\nB30/RTX6000 Positioning Ambiguous, Still Disagreement on Whether It Competes with H20\nThere is still debate on whether RTX6000 (formerly B30) will be affected by demand diversion from the H20 restart. This product does not contain high-bandwidth memory, has simple packaging, low production difficulty, and can be manufactured on the same line as gaming chips, but its market role is not yet clear, and some customers still view it as an intermediate transitional solution.\n\nLong-term Chinese Demand Has Structural Upside, Especially Benefiting AMD and ASIC Vendors\nAlthough the short-term impact is mainly concentrated on Nvidia, AMD may have a higher market share in the Chinese AI market, and if licenses are restored, it will bring greater marginal benefits. In addition, as ByteDance is Broadcom's third-largest ASIC customer, the loosening of licenses will also benefit its deployment in the $600-900 billion TAM by 2027. At the same time, H20 connector suppliers such as ALAB and MRVL are also expected to benefit, although the short-term numerical impact is limited.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00046299595545273853, "ret_signed_m1d": 0.00046299595545273853, "alpha_spy_m1d": 0.003795047389208106, "alpha_c_m1d": -0.004791443777860005, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02008490882389835, "ret_signed_p1d": 0.02008490882389835, "alpha_spy_p1d": 0.0139652188463808, "alpha_c_p1d": 0.011684768417666413, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": 0.010256191095350475, "ret_signed_p1w": 0.010256191095350475, "alpha_spy_p1w": -0.005747930704830928, "alpha_c_p1w": 0.019485985450169108, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.10832957704729251, "ret_signed_p1m": 0.10832957704729251, "alpha_spy_p1m": 0.07512009493472194, "alpha_c_p1m": 0.06463485710682448, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.2724303833772648, "ret_signed_p3m": 0.2724303833772648, "alpha_spy_p3m": 0.2072958825292388, "alpha_c_p3m": 0.09796272966932018, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.18830152622528895, "ret_signed_p6m": 0.18830152622528895, "alpha_spy_p6m": 0.07737145877764973, "alpha_c_p6m": -0.12585519561724556, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.3925, -0.4095, -0.3975, -0.3715, -0.3315, -0.3168, -0.3162, -0.3208, -0.3162, -0.2989, -0.2765, -0.2869, -0.2891, -0.2724, -0.2618, -0.2603, -0.2128, -0.1743, -0.1753, -0.1735, -0.1878, -0.1806, -0.1769, -0.1838, -0.181, -0.1874, -0.1874, -0.1628, -0.1494, -0.1403, -0.14, -0.1164, -0.0875, -0.0724, -0.0765, -0.1227, -0.1321, -0.1309, -0.1015, -0.0902, -0.1164, -0.1043, -0.114, -0.1073, -0.1073, -0.1098, -0.0963, -0.0607, -0.0575, -0.0379, -0.0408, -0.0184, -0.0572, -0.0389, -0.02, -0.02, -0.0236, -0.0321, -0.0104, -0.0193, -0.0229, -0.0186, 0.0005, 0.0, 0.0201, 0.009, 0.0264, -0.0079, 0.0103, 0.0281, 0.0334, 0.048, 0.0592, 0.0777, 0.0459, 0.0279, 0.0602, 0.0432, 0.0743, 0.0817, 0.086, 0.0822, 0.114, 0.1007, 0.1083, 0.0909, 0.0889, 0.0502, 0.0369, 0.0313, 0.047, 0.0478, 0.0613, 0.0692, 0.0991, 0.059, 0.059, 0.0621, 0.0768, 0.0901, 0.1926, 0.2309, 0.1989, 0.3161, 0.2807, 0.2815, 0.2966, 0.282, 0.2328, 0.2298, 0.2284, 0.2085, 0.2091, 0.2104, 0.1989, 0.1933, 0.1697, 0.1769, 0.1893, 0.2064, 0.207, 0.1968, 0.2001, 0.2325, 0.2308, 0.158, 0.2724, 0.2276, 0.2533, 0.2633, 0.2461, 0.2458, 0.2223, 0.2139, 0.2282, 0.2633, 0.2915, 0.3305, 0.3769, 0.343, 0.3186, 0.2933, 0.2555, 0.2806, 0.2685, 0.2465, 0.2785, 0.2555, 0.2672, 0.2128, 0.2216, 0.2223, 0.2146, 0.2643, 0.2372, 0.2136, 0.3483, 0.3735, 0.4182, 0.4182, 0.4375, 0.3772, 0.3611, 0.3577, 0.3592, 0.3921, 0.4308, 0.4493, 0.4732, 0.4496, 0.284, 0.2122, 0.2175, 0.163, 0.1768, 0.2141, 0.2204, 0.2485, 0.2517, 0.2517, 0.2585, 0.2487, 0.2504, 0.237, 0.237, 0.2424, 0.2274, 0.2287, 0.2277, 0.1883]}
{"unit_id": "orig:1945588486666588437", "ticker_idx": 3, "ts": "2025-07-16T20:56:57+00:00", "symbol": "ALAB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "H20 connector suppliers such as ALAB and MRVL are also expected to benefit", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley (adopted): H20/RTX PRO China license resumption is structurally positive for NVDA; AMD, AVGO, ALAB, MRVL also benefit from licensing loosening.", "resolved_tickers": ["ALAB"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Expected to Bring Structural Revenue Upside, but Short-term Supply and Licensing Processes Remain Uncertain\nMorgan Stanley (J. Moore, 25/07/16)\n\nMorgan Stanley noted that the U.S. government is expected to issue licenses to Nvidia to resume exports of H20 and RTX PRO GPU products to the Chinese market. Although this change still faces short-term supply and process obstacles, it is generally a significant positive for Nvidia's stock, especially considering that China AI revenue was set at a zero baseline in previous financial reports.\n\nU.S. Government Plans to Issue H20 Export Licenses, Potentially Restarting China Market Revenue Contribution\nNvidia announced it is awaiting U.S. government approval to resume sales of H20 and RTX Pro products in China. This license is expected to receive official support, and if implemented, it will reverse the stagnation in the Chinese market caused by previous export restrictions. Before 2023, China accounted for approximately 25% of Nvidia's data center revenue, which rebounded to 11% to 14% after the introduction of H20.\n\nLimited H20 Spot Supply, Short-term Supply Constraints May Restrict Revenue Recovery Pace\nAlthough Nvidia had previously made a $4.5 billion impairment charge for export restrictions, a significant portion of this included costs for canceled purchase commitments and some scrapped semi-finished products. Additionally, the company has shifted its production focus to the Blackwell architecture, and H20 spot resources are limited, facing certain cycle pressures to restart production.\n\nExport Licensing Process Still Requires Individual Customer Applications, With Delays and Additional Restrictions\nWhile the U.S. government has expressed support for issuing export licenses, the specific process requires each Chinese customer to submit an application to the Department of Commerce. Even under a \"presumed approval\" framework, the relevant process may still be lengthy, and it is not ruled out that individual licenses may come with further performance or usage restrictions.\n\nB30/RTX6000 Positioning Ambiguous, Still Disagreement on Whether It Competes with H20\nThere is still debate on whether RTX6000 (formerly B30) will be affected by demand diversion from the H20 restart. This product does not contain high-bandwidth memory, has simple packaging, low production difficulty, and can be manufactured on the same line as gaming chips, but its market role is not yet clear, and some customers still view it as an intermediate transitional solution.\n\nLong-term Chinese Demand Has Structural Upside, Especially Benefiting AMD and ASIC Vendors\nAlthough the short-term impact is mainly concentrated on Nvidia, AMD may have a higher market share in the Chinese AI market, and if licenses are restored, it will bring greater marginal benefits. In addition, as ByteDance is Broadcom's third-largest ASIC customer, the loosening of licenses will also benefit its deployment in the $600-900 billion TAM by 2027. At the same time, H20 connector suppliers such as ALAB and MRVL are also expected to benefit, although the short-term numerical impact is limited.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004568176612927299, "ret_signed_m1d": 0.004568176612927299, "alpha_spy_m1d": 0.007900228046682667, "alpha_c_m1d": -0.0006862631203854441, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06536865724651397, "ret_signed_p1d": 0.06536865724651397, "alpha_spy_p1d": 0.05924896726899642, "alpha_c_p1d": 0.056968516840282035, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": 0.29954318233870736, "ret_signed_p1w": 0.29954318233870736, "alpha_spy_p1w": 0.28353906053852596, "alpha_c_p1w": 0.308772976693526, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 1.0740700171607434, "ret_signed_p1m": 1.0740700171607434, "alpha_spy_p1m": 1.0408605350481728, "alpha_c_p1m": 1.0303752972202753, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 1.17021963759962, "ret_signed_p3m": 1.17021963759962, "alpha_spy_p3m": 1.1050851367515941, "alpha_c_p3m": 0.9957519838916755, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.7046986248170239, "ret_signed_p6m": 0.7046986248170239, "alpha_spy_p6m": 0.5937685573693847, "alpha_c_p6m": 0.39054190297448943, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.3587, -0.4081, -0.3984, -0.3458, -0.2979, -0.2809, -0.2813, -0.2859, -0.2896, -0.2471, -0.2248, -0.2152, -0.2238, -0.2268, -0.2208, -0.2256, -0.1268, -0.0455, -0.0139, 0.0027, -0.015, -0.018, 0.0274, 0.0025, 0.0257, 0.0291, 0.0291, 0.0609, 0.0492, 0.0508, -0.0133, 0.0107, 0.0343, 0.0357, -0.0134, -0.0129, 0.0045, -0.0052, 0.0283, 0.0261, -0.024, 0.0365, 0.0106, 0.0826, 0.0826, 0.0123, -0.0652, -0.0509, -0.0251, 0.0655, -0.01, -0.0165, -0.0357, -0.0367, -0.0124, -0.0124, -0.0252, 0.0039, 0.0861, 0.0553, 0.0431, -0.0176, 0.0046, 0.0, 0.0654, 0.1108, 0.3258, 0.2716, 0.2995, 0.3235, 0.3295, 0.3492, 0.2879, 0.4017, 0.4872, 0.4259, 0.5002, 0.4742, 0.8968, 0.8587, 0.95, 0.9516, 1.0883, 1.1062, 1.0741, 1.0214, 1.0277, 0.8606, 0.8757, 0.9309, 0.9479, 0.8942, 0.9421, 0.9491, 1.0573, 0.9817, 0.9817, 0.8951, 0.9209, 1.0443, 1.0796, 1.3504, 1.3329, 1.5057, 1.5332, 1.4962, 1.5157, 1.5972, 1.7124, 1.7396, 1.667, 1.581, 1.5028, 1.2327, 1.1823, 1.1512, 1.1623, 1.1296, 1.1155, 1.2797, 1.1834, 1.4017, 1.3069, 1.3859, 1.4519, 1.2429, 1.1702, 0.7571, 0.7562, 0.7789, 0.7381, 0.7001, 0.7132, 0.6843, 0.7799, 0.7943, 0.8521, 0.83, 0.8884, 0.8441, 1.0305, 1.0835, 0.9503, 0.9789, 0.771, 0.8, 0.8897, 0.8224, 0.7162, 0.5714, 0.5699, 0.5379, 0.5175, 0.5446, 0.515, 0.5423, 0.607, 0.5747, 0.6774, 0.6774, 0.7138, 0.7967, 0.5547, 0.6587, 0.6588, 0.7536, 0.9115, 0.8173, 0.7872, 0.8893, 0.619, 0.5625, 0.5765, 0.5253, 0.5867, 0.7881, 0.8775, 0.8363, 0.8487, 0.8487, 0.8192, 0.8502, 0.8582, 0.8094, 0.8094, 0.953, 0.8176, 0.7513, 0.803, 0.7047]}
{"unit_id": "orig:1945588486666588437", "ticker_idx": 4, "ts": "2025-07-16T20:56:57+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "H20 connector suppliers such as ALAB and MRVL are also expected to benefit", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley (adopted): H20/RTX PRO China license resumption is structurally positive for NVDA; AMD, AVGO, ALAB, MRVL also benefit from licensing loosening.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nvidia: Resumption of Shipments to China Expected to Bring Structural Revenue Upside, but Short-term Supply and Licensing Processes Remain Uncertain\nMorgan Stanley (J. Moore, 25/07/16)\n\nMorgan Stanley noted that the U.S. government is expected to issue licenses to Nvidia to resume exports of H20 and RTX PRO GPU products to the Chinese market. Although this change still faces short-term supply and process obstacles, it is generally a significant positive for Nvidia's stock, especially considering that China AI revenue was set at a zero baseline in previous financial reports.\n\nU.S. Government Plans to Issue H20 Export Licenses, Potentially Restarting China Market Revenue Contribution\nNvidia announced it is awaiting U.S. government approval to resume sales of H20 and RTX Pro products in China. This license is expected to receive official support, and if implemented, it will reverse the stagnation in the Chinese market caused by previous export restrictions. Before 2023, China accounted for approximately 25% of Nvidia's data center revenue, which rebounded to 11% to 14% after the introduction of H20.\n\nLimited H20 Spot Supply, Short-term Supply Constraints May Restrict Revenue Recovery Pace\nAlthough Nvidia had previously made a $4.5 billion impairment charge for export restrictions, a significant portion of this included costs for canceled purchase commitments and some scrapped semi-finished products. Additionally, the company has shifted its production focus to the Blackwell architecture, and H20 spot resources are limited, facing certain cycle pressures to restart production.\n\nExport Licensing Process Still Requires Individual Customer Applications, With Delays and Additional Restrictions\nWhile the U.S. government has expressed support for issuing export licenses, the specific process requires each Chinese customer to submit an application to the Department of Commerce. Even under a \"presumed approval\" framework, the relevant process may still be lengthy, and it is not ruled out that individual licenses may come with further performance or usage restrictions.\n\nB30/RTX6000 Positioning Ambiguous, Still Disagreement on Whether It Competes with H20\nThere is still debate on whether RTX6000 (formerly B30) will be affected by demand diversion from the H20 restart. This product does not contain high-bandwidth memory, has simple packaging, low production difficulty, and can be manufactured on the same line as gaming chips, but its market role is not yet clear, and some customers still view it as an intermediate transitional solution.\n\nLong-term Chinese Demand Has Structural Upside, Especially Benefiting AMD and ASIC Vendors\nAlthough the short-term impact is mainly concentrated on Nvidia, AMD may have a higher market share in the Chinese AI market, and if licenses are restored, it will bring greater marginal benefits. In addition, as ByteDance is Broadcom's third-largest ASIC customer, the loosening of licenses will also benefit its deployment in the $600-900 billion TAM by 2027. At the same time, H20 connector suppliers such as ALAB and MRVL are also expected to benefit, although the short-term numerical impact is limited.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.022018465220547734, "ret_signed_m1d": 0.022018465220547734, "alpha_spy_m1d": 0.0253505166543031, "alpha_c_m1d": 0.01676402548723499, "bench_spy_m1d": -0.0033320514337553675, "bench_c_m1d": 0.0052544397333127435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01637276023404044, "ret_signed_p1d": 0.01637276023404044, "alpha_spy_p1d": 0.01025307025652289, "alpha_c_p1d": 0.007972619827808503, "bench_spy_p1d": 0.006119689977517551, "bench_c_p1d": 0.008400140406231937, "ret_p1w": 0.03415678488483054, "ret_signed_p1w": 0.03415678488483054, "alpha_spy_p1w": 0.018152663084649134, "alpha_c_p1w": 0.04338657923964917, "bench_spy_p1w": 0.016004121800181403, "bench_c_p1w": -0.009229794354818632, "ret_p1m": 0.11559648388988975, "ret_signed_p1m": 0.11559648388988975, "alpha_spy_p1m": 0.08238700177731917, "alpha_c_p1m": 0.07190176394942172, "bench_spy_p1m": 0.033209482112570576, "bench_c_p1m": 0.04369471994046803, "ret_p3m": 0.2625151405336048, "ret_signed_p3m": 0.2625151405336048, "alpha_spy_p3m": 0.19738063968557884, "alpha_c_p3m": 0.08804748682566022, "bench_spy_p3m": 0.06513450084802597, "bench_c_p3m": 0.1744676537079446, "ret_p6m": 0.17862037440313094, "ret_signed_p6m": 0.17862037440313094, "alpha_spy_p6m": 0.06769030695549172, "alpha_c_p6m": -0.13553634743940357, "bench_spy_p6m": 0.11093006744763922, "bench_c_p6m": 0.3141567218425345, "price_path": [-0.2709, -0.3036, -0.2861, -0.2417, -0.1915, -0.1691, -0.1722, -0.1724, -0.1768, -0.1404, -0.121, -0.1259, -0.1366, -0.2059, -0.1873, -0.1588, -0.0904, -0.0765, -0.0706, -0.0805, -0.1008, -0.1177, -0.1338, -0.1523, -0.1277, -0.1441, -0.1441, -0.1, -0.0891, -0.1012, -0.1512, -0.1331, -0.1206, -0.065, -0.0811, -0.0361, -0.0249, -0.0292, -0.0376, -0.0179, -0.0524, -0.0069, -0.0129, 0.057, 0.057, 0.0367, -0.0018, 0.0607, 0.0708, 0.1278, 0.0882, 0.0916, 0.0752, 0.0471, 0.0602, 0.0602, 0.0091, 0.0147, 0.0191, 0.0346, 0.0263, 0.0234, 0.022, 0.0, 0.0164, 0.0536, 0.0312, 0.0161, 0.0342, 0.045, 0.0474, 0.0714, 0.0775, 0.1537, 0.1344, 0.0508, 0.0802, 0.0816, 0.0631, 0.0706, 0.0916, 0.0908, 0.0982, 0.1195, 0.1156, 0.0754, 0.0831, 0.0172, 0.0052, 0.0051, 0.0303, 0.0296, 0.0481, 0.0556, 0.09, -0.1126, -0.1126, -0.0882, -0.1205, -0.0953, -0.1061, -0.0685, -0.0566, -0.0529, -0.0601, -0.0494, -0.0483, -0.0281, 0.0018, 0.0477, 0.0481, 0.0661, 0.0532, 0.1304, 0.1829, 0.1739, 0.1629, 0.1866, 0.1841, 0.2167, 0.2169, 0.255, 0.2275, 0.3056, 0.2799, 0.2091, 0.2625, 0.2177, 0.2555, 0.2461, 0.2422, 0.2124, 0.1901, 0.1447, 0.169, 0.1882, 0.2529, 0.2494, 0.2732, 0.2509, 0.324, 0.2764, 0.2371, 0.3121, 0.3182, 0.2841, 0.3168, 0.2617, 0.2617, 0.2361, 0.221, 0.1786, 0.1113, 0.1485, 0.083, 0.0939, 0.1834, 0.1783, 0.2389, 0.2389, 0.2627, 0.2867, 0.3119, 0.4152, 0.3868, 0.397, 0.2994, 0.2556, 0.306, 0.2631, 0.1925, 0.1901, 0.1874, 0.1539, 0.193, 0.1877, 0.1977, 0.2384, 0.2216, 0.2216, 0.2194, 0.2112, 0.2254, 0.2002, 0.2002, 0.2625, 0.2744, 0.2461, 0.1954, 0.1786]}
{"unit_id": "thread:1967749068467454314", "ticker_idx": 0, "ts": "2025-09-16T00:35:11+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": "AI infrastructure/GPU", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "An AI infrastructure investment of $3 million in the GB200 NVL72 can generate $30 million in revenue over three years.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish NVDA: sharing Dion Harris's 10x ROI case for GB200 NVL72, implying strong durable demand for Blackwell.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA Product Leader Dion Harris:\n\nAn AI infrastructure investment of $3 million in the GB200 NVL72 can generate $30 million in revenue over three years.\n\nEven if you received GPUs with only one-quarter the performance of Blackwell “for free,” you would still incur an opportunity cost (lost revenue) of more than $20 million.\n\n$NVDA\n\n---\n\nhttps://t.co/oCOmLGvXMx", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01641119410161429, "ret_signed_m1d": 0.01641119410161429, "alpha_spy_m1d": 0.01503250314456972, "alpha_c_m1d": 0.016541333006460546, "bench_spy_m1d": 0.0013786909570445705, "bench_c_m1d": -0.00013013890484625534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0262466362492908, "ret_signed_p1d": -0.0262466362492908, "alpha_spy_p1d": -0.025004200956742162, "alpha_c_p1d": -0.019866641209715485, "bench_spy_p1d": -0.0012424352925486382, "bench_c_p1d": -0.006379995039575315, "ret_p1w": 0.020299523250129736, "ret_signed_p1w": 0.020299523250129736, "alpha_spy_p1w": 0.012649900821011029, "alpha_c_p1w": -0.02686696117888987, "bench_spy_p1w": 0.007649622429118708, "bench_c_p1w": 0.047166484429019606, "ret_p1m": 0.028305071049989472, "ret_signed_p1m": 0.028305071049989472, "alpha_spy_p1m": 0.017677595528911905, "alpha_c_p1m": -0.08422382780846505, "bench_spy_p1m": 0.010627475521077567, "bench_c_p1m": 0.11252889885845452, "ret_p3m": 0.0008562195203041245, "ret_signed_p3m": 0.0008562195203041245, "alpha_spy_p3m": -0.03497724826562987, "alpha_c_p3m": -0.15184066016528197, "bench_spy_p3m": 0.03583346778593399, "bench_c_p3m": 0.1526968796855861, "ret_p6m": 0.06387474412457306, "ret_signed_p6m": 0.06387474412457306, "alpha_spy_p6m": 0.033254897428516506, "alpha_c_p6m": -0.24558082385252722, "bench_spy_p6m": 0.030619846696056552, "bench_c_p6m": 0.3094555679771003, "price_path": [-0.1682, -0.1775, -0.1757, -0.1543, -0.1177, -0.1136, -0.098, -0.0966, -0.1234, -0.1009, -0.0889, -0.0889, -0.0952, -0.0851, -0.0687, -0.0617, -0.057, -0.0619, -0.024, -0.0201, -0.0108, -0.0142, -0.0201, -0.0449, -0.0235, -0.0066, -0.0079, 0.0106, 0.0035, 0.025, 0.017, -0.0067, 0.0292, 0.0193, 0.0259, 0.0336, 0.0447, 0.041, 0.0473, 0.0383, 0.0408, 0.0318, 0.0407, 0.0043, 0.0029, 0.0005, 0.0177, 0.0281, 0.0393, 0.0384, 0.0302, -0.0041, -0.0041, -0.0235, -0.0244, -0.0185, -0.045, -0.0376, -0.0236, 0.014, 0.0131, 0.0168, 0.0164, 0.0, -0.0262, 0.0078, 0.0102, 0.0499, 0.0203, 0.012, 0.0161, 0.0189, 0.0399, 0.0669, 0.0707, 0.0801, 0.0728, 0.061, 0.0581, 0.0814, 0.1012, 0.0473, 0.0769, 0.0294, 0.0283, 0.0396, 0.0477, 0.0444, 0.0359, 0.0309, 0.0416, 0.0651, 0.095, 0.1495, 0.1839, 0.1602, 0.1579, 0.183, 0.1362, 0.1163, 0.0755, 0.0759, 0.1382, 0.1045, 0.1082, 0.0685, 0.0874, 0.067, 0.0371, 0.0666, 0.0329, 0.0229, 0.0439, 0.0168, 0.0308, 0.0308, 0.0121, 0.0288, 0.0376, 0.0269, 0.0487, 0.0431, 0.0611, 0.0578, 0.051, 0.0347, 0.0009, 0.0081, 0.0163, -0.0225, -0.0042, 0.035, 0.0504, 0.082, 0.0786, 0.0786, 0.0896, 0.0763, 0.0725, 0.0665, 0.0665, 0.0799, 0.0758, 0.0707, 0.0814, 0.0582, 0.0571, 0.0576, 0.0626, 0.0473, 0.0697, 0.065, 0.065, 0.0183, 0.0483, 0.057, 0.0732, 0.0663, 0.0781, 0.0952, 0.1009, 0.093, 0.0614, 0.0313, -0.0039, -0.0171, 0.0603, 0.0867, 0.0782, 0.0868, 0.069, 0.0454, 0.0454, 0.0578, 0.075, 0.0745, 0.0855, 0.0954, 0.1028, 0.1183, 0.0573, 0.0133, 0.0435, 0.0296, 0.0467, 0.0484, 0.0169, 0.0445, 0.0566, 0.0639]}
{"unit_id": "orig:1948532372523323900", "ticker_idx": 0, "ts": "2025-07-24T23:54:54+00:00", "symbol": "ASMPT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rating and Target Price Upgraded... more upside risk than downside risk", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley upgrades ASMPT rating and PT: AI data center demand driving mainstream equipment recovery, TCB growth, auto/industrial at trough with limited downside.", "resolved_tickers": ["0522.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "ASMPT listed on Hong Kong Exchange 0522.HK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "ASMPT Quarterly Report: AI Data Centers Drive Mainstream Equipment Recovery, Optimistic on TCB and China's Advanced Packaging Growth Potential\nMorgan Stanley (D. Dai, 25/07/24)\n\nMorgan Stanley states that as AI data center construction begins to drive demand for mainstream semiconductor equipment, ASMPT is poised for an inflection point in its performance. Although the automotive and industrial sectors remain at a low point, overall risks are converging, and the company is performing positively in TCB and the Chinese market.\n\nAI Data Centers Drive Mainstream Equipment Demand, Reshaping Growth Expectations\n\nPrevious conservative ratings for ASMPT stemmed from the slow recovery of the mainstream semiconductor market, while the TCB business accounted for only about 10% in 2024, making it difficult to fully offset the impact. Now, AI data centers are beginning to drive demand for mainstream equipment, including wire bonders and molding equipment. The company is expected to benefit from NVIDIA's dominant 800V HVDC data center power architecture upgrade, and supply chain customers are willing to invest in new technologies to meet its demands.\n\nTCB Order Performance Steady, China Market Provides Future Growth Space\n\nASMPT continues its positive trend in TCB. C2S solution orders are growing among packaging partners of leading foundries, and AOR technology shipments for C2W are expected to increase in 2025. Although NVIDIA's Rubin platform still uses traditional processes, some HBM customers are already using ASMPT's TCB tools for small-scale production in HBM4 12H. The company also benefits from advanced packaging and HBM investments in China, giving it a competitive advantage.\n\nAutomotive and Industrial Sectors Hit Historical Lows, Limited Future Downside\n\nAffected by weak demand, the automotive and industrial businesses' share of revenue has fallen from 41% in 1H22 to the current 23%, continuously dragging down the company's profit performance. Management expects short-term demand to remain weak, but as Tier 1 inventory gradually depletes and the Chinese automotive market performs steadily, Morgan Stanley believes this business segment is likely to rebound between 2026 and 2027.\n\nAttractive Valuation, Rating and Target Price Upgraded\n\nASMPT, as a relatively lagging stock in the advanced packaging sector, has been constrained by weakness in the mainstream market and automotive/industrial sectors. Currently, AI servers are driving demand for mainstream tools, and the automotive/industrial segment is at a trough, indicating more upside risk than downside risk.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09971502760505369, "ret_signed_m1d": -0.09971502760505369, "alpha_spy_m1d": -0.09938408464172965, "alpha_c_m1d": -0.09558020778025567, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005698130209061292, "ret_signed_p1d": 0.005698130209061292, "alpha_spy_p1d": 0.0014737805687421446, "alpha_c_p1d": 0.006775312838797287, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": -0.04700847614986836, "ret_signed_p1w": -0.04700847614986836, "alpha_spy_p1w": -0.04332017455187276, "alpha_c_p1w": -0.05041361563103153, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.0033671368367119703, "ret_signed_p1m": -0.0033671368367119703, "alpha_spy_p1m": -0.02053251090801489, "alpha_c_p1m": -0.023659057903168468, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.22756009846217906, "ret_signed_p3m": 0.22756009846217906, "alpha_spy_p3m": 0.16651058140023078, "alpha_c_p3m": 0.027768621107319147, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 0.39914681777348404, "ret_signed_p6m": 0.39914681777348404, "alpha_spy_p6m": 0.3026697158054068, "alpha_c_p6m": 0.0036086960706738225, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.2756, -0.2656, -0.2607, -0.2607, -0.2437, -0.2437, -0.2458, -0.2536, -0.2451, -0.2409, -0.2111, -0.2066, -0.1944, -0.1944, -0.2044, -0.2393, -0.2336, -0.2308, -0.2407, -0.2536, -0.2443, -0.2543, -0.2457, -0.2329, -0.2464, -0.2372, -0.2244, -0.2258, -0.208, -0.213, -0.188, -0.1823, -0.2066, -0.2258, -0.2258, -0.2229, -0.2279, -0.2443, -0.2543, -0.2443, -0.2187, -0.2094, -0.1973, -0.2016, -0.1759, -0.1802, -0.1802, -0.1845, -0.1809, -0.1624, -0.1738, -0.1581, -0.1638, -0.1631, -0.1403, -0.1368, -0.1182, -0.1268, -0.1083, -0.1132, -0.0997, -0.1282, -0.0997, 0.0, 0.0057, -0.0192, -0.0121, -0.0477, -0.047, -0.0712, -0.0377, -0.037, -0.0363, -0.0107, -0.0171, 0.0038, 0.0059, 0.0267, -0.0005, 0.0259, 0.0024, -0.0463, -0.0484, -0.0556, -0.0034, 0.0724, 0.0395, 0.0009, 0.0395, 0.0081, 0.0152, -0.027, -0.0348, -0.0413, -0.0284, -0.0234, -0.0205, -0.0155, 0.0152, 0.0145, 0.0045, 0.0252, 0.0188, 0.0932, 0.1003, 0.1332, 0.1296, 0.1997, 0.1646, 0.1275, 0.1375, 0.1732, 0.1732, 0.2154, 0.1997, 0.2769, 0.2769, 0.2883, 0.3634, 0.3269, 0.2848, 0.2004, 0.2554, 0.2626, 0.204, 0.2447, 0.2276, 0.1947, 0.1904, 0.2869, 0.3405, 0.2476, 0.2476, 0.199, 0.1725, 0.1811, 0.1675, 0.1661, 0.1768, 0.1604, 0.1775, 0.1568, 0.1339, 0.1453, 0.1139, 0.1146, 0.0867, 0.0731, 0.0846, 0.0181, 0.0131, 0.0481, 0.0367, 0.0588, 0.0767, 0.0939, 0.0932, 0.106, 0.116, 0.1239, 0.1132, 0.131, 0.1024, 0.1275, 0.126, 0.0753, 0.0688, 0.086, 0.0789, 0.0631, 0.106, 0.1074, 0.1067, 0.1067, 0.1067, 0.0889, 0.1046, 0.111, 0.111, 0.1611, 0.1754, 0.219, 0.294, 0.2726, 0.2862, 0.2876, 0.3341, 0.3491, 0.3627, 0.3991]}
{"unit_id": "quote:1973640639263941036", "ticker_idx": 0, "ts": "2025-10-02T06:46:11+00:00", "symbol": "3661.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "2026/2027/2028 EPS are projected at $180 / $240 / $300, significantly above market expectations. Current valuation remains inexpensive, and the recent stock price correction offers a very attractive buying opportunity.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Alchip (3661): Trn3 no delay, EPS $180/$240/$300 well above consensus; correction is attractive buying opportunity.", "resolved_tickers": ["3661.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Alchip (3661) Trn3 Has No Delay; 2026/27 Earnings to Beat Market Expectations\n\n    1. 2026 Trn3 Volume Above Expectations\n    • Estimated at about 3–4 million units, especially with demand projections boosted following the release of Sonnet 4.5 for inference applications.\n\n    • Contrary to market rumors, there has been no delay: Trn3-1 taped out in Q1 2025 and will enter mass production in Q1 2026; Trn3-2 taped out in Q3 2025 and will enter mass production in Q2 2026.\n\n    2.    Capacity Planning With Annapurna\n    • Management recently returned from the U.S. after discussions with customers about capacity planning.\n\n    • It is reported that Annapurna has recently requested Alchip to prepare more capacity.\n\n    • All turnkey production for Trn-2 will be handled by Alchip.\n\n    • Market concerns about Annapurna self-fabbing chips have had no significant impact on Alchip’s demand.\n\n    3. Trn4 Timeline\n    • Customer frontend design for Trn4 is expected to be finalized in Q4 2025.\n\n    • Netlist files will then be delivered to Alchip to begin backend design.\n\n    4. Earnings Outlook\n\n    • 2026/2027/2028 EPS are projected at $180 / $240 / $300, significantly above market expectations.\n\n    • Current valuation remains inexpensive, and the recent stock price correction offers a very attractive buying opportunity.\n\n$MRVL", "reply_text": null, "parent_text": null, "quoted_text": "轉 \n世芯3661 Trn3沒有遞延，2026/27年獲利優於市場預期\n1. 2026年Trn3量優於市場預期(約3-4m顆)，尤其在Sonnet 4.5發布後對推理需求量預測提升。沒有如市場傳聞的遞延，其中Trn3-1已於Q1 25 tape out，Q1 26量產。Trn3-2 已於Q3 25 tape out，Q2 26量產。\n2.", "parent_author": null, "ret_m1d": 0.03348554033485551, "ret_signed_m1d": 0.03348554033485551, "alpha_spy_m1d": 0.034635927024197266, "alpha_c_m1d": 0.04693753465691053, "bench_spy_m1d": -0.0011503866893417536, "bench_c_m1d": -0.013451994322055016, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007610350076103556, "ret_signed_p1d": 0.007610350076103556, "alpha_spy_p1d": 0.0076251174383251685, "alpha_c_p1d": 0.012222503788977823, "bench_spy_p1d": -1.4767362221612679e-05, "bench_c_p1d": -0.004612153712874267, "ret_p1w": 0.07001522070015231, "ret_signed_p1w": 0.07001522070015231, "alpha_spy_p1w": 0.06711623156737168, "alpha_c_p1w": 0.04940859577938972, "bench_spy_p1w": 0.0028989891327806383, "bench_c_p1w": 0.020606624920762595, "ret_p1m": 0.06392694063926951, "ret_signed_p1m": 0.06392694063926951, "alpha_spy_p1m": 0.04474028475797498, "alpha_c_p1m": -0.00933462314149458, "bench_spy_p1m": 0.019186655881294534, "bench_c_p1m": 0.0732615637807641, "ret_p3m": 0.060882800608828, "ret_signed_p3m": 0.060882800608828, "alpha_spy_p3m": 0.031265984307200956, "alpha_c_p3m": -0.016578393403191782, "bench_spy_p3m": 0.029616816301627047, "bench_c_p3m": 0.07746119401201979, "ret_p6m": -0.07458143074581436, "ret_signed_p6m": -0.07458143074581436, "alpha_spy_p6m": -0.02748271753340037, "alpha_c_p6m": -0.18448717121188807, "bench_spy_p6m": -0.04709871321241399, "bench_c_p6m": 0.10990574046607371, "price_path": [-0.029, -0.0305, 0.0147, 0.0434, 0.0343, -0.0018, 0.0162, 0.0856, 0.1323, 0.1851, 0.1399, 0.1218, 0.1535, 0.1987, 0.1716, 0.17, 0.1504, 0.1308, 0.1716, 0.1716, 0.1323, 0.1308, 0.1188, 0.1354, 0.158, 0.1791, 0.2288, 0.2424, 0.2816, 0.2982, 0.2756, 0.2168, 0.1037, 0.1414, 0.1127, 0.1746, 0.2002, 0.2304, 0.2062, 0.2228, 0.1806, 0.1705, 0.2116, 0.172, 0.2085, 0.21, 0.1918, 0.1994, 0.1948, 0.1537, 0.1339, 0.1202, 0.1446, 0.1324, 0.1416, 0.1446, 0.1263, 0.0381, 0.0259, -0.003, -0.003, 0.0548, 0.0335, 0.0, 0.0076, 0.0076, 0.0533, 0.0487, 0.07, 0.07, 0.0259, -0.0639, -0.0624, -0.0578, -0.0609, -0.0152, 0.0213, -0.0198, -0.0411, -0.0411, -0.067, -0.067, -0.0107, 0.0213, 0.0639, 0.0502, 0.0487, 0.0167, 0.1172, 0.0852, 0.1309, 0.1065, 0.0715, 0.0731, 0.0685, 0.0989, -0.0107, 0.0107, -0.0639, -0.0898, -0.0959, -0.102, -0.0533, 0.0, 0.0061, 0.0015, 0.0411, -0.0289, -0.035, -0.0396, -0.0502, -0.0076, 0.0259, 0.003, -0.0198, -0.035, -0.0822, -0.0259, -0.035, -0.0213, -0.0183, -0.0198, -0.032, -0.032, -0.0213, 0.032, 0.0609, 0.0685, 0.0685, 0.1065, 0.1111, 0.137, 0.1202, 0.0959, 0.0868, 0.105, 0.1187, 0.0533, 0.0107, 0.0167, 0.0457, 0.0335, -0.0137, -0.003, 0.0441, 0.0715, 0.0441, 0.0259, -0.0046, -0.0487, -0.0441, 0.0107, 0.0304, -0.0122, -0.0518, 0.0, 0.0046, 0.0289, 0.0289, 0.0289, 0.0289, 0.0289, 0.0289, 0.0289, 0.0289, 0.0137, 0.0441, 0.0594, 0.0624, 0.0624, 0.0472, 0.0213, -0.0502, -0.0046, 0.0076, -0.0883, -0.0685, 0.0244, 0.0091, -0.0289, -0.0487, -0.0289, 0.0046, -0.0198, 0.0274, -0.0289, -0.0533, -0.0411, -0.067, -0.0746]}
{"unit_id": "orig:1960176170588364966", "ticker_idx": 0, "ts": "2025-08-26T03:03:12+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman currently maintains a 'Buy' rating on Nvidia (NVDA) with a 12-month price target of $200", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman's tactically cautious view: NVDA, AVGO, AMD, MRVL all likely to underperform or consolidate in H2 2025 amid lack of new hard catalysts; NVDA Buy maintained with $200 PT on 12-month basis.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Nvidia’s stock likely to struggle to outperform large-cap peers in the second half… autumn will be a season of harvest\n\nAmid the AI frenzy, Nvidia’s stock seems to follow a certain seasonal rhythm of “harvest in autumn, conserve in winter.”\n\nIn a report released on August 24, Goldman Sachs presented a tactically cautious view on Nvidia’s stock. While maintaining a stance of being “very optimistic” about Nvidia’s long-term growth outlook, the bank argued that based on historical trading patterns, the stock is unlikely to outperform large-cap peers over the next few months.\n\nThe bank noted that Nvidia’s stock performance shows a pattern: it rises in the first half of the year due to clear positive data but comes under pressure in the second half due to a lack of new hard catalysts. For example, its stock surged 149% in the first half of 2024, but only rose 12% in the second half.\n\nAnalysts believe that until new quantitative data points emerge that are sufficient to drive earnings expectations higher, uncertainty over the long-term outlook (i.e., into 2027) will suppress stock trends through the end of 2025, similar to trading patterns in the second half of 2023 and 2024.\n\nFor investors, this means Nvidia may be entering a tactical consolidation period, an “AI autumn.” Goldman Sachs pointed out that from now until January 2026, when hyperscale computing customers provide formal capital expenditure guidance, the market will closely monitor three potential drivers: any quantitative comments from hyperscalers in October, the company’s disclosure on the timing of its next-generation “Rubin” platform launch, and further clarity on China operations and U.S. export controls.\n\nHowever, Goldman emphasized that this is only a short-term tactical judgment. The report reiterated that based on positive capex commentary from hyperscale customers and demand prospects from non-traditional customers, it remains “very optimistic” about Nvidia achieving extraordinary growth in 2026.\n\nHistory repeating? “Surge in the first half, consolidation in the second half”\n\nGoldman’s analysis revealed a distinct pattern in Nvidia’s stock performance. Nvidia’s stock usually performs strongly in the first half of each year, driven mainly by hyperscale customers like Amazon, Microsoft, Meta, and Google releasing annual capex “quantitative guidance.” These concrete numbers provide the market with solid growth expectations, boosting earnings forecasts and the stock price.\n\nHowever, in the second half, the situation often changes. Although these customers continue to make qualitative comments on capex direction, the lack of new hard data to further raise expectations allows other factors to weigh more heavily on the stock.\n\nThis pattern was especially evident in the past two years:\n\n2024 review: Nvidia’s stock surged 149% in the first half, thanks to better-than-expected capex guidance early in the year. But in the second half, market concerns about a “peak” in 2025 capex, competition from ASICs, and doubts over the timing of Blackwell’s launch emerged, leading to only a 12% gain in the second half.\n\n2023 review: Similarly, after ChatGPT ignited the AI narrative, Nvidia’s stock soared 189% in the first half of 2023. But in the second half, even with strong fundamentals, investors began questioning the sustainability of the spending boom, compressing valuations and limiting second-half gains to 17%.\n\nThree key variables determining performance before year-end\n\nLooking ahead to the remainder of 2025, Goldman believes that in the absence of new annual guidance from major customers, Nvidia’s stock performance will depend on the progress of three key variables.\n\n1. Hyperscale customer comments in October earnings reports.\nThe market will be searching for any quantitative clues on 2026 capex scale, not just directional descriptions.\n\n2. Timing of Nvidia’s next-generation “Rubin” architecture launch.\nAny clear information about the platform’s rollout will shape views on its future technological leadership and growth cycle.\n\n3. Clarity on China business.\nWith U.S. export control policies evolving, further visibility on China market prospects will significantly impact investor sentiment. Goldman believes that if no substantive progress is made on these fronts, the stock may come under pressure due to a lack of catalysts.\n\nDespite issuing a short-term caution, Goldman has not changed its long-term positive view on Nvidia. The report reiterated that based on positive commentary from cloud service providers and demand prospects from non-traditional customers, they remain “very optimistic” about Nvidia delivering excess growth in 2026.\n\nGoldman currently maintains a “Buy” rating on Nvidia (NVDA) with a 12-month price target of $200, based on standardized EPS of $5.75 and a P/E multiple of 35x.\n\nDivergence among AI peers: Broadcom aligned, AMD and Marvell may consolidate\n\nThe Goldman report also provided tactical outlooks for several other AI-related semiconductor companies.\n\nBroadcom: Expected to show trading dynamics in the second half of 2025 similar to Nvidia, as the market has largely digested its AI business guidance for fiscal 2026 and 2027. Incremental data from new XPU customers and AI networking will be key to whether the stock can outperform the market.\n\nAMD: Its potential growth in data center GPUs for 2026 and short-term strength in PC and server CPUs are already reflected in the current share price. The company’s investor day in November will be the “litmus test” for data center GPU revenue expectations and its 2026 outlook.\n\nMarvell: Expected to remain range-bound for the rest of the year. Visibility into growth from Amazon’s custom computing business and Microsoft’s business in the second half of 2026 will be key to a breakout.\n\n$NVDA $AMD $AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01078283957708237, "ret_signed_m1d": 0.01078283957708237, "alpha_spy_m1d": 0.006613261402572013, "alpha_c_m1d": 0.0012462736630559634, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0009352252999643706, "ret_signed_p1d": 0.0009352252999643706, "alpha_spy_p1d": 0.003213738581973713, "alpha_c_p1d": 0.004001740074807136, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.06046102686377797, "ret_signed_p1w": 0.06046102686377797, "alpha_spy_p1w": 0.0528815021516118, "alpha_c_p1w": 0.02585281936163819, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": -0.026352128487647786, "ret_signed_p1m": 0.026352128487647786, "alpha_spy_p1m": 0.0539000810695367, "alpha_c_p1m": 0.1084413858754818, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": -0.01584366868256315, "ret_signed_p3m": 0.01584366868256315, "alpha_spy_p3m": 0.040174301739457574, "alpha_c_p3m": 0.1148496148905338, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.034279923956166636, "ret_signed_p6m": -0.034279923956166636, "alpha_spy_p6m": 0.03557288954310378, "alpha_c_p6m": 0.3602733243718008, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.2343, -0.2566, -0.2443, -0.2231, -0.2193, -0.2299, -0.2204, -0.2154, -0.2081, -0.2142, -0.2023, -0.219, -0.204, -0.2071, -0.1996, -0.1996, -0.2086, -0.2069, -0.1863, -0.1511, -0.1472, -0.1321, -0.1308, -0.1566, -0.1349, -0.1234, -0.1234, -0.1294, -0.1198, -0.1039, -0.0972, -0.0927, -0.0974, -0.0609, -0.0572, -0.0482, -0.0515, -0.0572, -0.0811, -0.0605, -0.0442, -0.0455, -0.0276, -0.0344, -0.0138, -0.0215, -0.0443, -0.0097, -0.0193, -0.0129, -0.0055, 0.0051, 0.0016, 0.0076, -0.001, 0.0014, -0.0073, 0.0013, -0.0337, -0.035, -0.0374, -0.0208, -0.0108, 0.0, -0.0009, -0.0088, -0.0418, -0.0418, -0.0605, -0.0613, -0.0556, -0.0811, -0.074, -0.0606, -0.0244, -0.0253, -0.0217, -0.0221, -0.0379, -0.0631, -0.0304, -0.028, 0.0102, -0.0183, -0.0264, -0.0224, -0.0196, 0.0005, 0.0265, 0.0302, 0.0392, 0.0322, 0.0208, 0.018, 0.0404, 0.0595, 0.0077, 0.0361, -0.0095, -0.0106, 0.0003, 0.008, 0.0048, -0.0033, -0.0081, 0.0022, 0.0248, 0.0535, 0.106, 0.1391, 0.1163, 0.1141, 0.1382, 0.0931, 0.074, 0.0348, 0.0352, 0.0951, 0.0627, 0.0662, 0.0281, 0.0463, 0.0266, -0.0022, 0.0262, -0.0062, -0.0158, 0.0043, -0.0217, -0.0083, -0.0083, -0.0262, -0.0101, -0.0016, -0.0119, 0.009, 0.0036, 0.0209, 0.0177, 0.0112, -0.0045, -0.037, -0.03, -0.0222, -0.0595, -0.0419, -0.0042, 0.0107, 0.041, 0.0377, 0.0377, 0.0483, 0.0356, 0.0319, 0.0261, 0.0261, 0.0391, 0.0351, 0.0302, 0.0405, 0.0181, 0.0171, 0.0176, 0.0223, 0.0076, 0.0292, 0.0247, 0.0247, -0.0202, 0.0086, 0.017, 0.0326, 0.026, 0.0373, 0.0538, 0.0592, 0.0516, 0.0212, -0.0078, -0.0416, -0.0543, 0.0201, 0.0456, 0.0374, 0.0457, 0.0286, 0.0058, 0.0058, 0.0177, 0.0343]}
{"unit_id": "orig:1960176170588364966", "ticker_idx": 1, "ts": "2025-08-26T03:03:12+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom: Expected to show trading dynamics in the second half of 2025 similar to Nvidia, as the market has largely digested its AI business guidance for fiscal 2026 and 2027", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman's tactically cautious view: NVDA, AVGO, AMD, MRVL all likely to underperform or consolidate in H2 2025 amid lack of new hard catalysts; NVDA Buy maintained with $200 PT on 12-month basis.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Nvidia’s stock likely to struggle to outperform large-cap peers in the second half… autumn will be a season of harvest\n\nAmid the AI frenzy, Nvidia’s stock seems to follow a certain seasonal rhythm of “harvest in autumn, conserve in winter.”\n\nIn a report released on August 24, Goldman Sachs presented a tactically cautious view on Nvidia’s stock. While maintaining a stance of being “very optimistic” about Nvidia’s long-term growth outlook, the bank argued that based on historical trading patterns, the stock is unlikely to outperform large-cap peers over the next few months.\n\nThe bank noted that Nvidia’s stock performance shows a pattern: it rises in the first half of the year due to clear positive data but comes under pressure in the second half due to a lack of new hard catalysts. For example, its stock surged 149% in the first half of 2024, but only rose 12% in the second half.\n\nAnalysts believe that until new quantitative data points emerge that are sufficient to drive earnings expectations higher, uncertainty over the long-term outlook (i.e., into 2027) will suppress stock trends through the end of 2025, similar to trading patterns in the second half of 2023 and 2024.\n\nFor investors, this means Nvidia may be entering a tactical consolidation period, an “AI autumn.” Goldman Sachs pointed out that from now until January 2026, when hyperscale computing customers provide formal capital expenditure guidance, the market will closely monitor three potential drivers: any quantitative comments from hyperscalers in October, the company’s disclosure on the timing of its next-generation “Rubin” platform launch, and further clarity on China operations and U.S. export controls.\n\nHowever, Goldman emphasized that this is only a short-term tactical judgment. The report reiterated that based on positive capex commentary from hyperscale customers and demand prospects from non-traditional customers, it remains “very optimistic” about Nvidia achieving extraordinary growth in 2026.\n\nHistory repeating? “Surge in the first half, consolidation in the second half”\n\nGoldman’s analysis revealed a distinct pattern in Nvidia’s stock performance. Nvidia’s stock usually performs strongly in the first half of each year, driven mainly by hyperscale customers like Amazon, Microsoft, Meta, and Google releasing annual capex “quantitative guidance.” These concrete numbers provide the market with solid growth expectations, boosting earnings forecasts and the stock price.\n\nHowever, in the second half, the situation often changes. Although these customers continue to make qualitative comments on capex direction, the lack of new hard data to further raise expectations allows other factors to weigh more heavily on the stock.\n\nThis pattern was especially evident in the past two years:\n\n2024 review: Nvidia’s stock surged 149% in the first half, thanks to better-than-expected capex guidance early in the year. But in the second half, market concerns about a “peak” in 2025 capex, competition from ASICs, and doubts over the timing of Blackwell’s launch emerged, leading to only a 12% gain in the second half.\n\n2023 review: Similarly, after ChatGPT ignited the AI narrative, Nvidia’s stock soared 189% in the first half of 2023. But in the second half, even with strong fundamentals, investors began questioning the sustainability of the spending boom, compressing valuations and limiting second-half gains to 17%.\n\nThree key variables determining performance before year-end\n\nLooking ahead to the remainder of 2025, Goldman believes that in the absence of new annual guidance from major customers, Nvidia’s stock performance will depend on the progress of three key variables.\n\n1. Hyperscale customer comments in October earnings reports.\nThe market will be searching for any quantitative clues on 2026 capex scale, not just directional descriptions.\n\n2. Timing of Nvidia’s next-generation “Rubin” architecture launch.\nAny clear information about the platform’s rollout will shape views on its future technological leadership and growth cycle.\n\n3. Clarity on China business.\nWith U.S. export control policies evolving, further visibility on China market prospects will significantly impact investor sentiment. Goldman believes that if no substantive progress is made on these fronts, the stock may come under pressure due to a lack of catalysts.\n\nDespite issuing a short-term caution, Goldman has not changed its long-term positive view on Nvidia. The report reiterated that based on positive commentary from cloud service providers and demand prospects from non-traditional customers, they remain “very optimistic” about Nvidia delivering excess growth in 2026.\n\nGoldman currently maintains a “Buy” rating on Nvidia (NVDA) with a 12-month price target of $200, based on standardized EPS of $5.75 and a P/E multiple of 35x.\n\nDivergence among AI peers: Broadcom aligned, AMD and Marvell may consolidate\n\nThe Goldman report also provided tactical outlooks for several other AI-related semiconductor companies.\n\nBroadcom: Expected to show trading dynamics in the second half of 2025 similar to Nvidia, as the market has largely digested its AI business guidance for fiscal 2026 and 2027. Incremental data from new XPU customers and AI networking will be key to whether the stock can outperform the market.\n\nAMD: Its potential growth in data center GPUs for 2026 and short-term strength in PC and server CPUs are already reflected in the current share price. The company’s investor day in November will be the “litmus test” for data center GPU revenue expectations and its 2026 outlook.\n\nMarvell: Expected to remain range-bound for the rest of the year. Visibility into growth from Amazon’s custom computing business and Microsoft’s business in the second half of 2026 will be key to a breakout.\n\n$NVDA $AMD $AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012684067074898997, "ret_signed_m1d": 0.012684067074898997, "alpha_spy_m1d": 0.00851448890038864, "alpha_c_m1d": 0.00314750116087259, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007516552853781278, "ret_signed_p1d": -0.007516552853781278, "alpha_spy_p1d": -0.005238039571771935, "alpha_c_p1d": -0.004450038078938512, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": 0.0007718210940537151, "ret_signed_p1w": -0.0007718210940537151, "alpha_spy_p1w": -0.008351345806219879, "alpha_c_p1w": -0.03538002859619349, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": 0.14053683895696434, "ret_signed_p1m": -0.14053683895696434, "alpha_spy_p1m": -0.11298888637507543, "alpha_c_p1m": -0.058447581569130325, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.14352844388343633, "ret_signed_p3m": -0.14352844388343633, "alpha_spy_p3m": -0.1191978108265419, "alpha_c_p3m": -0.04452249767546568, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.12318600414604086, "ret_signed_p6m": -0.12318600414604086, "alpha_spy_p6m": -0.053333190646770445, "alpha_c_p6m": 0.2713672441819266, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.19, -0.1896, -0.1674, -0.1401, -0.126, -0.1298, -0.1733, -0.1822, -0.181, -0.1533, -0.1428, -0.1674, -0.156, -0.1652, -0.1589, -0.1589, -0.1611, -0.1485, -0.1149, -0.1119, -0.0934, -0.0962, -0.075, -0.1116, -0.0943, -0.0766, -0.0766, -0.08, -0.088, -0.0675, -0.0759, -0.0793, -0.0752, -0.0573, -0.0577, -0.0388, -0.0492, -0.0329, -0.0652, -0.0481, -0.0312, -0.0263, -0.0124, -0.002, 0.0155, -0.0145, -0.0314, -0.001, -0.017, 0.0123, 0.0193, 0.0234, 0.0198, 0.0497, 0.0372, 0.0444, 0.028, 0.026, -0.0104, -0.023, -0.0282, -0.0135, -0.0127, 0.0, 0.0075, 0.0357, -0.0021, -0.0021, 0.0008, 0.0147, 0.0271, 0.1238, 0.1599, 0.1297, 0.2401, 0.2068, 0.2076, 0.2217, 0.208, 0.1616, 0.1589, 0.1575, 0.1388, 0.1393, 0.1405, 0.1297, 0.1245, 0.1022, 0.1089, 0.1206, 0.1367, 0.1374, 0.1277, 0.1308, 0.1613, 0.1597, 0.0912, 0.199, 0.1567, 0.1809, 0.1904, 0.1742, 0.1739, 0.1518, 0.1439, 0.1573, 0.1904, 0.217, 0.2537, 0.2974, 0.2654, 0.2425, 0.2187, 0.183, 0.2067, 0.1953, 0.1746, 0.2047, 0.1831, 0.194, 0.1428, 0.1511, 0.1518, 0.1445, 0.1913, 0.1658, 0.1435, 0.2705, 0.2942, 0.3364, 0.3364, 0.3545, 0.2977, 0.2826, 0.2794, 0.2808, 0.3117, 0.3482, 0.3657, 0.3881, 0.3659, 0.2098, 0.1422, 0.1472, 0.0959, 0.1088, 0.1441, 0.1499, 0.1764, 0.1795, 0.1795, 0.1859, 0.1767, 0.1782, 0.1656, 0.1656, 0.1707, 0.1566, 0.1577, 0.1568, 0.1197, 0.1618, 0.1862, 0.1942, 0.1447, 0.1552, 0.1845, 0.1845, 0.1201, 0.1073, 0.0962, 0.0779, 0.094, 0.1208, 0.1223, 0.1138, 0.1157, 0.1151, 0.0788, 0.0374, 0.0457, 0.1212, 0.1583, 0.1465, 0.1543, 0.1153, 0.0951, 0.0951, 0.1199, 0.1232]}
{"unit_id": "orig:1960176170588364966", "ticker_idx": 2, "ts": "2025-08-26T03:03:12+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Its potential growth in data center GPUs for 2026 and short-term strength in PC and server CPUs are already reflected in the current share price", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman's tactically cautious view: NVDA, AVGO, AMD, MRVL all likely to underperform or consolidate in H2 2025 amid lack of new hard catalysts; NVDA Buy maintained with $200 PT on 12-month basis.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Nvidia’s stock likely to struggle to outperform large-cap peers in the second half… autumn will be a season of harvest\n\nAmid the AI frenzy, Nvidia’s stock seems to follow a certain seasonal rhythm of “harvest in autumn, conserve in winter.”\n\nIn a report released on August 24, Goldman Sachs presented a tactically cautious view on Nvidia’s stock. While maintaining a stance of being “very optimistic” about Nvidia’s long-term growth outlook, the bank argued that based on historical trading patterns, the stock is unlikely to outperform large-cap peers over the next few months.\n\nThe bank noted that Nvidia’s stock performance shows a pattern: it rises in the first half of the year due to clear positive data but comes under pressure in the second half due to a lack of new hard catalysts. For example, its stock surged 149% in the first half of 2024, but only rose 12% in the second half.\n\nAnalysts believe that until new quantitative data points emerge that are sufficient to drive earnings expectations higher, uncertainty over the long-term outlook (i.e., into 2027) will suppress stock trends through the end of 2025, similar to trading patterns in the second half of 2023 and 2024.\n\nFor investors, this means Nvidia may be entering a tactical consolidation period, an “AI autumn.” Goldman Sachs pointed out that from now until January 2026, when hyperscale computing customers provide formal capital expenditure guidance, the market will closely monitor three potential drivers: any quantitative comments from hyperscalers in October, the company’s disclosure on the timing of its next-generation “Rubin” platform launch, and further clarity on China operations and U.S. export controls.\n\nHowever, Goldman emphasized that this is only a short-term tactical judgment. The report reiterated that based on positive capex commentary from hyperscale customers and demand prospects from non-traditional customers, it remains “very optimistic” about Nvidia achieving extraordinary growth in 2026.\n\nHistory repeating? “Surge in the first half, consolidation in the second half”\n\nGoldman’s analysis revealed a distinct pattern in Nvidia’s stock performance. Nvidia’s stock usually performs strongly in the first half of each year, driven mainly by hyperscale customers like Amazon, Microsoft, Meta, and Google releasing annual capex “quantitative guidance.” These concrete numbers provide the market with solid growth expectations, boosting earnings forecasts and the stock price.\n\nHowever, in the second half, the situation often changes. Although these customers continue to make qualitative comments on capex direction, the lack of new hard data to further raise expectations allows other factors to weigh more heavily on the stock.\n\nThis pattern was especially evident in the past two years:\n\n2024 review: Nvidia’s stock surged 149% in the first half, thanks to better-than-expected capex guidance early in the year. But in the second half, market concerns about a “peak” in 2025 capex, competition from ASICs, and doubts over the timing of Blackwell’s launch emerged, leading to only a 12% gain in the second half.\n\n2023 review: Similarly, after ChatGPT ignited the AI narrative, Nvidia’s stock soared 189% in the first half of 2023. But in the second half, even with strong fundamentals, investors began questioning the sustainability of the spending boom, compressing valuations and limiting second-half gains to 17%.\n\nThree key variables determining performance before year-end\n\nLooking ahead to the remainder of 2025, Goldman believes that in the absence of new annual guidance from major customers, Nvidia’s stock performance will depend on the progress of three key variables.\n\n1. Hyperscale customer comments in October earnings reports.\nThe market will be searching for any quantitative clues on 2026 capex scale, not just directional descriptions.\n\n2. Timing of Nvidia’s next-generation “Rubin” architecture launch.\nAny clear information about the platform’s rollout will shape views on its future technological leadership and growth cycle.\n\n3. Clarity on China business.\nWith U.S. export control policies evolving, further visibility on China market prospects will significantly impact investor sentiment. Goldman believes that if no substantive progress is made on these fronts, the stock may come under pressure due to a lack of catalysts.\n\nDespite issuing a short-term caution, Goldman has not changed its long-term positive view on Nvidia. The report reiterated that based on positive commentary from cloud service providers and demand prospects from non-traditional customers, they remain “very optimistic” about Nvidia delivering excess growth in 2026.\n\nGoldman currently maintains a “Buy” rating on Nvidia (NVDA) with a 12-month price target of $200, based on standardized EPS of $5.75 and a P/E multiple of 35x.\n\nDivergence among AI peers: Broadcom aligned, AMD and Marvell may consolidate\n\nThe Goldman report also provided tactical outlooks for several other AI-related semiconductor companies.\n\nBroadcom: Expected to show trading dynamics in the second half of 2025 similar to Nvidia, as the market has largely digested its AI business guidance for fiscal 2026 and 2027. Incremental data from new XPU customers and AI networking will be key to whether the stock can outperform the market.\n\nAMD: Its potential growth in data center GPUs for 2026 and short-term strength in PC and server CPUs are already reflected in the current share price. The company’s investor day in November will be the “litmus test” for data center GPU revenue expectations and its 2026 outlook.\n\nMarvell: Expected to remain range-bound for the rest of the year. Visibility into growth from Amazon’s custom computing business and Microsoft’s business in the second half of 2026 will be key to a breakout.\n\n$NVDA $AMD $AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019565445939085047, "ret_signed_m1d": 0.019565445939085047, "alpha_spy_m1d": 0.015395867764574689, "alpha_c_m1d": 0.01002888002505864, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0030609157398324705, "ret_signed_p1d": -0.0030609157398324705, "alpha_spy_p1d": -0.0007824024578231281, "alpha_c_p1d": 5.599035010295239e-06, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.025807153516866133, "ret_signed_p1w": 0.025807153516866133, "alpha_spy_p1w": 0.01822762880469997, "alpha_c_p1w": -0.008801053985273644, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": -0.034449588300239364, "ret_signed_p1m": 0.034449588300239364, "alpha_spy_p1m": 0.06199754088212828, "alpha_c_p1m": 0.11653884568807338, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.22302247479946158, "ret_signed_p3m": -0.22302247479946158, "alpha_spy_p3m": -0.19869184174256715, "alpha_c_p3m": -0.12401652859149093, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.2010563016547846, "ret_signed_p6m": -0.2010563016547846, "alpha_spy_p6m": -0.1312034881555142, "alpha_c_p6m": 0.19349694667318285, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.3216, -0.3354, -0.312, -0.2959, -0.2883, -0.3057, -0.3027, -0.2694, -0.2604, -0.273, -0.2888, -0.3028, -0.2414, -0.2372, -0.239, -0.239, -0.2303, -0.2223, -0.1692, -0.1394, -0.1377, -0.1369, -0.1484, -0.1831, -0.1686, -0.1723, -0.1723, -0.191, -0.1728, -0.1693, -0.1348, -0.1212, -0.1223, -0.0661, -0.0393, -0.0373, -0.0578, -0.0577, -0.0714, -0.0478, -0.027, -0.0009, 0.0423, 0.0649, 0.0774, 0.0582, 0.0305, 0.061, 0.0462, -0.021, 0.0347, 0.0369, 0.034, 0.05, 0.1068, 0.086, 0.0654, 0.0571, -0.0004, -0.0085, -0.0175, 0.0068, -0.0196, 0.0, 0.0031, 0.0118, -0.0239, -0.0239, -0.0258, -0.0269, -0.029, -0.0929, -0.0913, -0.0648, -0.0425, -0.0657, -0.0483, -0.0328, -0.037, -0.0448, -0.0522, -0.0554, -0.041, -0.0343, -0.0344, -0.0321, -0.043, -0.0316, -0.029, -0.0157, 0.0187, -0.0117, 0.2226, 0.2694, 0.4138, 0.3977, 0.2898, 0.2989, 0.3089, 0.432, 0.4078, 0.3989, 0.4438, 0.4286, 0.3818, 0.4103, 0.5179, 0.5585, 0.5485, 0.5864, 0.5295, 0.5372, 0.5583, 0.5007, 0.5384, 0.4266, 0.4016, 0.4643, 0.4255, 0.5538, 0.4882, 0.4813, 0.4435, 0.3821, 0.3417, 0.2365, 0.223, 0.2907, 0.2371, 0.2858, 0.2858, 0.3055, 0.3189, 0.2918, 0.306, 0.2962, 0.3082, 0.327, 0.3301, 0.3289, 0.329, 0.265, 0.2458, 0.2554, 0.189, 0.2067, 0.2809, 0.2901, 0.2898, 0.2906, 0.2906, 0.2903, 0.294, 0.2924, 0.2853, 0.2853, 0.3412, 0.3269, 0.2865, 0.2605, 0.2284, 0.2194, 0.2465, 0.3262, 0.342, 0.3679, 0.3914, 0.3914, 0.3919, 0.4992, 0.5228, 0.5585, 0.5083, 0.5126, 0.5169, 0.5135, 0.4208, 0.478, 0.4531, 0.2015, 0.1553, 0.251, 0.2964, 0.2818, 0.2818, 0.236, 0.2443, 0.2443, 0.2188, 0.2011]}
{"unit_id": "orig:1960176170588364966", "ticker_idx": 3, "ts": "2025-08-26T03:03:12+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Marvell: Expected to remain range-bound for the rest of the year", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman's tactically cautious view: NVDA, AVGO, AMD, MRVL all likely to underperform or consolidate in H2 2025 amid lack of new hard catalysts; NVDA Buy maintained with $200 PT on 12-month basis.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Nvidia’s stock likely to struggle to outperform large-cap peers in the second half… autumn will be a season of harvest\n\nAmid the AI frenzy, Nvidia’s stock seems to follow a certain seasonal rhythm of “harvest in autumn, conserve in winter.”\n\nIn a report released on August 24, Goldman Sachs presented a tactically cautious view on Nvidia’s stock. While maintaining a stance of being “very optimistic” about Nvidia’s long-term growth outlook, the bank argued that based on historical trading patterns, the stock is unlikely to outperform large-cap peers over the next few months.\n\nThe bank noted that Nvidia’s stock performance shows a pattern: it rises in the first half of the year due to clear positive data but comes under pressure in the second half due to a lack of new hard catalysts. For example, its stock surged 149% in the first half of 2024, but only rose 12% in the second half.\n\nAnalysts believe that until new quantitative data points emerge that are sufficient to drive earnings expectations higher, uncertainty over the long-term outlook (i.e., into 2027) will suppress stock trends through the end of 2025, similar to trading patterns in the second half of 2023 and 2024.\n\nFor investors, this means Nvidia may be entering a tactical consolidation period, an “AI autumn.” Goldman Sachs pointed out that from now until January 2026, when hyperscale computing customers provide formal capital expenditure guidance, the market will closely monitor three potential drivers: any quantitative comments from hyperscalers in October, the company’s disclosure on the timing of its next-generation “Rubin” platform launch, and further clarity on China operations and U.S. export controls.\n\nHowever, Goldman emphasized that this is only a short-term tactical judgment. The report reiterated that based on positive capex commentary from hyperscale customers and demand prospects from non-traditional customers, it remains “very optimistic” about Nvidia achieving extraordinary growth in 2026.\n\nHistory repeating? “Surge in the first half, consolidation in the second half”\n\nGoldman’s analysis revealed a distinct pattern in Nvidia’s stock performance. Nvidia’s stock usually performs strongly in the first half of each year, driven mainly by hyperscale customers like Amazon, Microsoft, Meta, and Google releasing annual capex “quantitative guidance.” These concrete numbers provide the market with solid growth expectations, boosting earnings forecasts and the stock price.\n\nHowever, in the second half, the situation often changes. Although these customers continue to make qualitative comments on capex direction, the lack of new hard data to further raise expectations allows other factors to weigh more heavily on the stock.\n\nThis pattern was especially evident in the past two years:\n\n2024 review: Nvidia’s stock surged 149% in the first half, thanks to better-than-expected capex guidance early in the year. But in the second half, market concerns about a “peak” in 2025 capex, competition from ASICs, and doubts over the timing of Blackwell’s launch emerged, leading to only a 12% gain in the second half.\n\n2023 review: Similarly, after ChatGPT ignited the AI narrative, Nvidia’s stock soared 189% in the first half of 2023. But in the second half, even with strong fundamentals, investors began questioning the sustainability of the spending boom, compressing valuations and limiting second-half gains to 17%.\n\nThree key variables determining performance before year-end\n\nLooking ahead to the remainder of 2025, Goldman believes that in the absence of new annual guidance from major customers, Nvidia’s stock performance will depend on the progress of three key variables.\n\n1. Hyperscale customer comments in October earnings reports.\nThe market will be searching for any quantitative clues on 2026 capex scale, not just directional descriptions.\n\n2. Timing of Nvidia’s next-generation “Rubin” architecture launch.\nAny clear information about the platform’s rollout will shape views on its future technological leadership and growth cycle.\n\n3. Clarity on China business.\nWith U.S. export control policies evolving, further visibility on China market prospects will significantly impact investor sentiment. Goldman believes that if no substantive progress is made on these fronts, the stock may come under pressure due to a lack of catalysts.\n\nDespite issuing a short-term caution, Goldman has not changed its long-term positive view on Nvidia. The report reiterated that based on positive commentary from cloud service providers and demand prospects from non-traditional customers, they remain “very optimistic” about Nvidia delivering excess growth in 2026.\n\nGoldman currently maintains a “Buy” rating on Nvidia (NVDA) with a 12-month price target of $200, based on standardized EPS of $5.75 and a P/E multiple of 35x.\n\nDivergence among AI peers: Broadcom aligned, AMD and Marvell may consolidate\n\nThe Goldman report also provided tactical outlooks for several other AI-related semiconductor companies.\n\nBroadcom: Expected to show trading dynamics in the second half of 2025 similar to Nvidia, as the market has largely digested its AI business guidance for fiscal 2026 and 2027. Incremental data from new XPU customers and AI networking will be key to whether the stock can outperform the market.\n\nAMD: Its potential growth in data center GPUs for 2026 and short-term strength in PC and server CPUs are already reflected in the current share price. The company’s investor day in November will be the “litmus test” for data center GPU revenue expectations and its 2026 outlook.\n\nMarvell: Expected to remain range-bound for the rest of the year. Visibility into growth from Amazon’s custom computing business and Microsoft’s business in the second half of 2026 will be key to a breakout.\n\n$NVDA $AMD $AVGO $MRVL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017640767834846915, "ret_signed_m1d": 0.017640767834846915, "alpha_spy_m1d": 0.013471189660336558, "alpha_c_m1d": 0.008104201920820508, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007137026755229758, "ret_signed_p1d": -0.007137026755229758, "alpha_spy_p1d": -0.004858513473220416, "alpha_c_p1d": -0.004070511980386993, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.13008355026345564, "ret_signed_p1w": 0.13008355026345564, "alpha_spy_p1w": 0.12250402555128947, "alpha_c_p1w": 0.09547534276131586, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": 0.07850780896973886, "ret_signed_p1m": -0.07850780896973886, "alpha_spy_p1m": -0.050959856387849944, "alpha_c_p1m": 0.0035814484180951567, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.043647678719026306, "ret_signed_p3m": -0.043647678719026306, "alpha_spy_p3m": -0.019317045662131882, "alpha_c_p3m": 0.055358267488944346, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.0665136061248921, "ret_signed_p6m": -0.0665136061248921, "alpha_spy_p6m": 0.003339207374378317, "alpha_c_p6m": 0.32803964220307535, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.1425, -0.1901, -0.1729, -0.1609, -0.1079, -0.1233, -0.0803, -0.0697, -0.0737, -0.0818, -0.063, -0.0959, -0.0525, -0.0583, 0.0085, 0.0085, -0.0109, -0.0476, 0.012, 0.0217, 0.076, 0.0382, 0.0414, 0.0258, -0.001, 0.0116, 0.0116, -0.0373, -0.0319, -0.0277, -0.0129, -0.0209, -0.0236, -0.0249, -0.0459, -0.0303, 0.0053, -0.0162, -0.0306, -0.0133, -0.003, -0.0007, 0.0222, 0.028, 0.1007, 0.0823, 0.0026, 0.0306, 0.0319, 0.0143, 0.0214, 0.0415, 0.0407, 0.0478, 0.0681, 0.0644, 0.026, 0.0334, -0.0295, -0.0409, -0.0411, -0.017, -0.0176, 0.0, 0.0071, 0.04, -0.1534, -0.1534, -0.1301, -0.1609, -0.1368, -0.1472, -0.1112, -0.0999, -0.0964, -0.1033, -0.0931, -0.092, -0.0727, -0.0442, -0.0004, 0.0, 0.0171, 0.0048, 0.0785, 0.1286, 0.12, 0.1095, 0.1321, 0.1297, 0.1608, 0.1611, 0.1974, 0.1712, 0.2456, 0.2211, 0.1536, 0.2045, 0.1618, 0.1978, 0.1889, 0.1851, 0.1567, 0.1354, 0.0922, 0.1153, 0.1337, 0.1954, 0.192, 0.2148, 0.1935, 0.2632, 0.2177, 0.1803, 0.2518, 0.2576, 0.2252, 0.2563, 0.2037, 0.2037, 0.1793, 0.1649, 0.1245, 0.0602, 0.0958, 0.0333, 0.0436, 0.1291, 0.1242, 0.182, 0.182, 0.2047, 0.2276, 0.2517, 0.3502, 0.3231, 0.3328, 0.2397, 0.1979, 0.246, 0.2051, 0.1377, 0.1354, 0.1329, 0.1009, 0.1382, 0.1331, 0.1427, 0.1815, 0.1655, 0.1655, 0.1634, 0.1556, 0.1691, 0.1451, 0.1451, 0.2045, 0.2159, 0.1889, 0.1405, 0.1245, 0.1222, 0.1178, 0.1199, 0.0951, 0.0839, 0.085, 0.085, 0.0761, 0.1132, 0.1206, 0.0819, 0.1027, 0.1183, 0.1276, 0.0969, 0.0642, 0.0607, 0.0186, -0.0058, 0.0007, 0.0826, 0.1105, 0.1059, 0.0969, 0.0549, 0.06, 0.06, 0.0654, 0.0665]}
{"unit_id": "quote:1948556756424294626", "ticker_idx": 0, "ts": "2025-07-25T01:31:47+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The H20 has definitely gone back into production.", "tweet_type": "prediction", "timeline": "months", "summary": "Author confirms H20 back in production, validating TSMC CoWoS capacity orders; bullish for both NVDA and TSM.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I heard about it a few days ago too.\nIt's true.\nThe H20 has definitely gone back into production.", "reply_text": null, "parent_text": null, "quoted_text": "Rumor: Nvidia has already started to prep supply chain partners for H20 production, including asking TSMC for 5,000 to 9,000 wafers of capacity, media report, adding H20 production takes around 4-months due to CoWoS advanced packaging. $NVDA $TSM #China #semiconductors", "parent_author": null, "ret_m1d": 0.0013832829299138538, "ret_signed_m1d": 0.0013832829299138538, "alpha_spy_m1d": 0.0055898625073081964, "alpha_c_m1d": 0.0003049387265330328, "bench_spy_m1d": -0.004206579577394343, "bench_c_m1d": 0.001078344203380821, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01873193435220344, "ret_signed_p1d": 0.01873193435220344, "alpha_spy_p1d": 0.018983034497395135, "alpha_c_p1d": 0.005096582295768037, "bench_spy_p1d": -0.0002511001451916961, "bench_c_p1d": 0.013635352056435401, "ret_p1w": 0.0012679653716589634, "ret_signed_p1w": 0.0012679653716589634, "alpha_spy_p1w": 0.025408746856370956, "alpha_c_p1w": 0.013581342185182144, "bench_spy_p1w": -0.024140781484711993, "bench_c_p1w": -0.01231337681352318, "ret_p1m": 0.036368835689368106, "ret_signed_p1m": 0.036368835689368106, "alpha_spy_p1m": 0.0279401216583286, "alpha_c_p1m": 0.014002696996435748, "bench_spy_p1m": 0.008428714031039508, "bench_c_p1m": 0.022366138692932358, "ret_p3m": 0.0391363691259623, "ret_signed_p3m": 0.0391363691259623, "alpha_spy_p3m": -0.011956761879331035, "alpha_c_p3m": -0.13864379858334153, "bench_spy_p3m": 0.051093131005293335, "bench_c_p3m": 0.17778016770930383, "ret_p6m": 0.07349194144893922, "ret_signed_p6m": 0.07349194144893922, "alpha_spy_p6m": -0.0183727423349187, "alpha_c_p6m": -0.32355105069800616, "bench_spy_p6m": 0.09186468378385793, "bench_c_p6m": 0.3970429921469454, "price_path": [-0.3717, -0.3723, -0.3568, -0.3401, -0.344, -0.3456, -0.3253, -0.3236, -0.3277, -0.2911, -0.2512, -0.22, -0.2229, -0.2197, -0.2187, -0.2255, -0.2404, -0.2345, -0.2433, -0.2433, -0.2191, -0.2231, -0.1978, -0.2212, -0.2082, -0.1861, -0.1821, -0.1932, -0.1832, -0.178, -0.1703, -0.1768, -0.1643, -0.1817, -0.1661, -0.1693, -0.1615, -0.1615, -0.1709, -0.169, -0.1476, -0.1106, -0.1065, -0.0908, -0.0894, -0.1164, -0.0937, -0.0816, -0.0816, -0.088, -0.0778, -0.0612, -0.0542, -0.0495, -0.0544, -0.0161, -0.0123, -0.0029, -0.0063, -0.0122, -0.0373, -0.0157, 0.0014, 0.0, 0.0187, 0.0116, 0.0333, 0.0252, 0.0013, 0.0375, 0.0274, 0.0341, 0.0419, 0.053, 0.0493, 0.0557, 0.0466, 0.0491, 0.0401, 0.049, 0.0123, 0.011, 0.0085, 0.0259, 0.0364, 0.0477, 0.0467, 0.0384, 0.0039, 0.0039, -0.0157, -0.0166, -0.0106, -0.0373, -0.0299, -0.0158, 0.0221, 0.0212, 0.025, 0.0246, 0.008, -0.0184, 0.0158, 0.0183, 0.0583, 0.0285, 0.0201, 0.0242, 0.0271, 0.0482, 0.0754, 0.0793, 0.0888, 0.0814, 0.0695, 0.0666, 0.09, 0.11, 0.0557, 0.0855, 0.0377, 0.0365, 0.048, 0.0561, 0.0527, 0.0442, 0.0391, 0.05, 0.0736, 0.1038, 0.1587, 0.1934, 0.1695, 0.1672, 0.1925, 0.1453, 0.1252, 0.0841, 0.0845, 0.1473, 0.1134, 0.1171, 0.0771, 0.0961, 0.0756, 0.0454, 0.0751, 0.0412, 0.0311, 0.0522, 0.025, 0.039, 0.039, 0.0202, 0.0371, 0.0459, 0.0352, 0.0571, 0.0515, 0.0696, 0.0662, 0.0594, 0.0429, 0.0089, 0.0162, 0.0244, -0.0146, 0.0038, 0.0433, 0.0589, 0.0907, 0.0872, 0.0872, 0.0983, 0.085, 0.081, 0.075, 0.075, 0.0886, 0.0844, 0.0793, 0.0901, 0.0666, 0.0656, 0.0661, 0.0711, 0.0557, 0.0782, 0.0735, 0.0735]}
{"unit_id": "quote:1948556756424294626", "ticker_idx": 1, "ts": "2025-07-25T01:31:47+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The H20 has definitely gone back into production.", "tweet_type": "prediction", "timeline": "months", "summary": "Author confirms H20 back in production, validating TSMC CoWoS capacity orders; bullish for both NVDA and TSM.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I heard about it a few days ago too.\nIt's true.\nThe H20 has definitely gone back into production.", "reply_text": null, "parent_text": null, "quoted_text": "Rumor: Nvidia has already started to prep supply chain partners for H20 production, including asking TSMC for 5,000 to 9,000 wafers of capacity, media report, adding H20 production takes around 4-months due to CoWoS advanced packaging. $NVDA $TSM #China #semiconductors", "parent_author": null, "ret_m1d": -0.016286599225248666, "ret_signed_m1d": -0.016286599225248666, "alpha_spy_m1d": -0.012080019647854323, "alpha_c_m1d": -0.017364943428629487, "bench_spy_m1d": -0.004206579577394343, "bench_c_m1d": 0.001078344203380821, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011604161996523032, "ret_signed_p1d": -0.011604161996523032, "alpha_spy_p1d": -0.011353061851331336, "alpha_c_p1d": -0.025239514052958434, "bench_spy_p1d": -0.0002511001451916961, "bench_c_p1d": 0.013635352056435401, "ret_p1w": -0.042304498359671094, "ret_signed_p1w": -0.042304498359671094, "alpha_spy_p1w": -0.0181637168749591, "alpha_c_p1w": -0.029991121546147914, "bench_spy_p1w": -0.024140781484711993, "bench_c_p1w": -0.01231337681352318, "ret_p1m": -0.04075732689530853, "ret_signed_p1m": -0.04075732689530853, "alpha_spy_p1m": -0.04918604092634804, "alpha_c_p1m": -0.06312346558824089, "bench_spy_p1m": 0.008428714031039508, "bench_c_p1m": 0.022366138692932358, "ret_p3m": 0.1799323187687103, "ret_signed_p3m": 0.1799323187687103, "alpha_spy_p3m": 0.12883918776341696, "alpha_c_p3m": 0.0021521510594064708, "bench_spy_p3m": 0.051093131005293335, "bench_c_p3m": 0.17778016770930383, "ret_p6m": 0.4023105170872894, "ret_signed_p6m": 0.4023105170872894, "alpha_spy_p6m": 0.31044583330343145, "alpha_c_p6m": 0.005267524940343993, "bench_spy_p6m": 0.09186468378385793, "bench_c_p6m": 0.3970429921469454, "price_path": [-0.3327, -0.3238, -0.2993, -0.2727, -0.2844, -0.3011, -0.2919, -0.2892, -0.2839, -0.2415, -0.213, -0.2099, -0.2121, -0.2121, -0.215, -0.2152, -0.2221, -0.2041, -0.2212, -0.2212, -0.198, -0.2043, -0.2002, -0.2157, -0.2096, -0.1983, -0.1789, -0.1751, -0.1676, -0.1602, -0.1381, -0.1314, -0.1228, -0.1405, -0.1218, -0.1291, -0.1307, -0.1307, -0.1469, -0.1436, -0.1039, -0.0931, -0.0879, -0.0693, -0.0778, -0.0852, -0.0489, -0.044, -0.044, -0.0669, -0.0722, -0.056, -0.0645, -0.0619, -0.0689, -0.0352, -0.0327, 0.0, -0.0212, -0.0275, -0.0448, -0.0215, -0.0163, 0.0, -0.0116, -0.0174, -0.011, -0.0162, -0.0423, -0.0269, -0.0535, -0.0579, -0.0121, -0.0154, -0.0143, -0.0053, -0.0169, -0.0187, -0.0274, -0.0171, -0.0525, -0.0692, -0.0744, -0.0513, -0.0408, -0.028, -0.0257, -0.0298, -0.06, -0.06, -0.0701, -0.0579, -0.0423, -0.0089, 0.0065, 0.0217, 0.0604, 0.0542, 0.0559, 0.0643, 0.0704, 0.0734, 0.0973, 0.0819, 0.1136, 0.1547, 0.1466, 0.13, 0.1165, 0.116, 0.1408, 0.1783, 0.1768, 0.1935, 0.2352, 0.201, 0.2438, 0.2249, 0.1464, 0.2372, 0.2088, 0.2446, 0.2247, 0.2053, 0.216, 0.2029, 0.1799, 0.1875, 0.2048, 0.2182, 0.2316, 0.2461, 0.2385, 0.2271, 0.2452, 0.201, 0.1994, 0.1814, 0.1702, 0.206, 0.1893, 0.187, 0.1526, 0.1633, 0.1519, 0.1351, 0.1533, 0.1335, 0.1235, 0.1626, 0.1628, 0.1843, 0.1843, 0.1907, 0.175, 0.193, 0.2068, 0.1965, 0.2038, 0.233, 0.2393, 0.2668, 0.2485, 0.1961, 0.1784, 0.1749, 0.1343, 0.1659, 0.1834, 0.2011, 0.2162, 0.2237, 0.2237, 0.2403, 0.2324, 0.2269, 0.2446, 0.2446, 0.309, 0.3198, 0.341, 0.3052, 0.3024, 0.3254, 0.3588, 0.3565, 0.3397, 0.3992, 0.4023, 0.4023]}
{"unit_id": "quote:1966056144574779617", "ticker_idx": 0, "ts": "2025-09-11T08:28:07+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This absolutely does not mean that one should short Micron. As wafers are increasingly consumed by HBM, it is almost certain that the price of general-purpose products like DDR5 will rise. Micron can achieve profit margins of over 50% on DDR5.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Micron: DDR5 price rises as wafers shift to HBM, with 50%+ margins possible even outside HBM.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "This absolutely does not mean that one should short Micron. Why is everyone thinking in such extreme terms?\n\nAs wafers are increasingly consumed by HBM, it is almost certain that the price of general-purpose products like DDR5 will rise.\n\nMicron’s decision to allocate part of its HBM capacity to DDR5 comes from the judgment that even selling DDR5, it can still generate sufficient high profits. Of course, not as much as selling HBM, but I believe Micron can achieve profit margins of over 50% on DDR5.", "reply_text": null, "parent_text": null, "quoted_text": "To give you some additional context on this,\n\nNVIDIA originally requested 9Gbps for HBM4.\n\nHowever, when Samsung failed on d1b and instead applied d1c and SF4 to HBM4, it turned out they could hit over 10Gbps.\n\nMicron said 10Gbps was impossible, while SK Hynix responded by saying", "parent_author": null, "ret_m1d": -0.0701998222349961, "ret_signed_m1d": -0.0701998222349961, "alpha_spy_m1d": -0.06195804882213063, "alpha_c_m1d": -0.060000074745877874, "bench_spy_m1d": -0.008241773412865472, "bench_c_m1d": -0.010199747489118227, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04423188285777435, "ret_signed_p1d": 0.04423188285777435, "alpha_spy_p1d": 0.044566495769602144, "alpha_c_p1d": 0.04294857964881138, "bench_spy_p1d": -0.00033461291182779185, "bench_c_p1d": 0.0012833032089629715, "ret_p1w": 0.1216709910368523, "ret_signed_p1w": 0.1216709910368523, "alpha_spy_p1w": 0.11463064177397886, "alpha_c_p1w": 0.07876630004540708, "bench_spy_p1w": 0.0070403492628734465, "bench_c_p1w": 0.04290469099144523, "ret_p1m": 0.206838857992798, "ret_signed_p1m": 0.206838857992798, "alpha_spy_p1m": 0.21109597661340285, "alpha_c_p1m": 0.13639496964752995, "bench_spy_p1m": -0.0042571186206048495, "bench_c_p1m": 0.07044388834526805, "ret_p3m": 0.6774794189004509, "ret_signed_p3m": 0.6774794189004509, "alpha_spy_p3m": 0.6359613039869827, "alpha_c_p3m": 0.4633840338465367, "bench_spy_p3m": 0.04151811491346824, "bench_c_p3m": 0.21409538505391423, "ret_p6m": 1.4618555232081976, "ret_signed_p6m": 1.4618555232081976, "alpha_spy_p6m": 1.4335624337463702, "alpha_c_p6m": 1.205828928484134, "bench_spy_p6m": 0.028293089461827403, "bench_c_p6m": 0.25602659472406364, "price_path": [-0.2048, -0.2015, -0.1917, -0.1917, -0.1799, -0.19, -0.1513, -0.1557, -0.164, -0.1722, -0.1822, -0.1979, -0.1922, -0.1886, -0.1886, -0.2036, -0.1737, -0.1882, -0.1824, -0.1729, -0.2123, -0.2023, -0.2267, -0.2478, -0.2403, -0.248, -0.2746, -0.2706, -0.258, -0.2611, -0.2611, -0.2564, -0.238, -0.2752, -0.3034, -0.2843, -0.2757, -0.2775, -0.257, -0.2104, -0.1783, -0.1516, -0.1747, -0.1679, -0.1973, -0.1795, -0.1894, -0.2216, -0.231, -0.2184, -0.2268, -0.2263, -0.218, -0.1897, -0.2096, -0.2096, -0.2131, -0.2115, -0.1751, -0.1275, -0.1269, -0.1018, -0.0702, 0.0, 0.0442, 0.0478, 0.0548, 0.0626, 0.1217, 0.0808, 0.0933, 0.1052, 0.074, 0.0416, 0.0445, 0.0885, 0.1112, 0.2097, 0.2204, 0.2482, 0.269, 0.234, 0.3061, 0.2781, 0.2068, 0.2811, 0.2431, 0.2756, 0.3459, 0.3449, 0.3741, 0.3443, 0.3189, 0.3737, 0.4555, 0.4627, 0.4747, 0.5061, 0.4887, 0.4871, 0.5597, 0.4489, 0.5783, 0.5838, 0.5811, 0.6833, 0.6023, 0.6275, 0.5747, 0.6403, 0.6079, 0.5185, 0.5014, 0.3382, 0.3781, 0.4881, 0.4921, 0.5302, 0.5302, 0.5715, 0.598, 0.5916, 0.5561, 0.5062, 0.5765, 0.6409, 0.6775, 0.7525, 0.7176, 0.6025, 0.5783, 0.5452, 0.4987, 0.6518, 0.7672, 0.8381, 0.836, 0.9052, 0.9052, 0.8926, 0.9571, 0.9455, 0.8975, 0.8975, 1.097, 1.0753, 1.2832, 1.2574, 1.1741, 1.2943, 1.2994, 1.248, 1.2162, 1.238, 1.4117, 1.4117, 1.4266, 1.5869, 1.6432, 1.657, 1.5868, 1.7274, 1.8939, 1.8973, 1.7582, 1.9106, 1.7886, 1.5224, 1.5456, 1.624, 1.5496, 1.4815, 1.7281, 1.7522, 1.7368, 1.7368, 1.6578, 1.7986, 1.7747, 1.8466, 1.7987, 1.779, 1.8521, 1.7628, 1.7415, 1.7435, 1.5242, 1.6644, 1.6397, 1.4619]}
{"unit_id": "quote:1956177797430566984", "ticker_idx": 0, "ts": "2025-08-15T02:15:05+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley confirmed that Rubin has not been delayed.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Author shares/endorses MS note confirming Rubin is on track with no delays; bullish read for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Morgan Stanley confirmed that Rubin has not been delayed.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": "MS on Rubin Timeline\n\n$nvda Rubin is on track: NVIDIA's first Rubin silicon should be out from the TSMC fab this October (first tape-out was in June). Meanwhile, our checks suggest NVIDIA's Rubin engineering sampling (ES) is on track for 4Q25, although the fine-tuning of the", "parent_author": null, "ret_m1d": 0.00870043536852072, "ret_signed_m1d": 0.00870043536852072, "alpha_spy_m1d": 0.0063537466714940205, "alpha_c_m1d": -0.01250706844600935, "bench_spy_m1d": 0.0023466886970267, "bench_c_m1d": 0.02120750381453007, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008644955017783396, "ret_signed_p1d": 0.008644955017783396, "alpha_spy_p1d": 0.008862586454974508, "alpha_c_p1d": 0.0045523079077973705, "bench_spy_p1d": -0.00021763143719111255, "bench_c_p1d": 0.004092647109986025, "ret_p1w": -0.013632520145879456, "ret_signed_p1w": -0.013632520145879456, "alpha_spy_p1w": -0.016538832538540604, "alpha_c_p1w": -0.006833979524653588, "bench_spy_p1w": 0.002906312392661148, "bench_c_p1w": -0.006798540621225868, "ret_p1m": -0.014907045886140446, "ret_signed_p1m": -0.014907045886140446, "alpha_spy_p1m": -0.042057883030574894, "alpha_c_p1m": -0.05387209882568955, "bench_spy_p1m": 0.027150837144434448, "bench_c_p1m": 0.038965052939549105, "ret_p3m": 0.07404233387567638, "ret_signed_p3m": 0.07404233387567638, "alpha_spy_p3m": 0.009025252916877502, "alpha_c_p3m": -0.12849441152245134, "bench_spy_p3m": 0.06501708095879888, "bench_c_p3m": 0.20253674539812772, "ret_p6m": 0.05326291275562611, "ret_signed_p6m": 0.05326291275562611, "alpha_spy_p6m": -0.03142278870589088, "alpha_c_p6m": -0.326460829940439, "bench_spy_p6m": 0.084685701461517, "bench_c_p6m": 0.3797237426960651, "price_path": [-0.2554, -0.2697, -0.2639, -0.2725, -0.2725, -0.2492, -0.253, -0.2287, -0.2512, -0.2387, -0.2175, -0.2136, -0.2243, -0.2147, -0.2096, -0.2023, -0.2085, -0.1965, -0.2132, -0.1982, -0.2013, -0.1938, -0.1938, -0.2028, -0.2011, -0.1804, -0.1449, -0.1409, -0.1258, -0.1245, -0.1505, -0.1286, -0.117, -0.117, -0.1231, -0.1133, -0.0974, -0.0906, -0.0861, -0.0908, -0.054, -0.0503, -0.0413, -0.0446, -0.0503, -0.0744, -0.0536, -0.0372, -0.0385, -0.0205, -0.0274, -0.0065, -0.0143, -0.0373, -0.0025, -0.0121, -0.0057, 0.0018, 0.0125, 0.0089, 0.015, 0.0063, 0.0087, 0.0, 0.0086, -0.0267, -0.028, -0.0303, -0.0136, -0.0035, 0.0073, 0.0064, -0.0016, -0.0347, -0.0347, -0.0536, -0.0545, -0.0487, -0.0744, -0.0673, -0.0537, -0.0173, -0.0181, -0.0145, -0.0149, -0.0308, -0.0563, -0.0233, -0.0209, 0.0176, -0.0111, -0.0192, -0.0152, -0.0125, 0.0078, 0.034, 0.0377, 0.0468, 0.0398, 0.0283, 0.0255, 0.0481, 0.0672, 0.0151, 0.0437, -0.0023, -0.0034, 0.0076, 0.0154, 0.0122, 0.004, -0.0009, 0.0095, 0.0323, 0.0612, 0.1141, 0.1474, 0.1244, 0.1222, 0.1465, 0.1011, 0.0819, 0.0423, 0.0427, 0.1031, 0.0705, 0.074, 0.0356, 0.0539, 0.0341, 0.0051, 0.0337, 0.0011, -0.0086, 0.0117, -0.0145, -0.001, -0.001, -0.0191, -0.0029, 0.0057, -0.0047, 0.0164, 0.011, 0.0284, 0.0252, 0.0186, 0.0028, -0.03, -0.0229, -0.015, -0.0526, -0.0349, 0.0031, 0.0181, 0.0487, 0.0453, 0.0453, 0.056, 0.0432, 0.0394, 0.0336, 0.0336, 0.0467, 0.0426, 0.0377, 0.0481, 0.0256, 0.0246, 0.025, 0.0298, 0.015, 0.0367, 0.0321, 0.0321, -0.0131, 0.016, 0.0244, 0.0401, 0.0335, 0.0448, 0.0615, 0.067, 0.0593, 0.0287, -0.0005, -0.0346, -0.0474, 0.0276, 0.0533]}
{"unit_id": "reply:1962517670127165528", "ticker_idx": 0, "ts": "2025-09-01T14:07:29+00:00", "symbol": "GEV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I have a feeling that GE Vernova will rise further.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Expects GE Vernova to continue rising.", "resolved_tickers": ["GEV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@zephyr_z9 I have a feeling that GE Vernova will rise further.", "reply_text": null, "parent_text": "Little to do with Data Centers", "quoted_text": null, "parent_author": "zephyr_z9", "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05430937649840151, "ret_signed_p1d": -0.05430937649840151, "alpha_spy_p1d": -0.04689908094252726, "alpha_c_p1d": -0.044770522369874555, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.009538854128526952, "ret_p1w": -0.02078398510483559, "ret_signed_p1w": -0.02078398510483559, "alpha_spy_p1w": -0.02664393655913122, "alpha_c_p1w": -0.015916024331698475, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": -0.004867960773137114, "ret_p1m": 0.003148744068362408, "ret_signed_p1m": 0.003148744068362408, "alpha_spy_p1m": -0.0324716688687996, "alpha_c_p1m": -0.01564730890352184, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.018796052971884247, "ret_p3m": -0.03752907528547622, "ret_signed_p3m": -0.03752907528547622, "alpha_spy_p3m": -0.09413614907992796, "alpha_c_p3m": -0.04687907736467434, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.009350002079198116, "ret_p6m": 0.43684801611767554, "ret_signed_p6m": 0.43684801611767554, "alpha_spy_p6m": 0.3651601315792352, "alpha_c_p6m": 0.26376825706124607, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.17307975905642947, "price_path": [-0.204, -0.2102, -0.2091, -0.2173, -0.2412, -0.2116, -0.2048, -0.2198, -0.2044, -0.2031, -0.2007, -0.2007, -0.2059, -0.1848, -0.167, -0.1803, -0.1735, -0.1526, -0.1371, -0.1749, -0.1764, -0.1569, -0.1569, -0.1353, -0.1357, -0.1263, -0.1205, -0.1208, -0.0949, -0.0874, -0.0849, -0.0702, -0.063, -0.0768, -0.1044, 0.0262, 0.0179, 0.0516, 0.0566, 0.0321, 0.0686, 0.0772, 0.071, 0.0812, 0.06, 0.0841, 0.0537, 0.0589, 0.0617, 0.0725, 0.0348, 0.0201, 0.0146, 0.0197, -0.0161, -0.0137, -0.0114, -0.0096, -0.0174, 0.0211, 0.0154, 0.0338, 0.0, 0.0, -0.0543, -0.0586, -0.0231, -0.0504, -0.0208, -0.0119, 0.0499, 0.0346, 0.0205, 0.0256, 0.0081, 0.003, -0.0032, 0.0183, 0.0512, 0.0333, 0.0261, -0.0089, -0.0127, -0.0172, 0.0031, -0.0111, -0.011, -0.0293, -0.0159, -0.0112, 0.0204, 0.0347, -0.0137, 0.0576, 0.0513, 0.0049, -0.0179, -0.0212, -0.0304, -0.0447, -0.0599, -0.0287, -0.0462, -0.0462, -0.0681, -0.0567, -0.0631, -0.045, -0.0513, -0.1057, -0.0865, -0.1021, -0.0613, -0.0537, -0.0598, -0.0609, -0.089, -0.0562, -0.0583, -0.0943, -0.0283, -0.0892, -0.0928, -0.0526, -0.0655, -0.0375, -0.0375, -0.0211, -0.0585, -0.0182, -0.0175, 0.0268, 0.0304, 0.015, 0.0205, 0.18, 0.1493, 0.0963, 0.112, 0.12, 0.0024, 0.0436, 0.0744, 0.0801, 0.0795, 0.0891, 0.0891, 0.0828, 0.0828, 0.0766, 0.0667, 0.0667, 0.1091, 0.112, 0.121, 0.0818, 0.0264, 0.0167, 0.0449, 0.065, 0.0521, 0.0489, 0.1132, 0.1132, 0.1186, 0.0909, 0.0807, 0.0743, 0.0878, 0.1314, 0.1622, 0.1717, 0.1864, 0.2331, 0.2744, 0.2188, 0.2046, 0.2729, 0.3091, 0.2916, 0.3453, 0.3337, 0.3101, 0.3101, 0.3379, 0.3353, 0.3632, 0.3562, 0.3584, 0.4368]}
{"unit_id": "thread:1965565561024471393", "ticker_idx": 0, "ts": "2025-09-09T23:58:42+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is likely to shift its HBM4 capacity to general-purpose DDR5 due to HBM4 speed issues... Micron's monopolistic position in CPU-targeted LPDDR5x and SoCamm (SoCAMM) is expected to weaken", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish MU: HBM4 speed issues cause delays vs Samsung/Hynix, capacity shift to DDR5, and monopoly in LPDDR5x/SoCAMM weakening.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Micron is likely to shift its HBM4 capacity to general-purpose DDR5 due to HBM4 speed issues\n\nMicron is projected to face delays in its CS schedule compared to Samsung Electronics and SK Hynix due to HBM4 speed issues for NVIDIA. Industry sources estimate that providing samples meeting NVIDIA’s speed conditions by 2025 will be difficult. While issues arose with HBM, Micron is expected to offset this through demand in server-oriented DRAM above 96GB and QLC eSSD.\n\nDue to insufficient supply of server DDR5, Micron is reported to shift part of its 2026 HBM sales plan toward DDR5. Considering NVIDIA-targeted HBM4 speed issues, Micron is likely to allocate additional capacity away from HBM and into DRAM. It is interpreted as a strategic decision, taking into account structural changes in the HBM market landscape, additional Capex burdens from HBM, etc.\n\nHowever, Micron’s monopolistic position in CPU-targeted LPDDR5x and SoCamm (SoCAMM) is expected to weaken. This is because SoCAMM, initially expected to be adopted in GB300, has been canceled, and Samsung Electronics and SK Hynix are both preparing to supply alternatives.\n\n$MU\n\n---\n\nSource: KIS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027950290206556927, "ret_signed_m1d": 0.027950290206556927, "alpha_spy_m1d": 0.025643702157598147, "alpha_c_m1d": 0.02361862796560732, "bench_spy_m1d": -0.0023065880489587798, "bench_c_m1d": -0.0043316622409496075, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03519664081729679, "ret_signed_p1d": -0.03519664081729679, "alpha_spy_p1d": -0.03230578653829208, "alpha_c_p1d": -0.02505598720091906, "bench_spy_p1d": 0.0028908542790047065, "bench_c_p1d": 0.010140653616377726, "ret_p1w": 0.17435658300196777, "ret_signed_p1w": -0.17435658300196777, "alpha_spy_p1w": -0.1594871825473323, "alpha_c_p1w": -0.1427928066765063, "bench_spy_p1w": 0.014869400454635473, "bench_c_p1w": 0.03156377632546148, "ret_p1m": 0.45417812020634907, "ret_signed_p1m": -0.45417812020634907, "alpha_spy_p1m": -0.4162802254861453, "alpha_c_p1m": -0.2921284154614521, "bench_spy_p1m": 0.03789789472020377, "bench_c_p1m": 0.162049704744897, "ret_p3m": 0.7551650530019045, "ret_signed_p3m": -0.7551650530019045, "alpha_spy_p3m": -0.6978694815847044, "alpha_c_p3m": -0.5314988405738577, "bench_spy_p3m": 0.0572955714172001, "bench_c_p3m": 0.22366621242804685, "ret_p6m": 1.966451561583359, "ret_signed_p6m": -1.966451561583359, "alpha_spy_p6m": -1.9068978622354222, "alpha_c_p6m": -1.6221654136615136, "bench_spy_p6m": 0.05955369934793686, "bench_c_p6m": 0.34428614792184553, "price_path": [-0.1417, -0.146, -0.1147, -0.111, -0.1001, -0.1001, -0.0869, -0.0982, -0.0551, -0.06, -0.0692, -0.0784, -0.0895, -0.1069, -0.1007, -0.0966, -0.0966, -0.1133, -0.08, -0.0961, -0.0897, -0.0792, -0.123, -0.1119, -0.1391, -0.1625, -0.1542, -0.1627, -0.1924, -0.1879, -0.1738, -0.1773, -0.1774, -0.1721, -0.1516, -0.193, -0.2245, -0.2031, -0.1936, -0.1957, -0.1728, -0.1209, -0.0852, -0.0554, -0.0811, -0.0736, -0.1063, -0.0864, -0.0975, -0.1333, -0.1438, -0.1298, -0.1392, -0.1386, -0.1293, -0.0979, -0.12, -0.12, -0.1239, -0.1222, -0.0816, -0.0286, -0.028, 0.0, 0.0352, 0.1134, 0.1626, 0.1666, 0.1744, 0.183, 0.2488, 0.2033, 0.2172, 0.2305, 0.1957, 0.1596, 0.1629, 0.2119, 0.2372, 0.3469, 0.3587, 0.3897, 0.4129, 0.3739, 0.4542, 0.423, 0.3436, 0.4263, 0.384, 0.4201, 0.4985, 0.4974, 0.5299, 0.4967, 0.4685, 0.5294, 0.6205, 0.6285, 0.6419, 0.6768, 0.6574, 0.6556, 0.7365, 0.6132, 0.7572, 0.7634, 0.7603, 0.8741, 0.7839, 0.812, 0.7532, 0.8263, 0.7902, 0.6906, 0.6716, 0.4899, 0.5343, 0.6568, 0.6613, 0.7037, 0.7037, 0.7497, 0.7791, 0.772, 0.7325, 0.677, 0.7552, 0.8269, 0.8676, 0.9512, 0.9123, 0.7842, 0.7572, 0.7203, 0.6686, 0.839, 0.9675, 1.0465, 1.0441, 1.1211, 1.1211, 1.1071, 1.1789, 1.166, 1.1126, 1.1126, 1.3347, 1.3105, 1.542, 1.5133, 1.4206, 1.5543, 1.5601, 1.5028, 1.4674, 1.4917, 1.685, 1.685, 1.7017, 1.8801, 1.9428, 1.9582, 1.88, 2.0365, 2.2219, 2.2257, 2.0709, 2.2405, 2.1046, 1.8083, 1.8341, 1.9214, 1.8386, 1.7628, 2.0373, 2.0642, 2.0471, 2.0471, 1.9591, 2.1158, 2.0892, 2.1693, 2.116, 2.0941, 2.1754, 2.0759, 2.0523, 2.0545, 1.8103, 1.9665]}
{"unit_id": "orig:1945980154133852223", "ticker_idx": 0, "ts": "2025-07-17T22:53:17+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain a positive view on Samsung Electronics and SK Hynix, and regard any stock price corrections as a buying opportunity", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA reiterates Buy on Samsung Electronics and SK Hynix; dips are buying opportunities on conservative CapEx, HBM demand visibility, and SK Hynix's HBM dominance.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250718_Korea Memory Sector Outlook: Read Across from TSMC's 2Q25 Call - CLSA\n\n(1) TSMC maintains an optimistic stance on demand from AI and HPC customers.\n\n(2) Furthermore, the company anticipates a moderate recovery in the non-AI market in 2025.\n\n(3) Notably, TSMC's remark that \"there is no change in customers' purchasing behavior\" is a positive signal that could boost investor sentiment regarding semiconductor demand.\n\n(4) However, the uncertainty surrounding US tariff negotiations and their direct and indirect impact on IT demand remains a significant uncertain factor.\n\n(5) For these reasons, TSMC has chosen to maintain its annual CapEx guidance unchanged.\n\n(6) Accordingly, memory companies are likely to maintain a conservative stance on CapEx, which could be positive in reducing oversupply concerns in 2H25/2H26.\n\n(7) Meanwhile, ASML withdrew its previous guidance for revenue growth in 2026 just one day prior.\n\n(8) This is due to customers being unable to finalize their investment plans because of tariff and geopolitical issues.\n\n(9) Nevertheless, TSMC's robust short-term demand comments and conservative CapEx plans are judged to be positive for memory supply and demand.\n\n(10) Hyperscaler CapEx forecasts are being revised upwards, which is increasing HBM demand visibility.\n\n(11) However, ASML's conservative stance and TSMC's conservative CapEx execution suggest that memory companies are likely to maintain a cautious CapEx approach.\n\n(12) Especially in HBM, performance and power efficiency are key competitive factors, and SK Hynix is expected to maintain its lead in these areas.\n\n(13) Furthermore, customization demand for HBM4E is expected to fully materialize from late 2026, which will be positive for strengthening SK Hynix's dominance.\n\n(14) Overall, we maintain a positive view on Samsung Electronics and SK Hynix, and regard any stock price corrections as a buying opportunity.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029985112110534562, "ret_signed_m1d": -0.029985112110534562, "alpha_spy_m1d": -0.023902644946360296, "alpha_c_m1d": -0.020983481692528505, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.009001630418006057, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005996856566729658, "ret_signed_p1d": 0.005996856566729658, "alpha_spy_p1d": 0.0067292694688401156, "alpha_c_p1d": 0.006647803097681093, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.0006509465309514351, "ret_p1w": -0.010494854396156716, "ret_signed_p1w": -0.010494854396156716, "alpha_spy_p1w": -0.0206534693722904, "alpha_c_p1w": -0.011759188336187165, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": 0.0012643339400304487, "ret_p1m": 0.07346315149621097, "ret_signed_p1m": 0.07346315149621097, "alpha_spy_p1m": 0.0489423634853976, "alpha_c_p1m": 0.055689353420388166, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.0177737980758228, "ret_p3m": 0.3794403778943689, "ret_signed_p3m": 0.3794403778943689, "alpha_spy_p3m": 0.3220778346839188, "alpha_c_p3m": 0.29971985062533135, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.07972052726903756, "ret_p6m": 1.1034312041965202, "ret_signed_p6m": 1.1034312041965202, "alpha_spy_p6m": 0.9919560591424046, "alpha_c_p6m": 0.9806256692495765, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.12280553494694368, "price_path": [-0.1745, -0.1804, -0.17, -0.17, -0.17, -0.1685, -0.1685, -0.173, -0.173, -0.1909, -0.1909, -0.1909, -0.1864, -0.1864, -0.1834, -0.1417, -0.1521, -0.1447, -0.1462, -0.1536, -0.1685, -0.167, -0.17, -0.1849, -0.1924, -0.1849, -0.1968, -0.167, -0.164, -0.1626, -0.1536, -0.1536, -0.1387, -0.1193, -0.1193, -0.1089, -0.1179, -0.1074, -0.1134, -0.1313, -0.1477, -0.1342, -0.1089, -0.1179, -0.1134, -0.1357, -0.0985, -0.0866, -0.103, -0.0885, -0.1034, -0.0975, -0.0885, -0.0435, -0.051, -0.075, -0.0795, -0.0945, -0.0855, -0.0615, -0.063, -0.045, -0.03, 0.0, 0.006, 0.0165, -0.0105, -0.0045, -0.0105, -0.012, 0.0555, 0.0585, 0.0885, 0.0705, 0.033, 0.045, 0.048, 0.0315, 0.057, 0.0765, 0.0645, 0.066, 0.078, 0.0735, 0.0735, 0.0495, 0.0495, 0.057, 0.0585, 0.0705, 0.072, 0.054, 0.0585, 0.0435, 0.045, 0.0135, 0.036, 0.0465, 0.051, 0.042, 0.051, 0.072, 0.0885, 0.1004, 0.1304, 0.1469, 0.1904, 0.1724, 0.2039, 0.2039, 0.2519, 0.2699, 0.2804, 0.2909, 0.2489, 0.268, 0.2635, 0.2951, 0.3516, 0.3516, 0.3516, 0.3516, 0.3516, 0.3516, 0.4216, 0.405, 0.3794, 0.4306, 0.4713, 0.4743, 0.4773, 0.4683, 0.4849, 0.4532, 0.4879, 0.5361, 0.4984, 0.5135, 0.5677, 0.6189, 0.6731, 0.5797, 0.515, 0.4939, 0.4743, 0.515, 0.5586, 0.5526, 0.5481, 0.4638, 0.515, 0.4728, 0.4532, 0.515, 0.4276, 0.4562, 0.4954, 0.5481, 0.5586, 0.5135, 0.518, 0.5571, 0.5737, 0.5827, 0.6324, 0.649, 0.6324, 0.6264, 0.6159, 0.64, 0.5782, 0.5481, 0.6249, 0.6204, 0.6008, 0.6641, 0.6791, 0.6731, 0.6731, 0.7619, 0.8083, 0.8144, 0.8144, 0.8144, 0.9445, 1.0898, 1.1019, 1.1337, 1.1004, 1.1034]}
{"unit_id": "orig:1945980154133852223", "ticker_idx": 1, "ts": "2025-07-17T22:53:17+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is expected to maintain its lead in these areas... we maintain a positive view on Samsung Electronics and SK Hynix, and regard any stock price corrections as a buying opportunity", "tweet_type": "analysis", "timeline": "quarters", "summary": "CLSA reiterates Buy on Samsung Electronics and SK Hynix; dips are buying opportunities on conservative CapEx, HBM demand visibility, and SK Hynix's HBM dominance.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250718_Korea Memory Sector Outlook: Read Across from TSMC's 2Q25 Call - CLSA\n\n(1) TSMC maintains an optimistic stance on demand from AI and HPC customers.\n\n(2) Furthermore, the company anticipates a moderate recovery in the non-AI market in 2025.\n\n(3) Notably, TSMC's remark that \"there is no change in customers' purchasing behavior\" is a positive signal that could boost investor sentiment regarding semiconductor demand.\n\n(4) However, the uncertainty surrounding US tariff negotiations and their direct and indirect impact on IT demand remains a significant uncertain factor.\n\n(5) For these reasons, TSMC has chosen to maintain its annual CapEx guidance unchanged.\n\n(6) Accordingly, memory companies are likely to maintain a conservative stance on CapEx, which could be positive in reducing oversupply concerns in 2H25/2H26.\n\n(7) Meanwhile, ASML withdrew its previous guidance for revenue growth in 2026 just one day prior.\n\n(8) This is due to customers being unable to finalize their investment plans because of tariff and geopolitical issues.\n\n(9) Nevertheless, TSMC's robust short-term demand comments and conservative CapEx plans are judged to be positive for memory supply and demand.\n\n(10) Hyperscaler CapEx forecasts are being revised upwards, which is increasing HBM demand visibility.\n\n(11) However, ASML's conservative stance and TSMC's conservative CapEx execution suggest that memory companies are likely to maintain a cautious CapEx approach.\n\n(12) Especially in HBM, performance and power efficiency are key competitive factors, and SK Hynix is expected to maintain its lead in these areas.\n\n(13) Furthermore, customization demand for HBM4E is expected to fully materialize from late 2026, which will be positive for strengthening SK Hynix's dominance.\n\n(14) Overall, we maintain a positive view on Samsung Electronics and SK Hynix, and regard any stock price corrections as a buying opportunity.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09833034315341949, "ret_signed_m1d": 0.09833034315341949, "alpha_spy_m1d": 0.10441281031759375, "alpha_c_m1d": 0.10666050899695856, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0018553554029104857, "ret_signed_p1d": -0.0018553554029104857, "alpha_spy_p1d": -0.0011229425008000282, "alpha_c_p1d": 0.0029438376375250375, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.010158614976133684, "alpha_c_p1w": 0.013403676442963919, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": 0.025973927680496, "ret_signed_p1m": 0.025973927680496, "alpha_spy_p1m": 0.0014531396696826349, "alpha_c_p1m": 0.012467303446648925, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.5291072123884737, "ret_signed_p3m": 0.5291072123884737, "alpha_spy_p3m": 0.4717446691780236, "alpha_c_p3m": 0.3858483827526906, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 1.7666355538912395, "ret_signed_p6m": 1.7666355538912395, "alpha_spy_p6m": 1.655160408837124, "alpha_c_p6m": 1.4282134234533452, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.3459, -0.3563, -0.3296, -0.3396, -0.317, -0.3259, -0.3303, -0.3426, -0.3426, -0.3111, -0.3111, -0.3111, -0.2933, -0.2952, -0.2959, -0.2777, -0.2648, -0.237, -0.2574, -0.2426, -0.2614, -0.2518, -0.2574, -0.2707, -0.2592, -0.2481, -0.25, -0.2296, -0.2134, -0.2412, -0.2301, -0.2301, -0.1929, -0.167, -0.167, -0.1503, -0.1447, -0.1095, -0.1262, -0.1262, -0.0798, -0.0761, -0.0853, -0.0872, -0.0464, -0.0371, 0.0334, 0.0612, 0.0872, 0.0538, 0.0835, 0.0594, 0.0353, 0.0334, 0.0037, 0.0056, 0.0464, 0.0427, 0.102, 0.0928, 0.1132, 0.1076, 0.0983, 0.0, -0.0019, 0.0111, -0.0037, -0.0019, 0.0, -0.013, -0.0278, -0.026, -0.0223, 0.0148, -0.0427, -0.0427, -0.0223, -0.0408, -0.0278, -0.0482, -0.0093, -0.0019, 0.0315, 0.026, 0.026, -0.0074, -0.0241, -0.0519, -0.0909, -0.0686, -0.0371, -0.0297, -0.0353, -0.0023, -0.0004, -0.0487, -0.032, -0.0246, -0.0134, 0.0163, 0.0293, 0.0702, 0.1296, 0.1408, 0.2207, 0.23, 0.2931, 0.2393, 0.321, 0.321, 0.3043, 0.3415, 0.3284, 0.3247, 0.2504, 0.2969, 0.2913, 0.3377, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.4697, 0.5904, 0.5421, 0.5291, 0.57, 0.6815, 0.7298, 0.8041, 0.7799, 0.7892, 0.7781, 0.8951, 0.988, 0.936, 1.0735, 1.1107, 1.0772, 1.3039, 1.1775, 1.1515, 1.2035, 1.1552, 1.2519, 1.3002, 1.2927, 1.2742, 1.0809, 1.2519, 1.1181, 1.0884, 1.1218, 0.936, 0.9323, 0.9286, 0.9471, 1.0229, 0.9709, 1.0006, 1.075, 1.0527, 1.0155, 1.0229, 1.1456, 1.1047, 1.1828, 1.101, 1.1233, 1.0601, 0.9709, 1.0489, 1.0527, 1.0341, 1.1568, 1.1717, 1.1865, 1.1865, 1.2274, 1.3799, 1.4208, 1.4208, 1.4208, 1.5175, 1.5881, 1.6997, 1.7592, 1.8113, 1.7666]}
{"unit_id": "thread:1947554940626309599", "ticker_idx": 0, "ts": "2025-07-22T07:10:56+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB200 servers are expected to begin smooth shipments starting in the second quarter of 2025, coupled with strong shipments of ASIC servers, signaling that US-based cloud service providers are poised to deliver impressive cloud performance", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply chain analysis signals accelerating CSP data center deployments via dual Nvidia Blackwell + ASIC engines; TSMC raised full-year guidance; ODMs Quanta, Wistron, Wiwynn all bullish on H2 server growth.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia and ASIC dual engines accelerate CSP data center deployments\n\nThe GB200 servers are expected to begin smooth shipments starting in the second quarter of 2025, coupled with strong shipments of ASIC servers, signaling that US-based cloud service providers (CSPs) are poised to deliver impressive cloud performance. According to the supply chain, despite tariff-related adjustments, clients have been actively stocking up, and current orders show no signs of easing. This indicates that the AI boom will continue heating up throughout the second half of 2025.\n\nCloud demand remains robust despite AWS layoffs\nUS CSPs are set to release their second-quarter financial reports soon. Recent news about layoffs at AWS has drawn attention to the future momentum of cloud business growth. Supply chain analysis suggests that, based on shipment trends in the second quarter, CSP demand remains robust and may be reflected positively in their upcoming quarterly earnings.\n\nSupply constraints ease as dual-system approach gains traction\nThe server supply chain notes that although GB200 shipments began at the end of 2024, they have been facing challenges, which also slowed down CSP cloud data center deployment progress. Amazon CEO Andy Jassy recently mentioned that if supply capacity improves, cloud business growth could be even stronger, citing difficulties in sourcing certain components such as motherboards.\n\nAccording to the server supply chain, all major CSPs are simultaneously adopting Nvidia's Blackwell chip servers alongside their own developed ASIC servers. ASIC server shipments have been smoother in comparison. With these two major systems powering forward together, CSP data center construction is expected to accelerate.\n\nTariff concerns fail to dampen growth outlook\nThe supply chain further explains that in response to reciprocal tariffs imposed by the Trump administration, clients aggressively stocked inventory in the first half of the year. While there were concerns about a potential slowdown in purchasing momentum during the second half, TSMC's recent earnings call reassured stakeholders. Despite tariff uncertainties, TSMC raised its full-year revenue growth forecast, driven primarily by strong demand from AI and other high-performance computing (HPC) applications.\n\nTSMC Chairman C.C. Wei stressed that client behavior remains unchanged for the second half of the year. Server original design manufacturers (ODMs) hold a positive outlook on orders for the latter half, expecting continued growth. The only variable is the generational transition of Nvidia GPUs, which may cause short-term fluctuations; however, server shipments are projected to grow both year-over-year and compared to the first half of the year.\n\nServer momentum drives ODM optimism\nQuanta and Wistron both assess that servers will remain the main growth driver in the second half. Wistron's President and CEO Jeff Lin noted that tariffs inevitably impact economic conditions, especially mature products like notebooks and smartphones. However, AI servers are still in the development phase, with clear customer demand, representing a rigid demand that is less affected by tariffs.\n\nWiwynn also reported strong shipments of both the GB series and ASIC servers in the second half. Wiwynn's President and CEO William Lin stated that benefiting from simultaneous shipments of Nvidia's GB series servers and ASIC servers, CSP cloud data center deployments will proceed more smoothly in the coming months.\n\n---\n\nhttps://t.co/jFZJytBZCV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02604320280392458, "ret_signed_m1d": 0.02604320280392458, "alpha_spy_m1d": 0.026186303486413043, "alpha_c_m1d": 0.008031506169791092, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.01801169663413349, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022450948681822958, "ret_signed_p1d": 0.022450948681822958, "alpha_spy_p1d": 0.013943407559708021, "alpha_c_p1d": 0.018105742857921214, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.004345205823901743, "ret_p1w": 0.05076932494933595, "ret_signed_p1w": 0.05076932494933595, "alpha_spy_p1w": 0.04059224323816113, "alpha_c_p1w": 0.024382599700092156, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.026386725249243792, "ret_p1m": 0.05011064756694128, "ret_signed_p1m": 0.05011064756694128, "alpha_spy_p1m": 0.035401522216518266, "alpha_c_p1m": 0.037530511849255044, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.012580135717686236, "ret_p3m": 0.09699049582170671, "ret_signed_p3m": 0.09699049582170671, "alpha_spy_p3m": 0.0375623093135562, "alpha_c_p3m": -0.10436204684135753, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.20135254266306424, "ret_p6m": 0.09657248375006633, "ret_signed_p6m": 0.09657248375006633, "alpha_spy_p6m": -0.007519951068894137, "alpha_c_p6m": -0.2685268222141701, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.36509930596423645, "price_path": [-0.3629, -0.3354, -0.3491, -0.3473, -0.3479, -0.3318, -0.3145, -0.3186, -0.3203, -0.2992, -0.2974, -0.3017, -0.2637, -0.2222, -0.1898, -0.1928, -0.1894, -0.1884, -0.1955, -0.211, -0.2048, -0.214, -0.214, -0.1888, -0.193, -0.1667, -0.191, -0.1776, -0.1546, -0.1504, -0.1619, -0.1516, -0.1461, -0.1382, -0.1449, -0.1319, -0.15, -0.1337, -0.1372, -0.129, -0.129, -0.1388, -0.1369, -0.1145, -0.0762, -0.0719, -0.0556, -0.0541, -0.0822, -0.0586, -0.046, -0.046, -0.0526, -0.0421, -0.0248, -0.0175, -0.0126, -0.0177, 0.022, 0.026, 0.0357, 0.0322, 0.026, 0.0, 0.0225, 0.0402, 0.0387, 0.0582, 0.0508, 0.0733, 0.0649, 0.0401, 0.0777, 0.0672, 0.0742, 0.0823, 0.0938, 0.09, 0.0966, 0.0872, 0.0897, 0.0803, 0.0897, 0.0515, 0.0501, 0.0476, 0.0656, 0.0765, 0.0882, 0.0872, 0.0787, 0.0428, 0.0428, 0.0225, 0.0215, 0.0277, -0.0001, 0.0077, 0.0223, 0.0617, 0.0608, 0.0647, 0.0642, 0.0471, 0.0196, 0.0552, 0.0578, 0.0993, 0.0683, 0.0596, 0.0639, 0.0669, 0.0888, 0.1171, 0.1211, 0.1309, 0.1233, 0.1109, 0.1079, 0.1323, 0.153, 0.0966, 0.1275, 0.0779, 0.0767, 0.0885, 0.097, 0.0935, 0.0847, 0.0794, 0.0906, 0.1152, 0.1465, 0.2036, 0.2396, 0.2148, 0.2124, 0.2386, 0.1896, 0.1688, 0.1261, 0.1265, 0.1918, 0.1565, 0.1603, 0.1188, 0.1386, 0.1172, 0.0859, 0.1167, 0.0815, 0.071, 0.093, 0.0647, 0.0793, 0.0793, 0.0597, 0.0772, 0.0865, 0.0753, 0.098, 0.0922, 0.111, 0.1075, 0.1004, 0.0833, 0.048, 0.0556, 0.0641, 0.0235, 0.0427, 0.0837, 0.0999, 0.1329, 0.1293, 0.1293, 0.1408, 0.127, 0.1229, 0.1167, 0.1167, 0.1308, 0.1264, 0.1211, 0.1323, 0.1079, 0.1069, 0.1074, 0.1126, 0.0966]}
{"unit_id": "thread:1947554940626309599", "ticker_idx": 1, "ts": "2025-07-22T07:10:56+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC raised its full-year revenue growth forecast, driven primarily by strong demand from AI and other high-performance computing (HPC) applications", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply chain analysis signals accelerating CSP data center deployments via dual Nvidia Blackwell + ASIC engines; TSMC raised full-year guidance; ODMs Quanta, Wistron, Wiwynn all bullish on H2 server growth.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia and ASIC dual engines accelerate CSP data center deployments\n\nThe GB200 servers are expected to begin smooth shipments starting in the second quarter of 2025, coupled with strong shipments of ASIC servers, signaling that US-based cloud service providers (CSPs) are poised to deliver impressive cloud performance. According to the supply chain, despite tariff-related adjustments, clients have been actively stocking up, and current orders show no signs of easing. This indicates that the AI boom will continue heating up throughout the second half of 2025.\n\nCloud demand remains robust despite AWS layoffs\nUS CSPs are set to release their second-quarter financial reports soon. Recent news about layoffs at AWS has drawn attention to the future momentum of cloud business growth. Supply chain analysis suggests that, based on shipment trends in the second quarter, CSP demand remains robust and may be reflected positively in their upcoming quarterly earnings.\n\nSupply constraints ease as dual-system approach gains traction\nThe server supply chain notes that although GB200 shipments began at the end of 2024, they have been facing challenges, which also slowed down CSP cloud data center deployment progress. Amazon CEO Andy Jassy recently mentioned that if supply capacity improves, cloud business growth could be even stronger, citing difficulties in sourcing certain components such as motherboards.\n\nAccording to the server supply chain, all major CSPs are simultaneously adopting Nvidia's Blackwell chip servers alongside their own developed ASIC servers. ASIC server shipments have been smoother in comparison. With these two major systems powering forward together, CSP data center construction is expected to accelerate.\n\nTariff concerns fail to dampen growth outlook\nThe supply chain further explains that in response to reciprocal tariffs imposed by the Trump administration, clients aggressively stocked inventory in the first half of the year. While there were concerns about a potential slowdown in purchasing momentum during the second half, TSMC's recent earnings call reassured stakeholders. Despite tariff uncertainties, TSMC raised its full-year revenue growth forecast, driven primarily by strong demand from AI and other high-performance computing (HPC) applications.\n\nTSMC Chairman C.C. Wei stressed that client behavior remains unchanged for the second half of the year. Server original design manufacturers (ODMs) hold a positive outlook on orders for the latter half, expecting continued growth. The only variable is the generational transition of Nvidia GPUs, which may cause short-term fluctuations; however, server shipments are projected to grow both year-over-year and compared to the first half of the year.\n\nServer momentum drives ODM optimism\nQuanta and Wistron both assess that servers will remain the main growth driver in the second half. Wistron's President and CEO Jeff Lin noted that tariffs inevitably impact economic conditions, especially mature products like notebooks and smartphones. However, AI servers are still in the development phase, with clear customer demand, representing a rigid demand that is less affected by tariffs.\n\nWiwynn also reported strong shipments of both the GB series and ASIC servers in the second half. Wiwynn's President and CEO William Lin stated that benefiting from simultaneous shipments of Nvidia's GB series servers and ASIC servers, CSP cloud data center deployments will proceed more smoothly in the coming months.\n\n---\n\nhttps://t.co/jFZJytBZCV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01769920682897208, "ret_signed_m1d": 0.01769920682897208, "alpha_spy_m1d": 0.017842307511460542, "alpha_c_m1d": -0.0003124898051614089, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.01801169663413349, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013274432415607551, "ret_signed_p1d": 0.013274432415607551, "alpha_spy_p1d": 0.004766891293492614, "alpha_c_p1d": 0.008929226591705808, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.004345205823901743, "ret_p1w": 0.0044247744133645295, "ret_signed_p1w": 0.0044247744133645295, "alpha_spy_p1w": -0.0057523072978102885, "alpha_c_p1w": -0.021961950835879263, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.026386725249243792, "ret_p1m": 0.0044247744133645295, "ret_signed_p1m": 0.0044247744133645295, "alpha_spy_p1m": -0.010284350937058484, "alpha_c_p1m": -0.008155361304321707, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.012580135717686236, "ret_p3m": 0.28831866760455704, "ret_signed_p3m": 0.28831866760455704, "alpha_spy_p3m": 0.22889048109640653, "alpha_c_p3m": 0.0869661249414928, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.20135254266306424, "ret_p6m": 0.5243920479612401, "ret_signed_p6m": 0.5243920479612401, "alpha_spy_p6m": 0.4202996131422796, "alpha_c_p6m": 0.1592927419970036, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.36509930596423645, "price_path": [-0.2386, -0.2175, -0.2095, -0.2051, -0.1999, -0.1999, -0.1628, -0.1734, -0.1893, -0.1822, -0.191, -0.1637, -0.1567, -0.1461, -0.1197, -0.125, -0.1205, -0.1329, -0.1382, -0.1276, -0.1346, -0.1329, -0.1399, -0.1496, -0.1479, -0.1479, -0.1479, -0.1664, -0.1628, -0.1276, -0.1205, -0.1232, -0.1144, -0.0791, -0.0615, -0.0752, -0.0885, -0.0929, -0.0752, -0.0664, -0.0841, -0.0664, -0.0973, -0.0708, -0.0531, -0.0487, -0.0442, -0.0619, -0.0398, -0.0398, -0.0354, -0.0398, -0.0442, -0.0442, -0.0354, -0.0265, -0.0265, -0.031, -0.0177, 0.0, 0.0, 0.0221, 0.0177, 0.0, 0.0133, 0.0133, 0.0133, 0.0133, 0.0044, 0.0221, 0.0265, 0.0265, 0.0044, 0.0177, -0.0044, 0.0442, 0.0398, 0.0442, 0.0442, 0.0619, 0.0398, 0.0442, 0.0442, 0.0487, 0.0044, 0.0177, 0.0044, 0.0354, 0.0398, 0.0531, 0.0265, 0.0265, 0.031, 0.0265, 0.0265, 0.0265, 0.0442, 0.0442, 0.0619, 0.0841, 0.0973, 0.115, 0.1106, 0.1373, 0.1239, 0.1417, 0.1239, 0.1506, 0.1906, 0.1906, 0.1728, 0.155, 0.155, 0.1595, 0.1773, 0.2128, 0.2439, 0.2439, 0.275, 0.2572, 0.2794, 0.2794, 0.2572, 0.2661, 0.3016, 0.3194, 0.2883, 0.315, 0.315, 0.2972, 0.2883, 0.2883, 0.315, 0.3105, 0.3372, 0.3372, 0.3327, 0.3416, 0.3372, 0.2972, 0.3016, 0.2972, 0.3105, 0.3016, 0.3105, 0.2972, 0.2705, 0.2839, 0.2483, 0.2395, 0.2928, 0.2306, 0.2217, 0.2572, 0.2794, 0.275, 0.2794, 0.2528, 0.2705, 0.2883, 0.2839, 0.2972, 0.3283, 0.315, 0.3372, 0.3104, 0.3194, 0.2926, 0.2792, 0.2748, 0.2748, 0.2748, 0.306, 0.3283, 0.3327, 0.3327, 0.3461, 0.3639, 0.355, 0.3818, 0.3818, 0.413, 0.4887, 0.5199, 0.4932, 0.5021, 0.4976, 0.5066, 0.5244, 0.5244]}
{"unit_id": "thread:1947554940626309599", "ticker_idx": 2, "ts": "2025-07-22T07:10:56+00:00", "symbol": "Quanta", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Quanta and Wistron both assess that servers will remain the main growth driver in the second half", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply chain analysis signals accelerating CSP data center deployments via dual Nvidia Blackwell + ASIC engines; TSMC raised full-year guidance; ODMs Quanta, Wistron, Wiwynn all bullish on H2 server growth.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta in server/ODM context = Quanta Computer, TWSE 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia and ASIC dual engines accelerate CSP data center deployments\n\nThe GB200 servers are expected to begin smooth shipments starting in the second quarter of 2025, coupled with strong shipments of ASIC servers, signaling that US-based cloud service providers (CSPs) are poised to deliver impressive cloud performance. According to the supply chain, despite tariff-related adjustments, clients have been actively stocking up, and current orders show no signs of easing. This indicates that the AI boom will continue heating up throughout the second half of 2025.\n\nCloud demand remains robust despite AWS layoffs\nUS CSPs are set to release their second-quarter financial reports soon. Recent news about layoffs at AWS has drawn attention to the future momentum of cloud business growth. Supply chain analysis suggests that, based on shipment trends in the second quarter, CSP demand remains robust and may be reflected positively in their upcoming quarterly earnings.\n\nSupply constraints ease as dual-system approach gains traction\nThe server supply chain notes that although GB200 shipments began at the end of 2024, they have been facing challenges, which also slowed down CSP cloud data center deployment progress. Amazon CEO Andy Jassy recently mentioned that if supply capacity improves, cloud business growth could be even stronger, citing difficulties in sourcing certain components such as motherboards.\n\nAccording to the server supply chain, all major CSPs are simultaneously adopting Nvidia's Blackwell chip servers alongside their own developed ASIC servers. ASIC server shipments have been smoother in comparison. With these two major systems powering forward together, CSP data center construction is expected to accelerate.\n\nTariff concerns fail to dampen growth outlook\nThe supply chain further explains that in response to reciprocal tariffs imposed by the Trump administration, clients aggressively stocked inventory in the first half of the year. While there were concerns about a potential slowdown in purchasing momentum during the second half, TSMC's recent earnings call reassured stakeholders. Despite tariff uncertainties, TSMC raised its full-year revenue growth forecast, driven primarily by strong demand from AI and other high-performance computing (HPC) applications.\n\nTSMC Chairman C.C. Wei stressed that client behavior remains unchanged for the second half of the year. Server original design manufacturers (ODMs) hold a positive outlook on orders for the latter half, expecting continued growth. The only variable is the generational transition of Nvidia GPUs, which may cause short-term fluctuations; however, server shipments are projected to grow both year-over-year and compared to the first half of the year.\n\nServer momentum drives ODM optimism\nQuanta and Wistron both assess that servers will remain the main growth driver in the second half. Wistron's President and CEO Jeff Lin noted that tariffs inevitably impact economic conditions, especially mature products like notebooks and smartphones. However, AI servers are still in the development phase, with clear customer demand, representing a rigid demand that is less affected by tariffs.\n\nWiwynn also reported strong shipments of both the GB series and ASIC servers in the second half. Wiwynn's President and CEO William Lin stated that benefiting from simultaneous shipments of Nvidia's GB series servers and ASIC servers, CSP cloud data center deployments will proceed more smoothly in the coming months.\n\n---\n\nhttps://t.co/jFZJytBZCV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04190476190476189, "ret_signed_m1d": 0.04190476190476189, "alpha_spy_m1d": 0.04204786258725035, "alpha_c_m1d": 0.03255434909335375, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.009350412811408138, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007619047619047636, "ret_signed_p1d": 0.007619047619047636, "alpha_spy_p1d": -0.0008884935030673002, "alpha_c_p1d": 0.001900565393569309, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.005718482225478327, "ret_p1w": 0.01904761904761898, "ret_signed_p1w": 0.01904761904761898, "alpha_spy_p1w": 0.008870537336444162, "alpha_c_p1w": -0.0013533246836809898, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.02040094373129997, "ret_p1m": -0.011428571428571455, "ret_signed_p1m": -0.011428571428571455, "alpha_spy_p1m": -0.026137696778994468, "alpha_c_p1m": -0.015446817152940162, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.0040182457243687075, "ret_p3m": 0.09904761904761905, "ret_signed_p3m": 0.09904761904761905, "alpha_spy_p3m": 0.039619432539468535, "alpha_c_p3m": -0.003600208352980916, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.10264782740059997, "ret_p6m": 0.08952380952380956, "ret_signed_p6m": 0.08952380952380956, "alpha_spy_p6m": -0.014568625295150905, "alpha_c_p6m": -0.03180625585733421, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.12133006538114377, "price_path": [-0.1637, -0.1455, -0.1092, -0.1183, -0.1328, -0.1328, -0.0873, -0.1037, -0.0764, -0.0764, -0.0855, -0.0674, -0.0601, -0.0474, 0.0054, 0.009, -0.0001, -0.0328, -0.031, -0.011, -0.0274, -0.0401, -0.0346, -0.0328, -0.0219, -0.0128, -0.0128, -0.0146, -0.0001, 0.0181, 0.0236, 0.0036, 0.0036, 0.0308, 0.029, 0.0417, 0.0454, 0.0399, 0.029, 0.029, 0.0145, 0.0163, 0.0254, 0.0526, 0.0654, 0.0545, 0.0381, 0.0457, 0.0705, 0.0743, 0.0667, 0.059, 0.0552, 0.0362, 0.0514, 0.04, 0.04, 0.0267, 0.0305, 0.0381, 0.0324, 0.0419, 0.0419, 0.0, 0.0076, 0.0248, 0.0229, 0.0362, 0.019, 0.0286, 0.0724, 0.0724, 0.0705, 0.0971, 0.099, 0.101, 0.1067, 0.1048, 0.0781, 0.0343, 0.0305, 0.0305, 0.0248, 0.021, -0.0114, -0.0133, -0.0152, 0.021, 0.0095, 0.0248, 0.0133, 0.0, -0.0248, -0.0343, -0.0343, -0.021, -0.0152, -0.0095, 0.0, 0.04, 0.0286, 0.0438, 0.0514, 0.0419, 0.04, 0.0648, 0.0705, 0.0705, 0.0914, 0.08, 0.0857, 0.0857, 0.0857, 0.1048, 0.0914, 0.12, 0.1429, 0.1429, 0.1752, 0.1657, 0.1352, 0.1352, 0.1219, 0.0819, 0.0895, 0.1105, 0.099, 0.1295, 0.1086, 0.1029, 0.1181, 0.1181, 0.179, 0.1581, 0.1733, 0.1657, 0.1448, 0.1333, 0.0952, 0.1029, 0.1162, 0.0933, 0.1105, 0.1105, 0.0971, 0.0857, 0.0648, 0.0514, 0.0286, 0.0171, 0.0438, 0.0286, 0.0362, 0.0476, 0.0648, 0.0724, 0.0743, 0.0933, 0.099, 0.1143, 0.1219, 0.1333, 0.1333, 0.1276, 0.1143, 0.0971, 0.0914, 0.0857, 0.04, 0.0095, -0.0095, 0.0095, 0.0171, 0.0076, 0.0057, 0.0057, 0.0019, 0.0019, 0.0019, 0.0362, 0.0362, 0.0552, 0.059, 0.0743, 0.0743, 0.04, 0.0781, 0.0686, 0.0629, 0.0895]}
{"unit_id": "thread:1947554940626309599", "ticker_idx": 3, "ts": "2025-07-22T07:10:56+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Quanta and Wistron both assess that servers will remain the main growth driver in the second half", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply chain analysis signals accelerating CSP data center deployments via dual Nvidia Blackwell + ASIC engines; TSMC raised full-year guidance; ODMs Quanta, Wistron, Wiwynn all bullish on H2 server growth.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia and ASIC dual engines accelerate CSP data center deployments\n\nThe GB200 servers are expected to begin smooth shipments starting in the second quarter of 2025, coupled with strong shipments of ASIC servers, signaling that US-based cloud service providers (CSPs) are poised to deliver impressive cloud performance. According to the supply chain, despite tariff-related adjustments, clients have been actively stocking up, and current orders show no signs of easing. This indicates that the AI boom will continue heating up throughout the second half of 2025.\n\nCloud demand remains robust despite AWS layoffs\nUS CSPs are set to release their second-quarter financial reports soon. Recent news about layoffs at AWS has drawn attention to the future momentum of cloud business growth. Supply chain analysis suggests that, based on shipment trends in the second quarter, CSP demand remains robust and may be reflected positively in their upcoming quarterly earnings.\n\nSupply constraints ease as dual-system approach gains traction\nThe server supply chain notes that although GB200 shipments began at the end of 2024, they have been facing challenges, which also slowed down CSP cloud data center deployment progress. Amazon CEO Andy Jassy recently mentioned that if supply capacity improves, cloud business growth could be even stronger, citing difficulties in sourcing certain components such as motherboards.\n\nAccording to the server supply chain, all major CSPs are simultaneously adopting Nvidia's Blackwell chip servers alongside their own developed ASIC servers. ASIC server shipments have been smoother in comparison. With these two major systems powering forward together, CSP data center construction is expected to accelerate.\n\nTariff concerns fail to dampen growth outlook\nThe supply chain further explains that in response to reciprocal tariffs imposed by the Trump administration, clients aggressively stocked inventory in the first half of the year. While there were concerns about a potential slowdown in purchasing momentum during the second half, TSMC's recent earnings call reassured stakeholders. Despite tariff uncertainties, TSMC raised its full-year revenue growth forecast, driven primarily by strong demand from AI and other high-performance computing (HPC) applications.\n\nTSMC Chairman C.C. Wei stressed that client behavior remains unchanged for the second half of the year. Server original design manufacturers (ODMs) hold a positive outlook on orders for the latter half, expecting continued growth. The only variable is the generational transition of Nvidia GPUs, which may cause short-term fluctuations; however, server shipments are projected to grow both year-over-year and compared to the first half of the year.\n\nServer momentum drives ODM optimism\nQuanta and Wistron both assess that servers will remain the main growth driver in the second half. Wistron's President and CEO Jeff Lin noted that tariffs inevitably impact economic conditions, especially mature products like notebooks and smartphones. However, AI servers are still in the development phase, with clear customer demand, representing a rigid demand that is less affected by tariffs.\n\nWiwynn also reported strong shipments of both the GB series and ASIC servers in the second half. Wiwynn's President and CEO William Lin stated that benefiting from simultaneous shipments of Nvidia's GB series servers and ASIC servers, CSP cloud data center deployments will proceed more smoothly in the coming months.\n\n---\n\nhttps://t.co/jFZJytBZCV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03493449781659397, "ret_signed_m1d": 0.03493449781659397, "alpha_spy_m1d": 0.03507759849908243, "alpha_c_m1d": 0.02558408500518583, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.009350412811408138, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013100436681222627, "ret_signed_p1d": 0.013100436681222627, "alpha_spy_p1d": 0.00459289555910769, "alpha_c_p1d": 0.0073819544557442995, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.005718482225478327, "ret_p1w": 0.05240174672489073, "ret_signed_p1w": 0.05240174672489073, "alpha_spy_p1w": 0.04222466501371591, "alpha_c_p1w": 0.03200080299359076, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.02040094373129997, "ret_p1m": -0.008733624454148492, "ret_signed_p1m": -0.008733624454148492, "alpha_spy_p1m": -0.023442749804571505, "alpha_c_p1m": -0.0127518701785172, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.0040182457243687075, "ret_p3m": 0.19213973799126638, "ret_signed_p3m": 0.19213973799126638, "alpha_spy_p3m": 0.13271155148311586, "alpha_c_p3m": 0.08949191059066641, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.10264782740059997, "ret_p6m": 0.27510917030567694, "ret_signed_p6m": 0.27510917030567694, "alpha_spy_p6m": 0.17101673548671648, "alpha_c_p6m": 0.15377910492453317, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.12133006538114377, "price_path": [-0.147, -0.1344, -0.1344, -0.1217, -0.1428, -0.1428, -0.109, -0.1259, -0.0795, -0.1006, -0.0964, -0.0964, -0.0837, -0.0837, -0.033, -0.0584, -0.0626, -0.0795, -0.0753, -0.0457, -0.0457, -0.0457, -0.0204, -0.0288, -0.0246, -0.0161, -0.0161, -0.0288, -0.0437, -0.0044, 0.0044, -0.0393, -0.0087, 0.0, 0.0044, 0.0175, 0.0349, 0.0393, 0.0306, 0.0131, 0.0087, 0.0349, 0.0349, 0.048, 0.0961, 0.0917, 0.0568, 0.0699, 0.048, 0.048, 0.0568, 0.0393, 0.0306, 0.0349, 0.0524, 0.048, 0.0437, 0.0218, 0.048, 0.0306, 0.0306, 0.0393, 0.0349, 0.0, 0.0131, 0.0218, 0.0175, 0.048, 0.0524, 0.0524, 0.0742, 0.0742, 0.0437, 0.083, 0.0699, 0.0786, 0.0873, 0.1135, 0.1004, 0.0262, 0.0175, 0.0175, 0.0131, 0.0087, -0.0087, -0.0262, -0.0218, -0.0087, -0.0131, 0.0044, 0.0044, -0.0131, -0.0437, -0.0393, -0.0393, -0.0131, 0.0087, 0.0, 0.0131, 0.048, 0.0393, 0.048, 0.0437, 0.0306, 0.0393, 0.048, 0.0568, 0.0611, 0.0699, 0.0611, 0.0961, 0.1179, 0.1179, 0.2271, 0.2751, 0.3057, 0.3406, 0.3406, 0.3624, 0.3144, 0.3188, 0.3188, 0.2707, 0.2052, 0.2052, 0.2183, 0.1921, 0.2402, 0.2445, 0.2533, 0.2489, 0.2489, 0.2795, 0.2926, 0.3319, 0.3537, 0.3144, 0.2533, 0.214, 0.2009, 0.2402, 0.2009, 0.2096, 0.262, 0.214, 0.2576, 0.2271, 0.2052, 0.1965, 0.1965, 0.2489, 0.214, 0.214, 0.2183, 0.2445, 0.2707, 0.262, 0.2314, 0.2489, 0.2664, 0.2751, 0.3188, 0.3013, 0.3057, 0.2882, 0.2576, 0.2533, 0.2358, 0.214, 0.2052, 0.2183, 0.2576, 0.2795, 0.2664, 0.2707, 0.2707, 0.2795, 0.2751, 0.2838, 0.3144, 0.3144, 0.3537, 0.3537, 0.3493, 0.3843, 0.3188, 0.3013, 0.2751, 0.2533, 0.2751]}
{"unit_id": "thread:1947554940626309599", "ticker_idx": 4, "ts": "2025-07-22T07:10:56+00:00", "symbol": "Wiwynn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Wiwynn also reported strong shipments of both the GB series and ASIC servers in the second half", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply chain analysis signals accelerating CSP data center deployments via dual Nvidia Blackwell + ASIC engines; TSMC raised full-year guidance; ODMs Quanta, Wistron, Wiwynn all bullish on H2 server growth.", "resolved_tickers": ["6669.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wiwynn Corp listed on TWSE as 6669.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia and ASIC dual engines accelerate CSP data center deployments\n\nThe GB200 servers are expected to begin smooth shipments starting in the second quarter of 2025, coupled with strong shipments of ASIC servers, signaling that US-based cloud service providers (CSPs) are poised to deliver impressive cloud performance. According to the supply chain, despite tariff-related adjustments, clients have been actively stocking up, and current orders show no signs of easing. This indicates that the AI boom will continue heating up throughout the second half of 2025.\n\nCloud demand remains robust despite AWS layoffs\nUS CSPs are set to release their second-quarter financial reports soon. Recent news about layoffs at AWS has drawn attention to the future momentum of cloud business growth. Supply chain analysis suggests that, based on shipment trends in the second quarter, CSP demand remains robust and may be reflected positively in their upcoming quarterly earnings.\n\nSupply constraints ease as dual-system approach gains traction\nThe server supply chain notes that although GB200 shipments began at the end of 2024, they have been facing challenges, which also slowed down CSP cloud data center deployment progress. Amazon CEO Andy Jassy recently mentioned that if supply capacity improves, cloud business growth could be even stronger, citing difficulties in sourcing certain components such as motherboards.\n\nAccording to the server supply chain, all major CSPs are simultaneously adopting Nvidia's Blackwell chip servers alongside their own developed ASIC servers. ASIC server shipments have been smoother in comparison. With these two major systems powering forward together, CSP data center construction is expected to accelerate.\n\nTariff concerns fail to dampen growth outlook\nThe supply chain further explains that in response to reciprocal tariffs imposed by the Trump administration, clients aggressively stocked inventory in the first half of the year. While there were concerns about a potential slowdown in purchasing momentum during the second half, TSMC's recent earnings call reassured stakeholders. Despite tariff uncertainties, TSMC raised its full-year revenue growth forecast, driven primarily by strong demand from AI and other high-performance computing (HPC) applications.\n\nTSMC Chairman C.C. Wei stressed that client behavior remains unchanged for the second half of the year. Server original design manufacturers (ODMs) hold a positive outlook on orders for the latter half, expecting continued growth. The only variable is the generational transition of Nvidia GPUs, which may cause short-term fluctuations; however, server shipments are projected to grow both year-over-year and compared to the first half of the year.\n\nServer momentum drives ODM optimism\nQuanta and Wistron both assess that servers will remain the main growth driver in the second half. Wistron's President and CEO Jeff Lin noted that tariffs inevitably impact economic conditions, especially mature products like notebooks and smartphones. However, AI servers are still in the development phase, with clear customer demand, representing a rigid demand that is less affected by tariffs.\n\nWiwynn also reported strong shipments of both the GB series and ASIC servers in the second half. Wiwynn's President and CEO William Lin stated that benefiting from simultaneous shipments of Nvidia's GB series servers and ASIC servers, CSP cloud data center deployments will proceed more smoothly in the coming months.\n\n---\n\nhttps://t.co/jFZJytBZCV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.022267206477732726, "ret_signed_m1d": 0.022267206477732726, "alpha_spy_m1d": 0.022410307160221188, "alpha_c_m1d": 0.012916793666324589, "bench_spy_m1d": -0.00014310068248846175, "bench_c_m1d": 0.009350412811408138, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010121457489878583, "ret_signed_p1d": -0.010121457489878583, "alpha_spy_p1d": -0.01862899861199352, "alpha_c_p1d": -0.01583993971535691, "bench_spy_p1d": 0.008507541122114937, "bench_c_p1d": 0.005718482225478327, "ret_p1w": 0.06477732793522262, "ret_signed_p1w": 0.06477732793522262, "alpha_spy_p1w": 0.0546002462240478, "alpha_c_p1w": 0.04437638420392265, "bench_spy_p1w": 0.010177081711174818, "bench_c_p1w": 0.02040094373129997, "ret_p1m": 0.2085020242914979, "ret_signed_p1m": 0.2085020242914979, "alpha_spy_p1m": 0.19379289894107488, "alpha_c_p1m": 0.20448377856712918, "bench_spy_p1m": 0.014709125350423014, "bench_c_p1m": 0.0040182457243687075, "ret_p3m": 0.5323886639676114, "ret_signed_p3m": 0.5323886639676114, "alpha_spy_p3m": 0.4729604774594609, "alpha_c_p3m": 0.42974083656701145, "bench_spy_p3m": 0.059428186508150516, "bench_c_p3m": 0.10264782740059997, "ret_p6m": 0.6032388663967612, "ret_signed_p6m": 0.6032388663967612, "alpha_spy_p6m": 0.4991464315778007, "alpha_c_p6m": 0.4819088010156174, "bench_spy_p6m": 0.10409243481896047, "bench_c_p6m": 0.12133006538114377, "price_path": [-0.2669, -0.2511, -0.2531, -0.2413, -0.2492, -0.2492, -0.1882, -0.1981, -0.1646, -0.1784, -0.1548, -0.1352, -0.1135, -0.1253, -0.0782, -0.0644, -0.0644, -0.1116, -0.0958, -0.0782, -0.0801, -0.1037, -0.0683, -0.086, -0.0644, -0.0487, -0.0487, -0.0526, -0.0251, -0.0035, -0.0231, -0.031, -0.031, -0.0231, -0.0074, -0.0035, 0.0083, -0.0074, 0.0024, 0.0064, -0.0054, -0.0035, 0.0024, 0.0182, 0.0972, 0.0607, 0.0243, 0.0243, -0.0081, 0.0081, 0.0, -0.0061, -0.0061, -0.0202, 0.0364, 0.0506, 0.0547, -0.002, 0.0202, 0.0142, -0.0061, 0.0486, 0.0223, 0.0, -0.0101, 0.0162, 0.0182, 0.0526, 0.0648, 0.0729, 0.1194, 0.1194, 0.0709, 0.1154, 0.1235, 0.1377, 0.17, 0.2854, 0.3239, 0.3704, 0.3745, 0.3543, 0.3866, 0.3381, 0.2085, 0.2267, 0.2024, 0.2348, 0.2368, 0.2085, 0.2045, 0.2004, 0.1478, 0.168, 0.1842, 0.1883, 0.2247, 0.2652, 0.2186, 0.3401, 0.3401, 0.3117, 0.3036, 0.2955, 0.2794, 0.2955, 0.3077, 0.2652, 0.2713, 0.247, 0.253, 0.2348, 0.2348, 0.3421, 0.336, 0.3421, 0.4028, 0.4028, 0.4211, 0.3907, 0.3644, 0.3644, 0.3887, 0.4211, 0.5364, 0.5628, 0.5324, 0.5709, 0.6457, 0.6457, 0.6619, 0.6619, 0.6417, 0.6538, 0.664, 0.7409, 0.7672, 0.747, 0.7409, 0.6862, 0.749, 0.6498, 0.7389, 0.9109, 0.8462, 0.8057, 0.7328, 0.7611, 0.6721, 0.6802, 0.7328, 0.6579, 0.6316, 0.6883, 0.8016, 0.8016, 0.8502, 0.7935, 0.8138, 0.8623, 0.8401, 0.8968, 0.8745, 0.83, 0.832, 0.83, 0.7834, 0.7368, 0.6842, 0.6903, 0.6903, 0.7348, 0.7206, 0.7227, 0.7166, 0.7166, 0.7409, 0.7611, 0.7794, 0.8158, 0.8158, 0.8138, 0.8381, 0.8684, 0.8907, 0.8826, 0.7166, 0.6579, 0.5789, 0.6032]}
{"unit_id": "orig:1959786357968224556", "ticker_idx": 0, "ts": "2025-08-25T01:14:13+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS raises its price target from $175 to $205, maintaining a Buy rating", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses UBS Buy on NVDA with PT raised to $205, driven by strong Blackwell ramp, CSP capex, and China revenue recovery.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA FY26Q2 Earnings Preview: Strong Upside Momentum\nUBS (T. Arcuri, 25/08/21)\n\nEarnings Outlook\n\nFQ2 (July) revenue is expected to be around $46.0 billion, about $1.0 billion above the prior quarter. FQ3 (October) guidance is $54.0–55.0 billion (ex-China), and could reach $57.0 billion including China. Compute business is expected to grow 20%–25% QoQ in FQ3, contributing about $7–8 billion incremental revenue; Networking is expected to be around $6.0 billion, with total data center revenue reasonably expected at $49.0 billion.\n\nChina Market and H20 Factors\n\nThe company may leverage already provisioned H20 inventory, while also adding Hopper wafer orders and developing a Blackwell version for China. H20 demand remains uncertain: on the one hand, customers are being discouraged from purchasing; on the other hand, projects such as Deepseek R2 were delayed due to lack of H20. Overall, China’s market is expected to restore several billion dollars of revenue per quarter, but a 15% license fee must be borne, which is expected to impact gross margin by 50–60bps. FQ3 gross margin is expected to be 73.5%.\n\nSupply Chain and Shipments\n\nGPU output increased +20% QoQ in CQ2 and is expected to rise another +30% in CQ3, reaching about 5 million units for the year, with an ASP of about $33,000, still below UBS Asia team’s forecast of 6.5 million units. Hon Hai expects Q3 cabinet shipments of about 4.6k units (vs 1.15k in Q2), while Quanta confirmed that customers are ramping Blackwell Ultra. The UBS team raised its 2025 cabinet shipment forecast from 27k to 28k, with potential to reach 30k.\n\nDemand and Capital Expenditure\n\nOverall demand is strong, with Texas data centers adding 40GW of new power demand. The four major CSPs’ 2026 Capex forecast was raised by $75 billion in July, reaching nearly $400 billion, with capital intensity expected to rise to 41%, almost double that of 2023.\n\nValuation and Price Target\n\nUBS estimates F2026/C2025E EPS at $4.53, F2027/C2026E EPS at $6.85, and F2028/C2027E EPS at $7.05. Based on a 29x PE, UBS raises its price target from $175 to $205, maintaining a Buy rating.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010121735824010547, "ret_signed_m1d": -0.010121735824010547, "alpha_spy_m1d": -0.014542348693156959, "alpha_c_m1d": -0.009169051936335437, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01090037658917331, "ret_signed_p1d": 0.01090037658917331, "alpha_spy_p1d": 0.006713340239282495, "alpha_c_p1d": 0.0012719889215084823, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.03131088170414875, "ret_signed_p1w": -0.03131088170414875, "alpha_spy_p1w": -0.035326711638982156, "alpha_c_p1w": -0.01896056982958516, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": -0.007618864936297265, "ret_signed_p1m": -0.007618864936297265, "alpha_spy_p1m": -0.042762646229730694, "alpha_c_p1m": -0.10213433451457077, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.00467260846353823, "ret_signed_p3m": 0.00467260846353823, "alpha_spy_p3m": -0.013801684658514723, "alpha_c_p3m": -0.10141057030732781, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.028812254552892602, "ret_signed_p6m": 0.028812254552892602, "alpha_spy_p6m": -0.04013497353668627, "alpha_c_p6m": -0.36196749018156815, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.2503, -0.226, -0.2485, -0.236, -0.2147, -0.2108, -0.2215, -0.2119, -0.2068, -0.1994, -0.2057, -0.1936, -0.2104, -0.1953, -0.1985, -0.1909, -0.1909, -0.2, -0.1982, -0.1775, -0.1418, -0.1379, -0.1227, -0.1214, -0.1474, -0.1255, -0.1138, -0.1138, -0.12, -0.1102, -0.0942, -0.0874, -0.0828, -0.0875, -0.0507, -0.0469, -0.0379, -0.0412, -0.0469, -0.0711, -0.0502, -0.0338, -0.0351, -0.017, -0.0239, -0.003, -0.0108, -0.0339, 0.0011, -0.0086, -0.0022, 0.0053, 0.0161, 0.0125, 0.0186, 0.0099, 0.0123, 0.0036, 0.0122, -0.0232, -0.0245, -0.0269, -0.0101, 0.0, 0.0109, 0.01, 0.002, -0.0313, -0.0313, -0.0502, -0.0511, -0.0453, -0.0711, -0.064, -0.0503, -0.0138, -0.0146, -0.011, -0.0114, -0.0274, -0.0529, -0.0198, -0.0174, 0.0212, -0.0076, -0.0157, -0.0117, -0.009, 0.0114, 0.0377, 0.0414, 0.0506, 0.0435, 0.0319, 0.0291, 0.0518, 0.071, 0.0187, 0.0474, 0.0013, 0.0002, 0.0112, 0.019, 0.0158, 0.0076, 0.0027, 0.0131, 0.0359, 0.065, 0.1181, 0.1515, 0.1284, 0.1262, 0.1506, 0.1051, 0.0857, 0.0461, 0.0464, 0.1071, 0.0743, 0.0779, 0.0393, 0.0577, 0.0378, 0.0087, 0.0374, 0.0047, -0.0051, 0.0153, -0.011, 0.0026, 0.0026, -0.0156, 0.0007, 0.0092, -0.0012, 0.02, 0.0146, 0.032, 0.0288, 0.0222, 0.0063, -0.0265, -0.0195, -0.0115, -0.0492, -0.0314, 0.0067, 0.0217, 0.0524, 0.0491, 0.0491, 0.0597, 0.0469, 0.0431, 0.0373, 0.0373, 0.0504, 0.0463, 0.0414, 0.0518, 0.0292, 0.0282, 0.0286, 0.0335, 0.0186, 0.0404, 0.0358, 0.0358, -0.0096, 0.0196, 0.0281, 0.0438, 0.0372, 0.0486, 0.0652, 0.0708, 0.0631, 0.0324, 0.0031, -0.0311, -0.044, 0.0313, 0.057, 0.0487, 0.0571, 0.0398, 0.0168, 0.0168, 0.0288]}
{"unit_id": "orig:1968273482514608458", "ticker_idx": 0, "ts": "2025-09-17T11:19:01+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Barclays: AI server revenue raised to $235B in CY25 and $324B in CY26. $NVDA", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Barclays raise in AI server revenue forecasts alongside NVDA ticker; implied bullish endorsement.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Barclays: AI server revenue raised to $235B in CY25 and $324B in CY26.\n\n$NVDA https://t.co/ubw1y7Qsx2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026954090456944746, "ret_signed_m1d": 0.026954090456944746, "alpha_spy_m1d": 0.02571010959867448, "alpha_c_m1d": 0.02053312971971666, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.034940451119733984, "ret_signed_p1d": 0.034940451119733984, "alpha_spy_p1d": 0.030268005977403645, "alpha_c_p1d": -0.003454338781990085, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.03922716396777459, "ret_signed_p1w": 0.03922716396777459, "alpha_spy_p1w": 0.03353400461123068, "alpha_c_p1w": -0.012730208137761156, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.06764932027350357, "ret_signed_p1m": 0.06764932027350357, "alpha_spy_p1m": 0.06265584178852701, "alpha_c_p1m": -0.057002521668529704, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.03529165061072326, "ret_signed_p3m": 0.03529165061072326, "alpha_spy_p3m": -0.0002634924218634094, "alpha_c_p3m": -0.12080992183568728, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.07557761069639479, "ret_signed_p6m": 0.07557761069639479, "alpha_spy_p6m": 0.05934507050279625, "alpha_c_p6m": -0.19989400037546523, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.1553, -0.1534, -0.1315, -0.0939, -0.0897, -0.0737, -0.0723, -0.0998, -0.0766, -0.0644, -0.0644, -0.0708, -0.0605, -0.0436, -0.0364, -0.0316, -0.0366, 0.0024, 0.0063, 0.0159, 0.0124, 0.0063, -0.0192, 0.0028, 0.0202, 0.0188, 0.0379, 0.0306, 0.0527, 0.0445, 0.0201, 0.057, 0.0467, 0.0536, 0.0615, 0.0728, 0.0691, 0.0755, 0.0663, 0.0688, 0.0596, 0.0688, 0.0314, 0.0299, 0.0275, 0.0452, 0.0558, 0.0674, 0.0664, 0.058, 0.0228, 0.0228, 0.0028, 0.0019, 0.008, -0.0193, -0.0117, 0.0027, 0.0413, 0.0404, 0.0442, 0.0438, 0.027, 0.0, 0.0349, 0.0375, 0.0782, 0.0478, 0.0392, 0.0435, 0.0464, 0.0679, 0.0957, 0.0995, 0.1092, 0.1018, 0.0896, 0.0866, 0.1105, 0.1308, 0.0756, 0.1059, 0.0572, 0.056, 0.0676, 0.0759, 0.0725, 0.0638, 0.0587, 0.0697, 0.0938, 0.1245, 0.1805, 0.2158, 0.1914, 0.1891, 0.2149, 0.1668, 0.1463, 0.1045, 0.1049, 0.1689, 0.1343, 0.1381, 0.0973, 0.1167, 0.0958, 0.065, 0.0953, 0.0608, 0.0504, 0.072, 0.0442, 0.0585, 0.0585, 0.0394, 0.0566, 0.0656, 0.0546, 0.0769, 0.0712, 0.0897, 0.0863, 0.0793, 0.0625, 0.0278, 0.0353, 0.0437, 0.0039, 0.0227, 0.0629, 0.0787, 0.1112, 0.1076, 0.1076, 0.1189, 0.1054, 0.1014, 0.0953, 0.0953, 0.1091, 0.1048, 0.0996, 0.1106, 0.0867, 0.0856, 0.0861, 0.0912, 0.0755, 0.0985, 0.0937, 0.0937, 0.0457, 0.0766, 0.0855, 0.1021, 0.0951, 0.1071, 0.1247, 0.1305, 0.1224, 0.09, 0.0591, 0.023, 0.0094, 0.0889, 0.116, 0.1072, 0.1161, 0.0978, 0.0736, 0.0736, 0.0863, 0.1039, 0.1035, 0.1147, 0.1249, 0.1325, 0.1485, 0.0858, 0.0406, 0.0716, 0.0574, 0.0749, 0.0767, 0.0443, 0.0726, 0.0851, 0.0926, 0.0756]}
{"unit_id": "orig:1959759519283888180", "ticker_idx": 0, "ts": "2025-08-24T23:27:34+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan sees no AI CAPEX peak, forecasts +20% growth in 2026; maintains positive view on TSMC, SK Hynix, Hon Hai, Delta; strong NVDA shipments expected with Vera Rubin in 2H26.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_Asia Tech Strategy: No signs of excesses in Gen-AI – JPMorgan\n\n[Key Takeaways]\n1. No signs of overheating in the AI industry at present.\n\n2. Top 4 CSP CAPEX forecast: $398bn in 2026 → $468bn in 2027 (YoY +18%).\n\n3. Adoption rate of reasoning models among ChatGPT users has increased 3–7x within just a few months.\n\n[Contents]\n\n1. Over the past week, concerns have re-emerged in the market about an AI CAPEX peak-out in 2026.\n\n2. The reasons are:\n- ROI still low due to limited AI adoption,\n- Concerns about companies’ ability to continuously finance CAPEX.\n\n3. On top of that, there is a lack of near-term momentum in NVDA/AMD’s GPU shipments to China.\n\n4. However, our view is different.\n\n5. We see no signs of overheating in AI investments.\n\n6. The top four hyperscalers are still funding CAPEX with growing OCF, and we see sufficient growth potential even into 2027.\n\n7. Even if OCF increases by just +10% in 2027 while FCF remains flat at 2026 levels, the top four hyperscalers could still raise CAPEX by +18%.\n\n8. In addition, large private AI labs and sovereign AI efforts are expanding (Human, Stargate, Anthropic, etc.).\n\n9. On the consumer side, the scope of AI is expanding significantly, and adoption of reasoning models is still in its early stages—with token consumption rising steeply.\n\n10. Currently, 7% of free ChatGPT users and 24% of paid users are leveraging reasoning models. Just a few months ago, the figures were only 1% and 7%, respectively.\n\n11. In 2027, we expect further growth driven by wider adoption of reasoning models and the emergence of enterprise/agentic AI workloads.\n\n12. China is also only at the beginning stage of CAPEX growth, implying additional demand potential.\n\n13. Overall, we forecast AI CAPEX to grow by at least YoY +20% in 2026.\n\n14. We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders, while also seeing opportunities in substrates, OSAT, and Chinese AI infrastructure companies.\n\n15. In 2026, the Google TPU supply chain will likely show the largest growth, followed by those related to NVIDIA, AMD, and Amazon.\n\n16. NVIDIA’s GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.\n\n17. For NVIDIA’s quarterly earnings this week, we expect topline growth of around QoQ +14%, excluding shipments to China.\n\n18. Positive factors: growth in inference tokens, rising demand beyond the U.S. top four hyperscalers, resumption of shipments to China.\n\nNegative factors: competition from ASIC adoption (e.g., TPU, Trainium).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010121735824010547, "ret_signed_m1d": -0.010121735824010547, "alpha_spy_m1d": -0.014542348693156959, "alpha_c_m1d": -0.009169051936335437, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01090037658917331, "ret_signed_p1d": 0.01090037658917331, "alpha_spy_p1d": 0.006713340239282495, "alpha_c_p1d": 0.0012719889215084823, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.03131088170414875, "ret_signed_p1w": -0.03131088170414875, "alpha_spy_p1w": -0.035326711638982156, "alpha_c_p1w": -0.01896056982958516, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": -0.007618864936297265, "ret_signed_p1m": -0.007618864936297265, "alpha_spy_p1m": -0.042762646229730694, "alpha_c_p1m": -0.10213433451457077, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.00467260846353823, "ret_signed_p3m": 0.00467260846353823, "alpha_spy_p3m": -0.013801684658514723, "alpha_c_p3m": -0.10141057030732781, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.028812254552892602, "ret_signed_p6m": 0.028812254552892602, "alpha_spy_p6m": -0.04013497353668627, "alpha_c_p6m": -0.36196749018156815, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.2503, -0.226, -0.2485, -0.236, -0.2147, -0.2108, -0.2215, -0.2119, -0.2068, -0.1994, -0.2057, -0.1936, -0.2104, -0.1953, -0.1985, -0.1909, -0.1909, -0.2, -0.1982, -0.1775, -0.1418, -0.1379, -0.1227, -0.1214, -0.1474, -0.1255, -0.1138, -0.1138, -0.12, -0.1102, -0.0942, -0.0874, -0.0828, -0.0875, -0.0507, -0.0469, -0.0379, -0.0412, -0.0469, -0.0711, -0.0502, -0.0338, -0.0351, -0.017, -0.0239, -0.003, -0.0108, -0.0339, 0.0011, -0.0086, -0.0022, 0.0053, 0.0161, 0.0125, 0.0186, 0.0099, 0.0123, 0.0036, 0.0122, -0.0232, -0.0245, -0.0269, -0.0101, 0.0, 0.0109, 0.01, 0.002, -0.0313, -0.0313, -0.0502, -0.0511, -0.0453, -0.0711, -0.064, -0.0503, -0.0138, -0.0146, -0.011, -0.0114, -0.0274, -0.0529, -0.0198, -0.0174, 0.0212, -0.0076, -0.0157, -0.0117, -0.009, 0.0114, 0.0377, 0.0414, 0.0506, 0.0435, 0.0319, 0.0291, 0.0518, 0.071, 0.0187, 0.0474, 0.0013, 0.0002, 0.0112, 0.019, 0.0158, 0.0076, 0.0027, 0.0131, 0.0359, 0.065, 0.1181, 0.1515, 0.1284, 0.1262, 0.1506, 0.1051, 0.0857, 0.0461, 0.0464, 0.1071, 0.0743, 0.0779, 0.0393, 0.0577, 0.0378, 0.0087, 0.0374, 0.0047, -0.0051, 0.0153, -0.011, 0.0026, 0.0026, -0.0156, 0.0007, 0.0092, -0.0012, 0.02, 0.0146, 0.032, 0.0288, 0.0222, 0.0063, -0.0265, -0.0195, -0.0115, -0.0492, -0.0314, 0.0067, 0.0217, 0.0524, 0.0491, 0.0491, 0.0597, 0.0469, 0.0431, 0.0373, 0.0373, 0.0504, 0.0463, 0.0414, 0.0518, 0.0292, 0.0282, 0.0286, 0.0335, 0.0186, 0.0404, 0.0358, 0.0358, -0.0096, 0.0196, 0.0281, 0.0438, 0.0372, 0.0486, 0.0652, 0.0708, 0.0631, 0.0324, 0.0031, -0.0311, -0.044, 0.0313, 0.057, 0.0487, 0.0571, 0.0398, 0.0168, 0.0168, 0.0288]}
{"unit_id": "orig:1959759519283888180", "ticker_idx": 1, "ts": "2025-08-24T23:27:34+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan sees no AI CAPEX peak, forecasts +20% growth in 2026; maintains positive view on TSMC, SK Hynix, Hon Hai, Delta; strong NVDA shipments expected with Vera Rubin in 2H26.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_Asia Tech Strategy: No signs of excesses in Gen-AI – JPMorgan\n\n[Key Takeaways]\n1. No signs of overheating in the AI industry at present.\n\n2. Top 4 CSP CAPEX forecast: $398bn in 2026 → $468bn in 2027 (YoY +18%).\n\n3. Adoption rate of reasoning models among ChatGPT users has increased 3–7x within just a few months.\n\n[Contents]\n\n1. Over the past week, concerns have re-emerged in the market about an AI CAPEX peak-out in 2026.\n\n2. The reasons are:\n- ROI still low due to limited AI adoption,\n- Concerns about companies’ ability to continuously finance CAPEX.\n\n3. On top of that, there is a lack of near-term momentum in NVDA/AMD’s GPU shipments to China.\n\n4. However, our view is different.\n\n5. We see no signs of overheating in AI investments.\n\n6. The top four hyperscalers are still funding CAPEX with growing OCF, and we see sufficient growth potential even into 2027.\n\n7. Even if OCF increases by just +10% in 2027 while FCF remains flat at 2026 levels, the top four hyperscalers could still raise CAPEX by +18%.\n\n8. In addition, large private AI labs and sovereign AI efforts are expanding (Human, Stargate, Anthropic, etc.).\n\n9. On the consumer side, the scope of AI is expanding significantly, and adoption of reasoning models is still in its early stages—with token consumption rising steeply.\n\n10. Currently, 7% of free ChatGPT users and 24% of paid users are leveraging reasoning models. Just a few months ago, the figures were only 1% and 7%, respectively.\n\n11. In 2027, we expect further growth driven by wider adoption of reasoning models and the emergence of enterprise/agentic AI workloads.\n\n12. China is also only at the beginning stage of CAPEX growth, implying additional demand potential.\n\n13. Overall, we forecast AI CAPEX to grow by at least YoY +20% in 2026.\n\n14. We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders, while also seeing opportunities in substrates, OSAT, and Chinese AI infrastructure companies.\n\n15. In 2026, the Google TPU supply chain will likely show the largest growth, followed by those related to NVIDIA, AMD, and Amazon.\n\n16. NVIDIA’s GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.\n\n17. For NVIDIA’s quarterly earnings this week, we expect topline growth of around QoQ +14%, excluding shipments to China.\n\n18. Positive factors: growth in inference tokens, rising demand beyond the U.S. top four hyperscalers, resumption of shipments to China.\n\nNegative factors: competition from ASIC adoption (e.g., TPU, Trainium).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02991460381475941, "ret_signed_m1d": -0.02991460381475941, "alpha_spy_m1d": -0.03433521668390582, "alpha_c_m1d": -0.0289619199270843, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004273499767392774, "ret_signed_p1d": 0.004273499767392774, "alpha_spy_p1d": 8.646341750195852e-05, "alpha_c_p1d": -0.005354887900272054, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.004273499767392774, "ret_signed_p1w": -0.004273499767392774, "alpha_spy_p1w": -0.008289329702226178, "alpha_c_p1w": 0.008076812107170817, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.14988038544903337, "ret_signed_p1m": 0.14988038544903337, "alpha_spy_p1m": 0.11473660415559994, "alpha_c_p1m": 0.05536491587075987, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.2485641825277558, "ret_signed_p3m": 0.2485641825277558, "alpha_spy_p3m": 0.23008988940570285, "alpha_c_p3m": 0.14248100375688977, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.6487768189024732, "ret_signed_p6m": 0.6487768189024732, "alpha_spy_p6m": 0.5798295908128943, "alpha_c_p6m": 0.25799707416801243, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.177, -0.177, -0.177, -0.1949, -0.1915, -0.1574, -0.1506, -0.1532, -0.1447, -0.1106, -0.0936, -0.1068, -0.1197, -0.1239, -0.1068, -0.0983, -0.1154, -0.0983, -0.1282, -0.1026, -0.0855, -0.0812, -0.0769, -0.094, -0.0726, -0.0726, -0.0684, -0.0726, -0.0769, -0.0769, -0.0684, -0.0598, -0.0598, -0.0641, -0.0513, -0.0342, -0.0342, -0.0128, -0.0171, -0.0342, -0.0214, -0.0214, -0.0214, -0.0214, -0.0299, -0.0128, -0.0085, -0.0085, -0.0299, -0.0171, -0.0385, 0.0085, 0.0043, 0.0085, 0.0085, 0.0256, 0.0043, 0.0085, 0.0085, 0.0128, -0.0299, -0.0171, -0.0299, 0.0, 0.0043, 0.0171, -0.0085, -0.0085, -0.0043, -0.0085, -0.0085, -0.0085, 0.0085, 0.0085, 0.0256, 0.047, 0.0598, 0.0769, 0.0726, 0.0984, 0.0855, 0.1027, 0.0855, 0.1113, 0.1499, 0.1499, 0.1327, 0.1156, 0.1156, 0.1198, 0.137, 0.1713, 0.2014, 0.2014, 0.2314, 0.2142, 0.2357, 0.2357, 0.2142, 0.2228, 0.2571, 0.2743, 0.2443, 0.27, 0.27, 0.2529, 0.2443, 0.2443, 0.27, 0.2657, 0.2915, 0.2915, 0.2872, 0.2958, 0.2915, 0.2529, 0.2571, 0.2529, 0.2657, 0.2571, 0.2657, 0.2529, 0.2271, 0.24, 0.2057, 0.1971, 0.2486, 0.1885, 0.1799, 0.2142, 0.2357, 0.2314, 0.2357, 0.2099, 0.2271, 0.2443, 0.24, 0.2529, 0.2829, 0.27, 0.2915, 0.2656, 0.2743, 0.2484, 0.2355, 0.2312, 0.2312, 0.2312, 0.2613, 0.2829, 0.2872, 0.2872, 0.3001, 0.3173, 0.3087, 0.3345, 0.3345, 0.3647, 0.4378, 0.468, 0.4421, 0.4508, 0.4464, 0.4551, 0.4723, 0.4723, 0.4551, 0.4981, 0.5153, 0.5282, 0.4981, 0.5153, 0.5239, 0.511, 0.5325, 0.567, 0.5541, 0.5282, 0.5196, 0.5498, 0.5368, 0.5196, 0.5325, 0.5627, 0.6186, 0.6488, 0.6488, 0.6488, 0.6488, 0.6488]}
{"unit_id": "orig:1959759519283888180", "ticker_idx": 2, "ts": "2025-08-24T23:27:34+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan sees no AI CAPEX peak, forecasts +20% growth in 2026; maintains positive view on TSMC, SK Hynix, Hon Hai, Delta; strong NVDA shipments expected with Vera Rubin in 2H26.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_Asia Tech Strategy: No signs of excesses in Gen-AI – JPMorgan\n\n[Key Takeaways]\n1. No signs of overheating in the AI industry at present.\n\n2. Top 4 CSP CAPEX forecast: $398bn in 2026 → $468bn in 2027 (YoY +18%).\n\n3. Adoption rate of reasoning models among ChatGPT users has increased 3–7x within just a few months.\n\n[Contents]\n\n1. Over the past week, concerns have re-emerged in the market about an AI CAPEX peak-out in 2026.\n\n2. The reasons are:\n- ROI still low due to limited AI adoption,\n- Concerns about companies’ ability to continuously finance CAPEX.\n\n3. On top of that, there is a lack of near-term momentum in NVDA/AMD’s GPU shipments to China.\n\n4. However, our view is different.\n\n5. We see no signs of overheating in AI investments.\n\n6. The top four hyperscalers are still funding CAPEX with growing OCF, and we see sufficient growth potential even into 2027.\n\n7. Even if OCF increases by just +10% in 2027 while FCF remains flat at 2026 levels, the top four hyperscalers could still raise CAPEX by +18%.\n\n8. In addition, large private AI labs and sovereign AI efforts are expanding (Human, Stargate, Anthropic, etc.).\n\n9. On the consumer side, the scope of AI is expanding significantly, and adoption of reasoning models is still in its early stages—with token consumption rising steeply.\n\n10. Currently, 7% of free ChatGPT users and 24% of paid users are leveraging reasoning models. Just a few months ago, the figures were only 1% and 7%, respectively.\n\n11. In 2027, we expect further growth driven by wider adoption of reasoning models and the emergence of enterprise/agentic AI workloads.\n\n12. China is also only at the beginning stage of CAPEX growth, implying additional demand potential.\n\n13. Overall, we forecast AI CAPEX to grow by at least YoY +20% in 2026.\n\n14. We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders, while also seeing opportunities in substrates, OSAT, and Chinese AI infrastructure companies.\n\n15. In 2026, the Google TPU supply chain will likely show the largest growth, followed by those related to NVIDIA, AMD, and Amazon.\n\n16. NVIDIA’s GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.\n\n17. For NVIDIA’s quarterly earnings this week, we expect topline growth of around QoQ +14%, excluding shipments to China.\n\n18. Positive factors: growth in inference tokens, rising demand beyond the U.S. top four hyperscalers, resumption of shipments to China.\n\nNegative factors: competition from ASIC adoption (e.g., TPU, Trainium).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032755225139682875, "ret_signed_m1d": -0.032755225139682875, "alpha_spy_m1d": -0.03717583800882929, "alpha_c_m1d": -0.031802541252007765, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007707168704449874, "ret_signed_p1d": 0.007707168704449874, "alpha_spy_p1d": 0.003520132354559058, "alpha_c_p1d": -0.0019212189632149546, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.01206254011757013, "ret_signed_p1w": -0.01206254011757013, "alpha_spy_p1w": -0.016078370052403534, "alpha_c_p1w": 0.00028777175699346014, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.39314614138507986, "ret_signed_p1m": 0.39314614138507986, "alpha_spy_p1m": 0.35800236009164643, "alpha_c_p1m": 0.29863067180680636, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 1.2035636253174942, "ret_signed_p3m": 1.2035636253174942, "alpha_spy_p3m": 1.1850893321954412, "alpha_c_p3m": 1.0974804465466281, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 2.39846717626965, "ret_signed_p6m": 2.39846717626965, "alpha_spy_p6m": 2.3295199481800712, "alpha_c_p6m": 2.0076874315351896, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1999, -0.183, -0.2119, -0.2004, -0.2004, -0.1618, -0.1349, -0.1349, -0.1175, -0.1118, -0.0751, -0.0925, -0.0925, -0.0443, -0.0405, -0.0501, -0.052, -0.0096, 0.0, 0.0732, 0.1021, 0.1291, 0.0944, 0.1252, 0.1002, 0.0751, 0.0732, 0.0424, 0.0443, 0.0867, 0.0829, 0.1445, 0.1349, 0.1561, 0.1503, 0.1407, 0.0385, 0.0366, 0.0501, 0.0347, 0.0366, 0.0385, 0.025, 0.0096, 0.0116, 0.0154, 0.0539, -0.0058, -0.0058, 0.0154, -0.0039, 0.0096, -0.0116, 0.0289, 0.0366, 0.0713, 0.0655, 0.0655, 0.0308, 0.0135, -0.0154, -0.0559, -0.0328, 0.0, 0.0077, 0.0019, 0.0362, 0.0381, -0.0121, 0.0053, 0.013, 0.0246, 0.0555, 0.069, 0.1114, 0.1732, 0.1848, 0.2677, 0.2774, 0.343, 0.287, 0.3719, 0.3719, 0.3546, 0.3931, 0.3796, 0.3758, 0.2986, 0.3468, 0.341, 0.3893, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.6517, 0.6015, 0.588, 0.6305, 0.7463, 0.7964, 0.8736, 0.8485, 0.8582, 0.8466, 0.9682, 1.0646, 1.0106, 1.1534, 1.192, 1.1573, 1.3927, 1.2615, 1.2344, 1.2885, 1.2383, 1.3386, 1.3888, 1.3811, 1.3618, 1.1611, 1.3386, 1.1997, 1.1688, 1.2036, 1.0106, 1.0067, 1.0029, 1.0222, 1.1009, 1.0468, 1.0777, 1.1549, 1.1318, 1.0931, 1.1009, 1.2283, 1.1858, 1.2669, 1.182, 1.2051, 1.1395, 1.0468, 1.1279, 1.1318, 1.1125, 1.2399, 1.2553, 1.2708, 1.2708, 1.3133, 1.4716, 1.5141, 1.5141, 1.5141, 1.6145, 1.6879, 1.8037, 1.8655, 1.9196, 1.8732, 1.8926, 1.8501, 1.8655, 1.8926, 1.9196, 1.9505, 1.8694, 1.8578, 1.9157, 1.9621, 1.8424, 2.0895, 2.2479, 2.3251, 2.5105, 2.2054, 2.5027, 2.4757, 2.2517, 2.2401, 2.4255, 2.383, 2.3212, 2.4294, 2.3985, 2.3985, 2.3985]}
{"unit_id": "orig:1959759519283888180", "ticker_idx": 3, "ts": "2025-08-24T23:27:34+00:00", "symbol": "Hon Hai", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan sees no AI CAPEX peak, forecasts +20% growth in 2026; maintains positive view on TSMC, SK Hynix, Hon Hai, Delta; strong NVDA shipments expected with Vera Rubin in 2H26.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250825_Asia Tech Strategy: No signs of excesses in Gen-AI – JPMorgan\n\n[Key Takeaways]\n1. No signs of overheating in the AI industry at present.\n\n2. Top 4 CSP CAPEX forecast: $398bn in 2026 → $468bn in 2027 (YoY +18%).\n\n3. Adoption rate of reasoning models among ChatGPT users has increased 3–7x within just a few months.\n\n[Contents]\n\n1. Over the past week, concerns have re-emerged in the market about an AI CAPEX peak-out in 2026.\n\n2. The reasons are:\n- ROI still low due to limited AI adoption,\n- Concerns about companies’ ability to continuously finance CAPEX.\n\n3. On top of that, there is a lack of near-term momentum in NVDA/AMD’s GPU shipments to China.\n\n4. However, our view is different.\n\n5. We see no signs of overheating in AI investments.\n\n6. The top four hyperscalers are still funding CAPEX with growing OCF, and we see sufficient growth potential even into 2027.\n\n7. Even if OCF increases by just +10% in 2027 while FCF remains flat at 2026 levels, the top four hyperscalers could still raise CAPEX by +18%.\n\n8. In addition, large private AI labs and sovereign AI efforts are expanding (Human, Stargate, Anthropic, etc.).\n\n9. On the consumer side, the scope of AI is expanding significantly, and adoption of reasoning models is still in its early stages—with token consumption rising steeply.\n\n10. Currently, 7% of free ChatGPT users and 24% of paid users are leveraging reasoning models. Just a few months ago, the figures were only 1% and 7%, respectively.\n\n11. In 2027, we expect further growth driven by wider adoption of reasoning models and the emergence of enterprise/agentic AI workloads.\n\n12. China is also only at the beginning stage of CAPEX growth, implying additional demand potential.\n\n13. Overall, we forecast AI CAPEX to grow by at least YoY +20% in 2026.\n\n14. We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders, while also seeing opportunities in substrates, OSAT, and Chinese AI infrastructure companies.\n\n15. In 2026, the Google TPU supply chain will likely show the largest growth, followed by those related to NVIDIA, AMD, and Amazon.\n\n16. NVIDIA’s GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.\n\n17. For NVIDIA’s quarterly earnings this week, we expect topline growth of around QoQ +14%, excluding shipments to China.\n\n18. Positive factors: growth in inference tokens, rising demand beyond the U.S. top four hyperscalers, resumption of shipments to China.\n\nNegative factors: competition from ASIC adoption (e.g., TPU, Trainium).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02409638554216864, "ret_signed_m1d": -0.02409638554216864, "alpha_spy_m1d": -0.028516998411315053, "alpha_c_m1d": -0.026349858935068027, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": 0.0022534733928993855, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0072289156626506035, "ret_signed_p1d": 0.0072289156626506035, "alpha_spy_p1d": 0.0030418793127597876, "alpha_c_p1d": 0.0026456672227093136, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.00458324843994129, "ret_p1w": -0.04337349397590362, "ret_signed_p1w": -0.04337349397590362, "alpha_spy_p1w": -0.047389323910737025, "alpha_c_p1w": -0.04574165390417062, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": 0.0023681599282669996, "ret_p1m": 0.06506024096385543, "ret_signed_p1m": 0.06506024096385543, "alpha_spy_p1m": 0.029916459670422002, "alpha_c_p1m": -0.006551670098842832, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.07161191106269826, "ret_p3m": 0.1397590361445784, "ret_signed_p3m": 0.1397590361445784, "alpha_spy_p3m": 0.12128474302252545, "alpha_c_p3m": 0.09900821850564889, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.040750817638929515, "ret_p6m": 0.09397590361445785, "ret_signed_p6m": 0.09397590361445785, "alpha_spy_p6m": 0.025028675524878974, "alpha_c_p6m": 0.02556433202390429, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.06841157159055355, "price_path": [-0.2931, -0.2745, -0.2745, -0.2908, -0.2955, -0.2745, -0.2769, -0.2885, -0.2815, -0.2769, -0.2745, -0.2676, -0.2722, -0.2792, -0.2792, -0.2745, -0.2838, -0.2769, -0.2745, -0.2536, -0.2466, -0.242, -0.2327, -0.2513, -0.2304, -0.2193, -0.2096, -0.2241, -0.2241, -0.2337, -0.2241, -0.2217, -0.2217, -0.2265, -0.2145, -0.2169, -0.2096, -0.2024, -0.2048, -0.2193, -0.2, -0.1614, -0.159, -0.1518, -0.1735, -0.1735, -0.1422, -0.1422, -0.1325, -0.1108, -0.1036, -0.0627, -0.0627, -0.0458, -0.0506, -0.0434, -0.0386, -0.0024, 0.012, 0.0024, -0.0337, -0.0072, -0.0241, 0.0, 0.0072, 0.0024, -0.0072, -0.0193, -0.0434, -0.0434, -0.0337, -0.0217, -0.012, -0.0193, 0.0, 0.012, 0.0337, 0.0482, 0.041, 0.0386, 0.0217, 0.0361, 0.0313, 0.041, 0.0651, 0.0747, 0.1084, 0.0578, 0.0578, 0.041, 0.0554, 0.0867, 0.0916, 0.0916, 0.1012, 0.0843, 0.0675, 0.0675, 0.0265, -0.0072, -0.0048, 0.0747, 0.0916, 0.1494, 0.1518, 0.1687, 0.1518, 0.1518, 0.188, 0.2048, 0.253, 0.2627, 0.241, 0.212, 0.1759, 0.1904, 0.1952, 0.1759, 0.2024, 0.188, 0.2096, 0.2145, 0.1614, 0.1373, 0.1084, 0.1036, 0.1398, 0.0843, 0.0602, 0.0554, 0.094, 0.1133, 0.0867, 0.0699, 0.0699, 0.094, 0.1012, 0.1133, 0.1181, 0.1325, 0.1253, 0.0892, 0.094, 0.0675, 0.0506, 0.0434, 0.041, 0.0675, 0.0795, 0.0795, 0.0795, 0.0795, 0.0867, 0.1133, 0.0988, 0.1108, 0.1108, 0.1181, 0.1301, 0.1373, 0.1494, 0.106, 0.1108, 0.1012, 0.0916, 0.1301, 0.1253, 0.1301, 0.106, 0.0771, 0.0554, 0.0771, 0.0675, 0.0795, 0.0867, 0.0867, 0.0795, 0.0627, 0.0337, 0.0434, 0.0554, 0.0386, 0.0361, 0.053, 0.0651, 0.094, 0.094, 0.094, 0.094, 0.094]}
{"unit_id": "orig:1959759519283888180", "ticker_idx": 4, "ts": "2025-08-24T23:27:34+00:00", "symbol": "Delta Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan sees no AI CAPEX peak, forecasts +20% growth in 2026; maintains positive view on TSMC, SK Hynix, Hon Hai, Delta; strong NVDA shipments expected with Vera Rubin in 2H26.", "resolved_tickers": ["2308.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Delta Electronics Taiwan 2308.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250825_Asia Tech Strategy: No signs of excesses in Gen-AI – JPMorgan\n\n[Key Takeaways]\n1. No signs of overheating in the AI industry at present.\n\n2. Top 4 CSP CAPEX forecast: $398bn in 2026 → $468bn in 2027 (YoY +18%).\n\n3. Adoption rate of reasoning models among ChatGPT users has increased 3–7x within just a few months.\n\n[Contents]\n\n1. Over the past week, concerns have re-emerged in the market about an AI CAPEX peak-out in 2026.\n\n2. The reasons are:\n- ROI still low due to limited AI adoption,\n- Concerns about companies’ ability to continuously finance CAPEX.\n\n3. On top of that, there is a lack of near-term momentum in NVDA/AMD’s GPU shipments to China.\n\n4. However, our view is different.\n\n5. We see no signs of overheating in AI investments.\n\n6. The top four hyperscalers are still funding CAPEX with growing OCF, and we see sufficient growth potential even into 2027.\n\n7. Even if OCF increases by just +10% in 2027 while FCF remains flat at 2026 levels, the top four hyperscalers could still raise CAPEX by +18%.\n\n8. In addition, large private AI labs and sovereign AI efforts are expanding (Human, Stargate, Anthropic, etc.).\n\n9. On the consumer side, the scope of AI is expanding significantly, and adoption of reasoning models is still in its early stages—with token consumption rising steeply.\n\n10. Currently, 7% of free ChatGPT users and 24% of paid users are leveraging reasoning models. Just a few months ago, the figures were only 1% and 7%, respectively.\n\n11. In 2027, we expect further growth driven by wider adoption of reasoning models and the emergence of enterprise/agentic AI workloads.\n\n12. China is also only at the beginning stage of CAPEX growth, implying additional demand potential.\n\n13. Overall, we forecast AI CAPEX to grow by at least YoY +20% in 2026.\n\n14. We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders, while also seeing opportunities in substrates, OSAT, and Chinese AI infrastructure companies.\n\n15. In 2026, the Google TPU supply chain will likely show the largest growth, followed by those related to NVIDIA, AMD, and Amazon.\n\n16. NVIDIA’s GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.\n\n17. For NVIDIA’s quarterly earnings this week, we expect topline growth of around QoQ +14%, excluding shipments to China.\n\n18. Positive factors: growth in inference tokens, rising demand beyond the U.S. top four hyperscalers, resumption of shipments to China.\n\nNegative factors: competition from ASIC adoption (e.g., TPU, Trainium).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017751479289940808, "ret_signed_m1d": -0.017751479289940808, "alpha_spy_m1d": -0.02217209215908722, "alpha_c_m1d": -0.020004952682840194, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": 0.0022534733928993855, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013313609467455523, "ret_signed_p1d": 0.013313609467455523, "alpha_spy_p1d": 0.009126573117564707, "alpha_c_p1d": 0.008730361027514233, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.00458324843994129, "ret_p1w": 0.01775147928994092, "ret_signed_p1w": 0.01775147928994092, "alpha_spy_p1w": 0.013735649355107515, "alpha_c_p1w": 0.01538331936167392, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": 0.0023681599282669996, "ret_p1m": 0.32100591715976323, "ret_signed_p1m": 0.32100591715976323, "alpha_spy_p1m": 0.2858621358663298, "alpha_c_p1m": 0.24939400609706497, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.07161191106269826, "ret_p3m": 0.41124260355029585, "ret_signed_p3m": 0.41124260355029585, "alpha_spy_p3m": 0.3927683104282429, "alpha_c_p3m": 0.37049178591136633, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.040750817638929515, "ret_p6m": 0.863905325443787, "ret_signed_p6m": 0.863905325443787, "alpha_spy_p6m": 0.7949580973542081, "alpha_c_p6m": 0.7954937538532334, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.06841157159055355, "price_path": [-0.4622, -0.4564, -0.4564, -0.4659, -0.4484, -0.4346, -0.4215, -0.4244, -0.4223, -0.4157, -0.4157, -0.4194, -0.4215, -0.4179, -0.4157, -0.4075, -0.4024, -0.4083, -0.4083, -0.3831, -0.3735, -0.3787, -0.3728, -0.3891, -0.3669, -0.3646, -0.321, -0.324, -0.3343, -0.3321, -0.3328, -0.2862, -0.2855, -0.3003, -0.2692, -0.2604, -0.2618, -0.247, -0.2382, -0.2411, -0.2308, -0.2337, -0.2352, -0.2278, -0.2234, -0.216, -0.1612, -0.1612, -0.1228, -0.0651, -0.0814, -0.0784, -0.0577, -0.0562, -0.0503, -0.0015, -0.0089, 0.0148, 0.0178, -0.0222, -0.0754, -0.0414, -0.0178, 0.0, 0.0133, 0.0651, 0.0488, 0.0518, 0.0178, 0.0133, 0.0059, 0.0222, 0.074, 0.074, 0.1598, 0.2382, 0.2293, 0.2382, 0.2367, 0.2441, 0.2367, 0.3195, 0.3136, 0.3417, 0.321, 0.318, 0.3077, 0.2544, 0.2544, 0.2633, 0.2885, 0.3225, 0.3935, 0.3935, 0.4601, 0.4512, 0.4778, 0.4778, 0.5237, 0.4675, 0.5015, 0.5163, 0.4778, 0.5015, 0.5089, 0.5089, 0.4941, 0.4941, 0.5828, 0.5902, 0.568, 0.4941, 0.4719, 0.4541, 0.4423, 0.4305, 0.4467, 0.4275, 0.4527, 0.4379, 0.4305, 0.4172, 0.3639, 0.355, 0.3284, 0.3225, 0.4112, 0.324, 0.321, 0.3462, 0.3979, 0.3935, 0.3787, 0.4393, 0.463, 0.4749, 0.4778, 0.4482, 0.4482, 0.4364, 0.432, 0.3979, 0.3876, 0.3639, 0.3328, 0.3166, 0.318, 0.3476, 0.4098, 0.395, 0.4127, 0.4127, 0.4142, 0.4231, 0.4216, 0.4246, 0.4246, 0.4719, 0.4941, 0.5533, 0.5237, 0.4941, 0.5015, 0.5607, 0.5237, 0.568, 0.5459, 0.6642, 0.6568, 0.6716, 0.679, 0.8417, 0.8639, 0.8195, 0.8491, 0.8935, 0.8565, 0.8047, 0.7382, 0.7973, 0.7899, 0.7012, 0.716, 0.7456, 0.753, 0.8639, 0.8639, 0.8639, 0.8639, 0.8639]}
{"unit_id": "orig:1959759519283888180", "ticker_idx": 5, "ts": "2025-08-24T23:27:34+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "lack of near-term momentum in NVDA/AMD's GPU shipments to China", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan sees no AI CAPEX peak, forecasts +20% growth in 2026; maintains positive view on TSMC, SK Hynix, Hon Hai, Delta; strong NVDA shipments expected with Vera Rubin in 2H26.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250825_Asia Tech Strategy: No signs of excesses in Gen-AI – JPMorgan\n\n[Key Takeaways]\n1. No signs of overheating in the AI industry at present.\n\n2. Top 4 CSP CAPEX forecast: $398bn in 2026 → $468bn in 2027 (YoY +18%).\n\n3. Adoption rate of reasoning models among ChatGPT users has increased 3–7x within just a few months.\n\n[Contents]\n\n1. Over the past week, concerns have re-emerged in the market about an AI CAPEX peak-out in 2026.\n\n2. The reasons are:\n- ROI still low due to limited AI adoption,\n- Concerns about companies’ ability to continuously finance CAPEX.\n\n3. On top of that, there is a lack of near-term momentum in NVDA/AMD’s GPU shipments to China.\n\n4. However, our view is different.\n\n5. We see no signs of overheating in AI investments.\n\n6. The top four hyperscalers are still funding CAPEX with growing OCF, and we see sufficient growth potential even into 2027.\n\n7. Even if OCF increases by just +10% in 2027 while FCF remains flat at 2026 levels, the top four hyperscalers could still raise CAPEX by +18%.\n\n8. In addition, large private AI labs and sovereign AI efforts are expanding (Human, Stargate, Anthropic, etc.).\n\n9. On the consumer side, the scope of AI is expanding significantly, and adoption of reasoning models is still in its early stages—with token consumption rising steeply.\n\n10. Currently, 7% of free ChatGPT users and 24% of paid users are leveraging reasoning models. Just a few months ago, the figures were only 1% and 7%, respectively.\n\n11. In 2027, we expect further growth driven by wider adoption of reasoning models and the emergence of enterprise/agentic AI workloads.\n\n12. China is also only at the beginning stage of CAPEX growth, implying additional demand potential.\n\n13. Overall, we forecast AI CAPEX to grow by at least YoY +20% in 2026.\n\n14. We maintain a positive view on TSMC, SK Hynix, Hon Hai, Delta as key leaders, while also seeing opportunities in substrates, OSAT, and Chinese AI infrastructure companies.\n\n15. In 2026, the Google TPU supply chain will likely show the largest growth, followed by those related to NVIDIA, AMD, and Amazon.\n\n16. NVIDIA’s GB200/300 shipments are expected to remain strong, with Vera Rubin set for mass production from 2H26. No signs of schedule delays so far.\n\n17. For NVIDIA’s quarterly earnings this week, we expect topline growth of around QoQ +14%, excluding shipments to China.\n\n18. Positive factors: growth in inference tokens, rising demand beyond the U.S. top four hyperscalers, resumption of shipments to China.\n\nNegative factors: competition from ASIC adoption (e.g., TPU, Trainium).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0269343405977287, "ret_signed_m1d": -0.0269343405977287, "alpha_spy_m1d": -0.02251372772858229, "alpha_c_m1d": -0.02788702448540381, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019955891862486697, "ret_signed_p1d": -0.019955891862486697, "alpha_spy_p1d": -0.01576885551259588, "alpha_c_p1d": -0.010327504194821868, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.004468632007906592, "ret_signed_p1w": 0.004468632007906592, "alpha_spy_p1w": 0.008484461942739996, "alpha_c_p1w": -0.007881679866656999, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": -0.015058806958104909, "ret_signed_p1m": 0.015058806958104909, "alpha_spy_p1m": 0.05020258825153834, "alpha_c_p1m": 0.10957427653637841, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.26114105963957845, "ret_signed_p3m": -0.26114105963957845, "alpha_spy_p3m": -0.2426667665175255, "alpha_c_p3m": -0.15505788086871242, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 0.24314398305766294, "ret_signed_p6m": -0.24314398305766294, "alpha_spy_p6m": -0.17419675496808407, "alpha_c_p6m": 0.14763576167679782, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.3091, -0.3081, -0.3222, -0.2983, -0.2819, -0.2741, -0.2918, -0.2887, -0.2548, -0.2456, -0.2584, -0.2746, -0.2889, -0.2263, -0.222, -0.2239, -0.2239, -0.215, -0.2068, -0.1526, -0.1222, -0.1205, -0.1197, -0.1314, -0.1668, -0.1521, -0.1558, -0.1558, -0.1748, -0.1563, -0.1527, -0.1175, -0.1037, -0.1048, -0.0474, -0.0201, -0.0181, -0.039, -0.0389, -0.0529, -0.0288, -0.0076, 0.019, 0.0631, 0.0862, 0.0989, 0.0793, 0.0511, 0.0821, 0.067, -0.0015, 0.0553, 0.0575, 0.0546, 0.0709, 0.1289, 0.1077, 0.0866, 0.0782, 0.0195, 0.0113, 0.0021, 0.0269, 0.0, 0.02, 0.0231, 0.032, -0.0045, -0.0045, -0.0064, -0.0075, -0.0096, -0.0748, -0.0732, -0.0462, -0.0234, -0.0471, -0.0293, -0.0135, -0.0178, -0.0257, -0.0333, -0.0365, -0.0219, -0.0151, -0.0152, -0.0128, -0.0239, -0.0122, -0.0096, 0.004, 0.039, 0.008, 0.247, 0.2947, 0.442, 0.4256, 0.3155, 0.3248, 0.335, 0.4606, 0.4358, 0.4268, 0.4726, 0.4571, 0.4093, 0.4385, 0.5482, 0.5896, 0.5794, 0.6181, 0.56, 0.5678, 0.5894, 0.5307, 0.5691, 0.4551, 0.4296, 0.4935, 0.454, 0.5848, 0.5179, 0.5108, 0.4723, 0.4097, 0.3685, 0.2611, 0.2474, 0.3164, 0.2618, 0.3115, 0.3115, 0.3316, 0.3452, 0.3176, 0.332, 0.3221, 0.3343, 0.3535, 0.3566, 0.3554, 0.3555, 0.2903, 0.2707, 0.2804, 0.2127, 0.2308, 0.3065, 0.3158, 0.3155, 0.3164, 0.3164, 0.3161, 0.3198, 0.3182, 0.311, 0.311, 0.368, 0.3533, 0.3121, 0.2856, 0.2529, 0.2437, 0.2714, 0.3527, 0.3688, 0.3952, 0.4191, 0.4191, 0.4197, 0.5291, 0.5532, 0.5896, 0.5384, 0.5428, 0.5471, 0.5437, 0.4491, 0.5075, 0.4821, 0.2255, 0.1784, 0.276, 0.3222, 0.3074, 0.3074, 0.2607, 0.2691, 0.2691, 0.2431]}
{"unit_id": "orig:1950327044174418135", "ticker_idx": 0, "ts": "2025-07-29T22:46:17+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD, PT Raised to $200 from $175 (BofA)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares BofA Buy on AMD with PT raised to $200; bullish on AI GPU revenue, CPU share gains, and above-consensus 2025/2026 estimates.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AMD, PT Raised to $200 from $175 (BofA)\n\n1. AI GPU Revenue Momentum\n\n- Potential to resume shipments of MI308 AI GPU to China in 2025 (premised on U.S. BIS approval).\n- 2025 revenue upside: Approx. $0.7–1.0bn.\n- 2026 revenue upside: Approx. $1.5–2.0bn.\n- Strong pricing for current-generation MI355X GPU ($20K+ vs. consensus $17K) → ASP improvement.\n- Increase in cloud Capex → Continued strong demand for AI.\n- In the total AI GPU market, AMD has the potential to reach approx. $12bn in revenue by 2026 (+70% YoY).\n\n2. PC Market Outlook: Risks Priced In\n\n- 2H25 PC demand is low compared to seasonality (similar to INTC).\n- INTC Q3/Q4: Expected +4%/+2% QoQ (down from traditional +9%/+4%).\n- AMD Q3/Q4: Expected +2%/+0% QoQ (lower than traditional +15%/+14% → already reflects conservative outlook).\n- Some Q2 PC demand was pulled forward due to tariff issues → a headwind for the second half.\n- Intel's 18A products to launch in Q4 2025, with a significant impact from 2026 → AMD's competitive advantage is expected to continue.\n\n3. Expectation of CPU Market Share Expansion\n\n- Total CPU (PC + Server) market share: Expected to grow from <20% in 2023 to >30% in 2026.\n- Competitor estimates: Intel 73% → 59%, ARM 8% → 11%.\n- AMD continues to have a competitive advantage over Intel in the x86-based PC and cloud markets.\n- PC: AMD commands a +17% premium over Intel.\n- Server: AMD commands a +64% premium over Intel.\n\n4. Potential for Upward Revision of Financial Guidance\n\n- Q2 revenue estimate: $7.5bn+ (consensus $7.4bn).\n- Q3 revenue estimate: $8.5bn+ (consensus $8.3bn).\n- Full-year 2025 revenue/EPS forecast: $33bn+ / $4.10+ (consensus $32.2bn / $4.01).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021302968481107243, "ret_signed_m1d": -0.021302968481107243, "alpha_spy_m1d": -0.02394768982272466, "alpha_c_m1d": -0.01621599866774548, "bench_spy_m1d": 0.0026447213416174176, "bench_c_m1d": -0.005086969813361764, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011665870361522446, "ret_signed_p1d": 0.011665870361522446, "alpha_spy_p1d": 0.012925104600601367, "alpha_c_p1d": 0.00019426848129167062, "bench_spy_p1d": -0.0012592342390789213, "bench_c_p1d": 0.011471601880230775, "ret_p1w": -0.01763979282995154, "ret_signed_p1w": -0.01763979282995154, "alpha_spy_p1w": -0.006164153625172086, "alpha_c_p1w": 0.0021620983748227296, "bench_spy_p1w": -0.011475639204779453, "bench_c_p1w": -0.01980189120477427, "ret_p1m": -0.05810413332246367, "ret_signed_p1m": -0.05810413332246367, "alpha_spy_p1m": -0.07600239037406875, "alpha_c_p1m": -0.07435542996569278, "bench_spy_p1m": 0.017898257051605082, "bench_c_p1m": 0.016251296643229107, "ret_p3m": 0.42538319820224224, "ret_signed_p3m": 0.42538319820224224, "alpha_spy_p3m": 0.35632850174106245, "alpha_c_p3m": 0.22654384086696489, "bench_spy_p3m": 0.06905469646117979, "bench_c_p3m": 0.19883935733527736, "ret_p6m": 0.407799817486173, "ret_signed_p6m": 0.407799817486173, "alpha_spy_p6m": 0.3226831987917187, "alpha_c_p6m": 0.031286814473518776, "bench_spy_p6m": 0.0851166186944543, "bench_c_p6m": 0.3765130030126542, "price_path": [-0.4553, -0.4432, -0.4331, -0.4442, -0.4344, -0.4268, -0.4204, -0.3907, -0.3662, -0.3366, -0.3519, -0.3397, -0.3534, -0.3603, -0.3685, -0.3761, -0.3783, -0.3783, -0.3544, -0.364, -0.363, -0.376, -0.354, -0.3389, -0.3317, -0.348, -0.3452, -0.314, -0.3055, -0.3173, -0.3322, -0.3454, -0.2877, -0.2837, -0.2854, -0.2854, -0.2773, -0.2697, -0.2198, -0.1918, -0.1903, -0.1895, -0.2003, -0.2329, -0.2193, -0.2228, -0.2228, -0.2403, -0.2233, -0.22, -0.1876, -0.1748, -0.1758, -0.123, -0.0978, -0.096, -0.1153, -0.1152, -0.128, -0.1059, -0.0863, -0.0618, -0.0213, 0.0, 0.0117, -0.0064, -0.0323, -0.0037, -0.0176, -0.0807, -0.0284, -0.0264, -0.0291, -0.014, 0.0393, 0.0198, 0.0004, -0.0073, -0.0614, -0.069, -0.0774, -0.0546, -0.0794, -0.061, -0.0581, -0.0499, -0.0835, -0.0835, -0.0852, -0.0863, -0.0882, -0.1482, -0.1467, -0.1218, -0.1009, -0.1227, -0.1063, -0.0917, -0.0957, -0.103, -0.11, -0.113, -0.0995, -0.0932, -0.0933, -0.0911, -0.1013, -0.0906, -0.0882, -0.0757, -0.0435, -0.072, 0.1481, 0.192, 0.3275, 0.3125, 0.2111, 0.2197, 0.2291, 0.3447, 0.3219, 0.3136, 0.3557, 0.3415, 0.2975, 0.3243, 0.4254, 0.4634, 0.4541, 0.4897, 0.4362, 0.4434, 0.4633, 0.4092, 0.4446, 0.3396, 0.3162, 0.375, 0.3386, 0.459, 0.3974, 0.3909, 0.3555, 0.2978, 0.2599, 0.1611, 0.1484, 0.212, 0.1617, 0.2074, 0.2074, 0.2259, 0.2385, 0.213, 0.2263, 0.2172, 0.2284, 0.2461, 0.249, 0.2479, 0.2479, 0.1879, 0.1699, 0.1788, 0.1165, 0.1331, 0.2028, 0.2114, 0.2111, 0.2119, 0.2119, 0.2116, 0.2151, 0.2136, 0.2069, 0.2069, 0.2594, 0.2459, 0.208, 0.1836, 0.1535, 0.145, 0.1705, 0.2453, 0.2601, 0.2845, 0.3065, 0.3065, 0.307, 0.4078]}
{"unit_id": "quote:1956135589042160075", "ticker_idx": 0, "ts": "2025-08-14T23:27:22+00:00", "symbol": "Intel", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think Intel's 18A will have a hard time succeeding. Especially from what I've heard from people at Intel, 18A can't even make chips with large dies like AI GPUs. When the die size gets bigger, the process errors become so severe that it becomes difficult to manufacture.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Intel foundry: 18A process has fatal yield issues with large dies, undermining its ability to win custom ASIC customers.", "resolved_tickers": ["INTC"], "resolution_method": "exchange_ticker", "resolution_reason": "Intel Corporation INTC US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I think Intel’s 18A will have a hard time succeeding.\n\nEspecially from what I’ve heard from people at Intel, 18A can’t even make chips with large dies like AI GPUs.\n\nWhen the die size gets bigger, the process errors become so severe that it becomes difficult to manufacture.", "reply_text": null, "parent_text": null, "quoted_text": "My intiuition still says, @Intel foundries best customer option is the custom ASIC, both custom CPUs and XPUs, from the likes of Google, Meta, Amazon, Microsoft etc.,  Those volumes, depending on year are in the 5-6 million range and $10-15b in foundry revenue.", "parent_author": null, "ret_m1d": -0.06873433591973854, "ret_signed_m1d": 0.06873433591973854, "alpha_spy_m1d": 0.06864137910667223, "alpha_c_m1d": 0.06661461648598932, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02933775514184722, "ret_signed_p1d": -0.02933775514184722, "alpha_spy_p1d": -0.03167894978386998, "alpha_c_p1d": -0.050104840905821235, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.015088038606142273, "ret_signed_p1w": 0.015088038606142273, "alpha_spy_p1w": 0.0005132883488409012, "alpha_c_p1w": -0.03280549400880228, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.009220423626437269, "ret_signed_p1m": -0.009220423626437269, "alpha_spy_p1m": 0.010098941625504398, "alpha_c_p1m": -0.0012712222676916074, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.5875943042382468, "ret_signed_p3m": -0.5875943042382468, "alpha_spy_p3m": -0.5256614934990818, "alpha_c_p3m": -0.424869082781, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 1.1202849479659958, "ret_signed_p6m": -1.1202849479659958, "alpha_spy_p6m": -1.043331595107278, "alpha_c_p6m": -0.7858262551188777, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.1039, -0.1085, -0.1329, -0.1387, -0.1597, -0.1597, -0.1387, -0.1463, -0.1513, -0.1806, -0.1727, -0.1496, -0.1513, -0.1622, -0.1593, -0.1417, -0.0746, -0.1333, -0.1295, -0.1559, -0.1308, -0.1282, -0.0993, -0.0993, -0.1165, -0.1119, -0.0549, -0.0696, -0.057, -0.049, -0.0612, -0.0423, -0.083, -0.0574, -0.0574, -0.078, -0.0113, -0.0176, -0.0017, -0.018, -0.0235, -0.0394, -0.049, -0.0444, -0.0319, -0.0251, -0.026, -0.0155, -0.0516, -0.1324, -0.1333, -0.1446, -0.1475, -0.1702, -0.1907, -0.1827, -0.1538, -0.1446, -0.1714, -0.1639, -0.1345, -0.0859, -0.0687, 0.0, 0.0293, -0.0084, 0.0608, -0.0134, -0.0151, 0.0394, 0.0289, 0.0205, 0.0415, 0.0448, 0.0205, 0.0205, 0.0147, 0.0059, 0.0314, 0.0264, 0.026, 0.0243, 0.0381, 0.0314, 0.0092, 0.0381, 0.0591, 0.0436, 0.2812, 0.2397, 0.2054, 0.2297, 0.3085, 0.4246, 0.4878, 0.4451, 0.4061, 0.5063, 0.5633, 0.5436, 0.5335, 0.5578, 0.5687, 0.5842, 0.5243, 0.5599, 0.4933, 0.557, 0.544, 0.5511, 0.5968, 0.5977, 0.5474, 0.5993, 0.6044, 0.6572, 0.7406, 0.7326, 0.6832, 0.676, 0.6555, 0.552, 0.6085, 0.5608, 0.5981, 0.6115, 0.5876, 0.588, 0.505, 0.4887, 0.4547, 0.4388, 0.4715, 0.4091, 0.4459, 0.5, 0.5017, 0.5427, 0.5427, 0.6999, 0.6769, 0.8219, 0.834, 0.6974, 0.7355, 0.689, 0.6974, 0.7091, 0.6559, 0.5847, 0.5721, 0.5637, 0.5109, 0.5205, 0.5432, 0.5243, 0.5235, 0.5155, 0.5155, 0.5172, 0.5373, 0.5633, 0.5465, 0.5465, 0.6505, 0.65, 0.6781, 0.7867, 0.723, 0.9091, 0.8466, 0.982, 1.0419, 1.0251, 0.9681, 0.9681, 1.0352, 1.2737, 1.2766, 0.8889, 0.7808, 0.8412, 1.0444, 1.0394, 0.9476, 1.0457, 1.0641, 1.0369, 1.0218, 1.1203]}
{"unit_id": "thread:1967391797501678001", "ticker_idx": 2, "ts": "2025-09-15T00:55:31+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Global leaders such as Besi and ASMPT… conducted preliminary tests in hopes of supplying sample equipment to SK Hynix, but have been unable to advance further", "tweet_type": "analysis", "timeline": "months", "summary": "Hanwha Semitec leads hybrid bonder race as sole SK Hynix sample supplier; Hanmi defending HBM turf via alliances; Besi/ASMPT locked out of SK Hynix memory hybrid bonder so far.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Hanwha Semitec is the only company to supply hybrid bonder samples to SK Hynix”… Hanwha Semitec takes the lead\n\nBeating out global semiconductor equipment giants, Hanwha Semitec has been carrying out exclusive R&D with SK Hynix, solidifying its position in the hybrid bonder market — a technology expected to become critical in high bandwidth memory (HBM) manufacturing. The company plans to reflect the expertise gained from joint development into its second-generation product, set for release early next year, and aims to secure an edge in upcoming mass production qualification tests. In response, Hanmi Semiconductor (042700), eager to defend its foothold in the HBM equipment market, has been building alliances with equipment makers such as TES (095610) and HPSP (403870). LG Electronics (066570), also expanding into the equipment field, has teamed up with domestic player Justem (417840), intensifying the three-way competition to capture the next-generation bonder market.\n\nAccording to industry sources on the 15th, Hanwha Semitec — a subsidiary of Hanwha Vision (489790) — is currently the only company to have supplied hybrid bonder sample equipment to SK Hynix. Since 2022, SK Hynix and Hanwha Semitec have been testing the tool’s performance, safety, and areas for improvement. The second-generation hybrid bonder, scheduled for launch in early 2026, incorporates the improvements identified in collaboration with SK Hynix. Hanwha Semitec also intends to supply samples of this second-generation equipment to SK Hynix and carry the process through to full-scale mass production quality verification.\n\nThe hybrid bonder is the next-generation HBM bonding tool that succeeds the TC bonder. While the TC bonder uses solder balls (tiny spheres of solder for electrical connections) to bond chips, the hybrid bonder directly bonds chips without solder balls. Hybrid bonding boosts data transfer speeds and allows more compact semiconductor packaging. With each HBM generation requiring higher stack heights, hybrid bonding is increasingly viewed as indispensable technology.\n\nThough Hanwha Semitec was a latecomer in the TC bonder race, it is now seen as moving ahead in the emerging hybrid bonder market. Global leaders such as Besi and ASMPT, which dominate the hybrid bonder market in non-memory areas like CMOS image sensors (CIS), also conducted preliminary tests in hopes of supplying sample equipment to SK Hynix, but have been unable to advance further.\n\nAn industry insider commented, “Hanwha Semitec is widely regarded as having the leading edge in hybrid bonder performance. By early next year, SK Hynix’s preferred equipment supplier should become clearer.”\n\nHanmi Semiconductor, which has led the TC bonder market, is also accelerating efforts to secure hybrid bonder capabilities to maintain leadership in the next-generation segment. Although its product launch timing is seen as somewhat late, the company has recently revealed partnerships with domestic equipment maker TES and is reportedly pursuing collaboration with HPSP as well. Unlike TC bonders, which primarily required back-end semiconductor process expertise, developing a hybrid bonder also demands front-end process capabilities such as cleaning, thin-film deposition, and annealing. Hanmi is aiming to accelerate development timelines by partnering with these companies.\n\nLG Electronics, which has entered the bonding equipment market, is likewise pushing ahead in technology development by working with Justem on government-funded hybrid bonding equipment projects. ASMPT and Besi are also stepping up efforts to extend their hybrid bonder expertise from system semiconductors into the memory sector.\n\n---\n\nhttps://t.co/NvKtwXScFU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05332202492433302, "ret_signed_m1d": 0.05332202492433302, "alpha_spy_m1d": 0.04802636000084637, "alpha_c_m1d": 0.04404378691155164, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013118928428609844, "ret_signed_p1d": 0.013118928428609844, "alpha_spy_p1d": 0.011742135643328155, "alpha_c_p1d": 0.013249084271795053, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.03935678528582942, "ret_signed_p1w": -0.03935678528582942, "alpha_spy_p1w": -0.027586943911767303, "alpha_c_p1w": 0.009476072586680706, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.18916628078843667, "ret_signed_p1m": -0.18916628078843667, "alpha_spy_p1m": -0.18439104595925793, "alpha_c_p1m": -0.10344829342033024, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.15869647799636333, "ret_signed_p3m": -0.15869647799636333, "alpha_spy_p3m": -0.11304624632208249, "alpha_c_p3m": 0.048745625824348604, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 0.4777825762560579, "ret_signed_p6m": -0.4777825762560579, "alpha_spy_p6m": -0.4472882412419439, "alpha_c_p6m": -0.18023954186093127, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [0.0643, 0.0554, 0.0423, 0.0639, 0.091, 0.0935, 0.0986, 0.1041, 0.0753, 0.0339, 0.0432, 0.0322, 0.0245, 0.0275, 0.0343, 0.0322, 0.0783, 0.0762, 0.055, 0.069, 0.0152, 0.0791, 0.0829, 0.1016, 0.0647, 0.0466, 0.0749, -0.0089, 0.0394, 0.0157, 0.036, 0.008, -0.0085, 0.0021, -0.0241, -0.0284, 0.0152, 0.022, 0.0288, 0.0474, 0.0453, 0.0381, 0.0034, -0.003, 0.0106, -0.0089, -0.0229, 0.0072, 0.0085, 0.0216, 0.014, 0.0135, -0.0258, -0.0224, -0.0952, -0.1045, -0.0931, -0.08, -0.0618, -0.0567, -0.0491, -0.0381, -0.0533, 0.0, -0.0131, -0.0212, 0.0563, 0.0284, 0.0394, 0.0741, 0.088, 0.0724, 0.036, 0.0698, 0.0736, 0.0817, 0.1303, 0.1045, 0.2421, 0.2425, 0.2387, 0.2306, 0.1875, 0.2082, 0.1892, 0.2315, 0.2281, 0.1993, 0.2289, 0.2213, 0.1722, 0.2209, 0.2446, 0.2526, 0.2408, 0.2497, 0.2446, 0.2493, 0.234, 0.2065, 0.1976, 0.1659, 0.1354, 0.1549, 0.1663, 0.1456, 0.1325, 0.1121, 0.102, 0.0762, 0.0927, 0.1007, 0.0394, 0.0529, 0.0546, 0.0969, 0.0994, 0.0994, 0.0982, 0.1117, 0.1507, 0.1638, 0.1811, 0.2429, 0.2129, 0.1887, 0.1587, 0.1181, 0.121, 0.1308, 0.0944, 0.1126, 0.1066, 0.1105, 0.1164, 0.1134, 0.1134, 0.1134, 0.1181, 0.1265, 0.132, 0.132, 0.262, 0.3013, 0.3618, 0.3441, 0.2789, 0.2789, 0.3724, 0.4152, 0.3669, 0.4697, 0.4634, 0.4287, 0.4634, 0.4731, 0.4812, 0.4816, 0.4799, 0.4875, 0.4342, 0.3758, 0.3923, 0.3931, 0.3555, 0.3335, 0.3606, 0.4008, 0.4168, 0.4308, 0.4465, 0.4278, 0.4913, 0.5108, 0.5243, 0.5827, 0.4642, 0.5616, 0.5882, 0.6276, 0.6725, 0.5988, 0.603, 0.5954, 0.5349, 0.6166, 0.5967, 0.3229, 0.4075, 0.4778]}
{"unit_id": "orig:1963460124376182933", "ticker_idx": 0, "ts": "2025-09-04T04:32:27+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM expects Ironwood to generate $9 billion in revenue for Broadcom within the next 6–7 months and to contribute over $15 billion in total lifetime revenue", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares JPM's bullish Ironwood revenue forecast for Broadcom: $9B in 6-7 months, $15B+ lifetime.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM expects Ironwood to generate $9 billion in revenue for Broadcom within the next 6–7 months and to contribute over $15 billion in total lifetime revenue.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012120219145787092, "ret_signed_m1d": -0.012120219145787092, "alpha_spy_m1d": -0.003832125291686861, "alpha_c_m1d": -0.00028698260536097475, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.011833236540426118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0940542281442045, "ret_signed_p1d": 0.0940542281442045, "alpha_spy_p1d": 0.09695037663273964, "alpha_c_p1d": 0.0822553153297425, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.011798912814461993, "ret_p1w": 0.17487734740507133, "ret_signed_p1w": 0.17487734740507133, "alpha_spy_p1w": 0.16176711226724771, "alpha_c_p1w": 0.12633662853529137, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.04854071886977995, "ret_p1m": 0.1073168217624092, "ret_signed_p1m": 0.1073168217624092, "alpha_spy_p1m": 0.07350882554024163, "alpha_c_p1m": -0.054209470898153755, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.16152629266056295, "ret_p3m": 0.2486890969666664, "ret_signed_p3m": 0.2486890969666664, "alpha_spy_p3m": 0.1958489427051786, "alpha_c_p3m": 0.008676588344810554, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.24001250862185586, "ret_p6m": 0.047729253740592004, "ret_signed_p6m": 0.047729253740592004, "alpha_spy_p6m": -0.015132112897848327, "alpha_c_p6m": -0.35858582602405065, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.40631507976464265, "price_path": [-0.2038, -0.2027, -0.1757, -0.1654, -0.1894, -0.1783, -0.1872, -0.1811, -0.1811, -0.1833, -0.171, -0.1383, -0.1354, -0.1174, -0.1201, -0.0995, -0.1351, -0.1183, -0.101, -0.101, -0.1043, -0.1121, -0.0921, -0.1003, -0.1036, -0.0996, -0.0822, -0.0826, -0.0642, -0.0744, -0.0584, -0.0899, -0.0732, -0.0568, -0.052, -0.0385, -0.0284, -0.0114, -0.0405, -0.057, -0.0274, -0.043, -0.0145, -0.0076, -0.0037, -0.0072, 0.022, 0.0098, 0.0168, 0.0008, -0.0011, -0.0366, -0.0488, -0.0539, -0.0395, -0.0388, -0.0264, -0.0191, 0.0083, -0.0285, -0.0285, -0.0257, -0.0121, 0.0, 0.0941, 0.1292, 0.0999, 0.2074, 0.1749, 0.1757, 0.1894, 0.1761, 0.1309, 0.1282, 0.1269, 0.1087, 0.1092, 0.1104, 0.0999, 0.0948, 0.0731, 0.0796, 0.091, 0.1067, 0.1073, 0.0979, 0.1009, 0.1306, 0.1291, 0.0624, 0.1673, 0.1262, 0.1497, 0.159, 0.1432, 0.1429, 0.1214, 0.1136, 0.1267, 0.1589, 0.1848, 0.2205, 0.2631, 0.232, 0.2096, 0.1864, 0.1517, 0.1748, 0.1637, 0.1435, 0.1728, 0.1518, 0.1625, 0.1126, 0.1207, 0.1213, 0.1143, 0.1598, 0.135, 0.1133, 0.2369, 0.26, 0.301, 0.301, 0.3187, 0.2634, 0.2487, 0.2455, 0.2469, 0.2771, 0.3126, 0.3296, 0.3514, 0.3298, 0.1779, 0.112, 0.1169, 0.0669, 0.0795, 0.1138, 0.1195, 0.1453, 0.1483, 0.1483, 0.1546, 0.1456, 0.1471, 0.1348, 0.1348, 0.1398, 0.126, 0.1271, 0.1263, 0.0901, 0.1311, 0.1548, 0.1627, 0.1144, 0.1247, 0.1532, 0.1532, 0.0905, 0.0781, 0.0672, 0.0494, 0.0651, 0.0911, 0.0926, 0.0844, 0.0863, 0.0856, 0.0503, 0.01, 0.0181, 0.0916, 0.1277, 0.1162, 0.1238, 0.0858, 0.0662, 0.0662, 0.0903, 0.0935, 0.0951, 0.0907, 0.0831, 0.0672, 0.0896, 0.0548, 0.0477]}
{"unit_id": "thread:1964848916903678385", "ticker_idx": 0, "ts": "2025-09-08T00:31:01+00:00", "symbol": "Xiaomi", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "지금은 샤오미가 더 저렴해보여서요", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Xiaomi is the only real Tesla rival among Chinese EV makers and currently looks cheaper; implied long.", "resolved_tickers": ["1810.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "Xiaomi Corp listed on HKEX as 1810.HK", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "ko", "tweet_text": "@jeff_in_every BYD가 다들 테슬라의 경쟁자라고 ���하지만...저는 BYD보다 샤오미가 유이한 테슬라의 경쟁자라고 봐서요 ㅎㅎ\n\n중국 메이커중에 엔비디아 칩 사서 자율주행 훈련시키면서 차도 꽤 품질 좋게 만드는 메이커는 샤오미가 유일하다고 생각합니다\n\n---\n\n@jeff_in_every 물론 테슬라가 가장 기술력 좋다고 생각하지만...지금은 샤오미가 더 저렴해보여서요", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020091296330978836, "ret_signed_m1d": -0.020091296330978836, "alpha_spy_m1d": -0.017640774232498657, "alpha_c_m1d": -0.012652507280170977, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.007438789050807859, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02648403219980744, "ret_signed_p1d": 0.02648403219980744, "alpha_spy_p1d": 0.024172111502197557, "alpha_c_p1d": 0.024558298372460152, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.0019257338273472868, "ret_p1w": 0.02648403219980744, "ret_signed_p1w": 0.02648403219980744, "alpha_spy_p1w": 0.007865908088656504, "alpha_c_p1w": -0.005574275743821522, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03205830794362896, "ret_p1m": -0.016438384034317965, "ret_signed_p1m": -0.016438384034317965, "alpha_spy_p1m": -0.05056923988421247, "alpha_c_p1m": -0.09655441863782499, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.08011603460350702, "ret_p3m": -0.23324201749340034, "ret_signed_p3m": -0.23324201749340034, "alpha_spy_p3m": -0.29097277388421994, "alpha_c_p3m": -0.33373699958224756, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.10049498208884722, "ret_p6m": -0.42319634842546017, "ret_signed_p6m": -0.42319634842546017, "alpha_spy_p6m": -0.4777593845038506, "alpha_c_p6m": -0.464509957141199, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.04131360871573886, "price_path": [-0.0119, -0.0466, -0.0511, -0.011, -0.0137, -0.0174, -0.0311, -0.0128, 0.0018, 0.0393, 0.0347, 0.0393, 0.0767, 0.095, 0.095, 0.0986, 0.0612, 0.0502, 0.0466, 0.0712, 0.0484, 0.0393, 0.0475, 0.0466, 0.053, 0.0466, 0.0256, 0.0411, 0.0548, 0.0584, 0.0667, 0.0676, 0.0484, 0.0374, 0.01, 0.0, -0.0292, -0.0247, -0.0037, -0.0082, -0.0137, -0.053, -0.0639, -0.0721, -0.0557, -0.0274, -0.0283, -0.0347, -0.0311, -0.0429, -0.0402, -0.0621, -0.0402, -0.0228, -0.0228, -0.0283, -0.0301, -0.0347, -0.0137, 0.0201, -0.0009, -0.0237, -0.0201, 0.0, 0.0265, 0.0037, -0.0018, 0.0073, 0.0265, 0.0311, 0.0566, 0.0384, 0.0356, 0.0247, 0.0128, 0.0393, 0.0858, -0.0018, -0.0219, -0.0137, -0.0137, 0.0192, 0.0046, -0.0164, -0.0164, -0.0164, -0.0256, -0.0493, -0.1036, -0.112, -0.0963, -0.1288, -0.1605, -0.139, -0.1514, -0.1536, -0.1463, -0.1613, -0.1635, -0.1795, -0.1795, -0.1934, -0.211, -0.1832, -0.2069, -0.2088, -0.2066, -0.2285, -0.2263, -0.215, -0.2018, -0.2055, -0.2263, -0.2336, -0.2552, -0.291, -0.3114, -0.3045, -0.2939, -0.2632, -0.2676, -0.2493, -0.2508, -0.2639, -0.2566, -0.2654, -0.2332, -0.2186, -0.2223, -0.2464, -0.2369, -0.2296, -0.2153, -0.2358, -0.253, -0.2471, -0.2658, -0.2595, -0.2731, -0.284, -0.284, -0.284, -0.284, -0.2953, -0.2811, -0.284, -0.284, -0.2643, -0.2815, -0.2921, -0.303, -0.3056, -0.3092, -0.2924, -0.3063, -0.31, -0.3085, -0.3224, -0.3337, -0.352, -0.3531, -0.3563, -0.3381, -0.3567, -0.3505, -0.3366, -0.3311, -0.3516, -0.3596, -0.368, -0.3797, -0.3622, -0.3574, -0.3571, -0.3501, -0.3224, -0.333, -0.3271, -0.3333, -0.3333, -0.3333, -0.3333, -0.3542, -0.3322, -0.3472, -0.3498, -0.3574, -0.3626, -0.3947, -0.4232]}
{"unit_id": "orig:1962486630822072540", "ticker_idx": 0, "ts": "2025-09-01T12:04:08+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the meeting was to discuss increasing the adoption of Micron's DRAM in the S26. $MU", "tweet_type": "news_reaction", "timeline": "months", "summary": "Rumor that Samsung MX may shift S26 DRAM supply to Micron due to Samsung DS LPDDR5X technical issues; bullish for MU share gains.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Rumor: The reason why Samsung MX Division President TM Roh recently met with Micron’s SJ is that Samsung DS Division’s 1bnm-class LPDDR5X is facing technical issues (such as overheating and performance degradation), and the meeting was to discuss increasing the adoption of Micron’s DRAM in the S26.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004453374548142985, "ret_signed_p1d": -0.004453374548142985, "alpha_spy_p1d": 0.0029569210077312613, "alpha_c_p1d": 0.008671450330633879, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.013124824878776864, "ret_p1w": 0.10461318165995759, "ret_signed_p1w": 0.10461318165995759, "alpha_spy_p1w": 0.09875323020566196, "alpha_c_p1w": 0.0831520329195703, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": 0.021461148740387292, "ret_p1m": 0.40593227727445536, "ret_signed_p1m": 0.40593227727445536, "alpha_spy_p1m": 0.37031186433729335, "alpha_c_p1m": 0.28167738281751853, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.12425489445693683, "ret_p3m": 0.9360069202668313, "ret_signed_p3m": 0.9360069202668313, "alpha_spy_p3m": 0.8793998464723796, "alpha_c_p3m": 0.7382048557319516, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.1978020645348797, "ret_p6m": 2.5160132606775507, "ret_signed_p6m": 2.5160132606775507, "alpha_spy_p6m": 2.44432537613911, "alpha_c_p6m": 2.067585030667101, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.4484282300104496, "price_path": [-0.1332, -0.1077, -0.0887, -0.0686, -0.0418, -0.026, -0.0247, -0.0296, 0.006, 0.0102, 0.0226, 0.0226, 0.0376, 0.0248, 0.0738, 0.0682, 0.0577, 0.0473, 0.0347, 0.0148, 0.022, 0.0266, 0.0266, 0.0076, 0.0455, 0.0271, 0.0345, 0.0464, -0.0034, 0.0092, -0.0217, -0.0483, -0.0388, -0.0486, -0.0823, -0.0771, -0.0612, -0.0651, -0.0652, -0.0592, -0.0359, -0.0829, -0.1187, -0.0944, -0.0836, -0.086, -0.06, -0.001, 0.0396, 0.0734, 0.0442, 0.0528, 0.0156, 0.0381, 0.0255, -0.0151, -0.0271, -0.0112, -0.0218, -0.0211, -0.0106, 0.0251, 0.0, 0.0, -0.0045, -0.0024, 0.0437, 0.1039, 0.1046, 0.1364, 0.1764, 0.2652, 0.3211, 0.3257, 0.3345, 0.3443, 0.4191, 0.3674, 0.3832, 0.3983, 0.3588, 0.3178, 0.3215, 0.3772, 0.4059, 0.5305, 0.544, 0.5793, 0.6056, 0.5613, 0.6525, 0.6171, 0.5269, 0.6208, 0.5728, 0.6138, 0.7029, 0.7016, 0.7385, 0.7008, 0.6687, 0.738, 0.8415, 0.8506, 0.8658, 0.9055, 0.8835, 0.8814, 0.9733, 0.8332, 0.9969, 1.0039, 1.0004, 1.1297, 1.0272, 1.0591, 0.9923, 1.0753, 1.0343, 0.9212, 0.8995, 0.6931, 0.7435, 0.8828, 0.8878, 0.936, 0.936, 0.9883, 1.0218, 1.0136, 0.9688, 0.9057, 0.9945, 1.0761, 1.1223, 1.2173, 1.1731, 1.0275, 0.9969, 0.9549, 0.8962, 1.0898, 1.2358, 1.3255, 1.3229, 1.4104, 1.4104, 1.3945, 1.476, 1.4614, 1.4007, 1.4007, 1.6531, 1.6256, 1.8887, 1.8561, 1.7507, 1.9027, 1.9092, 1.8441, 1.8039, 1.8315, 2.0512, 2.0512, 2.0701, 2.2729, 2.3442, 2.3616, 2.2728, 2.4507, 2.6613, 2.6656, 2.4897, 2.6825, 2.528, 2.1913, 2.2206, 2.3199, 2.2257, 2.1395, 2.4515, 2.482, 2.4626, 2.4626, 2.3627, 2.5407, 2.5105, 2.6015, 2.5409, 2.516]}
{"unit_id": "quote:1970982114708554147", "ticker_idx": 0, "ts": "2025-09-24T22:42:09+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Structural demand expansion will emerge as dependency on pure HBM gives way to the combined growth of SOCAMM (server CPU memory) and Edge AI demand", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish global DRAM: 4-year memory supercycle 2024–2027, AI-centric demand reshaping TAM via HBM, SOCAMM, and Edge AI to 53% AI share by 2027.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Changes in DRAM TAM Share from HBM, SOCAMM, and Edge AI\n\n• In 2023, AI DRAM TAM was essentially 100% HBM. By 2025, it will account for 33% of total DRAM, and by 2027, it is expected to expand to 53%.\n    \n• Structural demand expansion will emerge as dependency on pure HBM gives way to the combined growth of SOCAMM (server CPU memory) and Edge AI demand.\n    \n• The structure of DRAM demand will rapidly be reshaped into an “AI-centric” model.", "reply_text": null, "parent_text": null, "quoted_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra https://t.co/IL9PiWxb6i", "parent_author": null, "ret_m1d": 0.010219261695896894, "ret_signed_m1d": 0.010219261695896894, "alpha_spy_m1d": 0.007027473142046511, "alpha_c_m1d": 0.008381803965575604, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008259317884715492, "ret_signed_p1d": -0.008259317884715492, "alpha_spy_p1d": -0.0036458203962434297, "alpha_c_p1d": -0.008103638291812413, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.04830357054726327, "ret_signed_p1w": 0.04830357054726327, "alpha_spy_p1w": 0.03718560612396956, "alpha_c_p1w": 0.009126939226759966, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.25085205846157504, "ret_signed_p1m": 0.25085205846157504, "alpha_spy_p1m": 0.2347273796390122, "alpha_c_p1m": 0.1770767101201443, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 0.5449006934175141, "ret_signed_p3m": 0.5449006934175141, "alpha_spy_p3m": 0.5059449215457171, "alpha_c_p3m": 0.42123818656422773, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 1.6535661009895588, "ret_signed_p6m": 1.6535661009895588, "alpha_spy_p6m": 1.65258338368522, "alpha_c_p6m": 1.4198802593573656, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.2413, -0.2409, -0.2503, -0.2522, -0.2398, -0.2492, -0.2597, -0.2413, -0.2506, -0.2316, -0.2248, -0.2322, -0.2259, -0.2319, -0.2553, -0.2518, -0.2482, -0.2673, -0.264, -0.2612, -0.2658, -0.252, -0.2492, -0.2348, -0.2417, -0.2747, -0.2656, -0.257, -0.2666, -0.2503, -0.2359, -0.2193, -0.2087, -0.2044, -0.2048, -0.2139, -0.223, -0.2303, -0.2453, -0.2576, -0.245, -0.2393, -0.242, -0.2396, -0.2265, -0.2318, -0.2521, -0.2432, -0.2381, -0.2228, -0.2029, -0.1971, -0.1736, -0.1446, -0.1169, -0.0753, -0.0676, -0.0382, -0.054, -0.007, -0.0197, -0.0075, 0.0102, 0.0, -0.0083, -0.0369, -0.0066, -0.0022, 0.0483, 0.0994, 0.1081, 0.1145, 0.1036, 0.126, 0.1173, 0.1437, 0.1503, 0.1286, 0.1623, 0.2227, 0.2353, 0.2638, 0.2461, 0.2449, 0.2509, 0.3146, 0.3527, 0.3336, 0.3817, 0.3998, 0.4042, 0.4978, 0.4074, 0.4241, 0.4334, 0.4154, 0.4819, 0.4803, 0.4846, 0.4624, 0.4123, 0.4585, 0.3862, 0.3683, 0.3422, 0.2852, 0.3258, 0.3363, 0.3666, 0.3883, 0.3763, 0.3932, 0.42, 0.4077, 0.3853, 0.4219, 0.477, 0.4738, 0.5151, 0.481, 0.4571, 0.4177, 0.3771, 0.4023, 0.4496, 0.4756, 0.5449, 0.5519, 0.5755, 0.5755, 0.605, 0.6754, 0.6836, 0.6687, 0.6687, 0.7888, 0.8376, 0.9333, 0.9485, 0.927, 0.9539, 0.9594, 0.9284, 0.9329, 0.9604, 1.0405, 1.0496, 1.0185, 1.0824, 1.1249, 1.1395, 1.0888, 1.2213, 1.3227, 1.3357, 1.3366, 1.2703, 1.3717, 1.2888, 1.2033, 1.2221, 1.2745, 1.2407, 1.3102, 1.3864, 1.3844, 1.3844, 1.3599, 1.4036, 1.4439, 1.5179, 1.5163, 1.5882, 1.6368, 1.7437, 1.6956, 1.6963, 1.4298, 1.299, 1.4538, 1.3693, 1.2684, 1.4489, 1.5052, 1.446, 1.4529, 1.5655, 1.6233, 1.7613, 1.6536]}
{"unit_id": "reply:1962529979209859228", "ticker_idx": 0, "ts": "2025-09-01T14:56:23+00:00", "symbol": "Korean defense", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Korean defense stocks are still trading at attractive valuations", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long Korean defense stocks; still at attractive valuations.", "resolved_tickers": ["047810.KS", "012450.KS", "064350.KS"], "resolution_method": "basket", "resolution_reason": "Korea defense basket: Korean Air, Hanwha Aerospace, Hyundai Rotem", "bench_c": "XLI", "bench_c_reason": "industry='Aerospace & Defense'", "bench_c_industry": "Aerospace & Defense", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@stevehou I agree. Korean defense stocks are still trading at attractive valuations. https://t.co/oNHGIgxXZW", "reply_text": null, "parent_text": "Korean defense and Asian defense in general is under-appreciated as an investment theme.", "quoted_text": null, "parent_author": "stevehou", "ret_m1d": -0.00541062137692605, "ret_signed_m1d": -0.00541062137692605, "alpha_spy_m1d": -0.00541062137692605, "alpha_c_m1d": -0.00541062137692605, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.030469587941108916, "ret_signed_p1d": 0.030469587941108916, "alpha_spy_p1d": 0.037879883496983166, "alpha_c_p1d": 0.04000844206963587, "bench_spy_p1d": -0.007410295555874247, "bench_c_p1d": -0.009538854128526952, "ret_p1w": 0.057688790206237885, "ret_signed_p1w": 0.057688790206237885, "alpha_spy_p1w": 0.051828838751942254, "alpha_c_p1w": 0.06255675097937499, "bench_spy_p1w": 0.005859951454295631, "bench_c_p1w": -0.004867960773137114, "ret_p1m": 0.16807350673992158, "ret_signed_p1m": 0.16807350673992158, "alpha_spy_p1m": 0.13245309380275958, "alpha_c_p1m": 0.14927745376803733, "bench_spy_p1m": 0.035620412937162005, "bench_c_p1m": 0.018796052971884247, "ret_p3m": 0.013392968465257874, "ret_signed_p3m": 0.013392968465257874, "alpha_spy_p3m": -0.04321410532919386, "alpha_c_p3m": 0.004042966386059757, "bench_spy_p3m": 0.056607073794451734, "bench_c_p3m": 0.009350002079198116, "ret_p6m": 0.49435824438080084, "ret_signed_p6m": 0.49435824438080084, "alpha_spy_p6m": 0.4226703598423605, "alpha_c_p6m": 0.32127848532437137, "bench_spy_p6m": 0.07168788453844033, "bench_c_p6m": 0.17307975905642947, "price_path": [-0.1106, -0.0827, -0.0827, -0.1115, -0.0471, -0.0606, -0.0186, 0.0165, 0.05, 0.0349, 0.0239, 0.0548, 0.0477, 0.0446, -0.0023, -0.035, -0.0014, -0.002, -0.0301, -0.0256, -0.0308, -0.0453, -0.101, -0.0819, -0.0763, -0.0411, -0.0439, -0.0796, -0.0703, -0.0384, -0.0406, -0.009, -0.0143, 0.0075, -0.0087, -0.0059, 0.0009, 0.0162, 0.0358, 0.0791, 0.037, 0.05, 0.0026, 0.0194, 0.0359, 0.0224, 0.0074, -0.0371, -0.0438, -0.066, -0.0474, -0.0426, -0.0426, -0.0426, -0.0879, -0.093, -0.0775, -0.0321, -0.0177, -0.0177, -0.0206, -0.0002, -0.0054, 0.0, 0.0305, 0.0398, 0.035, 0.0272, 0.0577, 0.0432, 0.059, 0.0983, 0.0912, 0.08, 0.1232, 0.152, 0.1209, 0.1209, 0.1353, 0.1291, 0.1694, 0.1511, 0.1307, 0.1474, 0.1681, 0.1666, 0.1739, 0.1739, 0.1739, 0.1739, 0.1739, 0.1739, 0.1426, 0.0874, 0.0417, 0.0743, 0.0519, 0.0225, 0.0607, 0.0799, 0.1322, 0.1621, 0.138, 0.1576, 0.1453, 0.1262, 0.1397, 0.1251, 0.1919, 0.1522, 0.0746, 0.0729, 0.0267, 0.0731, 0.0494, 0.0595, 0.0552, 0.0566, 0.0944, 0.0614, 0.0344, 0.0457, 0.0136, -0.0146, 0.0074, 0.0216, 0.0134, 0.0033, -0.0269, -0.0301, 0.0008, -0.0068, 0.0203, 0.0666, 0.0716, 0.0495, 0.0308, 0.0837, 0.0475, 0.0205, 0.0098, 0.0035, 0.0209, 0.043, 0.0608, 0.044, 0.044, 0.0213, 0.0962, 0.0793, 0.0793, 0.0793, 0.0992, 0.1599, 0.2186, 0.2073, 0.2753, 0.3596, 0.3718, 0.4339, 0.4031, 0.4497, 0.4056, 0.4512, 0.4503, 0.4419, 0.4221, 0.4014, 0.4296, 0.4031, 0.4543, 0.4723, 0.4771, 0.3973, 0.4456, 0.4928, 0.387, 0.3492, 0.3863, 0.3427, 0.3451, 0.3626, 0.3851, 0.3851, 0.3851, 0.3851, 0.4114, 0.4718, 0.464, 0.4944]}
{"unit_id": "orig:1958711200889589957", "ticker_idx": 0, "ts": "2025-08-22T02:01:56+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix has established itself as NVIDIA's main partner starting from the HBM3e generation and is ahead in both technology and production for HBM4. It is scheduled to be the priority supplier for NVIDIA's next-generation Rubin GPU, securing market trust and a basis for market share expansion.", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA note: SK Hynix dominant in HBM4 (60% share, Rubin priority supplier); Samsung delayed and losing NVIDIA share; Micron benefits from 2027 supply tightness; HBM4 price premium 20-40%+ supports DRAM re-rating.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Memory Tech (BofA)\n\n1. HBM Competitive Landscape\n\n1. HBM4 Mass Production Timing and Market Share Outlook\n- SK Hynix to begin early mass production of HBM4 for NVIDIA’s next-generation GPU Rubin in 4Q25, while Samsung is expected in 1Q–2Q26.\n\n- 2025 will remain centered on HBM3e (led especially by SK Hynix), with full-scale HBM4 competition beginning in 2026.\n\n- 2026 market share outlook: SK Hynix ~60%, Samsung ~20%, Micron the remainder.\n\n2. Price Premium Structure\n- Initial HBM4 shipments (4Q25–1H26) could carry more than a 40% price premium over HBM3e (current estimate 30%).\n\n- Even during large-scale shipments in 2H26, a premium of 20%+ is expected (higher than the previously expected 15–20%).\n\n- In 2027, with Rubin GPU memory capacity increasing (288GB → 384GB/1TB), potential supply shortages could maintain the price premium.\n\n3. Samsung Electronics Situation\n- Samsung’s large-scale mass production of HBM4 is delayed, and media reports of production starting at the end of 2025 (Nov–Dec) are considered lacking in basis.\n\n- Samsung’s 12-layer HBM3e production volume is focused on secondary customers rather than NVIDIA as a main customer, indicating delays in obtaining full qualification from NVIDIA.\n\n- Current 12-Hi HBM3e capacity expansion at Samsung is more centered on non-NVIDIA customers.\n\n4. SK Hynix Advantage Factors\n- SK Hynix has established itself as NVIDIA’s main partner starting from the HBM3e generation and is ahead in both technology and production for HBM4.\n\n- It is scheduled to be the priority supplier for NVIDIA’s next-generation Rubin GPU, securing market trust and a basis for market share expansion.\n\n- Based on the 1c node, DDR5 will also be launched in the market starting 4Q25, enabling reliable customer response in both HBM and DDR5.\n\n5. Structural Characteristics of the Market\n- HBM4 competition is viewed as a “win-win structure”: SK Hynix will maintain its lead, but Samsung will also secure some share.\n\n- With the significant increase in Rubin GPU’s memory demand in 2027, overall HBM demand could exceed supply, benefiting all three major players.\n\n- With price premiums and strong demand, HBM4 is expected to be the key driver for DRAM makers to defend profit margins and for stock re-rating.\n\n2. CHIPS Act and Equity Stakes\n\n- Samsung Electronics and SK Hynix have low likelihood of transferring or newly issuing equity stakes to the U.S. government.\n\n- Strong opposition is likely due to shareholder value dilution and concerns over U.S. government intervention in management.\n\n- The unfavorable scenario is if CHIPS Act subsidies are injected directly into U.S. fab assets rather than into the parent company’s equity.\n\n- Samsung holds over KRW 100 trillion in cash, and SK Hynix, with record-high profits, may be able to operate U.S. plants even without subsidies.\n\n3. DRAM and Server Demand\n\n- Beyond HBM competition issues, strong orders for server DDR5 have been confirmed.\n\n- Demand is segmented: 96GB in China, 128–256GB modules in the U.S.\n\n- 4Q server DDR5 bit growth expected at 10%+, with supply limited to below that level → ASP projected to rise 5%+.\n\n- SK Hynix’s 1c DDR5 expected to enter the market in 4Q, while Samsung’s 1c node is focused on HBM4 sampling.\n\n$MU $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023904440996469822, "ret_signed_m1d": -0.023904440996469822, "alpha_spy_m1d": -0.00877982151088852, "alpha_c_m1d": -0.0028581665901818987, "bench_spy_m1d": -0.015124619485581303, "bench_c_m1d": -0.021046274406287924, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03386446325793102, "ret_signed_p1d": 0.03386446325793102, "alpha_spy_p1d": 0.038265620315549476, "alpha_c_p1d": 0.032910870898179345, "bench_spy_p1d": -0.0044011570576184544, "bench_c_p1d": 0.000953592359751676, "ret_p1w": 0.07326097337444981, "ret_signed_p1w": 0.07326097337444981, "alpha_spy_p1w": 0.0736639747954948, "alpha_c_p1w": 0.08466947005230585, "bench_spy_p1w": -0.0004030014210449906, "bench_c_p1w": -0.01140849667785604, "ret_p1m": 0.40042618833488675, "ret_signed_p1m": 0.40042618833488675, "alpha_spy_p1m": 0.3641975528948369, "alpha_c_p1m": 0.3032664173596862, "bench_spy_p1m": 0.03622863544004984, "bench_c_p1m": 0.09715977097520057, "ret_p3m": 1.242277735294524, "ret_signed_p3m": 1.242277735294524, "alpha_spy_p3m": 1.212591076768645, "alpha_c_p3m": 1.0863044492248435, "bench_spy_p3m": 0.02968665852587904, "bench_c_p3m": 0.15597328606968053, "ret_p6m": 2.513554443093718, "ret_signed_p6m": 2.513554443093718, "alpha_spy_p6m": 2.45102617337865, "alpha_c_p6m": 2.120731103923319, "bench_spy_p6m": 0.06252826971506797, "bench_c_p6m": 0.3928233391703986, "price_path": [-0.1947, -0.1728, -0.1554, -0.1853, -0.1733, -0.1733, -0.1335, -0.1056, -0.1056, -0.0876, -0.0817, -0.0438, -0.0618, -0.0618, -0.012, -0.008, -0.0179, -0.0199, 0.0239, 0.0339, 0.1096, 0.1394, 0.1673, 0.1315, 0.1633, 0.1375, 0.1116, 0.1096, 0.0777, 0.0797, 0.1235, 0.1195, 0.1833, 0.1733, 0.1952, 0.1892, 0.1793, 0.0737, 0.0717, 0.0857, 0.0697, 0.0717, 0.0737, 0.0598, 0.0438, 0.0458, 0.0498, 0.0896, 0.0279, 0.0279, 0.0498, 0.0299, 0.0438, 0.0219, 0.0637, 0.0717, 0.1076, 0.1016, 0.1016, 0.0657, 0.0478, 0.0179, -0.0239, 0.0, 0.0339, 0.0418, 0.0359, 0.0713, 0.0733, 0.0214, 0.0393, 0.0473, 0.0593, 0.0912, 0.1052, 0.1491, 0.2129, 0.2249, 0.3107, 0.3206, 0.3885, 0.3306, 0.4184, 0.4184, 0.4004, 0.4403, 0.4264, 0.4224, 0.3426, 0.3924, 0.3865, 0.4363, 0.578, 0.578, 0.578, 0.578, 0.578, 0.578, 0.7076, 0.6558, 0.6418, 0.6857, 0.8054, 0.8573, 0.9371, 0.9111, 0.9211, 0.9091, 1.0348, 1.1346, 1.0787, 1.2263, 1.2662, 1.2303, 1.4737, 1.338, 1.3101, 1.366, 1.3141, 1.4178, 1.4697, 1.4617, 1.4418, 1.2343, 1.4178, 1.2742, 1.2423, 1.2782, 1.0787, 1.0747, 1.0707, 1.0907, 1.172, 1.1161, 1.1481, 1.2279, 1.204, 1.164, 1.172, 1.3038, 1.2599, 1.3437, 1.2559, 1.2798, 1.2119, 1.1161, 1.2, 1.204, 1.184, 1.3158, 1.3317, 1.3477, 1.3477, 1.3916, 1.5553, 1.5992, 1.5992, 1.5992, 1.703, 1.7789, 1.8987, 1.9626, 2.0185, 1.9706, 1.9905, 1.9466, 1.9626, 1.9905, 2.0185, 2.0504, 1.9666, 1.9546, 2.0145, 2.0624, 1.9386, 2.1941, 2.3578, 2.4377, 2.6293, 2.3139, 2.6214, 2.5934, 2.3618, 2.3499, 2.5415, 2.4976, 2.4337, 2.5455, 2.5136, 2.5136]}
{"unit_id": "orig:1958711200889589957", "ticker_idx": 1, "ts": "2025-08-22T02:01:56+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's large-scale mass production of HBM4 is delayed, and media reports of production starting at the end of 2025 (Nov–Dec) are considered lacking in basis. Samsung's 12-layer HBM3e production volume is focused on secondary customers rather than NVIDIA", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA note: SK Hynix dominant in HBM4 (60% share, Rubin priority supplier); Samsung delayed and losing NVIDIA share; Micron benefits from 2027 supply tightness; HBM4 price premium 20-40%+ supports DRAM re-rating.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Global Memory Tech (BofA)\n\n1. HBM Competitive Landscape\n\n1. HBM4 Mass Production Timing and Market Share Outlook\n- SK Hynix to begin early mass production of HBM4 for NVIDIA’s next-generation GPU Rubin in 4Q25, while Samsung is expected in 1Q–2Q26.\n\n- 2025 will remain centered on HBM3e (led especially by SK Hynix), with full-scale HBM4 competition beginning in 2026.\n\n- 2026 market share outlook: SK Hynix ~60%, Samsung ~20%, Micron the remainder.\n\n2. Price Premium Structure\n- Initial HBM4 shipments (4Q25–1H26) could carry more than a 40% price premium over HBM3e (current estimate 30%).\n\n- Even during large-scale shipments in 2H26, a premium of 20%+ is expected (higher than the previously expected 15–20%).\n\n- In 2027, with Rubin GPU memory capacity increasing (288GB → 384GB/1TB), potential supply shortages could maintain the price premium.\n\n3. Samsung Electronics Situation\n- Samsung’s large-scale mass production of HBM4 is delayed, and media reports of production starting at the end of 2025 (Nov–Dec) are considered lacking in basis.\n\n- Samsung’s 12-layer HBM3e production volume is focused on secondary customers rather than NVIDIA as a main customer, indicating delays in obtaining full qualification from NVIDIA.\n\n- Current 12-Hi HBM3e capacity expansion at Samsung is more centered on non-NVIDIA customers.\n\n4. SK Hynix Advantage Factors\n- SK Hynix has established itself as NVIDIA’s main partner starting from the HBM3e generation and is ahead in both technology and production for HBM4.\n\n- It is scheduled to be the priority supplier for NVIDIA’s next-generation Rubin GPU, securing market trust and a basis for market share expansion.\n\n- Based on the 1c node, DDR5 will also be launched in the market starting 4Q25, enabling reliable customer response in both HBM and DDR5.\n\n5. Structural Characteristics of the Market\n- HBM4 competition is viewed as a “win-win structure”: SK Hynix will maintain its lead, but Samsung will also secure some share.\n\n- With the significant increase in Rubin GPU’s memory demand in 2027, overall HBM demand could exceed supply, benefiting all three major players.\n\n- With price premiums and strong demand, HBM4 is expected to be the key driver for DRAM makers to defend profit margins and for stock re-rating.\n\n2. CHIPS Act and Equity Stakes\n\n- Samsung Electronics and SK Hynix have low likelihood of transferring or newly issuing equity stakes to the U.S. government.\n\n- Strong opposition is likely due to shareholder value dilution and concerns over U.S. government intervention in management.\n\n- The unfavorable scenario is if CHIPS Act subsidies are injected directly into U.S. fab assets rather than into the parent company’s equity.\n\n- Samsung holds over KRW 100 trillion in cash, and SK Hynix, with record-high profits, may be able to operate U.S. plants even without subsidies.\n\n3. DRAM and Server Demand\n\n- Beyond HBM competition issues, strong orders for server DDR5 have been confirmed.\n\n- Demand is segmented: 96GB in China, 128–256GB modules in the U.S.\n\n- 4Q server DDR5 bit growth expected at 10%+, with supply limited to below that level → ASP projected to rise 5%+.\n\n- SK Hynix’s 1c DDR5 expected to enter the market in 4Q, while Samsung’s 1c node is focused on HBM4 sampling.\n\n$MU $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01120454472261545, "ret_signed_m1d": 0.01120454472261545, "alpha_spy_m1d": -0.003920074762965853, "alpha_c_m1d": -0.0021708827940009057, "bench_spy_m1d": -0.015124619485581303, "bench_c_m1d": -0.013375427516616356, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0014006372589652205, "ret_signed_p1d": -0.0014006372589652205, "alpha_spy_p1d": -0.005801794316583675, "alpha_c_p1d": -0.003649043927261175, "bench_spy_p1d": -0.0044011570576184544, "bench_c_p1d": -0.0022484066682959547, "ret_p1w": -0.023809505364553618, "ret_signed_p1w": 0.023809505364553618, "alpha_spy_p1w": 0.023406503943508628, "alpha_c_p1w": 0.023923934037950567, "bench_spy_p1w": -0.0004030014210449906, "bench_c_p1w": 0.00011442867339694907, "ret_p1m": 0.16946781206980566, "ret_signed_p1m": -0.16946781206980566, "alpha_spy_p1m": -0.13323917662975582, "alpha_c_p1m": -0.09095526338945437, "bench_spy_p1m": 0.03622863544004984, "bench_c_p1m": 0.07851254868035129, "ret_p3m": 0.3575707620070947, "ret_signed_p3m": -0.3575707620070947, "alpha_spy_p3m": -0.32788410348121566, "alpha_c_p3m": -0.28550650260056143, "bench_spy_p3m": 0.02968665852587904, "bench_c_p3m": 0.07206425940653327, "ret_p6m": 1.5615295985883844, "ret_signed_p6m": -1.5615295985883844, "alpha_spy_p6m": -1.4990013288733164, "alpha_c_p6m": -1.4949088746448733, "bench_spy_p6m": 0.06252826971506797, "bench_c_p6m": 0.06662072394351104, "price_path": [-0.2497, -0.2219, -0.2191, -0.2177, -0.2093, -0.2093, -0.1954, -0.1773, -0.1773, -0.1676, -0.1759, -0.1662, -0.1717, -0.1885, -0.2038, -0.1912, -0.1676, -0.1759, -0.1717, -0.1926, -0.1578, -0.1467, -0.162, -0.1485, -0.1625, -0.1569, -0.1485, -0.1064, -0.1134, -0.1359, -0.1401, -0.1541, -0.1457, -0.1232, -0.1246, -0.1078, -0.0938, -0.0658, -0.0602, -0.0504, -0.0756, -0.07, -0.0756, -0.077, -0.014, -0.0112, 0.0168, 0.0, -0.035, -0.0238, -0.021, -0.0364, -0.0126, 0.0056, -0.0056, -0.0042, 0.007, 0.0028, 0.0028, -0.0196, -0.0196, -0.0126, -0.0112, 0.0, 0.0014, -0.0154, -0.0112, -0.0252, -0.0238, -0.0532, -0.0322, -0.0224, -0.0182, -0.0266, -0.0182, 0.0014, 0.0168, 0.028, 0.056, 0.0714, 0.112, 0.0952, 0.1246, 0.1246, 0.1695, 0.1863, 0.1961, 0.2059, 0.1667, 0.1845, 0.1803, 0.2099, 0.2626, 0.2626, 0.2626, 0.2626, 0.2626, 0.2626, 0.328, 0.3126, 0.2886, 0.3365, 0.3745, 0.3773, 0.3801, 0.3716, 0.3871, 0.3576, 0.3899, 0.4349, 0.3998, 0.4138, 0.4645, 0.5123, 0.563, 0.4757, 0.4152, 0.3956, 0.3773, 0.4152, 0.456, 0.4504, 0.4462, 0.3674, 0.4152, 0.3759, 0.3576, 0.4152, 0.3337, 0.3604, 0.397, 0.4462, 0.456, 0.4138, 0.4181, 0.4546, 0.4701, 0.4786, 0.525, 0.5405, 0.525, 0.5194, 0.5095, 0.532, 0.4743, 0.4462, 0.5179, 0.5137, 0.4954, 0.5545, 0.5686, 0.563, 0.563, 0.646, 0.6893, 0.695, 0.695, 0.695, 0.8165, 0.9522, 0.9636, 0.9932, 0.9621, 0.965, 0.9621, 0.9452, 0.9833, 1.0342, 1.1049, 1.1106, 1.0526, 1.1134, 1.153, 1.1502, 1.1502, 1.2548, 1.2958, 1.2717, 1.2689, 1.1261, 1.3679, 1.3905, 1.2519, 1.242, 1.3523, 1.3438, 1.3721, 1.5248, 1.5615, 1.5615]}
{"unit_id": "orig:1958711200889589957", "ticker_idx": 2, "ts": "2025-08-22T02:01:56+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "With the significant increase in Rubin GPU's memory demand in 2027, overall HBM demand could exceed supply, benefiting all three major players", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA note: SK Hynix dominant in HBM4 (60% share, Rubin priority supplier); Samsung delayed and losing NVIDIA share; Micron benefits from 2027 supply tightness; HBM4 price premium 20-40%+ supports DRAM re-rating.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Global Memory Tech (BofA)\n\n1. HBM Competitive Landscape\n\n1. HBM4 Mass Production Timing and Market Share Outlook\n- SK Hynix to begin early mass production of HBM4 for NVIDIA’s next-generation GPU Rubin in 4Q25, while Samsung is expected in 1Q–2Q26.\n\n- 2025 will remain centered on HBM3e (led especially by SK Hynix), with full-scale HBM4 competition beginning in 2026.\n\n- 2026 market share outlook: SK Hynix ~60%, Samsung ~20%, Micron the remainder.\n\n2. Price Premium Structure\n- Initial HBM4 shipments (4Q25–1H26) could carry more than a 40% price premium over HBM3e (current estimate 30%).\n\n- Even during large-scale shipments in 2H26, a premium of 20%+ is expected (higher than the previously expected 15–20%).\n\n- In 2027, with Rubin GPU memory capacity increasing (288GB → 384GB/1TB), potential supply shortages could maintain the price premium.\n\n3. Samsung Electronics Situation\n- Samsung’s large-scale mass production of HBM4 is delayed, and media reports of production starting at the end of 2025 (Nov–Dec) are considered lacking in basis.\n\n- Samsung’s 12-layer HBM3e production volume is focused on secondary customers rather than NVIDIA as a main customer, indicating delays in obtaining full qualification from NVIDIA.\n\n- Current 12-Hi HBM3e capacity expansion at Samsung is more centered on non-NVIDIA customers.\n\n4. SK Hynix Advantage Factors\n- SK Hynix has established itself as NVIDIA’s main partner starting from the HBM3e generation and is ahead in both technology and production for HBM4.\n\n- It is scheduled to be the priority supplier for NVIDIA’s next-generation Rubin GPU, securing market trust and a basis for market share expansion.\n\n- Based on the 1c node, DDR5 will also be launched in the market starting 4Q25, enabling reliable customer response in both HBM and DDR5.\n\n5. Structural Characteristics of the Market\n- HBM4 competition is viewed as a “win-win structure”: SK Hynix will maintain its lead, but Samsung will also secure some share.\n\n- With the significant increase in Rubin GPU’s memory demand in 2027, overall HBM demand could exceed supply, benefiting all three major players.\n\n- With price premiums and strong demand, HBM4 is expected to be the key driver for DRAM makers to defend profit margins and for stock re-rating.\n\n2. CHIPS Act and Equity Stakes\n\n- Samsung Electronics and SK Hynix have low likelihood of transferring or newly issuing equity stakes to the U.S. government.\n\n- Strong opposition is likely due to shareholder value dilution and concerns over U.S. government intervention in management.\n\n- The unfavorable scenario is if CHIPS Act subsidies are injected directly into U.S. fab assets rather than into the parent company’s equity.\n\n- Samsung holds over KRW 100 trillion in cash, and SK Hynix, with record-high profits, may be able to operate U.S. plants even without subsidies.\n\n3. DRAM and Server Demand\n\n- Beyond HBM competition issues, strong orders for server DDR5 have been confirmed.\n\n- Demand is segmented: 96GB in China, 128–256GB modules in the U.S.\n\n- 4Q server DDR5 bit growth expected at 10%+, with supply limited to below that level → ASP projected to rise 5%+.\n\n- SK Hynix’s 1c DDR5 expected to enter the market in 4Q, while Samsung’s 1c node is focused on HBM4 sampling.\n\n$MU $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01606055756302427, "ret_signed_m1d": -0.01606055756302427, "alpha_spy_m1d": -0.000935938077442966, "alpha_c_m1d": 0.0049857168432636545, "bench_spy_m1d": -0.015124619485581303, "bench_c_m1d": -0.021046274406287924, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010707038375349476, "ret_signed_p1d": -0.010707038375349476, "alpha_spy_p1d": -0.006305881317731021, "alpha_c_p1d": -0.011660630735101152, "bench_spy_p1d": -0.0044011570576184544, "bench_c_p1d": 0.000953592359751676, "ret_p1w": 0.011301823342826589, "ret_signed_p1w": 0.011301823342826589, "alpha_spy_p1w": 0.01170482476387158, "alpha_c_p1w": 0.02271032002068263, "bench_spy_p1w": -0.0004030014210449906, "bench_c_p1w": -0.01140849667785604, "ret_p1m": 0.3988782774933848, "ret_signed_p1m": 0.3988782774933848, "alpha_spy_p1m": 0.36264964205333494, "alpha_c_p1m": 0.3017185065181842, "bench_spy_p1m": 0.03622863544004984, "bench_c_p1m": 0.09715977097520057, "ret_p3m": 0.9209845603275635, "ret_signed_p3m": 0.9209845603275635, "alpha_spy_p3m": 0.8912979018016844, "alpha_c_p3m": 0.765011274257883, "bench_spy_p3m": 0.02968665852587904, "bench_c_p3m": 0.15597328606968053, "ret_p6m": 2.5017352385583753, "ret_signed_p6m": 2.5017352385583753, "alpha_spy_p6m": 2.4392069688433073, "alpha_c_p6m": 2.1089118993879765, "bench_spy_p6m": 0.06252826971506797, "bench_c_p6m": 0.3928233391703986, "price_path": [-0.1818, -0.1835, -0.1782, -0.1981, -0.1665, -0.1319, -0.1234, -0.0976, -0.0784, -0.0581, -0.031, -0.0149, -0.0137, -0.0186, 0.0174, 0.0216, 0.0342, 0.0342, 0.0493, 0.0364, 0.0859, 0.0803, 0.0697, 0.0592, 0.0463, 0.0263, 0.0335, 0.0382, 0.0382, 0.019, 0.0573, 0.0387, 0.0461, 0.0582, 0.0079, 0.0206, -0.0106, -0.0376, -0.028, -0.0378, -0.0719, -0.0667, -0.0506, -0.0546, -0.0546, -0.0486, -0.025, -0.0726, -0.1088, -0.0842, -0.0732, -0.0756, -0.0494, 0.0103, 0.0513, 0.0856, 0.056, 0.0647, 0.0271, 0.0499, 0.0371, -0.004, -0.0161, 0.0, -0.0107, -0.01, 0.0006, 0.0367, 0.0113, 0.0113, 0.0068, 0.0088, 0.0555, 0.1163, 0.1171, 0.1492, 0.1897, 0.2795, 0.3361, 0.3407, 0.3496, 0.3595, 0.4352, 0.3828, 0.3989, 0.4141, 0.3742, 0.3327, 0.3364, 0.3928, 0.4218, 0.5478, 0.5614, 0.5971, 0.6237, 0.5789, 0.6712, 0.6354, 0.5441, 0.6391, 0.5906, 0.6321, 0.7221, 0.7208, 0.7582, 0.7201, 0.6876, 0.7576, 0.8623, 0.8715, 0.8869, 0.927, 0.9047, 0.9027, 0.9956, 0.8539, 1.0194, 1.0265, 1.023, 1.1538, 1.0501, 1.0824, 1.0148, 1.0988, 1.0573, 0.9429, 0.921, 0.7122, 0.7633, 0.9041, 0.9092, 0.9579, 0.9579, 1.0108, 1.0446, 1.0364, 0.991, 0.9272, 1.0171, 1.0995, 1.1463, 1.2423, 1.1977, 1.0504, 1.0194, 0.977, 0.9176, 1.1134, 1.2611, 1.3518, 1.3491, 1.4376, 1.4376, 1.4216, 1.504, 1.4892, 1.4278, 1.4278, 1.6831, 1.6553, 1.9213, 1.8883, 1.7818, 1.9355, 1.9421, 1.8763, 1.8356, 1.8635, 2.0857, 2.0857, 2.1048, 2.3099, 2.382, 2.3996, 2.3097, 2.4897, 2.7027, 2.707, 2.5291, 2.7241, 2.5679, 2.2273, 2.257, 2.3574, 2.2622, 2.175, 2.4905, 2.5214, 2.5017, 2.5017]}
{"unit_id": "thread:1970633506616418332", "ticker_idx": 0, "ts": "2025-09-23T23:36:55+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron also projected that its market share in HBM4 would rise compared to 2025… In its fourth-quarter earnings report, Micron posted an earnings surprise… EPS came in at $3.03… far exceeding Wall Street consensus estimates", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares bullish Micron news: HBM4 elimination rumors denied, beat-and-raise quarter, rising HBM market share projected, $100B HBM market by 2030.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Micron CEO Denies Rumors of HBM4 Elimination… “Mass Production to Begin in First Half of Next Year”\n\nMicron has effectively denied rumors circulating in the Korean securities market that the company had been eliminated from the HBM4 competition.\n\nOn the earnings conference call held on the 23rd (local time), Sanjay Mehrotra, CEO of U.S. memory chipmaker Micron, said:\n“Micron plans to transition to HBM4 in 2026, and we can provide industry-leading data processing speeds of 11 gigabits per second (Gbps) to meet customer requirements,” adding that “the first shipments will be in the first half of 2026.”\n\nMicron also projected that its market share in HBM4 would rise compared to 2025. The company stated that it is currently in discussions with customers about specifications and volumes, and that it plans to finalize sales contracts for the remaining supply of its entire 2026 HBM capacity within the next few months. Regarding HBM3E, Micron announced that pricing negotiations for the bulk of its supply have already been completed with most customers.\n\nRecently, in the semiconductor industry—especially among securities analysts—speculation had been spreading that Micron failed to meet NVIDIA’s requirement of achieving 10Gbps performance in HBM4.\n\nHowever, given the disclosed mass production schedule, it is possible Micron hurriedly redesigned its product to meet NVIDIA’s demands, since its timeline specifies production next year rather than this year. The company also said that for the next generation HBM4E, it will cooperate with TSMC to produce the base logic die.\n\nThe CEO also offered an optimistic outlook for the HBM market:\n“By 2030, the HBM market will reach $100 billion,” said Sanjay, noting that “HBM’s growth trajectory is clearer than that of general DRAM, and this trend will continue in 2026 as well.”\n\nIn its fourth-quarter earnings report (June–August), Micron posted an earnings surprise. The company recorded revenue of $11.32 billion (about 15 trillion KRW) and operating profit of $3.96 billion (about 5 trillion KRW), up 46% and 126.6% year-over-year, respectively. Earnings per share (EPS) came in at $3.03, up 156.8% compared to the previous year, far exceeding Wall Street consensus estimates of $11.12 billion in revenue and EPS of $2.87.\n\nDRAM sales drove the strong results. Micron’s Q4 DRAM revenue rose about 70% year-on-year to $8.98 billion. In contrast, NAND revenue fell 5% year-on-year to $2.25 billion, as the recent uptick in NAND prices had not yet been reflected.\n\nFor its fiscal Q1 2026 guidance, Micron projected revenue of $12.2–12.8 billion and adjusted EPS of $3.60–3.90.\n\nSanjay emphasized:\n“As the only memory manufacturer in the United States, Micron is in a unique position to take advantage of the upcoming opportunities in the AI industry.”\n\n$MU\n\n---\n\nhttps://t.co/KWUhHb5CMn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010756522195869866, "ret_signed_m1d": -0.010756522195869866, "alpha_spy_m1d": -0.016229790634715746, "alpha_c_m1d": -0.012217492831149812, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02824341205600267, "ret_signed_p1d": -0.02824341205600267, "alpha_spy_p1d": -0.025061778603388984, "alpha_c_p1d": -0.02640932438425203, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.0054684811190846006, "ret_signed_p1w": 0.0054684811190846006, "alpha_spy_p1w": 0.0009903523963925576, "alpha_c_p1w": -0.009016948046868034, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.19340362628470564, "ret_signed_p1m": 0.19340362628470564, "alpha_spy_p1m": 0.18648281458315097, "alpha_c_p1m": 0.1408702510561941, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.5989816019108829, "ret_signed_p3m": 0.5989816019108829, "alpha_spy_p3m": 0.5697434492981737, "alpha_c_p3m": 0.49164560331085694, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 1.7775123773604586, "ret_signed_p6m": 1.7775123773604586, "alpha_spy_p6m": 1.777249431682111, "alpha_c_p6m": 1.5499869217099527, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.2435, -0.251, -0.2601, -0.2742, -0.2691, -0.2658, -0.2658, -0.2794, -0.2523, -0.2654, -0.2602, -0.2517, -0.2872, -0.2782, -0.3003, -0.3194, -0.3126, -0.3196, -0.3437, -0.34, -0.3286, -0.3314, -0.3315, -0.3272, -0.3105, -0.3442, -0.3697, -0.3524, -0.3446, -0.3463, -0.3277, -0.2856, -0.2565, -0.2323, -0.2532, -0.2471, -0.2737, -0.2576, -0.2666, -0.2957, -0.3042, -0.2928, -0.3004, -0.2999, -0.2924, -0.2669, -0.2848, -0.2848, -0.288, -0.2866, -0.2536, -0.2106, -0.21, -0.1873, -0.1587, -0.0952, -0.0552, -0.0519, -0.0456, -0.0386, 0.0149, -0.0221, -0.0108, 0.0, -0.0282, -0.0576, -0.0549, -0.0151, 0.0055, 0.0946, 0.1042, 0.1294, 0.1482, 0.1166, 0.1818, 0.1565, 0.092, 0.1591, 0.1248, 0.1541, 0.2178, 0.2169, 0.2433, 0.2164, 0.1934, 0.243, 0.317, 0.3235, 0.3343, 0.3627, 0.347, 0.3455, 0.4113, 0.311, 0.4281, 0.4331, 0.4306, 0.5231, 0.4498, 0.4726, 0.4248, 0.4842, 0.4548, 0.374, 0.3585, 0.2108, 0.2469, 0.3465, 0.3501, 0.3846, 0.3846, 0.422, 0.4459, 0.4401, 0.408, 0.3629, 0.4264, 0.4847, 0.5178, 0.5857, 0.5541, 0.45, 0.4281, 0.3981, 0.3561, 0.4945, 0.599, 0.6631, 0.6612, 0.7238, 0.7238, 0.7124, 0.7708, 0.7603, 0.7169, 0.7169, 0.8974, 0.8777, 1.0659, 1.0425, 0.9672, 1.0759, 1.0806, 1.034, 1.0052, 1.025, 1.1821, 1.1821, 1.1956, 1.3407, 1.3916, 1.4041, 1.3406, 1.4678, 1.6184, 1.6215, 1.4957, 1.6336, 1.5231, 1.2823, 1.3033, 1.3742, 1.3069, 1.2453, 1.4684, 1.4902, 1.4763, 1.4763, 1.4049, 1.5322, 1.5105, 1.5756, 1.5323, 1.5145, 1.5806, 1.4998, 1.4806, 1.4824, 1.2839, 1.4108, 1.3884, 1.2275, 1.3419, 1.4249, 1.5186, 1.4384, 1.5634, 1.6576, 1.7773, 1.7775]}
{"unit_id": "orig:1970349759954329958", "ticker_idx": 0, "ts": "2025-09-23T04:49:24+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics OP of KRW 46 trillion, target P/B of 1.4x, TP KRW 93,000", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish memory cycle; Samsung TP KRW 93,000 and SK Hynix TP KRW 430,000 on 2026 OP forecasts as commodity DRAM/NAND upcycle extends into 1H next year.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "The Commodity Fever\n[Samsung Securities, Semiconductor/IT – Jongwook Lee]\n\nThe recent rally in memory stock prices is due to the rise in commodity prices.\nMany investors believe that the memory price upcycle is still at an early stage, and we are also witnessing strong FOMO for Asian tech stocks, not just in Korea.\nIt may be useful to calmly consider how long this rally can last, and what events could mark the turning point.\n\n1. Why is it rising?\nIt is still more about supply than demand.\nSince supply has been aligned with conservative demand assumptions, even a small increase in demand can trigger a sharp price reaction.\nThere’s also the bad memory of last year when DRAM prices fell because server-side commodity demand didn’t materialize as consumer demand turned out weak.\nYield deterioration and supply disruptions due to technical issues are also commonplace.\nThis time, however, there has been a modest increase in demand from general servers.\n\n2. Why now?\nWe believe that with stronger cloud investment and capacity build-outs for AI inference, commodity memory demand crossed a tipping point after 3Q.\nAs a result, memory makers gained pricing power in 4Q negotiations, both on inventory and demand fronts.\nThe decisive factor for investors was that HDD price hikes and Samsung’s QLC disruptions spread the rally further into NAND prices.\n\n3. How long will it last?\nWe think the probability of memory price increases lasting into the first half of next year has risen significantly.\nDirectly speaking, prices can keep rising repeatedly until memory suppliers decide on new commodity capacity expansion.\nWe see three conditions for the end of this commodity rally:\n    1. new capacity expansion,\n    2. suppliers overcoming technical hurdles, and\n    3. the end of interest rate cuts or a shift to rate hikes.\n\n4. Profit outlook\nWe forecast for 2026:\n    • Samsung Electronics OP of KRW 46 trillion, target P/B of 1.4x, TP KRW 93,000\n    • SK Hynix OP of KRW 51 trillion, target P/B of 2.1x, TP KRW 430,000\n\nThe key points are:\n    1. Unlike HBM, all memory players will benefit, which means relative “catch-up” pricing will work, and\n    2. current earnings and target prices are in an upward revision trend; more importantly, this signals a turning point in memory capex direction.\n\nPlease refer to the full report for further details.\n\nThank you.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014167682693315031, "ret_signed_m1d": -0.014167682693315031, "alpha_spy_m1d": -0.01964095113216091, "alpha_c_m1d": -0.02287516051591476, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.00870747782259973, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00826445047379698, "ret_signed_p1d": 0.00826445047379698, "alpha_spy_p1d": 0.011446083926410666, "alpha_c_p1d": 0.014438117118731553, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.006173666644934572, "ret_p1w": -0.00502563521762156, "ret_signed_p1w": -0.00502563521762156, "alpha_spy_p1w": -0.009503763940313603, "alpha_c_p1w": -0.010878234503607032, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.005852599285985471, "ret_p1m": 0.1693023341729294, "ret_signed_p1m": 0.1693023341729294, "alpha_spy_p1m": 0.16238152247137472, "alpha_c_p1m": 0.150495649812612, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.018806684360317405, "ret_p3m": 0.2606170159033787, "ret_signed_p3m": 0.2606170159033787, "alpha_spy_p3m": 0.23137886329066948, "alpha_c_p3m": 0.22857067124980013, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.032046344653578585, "ret_p6m": 1.4846316970647475, "ret_signed_p6m": 1.4846316970647475, "alpha_spy_p6m": 1.4843687513864, "alpha_c_p6m": 1.4984832418519514, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": -0.013851544787203851, "price_path": [-0.2936, -0.2822, -0.294, -0.2893, -0.2822, -0.2468, -0.2527, -0.2715, -0.2751, -0.2869, -0.2798, -0.2609, -0.2621, -0.2479, -0.2361, -0.2125, -0.2078, -0.1995, -0.2208, -0.2161, -0.2208, -0.222, -0.1688, -0.1665, -0.1429, -0.157, -0.1865, -0.1771, -0.1747, -0.1877, -0.1677, -0.1523, -0.1617, -0.1606, -0.1511, -0.1547, -0.1547, -0.1736, -0.1736, -0.1677, -0.1665, -0.157, -0.1558, -0.17, -0.1665, -0.1783, -0.1771, -0.2019, -0.1842, -0.1759, -0.1724, -0.1795, -0.1724, -0.1558, -0.1429, -0.1334, -0.1098, -0.0968, -0.0626, -0.0767, -0.0519, -0.0519, -0.0142, 0.0, 0.0083, 0.0165, -0.0165, -0.0015, -0.005, 0.0199, 0.0643, 0.0643, 0.0643, 0.0643, 0.0643, 0.0643, 0.1195, 0.1064, 0.0863, 0.1266, 0.1586, 0.161, 0.1634, 0.1563, 0.1693, 0.1444, 0.1717, 0.2096, 0.18, 0.1918, 0.2345, 0.2748, 0.3175, 0.244, 0.193, 0.1764, 0.161, 0.193, 0.2274, 0.2227, 0.2191, 0.1527, 0.193, 0.1598, 0.1444, 0.193, 0.1242, 0.1468, 0.1776, 0.2191, 0.2274, 0.1918, 0.1954, 0.2262, 0.2393, 0.2464, 0.2855, 0.2986, 0.2855, 0.2808, 0.2725, 0.2915, 0.2428, 0.2191, 0.2796, 0.276, 0.2606, 0.3104, 0.3223, 0.3175, 0.3175, 0.3875, 0.424, 0.4288, 0.4288, 0.4288, 0.5313, 0.6457, 0.6552, 0.6803, 0.654, 0.6564, 0.654, 0.6397, 0.6719, 0.7148, 0.7744, 0.7792, 0.7303, 0.7815, 0.8149, 0.8125, 0.8125, 0.9007, 0.9353, 0.915, 0.9126, 0.7923, 0.996, 1.0151, 0.8983, 0.89, 0.9829, 0.9758, 0.9996, 1.1283, 1.1593, 1.1593, 1.1593, 1.1593, 1.2642, 1.2654, 1.2999, 1.3833, 1.425, 1.5978, 1.58, 1.58, 1.3249, 1.0521, 1.2832, 1.2427, 1.0675, 1.2391, 1.2642, 1.2391, 1.1867, 1.2487, 1.3106, 1.4846]}
{"unit_id": "orig:1970349759954329958", "ticker_idx": 1, "ts": "2025-09-23T04:49:24+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix OP of KRW 51 trillion, target P/B of 2.1x, TP KRW 430,000", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish memory cycle; Samsung TP KRW 93,000 and SK Hynix TP KRW 430,000 on 2026 OP forecasts as commodity DRAM/NAND upcycle extends into 1H next year.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "The Commodity Fever\n[Samsung Securities, Semiconductor/IT – Jongwook Lee]\n\nThe recent rally in memory stock prices is due to the rise in commodity prices.\nMany investors believe that the memory price upcycle is still at an early stage, and we are also witnessing strong FOMO for Asian tech stocks, not just in Korea.\nIt may be useful to calmly consider how long this rally can last, and what events could mark the turning point.\n\n1. Why is it rising?\nIt is still more about supply than demand.\nSince supply has been aligned with conservative demand assumptions, even a small increase in demand can trigger a sharp price reaction.\nThere’s also the bad memory of last year when DRAM prices fell because server-side commodity demand didn’t materialize as consumer demand turned out weak.\nYield deterioration and supply disruptions due to technical issues are also commonplace.\nThis time, however, there has been a modest increase in demand from general servers.\n\n2. Why now?\nWe believe that with stronger cloud investment and capacity build-outs for AI inference, commodity memory demand crossed a tipping point after 3Q.\nAs a result, memory makers gained pricing power in 4Q negotiations, both on inventory and demand fronts.\nThe decisive factor for investors was that HDD price hikes and Samsung’s QLC disruptions spread the rally further into NAND prices.\n\n3. How long will it last?\nWe think the probability of memory price increases lasting into the first half of next year has risen significantly.\nDirectly speaking, prices can keep rising repeatedly until memory suppliers decide on new commodity capacity expansion.\nWe see three conditions for the end of this commodity rally:\n    1. new capacity expansion,\n    2. suppliers overcoming technical hurdles, and\n    3. the end of interest rate cuts or a shift to rate hikes.\n\n4. Profit outlook\nWe forecast for 2026:\n    • Samsung Electronics OP of KRW 46 trillion, target P/B of 1.4x, TP KRW 93,000\n    • SK Hynix OP of KRW 51 trillion, target P/B of 2.1x, TP KRW 430,000\n\nThe key points are:\n    1. Unlike HBM, all memory players will benefit, which means relative “catch-up” pricing will work, and\n    2. current earnings and target prices are in an upward revision trend; more importantly, this signals a turning point in memory capex direction.\n\nPlease refer to the full report for further details.\n\nThank you.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027700775241169207, "ret_signed_m1d": -0.027700775241169207, "alpha_spy_m1d": -0.033174043680015086, "alpha_c_m1d": -0.029161745876449152, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009695288694700488, "ret_signed_p1d": -0.009695288694700488, "alpha_spy_p1d": -0.006513655242086802, "alpha_c_p1d": -0.007861201022949849, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.037396063935869694, "ret_signed_p1w": -0.037396063935869694, "alpha_spy_p1w": -0.04187419265856174, "alpha_c_p1w": -0.05188149310182233, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.33379508814861514, "ret_signed_p1m": 0.33379508814861514, "alpha_spy_p1m": 0.32687427644706046, "alpha_c_p1m": 0.2812617129201036, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.5163205834724471, "ret_signed_p3m": 0.5163205834724471, "alpha_spy_p3m": 0.4870824308597379, "alpha_c_p3m": 0.4089845848724212, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 1.9327045667287934, "ret_signed_p6m": 1.9327045667287934, "alpha_spy_p6m": 1.9324416210504458, "alpha_c_p6m": 1.7051791110782875, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.1895, -0.2144, -0.1923, -0.2103, -0.2283, -0.2296, -0.2518, -0.2504, -0.22, -0.2227, -0.1785, -0.1854, -0.1702, -0.1743, -0.1812, -0.2545, -0.2559, -0.2462, -0.2573, -0.2559, -0.2545, -0.2642, -0.2753, -0.2739, -0.2711, -0.2435, -0.2863, -0.2863, -0.2711, -0.285, -0.2753, -0.2905, -0.2615, -0.2559, -0.231, -0.2352, -0.2352, -0.2601, -0.2725, -0.2933, -0.3223, -0.3057, -0.2822, -0.2767, -0.2808, -0.2562, -0.2548, -0.2909, -0.2784, -0.2729, -0.2645, -0.2424, -0.2327, -0.2022, -0.1579, -0.1496, -0.09, -0.0831, -0.036, -0.0762, -0.0152, -0.0152, -0.0277, 0.0, -0.0097, -0.0125, -0.0679, -0.0332, -0.0374, -0.0028, 0.0956, 0.0956, 0.0956, 0.0956, 0.0956, 0.0956, 0.1856, 0.1496, 0.1399, 0.1704, 0.2535, 0.2895, 0.3449, 0.3269, 0.3338, 0.3255, 0.4127, 0.482, 0.4432, 0.5457, 0.5734, 0.5485, 0.7175, 0.6233, 0.6039, 0.6427, 0.6066, 0.6787, 0.7147, 0.7091, 0.6953, 0.5512, 0.6787, 0.5789, 0.5568, 0.5817, 0.4432, 0.4404, 0.4377, 0.4515, 0.508, 0.4692, 0.4914, 0.5468, 0.5302, 0.5025, 0.508, 0.5995, 0.569, 0.6272, 0.5662, 0.5829, 0.5357, 0.4692, 0.5274, 0.5302, 0.5163, 0.6078, 0.6189, 0.63, 0.63, 0.6605, 0.7741, 0.8046, 0.8046, 0.8046, 0.8767, 0.9294, 1.0125, 1.0569, 1.0957, 1.0624, 1.0763, 1.0458, 1.0569, 1.0763, 1.0957, 1.1179, 1.0596, 1.0513, 1.0929, 1.1262, 1.0402, 1.2177, 1.3313, 1.3867, 1.5198, 1.3008, 1.5143, 1.4949, 1.3341, 1.3258, 1.4588, 1.4283, 1.384, 1.4616, 1.4394, 1.4394, 1.4394, 1.4394, 1.4782, 1.6307, 1.6362, 1.7859, 1.822, 2.0521, 1.9466, 1.9466, 1.6078, 1.3578, 1.6133, 1.5661, 1.3217, 1.605, 1.6522, 1.5828, 1.5272, 1.705, 1.6939, 1.9327]}
{"unit_id": "orig:1951093817614147811", "ticker_idx": 0, "ts": "2025-08-01T01:33:10+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The investment decision is a matter of when to start accumulating shares for the 2027 story", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish AAPL long-term on 2027 20th-anniversary iPhone cycle; near-term headwinds from tariffs and AI lag, but accumulation timing is the question.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "What Apple’s Earnings Don’t Tell You\n\n[Samsung Securities IT, Semiconductor/Lee Jong-wook]\n\n1. The Truth Behind the Revenue Surprise\n\nThis quarter's earnings surprise is based on a revenue surprise. Although product revenue grew by 8.2%, the product Gross Profit Margin (GPM) decreased by 0.8%pts. The revenue surprise (+9.6% y-y) can be broken down into three main areas:\n\n1. iPhone and MacBook sales (+13.7%)\n2. Revenue growth outside the U.S./Europe (+12.3%)\n3. Service revenue growth (+13.3%)Growth in the U.S. (+9.3%) and China (+4.4%) lagged the company-wide growth rate.\n\nHowever, we believe the breakdown should be different.\n\n- 1. Dollar Depreciation\nThe trade-weighted value of the U.S. dollar fell by 7.0% year-over-year. Considering its value fell only 0.9% against the Yuan (which has a 19% weight), the dollar's value excluding China dropped by 8.4%. Of the 12.3% revenue growth in regions outside the U.S. and China, 70% can be explained by foreign exchange rates. The difference in growth rates between China and ex-China regions is 7.9%pts, while the difference in currency fluctuations is 7.5%pts. What appears to be sluggishness in China is largely explained by the difference in exchange rates. Indeed, Apple mentioned the contribution from the weaker dollar during the Q&A session.\n\n- 2. Inventory Front-loading Due to U.S. Tariffs\nIt appears that Apple's sell-in within the U.S. increased significantly after the tariff announcement in April. Global iPhone distribution inventory rose by 3.7 million units during the two months of April and May, a meaningful increase compared to 2024 (+0.6 million units) or 2023 (-0.4 million units). We suspect most of this occurred in the U.S. channel and is the reason why the U.S. grew during an unusual off-season period.\n\n - 3. Sales Growth\nThe places where Apple was truly competitive were Japan and India. The iPhone's market share in India is 10%, about half of its global share, and it saw benefits from the iPhone 16e launch. In Japan, where the Google Pixel usually sells well, Pixel sales were halted due to LTE patent infringement, and the Sony Xperia is undergoing free replacements for hardware defects. The iPhone is seeing a reflective benefit.\n\n2. AI Competitiveness\n\n- 1. M&A, As Quickly as Possible\nApple stated in an interview today that it is open to M&A of AI companies, which is welcome news from an investor's perspective. As we have mentioned, Apple's core is not technology but design and ecosystem, and it has shopped for most of its technology via M&A until now. This is not unusual at all. Rather, we believe that Apple solving AI through partnerships does not align with its philosophy, and it should be internalized through M&A.Considering past cases where productization occurred 2-3 years after an M&A, they are a bit late. However, the hopeful sign is that Apple is becoming more aggressive as it is increasingly backed into a corner.\n\n- 2. Head of AI Shifts from Siri Lead to Vision Pro Lead\nThe new head of Apple AI, Mike Rockwell, was formerly the head of Apple Vision Pro development. Before Apple, he was an engineer at Dolby, developing 3D spatial environments and audio. His current interest is known to be the personalization of AI. He seems appropriate for bringing AI into the realm of consumer devices.\n\n3. Tariffs\n\n- 1. Who Will Be the Scapegoat?\nTariffs on semiconductor items are imminent. This is a period of increased volatility for Apple. Now that the decision has been made to impose tariffs on semiconductors, someone must be sacrificed. To simplify, it's a game of whether the scapegoat will be NVIDIA (AI), Apple (smartphones), or Dell & HP (PC, servers). Apple's priority for relief does not seem very high. Of course, tariffs on the iPhone raise inflation concerns, but if inflation were the U.S. government's top priority, it would not have pursued a tariff policy in the first place.\n\n- 2. Hopes Are on 2027\nAs expected, our hopes are also pinned on the 20th-anniversary iPhone in 2027. We believe Apple's capabilities, from hardware to AI software, will be concentrated in 2027. The investment decision is a matter of when to start accumulating shares for the 2027 story. It is difficult to have high expectations for this year's iPhone.\n\nThank you.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025644763850577235, "ret_signed_m1d": 0.025644763850577235, "alpha_spy_m1d": 0.008981073909480664, "alpha_c_m1d": 0.0035885126325636563, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.02205625121801358, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004792908693340925, "ret_signed_p1d": 0.004792908693340925, "alpha_spy_p1d": -0.010406905495620844, "alpha_c_p1d": -0.015512978834021585, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02030588752736251, "ret_p1w": 0.13326417762117826, "ret_signed_p1w": 0.13326417762117826, "alpha_spy_p1w": 0.10839759564928442, "alpha_c_p1w": 0.09883761691001935, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.034426560711158904, "ret_p1m": 0.14835179516828956, "ret_signed_p1m": 0.14835179516828956, "alpha_spy_p1m": 0.11082673721124481, "alpha_c_p1m": 0.12742347664381581, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.020928318524473744, "ret_p3m": 0.33415387615268677, "ret_signed_p3m": 0.33415387615268677, "alpha_spy_p3m": 0.2254621411005353, "alpha_c_p3m": 0.14957030716290953, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.18458356898977724, "ret_p6m": 0.26468879513203136, "ret_signed_p6m": 0.26468879513203136, "alpha_spy_p6m": 0.1440826994844926, "alpha_c_p6m": 0.12500226339693055, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.13968653173510082, "price_path": [-0.0204, -0.0316, -0.0254, -0.0203, 0.0416, 0.0521, 0.0492, 0.0448, 0.0439, 0.0316, 0.0221, -0.0014, -0.005, -0.0351, -0.0351, -0.0107, -0.0097, -0.012, -0.0076, -0.0034, 0.0044, 0.0022, -0.0086, 0.0076, -0.0046, 0.0014, -0.0178, -0.0157, -0.0293, -0.0196, -0.0333, -0.0287, -0.0287, -0.0068, -0.0043, -0.0103, -0.0041, -0.0068, -0.0064, 0.0138, 0.0269, 0.0497, 0.0552, 0.0552, 0.0374, 0.0377, 0.0433, 0.0496, 0.0434, 0.0308, 0.0333, 0.0384, 0.0378, 0.0435, 0.0499, 0.0594, 0.0582, 0.0562, 0.0568, 0.0577, 0.0439, 0.033, 0.0256, 0.0, 0.0048, 0.0027, 0.0537, 0.0872, 0.1333, 0.1238, 0.136, 0.1542, 0.1515, 0.1456, 0.1422, 0.1405, 0.118, 0.1125, 0.1267, 0.1237, 0.1344, 0.1402, 0.1504, 0.1484, 0.1484, 0.1364, 0.1797, 0.1861, 0.1857, 0.1767, 0.1593, 0.1219, 0.1379, 0.1579, 0.1709, 0.1781, 0.1822, 0.1767, 0.2144, 0.2668, 0.2586, 0.2481, 0.2707, 0.2637, 0.2586, 0.2596, 0.2637, 0.272, 0.2764, 0.2698, 0.2688, 0.2766, 0.2567, 0.2133, 0.2251, 0.2257, 0.2334, 0.2241, 0.248, 0.2973, 0.2999, 0.2785, 0.2841, 0.3001, 0.3298, 0.3307, 0.3342, 0.3426, 0.3375, 0.3309, 0.3358, 0.3363, 0.3345, 0.3281, 0.3341, 0.3629, 0.3541, 0.3515, 0.3489, 0.3244, 0.3243, 0.3298, 0.3184, 0.3443, 0.3662, 0.3714, 0.3743, 0.3743, 0.3808, 0.4018, 0.4171, 0.407, 0.3899, 0.3804, 0.376, 0.3725, 0.3804, 0.3767, 0.3779, 0.3573, 0.3598, 0.346, 0.3478, 0.3551, 0.3417, 0.3486, 0.3558, 0.3558, 0.3538, 0.3556, 0.3522, 0.3461, 0.3461, 0.3419, 0.3234, 0.2991, 0.2891, 0.2827, 0.2843, 0.2887, 0.2926, 0.2872, 0.2786, 0.2653, 0.2653, 0.2216, 0.2263, 0.2297, 0.2282, 0.2647]}
{"unit_id": "orig:1963211962470584460", "ticker_idx": 0, "ts": "2025-09-03T12:06:21+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it remains his top pick in semiconductors", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM/Harlan Sur's bullish AVGO preview: beat + guidance raise expected on AI ASIC/networking strength; FY25 AI rev >$20B, FY26 >$33B; top semi pick.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM (25/09/02)\n\nAVGO: Earnings Preview — Strong AI demand expected to drive earnings beat and guidance raise; accelerated adoption from Google and Meta + continued AI networking growth.\n\nBroadcom will announce its F3Q25 results after U.S. market close next Thursday. Harlan Sur’s preview highlights:\n\n1. Optimistic demand outlook for AI ASIC and networking solutions;\n\n2. Continued improvement in non-AI semiconductor businesses (enterprise, server/storage, broadband);\n\n3. Further release of revenue synergies from VMware.He expects the company to deliver a beat, with AI revenue between $5.2–5.4 billion (consensus: $5.1 billion; buy-side mean: $5.26 billion), benefiting from the early shipment of Google’s TPU v6 3nm.\n\nIn addition, he expects AI semiconductor revenue guidance to grow 8–10% (over $6.0 billion), mainly driven by sustained strong AI networking demand, continued adoption of Google’s TPU v6 3nm, and ramping of Meta’s MTIA (inference) project. He estimates that Google’s TPU v6 3nm will contribute $9.0 billion in revenue over the next 6–7 months, with lifetime revenue exceeding $15 billion. On the networking side, demand for Tomahawk 5 / Jericho 3 AI remains strong, while Tomahawk 6 (3nm) is also accelerating its adoption.\n\nHarlan’s model forecasts FY25 AI revenue to exceed $20 billion (+60% YoY), and FY26 AI revenue to surpass $33 billion, benefiting from new products and projects (Meta MTIA 3nm ASIC, TPU v7 3nm, SoftBank/ARM and OpenAI 2nm/3nm ASICs), as well as the ramp of Tomahawk 6. Cyclical trends in broadband, storage, and enterprise markets also support non-AI semiconductor businesses, while in software infrastructure Harlan continues to see strong momentum (particularly VMware renewal performance, driving further transition and upsell into the VCF full stack).\n\nGiven Broadcom’s deep exposure to AI infrastructure spend, diversified end-market coverage, and industry-leading gross margins, operating margins, and free cash flow margins, it remains his top pick in semiconductors.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013724127508028428, "ret_signed_m1d": -0.013724127508028428, "alpha_spy_m1d": -0.008333781098861182, "alpha_c_m1d": -0.013898763145839599, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": 0.0001746356378111713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012268921158915536, "ret_signed_p1d": 0.012268921158915536, "alpha_spy_p1d": 0.00391156071710097, "alpha_c_p1d": 0.0002939823348269144, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.011974938824088621, "ret_p1w": 0.2221633296323029, "ret_signed_p1w": 0.2221633296323029, "alpha_spy_p1w": 0.20900577458955505, "alpha_c_p1w": 0.17188932054526473, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.05027400908703816, "ret_p1m": 0.12027300267929708, "ret_signed_p1m": 0.12027300267929708, "alpha_spy_p1m": 0.07780970597196557, "alpha_c_p1m": -0.060608905154126536, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.18088190783342362, "ret_p3m": 0.2789491617668358, "ret_signed_p3m": 0.2789491617668358, "alpha_spy_p3m": 0.21927278526069127, "alpha_c_p3m": 0.04667596151854103, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.2322732002482948, "ret_p6m": 0.06771962151439914, "ret_signed_p6m": 0.06771962151439914, "alpha_spy_p6m": -0.009195747159325585, "alpha_c_p6m": -0.3751879913168956, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.44290761283129476, "price_path": [-0.1853, -0.1941, -0.1929, -0.1656, -0.1552, -0.1795, -0.1683, -0.1773, -0.171, -0.171, -0.1733, -0.1608, -0.1277, -0.1248, -0.1066, -0.1093, -0.0884, -0.1245, -0.1074, -0.09, -0.09, -0.0933, -0.1012, -0.081, -0.0893, -0.0926, -0.0886, -0.0709, -0.0714, -0.0527, -0.063, -0.0469, -0.0787, -0.0618, -0.0452, -0.0404, -0.0268, -0.0164, 0.0008, -0.0287, -0.0455, -0.0154, -0.0313, -0.0024, 0.0045, 0.0085, 0.005, 0.0345, 0.0222, 0.0292, 0.0131, 0.0111, -0.0247, -0.0371, -0.0423, -0.0277, -0.027, -0.0145, -0.0071, 0.0207, -0.0165, -0.0165, -0.0137, 0.0, 0.0123, 0.1075, 0.1431, 0.1134, 0.2222, 0.1893, 0.1901, 0.204, 0.1905, 0.1448, 0.1421, 0.1407, 0.1223, 0.1228, 0.124, 0.1134, 0.1082, 0.0862, 0.0929, 0.1044, 0.1203, 0.1209, 0.1114, 0.1144, 0.1445, 0.1429, 0.0754, 0.1816, 0.14, 0.1638, 0.1732, 0.1572, 0.1569, 0.1351, 0.1273, 0.1405, 0.1731, 0.1993, 0.2355, 0.2786, 0.2471, 0.2245, 0.201, 0.1659, 0.1892, 0.1779, 0.1575, 0.1872, 0.1659, 0.1767, 0.1262, 0.1345, 0.1351, 0.128, 0.1741, 0.1489, 0.127, 0.2521, 0.2755, 0.317, 0.317, 0.3349, 0.2789, 0.264, 0.2608, 0.2622, 0.2927, 0.3287, 0.3459, 0.368, 0.3462, 0.1923, 0.1257, 0.1306, 0.08, 0.0928, 0.1275, 0.1333, 0.1594, 0.1624, 0.1624, 0.1687, 0.1596, 0.1611, 0.1487, 0.1487, 0.1537, 0.1398, 0.141, 0.1401, 0.1035, 0.145, 0.169, 0.1769, 0.1281, 0.1385, 0.1673, 0.1673, 0.1039, 0.0913, 0.0803, 0.0622, 0.0782, 0.1045, 0.106, 0.0977, 0.0996, 0.099, 0.0632, 0.0224, 0.0306, 0.105, 0.1415, 0.1299, 0.1376, 0.0992, 0.0792, 0.0792, 0.1037, 0.1069, 0.1085, 0.1041, 0.0964, 0.0803, 0.1029, 0.0677]}
{"unit_id": "orig:1943952744748028315", "ticker_idx": 0, "ts": "2025-07-12T08:37:05+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi adjusted its iPhone shipment forecasts for the June/September quarters to 45 million and 50 million units respectively, reflecting short-term demand driven by promotions and a tariff buffer in the Chinese market", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi note on AAPL: short-term China iPhone demand supported by promotions/tariff buffer, App Store growing low double-digits; regulatory/AI risks remain mid-to-long-term headwinds.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Apple Outlook: Short-Term Sales Benefit from Promotions and Tariff Buffer, AI Strategy and Regulatory Risks Remain Mid-to-Long-Term Focus\nCiti (A. Malik, 25/07/10)\n\nCiti notes that although iPhone short-term sales in China have benefited from tariff suspensions and 618 promotions, full-year demand remains affected by delayed AI features and tariff uncertainties. Meanwhile, Apple has shown positive signals in AI collaboration and App Store service growth, but regulatory pressure still poses a long-term risk.\n\nShort-term Improvement in China Sales, But Full-Year Demand Remains Uncertain\nCiti adjusted its iPhone shipment forecasts for the June/September quarters to 45 million and 50 million units respectively, reflecting short-term demand driven by promotions and a tariff buffer in the Chinese market. However, full-year forecasts remain at 226M/234M for CY25/CY26, representing -0.5% and +3.1% annual growth, primarily due to concerns over delayed AI features and uncertainties arising from the Section 232 tariff decision.\n\nApp Store Services Maintain Steady Growth Trend\nSensor Tower data shows that App Store revenue grew 12% year-over-year in Q2 2024. Although Apple has not provided specific service revenue guidance, Citi still expects low double-digit growth to be maintained. No significant signs of slowdown due to DMA reforms have been observed in the EU region yet, and the services business remains an important support.\n\nAI Strategy More Open, But Large-Scale M&A Unlikely\nMarket rumors suggest Apple is considering adopting external LLM models to enhance Siri functionality, while also exploring third-party search AI integration in Safari. Although Apple opened its self-developed models for developer access at WWDC, indicating it retains an independent path, a more active trend of external cooperation, if realized, could boost investor sentiment. However, Apple has no history of large-scale M&A, and it is not expected to advance its AI布局 (AI layout) through major M&A.\n\nRegulatory Risks Remain Key Downside Factor\nCurrent regulatory risks, particularly the traffic acquisition cost (TAC) agreement with Google, are a key focus for investors. The Google antitrust case could trigger changes in Apple's revenue structure and cooperation model, leading to mid-to-long-term valuation pressure.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012175345536607374, "ret_signed_m1d": 0.012175345536607374, "alpha_spy_m1d": 0.014079828785932391, "alpha_c_m1d": 0.011353880013827533, "bench_spy_m1d": -0.0019044832493250174, "bench_c_m1d": 0.0008214655227798406, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00234878684197648, "ret_signed_p1d": 0.00234878684197648, "alpha_spy_p1d": 0.00662195389260356, "alpha_c_p1d": -0.006530975584464915, "bench_spy_p1d": -0.00427316705062708, "bench_c_p1d": 0.008879762426441395, "ret_p1w": 0.018502515843993983, "ret_signed_p1w": 0.018502515843993983, "alpha_spy_p1w": 0.012164483859275599, "alpha_c_p1w": -0.003364156831690668, "bench_spy_p1w": 0.006338031984718384, "bench_c_p1w": 0.02186667267568465, "ret_p1m": 0.10205459001825079, "ret_signed_p1m": 0.10205459001825079, "alpha_spy_p1m": 0.07343768994599031, "alpha_c_p1m": 0.05288388244679232, "bench_spy_p1m": 0.028616900072260476, "bench_c_p1m": 0.04917070757145847, "ret_p3m": 0.21909835546736578, "ret_signed_p3m": 0.21909835546736578, "alpha_spy_p3m": 0.14193762356338002, "alpha_c_p3m": 0.08244933070222094, "bench_spy_p3m": 0.07716073190398576, "bench_c_p3m": 0.13664902476514484, "ret_p6m": 0.2602452296187705, "ret_signed_p6m": 0.2602452296187705, "alpha_spy_p6m": 0.14666199443055605, "alpha_c_p6m": 0.10971177744751737, "bench_spy_p6m": 0.11358323518821445, "bench_c_p6m": 0.15053345217125313, "price_path": [-0.07, -0.057, -0.057, -0.0753, -0.0438, -0.0206, -0.0025, 0.0018, 0.006, 0.0111, 0.0173, 0.0212, -0.017, -0.0479, -0.0497, -0.0605, -0.0546, -0.0496, 0.0104, 0.0207, 0.0178, 0.0136, 0.0127, 0.0008, -0.0084, -0.0313, -0.0348, -0.064, -0.064, -0.0403, -0.0393, -0.0416, -0.0372, -0.0332, -0.0256, -0.0278, -0.0383, -0.0225, -0.0344, -0.0285, -0.0472, -0.0452, -0.0583, -0.0489, -0.0622, -0.0577, -0.0577, -0.0365, -0.0341, -0.0399, -0.0338, -0.0365, -0.0361, -0.0165, -0.0038, 0.0183, 0.0236, 0.0236, 0.0064, 0.0067, 0.0121, 0.0182, 0.0122, 0.0, 0.0023, 0.0074, 0.0067, 0.0123, 0.0185, 0.0277, 0.0265, 0.0246, 0.0252, 0.026, 0.0127, 0.0021, -0.005, -0.0299, -0.0253, -0.0273, 0.0222, 0.0547, 0.0994, 0.0902, 0.1021, 0.1197, 0.1171, 0.1114, 0.108, 0.1064, 0.0846, 0.0793, 0.093, 0.0901, 0.1004, 0.1061, 0.116, 0.114, 0.114, 0.1024, 0.1444, 0.1507, 0.1502, 0.1415, 0.1246, 0.0883, 0.1039, 0.1233, 0.1359, 0.1428, 0.1469, 0.1415, 0.1781, 0.2289, 0.221, 0.2108, 0.2327, 0.2259, 0.221, 0.2219, 0.2259, 0.2339, 0.2382, 0.2318, 0.2308, 0.2384, 0.2191, 0.177, 0.1885, 0.189, 0.1965, 0.1875, 0.2107, 0.2584, 0.261, 0.2403, 0.2457, 0.2612, 0.29, 0.2909, 0.2942, 0.3024, 0.2975, 0.2911, 0.2959, 0.2964, 0.2946, 0.2883, 0.2942, 0.3222, 0.3136, 0.3111, 0.3085, 0.2847, 0.2846, 0.29, 0.2789, 0.3041, 0.3254, 0.3304, 0.3332, 0.3332, 0.3395, 0.3599, 0.3747, 0.3649, 0.3483, 0.3391, 0.3348, 0.3314, 0.3391, 0.3355, 0.3367, 0.3167, 0.3191, 0.3058, 0.3075, 0.3146, 0.3016, 0.3083, 0.3152, 0.3152, 0.3133, 0.315, 0.3117, 0.3059, 0.3059, 0.3018, 0.2838, 0.2602]}
{"unit_id": "thread:1945998773584605561", "ticker_idx": 0, "ts": "2025-07-18T00:07:17+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Applied Materials Bolsters Advanced Semiconductor Packaging… Designating it as a new growth engine… is expanding its investment in advanced semiconductor packaging", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author shares bullish report on AMAT expanding into advanced packaging as new growth engine, with Besi as key collaborator and AMAT as largest Besi shareholder.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Applied Materials Bolsters Advanced Semiconductor Packaging\n\nApplied Materials, the world's largest semiconductor equipment company, is expanding its investment in advanced semiconductor packaging. Designating it as a new growth engine, the company has embarked on securing talent and developing new equipment solutions. This is interpreted as a strategic response to the increasing demand for advanced semiconductor packaging driven by the expansion of the Artificial Intelligence (AI) market.\n\nAccording to industry sources on the 17th, Applied Materials has internally decided to make advanced semiconductor packaging a future growth driver and is moving to significantly enhance its competitiveness. To this end, it is understood to have initiated comprehensive efforts, from recruitment to new equipment.\n\nFor talent, aggressive recruitment is underway with no limitations on scale. Multiple industry insiders stated in unison, \"Applied Materials is contacting the entire industry to secure high-level talent in advanced packaging. They are gathering so many people that it's as if they are 'sucking up' packaging personnel.\"\n\nIn particular, it has been identified that the expansion of personnel is centered around the Advanced Packaging Technology (APT) team in Taiwan. Applied Materials' Taiwan APT team is currently the organization leading major equipment development and is collaborating with global material and component companies, including those in Korea.\n\nIt is analyzed that R&D is being conducted with demand in mind, given Taiwan's strong semiconductor packaging ecosystem. Synergy is also expected with Applied Materials' Singapore Advanced Packaging R&D Center.\n\nThe company also plans to expand its packaging equipment portfolio. Currently, it is developing various packaging equipment, starting with lithography machines for semiconductor substrates. Development of hybrid bonding equipment, expected to be used in next-generation High Bandwidth Memory (HBM), 400-layer (or more) NAND flash memory, and 3D DRAM, is also in full swing. For this, Applied Materials is cooperating with BE Semiconductor Industries (Besi), a Dutch packaging equipment company. Applied Materials, which became the largest shareholder of Besi by acquiring a stake this year, has established a full-fledged collaborative system between the two companies.\n\nApplied Materials' move comes as the advanced semiconductor packaging market is expected to grow rapidly with the proliferation of AI. The semiconductor industry is heavily investing in packaging technology to enhance performance by expanding bandwidth and increasing integration in advanced semiconductors like AI chips. This aims to overcome various challenges that could not be solved by existing front-end processes through packaging technology.\n\nUntil now, Applied Materials has had an equipment portfolio primarily focused on front-end processes. It boasts virtually all front-end process technologies, excluding lithography, such as thin-film deposition, etching, circuit control, polishing, and metrology. By expanding its scope to advanced packaging (back-end processes), including substrate processes and wafer bonding, it is analyzed that the company intends to significantly expand its equipment portfolio.\n\nAccording to market research firm Global Information, the semiconductor packaging market is expected to grow from $104.1 billion (approximately 144.5 trillion KRW) this year to over 7% annual growth, surpassing $146.4 billion (approximately 203.2 trillion KRW) by 2030. An industry insider familiar with the matter stated, \"I understand that Applied Materials is considering additional equity investments or mergers and acquisitions (M&A) to strengthen advanced packaging. They are building diversified strategies to enhance packaging capabilities.\"\n\n$AMAT\n\n---\n\nhttps://t.co/yuegOOUAEX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010922060242036347, "ret_signed_m1d": 0.010922060242036347, "alpha_spy_m1d": 0.01018911051809157, "alpha_c_m1d": 0.006099723878486341, "bench_spy_m1d": 0.0007329497239447758, "bench_c_m1d": 0.004822336363550006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01139478830427021, "ret_signed_p1d": 0.01139478830427021, "alpha_spy_p1d": 0.009498613138603673, "alpha_c_p1d": 0.010705898173441142, "bench_spy_p1d": 0.001896175165666536, "bench_c_p1d": 0.0006888901308290674, "ret_p1w": -0.024942155987023873, "ret_signed_p1w": -0.024942155987023873, "alpha_spy_p1w": -0.040111557312355495, "alpha_c_p1w": -0.015228309537923979, "bench_spy_p1w": 0.015169401325331622, "bench_c_p1w": -0.009713846449099894, "ret_p1m": -0.14130428553775787, "ret_signed_p1m": -0.14130428553775787, "alpha_spy_p1m": -0.16635286442150876, "alpha_c_p1m": -0.16386630726632057, "bench_spy_p1m": 0.025048578883750894, "bench_c_p1m": 0.022562021728562698, "ret_p3m": 0.1984470758052579, "ret_signed_p3m": 0.1984470758052579, "alpha_spy_p3m": 0.13561181481316975, "alpha_c_p3m": 0.02115387729038387, "bench_spy_p3m": 0.06283526099208814, "bench_c_p3m": 0.17729319851487402, "ret_p6m": 0.6211118076525448, "ret_signed_p6m": 0.6211118076525448, "alpha_spy_p6m": 0.5070752074407712, "alpha_c_p6m": 0.27125969811764317, "bench_spy_p6m": 0.11403660021177364, "bench_c_p6m": 0.34985210953490165, "price_path": [-0.2758, -0.2482, -0.2137, -0.2065, -0.2104, -0.2169, -0.2109, -0.22, -0.1879, -0.1904, -0.1987, -0.1842, -0.1832, -0.1852, -0.1204, -0.094, -0.0882, -0.085, -0.1331, -0.1311, -0.1322, -0.1505, -0.1571, -0.1729, -0.1729, -0.1502, -0.1515, -0.1626, -0.1769, -0.1742, -0.1507, -0.1497, -0.1378, -0.1244, -0.1084, -0.0875, -0.0927, -0.0811, -0.1042, -0.0729, -0.0859, -0.0924, -0.0924, -0.1102, -0.097, -0.0539, -0.0387, -0.0363, -0.038, -0.0387, -0.0351, -0.0023, 0.0032, 0.0032, 0.0018, 0.0239, 0.026, 0.0399, 0.0393, 0.035, 0.0465, 0.0229, 0.0109, 0.0, 0.0114, -0.0173, -0.018, -0.0122, -0.0249, -0.0009, -0.0107, -0.0055, -0.0545, -0.0549, -0.04, -0.0593, -0.0646, -0.0383, -0.0292, -0.0318, -0.0104, -0.0022, -0.0116, -0.1506, -0.1413, -0.1482, -0.1548, -0.1583, -0.1443, -0.147, -0.1337, -0.1343, -0.1297, -0.1534, -0.1534, -0.1702, -0.1772, -0.1667, -0.143, -0.1466, -0.139, -0.1394, -0.104, -0.1164, -0.0999, -0.0861, -0.062, -0.0007, 0.0011, 0.0559, 0.0578, 0.0608, 0.0511, 0.0739, 0.0793, 0.0782, 0.1466, 0.1774, 0.1455, 0.1791, 0.1141, 0.1454, 0.1601, 0.1056, 0.1558, 0.149, 0.1984, 0.1992, 0.1848, 0.2013, 0.1901, 0.1615, 0.2031, 0.2046, 0.2182, 0.1988, 0.2415, 0.2246, 0.2275, 0.2518, 0.2122, 0.2685, 0.2298, 0.2116, 0.2379, 0.2042, 0.215, 0.1755, 0.1902, 0.2044, 0.1855, 0.2382, 0.162, 0.182, 0.2184, 0.2793, 0.3189, 0.3189, 0.331, 0.3442, 0.4, 0.4174, 0.4217, 0.4141, 0.4149, 0.4095, 0.4518, 0.4252, 0.3677, 0.3786, 0.3657, 0.31, 0.3376, 0.3529, 0.3666, 0.3731, 0.376, 0.376, 0.3819, 0.3879, 0.3717, 0.356, 0.356, 0.4187, 0.5002, 0.5619, 0.5418, 0.486, 0.5891, 0.6211]}
{"unit_id": "thread:1945998773584605561", "ticker_idx": 1, "ts": "2025-07-18T00:07:17+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Applied Materials is cooperating with BE Semiconductor Industries (Besi), a Dutch packaging equipment company. Applied Materials, which became the largest shareholder of Besi by acquiring a stake this year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author shares bullish report on AMAT expanding into advanced packaging as new growth engine, with Besi as key collaborator and AMAT as largest Besi shareholder.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: Applied Materials Bolsters Advanced Semiconductor Packaging\n\nApplied Materials, the world's largest semiconductor equipment company, is expanding its investment in advanced semiconductor packaging. Designating it as a new growth engine, the company has embarked on securing talent and developing new equipment solutions. This is interpreted as a strategic response to the increasing demand for advanced semiconductor packaging driven by the expansion of the Artificial Intelligence (AI) market.\n\nAccording to industry sources on the 17th, Applied Materials has internally decided to make advanced semiconductor packaging a future growth driver and is moving to significantly enhance its competitiveness. To this end, it is understood to have initiated comprehensive efforts, from recruitment to new equipment.\n\nFor talent, aggressive recruitment is underway with no limitations on scale. Multiple industry insiders stated in unison, \"Applied Materials is contacting the entire industry to secure high-level talent in advanced packaging. They are gathering so many people that it's as if they are 'sucking up' packaging personnel.\"\n\nIn particular, it has been identified that the expansion of personnel is centered around the Advanced Packaging Technology (APT) team in Taiwan. Applied Materials' Taiwan APT team is currently the organization leading major equipment development and is collaborating with global material and component companies, including those in Korea.\n\nIt is analyzed that R&D is being conducted with demand in mind, given Taiwan's strong semiconductor packaging ecosystem. Synergy is also expected with Applied Materials' Singapore Advanced Packaging R&D Center.\n\nThe company also plans to expand its packaging equipment portfolio. Currently, it is developing various packaging equipment, starting with lithography machines for semiconductor substrates. Development of hybrid bonding equipment, expected to be used in next-generation High Bandwidth Memory (HBM), 400-layer (or more) NAND flash memory, and 3D DRAM, is also in full swing. For this, Applied Materials is cooperating with BE Semiconductor Industries (Besi), a Dutch packaging equipment company. Applied Materials, which became the largest shareholder of Besi by acquiring a stake this year, has established a full-fledged collaborative system between the two companies.\n\nApplied Materials' move comes as the advanced semiconductor packaging market is expected to grow rapidly with the proliferation of AI. The semiconductor industry is heavily investing in packaging technology to enhance performance by expanding bandwidth and increasing integration in advanced semiconductors like AI chips. This aims to overcome various challenges that could not be solved by existing front-end processes through packaging technology.\n\nUntil now, Applied Materials has had an equipment portfolio primarily focused on front-end processes. It boasts virtually all front-end process technologies, excluding lithography, such as thin-film deposition, etching, circuit control, polishing, and metrology. By expanding its scope to advanced packaging (back-end processes), including substrate processes and wafer bonding, it is analyzed that the company intends to significantly expand its equipment portfolio.\n\nAccording to market research firm Global Information, the semiconductor packaging market is expected to grow from $104.1 billion (approximately 144.5 trillion KRW) this year to over 7% annual growth, surpassing $146.4 billion (approximately 203.2 trillion KRW) by 2030. An industry insider familiar with the matter stated, \"I understand that Applied Materials is considering additional equity investments or mergers and acquisitions (M&A) to strengthen advanced packaging. They are building diversified strategies to enhance packaging capabilities.\"\n\n$AMAT\n\n---\n\nhttps://t.co/yuegOOUAEX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0035169459286064075, "ret_signed_m1d": -0.0035169459286064075, "alpha_spy_m1d": -0.004249895652551183, "alpha_c_m1d": -0.008339282292156414, "bench_spy_m1d": 0.0007329497239447758, "bench_c_m1d": 0.004822336363550006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017194224556985604, "ret_signed_p1d": 0.017194224556985604, "alpha_spy_p1d": 0.015298049391319068, "alpha_c_p1d": 0.016505334426156537, "bench_spy_p1d": 0.001896175165666536, "bench_c_p1d": 0.0006888901308290674, "ret_p1w": -0.08479870746896778, "ret_signed_p1w": -0.08479870746896778, "alpha_spy_p1w": -0.0999681087942994, "alpha_c_p1w": -0.07508486101986789, "bench_spy_p1w": 0.015169401325331622, "bench_c_p1w": -0.009713846449099894, "ret_p1m": -0.07932779601761619, "ret_signed_p1m": -0.07932779601761619, "alpha_spy_p1m": -0.10437637490136709, "alpha_c_p1m": -0.10188981774617889, "bench_spy_p1m": 0.025048578883750894, "bench_c_p1m": 0.022562021728562698, "ret_p3m": 0.13716293134670976, "ret_signed_p3m": 0.13716293134670976, "alpha_spy_p3m": 0.07432767035462162, "alpha_c_p3m": -0.04013026716816426, "bench_spy_p3m": 0.06283526099208814, "bench_c_p3m": 0.17729319851487402, "ret_p6m": 0.26729185082849716, "ret_signed_p6m": 0.26729185082849716, "alpha_spy_p6m": 0.15325525061672352, "alpha_c_p6m": -0.08256025870640449, "bench_spy_p6m": 0.11403660021177364, "bench_c_p6m": 0.34985210953490165, "price_path": [-0.2883, -0.2824, -0.2688, -0.2416, -0.2363, -0.2294, -0.2635, -0.2635, -0.2208, -0.2177, -0.2275, -0.1872, -0.1708, -0.1567, -0.1012, -0.0723, -0.061, -0.0973, -0.1145, -0.1336, -0.152, -0.1493, -0.1551, -0.177, -0.1598, -0.1446, -0.1489, -0.1364, -0.1673, -0.1876, -0.1657, -0.1563, -0.1356, -0.1122, -0.1024, -0.0598, -0.0742, -0.0406, -0.016, -0.0074, -0.0086, -0.0172, -0.0254, -0.0375, -0.0176, 0.0074, 0.0098, 0.0145, 0.0195, -0.007, -0.0453, -0.0367, -0.0469, -0.0539, -0.0512, -0.0449, -0.0469, -0.0043, -0.0063, -0.0258, -0.0129, -0.0625, -0.0035, 0.0, 0.0172, -0.0168, -0.0336, -0.0074, -0.0848, -0.0403, -0.0621, -0.0434, -0.0692, -0.0844, -0.0746, -0.0989, -0.1028, -0.0625, -0.0563, -0.05, -0.0328, -0.0348, -0.0414, -0.0735, -0.0793, -0.0668, -0.0848, -0.0977, -0.0699, -0.0688, -0.0567, -0.0637, -0.0641, -0.1004, -0.0973, -0.1645, -0.1731, -0.1626, -0.1504, -0.1336, -0.129, -0.1219, -0.1118, -0.1258, -0.0766, -0.0887, -0.0961, -0.0246, -0.0504, -0.0403, -0.0082, 0.0047, -0.0098, -0.0434, -0.0121, -0.0086, -0.0012, 0.0438, 0.0199, 0.1469, 0.1473, 0.1438, 0.1364, 0.0965, 0.1157, 0.0981, 0.1372, 0.134, 0.1075, 0.1348, 0.1278, 0.0825, 0.1274, 0.1493, 0.1567, 0.1458, 0.154, 0.1493, 0.1536, 0.1395, 0.1141, 0.1059, 0.0766, 0.0485, 0.0664, 0.077, 0.0578, 0.0457, 0.027, 0.0176, -0.0063, 0.009, 0.0164, -0.0403, -0.0277, -0.0262, 0.0129, 0.0152, 0.0152, 0.0141, 0.0266, 0.0625, 0.0746, 0.0907, 0.1477, 0.12, 0.0977, 0.0699, 0.0324, 0.0352, 0.0442, 0.0106, 0.0274, 0.0219, 0.0254, 0.0309, 0.0281, 0.0281, 0.0281, 0.0324, 0.0403, 0.0453, 0.0453, 0.1653, 0.2016, 0.2575, 0.2411, 0.1809, 0.1809, 0.2673]}
{"unit_id": "quote:1967742648636321850", "ticker_idx": 0, "ts": "2025-09-16T00:09:40+00:00", "symbol": "ASML", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This is bad news for ASML", "tweet_type": "analysis", "timeline": "unspecified", "summary": "TSMC 2nm keeps EUV layer count same as 3nm, boosting customer adoption; bearish ASML on reduced EUV intensity, bullish TSMC on rising 2nm demand.", "resolved_tickers": ["ASML"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting. According to local reports in Taiwan, although TSMC’s 2nm adopts a GAA structure, the number of EUV layers is the same as in the 3nm process, so it’s not that expensive.\n\nAs a result, customers’ willingness to adopt it has clearly increased.\n\nThis is bad news for ASML, but good news for TSMC.", "reply_text": null, "parent_text": null, "quoted_text": "Apple’s iPhone 18 will come with an Apple C2 modem chip for the first time, while the A20 processor inside will be made on TSMC’s 2nm manufacturing process and WMCM advanced packaging, media report, citing unnamed supply chain sources. The MacBook’s M6 chip will also be made on", "parent_author": null, "ret_m1d": -0.012659092575970865, "ret_signed_m1d": 0.012659092575970865, "alpha_spy_m1d": 0.014037783533015435, "alpha_c_m1d": 0.01252895367112461, "bench_spy_m1d": 0.0013786909570445705, "bench_c_m1d": -0.00013013890484625534, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007001156020913113, "ret_signed_p1d": 0.007001156020913113, "alpha_spy_p1d": 0.005758720728364475, "alpha_c_p1d": 0.0006211609813377983, "bench_spy_p1d": -0.0012424352925486382, "bench_c_p1d": -0.006379995039575315, "ret_p1w": 0.09686707855004939, "ret_signed_p1w": -0.09686707855004939, "alpha_spy_p1w": -0.08921745612093068, "alpha_c_p1w": -0.049700594121029784, "bench_spy_p1w": 0.007649622429118708, "bench_c_p1w": 0.047166484429019606, "ret_p1m": 0.14957534567120812, "ret_signed_p1m": -0.14957534567120812, "alpha_spy_p1m": -0.13894787015013055, "alpha_c_p1m": -0.0370464468127536, "bench_spy_p1m": 0.010627475521077567, "bench_c_p1m": 0.11252889885845452, "ret_p3m": 0.23263094294285902, "ret_signed_p3m": -0.23263094294285902, "alpha_spy_p3m": -0.19679747515692503, "alpha_c_p3m": -0.07993406325727292, "bench_spy_p3m": 0.03583346778593399, "bench_c_p3m": 0.1526968796855861, "ret_p6m": 0.5834936977651426, "ret_signed_p6m": -0.5834936977651426, "alpha_spy_p6m": -0.552873851069086, "alpha_c_p6m": -0.2740381297880423, "bench_spy_p6m": 0.030619846696056552, "bench_c_p6m": 0.3094555679771003, "price_path": [-0.1351, -0.1409, -0.1146, -0.0764, -0.0743, -0.0937, -0.0962, -0.09, -0.1024, -0.092, -0.0978, -0.0978, -0.1085, -0.0983, -0.0918, -0.0892, -0.0894, -0.0839, -0.0654, -0.1433, -0.1541, -0.1659, -0.1828, -0.1989, -0.1859, -0.1767, -0.1924, -0.1711, -0.1821, -0.1787, -0.2091, -0.2147, -0.2038, -0.2149, -0.2134, -0.1882, -0.1777, -0.1789, -0.1555, -0.1399, -0.1403, -0.1551, -0.149, -0.1535, -0.1468, -0.1628, -0.1406, -0.1411, -0.1312, -0.1234, -0.1309, -0.1546, -0.1546, -0.1737, -0.1612, -0.1423, -0.1101, -0.0935, -0.0834, -0.0971, -0.0845, -0.0735, -0.0127, 0.0, -0.007, 0.0562, 0.0612, 0.0904, 0.0969, 0.078, 0.081, 0.0832, 0.0958, 0.1021, 0.1421, 0.1728, 0.1751, 0.1877, 0.141, 0.1245, 0.1163, 0.0658, 0.1209, 0.1193, 0.1496, 0.1607, 0.1717, 0.1864, 0.1669, 0.1516, 0.1799, 0.1761, 0.2067, 0.1982, 0.2212, 0.2265, 0.208, 0.2166, 0.1748, 0.1903, 0.1737, 0.1598, 0.1847, 0.166, 0.183, 0.1631, 0.1484, 0.1632, 0.1451, 0.1853, 0.1188, 0.1023, 0.1265, 0.1441, 0.1872, 0.1872, 0.2089, 0.2408, 0.2645, 0.3011, 0.266, 0.2539, 0.2769, 0.2675, 0.2765, 0.2805, 0.2326, 0.2406, 0.2272, 0.158, 0.1818, 0.2043, 0.2054, 0.211, 0.2151, 0.2151, 0.2234, 0.2157, 0.2227, 0.2201, 0.2201, 0.3272, 0.4007, 0.4166, 0.401, 0.362, 0.4528, 0.4611, 0.4485, 0.4412, 0.5186, 0.5494, 0.5494, 0.5123, 0.5511, 0.5909, 0.5841, 0.6118, 0.6589, 0.6227, 0.6595, 0.6228, 0.6438, 0.5919, 0.5272, 0.5398, 0.6114, 0.6302, 0.6143, 0.6394, 0.6065, 0.6063, 0.6063, 0.6213, 0.6772, 0.666, 0.6782, 0.6969, 0.7104, 0.7432, 0.6716, 0.6564, 0.6256, 0.5541, 0.598, 0.5626, 0.4763, 0.5501, 0.5797, 0.5835]}
{"unit_id": "quote:1967742648636321850", "ticker_idx": 1, "ts": "2025-09-16T00:09:40+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "good news for TSMC", "tweet_type": "analysis", "timeline": "unspecified", "summary": "TSMC 2nm keeps EUV layer count same as 3nm, boosting customer adoption; bearish ASML on reduced EUV intensity, bullish TSMC on rising 2nm demand.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting. According to local reports in Taiwan, although TSMC’s 2nm adopts a GAA structure, the number of EUV layers is the same as in the 3nm process, so it’s not that expensive.\n\nAs a result, customers’ willingness to adopt it has clearly increased.\n\nThis is bad news for ASML, but good news for TSMC.", "reply_text": null, "parent_text": null, "quoted_text": "Apple’s iPhone 18 will come with an Apple C2 modem chip for the first time, while the A20 processor inside will be made on TSMC’s 2nm manufacturing process and WMCM advanced packaging, media report, citing unnamed supply chain sources. The MacBook’s M6 chip will also be made on", "parent_author": null, "ret_m1d": -0.02343750374990272, "ret_signed_m1d": -0.02343750374990272, "alpha_spy_m1d": -0.024816194706947292, "alpha_c_m1d": -0.023307364845056466, "bench_spy_m1d": 0.0013786909570445705, "bench_c_m1d": -0.00013013890484625534, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011718703876197, "ret_signed_p1d": -0.011718703876197, "alpha_spy_p1d": -0.010476268583648363, "alpha_c_p1d": -0.005338708836621686, "bench_spy_p1d": -0.0012424352925486382, "bench_c_p1d": -0.006379995039575315, "ret_p1w": 0.04687500749980544, "ret_signed_p1w": 0.04687500749980544, "alpha_spy_p1w": 0.039225385070686736, "alpha_c_p1w": -0.0002914769292141628, "bench_spy_p1w": 0.007649622429118708, "bench_c_p1w": 0.047166484429019606, "ret_p1m": 0.14453116112730635, "ret_signed_p1m": 0.14453116112730635, "alpha_spy_p1m": 0.13390368560622878, "alpha_c_p1m": 0.03200226226885183, "bench_spy_p1m": 0.010627475521077567, "bench_c_p1m": 0.11252889885845452, "ret_p3m": 0.16010416497671054, "ret_signed_p3m": 0.16010416497671054, "alpha_spy_p3m": 0.12427069719077655, "alpha_c_p3m": 0.0074072852911244436, "bench_spy_p3m": 0.03583346778593399, "bench_c_p3m": 0.1526968796855861, "ret_p6m": 0.5206770473892661, "ret_signed_p6m": 0.5206770473892661, "alpha_spy_p6m": 0.49005720069320957, "alpha_c_p6m": 0.21122147941216585, "bench_spy_p6m": 0.030619846696056552, "bench_c_p6m": 0.3094555679771003, "price_path": [-0.1946, -0.1791, -0.2063, -0.183, -0.1674, -0.1635, -0.1596, -0.1752, -0.1557, -0.1557, -0.1518, -0.1557, -0.1596, -0.1596, -0.1518, -0.144, -0.144, -0.1479, -0.1363, -0.1207, -0.1207, -0.1013, -0.1051, -0.1207, -0.109, -0.109, -0.109, -0.109, -0.1168, -0.1013, -0.0974, -0.0974, -0.1168, -0.1051, -0.1246, -0.0818, -0.0857, -0.0818, -0.0818, -0.0662, -0.0857, -0.0818, -0.0818, -0.0779, -0.1168, -0.1051, -0.1168, -0.0896, -0.0857, -0.074, -0.0974, -0.0974, -0.0935, -0.0974, -0.0974, -0.0974, -0.0818, -0.0818, -0.0662, -0.0468, -0.0351, -0.0195, -0.0234, 0.0, -0.0117, 0.0039, -0.0117, 0.0117, 0.0469, 0.0469, 0.0312, 0.0156, 0.0156, 0.0195, 0.0352, 0.0664, 0.0938, 0.0938, 0.1211, 0.1055, 0.125, 0.125, 0.1055, 0.1133, 0.1445, 0.1602, 0.1328, 0.1563, 0.1563, 0.1406, 0.1328, 0.1328, 0.1563, 0.1523, 0.1758, 0.1758, 0.1719, 0.1797, 0.1758, 0.1406, 0.1445, 0.1406, 0.1523, 0.1445, 0.1523, 0.1406, 0.1172, 0.1289, 0.0977, 0.0898, 0.1367, 0.082, 0.0742, 0.1055, 0.125, 0.1211, 0.125, 0.1016, 0.1172, 0.1328, 0.1289, 0.1406, 0.168, 0.1563, 0.1758, 0.1523, 0.1601, 0.1366, 0.1248, 0.1209, 0.1209, 0.1209, 0.1483, 0.1679, 0.1719, 0.1719, 0.1836, 0.1993, 0.1915, 0.215, 0.215, 0.2424, 0.309, 0.3365, 0.313, 0.3208, 0.3169, 0.3247, 0.3404, 0.3404, 0.3247, 0.3639, 0.3796, 0.3913, 0.3639, 0.3796, 0.3874, 0.3757, 0.3953, 0.4266, 0.4149, 0.3913, 0.3835, 0.4109, 0.3992, 0.3835, 0.3953, 0.4227, 0.4736, 0.5011, 0.5011, 0.5011, 0.5011, 0.5011, 0.5011, 0.5011, 0.5011, 0.4893, 0.5403, 0.5795, 0.5638, 0.5638, 0.5481, 0.5168, 0.4619, 0.4893, 0.4815, 0.4188, 0.4501, 0.5207]}
{"unit_id": "thread:1970317006529503505", "ticker_idx": 0, "ts": "2025-09-23T02:39:15+00:00", "symbol": "Nanya Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nanya's DDR4 output through year-end is already sold out... August revenue reached NT$6.763 billion (US$224 million), up 141% year-over-year, while net profit rose 210%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays Digitimes analysis: DDR4 tightness benefits Nanya and Winbond; DDR5/LPDDR5X prices seen rising 5–10% per quarter from Q4 into 2026.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya Technology Corp listed on TWSE as 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "China policy containment eases DDR5 price fears as Taiwan's DDR4 supply sells out- Digitimes\n\nChina's AI-driven supply chain expansion is reshaping the DRAM landscape. CXMT, the country's top DRAM producer, will pivot entirely to DDR5, LPDDR5, and high-bandwidth memory (HBM) in 2025. While fears of oversupply had raised concerns of DDR5 price swings, industry sources say Beijing is ring-fencing advanced DRAM for domestic use. This policy curbs export leakage and is expected to stabilize DDR5 pricing, with a gradual market upturn likely in 2026.\n\nDDR4 supply tightens as major players exit\nGlobal DRAM makers are phasing out DDR4, with CXMT joining the retreat. The contraction has tightened supply, benefiting Taiwanese suppliers Nanya Technology and Winbond Electronics. Even as DDR4 contract prices surge, supply-chain executives caution that quotes are largely academic — Nanya's DDR4 output through year-end is already sold out.\n\nWith only residual lines reserved for servers and industrial clients, DDR4 supply is critically tight. Taiwanese vendors cannot fully cover demand, leaving a supply gap projected at over 10% into early 2026.\n\nNanya's capacity is in high demand. DDR4 8Gb prices jumped 50% to US$3.9 by late July and further to about US$5.5 in September — over 40% higher. August revenue reached NT$6.763 billion (US$224 million), up 141% year-over-year, while net profit rose 210% to NT$811 million, marking a sharp rebound from losses.\n\nFourth-quarter contract prices are not finalized, but industry forecasts point to a 20% rise, slower than in the third quarter. With nearly all supplies already booked, buyers are scrambling to secure allocations ahead of time.\n\nDDR5 and LPDDR5X outlook\nDDR4's surge contrasts with DDR5's relative stability. Spot prices for DDR4 16Gb have overtaken DDR5 equivalents, reflecting the shift in capacity toward DDR5 and stable mainstream supply. However, CXMT's ramp into DDR5 raises concerns about pricing, given its history of undercutting DDR4 with steep discounts.\n\nCXMT's DDR5, slated for 2025, will use 1Y/1Z nodes comparable to top global peers, though yields will take time to mature. Capacity could hit 300,000 wafers per month by late 2025, fueling worries of oversupply and price pressure into 2026.\n\nNevertheless, sources say China's localization policy is keeping most of CXMT's output at home. Ongoing work to raise HBM2 yields and sample HBM3 by late 2025, coupled with rising LPDDR5X demand from edge AI and cloud providers, reduces the risk of near-term DDR5 oversupply. Smartphone demand is weak, but embedded and AI workloads are driving strong order growth.\n\nPrices set to climb into 2026\nIndustry insiders expect no significant price cuts for DDR5 or LPDDR5X. Despite yield challenges, pricing remains profitable, and tightening supply has already pushed spot prices higher in the latter half of the third quarter.\n\nDDR5's rally should be more measured than DDR4's, with quarterly gains of 5–10% expected from the fourth quarter. LPDDR4 remains tight due to mid-tier smartphone adoption, while the industry's shift to LPDDR5X in 2026 is set to add further upward momentum.\n\n---\n\nhttps://t.co/XyeaNWVIS3", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0024690981264468004, "ret_signed_m1d": -0.0024690981264468004, "alpha_spy_m1d": -0.00794236656529268, "alpha_c_m1d": -0.003930068761726746, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06172839506172845, "ret_signed_p1d": -0.06172839506172845, "alpha_spy_p1d": -0.05854676160911476, "alpha_c_p1d": -0.05989430738997781, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.09876543209876543, "ret_signed_p1w": -0.09876543209876543, "alpha_spy_p1w": -0.10324356082145747, "alpha_c_p1w": -0.11325086126471806, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.3580246913580247, "ret_signed_p1m": 0.3580246913580247, "alpha_spy_p1m": 0.35110387965647005, "alpha_c_p1m": 0.3054913161295132, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 1.1543209876543208, "ret_signed_p3m": 1.1543209876543208, "alpha_spy_p3m": 1.1250828350416116, "alpha_c_p3m": 1.0469849890542948, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 2.4074074074074074, "ret_signed_p6m": 2.4074074074074074, "alpha_spy_p6m": 2.40714446172906, "alpha_c_p6m": 2.1798819517569017, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.3346, -0.3444, -0.3679, -0.3889, -0.3895, -0.3753, -0.4019, -0.3963, -0.4012, -0.4025, -0.4179, -0.4698, -0.4642, -0.4846, -0.4889, -0.4679, -0.4784, -0.4821, -0.4994, -0.4685, -0.4691, -0.466, -0.4654, -0.479, -0.4457, -0.45, -0.45, -0.4599, -0.4519, -0.4648, -0.479, -0.4574, -0.4593, -0.4142, -0.4315, -0.4278, -0.4074, -0.4154, -0.4352, -0.45, -0.4265, -0.4401, -0.4204, -0.4142, -0.4099, -0.4241, -0.4198, -0.4228, -0.4364, -0.4278, -0.4049, -0.3457, -0.3395, -0.3346, -0.3284, -0.3025, -0.2914, -0.221, -0.1457, -0.0951, -0.0123, -0.0222, -0.0025, 0.0, -0.0617, -0.0543, -0.1235, -0.1235, -0.0988, -0.0593, -0.016, 0.0247, 0.0247, 0.1259, 0.121, 0.216, 0.216, 0.1802, 0.1, 0.0654, 0.1716, 0.284, 0.321, 0.3025, 0.358, 0.3519, 0.3519, 0.4815, 0.6049, 0.642, 0.6605, 0.6358, 0.6914, 0.6235, 0.7037, 0.8272, 0.8148, 0.9938, 1.0309, 1.0123, 1.0185, 0.9568, 1.0556, 0.9815, 0.9753, 0.9198, 0.7284, 0.7654, 0.7716, 0.6852, 0.8025, 0.8025, 0.8457, 0.8457, 0.8642, 0.8642, 0.8889, 1.0185, 1.0, 0.9259, 0.9259, 1.0, 1.0062, 0.9506, 1.0309, 1.0988, 1.1543, 1.2222, 1.179, 1.3333, 1.3333, 1.3333, 1.3272, 1.4012, 1.3827, 1.3827, 1.5556, 1.5679, 1.821, 1.9753, 1.9815, 1.6852, 1.9506, 1.8395, 1.9198, 1.9321, 2.0864, 2.3951, 2.358, 2.0988, 2.2901, 2.3519, 2.5309, 2.4198, 2.6667, 2.6667, 3.0309, 2.6296, 2.4259, 2.642, 2.3951, 2.2593, 2.3889, 2.3457, 2.4321, 2.4321, 2.4321, 2.4321, 2.4321, 2.4321, 2.4321, 2.4321, 2.5679, 2.679, 2.5556, 2.5247, 2.5247, 2.5185, 2.1667, 1.8951, 2.179, 1.8765, 1.5926, 1.8519, 2.1358, 1.9568, 1.9383, 2.2284, 2.3519, 2.4074]}
{"unit_id": "thread:1970317006529503505", "ticker_idx": 1, "ts": "2025-09-23T02:39:15+00:00", "symbol": "Winbond Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The contraction has tightened supply, benefiting Taiwanese suppliers Nanya Technology and Winbond Electronics", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays Digitimes analysis: DDR4 tightness benefits Nanya and Winbond; DDR5/LPDDR5X prices seen rising 5–10% per quarter from Q4 into 2026.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "China policy containment eases DDR5 price fears as Taiwan's DDR4 supply sells out- Digitimes\n\nChina's AI-driven supply chain expansion is reshaping the DRAM landscape. CXMT, the country's top DRAM producer, will pivot entirely to DDR5, LPDDR5, and high-bandwidth memory (HBM) in 2025. While fears of oversupply had raised concerns of DDR5 price swings, industry sources say Beijing is ring-fencing advanced DRAM for domestic use. This policy curbs export leakage and is expected to stabilize DDR5 pricing, with a gradual market upturn likely in 2026.\n\nDDR4 supply tightens as major players exit\nGlobal DRAM makers are phasing out DDR4, with CXMT joining the retreat. The contraction has tightened supply, benefiting Taiwanese suppliers Nanya Technology and Winbond Electronics. Even as DDR4 contract prices surge, supply-chain executives caution that quotes are largely academic — Nanya's DDR4 output through year-end is already sold out.\n\nWith only residual lines reserved for servers and industrial clients, DDR4 supply is critically tight. Taiwanese vendors cannot fully cover demand, leaving a supply gap projected at over 10% into early 2026.\n\nNanya's capacity is in high demand. DDR4 8Gb prices jumped 50% to US$3.9 by late July and further to about US$5.5 in September — over 40% higher. August revenue reached NT$6.763 billion (US$224 million), up 141% year-over-year, while net profit rose 210% to NT$811 million, marking a sharp rebound from losses.\n\nFourth-quarter contract prices are not finalized, but industry forecasts point to a 20% rise, slower than in the third quarter. With nearly all supplies already booked, buyers are scrambling to secure allocations ahead of time.\n\nDDR5 and LPDDR5X outlook\nDDR4's surge contrasts with DDR5's relative stability. Spot prices for DDR4 16Gb have overtaken DDR5 equivalents, reflecting the shift in capacity toward DDR5 and stable mainstream supply. However, CXMT's ramp into DDR5 raises concerns about pricing, given its history of undercutting DDR4 with steep discounts.\n\nCXMT's DDR5, slated for 2025, will use 1Y/1Z nodes comparable to top global peers, though yields will take time to mature. Capacity could hit 300,000 wafers per month by late 2025, fueling worries of oversupply and price pressure into 2026.\n\nNevertheless, sources say China's localization policy is keeping most of CXMT's output at home. Ongoing work to raise HBM2 yields and sample HBM3 by late 2025, coupled with rising LPDDR5X demand from edge AI and cloud providers, reduces the risk of near-term DDR5 oversupply. Smartphone demand is weak, but embedded and AI workloads are driving strong order growth.\n\nPrices set to climb into 2026\nIndustry insiders expect no significant price cuts for DDR5 or LPDDR5X. Despite yield challenges, pricing remains profitable, and tightening supply has already pushed spot prices higher in the latter half of the third quarter.\n\nDDR5's rally should be more measured than DDR4's, with quarterly gains of 5–10% expected from the fourth quarter. LPDDR4 remains tight due to mid-tier smartphone adoption, while the industry's shift to LPDDR5X in 2026 is set to add further upward momentum.\n\n---\n\nhttps://t.co/XyeaNWVIS3", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06035664212583547, "ret_signed_m1d": -0.06035664212583547, "alpha_spy_m1d": -0.06582991056468135, "alpha_c_m1d": -0.061817612761115415, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03429362078308684, "ret_signed_p1d": -0.03429362078308684, "alpha_spy_p1d": -0.031111987330473156, "alpha_c_p1d": -0.0324595331113362, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.06584363828369677, "ret_signed_p1w": -0.06584363828369677, "alpha_spy_p1w": -0.07032176700638881, "alpha_c_p1w": -0.0803290674496494, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.24691356466120307, "ret_signed_p1m": 0.24691356466120307, "alpha_spy_p1m": 0.2399927529596484, "alpha_c_p1m": 0.19438018943269153, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.9561042411442022, "ret_signed_p3m": 0.9561042411442022, "alpha_spy_p3m": 0.926866088531493, "alpha_c_p3m": 0.8487682425441763, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 2.5116596564283626, "ret_signed_p6m": 2.5116596564283626, "alpha_spy_p6m": 2.511396710750015, "alpha_c_p6m": 2.2841342007778564, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.428, -0.4198, -0.4472, -0.4554, -0.4636, -0.4582, -0.4746, -0.476, -0.4815, -0.4801, -0.4925, -0.4938, -0.4993, -0.5089, -0.5158, -0.5089, -0.5158, -0.5226, -0.5295, -0.5171, -0.524, -0.5213, -0.5199, -0.5226, -0.5103, -0.524, -0.524, -0.524, -0.5171, -0.5322, -0.5473, -0.5213, -0.5226, -0.4938, -0.5034, -0.5062, -0.4952, -0.476, -0.4897, -0.5062, -0.4952, -0.5007, -0.4952, -0.4856, -0.4568, -0.4623, -0.4595, -0.4431, -0.4527, -0.4486, -0.4376, -0.3813, -0.3251, -0.3265, -0.3155, -0.3141, -0.3018, -0.2647, -0.2332, -0.1934, -0.1139, -0.0809, -0.0604, 0.0, -0.0343, -0.0837, -0.1385, -0.1385, -0.0658, -0.0727, -0.0123, 0.0645, 0.0645, 0.1701, 0.1317, 0.192, 0.192, 0.1797, 0.1166, 0.0974, 0.2071, 0.2058, 0.2716, 0.2236, 0.2469, 0.2702, 0.2702, 0.3964, 0.4815, 0.4815, 0.5144, 0.487, 0.5528, 0.4623, 0.5364, 0.6187, 0.594, 0.7531, 0.7695, 0.7778, 0.7613, 0.6543, 0.8162, 0.6845, 0.6296, 0.5912, 0.4321, 0.4979, 0.5775, 0.4705, 0.561, 0.5912, 0.5802, 0.5528, 0.5309, 0.5857, 0.6955, 0.8628, 0.9561, 0.8519, 0.8765, 1.0466, 0.9506, 0.9369, 1.0274, 1.0439, 0.9561, 1.0055, 1.0027, 1.1097, 1.1097, 1.0988, 1.1097, 1.2963, 1.2661, 1.2661, 1.4911, 1.62, 1.8532, 1.9218, 1.9767, 1.6831, 1.7984, 1.738, 1.738, 1.7215, 1.9218, 2.1962, 2.1001, 1.8944, 1.8669, 1.8258, 2.1001, 2.1687, 2.4842, 2.5665, 2.5254, 2.1824, 1.8944, 1.9767, 1.8121, 1.6831, 1.9492, 1.8395, 1.8944, 1.8944, 1.8944, 1.8944, 1.8944, 1.8944, 1.8944, 1.8944, 2.155, 2.3471, 2.2373, 2.3608, 2.3608, 2.4156, 2.0864, 1.8532, 2.0727, 1.9218, 1.7709, 1.8944, 2.1276, 1.9904, 1.9904, 2.2099, 2.3882, 2.5117]}
{"unit_id": "reply:1969675493596414261", "ticker_idx": 0, "ts": "2025-09-21T08:10:07+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "We believe the pause in Lithography intensity is temporary, and as such see an increase in EUV exposed layers in the subsequent nodes post the 2nm process", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Relays GS thesis that EUV intensity pause is temporary; more EUV layers expected post-2nm, bullish for ASML.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "We believe the pause in Lithography intensity is temporary, and as such see an increase in EUV exposed layers in the subsequent nodes post the 2nm process (given it is materially difficult to execute an architectural transition in tandem with incremental shrinkage- GS\n\nhttps://t.co/NMPPqCIjBX", "reply_text": null, "parent_text": "N2 and N3 have same number of EUV layers\nASML will benefit as overall demand expands\nHigh NA adoption is key for ASML\n\nRn the biggest opportunity is to bet on the suppliers of TSMC's Advanced Packaging", "quoted_text": null, "parent_author": "zephyr_z9", "ret_m1d": -0.02109022638613367, "ret_signed_m1d": -0.02109022638613367, "alpha_spy_m1d": -0.016381495461561313, "alpha_c_m1d": -0.0010386686311859572, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011100103218230739, "ret_signed_p1d": 0.011100103218230739, "alpha_spy_p1d": 0.0165435780580383, "alpha_c_p1d": 0.01255894253211165, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.018130279585259368, "ret_signed_p1w": 0.018130279585259368, "alpha_spy_p1w": 0.02286901884312753, "alpha_c_p1w": 0.016609392328177597, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.09743463457121293, "ret_signed_p1m": 0.09743463457121293, "alpha_spy_p1m": 0.09076146241699479, "alpha_c_p1m": 0.025640318276833884, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.10295464172134006, "ret_signed_p3m": 0.10295464172134006, "alpha_spy_p3m": 0.08851345349811957, "alpha_c_p3m": 0.025107562820681384, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 0.476951397071127, "ret_signed_p6m": 0.476951397071127, "alpha_spy_p6m": 0.46805558315077667, "alpha_c_p6m": 0.24122204392806146, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.1457, -0.1667, -0.1605, -0.1665, -0.1781, -0.1665, -0.1671, -0.1889, -0.1739, -0.1649, -0.1638, -0.1512, -0.1587, -0.1546, -0.1314, -0.2302, -0.2002, -0.2207, -0.2323, -0.259, -0.2538, -0.2424, -0.2559, -0.2251, -0.2316, -0.2261, -0.2438, -0.2662, -0.257, -0.2636, -0.2687, -0.2468, -0.2391, -0.2267, -0.2182, -0.2099, -0.2071, -0.2147, -0.2123, -0.2078, -0.2129, -0.2174, -0.203, -0.2024, -0.1976, -0.1857, -0.193, -0.2148, -0.2155, -0.2388, -0.2288, -0.2013, -0.1857, -0.1655, -0.1574, -0.1611, -0.1539, -0.1492, -0.1004, -0.0961, -0.091, -0.0207, -0.0211, 0.0, 0.0111, -0.0035, 0.0011, 0.0025, 0.0181, 0.0213, 0.0377, 0.0824, 0.0855, 0.1067, 0.0789, 0.0506, 0.0481, 0.0038, 0.0409, 0.0442, 0.0767, 0.0821, 0.08, 0.1088, 0.0974, 0.0731, 0.0948, 0.1031, 0.12, 0.1211, 0.1359, 0.1583, 0.1343, 0.1447, 0.1342, 0.1201, 0.1062, 0.0759, 0.0955, 0.0958, 0.1018, 0.0891, 0.08, 0.0828, 0.0695, 0.0954, 0.0994, 0.0304, 0.0611, 0.0629, 0.1235, 0.1085, 0.1162, 0.145, 0.1603, 0.1903, 0.1828, 0.1757, 0.1901, 0.1773, 0.1688, 0.1625, 0.1432, 0.1503, 0.1226, 0.0799, 0.103, 0.114, 0.107, 0.1159, 0.1107, 0.1107, 0.1107, 0.1204, 0.1347, 0.1384, 0.1384, 0.2186, 0.3013, 0.3109, 0.2995, 0.2516, 0.3363, 0.3423, 0.3623, 0.3396, 0.4201, 0.4421, 0.3843, 0.4085, 0.4268, 0.4535, 0.4557, 0.4555, 0.5044, 0.4757, 0.4727, 0.5019, 0.5133, 0.4708, 0.4092, 0.4204, 0.475, 0.4906, 0.476, 0.4943, 0.4596, 0.4727, 0.4784, 0.4836, 0.54, 0.5319, 0.5534, 0.5455, 0.5631, 0.594, 0.5247, 0.5259, 0.4975, 0.4374, 0.4844, 0.4673, 0.4191, 0.4198, 0.4846, 0.4831, 0.4604, 0.4594, 0.4814, 0.477]}
{"unit_id": "orig:1946461936285278442", "ticker_idx": 0, "ts": "2025-07-19T06:47:43+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML will face challenges in 2026, with an expected flat growth from 2024-2026 due to a slowdown in EUV spending", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley flags ASML 2026 flat growth on EUV slowdown, TSMC flat CapEx, and WFE sector headwinds into summer 2025.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "ASML: Facing Challenges in 2026, Slowdown in EUV Spending, TSMC CapEx Remains Flat\nMorgan Stanley (L. Simpson, 25/07/17)\n\nMorgan Stanley notes that ASML will face challenges in 2026, with an expected flat growth from 2024-2026 due to a slowdown in EUV spending. While TSMC's sales are strong (expected to grow over 30% year-on-year in 2025), the company maintains a cautious outlook on its 2026 capital expenditure (CapEx). ASML expects a reduction in EUV equipment shipments in 2026, specifically only 21 units of TSMC's EUV Low Numerical Aperture (LNA) tools.\n\nASML Faces Slowdown in EUV Spending: According to Morgan Stanley's analysis, ASML will experience flat growth between 2024 and 2026, mainly due to a slowdown in EUV spending caused by several factors. These factors include fewer layers, limited options for fabs other than TSMC, slow adoption of High Numerical Aperture (HNA), and the end of the tool upgrade cycle. The uncertainty surrounding Samsung's HBM4 approval by NVIDIA is also expected to exacerbate this trend.\n\nTSMC CapEx Outlook Remains Flat: Despite strong N2 demand, TSMC's capital expenditure is expected to remain around $40 billion in 2026, consistent with 2025. Morgan Stanley believes that even with TSMC's strong sales growth, its CapEx will not increase in sync with sales growth due to potentially decreasing capital intensity. TSMC is expected to focus on deposition and packaging, while the demand growth for EUV equipment will be constrained by flat layer counts and bottlenecks in capacity expansion.\n\nExpected Decrease in EUV LNA Tool Shipments: Due to CapEx limitations, TSMC is expected to purchase only 21 units of EUV LNA tools from ASML in 2026, significantly lower than anticipated. Morgan Stanley points out that while TSMC plans some N2/A16 process development, 40 chip tape-outs are not expected to lead to a large-scale capacity expansion.\n\nTSMC's Future Capacity Expansion Limited: Morgan Stanley expects an increase in TSMC's N2/A16 capacity in 2026, with N2 capacity projected to rise from 40k wafers/month in 2025 to 50k wafers/month. However, due to the lack of significant growth in EUV spending, TSMC's overall capacity expansion in 2026 is still expected to remain at a low level. Despite strong demand, TSMC may not significantly expand its EUV equipment capacity.\n\nWFE Sector Faces Challenges: Morgan Stanley believes that summer 2025 may pose challenges for Wafer Fab Equipment (WFE) companies, especially considering ASML's uncertain guidance for the 2026 fiscal year and the uncertainty in HBM spending. Despite TSMC's strong sales growth in 2025, WFE companies may face challenges due to a cautious outlook on capital expenditure, particularly against the backdrop of slowing EUV spending.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015061663527752067, "ret_signed_m1d": -0.015061663527752067, "alpha_spy_m1d": -0.01695425001791695, "alpha_c_m1d": -0.01575007941566997, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.034769939171927944, "ret_signed_p1d": 0.034769939171927944, "alpha_spy_p1d": 0.03491306033515251, "alpha_c_p1d": 0.01707692376386738, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": 0.009399590799242796, "ret_signed_p1w": -0.009399590799242796, "alpha_spy_p1w": 0.0035940878855227343, "alpha_c_p1w": -0.006301561240879661, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.031891338518251544, "ret_signed_p1m": -0.031891338518251544, "alpha_spy_p1m": -0.014333249876912868, "alpha_c_p1m": -0.030652126694906245, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.4095930024797507, "ret_signed_p3m": -0.4095930024797507, "alpha_spy_p3m": -0.3559937352757694, "alpha_c_p3m": -0.22787818410910976, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.7745160888057321, "ret_signed_p6m": -0.7745160888057321, "alpha_spy_p6m": -0.6648111719476744, "alpha_c_p6m": -0.422589203989312, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.0721, -0.0549, -0.0564, -0.0635, -0.0627, -0.0667, -0.0667, -0.0244, -0.0277, -0.0362, -0.0311, 0.0103, 0.0043, 0.0686, 0.0955, 0.1001, 0.0931, 0.071, 0.0614, 0.0575, 0.0651, 0.0524, 0.0296, 0.0479, 0.0673, 0.0588, 0.0663, 0.0477, 0.0354, 0.0465, 0.0453, 0.0559, 0.0697, 0.0897, 0.0968, 0.1059, 0.0841, 0.0649, 0.0737, 0.0671, 0.0655, 0.0465, 0.0421, 0.0715, 0.1078, 0.1128, 0.0854, 0.0936, 0.0857, 0.0707, 0.0857, 0.0849, 0.0566, 0.0761, 0.0878, 0.0892, 0.1056, 0.0958, 0.1013, 0.1314, 0.0027, 0.0418, 0.0151, 0.0, -0.0348, -0.028, -0.0131, -0.0308, 0.0094, 0.0009, 0.0081, -0.015, -0.0441, -0.0322, -0.0407, -0.0475, -0.0189, -0.0089, 0.0073, 0.0184, 0.0292, 0.0329, 0.0229, 0.0261, 0.0319, 0.0253, 0.0194, 0.0382, 0.039, 0.0452, 0.0606, 0.0512, 0.0227, 0.0219, -0.0084, 0.0046, 0.0404, 0.0606, 0.087, 0.0976, 0.0928, 0.1021, 0.1082, 0.1718, 0.1774, 0.184, 0.2756, 0.2751, 0.3026, 0.3171, 0.2981, 0.304, 0.3058, 0.3262, 0.3304, 0.3518, 0.4099, 0.4139, 0.4416, 0.4054, 0.3685, 0.3653, 0.3076, 0.3559, 0.3601, 0.4025, 0.4096, 0.4069, 0.4443, 0.4295, 0.3979, 0.4261, 0.4369, 0.4589, 0.4604, 0.4797, 0.5088, 0.4776, 0.4911, 0.4774, 0.4591, 0.4409, 0.4015, 0.427, 0.4274, 0.4353, 0.4187, 0.4068, 0.4105, 0.3931, 0.4269, 0.432, 0.3422, 0.3821, 0.3846, 0.4634, 0.4439, 0.4539, 0.4914, 0.5114, 0.5505, 0.5407, 0.5315, 0.5502, 0.5336, 0.5225, 0.5143, 0.4892, 0.4983, 0.4623, 0.4066, 0.4367, 0.451, 0.442, 0.4536, 0.4468, 0.4468, 0.4468, 0.4594, 0.4781, 0.4829, 0.4829, 0.5873, 0.695, 0.7076, 0.6928, 0.6303, 0.7407, 0.7484, 0.7745]}
{"unit_id": "orig:1946461936285278442", "ticker_idx": 1, "ts": "2025-07-19T06:47:43+00:00", "symbol": "WFE", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "summer 2025 may pose challenges for Wafer Fab Equipment (WFE) companies", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley flags ASML 2026 flat growth on EUV slowdown, TSMC flat CapEx, and WFE sector headwinds into summer 2025.", "resolved_tickers": ["AMAT", "LRCX", "KLAC", "ASML", "8035.T"], "resolution_method": "basket", "resolution_reason": "Wafer Fab Equipment basket: AMAT, Lam, KLA, ASML, TEL", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "ASML: Facing Challenges in 2026, Slowdown in EUV Spending, TSMC CapEx Remains Flat\nMorgan Stanley (L. Simpson, 25/07/17)\n\nMorgan Stanley notes that ASML will face challenges in 2026, with an expected flat growth from 2024-2026 due to a slowdown in EUV spending. While TSMC's sales are strong (expected to grow over 30% year-on-year in 2025), the company maintains a cautious outlook on its 2026 capital expenditure (CapEx). ASML expects a reduction in EUV equipment shipments in 2026, specifically only 21 units of TSMC's EUV Low Numerical Aperture (LNA) tools.\n\nASML Faces Slowdown in EUV Spending: According to Morgan Stanley's analysis, ASML will experience flat growth between 2024 and 2026, mainly due to a slowdown in EUV spending caused by several factors. These factors include fewer layers, limited options for fabs other than TSMC, slow adoption of High Numerical Aperture (HNA), and the end of the tool upgrade cycle. The uncertainty surrounding Samsung's HBM4 approval by NVIDIA is also expected to exacerbate this trend.\n\nTSMC CapEx Outlook Remains Flat: Despite strong N2 demand, TSMC's capital expenditure is expected to remain around $40 billion in 2026, consistent with 2025. Morgan Stanley believes that even with TSMC's strong sales growth, its CapEx will not increase in sync with sales growth due to potentially decreasing capital intensity. TSMC is expected to focus on deposition and packaging, while the demand growth for EUV equipment will be constrained by flat layer counts and bottlenecks in capacity expansion.\n\nExpected Decrease in EUV LNA Tool Shipments: Due to CapEx limitations, TSMC is expected to purchase only 21 units of EUV LNA tools from ASML in 2026, significantly lower than anticipated. Morgan Stanley points out that while TSMC plans some N2/A16 process development, 40 chip tape-outs are not expected to lead to a large-scale capacity expansion.\n\nTSMC's Future Capacity Expansion Limited: Morgan Stanley expects an increase in TSMC's N2/A16 capacity in 2026, with N2 capacity projected to rise from 40k wafers/month in 2025 to 50k wafers/month. However, due to the lack of significant growth in EUV spending, TSMC's overall capacity expansion in 2026 is still expected to remain at a low level. Despite strong demand, TSMC may not significantly expand its EUV equipment capacity.\n\nWFE Sector Faces Challenges: Morgan Stanley believes that summer 2025 may pose challenges for Wafer Fab Equipment (WFE) companies, especially considering ASML's uncertain guidance for the 2026 fiscal year and the uncertainty in HBM spending. Despite TSMC's strong sales growth in 2025, WFE companies may face challenges due to a cautious outlook on capital expenditure, particularly against the backdrop of slowing EUV spending.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.001651729316419881, "ret_signed_m1d": 0.001651729316419881, "alpha_spy_m1d": -0.0002408571737450016, "alpha_c_m1d": 0.0009633134285019773, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029921285827401833, "ret_signed_p1d": 0.029921285827401833, "alpha_spy_p1d": 0.030064406990626402, "alpha_c_p1d": 0.01222827041934127, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.012548121060811334, "ret_signed_p1w": 0.012548121060811334, "alpha_spy_p1w": 0.025541799745576865, "alpha_c_p1w": 0.01564615061917447, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": -0.08751375989615376, "ret_signed_p1m": 0.08751375989615376, "alpha_spy_p1m": 0.10507184853749243, "alpha_c_p1m": 0.08875297171949906, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.25589486946046835, "ret_signed_p3m": -0.25589486946046835, "alpha_spy_p3m": -0.20229560225648702, "alpha_c_p3m": -0.07418005108982739, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.7018525750416377, "ret_signed_p6m": -0.7018525750416377, "alpha_spy_p6m": -0.59214765818358, "alpha_c_p6m": -0.3499256902252176, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.2573, -0.2254, -0.215, -0.2167, -0.2224, -0.2158, -0.2236, -0.2011, -0.2058, -0.2128, -0.1918, -0.1847, -0.1837, -0.1301, -0.0997, -0.0942, -0.0973, -0.1183, -0.1221, -0.1188, -0.1318, -0.139, -0.149, -0.1447, -0.1222, -0.1279, -0.1233, -0.1475, -0.1426, -0.134, -0.1308, -0.1186, -0.1061, -0.0883, -0.0698, -0.0597, -0.0539, -0.0799, -0.0553, -0.0584, -0.0641, -0.0694, -0.0824, -0.0718, -0.0328, -0.0219, -0.0153, -0.0101, -0.006, -0.0131, 0.0011, 0.0049, 0.0058, -0.0014, 0.0112, 0.0134, 0.0174, 0.0194, 0.0115, 0.0309, 0.0086, 0.0043, -0.0017, 0.0, -0.0299, -0.0232, -0.0169, -0.0257, -0.0125, -0.0204, -0.0145, -0.0493, -0.0815, -0.0651, -0.0826, -0.0897, -0.0708, -0.0588, -0.0601, -0.0366, -0.0241, -0.0266, -0.0886, -0.0876, -0.0875, -0.0911, -0.1021, -0.0934, -0.0888, -0.0771, -0.0745, -0.0705, -0.094, -0.0968, -0.116, -0.1163, -0.097, -0.0716, -0.0606, -0.0508, -0.0466, -0.0128, -0.0004, 0.0275, 0.0393, 0.0539, 0.1109, 0.1144, 0.1555, 0.1569, 0.1505, 0.1502, 0.1483, 0.1595, 0.1684, 0.216, 0.2548, 0.2434, 0.283, 0.2261, 0.2218, 0.2209, 0.1606, 0.2058, 0.1991, 0.2467, 0.2559, 0.2545, 0.2859, 0.2746, 0.2508, 0.2806, 0.2972, 0.3271, 0.3227, 0.3605, 0.3583, 0.3548, 0.3714, 0.3415, 0.3716, 0.3489, 0.3303, 0.3708, 0.3396, 0.3482, 0.3134, 0.2819, 0.2976, 0.2656, 0.3018, 0.2503, 0.2383, 0.2751, 0.3031, 0.3312, 0.3385, 0.3486, 0.3515, 0.3797, 0.4111, 0.4048, 0.4, 0.4161, 0.4221, 0.4369, 0.4313, 0.3729, 0.3885, 0.3791, 0.3263, 0.3602, 0.3951, 0.4226, 0.4245, 0.4337, 0.4378, 0.447, 0.4371, 0.4285, 0.4135, 0.4135, 0.4923, 0.5811, 0.6334, 0.6164, 0.573, 0.6753, 0.6937, 0.7019]}
{"unit_id": "orig:1946461936285278442", "ticker_idx": 2, "ts": "2025-07-19T06:47:43+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's capital expenditure is expected to remain around $40 billion in 2026, consistent with 2025", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley flags ASML 2026 flat growth on EUV slowdown, TSMC flat CapEx, and WFE sector headwinds into summer 2025.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "ASML: Facing Challenges in 2026, Slowdown in EUV Spending, TSMC CapEx Remains Flat\nMorgan Stanley (L. Simpson, 25/07/17)\n\nMorgan Stanley notes that ASML will face challenges in 2026, with an expected flat growth from 2024-2026 due to a slowdown in EUV spending. While TSMC's sales are strong (expected to grow over 30% year-on-year in 2025), the company maintains a cautious outlook on its 2026 capital expenditure (CapEx). ASML expects a reduction in EUV equipment shipments in 2026, specifically only 21 units of TSMC's EUV Low Numerical Aperture (LNA) tools.\n\nASML Faces Slowdown in EUV Spending: According to Morgan Stanley's analysis, ASML will experience flat growth between 2024 and 2026, mainly due to a slowdown in EUV spending caused by several factors. These factors include fewer layers, limited options for fabs other than TSMC, slow adoption of High Numerical Aperture (HNA), and the end of the tool upgrade cycle. The uncertainty surrounding Samsung's HBM4 approval by NVIDIA is also expected to exacerbate this trend.\n\nTSMC CapEx Outlook Remains Flat: Despite strong N2 demand, TSMC's capital expenditure is expected to remain around $40 billion in 2026, consistent with 2025. Morgan Stanley believes that even with TSMC's strong sales growth, its CapEx will not increase in sync with sales growth due to potentially decreasing capital intensity. TSMC is expected to focus on deposition and packaging, while the demand growth for EUV equipment will be constrained by flat layer counts and bottlenecks in capacity expansion.\n\nExpected Decrease in EUV LNA Tool Shipments: Due to CapEx limitations, TSMC is expected to purchase only 21 units of EUV LNA tools from ASML in 2026, significantly lower than anticipated. Morgan Stanley points out that while TSMC plans some N2/A16 process development, 40 chip tape-outs are not expected to lead to a large-scale capacity expansion.\n\nTSMC's Future Capacity Expansion Limited: Morgan Stanley expects an increase in TSMC's N2/A16 capacity in 2026, with N2 capacity projected to rise from 40k wafers/month in 2025 to 50k wafers/month. However, due to the lack of significant growth in EUV spending, TSMC's overall capacity expansion in 2026 is still expected to remain at a low level. Despite strong demand, TSMC may not significantly expand its EUV equipment capacity.\n\nWFE Sector Faces Challenges: Morgan Stanley believes that summer 2025 may pose challenges for Wafer Fab Equipment (WFE) companies, especially considering ASML's uncertain guidance for the 2026 fiscal year and the uncertainty in HBM spending. Despite TSMC's strong sales growth in 2025, WFE companies may face challenges due to a cautious outlook on capital expenditure, particularly against the backdrop of slowing EUV spending.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004347821422747922, "ret_signed_m1d": -0.004347821422747922, "alpha_spy_m1d": -0.006240407912912804, "alpha_c_m1d": -0.005036237310665825, "bench_spy_m1d": -0.0018925864901648826, "bench_c_m1d": -0.0006884158879179036, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01739139296779124, "ret_signed_p1d": 0.01739139296779124, "alpha_spy_p1d": 0.01753451413101581, "alpha_c_p1d": -0.0003016224402693224, "bench_spy_p1d": 0.0001431211632245688, "bench_c_p1d": -0.017693015408060564, "ret_p1w": -0.004347821422747811, "ret_signed_p1w": 0.004347821422747811, "alpha_spy_p1w": 0.01734150010751334, "alpha_c_p1w": 0.007445850981110946, "bench_spy_p1w": 0.01299367868476553, "bench_c_p1w": 0.0030980295583631356, "ret_p1m": 0.030434749959234786, "ret_signed_p1m": -0.030434749959234786, "alpha_spy_p1m": -0.01287666131789611, "alpha_c_p1m": -0.029195538135889487, "bench_spy_p1m": 0.017558088641338676, "bench_c_p1m": 0.0012392118233452987, "ret_p3m": 0.29646962779241504, "ret_signed_p3m": -0.29646962779241504, "alpha_spy_p3m": -0.2428703605884337, "alpha_c_p3m": -0.11475480942177407, "bench_spy_p3m": 0.05359926720398134, "bench_c_p3m": 0.18171481837064096, "ret_p6m": 0.4978807468181701, "ret_signed_p6m": -0.4978807468181701, "alpha_spy_p6m": -0.38817582996011235, "alpha_c_p6m": -0.14595386200174998, "bench_spy_p6m": 0.10970491685805772, "bench_c_p6m": 0.3519268848164201, "price_path": [-0.2441, -0.2519, -0.2311, -0.2233, -0.219, -0.2138, -0.2138, -0.1774, -0.1878, -0.2034, -0.1965, -0.2051, -0.1783, -0.1713, -0.161, -0.135, -0.1402, -0.1358, -0.148, -0.1532, -0.1428, -0.1497, -0.148, -0.1549, -0.1644, -0.1627, -0.1627, -0.1627, -0.1809, -0.1774, -0.1428, -0.1358, -0.1384, -0.1298, -0.0951, -0.0778, -0.0913, -0.1043, -0.1087, -0.0913, -0.0826, -0.1, -0.0826, -0.113, -0.087, -0.0696, -0.0652, -0.0609, -0.0783, -0.0565, -0.0565, -0.0522, -0.0565, -0.0609, -0.0609, -0.0522, -0.0435, -0.0435, -0.0478, -0.0348, -0.0174, -0.0174, 0.0043, 0.0, -0.0174, -0.0043, -0.0043, -0.0043, -0.0043, -0.013, 0.0043, 0.0087, 0.0087, -0.013, 0.0, -0.0217, 0.0261, 0.0217, 0.0261, 0.0261, 0.0435, 0.0217, 0.0261, 0.0261, 0.0304, -0.013, 0.0, -0.013, 0.0174, 0.0217, 0.0348, 0.0087, 0.0087, 0.013, 0.0087, 0.0087, 0.0087, 0.0261, 0.0261, 0.0435, 0.0652, 0.0783, 0.0957, 0.0913, 0.1175, 0.1044, 0.1219, 0.1044, 0.1306, 0.1699, 0.1699, 0.1524, 0.135, 0.135, 0.1393, 0.1568, 0.1917, 0.2223, 0.2223, 0.2528, 0.2354, 0.2572, 0.2572, 0.2354, 0.2441, 0.279, 0.2965, 0.2659, 0.2921, 0.2921, 0.2746, 0.2659, 0.2659, 0.2921, 0.2877, 0.3139, 0.3139, 0.3096, 0.3183, 0.3139, 0.2746, 0.279, 0.2746, 0.2877, 0.279, 0.2877, 0.2746, 0.2485, 0.2615, 0.2266, 0.2179, 0.2703, 0.2092, 0.2004, 0.2354, 0.2572, 0.2528, 0.2572, 0.231, 0.2485, 0.2659, 0.2615, 0.2746, 0.3052, 0.2921, 0.3139, 0.2877, 0.2964, 0.2701, 0.257, 0.2526, 0.2526, 0.2526, 0.2833, 0.3052, 0.3096, 0.3096, 0.3227, 0.3402, 0.3314, 0.3577, 0.3577, 0.3884, 0.4628, 0.4935, 0.4672, 0.476, 0.4716, 0.4804, 0.4979]}
{"unit_id": "orig:1963211378594119865", "ticker_idx": 0, "ts": "2025-09-03T12:04:02+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley maintains an \"Overweight\" rating and raises its price target from $338 to $357", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Overweight on AVGO with PT raised to $357 on AI ASIC/networking strength and 2026 revenue upgrade.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Broadcom F3Q25 Outlook:\n\nMorgan Stanley (25/09/02)\n\nBroadcom (AVGO) continues to see improving performance outlook driven by strong growth in AI ASIC and networking businesses. Morgan Stanley believes the company’s 2026 AI revenue forecast should be raised to $31.6 billion, up 53% year-over-year, with ASIC revenue expected to grow 61% to $20.8 billion, mainly supported by new customer adoption and rising TPU demand. This scale aligns with the company’s 2027 serviceable available market (SAM) target of $60–90 billion, implying that 2026 revenue must exceed $20 billion to support rapid expansion the following year. Although access to the China market remains uncertain, the largest ASIC project, TPU, is performing steadily, Meta’s MTIA project is progressing as planned, and surging inference demand combined with CoWoS capacity outlook makes Morgan Stanley more optimistic about Broadcom’s ASIC business.\n\nIn AI networking, Ethernet continues to gain market share. Morgan Stanley expects 2026 AI networking revenue to grow 39% year-over-year to $10.8 billion. Arista’s results show that major cloud customers are accelerating Ethernet cluster adoption, further confirming the view that Ethernet, with its open ecosystem and standardization advantages, will dominate the scale-out AI networking market.\n\nOn specific forecasts, Morgan Stanley raised FY26 AI revenue to $28.3 billion and the company’s total semiconductor revenue to $46.3 billion. Forecasts for 2025/2026 revenue and EPS are $62.7 billion / $6.68 and $75.1 billion / $8.07, respectively. For the July 2025 quarter, AI revenue is expected at $5.1 billion, up 16% sequentially, benefiting from double-digit growth in both ASIC and networking; for the October quarter, AI revenue is expected to grow 5% sequentially, with networking potentially providing further upside. The company’s new ASIC project pipeline continues to contribute momentum, with TPU likely to exceed expectations thanks to better pricing and stronger customer adoption.\n\nIn non-AI businesses, Morgan Stanley expects cyclical recovery in semiconductors to gradually materialize, with enterprise networking, broadband, storage, and wireless seeing flat to slight sequential growth; software trends may fluctuate, but overall remain at low single-digit growth, while VMware asset integration remains steady. For the July quarter, total company revenue is expected at $15.876 billion, up 21.5% year-over-year, in line with consensus; gross margin at ~78.1%, EPS at $1.68, slightly above consensus. For the October quarter, revenue is expected at $16.9 billion, up 20.3% year-over-year, with AI revenue at $5.4 billion, gross margin at 77.8%, and EPS at $1.82, again slightly above consensus.\n\nOverall, Morgan Stanley believes Broadcom is well-positioned in the AI infrastructure investment wave, with new ASIC adoption and Ethernet ramping serving as key growth drivers. While long-term uncertainties exist, its ambitious SAM target appears achievable. Morgan Stanley maintains an “Overweight” rating and raises its price target from $338 to $357, based on a CY26 EPS of $7.13 and a 50x multiple, reflecting strong long-term growth prospects and solid execution.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013724127508028428, "ret_signed_m1d": -0.013724127508028428, "alpha_spy_m1d": -0.008333781098861182, "alpha_c_m1d": -0.013898763145839599, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": 0.0001746356378111713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012268921158915536, "ret_signed_p1d": 0.012268921158915536, "alpha_spy_p1d": 0.00391156071710097, "alpha_c_p1d": 0.0002939823348269144, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.011974938824088621, "ret_p1w": 0.2221633296323029, "ret_signed_p1w": 0.2221633296323029, "alpha_spy_p1w": 0.20900577458955505, "alpha_c_p1w": 0.17188932054526473, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.05027400908703816, "ret_p1m": 0.12027300267929708, "ret_signed_p1m": 0.12027300267929708, "alpha_spy_p1m": 0.07780970597196557, "alpha_c_p1m": -0.060608905154126536, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.18088190783342362, "ret_p3m": 0.2789491617668358, "ret_signed_p3m": 0.2789491617668358, "alpha_spy_p3m": 0.21927278526069127, "alpha_c_p3m": 0.04667596151854103, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.2322732002482948, "ret_p6m": 0.06771962151439914, "ret_signed_p6m": 0.06771962151439914, "alpha_spy_p6m": -0.009195747159325585, "alpha_c_p6m": -0.3751879913168956, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.44290761283129476, "price_path": [-0.1853, -0.1941, -0.1929, -0.1656, -0.1552, -0.1795, -0.1683, -0.1773, -0.171, -0.171, -0.1733, -0.1608, -0.1277, -0.1248, -0.1066, -0.1093, -0.0884, -0.1245, -0.1074, -0.09, -0.09, -0.0933, -0.1012, -0.081, -0.0893, -0.0926, -0.0886, -0.0709, -0.0714, -0.0527, -0.063, -0.0469, -0.0787, -0.0618, -0.0452, -0.0404, -0.0268, -0.0164, 0.0008, -0.0287, -0.0455, -0.0154, -0.0313, -0.0024, 0.0045, 0.0085, 0.005, 0.0345, 0.0222, 0.0292, 0.0131, 0.0111, -0.0247, -0.0371, -0.0423, -0.0277, -0.027, -0.0145, -0.0071, 0.0207, -0.0165, -0.0165, -0.0137, 0.0, 0.0123, 0.1075, 0.1431, 0.1134, 0.2222, 0.1893, 0.1901, 0.204, 0.1905, 0.1448, 0.1421, 0.1407, 0.1223, 0.1228, 0.124, 0.1134, 0.1082, 0.0862, 0.0929, 0.1044, 0.1203, 0.1209, 0.1114, 0.1144, 0.1445, 0.1429, 0.0754, 0.1816, 0.14, 0.1638, 0.1732, 0.1572, 0.1569, 0.1351, 0.1273, 0.1405, 0.1731, 0.1993, 0.2355, 0.2786, 0.2471, 0.2245, 0.201, 0.1659, 0.1892, 0.1779, 0.1575, 0.1872, 0.1659, 0.1767, 0.1262, 0.1345, 0.1351, 0.128, 0.1741, 0.1489, 0.127, 0.2521, 0.2755, 0.317, 0.317, 0.3349, 0.2789, 0.264, 0.2608, 0.2622, 0.2927, 0.3287, 0.3459, 0.368, 0.3462, 0.1923, 0.1257, 0.1306, 0.08, 0.0928, 0.1275, 0.1333, 0.1594, 0.1624, 0.1624, 0.1687, 0.1596, 0.1611, 0.1487, 0.1487, 0.1537, 0.1398, 0.141, 0.1401, 0.1035, 0.145, 0.169, 0.1769, 0.1281, 0.1385, 0.1673, 0.1673, 0.1039, 0.0913, 0.0803, 0.0622, 0.0782, 0.1045, 0.106, 0.0977, 0.0996, 0.099, 0.0632, 0.0224, 0.0306, 0.105, 0.1415, 0.1299, 0.1376, 0.0992, 0.0792, 0.0792, 0.1037, 0.1069, 0.1085, 0.1041, 0.0964, 0.0803, 0.1029, 0.0677]}
{"unit_id": "thread:1960159539753550293", "ticker_idx": 0, "ts": "2025-08-26T01:57:06+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "at a minimum, an additional $13 billion in revenue will be added for Nvidia", "tweet_type": "news_reaction", "timeline": "years", "summary": "B40 China demand of 2M/5M units adds $13B+ incremental revenue; bullish on Nvidia.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Instead, Morgan Stanley noted that Chinese customers are showing strong interest in Nvidia’s B40 chip, projecting demand of 2 million units this year and 5 million units next year.\n\n$NVDA\n\n---\n\nThis means that, at a minimum, an additional $13 billion in revenue will be added for Nvidia! (According to a previous Reuters report, the estimated unit price of the B40 was $6,500.)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01078283957708237, "ret_signed_m1d": -0.01078283957708237, "alpha_spy_m1d": -0.006613261402572013, "alpha_c_m1d": -0.0012462736630559634, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0009352252999643706, "ret_signed_p1d": -0.0009352252999643706, "alpha_spy_p1d": -0.003213738581973713, "alpha_c_p1d": -0.004001740074807136, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.06046102686377797, "ret_signed_p1w": -0.06046102686377797, "alpha_spy_p1w": -0.0528815021516118, "alpha_c_p1w": -0.02585281936163819, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": -0.026352128487647786, "ret_signed_p1m": -0.026352128487647786, "alpha_spy_p1m": -0.0539000810695367, "alpha_c_p1m": -0.1084413858754818, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": -0.01584366868256315, "ret_signed_p3m": -0.01584366868256315, "alpha_spy_p3m": -0.040174301739457574, "alpha_c_p3m": -0.1148496148905338, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.034279923956166636, "ret_signed_p6m": 0.034279923956166636, "alpha_spy_p6m": -0.03557288954310378, "alpha_c_p6m": -0.3602733243718008, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.2343, -0.2566, -0.2443, -0.2231, -0.2193, -0.2299, -0.2204, -0.2154, -0.2081, -0.2142, -0.2023, -0.219, -0.204, -0.2071, -0.1996, -0.1996, -0.2086, -0.2069, -0.1863, -0.1511, -0.1472, -0.1321, -0.1308, -0.1566, -0.1349, -0.1234, -0.1234, -0.1294, -0.1198, -0.1039, -0.0972, -0.0927, -0.0974, -0.0609, -0.0572, -0.0482, -0.0515, -0.0572, -0.0811, -0.0605, -0.0442, -0.0455, -0.0276, -0.0344, -0.0138, -0.0215, -0.0443, -0.0097, -0.0193, -0.0129, -0.0055, 0.0051, 0.0016, 0.0076, -0.001, 0.0014, -0.0073, 0.0013, -0.0337, -0.035, -0.0374, -0.0208, -0.0108, 0.0, -0.0009, -0.0088, -0.0418, -0.0418, -0.0605, -0.0613, -0.0556, -0.0811, -0.074, -0.0606, -0.0244, -0.0253, -0.0217, -0.0221, -0.0379, -0.0631, -0.0304, -0.028, 0.0102, -0.0183, -0.0264, -0.0224, -0.0196, 0.0005, 0.0265, 0.0302, 0.0392, 0.0322, 0.0208, 0.018, 0.0404, 0.0595, 0.0077, 0.0361, -0.0095, -0.0106, 0.0003, 0.008, 0.0048, -0.0033, -0.0081, 0.0022, 0.0248, 0.0535, 0.106, 0.1391, 0.1163, 0.1141, 0.1382, 0.0931, 0.074, 0.0348, 0.0352, 0.0951, 0.0627, 0.0662, 0.0281, 0.0463, 0.0266, -0.0022, 0.0262, -0.0062, -0.0158, 0.0043, -0.0217, -0.0083, -0.0083, -0.0262, -0.0101, -0.0016, -0.0119, 0.009, 0.0036, 0.0209, 0.0177, 0.0112, -0.0045, -0.037, -0.03, -0.0222, -0.0595, -0.0419, -0.0042, 0.0107, 0.041, 0.0377, 0.0377, 0.0483, 0.0356, 0.0319, 0.0261, 0.0261, 0.0391, 0.0351, 0.0302, 0.0405, 0.0181, 0.0171, 0.0176, 0.0223, 0.0076, 0.0292, 0.0247, 0.0247, -0.0202, 0.0086, 0.017, 0.0326, 0.026, 0.0373, 0.0538, 0.0592, 0.0516, 0.0212, -0.0078, -0.0416, -0.0543, 0.0201, 0.0456, 0.0374, 0.0457, 0.0286, 0.0058, 0.0058, 0.0177, 0.0343]}
{"unit_id": "orig:1958748289639317807", "ticker_idx": 0, "ts": "2025-08-22T04:29:18+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "risk factors have emerged for Micron's HBM4 CS and mass production schedule... if Micron's HBM4 supply to Nvidia becomes difficult or is delayed compared to the original schedule, this could lead to an expansion of SK Hynix's share as Nvidia's primary supplier", "tweet_type": "analysis", "timeline": "quarters", "summary": "Micron faces HBM4 risk from NVIDIA's raised 10Gbps speed requirement; DRAM-process base die disadvantaged vs SK Hynix/Samsung foundry approach, threatening Micron's schedule and share.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron, Risk Factors in HBM4\n\nWe estimate that risk factors have emerged for Micron’s HBM4 CS and mass production schedule.\n\nThe reason is that Nvidia has recently raised the speed margin requirements to 10Gbps, higher than before, and presented them to all three DRAM makers. Unlike SK Hynix and Samsung Electronics, which will produce the base die of HBM4 through foundry processes, Micron continues to produce the base die through DRAM processes. Because of this, it is assumed that achieving 10Gbps speed may be difficult. DRAM’s 1b nano corresponds to the 12nm class, the same linewidth as TSMC’s 12nm foundry that SK Hynix uses for HBM4 base die production. However, DRAM transistors have a planar structure, and compared to the 3D FinFET of foundries, they are inferior in performance and power ratio, imposing restrictions on high-speed logic implementation. For this reason, SK Hynix decided to use TSMC’s 12nm foundry to produce base dies in preparation for the heightened performance requirements of HBM4. Samsung Electronics will also produce base dies for HBM4 using Samsung Foundry’s 4nm process.\n\nIn HBM3e, Nvidia’s speed margin was 8.5Gbps, but Micron reportedly already achieved 9.2Gbps with 1b nano. Nvidia’s HBM4 speed requirement was initially 9.0Gbps. Since Micron had already achieved 9.2Gbps in HBM3e with 1b nano, it would have judged that it could meet speed requirements in HBM4 as well. However, it has been identified that Nvidia recently demanded all three DRAM makers raise it further to 10Gbps. The reason lies in HBM4’s bandwidth (the amount of data transferable per unit time). One of the biggest bottlenecks in AI systems is still the long time GPUs spend waiting for data from HBM. Therefore, increasing HBM’s bandwidth to reduce GPU idle time is very important for improving AI system performance and achieving cost efficiency. Although increasing pin speed raises power consumption, while HBM4 was initially conceived with lower pin speeds, Nvidia ultimately seems to have judged that increasing bandwidth is more important than power efficiency for AI system improvement.\n\nMicron DRAM is known for its good low-power characteristics, but as described above, since it produces HBM4 base dies with DRAM processes just like in HBM3e, it may be at a disadvantage in terms of speed. For these reasons, if Micron’s HBM4 supply to Nvidia becomes difficult or is delayed compared to the original schedule, this could lead to an expansion of SK Hynix’s share as Nvidia’s primary supplier.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01606055756302427, "ret_signed_m1d": 0.01606055756302427, "alpha_spy_m1d": 0.000935938077442966, "alpha_c_m1d": -0.0049857168432636545, "bench_spy_m1d": -0.015124619485581303, "bench_c_m1d": -0.021046274406287924, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010707038375349476, "ret_signed_p1d": 0.010707038375349476, "alpha_spy_p1d": 0.006305881317731021, "alpha_c_p1d": 0.011660630735101152, "bench_spy_p1d": -0.0044011570576184544, "bench_c_p1d": 0.000953592359751676, "ret_p1w": 0.011301823342826589, "ret_signed_p1w": -0.011301823342826589, "alpha_spy_p1w": -0.01170482476387158, "alpha_c_p1w": -0.02271032002068263, "bench_spy_p1w": -0.0004030014210449906, "bench_c_p1w": -0.01140849667785604, "ret_p1m": 0.3988782774933848, "ret_signed_p1m": -0.3988782774933848, "alpha_spy_p1m": -0.36264964205333494, "alpha_c_p1m": -0.3017185065181842, "bench_spy_p1m": 0.03622863544004984, "bench_c_p1m": 0.09715977097520057, "ret_p3m": 0.9209845603275635, "ret_signed_p3m": -0.9209845603275635, "alpha_spy_p3m": -0.8912979018016844, "alpha_c_p3m": -0.765011274257883, "bench_spy_p3m": 0.02968665852587904, "bench_c_p3m": 0.15597328606968053, "ret_p6m": 2.5017352385583753, "ret_signed_p6m": -2.5017352385583753, "alpha_spy_p6m": -2.4392069688433073, "alpha_c_p6m": -2.1089118993879765, "bench_spy_p6m": 0.06252826971506797, "bench_c_p6m": 0.3928233391703986, "price_path": [-0.1818, -0.1835, -0.1782, -0.1981, -0.1665, -0.1319, -0.1234, -0.0976, -0.0784, -0.0581, -0.031, -0.0149, -0.0137, -0.0186, 0.0174, 0.0216, 0.0342, 0.0342, 0.0493, 0.0364, 0.0859, 0.0803, 0.0697, 0.0592, 0.0463, 0.0263, 0.0335, 0.0382, 0.0382, 0.019, 0.0573, 0.0387, 0.0461, 0.0582, 0.0079, 0.0206, -0.0106, -0.0376, -0.028, -0.0378, -0.0719, -0.0667, -0.0506, -0.0546, -0.0546, -0.0486, -0.025, -0.0726, -0.1088, -0.0842, -0.0732, -0.0756, -0.0494, 0.0103, 0.0513, 0.0856, 0.056, 0.0647, 0.0271, 0.0499, 0.0371, -0.004, -0.0161, 0.0, -0.0107, -0.01, 0.0006, 0.0367, 0.0113, 0.0113, 0.0068, 0.0088, 0.0555, 0.1163, 0.1171, 0.1492, 0.1897, 0.2795, 0.3361, 0.3407, 0.3496, 0.3595, 0.4352, 0.3828, 0.3989, 0.4141, 0.3742, 0.3327, 0.3364, 0.3928, 0.4218, 0.5478, 0.5614, 0.5971, 0.6237, 0.5789, 0.6712, 0.6354, 0.5441, 0.6391, 0.5906, 0.6321, 0.7221, 0.7208, 0.7582, 0.7201, 0.6876, 0.7576, 0.8623, 0.8715, 0.8869, 0.927, 0.9047, 0.9027, 0.9956, 0.8539, 1.0194, 1.0265, 1.023, 1.1538, 1.0501, 1.0824, 1.0148, 1.0988, 1.0573, 0.9429, 0.921, 0.7122, 0.7633, 0.9041, 0.9092, 0.9579, 0.9579, 1.0108, 1.0446, 1.0364, 0.991, 0.9272, 1.0171, 1.0995, 1.1463, 1.2423, 1.1977, 1.0504, 1.0194, 0.977, 0.9176, 1.1134, 1.2611, 1.3518, 1.3491, 1.4376, 1.4376, 1.4216, 1.504, 1.4892, 1.4278, 1.4278, 1.6831, 1.6553, 1.9213, 1.8883, 1.7818, 1.9355, 1.9421, 1.8763, 1.8356, 1.8635, 2.0857, 2.0857, 2.1048, 2.3099, 2.382, 2.3996, 2.3097, 2.4897, 2.7027, 2.707, 2.5291, 2.7241, 2.5679, 2.2273, 2.257, 2.3574, 2.2622, 2.175, 2.4905, 2.5214, 2.5017, 2.5017]}
{"unit_id": "orig:1958748289639317807", "ticker_idx": 1, "ts": "2025-08-22T04:29:18+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "this could lead to an expansion of SK Hynix's share as Nvidia's primary supplier", "tweet_type": "analysis", "timeline": "quarters", "summary": "Micron faces HBM4 risk from NVIDIA's raised 10Gbps speed requirement; DRAM-process base die disadvantaged vs SK Hynix/Samsung foundry approach, threatening Micron's schedule and share.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron, Risk Factors in HBM4\n\nWe estimate that risk factors have emerged for Micron’s HBM4 CS and mass production schedule.\n\nThe reason is that Nvidia has recently raised the speed margin requirements to 10Gbps, higher than before, and presented them to all three DRAM makers. Unlike SK Hynix and Samsung Electronics, which will produce the base die of HBM4 through foundry processes, Micron continues to produce the base die through DRAM processes. Because of this, it is assumed that achieving 10Gbps speed may be difficult. DRAM’s 1b nano corresponds to the 12nm class, the same linewidth as TSMC’s 12nm foundry that SK Hynix uses for HBM4 base die production. However, DRAM transistors have a planar structure, and compared to the 3D FinFET of foundries, they are inferior in performance and power ratio, imposing restrictions on high-speed logic implementation. For this reason, SK Hynix decided to use TSMC’s 12nm foundry to produce base dies in preparation for the heightened performance requirements of HBM4. Samsung Electronics will also produce base dies for HBM4 using Samsung Foundry’s 4nm process.\n\nIn HBM3e, Nvidia’s speed margin was 8.5Gbps, but Micron reportedly already achieved 9.2Gbps with 1b nano. Nvidia’s HBM4 speed requirement was initially 9.0Gbps. Since Micron had already achieved 9.2Gbps in HBM3e with 1b nano, it would have judged that it could meet speed requirements in HBM4 as well. However, it has been identified that Nvidia recently demanded all three DRAM makers raise it further to 10Gbps. The reason lies in HBM4’s bandwidth (the amount of data transferable per unit time). One of the biggest bottlenecks in AI systems is still the long time GPUs spend waiting for data from HBM. Therefore, increasing HBM’s bandwidth to reduce GPU idle time is very important for improving AI system performance and achieving cost efficiency. Although increasing pin speed raises power consumption, while HBM4 was initially conceived with lower pin speeds, Nvidia ultimately seems to have judged that increasing bandwidth is more important than power efficiency for AI system improvement.\n\nMicron DRAM is known for its good low-power characteristics, but as described above, since it produces HBM4 base dies with DRAM processes just like in HBM3e, it may be at a disadvantage in terms of speed. For these reasons, if Micron’s HBM4 supply to Nvidia becomes difficult or is delayed compared to the original schedule, this could lead to an expansion of SK Hynix’s share as Nvidia’s primary supplier.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023904440996469822, "ret_signed_m1d": -0.023904440996469822, "alpha_spy_m1d": -0.00877982151088852, "alpha_c_m1d": -0.0028581665901818987, "bench_spy_m1d": -0.015124619485581303, "bench_c_m1d": -0.021046274406287924, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03386446325793102, "ret_signed_p1d": 0.03386446325793102, "alpha_spy_p1d": 0.038265620315549476, "alpha_c_p1d": 0.032910870898179345, "bench_spy_p1d": -0.0044011570576184544, "bench_c_p1d": 0.000953592359751676, "ret_p1w": 0.07326097337444981, "ret_signed_p1w": 0.07326097337444981, "alpha_spy_p1w": 0.0736639747954948, "alpha_c_p1w": 0.08466947005230585, "bench_spy_p1w": -0.0004030014210449906, "bench_c_p1w": -0.01140849667785604, "ret_p1m": 0.40042618833488675, "ret_signed_p1m": 0.40042618833488675, "alpha_spy_p1m": 0.3641975528948369, "alpha_c_p1m": 0.3032664173596862, "bench_spy_p1m": 0.03622863544004984, "bench_c_p1m": 0.09715977097520057, "ret_p3m": 1.242277735294524, "ret_signed_p3m": 1.242277735294524, "alpha_spy_p3m": 1.212591076768645, "alpha_c_p3m": 1.0863044492248435, "bench_spy_p3m": 0.02968665852587904, "bench_c_p3m": 0.15597328606968053, "ret_p6m": 2.513554443093718, "ret_signed_p6m": 2.513554443093718, "alpha_spy_p6m": 2.45102617337865, "alpha_c_p6m": 2.120731103923319, "bench_spy_p6m": 0.06252826971506797, "bench_c_p6m": 0.3928233391703986, "price_path": [-0.1947, -0.1728, -0.1554, -0.1853, -0.1733, -0.1733, -0.1335, -0.1056, -0.1056, -0.0876, -0.0817, -0.0438, -0.0618, -0.0618, -0.012, -0.008, -0.0179, -0.0199, 0.0239, 0.0339, 0.1096, 0.1394, 0.1673, 0.1315, 0.1633, 0.1375, 0.1116, 0.1096, 0.0777, 0.0797, 0.1235, 0.1195, 0.1833, 0.1733, 0.1952, 0.1892, 0.1793, 0.0737, 0.0717, 0.0857, 0.0697, 0.0717, 0.0737, 0.0598, 0.0438, 0.0458, 0.0498, 0.0896, 0.0279, 0.0279, 0.0498, 0.0299, 0.0438, 0.0219, 0.0637, 0.0717, 0.1076, 0.1016, 0.1016, 0.0657, 0.0478, 0.0179, -0.0239, 0.0, 0.0339, 0.0418, 0.0359, 0.0713, 0.0733, 0.0214, 0.0393, 0.0473, 0.0593, 0.0912, 0.1052, 0.1491, 0.2129, 0.2249, 0.3107, 0.3206, 0.3885, 0.3306, 0.4184, 0.4184, 0.4004, 0.4403, 0.4264, 0.4224, 0.3426, 0.3924, 0.3865, 0.4363, 0.578, 0.578, 0.578, 0.578, 0.578, 0.578, 0.7076, 0.6558, 0.6418, 0.6857, 0.8054, 0.8573, 0.9371, 0.9111, 0.9211, 0.9091, 1.0348, 1.1346, 1.0787, 1.2263, 1.2662, 1.2303, 1.4737, 1.338, 1.3101, 1.366, 1.3141, 1.4178, 1.4697, 1.4617, 1.4418, 1.2343, 1.4178, 1.2742, 1.2423, 1.2782, 1.0787, 1.0747, 1.0707, 1.0907, 1.172, 1.1161, 1.1481, 1.2279, 1.204, 1.164, 1.172, 1.3038, 1.2599, 1.3437, 1.2559, 1.2798, 1.2119, 1.1161, 1.2, 1.204, 1.184, 1.3158, 1.3317, 1.3477, 1.3477, 1.3916, 1.5553, 1.5992, 1.5992, 1.5992, 1.703, 1.7789, 1.8987, 1.9626, 2.0185, 1.9706, 1.9905, 1.9466, 1.9626, 1.9905, 2.0185, 2.0504, 1.9666, 1.9546, 2.0145, 2.0624, 1.9386, 2.1941, 2.3578, 2.4377, 2.6293, 2.3139, 2.6214, 2.5934, 2.3618, 2.3499, 2.5415, 2.4976, 2.4337, 2.5455, 2.5136, 2.5136]}
{"unit_id": "orig:1968174930836824214", "ticker_idx": 0, "ts": "2025-09-17T04:47:25+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we revise Broadcom's CoWoS booking in 2026 from the original 145k at TSMC to 205k", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley note: AVGO CoWoS allocation revised sharply up to 205k; MediaTek PT cut to NT$1800 on delayed TPU v8 tape-out.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s comments on ASIC CoWoS allocation:\n\nUpdates on our 2026 CoWoS allocation estimates: We keep our 2026-end CoWoS capacity as 93kpwm at TSMC, but fine-tune the allocation based on recent industry dynamics. Numbers are based on our Asia supply chain checks:\n\n- OpenAI: a small CoWoS booking (around 10k) for OpenAI in 2026.\n\n- Google: The 3nm TPU v7 from Broadcom shows volume upside in 2026, and we revise Broadcom’s CoWoS booking in 2026 from the original 145k at TSMC to 205k. This implies TPU volume may reach 3mn from Broadcom.\n\nThe 3nm TPU v8 from MediaTek should tape-out slightly later in October. Given tape-out delay, we now cut our estimate to around 200k-300k (previously 300k-400k) of TPU v8. We revise down MediaTek's target price to NT$1800, accordingly.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0399514585660008, "ret_signed_m1d": 0.0399514585660008, "alpha_spy_m1d": 0.03870747770773053, "alpha_c_m1d": 0.03353049782877271, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0023687311878447526, "ret_signed_p1d": -0.0023687311878447526, "alpha_spy_p1d": -0.007041176330175092, "alpha_c_p1d": -0.04076352108956882, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": -0.018137426839401893, "ret_signed_p1w": -0.018137426839401893, "alpha_spy_p1w": -0.023830586195945802, "alpha_c_p1w": -0.07009479894493764, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.024805091264776147, "ret_signed_p1m": 0.024805091264776147, "alpha_spy_p1m": 0.019811612779799592, "alpha_c_p1m": -0.09984675067725712, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": -0.016690617604364744, "ret_signed_p3m": -0.016690617604364744, "alpha_spy_p3m": -0.05224576063695141, "alpha_c_p3m": -0.17279219005077529, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": -0.02594219927061947, "ret_signed_p6m": -0.02594219927061947, "alpha_spy_p6m": -0.042174739464218014, "alpha_c_p6m": -0.3014138103424795, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.2778, -0.2669, -0.238, -0.2355, -0.2195, -0.2219, -0.2037, -0.2352, -0.2203, -0.2051, -0.2051, -0.208, -0.2148, -0.1972, -0.2044, -0.2074, -0.2039, -0.1884, -0.1888, -0.1725, -0.1815, -0.1674, -0.1952, -0.1805, -0.166, -0.1617, -0.1498, -0.1408, -0.1258, -0.1516, -0.1662, -0.14, -0.1538, -0.1285, -0.1225, -0.119, -0.1221, -0.0963, -0.1071, -0.1009, -0.1151, -0.1167, -0.1481, -0.1589, -0.1634, -0.1507, -0.15, -0.1391, -0.1327, -0.1084, -0.1409, -0.1409, -0.1385, -0.1265, -0.1158, -0.0326, -0.0015, -0.0274, 0.0676, 0.0389, 0.0396, 0.0518, 0.04, 0.0, -0.0024, -0.0036, -0.0196, -0.0192, -0.0181, -0.0274, -0.032, -0.0512, -0.0453, -0.0353, -0.0214, -0.0209, -0.0292, -0.0265, -0.0002, -0.0016, -0.0606, 0.0322, -0.0042, 0.0166, 0.0248, 0.0109, 0.0106, -0.0084, -0.0153, -0.0037, 0.0247, 0.0477, 0.0793, 0.1169, 0.0894, 0.0696, 0.0491, 0.0184, 0.0388, 0.029, 0.0111, 0.0371, 0.0185, 0.0279, -0.0162, -0.009, -0.0085, -0.0147, 0.0256, 0.0036, -0.0156, 0.0937, 0.1142, 0.1504, 0.1504, 0.166, 0.1172, 0.1042, 0.1014, 0.1026, 0.1292, 0.1607, 0.1757, 0.195, 0.1759, 0.0415, -0.0167, -0.0124, -0.0566, -0.0454, -0.0151, -0.0101, 0.0128, 0.0154, 0.0154, 0.0209, 0.013, 0.0143, 0.0034, 0.0034, 0.0078, -0.0043, -0.0033, -0.0041, -0.0361, 0.0002, 0.0211, 0.0281, -0.0146, -0.0055, 0.0197, 0.0197, -0.0357, -0.0467, -0.0563, -0.0721, -0.0582, -0.0352, -0.0339, -0.0411, -0.0395, -0.04, -0.0713, -0.1069, -0.0998, -0.0348, -0.0028, -0.013, -0.0063, -0.0399, -0.0573, -0.0573, -0.0359, -0.0331, -0.0317, -0.0356, -0.0423, -0.0563, -0.0366, -0.0673, -0.0735, -0.0757, -0.0901, -0.0794, -0.0352, -0.0419, 0.0024, -0.0068, -0.0097, -0.0259]}
{"unit_id": "orig:1968174930836824214", "ticker_idx": 1, "ts": "2025-09-17T04:47:25+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we revise down MediaTek's target price to NT$1800, accordingly", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley note: AVGO CoWoS allocation revised sharply up to 205k; MediaTek PT cut to NT$1800 on delayed TPU v8 tape-out.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s comments on ASIC CoWoS allocation:\n\nUpdates on our 2026 CoWoS allocation estimates: We keep our 2026-end CoWoS capacity as 93kpwm at TSMC, but fine-tune the allocation based on recent industry dynamics. Numbers are based on our Asia supply chain checks:\n\n- OpenAI: a small CoWoS booking (around 10k) for OpenAI in 2026.\n\n- Google: The 3nm TPU v7 from Broadcom shows volume upside in 2026, and we revise Broadcom’s CoWoS booking in 2026 from the original 145k at TSMC to 205k. This implies TPU volume may reach 3mn from Broadcom.\n\nThe 3nm TPU v8 from MediaTek should tape-out slightly later in October. Given tape-out delay, we now cut our estimate to around 200k-300k (previously 300k-400k) of TPU v8. We revise down MediaTek's target price to NT$1800, accordingly.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016556254552547722, "ret_signed_m1d": -0.016556254552547722, "alpha_spy_m1d": -0.015312273694277456, "alpha_c_m1d": -0.010135293815319635, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.004672445142330339, "alpha_c_p1d": 0.03839478990172407, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": -0.11920528000346775, "ret_signed_p1w": 0.11920528000346775, "alpha_spy_p1w": 0.12489843936001166, "alpha_c_p1w": 0.1711626521090035, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": -0.11920528000346775, "ret_signed_p1m": 0.11920528000346775, "alpha_spy_p1m": 0.1241987584884443, "alpha_c_p1m": 0.24385712194550102, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": -0.05960268120592127, "ret_signed_p3m": 0.05960268120592127, "alpha_spy_p3m": 0.09515782423850794, "alpha_c_p3m": 0.2157042536523318, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.2050345913273921, "ret_signed_p6m": -0.2050345913273921, "alpha_spy_p6m": -0.18880205113379356, "alpha_c_p6m": 0.07043701974446792, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.1882, -0.1849, -0.1752, -0.1589, -0.1622, -0.1654, -0.1882, -0.1719, -0.1589, -0.1589, -0.1457, -0.1523, -0.1589, -0.1026, -0.0728, -0.0596, -0.0828, -0.0762, -0.0662, -0.0795, -0.0662, -0.0563, -0.053, -0.053, -0.053, -0.053, -0.0728, -0.0993, -0.0861, -0.0927, -0.0927, -0.1192, -0.1126, -0.1258, -0.1026, -0.106, -0.1026, -0.0993, -0.0927, -0.0795, -0.0927, -0.0662, -0.0795, -0.1026, -0.096, -0.096, -0.0662, -0.0728, -0.0728, -0.0828, -0.0927, -0.0993, -0.0993, -0.0762, -0.0828, -0.0497, -0.053, 0.0, -0.0132, -0.0199, -0.0166, -0.0166, 0.0166, 0.0, 0.0, -0.0464, -0.0695, -0.0894, -0.1192, -0.1093, -0.1325, -0.1325, -0.1291, -0.149, -0.1523, -0.1258, -0.1258, -0.1225, -0.1159, -0.1093, -0.1093, -0.1291, -0.1358, -0.1291, -0.1192, -0.1192, -0.1126, -0.1126, -0.1192, -0.1424, -0.1424, -0.1159, -0.1291, -0.1391, -0.1325, -0.1325, -0.1391, -0.1291, -0.1192, -0.1457, -0.1656, -0.1722, -0.1788, -0.1755, -0.1755, -0.1854, -0.1854, -0.2252, -0.2318, -0.2152, -0.2417, -0.2384, -0.2152, -0.1391, -0.1126, -0.0762, -0.043, -0.0629, -0.0728, -0.0695, -0.0563, -0.0464, -0.0596, -0.0331, -0.0762, -0.0695, -0.0596, -0.0596, -0.0563, -0.0596, -0.0662, -0.0728, -0.0795, -0.0861, -0.0861, -0.0828, -0.0596, -0.0596, -0.053, -0.053, -0.0265, 0.0099, 0.0025, 0.0059, -0.0245, -0.0414, -0.0245, 0.0025, 0.0126, 0.0059, 0.016, 0.0025, 0.0025, -0.011, 0.0025, 0.1004, 0.1949, 0.1983, 0.2017, 0.2017, 0.1882, 0.151, 0.2118, 0.2152, 0.1949, 0.1544, 0.2354, 0.2455, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2523, 0.2118, 0.232, 0.2624, 0.313, 0.313, 0.2827, 0.2253, 0.1612, 0.1983, 0.1915, 0.124, 0.151, 0.1915, 0.205]}
{"unit_id": "orig:1970981381628133811", "ticker_idx": 0, "ts": "2025-09-24T22:39:15+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix leads HBM4 qualification with early customer sample shipments… SK hynix: raised to ₩460,000", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM 4-year memory supercycle thesis: SK Hynix PT ₩460K, Samsung PT ₩100K; HBM4 TAM $90B by 2027, DRAM prices rising 4 consecutive years.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra GPUs will drive 4× memory content growth compared to the previous generation.\n\n2. HBM Outlook\n    • In 2026, HBM3E is expected to face oversupply and downward price pressure, but HBM4 supply constraints could support a 35% price premium.\n    • HBM4 TAM projected to accelerate to $90 billion by 2027.\n    • SK hynix leads HBM4 qualification with early customer sample shipments; Samsung and Micron are expected to compete with less than 40% market share.\n    • Rubin Ultra’s surge in HBM demand → simultaneous growth in pricing and TAM from 4× content increase.\n\n3. DRAM\n    • For the first time ever, DRAM prices are forecast to rise for four consecutive years (2024–2027).\n    • By 2027, HBM will account for 43% of DRAM TAM → reduces price volatility and provides downside margin protection.\n    • Broader demand diversification from new applications such as SOCAMM2 and GDDR7.\n    • DRAM TAM expected to grow at a 27% CAGR through 2027.\n\n4. NAND\n    • Two years of underinvestment and HDD shortages → expanding adoption of eSSD.\n    • NAND ASP projected to rise +7% by 2026, though it is still too early to classify NAND as a structural AI storage essential.\n\n5. Valuation\n    • Stock prices have surpassed prior peaks within just 1–2 months.\n    • While current valuations are at past cycle highs, structurally improved margins mean simple historical comparisons are inappropriate.\n    • Top picks: SK hynix > Samsung Electronics > Nanya.\n\n6. CapEx and Technology Trends\n    • Memory CapEx expected to rise 7–12% in 2026–2027 to meet demand surge.\n    • Investments concentrated in DRAM, prioritizing HBM; NAND focus remains on utilization improvements and gradual transition investments.\n    • From HBM5 generation (post-2027), hybrid bonding adoption expected to accelerate → key equipment vendors to benefit.\n\n7. Target Price Revisions\n    • SK hynix: raised to ₩460,000\n    • Samsung Electronics: raised to ₩100,000", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009790207583604715, "ret_signed_m1d": 0.009790207583604715, "alpha_spy_m1d": 0.006598419029754332, "alpha_c_m1d": 0.007952749853283425, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0027972146883529225, "ret_signed_p1d": -0.0027972146883529225, "alpha_spy_p1d": 0.00181628280011914, "alpha_c_p1d": -0.002641535095449843, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.006993080546512598, "ret_signed_p1w": 0.006993080546512598, "alpha_spy_p1w": -0.004124883876781116, "alpha_c_p1w": -0.03218355077399071, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.33846166252178467, "ret_signed_p1m": 0.33846166252178467, "alpha_spy_p1m": 0.3223369836992218, "alpha_c_p1m": 0.26468631418035393, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 0.6235395875604981, "ret_signed_p3m": 0.6235395875604981, "alpha_spy_p3m": 0.5845838156887011, "alpha_c_p3m": 0.4998770807072117, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 1.8408282027398206, "ret_signed_p6m": 1.8408282027398206, "alpha_spy_p6m": 1.8398454854354818, "alpha_c_p6m": 1.6071423611076274, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.2067, -0.1844, -0.2026, -0.2207, -0.2221, -0.2444, -0.2431, -0.2123, -0.2151, -0.1704, -0.1774, -0.162, -0.1662, -0.1732, -0.2472, -0.2486, -0.2389, -0.25, -0.2486, -0.2472, -0.257, -0.2682, -0.2668, -0.264, -0.2361, -0.2794, -0.2794, -0.264, -0.278, -0.2682, -0.2836, -0.2542, -0.2486, -0.2235, -0.2277, -0.2277, -0.2528, -0.2654, -0.2863, -0.3157, -0.2989, -0.2752, -0.2696, -0.2738, -0.249, -0.2476, -0.2839, -0.2713, -0.2657, -0.2573, -0.235, -0.2252, -0.1944, -0.1497, -0.1413, -0.0811, -0.0741, -0.0266, -0.0671, -0.0056, -0.0056, -0.0182, 0.0098, 0.0, -0.0028, -0.0587, -0.0238, -0.028, 0.007, 0.1063, 0.1063, 0.1063, 0.1063, 0.1063, 0.1063, 0.1972, 0.1608, 0.151, 0.1818, 0.2657, 0.3021, 0.358, 0.3399, 0.3469, 0.3385, 0.4266, 0.4965, 0.4573, 0.5608, 0.5888, 0.5636, 0.7343, 0.6392, 0.6196, 0.6587, 0.6224, 0.6951, 0.7315, 0.7259, 0.7119, 0.5664, 0.6951, 0.5944, 0.572, 0.5972, 0.4573, 0.4545, 0.4517, 0.4657, 0.5228, 0.4836, 0.506, 0.562, 0.5452, 0.5172, 0.5228, 0.6151, 0.5844, 0.6431, 0.5816, 0.5983, 0.5508, 0.4836, 0.5424, 0.5452, 0.5312, 0.6235, 0.6347, 0.6459, 0.6459, 0.6767, 0.7915, 0.8223, 0.8223, 0.8223, 0.8951, 0.9482, 1.0322, 1.077, 1.1162, 1.0826, 1.0966, 1.0658, 1.077, 1.0966, 1.1162, 1.1386, 1.0798, 1.0714, 1.1134, 1.147, 1.0602, 1.2394, 1.3541, 1.4101, 1.5445, 1.3233, 1.5389, 1.5193, 1.3569, 1.3485, 1.4829, 1.4521, 1.4073, 1.4857, 1.4633, 1.4633, 1.4633, 1.4633, 1.5025, 1.6564, 1.662, 1.8132, 1.8496, 2.082, 1.9754, 1.9754, 1.6333, 1.3809, 1.6389, 1.5912, 1.3445, 1.6305, 1.6782, 1.6081, 1.552, 1.7315, 1.7202, 1.9614, 1.8408]}
{"unit_id": "orig:1970981381628133811", "ticker_idx": 1, "ts": "2025-09-24T22:39:15+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics: raised to ₩100,000", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM 4-year memory supercycle thesis: SK Hynix PT ₩460K, Samsung PT ₩100K; HBM4 TAM $90B by 2027, DRAM prices rising 4 consecutive years.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra GPUs will drive 4× memory content growth compared to the previous generation.\n\n2. HBM Outlook\n    • In 2026, HBM3E is expected to face oversupply and downward price pressure, but HBM4 supply constraints could support a 35% price premium.\n    • HBM4 TAM projected to accelerate to $90 billion by 2027.\n    • SK hynix leads HBM4 qualification with early customer sample shipments; Samsung and Micron are expected to compete with less than 40% market share.\n    • Rubin Ultra’s surge in HBM demand → simultaneous growth in pricing and TAM from 4× content increase.\n\n3. DRAM\n    • For the first time ever, DRAM prices are forecast to rise for four consecutive years (2024–2027).\n    • By 2027, HBM will account for 43% of DRAM TAM → reduces price volatility and provides downside margin protection.\n    • Broader demand diversification from new applications such as SOCAMM2 and GDDR7.\n    • DRAM TAM expected to grow at a 27% CAGR through 2027.\n\n4. NAND\n    • Two years of underinvestment and HDD shortages → expanding adoption of eSSD.\n    • NAND ASP projected to rise +7% by 2026, though it is still too early to classify NAND as a structural AI storage essential.\n\n5. Valuation\n    • Stock prices have surpassed prior peaks within just 1–2 months.\n    • While current valuations are at past cycle highs, structurally improved margins mean simple historical comparisons are inappropriate.\n    • Top picks: SK hynix > Samsung Electronics > Nanya.\n\n6. CapEx and Technology Trends\n    • Memory CapEx expected to rise 7–12% in 2026–2027 to meet demand surge.\n    • Investments concentrated in DRAM, prioritizing HBM; NAND focus remains on utilization improvements and gradual transition investments.\n    • From HBM5 generation (post-2027), hybrid bonding adoption expected to accelerate → key equipment vendors to benefit.\n\n7. Target Price Revisions\n    • SK hynix: raised to ₩460,000\n    • Samsung Electronics: raised to ₩100,000", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008196709176757522, "ret_signed_m1d": -0.008196709176757522, "alpha_spy_m1d": -0.011388497730607905, "alpha_c_m1d": -0.014408726747365486, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.006212017570607964, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008196709176757633, "ret_signed_p1d": 0.008196709176757633, "alpha_spy_p1d": 0.012810206665229695, "alpha_c_p1d": 0.009956187442882514, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.001759478266124881, "ret_p1w": 0.011518714093654125, "ret_signed_p1w": 0.011518714093654125, "alpha_spy_p1w": 0.00040074967036041187, "alpha_c_p1w": -0.010492785369283109, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.022011499462937234, "ret_p1m": 0.13501813070950797, "ret_signed_p1m": 0.13501813070950797, "alpha_spy_p1m": 0.11889345188694511, "alpha_c_p1m": 0.09695573252480783, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.03806239818470014, "ret_p3m": 0.2996839270034226, "ret_signed_p3m": 0.2996839270034226, "alpha_spy_p3m": 0.2607281551316256, "alpha_c_p3m": 0.2556236770700404, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.0440602499333822, "ret_p6m": 1.369713804043995, "ret_signed_p6m": 1.369713804043995, "alpha_spy_p6m": 1.368731086739656, "alpha_c_p6m": 1.3740589857250702, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": -0.0043451816810753385, "price_path": [-0.2881, -0.2998, -0.2951, -0.2881, -0.2529, -0.2588, -0.2775, -0.281, -0.2927, -0.2857, -0.267, -0.2681, -0.2541, -0.2424, -0.219, -0.2143, -0.2061, -0.2272, -0.2225, -0.2272, -0.2283, -0.1756, -0.1733, -0.1499, -0.1639, -0.1932, -0.1838, -0.1815, -0.1944, -0.1745, -0.1593, -0.1686, -0.1674, -0.1581, -0.1616, -0.1616, -0.1803, -0.1803, -0.1745, -0.1733, -0.1639, -0.1628, -0.1768, -0.1733, -0.185, -0.1838, -0.2084, -0.1909, -0.1827, -0.1792, -0.1862, -0.1792, -0.1628, -0.1499, -0.1405, -0.1171, -0.1042, -0.0703, -0.0843, -0.0597, -0.0597, -0.0222, -0.0082, 0.0, 0.0082, -0.0246, -0.0097, -0.0132, 0.0115, 0.0556, 0.0556, 0.0556, 0.0556, 0.0556, 0.0556, 0.1103, 0.0974, 0.0774, 0.1174, 0.1491, 0.1515, 0.1538, 0.1468, 0.1597, 0.135, 0.1621, 0.1997, 0.1703, 0.1821, 0.2244, 0.2644, 0.3067, 0.2338, 0.1832, 0.1668, 0.1515, 0.1832, 0.2174, 0.2126, 0.2091, 0.1433, 0.1832, 0.1503, 0.135, 0.1832, 0.115, 0.1374, 0.168, 0.2091, 0.2174, 0.1821, 0.1856, 0.2162, 0.2291, 0.2362, 0.275, 0.2879, 0.275, 0.2703, 0.262, 0.2809, 0.2326, 0.2091, 0.2691, 0.2656, 0.2503, 0.2997, 0.3114, 0.3067, 0.3067, 0.3761, 0.4124, 0.4171, 0.4171, 0.4171, 0.5187, 0.6322, 0.6417, 0.6665, 0.6405, 0.6428, 0.6405, 0.6263, 0.6582, 0.7008, 0.7599, 0.7646, 0.7161, 0.7669, 0.8, 0.7977, 0.7977, 0.8851, 0.9194, 0.8993, 0.897, 0.7776, 0.9797, 0.9986, 0.8828, 0.8745, 0.9667, 0.9596, 0.9832, 1.1109, 1.1416, 1.1416, 1.1416, 1.1416, 1.2456, 1.2468, 1.2811, 1.3638, 1.4052, 1.5765, 1.5588, 1.5588, 1.3059, 1.0352, 1.2645, 1.2243, 1.0506, 1.2208, 1.2456, 1.2208, 1.1688, 1.2302, 1.2917, 1.4643, 1.3697]}
{"unit_id": "orig:1970981381628133811", "ticker_idx": 2, "ts": "2025-09-24T22:39:15+00:00", "symbol": "Nanya", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top picks: SK hynix > Samsung Electronics > Nanya", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM 4-year memory supercycle thesis: SK Hynix PT ₩460K, Samsung PT ₩100K; HBM4 TAM $90B by 2027, DRAM prices rising 4 consecutive years.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya in memory context = Nanya Technology, TWSE 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra GPUs will drive 4× memory content growth compared to the previous generation.\n\n2. HBM Outlook\n    • In 2026, HBM3E is expected to face oversupply and downward price pressure, but HBM4 supply constraints could support a 35% price premium.\n    • HBM4 TAM projected to accelerate to $90 billion by 2027.\n    • SK hynix leads HBM4 qualification with early customer sample shipments; Samsung and Micron are expected to compete with less than 40% market share.\n    • Rubin Ultra’s surge in HBM demand → simultaneous growth in pricing and TAM from 4× content increase.\n\n3. DRAM\n    • For the first time ever, DRAM prices are forecast to rise for four consecutive years (2024–2027).\n    • By 2027, HBM will account for 43% of DRAM TAM → reduces price volatility and provides downside margin protection.\n    • Broader demand diversification from new applications such as SOCAMM2 and GDDR7.\n    • DRAM TAM expected to grow at a 27% CAGR through 2027.\n\n4. NAND\n    • Two years of underinvestment and HDD shortages → expanding adoption of eSSD.\n    • NAND ASP projected to rise +7% by 2026, though it is still too early to classify NAND as a structural AI storage essential.\n\n5. Valuation\n    • Stock prices have surpassed prior peaks within just 1–2 months.\n    • While current valuations are at past cycle highs, structurally improved margins mean simple historical comparisons are inappropriate.\n    • Top picks: SK hynix > Samsung Electronics > Nanya.\n\n6. CapEx and Technology Trends\n    • Memory CapEx expected to rise 7–12% in 2026–2027 to meet demand surge.\n    • Investments concentrated in DRAM, prioritizing HBM; NAND focus remains on utilization improvements and gradual transition investments.\n    • From HBM5 generation (post-2027), hybrid bonding adoption expected to accelerate → key equipment vendors to benefit.\n\n7. Target Price Revisions\n    • SK hynix: raised to ₩460,000\n    • Samsung Electronics: raised to ₩100,000", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06578947368421062, "ret_signed_m1d": 0.06578947368421062, "alpha_spy_m1d": 0.06259768513036024, "alpha_c_m1d": 0.06395201595388933, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007894716764751175, "ret_signed_p1d": 0.007894716764751175, "alpha_spy_p1d": 0.012508214253223238, "alpha_c_p1d": 0.008050396357654255, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.002631538792660315, "ret_signed_p1w": 0.002631538792660315, "alpha_spy_p1w": -0.008486425630633398, "alpha_c_p1w": -0.03654509252784299, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.4407894736842106, "ret_signed_p1m": 0.4407894736842106, "alpha_spy_p1m": 0.42466479486164777, "alpha_c_p1m": 0.3670141253427799, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 1.3684210526315788, "ret_signed_p3m": 1.3684210526315788, "alpha_spy_p3m": 1.3294652807597818, "alpha_c_p3m": 1.2447585457782924, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 2.4210526315789473, "ret_signed_p6m": 2.4210526315789473, "alpha_spy_p6m": 2.4200699142746087, "alpha_c_p6m": 2.187366789946754, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.3013, -0.3263, -0.3487, -0.3493, -0.3342, -0.3625, -0.3566, -0.3618, -0.3632, -0.3796, -0.4349, -0.4289, -0.4507, -0.4553, -0.4329, -0.4441, -0.448, -0.4664, -0.4336, -0.4342, -0.4309, -0.4303, -0.4447, -0.4092, -0.4138, -0.4138, -0.4243, -0.4158, -0.4296, -0.4447, -0.4217, -0.4237, -0.3757, -0.3941, -0.3901, -0.3684, -0.377, -0.398, -0.4138, -0.3888, -0.4033, -0.3822, -0.3757, -0.3711, -0.3862, -0.3816, -0.3849, -0.3993, -0.3901, -0.3658, -0.3026, -0.2961, -0.2908, -0.2842, -0.2566, -0.2447, -0.1697, -0.0895, -0.0355, 0.0526, 0.0421, 0.0632, 0.0658, 0.0, 0.0079, -0.0658, -0.0658, -0.0395, 0.0026, 0.0487, 0.0921, 0.0921, 0.2, 0.1947, 0.2961, 0.2961, 0.2579, 0.1724, 0.1355, 0.2487, 0.3684, 0.4079, 0.3882, 0.4474, 0.4408, 0.4408, 0.5789, 0.7105, 0.75, 0.7697, 0.7434, 0.8026, 0.7303, 0.8158, 0.9474, 0.9342, 1.125, 1.1645, 1.1447, 1.1513, 1.0855, 1.1908, 1.1118, 1.1053, 1.0461, 0.8421, 0.8816, 0.8882, 0.7961, 0.9211, 0.9211, 0.9671, 0.9671, 0.9868, 0.9868, 1.0132, 1.1513, 1.1316, 1.0526, 1.0526, 1.1316, 1.1382, 1.0789, 1.1645, 1.2368, 1.2961, 1.3684, 1.3224, 1.4868, 1.4868, 1.4868, 1.4803, 1.5592, 1.5395, 1.5395, 1.7237, 1.7368, 2.0066, 2.1711, 2.1776, 1.8618, 2.1447, 2.0263, 2.1118, 2.125, 2.2895, 2.6184, 2.5789, 2.3026, 2.5066, 2.5724, 2.7632, 2.6447, 2.9079, 2.9079, 3.2961, 2.8684, 2.6513, 2.8816, 2.6184, 2.4737, 2.6118, 2.5658, 2.6579, 2.6579, 2.6579, 2.6579, 2.6579, 2.6579, 2.6579, 2.6579, 2.8026, 2.9211, 2.7895, 2.7566, 2.7566, 2.75, 2.375, 2.0855, 2.3882, 2.0658, 1.7632, 2.0395, 2.3421, 2.1513, 2.1316, 2.4408, 2.5724, 2.6316, 2.4211]}
{"unit_id": "orig:1970981381628133811", "ticker_idx": 3, "ts": "2025-09-24T22:39:15+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and Micron are expected to compete with less than 40% market share", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM 4-year memory supercycle thesis: SK Hynix PT ₩460K, Samsung PT ₩100K; HBM4 TAM $90B by 2027, DRAM prices rising 4 consecutive years.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra GPUs will drive 4× memory content growth compared to the previous generation.\n\n2. HBM Outlook\n    • In 2026, HBM3E is expected to face oversupply and downward price pressure, but HBM4 supply constraints could support a 35% price premium.\n    • HBM4 TAM projected to accelerate to $90 billion by 2027.\n    • SK hynix leads HBM4 qualification with early customer sample shipments; Samsung and Micron are expected to compete with less than 40% market share.\n    • Rubin Ultra’s surge in HBM demand → simultaneous growth in pricing and TAM from 4× content increase.\n\n3. DRAM\n    • For the first time ever, DRAM prices are forecast to rise for four consecutive years (2024–2027).\n    • By 2027, HBM will account for 43% of DRAM TAM → reduces price volatility and provides downside margin protection.\n    • Broader demand diversification from new applications such as SOCAMM2 and GDDR7.\n    • DRAM TAM expected to grow at a 27% CAGR through 2027.\n\n4. NAND\n    • Two years of underinvestment and HDD shortages → expanding adoption of eSSD.\n    • NAND ASP projected to rise +7% by 2026, though it is still too early to classify NAND as a structural AI storage essential.\n\n5. Valuation\n    • Stock prices have surpassed prior peaks within just 1–2 months.\n    • While current valuations are at past cycle highs, structurally improved margins mean simple historical comparisons are inappropriate.\n    • Top picks: SK hynix > Samsung Electronics > Nanya.\n\n6. CapEx and Technology Trends\n    • Memory CapEx expected to rise 7–12% in 2026–2027 to meet demand surge.\n    • Investments concentrated in DRAM, prioritizing HBM; NAND focus remains on utilization improvements and gradual transition investments.\n    • From HBM5 generation (post-2027), hybrid bonding adoption expected to accelerate → key equipment vendors to benefit.\n\n7. Target Price Revisions\n    • SK hynix: raised to ₩460,000\n    • Samsung Electronics: raised to ₩100,000", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029064286680843487, "ret_signed_m1d": 0.029064286680843487, "alpha_spy_m1d": 0.025872498126993104, "alpha_c_m1d": 0.027226828950522197, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03017744814255119, "ret_signed_p1d": -0.03017744814255119, "alpha_spy_p1d": -0.025563950654079126, "alpha_c_p1d": -0.03002176854964811, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.1263989170016231, "ret_signed_p1w": 0.1263989170016231, "alpha_spy_p1w": 0.11528095257832938, "alpha_c_p1w": 0.08722228568111978, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.27907638215343256, "ret_signed_p1m": 0.27907638215343256, "alpha_spy_p1m": 0.2629517033308697, "alpha_c_p1m": 0.20530103381200182, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 0.7114785656886218, "ret_signed_p3m": 0.7114785656886218, "alpha_spy_p3m": 0.6725227938168248, "alpha_c_p3m": 0.5878160588353354, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 1.7501562961848611, "ret_signed_p6m": 1.7501562961848611, "alpha_spy_p6m": 1.7491735788805223, "alpha_c_p6m": 1.516470454552668, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.2292, -0.2385, -0.2531, -0.2479, -0.2445, -0.2445, -0.2584, -0.2306, -0.2441, -0.2387, -0.2299, -0.2665, -0.2573, -0.28, -0.2996, -0.2926, -0.2998, -0.3246, -0.3208, -0.3091, -0.312, -0.312, -0.3076, -0.2905, -0.3251, -0.3514, -0.3336, -0.3256, -0.3273, -0.3082, -0.2648, -0.2349, -0.21, -0.2315, -0.2252, -0.2526, -0.236, -0.2453, -0.2752, -0.284, -0.2723, -0.2801, -0.2796, -0.2718, -0.2456, -0.2641, -0.2641, -0.2673, -0.2658, -0.2319, -0.1876, -0.1871, -0.1637, -0.1343, -0.0689, -0.0277, -0.0244, -0.0179, -0.0106, 0.0444, 0.0063, 0.018, 0.0291, 0.0, -0.0302, -0.0275, 0.0135, 0.0347, 0.1264, 0.1363, 0.1623, 0.1816, 0.149, 0.2161, 0.1901, 0.1237, 0.1928, 0.1575, 0.1877, 0.2532, 0.2523, 0.2794, 0.2517, 0.2281, 0.2791, 0.3552, 0.3619, 0.3731, 0.4023, 0.3861, 0.3846, 0.4523, 0.3491, 0.4696, 0.4747, 0.4722, 0.5674, 0.4919, 0.5154, 0.4662, 0.5273, 0.4971, 0.4139, 0.3979, 0.246, 0.2832, 0.3856, 0.3893, 0.4248, 0.4248, 0.4633, 0.4879, 0.4819, 0.4489, 0.4025, 0.4679, 0.5279, 0.5619, 0.6318, 0.5993, 0.4921, 0.4696, 0.4387, 0.3955, 0.538, 0.6455, 0.7115, 0.7095, 0.7739, 0.7739, 0.7622, 0.8222, 0.8115, 0.7668, 0.7668, 0.9525, 0.9323, 1.1259, 1.1019, 1.0243, 1.1362, 1.141, 1.0931, 1.0635, 1.0838, 1.2455, 1.2455, 1.2595, 1.4087, 1.4611, 1.4739, 1.4086, 1.5395, 1.6945, 1.6977, 1.5682, 1.7101, 1.5965, 1.3486, 1.3702, 1.4432, 1.374, 1.3105, 1.5401, 1.5626, 1.5483, 1.5483, 1.4748, 1.6058, 1.5835, 1.6505, 1.6059, 1.5876, 1.6556, 1.5724, 1.5527, 1.5545, 1.3503, 1.4809, 1.4579, 1.2923, 1.41, 1.4954, 1.5918, 1.5092, 1.6379, 1.7349, 1.858, 1.8582, 1.7502]}
{"unit_id": "orig:1970981381628133811", "ticker_idx": 4, "ts": "2025-09-24T22:39:15+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM prices are forecast to rise for four consecutive years (2024–2027)… DRAM TAM expected to grow at a 27% CAGR through 2027", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM 4-year memory supercycle thesis: SK Hynix PT ₩460K, Samsung PT ₩100K; HBM4 TAM $90B by 2027, DRAM prices rising 4 consecutive years.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra GPUs will drive 4× memory content growth compared to the previous generation.\n\n2. HBM Outlook\n    • In 2026, HBM3E is expected to face oversupply and downward price pressure, but HBM4 supply constraints could support a 35% price premium.\n    • HBM4 TAM projected to accelerate to $90 billion by 2027.\n    • SK hynix leads HBM4 qualification with early customer sample shipments; Samsung and Micron are expected to compete with less than 40% market share.\n    • Rubin Ultra’s surge in HBM demand → simultaneous growth in pricing and TAM from 4× content increase.\n\n3. DRAM\n    • For the first time ever, DRAM prices are forecast to rise for four consecutive years (2024–2027).\n    • By 2027, HBM will account for 43% of DRAM TAM → reduces price volatility and provides downside margin protection.\n    • Broader demand diversification from new applications such as SOCAMM2 and GDDR7.\n    • DRAM TAM expected to grow at a 27% CAGR through 2027.\n\n4. NAND\n    • Two years of underinvestment and HDD shortages → expanding adoption of eSSD.\n    • NAND ASP projected to rise +7% by 2026, though it is still too early to classify NAND as a structural AI storage essential.\n\n5. Valuation\n    • Stock prices have surpassed prior peaks within just 1–2 months.\n    • While current valuations are at past cycle highs, structurally improved margins mean simple historical comparisons are inappropriate.\n    • Top picks: SK hynix > Samsung Electronics > Nanya.\n\n6. CapEx and Technology Trends\n    • Memory CapEx expected to rise 7–12% in 2026–2027 to meet demand surge.\n    • Investments concentrated in DRAM, prioritizing HBM; NAND focus remains on utilization improvements and gradual transition investments.\n    • From HBM5 generation (post-2027), hybrid bonding adoption expected to accelerate → key equipment vendors to benefit.\n\n7. Target Price Revisions\n    • SK hynix: raised to ₩460,000\n    • Samsung Electronics: raised to ₩100,000", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010219261695896894, "ret_signed_m1d": 0.010219261695896894, "alpha_spy_m1d": 0.007027473142046511, "alpha_c_m1d": 0.008381803965575604, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008259317884715492, "ret_signed_p1d": -0.008259317884715492, "alpha_spy_p1d": -0.0036458203962434297, "alpha_c_p1d": -0.008103638291812413, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.04830357054726327, "ret_signed_p1w": 0.04830357054726327, "alpha_spy_p1w": 0.03718560612396956, "alpha_c_p1w": 0.009126939226759966, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.25085205846157504, "ret_signed_p1m": 0.25085205846157504, "alpha_spy_p1m": 0.2347273796390122, "alpha_c_p1m": 0.1770767101201443, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 0.5449006934175141, "ret_signed_p3m": 0.5449006934175141, "alpha_spy_p3m": 0.5059449215457171, "alpha_c_p3m": 0.42123818656422773, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 1.6535661009895588, "ret_signed_p6m": 1.6535661009895588, "alpha_spy_p6m": 1.65258338368522, "alpha_c_p6m": 1.4198802593573656, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.2413, -0.2409, -0.2503, -0.2522, -0.2398, -0.2492, -0.2597, -0.2413, -0.2506, -0.2316, -0.2248, -0.2322, -0.2259, -0.2319, -0.2553, -0.2518, -0.2482, -0.2673, -0.264, -0.2612, -0.2658, -0.252, -0.2492, -0.2348, -0.2417, -0.2747, -0.2656, -0.257, -0.2666, -0.2503, -0.2359, -0.2193, -0.2087, -0.2044, -0.2048, -0.2139, -0.223, -0.2303, -0.2453, -0.2576, -0.245, -0.2393, -0.242, -0.2396, -0.2265, -0.2318, -0.2521, -0.2432, -0.2381, -0.2228, -0.2029, -0.1971, -0.1736, -0.1446, -0.1169, -0.0753, -0.0676, -0.0382, -0.054, -0.007, -0.0197, -0.0075, 0.0102, 0.0, -0.0083, -0.0369, -0.0066, -0.0022, 0.0483, 0.0994, 0.1081, 0.1145, 0.1036, 0.126, 0.1173, 0.1437, 0.1503, 0.1286, 0.1623, 0.2227, 0.2353, 0.2638, 0.2461, 0.2449, 0.2509, 0.3146, 0.3527, 0.3336, 0.3817, 0.3998, 0.4042, 0.4978, 0.4074, 0.4241, 0.4334, 0.4154, 0.4819, 0.4803, 0.4846, 0.4624, 0.4123, 0.4585, 0.3862, 0.3683, 0.3422, 0.2852, 0.3258, 0.3363, 0.3666, 0.3883, 0.3763, 0.3932, 0.42, 0.4077, 0.3853, 0.4219, 0.477, 0.4738, 0.5151, 0.481, 0.4571, 0.4177, 0.3771, 0.4023, 0.4496, 0.4756, 0.5449, 0.5519, 0.5755, 0.5755, 0.605, 0.6754, 0.6836, 0.6687, 0.6687, 0.7888, 0.8376, 0.9333, 0.9485, 0.927, 0.9539, 0.9594, 0.9284, 0.9329, 0.9604, 1.0405, 1.0496, 1.0185, 1.0824, 1.1249, 1.1395, 1.0888, 1.2213, 1.3227, 1.3357, 1.3366, 1.2703, 1.3717, 1.2888, 1.2033, 1.2221, 1.2745, 1.2407, 1.3102, 1.3864, 1.3844, 1.3844, 1.3599, 1.4036, 1.4439, 1.5179, 1.5163, 1.5882, 1.6368, 1.7437, 1.6956, 1.6963, 1.4298, 1.299, 1.4538, 1.3693, 1.2684, 1.4489, 1.5052, 1.446, 1.4529, 1.5655, 1.6233, 1.7613, 1.6536]}
{"unit_id": "orig:1970981381628133811", "ticker_idx": 5, "ts": "2025-09-24T22:39:15+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND ASP projected to rise +7% by 2026", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM 4-year memory supercycle thesis: SK Hynix PT ₩460K, Samsung PT ₩100K; HBM4 TAM $90B by 2027, DRAM prices rising 4 consecutive years.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Entering a 4-Year Memory Supercycle (JPM)\n\n1. Global Memory Cycle\n    • With surging datacenter and AI compute demand, a four-year upcycle is expected from 2024–2027.\n    • Global memory TAM revised up by 6–24% versus prior estimates.\n    • Introduction of HBM4E in Rubin Ultra GPUs will drive 4× memory content growth compared to the previous generation.\n\n2. HBM Outlook\n    • In 2026, HBM3E is expected to face oversupply and downward price pressure, but HBM4 supply constraints could support a 35% price premium.\n    • HBM4 TAM projected to accelerate to $90 billion by 2027.\n    • SK hynix leads HBM4 qualification with early customer sample shipments; Samsung and Micron are expected to compete with less than 40% market share.\n    • Rubin Ultra’s surge in HBM demand → simultaneous growth in pricing and TAM from 4× content increase.\n\n3. DRAM\n    • For the first time ever, DRAM prices are forecast to rise for four consecutive years (2024–2027).\n    • By 2027, HBM will account for 43% of DRAM TAM → reduces price volatility and provides downside margin protection.\n    • Broader demand diversification from new applications such as SOCAMM2 and GDDR7.\n    • DRAM TAM expected to grow at a 27% CAGR through 2027.\n\n4. NAND\n    • Two years of underinvestment and HDD shortages → expanding adoption of eSSD.\n    • NAND ASP projected to rise +7% by 2026, though it is still too early to classify NAND as a structural AI storage essential.\n\n5. Valuation\n    • Stock prices have surpassed prior peaks within just 1–2 months.\n    • While current valuations are at past cycle highs, structurally improved margins mean simple historical comparisons are inappropriate.\n    • Top picks: SK hynix > Samsung Electronics > Nanya.\n\n6. CapEx and Technology Trends\n    • Memory CapEx expected to rise 7–12% in 2026–2027 to meet demand surge.\n    • Investments concentrated in DRAM, prioritizing HBM; NAND focus remains on utilization improvements and gradual transition investments.\n    • From HBM5 generation (post-2027), hybrid bonding adoption expected to accelerate → key equipment vendors to benefit.\n\n7. Target Price Revisions\n    • SK hynix: raised to ₩460,000\n    • Samsung Electronics: raised to ₩100,000", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014041406951468027, "ret_signed_m1d": 0.014041406951468027, "alpha_spy_m1d": 0.010849618397617644, "alpha_c_m1d": 0.012203949221146737, "bench_spy_m1d": 0.003191788553850383, "bench_c_m1d": 0.0018374577303212902, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018478702861400787, "ret_signed_p1d": -0.018478702861400787, "alpha_spy_p1d": -0.013865205372928725, "alpha_c_p1d": -0.018323023268497707, "bench_spy_p1d": -0.0046134974884720625, "bench_c_p1d": -0.00015567959290307964, "ret_p1w": 0.061735662746971884, "ret_signed_p1w": 0.061735662746971884, "alpha_spy_p1w": 0.05061769832367817, "alpha_c_p1w": 0.022559031426468576, "bench_spy_p1w": 0.011117964423293714, "bench_c_p1w": 0.03917663132050331, "ret_p1m": 0.382957950023454, "ret_signed_p1m": 0.382957950023454, "alpha_spy_p1m": 0.36683327120089115, "alpha_c_p1m": 0.3091826016820233, "bench_spy_p1m": 0.016124678822562855, "bench_c_p1m": 0.07377534834143074, "ret_p3m": 0.8189522880393127, "ret_signed_p3m": 0.8189522880393127, "alpha_spy_p3m": 0.7799965161675158, "alpha_c_p3m": 0.6952897811860264, "bench_spy_p3m": 0.038955771871796996, "bench_c_p3m": 0.12366250685328639, "ret_p6m": 3.042383605738995, "ret_signed_p6m": 3.042383605738995, "alpha_spy_p6m": 3.041400888434656, "alpha_c_p6m": 2.808697764106802, "bench_spy_p6m": 0.0009827173043388537, "bench_c_p6m": 0.2336858416321932, "price_path": [-0.3484, -0.3526, -0.3646, -0.3613, -0.348, -0.3604, -0.3658, -0.3491, -0.3484, -0.3416, -0.342, -0.3558, -0.3509, -0.3572, -0.3724, -0.3715, -0.3705, -0.3806, -0.3698, -0.3688, -0.3751, -0.3681, -0.3654, -0.3549, -0.3589, -0.3841, -0.3779, -0.3745, -0.3813, -0.3757, -0.3572, -0.3492, -0.3256, -0.3243, -0.3261, -0.3357, -0.344, -0.349, -0.363, -0.3703, -0.357, -0.3532, -0.352, -0.3492, -0.3342, -0.3291, -0.338, -0.3379, -0.3307, -0.3024, -0.2606, -0.2498, -0.2346, -0.1979, -0.1394, -0.0932, -0.0807, -0.0494, -0.0648, -0.0235, -0.024, -0.0035, 0.014, 0.0, -0.0185, -0.05, 0.0123, 0.0205, 0.0617, 0.1269, 0.1718, 0.1668, 0.1429, 0.1774, 0.1815, 0.1694, 0.2088, 0.1746, 0.2418, 0.3006, 0.287, 0.3437, 0.3312, 0.3367, 0.383, 0.5165, 0.5616, 0.5308, 0.6457, 0.6726, 0.6792, 0.7508, 0.669, 0.7143, 0.7408, 0.8155, 0.9641, 0.9654, 0.9929, 0.8917, 0.7616, 0.8589, 0.737, 0.7515, 0.6559, 0.5776, 0.6555, 0.6417, 0.5896, 0.6294, 0.6531, 0.6158, 0.6358, 0.5981, 0.625, 0.6919, 0.748, 0.7229, 0.7582, 0.7609, 0.6869, 0.6267, 0.5966, 0.6315, 0.6937, 0.7388, 0.819, 0.8239, 0.8748, 0.8825, 0.9252, 0.9259, 0.913, 0.8984, 0.8984, 1.0463, 1.1102, 1.3291, 1.3901, 1.3517, 1.4412, 1.4682, 1.4909, 1.4733, 1.5468, 1.6489, 1.6729, 1.7332, 1.9204, 2.0071, 1.9334, 1.9015, 2.0428, 2.2161, 2.2594, 2.4194, 2.4368, 2.6562, 2.3888, 2.2654, 2.3027, 2.2971, 2.191, 2.3483, 2.5501, 2.6089, 2.5869, 2.503, 2.5104, 2.5692, 2.6436, 2.6758, 2.7319, 2.7163, 2.8104, 2.7472, 2.7273, 2.4068, 2.3571, 2.4256, 2.2851, 2.269, 2.4999, 2.6801, 2.5657, 2.6483, 2.8627, 2.8899, 3.1118, 3.0424]}
{"unit_id": "orig:1970324513415143614", "ticker_idx": 0, "ts": "2025-09-23T03:09:05+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period… this helps to flesh out multi-year visibility to the substantial growth headroom", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS: OpenAI-NVDA deal ~$400B revenue, multi-year visibility bullish; AVGO ASIC intact; AMD most negatively impacted by NVDA dominance.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) UBS Commentary on the OpenAI–NVDA Deal:\n\nWhile timing will, to a large degree, depend on OpenAI's ramp plan, we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period. Given the sheer size of the numbers, it does seem likely there could be some overlap with Oracle's capacity allocated to OpenAI, but our conversation with management suggested this is largely incremental. For NVDA, this helps to flesh out multi-year visibility to the substantial growth headroom it laid out last earnings call when it highlighted a $3-4T TAM by 2030 (of which NVDA's share has been ~30-35%) and we believe it was OpenAI approaching NVDA for help rather than any sort of vendor financing effort on the part of NVDA. As for AVGO, this could ultimately impact the size of its ASIC revenue from OpenAI, but we have seen no signs of any weakening in what we believe to ultimately be a volume target bigger than what Google does with AVGO. Ultimately, this feels the most negative for AMD as it solidifies NVDA's market dominance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02903106179928261, "ret_signed_m1d": 0.02903106179928261, "alpha_spy_m1d": 0.02355779336043673, "alpha_c_m1d": 0.027570091164002664, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008182476268150385, "ret_signed_p1d": -0.008182476268150385, "alpha_spy_p1d": -0.005000842815536699, "alpha_c_p1d": -0.0063483885963997455, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.04567619863458905, "ret_signed_p1w": 0.04567619863458905, "alpha_spy_p1w": 0.041198069911897006, "alpha_c_p1w": 0.031190769468636415, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.01036827301103349, "ret_signed_p1m": 0.01036827301103349, "alpha_spy_p1m": 0.0034474613094788165, "alpha_c_p1m": -0.042165102217478045, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.014403919022290212, "ret_signed_p3m": 0.014403919022290212, "alpha_spy_p3m": -0.014834233590419021, "alpha_c_p3m": -0.09293207957773575, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 0.011151779837404119, "ret_signed_p6m": 0.011151779837404119, "alpha_spy_p6m": 0.010888834159056504, "alpha_c_p6m": -0.21637367581310185, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.1312, -0.1159, -0.1146, -0.1409, -0.1188, -0.107, -0.107, -0.1132, -0.1033, -0.0872, -0.0804, -0.0758, -0.0805, -0.0434, -0.0396, -0.0305, -0.0338, -0.0396, -0.0639, -0.0429, -0.0263, -0.0277, -0.0095, -0.0164, 0.0047, -0.0032, -0.0265, 0.0087, -0.001, 0.0055, 0.0131, 0.0239, 0.0203, 0.0265, 0.0177, 0.0201, 0.0113, 0.02, -0.0157, -0.017, -0.0194, -0.0025, 0.0077, 0.0187, 0.0177, 0.0097, -0.0239, -0.0239, -0.0429, -0.0438, -0.038, -0.064, -0.0568, -0.043, -0.0062, -0.0071, -0.0034, -0.0038, -0.0199, -0.0456, -0.0123, -0.0099, 0.029, 0.0, -0.0082, -0.0041, -0.0013, 0.0192, 0.0457, 0.0494, 0.0586, 0.0515, 0.0398, 0.037, 0.0599, 0.0792, 0.0265, 0.0554, 0.009, 0.0078, 0.0189, 0.0268, 0.0236, 0.0153, 0.0104, 0.0209, 0.0439, 0.0732, 0.1267, 0.1603, 0.1371, 0.1348, 0.1594, 0.1135, 0.094, 0.0541, 0.0545, 0.1156, 0.0826, 0.0861, 0.0472, 0.0658, 0.0458, 0.0164, 0.0453, 0.0124, 0.0025, 0.0231, -0.0034, 0.0103, 0.0103, -0.008, 0.0084, 0.017, 0.0065, 0.0278, 0.0224, 0.04, 0.0367, 0.03, 0.0141, -0.0191, -0.0119, -0.0039, -0.0419, -0.024, 0.0144, 0.0295, 0.0605, 0.0571, 0.0571, 0.0679, 0.0549, 0.0511, 0.0453, 0.0453, 0.0585, 0.0544, 0.0494, 0.0599, 0.0371, 0.0361, 0.0365, 0.0414, 0.0265, 0.0484, 0.0438, 0.0438, -0.002, 0.0275, 0.036, 0.0518, 0.0451, 0.0566, 0.0734, 0.079, 0.0712, 0.0403, 0.0108, -0.0237, -0.0367, 0.0392, 0.0651, 0.0567, 0.0652, 0.0478, 0.0246, 0.0246, 0.0367, 0.0536, 0.0531, 0.0639, 0.0736, 0.0809, 0.0961, 0.0363, -0.0069, 0.0228, 0.0091, 0.0259, 0.0276, -0.0034, 0.0237, 0.0356, 0.0427, 0.0265, 0.0103, 0.027, 0.0197, 0.0112]}
{"unit_id": "orig:1970324513415143614", "ticker_idx": 1, "ts": "2025-09-23T03:09:05+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we have seen no signs of any weakening in what we believe to ultimately be a volume target bigger than what Google does with AVGO", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS: OpenAI-NVDA deal ~$400B revenue, multi-year visibility bullish; AVGO ASIC intact; AMD most negatively impacted by NVDA dominance.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) UBS Commentary on the OpenAI–NVDA Deal:\n\nWhile timing will, to a large degree, depend on OpenAI's ramp plan, we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period. Given the sheer size of the numbers, it does seem likely there could be some overlap with Oracle's capacity allocated to OpenAI, but our conversation with management suggested this is largely incremental. For NVDA, this helps to flesh out multi-year visibility to the substantial growth headroom it laid out last earnings call when it highlighted a $3-4T TAM by 2030 (of which NVDA's share has been ~30-35%) and we believe it was OpenAI approaching NVDA for help rather than any sort of vendor financing effort on the part of NVDA. As for AVGO, this could ultimately impact the size of its ASIC revenue from OpenAI, but we have seen no signs of any weakening in what we believe to ultimately be a volume target bigger than what Google does with AVGO. Ultimately, this feels the most negative for AMD as it solidifies NVDA's market dominance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00044250703604264974, "ret_signed_m1d": -0.00044250703604264974, "alpha_spy_m1d": -0.005915775474888529, "alpha_c_m1d": -0.0019034776713225954, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.001091668021910408, "ret_signed_p1d": 0.001091668021910408, "alpha_spy_p1d": 0.004273301474524094, "alpha_c_p1d": 0.0029257556936610474, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": -0.026641906763270273, "ret_signed_p1w": -0.026641906763270273, "alpha_spy_p1w": -0.031120035485962316, "alpha_c_p1w": -0.04112733592922291, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.004012485659201204, "ret_signed_p1m": 0.004012485659201204, "alpha_spy_p1m": -0.0029083260423534707, "alpha_c_p1m": -0.04852088956931033, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.004189488473618219, "ret_signed_p3m": 0.004189488473618219, "alpha_spy_p3m": -0.025048664139091015, "alpha_c_p3m": -0.10314651012640774, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": -0.06610467519757557, "ret_signed_p6m": -0.06610467519757557, "alpha_spy_p6m": -0.06636762087592318, "alpha_c_p6m": -0.29363013084808154, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.2043, -0.2067, -0.1881, -0.2203, -0.2051, -0.1895, -0.1895, -0.1924, -0.1995, -0.1815, -0.1889, -0.1919, -0.1883, -0.1725, -0.1729, -0.1563, -0.1655, -0.1511, -0.1795, -0.1644, -0.1497, -0.1453, -0.1332, -0.124, -0.1087, -0.135, -0.1499, -0.1231, -0.1372, -0.1115, -0.1053, -0.1018, -0.1049, -0.0786, -0.0896, -0.0833, -0.0977, -0.0994, -0.1314, -0.1424, -0.147, -0.1341, -0.1334, -0.1223, -0.1157, -0.0909, -0.1241, -0.1241, -0.1216, -0.1094, -0.0984, -0.0136, 0.0181, -0.0084, 0.0885, 0.0592, 0.0599, 0.0724, 0.0603, 0.0196, 0.0172, 0.016, -0.0004, 0.0, 0.0011, -0.0084, -0.013, -0.0326, -0.0266, -0.0164, -0.0022, -0.0017, -0.0102, -0.0075, 0.0194, 0.0179, -0.0422, 0.0524, 0.0153, 0.0366, 0.0449, 0.0307, 0.0304, 0.011, 0.004, 0.0158, 0.0448, 0.0682, 0.1004, 0.1388, 0.1107, 0.0905, 0.0697, 0.0384, 0.0591, 0.0491, 0.0309, 0.0574, 0.0384, 0.048, 0.0031, 0.0104, 0.0109, 0.0046, 0.0457, 0.0232, 0.0037, 0.1151, 0.136, 0.173, 0.173, 0.1889, 0.1391, 0.1258, 0.1229, 0.1242, 0.1514, 0.1834, 0.1987, 0.2184, 0.1989, 0.0619, 0.0026, 0.007, -0.0381, -0.0267, 0.0042, 0.0093, 0.0326, 0.0353, 0.0353, 0.0409, 0.0328, 0.0342, 0.0231, 0.0231, 0.0276, 0.0152, 0.0162, 0.0154, -0.0172, 0.0197, 0.0411, 0.0482, 0.0047, 0.014, 0.0397, 0.0397, -0.0168, -0.0281, -0.0378, -0.0539, -0.0397, -0.0163, -0.0149, -0.0224, -0.0207, -0.0212, -0.0531, -0.0894, -0.0821, -0.0159, 0.0167, 0.0063, 0.0132, -0.0211, -0.0388, -0.0388, -0.017, -0.0141, -0.0127, -0.0167, -0.0235, -0.0378, -0.0177, -0.049, -0.0554, -0.0576, -0.0723, -0.0614, -0.0163, -0.0231, 0.022, 0.0127, 0.0097, -0.0069, -0.0477, -0.0395, -0.0502, -0.0661]}
{"unit_id": "orig:1970324513415143614", "ticker_idx": 2, "ts": "2025-09-23T03:09:05+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this feels the most negative for AMD as it solidifies NVDA's market dominance", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS: OpenAI-NVDA deal ~$400B revenue, multi-year visibility bullish; AVGO ASIC intact; AMD most negatively impacted by NVDA dominance.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) UBS Commentary on the OpenAI–NVDA Deal:\n\nWhile timing will, to a large degree, depend on OpenAI's ramp plan, we estimate this deal ultimately translates into ~$400B of NVDA revenue over a multi-year period. Given the sheer size of the numbers, it does seem likely there could be some overlap with Oracle's capacity allocated to OpenAI, but our conversation with management suggested this is largely incremental. For NVDA, this helps to flesh out multi-year visibility to the substantial growth headroom it laid out last earnings call when it highlighted a $3-4T TAM by 2030 (of which NVDA's share has been ~30-35%) and we believe it was OpenAI approaching NVDA for help rather than any sort of vendor financing effort on the part of NVDA. As for AVGO, this could ultimately impact the size of its ASIC revenue from OpenAI, but we have seen no signs of any weakening in what we believe to ultimately be a volume target bigger than what Google does with AVGO. Ultimately, this feels the most negative for AMD as it solidifies NVDA's market dominance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006898698896568578, "ret_signed_m1d": 0.006898698896568578, "alpha_spy_m1d": 0.012371967335414458, "alpha_c_m1d": 0.008359669531848524, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00012423253219473462, "ret_signed_p1d": 0.00012423253219473462, "alpha_spy_p1d": -0.003057400920418951, "alpha_c_p1d": -0.0017098551395559047, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.005531382370474347, "ret_signed_p1w": -0.005531382370474347, "alpha_spy_p1w": -0.001053253647782304, "alpha_c_p1w": 0.008954046795478288, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.4308887785021198, "ret_signed_p1m": -0.4308887785021198, "alpha_spy_p1m": -0.42396796680056514, "alpha_c_p1m": -0.3783554032736083, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.32647607689216107, "ret_signed_p3m": -0.32647607689216107, "alpha_spy_p3m": -0.29723792427945184, "alpha_c_p3m": -0.2191400782921351, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 0.23965204648914118, "ret_signed_p6m": -0.23965204648914118, "alpha_spy_p6m": -0.23938910081079356, "alpha_c_p6m": -0.01212659083863521, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.107, -0.1062, -0.1181, -0.1541, -0.1391, -0.1429, -0.1429, -0.1622, -0.1434, -0.1398, -0.104, -0.09, -0.0911, -0.0329, -0.0051, -0.003, -0.0243, -0.0242, -0.0384, -0.014, 0.0076, 0.0346, 0.0793, 0.1028, 0.1157, 0.0958, 0.0671, 0.0987, 0.0833, 0.0138, 0.0715, 0.0737, 0.0707, 0.0873, 0.1462, 0.1246, 0.1032, 0.0947, 0.0351, 0.0267, 0.0175, 0.0426, 0.0153, 0.0356, 0.0387, 0.0477, 0.0108, 0.0108, 0.0088, 0.0076, 0.0055, -0.0607, -0.059, -0.0316, -0.0085, -0.0325, -0.0145, 0.0016, -0.0027, -0.0108, -0.0185, -0.0218, -0.0069, 0.0, -0.0001, 0.0023, -0.0089, 0.0029, 0.0055, 0.0193, 0.0549, 0.0234, 0.2661, 0.3145, 0.464, 0.4474, 0.3356, 0.3451, 0.3554, 0.4829, 0.4578, 0.4486, 0.4951, 0.4794, 0.4309, 0.4605, 0.5719, 0.6139, 0.6035, 0.6428, 0.5838, 0.5918, 0.6137, 0.5541, 0.5931, 0.4773, 0.4515, 0.5163, 0.4762, 0.609, 0.5411, 0.5339, 0.4948, 0.4313, 0.3894, 0.2804, 0.2665, 0.3365, 0.2811, 0.3315, 0.3315, 0.352, 0.3658, 0.3377, 0.3524, 0.3423, 0.3547, 0.3742, 0.3774, 0.3761, 0.3762, 0.31, 0.2901, 0.3, 0.2313, 0.2496, 0.3265, 0.3359, 0.3356, 0.3365, 0.3365, 0.3362, 0.34, 0.3383, 0.331, 0.331, 0.3889, 0.374, 0.3322, 0.3053, 0.2721, 0.2627, 0.2908, 0.3733, 0.3897, 0.4165, 0.4408, 0.4408, 0.4414, 0.5525, 0.5769, 0.6139, 0.5619, 0.5664, 0.5708, 0.5673, 0.4713, 0.5306, 0.5047, 0.2442, 0.1964, 0.2955, 0.3424, 0.3273, 0.3274, 0.2799, 0.2885, 0.2885, 0.2622, 0.2438, 0.264, 0.2439, 0.2219, 0.329, 0.3105, 0.2659, 0.2443, 0.2344, 0.1868, 0.2559, 0.2396, 0.196, 0.2597, 0.2631, 0.273, 0.229, 0.2019, 0.2218, 0.2201, 0.2397]}
{"unit_id": "orig:1969940613828403290", "ticker_idx": 0, "ts": "2025-09-22T01:43:36+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this certification will serve as a signal that Samsung's HBM competitiveness has been strengthened", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "IM Securities sees Samsung HBM3E NVIDIA cert likely this week/next, limited near-term sales but a positive signal for HBM competitiveness recovery.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electronics HBM3E 12-Hi NVIDIA Certification News – IM Securities Comment by Myungsub Song\n\nThe official announcement is expected this week or next, and as we have mentioned before, the likelihood of passing appears to be very high.\n\nUnlike in the past, when NVIDIA never disclosed the status of certifications, it is reported that during this re-certification process, NVIDIA has been communicating with Samsung almost every week.\n\nHowever, as the certification period has been prolonged, even if certification is passed, sales volume is expected to remain quite limited. That said, this certification will serve as a signal that Samsung’s HBM competitiveness has been strengthened.\n\nFor HBM4, engineering samples have already been delivered to NVIDIA, and the final completed samples are scheduled to be sent around November. Therefore, the final certification results are expected to be announced around January or February next year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03832337845322764, "ret_signed_m1d": -0.03832337845322764, "alpha_spy_m1d": -0.033614647528655284, "alpha_c_m1d": -0.02382532931420156, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.014498049139026081, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014371290578119256, "ret_signed_p1d": 0.014371290578119256, "alpha_spy_p1d": 0.019814765417926816, "alpha_c_p1d": 0.023003602734063433, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.008632312155944177, "ret_p1w": 0.012882246056397761, "ret_signed_p1w": 0.012882246056397761, "alpha_spy_p1w": 0.017620985314265925, "alpha_c_p1w": 0.021125457829964778, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": -0.008243211773567016, "ret_p1m": 0.1728743309827534, "ret_signed_p1m": 0.1728743309827534, "alpha_spy_p1m": 0.16620115882853526, "alpha_c_p1m": 0.15277958060082542, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.020094750381927984, "ret_p3m": 0.294372036283288, "ret_signed_p3m": 0.294372036283288, "alpha_spy_p3m": 0.2799308480600675, "alpha_c_p3m": 0.29288622280311416, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.0014858134801738476, "ret_p6m": 1.343854873661173, "ret_signed_p6m": 1.343854873661173, "alpha_spy_p6m": 1.3349590597408227, "alpha_c_p6m": 1.3550227258714505, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": -0.011167852210277496, "price_path": [-0.2703, -0.2834, -0.2719, -0.2838, -0.279, -0.2719, -0.2359, -0.2419, -0.2611, -0.2647, -0.2766, -0.2695, -0.2503, -0.2515, -0.2371, -0.2251, -0.2012, -0.1964, -0.188, -0.2096, -0.2048, -0.2096, -0.2108, -0.1569, -0.1545, -0.1305, -0.1449, -0.1749, -0.1653, -0.1629, -0.176, -0.1557, -0.1401, -0.1497, -0.1485, -0.1389, -0.1425, -0.1425, -0.1617, -0.1617, -0.1557, -0.1545, -0.1449, -0.1437, -0.1581, -0.1545, -0.1665, -0.1653, -0.1904, -0.1725, -0.1641, -0.1605, -0.1677, -0.1605, -0.1437, -0.1305, -0.121, -0.097, -0.0838, -0.0491, -0.0635, -0.0383, -0.0383, 0.0, 0.0144, 0.0228, 0.0311, -0.0024, 0.0129, 0.0093, 0.0345, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.0796, 0.1356, 0.1224, 0.1019, 0.1428, 0.1753, 0.1777, 0.1801, 0.1729, 0.1861, 0.1608, 0.1885, 0.227, 0.1969, 0.209, 0.2523, 0.2932, 0.3365, 0.2619, 0.2102, 0.1933, 0.1777, 0.2102, 0.2451, 0.2402, 0.2366, 0.1693, 0.2102, 0.1765, 0.1608, 0.2102, 0.1404, 0.1633, 0.1945, 0.2366, 0.2451, 0.209, 0.2126, 0.2438, 0.2571, 0.2643, 0.304, 0.3172, 0.304, 0.2992, 0.2908, 0.31, 0.2607, 0.2366, 0.298, 0.2944, 0.2787, 0.3293, 0.3413, 0.3365, 0.3365, 0.4074, 0.4445, 0.4493, 0.4493, 0.4493, 0.5533, 0.6693, 0.679, 0.7044, 0.6778, 0.6802, 0.6778, 0.6633, 0.6959, 0.7395, 0.7999, 0.8047, 0.7552, 0.8071, 0.841, 0.8386, 0.8386, 0.928, 0.9631, 0.9425, 0.9401, 0.818, 1.0247, 1.0441, 0.9256, 0.9172, 1.0114, 1.0042, 1.0284, 1.1589, 1.1903, 1.1903, 1.1903, 1.1903, 1.2967, 1.2979, 1.333, 1.4176, 1.4599, 1.6352, 1.617, 1.617, 1.3584, 1.0815, 1.3161, 1.275, 1.0973, 1.2713, 1.2967, 1.2713, 1.2181, 1.281, 1.3439]}
{"unit_id": "orig:1963048273419382958", "ticker_idx": 0, "ts": "2025-09-03T01:15:54+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory companies stand to benefit minor positive", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities: VEU termination is a minor positive for memory (Samsung, SK Hynix) via potential supply delays pushing product prices higher.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Impact of VEU Renewal Termination on Memory\n[Samsung Securities – Semiconductor, IT / Jong-Wook Lee]\n\nOver the weekend, the VEU renewal for Samsung Electronics and SK hynix was terminated, and last night the termination of TSMC’s VEU renewal was also mentioned. Below, we share our outlook and investment ideas.\n\n1. If the VEU is abolished, it is not Samsung or SK that apply for export licenses, but AMAT and Lam Research. The inconvenience falls on the equipment makers. The VEU is a system established for the convenience of U.S. equipment companies’ exports.\n\n2. The U.S. Department of Commerce (BIS) also recognizes the necessity of the VEU. There is a high possibility that the VEU will be extended.\n\n3. However, the current U.S. administration emphasizes that the VEU can be abolished at any time, and may use it as a political tool. This process will increase uncertainty.\n\n4. If the VEU is permanently abolished, Samsung and SK’s investment decision-making and subsequent execution could be delayed by about 2–3 months. If supply is delayed, the possibility of product price increases rises.\n\n5. Even if the VEU is abolished, it is unlikely that inconvenient U.S. equipment will be replaced with Chinese or Korean equipment. The benefits for equipment that is not subject to the VEU, or for Chinese equipment, will be limited.\n\nTherefore, regarding this news, memory companies stand to benefit minor positive, while for TSMC and equipment makers the impact is neutral.\n\nThank you.\n\n(Published September 3, 2025)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010028639022655161, "ret_signed_m1d": -0.010028639022655161, "alpha_spy_m1d": -0.0046382926134879154, "alpha_c_m1d": -0.0040978347870296306, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": -0.005930804235625531, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004298117531614576, "ret_signed_p1d": 0.004298117531614576, "alpha_spy_p1d": -0.004059242910199989, "alpha_c_p1d": -0.0006378882629545224, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.004936005794569098, "ret_p1w": 0.0401146692972878, "ret_signed_p1w": 0.0401146692972878, "alpha_spy_p1w": 0.02695711425453995, "alpha_c_p1w": 0.006443350702212802, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.033671318595074995, "ret_p1m": 0.29155346943340743, "ret_signed_p1m": 0.29155346943340743, "alpha_spy_p1m": 0.24909017272607592, "alpha_c_p1m": 0.19521606287821913, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.0963374065551883, "ret_p3m": 0.4505693276497549, "ret_signed_p3m": 0.4505693276497549, "alpha_spy_p3m": 0.39089295114361033, "alpha_c_p3m": 0.3535039654645702, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.09706536218518469, "ret_p6m": 2.1523940606348493, "ret_signed_p6m": 2.1523940606348493, "alpha_spy_p6m": 2.0754786919611243, "alpha_c_p6m": 2.070279047631538, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.08211501300331125, "price_path": [-0.1585, -0.1485, -0.157, -0.1471, -0.1528, -0.1698, -0.1855, -0.1727, -0.1485, -0.157, -0.1528, -0.1741, -0.1385, -0.1271, -0.1428, -0.1289, -0.1433, -0.1375, -0.1289, -0.086, -0.0931, -0.116, -0.1203, -0.1347, -0.1261, -0.1032, -0.1046, -0.0874, -0.0731, -0.0444, -0.0387, -0.0287, -0.0544, -0.0487, -0.0544, -0.0559, 0.0086, 0.0115, 0.0401, 0.0229, -0.0129, -0.0014, 0.0014, -0.0143, 0.01, 0.0287, 0.0172, 0.0186, 0.0301, 0.0258, 0.0258, 0.0029, 0.0029, 0.01, 0.0115, 0.0229, 0.0244, 0.0072, 0.0115, -0.0029, -0.0014, -0.0315, -0.01, 0.0, 0.0043, -0.0043, 0.0043, 0.0244, 0.0401, 0.0516, 0.0802, 0.096, 0.1375, 0.1203, 0.1504, 0.1504, 0.1963, 0.2135, 0.2235, 0.2335, 0.1934, 0.2117, 0.2074, 0.2376, 0.2916, 0.2916, 0.2916, 0.2916, 0.2916, 0.2916, 0.3585, 0.3426, 0.3182, 0.3671, 0.406, 0.4088, 0.4117, 0.4031, 0.4189, 0.3887, 0.4218, 0.4678, 0.4319, 0.4463, 0.4981, 0.547, 0.5988, 0.5096, 0.4477, 0.4275, 0.4088, 0.4477, 0.4894, 0.4837, 0.4794, 0.3988, 0.4477, 0.4074, 0.3887, 0.4477, 0.3642, 0.3916, 0.429, 0.4794, 0.4894, 0.4463, 0.4506, 0.488, 0.5038, 0.5124, 0.5599, 0.5758, 0.5599, 0.5542, 0.5441, 0.5671, 0.5081, 0.4794, 0.5527, 0.5484, 0.5297, 0.5902, 0.6045, 0.5988, 0.5988, 0.6837, 0.728, 0.7338, 0.7338, 0.7338, 0.8582, 0.997, 1.0086, 1.0389, 1.0071, 1.01, 1.0071, 0.9898, 1.0288, 1.0809, 1.1532, 1.159, 1.0997, 1.1618, 1.2023, 1.1994, 1.1994, 1.3065, 1.3484, 1.3238, 1.3209, 1.1749, 1.4221, 1.4453, 1.3036, 1.2934, 1.4062, 1.3976, 1.4265, 1.5826, 1.6202, 1.6202, 1.6202, 1.6202, 1.7475, 1.7489, 1.7909, 1.8921, 1.9427, 2.1524]}
{"unit_id": "orig:1963048273419382958", "ticker_idx": 1, "ts": "2025-09-03T01:15:54+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory companies stand to benefit minor positive", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities: VEU termination is a minor positive for memory (Samsung, SK Hynix) via potential supply delays pushing product prices higher.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Impact of VEU Renewal Termination on Memory\n[Samsung Securities – Semiconductor, IT / Jong-Wook Lee]\n\nOver the weekend, the VEU renewal for Samsung Electronics and SK hynix was terminated, and last night the termination of TSMC’s VEU renewal was also mentioned. Below, we share our outlook and investment ideas.\n\n1. If the VEU is abolished, it is not Samsung or SK that apply for export licenses, but AMAT and Lam Research. The inconvenience falls on the equipment makers. The VEU is a system established for the convenience of U.S. equipment companies’ exports.\n\n2. The U.S. Department of Commerce (BIS) also recognizes the necessity of the VEU. There is a high possibility that the VEU will be extended.\n\n3. However, the current U.S. administration emphasizes that the VEU can be abolished at any time, and may use it as a political tool. This process will increase uncertainty.\n\n4. If the VEU is permanently abolished, Samsung and SK’s investment decision-making and subsequent execution could be delayed by about 2–3 months. If supply is delayed, the possibility of product price increases rises.\n\n5. Even if the VEU is abolished, it is unlikely that inconvenient U.S. equipment will be replaced with Chinese or Korean equipment. The benefits for equipment that is not subject to the VEU, or for Chinese equipment, will be limited.\n\nTherefore, regarding this news, memory companies stand to benefit minor positive, while for TSMC and equipment makers the impact is neutral.\n\nThank you.\n\n(Published September 3, 2025)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00761902022017813, "ret_signed_m1d": -0.00761902022017813, "alpha_spy_m1d": -0.002228673811010884, "alpha_c_m1d": -0.007793655857989301, "bench_spy_m1d": -0.005390346409167246, "bench_c_m1d": 0.0001746356378111713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011428619859768174, "ret_signed_p1d": 0.011428619859768174, "alpha_spy_p1d": 0.003071259417953609, "alpha_c_p1d": -0.0005463189643204469, "bench_spy_p1d": 0.008357360441814565, "bench_c_p1d": 0.011974938824088621, "ret_p1w": 0.1580953708831201, "ret_signed_p1w": 0.1580953708831201, "alpha_spy_p1w": 0.14493781584037224, "alpha_c_p1w": 0.10782136179608193, "bench_spy_p1w": 0.013157555042747848, "bench_c_p1w": 0.05027400908703816, "ret_p1m": 0.5066667844476991, "ret_signed_p1m": 0.5066667844476991, "alpha_spy_p1m": 0.46420348774036757, "alpha_c_p1m": 0.32578487661427546, "bench_spy_p1m": 0.04246329670733151, "bench_c_p1m": 0.18088190783342362, "ret_p3m": 1.0509914675998107, "ret_signed_p3m": 1.0509914675998107, "alpha_spy_p3m": 0.9913150910936661, "alpha_c_p3m": 0.8187182673515159, "bench_spy_p3m": 0.059676376506144546, "bench_c_p3m": 0.2322732002482948, "ret_p6m": 3.197396076244278, "ret_signed_p6m": 3.197396076244278, "alpha_spy_p6m": 3.120480707570553, "alpha_c_p6m": 2.754488463412983, "bench_spy_p6m": 0.07691536867372473, "bench_c_p6m": 0.44290761283129476, "price_path": [-0.146, -0.1289, -0.1232, -0.087, -0.1042, -0.1042, -0.0566, -0.0528, -0.0623, -0.0642, -0.0224, -0.0129, 0.0594, 0.088, 0.1146, 0.0803, 0.1108, 0.0861, 0.0613, 0.0594, 0.029, 0.0309, 0.0727, 0.0689, 0.1298, 0.1203, 0.1412, 0.1355, 0.126, 0.0252, 0.0233, 0.0366, 0.0214, 0.0233, 0.0252, 0.0119, -0.0033, -0.0014, 0.0024, 0.0404, -0.0186, -0.0186, 0.0024, -0.0167, -0.0033, -0.0243, 0.0157, 0.0233, 0.0575, 0.0518, 0.0518, 0.0176, 0.0005, -0.0281, -0.068, -0.0452, -0.0129, -0.0052, -0.011, 0.0229, 0.0248, -0.0248, -0.0076, 0.0, 0.0114, 0.0419, 0.0552, 0.0971, 0.1581, 0.1695, 0.2514, 0.261, 0.3257, 0.2705, 0.3543, 0.3543, 0.3371, 0.3752, 0.3619, 0.3581, 0.2819, 0.3295, 0.3238, 0.3714, 0.5067, 0.5067, 0.5067, 0.5067, 0.5067, 0.5067, 0.6305, 0.581, 0.5676, 0.6095, 0.7238, 0.7733, 0.8495, 0.8248, 0.8343, 0.8229, 0.9429, 1.0381, 0.9848, 1.1257, 1.1638, 1.1295, 1.3619, 1.2324, 1.2057, 1.259, 1.2095, 1.3086, 1.3581, 1.3505, 1.3314, 1.1333, 1.3086, 1.1714, 1.141, 1.1752, 0.9848, 0.981, 0.9771, 0.9962, 1.0739, 1.0205, 1.051, 1.1272, 1.1044, 1.0662, 1.0739, 1.1997, 1.1577, 1.2378, 1.1539, 1.1768, 1.112, 1.0205, 1.1006, 1.1044, 1.0853, 1.2111, 1.2264, 1.2416, 1.2416, 1.2835, 1.4398, 1.4818, 1.4818, 1.4818, 1.5809, 1.6533, 1.7677, 1.8287, 1.8821, 1.8363, 1.8554, 1.8134, 1.8287, 1.8554, 1.8821, 1.9126, 1.8325, 1.8211, 1.8783, 1.924, 1.8058, 2.0498, 2.2061, 2.2823, 2.4653, 2.1642, 2.4577, 2.431, 2.2099, 2.1985, 2.3815, 2.3395, 2.2785, 2.3853, 2.3548, 2.3548, 2.3548, 2.3548, 2.4082, 2.6178, 2.6255, 2.8313, 2.8809, 3.1974]}
{"unit_id": "orig:1945086062835196258", "ticker_idx": 0, "ts": "2025-07-15T11:40:29+00:00", "symbol": "ASM International", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley maintains a constructive view but remains tactically cautious", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: Besi upside on TCB/packaging, ASML raised on China demand; ASMI constructive but cautious on FX; VAT most at risk with order/margin pressure.", "resolved_tickers": ["ASMIY"], "resolution_method": "exchange_ticker", "resolution_reason": "ASM International Amsterdam; US OTC ADR ASMIY", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "European Equipment Outlook: FX Pressure and China Order Pull to Dominate This Earnings Season\n\nMorgan Stanley (N. Putten, 25/07/15)\n\nMorgan Stanley believes that unfavorable exchange rate impacts and potential demand recovery in the Chinese market will be the main conflicting factors in this round of European semiconductor equipment stock earnings reports. ASMI will experience both benefits and pressures, Besi may have upside potential beyond expectations, and VAT faces significant risks.\n\nASMI May Be Dragged Down by FX, Full-Year Outlook Still Supported by China Recovery\n\nASMI's 2Q order flow was significantly affected by the weakening exchange rate (USD vs. EUR), with order forecasts below market consensus. However, a recovery in Chinese demand is expected to provide rebound momentum in the second half. Morgan Stanley maintains a constructive view but remains tactically cautious.\n\nBesi Orders Have Upside Potential, Mainly from TCB and Flip-Chip Packaging Demand\n\nBesi is impacted by exchange rates by approximately 5%, but this may be offset by packaging orders related to NVIDIA B30 and RTX 6000 Pro. The $20 million TCB order announced in May will also provide support.\n\nVAT Group May Be Under Most Pressure, Both Orders and Margins Affected\n\nDue to exchange rate drag on orders by approximately 5%, high inventory limiting new demand, and emerging gross margin pressure, VAT Group may find it difficult to maintain its full-year growth guidance, with order forecasts below market consensus.\n\nAixtron Maintains Neutral Outlook, No Significant Change in Orders and Full-Year Guidance\n\nAixtron's outlook remains neutral. The company is denominated in Euros, so the exchange rate impact is relatively controllable. Orders and full-year forecasts are expected to remain flat.\n\nOverall, FX Drags on ASMI and VAT, ASML and China-Related Exposure May Expand\n\nA weaker USD has a high single-digit percentage negative impact on primarily USD-denominated companies like ASMI and VAT. However, ASML and ASMI may gain some buffer from new demand in the Chinese market. Morgan Stanley has thus raised its 2025 forecasts for both companies.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0029756094451346415, "ret_signed_m1d": -0.0029756094451346415, "alpha_spy_m1d": -0.007267114815106335, "alpha_c_m1d": 0.01586889149892734, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.032536079040410226, "ret_signed_p1d": -0.032536079040410226, "alpha_spy_p1d": -0.035879270158925336, "alpha_c_p1d": -0.02730910413174259, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.089445680435841, "ret_signed_p1w": -0.089445680435841, "alpha_spy_p1w": -0.10024706385665638, "alpha_c_p1w": -0.07077308552013284, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.18391575222339007, "ret_signed_p1m": -0.18391575222339007, "alpha_spy_p1m": -0.22048308614645573, "alpha_c_p1m": -0.21995432991523, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": -0.01105668897560308, "ret_signed_p3m": -0.01105668897560308, "alpha_spy_p3m": -0.06360180711616714, "alpha_c_p3m": -0.12983249067036706, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.19547644413774856, "ret_signed_p6m": 0.19547644413774856, "alpha_spy_p6m": 0.08071914085656151, "alpha_c_p6m": -0.13264605569803023, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2585, -0.2585, -0.2861, -0.2791, -0.2636, -0.2286, -0.2047, -0.2026, -0.1773, -0.2119, -0.2115, -0.1847, -0.1978, -0.2086, -0.1788, -0.1844, -0.1811, -0.1215, -0.0942, -0.0955, -0.1062, -0.1121, -0.1192, -0.1081, -0.1319, -0.1199, -0.1294, -0.1294, -0.0987, -0.1109, -0.0834, -0.1152, -0.112, -0.0975, -0.088, -0.0885, -0.0626, -0.039, -0.003, -0.0035, -0.0114, -0.0284, -0.0196, -0.03, -0.0344, -0.0344, -0.0438, -0.0152, 0.0132, 0.0299, 0.0274, 0.0318, 0.0421, 0.01, -0.0052, -0.0176, -0.0176, -0.0265, -0.0178, -0.0197, 0.003, -0.0033, -0.003, 0.0, -0.0325, -0.0217, -0.0185, -0.0133, -0.0894, -0.1269, -0.1569, -0.1856, -0.1552, -0.167, -0.1642, -0.2129, -0.2166, -0.2105, -0.2163, -0.228, -0.2083, -0.2014, -0.2111, -0.1823, -0.1839, -0.187, -0.213, -0.2187, -0.2215, -0.2262, -0.2407, -0.2223, -0.2282, -0.2205, -0.2205, -0.1867, -0.2195, -0.2195, -0.2314, -0.2363, -0.2218, -0.2046, -0.193, -0.194, -0.1999, -0.1858, -0.1882, -0.1372, -0.118, -0.1106, -0.0451, -0.0566, -0.0325, -0.0462, -0.0434, -0.0532, -0.0605, -0.0419, -0.0228, 0.0032, 0.0559, 0.0414, 0.064, 0.0141, 0.0386, 0.0269, -0.0111, 0.024, 0.0248, 0.0302, 0.0323, 0.0304, 0.035, 0.033, 0.0107, 0.0359, 0.0512, 0.0697, -0.0081, 0.0392, 0.0604, 0.0545, 0.0618, 0.0316, 0.0291, 0.0183, 0.0114, 0.0202, 0.009, -0.0163, -0.0346, -0.0412, -0.0629, -0.0788, -0.0677, -0.097, -0.0976, -0.1066, -0.1103, -0.1016, -0.1016, -0.1019, -0.1135, -0.0885, -0.0509, -0.0404, -0.0225, 0.0476, 0.0437, 0.0485, 0.0466, -0.0183, -0.0189, -0.007, -0.038, -0.0173, -0.0308, -0.0192, -0.0127, -0.0157, -0.0157, -0.0108, -0.0071, -0.01, -0.0188, -0.0188, 0.0623, 0.0965, 0.1875, 0.1955]}
{"unit_id": "orig:1945086062835196258", "ticker_idx": 1, "ts": "2025-07-15T11:40:29+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Besi Orders Have Upside Potential, Mainly from TCB and Flip-Chip Packaging Demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: Besi upside on TCB/packaging, ASML raised on China demand; ASMI constructive but cautious on FX; VAT most at risk with order/margin pressure.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "European Equipment Outlook: FX Pressure and China Order Pull to Dominate This Earnings Season\n\nMorgan Stanley (N. Putten, 25/07/15)\n\nMorgan Stanley believes that unfavorable exchange rate impacts and potential demand recovery in the Chinese market will be the main conflicting factors in this round of European semiconductor equipment stock earnings reports. ASMI will experience both benefits and pressures, Besi may have upside potential beyond expectations, and VAT faces significant risks.\n\nASMI May Be Dragged Down by FX, Full-Year Outlook Still Supported by China Recovery\n\nASMI's 2Q order flow was significantly affected by the weakening exchange rate (USD vs. EUR), with order forecasts below market consensus. However, a recovery in Chinese demand is expected to provide rebound momentum in the second half. Morgan Stanley maintains a constructive view but remains tactically cautious.\n\nBesi Orders Have Upside Potential, Mainly from TCB and Flip-Chip Packaging Demand\n\nBesi is impacted by exchange rates by approximately 5%, but this may be offset by packaging orders related to NVIDIA B30 and RTX 6000 Pro. The $20 million TCB order announced in May will also provide support.\n\nVAT Group May Be Under Most Pressure, Both Orders and Margins Affected\n\nDue to exchange rate drag on orders by approximately 5%, high inventory limiting new demand, and emerging gross margin pressure, VAT Group may find it difficult to maintain its full-year growth guidance, with order forecasts below market consensus.\n\nAixtron Maintains Neutral Outlook, No Significant Change in Orders and Full-Year Guidance\n\nAixtron's outlook remains neutral. The company is denominated in Euros, so the exchange rate impact is relatively controllable. Orders and full-year forecasts are expected to remain flat.\n\nOverall, FX Drags on ASMI and VAT, ASML and China-Related Exposure May Expand\n\nA weaker USD has a high single-digit percentage negative impact on primarily USD-denominated companies like ASMI and VAT. However, ASML and ASMI may gain some buffer from new demand in the Chinese market. Morgan Stanley has thus raised its 2025 forecasts for both companies.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013064138431212902, "ret_signed_m1d": -0.013064138431212902, "alpha_spy_m1d": -0.017355643801184595, "alpha_c_m1d": 0.00578036251284908, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05027718302030382, "ret_signed_p1d": -0.05027718302030382, "alpha_spy_p1d": -0.05362037413881893, "alpha_c_p1d": -0.04505020811163618, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.003958802196844968, "ret_signed_p1w": -0.003958802196844968, "alpha_spy_p1w": -0.014760185617660349, "alpha_c_p1w": 0.014713792718863195, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.022169474665580724, "ret_signed_p1m": -0.022169474665580724, "alpha_spy_p1m": -0.05873680858864638, "alpha_c_p1m": -0.05820805235742066, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.11084713017690517, "ret_signed_p3m": 0.11084713017690517, "alpha_spy_p3m": 0.05830201203634111, "alpha_c_p3m": -0.00792867151785881, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.25732384485191284, "ret_signed_p6m": 0.25732384485191284, "alpha_spy_p6m": 0.1425665415707258, "alpha_c_p6m": -0.07079865498386595, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2728, -0.2728, -0.2728, -0.279, -0.273, -0.2592, -0.2317, -0.2263, -0.2193, -0.2538, -0.2538, -0.2106, -0.2074, -0.2174, -0.1766, -0.1599, -0.1457, -0.0895, -0.0602, -0.0487, -0.0855, -0.1029, -0.1223, -0.1409, -0.1382, -0.1441, -0.1663, -0.1489, -0.1334, -0.1378, -0.1251, -0.1564, -0.177, -0.1548, -0.1453, -0.1243, -0.1006, -0.0907, -0.0475, -0.0622, -0.0281, -0.0032, 0.0055, 0.0044, -0.0044, -0.0127, -0.0249, -0.0048, 0.0206, 0.023, 0.0277, 0.0329, 0.0059, -0.0329, -0.0241, -0.0344, -0.0416, -0.0388, -0.0325, -0.0344, 0.0087, 0.0067, -0.0131, 0.0, -0.0503, 0.0095, 0.0131, 0.0305, -0.004, -0.021, 0.0055, -0.0728, -0.0277, -0.0499, -0.0309, -0.057, -0.0724, -0.0625, -0.0871, -0.0911, -0.0503, -0.0439, -0.0376, -0.0202, -0.0222, -0.0289, -0.0614, -0.0673, -0.0546, -0.0728, -0.0859, -0.0578, -0.0566, -0.0443, -0.0515, -0.0519, -0.0887, -0.0855, -0.1536, -0.1623, -0.1516, -0.1394, -0.1223, -0.1176, -0.1105, -0.1002, -0.1144, -0.0645, -0.0768, -0.0843, -0.0119, -0.038, -0.0277, 0.0048, 0.0178, 0.0032, -0.0309, 0.0008, 0.0044, 0.0119, 0.0574, 0.0333, 0.1619, 0.1623, 0.1587, 0.1512, 0.1108, 0.1302, 0.1124, 0.152, 0.1489, 0.1219, 0.1496, 0.1425, 0.0966, 0.1421, 0.1643, 0.1718, 0.1607, 0.169, 0.1643, 0.1686, 0.1544, 0.1287, 0.1203, 0.0907, 0.0622, 0.0804, 0.0911, 0.0717, 0.0594, 0.0404, 0.0309, 0.0067, 0.0222, 0.0297, -0.0277, -0.015, -0.0135, 0.0261, 0.0285, 0.0285, 0.0273, 0.04, 0.0764, 0.0887, 0.1049, 0.1627, 0.1346, 0.112, 0.0839, 0.0459, 0.0487, 0.0578, 0.0238, 0.0408, 0.0352, 0.0388, 0.0443, 0.0416, 0.0416, 0.0416, 0.0459, 0.0538, 0.059, 0.059, 0.1805, 0.2173, 0.274, 0.2573]}
{"unit_id": "orig:1945086062835196258", "ticker_idx": 2, "ts": "2025-07-15T11:40:29+00:00", "symbol": "VAT Group", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "VAT Group may find it difficult to maintain its full-year growth guidance, with order forecasts below market consensus", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: Besi upside on TCB/packaging, ASML raised on China demand; ASMI constructive but cautious on FX; VAT most at risk with order/margin pressure.", "resolved_tickers": ["VACN.SW"], "resolution_method": "exchange_ticker", "resolution_reason": "VAT Group listed on SIX Swiss Exchange as VACN.SW", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "original", "lang": "en", "tweet_text": "European Equipment Outlook: FX Pressure and China Order Pull to Dominate This Earnings Season\n\nMorgan Stanley (N. Putten, 25/07/15)\n\nMorgan Stanley believes that unfavorable exchange rate impacts and potential demand recovery in the Chinese market will be the main conflicting factors in this round of European semiconductor equipment stock earnings reports. ASMI will experience both benefits and pressures, Besi may have upside potential beyond expectations, and VAT faces significant risks.\n\nASMI May Be Dragged Down by FX, Full-Year Outlook Still Supported by China Recovery\n\nASMI's 2Q order flow was significantly affected by the weakening exchange rate (USD vs. EUR), with order forecasts below market consensus. However, a recovery in Chinese demand is expected to provide rebound momentum in the second half. Morgan Stanley maintains a constructive view but remains tactically cautious.\n\nBesi Orders Have Upside Potential, Mainly from TCB and Flip-Chip Packaging Demand\n\nBesi is impacted by exchange rates by approximately 5%, but this may be offset by packaging orders related to NVIDIA B30 and RTX 6000 Pro. The $20 million TCB order announced in May will also provide support.\n\nVAT Group May Be Under Most Pressure, Both Orders and Margins Affected\n\nDue to exchange rate drag on orders by approximately 5%, high inventory limiting new demand, and emerging gross margin pressure, VAT Group may find it difficult to maintain its full-year growth guidance, with order forecasts below market consensus.\n\nAixtron Maintains Neutral Outlook, No Significant Change in Orders and Full-Year Guidance\n\nAixtron's outlook remains neutral. The company is denominated in Euros, so the exchange rate impact is relatively controllable. Orders and full-year forecasts are expected to remain flat.\n\nOverall, FX Drags on ASMI and VAT, ASML and China-Related Exposure May Expand\n\nA weaker USD has a high single-digit percentage negative impact on primarily USD-denominated companies like ASMI and VAT. However, ASML and ASMI may gain some buffer from new demand in the Chinese market. Morgan Stanley has thus raised its 2025 forecasts for both companies.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027501499809033048, "ret_signed_m1d": 0.027501499809033048, "alpha_spy_m1d": 0.03179300517900474, "alpha_c_m1d": 0.03504083358510024, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": 0.007539333776067192, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03481563329718851, "ret_signed_p1d": 0.03481563329718851, "alpha_spy_p1d": 0.03815882441570362, "alpha_c_p1d": 0.03835174367452443, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": 0.0035361103773359215, "ret_p1w": -0.03393789028248462, "ret_signed_p1w": 0.03393789028248462, "alpha_spy_p1w": 0.0447392737033, "alpha_c_p1w": 0.04154398388037206, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": 0.007606093597887442, "ret_p1m": -0.1711527480713061, "ret_signed_p1m": 0.1711527480713061, "alpha_spy_p1m": 0.20772008199437175, "alpha_c_p1m": 0.18849991231157726, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.017347164240271162, "ret_p3m": 0.018431880291470026, "ret_signed_p3m": -0.018431880291470026, "alpha_spy_p3m": 0.034113237849094036, "alpha_c_p3m": -0.014166555157765215, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.0042653251337048115, "ret_p6m": 0.3007607407384645, "ret_signed_p6m": -0.3007607407384645, "alpha_spy_p6m": -0.18600343745727743, "alpha_c_p6m": -0.23208247411557736, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.06867826662288712, "price_path": [-0.2234, -0.2234, -0.2234, -0.2223, -0.187, -0.1661, -0.1555, -0.1558, -0.1578, -0.1573, -0.1573, -0.1238, -0.1401, -0.1445, -0.1477, -0.1164, -0.1179, -0.0603, -0.0357, -0.0386, -0.0494, -0.0667, -0.0658, -0.091, -0.0889, -0.0922, -0.108, -0.0793, -0.0708, -0.0731, -0.0731, -0.0851, -0.0913, -0.0793, -0.0582, -0.0456, -0.0471, -0.0471, -0.0111, -0.0023, -0.0129, -0.0228, -0.0067, -0.0146, -0.0117, -0.0222, -0.0392, -0.036, -0.0176, -0.0199, -0.0237, -0.0018, -0.0208, -0.0354, -0.0129, -0.0129, -0.0252, -0.0278, -0.0135, -0.0255, 0.007, -0.0099, -0.0275, 0.0, -0.0348, 0.0044, -0.0056, -0.0082, -0.0339, -0.1276, -0.1495, -0.1539, -0.141, -0.1398, -0.1404, -0.1589, -0.1589, -0.1621, -0.172, -0.1864, -0.1668, -0.1656, -0.1688, -0.1609, -0.1712, -0.1744, -0.1984, -0.2095, -0.1972, -0.2045, -0.2101, -0.194, -0.203, -0.1995, -0.2077, -0.213, -0.2352, -0.2142, -0.2461, -0.2285, -0.218, -0.1916, -0.1597, -0.1577, -0.1606, -0.1398, -0.1047, -0.0939, -0.0775, -0.0875, -0.002, -0.0176, 0.012, 0.0237, 0.0146, -0.0061, -0.0831, -0.0854, -0.0802, -0.0521, 0.0415, 0.0295, 0.0854, 0.0664, 0.0544, 0.0524, 0.0184, 0.0483, 0.0293, 0.0494, 0.0138, -0.0111, -0.0105, -0.0138, -0.0483, -0.019, 0.0073, 0.0035, -0.0099, -0.0006, 0.0433, 0.0249, 0.0079, -0.031, -0.0132, -0.026, -0.0404, -0.0293, 0.0018, -0.0047, -0.0418, -0.0328, -0.0588, -0.0919, -0.0535, -0.0135, -0.0644, -0.0249, -0.0202, 0.0184, 0.0278, 0.0339, 0.067, 0.0837, 0.1618, 0.1747, 0.1439, 0.153, 0.1404, 0.1486, 0.1407, 0.1437, 0.1483, 0.1156, 0.0936, 0.1235, 0.1144, 0.1214, 0.1246, 0.1246, 0.1246, 0.1246, 0.1258, 0.129, 0.129, 0.129, 0.129, 0.2657, 0.2876, 0.3008]}
{"unit_id": "orig:1945086062835196258", "ticker_idx": 3, "ts": "2025-07-15T11:40:29+00:00", "symbol": "ASML", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML and ASMI may gain some buffer from new demand in the Chinese market. Morgan Stanley has thus raised its 2025 forecasts for both companies", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: Besi upside on TCB/packaging, ASML raised on China demand; ASMI constructive but cautious on FX; VAT most at risk with order/margin pressure.", "resolved_tickers": ["ASML"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "European Equipment Outlook: FX Pressure and China Order Pull to Dominate This Earnings Season\n\nMorgan Stanley (N. Putten, 25/07/15)\n\nMorgan Stanley believes that unfavorable exchange rate impacts and potential demand recovery in the Chinese market will be the main conflicting factors in this round of European semiconductor equipment stock earnings reports. ASMI will experience both benefits and pressures, Besi may have upside potential beyond expectations, and VAT faces significant risks.\n\nASMI May Be Dragged Down by FX, Full-Year Outlook Still Supported by China Recovery\n\nASMI's 2Q order flow was significantly affected by the weakening exchange rate (USD vs. EUR), with order forecasts below market consensus. However, a recovery in Chinese demand is expected to provide rebound momentum in the second half. Morgan Stanley maintains a constructive view but remains tactically cautious.\n\nBesi Orders Have Upside Potential, Mainly from TCB and Flip-Chip Packaging Demand\n\nBesi is impacted by exchange rates by approximately 5%, but this may be offset by packaging orders related to NVIDIA B30 and RTX 6000 Pro. The $20 million TCB order announced in May will also provide support.\n\nVAT Group May Be Under Most Pressure, Both Orders and Margins Affected\n\nDue to exchange rate drag on orders by approximately 5%, high inventory limiting new demand, and emerging gross margin pressure, VAT Group may find it difficult to maintain its full-year growth guidance, with order forecasts below market consensus.\n\nAixtron Maintains Neutral Outlook, No Significant Change in Orders and Full-Year Guidance\n\nAixtron's outlook remains neutral. The company is denominated in Euros, so the exchange rate impact is relatively controllable. Orders and full-year forecasts are expected to remain flat.\n\nOverall, FX Drags on ASMI and VAT, ASML and China-Related Exposure May Expand\n\nA weaker USD has a high single-digit percentage negative impact on primarily USD-denominated companies like ASMI and VAT. However, ASML and ASMI may gain some buffer from new demand in the Chinese market. Morgan Stanley has thus raised its 2025 forecasts for both companies.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019793023841228274, "ret_signed_m1d": -0.019793023841228274, "alpha_spy_m1d": -0.024084529211199968, "alpha_c_m1d": -0.0009485228971662929, "bench_spy_m1d": 0.004291505369971693, "bench_c_m1d": -0.01884450094406198, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08331522106336908, "ret_signed_p1d": -0.08331522106336908, "alpha_spy_p1d": -0.08665841218188419, "alpha_c_p1d": -0.07808824615470145, "bench_spy_p1d": 0.0033431911185151097, "bench_c_p1d": -0.0052269749086676365, "ret_p1w": -0.14281559745975791, "ret_signed_p1w": -0.14281559745975791, "alpha_spy_p1w": -0.1536169808805733, "alpha_c_p1w": -0.12414300254404975, "bench_spy_p1w": 0.01080138342081538, "bench_c_p1w": -0.018672594915708163, "ret_p1m": -0.07962562461171974, "ret_signed_p1m": -0.07962562461171974, "alpha_spy_p1m": -0.1161929585347854, "alpha_c_p1m": -0.11566420230355967, "bench_spy_p1m": 0.03656733392306566, "bench_c_p1m": 0.036038577691839935, "ret_p3m": 0.1403911443107939, "ret_signed_p3m": 0.1403911443107939, "alpha_spy_p3m": 0.08784602617022985, "alpha_c_p3m": 0.02161534261602993, "bench_spy_p3m": 0.05254511814056406, "bench_c_p3m": 0.11877580169476398, "ret_p6m": 0.49907799327808067, "ret_signed_p6m": 0.49907799327808067, "alpha_spy_p6m": 0.3843206899968936, "alpha_c_p6m": 0.17095549344230188, "bench_spy_p6m": 0.11475730328118705, "bench_c_p6m": 0.3281224998357788, "price_path": [-0.2246, -0.2246, -0.2433, -0.2264, -0.203, -0.1824, -0.1796, -0.1826, -0.1907, -0.1883, -0.1899, -0.1612, -0.1697, -0.1738, -0.149, -0.1397, -0.1419, -0.0885, -0.0718, -0.0657, -0.0786, -0.091, -0.0898, -0.0902, -0.1043, -0.1, -0.11, -0.11, -0.0805, -0.093, -0.0923, -0.1048, -0.0929, -0.0921, -0.0975, -0.0914, -0.0851, -0.0642, -0.0462, -0.0473, -0.0447, -0.0754, -0.0581, -0.0767, -0.0746, -0.0746, -0.0808, -0.0526, -0.0117, -0.0095, -0.0303, -0.0329, -0.0263, -0.0395, -0.0285, -0.0347, -0.0347, -0.0461, -0.0351, -0.0282, -0.0254, -0.0256, -0.0198, 0.0, -0.0833, -0.0949, -0.1075, -0.1256, -0.1428, -0.1289, -0.119, -0.1358, -0.113, -0.1248, -0.1212, -0.1538, -0.1597, -0.1481, -0.1599, -0.1583, -0.1313, -0.1201, -0.1214, -0.0964, -0.0796, -0.0801, -0.096, -0.0894, -0.0942, -0.087, -0.1042, -0.0805, -0.081, -0.0703, -0.062, -0.07, -0.0954, -0.0954, -0.1158, -0.1025, -0.0822, -0.0478, -0.0301, -0.0193, -0.0339, -0.0204, -0.0086, 0.0565, 0.07, 0.0625, 0.1302, 0.1355, 0.1667, 0.1737, 0.1535, 0.1567, 0.1591, 0.1726, 0.1792, 0.2221, 0.2549, 0.2574, 0.2709, 0.2209, 0.2033, 0.1944, 0.1404, 0.1994, 0.1976, 0.2301, 0.242, 0.2538, 0.2695, 0.2486, 0.2322, 0.2625, 0.2584, 0.2912, 0.282, 0.3067, 0.3124, 0.2926, 0.3018, 0.2571, 0.2737, 0.2559, 0.241, 0.2676, 0.2476, 0.2658, 0.2445, 0.2288, 0.2447, 0.2252, 0.2683, 0.1971, 0.1795, 0.2054, 0.2242, 0.2703, 0.2703, 0.2935, 0.3277, 0.353, 0.3922, 0.3546, 0.3417, 0.3663, 0.3563, 0.3659, 0.3702, 0.3189, 0.3274, 0.3131, 0.2391, 0.2646, 0.2886, 0.2898, 0.2957, 0.3002, 0.3002, 0.3091, 0.3008, 0.3083, 0.3055, 0.3055, 0.4201, 0.4987, 0.5158, 0.4991]}
{"unit_id": "orig:1970449673979941047", "ticker_idx": 0, "ts": "2025-09-23T11:26:26+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$35–40 billion could flow into NVIDIA's revenue... structurally expanding demand for NVIDIA", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley view: NVIDIA–OpenAI deal represents long-term upside via $35–40B revenue potential and structural demand expansion from compute arms race.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "On the NVIDIA–OpenAI Deal (Morgan Stanley)\n\n1. Financial Implications\n    • Estimated AI infrastructure build-out cost: $50–60 billion per gigawatt (GW)\n    • Of this, around $35–40 billion could flow into NVIDIA’s revenue (realized over a long-term horizon)\n    • Historical context: the world’s largest GPU clusters today are only a few hundred megawatts in scale; 10 GW represents a disruptive scale by industry standards\n    • Morgan Stanley is not yet incorporating these numbers into earnings estimates, viewing them instead as a long-term upside driver. The emphasis is on growth potential over time rather than immediate, confirmed revenues\n\n2. Industry Implications\n    • OpenAI’s aggressive infrastructure investment—at a scale far exceeding rivals—intensifies the AGI race\n    • As the “compute arms race” accelerates, large tech and cloud companies could pursue similar investments, structurally expanding demand for NVIDIA\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02903106179928261, "ret_signed_m1d": 0.02903106179928261, "alpha_spy_m1d": 0.02355779336043673, "alpha_c_m1d": 0.027570091164002664, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008182476268150385, "ret_signed_p1d": -0.008182476268150385, "alpha_spy_p1d": -0.005000842815536699, "alpha_c_p1d": -0.0063483885963997455, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.04567619863458905, "ret_signed_p1w": 0.04567619863458905, "alpha_spy_p1w": 0.041198069911897006, "alpha_c_p1w": 0.031190769468636415, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.01036827301103349, "ret_signed_p1m": 0.01036827301103349, "alpha_spy_p1m": 0.0034474613094788165, "alpha_c_p1m": -0.042165102217478045, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.014403919022290212, "ret_signed_p3m": 0.014403919022290212, "alpha_spy_p3m": -0.014834233590419021, "alpha_c_p3m": -0.09293207957773575, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 0.011151779837404119, "ret_signed_p6m": 0.011151779837404119, "alpha_spy_p6m": 0.010888834159056504, "alpha_c_p6m": -0.21637367581310185, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.1312, -0.1159, -0.1146, -0.1409, -0.1188, -0.107, -0.107, -0.1132, -0.1033, -0.0872, -0.0804, -0.0758, -0.0805, -0.0434, -0.0396, -0.0305, -0.0338, -0.0396, -0.0639, -0.0429, -0.0263, -0.0277, -0.0095, -0.0164, 0.0047, -0.0032, -0.0265, 0.0087, -0.001, 0.0055, 0.0131, 0.0239, 0.0203, 0.0265, 0.0177, 0.0201, 0.0113, 0.02, -0.0157, -0.017, -0.0194, -0.0025, 0.0077, 0.0187, 0.0177, 0.0097, -0.0239, -0.0239, -0.0429, -0.0438, -0.038, -0.064, -0.0568, -0.043, -0.0062, -0.0071, -0.0034, -0.0038, -0.0199, -0.0456, -0.0123, -0.0099, 0.029, 0.0, -0.0082, -0.0041, -0.0013, 0.0192, 0.0457, 0.0494, 0.0586, 0.0515, 0.0398, 0.037, 0.0599, 0.0792, 0.0265, 0.0554, 0.009, 0.0078, 0.0189, 0.0268, 0.0236, 0.0153, 0.0104, 0.0209, 0.0439, 0.0732, 0.1267, 0.1603, 0.1371, 0.1348, 0.1594, 0.1135, 0.094, 0.0541, 0.0545, 0.1156, 0.0826, 0.0861, 0.0472, 0.0658, 0.0458, 0.0164, 0.0453, 0.0124, 0.0025, 0.0231, -0.0034, 0.0103, 0.0103, -0.008, 0.0084, 0.017, 0.0065, 0.0278, 0.0224, 0.04, 0.0367, 0.03, 0.0141, -0.0191, -0.0119, -0.0039, -0.0419, -0.024, 0.0144, 0.0295, 0.0605, 0.0571, 0.0571, 0.0679, 0.0549, 0.0511, 0.0453, 0.0453, 0.0585, 0.0544, 0.0494, 0.0599, 0.0371, 0.0361, 0.0365, 0.0414, 0.0265, 0.0484, 0.0438, 0.0438, -0.002, 0.0275, 0.036, 0.0518, 0.0451, 0.0566, 0.0734, 0.079, 0.0712, 0.0403, 0.0108, -0.0237, -0.0367, 0.0392, 0.0651, 0.0567, 0.0652, 0.0478, 0.0246, 0.0246, 0.0367, 0.0536, 0.0531, 0.0639, 0.0736, 0.0809, 0.0961, 0.0363, -0.0069, 0.0228, 0.0091, 0.0259, 0.0276, -0.0034, 0.0237, 0.0356, 0.0427, 0.0265, 0.0103, 0.027, 0.0197, 0.0112]}
{"unit_id": "orig:1967534576957628825", "ticker_idx": 0, "ts": "2025-09-15T10:22:52+00:00", "symbol": "commodity memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Commodity Memory Supply Shortage Expected to Deepen Further", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: Nvidia SOCAMM2 transition deepens commodity DRAM shortage; SK Hynix, Samsung, MU, Simmtech, ISC, Hana Micron all expected to benefit.", "resolved_tickers": ["MU", "2408.TW", "2344.TW", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global commodity memory basket: major DRAM producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250915_Global Semiconductors: Commodity Memory Supply Shortage Expected to Deepen Further as Nvidia Pivots to SOCAMM2 - CITI\n\n1. According to Korea’s ETNews, NVIDIA has halted its existing SOCAMM1 project and shifted to SOCAMM2, currently undergoing product sampling in collaboration with SEC, SKH, and MU.\n\n2. SOCAMM2 is expected to be commercialized in early 2026 [with the potential integration of LPDDR6 in the future].\n\n3. This transition is anticipated to further exacerbate the shortage of commodity memory supply.\n\n4. Compared to the previous SOCAMM, SOCAMM2 does not offer a major difference in data transfer speed, but it features improved interfaces optimized for ARM-based CPUs.\n\n5. The data transfer speed reaches 8,533 MT/s, which is approximately 33% faster than standard server DRAM (6,400 MT/s).\n\n6. However, SOCAMM2 has 128 I/O pins, double that of standard server DRAM, significantly enhancing large-scale data processing and scalability.\n\n7. SOCAMM demand in 2026 is projected to reach around 27 billion GB, with each VR200 device expected to require about 3TB of SOCAMM.\n\n8. This implies a sharp increase in memory content per device.\n\n9. Overall, SOCAMM demand in 2026 is expected to account for about 7% of total DRAM demand and around 32% of mobile DRAM demand.\n\n10. Both SK Hynix and Samsung Electronics are expected to benefit, with related beneficiaries including Simmtech, ISC, and Hana Micron.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03321207805643149, "ret_signed_m1d": 0.03321207805643149, "alpha_spy_m1d": 0.027916413132944837, "alpha_c_m1d": 0.02393384004365011, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04710117369436051, "ret_signed_p1d": -0.04710117369436051, "alpha_spy_p1d": -0.0484779664796422, "alpha_c_p1d": -0.0469710178511753, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.15076728452487226, "ret_signed_p1w": -0.15076728452487226, "alpha_spy_p1w": -0.13899744315081014, "alpha_c_p1w": -0.10193442665236213, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.31260480496495574, "ret_signed_p1m": -0.31260480496495574, "alpha_spy_p1m": -0.307829570135777, "alpha_c_p1m": -0.2268868175968493, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.9561564484780662, "ret_signed_p3m": -0.9561564484780662, "alpha_spy_p3m": -0.9105062168037854, "alpha_c_p3m": -0.7487143446573543, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 2.0950730411589347, "ret_signed_p6m": -2.0950730411589347, "alpha_spy_p6m": -2.064578706144821, "alpha_c_p6m": -1.797530006763808, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.2076, -0.2058, -0.2057, -0.2067, -0.1834, -0.1816, -0.1808, -0.1855, -0.1987, -0.2122, -0.2159, -0.2025, -0.2199, -0.2256, -0.2168, -0.2228, -0.2178, -0.227, -0.2315, -0.2352, -0.2417, -0.2492, -0.2516, -0.252, -0.2705, -0.2571, -0.2574, -0.2589, -0.249, -0.2515, -0.2302, -0.2393, -0.2606, -0.2573, -0.2479, -0.2616, -0.2589, -0.2373, -0.2278, -0.2018, -0.2057, -0.2059, -0.2033, -0.2064, -0.2198, -0.2374, -0.2363, -0.2332, -0.2228, -0.2204, -0.21, -0.207, -0.2084, -0.218, -0.2182, -0.2115, -0.1931, -0.1502, -0.1296, -0.1136, -0.0904, -0.0661, -0.0332, 0.0, 0.0471, 0.0605, 0.1334, 0.1321, 0.1508, 0.1793, 0.1478, 0.1314, 0.0798, 0.0991, 0.1279, 0.168, 0.2289, 0.2655, 0.2695, 0.3175, 0.3196, 0.3551, 0.3733, 0.3642, 0.3126, 0.3203, 0.416, 0.4527, 0.4983, 0.4693, 0.4894, 0.4973, 0.538, 0.6305, 0.6726, 0.7131, 0.739, 0.7284, 0.8207, 0.7207, 0.7706, 0.8305, 0.8088, 0.9404, 0.9544, 0.9544, 0.9376, 0.8591, 0.959, 0.858, 0.83, 0.7903, 0.6601, 0.7129, 0.7431, 0.7113, 0.7802, 0.7799, 0.7987, 0.809, 0.8002, 0.8012, 0.8606, 0.9746, 0.9926, 0.9712, 0.9562, 1.0073, 0.9571, 0.913, 0.9755, 1.0264, 1.0324, 1.1078, 1.1006, 1.1839, 1.1839, 1.2006, 1.2472, 1.3225, 1.3004, 1.3004, 1.4824, 1.5533, 1.7417, 1.8103, 1.8135, 1.6738, 1.7768, 1.7122, 1.7363, 1.7529, 1.8975, 2.0573, 2.001, 1.9186, 1.9875, 2.0015, 2.0899, 2.1651, 2.3785, 2.4092, 2.4935, 2.2518, 2.1896, 2.2166, 2.0519, 1.9933, 2.1344, 2.0722, 2.152, 2.202, 2.2011, 2.2011, 2.186, 2.2129, 2.24, 2.2873, 2.3927, 2.5209, 2.4904, 2.5874, 2.5564, 2.5701, 2.218, 1.9967, 2.2315, 2.0596, 1.8776, 2.0951]}
{"unit_id": "orig:1967534576957628825", "ticker_idx": 1, "ts": "2025-09-15T10:22:52+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Both SK Hynix and Samsung Electronics are expected to benefit", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: Nvidia SOCAMM2 transition deepens commodity DRAM shortage; SK Hynix, Samsung, MU, Simmtech, ISC, Hana Micron all expected to benefit.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250915_Global Semiconductors: Commodity Memory Supply Shortage Expected to Deepen Further as Nvidia Pivots to SOCAMM2 - CITI\n\n1. According to Korea’s ETNews, NVIDIA has halted its existing SOCAMM1 project and shifted to SOCAMM2, currently undergoing product sampling in collaboration with SEC, SKH, and MU.\n\n2. SOCAMM2 is expected to be commercialized in early 2026 [with the potential integration of LPDDR6 in the future].\n\n3. This transition is anticipated to further exacerbate the shortage of commodity memory supply.\n\n4. Compared to the previous SOCAMM, SOCAMM2 does not offer a major difference in data transfer speed, but it features improved interfaces optimized for ARM-based CPUs.\n\n5. The data transfer speed reaches 8,533 MT/s, which is approximately 33% faster than standard server DRAM (6,400 MT/s).\n\n6. However, SOCAMM2 has 128 I/O pins, double that of standard server DRAM, significantly enhancing large-scale data processing and scalability.\n\n7. SOCAMM demand in 2026 is projected to reach around 27 billion GB, with each VR200 device expected to require about 3TB of SOCAMM.\n\n8. This implies a sharp increase in memory content per device.\n\n9. Overall, SOCAMM demand in 2026 is expected to account for about 7% of total DRAM demand and around 32% of mobile DRAM demand.\n\n10. Both SK Hynix and Samsung Electronics are expected to benefit, with related beneficiaries including Simmtech, ISC, and Hana Micron.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007552760263550007, "ret_signed_m1d": -0.007552760263550007, "alpha_spy_m1d": -0.002257095340063353, "alpha_c_m1d": 0.0017254777492313744, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05135954607514148, "ret_signed_p1d": 0.05135954607514148, "alpha_spy_p1d": 0.05273633886042317, "alpha_c_p1d": 0.05122939023195627, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.06042302879498718, "ret_signed_p1w": 0.06042302879498718, "alpha_spy_p1w": 0.04865318742092506, "alpha_c_p1w": 0.01159017092247705, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.24320245896160975, "ret_signed_p1m": 0.24320245896160975, "alpha_spy_p1m": 0.238427224132431, "alpha_c_p1m": 0.15748447159350332, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.7081711738099534, "ret_signed_p3m": 0.7081711738099534, "alpha_spy_p3m": 0.6625209421356726, "alpha_c_p3m": 0.5007290699892415, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 1.8410994136412615, "ret_signed_p6m": 1.8410994136412615, "alpha_spy_p6m": 1.8106050786271475, "alpha_c_p6m": 1.5435563792461349, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.2564, -0.2579, -0.2247, -0.2171, -0.1598, -0.1372, -0.1161, -0.1432, -0.1191, -0.1387, -0.1583, -0.1598, -0.184, -0.1824, -0.1493, -0.1523, -0.104, -0.1116, -0.095, -0.0995, -0.107, -0.187, -0.1885, -0.1779, -0.19, -0.1885, -0.187, -0.1975, -0.2096, -0.2081, -0.2051, -0.1749, -0.2217, -0.2217, -0.2051, -0.2202, -0.2096, -0.2262, -0.1945, -0.1885, -0.1613, -0.1659, -0.1659, -0.193, -0.2066, -0.2292, -0.2609, -0.2428, -0.2171, -0.2111, -0.2156, -0.1888, -0.1873, -0.2266, -0.213, -0.2069, -0.1979, -0.1737, -0.1631, -0.1299, -0.0816, -0.0725, -0.0076, 0.0, 0.0514, 0.0076, 0.074, 0.074, 0.0604, 0.0906, 0.0801, 0.077, 0.0166, 0.0544, 0.0498, 0.0876, 0.1949, 0.1949, 0.1949, 0.1949, 0.1949, 0.1949, 0.2931, 0.2538, 0.2432, 0.2764, 0.3671, 0.4063, 0.4668, 0.4471, 0.4547, 0.4456, 0.5408, 0.6163, 0.574, 0.6858, 0.716, 0.6888, 0.8731, 0.7704, 0.7492, 0.7915, 0.7523, 0.8308, 0.8701, 0.864, 0.8489, 0.6918, 0.8308, 0.7221, 0.6979, 0.7251, 0.574, 0.571, 0.568, 0.5831, 0.6447, 0.6024, 0.6265, 0.687, 0.6689, 0.6386, 0.6447, 0.7445, 0.7112, 0.7747, 0.7082, 0.7263, 0.6749, 0.6024, 0.6658, 0.6689, 0.6538, 0.7535, 0.7656, 0.7777, 0.7777, 0.811, 0.9349, 0.9682, 0.9682, 0.9682, 1.0468, 1.1042, 1.1949, 1.2433, 1.2856, 1.2493, 1.2645, 1.2312, 1.2433, 1.2645, 1.2856, 1.3098, 1.2463, 1.2373, 1.2826, 1.3189, 1.2252, 1.4186, 1.5426, 1.6031, 1.7482, 1.5093, 1.7421, 1.721, 1.5456, 1.5366, 1.6817, 1.6484, 1.6, 1.6847, 1.6605, 1.6605, 1.6605, 1.6605, 1.7028, 1.8691, 1.8752, 2.0384, 2.0777, 2.3288, 2.2137, 2.2137, 1.8441, 1.5715, 1.8502, 1.7987, 1.5322, 1.8411]}
{"unit_id": "orig:1967534576957628825", "ticker_idx": 2, "ts": "2025-09-15T10:22:52+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Both SK Hynix and Samsung Electronics are expected to benefit", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: Nvidia SOCAMM2 transition deepens commodity DRAM shortage; SK Hynix, Samsung, MU, Simmtech, ISC, Hana Micron all expected to benefit.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "250915_Global Semiconductors: Commodity Memory Supply Shortage Expected to Deepen Further as Nvidia Pivots to SOCAMM2 - CITI\n\n1. According to Korea’s ETNews, NVIDIA has halted its existing SOCAMM1 project and shifted to SOCAMM2, currently undergoing product sampling in collaboration with SEC, SKH, and MU.\n\n2. SOCAMM2 is expected to be commercialized in early 2026 [with the potential integration of LPDDR6 in the future].\n\n3. This transition is anticipated to further exacerbate the shortage of commodity memory supply.\n\n4. Compared to the previous SOCAMM, SOCAMM2 does not offer a major difference in data transfer speed, but it features improved interfaces optimized for ARM-based CPUs.\n\n5. The data transfer speed reaches 8,533 MT/s, which is approximately 33% faster than standard server DRAM (6,400 MT/s).\n\n6. However, SOCAMM2 has 128 I/O pins, double that of standard server DRAM, significantly enhancing large-scale data processing and scalability.\n\n7. SOCAMM demand in 2026 is projected to reach around 27 billion GB, with each VR200 device expected to require about 3TB of SOCAMM.\n\n8. This implies a sharp increase in memory content per device.\n\n9. Overall, SOCAMM demand in 2026 is expected to account for about 7% of total DRAM demand and around 32% of mobile DRAM demand.\n\n10. Both SK Hynix and Samsung Electronics are expected to benefit, with related beneficiaries including Simmtech, ISC, and Hana Micron.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014379050266663418, "ret_signed_m1d": -0.014379050266663418, "alpha_spy_m1d": -0.009083385343176764, "alpha_c_m1d": -0.005451937235376625, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.008927113031286793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03790862122232053, "ret_signed_p1d": 0.03790862122232053, "alpha_spy_p1d": 0.039285414007602215, "alpha_c_p1d": 0.04138439935144289, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": -0.0034757781291223644, "ret_p1w": 0.09150334763684276, "ret_signed_p1w": 0.09150334763684276, "alpha_spy_p1w": 0.07973350626278064, "alpha_c_p1w": 0.05600184578992917, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.035501501846913586, "ret_p1m": 0.20272799943230813, "ret_signed_p1m": 0.20272799943230813, "alpha_spy_p1m": 0.19795276460312938, "alpha_c_p1m": 0.17143938193303665, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.031288617499271476, "ret_p3m": 0.40887237642603425, "ret_signed_p3m": 0.40887237642603425, "alpha_spy_p3m": 0.3632221447517534, "alpha_c_p3m": 0.324720630975325, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.08415174545070925, "ret_p6m": 1.4791613363774125, "ret_signed_p6m": 1.4791613363774125, "alpha_spy_p6m": 1.4486670013632985, "alpha_c_p6m": 1.4536098155675308, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.025551520809881767, "price_path": [-0.2231, -0.2309, -0.227, -0.2465, -0.214, -0.2036, -0.2179, -0.2052, -0.2183, -0.2131, -0.2052, -0.166, -0.1725, -0.1935, -0.1974, -0.2105, -0.2026, -0.1817, -0.183, -0.1673, -0.1542, -0.1281, -0.1229, -0.1137, -0.1373, -0.132, -0.1373, -0.1386, -0.0797, -0.0771, -0.051, -0.0667, -0.0993, -0.0889, -0.0863, -0.1007, -0.0784, -0.0614, -0.0719, -0.0706, -0.0601, -0.0641, -0.0641, -0.085, -0.085, -0.0784, -0.0771, -0.0667, -0.0654, -0.081, -0.0771, -0.0902, -0.0889, -0.1163, -0.0967, -0.0876, -0.0837, -0.0915, -0.0837, -0.0654, -0.051, -0.0405, -0.0144, 0.0, 0.0379, 0.0222, 0.0497, 0.0497, 0.0915, 0.1072, 0.1163, 0.1255, 0.0889, 0.1056, 0.1016, 0.1292, 0.1784, 0.1784, 0.1784, 0.1784, 0.1784, 0.1784, 0.2395, 0.225, 0.2027, 0.2474, 0.2828, 0.2854, 0.2881, 0.2802, 0.2946, 0.2671, 0.2973, 0.3393, 0.3065, 0.3196, 0.3669, 0.4115, 0.4588, 0.3774, 0.3209, 0.3025, 0.2854, 0.3209, 0.359, 0.3537, 0.3498, 0.2763, 0.3209, 0.2841, 0.2671, 0.3209, 0.2447, 0.2697, 0.3038, 0.3498, 0.359, 0.3196, 0.3235, 0.3577, 0.3721, 0.38, 0.4233, 0.4378, 0.4233, 0.4181, 0.4089, 0.4299, 0.376, 0.3498, 0.4168, 0.4128, 0.3957, 0.4509, 0.464, 0.4588, 0.4588, 0.5362, 0.5767, 0.582, 0.582, 0.582, 0.6954, 0.8221, 0.8327, 0.8604, 0.8313, 0.834, 0.8313, 0.8155, 0.8511, 0.8986, 0.9646, 0.9699, 0.9158, 0.9725, 1.0095, 1.0068, 1.0068, 1.1045, 1.1427, 1.1203, 1.1176, 0.9844, 1.21, 1.2311, 1.1018, 1.0926, 1.1955, 1.1876, 1.214, 1.3565, 1.3908, 1.3908, 1.3908, 1.3908, 1.5069, 1.5082, 1.5465, 1.6388, 1.685, 1.8763, 1.8565, 1.8565, 1.5742, 1.272, 1.528, 1.4831, 1.2892, 1.4792]}
{"unit_id": "orig:1967534576957628825", "ticker_idx": 3, "ts": "2025-09-15T10:22:52+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "currently undergoing product sampling in collaboration with SEC, SKH, and MU", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: Nvidia SOCAMM2 transition deepens commodity DRAM shortage; SK Hynix, Samsung, MU, Simmtech, ISC, Hana Micron all expected to benefit.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250915_Global Semiconductors: Commodity Memory Supply Shortage Expected to Deepen Further as Nvidia Pivots to SOCAMM2 - CITI\n\n1. According to Korea’s ETNews, NVIDIA has halted its existing SOCAMM1 project and shifted to SOCAMM2, currently undergoing product sampling in collaboration with SEC, SKH, and MU.\n\n2. SOCAMM2 is expected to be commercialized in early 2026 [with the potential integration of LPDDR6 in the future].\n\n3. This transition is anticipated to further exacerbate the shortage of commodity memory supply.\n\n4. Compared to the previous SOCAMM, SOCAMM2 does not offer a major difference in data transfer speed, but it features improved interfaces optimized for ARM-based CPUs.\n\n5. The data transfer speed reaches 8,533 MT/s, which is approximately 33% faster than standard server DRAM (6,400 MT/s).\n\n6. However, SOCAMM2 has 128 I/O pins, double that of standard server DRAM, significantly enhancing large-scale data processing and scalability.\n\n7. SOCAMM demand in 2026 is projected to reach around 27 billion GB, with each VR200 device expected to require about 3TB of SOCAMM.\n\n8. This implies a sharp increase in memory content per device.\n\n9. Overall, SOCAMM demand in 2026 is expected to account for about 7% of total DRAM demand and around 32% of mobile DRAM demand.\n\n10. Both SK Hynix and Samsung Electronics are expected to benefit, with related beneficiaries including Simmtech, ISC, and Hana Micron.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0034227847689322077, "ret_signed_m1d": -0.0034227847689322077, "alpha_spy_m1d": 0.001872880154554446, "alpha_c_m1d": 0.0058554532438491735, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00665524802138262, "ret_signed_p1d": 0.00665524802138262, "alpha_spy_p1d": 0.008032040806664309, "alpha_c_p1d": 0.006525092178197411, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.04341761654698839, "ret_signed_p1w": 0.04341761654698839, "alpha_spy_p1w": 0.031647775172926274, "alpha_c_p1w": -0.005415241325521736, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.18639236993776254, "ret_signed_p1m": 0.18639236993776254, "alpha_spy_p1m": 0.1816171351085838, "alpha_c_p1m": 0.1006743825696561, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.6392333132583683, "ret_signed_p3m": 0.6392333132583683, "alpha_spy_p3m": 0.5935830815840875, "alpha_c_p3m": 0.43179120943765636, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 1.5576811730081666, "ret_signed_p6m": 1.5576811730081666, "alpha_spy_p6m": 1.5271868379940525, "alpha_c_p6m": 1.26013813861304, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.2286, -0.2286, -0.2173, -0.2269, -0.19, -0.1942, -0.2021, -0.21, -0.2195, -0.2345, -0.2291, -0.2256, -0.2256, -0.2399, -0.2114, -0.2252, -0.2197, -0.2107, -0.2482, -0.2387, -0.262, -0.2821, -0.275, -0.2823, -0.3077, -0.3039, -0.2918, -0.2948, -0.2949, -0.2904, -0.2727, -0.3082, -0.3352, -0.3169, -0.3087, -0.3105, -0.2909, -0.2464, -0.2158, -0.1903, -0.2123, -0.2059, -0.2339, -0.2169, -0.2264, -0.2571, -0.2661, -0.2541, -0.2621, -0.2616, -0.2537, -0.2267, -0.2457, -0.2457, -0.249, -0.2475, -0.2127, -0.1673, -0.1668, -0.1428, -0.1126, -0.0456, -0.0034, 0.0, 0.0067, 0.0141, 0.0705, 0.0314, 0.0434, 0.0548, 0.025, -0.006, -0.0032, 0.0389, 0.0605, 0.1545, 0.1647, 0.1913, 0.2111, 0.1777, 0.2465, 0.2198, 0.1518, 0.2226, 0.1864, 0.2173, 0.2845, 0.2836, 0.3114, 0.283, 0.2588, 0.311, 0.3891, 0.3959, 0.4074, 0.4374, 0.4207, 0.4192, 0.4885, 0.3828, 0.5063, 0.5116, 0.509, 0.6065, 0.5292, 0.5532, 0.5028, 0.5655, 0.5345, 0.4492, 0.4329, 0.2772, 0.3152, 0.4202, 0.424, 0.4604, 0.4604, 0.4998, 0.5251, 0.5189, 0.4851, 0.4375, 0.5045, 0.566, 0.6009, 0.6725, 0.6392, 0.5294, 0.5063, 0.4747, 0.4303, 0.5764, 0.6865, 0.7542, 0.7522, 0.8182, 0.8182, 0.8062, 0.8677, 0.8567, 0.8109, 0.8109, 1.0013, 0.9806, 1.179, 1.1544, 1.0749, 1.1896, 1.1945, 1.1454, 1.1151, 1.1359, 1.3016, 1.3016, 1.3159, 1.4689, 1.5226, 1.5357, 1.4687, 1.6029, 1.7618, 1.765, 1.6324, 1.7778, 1.6613, 1.4072, 1.4294, 1.5043, 1.4333, 1.3682, 1.6036, 1.6266, 1.6119, 1.6119, 1.5366, 1.6709, 1.648, 1.7167, 1.671, 1.6522, 1.722, 1.6367, 1.6164, 1.6183, 1.409, 1.5428, 1.5192, 1.3495, 1.4702, 1.5577]}
{"unit_id": "orig:1967534576957628825", "ticker_idx": 5, "ts": "2025-09-15T10:22:52+00:00", "symbol": "ISC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "related beneficiaries including Simmtech, ISC, and Hana Micron", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: Nvidia SOCAMM2 transition deepens commodity DRAM shortage; SK Hynix, Samsung, MU, Simmtech, ISC, Hana Micron all expected to benefit.", "resolved_tickers": ["095340.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "ISC Co. 095340.KQ Korea KOSDAQ socket/test", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "250915_Global Semiconductors: Commodity Memory Supply Shortage Expected to Deepen Further as Nvidia Pivots to SOCAMM2 - CITI\n\n1. According to Korea’s ETNews, NVIDIA has halted its existing SOCAMM1 project and shifted to SOCAMM2, currently undergoing product sampling in collaboration with SEC, SKH, and MU.\n\n2. SOCAMM2 is expected to be commercialized in early 2026 [with the potential integration of LPDDR6 in the future].\n\n3. This transition is anticipated to further exacerbate the shortage of commodity memory supply.\n\n4. Compared to the previous SOCAMM, SOCAMM2 does not offer a major difference in data transfer speed, but it features improved interfaces optimized for ARM-based CPUs.\n\n5. The data transfer speed reaches 8,533 MT/s, which is approximately 33% faster than standard server DRAM (6,400 MT/s).\n\n6. However, SOCAMM2 has 128 I/O pins, double that of standard server DRAM, significantly enhancing large-scale data processing and scalability.\n\n7. SOCAMM demand in 2026 is projected to reach around 27 billion GB, with each VR200 device expected to require about 3TB of SOCAMM.\n\n8. This implies a sharp increase in memory content per device.\n\n9. Overall, SOCAMM demand in 2026 is expected to account for about 7% of total DRAM demand and around 32% of mobile DRAM demand.\n\n10. Both SK Hynix and Samsung Electronics are expected to benefit, with related beneficiaries including Simmtech, ISC, and Hana Micron.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028328656143128184, "ret_signed_m1d": -0.028328656143128184, "alpha_spy_m1d": -0.02303299121964153, "alpha_c_m1d": -0.02163590750409805, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.006692748639030133, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06373930898070723, "ret_signed_p1d": 0.06373930898070723, "alpha_spy_p1d": 0.06511610176598892, "alpha_c_p1d": 0.04479753595297087, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.018941773027736364, "ret_p1w": 0.11473089003833792, "ret_signed_p1w": 0.11473089003833792, "alpha_spy_p1w": 0.1029610486642758, "alpha_c_p1w": 0.08480294972872704, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.02992794030961088, "ret_p1m": 0.1218129982936762, "ret_signed_p1m": 0.1218129982936762, "alpha_spy_p1m": 0.11703776346449746, "alpha_c_p1m": 0.077615560731223, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.0441974375624532, "ret_p3m": 0.5410764398466483, "ret_signed_p3m": 0.5410764398466483, "alpha_spy_p3m": 0.4954262081723675, "alpha_c_p3m": 0.3419352023806528, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.19914123746599555, "ret_p6m": 1.6503296903125544, "ret_signed_p6m": 1.6503296903125544, "alpha_spy_p6m": 1.6198353552984404, "alpha_c_p6m": 0.9679634704201712, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.6823662198923832, "price_path": [-0.2465, -0.2422, -0.2436, -0.2734, -0.2436, -0.1218, -0.1218, -0.1586, -0.153, -0.1686, -0.1813, -0.1317, -0.1416, -0.1346, -0.1431, -0.1473, -0.1558, -0.1388, -0.1076, -0.0694, -0.0864, -0.153, -0.1289, -0.1346, -0.1657, -0.1714, -0.1657, -0.1742, -0.1317, -0.1572, -0.1813, -0.1431, -0.1756, -0.1785, -0.1586, -0.1728, -0.1586, -0.1572, -0.1487, -0.1728, -0.0878, -0.0878, -0.0878, -0.1346, -0.136, -0.1686, -0.1898, -0.1884, -0.1799, -0.1728, -0.1799, -0.1771, -0.1756, -0.1983, -0.2139, -0.1827, -0.1034, -0.085, -0.0453, -0.0411, -0.0297, -0.0227, -0.0283, 0.0, 0.0637, 0.0439, 0.0666, 0.0666, 0.1147, 0.1261, 0.1374, 0.1091, 0.0907, 0.119, 0.102, 0.1317, 0.153, 0.153, 0.153, 0.153, 0.153, 0.153, 0.1431, 0.1501, 0.1218, 0.1416, 0.1983, 0.119, 0.1275, 0.0892, 0.1076, 0.1076, 0.1374, 0.1331, 0.1261, 0.1261, 0.1317, 0.1289, 0.1416, 0.1544, 0.0482, 0.1331, 0.0836, 0.1289, 0.1204, 0.1161, 0.0949, 0.0482, 0.1785, 0.2479, 0.2507, 0.2465, 0.1374, 0.2493, 0.4533, 0.4773, 0.4589, 0.585, 0.5737, 0.5694, 0.6147, 0.5623, 0.551, 0.5595, 0.5538, 0.5283, 0.5411, 0.5071, 0.4589, 0.3952, 0.4164, 0.3909, 0.4363, 0.5113, 0.5198, 0.4943, 0.4943, 0.4887, 0.5565, 0.5865, 0.5865, 0.5865, 0.7136, 0.6022, 0.665, 0.5651, 0.5865, 0.5636, 0.5579, 0.5551, 0.5308, 0.5065, 0.5037, 0.4665, 0.418, 0.4251, 0.5179, 0.4894, 0.6122, 0.6836, 0.8792, 0.9806, 1.152, 0.9592, 1.1377, 1.0434, 1.1248, 1.3347, 1.5104, 1.4119, 1.4219, 1.5075, 1.4376, 1.4376, 1.4376, 1.4376, 1.6989, 1.6703, 1.5532, 1.6061, 1.6046, 1.7517, 1.7531, 1.7531, 1.5775, 1.3405, 1.8845, 2.0202, 1.7346, 1.6503]}
{"unit_id": "orig:1967534576957628825", "ticker_idx": 6, "ts": "2025-09-15T10:22:52+00:00", "symbol": "Hana Micron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "related beneficiaries including Simmtech, ISC, and Hana Micron", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi: Nvidia SOCAMM2 transition deepens commodity DRAM shortage; SK Hynix, Samsung, MU, Simmtech, ISC, Hana Micron all expected to benefit.", "resolved_tickers": ["067310.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "Hana Micron 067310.KQ Korea KOSDAQ semiconductor pkg", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "250915_Global Semiconductors: Commodity Memory Supply Shortage Expected to Deepen Further as Nvidia Pivots to SOCAMM2 - CITI\n\n1. According to Korea’s ETNews, NVIDIA has halted its existing SOCAMM1 project and shifted to SOCAMM2, currently undergoing product sampling in collaboration with SEC, SKH, and MU.\n\n2. SOCAMM2 is expected to be commercialized in early 2026 [with the potential integration of LPDDR6 in the future].\n\n3. This transition is anticipated to further exacerbate the shortage of commodity memory supply.\n\n4. Compared to the previous SOCAMM, SOCAMM2 does not offer a major difference in data transfer speed, but it features improved interfaces optimized for ARM-based CPUs.\n\n5. The data transfer speed reaches 8,533 MT/s, which is approximately 33% faster than standard server DRAM (6,400 MT/s).\n\n6. However, SOCAMM2 has 128 I/O pins, double that of standard server DRAM, significantly enhancing large-scale data processing and scalability.\n\n7. SOCAMM demand in 2026 is projected to reach around 27 billion GB, with each VR200 device expected to require about 3TB of SOCAMM.\n\n8. This implies a sharp increase in memory content per device.\n\n9. Overall, SOCAMM demand in 2026 is expected to account for about 7% of total DRAM demand and around 32% of mobile DRAM demand.\n\n10. Both SK Hynix and Samsung Electronics are expected to benefit, with related beneficiaries including Simmtech, ISC, and Hana Micron.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.043173160780352715, "ret_signed_m1d": -0.043173160780352715, "alpha_spy_m1d": -0.03787749585686606, "alpha_c_m1d": -0.03648041214132258, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.006692748639030133, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015110611558032394, "ret_signed_p1d": 0.015110611558032394, "alpha_spy_p1d": 0.016487404343314083, "alpha_c_p1d": -0.0038311614697039698, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.018941773027736364, "ret_p1w": 0.09282244894011682, "ret_signed_p1w": 0.09282244894011682, "alpha_spy_p1w": 0.0810526075660547, "alpha_c_p1w": 0.06289450863050594, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.02992794030961088, "ret_p1m": 0.40313010357047396, "ret_signed_p1m": 0.40313010357047396, "alpha_spy_p1m": 0.3983548687412952, "alpha_c_p1m": 0.35893266600802076, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.0441974375624532, "ret_p3m": 0.32218030819686194, "ret_signed_p3m": 0.32218030819686194, "alpha_spy_p3m": 0.2765300765225811, "alpha_c_p3m": 0.1230390707308664, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.19914123746599555, "ret_p6m": 0.8481115065045073, "ret_signed_p6m": 0.8481115065045073, "alpha_spy_p6m": 0.8176171714903933, "alpha_c_p6m": 0.16574528661212407, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.6823662198923832, "price_path": [-0.3659, -0.3578, -0.36, -0.3794, -0.3389, -0.3162, -0.3287, -0.3373, -0.3551, -0.3567, -0.3664, -0.3184, -0.3411, -0.3535, -0.3761, -0.3815, -0.3869, -0.3821, -0.4047, -0.4015, -0.4004, -0.4021, -0.3988, -0.4091, -0.4253, -0.4172, -0.4172, -0.4193, -0.3929, -0.3961, -0.3162, -0.3044, -0.3448, -0.3519, -0.3546, -0.3654, -0.3605, -0.3448, -0.34, -0.3486, -0.3216, -0.3324, -0.3324, -0.3421, -0.3551, -0.3751, -0.3702, -0.3708, -0.3562, -0.3465, -0.3108, -0.2536, -0.2472, -0.2628, -0.2666, -0.2655, -0.2682, -0.2455, -0.2434, -0.1495, -0.088, -0.0766, -0.0432, 0.0, 0.0151, 0.0086, 0.0847, 0.0847, 0.0928, 0.1009, 0.1495, 0.1576, 0.1792, 0.2331, 0.1954, 0.2979, 0.3357, 0.3357, 0.3357, 0.3357, 0.3357, 0.3357, 0.3141, 0.4976, 0.4031, 0.4301, 0.4004, 0.4004, 0.4004, 0.4004, 0.3923, 0.3977, 0.4463, 0.3815, 0.395, 0.4895, 0.6622, 0.5866, 0.5893, 0.5704, 0.5084, 0.5057, 0.4895, 0.3141, 0.333, 0.3654, 0.4031, 0.3222, 0.4031, 0.3977, 0.3681, 0.3357, 0.2196, 0.2628, 0.3438, 0.3006, 0.3761, 0.3869, 0.3654, 0.3627, 0.3708, 0.3168, 0.3654, 0.3546, 0.3438, 0.3627, 0.3222, 0.3303, 0.2952, 0.2358, 0.3195, 0.3141, 0.2601, 0.3734, 0.3222, 0.2979, 0.2979, 0.3546, 0.3638, 0.3935, 0.3935, 0.3935, 0.4422, 0.4855, 0.564, 0.4855, 0.4882, 0.4612, 0.4449, 0.4882, 0.5613, 0.5423, 0.5126, 0.4585, 0.4071, 0.4043, 0.3854, 0.3989, 0.5288, 0.8102, 0.9618, 0.9888, 1.0429, 0.9672, 0.9996, 0.9509, 0.8481, 0.794, 0.8995, 0.8427, 0.8211, 0.8833, 0.7913, 0.7913, 0.7913, 0.7913, 0.8779, 0.886, 0.8319, 0.9266, 0.8752, 1.0105, 0.9591, 0.9591, 0.8427, 0.6343, 0.8941, 0.9563, 0.7291, 0.8481]}
{"unit_id": "orig:1964243357687042057", "ticker_idx": 0, "ts": "2025-09-06T08:24:45+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We reiterate our Buy rating and keep AVGO as our top pick… we lift our PO from $300 to $400, applying 37x FY26 P/E", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on AVGO, raises PO to $400; 4th mega customer (OpenAI), FY26 AI growth ~110%, EPS raised 21–35%.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) Broadcom: “Shock-and-Awe” results — accelerating AI pipeline, powerful operating leverage, Price Objective (PO) $400\n\n1. Accelerating AI pipeline; estimates raised 20–35%; EPS potential of $17–$20\nWe reiterate our Buy rating and keep AVGO as our top pick. We see a high likelihood that a fourth mega customer (reported by the press as OpenAI) joins Broadcom’s custom AI chip (XPU) business alongside the existing three (Google, Meta, ByteDance). With roughly $10B of incremental revenue in 2HFY26E, we lift our FY26 AI growth outlook from YoY 55–60% to about 110%. AVGO also indicated growth could further accelerate in FY27E on additional programs and new customers. Meanwhile, opex is expected to grow a modest 5–8%, creating substantial operating leverage and lining up with a period when AVGO deploys more cash to debt paydown. Net, we raise our FY26/FY27 EPS estimates by 21%/35% to $9.62/$13.36. Note: our FY27 assumes a more conservative AI growth of 50–65% YoY. If FY27 were to grow at a FY26-like 110%+ pace, EPS potential could approach $17–$20 on leverage. Reflecting faster EPS growth and accelerating profit, we lift our PO from $300 to $400, applying 37x FY26 P/E (prior 35x).\n\n2. As the AI pie expands, AVGO’s share rises\nWe agree AVGO is taking more share, but the overall AI market is also expanding toward the multi-trillion-dollar TAM cited by NVDA. Assuming AVGO and NVDA are the only two suppliers for now, we estimate AVGO’s CY25E share of computing + networking AI at ~11% ($23B out of $205B consolidated data-center revenue). Taking AVGO’s commentary at face value, AI revenue could approach $100B by CY27E, implying share could more than double to 24% ($100B of $418B). While such a scenario could pressure NVDA’s stock in the short term, we see greater risk that our NVDA CY27 data-center revenue forecast ($318B in our model) is actually conservative. AVGO’s custom silicon, networking, and scale are compelling, but they are (mostly) limited to internal workloads—not enterprise, not government, and not emerging-cloud facing—and exclude software or developer support. Over time, the evolution of “internal-use” vs “external-facing” markets could shift share between NVDA and AVGO.\n\n3. Risks: customer concentration; ASIC/networking competition\nWe note Google accounts for 30%+ of FY25E semiconductor revenue. Concentration risk is unavoidable in capital-intensive AI, but AVGO’s exposure to Google (15–20% of total FY25 revenue; 30%+ on semiconductor revenue alone, by our estimate) is notable. Unexpected order changes, new competitors, or content shifts (e.g., memory disaggregation) could raise risk. Second, NVDA’s improved networking stack is intensifying competition in networking, especially in clusters where NVDA’s rack-scale solutions are deployed.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031129663717725742, "ret_signed_m1d": -0.031129663717725742, "alpha_spy_m1d": -0.028679141619245563, "alpha_c_m1d": -0.020203014982361656, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025979983430521436, "ret_signed_p1d": -0.025979983430521436, "alpha_spy_p1d": -0.028291904128131318, "alpha_c_p1d": -0.0303304905991022, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.05334865965704916, "ret_signed_p1w": 0.05334865965704916, "alpha_spy_p1w": 0.03473053554589822, "alpha_c_p1w": 0.017431888348674818, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": -0.025064755049650667, "ret_signed_p1m": -0.025064755049650667, "alpha_spy_p1m": -0.05919561089954517, "alpha_c_p1m": -0.16175038093886418, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.1042466170647871, "ret_signed_p3m": 0.1042466170647871, "alpha_spy_p3m": 0.04651586067396751, "alpha_c_p3m": -0.11523275297268976, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": -0.08873355372693392, "ret_signed_p6m": -0.08873355372693392, "alpha_spy_p6m": -0.14329658980532434, "alpha_c_p6m": -0.41166909921848605, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.27, -0.2609, -0.2822, -0.2724, -0.2802, -0.2748, -0.2748, -0.2768, -0.2658, -0.2369, -0.2343, -0.2184, -0.2207, -0.2025, -0.2341, -0.2192, -0.2039, -0.2039, -0.2068, -0.2137, -0.196, -0.2032, -0.2062, -0.2027, -0.1872, -0.1876, -0.1713, -0.1803, -0.1662, -0.194, -0.1793, -0.1647, -0.1605, -0.1486, -0.1395, -0.1245, -0.1503, -0.1649, -0.1387, -0.1525, -0.1272, -0.1212, -0.1177, -0.1208, -0.095, -0.1058, -0.0996, -0.1137, -0.1154, -0.1468, -0.1576, -0.1622, -0.1494, -0.1488, -0.1378, -0.1313, -0.107, -0.1396, -0.1396, -0.1372, -0.1252, -0.1144, -0.0311, 0.0, -0.026, 0.0692, 0.0404, 0.0411, 0.0533, 0.0415, 0.0015, -0.0009, -0.0021, -0.0182, -0.0177, -0.0167, -0.026, -0.0305, -0.0497, -0.0439, -0.0338, -0.0199, -0.0194, -0.0277, -0.0251, 0.0013, -0.0001, -0.0592, 0.0337, -0.0027, 0.0182, 0.0263, 0.0124, 0.0121, -0.007, -0.0138, -0.0022, 0.0263, 0.0492, 0.0809, 0.1186, 0.091, 0.0712, 0.0507, 0.0199, 0.0403, 0.0305, 0.0127, 0.0386, 0.02, 0.0294, -0.0147, -0.0075, -0.007, -0.0132, 0.0271, 0.0051, -0.0141, 0.0953, 0.1158, 0.1522, 0.1522, 0.1678, 0.1189, 0.1058, 0.103, 0.1042, 0.1309, 0.1624, 0.1775, 0.1968, 0.1777, 0.0431, -0.0152, -0.0109, -0.0552, -0.044, -0.0136, -0.0086, 0.0143, 0.0169, 0.0169, 0.0224, 0.0145, 0.0158, 0.0049, 0.0049, 0.0094, -0.0028, -0.0018, -0.0026, -0.0346, 0.0017, 0.0227, 0.0296, -0.0131, -0.004, 0.0212, 0.0212, -0.0343, -0.0453, -0.0549, -0.0707, -0.0568, -0.0337, -0.0324, -0.0397, -0.038, -0.0386, -0.0699, -0.1055, -0.0984, -0.0333, -0.0013, -0.0115, -0.0048, -0.0384, -0.0558, -0.0558, -0.0344, -0.0316, -0.0302, -0.0341, -0.0408, -0.0549, -0.0351, -0.0659, -0.0722, -0.0743, -0.0887]}
{"unit_id": "quote:1959879155656618364", "ticker_idx": 0, "ts": "2025-08-25T07:22:58+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Excluding AMEC equipment from TSMC fabs is a positive for Lam Research.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long LRCX as beneficiary of TSMC excluding AMEC equipment from its fabs to qualify for US subsidies.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Excluding AMEC equipment from TSMC fabs is a positive for Lam Research.\n\n$LRCX https://t.co/nefzmJEALB", "reply_text": null, "parent_text": null, "quoted_text": "Interesting.\n\nNikkei reported that in order to receive U.S. subsidies, TSMC decided not to use Chinese equipment at its Taiwan fabs, rather than removing Chinese equipment from its U.S. fabs.\n\nThis will be a major blow to AMEC. https://t.co/OO9hpqRD4E", "parent_author": null, "ret_m1d": -0.011848275762884608, "ret_signed_m1d": -0.011848275762884608, "alpha_spy_m1d": -0.01626888863203102, "alpha_c_m1d": -0.010895591875209498, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023203031063404156, "ret_signed_p1d": 0.023203031063404156, "alpha_spy_p1d": 0.01901599471351334, "alpha_c_p1d": 0.013574643395739328, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.011157120035606094, "ret_signed_p1w": -0.011157120035606094, "alpha_spy_p1w": -0.015172949970439498, "alpha_c_p1w": 0.001193191838957497, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.3026263690556614, "ret_signed_p1m": 0.3026263690556614, "alpha_spy_p1m": 0.267482587762228, "alpha_c_p1m": 0.20811089947738792, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 0.3809800090150748, "ret_signed_p3m": 0.3809800090150748, "alpha_spy_p3m": 0.36250571589302183, "alpha_c_p3m": 0.27489683024420875, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 1.334457000705083, "ret_signed_p6m": 1.334457000705083, "alpha_spy_p6m": 1.2655097726155042, "alpha_c_p6m": 0.9436772559706224, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1727, -0.1711, -0.2043, -0.1876, -0.176, -0.1651, -0.1637, -0.1495, -0.1303, -0.1037, -0.1042, -0.0972, -0.1183, -0.08, -0.0874, -0.0893, -0.0893, -0.1065, -0.0955, -0.0558, -0.0519, -0.0438, -0.0403, -0.0389, -0.0441, -0.0242, -0.0244, -0.0244, -0.031, -0.0143, -0.0145, -0.0022, 0.0044, -0.0164, -0.0021, -0.009, -0.0048, -0.0061, 0.0045, -0.0354, -0.0413, -0.0346, -0.0427, -0.0263, -0.0231, -0.0216, -0.0636, -0.0485, -0.0283, -0.0454, -0.0527, -0.021, 0.0046, 0.0071, 0.0395, 0.0539, 0.0602, -0.0175, -0.0237, -0.0094, -0.021, -0.0283, -0.0118, 0.0, 0.0232, 0.0236, 0.0277, -0.0112, -0.0112, -0.042, -0.035, -0.0085, 0.0165, 0.0374, 0.0424, 0.06, 0.1412, 0.1548, 0.177, 0.1895, 0.2036, 0.2472, 0.2532, 0.3053, 0.3026, 0.2695, 0.2676, 0.2696, 0.2969, 0.3247, 0.4126, 0.4542, 0.4425, 0.4756, 0.3885, 0.4102, 0.3949, 0.2997, 0.3634, 0.3683, 0.4323, 0.4085, 0.4, 0.4251, 0.4349, 0.3974, 0.4596, 0.5006, 0.5522, 0.5396, 0.5895, 0.5929, 0.5578, 0.5952, 0.5411, 0.6329, 0.6046, 0.5765, 0.6459, 0.5748, 0.5969, 0.5168, 0.4668, 0.4588, 0.4171, 0.4721, 0.381, 0.4113, 0.4877, 0.5031, 0.5348, 0.5348, 0.5433, 0.5314, 0.565, 0.583, 0.5567, 0.5726, 0.6127, 0.6431, 0.6674, 0.6718, 0.5907, 0.6281, 0.6178, 0.5358, 0.6321, 0.7071, 0.7367, 0.7357, 0.7572, 0.7572, 0.7646, 0.7428, 0.7221, 0.6963, 0.6963, 0.8338, 0.93, 1.0509, 1.0124, 0.9914, 1.1638, 1.184, 1.1244, 1.069, 1.155, 1.2094, 1.2094, 1.204, 1.2632, 1.187, 1.1597, 1.2085, 1.363, 1.3741, 1.4592, 1.3134, 1.3535, 1.2802, 1.0788, 1.1138, 1.2892, 1.272, 1.2456, 1.3299, 1.2919, 1.334, 1.334, 1.3345]}
{"unit_id": "orig:1966667119967531389", "ticker_idx": 0, "ts": "2025-09-13T00:55:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA REQUESTED TSMC TO MOVE UP THE RUBIN PLATFORM SCHEDULE", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "NVIDIA pulling forward Rubin platform schedule at TSMC signals accelerating demand; bullish for both NVDA and TSM.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA REQUESTED TSMC TO MOVE UP THE RUBIN PLATFORM SCHEDULE\n\n$NVDA $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0003938101323475163, "ret_signed_m1d": 0.0003938101323475163, "alpha_spy_m1d": 0.00568947505583417, "alpha_c_m1d": 0.009672048145128898, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016146215426247723, "ret_signed_p1d": -0.016146215426247723, "alpha_spy_p1d": -0.014769422640966035, "alpha_c_p1d": -0.016276371269432932, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.03296766884920643, "ret_signed_p1w": 0.03296766884920643, "alpha_spy_p1w": 0.021197827475144315, "alpha_c_p1w": -0.015865189023303694, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.012826946331671918, "ret_signed_p1m": 0.012826946331671918, "alpha_spy_p1m": 0.008051711502493175, "alpha_c_p1m": -0.07289104103643451, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.017946907319828842, "ret_signed_p3m": 0.017946907319828842, "alpha_spy_p3m": -0.027703324354451997, "alpha_c_p3m": -0.1894951965008831, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 0.03955160419463177, "ret_signed_p6m": 0.03955160419463177, "alpha_spy_p6m": 0.009057269180517746, "alpha_c_p6m": -0.25799143020049486, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.1816, -0.1816, -0.1908, -0.189, -0.168, -0.1319, -0.1279, -0.1126, -0.1112, -0.1376, -0.1154, -0.1036, -0.1036, -0.1098, -0.0999, -0.0837, -0.0768, -0.0722, -0.077, -0.0397, -0.0359, -0.0268, -0.0301, -0.0359, -0.0604, -0.0393, -0.0226, -0.024, -0.0057, -0.0127, 0.0085, 0.0006, -0.0227, 0.0126, 0.0028, 0.0093, 0.0169, 0.0278, 0.0242, 0.0304, 0.0215, 0.024, 0.0151, 0.0239, -0.0119, -0.0133, -0.0156, 0.0013, 0.0115, 0.0226, 0.0216, 0.0136, -0.0201, -0.0201, -0.0393, -0.0402, -0.0343, -0.0604, -0.0532, -0.0394, -0.0024, -0.0033, 0.0004, 0.0, -0.0161, -0.042, -0.0085, -0.0061, 0.033, 0.0038, -0.0044, -0.0003, 0.0025, 0.0231, 0.0497, 0.0534, 0.0627, 0.0555, 0.0438, 0.041, 0.0639, 0.0834, 0.0304, 0.0595, 0.0128, 0.0117, 0.0228, 0.0308, 0.0275, 0.0192, 0.0142, 0.0248, 0.0479, 0.0773, 0.131, 0.1648, 0.1414, 0.1392, 0.1639, 0.1178, 0.0982, 0.0581, 0.0585, 0.1198, 0.0867, 0.0903, 0.0513, 0.0699, 0.0498, 0.0203, 0.0493, 0.0163, 0.0064, 0.027, 0.0004, 0.0141, 0.0141, -0.0042, 0.0122, 0.0209, 0.0104, 0.0317, 0.0263, 0.0439, 0.0407, 0.034, 0.0179, -0.0153, -0.0082, -0.0001, -0.0383, -0.0203, 0.0183, 0.0335, 0.0645, 0.0612, 0.0612, 0.072, 0.059, 0.0551, 0.0493, 0.0493, 0.0625, 0.0584, 0.0534, 0.064, 0.0411, 0.0401, 0.0405, 0.0454, 0.0304, 0.0524, 0.0478, 0.0478, 0.0019, 0.0314, 0.0399, 0.0559, 0.0491, 0.0606, 0.0775, 0.0831, 0.0753, 0.0443, 0.0146, -0.02, -0.033, 0.0432, 0.0692, 0.0608, 0.0693, 0.0518, 0.0285, 0.0285, 0.0407, 0.0576, 0.0572, 0.068, 0.0777, 0.085, 0.1003, 0.0402, -0.0031, 0.0267, 0.013, 0.0298, 0.0315, 0.0004, 0.0276, 0.0396]}
{"unit_id": "orig:1966667119967531389", "ticker_idx": 1, "ts": "2025-09-13T00:55:54+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA REQUESTED TSMC TO MOVE UP THE RUBIN PLATFORM SCHEDULE", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "NVIDIA pulling forward Rubin platform schedule at TSMC signals accelerating demand; bullish for both NVDA and TSM.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA REQUESTED TSMC TO MOVE UP THE RUBIN PLATFORM SCHEDULE\n\n$NVDA $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007843080909536448, "ret_signed_m1d": -0.007843080909536448, "alpha_spy_m1d": -0.0025474159860497947, "alpha_c_m1d": 0.0014351571032449328, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005764539631195165, "ret_signed_p1d": 0.005764539631195165, "alpha_spy_p1d": 0.007141332416476853, "alpha_c_p1d": 0.005634383788009956, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.046331366988609624, "ret_signed_p1w": 0.046331366988609624, "alpha_spy_p1w": 0.034561525614547506, "alpha_c_p1w": -0.0025014908839005034, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.13579331934291772, "ret_signed_p1m": 0.13579331934291772, "alpha_spy_p1m": 0.13101808451373897, "alpha_c_p1m": 0.05007533197481129, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.1731475879723252, "ret_signed_p3m": 0.1731475879723252, "alpha_spy_p3m": 0.12749735629804437, "alpha_c_p3m": -0.03429451584838672, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 0.3356988434058399, "ret_signed_p6m": 0.3356988434058399, "alpha_spy_p6m": 0.30520450839172586, "alpha_c_p6m": 0.03815580901071325, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.1832, -0.1832, -0.1984, -0.1953, -0.158, -0.1478, -0.143, -0.1255, -0.1335, -0.1404, -0.1063, -0.1017, -0.1017, -0.1232, -0.1282, -0.113, -0.121, -0.1185, -0.1251, -0.0935, -0.0911, -0.0604, -0.0803, -0.0862, -0.1025, -0.0805, -0.0757, -0.0604, -0.0713, -0.0767, -0.0707, -0.0756, -0.1001, -0.0856, -0.1106, -0.1148, -0.0718, -0.0748, -0.0738, -0.0654, -0.0763, -0.078, -0.0861, -0.0764, -0.1097, -0.1254, -0.1303, -0.1086, -0.0987, -0.0867, -0.0845, -0.0884, -0.1167, -0.1167, -0.1262, -0.1147, -0.1001, -0.0688, -0.0543, -0.04, -0.0036, -0.0094, -0.0078, 0.0, 0.0058, 0.0086, 0.031, 0.0165, 0.0463, 0.085, 0.0773, 0.0618, 0.0491, 0.0486, 0.0719, 0.1071, 0.1057, 0.1214, 0.1606, 0.1285, 0.1687, 0.1509, 0.0771, 0.1625, 0.1358, 0.1695, 0.1508, 0.1325, 0.1425, 0.1303, 0.1087, 0.1158, 0.132, 0.1447, 0.1572, 0.1709, 0.1637, 0.153, 0.17, 0.1285, 0.127, 0.1101, 0.0996, 0.1332, 0.1175, 0.1154, 0.0831, 0.0931, 0.0823, 0.0666, 0.0837, 0.065, 0.0557, 0.0924, 0.0926, 0.1128, 0.1128, 0.1188, 0.1041, 0.121, 0.1339, 0.1242, 0.1311, 0.1586, 0.1645, 0.1903, 0.1731, 0.1239, 0.1073, 0.104, 0.0658, 0.0955, 0.112, 0.1286, 0.1427, 0.1499, 0.1499, 0.1654, 0.158, 0.1529, 0.1695, 0.1695, 0.2299, 0.2401, 0.26, 0.2264, 0.2238, 0.2454, 0.2767, 0.2746, 0.2588, 0.3147, 0.3177, 0.3177, 0.259, 0.255, 0.2598, 0.2887, 0.2804, 0.302, 0.3173, 0.3067, 0.2721, 0.3136, 0.2921, 0.2535, 0.2727, 0.3425, 0.3677, 0.3927, 0.4396, 0.4166, 0.4099, 0.4099, 0.4015, 0.3941, 0.3869, 0.4259, 0.424, 0.4845, 0.4921, 0.4501, 0.4415, 0.4204, 0.3589, 0.3755, 0.3618, 0.3041, 0.3419, 0.3357]}
{"unit_id": "orig:1971343442774196241", "ticker_idx": 0, "ts": "2025-09-25T22:37:57+00:00", "symbol": "CDNS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long $CDNS", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long CDNS in context of TSMC leveraging AI to improve chip energy efficiency tenfold.", "resolved_tickers": ["CDNS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Application", "unit_kind": "original", "lang": "en", "tweet_text": "At the Open Innovation Platform Ecosystem Forum held in Silicon Valley, TSMC announced that by leveraging AI technology in semiconductor design, it has improved chip energy efficiency by about tenfold. (Economic Daily News, Taiwan)\n\nhttps://t.co/qnBix5mwf3\n\nLong $CDNS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01680622667781928, "ret_signed_m1d": 0.01680622667781928, "alpha_spy_m1d": 0.012171346179810305, "alpha_c_m1d": 0.012171346179810305, "bench_spy_m1d": 0.004634880498008975, "bench_c_m1d": 0.004634880498008975, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0027061249177861013, "ret_signed_p1d": -0.0027061249177861013, "alpha_spy_p1d": -0.008435269284780333, "alpha_c_p1d": -0.008435269284780333, "bench_spy_p1d": 0.005729144366994232, "bench_c_p1d": 0.005729144366994232, "ret_p1w": -0.010881351765466118, "ret_signed_p1w": -0.010881351765466118, "alpha_spy_p1w": -0.02785564080877012, "alpha_c_p1w": -0.02785564080877012, "bench_spy_p1w": 0.016974289043304003, "bench_c_p1w": 0.016974289043304003, "ret_p1m": -0.016977130680015384, "ret_signed_p1m": -0.016977130680015384, "alpha_spy_p1m": -0.04615426682421975, "alpha_c_p1m": -0.04615426682421975, "bench_spy_p1m": 0.029177136144204363, "bench_c_p1m": 0.029177136144204363, "ret_p3m": -0.09721986304382635, "ret_signed_p3m": -0.09721986304382635, "alpha_spy_p3m": -0.1457615799066181, "alpha_c_p3m": -0.1457615799066181, "bench_spy_p3m": 0.04854171686279174, "bench_c_p3m": 0.04854171686279174, "ret_p6m": -0.19130634116437606, "ret_signed_p6m": -0.19130634116437606, "alpha_spy_p6m": -0.1825121199521862, "alpha_c_p6m": -0.1825121199521862, "bench_spy_p6m": -0.008794221212189868, "bench_c_p6m": -0.008794221212189868, "price_path": [-0.1222, -0.1185, -0.1143, -0.0691, -0.0691, -0.0876, -0.0791, -0.0802, -0.0809, -0.0868, -0.0956, -0.0941, -0.1039, -0.086, -0.1011, -0.099, -0.0896, -0.0701, -0.0753, -0.0538, -0.0493, 0.0433, 0.0569, 0.0385, 0.0168, 0.0394, 0.0269, 0.0259, 0.0093, 0.0028, -0.0056, 0.0073, -0.0055, -0.0061, -0.0034, 0.0148, -0.0119, -0.016, -0.0099, -0.0041, -0.0167, -0.02, -0.0119, 0.0092, -0.0018, -0.0018, -0.0235, -0.0107, -0.0049, -0.0001, 0.0273, 0.0305, -0.0357, 0.0104, -0.0216, 0.0013, -0.0058, -0.0108, 0.0394, 0.0635, 0.0636, 0.0434, 0.0168, 0.0, -0.0027, -0.0072, 0.0006, 0.0026, -0.0109, -0.0108, 0.0066, -0.0159, -0.003, -0.0068, -0.0685, -0.0536, -0.0721, -0.0768, -0.0764, -0.071, -0.061, -0.0502, -0.0585, -0.0393, -0.017, 0.001, -0.0278, -0.0274, -0.0426, -0.0352, -0.0446, -0.0508, -0.0667, -0.0758, -0.0741, -0.063, -0.0927, -0.1012, -0.0999, -0.1029, -0.1133, -0.1363, -0.1262, -0.1393, -0.1438, -0.1327, -0.135, -0.1274, -0.1274, -0.1117, -0.118, -0.0943, -0.0426, -0.0392, -0.0385, -0.0392, -0.0455, -0.037, -0.0445, -0.0793, -0.0929, -0.0898, -0.1084, -0.1024, -0.103, -0.0954, -0.0972, -0.0949, -0.0949, -0.0916, -0.095, -0.101, -0.1096, -0.1096, -0.1158, -0.142, -0.1037, -0.0869, -0.0919, -0.0677, -0.0728, -0.0798, -0.1079, -0.0868, -0.0957, -0.0957, -0.1255, -0.106, -0.0968, -0.0933, -0.0827, -0.0931, -0.0871, -0.1378, -0.1558, -0.1762, -0.2352, -0.2269, -0.2305, -0.1924, -0.1711, -0.1483, -0.1464, -0.1787, -0.147, -0.147, -0.1926, -0.1312, -0.1552, -0.156, -0.203, -0.1721, -0.1402, -0.1523, -0.1415, -0.1359, -0.1437, -0.13, -0.1459, -0.1542, -0.151, -0.1646, -0.1621, -0.173, -0.1824, -0.1662, -0.1632, -0.175, -0.1813, -0.1913]}
{"unit_id": "orig:1963410090532491617", "ticker_idx": 0, "ts": "2025-09-04T01:13:38+00:00", "symbol": "Elite Material", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EMC should benefit the most; re-iterate Buy - Goldman Sachs", "tweet_type": "news_reaction", "timeline": "years", "summary": "Goldman reiterates Buy on Elite Material; expects EMC to capture 40–45% of AI GPU CCL TAM in Rubin generation with 100% Switch Tray share, dismissing market-share loss fears.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Elite Material Co. 2383.TW Taiwan CCL", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "250904_Elite Material: Rubin CCL upgrade is in-line with our expectation, where EMC should benefit the most; re-iterate Buy - Goldman Sachs\n\n(1) According to our recent channel checks, Vera Rubin’s Computing/Switch Tray has been confirmed as follows:\n■ Computing Tray: M6/M8.5 CCL + HVLP-4 copper foil\n■ Switch Tray: M9-class CCL\n\n(2) We maintain our Rubin CCL TAM forecast at $275mn in 2026 and $2bn in 2027.\n\n(3) Here, EMC is expected to capture 40–45% market share of AI GPU CCL TAM.\n■ Computing Tray: 15% in M8.5\n■ Switch Tray: 100% in M9\n\n(4) This figure is even higher than EMC’s GB300 market share of 35% [AI rack-level CCL revenue].\n\n(5) In addition, another high-power consumption version of AI GPU appears to be at the sampling stage within the supply chain, which could lead to additional CCL upgrade demand in the mid-to-long term.\n\n(6) Furthermore, Rubin’s H/W design is still only in the early stages, and based on past cases, the final design may change after rack-level testing.\n\n(7) The CCL specifications are expected to be finalized around June–July 2026.\n\n(8) Recently, there has been investor concern that EMC might lose market share in the Rubin generation compared with the Blackwell generation.\n\n(9) This is attributed to the entry of mainland Chinese companies into the supply chain.\n\n(10) However, we judge the downside risk to EMC’s market share to be limited. In particular, EMC secured 70% market share in Switch Tray for GB300, and we expect it to recover to 100% in the Rubin generation.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04484304932735417, "ret_signed_m1d": 0.04484304932735417, "alpha_spy_m1d": 0.0531311431814544, "alpha_c_m1d": 0.04975481063962428, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.004911761312270113, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04484304932735417, "ret_signed_p1d": 0.04484304932735417, "alpha_spy_p1d": 0.04773919781588931, "alpha_c_p1d": 0.04400550308235318, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.0008375462450009863, "ret_p1w": 0.188340807174888, "ret_signed_p1w": 0.188340807174888, "alpha_spy_p1w": 0.17523057203706438, "alpha_c_p1w": 0.15628176757291556, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.03205903960197243, "ret_p1m": 0.17040358744394624, "ret_signed_p1m": 0.17040358744394624, "alpha_spy_p1m": 0.13659559122177867, "alpha_c_p1m": 0.08494097060134509, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.08546261684260115, "ret_p3m": 0.29147982062780264, "ret_signed_p3m": 0.29147982062780264, "alpha_spy_p3m": 0.23863966636631484, "alpha_c_p3m": 0.18855648206743725, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.1029233385603654, "ret_p6m": 1.188340807174888, "ret_signed_p6m": 1.188340807174888, "alpha_spy_p6m": 1.1254794405364477, "alpha_c_p6m": 1.1287226577962577, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.0596181493786303, "price_path": [-0.2741, -0.2121, -0.2272, -0.2405, -0.2387, -0.2219, -0.2325, -0.252, -0.252, -0.2582, -0.2617, -0.2387, -0.2405, -0.2431, -0.2334, -0.2192, -0.2033, -0.2006, -0.1785, -0.1829, -0.2024, -0.1564, -0.1528, -0.1528, -0.167, -0.1785, -0.1413, -0.1298, -0.1387, -0.0926, -0.1015, -0.1236, -0.1254, -0.1245, -0.1201, -0.0926, -0.1015, -0.1103, -0.0218, -0.0218, -0.0174, -0.0041, -0.0174, 0.0136, 0.018, 0.0225, 0.0402, 0.0623, 0.0711, 0.1331, 0.1552, 0.0889, 0.018, 0.0933, 0.0667, 0.1066, 0.1066, 0.1331, 0.1211, 0.0942, 0.0359, 0.0314, 0.0448, 0.0, 0.0448, 0.0269, 0.0987, 0.1256, 0.1883, 0.2063, 0.1794, 0.1525, 0.1794, 0.1973, 0.157, 0.1659, 0.2063, 0.1614, 0.1256, 0.0987, 0.0987, 0.0987, 0.1031, 0.1614, 0.1704, 0.1704, 0.1076, 0.0717, 0.1076, 0.1076, 0.0762, 0.0493, 0.0224, 0.0224, -0.009, 0.009, 0.0135, 0.0045, 0.0, 0.0, 0.0448, 0.1256, 0.1211, 0.1435, 0.2197, 0.2646, 0.3184, 0.2915, 0.3004, 0.2287, 0.2197, 0.1839, 0.2063, 0.2332, 0.2377, 0.3004, 0.2108, 0.1928, 0.2825, 0.1749, 0.1839, 0.3004, 0.3184, 0.3408, 0.3677, 0.3229, 0.2915, 0.2825, 0.278, 0.3408, 0.3318, 0.3453, 0.435, 0.4305, 0.4395, 0.4395, 0.3408, 0.3408, 0.3453, 0.3901, 0.435, 0.4484, 0.5157, 0.5157, 0.4709, 0.5022, 0.4753, 0.4753, 0.4753, 0.426, 0.4529, 0.5202, 0.417, 0.3857, 0.4036, 0.4439, 0.4664, 0.4978, 0.4081, 0.4574, 0.417, 0.4484, 0.3991, 0.4888, 0.6143, 0.5605, 0.6323, 0.704, 0.6054, 0.565, 0.6009, 0.6637, 0.8296, 0.7489, 0.7265, 0.8655, 0.8879, 0.9686, 0.9686, 0.9686, 0.9686, 0.9686, 0.9686, 0.9686, 0.9686, 0.9507, 1.0717, 1.1704, 1.1883, 1.1883]}
{"unit_id": "thread:1959595231202161024", "ticker_idx": 0, "ts": "2025-08-24T12:34:45+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix—which outsourced the base die production to TSMC, raising costs by as much as 30%—would inevitably suffer a blow to its profitability", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish SK Hynix: HBM4 price hike likely only ~10% vs 20–40% expected, while TSMC outsourcing raises costs 30%, crushing profitability.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "According to reports from the Korean media, the price increase for HBM4 may be limited to around 10%.\n\nThe industry had been expecting an increase of 20–40%.\n\nIf the price hike really ends up being only 10%, SK Hynix—which outsourced the base die production to TSMC, raising costs by as much as 30%—would inevitably suffer a blow to its profitability.\n\n---\n\nhttps://t.co/L4RrJWrwso", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032755225139682875, "ret_signed_m1d": 0.032755225139682875, "alpha_spy_m1d": 0.03717583800882929, "alpha_c_m1d": 0.031802541252007765, "bench_spy_m1d": 0.004420612869146412, "bench_c_m1d": -0.0009526838876751098, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007707168704449874, "ret_signed_p1d": -0.007707168704449874, "alpha_spy_p1d": -0.003520132354559058, "alpha_c_p1d": 0.0019212189632149546, "bench_spy_p1d": 0.004187036349890816, "bench_c_p1d": 0.009628387667664828, "ret_p1w": -0.01206254011757013, "ret_signed_p1w": 0.01206254011757013, "alpha_spy_p1w": 0.016078370052403534, "alpha_c_p1w": -0.00028777175699346014, "bench_spy_p1w": 0.004015829934833404, "bench_c_p1w": -0.012350311874563591, "ret_p1m": 0.39314614138507986, "ret_signed_p1m": -0.39314614138507986, "alpha_spy_p1m": -0.35800236009164643, "alpha_c_p1m": -0.29863067180680636, "bench_spy_p1m": 0.03514378129343343, "bench_c_p1m": 0.0945154695782735, "ret_p3m": 1.2035636253174942, "ret_signed_p3m": -1.2035636253174942, "alpha_spy_p3m": -1.1850893321954412, "alpha_c_p3m": -1.0974804465466281, "bench_spy_p3m": 0.018474293122052954, "bench_c_p3m": 0.10608317877086604, "ret_p6m": 2.39846717626965, "ret_signed_p6m": -2.39846717626965, "alpha_spy_p6m": -2.3295199481800712, "alpha_c_p6m": -2.0076874315351896, "bench_spy_p6m": 0.06894722808957887, "bench_c_p6m": 0.39077974473446075, "price_path": [-0.1999, -0.183, -0.2119, -0.2004, -0.2004, -0.1618, -0.1349, -0.1349, -0.1175, -0.1118, -0.0751, -0.0925, -0.0925, -0.0443, -0.0405, -0.0501, -0.052, -0.0096, 0.0, 0.0732, 0.1021, 0.1291, 0.0944, 0.1252, 0.1002, 0.0751, 0.0732, 0.0424, 0.0443, 0.0867, 0.0829, 0.1445, 0.1349, 0.1561, 0.1503, 0.1407, 0.0385, 0.0366, 0.0501, 0.0347, 0.0366, 0.0385, 0.025, 0.0096, 0.0116, 0.0154, 0.0539, -0.0058, -0.0058, 0.0154, -0.0039, 0.0096, -0.0116, 0.0289, 0.0366, 0.0713, 0.0655, 0.0655, 0.0308, 0.0135, -0.0154, -0.0559, -0.0328, 0.0, 0.0077, 0.0019, 0.0362, 0.0381, -0.0121, 0.0053, 0.013, 0.0246, 0.0555, 0.069, 0.1114, 0.1732, 0.1848, 0.2677, 0.2774, 0.343, 0.287, 0.3719, 0.3719, 0.3546, 0.3931, 0.3796, 0.3758, 0.2986, 0.3468, 0.341, 0.3893, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.5263, 0.6517, 0.6015, 0.588, 0.6305, 0.7463, 0.7964, 0.8736, 0.8485, 0.8582, 0.8466, 0.9682, 1.0646, 1.0106, 1.1534, 1.192, 1.1573, 1.3927, 1.2615, 1.2344, 1.2885, 1.2383, 1.3386, 1.3888, 1.3811, 1.3618, 1.1611, 1.3386, 1.1997, 1.1688, 1.2036, 1.0106, 1.0067, 1.0029, 1.0222, 1.1009, 1.0468, 1.0777, 1.1549, 1.1318, 1.0931, 1.1009, 1.2283, 1.1858, 1.2669, 1.182, 1.2051, 1.1395, 1.0468, 1.1279, 1.1318, 1.1125, 1.2399, 1.2553, 1.2708, 1.2708, 1.3133, 1.4716, 1.5141, 1.5141, 1.5141, 1.6145, 1.6879, 1.8037, 1.8655, 1.9196, 1.8732, 1.8926, 1.8501, 1.8655, 1.8926, 1.9196, 1.9505, 1.8694, 1.8578, 1.9157, 1.9621, 1.8424, 2.0895, 2.2479, 2.3251, 2.5105, 2.2054, 2.5027, 2.4757, 2.2517, 2.2401, 2.4255, 2.383, 2.3212, 2.4294, 2.3985, 2.3985, 2.3985]}
{"unit_id": "orig:1951161324299530632", "ticker_idx": 0, "ts": "2025-08-01T06:01:25+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura: Raises Samsung's target price; Samsung likely to supply HBM4 to Nvidia in Q2 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura raises Samsung target price; HBM4 Nvidia supply win in Q2 2026 is the bullish catalyst.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura: Raises Samsung’s target price; Samsung likely to supply HBM4 to Nvidia in Q2 2026. https://t.co/e3zlMoX0NK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03628451801845611, "ret_signed_m1d": 0.03628451801845611, "alpha_spy_m1d": 0.019620828077359542, "alpha_c_m1d": 0.014228266800442535, "bench_spy_m1d": 0.01666368994109657, "bench_c_m1d": 0.02205625121801358, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01161109622749179, "ret_signed_p1d": 0.01161109622749179, "alpha_spy_p1d": -0.0035887179614699782, "alpha_c_p1d": -0.008694791299870719, "bench_spy_p1d": 0.015199814188961769, "bench_c_p1d": 0.02030588752736251, "ret_p1w": 0.042090123474913144, "ret_signed_p1w": 0.042090123474913144, "alpha_spy_p1w": 0.017223541503019302, "alpha_c_p1w": 0.007663562763754239, "bench_spy_p1w": 0.024866581971893842, "bench_c_p1w": 0.034426560711158904, "ret_p1m": -0.01886781633450707, "ret_signed_p1m": -0.01886781633450707, "alpha_spy_p1m": -0.05639287429155182, "alpha_c_p1m": -0.039796134858980814, "bench_spy_p1m": 0.03752505795704475, "bench_c_p1m": 0.020928318524473744, "ret_p3m": 0.4651436595476073, "ret_signed_p3m": 0.4651436595476073, "alpha_spy_p3m": 0.3564519244954558, "alpha_c_p3m": 0.28056009055783004, "bench_spy_p3m": 0.10869173505215146, "bench_c_p3m": 0.18458356898977724, "ret_p6m": 1.2281755934741243, "ret_signed_p6m": 1.2281755934741243, "alpha_spy_p6m": 1.1075694978265855, "alpha_c_p6m": 1.0884890617390235, "bench_spy_p6m": 0.12060609564753877, "bench_c_p6m": 0.13968653173510082, "price_path": [-0.2167, -0.2124, -0.2124, -0.2095, -0.1691, -0.1792, -0.172, -0.1734, -0.1806, -0.1951, -0.1936, -0.1965, -0.2109, -0.2181, -0.2109, -0.2225, -0.1936, -0.1907, -0.1893, -0.1806, -0.1806, -0.1662, -0.1475, -0.1475, -0.1374, -0.146, -0.1359, -0.1417, -0.159, -0.1749, -0.1619, -0.1374, -0.146, -0.1417, -0.1633, -0.1273, -0.1157, -0.1316, -0.1176, -0.1321, -0.1263, -0.1176, -0.074, -0.0813, -0.1045, -0.1089, -0.1234, -0.1147, -0.0914, -0.0929, -0.0755, -0.061, -0.0319, -0.0261, -0.016, -0.0421, -0.0363, -0.0421, -0.0435, 0.0218, 0.0247, 0.0537, 0.0363, 0.0, 0.0116, 0.0145, -0.0015, 0.0232, 0.0421, 0.0305, 0.0319, 0.0435, 0.0392, 0.0392, 0.016, 0.016, 0.0232, 0.0247, 0.0363, 0.0377, 0.0203, 0.0247, 0.0102, 0.0116, -0.0189, 0.0029, 0.0131, 0.0174, 0.0087, 0.0174, 0.0377, 0.0537, 0.0653, 0.0943, 0.1103, 0.1524, 0.135, 0.1655, 0.1655, 0.2119, 0.2293, 0.2395, 0.2496, 0.209, 0.2275, 0.2231, 0.2538, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3084, 0.3762, 0.3602, 0.3354, 0.385, 0.4243, 0.4272, 0.4302, 0.4214, 0.4374, 0.4068, 0.4404, 0.487, 0.4506, 0.4651, 0.5176, 0.5672, 0.6197, 0.5293, 0.4666, 0.4462, 0.4272, 0.4666, 0.5089, 0.503, 0.4987, 0.417, 0.4666, 0.4258, 0.4068, 0.4666, 0.382, 0.4097, 0.4476, 0.4987, 0.5089, 0.4651, 0.4695, 0.5074, 0.5235, 0.5322, 0.5803, 0.5964, 0.5803, 0.5745, 0.5643, 0.5876, 0.5278, 0.4987, 0.573, 0.5687, 0.5497, 0.6109, 0.6255, 0.6197, 0.6197, 0.7057, 0.7506, 0.7565, 0.7565, 0.7565, 0.8824, 1.0231, 1.0348, 1.0656, 1.0333, 1.0363, 1.0333, 1.0158, 1.0553, 1.1081, 1.1813, 1.1872, 1.1271, 1.1901, 1.2311, 1.2282, 1.2282]}
{"unit_id": "thread:1970512635105026198", "ticker_idx": 0, "ts": "2025-09-23T15:36:37+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Seeing statements in the media like \"I don't understand why Micron is the cheapest stock in the Philadelphia Semiconductor Index in terms of PER\" makes me feel uneasy.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Micron: 'cheap on PER' narrative signals memory cycle is near peak, not a buying opportunity.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "One piece of wisdom my seniors taught me is this: when memory stocks are described as “cheap on a PER basis,” it usually means the memory cycle is right near its peak.\n\nSeeing statements in the media like “I don’t understand why Micron is the cheapest stock in the Philadelphia Semiconductor Index in terms of PER” makes me feel uneasy.\n\n---\n\nhttps://t.co/ftcyjck5YL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010756522195869866, "ret_signed_m1d": 0.010756522195869866, "alpha_spy_m1d": 0.016229790634715746, "alpha_c_m1d": 0.012217492831149812, "bench_spy_m1d": 0.00547326843884588, "bench_c_m1d": 0.0014609706352799456, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02824341205600267, "ret_signed_p1d": 0.02824341205600267, "alpha_spy_p1d": 0.025061778603388984, "alpha_c_p1d": 0.02640932438425203, "bench_spy_p1d": -0.003181633452613686, "bench_c_p1d": -0.0018340876717506394, "ret_p1w": 0.0054684811190846006, "ret_signed_p1w": -0.0054684811190846006, "alpha_spy_p1w": -0.0009903523963925576, "alpha_c_p1w": 0.009016948046868034, "bench_spy_p1w": 0.004478128722692043, "bench_c_p1w": 0.014485429165952635, "ret_p1m": 0.19340362628470564, "ret_signed_p1m": -0.19340362628470564, "alpha_spy_p1m": -0.18648281458315097, "alpha_c_p1m": -0.1408702510561941, "bench_spy_p1m": 0.006920811701554674, "bench_c_p1m": 0.052533375228511536, "ret_p3m": 0.5989816019108829, "ret_signed_p3m": -0.5989816019108829, "alpha_spy_p3m": -0.5697434492981737, "alpha_c_p3m": -0.49164560331085694, "bench_spy_p3m": 0.029238152612709234, "bench_c_p3m": 0.10733599860002596, "ret_p6m": 1.7775123773604586, "ret_signed_p6m": -1.7775123773604586, "alpha_spy_p6m": -1.777249431682111, "alpha_c_p6m": -1.5499869217099527, "bench_spy_p6m": 0.00026294567834761473, "bench_c_p6m": 0.22752545565050597, "price_path": [-0.2435, -0.251, -0.2601, -0.2742, -0.2691, -0.2658, -0.2658, -0.2794, -0.2523, -0.2654, -0.2602, -0.2517, -0.2872, -0.2782, -0.3003, -0.3194, -0.3126, -0.3196, -0.3437, -0.34, -0.3286, -0.3314, -0.3315, -0.3272, -0.3105, -0.3442, -0.3697, -0.3524, -0.3446, -0.3463, -0.3277, -0.2856, -0.2565, -0.2323, -0.2532, -0.2471, -0.2737, -0.2576, -0.2666, -0.2957, -0.3042, -0.2928, -0.3004, -0.2999, -0.2924, -0.2669, -0.2848, -0.2848, -0.288, -0.2866, -0.2536, -0.2106, -0.21, -0.1873, -0.1587, -0.0952, -0.0552, -0.0519, -0.0456, -0.0386, 0.0149, -0.0221, -0.0108, 0.0, -0.0282, -0.0576, -0.0549, -0.0151, 0.0055, 0.0946, 0.1042, 0.1294, 0.1482, 0.1166, 0.1818, 0.1565, 0.092, 0.1591, 0.1248, 0.1541, 0.2178, 0.2169, 0.2433, 0.2164, 0.1934, 0.243, 0.317, 0.3235, 0.3343, 0.3627, 0.347, 0.3455, 0.4113, 0.311, 0.4281, 0.4331, 0.4306, 0.5231, 0.4498, 0.4726, 0.4248, 0.4842, 0.4548, 0.374, 0.3585, 0.2108, 0.2469, 0.3465, 0.3501, 0.3846, 0.3846, 0.422, 0.4459, 0.4401, 0.408, 0.3629, 0.4264, 0.4847, 0.5178, 0.5857, 0.5541, 0.45, 0.4281, 0.3981, 0.3561, 0.4945, 0.599, 0.6631, 0.6612, 0.7238, 0.7238, 0.7124, 0.7708, 0.7603, 0.7169, 0.7169, 0.8974, 0.8777, 1.0659, 1.0425, 0.9672, 1.0759, 1.0806, 1.034, 1.0052, 1.025, 1.1821, 1.1821, 1.1956, 1.3407, 1.3916, 1.4041, 1.3406, 1.4678, 1.6184, 1.6215, 1.4957, 1.6336, 1.5231, 1.2823, 1.3033, 1.3742, 1.3069, 1.2453, 1.4684, 1.4902, 1.4763, 1.4763, 1.4049, 1.5322, 1.5105, 1.5756, 1.5323, 1.5145, 1.5806, 1.4998, 1.4806, 1.4824, 1.2839, 1.4108, 1.3884, 1.2275, 1.3419, 1.4249, 1.5186, 1.4384, 1.5634, 1.6576, 1.7773, 1.7775]}
{"unit_id": "orig:1970254513195098319", "ticker_idx": 0, "ts": "2025-09-22T22:30:56+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rating upgraded (EW → OW), CY26 EPS revised from $9.58 → $10.45, target price $209. With DRAM exposure at 28%, the highest among peers, AMAT is set to benefit most from the memory cycle recovery.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises 2026 WFE outlook to +10%/$128.3B; upgrades AMAT to OW (PT $209) and LRCX to EW (PT $125) on memory/NAND recovery.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "2026 Semiconductor Equipment Market Outlook Raised (Morgan Stanley)\n\n1. Market Outlook\n    • The 2026 WFE (Wafer Fab Equipment) market forecast has been revised upward from +5% to +10% growth, with the market size now estimated at $128.3B.\n    • Memory is the key driver of the upward revision: DRAM WFE projected at $34.9B (+18%), NAND WFE at $13.8B (+35%).\n    • For DRAM, Samsung and SK Hynix’s greenfield investments and strong DDR5/HBM demand are the main catalysts, while NAND is supported by surging eSSD demand (+50% in 2026).\n    • Memory supply-demand improvement creates upside risk for memory Capex in 2026/27.\n\n2. Company-Level Changes\n\n1) Applied Materials (AMAT)\n    • Rating upgraded (EW → OW), CY26 EPS revised from $9.58 → $10.45, target price $209.\n    • With DRAM exposure at 28%, the highest among peers, AMAT is set to benefit most from the memory cycle recovery.\n    • Risks related to China and ICAPS (power semiconductors/displays) largely resolved.\n    • Currently trading at a 25% valuation discount vs. LAM, but the target price reflects only a 15% discount → room for re-rating.\n\n2) Lam Research (LRCX)\n    • Rating upgraded (UW → EW), CY26 EPS revised from $5.12 → $5.43, target price $125.\n    • Supported by NAND recovery and non-China logic growth (+16%), WFE is expected to outperform for a third consecutive year.\n    • However, with current valuation at 23x already reflecting optimism, further upgrades would require revenue above $24B and EPS above $6.\n\n3. Regional WFE Outlook\n    • Korea: $28.2B (+20%), strong growth led by Samsung and SK Hynix.\n    • Taiwan: $28.2B (+15%), sustained growth driven by TSMC.\n    • China: $39.8B, flat (0% growth) due to U.S. export restrictions.\n    • Japan: $11.6B (+10%).\n    • North America & Europe: $12.8B and $3.8B respectively, modest growth.\n\n$AMAT $LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05196488656359266, "ret_signed_m1d": -0.05196488656359266, "alpha_spy_m1d": -0.0472561556390203, "alpha_c_m1d": -0.03191332880864495, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0017453224875789708, "ret_signed_p1d": 0.0017453224875789708, "alpha_spy_p1d": 0.007188797327386531, "alpha_c_p1d": 0.0032041618014598816, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.022092462345511876, "ret_signed_p1w": 0.022092462345511876, "alpha_spy_p1w": 0.02683120160338004, "alpha_c_p1w": 0.020571575088430105, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.127069551261644, "ret_signed_p1m": 0.127069551261644, "alpha_spy_p1m": 0.12039637910742584, "alpha_c_p1m": 0.05527523496726494, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.26669106989399727, "ret_signed_p3m": 0.26669106989399727, "alpha_spy_p3m": 0.2522498816707768, "alpha_c_p3m": 0.1888439909933386, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 0.763371754530642, "ret_signed_p6m": 0.763371754530642, "alpha_spy_p6m": 0.7544759406102917, "alpha_c_p6m": 0.5276424013875765, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.0896, -0.0874, -0.0889, -0.0896, -0.0862, -0.0551, -0.05, -0.05, -0.0513, -0.0304, -0.0284, -0.0152, -0.0157, -0.0199, -0.009, -0.0313, -0.0426, -0.053, -0.0422, -0.0694, -0.07, -0.0645, -0.0766, -0.0538, -0.0631, -0.0582, -0.1046, -0.1049, -0.0909, -0.1091, -0.1141, -0.0892, -0.0807, -0.0831, -0.0629, -0.055, -0.0639, -0.1956, -0.1868, -0.1933, -0.1996, -0.2029, -0.1897, -0.1922, -0.1796, -0.1802, -0.1758, -0.1983, -0.1983, -0.2142, -0.2208, -0.2109, -0.1884, -0.1919, -0.1846, -0.185, -0.1515, -0.1632, -0.1476, -0.1346, -0.1117, -0.0537, -0.052, 0.0, 0.0017, 0.0046, -0.0046, 0.017, 0.0221, 0.021, 0.0859, 0.1151, 0.0848, 0.1166, 0.0551, 0.0847, 0.0986, 0.047, 0.0946, 0.0881, 0.1349, 0.1356, 0.122, 0.1377, 0.1271, 0.0999, 0.1394, 0.1408, 0.1537, 0.1352, 0.1757, 0.1597, 0.1625, 0.1855, 0.148, 0.2013, 0.1646, 0.1474, 0.1724, 0.1404, 0.1507, 0.1133, 0.1271, 0.1406, 0.1227, 0.1726, 0.1004, 0.1193, 0.1538, 0.2115, 0.2491, 0.2491, 0.2604, 0.2729, 0.3258, 0.3423, 0.3463, 0.3391, 0.3399, 0.3348, 0.3749, 0.3497, 0.2952, 0.3055, 0.2934, 0.2406, 0.2667, 0.2812, 0.2942, 0.3003, 0.3031, 0.3031, 0.3087, 0.3144, 0.299, 0.2841, 0.2841, 0.3435, 0.4207, 0.4791, 0.4601, 0.4073, 0.5049, 0.5352, 0.5234, 0.5085, 0.5944, 0.634, 0.634, 0.5901, 0.6252, 0.5929, 0.6109, 0.5963, 0.6625, 0.6827, 0.7056, 0.6106, 0.641, 0.5923, 0.4871, 0.519, 0.6115, 0.6518, 0.6443, 0.6983, 0.6409, 0.7734, 0.7734, 0.7945, 0.8453, 0.8503, 0.878, 0.8689, 0.8908, 0.976, 0.8797, 0.8626, 0.862, 0.7577, 0.7899, 0.7337, 0.6247, 0.6957, 0.7305, 0.7564, 0.6874, 0.7087, 0.732, 0.7634]}
{"unit_id": "orig:1970254513195098319", "ticker_idx": 1, "ts": "2025-09-22T22:30:56+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rating upgraded (UW → EW), CY26 EPS revised from $5.12 → $5.43, target price $125.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises 2026 WFE outlook to +10%/$128.3B; upgrades AMAT to OW (PT $209) and LRCX to EW (PT $125) on memory/NAND recovery.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "2026 Semiconductor Equipment Market Outlook Raised (Morgan Stanley)\n\n1. Market Outlook\n    • The 2026 WFE (Wafer Fab Equipment) market forecast has been revised upward from +5% to +10% growth, with the market size now estimated at $128.3B.\n    • Memory is the key driver of the upward revision: DRAM WFE projected at $34.9B (+18%), NAND WFE at $13.8B (+35%).\n    • For DRAM, Samsung and SK Hynix’s greenfield investments and strong DDR5/HBM demand are the main catalysts, while NAND is supported by surging eSSD demand (+50% in 2026).\n    • Memory supply-demand improvement creates upside risk for memory Capex in 2026/27.\n\n2. Company-Level Changes\n\n1) Applied Materials (AMAT)\n    • Rating upgraded (EW → OW), CY26 EPS revised from $9.58 → $10.45, target price $209.\n    • With DRAM exposure at 28%, the highest among peers, AMAT is set to benefit most from the memory cycle recovery.\n    • Risks related to China and ICAPS (power semiconductors/displays) largely resolved.\n    • Currently trading at a 25% valuation discount vs. LAM, but the target price reflects only a 15% discount → room for re-rating.\n\n2) Lam Research (LRCX)\n    • Rating upgraded (UW → EW), CY26 EPS revised from $5.12 → $5.43, target price $125.\n    • Supported by NAND recovery and non-China logic growth (+16%), WFE is expected to outperform for a third consecutive year.\n    • However, with current valuation at 23x already reflecting optimism, further upgrades would require revenue above $24B and EPS above $6.\n\n3. Regional WFE Outlook\n    • Korea: $28.2B (+20%), strong growth led by Samsung and SK Hynix.\n    • Taiwan: $28.2B (+15%), sustained growth driven by TSMC.\n    • China: $39.8B, flat (0% growth) due to U.S. export restrictions.\n    • Japan: $11.6B (+10%).\n    • North America & Europe: $12.8B and $3.8B respectively, modest growth.\n\n$AMAT $LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.039939438025515916, "ret_signed_m1d": -0.039939438025515916, "alpha_spy_m1d": -0.03523070710094356, "alpha_c_m1d": -0.019887880270568203, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0020423922223477, "ret_signed_p1d": -0.0020423922223477, "alpha_spy_p1d": 0.0034010826174598607, "alpha_c_p1d": -0.000583552908466789, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": -0.006438232654569442, "ret_signed_p1w": -0.006438232654569442, "alpha_spy_p1w": -0.0016994933967012793, "alpha_c_p1w": -0.007959119911651213, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.09929204028372207, "ret_signed_p1m": 0.09929204028372207, "alpha_spy_p1m": 0.09261886812950393, "alpha_c_p1m": 0.027497723989343026, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.25035483810359294, "ret_signed_p3m": 0.25035483810359294, "alpha_spy_p3m": 0.23591364988037244, "alpha_c_p3m": 0.17250775920293426, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 0.7213548401912171, "ret_signed_p6m": 0.7213548401912171, "alpha_spy_p6m": 0.7124590262708668, "alpha_c_p6m": 0.48562548704815156, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2737, -0.2675, -0.2648, -0.2637, -0.2677, -0.2524, -0.2526, -0.2526, -0.2576, -0.2449, -0.245, -0.2356, -0.2305, -0.2464, -0.2355, -0.2408, -0.2376, -0.2386, -0.2304, -0.261, -0.2655, -0.2604, -0.2666, -0.254, -0.2516, -0.2505, -0.2826, -0.271, -0.2556, -0.2687, -0.2743, -0.25, -0.2303, -0.2284, -0.2036, -0.1926, -0.1877, -0.2473, -0.252, -0.2411, -0.25, -0.2556, -0.243, -0.2339, -0.2161, -0.2158, -0.2126, -0.2424, -0.2424, -0.266, -0.2607, -0.2404, -0.2213, -0.2052, -0.2014, -0.1879, -0.1257, -0.1153, -0.0983, -0.0887, -0.0779, -0.0445, -0.0399, 0.0, -0.002, -0.0274, -0.0289, -0.0274, -0.0064, 0.0149, 0.0822, 0.1141, 0.1051, 0.1304, 0.0637, 0.0803, 0.0687, -0.0043, 0.0445, 0.0483, 0.0973, 0.0791, 0.0725, 0.0918, 0.0993, 0.0706, 0.1182, 0.1496, 0.1892, 0.1795, 0.2178, 0.2203, 0.1934, 0.2221, 0.1807, 0.251, 0.2293, 0.2078, 0.261, 0.2065, 0.2234, 0.162, 0.1237, 0.1176, 0.0856, 0.1278, 0.058, 0.0812, 0.1398, 0.1515, 0.1758, 0.1758, 0.1824, 0.1732, 0.199, 0.2128, 0.1926, 0.2048, 0.2355, 0.2588, 0.2774, 0.2808, 0.2186, 0.2473, 0.2394, 0.1766, 0.2504, 0.3078, 0.3305, 0.3298, 0.3462, 0.3462, 0.3519, 0.3352, 0.3193, 0.2995, 0.2995, 0.4049, 0.4786, 0.5712, 0.5417, 0.5256, 0.6577, 0.6732, 0.6275, 0.5851, 0.651, 0.6926, 0.6926, 0.6885, 0.7339, 0.6755, 0.6545, 0.692, 0.8103, 0.8188, 0.884, 0.7724, 0.803, 0.7469, 0.5926, 0.6194, 0.7538, 0.7406, 0.7204, 0.785, 0.7559, 0.7881, 0.7881, 0.7885, 0.8227, 0.8022, 0.8594, 0.8392, 0.8543, 0.894, 0.815, 0.7756, 0.7537, 0.6495, 0.6949, 0.6317, 0.5151, 0.6049, 0.6359, 0.6636, 0.5923, 0.6129, 0.6676, 0.7214]}
{"unit_id": "thread:1964989117378580519", "ticker_idx": 0, "ts": "2025-09-08T09:48:07+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BROADCOM WINS ORDERS FOR BYTEDANCE'S 2ND-GEN ASIC, APPLE'S CUSTOM ASIC, AND XAI'S CLOUD ASIC", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares bullish news of Broadcom winning major ASIC orders from ByteDance, Apple, and xAI.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "BROADCOM WINS ORDERS FOR BYTEDANCE’S 2ND-GEN ASIC, APPLE’S CUSTOM ASIC, AND XAI’S CLOUD ASIC\n\n$AVGO\n\n---\n\nhttps://t.co/BDbOIrZt5e", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031129663717725742, "ret_signed_m1d": -0.031129663717725742, "alpha_spy_m1d": -0.028679141619245563, "alpha_c_m1d": -0.020203014982361656, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025979983430521436, "ret_signed_p1d": -0.025979983430521436, "alpha_spy_p1d": -0.028291904128131318, "alpha_c_p1d": -0.0303304905991022, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.05334865965704916, "ret_signed_p1w": 0.05334865965704916, "alpha_spy_p1w": 0.03473053554589822, "alpha_c_p1w": 0.017431888348674818, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": -0.025064755049650667, "ret_signed_p1m": -0.025064755049650667, "alpha_spy_p1m": -0.05919561089954517, "alpha_c_p1m": -0.16175038093886418, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.1042466170647871, "ret_signed_p3m": 0.1042466170647871, "alpha_spy_p3m": 0.04651586067396751, "alpha_c_p3m": -0.11523275297268976, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": -0.08873355372693392, "ret_signed_p6m": -0.08873355372693392, "alpha_spy_p6m": -0.14329658980532434, "alpha_c_p6m": -0.41166909921848605, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.27, -0.2609, -0.2822, -0.2724, -0.2802, -0.2748, -0.2748, -0.2768, -0.2658, -0.2369, -0.2343, -0.2184, -0.2207, -0.2025, -0.2341, -0.2192, -0.2039, -0.2039, -0.2068, -0.2137, -0.196, -0.2032, -0.2062, -0.2027, -0.1872, -0.1876, -0.1713, -0.1803, -0.1662, -0.194, -0.1793, -0.1647, -0.1605, -0.1486, -0.1395, -0.1245, -0.1503, -0.1649, -0.1387, -0.1525, -0.1272, -0.1212, -0.1177, -0.1208, -0.095, -0.1058, -0.0996, -0.1137, -0.1154, -0.1468, -0.1576, -0.1622, -0.1494, -0.1488, -0.1378, -0.1313, -0.107, -0.1396, -0.1396, -0.1372, -0.1252, -0.1144, -0.0311, 0.0, -0.026, 0.0692, 0.0404, 0.0411, 0.0533, 0.0415, 0.0015, -0.0009, -0.0021, -0.0182, -0.0177, -0.0167, -0.026, -0.0305, -0.0497, -0.0439, -0.0338, -0.0199, -0.0194, -0.0277, -0.0251, 0.0013, -0.0001, -0.0592, 0.0337, -0.0027, 0.0182, 0.0263, 0.0124, 0.0121, -0.007, -0.0138, -0.0022, 0.0263, 0.0492, 0.0809, 0.1186, 0.091, 0.0712, 0.0507, 0.0199, 0.0403, 0.0305, 0.0127, 0.0386, 0.02, 0.0294, -0.0147, -0.0075, -0.007, -0.0132, 0.0271, 0.0051, -0.0141, 0.0953, 0.1158, 0.1522, 0.1522, 0.1678, 0.1189, 0.1058, 0.103, 0.1042, 0.1309, 0.1624, 0.1775, 0.1968, 0.1777, 0.0431, -0.0152, -0.0109, -0.0552, -0.044, -0.0136, -0.0086, 0.0143, 0.0169, 0.0169, 0.0224, 0.0145, 0.0158, 0.0049, 0.0049, 0.0094, -0.0028, -0.0018, -0.0026, -0.0346, 0.0017, 0.0227, 0.0296, -0.0131, -0.004, 0.0212, 0.0212, -0.0343, -0.0453, -0.0549, -0.0707, -0.0568, -0.0337, -0.0324, -0.0397, -0.038, -0.0386, -0.0699, -0.1055, -0.0984, -0.0333, -0.0013, -0.0115, -0.0048, -0.0384, -0.0558, -0.0558, -0.0344, -0.0316, -0.0302, -0.0341, -0.0408, -0.0549, -0.0351, -0.0659, -0.0722, -0.0743, -0.0887]}
{"unit_id": "thread:1972974037778530469", "ticker_idx": 2, "ts": "2025-09-30T10:37:21+00:00", "symbol": "ASMPT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASMPT has gained traction after winning supply orders with Micron, and SK hynix appears to have decided to reintroduce them… Industry observers expect SK hynix to adopt a dual-vendor model for HBM4 TC bonders, using both Hanmi Semiconductor and ASMPT", "tweet_type": "analysis", "timeline": "months", "summary": "SK hynix reverting to Hanmi as primary HBM4 TC bonder supplier with ASMPT as dual source; Hanwha Semitek losing ground on technical immaturity.", "resolved_tickers": ["0522.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "ASMPT listed on Hong Kong Exchange 0522.HK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Surrender to Hanmi” — SK hynix’s TC Bonder Supply Chain Diversification Stalls\n\nSK hynix is reorganizing its supply chain for TC bonders used in high-bandwidth memory (HBM), but the move is effectively faltering. The company has been pushing to diversify its supply chain away from its long-standing sole vendor model with Hanmi Semiconductor, but with its new partner Hanwha Semitek facing questions over technical maturity, the company is readjusting its strategy. Investment in TC bonders has also slowed, with expected equipment purchases this year likely to shrink to around 80 units, down from the original estimate of 120.\n\nAccording to industry sources, starting with the next-generation HBM4, SK hynix is considering operating its TC bonder supply chain mainly around Hanmi Semiconductor while also seeking to rekindle cooperation with Hong Kong’s ASMPT. The plan is to strengthen its long-standing partnership with Hanmi while mitigating the risks of having a single vendor through ASMPT rather than Hanwha Semitek.\n\nSK hynix had long received exclusive TC bonder supply from Hanmi Semiconductor but began moving toward vendor diversification last year as part of its risk management strategy. This year, it formally brought Hanwha Semitek into the supply chain, prioritizing its orders and giving the company substantial backing.\n\nThe order volume reflected this shift. In March and May, Hanwha Semitek secured three contracts worth a total of 89 billion KRW (including VAT): 23.1 billion KRW on March 14, another 23.1 billion KRW on March 16, and 42.8 billion KRW on May 16. Over the same period, Hanmi Semiconductor received just one order of 42.8 billion KRW, or about half as much. This caused friction between the two companies, as SK hynix allocated more orders to inexperienced Hanwha Semitek than to its established partner.\n\nRecently, however, sentiment has shifted toward giving Hanwha Semitek more time as a latecomer, and SK hynix appears to be stepping back. Hanwha Vision’s semiannual report shows that as of the first half of this year, Hanwha Semitek pledged land and buildings valued at 111.9 billion KRW as collateral to SK hynix. It is unusual for a vendor to provide such collateral to a customer, interpreted as SK hynix putting a safeguard in place for their transactions.\n\nHanwha Semitek also depends heavily on imported core parts such as TC bonder heads and motors. But its suppliers face capacity shortages, limiting their ability to meet demand. This has further influenced SK hynix’s decision to reinforce its relationships with established vendors as it transitions from HBM3E to HBM4.\n\nTo prepare for dual sourcing risks, the company is exploring reinstating ASMPT into its supply chain. ASMPT had previously received orders from SK hynix but deliveries were halted due to delays and stability issues. Earlier this year, SK hynix resumed qualification tests of ASMPT equipment, but progress was slow due to weak data processing speed. Recently, however, ASMPT has gained traction after winning supply orders with Micron, and SK hynix appears to have decided to reintroduce them.\n\nIndustry observers expect SK hynix to adopt a dual-vendor model for HBM4 TC bonders, using both Hanmi Semiconductor and ASMPT. SK hynix has not yet passed NVIDIA’s HBM4 qualification test but is expected to gain approval in November. Once it receives the purchase order from NVIDIA, mass production could begin by year-end.\n\nThat said, new TC bonder orders for HBM4 will likely be very limited this year. Although forecasts earlier in the year suggested SK hynix would order about 20 units in the second half, current expectations point to much lower volumes. With limited space in its fabs and HBM4 able to run on upgraded HBM3E TC bonders, demand for entirely new machines is minimal.\n\nAn industry insider noted, “Even if SK hynix places an order for mass production test equipment in October or November, it will be very small, around four units.” If so, Hanmi Semiconductor is expected to handle three units and ASMPT one.\n\nThe insider added, “The price of TC bonders for HBM4 is likely to rise about 40% from the previous generation. Even if upgrades rather than new orders are made, Hanmi Semiconductor is expected to fully reflect these costs.”\n\n---\n\nhttps://t.co/ynJ51zsqiI", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030469169274680907, "ret_signed_m1d": -0.030469169274680907, "alpha_spy_m1d": -0.02671643855218353, "alpha_c_m1d": -0.019132043110706043, "bench_spy_m1d": -0.003752730722497377, "bench_c_m1d": -0.011337126163974864, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0034075694259330103, "alpha_c_p1d": -0.022459919532809014, "bench_spy_p1d": 0.0034075694259330103, "bench_c_p1d": 0.022459919532809014, "ret_p1w": 0.08836081552367969, "ret_signed_p1w": 0.08836081552367969, "alpha_spy_p1w": 0.08394761598810874, "alpha_c_p1w": 0.05560551163119909, "bench_spy_p1w": 0.004413199535570955, "bench_c_p1w": 0.0327553038924806, "ret_p1m": 0.06337601061138454, "ret_signed_p1m": 0.06337601061138454, "alpha_spy_p1m": 0.031537783454156365, "alpha_c_p1m": -0.06608231085235183, "bench_spy_p1m": 0.03183822715722817, "bench_c_p1m": 0.12945832146373637, "ret_p3m": -0.056672763420673955, "ret_signed_p3m": -0.056672763420673955, "alpha_spy_p3m": -0.09595609947371131, "alpha_c_p3m": -0.18119298860424282, "bench_spy_p3m": 0.039283336053037354, "bench_c_p3m": 0.12452022518356887, "ret_p6m": 0.33455212256703115, "ret_signed_p6m": 0.33455212256703115, "alpha_spy_p6m": 0.342988395388933, "alpha_c_p6m": 0.1081101342002837, "bench_spy_p6m": -0.008436272821901847, "bench_c_p6m": 0.22644198836674745, "price_path": [-0.3018, -0.2861, -0.2958, -0.2824, -0.2873, -0.2867, -0.2672, -0.2642, -0.2484, -0.2557, -0.2399, -0.2442, -0.2326, -0.2569, -0.2326, -0.1476, -0.1428, -0.164, -0.158, -0.1883, -0.1877, -0.2084, -0.1798, -0.1792, -0.1786, -0.1568, -0.1622, -0.1444, -0.1426, -0.1249, -0.1481, -0.1255, -0.1456, -0.1871, -0.1889, -0.195, -0.1505, -0.0859, -0.114, -0.1469, -0.114, -0.1408, -0.1347, -0.1706, -0.1773, -0.1828, -0.1718, -0.1676, -0.1651, -0.1609, -0.1347, -0.1353, -0.1438, -0.1261, -0.1316, -0.0683, -0.0622, -0.0341, -0.0372, 0.0225, -0.0073, -0.039, -0.0305, 0.0, 0.0, 0.036, 0.0225, 0.0884, 0.0884, 0.0981, 0.1621, 0.131, 0.0951, 0.0232, 0.0701, 0.0762, 0.0262, 0.0609, 0.0463, 0.0183, 0.0146, 0.0969, 0.1426, 0.0634, 0.0634, 0.0219, -0.0006, 0.0067, -0.0049, -0.0061, 0.003, -0.011, 0.0037, -0.014, -0.0335, -0.0238, -0.0506, -0.05, -0.0737, -0.0853, -0.0756, -0.1322, -0.1365, -0.1066, -0.1164, -0.0975, -0.0823, -0.0676, -0.0683, -0.0573, -0.0488, -0.042, -0.0512, -0.036, -0.0603, -0.039, -0.0402, -0.0835, -0.089, -0.0743, -0.0804, -0.0938, -0.0573, -0.0561, -0.0567, -0.0567, -0.0567, -0.0719, -0.0585, -0.053, -0.053, -0.0104, 0.0018, 0.039, 0.103, 0.0847, 0.0963, 0.0975, 0.1371, 0.1499, 0.1615, 0.1926, 0.1871, 0.1895, 0.2407, 0.3187, 0.27, 0.2529, 0.2956, 0.3346, 0.27, 0.2663, 0.2169, 0.27, 0.2566, 0.1737, 0.1999, 0.231, 0.2297, 0.2797, 0.2736, 0.27, 0.2553, 0.2553, 0.2553, 0.2553, 0.2785, 0.3443, 0.3601, 0.3467, 0.3894, 0.3662, 0.3211, 0.2602, 0.3163, 0.3699, 0.3541, 0.309, 0.365, 0.348, 0.3321, 0.3187, 0.3102, 0.2651, 0.3285, 0.3077, 0.2821, 0.2505, 0.2712, 0.3346]}
{"unit_id": "reply:1950400314844336623", "ticker_idx": 0, "ts": "2025-07-30T03:37:26+00:00", "symbol": "삼성전기", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "테슬라의 세컨~써드 벤더에서 퍼스트 벤더가 될 것 같은...", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Samsung Electro-Mechanics likely to be upgraded from secondary/tertiary to primary Tesla vendor; implied bullish.", "resolved_tickers": ["009150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "삼성전기 = Samsung Electro-Mechanics, KOSPI 009150.KS", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@Alisvolatprop12 테슬라의 세컨~써드 벤더에서 퍼스트 벤더가 될 것 같은...", "reply_text": null, "parent_text": "삼성전기의 폭등 이유\n\n1. 과거 테슬라 HWD에 FC BGA를 공급한 이력 존재\n\n2. 삼성전기는 전장용 카메라 모듈도 다수 공급", "quoted_text": null, "parent_author": "Alisvolatprop12", "ret_m1d": -0.09545755015471191, "ret_signed_m1d": -0.09545755015471191, "alpha_spy_m1d": -0.09671837206390821, "alpha_c_m1d": -0.09322834178300188, "bench_spy_m1d": 0.0012608219091962969, "bench_c_m1d": -0.002229208371710034, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02106651530426662, "ret_signed_p1d": -0.02106651530426662, "alpha_spy_p1d": -0.01731538579371894, "alpha_c_p1d": -0.013736846689473725, "bench_spy_p1d": -0.0037511295105476794, "bench_c_p1d": -0.007329668614792895, "ret_p1w": 0.024358190759807163, "ret_signed_p1w": 0.024358190759807163, "alpha_spy_p1w": 0.027006149607883945, "alpha_c_p1w": 0.03051664515063912, "bench_spy_p1w": -0.0026479588480767813, "bench_c_p1w": -0.006158454390831958, "ret_p1m": 0.06583278247163626, "ret_signed_p1m": 0.06583278247163626, "alpha_spy_p1m": 0.043041845099637044, "alpha_c_p1m": 0.05880538136092195, "bench_spy_p1m": 0.02279093737199922, "bench_c_p1m": 0.007027401110714315, "ret_p3m": 0.5042791780922027, "ret_signed_p3m": 0.5042791780922027, "alpha_spy_p3m": 0.42124829261220276, "alpha_c_p3m": 0.37316141578397644, "bench_spy_p3m": 0.08303088547999993, "bench_c_p3m": 0.13111776230822625, "ret_p6m": 0.880126958396378, "ret_signed_p6m": 0.880126958396378, "alpha_spy_p6m": 0.7879672410936831, "alpha_c_p6m": 0.7823014553079473, "bench_spy_p6m": 0.0921597173026949, "bench_c_p6m": 0.09782550308843074, "price_path": [-0.2258, -0.2258, -0.2258, -0.2212, -0.2245, -0.2284, -0.1718, -0.1705, -0.1685, -0.1955, -0.1929, -0.2054, -0.214, -0.2054, -0.2179, -0.2278, -0.2028, -0.2179, -0.1883, -0.1837, -0.1955, -0.2047, -0.2047, -0.1988, -0.131, -0.131, -0.1442, -0.1455, -0.1284, -0.135, -0.1613, -0.1646, -0.1396, -0.1244, -0.1422, -0.1224, -0.1422, -0.1001, -0.0843, -0.0856, -0.108, -0.1132, -0.1099, -0.1145, -0.0632, -0.0876, -0.0856, -0.1001, -0.1086, -0.0974, -0.0974, -0.0948, -0.0823, -0.0915, -0.0928, -0.0869, -0.0961, -0.1145, -0.1053, -0.1106, -0.1119, -0.0882, -0.0955, 0.0, -0.0211, -0.0566, -0.0105, 0.025, 0.0244, 0.0224, 0.054, 0.052, 0.0388, 0.0724, 0.0704, 0.0704, 0.0415, 0.0467, 0.0303, 0.0191, 0.0184, 0.0375, 0.05, 0.0619, 0.0658, 0.0566, 0.0678, 0.1185, 0.1468, 0.1745, 0.1731, 0.1764, 0.187, 0.1883, 0.2377, 0.2535, 0.2442, 0.2699, 0.2646, 0.2837, 0.2837, 0.3627, 0.343, 0.3265, 0.3167, 0.262, 0.3009, 0.2739, 0.2725, 0.2936, 0.2936, 0.2936, 0.2936, 0.2936, 0.2936, 0.3858, 0.3397, 0.3048, 0.3167, 0.3265, 0.3627, 0.3989, 0.422, 0.3957, 0.4615, 0.4615, 0.5043, 0.5273, 0.5273, 0.4812, 0.6129, 0.5997, 0.5734, 0.5109, 0.5273, 0.4648, 0.4747, 0.4977, 0.4911, 0.4582, 0.4022, 0.4187, 0.343, 0.4154, 0.4384, 0.4319, 0.4878, 0.5899, 0.6919, 0.6787, 0.6524, 0.6458, 0.659, 0.7182, 0.7446, 0.7314, 0.7248, 0.7742, 0.7676, 0.7742, 0.7907, 0.7347, 0.6787, 0.7018, 0.6722, 0.6656, 0.6754, 0.6557, 0.6689, 0.6689, 0.7051, 0.7107, 0.6941, 0.6941, 0.6941, 0.7938, 0.7938, 0.7805, 0.7772, 0.807, 0.7672, 0.8536, 0.92, 0.8901, 0.9233, 0.9233, 0.9399, 0.8735, 0.8602, 0.8801]}
{"unit_id": "orig:1958018622837973044", "ticker_idx": 0, "ts": "2025-08-20T04:09:52+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "There is a statistically significant relationship between low active holdings and future stock performance... Leading indicators of compute demand remain exceptionally strong, with no signs of slowing... we continue to view Nvidia as a high-quality asset in the current AI-dominated era", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley's bullish NVDA thesis: most underweight large-cap tech stock with strong fundamentals and historically underweight stocks outperform as holdings catch up to index weight.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Surprised? Nvidia is the “most underweight” large-cap tech stock\n\nAs the world’s most valuable company and the “leader” of the AI concept, Nvidia has actually become the large-cap tech stock that institutional investors are least willing to overweight.\n\nMorgan Stanley’s latest research shows that although Nvidia has risen to the top as the world’s most valuable company, institutional investors’ actual holdings still significantly lag its market position. Analyst Erik Woodring pointed out in the research report that Nvidia has now become “the most underweight large-cap tech stock.”\n\nData shows that Nvidia’s weight in the S&P 500 Index has reached 7.37%, but its share in institutional investors’ average portfolios is only 4.2%, leaving an adjusted underweight gap of as much as 2.41 percentage points. This gap ranks first among the 15 major tech companies tracked by Morgan Stanley.\n\nAnalysts believe that this mismatch between holdings and index weight highlights Nvidia’s unique position. Although the stock has surged nearly 1,300% over the past five years on the back of the AI boom, its rapid rise, along with geopolitical and supply chain-related risks, has caused some investors to remain cautious about significantly increasing their holdings.\n\nAmong other large tech stocks tracked by Morgan Stanley, Microsoft, Apple, and Amazon are also underweight, though to a lesser extent than Nvidia. Microsoft is underweight by 2.39 percentage points, Apple by 1.66 percentage points, and Amazon by 1.40 percentage points.\n\nIn contrast, institutional investors are overweight in certain tech stocks. Intuit is overweight by 0.83 percentage points, Oracle by 0.32 percentage points, and Dell by 0.25 percentage points.\n\nHistorical experience shows that underweight stocks often perform better over time, as investors gradually increase their holdings to match their index weight. Analysts said: “There is a statistically significant relationship between low active holdings and future stock performance.”\n\nFundamentals remain solid\nDespite being underweight, Morgan Stanley analysts remain optimistic about Nvidia’s fundamentals. In the report, analysts wrote:\n\n“Leading indicators of compute demand remain exceptionally strong, with no signs of slowing. As supply chain constraints on rack-level solutions gradually ease, and with the U.S. government advancing export license approvals for China, we continue to view Nvidia as a high-quality asset in the current AI-dominated era.”\n\nNvidia’s stock has risen 35% over the past year, outperforming the S&P 500 Index’s roughly 10% gain. Market optimism is mainly driven by demand for its graphics processing units (GPUs), which are widely used in AI and cloud enterprise applications.\n\nHowever, not everyone agrees with the optimistic outlook for large-cap stocks. Apollo Management’s chief economist Torsten Sløk previously stated that the current valuations of large-cap tech stocks and the overall index may be difficult to sustain. He noted that the price-to-earnings ratios of the top 10 companies in the S&P 500, including Meta and Nvidia, have already exceeded the levels seen during the 1999 internet bubble.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0013682987641980837, "ret_signed_m1d": 0.0013682987641980837, "alpha_spy_m1d": -0.0012958637691051766, "alpha_c_m1d": -0.00524164035185648, "bench_spy_m1d": 0.0026641625333032604, "bench_c_m1d": 0.006609939116054564, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0023945663416949925, "ret_signed_p1d": -0.0023945663416949925, "alpha_spy_p1d": 0.0016172559458635938, "alpha_c_p1d": 0.002796513053215821, "bench_spy_p1d": -0.004011822287558586, "bench_c_p1d": -0.005191079394910814, "ret_p1w": 0.03534780508381363, "ret_signed_p1w": 0.03534780508381363, "alpha_spy_p1w": 0.02199583544208994, "alpha_c_p1w": 0.005239862414518237, "bench_spy_p1w": 0.013351969641723693, "bench_c_p1w": 0.030107942669295396, "ret_p1m": 0.004845775344937353, "ret_signed_p1m": 0.004845775344937353, "alpha_spy_p1m": -0.033000399871087494, "alpha_c_p1m": -0.09208809771939919, "bench_spy_p1m": 0.03784617521602485, "bench_c_p1m": 0.09693387306433654, "ret_p3m": 0.06391406533865873, "ret_signed_p3m": 0.06391406533865873, "alpha_spy_p3m": 0.01783183793325338, "alpha_c_p3m": -0.1137575130778925, "bench_spy_p3m": 0.04608222740540535, "bench_c_p3m": 0.17767157841655123, "ret_p6m": 0.0659119590321291, "ret_signed_p6m": 0.0659119590321291, "alpha_spy_p6m": -0.007848851859500572, "alpha_c_p6m": -0.3438805355145891, "bench_spy_p6m": 0.07376081089162967, "bench_c_p6m": 0.4097924945467182, "price_path": [-0.2515, -0.2515, -0.2275, -0.2315, -0.2065, -0.2296, -0.2168, -0.1949, -0.1909, -0.2019, -0.1921, -0.1869, -0.1793, -0.1857, -0.1733, -0.1906, -0.1751, -0.1783, -0.1706, -0.1706, -0.1799, -0.1781, -0.1568, -0.1202, -0.1162, -0.1006, -0.0993, -0.126, -0.1035, -0.0916, -0.0916, -0.0978, -0.0878, -0.0714, -0.0644, -0.0597, -0.0646, -0.0268, -0.023, -0.0137, -0.017, -0.0229, -0.0477, -0.0263, -0.0095, -0.0108, 0.0077, 0.0006, 0.0221, 0.0141, -0.0096, 0.0262, 0.0163, 0.0229, 0.0306, 0.0416, 0.038, 0.0442, 0.0353, 0.0377, 0.0288, 0.0377, 0.0014, 0.0, -0.0024, 0.0148, 0.0251, 0.0363, 0.0353, 0.0272, -0.007, -0.007, -0.0263, -0.0273, -0.0213, -0.0478, -0.0404, -0.0265, 0.011, 0.0101, 0.0139, 0.0135, -0.0029, -0.0291, 0.0048, 0.0073, 0.0469, 0.0173, 0.009, 0.0131, 0.016, 0.0368, 0.0638, 0.0676, 0.077, 0.0697, 0.0579, 0.055, 0.0782, 0.098, 0.0443, 0.0737, 0.0265, 0.0253, 0.0366, 0.0446, 0.0413, 0.0329, 0.0279, 0.0386, 0.062, 0.0918, 0.1462, 0.1805, 0.1568, 0.1545, 0.1795, 0.1328, 0.113, 0.0724, 0.0728, 0.1349, 0.1013, 0.105, 0.0654, 0.0843, 0.0639, 0.034, 0.0635, 0.0299, 0.0199, 0.0408, 0.0139, 0.0278, 0.0278, 0.0092, 0.0258, 0.0346, 0.0239, 0.0456, 0.0401, 0.058, 0.0547, 0.0479, 0.0316, -0.0021, 0.0052, 0.0133, -0.0253, -0.0071, 0.032, 0.0474, 0.0789, 0.0754, 0.0754, 0.0864, 0.0732, 0.0693, 0.0634, 0.0634, 0.0768, 0.0726, 0.0676, 0.0783, 0.0551, 0.0541, 0.0545, 0.0595, 0.0442, 0.0665, 0.0619, 0.0619, 0.0153, 0.0453, 0.0539, 0.0701, 0.0632, 0.0749, 0.092, 0.0977, 0.0898, 0.0583, 0.0283, -0.0068, -0.02, 0.0572, 0.0836, 0.075, 0.0836, 0.0659]}
{"unit_id": "orig:1967523528737222915", "ticker_idx": 0, "ts": "2025-09-15T09:38:58+00:00", "symbol": "WDC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "As demand for all capacities reaches an all-time high, the company will gradually raise prices for all HDD products, effective immediately.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish WDC: all-time-high HDD demand enabling immediate price increases across all capacities.", "resolved_tickers": ["WDC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "WDC\n\n- As demand for all capacities reaches an all-time high, the company will gradually raise prices for all HDD products, effective immediately.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04619590823579178, "ret_signed_m1d": -0.04619590823579178, "alpha_spy_m1d": -0.04090024331230513, "alpha_c_m1d": -0.03726879520450499, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.008927113031286793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006836543185559707, "ret_signed_p1d": 0.006836543185559707, "alpha_spy_p1d": 0.008213335970841396, "alpha_c_p1d": 0.010312321314682071, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": -0.0034757781291223644, "ret_p1w": 0.09786115547219287, "ret_signed_p1w": 0.09786115547219287, "alpha_spy_p1w": 0.08609131409813076, "alpha_c_p1w": 0.06235965362527929, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.035501501846913586, "ret_p1m": 0.10489302419970059, "ret_signed_p1m": 0.10489302419970059, "alpha_spy_p1m": 0.10011778937052185, "alpha_c_p1m": 0.07360440670042911, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.031288617499271476, "ret_p3m": 0.8297735663661057, "ret_signed_p3m": 0.8297735663661057, "alpha_spy_p3m": 0.7841233346918248, "alpha_c_p3m": 0.7456218209153964, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.08415174545070925, "ret_p6m": 1.6033946206390755, "ret_signed_p6m": 1.6033946206390755, "alpha_spy_p6m": 1.5729002856249614, "alpha_c_p6m": 1.5778430998291937, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.025551520809881767, "price_path": [-0.4226, -0.4226, -0.4216, -0.411, -0.3945, -0.3898, -0.3804, -0.3826, -0.3758, -0.3772, -0.3583, -0.3554, -0.3554, -0.3638, -0.3755, -0.3694, -0.3653, -0.3548, -0.3471, -0.3412, -0.351, -0.3462, -0.3366, -0.3294, -0.3458, -0.3238, -0.3267, -0.3286, -0.327, -0.3112, -0.3032, -0.2324, -0.2532, -0.246, -0.2602, -0.2803, -0.2738, -0.2687, -0.2719, -0.2595, -0.2579, -0.2563, -0.2678, -0.2558, -0.26, -0.2621, -0.2717, -0.2491, -0.2272, -0.2224, -0.213, -0.1997, -0.2163, -0.2163, -0.2009, -0.1611, -0.1162, -0.1011, -0.0889, -0.0767, -0.072, -0.0609, -0.0462, 0.0, 0.0068, -0.0142, 0.027, 0.0414, 0.0979, 0.0768, 0.0713, 0.0471, 0.0439, 0.1402, 0.1726, 0.2754, 0.282, 0.2824, 0.2236, 0.1713, 0.1835, 0.1691, 0.1273, 0.1609, 0.1049, 0.1763, 0.2298, 0.2325, 0.1869, 0.1858, 0.1766, 0.2279, 0.2641, 0.2371, 0.22, 0.3808, 0.3491, 0.467, 0.5433, 0.4863, 0.5636, 0.5978, 0.5916, 0.7015, 0.6602, 0.6223, 0.5349, 0.5415, 0.5866, 0.4929, 0.5038, 0.3696, 0.3594, 0.4741, 0.5178, 0.5406, 0.5406, 0.5952, 0.5972, 0.5626, 0.5196, 0.5737, 0.6508, 0.6595, 0.6572, 0.7785, 0.8298, 0.7236, 0.6816, 0.7064, 0.6251, 0.7106, 0.77, 0.7277, 0.7423, 0.7551, 0.7551, 0.7745, 0.7563, 0.7209, 0.6838, 0.6838, 0.8347, 0.8364, 1.1443, 0.9537, 0.8345, 0.9594, 1.0735, 1.0917, 1.1015, 1.1709, 1.1651, 1.1651, 1.1794, 1.3644, 1.378, 1.3106, 1.3542, 1.4696, 1.7339, 1.7213, 1.4459, 1.6413, 1.8369, 1.6333, 1.5432, 1.7621, 1.7954, 1.5664, 1.6757, 1.7769, 1.7523, 1.7523, 1.777, 1.8987, 1.7825, 1.7908, 1.7409, 1.6447, 1.8439, 1.7588, 1.7339, 1.6399, 1.4496, 1.5541, 1.5331, 1.3983, 1.5627, 1.6034]}
{"unit_id": "orig:1945828530765570056", "ticker_idx": 0, "ts": "2025-07-17T12:50:48+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "In the short term, the stock price may rise by 3% to 5%, reflecting the strong Q3 revenue growth guidance and robust margin outlook. However, due to the conservative Q4 guidance, the stock is likely to trade within a narrow range for the next two to three months.", "tweet_type": "news_reaction", "timeline": "months", "summary": "JPMorgan shared/adopted: TSM +3-5% near-term on strong Q3 guidance/margins, then range-bound 2-3 months on conservative Q4 outlook; 2025 EPS raised ~3-5%.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC - JP Morgan\n\nTSMC's Q2 2025 earnings announcement showed profitability and operating profit margins exceeding market expectations. The company's performance was robust, particularly due to high margins and non-operating income, despite foreign exchange headwinds. The Q3 2025 guidance was set at approximately an 8% increase quarter-over-quarter, which is higher than our firm's estimated 4% increase. This performance improvement was primarily driven by seasonal demand from smartphones, while the HPC segment is expected to remain flat after a surge in the first half of the year. The gross margin guidance of around 56.5% also indicates sound profitability despite the strong appreciation of the New Taiwan Dollar. A positive outlook is maintained for data center AI demand, and CoWoS supply shortages are expected to persist until 2026.\n\nTSMC upgraded its full-year 2025 revenue guidance to roughly 30% growth, but anticipates Q4 2025 revenue to decline by approximately 10% quarter-over-quarter. This is a significant decrease even considering typical seasonality. JPMorgan analyzes this as a result of conservative assumptions regarding tariff impacts and the pulled-forward demand from Q2 and Q3. In terms of long-term growth, the N2 process and AI demand were cited as key drivers, with the ramp-up of N3 and N2 nodes at the Arizona fab also accelerating. Particularly, N2 is experiencing strong demand from both HPC and AI, with wafer input expected to begin by the end of 2025 and revenue to fully materialize by mid-2026. The average selling price (ASP) in Q2 increased by 16% year-over-year, and double-digit ASP growth is projected to continue in the future.\n\nHowever, the Q4 revenue guidance reflects a sharper decline than expected, and demand recovery in non-AI segments remains weak. The effect of China's consumer subsidies may also be short-lived, and a somewhat cautious tone was observed regarding expansion plans in Japan and Europe. Capital expenditure plans beyond 2026 were not specified, and the company is assessed to have taken a more conservative stance than market expectations. This presents a slight disconnect with the strong anticipation for N2 process expansion.\n\nJPMorgan expects to upgrade its 2025 earnings forecast by about 3% to 5% based on this announcement, but a significant upward revision of profit forecasts will be challenging until the extent of the Q4 earnings decline becomes clear. In the short term, the stock price may rise by 3% to 5%, reflecting the strong Q3 revenue growth guidance and robust margin outlook. However, due to the conservative Q4 guidance, the stock is likely to trade within a narrow range for the next two to three months.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032736137765441486, "ret_signed_m1d": -0.032736137765441486, "alpha_spy_m1d": -0.02665367060126722, "alpha_c_m1d": -0.024405971921902414, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.008330165843539072, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021172647928687227, "ret_signed_p1d": -0.021172647928687227, "alpha_spy_p1d": -0.02044023502657677, "alpha_c_p1d": -0.016373454888251704, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.004799193040435523, "ret_p1w": -0.016286599225248666, "ret_signed_p1w": -0.016286599225248666, "alpha_spy_p1w": -0.02644521420138235, "alpha_c_p1w": -0.002882922782284747, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": -0.013403676442963919, "ret_p1m": -0.02736158446200665, "ret_signed_p1m": -0.02736158446200665, "alpha_spy_p1m": -0.051882372472820015, "alpha_c_p1m": -0.040868208695853725, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.013506624233847075, "ret_p3m": 0.20876894271373292, "ret_signed_p3m": 0.20876894271373292, "alpha_spy_p3m": 0.15140639950328283, "alpha_c_p3m": 0.06551011307794985, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.14325882963578307, "ret_p6m": 0.3254373776897599, "ret_signed_p6m": 0.3254373776897599, "alpha_spy_p6m": 0.2139622326356443, "alpha_c_p6m": -0.012984752748134376, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.33842213043789426, "price_path": [-0.4002, -0.3858, -0.3598, -0.334, -0.3302, -0.3371, -0.3327, -0.3238, -0.2993, -0.2727, -0.2844, -0.3011, -0.2919, -0.2892, -0.2839, -0.2415, -0.213, -0.2099, -0.2121, -0.2121, -0.215, -0.2152, -0.2221, -0.2041, -0.2212, -0.2212, -0.198, -0.2043, -0.2002, -0.2157, -0.2096, -0.1983, -0.1789, -0.1751, -0.1676, -0.1602, -0.1381, -0.1314, -0.1228, -0.1405, -0.1218, -0.1291, -0.1307, -0.1307, -0.1469, -0.1436, -0.1039, -0.0931, -0.0879, -0.0693, -0.0778, -0.0852, -0.0489, -0.044, -0.044, -0.0669, -0.0722, -0.056, -0.0645, -0.0619, -0.0689, -0.0352, -0.0327, 0.0, -0.0212, -0.0275, -0.0448, -0.0215, -0.0163, 0.0, -0.0116, -0.0174, -0.011, -0.0162, -0.0423, -0.0269, -0.0535, -0.0579, -0.0121, -0.0154, -0.0143, -0.0053, -0.0169, -0.0187, -0.0274, -0.0171, -0.0525, -0.0692, -0.0744, -0.0513, -0.0408, -0.028, -0.0257, -0.0298, -0.06, -0.06, -0.0701, -0.0579, -0.0423, -0.0089, 0.0065, 0.0217, 0.0604, 0.0542, 0.0559, 0.0643, 0.0704, 0.0734, 0.0973, 0.0819, 0.1136, 0.1547, 0.1466, 0.13, 0.1165, 0.116, 0.1408, 0.1783, 0.1768, 0.1935, 0.2352, 0.201, 0.2438, 0.2249, 0.1464, 0.2372, 0.2088, 0.2446, 0.2247, 0.2053, 0.216, 0.2029, 0.1799, 0.1875, 0.2048, 0.2182, 0.2316, 0.2461, 0.2385, 0.2271, 0.2452, 0.201, 0.1994, 0.1814, 0.1702, 0.206, 0.1893, 0.187, 0.1526, 0.1633, 0.1519, 0.1351, 0.1533, 0.1335, 0.1235, 0.1626, 0.1628, 0.1843, 0.1843, 0.1907, 0.175, 0.193, 0.2068, 0.1965, 0.2038, 0.233, 0.2393, 0.2668, 0.2485, 0.1961, 0.1784, 0.1749, 0.1343, 0.1659, 0.1834, 0.2011, 0.2162, 0.2237, 0.2237, 0.2403, 0.2324, 0.2269, 0.2446, 0.2446, 0.309, 0.3198, 0.341, 0.3052, 0.3024, 0.3254]}
{"unit_id": "quote:1950103088616132918", "ticker_idx": 0, "ts": "2025-07-29T07:56:22+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The CoWoS allocation for AMD's AI accelerators in 2026 is double that of 2025.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish AMD: CoWoS packaging allocation for AI accelerators doubling from 2025 to 2026, signaling strong demand ramp.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I’d like to draw attention to AMD in this chart.\n\nThe CoWoS allocation for AMD’s AI accelerators in 2026 is double that of 2025.\n\n$AMD https://t.co/AtdU4XtlaQ", "reply_text": null, "parent_text": null, "quoted_text": "Morgan Stanley's Detailed Analysis of the TSMC CoWoS Capacity Battle: NVIDIA Secures 60%, Cloud AI Chip Market to Surge 40-50% by 2026\n\nAccording to a recent research report from Morgan Stanley, a detailed forecast has been released regarding the allocation of CoWoS capacity https://t.co/unpdIBsicl", "parent_author": null, "ret_m1d": -0.021302968481107243, "ret_signed_m1d": -0.021302968481107243, "alpha_spy_m1d": -0.02394768982272466, "alpha_c_m1d": -0.01621599866774548, "bench_spy_m1d": 0.0026447213416174176, "bench_c_m1d": -0.005086969813361764, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011665870361522446, "ret_signed_p1d": 0.011665870361522446, "alpha_spy_p1d": 0.012925104600601367, "alpha_c_p1d": 0.00019426848129167062, "bench_spy_p1d": -0.0012592342390789213, "bench_c_p1d": 0.011471601880230775, "ret_p1w": -0.01763979282995154, "ret_signed_p1w": -0.01763979282995154, "alpha_spy_p1w": -0.006164153625172086, "alpha_c_p1w": 0.0021620983748227296, "bench_spy_p1w": -0.011475639204779453, "bench_c_p1w": -0.01980189120477427, "ret_p1m": -0.05810413332246367, "ret_signed_p1m": -0.05810413332246367, "alpha_spy_p1m": -0.07600239037406875, "alpha_c_p1m": -0.07435542996569278, "bench_spy_p1m": 0.017898257051605082, "bench_c_p1m": 0.016251296643229107, "ret_p3m": 0.42538319820224224, "ret_signed_p3m": 0.42538319820224224, "alpha_spy_p3m": 0.35632850174106245, "alpha_c_p3m": 0.22654384086696489, "bench_spy_p3m": 0.06905469646117979, "bench_c_p3m": 0.19883935733527736, "ret_p6m": 0.407799817486173, "ret_signed_p6m": 0.407799817486173, "alpha_spy_p6m": 0.3226831987917187, "alpha_c_p6m": 0.031286814473518776, "bench_spy_p6m": 0.0851166186944543, "bench_c_p6m": 0.3765130030126542, "price_path": [-0.4553, -0.4432, -0.4331, -0.4442, -0.4344, -0.4268, -0.4204, -0.3907, -0.3662, -0.3366, -0.3519, -0.3397, -0.3534, -0.3603, -0.3685, -0.3761, -0.3783, -0.3783, -0.3544, -0.364, -0.363, -0.376, -0.354, -0.3389, -0.3317, -0.348, -0.3452, -0.314, -0.3055, -0.3173, -0.3322, -0.3454, -0.2877, -0.2837, -0.2854, -0.2854, -0.2773, -0.2697, -0.2198, -0.1918, -0.1903, -0.1895, -0.2003, -0.2329, -0.2193, -0.2228, -0.2228, -0.2403, -0.2233, -0.22, -0.1876, -0.1748, -0.1758, -0.123, -0.0978, -0.096, -0.1153, -0.1152, -0.128, -0.1059, -0.0863, -0.0618, -0.0213, 0.0, 0.0117, -0.0064, -0.0323, -0.0037, -0.0176, -0.0807, -0.0284, -0.0264, -0.0291, -0.014, 0.0393, 0.0198, 0.0004, -0.0073, -0.0614, -0.069, -0.0774, -0.0546, -0.0794, -0.061, -0.0581, -0.0499, -0.0835, -0.0835, -0.0852, -0.0863, -0.0882, -0.1482, -0.1467, -0.1218, -0.1009, -0.1227, -0.1063, -0.0917, -0.0957, -0.103, -0.11, -0.113, -0.0995, -0.0932, -0.0933, -0.0911, -0.1013, -0.0906, -0.0882, -0.0757, -0.0435, -0.072, 0.1481, 0.192, 0.3275, 0.3125, 0.2111, 0.2197, 0.2291, 0.3447, 0.3219, 0.3136, 0.3557, 0.3415, 0.2975, 0.3243, 0.4254, 0.4634, 0.4541, 0.4897, 0.4362, 0.4434, 0.4633, 0.4092, 0.4446, 0.3396, 0.3162, 0.375, 0.3386, 0.459, 0.3974, 0.3909, 0.3555, 0.2978, 0.2599, 0.1611, 0.1484, 0.212, 0.1617, 0.2074, 0.2074, 0.2259, 0.2385, 0.213, 0.2263, 0.2172, 0.2284, 0.2461, 0.249, 0.2479, 0.2479, 0.1879, 0.1699, 0.1788, 0.1165, 0.1331, 0.2028, 0.2114, 0.2111, 0.2119, 0.2119, 0.2116, 0.2151, 0.2136, 0.2069, 0.2069, 0.2594, 0.2459, 0.208, 0.1836, 0.1535, 0.145, 0.1705, 0.2453, 0.2601, 0.2845, 0.3065, 0.3065, 0.307, 0.4078]}
{"unit_id": "orig:1945827138789330955", "ticker_idx": 0, "ts": "2025-07-17T12:45:16+00:00", "symbol": "CRWV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Initiating Coverage with 'Reduce' Target Price: $32.00 → Represents a 77% downside from the current stock price of $140.59", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "HSBC initiates CoreWeave with Reduce, PT $32 (77% downside); DCF shows excessive expectations, weak FCF, high leverage, customer concentration risk.", "resolved_tickers": ["CRWV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "HSBC has set a price target of $32 for CoreWeave.\n\nHSBC's Investment Opinion: Initiating Coverage with 'Reduce'\nTarget Price: $32.00 → Represents a 77% downside from the current stock price of $140.59.\n\nAccording to DCF-based valuation, the current stock price reflects excessive expectations.\n\nKey Investment Summary\n1. Business Overview and Market Position\nCoreWeave is an AI-specialized GPU cloud infrastructure company, deploying over 250,000 Nvidia GPUs across 32 data centers in North America and Europe.\n\nKey Customers:\nMicrosoft (62% of 2024 revenue, 72% as of 1Q25)\nOpenAI (Total contract value of approximately $15.9B, but no upfront payments)\nNvidia (Approximately 4% of 2024 revenue)\n\n2. Growth and Outlook\n\nThe GPU cloud market is projected to grow at a CAGR of 55% from 2024-27, reaching a market size of $200B by 2027.\n\nCoreWeave is expected to expand its market share from 4% in 2024 to 9% by 2027 in this market.\n\nHowever, cash flow (FCF) and profitability are weak, specifically:\nEBITDA margin consensus is 10%p lower than estimated for 2025-30.\nMaintenance CAPEX is expected to increase (reaching 35% of revenue after 2030).\nNet debt/equity ratio is projected to approach 10x after 2025.\nCovenant breach risk in 4Q25.\nStructural Risks and Rationale for Investment Opinion\n\n1. Customer Concentration and Commoditization Risk\n\nMicrosoft and OpenAI account for most of the revenue → significant risk if customer diversification fails.\n\nNeither company uses CoreWeave's software stack, meaning no customer lock-in effect.\n\nMicrosoft is only using bare metal; contract termination is possible if GPU supply improves.\n\nOpenAI's contract is large, but risk is shifted without upfront payments.\n\n2. High Funding Costs and Financial Leverage\n\nAverage interest rate of 13% in 1Q25 (including estimates).\nLow credit ratings of customers other than Microsoft → increased equity contribution required for SPV formation.\nNeeds to raise $42B in funding from 2025 to 2030, which cannot be covered by internal cash flow.\n\nBased on cash and remaining credit, an additional $7.6B is needed for 2025 alone.\n\n3. Structural Limitations of GPU Cloud\n\nGPU cloud is difficult to differentiate from general CPU cloud (low software demand and intensified price competition).\n\nLower asset turnover ratio compared to AWS and Azure (CRWV: 32%, AWS: 76%).\nVery high depreciation expense (54% of total operating expenses as of 1Q25).\n\n4. Intensified Competition\n\nFierce competition with Oracle, Lambda Labs, Crusoe, Applied Digital, Nebius, etc.\n\nNvidia prioritizes GPU allocation to these Neoclouds, encouraging competition → supplier (Nvidia) has superior bargaining power.\n\nPrice competition is inevitable: Operating at prices 30~50% cheaper than Hyperscalers.\n\nKey Financial Estimates (HSBC Basis)\n\nRevenue\n2024: $1,915 million\n2025e: $5,010 million\n2026e: $11,385 million\n2027e: $18,054 million\n\nEBITDA\n2024: $1,188 million\n2025e: $2,819 million\n2026e: $6,891 million\n2027e: $10,983 million\n\nNet Income\n2024: -$863 million (Net Loss)\n2025e: -$624 million (Net Loss)\n2026e: -$305 million (Net Loss)\n2027e: $733 million (Expected to turn profitable)\n\nFree Cash Flow (FCF)\n2024: -$5,953 million\n2025e: -$17,739 million\n2026e: -$5,405 million\n2027e: -$4,505 million (FCF is projected to remain negative through 2027.)\n\nNet Debt\n2024: $6,153 million\n2025e: $22,665 million\n2026e: $28,070 million\n2027e: $32,575 million (Net debt is projected to increase significantly and continuously.)\n\nROE (Return on Equity)\n2024: -11.0%\n2025e: -21.2%\n2026e: -9.2%\n2027e: 27.4% (ROE is only expected to turn positive after 2027.)\n\nNet Debt/Equity\n2024: 4.7x\n2025e: 8.1x\n2026e: 11.0x\n2027e: 9.6x (This ratio remains very high, indicating substantial financial leverage.)\n\nHSBC Conclusion\nFCF positive conversion is difficult even in the long term.\n\nDepreciation and replacement costs of GPU assets will erode real profitability.\n\nThe stock price reflects overly optimistic expectations, and the DCF-based value is around $32.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0819150292890023, "ret_signed_m1d": -0.0819150292890023, "alpha_spy_m1d": -0.08799749645317656, "alpha_c_m1d": -0.08799749645317656, "bench_spy_m1d": -0.006082467164174266, "bench_c_m1d": -0.006082467164174266, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06951069967563106, "ret_signed_p1d": 0.06951069967563106, "alpha_spy_p1d": 0.0687782867735206, "alpha_c_p1d": 0.0687782867735206, "bench_spy_p1d": -0.0007324129021104575, "bench_c_p1d": -0.0007324129021104575, "ret_p1w": -0.09235312074593904, "ret_signed_p1w": 0.09235312074593904, "alpha_spy_p1w": 0.10251173572207273, "alpha_c_p1w": 0.10251173572207273, "bench_spy_p1w": 0.010158614976133684, "bench_c_p1w": 0.010158614976133684, "ret_p1m": -0.24385450310836787, "ret_signed_p1m": 0.24385450310836787, "alpha_spy_p1m": 0.26837529111918124, "alpha_c_p1m": 0.26837529111918124, "bench_spy_p1m": 0.024520788010813366, "bench_c_p1m": 0.024520788010813366, "ret_p3m": 0.01399282014053882, "ret_signed_p3m": -0.01399282014053882, "alpha_spy_p3m": 0.04336972306991127, "alpha_c_p3m": 0.04336972306991127, "bench_spy_p3m": 0.05736254321045009, "bench_c_p3m": 0.05736254321045009, "ret_p6m": -0.39384316375469375, "ret_signed_p6m": 0.39384316375469375, "alpha_spy_p6m": 0.5053183088088093, "alpha_c_p6m": 0.5053183088088093, "bench_spy_p6m": 0.1114751450541156, "bench_c_p6m": 0.1114751450541156, "price_path": [-0.7321, -0.7087, -0.6852, -0.6841, -0.6858, -0.6822, -0.6761, -0.6876, -0.6648, -0.6099, -0.6142, -0.5893, -0.5946, -0.584, -0.6115, -0.5576, -0.5215, -0.4898, -0.5025, -0.3926, -0.3451, -0.3174, -0.1877, -0.2424, -0.2229, -0.2229, -0.0623, -0.1215, -0.2016, -0.1581, -0.0908, 0.1382, 0.2336, 0.0215, 0.0601, 0.2261, 0.1716, 0.1323, 0.1268, 0.1133, 0.1989, 0.3004, 0.2858, 0.2858, 0.3885, 0.3137, 0.3056, 0.2064, 0.1957, 0.2101, 0.2333, 0.1795, 0.1479, 0.2495, 0.2495, 0.2079, 0.1455, 0.1576, 0.046, -0.0482, 0.0012, 0.0634, 0.0819, 0.0, -0.0695, -0.0552, -0.0185, -0.0466, -0.0924, -0.1255, -0.1659, -0.1775, -0.2218, -0.1368, -0.2123, -0.1982, -0.1541, -0.1662, -0.0842, -0.0201, 0.0573, 0.1251, -0.1093, -0.2474, -0.2439, -0.2678, -0.2974, -0.3078, -0.3133, -0.2891, -0.3013, -0.3088, -0.2668, -0.2225, -0.2206, -0.2206, -0.294, -0.3202, -0.3383, -0.3261, -0.2924, -0.242, -0.114, -0.1476, -0.1532, -0.0888, -0.1018, -0.0858, -0.0818, -0.0556, 0.0077, -0.01, 0.009, -0.042, -0.0898, -0.0733, 0.0351, 0.0366, 0.0438, 0.0195, 0.0124, -0.0256, 0.0588, 0.0822, 0.047, 0.0712, 0.014, 0.0532, 0.0721, 0.0352, -0.039, -0.0541, -0.0808, -0.0671, 0.0026, 0.0291, 0.0196, 0.0584, -0.0087, 0.0113, -0.0446, -0.1245, -0.1346, -0.1912, -0.2133, -0.2012, -0.3314, -0.3538, -0.4075, -0.4149, -0.4302, -0.4335, -0.4333, -0.4765, -0.4581, -0.4433, -0.4608, -0.4381, -0.4381, -0.4469, -0.4171, -0.4249, -0.3997, -0.3514, -0.3321, -0.3477, -0.3143, -0.3332, -0.3391, -0.4056, -0.4528, -0.4743, -0.5118, -0.4881, -0.3722, -0.3584, -0.3929, -0.4034, -0.4034, -0.422, -0.4333, -0.441, -0.4584, -0.4584, -0.4, -0.4187, -0.4105, -0.4162, -0.4169, -0.3938]}
{"unit_id": "orig:1969551230881185866", "ticker_idx": 0, "ts": "2025-09-20T23:56:20+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is rare for mainstream DRAM product lines to simultaneously reach new peaks, which may indicate a potential structural shift in the industry cycle", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish DRAM: DDR4/DDR5 at all-time highs driven by HBM supply shift, AI demand, and restocking; may signal a structural industry upcycle.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Long-Term Trend of DRAM Spot Prices (2000–2025)\n\n    • The current mainstream products, 16Gb DDR5 is priced around $7, and 16Gb DDR4 is around $10, reaching all-time high levels.\n\n    • This surpasses the previous price peaks seen in past DRAM cycles (early 2000s, 2017–2018).\n\n    • Compared to the 2023–2024 trough (DDR5 at around $4, DDR4 at $2.5–3.5), prices have more than doubled, confirming a strong upcycle.\n\n    • The sharp price increase is driven by:\n\n① Supply shortages of DDR5 due to memory makers shifting production to HBM\n\n② Server and PC replacement cycles alongside expanding AI-related demand\n\n③ Restocking of channel inventories\n\n    • From a historical perspective, it is rare for mainstream DRAM product lines to simultaneously reach new peaks, which may indicate a potential structural shift in the industry cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012327960751177628, "ret_signed_m1d": -0.012327960751177628, "alpha_spy_m1d": -0.00761922982660527, "alpha_c_m1d": 0.007723597003770085, "bench_spy_m1d": -0.004708730924572357, "bench_c_m1d": -0.020051557754947713, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01791158103942525, "ret_signed_p1d": 0.01791158103942525, "alpha_spy_p1d": 0.02335505587923281, "alpha_c_p1d": 0.01937042035330616, "bench_spy_p1d": -0.0054434748398075605, "bench_c_p1d": -0.0014588393138809108, "ret_p1w": 0.0009367612410294823, "ret_signed_p1w": 0.0009367612410294823, "alpha_spy_p1w": 0.005675500498897645, "alpha_c_p1w": -0.0005841260160522888, "bench_spy_p1w": -0.004738739257868163, "bench_c_p1w": 0.001520887257081771, "ret_p1m": 0.25571535433485715, "ret_signed_p1m": 0.25571535433485715, "alpha_spy_p1m": 0.249042182180639, "alpha_c_p1m": 0.1839210380404781, "bench_spy_p1m": 0.006673172154218143, "bench_c_p1m": 0.07179431629437905, "ret_p3m": 0.4596447512769175, "ret_signed_p3m": 0.4596447512769175, "alpha_spy_p3m": 0.445203563053697, "alpha_c_p3m": 0.38179767237625883, "bench_spy_p3m": 0.014441188223220491, "bench_c_p3m": 0.07784707890065867, "ret_p6m": 1.6406466633660866, "ret_signed_p6m": 1.6406466633660866, "alpha_spy_p6m": 1.6317508494457362, "alpha_c_p6m": 1.404917310223021, "bench_spy_p6m": 0.008895813920350326, "bench_c_p6m": 0.23572935314306553, "price_path": [-0.2281, -0.2284, -0.2356, -0.235, -0.2444, -0.2464, -0.2338, -0.2434, -0.2539, -0.2355, -0.2449, -0.2256, -0.2187, -0.2258, -0.2194, -0.2253, -0.2488, -0.2454, -0.2417, -0.2608, -0.2574, -0.2547, -0.2594, -0.2452, -0.2425, -0.228, -0.2346, -0.2679, -0.2589, -0.2502, -0.2599, -0.2436, -0.2294, -0.2129, -0.2024, -0.1977, -0.1983, -0.2072, -0.2167, -0.224, -0.2389, -0.2514, -0.2387, -0.2328, -0.2355, -0.2332, -0.2201, -0.2253, -0.246, -0.2369, -0.2317, -0.2165, -0.1968, -0.1909, -0.1672, -0.138, -0.1106, -0.0687, -0.0608, -0.031, -0.0472, 0.0001, -0.0123, 0.0, 0.0179, 0.0079, -0.0002, -0.0295, 0.0009, 0.0052, 0.0556, 0.1075, 0.116, 0.1224, 0.1117, 0.1337, 0.1252, 0.1529, 0.1588, 0.1371, 0.1711, 0.2318, 0.2447, 0.2734, 0.2557, 0.2548, 0.2602, 0.3243, 0.363, 0.3434, 0.3921, 0.4107, 0.4153, 0.5098, 0.4189, 0.4345, 0.4438, 0.4254, 0.4921, 0.4914, 0.4956, 0.4735, 0.4217, 0.4691, 0.3964, 0.3784, 0.3537, 0.2951, 0.3353, 0.346, 0.3764, 0.3985, 0.3858, 0.4027, 0.4301, 0.4181, 0.3957, 0.4323, 0.4877, 0.484, 0.5252, 0.4909, 0.4679, 0.4279, 0.387, 0.4132, 0.4596, 0.4849, 0.5547, 0.5619, 0.5851, 0.5851, 0.6154, 0.6864, 0.6949, 0.6803, 0.6803, 0.8005, 0.8506, 0.9457, 0.9615, 0.9406, 0.9666, 0.9721, 0.9412, 0.9462, 0.974, 1.0537, 1.0629, 1.031, 1.0943, 1.1371, 1.1518, 1.101, 1.2345, 1.3359, 1.3491, 1.3515, 1.2822, 1.3871, 1.3057, 1.2182, 1.2364, 1.2908, 1.2571, 1.3252, 1.4026, 1.4008, 1.4008, 1.3768, 1.4197, 1.4611, 1.5357, 1.5347, 1.6082, 1.657, 1.7671, 1.7184, 1.719, 1.4497, 1.3145, 1.4727, 1.3886, 1.2842, 1.4673, 1.5235, 1.4642, 1.4695, 1.5832, 1.6406]}
{"unit_id": "quote:1968193488731906253", "ticker_idx": 0, "ts": "2025-09-17T06:01:09+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB200 to Rubin and VR200 progressing smoothly overall... target of delivering just over 30,000 GB racks in 2025 remains unchanged... further expanding Nvidia's competitive advantage", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF Securities Buy on NVDA (GB200 yield improving, Rubin ahead of schedule) and Hold/downgrade on AMD (MI355 demand disappoints).", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "GF Securities’ comments on GB200/GB300 production:\n\nGB200 to Rubin and VR200 progressing smoothly overall: According to exchanges during SEMICON Taiwan, despite concerns about GB200 production issues, we observed that major ODMs’ first-pass yield has improved from 84–85% in June to around 90% in September, indicating that the target of delivering just over 30,000 GB racks in 2025 remains unchanged. We expect production to ramp to 11,000 units in 3Q25 and 16,000 units in 4Q25. In 2026, we anticipate GB300 will become the main product, with performance far surpassing competitors, consistent with our previous downgrade report on AMD (Hold), which cited lower-than-expected demand for the MI355.\n\nAs for Rubin, the next key milestones are the status of collaboration with Jentech (3653 TT) on the micro-channel lid and TSMC’s packaging progress in mid-November, both of which are currently proceeding smoothly. We expect Rubin packaging to be completed ahead of schedule by mid-2026, further expanding Nvidia’s competitive advantage.\n\n$NVDA $AMD", "reply_text": null, "parent_text": null, "quoted_text": "轉\n\n廣發海外電子通信》英偉達 (NVDA Buy) 季中更新，近期參訪要點", "parent_author": null, "ret_m1d": 0.026954090456944746, "ret_signed_m1d": 0.026954090456944746, "alpha_spy_m1d": 0.02571010959867448, "alpha_c_m1d": 0.02053312971971666, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.034940451119733984, "ret_signed_p1d": 0.034940451119733984, "alpha_spy_p1d": 0.030268005977403645, "alpha_c_p1d": -0.003454338781990085, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.03922716396777459, "ret_signed_p1w": 0.03922716396777459, "alpha_spy_p1w": 0.03353400461123068, "alpha_c_p1w": -0.012730208137761156, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.06764932027350357, "ret_signed_p1m": 0.06764932027350357, "alpha_spy_p1m": 0.06265584178852701, "alpha_c_p1m": -0.057002521668529704, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.03529165061072326, "ret_signed_p3m": 0.03529165061072326, "alpha_spy_p3m": -0.0002634924218634094, "alpha_c_p3m": -0.12080992183568728, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.07557761069639479, "ret_signed_p6m": 0.07557761069639479, "alpha_spy_p6m": 0.05934507050279625, "alpha_c_p6m": -0.19989400037546523, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.1553, -0.1534, -0.1315, -0.0939, -0.0897, -0.0737, -0.0723, -0.0998, -0.0766, -0.0644, -0.0644, -0.0708, -0.0605, -0.0436, -0.0364, -0.0316, -0.0366, 0.0024, 0.0063, 0.0159, 0.0124, 0.0063, -0.0192, 0.0028, 0.0202, 0.0188, 0.0379, 0.0306, 0.0527, 0.0445, 0.0201, 0.057, 0.0467, 0.0536, 0.0615, 0.0728, 0.0691, 0.0755, 0.0663, 0.0688, 0.0596, 0.0688, 0.0314, 0.0299, 0.0275, 0.0452, 0.0558, 0.0674, 0.0664, 0.058, 0.0228, 0.0228, 0.0028, 0.0019, 0.008, -0.0193, -0.0117, 0.0027, 0.0413, 0.0404, 0.0442, 0.0438, 0.027, 0.0, 0.0349, 0.0375, 0.0782, 0.0478, 0.0392, 0.0435, 0.0464, 0.0679, 0.0957, 0.0995, 0.1092, 0.1018, 0.0896, 0.0866, 0.1105, 0.1308, 0.0756, 0.1059, 0.0572, 0.056, 0.0676, 0.0759, 0.0725, 0.0638, 0.0587, 0.0697, 0.0938, 0.1245, 0.1805, 0.2158, 0.1914, 0.1891, 0.2149, 0.1668, 0.1463, 0.1045, 0.1049, 0.1689, 0.1343, 0.1381, 0.0973, 0.1167, 0.0958, 0.065, 0.0953, 0.0608, 0.0504, 0.072, 0.0442, 0.0585, 0.0585, 0.0394, 0.0566, 0.0656, 0.0546, 0.0769, 0.0712, 0.0897, 0.0863, 0.0793, 0.0625, 0.0278, 0.0353, 0.0437, 0.0039, 0.0227, 0.0629, 0.0787, 0.1112, 0.1076, 0.1076, 0.1189, 0.1054, 0.1014, 0.0953, 0.0953, 0.1091, 0.1048, 0.0996, 0.1106, 0.0867, 0.0856, 0.0861, 0.0912, 0.0755, 0.0985, 0.0937, 0.0937, 0.0457, 0.0766, 0.0855, 0.1021, 0.0951, 0.1071, 0.1247, 0.1305, 0.1224, 0.09, 0.0591, 0.023, 0.0094, 0.0889, 0.116, 0.1072, 0.1161, 0.0978, 0.0736, 0.0736, 0.0863, 0.1039, 0.1035, 0.1147, 0.1249, 0.1325, 0.1485, 0.0858, 0.0406, 0.0716, 0.0574, 0.0749, 0.0767, 0.0443, 0.0726, 0.0851, 0.0926, 0.0756]}
{"unit_id": "quote:1968193488731906253", "ticker_idx": 1, "ts": "2025-09-17T06:01:09+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "consistent with our previous downgrade report on AMD (Hold), which cited lower-than-expected demand for the MI355", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF Securities Buy on NVDA (GB200 yield improving, Rubin ahead of schedule) and Hold/downgrade on AMD (MI355 demand disappoints).", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "GF Securities’ comments on GB200/GB300 production:\n\nGB200 to Rubin and VR200 progressing smoothly overall: According to exchanges during SEMICON Taiwan, despite concerns about GB200 production issues, we observed that major ODMs’ first-pass yield has improved from 84–85% in June to around 90% in September, indicating that the target of delivering just over 30,000 GB racks in 2025 remains unchanged. We expect production to ramp to 11,000 units in 3Q25 and 16,000 units in 4Q25. In 2026, we anticipate GB300 will become the main product, with performance far surpassing competitors, consistent with our previous downgrade report on AMD (Hold), which cited lower-than-expected demand for the MI355.\n\nAs for Rubin, the next key milestones are the status of collaboration with Jentech (3653 TT) on the micro-channel lid and TSMC’s packaging progress in mid-November, both of which are currently proceeding smoothly. We expect Rubin packaging to be completed ahead of schedule by mid-2026, further expanding Nvidia’s competitive advantage.\n\n$NVDA $AMD", "reply_text": null, "parent_text": null, "quoted_text": "轉\n\n廣發海外電子通信》英偉達 (NVDA Buy) 季中更新，近期參訪要點", "parent_author": null, "ret_m1d": 0.008167900363446057, "ret_signed_m1d": -0.008167900363446057, "alpha_spy_m1d": -0.006923919505175791, "alpha_c_m1d": -0.0017469396262179693, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0077909365709525336, "ret_signed_p1d": 0.0077909365709525336, "alpha_spy_p1d": 0.012463381713282873, "alpha_c_p1d": 0.0461857264726766, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.01080674278165139, "ret_signed_p1w": -0.01080674278165139, "alpha_spy_p1w": -0.005113583425107482, "alpha_c_p1w": 0.04115062932388436, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.47373707063085835, "ret_signed_p1m": -0.47373707063085835, "alpha_spy_p1m": -0.4687435921458818, "alpha_c_p1m": -0.3490852286888251, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.3042221478691287, "ret_signed_p3m": -0.3042221478691287, "alpha_spy_p3m": -0.26866700483654204, "alpha_c_p3m": -0.14812057542271817, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.24239759326066967, "ret_signed_p6m": -0.24239759326066967, "alpha_spy_p6m": -0.22616505306707113, "alpha_c_p6m": 0.03307401781119035, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.1943, -0.1859, -0.1302, -0.099, -0.0973, -0.0964, -0.1084, -0.1448, -0.1297, -0.1335, -0.1335, -0.1531, -0.1341, -0.1304, -0.0942, -0.08, -0.0812, -0.0223, 0.0058, 0.0079, -0.0136, -0.0136, -0.0279, -0.0032, 0.0186, 0.0459, 0.0911, 0.1149, 0.1279, 0.1078, 0.0788, 0.1107, 0.0952, 0.0249, 0.0832, 0.0854, 0.0824, 0.0992, 0.1587, 0.1369, 0.1153, 0.1067, 0.0464, 0.0379, 0.0286, 0.054, 0.0264, 0.0469, 0.0501, 0.0592, 0.0218, 0.0218, 0.0199, 0.0187, 0.0165, -0.0504, -0.0487, -0.021, 0.0024, -0.0219, -0.0037, 0.0126, 0.0082, 0.0, -0.0078, -0.0111, 0.004, 0.0109, 0.0108, 0.0133, 0.0019, 0.0138, 0.0165, 0.0305, 0.0664, 0.0346, 0.2799, 0.3289, 0.48, 0.4632, 0.3502, 0.3598, 0.3703, 0.4991, 0.4737, 0.4644, 0.5114, 0.4955, 0.4465, 0.4764, 0.5891, 0.6315, 0.6211, 0.6608, 0.6012, 0.6092, 0.6314, 0.5711, 0.6105, 0.4935, 0.4673, 0.5329, 0.4923, 0.6266, 0.5579, 0.5507, 0.5112, 0.4469, 0.4046, 0.2944, 0.2803, 0.3512, 0.2951, 0.3461, 0.3461, 0.3667, 0.3807, 0.3523, 0.3672, 0.357, 0.3695, 0.3892, 0.3924, 0.3912, 0.3912, 0.3243, 0.3042, 0.3142, 0.2447, 0.2633, 0.341, 0.3505, 0.3502, 0.3511, 0.3511, 0.3508, 0.3547, 0.353, 0.3456, 0.3456, 0.4041, 0.389, 0.3468, 0.3196, 0.286, 0.2765, 0.3049, 0.3884, 0.4049, 0.432, 0.4566, 0.4566, 0.4571, 0.5695, 0.5942, 0.6316, 0.579, 0.5835, 0.588, 0.5844, 0.4874, 0.5473, 0.5212, 0.2578, 0.2095, 0.3096, 0.3571, 0.3419, 0.3419, 0.2939, 0.3026, 0.3026, 0.2759, 0.2574, 0.2778, 0.2575, 0.2352, 0.3436, 0.3248, 0.2797, 0.2579, 0.2479, 0.1997, 0.2696, 0.2531, 0.209, 0.2734, 0.2769, 0.2869, 0.2424]}
{"unit_id": "thread:1947811839360045353", "ticker_idx": 0, "ts": "2025-07-23T00:11:45+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "as mass production for its major CSP clients' products accelerates from 2026 onwards, there will be a significant contribution to MediaTek's revenue", "tweet_type": "analysis", "timeline": "years", "summary": "MediaTek winning Meta's Arke ASIC order signals growing CSP revenue from 2026+; Broadcom retains dominant position across major CSP clients.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: MediaTek Wins Major Order for Meta's New 2nm ASIC, Aiming for Mass Production in H1 2027\n\nMediaTek and Broadcom are in continuous competition for Meta's new Application-Specific Integrated Circuit (ASIC) projects, with industry sources emphasizing that their performances are neck and neck. However, recent reports suggest that MediaTek is poised to win a major order for one of Meta's upcoming 2nm ASICs, codenamed \"Arke.\" This chip, focused on post-training and inference, is expected to enter mass production in the first half of 2027.\n\nIC design industry sources reveal that if MediaTek wins the competition for this product, it would be the company's second major order from a Cloud Service Provider (CSP) client. MediaTek has traditionally refrained from commenting on any ASIC client-related news and maintains the same position on this matter.\n\nAccording to sources familiar with the ASIC industry, the \"Arke\" chip was not actually part of Meta's initial roadmap. Following \"Iris,\" which is scheduled for mass production at the end of 2025, the original plan was to launch another product, \"Olympus,\" using the N2P process.\n\nHowever, considering practical needs and cost-effectiveness, a chip specifically for inference, \"Arke,\" was added to the product launch schedule. \"Olympus\" will now be repositioned as a training-focused ASIC to compete with NVIDIA's future GPUs, and its launch schedule will also be pushed back to 2028.\n\nOf course, the timelines for both \"Arke\" in 2027 and \"Olympus\" in 2028 will ultimately depend on actual market conditions.\n\nMeta's previous products, and even the postponed \"Olympus,\" were reportedly all handled by the ASIC market leader, Broadcom. However, a cooperative relationship already exists between MediaTek and Meta. Meta's previously promoted proprietary smart glasses chip was a collaborative effort with MediaTek, establishing a working foundation between the two in the ASIC field.\n\nTherefore, it is not entirely surprising that MediaTek is expected to gain Meta's favor with the new \"Arke\" chip.\n\nIndustry insiders state that after securing Google as a major client, MediaTek indeed needs to broaden its collaborations to establish its influence in the cloud ASIC market.\n\nThe market has also recently observed that the design roadmaps and plans of major CSPs for ASICs are diverging from previous patterns. Although the demand for cloud AI remains immense and supply is extremely tight, the attitude of major CSPs towards cost-effectiveness has shifted.\n\nPreviously, the primary consideration was whether the technology met the requirements and could be integrated; technical specifications and launch speed were prioritized, with cost being a much lower consideration.\n\nThe situation is different now. As CSPs gain a clearer understanding of the actual cloud AI market dynamics and the technical details of chip design, they are beginning to seek the development of more practical and cost-effective products. It is in this environment that MediaTek's advantages in cost control are becoming more apparent.\n\nFor the publicly announced ASIC projects by major CSPs and key AI players, the furthest mass production plans currently extend to around 2028. The winners for most of these projects have already been decided, with only a very small number of deals yet to be finalized.\n\nCurrently, Broadcom remains the biggest winner, securing substantial projects from not only Google and Meta but also other key players like OpenAI, Apple, and ByteDance. However, MediaTek has successfully carved out a niche with Google and Meta, two clients where Broadcom previously held a dominant position. It is anticipated that as mass production for its major CSP clients' products accelerates from 2026 onwards, there will be a significant contribution to MediaTek's revenue.\n\n---\n\nhttps://t.co/c4Hwr2Em8F", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.008435773432739047, "alpha_c_m1d": 0.004326406696328222, "bench_spy_m1d": -0.008435773432739047, "bench_c_m1d": -0.004326406696328222, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.00033105252282705777, "alpha_c_p1d": -0.004151987545212288, "bench_spy_p1d": 0.00033105252282705777, "bench_c_p1d": 0.004151987545212288, "ret_p1w": -0.03496495889979201, "ret_signed_p1w": -0.03496495889979201, "alpha_spy_p1w": -0.03535909677584015, "alpha_c_p1w": -0.06863447722523142, "bench_spy_p1w": 0.00039413787604813955, "bench_c_p1w": 0.03366951832543941, "ret_p1m": -0.045454533588367596, "ret_signed_p1m": -0.045454533588367596, "alpha_spy_p1m": -0.04756731059494035, "alpha_c_p1m": -0.04842019320250501, "bench_spy_p1m": 0.0021127770065727525, "bench_c_p1m": 0.002965659614137417, "ret_p3m": -0.0629370130382636, "ret_signed_p3m": -0.0629370130382636, "alpha_spy_p3m": -0.1243537101395693, "alpha_c_p3m": -0.2744090722224306, "bench_spy_p3m": 0.061416697101305706, "bench_c_p3m": 0.211472059184167, "ret_p6m": 0.06215645585618468, "ret_signed_p6m": 0.06215645585618468, "alpha_spy_p6m": -0.035603422072686186, "alpha_c_p6m": -0.3252462090719157, "bench_spy_p6m": 0.09775987792887086, "bench_c_p6m": 0.3874026649281004, "price_path": [-0.0536, -0.0742, -0.0605, -0.0742, -0.0742, -0.1085, -0.1119, -0.1222, -0.1256, -0.1153, -0.0913, -0.0982, -0.0707, -0.0673, -0.0605, -0.0639, -0.1016, -0.0982, -0.081, -0.0913, -0.0982, -0.1222, -0.1222, -0.1222, -0.1359, -0.1359, -0.1359, -0.1256, -0.1256, -0.1462, -0.1222, -0.1187, -0.0845, -0.0982, -0.1187, -0.1427, -0.1393, -0.1256, -0.1187, -0.1325, -0.1427, -0.1393, -0.129, -0.1119, -0.1153, -0.1187, -0.1427, -0.1256, -0.1119, -0.1119, -0.0979, -0.1049, -0.1119, -0.0524, -0.021, -0.007, -0.0315, -0.0245, -0.014, -0.028, -0.014, -0.0035, 0.0, 0.0, 0.0, 0.0, -0.021, -0.049, -0.035, -0.042, -0.042, -0.0699, -0.0629, -0.0769, -0.0524, -0.0559, -0.0524, -0.049, -0.042, -0.028, -0.042, -0.014, -0.028, -0.0524, -0.0455, -0.0455, -0.014, -0.021, -0.021, -0.0315, -0.042, -0.049, -0.049, -0.0245, -0.0315, 0.0035, 0.0, 0.0559, 0.042, 0.035, 0.0385, 0.0385, 0.0734, 0.0559, 0.0559, 0.007, -0.0175, -0.0385, -0.0699, -0.0594, -0.0839, -0.0839, -0.0804, -0.1014, -0.1049, -0.0769, -0.0769, -0.0734, -0.0664, -0.0594, -0.0594, -0.0804, -0.0874, -0.0804, -0.0699, -0.0699, -0.0629, -0.0629, -0.0699, -0.0944, -0.0944, -0.0664, -0.0804, -0.0909, -0.0839, -0.0839, -0.0909, -0.0804, -0.0699, -0.0979, -0.1189, -0.1259, -0.1329, -0.1294, -0.1294, -0.1399, -0.1399, -0.1818, -0.1888, -0.1713, -0.1993, -0.1958, -0.1713, -0.0909, -0.0629, -0.0245, 0.0105, -0.0105, -0.021, -0.0175, -0.0035, 0.007, -0.007, 0.021, -0.0245, -0.0175, -0.007, -0.007, -0.0035, -0.007, -0.014, -0.021, -0.028, -0.035, -0.035, -0.0315, -0.007, -0.007, 0.0, 0.0, 0.028, 0.0664, 0.0586, 0.0622, 0.0301, 0.0123, 0.0301, 0.0586, 0.0693, 0.0622]}
{"unit_id": "thread:1947811839360045353", "ticker_idx": 1, "ts": "2025-07-23T00:11:45+00:00", "symbol": "Broadcom", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom remains the biggest winner, securing substantial projects from not only Google and Meta but also other key players like OpenAI, Apple, and ByteDance", "tweet_type": "analysis", "timeline": "years", "summary": "MediaTek winning Meta's Arke ASIC order signals growing CSP revenue from 2026+; Broadcom retains dominant position across major CSP clients.", "resolved_tickers": ["AVGO"], "resolution_method": "exchange_ticker", "resolution_reason": "Broadcom Inc. AVGO US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Report: MediaTek Wins Major Order for Meta's New 2nm ASIC, Aiming for Mass Production in H1 2027\n\nMediaTek and Broadcom are in continuous competition for Meta's new Application-Specific Integrated Circuit (ASIC) projects, with industry sources emphasizing that their performances are neck and neck. However, recent reports suggest that MediaTek is poised to win a major order for one of Meta's upcoming 2nm ASICs, codenamed \"Arke.\" This chip, focused on post-training and inference, is expected to enter mass production in the first half of 2027.\n\nIC design industry sources reveal that if MediaTek wins the competition for this product, it would be the company's second major order from a Cloud Service Provider (CSP) client. MediaTek has traditionally refrained from commenting on any ASIC client-related news and maintains the same position on this matter.\n\nAccording to sources familiar with the ASIC industry, the \"Arke\" chip was not actually part of Meta's initial roadmap. Following \"Iris,\" which is scheduled for mass production at the end of 2025, the original plan was to launch another product, \"Olympus,\" using the N2P process.\n\nHowever, considering practical needs and cost-effectiveness, a chip specifically for inference, \"Arke,\" was added to the product launch schedule. \"Olympus\" will now be repositioned as a training-focused ASIC to compete with NVIDIA's future GPUs, and its launch schedule will also be pushed back to 2028.\n\nOf course, the timelines for both \"Arke\" in 2027 and \"Olympus\" in 2028 will ultimately depend on actual market conditions.\n\nMeta's previous products, and even the postponed \"Olympus,\" were reportedly all handled by the ASIC market leader, Broadcom. However, a cooperative relationship already exists between MediaTek and Meta. Meta's previously promoted proprietary smart glasses chip was a collaborative effort with MediaTek, establishing a working foundation between the two in the ASIC field.\n\nTherefore, it is not entirely surprising that MediaTek is expected to gain Meta's favor with the new \"Arke\" chip.\n\nIndustry insiders state that after securing Google as a major client, MediaTek indeed needs to broaden its collaborations to establish its influence in the cloud ASIC market.\n\nThe market has also recently observed that the design roadmaps and plans of major CSPs for ASICs are diverging from previous patterns. Although the demand for cloud AI remains immense and supply is extremely tight, the attitude of major CSPs towards cost-effectiveness has shifted.\n\nPreviously, the primary consideration was whether the technology met the requirements and could be integrated; technical specifications and launch speed were prioritized, with cost being a much lower consideration.\n\nThe situation is different now. As CSPs gain a clearer understanding of the actual cloud AI market dynamics and the technical details of chip design, they are beginning to seek the development of more practical and cost-effective products. It is in this environment that MediaTek's advantages in cost control are becoming more apparent.\n\nFor the publicly announced ASIC projects by major CSPs and key AI players, the furthest mass production plans currently extend to around 2028. The winners for most of these projects have already been decided, with only a very small number of deals yet to be finalized.\n\nCurrently, Broadcom remains the biggest winner, securing substantial projects from not only Google and Meta but also other key players like OpenAI, Apple, and ByteDance. However, MediaTek has successfully carved out a niche with Google and Meta, two clients where Broadcom previously held a dominant position. It is anticipated that as mass production for its major CSP clients' products accelerates from 2026 onwards, there will be a significant contribution to MediaTek's revenue.\n\n---\n\nhttps://t.co/c4Hwr2Em8F", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017977499158006283, "ret_signed_m1d": -0.017977499158006283, "alpha_spy_m1d": -0.009541725725267236, "alpha_c_m1d": -0.01365109246167806, "bench_spy_m1d": -0.008435773432739047, "bench_c_m1d": -0.004326406696328222, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01769522975528459, "ret_signed_p1d": 0.01769522975528459, "alpha_spy_p1d": 0.017364177232457534, "alpha_c_p1d": 0.013543242210072304, "bench_spy_p1d": 0.00033105252282705777, "bench_c_p1d": 0.004151987545212288, "ret_p1w": 0.06672777274510944, "ret_signed_p1w": 0.06672777274510944, "alpha_spy_p1w": 0.0663336348690613, "alpha_c_p1w": 0.03305825441967003, "bench_spy_p1w": 0.00039413787604813955, "bench_c_p1w": 0.03366951832543941, "ret_p1m": 0.020832542189213932, "ret_signed_p1m": 0.020832542189213932, "alpha_spy_p1m": 0.01871976518264118, "alpha_c_p1m": 0.017866882575076515, "bench_spy_p1m": 0.0021127770065727525, "bench_c_p1m": 0.002965659614137417, "ret_p3m": 0.23317118800797543, "ret_signed_p3m": 0.23317118800797543, "alpha_spy_p3m": 0.17175449090666972, "alpha_c_p3m": 0.021699128823808422, "bench_spy_p3m": 0.061416697101305706, "bench_c_p3m": 0.211472059184167, "ret_p6m": 0.21352582180805402, "ret_signed_p6m": 0.21352582180805402, "alpha_spy_p6m": 0.11576594387918315, "alpha_c_p6m": -0.17387684312004636, "bench_spy_p6m": 0.09775987792887086, "bench_c_p6m": 0.3874026649281004, "price_path": [-0.3237, -0.3231, -0.3277, -0.3231, -0.3061, -0.2839, -0.2941, -0.2963, -0.2797, -0.2693, -0.2678, -0.2208, -0.1826, -0.1837, -0.1819, -0.196, -0.1889, -0.1853, -0.1921, -0.1893, -0.1957, -0.1957, -0.1713, -0.158, -0.1491, -0.1487, -0.1254, -0.0967, -0.0819, -0.0859, -0.1316, -0.1409, -0.1397, -0.1106, -0.0995, -0.1254, -0.1134, -0.123, -0.1164, -0.1164, -0.1188, -0.1055, -0.0702, -0.0671, -0.0477, -0.0505, -0.0283, -0.0668, -0.0486, -0.03, -0.03, -0.0335, -0.0419, -0.0204, -0.0292, -0.0328, -0.0285, -0.0097, -0.0102, 0.0097, -0.0012, 0.0159, -0.018, 0.0, 0.0177, 0.0229, 0.0374, 0.0484, 0.0667, 0.0353, 0.0174, 0.0495, 0.0326, 0.0634, 0.0707, 0.075, 0.0712, 0.1027, 0.0895, 0.0971, 0.0798, 0.0778, 0.0396, 0.0264, 0.0208, 0.0363, 0.0372, 0.0505, 0.0584, 0.088, 0.0483, 0.0483, 0.0513, 0.0659, 0.079, 0.1805, 0.2184, 0.1868, 0.3027, 0.2677, 0.2685, 0.2834, 0.269, 0.2202, 0.2173, 0.2159, 0.1963, 0.1968, 0.1981, 0.1868, 0.1812, 0.1578, 0.1649, 0.1772, 0.1941, 0.1948, 0.1846, 0.1879, 0.22, 0.2183, 0.1463, 0.2595, 0.2151, 0.2406, 0.2505, 0.2335, 0.2332, 0.2099, 0.2016, 0.2157, 0.2504, 0.2784, 0.317, 0.3629, 0.3293, 0.3052, 0.2802, 0.2427, 0.2676, 0.2556, 0.2338, 0.2655, 0.2428, 0.2543, 0.2005, 0.2092, 0.2099, 0.2023, 0.2515, 0.2246, 0.2013, 0.3346, 0.3595, 0.4038, 0.4038, 0.4229, 0.3633, 0.3473, 0.3439, 0.3454, 0.3779, 0.4163, 0.4346, 0.4582, 0.4349, 0.2709, 0.1999, 0.2051, 0.1512, 0.1648, 0.2018, 0.208, 0.2358, 0.239, 0.239, 0.2458, 0.2361, 0.2377, 0.2244, 0.2244, 0.2298, 0.2149, 0.2162, 0.2152, 0.1762, 0.2204, 0.246, 0.2545, 0.2025, 0.2135]}
{"unit_id": "orig:1960482919618843121", "ticker_idx": 0, "ts": "2025-08-26T23:22:06+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CLSA initiates coverage on Micron. Target at 2.4x 2026E BPS. Target price $155", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares CLSA initiation on Micron with $155 PT; bullish on HBM ramp, rising ASP through 2026, margin expansion from HBM mix.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250827_Micron: Riding the crest of the AI wave - CLSA\n\n[Key Takeaways]\n\n(1) CLSA initiates coverage on Micron. Target at 2.4x 2026E BPS. Target price $155\n\n(2) Expects ASP to rise through 2026, driven by strong HBM growth based on robust AI CAPEX and limited supply of commodity DRAM products\n\n(3) Forecasts mass production of HBM4 to begin in the second half of 2Q26, with margin improvement expected from the increasing revenue mix of HBM\n\n[Contents]\n\n(1) Global AI infrastructure investment growth is driving HBM demand\n\n(2) In 2025, Micron’s HBM3E 12-hi output is ramping faster than expected, proving strong execution by supplying to four major customers\n\n(3) Micron is currently undergoing HBM4 qualification testing, with mass production scheduled to begin in the second half of 2Q26\n\n(4) We expect Micron’s HBM revenue mix within DRAM to rise from 21% in 2025 to 31% in 2026 and 39% in 2027. Accordingly, margin improvements will also appear\n\n■ HBM Rev: ’24(A) $0.6bn → ’25(F) $7.8bn → ’26(F) $13.7bn → ’27(F) $20.1bn\n■ HBM Mix: ’24(A) 3% → ’25(F) 21% → ’26(F) 31% → ’27(F) 39%\n■ HBM M/S: ’24(A) 6% → ’25(F) 21% → ’26(F) 25% → ’27(F) 28%\n■ TSV CAPA: ’24(A) 25K → ’25(F) 65K → ’26(F) 95K\n(5) Micron has secured about 25% market share within NVIDIA. However, Samsung Electronics is expected to expand its share, and Micron’s is projected to decline to around 20% in 2026\n\n(6) Nevertheless, Micron is expected to offset this through transactions with major ASIC customers\n\n(7) Meanwhile, we forecast global HBM revenue to grow at a CAGR of +35% through 2030, reaching around $105bn, or about 46% of total DRAM revenue mix\n\n(8) As for commodity products, due to low inventory levels at both memory suppliers and customers, we forecast strong DRAM ASP from 2H25 through 2026\n\n(9) In 1H26, DRAM ASP may weaken slightly due to seasonal low demand, but the downtrend is expected to be short-lived, given tight supply and low inventories\n\n(10) All memory IDMs remain focused on HBM CAPA. Since HBM products consume 3–4x more wafer input than standard products, supply remains tight, which will continue to support ASP for commodity products\n\n(11) For NAND, suppliers’ restrained CAPEX stance and controlled supply are leading to a gradual recovery\n\n(12) Demand for data center eSSD is rebounding, and content growth is expected to increase as AI features are increasingly adopted in smartphones and PCs\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0006867236402662291, "ret_signed_m1d": -0.0006867236402662291, "alpha_spy_m1d": 0.0034828545342441286, "alpha_c_m1d": 0.008849842273760178, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010729638785070383, "ret_signed_p1d": 0.010729638785070383, "alpha_spy_p1d": 0.008451125503061041, "alpha_c_p1d": 0.007663124010227618, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": 0.016995737866876803, "ret_signed_p1w": 0.016995737866876803, "alpha_spy_p1w": 0.024575262579042967, "alpha_c_p1w": 0.05160394536901658, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": 0.38806870305705155, "ret_signed_p1m": 0.38806870305705155, "alpha_spy_p1m": 0.36052075047516263, "alpha_c_p1m": 0.30597944566921753, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.7811146604839896, "ret_signed_p3m": 0.7811146604839896, "alpha_spy_p3m": 0.7567840274270952, "alpha_c_p3m": 0.6821087142760189, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 2.61702817540985, "ret_signed_p6m": 2.61702817540985, "alpha_spy_p6m": 2.54717536191058, "alpha_c_p6m": 2.2224749270818824, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.1699, -0.1899, -0.158, -0.1231, -0.1146, -0.0885, -0.069, -0.0485, -0.0212, -0.005, -0.0037, -0.0087, 0.0277, 0.032, 0.0447, 0.0447, 0.0599, 0.0469, 0.0969, 0.0912, 0.0805, 0.0699, 0.0569, 0.0367, 0.044, 0.0487, 0.0487, 0.0294, 0.068, 0.0493, 0.0567, 0.0689, 0.0181, 0.031, -0.0006, -0.0278, -0.0181, -0.0281, -0.0625, -0.0573, -0.0409, -0.045, -0.0451, -0.039, -0.0151, -0.0632, -0.0997, -0.0749, -0.0639, -0.0663, -0.0397, 0.0205, 0.062, 0.0966, 0.0667, 0.0755, 0.0375, 0.0605, 0.0476, 0.0061, -0.0061, 0.0101, -0.0007, 0.0, 0.0107, 0.0472, 0.0215, 0.0215, 0.017, 0.0191, 0.0662, 0.1276, 0.1284, 0.1609, 0.2017, 0.2924, 0.3496, 0.3542, 0.3633, 0.3733, 0.4497, 0.3968, 0.413, 0.4284, 0.3881, 0.3462, 0.35, 0.4069, 0.4362, 0.5635, 0.5773, 0.6133, 0.6402, 0.5949, 0.6881, 0.6519, 0.5598, 0.6557, 0.6067, 0.6486, 0.7395, 0.7383, 0.776, 0.7375, 0.7047, 0.7754, 0.8812, 0.8905, 0.906, 0.9465, 0.924, 0.922, 1.0159, 0.8727, 1.0399, 1.047, 1.0435, 1.1756, 1.0709, 1.1035, 1.0352, 1.12, 1.0781, 0.9626, 0.9404, 0.7296, 0.7811, 0.9233, 0.9285, 0.9777, 0.9777, 1.0311, 1.0653, 1.057, 1.0112, 0.9467, 1.0375, 1.1208, 1.1681, 1.265, 1.2199, 1.0712, 1.0399, 0.997, 0.937, 1.1348, 1.284, 1.3756, 1.3729, 1.4623, 1.4623, 1.4461, 1.5294, 1.5144, 1.4524, 1.4524, 1.7103, 1.6822, 1.9509, 1.9176, 1.8099, 1.9652, 1.9719, 1.9054, 1.8643, 1.8925, 2.1169, 2.1169, 2.1363, 2.3434, 2.4162, 2.434, 2.3433, 2.525, 2.7402, 2.7445, 2.5649, 2.7618, 2.6041, 2.26, 2.29, 2.3914, 2.2952, 2.2072, 2.5259, 2.5571, 2.5372, 2.5372, 2.4351, 2.617]}
{"unit_id": "orig:1948531955068535219", "ticker_idx": 0, "ts": "2025-07-24T23:53:14+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi believes Amazon will deliver better-than-expected revenue and operating profit performance in its Q2 2025 earnings report, with a key focus on the uplift from AWS infrastructure deployment for second-half growth", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi Buy-thesis on Amazon: above-consensus Q2 + H2 AWS re-acceleration, Prime Day retail rebound, ad margin expansion, and OP beats drive sustained upside.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Amazon Outlook: AWS Growth to Drive H2 Acceleration, Retail and Advertising Margins Continue to Improve\nCiti (R. Josey, 25/07/22)\n\nCiti believes Amazon will deliver better-than-expected revenue and operating profit performance in its Q2 2025 earnings report, with a key focus on the uplift from AWS infrastructure deployment for second-half growth. Meanwhile, Prime Day is expected to drive a retail rebound, strong advertising demand, and improved operational efficiency, all contributing to margin expansion.\n\nAWS Infrastructure Deployment Lays Foundation for H2 Growth\n\nCiti expects Q2 2025 AWS revenue to grow 16.5% year-over-year, slightly below buyer expectations of 17%, but emphasizes that management's commentary on second-half growth is more crucial. Especially with sustained GenAI demand and the Anthropic contract expanding to $4 billion ARR, AWS is poised for re-acceleration. A significant portion of the approximately $74 billion in 2025 CapEx will also be allocated to AWS infrastructure construction.\n\nRetail Demand Rebounds, Prime Day Provides Additional Growth Momentum\n\nPrime Day's contribution to retail growth is over 500 basis points, with total retail revenue expected to grow +9% year-over-year in North America and +11% internationally. Citi credit card data shows continued acceleration in \"multi-category store\" spending in Q2, with year-over-year growth rising to 13.4% in mid-July, confirming the trend of rebounding consumer momentum.\n\nAdvertising Business Continues Strong Growth, Driving High-Margin Structure Improvement\n\nQ2 2025 advertising revenue is expected to grow 17.5% year-over-year, excluding currency effects. Channel checks indicate strong demand for Sponsored Listings, Prime Video, and DSP advertising. Notably, partnerships with Roku and Disney have expanded advertising distribution channels, with the advertising business providing a substantial boost to overall margins.\n\nOperating Profit Continues to Exceed Upper Guidance, Operational Efficiency Continues to Improve\n\nCiti forecasts Q2 2025 operating profit at $17.1 billion, with a margin of 10.5%, close to the upper end of guidance. Amazon has exceeded the high end of its guidance by over $1.2 billion in six of the past eight quarters. Operating cost control is good, with automation deployment (such as Vulcan robots) improving efficiency. Coupled with the increasing proportion of advertising and AWS, these factors jointly drive continuous improvement in profitability.\n\nNorth American Retail Margins Gradually Expanding, Long-Term Trend Optimistic\n\nWhile Citi conservatively estimates that North American retail (excluding advertising) margins will only reach 5% after 2030, there has been positive progress, including $4 billion investments in suburban fulfillment centers and robot deployments exceeding the number of warehouse employees. These measures are expected to drive continuous upward operational efficiency. Even with additional costs from the Kuiper project, overall retail margins are likely to continue improving.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016965949764856436, "ret_signed_m1d": -0.016965949764856436, "alpha_spy_m1d": -0.016635006801532404, "alpha_c_m1d": -0.033797173951574555, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": 0.01683122418671812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00340177109187767, "ret_signed_p1d": -0.00340177109187767, "alpha_spy_p1d": -0.007626120732196817, "alpha_c_p1d": -0.012692623925518287, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": 0.009290852833640617, "ret_p1w": 0.008095443816044323, "ret_signed_p1w": 0.008095443816044323, "alpha_spy_p1w": 0.011783745414039926, "alpha_c_p1w": 0.01424441211134142, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": -0.006148968295297097, "ret_p1m": -0.01459759485000256, "ret_signed_p1m": -0.01459759485000256, "alpha_spy_p1m": -0.03176296892130548, "alpha_c_p1m": -0.060827443614133614, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.046229848764131054, "ret_p3m": -0.04392196157213557, "ret_signed_p3m": -0.04392196157213557, "alpha_spy_p3m": -0.10497147863408385, "alpha_c_p3m": -0.11928930575565744, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.07536734418352187, "ret_p6m": 0.029668860682975806, "ret_signed_p6m": 0.029668860682975806, "alpha_spy_p6m": -0.06680824128510143, "alpha_c_p6m": -0.07229267322179722, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.10196153390477303, "price_path": [-0.1917, -0.1931, -0.2059, -0.181, -0.1819, -0.1976, -0.2033, -0.1874, -0.1729, -0.1687, -0.1016, -0.0898, -0.0946, -0.1165, -0.1147, -0.1123, -0.1213, -0.134, -0.1254, -0.1345, -0.1345, -0.1129, -0.1185, -0.1142, -0.1172, -0.1101, -0.1142, -0.1077, -0.1047, -0.0804, -0.0657, -0.063, -0.0819, -0.0818, -0.0867, -0.0695, -0.075, -0.0849, -0.0849, -0.0971, -0.1023, -0.0838, -0.0872, -0.0651, -0.0385, -0.0553, -0.0507, -0.053, -0.038, -0.038, -0.0377, -0.0554, -0.0417, -0.0429, -0.031, -0.0282, -0.0253, -0.0389, -0.036, -0.0263, -0.0126, -0.0205, -0.017, 0.0, -0.0034, 0.0024, -0.0053, -0.0088, 0.0081, -0.0753, -0.0886, -0.0796, -0.0427, -0.0392, -0.0411, -0.0471, -0.0463, -0.033, -0.0054, -0.0052, -0.0032, -0.0182, -0.0363, -0.0443, -0.0146, -0.0185, -0.0152, -0.0134, -0.0027, -0.0139, -0.0139, -0.0297, -0.0269, 0.0149, 0.0004, 0.0155, 0.0259, -0.0082, -0.0098, -0.0176, -0.0034, 0.0078, -0.0026, -0.0043, -0.0032, -0.0198, -0.0496, -0.0518, -0.0606, -0.0536, -0.0433, -0.0545, -0.05, -0.0423, -0.0548, -0.0488, -0.045, -0.0302, -0.0193, -0.0683, -0.0524, -0.0682, -0.0717, -0.0765, -0.0826, -0.0678, -0.0439, -0.0615, -0.048, -0.0345, -0.0226, -0.0128, -0.0083, -0.0403, 0.0516, 0.0937, 0.0736, 0.0774, 0.0465, 0.0524, 0.0696, 0.0726, 0.0515, 0.023, 0.0106, 0.0028, -0.0417, -0.0411, -0.065, -0.0497, -0.0256, -0.011, -0.0132, -0.0132, 0.0043, 0.0071, 0.0094, 0.0006, -0.0134, -0.0116, -0.023, -0.0186, -0.0019, -0.0084, -0.026, -0.0417, -0.0416, -0.0472, -0.0236, -0.021, -0.0164, -0.0004, 0.0006, 0.0006, 0.0012, -0.0007, 0.0013, -0.0061, -0.0061, -0.0247, 0.0036, 0.0375, 0.0402, 0.0605, 0.0652, 0.0613, 0.0447, 0.019, 0.0256, 0.0297]}
{"unit_id": "thread:1960261461600878690", "ticker_idx": 0, "ts": "2025-08-26T08:42:06+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's cooperation with Intel perfectly aligns with Trump's 'Save Intel' policy… closer collaboration with Intel could further accelerate this momentum", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Samsung–Intel foundry/packaging alliance aligns with Trump policy and could accelerate both companies' pursuit of TSMC, bullish for both names.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung–Intel Foundry Alliance Pursuit: From Packaging to Glass Substrates to Catch Up with TSMC\n\nAs Samsung Electronics Chairman Lee Jae-yong visits the United States as part of the economic delegation to the Korea–U.S. summit, speculation is growing that Samsung is actively pursuing a strategic alliance with Intel to strengthen its foundry business.\n\nIt is reported that Samsung is considering investing in the packaging sector, where Intel holds relative strengths, and utilizing Intel’s packaging production lines in the U.S. Samsung is also said to be reviewing ways to leverage Intel’s semiconductor glass substrate technology.\n\nSamsung Electronics has already invested $37 billion (about ₩51 trillion) to build a foundry plant in Taylor, Texas, and has been reviewing additional U.S. investments. Analysts suggest the company may be pushing for packaging production line investments in cooperation with Intel.\n\nAccording to semiconductor industry reports compiled on the 26th, Samsung Electronics is expected to tighten its pursuit of Taiwan’s TSMC, the global foundry market leader, by strengthening post-process packaging cooperation with Intel.\n\nAn industry insider said, “Samsung Electronics is deliberating the possibility of a strategic partnership with Intel, most likely in the packaging field,” adding, “Intel has strengths in packaging and is also ahead in glass substrate technology.”\n\nFoundry processes are divided into the front-end (fabricating circuits on wafers) and the back-end (packaging and testing the fabricated chips). While Samsung Foundry is considered ahead of Intel in front-end technology, Intel holds advantages in back-end packaging.\n\nSamsung Foundry has mostly outsourced wafer packaging and testing to OSAT (Outsourced Semiconductor Assembly and Test) companies after front-end wafer production. TSMC also collaborates with Taiwanese OSAT companies, but it possesses the industry’s leading back-end packaging technology.\n\nThe “Chip-on-Wafer-on-Substrate (CoWoS)” technology, which became famous for use in Nvidia AI semiconductors, is one of TSMC’s flagship packaging technologies. Intel is also evaluated as having strengths in post-process packaging, alongside TSMC.\n\nA source familiar with Intel noted, “The core of the Samsung–Intel cooperation will be packaging,” adding, “Intel Foundry is applying hybrid bonding packaging technology to its 18A (2nm-class) process production. While it lags in front-end node technology, it has a clear advantage in packaging.”\n\nHybrid bonding is a packaging technology that stacks chips through copper interconnections without “bumps” between chips. Although it is not yet used in memory semiconductors such as HBM (High Bandwidth Memory), Intel and TSMC already use this technology in packaging CPUs and image sensors (CIS).\n\nIn market share calculations that only account for front-end foundry processes, Samsung ranks second after TSMC. However, when including both front-end and back-end processes, Intel surpasses Samsung.\n\nAccording to Counterpoint Research’s Q1 2025 revenue-based integrated market share survey (front- and back-end combined), TSMC ranks first with a 35.3% share, Intel is second at 6.5%, and Samsung is fourth with 5.9%, behind Taiwanese back-end firm ASE.\n\nObservers believe Samsung and Intel’s cooperation could extend to next-generation semiconductor “glass substrates.”\n\nGlass substrates are drawing attention for high-performance AI semiconductors due to smoother surfaces, lower coefficients of thermal expansion, higher thermal stability, thinner thickness, and improved electrical performance compared to conventional plastic substrates.\n\nAn industry official said, “Intel has recently shifted its glass core substrate (GCS) technology into a licensing model to generate revenue,” adding, “After more than a decade of developing glass substrates, Intel’s move to open up the technology makes cooperation with Samsung all the more feasible.”\n\nSamsung Electro-Mechanics recently hired Kang Doo-an, a former senior engineer at Intel who worked extensively on glass substrates, as Executive Vice President overseeing technology marketing and application engineering. Kang is tasked with analyzing packaging technology for the semiconductor glass substrate market and establishing R&D strategies at Samsung Electro-Mechanics.\n\nSamsung–Intel foundry cooperation is seen as a potential Korea–U.S. semiconductor collaboration model aimed at catching up with TSMC, which controls more than 70% of the global market.\n\nThe cooperation between Samsung and Intel could take the form of equity investments or a joint venture (JV), where both sides contribute capital, technology, and manpower for specific business purposes.\n\nSingapore-based investment media outlet Smartkarma reported, “From a political standpoint, Samsung’s cooperation with Intel perfectly aligns with Trump’s ‘Save Intel’ policy,” adding, “Samsung’s 2nm AI6 order from Tesla was already a groundbreaking step, and closer collaboration with Intel could further accelerate this momentum.”\n\n$INTC\n\n---\n\nhttps://t.co/msTikUVXWd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00821350523439901, "ret_signed_m1d": 0.00821350523439901, "alpha_spy_m1d": 0.012383083408909368, "alpha_c_m1d": 0.017750071148425417, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.009536565914026407, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02053388058180472, "ret_signed_p1d": 0.02053388058180472, "alpha_spy_p1d": 0.018255367299795378, "alpha_c_p1d": 0.017467365806961954, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.0030665147748427657, "ret_p1w": -0.005749539827671235, "ret_signed_p1w": -0.005749539827671235, "alpha_spy_p1w": 0.0018299848844949285, "alpha_c_p1w": 0.02885866767446854, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.034608207502139776, "ret_p1m": 0.28213547532889627, "ret_signed_p1m": 0.28213547532889627, "alpha_spy_p1m": 0.25458752274700736, "alpha_c_p1m": 0.20004621794106225, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.08208925738783401, "ret_p3m": 0.4168377601445292, "ret_signed_p3m": 0.4168377601445292, "alpha_spy_p3m": 0.3925071270876348, "alpha_c_p3m": 0.31783181393655857, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.09900594620797065, "ret_p6m": 0.8669403848990316, "ret_signed_p6m": 0.8669403848990316, "alpha_spy_p6m": 0.7970875713997612, "alpha_c_p6m": 0.4723871365710641, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.39455324832796745, "price_path": [-0.1684, -0.1971, -0.1893, -0.1667, -0.1684, -0.1791, -0.1762, -0.1589, -0.0932, -0.1507, -0.147, -0.1729, -0.1483, -0.1458, -0.1175, -0.1175, -0.1343, -0.1298, -0.0739, -0.0883, -0.076, -0.0682, -0.0801, -0.0616, -0.1014, -0.0764, -0.0764, -0.0965, -0.0312, -0.0374, -0.0218, -0.0378, -0.0431, -0.0587, -0.0682, -0.0637, -0.0513, -0.0448, -0.0456, -0.0353, -0.0706, -0.1499, -0.1507, -0.1618, -0.1647, -0.1869, -0.207, -0.1992, -0.1708, -0.1618, -0.1881, -0.1807, -0.152, -0.1043, -0.0875, -0.0201, 0.0086, -0.0283, 0.0394, -0.0333, -0.0349, 0.0185, 0.0082, 0.0, 0.0205, 0.0238, 0.0, 0.0, -0.0057, -0.0144, 0.0107, 0.0057, 0.0053, 0.0037, 0.0172, 0.0107, -0.0111, 0.0172, 0.0378, 0.0226, 0.2554, 0.2148, 0.1811, 0.2049, 0.2821, 0.3959, 0.4579, 0.416, 0.3778, 0.476, 0.5318, 0.5125, 0.5027, 0.5265, 0.5372, 0.5524, 0.4936, 0.5285, 0.4632, 0.5257, 0.5129, 0.5199, 0.5647, 0.5655, 0.5162, 0.5671, 0.5721, 0.6238, 0.7055, 0.6977, 0.6493, 0.6423, 0.6222, 0.5207, 0.5762, 0.5294, 0.5659, 0.5791, 0.5556, 0.5561, 0.4747, 0.4587, 0.4255, 0.4099, 0.4419, 0.3807, 0.4168, 0.4698, 0.4715, 0.5117, 0.5117, 0.6657, 0.6431, 0.7852, 0.7971, 0.6632, 0.7006, 0.655, 0.6632, 0.6747, 0.6226, 0.5528, 0.5405, 0.5322, 0.4805, 0.4899, 0.5121, 0.4936, 0.4928, 0.485, 0.485, 0.4867, 0.5064, 0.5318, 0.5154, 0.5154, 0.6172, 0.6168, 0.6444, 0.7507, 0.6883, 0.8706, 0.8094, 0.9421, 1.0008, 0.9844, 0.9285, 0.9285, 0.9943, 1.2279, 1.2308, 0.8509, 0.745, 0.8041, 1.0033, 0.9984, 0.9084, 1.0045, 1.0226, 0.9959, 0.9811, 1.0776, 1.0632, 0.9355, 0.9832, 0.9088, 0.9216, 0.9216, 0.8965, 0.8669]}
{"unit_id": "thread:1960261461600878690", "ticker_idx": 1, "ts": "2025-08-26T08:42:06+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung–Intel foundry cooperation is seen as a potential Korea–U.S. semiconductor collaboration model aimed at catching up with TSMC", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Samsung–Intel foundry/packaging alliance aligns with Trump policy and could accelerate both companies' pursuit of TSMC, bullish for both names.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung–Intel Foundry Alliance Pursuit: From Packaging to Glass Substrates to Catch Up with TSMC\n\nAs Samsung Electronics Chairman Lee Jae-yong visits the United States as part of the economic delegation to the Korea–U.S. summit, speculation is growing that Samsung is actively pursuing a strategic alliance with Intel to strengthen its foundry business.\n\nIt is reported that Samsung is considering investing in the packaging sector, where Intel holds relative strengths, and utilizing Intel’s packaging production lines in the U.S. Samsung is also said to be reviewing ways to leverage Intel’s semiconductor glass substrate technology.\n\nSamsung Electronics has already invested $37 billion (about ₩51 trillion) to build a foundry plant in Taylor, Texas, and has been reviewing additional U.S. investments. Analysts suggest the company may be pushing for packaging production line investments in cooperation with Intel.\n\nAccording to semiconductor industry reports compiled on the 26th, Samsung Electronics is expected to tighten its pursuit of Taiwan’s TSMC, the global foundry market leader, by strengthening post-process packaging cooperation with Intel.\n\nAn industry insider said, “Samsung Electronics is deliberating the possibility of a strategic partnership with Intel, most likely in the packaging field,” adding, “Intel has strengths in packaging and is also ahead in glass substrate technology.”\n\nFoundry processes are divided into the front-end (fabricating circuits on wafers) and the back-end (packaging and testing the fabricated chips). While Samsung Foundry is considered ahead of Intel in front-end technology, Intel holds advantages in back-end packaging.\n\nSamsung Foundry has mostly outsourced wafer packaging and testing to OSAT (Outsourced Semiconductor Assembly and Test) companies after front-end wafer production. TSMC also collaborates with Taiwanese OSAT companies, but it possesses the industry’s leading back-end packaging technology.\n\nThe “Chip-on-Wafer-on-Substrate (CoWoS)” technology, which became famous for use in Nvidia AI semiconductors, is one of TSMC’s flagship packaging technologies. Intel is also evaluated as having strengths in post-process packaging, alongside TSMC.\n\nA source familiar with Intel noted, “The core of the Samsung–Intel cooperation will be packaging,” adding, “Intel Foundry is applying hybrid bonding packaging technology to its 18A (2nm-class) process production. While it lags in front-end node technology, it has a clear advantage in packaging.”\n\nHybrid bonding is a packaging technology that stacks chips through copper interconnections without “bumps” between chips. Although it is not yet used in memory semiconductors such as HBM (High Bandwidth Memory), Intel and TSMC already use this technology in packaging CPUs and image sensors (CIS).\n\nIn market share calculations that only account for front-end foundry processes, Samsung ranks second after TSMC. However, when including both front-end and back-end processes, Intel surpasses Samsung.\n\nAccording to Counterpoint Research’s Q1 2025 revenue-based integrated market share survey (front- and back-end combined), TSMC ranks first with a 35.3% share, Intel is second at 6.5%, and Samsung is fourth with 5.9%, behind Taiwanese back-end firm ASE.\n\nObservers believe Samsung and Intel’s cooperation could extend to next-generation semiconductor “glass substrates.”\n\nGlass substrates are drawing attention for high-performance AI semiconductors due to smoother surfaces, lower coefficients of thermal expansion, higher thermal stability, thinner thickness, and improved electrical performance compared to conventional plastic substrates.\n\nAn industry official said, “Intel has recently shifted its glass core substrate (GCS) technology into a licensing model to generate revenue,” adding, “After more than a decade of developing glass substrates, Intel’s move to open up the technology makes cooperation with Samsung all the more feasible.”\n\nSamsung Electro-Mechanics recently hired Kang Doo-an, a former senior engineer at Intel who worked extensively on glass substrates, as Executive Vice President overseeing technology marketing and application engineering. Kang is tasked with analyzing packaging technology for the semiconductor glass substrate market and establishing R&D strategies at Samsung Electro-Mechanics.\n\nSamsung–Intel foundry cooperation is seen as a potential Korea–U.S. semiconductor collaboration model aimed at catching up with TSMC, which controls more than 70% of the global market.\n\nThe cooperation between Samsung and Intel could take the form of equity investments or a joint venture (JV), where both sides contribute capital, technology, and manpower for specific business purposes.\n\nSingapore-based investment media outlet Smartkarma reported, “From a political standpoint, Samsung’s cooperation with Intel perfectly aligns with Trump’s ‘Save Intel’ policy,” adding, “Samsung’s 2nm AI6 order from Tesla was already a groundbreaking step, and closer collaboration with Intel could further accelerate this momentum.”\n\n$INTC\n\n---\n\nhttps://t.co/msTikUVXWd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01706974029297892, "ret_signed_m1d": 0.01706974029297892, "alpha_spy_m1d": 0.021239318467489277, "alpha_c_m1d": 0.02163207840389092, "bench_spy_m1d": -0.004169578174510358, "bench_c_m1d": -0.004562338110912001, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004267322671752005, "ret_signed_p1d": 0.004267322671752005, "alpha_spy_p1d": 0.001988809389742663, "alpha_c_p1d": -0.0011312033402968158, "bench_spy_p1d": 0.0022785132820093423, "bench_c_p1d": 0.005398526012048821, "ret_p1w": -0.01706974029297903, "ret_signed_p1w": -0.01706974029297903, "alpha_spy_p1w": -0.009490215580812866, "alpha_c_p1w": -0.004789613851303387, "bench_spy_p1w": -0.0075795247121661635, "bench_c_p1w": -0.012280126441675643, "ret_p1m": 0.21479374494685044, "ret_signed_p1m": 0.21479374494685044, "alpha_spy_p1m": 0.18724579236496153, "alpha_c_p1m": 0.15465648105664287, "bench_spy_p1m": 0.027547952581888913, "bench_c_p1m": 0.06013726389020757, "ret_p3m": 0.3545228506052651, "ret_signed_p3m": 0.3545228506052651, "alpha_spy_p3m": 0.3301922175483707, "alpha_c_p3m": 0.3145231767861125, "bench_spy_p3m": 0.024330633056894424, "bench_c_p3m": 0.03999967381915259, "ret_p6m": 1.6016103312258427, "ret_signed_p6m": 1.6016103312258427, "alpha_spy_p6m": 1.5317575177265723, "alpha_c_p6m": 1.5271694547080097, "bench_spy_p6m": 0.06985281349927042, "bench_c_p6m": 0.07444087651783304, "price_path": [-0.2069, -0.2054, -0.197, -0.197, -0.1828, -0.1644, -0.1644, -0.1545, -0.163, -0.1531, -0.1588, -0.1758, -0.1913, -0.1786, -0.1545, -0.163, -0.1588, -0.18, -0.1446, -0.1333, -0.1489, -0.1351, -0.1494, -0.1437, -0.1351, -0.0925, -0.0996, -0.1223, -0.1266, -0.1408, -0.1323, -0.1095, -0.111, -0.0939, -0.0797, -0.0512, -0.0455, -0.0356, -0.0612, -0.0555, -0.0612, -0.0626, 0.0014, 0.0043, 0.0327, 0.0156, -0.0199, -0.0085, -0.0057, -0.0213, 0.0028, 0.0213, 0.01, 0.0114, 0.0228, 0.0185, 0.0185, -0.0043, -0.0043, 0.0028, 0.0043, 0.0156, 0.0171, 0.0, 0.0043, -0.01, -0.0085, -0.0384, -0.0171, -0.0071, -0.0028, -0.0114, -0.0028, 0.0171, 0.0327, 0.0441, 0.0725, 0.0882, 0.1294, 0.1124, 0.1422, 0.1422, 0.1878, 0.2048, 0.2148, 0.2248, 0.1849, 0.2031, 0.1988, 0.2288, 0.2824, 0.2824, 0.2824, 0.2824, 0.2824, 0.2824, 0.3488, 0.3331, 0.3088, 0.3574, 0.396, 0.3988, 0.4017, 0.3931, 0.4088, 0.3788, 0.4117, 0.4574, 0.4217, 0.436, 0.4874, 0.536, 0.5874, 0.4988, 0.4374, 0.4174, 0.3988, 0.4374, 0.4788, 0.4731, 0.4688, 0.3888, 0.4374, 0.3974, 0.3788, 0.4374, 0.3545, 0.3817, 0.4188, 0.4688, 0.4788, 0.436, 0.4403, 0.4774, 0.4931, 0.5017, 0.5488, 0.5646, 0.5488, 0.5431, 0.5331, 0.556, 0.4974, 0.4688, 0.5417, 0.5374, 0.5188, 0.5788, 0.5931, 0.5874, 0.5874, 0.6717, 0.7157, 0.7215, 0.7215, 0.7215, 0.845, 0.9828, 0.9943, 1.0244, 0.9928, 0.9957, 0.9928, 0.9756, 1.0144, 1.0661, 1.1379, 1.1436, 1.0847, 1.1465, 1.1867, 1.1838, 1.1838, 1.29, 1.3317, 1.3073, 1.3044, 1.1594, 1.4049, 1.4279, 1.2872, 1.2771, 1.3891, 1.3805, 1.4092, 1.5643, 1.6016, 1.6016, 1.6016, 1.6016]}
{"unit_id": "orig:1955810102059131086", "ticker_idx": 0, "ts": "2025-08-14T01:54:00+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron and Nan Ya PCB (NYPCB) are expected to benefit the most due to their focus on high-end ASIC customers, with their long-term revenue share from this segment significantly revised upwards", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman raises AI server ABF TAM 35-55%; Unimicron, NYPCB top beneficiaries on ASIC share gains; Ibiden entering high-end ASIC substrates from early 2026.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "ABF: ASIC Servers to Become Key Growth Driver, Material Shortage Supports Price Outlook\nGoldman Sachs (C. Wang, 07/08/25)\n\nGoldman Sachs is raising its 2026/27 Total Addressable Market (TAM) forecast for the AI server ABF (Ajinomoto Build-up Film) market by 35% and 55%, respectively. The core logic is the strong demand for ASIC AI servers and their higher Average Selling Price (ASP) compared to GPU AI substrates. Unimicron and Nan Ya PCB (NYPCB) are expected to benefit the most due to their focus on high-end ASIC customers, with their long-term revenue share from this segment significantly revised upwards.\n\nRaised Forecasts for AI Server ABF Market and ASIC Shipments\nThe 2026/27 shipment forecast for ASIC AI server substrates is 3.4 million and 4.7 million units, respectively, with corresponding ASPs of $180 and $220. These ASPs are higher than those for GPU AI substrates ($143 and $183, respectively). Unimicron and NYPCB's key customers include Broadcom, Alchip, Marvell, and others. It is projected that by 2027, the revenue share from AI servers for these two companies will reach 16% and 20%, respectively, up from previous estimates of 13% and 12%.\n\nIbiden Poised to Enter the High-End ASIC Substrate Market\nLeveraging its technological leadership and a 40% capacity expansion plan, Ibiden is expected to increase its ASIC substrate revenue starting from early 2026, with a focus on the ultra-high-end segment (e.g., substrates with embedded components).\n\nT-glass Material Shortage Pushes Prices Upward\nGoldman Sachs believes that lead times for high-end ABF substrates may be extended, and prices are expected to be raised during contract negotiations from late September to early October. Although the capacity utilization rate is only projected to be around 75% for the period from 2H25 to 1H26, which may limit the earnings boost from price hikes after 2H26, it will still have a positive short-term impact on manufacturers' performance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 9.295681306631387e-05, "alpha_c_m1d": -0.0021661594326303213, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": 0.0021661594326303213, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018115942028985477, "ret_signed_p1d": -0.018115942028985477, "alpha_spy_p1d": -0.015774747386962717, "alpha_c_p1d": -0.010533525653932618, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.007582416375052858, "ret_p1w": -0.03260869565217395, "ret_signed_p1w": -0.03260869565217395, "alpha_spy_p1w": -0.018033945394872575, "alpha_c_p1w": 0.0003349647463984917, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.03294366039857244, "ret_p1m": 0.047101449275362306, "ret_signed_p1m": 0.047101449275362306, "alpha_spy_p1m": 0.02778208402342064, "alpha_c_p1m": 0.03533583992614164, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.011765609349220663, "ret_p3m": 0.19720359470533277, "ret_signed_p3m": 0.19720359470533277, "alpha_spy_p3m": 0.13527078396616776, "alpha_c_p3m": 0.10137367192576341, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.09582992277956937, "ret_p6m": 1.4637681159420288, "ret_signed_p6m": 1.4637681159420288, "alpha_spy_p6m": 1.386814763083311, "alpha_c_p6m": 1.4065410838471157, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.05722703209491309, "price_path": [-0.2913, -0.2806, -0.2663, -0.2699, -0.2412, -0.2555, -0.209, -0.2162, -0.2412, -0.2412, -0.277, -0.2699, -0.2627, -0.2627, -0.2627, -0.2842, -0.2341, -0.2197, -0.2341, -0.2663, -0.2627, -0.2734, -0.2591, -0.2806, -0.2842, -0.2126, -0.2018, -0.184, -0.184, -0.1911, -0.184, -0.1983, -0.1983, -0.1625, -0.1553, -0.1768, -0.1696, -0.1231, -0.0906, -0.1014, -0.1304, -0.0435, -0.0181, -0.0326, -0.0362, -0.0326, -0.0688, -0.0471, -0.0399, -0.0507, -0.0362, 0.0145, 0.029, -0.0036, -0.0036, -0.0217, -0.0181, -0.029, -0.0145, -0.0072, 0.0326, 0.0036, 0.0, 0.0, -0.0181, 0.029, 0.0072, -0.0326, -0.0326, -0.0217, 0.0145, 0.029, 0.1014, 0.087, 0.0471, -0.0254, -0.0507, -0.0399, -0.0326, -0.029, -0.0326, -0.0217, 0.0254, 0.0326, 0.0471, 0.029, 0.1014, 0.0688, 0.0761, 0.1413, 0.1486, 0.1522, 0.1377, 0.1377, 0.0797, 0.0797, 0.0978, 0.1232, 0.1087, 0.1304, 0.1304, 0.1449, 0.1594, 0.1667, 0.1667, 0.1123, 0.058, 0.1486, 0.1993, 0.163, 0.1594, 0.1594, 0.1522, 0.1304, 0.1304, 0.1848, 0.1775, 0.2029, 0.1594, 0.1848, 0.2935, 0.221, 0.1792, 0.19, 0.1613, 0.1828, 0.1972, 0.2044, 0.2835, 0.2428, 0.2138, 0.2319, 0.163, 0.2428, 0.1594, 0.1884, 0.2754, 0.2935, 0.3551, 0.3514, 0.4275, 0.4203, 0.4239, 0.4384, 0.558, 0.5978, 0.6377, 0.6486, 0.6304, 0.6304, 0.6051, 0.5507, 0.5362, 0.5, 0.558, 0.5978, 0.5616, 0.5725, 0.5725, 0.5978, 0.5978, 0.5906, 0.5942, 0.5942, 0.587, 0.6667, 0.6703, 0.6377, 0.6486, 0.6341, 0.7645, 0.8406, 0.8116, 0.808, 0.9855, 1.1812, 1.2319, 1.3261, 1.3696, 1.4565, 1.413, 1.5652, 1.663, 1.4964, 1.7428, 1.6449, 1.7899, 1.8116, 1.6486, 1.4638]}
{"unit_id": "orig:1955810102059131086", "ticker_idx": 1, "ts": "2025-08-14T01:54:00+00:00", "symbol": "Nan Ya PCB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron and Nan Ya PCB (NYPCB) are expected to benefit the most due to their focus on high-end ASIC customers, with their long-term revenue share from this segment significantly revised upwards", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman raises AI server ABF TAM 35-55%; Unimicron, NYPCB top beneficiaries on ASIC share gains; Ibiden entering high-end ASIC substrates from early 2026.", "resolved_tickers": ["8046.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nan Ya PCB (NYPCB) listed on TWSE as 8046.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "ABF: ASIC Servers to Become Key Growth Driver, Material Shortage Supports Price Outlook\nGoldman Sachs (C. Wang, 07/08/25)\n\nGoldman Sachs is raising its 2026/27 Total Addressable Market (TAM) forecast for the AI server ABF (Ajinomoto Build-up Film) market by 35% and 55%, respectively. The core logic is the strong demand for ASIC AI servers and their higher Average Selling Price (ASP) compared to GPU AI substrates. Unimicron and Nan Ya PCB (NYPCB) are expected to benefit the most due to their focus on high-end ASIC customers, with their long-term revenue share from this segment significantly revised upwards.\n\nRaised Forecasts for AI Server ABF Market and ASIC Shipments\nThe 2026/27 shipment forecast for ASIC AI server substrates is 3.4 million and 4.7 million units, respectively, with corresponding ASPs of $180 and $220. These ASPs are higher than those for GPU AI substrates ($143 and $183, respectively). Unimicron and NYPCB's key customers include Broadcom, Alchip, Marvell, and others. It is projected that by 2027, the revenue share from AI servers for these two companies will reach 16% and 20%, respectively, up from previous estimates of 13% and 12%.\n\nIbiden Poised to Enter the High-End ASIC Substrate Market\nLeveraging its technological leadership and a 40% capacity expansion plan, Ibiden is expected to increase its ASIC substrate revenue starting from early 2026, with a focus on the ultra-high-end segment (e.g., substrates with embedded components).\n\nT-glass Material Shortage Pushes Prices Upward\nGoldman Sachs believes that lead times for high-end ABF substrates may be extended, and prices are expected to be raised during contract negotiations from late September to early October. Although the capacity utilization rate is only projected to be around 75% for the period from 2H25 to 1H26, which may limit the earnings boost from price hikes after 2H26, it will still have a positive short-term impact on manufacturers' performance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00877192982456143, "ret_signed_m1d": 0.00877192982456143, "alpha_spy_m1d": 0.008864886637627745, "alpha_c_m1d": 0.0066057703919311095, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": 0.0021661594326303213, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00877192982456143, "ret_signed_p1d": -0.00877192982456143, "alpha_spy_p1d": -0.006430735182538672, "alpha_c_p1d": -0.0011895134495085724, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.007582416375052858, "ret_p1w": -0.011695906432748537, "ret_signed_p1w": -0.011695906432748537, "alpha_spy_p1w": 0.0028788438245528347, "alpha_c_p1w": 0.0212477539658239, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.03294366039857244, "ret_p1m": 0.07602339181286544, "ret_signed_p1m": 0.07602339181286544, "alpha_spy_p1m": 0.05670402656092377, "alpha_c_p1m": 0.06425778246364477, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.011765609349220663, "ret_p3m": 0.6900584795321638, "ret_signed_p3m": 0.6900584795321638, "alpha_spy_p3m": 0.6281256687929988, "alpha_c_p3m": 0.5942285567525945, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.09582992277956937, "ret_p6m": 1.2163742690058479, "ret_signed_p6m": 1.2163742690058479, "alpha_spy_p6m": 1.13942091614713, "alpha_c_p6m": 1.1591472369109348, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.05722703209491309, "price_path": [-0.4087, -0.3971, -0.3798, -0.4, -0.3885, -0.3942, -0.3363, -0.3827, -0.3914, -0.3914, -0.4325, -0.4279, -0.3914, -0.3856, -0.3798, -0.3971, -0.3682, -0.3537, -0.3624, -0.3827, -0.3624, -0.374, -0.3682, -0.3827, -0.3914, -0.3769, -0.3508, -0.3479, -0.345, -0.3479, -0.3508, -0.3537, -0.3392, -0.2749, -0.2632, -0.269, -0.2602, -0.1871, -0.1813, -0.155, -0.1608, -0.0789, 0.0117, -0.0058, 0.0, -0.0175, -0.0468, 0.0029, 0.0234, 0.0556, 0.0789, 0.0351, 0.0526, 0.038, 0.038, 0.0322, 0.038, 0.0175, 0.0088, -0.0146, 0.0585, 0.0351, 0.0088, 0.0, -0.0088, 0.0702, 0.0526, -0.0234, -0.0117, 0.0088, 0.0877, 0.0936, 0.2018, 0.3216, 0.2865, 0.1754, 0.152, 0.114, 0.0848, 0.0965, 0.0877, 0.076, 0.0819, 0.0497, 0.076, 0.0468, 0.1491, 0.1316, 0.1696, 0.2865, 0.3772, 0.3713, 0.2895, 0.3392, 0.2807, 0.2807, 0.3216, 0.3333, 0.3509, 0.4854, 0.4854, 0.4795, 0.5234, 0.5497, 0.5497, 0.5643, 0.4971, 0.6462, 0.6462, 0.6404, 0.6228, 0.5994, 0.6111, 0.5936, 0.5936, 0.6287, 0.614, 0.6988, 0.6491, 0.6988, 0.6667, 0.6374, 0.5906, 0.5965, 0.5965, 0.6988, 0.6901, 0.6637, 0.7076, 0.652, 0.5936, 0.5526, 0.4678, 0.5731, 0.462, 0.4444, 0.462, 0.5029, 0.5234, 0.5351, 0.5585, 0.5322, 0.5731, 0.5117, 0.5789, 0.5789, 0.6608, 0.6374, 0.6257, 0.5877, 0.5556, 0.4561, 0.424, 0.3713, 0.4006, 0.4269, 0.3977, 0.3977, 0.3977, 0.4415, 0.4064, 0.4064, 0.4094, 0.4094, 0.386, 0.4123, 0.4298, 0.4123, 0.386, 0.4064, 0.4883, 0.5673, 0.5292, 0.5409, 0.693, 0.8596, 0.9971, 1.1111, 1.1667, 1.2339, 1.1637, 1.3099, 1.3158, 1.1871, 1.2719, 1.152, 1.307, 1.3977, 1.3246, 1.2164]}
{"unit_id": "orig:1955810102059131086", "ticker_idx": 2, "ts": "2025-08-14T01:54:00+00:00", "symbol": "Ibiden", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Ibiden is expected to increase its ASIC substrate revenue starting from early 2026, with a focus on the ultra-high-end segment", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman raises AI server ABF TAM 35-55%; Unimicron, NYPCB top beneficiaries on ASIC share gains; Ibiden entering high-end ASIC substrates from early 2026.", "resolved_tickers": ["4062.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ibiden Co. Ltd 4062.T Japan ABF substrate", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "ABF: ASIC Servers to Become Key Growth Driver, Material Shortage Supports Price Outlook\nGoldman Sachs (C. Wang, 07/08/25)\n\nGoldman Sachs is raising its 2026/27 Total Addressable Market (TAM) forecast for the AI server ABF (Ajinomoto Build-up Film) market by 35% and 55%, respectively. The core logic is the strong demand for ASIC AI servers and their higher Average Selling Price (ASP) compared to GPU AI substrates. Unimicron and Nan Ya PCB (NYPCB) are expected to benefit the most due to their focus on high-end ASIC customers, with their long-term revenue share from this segment significantly revised upwards.\n\nRaised Forecasts for AI Server ABF Market and ASIC Shipments\nThe 2026/27 shipment forecast for ASIC AI server substrates is 3.4 million and 4.7 million units, respectively, with corresponding ASPs of $180 and $220. These ASPs are higher than those for GPU AI substrates ($143 and $183, respectively). Unimicron and NYPCB's key customers include Broadcom, Alchip, Marvell, and others. It is projected that by 2027, the revenue share from AI servers for these two companies will reach 16% and 20%, respectively, up from previous estimates of 13% and 12%.\n\nIbiden Poised to Enter the High-End ASIC Substrate Market\nLeveraging its technological leadership and a 40% capacity expansion plan, Ibiden is expected to increase its ASIC substrate revenue starting from early 2026, with a focus on the ultra-high-end segment (e.g., substrates with embedded components).\n\nT-glass Material Shortage Pushes Prices Upward\nGoldman Sachs believes that lead times for high-end ABF substrates may be extended, and prices are expected to be raised during contract negotiations from late September to early October. Although the capacity utilization rate is only projected to be around 75% for the period from 2H25 to 1H26, which may limit the earnings boost from price hikes after 2H26, it will still have a positive short-term impact on manufacturers' performance.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.001654323889319853, "ret_signed_m1d": -0.001654323889319853, "alpha_spy_m1d": -0.0015613670762535392, "alpha_c_m1d": -0.0038204833219501744, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": 0.0021661594326303213, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03609564798621445, "ret_signed_p1d": 0.03609564798621445, "alpha_spy_p1d": 0.03843684262823721, "alpha_c_p1d": 0.04367806436126731, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.007582416375052858, "ret_p1w": 0.022860613966683596, "ret_signed_p1w": 0.022860613966683596, "alpha_spy_p1w": 0.03743536422398497, "alpha_c_p1w": 0.055804274365256035, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.03294366039857244, "ret_p1m": 0.2613927707869328, "ret_signed_p1m": 0.2613927707869328, "alpha_spy_p1m": 0.24207340553499113, "alpha_c_p1m": 0.24962716143771213, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.011765609349220663, "ret_p3m": 1.0183408425410816, "ret_signed_p3m": 1.0183408425410816, "alpha_spy_p3m": 0.9564080318019166, "alpha_c_p3m": 0.9225109197615122, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.09582992277956937, "ret_p6m": 1.1383096265986286, "ret_signed_p6m": 1.1383096265986286, "alpha_spy_p6m": 1.0613562737399107, "alpha_c_p6m": 1.0810825945037155, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.05722703209491309, "price_path": [-0.2393, -0.2094, -0.2148, -0.2184, -0.2179, -0.1897, -0.1627, -0.1299, -0.1053, -0.111, -0.1014, -0.1143, -0.1068, -0.1305, -0.1235, -0.1367, -0.1387, -0.1122, -0.133, -0.134, -0.1033, -0.0838, -0.065, -0.0692, -0.0994, -0.1232, -0.1063, -0.0747, -0.0579, -0.0397, -0.0454, -0.0382, -0.0725, -0.0465, -0.0439, -0.042, -0.0068, -0.0322, -0.0486, -0.0651, -0.0675, -0.0323, -0.0421, -0.0379, -0.0532, -0.0532, -0.0602, -0.076, -0.0597, -0.0573, -0.0435, -0.0629, -0.0469, -0.0284, -0.0546, -0.0135, 0.035, 0.0191, 0.0041, -0.0068, -0.0068, 0.0033, -0.0017, 0.0, 0.0361, 0.048, 0.055, 0.0125, 0.0229, 0.0615, 0.0859, 0.1129, 0.1032, 0.1257, 0.0892, 0.0543, 0.0532, 0.0495, 0.048, 0.085, 0.0929, 0.0826, 0.1761, 0.2728, 0.2614, 0.2614, 0.2644, 0.1939, 0.1975, 0.2855, 0.3575, 0.3575, 0.3268, 0.3777, 0.3471, 0.3702, 0.3557, 0.3492, 0.3814, 0.391, 0.4638, 0.4555, 0.4939, 0.5596, 0.5015, 0.5015, 0.4755, 0.5407, 0.5769, 0.5181, 0.54, 0.509, 0.5339, 0.4585, 0.586, 0.6381, 0.9067, 1.0485, 1.0515, 1.2055, 1.2055, 1.1662, 1.0372, 1.0908, 0.9346, 0.967, 1.0183, 1.0681, 1.0553, 0.881, 1.0017, 0.832, 0.7565, 0.8916, 0.7097, 0.7097, 0.7097, 0.6456, 0.7279, 0.7927, 0.7082, 0.6856, 0.7814, 0.7942, 0.9172, 0.9225, 0.9512, 0.9029, 0.9135, 0.9248, 0.7942, 0.7165, 0.7595, 0.6833, 0.758, 0.8486, 0.8976, 1.0025, 0.9519, 0.998, 0.9998, 1.0315, 1.0315, 1.0315, 1.0315, 1.1603, 1.1163, 1.1797, 1.0677, 1.0263, 1.0263, 1.1941, 1.2442, 1.2581, 1.3647, 1.3635, 1.3939, 1.5101, 1.5793, 1.5002, 1.3906, 1.4145, 1.4899, 1.492, 1.4836, 1.333, 1.5325, 1.173, 1.1474, 1.1383]}
{"unit_id": "quote:1953338821816533001", "ticker_idx": 0, "ts": "2025-08-07T06:14:01+00:00", "symbol": "2449.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "King Yuan secured the burn-in test equipment deal with Nvidia, beating out Advantest", "tweet_type": "news_reaction", "timeline": "months", "summary": "Long King Yuan (2449.TW): rumored to have locked in burn-in test deal with Nvidia for Rubin, with August order price hikes.", "resolved_tickers": ["2449.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Rumor: King Yuan secured the burn-in test equipment deal with Nvidia, beating out Advantest.", "reply_text": null, "parent_text": null, "quoted_text": "轉傳\n\n測試設備\n京元電（2449）預計8月份就會跟輝達完成簽約\n同樣的京元電在這次Rubin晶片依然拿下全部的測試訂單\n對手日月光burn-in測試不順、加上愛德萬burn-in機台沒有通過NV認證，目前在老化測試跟FT還是京元電獨拿\n\n由於老化測試產能嚴重不足\n從公司的說法來看\n京元電8月份訂單全面漲價", "parent_author": null, "ret_m1d": -0.04526748971193417, "ret_signed_m1d": -0.04526748971193417, "alpha_spy_m1d": -0.04610581599171548, "alpha_c_m1d": -0.029810008825579604, "bench_spy_m1d": 0.0008383262797813096, "bench_c_m1d": -0.015457480886354569, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09876543209876543, "ret_signed_p1d": 0.09876543209876543, "alpha_spy_p1d": 0.09096785864007217, "alpha_c_p1d": 0.09048704980608702, "bench_spy_p1d": 0.007797573458693252, "bench_c_p1d": 0.00827838229267841, "ret_p1w": 0.14814814814814814, "ret_signed_p1w": 0.14814814814814814, "alpha_spy_p1w": 0.12806112616133536, "alpha_c_p1w": 0.11105000474304583, "bench_spy_p1w": 0.02008702198681278, "bench_c_p1w": 0.03709814340510231, "ret_p1m": 0.3168724279835391, "ret_signed_p1m": 0.3168724279835391, "alpha_spy_p1m": 0.2931634003164354, "alpha_c_p1m": 0.30945273620990155, "bench_spy_p1m": 0.0237090276671037, "bench_c_p1m": 0.007419691773637549, "ret_p3m": 0.7613168724279835, "ret_signed_p3m": 0.7613168724279835, "alpha_spy_p3m": 0.6903605505803985, "alpha_c_p3m": 0.5495827813102567, "bench_spy_p3m": 0.07095632184758505, "bench_c_p3m": 0.21173409111772679, "ret_p6m": 1.4362139917695473, "ret_signed_p6m": 1.4362139917695473, "alpha_spy_p6m": 1.3354803905599275, "alpha_c_p6m": 1.046012653764416, "bench_spy_p6m": 0.10073360120961983, "bench_c_p6m": 0.39020133800513124, "price_path": [-0.2571, -0.2745, -0.2041, -0.2223, -0.2254, -0.2429, -0.2421, -0.2254, -0.231, -0.2334, -0.2413, -0.2452, -0.2452, -0.2318, -0.2318, -0.2674, -0.2421, -0.2254, -0.2136, -0.2159, -0.1922, -0.1803, -0.212, -0.208, -0.1961, -0.1882, -0.1843, -0.1882, -0.2041, -0.208, -0.2001, -0.1724, -0.1724, -0.1882, -0.1684, -0.1922, -0.1843, -0.1803, -0.1486, -0.1526, -0.1447, -0.1763, -0.1605, -0.1646, -0.1523, -0.1811, -0.1481, -0.0988, -0.0823, -0.0988, -0.0576, -0.0864, -0.0823, -0.0576, -0.0782, -0.0535, -0.0535, -0.0576, -0.0288, -0.0288, -0.0329, -0.0247, -0.0453, 0.0, 0.0988, 0.1523, 0.1605, 0.1523, 0.1481, 0.1399, 0.1605, 0.1276, 0.0494, 0.0947, 0.0823, 0.1152, 0.1481, 0.144, 0.1687, 0.284, 0.3045, 0.251, 0.2634, 0.2428, 0.3169, 0.3909, 0.4033, 0.4198, 0.3663, 0.3663, 0.3704, 0.3663, 0.358, 0.3786, 0.3951, 0.4074, 0.3745, 0.3663, 0.3663, 0.2798, 0.2798, 0.3498, 0.358, 0.3786, 0.3909, 0.3909, 0.4198, 0.4321, 0.4486, 0.4486, 0.3992, 0.3292, 0.3663, 0.4856, 0.4403, 0.465, 0.5144, 0.4527, 0.4403, 0.4403, 0.5473, 0.5514, 0.7037, 0.6708, 0.786, 0.8436, 0.7613, 0.8148, 0.7901, 0.7572, 0.786, 0.7531, 0.7325, 0.7737, 0.716, 0.7572, 0.6296, 0.6173, 0.7407, 0.6831, 0.6749, 0.6872, 0.7449, 0.8272, 0.8848, 0.7901, 0.8642, 0.8601, 0.8148, 0.823, 0.8765, 0.8807, 0.8848, 0.823, 0.893, 0.823, 0.7778, 0.7695, 0.716, 0.7737, 0.8189, 0.8107, 0.8025, 0.8025, 0.8354, 1.0165, 1.0041, 1.037, 1.037, 1.2016, 1.1646, 1.1975, 1.1934, 1.1646, 1.1687, 1.2222, 1.2263, 1.2263, 1.2016, 1.251, 1.2346, 1.2593, 1.1975, 1.2593, 1.358, 1.4486, 1.4691, 1.5597, 1.5103, 1.4362]}
{"unit_id": "orig:1965189663066607935", "ticker_idx": 0, "ts": "2025-09-08T23:05:01+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Data Center Compute revenue (ex-China) expected to grow +17% QoQ vs. +19%... Vera Rubin is on track as planned; 'Scale Across' could emerge as a new competitive edge... GPM will recover to mid-70% levels by year-end", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish NVDA: data center compute growing +17% QoQ ex-China, Rubin on schedule, GPM recovering to mid-70s, strong demand from hyperscalers and sovereigns.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "250909_NVIDIA: Communacopia + Technology Conference 2025 - Goldman Sachs\n\n[Key Takeaways]\n(1) Data Center Compute revenue (ex-China) expected to grow +17% QoQ vs. +19%\n(2) Vera Rubin is on track as planned; “Scale Across” could emerge as a new competitive edge\n(3) Strong demand not only from hyperscalers but also from non-traditional customers, including sovereigns\n(4) Shipment timing for China-related products remains somewhat uncertain\n\n[Contents]\n\n(1) AI CAPEX\n\n- AI infrastructure investment to reach $3 trillion by 2030\n\n- Key drivers: 1) Hyperscalers, 2) Non-traditional customers\n\n(2) Near-term Business\n\n- Ex-China, Data Center Compute revenue expected to grow +17% in CY3Q25(F) [an upturn vs. +12% growth in the prior quarter]\n\n- Rubin family remains on schedule, with mass production slated for mid-2026\n\n(3) China Business\n\n- EVP Kress noted that NVIDIA has secured H20 shipment licenses for several major customers in China\n\n- However, due to political uncertainty, actual shipment timing is unclear; if resolved, shipments could contribute $2–5bn revenue in 3Q\n\n(4) Competitive Landscape\n\n- Strong emphasis that Training and Inference are not necessarily separate workloads\n\n- To enable accurate responses and Agentic AI, continued compute resources are required not only for inference but also for post-processing [tuning/adaptation/reinforcement, etc.]\n\n(5) Product Launch Cycle\n\n- NVIDIA’s annual product launch cycle delivers new models and connectivity solutions, providing customers with economic benefits\n\n- Six chips comprising the Rubin architecture have already taped out; “Scale Across” will be a new source of competitive strength for Rubin going forward\n\n(6) Financials & Capital Allocation\n\n- Reaffirmed that GPM will recover to mid-70% levels by year-end\n\n- Plans to continue share buybacks/cancellations, while also investing in ecosystem partners such as technology innovation companies\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007664302707223625, "ret_signed_m1d": -0.007664302707223625, "alpha_spy_m1d": -0.005213780608743446, "alpha_c_m1d": 0.0032623460281404615, "bench_spy_m1d": -0.002450522098480179, "bench_c_m1d": -0.010926648735364086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014556426417988044, "ret_signed_p1d": 0.014556426417988044, "alpha_spy_p1d": 0.012244505720378163, "alpha_c_p1d": 0.01020591924940728, "bench_spy_p1d": 0.0023119206976098816, "bench_c_p1d": 0.004350507168580764, "ret_p1w": 0.0561465529724674, "ret_signed_p1w": 0.0561465529724674, "alpha_spy_p1w": 0.037528428861316465, "alpha_c_p1w": 0.02022978166409306, "bench_spy_p1w": 0.018618124111150935, "bench_c_p1w": 0.03591677130837434, "ret_p1m": 0.09946193196537245, "ret_signed_p1m": 0.09946193196537245, "alpha_spy_p1m": 0.06533107611547795, "alpha_c_p1m": -0.03722369392384106, "bench_spy_p1m": 0.034130855849894504, "bench_c_p1m": 0.1366856258892135, "ret_p3m": 0.08965935717095963, "ret_signed_p3m": 0.08965935717095963, "alpha_spy_p3m": 0.03192860078014004, "alpha_c_p3m": -0.12982001286651723, "bench_spy_p3m": 0.05773075639081959, "bench_c_p3m": 0.21947937003747686, "ret_p6m": 0.06987213437582929, "ret_signed_p6m": 0.06987213437582929, "alpha_spy_p6m": 0.015309098297438872, "alpha_c_p6m": -0.25306341111572284, "bench_spy_p6m": 0.054563036078390414, "bench_c_p6m": 0.3229355454915521, "price_path": [-0.1514, -0.1385, -0.1565, -0.1403, -0.1437, -0.1356, -0.1356, -0.1453, -0.1434, -0.1213, -0.0832, -0.079, -0.0627, -0.0613, -0.0892, -0.0657, -0.0533, -0.0533, -0.0598, -0.0494, -0.0323, -0.025, -0.0201, -0.0252, 0.0142, 0.0182, 0.0279, 0.0244, 0.0182, -0.0076, 0.0147, 0.0323, 0.0308, 0.0501, 0.0428, 0.0651, 0.0568, 0.0321, 0.0695, 0.0591, 0.066, 0.074, 0.0855, 0.0817, 0.0882, 0.0789, 0.0815, 0.0721, 0.0814, 0.0436, 0.0421, 0.0396, 0.0575, 0.0683, 0.08, 0.079, 0.0705, 0.0349, 0.0349, 0.0147, 0.0137, 0.0199, -0.0077, 0.0, 0.0146, 0.0536, 0.0527, 0.0566, 0.0561, 0.0391, 0.0118, 0.0472, 0.0497, 0.091, 0.0602, 0.0515, 0.0558, 0.0588, 0.0805, 0.1086, 0.1125, 0.1223, 0.1148, 0.1024, 0.0995, 0.1236, 0.1442, 0.0883, 0.119, 0.0697, 0.0685, 0.0803, 0.0886, 0.0852, 0.0764, 0.0712, 0.0823, 0.1067, 0.1378, 0.1945, 0.2302, 0.2055, 0.2031, 0.2292, 0.1806, 0.1599, 0.1175, 0.1179, 0.1827, 0.1477, 0.1515, 0.1103, 0.1299, 0.1087, 0.0776, 0.1083, 0.0733, 0.0629, 0.0847, 0.0566, 0.0711, 0.0711, 0.0517, 0.069, 0.0782, 0.0671, 0.0897, 0.0839, 0.1026, 0.0991, 0.092, 0.0751, 0.04, 0.0475, 0.056, 0.0157, 0.0348, 0.0755, 0.0915, 0.1243, 0.1207, 0.1207, 0.1321, 0.1184, 0.1144, 0.1082, 0.1082, 0.1222, 0.1178, 0.1126, 0.1237, 0.0995, 0.0985, 0.0989, 0.1041, 0.0882, 0.1115, 0.1066, 0.1066, 0.0581, 0.0893, 0.0983, 0.1152, 0.108, 0.1202, 0.138, 0.1439, 0.1357, 0.1029, 0.0716, 0.0351, 0.0213, 0.1017, 0.1292, 0.1203, 0.1293, 0.1108, 0.0863, 0.0863, 0.0991, 0.117, 0.1165, 0.1279, 0.1382, 0.1459, 0.162, 0.0986, 0.0529, 0.0843, 0.0699]}
{"unit_id": "thread:1949810451812163611", "ticker_idx": 0, "ts": "2025-07-28T12:33:32+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "due to intensifying competition and an unsustainably high margin structure, HBM Average Selling Price (ASP) will decline significantly next year (2026)", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish SK Hynix: Goldman sees significant HBM ASP decline in 2026 due to intensifying competition and unsustainably high margin structure.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Goldman Sachs) A frequent question I get is, 'Why are semiconductor equipment stocks (semicaps) underperforming when AI is so hot?' Simply put, AI's contribution to semiconductor equipment companies' earnings is not yet large enough. Furthermore, it's not driving the transition to leading-edge processes (Nvidia's latest Blackwell GPU architecture uses a 4nm process).\n\nOn top of this, Intel did not stick to its previous stance on the 14A (1.4nm) process in its earnings announcement (previously, there were observations that it might transition to 14A to regain momentum), and combined with Samsung's struggles, TSMC has become the only alternative in AI chip manufacturing. (For reference, Samsung recently succeeded in winning the order for Tesla's next-generation AI6 chip, presumed to be 2nm). When Elon Musk tweeted, \"To accelerate progress, I will personally oversee the production line, and conveniently, the fab is not far from my house,\" I wonder if they knew what was coming. 😉\n\n---\n\nGoldman Sachs) Speaking of semiconductor equipment stocks, our report on SK Hynix published last week garnered a lot of attention. This was because Goldman Sachs (GS) pointed out the risk of HBM price declines in 2026. There's an interesting point in the analysis from Goldman Sachs Research (GIR). SK Hynix did not provide specific guidance (earnings forecasts) for 2026 (Note: SK Hynix already mentioned in May 2024 that almost all of its 2025 HBM volume was sold out). They only stated that negotiations are proceeding as planned and that they are currently discussing product mix and pricing with customers. We continue to believe that \"due to intensifying competition and an unsustainably high margin structure, HBM Average Selling Price (ASP) will decline significantly next year (2026).\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01526695385491017, "ret_signed_m1d": -0.01526695385491017, "alpha_spy_m1d": -0.015015790642599391, "alpha_c_m1d": -0.028718884106526232, "bench_spy_m1d": 0.00025116321231077876, "bench_c_m1d": -0.013451930251616062, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0019083467743843396, "ret_signed_p1d": -0.0019083467743843396, "alpha_spy_p1d": -0.00454609201487044, "alpha_c_p1d": 0.0032046326107659784, "bench_spy_p1d": -0.0026377452404861, "bench_c_p1d": 0.005112979385150318, "ret_p1w": -0.015267253287969629, "ret_signed_p1w": 0.015267253287969629, "alpha_spy_p1w": 0.006208176008847044, "alpha_c_p1w": 0.011080707277195567, "bench_spy_p1w": -0.009059077279122585, "bench_c_p1w": -0.0041865460107740615, "ret_p1m": -0.0019084066609962314, "ret_signed_p1m": 0.0019084066609962314, "alpha_spy_p1m": 0.014813789062543714, "alpha_c_p1m": 0.020233067615690103, "bench_spy_p1m": 0.012905382401547483, "bench_c_p1m": 0.018324660954693872, "ret_p3m": 0.8289738974021368, "ret_signed_p3m": -0.8289738974021368, "alpha_spy_p3m": -0.7713823371045563, "alpha_c_p3m": -0.6457613239605728, "bench_spy_p3m": 0.0575915602975805, "bench_c_p3m": 0.18321257344156394, "ret_p6m": 1.8420077992730124, "ret_signed_p6m": -1.8420077992730124, "alpha_spy_p6m": -1.7721013172230644, "alpha_c_p6m": -1.4981804203147, "bench_spy_p6m": 0.06990648204994798, "bench_c_p6m": 0.3438273789583124, "price_path": [-0.3237, -0.3237, -0.2914, -0.2914, -0.2914, -0.2731, -0.275, -0.2757, -0.2571, -0.2437, -0.2152, -0.2361, -0.2209, -0.2403, -0.2304, -0.2361, -0.2498, -0.238, -0.2266, -0.2285, -0.2075, -0.1908, -0.2195, -0.208, -0.208, -0.1698, -0.1431, -0.1431, -0.126, -0.1202, -0.084, -0.1011, -0.1011, -0.0534, -0.0496, -0.0592, -0.0611, -0.0191, -0.0095, 0.063, 0.0916, 0.1183, 0.084, 0.1145, 0.0897, 0.0649, 0.063, 0.0324, 0.0344, 0.0763, 0.0725, 0.1336, 0.124, 0.145, 0.1393, 0.1298, 0.0286, 0.0267, 0.0401, 0.0248, 0.0267, 0.0286, 0.0153, 0.0, 0.0019, 0.0057, 0.0439, -0.0153, -0.0153, 0.0057, -0.0134, 0.0, -0.021, 0.0191, 0.0267, 0.0611, 0.0553, 0.0553, 0.021, 0.0038, -0.0248, -0.0649, -0.042, -0.0095, -0.0019, -0.0076, 0.0263, 0.0282, -0.0215, -0.0043, 0.0034, 0.0148, 0.0454, 0.0588, 0.1008, 0.162, 0.1734, 0.2556, 0.2652, 0.3302, 0.2747, 0.3588, 0.3588, 0.3416, 0.3799, 0.3665, 0.3627, 0.2862, 0.334, 0.3283, 0.376, 0.5117, 0.5117, 0.5117, 0.5117, 0.5117, 0.5117, 0.6359, 0.5863, 0.5729, 0.6149, 0.7296, 0.7793, 0.8557, 0.8309, 0.8404, 0.829, 0.9494, 1.0449, 0.9914, 1.1328, 1.1711, 1.1367, 1.3698, 1.2399, 1.2131, 1.2666, 1.2169, 1.3163, 1.366, 1.3584, 1.3393, 1.1405, 1.3163, 1.1787, 1.1481, 1.1825, 0.9914, 0.9876, 0.9838, 1.0029, 1.0808, 1.0273, 1.0579, 1.1344, 1.1114, 1.0732, 1.0808, 1.2071, 1.165, 1.2453, 1.1611, 1.1841, 1.1191, 1.0273, 1.1076, 1.1114, 1.0923, 1.2185, 1.2338, 1.2491, 1.2491, 1.2912, 1.448, 1.4901, 1.4901, 1.4901, 1.5896, 1.6622, 1.777, 1.8382, 1.8917, 1.8458, 1.865, 1.8229, 1.8382, 1.865, 1.8917, 1.9223, 1.842]}
{"unit_id": "thread:1967412543087108351", "ticker_idx": 0, "ts": "2025-09-15T02:17:57+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron has paused NAND Flash and DRAM pricing quotes for a week, while market sources anticipate potential NAND Flash price rises of up to 30%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Sharing Digitimes report: Micron halting quotes ahead of broad NAND/DRAM price hikes up to 30%, signaling new memory upcycle into 2026; bullish for MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Micron halts NAND and DRAM quotes after SanDisk price hike, sharper increases looming- Digitimes\n\nMemory manufacturers are swiftly following SanDisk's announcement of a 10% price increase for NAND products, signaling a broader industry move toward higher pricing amid expected supply shortages starting in 2026. Micron has paused NAND Flash and DRAM pricing quotes for a week, while market sources anticipate potential NAND Flash price rises of up to 30%.\n\nIndustry insiders indicate that Micron's pricing suspension and rumored sharp price hikes are tied to anticipated growth in CSP's storage needs beginning in 2026. Though consumer demand remains weak, the anticipated surge in data center capacity—especially for AI and enterprise storage—is expected to tighten NAND Flash supplies, initiating a new price increase cycle.\n\nPhison Electronics, a Taiwanese NAND controller manufacturer, has also halted pricing updates pending further direction from NAND suppliers, a move widely interpreted in the industry as an early step in the pricing negotiation process. This coordinated halt signals a consensus on increasing prices, while the precise magnitude of the hikes remains uncertain ahead of the fourth quarter 2025 contract negotiations.\n\nMicron's planned price hikes could significantly surpass the initial 10% level seen with SanDisk, with estimates ranging from 10 to 30%. While some memory module makers expect a 10 to 15% NAND Flash price increase in the fourth quarter, reports from the China market suggest that Micron has hinted at even steeper rises, possibly up to 30%, for storage products. Unlike SanDisk, Micron has not formally notified all customers, indicating these early moves may be testing market responses before finalizing increases.\n\nPrice suspensions and hikes extend beyond NAND to DRAM products as well. Taiwanese memory companies report verbal notifications from Micron that DRAM pricing—including DDR4, DDR5, LPDDR4, and LPDDR5—will be suspended for around a week, canceling previously agreed prices. Analysts expect DRAM price hikes of at least 10% based on historical trends, with certain segments such as automotive electronics potentially facing increases as high as 70%.\n\nThe broader memory supply chain notes that DDR4 shortages have already driven significant price momentum, with DDR5 prices steadily rising. Expectations are that typical seasonal price slowdowns in late 2025 may give way to sustained or increased pricing pressure, largely fueled by CSP investments in AI ASICs and infrastructure. This demand surge is projected to substantially increase server NAND Flash consumption in 2026, intensifying supply constraints in the second half of next year.\n\nLeading CSPs have reportedly begun negotiating bulk procurement orders for 2026, with data centers placing enterprise SSD orders amounting to tens of exabytes (EB). Given global NAND consumption sits between 800 and 1000 EB annually, and servers accounted for roughly 30% in 2025, a shift toward higher-capacity enterprise SSDs is expected to reshape supply-demand dynamics. This transition is anticipated to boost demand for QLC NAND, a lower-cost, high-density flash technology favored for massive data storage.\n\nDespite the likelihood of continued softness in consumer NAND applications early next year, industry executives remain optimistic about the NAND market's prospects. Khein-Seng Pua, CEO of Phison Electronics, recently noted that the NAND industry is entering a shortage phase in 2026, driven by robust data center and AI usage. Some NAND manufacturers have already begun implementing price increase strategies, while Phison plans to engage closely with the supply chain to maintain stable raw material supplies amid tightening market conditions.\n\nMicron is expected to publish its financial report on September 23, after which it will enter a quiet period and refrain from commenting on market speculation related to its pricing actions. Meanwhile, the developments unfolding across NAND and DRAM suppliers illustrate a supply chain bracing for higher costs and constrained availability that could define the memory market landscape throughout 2026.\n\n$MU\n\n---\n\nhttps://t.co/mgSKe2h8eK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0034227847689322077, "ret_signed_m1d": -0.0034227847689322077, "alpha_spy_m1d": 0.001872880154554446, "alpha_c_m1d": 0.0058554532438491735, "bench_spy_m1d": -0.005295664923486654, "bench_c_m1d": -0.009278238012781381, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00665524802138262, "ret_signed_p1d": 0.00665524802138262, "alpha_spy_p1d": 0.008032040806664309, "alpha_c_p1d": 0.006525092178197411, "bench_spy_p1d": -0.0013767927852816886, "bench_c_p1d": 0.000130155843185209, "ret_p1w": 0.04341761654698839, "ret_signed_p1w": 0.04341761654698839, "alpha_spy_p1w": 0.031647775172926274, "alpha_c_p1w": -0.005415241325521736, "bench_spy_p1w": 0.011769841374062118, "bench_c_p1w": 0.04883285787251013, "ret_p1m": 0.18639236993776254, "ret_signed_p1m": 0.18639236993776254, "alpha_spy_p1m": 0.1816171351085838, "alpha_c_p1m": 0.1006743825696561, "bench_spy_p1m": 0.004775234829178743, "bench_c_p1m": 0.08571798736810643, "ret_p3m": 0.6392333132583683, "ret_signed_p3m": 0.6392333132583683, "alpha_spy_p3m": 0.5935830815840875, "alpha_c_p3m": 0.43179120943765636, "bench_spy_p3m": 0.04565023167428084, "bench_c_p3m": 0.20744210382071193, "ret_p6m": 1.5576811730081666, "ret_signed_p6m": 1.5576811730081666, "alpha_spy_p6m": 1.5271868379940525, "alpha_c_p6m": 1.26013813861304, "bench_spy_p6m": 0.030494335014114027, "bench_c_p6m": 0.29754303439512664, "price_path": [-0.2286, -0.2286, -0.2173, -0.2269, -0.19, -0.1942, -0.2021, -0.21, -0.2195, -0.2345, -0.2291, -0.2256, -0.2256, -0.2399, -0.2114, -0.2252, -0.2197, -0.2107, -0.2482, -0.2387, -0.262, -0.2821, -0.275, -0.2823, -0.3077, -0.3039, -0.2918, -0.2948, -0.2949, -0.2904, -0.2727, -0.3082, -0.3352, -0.3169, -0.3087, -0.3105, -0.2909, -0.2464, -0.2158, -0.1903, -0.2123, -0.2059, -0.2339, -0.2169, -0.2264, -0.2571, -0.2661, -0.2541, -0.2621, -0.2616, -0.2537, -0.2267, -0.2457, -0.2457, -0.249, -0.2475, -0.2127, -0.1673, -0.1668, -0.1428, -0.1126, -0.0456, -0.0034, 0.0, 0.0067, 0.0141, 0.0705, 0.0314, 0.0434, 0.0548, 0.025, -0.006, -0.0032, 0.0389, 0.0605, 0.1545, 0.1647, 0.1913, 0.2111, 0.1777, 0.2465, 0.2198, 0.1518, 0.2226, 0.1864, 0.2173, 0.2845, 0.2836, 0.3114, 0.283, 0.2588, 0.311, 0.3891, 0.3959, 0.4074, 0.4374, 0.4207, 0.4192, 0.4885, 0.3828, 0.5063, 0.5116, 0.509, 0.6065, 0.5292, 0.5532, 0.5028, 0.5655, 0.5345, 0.4492, 0.4329, 0.2772, 0.3152, 0.4202, 0.424, 0.4604, 0.4604, 0.4998, 0.5251, 0.5189, 0.4851, 0.4375, 0.5045, 0.566, 0.6009, 0.6725, 0.6392, 0.5294, 0.5063, 0.4747, 0.4303, 0.5764, 0.6865, 0.7542, 0.7522, 0.8182, 0.8182, 0.8062, 0.8677, 0.8567, 0.8109, 0.8109, 1.0013, 0.9806, 1.179, 1.1544, 1.0749, 1.1896, 1.1945, 1.1454, 1.1151, 1.1359, 1.3016, 1.3016, 1.3159, 1.4689, 1.5226, 1.5357, 1.4687, 1.6029, 1.7618, 1.765, 1.6324, 1.7778, 1.6613, 1.4072, 1.4294, 1.5043, 1.4333, 1.3682, 1.6036, 1.6266, 1.6119, 1.6119, 1.5366, 1.6709, 1.648, 1.7167, 1.671, 1.6522, 1.722, 1.6367, 1.6164, 1.6183, 1.409, 1.5428, 1.5192, 1.3495, 1.4702, 1.5577]}
{"unit_id": "quote:1963506999167258836", "ticker_idx": 0, "ts": "2025-09-04T07:38:43+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Chinese CSPs will choose Nvidia chips not because of supply shortages, but because of performance", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish NVDA: China demand to shift from supply-constrained to performance-driven, sustaining Chinese CSP adoption.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "However, the fact that Chinese CSPs are choosing Nvidia chips due to supply constraints is expected to change by next year.\n\nAs the FT previously reported, China is planning to expand its fab capacity to produce an enormous number of semiconductor chips.\n\nChinese CSPs will choose Nvidia chips not because of supply shortages, but because of performance.", "reply_text": null, "parent_text": null, "quoted_text": "Nvidia is bound to succeed in China.\n\nChina’s competing chips are still inferior to Nvidia’s, and now Nvidia is about to launch a chip in China that is six times more powerful than the H20.\n\n$NVDA https://t.co/DiWOmLlsNw", "parent_author": null, "ret_m1d": -0.006058477205052815, "ret_signed_m1d": -0.006058477205052815, "alpha_spy_m1d": 0.0022296166490474167, "alpha_c_m1d": 0.005774759335373303, "bench_spy_m1d": -0.008288093854100231, "bench_c_m1d": -0.011833236540426118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027030129068697395, "ret_signed_p1d": -0.027030129068697395, "alpha_spy_p1d": -0.02413398058016225, "alpha_c_p1d": -0.03882904188315939, "bench_spy_p1d": -0.002896148488535144, "bench_c_p1d": 0.011798912814461993, "ret_p1w": 0.03215653285758835, "ret_signed_p1w": 0.03215653285758835, "alpha_spy_p1w": 0.019046297719764738, "alpha_c_p1w": -0.016384186012191604, "bench_spy_p1w": 0.013110235137823611, "bench_c_p1w": 0.04854071886977995, "ret_p1m": 0.0930360228950089, "ret_signed_p1m": 0.0930360228950089, "alpha_spy_p1m": 0.05922802667284133, "alpha_c_p1m": -0.06849026976555406, "bench_spy_p1m": 0.03380799622216757, "bench_c_p1m": 0.16152629266056295, "ret_p3m": 0.057149151572260415, "ret_signed_p3m": 0.057149151572260415, "alpha_spy_p3m": 0.004308997310772611, "alpha_c_p3m": -0.18286335704959544, "bench_spy_p3m": 0.052840154261487804, "bench_c_p3m": 0.24001250862185586, "ret_p6m": 0.03233051870988013, "ret_signed_p6m": 0.03233051870988013, "alpha_spy_p6m": -0.0305308479285602, "alpha_c_p6m": -0.3739845610547625, "bench_spy_p6m": 0.06286136663844033, "bench_c_p6m": 0.40631507976464265, "price_path": [-0.1692, -0.1614, -0.1679, -0.1553, -0.173, -0.1571, -0.1604, -0.1525, -0.1525, -0.162, -0.1601, -0.1384, -0.1011, -0.0969, -0.081, -0.0796, -0.107, -0.0839, -0.0718, -0.0718, -0.0782, -0.0679, -0.0511, -0.044, -0.0393, -0.0442, -0.0056, -0.0017, 0.0078, 0.0044, -0.0016, -0.027, -0.0051, 0.0121, 0.0107, 0.0297, 0.0224, 0.0443, 0.0362, 0.012, 0.0486, 0.0384, 0.0452, 0.0531, 0.0643, 0.0606, 0.067, 0.0578, 0.0604, 0.0512, 0.0603, 0.0232, 0.0218, 0.0193, 0.0369, 0.0475, 0.0589, 0.0579, 0.0496, 0.0147, 0.0147, -0.0051, -0.0061, 0.0, -0.027, -0.0195, -0.0052, 0.033, 0.0322, 0.0359, 0.0355, 0.0188, -0.0079, 0.0267, 0.0292, 0.0697, 0.0395, 0.031, 0.0352, 0.0381, 0.0594, 0.087, 0.0908, 0.1004, 0.093, 0.0809, 0.078, 0.1017, 0.1219, 0.0671, 0.0971, 0.0488, 0.0477, 0.0592, 0.0674, 0.064, 0.0554, 0.0503, 0.0612, 0.0851, 0.1156, 0.1712, 0.2062, 0.182, 0.1797, 0.2052, 0.1575, 0.1373, 0.0957, 0.0961, 0.1596, 0.1253, 0.129, 0.0886, 0.1079, 0.0871, 0.0566, 0.0866, 0.0524, 0.0421, 0.0635, 0.0359, 0.0502, 0.0502, 0.0312, 0.0482, 0.0571, 0.0463, 0.0684, 0.0627, 0.081, 0.0777, 0.0707, 0.0541, 0.0197, 0.0271, 0.0354, -0.0041, 0.0146, 0.0545, 0.0702, 0.1024, 0.0989, 0.0989, 0.1101, 0.0966, 0.0926, 0.0866, 0.0866, 0.1003, 0.096, 0.0909, 0.1018, 0.0781, 0.077, 0.0775, 0.0826, 0.067, 0.0898, 0.085, 0.085, 0.0375, 0.068, 0.0769, 0.0934, 0.0864, 0.0983, 0.1158, 0.1216, 0.1135, 0.0814, 0.0507, 0.0149, 0.0014, 0.0802, 0.1072, 0.0985, 0.1073, 0.0891, 0.0651, 0.0651, 0.0777, 0.0952, 0.0947, 0.1059, 0.116, 0.1236, 0.1394, 0.0772, 0.0323]}
{"unit_id": "orig:1957273211642712418", "ticker_idx": 0, "ts": "2025-08-18T02:47:52+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bank of America's analyst team led by Wamsi Mohan reaffirmed a \"Buy\" rating on Apple and maintained a $250 price target", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses BofA Buy on Apple with $250 PT, as services overtake iPhone in gross profit contribution.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Why BofA Is Bullish on Apple: Apple Is No Longer Just a Hardware Company but a Service Empire\n\nApple’s profit core is quietly shifting from hardware sales to the services business.\n\nBank of America pointed out in a research report released on August 15 that Apple is undergoing a meaningful transformation. The report projects that starting in fiscal year 2025, services will overtake the iPhone to become the company’s largest contributor to gross profit.\n\nAccording to the report, in fiscal year 2025 the services business is expected to account for 42% of Apple’s annual gross profit, with the iPhone trailing at 41%. This marks the first historic moment when services surpass the iPhone.\n\nMore importantly, this gap is expected to widen further going forward. In fiscal year 2027, the contribution from services is projected to expand to 44%, while the iPhone will decline to 39%.\n\nTwo Key Drivers\n\n1. Higher Profitability\nThe profitability of services far exceeds that of hardware. As of the third quarter of fiscal year 2025, the gross margin of the services business reaches 75.6%, while the products business is only 34.5%.\n\n2. Faster Growth Rate\nBank of America’s model expects services revenue to maintain a “low double-digit” (about 12%) annual growth rate over the next few years, while iPhone revenue growth is expected to be in the “mid single digits” (about 6%).\n\nThanks to higher margins and faster revenue growth, the services business is likely to continue contributing more to the annual year-over-year increase in gross profit than the products business.\n\n⸻\n\nFactors That Raise the Valuation Multiple\n\nBecause the services business has higher margins and lower cyclicality, the analysis is that as its profit contribution grows, Apple deserves a higher valuation. This is the key logic behind Bank of America’s $250 price target and maintained “Buy” rating.\n\n⸻\n\nNature of Earnings: From Cyclical to Secular\n\nThe report emphasizes that iPhone sales exhibit a degree of “cyclicality,” being susceptible to product launch cycles and the macro consumption environment. In contrast, services revenue is regarded as “secular,” with more stable and predictable growth.\n\nIn addition, thanks to the optimization of the business mix, the company’s overall gross margin has been steadily moving toward 50%, after hovering around 40% for more than a decade. As the share of services grows, cash flow and earnings elasticity are significantly strengthened.\n\nThe report summarizes this as follows:\n\n“If a higher share of gross profit comes from the stable and secular services business, it naturally deserves a higher valuation multiple.”\n\n⸻\n\nConclusion\n\nBank of America’s analyst team led by Wamsi Mohan reaffirmed a “Buy” rating on Apple and maintained a $250 price target.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0030317614556589767, "ret_signed_m1d": 0.0030317614556589767, "alpha_spy_m1d": 0.0028140826447153877, "alpha_c_m1d": 0.005172399117579052, "bench_spy_m1d": 0.00021767881094358899, "bench_c_m1d": -0.002140637661920075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0014291945916864446, "ret_signed_p1d": -0.0014291945916864446, "alpha_spy_p1d": 0.003995935474991219, "alpha_c_p1d": 0.016034229181152382, "bench_spy_p1d": -0.005425130066677664, "bench_c_p1d": -0.017463423772838826, "ret_p1w": -0.01615477001067933, "ret_signed_p1w": -0.01615477001067933, "alpha_spy_p1w": -0.01486448483899816, "alpha_c_p1w": 0.0005200401777525343, "bench_spy_p1w": -0.0012902851716811714, "bench_c_p1w": -0.016674810188431866, "ret_p1m": 0.031443606493189336, "ret_signed_p1m": 0.031443606493189336, "alpha_spy_p1m": 0.005483662073526796, "alpha_c_p1m": 0.008534456665354018, "bench_spy_p1m": 0.02595994441966254, "bench_c_p1m": 0.022909149827835318, "ret_p3m": 0.1833108245402355, "ret_signed_p3m": 0.1833108245402355, "alpha_spy_p3m": 0.13573867525715944, "alpha_c_p3m": 0.10561403190636742, "bench_spy_p3m": 0.04757214928307607, "bench_c_p3m": 0.07769679263386808, "ret_p6m": 0.1875856174710473, "ret_signed_p6m": 0.1875856174710473, "alpha_spy_p6m": 0.10552491579703505, "alpha_c_p6m": 0.11386560993128914, "bench_spy_p6m": 0.08206070167401225, "bench_c_p6m": 0.07372000753975816, "price_path": [-0.1257, -0.1289, -0.1552, -0.1552, -0.1339, -0.133, -0.135, -0.1311, -0.1274, -0.1206, -0.1226, -0.132, -0.1178, -0.1285, -0.1232, -0.14, -0.1382, -0.1501, -0.1416, -0.1536, -0.1496, -0.1496, -0.1304, -0.1283, -0.1335, -0.128, -0.1304, -0.1301, -0.1124, -0.1009, -0.081, -0.0761, -0.0761, -0.0917, -0.0915, -0.0866, -0.0811, -0.0865, -0.0975, -0.0954, -0.0908, -0.0914, -0.0864, -0.0808, -0.0725, -0.0736, -0.0752, -0.0747, -0.074, -0.086, -0.0956, -0.102, -0.1245, -0.1203, -0.1221, -0.0774, -0.0481, -0.0078, -0.0161, -0.0054, 0.0106, 0.0082, 0.003, 0.0, -0.0014, -0.0211, -0.0259, -0.0136, -0.0162, -0.0068, -0.0017, 0.0072, 0.0054, 0.0054, -0.0051, 0.0328, 0.0385, 0.0381, 0.0303, 0.015, -0.0178, -0.0037, 0.0138, 0.0252, 0.0314, 0.0351, 0.0303, 0.0633, 0.1091, 0.102, 0.0928, 0.1125, 0.1064, 0.102, 0.1028, 0.1064, 0.1136, 0.1175, 0.1117, 0.1108, 0.1177, 0.1003, 0.0623, 0.0726, 0.0731, 0.0799, 0.0717, 0.0927, 0.1358, 0.1381, 0.1194, 0.1243, 0.1383, 0.1642, 0.1651, 0.1681, 0.1755, 0.171, 0.1653, 0.1696, 0.17, 0.1684, 0.1628, 0.1681, 0.1933, 0.1856, 0.1833, 0.181, 0.1595, 0.1594, 0.1643, 0.1543, 0.177, 0.1962, 0.2007, 0.2033, 0.2033, 0.2089, 0.2273, 0.2407, 0.2319, 0.2169, 0.2086, 0.2047, 0.2016, 0.2086, 0.2053, 0.2064, 0.1883, 0.1905, 0.1785, 0.18, 0.1864, 0.1747, 0.1808, 0.187, 0.187, 0.1853, 0.1868, 0.1839, 0.1786, 0.1786, 0.1749, 0.1586, 0.1374, 0.1286, 0.123, 0.1244, 0.1283, 0.1317, 0.127, 0.1194, 0.1078, 0.1078, 0.0695, 0.0736, 0.0767, 0.0753, 0.1073, 0.1197, 0.1117, 0.1197, 0.1249, 0.1706, 0.1683, 0.1987, 0.1961, 0.2057, 0.1917, 0.1876]}
{"unit_id": "thread:1957606932346466544", "ticker_idx": 0, "ts": "2025-08-19T00:53:57+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has reached its target yield earlier than planned in the development stage of its next-generation DRAM product, the 10-nano 6th generation (1c) DRAM, and is preparing for initial mass production... could start ramp-up 3 to as much as 6 months ahead of SK hynix", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung 1c DRAM yield ahead of schedule, HBM4 ramp could beat SK Hynix by 3-6 months; foundry wins Tesla/Apple orders and US tariff policy acts as tailwind.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung, Why the Confidence in HBM4 Next Year? Higher-than-expected Yield and Ramp-up Timing\n\nWhile Samsung Electronics is solidifying its foundation for the revival of its semiconductor business, Chairman Lee Jae-yong recently returned from visiting big tech companies in San Francisco and Silicon Valley for two weeks, leaving behind a meaningful remark: “We have prepared for next year’s business.” Analysts interpret this as the company chief personally stepping in to reinforce the customer base for a full revival, as Samsung’s semiconductor division faces difficulties from memory to foundry (semiconductor contract manufacturing).\n\nLee’s mention of next year’s business is widely seen as a step to validate the results Samsung has been preparing since the first half of this year with major customers in the second half, and then begin mass supply starting next year. In fact, after showing sluggish performance for 1–2 years in general-purpose memory, high-bandwidth memory (HBM), and foundry, Samsung has recently aligned the focus of all its businesses toward next year, steadily completing next-generation technologies one by one.\n\n◇ Next-generation DRAM business with repeated redesigns and verification… “Faster than competitors”\n\nAccording to industry sources on the 19th, Samsung Electronics has reached its target yield earlier than planned in the development stage of its next-generation DRAM product, the 10-nano 6th generation (1c) DRAM, and is preparing for initial mass production. Samsung originally expected to begin mass production at the same time as SK hynix or about 1–2 months earlier, but at the current pace, it could start ramp-up (large-scale mass production) 3 to as much as 6 months ahead of SK hynix.\n\nThe reason 1c DRAM is important is because it is mounted on the next-generation HBM, HBM4 (6th generation HBM), and serves as a key material that will determine the success of the HBM market. A source familiar with Samsung Electronics said, “Hwang Sang-jun, head of DRAM development, who enjoys the full trust of Vice Chairman Jun-Young Jeon, has been taking cautious steps, like knocking on a stone bridge multiple times before crossing it,” and added, “To avoid repeating mistakes made with 1a and 1b DRAM, he carried out several rounds of redesign and verification early on, and recently, encouraging results have been reported internally.”\n\nSamsung, which repeatedly stumbled with its HBM3E (5th generation HBM) 12-layer product, is also seeing momentum toward supplying its largest customer, Nvidia. Global securities firms such as Morgan Stanley (U.S.) and GF Securities (Hong Kong) are lending weight to the “August test pass” speculation regarding Samsung’s HBM3E. In the case of HBM4, it will be completed through Samsung Electronics’ foundry division process, making it a critical turning point where the memory and foundry businesses can create synergy.\n\nAn industry insider said, “The investment boom in generative artificial intelligence (AI) is creating immediate market reactions and demand for server memory technology upgrades,” adding, “If Samsung succeeds in mass-producing state-of-the-art DRAM based on 1c technology, it could demonstrate a ‘scale’ advantage in production capacity over SK hynix and Micron, thereby monopolizing larger profits.”\n\n◇ Will foundry turn into a cash cow?\n\nThe foundry division, which has been cited as the cause of Samsung Electronics’ semiconductor slump with quarterly losses in the trillions of won, has recently shown signs of recovery. Right before Lee’s departure for the U.S., Samsung announced a 22.7648 trillion won foundry contract with Tesla, and during his U.S. trip, news also emerged of winning orders from Apple. While Apple did not specify the type of chip, the industry expects it to be an image sensor (CIS) for the next-generation iPhone.\n\nPresident Trump’s semiconductor tariff policy has, in effect, become advantageous for Samsung Electronics, which already operates fabs in the U.S. Unlike TSMC, which is rushing to build fabs in Arizona and elsewhere, Samsung has been running its Texas foundry for over a decade. More recently, it is preparing to operate its Taylor plant. Having cooperated with local companies for a long time, Samsung is said to have gained an edge in securing orders for automotive chips and other sensor products.\n\nIn 2014, when Samsung beat out TSMC to secure Qualcomm’s Snapdragon 14-nano mobile application processor (AP) order, it was also evaluated as a successful result of its localization strategy. At that time, Samsung signed a production license agreement with U.S. foundry company GlobalFoundries (GF) to expand production capacity and stabilize manufacturing, an unconventional move that helped secure the deal with Qualcomm.\n\nNot only in Taiwan and Japan, but also in Europe, many semiconductor companies have relied on either in-house production or outsourcing to TSMC. However, with the launch of the Trump administration’s second term, it is highly likely they will look for partners to manufacture chips locally in the U.S. In Japan and Europe, there are numerous companies designing chips in fields ranging from sensors to automotive, consumer electronics, and security. While Samsung Foundry has traditionally focused on major customers and advanced process orders, it is said to have recently shifted course, building a solid performance base by expanding its portfolio through mature processes.\n\n---\n\nhttps://t.co/hRoUdxbbgu", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.0054547226465127, "alpha_c_m1d": -0.017773815443997476, "bench_spy_m1d": 0.0054547226465127, "bench_c_m1d": 0.017773815443997476, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00714291197185335, "ret_signed_p1d": 0.00714291197185335, "alpha_spy_p1d": 0.0097999956025, "alpha_c_p1d": 0.013908524026322144, "bench_spy_p1d": -0.0026570836306466505, "bench_c_p1d": -0.006765612054468795, "ret_p1w": 0.004285837489693689, "ret_signed_p1w": 0.004285837489693689, "alpha_spy_p1w": -0.004076005369385571, "alpha_c_p1w": -0.0011037198608323529, "bench_spy_p1w": 0.00836184285907926, "bench_c_p1w": 0.005389557350526042, "ret_p1m": 0.11714287584922056, "ret_signed_p1m": 0.11714287584922056, "alpha_spy_p1m": 0.0868682464041084, "alpha_c_p1m": 0.07964588293105934, "bench_spy_p1m": 0.030274629445112167, "bench_c_p1m": 0.03749699291816122, "ret_p3m": 0.3947668690437165, "ret_signed_p3m": 0.3947668690437165, "alpha_spy_p3m": 0.34165280724225444, "alpha_c_p3m": 0.29194469784981547, "bench_spy_p3m": 0.053114061801462054, "bench_c_p3m": 0.10282217119390102, "ret_p6m": 1.419543157063916, "ret_signed_p6m": 1.419543157063916, "alpha_spy_p6m": 1.3318314978263535, "alpha_c_p6m": 1.3235192746852633, "bench_spy_p6m": 0.08771165923756263, "bench_c_p6m": 0.09602388237865278, "price_path": [-0.2233, -0.2304, -0.2233, -0.2347, -0.2063, -0.2035, -0.202, -0.1935, -0.1935, -0.1793, -0.1609, -0.1609, -0.1509, -0.1594, -0.1495, -0.1552, -0.1722, -0.1878, -0.1751, -0.1509, -0.1594, -0.1552, -0.1765, -0.141, -0.1296, -0.1452, -0.1314, -0.1457, -0.14, -0.1314, -0.0886, -0.0957, -0.1186, -0.1229, -0.1371, -0.1286, -0.1057, -0.1071, -0.09, -0.0757, -0.0471, -0.0414, -0.0314, -0.0571, -0.0514, -0.0571, -0.0586, 0.0057, 0.0086, 0.0371, 0.02, -0.0157, -0.0043, -0.0014, -0.0171, 0.0071, 0.0257, 0.0143, 0.0157, 0.0271, 0.0229, 0.0229, 0.0, 0.0, 0.0071, 0.0086, 0.02, 0.0214, 0.0043, 0.0086, -0.0057, -0.0043, -0.0343, -0.0129, -0.0029, 0.0014, -0.0071, 0.0014, 0.0214, 0.0371, 0.0486, 0.0771, 0.0929, 0.1343, 0.1171, 0.1471, 0.1471, 0.1929, 0.21, 0.22, 0.23, 0.19, 0.2082, 0.2039, 0.2341, 0.2879, 0.2879, 0.2879, 0.2879, 0.2879, 0.2879, 0.3546, 0.3388, 0.3144, 0.3632, 0.4019, 0.4048, 0.4077, 0.3991, 0.4149, 0.3847, 0.4177, 0.4636, 0.4278, 0.4421, 0.4938, 0.5426, 0.5942, 0.5053, 0.4436, 0.4235, 0.4048, 0.4436, 0.4852, 0.4794, 0.4751, 0.3948, 0.4436, 0.4034, 0.3847, 0.4436, 0.3603, 0.3876, 0.4249, 0.4751, 0.4852, 0.4421, 0.4464, 0.4837, 0.4995, 0.5081, 0.5555, 0.5713, 0.5555, 0.5497, 0.5397, 0.5627, 0.5038, 0.4751, 0.5483, 0.544, 0.5253, 0.5856, 0.6, 0.5942, 0.5942, 0.6789, 0.7231, 0.7289, 0.7289, 0.7289, 0.8529, 0.9913, 1.0028, 1.0331, 1.0014, 1.0043, 1.0014, 0.9841, 1.023, 1.0749, 1.147, 1.1528, 1.0937, 1.1557, 1.196, 1.1932, 1.1932, 1.2999, 1.3417, 1.3172, 1.3143, 1.1686, 1.4152, 1.4383, 1.297, 1.2869, 1.3994, 1.3907, 1.4195]}
{"unit_id": "thread:1957606932346466544", "ticker_idx": 1, "ts": "2025-08-19T00:53:57+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung originally expected to begin mass production at the same time as SK hynix or about 1–2 months earlier, but at the current pace, it could start ramp-up 3 to as much as 6 months ahead of SK hynix", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung 1c DRAM yield ahead of schedule, HBM4 ramp could beat SK Hynix by 3-6 months; foundry wins Tesla/Apple orders and US tariff policy acts as tailwind.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung, Why the Confidence in HBM4 Next Year? Higher-than-expected Yield and Ramp-up Timing\n\nWhile Samsung Electronics is solidifying its foundation for the revival of its semiconductor business, Chairman Lee Jae-yong recently returned from visiting big tech companies in San Francisco and Silicon Valley for two weeks, leaving behind a meaningful remark: “We have prepared for next year’s business.” Analysts interpret this as the company chief personally stepping in to reinforce the customer base for a full revival, as Samsung’s semiconductor division faces difficulties from memory to foundry (semiconductor contract manufacturing).\n\nLee’s mention of next year’s business is widely seen as a step to validate the results Samsung has been preparing since the first half of this year with major customers in the second half, and then begin mass supply starting next year. In fact, after showing sluggish performance for 1–2 years in general-purpose memory, high-bandwidth memory (HBM), and foundry, Samsung has recently aligned the focus of all its businesses toward next year, steadily completing next-generation technologies one by one.\n\n◇ Next-generation DRAM business with repeated redesigns and verification… “Faster than competitors”\n\nAccording to industry sources on the 19th, Samsung Electronics has reached its target yield earlier than planned in the development stage of its next-generation DRAM product, the 10-nano 6th generation (1c) DRAM, and is preparing for initial mass production. Samsung originally expected to begin mass production at the same time as SK hynix or about 1–2 months earlier, but at the current pace, it could start ramp-up (large-scale mass production) 3 to as much as 6 months ahead of SK hynix.\n\nThe reason 1c DRAM is important is because it is mounted on the next-generation HBM, HBM4 (6th generation HBM), and serves as a key material that will determine the success of the HBM market. A source familiar with Samsung Electronics said, “Hwang Sang-jun, head of DRAM development, who enjoys the full trust of Vice Chairman Jun-Young Jeon, has been taking cautious steps, like knocking on a stone bridge multiple times before crossing it,” and added, “To avoid repeating mistakes made with 1a and 1b DRAM, he carried out several rounds of redesign and verification early on, and recently, encouraging results have been reported internally.”\n\nSamsung, which repeatedly stumbled with its HBM3E (5th generation HBM) 12-layer product, is also seeing momentum toward supplying its largest customer, Nvidia. Global securities firms such as Morgan Stanley (U.S.) and GF Securities (Hong Kong) are lending weight to the “August test pass” speculation regarding Samsung’s HBM3E. In the case of HBM4, it will be completed through Samsung Electronics’ foundry division process, making it a critical turning point where the memory and foundry businesses can create synergy.\n\nAn industry insider said, “The investment boom in generative artificial intelligence (AI) is creating immediate market reactions and demand for server memory technology upgrades,” adding, “If Samsung succeeds in mass-producing state-of-the-art DRAM based on 1c technology, it could demonstrate a ‘scale’ advantage in production capacity over SK hynix and Micron, thereby monopolizing larger profits.”\n\n◇ Will foundry turn into a cash cow?\n\nThe foundry division, which has been cited as the cause of Samsung Electronics’ semiconductor slump with quarterly losses in the trillions of won, has recently shown signs of recovery. Right before Lee’s departure for the U.S., Samsung announced a 22.7648 trillion won foundry contract with Tesla, and during his U.S. trip, news also emerged of winning orders from Apple. While Apple did not specify the type of chip, the industry expects it to be an image sensor (CIS) for the next-generation iPhone.\n\nPresident Trump’s semiconductor tariff policy has, in effect, become advantageous for Samsung Electronics, which already operates fabs in the U.S. Unlike TSMC, which is rushing to build fabs in Arizona and elsewhere, Samsung has been running its Texas foundry for over a decade. More recently, it is preparing to operate its Taylor plant. Having cooperated with local companies for a long time, Samsung is said to have gained an edge in securing orders for automotive chips and other sensor products.\n\nIn 2014, when Samsung beat out TSMC to secure Qualcomm’s Snapdragon 14-nano mobile application processor (AP) order, it was also evaluated as a successful result of its localization strategy. At that time, Samsung signed a production license agreement with U.S. foundry company GlobalFoundries (GF) to expand production capacity and stabilize manufacturing, an unconventional move that helped secure the deal with Qualcomm.\n\nNot only in Taiwan and Japan, but also in Europe, many semiconductor companies have relied on either in-house production or outsourcing to TSMC. However, with the launch of the Trump administration’s second term, it is highly likely they will look for partners to manufacture chips locally in the U.S. In Japan and Europe, there are numerous companies designing chips in fields ranging from sensors to automotive, consumer electronics, and security. While Samsung Foundry has traditionally focused on major customers and advanced process orders, it is said to have recently shifted course, building a solid performance base by expanding its portfolio through mature processes.\n\n---\n\nhttps://t.co/hRoUdxbbgu", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017110114290354028, "ret_signed_m1d": -0.017110114290354028, "alpha_spy_m1d": -0.011655391643841329, "alpha_c_m1d": 0.003483228159278262, "bench_spy_m1d": 0.0054547226465127, "bench_c_m1d": 0.02059334244963229, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0285171355588133, "ret_signed_p1d": 0.0285171355588133, "alpha_spy_p1d": 0.02586005192816665, "alpha_c_p1d": 0.021950600837470358, "bench_spy_p1d": -0.0026570836306466505, "bench_c_p1d": -0.006566534721342943, "ret_p1w": -0.005703450975324653, "ret_signed_p1w": 0.005703450975324653, "alpha_spy_p1w": 0.014065293834403914, "alpha_c_p1w": 0.025918649093801482, "bench_spy_p1w": 0.00836184285907926, "bench_c_p1w": 0.02021519811847683, "ret_p1m": 0.2698924117273418, "ret_signed_p1m": -0.2698924117273418, "alpha_spy_p1m": -0.23961778228222963, "alpha_c_p1m": -0.22045453964116057, "bench_spy_p1m": 0.030274629445112167, "bench_c_p1m": 0.04943787208618122, "ret_p3m": 1.1323529227523972, "ret_signed_p3m": -1.1323529227523972, "alpha_spy_p3m": -1.0792388609509351, "alpha_c_p3m": -0.9463247506313694, "bench_spy_p3m": 0.053114061801462054, "bench_c_p3m": 0.18602817212102774, "ret_p6m": 2.277030241755184, "ret_signed_p6m": -2.277030241755184, "alpha_spy_p6m": -2.189318582517622, "alpha_c_p6m": -1.8465952747394103, "bench_spy_p6m": 0.08771165923756263, "bench_c_p6m": 0.4304349670157739, "price_path": [-0.2527, -0.2409, -0.2295, -0.2314, -0.2106, -0.1939, -0.2224, -0.211, -0.211, -0.173, -0.1464, -0.1464, -0.1293, -0.1236, -0.0875, -0.1046, -0.1046, -0.057, -0.0532, -0.0627, -0.0646, -0.0228, -0.0133, 0.0589, 0.0875, 0.1141, 0.0798, 0.1103, 0.0856, 0.0608, 0.0589, 0.0285, 0.0304, 0.0722, 0.0684, 0.1293, 0.1198, 0.1407, 0.135, 0.1255, 0.0247, 0.0228, 0.0361, 0.0209, 0.0228, 0.0247, 0.0114, -0.0038, -0.0019, 0.0019, 0.0399, -0.019, -0.019, 0.0019, -0.0171, -0.0038, -0.0247, 0.0152, 0.0228, 0.057, 0.0513, 0.0513, 0.0171, 0.0, -0.0285, -0.0684, -0.0456, -0.0133, -0.0057, -0.0114, 0.0224, 0.0243, -0.0252, -0.0081, -0.0005, 0.011, 0.0414, 0.0548, 0.0966, 0.1576, 0.169, 0.2509, 0.2604, 0.3251, 0.2699, 0.3537, 0.3537, 0.3365, 0.3746, 0.3613, 0.3575, 0.2813, 0.3289, 0.3232, 0.3708, 0.506, 0.506, 0.506, 0.506, 0.506, 0.506, 0.6297, 0.5802, 0.5669, 0.6088, 0.723, 0.7725, 0.8487, 0.8239, 0.8334, 0.822, 0.942, 1.0372, 0.9838, 1.1247, 1.1628, 1.1285, 1.3608, 1.2314, 1.2047, 1.258, 1.2085, 1.3075, 1.357, 1.3494, 1.3304, 1.1324, 1.3075, 1.1704, 1.14, 1.1742, 0.9838, 0.98, 0.9762, 0.9953, 1.0729, 1.0196, 1.05, 1.1263, 1.1034, 1.0653, 1.0729, 1.1987, 1.1567, 1.2368, 1.1529, 1.1758, 1.111, 1.0196, 1.0996, 1.1034, 1.0843, 1.2101, 1.2253, 1.2406, 1.2406, 1.2825, 1.4387, 1.4806, 1.4806, 1.4806, 1.5797, 1.6521, 1.7664, 1.8274, 1.8807, 1.835, 1.8541, 1.8121, 1.8274, 1.8541, 1.8807, 1.9112, 1.8312, 1.8198, 1.8769, 1.9227, 1.8045, 2.0484, 2.2046, 2.2808, 2.4637, 2.1627, 2.4561, 2.4295, 2.2084, 2.197, 2.3799, 2.338, 2.277]}
{"unit_id": "orig:1955808837258367336", "ticker_idx": 0, "ts": "2025-08-14T01:48:58+00:00", "symbol": "semiconductors", "kind": "sector", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS expects a stronger-than-anticipated performance in the fourth quarter of 2025 (4Q25) and is significantly raising its DDR and NAND contract price assumptions, reaffirming its view of a modest up-cycle leading into 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DDR/NAND price forecasts to +5% QoQ in 4Q25 and sees a modest up-cycle into 2026; bullish memory sector, mixed on individual names with Samsung lagging on HBM certification.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "DRAM semiconductor basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Memory Commentary: Returning to a Modest Up-cycle Before 2026\nUBS (N. Gaudois, 12/08/25)\n\nUBS expects a stronger-than-anticipated performance in the fourth quarter of 2025 (4Q25) and is significantly raising its DDR and NAND contract price assumptions, reaffirming its view of a modest up-cycle leading into 2026. Growth in HBM demand, improving end-market demand, and capacity constraints at some competitors will support price trends.\n\n4Q25 Pricing Expectations Raised\nIn 3Q25, average contract prices for both DDR and NAND rose by approximately 3% quarter-over-quarter (QoQ). Current inventory levels are around 9-10 weeks for PC/mobile DDR, 13-14 weeks for server DDR, and 8-13 weeks for NAND. Demand from smartphones and PCs is stable, with initial signs for Q4 being positive. A decreased risk from U.S. Section 232 tariffs also provides support. A recovery in demand from hyperscale customers for DDR and enterprise SSDs has prompted UBS to raise its 4Q25 QoQ contract price forecast for DDR and NAND from the previous -3% and -5%, respectively, to +5%.\n\nModest Price Up-cycle Before 2026\nUBS forecasts that DDR and NAND contract prices will continue to see small sequential increases each quarter until 3Q26. The reasons for this include: HBM demand diverting production capacity, improving end-market demand (e.g., servers), and limited yields for CXMT's 16nm DDR5 making a repeat of last year's DDR4 oversupply unlikely. In the long term, the rising proportion of HBM should help mitigate the volatility of traditional memory cycles. This current DDR contract price cycle is expected to see a 23% increase from trough to peak. The next down-cycle (2027-28) may see a 43% decline, followed by a projected 28% increase in the subsequent up-cycle (2028-29).\n\nHBM Demand Forecasts Raised\nThe 2025 HBM bit demand forecast has been raised by 4% to 16.9 billion Gb (+96% YoY), and the 2026 forecast is raised by 3% to 26.1 billion Gb (+55% YoY). This is driven by upward revisions to shipment expectations for Nvidia's AI GPUs, especially Blackwell Ultra and H20. Nvidia's AI GPU procurement volume is estimated to be 6.5 million units in 2025 and 7.0 million units in 2026.\n\nSupply-Side Dynamics\nSK hynix's 2026 HBM bit shipment forecast is revised down to 16.1 billion Gb (+27% YoY), primarily due to constraints on its capacity expansion. Micron's shipments for 2025 and 2026 are projected to be 5.3 billion and 8.9 billion Gb (+67% YoY), respectively, with its new Boise factory not expected to produce its first wafers until the second half of 2027. Samsung's shipments for 2025 and 2026 are projected at 4.6 billion and 7.6 billion Gb (-9% and +63% YoY, respectively); it has not yet received Nvidia's certification for HBM3E 12-Hi, and its HBM4 wafer input is not expected to begin before January-February 2026. UBS believes the HBM supply-demand situation will be largely balanced in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0004912311710793352, "ret_signed_m1d": 0.0004912311710793352, "alpha_spy_m1d": 0.0005841879841456491, "alpha_c_m1d": 0.0026109506048285582, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011759388231049753, "ret_signed_p1d": -0.011759388231049753, "alpha_spy_p1d": -0.009418193589026994, "alpha_c_p1d": 0.00900769753292426, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.0679048657492165, "ret_signed_p1w": -0.0679048657492165, "alpha_spy_p1w": -0.05333011549191513, "alpha_c_p1w": -0.020011333134271947, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.16592749470154425, "ret_signed_p1m": 0.16592749470154425, "alpha_spy_p1m": 0.14660812944960258, "alpha_c_p1m": 0.1579782933427986, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.8731774285061729, "ret_signed_p3m": 0.8731774285061729, "alpha_spy_p3m": 0.8112446177670078, "alpha_c_p3m": 0.7104522070489261, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 1.8100526364114333, "ret_signed_p6m": 1.8100526364114333, "alpha_spy_p6m": 1.7330992835527155, "alpha_c_p6m": 1.4755939435643153, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2396, -0.2375, -0.2463, -0.2579, -0.2604, -0.2544, -0.2507, -0.2354, -0.2275, -0.2423, -0.2261, -0.2152, -0.1959, -0.1734, -0.1673, -0.1523, -0.1448, -0.1251, -0.132, -0.1391, -0.1178, -0.1111, -0.1023, -0.1057, -0.0863, -0.0943, -0.0443, -0.0333, -0.0333, -0.043, -0.042, -0.0542, -0.057, -0.0422, -0.0542, -0.067, -0.0432, -0.0548, -0.0304, -0.0222, -0.0318, -0.024, -0.0322, -0.0633, -0.059, -0.0546, -0.0785, -0.0744, -0.0706, -0.0765, -0.0604, -0.057, -0.0391, -0.0475, -0.0892, -0.0778, -0.0668, -0.0787, -0.0583, -0.0402, -0.0184, -0.0048, 0.0005, 0.0, -0.0118, -0.0229, -0.0323, -0.0519, -0.0679, -0.0519, -0.0446, -0.0475, -0.0446, -0.0272, -0.0341, -0.0596, -0.0486, -0.0423, -0.0227, 0.0033, 0.0105, 0.0404, 0.0775, 0.1129, 0.1659, 0.1755, 0.2123, 0.1923, 0.2524, 0.236, 0.2505, 0.2729, 0.2594, 0.2485, 0.2125, 0.2511, 0.257, 0.3214, 0.386, 0.3972, 0.4055, 0.3915, 0.4204, 0.4092, 0.4416, 0.4505, 0.4231, 0.4653, 0.5423, 0.5586, 0.5953, 0.5728, 0.5707, 0.5792, 0.6608, 0.7088, 0.685, 0.747, 0.7689, 0.7733, 0.8929, 0.7784, 0.8017, 0.8143, 0.7914, 0.8764, 0.8732, 0.879, 0.8504, 0.7877, 0.8462, 0.7538, 0.7312, 0.6959, 0.6244, 0.6761, 0.6887, 0.7263, 0.7542, 0.7398, 0.7615, 0.7952, 0.7789, 0.7497, 0.7957, 0.8665, 0.8627, 0.9163, 0.8724, 0.8411, 0.7917, 0.74, 0.7706, 0.8317, 0.8659, 0.9538, 0.9624, 0.9931, 0.9931, 1.029, 1.1187, 1.1293, 1.11, 1.11, 1.2618, 1.3211, 1.4444, 1.4633, 1.4365, 1.4711, 1.4783, 1.4387, 1.4435, 1.4776, 1.5791, 1.5907, 1.552, 1.6328, 1.6867, 1.7057, 1.6402, 1.8086, 1.9384, 1.956, 1.9573, 1.8755, 1.9999, 1.8924, 1.7855, 1.8101]}
{"unit_id": "orig:1955808837258367336", "ticker_idx": 1, "ts": "2025-08-14T01:48:58+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's shipments for 2025 and 2026 are projected to be 5.3 billion and 8.9 billion Gb (+67% YoY), respectively", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DDR/NAND price forecasts to +5% QoQ in 4Q25 and sees a modest up-cycle into 2026; bullish memory sector, mixed on individual names with Samsung lagging on HBM certification.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Memory Commentary: Returning to a Modest Up-cycle Before 2026\nUBS (N. Gaudois, 12/08/25)\n\nUBS expects a stronger-than-anticipated performance in the fourth quarter of 2025 (4Q25) and is significantly raising its DDR and NAND contract price assumptions, reaffirming its view of a modest up-cycle leading into 2026. Growth in HBM demand, improving end-market demand, and capacity constraints at some competitors will support price trends.\n\n4Q25 Pricing Expectations Raised\nIn 3Q25, average contract prices for both DDR and NAND rose by approximately 3% quarter-over-quarter (QoQ). Current inventory levels are around 9-10 weeks for PC/mobile DDR, 13-14 weeks for server DDR, and 8-13 weeks for NAND. Demand from smartphones and PCs is stable, with initial signs for Q4 being positive. A decreased risk from U.S. Section 232 tariffs also provides support. A recovery in demand from hyperscale customers for DDR and enterprise SSDs has prompted UBS to raise its 4Q25 QoQ contract price forecast for DDR and NAND from the previous -3% and -5%, respectively, to +5%.\n\nModest Price Up-cycle Before 2026\nUBS forecasts that DDR and NAND contract prices will continue to see small sequential increases each quarter until 3Q26. The reasons for this include: HBM demand diverting production capacity, improving end-market demand (e.g., servers), and limited yields for CXMT's 16nm DDR5 making a repeat of last year's DDR4 oversupply unlikely. In the long term, the rising proportion of HBM should help mitigate the volatility of traditional memory cycles. This current DDR contract price cycle is expected to see a 23% increase from trough to peak. The next down-cycle (2027-28) may see a 43% decline, followed by a projected 28% increase in the subsequent up-cycle (2028-29).\n\nHBM Demand Forecasts Raised\nThe 2025 HBM bit demand forecast has been raised by 4% to 16.9 billion Gb (+96% YoY), and the 2026 forecast is raised by 3% to 26.1 billion Gb (+55% YoY). This is driven by upward revisions to shipment expectations for Nvidia's AI GPUs, especially Blackwell Ultra and H20. Nvidia's AI GPU procurement volume is estimated to be 6.5 million units in 2025 and 7.0 million units in 2026.\n\nSupply-Side Dynamics\nSK hynix's 2026 HBM bit shipment forecast is revised down to 16.1 billion Gb (+27% YoY), primarily due to constraints on its capacity expansion. Micron's shipments for 2025 and 2026 are projected to be 5.3 billion and 8.9 billion Gb (+67% YoY), respectively, with its new Boise factory not expected to produce its first wafers until the second half of 2027. Samsung's shipments for 2025 and 2026 are projected at 4.6 billion and 7.6 billion Gb (-9% and +63% YoY, respectively); it has not yet received Nvidia's certification for HBM3E 12-Hi, and its HBM4 wafer input is not expected to begin before January-February 2026. UBS believes the HBM supply-demand situation will be largely balanced in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008141189984252506, "ret_signed_m1d": -0.008141189984252506, "alpha_spy_m1d": -0.008048233171186192, "alpha_c_m1d": -0.006021470550503283, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03527816469314926, "ret_signed_p1d": -0.03527816469314926, "alpha_spy_p1d": -0.0329369700511265, "alpha_c_p1d": -0.014511078929175247, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.07582409721739869, "ret_signed_p1w": -0.07582409721739869, "alpha_spy_p1w": -0.06124934696009732, "alpha_c_p1w": -0.027930564602454133, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.2549285083860031, "ret_signed_p1m": 0.2549285083860031, "alpha_spy_p1m": 0.23560914313406145, "alpha_c_p1m": 0.24697930702725746, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 0.9256204500769809, "ret_signed_p3m": 0.9256204500769809, "alpha_spy_p3m": 0.8636876393378159, "alpha_c_p3m": 0.7628952286197341, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 2.153457720433874, "ret_signed_p6m": 2.153457720433874, "alpha_spy_p6m": 2.076504367575156, "alpha_c_p6m": 1.8189990275867558, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2134, -0.2178, -0.2358, -0.2438, -0.2555, -0.2555, -0.2315, -0.2331, -0.2281, -0.2468, -0.2171, -0.1847, -0.1767, -0.1524, -0.1343, -0.1153, -0.0899, -0.0748, -0.0736, -0.0782, -0.0444, -0.0404, -0.0286, -0.0286, -0.0144, -0.0265, 0.02, 0.0147, 0.0047, -0.0052, -0.0172, -0.036, -0.0292, -0.0249, -0.0249, -0.0429, -0.0069, -0.0243, -0.0174, -0.0061, -0.0533, -0.0413, -0.0707, -0.096, -0.087, -0.0963, -0.1283, -0.1234, -0.1082, -0.112, -0.1121, -0.1064, -0.0842, -0.1289, -0.1629, -0.1398, -0.1295, -0.1318, -0.1071, -0.0511, -0.0125, 0.0196, -0.0081, 0.0, -0.0353, -0.0139, -0.0259, -0.0645, -0.0758, -0.0607, -0.0708, -0.0702, -0.0602, -0.0263, -0.0501, -0.0501, -0.0544, -0.0524, -0.0086, 0.0485, 0.0492, 0.0794, 0.1174, 0.2018, 0.2549, 0.2592, 0.2676, 0.277, 0.348, 0.2988, 0.3139, 0.3282, 0.2907, 0.2517, 0.2552, 0.3082, 0.3355, 0.4538, 0.4666, 0.5001, 0.5251, 0.483, 0.5697, 0.536, 0.4503, 0.5396, 0.494, 0.5329, 0.6175, 0.6163, 0.6514, 0.6156, 0.5851, 0.6509, 0.7492, 0.7578, 0.7723, 0.81, 0.7891, 0.7871, 0.8744, 0.7413, 0.8968, 0.9034, 0.9001, 1.023, 0.9256, 0.9559, 0.8924, 0.9713, 0.9323, 0.8249, 0.8043, 0.6082, 0.6562, 0.7884, 0.7932, 0.839, 0.839, 0.8886, 0.9204, 0.9127, 0.8701, 0.8101, 0.8946, 0.972, 1.0159, 1.1061, 1.0642, 0.9259, 0.8968, 0.8569, 0.8011, 0.985, 1.1238, 1.209, 1.2064, 1.2896, 1.2896, 1.2745, 1.3519, 1.338, 1.2803, 1.2803, 1.5201, 1.494, 1.7439, 1.7129, 1.6128, 1.7572, 1.7634, 1.7016, 1.6634, 1.6896, 1.8983, 1.8983, 1.9162, 2.1089, 2.1765, 2.1931, 2.1087, 2.2777, 2.4778, 2.4818, 2.3148, 2.4979, 2.3512, 2.0313, 2.0592, 2.1535]}
{"unit_id": "orig:1955808837258367336", "ticker_idx": 2, "ts": "2025-08-14T01:48:58+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix's 2026 HBM bit shipment forecast is revised down to 16.1 billion Gb (+27% YoY)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DDR/NAND price forecasts to +5% QoQ in 4Q25 and sees a modest up-cycle into 2026; bullish memory sector, mixed on individual names with Samsung lagging on HBM certification.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Memory Commentary: Returning to a Modest Up-cycle Before 2026\nUBS (N. Gaudois, 12/08/25)\n\nUBS expects a stronger-than-anticipated performance in the fourth quarter of 2025 (4Q25) and is significantly raising its DDR and NAND contract price assumptions, reaffirming its view of a modest up-cycle leading into 2026. Growth in HBM demand, improving end-market demand, and capacity constraints at some competitors will support price trends.\n\n4Q25 Pricing Expectations Raised\nIn 3Q25, average contract prices for both DDR and NAND rose by approximately 3% quarter-over-quarter (QoQ). Current inventory levels are around 9-10 weeks for PC/mobile DDR, 13-14 weeks for server DDR, and 8-13 weeks for NAND. Demand from smartphones and PCs is stable, with initial signs for Q4 being positive. A decreased risk from U.S. Section 232 tariffs also provides support. A recovery in demand from hyperscale customers for DDR and enterprise SSDs has prompted UBS to raise its 4Q25 QoQ contract price forecast for DDR and NAND from the previous -3% and -5%, respectively, to +5%.\n\nModest Price Up-cycle Before 2026\nUBS forecasts that DDR and NAND contract prices will continue to see small sequential increases each quarter until 3Q26. The reasons for this include: HBM demand diverting production capacity, improving end-market demand (e.g., servers), and limited yields for CXMT's 16nm DDR5 making a repeat of last year's DDR4 oversupply unlikely. In the long term, the rising proportion of HBM should help mitigate the volatility of traditional memory cycles. This current DDR contract price cycle is expected to see a 23% increase from trough to peak. The next down-cycle (2027-28) may see a 43% decline, followed by a projected 28% increase in the subsequent up-cycle (2028-29).\n\nHBM Demand Forecasts Raised\nThe 2025 HBM bit demand forecast has been raised by 4% to 16.9 billion Gb (+96% YoY), and the 2026 forecast is raised by 3% to 26.1 billion Gb (+55% YoY). This is driven by upward revisions to shipment expectations for Nvidia's AI GPUs, especially Blackwell Ultra and H20. Nvidia's AI GPU procurement volume is estimated to be 6.5 million units in 2025 and 7.0 million units in 2026.\n\nSupply-Side Dynamics\nSK hynix's 2026 HBM bit shipment forecast is revised down to 16.1 billion Gb (+27% YoY), primarily due to constraints on its capacity expansion. Micron's shipments for 2025 and 2026 are projected to be 5.3 billion and 8.9 billion Gb (+67% YoY), respectively, with its new Boise factory not expected to produce its first wafers until the second half of 2027. Samsung's shipments for 2025 and 2026 are projected at 4.6 billion and 7.6 billion Gb (-9% and +63% YoY, respectively); it has not yet received Nvidia's certification for HBM3E 12-Hi, and its HBM4 wafer input is not expected to begin before January-February 2026. UBS believes the HBM supply-demand situation will be largely balanced in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005425040167723383, "ret_signed_m1d": 0.005425040167723383, "alpha_spy_m1d": 0.005517996980789697, "alpha_c_m1d": 0.007544759601472606, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": -0.002119719433749223, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.002341194642022759, "alpha_c_p1d": 0.020767085763974014, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.020767085763974014, "ret_p1w": -0.1139239141549615, "ret_signed_p1w": -0.1139239141549615, "alpha_spy_p1w": -0.09934916389766013, "alpha_c_p1w": -0.06603038154001695, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.047893532614944556, "ret_p1m": 0.1897814129073987, "ret_signed_p1m": 0.1897814129073987, "alpha_spy_p1m": 0.17046204765545703, "alpha_c_p1m": 0.18183221154865303, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.007949201358745661, "ret_p3m": 1.241931755103375, "ret_signed_p3m": 1.241931755103375, "alpha_spy_p3m": 1.17999894436421, "alpha_c_p3m": 1.0792065336461283, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.16272522145724677, "ret_p6m": 2.0409176796884956, "ret_signed_p6m": 2.0409176796884956, "alpha_spy_p6m": 1.9639643268297777, "alpha_c_p6m": 1.7064589868413775, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.33445869284711804, "price_path": [-0.2801, -0.2708, -0.2762, -0.2892, -0.278, -0.2671, -0.269, -0.2491, -0.2333, -0.2604, -0.2495, -0.2495, -0.2134, -0.1881, -0.1881, -0.1718, -0.1664, -0.132, -0.1483, -0.1483, -0.1031, -0.0995, -0.1085, -0.1103, -0.0705, -0.0615, 0.0072, 0.0344, 0.0597, 0.0271, 0.0561, 0.0325, 0.009, 0.0072, -0.0217, -0.0199, 0.0199, 0.0163, 0.0741, 0.0651, 0.085, 0.0796, 0.0705, -0.0253, -0.0271, -0.0145, -0.0289, -0.0271, -0.0253, -0.038, -0.0524, -0.0506, -0.047, -0.0108, -0.0669, -0.0669, -0.047, -0.0651, -0.0524, -0.0723, -0.0344, -0.0271, 0.0054, 0.0, 0.0, -0.0325, -0.0488, -0.0759, -0.1139, -0.0922, -0.0615, -0.0542, -0.0597, -0.0275, -0.0257, -0.0728, -0.0565, -0.0493, -0.0384, -0.0094, 0.0033, 0.0431, 0.101, 0.1119, 0.1898, 0.1988, 0.2604, 0.2079, 0.2876, 0.2876, 0.2713, 0.3075, 0.2948, 0.2912, 0.2188, 0.264, 0.2586, 0.3039, 0.4324, 0.4324, 0.4324, 0.4324, 0.4324, 0.4324, 0.5502, 0.5031, 0.4904, 0.5302, 0.6389, 0.686, 0.7584, 0.7349, 0.7439, 0.7331, 0.8471, 0.9377, 0.887, 1.021, 1.0572, 1.0246, 1.2456, 1.1224, 1.0971, 1.1478, 1.1007, 1.1948, 1.2419, 1.2347, 1.2166, 1.0282, 1.1948, 1.0645, 1.0355, 1.0681, 0.887, 0.8834, 0.8797, 0.8979, 0.9717, 0.921, 0.95, 1.0224, 1.0007, 0.9645, 0.9717, 1.0913, 1.0514, 1.1276, 1.0478, 1.0696, 1.0079, 0.921, 0.9971, 1.0007, 0.9826, 1.1022, 1.1167, 1.1312, 1.1312, 1.171, 1.3197, 1.3595, 1.3595, 1.3595, 1.4538, 1.5226, 1.6314, 1.6893, 1.7401, 1.6966, 1.7147, 1.6748, 1.6893, 1.7147, 1.7401, 1.7691, 1.693, 1.6821, 1.7365, 1.78, 1.6676, 1.8996, 2.0482, 2.1207, 2.2946, 2.0083, 2.2874, 2.262, 2.0518, 2.0409]}
{"unit_id": "orig:1955808837258367336", "ticker_idx": 3, "ts": "2025-08-14T01:48:58+00:00", "symbol": "Samsung", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's shipments for 2025 and 2026 are projected at 4.6 billion and 7.6 billion Gb (-9% and +63% YoY, respectively); it has not yet received Nvidia's certification for HBM3E 12-Hi", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DDR/NAND price forecasts to +5% QoQ in 4Q25 and sees a modest up-cycle into 2026; bullish memory sector, mixed on individual names with Samsung lagging on HBM certification.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung in semiconductor context = Samsung Electronics, 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Semiconductor Memory Commentary: Returning to a Modest Up-cycle Before 2026\nUBS (N. Gaudois, 12/08/25)\n\nUBS expects a stronger-than-anticipated performance in the fourth quarter of 2025 (4Q25) and is significantly raising its DDR and NAND contract price assumptions, reaffirming its view of a modest up-cycle leading into 2026. Growth in HBM demand, improving end-market demand, and capacity constraints at some competitors will support price trends.\n\n4Q25 Pricing Expectations Raised\nIn 3Q25, average contract prices for both DDR and NAND rose by approximately 3% quarter-over-quarter (QoQ). Current inventory levels are around 9-10 weeks for PC/mobile DDR, 13-14 weeks for server DDR, and 8-13 weeks for NAND. Demand from smartphones and PCs is stable, with initial signs for Q4 being positive. A decreased risk from U.S. Section 232 tariffs also provides support. A recovery in demand from hyperscale customers for DDR and enterprise SSDs has prompted UBS to raise its 4Q25 QoQ contract price forecast for DDR and NAND from the previous -3% and -5%, respectively, to +5%.\n\nModest Price Up-cycle Before 2026\nUBS forecasts that DDR and NAND contract prices will continue to see small sequential increases each quarter until 3Q26. The reasons for this include: HBM demand diverting production capacity, improving end-market demand (e.g., servers), and limited yields for CXMT's 16nm DDR5 making a repeat of last year's DDR4 oversupply unlikely. In the long term, the rising proportion of HBM should help mitigate the volatility of traditional memory cycles. This current DDR contract price cycle is expected to see a 23% increase from trough to peak. The next down-cycle (2027-28) may see a 43% decline, followed by a projected 28% increase in the subsequent up-cycle (2028-29).\n\nHBM Demand Forecasts Raised\nThe 2025 HBM bit demand forecast has been raised by 4% to 16.9 billion Gb (+96% YoY), and the 2026 forecast is raised by 3% to 26.1 billion Gb (+55% YoY). This is driven by upward revisions to shipment expectations for Nvidia's AI GPUs, especially Blackwell Ultra and H20. Nvidia's AI GPU procurement volume is estimated to be 6.5 million units in 2025 and 7.0 million units in 2026.\n\nSupply-Side Dynamics\nSK hynix's 2026 HBM bit shipment forecast is revised down to 16.1 billion Gb (+27% YoY), primarily due to constraints on its capacity expansion. Micron's shipments for 2025 and 2026 are projected to be 5.3 billion and 8.9 billion Gb (+67% YoY), respectively, with its new Boise factory not expected to produce its first wafers until the second half of 2027. Samsung's shipments for 2025 and 2026 are projected at 4.6 billion and 7.6 billion Gb (-9% and +63% YoY, respectively); it has not yet received Nvidia's certification for HBM3E 12-Hi, and its HBM4 wafer input is not expected to begin before January-February 2026. UBS believes the HBM supply-demand situation will be largely balanced in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004189843329767129, "ret_signed_m1d": -0.004189843329767129, "alpha_spy_m1d": -0.004282800142833443, "alpha_c_m1d": -0.0020236838971368076, "bench_spy_m1d": -9.295681306631387e-05, "bench_c_m1d": 0.0021661594326303213, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": -0.002341194642022759, "alpha_c_p1d": -0.007582416375052858, "bench_spy_p1d": -0.002341194642022759, "bench_c_p1d": -0.007582416375052858, "ret_p1w": -0.013966585875289317, "ret_signed_p1w": 0.013966585875289317, "alpha_spy_p1w": -0.0006081643820120552, "alpha_c_p1w": -0.01897707452328312, "bench_spy_p1w": -0.014574750257301372, "bench_c_p1w": -0.03294366039857244, "ret_p1m": 0.05307256281123096, "ret_signed_p1m": -0.05307256281123096, "alpha_spy_p1m": -0.033753197559289294, "alpha_c_p1m": -0.0413069534620103, "bench_spy_p1m": 0.019319365251941667, "bench_c_p1m": 0.011765609349220663, "ret_p3m": 0.45198008033816284, "ret_signed_p3m": -0.45198008033816284, "alpha_spy_p3m": -0.3900472695989978, "alpha_c_p3m": -0.35615015755859347, "bench_spy_p3m": 0.06193281073916501, "bench_c_p3m": 0.09582992277956937, "ret_p6m": 1.2357825091119299, "ret_signed_p6m": -1.2357825091119299, "alpha_spy_p6m": -1.158829156253212, "alpha_c_p6m": -1.1785554770170168, "bench_spy_p6m": 0.07695335285871785, "bench_c_p6m": 0.05722703209491309, "price_path": [-0.2254, -0.224, -0.2268, -0.2407, -0.2476, -0.2407, -0.2518, -0.224, -0.2213, -0.2199, -0.2115, -0.2115, -0.1977, -0.1796, -0.1796, -0.1699, -0.1782, -0.1685, -0.1741, -0.1907, -0.206, -0.1935, -0.1699, -0.1782, -0.1741, -0.1949, -0.1602, -0.1491, -0.1643, -0.1508, -0.1648, -0.1592, -0.1508, -0.1089, -0.1159, -0.1383, -0.1425, -0.1564, -0.148, -0.1257, -0.1271, -0.1103, -0.0964, -0.0684, -0.0628, -0.0531, -0.0782, -0.0726, -0.0782, -0.0796, -0.0168, -0.014, 0.014, -0.0028, -0.0377, -0.0265, -0.0237, -0.0391, -0.0154, 0.0028, -0.0084, -0.007, 0.0042, 0.0, 0.0, -0.0223, -0.0223, -0.0154, -0.014, -0.0028, -0.0014, -0.0182, -0.014, -0.0279, -0.0265, -0.0559, -0.0349, -0.0251, -0.0209, -0.0293, -0.0209, -0.0014, 0.014, 0.0251, 0.0531, 0.0684, 0.1089, 0.0922, 0.1215, 0.1215, 0.1662, 0.183, 0.1927, 0.2025, 0.1634, 0.1812, 0.177, 0.2065, 0.2591, 0.2591, 0.2591, 0.2591, 0.2591, 0.2591, 0.3243, 0.3089, 0.285, 0.3327, 0.3706, 0.3734, 0.3762, 0.3678, 0.3832, 0.3538, 0.386, 0.4309, 0.3959, 0.4099, 0.4604, 0.5081, 0.5586, 0.4716, 0.4113, 0.3917, 0.3734, 0.4113, 0.452, 0.4464, 0.4422, 0.3636, 0.4113, 0.372, 0.3538, 0.4113, 0.3299, 0.3566, 0.3931, 0.4422, 0.452, 0.4099, 0.4141, 0.4506, 0.466, 0.4744, 0.5207, 0.5362, 0.5207, 0.5151, 0.5053, 0.5277, 0.4702, 0.4422, 0.5137, 0.5095, 0.4913, 0.5502, 0.5642, 0.5586, 0.5586, 0.6414, 0.6846, 0.6902, 0.6902, 0.6902, 0.8115, 0.9468, 0.9581, 0.9877, 0.9567, 0.9595, 0.9567, 0.9397, 0.9778, 1.0286, 1.099, 1.1047, 1.0469, 1.1075, 1.147, 1.1442, 1.1442, 1.2485, 1.2894, 1.2654, 1.2626, 1.1202, 1.3612, 1.3838, 1.2457, 1.2358]}
{"unit_id": "orig:1948199700701859947", "ticker_idx": 0, "ts": "2025-07-24T01:52:59+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "2026 Could Be a Key Year for Valuation Revisions … the preliminary forecast for 2026 is a 9% year-over-year decline, but the risk is skewed to the upside. If the order structure proves favorable, there's potential for upward revisions to the outlook", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley bullish note on KLA: DRAM-driven process control intensity rising, 2026 forecast conservative with upside risk skewed higher.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "KLA: Focusing on DRAM Upside Potential, 2026 Could Be a Key Year for Valuation Revisions\nMorgan Stanley (S. Brett, 2025/07/20)\n\nMorgan Stanley notes that while KLA's September quarter guidance was slightly better than market expectations, it wasn't a substantial catalyst. Market focus has shifted to the strength of DRAM-driven demand in the second half of the year and its potential impact on upward revisions to 2026 forecasts.\n\nSeptember Quarter Guidance Slightly Raised, Meets Expectations but Unlikely to Be a Valuation Catalyst\nKLA expects September quarter revenue to see a low single-digit sequential increase (approx. $3.1 billion), which is better than the market's expectation of a 1% sequential decline but broadly in line with buyer expectations. As the market generally anticipated China and DRAM to drive second-half growth, the current guidance is unlikely to lead to further valuation re-rating.\n\nDRAM is the Market Focus, Determining the Upside for 2026 Earnings Revisions\nMorgan Stanley believes that the market will not reward valuation increases driven by China, but it will react positively to DRAM-led growth. SK Hynix might be a major client in the second half. If its spending is for improving yields on existing production lines (rather than new capacity builds), it would resemble TSMC's strategy of reinvesting in older process nodes.\n\nDeclining Yields Drive Demand for Process Control, KLA Benefits Significantly\nThe application of EUV and HBM in DRAM is lowering yields, prompting manufacturers to increase investment in process control. KLA notes that the intensity of process control in DRAM has increased from 8-9% to 9-10%. This is expected to further increase by 100-200 bps with higher HBM penetration, which is favorable for its product demand.\n\nDecreased DRAM Output Efficiency Further Boosts Demand for KLA\nDRAM bit output per wafer is estimated to decline year-over-year in 2024, marking the first such decrease since 2012. This is mainly due to yield issues at EUV nodes and low HBM die utilization (estimated die discount of 3x in 2024, reducing bit output by 8%). Manufacturers need to either improve yields or expand capacity, both of which present opportunities for KLA.\n\n2025 Market Share Expected to Rise, 2026 Forecast Remains Conservative\nInstitutions expect KLA's DRAM revenue to grow by 28% year-over-year in 2025, significantly higher than the industry's +5% increase, with market share rising from 5.8% to 7.4%. The preliminary forecast for 2026 is a 9% year-over-year decline, but the risk is skewed to the upside. If the order structure proves favorable, there's potential for upward revisions to the outlook.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007841381247291834, "ret_signed_m1d": -0.007841381247291834, "alpha_spy_m1d": -0.007510438283967802, "alpha_c_m1d": -0.003706561422493815, "bench_spy_m1d": -0.0003309429633240324, "bench_c_m1d": -0.0041348198247980195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002311467905428133, "ret_signed_p1d": -0.002311467905428133, "alpha_spy_p1d": -0.00653581754574728, "alpha_c_p1d": -0.0012342852756921374, "bench_spy_p1d": 0.004224349640319147, "bench_c_p1d": -0.0010771826297359954, "ret_p1w": -0.02781520050861397, "ret_signed_p1w": -0.02781520050861397, "alpha_spy_p1w": -0.024126898910618366, "alpha_c_p1w": -0.031220339989777135, "bench_spy_p1w": -0.0036883015979956024, "bench_c_p1w": 0.0034051394811631663, "ret_p1m": -0.03539780442140339, "ret_signed_p1m": -0.03539780442140339, "alpha_spy_p1m": -0.05256317849270631, "alpha_c_p1m": -0.05568972548785989, "bench_spy_p1m": 0.01716537407130292, "bench_c_p1m": 0.020291921066456498, "ret_p3m": 0.27179011819785526, "ret_signed_p3m": 0.27179011819785526, "alpha_spy_p3m": 0.21074060113590698, "alpha_c_p3m": 0.07199864084299534, "bench_spy_p3m": 0.06104951706194828, "bench_c_p3m": 0.19979147735485991, "ret_p6m": 0.7406580974614276, "ret_signed_p6m": 0.7406580974614276, "alpha_spy_p6m": 0.6441809954933504, "alpha_c_p6m": 0.3451199757586174, "bench_spy_p6m": 0.09647710196807724, "bench_c_p6m": 0.3955381217028102, "price_path": [-0.2346, -0.241, -0.2247, -0.2538, -0.2297, -0.2351, -0.2504, -0.2339, -0.224, -0.2264, -0.161, -0.1219, -0.1118, -0.1107, -0.1271, -0.1296, -0.126, -0.1385, -0.1512, -0.1626, -0.1626, -0.1273, -0.1401, -0.144, -0.1629, -0.1568, -0.142, -0.135, -0.1242, -0.1064, -0.0828, -0.0526, -0.0356, -0.0323, -0.0404, -0.0131, -0.0124, -0.0365, -0.0365, -0.0599, -0.053, -0.0168, -0.0119, -0.0014, -0.0158, -0.0093, -0.0059, 0.0187, 0.0226, 0.0226, 0.0093, 0.0166, 0.021, 0.027, 0.0226, 0.0195, 0.0358, 0.0324, 0.0364, 0.0298, 0.0371, -0.0132, -0.0078, 0.0, -0.0023, 0.021, 0.0132, 0.023, -0.0278, -0.0194, 0.0127, -0.023, -0.0176, 0.0087, 0.0117, 0.0066, 0.0347, 0.0501, 0.0567, -0.0323, -0.0209, -0.029, -0.0264, -0.0331, -0.0354, -0.0251, -0.0156, -0.0148, -0.0091, -0.0335, -0.0335, -0.0619, -0.0646, -0.0321, 0.0032, 0.0075, 0.0172, 0.0337, 0.0632, 0.0685, 0.0961, 0.0979, 0.0972, 0.1601, 0.158, 0.1874, 0.1873, 0.1845, 0.1739, 0.1796, 0.1795, 0.1955, 0.2512, 0.2627, 0.2209, 0.2632, 0.2023, 0.1778, 0.1676, 0.0893, 0.1361, 0.1369, 0.2048, 0.2179, 0.2266, 0.2778, 0.2718, 0.2351, 0.2846, 0.311, 0.3468, 0.3368, 0.3692, 0.346, 0.3397, 0.3513, 0.3228, 0.3601, 0.3372, 0.3227, 0.35, 0.32, 0.3289, 0.2876, 0.2573, 0.2587, 0.2469, 0.2962, 0.224, 0.2181, 0.262, 0.2722, 0.2868, 0.2868, 0.3051, 0.2847, 0.321, 0.3453, 0.3413, 0.3483, 0.3596, 0.3607, 0.3755, 0.3836, 0.3255, 0.3602, 0.3582, 0.3012, 0.3571, 0.383, 0.4052, 0.4086, 0.4178, 0.4178, 0.4207, 0.3993, 0.3808, 0.349, 0.349, 0.415, 0.5015, 0.5488, 0.5096, 0.4706, 0.5543, 0.5856, 0.6008, 0.5926, 0.7153, 0.7407]}
{"unit_id": "orig:1968165867134300465", "ticker_idx": 0, "ts": "2025-09-17T04:11:24+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "~7% and 5% EPS upside for LAM and AMAT in 2026, respectively", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises 2026 semi-equipment bull case to $50B on stronger memory prices, implying ~7% EPS upside for LRCX and ~5% for AMAT.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Memory Semiconductor Equipment: Stronger Memory Prices Push 2026 Bull Case Expectations Higher\n\nMorgan Stanley says that strengthening memory prices are pushing up the 2026 semiconductor equipment bull case scenario to USD 50 billion (vs. previous estimate of USD 43.8 billion), which implies around 7% and 5% EPS upside for LAM and AMAT in 2026, respectively.\n\nThe firm noted that 2026 DRAM bit growth expectations have been raised to 22%, corresponding to potential DRAM equipment investment of USD 35 billion; NAND bit growth was revised up to 26%, corresponding to NAND equipment investment reaching USD 15 billion.\n\nAlthough capital expenditure in the second half of 2025 is unlikely to see a clear near-term upside, the tight NAND supply–demand balance and the trend of eSSD replacing HDDs are reinforcing the possibility of ramped-up investment in 2026.\n\n(Morgan Stanley, S. Brett, 16-Sep-2025)\n\n$LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01173095054577944, "ret_signed_m1d": -0.01173095054577944, "alpha_spy_m1d": -0.012974931404049705, "alpha_c_m1d": -0.018151911283007527, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03625926738363949, "ret_signed_p1d": 0.03625926738363949, "alpha_spy_p1d": 0.031586822241309154, "alpha_c_p1d": -0.002135522518084576, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.054744855143045834, "ret_signed_p1w": 0.054744855143045834, "alpha_spy_p1w": 0.049051695786501925, "alpha_c_p1w": 0.0027874830375100856, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.1702308192326889, "ret_signed_p1m": 0.1702308192326889, "alpha_spy_p1m": 0.16523734074771235, "alpha_c_p1m": 0.04557897729065563, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.35271098642796184, "ret_signed_p3m": 0.35271098642796184, "alpha_spy_p3m": 0.31715584339537517, "alpha_c_p3m": 0.1966094139815513, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.7268346853718224, "ret_signed_p6m": 0.7268346853718224, "alpha_spy_p6m": 0.7106021451782238, "alpha_c_p6m": 0.45136307429996236, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.2577, -0.2485, -0.2155, -0.2123, -0.2056, -0.2026, -0.2015, -0.2058, -0.1893, -0.1894, -0.1894, -0.1949, -0.181, -0.1812, -0.171, -0.1655, -0.1828, -0.1709, -0.1766, -0.1732, -0.1742, -0.1654, -0.1986, -0.2034, -0.1979, -0.2046, -0.191, -0.1884, -0.1871, -0.222, -0.2094, -0.1927, -0.2069, -0.213, -0.1866, -0.1653, -0.1632, -0.1363, -0.1244, -0.1191, -0.1837, -0.1888, -0.1769, -0.1866, -0.1927, -0.179, -0.1692, -0.1499, -0.1495, -0.1461, -0.1784, -0.1784, -0.204, -0.1982, -0.1762, -0.1555, -0.1381, -0.134, -0.1193, -0.0518, -0.0405, -0.0221, -0.0117, 0.0, 0.0363, 0.0412, 0.0845, 0.0823, 0.0547, 0.0532, 0.0548, 0.0775, 0.1006, 0.1737, 0.2082, 0.1985, 0.226, 0.1536, 0.1716, 0.159, 0.0798, 0.1327, 0.1369, 0.19, 0.1702, 0.1632, 0.184, 0.1922, 0.161, 0.2127, 0.2468, 0.2897, 0.2791, 0.3207, 0.3234, 0.2943, 0.3253, 0.2805, 0.3567, 0.3331, 0.3098, 0.3675, 0.3084, 0.3268, 0.2602, 0.2186, 0.2121, 0.1774, 0.2231, 0.1474, 0.1725, 0.2361, 0.2488, 0.2752, 0.2752, 0.2823, 0.2723, 0.3003, 0.3152, 0.2933, 0.3066, 0.3399, 0.3651, 0.3853, 0.389, 0.3216, 0.3527, 0.3441, 0.276, 0.356, 0.4183, 0.4429, 0.4421, 0.46, 0.46, 0.4661, 0.448, 0.4308, 0.4094, 0.4094, 0.5236, 0.6035, 0.7039, 0.672, 0.6545, 0.7978, 0.8146, 0.765, 0.719, 0.7905, 0.8357, 0.8357, 0.8311, 0.8804, 0.8171, 0.7943, 0.8349, 0.9633, 0.9725, 1.0432, 0.9221, 0.9554, 0.8945, 0.7272, 0.7562, 0.9019, 0.8877, 0.8657, 0.9358, 0.9043, 0.9392, 0.9392, 0.9396, 0.9767, 0.9545, 1.0165, 0.9947, 1.011, 1.054, 0.9683, 0.9257, 0.9019, 0.7888, 0.8381, 0.7696, 0.6431, 0.7405, 0.7741, 0.8042, 0.7268]}
{"unit_id": "orig:1968165867134300465", "ticker_idx": 1, "ts": "2025-09-17T04:11:24+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "5% EPS upside for LAM and AMAT in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises 2026 semi-equipment bull case to $50B on stronger memory prices, implying ~7% EPS upside for LRCX and ~5% for AMAT.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Memory Semiconductor Equipment: Stronger Memory Prices Push 2026 Bull Case Expectations Higher\n\nMorgan Stanley says that strengthening memory prices are pushing up the 2026 semiconductor equipment bull case scenario to USD 50 billion (vs. previous estimate of USD 43.8 billion), which implies around 7% and 5% EPS upside for LAM and AMAT in 2026, respectively.\n\nThe firm noted that 2026 DRAM bit growth expectations have been raised to 22%, corresponding to potential DRAM equipment investment of USD 35 billion; NAND bit growth was revised up to 26%, corresponding to NAND equipment investment reaching USD 15 billion.\n\nAlthough capital expenditure in the second half of 2025 is unlikely to see a clear near-term upside, the tight NAND supply–demand balance and the trend of eSSD replacing HDDs are reinforcing the possibility of ramped-up investment in 2026.\n\n(Morgan Stanley, S. Brett, 16-Sep-2025)\n\n$LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025767826187439447, "ret_signed_m1d": -0.025767826187439447, "alpha_spy_m1d": -0.027011807045709713, "alpha_c_m1d": -0.032188786924667534, "bench_spy_m1d": 0.001243980858270266, "bench_c_m1d": 0.0064209607372280875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06528924261639157, "ret_signed_p1d": 0.06528924261639157, "alpha_spy_p1d": 0.060616797474061235, "alpha_c_p1d": 0.026894452714667505, "bench_spy_p1d": 0.004672445142330339, "bench_c_p1d": 0.03839478990172407, "ret_p1w": 0.1308594148283848, "ret_signed_p1w": 0.1308594148283848, "alpha_spy_p1w": 0.1251662554718409, "alpha_c_p1w": 0.07890204272284906, "bench_spy_p1w": 0.005693159356543909, "bench_c_p1w": 0.05195737210553575, "ret_p1m": 0.2783921087091703, "ret_signed_p1m": 0.2783921087091703, "alpha_spy_p1m": 0.27339863022419375, "alpha_c_p1m": 0.15374026676713703, "bench_spy_p1m": 0.0049934784849765546, "bench_c_p1m": 0.12465184194203327, "ret_p3m": 0.4696126673520833, "ret_signed_p3m": 0.4696126673520833, "alpha_spy_p3m": 0.4340575243194966, "alpha_c_p3m": 0.31351109490567275, "bench_spy_p3m": 0.03555514303258667, "bench_c_p3m": 0.15610157244641054, "ret_p6m": 0.8994697098956654, "ret_signed_p6m": 0.8994697098956654, "alpha_spy_p6m": 0.8832371697020669, "alpha_c_p6m": 0.6239980988238054, "bench_spy_p6m": 0.016232540193598544, "bench_c_p6m": 0.27547161107186, "price_path": [-0.0514, -0.0374, 0.0086, 0.0248, 0.0273, 0.0256, 0.0248, 0.0287, 0.0636, 0.0695, 0.0695, 0.068, 0.0915, 0.0938, 0.1085, 0.108, 0.1033, 0.1156, 0.0905, 0.0777, 0.0661, 0.0782, 0.0476, 0.0469, 0.0531, 0.0395, 0.0651, 0.0547, 0.0602, 0.0079, 0.0076, 0.0234, 0.0029, -0.0028, 0.0252, 0.0349, 0.0321, 0.0549, 0.0638, 0.0537, -0.0945, -0.0846, -0.0919, -0.099, -0.1027, -0.0878, -0.0906, -0.0765, -0.0771, -0.0722, -0.0975, -0.0975, -0.1154, -0.1228, -0.1117, -0.0863, -0.0903, -0.0821, -0.0826, -0.0448, -0.058, -0.0404, -0.0258, 0.0, 0.0653, 0.0672, 0.1257, 0.1277, 0.1309, 0.1205, 0.1448, 0.1506, 0.1494, 0.2224, 0.2552, 0.2212, 0.257, 0.1877, 0.2211, 0.2367, 0.1786, 0.2321, 0.2249, 0.2776, 0.2784, 0.2631, 0.2807, 0.2687, 0.2382, 0.2826, 0.2842, 0.2987, 0.2779, 0.3235, 0.3055, 0.3086, 0.3345, 0.2923, 0.3523, 0.311, 0.2916, 0.3197, 0.2837, 0.2953, 0.2532, 0.2688, 0.2839, 0.2638, 0.32, 0.2388, 0.26, 0.2988, 0.3638, 0.4061, 0.4061, 0.4189, 0.4329, 0.4924, 0.511, 0.5156, 0.5075, 0.5084, 0.5026, 0.5477, 0.5193, 0.458, 0.4696, 0.4559, 0.3965, 0.4259, 0.4423, 0.4569, 0.4638, 0.4669, 0.4669, 0.4732, 0.4796, 0.4623, 0.4455, 0.4455, 0.5124, 0.5993, 0.665, 0.6436, 0.5842, 0.6941, 0.7282, 0.7149, 0.6981, 0.7948, 0.8394, 0.8394, 0.79, 0.8294, 0.7932, 0.8133, 0.7969, 0.8715, 0.8942, 0.92, 0.813, 0.8472, 0.7925, 0.674, 0.7099, 0.8141, 0.8594, 0.851, 0.9118, 0.8472, 0.9963, 0.9963, 1.0201, 1.0773, 1.0828, 1.1141, 1.1038, 1.1285, 1.2243, 1.116, 1.0968, 1.0961, 0.9786, 1.0149, 0.9516, 0.8289, 0.9089, 0.948, 0.9772, 0.8995]}
{"unit_id": "orig:1953313791137202271", "ticker_idx": 0, "ts": "2025-08-07T04:34:33+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We are lowering our price target to $168 and maintaining our Equal-weight rating due to decreased confidence in China demand and a lack of overall AI upside drivers.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley cuts AMD PT to $168, maintains Equal-weight; lowered confidence in China demand and no clear AI upside catalysts make current valuation hard to sustain.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AMD: Weaker 3Q EPS Guidance\nMorgan Stanley (J. Moore, 08/25/25)\n\nEach business segment performed strongly this quarter, but the timing of a recovery in the Chinese market is more ambiguous than expected. The GPU business lacks clear short-term upside, which may make it difficult to sustain the current high valuation premium. However, the MI400 remains the core source of market optimism for 2025 and is the key value driver for the stock.\n\nKey Points\n\n- AMD's third-quarter revenue guidance of $8.77 billion exceeded market expectations, but the EPS upside was weakened by rising operating expenses.\n- Shipments of the MI308 to China were not included in the guidance, mainly due to the uncertain timing of license approvals. This was in line with expectations, but management's tone was more conservative than we anticipated.\n- The MI400 series is still viewed as a significant milestone in the AI field, and the company remains optimistic about the product's potential.\n- We are lowering our price target to $168 and maintaining our Equal-weight rating due to decreased confidence in China demand and a lack of overall AI upside drivers.\n\nRevenue Growth Remains Strong, but It's Hard to Sustain Optimistic Expectations\n\nAlthough AMD's third-quarter guidance was well above market expectations, several factors make the revenue growth appear less attractive:\n\n1. The outperformance in the second quarter was primarily driven by the console gaming business, which is one of AMD's lowest-quality segments.\n\n2. Data Center GPUs are expected to grow year-over-year in the third quarter, which we estimate will exceed $1.6 billion. Combined with management's mention of strong server growth and growth in the Embedded and Client businesses, we believe the overall figure will not be $100-200 million higher than this level. Therefore, we are only raising our full-year Instinct revenue forecast by $250 million. Uncertainty remains for the fourth quarter, and the company has not provided clear guidance for this business. Considering the stock's forward P/E multiple has increased 15x over the past three months due to rising expectations for the AI business, we would have preferred to see an upward revision to the 2025 AI-related revenue forecast rather than for the gaming business, especially given the tight supply-demand situation for AI computing power.\n\n3. Operating expenses exceeded expectations, limiting the extent of EPS improvement.\n\nSales to China Not Yet Included in Guidance, Management Tone More Cautious\n\nAMD did not include any potential impact from selling the MI308 to China in its guidance because license approval remains uncertain. This is similar to NVIDIA's statements. AMD did not suggest that it could quickly convert inventory into shipments once licenses are approved, and the phrase \"could take several quarters\" mentioned in the conference call implies a much longer timeframe than we expected. It's worth remembering that in April this year, the company had estimated the China market gap to be $700 million. We believe this conveys a degree of conservatism and also reflects that the company is less optimistic about the overall opportunity in the Chinese market than we previously anticipated, even though this could have been a multi-billion dollar annual opportunity for AMD.\n\nMI400 Remains a Key Driver of Future Value\n\nThe MI400, as the next-generation product, remains the core factor driving the stock price. Management's comments on the product were generally positive and in line with expectations. On the conference call, the company reiterated that its performance is expected to meet or exceed the level of Rubin and create \"billions of dollars\" in AI revenue in the future.\n\nHowever, aside from this \"billions of dollars\" statement, the company did not provide guidance for its AI business beyond the current quarter. It did not offer a more specific breakdown of this year's \"strong double-digit growth,\" nor did it specify when these \"billions of dollars\" would be realized. We do not see this as a sign of caution, but rather as the company's reluctance to define success or failure based on the performance of a single year.\n\nOther Businesses Besides AI\n\nThe Embedded business is continuously recovering and is expected to return to growth in the second half of the year, which will help gross margin performance. The company is also ramping up the MI355.\n\nThe Gaming business has performed remarkably well over the past two quarters and is set to grow approximately 2.5x year-over-year, mainly due to an acceleration in console shipments after a prolonged inventory destocking cycle. However, without the launch of a new generation of consoles, we believe it is unrealistic for annual revenue to return to the $6 billion level of 2022/2023.\n\nThe PC business also remains strong. Although we expect a digestion period after several quarters of strong performance, we have not yet seen this trend. We remain conservative on this market. Intel has clearly stated that it sees signs of pull-in demand in the market, and if subsequent unit shipment growth is lower than expected, we would be concerned about Intel's more aggressive competitive approach. But what is certain is that AMD is continuously gaining market share, and this trend is unlikely to change in the short term.\n\nDetailed Segment Data\n\nThe company's revenue for the quarter was $7.685 billion, up 3.3% sequentially and 31.7% year-over-year, higher than the market consensus of $7.415 billion and our estimate of $7.393 billion. \n\nBy business segment:\n\n- Data Center revenue was $3.24 billion, down 11.8% sequentially but up 14.3% year-over-year.\n- Client revenue was $2.499 billion, up 8.9% sequentially and 67.5% year-over-year.\n- Gaming revenue was $1.122 billion, up 73.4% sequentially and 73.1% year-over-year.\n- Embedded revenue was $824 million, up 0.1% sequentially but down 4.3% year-over-year.\n- Gross margin was 43.2%, in line with the market consensus (43.1%) but slightly below our expectation of 43.4%. Earnings per share (EPS) were $0.48, consistent with both market and our expectations ($0.48 and $0.49, respectively).\n\nThird Quarter Outlook\n\n- The company guides for third-quarter revenue of $8.7 billion (midpoint), up 13.2% sequentially, which is higher than the market consensus of $8.323 billion and our estimate of $8.280 billion.\n- The company expects a third-quarter gross margin of 54.0%, a significant increase of 1080 basis points sequentially and 44 basis points year-over-year, in line with the market's expectation of 53.9% but below our expectation of 55.9%.\n\nForecast Adjustments\n\n- We now forecast Q3 revenue of $8.73 billion, a non-GAAP gross margin of 54.1%, a non-GAAP operating margin of 24.8%, and EPS of $1.16.\n\n- For fiscal year 2025, we are raising our forecast to $33.419 billion in revenue, a 52.0% non-GAAP gross margin, and EPS of $4.28 (previously $32.413 billion / 52.5% / $4.20).\n\n- For fiscal year 2026, we expect revenue growth of 20% to a total of $40.044 billion, a non-GAAP gross margin of 55.9%, and EPS of $6.13 (previously $38.920 billion / 55.7% / $6.12).\n\nView on the Stock\n\nOur previous investment thesis was that AMD could maintain a high valuation premium based on its AI business, especially with the potential upside from the China market and traditional customers. However, the current revenue performance has not led us to significantly raise our 2025 AI revenue expectations. The China market could still bring upside, but management's tone on this market is more cautious than expected. Against the backdrop of the current unprecedented AI investment frenzy, AMD's lack of short-term catalysts makes it difficult for us to get excited, especially considering the stock's current valuation is still higher than companies that have already started to deliver on AI upside.\n\nThe future MI400 could improve AMD's position, but this remains to be seen amidst the rapid advancements of ASICs and NVIDIA. Although AMD remains strong in its other business areas (partly thanks to Intel's disarray), the uncertainty in AI leads us to maintain an Equal-weight rating. We would like to see more explicit enthusiasm from customers for AMD's mid-to-long-term AI product roadmap, which is not yet apparent.\n\nOur previous price target was based on a median 2026 EPS of $5.49, corresponding to a 34x P/E multiple. We are now raising our EPS estimate to $5.59 but lowering the valuation multiple to 30x (or 27x on a non-GAAP basis), primarily due to decreased confidence in the China market as a driver, which was one of the reasons for our previous premium valuation. The current valuation is still higher than its peers but lower than AI leaders NVIDIA and Broadcom. Therefore, we are lowering our price target to $168.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.053828301089551944, "ret_signed_m1d": 0.053828301089551944, "alpha_spy_m1d": 0.054666627369333254, "alpha_c_m1d": 0.038370820203197376, "bench_spy_m1d": 0.0008383262797813096, "bench_c_m1d": -0.015457480886354569, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002088170667614575, "ret_signed_p1d": -0.002088170667614575, "alpha_spy_p1d": 0.005709402791078677, "alpha_c_p1d": 0.006190211625063835, "bench_spy_p1d": 0.007797573458693252, "bench_c_p1d": 0.00827838229267841, "ret_p1w": 0.049593986974799886, "ret_signed_p1w": -0.049593986974799886, "alpha_spy_p1w": -0.029506964987987105, "alpha_c_p1w": -0.012495843569697573, "bench_spy_p1w": 0.02008702198681278, "bench_c_p1w": 0.03709814340510231, "ret_p1m": -0.12331783793218265, "ret_signed_p1m": 0.12331783793218265, "alpha_spy_p1m": 0.14702686559928635, "alpha_c_p1m": 0.1307375297058202, "bench_spy_p1m": 0.0237090276671037, "bench_c_p1m": 0.007419691773637549, "ret_p3m": 0.45040610153326055, "ret_signed_p3m": -0.45040610153326055, "alpha_spy_p3m": -0.3794497796856755, "alpha_c_p3m": -0.23867201041553376, "bench_spy_p3m": 0.07095632184758505, "bench_c_p3m": 0.21173409111772679, "ret_p6m": 0.37314387533958326, "ret_signed_p6m": -0.37314387533958326, "alpha_spy_p6m": -0.27241027412996344, "alpha_c_p6m": 0.017057462665547973, "bench_spy_p6m": 0.10073360120961983, "bench_c_p6m": 0.39020133800513124, "price_path": [-0.3729, -0.3477, -0.3172, -0.333, -0.3204, -0.3345, -0.3416, -0.35, -0.3578, -0.3602, -0.3602, -0.3355, -0.3454, -0.3444, -0.3577, -0.3351, -0.3195, -0.3122, -0.3289, -0.326, -0.2939, -0.2852, -0.2973, -0.3126, -0.3262, -0.2669, -0.2628, -0.2646, -0.2646, -0.2561, -0.2484, -0.197, -0.1682, -0.1666, -0.1658, -0.1769, -0.2105, -0.1965, -0.2001, -0.2001, -0.2181, -0.2006, -0.1972, -0.1638, -0.1507, -0.1517, -0.0974, -0.0715, -0.0695, -0.0894, -0.0893, -0.1026, -0.0798, -0.0596, -0.0344, 0.0073, 0.0292, 0.0412, 0.0227, -0.0041, 0.0254, 0.0111, -0.0538, 0.0, 0.0021, -0.0007, 0.0148, 0.0697, 0.0496, 0.0296, 0.0217, -0.0339, -0.0418, -0.0504, -0.0269, -0.0524, -0.0335, -0.0306, -0.0222, -0.0567, -0.0567, -0.0585, -0.0596, -0.0615, -0.1233, -0.1218, -0.0962, -0.0746, -0.097, -0.0802, -0.0652, -0.0693, -0.0768, -0.084, -0.0871, -0.0731, -0.0667, -0.0668, -0.0646, -0.0751, -0.064, -0.0615, -0.0487, -0.0155, -0.0448, 0.1816, 0.2269, 0.3664, 0.3509, 0.2465, 0.2553, 0.265, 0.384, 0.3606, 0.352, 0.3954, 0.3807, 0.3354, 0.3631, 0.4671, 0.5062, 0.4966, 0.5332, 0.4782, 0.4856, 0.5061, 0.4504, 0.4868, 0.3788, 0.3546, 0.4152, 0.3777, 0.5017, 0.4383, 0.4316, 0.3951, 0.3358, 0.2967, 0.195, 0.182, 0.2474, 0.1956, 0.2427, 0.2427, 0.2618, 0.2747, 0.2485, 0.2622, 0.2528, 0.2643, 0.2825, 0.2855, 0.2843, 0.2844, 0.2226, 0.2041, 0.2133, 0.1491, 0.1662, 0.238, 0.2468, 0.2465, 0.2473, 0.2473, 0.247, 0.2506, 0.2491, 0.2422, 0.2422, 0.2962, 0.2824, 0.2433, 0.2182, 0.1872, 0.1785, 0.2047, 0.2817, 0.297, 0.322, 0.3447, 0.3447, 0.3452, 0.449, 0.4718, 0.5063, 0.4577, 0.4619, 0.466, 0.4628, 0.3731]}
{"unit_id": "orig:1990410742030561599", "ticker_idx": 0, "ts": "2025-11-17T13:24:35+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The probability of a beat-and-raise is very high", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares JPM NVDA outlook: F4Q rev $63-64B above $61.5B consensus, beat-and-raise probability very high.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM’s NVIDIA outlook:\n    \n• For FY26, total rack shipments are expected to be around 28,000–30,000 units.\n    \n• J.P. Morgan expects F4Q revenue of $63–64B, above the F3Q revenue consensus of about $55.8B.\n\n(Street consensus: $61.5B.)\n    \n• The probability of a beat-and-raise is very high.\n\nAnd… what really stands out to me is this:\n\n“We expect the average selling price (ASP) of HBM4 to rise by 30–40% versus HBM3e, but what is more noteworthy is that we see LPDDR price increases as an even bigger source of pressure. NVDA is already going to reflect higher HBM ASPs in how it prices Rubin and lock in HBM4 prices with memory manufacturers in advance, but in the case of LPDDR it is likely to remain exposed to floating market prices. As far as we can tell, in the current pricing environment long-term supply agreements (LTAs) allow you to secure volume but not price.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019131846405796793, "ret_signed_m1d": 0.019131846405796793, "alpha_spy_m1d": 0.009727708599916118, "alpha_c_m1d": 0.005379143182799417, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.013752703222997376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028081462747415187, "ret_signed_p1d": -0.028081462747415187, "alpha_spy_p1d": -0.019684002189208627, "alpha_c_p1d": -0.007482093889263797, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.02059936885815139, "ret_p1w": -0.021704126091018905, "ret_signed_p1w": -0.021704126091018905, "alpha_spy_p1w": -0.026300981222813657, "alpha_c_p1w": -0.018236624138941115, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.0034675019520777894, "ret_p1m": -0.04753536547217685, "ret_signed_p1m": -0.04753536547217685, "alpha_spy_p1m": -0.06736507223466526, "alpha_c_p1m": -0.08174053694917105, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.0342051714769942, "ret_p3m": 0.0018778713042340023, "ret_signed_p3m": 0.0018778713042340023, "alpha_spy_p3m": -0.024581409583154246, "alpha_c_p3m": -0.19522369791446703, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.19710156921870103, "ret_p6m": 0.18330246755822355, "ret_signed_p6m": 0.18330246755822355, "alpha_spy_p6m": 0.06806037629574302, "alpha_c_p6m": -0.4711095881866425, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.654412055744866, "price_path": [-0.0601, -0.0623, -0.0462, -0.0364, -0.0259, -0.0269, -0.0345, -0.0666, -0.0666, -0.0848, -0.0857, -0.0801, -0.105, -0.0981, -0.0849, -0.0497, -0.0505, -0.0471, -0.0474, -0.0628, -0.0874, -0.0555, -0.0532, -0.016, -0.0438, -0.0516, -0.0477, -0.0451, -0.0255, -0.0001, 0.0034, 0.0123, 0.0055, -0.0057, -0.0084, 0.0135, 0.032, -0.0184, 0.0092, -0.0352, -0.0363, -0.0257, -0.0181, -0.0212, -0.0292, -0.0339, -0.0238, -0.0018, 0.0262, 0.0773, 0.1095, 0.0873, 0.0852, 0.1087, 0.0648, 0.0461, 0.0079, 0.0083, 0.0667, 0.0352, 0.0386, 0.0014, 0.0191, 0.0, -0.0281, -0.0004, -0.0319, -0.0414, -0.0217, -0.0471, -0.034, -0.034, -0.0514, -0.0358, -0.0275, -0.0376, -0.0172, -0.0224, -0.0056, -0.0087, -0.0151, -0.0303, -0.062, -0.0552, -0.0475, -0.0839, -0.0667, -0.03, -0.0155, 0.014, 0.0108, 0.0108, 0.0211, 0.0087, 0.0051, -0.0005, -0.0005, 0.0121, 0.0082, 0.0035, 0.0135, -0.0083, -0.0093, -0.0088, -0.0042, -0.0185, 0.0025, -0.0019, -0.0019, -0.0457, -0.0175, -0.0094, 0.0058, -0.0006, 0.0103, 0.0264, 0.0317, 0.0243, -0.0053, -0.0335, -0.0665, -0.0788, -0.0063, 0.0185, 0.0105, 0.0185, 0.0019, -0.0203, -0.0203, -0.0087, 0.0075, 0.007, 0.0173, 0.0266, 0.0336, 0.0481, -0.0091, -0.0504, -0.022, -0.035, -0.019, -0.0174, -0.047, -0.0211, -0.0098, -0.0029, -0.0184, -0.0339, -0.018, -0.0249, -0.0331, -0.043, -0.0744, -0.0586, -0.061, -0.0423, -0.0822, -0.1022, -0.1147, -0.0653, -0.058, -0.0493, -0.0493, -0.0479, -0.0454, -0.0241, -0.0143, 0.011, 0.0146, 0.0532, 0.0659, 0.0631, 0.0809, 0.083, 0.0713, 0.0853, 0.07, 0.1163, 0.161, 0.1425, 0.1215, 0.0696, 0.0636, 0.0638, 0.0532, 0.1139, 0.1336, 0.1534, 0.1761, 0.1833]}
{"unit_id": "thread:1984213326780629083", "ticker_idx": 0, "ts": "2025-10-31T10:58:16+00:00", "symbol": "SK hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Companies experienced in stacking memory — such as SK hynix and Samsung Electronics — are best positioned to develop HBF", "tweet_type": "analysis", "timeline": "years", "summary": "Relays bullish thesis: memory era incoming via HBF (2027-28); SK hynix and Samsung best positioned; Micron/SanDisk potential Nvidia acquisition targets.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Father of HBM: Nvidia May Acquire a Memory Company Like SanDisk”\n\nProfessor Jung-Ho Kim, known as the “father of HBM (High Bandwidth Memory)” and a professor in the Department of Electrical and Electronic Engineering at KAIST, appeared in a YouTube video and made a bold claim — that Nvidia could acquire a memory company such as SanDisk or Micron.\n\n(He previously recommended buying SK hynix stock back in 2017.)\n\n1. Shift of Market Power: From GPU to Memory\n\nKim argues that as the AI era unfolds, the driving force of the stock market is shifting from GPU (Graphics Processing Units) to memory.\n\nIf the GPU is the castella cake, memory is the whipped cream — to improve AI performance, memory upgrades and capacity expansion are essential.\n\n2. Memory Determines AI Speed\n\nThe speed at which generative AI produces words in response to prompts is determined by memory, and this is even more critical during inference than during training.\n\n3. Explosive Memory Pressure\n\nAs AI model sizes (parameters), user counts, and data types (text → image → video) grow exponentially, the demand pressure on memory capacity is reaching its peak.\n\n4. From HBM to HBF Era\n\nThe professor predicts that the HBM (High Bandwidth Memory) era is ending, and the HBF (High Bandwidth Flash) era is about to begin — which will have an even greater impact on Samsung Electronics and SK hynix stock prices in the future.\n\n5. What Is HBF?\n\nThe “F” in HBF stands for Flash Memory.\n\nHBF will serve as the device that stores AI’s core knowledge — the “encrypted books” or priors, also known as KV (Key-Value) caches.\n\n6. Characteristics of Priors\n\nThese “encrypted books” require massive capacity and are read-intensive (much more reading than writing).\n\nCompared with HBM (~100GB beside the GPU), stacking the same 16 layers of HBF provides roughly 10 times the capacity.\n\n7. Structure of HBF\nHBF is built by stacking NAND Flash, just as HBM stacks DRAM.\n\nSince NAND already uses 3D stacking technologies (128 layers, 256 layers, etc.), stacking 16 layers of HBF could yield 2,000–3,000 total layers.\n\n8. Hierarchical Memory Structure in AI Computing\n\nSRAM (inside GPU): fastest, smallest capacity.\nHBM (bookshelf beside desk): used for encoder/decoder computation, fast read/write.\n\nHBF (basement library): used for inference, refilling HBM when needed — acts as first-level backup.\n\nNetwork storage (public library): vast memory storage connected via optical networks.\n\n9. Technical Differences: DRAM vs NAND\nDRAM (HBM) leaks charge, requiring refresh cycles, but supports unlimited reads/writes.\n\nNAND Flash (HBF) traps electrons like a prison, retaining data without refresh, but uses tunneling during writes, damaging cell walls over time — limiting its endurance to about 100,000 cycles.\n\n10. Commercialization Timeline\nHBF products are expected to launch around 2027–2028.\n\n11. Competitive Landscape\nCompanies experienced in stacking memory — such as SK hynix and Samsung Electronics — are best positioned to develop HBF.\nUnlike HBM, where SK hynix led, the NAND Flash market (Samsung ~30%, SK hynix ~20%) creates a more level playing field, meaning both firms will compete on equal footing.\n\nKim also forecasts that AI memory bandwidth and capacity requirements will double every two years, implying a 1,000× increase over 20 years, which could expand revenues from hundreds of billions to tens of trillions of dollars.\n\n12. Nvidia’s Potential Strategy\nHe argues that Nvidia, in its effort to reduce dependence on external suppliers (Samsung, SK hynix), may avoid using both HBM and HBF simultaneously — possibly shifting entirely to HBF — and could therefore acquire a memory company such as Micron or SanDisk.\n\n13. The Future Is Memory\n\nWith GPU performance growth stagnating (the slowdown of Moore’s Law), innovation in memory technologies like HBM and HBF will drive further AI computing advancements.\n\nHe concludes that this signals the arrival of a “Memory Era”, where memory companies will power the next wave of stock market gains.\n\n---\n\nFull video(in Korean):\n\nhttps://t.co/oBiJYG7Nkq", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01610007841118355, "ret_signed_m1d": 0.01610007841118355, "alpha_spy_m1d": 0.019369633962219357, "alpha_c_m1d": 0.014144188394368706, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": 0.001955890016814843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10912350108961721, "ret_signed_p1d": 0.10912350108961721, "alpha_spy_p1d": 0.1072468140016527, "alpha_c_p1d": 0.1006390670867805, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.008484434002836716, "ret_p1w": 0.03756707385024027, "ret_signed_p1w": 0.03756707385024027, "alpha_spy_p1w": 0.05382674549794686, "alpha_c_p1w": 0.07861158691845949, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04104451306821921, "ret_p1m": -0.03687792093991171, "ret_signed_p1m": -0.03687792093991171, "alpha_spy_p1m": -0.03425354688475746, "alpha_c_p1m": -0.00916599350196401, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.027711927437947703, "ret_p3m": 0.5055496597062261, "ret_signed_p3m": 0.5055496597062261, "alpha_spy_p3m": 0.48294922819329034, "alpha_c_p3m": 0.35430068639247914, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.15124897331374698, "ret_p6m": 1.31719330565361, "ret_signed_p6m": 1.31719330565361, "alpha_spy_p6m": 1.2626788441025214, "alpha_c_p6m": 0.9182750577074041, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.3989182479462059, "price_path": [-0.5293, -0.5382, -0.532, -0.5418, -0.5231, -0.5195, -0.5034, -0.5061, -0.5061, -0.5222, -0.5302, -0.5436, -0.5623, -0.5516, -0.5364, -0.5329, -0.5356, -0.5197, -0.5188, -0.542, -0.534, -0.5304, -0.525, -0.5107, -0.5045, -0.4848, -0.4562, -0.4508, -0.4123, -0.4079, -0.3775, -0.4034, -0.364, -0.364, -0.3721, -0.3542, -0.3605, -0.3623, -0.398, -0.3757, -0.3784, -0.356, -0.2925, -0.2925, -0.2925, -0.2925, -0.2925, -0.2925, -0.2343, -0.2576, -0.2639, -0.2442, -0.1905, -0.1673, -0.1315, -0.1431, -0.1386, -0.144, -0.0877, -0.0429, -0.068, -0.0018, 0.0161, 0.0, 0.1091, 0.0483, 0.0358, 0.0608, 0.0376, 0.0841, 0.1073, 0.1038, 0.0948, 0.0018, 0.0841, 0.0197, 0.0054, 0.0215, -0.068, -0.0698, -0.0716, -0.0626, -0.0261, -0.0512, -0.0369, -0.0011, -0.0118, -0.0297, -0.0261, 0.0329, 0.0132, 0.0508, 0.0115, 0.0222, -0.0082, -0.0512, -0.0136, -0.0118, -0.0208, 0.0383, 0.0455, 0.0526, 0.0526, 0.0723, 0.1457, 0.1654, 0.1654, 0.1654, 0.212, 0.246, 0.2997, 0.3283, 0.3534, 0.3319, 0.3409, 0.3212, 0.3283, 0.3409, 0.3534, 0.3677, 0.3301, 0.3247, 0.3516, 0.3731, 0.3176, 0.4322, 0.5055, 0.5414, 0.6273, 0.4859, 0.6237, 0.6112, 0.5073, 0.502, 0.5879, 0.5682, 0.5396, 0.5897, 0.5754, 0.5754, 0.5754, 0.5754, 0.6004, 0.6989, 0.7025, 0.7991, 0.8224, 0.971, 0.9029, 0.9029, 0.6841, 0.5227, 0.6877, 0.6572, 0.4994, 0.6823, 0.7128, 0.6679, 0.6321, 0.7469, 0.7397, 0.8939, 0.8168, 0.806, 0.6733, 0.7684, 0.7845, 0.6733, 0.6733, 0.5657, 0.4473, 0.6088, 0.4886, 0.5711, 0.589, 0.6428, 0.8527, 0.7899, 0.8419, 0.8652, 0.9782, 1.0374, 1.0715, 1.0231, 1.0912, 1.1952, 1.1934, 1.197, 1.1916, 1.3172]}
{"unit_id": "thread:1984213326780629083", "ticker_idx": 1, "ts": "2025-10-31T10:58:16+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Companies experienced in stacking memory — such as SK hynix and Samsung Electronics — are best positioned to develop HBF", "tweet_type": "analysis", "timeline": "years", "summary": "Relays bullish thesis: memory era incoming via HBF (2027-28); SK hynix and Samsung best positioned; Micron/SanDisk potential Nvidia acquisition targets.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Father of HBM: Nvidia May Acquire a Memory Company Like SanDisk”\n\nProfessor Jung-Ho Kim, known as the “father of HBM (High Bandwidth Memory)” and a professor in the Department of Electrical and Electronic Engineering at KAIST, appeared in a YouTube video and made a bold claim — that Nvidia could acquire a memory company such as SanDisk or Micron.\n\n(He previously recommended buying SK hynix stock back in 2017.)\n\n1. Shift of Market Power: From GPU to Memory\n\nKim argues that as the AI era unfolds, the driving force of the stock market is shifting from GPU (Graphics Processing Units) to memory.\n\nIf the GPU is the castella cake, memory is the whipped cream — to improve AI performance, memory upgrades and capacity expansion are essential.\n\n2. Memory Determines AI Speed\n\nThe speed at which generative AI produces words in response to prompts is determined by memory, and this is even more critical during inference than during training.\n\n3. Explosive Memory Pressure\n\nAs AI model sizes (parameters), user counts, and data types (text → image → video) grow exponentially, the demand pressure on memory capacity is reaching its peak.\n\n4. From HBM to HBF Era\n\nThe professor predicts that the HBM (High Bandwidth Memory) era is ending, and the HBF (High Bandwidth Flash) era is about to begin — which will have an even greater impact on Samsung Electronics and SK hynix stock prices in the future.\n\n5. What Is HBF?\n\nThe “F” in HBF stands for Flash Memory.\n\nHBF will serve as the device that stores AI’s core knowledge — the “encrypted books” or priors, also known as KV (Key-Value) caches.\n\n6. Characteristics of Priors\n\nThese “encrypted books” require massive capacity and are read-intensive (much more reading than writing).\n\nCompared with HBM (~100GB beside the GPU), stacking the same 16 layers of HBF provides roughly 10 times the capacity.\n\n7. Structure of HBF\nHBF is built by stacking NAND Flash, just as HBM stacks DRAM.\n\nSince NAND already uses 3D stacking technologies (128 layers, 256 layers, etc.), stacking 16 layers of HBF could yield 2,000–3,000 total layers.\n\n8. Hierarchical Memory Structure in AI Computing\n\nSRAM (inside GPU): fastest, smallest capacity.\nHBM (bookshelf beside desk): used for encoder/decoder computation, fast read/write.\n\nHBF (basement library): used for inference, refilling HBM when needed — acts as first-level backup.\n\nNetwork storage (public library): vast memory storage connected via optical networks.\n\n9. Technical Differences: DRAM vs NAND\nDRAM (HBM) leaks charge, requiring refresh cycles, but supports unlimited reads/writes.\n\nNAND Flash (HBF) traps electrons like a prison, retaining data without refresh, but uses tunneling during writes, damaging cell walls over time — limiting its endurance to about 100,000 cycles.\n\n10. Commercialization Timeline\nHBF products are expected to launch around 2027–2028.\n\n11. Competitive Landscape\nCompanies experienced in stacking memory — such as SK hynix and Samsung Electronics — are best positioned to develop HBF.\nUnlike HBM, where SK hynix led, the NAND Flash market (Samsung ~30%, SK hynix ~20%) creates a more level playing field, meaning both firms will compete on equal footing.\n\nKim also forecasts that AI memory bandwidth and capacity requirements will double every two years, implying a 1,000× increase over 20 years, which could expand revenues from hundreds of billions to tens of trillions of dollars.\n\n12. Nvidia’s Potential Strategy\nHe argues that Nvidia, in its effort to reduce dependence on external suppliers (Samsung, SK hynix), may avoid using both HBM and HBF simultaneously — possibly shifting entirely to HBF — and could therefore acquire a memory company such as Micron or SanDisk.\n\n13. The Future Is Memory\n\nWith GPU performance growth stagnating (the slowdown of Moore’s Law), innovation in memory technologies like HBM and HBF will drive further AI computing advancements.\n\nHe concludes that this signals the arrival of a “Memory Era”, where memory companies will power the next wave of stock market gains.\n\n---\n\nFull video(in Korean):\n\nhttps://t.co/oBiJYG7Nkq", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031627900169409084, "ret_signed_m1d": -0.031627900169409084, "alpha_spy_m1d": -0.028358344618373277, "alpha_c_m1d": -0.030663475197112433, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": -0.0009644249722966514, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03348832614351105, "ret_signed_p1d": 0.03348832614351105, "alpha_spy_p1d": 0.031611639055546537, "alpha_c_p1d": 0.029397560278537016, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.004090765864974033, "ret_p1w": -0.08930227622821552, "ret_signed_p1w": -0.08930227622821552, "alpha_spy_p1w": -0.07304260458050893, "alpha_c_p1w": -0.04766325858196607, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04163901764624944, "ret_p1m": -0.0623255507623407, "ret_signed_p1m": -0.0623255507623407, "alpha_spy_p1m": -0.05970117670718644, "alpha_c_p1m": -0.014666917306523919, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.04765863345581678, "ret_p3m": 0.5180414411843266, "ret_signed_p3m": 0.5180414411843266, "alpha_spy_p3m": 0.4954410096713908, "alpha_c_p3m": 0.5239191818977161, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": -0.005877740713389534, "ret_p6m": 1.102867513345993, "ret_signed_p6m": 1.102867513345993, "alpha_spy_p6m": 1.0483530517949045, "alpha_c_p6m": 1.0318717496508736, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.07099576369511951, "price_path": [-0.3527, -0.3628, -0.3471, -0.3351, -0.3425, -0.3415, -0.3341, -0.3369, -0.3369, -0.3517, -0.3517, -0.3471, -0.3462, -0.3388, -0.3378, -0.349, -0.3462, -0.3554, -0.3545, -0.374, -0.3601, -0.3536, -0.3508, -0.3564, -0.3508, -0.3378, -0.3277, -0.3202, -0.3017, -0.2915, -0.2647, -0.2758, -0.2563, -0.2563, -0.2267, -0.2156, -0.2091, -0.2026, -0.2286, -0.2167, -0.2195, -0.2, -0.1651, -0.1651, -0.1651, -0.1651, -0.1651, -0.1651, -0.1219, -0.1321, -0.1479, -0.1163, -0.0912, -0.0893, -0.0874, -0.093, -0.0828, -0.1023, -0.0809, -0.0512, -0.0744, -0.0651, -0.0316, 0.0, 0.0335, -0.0242, -0.0642, -0.0772, -0.0893, -0.0642, -0.0372, -0.0409, -0.0437, -0.0958, -0.0642, -0.0902, -0.1023, -0.0642, -0.1181, -0.1005, -0.0763, -0.0437, -0.0372, -0.0651, -0.0623, -0.0381, -0.0279, -0.0223, 0.0084, 0.0186, 0.0084, 0.0047, -0.0019, 0.013, -0.0251, -0.0437, 0.0037, 0.0009, -0.0112, 0.0279, 0.0372, 0.0335, 0.0335, 0.0884, 0.117, 0.1208, 0.1208, 0.1208, 0.2012, 0.2909, 0.2984, 0.318, 0.2974, 0.2993, 0.2974, 0.2862, 0.3115, 0.3451, 0.3918, 0.3956, 0.3573, 0.3975, 0.4236, 0.4218, 0.4218, 0.4909, 0.518, 0.5022, 0.5003, 0.4059, 0.5657, 0.5807, 0.4891, 0.4825, 0.5554, 0.5498, 0.5685, 0.6695, 0.6938, 0.6938, 0.6938, 0.6938, 0.776, 0.777, 0.8041, 0.8695, 0.9022, 1.0378, 1.0237, 1.0237, 0.8237, 0.6096, 0.791, 0.7592, 0.6218, 0.7564, 0.776, 0.7564, 0.7153, 0.7639, 0.8125, 0.949, 0.8742, 0.8639, 0.7414, 0.7732, 0.7667, 0.6835, 0.6835, 0.6514, 0.5661, 0.7764, 0.6711, 0.7441, 0.8087, 0.8406, 0.9717, 0.9108, 0.9296, 0.8827, 0.9343, 0.9764, 1.0373, 1.0232, 1.0092, 1.0513, 1.0373, 1.1029, 1.056, 1.1029]}
{"unit_id": "thread:1984213326780629083", "ticker_idx": 2, "ts": "2025-10-31T10:58:16+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia… may… acquire a memory company such as Micron or SanDisk", "tweet_type": "analysis", "timeline": "years", "summary": "Relays bullish thesis: memory era incoming via HBF (2027-28); SK hynix and Samsung best positioned; Micron/SanDisk potential Nvidia acquisition targets.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Father of HBM: Nvidia May Acquire a Memory Company Like SanDisk”\n\nProfessor Jung-Ho Kim, known as the “father of HBM (High Bandwidth Memory)” and a professor in the Department of Electrical and Electronic Engineering at KAIST, appeared in a YouTube video and made a bold claim — that Nvidia could acquire a memory company such as SanDisk or Micron.\n\n(He previously recommended buying SK hynix stock back in 2017.)\n\n1. Shift of Market Power: From GPU to Memory\n\nKim argues that as the AI era unfolds, the driving force of the stock market is shifting from GPU (Graphics Processing Units) to memory.\n\nIf the GPU is the castella cake, memory is the whipped cream — to improve AI performance, memory upgrades and capacity expansion are essential.\n\n2. Memory Determines AI Speed\n\nThe speed at which generative AI produces words in response to prompts is determined by memory, and this is even more critical during inference than during training.\n\n3. Explosive Memory Pressure\n\nAs AI model sizes (parameters), user counts, and data types (text → image → video) grow exponentially, the demand pressure on memory capacity is reaching its peak.\n\n4. From HBM to HBF Era\n\nThe professor predicts that the HBM (High Bandwidth Memory) era is ending, and the HBF (High Bandwidth Flash) era is about to begin — which will have an even greater impact on Samsung Electronics and SK hynix stock prices in the future.\n\n5. What Is HBF?\n\nThe “F” in HBF stands for Flash Memory.\n\nHBF will serve as the device that stores AI’s core knowledge — the “encrypted books” or priors, also known as KV (Key-Value) caches.\n\n6. Characteristics of Priors\n\nThese “encrypted books” require massive capacity and are read-intensive (much more reading than writing).\n\nCompared with HBM (~100GB beside the GPU), stacking the same 16 layers of HBF provides roughly 10 times the capacity.\n\n7. Structure of HBF\nHBF is built by stacking NAND Flash, just as HBM stacks DRAM.\n\nSince NAND already uses 3D stacking technologies (128 layers, 256 layers, etc.), stacking 16 layers of HBF could yield 2,000–3,000 total layers.\n\n8. Hierarchical Memory Structure in AI Computing\n\nSRAM (inside GPU): fastest, smallest capacity.\nHBM (bookshelf beside desk): used for encoder/decoder computation, fast read/write.\n\nHBF (basement library): used for inference, refilling HBM when needed — acts as first-level backup.\n\nNetwork storage (public library): vast memory storage connected via optical networks.\n\n9. Technical Differences: DRAM vs NAND\nDRAM (HBM) leaks charge, requiring refresh cycles, but supports unlimited reads/writes.\n\nNAND Flash (HBF) traps electrons like a prison, retaining data without refresh, but uses tunneling during writes, damaging cell walls over time — limiting its endurance to about 100,000 cycles.\n\n10. Commercialization Timeline\nHBF products are expected to launch around 2027–2028.\n\n11. Competitive Landscape\nCompanies experienced in stacking memory — such as SK hynix and Samsung Electronics — are best positioned to develop HBF.\nUnlike HBM, where SK hynix led, the NAND Flash market (Samsung ~30%, SK hynix ~20%) creates a more level playing field, meaning both firms will compete on equal footing.\n\nKim also forecasts that AI memory bandwidth and capacity requirements will double every two years, implying a 1,000× increase over 20 years, which could expand revenues from hundreds of billions to tens of trillions of dollars.\n\n12. Nvidia’s Potential Strategy\nHe argues that Nvidia, in its effort to reduce dependence on external suppliers (Samsung, SK hynix), may avoid using both HBM and HBF simultaneously — possibly shifting entirely to HBF — and could therefore acquire a memory company such as Micron or SanDisk.\n\n13. The Future Is Memory\n\nWith GPU performance growth stagnating (the slowdown of Moore’s Law), innovation in memory technologies like HBM and HBF will drive further AI computing advancements.\n\nHe concludes that this signals the arrival of a “Memory Era”, where memory companies will power the next wave of stock market gains.\n\n---\n\nFull video(in Korean):\n\nhttps://t.co/oBiJYG7Nkq", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001072551239470254, "ret_signed_m1d": 0.001072551239470254, "alpha_spy_m1d": 0.004342106790506062, "alpha_c_m1d": -0.0008833387773445889, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": 0.001955890016814843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048844826963758026, "ret_signed_p1d": 0.048844826963758026, "alpha_spy_p1d": 0.04696813987579351, "alpha_c_p1d": 0.04036039296092131, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.008484434002836716, "ret_p1w": 0.06323452544896702, "ret_signed_p1w": 0.06323452544896702, "alpha_spy_p1w": 0.0794941970966736, "alpha_c_p1w": 0.10427903851718623, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04104451306821921, "ret_p1m": 0.0745855107445188, "ret_signed_p1m": 0.0745855107445188, "alpha_spy_p1m": 0.07720988479967306, "alpha_c_p1m": 0.1022974381824665, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.027711927437947703, "ret_p3m": 0.9459974955150556, "ret_signed_p3m": 0.9459974955150556, "alpha_spy_p3m": 0.9233970640021198, "alpha_c_p3m": 0.7947485222013086, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.15124897331374698, "ret_p6m": 1.3461248968037958, "ret_signed_p6m": 1.3461248968037958, "alpha_spy_p6m": 1.2916104352527071, "alpha_c_p6m": 0.9472066488575899, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.3989182479462059, "price_path": [-0.5129, -0.5142, -0.5004, -0.469, -0.4475, -0.4295, -0.445, -0.4404, -0.4602, -0.4482, -0.4549, -0.4765, -0.4829, -0.4744, -0.4801, -0.4797, -0.4741, -0.4551, -0.4685, -0.4685, -0.4709, -0.4698, -0.4453, -0.4133, -0.4129, -0.396, -0.3747, -0.3275, -0.2978, -0.2954, -0.2907, -0.2855, -0.2457, -0.2732, -0.2648, -0.2568, -0.2778, -0.2996, -0.2976, -0.268, -0.2527, -0.1865, -0.1794, -0.1606, -0.1466, -0.1702, -0.1217, -0.1405, -0.1885, -0.1385, -0.1641, -0.1422, -0.0949, -0.0956, -0.076, -0.096, -0.1131, -0.0762, -0.0212, -0.0164, -0.0083, 0.0128, 0.0011, 0.0, 0.0488, -0.0257, 0.0614, 0.0651, 0.0632, 0.132, 0.0775, 0.0944, 0.0589, 0.1031, 0.0812, 0.0211, 0.0096, -0.1001, -0.0733, 0.0007, 0.0034, 0.029, 0.029, 0.0568, 0.0746, 0.0703, 0.0464, 0.0129, 0.0601, 0.1035, 0.128, 0.1785, 0.155, 0.0776, 0.0614, 0.0391, 0.0078, 0.1107, 0.1884, 0.236, 0.2346, 0.2811, 0.2811, 0.2727, 0.316, 0.3083, 0.276, 0.276, 0.4101, 0.3955, 0.5354, 0.518, 0.462, 0.5428, 0.5463, 0.5117, 0.4903, 0.505, 0.6217, 0.6217, 0.6318, 0.7396, 0.7775, 0.7867, 0.7395, 0.8341, 0.946, 0.9483, 0.8548, 0.9573, 0.8752, 0.6962, 0.7118, 0.7645, 0.7145, 0.6687, 0.8345, 0.8507, 0.8404, 0.8404, 0.7873, 0.8819, 0.8658, 0.9142, 0.882, 0.8688, 0.9179, 0.8578, 0.8436, 0.8449, 0.6974, 0.7917, 0.7751, 0.6555, 0.7405, 0.8022, 0.8718, 0.8122, 0.9051, 0.9751, 1.0641, 1.0642, 0.9862, 0.8907, 0.8077, 0.7683, 0.7082, 0.5891, 0.597, 0.4393, 0.511, 0.6452, 0.638, 0.638, 0.6896, 0.6887, 0.8191, 0.8852, 0.8811, 0.9078, 1.0827, 1.0405, 1.045, 1.0353, 1.0056, 1.0099, 1.1803, 1.1545, 1.2216, 1.3461]}
{"unit_id": "orig:1988755845027623247", "ticker_idx": 0, "ts": "2025-11-12T23:48:37+00:00", "symbol": "Infineon Technologies", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the AI power semiconductor market is mainly shared by three players: Infineon, and Renesas and Monolithic Power... the technological leaders have already moved to secure an early lead", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Barclays thesis: Infineon among top beneficiaries of 10x power semiconductor content growth per rack as NVIDIA's 800V architecture reshapes the market from 2027.", "resolved_tickers": ["IFX.DE"], "resolution_method": "exchange_ticker", "resolution_reason": "Infineon Technologies IFX.DE Germany", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Why are power semiconductors the future? Barclays:\n\nThe power consumption of AI chips is rising rapidly. NVIDIA’s current Blackwell-generation GPU chips have already reached around 1.2 kW, and the power consumption of GB200 rack-level servers is about 120 kW.\n\nEven more striking is the Rubin Ultra chip scheduled for release in 2027. The power consumption of this GPU is expected to reach three times that of Blackwell, and the power consumption per server rack will rise to around 600 kW. By the time of the Feynman architecture, power is expected to increase by another 1.5 times, reaching more than eight times the current Blackwell level, with rack-level server power consumption likely to hit the megawatt (1 MW) range.\n\nTo address this issue, NVIDIA has announced a new power system called “Kyber,” which adopts an 800 V architecture. This will significantly increase the content of power semiconductors, and all major power semiconductor manufacturers are now competing to enter this supply chain.\n\nPower semiconductor content to grow 10x, reaching $140,000 per rack\n\nThe report expects GPU power demand to grow about fivefold over the next few years, while rack power demand will grow tenfold, with 1 MW-per-rack demand already coming into view.\n\nCurrently, the value of power semiconductors in a single rack is around $12,000–$15,000. Taking into account the reference design and the increase in power demand, the report projects that by 2027, the power semiconductor content (on a BoM basis) will grow more than sixfold, and by 2029 it will increase by more than another 50%, reaching about $140,000 per rack (at today’s prices).\n\nWhat is noteworthy is that even under the most optimistic assumptions, power semiconductors still account for less than 1% of total rack spending. This clearly illustrates how much “leverage” power semiconductors have within the overall system.\n\n800 V architecture reshapes the power semiconductor market landscape\n\nThe traditional 54 V power delivery architecture scales poorly once rack power demand exceeds 200 kW. NVIDIA’s proposed 800 V data center architecture is a fundamentally different approach that overturns this limitation.\n\nIn the new architecture, it is no longer a matter of converting 54 V DC down to 1 V, but implementing a large-step conversion from 800 V down to 1 V. This change will significantly reduce, or in some cases entirely replace, the role of PSUs (power supply units) in data centers. As a result, it can save space, while also reducing power loss and copper usage.\n\nThis new architecture is expected to begin deployment from 2027, coinciding with the launch timing of NVIDIA’s Rubin Ultra chips.\n\nNew opportunities for wide-bandgap semiconductors, with companies like Infineon set to benefit\n\nThe report points out that in recent years, silicon carbide (SiC) has struggled due to oversupply, while gallium nitride (GaN) has been mostly confined to consumer applications. However, the surge in AI power demand is creating major new opportunities for both of these materials.\n\nIn future server architectures, GaN could account for up to 30% of the bill of materials (BoM), while SiC is expected to account for around 10–15%.\n\nA hundred-billion-dollar-scale market is taking shape, with the competitive landscape likely to be reshaped\n\nThe report estimates that if AI data center capacity increases by around 16 GW every month, the annual market size for power semiconductors will reach 1.4–8.5 billion dollars. This projection is based on the assumption that all newly added capacity adopts the new 800 V architecture.\n\nAt present, the AI power semiconductor market is mainly shared by three players: Infineon, and Renesas and Monolithic Power, which are outside the coverage of this report. However, NVIDIA has already released a long list of potential suppliers for its 800 V architecture, with a total of 14 companies named as potential vendors.\n\nAs the 800 V data center era begins in earnest, competition among power semiconductor suppliers is expected to intensify. The report assumes that, over the forecast period, AI power semiconductor prices will see a low- to mid-single-digit annual decline.\n\nThis “energy efficiency revolution” triggered by AI compute demand is reshaping the structure of the power semiconductor industry. As the 800 V architecture is gradually commercialized from 2027 onward, a hundred-billion-dollar-scale new market is quietly taking shape, and the technological leaders have already moved to secure an early lead.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06473423841051873, "ret_signed_m1d": -0.06473423841051873, "alpha_spy_m1d": -0.06417814921133214, "alpha_c_m1d": -0.05213335579336509, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.016594021470737852, "alpha_c_p1d": 0.03012398366125879, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.09523801438783874, "ret_signed_p1w": -0.09523801438783874, "alpha_spy_p1w": -0.06487434049485685, "alpha_c_p1w": -0.04998169196849689, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": 0.009523801438783952, "ret_signed_p1m": 0.009523801438783952, "alpha_spy_p1m": 0.0010512332264629443, "alpha_c_p1m": -0.03367937228394369, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.15914433606585465, "ret_signed_p3m": 0.15914433606585465, "alpha_spy_p3m": 0.1406764439969208, "alpha_c_p3m": 0.011799984239602823, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 0.6551509271404623, "ret_signed_p6m": 0.6551509271404623, "alpha_spy_p6m": 0.5785235241497786, "alpha_c_p6m": 0.13128355787704105, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [0.0119, 0.012, 0.0214, 0.0048, -0.0021, 0.0173, 0.0119, 0.0018, -0.009, 0.0025, -0.0349, -0.0431, -0.0859, -0.13, -0.1366, -0.1284, -0.1146, -0.1143, -0.1213, -0.1188, -0.1234, -0.0987, -0.1104, -0.0987, -0.0701, -0.0893, -0.0893, -0.0669, -0.0751, -0.0803, -0.0983, -0.0879, -0.0879, -0.0738, -0.0549, -0.0576, -0.0592, -0.0741, -0.0874, -0.0841, -0.1161, -0.1005, -0.1132, -0.1026, -0.0818, -0.1039, -0.0587, -0.0272, -0.0556, -0.0774, -0.0754, -0.0534, -0.054, -0.0511, -0.0497, -0.0526, -0.0446, -0.0562, -0.0545, -0.0636, -0.0636, -0.0795, -0.0647, 0.0, 0.0, -0.0257, -0.0509, -0.0894, -0.0952, -0.0915, -0.1251, -0.0928, -0.0738, -0.0396, -0.0142, 0.0029, -0.0228, -0.0123, 0.0144, 0.0076, 0.0353, 0.0382, 0.0353, 0.0134, 0.0095, 0.0018, 0.0018, -0.0062, -0.0062, -0.0062, -0.0115, 0.0123, 0.0137, 0.0137, 0.0137, 0.0137, 0.0142, 0.0415, 0.0415, 0.0415, 0.0563, 0.1017, 0.1546, 0.1631, 0.1197, 0.147, 0.1502, 0.1742, 0.1429, 0.1485, 0.1575, 0.1217, 0.1173, 0.134, 0.1756, 0.1676, 0.1554, 0.1683, 0.1986, 0.1539, 0.1487, 0.1561, 0.1339, 0.1126, 0.143, 0.1604, 0.1591, 0.1833, 0.2011, 0.1801, 0.2011, 0.193, 0.2306, 0.2723, 0.2653, 0.2574, 0.2766, 0.2985, 0.3095, 0.301, 0.2747, 0.2355, 0.1594, 0.2224, 0.1767, 0.0966, 0.0835, 0.1493, 0.1558, 0.1202, 0.1081, 0.097, 0.1026, 0.1119, 0.0319, 0.0472, 0.0707, 0.0658, 0.1032, 0.0817, 0.0412, 0.0382, 0.0571, 0.1169, 0.0838, 0.0838, 0.0838, 0.0644, 0.1903, 0.1792, 0.1913, 0.1942, 0.2366, 0.2439, 0.2798, 0.3603, 0.3351, 0.3279, 0.3739, 0.4841, 0.5063, 0.491, 0.4688, 0.5494, 0.5892, 0.5892, 0.5803, 0.683, 0.6476, 0.6552]}
{"unit_id": "orig:1985550014312300867", "ticker_idx": 0, "ts": "2025-11-04T03:29:47+00:00", "symbol": "Silicon Motion Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the most severe, panic-level, large-scale shortage in the history of memory is approaching, and the NAND Flash supply shortfall is expected to exceed 80%", "tweet_type": "analysis", "timeline": "years", "summary": "Panic-level NAND/DRAM/HBM shortage approaching with 70-80% supply gap persisting through 2026+, bullish on memory broadly — NAND, HBM3E, SSDs, HDDs all in simultaneous shortage driven by AI inference surge.", "resolved_tickers": ["SIMO"], "resolution_method": "exchange_ticker", "resolution_reason": "Silicon Motion Technology listed on Nasdaq as SIMO", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) Gou Jiazhang of Silicon Motion (慧榮): The most severe, panic-level, large-scale shortage in the history of memory is approaching, and the NAND Flash supply shortfall is expected to exceed 80%.\n\nGou Jiazhang of Silicon Motion (慧榮) stated that this time, all types of memory—DRAM, NAND, SSD, and HDD—are, for the first time ever, experiencing a simultaneous, large-scale shortage. The growth of AI inference has already surpassed AI training, and inference drives demand for memory—especially high-capacity SSDs—much more strongly. He asserted that the current shortage is nearing a “panic” level and that the supply shortfall will not be resolved throughout 2026. He said this, too, is a conservative view, and that if this trend continues, smaller, less competitive firms could face an existential crisis.\n\nAI inference is driving a surge in HBM3E demand\n\nSharing takeaways from a recent business trip to Korea, Gou said customers kept coming to discuss only capacity and never price; the mood was that simply getting a capacity allocation would be fortunate. He said HBM demand is very strong, and since AI inference demand emerged, HBM3E demand has expanded. It is not only NVIDIA’s GPUs that require HBM3E; all cloud providers’ in-house ASICs also need HBM3E. At present, Samsung, SK hynix, and Micron are prioritizing capacity allocation to HBM and DDR5. From a capacity-consumption perspective, each unit of HBM requires roughly three times the number of wafers compared with DDR5. As a result, HBM is in short supply now—and so is DDR5.\n\nThe HDD shortage is shifting demand en masse to the NAND industry\n\nIn the China market, starting from the second quarter of this year, SSD demand has rapidly stepped up from the past 16TB/32TB to 64TB/128TB. Looking closely at the cause, because sufficient HDD volumes could not be secured, high-capacity SSDs were purchased as substitutes. Gou also analyzed that, thanks to strong AI inference demand, even if you add up all NAND Flash capacity worldwide, it still cannot meet even one-tenth of data centers’ annual warm-data demand. Therefore, over the next 10 years it will be impossible for SSDs to replace HDDs. He added that timing is the only question.\n\nTracing the current HDD shortage back, during the pandemic the blow from excessive inventories made HDD makers’ business models more conservative, shifting them to expand capacity only to the extent of the orders they received. This has made it difficult to absorb the sudden volume surge from the AI boom all at once. On top of this, the North American hyperscalers’ demand for ultra-high-capacity SSDs has increased to an even more astonishing level. The industry must wait for NAND technology to transition from TLC to QLC; at the earliest, in 2027 the global bit output of QLC may overtake that of TLC NAND.\n\nThe NAND shortage could trigger a reshuffle of the global smartphone market\n\nApple has removed the 128GB model in the iPhone 17 and pushed capacity straight up to 2TB, and Samsung is expected to follow. For 2026, overall smartphone bit growth is forecast at about 32%. In the context of a NAND Flash shortage, major vendors will defend premium smartphones from impact, but low-end models may see capacity cut. A recent example is smartphone storage in Africa and Southeast Asia falling from 128GB to 64GB.\n\nThe NAND Flash shortage will also affect the competitive landscape of the global smartphone market. Apple need not worry about shortages, and Samsung can be supported by its in-house NAND. The focus should be on branded smartphones in the China-based Android camp. Although the local NAND giant YMTC長江存儲 is actively expanding, it is estimated it can meet only 50% of local vendors’ demand. Recently, Xiaomi announced price hikes for new models citing rising memory prices. Going forward, whether smartphone vendors can secure sufficient NAND chips—and what choices they make between specifications and cost/selling price—will determine how market shares shift.\n\nQuestions about an AI bubble\n\nGou stated that concerns about an AI bubble will not materialize in the short term. From cloud AI to edge AI, inference consumes memory in very large quantities. The supply gap in the NAND market is 70%–80%, already at a panic-level shortage. Some overbooking naturally exists, but the key is how the major memory makers allocate and plan capacity. Including customer groups such as cloud, smartphones, autos, and PCs, many Chinese automakers still have not recognized the severity of the shortage. Moreover, all capacity cannot be diverted solely to cloud providers; doing so would lead to fragmentation of the technology industry, which is undesirable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.058186206801379425, "ret_signed_m1d": 0.058186206801379425, "alpha_spy_m1d": 0.046190474972560125, "alpha_c_m1d": 0.02037014513060642, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.037816061670773005, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07010511815913367, "ret_signed_p1d": 0.07010511815913367, "alpha_spy_p1d": 0.06663972896213743, "alpha_c_p1d": 0.05099866470761283, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.019106453451520844, "ret_p1w": 0.009968598788362426, "ret_signed_p1w": 0.009968598788362426, "alpha_spy_p1w": -0.0015235193852181084, "alpha_c_p1w": 0.01481613315847452, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.004847534370112094, "ret_p1m": -0.008796604704787159, "ret_signed_p1m": -0.008796604704787159, "alpha_spy_p1m": -0.021606845261213037, "alpha_c_p1m": -0.041708464623713803, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": 0.032911859918926645, "ret_p3m": 0.29613880932995285, "ret_signed_p3m": 0.29613880932995285, "alpha_spy_p3m": 0.26833442669120644, "alpha_c_p3m": 0.14885629231932462, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": 0.14728251701062822, "ret_p6m": 1.3794472399844606, "ret_signed_p6m": 1.3794472399844606, "alpha_spy_p6m": 1.3196289372461114, "alpha_c_p6m": 0.958837026873451, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.42061021311100966, "price_path": [-0.1928, -0.1724, -0.1869, -0.1613, -0.1696, -0.1821, -0.2018, -0.1853, -0.2019, -0.1967, -0.196, -0.1718, -0.1583, -0.144, -0.1487, -0.1034, -0.1366, -0.1366, -0.1353, -0.1387, -0.1205, -0.1074, -0.0806, -0.0908, -0.0931, -0.0348, -0.0386, -0.0247, -0.0217, -0.0415, -0.01, 0.0009, 0.03, -0.0102, 0.0016, -0.0215, -0.0184, 0.0163, 0.0273, 0.0543, 0.106, 0.0711, 0.0834, 0.0234, 0.017, 0.0204, -0.0703, -0.0273, -0.0385, 0.0039, 0.0008, 0.02, 0.0416, 0.0196, -0.0011, 0.0297, 0.0741, 0.1157, 0.0884, 0.09, 0.0831, 0.0631, 0.0582, 0.0, 0.0701, 0.0334, 0.0154, 0.0234, 0.01, -0.0112, -0.0617, -0.0547, -0.0685, -0.0787, -0.0758, -0.1264, -0.1168, -0.0863, -0.0768, -0.0455, -0.0455, -0.0305, -0.0332, -0.0341, -0.0088, -0.008, 0.0049, 0.0119, 0.0307, 0.0432, 0.0164, -0.0441, -0.0511, -0.0682, -0.0787, -0.0566, -0.0328, -0.0241, -0.0232, -0.0292, -0.0292, -0.0166, -0.0286, -0.0141, 0.0103, 0.0103, 0.0218, 0.0199, 0.1466, 0.3201, 0.21, 0.2328, 0.2568, 0.2376, 0.2178, 0.2357, 0.2301, 0.2301, 0.2316, 0.2839, 0.2563, 0.2307, 0.2331, 0.2668, 0.2762, 0.2661, 0.2961, 0.3051, 0.3124, 0.399, 0.3752, 0.3956, 0.4991, 0.4312, 0.5309, 0.5135, 0.4909, 0.4909, 0.4371, 0.4524, 0.4475, 0.4756, 0.4896, 0.4866, 0.5645, 0.434, 0.4138, 0.4133, 0.3066, 0.338, 0.3539, 0.2947, 0.3306, 0.3432, 0.3526, 0.314, 0.3485, 0.3781, 0.3804, 0.3719, 0.3948, 0.3539, 0.3536, 0.2979, 0.2136, 0.2169, 0.2041, 0.1536, 0.2285, 0.2798, 0.2461, 0.2461, 0.2755, 0.2801, 0.3319, 0.3894, 0.3862, 0.3784, 0.4238, 0.521, 0.4988, 0.5284, 0.5174, 0.5584, 0.5529, 0.5535, 0.6789, 0.6141, 0.632, 1.3794]}
{"unit_id": "orig:1985550014312300867", "ticker_idx": 2, "ts": "2025-11-04T03:29:47+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM demand is very strong, and since AI inference demand emerged, HBM3E demand has expanded", "tweet_type": "analysis", "timeline": "years", "summary": "Panic-level NAND/DRAM/HBM shortage approaching with 70-80% supply gap persisting through 2026+, bullish on memory broadly — NAND, HBM3E, SSDs, HDDs all in simultaneous shortage driven by AI inference surge.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) Gou Jiazhang of Silicon Motion (慧榮): The most severe, panic-level, large-scale shortage in the history of memory is approaching, and the NAND Flash supply shortfall is expected to exceed 80%.\n\nGou Jiazhang of Silicon Motion (慧榮) stated that this time, all types of memory—DRAM, NAND, SSD, and HDD—are, for the first time ever, experiencing a simultaneous, large-scale shortage. The growth of AI inference has already surpassed AI training, and inference drives demand for memory—especially high-capacity SSDs—much more strongly. He asserted that the current shortage is nearing a “panic” level and that the supply shortfall will not be resolved throughout 2026. He said this, too, is a conservative view, and that if this trend continues, smaller, less competitive firms could face an existential crisis.\n\nAI inference is driving a surge in HBM3E demand\n\nSharing takeaways from a recent business trip to Korea, Gou said customers kept coming to discuss only capacity and never price; the mood was that simply getting a capacity allocation would be fortunate. He said HBM demand is very strong, and since AI inference demand emerged, HBM3E demand has expanded. It is not only NVIDIA’s GPUs that require HBM3E; all cloud providers’ in-house ASICs also need HBM3E. At present, Samsung, SK hynix, and Micron are prioritizing capacity allocation to HBM and DDR5. From a capacity-consumption perspective, each unit of HBM requires roughly three times the number of wafers compared with DDR5. As a result, HBM is in short supply now—and so is DDR5.\n\nThe HDD shortage is shifting demand en masse to the NAND industry\n\nIn the China market, starting from the second quarter of this year, SSD demand has rapidly stepped up from the past 16TB/32TB to 64TB/128TB. Looking closely at the cause, because sufficient HDD volumes could not be secured, high-capacity SSDs were purchased as substitutes. Gou also analyzed that, thanks to strong AI inference demand, even if you add up all NAND Flash capacity worldwide, it still cannot meet even one-tenth of data centers’ annual warm-data demand. Therefore, over the next 10 years it will be impossible for SSDs to replace HDDs. He added that timing is the only question.\n\nTracing the current HDD shortage back, during the pandemic the blow from excessive inventories made HDD makers’ business models more conservative, shifting them to expand capacity only to the extent of the orders they received. This has made it difficult to absorb the sudden volume surge from the AI boom all at once. On top of this, the North American hyperscalers’ demand for ultra-high-capacity SSDs has increased to an even more astonishing level. The industry must wait for NAND technology to transition from TLC to QLC; at the earliest, in 2027 the global bit output of QLC may overtake that of TLC NAND.\n\nThe NAND shortage could trigger a reshuffle of the global smartphone market\n\nApple has removed the 128GB model in the iPhone 17 and pushed capacity straight up to 2TB, and Samsung is expected to follow. For 2026, overall smartphone bit growth is forecast at about 32%. In the context of a NAND Flash shortage, major vendors will defend premium smartphones from impact, but low-end models may see capacity cut. A recent example is smartphone storage in Africa and Southeast Asia falling from 128GB to 64GB.\n\nThe NAND Flash shortage will also affect the competitive landscape of the global smartphone market. Apple need not worry about shortages, and Samsung can be supported by its in-house NAND. The focus should be on branded smartphones in the China-based Android camp. Although the local NAND giant YMTC長江存儲 is actively expanding, it is estimated it can meet only 50% of local vendors’ demand. Recently, Xiaomi announced price hikes for new models citing rising memory prices. Going forward, whether smartphone vendors can secure sufficient NAND chips—and what choices they make between specifications and cost/selling price—will determine how market shares shift.\n\nQuestions about an AI bubble\n\nGou stated that concerns about an AI bubble will not materialize in the short term. From cloud AI to edge AI, inference consumes memory in very large quantities. The supply gap in the NAND market is 70%–80%, already at a panic-level shortage. Some overbooking naturally exists, but the key is how the major memory makers allocate and plan capacity. Including customer groups such as cloud, smartphones, autos, and PCs, many Chinese automakers still have not recognized the severity of the shortage. Moreover, all capacity cannot be diverted solely to cloud providers; doing so would lead to fragmentation of the technology industry, which is undesirable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.058020619072690405, "ret_signed_m1d": 0.058020619072690405, "alpha_spy_m1d": 0.046024887243871104, "alpha_c_m1d": 0.0202045574019174, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.037816061670773005, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01194539011893414, "ret_signed_p1d": -0.01194539011893414, "alpha_spy_p1d": -0.015410779315930379, "alpha_c_p1d": -0.031051843570454984, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.019106453451520844, "ret_p1w": 0.05631396671118849, "ret_signed_p1w": 0.05631396671118849, "alpha_spy_p1w": 0.044821848537607956, "alpha_c_p1w": 0.061161501081300584, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.004847534370112094, "ret_p1m": -0.05734578661241485, "ret_signed_p1m": -0.05734578661241485, "alpha_spy_p1m": -0.07015602716884073, "alpha_c_p1m": -0.0902576465313415, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": 0.032911859918926645, "ret_p3m": 0.5523055581490084, "ret_signed_p3m": 0.5523055581490084, "alpha_spy_p3m": 0.524501175510262, "alpha_c_p3m": 0.40502304113838017, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": 0.14728251701062822, "ret_p6m": 1.2121395194187587, "ret_signed_p6m": 1.2121395194187587, "alpha_spy_p6m": 1.1523212166804095, "alpha_c_p6m": 0.7915293063077491, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.42061021311100966, "price_path": [-0.5535, -0.5629, -0.545, -0.5416, -0.5263, -0.5288, -0.5288, -0.5442, -0.5518, -0.5646, -0.5825, -0.5723, -0.5578, -0.5544, -0.557, -0.5418, -0.541, -0.5631, -0.5555, -0.552, -0.5469, -0.5333, -0.5273, -0.5085, -0.4812, -0.4761, -0.4394, -0.4352, -0.4061, -0.4309, -0.3933, -0.3933, -0.401, -0.384, -0.3899, -0.3916, -0.4258, -0.4044, -0.407, -0.3857, -0.3251, -0.3251, -0.3251, -0.3251, -0.3251, -0.3251, -0.2696, -0.2918, -0.2978, -0.279, -0.2278, -0.2056, -0.1715, -0.1826, -0.1783, -0.1834, -0.1297, -0.087, -0.1109, -0.0478, -0.0307, -0.0461, 0.058, 0.0, -0.0119, 0.0119, -0.0102, 0.0341, 0.0563, 0.0529, 0.0444, -0.0444, 0.0341, -0.0273, -0.041, -0.0256, -0.1109, -0.1126, -0.1143, -0.1058, -0.071, -0.0949, -0.0813, -0.0471, -0.0573, -0.0744, -0.071, -0.0147, -0.0334, 0.0024, -0.0351, -0.0249, -0.0539, -0.0949, -0.0591, -0.0573, -0.0659, -0.0095, -0.0027, 0.0041, 0.0041, 0.0229, 0.0929, 0.1117, 0.1117, 0.1117, 0.1561, 0.1886, 0.2398, 0.2671, 0.291, 0.2705, 0.2791, 0.2603, 0.2671, 0.2791, 0.291, 0.3047, 0.2688, 0.2637, 0.2893, 0.3098, 0.2569, 0.3662, 0.4362, 0.4703, 0.5523, 0.4174, 0.5489, 0.5369, 0.4379, 0.4328, 0.5147, 0.496, 0.4686, 0.5164, 0.5028, 0.5028, 0.5028, 0.5028, 0.5267, 0.6206, 0.624, 0.7162, 0.7384, 0.8802, 0.8152, 0.8152, 0.6065, 0.4525, 0.6099, 0.5808, 0.4303, 0.6048, 0.6339, 0.5911, 0.5569, 0.6664, 0.6595, 0.8067, 0.7331, 0.7228, 0.5962, 0.6869, 0.7023, 0.5962, 0.5962, 0.4936, 0.3807, 0.5346, 0.42, 0.4987, 0.5158, 0.5671, 0.7673, 0.7074, 0.7571, 0.7793, 0.8871, 0.9435, 0.976, 0.9298, 0.9949, 1.0941, 1.0924, 1.0958, 1.0907, 1.2104, 1.2241, 1.2121]}
{"unit_id": "orig:1985550014312300867", "ticker_idx": 3, "ts": "2025-11-04T03:29:47+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung, SK hynix, and Micron are prioritizing capacity allocation to HBM and DDR5", "tweet_type": "analysis", "timeline": "years", "summary": "Panic-level NAND/DRAM/HBM shortage approaching with 70-80% supply gap persisting through 2026+, bullish on memory broadly — NAND, HBM3E, SSDs, HDDs all in simultaneous shortage driven by AI inference surge.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) Gou Jiazhang of Silicon Motion (慧榮): The most severe, panic-level, large-scale shortage in the history of memory is approaching, and the NAND Flash supply shortfall is expected to exceed 80%.\n\nGou Jiazhang of Silicon Motion (慧榮) stated that this time, all types of memory—DRAM, NAND, SSD, and HDD—are, for the first time ever, experiencing a simultaneous, large-scale shortage. The growth of AI inference has already surpassed AI training, and inference drives demand for memory—especially high-capacity SSDs—much more strongly. He asserted that the current shortage is nearing a “panic” level and that the supply shortfall will not be resolved throughout 2026. He said this, too, is a conservative view, and that if this trend continues, smaller, less competitive firms could face an existential crisis.\n\nAI inference is driving a surge in HBM3E demand\n\nSharing takeaways from a recent business trip to Korea, Gou said customers kept coming to discuss only capacity and never price; the mood was that simply getting a capacity allocation would be fortunate. He said HBM demand is very strong, and since AI inference demand emerged, HBM3E demand has expanded. It is not only NVIDIA’s GPUs that require HBM3E; all cloud providers’ in-house ASICs also need HBM3E. At present, Samsung, SK hynix, and Micron are prioritizing capacity allocation to HBM and DDR5. From a capacity-consumption perspective, each unit of HBM requires roughly three times the number of wafers compared with DDR5. As a result, HBM is in short supply now—and so is DDR5.\n\nThe HDD shortage is shifting demand en masse to the NAND industry\n\nIn the China market, starting from the second quarter of this year, SSD demand has rapidly stepped up from the past 16TB/32TB to 64TB/128TB. Looking closely at the cause, because sufficient HDD volumes could not be secured, high-capacity SSDs were purchased as substitutes. Gou also analyzed that, thanks to strong AI inference demand, even if you add up all NAND Flash capacity worldwide, it still cannot meet even one-tenth of data centers’ annual warm-data demand. Therefore, over the next 10 years it will be impossible for SSDs to replace HDDs. He added that timing is the only question.\n\nTracing the current HDD shortage back, during the pandemic the blow from excessive inventories made HDD makers’ business models more conservative, shifting them to expand capacity only to the extent of the orders they received. This has made it difficult to absorb the sudden volume surge from the AI boom all at once. On top of this, the North American hyperscalers’ demand for ultra-high-capacity SSDs has increased to an even more astonishing level. The industry must wait for NAND technology to transition from TLC to QLC; at the earliest, in 2027 the global bit output of QLC may overtake that of TLC NAND.\n\nThe NAND shortage could trigger a reshuffle of the global smartphone market\n\nApple has removed the 128GB model in the iPhone 17 and pushed capacity straight up to 2TB, and Samsung is expected to follow. For 2026, overall smartphone bit growth is forecast at about 32%. In the context of a NAND Flash shortage, major vendors will defend premium smartphones from impact, but low-end models may see capacity cut. A recent example is smartphone storage in Africa and Southeast Asia falling from 128GB to 64GB.\n\nThe NAND Flash shortage will also affect the competitive landscape of the global smartphone market. Apple need not worry about shortages, and Samsung can be supported by its in-house NAND. The focus should be on branded smartphones in the China-based Android camp. Although the local NAND giant YMTC長江存儲 is actively expanding, it is estimated it can meet only 50% of local vendors’ demand. Recently, Xiaomi announced price hikes for new models citing rising memory prices. Going forward, whether smartphone vendors can secure sufficient NAND chips—and what choices they make between specifications and cost/selling price—will determine how market shares shift.\n\nQuestions about an AI bubble\n\nGou stated that concerns about an AI bubble will not materialize in the short term. From cloud AI to edge AI, inference consumes memory in very large quantities. The supply gap in the NAND market is 70%–80%, already at a panic-level shortage. Some overbooking naturally exists, but the key is how the major memory makers allocate and plan capacity. Including customer groups such as cloud, smartphones, autos, and PCs, many Chinese automakers still have not recognized the severity of the shortage. Moreover, all capacity cannot be diverted solely to cloud providers; doing so would lead to fragmentation of the technology industry, which is undesirable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07645741282895635, "ret_signed_m1d": 0.07645741282895635, "alpha_spy_m1d": 0.06446168100013705, "alpha_c_m1d": 0.038641351158183346, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.037816061670773005, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08929970527516695, "ret_signed_p1d": 0.08929970527516695, "alpha_spy_p1d": 0.08583431607817071, "alpha_c_p1d": 0.07019325182364611, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.019106453451520844, "ret_p1w": 0.10585698189947701, "ret_signed_p1w": 0.10585698189947701, "alpha_spy_p1w": 0.09436486372589648, "alpha_c_p1w": 0.1107045162695891, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.004847534370112094, "ret_p1m": 0.073980710015223, "ret_signed_p1m": 0.073980710015223, "alpha_spy_p1m": 0.061170469458797117, "alpha_c_p1m": 0.04106885009629635, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": 0.032911859918926645, "ret_p3m": 0.9036261422706069, "ret_signed_p3m": 0.9036261422706069, "alpha_spy_p3m": 0.8758217596318605, "alpha_c_p3m": 0.7563436252599787, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": 0.14728251701062822, "ret_p6m": 1.379889789526489, "ret_signed_p6m": 1.379889789526489, "alpha_spy_p6m": 1.3200714867881398, "alpha_c_p6m": 0.9592795764154793, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.42061021311100966, "price_path": [-0.4872, -0.455, -0.4329, -0.4144, -0.4304, -0.4257, -0.446, -0.4337, -0.4406, -0.4627, -0.4693, -0.4606, -0.4664, -0.466, -0.4603, -0.4408, -0.4545, -0.4545, -0.4569, -0.4558, -0.4307, -0.3978, -0.3974, -0.3801, -0.3583, -0.3098, -0.2793, -0.2768, -0.272, -0.2667, -0.2259, -0.2541, -0.2454, -0.2372, -0.2588, -0.2811, -0.2791, -0.2487, -0.2331, -0.1651, -0.1578, -0.1385, -0.1242, -0.1483, -0.0986, -0.1179, -0.1671, -0.1159, -0.142, -0.1197, -0.0711, -0.0718, -0.0516, -0.0722, -0.0897, -0.0519, 0.0045, 0.0095, 0.0178, 0.0394, 0.0274, 0.0263, 0.0765, 0.0, 0.0893, 0.0931, 0.0912, 0.1618, 0.1059, 0.1232, 0.0868, 0.1321, 0.1097, 0.048, 0.0362, -0.0764, -0.0489, 0.0271, 0.0298, 0.0561, 0.0561, 0.0846, 0.1029, 0.0984, 0.074, 0.0395, 0.088, 0.1325, 0.1577, 0.2095, 0.1854, 0.106, 0.0893, 0.0664, 0.0344, 0.14, 0.2196, 0.2686, 0.2671, 0.3149, 0.3149, 0.3062, 0.3507, 0.3427, 0.3096, 0.3096, 0.4473, 0.4323, 0.5758, 0.558, 0.5005, 0.5834, 0.587, 0.5515, 0.5295, 0.5446, 0.6644, 0.6644, 0.6748, 0.7854, 0.8242, 0.8337, 0.7853, 0.8823, 0.9972, 0.9996, 0.9036, 1.0088, 0.9245, 0.7408, 0.7568, 0.811, 0.7596, 0.7126, 0.8828, 0.8995, 0.8889, 0.8889, 0.8343, 0.9315, 0.915, 0.9646, 0.9316, 0.918, 0.9684, 0.9067, 0.8921, 0.8935, 0.7421, 0.8389, 0.8218, 0.6991, 0.7863, 0.8496, 0.9211, 0.8599, 0.9552, 1.0271, 1.1184, 1.1186, 1.0385, 0.9404, 0.8553, 0.8148, 0.7532, 0.631, 0.6391, 0.4772, 0.5508, 0.6885, 0.6812, 0.6812, 0.734, 0.7332, 0.867, 0.9349, 0.9306, 0.958, 1.1375, 1.0942, 1.0988, 1.0889, 1.0584, 1.0628, 1.2377, 1.2112, 1.2801, 1.4079, 1.3148, 1.3799]}
{"unit_id": "orig:1985550014312300867", "ticker_idx": 4, "ts": "2025-11-04T03:29:47+00:00", "symbol": "Western Digital", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "HDD shortage is shifting demand en masse to the NAND industry", "tweet_type": "analysis", "timeline": "years", "summary": "Panic-level NAND/DRAM/HBM shortage approaching with 70-80% supply gap persisting through 2026+, bullish on memory broadly — NAND, HBM3E, SSDs, HDDs all in simultaneous shortage driven by AI inference surge.", "resolved_tickers": ["WDC"], "resolution_method": "exchange_ticker", "resolution_reason": "Western Digital listed on Nasdaq as WDC", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) Gou Jiazhang of Silicon Motion (慧榮): The most severe, panic-level, large-scale shortage in the history of memory is approaching, and the NAND Flash supply shortfall is expected to exceed 80%.\n\nGou Jiazhang of Silicon Motion (慧榮) stated that this time, all types of memory—DRAM, NAND, SSD, and HDD—are, for the first time ever, experiencing a simultaneous, large-scale shortage. The growth of AI inference has already surpassed AI training, and inference drives demand for memory—especially high-capacity SSDs—much more strongly. He asserted that the current shortage is nearing a “panic” level and that the supply shortfall will not be resolved throughout 2026. He said this, too, is a conservative view, and that if this trend continues, smaller, less competitive firms could face an existential crisis.\n\nAI inference is driving a surge in HBM3E demand\n\nSharing takeaways from a recent business trip to Korea, Gou said customers kept coming to discuss only capacity and never price; the mood was that simply getting a capacity allocation would be fortunate. He said HBM demand is very strong, and since AI inference demand emerged, HBM3E demand has expanded. It is not only NVIDIA’s GPUs that require HBM3E; all cloud providers’ in-house ASICs also need HBM3E. At present, Samsung, SK hynix, and Micron are prioritizing capacity allocation to HBM and DDR5. From a capacity-consumption perspective, each unit of HBM requires roughly three times the number of wafers compared with DDR5. As a result, HBM is in short supply now—and so is DDR5.\n\nThe HDD shortage is shifting demand en masse to the NAND industry\n\nIn the China market, starting from the second quarter of this year, SSD demand has rapidly stepped up from the past 16TB/32TB to 64TB/128TB. Looking closely at the cause, because sufficient HDD volumes could not be secured, high-capacity SSDs were purchased as substitutes. Gou also analyzed that, thanks to strong AI inference demand, even if you add up all NAND Flash capacity worldwide, it still cannot meet even one-tenth of data centers’ annual warm-data demand. Therefore, over the next 10 years it will be impossible for SSDs to replace HDDs. He added that timing is the only question.\n\nTracing the current HDD shortage back, during the pandemic the blow from excessive inventories made HDD makers’ business models more conservative, shifting them to expand capacity only to the extent of the orders they received. This has made it difficult to absorb the sudden volume surge from the AI boom all at once. On top of this, the North American hyperscalers’ demand for ultra-high-capacity SSDs has increased to an even more astonishing level. The industry must wait for NAND technology to transition from TLC to QLC; at the earliest, in 2027 the global bit output of QLC may overtake that of TLC NAND.\n\nThe NAND shortage could trigger a reshuffle of the global smartphone market\n\nApple has removed the 128GB model in the iPhone 17 and pushed capacity straight up to 2TB, and Samsung is expected to follow. For 2026, overall smartphone bit growth is forecast at about 32%. In the context of a NAND Flash shortage, major vendors will defend premium smartphones from impact, but low-end models may see capacity cut. A recent example is smartphone storage in Africa and Southeast Asia falling from 128GB to 64GB.\n\nThe NAND Flash shortage will also affect the competitive landscape of the global smartphone market. Apple need not worry about shortages, and Samsung can be supported by its in-house NAND. The focus should be on branded smartphones in the China-based Android camp. Although the local NAND giant YMTC長江存儲 is actively expanding, it is estimated it can meet only 50% of local vendors’ demand. Recently, Xiaomi announced price hikes for new models citing rising memory prices. Going forward, whether smartphone vendors can secure sufficient NAND chips—and what choices they make between specifications and cost/selling price—will determine how market shares shift.\n\nQuestions about an AI bubble\n\nGou stated that concerns about an AI bubble will not materialize in the short term. From cloud AI to edge AI, inference consumes memory in very large quantities. The supply gap in the NAND market is 70%–80%, already at a panic-level shortage. Some overbooking naturally exists, but the key is how the major memory makers allocate and plan capacity. Including customer groups such as cloud, smartphones, autos, and PCs, many Chinese automakers still have not recognized the severity of the shortage. Moreover, all capacity cannot be diverted solely to cloud providers; doing so would lead to fragmentation of the technology industry, which is undesirable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.038375730185107315, "ret_signed_m1d": 0.038375730185107315, "alpha_spy_m1d": 0.026379998356288015, "alpha_c_m1d": 0.011261313305086063, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.027114416880021253, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05204373520654815, "ret_signed_p1d": 0.05204373520654815, "alpha_spy_p1d": 0.04857834600955191, "alpha_c_p1d": 0.048131414336894496, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.0039123208696536516, "ret_p1w": 0.11703261928532016, "ret_signed_p1w": 0.11703261928532016, "alpha_spy_p1w": 0.10554050111173963, "alpha_c_p1w": 0.12019649887841133, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.003163879593091168, "ret_p1m": 0.022407737452221133, "ret_signed_p1m": 0.022407737452221133, "alpha_spy_p1m": 0.009597496895795254, "alpha_c_p1m": 0.03584584492959608, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": -0.01343810747737495, "ret_p3m": 0.6456249934742153, "ret_signed_p3m": 0.6456249934742153, "alpha_spy_p3m": 0.6178206108354689, "alpha_c_p3m": 0.6651647348835379, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": -0.019539741409322575, "ret_p6m": 1.7157946478124364, "ret_signed_p6m": 1.7157946478124364, "alpha_spy_p6m": 1.6559763450740872, "alpha_c_p6m": 1.630202595261427, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.08559205255100943, "price_path": [-0.5114, -0.5079, -0.5101, -0.5018, -0.5007, -0.4996, -0.5073, -0.4993, -0.5021, -0.5035, -0.51, -0.4948, -0.48, -0.4768, -0.4705, -0.4615, -0.4727, -0.4727, -0.4624, -0.4355, -0.4054, -0.3952, -0.387, -0.3788, -0.3756, -0.3682, -0.3583, -0.3272, -0.3226, -0.3367, -0.309, -0.2993, -0.2613, -0.2755, -0.2792, -0.2955, -0.2977, -0.2329, -0.2111, -0.1419, -0.1375, -0.1371, -0.1768, -0.2119, -0.2037, -0.2134, -0.2416, -0.219, -0.2566, -0.2086, -0.1726, -0.1707, -0.2014, -0.2022, -0.2084, -0.1739, -0.1495, -0.1676, -0.1791, -0.071, -0.0923, -0.0129, 0.0384, 0.0, 0.052, 0.075, 0.0708, 0.1448, 0.117, 0.0915, 0.0327, 0.0371, 0.0675, 0.0045, 0.0118, -0.0785, -0.0854, -0.0082, 0.0212, 0.0365, 0.0365, 0.0733, 0.0746, 0.0513, 0.0224, 0.0588, 0.1107, 0.1165, 0.115, 0.1966, 0.2311, 0.1597, 0.1314, 0.1481, 0.0934, 0.1509, 0.1909, 0.1625, 0.1723, 0.1809, 0.1809, 0.1939, 0.1817, 0.1578, 0.1329, 0.1329, 0.2344, 0.2356, 0.4427, 0.3145, 0.2343, 0.3183, 0.3951, 0.4074, 0.4139, 0.4606, 0.4567, 0.4567, 0.4664, 0.5908, 0.6, 0.5546, 0.5839, 0.6616, 0.8394, 0.8309, 0.6456, 0.7772, 0.9087, 0.7718, 0.7111, 0.8584, 0.8808, 0.7267, 0.8002, 0.8684, 0.8518, 0.8518, 0.8684, 0.9503, 0.8721, 0.8777, 0.8442, 0.7794, 0.9134, 0.8562, 0.8394, 0.7762, 0.6481, 0.7184, 0.7043, 0.6136, 0.7242, 0.7516, 0.7687, 0.7185, 0.7916, 0.8831, 1.0647, 1.0061, 1.0853, 0.9285, 0.9396, 0.9808, 0.9485, 0.7985, 0.8116, 0.6559, 0.7797, 0.9589, 0.9408, 0.9408, 1.0012, 1.0526, 1.229, 1.2231, 1.2596, 1.3039, 1.4096, 1.4016, 1.3798, 1.451, 1.4615, 1.5253, 1.5601, 1.6524, 1.6582, 1.6366, 1.5726, 1.7158]}
{"unit_id": "orig:1985550014312300867", "ticker_idx": 5, "ts": "2025-11-04T03:29:47+00:00", "symbol": "Seagate", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "HDD makers' business models more conservative, shifting them to expand capacity only to the extent of the orders they received", "tweet_type": "analysis", "timeline": "years", "summary": "Panic-level NAND/DRAM/HBM shortage approaching with 70-80% supply gap persisting through 2026+, bullish on memory broadly — NAND, HBM3E, SSDs, HDDs all in simultaneous shortage driven by AI inference surge.", "resolved_tickers": ["STX"], "resolution_method": "exchange_ticker", "resolution_reason": "Seagate Technology listed on Nasdaq as STX", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "(Must read) Gou Jiazhang of Silicon Motion (慧榮): The most severe, panic-level, large-scale shortage in the history of memory is approaching, and the NAND Flash supply shortfall is expected to exceed 80%.\n\nGou Jiazhang of Silicon Motion (慧榮) stated that this time, all types of memory—DRAM, NAND, SSD, and HDD—are, for the first time ever, experiencing a simultaneous, large-scale shortage. The growth of AI inference has already surpassed AI training, and inference drives demand for memory—especially high-capacity SSDs—much more strongly. He asserted that the current shortage is nearing a “panic” level and that the supply shortfall will not be resolved throughout 2026. He said this, too, is a conservative view, and that if this trend continues, smaller, less competitive firms could face an existential crisis.\n\nAI inference is driving a surge in HBM3E demand\n\nSharing takeaways from a recent business trip to Korea, Gou said customers kept coming to discuss only capacity and never price; the mood was that simply getting a capacity allocation would be fortunate. He said HBM demand is very strong, and since AI inference demand emerged, HBM3E demand has expanded. It is not only NVIDIA’s GPUs that require HBM3E; all cloud providers’ in-house ASICs also need HBM3E. At present, Samsung, SK hynix, and Micron are prioritizing capacity allocation to HBM and DDR5. From a capacity-consumption perspective, each unit of HBM requires roughly three times the number of wafers compared with DDR5. As a result, HBM is in short supply now—and so is DDR5.\n\nThe HDD shortage is shifting demand en masse to the NAND industry\n\nIn the China market, starting from the second quarter of this year, SSD demand has rapidly stepped up from the past 16TB/32TB to 64TB/128TB. Looking closely at the cause, because sufficient HDD volumes could not be secured, high-capacity SSDs were purchased as substitutes. Gou also analyzed that, thanks to strong AI inference demand, even if you add up all NAND Flash capacity worldwide, it still cannot meet even one-tenth of data centers’ annual warm-data demand. Therefore, over the next 10 years it will be impossible for SSDs to replace HDDs. He added that timing is the only question.\n\nTracing the current HDD shortage back, during the pandemic the blow from excessive inventories made HDD makers’ business models more conservative, shifting them to expand capacity only to the extent of the orders they received. This has made it difficult to absorb the sudden volume surge from the AI boom all at once. On top of this, the North American hyperscalers’ demand for ultra-high-capacity SSDs has increased to an even more astonishing level. The industry must wait for NAND technology to transition from TLC to QLC; at the earliest, in 2027 the global bit output of QLC may overtake that of TLC NAND.\n\nThe NAND shortage could trigger a reshuffle of the global smartphone market\n\nApple has removed the 128GB model in the iPhone 17 and pushed capacity straight up to 2TB, and Samsung is expected to follow. For 2026, overall smartphone bit growth is forecast at about 32%. In the context of a NAND Flash shortage, major vendors will defend premium smartphones from impact, but low-end models may see capacity cut. A recent example is smartphone storage in Africa and Southeast Asia falling from 128GB to 64GB.\n\nThe NAND Flash shortage will also affect the competitive landscape of the global smartphone market. Apple need not worry about shortages, and Samsung can be supported by its in-house NAND. The focus should be on branded smartphones in the China-based Android camp. Although the local NAND giant YMTC長江存儲 is actively expanding, it is estimated it can meet only 50% of local vendors’ demand. Recently, Xiaomi announced price hikes for new models citing rising memory prices. Going forward, whether smartphone vendors can secure sufficient NAND chips—and what choices they make between specifications and cost/selling price—will determine how market shares shift.\n\nQuestions about an AI bubble\n\nGou stated that concerns about an AI bubble will not materialize in the short term. From cloud AI to edge AI, inference consumes memory in very large quantities. The supply gap in the NAND market is 70%–80%, already at a panic-level shortage. Some overbooking naturally exists, but the key is how the major memory makers allocate and plan capacity. Including customer groups such as cloud, smartphones, autos, and PCs, many Chinese automakers still have not recognized the severity of the shortage. Moreover, all capacity cannot be diverted solely to cloud providers; doing so would lead to fragmentation of the technology industry, which is undesirable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06058784629295055, "ret_signed_m1d": 0.06058784629295055, "alpha_spy_m1d": 0.04859211446413125, "alpha_c_m1d": 0.0334734294129293, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.027114416880021253, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10140579923400606, "ret_signed_p1d": 0.10140579923400606, "alpha_spy_p1d": 0.09794041003700982, "alpha_c_p1d": 0.0974934783643524, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.0039123208696536516, "ret_p1w": 0.15025167914636461, "ret_signed_p1w": 0.15025167914636461, "alpha_spy_p1w": 0.13875956097278408, "alpha_c_p1w": 0.15341555873945578, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.003163879593091168, "ret_p1m": 0.03310965934256149, "ret_signed_p1m": 0.03310965934256149, "alpha_spy_p1m": 0.020299418786135615, "alpha_c_p1m": 0.04654776681993644, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": -0.01343810747737495, "ret_p3m": 0.632556893148926, "ret_signed_p3m": 0.632556893148926, "alpha_spy_p3m": 0.6047525105101796, "alpha_c_p3m": 0.6520966345582486, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": -0.019539741409322575, "ret_p6m": 1.580528879995358, "ret_signed_p6m": 1.580528879995358, "alpha_spy_p6m": 1.5207105772570089, "alpha_c_p6m": 1.4949368274443486, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.08559205255100943, "price_path": [-0.4104, -0.401, -0.3961, -0.3805, -0.3752, -0.38, -0.3852, -0.3682, -0.3712, -0.3693, -0.3845, -0.3661, -0.3471, -0.3421, -0.3342, -0.3137, -0.3335, -0.3335, -0.3212, -0.298, -0.2675, -0.2509, -0.2466, -0.2372, -0.2314, -0.2164, -0.2197, -0.1595, -0.1594, -0.1505, -0.1375, -0.1192, -0.087, -0.0917, -0.1094, -0.1247, -0.134, -0.0877, -0.0572, 0.0258, 0.0174, 0.0096, -0.0302, -0.1013, -0.104, -0.1145, -0.1438, -0.1233, -0.1548, -0.1238, -0.0973, -0.0998, -0.1437, -0.143, -0.1411, -0.0957, -0.0649, -0.0801, -0.1094, 0.0609, 0.0717, 0.022, 0.0606, 0.0, 0.1014, 0.1122, 0.1157, 0.1742, 0.1503, 0.1313, 0.0486, 0.0313, 0.0439, 0.0139, 0.035, -0.0395, -0.0515, 0.012, 0.046, 0.0875, 0.0875, 0.1051, 0.0788, 0.0659, 0.0331, 0.0609, 0.1135, 0.1399, 0.1297, 0.1939, 0.2295, 0.1488, 0.1406, 0.1508, 0.1089, 0.1662, 0.1836, 0.1297, 0.1295, 0.1423, 0.1423, 0.1461, 0.1264, 0.1216, 0.1028, 0.1028, 0.1514, 0.1606, 0.3231, 0.2344, 0.1391, 0.2174, 0.2873, 0.2752, 0.2505, 0.2827, 0.3064, 0.3064, 0.3054, 0.3784, 0.3876, 0.3859, 0.4347, 0.4887, 0.7737, 0.7882, 0.6326, 0.7337, 0.7798, 0.6764, 0.6236, 0.7192, 0.7019, 0.5867, 0.6308, 0.7266, 0.7058, 0.7058, 0.6656, 0.6984, 0.6377, 0.6463, 0.6314, 0.5858, 0.6893, 0.6405, 0.6332, 0.5198, 0.4321, 0.5017, 0.471, 0.4128, 0.499, 0.5389, 0.5456, 0.4976, 0.5365, 0.5969, 0.6862, 0.6289, 0.7403, 0.6467, 0.6179, 0.7017, 0.6576, 0.5195, 0.5246, 0.4538, 0.5715, 0.6973, 0.7223, 0.7223, 0.8184, 0.8802, 0.9909, 1.0088, 1.0183, 1.059, 1.1398, 1.0843, 1.1333, 1.1972, 1.1651, 1.246, 1.3261, 1.3572, 1.3517, 1.3902, 1.3227, 1.5805]}
{"unit_id": "orig:1975535664025509902", "ticker_idx": 0, "ts": "2025-10-07T12:16:20+00:00", "symbol": "BTC-USD", "kind": "crypto", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "alternative assets like Bitcoin and gold are once again back in focus", "tweet_type": "analysis", "timeline": "months", "summary": "Relays/endorses Goldman MD Molavi's bullish thesis: don't fight the AI capex wave, rate cuts incoming, tech-driven bull intact; gold and BTC outperforming in real terms.", "resolved_tickers": ["BTC-USD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": null, "bench_c_reason": "no_country_fallback (all FX/crypto/index)", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "“Pessimists sound smart, but optimists make money.”\n\n“The market may be noisy, but the real rewards go to the optimists.”\n\nBobby Molavi, Managing Director and veteran trader at Goldman Sachs, recently wrote these words.\nHe emphasized, “There are indeed signs of a bubble in the market, but the main pillars supporting this long bull run—especially the massive capital expenditure boom triggered by AI—are crystal clear.”\n\nMolavi warned, “Until the long-term AI narrative becomes reality, fighting against that enormous flow of capital is futile.”\nIn other words, what investors need now is “to see the signals, ignore the noise, and recognize that the success or failure of AI won’t be determined in the short term.”\n\nIt is precisely these optimists who are crushing the pessimists in today’s “tech-driven super-bull market.”\n\n⸻\n\n■ U.S. Stock Market: Staying Above the 50-Day Moving Average for Over 100 Days\n\nRecently, U.S. equities have shown astonishing resilience.\nBoth the S&P 500 and Nasdaq have remained above their 50-day moving averages for more than 100 trading days, setting fresh all-time highs.\n\nInvestor enthusiasm has reached extreme levels:\n\n    • Retail investors: Net buyers for 21 of the past 24 weeks\n    • ETFs: Inflows on 183 of the last 185 trading days\n    • Call-option volume: Record-breaking 40 million contracts in a single day\n\nParticularly, “heavily shorted stocks” and “unprofitable tech names” have surged, while future-theme sectors—nuclear energy, quantum tech, drones, AI—continue to gain strong momentum.\n\nGoldman Sachs’ High/Low Short Interest Index (GSPUWSHI) recorded the most powerful short-covering rally since the meme-stock frenzy of 2021.\n\n■ The AI Debate Is Still “Ongoing”… But Capital Has Already Moved\n\nThe debate over whether AI is a “revolution” or a “bubble” may persist for several quarters—or even longer.\nHowever, Molavi points out, “What matters more than the debate is where the money is actually going.”\n\nCiting one study, he noted:\n\n“By 2029, cumulative capital expenditures by hyperscale cloud companies are projected to reach $2.8 trillion,\nwhile total global CAPEX is expected to hit $5.5 trillion.”\n\nMolavi likened this flow of capital to “a massive oil tanker—once it starts moving, it takes a long time to stop or change course.”\nThus, short-term corrections of one or two quarters are meaningless within a five-year investment cycle.\n\nHe concluded, “Just as the old saying went ‘Don’t fight the Fed,’ the new mantra is now ‘Don’t fight the CAPEX.’”\n\n⸻\n\n■ Three Tailwinds: Rates, Earnings, and Labor-Market Shifts\n\nBeyond AI, Molavi highlighted three other tailwinds sustaining the market:\n\n1. Interest Rates:\nHe sees a rate-cut scenario as “virtually certain.” Whether driven by cooling inflation, slowing growth, or political pressure, lower rates would inject fresh liquidity into asset markets.\n\n2. Corporate Earnings:\nAI not only boosts margins through productivity and cost savings, but also compels companies to prove efficiency simply by claiming AI adoption—forcing leaner cost structures.\n    \n3. Employment:\nWhile AI-driven efficiency may lift unemployment, this does not signal economic weakness.\nInstead, it represents a re-pricing of the market around productivity, as investors increasingly value output over headcount.\n\n⸻\n\n■ The Market Is Already in “Party Mode”\n\nMarket exuberance has reached historical extremes.\nAccording to Goldman Sachs, average daily call-option volume over the past 20 days has hit 40 million, more than double three years ago.\n\n• On certain days, calls made up 65% of all option trades.\n    \n• Even as the S&P 500 and Nasdaq hit new highs, 57% and 51% of their constituents, respectively, fell.\n\nIn short, a handful of mega-cap tech stocks are driving the market.\nTech and tech-related equities now account for 56% of total U.S. market capitalization, while defensive sectors have fallen to a record-low 16% share.\n\nMolavi captured the moment vividly:\n\n“It’s like a party at two in the morning.\nNobody thinks it’s time to go home.\nBut the hangover—though delayed—always comes.”\n\n⸻\n\n■ “Fake Prosperity” Under a Weak Dollar: What Should Assets Be Measured In?\n\nMolavi offered another provocative angle:\n“This bull market may not be pure price appreciation—it could simply be the reflection of fiat-currency debasement.”\n\nSince the pandemic (in USD terms):\n    • Nasdaq +165%\n    • S&P 500 +102%\n    • In gold terms:\n    • Nasdaq +7%\n    • S&P 500 –18%\n    • In Bitcoin terms:\n    • Nasdaq –78%\n    • S&P 500 –84%\n\nIn other words, the rally measured in dollars may be an illusion born of currency weakness, while real value has appreciated far more in non-dollar assets such as gold and Bitcoin.\n\nMolavi concluded:\n\n“When assessing investment returns,\nthe unit of account matters more than ever.\nThat’s why alternative assets like Bitcoin and gold\nare once again back in focus.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.027180822161542606, "ret_signed_m1d": 0.027180822161542606, "alpha_spy_m1d": 0.023459533260817622, "alpha_c_m1d": null, "bench_spy_m1d": 0.003721288900724984, "bench_c_m1d": null, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": null, "bench_spy_t0": 0.0, "bench_c_t0": null, "ret_p1d": 0.01567280940669602, "ret_signed_p1d": 0.01567280940669602, "alpha_spy_p1d": 0.009709756993903751, "alpha_c_p1d": null, "bench_spy_p1d": 0.005963052412792269, "bench_c_p1d": null, "ret_p1w": -0.06860950083099737, "ret_signed_p1w": -0.06860950083099737, "alpha_spy_p1w": -0.05831236006334861, "alpha_c_p1w": null, "bench_spy_p1w": -0.01029714076764876, "bench_c_p1w": null, "ret_p1m": -0.14458087234880568, "ret_signed_p1m": -0.14458087234880568, "alpha_spy_p1m": -0.1572243561760832, "alpha_c_p1m": null, "bench_spy_p1m": 0.012643483827277535, "bench_c_p1m": null, "ret_p3m": -0.25941810435078283, "ret_signed_p3m": -0.25941810435078283, "alpha_spy_p3m": -0.28343277924645327, "alpha_c_p3m": null, "bench_spy_p3m": 0.02401467489567044, "bench_c_p3m": null, "ret_p6m": -0.43945838152702177, "ret_signed_p6m": -0.43945838152702177, "alpha_spy_p6m": -0.4242905859363707, "alpha_c_p6m": null, "bench_spy_p6m": -0.015167795590651068, "bench_c_p6m": null, "price_path": [-0.045, -0.0324, -0.0132, -0.0303, -0.0223, -0.0178, -0.0284, -0.033, -0.012, -0.0222, -0.0254, -0.0314, -0.029, -0.0291, -0.0298, -0.0469, -0.067, -0.0525, -0.0602, -0.0529, -0.0326, -0.0392, -0.0224, -0.0105, 0.0156, -0.0255, -0.0334, -0.0428, -0.071, -0.0591, -0.0744, -0.0377, -0.0933, -0.0794, -0.0842, -0.0733, -0.1074, -0.1005, -0.0844, -0.0801, -0.0883, -0.0889, -0.0772, -0.0817, -0.0617, -0.0489, -0.044, -0.0495, -0.0379, -0.041, -0.0355, -0.0474, -0.0717, -0.0777, -0.0669, -0.1021, -0.0967, -0.0581, -0.0609, -0.0231, -0.0063, 0.0067, 0.0272, 0.0, 0.0157, 0.0021, -0.0678, -0.0509, -0.0686, -0.0878, -0.1092, -0.1234, -0.0894, -0.1068, -0.1133, -0.0937, -0.0858, -0.0604, -0.0699, -0.0938, -0.1082, -0.0979, -0.1227, -0.1635, -0.1446, -0.1659, -0.1489, -0.1273, -0.1519, -0.1629, -0.1791, -0.2228, -0.2417, -0.2347, -0.2469, -0.2867, -0.2994, -0.2732, -0.2808, -0.2547, -0.2484, -0.2514, -0.2893, -0.2478, -0.2299, -0.2413, -0.264, -0.2537, -0.2368, -0.2423, -0.2383, -0.2567, -0.2884, -0.2767, -0.2907, -0.2963, -0.2746, -0.2714, -0.2803, -0.2786, -0.2817, -0.2812, -0.2825, -0.2719, -0.2795, -0.2694, -0.2594, -0.227, -0.2283, -0.2482, -0.2505, -0.2547, -0.2491, -0.2151, -0.2019, -0.2133, -0.2135, -0.2379, -0.2729, -0.2641, -0.2634, -0.263, -0.2732, -0.2664, -0.2657, -0.3037, -0.3073, -0.3521, -0.3773, -0.3988, -0.4837, -0.4191, -0.4226, -0.4336, -0.4484, -0.4547, -0.433, -0.4332, -0.4443, -0.4531, -0.4487, -0.4401, -0.468, -0.4724, -0.4404, -0.4446, -0.4575, -0.4337, -0.4377, -0.4013, -0.4167, -0.439, -0.4368, -0.4242, -0.422, -0.4196, -0.4157, -0.3836, -0.3913, -0.4134, -0.4244, -0.4193, -0.4161, -0.4194, -0.4129, -0.4336, -0.4538, -0.4509, -0.4382, -0.4395]}
{"unit_id": "orig:1975535664025509902", "ticker_idx": 1, "ts": "2025-10-07T12:16:20+00:00", "symbol": "gold", "kind": "commodity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "alternative assets like Bitcoin and gold are once again back in focus", "tweet_type": "analysis", "timeline": "months", "summary": "Relays/endorses Goldman MD Molavi's bullish thesis: don't fight the AI capex wave, rate cuts incoming, tech-driven bull intact; gold and BTC outperforming in real terms.", "resolved_tickers": ["GLD"], "resolution_method": "etf", "resolution_reason": "Gold commodity proxied by GLD ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "“Pessimists sound smart, but optimists make money.”\n\n“The market may be noisy, but the real rewards go to the optimists.”\n\nBobby Molavi, Managing Director and veteran trader at Goldman Sachs, recently wrote these words.\nHe emphasized, “There are indeed signs of a bubble in the market, but the main pillars supporting this long bull run—especially the massive capital expenditure boom triggered by AI—are crystal clear.”\n\nMolavi warned, “Until the long-term AI narrative becomes reality, fighting against that enormous flow of capital is futile.”\nIn other words, what investors need now is “to see the signals, ignore the noise, and recognize that the success or failure of AI won’t be determined in the short term.”\n\nIt is precisely these optimists who are crushing the pessimists in today’s “tech-driven super-bull market.”\n\n⸻\n\n■ U.S. Stock Market: Staying Above the 50-Day Moving Average for Over 100 Days\n\nRecently, U.S. equities have shown astonishing resilience.\nBoth the S&P 500 and Nasdaq have remained above their 50-day moving averages for more than 100 trading days, setting fresh all-time highs.\n\nInvestor enthusiasm has reached extreme levels:\n\n    • Retail investors: Net buyers for 21 of the past 24 weeks\n    • ETFs: Inflows on 183 of the last 185 trading days\n    • Call-option volume: Record-breaking 40 million contracts in a single day\n\nParticularly, “heavily shorted stocks” and “unprofitable tech names” have surged, while future-theme sectors—nuclear energy, quantum tech, drones, AI—continue to gain strong momentum.\n\nGoldman Sachs’ High/Low Short Interest Index (GSPUWSHI) recorded the most powerful short-covering rally since the meme-stock frenzy of 2021.\n\n■ The AI Debate Is Still “Ongoing”… But Capital Has Already Moved\n\nThe debate over whether AI is a “revolution” or a “bubble” may persist for several quarters—or even longer.\nHowever, Molavi points out, “What matters more than the debate is where the money is actually going.”\n\nCiting one study, he noted:\n\n“By 2029, cumulative capital expenditures by hyperscale cloud companies are projected to reach $2.8 trillion,\nwhile total global CAPEX is expected to hit $5.5 trillion.”\n\nMolavi likened this flow of capital to “a massive oil tanker—once it starts moving, it takes a long time to stop or change course.”\nThus, short-term corrections of one or two quarters are meaningless within a five-year investment cycle.\n\nHe concluded, “Just as the old saying went ‘Don’t fight the Fed,’ the new mantra is now ‘Don’t fight the CAPEX.’”\n\n⸻\n\n■ Three Tailwinds: Rates, Earnings, and Labor-Market Shifts\n\nBeyond AI, Molavi highlighted three other tailwinds sustaining the market:\n\n1. Interest Rates:\nHe sees a rate-cut scenario as “virtually certain.” Whether driven by cooling inflation, slowing growth, or political pressure, lower rates would inject fresh liquidity into asset markets.\n\n2. Corporate Earnings:\nAI not only boosts margins through productivity and cost savings, but also compels companies to prove efficiency simply by claiming AI adoption—forcing leaner cost structures.\n    \n3. Employment:\nWhile AI-driven efficiency may lift unemployment, this does not signal economic weakness.\nInstead, it represents a re-pricing of the market around productivity, as investors increasingly value output over headcount.\n\n⸻\n\n■ The Market Is Already in “Party Mode”\n\nMarket exuberance has reached historical extremes.\nAccording to Goldman Sachs, average daily call-option volume over the past 20 days has hit 40 million, more than double three years ago.\n\n• On certain days, calls made up 65% of all option trades.\n    \n• Even as the S&P 500 and Nasdaq hit new highs, 57% and 51% of their constituents, respectively, fell.\n\nIn short, a handful of mega-cap tech stocks are driving the market.\nTech and tech-related equities now account for 56% of total U.S. market capitalization, while defensive sectors have fallen to a record-low 16% share.\n\nMolavi captured the moment vividly:\n\n“It’s like a party at two in the morning.\nNobody thinks it’s time to go home.\nBut the hangover—though delayed—always comes.”\n\n⸻\n\n■ “Fake Prosperity” Under a Weak Dollar: What Should Assets Be Measured In?\n\nMolavi offered another provocative angle:\n“This bull market may not be pure price appreciation—it could simply be the reflection of fiat-currency debasement.”\n\nSince the pandemic (in USD terms):\n    • Nasdaq +165%\n    • S&P 500 +102%\n    • In gold terms:\n    • Nasdaq +7%\n    • S&P 500 –18%\n    • In Bitcoin terms:\n    • Nasdaq –78%\n    • S&P 500 –84%\n\nIn other words, the rally measured in dollars may be an illusion born of currency weakness, while real value has appreciated far more in non-dollar assets such as gold and Bitcoin.\n\nMolavi concluded:\n\n“When assessing investment returns,\nthe unit of account matters more than ever.\nThat’s why alternative assets like Bitcoin and gold\nare once again back in focus.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00513297884750108, "ret_signed_m1d": -0.00513297884750108, "alpha_spy_m1d": -0.008854267748226063, "alpha_c_m1d": -0.008854267748226063, "bench_spy_m1d": 0.003721288900724984, "bench_c_m1d": 0.003721288900724984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01649095687844504, "ret_signed_p1d": 0.01649095687844504, "alpha_spy_p1d": 0.010527904465652771, "alpha_c_p1d": 0.010527904465652771, "bench_spy_p1d": 0.005963052412792269, "bench_c_p1d": 0.005963052412792269, "ret_p1w": 0.039671267383504016, "ret_signed_p1w": 0.039671267383504016, "alpha_spy_p1w": 0.04996840815115278, "alpha_c_p1w": 0.04996840815115278, "bench_spy_p1w": -0.01029714076764876, "bench_c_p1w": -0.01029714076764876, "ret_p1m": 0.0006825752015897635, "ret_signed_p1m": 0.0006825752015897635, "alpha_spy_p1m": -0.011960908625687772, "alpha_c_p1m": -0.011960908625687772, "bench_spy_p1m": 0.012643483827277535, "bench_c_p1m": 0.012643483827277535, "ret_p3m": 0.0874242018236222, "ret_signed_p3m": 0.0874242018236222, "alpha_spy_p3m": 0.06340952692795176, "alpha_c_p3m": 0.06340952692795176, "bench_spy_p3m": 0.02401467489567044, "bench_c_p3m": 0.02401467489567044, "ret_p6m": 0.19538031903724895, "ret_signed_p6m": 0.19538031903724895, "alpha_spy_p6m": 0.21054811462790002, "alpha_c_p6m": 0.21054811462790002, "bench_spy_p6m": -0.015167795590651068, "bench_c_p6m": -0.015167795590651068, "price_path": [-0.164, -0.156, -0.159, -0.1625, -0.1584, -0.1602, -0.158, -0.1451, -0.137, -0.1477, -0.1529, -0.1607, -0.1661, -0.1638, -0.1783, -0.1728, -0.156, -0.1511, -0.1504, -0.1522, -0.1451, -0.1453, -0.1576, -0.1583, -0.1558, -0.1611, -0.1606, -0.1619, -0.1665, -0.1581, -0.161, -0.152, -0.1541, -0.1479, -0.1462, -0.1399, -0.1316, -0.1316, -0.111, -0.1041, -0.108, -0.0961, -0.0858, -0.0879, -0.0846, -0.086, -0.0842, -0.0747, -0.0728, -0.08, -0.0837, -0.0739, -0.0579, -0.0541, -0.0626, -0.0587, -0.0533, -0.0377, -0.0295, -0.0279, -0.0313, -0.0235, -0.0051, 0.0, 0.0165, -0.0023, 0.0078, 0.0323, 0.0397, 0.0577, 0.0824, 0.0621, 0.1007, 0.03, 0.0301, 0.0342, 0.0307, 0.002, -0.0051, -0.0089, 0.0106, 0.0051, 0.0069, -0.0108, 0.0007, -0.0005, 0.0056, 0.0331, 0.0372, 0.0539, 0.0454, 0.0265, 0.0147, 0.0221, 0.0238, 0.0235, 0.0219, 0.0381, 0.0377, 0.046, 0.046, 0.059, 0.0641, 0.0573, 0.0563, 0.057, 0.0551, 0.0523, 0.0577, 0.0622, 0.0737, 0.0797, 0.0807, 0.0809, 0.0902, 0.0882, 0.0894, 0.1146, 0.1294, 0.1247, 0.1247, 0.1378, 0.0883, 0.0891, 0.082, 0.082, 0.0874, 0.116, 0.1281, 0.1173, 0.1235, 0.1316, 0.1528, 0.1512, 0.1629, 0.1558, 0.1502, 0.1502, 0.1938, 0.2112, 0.2335, 0.2505, 0.2688, 0.2999, 0.3503, 0.354, 0.2148, 0.1662, 0.2403, 0.2395, 0.2065, 0.2435, 0.2751, 0.2625, 0.2768, 0.2324, 0.2631, 0.2631, 0.2237, 0.2512, 0.2547, 0.2795, 0.314, 0.2958, 0.2926, 0.3037, 0.3208, 0.3378, 0.2782, 0.2882, 0.2727, 0.2928, 0.2901, 0.3047, 0.3003, 0.2747, 0.2582, 0.2571, 0.2539, 0.2143, 0.1642, 0.1287, 0.1032, 0.1034, 0.1366, 0.0939, 0.1323, 0.1319, 0.1748, 0.1954]}
{"unit_id": "orig:1975535664025509902", "ticker_idx": 2, "ts": "2025-10-07T12:16:20+00:00", "symbol": "S&P500", "kind": "index", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "optimists make money… the main pillars supporting this long bull run—especially the massive capital expenditure boom triggered by AI—are crystal clear", "tweet_type": "analysis", "timeline": "months", "summary": "Relays/endorses Goldman MD Molavi's bullish thesis: don't fight the AI capex wave, rate cuts incoming, tech-driven bull intact; gold and BTC outperforming in real terms.", "resolved_tickers": ["SPY"], "resolution_method": "etf", "resolution_reason": "S&P 500 best proxied by SPY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "“Pessimists sound smart, but optimists make money.”\n\n“The market may be noisy, but the real rewards go to the optimists.”\n\nBobby Molavi, Managing Director and veteran trader at Goldman Sachs, recently wrote these words.\nHe emphasized, “There are indeed signs of a bubble in the market, but the main pillars supporting this long bull run—especially the massive capital expenditure boom triggered by AI—are crystal clear.”\n\nMolavi warned, “Until the long-term AI narrative becomes reality, fighting against that enormous flow of capital is futile.”\nIn other words, what investors need now is “to see the signals, ignore the noise, and recognize that the success or failure of AI won’t be determined in the short term.”\n\nIt is precisely these optimists who are crushing the pessimists in today’s “tech-driven super-bull market.”\n\n⸻\n\n■ U.S. Stock Market: Staying Above the 50-Day Moving Average for Over 100 Days\n\nRecently, U.S. equities have shown astonishing resilience.\nBoth the S&P 500 and Nasdaq have remained above their 50-day moving averages for more than 100 trading days, setting fresh all-time highs.\n\nInvestor enthusiasm has reached extreme levels:\n\n    • Retail investors: Net buyers for 21 of the past 24 weeks\n    • ETFs: Inflows on 183 of the last 185 trading days\n    • Call-option volume: Record-breaking 40 million contracts in a single day\n\nParticularly, “heavily shorted stocks” and “unprofitable tech names” have surged, while future-theme sectors—nuclear energy, quantum tech, drones, AI—continue to gain strong momentum.\n\nGoldman Sachs’ High/Low Short Interest Index (GSPUWSHI) recorded the most powerful short-covering rally since the meme-stock frenzy of 2021.\n\n■ The AI Debate Is Still “Ongoing”… But Capital Has Already Moved\n\nThe debate over whether AI is a “revolution” or a “bubble” may persist for several quarters—or even longer.\nHowever, Molavi points out, “What matters more than the debate is where the money is actually going.”\n\nCiting one study, he noted:\n\n“By 2029, cumulative capital expenditures by hyperscale cloud companies are projected to reach $2.8 trillion,\nwhile total global CAPEX is expected to hit $5.5 trillion.”\n\nMolavi likened this flow of capital to “a massive oil tanker—once it starts moving, it takes a long time to stop or change course.”\nThus, short-term corrections of one or two quarters are meaningless within a five-year investment cycle.\n\nHe concluded, “Just as the old saying went ‘Don’t fight the Fed,’ the new mantra is now ‘Don’t fight the CAPEX.’”\n\n⸻\n\n■ Three Tailwinds: Rates, Earnings, and Labor-Market Shifts\n\nBeyond AI, Molavi highlighted three other tailwinds sustaining the market:\n\n1. Interest Rates:\nHe sees a rate-cut scenario as “virtually certain.” Whether driven by cooling inflation, slowing growth, or political pressure, lower rates would inject fresh liquidity into asset markets.\n\n2. Corporate Earnings:\nAI not only boosts margins through productivity and cost savings, but also compels companies to prove efficiency simply by claiming AI adoption—forcing leaner cost structures.\n    \n3. Employment:\nWhile AI-driven efficiency may lift unemployment, this does not signal economic weakness.\nInstead, it represents a re-pricing of the market around productivity, as investors increasingly value output over headcount.\n\n⸻\n\n■ The Market Is Already in “Party Mode”\n\nMarket exuberance has reached historical extremes.\nAccording to Goldman Sachs, average daily call-option volume over the past 20 days has hit 40 million, more than double three years ago.\n\n• On certain days, calls made up 65% of all option trades.\n    \n• Even as the S&P 500 and Nasdaq hit new highs, 57% and 51% of their constituents, respectively, fell.\n\nIn short, a handful of mega-cap tech stocks are driving the market.\nTech and tech-related equities now account for 56% of total U.S. market capitalization, while defensive sectors have fallen to a record-low 16% share.\n\nMolavi captured the moment vividly:\n\n“It’s like a party at two in the morning.\nNobody thinks it’s time to go home.\nBut the hangover—though delayed—always comes.”\n\n⸻\n\n■ “Fake Prosperity” Under a Weak Dollar: What Should Assets Be Measured In?\n\nMolavi offered another provocative angle:\n“This bull market may not be pure price appreciation—it could simply be the reflection of fiat-currency debasement.”\n\nSince the pandemic (in USD terms):\n    • Nasdaq +165%\n    • S&P 500 +102%\n    • In gold terms:\n    • Nasdaq +7%\n    • S&P 500 –18%\n    • In Bitcoin terms:\n    • Nasdaq –78%\n    • S&P 500 –84%\n\nIn other words, the rally measured in dollars may be an illusion born of currency weakness, while real value has appreciated far more in non-dollar assets such as gold and Bitcoin.\n\nMolavi concluded:\n\n“When assessing investment returns,\nthe unit of account matters more than ever.\nThat’s why alternative assets like Bitcoin and gold\nare once again back in focus.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003721288900724984, "ret_signed_m1d": 0.003721288900724984, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.003721288900724984, "bench_c_m1d": 0.003721288900724984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005963052412792269, "ret_signed_p1d": 0.005963052412792269, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.005963052412792269, "bench_c_p1d": 0.005963052412792269, "ret_p1w": -0.01029714076764876, "ret_signed_p1w": -0.01029714076764876, "alpha_spy_p1w": 0.0, "alpha_c_p1w": 0.0, "bench_spy_p1w": -0.01029714076764876, "bench_c_p1w": -0.01029714076764876, "ret_p1m": 0.012643483827277535, "ret_signed_p1m": 0.012643483827277535, "alpha_spy_p1m": 0.0, "alpha_c_p1m": 0.0, "bench_spy_p1m": 0.012643483827277535, "bench_c_p1m": 0.012643483827277535, "ret_p3m": 0.02401467489567044, "ret_signed_p3m": 0.02401467489567044, "alpha_spy_p3m": 0.0, "alpha_c_p3m": 0.0, "bench_spy_p3m": 0.02401467489567044, "bench_c_p3m": 0.02401467489567044, "ret_p6m": -0.015167795590651068, "ret_signed_p6m": -0.015167795590651068, "alpha_spy_p6m": 0.0, "alpha_c_p6m": 0.0, "bench_spy_p6m": -0.015167795590651068, "bench_c_p6m": -0.015167795590651068, "price_path": [-0.0673, -0.0706, -0.0688, -0.0728, -0.0697, -0.064, -0.0647, -0.0629, -0.0628, -0.0548, -0.0545, -0.0505, -0.0507, -0.0532, -0.0544, -0.058, -0.0734, -0.0593, -0.0641, -0.0569, -0.0577, -0.0504, -0.0522, -0.0422, -0.0389, -0.0388, -0.041, -0.0412, -0.0464, -0.049, -0.0528, -0.0383, -0.0425, -0.0385, -0.0363, -0.0329, -0.0386, -0.0386, -0.0458, -0.0406, -0.0326, -0.0354, -0.033, -0.0308, -0.028, -0.0199, -0.0202, -0.015, -0.0164, -0.0176, -0.013, -0.0081, -0.0034, -0.0088, -0.012, -0.0165, -0.0109, -0.0081, -0.0044, -0.001, 0.0001, 0.0001, 0.0037, 0.0, 0.006, 0.003, -0.0241, -0.0091, -0.0103, -0.0059, -0.0127, -0.0071, 0.0033, 0.0032, -0.002, 0.0039, 0.0122, 0.0241, 0.0268, 0.0273, 0.016, 0.0193, 0.0213, 0.0091, 0.0126, 0.0018, 0.0028, 0.0184, 0.0207, 0.0213, 0.0044, 0.0042, -0.0052, -0.0135, -0.0097, -0.0248, -0.0151, -0.0006, 0.0088, 0.0158, 0.0158, 0.0213, 0.0167, 0.0185, 0.0221, 0.0228, 0.0248, 0.0217, 0.0208, 0.0276, 0.03, 0.0189, 0.0174, 0.0146, 0.0034, 0.011, 0.0201, 0.0265, 0.0312, 0.0348, 0.0348, 0.0347, 0.031, 0.0298, 0.0221, 0.0221, 0.024, 0.0308, 0.037, 0.0336, 0.0335, 0.0404, 0.042, 0.0399, 0.0348, 0.0376, 0.0367, 0.0367, 0.0156, 0.0274, 0.0327, 0.0331, 0.0383, 0.0425, 0.0424, 0.0403, 0.0372, 0.0424, 0.0335, 0.0285, 0.0157, 0.0352, 0.0402, 0.0374, 0.0372, 0.0212, 0.0219, 0.0219, 0.0235, 0.0287, 0.026, 0.0334, 0.0228, 0.0303, 0.039, 0.0332, 0.0282, 0.0288, 0.0198, 0.027, 0.0212, 0.0078, 0.0167, 0.015, 0.0138, -0.0016, -0.0073, 0.0028, 0.0055, -0.0086, -0.011, -0.0252, -0.015, -0.0183, -0.0128, -0.0304, -0.047, -0.0501, -0.0225, -0.0152]}
{"unit_id": "orig:1975535664025509902", "ticker_idx": 3, "ts": "2025-10-07T12:16:20+00:00", "symbol": "QQQ", "kind": "etf", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "tech and tech-related equities now account for 56% of total U.S. market capitalization… future-theme sectors—nuclear energy, quantum tech, drones, AI—continue to gain strong momentum", "tweet_type": "analysis", "timeline": "months", "summary": "Relays/endorses Goldman MD Molavi's bullish thesis: don't fight the AI capex wave, rate cuts incoming, tech-driven bull intact; gold and BTC outperforming in real terms.", "resolved_tickers": ["QQQ"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "“Pessimists sound smart, but optimists make money.”\n\n“The market may be noisy, but the real rewards go to the optimists.”\n\nBobby Molavi, Managing Director and veteran trader at Goldman Sachs, recently wrote these words.\nHe emphasized, “There are indeed signs of a bubble in the market, but the main pillars supporting this long bull run—especially the massive capital expenditure boom triggered by AI—are crystal clear.”\n\nMolavi warned, “Until the long-term AI narrative becomes reality, fighting against that enormous flow of capital is futile.”\nIn other words, what investors need now is “to see the signals, ignore the noise, and recognize that the success or failure of AI won’t be determined in the short term.”\n\nIt is precisely these optimists who are crushing the pessimists in today’s “tech-driven super-bull market.”\n\n⸻\n\n■ U.S. Stock Market: Staying Above the 50-Day Moving Average for Over 100 Days\n\nRecently, U.S. equities have shown astonishing resilience.\nBoth the S&P 500 and Nasdaq have remained above their 50-day moving averages for more than 100 trading days, setting fresh all-time highs.\n\nInvestor enthusiasm has reached extreme levels:\n\n    • Retail investors: Net buyers for 21 of the past 24 weeks\n    • ETFs: Inflows on 183 of the last 185 trading days\n    • Call-option volume: Record-breaking 40 million contracts in a single day\n\nParticularly, “heavily shorted stocks” and “unprofitable tech names” have surged, while future-theme sectors—nuclear energy, quantum tech, drones, AI—continue to gain strong momentum.\n\nGoldman Sachs’ High/Low Short Interest Index (GSPUWSHI) recorded the most powerful short-covering rally since the meme-stock frenzy of 2021.\n\n■ The AI Debate Is Still “Ongoing”… But Capital Has Already Moved\n\nThe debate over whether AI is a “revolution” or a “bubble” may persist for several quarters—or even longer.\nHowever, Molavi points out, “What matters more than the debate is where the money is actually going.”\n\nCiting one study, he noted:\n\n“By 2029, cumulative capital expenditures by hyperscale cloud companies are projected to reach $2.8 trillion,\nwhile total global CAPEX is expected to hit $5.5 trillion.”\n\nMolavi likened this flow of capital to “a massive oil tanker—once it starts moving, it takes a long time to stop or change course.”\nThus, short-term corrections of one or two quarters are meaningless within a five-year investment cycle.\n\nHe concluded, “Just as the old saying went ‘Don’t fight the Fed,’ the new mantra is now ‘Don’t fight the CAPEX.’”\n\n⸻\n\n■ Three Tailwinds: Rates, Earnings, and Labor-Market Shifts\n\nBeyond AI, Molavi highlighted three other tailwinds sustaining the market:\n\n1. Interest Rates:\nHe sees a rate-cut scenario as “virtually certain.” Whether driven by cooling inflation, slowing growth, or political pressure, lower rates would inject fresh liquidity into asset markets.\n\n2. Corporate Earnings:\nAI not only boosts margins through productivity and cost savings, but also compels companies to prove efficiency simply by claiming AI adoption—forcing leaner cost structures.\n    \n3. Employment:\nWhile AI-driven efficiency may lift unemployment, this does not signal economic weakness.\nInstead, it represents a re-pricing of the market around productivity, as investors increasingly value output over headcount.\n\n⸻\n\n■ The Market Is Already in “Party Mode”\n\nMarket exuberance has reached historical extremes.\nAccording to Goldman Sachs, average daily call-option volume over the past 20 days has hit 40 million, more than double three years ago.\n\n• On certain days, calls made up 65% of all option trades.\n    \n• Even as the S&P 500 and Nasdaq hit new highs, 57% and 51% of their constituents, respectively, fell.\n\nIn short, a handful of mega-cap tech stocks are driving the market.\nTech and tech-related equities now account for 56% of total U.S. market capitalization, while defensive sectors have fallen to a record-low 16% share.\n\nMolavi captured the moment vividly:\n\n“It’s like a party at two in the morning.\nNobody thinks it’s time to go home.\nBut the hangover—though delayed—always comes.”\n\n⸻\n\n■ “Fake Prosperity” Under a Weak Dollar: What Should Assets Be Measured In?\n\nMolavi offered another provocative angle:\n“This bull market may not be pure price appreciation—it could simply be the reflection of fiat-currency debasement.”\n\nSince the pandemic (in USD terms):\n    • Nasdaq +165%\n    • S&P 500 +102%\n    • In gold terms:\n    • Nasdaq +7%\n    • S&P 500 –18%\n    • In Bitcoin terms:\n    • Nasdaq –78%\n    • S&P 500 –84%\n\nIn other words, the rally measured in dollars may be an illusion born of currency weakness, while real value has appreciated far more in non-dollar assets such as gold and Bitcoin.\n\nMolavi concluded:\n\n“When assessing investment returns,\nthe unit of account matters more than ever.\nThat’s why alternative assets like Bitcoin and gold\nare once again back in focus.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0052935032180592145, "ret_signed_m1d": 0.0052935032180592145, "alpha_spy_m1d": 0.0015722143173342307, "alpha_c_m1d": 0.0015722143173342307, "bench_spy_m1d": 0.003721288900724984, "bench_c_m1d": 0.003721288900724984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011463807753162802, "ret_signed_p1d": 0.011463807753162802, "alpha_spy_p1d": 0.005500755340370533, "alpha_c_p1d": 0.005500755340370533, "bench_spy_p1d": 0.005963052412792269, "bench_c_p1d": 0.005963052412792269, "ret_p1w": -0.010769006986574015, "ret_signed_p1w": -0.010769006986574015, "alpha_spy_p1w": -0.00047186621892525427, "alpha_c_p1w": -0.00047186621892525427, "bench_spy_p1w": -0.01029714076764876, "bench_c_p1w": -0.01029714076764876, "ret_p1m": 0.031050022719701698, "ret_signed_p1m": 0.031050022719701698, "alpha_spy_p1m": 0.018406538892424162, "alpha_c_p1m": 0.018406538892424162, "bench_spy_p1m": 0.012643483827277535, "bench_c_p1m": 0.012643483827277535, "ret_p3m": 0.015549709988500027, "ret_signed_p3m": 0.015549709988500027, "alpha_spy_p3m": -0.008464964907170414, "alpha_c_p3m": -0.008464964907170414, "bench_spy_p3m": 0.02401467489567044, "bench_c_p3m": 0.02401467489567044, "ret_p6m": -0.030949811070674405, "ret_signed_p6m": -0.030949811070674405, "alpha_spy_p6m": -0.015782015480023337, "alpha_c_p6m": -0.015782015480023337, "bench_spy_p6m": -0.015167795590651068, "bench_c_p6m": -0.015167795590651068, "price_path": [-0.0822, -0.0843, -0.081, -0.0801, -0.0792, -0.0717, -0.0726, -0.0678, -0.0726, -0.0684, -0.0664, -0.0642, -0.0613, -0.0627, -0.0615, -0.0664, -0.0848, -0.0679, -0.0743, -0.0626, -0.0594, -0.0507, -0.0535, -0.0416, -0.0411, -0.0418, -0.0461, -0.0464, -0.0594, -0.065, -0.0693, -0.0549, -0.0577, -0.0539, -0.0524, -0.0465, -0.0575, -0.0575, -0.0654, -0.0581, -0.0495, -0.0482, -0.0435, -0.0408, -0.0405, -0.0349, -0.0307, -0.0224, -0.0232, -0.0251, -0.0163, -0.0097, -0.0038, -0.0104, -0.0139, -0.0182, -0.0141, -0.0096, -0.0068, -0.0021, 0.002, -0.0022, 0.0053, 0.0, 0.0115, 0.0102, -0.0248, -0.0041, -0.0108, -0.0038, -0.0075, -0.001, 0.0116, 0.0114, 0.0016, 0.01, 0.0208, 0.039, 0.047, 0.0517, 0.0356, 0.0406, 0.0456, 0.0244, 0.0311, 0.0118, 0.0087, 0.031, 0.0282, 0.0274, 0.0064, 0.0072, -0.0014, -0.0136, -0.0077, -0.0312, -0.0239, 0.0011, 0.0072, 0.0161, 0.0161, 0.0244, 0.0209, 0.0289, 0.0314, 0.0305, 0.0347, 0.0327, 0.034, 0.0382, 0.0349, 0.0151, 0.01, 0.012, -0.0068, 0.0076, 0.0207, 0.0256, 0.0304, 0.0335, 0.0335, 0.0334, 0.0284, 0.026, 0.0175, 0.0175, 0.0155, 0.0236, 0.0326, 0.0336, 0.0277, 0.038, 0.0388, 0.0373, 0.0262, 0.0299, 0.029, 0.029, 0.0072, 0.0208, 0.0282, 0.0315, 0.036, 0.0454, 0.0488, 0.0426, 0.03, 0.0371, 0.0212, 0.0033, -0.0111, 0.0098, 0.0175, 0.0128, 0.0155, -0.0051, -0.003, -0.003, -0.004, 0.0034, -0.0004, 0.0084, -0.0038, 0.0069, 0.0214, 0.0091, 0.0059, 0.0072, -0.0036, 0.0116, 0.0086, -0.0066, 0.0067, 0.0067, 0.0066, -0.0107, -0.0166, -0.0056, -0.0007, -0.0146, -0.0177, -0.0359, -0.0248, -0.0315, -0.0251, -0.0484, -0.067, -0.0741, -0.0428, -0.0309]}
{"unit_id": "quote:1987451533211181376", "ticker_idx": 0, "ts": "2025-11-09T09:25:45+00:00", "symbol": "memory semiconductors", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "2026 should be safe. There are no cleanrooms available to expand capacity immediately", "tweet_type": "analysis", "timeline": "years", "summary": "Memory sector safe through 2026 but faces downturn from H1 2027 as new capacity from Micron Idaho and SK Hynix Yongin comes online; chaebol CAPEX behavior makes cycle end inevitable.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory semiconductor basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Thought: The end of the memory supercycle is inevitable.\n\nAs @evrgn11112231 correctly pointed out, for the cycle to be sustained, the majority of excess cash in an upturn must be returned via dividends/buybacks, thereby suppressing excessive reinvestment into WFE/greenfield.\n\nHowever, after thinking it through, I’ve concluded that it will be difficult to prevent the enormous profits in this supercycle from being reinvested into WFE.\n\nLet me explain why.\n\nAmong the Big Three in memory, Samsung and SK hynix are owned by Korean chaebol.\n\nDo you know what characterizes Korean chaebol? Strong leadership.\n\nPut bluntly, it means they can control companies with small equity stakes and make unilateral decisions without being beholden to shareholders.\n\nI believe this helped Korea establish a solid position in the memory semiconductor industry, but conversely, I want to point out that it has trapped the memory industry in the swamp of an eternal cycle.\n\nIn other words, even when they make a lot during an upturn, chaebol-owned semiconductor companies do not pay additional dividends or conduct share buybacks.\n\nInstead, based on the enormous money earned in the upturn, they have executed massive CAPEX—and will continue to do so.\n\nThis inevitably leads to oversupply and brings about a downturn.\n\nEven if Micron tries to manage for profitability on its own, if the two chaebol expand capacity with huge sums, Micron will also have no choice but to pour in huge sums and expand capacity. You can’t walk when everyone else is running.\n\nIn the end, memory falls back into a slump.\n\nThat said, as I’ve said before, 2026 should be safe. There are no cleanrooms available to expand capacity immediately, so fabs have to be built.\n\nHowever, from the first half of 2027, when Micron’s Idaho and SK hynix’s Yongin come online, we may have to worry about a memory downturn.", "reply_text": null, "parent_text": null, "quoted_text": "Also for “this time to be different” you aren’t just making a bet on capacity discipline IMO, but rather also what happens to pricing and industry returns when demand inevitably softens and producers are sitting on excess inventory and underutilized fabs.\n\nWhat does “this time is", "parent_author": null, "ret_m1d": -0.04348723941581648, "ret_signed_m1d": -0.04348723941581648, "alpha_spy_m1d": -0.028122561735264763, "alpha_c_m1d": -0.013748396878358972, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0007181129709183942, "ret_signed_p1d": 0.0007181129709183942, "alpha_spy_p1d": -0.0015710367315273643, "alpha_c_p1d": 0.02229065484291859, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.014936176078392297, "ret_signed_p1w": -0.014936176078392297, "alpha_spy_p1w": 0.008206097822922994, "alpha_c_p1w": 0.036598238984896746, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": 0.0029076674749224374, "ret_signed_p1m": 0.0029076674749224374, "alpha_spy_p1m": 0.0005598770588148838, "alpha_c_p1m": -0.02554904742859332, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.49794942706365525, "ret_signed_p3m": 0.49794942706365525, "alpha_spy_p3m": 0.5006169039510169, "alpha_c_p3m": 0.4325470103123638, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 1.4168680594793785, "ret_signed_p6m": 1.4168680594793785, "alpha_spy_p6m": 1.3487019680802148, "alpha_c_p6m": 0.9555218996935289, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.4467, -0.4472, -0.453, -0.4597, -0.4641, -0.473, -0.4803, -0.4718, -0.4685, -0.4713, -0.4694, -0.4623, -0.4656, -0.4797, -0.473, -0.4692, -0.4594, -0.4475, -0.4435, -0.4279, -0.4092, -0.391, -0.3638, -0.3581, -0.3378, -0.3482, -0.3175, -0.3256, -0.315, -0.3032, -0.309, -0.3136, -0.3333, -0.3135, -0.3108, -0.2775, -0.2434, -0.2379, -0.2338, -0.2407, -0.2264, -0.232, -0.2128, -0.2089, -0.224, -0.2002, -0.1609, -0.1532, -0.1358, -0.1473, -0.1473, -0.145, -0.1039, -0.0781, -0.0917, -0.0618, -0.0478, -0.0419, 0.018, -0.0432, -0.0356, -0.0315, -0.0435, 0.0, 0.0007, 0.0033, -0.0109, -0.0451, -0.0149, -0.0617, -0.0738, -0.0876, -0.1264, -0.0989, -0.09, -0.0681, -0.0546, -0.0641, -0.0534, -0.0351, -0.0418, -0.0551, -0.0292, 0.0054, 0.0029, 0.028, 0.0067, -0.0075, -0.0353, -0.0617, -0.0424, -0.0125, 0.0033, 0.0494, 0.0545, 0.069, 0.069, 0.0922, 0.1377, 0.1428, 0.1333, 0.1333, 0.2158, 0.2539, 0.3142, 0.3249, 0.3088, 0.3267, 0.3298, 0.3095, 0.3144, 0.3346, 0.3895, 0.3952, 0.373, 0.4174, 0.4461, 0.4548, 0.4238, 0.5115, 0.5767, 0.5827, 0.5809, 0.534, 0.6091, 0.5579, 0.4979, 0.5095, 0.5472, 0.5256, 0.5723, 0.6284, 0.6297, 0.6297, 0.614, 0.6419, 0.6742, 0.719, 0.7203, 0.7694, 0.8027, 0.879, 0.8488, 0.8492, 0.6673, 0.5692, 0.6796, 0.6237, 0.5512, 0.6736, 0.7105, 0.6721, 0.6738, 0.747, 0.7883, 0.8844, 0.8111, 0.776, 0.6671, 0.6961, 0.681, 0.5821, 0.5844, 0.4935, 0.4478, 0.6119, 0.5353, 0.5867, 0.6304, 0.658, 0.8077, 0.7861, 0.8076, 0.8059, 0.9105, 0.9313, 0.9648, 0.9421, 0.9493, 0.9975, 1.0422, 1.059, 1.0604, 1.1524, 1.1218, 1.1499, 1.126, 1.159, 1.3329, 1.4169]}
{"unit_id": "quote:1987451533211181376", "ticker_idx": 1, "ts": "2025-11-09T09:25:45+00:00", "symbol": "memory semiconductors", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "from the first half of 2027, when Micron's Idaho and SK hynix's Yongin come online, we may have to worry about a memory downturn", "tweet_type": "analysis", "timeline": "years", "summary": "Memory sector safe through 2026 but faces downturn from H1 2027 as new capacity from Micron Idaho and SK Hynix Yongin comes online; chaebol CAPEX behavior makes cycle end inevitable.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory semiconductor basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Thought: The end of the memory supercycle is inevitable.\n\nAs @evrgn11112231 correctly pointed out, for the cycle to be sustained, the majority of excess cash in an upturn must be returned via dividends/buybacks, thereby suppressing excessive reinvestment into WFE/greenfield.\n\nHowever, after thinking it through, I’ve concluded that it will be difficult to prevent the enormous profits in this supercycle from being reinvested into WFE.\n\nLet me explain why.\n\nAmong the Big Three in memory, Samsung and SK hynix are owned by Korean chaebol.\n\nDo you know what characterizes Korean chaebol? Strong leadership.\n\nPut bluntly, it means they can control companies with small equity stakes and make unilateral decisions without being beholden to shareholders.\n\nI believe this helped Korea establish a solid position in the memory semiconductor industry, but conversely, I want to point out that it has trapped the memory industry in the swamp of an eternal cycle.\n\nIn other words, even when they make a lot during an upturn, chaebol-owned semiconductor companies do not pay additional dividends or conduct share buybacks.\n\nInstead, based on the enormous money earned in the upturn, they have executed massive CAPEX—and will continue to do so.\n\nThis inevitably leads to oversupply and brings about a downturn.\n\nEven if Micron tries to manage for profitability on its own, if the two chaebol expand capacity with huge sums, Micron will also have no choice but to pour in huge sums and expand capacity. You can’t walk when everyone else is running.\n\nIn the end, memory falls back into a slump.\n\nThat said, as I’ve said before, 2026 should be safe. There are no cleanrooms available to expand capacity immediately, so fabs have to be built.\n\nHowever, from the first half of 2027, when Micron’s Idaho and SK hynix’s Yongin come online, we may have to worry about a memory downturn.", "reply_text": null, "parent_text": null, "quoted_text": "Also for “this time to be different” you aren’t just making a bet on capacity discipline IMO, but rather also what happens to pricing and industry returns when demand inevitably softens and producers are sitting on excess inventory and underutilized fabs.\n\nWhat does “this time is", "parent_author": null, "ret_m1d": -0.04348723941581648, "ret_signed_m1d": 0.04348723941581648, "alpha_spy_m1d": 0.028122561735264763, "alpha_c_m1d": 0.013748396878358972, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0007181129709183942, "ret_signed_p1d": -0.0007181129709183942, "alpha_spy_p1d": 0.0015710367315273643, "alpha_c_p1d": -0.02229065484291859, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.014936176078392297, "ret_signed_p1w": 0.014936176078392297, "alpha_spy_p1w": -0.008206097822922994, "alpha_c_p1w": -0.036598238984896746, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": 0.0029076674749224374, "ret_signed_p1m": -0.0029076674749224374, "alpha_spy_p1m": -0.0005598770588148838, "alpha_c_p1m": 0.02554904742859332, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.49794942706365525, "ret_signed_p3m": -0.49794942706365525, "alpha_spy_p3m": -0.5006169039510169, "alpha_c_p3m": -0.4325470103123638, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 1.4168680594793785, "ret_signed_p6m": -1.4168680594793785, "alpha_spy_p6m": -1.3487019680802148, "alpha_c_p6m": -0.9555218996935289, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.4467, -0.4472, -0.453, -0.4597, -0.4641, -0.473, -0.4803, -0.4718, -0.4685, -0.4713, -0.4694, -0.4623, -0.4656, -0.4797, -0.473, -0.4692, -0.4594, -0.4475, -0.4435, -0.4279, -0.4092, -0.391, -0.3638, -0.3581, -0.3378, -0.3482, -0.3175, -0.3256, -0.315, -0.3032, -0.309, -0.3136, -0.3333, -0.3135, -0.3108, -0.2775, -0.2434, -0.2379, -0.2338, -0.2407, -0.2264, -0.232, -0.2128, -0.2089, -0.224, -0.2002, -0.1609, -0.1532, -0.1358, -0.1473, -0.1473, -0.145, -0.1039, -0.0781, -0.0917, -0.0618, -0.0478, -0.0419, 0.018, -0.0432, -0.0356, -0.0315, -0.0435, 0.0, 0.0007, 0.0033, -0.0109, -0.0451, -0.0149, -0.0617, -0.0738, -0.0876, -0.1264, -0.0989, -0.09, -0.0681, -0.0546, -0.0641, -0.0534, -0.0351, -0.0418, -0.0551, -0.0292, 0.0054, 0.0029, 0.028, 0.0067, -0.0075, -0.0353, -0.0617, -0.0424, -0.0125, 0.0033, 0.0494, 0.0545, 0.069, 0.069, 0.0922, 0.1377, 0.1428, 0.1333, 0.1333, 0.2158, 0.2539, 0.3142, 0.3249, 0.3088, 0.3267, 0.3298, 0.3095, 0.3144, 0.3346, 0.3895, 0.3952, 0.373, 0.4174, 0.4461, 0.4548, 0.4238, 0.5115, 0.5767, 0.5827, 0.5809, 0.534, 0.6091, 0.5579, 0.4979, 0.5095, 0.5472, 0.5256, 0.5723, 0.6284, 0.6297, 0.6297, 0.614, 0.6419, 0.6742, 0.719, 0.7203, 0.7694, 0.8027, 0.879, 0.8488, 0.8492, 0.6673, 0.5692, 0.6796, 0.6237, 0.5512, 0.6736, 0.7105, 0.6721, 0.6738, 0.747, 0.7883, 0.8844, 0.8111, 0.776, 0.6671, 0.6961, 0.681, 0.5821, 0.5844, 0.4935, 0.4478, 0.6119, 0.5353, 0.5867, 0.6304, 0.658, 0.8077, 0.7861, 0.8076, 0.8059, 0.9105, 0.9313, 0.9648, 0.9421, 0.9493, 0.9975, 1.0422, 1.059, 1.0604, 1.1524, 1.1218, 1.1499, 1.126, 1.159, 1.3329, 1.4169]}
{"unit_id": "quote:1987451533211181376", "ticker_idx": 2, "ts": "2025-11-09T09:25:45+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "from the first half of 2027, when Micron's Idaho and SK hynix's Yongin come online, we may have to worry about a memory downturn", "tweet_type": "analysis", "timeline": "years", "summary": "Memory sector safe through 2026 but faces downturn from H1 2027 as new capacity from Micron Idaho and SK Hynix Yongin comes online; chaebol CAPEX behavior makes cycle end inevitable.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Thought: The end of the memory supercycle is inevitable.\n\nAs @evrgn11112231 correctly pointed out, for the cycle to be sustained, the majority of excess cash in an upturn must be returned via dividends/buybacks, thereby suppressing excessive reinvestment into WFE/greenfield.\n\nHowever, after thinking it through, I’ve concluded that it will be difficult to prevent the enormous profits in this supercycle from being reinvested into WFE.\n\nLet me explain why.\n\nAmong the Big Three in memory, Samsung and SK hynix are owned by Korean chaebol.\n\nDo you know what characterizes Korean chaebol? Strong leadership.\n\nPut bluntly, it means they can control companies with small equity stakes and make unilateral decisions without being beholden to shareholders.\n\nI believe this helped Korea establish a solid position in the memory semiconductor industry, but conversely, I want to point out that it has trapped the memory industry in the swamp of an eternal cycle.\n\nIn other words, even when they make a lot during an upturn, chaebol-owned semiconductor companies do not pay additional dividends or conduct share buybacks.\n\nInstead, based on the enormous money earned in the upturn, they have executed massive CAPEX—and will continue to do so.\n\nThis inevitably leads to oversupply and brings about a downturn.\n\nEven if Micron tries to manage for profitability on its own, if the two chaebol expand capacity with huge sums, Micron will also have no choice but to pour in huge sums and expand capacity. You can’t walk when everyone else is running.\n\nIn the end, memory falls back into a slump.\n\nThat said, as I’ve said before, 2026 should be safe. There are no cleanrooms available to expand capacity immediately, so fabs have to be built.\n\nHowever, from the first half of 2027, when Micron’s Idaho and SK hynix’s Yongin come online, we may have to worry about a memory downturn.", "reply_text": null, "parent_text": null, "quoted_text": "Also for “this time to be different” you aren’t just making a bet on capacity discipline IMO, but rather also what happens to pricing and industry returns when demand inevitably softens and producers are sitting on excess inventory and underutilized fabs.\n\nWhat does “this time is", "parent_author": null, "ret_m1d": -0.06071854280899214, "ret_signed_m1d": 0.06071854280899214, "alpha_spy_m1d": 0.04535386512844042, "alpha_c_m1d": 0.030979700271534627, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04812476000954269, "ret_signed_p1d": 0.04812476000954269, "alpha_spy_p1d": 0.05041390971198845, "alpha_c_p1d": 0.026552218137542494, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.04480852823517689, "ret_signed_p1w": 0.04480852823517689, "alpha_spy_p1w": 0.0216662543338616, "alpha_c_p1w": -0.006725886828112149, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": -0.0034741303475329666, "ret_signed_p1m": 0.0034741303475329666, "alpha_spy_p1m": 0.00582192076364052, "alpha_c_p1m": 0.03193084525104872, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.5122176055763579, "ret_signed_p3m": -0.5122176055763579, "alpha_spy_p3m": -0.5148850824637196, "alpha_c_p3m": -0.44681518882506643, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 1.5295211549421501, "ret_signed_p6m": -1.5295211549421501, "alpha_spy_p6m": -1.4613550635429864, "alpha_c_p6m": -1.0681749951563004, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.5097, -0.5057, -0.5231, -0.5125, -0.5185, -0.5376, -0.5432, -0.5357, -0.5407, -0.5404, -0.5354, -0.5187, -0.5305, -0.5305, -0.5325, -0.5316, -0.5099, -0.4817, -0.4813, -0.4664, -0.4476, -0.4059, -0.3797, -0.3775, -0.3734, -0.3688, -0.3337, -0.358, -0.3505, -0.3434, -0.362, -0.3812, -0.3795, -0.3533, -0.3399, -0.2813, -0.275, -0.2585, -0.2461, -0.2669, -0.2241, -0.2407, -0.2831, -0.239, -0.2615, -0.2422, -0.2004, -0.201, -0.1837, -0.2014, -0.2165, -0.1839, -0.1353, -0.1311, -0.1239, -0.1053, -0.1156, -0.1166, -0.0734, -0.1392, -0.0624, -0.0591, -0.0607, 0.0, -0.0481, -0.0332, -0.0645, -0.0255, -0.0448, -0.0979, -0.1081, -0.205, -0.1813, -0.1159, -0.1136, -0.091, -0.091, -0.0664, -0.0507, -0.0545, -0.0756, -0.1052, -0.0635, -0.0252, -0.0035, 0.0411, 0.0204, -0.048, -0.0624, -0.0821, -0.1097, -0.0188, 0.0498, 0.0919, 0.0907, 0.1318, 0.1318, 0.1243, 0.1626, 0.1557, 0.1272, 0.1272, 0.2457, 0.2328, 0.3564, 0.341, 0.2916, 0.3629, 0.366, 0.3354, 0.3166, 0.3295, 0.4327, 0.4327, 0.4416, 0.5368, 0.5702, 0.5784, 0.5367, 0.6202, 0.7191, 0.7211, 0.6386, 0.7291, 0.6566, 0.4984, 0.5122, 0.5588, 0.5146, 0.4741, 0.6206, 0.635, 0.6258, 0.6258, 0.5789, 0.6625, 0.6483, 0.6911, 0.6626, 0.6509, 0.6943, 0.6412, 0.6286, 0.6298, 0.4995, 0.5828, 0.5681, 0.4625, 0.5376, 0.5921, 0.6536, 0.6009, 0.683, 0.7449, 0.8234, 0.8236, 0.7546, 0.6702, 0.597, 0.5621, 0.5091, 0.4039, 0.4108, 0.2715, 0.3349, 0.4534, 0.4471, 0.4471, 0.4926, 0.4919, 0.607, 0.6654, 0.6618, 0.6854, 0.8399, 0.8026, 0.8066, 0.798, 0.7718, 0.7756, 0.9261, 0.9033, 0.9626, 1.0726, 0.9925, 1.0485, 1.0434, 1.1423, 1.2776, 1.5295]}
{"unit_id": "quote:1987451533211181376", "ticker_idx": 3, "ts": "2025-11-09T09:25:45+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "from the first half of 2027, when Micron's Idaho and SK hynix's Yongin come online, we may have to worry about a memory downturn", "tweet_type": "analysis", "timeline": "years", "summary": "Memory sector safe through 2026 but faces downturn from H1 2027 as new capacity from Micron Idaho and SK Hynix Yongin comes online; chaebol CAPEX behavior makes cycle end inevitable.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Thought: The end of the memory supercycle is inevitable.\n\nAs @evrgn11112231 correctly pointed out, for the cycle to be sustained, the majority of excess cash in an upturn must be returned via dividends/buybacks, thereby suppressing excessive reinvestment into WFE/greenfield.\n\nHowever, after thinking it through, I’ve concluded that it will be difficult to prevent the enormous profits in this supercycle from being reinvested into WFE.\n\nLet me explain why.\n\nAmong the Big Three in memory, Samsung and SK hynix are owned by Korean chaebol.\n\nDo you know what characterizes Korean chaebol? Strong leadership.\n\nPut bluntly, it means they can control companies with small equity stakes and make unilateral decisions without being beholden to shareholders.\n\nI believe this helped Korea establish a solid position in the memory semiconductor industry, but conversely, I want to point out that it has trapped the memory industry in the swamp of an eternal cycle.\n\nIn other words, even when they make a lot during an upturn, chaebol-owned semiconductor companies do not pay additional dividends or conduct share buybacks.\n\nInstead, based on the enormous money earned in the upturn, they have executed massive CAPEX—and will continue to do so.\n\nThis inevitably leads to oversupply and brings about a downturn.\n\nEven if Micron tries to manage for profitability on its own, if the two chaebol expand capacity with huge sums, Micron will also have no choice but to pour in huge sums and expand capacity. You can’t walk when everyone else is running.\n\nIn the end, memory falls back into a slump.\n\nThat said, as I’ve said before, 2026 should be safe. There are no cleanrooms available to expand capacity immediately, so fabs have to be built.\n\nHowever, from the first half of 2027, when Micron’s Idaho and SK hynix’s Yongin come online, we may have to worry about a memory downturn.", "reply_text": null, "parent_text": null, "quoted_text": "Also for “this time to be different” you aren’t just making a bet on capacity discipline IMO, but rather also what happens to pricing and industry returns when demand inevitably softens and producers are sitting on excess inventory and underutilized fabs.\n\nWhat does “this time is", "parent_author": null, "ret_m1d": -0.04290426551341642, "ret_signed_m1d": 0.04290426551341642, "alpha_spy_m1d": 0.027539587832864698, "alpha_c_m1d": 0.013165422975958907, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02145208104825036, "ret_signed_p1d": -0.02145208104825036, "alpha_spy_p1d": -0.0191629313458046, "alpha_c_p1d": -0.043024622920250555, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": 0.0, "ret_signed_p1w": -0.0, "alpha_spy_p1w": -0.02314227390131529, "alpha_c_p1w": -0.05153441506328904, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": -0.06533766967354793, "ret_signed_p1m": 0.06533766967354793, "alpha_spy_p1m": 0.06768546008965548, "alpha_c_p1m": 0.09379438457706368, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.3904341214598579, "ret_signed_p3m": -0.3904341214598579, "alpha_spy_p3m": -0.3931015983472196, "alpha_c_p3m": -0.32503170470856646, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 1.3939082059047334, "ret_signed_p6m": -1.3939082059047334, "alpha_spy_p6m": -1.3257421145055697, "alpha_c_p6m": -0.9325620461188837, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.5419, -0.5444, -0.5444, -0.5592, -0.5666, -0.579, -0.5963, -0.5864, -0.5724, -0.5691, -0.5716, -0.5569, -0.5561, -0.5776, -0.5701, -0.5668, -0.5619, -0.5487, -0.5429, -0.5248, -0.4983, -0.4934, -0.4579, -0.4538, -0.4257, -0.4497, -0.4134, -0.4134, -0.4208, -0.4043, -0.4101, -0.4117, -0.4447, -0.4241, -0.4266, -0.4059, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.2937, -0.3152, -0.321, -0.3028, -0.2533, -0.2318, -0.1988, -0.2096, -0.2054, -0.2104, -0.1584, -0.1172, -0.1403, -0.0792, -0.0627, -0.0776, 0.0231, -0.033, -0.0446, -0.0215, -0.0429, 0.0, 0.0215, 0.0182, 0.0099, -0.0759, 0.0, -0.0594, -0.0726, -0.0578, -0.1403, -0.1419, -0.1436, -0.1353, -0.1017, -0.1248, -0.1116, -0.0785, -0.0885, -0.105, -0.1017, -0.0472, -0.0653, -0.0307, -0.067, -0.0571, -0.0852, -0.1248, -0.0901, -0.0885, -0.0967, -0.0422, -0.0356, -0.029, -0.029, -0.0108, 0.0569, 0.075, 0.075, 0.075, 0.118, 0.1493, 0.1989, 0.2253, 0.2484, 0.2286, 0.2369, 0.2187, 0.2253, 0.2369, 0.2484, 0.2616, 0.227, 0.222, 0.2468, 0.2666, 0.2154, 0.3211, 0.3888, 0.4218, 0.5011, 0.3706, 0.4978, 0.4862, 0.3904, 0.3855, 0.4647, 0.4466, 0.4202, 0.4664, 0.4532, 0.4532, 0.4532, 0.4532, 0.4763, 0.5671, 0.5704, 0.6596, 0.6811, 0.8182, 0.7553, 0.7553, 0.5535, 0.4046, 0.5568, 0.5287, 0.3831, 0.5518, 0.5799, 0.5386, 0.5055, 0.6114, 0.6048, 0.747, 0.6759, 0.666, 0.5435, 0.6312, 0.6461, 0.5435, 0.5435, 0.4443, 0.3351, 0.484, 0.3731, 0.4492, 0.4658, 0.5154, 0.709, 0.6511, 0.6991, 0.7206, 0.8248, 0.8794, 0.9108, 0.8662, 0.929, 1.025, 1.0233, 1.0266, 1.0217, 1.1375, 1.1507, 1.1391, 1.1276, 1.1276, 1.3939, 1.3939]}
{"unit_id": "orig:1989479922910597530", "ticker_idx": 0, "ts": "2025-11-14T23:45:51+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC is once again accelerating CoWoS capacity expansion... CoWoS output in 2026 is projected to grow 56% YoY", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM raises CoWoS capacity estimates; demand revisions up for NVDA, Broadcom, AMD; TSMC output +56% YoY in 2026 with supply remaining structurally tight, benefiting OSATs.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS capacity significantly revised up; AI packaging to remain tight in 2026–27\nJPM (G. Hariharan, 25/11/13)\n\nJPM notes that TSMC is once again accelerating CoWoS capacity expansion, and now expects capacity to reach 105k wafers per month by end-2026 (vs. the previous estimate of 95k wpm), mainly due to the earlier-than-expected ramp of the AP8 fab and surging demand from NVDA and ASIC customers. Capacity is expected to further rise to 141k wafers/month in 2027, while non-TSMC capacity will begin to contribute after 2H26, accounting for 9%/11% of total capacity in 2026/27, respectively. Driven by strong demand, TSMC’s CoWoS output in 2026 is projected to grow 56% YoY, but overall supply will remain structurally tight.\n\nNvidia: GB300 demand has been revised up meaningfully in recent months, pushing up its 2026 CoWoS usage to 700k wafers. Rubin will start to ramp from 2Q26, with a smooth volume ramp expected in 2H26; its package area will be 50–60% larger than Blackwell, reducing the number of dies that can be cut from a single CoWoS wafer to 9–10. As Rubin CPX does not use HBM, it is more likely to adopt FCBGA rather than the CoWoS-S approach suggested in some market rumors.\n\nBroadcom: Demand has likewise been continuously revised up, with its 2026 requirement now expected at 20k wafers, mainly driven by Google TPU v6p/v7p, Anthropic orders, and the migration of Tomahawk 6 to CoWoS-S. MediaTek’s TPU demand in 2027 is also projected to be as high as 40k wafers.\n\nAMD: Supported by strong demand for MI450 and the Venice CPU, its 2026 usage has been revised up from 80k to 90k wafers, but there is still potential delay risk as production is concentrated in 2H26.\n\nAWS: Trainium 3 demand is stable, with around 80k wafers required in 2026.\n\nJPM emphasizes that TSMC is becoming more stringent in approving new projects, and from 2026 roughly 70% of back-end steps for CoWoS-L will be outsourced to ASE/SPIL. It is also accelerating the outsourcing of probe testing to Ardentec, ASE and Kyec; from 2027 onward, some AI accelerator projects will be further outsourced to OSATs.\n\nSoIC capacity is also being brought forward (22k/30k wafers in 2026/27) to meet AMD and Apple demand, with 2027–28 set to mark the period when 2nm ASICs are widely adopted. To address demand for larger packages, TSMC is actively pushing ahead with its CoPoS process on 310 × 310mm glass panels, with mass production expected in 2028, while CoWoP, given its technical difficulty, is still more likely to exist as a lower-cost, non-leading-edge solution.\n\nWMCM is expected to be first adopted by the iPhone A20 Pro in 2026, and then gradually extended to the full iPhone/iPad/MacBook lineup, with potential future outsourcing to SPIL and Amkor.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02097902398349083, "ret_signed_m1d": 0.02097902398349083, "alpha_spy_m1d": 0.020815411119116556, "alpha_c_m1d": 0.021442852955098513, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010489469027826104, "ret_signed_p1d": 0.010489469027826104, "alpha_spy_p1d": 0.019805992958788265, "alpha_c_p1d": 0.024055601260345782, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.031468493011317045, "ret_signed_p1w": -0.031468493011317045, "alpha_spy_p1w": -0.01227006267166364, "alpha_c_p1w": 0.02317237008149131, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": 0.01736593023071964, "ret_signed_p1m": 0.01736593023071964, "alpha_spy_p1m": 0.004269410151191533, "alpha_c_p1m": -0.005591873489850796, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.3436246527656046, "ret_signed_p3m": 0.3436246527656046, "alpha_spy_p3m": 0.3107719935038742, "alpha_c_p3m": 0.1375529922099683, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.5732631062294933, "ret_signed_p6m": 0.5732631062294933, "alpha_spy_p6m": 0.4667348707234371, "alpha_c_p6m": -0.1024955074835876, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.1746, -0.2095, -0.199, -0.2095, -0.1851, -0.1816, -0.1711, -0.192, -0.192, -0.1886, -0.192, -0.192, -0.192, -0.1781, -0.1781, -0.1642, -0.1468, -0.1363, -0.1224, -0.1259, -0.1049, -0.1154, -0.1014, -0.1154, -0.0944, -0.0629, -0.0629, -0.0769, -0.0909, -0.0909, -0.0874, -0.0734, -0.0455, -0.021, -0.021, 0.0035, -0.0105, 0.007, 0.007, -0.0105, -0.0035, 0.0245, 0.0385, 0.014, 0.035, 0.035, 0.021, 0.014, 0.014, 0.035, 0.0315, 0.0524, 0.0524, 0.049, 0.0559, 0.0524, 0.021, 0.0245, 0.021, 0.0315, 0.0245, 0.0315, 0.021, 0.0, 0.0105, -0.0175, -0.0245, 0.0175, -0.0315, -0.0385, -0.0105, 0.007, 0.0035, 0.007, -0.014, 0.0, 0.014, 0.0105, 0.021, 0.0455, 0.035, 0.0524, 0.0314, 0.0384, 0.0174, 0.0068, 0.0033, 0.0033, 0.0033, 0.0279, 0.0454, 0.0489, 0.0489, 0.0595, 0.0735, 0.0665, 0.0875, 0.0875, 0.1121, 0.1717, 0.1963, 0.1752, 0.1822, 0.1787, 0.1858, 0.1998, 0.1998, 0.1858, 0.2208, 0.2349, 0.2454, 0.2208, 0.2349, 0.2419, 0.2314, 0.2489, 0.277, 0.2664, 0.2454, 0.2384, 0.2629, 0.2524, 0.2384, 0.2489, 0.2735, 0.3191, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3331, 0.3787, 0.4138, 0.3998, 0.3998, 0.3857, 0.3577, 0.3085, 0.3331, 0.3261, 0.27, 0.298, 0.3612, 0.3226, 0.3085, 0.2945, 0.3163, 0.341, 0.3023, 0.2952, 0.2741, 0.2741, 0.2987, 0.2952, 0.2811, 0.253, 0.2389, 0.3058, 0.2741, 0.2741, 0.2741, 0.3093, 0.3726, 0.3762, 0.4078, 0.4008, 0.4466, 0.4642, 0.4677, 0.429, 0.4254, 0.443, 0.443, 0.4642, 0.5381, 0.5944, 0.5592, 0.5345, 0.5029, 0.5029, 0.6014, 0.5838, 0.5838, 0.6261, 0.612, 0.5733]}
{"unit_id": "orig:1989479922910597530", "ticker_idx": 1, "ts": "2025-11-14T23:45:51+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GB300 demand has been revised up meaningfully in recent months, pushing up its 2026 CoWoS usage to 700k wafers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM raises CoWoS capacity estimates; demand revisions up for NVDA, Broadcom, AMD; TSMC output +56% YoY in 2026 with supply remaining structurally tight, benefiting OSATs.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS capacity significantly revised up; AI packaging to remain tight in 2026–27\nJPM (G. Hariharan, 25/11/13)\n\nJPM notes that TSMC is once again accelerating CoWoS capacity expansion, and now expects capacity to reach 105k wafers per month by end-2026 (vs. the previous estimate of 95k wpm), mainly due to the earlier-than-expected ramp of the AP8 fab and surging demand from NVDA and ASIC customers. Capacity is expected to further rise to 141k wafers/month in 2027, while non-TSMC capacity will begin to contribute after 2H26, accounting for 9%/11% of total capacity in 2026/27, respectively. Driven by strong demand, TSMC’s CoWoS output in 2026 is projected to grow 56% YoY, but overall supply will remain structurally tight.\n\nNvidia: GB300 demand has been revised up meaningfully in recent months, pushing up its 2026 CoWoS usage to 700k wafers. Rubin will start to ramp from 2Q26, with a smooth volume ramp expected in 2H26; its package area will be 50–60% larger than Blackwell, reducing the number of dies that can be cut from a single CoWoS wafer to 9–10. As Rubin CPX does not use HBM, it is more likely to adopt FCBGA rather than the CoWoS-S approach suggested in some market rumors.\n\nBroadcom: Demand has likewise been continuously revised up, with its 2026 requirement now expected at 20k wafers, mainly driven by Google TPU v6p/v7p, Anthropic orders, and the migration of Tomahawk 6 to CoWoS-S. MediaTek’s TPU demand in 2027 is also projected to be as high as 40k wafers.\n\nAMD: Supported by strong demand for MI450 and the Venice CPU, its 2026 usage has been revised up from 80k to 90k wafers, but there is still potential delay risk as production is concentrated in 2H26.\n\nAWS: Trainium 3 demand is stable, with around 80k wafers required in 2026.\n\nJPM emphasizes that TSMC is becoming more stringent in approving new projects, and from 2026 roughly 70% of back-end steps for CoWoS-L will be outsourced to ASE/SPIL. It is also accelerating the outsourcing of probe testing to Ardentec, ASE and Kyec; from 2027 onward, some AI accelerator projects will be further outsourced to OSATs.\n\nSoIC capacity is also being brought forward (22k/30k wafers in 2026/27) to meet AMD and Apple demand, with 2027–28 set to mark the period when 2nm ASICs are widely adopted. To address demand for larger packages, TSMC is actively pushing ahead with its CoPoS process on 310 × 310mm glass panels, with mass production expected in 2028, while CoWoP, given its technical difficulty, is still more likely to exist as a lower-cost, non-leading-edge solution.\n\nWMCM is expected to be first adopted by the iPhone A20 Pro in 2026, and then gradually extended to the full iPhone/iPad/MacBook lineup, with potential future outsourcing to SPIL and Amkor.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017405531691603637, "ret_signed_m1d": -0.017405531691603637, "alpha_spy_m1d": -0.01756914455597791, "alpha_c_m1d": -0.016941702719995955, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018772690180637097, "ret_signed_p1d": -0.018772690180637097, "alpha_spy_p1d": -0.009456166249674935, "alpha_c_p1d": -0.0052065579481174185, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.05936791649684514, "ret_signed_p1w": -0.05936791649684514, "alpha_spy_p1w": -0.040169486157191736, "alpha_c_p1w": -0.004727053404036785, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.07293574274855108, "ret_signed_p1m": -0.07293574274855108, "alpha_spy_p1m": -0.08603226282807919, "alpha_c_p1m": -0.09589354646912152, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.0005753669288237617, "ret_signed_p3m": -0.0005753669288237617, "alpha_spy_p3m": -0.03342802619055418, "alpha_c_p3m": -0.20664702748446007, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.15404161563996577, "ret_signed_p6m": 0.15404161563996577, "alpha_spy_p6m": 0.0475133801339096, "alpha_c_p6m": -0.5217169980731151, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.0765, -0.0777, -0.0799, -0.0641, -0.0545, -0.0442, -0.0451, -0.0526, -0.0841, -0.0841, -0.102, -0.1029, -0.0974, -0.1218, -0.115, -0.1021, -0.0676, -0.0684, -0.0649, -0.0653, -0.0804, -0.1045, -0.0733, -0.071, -0.0345, -0.0617, -0.0694, -0.0656, -0.063, -0.0438, -0.0189, -0.0154, -0.0067, -0.0134, -0.0243, -0.027, -0.0056, 0.0126, -0.0369, -0.0097, -0.0533, -0.0544, -0.044, -0.0365, -0.0396, -0.0474, -0.052, -0.0421, -0.0206, 0.0069, 0.0571, 0.0887, 0.0669, 0.0648, 0.0879, 0.0448, 0.0265, -0.011, -0.0106, 0.0467, 0.0157, 0.0191, -0.0174, 0.0, -0.0188, -0.0463, -0.0192, -0.0501, -0.0594, -0.0401, -0.0649, -0.0521, -0.0521, -0.0693, -0.0539, -0.0458, -0.0556, -0.0357, -0.0408, -0.0242, -0.0273, -0.0335, -0.0485, -0.0796, -0.0729, -0.0654, -0.1011, -0.0842, -0.0482, -0.034, -0.005, -0.0081, -0.0081, 0.0019, -0.0102, -0.0138, -0.0192, -0.0192, -0.0069, -0.0107, -0.0154, -0.0055, -0.0269, -0.0279, -0.0274, -0.0229, -0.0369, -0.0164, -0.0207, -0.0207, -0.0636, -0.036, -0.028, -0.0131, -0.0194, -0.0086, 0.0072, 0.0124, 0.0051, -0.0239, -0.0516, -0.084, -0.0961, -0.025, -0.0006, -0.0085, -0.0006, -0.0169, -0.0386, -0.0386, -0.0273, -0.0115, -0.0119, -0.0018, 0.0073, 0.0141, 0.0284, -0.0277, -0.0682, -0.0404, -0.0532, -0.0374, -0.0359, -0.0649, -0.0395, -0.0283, -0.0217, -0.0369, -0.0521, -0.0364, -0.0432, -0.0513, -0.0609, -0.0918, -0.0763, -0.0786, -0.0603, -0.0994, -0.119, -0.1314, -0.0828, -0.0757, -0.0671, -0.0671, -0.0658, -0.0634, -0.0424, -0.0328, -0.008, -0.0044, 0.0335, 0.0459, 0.0431, 0.0606, 0.0626, 0.0512, 0.065, 0.0499, 0.0953, 0.1392, 0.1211, 0.1005, 0.0495, 0.0437, 0.0438, 0.0334, 0.093, 0.1123, 0.1317, 0.154]}
{"unit_id": "orig:1989479922910597530", "ticker_idx": 2, "ts": "2025-11-14T23:45:51+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Demand has likewise been continuously revised up, with its 2026 requirement now expected at 20k wafers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM raises CoWoS capacity estimates; demand revisions up for NVDA, Broadcom, AMD; TSMC output +56% YoY in 2026 with supply remaining structurally tight, benefiting OSATs.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS capacity significantly revised up; AI packaging to remain tight in 2026–27\nJPM (G. Hariharan, 25/11/13)\n\nJPM notes that TSMC is once again accelerating CoWoS capacity expansion, and now expects capacity to reach 105k wafers per month by end-2026 (vs. the previous estimate of 95k wpm), mainly due to the earlier-than-expected ramp of the AP8 fab and surging demand from NVDA and ASIC customers. Capacity is expected to further rise to 141k wafers/month in 2027, while non-TSMC capacity will begin to contribute after 2H26, accounting for 9%/11% of total capacity in 2026/27, respectively. Driven by strong demand, TSMC’s CoWoS output in 2026 is projected to grow 56% YoY, but overall supply will remain structurally tight.\n\nNvidia: GB300 demand has been revised up meaningfully in recent months, pushing up its 2026 CoWoS usage to 700k wafers. Rubin will start to ramp from 2Q26, with a smooth volume ramp expected in 2H26; its package area will be 50–60% larger than Blackwell, reducing the number of dies that can be cut from a single CoWoS wafer to 9–10. As Rubin CPX does not use HBM, it is more likely to adopt FCBGA rather than the CoWoS-S approach suggested in some market rumors.\n\nBroadcom: Demand has likewise been continuously revised up, with its 2026 requirement now expected at 20k wafers, mainly driven by Google TPU v6p/v7p, Anthropic orders, and the migration of Tomahawk 6 to CoWoS-S. MediaTek’s TPU demand in 2027 is also projected to be as high as 40k wafers.\n\nAMD: Supported by strong demand for MI450 and the Venice CPU, its 2026 usage has been revised up from 80k to 90k wafers, but there is still potential delay risk as production is concentrated in 2H26.\n\nAWS: Trainium 3 demand is stable, with around 80k wafers required in 2026.\n\nJPM emphasizes that TSMC is becoming more stringent in approving new projects, and from 2026 roughly 70% of back-end steps for CoWoS-L will be outsourced to ASE/SPIL. It is also accelerating the outsourcing of probe testing to Ardentec, ASE and Kyec; from 2027 onward, some AI accelerator projects will be further outsourced to OSATs.\n\nSoIC capacity is also being brought forward (22k/30k wafers in 2026/27) to meet AMD and Apple demand, with 2027–28 set to mark the period when 2nm ASICs are widely adopted. To address demand for larger packages, TSMC is actively pushing ahead with its CoPoS process on 310 × 310mm glass panels, with mass production expected in 2028, while CoWoP, given its technical difficulty, is still more likely to exist as a lower-cost, non-leading-edge solution.\n\nWMCM is expected to be first adopted by the iPhone A20 Pro in 2026, and then gradually extended to the full iPhone/iPad/MacBook lineup, with potential future outsourcing to SPIL and Amkor.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007241573924393063, "ret_signed_m1d": -0.007241573924393063, "alpha_spy_m1d": -0.007405186788767337, "alpha_c_m1d": -0.006777744952785381, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0005548968528883425, "ret_signed_p1d": 0.0005548968528883425, "alpha_spy_p1d": 0.009871420783850504, "alpha_c_p1d": 0.014121029085408021, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.006599174794477025, "ret_signed_p1w": -0.006599174794477025, "alpha_spy_p1w": 0.01259925554517638, "alpha_c_p1w": 0.04804168829833133, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.007738046455298853, "ret_signed_p1m": -0.007738046455298853, "alpha_spy_p1m": -0.02083456653482696, "alpha_c_p1m": -0.030695850175869288, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.0027912152724218764, "ret_signed_p3m": 0.0027912152724218764, "alpha_spy_p3m": -0.030061443989308545, "alpha_c_p3m": -0.20328044528321443, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.25605975636073564, "ret_signed_p6m": 0.25605975636073564, "alpha_spy_p6m": 0.14953152085467947, "alpha_c_p6m": -0.41969885735234524, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.1403, -0.1512, -0.1558, -0.143, -0.1423, -0.1313, -0.1248, -0.1003, -0.1331, -0.1331, -0.1306, -0.1185, -0.1077, -0.0238, 0.0076, -0.0186, 0.0773, 0.0483, 0.049, 0.0613, 0.0494, 0.0091, 0.0067, 0.0055, -0.0107, -0.0103, -0.0092, -0.0186, -0.0232, -0.0425, -0.0366, -0.0265, -0.0125, -0.0119, -0.0204, -0.0177, 0.0089, 0.0075, -0.0521, 0.0416, 0.0049, 0.0259, 0.0341, 0.0201, 0.0198, 0.0006, -0.0063, 0.0053, 0.0341, 0.0572, 0.0891, 0.1271, 0.0993, 0.0793, 0.0587, 0.0277, 0.0482, 0.0383, 0.0204, 0.0465, 0.0277, 0.0373, -0.0072, 0.0, 0.0006, -0.0057, 0.0349, 0.0127, -0.0066, 0.1037, 0.1243, 0.1609, 0.1609, 0.1767, 0.1274, 0.1142, 0.1114, 0.1126, 0.1395, 0.1712, 0.1864, 0.2059, 0.1866, 0.051, -0.0077, -0.0034, -0.048, -0.0367, -0.0061, -0.001, 0.022, 0.0246, 0.0246, 0.0302, 0.0222, 0.0235, 0.0126, 0.0126, 0.017, 0.0047, 0.0057, 0.005, -0.0273, 0.0093, 0.0304, 0.0375, -0.0056, 0.0036, 0.029, 0.029, -0.0269, -0.0381, -0.0477, -0.0637, -0.0496, -0.0264, -0.0251, -0.0324, -0.0307, -0.0313, -0.0628, -0.0988, -0.0916, -0.026, 0.0062, -0.004, 0.0028, -0.0311, -0.0487, -0.0487, -0.0271, -0.0243, -0.0229, -0.0268, -0.0335, -0.0477, -0.0278, -0.0588, -0.0651, -0.0672, -0.0818, -0.071, -0.0264, -0.0331, 0.0115, 0.0023, -0.0007, -0.0171, -0.0575, -0.0494, -0.06, -0.0757, -0.0643, -0.0916, -0.0545, -0.0668, -0.0653, -0.0929, -0.1185, -0.1398, -0.0926, -0.0809, -0.0778, -0.0778, -0.0782, -0.0209, 0.028, 0.0405, 0.0893, 0.1133, 0.1164, 0.1631, 0.1682, 0.1919, 0.1716, 0.1791, 0.2391, 0.2312, 0.2394, 0.2261, 0.1722, 0.1887, 0.2238, 0.2351, 0.2211, 0.2529, 0.2473, 0.2095, 0.2607, 0.2561]}
{"unit_id": "orig:1989479922910597530", "ticker_idx": 3, "ts": "2025-11-14T23:45:51+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "its 2026 usage has been revised up from 80k to 90k wafers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM raises CoWoS capacity estimates; demand revisions up for NVDA, Broadcom, AMD; TSMC output +56% YoY in 2026 with supply remaining structurally tight, benefiting OSATs.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS capacity significantly revised up; AI packaging to remain tight in 2026–27\nJPM (G. Hariharan, 25/11/13)\n\nJPM notes that TSMC is once again accelerating CoWoS capacity expansion, and now expects capacity to reach 105k wafers per month by end-2026 (vs. the previous estimate of 95k wpm), mainly due to the earlier-than-expected ramp of the AP8 fab and surging demand from NVDA and ASIC customers. Capacity is expected to further rise to 141k wafers/month in 2027, while non-TSMC capacity will begin to contribute after 2H26, accounting for 9%/11% of total capacity in 2026/27, respectively. Driven by strong demand, TSMC’s CoWoS output in 2026 is projected to grow 56% YoY, but overall supply will remain structurally tight.\n\nNvidia: GB300 demand has been revised up meaningfully in recent months, pushing up its 2026 CoWoS usage to 700k wafers. Rubin will start to ramp from 2Q26, with a smooth volume ramp expected in 2H26; its package area will be 50–60% larger than Blackwell, reducing the number of dies that can be cut from a single CoWoS wafer to 9–10. As Rubin CPX does not use HBM, it is more likely to adopt FCBGA rather than the CoWoS-S approach suggested in some market rumors.\n\nBroadcom: Demand has likewise been continuously revised up, with its 2026 requirement now expected at 20k wafers, mainly driven by Google TPU v6p/v7p, Anthropic orders, and the migration of Tomahawk 6 to CoWoS-S. MediaTek’s TPU demand in 2027 is also projected to be as high as 40k wafers.\n\nAMD: Supported by strong demand for MI450 and the Venice CPU, its 2026 usage has been revised up from 80k to 90k wafers, but there is still potential delay risk as production is concentrated in 2H26.\n\nAWS: Trainium 3 demand is stable, with around 80k wafers required in 2026.\n\nJPM emphasizes that TSMC is becoming more stringent in approving new projects, and from 2026 roughly 70% of back-end steps for CoWoS-L will be outsourced to ASE/SPIL. It is also accelerating the outsourcing of probe testing to Ardentec, ASE and Kyec; from 2027 onward, some AI accelerator projects will be further outsourced to OSATs.\n\nSoIC capacity is also being brought forward (22k/30k wafers in 2026/27) to meet AMD and Apple demand, with 2027–28 set to mark the period when 2nm ASICs are widely adopted. To address demand for larger packages, TSMC is actively pushing ahead with its CoPoS process on 310 × 310mm glass panels, with mass production expected in 2028, while CoWoP, given its technical difficulty, is still more likely to exist as a lower-cost, non-leading-edge solution.\n\nWMCM is expected to be first adopted by the iPhone A20 Pro in 2026, and then gradually extended to the full iPhone/iPad/MacBook lineup, with potential future outsourcing to SPIL and Amkor.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0046594917817315284, "ret_signed_m1d": 0.0046594917817315284, "alpha_spy_m1d": 0.004495878917357254, "alpha_c_m1d": 0.005123320753339211, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02548516408716195, "ret_signed_p1d": -0.02548516408716195, "alpha_spy_p1d": -0.016168640156199787, "alpha_c_p1d": -0.01191903185464227, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.1743446343541305, "ret_signed_p1w": -0.1743446343541305, "alpha_spy_p1w": -0.1551462040144771, "alpha_c_p1w": -0.11970377126132214, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.1589481630225522, "ret_signed_p1m": -0.1589481630225522, "alpha_spy_p1m": -0.17204468310208032, "alpha_c_p1m": -0.18190596674312265, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.13463796465396471, "ret_signed_p3m": -0.13463796465396471, "alpha_spy_p3m": -0.16749062391569514, "alpha_c_p3m": -0.340709625209601, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.8588793528754974, "ret_signed_p6m": 0.8588793528754974, "alpha_spy_p6m": 0.7523511173694413, "alpha_c_p6m": 0.18312073916241656, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.3252, -0.3307, -0.3367, -0.3203, -0.3381, -0.3249, -0.3228, -0.317, -0.3411, -0.3411, -0.3423, -0.3431, -0.3445, -0.3876, -0.3865, -0.3687, -0.3536, -0.3693, -0.3575, -0.347, -0.3499, -0.3551, -0.3602, -0.3623, -0.3526, -0.3481, -0.3482, -0.3466, -0.3539, -0.3462, -0.3445, -0.3355, -0.3123, -0.3328, -0.1746, -0.143, -0.0456, -0.0564, -0.1293, -0.1231, -0.1164, -0.0333, -0.0496, -0.0556, -0.0253, -0.0356, -0.0672, -0.0479, 0.0248, 0.0521, 0.0454, 0.071, 0.0325, 0.0377, 0.052, 0.0131, 0.0386, -0.0369, -0.0538, -0.0115, -0.0376, 0.0489, 0.0047, 0.0, -0.0255, -0.0669, -0.0942, -0.1653, -0.1743, -0.1287, -0.1648, -0.132, -0.132, -0.1186, -0.1096, -0.1279, -0.1184, -0.1249, -0.1169, -0.1041, -0.1021, -0.1029, -0.1028, -0.146, -0.1589, -0.1525, -0.1973, -0.1854, -0.1352, -0.1291, -0.1293, -0.1287, -0.1287, -0.1289, -0.1264, -0.1275, -0.1323, -0.1323, -0.0946, -0.1043, -0.1315, -0.1491, -0.1707, -0.1768, -0.1585, -0.1047, -0.094, -0.0765, -0.0607, -0.0607, -0.0603, 0.0121, 0.028, 0.0521, 0.0182, 0.0211, 0.024, 0.0218, -0.0408, -0.0022, -0.019, -0.1889, -0.22, -0.1555, -0.1248, -0.1347, -0.1346, -0.1656, -0.16, -0.16, -0.1772, -0.1892, -0.176, -0.1891, -0.2034, -0.1336, -0.1457, -0.1747, -0.1888, -0.1953, -0.2263, -0.1813, -0.1919, -0.2203, -0.1788, -0.1766, -0.1701, -0.1988, -0.2164, -0.2035, -0.2046, -0.1918, -0.1683, -0.1843, -0.1788, -0.1679, -0.1075, -0.1744, -0.1816, -0.2057, -0.1758, -0.1483, -0.1188, -0.1188, -0.1079, -0.1024, -0.0607, -0.0412, -0.0072, 0.0001, 0.0335, 0.0458, 0.1274, 0.128, 0.114, 0.1527, 0.2295, 0.2371, 0.4092, 0.3558, 0.3095, 0.3659, 0.4363, 0.4608, 0.3838, 0.4394, 0.7073, 0.655, 0.8443, 0.8589]}
{"unit_id": "orig:1989479922910597530", "ticker_idx": 4, "ts": "2025-11-14T23:45:51+00:00", "symbol": "ASE Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "roughly 70% of back-end steps for CoWoS-L will be outsourced to ASE/SPIL", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM raises CoWoS capacity estimates; demand revisions up for NVDA, Broadcom, AMD; TSMC output +56% YoY in 2026 with supply remaining structurally tight, benefiting OSATs.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CoWoS capacity significantly revised up; AI packaging to remain tight in 2026–27\nJPM (G. Hariharan, 25/11/13)\n\nJPM notes that TSMC is once again accelerating CoWoS capacity expansion, and now expects capacity to reach 105k wafers per month by end-2026 (vs. the previous estimate of 95k wpm), mainly due to the earlier-than-expected ramp of the AP8 fab and surging demand from NVDA and ASIC customers. Capacity is expected to further rise to 141k wafers/month in 2027, while non-TSMC capacity will begin to contribute after 2H26, accounting for 9%/11% of total capacity in 2026/27, respectively. Driven by strong demand, TSMC’s CoWoS output in 2026 is projected to grow 56% YoY, but overall supply will remain structurally tight.\n\nNvidia: GB300 demand has been revised up meaningfully in recent months, pushing up its 2026 CoWoS usage to 700k wafers. Rubin will start to ramp from 2Q26, with a smooth volume ramp expected in 2H26; its package area will be 50–60% larger than Blackwell, reducing the number of dies that can be cut from a single CoWoS wafer to 9–10. As Rubin CPX does not use HBM, it is more likely to adopt FCBGA rather than the CoWoS-S approach suggested in some market rumors.\n\nBroadcom: Demand has likewise been continuously revised up, with its 2026 requirement now expected at 20k wafers, mainly driven by Google TPU v6p/v7p, Anthropic orders, and the migration of Tomahawk 6 to CoWoS-S. MediaTek’s TPU demand in 2027 is also projected to be as high as 40k wafers.\n\nAMD: Supported by strong demand for MI450 and the Venice CPU, its 2026 usage has been revised up from 80k to 90k wafers, but there is still potential delay risk as production is concentrated in 2H26.\n\nAWS: Trainium 3 demand is stable, with around 80k wafers required in 2026.\n\nJPM emphasizes that TSMC is becoming more stringent in approving new projects, and from 2026 roughly 70% of back-end steps for CoWoS-L will be outsourced to ASE/SPIL. It is also accelerating the outsourcing of probe testing to Ardentec, ASE and Kyec; from 2027 onward, some AI accelerator projects will be further outsourced to OSATs.\n\nSoIC capacity is also being brought forward (22k/30k wafers in 2026/27) to meet AMD and Apple demand, with 2027–28 set to mark the period when 2nm ASICs are widely adopted. To address demand for larger packages, TSMC is actively pushing ahead with its CoPoS process on 310 × 310mm glass panels, with mass production expected in 2028, while CoWoP, given its technical difficulty, is still more likely to exist as a lower-cost, non-leading-edge solution.\n\nWMCM is expected to be first adopted by the iPhone A20 Pro in 2026, and then gradually extended to the full iPhone/iPad/MacBook lineup, with potential future outsourcing to SPIL and Amkor.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001368956757951123, "ret_signed_m1d": 0.001368956757951123, "alpha_spy_m1d": 0.001205343893576849, "alpha_c_m1d": 0.0018327857295588057, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021218307544634496, "ret_signed_p1d": -0.021218307544634496, "alpha_spy_p1d": -0.011901783613672334, "alpha_c_p1d": -0.007652175312114817, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.05407255170550063, "ret_signed_p1w": -0.05407255170550063, "alpha_spy_p1w": -0.03487412136584722, "alpha_c_p1w": 0.0005683113873077295, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": 0.06639290142525689, "ret_signed_p1m": 0.06639290142525689, "alpha_spy_p1m": 0.05329638134572878, "alpha_c_p1m": 0.043435097704686454, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.6249144646809222, "ret_signed_p3m": 0.6249144646809222, "alpha_spy_p3m": 0.5920618054191917, "alpha_c_p3m": 0.41884280412528585, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 1.4161533242012978, "ret_signed_p6m": 1.4161533242012978, "alpha_spy_p6m": 1.3096250886952416, "alpha_c_p6m": 0.7403947104882169, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.3272, -0.3368, -0.3409, -0.3251, -0.3238, -0.3128, -0.3142, -0.3176, -0.3217, -0.3217, -0.2964, -0.293, -0.2943, -0.2834, -0.2122, -0.2341, -0.2266, -0.2396, -0.2437, -0.2334, -0.2266, -0.2266, -0.2149, -0.2156, -0.1985, -0.1992, -0.2101, -0.2156, -0.2355, -0.2382, -0.2409, -0.2382, -0.2368, -0.2416, -0.2204, -0.2211, -0.1882, -0.1985, -0.2416, -0.1971, -0.2252, -0.1834, -0.1389, -0.1294, -0.1109, -0.1321, -0.1472, -0.1225, -0.1116, -0.0945, -0.0856, -0.0123, 0.0294, 0.0958, 0.1081, 0.052, 0.0835, 0.052, 0.0418, 0.0527, 0.0253, 0.026, 0.0014, 0.0, -0.0212, -0.0301, -0.037, -0.0589, -0.0541, -0.0308, -0.0294, -0.0144, -0.0144, 0.0233, 0.0151, 0.0452, 0.0541, 0.0376, 0.0411, 0.0773, 0.0931, 0.1246, 0.1225, 0.0787, 0.0664, 0.0472, 0.0096, 0.0178, 0.0376, 0.0507, 0.063, 0.0643, 0.0643, 0.076, 0.0965, 0.0972, 0.102, 0.102, 0.154, 0.1608, 0.1978, 0.206, 0.1862, 0.2074, 0.2505, 0.2799, 0.2923, 0.3135, 0.3285, 0.3285, 0.2936, 0.2991, 0.3046, 0.3272, 0.3573, 0.3737, 0.3867, 0.3539, 0.2991, 0.3285, 0.3272, 0.2998, 0.3867, 0.4298, 0.5161, 0.5394, 0.6249, 0.6057, 0.6016, 0.6016, 0.6064, 0.5975, 0.5866, 0.64, 0.6071, 0.6988, 0.7187, 0.6585, 0.6626, 0.6564, 0.527, 0.5127, 0.5099, 0.4456, 0.4825, 0.486, 0.499, 0.4367, 0.4716, 0.473, 0.4942, 0.4825, 0.499, 0.4586, 0.4839, 0.4736, 0.5332, 0.4668, 0.4716, 0.4298, 0.4839, 0.5387, 0.525, 0.525, 0.5435, 0.5188, 0.6523, 0.6776, 0.7002, 0.8172, 0.8392, 0.8638, 0.8891, 0.9569, 1.0041, 1.026, 1.05, 1.0404, 1.1903, 1.1116, 1.0589, 1.0938, 1.1499, 1.1622, 1.2094, 1.2888, 1.3381, 1.282, 1.3429, 1.4162]}
{"unit_id": "orig:2000699857301127328", "ticker_idx": 0, "ts": "2025-12-15T22:49:51+00:00", "symbol": "commodity DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I want to state that it is difficult to foresee a decline in commodity DRAM prices within 2026", "tweet_type": "analysis", "timeline": "years", "summary": "Commodity DRAM prices stay elevated through 2026: HBM supply crunch limits overall DRAM supply even if general-purpose server demand cools.", "resolved_tickers": ["MU", "2408.TW", "2344.TW"], "resolution_method": "basket", "resolution_reason": "Commodity DRAM basket: Micron, Nanya, Winbond", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "The strength of commodity DRAM this year can literally be explained by the replacement cycle of general-purpose servers.\n\nDriven by the mindset of \"Let's replace the general-purpose servers that are due for a refresh while we're aggressively building out AI servers,\" commodity DRAM sales have been strong. (Amazon played a major role in this.)\n\nThis is data from Goldman Sachs. As you can see, there is significant growth in the TAM for general-purpose servers in 2025 compared to 2024.\n\nThen, you might ask this question:\n\n\"Looking at this data, the growth of general-purpose servers appears to be slowing down again. Won't the current surge in commodity DRAM prices come to an end soon?\"\n\nThat is a great question. However, I would like to confidently tell you: No.\n\nThe reason is singular: even if the demand for general-purpose servers remains at current levels, the overall DRAM supply is not increasing.\n\nThe shortage of HBM supply remains persistent, and cleanroom space is scarce.\n\nStructurally, it is difficult for supply to increase significantly until the second quarter of 2027.\n\nEven taking a conservative view, I want to state that it is difficult to foresee a decline in commodity DRAM prices within 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020491907515382397, "ret_signed_m1d": 0.020491907515382397, "alpha_spy_m1d": 0.018978823668354978, "alpha_c_m1d": 0.017034833551626345, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01857842057810229, "ret_signed_p1d": -0.01857842057810229, "alpha_spy_p1d": -0.01584612298282358, "alpha_c_p1d": -0.015858077329077364, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.10013702308452059, "ret_signed_p1w": 0.10013702308452059, "alpha_spy_p1w": 0.09114129777181816, "alpha_c_p1w": 0.07769653341232623, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4144405867282879, "ret_signed_p1m": 0.4144405867282879, "alpha_spy_p1m": 0.39227309682257694, "alpha_c_p1m": 0.3015239615188079, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.5714414764317799, "ret_signed_p3m": 0.5714414764317799, "alpha_spy_p3m": 0.590100649696472, "alpha_c_p3m": 0.4681892696444021, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4874, -0.4476, -0.4522, -0.4428, -0.4295, -0.4523, -0.4663, -0.4866, -0.4773, -0.4559, -0.4297, -0.41, -0.3842, -0.3798, -0.3524, -0.3445, -0.3243, -0.3394, -0.3317, -0.3639, -0.3661, -0.3148, -0.2966, -0.273, -0.2906, -0.2827, -0.2682, -0.2509, -0.2063, -0.1687, -0.1559, -0.1509, -0.16, -0.1242, -0.1744, -0.121, -0.0853, -0.0922, -0.0136, -0.0218, -0.0181, -0.0311, -0.0457, -0.0085, -0.0622, -0.0762, -0.1265, -0.177, -0.1364, -0.1209, -0.1455, -0.1106, -0.0967, -0.0858, -0.0919, -0.1, -0.1012, -0.0635, 0.0003, 0.0209, 0.0066, 0.0034, 0.0205, 0.0, -0.0186, 0.0004, 0.0468, 0.0654, 0.1001, 0.092, 0.1506, 0.1506, 0.146, 0.1605, 0.2023, 0.1839, 0.1839, 0.2932, 0.3127, 0.4385, 0.4704, 0.4632, 0.3892, 0.4541, 0.4144, 0.4211, 0.4249, 0.5214, 0.6196, 0.6002, 0.5558, 0.5948, 0.6009, 0.6627, 0.6857, 0.8158, 0.8306, 0.8547, 0.7616, 0.6528, 0.6465, 0.5823, 0.5542, 0.6055, 0.5652, 0.641, 0.6461, 0.6429, 0.6429, 0.6262, 0.6559, 0.6509, 0.666, 0.723, 0.7702, 0.7463, 0.7434, 0.7389, 0.7477, 0.5867, 0.5313, 0.6108, 0.4972, 0.4509, 0.5345, 0.6434, 0.5714, 0.5975, 0.7053, 0.7842, 0.8146, 0.7291, 0.5895, 0.5292, 0.4659, 0.4615, 0.4117, 0.3873, 0.3426, 0.3068, 0.3883, 0.3584, 0.3584, 0.3745, 0.3743, 0.4793, 0.4415, 0.4725, 0.5066, 0.546, 0.4998, 0.5025, 0.4838, 0.4584, 0.5118, 0.561, 0.5171, 0.536, 0.644, 0.6456, 0.6472, 0.5917, 0.6269, 0.7455, 0.891, 1.0261, 1.0343, 1.1153, 1.2884, 1.3079, 1.3623, 1.421, 1.2672, 1.198, 1.0983, 1.1366, 1.2088, 1.2809, 1.2677, 1.5455, 1.6746, 1.6405, 1.8203, null, null, null, null, null, null, null]}
{"unit_id": "quote:1988121569621930207", "ticker_idx": 0, "ts": "2025-11-11T05:48:14+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi: Maintains Buy on NVIDIA, raises target price from $210 to $220", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares Citi Buy reiteration on NVDA with PT raised to $220, ahead of Nov 19 earnings beat-and-raise call.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "More and more of the sell side is shelving the “AI capex peaks in 2027” thesis and turning bullish on AI again.\n\nIs that a good sign?\n\nA true bubble is when everyone is cheering—unlike now, when a sizable number of people are still pessimistic. https://t.co/gRykazZ1UE", "reply_text": null, "parent_text": null, "quoted_text": "Citi: Maintains Buy on NVIDIA, raises target price from $210 to $220\n\nCiti presented a “30-day short-term bullish” view on NVIDIA, betting that the earnings scheduled to be announced on November 19 will deliver a “beat and raise” (revenue surprise and guidance increase). https://t.co/VDe3WCkZyy", "parent_author": null, "ret_m1d": 0.030492882347475936, "ret_signed_m1d": 0.030492882347475936, "alpha_spy_m1d": 0.03277680381178116, "alpha_c_m1d": 0.00844470525187968, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003313360000960719, "ret_signed_p1d": 0.003313360000960719, "alpha_spy_p1d": 0.002756961394518642, "alpha_c_p1d": -0.00944833118887245, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.061089260286197655, "ret_signed_p1w": -0.061089260286197655, "alpha_spy_p1w": -0.02753153112227391, "alpha_c_p1w": -0.010498217267669863, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": -0.04850783058703967, "ret_signed_p1m": -0.04850783058703967, "alpha_spy_p1m": -0.055198993411596176, "alpha_c_p1m": -0.11416803762039751, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": -0.04006875267429444, "ret_signed_p3m": -0.04006875267429444, "alpha_spy_p3m": -0.05421328419987903, "alpha_c_p3m": -0.1877678549390902, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.07606554636507679, "ret_signed_p6m": 0.07606554636507679, "alpha_spy_p6m": -0.004473938298911406, "alpha_c_p6m": -0.4948521164189732, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.0577, -0.0659, -0.0578, -0.0908, -0.092, -0.0942, -0.0786, -0.0692, -0.059, -0.0599, -0.0673, -0.0983, -0.0983, -0.1159, -0.1167, -0.1114, -0.1354, -0.1287, -0.116, -0.082, -0.0828, -0.0794, -0.0798, -0.0946, -0.1184, -0.0876, -0.0854, -0.0494, -0.0763, -0.0838, -0.0801, -0.0775, -0.0586, -0.0341, -0.0306, -0.0221, -0.0287, -0.0394, -0.042, -0.021, -0.0031, -0.0518, -0.0251, -0.068, -0.069, -0.0588, -0.0515, -0.0545, -0.0621, -0.0667, -0.0569, -0.0357, -0.0086, 0.0407, 0.0719, 0.0504, 0.0483, 0.071, 0.0286, 0.0106, -0.0263, -0.0259, 0.0305, 0.0, 0.0033, -0.0326, -0.0155, -0.034, -0.0611, -0.0344, -0.0648, -0.0739, -0.0549, -0.0794, -0.0668, -0.0668, -0.0837, -0.0685, -0.0606, -0.0703, -0.0506, -0.0556, -0.0393, -0.0423, -0.0485, -0.0633, -0.0939, -0.0873, -0.0799, -0.115, -0.0984, -0.063, -0.049, -0.0204, -0.0235, -0.0235, -0.0136, -0.0255, -0.029, -0.0344, -0.0344, -0.0223, -0.026, -0.0306, -0.0209, -0.042, -0.0429, -0.0425, -0.038, -0.0518, -0.0316, -0.0358, -0.0358, -0.0781, -0.0509, -0.043, -0.0284, -0.0346, -0.024, -0.0084, -0.0033, -0.0105, -0.039, -0.0663, -0.0982, -0.1101, -0.0401, -0.0161, -0.0239, -0.016, -0.0321, -0.0535, -0.0535, -0.0423, -0.0268, -0.0272, -0.0172, -0.0083, -0.0015, 0.0125, -0.0428, -0.0826, -0.0552, -0.0678, -0.0523, -0.0508, -0.0794, -0.0544, -0.0434, -0.0368, -0.0518, -0.0667, -0.0514, -0.058, -0.066, -0.0755, -0.1058, -0.0906, -0.0929, -0.0749, -0.1134, -0.1326, -0.1448, -0.097, -0.09, -0.0815, -0.0815, -0.0802, -0.0779, -0.0573, -0.0478, -0.0233, -0.0198, 0.0175, 0.0297, 0.027, 0.0442, 0.0462, 0.0349, 0.0485, 0.0337, 0.0783, 0.1215, 0.1037, 0.0834, 0.0333, 0.0275, 0.0277, 0.0174, 0.0761]}
{"unit_id": "orig:1988743719106101491", "ticker_idx": 0, "ts": "2025-11-12T23:00:26+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement)", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA reaffirms AI/DRAM/HBM super-cycle; key longs SK Hynix, Samsung, Micron on HBM/DDR5 mix; legacy DRAM names also benefit from price rebound through 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Confirmation of Strong Memory Market: Reaffirming an AI·DRAM·HBM-Driven Super-Boom Cycle (BofA)\n\n1. Industry Overview\n\n• All three experts — Nanya Tech CEO, an AI semiconductor specialist, and a TrendForce analyst — expressed a bullish view on the market outlook based on improving prices and supply–demand dynamics.\n• The global memory market is experiencing structural demand expansion centered on AI servers, which aligns with BofA’s super-cycle thesis.\n• In particular, TrendForce significantly raised its forecast for DRAM and NAND price increases compared to prior estimates, projecting that demand will far exceed supply even in 2026.\n\n2. TrendForce View\n\n• 4Q25 DRAM +30–35% (previous +18–23%), NAND +25–30% (previous +20–25%) — upward revision.\n• 1Q26 DRAM +20–25% (previous +8–13%), NAND +20–25% (previous +18–23%).\n• The 2026 supply-demand sufficiency ratio is expected to remain low — demand far stronger than supply.\n• DRAM makers are intentionally slowing shipments and building inventory strategically to induce higher selling prices later.\n• Rising memory prices may lead to higher end-product prices in 2026, including notebooks and smartphones.\n• Despite the 50–60% price premium of HBM4, the average HBM ASP in 2026 is still expected to decline about 10% YoY (although a shift from HBM to DDR5 could cause price increases).\n\n3. Nanya Tech View\n\n• Price uptrend expected to continue in 4Q; October revenue approached the peaks of 2018/2021.\n• DDR5-to-DDR4 production conversion is relatively easy, enabling flexible response to customer demand.\n• New fab expansions planned for 2026 will realistically only begin in 2H27, meaning bit growth in 2026 will be virtually flat.\n• Pricing negotiations are shifting from quarterly to monthly, reflecting a supplier-dominant market.\n• Channel inventory remains low, leaving room for further price increases.\n\n4. AI Semiconductor Expert Robert Castellano’s View\n\n• OpenAI’s massive AI infrastructure investment is driving structural demand for DRAM/HBM.\n• OpenAI does not own its own datacenters; CapEx expansion is carried out through partners (cloud providers, semiconductor companies, building leases) — approximately USD 1.4 trillion commitment over 8 years.\n• OpenAI’s annualized revenue is growing to around USD 28 billion.\n• U.S. big tech’s AI server chips (GPU+HBM) represent a shift toward “high-value, high-density memory,” unlike the dot-com/cloud bubbles of the past.\n• AI data center and sovereign AI infrastructure investment is expected to deepen HBM supply shortages.\n• Memory makers, unlike in past cycles, are pursuing “profit-centric” strategies and refraining from aggressive capacity expansions; simultaneous bottlenecks in HBM and DDR5 are tightening overall DRAM supply.\n\n5. Investment Implications\n\n• AI-driven demand for DRAM/HBM is expected to maintain structural growth for the next 1–2 years.\n• Supply increase remains limited, and memory makers are shifting toward price-focused profit optimization.\n• During the HBM4 transition period (2026–27), DDR5 production capacity is likely to shrink, supporting continued strength in commodity DRAM pricing.\n• Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement).\n• Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound.\n• The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0032414031765386486, "ret_signed_m1d": 0.0032414031765386486, "alpha_spy_m1d": 0.003797492375725242, "alpha_c_m1d": 0.01584228579369229, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008103711087697363, "ret_signed_p1d": -0.008103711087697363, "alpha_spy_p1d": 0.008490310383040489, "alpha_c_p1d": 0.022020272573561428, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.08914102511102207, "ret_signed_p1w": -0.08914102511102207, "alpha_spy_p1w": -0.058777351218040175, "alpha_c_p1w": -0.04388470269168021, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": -0.08362296077883868, "ret_signed_p1m": -0.08362296077883868, "alpha_spy_p1m": -0.09209552899115969, "alpha_c_p1m": -0.12682613450156632, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.4386309601742753, "ret_signed_p3m": 0.4386309601742753, "alpha_spy_p3m": 0.42016306810534143, "alpha_c_p3m": 0.29128660834802345, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6875831566891168, "ret_signed_p6m": 1.6875831566891168, "alpha_spy_p6m": 1.6109557536984331, "alpha_c_p6m": 1.1637157874256956, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.5525, -0.5671, -0.5744, -0.5865, -0.6035, -0.5938, -0.58, -0.5768, -0.5792, -0.5648, -0.564, -0.5851, -0.5778, -0.5746, -0.5697, -0.5567, -0.5511, -0.5332, -0.5073, -0.5024, -0.4676, -0.4635, -0.436, -0.4595, -0.4238, -0.4238, -0.4311, -0.4149, -0.4206, -0.4222, -0.4546, -0.4344, -0.4368, -0.4165, -0.359, -0.359, -0.359, -0.359, -0.359, -0.359, -0.3063, -0.3274, -0.3331, -0.3152, -0.2666, -0.2455, -0.2131, -0.2237, -0.2196, -0.2245, -0.1734, -0.1329, -0.1556, -0.0956, -0.0794, -0.094, 0.0049, -0.0502, -0.0616, -0.0389, -0.06, -0.0178, 0.0032, 0.0, -0.0081, -0.0924, -0.0178, -0.0762, -0.0891, -0.0746, -0.1556, -0.1572, -0.1588, -0.1507, -0.1177, -0.1404, -0.1274, -0.095, -0.1047, -0.1209, -0.1177, -0.0642, -0.082, -0.0479, -0.0836, -0.0739, -0.1015, -0.1404, -0.1063, -0.1047, -0.1128, -0.0593, -0.0528, -0.0463, -0.0463, -0.0285, 0.038, 0.0559, 0.0559, 0.0559, 0.098, 0.1288, 0.1775, 0.2035, 0.2262, 0.2067, 0.2148, 0.197, 0.2035, 0.2148, 0.2262, 0.2391, 0.2051, 0.2002, 0.2245, 0.244, 0.1937, 0.2975, 0.364, 0.3965, 0.4743, 0.3462, 0.4711, 0.4597, 0.3656, 0.3608, 0.4386, 0.4208, 0.3948, 0.4403, 0.4273, 0.4273, 0.4273, 0.4273, 0.45, 0.5392, 0.5424, 0.63, 0.6511, 0.7858, 0.724, 0.724, 0.5258, 0.3795, 0.529, 0.5014, 0.3584, 0.5242, 0.5518, 0.5112, 0.4787, 0.5827, 0.5762, 0.7159, 0.646, 0.6363, 0.516, 0.6022, 0.6168, 0.516, 0.516, 0.4185, 0.3113, 0.4575, 0.3487, 0.4234, 0.4397, 0.4884, 0.6785, 0.6216, 0.6688, 0.6899, 0.7923, 0.8459, 0.8768, 0.8329, 0.8946, 0.9889, 0.9873, 0.9905, 0.9856, 1.0994, 1.1124, 1.101, 1.0896, 1.0896, 1.3512, 1.3512, 1.6015, 1.6876]}
{"unit_id": "orig:1988743719106101491", "ticker_idx": 1, "ts": "2025-11-12T23:00:26+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement)", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA reaffirms AI/DRAM/HBM super-cycle; key longs SK Hynix, Samsung, Micron on HBM/DDR5 mix; legacy DRAM names also benefit from price rebound through 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Confirmation of Strong Memory Market: Reaffirming an AI·DRAM·HBM-Driven Super-Boom Cycle (BofA)\n\n1. Industry Overview\n\n• All three experts — Nanya Tech CEO, an AI semiconductor specialist, and a TrendForce analyst — expressed a bullish view on the market outlook based on improving prices and supply–demand dynamics.\n• The global memory market is experiencing structural demand expansion centered on AI servers, which aligns with BofA’s super-cycle thesis.\n• In particular, TrendForce significantly raised its forecast for DRAM and NAND price increases compared to prior estimates, projecting that demand will far exceed supply even in 2026.\n\n2. TrendForce View\n\n• 4Q25 DRAM +30–35% (previous +18–23%), NAND +25–30% (previous +20–25%) — upward revision.\n• 1Q26 DRAM +20–25% (previous +8–13%), NAND +20–25% (previous +18–23%).\n• The 2026 supply-demand sufficiency ratio is expected to remain low — demand far stronger than supply.\n• DRAM makers are intentionally slowing shipments and building inventory strategically to induce higher selling prices later.\n• Rising memory prices may lead to higher end-product prices in 2026, including notebooks and smartphones.\n• Despite the 50–60% price premium of HBM4, the average HBM ASP in 2026 is still expected to decline about 10% YoY (although a shift from HBM to DDR5 could cause price increases).\n\n3. Nanya Tech View\n\n• Price uptrend expected to continue in 4Q; October revenue approached the peaks of 2018/2021.\n• DDR5-to-DDR4 production conversion is relatively easy, enabling flexible response to customer demand.\n• New fab expansions planned for 2026 will realistically only begin in 2H27, meaning bit growth in 2026 will be virtually flat.\n• Pricing negotiations are shifting from quarterly to monthly, reflecting a supplier-dominant market.\n• Channel inventory remains low, leaving room for further price increases.\n\n4. AI Semiconductor Expert Robert Castellano’s View\n\n• OpenAI’s massive AI infrastructure investment is driving structural demand for DRAM/HBM.\n• OpenAI does not own its own datacenters; CapEx expansion is carried out through partners (cloud providers, semiconductor companies, building leases) — approximately USD 1.4 trillion commitment over 8 years.\n• OpenAI’s annualized revenue is growing to around USD 28 billion.\n• U.S. big tech’s AI server chips (GPU+HBM) represent a shift toward “high-value, high-density memory,” unlike the dot-com/cloud bubbles of the past.\n• AI data center and sovereign AI infrastructure investment is expected to deepen HBM supply shortages.\n• Memory makers, unlike in past cycles, are pursuing “profit-centric” strategies and refraining from aggressive capacity expansions; simultaneous bottlenecks in HBM and DDR5 are tightening overall DRAM supply.\n\n5. Investment Implications\n\n• AI-driven demand for DRAM/HBM is expected to maintain structural growth for the next 1–2 years.\n• Supply increase remains limited, and memory makers are shifting toward price-focused profit optimization.\n• During the HBM4 transition period (2026–27), DDR5 production capacity is likely to shrink, supporting continued strength in commodity DRAM pricing.\n• Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement).\n• Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound.\n• The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0038797231644660535, "ret_signed_m1d": 0.0038797231644660535, "alpha_spy_m1d": 0.004435812363652647, "alpha_c_m1d": 0.0069758299120047385, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.003096106747538685, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0029097732978722313, "ret_signed_p1d": -0.0029097732978722313, "alpha_spy_p1d": 0.01368424817286562, "alpha_c_p1d": 0.022029102618702723, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.024938875916574954, "ret_p1w": -0.06401547036464372, "ret_signed_p1w": -0.06401547036464372, "alpha_spy_p1w": -0.03365179647166183, "alpha_c_p1w": -0.019955842120262357, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.044059628244381366, "ret_p1m": 0.040737131377847735, "ret_signed_p1m": 0.040737131377847735, "alpha_spy_p1m": 0.03226456316552673, "alpha_c_p1m": 0.033864517788774195, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.00687261358907354, "ret_p3m": 0.6218127929611794, "ret_signed_p3m": 0.6218127929611794, "alpha_spy_p3m": 0.6033449008922456, "alpha_c_p3m": 0.6448977738396983, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": -0.023084980878518846, "ret_p6m": 1.651643939888746, "ret_signed_p6m": 1.651643939888746, "alpha_spy_p6m": 1.5750165368980622, "alpha_c_p6m": 1.493786849312865, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.15785709057588093, "price_path": [-0.3086, -0.3241, -0.3241, -0.3192, -0.3183, -0.3105, -0.3096, -0.3212, -0.3183, -0.3279, -0.327, -0.3472, -0.3328, -0.326, -0.3231, -0.3289, -0.3231, -0.3096, -0.299, -0.2912, -0.2719, -0.2613, -0.2333, -0.2449, -0.2246, -0.2246, -0.1937, -0.1821, -0.1754, -0.1686, -0.1956, -0.1833, -0.1862, -0.1659, -0.1295, -0.1295, -0.1295, -0.1295, -0.1295, -0.1295, -0.0844, -0.0951, -0.1115, -0.0786, -0.0524, -0.0504, -0.0485, -0.0543, -0.0436, -0.064, -0.0417, -0.0107, -0.0349, -0.0252, 0.0097, 0.0427, 0.0776, 0.0175, -0.0242, -0.0378, -0.0504, -0.0242, 0.0039, 0.0, -0.0029, -0.0572, -0.0242, -0.0514, -0.064, -0.0242, -0.0805, -0.0621, -0.0369, -0.0029, 0.0039, -0.0252, -0.0223, 0.0029, 0.0136, 0.0194, 0.0514, 0.0621, 0.0514, 0.0475, 0.0407, 0.0563, 0.0165, -0.0029, 0.0466, 0.0436, 0.031, 0.0718, 0.0815, 0.0776, 0.0776, 0.1348, 0.1647, 0.1686, 0.1686, 0.1686, 0.2524, 0.346, 0.3538, 0.3743, 0.3528, 0.3548, 0.3528, 0.3411, 0.3674, 0.4025, 0.4512, 0.4551, 0.4152, 0.4571, 0.4844, 0.4824, 0.4824, 0.5546, 0.5828, 0.5663, 0.5643, 0.4659, 0.6325, 0.6481, 0.5526, 0.5458, 0.6218, 0.616, 0.6355, 0.7407, 0.7661, 0.7661, 0.7661, 0.7661, 0.8518, 0.8528, 0.8811, 0.9493, 0.9834, 1.1247, 1.1101, 1.1101, 0.9015, 0.6783, 0.8674, 0.8343, 0.691, 0.8314, 0.8518, 0.8314, 0.7885, 0.8392, 0.8898, 1.0321, 0.9542, 0.9434, 0.8158, 0.8489, 0.8421, 0.7553, 0.7553, 0.7219, 0.633, 0.8522, 0.7424, 0.8185, 0.8859, 0.9191, 1.0559, 0.9924, 1.0119, 0.9631, 1.0168, 1.0608, 1.1242, 1.1096, 1.0949, 1.1389, 1.1242, 1.1926, 1.1438, 1.1926, 1.1682, 1.2073, 1.1535, 1.1535, 1.2707, 1.2707, 1.5979, 1.6516]}
{"unit_id": "orig:1988743719106101491", "ticker_idx": 2, "ts": "2025-11-12T23:00:26+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement)", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA reaffirms AI/DRAM/HBM super-cycle; key longs SK Hynix, Samsung, Micron on HBM/DDR5 mix; legacy DRAM names also benefit from price rebound through 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Confirmation of Strong Memory Market: Reaffirming an AI·DRAM·HBM-Driven Super-Boom Cycle (BofA)\n\n1. Industry Overview\n\n• All three experts — Nanya Tech CEO, an AI semiconductor specialist, and a TrendForce analyst — expressed a bullish view on the market outlook based on improving prices and supply–demand dynamics.\n• The global memory market is experiencing structural demand expansion centered on AI servers, which aligns with BofA’s super-cycle thesis.\n• In particular, TrendForce significantly raised its forecast for DRAM and NAND price increases compared to prior estimates, projecting that demand will far exceed supply even in 2026.\n\n2. TrendForce View\n\n• 4Q25 DRAM +30–35% (previous +18–23%), NAND +25–30% (previous +20–25%) — upward revision.\n• 1Q26 DRAM +20–25% (previous +8–13%), NAND +20–25% (previous +18–23%).\n• The 2026 supply-demand sufficiency ratio is expected to remain low — demand far stronger than supply.\n• DRAM makers are intentionally slowing shipments and building inventory strategically to induce higher selling prices later.\n• Rising memory prices may lead to higher end-product prices in 2026, including notebooks and smartphones.\n• Despite the 50–60% price premium of HBM4, the average HBM ASP in 2026 is still expected to decline about 10% YoY (although a shift from HBM to DDR5 could cause price increases).\n\n3. Nanya Tech View\n\n• Price uptrend expected to continue in 4Q; October revenue approached the peaks of 2018/2021.\n• DDR5-to-DDR4 production conversion is relatively easy, enabling flexible response to customer demand.\n• New fab expansions planned for 2026 will realistically only begin in 2H27, meaning bit growth in 2026 will be virtually flat.\n• Pricing negotiations are shifting from quarterly to monthly, reflecting a supplier-dominant market.\n• Channel inventory remains low, leaving room for further price increases.\n\n4. AI Semiconductor Expert Robert Castellano’s View\n\n• OpenAI’s massive AI infrastructure investment is driving structural demand for DRAM/HBM.\n• OpenAI does not own its own datacenters; CapEx expansion is carried out through partners (cloud providers, semiconductor companies, building leases) — approximately USD 1.4 trillion commitment over 8 years.\n• OpenAI’s annualized revenue is growing to around USD 28 billion.\n• U.S. big tech’s AI server chips (GPU+HBM) represent a shift toward “high-value, high-density memory,” unlike the dot-com/cloud bubbles of the past.\n• AI data center and sovereign AI infrastructure investment is expected to deepen HBM supply shortages.\n• Memory makers, unlike in past cycles, are pursuing “profit-centric” strategies and refraining from aggressive capacity expansions; simultaneous bottlenecks in HBM and DDR5 are tightening overall DRAM supply.\n\n5. Investment Implications\n\n• AI-driven demand for DRAM/HBM is expected to maintain structural growth for the next 1–2 years.\n• Supply increase remains limited, and memory makers are shifting toward price-focused profit optimization.\n• During the HBM4 transition period (2026–27), DDR5 production capacity is likely to shrink, supporting continued strength in commodity DRAM pricing.\n• Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement).\n• Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound.\n• The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015475712104625061, "ret_signed_m1d": -0.015475712104625061, "alpha_spy_m1d": -0.014919622905438468, "alpha_c_m1d": -0.0028748294874714198, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03246222316673142, "ret_signed_p1d": -0.03246222316673142, "alpha_spy_p1d": -0.015868201695993567, "alpha_c_p1d": -0.002338239505472628, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.07750103077046255, "ret_signed_p1w": -0.07750103077046255, "alpha_spy_p1w": -0.04713735687748066, "alpha_c_p1w": -0.0322447083511207, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": 0.055369522168985386, "ret_signed_p1m": 0.055369522168985386, "alpha_spy_p1m": 0.04689695395666438, "alpha_c_p1m": 0.012166348446257746, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.5665779193876388, "ret_signed_p3m": 0.5665779193876388, "alpha_spy_p3m": 0.548110027318705, "alpha_c_p3m": 0.41923356756138697, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6425602317905326, "ret_signed_p6m": 1.6425602317905326, "alpha_spy_p6m": 1.5659328287998489, "alpha_c_p6m": 1.1186928625271113, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.5068, -0.4958, -0.5019, -0.5217, -0.5275, -0.5198, -0.5249, -0.5246, -0.5195, -0.5021, -0.5144, -0.5144, -0.5165, -0.5155, -0.4931, -0.4639, -0.4635, -0.4481, -0.4287, -0.3856, -0.3584, -0.3562, -0.3519, -0.3471, -0.3108, -0.3359, -0.3282, -0.3209, -0.3401, -0.36, -0.3582, -0.3312, -0.3172, -0.2567, -0.2502, -0.233, -0.2203, -0.2418, -0.1975, -0.2147, -0.2585, -0.2129, -0.2362, -0.2163, -0.173, -0.1736, -0.1557, -0.174, -0.1896, -0.1559, -0.1057, -0.1013, -0.0939, -0.0746, -0.0853, -0.0863, -0.0416, -0.1097, -0.0302, -0.0268, -0.0285, 0.0343, -0.0155, 0.0, -0.0325, 0.0079, -0.012, -0.067, -0.0775, -0.1777, -0.1532, -0.0856, -0.0832, -0.0598, -0.0598, -0.0344, -0.0181, -0.0221, -0.0439, -0.0745, -0.0314, 0.0082, 0.0307, 0.0768, 0.0554, -0.0154, -0.0302, -0.0506, -0.0791, 0.0149, 0.0858, 0.1294, 0.1281, 0.1706, 0.1706, 0.1629, 0.2025, 0.1954, 0.1659, 0.1659, 0.2885, 0.2751, 0.4029, 0.387, 0.3359, 0.4097, 0.4129, 0.3812, 0.3617, 0.3751, 0.4818, 0.4818, 0.491, 0.5895, 0.6241, 0.6325, 0.5894, 0.6758, 0.7781, 0.7802, 0.6948, 0.7884, 0.7134, 0.5498, 0.5641, 0.6123, 0.5666, 0.5247, 0.6762, 0.691, 0.6816, 0.6816, 0.6331, 0.7196, 0.7049, 0.7491, 0.7196, 0.7075, 0.7524, 0.6975, 0.6845, 0.6857, 0.551, 0.6371, 0.6219, 0.5127, 0.5904, 0.6467, 0.7103, 0.6558, 0.7407, 0.8047, 0.886, 0.8861, 0.8148, 0.7275, 0.6517, 0.6157, 0.5608, 0.452, 0.4592, 0.3151, 0.3806, 0.5033, 0.4967, 0.4967, 0.5438, 0.543, 0.6622, 0.7226, 0.7188, 0.7432, 0.903, 0.8645, 0.8685, 0.8597, 0.8325, 0.8365, 0.9922, 0.9686, 1.0299, 1.1437, 1.0609, 1.1188, 1.1135, 1.2158, 1.3558, 1.6163, 1.7241, 1.6426]}
{"unit_id": "orig:1988743719106101491", "ticker_idx": 3, "ts": "2025-11-12T23:00:26+00:00", "symbol": "Nanya Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA reaffirms AI/DRAM/HBM super-cycle; key longs SK Hynix, Samsung, Micron on HBM/DDR5 mix; legacy DRAM names also benefit from price rebound through 2026.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya Technology Corp listed on TWSE as 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Confirmation of Strong Memory Market: Reaffirming an AI·DRAM·HBM-Driven Super-Boom Cycle (BofA)\n\n1. Industry Overview\n\n• All three experts — Nanya Tech CEO, an AI semiconductor specialist, and a TrendForce analyst — expressed a bullish view on the market outlook based on improving prices and supply–demand dynamics.\n• The global memory market is experiencing structural demand expansion centered on AI servers, which aligns with BofA’s super-cycle thesis.\n• In particular, TrendForce significantly raised its forecast for DRAM and NAND price increases compared to prior estimates, projecting that demand will far exceed supply even in 2026.\n\n2. TrendForce View\n\n• 4Q25 DRAM +30–35% (previous +18–23%), NAND +25–30% (previous +20–25%) — upward revision.\n• 1Q26 DRAM +20–25% (previous +8–13%), NAND +20–25% (previous +18–23%).\n• The 2026 supply-demand sufficiency ratio is expected to remain low — demand far stronger than supply.\n• DRAM makers are intentionally slowing shipments and building inventory strategically to induce higher selling prices later.\n• Rising memory prices may lead to higher end-product prices in 2026, including notebooks and smartphones.\n• Despite the 50–60% price premium of HBM4, the average HBM ASP in 2026 is still expected to decline about 10% YoY (although a shift from HBM to DDR5 could cause price increases).\n\n3. Nanya Tech View\n\n• Price uptrend expected to continue in 4Q; October revenue approached the peaks of 2018/2021.\n• DDR5-to-DDR4 production conversion is relatively easy, enabling flexible response to customer demand.\n• New fab expansions planned for 2026 will realistically only begin in 2H27, meaning bit growth in 2026 will be virtually flat.\n• Pricing negotiations are shifting from quarterly to monthly, reflecting a supplier-dominant market.\n• Channel inventory remains low, leaving room for further price increases.\n\n4. AI Semiconductor Expert Robert Castellano’s View\n\n• OpenAI’s massive AI infrastructure investment is driving structural demand for DRAM/HBM.\n• OpenAI does not own its own datacenters; CapEx expansion is carried out through partners (cloud providers, semiconductor companies, building leases) — approximately USD 1.4 trillion commitment over 8 years.\n• OpenAI’s annualized revenue is growing to around USD 28 billion.\n• U.S. big tech’s AI server chips (GPU+HBM) represent a shift toward “high-value, high-density memory,” unlike the dot-com/cloud bubbles of the past.\n• AI data center and sovereign AI infrastructure investment is expected to deepen HBM supply shortages.\n• Memory makers, unlike in past cycles, are pursuing “profit-centric” strategies and refraining from aggressive capacity expansions; simultaneous bottlenecks in HBM and DDR5 are tightening overall DRAM supply.\n\n5. Investment Implications\n\n• AI-driven demand for DRAM/HBM is expected to maintain structural growth for the next 1–2 years.\n• Supply increase remains limited, and memory makers are shifting toward price-focused profit optimization.\n• During the HBM4 transition period (2026–27), DDR5 production capacity is likely to shrink, supporting continued strength in commodity DRAM pricing.\n• Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement).\n• Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound.\n• The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009202453987730008, "ret_signed_m1d": 0.009202453987730008, "alpha_spy_m1d": 0.009758543186916602, "alpha_c_m1d": 0.02180333660488365, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0030674846625766694, "ret_signed_p1d": 0.0030674846625766694, "alpha_spy_p1d": 0.01966150613331452, "alpha_c_p1d": 0.03319146832383546, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.018404907975460127, "ret_signed_p1w": -0.018404907975460127, "alpha_spy_p1w": 0.011958765917521763, "alpha_c_p1w": 0.026851414443881727, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": -0.042944785276073594, "ret_signed_p1m": -0.042944785276073594, "alpha_spy_p1m": -0.0514173534883946, "alpha_c_p1m": -0.08614795899880123, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.6840490797546013, "ret_signed_p3m": 0.6840490797546013, "alpha_spy_p3m": 0.6655811876856674, "alpha_c_p3m": 0.5367047279283494, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 0.7607361963190185, "ret_signed_p6m": 0.7607361963190185, "alpha_spy_p6m": 0.6841087933283347, "alpha_c_p6m": 0.2368688270555972, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.7055, -0.7095, -0.7193, -0.7267, -0.715, -0.7218, -0.712, -0.7089, -0.7067, -0.7138, -0.7117, -0.7132, -0.7199, -0.7156, -0.7043, -0.6748, -0.6718, -0.6693, -0.6663, -0.6534, -0.6479, -0.6129, -0.5755, -0.5503, -0.5092, -0.5141, -0.5043, -0.5031, -0.5337, -0.5301, -0.5644, -0.5644, -0.5521, -0.5325, -0.511, -0.4908, -0.4908, -0.4405, -0.4429, -0.3957, -0.3957, -0.4135, -0.4534, -0.4706, -0.4178, -0.362, -0.3436, -0.3528, -0.3252, -0.3282, -0.3282, -0.2638, -0.2025, -0.184, -0.1748, -0.1871, -0.1595, -0.1933, -0.1534, -0.092, -0.0982, -0.0092, 0.0092, 0.0, 0.0031, -0.0276, 0.0215, -0.0153, -0.0184, -0.046, -0.1411, -0.1227, -0.1196, -0.1626, -0.1043, -0.1043, -0.0828, -0.0828, -0.0736, -0.0736, -0.0613, 0.0031, -0.0061, -0.0429, -0.0429, -0.0061, -0.0031, -0.0307, 0.0092, 0.0429, 0.0706, 0.1043, 0.0828, 0.1595, 0.1595, 0.1595, 0.1564, 0.1933, 0.184, 0.184, 0.2699, 0.2761, 0.4018, 0.4785, 0.4816, 0.3344, 0.4663, 0.411, 0.4509, 0.4571, 0.5337, 0.6871, 0.6687, 0.5399, 0.635, 0.6656, 0.7546, 0.6994, 0.8221, 0.8221, 1.0031, 0.8037, 0.7025, 0.8098, 0.6871, 0.6196, 0.684, 0.6626, 0.7055, 0.7055, 0.7055, 0.7055, 0.7055, 0.7055, 0.7055, 0.7055, 0.773, 0.8282, 0.7669, 0.7515, 0.7515, 0.7485, 0.5736, 0.4387, 0.5798, 0.4294, 0.2883, 0.4172, 0.5583, 0.4693, 0.4601, 0.6043, 0.6656, 0.6933, 0.5951, 0.4356, 0.3681, 0.3282, 0.3896, 0.3834, 0.3466, 0.3528, 0.2178, 0.2945, 0.2301, 0.2301, 0.2301, 0.2638, 0.3681, 0.2577, 0.319, 0.3834, 0.3497, 0.2975, 0.3098, 0.2883, 0.2699, 0.3497, 0.3436, 0.2761, 0.2638, 0.3896, 0.4571, 0.4417, 0.3221, 0.3221, 0.454, 0.5736, 0.7301, 0.7607]}
{"unit_id": "orig:1988743719106101491", "ticker_idx": 4, "ts": "2025-11-12T23:00:26+00:00", "symbol": "Winbond", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA reaffirms AI/DRAM/HBM super-cycle; key longs SK Hynix, Samsung, Micron on HBM/DDR5 mix; legacy DRAM names also benefit from price rebound through 2026.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Confirmation of Strong Memory Market: Reaffirming an AI·DRAM·HBM-Driven Super-Boom Cycle (BofA)\n\n1. Industry Overview\n\n• All three experts — Nanya Tech CEO, an AI semiconductor specialist, and a TrendForce analyst — expressed a bullish view on the market outlook based on improving prices and supply–demand dynamics.\n• The global memory market is experiencing structural demand expansion centered on AI servers, which aligns with BofA’s super-cycle thesis.\n• In particular, TrendForce significantly raised its forecast for DRAM and NAND price increases compared to prior estimates, projecting that demand will far exceed supply even in 2026.\n\n2. TrendForce View\n\n• 4Q25 DRAM +30–35% (previous +18–23%), NAND +25–30% (previous +20–25%) — upward revision.\n• 1Q26 DRAM +20–25% (previous +8–13%), NAND +20–25% (previous +18–23%).\n• The 2026 supply-demand sufficiency ratio is expected to remain low — demand far stronger than supply.\n• DRAM makers are intentionally slowing shipments and building inventory strategically to induce higher selling prices later.\n• Rising memory prices may lead to higher end-product prices in 2026, including notebooks and smartphones.\n• Despite the 50–60% price premium of HBM4, the average HBM ASP in 2026 is still expected to decline about 10% YoY (although a shift from HBM to DDR5 could cause price increases).\n\n3. Nanya Tech View\n\n• Price uptrend expected to continue in 4Q; October revenue approached the peaks of 2018/2021.\n• DDR5-to-DDR4 production conversion is relatively easy, enabling flexible response to customer demand.\n• New fab expansions planned for 2026 will realistically only begin in 2H27, meaning bit growth in 2026 will be virtually flat.\n• Pricing negotiations are shifting from quarterly to monthly, reflecting a supplier-dominant market.\n• Channel inventory remains low, leaving room for further price increases.\n\n4. AI Semiconductor Expert Robert Castellano’s View\n\n• OpenAI’s massive AI infrastructure investment is driving structural demand for DRAM/HBM.\n• OpenAI does not own its own datacenters; CapEx expansion is carried out through partners (cloud providers, semiconductor companies, building leases) — approximately USD 1.4 trillion commitment over 8 years.\n• OpenAI’s annualized revenue is growing to around USD 28 billion.\n• U.S. big tech’s AI server chips (GPU+HBM) represent a shift toward “high-value, high-density memory,” unlike the dot-com/cloud bubbles of the past.\n• AI data center and sovereign AI infrastructure investment is expected to deepen HBM supply shortages.\n• Memory makers, unlike in past cycles, are pursuing “profit-centric” strategies and refraining from aggressive capacity expansions; simultaneous bottlenecks in HBM and DDR5 are tightening overall DRAM supply.\n\n5. Investment Implications\n\n• AI-driven demand for DRAM/HBM is expected to maintain structural growth for the next 1–2 years.\n• Supply increase remains limited, and memory makers are shifting toward price-focused profit optimization.\n• During the HBM4 transition period (2026–27), DDR5 production capacity is likely to shrink, supporting continued strength in commodity DRAM pricing.\n• Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement).\n• Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound.\n• The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00462965319084141, "ret_signed_m1d": -0.00462965319084141, "alpha_spy_m1d": -0.004073563991654816, "alpha_c_m1d": 0.007971229426312232, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00925930638168282, "ret_signed_p1d": -0.00925930638168282, "alpha_spy_p1d": 0.007334715089055033, "alpha_c_p1d": 0.02086467727957597, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.08333328401916906, "ret_signed_p1w": -0.08333328401916906, "alpha_spy_p1w": -0.05296961012618717, "alpha_c_p1w": -0.03807696159982721, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": 0.055555601582108816, "ret_signed_p1m": 0.055555601582108816, "alpha_spy_p1m": 0.04708303336978781, "alpha_c_p1m": 0.012352427859381176, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.6589506569179271, "ret_signed_p3m": 0.6589506569179271, "alpha_spy_p3m": 0.6404827648489932, "alpha_c_p3m": 0.5116063050916753, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 0.768469998860251, "ret_signed_p6m": 0.768469998860251, "alpha_spy_p6m": 0.6918425958695673, "alpha_c_p6m": 0.24460262959682977, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.716, -0.7052, -0.713, -0.7222, -0.716, -0.7191, -0.716, -0.7106, -0.6944, -0.6975, -0.696, -0.6867, -0.6921, -0.6898, -0.6836, -0.652, -0.6204, -0.6211, -0.615, -0.6142, -0.6073, -0.5864, -0.5687, -0.5463, -0.5015, -0.483, -0.4715, -0.4375, -0.4568, -0.4846, -0.5154, -0.5154, -0.4745, -0.4784, -0.4444, -0.4012, -0.4012, -0.3418, -0.3634, -0.3295, -0.3295, -0.3364, -0.3719, -0.3827, -0.321, -0.3218, -0.2847, -0.3117, -0.2986, -0.2855, -0.2855, -0.2145, -0.1667, -0.1667, -0.1481, -0.1636, -0.1265, -0.1775, -0.1358, -0.0895, -0.1034, -0.0139, -0.0046, 0.0, -0.0093, -0.0694, 0.0216, -0.0525, -0.0833, -0.1049, -0.1944, -0.1574, -0.1127, -0.1728, -0.1219, -0.1049, -0.1111, -0.1265, -0.1389, -0.108, -0.0463, 0.0478, 0.1003, 0.0417, 0.0556, 0.1512, 0.0972, 0.0895, 0.1404, 0.1497, 0.1003, 0.1281, 0.1265, 0.1867, 0.1867, 0.1806, 0.1867, 0.2917, 0.2747, 0.2747, 0.4012, 0.4738, 0.6049, 0.6435, 0.6744, 0.5093, 0.5741, 0.5401, 0.5401, 0.5309, 0.6435, 0.7978, 0.7438, 0.6281, 0.6127, 0.5895, 0.7438, 0.7824, 0.9599, 1.0062, 0.983, 0.7901, 0.6281, 0.6744, 0.5818, 0.5093, 0.659, 0.5972, 0.6281, 0.6281, 0.6281, 0.6281, 0.6281, 0.6281, 0.6281, 0.6281, 0.7747, 0.8827, 0.821, 0.8904, 0.8904, 0.9213, 0.7361, 0.6049, 0.7284, 0.6435, 0.5586, 0.6281, 0.7593, 0.6821, 0.6821, 0.8056, 0.9059, 0.9753, 0.8827, 0.6975, 0.659, 0.5355, 0.5154, 0.4815, 0.4334, 0.4427, 0.3993, 0.4443, 0.4241, 0.4241, 0.4241, 0.3869, 0.483, 0.4117, 0.4505, 0.4644, 0.4505, 0.3993, 0.39, 0.362, 0.3295, 0.4132, 0.4055, 0.362, 0.3682, 0.4567, 0.4815, 0.438, 0.3931, 0.3931, 0.4799, 0.5327, 0.6831, 0.7685]}
{"unit_id": "orig:1988743719106101491", "ticker_idx": 5, "ts": "2025-11-12T23:00:26+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx", "tweet_type": "news_reaction", "timeline": "years", "summary": "BofA reaffirms AI/DRAM/HBM super-cycle; key longs SK Hynix, Samsung, Micron on HBM/DDR5 mix; legacy DRAM names also benefit from price rebound through 2026.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Confirmation of Strong Memory Market: Reaffirming an AI·DRAM·HBM-Driven Super-Boom Cycle (BofA)\n\n1. Industry Overview\n\n• All three experts — Nanya Tech CEO, an AI semiconductor specialist, and a TrendForce analyst — expressed a bullish view on the market outlook based on improving prices and supply–demand dynamics.\n• The global memory market is experiencing structural demand expansion centered on AI servers, which aligns with BofA’s super-cycle thesis.\n• In particular, TrendForce significantly raised its forecast for DRAM and NAND price increases compared to prior estimates, projecting that demand will far exceed supply even in 2026.\n\n2. TrendForce View\n\n• 4Q25 DRAM +30–35% (previous +18–23%), NAND +25–30% (previous +20–25%) — upward revision.\n• 1Q26 DRAM +20–25% (previous +8–13%), NAND +20–25% (previous +18–23%).\n• The 2026 supply-demand sufficiency ratio is expected to remain low — demand far stronger than supply.\n• DRAM makers are intentionally slowing shipments and building inventory strategically to induce higher selling prices later.\n• Rising memory prices may lead to higher end-product prices in 2026, including notebooks and smartphones.\n• Despite the 50–60% price premium of HBM4, the average HBM ASP in 2026 is still expected to decline about 10% YoY (although a shift from HBM to DDR5 could cause price increases).\n\n3. Nanya Tech View\n\n• Price uptrend expected to continue in 4Q; October revenue approached the peaks of 2018/2021.\n• DDR5-to-DDR4 production conversion is relatively easy, enabling flexible response to customer demand.\n• New fab expansions planned for 2026 will realistically only begin in 2H27, meaning bit growth in 2026 will be virtually flat.\n• Pricing negotiations are shifting from quarterly to monthly, reflecting a supplier-dominant market.\n• Channel inventory remains low, leaving room for further price increases.\n\n4. AI Semiconductor Expert Robert Castellano’s View\n\n• OpenAI’s massive AI infrastructure investment is driving structural demand for DRAM/HBM.\n• OpenAI does not own its own datacenters; CapEx expansion is carried out through partners (cloud providers, semiconductor companies, building leases) — approximately USD 1.4 trillion commitment over 8 years.\n• OpenAI’s annualized revenue is growing to around USD 28 billion.\n• U.S. big tech’s AI server chips (GPU+HBM) represent a shift toward “high-value, high-density memory,” unlike the dot-com/cloud bubbles of the past.\n• AI data center and sovereign AI infrastructure investment is expected to deepen HBM supply shortages.\n• Memory makers, unlike in past cycles, are pursuing “profit-centric” strategies and refraining from aggressive capacity expansions; simultaneous bottlenecks in HBM and DDR5 are tightening overall DRAM supply.\n\n5. Investment Implications\n\n• AI-driven demand for DRAM/HBM is expected to maintain structural growth for the next 1–2 years.\n• Supply increase remains limited, and memory makers are shifting toward price-focused profit optimization.\n• During the HBM4 transition period (2026–27), DDR5 production capacity is likely to shrink, supporting continued strength in commodity DRAM pricing.\n• Key beneficiaries: SK hynix, Samsung Electronics, Micron (HBM supply expansion, DDR5/HBM mix improvement).\n• Legacy DRAM makers (Nanya, Winbond, etc.) may also benefit from short-term price rebound.\n• The likelihood of a memory super-cycle through 2026 is strengthening, with the industry entering a high-profit phase supported by prolonged AI CapEx.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0027848619212067862, "ret_signed_m1d": -0.0027848619212067862, "alpha_spy_m1d": -0.0022287727220201927, "alpha_c_m1d": 0.009816020695946856, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014491902517433672, "ret_signed_p1d": -0.014491902517433672, "alpha_spy_p1d": 0.0021021189533041804, "alpha_c_p1d": 0.015632081143825118, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.07688584208204279, "ret_signed_p1w": -0.07688584208204279, "alpha_spy_p1w": -0.046522168189060895, "alpha_c_p1w": -0.03162951966270093, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": 0.004161230922664812, "ret_signed_p1m": 0.004161230922664812, "alpha_spy_p1m": -0.004311337289656196, "alpha_c_p1m": -0.039041942800062825, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.5423405575076978, "ret_signed_p3m": 0.5423405575076978, "alpha_spy_p3m": 0.523872665438764, "alpha_c_p3m": 0.394996205681446, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6605957761227985, "ret_signed_p6m": 1.6605957761227985, "alpha_spy_p6m": 1.5839683731321148, "alpha_c_p6m": 1.1367284068593773, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.456, -0.4623, -0.4668, -0.4758, -0.4831, -0.4747, -0.4715, -0.4742, -0.4723, -0.465, -0.4684, -0.4822, -0.4757, -0.472, -0.462, -0.4498, -0.4459, -0.4303, -0.4116, -0.3931, -0.366, -0.3603, -0.3404, -0.3505, -0.3197, -0.3281, -0.3177, -0.306, -0.312, -0.3169, -0.3362, -0.3163, -0.3134, -0.2797, -0.2462, -0.2405, -0.2362, -0.2434, -0.2286, -0.2344, -0.2164, -0.2118, -0.2269, -0.2034, -0.164, -0.1565, -0.1391, -0.1507, -0.1509, -0.1481, -0.1069, -0.0816, -0.0948, -0.0651, -0.0517, -0.0459, 0.0136, -0.0475, -0.0387, -0.0345, -0.0463, -0.0026, -0.0028, 0.0, -0.0145, -0.0472, -0.018, -0.0648, -0.0769, -0.0922, -0.1298, -0.1016, -0.093, -0.0711, -0.0579, -0.0667, -0.056, -0.0381, -0.045, -0.0587, -0.0325, 0.0021, 0.0, 0.0255, 0.0042, -0.011, -0.0384, -0.0646, -0.0463, -0.0154, 0.0014, 0.0473, 0.0523, 0.0673, 0.0673, 0.0897, 0.1351, 0.1399, 0.1301, 0.1301, 0.213, 0.25, 0.3114, 0.3216, 0.3049, 0.3237, 0.3268, 0.3064, 0.3109, 0.3308, 0.3864, 0.392, 0.3704, 0.4156, 0.4443, 0.453, 0.4219, 0.5093, 0.575, 0.581, 0.5778, 0.5335, 0.6057, 0.5526, 0.4941, 0.5063, 0.5423, 0.5205, 0.5688, 0.624, 0.625, 0.625, 0.6088, 0.6376, 0.6689, 0.7137, 0.7144, 0.7623, 0.7957, 0.8693, 0.8395, 0.84, 0.6594, 0.565, 0.6728, 0.6161, 0.5466, 0.6674, 0.7046, 0.6661, 0.6693, 0.7422, 0.784, 0.8781, 0.805, 0.7691, 0.6612, 0.6889, 0.6732, 0.5745, 0.5769, 0.4852, 0.4416, 0.6044, 0.5292, 0.5796, 0.6231, 0.6502, 0.7989, 0.7789, 0.7998, 0.7987, 0.904, 0.9237, 0.9565, 0.9341, 0.9407, 0.9881, 1.0346, 1.0506, 1.0531, 1.1452, 1.1138, 1.1423, 1.1189, 1.153, 1.3259, 1.4128, 1.6412, 1.6606]}
{"unit_id": "orig:1976242882341425580", "ticker_idx": 0, "ts": "2025-10-09T11:06:34+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA remains our Top Pick and continues to offer the greatest upside potential. Accordingly, we raise our price target from $240 to $300. This implies more than 60% near-term upside", "tweet_type": "news_reaction", "timeline": "years", "summary": "Cantor Fitzgerald raises NVDA PT to $300 (27x 2027E EPS $11); Top Pick with 60%+ upside, EPS estimates 50% above consensus through 2027.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Cantor Fitzgerald) NVIDIA: Raises Target Price to $300 Amid Accelerating AI Infrastructure Growth\n\n“Last week, we hosted an investor meeting in New York with NVIDIA’s senior management, including CEO Jensen Huang, CFO Colette Kress, IR head Toshiya Hari, and other executives. Compared to a year ago, much has changed—exponentially. As expected, Jensen Huang was highly optimistic, and with good reason. We are still in the early innings of a multi-trillion-dollar AI infrastructure build-out. Even just among hyperscalers, we see substantial visibility into hundreds of billions of dollars of demand over the next several years—not to mention other growth engines such as neo-cloud, enterprise AI, and physical AI. Thus, this is not a bubble; we are at the beginning of this investment cycle.”\n\n“One particularly notable point from the meeting was the new partnership with OpenAI, through which NVIDIA aims to help build OpenAI into a self-hosted hyperscaler. The core of this vision is to eliminate ‘margin stacking’ by server ODMs (original design manufacturers) and CSPs (cloud service providers), thereby narrowing the cost gap between NVIDIA GPUs and ASICs to roughly 15% on average. This is a true win-win for both parties and could put further pressure on the ASIC market. With full-stack solutions like CUDA-X, NVIDIA continues its annual cycle of ‘Extreme Co-Design,’ enabling end-to-end optimization of the entire AI infrastructure (nobody just wants to buy chips—they want large-scale AI deployments). Simply put, Jensen Huang and his team are more focused, more competitive, and running harder than ever before—to ‘produce results,’ as he puts it. We find it hard to envision a scenario where NVIDIA fails to capture at least 75% of the AI accelerator market in the long term. After three days of meetings, our conviction in the growth of AI infrastructure and NVIDIA’s share into the 2030s has only strengthened.”\n\n“We continue to believe that achieving EPS of $8 by 2026 (vs. consensus of $6.26) is highly attainable, followed by $11 in 2027 (vs. consensus of $7.36). Given that long-term consensus estimates may still be 50% too low, NVIDIA remains our Top Pick and continues to offer the greatest upside potential. Accordingly, we raise our price target from $240 (based on 30× 2026E EPS of $8) to $300 (based on 27× 2027E EPS of $11). This implies more than 60% near-term upside, with even greater long-term potential as the company pursues its ambitious goal of $50 EPS by 2030 amid a $3–4 trillion AI infrastructure market. To borrow from the Boss—Bruce Springsteen—baby, NVIDIA was ‘Born to Run’… and it’s still running.”\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017967518834077323, "ret_signed_m1d": -0.017967518834077323, "alpha_spy_m1d": -0.020872944255104775, "alpha_c_m1d": -0.020458839029821707, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04886534322137259, "ret_signed_p1d": -0.04886534322137259, "alpha_spy_p1d": -0.021837515322282952, "alpha_c_p1d": 0.008693783732913518, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": -0.05587578994716014, "ret_signed_p1w": -0.05587578994716014, "alpha_spy_p1w": -0.04020151001351813, "alpha_c_p1w": -0.05034290246324635, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": -0.02295266728066714, "ret_signed_p1m": -0.02295266728066714, "alpha_spy_p1m": -0.022669514513621647, "alpha_c_p1m": -0.031382399529811145, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": -0.027624079534145696, "ret_signed_p3m": -0.027624079534145696, "alpha_spy_p3m": -0.0614375135049966, "alpha_c_p3m": -0.15397074389281917, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": -0.07872733891956518, "ret_signed_p6m": -0.07872733891956518, "alpha_spy_p6m": -0.061450264333961946, "alpha_c_p6m": -0.21873130460046342, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.148, -0.1136, -0.1101, -0.1017, -0.1047, -0.1101, -0.1327, -0.1132, -0.0978, -0.0991, -0.0822, -0.0886, -0.0691, -0.0764, -0.0979, -0.0653, -0.0744, -0.0683, -0.0613, -0.0513, -0.0546, -0.0489, -0.0571, -0.0548, -0.063, -0.0549, -0.088, -0.0892, -0.0914, -0.0758, -0.0663, -0.0561, -0.057, -0.0644, -0.0955, -0.0955, -0.1132, -0.114, -0.1086, -0.1327, -0.126, -0.1133, -0.0792, -0.08, -0.0766, -0.077, -0.0919, -0.1157, -0.0848, -0.0826, -0.0465, -0.0734, -0.081, -0.0773, -0.0747, -0.0557, -0.0311, -0.0277, -0.0191, -0.0257, -0.0365, -0.0391, -0.018, 0.0, -0.0489, -0.0221, -0.0651, -0.0662, -0.0559, -0.0486, -0.0516, -0.0593, -0.0638, -0.0541, -0.0328, -0.0056, 0.0439, 0.0751, 0.0536, 0.0515, 0.0743, 0.0318, 0.0137, -0.0233, -0.023, 0.0337, 0.0031, 0.0064, -0.0297, -0.0125, -0.031, -0.0582, -0.0314, -0.062, -0.0711, -0.052, -0.0766, -0.0639, -0.0639, -0.0809, -0.0657, -0.0577, -0.0674, -0.0477, -0.0527, -0.0364, -0.0394, -0.0456, -0.0604, -0.0911, -0.0845, -0.0771, -0.1123, -0.0957, -0.0601, -0.0461, -0.0174, -0.0205, -0.0205, -0.0105, -0.0225, -0.0261, -0.0315, -0.0315, -0.0193, -0.0231, -0.0276, -0.0179, -0.039, -0.04, -0.0396, -0.0351, -0.0489, -0.0286, -0.0329, -0.0329, -0.0752, -0.048, -0.0401, -0.0254, -0.0316, -0.021, -0.0054, -0.0003, -0.0074, -0.0361, -0.0635, -0.0954, -0.1074, -0.0371, -0.0131, -0.0209, -0.013, -0.0292, -0.0506, -0.0506, -0.0394, -0.0238, -0.0242, -0.0142, -0.0052, 0.0015, 0.0156, -0.0398, -0.0798, -0.0523, -0.065, -0.0494, -0.0479, -0.0765, -0.0515, -0.0405, -0.0339, -0.0489, -0.0639, -0.0484, -0.0551, -0.0631, -0.0727, -0.1031, -0.0878, -0.0901, -0.072, -0.1107, -0.13, -0.1422, -0.0943, -0.0872, -0.0787, -0.0787]}
{"unit_id": "orig:1975520390018699724", "ticker_idx": 0, "ts": "2025-10-07T11:15:39+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Major win for AMD... the deal still serves as strong validation of AMD's data-center AI roadmap... AMD believes this project provides clear visibility toward achieving its initial $10 billion annual AI revenue target in 2027 and potentially exceeding $100 billion in the following years", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Wolfe Research's bullish AMD analysis: 6GW OpenAI deal validates AI roadmap, $10B-$15B per GW revenue, $1.25-$1.70 EPS accretion from first tranche alone.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Wolfe Research) AMD: OpenAI to Deploy 6GW of AMD GPUs\n\nThis morning, AMD announced a 6-gigawatt (GW) GPU supply agreement with OpenAI, spanning the next five years. According to AMD, every 1GW deployment represents “significant double-digit billions” in revenue. The first batch is expected to begin in the second half of 2026, following the mid-2026 launch of the MI450-Helios rack. AMD believes this project provides clear visibility toward achieving its initial $10 billion annual AI revenue target in 2027 and potentially exceeding $100 billion in the following years. The company noted that gross margins (GM) for this business will be consistent with previous guidance — slightly below the corporate average — but emphasized that the deal will generate meaningful operating leverage and EPS accretion.\n\n⸻\n\nKey Takeaways\n\n● Major win for AMD.\nThe first 1GW batch begins deployment in late 2026. Although the MI450 specifications are not fully disclosed, assuming similar power consumption to NVIDIA’s VR200 and rack pricing of roughly $2 million–$3 million (vs $3 million for Blackwell and $4 million for Rubin), the first 1GW could yield $10 billion–$15 billion in revenue. This figure exceeds AMD’s expected 2025 data-center AI revenue of roughly $6.5 billion, making the contract highly significant. AMD has guided for “tens of billions” in data-center AI revenue by 2027. Accounting for warrant dilution and assuming margins near the corporate operating margin (OM), Wolfe estimates incremental EPS of $1.25–$1.70 per share from the first 1GW tranche alone.\n\n● Funding mechanism raises mild concerns.\nAs part of the deal, AMD will issue stock warrants to OpenAI, exercisable in tranches tied to GPU deployment volume. Wolfe views this as a form of equity-financed transaction — in contrast to NVIDIA’s prior OpenAI partnership, where NVIDIA indirectly helped fund OpenAI’s hardware purchases. Assuming the first tranche equals 26.7 million shares (one-sixth of total warrants), this represents a $6 billion transfer of value to OpenAI at current share prices, which OpenAI could liquidate to pay for GPU purchases.\n\n● Validates AMD as a key AI player.\nThe 6GW agreement follows NVIDIA’s reported 10GW OpenAI contract and Broadcom’s $10 billion custom-chip deal disclosed in its September earnings (believed also linked to OpenAI). Although the unique financing structure likely influenced the award, the deal still serves as strong validation of AMD’s data-center AI roadmap — OpenAI would not have signed if AMD’s solution were unproven.\n\n● Rack-level assumptions.\nWhile MI450 specs and pricing remain undisclosed, Wolfe assumes the following based on NVIDIA’s VR200: 1.8 kW per GPU, 72 GPUs per rack, and 200 kW per rack including overhead. That implies ~5,000 racks per GW. Using a $2 million–$3 million rack price (vs $3 million for GB200, $4 million for VR200), AMD would generate $10 billion–$15 billion per GW — broadly consistent with AMD’s own “significant double-digit billions” comment.\n\n● Warrants.\nAMD will issue up to 160 million common-stock warrants to OpenAI, vesting over time as OpenAI scales AMD deployments to 6GW. The first tranche vests with the initial 1GW MI450 rollout; later tranches vest as deployments expand and specific technical/commercial milestones are achieved. Vesting is also tied to AMD share-price targets, with the final tranche activating at $600 per share. Assuming proportional allocation, the first tranche equals 26.7 million shares worth $5.8 billion — a material portion of GPU contract value.\n\n● EPS impact.\nAfter factoring in dilution from 26.7 million additional shares and assuming $10 billion–$15 billion incremental revenue at a 25% operating margin, Wolfe estimates $2.2 billion–$2.8 billion in incremental net income — translating to $1.25–$1.70 of annualized EPS accretion from the first 1GW segment.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03687763224218066, "ret_signed_m1d": -0.03687763224218066, "alpha_spy_m1d": -0.04059892114290564, "alpha_c_m1d": -0.0556582260463776, "bench_spy_m1d": 0.003721288900724984, "bench_c_m1d": 0.018780593804196943, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11370622512583206, "ret_signed_p1d": 0.11370622512583206, "alpha_spy_p1d": 0.10774317271303979, "alpha_c_p1d": 0.08694457376543374, "bench_spy_p1d": 0.005963052412792269, "bench_c_p1d": 0.02676165136039832, "ret_p1w": 0.031109649671150708, "ret_signed_p1w": 0.031109649671150708, "alpha_spy_p1w": 0.04140679043879947, "alpha_c_p1w": 0.04164216067572424, "bench_spy_p1w": -0.01029714076764876, "bench_c_p1w": -0.01053251100457353, "ret_p1m": 0.2119048424635135, "ret_signed_p1m": 0.2119048424635135, "alpha_spy_p1m": 0.19926135863623595, "alpha_c_p1m": 0.14529748446485868, "bench_spy_p1m": 0.012643483827277535, "bench_c_p1m": 0.0666073579986548, "ret_p3m": 0.05654582300828648, "ret_signed_p3m": 0.05654582300828648, "alpha_spy_p3m": 0.03253114811261604, "alpha_c_p3m": -0.05445121201243741, "bench_spy_p3m": 0.02401467489567044, "bench_c_p3m": 0.11099703502072389, "ret_p6m": -0.006146223945586393, "ret_signed_p6m": -0.006146223945586393, "alpha_spy_p6m": 0.009021571645064674, "alpha_c_p6m": -0.17270803325050788, "bench_spy_p6m": -0.015167795590651068, "bench_c_p6m": 0.16656180930492148, "price_path": [-0.3184, -0.3077, -0.3086, -0.2643, -0.2432, -0.2416, -0.2578, -0.2577, -0.2685, -0.2499, -0.2335, -0.2129, -0.179, -0.1611, -0.1513, -0.1664, -0.1882, -0.1642, -0.1759, -0.2288, -0.1849, -0.1832, -0.1855, -0.1729, -0.1281, -0.1445, -0.1607, -0.1672, -0.2126, -0.2189, -0.226, -0.2068, -0.2276, -0.2122, -0.2098, -0.203, -0.2311, -0.2311, -0.2326, -0.2335, -0.2351, -0.2854, -0.2841, -0.2633, -0.2457, -0.264, -0.2503, -0.2381, -0.2414, -0.2475, -0.2534, -0.2559, -0.2445, -0.2393, -0.2394, -0.2375, -0.2461, -0.2371, -0.2351, -0.2246, -0.1975, -0.2215, -0.0369, 0.0, 0.1137, 0.1011, 0.016, 0.0232, 0.0311, 0.1281, 0.109, 0.102, 0.1373, 0.1254, 0.0885, 0.111, 0.1958, 0.2277, 0.2198, 0.2497, 0.2049, 0.2109, 0.2276, 0.1822, 0.2119, 0.1238, 0.1042, 0.1535, 0.123, 0.224, 0.1723, 0.1669, 0.1372, 0.0888, 0.0569, -0.026, -0.0365, 0.0167, -0.0254, 0.0129, 0.0129, 0.0285, 0.039, 0.0176, 0.0288, 0.0211, 0.0305, 0.0454, 0.0478, 0.0469, 0.0469, -0.0035, -0.0186, -0.0111, -0.0634, -0.0494, 0.0091, 0.0163, 0.016, 0.0167, 0.0167, 0.0165, 0.0194, 0.0181, 0.0125, 0.0125, 0.0565, 0.0452, 0.0134, -0.007, -0.0323, -0.0394, -0.0181, 0.0447, 0.0572, 0.0776, 0.0961, 0.0961, 0.0965, 0.181, 0.1996, 0.2277, 0.1882, 0.1916, 0.1949, 0.1923, 0.1192, 0.1643, 0.1447, -0.0535, -0.0899, -0.0145, 0.0212, 0.0097, 0.0098, -0.0263, -0.0198, -0.0198, -0.0399, -0.0539, -0.0385, -0.0537, -0.0705, 0.011, -0.0031, -0.037, -0.0534, -0.0609, -0.0972, -0.0446, -0.057, -0.0902, -0.0417, -0.0391, -0.0316, -0.0651, -0.0857, -0.0706, -0.0719, -0.057, -0.0295, -0.0481, -0.0417, -0.029, 0.0414, -0.0366, -0.045, -0.0731, -0.0382, -0.0061]}
{"unit_id": "quote:1980431514258272356", "ticker_idx": 1, "ts": "2025-10-21T00:30:42+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom's custom ASIC project for OpenAI is expected to enter small-scale production between late 2026 and early 2027", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Positive views on Phison (QLC NAND adoption), Broadcom (OpenAI ASIC ramp), and Alchip (Trainium3 no cuts).", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• Meta:\nMeta plans to prioritize QLC NAND adoption in its data centers to reduce costs.\nAs a result, analysts reaffirmed a positive view on Phison (8299).\n\n• ASIC Segment:\nBroadcom’s custom ASIC project for OpenAI is expected to enter small-scale production between late 2026 and early 2027.\n\n• AWS Trainium3:\nThere are no signs of production cuts.\nAnalysts maintain a positive outlook on Alchip (3661).", "reply_text": null, "parent_text": null, "quoted_text": "早安！10/21 外電綜合整理\n\n- OCP大會：\n美系大行整理此次大會的四大重點為：AMD 的 Helios 機櫃（ORW）、800V 直流電架構、Google Deschutes 2MW 冷卻配電單元（CDU）以及 ESUN 網路。\n\n* Helios 機櫃採用雙寬設計（2T），明年下半年開始出貨。其中，Meta 與 AMD", "parent_author": null, "ret_m1d": 0.019202600716187623, "ret_signed_m1d": 0.019202600716187623, "alpha_spy_m1d": 0.019187604571294203, "alpha_c_m1d": 0.0136421652007499, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0068873625963382645, "ret_signed_p1d": -0.0068873625963382645, "alpha_spy_p1d": -0.0016885466334175536, "alpha_c_p1d": 0.012516004091333688, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.08845491359008517, "ret_signed_p1w": 0.08845491359008517, "alpha_spy_p1w": 0.0649627210951802, "alpha_c_p1w": 0.036876554890219904, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.034319674881862605, "ret_signed_p1m": 0.034319674881862605, "alpha_spy_p1m": 0.04722001708657386, "alpha_c_p1m": 0.05129035103153057, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.02837487160635943, "ret_signed_p3m": 0.02837487160635943, "alpha_spy_p3m": -0.005014319318705107, "alpha_c_p3m": -0.13477568250451522, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.16241421087236496, "ret_signed_p6m": 0.16241421087236496, "alpha_spy_p6m": 0.11379808072630992, "alpha_c_p6m": -0.15357065821803229, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.1589, -0.1546, -0.1426, -0.1335, -0.1184, -0.1443, -0.1591, -0.1326, -0.1466, -0.1211, -0.115, -0.1115, -0.1146, -0.0886, -0.0995, -0.0933, -0.1075, -0.1092, -0.1408, -0.1517, -0.1563, -0.1435, -0.1428, -0.1318, -0.1253, -0.1008, -0.1336, -0.1336, -0.1311, -0.119, -0.1082, -0.0243, 0.007, -0.0192, 0.0767, 0.0477, 0.0484, 0.0607, 0.0488, 0.0085, 0.0061, 0.0049, -0.0113, -0.0109, -0.0098, -0.0191, -0.0237, -0.0431, -0.0372, -0.0271, -0.0131, -0.0125, -0.0209, -0.0182, 0.0083, 0.0069, -0.0526, 0.041, 0.0043, 0.0253, 0.0335, 0.0195, 0.0192, 0.0, -0.0069, 0.0048, 0.0335, 0.0566, 0.0885, 0.1264, 0.0987, 0.0787, 0.058, 0.0271, 0.0476, 0.0377, 0.0198, 0.0459, 0.0271, 0.0367, -0.0078, -0.0006, -0.0, -0.0063, 0.0343, 0.0121, -0.0072, 0.103, 0.1237, 0.1602, 0.1602, 0.176, 0.1267, 0.1136, 0.1108, 0.112, 0.1389, 0.1705, 0.1857, 0.2052, 0.1859, 0.0504, -0.0083, -0.004, -0.0486, -0.0373, -0.0067, -0.0016, 0.0214, 0.024, 0.024, 0.0296, 0.0216, 0.0229, 0.012, 0.012, 0.0164, 0.0041, 0.0052, 0.0044, -0.0279, 0.0087, 0.0298, 0.0369, -0.0062, 0.003, 0.0284, 0.0284, -0.0275, -0.0386, -0.0483, -0.0642, -0.0502, -0.0269, -0.0256, -0.033, -0.0313, -0.0319, -0.0634, -0.0993, -0.0921, -0.0266, 0.0057, -0.0046, 0.0022, -0.0317, -0.0492, -0.0492, -0.0277, -0.0248, -0.0234, -0.0274, -0.0341, -0.0483, -0.0283, -0.0594, -0.0657, -0.0678, -0.0824, -0.0716, -0.027, -0.0337, 0.0109, 0.0017, -0.0013, -0.0176, -0.058, -0.05, -0.0605, -0.0762, -0.0648, -0.0921, -0.055, -0.0674, -0.0659, -0.0934, -0.119, -0.1403, -0.0931, -0.0815, -0.0783, -0.0783, -0.0787, -0.0214, 0.0274, 0.0399, 0.0887, 0.1127, 0.1157, 0.1624]}
{"unit_id": "quote:1980431514258272356", "ticker_idx": 2, "ts": "2025-10-21T00:30:42+00:00", "symbol": "3661.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Analysts maintain a positive outlook on Alchip (3661)", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Positive views on Phison (QLC NAND adoption), Broadcom (OpenAI ASIC ramp), and Alchip (Trainium3 no cuts).", "resolved_tickers": ["3661.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• Meta:\nMeta plans to prioritize QLC NAND adoption in its data centers to reduce costs.\nAs a result, analysts reaffirmed a positive view on Phison (8299).\n\n• ASIC Segment:\nBroadcom’s custom ASIC project for OpenAI is expected to enter small-scale production between late 2026 and early 2027.\n\n• AWS Trainium3:\nThere are no signs of production cuts.\nAnalysts maintain a positive outlook on Alchip (3661).", "reply_text": null, "parent_text": null, "quoted_text": "早安！10/21 外電綜合整理\n\n- OCP大會：\n美系大行整理此次大會的四大重點為：AMD 的 Helios 機櫃（ORW）、800V 直流電架構、Google Deschutes 2MW 冷卻配電單元（CDU）以及 ESUN 網路。\n\n* Helios 機櫃採用雙寬設計（2T），明年下半年開始出貨。其中，Meta 與 AMD", "parent_author": null, "ret_m1d": -0.03576751117734722, "ret_signed_m1d": -0.03576751117734722, "alpha_spy_m1d": -0.03578250732224064, "alpha_c_m1d": -0.041327946692784945, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04023845007451565, "ret_signed_p1d": -0.04023845007451565, "alpha_spy_p1d": -0.03503963411159494, "alpha_c_p1d": -0.0208350833868437, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": -0.08643815201192251, "ret_signed_p1w": -0.08643815201192251, "alpha_spy_p1w": -0.10993034450682748, "alpha_c_p1w": -0.13801651071178778, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": -0.010432190760059634, "ret_signed_p1m": -0.010432190760059634, "alpha_spy_p1m": 0.002468151444651623, "alpha_c_p1m": 0.006538485389608328, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": -0.0044709388971684305, "ret_signed_p3m": -0.0044709388971684305, "alpha_spy_p3m": -0.03786012982223297, "alpha_c_p3m": -0.16762149300804308, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": -0.0432190760059612, "ret_signed_p6m": -0.0432190760059612, "alpha_spy_p6m": -0.09183520615201624, "alpha_c_p6m": -0.35920394509635845, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [0.1737, 0.1471, 0.1456, 0.1264, 0.1072, 0.1471, 0.1471, 0.1087, 0.1072, 0.0954, 0.1117, 0.1338, 0.1545, 0.2032, 0.2165, 0.2549, 0.2711, 0.249, 0.1914, 0.0807, 0.1176, 0.0895, 0.1501, 0.1752, 0.2047, 0.1811, 0.1973, 0.156, 0.1461, 0.1863, 0.1475, 0.1833, 0.1848, 0.1669, 0.1744, 0.1699, 0.1297, 0.1103, 0.0969, 0.1207, 0.1088, 0.1177, 0.1207, 0.1028, 0.0164, 0.0045, -0.0238, -0.0238, 0.0328, 0.0119, -0.0209, -0.0134, -0.0134, 0.0313, 0.0268, 0.0477, 0.0477, 0.0045, -0.0835, -0.082, -0.0775, -0.0805, -0.0358, 0.0, -0.0402, -0.0611, -0.0611, -0.0864, -0.0864, -0.0313, 0.0, 0.0417, 0.0283, 0.0268, -0.0045, 0.0939, 0.0626, 0.1073, 0.0835, 0.0492, 0.0507, 0.0462, 0.076, -0.0313, -0.0104, -0.0835, -0.1088, -0.1148, -0.1207, -0.073, -0.0209, -0.0149, -0.0194, 0.0194, -0.0492, -0.0551, -0.0596, -0.07, -0.0283, 0.0045, -0.0179, -0.0402, -0.0551, -0.1013, -0.0462, -0.0551, -0.0417, -0.0387, -0.0402, -0.0522, -0.0522, -0.0417, 0.0104, 0.0387, 0.0462, 0.0462, 0.0835, 0.0879, 0.1133, 0.0969, 0.073, 0.0641, 0.082, 0.0954, 0.0313, -0.0104, -0.0045, 0.0238, 0.0119, -0.0343, -0.0238, 0.0224, 0.0492, 0.0224, 0.0045, -0.0253, -0.0686, -0.0641, -0.0104, 0.0089, -0.0328, -0.0715, -0.0209, -0.0164, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, -0.0075, 0.0224, 0.0373, 0.0402, 0.0402, 0.0253, 0.0, -0.07, -0.0253, -0.0134, -0.1073, -0.0879, 0.003, -0.0119, -0.0492, -0.0686, -0.0492, -0.0164, -0.0402, 0.006, -0.0492, -0.073, -0.0611, -0.0864, -0.0939, -0.1803, -0.2593, -0.1863, -0.1937, -0.1937, -0.1937, -0.1937, -0.1475, -0.1371, -0.0984, -0.079, -0.0462, -0.0432]}
{"unit_id": "quote:1987769511672758543", "ticker_idx": 0, "ts": "2025-11-10T06:29:17+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs has doubled its target price for SanDisk, citing strong confidence that the ongoing NAND supply shortage will significantly drive pricing power and profit growth", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/endorses Goldman Sachs doubling SanDisk PT to $280, bullish on NAND supply shortage persisting through 2026 + AI server demand, EPS estimates up 79%.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "Goldman Sachs has doubled its target price for SanDisk, citing strong confidence that the ongoing NAND supply shortage will significantly drive pricing power and profit growth.\n\nThe firm raised its target price to $280 and assessed that the supply shortage in the NAND flash memory market will persist through 2026. This environment grants manufacturers substantial pricing power, which is expected to translate directly into soaring profit margins.\n\nAs a result, Goldman Sachs has revised its EPS forecasts for 2025–2027 upward by an average of 79%.\n\nThe key driver is the tightening of production capacity due to surging demand for AI servers. Goldman Sachs pointed out that market participants have not yet fully priced in the substantial earnings leverage embedded in this supercycle.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": "高盛直接將 SanDisk 目標價翻倍！NAND 供不應求將極大推動定價能力和利潤率增長\n\n高盛將SanDisk目標價上調，稱 NAND 快閃記憶體市場供不應求將持續至 2026 年，賦予製造商強大定價權，並最終體現在利潤率的飆升上，將 2025-2027 年 EPS 預測平均上調 79%，AI", "parent_author": null, "ret_m1d": -0.10625122292397904, "ret_signed_m1d": -0.10625122292397904, "alpha_spy_m1d": -0.09088654524342732, "alpha_c_m1d": -0.08131297184988995, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02493825107408909, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013547207314287046, "ret_signed_p1d": 0.013547207314287046, "alpha_spy_p1d": 0.011258057611841288, "alpha_c_p1d": 0.022074166274562823, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.008526958960275777, "ret_p1w": -0.0077253488707415, "ret_signed_p1w": -0.0077253488707415, "alpha_spy_p1w": 0.015416925030573791, "alpha_c_p1w": 0.032507417113539994, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.04023276598428149, "ret_p1m": -0.18096661050232876, "ret_signed_p1m": -0.18096661050232876, "alpha_spy_p1m": -0.18331440091843632, "alpha_c_p1m": -0.18269246359947766, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.0017258530971488995, "ret_p3m": 1.1504011418436924, "ret_signed_p3m": 1.1504011418436924, "alpha_spy_p3m": 1.1530686187310542, "alpha_c_p3m": 1.2311324843900666, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": -0.08073134254637426, "ret_p6m": 4.248441433928621, "ret_signed_p6m": 4.248441433928621, "alpha_spy_p6m": 4.180275342529457, "alpha_c_p6m": 4.124443957627648, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.12399747630097302, "price_path": [-0.8246, -0.8258, -0.8338, -0.8301, -0.8336, -0.8343, -0.8302, -0.8269, -0.8254, -0.8233, -0.8192, -0.8102, -0.8042, -0.8042, -0.8094, -0.8022, -0.7667, -0.7442, -0.7369, -0.7369, -0.7241, -0.6854, -0.6786, -0.6638, -0.6583, -0.6493, -0.631, -0.6185, -0.6159, -0.6029, -0.6274, -0.6481, -0.6375, -0.5764, -0.5813, -0.548, -0.5367, -0.5208, -0.5478, -0.5486, -0.5078, -0.516, -0.5637, -0.4976, -0.5249, -0.4615, -0.4616, -0.4769, -0.4475, -0.4428, -0.4516, -0.3766, -0.3052, -0.3413, -0.3449, -0.2373, -0.2692, -0.2561, -0.2274, -0.2739, -0.192, -0.2249, -0.1063, 0.0, 0.0135, 0.0565, -0.091, -0.0515, -0.0077, -0.0859, -0.0821, -0.2687, -0.2526, -0.153, -0.1771, -0.1975, -0.1975, -0.1667, -0.2156, -0.2336, -0.2746, -0.2039, -0.1473, -0.1585, -0.181, -0.131, -0.0983, -0.2305, -0.2466, -0.2188, -0.2281, -0.181, -0.1132, -0.1004, -0.086, -0.0667, -0.0667, -0.0668, -0.0884, -0.1035, -0.1141, -0.1141, 0.0272, 0.0229, 0.3048, 0.3195, 0.2485, 0.4085, 0.4528, 0.4548, 0.4473, 0.5273, 0.5436, 0.5436, 0.6911, 0.8708, 0.8789, 0.7684, 0.757, 0.7967, 0.9691, 1.0127, 1.1506, 1.4827, 1.5957, 1.1816, 1.1504, 1.2316, 1.1773, 1.0214, 1.2368, 1.3523, 1.3383, 1.3383, 1.2041, 1.2407, 1.3179, 1.4257, 1.4874, 1.383, 1.3601, 1.4329, 1.3712, 1.3104, 1.1101, 1.2357, 1.1108, 0.968, 1.1972, 1.3097, 1.4461, 1.3095, 1.4692, 1.626, 1.6877, 1.8128, 1.8815, 1.6487, 1.6217, 1.6217, 1.5298, 1.2511, 1.2983, 1.1366, 1.3711, 1.5853, 1.6184, 1.6184, 1.7043, 1.6527, 1.9143, 2.1781, 2.1788, 2.5548, 2.5248, 2.3279, 2.4315, 2.4372, 2.4074, 2.3719, 2.6539, 2.4799, 2.6943, 2.994, 2.7408, 2.9717, 3.0922, 3.4299, 3.6869, 4.2484]}
{"unit_id": "reply:2000164445059010667", "ticker_idx": 0, "ts": "2025-12-14T11:22:19+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm positive on it through the second quarter of 2026.", "tweet_type": "prediction", "timeline": "quarters", "summary": "Bullish on MU (and Samsung) through Q2 2026 on Blackwell/Rubin demand and consumer business exit.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@EdwardDiGi I’m positive on it through the second quarter of 2026.", "reply_text": null, "parent_text": "@jukan05 Yeah that seems the sentiment feeling. But as nobody has modelled Blackwell and Rubin demand, plus H200 and MU and Samsung exiting consumer business \n\nI mean, all this is not mega bullish for 2026?", "quoted_text": null, "parent_author": "EdwardDiGi", "ret_m1d": 0.015326278026813522, "ret_signed_m1d": 0.015326278026813522, "alpha_spy_m1d": 0.013813194179786104, "alpha_c_m1d": 0.01186920406305747, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021010569585278582, "ret_signed_p1d": -0.021010569585278582, "alpha_spy_p1d": -0.018278271989999872, "alpha_c_p1d": -0.018290226336253657, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.1645894394682299, "ret_signed_p1w": 0.1645894394682299, "alpha_spy_p1w": 0.15559371415552747, "alpha_c_p1w": 0.14214894979603554, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4242803264561392, "ret_signed_p1m": 0.4242803264561392, "alpha_spy_p1m": 0.4021128365504283, "alpha_c_p1m": 0.31136370124665924, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.7074262303029928, "ret_signed_p3m": 0.7074262303029928, "alpha_spy_p3m": 0.7260854035676849, "alpha_c_p3m": 0.604174023515615, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3268, -0.2893, -0.3152, -0.3073, -0.2998, -0.3195, -0.3401, -0.3382, -0.3103, -0.2959, -0.2335, -0.2268, -0.2091, -0.196, -0.2181, -0.1725, -0.1902, -0.2354, -0.1883, -0.2124, -0.1918, -0.1472, -0.1479, -0.1294, -0.1483, -0.1643, -0.1296, -0.0778, -0.0733, -0.0656, -0.0458, -0.0568, -0.0578, -0.0118, -0.082, 0.0, 0.0035, 0.0018, 0.0665, 0.0152, 0.0312, -0.0023, 0.0393, 0.0187, -0.0379, -0.0488, -0.1521, -0.1269, -0.0571, -0.0546, -0.0305, -0.0305, -0.0043, 0.0125, 0.0084, -0.0141, -0.0457, -0.0012, 0.0397, 0.0628, 0.1104, 0.0883, 0.0153, 0.0, -0.021, -0.0504, 0.0465, 0.1197, 0.1646, 0.1632, 0.2071, 0.2071, 0.1991, 0.24, 0.2326, 0.2022, 0.2022, 0.3286, 0.3148, 0.4466, 0.4303, 0.3775, 0.4536, 0.4569, 0.4243, 0.4041, 0.418, 0.528, 0.528, 0.5375, 0.639, 0.6747, 0.6834, 0.6389, 0.728, 0.8335, 0.8356, 0.7476, 0.8441, 0.7668, 0.5981, 0.6128, 0.6625, 0.6154, 0.5722, 0.7284, 0.7437, 0.734, 0.734, 0.684, 0.7731, 0.758, 0.8035, 0.7732, 0.7608, 0.807, 0.7504, 0.737, 0.7383, 0.5993, 0.6881, 0.6725, 0.5598, 0.6399, 0.698, 0.7636, 0.7074, 0.795, 0.861, 0.9447, 0.9449, 0.8714, 0.7814, 0.7032, 0.6661, 0.6094, 0.4973, 0.5047, 0.3561, 0.4237, 0.5501, 0.5433, 0.5433, 0.5919, 0.5911, 0.714, 0.7762, 0.7724, 0.7975, 0.9623, 0.9226, 0.9268, 0.9177, 0.8896, 0.8937, 1.0542, 1.03, 1.0932, 1.2105, 1.1251, 1.1848, 1.1793, 1.2849, 1.4292, 1.6978, 1.809, 1.7249, 2.1471, 2.3515, 2.2304, 2.3865, 2.2701, 2.0537, 1.872, 1.9445, 2.0846, 2.2115, 2.1647, 2.1647, 2.7752, 2.9123, 2.8917, 3.0918, null, null, null, null, null, null, null]}
{"unit_id": "thread:2002671016108253209", "ticker_idx": 0, "ts": "2025-12-21T09:22:32+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the gross profit margins of Samsung's memory division and SK hynix in Q4 this year are expected to be around 63–67% respectively, surpassing TSMC", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory semis entering a golden age: Samsung, Hynix, and Micron gross margins set to surpass TSMC in Q4/Q1, driven by AI inference demand surge.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory semiconductors to overtake TSMC in profits\n\nThe profitability of Samsung Electronics’ memory division and SK hynix this year in the fourth quarter will overtake Taiwan’s TSMC, the world’s No. 1 foundry (contract chipmaker). It will be the first time in seven years—since the fourth quarter of 2018—that memory profitability surpasses foundry profitability. As the center of the artificial intelligence (AI) industry shifts from GPU-led “training” to “inference,” where storing data and retrieving it whenever needed becomes crucial, memory demand is surging. Evaluations are emerging that a full-fledged “memory-centric” era—where high-performance DRAM and NAND flash determine the pace of AI advancement—has begun.\n\nAccording to the semiconductor and investment industries on the 21st, the gross profit margins of Samsung’s memory division and SK hynix in Q4 this year are expected to be around 63–67% respectively, surpassing TSMC (company’s official guidance: 60%). Gross profit margin, calculated by dividing gross profit (revenue − cost) by revenue, is considered a key profitability indicator. Micron, the No. 3 memory company in the U.S., officially announced on the 17th that its gross profit margin of 56% in fiscal 2026 first quarter (Sep–Nov) will jump to 67% in the second quarter (Dec–Feb next year). This means that in the first quarter of next year, Micron, too, will surpass TSMC in profitability.\n\nOver the past two to three years, TSMC has taken charge of manufacturing AI accelerators for Nvidia, AMD, Broadcom, and others by leveraging cutting-edge processes at 3 nanometers (nm; 1 nm = one-billionth of a meter) and below, as well as state-of-the-art packaging. As AI accelerators determined AI training performance, TSMC became a main protagonist of the AI semiconductor era alongside Nvidia, which designs GPUs.\n\nThe era of AI inference brings a “golden age of memory” … “$440 billion market next year” Memory gross profit margins surpass \nfoundry\n\n“An ‘Nvidia moment’ (a moment when a specific company reshapes the industry framework, like Nvidia leading the AI semiconductor revolution) has begun for the memory semiconductor industry.”\n\nAfter Micron, the world’s third-largest memory semiconductor company, presented “67%” as its gross profit margin (gross profit ÷ revenue) forecast for fiscal 2026 second quarter (Dec–Feb next year) on the 17th, such reactions emerged in the semiconductor industry. It was forecasting a profit margin comparable to Nvidia (72%), the “lead actor,” by surpassing TSMC (60%), despite having been seen as merely a “supporting actor” in AI. A semiconductor industry official said, “A ‘memory-centric’ era in which memory determines the pace of AI industry development is opening in earnest.”\n\n---\n\nhttps://t.co/4tPX8gocTK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03800910447247974, "ret_signed_m1d": -0.03800910447247974, "alpha_spy_m1d": -0.03181778345714559, "alpha_c_m1d": -0.03264273289239972, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.0053663715800800205, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009049725757364735, "ret_signed_p1d": 0.009049725757364735, "alpha_spy_p1d": 0.004479270938931856, "alpha_c_p1d": 0.0036074705593895917, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.005442255197975143, "ret_p1w": 0.08670506815824552, "ret_signed_p1w": 0.08670506815824552, "alpha_spy_p1w": 0.082295296882958, "alpha_c_p1w": 0.08181389010770301, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.004891178050542511, "ret_p1m": 0.32041476800311086, "ret_signed_p1m": 0.32041476800311086, "alpha_spy_p1m": 0.33100138268862156, "alpha_c_p1m": 0.34328604833761345, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": -0.022871280334502586, "ret_p3m": 0.8233000768945837, "ret_signed_p3m": 0.8233000768945837, "alpha_spy_p3m": 0.8598493272048692, "alpha_c_p3m": 0.8696627572260551, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": -0.046362680331471395, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2306, -0.2243, -0.2495, -0.238, -0.2407, -0.2217, -0.1878, -0.1878, -0.1878, -0.1878, -0.1878, -0.1878, -0.1457, -0.1557, -0.171, -0.1403, -0.1158, -0.114, -0.1122, -0.1176, -0.1077, -0.1267, -0.1059, -0.0769, -0.0995, -0.0905, -0.0579, -0.0271, 0.0054, -0.0507, -0.0896, -0.1023, -0.114, -0.0896, -0.0633, -0.067, -0.0697, -0.1204, -0.0896, -0.1149, -0.1267, -0.0896, -0.1421, -0.1249, -0.1014, -0.0697, -0.0633, -0.0905, -0.0878, -0.0643, -0.0543, -0.0489, -0.019, -0.009, -0.019, -0.0226, -0.029, -0.0145, -0.0516, -0.0697, -0.0235, -0.0262, -0.038, 0.0, 0.009, 0.0054, 0.0054, 0.0588, 0.0867, 0.0903, 0.0903, 0.0903, 0.1685, 0.2558, 0.2631, 0.2822, 0.2622, 0.264, 0.2622, 0.2513, 0.2759, 0.3086, 0.3541, 0.3577, 0.3204, 0.3595, 0.385, 0.3832, 0.3832, 0.4505, 0.4768, 0.4614, 0.4595, 0.3677, 0.5232, 0.5378, 0.4486, 0.4423, 0.5132, 0.5077, 0.5259, 0.6241, 0.6478, 0.6478, 0.6478, 0.6478, 0.7278, 0.7287, 0.7551, 0.8188, 0.8506, 0.9824, 0.9688, 0.9688, 0.7742, 0.5659, 0.7424, 0.7114, 0.5778, 0.7087, 0.7278, 0.7087, 0.6687, 0.716, 0.7633, 0.8961, 0.8233, 0.8133, 0.6942, 0.7251, 0.7187, 0.6378, 0.6378, 0.6065, 0.5236, 0.7282, 0.6257, 0.6968, 0.7596, 0.7906, 0.9182, 0.859, 0.8772, 0.8316, 0.8817, 0.9228, 0.982, 0.9683, 0.9547, 0.9957, 0.982, 1.0458, 1.0002, 1.0458, 1.023, 1.0594, 1.0093, 1.0093, 1.1187, 1.1187, 1.4239, 1.4741, 1.4467, 1.6016, 1.5424, 1.588, 1.6973, 1.465, 1.5606, 1.5105, 1.5151, 1.7292, 1.6654, 1.6654, 1.7247, 1.7976, 1.7292, 1.8887, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2002671016108253209", "ticker_idx": 1, "ts": "2025-12-21T09:22:32+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the gross profit margins of Samsung's memory division and SK hynix in Q4 this year are expected to be around 63–67% respectively, surpassing TSMC", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory semis entering a golden age: Samsung, Hynix, and Micron gross margins set to surpass TSMC in Q4/Q1, driven by AI inference demand surge.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory semiconductors to overtake TSMC in profits\n\nThe profitability of Samsung Electronics’ memory division and SK hynix this year in the fourth quarter will overtake Taiwan’s TSMC, the world’s No. 1 foundry (contract chipmaker). It will be the first time in seven years—since the fourth quarter of 2018—that memory profitability surpasses foundry profitability. As the center of the artificial intelligence (AI) industry shifts from GPU-led “training” to “inference,” where storing data and retrieving it whenever needed becomes crucial, memory demand is surging. Evaluations are emerging that a full-fledged “memory-centric” era—where high-performance DRAM and NAND flash determine the pace of AI advancement—has begun.\n\nAccording to the semiconductor and investment industries on the 21st, the gross profit margins of Samsung’s memory division and SK hynix in Q4 this year are expected to be around 63–67% respectively, surpassing TSMC (company’s official guidance: 60%). Gross profit margin, calculated by dividing gross profit (revenue − cost) by revenue, is considered a key profitability indicator. Micron, the No. 3 memory company in the U.S., officially announced on the 17th that its gross profit margin of 56% in fiscal 2026 first quarter (Sep–Nov) will jump to 67% in the second quarter (Dec–Feb next year). This means that in the first quarter of next year, Micron, too, will surpass TSMC in profitability.\n\nOver the past two to three years, TSMC has taken charge of manufacturing AI accelerators for Nvidia, AMD, Broadcom, and others by leveraging cutting-edge processes at 3 nanometers (nm; 1 nm = one-billionth of a meter) and below, as well as state-of-the-art packaging. As AI accelerators determined AI training performance, TSMC became a main protagonist of the AI semiconductor era alongside Nvidia, which designs GPUs.\n\nThe era of AI inference brings a “golden age of memory” … “$440 billion market next year” Memory gross profit margins surpass \nfoundry\n\n“An ‘Nvidia moment’ (a moment when a specific company reshapes the industry framework, like Nvidia leading the AI semiconductor revolution) has begun for the memory semiconductor industry.”\n\nAfter Micron, the world’s third-largest memory semiconductor company, presented “67%” as its gross profit margin (gross profit ÷ revenue) forecast for fiscal 2026 second quarter (Dec–Feb next year) on the 17th, such reactions emerged in the semiconductor industry. It was forecasting a profit margin comparable to Nvidia (72%), the “lead actor,” by surpassing TSMC (60%), despite having been seen as merely a “supporting actor” in AI. A semiconductor industry official said, “A ‘memory-centric’ era in which memory determines the pace of AI industry development is opening in earnest.”\n\n---\n\nhttps://t.co/4tPX8gocTK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05689661744026986, "ret_signed_m1d": -0.05689661744026986, "alpha_spy_m1d": -0.05070529642493571, "alpha_c_m1d": -0.044177651491505765, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.012718965948764094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006896612041494166, "ret_signed_p1d": 0.006896612041494166, "alpha_spy_p1d": 0.002326157223061287, "alpha_c_p1d": -0.0027224909266945208, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.009619102968188686, "ret_p1w": 0.10344831681829847, "ret_signed_p1w": 0.10344831681829847, "alpha_spy_p1w": 0.09903854554301095, "alpha_c_p1w": 0.0909101486364996, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.01253816818179887, "ret_p1m": 0.2810344572540595, "ret_signed_p1m": 0.2810344572540595, "alpha_spy_p1m": 0.2916210719395702, "alpha_c_p1m": 0.19571335244480492, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": 0.08532110480925459, "ret_p3m": 0.7497745201325205, "ret_signed_p3m": 0.7497745201325205, "alpha_spy_p3m": 0.786323770442806, "alpha_c_p3m": 0.6518595909541705, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": 0.0979149291783501, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3841, -0.3858, -0.4202, -0.3987, -0.4013, -0.3798, -0.3186, -0.3186, -0.3186, -0.3186, -0.3186, -0.3186, -0.2626, -0.285, -0.291, -0.2721, -0.2204, -0.198, -0.1635, -0.1747, -0.1704, -0.1756, -0.1213, -0.0782, -0.1024, -0.0386, -0.0214, -0.0369, 0.0682, 0.0096, -0.0024, 0.0217, -0.0007, 0.0441, 0.0665, 0.063, 0.0544, -0.0352, 0.0441, -0.0179, -0.0317, -0.0162, -0.1024, -0.1041, -0.1058, -0.0972, -0.0621, -0.0862, -0.0724, -0.0379, -0.0483, -0.0655, -0.0621, -0.0052, -0.0241, 0.0121, -0.0259, -0.0155, -0.0448, -0.0862, -0.05, -0.0483, -0.0569, 0.0, 0.0069, 0.0138, 0.0138, 0.0328, 0.1034, 0.1224, 0.1224, 0.1224, 0.1672, 0.2, 0.2517, 0.2793, 0.3034, 0.2828, 0.2914, 0.2724, 0.2793, 0.2914, 0.3034, 0.3172, 0.281, 0.2759, 0.3017, 0.3224, 0.269, 0.3793, 0.45, 0.4845, 0.5672, 0.431, 0.5638, 0.5517, 0.4517, 0.4466, 0.5293, 0.5103, 0.4828, 0.531, 0.5172, 0.5172, 0.5172, 0.5172, 0.5414, 0.6362, 0.6397, 0.7328, 0.7552, 0.8983, 0.8327, 0.8327, 0.622, 0.4665, 0.6254, 0.596, 0.444, 0.6202, 0.6496, 0.6064, 0.5719, 0.6824, 0.6755, 0.824, 0.7498, 0.7394, 0.6116, 0.7031, 0.7187, 0.6116, 0.6116, 0.5079, 0.3939, 0.5494, 0.4337, 0.5131, 0.5304, 0.5822, 0.7843, 0.7239, 0.774, 0.7964, 0.9052, 0.9622, 0.9951, 0.9484, 1.0141, 1.1142, 1.1125, 1.116, 1.1108, 1.2317, 1.2455, 1.2334, 1.2213, 1.2213, 1.4994, 1.4994, 1.7654, 1.857, 1.9123, 2.2474, 2.1696, 2.4132, 2.4028, 2.142, 2.1783, 2.0142, 2.0142, 2.351, 2.3527, 2.3527, 2.5445, 2.8744, 2.9545, 3.0305, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2002671016108253209", "ticker_idx": 2, "ts": "2025-12-21T09:22:32+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron, too, will surpass TSMC in profitability… An 'Nvidia moment' has begun for the memory semiconductor industry", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory semis entering a golden age: Samsung, Hynix, and Micron gross margins set to surpass TSMC in Q4/Q1, driven by AI inference demand surge.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory semiconductors to overtake TSMC in profits\n\nThe profitability of Samsung Electronics’ memory division and SK hynix this year in the fourth quarter will overtake Taiwan’s TSMC, the world’s No. 1 foundry (contract chipmaker). It will be the first time in seven years—since the fourth quarter of 2018—that memory profitability surpasses foundry profitability. As the center of the artificial intelligence (AI) industry shifts from GPU-led “training” to “inference,” where storing data and retrieving it whenever needed becomes crucial, memory demand is surging. Evaluations are emerging that a full-fledged “memory-centric” era—where high-performance DRAM and NAND flash determine the pace of AI advancement—has begun.\n\nAccording to the semiconductor and investment industries on the 21st, the gross profit margins of Samsung’s memory division and SK hynix in Q4 this year are expected to be around 63–67% respectively, surpassing TSMC (company’s official guidance: 60%). Gross profit margin, calculated by dividing gross profit (revenue − cost) by revenue, is considered a key profitability indicator. Micron, the No. 3 memory company in the U.S., officially announced on the 17th that its gross profit margin of 56% in fiscal 2026 first quarter (Sep–Nov) will jump to 67% in the second quarter (Dec–Feb next year). This means that in the first quarter of next year, Micron, too, will surpass TSMC in profitability.\n\nOver the past two to three years, TSMC has taken charge of manufacturing AI accelerators for Nvidia, AMD, Broadcom, and others by leveraging cutting-edge processes at 3 nanometers (nm; 1 nm = one-billionth of a meter) and below, as well as state-of-the-art packaging. As AI accelerators determined AI training performance, TSMC became a main protagonist of the AI semiconductor era alongside Nvidia, which designs GPUs.\n\nThe era of AI inference brings a “golden age of memory” … “$440 billion market next year” Memory gross profit margins surpass \nfoundry\n\n“An ‘Nvidia moment’ (a moment when a specific company reshapes the industry framework, like Nvidia leading the AI semiconductor revolution) has begun for the memory semiconductor industry.”\n\nAfter Micron, the world’s third-largest memory semiconductor company, presented “67%” as its gross profit margin (gross profit ÷ revenue) forecast for fiscal 2026 second quarter (Dec–Feb next year) on the 17th, such reactions emerged in the semiconductor industry. It was forecasting a profit margin comparable to Nvidia (72%), the “lead actor,” by surpassing TSMC (60%), despite having been seen as merely a “supporting actor” in AI. A semiconductor industry official said, “A ‘memory-centric’ era in which memory determines the pace of AI industry development is opening in earnest.”\n\n---\n\nhttps://t.co/4tPX8gocTK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.038576997355407316, "ret_signed_m1d": -0.038576997355407316, "alpha_spy_m1d": -0.032385676340073166, "alpha_c_m1d": -0.025858031406643223, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.012718965948764094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0011570437939640854, "ret_signed_p1d": -0.0011570437939640854, "alpha_spy_p1d": -0.005727498612396964, "alpha_c_p1d": -0.010776146762152772, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.009619102968188686, "ret_p1w": 0.06471284763628149, "ret_signed_p1w": 0.06471284763628149, "alpha_spy_p1w": 0.06030307636099397, "alpha_c_p1w": 0.05217467945448262, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.01253816818179887, "ret_p1m": 0.3201759263113946, "ret_signed_p1m": 0.3201759263113946, "alpha_spy_p1m": 0.3307625409969053, "alpha_c_p1m": 0.23485482150214, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": 0.08532110480925459, "ret_p3m": 0.6068891257649625, "ret_signed_p3m": 0.6068891257649625, "alpha_spy_p3m": 0.6434383760752479, "alpha_c_p3m": 0.5089741965866124, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": 0.0979149291783501, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4157, -0.4333, -0.4318, -0.4078, -0.3954, -0.3419, -0.3361, -0.3209, -0.3096, -0.3286, -0.2894, -0.3046, -0.3434, -0.303, -0.3237, -0.306, -0.2678, -0.2683, -0.2524, -0.2686, -0.2824, -0.2526, -0.2081, -0.2042, -0.1977, -0.1806, -0.1901, -0.191, -0.1515, -0.2117, -0.1413, -0.1383, -0.1398, -0.0842, -0.1283, -0.1146, -0.1433, -0.1076, -0.1252, -0.1739, -0.1832, -0.272, -0.2503, -0.1904, -0.1882, -0.1675, -0.1675, -0.145, -0.1306, -0.1341, -0.1534, -0.1806, -0.1423, -0.1073, -0.0874, -0.0466, -0.0655, -0.1282, -0.1413, -0.1594, -0.1846, -0.1014, -0.0386, 0.0, -0.0012, 0.0365, 0.0365, 0.0296, 0.0647, 0.0584, 0.0323, 0.0323, 0.1408, 0.129, 0.2422, 0.2281, 0.1828, 0.2482, 0.251, 0.223, 0.2057, 0.2176, 0.312, 0.312, 0.3202, 0.4074, 0.438, 0.4455, 0.4073, 0.4838, 0.5744, 0.5762, 0.5006, 0.5835, 0.5171, 0.3723, 0.3849, 0.4276, 0.3871, 0.35, 0.4842, 0.4973, 0.4889, 0.4889, 0.446, 0.5225, 0.5095, 0.5487, 0.5226, 0.5119, 0.5517, 0.503, 0.4915, 0.4926, 0.3733, 0.4496, 0.4361, 0.3393, 0.4081, 0.458, 0.5144, 0.4661, 0.5413, 0.598, 0.6699, 0.67, 0.6069, 0.5296, 0.4625, 0.4306, 0.382, 0.2857, 0.292, 0.1644, 0.2225, 0.331, 0.3252, 0.3252, 0.3669, 0.3663, 0.4717, 0.5252, 0.5219, 0.5435, 0.685, 0.6508, 0.6545, 0.6466, 0.6226, 0.6261, 0.7639, 0.7431, 0.7974, 0.8981, 0.8247, 0.876, 0.8713, 0.962, 1.0858, 1.3165, 1.412, 1.3398, 1.7023, 1.8779, 1.7738, 1.9079, 1.8079, 1.6221, 1.4661, 1.5283, 1.6487, 1.7576, 1.7174, 1.7174, 2.2417, 2.3594, 2.3417, 2.5135, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1975207274437370228", "ticker_idx": 0, "ts": "2025-10-06T14:31:26+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We see substantial upside potential in NVIDIA's 2026 estimates and consider the risk/reward attractive... we raise our estimates and price target within this report. 12-Month Price Target: $210.00", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman Sachs raises NVDA PT to $210, sees substantial 2026 upside; risk/reward attractive despite circular-revenue concerns.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) NVIDIA: Impact of OpenAI and Strategic Investments on Business\n\n12-Month Price Target: $210.00\n\nNVIDIA has announced a series of strategic investments and partnerships with companies including OpenAI. This announcement has sparked investor debate over the nature of these transactions — particularly the extent to which NVIDIA’s equity investments in partner firms may be reused by those firms to purchase NVIDIA GPUs, allowing NVIDIA to recognize this as “circular revenue.”\n\nThe sustainability of OpenAI’s infrastructure spending will increasingly depend on equity and debt financing. Considering OpenAI’s publicly stated plans for infrastructure investment — including building its own data centers using NVIDIA GPUs, funding the Stargate project, and its Oracle partnership — the firm’s reliance on capital raising is likely to rise significantly, potentially reaching up to $75 billion in 2026 alone. Thus far, such initiatives (e.g., Oracle’s $18 billion debt financing to support OpenAI) have met strong market demand.\n\nWe view potential “circular revenue” from these strategic investments as a factor that could dilute NVIDIA’s valuation multiple. As with any strategic investment in customers with limited self-funding capacity, investors are likely to focus on the fundamentals of the investee companies. When the investor is also the supplier, dual roles can create a circular nature of revenue, warranting closer scrutiny.\n\nHowever, we also believe NVIDIA’s investments are strategically significant as ecosystem-building moves. Specifically, NVIDIA’s actions have clear strategic merit in the following ways:\n\n1. They strengthen the CUDA software ecosystem among developers.\n\n2. They act as a signaling mechanism to other investors, conveying NVIDIA’s conviction in the size of the market opportunity.\n\nWe see substantial upside potential in NVIDIA’s 2026 estimates and consider the risk/reward attractive, given that “circular” transactions are expected to account for less than 15% of projected 2027 revenue. Alongside strong growth momentum from core customers, NVIDIA also benefits from incremental upside driven by nontraditional clients. Accordingly, we raise our estimates and price target within this report.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011210528125265684, "ret_signed_m1d": 0.011210528125265684, "alpha_spy_m1d": 0.014783941955614721, "alpha_c_m1d": 0.030722476669672427, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.019511948544406743, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002694803694029413, "ret_signed_p1d": -0.002694803694029413, "alpha_spy_p1d": 0.001012688556932484, "alpha_c_p1d": 0.015739581411471026, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.01843438510550044, "ret_p1w": 0.014983368445398382, "ret_signed_p1w": 0.014983368445398382, "alpha_spy_p1w": 0.02774375888322156, "alpha_c_p1w": 0.025554758783950526, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.010571390338552145, "ret_p1m": 0.07087422544133548, "ret_signed_p1m": 0.07087422544133548, "alpha_spy_p1m": 0.06546922909627484, "alpha_c_p1m": 0.04355749747916593, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.02731672796216955, "ret_p3m": 0.005230126727494788, "ret_signed_p3m": 0.005230126727494788, "alpha_spy_p3m": -0.013121342989518503, "alpha_c_p3m": -0.04681305818522641, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.052043184912721197, "ret_p6m": -0.05993775072555885, "ret_signed_p6m": -0.05993775072555885, "alpha_spy_p6m": -0.03378130835341453, "alpha_c_p6m": -0.1799592574190313, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.12002150669347245, "price_path": [-0.1222, -0.1156, -0.1112, -0.1158, -0.08, -0.0764, -0.0676, -0.0708, -0.0764, -0.0998, -0.0796, -0.0637, -0.0649, -0.0474, -0.0541, -0.0338, -0.0414, -0.0638, -0.0299, -0.0393, -0.033, -0.0258, -0.0154, -0.0188, -0.0129, -0.0213, -0.019, -0.0275, -0.0191, -0.0534, -0.0547, -0.057, -0.0407, -0.0309, -0.0204, -0.0213, -0.029, -0.0613, -0.0613, -0.0796, -0.0805, -0.0749, -0.0999, -0.0929, -0.0797, -0.0443, -0.0451, -0.0416, -0.042, -0.0575, -0.0822, -0.0501, -0.0478, -0.0104, -0.0383, -0.0462, -0.0423, -0.0396, -0.0199, 0.0056, 0.0092, 0.0181, 0.0112, 0.0, -0.0027, 0.0192, 0.0379, -0.0128, 0.015, -0.0297, -0.0308, -0.0201, -0.0125, -0.0156, -0.0236, -0.0283, -0.0182, 0.0039, 0.0321, 0.0835, 0.1159, 0.0935, 0.0914, 0.115, 0.0709, 0.0521, 0.0137, 0.0141, 0.0728, 0.0411, 0.0445, 0.0071, 0.025, 0.0057, -0.0225, 0.0053, -0.0264, -0.0359, -0.0161, -0.0416, -0.0285, -0.0285, -0.046, -0.0303, -0.022, -0.0321, -0.0116, -0.0168, 0.0001, -0.003, -0.0094, -0.0248, -0.0566, -0.0498, -0.0421, -0.0786, -0.0614, -0.0245, -0.0099, 0.0198, 0.0166, 0.0166, 0.027, 0.0145, 0.0108, 0.0052, 0.0052, 0.0179, 0.014, 0.0092, 0.0193, -0.0026, -0.0036, -0.0032, 0.0015, -0.0129, 0.0082, 0.0038, 0.0038, -0.0402, -0.0119, -0.0037, 0.0115, 0.0051, 0.0161, 0.0323, 0.0376, 0.0302, 0.0004, -0.028, -0.0611, -0.0736, -0.0006, 0.0243, 0.0162, 0.0244, 0.0076, -0.0147, -0.0147, -0.003, 0.0132, 0.0128, 0.0231, 0.0324, 0.0395, 0.0541, -0.0034, -0.045, -0.0164, -0.0295, -0.0134, -0.0118, -0.0416, -0.0155, -0.0041, 0.0028, -0.0128, -0.0284, -0.0124, -0.0193, -0.0276, -0.0375, -0.0691, -0.0533, -0.0556, -0.0369, -0.077, -0.097, -0.1097, -0.0599]}
{"unit_id": "quote:1976536147049513002", "ticker_idx": 0, "ts": "2025-10-10T06:31:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'd say there's still substantial upside left for NVIDIA", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Author believes UBS's 678K CoWoS wafer estimate is too conservative vs supply-chain checks; sees substantial upside for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "UBS’s estimate of 678,000 wafers still looks conservative compared to what I’ve heard from the supply chain…\n\nI’d say there’s still substantial upside left for NVIDIA. https://t.co/ry47yjaqXf", "reply_text": null, "parent_text": null, "quoted_text": "UBS Raises CoWoS Demand Forecasts for NVIDIA and Broadcom, With Rubin and CPX as Key Growth Drivers\n\nUBS (T. Arcuri, Oct 8, 2025)\n\nUBS has raised its CoWoS demand forecasts for NVIDIA and Broadcom in its latest report. The firm had previously revised TSMC’s CoWoS capacity https://t.co/hpDhkezpdG", "parent_author": null, "ret_m1d": 0.05137584134182771, "ret_signed_m1d": 0.05137584134182771, "alpha_spy_m1d": 0.023597217580441754, "alpha_c_m1d": -0.009698681846317392, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02817212030804872, "ret_signed_p1d": 0.02817212030804872, "alpha_spy_p1d": 0.012828045614839967, "alpha_c_p1d": -0.016119988067143876, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.0003276050769120964, "ret_signed_p1w": 0.0003276050769120964, "alpha_spy_p1w": -0.017083865558432887, "alpha_c_p1w": -0.05343148722476987, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.0867547734359273, "ret_signed_p1m": 0.0867547734359273, "alpha_spy_p1m": 0.04323380469821747, "alpha_c_p1m": -0.016060782030970833, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.03254271530285635, "ret_signed_p3m": 0.03254271530285635, "alpha_spy_p3m": -0.026563715450941272, "alpha_c_p3m": -0.15457850936422024, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": -0.030031104152419386, "ret_signed_p6m": -0.030031104152419386, "alpha_spy_p6m": -0.04482688746932173, "alpha_c_p6m": -0.25094505407484824, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.0681, -0.0644, -0.0555, -0.0587, -0.0644, -0.0881, -0.0676, -0.0515, -0.0528, -0.0351, -0.0418, -0.0213, -0.0289, -0.0516, -0.0173, -0.0268, -0.0205, -0.0131, -0.0026, -0.0061, -0.0001, -0.0086, -0.0063, -0.0149, -0.0063, -0.0411, -0.0424, -0.0447, -0.0283, -0.0183, -0.0076, -0.0086, -0.0164, -0.0491, -0.0491, -0.0676, -0.0685, -0.0628, -0.0882, -0.0811, -0.0678, -0.0319, -0.0327, -0.0292, -0.0295, -0.0452, -0.0703, -0.0378, -0.0354, 0.0025, -0.0258, -0.0338, -0.0299, -0.0271, -0.0072, 0.0187, 0.0223, 0.0313, 0.0244, 0.013, 0.0103, 0.0325, 0.0514, 0.0, 0.0282, -0.0171, -0.0182, -0.0074, 0.0003, -0.0028, -0.0109, -0.0157, -0.0055, 0.0169, 0.0455, 0.0976, 0.1304, 0.1077, 0.1055, 0.1295, 0.0848, 0.0658, 0.0269, 0.0272, 0.0868, 0.0546, 0.0581, 0.0202, 0.0383, 0.0188, -0.0098, 0.0183, -0.0138, -0.0234, -0.0033, -0.0292, -0.0158, -0.0158, -0.0336, -0.0177, -0.0093, -0.0195, 0.0013, -0.004, 0.0131, 0.0099, 0.0034, -0.0121, -0.0444, -0.0375, -0.0296, -0.0667, -0.0492, -0.0118, 0.0029, 0.0331, 0.0298, 0.0298, 0.0403, 0.0277, 0.024, 0.0183, 0.0183, 0.0311, 0.0271, 0.0223, 0.0325, 0.0103, 0.0093, 0.0098, 0.0145, -0.0001, 0.0213, 0.0168, 0.0168, -0.0277, 0.0009, 0.0092, 0.0247, 0.0181, 0.0293, 0.0457, 0.0511, 0.0436, 0.0134, -0.0153, -0.0489, -0.0615, 0.0123, 0.0376, 0.0294, 0.0377, 0.0207, -0.0019, -0.0019, 0.0099, 0.0264, 0.0259, 0.0364, 0.0459, 0.053, 0.0678, 0.0095, -0.0325, -0.0037, -0.0169, -0.0006, 0.001, -0.0291, -0.0027, 0.0088, 0.0158, 0.0, -0.0158, 0.0004, -0.0066, -0.015, -0.025, -0.057, -0.041, -0.0434, -0.0244, -0.065, -0.0853, -0.0981, -0.0477, -0.0404, -0.0314, -0.0314, -0.03]}
{"unit_id": "orig:1977271008710279458", "ticker_idx": 0, "ts": "2025-10-12T07:11:59+00:00", "symbol": "gold", "kind": "commodity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If History Repeats, Gold Could Reach $6,000 Next Spring", "tweet_type": "news_reaction", "timeline": "months", "summary": "Relays BofA Hartnett's bullish gold call: $6,000 by next spring on dollar devaluation, Fed chair change, and historical 300% bull-market pattern.", "resolved_tickers": ["GLD"], "resolution_method": "etf", "resolution_reason": "Gold commodity proxied by GLD ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "BofA Hartnett: “If History Repeats, Gold Could Reach $6,000 Next Spring”\n\n• Hartnett noted that although gold prices surpassed $4,000 last week, there are still very few long-term investors betting on gold. According to his calculations, gold accounts for only 0.5% of personal assets and 2.3% of institutional assets among Bank of America clients.\n\n• He argued that the early nomination of a new Fed Chair, stimulus and asset-bubble-style policies (citing Argentina’s bailout case), and the coming revaluation of gold (as seen in 1934 and 1973) could all act as favorable catalysts for a U.S. dollar devaluation (“devalue trade”).\n\n• Hartnett stated, “History doesn’t guarantee the future, but in the past four gold bull markets, prices rose an average of 300% over about 43 months.” If the same pattern repeats, he projected that gold prices could reach $6,000 by next spring.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023724513495688604, "ret_signed_m1d": -0.023724513495688604, "alpha_spy_m1d": -0.008612321406671897, "alpha_c_m1d": -0.008612321406671897, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.015112192089016707, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007141189222626965, "ret_signed_p1d": 0.007141189222626965, "alpha_spy_p1d": 0.008362844026083671, "alpha_c_p1d": 0.008362844026083671, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.0012216548034567065, "ret_p1w": 0.06628050940601504, "ret_signed_p1w": 0.06628050940601504, "alpha_spy_p1w": 0.05382270398907396, "alpha_c_p1w": 0.05382270398907396, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.012457805416941081, "ret_p1m": 0.004707870603659403, "ret_signed_p1m": 0.004707870603659403, "alpha_spy_p1m": -0.025395884883187003, "alpha_c_p1m": -0.025395884883187003, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.030103755486846406, "ret_p3m": 0.08833874003541609, "ret_signed_p3m": 0.08833874003541609, "alpha_spy_p3m": 0.04534363794910434, "alpha_c_p3m": 0.04534363794910434, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.04299510208631174, "ret_p6m": 0.14208257753716058, "ret_signed_p6m": 0.14208257753716058, "alpha_spy_p6m": 0.1421827465654525, "alpha_c_p6m": 0.1421827465654525, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": -0.00010016902829190766, "price_path": [-0.1847, -0.1865, -0.1843, -0.1718, -0.164, -0.1743, -0.1794, -0.187, -0.1922, -0.19, -0.204, -0.1987, -0.1824, -0.1777, -0.177, -0.1788, -0.1718, -0.172, -0.1839, -0.1847, -0.1822, -0.1874, -0.1869, -0.1882, -0.1926, -0.1844, -0.1873, -0.1786, -0.1805, -0.1746, -0.1729, -0.1668, -0.1587, -0.1587, -0.1389, -0.1321, -0.1359, -0.1244, -0.1144, -0.1165, -0.1133, -0.1146, -0.1129, -0.1036, -0.1018, -0.1088, -0.1123, -0.1029, -0.0874, -0.0837, -0.092, -0.0882, -0.0829, -0.0678, -0.0598, -0.0583, -0.0616, -0.0541, -0.0363, -0.0313, -0.0153, -0.0335, -0.0237, 0.0, 0.0071, 0.0246, 0.0486, 0.0288, 0.0663, -0.0022, -0.0021, 0.0019, -0.0015, -0.0293, -0.0363, -0.0399, -0.0211, -0.0264, -0.0246, -0.0417, -0.0306, -0.0318, -0.0259, 0.0008, 0.0047, 0.0209, 0.0126, -0.0056, -0.017, -0.0099, -0.0083, -0.0086, -0.0101, 0.0056, 0.0053, 0.0133, 0.0133, 0.0259, 0.0308, 0.0242, 0.0232, 0.0239, 0.0221, 0.0194, 0.0246, 0.029, 0.0401, 0.0459, 0.0468, 0.0471, 0.0561, 0.0542, 0.0554, 0.0797, 0.094, 0.0895, 0.0895, 0.1022, 0.0542, 0.055, 0.0482, 0.0482, 0.0534, 0.0811, 0.0928, 0.0824, 0.0883, 0.0962, 0.1167, 0.1152, 0.1266, 0.1197, 0.1143, 0.1143, 0.1564, 0.1733, 0.1949, 0.2114, 0.2291, 0.2592, 0.308, 0.3116, 0.1768, 0.1297, 0.2015, 0.2007, 0.1687, 0.2046, 0.2352, 0.223, 0.2368, 0.1939, 0.2236, 0.2236, 0.1854, 0.2121, 0.2155, 0.2394, 0.2729, 0.2553, 0.2521, 0.2629, 0.2795, 0.296, 0.2382, 0.2479, 0.2329, 0.2524, 0.2498, 0.2639, 0.2596, 0.2348, 0.2189, 0.2178, 0.2147, 0.1763, 0.1278, 0.0933, 0.0686, 0.0689, 0.101, 0.0596, 0.0968, 0.0965, 0.1381, 0.158, 0.1357, 0.1357, 0.1311, 0.1421]}
{"unit_id": "orig:1980508499441549804", "ticker_idx": 0, "ts": "2025-10-21T05:36:36+00:00", "symbol": "HBM memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "demand for HBM (high-bandwidth memory) is surging, and the report projects that global HBM consumption will reach 26 billion GB in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: chip/packaging bottlenecks resolved; investment shifts to downstream HBM, power, cooling, racks, and networking; NVDA dominates AI demand; foundries/CoWoS de-rated as growth drivers.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM memory basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "OCP Conference Focus: Foundries and CoWoS Are No Longer the Bottleneck; AI Chip Bottlenecks Lie in Downstream Value Chains Such as Memory, Racks, and Power\n\nThe AI semiconductor industry will see another year of strong growth in 2026, but the investment logic in AI hardware is undergoing a profound shift.\n\nOn October 20, a Morgan Stanley report pointed out that over the past two years the market’s focus had been concentrated on upstream capacity bottlenecks such as TSMC’s CoWoS packaging and leading-edge process nodes.\n\nHowever, according to the latest remarks from NVIDIA and TSMC and the signals from the 2025 OCP Conference, this phase has already changed. Through large-scale expansions, the chip manufacturing and packaging stages are no longer the key factors constraining AI progress.\n\nMorgan Stanley emphasized that the real bottlenecks are moving downstream and are concentrating in the associated infrastructure—data center space, power supply, liquid cooling, high-bandwidth memory (HBM), server racks, and optical modules.\n\nThe report sees this as meaning that investment opportunities are spreading from upstream foundries and packaging to a broader downstream supply chain. Going forward, data centers that cannot secure sufficient power and physical space will become laggards in the competition for AI compute.\n\nUpstream Capacity Is No Longer the Sole Bottleneck; Chip Manufacturing and Packaging Have Already Expanded Significantly\n\nThere was a time when there were major doubts about whether AI chips could be supplied in sufficient quantity, and TSMC’s advanced CoWoS packaging capability was regarded as the critical bottleneck. But the latest industry trends show that the situation has changed markedly.\n\nOn its recent earnings conference call, TSMC stated that “AI demand is stronger than we imagined three months ago,” and said the company is working to “narrow the demand–supply gap.” Crucially, TSMC said the lead time to expand CoWoS capacity is only six months, which gives the supply side very significant flexibility.\n\nAlthough wafer front-end capacity at leading nodes such as 4 nm and 3 nm remains tight, AI semiconductors clearly have higher priority than cryptocurrency ASICs or Android smartphone SoCs.\n\nNVIDIA CEO Jensen Huang likewise made it clear in recent exchanges that semiconductor capacity is not the limiting factor it once was. After the surge in demand a few years ago, the manufacturing and packaging stages across the supply chain have “expanded dramatically,” and the company is confident it can meet customer demand.\n\nIn sum, even as overall demand continues to grow rapidly, global CoWoS demand in 2026 is projected to reach 1.154 million wafer starts, up 70% year over year, and the supply side’s ability to respond quickly has been noticeably strengthened.\n\nBottlenecks Shift: Downstream Infrastructure Emerges as the New Challenge\n\nIf chip supply is no longer the toughest problem, the bottlenecks naturally move downstream.\n\nNVIDIA pointed out that the bigger constraints now stem from the availability of data center space, power, and related infrastructure, and that the build cycles in these areas are far longer than those for chip manufacturing.\n\nThe OCP Conference contents corroborate this. As AI clusters move toward “hundreds-of-thousands-GPU” scale, data center design philosophy is being re-architected:\n\nPower and Cooling:\nDeploying large-scale GPU clusters entails enormous power consumption and heat/cooling challenges. At the OCP Conference, liquid cooling has become the default configuration for new AI racks, and demand for power-delivery solutions such as high-voltage DC (HVDC 800 V) is steadily increasing.\nThis benefits companies such as Aspeed, whose BMCs (baseboard management controllers) are expanding beyond servers to a wide range of equipment including cooling.\n\nStorage and Memory:\nAI workloads impose extreme demands on data storage and access speed. Meta has made it clear that, for cost reasons, it will prioritize QLC NAND flash in its data centers. At the same time, Seagate noted that HDDs (mechanical hard drives) will still keep about 95% of total capacity online to meet the needs of large and remote data centers.\nAbove all, demand for HBM (high-bandwidth memory) is surging, and the report projects that global HBM consumption will reach 26 billion GB in 2026, with NVIDIA alone consuming 54% of it. Such highly concentrated, intense demand makes HBM supply a key variable affecting AI server shipments.\n\nRacks and Networking:\nTo realize ultra-large deployments, OCP has put forward standardized blueprints such as the “AI Open Data Center” and “AI Open Cluster Design,” covering racks, liquid cooling, power interfaces, and more.\nOn the networking side, Alibaba stated that pluggable optics remain the best choice due to total cost of ownership and flexibility, and technologies such as LPO (linear-drive pluggable optics) are also drawing attention.\nCPO/NPO are expected to be realized in 2028 as manufacturing processes mature.\n\nDemand Outlook Points to Explosive Growth in Downstream Components\nThe importance of downstream infrastructure is validated by demand forecast data.\n\nMorgan Stanley analysts expect global cloud services capex to increase 31% year over year in 2026 to $582 billion, far exceeding the market’s common expectation of 16%.\n\nAssuming a higher capex mix for AI servers, this implies that AI server capex could grow about 70% year over year in 2026.\n\nOn the demand side, major AI big techs continue “stockpiling.” The report breaks down AI chip demand in 2026 in detail:\n\nCoWoS capacity consumption: NVIDIA 59%, Broadcom 18%, AMD 9%, AWS 6%.\n\nAI compute wafer consumption: NVIDIA leads with 55%, followed by Google 22%, AMD 6%, and AWS 6%.\n\nIn summary, the signals conveyed by the OCP Conference and cross-validation with industry-chain data clearly point to a new direction for AI hardware investment. As capacity bottlenecks at the chip manufacturing and packaging stages are gradually resolved, the market’s focus will inevitably shift to the infrastructure that supports ultra-large-scale AI compute.\n\nFor investors, the report sees this as a call to broaden their view from individual chip companies to the entire data center ecosystem and to find the “pick-and-shovel” players with core competitiveness in downstream stages such as power, cooling, storage, memory, and networking.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013956793581783552, "ret_signed_m1d": 0.013956793581783552, "alpha_spy_m1d": 0.013941797436890132, "alpha_c_m1d": 0.00839635806634583, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0007941345459554124, "ret_signed_p1d": -0.0007941345459554124, "alpha_spy_p1d": 0.004404681416965299, "alpha_c_p1d": 0.01860923214171654, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.06839506472916757, "ret_signed_p1w": 0.06839506472916757, "alpha_spy_p1w": 0.04490287223426259, "alpha_c_p1w": 0.016816706029302297, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.0932779612689435, "ret_signed_p1m": 0.0932779612689435, "alpha_spy_p1m": 0.10617830347365476, "alpha_c_p1m": 0.11024863741861146, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.6359880167447961, "ret_signed_p3m": 0.6359880167447961, "alpha_spy_p3m": 0.6025988258197316, "alpha_c_p3m": 0.47283746263392146, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 1.27133163964156, "ret_signed_p6m": 1.27133163964156, "alpha_spy_p6m": 1.222715509495505, "alpha_c_p6m": 0.9553467705511627, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.4041, -0.4076, -0.3951, -0.3929, -0.3808, -0.3872, -0.4135, -0.406, -0.3994, -0.4071, -0.3938, -0.3816, -0.3691, -0.3607, -0.3575, -0.3578, -0.3651, -0.3724, -0.378, -0.3895, -0.3988, -0.3888, -0.3846, -0.3872, -0.3851, -0.3754, -0.3796, -0.3958, -0.3884, -0.3843, -0.3721, -0.3568, -0.3522, -0.3335, -0.3108, -0.2886, -0.2558, -0.2495, -0.226, -0.2383, -0.2012, -0.2113, -0.2004, -0.1865, -0.1942, -0.2006, -0.2233, -0.1994, -0.1958, -0.1555, -0.1153, -0.1084, -0.1033, -0.112, -0.0941, -0.101, -0.0802, -0.0746, -0.0922, -0.0649, -0.0174, -0.0079, 0.014, 0.0, -0.0008, 0.0035, 0.0536, 0.0837, 0.0684, 0.1053, 0.1203, 0.1253, 0.198, 0.1257, 0.1382, 0.1445, 0.1304, 0.183, 0.1819, 0.1854, 0.1678, 0.1287, 0.1643, 0.1075, 0.0933, 0.0731, 0.0284, 0.0615, 0.0706, 0.0955, 0.1121, 0.1023, 0.1155, 0.1367, 0.1275, 0.1102, 0.1403, 0.1831, 0.1807, 0.2126, 0.1862, 0.1673, 0.1354, 0.1037, 0.1242, 0.1618, 0.1825, 0.2375, 0.2431, 0.2617, 0.2617, 0.2864, 0.3415, 0.3476, 0.3357, 0.3357, 0.4329, 0.4737, 0.5489, 0.5609, 0.5424, 0.5645, 0.5686, 0.544, 0.5482, 0.5709, 0.636, 0.6429, 0.6179, 0.6704, 0.7044, 0.7155, 0.6765, 0.7813, 0.8611, 0.8701, 0.8683, 0.8164, 0.8985, 0.8331, 0.7648, 0.7798, 0.8215, 0.7949, 0.8518, 0.9144, 0.9139, 0.9139, 0.8943, 0.9292, 0.9633, 1.0198, 1.0193, 1.076, 1.1152, 1.2007, 1.1638, 1.1643, 0.9513, 0.8446, 0.9693, 0.9016, 0.8211, 0.9645, 1.0092, 0.9626, 0.9677, 1.0561, 1.104, 1.2142, 1.1279, 1.0847, 0.9575, 0.9916, 0.9733, 0.8556, 0.8585, 0.7467, 0.6958, 0.8853, 0.7972, 0.8562, 0.9059, 0.9382, 1.1161, 1.0937, 1.1193, 1.121, 1.2484, 1.2713]}
{"unit_id": "orig:1980508499441549804", "ticker_idx": 1, "ts": "2025-10-21T05:36:36+00:00", "symbol": "power and cooling", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the real bottlenecks are moving downstream and are concentrating in the associated infrastructure—data center space, power supply, liquid cooling, high-bandwidth memory (HBM), server racks, and optical modules", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: chip/packaging bottlenecks resolved; investment shifts to downstream HBM, power, cooling, racks, and networking; NVDA dominates AI demand; foundries/CoWoS de-rated as growth drivers.", "resolved_tickers": ["VRT", "SMCI", "GRMN"], "resolution_method": "basket", "resolution_reason": "Power and cooling basket: Vertiv, SMCI; no single ETF", "bench_c": "XLI", "bench_c_reason": "industry='Electrical Equipment & Parts'", "bench_c_industry": "Electrical Equipment & Parts", "unit_kind": "original", "lang": "en", "tweet_text": "OCP Conference Focus: Foundries and CoWoS Are No Longer the Bottleneck; AI Chip Bottlenecks Lie in Downstream Value Chains Such as Memory, Racks, and Power\n\nThe AI semiconductor industry will see another year of strong growth in 2026, but the investment logic in AI hardware is undergoing a profound shift.\n\nOn October 20, a Morgan Stanley report pointed out that over the past two years the market’s focus had been concentrated on upstream capacity bottlenecks such as TSMC’s CoWoS packaging and leading-edge process nodes.\n\nHowever, according to the latest remarks from NVIDIA and TSMC and the signals from the 2025 OCP Conference, this phase has already changed. Through large-scale expansions, the chip manufacturing and packaging stages are no longer the key factors constraining AI progress.\n\nMorgan Stanley emphasized that the real bottlenecks are moving downstream and are concentrating in the associated infrastructure—data center space, power supply, liquid cooling, high-bandwidth memory (HBM), server racks, and optical modules.\n\nThe report sees this as meaning that investment opportunities are spreading from upstream foundries and packaging to a broader downstream supply chain. Going forward, data centers that cannot secure sufficient power and physical space will become laggards in the competition for AI compute.\n\nUpstream Capacity Is No Longer the Sole Bottleneck; Chip Manufacturing and Packaging Have Already Expanded Significantly\n\nThere was a time when there were major doubts about whether AI chips could be supplied in sufficient quantity, and TSMC’s advanced CoWoS packaging capability was regarded as the critical bottleneck. But the latest industry trends show that the situation has changed markedly.\n\nOn its recent earnings conference call, TSMC stated that “AI demand is stronger than we imagined three months ago,” and said the company is working to “narrow the demand–supply gap.” Crucially, TSMC said the lead time to expand CoWoS capacity is only six months, which gives the supply side very significant flexibility.\n\nAlthough wafer front-end capacity at leading nodes such as 4 nm and 3 nm remains tight, AI semiconductors clearly have higher priority than cryptocurrency ASICs or Android smartphone SoCs.\n\nNVIDIA CEO Jensen Huang likewise made it clear in recent exchanges that semiconductor capacity is not the limiting factor it once was. After the surge in demand a few years ago, the manufacturing and packaging stages across the supply chain have “expanded dramatically,” and the company is confident it can meet customer demand.\n\nIn sum, even as overall demand continues to grow rapidly, global CoWoS demand in 2026 is projected to reach 1.154 million wafer starts, up 70% year over year, and the supply side’s ability to respond quickly has been noticeably strengthened.\n\nBottlenecks Shift: Downstream Infrastructure Emerges as the New Challenge\n\nIf chip supply is no longer the toughest problem, the bottlenecks naturally move downstream.\n\nNVIDIA pointed out that the bigger constraints now stem from the availability of data center space, power, and related infrastructure, and that the build cycles in these areas are far longer than those for chip manufacturing.\n\nThe OCP Conference contents corroborate this. As AI clusters move toward “hundreds-of-thousands-GPU” scale, data center design philosophy is being re-architected:\n\nPower and Cooling:\nDeploying large-scale GPU clusters entails enormous power consumption and heat/cooling challenges. At the OCP Conference, liquid cooling has become the default configuration for new AI racks, and demand for power-delivery solutions such as high-voltage DC (HVDC 800 V) is steadily increasing.\nThis benefits companies such as Aspeed, whose BMCs (baseboard management controllers) are expanding beyond servers to a wide range of equipment including cooling.\n\nStorage and Memory:\nAI workloads impose extreme demands on data storage and access speed. Meta has made it clear that, for cost reasons, it will prioritize QLC NAND flash in its data centers. At the same time, Seagate noted that HDDs (mechanical hard drives) will still keep about 95% of total capacity online to meet the needs of large and remote data centers.\nAbove all, demand for HBM (high-bandwidth memory) is surging, and the report projects that global HBM consumption will reach 26 billion GB in 2026, with NVIDIA alone consuming 54% of it. Such highly concentrated, intense demand makes HBM supply a key variable affecting AI server shipments.\n\nRacks and Networking:\nTo realize ultra-large deployments, OCP has put forward standardized blueprints such as the “AI Open Data Center” and “AI Open Cluster Design,” covering racks, liquid cooling, power interfaces, and more.\nOn the networking side, Alibaba stated that pluggable optics remain the best choice due to total cost of ownership and flexibility, and technologies such as LPO (linear-drive pluggable optics) are also drawing attention.\nCPO/NPO are expected to be realized in 2028 as manufacturing processes mature.\n\nDemand Outlook Points to Explosive Growth in Downstream Components\nThe importance of downstream infrastructure is validated by demand forecast data.\n\nMorgan Stanley analysts expect global cloud services capex to increase 31% year over year in 2026 to $582 billion, far exceeding the market’s common expectation of 16%.\n\nAssuming a higher capex mix for AI servers, this implies that AI server capex could grow about 70% year over year in 2026.\n\nOn the demand side, major AI big techs continue “stockpiling.” The report breaks down AI chip demand in 2026 in detail:\n\nCoWoS capacity consumption: NVIDIA 59%, Broadcom 18%, AMD 9%, AWS 6%.\n\nAI compute wafer consumption: NVIDIA leads with 55%, followed by Google 22%, AMD 6%, and AWS 6%.\n\nIn summary, the signals conveyed by the OCP Conference and cross-validation with industry-chain data clearly point to a new direction for AI hardware investment. As capacity bottlenecks at the chip manufacturing and packaging stages are gradually resolved, the market’s focus will inevitably shift to the infrastructure that supports ultra-large-scale AI compute.\n\nFor investors, the report sees this as a call to broaden their view from individual chip companies to the entire data center ecosystem and to find the “pick-and-shovel” players with core competitiveness in downstream stages such as power, cooling, storage, memory, and networking.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00018022142695617305, "ret_signed_m1d": 0.00018022142695617305, "alpha_spy_m1d": 0.0001652252820627531, "alpha_c_m1d": 0.008961228217128192, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": -0.008781006790172019, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02525445902982811, "ret_signed_p1d": -0.02525445902982811, "alpha_spy_p1d": -0.020055643066907398, "alpha_c_p1d": -0.012276659186739442, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.012977799843088667, "ret_p1w": 0.010804180130794983, "ret_signed_p1w": 0.010804180130794983, "alpha_spy_p1w": -0.01268801236410999, "alpha_c_p1w": 0.012095516321735889, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": -0.001291336190940906, "ret_p1m": -0.21800628327410365, "ret_signed_p1m": -0.21800628327410365, "alpha_spy_p1m": -0.2051059410693924, "alpha_c_p1m": -0.18875780604161174, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.02924847723249191, "ret_p3m": -0.1819124074871897, "ret_signed_p3m": -0.1819124074871897, "alpha_spy_p3m": -0.21530159841225424, "alpha_c_p3m": -0.2632116665699148, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.08129925908272506, "ret_p6m": 0.09601487244026286, "ret_signed_p6m": 0.09601487244026286, "alpha_spy_p6m": 0.04739874229420782, "alpha_c_p6m": -0.016129536736254058, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.11214440917651691, "price_path": [-0.125, -0.0954, -0.0508, -0.0553, -0.0634, -0.0748, -0.0967, -0.0807, -0.0857, -0.1451, -0.148, -0.155, -0.156, -0.1359, -0.1453, -0.1602, -0.1635, -0.1551, -0.1826, -0.1937, -0.1981, -0.1813, -0.1862, -0.1791, -0.1701, -0.1583, -0.1847, -0.1847, -0.2061, -0.2037, -0.1989, -0.2042, -0.2078, -0.1845, -0.1606, -0.1554, -0.1536, -0.1476, -0.1525, -0.1555, -0.1353, -0.1358, -0.1117, -0.1275, -0.1354, -0.1384, -0.1381, -0.1192, -0.0943, -0.0349, -0.0307, -0.0376, -0.0157, -0.0223, 0.0205, 0.0148, -0.0273, 0.0076, -0.0124, 0.003, -0.0063, -0.0207, 0.0002, 0.0, -0.0253, -0.0275, -0.0194, 0.0149, 0.0108, -0.0084, -0.0391, -0.0326, -0.0472, -0.0945, -0.1083, -0.1352, -0.1486, -0.1286, -0.1577, -0.1755, -0.2159, -0.1951, -0.2231, -0.2216, -0.218, -0.2556, -0.2445, -0.2201, -0.2232, -0.2139, -0.2139, -0.1923, -0.1939, -0.1932, -0.1894, -0.1756, -0.1585, -0.1639, -0.1786, -0.1621, -0.1728, -0.2183, -0.2247, -0.2252, -0.2644, -0.2576, -0.2374, -0.2198, -0.2251, -0.2229, -0.2229, -0.2203, -0.2273, -0.233, -0.2428, -0.2428, -0.2071, -0.2139, -0.2017, -0.2097, -0.2272, -0.2219, -0.2108, -0.2117, -0.2204, -0.2088, -0.1819, -0.1819, -0.2075, -0.1853, -0.1826, -0.1845, -0.1914, -0.1766, -0.1683, -0.1721, -0.1993, -0.1813, -0.1888, -0.1731, -0.2087, -0.1484, -0.1364, -0.1389, -0.056, -0.0931, -0.0809, -0.0809, -0.0633, -0.0391, -0.0214, -0.0064, -0.016, 0.0072, 0.0401, 0.0293, 0.0199, 0.0251, -0.0132, 0.0124, -0.0076, -0.0241, 0.0219, 0.0299, 0.0195, 0.0081, -0.007, 0.0127, 0.0181, -0.0007, 0.0152, -0.0763, -0.0611, -0.0256, 0.0005, -0.0677, -0.0796, -0.1229, -0.0732, -0.0506, -0.0425, -0.0425, -0.0476, -0.0391, 0.0163, 0.033, 0.0602, 0.08, 0.1115, 0.096]}
{"unit_id": "orig:1980508499441549804", "ticker_idx": 2, "ts": "2025-10-21T05:36:36+00:00", "symbol": "TSMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foundries and CoWoS Are No Longer the Bottleneck", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: chip/packaging bottlenecks resolved; investment shifts to downstream HBM, power, cooling, racks, and networking; NVDA dominates AI demand; foundries/CoWoS de-rated as growth drivers.", "resolved_tickers": ["TSM"], "resolution_method": "exchange_ticker", "resolution_reason": "TSMC listed on NYSE as TSM (ADR); also 2330.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "OCP Conference Focus: Foundries and CoWoS Are No Longer the Bottleneck; AI Chip Bottlenecks Lie in Downstream Value Chains Such as Memory, Racks, and Power\n\nThe AI semiconductor industry will see another year of strong growth in 2026, but the investment logic in AI hardware is undergoing a profound shift.\n\nOn October 20, a Morgan Stanley report pointed out that over the past two years the market’s focus had been concentrated on upstream capacity bottlenecks such as TSMC’s CoWoS packaging and leading-edge process nodes.\n\nHowever, according to the latest remarks from NVIDIA and TSMC and the signals from the 2025 OCP Conference, this phase has already changed. Through large-scale expansions, the chip manufacturing and packaging stages are no longer the key factors constraining AI progress.\n\nMorgan Stanley emphasized that the real bottlenecks are moving downstream and are concentrating in the associated infrastructure—data center space, power supply, liquid cooling, high-bandwidth memory (HBM), server racks, and optical modules.\n\nThe report sees this as meaning that investment opportunities are spreading from upstream foundries and packaging to a broader downstream supply chain. Going forward, data centers that cannot secure sufficient power and physical space will become laggards in the competition for AI compute.\n\nUpstream Capacity Is No Longer the Sole Bottleneck; Chip Manufacturing and Packaging Have Already Expanded Significantly\n\nThere was a time when there were major doubts about whether AI chips could be supplied in sufficient quantity, and TSMC’s advanced CoWoS packaging capability was regarded as the critical bottleneck. But the latest industry trends show that the situation has changed markedly.\n\nOn its recent earnings conference call, TSMC stated that “AI demand is stronger than we imagined three months ago,” and said the company is working to “narrow the demand–supply gap.” Crucially, TSMC said the lead time to expand CoWoS capacity is only six months, which gives the supply side very significant flexibility.\n\nAlthough wafer front-end capacity at leading nodes such as 4 nm and 3 nm remains tight, AI semiconductors clearly have higher priority than cryptocurrency ASICs or Android smartphone SoCs.\n\nNVIDIA CEO Jensen Huang likewise made it clear in recent exchanges that semiconductor capacity is not the limiting factor it once was. After the surge in demand a few years ago, the manufacturing and packaging stages across the supply chain have “expanded dramatically,” and the company is confident it can meet customer demand.\n\nIn sum, even as overall demand continues to grow rapidly, global CoWoS demand in 2026 is projected to reach 1.154 million wafer starts, up 70% year over year, and the supply side’s ability to respond quickly has been noticeably strengthened.\n\nBottlenecks Shift: Downstream Infrastructure Emerges as the New Challenge\n\nIf chip supply is no longer the toughest problem, the bottlenecks naturally move downstream.\n\nNVIDIA pointed out that the bigger constraints now stem from the availability of data center space, power, and related infrastructure, and that the build cycles in these areas are far longer than those for chip manufacturing.\n\nThe OCP Conference contents corroborate this. As AI clusters move toward “hundreds-of-thousands-GPU” scale, data center design philosophy is being re-architected:\n\nPower and Cooling:\nDeploying large-scale GPU clusters entails enormous power consumption and heat/cooling challenges. At the OCP Conference, liquid cooling has become the default configuration for new AI racks, and demand for power-delivery solutions such as high-voltage DC (HVDC 800 V) is steadily increasing.\nThis benefits companies such as Aspeed, whose BMCs (baseboard management controllers) are expanding beyond servers to a wide range of equipment including cooling.\n\nStorage and Memory:\nAI workloads impose extreme demands on data storage and access speed. Meta has made it clear that, for cost reasons, it will prioritize QLC NAND flash in its data centers. At the same time, Seagate noted that HDDs (mechanical hard drives) will still keep about 95% of total capacity online to meet the needs of large and remote data centers.\nAbove all, demand for HBM (high-bandwidth memory) is surging, and the report projects that global HBM consumption will reach 26 billion GB in 2026, with NVIDIA alone consuming 54% of it. Such highly concentrated, intense demand makes HBM supply a key variable affecting AI server shipments.\n\nRacks and Networking:\nTo realize ultra-large deployments, OCP has put forward standardized blueprints such as the “AI Open Data Center” and “AI Open Cluster Design,” covering racks, liquid cooling, power interfaces, and more.\nOn the networking side, Alibaba stated that pluggable optics remain the best choice due to total cost of ownership and flexibility, and technologies such as LPO (linear-drive pluggable optics) are also drawing attention.\nCPO/NPO are expected to be realized in 2028 as manufacturing processes mature.\n\nDemand Outlook Points to Explosive Growth in Downstream Components\nThe importance of downstream infrastructure is validated by demand forecast data.\n\nMorgan Stanley analysts expect global cloud services capex to increase 31% year over year in 2026 to $582 billion, far exceeding the market’s common expectation of 16%.\n\nAssuming a higher capex mix for AI servers, this implies that AI server capex could grow about 70% year over year in 2026.\n\nOn the demand side, major AI big techs continue “stockpiling.” The report breaks down AI chip demand in 2026 in detail:\n\nCoWoS capacity consumption: NVIDIA 59%, Broadcom 18%, AMD 9%, AWS 6%.\n\nAI compute wafer consumption: NVIDIA leads with 55%, followed by Google 22%, AMD 6%, and AWS 6%.\n\nIn summary, the signals conveyed by the OCP Conference and cross-validation with industry-chain data clearly point to a new direction for AI hardware investment. As capacity bottlenecks at the chip manufacturing and packaging stages are gradually resolved, the market’s focus will inevitably shift to the infrastructure that supports ultra-large-scale AI compute.\n\nFor investors, the report sees this as a call to broaden their view from individual chip companies to the entire data center ecosystem and to find the “pick-and-shovel” players with core competitiveness in downstream stages such as power, cooling, storage, memory, and networking.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010831595550414352, "ret_signed_m1d": -0.010831595550414352, "alpha_spy_m1d": -0.010816599405520932, "alpha_c_m1d": -0.00527116003497663, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019116480392036284, "ret_signed_p1d": 0.019116480392036284, "alpha_spy_p1d": 0.013917664429115573, "alpha_c_p1d": -0.0002868862956356688, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.02383615779061965, "ret_signed_p1w": -0.02383615779061965, "alpha_spy_p1w": -0.0003439652957146766, "alpha_c_p1w": 0.027742200909245618, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": -0.04122103744337835, "ret_signed_p1m": 0.04122103744337835, "alpha_spy_p1m": 0.028320695238667093, "alpha_c_p1m": 0.024250361293710387, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.16574760577726777, "ret_signed_p3m": -0.16574760577726777, "alpha_spy_p3m": -0.13235841485220323, "alpha_c_p3m": -0.0025970516663931242, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.28067779889675193, "ret_signed_p6m": -0.28067779889675193, "alpha_spy_p6m": -0.2320616687506969, "alpha_c_p6m": 0.03530707019364532, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.1822, -0.1687, -0.1783, -0.1831, -0.1778, -0.1822, -0.2039, -0.191, -0.2131, -0.2169, -0.1788, -0.1815, -0.1806, -0.1731, -0.1828, -0.1843, -0.1914, -0.1829, -0.2124, -0.2262, -0.2305, -0.2114, -0.2026, -0.192, -0.1901, -0.1935, -0.2186, -0.2186, -0.2269, -0.2168, -0.2039, -0.1761, -0.1633, -0.1507, -0.1185, -0.1236, -0.1222, -0.1153, -0.1102, -0.1077, -0.0878, -0.1006, -0.0743, -0.0401, -0.0469, -0.0606, -0.0718, -0.0723, -0.0517, -0.0205, -0.0217, -0.0079, 0.0268, -0.0016, 0.034, 0.0182, -0.047, 0.0285, 0.0049, 0.0346, 0.0181, 0.0019, 0.0108, 0.0, -0.0191, -0.0128, 0.0015, 0.0127, 0.0238, 0.0359, 0.0296, 0.0201, 0.0351, -0.0016, -0.003, -0.0179, -0.0272, 0.0026, -0.0113, -0.0132, -0.0418, -0.0329, -0.0424, -0.0564, -0.0412, -0.0578, -0.066, -0.0335, -0.0334, -0.0154, -0.0154, -0.0102, -0.0232, -0.0082, 0.0032, -0.0054, 0.0007, 0.025, 0.0302, 0.0531, 0.0379, -0.0057, -0.0203, -0.0233, -0.0571, -0.0308, -0.0162, -0.0015, 0.011, 0.0173, 0.0173, 0.0311, 0.0245, 0.02, 0.0346, 0.0346, 0.0882, 0.0971, 0.1148, 0.085, 0.0827, 0.1018, 0.1296, 0.1276, 0.1137, 0.1632, 0.1657, 0.1657, 0.1139, 0.1103, 0.1146, 0.1401, 0.1328, 0.1519, 0.1654, 0.156, 0.1254, 0.1622, 0.1431, 0.109, 0.126, 0.1877, 0.21, 0.2322, 0.2736, 0.2532, 0.2473, 0.2473, 0.24, 0.2334, 0.227, 0.2616, 0.2599, 0.3133, 0.3201, 0.2829, 0.2753, 0.2567, 0.2023, 0.217, 0.2048, 0.1538, 0.1872, 0.1817, 0.2071, 0.1464, 0.1518, 0.1584, 0.1813, 0.1594, 0.1567, 0.1241, 0.1555, 0.1719, 0.1873, 0.1134, 0.1156, 0.0806, 0.1538, 0.1659, 0.1576, 0.1576, 0.1668, 0.179, 0.2493, 0.2479, 0.2653, 0.2618, 0.297, 0.2807]}
{"unit_id": "orig:1980508499441549804", "ticker_idx": 3, "ts": "2025-10-21T05:36:36+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA alone consuming 54% of it. Such highly concentrated, intense demand makes HBM supply a key variable affecting AI server shipments", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley: chip/packaging bottlenecks resolved; investment shifts to downstream HBM, power, cooling, racks, and networking; NVDA dominates AI demand; foundries/CoWoS de-rated as growth drivers.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "OCP Conference Focus: Foundries and CoWoS Are No Longer the Bottleneck; AI Chip Bottlenecks Lie in Downstream Value Chains Such as Memory, Racks, and Power\n\nThe AI semiconductor industry will see another year of strong growth in 2026, but the investment logic in AI hardware is undergoing a profound shift.\n\nOn October 20, a Morgan Stanley report pointed out that over the past two years the market’s focus had been concentrated on upstream capacity bottlenecks such as TSMC’s CoWoS packaging and leading-edge process nodes.\n\nHowever, according to the latest remarks from NVIDIA and TSMC and the signals from the 2025 OCP Conference, this phase has already changed. Through large-scale expansions, the chip manufacturing and packaging stages are no longer the key factors constraining AI progress.\n\nMorgan Stanley emphasized that the real bottlenecks are moving downstream and are concentrating in the associated infrastructure—data center space, power supply, liquid cooling, high-bandwidth memory (HBM), server racks, and optical modules.\n\nThe report sees this as meaning that investment opportunities are spreading from upstream foundries and packaging to a broader downstream supply chain. Going forward, data centers that cannot secure sufficient power and physical space will become laggards in the competition for AI compute.\n\nUpstream Capacity Is No Longer the Sole Bottleneck; Chip Manufacturing and Packaging Have Already Expanded Significantly\n\nThere was a time when there were major doubts about whether AI chips could be supplied in sufficient quantity, and TSMC’s advanced CoWoS packaging capability was regarded as the critical bottleneck. But the latest industry trends show that the situation has changed markedly.\n\nOn its recent earnings conference call, TSMC stated that “AI demand is stronger than we imagined three months ago,” and said the company is working to “narrow the demand–supply gap.” Crucially, TSMC said the lead time to expand CoWoS capacity is only six months, which gives the supply side very significant flexibility.\n\nAlthough wafer front-end capacity at leading nodes such as 4 nm and 3 nm remains tight, AI semiconductors clearly have higher priority than cryptocurrency ASICs or Android smartphone SoCs.\n\nNVIDIA CEO Jensen Huang likewise made it clear in recent exchanges that semiconductor capacity is not the limiting factor it once was. After the surge in demand a few years ago, the manufacturing and packaging stages across the supply chain have “expanded dramatically,” and the company is confident it can meet customer demand.\n\nIn sum, even as overall demand continues to grow rapidly, global CoWoS demand in 2026 is projected to reach 1.154 million wafer starts, up 70% year over year, and the supply side’s ability to respond quickly has been noticeably strengthened.\n\nBottlenecks Shift: Downstream Infrastructure Emerges as the New Challenge\n\nIf chip supply is no longer the toughest problem, the bottlenecks naturally move downstream.\n\nNVIDIA pointed out that the bigger constraints now stem from the availability of data center space, power, and related infrastructure, and that the build cycles in these areas are far longer than those for chip manufacturing.\n\nThe OCP Conference contents corroborate this. As AI clusters move toward “hundreds-of-thousands-GPU” scale, data center design philosophy is being re-architected:\n\nPower and Cooling:\nDeploying large-scale GPU clusters entails enormous power consumption and heat/cooling challenges. At the OCP Conference, liquid cooling has become the default configuration for new AI racks, and demand for power-delivery solutions such as high-voltage DC (HVDC 800 V) is steadily increasing.\nThis benefits companies such as Aspeed, whose BMCs (baseboard management controllers) are expanding beyond servers to a wide range of equipment including cooling.\n\nStorage and Memory:\nAI workloads impose extreme demands on data storage and access speed. Meta has made it clear that, for cost reasons, it will prioritize QLC NAND flash in its data centers. At the same time, Seagate noted that HDDs (mechanical hard drives) will still keep about 95% of total capacity online to meet the needs of large and remote data centers.\nAbove all, demand for HBM (high-bandwidth memory) is surging, and the report projects that global HBM consumption will reach 26 billion GB in 2026, with NVIDIA alone consuming 54% of it. Such highly concentrated, intense demand makes HBM supply a key variable affecting AI server shipments.\n\nRacks and Networking:\nTo realize ultra-large deployments, OCP has put forward standardized blueprints such as the “AI Open Data Center” and “AI Open Cluster Design,” covering racks, liquid cooling, power interfaces, and more.\nOn the networking side, Alibaba stated that pluggable optics remain the best choice due to total cost of ownership and flexibility, and technologies such as LPO (linear-drive pluggable optics) are also drawing attention.\nCPO/NPO are expected to be realized in 2028 as manufacturing processes mature.\n\nDemand Outlook Points to Explosive Growth in Downstream Components\nThe importance of downstream infrastructure is validated by demand forecast data.\n\nMorgan Stanley analysts expect global cloud services capex to increase 31% year over year in 2026 to $582 billion, far exceeding the market’s common expectation of 16%.\n\nAssuming a higher capex mix for AI servers, this implies that AI server capex could grow about 70% year over year in 2026.\n\nOn the demand side, major AI big techs continue “stockpiling.” The report breaks down AI chip demand in 2026 in detail:\n\nCoWoS capacity consumption: NVIDIA 59%, Broadcom 18%, AMD 9%, AWS 6%.\n\nAI compute wafer consumption: NVIDIA leads with 55%, followed by Google 22%, AMD 6%, and AWS 6%.\n\nIn summary, the signals conveyed by the OCP Conference and cross-validation with industry-chain data clearly point to a new direction for AI hardware investment. As capacity bottlenecks at the chip manufacturing and packaging stages are gradually resolved, the market’s focus will inevitably shift to the infrastructure that supports ultra-large-scale AI compute.\n\nFor investors, the report sees this as a call to broaden their view from individual chip companies to the entire data center ecosystem and to find the “pick-and-shovel” players with core competitiveness in downstream stages such as power, cooling, storage, memory, and networking.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008169605123257151, "ret_signed_m1d": 0.008169605123257151, "alpha_spy_m1d": 0.008154608978363731, "alpha_c_m1d": 0.002609169607819428, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004857555235790012, "ret_signed_p1d": -0.004857555235790012, "alpha_spy_p1d": 0.0003412607271306989, "alpha_c_p1d": 0.01454581145188194, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.10968210622045049, "ret_signed_p1w": 0.10968210622045049, "alpha_spy_p1w": 0.08618991372554552, "alpha_c_p1w": 0.058103747520585225, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.029587072846396723, "ret_signed_p1m": 0.029587072846396723, "alpha_spy_p1m": 0.04248741505110798, "alpha_c_p1m": 0.046557748996064685, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.02804350496049346, "ret_signed_p3m": 0.02804350496049346, "alpha_spy_p3m": -0.005345685964571079, "alpha_c_p3m": -0.1351070491503812, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.097879422110557, "ret_signed_p6m": 0.097879422110557, "alpha_spy_p6m": 0.049263291964501965, "alpha_c_p6m": -0.21810544697984025, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.041, -0.0423, -0.0244, -0.0312, -0.0105, -0.0182, -0.0411, -0.0065, -0.0161, -0.0097, -0.0022, 0.0084, 0.0049, 0.011, 0.0023, 0.0047, -0.004, 0.0046, -0.0305, -0.0318, -0.0342, -0.0176, -0.0075, 0.0033, 0.0024, -0.0055, -0.0386, -0.0386, -0.0574, -0.0582, -0.0525, -0.0781, -0.071, -0.0575, -0.0212, -0.022, -0.0184, -0.0188, -0.0347, -0.06, -0.0272, -0.0248, 0.0135, -0.0151, -0.0231, -0.0192, -0.0164, 0.0038, 0.0299, 0.0336, 0.0427, 0.0357, 0.0242, 0.0214, 0.0439, 0.063, 0.011, 0.0395, -0.0062, -0.0073, 0.0036, 0.0114, 0.0082, 0.0, -0.0049, 0.0055, 0.0282, 0.057, 0.1097, 0.1429, 0.1199, 0.1177, 0.142, 0.0968, 0.0776, 0.0382, 0.0386, 0.0988, 0.0662, 0.0698, 0.0315, 0.0497, 0.03, 0.0011, 0.0296, -0.0029, -0.0126, 0.0077, -0.0184, -0.005, -0.005, -0.023, -0.0068, 0.0017, -0.0087, 0.0123, 0.007, 0.0243, 0.0211, 0.0145, -0.0012, -0.0338, -0.0268, -0.0189, -0.0564, -0.0387, -0.0009, 0.014, 0.0445, 0.0412, 0.0412, 0.0518, 0.039, 0.0353, 0.0295, 0.0295, 0.0425, 0.0385, 0.0336, 0.0439, 0.0215, 0.0205, 0.0209, 0.0257, 0.011, 0.0326, 0.028, 0.028, -0.017, 0.012, 0.0204, 0.036, 0.0294, 0.0407, 0.0572, 0.0627, 0.0551, 0.0246, -0.0045, -0.0384, -0.0512, 0.0235, 0.0491, 0.0408, 0.0491, 0.032, 0.0092, 0.0092, 0.0211, 0.0377, 0.0373, 0.0479, 0.0574, 0.0646, 0.0795, 0.0206, -0.0219, 0.0073, -0.0061, 0.0104, 0.0121, -0.0184, 0.0083, 0.02, 0.027, 0.011, -0.0049, 0.0115, 0.0044, -0.0041, -0.0142, -0.0466, -0.0304, -0.0328, -0.0136, -0.0547, -0.0752, -0.0882, -0.0372, -0.0298, -0.0207, -0.0207, -0.0193, -0.0168, 0.0052, 0.0153, 0.0413, 0.0451, 0.0849, 0.0979]}
{"unit_id": "orig:1982998762416193687", "ticker_idx": 0, "ts": "2025-10-28T02:32:01+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM raises Samsung Electronics' price target to ₩135,000", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "JPM raises Samsung Electronics PT to ₩135,000; author shares bullish rating.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "JPM raises Samsung Electronics’ price target to ₩135,000.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02512561264602753, "ret_signed_m1d": 0.02512561264602753, "alpha_spy_m1d": 0.02777457595882371, "alpha_c_m1d": 0.034928223696123806, "bench_spy_m1d": -0.002648963312796182, "bench_c_m1d": -0.009802611050096277, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010050197620864676, "ret_signed_p1d": 0.010050197620864676, "alpha_spy_p1d": 0.009569899922058811, "alpha_c_p1d": 0.002863683604111822, "bench_spy_p1d": 0.0004802976988058649, "bench_c_p1d": 0.007186514016752854, "ret_p1w": 0.05427132015291658, "ret_signed_p1w": 0.05427132015291658, "alpha_spy_p1w": 0.07147501908845322, "alpha_c_p1w": 0.08083107958630564, "bench_spy_p1w": -0.017203698935536638, "bench_c_p1w": -0.026559759433389063, "ret_p1m": 0.03316580236775035, "ret_signed_p1m": 0.03316580236775035, "alpha_spy_p1m": 0.043907192254231786, "alpha_c_p1m": 0.09337241667261431, "bench_spy_p1m": -0.010741389886481434, "bench_c_p1m": -0.06020661430486396, "ret_p3m": 0.5360740797372971, "ret_signed_p3m": 0.5360740797372971, "alpha_spy_p3m": 0.5299515348527803, "alpha_c_p3m": 0.5736282570353839, "bench_spy_p3m": 0.006122544884516801, "bench_c_p3m": -0.037554177298086744, "ret_p6m": 1.2011021481381277, "ret_signed_p6m": 1.2011021481381277, "alpha_spy_p6m": 1.1600582260928742, "alpha_c_p6m": 1.1511177095272758, "bench_spy_p6m": 0.04104392204525342, "bench_c_p6m": 0.04998443861085189, "price_path": [-0.2856, -0.3106, -0.3026, -0.3006, -0.3116, -0.2946, -0.2816, -0.2896, -0.2886, -0.2806, -0.2836, -0.2836, -0.2996, -0.2996, -0.2946, -0.2936, -0.2856, -0.2846, -0.2966, -0.2936, -0.3036, -0.3026, -0.3236, -0.3086, -0.3016, -0.2986, -0.3046, -0.2986, -0.2846, -0.2736, -0.2656, -0.2456, -0.2346, -0.2056, -0.2176, -0.1965, -0.1965, -0.1645, -0.1525, -0.1455, -0.1385, -0.1665, -0.1538, -0.1568, -0.1357, -0.098, -0.098, -0.098, -0.098, -0.098, -0.098, -0.0513, -0.0623, -0.0794, -0.0452, -0.0181, -0.0161, -0.0141, -0.0201, -0.009, -0.0302, -0.007, 0.0251, 0.0, 0.0101, 0.0462, 0.0804, 0.1166, 0.0543, 0.0111, -0.003, -0.0161, 0.0111, 0.0402, 0.0362, 0.0332, -0.0231, 0.0111, -0.0171, -0.0302, 0.0111, -0.0472, -0.0281, -0.002, 0.0332, 0.0402, 0.0101, 0.0131, 0.0392, 0.0503, 0.0563, 0.0894, 0.1005, 0.0894, 0.0854, 0.0784, 0.0945, 0.0533, 0.0332, 0.0844, 0.0814, 0.0683, 0.1106, 0.1206, 0.1166, 0.1166, 0.1759, 0.2068, 0.2109, 0.2109, 0.2109, 0.2977, 0.3947, 0.4028, 0.424, 0.4018, 0.4038, 0.4018, 0.3896, 0.4169, 0.4533, 0.5038, 0.5078, 0.4664, 0.5098, 0.5381, 0.5361, 0.5361, 0.6108, 0.6401, 0.6229, 0.6209, 0.5189, 0.6916, 0.7078, 0.6088, 0.6017, 0.6805, 0.6744, 0.6946, 0.8037, 0.83, 0.83, 0.83, 0.83, 0.9188, 0.9198, 0.9491, 1.0198, 1.0552, 1.2016, 1.1865, 1.1865, 0.9703, 0.7391, 0.935, 0.9007, 0.7522, 0.8976, 0.9188, 0.8976, 0.8532, 0.9057, 0.9582, 1.1057, 1.0249, 1.0138, 0.8815, 0.9158, 0.9087, 0.8188, 0.8188, 0.7842, 0.6921, 0.9193, 0.8054, 0.8843, 0.9542, 0.9886, 1.1303, 1.0645, 1.0847, 1.0341, 1.0898, 1.1353, 1.2011, 1.1859, 1.1707, 1.2163, 1.2011]}
{"unit_id": "orig:1977301691839230406", "ticker_idx": 0, "ts": "2025-10-12T09:13:54+00:00", "symbol": "5803.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe that this differentiated new product launch will further strengthen its dominant market position", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Goldman endorses Fujikura's new fusion splicer launch as further strengthening its dominant 50%+ global market position.", "resolved_tickers": ["5803.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Conglomerates'", "bench_c_industry": "Conglomerates", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) Fujikura (5803.T): Launch of New Optical Fiber Fusion Splicer and Fiber Cleaver; Enhanced Competitiveness Through Reduced Work Time\n\nOn October 10, Fujikura announced that it would begin sales in October of a new core-alignment-type optical fiber fusion splicer and a new fiber cleaver. These products are used for long-distance optical fiber connections in fields such as telecommunications infrastructure and data center interconnect (DCI). Unlike existing products, they can splice two fibers simultaneously, reducing splicing work time by 20%.\n\nFujikura holds the world’s No.1 market share in the fusion splicer segment (over 50% global share), and we believe that this differentiated new product launch will further strengthen its dominant market position. Since optical fiber fusion splicing is a labor-intensive process, the work-time reduction provided by these products will bring substantial cost-saving benefits to customers in terms of installation efficiency (for more details on fusion splicers, refer to our March 7, 2025 report on Fujikura’s Sakura Plant visit).\n\n⸻\n\nKey Details\n    \n• Fujikura announced that it will begin sales in October of the new core-alignment-type fusion splicer “100S” and the new fiber cleaver “CT60.”\n    \n• The “100S” employs a core alignment method (aligning the central axis of the light-transmitting core) to splice single-core optical fibers used for long-distance transmission. Unlike previous models, it enables simultaneous splicing of two fibers.\n    \n• According to the company, when used in combination with the single-fiber stripper “SS05” (capable of dual-fiber simultaneous coating removal) and the new fiber cleaver “CT60” (capable of dual-fiber simultaneous cutting), total splicing work time can be reduced by 20%.\n    \n• The “100S” is also newly equipped with GPS functionality, allowing splicing test results to be recorded along with location data. This is expected to improve installation management and enhance traceability.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.015112192089016707, "alpha_c_m1d": 0.010299879851799698, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.010299879851799698, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.049214923518699716, "ret_signed_p1d": -0.049214923518699716, "alpha_spy_p1d": -0.04799326871524301, "alpha_c_p1d": -0.06096718547178315, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": 0.011752261953083432, "ret_p1w": 0.03834534477686624, "ret_signed_p1w": 0.03834534477686624, "alpha_spy_p1w": 0.02588753935992516, "alpha_c_p1w": 0.024744430628695913, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.013600914148170329, "ret_p1m": 0.16455313252937098, "ret_signed_p1m": 0.16455313252937098, "alpha_spy_p1m": 0.13444937704252458, "alpha_c_p1m": 0.1461984978622155, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.01835463466715548, "ret_p3m": 0.061896128342474155, "ret_signed_p3m": 0.061896128342474155, "alpha_spy_p3m": 0.018901026256162412, "alpha_c_p3m": 0.0007672340181485993, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.061128894324325556, "ret_p6m": 0.5908181557212777, "ret_signed_p6m": 0.5908181557212777, "alpha_spy_p6m": 0.5909183247495696, "alpha_c_p6m": 0.49940232494898384, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.09141583077229387, "price_path": [-0.5229, -0.5192, -0.5151, -0.5151, -0.4936, -0.4986, -0.4779, -0.4688, -0.4713, -0.4638, -0.416, -0.3777, -0.3828, -0.384, -0.3597, -0.3444, -0.3096, -0.3345, -0.3345, -0.3024, -0.2802, -0.2859, -0.2638, -0.2563, -0.2796, -0.3204, -0.3135, -0.3111, -0.3105, -0.3087, -0.2997, -0.2611, -0.2341, -0.2554, -0.2548, -0.2503, -0.2119, -0.1999, -0.187, -0.226, -0.1825, -0.1609, -0.1543, -0.1543, -0.1621, -0.193, -0.184, -0.1618, -0.1453, -0.1453, -0.1261, -0.1273, -0.1543, -0.1135, -0.1271, -0.1368, -0.1452, -0.1464, -0.1244, -0.0767, -0.0359, 0.0136, 0.0, 0.0, -0.0492, -0.0242, 0.0145, -0.0069, 0.0383, 0.0069, 0.0226, 0.0414, 0.0824, 0.1688, 0.1591, 0.2144, 0.2563, 0.2784, 0.2784, 0.2506, 0.1845, 0.2998, 0.2328, 0.2385, 0.1646, 0.2123, 0.2503, 0.1703, 0.1981, 0.0794, 0.0773, 0.1407, 0.0462, 0.0462, 0.0495, 0.0891, 0.093, 0.0839, -0.013, 0.0024, 0.0284, 0.0236, 0.0417, 0.1144, 0.1153, 0.1147, 0.1002, 0.1171, 0.0682, -0.0036, -0.0021, -0.0362, 0.0042, 0.074, 0.0773, 0.0912, 0.064, 0.0362, 0.0752, 0.0531, 0.0531, 0.0531, 0.0531, 0.1138, 0.1171, 0.1018, 0.0619, 0.0317, 0.0317, 0.0365, 0.0444, 0.0447, 0.0694, 0.0441, 0.0193, 0.0827, 0.0969, 0.0812, 0.0845, 0.1093, 0.212, 0.1757, 0.1818, 0.1993, 0.3152, 0.3774, 0.3077, 0.3524, 0.3258, 0.3925, 0.3925, 0.3919, 0.3161, 0.295, 0.2935, 0.349, 0.3859, 0.3829, 0.3829, 0.5211, 0.6199, 0.6585, 0.6162, 0.7165, 0.6516, 0.5323, 0.5894, 0.5157, 0.3656, 0.4532, 0.5489, 0.5519, 0.6014, 0.5477, 0.4858, 0.5522, 0.5296, 0.5296, 0.4692, 0.5459, 0.6591, 0.6978, 0.6685, 0.64, 0.4889, 0.6185, 0.5628, 0.6804, 0.6745, 0.5908]}
{"unit_id": "orig:1986600821312934086", "ticker_idx": 0, "ts": "2025-11-07T01:05:19+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We expect the NAND market to remain undersupplied through 2026. We raised our 2026 data-center demand growth outlook from +25% to +40%. NAND prices were higher at the end of the quarter than at the beginning, and that trend is expected to carry into next quarter.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SNDK: NAND undersupply through 2026, data-center demand outlook raised to +40%, rising prices carrying into next quarter, QLC mix expanding.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "SanDisk Earnings Call — Q&A \n\nQ: You mentioned the contract structure with customers has changed versus the past. What exactly has changed? And what’s the bit-shipment outlook for 2025–2026?\nA1: In the past, most contracts were on a quarterly basis. Now, data-center customers are sharing demand forecasts out through 2027 and are asking for long-term supply agreements.\n\n⸻\n\nQ: What does supply look like over the next few years? Any plans to add new wafer capacity?\nA1: We expect the NAND market to remain undersupplied through 2026. Demand is extending far enough that customers are already discussing supply for 2027. At present, we have no plans to add additional wafer capacity.\nA2: We raised our 2026 data-center demand growth outlook from +25% to +40%.\n\n⸻\n\nQ: Is the HDD shortage affecting enterprise SSD demand?\nA1: HDDs and SSDs are more complementary than substitutes. With AI proliferating, a larger share of data is moving to SSDs.\nA2: Over the long term, SSD demand will grow much faster. Starting in 2026, data centers are expected to become the largest end market for NAND, replacing the traditional mobile-centric mix.\n\n⸻\n\nQ: Within data-center demand, how do you see the mix of QLC vs. TLC? And what role does “Big Sage” play?\nA1: Both QLC (cost-optimized, high-capacity storage) and TLC (higher-performance compute) have solid demand. QLC currently accounts for about 20% of total revenue and is expected to reach around 40% by 2026. The Big Sage node is optimized to deliver efficient, high-performance QLC.\n\n⸻\n\nQ: Why is gross margin jumping from 33% to 41–43% this quarter?\nA1: Pricing. NAND prices were higher at the end of the quarter than at the beginning, and that trend is expected to carry into next quarter.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.13274592389044837, "ret_signed_m1d": -0.13274592389044837, "alpha_spy_m1d": -0.1317623846109488, "alpha_c_m1d": -0.1362163518175009, "bench_spy_m1d": -0.0009835392794995679, "bench_c_m1d": 0.003470427927052544, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11888264985557728, "ret_signed_p1d": 0.11888264985557728, "alpha_spy_p1d": 0.10327821506406676, "alpha_c_p1d": 0.09330657623607563, "bench_spy_p1d": 0.015604434791510524, "bench_c_p1d": 0.02557607361950165, "ret_p1w": 0.06129951643757359, "ret_signed_p1w": 0.06129951643757359, "alpha_spy_p1w": 0.05986862276529625, "alpha_c_p1w": 0.06133413682082334, "bench_spy_p1w": 0.0014308936722773336, "bench_c_p1w": -3.4620383249750475e-05, "ret_p1m": -0.05850173190572194, "ret_signed_p1m": -0.05850173190572194, "alpha_spy_p1m": -0.07737002409419147, "alpha_c_p1m": -0.08314082503009268, "bench_spy_p1m": 0.018868292188469526, "bench_c_p1m": 0.024639093124370737, "ret_p3m": 1.4409136388077375, "ret_signed_p3m": 1.4409136388077375, "alpha_spy_p3m": 1.4152080516075256, "alpha_c_p3m": 1.480825524622492, "bench_spy_p3m": 0.025705587200211877, "bench_c_p3m": -0.03991188581475458, "ret_p6m": 4.244112275583879, "ret_signed_p6m": 4.244112275583879, "alpha_spy_p6m": 4.1679115384280045, "alpha_c_p6m": 4.116283307073772, "bench_spy_p6m": 0.07620073715587439, "bench_c_p6m": 0.1278289685101064, "price_path": [-0.8045, -0.8037, -0.8051, -0.814, -0.8099, -0.8138, -0.8146, -0.81, -0.8064, -0.8047, -0.8023, -0.7977, -0.7876, -0.7809, -0.7809, -0.7867, -0.7786, -0.739, -0.7138, -0.7056, -0.7056, -0.6913, -0.648, -0.6403, -0.6238, -0.6177, -0.6076, -0.5871, -0.5732, -0.5702, -0.5557, -0.5831, -0.6063, -0.5945, -0.5261, -0.5315, -0.4942, -0.4816, -0.4638, -0.494, -0.4949, -0.4493, -0.4585, -0.5118, -0.4379, -0.4685, -0.3974, -0.3976, -0.4147, -0.3818, -0.3766, -0.3864, -0.3024, -0.2226, -0.263, -0.267, -0.1467, -0.1823, -0.1677, -0.1356, -0.1875, -0.096, -0.1327, 0.0, 0.1189, 0.134, 0.1821, 0.0171, 0.0613, 0.1102, 0.0228, 0.0271, -0.1817, -0.1637, -0.0523, -0.0793, -0.1021, -0.1021, -0.0676, -0.1224, -0.1425, -0.1883, -0.1093, -0.046, -0.0585, -0.0836, -0.0276, 0.0089, -0.1391, -0.157, -0.126, -0.1363, -0.0836, -0.0078, 0.0066, 0.0226, 0.0443, 0.0443, 0.0441, 0.0199, 0.0031, -0.0088, -0.0088, 0.1493, 0.1445, 0.46, 0.4764, 0.3969, 0.576, 0.6255, 0.6277, 0.6194, 0.7089, 0.7272, 0.7272, 0.8921, 1.0932, 1.1022, 0.9786, 0.9659, 1.0103, 1.2032, 1.252, 1.4063, 1.7779, 1.9043, 1.4409, 1.406, 1.4969, 1.4361, 1.2617, 1.5027, 1.6319, 1.6163, 1.6163, 1.4661, 1.5071, 1.5935, 1.7141, 1.7831, 1.6663, 1.6406, 1.7221, 1.6531, 1.5851, 1.361, 1.5015, 1.3617, 1.202, 1.4584, 1.5843, 1.7369, 1.584, 1.7627, 1.9382, 2.0072, 2.1472, 2.224, 1.9635, 1.9334, 1.9334, 1.8305, 1.5187, 1.5715, 1.3906, 1.653, 1.8926, 1.9296, 1.9296, 2.0258, 1.9681, 2.2608, 2.5559, 2.5567, 2.9774, 2.9438, 2.7236, 2.8394, 2.8458, 2.8125, 2.7727, 3.0883, 2.8936, 3.1335, 3.4688, 3.1855, 3.4438, 3.5787, 3.9566, 4.2441]}
{"unit_id": "orig:1994256814461075511", "ticker_idx": 0, "ts": "2025-11-28T04:07:30+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is estimated that both Spot Price and Contract Price will continue their upward trend throughout 2026", "tweet_type": "analysis", "timeline": "months", "summary": "KIS Memory Q&A: DRAM and NAND ASP both rising through 2026; supply constrained, demand robust; tagging MU and SNDK as beneficiaries.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Key Q&A from KIS Memory Semiconductor Asia Marketing\n\nQ. The premium gap between DRAM Spot Price and Contract Price has widened abnormally. Why?\n\nA. The DRAM Spot Price is currently not a price that reflects the actual market situation. This is because there is almost no DRAM inventory in the channel market, so the prices of a few abnormal transactions with extremely low trading volumes are being reflected in the Spot Price. Major suppliers like Samsung Electronics and SK Hynix are forced to drastically reduce channel supply because they lack sufficient volume for their contract customers. Channel suppliers such as A-Data and Phison are accumulating inventory and sharply reducing channel circulation, judging that DRAM and NAND price increases will continue in the future.\n\nBecause of this, the Spot Price may face a short-term correction, but this does not imply a change in the industry cycle; rather, it should be viewed as a short-term adjustment due to increased volatility in the Spot Price. In terms of directionality, it is estimated that both Spot Price and Contract Price will continue their upward trend throughout 2026. However, considering the volatility of the Spot Price, it is judged that it would be somewhat unreasonable for the Contract Price to rise to the level of the Spot Price.\n\nQ. What is the outlook for HBM prices in 2026?\n\nA. We have consistently mentioned through our reports that a decline in HBM3e prices is inevitable. This is not due to intensifying competition, but rather a process of the HBM3e price premium dissipating due to the launch of the new product, HBM4. Nvidia’s HBM3e requirements have actually increased, and the HBM supply shortage continues to intensify. Furthermore, in a situation where the margins of conventional DRAM could surpass those of HBM due to rising conventional DRAM prices, memory suppliers have no incentive to lower prices to compete in HBM.\n\nWhile the blended ASP of HBM in 2026 is bound to decline compared to 2025 as the proportion of HBM3e expands, it is necessary to understand that, as mentioned above, this is a process of HBM3e premium dissipation and not due to HBM competition overheating or oversupply.\n\nQ. Until when and to what extent can the DRAM ASP (Average Selling Price) rise?\n\nA. It is known that the supply fulfillment rate against server customer demand is at the 60% level. Considering this, a sharp rise in ASP is natural. However, the question to consider is up to what level clients can absorb the ASP increase in terms of BOM (Bill of Materials) cost. Simply comparing it to the previous server super-cycle prices of 2017–2018, server DRAM ASP rose 100% in one year, reaching $1 per Gb.\n\nThis could be considered a primary resistance line, but considering that current demand is not inventory restocking demand, the possibility of ASP rising beyond this cannot be ruled out. However, since the average price of HBM3e in 2026 is $1.25–$1.35, it is judged that the possibility of conventional DRAM prices rising to exceed HBM prices is low.\n\nQ. What is the impact of the increased burden on clients' BOM costs due to the rise in DRAM ASP?\n\nA. Mobile and PC sectors, which have relatively lower margins compared to data centers, are facing pressure from increased BOM costs. Among mobile suppliers, Apple and Samsung Electronics find it relatively easier to procure DRAM, and considering that Chinese mobile companies can procure some from CXMT, it is estimated that the burden on mobile is less than that on PC.\n\nIn the case of PC OEMs, operating profit margins are only 5–7%, so margin pressure is most severe when DRAM prices rise. We see only two options: passing the price increase on to consumer prices or downgrading specs to sell cheaper models. If PC consumer prices increase, PC shipments may decrease due to reduced consumer purchasing desire, and if specs are downgraded, DRAM content per unit may decrease. Both are negative for DRAM demand, but since PCs account for only about 10% of DRAM, the impact on overall DRAM demand is estimated to be somewhat insignificant.\n\nQ. All three DRAM makers are revising their Capex upward. Is there no risk of supply increase?\n\nA. Most of the Capex of the three DRAM makers will be used for mid-to-long-term infrastructure investment. It is estimated that the supply increase resulting from this will contribute from late 2027 or 2028 at the earliest. Samsung Electronics' Pyeongtaek P5 is scheduled for completion in late 2027, and SK Hynix's Yongin Y1 is scheduled for completion in the third quarter of 2027; considering the lead time for equipment move-in, full-scale production contribution will be possible from 2028. All three DRAM makers emphasized that the Capex increase is for securing mid-to-long-term competitiveness, and notably, SK Hynix mentioned in its conference call that it would maintain \"Capex Discipline.\"\n\nQ. What is the current status of CXMT?\n\nA. It is judged that the market impact of CXMT's supply increase will be limited until 2026. CXMT's market share is 16% based on Capa (capacity) as of the end of 2025, but it is expected to be 6–7% based on shipments. The inability to fully utilize Capa is due to relatively low yields, limited applications centered on mobile, and restrictions on sales to non-China customers. Until the IPO in the first quarter, they appear likely to maximize sales of DDR4 and LPDDR4 to maximize profitability.\n\nHowever, they plan to accelerate the transition to the 1z node within 2026, and the center of gravity will shift to DDR5 and LPDDR5 from 2027 onwards. Nevertheless, in 2026, DRAM supply growth will be limited compared to Capa growth rates due to the node transition within the year. Capa expansion somewhat stalled in 2025 but is seen to resume from 2026. Currently, CXMT's Capa is at the level of 300K wafers per month, and they have a goal to double this by 2030.\n\nQ. Can the trend of rising NAND ASP continue like DRAM?\n\nA. The current shortage of NAND supply is because NAND suppliers have been reducing supply for three years. Unlike DRAM, where physical cleanroom space is insufficient, NAND still has room. However, it is important that NAND suppliers have absolutely no intention of increasing utilization rates. This is because NAND suppliers implicitly agree that the demand structure for NAND is weaker compared to DRAM, and increasing supply could quickly turn the industry into a downturn.\n\nSince 2026 is the year when node transition from V7 to V9 takes place, the decline in utilization rates due to process transition will also act as a constraint on NAND supply. Therefore, it is estimated that NAND ASP will also continue its upward trend throughout 2026.\n\n$MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009987782549662002, "ret_signed_m1d": 0.009987782549662002, "alpha_spy_m1d": 0.015416556434493097, "alpha_c_m1d": 0.02296048327258645, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011636503345092844, "ret_signed_p1d": 0.011636503345092844, "alpha_spy_p1d": 0.016201918947644035, "alpha_c_p1d": 0.009706231242160671, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.03605042281291615, "ret_signed_p1w": 0.03605042281291615, "alpha_spy_p1w": 0.03268483450278015, "alpha_c_p1w": 0.0015891610355884822, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.21589458983864432, "ret_signed_p1m": 0.21589458983864432, "alpha_spy_p1m": 0.2063941457193116, "alpha_c_p1m": 0.1788125620075495, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.9234377538180883, "ret_signed_p3m": 0.9234377538180883, "alpha_spy_p3m": 0.9061588216374324, "alpha_c_p3m": 0.7099540736678168, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 2.2593443363820978, "ret_signed_p6m": 2.2593443363820978, "alpha_spy_p6m": 2.1620416483755474, "alpha_c_p6m": 1.618282487728361, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.4412, -0.4373, -0.4267, -0.4136, -0.4093, -0.3924, -0.372, -0.3526, -0.3231, -0.3171, -0.2954, -0.3068, -0.2735, -0.2822, -0.2718, -0.259, -0.2655, -0.2707, -0.2919, -0.2705, -0.2676, -0.2319, -0.1949, -0.189, -0.1846, -0.192, -0.1767, -0.1827, -0.1619, -0.158, -0.1739, -0.1488, -0.1061, -0.0975, -0.078, -0.0904, -0.0906, -0.0878, -0.0431, -0.0152, -0.0297, 0.0035, 0.018, 0.0233, 0.089, 0.0235, 0.0323, 0.0377, 0.0246, 0.0716, 0.0722, 0.0749, 0.0596, 0.0223, 0.0556, 0.0047, -0.0083, -0.0236, -0.0658, -0.0368, -0.028, -0.0051, 0.01, 0.0, 0.0116, 0.0315, 0.0238, 0.0089, 0.0361, 0.0741, 0.0713, 0.0991, 0.0755, 0.0602, 0.0308, 0.002, 0.0223, 0.0544, 0.0714, 0.1212, 0.1265, 0.1424, 0.1424, 0.1662, 0.2159, 0.2217, 0.2115, 0.2115, 0.2988, 0.3382, 0.4038, 0.4154, 0.3992, 0.4178, 0.4214, 0.3996, 0.4043, 0.4254, 0.4833, 0.4896, 0.4659, 0.5124, 0.5431, 0.5529, 0.5185, 0.6132, 0.684, 0.6916, 0.6917, 0.6406, 0.7202, 0.6646, 0.6004, 0.6128, 0.6532, 0.6299, 0.6788, 0.7375, 0.7379, 0.7379, 0.7211, 0.751, 0.784, 0.8342, 0.835, 0.8881, 0.9234, 1.005, 0.9716, 0.972, 0.7773, 0.674, 0.7914, 0.7316, 0.654, 0.7857, 0.8254, 0.7838, 0.7859, 0.8656, 0.9085, 1.0114, 0.933, 0.8954, 0.779, 0.8113, 0.7957, 0.6894, 0.6919, 0.5928, 0.5435, 0.7175, 0.6355, 0.6905, 0.7361, 0.7661, 0.9277, 0.9048, 0.9284, 0.9283, 1.0416, 1.0641, 1.0992, 1.0741, 1.0837, 1.1366, 1.1847, 1.2012, 1.2038, 1.3039, 1.272, 1.3009, 1.2763, 1.3116, 1.5015, 1.5914, 1.8375, 1.8611, 2.0126, 2.2601, 2.1694, 2.3273, 2.3246, 2.0718, 2.0593, 2.0053, 2.0539, 2.2978, 2.2593, 2.2593]}
{"unit_id": "orig:1994256814461075511", "ticker_idx": 1, "ts": "2025-11-28T04:07:30+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is estimated that NAND ASP will also continue its upward trend throughout 2026", "tweet_type": "analysis", "timeline": "months", "summary": "KIS Memory Q&A: DRAM and NAND ASP both rising through 2026; supply constrained, demand robust; tagging MU and SNDK as beneficiaries.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket: dominant producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Key Q&A from KIS Memory Semiconductor Asia Marketing\n\nQ. The premium gap between DRAM Spot Price and Contract Price has widened abnormally. Why?\n\nA. The DRAM Spot Price is currently not a price that reflects the actual market situation. This is because there is almost no DRAM inventory in the channel market, so the prices of a few abnormal transactions with extremely low trading volumes are being reflected in the Spot Price. Major suppliers like Samsung Electronics and SK Hynix are forced to drastically reduce channel supply because they lack sufficient volume for their contract customers. Channel suppliers such as A-Data and Phison are accumulating inventory and sharply reducing channel circulation, judging that DRAM and NAND price increases will continue in the future.\n\nBecause of this, the Spot Price may face a short-term correction, but this does not imply a change in the industry cycle; rather, it should be viewed as a short-term adjustment due to increased volatility in the Spot Price. In terms of directionality, it is estimated that both Spot Price and Contract Price will continue their upward trend throughout 2026. However, considering the volatility of the Spot Price, it is judged that it would be somewhat unreasonable for the Contract Price to rise to the level of the Spot Price.\n\nQ. What is the outlook for HBM prices in 2026?\n\nA. We have consistently mentioned through our reports that a decline in HBM3e prices is inevitable. This is not due to intensifying competition, but rather a process of the HBM3e price premium dissipating due to the launch of the new product, HBM4. Nvidia’s HBM3e requirements have actually increased, and the HBM supply shortage continues to intensify. Furthermore, in a situation where the margins of conventional DRAM could surpass those of HBM due to rising conventional DRAM prices, memory suppliers have no incentive to lower prices to compete in HBM.\n\nWhile the blended ASP of HBM in 2026 is bound to decline compared to 2025 as the proportion of HBM3e expands, it is necessary to understand that, as mentioned above, this is a process of HBM3e premium dissipation and not due to HBM competition overheating or oversupply.\n\nQ. Until when and to what extent can the DRAM ASP (Average Selling Price) rise?\n\nA. It is known that the supply fulfillment rate against server customer demand is at the 60% level. Considering this, a sharp rise in ASP is natural. However, the question to consider is up to what level clients can absorb the ASP increase in terms of BOM (Bill of Materials) cost. Simply comparing it to the previous server super-cycle prices of 2017–2018, server DRAM ASP rose 100% in one year, reaching $1 per Gb.\n\nThis could be considered a primary resistance line, but considering that current demand is not inventory restocking demand, the possibility of ASP rising beyond this cannot be ruled out. However, since the average price of HBM3e in 2026 is $1.25–$1.35, it is judged that the possibility of conventional DRAM prices rising to exceed HBM prices is low.\n\nQ. What is the impact of the increased burden on clients' BOM costs due to the rise in DRAM ASP?\n\nA. Mobile and PC sectors, which have relatively lower margins compared to data centers, are facing pressure from increased BOM costs. Among mobile suppliers, Apple and Samsung Electronics find it relatively easier to procure DRAM, and considering that Chinese mobile companies can procure some from CXMT, it is estimated that the burden on mobile is less than that on PC.\n\nIn the case of PC OEMs, operating profit margins are only 5–7%, so margin pressure is most severe when DRAM prices rise. We see only two options: passing the price increase on to consumer prices or downgrading specs to sell cheaper models. If PC consumer prices increase, PC shipments may decrease due to reduced consumer purchasing desire, and if specs are downgraded, DRAM content per unit may decrease. Both are negative for DRAM demand, but since PCs account for only about 10% of DRAM, the impact on overall DRAM demand is estimated to be somewhat insignificant.\n\nQ. All three DRAM makers are revising their Capex upward. Is there no risk of supply increase?\n\nA. Most of the Capex of the three DRAM makers will be used for mid-to-long-term infrastructure investment. It is estimated that the supply increase resulting from this will contribute from late 2027 or 2028 at the earliest. Samsung Electronics' Pyeongtaek P5 is scheduled for completion in late 2027, and SK Hynix's Yongin Y1 is scheduled for completion in the third quarter of 2027; considering the lead time for equipment move-in, full-scale production contribution will be possible from 2028. All three DRAM makers emphasized that the Capex increase is for securing mid-to-long-term competitiveness, and notably, SK Hynix mentioned in its conference call that it would maintain \"Capex Discipline.\"\n\nQ. What is the current status of CXMT?\n\nA. It is judged that the market impact of CXMT's supply increase will be limited until 2026. CXMT's market share is 16% based on Capa (capacity) as of the end of 2025, but it is expected to be 6–7% based on shipments. The inability to fully utilize Capa is due to relatively low yields, limited applications centered on mobile, and restrictions on sales to non-China customers. Until the IPO in the first quarter, they appear likely to maximize sales of DDR4 and LPDDR4 to maximize profitability.\n\nHowever, they plan to accelerate the transition to the 1z node within 2026, and the center of gravity will shift to DDR5 and LPDDR5 from 2027 onwards. Nevertheless, in 2026, DRAM supply growth will be limited compared to Capa growth rates due to the node transition within the year. Capa expansion somewhat stalled in 2025 but is seen to resume from 2026. Currently, CXMT's Capa is at the level of 300K wafers per month, and they have a goal to double this by 2030.\n\nQ. Can the trend of rising NAND ASP continue like DRAM?\n\nA. The current shortage of NAND supply is because NAND suppliers have been reducing supply for three years. Unlike DRAM, where physical cleanroom space is insufficient, NAND still has room. However, it is important that NAND suppliers have absolutely no intention of increasing utilization rates. This is because NAND suppliers implicitly agree that the demand structure for NAND is weaker compared to DRAM, and increasing supply could quickly turn the industry into a downturn.\n\nSince 2026 is the year when node transition from V7 to V9 takes place, the decline in utilization rates due to process transition will also act as a constraint on NAND supply. Therefore, it is estimated that NAND ASP will also continue its upward trend throughout 2026.\n\n$MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008979101929613643, "ret_signed_m1d": -0.008979101929613643, "alpha_spy_m1d": -0.0035503280447825476, "alpha_c_m1d": 0.0039935987933108034, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01590302580181129, "ret_signed_p1d": -0.01590302580181129, "alpha_spy_p1d": -0.011337610199260096, "alpha_c_p1d": -0.017833297904743462, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.026789439838736894, "ret_signed_p1w": 0.026789439838736894, "alpha_spy_p1w": 0.02342385152860089, "alpha_c_p1w": -0.007671821938590773, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.17530312216126093, "ret_signed_p1m": 0.17530312216126093, "alpha_spy_p1m": 0.16580267804192822, "alpha_c_p1m": 0.1382210943301661, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 1.1769191797579022, "ret_signed_p3m": 1.1769191797579022, "alpha_spy_p3m": 1.1596402475772463, "alpha_c_p3m": 0.9634354996076306, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 3.6717877537346326, "ret_signed_p6m": 3.6717877537346326, "alpha_spy_p6m": 3.5744850657280822, "alpha_c_p6m": 3.030725905080896, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.5634, -0.5591, -0.5442, -0.5216, -0.5154, -0.5048, -0.4829, -0.4509, -0.4223, -0.4152, -0.3952, -0.4055, -0.3794, -0.3814, -0.3684, -0.3576, -0.3646, -0.374, -0.3947, -0.3612, -0.3564, -0.3306, -0.2907, -0.267, -0.2678, -0.2814, -0.2629, -0.2612, -0.2615, -0.2432, -0.2627, -0.2257, -0.1881, -0.1934, -0.1622, -0.1708, -0.1673, -0.1464, -0.0724, -0.0424, -0.0612, -0.0005, 0.018, 0.0227, 0.069, 0.0172, 0.0375, 0.0532, 0.0853, 0.1662, 0.1671, 0.1801, 0.1309, 0.0542, 0.1089, 0.0404, 0.0457, 0.0024, -0.0468, -0.0055, -0.0098, -0.0322, -0.009, 0.0, -0.0159, -0.0012, -0.02, -0.0107, 0.0268, 0.0623, 0.0493, 0.0694, 0.0661, 0.0311, -0.0051, -0.0266, -0.0036, 0.0314, 0.0543, 0.1039, 0.1069, 0.1349, 0.139, 0.1664, 0.1753, 0.1701, 0.1614, 0.1614, 0.2477, 0.2897, 0.4021, 0.4358, 0.4156, 0.4586, 0.4713, 0.4799, 0.4724, 0.5116, 0.5741, 0.5877, 0.6087, 0.7073, 0.7576, 0.7247, 0.7038, 0.7915, 0.8858, 0.9073, 0.9853, 0.9706, 1.097, 0.9668, 0.8916, 0.9093, 0.9164, 0.8638, 0.9448, 1.0573, 1.0897, 1.0781, 1.0373, 1.0438, 1.0786, 1.1199, 1.1352, 1.1783, 1.1769, 1.2386, 1.2031, 1.1952, 1.0024, 0.9502, 1.0131, 0.9361, 0.9031, 1.0419, 1.1369, 1.0763, 1.1121, 1.2305, 1.2497, 1.3793, 1.3268, 1.2484, 1.153, 1.1646, 1.1618, 1.0039, 0.9977, 0.8942, 0.9009, 1.1145, 1.0583, 1.1073, 1.1755, 1.1911, 1.443, 1.4946, 1.5609, 1.6751, 1.8151, 1.7263, 1.8033, 1.7186, 1.7168, 1.7874, 1.9278, 1.9082, 1.9442, 2.0951, 2.0333, 2.1061, 2.1466, 2.2244, 2.4, 2.5887, 2.7397, 2.84, 3.153, 3.3191, 3.1825, 3.3665, 3.2637, 3.0491, 3.1236, 3.1005, 3.1703, 3.5368, 3.5006, 3.6718]}
{"unit_id": "orig:1994256814461075511", "ticker_idx": 2, "ts": "2025-11-28T04:07:30+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU", "tweet_type": "analysis", "timeline": "months", "summary": "KIS Memory Q&A: DRAM and NAND ASP both rising through 2026; supply constrained, demand robust; tagging MU and SNDK as beneficiaries.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Key Q&A from KIS Memory Semiconductor Asia Marketing\n\nQ. The premium gap between DRAM Spot Price and Contract Price has widened abnormally. Why?\n\nA. The DRAM Spot Price is currently not a price that reflects the actual market situation. This is because there is almost no DRAM inventory in the channel market, so the prices of a few abnormal transactions with extremely low trading volumes are being reflected in the Spot Price. Major suppliers like Samsung Electronics and SK Hynix are forced to drastically reduce channel supply because they lack sufficient volume for their contract customers. Channel suppliers such as A-Data and Phison are accumulating inventory and sharply reducing channel circulation, judging that DRAM and NAND price increases will continue in the future.\n\nBecause of this, the Spot Price may face a short-term correction, but this does not imply a change in the industry cycle; rather, it should be viewed as a short-term adjustment due to increased volatility in the Spot Price. In terms of directionality, it is estimated that both Spot Price and Contract Price will continue their upward trend throughout 2026. However, considering the volatility of the Spot Price, it is judged that it would be somewhat unreasonable for the Contract Price to rise to the level of the Spot Price.\n\nQ. What is the outlook for HBM prices in 2026?\n\nA. We have consistently mentioned through our reports that a decline in HBM3e prices is inevitable. This is not due to intensifying competition, but rather a process of the HBM3e price premium dissipating due to the launch of the new product, HBM4. Nvidia’s HBM3e requirements have actually increased, and the HBM supply shortage continues to intensify. Furthermore, in a situation where the margins of conventional DRAM could surpass those of HBM due to rising conventional DRAM prices, memory suppliers have no incentive to lower prices to compete in HBM.\n\nWhile the blended ASP of HBM in 2026 is bound to decline compared to 2025 as the proportion of HBM3e expands, it is necessary to understand that, as mentioned above, this is a process of HBM3e premium dissipation and not due to HBM competition overheating or oversupply.\n\nQ. Until when and to what extent can the DRAM ASP (Average Selling Price) rise?\n\nA. It is known that the supply fulfillment rate against server customer demand is at the 60% level. Considering this, a sharp rise in ASP is natural. However, the question to consider is up to what level clients can absorb the ASP increase in terms of BOM (Bill of Materials) cost. Simply comparing it to the previous server super-cycle prices of 2017–2018, server DRAM ASP rose 100% in one year, reaching $1 per Gb.\n\nThis could be considered a primary resistance line, but considering that current demand is not inventory restocking demand, the possibility of ASP rising beyond this cannot be ruled out. However, since the average price of HBM3e in 2026 is $1.25–$1.35, it is judged that the possibility of conventional DRAM prices rising to exceed HBM prices is low.\n\nQ. What is the impact of the increased burden on clients' BOM costs due to the rise in DRAM ASP?\n\nA. Mobile and PC sectors, which have relatively lower margins compared to data centers, are facing pressure from increased BOM costs. Among mobile suppliers, Apple and Samsung Electronics find it relatively easier to procure DRAM, and considering that Chinese mobile companies can procure some from CXMT, it is estimated that the burden on mobile is less than that on PC.\n\nIn the case of PC OEMs, operating profit margins are only 5–7%, so margin pressure is most severe when DRAM prices rise. We see only two options: passing the price increase on to consumer prices or downgrading specs to sell cheaper models. If PC consumer prices increase, PC shipments may decrease due to reduced consumer purchasing desire, and if specs are downgraded, DRAM content per unit may decrease. Both are negative for DRAM demand, but since PCs account for only about 10% of DRAM, the impact on overall DRAM demand is estimated to be somewhat insignificant.\n\nQ. All three DRAM makers are revising their Capex upward. Is there no risk of supply increase?\n\nA. Most of the Capex of the three DRAM makers will be used for mid-to-long-term infrastructure investment. It is estimated that the supply increase resulting from this will contribute from late 2027 or 2028 at the earliest. Samsung Electronics' Pyeongtaek P5 is scheduled for completion in late 2027, and SK Hynix's Yongin Y1 is scheduled for completion in the third quarter of 2027; considering the lead time for equipment move-in, full-scale production contribution will be possible from 2028. All three DRAM makers emphasized that the Capex increase is for securing mid-to-long-term competitiveness, and notably, SK Hynix mentioned in its conference call that it would maintain \"Capex Discipline.\"\n\nQ. What is the current status of CXMT?\n\nA. It is judged that the market impact of CXMT's supply increase will be limited until 2026. CXMT's market share is 16% based on Capa (capacity) as of the end of 2025, but it is expected to be 6–7% based on shipments. The inability to fully utilize Capa is due to relatively low yields, limited applications centered on mobile, and restrictions on sales to non-China customers. Until the IPO in the first quarter, they appear likely to maximize sales of DDR4 and LPDDR4 to maximize profitability.\n\nHowever, they plan to accelerate the transition to the 1z node within 2026, and the center of gravity will shift to DDR5 and LPDDR5 from 2027 onwards. Nevertheless, in 2026, DRAM supply growth will be limited compared to Capa growth rates due to the node transition within the year. Capa expansion somewhat stalled in 2025 but is seen to resume from 2026. Currently, CXMT's Capa is at the level of 300K wafers per month, and they have a goal to double this by 2030.\n\nQ. Can the trend of rising NAND ASP continue like DRAM?\n\nA. The current shortage of NAND supply is because NAND suppliers have been reducing supply for three years. Unlike DRAM, where physical cleanroom space is insufficient, NAND still has room. However, it is important that NAND suppliers have absolutely no intention of increasing utilization rates. This is because NAND suppliers implicitly agree that the demand structure for NAND is weaker compared to DRAM, and increasing supply could quickly turn the industry into a downturn.\n\nSince 2026 is the year when node transition from V7 to V9 takes place, the decline in utilization rates due to process transition will also act as a constraint on NAND supply. Therefore, it is estimated that NAND ASP will also continue its upward trend throughout 2026.\n\n$MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026302486440269668, "ret_signed_m1d": -0.026302486440269668, "alpha_spy_m1d": -0.020873712555438573, "alpha_c_m1d": -0.013329785717345222, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016830145448651246, "ret_signed_p1d": 0.016830145448651246, "alpha_spy_p1d": 0.02139556105120244, "alpha_c_p1d": 0.014899873345719072, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.003129183432686311, "ret_signed_p1w": 0.003129183432686311, "alpha_spy_p1w": -0.00023640487744969185, "alpha_c_p1w": -0.031332078344641356, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.24530160753597463, "ret_signed_p1m": 0.24530160753597463, "alpha_spy_p1m": 0.23580116341664192, "alpha_c_p1m": 0.2082195797048798, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.8148397275073795, "ret_signed_p3m": 0.8148397275073795, "alpha_spy_p3m": 0.7975607953267236, "alpha_c_p3m": 0.601356047357108, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 2.178361776639486, "ret_signed_p6m": 2.178361776639486, "alpha_spy_p6m": 2.0810590886329354, "alpha_c_p6m": 1.537299927985749, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.4993, -0.4983, -0.4751, -0.4448, -0.4444, -0.4285, -0.4084, -0.3637, -0.3355, -0.3333, -0.3288, -0.3239, -0.2863, -0.3123, -0.3043, -0.2967, -0.3166, -0.3372, -0.3354, -0.3074, -0.2929, -0.2302, -0.2235, -0.2057, -0.1925, -0.2148, -0.1689, -0.1867, -0.2321, -0.1848, -0.209, -0.1883, -0.1436, -0.1442, -0.1256, -0.1446, -0.1607, -0.1259, -0.0738, -0.0693, -0.0616, -0.0417, -0.0527, -0.0537, -0.0075, -0.078, 0.0043, 0.0078, 0.0061, 0.0711, 0.0196, 0.0356, 0.002, 0.0438, 0.0231, -0.0337, -0.0447, -0.1485, -0.1231, -0.0531, -0.0505, -0.0263, -0.0263, 0.0, 0.0168, 0.0127, -0.0098, -0.0416, 0.0031, 0.0441, 0.0674, 0.1151, 0.0929, 0.0197, 0.0043, -0.0168, -0.0463, 0.051, 0.1245, 0.1696, 0.1683, 0.2123, 0.2123, 0.2043, 0.2453, 0.2379, 0.2074, 0.2074, 0.3344, 0.3205, 0.4528, 0.4364, 0.3834, 0.4599, 0.4632, 0.4304, 0.4102, 0.4241, 0.5346, 0.5346, 0.5441, 0.6461, 0.6819, 0.6907, 0.646, 0.7355, 0.8414, 0.8436, 0.7551, 0.8521, 0.7744, 0.605, 0.6198, 0.6697, 0.6224, 0.579, 0.7359, 0.7513, 0.7415, 0.7415, 0.6912, 0.7808, 0.7656, 0.8113, 0.7809, 0.7683, 0.8148, 0.758, 0.7445, 0.7458, 0.6062, 0.6954, 0.6797, 0.5665, 0.647, 0.7053, 0.7712, 0.7148, 0.8027, 0.869, 0.9531, 0.9533, 0.8794, 0.789, 0.7106, 0.6732, 0.6164, 0.5037, 0.5112, 0.3619, 0.4298, 0.5568, 0.55, 0.55, 0.5987, 0.598, 0.7214, 0.7839, 0.78, 0.8053, 0.9708, 0.9308, 0.9351, 0.9259, 0.8978, 0.9019, 1.0631, 1.0387, 1.1022, 1.22, 1.1342, 1.1942, 1.1887, 1.2947, 1.4396, 1.7094, 1.8211, 1.7366, 2.1606, 2.366, 2.2443, 2.4011, 2.2842, 2.0669, 1.8844, 1.9572, 2.0979, 2.2253, 2.1784, 2.1784]}
{"unit_id": "orig:1994256814461075511", "ticker_idx": 3, "ts": "2025-11-28T04:07:30+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$SNDK", "tweet_type": "analysis", "timeline": "months", "summary": "KIS Memory Q&A: DRAM and NAND ASP both rising through 2026; supply constrained, demand robust; tagging MU and SNDK as beneficiaries.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Summary of Key Q&A from KIS Memory Semiconductor Asia Marketing\n\nQ. The premium gap between DRAM Spot Price and Contract Price has widened abnormally. Why?\n\nA. The DRAM Spot Price is currently not a price that reflects the actual market situation. This is because there is almost no DRAM inventory in the channel market, so the prices of a few abnormal transactions with extremely low trading volumes are being reflected in the Spot Price. Major suppliers like Samsung Electronics and SK Hynix are forced to drastically reduce channel supply because they lack sufficient volume for their contract customers. Channel suppliers such as A-Data and Phison are accumulating inventory and sharply reducing channel circulation, judging that DRAM and NAND price increases will continue in the future.\n\nBecause of this, the Spot Price may face a short-term correction, but this does not imply a change in the industry cycle; rather, it should be viewed as a short-term adjustment due to increased volatility in the Spot Price. In terms of directionality, it is estimated that both Spot Price and Contract Price will continue their upward trend throughout 2026. However, considering the volatility of the Spot Price, it is judged that it would be somewhat unreasonable for the Contract Price to rise to the level of the Spot Price.\n\nQ. What is the outlook for HBM prices in 2026?\n\nA. We have consistently mentioned through our reports that a decline in HBM3e prices is inevitable. This is not due to intensifying competition, but rather a process of the HBM3e price premium dissipating due to the launch of the new product, HBM4. Nvidia’s HBM3e requirements have actually increased, and the HBM supply shortage continues to intensify. Furthermore, in a situation where the margins of conventional DRAM could surpass those of HBM due to rising conventional DRAM prices, memory suppliers have no incentive to lower prices to compete in HBM.\n\nWhile the blended ASP of HBM in 2026 is bound to decline compared to 2025 as the proportion of HBM3e expands, it is necessary to understand that, as mentioned above, this is a process of HBM3e premium dissipation and not due to HBM competition overheating or oversupply.\n\nQ. Until when and to what extent can the DRAM ASP (Average Selling Price) rise?\n\nA. It is known that the supply fulfillment rate against server customer demand is at the 60% level. Considering this, a sharp rise in ASP is natural. However, the question to consider is up to what level clients can absorb the ASP increase in terms of BOM (Bill of Materials) cost. Simply comparing it to the previous server super-cycle prices of 2017–2018, server DRAM ASP rose 100% in one year, reaching $1 per Gb.\n\nThis could be considered a primary resistance line, but considering that current demand is not inventory restocking demand, the possibility of ASP rising beyond this cannot be ruled out. However, since the average price of HBM3e in 2026 is $1.25–$1.35, it is judged that the possibility of conventional DRAM prices rising to exceed HBM prices is low.\n\nQ. What is the impact of the increased burden on clients' BOM costs due to the rise in DRAM ASP?\n\nA. Mobile and PC sectors, which have relatively lower margins compared to data centers, are facing pressure from increased BOM costs. Among mobile suppliers, Apple and Samsung Electronics find it relatively easier to procure DRAM, and considering that Chinese mobile companies can procure some from CXMT, it is estimated that the burden on mobile is less than that on PC.\n\nIn the case of PC OEMs, operating profit margins are only 5–7%, so margin pressure is most severe when DRAM prices rise. We see only two options: passing the price increase on to consumer prices or downgrading specs to sell cheaper models. If PC consumer prices increase, PC shipments may decrease due to reduced consumer purchasing desire, and if specs are downgraded, DRAM content per unit may decrease. Both are negative for DRAM demand, but since PCs account for only about 10% of DRAM, the impact on overall DRAM demand is estimated to be somewhat insignificant.\n\nQ. All three DRAM makers are revising their Capex upward. Is there no risk of supply increase?\n\nA. Most of the Capex of the three DRAM makers will be used for mid-to-long-term infrastructure investment. It is estimated that the supply increase resulting from this will contribute from late 2027 or 2028 at the earliest. Samsung Electronics' Pyeongtaek P5 is scheduled for completion in late 2027, and SK Hynix's Yongin Y1 is scheduled for completion in the third quarter of 2027; considering the lead time for equipment move-in, full-scale production contribution will be possible from 2028. All three DRAM makers emphasized that the Capex increase is for securing mid-to-long-term competitiveness, and notably, SK Hynix mentioned in its conference call that it would maintain \"Capex Discipline.\"\n\nQ. What is the current status of CXMT?\n\nA. It is judged that the market impact of CXMT's supply increase will be limited until 2026. CXMT's market share is 16% based on Capa (capacity) as of the end of 2025, but it is expected to be 6–7% based on shipments. The inability to fully utilize Capa is due to relatively low yields, limited applications centered on mobile, and restrictions on sales to non-China customers. Until the IPO in the first quarter, they appear likely to maximize sales of DDR4 and LPDDR4 to maximize profitability.\n\nHowever, they plan to accelerate the transition to the 1z node within 2026, and the center of gravity will shift to DDR5 and LPDDR5 from 2027 onwards. Nevertheless, in 2026, DRAM supply growth will be limited compared to Capa growth rates due to the node transition within the year. Capa expansion somewhat stalled in 2025 but is seen to resume from 2026. Currently, CXMT's Capa is at the level of 300K wafers per month, and they have a goal to double this by 2030.\n\nQ. Can the trend of rising NAND ASP continue like DRAM?\n\nA. The current shortage of NAND supply is because NAND suppliers have been reducing supply for three years. Unlike DRAM, where physical cleanroom space is insufficient, NAND still has room. However, it is important that NAND suppliers have absolutely no intention of increasing utilization rates. This is because NAND suppliers implicitly agree that the demand structure for NAND is weaker compared to DRAM, and increasing supply could quickly turn the industry into a downturn.\n\nSince 2026 is the year when node transition from V7 to V9 takes place, the decline in utilization rates due to process transition will also act as a constraint on NAND supply. Therefore, it is estimated that NAND ASP will also continue its upward trend throughout 2026.\n\n$MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03690436016756238, "ret_signed_m1d": -0.03690436016756238, "alpha_spy_m1d": -0.03147558628273128, "alpha_c_m1d": -0.028379522796800005, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.008524837370762373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05871551720720969, "ret_signed_p1d": -0.05871551720720969, "alpha_spy_p1d": -0.054150101604658496, "alpha_c_p1d": -0.05916980883035661, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0004542916231469185, "ret_p1w": 0.023244367922701237, "ret_signed_p1w": 0.023244367922701237, "alpha_spy_p1w": 0.019878779612565234, "alpha_c_p1w": -0.0011425067808037426, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.02438687470350498, "ret_p1m": 0.09391795653595958, "ret_signed_p1m": 0.09391795653595958, "alpha_spy_p1m": 0.08441751241662687, "alpha_c_p1m": 0.07308597526102845, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.020831981274931133, "ret_p3m": 1.8322286292552974, "ret_signed_p3m": 1.8322286292552974, "alpha_spy_p3m": 1.8149496970746415, "alpha_c_p3m": 1.8314116741382191, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.0008169551170782441, "ret_p6m": 5.622581285786706, "ret_signed_p6m": 5.622581285786706, "alpha_spy_p6m": 5.525278597780155, "alpha_c_p6m": 5.358600752434013, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.26398053335269256, "price_path": [-0.7713, -0.7626, -0.7201, -0.693, -0.6843, -0.6842, -0.6689, -0.6224, -0.6143, -0.5965, -0.59, -0.5791, -0.5572, -0.5422, -0.5391, -0.5235, -0.5529, -0.5777, -0.565, -0.4917, -0.4975, -0.4575, -0.444, -0.4249, -0.4573, -0.4583, -0.4094, -0.4192, -0.4764, -0.3971, -0.4299, -0.3537, -0.3539, -0.3723, -0.337, -0.3314, -0.3419, -0.2518, -0.1662, -0.2096, -0.2139, -0.0847, -0.123, -0.1073, -0.0729, -0.1286, -0.0304, -0.0698, 0.0726, 0.2001, 0.2163, 0.2679, 0.0909, 0.1383, 0.1908, 0.097, 0.1016, -0.1224, -0.1031, 0.0165, -0.0125, -0.0369, -0.0369, 0.0, -0.0587, -0.0803, -0.1294, -0.0447, 0.0232, 0.0098, -0.0171, 0.0429, 0.0821, -0.0766, -0.0959, -0.0626, -0.0737, -0.0171, 0.0642, 0.0796, 0.0968, 0.12, 0.12, 0.1199, 0.0939, 0.0759, 0.0631, 0.0631, 0.2327, 0.2275, 0.5659, 0.5835, 0.4983, 0.6903, 0.7434, 0.7458, 0.7369, 0.8329, 0.8525, 0.8525, 1.0294, 1.2451, 1.2547, 1.1221, 1.1086, 1.1562, 1.3631, 1.4154, 1.5808, 1.9794, 2.115, 1.618, 1.5806, 1.678, 1.6129, 1.4258, 1.6843, 1.8229, 1.8062, 1.8062, 1.6451, 1.689, 1.7817, 1.911, 1.985, 1.8597, 1.8322, 1.9197, 1.8456, 1.7727, 1.5323, 1.683, 1.5331, 1.3617, 1.6367, 1.7718, 1.9355, 1.7715, 1.9632, 2.1513, 2.2254, 2.3755, 2.4579, 2.1786, 2.1462, 2.1462, 2.0359, 1.7014, 1.7581, 1.564, 1.8455, 2.1025, 2.1422, 2.1422, 2.2454, 2.1834, 2.4974, 2.8139, 2.8148, 3.2659, 3.2299, 2.9937, 3.118, 3.1248, 3.0891, 3.0464, 3.3849, 3.1761, 3.4334, 3.7931, 3.4892, 3.7663, 3.9109, 4.3162, 4.6246, 5.2985, 5.3149, 5.0013, 5.9972, 5.931, 5.5031, 5.4817, 5.1928, 5.3042, 4.9701, 5.1953, 5.2368, 5.9072, 5.6226, 5.6226]}
{"unit_id": "orig:1990925591156306109", "ticker_idx": 0, "ts": "2025-11-18T23:30:25+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SKH remains our top pick, backed by 1) its dominant market share in the high-growth, high-margin HBM market, and 2) its strengthening technological moat. We raise our TP from KRW 710,000 to KRW 845,000.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "CLSA reiterates SK Hynix as top pick, raises TP to KRW 845K on HBM dominance, DRAM/NAND ASP upgrades, and OP estimates +21-33% above consensus through 2027.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251119_SK Hynix: Solid positioning and superb execution - CLSA\n\n(1) SKH remains our top pick, backed by 1) its dominant market share in the high-growth, high-margin HBM market, and 2) its strengthening technological moat.\n\n(2) We expect blended DRAM/NAND ASPs to rise more than previously expected through 2026–27 and raise OP by +6% and +26%, respectively.\n■ 26(F): KRW 83.0trn → KRW 87.8trn\n■ 27(F): KRW 85.8trn → KRW 108.6trn\n\n(3) The revised estimates are +21% and +33% above current market expectations, respectively.\n\n(4) Accordingly, we raise our TP from KRW 710,000 to KRW 845,000.\n\n(5) The target price is based on 2.5x 2Q26 BVPS, which is roughly 2x higher than the past 10-year average 12M(F) P/B of 1.2x.\n\n(6) In particular, SKH’s book value is projected to grow by +49% in 2025, +76% in 2026, and +49% in 2027.\n\n(7) DRAM ASP outlook\n■ 26(F): 1Q +10% → 2Q +4% → 3Q +5% → 4Q +5%\n■ 27(F): 1Q +2% → 2Q Flat → 3Q -4% → 4Q -4%\n\n(8) NAND ASP outlook\n■ 26(F): 1Q +12% → 2Q +4% → 3Q +5% → 4Q +3%\n■ 27(F): 1Q +1% → 2Q Flat → 3Q -5% → 4Q -5%\n\n(9) CAPEX outlook\n■ 26(F): +30%\n■ 27(F): +20%\n\n(10) FCF and shareholder return outlook\n■ The company shifted to a net cash position in 3Q25, and as FCF expands sharply in 2026–27, we see a high likelihood of 1) higher dividends and 2) increased shareholder returns via share buybacks/cancellations.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06315794566036503, "ret_signed_m1d": 0.06315794566036503, "alpha_spy_m1d": 0.05468937057691092, "alpha_c_m1d": 0.04212531794586649, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.021032627714498542, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014035037953128104, "ret_signed_p1d": -0.014035037953128104, "alpha_spy_p1d": -0.017898237736715283, "alpha_c_p1d": -0.03248742819980832, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.018452390246680217, "ret_p1w": -0.08947362807891579, "ret_signed_p1w": -0.08947362807891579, "alpha_spy_p1w": -0.1121071450451262, "alpha_c_p1w": -0.10963625688642387, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.020162628807508076, "ret_p1m": -0.03264096554147555, "ret_signed_p1m": -0.03264096554147555, "alpha_spy_p1m": -0.04979032527211624, "alpha_c_p1m": -0.050493210849571146, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": 0.017852245308095593, "ret_p3m": 0.5449654244903281, "ret_signed_p3m": 0.5449654244903281, "alpha_spy_p3m": 0.5090842465951997, "alpha_c_p3m": 0.3178399570909469, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.22712546739938122, "ret_p6m": 2.4755510784520043, "ret_signed_p6m": 2.4755510784520043, "alpha_spy_p6m": 2.344572025172618, "alpha_c_p6m": 1.7526032958378377, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.7229477826141666, "price_path": [-0.5708, -0.5603, -0.5454, -0.5419, -0.5445, -0.5289, -0.5281, -0.5509, -0.543, -0.5395, -0.5342, -0.5202, -0.514, -0.4947, -0.4667, -0.4614, -0.4237, -0.4193, -0.3895, -0.4149, -0.3763, -0.3763, -0.3842, -0.3667, -0.3728, -0.3746, -0.4096, -0.3877, -0.3904, -0.3684, -0.3061, -0.3061, -0.3061, -0.3061, -0.3061, -0.3061, -0.2491, -0.2719, -0.2781, -0.2588, -0.2061, -0.1833, -0.1482, -0.1596, -0.1553, -0.1605, -0.1053, -0.0614, -0.086, -0.0211, -0.0035, -0.0193, 0.0877, 0.0281, 0.0158, 0.0404, 0.0175, 0.0632, 0.086, 0.0825, 0.0737, -0.0175, 0.0632, 0.0, -0.014, 0.0018, -0.086, -0.0877, -0.0895, -0.0807, -0.0449, -0.0695, -0.0555, -0.0204, -0.0309, -0.0484, -0.0449, 0.013, -0.0063, 0.0306, -0.0081, 0.0025, -0.0274, -0.0695, -0.0326, -0.0309, -0.0397, 0.0183, 0.0253, 0.0323, 0.0323, 0.0516, 0.1236, 0.1429, 0.1429, 0.1429, 0.1886, 0.2219, 0.2746, 0.3027, 0.3273, 0.3062, 0.315, 0.2957, 0.3027, 0.315, 0.3273, 0.3413, 0.3044, 0.2992, 0.3255, 0.3466, 0.2922, 0.4045, 0.4765, 0.5116, 0.5959, 0.4572, 0.5924, 0.5801, 0.4783, 0.473, 0.5573, 0.5379, 0.5099, 0.559, 0.545, 0.545, 0.545, 0.545, 0.5695, 0.6661, 0.6696, 0.7644, 0.7872, 0.933, 0.8662, 0.8662, 0.6516, 0.4933, 0.6551, 0.6252, 0.4704, 0.6498, 0.6797, 0.6358, 0.6006, 0.7132, 0.7061, 0.8574, 0.7817, 0.7712, 0.641, 0.7343, 0.7501, 0.641, 0.641, 0.5355, 0.4194, 0.5777, 0.4599, 0.5408, 0.5584, 0.6111, 0.8169, 0.7554, 0.8064, 0.8292, 0.94, 0.9981, 1.0315, 0.984, 1.0509, 1.1529, 1.1511, 1.1546, 1.1494, 1.2725, 1.2865, 1.2742, 1.2619, 1.2619, 1.5451, 1.5451, 1.816, 1.9092, 1.9655, 2.3067, 2.2275, 2.4756]}
{"unit_id": "thread:1990921592118116488", "ticker_idx": 0, "ts": "2025-11-18T23:14:31+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics, which operates about 70% of its total DRAM production capacity as commodity DRAM, will be the direct beneficiary", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung is the direct beneficiary of rising commodity DRAM prices given 70% of DRAM capacity there; SK Hynix faces margin pressure as HBM-focused strategy may be outpaced by commodity DRAM profitability in 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung puts more weight on expanding production of commodity memory than HBM… Aiming to maximize profitability\n\nSamsung Electronics has set the core goal of next year’s DS (Device Solutions) semiconductor business as “maximizing profitability” and has begun to reorganize its strategy accordingly, with Han Jin-man, head of the Foundry Business Unit, being mentioned as a candidate to head the Memory Business Unit. Until Samsung’s high-bandwidth memory (HBM) technology improves and selling prices can be fully maximized, the company is sketching out a scenario in which it focuses more heavily on commodity DRAM, where profitability can improve more quickly.\n\nRecently in the industry, Han Jin-man, who previously served as Head of Strategic Marketing Office in the Memory Business Unit, has been mentioned as a candidate for the Memory Business Unit chief, and this is being interpreted as part of the same trend. The view is that if Han, who comes from a “strategic marketing” background, takes over the Memory Business Unit and strengthens price-negotiation power for DRAM and NAND to maximize profitability, DS division operating profit could approach 100 trillion won next year and the year after.\n\nAccording to a source familiar with the company’s situation, Samsung Electronics plans to keep its existing HBM production target for next year, while expanding production of commodity DRAM. To this end, it is preparing a plan to further increase capacity for 10-nanometer-class 5th-generation (1b) DRAM and allocate a significant portion of that capacity to commodity DRAM. Internally, it is also reviewing the option of partially utilizing NAND flash production lines.\n\nA company official stated, “Next year, the absolute top priority is to maximize profitability,” adding, “We have recently been outperformed in technological competitiveness by SK hynix and have fallen behind, and at this point our competitiveness is not yet sufficient to aim for a rapid reversal through HBM alone.” The official went on, “Internally, the stance that we must demonstrate a clear advantage on the profitability front has become stronger,” and added, “The intention is to regain the number-one position in operating profit, overtake SK hynix, and project confidence to the market.”\n\nThe reason Samsung Electronics is seeking to boost profitability in commodity DRAM products such as DDR5, LPDDR5X, and GDDR7 is that their prices have been on a clearly rising trend recently. As the three memory players (Samsung Electronics, SK hynix, and Micron) have concentrated their DRAM capacity on HBM, supply of commodity DRAM has declined, creating strong upward pressure on prices. According to semiconductor distributor Fusion Worldwide, the November contract price of Samsung’s 32GB DDR5 memory chip module reached 239 dollars, up 60% from 149 dollars in September.\n\nIn contrast, Samsung Electronics’ HBM business is still viewed as lacking sufficient market position relative to competitors to expect high profitability. From next year, the price decline of the previous-generation HBM3E 12-high products is already taken as a given, and solid results in HBM4 will be essential to secure profitability. However, Samsung has yet to finalize an HBM4 supply agreement with Nvidia and still lacks sufficient market share. Even if Samsung were to receive the same price as SK hynix for HBM4, taking current yield and cost structures into account, profitability would still fall short of the level of commodity DRAM.\n\nAccordingly, Samsung Electronics plans to maintain a strategy of gradually taking share from SK hynix in HBM through technological improvements, while preparing a separate plan to maximize profitability in DRAM and NAND, where it already has strengths in market dominance. For HBM, the idea is to prioritize market share even at the cost of profitability, and to offset the resulting losses with earnings from commodity DRAM.\n\nA semiconductor industry source said, “Samsung Electronics manufactures the base die for HBM4 in-house, but compared with SK hynix, which outsources base-die production to TSMC, its cost structure is not that favorable.” The source added that while SK hynix has seen total HBM4 costs rise by around 30% compared with HBM3E, it has still managed to secure a similar level of price premium from Nvidia and defend its profitability. Samsung, meanwhile, is applying the 1c DRAM process to HBM4 core dies, but small-scale wafer test yields are reportedly only around 50%, meaning significant improvement is required before moving to mass production.\n\nAgainst this backdrop, there is talk that in the upcoming Samsung executive reshuffle, Han Jin-man will be positioned as the next candidate for head of the Memory Business Unit. Han served as Vice President of the Strategic Marketing Office in the Memory Business Unit in 2020 and is evaluated as having strong capabilities in sales, including price negotiations. Even in the Foundry Business Unit, which had relatively few big-tech references, he proved his performance by successively securing Apple and Tesla as customers.\n\nThe aforementioned company official said, “The reason Han has recently emerged as a candidate for Memory Business Unit chief is also his price-negotiation capability,” and continued, “The strategy is to use Han’s negotiation skills to extract as much profit as possible by seizing the opportunity presented by the sharp price surge in commodity DRAM next year.”\n\nIn fact, during its recent third-quarter earnings conference call, Samsung Electronics stated that it was “internally reviewing the possibility of increasing HBM production,” but at the same time noted, “With the recent rise in commodity DRAM prices, profitability is improving steeply, so we will determine the scale of additional expansions at an appropriate level while monitoring market conditions, taking into account relative profitability between HBM and commodity DRAM.”\n\nThe plan to further expand 1b DRAM capacity is expected to take the form of conversion investments on existing production lines. The 10-nanometer-class DRAM process has evolved in the order of 1x (first generation) → 1y (second generation) → 1z (third generation) → 1a (fourth generation) → 1b (fifth generation) → 1c (sixth generation). Samsung Electronics is expected to carry out conversion investments on matured-node lines such as 1z, and allocate a large portion of the additional capacity secured to commodity DRAM.\n\nThe company is also reviewing a plan to convert part of its NAND capacity to commodity DRAM. Samsung has displayed a rather conservative attitude toward NAND capacity expansion, stating in its third-quarter conference call that it would “focus on converting existing lines to advanced nodes” in NAND. Internally, it is reportedly even discussing a plan to reduce NAND capacity.\n\nA semiconductor industry official added, “Samsung Electronics has had many experiences in the past of suffering losses after being swayed by capacity-expansion demands from customers such as Apple, so it tends to be conservative about the NAND business,” and continued, “The company also believes the recent upward trend in NAND prices will not be sustainable for long, so it appears to be considering converting part of that capacity.”\n\nMeanwhile, the securities industry currently estimates next year’s operating profit at about 76.2045 trillion won for Samsung Electronics and 70.2742 trillion won for SK hynix. Foreign securities firm Morgan Stanley, in a report titled “Memory – Maximum Pricing Power” published on the 10th (local time), projected Samsung Electronics’ operating profit for next year at 116.448 trillion won, with the semiconductor (DS) division accounting for 94.625 trillion won of that. For 2027, it forecast operating profit of 135.22 trillion won, with semiconductor operating profit of 109.896 trillion won.\n\nBy contrast, SK hynix plans to maintain its management stance of making profitability maximization through HBM its top priority next year as well. Some in the industry expect SK hynix’s HBM operating margin to reach as high as 70%. In the Q&A session of its third-quarter conference call, SK hynix said regarding the possibility of adjusting capacity mix, “There is a possibility that general DRAM margins could become similar to those of HBM, but we will not immediately adjust our capacity mix based solely on temporary changes in profitability,” thereby officially declaring its policy of focusing on HBM.\n\nKim Dong-won, an analyst at KB Securities, stated, “We expect DDR5 margins to surpass those of HBM in 2026, leading to a reversal in profitability,” and added, “Samsung Electronics, which operates about 70% of its total DRAM production capacity as commodity DRAM, will be the direct beneficiary.”\n\n---\n\nhttps://t.co/APe7xxnb8v", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.028629816550031828, "ret_signed_m1d": 0.028629816550031828, "alpha_spy_m1d": 0.020161241466577717, "alpha_c_m1d": 0.012108339584125272, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.016521476965906556, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013292434936019282, "ret_signed_p1d": -0.013292434936019282, "alpha_spy_p1d": -0.01715563471960646, "alpha_c_p1d": -0.020245111658979686, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.006952676722960405, "ret_p1w": 0.015337381614012546, "ret_signed_p1w": 0.015337381614012546, "alpha_spy_p1w": -0.007296135352197863, "alpha_c_p1w": 0.010212503775267834, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.0051248778387447125, "ret_p1m": 0.10327185837897668, "ret_signed_p1m": 0.10327185837897668, "alpha_spy_p1m": 0.08612249864833599, "alpha_c_p1m": 0.10416788123293341, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": -0.000896022853956735, "ret_p3m": 0.8617671386246695, "ret_signed_p3m": 0.8617671386246695, "alpha_spy_p3m": 0.825885960729541, "alpha_c_p3m": 0.8599273854575797, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.0018397531670897305, "ret_p6m": 1.924041286961046, "ret_signed_p6m": 1.924041286961046, "alpha_spy_p6m": 1.7930622336816595, "alpha_c_p6m": 1.6529344925524478, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.2711067944085981, "price_path": [-0.2813, -0.2732, -0.2722, -0.2844, -0.2813, -0.2915, -0.2905, -0.3119, -0.2966, -0.2895, -0.2864, -0.2925, -0.2864, -0.2722, -0.261, -0.2528, -0.2325, -0.2213, -0.1917, -0.204, -0.1826, -0.1826, -0.15, -0.1378, -0.1307, -0.1235, -0.152, -0.1391, -0.1421, -0.1207, -0.0823, -0.0823, -0.0823, -0.0823, -0.0823, -0.0823, -0.0348, -0.046, -0.0634, -0.0286, -0.001, 0.001, 0.0031, -0.0031, 0.0082, -0.0133, 0.0102, 0.0429, 0.0174, 0.0276, 0.0644, 0.0992, 0.136, 0.0726, 0.0286, 0.0143, 0.001, 0.0286, 0.0583, 0.0542, 0.0511, -0.0061, 0.0286, 0.0, -0.0133, 0.0286, -0.0307, -0.0112, 0.0153, 0.0511, 0.0583, 0.0276, 0.0307, 0.0573, 0.0685, 0.0746, 0.1084, 0.1196, 0.1084, 0.1043, 0.0971, 0.1135, 0.0716, 0.0511, 0.1033, 0.1002, 0.0869, 0.1299, 0.1401, 0.136, 0.136, 0.1963, 0.2278, 0.2319, 0.2319, 0.2319, 0.3203, 0.4189, 0.4271, 0.4487, 0.4261, 0.4282, 0.4261, 0.4138, 0.4415, 0.4785, 0.5299, 0.534, 0.4919, 0.5361, 0.5648, 0.5628, 0.5628, 0.6388, 0.6686, 0.6511, 0.6491, 0.5453, 0.721, 0.7374, 0.6368, 0.6296, 0.7097, 0.7035, 0.7241, 0.8351, 0.8618, 0.8618, 0.8618, 0.8618, 0.9522, 0.9532, 0.983, 1.0549, 1.0909, 1.2399, 1.2245, 1.2245, 1.0046, 0.7693, 0.9686, 0.9337, 0.7827, 0.9306, 0.9522, 0.9306, 0.8854, 0.9388, 0.9923, 1.1423, 1.0601, 1.0488, 0.9142, 0.9491, 0.9419, 0.8505, 0.8505, 0.8152, 0.7215, 0.9526, 0.8368, 0.9171, 0.9881, 1.0231, 1.1673, 1.1004, 1.121, 1.0695, 1.1261, 1.1724, 1.2394, 1.2239, 1.2085, 1.2548, 1.2394, 1.3114, 1.26, 1.3114, 1.2857, 1.3269, 1.2703, 1.2703, 1.3938, 1.3938, 1.7387, 1.7953, 1.7645, 1.9395, 1.8726, 1.924]}
{"unit_id": "thread:1990921592118116488", "ticker_idx": 1, "ts": "2025-11-18T23:14:31+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has yet to finalize an HBM4 supply agreement with Nvidia and still lacks sufficient market share. Even if Samsung were to receive the same price as SK hynix for HBM4, taking current yield and cost structures into account, profitability would still fall short of the level of commodity DRAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung is the direct beneficiary of rising commodity DRAM prices given 70% of DRAM capacity there; SK Hynix faces margin pressure as HBM-focused strategy may be outpaced by commodity DRAM profitability in 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung puts more weight on expanding production of commodity memory than HBM… Aiming to maximize profitability\n\nSamsung Electronics has set the core goal of next year’s DS (Device Solutions) semiconductor business as “maximizing profitability” and has begun to reorganize its strategy accordingly, with Han Jin-man, head of the Foundry Business Unit, being mentioned as a candidate to head the Memory Business Unit. Until Samsung’s high-bandwidth memory (HBM) technology improves and selling prices can be fully maximized, the company is sketching out a scenario in which it focuses more heavily on commodity DRAM, where profitability can improve more quickly.\n\nRecently in the industry, Han Jin-man, who previously served as Head of Strategic Marketing Office in the Memory Business Unit, has been mentioned as a candidate for the Memory Business Unit chief, and this is being interpreted as part of the same trend. The view is that if Han, who comes from a “strategic marketing” background, takes over the Memory Business Unit and strengthens price-negotiation power for DRAM and NAND to maximize profitability, DS division operating profit could approach 100 trillion won next year and the year after.\n\nAccording to a source familiar with the company’s situation, Samsung Electronics plans to keep its existing HBM production target for next year, while expanding production of commodity DRAM. To this end, it is preparing a plan to further increase capacity for 10-nanometer-class 5th-generation (1b) DRAM and allocate a significant portion of that capacity to commodity DRAM. Internally, it is also reviewing the option of partially utilizing NAND flash production lines.\n\nA company official stated, “Next year, the absolute top priority is to maximize profitability,” adding, “We have recently been outperformed in technological competitiveness by SK hynix and have fallen behind, and at this point our competitiveness is not yet sufficient to aim for a rapid reversal through HBM alone.” The official went on, “Internally, the stance that we must demonstrate a clear advantage on the profitability front has become stronger,” and added, “The intention is to regain the number-one position in operating profit, overtake SK hynix, and project confidence to the market.”\n\nThe reason Samsung Electronics is seeking to boost profitability in commodity DRAM products such as DDR5, LPDDR5X, and GDDR7 is that their prices have been on a clearly rising trend recently. As the three memory players (Samsung Electronics, SK hynix, and Micron) have concentrated their DRAM capacity on HBM, supply of commodity DRAM has declined, creating strong upward pressure on prices. According to semiconductor distributor Fusion Worldwide, the November contract price of Samsung’s 32GB DDR5 memory chip module reached 239 dollars, up 60% from 149 dollars in September.\n\nIn contrast, Samsung Electronics’ HBM business is still viewed as lacking sufficient market position relative to competitors to expect high profitability. From next year, the price decline of the previous-generation HBM3E 12-high products is already taken as a given, and solid results in HBM4 will be essential to secure profitability. However, Samsung has yet to finalize an HBM4 supply agreement with Nvidia and still lacks sufficient market share. Even if Samsung were to receive the same price as SK hynix for HBM4, taking current yield and cost structures into account, profitability would still fall short of the level of commodity DRAM.\n\nAccordingly, Samsung Electronics plans to maintain a strategy of gradually taking share from SK hynix in HBM through technological improvements, while preparing a separate plan to maximize profitability in DRAM and NAND, where it already has strengths in market dominance. For HBM, the idea is to prioritize market share even at the cost of profitability, and to offset the resulting losses with earnings from commodity DRAM.\n\nA semiconductor industry source said, “Samsung Electronics manufactures the base die for HBM4 in-house, but compared with SK hynix, which outsources base-die production to TSMC, its cost structure is not that favorable.” The source added that while SK hynix has seen total HBM4 costs rise by around 30% compared with HBM3E, it has still managed to secure a similar level of price premium from Nvidia and defend its profitability. Samsung, meanwhile, is applying the 1c DRAM process to HBM4 core dies, but small-scale wafer test yields are reportedly only around 50%, meaning significant improvement is required before moving to mass production.\n\nAgainst this backdrop, there is talk that in the upcoming Samsung executive reshuffle, Han Jin-man will be positioned as the next candidate for head of the Memory Business Unit. Han served as Vice President of the Strategic Marketing Office in the Memory Business Unit in 2020 and is evaluated as having strong capabilities in sales, including price negotiations. Even in the Foundry Business Unit, which had relatively few big-tech references, he proved his performance by successively securing Apple and Tesla as customers.\n\nThe aforementioned company official said, “The reason Han has recently emerged as a candidate for Memory Business Unit chief is also his price-negotiation capability,” and continued, “The strategy is to use Han’s negotiation skills to extract as much profit as possible by seizing the opportunity presented by the sharp price surge in commodity DRAM next year.”\n\nIn fact, during its recent third-quarter earnings conference call, Samsung Electronics stated that it was “internally reviewing the possibility of increasing HBM production,” but at the same time noted, “With the recent rise in commodity DRAM prices, profitability is improving steeply, so we will determine the scale of additional expansions at an appropriate level while monitoring market conditions, taking into account relative profitability between HBM and commodity DRAM.”\n\nThe plan to further expand 1b DRAM capacity is expected to take the form of conversion investments on existing production lines. The 10-nanometer-class DRAM process has evolved in the order of 1x (first generation) → 1y (second generation) → 1z (third generation) → 1a (fourth generation) → 1b (fifth generation) → 1c (sixth generation). Samsung Electronics is expected to carry out conversion investments on matured-node lines such as 1z, and allocate a large portion of the additional capacity secured to commodity DRAM.\n\nThe company is also reviewing a plan to convert part of its NAND capacity to commodity DRAM. Samsung has displayed a rather conservative attitude toward NAND capacity expansion, stating in its third-quarter conference call that it would “focus on converting existing lines to advanced nodes” in NAND. Internally, it is reportedly even discussing a plan to reduce NAND capacity.\n\nA semiconductor industry official added, “Samsung Electronics has had many experiences in the past of suffering losses after being swayed by capacity-expansion demands from customers such as Apple, so it tends to be conservative about the NAND business,” and continued, “The company also believes the recent upward trend in NAND prices will not be sustainable for long, so it appears to be considering converting part of that capacity.”\n\nMeanwhile, the securities industry currently estimates next year’s operating profit at about 76.2045 trillion won for Samsung Electronics and 70.2742 trillion won for SK hynix. Foreign securities firm Morgan Stanley, in a report titled “Memory – Maximum Pricing Power” published on the 10th (local time), projected Samsung Electronics’ operating profit for next year at 116.448 trillion won, with the semiconductor (DS) division accounting for 94.625 trillion won of that. For 2027, it forecast operating profit of 135.22 trillion won, with semiconductor operating profit of 109.896 trillion won.\n\nBy contrast, SK hynix plans to maintain its management stance of making profitability maximization through HBM its top priority next year as well. Some in the industry expect SK hynix’s HBM operating margin to reach as high as 70%. In the Q&A session of its third-quarter conference call, SK hynix said regarding the possibility of adjusting capacity mix, “There is a possibility that general DRAM margins could become similar to those of HBM, but we will not immediately adjust our capacity mix based solely on temporary changes in profitability,” thereby officially declaring its policy of focusing on HBM.\n\nKim Dong-won, an analyst at KB Securities, stated, “We expect DDR5 margins to surpass those of HBM in 2026, leading to a reversal in profitability,” and added, “Samsung Electronics, which operates about 70% of its total DRAM production capacity as commodity DRAM, will be the direct beneficiary.”\n\n---\n\nhttps://t.co/APe7xxnb8v", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06315794566036503, "ret_signed_m1d": -0.06315794566036503, "alpha_spy_m1d": -0.05468937057691092, "alpha_c_m1d": -0.04212531794586649, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.021032627714498542, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014035037953128104, "ret_signed_p1d": 0.014035037953128104, "alpha_spy_p1d": 0.017898237736715283, "alpha_c_p1d": 0.03248742819980832, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.018452390246680217, "ret_p1w": -0.08947362807891579, "ret_signed_p1w": 0.08947362807891579, "alpha_spy_p1w": 0.1121071450451262, "alpha_c_p1w": 0.10963625688642387, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.020162628807508076, "ret_p1m": -0.03264096554147555, "ret_signed_p1m": 0.03264096554147555, "alpha_spy_p1m": 0.04979032527211624, "alpha_c_p1m": 0.050493210849571146, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": 0.017852245308095593, "ret_p3m": 0.5449654244903281, "ret_signed_p3m": -0.5449654244903281, "alpha_spy_p3m": -0.5090842465951997, "alpha_c_p3m": -0.3178399570909469, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.22712546739938122, "ret_p6m": 2.4755510784520043, "ret_signed_p6m": -2.4755510784520043, "alpha_spy_p6m": -2.344572025172618, "alpha_c_p6m": -1.7526032958378377, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.7229477826141666, "price_path": [-0.5708, -0.5603, -0.5454, -0.5419, -0.5445, -0.5289, -0.5281, -0.5509, -0.543, -0.5395, -0.5342, -0.5202, -0.514, -0.4947, -0.4667, -0.4614, -0.4237, -0.4193, -0.3895, -0.4149, -0.3763, -0.3763, -0.3842, -0.3667, -0.3728, -0.3746, -0.4096, -0.3877, -0.3904, -0.3684, -0.3061, -0.3061, -0.3061, -0.3061, -0.3061, -0.3061, -0.2491, -0.2719, -0.2781, -0.2588, -0.2061, -0.1833, -0.1482, -0.1596, -0.1553, -0.1605, -0.1053, -0.0614, -0.086, -0.0211, -0.0035, -0.0193, 0.0877, 0.0281, 0.0158, 0.0404, 0.0175, 0.0632, 0.086, 0.0825, 0.0737, -0.0175, 0.0632, 0.0, -0.014, 0.0018, -0.086, -0.0877, -0.0895, -0.0807, -0.0449, -0.0695, -0.0555, -0.0204, -0.0309, -0.0484, -0.0449, 0.013, -0.0063, 0.0306, -0.0081, 0.0025, -0.0274, -0.0695, -0.0326, -0.0309, -0.0397, 0.0183, 0.0253, 0.0323, 0.0323, 0.0516, 0.1236, 0.1429, 0.1429, 0.1429, 0.1886, 0.2219, 0.2746, 0.3027, 0.3273, 0.3062, 0.315, 0.2957, 0.3027, 0.315, 0.3273, 0.3413, 0.3044, 0.2992, 0.3255, 0.3466, 0.2922, 0.4045, 0.4765, 0.5116, 0.5959, 0.4572, 0.5924, 0.5801, 0.4783, 0.473, 0.5573, 0.5379, 0.5099, 0.559, 0.545, 0.545, 0.545, 0.545, 0.5695, 0.6661, 0.6696, 0.7644, 0.7872, 0.933, 0.8662, 0.8662, 0.6516, 0.4933, 0.6551, 0.6252, 0.4704, 0.6498, 0.6797, 0.6358, 0.6006, 0.7132, 0.7061, 0.8574, 0.7817, 0.7712, 0.641, 0.7343, 0.7501, 0.641, 0.641, 0.5355, 0.4194, 0.5777, 0.4599, 0.5408, 0.5584, 0.6111, 0.8169, 0.7554, 0.8064, 0.8292, 0.94, 0.9981, 1.0315, 0.984, 1.0509, 1.1529, 1.1511, 1.1546, 1.1494, 1.2725, 1.2865, 1.2742, 1.2619, 1.2619, 1.5451, 1.5451, 1.816, 1.9092, 1.9655, 2.3067, 2.2275, 2.4756]}
{"unit_id": "quote:1984158590983401874", "ticker_idx": 0, "ts": "2025-10-31T07:20:46+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek significantly raises its ASIC business targets", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bullish MediaTek on significantly raised ASIC revenue targets ($1B CY26, multi-billions CY27+).", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "MediaTek significantly raises its ASIC business targets", "reply_text": null, "parent_text": null, "quoted_text": "Reitarates ASIC revenue target of $1B in CY26 and multiple billions in CY27. More complex follow on project is expected to generate revenue from '28 and beyond.", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0032695555510358076, "alpha_c_m1d": -0.001955890016814843, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": 0.001955890016814843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007633571108659365, "ret_signed_p1d": -0.007633571108659365, "alpha_spy_p1d": -0.009510258196623878, "alpha_c_p1d": -0.01611800511149608, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.008484434002836716, "ret_p1w": -0.03816795053310651, "ret_signed_p1w": -0.03816795053310651, "alpha_spy_p1w": -0.021908278885399923, "alpha_c_p1w": 0.0028765625351127033, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04104451306821921, "ret_p1m": 0.10305349493633065, "ret_signed_p1m": 0.10305349493633065, "alpha_spy_p1m": 0.1056778689914849, "alpha_c_p1m": 0.13076542237427835, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.027711927437947703, "ret_p3m": 0.38511861140096193, "ret_signed_p3m": 0.38511861140096193, "alpha_spy_p3m": 0.36251817988802615, "alpha_c_p3m": 0.23386963808721495, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.15124897331374698, "ret_p6m": 0.8948111341355856, "ret_signed_p6m": 0.8948111341355856, "alpha_spy_p6m": 0.8402966725844969, "alpha_c_p6m": 0.49589288618937966, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.3989182479462059, "price_path": [0.0229, 0.0076, 0.0344, 0.0305, 0.0344, 0.0382, 0.0458, 0.0611, 0.0458, 0.0763, 0.0611, 0.0344, 0.042, 0.042, 0.0763, 0.0687, 0.0687, 0.0573, 0.0458, 0.0382, 0.0382, 0.0649, 0.0573, 0.0954, 0.0916, 0.1527, 0.1374, 0.1298, 0.1336, 0.1336, 0.1718, 0.1527, 0.1527, 0.0992, 0.0725, 0.0496, 0.0153, 0.0267, 0.0, 0.0, 0.0038, -0.0191, -0.0229, 0.0076, 0.0076, 0.0115, 0.0191, 0.0267, 0.0267, 0.0038, -0.0038, 0.0038, 0.0153, 0.0153, 0.0229, 0.0229, 0.0153, -0.0115, -0.0115, 0.0191, 0.0038, -0.0076, 0.0, 0.0, -0.0076, 0.0038, 0.0153, -0.0153, -0.0382, -0.0458, -0.0534, -0.0496, -0.0496, -0.0611, -0.0611, -0.1069, -0.1145, -0.0954, -0.126, -0.1221, -0.0954, -0.0076, 0.0229, 0.0649, 0.1031, 0.0802, 0.0687, 0.0725, 0.0878, 0.0992, 0.084, 0.1145, 0.0649, 0.0725, 0.084, 0.084, 0.0878, 0.084, 0.0763, 0.0687, 0.0611, 0.0534, 0.0534, 0.0573, 0.084, 0.084, 0.0916, 0.0916, 0.1221, 0.1641, 0.1556, 0.1595, 0.1244, 0.105, 0.1244, 0.1556, 0.1672, 0.1595, 0.1711, 0.1556, 0.1556, 0.14, 0.1556, 0.2684, 0.3773, 0.3812, 0.3851, 0.3851, 0.3696, 0.3268, 0.3968, 0.4007, 0.3773, 0.3306, 0.424, 0.4357, 0.4435, 0.4435, 0.4435, 0.4435, 0.4435, 0.4435, 0.4435, 0.4435, 0.3968, 0.4201, 0.4552, 0.5135, 0.5135, 0.4785, 0.4124, 0.3384, 0.3812, 0.3734, 0.2956, 0.3268, 0.3734, 0.389, 0.3384, 0.3306, 0.3462, 0.3462, 0.3073, 0.3229, 0.2645, 0.2606, 0.2606, 0.2373, 0.2334, 0.175, 0.1595, 0.14, 0.14, 0.14, 0.14, 0.1439, 0.2295, 0.2256, 0.2256, 0.2606, 0.3384, 0.3929, 0.4746, 0.498, 0.4785, 0.6263, 0.7859, 0.7236, 0.8948, 0.8948]}
{"unit_id": "orig:1988220681281310853", "ticker_idx": 0, "ts": "2025-11-11T12:22:04+00:00", "symbol": "CRWV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DA DAVIDSON) CoreWeave, Inc. Price Target $36, Underperform", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares DA Davidson Underperform rating on CoreWeave PT $36, with detailed bearish thesis on negative return on capital, customer concentration risk, and heavy insider selling.", "resolved_tickers": ["CRWV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "DA DAVIDSON) CoreWeave, Inc. Price Target $36, Underperform\n\nWill the order backlog convert (into revenue)? Must it?\nWe understand we may not get a chance to ask questions today (perhaps we should have behaved more kindly). So for CoreWeave, which reports earnings this evening, we’ve listed several good questions others might consider asking. We fully expect the company to report better-than-expected revenue and profits, but we question how important that is in light of the answers to the questions below.\n    \n1. Where do you rank within OpenAI’s spending? Since your last earnings report, your key customer (OpenAI) has announced several other contracts—most notably $300 billion with Oracle, $250 billion with Microsoft, and $38 billion with Amazon. If, for any reason, OpenAI cannot fulfill all $1.4 trillion of its commitments, how could you get paid ahead of Microsoft, Amazon, and Oracle?\n    \n2. What is your return on capital? By our calculation, last quarter’s return was 5% (annualizing $200 million of adjusted EBIT on $16.6 billion of PP&E). Your contract terms are fixed up front and you have already achieved scale; how can you improve this return toward peers’ 15–20%?\n    \n3. What is your cost of capital? By our calculation, last quarter’s weighted average cost was 10% (annualizing $267 million of interest expense on $11.0 billion of debt). We can see you are gradually lowering this with additional sub-9% debt, but can you bring it down to Meta- or Oracle-like levels as the likelihood of your primary customer honoring its commitments appears increasingly uncertain?\n    \n4. If the answer to Question 3 (cost of capital) remains higher than Question 2 (return on capital), why would an investor want to buy the stock of a company generating negative returns?\n    \n5. Do you have sufficient capital to build out to the level promised to OpenAI? If, for any reason, you cannot, wouldn’t they be able to cancel the contract? Doesn’t your S-1 say so? If so, what happens next?\n    \n6. If you still believe in what you are doing, why are you selling so much (see pages 2–15)? In 25 years of doing this, we don’t recall seeing this many Form 4 (insider sale) filings over a three-month period.\n\n$CRWV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.19481843352880257, "ret_signed_m1d": -0.19481843352880257, "alpha_spy_m1d": -0.1971023549931078, "alpha_c_m1d": -0.1971023549931078, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": -0.0022839214643052275, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03348794099911834, "ret_signed_p1d": 0.03348794099911834, "alpha_spy_p1d": 0.034044339605560414, "alpha_c_p1d": 0.034044339605560414, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.0005563986064420767, "ret_p1w": -0.15261905144157495, "ret_signed_p1w": 0.15261905144157495, "alpha_spy_p1w": 0.1190613222776512, "alpha_c_p1w": 0.1190613222776512, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.033557729163923744, "ret_p1m": -0.002602055991936081, "ret_signed_p1m": 0.002602055991936081, "alpha_spy_p1m": 0.009293218816492588, "alpha_c_p1m": 0.009293218816492588, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.006691162824556507, "ret_p3m": 0.01764902782402822, "ret_signed_p3m": -0.01764902782402822, "alpha_spy_p3m": -0.0035044962984436268, "alpha_c_p3m": -0.0035044962984436268, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.014144531525584592, "ret_p6m": 0.5610362787681864, "ret_signed_p6m": -0.5610362787681864, "alpha_spy_p6m": -0.4804967941041982, "alpha_c_p6m": -0.4804967941041982, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.08053948466398819, "price_path": [0.1257, 0.131, 0.0951, 0.0509, 0.0354, 0.0272, 0.0634, 0.0451, 0.0339, 0.0966, 0.1629, 0.1657, 0.1657, 0.056, 0.0169, -0.0103, 0.0079, 0.0584, 0.1338, 0.3253, 0.2749, 0.2667, 0.3629, 0.3435, 0.3673, 0.3733, 0.4126, 0.5073, 0.4808, 0.5092, 0.433, 0.3615, 0.3861, 0.5483, 0.5505, 0.5613, 0.5249, 0.5143, 0.4575, 0.5837, 0.6187, 0.5661, 0.6022, 0.5167, 0.5753, 0.6036, 0.5485, 0.4375, 0.4149, 0.3749, 0.3954, 0.4996, 0.5393, 0.5251, 0.5831, 0.4827, 0.5127, 0.4291, 0.3095, 0.2945, 0.2098, 0.1767, 0.1948, 0.0, -0.0335, -0.1137, -0.1248, -0.1478, -0.1526, -0.1524, -0.217, -0.1894, -0.1673, -0.1935, -0.1595, -0.1595, -0.1728, -0.1282, -0.1398, -0.1022, -0.0299, -0.001, -0.0243, 0.0257, -0.0026, -0.0114, -0.1109, -0.1815, -0.2137, -0.2697, -0.2343, -0.061, -0.0403, -0.092, -0.1077, -0.1077, -0.1354, -0.1524, -0.1639, -0.1898, -0.1898, -0.1026, -0.1304, -0.1182, -0.1268, -0.1278, -0.0933, 0.0174, -0.0103, 0.016, 0.0749, 0.1453, 0.1453, 0.0773, 0.064, 0.0385, 0.0519, 0.1122, 0.2316, 0.1995, 0.126, 0.0543, 0.0062, 0.0189, -0.0671, -0.1554, 0.0176, 0.095, 0.076, 0.0765, 0.0827, 0.0865, 0.0865, 0.0295, 0.0799, 0.099, 0.0097, 0.0277, 0.1234, 0.1088, 0.1045, -0.0999, -0.117, -0.1653, -0.1006, -0.1535, -0.1742, -0.1582, -0.1524, -0.0727, -0.0965, -0.0824, -0.0286, -0.0709, -0.063, -0.0875, -0.0783, -0.0727, -0.0608, -0.0092, -0.0898, -0.1536, -0.2177, -0.1235, -0.1126, -0.0696, -0.0696, -0.0843, -0.0356, 0.0058, 0.0408, 0.154, 0.2475, 0.3259, 0.3428, 0.3526, 0.322, 0.3285, 0.3029, 0.3864, 0.3284, 0.2461, 0.2678, 0.1939, 0.2919, 0.2626, 0.3464, 0.4191, 0.4469, 0.561]}
{"unit_id": "reply:2000476821058683082", "ticker_idx": 0, "ts": "2025-12-15T08:03:35+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron will generate solid profits even without HBM4. Why sell?", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Micron; solid profits expected even without HBM4, no reason to sell.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@DhaniSriram Micron will generate solid profits even without HBM4. Why sell?", "reply_text": null, "parent_text": "@jukan05 Would you sell Micron or you’re still bullish because of the memory super cycle?", "quoted_text": null, "parent_author": "DhaniSriram", "ret_m1d": 0.015326278026813522, "ret_signed_m1d": 0.015326278026813522, "alpha_spy_m1d": 0.013813194179786104, "alpha_c_m1d": 0.01186920406305747, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021010569585278582, "ret_signed_p1d": -0.021010569585278582, "alpha_spy_p1d": -0.018278271989999872, "alpha_c_p1d": -0.018290226336253657, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.1645894394682299, "ret_signed_p1w": 0.1645894394682299, "alpha_spy_p1w": 0.15559371415552747, "alpha_c_p1w": 0.14214894979603554, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4242803264561392, "ret_signed_p1m": 0.4242803264561392, "alpha_spy_p1m": 0.4021128365504283, "alpha_c_p1m": 0.31136370124665924, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.7074262303029928, "ret_signed_p3m": 0.7074262303029928, "alpha_spy_p3m": 0.7260854035676849, "alpha_c_p3m": 0.604174023515615, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3268, -0.2893, -0.3152, -0.3073, -0.2998, -0.3195, -0.3401, -0.3382, -0.3103, -0.2959, -0.2335, -0.2268, -0.2091, -0.196, -0.2181, -0.1725, -0.1902, -0.2354, -0.1883, -0.2124, -0.1918, -0.1472, -0.1479, -0.1294, -0.1483, -0.1643, -0.1296, -0.0778, -0.0733, -0.0656, -0.0458, -0.0568, -0.0578, -0.0118, -0.082, 0.0, 0.0035, 0.0018, 0.0665, 0.0152, 0.0312, -0.0023, 0.0393, 0.0187, -0.0379, -0.0488, -0.1521, -0.1269, -0.0571, -0.0546, -0.0305, -0.0305, -0.0043, 0.0125, 0.0084, -0.0141, -0.0457, -0.0012, 0.0397, 0.0628, 0.1104, 0.0883, 0.0153, 0.0, -0.021, -0.0504, 0.0465, 0.1197, 0.1646, 0.1632, 0.2071, 0.2071, 0.1991, 0.24, 0.2326, 0.2022, 0.2022, 0.3286, 0.3148, 0.4466, 0.4303, 0.3775, 0.4536, 0.4569, 0.4243, 0.4041, 0.418, 0.528, 0.528, 0.5375, 0.639, 0.6747, 0.6834, 0.6389, 0.728, 0.8335, 0.8356, 0.7476, 0.8441, 0.7668, 0.5981, 0.6128, 0.6625, 0.6154, 0.5722, 0.7284, 0.7437, 0.734, 0.734, 0.684, 0.7731, 0.758, 0.8035, 0.7732, 0.7608, 0.807, 0.7504, 0.737, 0.7383, 0.5993, 0.6881, 0.6725, 0.5598, 0.6399, 0.698, 0.7636, 0.7074, 0.795, 0.861, 0.9447, 0.9449, 0.8714, 0.7814, 0.7032, 0.6661, 0.6094, 0.4973, 0.5047, 0.3561, 0.4237, 0.5501, 0.5433, 0.5433, 0.5919, 0.5911, 0.714, 0.7762, 0.7724, 0.7975, 0.9623, 0.9226, 0.9268, 0.9177, 0.8896, 0.8937, 1.0542, 1.03, 1.0932, 1.2105, 1.1251, 1.1848, 1.1793, 1.2849, 1.4292, 1.6978, 1.809, 1.7249, 2.1471, 2.3515, 2.2304, 2.3865, 2.2701, 2.0537, 1.872, 1.9445, 2.0846, 2.2115, 2.1647, 2.1647, 2.7752, 2.9123, 2.8917, 3.0918, null, null, null, null, null, null, null]}
{"unit_id": "quote:1981524615454478462", "ticker_idx": 0, "ts": "2025-10-24T00:54:17+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "That's a negative development for Advantest.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Korean firms displacing Advantest in SK Hynix HBM4 test equipment supply chain; bearish for Advantest.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "According to Korean media reports, Korean companies are making a large-scale entry into SK hynix’s HBM4 test equipment supply chain.\nThat’s a negative development for Advantest.", "reply_text": null, "parent_text": null, "quoted_text": "SK hynix’s HBM4 Test Equipment Supply Chain Takes Shape… Korean Equipment Makers Gaining Ground\n\nSK hynix is preparing investments in test equipment for full-scale HBM4 mass production next year. The company is expected to complete supply-chain build-out by the end of the year, https://t.co/3nubPrnMzA", "parent_author": null, "ret_m1d": -0.03600702753772922, "ret_signed_m1d": 0.03600702753772922, "alpha_spy_m1d": 0.027900705922892732, "alpha_c_m1d": 0.0179514386840951, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.01805558885363412, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06528102994215201, "ret_signed_p1d": -0.06528102994215201, "alpha_spy_p1d": -0.05348331791900307, "alpha_c_p1d": -0.040219883772402154, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.02506114616974986, "ret_p1w": 0.3545082011795848, "ret_signed_p1w": -0.3545082011795848, "alpha_spy_p1w": -0.3474059363056461, "alpha_c_p1w": -0.32067565820944366, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.033832542970141155, "ret_p1m": 0.07230685006297333, "ret_signed_p1m": -0.07230685006297333, "alpha_spy_p1m": -0.08488719703445702, "alpha_c_p1m": -0.10653833065246987, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.034231480589496543, "ret_p3m": 0.27459017003522135, "ret_signed_p3m": -0.27459017003522135, "alpha_spy_p3m": -0.2595657542719836, "alpha_c_p3m": -0.12638545470763862, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": 0.14820471532758273, "ret_p6m": 0.58881071989207, "ret_signed_p6m": -0.58881071989207, "alpha_spy_p6m": -0.5363848733071908, "alpha_c_p6m": -0.2634031623083104, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.3254075575837596, "price_path": [-0.395, -0.4014, -0.3952, -0.4031, -0.4128, -0.4131, -0.411, -0.4093, -0.3874, -0.3874, -0.3485, -0.3131, -0.345, -0.34, -0.3304, -0.3333, -0.371, -0.3605, -0.371, -0.3643, -0.3628, -0.3342, -0.3245, -0.3178, -0.3719, -0.3768, -0.3751, -0.3456, -0.3316, -0.3023, -0.2573, -0.2331, -0.1995, -0.1849, -0.1849, -0.1744, -0.191, -0.151, -0.1224, -0.0943, -0.0943, -0.0999, -0.1218, -0.1545, -0.1133, -0.1423, -0.1408, -0.1191, -0.0814, 0.0474, 0.0542, 0.0518, 0.0542, 0.0442, 0.0442, -0.0047, 0.0167, 0.024, -0.0123, 0.0252, 0.0187, 0.0012, -0.036, 0.0, 0.0653, 0.0609, 0.2951, 0.3039, 0.3545, 0.3545, 0.2752, 0.1994, 0.2371, 0.1686, 0.2131, 0.1636, 0.161, 0.2102, 0.1434, 0.171, 0.1276, 0.1212, 0.2198, 0.0723, 0.0723, 0.1171, 0.1393, 0.195, 0.2046, 0.1534, 0.1598, 0.2213, 0.2119, 0.1827, 0.1856, 0.1859, 0.1797, 0.2319, 0.2169, 0.1388, 0.1227, 0.1388, 0.101, 0.1238, 0.1739, 0.1519, 0.1803, 0.159, 0.1853, 0.1581, 0.1496, 0.1496, 0.1496, 0.1496, 0.2398, 0.262, 0.2064, 0.1771, 0.1859, 0.1859, 0.2872, 0.3501, 0.3167, 0.3349, 0.2974, 0.2579, 0.2746, 0.3378, 0.3776, 0.3803, 0.4611, 0.4953, 0.5726, 0.4933, 0.423, 0.524, 0.4915, 0.4198, 0.4362, 0.6016, 0.6203, 0.6203, 0.5697, 0.5884, 0.5867, 0.5688, 0.5785, 0.522, 0.4915, 0.4915, 0.5585, 0.6754, 0.6467, 0.572, 0.5105, 0.5061, 0.4344, 0.495, 0.5053, 0.3393, 0.4095, 0.457, 0.4344, 0.3847, 0.4133, 0.3785, 0.4713, 0.404, 0.404, 0.3308, 0.3083, 0.3715, 0.3446, 0.2922, 0.2268, 0.1919, 0.3191, 0.2385, 0.2637, 0.2857, 0.3015, 0.4786, 0.454, 0.4651, 0.4587, 0.5829, 0.6172, 0.6782, 0.634, 0.5888]}
{"unit_id": "orig:1981572864152363047", "ticker_idx": 0, "ts": "2025-10-24T04:06:01+00:00", "symbol": "Doosan", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "My pick: Doosan", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long Doosan as top CCL pick on M9-driven growth.", "resolved_tickers": ["006280.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Electronics CCL supplier context 006280.KS", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "original", "lang": "en", "tweet_text": "Lately, I’ve been researching CCL (Copper Clad Laminate) companies.\n\nM9 is going to drive tremendous growth.\n\nMy pick: Doosan. https://t.co/DzmlwgmTLH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.008106321614836487, "alpha_c_m1d": 0.021510080132933984, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.021510080132933984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.030279437845425416, "ret_signed_p1d": 0.030279437845425416, "alpha_spy_p1d": 0.018481725822276474, "alpha_c_p1d": 0.005727614655501201, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.024551823189924216, "ret_p1w": 0.006987600281033179, "ret_signed_p1w": 0.006987600281033179, "alpha_spy_p1w": -0.00011466459290554454, "alpha_c_p1w": -0.04678751530655889, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.05377511558759207, "ret_p1m": 0.013198732459013263, "ret_signed_p1m": 0.013198732459013263, "alpha_spy_p1m": 0.025779079430496954, "alpha_c_p1m": 0.03221011384747452, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.019011381388461257, "ret_p3m": 0.1715838895550399, "ret_signed_p3m": 0.1715838895550399, "alpha_spy_p3m": 0.15655947379180213, "alpha_c_p3m": -0.11312844695752067, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": 0.28471233651256056, "ret_p6m": 0.1582448610603635, "ret_signed_p6m": 0.1582448610603635, "alpha_spy_p6m": 0.10581901447548425, "alpha_c_p6m": -0.5089042291793979, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.6671490902397614, "price_path": [0.1312, 0.1289, 0.1281, 0.0668, 0.0543, 0.073, 0.0644, 0.0691, 0.045, 0.0427, 0.0388, 0.0512, 0.045, 0.045, 0.0342, 0.0171, 0.0171, 0.0116, 0.014, 0.0419, 0.0334, 0.0334, 0.0233, 0.007, -0.0054, 0.0008, 0.0179, 0.0241, 0.0171, 0.0388, 0.0295, 0.0326, 0.0349, 0.0404, 0.0458, 0.0342, 0.0225, 0.0194, 0.0194, 0.0831, 0.0963, 0.0707, 0.0629, 0.0225, 0.0155, 0.0031, 0.0039, 0.0272, 0.0272, 0.0272, 0.0272, 0.0272, 0.0272, 0.0109, 0.0054, -0.0008, 0.014, 0.0047, -0.0047, 0.0148, 0.0148, 0.0264, 0.0, 0.0, 0.0303, 0.0349, 0.0179, 0.0016, 0.007, 0.0148, 0.0194, 0.0217, 0.0303, -0.0031, 0.0241, 0.021, 0.0629, 0.0815, 0.0823, 0.0753, 0.0404, 0.028, 0.0396, 0.014, 0.0132, 0.007, 0.0427, 0.0427, 0.0528, 0.0652, 0.0675, 0.0714, 0.0652, 0.0512, 0.0334, 0.038, 0.0489, 0.1731, 0.2578, 0.2958, 0.2795, 0.2834, 0.2927, 0.2966, 0.2741, 0.2399, 0.2174, 0.2174, 0.1902, 0.1592, 0.2345, 0.2345, 0.2345, 0.2376, 0.25, 0.25, 0.2492, 0.229, 0.2143, 0.2422, 0.25, 0.2648, 0.3238, 0.2446, 0.2306, 0.2306, 0.1716, 0.1856, 0.2539, 0.2663, 0.3393, 0.3207, 0.3144, 0.2803, 0.2166, 0.2679, 0.2717, 0.3199, 0.2756, 0.3152, 0.3129, 0.3571, 0.3998, 0.3533, 0.3533, 0.3533, 0.3533, 0.3781, 0.3401, 0.3439, 0.3734, 0.309, 0.2927, 0.3129, 0.3129, 0.2585, 0.0769, 0.1615, 0.1809, 0.1196, 0.1762, 0.1755, 0.1584, 0.1817, 0.1607, 0.1514, 0.1716, 0.1273, 0.1405, 0.1087, 0.1219, 0.1568, 0.1514, 0.1724, 0.1371, 0.1206, 0.1645, 0.1065, 0.1026, 0.0979, 0.0853, 0.1183, 0.1316, 0.1316, 0.1159, 0.1292, 0.1606, 0.1731, 0.1857, 0.1582]}
{"unit_id": "quote:1995705226188521479", "ticker_idx": 0, "ts": "2025-12-02T04:02:58+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GFHK believes Intel could win projects on 14A for Apple's non-Pro smartphone SoC, Nvidia and AMD's server projects, and a Tier-1 Ethernet switch vendor.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish INTC on potential 14A foundry wins with Apple, Nvidia, AMD, and a Tier-1 Ethernet vendor.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Very interesting.\nGFHK believes Intel could win projects on 14A for Apple’s non-Pro smartphone SoC, Nvidia and AMD’s server projects, and a Tier-1 Ethernet switch vendor.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": "Turn out it's many things to shock the market but not negatively Lol \n\nGFHK: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple", "parent_author": null, "ret_m1d": -0.07959518753181105, "ret_signed_m1d": -0.07959518753181105, "alpha_spy_m1d": -0.0777464025943061, "alpha_c_m1d": -0.061594491184534084, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.018000696347276968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006671200659915444, "ret_signed_p1d": 0.006671200659915444, "alpha_spy_p1d": 0.0032084328666550643, "alpha_c_p1d": -0.00707280041390379, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.013744001073819234, "ret_p1w": -0.06832300752935394, "ret_signed_p1w": -0.06832300752935394, "alpha_spy_p1w": -0.07053845118369928, "alpha_c_p1w": -0.09494849636071345, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02662548883135951, "ret_p1m": -0.15113870509152816, "ret_signed_p1m": -0.15113870509152816, "alpha_spy_p1m": -0.15466749126834134, "alpha_c_p1m": -0.15620391610621576, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": 0.005065211014687598, "ret_p3m": 0.049229338153991886, "ret_signed_p3m": 0.049229338153991886, "alpha_spy_p3m": 0.03971107240739147, "alpha_c_p3m": -0.08488428568970208, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": 0.13411362384369396, "ret_p6m": 1.8012420801375995, "ret_signed_p6m": 1.8012420801375995, "alpha_spy_p6m": 1.6938321515491355, "alpha_c_p6m": 1.1392968796703244, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.661945200467275, "price_path": [-0.4339, -0.4366, -0.4369, -0.4378, -0.4302, -0.4339, -0.4461, -0.4302, -0.4187, -0.4272, -0.2968, -0.3195, -0.3384, -0.3251, -0.2818, -0.2181, -0.1833, -0.2068, -0.2282, -0.1732, -0.1419, -0.1527, -0.1583, -0.1449, -0.1389, -0.1304, -0.1633, -0.1438, -0.1804, -0.1454, -0.1525, -0.1486, -0.1235, -0.1231, -0.1507, -0.1222, -0.1194, -0.0904, -0.0446, -0.049, -0.0761, -0.0801, -0.0913, -0.1481, -0.1171, -0.1433, -0.1228, -0.1155, -0.1286, -0.1284, -0.1739, -0.1829, -0.2015, -0.2103, -0.1923, -0.2266, -0.2063, -0.1767, -0.1758, -0.1532, -0.1532, -0.0669, -0.0796, 0.0, 0.0067, -0.0683, -0.0474, -0.0729, -0.0683, -0.0619, -0.0911, -0.1302, -0.1371, -0.1417, -0.1707, -0.1654, -0.153, -0.1633, -0.1638, -0.1682, -0.1682, -0.1672, -0.1562, -0.1419, -0.1511, -0.1511, -0.0941, -0.0943, -0.0789, -0.0193, -0.0543, 0.0478, 0.0136, 0.0879, 0.1208, 0.1116, 0.0803, 0.0803, 0.1171, 0.248, 0.2496, 0.0368, -0.0225, 0.0106, 0.1222, 0.1194, 0.069, 0.1228, 0.133, 0.118, 0.1097, 0.1638, 0.1557, 0.0842, 0.1109, 0.0692, 0.0764, 0.0764, 0.0623, 0.0458, 0.0265, 0.0147, 0.0037, 0.061, 0.0784, 0.0458, 0.0492, 0.0467, -0.0085, 0.0485, 0.0571, -0.0012, 0.0485, 0.0761, 0.1037, 0.0409, 0.0529, 0.0527, 0.0136, 0.0359, 0.0623, 0.0092, 0.0124, 0.0136, 0.0853, 0.0145, -0.0078, -0.0525, 0.0152, 0.1049, 0.159, 0.159, 0.1682, 0.2172, 0.3561, 0.4198, 0.435, 0.4994, 0.4679, 0.4939, 0.5758, 0.5758, 0.5114, 0.5243, 0.5015, 0.5362, 0.8988, 0.9551, 0.9443, 1.1797, 1.1735, 1.2917, 1.2034, 1.4879, 1.5997, 1.5217, 1.8737, 1.9777, 1.7746, 1.7672, 1.6669, 1.5022, 1.4884, 1.5489, 1.7366, 1.726, 1.7568, 1.7568, 1.8415, 1.8012]}
{"unit_id": "thread:1983713013757784375", "ticker_idx": 0, "ts": "2025-10-30T01:50:12+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Demand remains solid, and inventory levels are expected to fall below normal levels, so we expect a wider increase in memory prices. We will continue to operate with a profitability-focused approach.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung's 3Q25 call signals rising memory prices, HBM3E bit shipments up 80%+ QoQ, HBM4 in mass production prep, and improving foundry profitability — all pointing to continued earnings recovery.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics 3Q25 Earnings Conference Call Key Q&A (1)\n\n1. What drove the strong 3Q performance, and what is the outlook for 4Q?\n\n• Data-center investments expanded far more than expected due to increasing inference demand, leading to strong server demand. Market prices are rising in mobile and PC as well. Our company also expanded sales mainly in AI-related products.\n\nDRAM shipment volume grew by the mid-teens %, led by D5 and GD7, while server SSD shipments increased by around 10 %. Inventory decreased, and results exceeded guidance. DRAM prices rose by the mid-teens % thanks to a higher sales mix of HBM, and NAND prices increased by the mid-single digits.\n\n• Demand remains solid, and inventory levels are expected to fall below normal levels, so we expect a wider increase in memory prices. We will continue to operate with a profitability-focused approach. We will expand 3E demand and respond to high-capacity server D5 and LPD5. Shipments are expected to increase by a single-digit %. NAND shipments are expected to decline by about 10 %, though SSD demand should remain strong.\n\n2. Update on share buybacks?\n\n• The program was fully completed at the end of September, earlier than the initially announced target and disclosure deadline. In addition to the annual KRW 9.8 trillion dividend, we successfully completed the share repurchase. We recognize investors’ interest in further shareholder-return plans and are considering ways to enhance long-term shareholder value. We plan to retire part of the repurchased shares as soon as possible.\n\n---\n\nSamsung Electronics 3Q25 Earnings Conference Call Key Q&A (2)\n\n3. Update on the HBM business?\n\n• Demand is increasing faster than supply, and we are expanding mass production and sales of HBM3E to all customers. Compared to the previous quarter, bit shipments expanded by more than 80%. Development of HBM4 products has been completed, and samples have been shipped to customers. Mass production shipment preparation is complete. Recently, as performance competition among customers has intensified, we set and prepared performance targets that exceed customer requirements in advance.\n\nAI performance enhancement requirements are increasing. We will continue to actively execute capacity investments for the 1c nanometer node.\n\nProduction plans for 2026 will be significantly expanded compared to this year, with customer demand already secured. However, as additional customer requests are being received, we are internally reviewing the possibility of further expansion. Profitability of conventional products is also improving, so we will expand while considering relative profitability and overall market conditions.\n\n4. Outlook for the 2026 memory market?\n\n• Memory growth momentum will continue next year. High-performance HBM4 will fully arrive, and server DRAM will continue to move toward higher capacities. Supply constraints will also persist. NAND demand will increase mainly for high-capacity products. As QLC demand strengthens, inventory will approach bottom levels earlier than expected. However, we are closely monitoring geopolitical issues. We will lead the market and respond proactively.\n\n5. Stock compensation program\n\n• Announced a program for creating long-term performance. The stock compensation system will be expanded from executives to all employees, with the aim of establishing a foundation for motivation and shared growth. The number of shares will be determined based on performance and future stock price increases, and distributed in installments over three years. Since January 2025, the stock grant system has been implemented to strengthen accountable management, and it is now being extended to all employees. Of the 8.4 trillion KRW in treasury shares, an appropriate portion will be retired at the right time, and the remainder will be used for the stock compensation program. The timing of additional share repurchases will be decided by the board of directors.\n\n---\n\nSamsung Electronics 3Q25 Earnings Conference Call Key Q&A (3)\n\n6. Background of reduced losses in the Foundry business?\n\n• In the first half of this year, one-off costs such as provisions for unsellable products due to the impact of advanced chips were reflected. Performance was sluggish due to sales of high-cost products and low utilization rates. As these factors have disappeared and utilization rates have improved, the loss has been significantly reduced.\n\nGoing forward, with the full-scale mass production of the first-generation 2nm process and the expansion of HPC and automotive products for major U.S. and Chinese customers, sales are expected to increase. Further performance improvement is anticipated through continuous utilization rate improvement and cost-efficiency activities.\n\n7. Profitability outlook for the smartphone division?\n\n• Memory prices began to rebound in earnest from the third quarter, and the rate of increase will expand in the next quarter. Despite rising cost burdens, sales will increase centered on the foldable lineup. We will also expand the Tab and Watch series to strengthen our dominance in the premium market. Process optimization efforts are also underway.\n\n8. Direction of memory investment?\n\n• Over the past few years, we have consistently invested in cleanroom capacity. We will continue facility investments in line with demand. For DRAM, facility and construction investments based on 1b and 1c nodes will increase to some extent. For NAND, the proportion of advanced process nodes will be gradually expanded.\n\n---\n\nSamsung Electronics 3Q25 Earnings Conference Call Key Q&A (4)\n\n9. Background of reduced foundry investment this year?\n\n• This year, although we invested in advanced processes, mass production lines mainly focused on supplementary investments for existing lines. In 2026, we will maintain a flexible investment stance while completing facility investments for the new Taylor fab in the U.S. We plan to expand investments to a level similar to 2024 while simultaneously preparing for the mass production of new processes such as the second-generation 2nm.\n\n10. Future display strategy?\n\n• OLED sales are expanding across various segments such as IT, automotive, and wearables. Demand for high-resolution displays is increasing and expected to continue. In 2026, we will expand production of competitive products using the new 8.6-generation line and strengthen our responsiveness to customers. For automotive displays, we will leverage competitiveness in rigid OLEDs to expand into volume zones and apply differentiated technologies. We will maintain a stable business portfolio.\n\n11. Exynos\n\n• We have a clear standard for customer experience in flagship products. This year, Exynos was adopted in the Flip7. It is difficult to confirm inclusion in the flagship S26 lineup for next year. Most device users this year are already utilizing Galaxy AI features, and we plan to continue expanding AI usage in the future.\n\n12. TV market conditions and future strategy?\n\n• Competition is intensifying. We will recover market share based on differentiated product competitiveness. We aim to secure technological leadership with Micro RGB and compete in the premium market. By expanding AI features and lineups, we will counter aggressive pricing from competitors and shift consumer focus toward value. We will launch an interactive AI platform and expand applicable lineups and countries next year. We also plan to target the mid- to low-end market and strengthen leadership in the ultra-large segment of 98 inches and above. The services business will continue to expand. TV Plus will implement a differentiation strategy through competitiveness in live broadcast content.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03458216044041573, "ret_signed_m1d": -0.03458216044041573, "alpha_spy_m1d": -0.045702609671820715, "alpha_c_m1d": -0.04703266506202186, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.012450504621606129, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032660895718641614, "ret_signed_p1d": 0.032660895718641614, "alpha_spy_p1d": 0.02938061510792611, "alpha_c_p1d": 0.03169553973292505, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": 0.000965355985716565, "ret_p1w": -0.047070154425128186, "ret_signed_p1w": -0.047070154425128186, "alpha_spy_p1w": -0.033066706022235715, "alpha_c_p1w": -0.00968542970643227, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.037384724718695916, "ret_p1m": -0.03458216044041573, "ret_signed_p1m": -0.03458216044041573, "alpha_spy_p1m": -0.039818781265619885, "alpha_c_p1m": 0.012589986288804145, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.04717214672921988, "ret_p3m": 0.5396287297994777, "ret_signed_p3m": 0.5396287297994777, "alpha_spy_p3m": 0.5135705887443416, "alpha_c_p3m": 0.5524151122899006, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": -0.012786382490422854, "ret_p6m": 1.1231849285306752, "ret_signed_p6m": 1.1231849285306752, "alpha_spy_p6m": 1.067030912629305, "alpha_c_p6m": 1.0534920535424417, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.06969287498823351, "price_path": [-0.3334, -0.3315, -0.342, -0.3258, -0.3133, -0.321, -0.32, -0.3124, -0.3153, -0.3153, -0.3306, -0.3306, -0.3258, -0.3248, -0.3172, -0.3162, -0.3277, -0.3248, -0.3344, -0.3334, -0.3535, -0.3392, -0.3325, -0.3296, -0.3353, -0.3296, -0.3162, -0.3057, -0.298, -0.2789, -0.2684, -0.2407, -0.2521, -0.2321, -0.2321, -0.2014, -0.19, -0.1833, -0.1766, -0.2034, -0.1912, -0.194, -0.1739, -0.1378, -0.1378, -0.1378, -0.1378, -0.1378, -0.1378, -0.0932, -0.1037, -0.1201, -0.0874, -0.0615, -0.0596, -0.0576, -0.0634, -0.0528, -0.073, -0.0509, -0.0202, -0.0442, -0.0346, 0.0, 0.0327, 0.0672, 0.0077, -0.0336, -0.0471, -0.0596, -0.0336, -0.0058, -0.0096, -0.0125, -0.0663, -0.0336, -0.0605, -0.073, -0.0336, -0.0893, -0.0711, -0.0461, -0.0125, -0.0058, -0.0346, -0.0317, -0.0067, 0.0038, 0.0096, 0.0413, 0.0519, 0.0413, 0.0375, 0.0307, 0.0461, 0.0067, -0.0125, 0.0365, 0.0336, 0.0211, 0.0615, 0.0711, 0.0672, 0.0672, 0.1239, 0.1535, 0.1574, 0.1574, 0.1574, 0.2404, 0.3331, 0.3408, 0.3611, 0.3398, 0.3417, 0.3398, 0.3282, 0.3543, 0.389, 0.4373, 0.4412, 0.4016, 0.4431, 0.4701, 0.4682, 0.4682, 0.5396, 0.5676, 0.5512, 0.5493, 0.4518, 0.6169, 0.6323, 0.5377, 0.5309, 0.6062, 0.6004, 0.6197, 0.724, 0.7491, 0.7491, 0.7491, 0.7491, 0.834, 0.835, 0.863, 0.9306, 0.9644, 1.1043, 1.0898, 1.0898, 0.8833, 0.6622, 0.8495, 0.8167, 0.6748, 0.8138, 0.834, 0.8138, 0.7713, 0.8215, 0.8717, 1.0126, 0.9354, 0.9248, 0.7983, 0.8311, 0.8244, 0.7385, 0.7385, 0.7053, 0.6173, 0.8345, 0.7256, 0.8011, 0.8678, 0.9007, 1.0361, 0.9733, 0.9926, 0.9442, 0.9974, 1.041, 1.1038, 1.0893, 1.0748, 1.1183, 1.1038, 1.1715, 1.1232]}
{"unit_id": "orig:1987498277521231872", "ticker_idx": 0, "ts": "2025-11-09T12:31:29+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "If you don't know which ASIC company will win, buy TSMC. All GPUs and ASICs are manufactured at TSMC.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Buy TSMC as the universal play on GPU/ASIC buildout regardless of which chip designer wins.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "If you don’t know which ASIC company will win, buy TSMC.\n\nAll GPUs and ASICs are manufactured at TSMC. https://t.co/ntGduZoXU9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029701595594541108, "ret_signed_m1d": -0.029701595594541108, "alpha_spy_m1d": -0.014336917913989389, "alpha_c_m1d": 3.7246942916402404e-05, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013885534393016874, "ret_signed_p1d": -0.013885534393016874, "alpha_spy_p1d": -0.016174684095462633, "alpha_c_p1d": 0.007687007478983321, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.044908027690838015, "ret_signed_p1w": -0.044908027690838015, "alpha_spy_p1w": -0.021765753789522724, "alpha_c_p1w": 0.006626387372451026, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": 0.02756810157796874, "ret_signed_p1m": 0.02756810157796874, "alpha_spy_p1m": 0.025220311161861186, "alpha_c_p1m": -0.0008886133255470163, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.12311733105084421, "ret_signed_p3m": 0.12311733105084421, "alpha_spy_p3m": 0.1257848079382059, "alpha_c_p3m": 0.05771491429955278, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 0.34314064370015984, "ret_signed_p6m": 0.34314064370015984, "alpha_spy_p6m": 0.2749745523009961, "alpha_c_p6m": -0.11820551608568985, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.1849, -0.1864, -0.1935, -0.185, -0.2144, -0.2282, -0.2325, -0.2134, -0.2046, -0.1941, -0.1921, -0.1956, -0.2206, -0.2206, -0.2289, -0.2188, -0.2059, -0.1782, -0.1655, -0.1529, -0.1207, -0.1259, -0.1245, -0.1176, -0.1125, -0.11, -0.0902, -0.103, -0.0767, -0.0425, -0.0493, -0.063, -0.0742, -0.0746, -0.0541, -0.023, -0.0242, -0.0104, 0.0241, -0.0042, 0.0313, 0.0156, -0.0495, 0.0258, 0.0023, 0.032, 0.0155, -0.0006, 0.0082, -0.0026, -0.0216, -0.0154, -0.001, 0.0101, 0.0212, 0.0333, 0.0269, 0.0175, 0.0325, -0.0041, -0.0055, -0.0204, -0.0297, 0.0, -0.0139, -0.0157, -0.0443, -0.0354, -0.0449, -0.0588, -0.0437, -0.0602, -0.0684, -0.036, -0.0359, -0.018, -0.018, -0.0127, -0.0257, -0.0108, 0.0006, -0.0079, -0.0019, 0.0224, 0.0276, 0.0504, 0.0352, -0.0083, -0.0229, -0.0258, -0.0595, -0.0333, -0.0188, -0.0041, 0.0084, 0.0147, 0.0147, 0.0284, 0.0219, 0.0173, 0.032, 0.032, 0.0854, 0.0943, 0.1119, 0.0822, 0.0799, 0.099, 0.1266, 0.1247, 0.1108, 0.1602, 0.1627, 0.1627, 0.111, 0.1075, 0.1117, 0.1372, 0.1298, 0.149, 0.1624, 0.1531, 0.1225, 0.1592, 0.1402, 0.1062, 0.1231, 0.1847, 0.2069, 0.229, 0.2704, 0.25, 0.2441, 0.2441, 0.2368, 0.2302, 0.2238, 0.2583, 0.2566, 0.31, 0.3167, 0.2796, 0.272, 0.2535, 0.1992, 0.2138, 0.2017, 0.1508, 0.1841, 0.1787, 0.204, 0.1434, 0.1489, 0.1554, 0.1782, 0.1564, 0.1537, 0.1212, 0.1526, 0.1689, 0.1842, 0.1105, 0.1127, 0.0778, 0.1509, 0.1629, 0.1546, 0.1546, 0.1638, 0.176, 0.2461, 0.2447, 0.2621, 0.2585, 0.2937, 0.2774, 0.2374, 0.2617, 0.2472, 0.2535, 0.3194, 0.3031, 0.3706, 0.3791, 0.3361, 0.3412, 0.3488, 0.3542, 0.3677, 0.3431]}
{"unit_id": "orig:1979331004738531465", "ticker_idx": 0, "ts": "2025-10-17T23:37:40+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPMorgan argued that Huang's forecast of explosive AI data center CapEx growth, while extremely aggressive, is entirely achievable from a financial standpoint", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPMorgan report endorsing Jensen Huang's $3–4T AI CapEx forecast as financially feasible, implying sustained demand tailwind for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Can Annual AI CapEx Grow from $600B to $3–4T? JPMorgan Says Jensen Huang’s “Boast” Is Financially Feasible\n\nRecently, Jensen Huang predicted that global annual spending on AI data centers will soar from about $600 billion this year to $3–4 trillion by 2030. This implies an average annual growth rate of 42% over the next five years — prompting skepticism that Huang might be exaggerating.\n\nHowever, JPMorgan, in a recent report, argued that Huang’s forecast of explosive AI data center CapEx growth, while extremely aggressive, is entirely achievable from a financial standpoint. The bank believes that through three major funding sources — internally generated cash flow, private equity/venture capital investment, and external financing through debt or equity issuance — the technology sector has sufficient capacity to support this unprecedented investment boom.\n\nFacing a $1.6 Trillion Annual Funding Gap\n\nJPMorgan’s calculations show that by 2030, the technology industry will face an annual funding gap of up to $1.6 trillion.\n\nThe analysis first focuses on the global technology industry (defined as companies in the MSCI Global Information Technology and Communication Services indices, plus Amazon and Tesla) and examines its cash flow and capital expenditure patterns. The report estimates that in 2025, the industry’s annual operating cash flow will be about $1.6 trillion, while CapEx, including R&D, will be around $1.3 trillion — leaving a $300 billion financial surplus.\n\nAssuming AI data center investment reaches $3.5 trillion by 2030 (the midpoint of Huang’s projection), and that non-AI CapEx of $700 billion in 2025 grows at 11% annually (the three-year average), total non-AI CapEx would rise to $1.1 trillion by 2030. This would push total tech CapEx to $4.6 trillion by that year.\n\nEven if operating cash flow grows 20% annually to reach $4 trillion, the sector would still face a $600 billion shortfall.\n\nWhen shareholder returns are included, the pressure intensifies. JPMorgan assumes shareholder payouts (dividends + buybacks) will increase from $700 billion this year to about $1 trillion by 2030. Accounting for these payouts, the annual total funding gap would widen to $1.6 trillion by 2030.\n\nPrivate Equity and Venture Capital: The Key to Filling the Gap\n\nTo bridge this trillion-dollar shortfall, private capital will be the first critical pillar. JPMorgan points out that PE, VC, and infrastructure funds are pouring into digital infrastructure and AI with unprecedented enthusiasm.\n\nAccording to the report, annual fundraising for digital infrastructure currently totals about $70 billion, while PE/VC investments in AI and machine learning already reach an annualized $260 billion, according to PitchBook data. Together, private markets could provide roughly $330 billion in AI-related funding in 2025.\n\nJPMorgan forecasts that if private capital grows at a conservative 10% annual rate, annual private investment could reach $531 billion by 2030 — narrowing the funding gap from $1.6 trillion to roughly $1.1 trillion.\n\nDebt Financing: Controlled Leverage Expansion\n\nThe remaining $1.1 trillion gap will likely be covered primarily through debt financing. JPMorgan believes the tech industry is fully capable of taking on this level of leverage without triggering systemic risk.\n\nBased on the structure of U.S. non-financial corporate financing, the report assumes 40% of new debt ($430 billion) will come from bank loans, and 60% ($640 billion) from bond issuance.\n\nThis raises the question of balance-sheet health. JPMorgan’s analysis indicates that even after this debt expansion, the sector’s net-debt-to-operating-cash-flow ratio would increase only from 0.7× today to 1.2× by 2030 — still far below the MSCI Global Index average of 2.2×. This suggests that tech companies maintain ample borrowing capacity to finance the future of AI.\n\nBeyond Financing: The Next Bottleneck Lies in Energy\n\nAlthough JPMorgan concludes that Huang’s AI CapEx projection is financially feasible, the report warns that other bottlenecks — particularly power generation and transmission capacity — will become the next critical challenge for sustaining this scale of growth.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007695698251601546, "ret_signed_m1d": -0.007695698251601546, "alpha_spy_m1d": -0.0020514364181644096, "alpha_c_m1d": -0.009066623492645198, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0031656175965260047, "ret_signed_p1d": -0.0031656175965260047, "alpha_spy_p1d": -0.013566092996147638, "alpha_c_p1d": -0.01597086120475333, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": 0.016591944087498867, "ret_signed_p1w": 0.016591944087498867, "alpha_spy_p1w": -0.0027641266814113497, "alpha_c_p1w": -0.007647617887912128, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": 0.018447737116893403, "ret_signed_p1m": 0.018447737116893403, "alpha_spy_p1m": 0.016521230409782373, "alpha_c_p1m": 0.025827499213636584, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": -0.00038105350341011146, "ret_signed_p3m": -0.00038105350341011146, "alpha_spy_p3m": -0.04253987925451863, "alpha_c_p3m": -0.13668306102086292, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.03335213474918097, "ret_signed_p6m": 0.03335213474918097, "alpha_spy_p6m": -0.00520442280174227, "alpha_c_p6m": -0.26384915257935515, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.0884, -0.0679, -0.0518, -0.0531, -0.0354, -0.0421, -0.0216, -0.0293, -0.0519, -0.0176, -0.0271, -0.0208, -0.0134, -0.0029, -0.0064, -0.0004, -0.009, -0.0066, -0.0152, -0.0067, -0.0414, -0.0427, -0.045, -0.0286, -0.0187, -0.008, -0.0089, -0.0167, -0.0494, -0.0494, -0.0679, -0.0688, -0.0631, -0.0885, -0.0814, -0.0681, -0.0322, -0.033, -0.0295, -0.0299, -0.0455, -0.0706, -0.0381, -0.0357, 0.0021, -0.0261, -0.0341, -0.0302, -0.0275, -0.0075, 0.0183, 0.0219, 0.0309, 0.024, 0.0127, 0.0099, 0.0321, 0.051, -0.0003, 0.0278, -0.0174, -0.0185, -0.0077, 0.0, -0.0032, -0.0112, -0.016, -0.0058, 0.0166, 0.0451, 0.0972, 0.13, 0.1074, 0.1052, 0.1291, 0.0844, 0.0654, 0.0265, 0.0269, 0.0864, 0.0543, 0.0577, 0.0199, 0.0379, 0.0184, -0.0102, 0.018, -0.0141, -0.0237, -0.0037, -0.0295, -0.0162, -0.0162, -0.0339, -0.018, -0.0096, -0.0198, 0.0009, -0.0044, 0.0128, 0.0096, 0.0031, -0.0124, -0.0447, -0.0378, -0.03, -0.067, -0.0495, -0.0121, 0.0026, 0.0328, 0.0295, 0.0295, 0.04, 0.0273, 0.0236, 0.018, 0.018, 0.0308, 0.0268, 0.022, 0.0322, 0.01, 0.009, 0.0094, 0.0142, -0.0004, 0.021, 0.0165, 0.0165, -0.0281, 0.0006, 0.0089, 0.0243, 0.0178, 0.029, 0.0454, 0.0508, 0.0432, 0.0131, -0.0157, -0.0492, -0.0618, 0.012, 0.0373, 0.0291, 0.0373, 0.0204, -0.0022, -0.0022, 0.0096, 0.026, 0.0256, 0.0361, 0.0455, 0.0526, 0.0674, 0.0092, -0.0329, -0.004, -0.0172, -0.0009, 0.0007, -0.0294, -0.0031, 0.0085, 0.0154, -0.0003, -0.0161, 0.0001, -0.0069, -0.0153, -0.0253, -0.0573, -0.0413, -0.0437, -0.0247, -0.0653, -0.0856, -0.0984, -0.048, -0.0407, -0.0317, -0.0317, -0.0303, -0.0278, -0.0061, 0.0039, 0.0296, 0.0334]}
{"unit_id": "thread:1997270783366959527", "ticker_idx": 0, "ts": "2025-12-06T11:43:56+00:00", "symbol": "AMKR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology", "tweet_type": "analysis", "timeline": "quarters", "summary": "Multiple packaging/testing names enter Google TPU supply chain with material revenue contributions expected in H2 2026: AMKR, ASE, KYEC, MPI, WinWay, Chroma ATE.", "resolved_tickers": ["AMKR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Google Weighs Amkor Participation to Address TPU CoWoS-S Bottlenecks\n\nGoogle DeepMind's latest AI model, Gemini 3, has made a stunning debut, outperforming ChatGPT, Claude, and other competitors, signaling a new phase in the generative AI landscape. This has drawn attention to the Google-developed TPU chips that train this model, which have also attracted interest from other US-based cloud service providers (CSPs).\n\nThe seventh-generation TPU chip, Ironwood, is designed by Broadcom and manufactured using TSMC's 3nm advanced process, paired with CoWoS-S advanced packaging technology. Mass production began in the third quarter of 2025, and it is expected to become the main Google ASIC shipment in 2026.\n\nDue to the continued tight supply of TSMC's CoWoS advanced packaging capacity in 2026, some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology.\n\nTaiwanese OSAT firms join Google TPU testing\nIn Taiwan's packaging and testing supply chain, ASE Technology Holding and King Yuan Electronics (KYEC) are reported to begin handling Google TPU wafers outsourced to TSMC for chip probing, as well as final testing and system-level testing orders. ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026.\n\nTo meet the demand for advanced testing of AI GPUs and AI ASICs, ASE and KYEC have increased their 2025 capital expenditures, continuing to purchase wafer testing machines from Japan's Advantest and US-based Teradyne. Taiwanese companies Hon Precision and Chroma ATE are key beneficiaries as suppliers of back-end product testing and system-level test equipment.\n\nProbe cards and sockets gain traction\nRegarding the testing interface, Google sources wafer test probe cards mainly from MPI Corporation and Italy's Technoprobe. MPI not only secured TPU orders but also entered the Axion CPU supply chain. MPI is Taiwan's leading probe card manufacturer, which uses microelectromechanical probe card technology to lead competitors in 3nm advanced testing. Through strong co-development with customers, it has surpassed Technoprobe and US-based FormFactor, becoming the main probe card supplier for CSPs' AI ASIC chips, with a market share of 60% to 80%.\n\nWinWay Technology is the world's largest test socket supplier, and its high-end test socket products have reportedly been adopted in the Google TPU supply chain, with strong AI ASIC demand expected in 2026. For advanced nodes below 7nm, WinWay's revenue share has reached nearly 80%, with about 40% coming from AI GPU and AI ASIC contributions.\n\nLeading load board manufacturer Chroma has also revealed that it is expected to enter the Google TPU supply chain. Analysis suggests that with MediaTek securing IC design service orders for 3nm v7e and 2nm v8e, Chroma is well-positioned to become a supplier of wafer and final product test load boards. The company has previously confirmed that its large-sized test load board products have achieved recognized technology and yield levels from major US-based CSP clients, indicating promising contributions in the second half of 2026.\n\nGrowing TPU demand\nIn addition to Google's original cloud service and Gemini computing demands, Anthropic will adopt TPU as a computing resource starting in 2026, and Apple recently announced plans to integrate the Gemini model into its Siri voice assistant, further boosting demand for Google TPU chips. DIGITIMES Research estimates that TPU shipments will reach 3.325 million units in 2026, nearly 1 million more than in 2025. Despite strong competition from AWS's Trainium camp, Google is expected to maintain its leading position in ASIC chip shipments.\n\n$AMKR\n\n---\n\nhttps://t.co/8pRKUqFMOW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03310220633504313, "ret_signed_m1d": -0.03310220633504313, "alpha_spy_m1d": -0.03611553673046919, "alpha_c_m1d": -0.021896215231824723, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0022366136144834625, "ret_signed_p1d": -0.0022366136144834625, "alpha_spy_p1d": -0.0013734686475218583, "alpha_c_p1d": -0.003457626498661326, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.01789307988465083, "ret_signed_p1w": -0.01789307988465083, "alpha_spy_p1w": -0.013650982312186533, "alpha_c_p1w": 0.024570591867216107, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.14717061197078962, "ret_signed_p1m": 0.14717061197078962, "alpha_spy_p1m": 0.13221484983389664, "alpha_c_p1m": 0.09215467858327209, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": -0.0033549204217252493, "ret_signed_p3m": -0.0033549204217252493, "alpha_spy_p3m": -0.0029061532788069444, "alpha_c_p3m": -0.07941023479724152, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4467, -0.4335, -0.4317, -0.4268, -0.3965, -0.4058, -0.3491, -0.3681, -0.3409, -0.3398, -0.3447, -0.3491, -0.352, -0.3634, -0.3661, -0.3422, -0.3384, -0.3442, -0.3177, -0.336, -0.3168, -0.3163, -0.3692, -0.3201, -0.3324, -0.3018, -0.2969, -0.3016, -0.2795, -0.2824, -0.3132, -0.2862, -0.2686, -0.2594, -0.2822, -0.2704, -0.2875, -0.2795, -0.1554, -0.2132, -0.1842, -0.2154, -0.2237, -0.2141, -0.2353, -0.2306, -0.2869, -0.2929, -0.3105, -0.315, -0.2985, -0.3262, -0.2846, -0.2603, -0.2456, -0.2206, -0.2206, -0.1878, -0.1576, -0.0927, -0.0174, -0.0342, -0.0331, 0.0, -0.0022, 0.0479, 0.0521, -0.0089, -0.0179, -0.0991, -0.1306, -0.121, -0.0966, -0.0877, -0.0861, -0.0883, -0.0883, -0.0921, -0.1033, -0.0964, -0.117, -0.117, -0.04, 0.0765, 0.1472, 0.1792, 0.1407, 0.1695, 0.1537, 0.1619, 0.0886, 0.1004, 0.0736, 0.0736, 0.0971, 0.1957, 0.1731, 0.1118, 0.127, 0.1349, 0.1398, 0.1183, 0.081, 0.0781, 0.0342, -0.0188, -0.0098, 0.104, 0.1747, 0.1957, 0.2563, 0.1539, 0.062, 0.062, 0.0476, 0.0454, 0.0812, 0.0722, 0.0528, 0.0854, 0.1393, 0.0852, 0.0696, 0.0693, -0.0027, 0.0237, -0.0034, -0.0693, -0.0333, -0.0172, -0.0177, -0.0752, -0.0366, 0.0001, 0.0384, 0.0503, 0.0786, 0.0266, 0.0335, 0.1263, 0.1189, 0.0116, -0.0039, -0.0756, 0.0091, 0.0418, 0.0465, 0.0465, 0.0539, 0.0671, 0.1747, 0.2356, 0.2988, 0.3544, 0.3741, 0.3383, 0.4095, 0.5097, 0.5561, 0.5734, 0.6191, 0.6339, 0.7502, 0.6946, 0.5991, 0.5823, 0.563, 0.5931, 0.5902, 0.719, 0.7304, 0.6195, 0.7168, 0.7186, 0.646, 0.672, 0.6155, 0.5765, 0.4799, 0.4687, 0.5348, 0.4768, 0.4734, 0.4734, 0.6462, 0.6177, 0.5816, 0.5588, null, null]}
{"unit_id": "thread:1997270783366959527", "ticker_idx": 1, "ts": "2025-12-06T11:43:56+00:00", "symbol": "ASE Technology Holding", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Multiple packaging/testing names enter Google TPU supply chain with material revenue contributions expected in H2 2026: AMKR, ASE, KYEC, MPI, WinWay, Chroma ATE.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ticker ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Google Weighs Amkor Participation to Address TPU CoWoS-S Bottlenecks\n\nGoogle DeepMind's latest AI model, Gemini 3, has made a stunning debut, outperforming ChatGPT, Claude, and other competitors, signaling a new phase in the generative AI landscape. This has drawn attention to the Google-developed TPU chips that train this model, which have also attracted interest from other US-based cloud service providers (CSPs).\n\nThe seventh-generation TPU chip, Ironwood, is designed by Broadcom and manufactured using TSMC's 3nm advanced process, paired with CoWoS-S advanced packaging technology. Mass production began in the third quarter of 2025, and it is expected to become the main Google ASIC shipment in 2026.\n\nDue to the continued tight supply of TSMC's CoWoS advanced packaging capacity in 2026, some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology.\n\nTaiwanese OSAT firms join Google TPU testing\nIn Taiwan's packaging and testing supply chain, ASE Technology Holding and King Yuan Electronics (KYEC) are reported to begin handling Google TPU wafers outsourced to TSMC for chip probing, as well as final testing and system-level testing orders. ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026.\n\nTo meet the demand for advanced testing of AI GPUs and AI ASICs, ASE and KYEC have increased their 2025 capital expenditures, continuing to purchase wafer testing machines from Japan's Advantest and US-based Teradyne. Taiwanese companies Hon Precision and Chroma ATE are key beneficiaries as suppliers of back-end product testing and system-level test equipment.\n\nProbe cards and sockets gain traction\nRegarding the testing interface, Google sources wafer test probe cards mainly from MPI Corporation and Italy's Technoprobe. MPI not only secured TPU orders but also entered the Axion CPU supply chain. MPI is Taiwan's leading probe card manufacturer, which uses microelectromechanical probe card technology to lead competitors in 3nm advanced testing. Through strong co-development with customers, it has surpassed Technoprobe and US-based FormFactor, becoming the main probe card supplier for CSPs' AI ASIC chips, with a market share of 60% to 80%.\n\nWinWay Technology is the world's largest test socket supplier, and its high-end test socket products have reportedly been adopted in the Google TPU supply chain, with strong AI ASIC demand expected in 2026. For advanced nodes below 7nm, WinWay's revenue share has reached nearly 80%, with about 40% coming from AI GPU and AI ASIC contributions.\n\nLeading load board manufacturer Chroma has also revealed that it is expected to enter the Google TPU supply chain. Analysis suggests that with MediaTek securing IC design service orders for 3nm v7e and 2nm v8e, Chroma is well-positioned to become a supplier of wafer and final product test load boards. The company has previously confirmed that its large-sized test load board products have achieved recognized technology and yield levels from major US-based CSP clients, indicating promising contributions in the second half of 2026.\n\nGrowing TPU demand\nIn addition to Google's original cloud service and Gemini computing demands, Anthropic will adopt TPU as a computing resource starting in 2026, and Apple recently announced plans to integrate the Gemini model into its Siri voice assistant, further boosting demand for Google TPU chips. DIGITIMES Research estimates that TPU shipments will reach 3.325 million units in 2026, nearly 1 million more than in 2025. Despite strong competition from AWS's Trainium camp, Google is expected to maintain its leading position in ASIC chip shipments.\n\n$AMKR\n\n---\n\nhttps://t.co/8pRKUqFMOW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.033672156332791414, "ret_signed_m1d": -0.033672156332791414, "alpha_spy_m1d": -0.036685486728217476, "alpha_c_m1d": -0.02246616522957301, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014612484069579157, "ret_signed_p1d": 0.014612484069579157, "alpha_spy_p1d": 0.015475629036540761, "alpha_c_p1d": 0.013391471185401294, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.01016517469750533, "ret_signed_p1w": -0.01016517469750533, "alpha_spy_p1w": -0.005923077125041032, "alpha_c_p1w": 0.03229849705436161, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.11181704285099903, "ret_signed_p1m": 0.11181704285099903, "alpha_spy_p1m": 0.09686128071410605, "alpha_c_p1m": 0.05680110946348149, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.4015247640867816, "ret_signed_p3m": 0.4015247640867816, "alpha_spy_p3m": 0.4019735312296999, "alpha_c_p3m": 0.32546944971126535, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2821, -0.2942, -0.298, -0.2884, -0.2821, -0.2821, -0.2713, -0.2719, -0.256, -0.2567, -0.2668, -0.2719, -0.2903, -0.2929, -0.2954, -0.2929, -0.2916, -0.2961, -0.2764, -0.277, -0.2465, -0.256, -0.2961, -0.2548, -0.2808, -0.2421, -0.2008, -0.1919, -0.1747, -0.1944, -0.2084, -0.1855, -0.1753, -0.1595, -0.1512, -0.0832, -0.0445, 0.0172, 0.0286, -0.0235, 0.0057, -0.0235, -0.033, -0.0229, -0.0483, -0.0476, -0.0705, -0.0718, -0.0915, -0.0997, -0.1061, -0.1264, -0.122, -0.1004, -0.0991, -0.0851, -0.0851, -0.0502, -0.0578, -0.0299, -0.0216, -0.0368, -0.0337, 0.0, 0.0146, 0.0438, 0.0419, 0.0013, -0.0102, -0.028, -0.0629, -0.0553, -0.0368, -0.0248, -0.0133, -0.0121, -0.0121, -0.0013, 0.0178, 0.0184, 0.0229, 0.0229, 0.0712, 0.0775, 0.1118, 0.1194, 0.101, 0.1207, 0.1607, 0.1881, 0.1995, 0.2192, 0.2332, 0.2332, 0.2008, 0.2058, 0.2109, 0.2319, 0.2598, 0.2751, 0.2872, 0.2567, 0.2058, 0.2332, 0.2319, 0.2065, 0.2872, 0.3272, 0.4072, 0.4288, 0.5083, 0.4905, 0.4867, 0.4867, 0.4911, 0.4828, 0.4727, 0.5222, 0.4917, 0.5769, 0.5953, 0.5394, 0.5432, 0.5375, 0.4174, 0.4041, 0.4015, 0.3418, 0.3761, 0.3793, 0.3914, 0.3335, 0.3659, 0.3672, 0.3869, 0.3761, 0.3914, 0.3539, 0.3774, 0.3679, 0.4231, 0.3615, 0.3659, 0.3272, 0.3774, 0.4282, 0.4155, 0.4155, 0.4327, 0.4098, 0.5337, 0.5572, 0.5781, 0.6868, 0.7071, 0.73, 0.7535, 0.8164, 0.8602, 0.8806, 0.9028, 0.8939, 1.033, 0.96, 0.9111, 0.9435, 0.9956, 1.007, 1.0508, 1.1245, 1.1703, 1.1182, 1.1747, 1.2427, 1.1779, 1.2535, 1.2395, 1.148, 1.0108, 0.9625, 1.0133, 1.0737, 1.2116, 1.2116, 1.4746, 1.4822, 1.5794, 1.4365, null, null]}
{"unit_id": "thread:1997270783366959527", "ticker_idx": 2, "ts": "2025-12-06T11:43:56+00:00", "symbol": "King Yuan Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Multiple packaging/testing names enter Google TPU supply chain with material revenue contributions expected in H2 2026: AMKR, ASE, KYEC, MPI, WinWay, Chroma ATE.", "resolved_tickers": ["2449.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "King Yuan Electronics 2449.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Google Weighs Amkor Participation to Address TPU CoWoS-S Bottlenecks\n\nGoogle DeepMind's latest AI model, Gemini 3, has made a stunning debut, outperforming ChatGPT, Claude, and other competitors, signaling a new phase in the generative AI landscape. This has drawn attention to the Google-developed TPU chips that train this model, which have also attracted interest from other US-based cloud service providers (CSPs).\n\nThe seventh-generation TPU chip, Ironwood, is designed by Broadcom and manufactured using TSMC's 3nm advanced process, paired with CoWoS-S advanced packaging technology. Mass production began in the third quarter of 2025, and it is expected to become the main Google ASIC shipment in 2026.\n\nDue to the continued tight supply of TSMC's CoWoS advanced packaging capacity in 2026, some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology.\n\nTaiwanese OSAT firms join Google TPU testing\nIn Taiwan's packaging and testing supply chain, ASE Technology Holding and King Yuan Electronics (KYEC) are reported to begin handling Google TPU wafers outsourced to TSMC for chip probing, as well as final testing and system-level testing orders. ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026.\n\nTo meet the demand for advanced testing of AI GPUs and AI ASICs, ASE and KYEC have increased their 2025 capital expenditures, continuing to purchase wafer testing machines from Japan's Advantest and US-based Teradyne. Taiwanese companies Hon Precision and Chroma ATE are key beneficiaries as suppliers of back-end product testing and system-level test equipment.\n\nProbe cards and sockets gain traction\nRegarding the testing interface, Google sources wafer test probe cards mainly from MPI Corporation and Italy's Technoprobe. MPI not only secured TPU orders but also entered the Axion CPU supply chain. MPI is Taiwan's leading probe card manufacturer, which uses microelectromechanical probe card technology to lead competitors in 3nm advanced testing. Through strong co-development with customers, it has surpassed Technoprobe and US-based FormFactor, becoming the main probe card supplier for CSPs' AI ASIC chips, with a market share of 60% to 80%.\n\nWinWay Technology is the world's largest test socket supplier, and its high-end test socket products have reportedly been adopted in the Google TPU supply chain, with strong AI ASIC demand expected in 2026. For advanced nodes below 7nm, WinWay's revenue share has reached nearly 80%, with about 40% coming from AI GPU and AI ASIC contributions.\n\nLeading load board manufacturer Chroma has also revealed that it is expected to enter the Google TPU supply chain. Analysis suggests that with MediaTek securing IC design service orders for 3nm v7e and 2nm v8e, Chroma is well-positioned to become a supplier of wafer and final product test load boards. The company has previously confirmed that its large-sized test load board products have achieved recognized technology and yield levels from major US-based CSP clients, indicating promising contributions in the second half of 2026.\n\nGrowing TPU demand\nIn addition to Google's original cloud service and Gemini computing demands, Anthropic will adopt TPU as a computing resource starting in 2026, and Apple recently announced plans to integrate the Gemini model into its Siri voice assistant, further boosting demand for Google TPU chips. DIGITIMES Research estimates that TPU shipments will reach 3.325 million units in 2026, nearly 1 million more than in 2025. Despite strong competition from AWS's Trainium camp, Google is expected to maintain its leading position in ASIC chip shipments.\n\n$AMKR\n\n---\n\nhttps://t.co/8pRKUqFMOW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028508771929824595, "ret_signed_m1d": -0.028508771929824595, "alpha_spy_m1d": -0.031522102325250656, "alpha_c_m1d": -0.01730278082660619, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002192982456140413, "ret_signed_p1d": 0.002192982456140413, "alpha_spy_p1d": 0.0030561274231020175, "alpha_c_p1d": 0.0009719695719625498, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.028508771929824595, "ret_signed_p1w": -0.028508771929824595, "alpha_spy_p1w": -0.024266674357360296, "alpha_c_p1w": 0.013954899822042344, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.17105263157894735, "ret_signed_p1m": 0.17105263157894735, "alpha_spy_p1m": 0.15609686944205436, "alpha_c_p1m": 0.11603669819142981, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.29385964912280693, "ret_signed_p3m": 0.29385964912280693, "alpha_spy_p3m": 0.29430841626572524, "alpha_c_p3m": 0.21780433474729066, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2434, -0.2719, -0.2719, -0.2697, -0.2719, -0.2763, -0.2654, -0.2566, -0.25, -0.2675, -0.2719, -0.2719, -0.318, -0.318, -0.2807, -0.2763, -0.2654, -0.2588, -0.2588, -0.2434, -0.2368, -0.2281, -0.2281, -0.2544, -0.2917, -0.2719, -0.2083, -0.2325, -0.2193, -0.193, -0.2259, -0.2325, -0.2325, -0.1754, -0.1732, -0.0921, -0.1096, -0.0482, -0.0175, -0.0614, -0.0329, -0.0461, -0.0636, -0.0482, -0.0658, -0.0768, -0.0548, -0.0855, -0.0636, -0.1316, -0.1382, -0.0724, -0.1031, -0.1075, -0.1009, -0.0702, -0.0263, 0.0044, -0.0461, -0.0066, -0.0088, -0.0329, -0.0285, 0.0, 0.0022, 0.0044, -0.0285, 0.0088, -0.0285, -0.0526, -0.057, -0.0855, -0.0548, -0.0307, -0.0351, -0.0395, -0.0395, -0.0219, 0.0746, 0.068, 0.0855, 0.0855, 0.1732, 0.1535, 0.1711, 0.1689, 0.1535, 0.1557, 0.1842, 0.1864, 0.1864, 0.1732, 0.1996, 0.1908, 0.2039, 0.1711, 0.2039, 0.2566, 0.3048, 0.3158, 0.364, 0.3377, 0.2982, 0.25, 0.2873, 0.3026, 0.2303, 0.2434, 0.2785, 0.3246, 0.375, 0.375, 0.375, 0.375, 0.375, 0.375, 0.375, 0.375, 0.3882, 0.4298, 0.4035, 0.4342, 0.4342, 0.386, 0.3377, 0.2193, 0.2939, 0.2478, 0.193, 0.2851, 0.3421, 0.3158, 0.2961, 0.2632, 0.2895, 0.4167, 0.386, 0.3509, 0.2544, 0.2215, 0.2654, 0.2632, 0.2215, 0.193, 0.1447, 0.2303, 0.136, 0.136, 0.136, 0.1754, 0.2281, 0.2061, 0.2171, 0.2939, 0.3202, 0.2763, 0.2434, 0.2061, 0.1886, 0.2325, 0.261, 0.2171, 0.2544, 0.2434, 0.2434, 0.2434, 0.3268, 0.3268, 0.4583, 0.5526, 0.4408, 0.4868, 0.364, 0.3136, 0.3202, 0.3377, 0.2982, 0.3136, 0.2939, 0.2675, 0.2171, 0.2544, 0.2982, 0.3904, 0.4846, 0.3838, 0.4035, 0.4408, null, null]}
{"unit_id": "thread:1997270783366959527", "ticker_idx": 3, "ts": "2025-12-06T11:43:56+00:00", "symbol": "MPI Corporation", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MPI not only secured TPU orders but also entered the Axion CPU supply chain", "tweet_type": "analysis", "timeline": "quarters", "summary": "Multiple packaging/testing names enter Google TPU supply chain with material revenue contributions expected in H2 2026: AMKR, ASE, KYEC, MPI, WinWay, Chroma ATE.", "resolved_tickers": ["6670.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MPI Corporation 6670.TW Taiwan probe card/test", "bench_c": "XLY", "bench_c_reason": "industry='Leisure'", "bench_c_industry": "Leisure", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Google Weighs Amkor Participation to Address TPU CoWoS-S Bottlenecks\n\nGoogle DeepMind's latest AI model, Gemini 3, has made a stunning debut, outperforming ChatGPT, Claude, and other competitors, signaling a new phase in the generative AI landscape. This has drawn attention to the Google-developed TPU chips that train this model, which have also attracted interest from other US-based cloud service providers (CSPs).\n\nThe seventh-generation TPU chip, Ironwood, is designed by Broadcom and manufactured using TSMC's 3nm advanced process, paired with CoWoS-S advanced packaging technology. Mass production began in the third quarter of 2025, and it is expected to become the main Google ASIC shipment in 2026.\n\nDue to the continued tight supply of TSMC's CoWoS advanced packaging capacity in 2026, some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology.\n\nTaiwanese OSAT firms join Google TPU testing\nIn Taiwan's packaging and testing supply chain, ASE Technology Holding and King Yuan Electronics (KYEC) are reported to begin handling Google TPU wafers outsourced to TSMC for chip probing, as well as final testing and system-level testing orders. ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026.\n\nTo meet the demand for advanced testing of AI GPUs and AI ASICs, ASE and KYEC have increased their 2025 capital expenditures, continuing to purchase wafer testing machines from Japan's Advantest and US-based Teradyne. Taiwanese companies Hon Precision and Chroma ATE are key beneficiaries as suppliers of back-end product testing and system-level test equipment.\n\nProbe cards and sockets gain traction\nRegarding the testing interface, Google sources wafer test probe cards mainly from MPI Corporation and Italy's Technoprobe. MPI not only secured TPU orders but also entered the Axion CPU supply chain. MPI is Taiwan's leading probe card manufacturer, which uses microelectromechanical probe card technology to lead competitors in 3nm advanced testing. Through strong co-development with customers, it has surpassed Technoprobe and US-based FormFactor, becoming the main probe card supplier for CSPs' AI ASIC chips, with a market share of 60% to 80%.\n\nWinWay Technology is the world's largest test socket supplier, and its high-end test socket products have reportedly been adopted in the Google TPU supply chain, with strong AI ASIC demand expected in 2026. For advanced nodes below 7nm, WinWay's revenue share has reached nearly 80%, with about 40% coming from AI GPU and AI ASIC contributions.\n\nLeading load board manufacturer Chroma has also revealed that it is expected to enter the Google TPU supply chain. Analysis suggests that with MediaTek securing IC design service orders for 3nm v7e and 2nm v8e, Chroma is well-positioned to become a supplier of wafer and final product test load boards. The company has previously confirmed that its large-sized test load board products have achieved recognized technology and yield levels from major US-based CSP clients, indicating promising contributions in the second half of 2026.\n\nGrowing TPU demand\nIn addition to Google's original cloud service and Gemini computing demands, Anthropic will adopt TPU as a computing resource starting in 2026, and Apple recently announced plans to integrate the Gemini model into its Siri voice assistant, further boosting demand for Google TPU chips. DIGITIMES Research estimates that TPU shipments will reach 3.325 million units in 2026, nearly 1 million more than in 2025. Despite strong competition from AWS's Trainium camp, Google is expected to maintain its leading position in ASIC chip shipments.\n\n$AMKR\n\n---\n\nhttps://t.co/8pRKUqFMOW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02208835341365467, "ret_signed_m1d": 0.02208835341365467, "alpha_spy_m1d": 0.01907502301822861, "alpha_c_m1d": 0.007341267121851125, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": 0.014747086291803546, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012048192771084376, "ret_signed_p1d": -0.012048192771084376, "alpha_spy_p1d": -0.011185047804122772, "alpha_c_p1d": -0.011963537527292512, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": -8.465524379186462e-05, "ret_p1w": 0.032128514056224855, "ret_signed_p1w": 0.032128514056224855, "alpha_spy_p1w": 0.03637061162868915, "alpha_c_p1w": 0.0004307732670756348, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": 0.03169774078914922, "ret_p1m": 0.028112449799196693, "ret_signed_p1m": 0.028112449799196693, "alpha_spy_p1m": 0.013156687662303712, "alpha_c_p1m": -2.5031615114468053e-05, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.02813748141431116, "ret_p3m": 0.0763052208835342, "ret_signed_p3m": 0.0763052208835342, "alpha_spy_p3m": 0.0767539880264525, "alpha_c_p3m": 0.08655194612461825, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": -0.010246725241084054, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0442, 0.0241, 0.0201, 0.0382, 0.0582, 0.0743, 0.0823, 0.0763, 0.0763, 0.0643, 0.0663, 0.0643, 0.0462, 0.0462, 0.0482, 0.0321, 0.0261, 0.0241, 0.0241, 0.0361, 0.0361, 0.0341, 0.0341, 0.0261, 0.0161, -0.004, 0.004, -0.004, -0.012, -0.0161, 0.0, 0.008, 0.008, -0.0201, -0.0361, -0.0321, -0.0181, -0.0261, -0.0141, -0.008, -0.0161, -0.0141, -0.006, 0.008, 0.0141, 0.0301, 0.0161, -0.0221, -0.0341, -0.0643, -0.0703, -0.0602, -0.0582, -0.0301, -0.0442, -0.0161, -0.0341, -0.0241, 0.002, -0.004, -0.004, 0.0261, 0.0221, 0.0, -0.012, 0.006, 0.0321, 0.0502, 0.0321, 0.0321, 0.0341, 0.0201, 0.0341, 0.0361, 0.0141, 0.0241, 0.0241, 0.0361, 0.0261, 0.0361, 0.0261, 0.0261, 0.004, 0.01, 0.0281, 0.0402, 0.0542, 0.0562, 0.0924, 0.1205, 0.1145, 0.1245, 0.1345, 0.1124, 0.1165, 0.1185, 0.1124, 0.1124, 0.0964, 0.0904, 0.0783, 0.0863, 0.0783, 0.0703, 0.0863, 0.1104, 0.1185, 0.1024, 0.0964, 0.0643, 0.0663, 0.0663, 0.0663, 0.0663, 0.0663, 0.0663, 0.0663, 0.0663, 0.0964, 0.0924, 0.0663, 0.0663, 0.0663, 0.0904, 0.0904, 0.0522, 0.0763, 0.0803, 0.0683, 0.0863, 0.1084, 0.0904, 0.1084, 0.1185, 0.1265, 0.1205, 0.0904, 0.1285, 0.1124, 0.1084, 0.1145, 0.0884, 0.0763, 0.0442, 0.0321, 0.0643, 0.0321, 0.0321, 0.0321, 0.0382, 0.0522, 0.0582, 0.0663, 0.0582, 0.0582, 0.0582, 0.0582, 0.0442, 0.0542, 0.0542, 0.0321, 0.0181, 0.012, 0.0161, 0.01, 0.0161, 0.008, 0.008, 0.0281, 0.0321, 0.0341, 0.0382, 0.0201, 0.0542, 0.1024, 0.0643, 0.0562, 0.0562, 0.0522, 0.0542, 0.0402, 0.0522, 0.0462, 0.0281, 0.0221, 0.0522, 0.0361, 0.0502, null, null]}
{"unit_id": "thread:1997270783366959527", "ticker_idx": 4, "ts": "2025-12-06T11:43:56+00:00", "symbol": "WinWay Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "strong AI ASIC demand expected in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Multiple packaging/testing names enter Google TPU supply chain with material revenue contributions expected in H2 2026: AMKR, ASE, KYEC, MPI, WinWay, Chroma ATE.", "resolved_tickers": ["6515.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "WinWay Technology listed on TWSE as 6515.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Google Weighs Amkor Participation to Address TPU CoWoS-S Bottlenecks\n\nGoogle DeepMind's latest AI model, Gemini 3, has made a stunning debut, outperforming ChatGPT, Claude, and other competitors, signaling a new phase in the generative AI landscape. This has drawn attention to the Google-developed TPU chips that train this model, which have also attracted interest from other US-based cloud service providers (CSPs).\n\nThe seventh-generation TPU chip, Ironwood, is designed by Broadcom and manufactured using TSMC's 3nm advanced process, paired with CoWoS-S advanced packaging technology. Mass production began in the third quarter of 2025, and it is expected to become the main Google ASIC shipment in 2026.\n\nDue to the continued tight supply of TSMC's CoWoS advanced packaging capacity in 2026, some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology.\n\nTaiwanese OSAT firms join Google TPU testing\nIn Taiwan's packaging and testing supply chain, ASE Technology Holding and King Yuan Electronics (KYEC) are reported to begin handling Google TPU wafers outsourced to TSMC for chip probing, as well as final testing and system-level testing orders. ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026.\n\nTo meet the demand for advanced testing of AI GPUs and AI ASICs, ASE and KYEC have increased their 2025 capital expenditures, continuing to purchase wafer testing machines from Japan's Advantest and US-based Teradyne. Taiwanese companies Hon Precision and Chroma ATE are key beneficiaries as suppliers of back-end product testing and system-level test equipment.\n\nProbe cards and sockets gain traction\nRegarding the testing interface, Google sources wafer test probe cards mainly from MPI Corporation and Italy's Technoprobe. MPI not only secured TPU orders but also entered the Axion CPU supply chain. MPI is Taiwan's leading probe card manufacturer, which uses microelectromechanical probe card technology to lead competitors in 3nm advanced testing. Through strong co-development with customers, it has surpassed Technoprobe and US-based FormFactor, becoming the main probe card supplier for CSPs' AI ASIC chips, with a market share of 60% to 80%.\n\nWinWay Technology is the world's largest test socket supplier, and its high-end test socket products have reportedly been adopted in the Google TPU supply chain, with strong AI ASIC demand expected in 2026. For advanced nodes below 7nm, WinWay's revenue share has reached nearly 80%, with about 40% coming from AI GPU and AI ASIC contributions.\n\nLeading load board manufacturer Chroma has also revealed that it is expected to enter the Google TPU supply chain. Analysis suggests that with MediaTek securing IC design service orders for 3nm v7e and 2nm v8e, Chroma is well-positioned to become a supplier of wafer and final product test load boards. The company has previously confirmed that its large-sized test load board products have achieved recognized technology and yield levels from major US-based CSP clients, indicating promising contributions in the second half of 2026.\n\nGrowing TPU demand\nIn addition to Google's original cloud service and Gemini computing demands, Anthropic will adopt TPU as a computing resource starting in 2026, and Apple recently announced plans to integrate the Gemini model into its Siri voice assistant, further boosting demand for Google TPU chips. DIGITIMES Research estimates that TPU shipments will reach 3.325 million units in 2026, nearly 1 million more than in 2025. Despite strong competition from AWS's Trainium camp, Google is expected to maintain its leading position in ASIC chip shipments.\n\n$AMKR\n\n---\n\nhttps://t.co/8pRKUqFMOW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015873015873015817, "ret_signed_m1d": 0.015873015873015817, "alpha_spy_m1d": 0.012859685477589755, "alpha_c_m1d": 0.02707900697623422, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06150793650793651, "ret_signed_p1d": 0.06150793650793651, "alpha_spy_p1d": 0.062371081474898116, "alpha_c_p1d": 0.06028692362375865, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": 0.10515873015873023, "ret_signed_p1w": 0.10515873015873023, "alpha_spy_p1w": 0.10940082773119453, "alpha_c_p1w": 0.14762240191059717, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.2420634920634921, "ret_signed_p1m": 0.2420634920634921, "alpha_spy_p1m": 0.2271077299265991, "alpha_c_p1m": 0.18704755867597456, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.8948412698412698, "ret_signed_p3m": 0.8948412698412698, "alpha_spy_p3m": 0.8952900369841881, "alpha_c_p3m": 0.8187859554657535, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3095, -0.3313, -0.3651, -0.3671, -0.3194, -0.3234, -0.256, -0.2679, -0.1984, -0.2004, -0.1944, -0.1984, -0.1647, -0.1647, -0.1012, -0.0992, -0.0992, -0.1131, -0.1131, -0.1567, -0.1349, -0.0893, -0.0893, -0.0992, -0.0972, -0.0258, -0.0357, -0.0397, -0.0496, -0.0278, -0.0218, -0.0298, -0.0298, -0.0179, -0.0496, 0.0437, 0.0079, 0.0437, -0.0317, -0.0397, -0.0417, -0.0079, -0.0556, -0.0635, -0.125, -0.131, -0.1329, -0.129, -0.125, -0.1429, -0.1508, -0.0675, -0.0833, -0.0179, -0.002, 0.0496, 0.0913, 0.123, 0.0119, 0.0159, 0.0258, 0.0198, 0.0159, 0.0, 0.0615, 0.0655, 0.0575, 0.1111, 0.1052, 0.0992, 0.0198, 0.0159, 0.0417, 0.0278, 0.0258, 0.0456, 0.0456, 0.1488, 0.1012, 0.1409, 0.127, 0.127, 0.1488, 0.2143, 0.2421, 0.3115, 0.2599, 0.3849, 0.377, 0.369, 0.3869, 0.369, 0.4028, 0.373, 0.4841, 0.5179, 0.4881, 0.5496, 0.6131, 0.6964, 0.5933, 0.5476, 0.5496, 0.6111, 0.7718, 0.8075, 0.756, 0.7897, 0.873, 0.9425, 0.9861, 0.9861, 0.9861, 0.9861, 0.9861, 0.9861, 0.9861, 0.9861, 0.8631, 0.8909, 0.9841, 1.0298, 1.0298, 0.8948, 0.877, 0.7956, 0.8948, 1.0833, 0.9762, 1.1726, 1.3413, 1.5119, 1.494, 1.4444, 1.627, 1.8651, 1.8611, 2.1468, 2.129, 2.25, 2.3532, 2.2321, 2.1409, 2.0456, 1.8373, 1.8929, 1.621, 1.621, 1.621, 1.8175, 2.0992, 2.1726, 2.1468, 2.3968, 2.3611, 2.5972, 2.8968, 3.0893, 3.2837, 3.252, 3.2579, 3.2579, 2.9921, 2.8016, 3.0218, 3.0417, 3.121, 3.121, 3.246, 2.9901, 3.0198, 2.9127, 2.8214, 3.004, 2.9484, 2.9187, 3.1825, 2.881, 3.0635, 2.7659, 2.6032, 2.7738, 2.4524, 2.5119, 2.5337, 2.8869, 2.619, 2.4385, null, null]}
{"unit_id": "thread:1997270783366959527", "ticker_idx": 5, "ts": "2025-12-06T11:43:56+00:00", "symbol": "Chroma ATE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Chroma is well-positioned to become a supplier of wafer and final product test load boards... indicating promising contributions in the second half of 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Multiple packaging/testing names enter Google TPU supply chain with material revenue contributions expected in H2 2026: AMKR, ASE, KYEC, MPI, WinWay, Chroma ATE.", "resolved_tickers": ["2360.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Chroma ATE Inc. 2360.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Scientific & Technical Instruments'", "bench_c_industry": "Scientific & Technical Instruments", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Google Weighs Amkor Participation to Address TPU CoWoS-S Bottlenecks\n\nGoogle DeepMind's latest AI model, Gemini 3, has made a stunning debut, outperforming ChatGPT, Claude, and other competitors, signaling a new phase in the generative AI landscape. This has drawn attention to the Google-developed TPU chips that train this model, which have also attracted interest from other US-based cloud service providers (CSPs).\n\nThe seventh-generation TPU chip, Ironwood, is designed by Broadcom and manufactured using TSMC's 3nm advanced process, paired with CoWoS-S advanced packaging technology. Mass production began in the third quarter of 2025, and it is expected to become the main Google ASIC shipment in 2026.\n\nDue to the continued tight supply of TSMC's CoWoS advanced packaging capacity in 2026, some orders are expected to overflow to US-based OSAT company Amkor, which will enter the Google TPU supply chain using a \"CoWoS-S-like\" technology.\n\nTaiwanese OSAT firms join Google TPU testing\nIn Taiwan's packaging and testing supply chain, ASE Technology Holding and King Yuan Electronics (KYEC) are reported to begin handling Google TPU wafers outsourced to TSMC for chip probing, as well as final testing and system-level testing orders. ASE and KYEC will focus on front-end and back-end testing, respectively, and are expected to see significant revenue contributions in the second half of 2026.\n\nTo meet the demand for advanced testing of AI GPUs and AI ASICs, ASE and KYEC have increased their 2025 capital expenditures, continuing to purchase wafer testing machines from Japan's Advantest and US-based Teradyne. Taiwanese companies Hon Precision and Chroma ATE are key beneficiaries as suppliers of back-end product testing and system-level test equipment.\n\nProbe cards and sockets gain traction\nRegarding the testing interface, Google sources wafer test probe cards mainly from MPI Corporation and Italy's Technoprobe. MPI not only secured TPU orders but also entered the Axion CPU supply chain. MPI is Taiwan's leading probe card manufacturer, which uses microelectromechanical probe card technology to lead competitors in 3nm advanced testing. Through strong co-development with customers, it has surpassed Technoprobe and US-based FormFactor, becoming the main probe card supplier for CSPs' AI ASIC chips, with a market share of 60% to 80%.\n\nWinWay Technology is the world's largest test socket supplier, and its high-end test socket products have reportedly been adopted in the Google TPU supply chain, with strong AI ASIC demand expected in 2026. For advanced nodes below 7nm, WinWay's revenue share has reached nearly 80%, with about 40% coming from AI GPU and AI ASIC contributions.\n\nLeading load board manufacturer Chroma has also revealed that it is expected to enter the Google TPU supply chain. Analysis suggests that with MediaTek securing IC design service orders for 3nm v7e and 2nm v8e, Chroma is well-positioned to become a supplier of wafer and final product test load boards. The company has previously confirmed that its large-sized test load board products have achieved recognized technology and yield levels from major US-based CSP clients, indicating promising contributions in the second half of 2026.\n\nGrowing TPU demand\nIn addition to Google's original cloud service and Gemini computing demands, Anthropic will adopt TPU as a computing resource starting in 2026, and Apple recently announced plans to integrate the Gemini model into its Siri voice assistant, further boosting demand for Google TPU chips. DIGITIMES Research estimates that TPU shipments will reach 3.325 million units in 2026, nearly 1 million more than in 2025. Despite strong competition from AWS's Trainium camp, Google is expected to maintain its leading position in ASIC chip shipments.\n\n$AMKR\n\n---\n\nhttps://t.co/8pRKUqFMOW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011180124223602483, "ret_signed_m1d": -0.011180124223602483, "alpha_spy_m1d": -0.014193454619028545, "alpha_c_m1d": -0.0042032378286138705, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.006976886394988613, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003726708074534235, "ret_signed_p1d": 0.003726708074534235, "alpha_spy_p1d": 0.004589853041495839, "alpha_c_p1d": 0.001084827478416539, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.002641880596117696, "ret_p1w": -0.044720496894409933, "ret_signed_p1w": -0.044720496894409933, "alpha_spy_p1w": -0.040478399321945635, "alpha_c_p1w": -0.00861679919436964, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.036103697700040294, "ret_p1m": 0.06335403726708067, "ret_signed_p1m": 0.06335403726708067, "alpha_spy_p1m": 0.048398275130187685, "alpha_c_p1m": 0.06848554570938681, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": -0.005131508442306143, "ret_p3m": 0.7453416149068324, "ret_signed_p3m": 0.7453416149068324, "alpha_spy_p3m": 0.7457903820497507, "alpha_c_p3m": 0.7943653420831709, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": -0.049023727176338516, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2795, -0.287, -0.2907, -0.3143, -0.3093, -0.2969, -0.2373, -0.2571, -0.2609, -0.2373, -0.2509, -0.2596, -0.2994, -0.2994, -0.2807, -0.2472, -0.2124, -0.2062, -0.2062, -0.195, -0.2149, -0.2273, -0.2273, -0.2, -0.1764, -0.1466, -0.1379, -0.1776, -0.1851, -0.159, -0.1391, -0.1404, -0.1404, -0.0832, -0.0783, -0.0211, -0.0012, 0.0174, 0.0075, -0.036, -0.0534, -0.0646, -0.0795, -0.0758, -0.0559, -0.0087, 0.0025, -0.0248, -0.0385, -0.0733, -0.0435, 0.0037, -0.0385, 0.0062, 0.0298, -0.0062, 0.0075, 0.0174, -0.0248, -0.0335, 0.0, -0.036, -0.0112, 0.0, 0.0037, -0.0062, -0.0112, 0.0, -0.0447, -0.0733, -0.0807, -0.0957, -0.0795, -0.0609, -0.0522, -0.0422, -0.0422, -0.0199, -0.0298, -0.0186, -0.0373, -0.0373, -0.0112, 0.0075, 0.0634, 0.1453, 0.1491, 0.1466, 0.1851, 0.2224, 0.2286, 0.2, 0.2286, 0.1801, 0.2795, 0.236, 0.2547, 0.2733, 0.2609, 0.2857, 0.2857, 0.2609, 0.2174, 0.2062, 0.2919, 0.2795, 0.2099, 0.2012, 0.2795, 0.2919, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.4348, 0.5217, 0.559, 0.7143, 0.7143, 0.882, 0.7329, 0.6646, 0.7453, 0.7081, 0.5404, 0.6894, 0.8571, 0.8571, 0.9006, 0.8012, 0.8261, 0.9441, 0.9814, 0.9689, 0.8882, 0.8385, 1.0186, 1.0373, 1.0062, 1.0186, 0.8199, 1.0, 0.9068, 0.9068, 0.9068, 0.9503, 1.1429, 1.2857, 1.2174, 1.3354, 1.2857, 1.5093, 1.5466, 1.8012, 1.5466, 1.5155, 1.5217, 1.4099, 1.3851, 1.3913, 1.5652, 1.5528, 1.6335, 1.6335, 1.8696, 1.7764, 1.913, 1.8882, 1.7702, 1.9317, 2.0311, 1.9876, 1.8075, 1.7826, 1.7143, 1.5528, 1.5652, 1.8199, 1.8199, 2.0559, 2.1056, 2.2609, 2.1925, 2.1304, null, null]}
{"unit_id": "thread:1975765955763708153", "ticker_idx": 0, "ts": "2025-10-08T03:31:26+00:00", "symbol": "DRAM", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM prices in the fourth quarter are expected to rise by more than 30%, with some specifications likely to exceed a 50% increase", "tweet_type": "analysis", "timeline": "quarters", "summary": "Broad-based DRAM shortage driving 30–50% price increases in Q4 2025–Q1 2026; bullish for Korean and US DRAM makers.", "resolved_tickers": ["005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: Samsung Electronics + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Due to shortages in legacy memory supply, Korean and U.S. DRAM manufacturers have decided to suspend new price quotations for enterprise customers for one week.\nAccording to supply chain sources, DRAM prices in the fourth quarter are expected to rise by more than 30%, with some specifications likely to exceed a 50% increase, further accelerating the simultaneous price surge of DDR3.\n\nVeteran analysts in the memory industry note that this rally is not limited to “specific specifications” but rather a broad-based shortage that has triggered an overall price adjustment.\nStarting in the fourth quarter, DDR4 contract prices are expected to rise further by an average of 30–50%, with the most likely period of increase occurring between Q4 2025 and Q1 2026.\n\n---\n\nhttps://t.co/6CQlGzZPXi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.005927705196020927, "alpha_c_m1d": 0.017234908638119495, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.017234908638119495, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0028970083792374535, "alpha_c_p1d": 0.0016856346179577875, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0016856346179577875, "ret_p1w": 0.06338191186015119, "ret_signed_p1w": 0.06338191186015119, "alpha_spy_p1w": 0.07517786463876874, "alpha_c_p1w": 0.08536411584531056, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.021982203985159376, "ret_p1m": 0.30233017297874143, "ret_signed_p1m": 0.30233017297874143, "alpha_spy_p1m": 0.306489926101976, "alpha_c_p1m": 0.3075934123216977, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": -0.00526323934295625, "ret_p3m": 0.6536283086689896, "ret_signed_p3m": 0.6536283086689896, "alpha_spy_p3m": 0.6289040638695929, "alpha_c_p3m": 0.6571072723974213, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": -0.0034789637284317054, "ret_p6m": 1.0527641081377057, "ret_signed_p6m": 1.0527641081377057, "alpha_spy_p6m": 1.0728881392727032, "alpha_c_p6m": 1.114543901219003, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": -0.06177979308129733, "price_path": [-0.281, -0.2746, -0.2699, -0.2675, -0.2898, -0.2883, -0.28, -0.295, -0.2921, -0.2937, -0.2987, -0.2788, -0.2771, -0.2647, -0.2587, -0.2922, -0.2877, -0.2797, -0.2921, -0.2782, -0.278, -0.2692, -0.2661, -0.2503, -0.2538, -0.2538, -0.2741, -0.2797, -0.2864, -0.2991, -0.2871, -0.2758, -0.28, -0.2802, -0.2745, -0.2733, -0.3014, -0.2874, -0.281, -0.2756, -0.2688, -0.261, -0.2393, -0.213, -0.2048, -0.1665, -0.1573, -0.1197, -0.1447, -0.1052, -0.1052, -0.0931, -0.0738, -0.0744, -0.0718, -0.1126, -0.0897, -0.0933, -0.0658, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.067, 0.0444, 0.0305, 0.0634, 0.1164, 0.1339, 0.1603, 0.1487, 0.158, 0.1425, 0.1952, 0.2446, 0.213, 0.2653, 0.298, 0.3056, 0.4028, 0.3252, 0.2924, 0.3023, 0.2787, 0.3266, 0.3592, 0.3544, 0.3464, 0.2495, 0.3266, 0.2655, 0.2481, 0.2823, 0.1868, 0.1961, 0.2093, 0.2352, 0.2648, 0.2304, 0.2422, 0.282, 0.2805, 0.2712, 0.2921, 0.34, 0.32, 0.3443, 0.3126, 0.3291, 0.2847, 0.2432, 0.2982, 0.2978, 0.2842, 0.3494, 0.36, 0.3628, 0.3628, 0.4096, 0.4787, 0.4948, 0.4948, 0.4948, 0.5758, 0.6536, 0.6961, 0.7281, 0.7335, 0.7194, 0.7246, 0.704, 0.7241, 0.7531, 0.79, 0.8024, 0.7528, 0.7731, 0.8078, 0.8218, 0.7826, 0.905, 0.9731, 0.9889, 1.0485, 0.892, 1.0852, 1.0853, 0.957, 0.9493, 1.0537, 1.0364, 1.0274, 1.1233, 1.1277, 1.1277, 1.1277, 1.1277, 1.1947, 1.2648, 1.2836, 1.3911, 1.4271, 1.6133, 1.5568, 1.5568, 1.2823, 1.0401, 1.2653, 1.2247, 1.0309, 1.2408, 1.2741, 1.2306, 1.1806, 1.2909, 1.3149, 1.5056, 1.4064, 1.3926, 1.2255, 1.3117, 1.3192, 1.1908, 1.1908, 1.0955, 0.9608, 1.2008, 1.0528]}
{"unit_id": "thread:1975765955763708153", "ticker_idx": 1, "ts": "2025-10-08T03:31:26+00:00", "symbol": "memory", "kind": "sector", "geography": "US", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Korean and U.S. DRAM manufacturers have decided to suspend new price quotations for enterprise customers for one week", "tweet_type": "analysis", "timeline": "quarters", "summary": "Broad-based DRAM shortage driving 30–50% price increases in Q4 2025–Q1 2026; bullish for Korean and US DRAM makers.", "resolved_tickers": ["MU"], "resolution_method": "exchange_ticker", "resolution_reason": "US DRAM memory = Micron Technology (MU)", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Due to shortages in legacy memory supply, Korean and U.S. DRAM manufacturers have decided to suspend new price quotations for enterprise customers for one week.\nAccording to supply chain sources, DRAM prices in the fourth quarter are expected to rise by more than 30%, with some specifications likely to exceed a 50% increase, further accelerating the simultaneous price surge of DDR3.\n\nVeteran analysts in the memory industry note that this rally is not limited to “specific specifications” but rather a broad-based shortage that has triggered an overall price adjustment.\nStarting in the fourth quarter, DDR4 contract prices are expected to rise further by an average of 30–50%, with the most likely period of increase occurring between Q4 2025 and Q1 2026.\n\n---\n\nhttps://t.co/6CQlGzZPXi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.055204972681472175, "ret_signed_m1d": -0.055204972681472175, "alpha_spy_m1d": -0.04927726748545125, "alpha_c_m1d": -0.029140840538495638, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.026064132142976537, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021420551013971667, "ret_signed_p1d": -0.021420551013971667, "alpha_spy_p1d": -0.018523542634734214, "alpha_c_p1d": -0.018935422070154084, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0024851289438175828, "ret_p1w": -0.02340488073003566, "ret_signed_p1w": -0.02340488073003566, "alpha_spy_p1w": -0.011608927951418102, "alpha_c_p1w": -0.011008515489854598, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.012396365240181062, "ret_p1m": 0.21262856709063205, "ret_signed_p1m": 0.21262856709063205, "alpha_spy_p1m": 0.21678832021386663, "alpha_c_p1m": 0.1981228409622089, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": 0.014505726128423158, "ret_p3m": 0.5888679864399216, "ret_signed_p3m": 0.5888679864399216, "alpha_spy_p3m": 0.5641437416405248, "alpha_c_p3m": 0.49439319217912403, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": 0.09447479426079752, "ret_p6m": 0.8649734258483615, "ret_signed_p6m": 0.8649734258483615, "alpha_spy_p6m": 0.885097456983359, "alpha_c_p6m": 0.7278025170186437, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": 0.13717090882971772, "price_path": [-0.3668, -0.3969, -0.3893, -0.408, -0.4241, -0.4183, -0.4242, -0.4446, -0.4415, -0.4319, -0.4343, -0.4343, -0.4307, -0.4166, -0.445, -0.4667, -0.452, -0.4454, -0.4469, -0.4312, -0.3955, -0.3709, -0.3504, -0.3681, -0.3629, -0.3854, -0.3718, -0.3794, -0.404, -0.4112, -0.4016, -0.408, -0.4076, -0.4013, -0.3796, -0.3949, -0.3949, -0.3975, -0.3963, -0.3684, -0.332, -0.3315, -0.3123, -0.2881, -0.2344, -0.2005, -0.1978, -0.1924, -0.1865, -0.1412, -0.1725, -0.1629, -0.1538, -0.1777, -0.2025, -0.2003, -0.1666, -0.1492, -0.0738, -0.0657, -0.0443, -0.0284, -0.0552, 0.0, -0.0214, -0.076, -0.0192, -0.0482, -0.0234, 0.0305, 0.0297, 0.0521, 0.0293, 0.0098, 0.0517, 0.1144, 0.1199, 0.1291, 0.1531, 0.1398, 0.1385, 0.1942, 0.1093, 0.2084, 0.2126, 0.2105, 0.2888, 0.2268, 0.2461, 0.2056, 0.2559, 0.231, 0.1626, 0.1495, 0.0246, 0.0551, 0.1394, 0.1424, 0.1716, 0.1716, 0.2032, 0.2235, 0.2185, 0.1914, 0.1532, 0.207, 0.2563, 0.2843, 0.3418, 0.3151, 0.2269, 0.2084, 0.183, 0.1475, 0.2646, 0.353, 0.4073, 0.4057, 0.4586, 0.4586, 0.449, 0.4984, 0.4895, 0.4528, 0.4528, 0.6055, 0.5889, 0.7481, 0.7283, 0.6646, 0.7565, 0.7605, 0.7211, 0.6968, 0.7135, 0.8464, 0.8464, 0.8579, 0.9806, 1.0237, 1.0342, 0.9805, 1.0882, 1.2156, 1.2182, 1.1118, 1.2284, 1.135, 0.9312, 0.9489, 1.009, 0.952, 0.8999, 1.0887, 1.1071, 1.0954, 1.0954, 1.0349, 1.1427, 1.1243, 1.1794, 1.1428, 1.1277, 1.1836, 1.1152, 1.099, 1.1005, 0.9326, 1.04, 1.021, 0.8849, 0.9817, 1.0519, 1.1312, 1.0633, 1.169, 1.2488, 1.35, 1.3502, 1.2614, 1.1526, 1.0582, 1.0133, 0.9449, 0.8093, 0.8183, 0.6387, 0.7204, 0.8732, 0.865]}
{"unit_id": "thread:1976950774547824810", "ticker_idx": 0, "ts": "2025-10-11T09:59:29+00:00", "symbol": "GEV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem", "tweet_type": "analysis", "timeline": "years", "summary": "GE Vernova, Siemens Energy, MHI entering a golden era on AI-driven gas turbine demand; specialty alloy maker ATI also benefits from tight upstream supply.", "resolved_tickers": ["GEV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI-Driven Power Demand Surge in the U.S. Brings Gas Turbines to the Forefront: How GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries Are Responding\n\nThe U.S. “power crunch” triggered by AI data centers has thrust the traditional industrial equipment — medium and large gas turbines (heavy gas turbines) — into the center of a new technological revolution.\n\nMarket demand is surging at a breakneck pace, and the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem.\n\nHowever, despite this overwhelming order volume, the three companies have not opted for full-scale capacity expansion.\n\nThe reason: the ghost of the early-2000s dot-com bubble collapse, which devastated the power industry, still lingers.\n\nBalancing historical lessons with present profit opportunities, all three firms have chosen a strategy of “limited expansion.”\n\n----------------------------------------------\n\n1. From a “Compute Race” to a “Power Race”: Explosive Demand and Policy Catch-Up\n\nThe essence of the AI race is competition in computing power, which fundamentally depends on massive electricity supply.\n\nAs a result, power demand for AI data centers is skyrocketing at an unprecedented rate. Gas turbines have emerged as the backbone of the U.S. power grid, prized for their high efficiency, operational flexibility, and lower emissions compared with coal.\n\nThe market response has been swift and intense. According to research firm Oxcap, since mid-2023, the cost of building new gas power plants has more than doubled, primarily due to soaring gas turbine prices.\n\nPower utilities and Big Tech firms have already booked orders through the late 2030s to secure future energy reliability.\n\nBeyond AI, manufacturing reshoring, the expansion of electric vehicles, and the broad industrial electrification trend are all pushing electricity demand higher.\n\nPolicy support has reinforced this momentum. The Trump administration has designated gas turbines as a “bridge solution” to fill the power gap until next-generation nuclear plants come online.\n\nAt the same time, cuts to clean-energy subsidies have weakened the outlook for renewables, further increasing reliance on natural gas.\n\nMedia analyses show that following the passage of the “One Big Beautiful Bill Act,” forecasts for new wind, solar, and battery capacity over the next five years were revised down by 23%, which in turn stimulated demand for gas-fired generation.\n\n2. Lessons from History: The “Ghost” of the 2000s Dot-Com Bubble\n\nDespite the current boom, all three manufacturers are taking an extremely cautious approach to expansion.\n\nThe reason lies in the industry’s strong cyclicality and the trauma of the early-2000s gas turbine collapse.\n\nChristian Bruch, CEO of Siemens Energy, stated:\n“This industry is inherently cyclical. Even today’s boom will eventually end.”\n\nIn the early 2000s, the Internet boom led to overestimation of electricity demand, triggering a debt-fueled gas power plant construction frenzy.\n\nAfter the bubble burst, major utilities such as Calpine went bankrupt, and turbine makers were left with severe overcapacity.\n\nBill Newsom, CEO of Mitsubishi Power Americas, summed up his company’s biggest concern:\n“Our greatest challenge is distinguishing how much of the current demand is real — and how much is an illusion.”\n\nAccording to Ryan Luther, Transition Director at Enverus,\n“If everyone restrains expansion, prices will hold. But if even one player overbuilds, the entire market could collapse.”\n\n3. The Strategy of “Limited Expansion”\n\nCaught between historical caution and immediate opportunity, the three giants have chosen partial and gradual expansion:\n\n• GE Vernova is investing over $300 million to increase its annual production of medium- and large-scale gas turbines from 55 to 80 units.\n\n• Siemens Energy is pursuing a 30–40% capacity expansion but has explicitly ruled out making high-risk bets on the 2030s market.\n\n• Mitsubishi Heavy Industries is committing hundreds of millions of dollars to expand production lines within the U.S.\n\nArtem Abramov, Vice President at Rystad Energy, noted:\n“Their ramp-up pace is dramatically slower than the surge in demand over the past two years. That shows just how determined they are to avoid overpromising.”\nThe bottleneck is no longer limited to assembly plants — it is now extending upstream.\n\nThe supply of special alloys, critical materials for gas turbine components, has become extremely tight.\n\nKimberly Fields, CEO of specialty alloy manufacturer ATI, commented:\n\n“Last year, one major turbine maker agreed to co-invest with us in a new plant. It’s all about securing material supply. For these companies, the real risk isn’t price — it’s whether they can get the materials at all.”\n\nUltimately, as AI ignites a new “power race,” one of the oldest technologies in modern energy has stepped back into the spotlight.\n\nBut standing at the center of it all, GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries are haunted by the lessons of two decades ago — and have made a cold, calculated choice:\n\n“This time, we won’t overreach.”\n\n---\n\nhttps://t.co/lSaAcduzlE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06739673611682917, "ret_signed_m1d": -0.06739673611682917, "alpha_spy_m1d": -0.05228454402781246, "alpha_c_m1d": -0.05709685626502947, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.010299879851799698, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005923632079480323, "ret_signed_p1d": -0.005923632079480323, "alpha_spy_p1d": -0.004701977276023617, "alpha_c_p1d": -0.017675894032563755, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": 0.011752261953083432, "ret_p1w": -0.08319673534341065, "ret_signed_p1w": -0.08319673534341065, "alpha_spy_p1w": -0.09565454076035174, "alpha_c_p1w": -0.09679764949158098, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.013600914148170329, "ret_p1m": -0.11095991720270704, "ret_signed_p1m": -0.11095991720270704, "alpha_spy_p1m": -0.14106367268955344, "alpha_c_p1m": -0.12931455186986252, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.01835463466715548, "ret_p3m": -0.029502619201022706, "ret_signed_p3m": -0.029502619201022706, "alpha_spy_p3m": -0.07249772128733445, "alpha_c_p3m": -0.09063151352534826, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.061128894324325556, "ret_p6m": 0.40740759604609544, "ret_signed_p6m": 0.40740759604609544, "alpha_spy_p6m": 0.40750776507438735, "alpha_c_p6m": 0.3159917652738016, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.09141583077229387, "price_path": [-0.1347, -0.1208, -0.114, -0.127, -0.1531, -0.0296, -0.0375, -0.0056, -0.0009, -0.024, 0.0104, 0.0186, 0.0127, 0.0224, 0.0023, 0.0251, -0.0037, 0.0013, 0.0039, 0.0142, -0.0215, -0.0354, -0.0406, -0.0358, -0.0696, -0.0674, -0.0652, -0.0635, -0.0709, -0.0345, -0.0399, -0.0225, -0.0544, -0.0544, -0.1058, -0.1098, -0.0763, -0.1021, -0.0741, -0.0656, -0.0072, -0.0218, -0.035, -0.0302, -0.0468, -0.0516, -0.0575, -0.0371, -0.006, -0.0229, -0.0297, -0.0628, -0.0665, -0.0707, -0.0514, -0.0649, -0.0648, -0.0822, -0.0695, -0.065, -0.0352, -0.0216, -0.0674, 0.0, -0.0059, -0.0498, -0.0713, -0.0744, -0.0832, -0.0967, -0.1111, -0.0815, -0.0981, -0.0981, -0.1188, -0.108, -0.1141, -0.097, -0.103, -0.1544, -0.1362, -0.1509, -0.1124, -0.1052, -0.111, -0.112, -0.1386, -0.1075, -0.1095, -0.1436, -0.0812, -0.1388, -0.1422, -0.1042, -0.1164, -0.0899, -0.0899, -0.0744, -0.1097, -0.0716, -0.071, -0.0291, -0.0257, -0.0402, -0.035, 0.1158, 0.0868, 0.0366, 0.0515, 0.059, -0.0521, -0.0132, 0.0159, 0.0213, 0.0208, 0.0298, 0.0298, 0.0239, 0.0239, 0.018, 0.0086, 0.0086, 0.0487, 0.0515, 0.06, 0.0229, -0.0295, -0.0386, -0.0119, 0.0071, -0.0051, -0.0081, 0.0526, 0.0526, 0.0577, 0.0315, 0.0219, 0.0159, 0.0286, 0.0698, 0.099, 0.1079, 0.1218, 0.166, 0.205, 0.1525, 0.139, 0.2036, 0.2379, 0.2213, 0.2721, 0.2611, 0.2388, 0.2388, 0.2651, 0.2626, 0.289, 0.2824, 0.2845, 0.3586, 0.3529, 0.3536, 0.3492, 0.3609, 0.3004, 0.2993, 0.2587, 0.2189, 0.282, 0.2961, 0.3091, 0.2851, 0.2433, 0.2778, 0.3043, 0.3266, 0.3559, 0.3152, 0.364, 0.4053, 0.4274, 0.3493, 0.3184, 0.2631, 0.3489, 0.3827, 0.3886, 0.3886, 0.3867, 0.4074]}
{"unit_id": "thread:1976950774547824810", "ticker_idx": 1, "ts": "2025-10-11T09:59:29+00:00", "symbol": "Siemens Energy", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem", "tweet_type": "analysis", "timeline": "years", "summary": "GE Vernova, Siemens Energy, MHI entering a golden era on AI-driven gas turbine demand; specialty alloy maker ATI also benefits from tight upstream supply.", "resolved_tickers": ["ENR.DE"], "resolution_method": "exchange_ticker", "resolution_reason": "Siemens Energy listed on Frankfurt exchange as ENR.DE", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI-Driven Power Demand Surge in the U.S. Brings Gas Turbines to the Forefront: How GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries Are Responding\n\nThe U.S. “power crunch” triggered by AI data centers has thrust the traditional industrial equipment — medium and large gas turbines (heavy gas turbines) — into the center of a new technological revolution.\n\nMarket demand is surging at a breakneck pace, and the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem.\n\nHowever, despite this overwhelming order volume, the three companies have not opted for full-scale capacity expansion.\n\nThe reason: the ghost of the early-2000s dot-com bubble collapse, which devastated the power industry, still lingers.\n\nBalancing historical lessons with present profit opportunities, all three firms have chosen a strategy of “limited expansion.”\n\n----------------------------------------------\n\n1. From a “Compute Race” to a “Power Race”: Explosive Demand and Policy Catch-Up\n\nThe essence of the AI race is competition in computing power, which fundamentally depends on massive electricity supply.\n\nAs a result, power demand for AI data centers is skyrocketing at an unprecedented rate. Gas turbines have emerged as the backbone of the U.S. power grid, prized for their high efficiency, operational flexibility, and lower emissions compared with coal.\n\nThe market response has been swift and intense. According to research firm Oxcap, since mid-2023, the cost of building new gas power plants has more than doubled, primarily due to soaring gas turbine prices.\n\nPower utilities and Big Tech firms have already booked orders through the late 2030s to secure future energy reliability.\n\nBeyond AI, manufacturing reshoring, the expansion of electric vehicles, and the broad industrial electrification trend are all pushing electricity demand higher.\n\nPolicy support has reinforced this momentum. The Trump administration has designated gas turbines as a “bridge solution” to fill the power gap until next-generation nuclear plants come online.\n\nAt the same time, cuts to clean-energy subsidies have weakened the outlook for renewables, further increasing reliance on natural gas.\n\nMedia analyses show that following the passage of the “One Big Beautiful Bill Act,” forecasts for new wind, solar, and battery capacity over the next five years were revised down by 23%, which in turn stimulated demand for gas-fired generation.\n\n2. Lessons from History: The “Ghost” of the 2000s Dot-Com Bubble\n\nDespite the current boom, all three manufacturers are taking an extremely cautious approach to expansion.\n\nThe reason lies in the industry’s strong cyclicality and the trauma of the early-2000s gas turbine collapse.\n\nChristian Bruch, CEO of Siemens Energy, stated:\n“This industry is inherently cyclical. Even today’s boom will eventually end.”\n\nIn the early 2000s, the Internet boom led to overestimation of electricity demand, triggering a debt-fueled gas power plant construction frenzy.\n\nAfter the bubble burst, major utilities such as Calpine went bankrupt, and turbine makers were left with severe overcapacity.\n\nBill Newsom, CEO of Mitsubishi Power Americas, summed up his company’s biggest concern:\n“Our greatest challenge is distinguishing how much of the current demand is real — and how much is an illusion.”\n\nAccording to Ryan Luther, Transition Director at Enverus,\n“If everyone restrains expansion, prices will hold. But if even one player overbuilds, the entire market could collapse.”\n\n3. The Strategy of “Limited Expansion”\n\nCaught between historical caution and immediate opportunity, the three giants have chosen partial and gradual expansion:\n\n• GE Vernova is investing over $300 million to increase its annual production of medium- and large-scale gas turbines from 55 to 80 units.\n\n• Siemens Energy is pursuing a 30–40% capacity expansion but has explicitly ruled out making high-risk bets on the 2030s market.\n\n• Mitsubishi Heavy Industries is committing hundreds of millions of dollars to expand production lines within the U.S.\n\nArtem Abramov, Vice President at Rystad Energy, noted:\n“Their ramp-up pace is dramatically slower than the surge in demand over the past two years. That shows just how determined they are to avoid overpromising.”\nThe bottleneck is no longer limited to assembly plants — it is now extending upstream.\n\nThe supply of special alloys, critical materials for gas turbine components, has become extremely tight.\n\nKimberly Fields, CEO of specialty alloy manufacturer ATI, commented:\n\n“Last year, one major turbine maker agreed to co-invest with us in a new plant. It’s all about securing material supply. For these companies, the real risk isn’t price — it’s whether they can get the materials at all.”\n\nUltimately, as AI ignites a new “power race,” one of the oldest technologies in modern energy has stepped back into the spotlight.\n\nBut standing at the center of it all, GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries are haunted by the lessons of two decades ago — and have made a cold, calculated choice:\n\n“This time, we won’t overreach.”\n\n---\n\nhttps://t.co/lSaAcduzlE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023041434301885988, "ret_signed_m1d": -0.023041434301885988, "alpha_spy_m1d": -0.00792924221286928, "alpha_c_m1d": -0.01274155445008629, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.010299879851799698, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021658968016493807, "ret_signed_p1d": -0.021658968016493807, "alpha_spy_p1d": -0.0204373132130371, "alpha_c_p1d": -0.03341122996957724, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": 0.011752261953083432, "ret_p1w": -0.06036865955922088, "ret_signed_p1w": -0.06036865955922088, "alpha_spy_p1w": -0.07282646497616196, "alpha_c_p1w": -0.07396957370739121, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.013600914148170329, "ret_p1m": -0.009216559597382212, "ret_signed_p1m": -0.009216559597382212, "alpha_spy_p1m": -0.03932031508422862, "alpha_c_p1m": -0.02757119426453769, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.01835463466715548, "ret_p3m": 0.1377879720278139, "ret_signed_p3m": 0.1377879720278139, "alpha_spy_p3m": 0.09479286994150216, "alpha_c_p3m": 0.07665907770348834, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.061128894324325556, "ret_p6m": 0.36746571057086386, "ret_signed_p6m": 0.36746571057086386, "alpha_spy_p6m": 0.36756587959915576, "alpha_c_p6m": 0.27604987979857, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.09141583077229387, "price_path": [-0.1408, -0.1194, -0.1335, -0.1335, -0.1609, -0.1609, -0.1045, -0.1073, -0.1242, -0.0977, -0.0599, -0.0599, -0.0975, -0.0868, -0.1001, -0.0907, -0.1124, -0.1152, -0.0828, -0.0828, -0.0924, -0.0924, -0.1259, -0.1076, -0.1292, -0.1543, -0.1429, -0.1438, -0.155, -0.1353, -0.1587, -0.1482, -0.1639, -0.174, -0.2221, -0.2079, -0.1786, -0.1928, -0.1928, -0.1615, -0.1231, -0.1331, -0.1414, -0.1176, -0.1333, -0.1521, -0.1228, -0.1222, -0.1222, -0.111, -0.0907, -0.1045, -0.0901, -0.0866, -0.0866, -0.0401, -0.0005, -0.0226, -0.0226, -0.0332, -0.0332, 0.0115, -0.023, 0.0, -0.0217, -0.024, -0.0295, -0.07, -0.0604, -0.0831, -0.1171, -0.0887, -0.0433, -0.0507, -0.0419, -0.0152, -0.0221, -0.0111, -0.0111, 0.0092, -0.0115, -0.0115, -0.0115, -0.0184, -0.0092, -0.0106, -0.0106, 0.0184, 0.053, -0.0143, 0.0041, 0.0332, -0.071, -0.071, -0.0152, -0.0152, 0.0544, 0.0627, 0.0373, 0.0696, 0.0512, 0.0793, 0.0788, 0.082, 0.0917, 0.0917, 0.1382, 0.0899, 0.0899, 0.1005, 0.1005, 0.1005, 0.1074, 0.0991, 0.1129, 0.1129, 0.1129, 0.1129, 0.1115, 0.1097, 0.1097, 0.1097, 0.1318, 0.1756, 0.1733, 0.1926, 0.1378, 0.1599, 0.1816, 0.2014, 0.1627, 0.1816, 0.2562, 0.2364, 0.223, 0.2258, 0.2673, 0.3065, 0.2982, 0.3106, 0.3134, 0.3189, 0.3323, 0.3636, 0.4276, 0.3677, 0.3442, 0.4009, 0.4553, 0.388, 0.5046, 0.4876, 0.494, 0.506, 0.4843, 0.5359, 0.5134, 0.523, 0.5272, 0.53, 0.5613, 0.5194, 0.5406, 0.5027, 0.4384, 0.5064, 0.4171, 0.3861, 0.3768, 0.4597, 0.431, 0.4129, 0.3324, 0.3592, 0.387, 0.4078, 0.3486, 0.3028, 0.3662, 0.4023, 0.4661, 0.3902, 0.3366, 0.3143, 0.3167, 0.4087, 0.3856, 0.3856, 0.3856, 0.3675]}
{"unit_id": "thread:1976950774547824810", "ticker_idx": 2, "ts": "2025-10-11T09:59:29+00:00", "symbol": "Mitsubishi Heavy Industries", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem", "tweet_type": "analysis", "timeline": "years", "summary": "GE Vernova, Siemens Energy, MHI entering a golden era on AI-driven gas turbine demand; specialty alloy maker ATI also benefits from tight upstream supply.", "resolved_tickers": ["7011.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Mitsubishi Heavy Industries 7011.T Japan", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI-Driven Power Demand Surge in the U.S. Brings Gas Turbines to the Forefront: How GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries Are Responding\n\nThe U.S. “power crunch” triggered by AI data centers has thrust the traditional industrial equipment — medium and large gas turbines (heavy gas turbines) — into the center of a new technological revolution.\n\nMarket demand is surging at a breakneck pace, and the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem.\n\nHowever, despite this overwhelming order volume, the three companies have not opted for full-scale capacity expansion.\n\nThe reason: the ghost of the early-2000s dot-com bubble collapse, which devastated the power industry, still lingers.\n\nBalancing historical lessons with present profit opportunities, all three firms have chosen a strategy of “limited expansion.”\n\n----------------------------------------------\n\n1. From a “Compute Race” to a “Power Race”: Explosive Demand and Policy Catch-Up\n\nThe essence of the AI race is competition in computing power, which fundamentally depends on massive electricity supply.\n\nAs a result, power demand for AI data centers is skyrocketing at an unprecedented rate. Gas turbines have emerged as the backbone of the U.S. power grid, prized for their high efficiency, operational flexibility, and lower emissions compared with coal.\n\nThe market response has been swift and intense. According to research firm Oxcap, since mid-2023, the cost of building new gas power plants has more than doubled, primarily due to soaring gas turbine prices.\n\nPower utilities and Big Tech firms have already booked orders through the late 2030s to secure future energy reliability.\n\nBeyond AI, manufacturing reshoring, the expansion of electric vehicles, and the broad industrial electrification trend are all pushing electricity demand higher.\n\nPolicy support has reinforced this momentum. The Trump administration has designated gas turbines as a “bridge solution” to fill the power gap until next-generation nuclear plants come online.\n\nAt the same time, cuts to clean-energy subsidies have weakened the outlook for renewables, further increasing reliance on natural gas.\n\nMedia analyses show that following the passage of the “One Big Beautiful Bill Act,” forecasts for new wind, solar, and battery capacity over the next five years were revised down by 23%, which in turn stimulated demand for gas-fired generation.\n\n2. Lessons from History: The “Ghost” of the 2000s Dot-Com Bubble\n\nDespite the current boom, all three manufacturers are taking an extremely cautious approach to expansion.\n\nThe reason lies in the industry’s strong cyclicality and the trauma of the early-2000s gas turbine collapse.\n\nChristian Bruch, CEO of Siemens Energy, stated:\n“This industry is inherently cyclical. Even today’s boom will eventually end.”\n\nIn the early 2000s, the Internet boom led to overestimation of electricity demand, triggering a debt-fueled gas power plant construction frenzy.\n\nAfter the bubble burst, major utilities such as Calpine went bankrupt, and turbine makers were left with severe overcapacity.\n\nBill Newsom, CEO of Mitsubishi Power Americas, summed up his company’s biggest concern:\n“Our greatest challenge is distinguishing how much of the current demand is real — and how much is an illusion.”\n\nAccording to Ryan Luther, Transition Director at Enverus,\n“If everyone restrains expansion, prices will hold. But if even one player overbuilds, the entire market could collapse.”\n\n3. The Strategy of “Limited Expansion”\n\nCaught between historical caution and immediate opportunity, the three giants have chosen partial and gradual expansion:\n\n• GE Vernova is investing over $300 million to increase its annual production of medium- and large-scale gas turbines from 55 to 80 units.\n\n• Siemens Energy is pursuing a 30–40% capacity expansion but has explicitly ruled out making high-risk bets on the 2030s market.\n\n• Mitsubishi Heavy Industries is committing hundreds of millions of dollars to expand production lines within the U.S.\n\nArtem Abramov, Vice President at Rystad Energy, noted:\n“Their ramp-up pace is dramatically slower than the surge in demand over the past two years. That shows just how determined they are to avoid overpromising.”\nThe bottleneck is no longer limited to assembly plants — it is now extending upstream.\n\nThe supply of special alloys, critical materials for gas turbine components, has become extremely tight.\n\nKimberly Fields, CEO of specialty alloy manufacturer ATI, commented:\n\n“Last year, one major turbine maker agreed to co-invest with us in a new plant. It’s all about securing material supply. For these companies, the real risk isn’t price — it’s whether they can get the materials at all.”\n\nUltimately, as AI ignites a new “power race,” one of the oldest technologies in modern energy has stepped back into the spotlight.\n\nBut standing at the center of it all, GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries are haunted by the lessons of two decades ago — and have made a cold, calculated choice:\n\n“This time, we won’t overreach.”\n\n---\n\nhttps://t.co/lSaAcduzlE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.015112192089016707, "alpha_c_m1d": 0.010299879851799698, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.010299879851799698, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013871656429483958, "ret_signed_p1d": -0.013871656429483958, "alpha_spy_p1d": -0.012650001626027252, "alpha_c_p1d": -0.02562391838256739, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": 0.011752261953083432, "ret_p1w": 0.09289080337149591, "ret_signed_p1w": 0.09289080337149591, "alpha_spy_p1w": 0.08043299795455483, "alpha_c_p1w": 0.07928988922332558, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.013600914148170329, "ret_p1m": 0.07406493194746955, "ret_signed_p1m": 0.07406493194746955, "alpha_spy_p1m": 0.04396117646062314, "alpha_c_p1m": 0.05571029728031407, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.01835463466715548, "ret_p3m": 0.055734433504486036, "ret_signed_p3m": 0.055734433504486036, "alpha_spy_p3m": 0.012739331418174293, "alpha_c_p3m": -0.005394460819839519, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.061128894324325556, "ret_p6m": 0.18841385186091753, "ret_signed_p6m": 0.18841385186091753, "alpha_spy_p6m": 0.18851402088920943, "alpha_c_p6m": 0.09699802108862365, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.09141583077229387, "price_path": [-0.1876, -0.2007, -0.2071, -0.2071, -0.1579, -0.1592, -0.1125, -0.1172, -0.1283, -0.1409, -0.1086, -0.1036, -0.1103, -0.1244, -0.0742, -0.0303, -0.0389, -0.0439, -0.0439, -0.0221, 0.0149, -0.0429, -0.0345, -0.0239, -0.0357, -0.056, -0.0599, -0.0564, -0.0599, -0.0705, -0.0715, -0.0614, -0.0732, -0.0765, -0.0848, -0.1337, -0.1261, -0.1093, -0.0797, -0.0772, -0.0767, -0.0587, -0.0572, -0.0572, -0.0621, -0.0856, -0.0915, -0.0868, -0.0908, -0.0908, -0.0424, -0.0498, -0.0471, -0.0277, -0.0391, -0.0756, -0.0884, -0.086, 0.0161, 0.0111, 0.052, 0.0386, 0.0, 0.0, -0.0139, 0.0213, 0.0572, 0.0627, 0.0929, 0.0572, 0.0768, 0.1028, 0.0964, 0.1095, 0.0939, 0.1018, 0.1241, 0.1526, 0.1526, 0.1385, 0.1035, 0.1439, 0.1021, 0.0805, 0.0741, 0.0718, 0.0852, 0.0339, 0.0453, -0.0037, -0.0265, 0.0206, -0.0451, -0.0451, -0.0396, -0.0263, -0.0384, -0.0223, -0.0334, -0.0436, -0.0374, 0.0027, 0.0114, 0.0424, 0.0567, 0.0515, 0.0032, 0.0297, 0.0102, -0.0171, -0.0317, -0.056, -0.0384, -0.0374, -0.0349, -0.0359, -0.0409, -0.0461, -0.0384, -0.0488, -0.0488, -0.0488, -0.0488, 0.031, 0.0552, 0.0312, 0.0557, 0.057, 0.057, 0.1087, 0.1667, 0.189, 0.1543, 0.2088, 0.1922, 0.1878, 0.1635, 0.1397, 0.1293, 0.1271, 0.0983, 0.1216, 0.1194, 0.1159, 0.1514, 0.1781, 0.161, 0.1935, 0.2311, 0.265, 0.265, 0.2373, 0.2462, 0.2286, 0.194, 0.2267, 0.2267, 0.2299, 0.2299, 0.1917, 0.1771, 0.214, 0.242, 0.2868, 0.2192, 0.1461, 0.1813, 0.1793, 0.1345, 0.1514, 0.1434, 0.1843, 0.1754, 0.187, 0.2091, 0.2385, 0.2009, 0.2009, 0.1189, 0.0973, 0.1263, 0.1174, 0.1338, 0.0975, 0.0491, 0.1402, 0.1703, 0.1931, 0.1762, 0.1884]}
{"unit_id": "thread:1976950774547824810", "ticker_idx": 3, "ts": "2025-10-11T09:59:29+00:00", "symbol": "ATI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "one major turbine maker agreed to co-invest with us in a new plant. It's all about securing material supply", "tweet_type": "analysis", "timeline": "years", "summary": "GE Vernova, Siemens Energy, MHI entering a golden era on AI-driven gas turbine demand; specialty alloy maker ATI also benefits from tight upstream supply.", "resolved_tickers": ["ATI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Metal Fabrication", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI-Driven Power Demand Surge in the U.S. Brings Gas Turbines to the Forefront: How GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries Are Responding\n\nThe U.S. “power crunch” triggered by AI data centers has thrust the traditional industrial equipment — medium and large gas turbines (heavy gas turbines) — into the center of a new technological revolution.\n\nMarket demand is surging at a breakneck pace, and the three giants of gas turbine manufacturing — GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries (MHI) — are entering a golden era, with orders and prices rising in tandem.\n\nHowever, despite this overwhelming order volume, the three companies have not opted for full-scale capacity expansion.\n\nThe reason: the ghost of the early-2000s dot-com bubble collapse, which devastated the power industry, still lingers.\n\nBalancing historical lessons with present profit opportunities, all three firms have chosen a strategy of “limited expansion.”\n\n----------------------------------------------\n\n1. From a “Compute Race” to a “Power Race”: Explosive Demand and Policy Catch-Up\n\nThe essence of the AI race is competition in computing power, which fundamentally depends on massive electricity supply.\n\nAs a result, power demand for AI data centers is skyrocketing at an unprecedented rate. Gas turbines have emerged as the backbone of the U.S. power grid, prized for their high efficiency, operational flexibility, and lower emissions compared with coal.\n\nThe market response has been swift and intense. According to research firm Oxcap, since mid-2023, the cost of building new gas power plants has more than doubled, primarily due to soaring gas turbine prices.\n\nPower utilities and Big Tech firms have already booked orders through the late 2030s to secure future energy reliability.\n\nBeyond AI, manufacturing reshoring, the expansion of electric vehicles, and the broad industrial electrification trend are all pushing electricity demand higher.\n\nPolicy support has reinforced this momentum. The Trump administration has designated gas turbines as a “bridge solution” to fill the power gap until next-generation nuclear plants come online.\n\nAt the same time, cuts to clean-energy subsidies have weakened the outlook for renewables, further increasing reliance on natural gas.\n\nMedia analyses show that following the passage of the “One Big Beautiful Bill Act,” forecasts for new wind, solar, and battery capacity over the next five years were revised down by 23%, which in turn stimulated demand for gas-fired generation.\n\n2. Lessons from History: The “Ghost” of the 2000s Dot-Com Bubble\n\nDespite the current boom, all three manufacturers are taking an extremely cautious approach to expansion.\n\nThe reason lies in the industry’s strong cyclicality and the trauma of the early-2000s gas turbine collapse.\n\nChristian Bruch, CEO of Siemens Energy, stated:\n“This industry is inherently cyclical. Even today’s boom will eventually end.”\n\nIn the early 2000s, the Internet boom led to overestimation of electricity demand, triggering a debt-fueled gas power plant construction frenzy.\n\nAfter the bubble burst, major utilities such as Calpine went bankrupt, and turbine makers were left with severe overcapacity.\n\nBill Newsom, CEO of Mitsubishi Power Americas, summed up his company’s biggest concern:\n“Our greatest challenge is distinguishing how much of the current demand is real — and how much is an illusion.”\n\nAccording to Ryan Luther, Transition Director at Enverus,\n“If everyone restrains expansion, prices will hold. But if even one player overbuilds, the entire market could collapse.”\n\n3. The Strategy of “Limited Expansion”\n\nCaught between historical caution and immediate opportunity, the three giants have chosen partial and gradual expansion:\n\n• GE Vernova is investing over $300 million to increase its annual production of medium- and large-scale gas turbines from 55 to 80 units.\n\n• Siemens Energy is pursuing a 30–40% capacity expansion but has explicitly ruled out making high-risk bets on the 2030s market.\n\n• Mitsubishi Heavy Industries is committing hundreds of millions of dollars to expand production lines within the U.S.\n\nArtem Abramov, Vice President at Rystad Energy, noted:\n“Their ramp-up pace is dramatically slower than the surge in demand over the past two years. That shows just how determined they are to avoid overpromising.”\nThe bottleneck is no longer limited to assembly plants — it is now extending upstream.\n\nThe supply of special alloys, critical materials for gas turbine components, has become extremely tight.\n\nKimberly Fields, CEO of specialty alloy manufacturer ATI, commented:\n\n“Last year, one major turbine maker agreed to co-invest with us in a new plant. It’s all about securing material supply. For these companies, the real risk isn’t price — it’s whether they can get the materials at all.”\n\nUltimately, as AI ignites a new “power race,” one of the oldest technologies in modern energy has stepped back into the spotlight.\n\nBut standing at the center of it all, GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries are haunted by the lessons of two decades ago — and have made a cold, calculated choice:\n\n“This time, we won’t overreach.”\n\n---\n\nhttps://t.co/lSaAcduzlE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03543025213869899, "ret_signed_m1d": -0.03543025213869899, "alpha_spy_m1d": -0.020318060049682285, "alpha_c_m1d": -0.020318060049682285, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.015112192089016707, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02169191792654801, "ret_signed_p1d": 0.02169191792654801, "alpha_spy_p1d": 0.022913572730004717, "alpha_c_p1d": 0.022913572730004717, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.0012216548034567065, "ret_p1w": -0.01892021729706317, "ret_signed_p1w": -0.01892021729706317, "alpha_spy_p1w": -0.03137802271400425, "alpha_c_p1w": -0.03137802271400425, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.012457805416941081, "ret_p1m": 0.1799228693804129, "ret_signed_p1m": 0.1799228693804129, "alpha_spy_p1m": 0.1498191138935665, "alpha_c_p1m": 0.1498191138935665, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.030103755486846406, "ret_p3m": 0.42913944975131835, "ret_signed_p3m": 0.42913944975131835, "alpha_spy_p3m": 0.3861443476650066, "alpha_c_p3m": 0.3861443476650066, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.04299510208631174, "ret_p6m": 0.7748854280685027, "ret_signed_p6m": 0.7748854280685027, "alpha_spy_p6m": 0.7749855970967946, "alpha_c_p6m": 0.7749855970967946, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": -0.00010016902829190766, "price_path": [0.0819, 0.1156, 0.1257, 0.1187, 0.1142, 0.1339, 0.124, 0.1447, 0.1511, 0.1341, 0.1357, -0.0728, -0.0754, -0.0885, -0.1026, -0.1026, -0.1114, -0.1198, -0.1212, -0.0904, -0.0923, -0.108, -0.1268, -0.1256, -0.1317, -0.1403, -0.1347, -0.1117, -0.1182, -0.0783, -0.075, -0.0544, -0.0656, -0.0656, -0.0557, -0.078, -0.0701, -0.0645, -0.066, -0.0976, -0.0794, -0.0844, -0.0824, -0.0666, -0.0423, -0.0547, -0.0301, -0.0271, -0.0375, -0.0589, -0.0719, -0.0671, -0.002, -0.0448, -0.0198, 0.0004, -0.0055, -0.0205, 0.0083, 0.0064, 0.0114, -0.0084, -0.0354, 0.0, 0.0217, 0.0041, -0.0212, -0.0319, -0.0189, 0.0161, -0.0099, 0.0792, 0.0756, 0.1052, 0.1874, 0.2413, 0.2193, 0.1927, 0.1884, 0.1571, 0.1628, 0.1594, 0.1768, 0.1822, 0.1799, 0.194, 0.1878, 0.1975, 0.19, 0.1832, 0.1825, 0.1424, 0.1626, 0.1765, 0.1966, 0.2087, 0.2087, 0.2148, 0.1947, 0.189, 0.1855, 0.2099, 0.2101, 0.2008, 0.1981, 0.2621, 0.3335, 0.3131, 0.3102, 0.3095, 0.3002, 0.3227, 0.3651, 0.395, 0.4005, 0.4056, 0.4056, 0.4094, 0.4055, 0.4, 0.383, 0.383, 0.4365, 0.4449, 0.4565, 0.4563, 0.4291, 0.4738, 0.4878, 0.4994, 0.4852, 0.5111, 0.4986, 0.4986, 0.4857, 0.5024, 0.4836, 0.4889, 0.4866, 0.4961, 0.4669, 0.4669, 0.4497, 0.4675, 0.5466, 0.5365, 0.5506, 0.6097, 0.6349, 0.6515, 0.6717, 0.6849, 0.7345, 0.7345, 0.7685, 0.7904, 0.8419, 0.9146, 0.9194, 0.9407, 0.9284, 0.971, 0.9714, 1.0055, 0.8915, 0.9501, 0.858, 0.8087, 0.8884, 0.9158, 0.8977, 0.7607, 0.7109, 0.7773, 0.7936, 0.8033, 0.778, 0.7088, 0.764, 0.8027, 0.7956, 0.7346, 0.6923, 0.6329, 0.753, 0.8227, 0.7671, 0.7671, 0.7831, 0.7749]}
{"unit_id": "orig:1991509375874449482", "ticker_idx": 0, "ts": "2025-11-20T14:10:10+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we estimate FY2027 revenue to reach USD 269 billion (a 19% upward revision from the prior USD 225 billion estimate)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Citi raises NVDA FY2027 revenue estimate to $269B (+19%), lifts FY2027 GPU unit shipments to 10.2M on rising CoWoS supply for Blackwell/Rubin.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi AI GPU Model\nAs data points increasingly suggest rising CoWoS supply for Blackwell and Rubin, we revise our GPU unit sales estimates (see memo by Citi TSMC analyst Laura Chen).\n    \n1. Units: We maintain FY2026 unit shipments at 7.1 million, but raise FY2027 unit shipments to 10.2 million (up 44% YoY versus the prior 8.7 million forecast).\n    \n2. Blackwell & Rubin: Between calendar 2025 and 2026, we expect combined Blackwell and Rubin unit shipments to increase from 14.2 million to 16.2 million.\n    \n3. Revenue outlook: Supported by higher unit shipments and a favorable product mix, we estimate FY2027 revenue to reach USD 269 billion (a 19% upward revision from the prior USD 225 billion estimate).\n    \n4. Rationale for the revision: The USD 44 billion upward adjustment to GPU revenue mainly reflects changes to our assumptions for total NVIDIA unit shipments, as well as the larger integrated scale of Blackwell and Rubin. For the 2025–2026 period, we update our combined Blackwell and Rubin revenue estimate from USD 358 billion to USD 417 billion (+16%).\n    \n5. Revenue mix: We expect AI GPU revenue to account for the high-80s to 90% of NVIDIA’s total data center revenue in FY2026–2027, with the remainder coming mainly from networking (Ethernet and InfiniBand), standalone CPUs, and software/services.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03255089940351219, "ret_signed_m1d": 0.03255089940351219, "alpha_spy_m1d": 0.017072637846835992, "alpha_c_m1d": -0.011558649598769621, "bench_spy_m1d": 0.015478261556676198, "bench_c_m1d": 0.04410954900228181, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009743076928405703, "ret_signed_p1d": -0.009743076928405703, "alpha_spy_p1d": -0.019704299905038525, "alpha_c_p1d": -0.01291139446864542, "bench_spy_p1d": 0.009961222976632822, "bench_c_p1d": 0.003168317540239718, "ret_p1w": -0.002103636162266831, "ret_signed_p1w": -0.002103636162266831, "alpha_spy_p1w": -0.04371090624740459, "alpha_c_p1w": -0.07165144068497487, "bench_spy_p1w": 0.04160727008513776, "bench_c_p1w": 0.06954780452270803, "ret_p1m": 0.0019933897789183863, "ret_signed_p1m": 0.0019933897789183863, "alpha_spy_p1m": -0.04409042435551447, "alpha_c_p1m": -0.09376179730878698, "bench_spy_p1m": 0.046083814134432854, "bench_c_p1m": 0.09575518708770536, "ret_p3m": 0.024027375570905285, "ret_signed_p3m": 0.024027375570905285, "alpha_spy_p3m": -0.025530021516462664, "alpha_c_p3m": -0.23336426678275912, "bench_spy_p3m": 0.04955739708736795, "bench_c_p3m": 0.2573916423536644, "ret_p6m": 0.2474798141406489, "ret_signed_p6m": 0.2474798141406489, "alpha_spy_p6m": 0.10825438381841268, "alpha_c_p6m": -0.4691339703335311, "bench_spy_p6m": 0.13922543032223622, "bench_c_p6m": 0.71661378447418, "price_path": [-0.0047, 0.0062, 0.0053, -0.0027, -0.0358, -0.0358, -0.0546, -0.0555, -0.0498, -0.0755, -0.0683, -0.0547, -0.0184, -0.0192, -0.0156, -0.016, -0.0319, -0.0573, -0.0244, -0.022, 0.0164, -0.0122, -0.0203, -0.0163, -0.0136, 0.0067, 0.0329, 0.0365, 0.0457, 0.0386, 0.0271, 0.0244, 0.0469, 0.066, 0.014, 0.0425, -0.0034, -0.0045, 0.0065, 0.0143, 0.0111, 0.0029, -0.002, 0.0084, 0.0311, 0.0601, 0.1129, 0.1461, 0.1232, 0.121, 0.1453, 0.0999, 0.0807, 0.0412, 0.0416, 0.1019, 0.0693, 0.0729, 0.0344, 0.0528, 0.033, 0.004, 0.0326, 0.0, -0.0097, 0.0106, -0.0156, -0.0021, -0.0021, -0.0202, -0.004, 0.0045, -0.0058, 0.0152, 0.0099, 0.0272, 0.024, 0.0174, 0.0017, -0.0311, -0.024, -0.0161, -0.0536, -0.0359, 0.002, 0.0169, 0.0475, 0.0442, 0.0442, 0.0548, 0.042, 0.0383, 0.0325, 0.0325, 0.0455, 0.0415, 0.0366, 0.0469, 0.0244, 0.0234, 0.0239, 0.0287, 0.0139, 0.0355, 0.031, 0.031, -0.0142, 0.0149, 0.0233, 0.039, 0.0323, 0.0437, 0.0603, 0.0658, 0.0581, 0.0276, -0.0016, -0.0357, -0.0484, 0.0265, 0.0521, 0.0438, 0.0522, 0.0349, 0.0121, 0.0121, 0.024, 0.0407, 0.0402, 0.0509, 0.0605, 0.0677, 0.0827, 0.0236, -0.019, 0.0102, -0.0032, 0.0133, 0.015, -0.0156, 0.0112, 0.0229, 0.03, 0.014, -0.002, 0.0144, 0.0073, -0.0012, -0.0114, -0.0438, -0.0276, -0.03, -0.0107, -0.0519, -0.0725, -0.0855, -0.0344, -0.027, -0.0179, -0.0179, -0.0165, -0.014, 0.0081, 0.0182, 0.0443, 0.0481, 0.088, 0.101, 0.0982, 0.1166, 0.1187, 0.1066, 0.1211, 0.1053, 0.1531, 0.1993, 0.1802, 0.1585, 0.1049, 0.0987, 0.0989, 0.0879, 0.1506, 0.171, 0.1915, 0.2149, 0.2223, 0.2503, 0.3052, 0.2475]}
{"unit_id": "orig:1980088657391550751", "ticker_idx": 0, "ts": "2025-10-20T01:48:18+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "maintains its 'Buy' rating on SK hynix, raising the target price to KRW 590,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DRAM/NAND price forecasts; upgrades Samsung to Buy PT KRW 112K, maintains Buy on SK Hynix PT KRW 590K, reinstates Buy on Nanya PT TWD 110 on demand surge through H1 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: DDR and NAND Prices Significantly Raised, SK hynix and Samsung to Benefit the Most\nUBS (T. Arcuri, Oct 20, 2025)\n\nUBS raised its forecast for DDR and NAND contract prices, expecting DRAM prices to continue rising from Q4 2025 through the first half of 2026: DDR contract prices are projected to increase by +17% QoQ in Q4 2025, +15% in Q1 2026, and +7% in Q2 2026; NAND contract prices are expected to rise by +15%, +10%, and +5%, respectively.\nAccording to institutional surveys, the top four U.S. cloud providers’ DDR demand in 2026 will double compared to 2025, while some smartphone customers’ demand is also expected to grow by about 20%. Supply and demand for enterprise SSDs are likewise tightening. UBS raised its forecast for 2026 server shipments to 18 million units (+5.7% YoY). DRAM end demand is expected to grow by 20.6% and 19.1% in 2025 and 2026, respectively—significantly outpacing supply growth—leading to an estimated supply-demand gap of around 3 percentage points.\n\nAgainst this backdrop, UBS raised its capital expenditure forecasts for SK hynix to KRW 28/35/36 trillion for 2025/26/27 (previously 25/31/32 trillion) and believes DRAM makers will struggle to achieve over 20% capacity growth. The firm maintains its “Buy” rating on SK hynix, raising the target price to KRW 590,000; upgrades Samsung Electronics to “Buy” with a target price of KRW 112,000; and reinstates coverage on Nanya Technology with a “Buy” rating and a target price of TWD 110.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04119462522047679, "ret_signed_m1d": -0.04119462522047679, "alpha_spy_m1d": -0.030901206271412796, "alpha_c_m1d": -0.028551282693331737, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013388283854285032, "ret_signed_p1d": -0.013388283854285032, "alpha_spy_p1d": -0.013373287934272504, "alpha_c_p1d": -0.007858595812623381, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.1019566925800015, "ret_signed_p1w": 0.1019566925800015, "alpha_spy_p1w": 0.08119100090416365, "alpha_c_p1w": 0.0653228619626689, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": 0.17404730285143066, "ret_signed_p1m": 0.17404730285143066, "alpha_spy_p1m": 0.1907610786521854, "alpha_c_p1m": 0.2141659301769061, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.5438442835073052, "ret_signed_p3m": 0.5438442835073052, "alpha_spy_p3m": 0.5096040261919479, "alpha_c_p3m": 0.39862376074724315, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 1.2777070106448454, "ret_signed_p6m": 1.2777070106448454, "alpha_spy_p6m": 1.2373163230380224, "alpha_c_p6m": 0.9718881072228924, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.4467, -0.4457, -0.4529, -0.4611, -0.4601, -0.458, -0.4375, -0.4694, -0.4694, -0.458, -0.4683, -0.4611, -0.4724, -0.4508, -0.4467, -0.4282, -0.4313, -0.4313, -0.4498, -0.4591, -0.4745, -0.4961, -0.4838, -0.4663, -0.4622, -0.4652, -0.447, -0.4459, -0.4727, -0.4634, -0.4593, -0.4531, -0.4367, -0.4295, -0.4068, -0.3738, -0.3677, -0.3234, -0.3182, -0.2832, -0.3131, -0.2678, -0.2678, -0.277, -0.2564, -0.2636, -0.2657, -0.3069, -0.2812, -0.2842, -0.2585, -0.1854, -0.1854, -0.1854, -0.1854, -0.1854, -0.1854, -0.1184, -0.1452, -0.1524, -0.1298, -0.068, -0.0412, 0.0, -0.0134, -0.0082, -0.0144, 0.0505, 0.102, 0.0731, 0.1493, 0.1699, 0.1514, 0.277, 0.207, 0.1926, 0.2214, 0.1946, 0.2482, 0.275, 0.2709, 0.2606, 0.1535, 0.2482, 0.174, 0.1576, 0.1761, 0.0731, 0.0711, 0.069, 0.0793, 0.1213, 0.0924, 0.1089, 0.1502, 0.1378, 0.1172, 0.1213, 0.1893, 0.1666, 0.2099, 0.1646, 0.1769, 0.1419, 0.0924, 0.1357, 0.1378, 0.1275, 0.1955, 0.2037, 0.212, 0.212, 0.2347, 0.3192, 0.3418, 0.3418, 0.3418, 0.3954, 0.4346, 0.4964, 0.5294, 0.5583, 0.5335, 0.5438, 0.5212, 0.5294, 0.5438, 0.5583, 0.5748, 0.5315, 0.5253, 0.5562, 0.5809, 0.517, 0.649, 0.7335, 0.7747, 0.8736, 0.7108, 0.8695, 0.8551, 0.7355, 0.7294, 0.8283, 0.8056, 0.7726, 0.8304, 0.8139, 0.8139, 0.8139, 0.8139, 0.8427, 0.9561, 0.9602, 1.0715, 1.0983, 1.2694, 1.191, 1.191, 0.939, 0.7532, 0.9432, 0.9081, 0.7263, 0.937, 0.9721, 0.9205, 0.8792, 1.0113, 1.0031, 1.1807, 1.0919, 1.0795, 0.9267, 1.0361, 1.0547, 0.9267, 0.9267, 0.8028, 0.6665, 0.8523, 0.714, 0.8089, 0.8296, 0.8915, 1.1332, 1.0609, 1.1208, 1.1476, 1.2777]}
{"unit_id": "orig:1980088657391550751", "ticker_idx": 1, "ts": "2025-10-20T01:48:18+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "upgrades Samsung Electronics to 'Buy' with a target price of KRW 112,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DRAM/NAND price forecasts; upgrades Samsung to Buy PT KRW 112K, maintains Buy on SK Hynix PT KRW 590K, reinstates Buy on Nanya PT TWD 110 on demand surge through H1 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: DDR and NAND Prices Significantly Raised, SK hynix and Samsung to Benefit the Most\nUBS (T. Arcuri, Oct 20, 2025)\n\nUBS raised its forecast for DDR and NAND contract prices, expecting DRAM prices to continue rising from Q4 2025 through the first half of 2026: DDR contract prices are projected to increase by +17% QoQ in Q4 2025, +15% in Q1 2026, and +7% in Q2 2026; NAND contract prices are expected to rise by +15%, +10%, and +5%, respectively.\nAccording to institutional surveys, the top four U.S. cloud providers’ DDR demand in 2026 will double compared to 2025, while some smartphone customers’ demand is also expected to grow by about 20%. Supply and demand for enterprise SSDs are likewise tightening. UBS raised its forecast for 2026 server shipments to 18 million units (+5.7% YoY). DRAM end demand is expected to grow by 20.6% and 19.1% in 2025 and 2026, respectively—significantly outpacing supply growth—leading to an estimated supply-demand gap of around 3 percentage points.\n\nAgainst this backdrop, UBS raised its capital expenditure forecasts for SK hynix to KRW 28/35/36 trillion for 2025/26/27 (previously 25/31/32 trillion) and believes DRAM makers will struggle to achieve over 20% capacity growth. The firm maintains its “Buy” rating on SK hynix, raising the target price to KRW 590,000; upgrades Samsung Electronics to “Buy” with a target price of KRW 112,000; and reinstates coverage on Nanya Technology with a “Buy” rating and a target price of TWD 110.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002038693068074693, "ret_signed_m1d": -0.002038693068074693, "alpha_spy_m1d": 0.0082547258809893, "alpha_c_m1d": 0.00899567002942181, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.011034363097496502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0061162395860097085, "ret_signed_p1d": -0.0061162395860097085, "alpha_spy_p1d": -0.00610124366599718, "alpha_c_p1d": -0.0066366592900064525, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": 0.000520419703996744, "ret_p1w": 0.03975535683183096, "ret_signed_p1w": 0.03975535683183096, "alpha_spy_p1w": 0.018989665155993096, "alpha_c_p1w": 0.0022454244184331174, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03750993241339784, "ret_p1m": -0.0030580796975584468, "ret_signed_p1m": -0.0030580796975584468, "alpha_spy_p1m": 0.01365569610319628, "alpha_c_p1m": 0.028726492666849124, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.03178457236440757, "ret_p3m": 0.47400115025816647, "ret_signed_p3m": 0.47400115025816647, "alpha_spy_p3m": 0.4397608929428092, "alpha_c_p3m": 0.46299705029383453, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.011004099964331937, "ret_p6m": 1.119605678926571, "ret_signed_p6m": 1.119605678926571, "alpha_spy_p6m": 1.079214991319748, "alpha_c_p6m": 1.090086169346484, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.029519509580086867, "price_path": [-0.3261, -0.3302, -0.3312, -0.2856, -0.2835, -0.2632, -0.2754, -0.3008, -0.2927, -0.2906, -0.3018, -0.2845, -0.2713, -0.2795, -0.2784, -0.2703, -0.2734, -0.2734, -0.2896, -0.2896, -0.2845, -0.2835, -0.2754, -0.2744, -0.2866, -0.2835, -0.2937, -0.2927, -0.314, -0.2987, -0.2916, -0.2886, -0.2947, -0.2886, -0.2744, -0.2632, -0.2551, -0.2348, -0.2236, -0.1942, -0.2064, -0.1851, -0.1851, -0.1526, -0.1404, -0.1333, -0.1262, -0.1546, -0.1417, -0.1448, -0.1233, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0377, -0.0489, -0.0663, -0.0316, -0.0041, -0.002, 0.0, -0.0061, 0.0051, -0.0163, 0.0071, 0.0398, 0.0143, 0.0245, 0.0612, 0.0958, 0.1325, 0.0693, 0.0255, 0.0112, -0.002, 0.0255, 0.055, 0.051, 0.0479, -0.0092, 0.0255, -0.0031, -0.0163, 0.0255, -0.0336, -0.0143, 0.0122, 0.0479, 0.055, 0.0245, 0.0275, 0.054, 0.0652, 0.0714, 0.105, 0.1162, 0.105, 0.1009, 0.0938, 0.1101, 0.0683, 0.0479, 0.0999, 0.0968, 0.0836, 0.1264, 0.1366, 0.1325, 0.1325, 0.1927, 0.2241, 0.2282, 0.2282, 0.2282, 0.3163, 0.4146, 0.4228, 0.4443, 0.4218, 0.4238, 0.4218, 0.4095, 0.4371, 0.474, 0.5252, 0.5293, 0.4873, 0.5314, 0.56, 0.558, 0.558, 0.6338, 0.6635, 0.6461, 0.644, 0.5406, 0.7157, 0.7321, 0.6317, 0.6246, 0.7045, 0.6983, 0.7188, 0.8294, 0.8561, 0.8561, 0.8561, 0.8561, 0.9462, 0.9472, 0.9769, 1.0486, 1.0845, 1.233, 1.2177, 1.2177, 0.9985, 0.7639, 0.9626, 0.9278, 0.7772, 0.9247, 0.9462, 0.9247, 0.8796, 0.9329, 0.9862, 1.1357, 1.0538, 1.0425, 0.9083, 0.9431, 0.936, 0.8448, 0.8448, 0.8096, 0.7162, 0.9466, 0.8312, 0.9112, 0.9821, 1.017, 1.1607, 1.0939, 1.1145, 1.0632, 1.1196]}
{"unit_id": "orig:1980088657391550751", "ticker_idx": 2, "ts": "2025-10-20T01:48:18+00:00", "symbol": "Nanya Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reinstates coverage on Nanya Technology with a 'Buy' rating and a target price of TWD 110", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises DRAM/NAND price forecasts; upgrades Samsung to Buy PT KRW 112K, maintains Buy on SK Hynix PT KRW 590K, reinstates Buy on Nanya PT TWD 110 on demand surge through H1 2026.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya Technology Corp listed on TWSE as 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: DDR and NAND Prices Significantly Raised, SK hynix and Samsung to Benefit the Most\nUBS (T. Arcuri, Oct 20, 2025)\n\nUBS raised its forecast for DDR and NAND contract prices, expecting DRAM prices to continue rising from Q4 2025 through the first half of 2026: DDR contract prices are projected to increase by +17% QoQ in Q4 2025, +15% in Q1 2026, and +7% in Q2 2026; NAND contract prices are expected to rise by +15%, +10%, and +5%, respectively.\nAccording to institutional surveys, the top four U.S. cloud providers’ DDR demand in 2026 will double compared to 2025, while some smartphone customers’ demand is also expected to grow by about 20%. Supply and demand for enterprise SSDs are likewise tightening. UBS raised its forecast for 2026 server shipments to 18 million units (+5.7% YoY). DRAM end demand is expected to grow by 20.6% and 19.1% in 2025 and 2026, respectively—significantly outpacing supply growth—leading to an estimated supply-demand gap of around 3 percentage points.\n\nAgainst this backdrop, UBS raised its capital expenditure forecasts for SK hynix to KRW 28/35/36 trillion for 2025/26/27 (previously 25/31/32 trillion) and believes DRAM makers will struggle to achieve over 20% capacity growth. The firm maintains its “Buy” rating on SK hynix, raising the target price to KRW 590,000; upgrades Samsung Electronics to “Buy” with a target price of KRW 112,000; and reinstates coverage on Nanya Technology with a “Buy” rating and a target price of TWD 110.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028037383177570097, "ret_signed_m1d": -0.028037383177570097, "alpha_spy_m1d": -0.017743964228506104, "alpha_c_m1d": -0.015394040650425045, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014018691588784993, "ret_signed_p1d": -0.014018691588784993, "alpha_spy_p1d": -0.014003695668772465, "alpha_c_p1d": -0.008489003547123342, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.12149532710280364, "ret_signed_p1w": 0.12149532710280364, "alpha_spy_p1w": 0.10072963542696578, "alpha_c_p1w": 0.08486149648547103, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": 0.5, "ret_signed_p1m": 0.5, "alpha_spy_p1m": 0.5167137758007547, "alpha_c_p1m": 0.5401186273254754, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 1.2196261682242993, "ret_signed_p3m": 1.2196261682242993, "alpha_spy_p3m": 1.185385910908942, "alpha_c_p3m": 1.0744056454642372, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 1.05607476635514, "ret_signed_p6m": 1.05607476635514, "alpha_spy_p6m": 1.015684078748317, "alpha_c_p6m": 0.750255862933187, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.5977, -0.5981, -0.5958, -0.5953, -0.6056, -0.5804, -0.5836, -0.5836, -0.5911, -0.585, -0.5949, -0.6056, -0.5893, -0.5907, -0.5565, -0.5696, -0.5668, -0.5514, -0.5575, -0.5724, -0.5836, -0.5659, -0.5762, -0.5612, -0.5565, -0.5533, -0.564, -0.5607, -0.5631, -0.5734, -0.5668, -0.5495, -0.5047, -0.5, -0.4963, -0.4916, -0.472, -0.4636, -0.4103, -0.3533, -0.315, -0.2523, -0.2598, -0.2449, -0.243, -0.2897, -0.2841, -0.3364, -0.3364, -0.3178, -0.2879, -0.2551, -0.2243, -0.2243, -0.1477, -0.1514, -0.0794, -0.0794, -0.1065, -0.1673, -0.1935, -0.1131, -0.028, 0.0, -0.014, 0.028, 0.0234, 0.0234, 0.1215, 0.215, 0.243, 0.257, 0.2383, 0.2804, 0.229, 0.2897, 0.3832, 0.3738, 0.5093, 0.5374, 0.5234, 0.528, 0.4813, 0.5561, 0.5, 0.4953, 0.4533, 0.3084, 0.3364, 0.3411, 0.2757, 0.3645, 0.3645, 0.3972, 0.3972, 0.4112, 0.4112, 0.4299, 0.528, 0.514, 0.4579, 0.4579, 0.514, 0.5187, 0.4766, 0.5374, 0.5888, 0.6308, 0.6822, 0.6495, 0.7664, 0.7664, 0.7664, 0.7617, 0.8178, 0.8037, 0.8037, 0.9346, 0.9439, 1.1355, 1.2523, 1.257, 1.0327, 1.2336, 1.1495, 1.2103, 1.2196, 1.3364, 1.5701, 1.5421, 1.3458, 1.4907, 1.5374, 1.6729, 1.5888, 1.7757, 1.7757, 2.0514, 1.7477, 1.5935, 1.757, 1.5701, 1.4673, 1.5654, 1.5327, 1.5981, 1.5981, 1.5981, 1.5981, 1.5981, 1.5981, 1.5981, 1.5981, 1.7009, 1.785, 1.6916, 1.6682, 1.6682, 1.6636, 1.3972, 1.1916, 1.4065, 1.1776, 0.9626, 1.1589, 1.3738, 1.2383, 1.2243, 1.4439, 1.5374, 1.5794, 1.4299, 1.1869, 1.0841, 1.0234, 1.1168, 1.1075, 1.0514, 1.0607, 0.8551, 0.972, 0.8738, 0.8738, 0.8738, 0.9252, 1.0841, 0.9159, 1.0093, 1.1075, 1.0561]}
{"unit_id": "orig:1998639142906015805", "ticker_idx": 0, "ts": "2025-12-10T06:21:19+00:00", "symbol": "ORCL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If you believe OpenAI will dominate the future, buy Oracle. Oracle is essentially a 3× levered play on OpenAI.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Buy Oracle as a 3× levered play on OpenAI's dominance of the future.", "resolved_tickers": ["ORCL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "This is an absolutely insane scale.\n\nIf you believe OpenAI will dominate the future, buy Oracle.\n\nOracle is essentially a 3× levered play on OpenAI. https://t.co/4EeKrPkIJs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006636482458458048, "ret_signed_m1d": -0.006636482458458048, "alpha_spy_m1d": -4.7911617059859246e-05, "alpha_c_m1d": -4.7911617059859246e-05, "bench_spy_m1d": -0.006588570841398189, "bench_c_m1d": -0.006588570841398189, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.10833587247291354, "ret_signed_p1d": -0.10833587247291354, "alpha_spy_p1d": -0.11066282060344079, "alpha_c_p1d": -0.11066282060344079, "bench_spy_p1d": 0.002326948130527251, "bench_c_p1d": 0.002326948130527251, "ret_p1w": -0.19976678926591762, "ret_signed_p1w": -0.19976678926591762, "alpha_spy_p1w": -0.17624912396667336, "alpha_c_p1w": -0.17624912396667336, "bench_spy_p1w": -0.02351766529924426, "bench_c_p1w": -0.02351766529924426, "ret_p1m": -0.1495897195619581, "ret_signed_p1m": -0.1495897195619581, "alpha_spy_p1m": -0.15537445994230625, "alpha_c_p1m": -0.15537445994230625, "bench_spy_p1m": 0.005784740380348152, "bench_c_p1m": 0.005784740380348152, "ret_p3m": -0.31859273885478356, "ret_signed_p3m": -0.31859273885478356, "alpha_spy_p3m": -0.30798180181468837, "alpha_c_p3m": -0.30798180181468837, "bench_spy_p3m": -0.010610937040095192, "bench_c_p3m": -0.010610937040095192, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.3079, 0.3525, 0.3727, 0.3492, 0.3278, 0.3817, 0.4689, 0.4048, 0.3808, 0.3041, 0.2689, 0.2657, 0.2589, 0.2937, 0.2927, 0.2809, 0.3053, 0.2724, 0.292, 0.3316, 0.3137, 0.3811, 0.3407, 0.3615, 0.4035, 0.3063, 0.2429, 0.2338, 0.2226, 0.2559, 0.2705, 0.2618, 0.2593, 0.2345, 0.1519, 0.1776, 0.1562, 0.1128, 0.1224, 0.0932, 0.0729, 0.0799, 0.0589, 0.0178, -0.0244, -0.0007, -0.0141, -0.0113, 0.0113, -0.0552, -0.1087, -0.1019, -0.1165, -0.0809, -0.0809, -0.0944, -0.099, -0.0982, -0.0685, -0.0389, -0.0243, -0.0111, -0.0066, 0.0, -0.1083, -0.1482, -0.1708, -0.1541, -0.1998, -0.1927, -0.1392, -0.1104, -0.1241, -0.1144, -0.1144, -0.1122, -0.1239, -0.1157, -0.126, -0.126, -0.1224, -0.1364, -0.1312, -0.1353, -0.1496, -0.1075, -0.0798, -0.0905, -0.1295, -0.1464, -0.1409, -0.1409, -0.1911, -0.2182, -0.1989, -0.2035, -0.1798, -0.2137, -0.2231, -0.2401, -0.2601, -0.2804, -0.3046, -0.3406, -0.3864, -0.3579, -0.296, -0.2811, -0.2934, -0.2965, -0.28, -0.28, -0.3078, -0.2979, -0.2962, -0.3342, -0.3647, -0.343, -0.3351, -0.3242, -0.3463, -0.329, -0.3301, -0.315, -0.3041, -0.3123, -0.3186, -0.3283, -0.2666, -0.2844, -0.3026, -0.2988, -0.3045, -0.3126, -0.3008, -0.327, -0.3061, -0.3387, -0.3435, -0.3579, -0.3721, -0.376, -0.3386, -0.3471, -0.3419, -0.3419, -0.3457, -0.3563, -0.3541, -0.378, -0.377, -0.2979, -0.2646, -0.2339, -0.1954, -0.2102, -0.1988, -0.1826, -0.1541, -0.2047, -0.2182, -0.2197, -0.2512, -0.2609, -0.2719, -0.2248, -0.1866, -0.1638, -0.1246, -0.1221, -0.1159, -0.1255, -0.1571, -0.1439, -0.1175, -0.1295, -0.1581, -0.1813, -0.1511, -0.1438, -0.1334, -0.1334, -0.129, -0.1385, -0.081, 0.0186, null, null, null, null]}
{"unit_id": "orig:1994912682458452136", "ticker_idx": 0, "ts": "2025-11-29T23:33:41+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "chipmakers are less interested in selling DRAM and NAND in 4Q assuming further price hike in 2026 and below-trend production volume increase", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA bullish on memory sector: DRAM/NAND price hikes continue into 2026, HBM4 ramp on track at Hynix, Samsung 1c node mostly for HBM4; spot price strength to persist into December.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA Simon Woo) Global Memory Tech\n\nWhat we learned this week: higher ASP but lower volume \nWe checked memory chipmakers’ pricing strategies vs OEMs’ reactions. Key takeaways: (1) conventional DRAM supply cut to OEM customers that do not accept double-digit% price hike (vs 3Q) QTD; (2) newly growing HBM orders with more favorable 2026 contract price (minimal price cut for HBM3e, 20%+ price premium for HBM4); (3) most Tier 1 OEMs still declined 30%+ DRAM contract price hike – broadly accepted 10-20% (vs 20- 30%+ among Tier 2/white-box OEMs); (4) commodity NAND contract price hike not yet as much as for DRAM – just 10-15% (high-end eSSD: still sub-10%) vs 3Q; and (5) minimal bit growth for most memory chipmakers (normal: 10%+/-) coupled with Tier 1 OEMs’ resistance to chip price hike. Overall, chipmakers are less interested in selling DRAM and NAND in 4Q assuming further price hike in 2026 and below-trend production volume increase (conventional DRAM: sub-10%; overall NAND: 15%+/-). That said, most OEMs seem to be asking for only high-single digit% contract price hike for 1Q. We now assume ex-HBM DRAM ASP at +22%/+7% QoQ for 4Q25/1Q26, to be conservative. \n\nFAQs: Google TPU, Samsung, Hynix, China local memory \nSome investors have raised concerns about Google’s more powerful TPU (consequently lower HBM demand from NVIDIA), Samsung Electronics’ more aggressive 1c node capacity expansion or HBM4 ramp-up (negative impact on SK Hynix), Hynix’s HBM4 redesign/ quality issue, and China’s local DRAM/NAND capacity expansion. Net-net, we do not see any negative implications (of Google, Samsung, Hynix, and even China memory) but assume positive impacts: Google’s new TPU v7 – HBM3e contents increased strongly from 64GB HBM3 (v6) to 192GB, even 9,216 units of v7 can be connected (like NVIDA’s NVL 144/192), Google continues using NVIDIA’s GPUs (mainly for high-end training) besides in-house TPUs (inferencing or also training); Samsung – 1c node DRAM ramp-up is mostly for HBM4 (minimal for conventional DRAM); this presents potential shortage of high-end LPDDR5 or GDDR7 (mostly 1b node for Samsung); Hynix – HBM4 actual sales already recorded (4Q) and production ramp-up on track; and China memory capacity – already large accounting for 10%+ of global DRAM and NAND, but actual chip production volume still low due to yield/quality issues (just low-single-digit% of global total, in our view). \n\nDecember spot price outlook \nAlmost all memory chipmakers and OEMs said current DRAM spot prices for 16Gb DDR4 (US$42; up again 10%+ this week, or 13x up YTD) and 16Gb DDR5 (US$27; also up 10% WoW) are abnormally high vs sub-US$10 contract prices on average (even 16Gb-equivalent HBM ASP is still sub-US$30). We also believe current DRAM spot price is based on speculative trading (buying the chip expecting higher selling price for the following month). Since all DRAM makers allude to near-term supply cut expecting higher 2026 contract price, at least spot-price strength will likely continue into Dec.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0114658706649311, "ret_signed_m1d": -0.0114658706649311, "alpha_spy_m1d": -0.016052224880577226, "alpha_c_m1d": -0.009539317334148256, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019644831246580636, "ret_signed_p1d": 0.019644831246580636, "alpha_spy_p1d": 0.01779262197246863, "alpha_c_p1d": 0.0013141702366183568, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.061888469616404373, "ret_signed_p1w": 0.061888469616404373, "alpha_spy_p1w": 0.05694930504065798, "alpha_c_p1w": 0.017719201095217185, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.20758480185732853, "ret_signed_p1m": 0.20758480185732853, "alpha_spy_p1m": 0.19469278646108146, "alpha_c_p1m": 0.17505854863278225, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 0.9828726284367323, "ret_signed_p3m": 0.9828726284367323, "alpha_spy_p3m": 0.9666044615229631, "alpha_c_p3m": 0.8119410423724093, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.5122622366497445, "ret_signed_p6m": 2.5122622366497445, "alpha_spy_p6m": 2.4026089571962377, "alpha_c_p6m": 1.8009815743518285, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.4432, -0.4328, -0.4199, -0.4156, -0.3989, -0.3788, -0.3597, -0.3306, -0.3246, -0.3031, -0.3144, -0.2816, -0.2901, -0.2797, -0.2671, -0.2735, -0.2786, -0.2995, -0.2784, -0.2756, -0.2404, -0.2038, -0.198, -0.1936, -0.2009, -0.1859, -0.1917, -0.1711, -0.1673, -0.183, -0.1582, -0.116, -0.1075, -0.0884, -0.1006, -0.1007, -0.0981, -0.0539, -0.0264, -0.0408, -0.008, 0.0064, 0.0118, 0.0766, 0.0119, 0.0204, 0.0256, 0.0127, 0.059, 0.0597, 0.0624, 0.0473, 0.0103, 0.0433, -0.0069, -0.0198, -0.0347, -0.0765, -0.0479, -0.039, -0.0164, -0.0015, -0.0115, 0.0, 0.0196, 0.0122, -0.0024, 0.0244, 0.0619, 0.0591, 0.0864, 0.0632, 0.0482, 0.019, -0.0094, 0.0108, 0.0424, 0.0591, 0.1082, 0.1135, 0.1291, 0.1291, 0.1528, 0.2019, 0.2076, 0.1976, 0.1976, 0.2839, 0.323, 0.3876, 0.3991, 0.3831, 0.4014, 0.4049, 0.3834, 0.3882, 0.4091, 0.4662, 0.4725, 0.449, 0.4949, 0.5252, 0.5349, 0.501, 0.5946, 0.6644, 0.6718, 0.6719, 0.6212, 0.7002, 0.6457, 0.582, 0.5942, 0.6343, 0.6113, 0.6595, 0.7178, 0.7182, 0.7182, 0.7018, 0.7311, 0.764, 0.8135, 0.8143, 0.867, 0.9019, 0.9829, 0.9499, 0.9503, 0.7577, 0.655, 0.7714, 0.7125, 0.6354, 0.7656, 0.8048, 0.7638, 0.7656, 0.8443, 0.8867, 0.9886, 0.9112, 0.8741, 0.7589, 0.7909, 0.7755, 0.6705, 0.673, 0.5754, 0.5264, 0.6986, 0.6174, 0.6719, 0.717, 0.7467, 0.9064, 0.8836, 0.9069, 0.9066, 1.0183, 1.0407, 1.0755, 1.0508, 1.0601, 1.1125, 1.1597, 1.1763, 1.1786, 1.2773, 1.2458, 1.2744, 1.25, 1.2847, 1.4721, 1.5606, 1.8043, 1.8278, 1.9767, 2.221, 2.1316, 2.2872, 2.2851, 2.0352, 2.0234, 1.9699, 2.0177, 2.2588, 2.2207, 2.2207, 2.5123]}
{"unit_id": "orig:1994912682458452136", "ticker_idx": 1, "ts": "2025-11-29T23:33:41+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hynix – HBM4 actual sales already recorded (4Q) and production ramp-up on track", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA bullish on memory sector: DRAM/NAND price hikes continue into 2026, HBM4 ramp on track at Hynix, Samsung 1c node mostly for HBM4; spot price strength to persist into December.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA Simon Woo) Global Memory Tech\n\nWhat we learned this week: higher ASP but lower volume \nWe checked memory chipmakers’ pricing strategies vs OEMs’ reactions. Key takeaways: (1) conventional DRAM supply cut to OEM customers that do not accept double-digit% price hike (vs 3Q) QTD; (2) newly growing HBM orders with more favorable 2026 contract price (minimal price cut for HBM3e, 20%+ price premium for HBM4); (3) most Tier 1 OEMs still declined 30%+ DRAM contract price hike – broadly accepted 10-20% (vs 20- 30%+ among Tier 2/white-box OEMs); (4) commodity NAND contract price hike not yet as much as for DRAM – just 10-15% (high-end eSSD: still sub-10%) vs 3Q; and (5) minimal bit growth for most memory chipmakers (normal: 10%+/-) coupled with Tier 1 OEMs’ resistance to chip price hike. Overall, chipmakers are less interested in selling DRAM and NAND in 4Q assuming further price hike in 2026 and below-trend production volume increase (conventional DRAM: sub-10%; overall NAND: 15%+/-). That said, most OEMs seem to be asking for only high-single digit% contract price hike for 1Q. We now assume ex-HBM DRAM ASP at +22%/+7% QoQ for 4Q25/1Q26, to be conservative. \n\nFAQs: Google TPU, Samsung, Hynix, China local memory \nSome investors have raised concerns about Google’s more powerful TPU (consequently lower HBM demand from NVIDIA), Samsung Electronics’ more aggressive 1c node capacity expansion or HBM4 ramp-up (negative impact on SK Hynix), Hynix’s HBM4 redesign/ quality issue, and China’s local DRAM/NAND capacity expansion. Net-net, we do not see any negative implications (of Google, Samsung, Hynix, and even China memory) but assume positive impacts: Google’s new TPU v7 – HBM3e contents increased strongly from 64GB HBM3 (v6) to 192GB, even 9,216 units of v7 can be connected (like NVIDA’s NVL 144/192), Google continues using NVIDIA’s GPUs (mainly for high-end training) besides in-house TPUs (inferencing or also training); Samsung – 1c node DRAM ramp-up is mostly for HBM4 (minimal for conventional DRAM); this presents potential shortage of high-end LPDDR5 or GDDR7 (mostly 1b node for Samsung); Hynix – HBM4 actual sales already recorded (4Q) and production ramp-up on track; and China memory capacity – already large accounting for 10%+ of global DRAM and NAND, but actual chip production volume still low due to yield/quality issues (just low-single-digit% of global total, in our view). \n\nDecember spot price outlook \nAlmost all memory chipmakers and OEMs said current DRAM spot prices for 16Gb DDR4 (US$42; up again 10%+ this week, or 13x up YTD) and 16Gb DDR5 (US$27; also up 10% WoW) are abnormally high vs sub-US$10 contract prices on average (even 16Gb-equivalent HBM ASP is still sub-US$30). We also believe current DRAM spot price is based on speculative trading (buying the chip expecting higher selling price for the following month). Since all DRAM makers allude to near-term supply cut expecting higher 2026 contract price, at least spot-price strength will likely continue into Dec.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01486978721660659, "ret_signed_m1d": -0.01486978721660659, "alpha_spy_m1d": -0.019456141432252716, "alpha_c_m1d": -0.012943233885823746, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.037174817256039994, "ret_signed_p1d": 0.037174817256039994, "alpha_spy_p1d": 0.03532260798192799, "alpha_c_p1d": 0.018844156246077715, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.07249076560395262, "ret_signed_p1w": 0.07249076560395262, "alpha_spy_p1w": 0.06755160102820623, "alpha_c_p1w": 0.028321497082765434, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.21003728679871325, "ret_signed_p1m": 0.21003728679871325, "alpha_spy_p1m": 0.19714527140246618, "alpha_c_p1m": 0.17751103357416698, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 1.0465204963316177, "ret_signed_p3m": 1.0465204963316177, "alpha_spy_p3m": 1.0302523294178485, "alpha_c_p3m": 0.8755889102672947, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.821164681653712, "ret_signed_p6m": 2.821164681653712, "alpha_spy_p6m": 2.7115114022002054, "alpha_c_p6m": 2.109884019355796, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.5124, -0.5069, -0.492, -0.4855, -0.4651, -0.4353, -0.4298, -0.3898, -0.3852, -0.3536, -0.3806, -0.3397, -0.3397, -0.3481, -0.3295, -0.336, -0.3378, -0.375, -0.3518, -0.3546, -0.3313, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.205, -0.2292, -0.2357, -0.2152, -0.1595, -0.1354, -0.0982, -0.1103, -0.1057, -0.1112, -0.0527, -0.0063, -0.0323, 0.0364, 0.055, 0.0383, 0.1516, 0.0884, 0.0754, 0.1014, 0.0773, 0.1256, 0.1497, 0.146, 0.1367, 0.0401, 0.1256, 0.0587, 0.0439, 0.0606, -0.0323, -0.0341, -0.036, -0.0267, 0.0112, -0.0149, 0.0, 0.0372, 0.026, 0.0074, 0.0112, 0.0725, 0.052, 0.0911, 0.0502, 0.0613, 0.0297, -0.0149, 0.0242, 0.026, 0.0167, 0.0781, 0.0855, 0.0929, 0.0929, 0.1134, 0.1896, 0.21, 0.21, 0.21, 0.2584, 0.2937, 0.3494, 0.3792, 0.4052, 0.3829, 0.3922, 0.3717, 0.3792, 0.3922, 0.4052, 0.4201, 0.381, 0.3755, 0.4033, 0.4257, 0.368, 0.487, 0.5632, 0.6004, 0.6896, 0.5428, 0.6859, 0.6729, 0.5651, 0.5595, 0.6487, 0.6283, 0.5985, 0.6506, 0.6357, 0.6357, 0.6357, 0.6357, 0.6617, 0.7639, 0.7677, 0.868, 0.8922, 1.0465, 0.9758, 0.9758, 0.7486, 0.581, 0.7523, 0.7206, 0.5568, 0.7467, 0.7784, 0.7318, 0.6946, 0.8137, 0.8063, 0.9664, 0.8864, 0.8752, 0.7374, 0.8361, 0.8529, 0.7374, 0.7374, 0.6257, 0.5028, 0.6704, 0.5456, 0.6313, 0.6499, 0.7057, 0.9236, 0.8584, 0.9124, 0.9367, 1.054, 1.1154, 1.1508, 1.1005, 1.1713, 1.2793, 1.2774, 1.2812, 1.2756, 1.4059, 1.4208, 1.4078, 1.3947, 1.3947, 1.6946, 1.6946, 1.9813, 2.08, 2.1396, 2.5009, 2.4171, 2.6796, 2.6685, 2.3873, 2.4264, 2.2495, 2.2495, 2.6126, 2.6145, 2.6145, 2.8212]}
{"unit_id": "orig:1994912682458452136", "ticker_idx": 2, "ts": "2025-11-29T23:33:41+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung – 1c node DRAM ramp-up is mostly for HBM4 (minimal for conventional DRAM); this presents potential shortage of high-end LPDDR5 or GDDR7", "tweet_type": "analysis", "timeline": "quarters", "summary": "BofA bullish on memory sector: DRAM/NAND price hikes continue into 2026, HBM4 ramp on track at Hynix, Samsung 1c node mostly for HBM4; spot price strength to persist into December.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "BofA Simon Woo) Global Memory Tech\n\nWhat we learned this week: higher ASP but lower volume \nWe checked memory chipmakers’ pricing strategies vs OEMs’ reactions. Key takeaways: (1) conventional DRAM supply cut to OEM customers that do not accept double-digit% price hike (vs 3Q) QTD; (2) newly growing HBM orders with more favorable 2026 contract price (minimal price cut for HBM3e, 20%+ price premium for HBM4); (3) most Tier 1 OEMs still declined 30%+ DRAM contract price hike – broadly accepted 10-20% (vs 20- 30%+ among Tier 2/white-box OEMs); (4) commodity NAND contract price hike not yet as much as for DRAM – just 10-15% (high-end eSSD: still sub-10%) vs 3Q; and (5) minimal bit growth for most memory chipmakers (normal: 10%+/-) coupled with Tier 1 OEMs’ resistance to chip price hike. Overall, chipmakers are less interested in selling DRAM and NAND in 4Q assuming further price hike in 2026 and below-trend production volume increase (conventional DRAM: sub-10%; overall NAND: 15%+/-). That said, most OEMs seem to be asking for only high-single digit% contract price hike for 1Q. We now assume ex-HBM DRAM ASP at +22%/+7% QoQ for 4Q25/1Q26, to be conservative. \n\nFAQs: Google TPU, Samsung, Hynix, China local memory \nSome investors have raised concerns about Google’s more powerful TPU (consequently lower HBM demand from NVIDIA), Samsung Electronics’ more aggressive 1c node capacity expansion or HBM4 ramp-up (negative impact on SK Hynix), Hynix’s HBM4 redesign/ quality issue, and China’s local DRAM/NAND capacity expansion. Net-net, we do not see any negative implications (of Google, Samsung, Hynix, and even China memory) but assume positive impacts: Google’s new TPU v7 – HBM3e contents increased strongly from 64GB HBM3 (v6) to 192GB, even 9,216 units of v7 can be connected (like NVIDA’s NVL 144/192), Google continues using NVIDIA’s GPUs (mainly for high-end training) besides in-house TPUs (inferencing or also training); Samsung – 1c node DRAM ramp-up is mostly for HBM4 (minimal for conventional DRAM); this presents potential shortage of high-end LPDDR5 or GDDR7 (mostly 1b node for Samsung); Hynix – HBM4 actual sales already recorded (4Q) and production ramp-up on track; and China memory capacity – already large accounting for 10%+ of global DRAM and NAND, but actual chip production volume still low due to yield/quality issues (just low-single-digit% of global total, in our view). \n\nDecember spot price outlook \nAlmost all memory chipmakers and OEMs said current DRAM spot prices for 16Gb DDR4 (US$42; up again 10%+ this week, or 13x up YTD) and 16Gb DDR5 (US$27; also up 10% WoW) are abnormally high vs sub-US$10 contract prices on average (even 16Gb-equivalent HBM ASP is still sub-US$30). We also believe current DRAM spot price is based on speculative trading (buying the chip expecting higher selling price for the following month). Since all DRAM makers allude to near-term supply cut expecting higher 2026 contract price, at least spot-price strength will likely continue into Dec.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002976244827510355, "ret_signed_m1d": -0.002976244827510355, "alpha_spy_m1d": -0.007562599043156482, "alpha_c_m1d": -0.0025221594915277867, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.00045408533598256806, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02579357553797723, "ret_signed_p1d": 0.02579357553797723, "alpha_spy_p1d": 0.023941366263865227, "alpha_c_p1d": 0.015491534515764682, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01030204102221255, "ret_p1w": 0.08630946109646254, "ret_signed_p1w": 0.08630946109646254, "alpha_spy_p1w": 0.08137029652071615, "alpha_c_p1w": 0.05519376840312207, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.03111569269334047, "ret_p1m": 0.1952663534648169, "ret_signed_p1m": 0.1952663534648169, "alpha_spy_p1m": 0.18237433806856984, "alpha_c_p1m": 0.1781156283389469, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.017150725125870014, "ret_p3m": 1.1732115801334566, "ret_signed_p3m": 1.1732115801334566, "alpha_spy_p3m": 1.1569434132196874, "alpha_c_p3m": 1.1868391630443735, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": -0.013627582910916813, "ret_p6m": 1.9868586647386945, "ret_signed_p6m": 1.9868586647386945, "alpha_spy_p6m": 1.8772053852851878, "alpha_c_p6m": 1.6901842707486512, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.29667439399004336, "price_path": [-0.3106, -0.3077, -0.3136, -0.3077, -0.2938, -0.283, -0.2751, -0.2553, -0.2444, -0.2158, -0.2277, -0.2069, -0.2069, -0.1753, -0.1635, -0.1565, -0.1496, -0.1773, -0.1647, -0.1677, -0.1468, -0.1096, -0.1096, -0.1096, -0.1096, -0.1096, -0.1096, -0.0635, -0.0744, -0.0913, -0.0575, -0.0308, -0.0288, -0.0268, -0.0327, -0.0218, -0.0427, -0.0198, 0.0119, -0.0129, -0.003, 0.0327, 0.0665, 0.1022, 0.0407, -0.002, -0.0159, -0.0288, -0.002, 0.0268, 0.0228, 0.0198, -0.0357, -0.002, -0.0298, -0.0427, -0.002, -0.0595, -0.0407, -0.0149, 0.0198, 0.0268, -0.003, 0.0, 0.0258, 0.0367, 0.0427, 0.0754, 0.0863, 0.0754, 0.0714, 0.0645, 0.0804, 0.0397, 0.0198, 0.0704, 0.0675, 0.0546, 0.0962, 0.1062, 0.1022, 0.1022, 0.1607, 0.1913, 0.1953, 0.1953, 0.1953, 0.281, 0.3767, 0.3847, 0.4056, 0.3837, 0.3857, 0.3837, 0.3717, 0.3986, 0.4345, 0.4844, 0.4884, 0.4475, 0.4903, 0.5183, 0.5163, 0.5163, 0.59, 0.6189, 0.602, 0.6, 0.4993, 0.6698, 0.6857, 0.588, 0.5811, 0.6588, 0.6528, 0.6728, 0.7804, 0.8064, 0.8064, 0.8064, 0.8064, 0.8941, 0.8951, 0.924, 0.9938, 1.0287, 1.1732, 1.1583, 1.1583, 0.9449, 0.7166, 0.91, 0.8761, 0.7296, 0.8731, 0.8941, 0.8731, 0.8293, 0.8811, 0.933, 1.0785, 0.9988, 0.9878, 0.8572, 0.8911, 0.8841, 0.7954, 0.7954, 0.7611, 0.6702, 0.8945, 0.7821, 0.86, 0.929, 0.9629, 1.1028, 1.0379, 1.0578, 1.0079, 1.0628, 1.1078, 1.1727, 1.1577, 1.1427, 1.1877, 1.1727, 1.2426, 1.1927, 1.2426, 1.2177, 1.2576, 1.2027, 1.2027, 1.3226, 1.3226, 1.6572, 1.7121, 1.6822, 1.852, 1.7871, 1.837, 1.9569, 1.7022, 1.807, 1.7521, 1.7571, 1.9919, 1.9219, 1.9219, 1.9869]}
{"unit_id": "orig:1978719330738245811", "ticker_idx": 0, "ts": "2025-10-16T07:07:05+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI chip demand remains far stronger than expected three months ago, and the overall demand momentum continues. The outlook for 2026 is positive, with AI and high-performance computing demand expected to be the key growth drivers.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "TSMC CEO: AI demand stronger than expected, 2nm mass production Q4, CoWoS capacity expansion 2026, positive 2026 outlook — bullish read for TSMC.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TSMC CEO remarks\n\n• Mass production of the 2nm process is scheduled to begin in the latter part of Q4.\n\n• The A16 node will enter mass production in the second half of next year.\n\n• AI chip demand remains far stronger than expected three months ago, and the overall demand momentum continues.\n\n• Regarding CAPEX, he stated that “it’s unlikely to suddenly decrease in any given year,” reaffirming TSMC’s commitment to maintaining a high level of investment for the coming years.\n\n• The company is making full-scale efforts to expand CoWoS advanced packaging capacity by 2026.\n\n• Current AI-related packaging and production capacity remains extremely tight, and the supply shortage persists.\n\n• The outlook for 2026 is positive, with AI and high-performance computing demand expected to be the key growth drivers.\n\n$TSM $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016241963220170907, "ret_signed_m1d": 0.016241963220170907, "alpha_spy_m1d": 0.009385005022910065, "alpha_c_m1d": 0.020669501165050286, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.004427537944879378, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015875169925897326, "ret_signed_p1d": -0.015875169925897326, "alpha_spy_p1d": -0.021551470284805307, "alpha_c_p1d": -0.014506121547831263, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": -0.0013690483780660623, "ret_p1w": -0.03038284766482957, "ret_signed_p1w": -0.03038284766482957, "alpha_spy_p1w": -0.04721495694949973, "alpha_c_p1w": -0.03475224586226555, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.004369398197435981, "ret_p1m": -0.05009338184747769, "ret_signed_p1m": -0.05009338184747769, "alpha_spy_p1m": -0.06718286283374841, "alpha_c_p1m": -0.05498719699999788, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.004893815152520187, "ret_p3m": 0.10760450308361191, "ret_signed_p3m": 0.10760450308361191, "alpha_spy_p3m": 0.05435309014084422, "alpha_c_p3m": -0.03643368938017488, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.1440381924637868, "ret_p6m": 0.24282145916799536, "ret_signed_p6m": 0.24282145916799536, "alpha_spy_p6m": 0.20847776072363544, "alpha_c_p6m": -0.03372800745778326, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.2765494666257786, "price_path": [-0.2059, -0.22, -0.201, -0.1968, -0.1835, -0.1929, -0.1977, -0.1924, -0.1967, -0.218, -0.2054, -0.2271, -0.2308, -0.1934, -0.196, -0.1951, -0.1878, -0.1973, -0.1988, -0.2058, -0.1974, -0.2264, -0.24, -0.2442, -0.2254, -0.2168, -0.2063, -0.2045, -0.2078, -0.2324, -0.2324, -0.2407, -0.2307, -0.218, -0.1908, -0.1782, -0.1658, -0.1341, -0.1392, -0.1378, -0.131, -0.126, -0.1236, -0.1041, -0.1166, -0.0907, -0.0571, -0.0638, -0.0773, -0.0883, -0.0887, -0.0685, -0.0379, -0.0391, -0.0255, 0.0085, -0.0194, 0.0156, 0.0001, -0.064, 0.0102, -0.013, 0.0162, 0.0, -0.0159, -0.0071, -0.0178, -0.0366, -0.0304, -0.0163, -0.0053, 0.0056, 0.0175, 0.0113, 0.002, 0.0167, -0.0193, -0.0207, -0.0354, -0.0445, -0.0152, -0.0289, -0.0307, -0.0588, -0.0501, -0.0595, -0.0731, -0.0583, -0.0745, -0.0826, -0.0507, -0.0506, -0.033, -0.033, -0.0278, -0.0406, -0.0258, -0.0146, -0.023, -0.0171, 0.0068, 0.0119, 0.0344, 0.0195, -0.0234, -0.0378, -0.0407, -0.0738, -0.048, -0.0337, -0.0192, -0.007, -0.0008, -0.0008, 0.0127, 0.0063, 0.0018, 0.0162, 0.0162, 0.0688, 0.0776, 0.095, 0.0657, 0.0635, 0.0823, 0.1095, 0.1076, 0.0939, 0.1425, 0.145, 0.145, 0.0941, 0.0906, 0.0948, 0.1198, 0.1126, 0.1314, 0.1447, 0.1355, 0.1054, 0.1415, 0.1228, 0.0893, 0.106, 0.1666, 0.1885, 0.2103, 0.251, 0.231, 0.2252, 0.2252, 0.2179, 0.2114, 0.2052, 0.2391, 0.2375, 0.29, 0.2966, 0.2601, 0.2526, 0.2343, 0.1809, 0.1953, 0.1833, 0.1333, 0.1661, 0.1607, 0.1857, 0.126, 0.1313, 0.1378, 0.1603, 0.1388, 0.1361, 0.1041, 0.135, 0.1511, 0.1662, 0.0936, 0.0957, 0.0614, 0.1333, 0.1452, 0.137, 0.137, 0.1461, 0.158, 0.2271, 0.2257, 0.2428]}
{"unit_id": "orig:1974260354009215219", "ticker_idx": 0, "ts": "2025-10-03T23:48:43+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The memory supercycle is coming…", "tweet_type": "prediction", "timeline": "quarters", "summary": "Memory supercycle imminent; bullish on the sector based on tightening fab discipline analogous to 2017 peak cycle.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Let’s look back at the peak of the memory cycle in 2017.\n\nAt that time, engineers at Samsung and SK Hynix had to write formal reports to their superiors if they made even a single mistake on a wafer that resulted in a defect or scrap. Every single chip immediately translated into money once it reached the market.\n\nToday, things aren’t quite as strict about each wafer as they were back then. But, as some say, “it might soon return to that state.”\n\nThe memory supercycle is coming…", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00744467243042916, "ret_signed_m1d": -0.00744467243042916, "alpha_spy_m1d": -0.0074594400107291054, "alpha_c_m1d": -0.012078196669325563, "bench_spy_m1d": 1.4767580299945848e-05, "bench_c_m1d": 0.004633524238896403, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005554675211845532, "ret_signed_p1d": 0.005554675211845532, "alpha_spy_p1d": 0.001968446301509423, "alpha_c_p1d": -0.014345565811428724, "bench_spy_p1d": 0.003586228910336109, "bench_c_p1d": 0.019900241023274257, "ret_p1w": 0.03360555004459109, "ret_signed_p1w": 0.03360555004459109, "alpha_spy_p1w": 0.05779833209161622, "alpha_c_p1w": 0.06728734331303321, "bench_spy_p1w": -0.02419278204702513, "bench_c_p1w": -0.03368179326844212, "ret_p1m": 0.35168436644079887, "ret_signed_p1m": 0.35168436644079887, "alpha_spy_p1m": 0.3305699369555361, "alpha_c_p1m": 0.2643016093338492, "bench_spy_p1m": 0.021114429485262765, "bench_c_p1m": 0.08738275710694965, "ret_p3m": 0.5032500104394616, "ret_signed_p3m": 0.5032500104394616, "alpha_spy_p3m": 0.48124649924086615, "alpha_c_p3m": 0.43027091258008443, "bench_spy_p3m": 0.02200351119859545, "bench_c_p3m": 0.07297909785937717, "ret_p6m": 0.968544223994483, "ret_signed_p6m": 0.968544223994483, "alpha_spy_p6m": 1.0188149018172465, "alpha_c_p6m": 0.8884145325966535, "bench_spy_p6m": -0.0502706778227634, "bench_c_p6m": 0.08012969139782955, "price_path": [-0.315, -0.3234, -0.3062, -0.2998, -0.3061, -0.3002, -0.3052, -0.3257, -0.3226, -0.3192, -0.3363, -0.3333, -0.331, -0.3351, -0.3219, -0.3195, -0.3063, -0.3123, -0.3421, -0.334, -0.3264, -0.3351, -0.3204, -0.3078, -0.2933, -0.2841, -0.2798, -0.2803, -0.2882, -0.2969, -0.3034, -0.3164, -0.3274, -0.316, -0.3108, -0.3134, -0.3113, -0.3, -0.3045, -0.3232, -0.3148, -0.3101, -0.2967, -0.2795, -0.2742, -0.253, -0.227, -0.2028, -0.1655, -0.1584, -0.1314, -0.146, -0.1039, -0.1149, -0.1035, -0.0874, -0.0961, -0.103, -0.1295, -0.1025, -0.0988, -0.0541, -0.0074, 0.0, 0.0056, -0.0038, 0.0155, 0.008, 0.0336, 0.0384, 0.019, 0.0495, 0.1037, 0.1151, 0.1405, 0.1248, 0.1242, 0.1285, 0.1855, 0.2203, 0.2025, 0.2457, 0.2629, 0.2675, 0.3517, 0.2704, 0.2831, 0.2912, 0.2747, 0.3339, 0.334, 0.3375, 0.3181, 0.271, 0.3138, 0.2491, 0.233, 0.2122, 0.1592, 0.1948, 0.2047, 0.2321, 0.2519, 0.2399, 0.2549, 0.2797, 0.2692, 0.2497, 0.2824, 0.3315, 0.3279, 0.3642, 0.3337, 0.314, 0.278, 0.2414, 0.2657, 0.3063, 0.3281, 0.3904, 0.397, 0.4173, 0.4173, 0.4452, 0.5084, 0.5161, 0.5033, 0.5033, 0.6106, 0.6566, 0.7404, 0.7549, 0.7362, 0.7589, 0.7638, 0.7363, 0.7412, 0.7664, 0.8373, 0.8456, 0.8166, 0.8729, 0.911, 0.9241, 0.8792, 0.9983, 1.0882, 1.0996, 1.1022, 1.0386, 1.1348, 1.0637, 0.9844, 1.0003, 1.05, 1.0203, 1.0801, 1.1505, 1.1493, 1.1493, 1.1282, 1.1658, 1.2041, 1.27, 1.2698, 1.3362, 1.3797, 1.48, 1.4366, 1.4372, 1.1956, 1.0716, 1.2151, 1.1406, 1.0451, 1.2095, 1.2594, 1.2067, 1.2103, 1.3116, 1.3629, 1.4902, 1.393, 1.3459, 1.2015, 1.2433, 1.2245, 1.0916, 1.0947, 0.9685]}
{"unit_id": "reply:1999771640876732803", "ticker_idx": 0, "ts": "2025-12-13T09:21:27+00:00", "symbol": "BABA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "저는 중국 테크 하나만 보유하라면 알리바바 ADR 선호합니다", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Alibaba ADR as top China tech pick: strongest domestic AI model (Qwen), multiple AI-monetizable platforms, in-house chip biz, growing Western adoption of Qwen.", "resolved_tickers": ["BABA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "1. 중국내에서 가장 강력한 AI모델 보유 (저는 이게 일시적인게 아니라 Qwen이 앞으로도 꽤 발전할 것 같다는 생각이 들어요.)\n\n2. AI로 수익을 내기 좋은 플랫폼 다수 보유 (배달, 이커머스 등)\n\n3. 자체 칩 비즈니스 보유해 추론용 칩 대응 가능\n\n4. 점점 더 서방 소프트웨어 회사에서 알리바바 Qwen 모델 채택 늘어남.\n\n등등...저는 중국 테크 하나만 보유하라면 알리바바 ADR 선호합니다.", "reply_text": null, "parent_text": "@jukan05 알리바바... 저번에 QnA에서 홍콩증시 기업 중 제일 좋게 본다고 하셨던 걸로 기억하는데...\n\n매력 포인트 몇 개 알려주실 수 있나요? 경청할게요! https://t.co/3F8EgHrAOY", "quoted_text": null, "parent_author": "StarAndCompass_", "ret_m1d": 0.03724429658393835, "ret_signed_m1d": 0.03724429658393835, "alpha_spy_m1d": 0.03573121273691093, "alpha_c_m1d": 0.045705769506583116, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": -0.008461472922644764, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005330155715087104, "ret_signed_p1d": -0.005330155715087104, "alpha_spy_p1d": -0.002597858119808394, "alpha_c_p1d": -0.0069730666948470965, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": 0.0016429109797599928, "ret_p1w": 0.005796591359879422, "ret_signed_p1w": 0.005796591359879422, "alpha_spy_p1w": -0.003199133952823008, "alpha_c_p1w": -0.0012060893744865187, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.00700268073436594, "ret_p1m": 0.11273235113454261, "ret_signed_p1m": 0.11273235113454261, "alpha_spy_p1m": 0.09056486122883167, "alpha_c_p1m": 0.09033731426233205, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.02239503687221056, "ret_p3m": -0.10586980996305728, "ret_signed_p3m": -0.10586980996305728, "alpha_spy_p3m": -0.08721063669836515, "alpha_c_p3m": -0.023811160514035645, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.08205864944902164, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1071, 0.0826, 0.0847, 0.0943, 0.0865, 0.1756, 0.1691, 0.1454, 0.1986, 0.1908, 0.2178, 0.2615, 0.2528, 0.2474, 0.2081, 0.2067, 0.1572, 0.0594, 0.1114, 0.0851, 0.1054, 0.0999, 0.113, 0.1558, 0.1105, 0.1051, 0.1453, 0.164, 0.1956, 0.1774, 0.1991, 0.1588, 0.1355, 0.1173, 0.0947, 0.0981, 0.1167, 0.1083, 0.1053, 0.0714, 0.0521, 0.065, 0.0247, 0.0508, 0.0642, 0.0586, 0.0213, 0.0189, 0.0709, 0.0461, 0.05, 0.05, 0.048, 0.0944, 0.0736, 0.0532, 0.049, 0.0548, 0.0536, 0.0391, 0.0582, 0.0454, 0.0372, 0.0, -0.0053, -0.02, -0.0185, -0.002, 0.0058, 0.0076, -0.0002, -0.0002, 0.0143, -0.0107, -0.0182, -0.0234, -0.0234, 0.0376, 0.0411, 0.0054, -0.0223, 0.0292, 0.0058, 0.1081, 0.1127, 0.132, 0.1389, 0.102, 0.102, 0.082, 0.1238, 0.1805, 0.1542, 0.1418, 0.1508, 0.1704, 0.161, 0.1297, 0.1219, 0.0903, 0.0603, 0.0511, 0.0828, 0.086, 0.1094, 0.0948, 0.0576, 0.0376, 0.0376, 0.0356, 0.0378, 0.0279, 0.029, 0.0179, 0.0201, 0.0146, -0.0136, -0.0398, -0.0502, -0.0966, -0.1121, -0.1315, -0.1286, -0.1163, -0.0882, -0.0919, -0.1059, -0.0991, -0.0891, -0.0901, -0.1043, -0.1678, -0.1844, -0.1601, -0.164, -0.1347, -0.1644, -0.1826, -0.1873, -0.1641, -0.1756, -0.1868, -0.1868, -0.1851, -0.2023, -0.165, -0.1493, -0.1516, -0.1471, -0.1249, -0.112, -0.0766, -0.0605, -0.0661, -0.098, -0.0911, -0.1225, -0.0951, -0.1171, -0.1282, -0.131, -0.1213, -0.1239, -0.1121, -0.1188, -0.0576, -0.0606, -0.0668, -0.0852, -0.102, -0.0285, -0.0598, -0.1166, -0.1121, -0.0963, -0.1041, -0.1241, -0.1339, -0.1339, -0.1374, -0.1488, -0.1594, -0.1724, null, null, null, null, null, null, null]}
{"unit_id": "orig:1982652729601479024", "ticker_idx": 0, "ts": "2025-10-27T03:37:01+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bank of America raises its target price for SK Hynix to 700,000 KRW", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "BofA raises SK Hynix PT to 700,000 KRW; author shares/adopts the bullish call.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bank of America raises its target price for SK Hynix to 700,000 KRW.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.046728937477062193, "ret_signed_m1d": -0.046728937477062193, "alpha_spy_m1d": -0.035068788533298756, "alpha_c_m1d": -0.022280497241673802, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026168160473419966, "ret_signed_p1d": -0.026168160473419966, "alpha_spy_p1d": -0.028824159430011043, "alpha_c_p1d": -0.034975154358737215, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.1588786568472491, "ret_signed_p1w": 0.1588786568472491, "alpha_spy_p1w": 0.1616513764325227, "alpha_c_p1w": 0.14176467213837873, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": -0.02990649421421021, "ret_signed_p1m": -0.02990649421421021, "alpha_spy_p1m": -0.014991962737800901, "alpha_c_p1m": 0.025463805722989563, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.4122257675720018, "ret_signed_p3m": 0.4122257675720018, "alpha_spy_p3m": 0.4037968003605552, "alpha_c_p3m": 0.28961249761491414, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 1.2937138055014525, "ret_signed_p6m": 1.2937138055014525, "alpha_spy_p6m": 1.2603692390393113, "alpha_c_p6m": 0.9987595414508477, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.5082, -0.4895, -0.5185, -0.5185, -0.5082, -0.5175, -0.511, -0.5213, -0.5017, -0.4979, -0.4811, -0.4839, -0.4839, -0.5007, -0.5091, -0.5231, -0.5427, -0.5315, -0.5157, -0.5119, -0.5147, -0.4981, -0.4972, -0.5215, -0.5131, -0.5093, -0.5037, -0.4888, -0.4822, -0.4617, -0.4318, -0.4262, -0.386, -0.3813, -0.3495, -0.3766, -0.3355, -0.3355, -0.3439, -0.3252, -0.3318, -0.3336, -0.371, -0.3477, -0.3505, -0.3271, -0.2607, -0.2607, -0.2607, -0.2607, -0.2607, -0.2607, -0.2, -0.2243, -0.2308, -0.2103, -0.1542, -0.1299, -0.0925, -0.1047, -0.1, -0.1056, -0.0467, 0.0, -0.0262, 0.043, 0.0617, 0.0449, 0.1589, 0.0953, 0.0822, 0.1084, 0.0841, 0.1327, 0.157, 0.1533, 0.1439, 0.0467, 0.1327, 0.0654, 0.0505, 0.0673, -0.0262, -0.028, -0.0299, -0.0206, 0.0176, -0.0086, 0.0063, 0.0437, 0.0325, 0.0138, 0.0176, 0.0793, 0.0587, 0.098, 0.0568, 0.0681, 0.0363, -0.0086, 0.0306, 0.0325, 0.0232, 0.0849, 0.0924, 0.0999, 0.0999, 0.1204, 0.1971, 0.2177, 0.2177, 0.2177, 0.2663, 0.3019, 0.358, 0.3879, 0.4141, 0.3917, 0.401, 0.3804, 0.3879, 0.401, 0.4141, 0.4291, 0.3898, 0.3842, 0.4122, 0.4347, 0.3767, 0.4964, 0.5731, 0.6105, 0.7003, 0.5525, 0.6965, 0.6834, 0.575, 0.5693, 0.6591, 0.6386, 0.6086, 0.661, 0.646, 0.646, 0.646, 0.646, 0.6722, 0.7751, 0.7788, 0.8799, 0.9042, 1.0595, 0.9883, 0.9883, 0.7596, 0.591, 0.7634, 0.7315, 0.5666, 0.7578, 0.7896, 0.7428, 0.7053, 0.8252, 0.8177, 0.9789, 0.8983, 0.8871, 0.7484, 0.8477, 0.8646, 0.7484, 0.7484, 0.636, 0.5123, 0.6809, 0.5554, 0.6416, 0.6603, 0.7165, 0.9358, 0.8702, 0.9245, 0.9489, 1.067, 1.1288, 1.1644, 1.1138, 1.185, 1.2937]}
{"unit_id": "quote:1976084995241558394", "ticker_idx": 0, "ts": "2025-10-09T00:39:11+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if customer demand proves accurate, TSMC's CoWoS capacity will be in short supply next year... capacity could expand further to 110–120k wafers per month", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish CoWoS supply-demand analysis: TSMC capacity tight into 2026, NVIDIA demand exceeds even expanded supply.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report:\n\nA major U.S. investment bank believes that if customer demand proves accurate, TSMC’s CoWoS capacity will be in short supply next year.\nLast month, the bank already raised its estimate of 2026 capacity to 100k wafers per month. Observations show that the cleanroom space for Phase 1 of the P8A fab is already ready; if demand is confirmed, capacity could expand further to 110–120k wafers per month.\n\nRegarding NVIDIA, even if TSMC allocates 50% of its CoWoS-L capacity to them next year, it would still fall short of NVIDIA’s total demand by about 20%.\nThe Rubin CPK (single-chip packaging with silicon capacitor, excluding HBM) will use CoWoS-S; Rubin CPX will also use CoWoS-S and will replace the B40.\nOverall, NVIDIA’s CoWoS demand for next year is estimated at 590k CoWoS-L and 15k CoWoS-S units.\n\nFor other ASICs, it is reaffirmed that AWS T3 will see mass shipments in 3Q next year.\nAs for Meta’s M1A ASIC, the probability of MediaTek being the supplier has dropped to 25%, due to the higher technical difficulty of the new packaging design.", "reply_text": null, "parent_text": null, "quoted_text": "早安！10/9 外電綜合整理\n\n- AI供應鏈追蹤報導：\n美系大行認為如果客戶需求屬實，明年台積電CoWoS產能將呈現供不應求狀況。\n上個月大行已上修明年產能至100k/月，現觀察P8A 廠第一期clean room 空間已準備好，若需求確定，將可擴增至 110～120k/月。\n\n關於Nvidia，就算明年台積電提供 5成的CoWoS-L", "parent_author": null, "ret_m1d": -0.017361126814406158, "ret_signed_m1d": -0.017361126814406158, "alpha_spy_m1d": -0.02026655223543361, "alpha_c_m1d": -0.019852447010150542, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.02702782789908964, "alpha_c_p1d": 0.05755912695428611, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": 0.031249960001040966, "ret_signed_p1w": 0.031249960001040966, "alpha_spy_p1w": 0.046924239934682976, "alpha_c_p1w": 0.03678284748495475, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": 0.013888833186634919, "ret_signed_p1m": 0.013888833186634919, "alpha_spy_p1m": 0.01417198595368041, "alpha_c_p1m": 0.005459100937490913, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": 0.18797445698479964, "ret_signed_p3m": 0.18797445698479964, "alpha_spy_p3m": 0.15416102301394874, "alpha_c_p3m": 0.061627792626126165, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": 0.2652487764585114, "ret_signed_p6m": 0.2652487764585114, "alpha_spy_p6m": 0.28252585104411465, "alpha_c_p6m": 0.12524481077761318, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.2426, -0.2322, -0.2184, -0.2184, -0.2011, -0.2046, -0.2184, -0.208, -0.208, -0.208, -0.208, -0.2149, -0.2011, -0.1977, -0.1977, -0.2149, -0.2046, -0.2219, -0.1838, -0.1873, -0.1838, -0.1838, -0.17, -0.1873, -0.1838, -0.1838, -0.1804, -0.2149, -0.2046, -0.2149, -0.1907, -0.1873, -0.1769, -0.1977, -0.1977, -0.1942, -0.1977, -0.1977, -0.1977, -0.1838, -0.1838, -0.17, -0.1527, -0.1423, -0.1285, -0.1319, -0.1111, -0.1215, -0.1076, -0.1215, -0.1007, -0.0694, -0.0694, -0.0833, -0.0972, -0.0972, -0.0938, -0.0799, -0.0521, -0.0278, -0.0278, -0.0035, -0.0174, 0.0, 0.0, -0.0174, -0.0104, 0.0174, 0.0312, 0.0069, 0.0278, 0.0278, 0.0139, 0.0069, 0.0069, 0.0278, 0.0243, 0.0451, 0.0451, 0.0417, 0.0486, 0.0451, 0.0139, 0.0174, 0.0139, 0.0243, 0.0174, 0.0243, 0.0139, -0.0069, 0.0035, -0.0243, -0.0313, 0.0104, -0.0382, -0.0451, -0.0174, 0.0, -0.0035, 0.0, -0.0208, -0.0069, 0.0069, 0.0035, 0.0139, 0.0382, 0.0278, 0.0451, 0.0242, 0.0312, 0.0103, -0.0002, -0.0036, -0.0036, -0.0036, 0.0208, 0.0382, 0.0417, 0.0417, 0.0521, 0.066, 0.0591, 0.08, 0.08, 0.1044, 0.1636, 0.188, 0.1671, 0.174, 0.1706, 0.1775, 0.1915, 0.1915, 0.1775, 0.2124, 0.2263, 0.2367, 0.2124, 0.2263, 0.2333, 0.2228, 0.2402, 0.2681, 0.2577, 0.2367, 0.2298, 0.2542, 0.2437, 0.2298, 0.2402, 0.2646, 0.3099, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3238, 0.3691, 0.404, 0.39, 0.39, 0.3761, 0.3482, 0.2995, 0.3238, 0.3169, 0.2611, 0.289, 0.3517, 0.3134, 0.2995, 0.2855, 0.3072, 0.3317, 0.2932, 0.2862, 0.2652, 0.2652, 0.2897, 0.2862, 0.2722, 0.2443, 0.2303, 0.2967, 0.2652, 0.2652]}
{"unit_id": "quote:1976084995241558394", "ticker_idx": 1, "ts": "2025-10-09T00:39:11+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's CoWoS demand for next year is estimated at 590k CoWoS-L and 15k CoWoS-S units", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish CoWoS supply-demand analysis: TSMC capacity tight into 2026, NVIDIA demand exceeds even expanded supply.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "AI Supply Chain Tracking Report:\n\nA major U.S. investment bank believes that if customer demand proves accurate, TSMC’s CoWoS capacity will be in short supply next year.\nLast month, the bank already raised its estimate of 2026 capacity to 100k wafers per month. Observations show that the cleanroom space for Phase 1 of the P8A fab is already ready; if demand is confirmed, capacity could expand further to 110–120k wafers per month.\n\nRegarding NVIDIA, even if TSMC allocates 50% of its CoWoS-L capacity to them next year, it would still fall short of NVIDIA’s total demand by about 20%.\nThe Rubin CPK (single-chip packaging with silicon capacitor, excluding HBM) will use CoWoS-S; Rubin CPX will also use CoWoS-S and will replace the B40.\nOverall, NVIDIA’s CoWoS demand for next year is estimated at 590k CoWoS-L and 15k CoWoS-S units.\n\nFor other ASICs, it is reaffirmed that AWS T3 will see mass shipments in 3Q next year.\nAs for Meta’s M1A ASIC, the probability of MediaTek being the supplier has dropped to 25%, due to the higher technical difficulty of the new packaging design.", "reply_text": null, "parent_text": null, "quoted_text": "早安！10/9 外電綜合整理\n\n- AI供應鏈追蹤報導：\n美系大行認為如果客戶需求屬實，明年台積電CoWoS產能將呈現供不應求狀況。\n上個月大行已上修明年產能至100k/月，現觀察P8A 廠第一期clean room 空間已準備好，若需求確定，將可擴增至 110～120k/月。\n\n關於Nvidia，就算明年台積電提供 5成的CoWoS-L", "parent_author": null, "ret_m1d": -0.017967518834077323, "ret_signed_m1d": -0.017967518834077323, "alpha_spy_m1d": -0.020872944255104775, "alpha_c_m1d": -0.020458839029821707, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04886534322137259, "ret_signed_p1d": -0.04886534322137259, "alpha_spy_p1d": -0.021837515322282952, "alpha_c_p1d": 0.008693783732913518, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": -0.05587578994716014, "ret_signed_p1w": -0.05587578994716014, "alpha_spy_p1w": -0.04020151001351813, "alpha_c_p1w": -0.05034290246324635, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": -0.02295266728066714, "ret_signed_p1m": -0.02295266728066714, "alpha_spy_p1m": -0.022669514513621647, "alpha_c_p1m": -0.031382399529811145, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": -0.027624079534145696, "ret_signed_p3m": -0.027624079534145696, "alpha_spy_p3m": -0.0614375135049966, "alpha_c_p3m": -0.15397074389281917, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": -0.07872733891956518, "ret_signed_p6m": -0.07872733891956518, "alpha_spy_p6m": -0.061450264333961946, "alpha_c_p6m": -0.21873130460046342, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.148, -0.1136, -0.1101, -0.1017, -0.1047, -0.1101, -0.1327, -0.1132, -0.0978, -0.0991, -0.0822, -0.0886, -0.0691, -0.0764, -0.0979, -0.0653, -0.0744, -0.0683, -0.0613, -0.0513, -0.0546, -0.0489, -0.0571, -0.0548, -0.063, -0.0549, -0.088, -0.0892, -0.0914, -0.0758, -0.0663, -0.0561, -0.057, -0.0644, -0.0955, -0.0955, -0.1132, -0.114, -0.1086, -0.1327, -0.126, -0.1133, -0.0792, -0.08, -0.0766, -0.077, -0.0919, -0.1157, -0.0848, -0.0826, -0.0465, -0.0734, -0.081, -0.0773, -0.0747, -0.0557, -0.0311, -0.0277, -0.0191, -0.0257, -0.0365, -0.0391, -0.018, 0.0, -0.0489, -0.0221, -0.0651, -0.0662, -0.0559, -0.0486, -0.0516, -0.0593, -0.0638, -0.0541, -0.0328, -0.0056, 0.0439, 0.0751, 0.0536, 0.0515, 0.0743, 0.0318, 0.0137, -0.0233, -0.023, 0.0337, 0.0031, 0.0064, -0.0297, -0.0125, -0.031, -0.0582, -0.0314, -0.062, -0.0711, -0.052, -0.0766, -0.0639, -0.0639, -0.0809, -0.0657, -0.0577, -0.0674, -0.0477, -0.0527, -0.0364, -0.0394, -0.0456, -0.0604, -0.0911, -0.0845, -0.0771, -0.1123, -0.0957, -0.0601, -0.0461, -0.0174, -0.0205, -0.0205, -0.0105, -0.0225, -0.0261, -0.0315, -0.0315, -0.0193, -0.0231, -0.0276, -0.0179, -0.039, -0.04, -0.0396, -0.0351, -0.0489, -0.0286, -0.0329, -0.0329, -0.0752, -0.048, -0.0401, -0.0254, -0.0316, -0.021, -0.0054, -0.0003, -0.0074, -0.0361, -0.0635, -0.0954, -0.1074, -0.0371, -0.0131, -0.0209, -0.013, -0.0292, -0.0506, -0.0506, -0.0394, -0.0238, -0.0242, -0.0142, -0.0052, 0.0015, 0.0156, -0.0398, -0.0798, -0.0523, -0.065, -0.0494, -0.0479, -0.0765, -0.0515, -0.0405, -0.0339, -0.0489, -0.0639, -0.0484, -0.0551, -0.0631, -0.0727, -0.1031, -0.0878, -0.0901, -0.072, -0.1107, -0.13, -0.1422, -0.0943, -0.0872, -0.0787, -0.0787]}
{"unit_id": "quote:1995319574393442673", "ticker_idx": 0, "ts": "2025-12-01T02:30:32+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "As long as Nvidia continues its stride forward, it is actually not easy for competitors to keep up. supply consistently falling short of demand, there is no reason for margins to come down.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish semis through 2030 on AI-driven profit reallocation; NVDA structurally dominant, SK Hynix HBM winner, AMD CPU resurgence from agentic AI, Apple beneficiary of memory cost advantage.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Looking back at the semiconductor market in 2025, there were truly too many fascinating stories. The overarching theme was:\n\nAI demand-driven valuation restructuring of semiconductor infrastructure + The rewriting of value distribution across the supply chain.\n\nStarting in 2024, semiconductor infrastructure began rapidly devouring the profits of the entire IT industry. The share of semiconductor net profit (EPS) within the IT sector of the S&P 500 rose from less than 20% to 40% in just two years, and it is still in a state of accelerated ascent.\n\nThe overall forward-looking profit margin for semiconductors has risen from 25% in 2023 to 43% by November 2025, significantly surpassing the average profit margins of several internet giants. This confirms that semiconductor margins exceeding those of the internet sector will be the \"New Normal.\" The profit distribution of the entire IT industry is flowing increasingly toward semiconductors.\n\nIt is worth noting that even during the semiconductor chip shortage of 2020–2022, when shortages were severe, semiconductor profit margins and their share of total IT profit distribution did not see such significant growth.\n\nThis is the first half of the story: The valuation system of semiconductor infrastructure is being reconstructed by AI demand; it no longer holds the subordinate infrastructure status it did during the Internet era.\n\nThe logic behind this phenomenon is the transformation of business models alongside technological characteristics:\n\nIn the Internet era, the network and compute costs for each request had extremely low marginal costs. Scaling effects were excellent, and the marginal cost of distribution was almost zero.\n\nIn the AI era, this \"near-zero marginal distribution cost\" characteristic favorable to scalability has encountered a fundamental and major challenge: Not only is training cost no longer a one-time expense but rather one that grows annually, but even for client-side AI inference requests, costs do not drop in sync with hardware compute deflation. This is because \"inference scaling\" has become a consensus, and vertical domains still require larger-scale flagship models to maintain competitiveness.\n\nPreviously, the biggest cost for internet companies was OPEX, specifically SDE labor costs. Now, for the first time in history, internet companies are burdened with balance sheets full of high depreciation costs, similar to semiconductor foundries. The business model is practically itching to shift slowly from \"Traffic × Conversion Rate\" to \"Gross Profit per Token.\"\n\nSimply put, the cost distribution shift from the Internet era to the AI era has added heavy hardware/compute CAPEX (MSFT 33%, Meta 38% in earnings reports) on top of the OPEX base. The winners of the previous era were Internet companies + CSPs + SaaS (the \"rent-collecting\" industries). In the AI era, compute (semiconductor/chip depreciation) has become the new rent-collecting industry. The profit distribution of the entire IT industry has undergone a drastic reallocation (EPS flow to semiconductors rising from 20% to 40% and climbing). This is the most important reason for the restructuring of the semiconductor infrastructure valuation system.\n\n--------------- How long can the \"New Normal\" of high semiconductor margins last?\n\nThe current high premiums come from an excessive backlog of semiconductor orders caused by an earlier \"cost-is-no-object\" arms race. However, obviously, hyperscalers do not want to be taken for a ride; they are all attempting to build in-house ASICs to lower costs. We can look at this problem from the perspective of the long-term compute distribution in 2030.\n\nIn the long run, OpenAI has already revealed the \"standard answer\": 10GW Nvidia, 10GW ASIC, 6GW AMD. Other hyperscalers have similar considerations for their split.\n\nFor example, on the inference side, they hope for ASIC >50%. Within the GPU segment, a 50/50 split between AMD and NV (legacy). For training, NV will still hold the lion's share (60%+), with the rest split between in-house ASICs and AMD.\n\nCalculated based on a 2030 split of 60% inference and 40% training, the market share comes out to NV 38%, ASIC 39%, and AMD 23%. This is almost entirely consistent with OpenAI’s ratios and can be considered a standard reference value.\n\nOf course, for Microsoft, Amazon, Google, and Anthropic, the AMD ratio will be noticeably lower than this standard answer/reference value, while xAI has no ASICs and uses only Nvidia + a small amount of AMD.\n\nAMD's risk lies in the ultra-long term (post-2030). As CSP in-house ASICs become more mature (except for Microsoft), the proportion of ASICs in inference may grow higher, making it hard to find incentives to buy large quantities of new GPUs unless they are sold cheaply enough.\n\nWhat about the recently hyped TPU? Is Meta shifting to TPUs? Will it significantly impact Nvidia’s margins?\n\nActually, Meta’s CAPEX is $72B this year, $110B next year, and the average annual CAPEX over the next six years may reach around $160B. Meanwhile, Meta’s 6-year $10B TPU order amounts to only $1.6B per year on average—and they are purchasing TPU cloud services, not bare TPUs.\n\nIn other words, this TPU order accounts for only 1% of Meta’s future 6-year CAPEX. They are not seriously considering large-scale deployment; it is likely just a means to bargain with Nvidia.\n\nFurthermore, looking at Meta’s recruitment ads over the recent months, we don’t see any hiring for TPU engineers, unlike Anthropic, which was hiring a bunch of TPU kernel engineers back in May before announcing large-scale TPU procurement for training in October.\n\nSo, whether the reason is diversifying suppliers, creating a fallback for delayed in-house ASICs, or delays with AMD’s MI350X, Meta’s purchase of TPUs basically has one consideration: increasing bargaining power when buying Nvidia GPUs. However, they can essentially only bargain on the inference share; the actual effect is very limited, as is the impact on Nvidia’s profit margins.\n\nKeep in mind that during the crypto bear market crash of 2022, NVDA inventory rose to 198 days, yet margins only retreated from 65% to 56%. Even with the \"double kill\" of PE and macroeconomics, the stock price only went from 300 to 100. Now, with supply consistently falling short of demand, there is no reason for margins to come down.\n\nAdditionally, the TPU v8 design is too conservative (it didn't use HBM4), and the TCO (Total Cost of Ownership) of the Kyber rack Rubin solution will be better than TPU v8. In the end, they will still have to rely on Nvidia, making it hard to bargain. As long as Nvidia continues its stride forward, it is actually not easy for competitors to keep up.\n\nIn summary, on one hand, bottlenecks across the entire supply chain (e.g., CoWoS expansion) are cautious, so the state of demand exceeding supply can last for many years.\n\nOn the other hand, there is a \"temporal mismatch\" between the profit curve of AI monetization and the hardware investment curve. The growth curve of the application side will lag by a few years. As long as this time lag between the application side and infrastructure side growth curves exists, semiconductors will continue to dominate profit distribution in the IT industry.\n\nLooking at OpenAI’s investment curve through 2030, this mismatch will last until around 2030. This means the dominant position in IT industry profit distribution brought about by the semiconductor industry’s super-expansion phase will last at least until 2030 based on current views.\n\nAnd high semiconductor margins might be sustained even longer, because the attribute has changed from \"one-time infrastructure\" in the internet era to \"rent-collecting infrastructure\" now.\n\nAI has not only fed the GPU; it has used the compute budget to lift every link in the chain that \"turns electricity into tokens\"—from memory, storage, interconnects, and fiber optics, to electricity, energy storage, and so on.\n\nThe first half covered \"Semiconductors swallowing IT profits.\" The second half covers how the \"AI Compute Value Spillover Effect\" reshapes the internal landscape of semiconductors: GPU Compute Growth -> Memory/Storage/Interconnect/CPU Bottlenecks -> Spillover Effect -> Structural Opportunities.\n\nThe more interesting story in 2025 is how huge industry dividends are creating structural new opportunities within the semiconductor sector. For instance, a super cluster requires several data center interconnects; fiber optic interconnect lengths reach the level of millions of miles. This is a new opportunity.\n\nThe most typical examples of new opportunities brought by structural trends in the semiconductor supply chain are Memory (DRAM/HBM) and Storage (SSD). The growth in HBM demand is so exaggerated that it is squeezing DDR4/5 capacity, leading the memory industry—historically known for its cyclical nature—to shout that \"the cycle no longer exists.\" SK Hynix, leading in HBM, is even envisioning annual profits of $100 billion a few years from now, firmly becoming a trillion-dollar market cap company.\n\nBehind these two sectors is a shift in structural trends: AI workloads are gradually extending from training to inference, and the proportion of inference is getting larger.\n\nInference is a purely memory-bound task; one could say bandwidth speed = tokens/s. As model sizes get larger and context lengths increase, the requirements for memory size also grow accordingly, leading to a surge in memory demand: Inference is Memory.\n\nThe memory on next-generation GPUs/ASICs has become an aesthetic of brute force. The sheer size of equipped memory is something unimaginable three years ago. Looking back, the 80GB of the H100 in 2022 looks like a toy; it has grown tenfold in just a few years:\n\n- Nvidia Ultra Rubin - 1024GB HBM\n\n- Qualcomm AI200 - 768GB LPDDR\n\n- AMD MI400x - 432GB HBM\n\nAnother potential explosion point for memory is on the Edge, namely edge LLMs for phones/PCs/cars/robots. In the last two years, mainstream flagship phones have upgraded from 6GB to 8GB/12GB/16GB, preparing in advance for a possible edge LLM ecosystem. After all, next-gen phone compute will reach the 150 TOPS level—solidly desktop class and very powerful. In terms of potential, the edge memory upgrade is a larger market than the cloud memory increment. The number of edge terminal devices is staggering—billions annually. Once the edge LLM ecosystem thrives, doubling memory usage will be easy, and various designs for edge low-power memory/processing-in-memory will follow.\nHowever, the software ecosystem for edge GenAI seems to lag significantly, progressing much slower than I imagined. This is likely because this area is still in a groping phase and does not have the clearly defined ROI of the cloud. Manufacturers are cautious with investments. When I was bullish on 2027 back in '23–'24, I may have been too optimistic.\nInternet -> Mobile Internet took 10–15 years. Edge GenAI/LLM might also need 7–10 years. It may be that we have to wait until the cloud ROI is mostly exploited and marginal returns diminish before edge GenAI/LLM gets the development resources to validate the edge ROI.\n\nAnother story of internal structural shifts in semiconductors in 2025 is NAND storage, specifically enterprise eSSD.\n\nThe source of the structural trend is the same: the growing demand for AI inference workloads. The memory dividend has also spilled over into SSD storage, and even HDD storage, because when memory isn't enough, high-speed SSDs are used as multi-level caches. The main logic is KV cache offloading to the next layer of SSD storage during AI inference memory overflow, as well as vector database search/indexing—all increasing the demand for SSD storage.\n\nMicron's earnings report put it precisely and bluntly: \"AI inference use cases such as KV cache tiering and vector database search and indexing, are driving demand for performance storage.\"\n\nAs for why storage prices only exploded in the fourth quarter, one needs to distinguish between contract prices and spot prices. Contract price increases are milder; even for the scarcest enterprise eSSDs, contracts rose about 25% in Q4. But as NAND capacity was slowly eaten up by contracts in 2025, spot prices created a strong visual shock, rising over 50% in a single month.\n\nAnother unverified logic is the explosion of multimodal AI, especially the demand for AI images and AI video, which would exacerbate storage shortages. I feel this storyline is \"promising for the future,\" but the current precision of video/images is perhaps not even at the GPT-3 level yet; achieving a breakout effect will still take some time.\n\n------------------------ So, what are the next semiconductor structural opportunities brought by trend shifts?\n\nWe must look at what the trend for AI inference demand is next. Undoubtedly, the proportion of agentic flow will get larger and larger. 2025 is not just the \"year of the agent,\" but the start of the \"decade of the agent.\"\n\nViewing agentic workloads from a CPU perspective: routing and tool processing happen on the CPU. If you profile common agentic frameworks like SWE-Agent, LangChain, or Toolformer, the CPU can account for up to 90% of end-to-end latency. Throughput bottlenecks are also stuck more on the CPU, and CPU energy consumption can even exceed 40% of the total.\n\nAgentic AI is currently more of a CPU bottleneck issue. In agentic frameworks, the CPU is the \"Orchestrator\" that is always busy. This will likely drive a new wave of recovery in CPU demand.\n\nIn AMD's Q2 2025 earnings report (August 5), Lisa Su explicitly stated this phenomenon: \"In particular, adoption of agentic AI is creating additional demand for general-purpose compute infrastructure, as customers quickly realize that each token generated by a GPU triggers multiple CPU-intensive tasks.\"\n\nIn the Q3 report, Lisa played her hand again regarding CPU TAM increasing due to Gen AI: \"Many customers are now planning substantially larger CPU build outs over the coming quarters to support increased demands from AI, serving as a powerful new catalyst for our server business.\"\n\nNvidia also views agent flow as a CPU demand. The CPU ratio in the GB200/300 architecture configurations is much larger than before—36 Grace CPUs : 72 Blackwell GPUs, reaching a 1:2 level. AMD's route involves using 1~4 256-core EPYCs to serve the MI400 series. Future hardware architectures for 72~128 GPUs will definitely develop towards optimizing agent workloads, such as agent task graph scheduling, load balancing, and CPU/GPU cooperative micro-batching.\n\nComparisons in compute power might eventually move away from the current pure GPU token rate comparison toward a whole-system full-stack agentic benchmark comparison.\n\n-------------------------- While semiconductor structural shifts bring opportunities, the next steps may also bring unexpected secondary effects.\n\nThe explosion in cloud AI data center demand has caused memory and storage prices to skyrocket, putting immense pressure on costs for consumer electronics. In 2026, this may evolve into a potential Black Swan event for the consumer electronics industry.\n\nThe recent plunge in PC manufacturer stocks is due to this reason. HP has already stated it will reduce memory configurations, implying a return to the era of 8GB RAM + 256GB storage for PCs.\n\nIf DRAM memory and storage prices continue to rise like this, we might see a ridiculous situation: spot prices for memory/storage becoming more expensive than CPUs and GPUs. The awkward part is that this could directly delay the progress of AI PCs that consumer electronics hopes for, as large memory is more favorable for AI PC performance.\n\nTo exaggerate a bit: every employee of PC and mobile phone manufacturers, and even employees of consumer electronics manufacturers, should buy storage and memory stocks as a career risk hedge.\n\nStarting early next year, memory and storage costs for the Android camp will become uncontainable. If Samsung and Xiaomi raise phone prices (prices in the US market have already risen significantly), the biggest beneficiary will be Apple.\n\nApple's memory and NAND capacities are locked in via exclusive long-term contracts at fixed prices. Incidentally, they also trapped Kioxia for a lot of non-increasing price capacity. This leads to Apple's cost advantage expanding further. Apple's growth in global mobile phone market share could be very significant; the coming period might be a glorious time for the iPhone.\n\nThe 2025 semiconductor market is truly full of wonderful stories. The grudges and relationships between Nvidia/AMD/TPU and various hyperscalers have caused the moods of onlookers placing their bets to fluctuate wildly.\n\nHBM/Memory manufacturers ate the \"memory-bound\" dividends; NAND manufacturers unexpectedly harvested the spillover effects of KV cache; and CPUs, after nearly a decade of silence, may return to the center of the growth narrative due to agent orchestration.\n\nIt is no longer just Nvidia/Broadcom and a few compute vendors dominating. Rather, every evolution following the value spillover of AI workloads—from training to inference, from text to multimodal, from single model calls to agentic flow—is rewriting the value distribution of the industry chain.\n\nThe boom in Cloud AI is squeezing the living space of consumer electronics—while PC manufacturers are forced to discuss returning to the \"8GB era,\" Apple sits back and reaps the benefits due to supply chain advantages. The secondary effects of this compute arms race may reshape the entire consumer electronics landscape in 2026 in unexpected ways.\n\nThe story of semiconductors is no longer a single thread, but a web that continuously reconstructs itself. And 2025 is likely just the first round of these alliances and rivalries.", "reply_text": null, "parent_text": null, "quoted_text": "回顾2025年半导体市场，真的是有太多太多精彩的故事，最大的主题就是:\n\nAI需求驱动导致半导体基建的估值体系重构 + 产业链的价值分配重写\n\n从2024年开始，半导体基建正在飞速吞噬整个IT产业利润，SP500里半导体净利润EPS在IT行业里占比，在两年时间从不到20%上升了到了40%，而且还在呈加速上升姿态 https://t.co/OAFOULgBFX", "parent_author": null, "ret_m1d": -0.0162295101850789, "ret_signed_m1d": -0.0162295101850789, "alpha_spy_m1d": -0.020815864400725026, "alpha_c_m1d": -0.014302956854296056, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008559328912786102, "ret_signed_p1d": 0.008559328912786102, "alpha_spy_p1d": 0.006707119638674097, "alpha_c_p1d": -0.009771332097176177, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.03134909133996655, "ret_signed_p1w": 0.03134909133996655, "alpha_spy_p1w": 0.02640992676422016, "alpha_c_p1w": -0.012820177181220638, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.042410128700367755, "ret_signed_p1m": 0.042410128700367755, "alpha_spy_p1m": 0.029518113304120686, "alpha_c_p1m": 0.00988387547582148, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 0.027680540949180266, "ret_signed_p3m": 0.027680540949180266, "alpha_spy_p3m": 0.011412374035411066, "alpha_c_p3m": -0.14325104511514275, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 0.19432850523535183, "ret_signed_p6m": 0.19432850523535183, "alpha_spy_p6m": 0.08467522578184505, "alpha_c_p6m": -0.5169521570625641, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.0517, -0.046, -0.0718, -0.0646, -0.051, -0.0145, -0.0153, -0.0117, -0.0121, -0.028, -0.0535, -0.0205, -0.0181, 0.0205, -0.0083, -0.0164, -0.0124, -0.0096, 0.0107, 0.037, 0.0407, 0.0499, 0.0428, 0.0312, 0.0285, 0.0511, 0.0703, 0.018, 0.0467, 0.0006, -0.0005, 0.0105, 0.0183, 0.0151, 0.0069, 0.002, 0.0124, 0.0352, 0.0643, 0.1173, 0.1507, 0.1277, 0.1254, 0.1498, 0.1043, 0.085, 0.0454, 0.0457, 0.1063, 0.0736, 0.0771, 0.0386, 0.057, 0.0371, 0.008, 0.0367, 0.004, -0.0058, 0.0146, -0.0117, 0.0019, 0.0019, -0.0162, 0.0, 0.0086, -0.0018, 0.0193, 0.0139, 0.0313, 0.0281, 0.0215, 0.0057, -0.0272, -0.0201, -0.0122, -0.0499, -0.0321, 0.006, 0.021, 0.0517, 0.0484, 0.0484, 0.059, 0.0462, 0.0424, 0.0366, 0.0366, 0.0497, 0.0456, 0.0407, 0.0511, 0.0285, 0.0275, 0.028, 0.0328, 0.018, 0.0397, 0.0351, 0.0351, -0.0102, 0.019, 0.0274, 0.0431, 0.0365, 0.0479, 0.0645, 0.07, 0.0624, 0.0317, 0.0024, -0.0318, -0.0446, 0.0306, 0.0563, 0.048, 0.0564, 0.0391, 0.0161, 0.0161, 0.0281, 0.0449, 0.0444, 0.0551, 0.0647, 0.0719, 0.087, 0.0277, -0.0151, 0.0143, 0.0008, 0.0174, 0.0191, -0.0116, 0.0152, 0.027, 0.0341, 0.018, 0.0019, 0.0185, 0.0113, 0.0028, -0.0074, -0.04, -0.0237, -0.0261, -0.0068, -0.0481, -0.0688, -0.0819, -0.0306, -0.0231, -0.014, -0.014, -0.0126, -0.01, 0.0121, 0.0223, 0.0485, 0.0523, 0.0923, 0.1054, 0.1026, 0.1211, 0.1232, 0.1111, 0.1256, 0.1097, 0.1577, 0.2041, 0.1849, 0.1631, 0.1093, 0.1031, 0.1033, 0.0923, 0.1553, 0.1757, 0.1962, 0.2198, 0.2272, 0.2553, 0.3104, 0.2525, 0.2358, 0.2263, 0.2422, 0.2202, 0.1969, 0.1969, 0.1943]}
{"unit_id": "quote:1995319574393442673", "ticker_idx": 1, "ts": "2025-12-01T02:30:32+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "SK Hynix, leading in HBM, is even envisioning annual profits of $100 billion a few years from now, firmly becoming a trillion-dollar market cap company.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish semis through 2030 on AI-driven profit reallocation; NVDA structurally dominant, SK Hynix HBM winner, AMD CPU resurgence from agentic AI, Apple beneficiary of memory cost advantage.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Looking back at the semiconductor market in 2025, there were truly too many fascinating stories. The overarching theme was:\n\nAI demand-driven valuation restructuring of semiconductor infrastructure + The rewriting of value distribution across the supply chain.\n\nStarting in 2024, semiconductor infrastructure began rapidly devouring the profits of the entire IT industry. The share of semiconductor net profit (EPS) within the IT sector of the S&P 500 rose from less than 20% to 40% in just two years, and it is still in a state of accelerated ascent.\n\nThe overall forward-looking profit margin for semiconductors has risen from 25% in 2023 to 43% by November 2025, significantly surpassing the average profit margins of several internet giants. This confirms that semiconductor margins exceeding those of the internet sector will be the \"New Normal.\" The profit distribution of the entire IT industry is flowing increasingly toward semiconductors.\n\nIt is worth noting that even during the semiconductor chip shortage of 2020–2022, when shortages were severe, semiconductor profit margins and their share of total IT profit distribution did not see such significant growth.\n\nThis is the first half of the story: The valuation system of semiconductor infrastructure is being reconstructed by AI demand; it no longer holds the subordinate infrastructure status it did during the Internet era.\n\nThe logic behind this phenomenon is the transformation of business models alongside technological characteristics:\n\nIn the Internet era, the network and compute costs for each request had extremely low marginal costs. Scaling effects were excellent, and the marginal cost of distribution was almost zero.\n\nIn the AI era, this \"near-zero marginal distribution cost\" characteristic favorable to scalability has encountered a fundamental and major challenge: Not only is training cost no longer a one-time expense but rather one that grows annually, but even for client-side AI inference requests, costs do not drop in sync with hardware compute deflation. This is because \"inference scaling\" has become a consensus, and vertical domains still require larger-scale flagship models to maintain competitiveness.\n\nPreviously, the biggest cost for internet companies was OPEX, specifically SDE labor costs. Now, for the first time in history, internet companies are burdened with balance sheets full of high depreciation costs, similar to semiconductor foundries. The business model is practically itching to shift slowly from \"Traffic × Conversion Rate\" to \"Gross Profit per Token.\"\n\nSimply put, the cost distribution shift from the Internet era to the AI era has added heavy hardware/compute CAPEX (MSFT 33%, Meta 38% in earnings reports) on top of the OPEX base. The winners of the previous era were Internet companies + CSPs + SaaS (the \"rent-collecting\" industries). In the AI era, compute (semiconductor/chip depreciation) has become the new rent-collecting industry. The profit distribution of the entire IT industry has undergone a drastic reallocation (EPS flow to semiconductors rising from 20% to 40% and climbing). This is the most important reason for the restructuring of the semiconductor infrastructure valuation system.\n\n--------------- How long can the \"New Normal\" of high semiconductor margins last?\n\nThe current high premiums come from an excessive backlog of semiconductor orders caused by an earlier \"cost-is-no-object\" arms race. However, obviously, hyperscalers do not want to be taken for a ride; they are all attempting to build in-house ASICs to lower costs. We can look at this problem from the perspective of the long-term compute distribution in 2030.\n\nIn the long run, OpenAI has already revealed the \"standard answer\": 10GW Nvidia, 10GW ASIC, 6GW AMD. Other hyperscalers have similar considerations for their split.\n\nFor example, on the inference side, they hope for ASIC >50%. Within the GPU segment, a 50/50 split between AMD and NV (legacy). For training, NV will still hold the lion's share (60%+), with the rest split between in-house ASICs and AMD.\n\nCalculated based on a 2030 split of 60% inference and 40% training, the market share comes out to NV 38%, ASIC 39%, and AMD 23%. This is almost entirely consistent with OpenAI’s ratios and can be considered a standard reference value.\n\nOf course, for Microsoft, Amazon, Google, and Anthropic, the AMD ratio will be noticeably lower than this standard answer/reference value, while xAI has no ASICs and uses only Nvidia + a small amount of AMD.\n\nAMD's risk lies in the ultra-long term (post-2030). As CSP in-house ASICs become more mature (except for Microsoft), the proportion of ASICs in inference may grow higher, making it hard to find incentives to buy large quantities of new GPUs unless they are sold cheaply enough.\n\nWhat about the recently hyped TPU? Is Meta shifting to TPUs? Will it significantly impact Nvidia’s margins?\n\nActually, Meta’s CAPEX is $72B this year, $110B next year, and the average annual CAPEX over the next six years may reach around $160B. Meanwhile, Meta’s 6-year $10B TPU order amounts to only $1.6B per year on average—and they are purchasing TPU cloud services, not bare TPUs.\n\nIn other words, this TPU order accounts for only 1% of Meta’s future 6-year CAPEX. They are not seriously considering large-scale deployment; it is likely just a means to bargain with Nvidia.\n\nFurthermore, looking at Meta’s recruitment ads over the recent months, we don’t see any hiring for TPU engineers, unlike Anthropic, which was hiring a bunch of TPU kernel engineers back in May before announcing large-scale TPU procurement for training in October.\n\nSo, whether the reason is diversifying suppliers, creating a fallback for delayed in-house ASICs, or delays with AMD’s MI350X, Meta’s purchase of TPUs basically has one consideration: increasing bargaining power when buying Nvidia GPUs. However, they can essentially only bargain on the inference share; the actual effect is very limited, as is the impact on Nvidia’s profit margins.\n\nKeep in mind that during the crypto bear market crash of 2022, NVDA inventory rose to 198 days, yet margins only retreated from 65% to 56%. Even with the \"double kill\" of PE and macroeconomics, the stock price only went from 300 to 100. Now, with supply consistently falling short of demand, there is no reason for margins to come down.\n\nAdditionally, the TPU v8 design is too conservative (it didn't use HBM4), and the TCO (Total Cost of Ownership) of the Kyber rack Rubin solution will be better than TPU v8. In the end, they will still have to rely on Nvidia, making it hard to bargain. As long as Nvidia continues its stride forward, it is actually not easy for competitors to keep up.\n\nIn summary, on one hand, bottlenecks across the entire supply chain (e.g., CoWoS expansion) are cautious, so the state of demand exceeding supply can last for many years.\n\nOn the other hand, there is a \"temporal mismatch\" between the profit curve of AI monetization and the hardware investment curve. The growth curve of the application side will lag by a few years. As long as this time lag between the application side and infrastructure side growth curves exists, semiconductors will continue to dominate profit distribution in the IT industry.\n\nLooking at OpenAI’s investment curve through 2030, this mismatch will last until around 2030. This means the dominant position in IT industry profit distribution brought about by the semiconductor industry’s super-expansion phase will last at least until 2030 based on current views.\n\nAnd high semiconductor margins might be sustained even longer, because the attribute has changed from \"one-time infrastructure\" in the internet era to \"rent-collecting infrastructure\" now.\n\nAI has not only fed the GPU; it has used the compute budget to lift every link in the chain that \"turns electricity into tokens\"—from memory, storage, interconnects, and fiber optics, to electricity, energy storage, and so on.\n\nThe first half covered \"Semiconductors swallowing IT profits.\" The second half covers how the \"AI Compute Value Spillover Effect\" reshapes the internal landscape of semiconductors: GPU Compute Growth -> Memory/Storage/Interconnect/CPU Bottlenecks -> Spillover Effect -> Structural Opportunities.\n\nThe more interesting story in 2025 is how huge industry dividends are creating structural new opportunities within the semiconductor sector. For instance, a super cluster requires several data center interconnects; fiber optic interconnect lengths reach the level of millions of miles. This is a new opportunity.\n\nThe most typical examples of new opportunities brought by structural trends in the semiconductor supply chain are Memory (DRAM/HBM) and Storage (SSD). The growth in HBM demand is so exaggerated that it is squeezing DDR4/5 capacity, leading the memory industry—historically known for its cyclical nature—to shout that \"the cycle no longer exists.\" SK Hynix, leading in HBM, is even envisioning annual profits of $100 billion a few years from now, firmly becoming a trillion-dollar market cap company.\n\nBehind these two sectors is a shift in structural trends: AI workloads are gradually extending from training to inference, and the proportion of inference is getting larger.\n\nInference is a purely memory-bound task; one could say bandwidth speed = tokens/s. As model sizes get larger and context lengths increase, the requirements for memory size also grow accordingly, leading to a surge in memory demand: Inference is Memory.\n\nThe memory on next-generation GPUs/ASICs has become an aesthetic of brute force. The sheer size of equipped memory is something unimaginable three years ago. Looking back, the 80GB of the H100 in 2022 looks like a toy; it has grown tenfold in just a few years:\n\n- Nvidia Ultra Rubin - 1024GB HBM\n\n- Qualcomm AI200 - 768GB LPDDR\n\n- AMD MI400x - 432GB HBM\n\nAnother potential explosion point for memory is on the Edge, namely edge LLMs for phones/PCs/cars/robots. In the last two years, mainstream flagship phones have upgraded from 6GB to 8GB/12GB/16GB, preparing in advance for a possible edge LLM ecosystem. After all, next-gen phone compute will reach the 150 TOPS level—solidly desktop class and very powerful. In terms of potential, the edge memory upgrade is a larger market than the cloud memory increment. The number of edge terminal devices is staggering—billions annually. Once the edge LLM ecosystem thrives, doubling memory usage will be easy, and various designs for edge low-power memory/processing-in-memory will follow.\nHowever, the software ecosystem for edge GenAI seems to lag significantly, progressing much slower than I imagined. This is likely because this area is still in a groping phase and does not have the clearly defined ROI of the cloud. Manufacturers are cautious with investments. When I was bullish on 2027 back in '23–'24, I may have been too optimistic.\nInternet -> Mobile Internet took 10–15 years. Edge GenAI/LLM might also need 7–10 years. It may be that we have to wait until the cloud ROI is mostly exploited and marginal returns diminish before edge GenAI/LLM gets the development resources to validate the edge ROI.\n\nAnother story of internal structural shifts in semiconductors in 2025 is NAND storage, specifically enterprise eSSD.\n\nThe source of the structural trend is the same: the growing demand for AI inference workloads. The memory dividend has also spilled over into SSD storage, and even HDD storage, because when memory isn't enough, high-speed SSDs are used as multi-level caches. The main logic is KV cache offloading to the next layer of SSD storage during AI inference memory overflow, as well as vector database search/indexing—all increasing the demand for SSD storage.\n\nMicron's earnings report put it precisely and bluntly: \"AI inference use cases such as KV cache tiering and vector database search and indexing, are driving demand for performance storage.\"\n\nAs for why storage prices only exploded in the fourth quarter, one needs to distinguish between contract prices and spot prices. Contract price increases are milder; even for the scarcest enterprise eSSDs, contracts rose about 25% in Q4. But as NAND capacity was slowly eaten up by contracts in 2025, spot prices created a strong visual shock, rising over 50% in a single month.\n\nAnother unverified logic is the explosion of multimodal AI, especially the demand for AI images and AI video, which would exacerbate storage shortages. I feel this storyline is \"promising for the future,\" but the current precision of video/images is perhaps not even at the GPT-3 level yet; achieving a breakout effect will still take some time.\n\n------------------------ So, what are the next semiconductor structural opportunities brought by trend shifts?\n\nWe must look at what the trend for AI inference demand is next. Undoubtedly, the proportion of agentic flow will get larger and larger. 2025 is not just the \"year of the agent,\" but the start of the \"decade of the agent.\"\n\nViewing agentic workloads from a CPU perspective: routing and tool processing happen on the CPU. If you profile common agentic frameworks like SWE-Agent, LangChain, or Toolformer, the CPU can account for up to 90% of end-to-end latency. Throughput bottlenecks are also stuck more on the CPU, and CPU energy consumption can even exceed 40% of the total.\n\nAgentic AI is currently more of a CPU bottleneck issue. In agentic frameworks, the CPU is the \"Orchestrator\" that is always busy. This will likely drive a new wave of recovery in CPU demand.\n\nIn AMD's Q2 2025 earnings report (August 5), Lisa Su explicitly stated this phenomenon: \"In particular, adoption of agentic AI is creating additional demand for general-purpose compute infrastructure, as customers quickly realize that each token generated by a GPU triggers multiple CPU-intensive tasks.\"\n\nIn the Q3 report, Lisa played her hand again regarding CPU TAM increasing due to Gen AI: \"Many customers are now planning substantially larger CPU build outs over the coming quarters to support increased demands from AI, serving as a powerful new catalyst for our server business.\"\n\nNvidia also views agent flow as a CPU demand. The CPU ratio in the GB200/300 architecture configurations is much larger than before—36 Grace CPUs : 72 Blackwell GPUs, reaching a 1:2 level. AMD's route involves using 1~4 256-core EPYCs to serve the MI400 series. Future hardware architectures for 72~128 GPUs will definitely develop towards optimizing agent workloads, such as agent task graph scheduling, load balancing, and CPU/GPU cooperative micro-batching.\n\nComparisons in compute power might eventually move away from the current pure GPU token rate comparison toward a whole-system full-stack agentic benchmark comparison.\n\n-------------------------- While semiconductor structural shifts bring opportunities, the next steps may also bring unexpected secondary effects.\n\nThe explosion in cloud AI data center demand has caused memory and storage prices to skyrocket, putting immense pressure on costs for consumer electronics. In 2026, this may evolve into a potential Black Swan event for the consumer electronics industry.\n\nThe recent plunge in PC manufacturer stocks is due to this reason. HP has already stated it will reduce memory configurations, implying a return to the era of 8GB RAM + 256GB storage for PCs.\n\nIf DRAM memory and storage prices continue to rise like this, we might see a ridiculous situation: spot prices for memory/storage becoming more expensive than CPUs and GPUs. The awkward part is that this could directly delay the progress of AI PCs that consumer electronics hopes for, as large memory is more favorable for AI PC performance.\n\nTo exaggerate a bit: every employee of PC and mobile phone manufacturers, and even employees of consumer electronics manufacturers, should buy storage and memory stocks as a career risk hedge.\n\nStarting early next year, memory and storage costs for the Android camp will become uncontainable. If Samsung and Xiaomi raise phone prices (prices in the US market have already risen significantly), the biggest beneficiary will be Apple.\n\nApple's memory and NAND capacities are locked in via exclusive long-term contracts at fixed prices. Incidentally, they also trapped Kioxia for a lot of non-increasing price capacity. This leads to Apple's cost advantage expanding further. Apple's growth in global mobile phone market share could be very significant; the coming period might be a glorious time for the iPhone.\n\nThe 2025 semiconductor market is truly full of wonderful stories. The grudges and relationships between Nvidia/AMD/TPU and various hyperscalers have caused the moods of onlookers placing their bets to fluctuate wildly.\n\nHBM/Memory manufacturers ate the \"memory-bound\" dividends; NAND manufacturers unexpectedly harvested the spillover effects of KV cache; and CPUs, after nearly a decade of silence, may return to the center of the growth narrative due to agent orchestration.\n\nIt is no longer just Nvidia/Broadcom and a few compute vendors dominating. Rather, every evolution following the value spillover of AI workloads—from training to inference, from text to multimodal, from single model calls to agentic flow—is rewriting the value distribution of the industry chain.\n\nThe boom in Cloud AI is squeezing the living space of consumer electronics—while PC manufacturers are forced to discuss returning to the \"8GB era,\" Apple sits back and reaps the benefits due to supply chain advantages. The secondary effects of this compute arms race may reshape the entire consumer electronics landscape in 2026 in unexpected ways.\n\nThe story of semiconductors is no longer a single thread, but a web that continuously reconstructs itself. And 2025 is likely just the first round of these alliances and rivalries.", "reply_text": null, "parent_text": null, "quoted_text": "回顾2025年半导体市场，真的是有太多太多精彩的故事，最大的主题就是:\n\nAI需求驱动导致半导体基建的估值体系重构 + 产业链的价值分配重写\n\n从2024年开始，半导体基建正在飞速吞噬整个IT产业利润，SP500里半导体净利润EPS在IT行业里占比，在两年时间从不到20%上升了到了40%，而且还在呈加速上升姿态 https://t.co/OAFOULgBFX", "parent_author": null, "ret_m1d": -0.01486978721660659, "ret_signed_m1d": -0.01486978721660659, "alpha_spy_m1d": -0.019456141432252716, "alpha_c_m1d": -0.012943233885823746, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.037174817256039994, "ret_signed_p1d": 0.037174817256039994, "alpha_spy_p1d": 0.03532260798192799, "alpha_c_p1d": 0.018844156246077715, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.07249076560395262, "ret_signed_p1w": 0.07249076560395262, "alpha_spy_p1w": 0.06755160102820623, "alpha_c_p1w": 0.028321497082765434, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.21003728679871325, "ret_signed_p1m": 0.21003728679871325, "alpha_spy_p1m": 0.19714527140246618, "alpha_c_p1m": 0.17751103357416698, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 1.0465204963316177, "ret_signed_p3m": 1.0465204963316177, "alpha_spy_p3m": 1.0302523294178485, "alpha_c_p3m": 0.8755889102672947, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.821164681653712, "ret_signed_p6m": 2.821164681653712, "alpha_spy_p6m": 2.7115114022002054, "alpha_c_p6m": 2.109884019355796, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.5124, -0.5069, -0.492, -0.4855, -0.4651, -0.4353, -0.4298, -0.3898, -0.3852, -0.3536, -0.3806, -0.3397, -0.3397, -0.3481, -0.3295, -0.336, -0.3378, -0.375, -0.3518, -0.3546, -0.3313, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.205, -0.2292, -0.2357, -0.2152, -0.1595, -0.1354, -0.0982, -0.1103, -0.1057, -0.1112, -0.0527, -0.0063, -0.0323, 0.0364, 0.055, 0.0383, 0.1516, 0.0884, 0.0754, 0.1014, 0.0773, 0.1256, 0.1497, 0.146, 0.1367, 0.0401, 0.1256, 0.0587, 0.0439, 0.0606, -0.0323, -0.0341, -0.036, -0.0267, 0.0112, -0.0149, 0.0, 0.0372, 0.026, 0.0074, 0.0112, 0.0725, 0.052, 0.0911, 0.0502, 0.0613, 0.0297, -0.0149, 0.0242, 0.026, 0.0167, 0.0781, 0.0855, 0.0929, 0.0929, 0.1134, 0.1896, 0.21, 0.21, 0.21, 0.2584, 0.2937, 0.3494, 0.3792, 0.4052, 0.3829, 0.3922, 0.3717, 0.3792, 0.3922, 0.4052, 0.4201, 0.381, 0.3755, 0.4033, 0.4257, 0.368, 0.487, 0.5632, 0.6004, 0.6896, 0.5428, 0.6859, 0.6729, 0.5651, 0.5595, 0.6487, 0.6283, 0.5985, 0.6506, 0.6357, 0.6357, 0.6357, 0.6357, 0.6617, 0.7639, 0.7677, 0.868, 0.8922, 1.0465, 0.9758, 0.9758, 0.7486, 0.581, 0.7523, 0.7206, 0.5568, 0.7467, 0.7784, 0.7318, 0.6946, 0.8137, 0.8063, 0.9664, 0.8864, 0.8752, 0.7374, 0.8361, 0.8529, 0.7374, 0.7374, 0.6257, 0.5028, 0.6704, 0.5456, 0.6313, 0.6499, 0.7057, 0.9236, 0.8584, 0.9124, 0.9367, 1.054, 1.1154, 1.1508, 1.1005, 1.1713, 1.2793, 1.2774, 1.2812, 1.2756, 1.4059, 1.4208, 1.4078, 1.3947, 1.3947, 1.6946, 1.6946, 1.9813, 2.08, 2.1396, 2.5009, 2.4171, 2.6796, 2.6685, 2.3873, 2.4264, 2.2495, 2.2495, 2.6126, 2.6145, 2.6145, 2.8212]}
{"unit_id": "quote:1995319574393442673", "ticker_idx": 2, "ts": "2025-12-01T02:30:32+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's Q2 2025 earnings report (August 5), Lisa Su explicitly stated this phenomenon: adoption of agentic AI is creating additional demand for general-purpose compute infrastructure", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish semis through 2030 on AI-driven profit reallocation; NVDA structurally dominant, SK Hynix HBM winner, AMD CPU resurgence from agentic AI, Apple beneficiary of memory cost advantage.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Looking back at the semiconductor market in 2025, there were truly too many fascinating stories. The overarching theme was:\n\nAI demand-driven valuation restructuring of semiconductor infrastructure + The rewriting of value distribution across the supply chain.\n\nStarting in 2024, semiconductor infrastructure began rapidly devouring the profits of the entire IT industry. The share of semiconductor net profit (EPS) within the IT sector of the S&P 500 rose from less than 20% to 40% in just two years, and it is still in a state of accelerated ascent.\n\nThe overall forward-looking profit margin for semiconductors has risen from 25% in 2023 to 43% by November 2025, significantly surpassing the average profit margins of several internet giants. This confirms that semiconductor margins exceeding those of the internet sector will be the \"New Normal.\" The profit distribution of the entire IT industry is flowing increasingly toward semiconductors.\n\nIt is worth noting that even during the semiconductor chip shortage of 2020–2022, when shortages were severe, semiconductor profit margins and their share of total IT profit distribution did not see such significant growth.\n\nThis is the first half of the story: The valuation system of semiconductor infrastructure is being reconstructed by AI demand; it no longer holds the subordinate infrastructure status it did during the Internet era.\n\nThe logic behind this phenomenon is the transformation of business models alongside technological characteristics:\n\nIn the Internet era, the network and compute costs for each request had extremely low marginal costs. Scaling effects were excellent, and the marginal cost of distribution was almost zero.\n\nIn the AI era, this \"near-zero marginal distribution cost\" characteristic favorable to scalability has encountered a fundamental and major challenge: Not only is training cost no longer a one-time expense but rather one that grows annually, but even for client-side AI inference requests, costs do not drop in sync with hardware compute deflation. This is because \"inference scaling\" has become a consensus, and vertical domains still require larger-scale flagship models to maintain competitiveness.\n\nPreviously, the biggest cost for internet companies was OPEX, specifically SDE labor costs. Now, for the first time in history, internet companies are burdened with balance sheets full of high depreciation costs, similar to semiconductor foundries. The business model is practically itching to shift slowly from \"Traffic × Conversion Rate\" to \"Gross Profit per Token.\"\n\nSimply put, the cost distribution shift from the Internet era to the AI era has added heavy hardware/compute CAPEX (MSFT 33%, Meta 38% in earnings reports) on top of the OPEX base. The winners of the previous era were Internet companies + CSPs + SaaS (the \"rent-collecting\" industries). In the AI era, compute (semiconductor/chip depreciation) has become the new rent-collecting industry. The profit distribution of the entire IT industry has undergone a drastic reallocation (EPS flow to semiconductors rising from 20% to 40% and climbing). This is the most important reason for the restructuring of the semiconductor infrastructure valuation system.\n\n--------------- How long can the \"New Normal\" of high semiconductor margins last?\n\nThe current high premiums come from an excessive backlog of semiconductor orders caused by an earlier \"cost-is-no-object\" arms race. However, obviously, hyperscalers do not want to be taken for a ride; they are all attempting to build in-house ASICs to lower costs. We can look at this problem from the perspective of the long-term compute distribution in 2030.\n\nIn the long run, OpenAI has already revealed the \"standard answer\": 10GW Nvidia, 10GW ASIC, 6GW AMD. Other hyperscalers have similar considerations for their split.\n\nFor example, on the inference side, they hope for ASIC >50%. Within the GPU segment, a 50/50 split between AMD and NV (legacy). For training, NV will still hold the lion's share (60%+), with the rest split between in-house ASICs and AMD.\n\nCalculated based on a 2030 split of 60% inference and 40% training, the market share comes out to NV 38%, ASIC 39%, and AMD 23%. This is almost entirely consistent with OpenAI’s ratios and can be considered a standard reference value.\n\nOf course, for Microsoft, Amazon, Google, and Anthropic, the AMD ratio will be noticeably lower than this standard answer/reference value, while xAI has no ASICs and uses only Nvidia + a small amount of AMD.\n\nAMD's risk lies in the ultra-long term (post-2030). As CSP in-house ASICs become more mature (except for Microsoft), the proportion of ASICs in inference may grow higher, making it hard to find incentives to buy large quantities of new GPUs unless they are sold cheaply enough.\n\nWhat about the recently hyped TPU? Is Meta shifting to TPUs? Will it significantly impact Nvidia’s margins?\n\nActually, Meta’s CAPEX is $72B this year, $110B next year, and the average annual CAPEX over the next six years may reach around $160B. Meanwhile, Meta’s 6-year $10B TPU order amounts to only $1.6B per year on average—and they are purchasing TPU cloud services, not bare TPUs.\n\nIn other words, this TPU order accounts for only 1% of Meta’s future 6-year CAPEX. They are not seriously considering large-scale deployment; it is likely just a means to bargain with Nvidia.\n\nFurthermore, looking at Meta’s recruitment ads over the recent months, we don’t see any hiring for TPU engineers, unlike Anthropic, which was hiring a bunch of TPU kernel engineers back in May before announcing large-scale TPU procurement for training in October.\n\nSo, whether the reason is diversifying suppliers, creating a fallback for delayed in-house ASICs, or delays with AMD’s MI350X, Meta’s purchase of TPUs basically has one consideration: increasing bargaining power when buying Nvidia GPUs. However, they can essentially only bargain on the inference share; the actual effect is very limited, as is the impact on Nvidia’s profit margins.\n\nKeep in mind that during the crypto bear market crash of 2022, NVDA inventory rose to 198 days, yet margins only retreated from 65% to 56%. Even with the \"double kill\" of PE and macroeconomics, the stock price only went from 300 to 100. Now, with supply consistently falling short of demand, there is no reason for margins to come down.\n\nAdditionally, the TPU v8 design is too conservative (it didn't use HBM4), and the TCO (Total Cost of Ownership) of the Kyber rack Rubin solution will be better than TPU v8. In the end, they will still have to rely on Nvidia, making it hard to bargain. As long as Nvidia continues its stride forward, it is actually not easy for competitors to keep up.\n\nIn summary, on one hand, bottlenecks across the entire supply chain (e.g., CoWoS expansion) are cautious, so the state of demand exceeding supply can last for many years.\n\nOn the other hand, there is a \"temporal mismatch\" between the profit curve of AI monetization and the hardware investment curve. The growth curve of the application side will lag by a few years. As long as this time lag between the application side and infrastructure side growth curves exists, semiconductors will continue to dominate profit distribution in the IT industry.\n\nLooking at OpenAI’s investment curve through 2030, this mismatch will last until around 2030. This means the dominant position in IT industry profit distribution brought about by the semiconductor industry’s super-expansion phase will last at least until 2030 based on current views.\n\nAnd high semiconductor margins might be sustained even longer, because the attribute has changed from \"one-time infrastructure\" in the internet era to \"rent-collecting infrastructure\" now.\n\nAI has not only fed the GPU; it has used the compute budget to lift every link in the chain that \"turns electricity into tokens\"—from memory, storage, interconnects, and fiber optics, to electricity, energy storage, and so on.\n\nThe first half covered \"Semiconductors swallowing IT profits.\" The second half covers how the \"AI Compute Value Spillover Effect\" reshapes the internal landscape of semiconductors: GPU Compute Growth -> Memory/Storage/Interconnect/CPU Bottlenecks -> Spillover Effect -> Structural Opportunities.\n\nThe more interesting story in 2025 is how huge industry dividends are creating structural new opportunities within the semiconductor sector. For instance, a super cluster requires several data center interconnects; fiber optic interconnect lengths reach the level of millions of miles. This is a new opportunity.\n\nThe most typical examples of new opportunities brought by structural trends in the semiconductor supply chain are Memory (DRAM/HBM) and Storage (SSD). The growth in HBM demand is so exaggerated that it is squeezing DDR4/5 capacity, leading the memory industry—historically known for its cyclical nature—to shout that \"the cycle no longer exists.\" SK Hynix, leading in HBM, is even envisioning annual profits of $100 billion a few years from now, firmly becoming a trillion-dollar market cap company.\n\nBehind these two sectors is a shift in structural trends: AI workloads are gradually extending from training to inference, and the proportion of inference is getting larger.\n\nInference is a purely memory-bound task; one could say bandwidth speed = tokens/s. As model sizes get larger and context lengths increase, the requirements for memory size also grow accordingly, leading to a surge in memory demand: Inference is Memory.\n\nThe memory on next-generation GPUs/ASICs has become an aesthetic of brute force. The sheer size of equipped memory is something unimaginable three years ago. Looking back, the 80GB of the H100 in 2022 looks like a toy; it has grown tenfold in just a few years:\n\n- Nvidia Ultra Rubin - 1024GB HBM\n\n- Qualcomm AI200 - 768GB LPDDR\n\n- AMD MI400x - 432GB HBM\n\nAnother potential explosion point for memory is on the Edge, namely edge LLMs for phones/PCs/cars/robots. In the last two years, mainstream flagship phones have upgraded from 6GB to 8GB/12GB/16GB, preparing in advance for a possible edge LLM ecosystem. After all, next-gen phone compute will reach the 150 TOPS level—solidly desktop class and very powerful. In terms of potential, the edge memory upgrade is a larger market than the cloud memory increment. The number of edge terminal devices is staggering—billions annually. Once the edge LLM ecosystem thrives, doubling memory usage will be easy, and various designs for edge low-power memory/processing-in-memory will follow.\nHowever, the software ecosystem for edge GenAI seems to lag significantly, progressing much slower than I imagined. This is likely because this area is still in a groping phase and does not have the clearly defined ROI of the cloud. Manufacturers are cautious with investments. When I was bullish on 2027 back in '23–'24, I may have been too optimistic.\nInternet -> Mobile Internet took 10–15 years. Edge GenAI/LLM might also need 7–10 years. It may be that we have to wait until the cloud ROI is mostly exploited and marginal returns diminish before edge GenAI/LLM gets the development resources to validate the edge ROI.\n\nAnother story of internal structural shifts in semiconductors in 2025 is NAND storage, specifically enterprise eSSD.\n\nThe source of the structural trend is the same: the growing demand for AI inference workloads. The memory dividend has also spilled over into SSD storage, and even HDD storage, because when memory isn't enough, high-speed SSDs are used as multi-level caches. The main logic is KV cache offloading to the next layer of SSD storage during AI inference memory overflow, as well as vector database search/indexing—all increasing the demand for SSD storage.\n\nMicron's earnings report put it precisely and bluntly: \"AI inference use cases such as KV cache tiering and vector database search and indexing, are driving demand for performance storage.\"\n\nAs for why storage prices only exploded in the fourth quarter, one needs to distinguish between contract prices and spot prices. Contract price increases are milder; even for the scarcest enterprise eSSDs, contracts rose about 25% in Q4. But as NAND capacity was slowly eaten up by contracts in 2025, spot prices created a strong visual shock, rising over 50% in a single month.\n\nAnother unverified logic is the explosion of multimodal AI, especially the demand for AI images and AI video, which would exacerbate storage shortages. I feel this storyline is \"promising for the future,\" but the current precision of video/images is perhaps not even at the GPT-3 level yet; achieving a breakout effect will still take some time.\n\n------------------------ So, what are the next semiconductor structural opportunities brought by trend shifts?\n\nWe must look at what the trend for AI inference demand is next. Undoubtedly, the proportion of agentic flow will get larger and larger. 2025 is not just the \"year of the agent,\" but the start of the \"decade of the agent.\"\n\nViewing agentic workloads from a CPU perspective: routing and tool processing happen on the CPU. If you profile common agentic frameworks like SWE-Agent, LangChain, or Toolformer, the CPU can account for up to 90% of end-to-end latency. Throughput bottlenecks are also stuck more on the CPU, and CPU energy consumption can even exceed 40% of the total.\n\nAgentic AI is currently more of a CPU bottleneck issue. In agentic frameworks, the CPU is the \"Orchestrator\" that is always busy. This will likely drive a new wave of recovery in CPU demand.\n\nIn AMD's Q2 2025 earnings report (August 5), Lisa Su explicitly stated this phenomenon: \"In particular, adoption of agentic AI is creating additional demand for general-purpose compute infrastructure, as customers quickly realize that each token generated by a GPU triggers multiple CPU-intensive tasks.\"\n\nIn the Q3 report, Lisa played her hand again regarding CPU TAM increasing due to Gen AI: \"Many customers are now planning substantially larger CPU build outs over the coming quarters to support increased demands from AI, serving as a powerful new catalyst for our server business.\"\n\nNvidia also views agent flow as a CPU demand. The CPU ratio in the GB200/300 architecture configurations is much larger than before—36 Grace CPUs : 72 Blackwell GPUs, reaching a 1:2 level. AMD's route involves using 1~4 256-core EPYCs to serve the MI400 series. Future hardware architectures for 72~128 GPUs will definitely develop towards optimizing agent workloads, such as agent task graph scheduling, load balancing, and CPU/GPU cooperative micro-batching.\n\nComparisons in compute power might eventually move away from the current pure GPU token rate comparison toward a whole-system full-stack agentic benchmark comparison.\n\n-------------------------- While semiconductor structural shifts bring opportunities, the next steps may also bring unexpected secondary effects.\n\nThe explosion in cloud AI data center demand has caused memory and storage prices to skyrocket, putting immense pressure on costs for consumer electronics. In 2026, this may evolve into a potential Black Swan event for the consumer electronics industry.\n\nThe recent plunge in PC manufacturer stocks is due to this reason. HP has already stated it will reduce memory configurations, implying a return to the era of 8GB RAM + 256GB storage for PCs.\n\nIf DRAM memory and storage prices continue to rise like this, we might see a ridiculous situation: spot prices for memory/storage becoming more expensive than CPUs and GPUs. The awkward part is that this could directly delay the progress of AI PCs that consumer electronics hopes for, as large memory is more favorable for AI PC performance.\n\nTo exaggerate a bit: every employee of PC and mobile phone manufacturers, and even employees of consumer electronics manufacturers, should buy storage and memory stocks as a career risk hedge.\n\nStarting early next year, memory and storage costs for the Android camp will become uncontainable. If Samsung and Xiaomi raise phone prices (prices in the US market have already risen significantly), the biggest beneficiary will be Apple.\n\nApple's memory and NAND capacities are locked in via exclusive long-term contracts at fixed prices. Incidentally, they also trapped Kioxia for a lot of non-increasing price capacity. This leads to Apple's cost advantage expanding further. Apple's growth in global mobile phone market share could be very significant; the coming period might be a glorious time for the iPhone.\n\nThe 2025 semiconductor market is truly full of wonderful stories. The grudges and relationships between Nvidia/AMD/TPU and various hyperscalers have caused the moods of onlookers placing their bets to fluctuate wildly.\n\nHBM/Memory manufacturers ate the \"memory-bound\" dividends; NAND manufacturers unexpectedly harvested the spillover effects of KV cache; and CPUs, after nearly a decade of silence, may return to the center of the growth narrative due to agent orchestration.\n\nIt is no longer just Nvidia/Broadcom and a few compute vendors dominating. Rather, every evolution following the value spillover of AI workloads—from training to inference, from text to multimodal, from single model calls to agentic flow—is rewriting the value distribution of the industry chain.\n\nThe boom in Cloud AI is squeezing the living space of consumer electronics—while PC manufacturers are forced to discuss returning to the \"8GB era,\" Apple sits back and reaps the benefits due to supply chain advantages. The secondary effects of this compute arms race may reshape the entire consumer electronics landscape in 2026 in unexpected ways.\n\nThe story of semiconductors is no longer a single thread, but a web that continuously reconstructs itself. And 2025 is likely just the first round of these alliances and rivalries.", "reply_text": null, "parent_text": null, "quoted_text": "回顾2025年半导体市场，真的是有太多太多精彩的故事，最大的主题就是:\n\nAI需求驱动导致半导体基建的估值体系重构 + 产业链的价值分配重写\n\n从2024年开始，半导体基建正在飞速吞噬整个IT产业利润，SP500里半导体净利润EPS在IT行业里占比，在两年时间从不到20%上升了到了40%，而且还在呈加速上升姿态 https://t.co/OAFOULgBFX", "parent_author": null, "ret_m1d": -0.010147414376048713, "ret_signed_m1d": -0.010147414376048713, "alpha_spy_m1d": -0.01473376859169484, "alpha_c_m1d": -0.00822086104526587, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020567842767812206, "ret_signed_p1d": -0.020567842767812206, "alpha_spy_p1d": -0.02242005204192421, "alpha_c_p1d": -0.038898503777774485, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.0061430930891002156, "ret_signed_p1w": 0.0061430930891002156, "alpha_spy_p1w": 0.0012039285133538247, "alpha_c_p1w": -0.03802617543208697, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": -0.020112842552914345, "ret_signed_p1m": -0.020112842552914345, "alpha_spy_p1m": -0.033004857949161415, "alpha_c_p1m": -0.05263909577746062, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": -0.07317074186837269, "ret_signed_p3m": -0.07317074186837269, "alpha_spy_p3m": -0.08943890878214189, "alpha_c_p3m": -0.2441023279326957, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 1.2929105717317593, "ret_signed_p6m": 1.2929105717317593, "alpha_spy_p6m": 1.1832572922782525, "alpha_c_p6m": 0.5816299094338433, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.2622, -0.2638, -0.3122, -0.311, -0.291, -0.274, -0.2916, -0.2784, -0.2667, -0.2698, -0.2758, -0.2814, -0.2838, -0.2729, -0.2678, -0.2679, -0.2662, -0.2744, -0.2657, -0.2638, -0.2537, -0.2277, -0.2507, -0.073, -0.0375, 0.0719, 0.0597, -0.0221, -0.0152, -0.0076, 0.0857, 0.0673, 0.0606, 0.0946, 0.0831, 0.0476, 0.0693, 0.1509, 0.1816, 0.1741, 0.2028, 0.1596, 0.1655, 0.1815, 0.1378, 0.1664, 0.0816, 0.0627, 0.1102, 0.0808, 0.1781, 0.1283, 0.1231, 0.0945, 0.0479, 0.0172, -0.0625, -0.0727, -0.0214, -0.062, -0.0251, -0.0251, -0.0101, 0.0, -0.0206, -0.0098, -0.0172, -0.0081, 0.0061, 0.0085, 0.0076, 0.0076, -0.0409, -0.0554, -0.0482, -0.0985, -0.0851, -0.0288, -0.0219, -0.0221, -0.0215, -0.0215, -0.0217, -0.0189, -0.0201, -0.0255, -0.0255, 0.0169, 0.006, -0.0246, -0.0443, -0.0686, -0.0755, -0.0549, 0.0055, 0.0175, 0.0371, 0.0549, 0.0549, 0.0553, 0.1367, 0.1546, 0.1817, 0.1436, 0.1468, 0.1501, 0.1475, 0.0772, 0.1206, 0.1017, -0.0891, -0.124, -0.0515, -0.0171, -0.0282, -0.0281, -0.0629, -0.0566, -0.0566, -0.0759, -0.0894, -0.0746, -0.0892, -0.1054, -0.0269, -0.0405, -0.0732, -0.089, -0.0962, -0.1311, -0.0805, -0.0924, -0.1244, -0.0777, -0.0752, -0.0679, -0.1002, -0.12, -0.1055, -0.1067, -0.0924, -0.0659, -0.0839, -0.0777, -0.0655, 0.0023, -0.0728, -0.0809, -0.1079, -0.0743, -0.0435, -0.0103, -0.0103, 0.0019, 0.0081, 0.0549, 0.0768, 0.115, 0.1232, 0.1607, 0.1746, 0.2662, 0.2668, 0.2511, 0.2945, 0.3809, 0.3894, 0.5827, 0.5227, 0.4707, 0.534, 0.6131, 0.6406, 0.5542, 0.6166, 0.9175, 0.8587, 1.0713, 1.0877, 1.0399, 1.0272, 1.0463, 0.9298, 0.9157, 0.8841, 1.0367, 1.0458, 1.1274, 1.1274, 1.2929]}
{"unit_id": "quote:1995319574393442673", "ticker_idx": 3, "ts": "2025-12-01T02:30:32+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Apple's memory and NAND capacities are locked in via exclusive long-term contracts at fixed prices. This leads to Apple's cost advantage expanding further. Apple's growth in global mobile phone market share could be very significant; the coming period might be a glorious time for the iPhone.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish semis through 2030 on AI-driven profit reallocation; NVDA structurally dominant, SK Hynix HBM winner, AMD CPU resurgence from agentic AI, Apple beneficiary of memory cost advantage.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Looking back at the semiconductor market in 2025, there were truly too many fascinating stories. The overarching theme was:\n\nAI demand-driven valuation restructuring of semiconductor infrastructure + The rewriting of value distribution across the supply chain.\n\nStarting in 2024, semiconductor infrastructure began rapidly devouring the profits of the entire IT industry. The share of semiconductor net profit (EPS) within the IT sector of the S&P 500 rose from less than 20% to 40% in just two years, and it is still in a state of accelerated ascent.\n\nThe overall forward-looking profit margin for semiconductors has risen from 25% in 2023 to 43% by November 2025, significantly surpassing the average profit margins of several internet giants. This confirms that semiconductor margins exceeding those of the internet sector will be the \"New Normal.\" The profit distribution of the entire IT industry is flowing increasingly toward semiconductors.\n\nIt is worth noting that even during the semiconductor chip shortage of 2020–2022, when shortages were severe, semiconductor profit margins and their share of total IT profit distribution did not see such significant growth.\n\nThis is the first half of the story: The valuation system of semiconductor infrastructure is being reconstructed by AI demand; it no longer holds the subordinate infrastructure status it did during the Internet era.\n\nThe logic behind this phenomenon is the transformation of business models alongside technological characteristics:\n\nIn the Internet era, the network and compute costs for each request had extremely low marginal costs. Scaling effects were excellent, and the marginal cost of distribution was almost zero.\n\nIn the AI era, this \"near-zero marginal distribution cost\" characteristic favorable to scalability has encountered a fundamental and major challenge: Not only is training cost no longer a one-time expense but rather one that grows annually, but even for client-side AI inference requests, costs do not drop in sync with hardware compute deflation. This is because \"inference scaling\" has become a consensus, and vertical domains still require larger-scale flagship models to maintain competitiveness.\n\nPreviously, the biggest cost for internet companies was OPEX, specifically SDE labor costs. Now, for the first time in history, internet companies are burdened with balance sheets full of high depreciation costs, similar to semiconductor foundries. The business model is practically itching to shift slowly from \"Traffic × Conversion Rate\" to \"Gross Profit per Token.\"\n\nSimply put, the cost distribution shift from the Internet era to the AI era has added heavy hardware/compute CAPEX (MSFT 33%, Meta 38% in earnings reports) on top of the OPEX base. The winners of the previous era were Internet companies + CSPs + SaaS (the \"rent-collecting\" industries). In the AI era, compute (semiconductor/chip depreciation) has become the new rent-collecting industry. The profit distribution of the entire IT industry has undergone a drastic reallocation (EPS flow to semiconductors rising from 20% to 40% and climbing). This is the most important reason for the restructuring of the semiconductor infrastructure valuation system.\n\n--------------- How long can the \"New Normal\" of high semiconductor margins last?\n\nThe current high premiums come from an excessive backlog of semiconductor orders caused by an earlier \"cost-is-no-object\" arms race. However, obviously, hyperscalers do not want to be taken for a ride; they are all attempting to build in-house ASICs to lower costs. We can look at this problem from the perspective of the long-term compute distribution in 2030.\n\nIn the long run, OpenAI has already revealed the \"standard answer\": 10GW Nvidia, 10GW ASIC, 6GW AMD. Other hyperscalers have similar considerations for their split.\n\nFor example, on the inference side, they hope for ASIC >50%. Within the GPU segment, a 50/50 split between AMD and NV (legacy). For training, NV will still hold the lion's share (60%+), with the rest split between in-house ASICs and AMD.\n\nCalculated based on a 2030 split of 60% inference and 40% training, the market share comes out to NV 38%, ASIC 39%, and AMD 23%. This is almost entirely consistent with OpenAI’s ratios and can be considered a standard reference value.\n\nOf course, for Microsoft, Amazon, Google, and Anthropic, the AMD ratio will be noticeably lower than this standard answer/reference value, while xAI has no ASICs and uses only Nvidia + a small amount of AMD.\n\nAMD's risk lies in the ultra-long term (post-2030). As CSP in-house ASICs become more mature (except for Microsoft), the proportion of ASICs in inference may grow higher, making it hard to find incentives to buy large quantities of new GPUs unless they are sold cheaply enough.\n\nWhat about the recently hyped TPU? Is Meta shifting to TPUs? Will it significantly impact Nvidia’s margins?\n\nActually, Meta’s CAPEX is $72B this year, $110B next year, and the average annual CAPEX over the next six years may reach around $160B. Meanwhile, Meta’s 6-year $10B TPU order amounts to only $1.6B per year on average—and they are purchasing TPU cloud services, not bare TPUs.\n\nIn other words, this TPU order accounts for only 1% of Meta’s future 6-year CAPEX. They are not seriously considering large-scale deployment; it is likely just a means to bargain with Nvidia.\n\nFurthermore, looking at Meta’s recruitment ads over the recent months, we don’t see any hiring for TPU engineers, unlike Anthropic, which was hiring a bunch of TPU kernel engineers back in May before announcing large-scale TPU procurement for training in October.\n\nSo, whether the reason is diversifying suppliers, creating a fallback for delayed in-house ASICs, or delays with AMD’s MI350X, Meta’s purchase of TPUs basically has one consideration: increasing bargaining power when buying Nvidia GPUs. However, they can essentially only bargain on the inference share; the actual effect is very limited, as is the impact on Nvidia’s profit margins.\n\nKeep in mind that during the crypto bear market crash of 2022, NVDA inventory rose to 198 days, yet margins only retreated from 65% to 56%. Even with the \"double kill\" of PE and macroeconomics, the stock price only went from 300 to 100. Now, with supply consistently falling short of demand, there is no reason for margins to come down.\n\nAdditionally, the TPU v8 design is too conservative (it didn't use HBM4), and the TCO (Total Cost of Ownership) of the Kyber rack Rubin solution will be better than TPU v8. In the end, they will still have to rely on Nvidia, making it hard to bargain. As long as Nvidia continues its stride forward, it is actually not easy for competitors to keep up.\n\nIn summary, on one hand, bottlenecks across the entire supply chain (e.g., CoWoS expansion) are cautious, so the state of demand exceeding supply can last for many years.\n\nOn the other hand, there is a \"temporal mismatch\" between the profit curve of AI monetization and the hardware investment curve. The growth curve of the application side will lag by a few years. As long as this time lag between the application side and infrastructure side growth curves exists, semiconductors will continue to dominate profit distribution in the IT industry.\n\nLooking at OpenAI’s investment curve through 2030, this mismatch will last until around 2030. This means the dominant position in IT industry profit distribution brought about by the semiconductor industry’s super-expansion phase will last at least until 2030 based on current views.\n\nAnd high semiconductor margins might be sustained even longer, because the attribute has changed from \"one-time infrastructure\" in the internet era to \"rent-collecting infrastructure\" now.\n\nAI has not only fed the GPU; it has used the compute budget to lift every link in the chain that \"turns electricity into tokens\"—from memory, storage, interconnects, and fiber optics, to electricity, energy storage, and so on.\n\nThe first half covered \"Semiconductors swallowing IT profits.\" The second half covers how the \"AI Compute Value Spillover Effect\" reshapes the internal landscape of semiconductors: GPU Compute Growth -> Memory/Storage/Interconnect/CPU Bottlenecks -> Spillover Effect -> Structural Opportunities.\n\nThe more interesting story in 2025 is how huge industry dividends are creating structural new opportunities within the semiconductor sector. For instance, a super cluster requires several data center interconnects; fiber optic interconnect lengths reach the level of millions of miles. This is a new opportunity.\n\nThe most typical examples of new opportunities brought by structural trends in the semiconductor supply chain are Memory (DRAM/HBM) and Storage (SSD). The growth in HBM demand is so exaggerated that it is squeezing DDR4/5 capacity, leading the memory industry—historically known for its cyclical nature—to shout that \"the cycle no longer exists.\" SK Hynix, leading in HBM, is even envisioning annual profits of $100 billion a few years from now, firmly becoming a trillion-dollar market cap company.\n\nBehind these two sectors is a shift in structural trends: AI workloads are gradually extending from training to inference, and the proportion of inference is getting larger.\n\nInference is a purely memory-bound task; one could say bandwidth speed = tokens/s. As model sizes get larger and context lengths increase, the requirements for memory size also grow accordingly, leading to a surge in memory demand: Inference is Memory.\n\nThe memory on next-generation GPUs/ASICs has become an aesthetic of brute force. The sheer size of equipped memory is something unimaginable three years ago. Looking back, the 80GB of the H100 in 2022 looks like a toy; it has grown tenfold in just a few years:\n\n- Nvidia Ultra Rubin - 1024GB HBM\n\n- Qualcomm AI200 - 768GB LPDDR\n\n- AMD MI400x - 432GB HBM\n\nAnother potential explosion point for memory is on the Edge, namely edge LLMs for phones/PCs/cars/robots. In the last two years, mainstream flagship phones have upgraded from 6GB to 8GB/12GB/16GB, preparing in advance for a possible edge LLM ecosystem. After all, next-gen phone compute will reach the 150 TOPS level—solidly desktop class and very powerful. In terms of potential, the edge memory upgrade is a larger market than the cloud memory increment. The number of edge terminal devices is staggering—billions annually. Once the edge LLM ecosystem thrives, doubling memory usage will be easy, and various designs for edge low-power memory/processing-in-memory will follow.\nHowever, the software ecosystem for edge GenAI seems to lag significantly, progressing much slower than I imagined. This is likely because this area is still in a groping phase and does not have the clearly defined ROI of the cloud. Manufacturers are cautious with investments. When I was bullish on 2027 back in '23–'24, I may have been too optimistic.\nInternet -> Mobile Internet took 10–15 years. Edge GenAI/LLM might also need 7–10 years. It may be that we have to wait until the cloud ROI is mostly exploited and marginal returns diminish before edge GenAI/LLM gets the development resources to validate the edge ROI.\n\nAnother story of internal structural shifts in semiconductors in 2025 is NAND storage, specifically enterprise eSSD.\n\nThe source of the structural trend is the same: the growing demand for AI inference workloads. The memory dividend has also spilled over into SSD storage, and even HDD storage, because when memory isn't enough, high-speed SSDs are used as multi-level caches. The main logic is KV cache offloading to the next layer of SSD storage during AI inference memory overflow, as well as vector database search/indexing—all increasing the demand for SSD storage.\n\nMicron's earnings report put it precisely and bluntly: \"AI inference use cases such as KV cache tiering and vector database search and indexing, are driving demand for performance storage.\"\n\nAs for why storage prices only exploded in the fourth quarter, one needs to distinguish between contract prices and spot prices. Contract price increases are milder; even for the scarcest enterprise eSSDs, contracts rose about 25% in Q4. But as NAND capacity was slowly eaten up by contracts in 2025, spot prices created a strong visual shock, rising over 50% in a single month.\n\nAnother unverified logic is the explosion of multimodal AI, especially the demand for AI images and AI video, which would exacerbate storage shortages. I feel this storyline is \"promising for the future,\" but the current precision of video/images is perhaps not even at the GPT-3 level yet; achieving a breakout effect will still take some time.\n\n------------------------ So, what are the next semiconductor structural opportunities brought by trend shifts?\n\nWe must look at what the trend for AI inference demand is next. Undoubtedly, the proportion of agentic flow will get larger and larger. 2025 is not just the \"year of the agent,\" but the start of the \"decade of the agent.\"\n\nViewing agentic workloads from a CPU perspective: routing and tool processing happen on the CPU. If you profile common agentic frameworks like SWE-Agent, LangChain, or Toolformer, the CPU can account for up to 90% of end-to-end latency. Throughput bottlenecks are also stuck more on the CPU, and CPU energy consumption can even exceed 40% of the total.\n\nAgentic AI is currently more of a CPU bottleneck issue. In agentic frameworks, the CPU is the \"Orchestrator\" that is always busy. This will likely drive a new wave of recovery in CPU demand.\n\nIn AMD's Q2 2025 earnings report (August 5), Lisa Su explicitly stated this phenomenon: \"In particular, adoption of agentic AI is creating additional demand for general-purpose compute infrastructure, as customers quickly realize that each token generated by a GPU triggers multiple CPU-intensive tasks.\"\n\nIn the Q3 report, Lisa played her hand again regarding CPU TAM increasing due to Gen AI: \"Many customers are now planning substantially larger CPU build outs over the coming quarters to support increased demands from AI, serving as a powerful new catalyst for our server business.\"\n\nNvidia also views agent flow as a CPU demand. The CPU ratio in the GB200/300 architecture configurations is much larger than before—36 Grace CPUs : 72 Blackwell GPUs, reaching a 1:2 level. AMD's route involves using 1~4 256-core EPYCs to serve the MI400 series. Future hardware architectures for 72~128 GPUs will definitely develop towards optimizing agent workloads, such as agent task graph scheduling, load balancing, and CPU/GPU cooperative micro-batching.\n\nComparisons in compute power might eventually move away from the current pure GPU token rate comparison toward a whole-system full-stack agentic benchmark comparison.\n\n-------------------------- While semiconductor structural shifts bring opportunities, the next steps may also bring unexpected secondary effects.\n\nThe explosion in cloud AI data center demand has caused memory and storage prices to skyrocket, putting immense pressure on costs for consumer electronics. In 2026, this may evolve into a potential Black Swan event for the consumer electronics industry.\n\nThe recent plunge in PC manufacturer stocks is due to this reason. HP has already stated it will reduce memory configurations, implying a return to the era of 8GB RAM + 256GB storage for PCs.\n\nIf DRAM memory and storage prices continue to rise like this, we might see a ridiculous situation: spot prices for memory/storage becoming more expensive than CPUs and GPUs. The awkward part is that this could directly delay the progress of AI PCs that consumer electronics hopes for, as large memory is more favorable for AI PC performance.\n\nTo exaggerate a bit: every employee of PC and mobile phone manufacturers, and even employees of consumer electronics manufacturers, should buy storage and memory stocks as a career risk hedge.\n\nStarting early next year, memory and storage costs for the Android camp will become uncontainable. If Samsung and Xiaomi raise phone prices (prices in the US market have already risen significantly), the biggest beneficiary will be Apple.\n\nApple's memory and NAND capacities are locked in via exclusive long-term contracts at fixed prices. Incidentally, they also trapped Kioxia for a lot of non-increasing price capacity. This leads to Apple's cost advantage expanding further. Apple's growth in global mobile phone market share could be very significant; the coming period might be a glorious time for the iPhone.\n\nThe 2025 semiconductor market is truly full of wonderful stories. The grudges and relationships between Nvidia/AMD/TPU and various hyperscalers have caused the moods of onlookers placing their bets to fluctuate wildly.\n\nHBM/Memory manufacturers ate the \"memory-bound\" dividends; NAND manufacturers unexpectedly harvested the spillover effects of KV cache; and CPUs, after nearly a decade of silence, may return to the center of the growth narrative due to agent orchestration.\n\nIt is no longer just Nvidia/Broadcom and a few compute vendors dominating. Rather, every evolution following the value spillover of AI workloads—from training to inference, from text to multimodal, from single model calls to agentic flow—is rewriting the value distribution of the industry chain.\n\nThe boom in Cloud AI is squeezing the living space of consumer electronics—while PC manufacturers are forced to discuss returning to the \"8GB era,\" Apple sits back and reaps the benefits due to supply chain advantages. The secondary effects of this compute arms race may reshape the entire consumer electronics landscape in 2026 in unexpected ways.\n\nThe story of semiconductors is no longer a single thread, but a web that continuously reconstructs itself. And 2025 is likely just the first round of these alliances and rivalries.", "reply_text": null, "parent_text": null, "quoted_text": "回顾2025年半导体市场，真的是有太多太多精彩的故事，最大的主题就是:\n\nAI需求驱动导致半导体基建的估值体系重构 + 产业链的价值分配重写\n\n从2024年开始，半导体基建正在飞速吞噬整个IT产业利润，SP500里半导体净利润EPS在IT行业里占比，在两年时间从不到20%上升了到了40%，而且还在呈加速上升姿态 https://t.co/OAFOULgBFX", "parent_author": null, "ret_m1d": -0.01501239062757509, "ret_signed_m1d": -0.01501239062757509, "alpha_spy_m1d": -0.019598744843221216, "alpha_c_m1d": -0.014558305291592522, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.00045408533598256806, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010914832876105063, "ret_signed_p1d": 0.010914832876105063, "alpha_spy_p1d": 0.009062623601993058, "alpha_c_p1d": 0.0006127918538925137, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01030204102221255, "ret_p1w": -0.018403424491690523, "ret_signed_p1w": -0.018403424491690523, "alpha_spy_p1w": -0.023342589067436914, "alpha_c_p1w": -0.04951911718503099, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.03111569269334047, "ret_p1m": -0.035393909206820506, "ret_signed_p1m": -0.035393909206820506, "alpha_spy_p1m": -0.048285924603067576, "alpha_c_p1m": -0.05254463433269052, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.017150725125870014, "ret_p3m": -0.03495090080175667, "ret_signed_p3m": -0.03495090080175667, "alpha_spy_p3m": -0.05121906771552587, "alpha_c_p3m": -0.021323317890839855, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": -0.013627582910916813, "ret_p6m": 0.09114384942984533, "ret_signed_p6m": 0.09114384942984533, "alpha_spy_p6m": -0.018509430023661455, "alpha_c_p6m": -0.20553054456019804, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.29667439399004336, "price_path": [-0.1585, -0.1538, -0.1542, -0.1605, -0.173, -0.1997, -0.1882, -0.174, -0.1647, -0.1596, -0.1566, -0.1605, -0.1337, -0.0963, -0.1021, -0.1096, -0.0935, -0.0985, -0.1021, -0.1014, -0.0985, -0.0926, -0.0895, -0.0942, -0.0949, -0.0893, -0.1035, -0.1345, -0.126, -0.1256, -0.1201, -0.1268, -0.1097, -0.0746, -0.0727, -0.088, -0.084, -0.0725, -0.0514, -0.0507, -0.0483, -0.0423, -0.0459, -0.0505, -0.0471, -0.0467, -0.048, -0.0526, -0.0483, -0.0277, -0.034, -0.0359, -0.0378, -0.0552, -0.0553, -0.0514, -0.0595, -0.041, -0.0254, -0.0217, -0.0196, -0.0196, -0.015, 0.0, 0.0109, 0.0037, -0.0085, -0.0153, -0.0184, -0.0209, -0.0153, -0.0179, -0.017, -0.0318, -0.03, -0.0398, -0.0385, -0.0333, -0.0428, -0.0379, -0.0328, -0.0328, -0.0343, -0.033, -0.0354, -0.0397, -0.0397, -0.0427, -0.056, -0.0733, -0.0804, -0.085, -0.0838, -0.0807, -0.0779, -0.0817, -0.0879, -0.0974, -0.0974, -0.1286, -0.1252, -0.1227, -0.1238, -0.0978, -0.0877, -0.0942, -0.0877, -0.0834, -0.0462, -0.0481, -0.0233, -0.0254, -0.0176, -0.029, -0.0324, -0.0259, -0.0746, -0.0957, -0.0957, -0.067, -0.0654, -0.0787, -0.0645, -0.0589, -0.0378, -0.0304, -0.035, -0.066, -0.064, -0.0675, -0.0718, -0.0797, -0.0897, -0.0812, -0.0778, -0.0779, -0.0957, -0.1157, -0.1061, -0.1011, -0.1163, -0.1198, -0.1232, -0.1108, -0.1103, -0.1068, -0.1059, -0.1203, -0.128, -0.1027, -0.0962, -0.0952, -0.0952, -0.0848, -0.1037, -0.0846, -0.079, -0.079, -0.0836, -0.0849, -0.058, -0.0687, -0.0446, -0.0346, -0.0589, -0.0342, -0.0333, -0.0416, -0.0538, -0.0429, -0.0448, -0.0406, -0.0095, -0.0212, 0.0048, 0.0165, 0.0163, 0.0371, 0.0358, 0.0433, 0.0577, 0.0553, 0.0625, 0.054, 0.058, 0.0696, 0.0793, 0.0929, 0.0929, 0.0911]}
{"unit_id": "quote:1995319574393442673", "ticker_idx": 4, "ts": "2025-12-01T02:30:32+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the dominant position in IT industry profit distribution brought about by the semiconductor industry's super-expansion phase will last at least until 2030 based on current views", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish semis through 2030 on AI-driven profit reallocation; NVDA structurally dominant, SK Hynix HBM winner, AMD CPU resurgence from agentic AI, Apple beneficiary of memory cost advantage.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "US semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "Looking back at the semiconductor market in 2025, there were truly too many fascinating stories. The overarching theme was:\n\nAI demand-driven valuation restructuring of semiconductor infrastructure + The rewriting of value distribution across the supply chain.\n\nStarting in 2024, semiconductor infrastructure began rapidly devouring the profits of the entire IT industry. The share of semiconductor net profit (EPS) within the IT sector of the S&P 500 rose from less than 20% to 40% in just two years, and it is still in a state of accelerated ascent.\n\nThe overall forward-looking profit margin for semiconductors has risen from 25% in 2023 to 43% by November 2025, significantly surpassing the average profit margins of several internet giants. This confirms that semiconductor margins exceeding those of the internet sector will be the \"New Normal.\" The profit distribution of the entire IT industry is flowing increasingly toward semiconductors.\n\nIt is worth noting that even during the semiconductor chip shortage of 2020–2022, when shortages were severe, semiconductor profit margins and their share of total IT profit distribution did not see such significant growth.\n\nThis is the first half of the story: The valuation system of semiconductor infrastructure is being reconstructed by AI demand; it no longer holds the subordinate infrastructure status it did during the Internet era.\n\nThe logic behind this phenomenon is the transformation of business models alongside technological characteristics:\n\nIn the Internet era, the network and compute costs for each request had extremely low marginal costs. Scaling effects were excellent, and the marginal cost of distribution was almost zero.\n\nIn the AI era, this \"near-zero marginal distribution cost\" characteristic favorable to scalability has encountered a fundamental and major challenge: Not only is training cost no longer a one-time expense but rather one that grows annually, but even for client-side AI inference requests, costs do not drop in sync with hardware compute deflation. This is because \"inference scaling\" has become a consensus, and vertical domains still require larger-scale flagship models to maintain competitiveness.\n\nPreviously, the biggest cost for internet companies was OPEX, specifically SDE labor costs. Now, for the first time in history, internet companies are burdened with balance sheets full of high depreciation costs, similar to semiconductor foundries. The business model is practically itching to shift slowly from \"Traffic × Conversion Rate\" to \"Gross Profit per Token.\"\n\nSimply put, the cost distribution shift from the Internet era to the AI era has added heavy hardware/compute CAPEX (MSFT 33%, Meta 38% in earnings reports) on top of the OPEX base. The winners of the previous era were Internet companies + CSPs + SaaS (the \"rent-collecting\" industries). In the AI era, compute (semiconductor/chip depreciation) has become the new rent-collecting industry. The profit distribution of the entire IT industry has undergone a drastic reallocation (EPS flow to semiconductors rising from 20% to 40% and climbing). This is the most important reason for the restructuring of the semiconductor infrastructure valuation system.\n\n--------------- How long can the \"New Normal\" of high semiconductor margins last?\n\nThe current high premiums come from an excessive backlog of semiconductor orders caused by an earlier \"cost-is-no-object\" arms race. However, obviously, hyperscalers do not want to be taken for a ride; they are all attempting to build in-house ASICs to lower costs. We can look at this problem from the perspective of the long-term compute distribution in 2030.\n\nIn the long run, OpenAI has already revealed the \"standard answer\": 10GW Nvidia, 10GW ASIC, 6GW AMD. Other hyperscalers have similar considerations for their split.\n\nFor example, on the inference side, they hope for ASIC >50%. Within the GPU segment, a 50/50 split between AMD and NV (legacy). For training, NV will still hold the lion's share (60%+), with the rest split between in-house ASICs and AMD.\n\nCalculated based on a 2030 split of 60% inference and 40% training, the market share comes out to NV 38%, ASIC 39%, and AMD 23%. This is almost entirely consistent with OpenAI’s ratios and can be considered a standard reference value.\n\nOf course, for Microsoft, Amazon, Google, and Anthropic, the AMD ratio will be noticeably lower than this standard answer/reference value, while xAI has no ASICs and uses only Nvidia + a small amount of AMD.\n\nAMD's risk lies in the ultra-long term (post-2030). As CSP in-house ASICs become more mature (except for Microsoft), the proportion of ASICs in inference may grow higher, making it hard to find incentives to buy large quantities of new GPUs unless they are sold cheaply enough.\n\nWhat about the recently hyped TPU? Is Meta shifting to TPUs? Will it significantly impact Nvidia’s margins?\n\nActually, Meta’s CAPEX is $72B this year, $110B next year, and the average annual CAPEX over the next six years may reach around $160B. Meanwhile, Meta’s 6-year $10B TPU order amounts to only $1.6B per year on average—and they are purchasing TPU cloud services, not bare TPUs.\n\nIn other words, this TPU order accounts for only 1% of Meta’s future 6-year CAPEX. They are not seriously considering large-scale deployment; it is likely just a means to bargain with Nvidia.\n\nFurthermore, looking at Meta’s recruitment ads over the recent months, we don’t see any hiring for TPU engineers, unlike Anthropic, which was hiring a bunch of TPU kernel engineers back in May before announcing large-scale TPU procurement for training in October.\n\nSo, whether the reason is diversifying suppliers, creating a fallback for delayed in-house ASICs, or delays with AMD’s MI350X, Meta’s purchase of TPUs basically has one consideration: increasing bargaining power when buying Nvidia GPUs. However, they can essentially only bargain on the inference share; the actual effect is very limited, as is the impact on Nvidia’s profit margins.\n\nKeep in mind that during the crypto bear market crash of 2022, NVDA inventory rose to 198 days, yet margins only retreated from 65% to 56%. Even with the \"double kill\" of PE and macroeconomics, the stock price only went from 300 to 100. Now, with supply consistently falling short of demand, there is no reason for margins to come down.\n\nAdditionally, the TPU v8 design is too conservative (it didn't use HBM4), and the TCO (Total Cost of Ownership) of the Kyber rack Rubin solution will be better than TPU v8. In the end, they will still have to rely on Nvidia, making it hard to bargain. As long as Nvidia continues its stride forward, it is actually not easy for competitors to keep up.\n\nIn summary, on one hand, bottlenecks across the entire supply chain (e.g., CoWoS expansion) are cautious, so the state of demand exceeding supply can last for many years.\n\nOn the other hand, there is a \"temporal mismatch\" between the profit curve of AI monetization and the hardware investment curve. The growth curve of the application side will lag by a few years. As long as this time lag between the application side and infrastructure side growth curves exists, semiconductors will continue to dominate profit distribution in the IT industry.\n\nLooking at OpenAI’s investment curve through 2030, this mismatch will last until around 2030. This means the dominant position in IT industry profit distribution brought about by the semiconductor industry’s super-expansion phase will last at least until 2030 based on current views.\n\nAnd high semiconductor margins might be sustained even longer, because the attribute has changed from \"one-time infrastructure\" in the internet era to \"rent-collecting infrastructure\" now.\n\nAI has not only fed the GPU; it has used the compute budget to lift every link in the chain that \"turns electricity into tokens\"—from memory, storage, interconnects, and fiber optics, to electricity, energy storage, and so on.\n\nThe first half covered \"Semiconductors swallowing IT profits.\" The second half covers how the \"AI Compute Value Spillover Effect\" reshapes the internal landscape of semiconductors: GPU Compute Growth -> Memory/Storage/Interconnect/CPU Bottlenecks -> Spillover Effect -> Structural Opportunities.\n\nThe more interesting story in 2025 is how huge industry dividends are creating structural new opportunities within the semiconductor sector. For instance, a super cluster requires several data center interconnects; fiber optic interconnect lengths reach the level of millions of miles. This is a new opportunity.\n\nThe most typical examples of new opportunities brought by structural trends in the semiconductor supply chain are Memory (DRAM/HBM) and Storage (SSD). The growth in HBM demand is so exaggerated that it is squeezing DDR4/5 capacity, leading the memory industry—historically known for its cyclical nature—to shout that \"the cycle no longer exists.\" SK Hynix, leading in HBM, is even envisioning annual profits of $100 billion a few years from now, firmly becoming a trillion-dollar market cap company.\n\nBehind these two sectors is a shift in structural trends: AI workloads are gradually extending from training to inference, and the proportion of inference is getting larger.\n\nInference is a purely memory-bound task; one could say bandwidth speed = tokens/s. As model sizes get larger and context lengths increase, the requirements for memory size also grow accordingly, leading to a surge in memory demand: Inference is Memory.\n\nThe memory on next-generation GPUs/ASICs has become an aesthetic of brute force. The sheer size of equipped memory is something unimaginable three years ago. Looking back, the 80GB of the H100 in 2022 looks like a toy; it has grown tenfold in just a few years:\n\n- Nvidia Ultra Rubin - 1024GB HBM\n\n- Qualcomm AI200 - 768GB LPDDR\n\n- AMD MI400x - 432GB HBM\n\nAnother potential explosion point for memory is on the Edge, namely edge LLMs for phones/PCs/cars/robots. In the last two years, mainstream flagship phones have upgraded from 6GB to 8GB/12GB/16GB, preparing in advance for a possible edge LLM ecosystem. After all, next-gen phone compute will reach the 150 TOPS level—solidly desktop class and very powerful. In terms of potential, the edge memory upgrade is a larger market than the cloud memory increment. The number of edge terminal devices is staggering—billions annually. Once the edge LLM ecosystem thrives, doubling memory usage will be easy, and various designs for edge low-power memory/processing-in-memory will follow.\nHowever, the software ecosystem for edge GenAI seems to lag significantly, progressing much slower than I imagined. This is likely because this area is still in a groping phase and does not have the clearly defined ROI of the cloud. Manufacturers are cautious with investments. When I was bullish on 2027 back in '23–'24, I may have been too optimistic.\nInternet -> Mobile Internet took 10–15 years. Edge GenAI/LLM might also need 7–10 years. It may be that we have to wait until the cloud ROI is mostly exploited and marginal returns diminish before edge GenAI/LLM gets the development resources to validate the edge ROI.\n\nAnother story of internal structural shifts in semiconductors in 2025 is NAND storage, specifically enterprise eSSD.\n\nThe source of the structural trend is the same: the growing demand for AI inference workloads. The memory dividend has also spilled over into SSD storage, and even HDD storage, because when memory isn't enough, high-speed SSDs are used as multi-level caches. The main logic is KV cache offloading to the next layer of SSD storage during AI inference memory overflow, as well as vector database search/indexing—all increasing the demand for SSD storage.\n\nMicron's earnings report put it precisely and bluntly: \"AI inference use cases such as KV cache tiering and vector database search and indexing, are driving demand for performance storage.\"\n\nAs for why storage prices only exploded in the fourth quarter, one needs to distinguish between contract prices and spot prices. Contract price increases are milder; even for the scarcest enterprise eSSDs, contracts rose about 25% in Q4. But as NAND capacity was slowly eaten up by contracts in 2025, spot prices created a strong visual shock, rising over 50% in a single month.\n\nAnother unverified logic is the explosion of multimodal AI, especially the demand for AI images and AI video, which would exacerbate storage shortages. I feel this storyline is \"promising for the future,\" but the current precision of video/images is perhaps not even at the GPT-3 level yet; achieving a breakout effect will still take some time.\n\n------------------------ So, what are the next semiconductor structural opportunities brought by trend shifts?\n\nWe must look at what the trend for AI inference demand is next. Undoubtedly, the proportion of agentic flow will get larger and larger. 2025 is not just the \"year of the agent,\" but the start of the \"decade of the agent.\"\n\nViewing agentic workloads from a CPU perspective: routing and tool processing happen on the CPU. If you profile common agentic frameworks like SWE-Agent, LangChain, or Toolformer, the CPU can account for up to 90% of end-to-end latency. Throughput bottlenecks are also stuck more on the CPU, and CPU energy consumption can even exceed 40% of the total.\n\nAgentic AI is currently more of a CPU bottleneck issue. In agentic frameworks, the CPU is the \"Orchestrator\" that is always busy. This will likely drive a new wave of recovery in CPU demand.\n\nIn AMD's Q2 2025 earnings report (August 5), Lisa Su explicitly stated this phenomenon: \"In particular, adoption of agentic AI is creating additional demand for general-purpose compute infrastructure, as customers quickly realize that each token generated by a GPU triggers multiple CPU-intensive tasks.\"\n\nIn the Q3 report, Lisa played her hand again regarding CPU TAM increasing due to Gen AI: \"Many customers are now planning substantially larger CPU build outs over the coming quarters to support increased demands from AI, serving as a powerful new catalyst for our server business.\"\n\nNvidia also views agent flow as a CPU demand. The CPU ratio in the GB200/300 architecture configurations is much larger than before—36 Grace CPUs : 72 Blackwell GPUs, reaching a 1:2 level. AMD's route involves using 1~4 256-core EPYCs to serve the MI400 series. Future hardware architectures for 72~128 GPUs will definitely develop towards optimizing agent workloads, such as agent task graph scheduling, load balancing, and CPU/GPU cooperative micro-batching.\n\nComparisons in compute power might eventually move away from the current pure GPU token rate comparison toward a whole-system full-stack agentic benchmark comparison.\n\n-------------------------- While semiconductor structural shifts bring opportunities, the next steps may also bring unexpected secondary effects.\n\nThe explosion in cloud AI data center demand has caused memory and storage prices to skyrocket, putting immense pressure on costs for consumer electronics. In 2026, this may evolve into a potential Black Swan event for the consumer electronics industry.\n\nThe recent plunge in PC manufacturer stocks is due to this reason. HP has already stated it will reduce memory configurations, implying a return to the era of 8GB RAM + 256GB storage for PCs.\n\nIf DRAM memory and storage prices continue to rise like this, we might see a ridiculous situation: spot prices for memory/storage becoming more expensive than CPUs and GPUs. The awkward part is that this could directly delay the progress of AI PCs that consumer electronics hopes for, as large memory is more favorable for AI PC performance.\n\nTo exaggerate a bit: every employee of PC and mobile phone manufacturers, and even employees of consumer electronics manufacturers, should buy storage and memory stocks as a career risk hedge.\n\nStarting early next year, memory and storage costs for the Android camp will become uncontainable. If Samsung and Xiaomi raise phone prices (prices in the US market have already risen significantly), the biggest beneficiary will be Apple.\n\nApple's memory and NAND capacities are locked in via exclusive long-term contracts at fixed prices. Incidentally, they also trapped Kioxia for a lot of non-increasing price capacity. This leads to Apple's cost advantage expanding further. Apple's growth in global mobile phone market share could be very significant; the coming period might be a glorious time for the iPhone.\n\nThe 2025 semiconductor market is truly full of wonderful stories. The grudges and relationships between Nvidia/AMD/TPU and various hyperscalers have caused the moods of onlookers placing their bets to fluctuate wildly.\n\nHBM/Memory manufacturers ate the \"memory-bound\" dividends; NAND manufacturers unexpectedly harvested the spillover effects of KV cache; and CPUs, after nearly a decade of silence, may return to the center of the growth narrative due to agent orchestration.\n\nIt is no longer just Nvidia/Broadcom and a few compute vendors dominating. Rather, every evolution following the value spillover of AI workloads—from training to inference, from text to multimodal, from single model calls to agentic flow—is rewriting the value distribution of the industry chain.\n\nThe boom in Cloud AI is squeezing the living space of consumer electronics—while PC manufacturers are forced to discuss returning to the \"8GB era,\" Apple sits back and reaps the benefits due to supply chain advantages. The secondary effects of this compute arms race may reshape the entire consumer electronics landscape in 2026 in unexpected ways.\n\nThe story of semiconductors is no longer a single thread, but a web that continuously reconstructs itself. And 2025 is likely just the first round of these alliances and rivalries.", "reply_text": null, "parent_text": null, "quoted_text": "回顾2025年半导体市场，真的是有太多太多精彩的故事，最大的主题就是:\n\nAI需求驱动导致半导体基建的估值体系重构 + 产业链的价值分配重写\n\n从2024年开始，半导体基建正在飞速吞噬整个IT产业利润，SP500里半导体净利润EPS在IT行业里占比，在两年时间从不到20%上升了到了40%，而且还在呈加速上升姿态 https://t.co/OAFOULgBFX", "parent_author": null, "ret_m1d": -0.001926553330782843, "ret_signed_m1d": -0.001926553330782843, "alpha_spy_m1d": -0.00651290754642897, "alpha_c_m1d": -0.00651290754642897, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": 0.004586354215646127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01833066100996228, "ret_signed_p1d": 0.01833066100996228, "alpha_spy_p1d": 0.016478451735850275, "alpha_c_p1d": 0.016478451735850275, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.0018522092741120044, "ret_p1w": 0.04416926852118719, "ret_signed_p1w": 0.04416926852118719, "alpha_spy_p1w": 0.0392301039454408, "alpha_c_p1w": 0.0392301039454408, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.004939164575746391, "ret_p1m": 0.032526253224546275, "ret_signed_p1m": 0.032526253224546275, "alpha_spy_p1m": 0.019634237828299206, "alpha_c_p1m": 0.019634237828299206, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.01289201539624707, "ret_p3m": 0.17093158606432302, "ret_signed_p3m": 0.17093158606432302, "alpha_spy_p3m": 0.15466341915055382, "alpha_c_p3m": 0.15466341915055382, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.0162681669137692, "ret_p6m": 0.711280662297916, "ret_signed_p6m": 0.711280662297916, "alpha_spy_p6m": 0.6016273828444092, "alpha_c_p6m": 0.6016273828444092, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.10965327945350678, "price_path": [-0.1885, -0.1788, -0.1691, -0.1599, -0.1563, -0.1477, -0.1389, -0.1378, -0.1297, -0.1296, -0.1352, -0.102, -0.1055, -0.0872, -0.0886, -0.0902, -0.0904, -0.0882, -0.0858, -0.0754, -0.0546, -0.0417, -0.0461, -0.0271, -0.0451, -0.0195, -0.022, -0.0783, -0.0374, -0.0551, -0.0317, -0.0274, -0.0287, -0.0163, -0.0217, -0.0407, -0.0231, -0.0052, 0.0198, 0.0288, 0.0443, 0.0305, 0.0285, 0.0372, -0.0006, 0.0185, -0.0053, -0.0137, 0.0165, -0.0054, 0.0073, -0.0231, -0.0226, -0.0359, -0.0557, -0.0383, -0.0789, -0.076, -0.0392, -0.0367, -0.0149, -0.0149, -0.0019, 0.0, 0.0183, 0.0323, 0.0245, 0.0325, 0.0442, 0.0454, 0.0599, 0.0508, 0.0033, -0.0002, -0.0029, -0.0389, -0.0162, 0.0093, 0.0223, 0.0321, 0.035, 0.035, 0.0398, 0.0351, 0.0325, 0.0235, 0.0235, 0.0609, 0.0731, 0.1016, 0.0942, 0.0771, 0.1062, 0.1103, 0.1127, 0.1037, 0.1266, 0.1379, 0.1379, 0.1095, 0.1423, 0.1448, 0.1371, 0.1334, 0.1574, 0.1841, 0.1866, 0.1466, 0.1594, 0.1302, 0.0857, 0.083, 0.1415, 0.1557, 0.1503, 0.1788, 0.1542, 0.1587, 0.1587, 0.1581, 0.1725, 0.1658, 0.1795, 0.1734, 0.1913, 0.2111, 0.1709, 0.1549, 0.155, 0.1114, 0.1342, 0.1236, 0.0816, 0.1208, 0.1292, 0.1397, 0.1031, 0.1008, 0.1195, 0.1279, 0.1188, 0.1224, 0.0934, 0.1122, 0.1214, 0.134, 0.0823, 0.0636, 0.0303, 0.0896, 0.114, 0.115, 0.115, 0.1254, 0.1365, 0.2019, 0.2229, 0.2416, 0.26, 0.2846, 0.2874, 0.2925, 0.3191, 0.3186, 0.3206, 0.3551, 0.3694, 0.4393, 0.4388, 0.396, 0.4198, 0.4401, 0.4489, 0.4403, 0.4855, 0.5624, 0.535, 0.6101, 0.6379, 0.5951, 0.6269, 0.6436, 0.5811, 0.5522, 0.5459, 0.6048, 0.6139, 0.6379, 0.6379, 0.7113]}
{"unit_id": "orig:1989252062845247953", "ticker_idx": 0, "ts": "2025-11-14T08:40:24+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]. In the U.S., we prefer Micron [memory] and KLA [SPE]", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan prefers SKH, Samsung, Tokyo Electron, Micron, KLA into 2026 on sustained memory up-cycle, HBM demand raised +15-21%, DRAM price upside risk, and disciplined CAPEX.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251114_Global Memory Market: Key considerations heading into 2026 and picturing bull/bear case scenarios - JPMorgan\n\n[Key Takeaways]\n\n(1) Whether the cycle can be sustained: We remain positive that the memory up-cycle will stay strong, supported by Token Economics\n■ Every +1% increase in ASP adds an additional KRW 0.17–0.23 tn in operating profit\n\n(2) Valuation: Given high margins and a long up-cycle, peak P/B of [2.6x] or higher is possible\n■ However, a full transition into a P/E-driven regime will take some time\n\n(3) HBM demand revision: We raise our 2026–2027 demand outlook by +15–21%\n■ Rubin Ultra Pro, scheduled for launch in 2H27, will feature 4x the HBM content versus existing products\n\n(4) SKH vs. SEC: SEC in the near term, SKH in the mid- to long-term\n■ SKH is expected to maintain its NVIDIA-centric Tier 1 status with 50%+ wallet share \n\n[Contents]\n\n(1) We remain positive on a stronger and longer memory up-cycle driven by AI inference-based Token Economics in an environment where memory supply remains constrained\n\n(2) Within Asian semiconductors, we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]. In the U.S., we prefer Micron [memory] and KLA [SPE]\n\n(3) In the near term, DRAM price momentum leans more towards upside risk\n\n(4) D5 spot prices have surged +177% over the past month, with the spread widening to as much as 263%\n\n(5) This implies further upside risk to our near-term DRAM contract price forecasts of +20% in 4Q25(F) and +10% in 1Q26(F)\n\n(6) Increasing SSD-based purchasing demand and supply disruptions [yield deterioration related to QLC migration, longer lead times until utilization levels fully recover] are also making the NAND price outlook stronger\n\n(7) We estimate that for every +1% ASP increase, the impact on SK hynix’s / Samsung’s earnings is roughly as follows:\n■ DRAM OP: +KRW 0.17–0.26 tn\n■ NAND OP: +KRW 0.07–0.09 tn\n\n(8) Margins are returning to historical peak levels, and narrowing the demand-supply gap will be the key topic going forward\n\n(9) For legacy DRAM in particular, the likelihood of reaching 70% OPM by 4Q25(F) is gradually increasing\n\n(10) If additional margin expansion continues into 1Q26(F), the memory industry will enter a new margin phase\n\n(11) Investor reactions are likely to be very mixed\n\n(12) Some will positively view high margins as evidence of a strong up-cycle, while others will question the sustainability of margins and the slowing pace of EPS upgrades\n\n(13) SK Group Chairman Chey Tae-won mentioned at the SK AI Summit that the top priority for memory makers going forward is to resolve supply-demand imbalances without causing oversupply\n\n(14) We judge that the sustainability of the up-cycle itself is more important for share prices than the exact peak level of margins or absolute earnings size\n\n(15) While memory makers’ CAPEX is increasing, we confirm that “discipline” is being maintained from several angles\n\n(16) Versus the combined SKH+SEC CAPEX growth of +50% in 2017, we expect 2026 CAPEX growth to be around +20%\n\n(17) We also believe most of this will be focused on advanced node shrink and on building new fabs scheduled to start operations in 2027\n■ SEC: Pyeongtaek P4-2 / P4-4 → ±200k wpm on a 1znm basis, 120k wpm on a 1cnm basis\n■ SKH: First phase of Yongin → Equivalent to six M15X-class fabs, 360k+ wpm on a 1cnm basis\n\n(18) Meanwhile, under Token Economics, the role of memory continues to be highlighted\n\n(19) As LLM parameter counts keep expanding from billions to trillions, higher-bandwidth HBM is required to load the models into systems\n\n(20) The AI inference process itself demands high-speed read/write operations and higher memory usage in the pre-fill and decode stages\n\n(21) Since GPUs are extremely expensive resources, they cannot be allowed to process only a single user query at a time\n\n(22) Therefore, improving efficiency — especially by off-loading some operations to other types of memory — is essential, and this has become even more important now that AI has entered an inference-centric era\n\n(23) As a result, hardware architects are allocating more compute resources to test-time compute, and the share of conventional DRAM consumption versus HBM is rising significantly\n\n(24) In particular, the Vera Rubin architecture has 2.4–6.0x higher dependence on conventional DRAM compared with Grace Blackwell\n\n(25) Most CSPs simultaneously run various AI inference and AI agent workloads and operate their businesses based on token pricing models\n\n(26) In this context, resolving data transfer bottlenecks — which are key to token generation — is of utmost importance\n\n(27) Even though memory TCO has more than doubled, CSPs continue to invest in these computing architectures, which is the fundamental reason server memory demand remains structurally strong\n\n(28) We also revise up our HBM demand outlook\n\n(29) Since our CoWoS outlook update, we have raised our 2026–2027 HBM bit demand forecast by +15–21%\n\n(30) This is because we believe HBM supply shortages will deepen over the next two years\n\n(31) This is inevitable given the increase in HBM content on GPUs such as Rubin and the significant upward revision in ASIC shipment forecasts\n■ Rubin Ultra Pro, scheduled for launch in mid-2027 → 4x HBM content versus existing products\n\n(32) Accordingly, we also see room for HBM3E discounts to narrow\n\n(33) GPU vs. ASIC: ASIC mix rising on stronger TPU volumes\n■ NVIDIA GPU shipments: ’26 6.3 mn → 7.2 mn / ’27 6.3 mn → 7.3 mn\n■ ASIC cnfgkfid: ’26 6.0 mn → 6.5 mn / ’27 6.8 mn → 8.7 mn\n\n(34) As TPU volumes are revised up significantly, ASIC bit-demand mix is projected to rise from 22% in 25(F) to 32–33% in 26–27(F)\n\n(35) However, since most ASICs adopt HBM3E rather than HBM4/HBM4E, we expect the ASIC share to fall again in 2028\n\n(36) Meanwhile, Samsung Electronics is aggressively expanding 1cnm HBM production capacity at P4-1/P4-2 based on HBM4 demand from AMD and some NVIDIA products. As a result, Samsung’s market share is projected to rise from 25% this year to 30% next year\n\n(37) Nevertheless, SK hynix is expected to maintain its Tier-1 vendor status in HBM with 50%+ wallet share, based on its strong partnership with NVIDIA\n\n(38) The following is the recent P/E vs. P/B valuation framework\n\n(39) The top three memory makers are currently trading at 2.6x P/B, near historical peak levels, with AI proliferation and expectations for an extended up-cycle raising the risk premium\n\n(40) Considering high margins and a long cycle, applying a valuation above peak P/B can be justified\n\n(41) However, for the valuation framework to shift firmly to a P/E-driven regime, industry-wide structural changes are needed, such as longer-term supply contracts, expansion of customized memory, and more stable margin structures\n\n(42) There is still a gap between share-price expectations and fundamentals, so we judge it is premature to move to a P/E-centric valuation logic\n\n(43) Lastly, on SEC vs. SKH, we prefer SKH from a mid- to long-term perspective\n\n(44) After risk-off phases, SKH has shown stronger share-price elasticity, driven by the expansion of high-margin products built on its technological competitiveness\n\n(45) From a short-term perspective into early 2026, however, SEC could outperform SKH on expectations of HBM share gains and narrowing margin gaps between the two\n\n(46) From a mid-term perspective, SK hynix is better positioned. This is because it has robust technological leadership in HBM and is the furthest ahead in MR-MUF-based 16-high products", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09285720201056402, "ret_signed_m1d": 0.09285720201056402, "alpha_spy_m1d": 0.09269358914618975, "alpha_c_m1d": 0.0933210309821717, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08214287073215654, "ret_signed_p1d": 0.08214287073215654, "alpha_spy_p1d": 0.0914593946631187, "alpha_c_p1d": 0.09570900296467622, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.06964281757401458, "ret_signed_p1w": -0.06964281757401458, "alpha_spy_p1w": -0.05044438723436118, "alpha_c_p1w": -0.015001954481206226, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.010005705267535414, "ret_signed_p1m": -0.010005705267535414, "alpha_spy_p1m": -0.02310222534706352, "alpha_c_p1m": -0.03296350898810585, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.536814195618829, "ret_signed_p3m": 0.536814195618829, "alpha_spy_p3m": 0.5039615363570986, "alpha_c_p3m": 0.33074253506319273, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.3657463705299318, "ret_signed_p6m": 2.3657463705299318, "alpha_spy_p6m": 2.2592181350238754, "alpha_c_p6m": 1.6899877568168509, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.531, -0.5444, -0.5631, -0.5524, -0.5373, -0.5337, -0.5364, -0.5205, -0.5196, -0.5429, -0.5348, -0.5312, -0.5259, -0.5116, -0.5054, -0.4857, -0.4571, -0.4518, -0.4134, -0.4089, -0.3786, -0.4045, -0.3652, -0.3652, -0.3732, -0.3554, -0.3616, -0.3634, -0.3991, -0.3768, -0.3795, -0.3571, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2357, -0.2589, -0.2652, -0.2455, -0.192, -0.1687, -0.133, -0.1446, -0.1402, -0.1455, -0.0893, -0.0446, -0.0696, -0.0036, 0.0143, -0.0018, 0.1071, 0.0464, 0.0339, 0.0589, 0.0357, 0.0821, 0.1054, 0.1018, 0.0929, 0.0, 0.0821, 0.0179, 0.0036, 0.0196, -0.0696, -0.0714, -0.0732, -0.0643, -0.0279, -0.0529, -0.0386, -0.0029, -0.0136, -0.0314, -0.0279, 0.0311, 0.0114, 0.049, 0.0097, 0.0204, -0.01, -0.0529, -0.0154, -0.0136, -0.0225, 0.0365, 0.0436, 0.0508, 0.0508, 0.0704, 0.1437, 0.1633, 0.1633, 0.1633, 0.2098, 0.2437, 0.2974, 0.3259, 0.351, 0.3295, 0.3385, 0.3188, 0.3259, 0.3385, 0.351, 0.3653, 0.3277, 0.3224, 0.3492, 0.3706, 0.3152, 0.4296, 0.5029, 0.5386, 0.6244, 0.4832, 0.6208, 0.6083, 0.5046, 0.4993, 0.5851, 0.5654, 0.5368, 0.5868, 0.5726, 0.5726, 0.5726, 0.5726, 0.5976, 0.6959, 0.6994, 0.7959, 0.8192, 0.9675, 0.8995, 0.8995, 0.6811, 0.52, 0.6847, 0.6542, 0.4967, 0.6793, 0.7097, 0.665, 0.6292, 0.7437, 0.7366, 0.8905, 0.8136, 0.8028, 0.6703, 0.7652, 0.7813, 0.6703, 0.6703, 0.5629, 0.4448, 0.6059, 0.4859, 0.5683, 0.5862, 0.6399, 0.8494, 0.7867, 0.8386, 0.8619, 0.9747, 1.0338, 1.0678, 1.0194, 1.0875, 1.1913, 1.1895, 1.1931, 1.1877, 1.3131, 1.3274, 1.3148, 1.3023, 1.3023, 1.5906, 1.5906, 1.8663, 1.9611, 2.0184, 2.3657]}
{"unit_id": "orig:1989252062845247953", "ticker_idx": 1, "ts": "2025-11-14T08:40:24+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we prefer SK hynix / Samsung Electronics [memory]", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan prefers SKH, Samsung, Tokyo Electron, Micron, KLA into 2026 on sustained memory up-cycle, HBM demand raised +15-21%, DRAM price upside risk, and disciplined CAPEX.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "251114_Global Memory Market: Key considerations heading into 2026 and picturing bull/bear case scenarios - JPMorgan\n\n[Key Takeaways]\n\n(1) Whether the cycle can be sustained: We remain positive that the memory up-cycle will stay strong, supported by Token Economics\n■ Every +1% increase in ASP adds an additional KRW 0.17–0.23 tn in operating profit\n\n(2) Valuation: Given high margins and a long up-cycle, peak P/B of [2.6x] or higher is possible\n■ However, a full transition into a P/E-driven regime will take some time\n\n(3) HBM demand revision: We raise our 2026–2027 demand outlook by +15–21%\n■ Rubin Ultra Pro, scheduled for launch in 2H27, will feature 4x the HBM content versus existing products\n\n(4) SKH vs. SEC: SEC in the near term, SKH in the mid- to long-term\n■ SKH is expected to maintain its NVIDIA-centric Tier 1 status with 50%+ wallet share \n\n[Contents]\n\n(1) We remain positive on a stronger and longer memory up-cycle driven by AI inference-based Token Economics in an environment where memory supply remains constrained\n\n(2) Within Asian semiconductors, we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]. In the U.S., we prefer Micron [memory] and KLA [SPE]\n\n(3) In the near term, DRAM price momentum leans more towards upside risk\n\n(4) D5 spot prices have surged +177% over the past month, with the spread widening to as much as 263%\n\n(5) This implies further upside risk to our near-term DRAM contract price forecasts of +20% in 4Q25(F) and +10% in 1Q26(F)\n\n(6) Increasing SSD-based purchasing demand and supply disruptions [yield deterioration related to QLC migration, longer lead times until utilization levels fully recover] are also making the NAND price outlook stronger\n\n(7) We estimate that for every +1% ASP increase, the impact on SK hynix’s / Samsung’s earnings is roughly as follows:\n■ DRAM OP: +KRW 0.17–0.26 tn\n■ NAND OP: +KRW 0.07–0.09 tn\n\n(8) Margins are returning to historical peak levels, and narrowing the demand-supply gap will be the key topic going forward\n\n(9) For legacy DRAM in particular, the likelihood of reaching 70% OPM by 4Q25(F) is gradually increasing\n\n(10) If additional margin expansion continues into 1Q26(F), the memory industry will enter a new margin phase\n\n(11) Investor reactions are likely to be very mixed\n\n(12) Some will positively view high margins as evidence of a strong up-cycle, while others will question the sustainability of margins and the slowing pace of EPS upgrades\n\n(13) SK Group Chairman Chey Tae-won mentioned at the SK AI Summit that the top priority for memory makers going forward is to resolve supply-demand imbalances without causing oversupply\n\n(14) We judge that the sustainability of the up-cycle itself is more important for share prices than the exact peak level of margins or absolute earnings size\n\n(15) While memory makers’ CAPEX is increasing, we confirm that “discipline” is being maintained from several angles\n\n(16) Versus the combined SKH+SEC CAPEX growth of +50% in 2017, we expect 2026 CAPEX growth to be around +20%\n\n(17) We also believe most of this will be focused on advanced node shrink and on building new fabs scheduled to start operations in 2027\n■ SEC: Pyeongtaek P4-2 / P4-4 → ±200k wpm on a 1znm basis, 120k wpm on a 1cnm basis\n■ SKH: First phase of Yongin → Equivalent to six M15X-class fabs, 360k+ wpm on a 1cnm basis\n\n(18) Meanwhile, under Token Economics, the role of memory continues to be highlighted\n\n(19) As LLM parameter counts keep expanding from billions to trillions, higher-bandwidth HBM is required to load the models into systems\n\n(20) The AI inference process itself demands high-speed read/write operations and higher memory usage in the pre-fill and decode stages\n\n(21) Since GPUs are extremely expensive resources, they cannot be allowed to process only a single user query at a time\n\n(22) Therefore, improving efficiency — especially by off-loading some operations to other types of memory — is essential, and this has become even more important now that AI has entered an inference-centric era\n\n(23) As a result, hardware architects are allocating more compute resources to test-time compute, and the share of conventional DRAM consumption versus HBM is rising significantly\n\n(24) In particular, the Vera Rubin architecture has 2.4–6.0x higher dependence on conventional DRAM compared with Grace Blackwell\n\n(25) Most CSPs simultaneously run various AI inference and AI agent workloads and operate their businesses based on token pricing models\n\n(26) In this context, resolving data transfer bottlenecks — which are key to token generation — is of utmost importance\n\n(27) Even though memory TCO has more than doubled, CSPs continue to invest in these computing architectures, which is the fundamental reason server memory demand remains structurally strong\n\n(28) We also revise up our HBM demand outlook\n\n(29) Since our CoWoS outlook update, we have raised our 2026–2027 HBM bit demand forecast by +15–21%\n\n(30) This is because we believe HBM supply shortages will deepen over the next two years\n\n(31) This is inevitable given the increase in HBM content on GPUs such as Rubin and the significant upward revision in ASIC shipment forecasts\n■ Rubin Ultra Pro, scheduled for launch in mid-2027 → 4x HBM content versus existing products\n\n(32) Accordingly, we also see room for HBM3E discounts to narrow\n\n(33) GPU vs. ASIC: ASIC mix rising on stronger TPU volumes\n■ NVIDIA GPU shipments: ’26 6.3 mn → 7.2 mn / ’27 6.3 mn → 7.3 mn\n■ ASIC cnfgkfid: ’26 6.0 mn → 6.5 mn / ’27 6.8 mn → 8.7 mn\n\n(34) As TPU volumes are revised up significantly, ASIC bit-demand mix is projected to rise from 22% in 25(F) to 32–33% in 26–27(F)\n\n(35) However, since most ASICs adopt HBM3E rather than HBM4/HBM4E, we expect the ASIC share to fall again in 2028\n\n(36) Meanwhile, Samsung Electronics is aggressively expanding 1cnm HBM production capacity at P4-1/P4-2 based on HBM4 demand from AMD and some NVIDIA products. As a result, Samsung’s market share is projected to rise from 25% this year to 30% next year\n\n(37) Nevertheless, SK hynix is expected to maintain its Tier-1 vendor status in HBM with 50%+ wallet share, based on its strong partnership with NVIDIA\n\n(38) The following is the recent P/E vs. P/B valuation framework\n\n(39) The top three memory makers are currently trading at 2.6x P/B, near historical peak levels, with AI proliferation and expectations for an extended up-cycle raising the risk premium\n\n(40) Considering high margins and a long cycle, applying a valuation above peak P/B can be justified\n\n(41) However, for the valuation framework to shift firmly to a P/E-driven regime, industry-wide structural changes are needed, such as longer-term supply contracts, expansion of customized memory, and more stable margin structures\n\n(42) There is still a gap between share-price expectations and fundamentals, so we judge it is premature to move to a P/E-centric valuation logic\n\n(43) Lastly, on SEC vs. SKH, we prefer SKH from a mid- to long-term perspective\n\n(44) After risk-off phases, SKH has shown stronger share-price elasticity, driven by the expansion of high-margin products built on its technological competitiveness\n\n(45) From a short-term perspective into early 2026, however, SEC could outperform SKH on expectations of HBM share gains and narrowing margin gaps between the two\n\n(46) From a mid-term perspective, SK hynix is better positioned. This is because it has robust technological leadership in HBM and is the furthest ahead in MR-MUF-based 16-high products", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05761317038957614, "ret_signed_m1d": 0.05761317038957614, "alpha_spy_m1d": 0.057449557525201866, "alpha_c_m1d": 0.06302709370298476, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03497941620794953, "ret_signed_p1d": 0.03497941620794953, "alpha_spy_p1d": 0.044295940138911694, "alpha_c_p1d": 0.05063102623894755, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.024691404986047938, "ret_signed_p1w": -0.024691404986047938, "alpha_spy_p1w": -0.005492974646394533, "alpha_c_p1w": 0.027191265659056163, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.07818927376678131, "ret_signed_p1m": 0.07818927376678131, "alpha_spy_p1m": 0.0650927536872532, "alpha_c_p1m": 0.09050924685115647, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": 0.7347294453035076, "ret_signed_p3m": 0.7347294453035076, "alpha_spy_p3m": 0.7018767860417772, "alpha_c_p3m": 0.7408938922948921, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 1.957630231150651, "ret_signed_p6m": 1.957630231150651, "alpha_spy_p6m": 1.8511019956445949, "alpha_c_p6m": 1.7195854032114168, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.283, -0.2779, -0.2769, -0.2687, -0.2677, -0.28, -0.2769, -0.2871, -0.2861, -0.3076, -0.2923, -0.2851, -0.282, -0.2882, -0.282, -0.2677, -0.2564, -0.2482, -0.2277, -0.2165, -0.1868, -0.199, -0.1775, -0.1775, -0.1448, -0.1325, -0.1253, -0.1181, -0.1468, -0.1337, -0.1368, -0.1152, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0288, -0.0401, -0.0576, -0.0226, 0.0051, 0.0072, 0.0093, 0.0031, 0.0144, -0.0072, 0.0165, 0.0494, 0.0237, 0.034, 0.071, 0.106, 0.143, 0.0792, 0.035, 0.0206, 0.0072, 0.035, 0.0648, 0.0607, 0.0576, 0.0, 0.035, 0.0062, -0.0072, 0.035, -0.0247, -0.0051, 0.0216, 0.0576, 0.0648, 0.034, 0.037, 0.0638, 0.0751, 0.0813, 0.1152, 0.1265, 0.1152, 0.1111, 0.1039, 0.1204, 0.0782, 0.0576, 0.1101, 0.107, 0.0936, 0.1368, 0.1471, 0.143, 0.143, 0.2037, 0.2354, 0.2395, 0.2395, 0.2395, 0.3284, 0.4277, 0.436, 0.4577, 0.4349, 0.437, 0.4349, 0.4225, 0.4504, 0.4876, 0.5393, 0.5435, 0.5011, 0.5455, 0.5745, 0.5724, 0.5724, 0.6489, 0.6789, 0.6613, 0.6593, 0.5548, 0.7316, 0.7482, 0.6469, 0.6396, 0.7203, 0.7141, 0.7347, 0.8464, 0.8733, 0.8733, 0.8733, 0.8733, 0.9642, 0.9653, 0.9952, 1.0676, 1.1038, 1.2537, 1.2382, 1.2382, 1.017, 0.7802, 0.9808, 0.9456, 0.7937, 0.9425, 0.9642, 0.9425, 0.897, 0.9508, 1.0046, 1.1555, 1.0728, 1.0614, 0.926, 0.9611, 0.9539, 0.8619, 0.8619, 0.8264, 0.7321, 0.9647, 0.8481, 0.9289, 1.0004, 1.0356, 1.1807, 1.1133, 1.1341, 1.0823, 1.1392, 1.1858, 1.2532, 1.2376, 1.2221, 1.2687, 1.2532, 1.3257, 1.2739, 1.3257, 1.2998, 1.3412, 1.2843, 1.2843, 1.4086, 1.4086, 1.7556, 1.8126, 1.7815, 1.9576]}
{"unit_id": "orig:1989252062845247953", "ticker_idx": 2, "ts": "2025-11-14T08:40:24+00:00", "symbol": "8035.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan prefers SKH, Samsung, Tokyo Electron, Micron, KLA into 2026 on sustained memory up-cycle, HBM demand raised +15-21%, DRAM price upside risk, and disciplined CAPEX.", "resolved_tickers": ["8035.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "251114_Global Memory Market: Key considerations heading into 2026 and picturing bull/bear case scenarios - JPMorgan\n\n[Key Takeaways]\n\n(1) Whether the cycle can be sustained: We remain positive that the memory up-cycle will stay strong, supported by Token Economics\n■ Every +1% increase in ASP adds an additional KRW 0.17–0.23 tn in operating profit\n\n(2) Valuation: Given high margins and a long up-cycle, peak P/B of [2.6x] or higher is possible\n■ However, a full transition into a P/E-driven regime will take some time\n\n(3) HBM demand revision: We raise our 2026–2027 demand outlook by +15–21%\n■ Rubin Ultra Pro, scheduled for launch in 2H27, will feature 4x the HBM content versus existing products\n\n(4) SKH vs. SEC: SEC in the near term, SKH in the mid- to long-term\n■ SKH is expected to maintain its NVIDIA-centric Tier 1 status with 50%+ wallet share \n\n[Contents]\n\n(1) We remain positive on a stronger and longer memory up-cycle driven by AI inference-based Token Economics in an environment where memory supply remains constrained\n\n(2) Within Asian semiconductors, we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]. In the U.S., we prefer Micron [memory] and KLA [SPE]\n\n(3) In the near term, DRAM price momentum leans more towards upside risk\n\n(4) D5 spot prices have surged +177% over the past month, with the spread widening to as much as 263%\n\n(5) This implies further upside risk to our near-term DRAM contract price forecasts of +20% in 4Q25(F) and +10% in 1Q26(F)\n\n(6) Increasing SSD-based purchasing demand and supply disruptions [yield deterioration related to QLC migration, longer lead times until utilization levels fully recover] are also making the NAND price outlook stronger\n\n(7) We estimate that for every +1% ASP increase, the impact on SK hynix’s / Samsung’s earnings is roughly as follows:\n■ DRAM OP: +KRW 0.17–0.26 tn\n■ NAND OP: +KRW 0.07–0.09 tn\n\n(8) Margins are returning to historical peak levels, and narrowing the demand-supply gap will be the key topic going forward\n\n(9) For legacy DRAM in particular, the likelihood of reaching 70% OPM by 4Q25(F) is gradually increasing\n\n(10) If additional margin expansion continues into 1Q26(F), the memory industry will enter a new margin phase\n\n(11) Investor reactions are likely to be very mixed\n\n(12) Some will positively view high margins as evidence of a strong up-cycle, while others will question the sustainability of margins and the slowing pace of EPS upgrades\n\n(13) SK Group Chairman Chey Tae-won mentioned at the SK AI Summit that the top priority for memory makers going forward is to resolve supply-demand imbalances without causing oversupply\n\n(14) We judge that the sustainability of the up-cycle itself is more important for share prices than the exact peak level of margins or absolute earnings size\n\n(15) While memory makers’ CAPEX is increasing, we confirm that “discipline” is being maintained from several angles\n\n(16) Versus the combined SKH+SEC CAPEX growth of +50% in 2017, we expect 2026 CAPEX growth to be around +20%\n\n(17) We also believe most of this will be focused on advanced node shrink and on building new fabs scheduled to start operations in 2027\n■ SEC: Pyeongtaek P4-2 / P4-4 → ±200k wpm on a 1znm basis, 120k wpm on a 1cnm basis\n■ SKH: First phase of Yongin → Equivalent to six M15X-class fabs, 360k+ wpm on a 1cnm basis\n\n(18) Meanwhile, under Token Economics, the role of memory continues to be highlighted\n\n(19) As LLM parameter counts keep expanding from billions to trillions, higher-bandwidth HBM is required to load the models into systems\n\n(20) The AI inference process itself demands high-speed read/write operations and higher memory usage in the pre-fill and decode stages\n\n(21) Since GPUs are extremely expensive resources, they cannot be allowed to process only a single user query at a time\n\n(22) Therefore, improving efficiency — especially by off-loading some operations to other types of memory — is essential, and this has become even more important now that AI has entered an inference-centric era\n\n(23) As a result, hardware architects are allocating more compute resources to test-time compute, and the share of conventional DRAM consumption versus HBM is rising significantly\n\n(24) In particular, the Vera Rubin architecture has 2.4–6.0x higher dependence on conventional DRAM compared with Grace Blackwell\n\n(25) Most CSPs simultaneously run various AI inference and AI agent workloads and operate their businesses based on token pricing models\n\n(26) In this context, resolving data transfer bottlenecks — which are key to token generation — is of utmost importance\n\n(27) Even though memory TCO has more than doubled, CSPs continue to invest in these computing architectures, which is the fundamental reason server memory demand remains structurally strong\n\n(28) We also revise up our HBM demand outlook\n\n(29) Since our CoWoS outlook update, we have raised our 2026–2027 HBM bit demand forecast by +15–21%\n\n(30) This is because we believe HBM supply shortages will deepen over the next two years\n\n(31) This is inevitable given the increase in HBM content on GPUs such as Rubin and the significant upward revision in ASIC shipment forecasts\n■ Rubin Ultra Pro, scheduled for launch in mid-2027 → 4x HBM content versus existing products\n\n(32) Accordingly, we also see room for HBM3E discounts to narrow\n\n(33) GPU vs. ASIC: ASIC mix rising on stronger TPU volumes\n■ NVIDIA GPU shipments: ’26 6.3 mn → 7.2 mn / ’27 6.3 mn → 7.3 mn\n■ ASIC cnfgkfid: ’26 6.0 mn → 6.5 mn / ’27 6.8 mn → 8.7 mn\n\n(34) As TPU volumes are revised up significantly, ASIC bit-demand mix is projected to rise from 22% in 25(F) to 32–33% in 26–27(F)\n\n(35) However, since most ASICs adopt HBM3E rather than HBM4/HBM4E, we expect the ASIC share to fall again in 2028\n\n(36) Meanwhile, Samsung Electronics is aggressively expanding 1cnm HBM production capacity at P4-1/P4-2 based on HBM4 demand from AMD and some NVIDIA products. As a result, Samsung’s market share is projected to rise from 25% this year to 30% next year\n\n(37) Nevertheless, SK hynix is expected to maintain its Tier-1 vendor status in HBM with 50%+ wallet share, based on its strong partnership with NVIDIA\n\n(38) The following is the recent P/E vs. P/B valuation framework\n\n(39) The top three memory makers are currently trading at 2.6x P/B, near historical peak levels, with AI proliferation and expectations for an extended up-cycle raising the risk premium\n\n(40) Considering high margins and a long cycle, applying a valuation above peak P/B can be justified\n\n(41) However, for the valuation framework to shift firmly to a P/E-driven regime, industry-wide structural changes are needed, such as longer-term supply contracts, expansion of customized memory, and more stable margin structures\n\n(42) There is still a gap between share-price expectations and fundamentals, so we judge it is premature to move to a P/E-centric valuation logic\n\n(43) Lastly, on SEC vs. SKH, we prefer SKH from a mid- to long-term perspective\n\n(44) After risk-off phases, SKH has shown stronger share-price elasticity, driven by the expansion of high-margin products built on its technological competitiveness\n\n(45) From a short-term perspective into early 2026, however, SEC could outperform SKH on expectations of HBM share gains and narrowing margin gaps between the two\n\n(46) From a mid-term perspective, SK hynix is better positioned. This is because it has robust technological leadership in HBM and is the furthest ahead in MR-MUF-based 16-high products", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06467826870829518, "ret_signed_m1d": 0.06467826870829518, "alpha_spy_m1d": 0.06451465584392091, "alpha_c_m1d": 0.06514209767990287, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.045525900336726766, "ret_signed_p1d": 0.045525900336726766, "alpha_spy_p1d": 0.05484242426768893, "alpha_c_p1d": 0.059092032569246444, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.052433250999218695, "ret_signed_p1w": -0.052433250999218695, "alpha_spy_p1w": -0.03323482065956529, "alpha_c_p1w": 0.0022076120935896615, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.022291983187549547, "ret_signed_p1m": -0.022291983187549547, "alpha_spy_p1m": -0.03538850326707765, "alpha_c_p1m": -0.04524978690811998, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.3029827331088488, "ret_signed_p3m": 0.3029827331088488, "alpha_spy_p3m": 0.2701300738471184, "alpha_c_p3m": 0.09691107255321252, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.6504554744015922, "ret_signed_p6m": 0.6504554744015922, "alpha_spy_p6m": 0.5439272388955361, "alpha_c_p6m": -0.025303139311488643, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.3404, -0.3495, -0.3652, -0.3753, -0.3713, -0.3707, -0.3679, -0.355, -0.3576, -0.3694, -0.3679, -0.38, -0.3757, -0.3708, -0.3597, -0.3466, -0.3426, -0.3346, -0.2979, -0.2979, -0.2846, -0.2462, -0.2087, -0.2039, -0.1731, -0.1731, -0.1528, -0.1383, -0.1749, -0.1678, -0.1724, -0.1896, -0.1257, -0.1055, -0.0402, -0.0553, -0.0815, -0.0675, -0.0807, -0.0807, -0.1071, -0.0873, -0.0502, -0.0556, -0.0138, -0.0358, -0.0352, -0.0666, -0.0496, -0.0386, -0.0126, 0.0201, 0.0358, 0.0732, 0.0732, 0.0923, 0.0477, 0.044, 0.0298, 0.0738, 0.0735, 0.0568, 0.0647, 0.0, 0.0455, -0.0116, -0.0311, 0.0204, -0.0524, -0.0524, -0.0235, -0.0214, 0.0104, -0.0016, -0.0069, -0.0173, 0.0292, 0.0619, 0.0405, 0.0411, 0.0546, 0.0399, 0.0235, -0.011, -0.0223, -0.0276, -0.0166, -0.0484, -0.0204, 0.0414, 0.0361, 0.043, 0.0609, 0.0779, 0.0747, 0.0776, 0.0776, 0.0776, 0.0776, 0.1595, 0.1727, 0.1984, 0.1504, 0.1903, 0.1903, 0.2882, 0.3284, 0.3372, 0.3234, 0.3243, 0.2898, 0.2939, 0.3344, 0.3099, 0.2901, 0.3228, 0.3752, 0.2976, 0.297, 0.2452, 0.3049, 0.2779, 0.2556, 0.2882, 0.2747, 0.303, 0.303, 0.2967, 0.3184, 0.2992, 0.3234, 0.3623, 0.4013, 0.3802, 0.3802, 0.3934, 0.4515, 0.4226, 0.3818, 0.3667, 0.3375, 0.2776, 0.3099, 0.3121, 0.222, 0.2568, 0.2725, 0.248, 0.2038, 0.2292, 0.2176, 0.265, 0.2349, 0.2349, 0.2025, 0.2286, 0.2672, 0.2725, 0.2336, 0.2309, 0.1798, 0.2448, 0.2049, 0.2176, 0.216, 0.2182, 0.344, 0.337, 0.3957, 0.3456, 0.3849, 0.3785, 0.4521, 0.3947, 0.4023, 0.4508, 0.4419, 0.4413, 0.453, 0.4926, 0.4308, 0.4308, 0.4068, 0.5037, 0.5037, 0.5037, 0.5037, 0.639, 0.6622, 0.6505]}
{"unit_id": "orig:1989252062845247953", "ticker_idx": 3, "ts": "2025-11-14T08:40:24+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "In the U.S., we prefer Micron [memory] and KLA [SPE]", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan prefers SKH, Samsung, Tokyo Electron, Micron, KLA into 2026 on sustained memory up-cycle, HBM demand raised +15-21%, DRAM price upside risk, and disciplined CAPEX.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251114_Global Memory Market: Key considerations heading into 2026 and picturing bull/bear case scenarios - JPMorgan\n\n[Key Takeaways]\n\n(1) Whether the cycle can be sustained: We remain positive that the memory up-cycle will stay strong, supported by Token Economics\n■ Every +1% increase in ASP adds an additional KRW 0.17–0.23 tn in operating profit\n\n(2) Valuation: Given high margins and a long up-cycle, peak P/B of [2.6x] or higher is possible\n■ However, a full transition into a P/E-driven regime will take some time\n\n(3) HBM demand revision: We raise our 2026–2027 demand outlook by +15–21%\n■ Rubin Ultra Pro, scheduled for launch in 2H27, will feature 4x the HBM content versus existing products\n\n(4) SKH vs. SEC: SEC in the near term, SKH in the mid- to long-term\n■ SKH is expected to maintain its NVIDIA-centric Tier 1 status with 50%+ wallet share \n\n[Contents]\n\n(1) We remain positive on a stronger and longer memory up-cycle driven by AI inference-based Token Economics in an environment where memory supply remains constrained\n\n(2) Within Asian semiconductors, we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]. In the U.S., we prefer Micron [memory] and KLA [SPE]\n\n(3) In the near term, DRAM price momentum leans more towards upside risk\n\n(4) D5 spot prices have surged +177% over the past month, with the spread widening to as much as 263%\n\n(5) This implies further upside risk to our near-term DRAM contract price forecasts of +20% in 4Q25(F) and +10% in 1Q26(F)\n\n(6) Increasing SSD-based purchasing demand and supply disruptions [yield deterioration related to QLC migration, longer lead times until utilization levels fully recover] are also making the NAND price outlook stronger\n\n(7) We estimate that for every +1% ASP increase, the impact on SK hynix’s / Samsung’s earnings is roughly as follows:\n■ DRAM OP: +KRW 0.17–0.26 tn\n■ NAND OP: +KRW 0.07–0.09 tn\n\n(8) Margins are returning to historical peak levels, and narrowing the demand-supply gap will be the key topic going forward\n\n(9) For legacy DRAM in particular, the likelihood of reaching 70% OPM by 4Q25(F) is gradually increasing\n\n(10) If additional margin expansion continues into 1Q26(F), the memory industry will enter a new margin phase\n\n(11) Investor reactions are likely to be very mixed\n\n(12) Some will positively view high margins as evidence of a strong up-cycle, while others will question the sustainability of margins and the slowing pace of EPS upgrades\n\n(13) SK Group Chairman Chey Tae-won mentioned at the SK AI Summit that the top priority for memory makers going forward is to resolve supply-demand imbalances without causing oversupply\n\n(14) We judge that the sustainability of the up-cycle itself is more important for share prices than the exact peak level of margins or absolute earnings size\n\n(15) While memory makers’ CAPEX is increasing, we confirm that “discipline” is being maintained from several angles\n\n(16) Versus the combined SKH+SEC CAPEX growth of +50% in 2017, we expect 2026 CAPEX growth to be around +20%\n\n(17) We also believe most of this will be focused on advanced node shrink and on building new fabs scheduled to start operations in 2027\n■ SEC: Pyeongtaek P4-2 / P4-4 → ±200k wpm on a 1znm basis, 120k wpm on a 1cnm basis\n■ SKH: First phase of Yongin → Equivalent to six M15X-class fabs, 360k+ wpm on a 1cnm basis\n\n(18) Meanwhile, under Token Economics, the role of memory continues to be highlighted\n\n(19) As LLM parameter counts keep expanding from billions to trillions, higher-bandwidth HBM is required to load the models into systems\n\n(20) The AI inference process itself demands high-speed read/write operations and higher memory usage in the pre-fill and decode stages\n\n(21) Since GPUs are extremely expensive resources, they cannot be allowed to process only a single user query at a time\n\n(22) Therefore, improving efficiency — especially by off-loading some operations to other types of memory — is essential, and this has become even more important now that AI has entered an inference-centric era\n\n(23) As a result, hardware architects are allocating more compute resources to test-time compute, and the share of conventional DRAM consumption versus HBM is rising significantly\n\n(24) In particular, the Vera Rubin architecture has 2.4–6.0x higher dependence on conventional DRAM compared with Grace Blackwell\n\n(25) Most CSPs simultaneously run various AI inference and AI agent workloads and operate their businesses based on token pricing models\n\n(26) In this context, resolving data transfer bottlenecks — which are key to token generation — is of utmost importance\n\n(27) Even though memory TCO has more than doubled, CSPs continue to invest in these computing architectures, which is the fundamental reason server memory demand remains structurally strong\n\n(28) We also revise up our HBM demand outlook\n\n(29) Since our CoWoS outlook update, we have raised our 2026–2027 HBM bit demand forecast by +15–21%\n\n(30) This is because we believe HBM supply shortages will deepen over the next two years\n\n(31) This is inevitable given the increase in HBM content on GPUs such as Rubin and the significant upward revision in ASIC shipment forecasts\n■ Rubin Ultra Pro, scheduled for launch in mid-2027 → 4x HBM content versus existing products\n\n(32) Accordingly, we also see room for HBM3E discounts to narrow\n\n(33) GPU vs. ASIC: ASIC mix rising on stronger TPU volumes\n■ NVIDIA GPU shipments: ’26 6.3 mn → 7.2 mn / ’27 6.3 mn → 7.3 mn\n■ ASIC cnfgkfid: ’26 6.0 mn → 6.5 mn / ’27 6.8 mn → 8.7 mn\n\n(34) As TPU volumes are revised up significantly, ASIC bit-demand mix is projected to rise from 22% in 25(F) to 32–33% in 26–27(F)\n\n(35) However, since most ASICs adopt HBM3E rather than HBM4/HBM4E, we expect the ASIC share to fall again in 2028\n\n(36) Meanwhile, Samsung Electronics is aggressively expanding 1cnm HBM production capacity at P4-1/P4-2 based on HBM4 demand from AMD and some NVIDIA products. As a result, Samsung’s market share is projected to rise from 25% this year to 30% next year\n\n(37) Nevertheless, SK hynix is expected to maintain its Tier-1 vendor status in HBM with 50%+ wallet share, based on its strong partnership with NVIDIA\n\n(38) The following is the recent P/E vs. P/B valuation framework\n\n(39) The top three memory makers are currently trading at 2.6x P/B, near historical peak levels, with AI proliferation and expectations for an extended up-cycle raising the risk premium\n\n(40) Considering high margins and a long cycle, applying a valuation above peak P/B can be justified\n\n(41) However, for the valuation framework to shift firmly to a P/E-driven regime, industry-wide structural changes are needed, such as longer-term supply contracts, expansion of customized memory, and more stable margin structures\n\n(42) There is still a gap between share-price expectations and fundamentals, so we judge it is premature to move to a P/E-centric valuation logic\n\n(43) Lastly, on SEC vs. SKH, we prefer SKH from a mid- to long-term perspective\n\n(44) After risk-off phases, SKH has shown stronger share-price elasticity, driven by the expansion of high-margin products built on its technological competitiveness\n\n(45) From a short-term perspective into early 2026, however, SEC could outperform SKH on expectations of HBM share gains and narrowing margin gaps between the two\n\n(46) From a mid-term perspective, SK hynix is better positioned. This is because it has robust technological leadership in HBM and is the furthest ahead in MR-MUF-based 16-high products", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04002758410929108, "ret_signed_m1d": -0.04002758410929108, "alpha_spy_m1d": -0.040191196973665355, "alpha_c_m1d": -0.0395637551376834, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01977073586105582, "ret_signed_p1d": -0.01977073586105582, "alpha_spy_p1d": -0.01045421193009366, "alpha_c_p1d": -0.006204603628536143, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.15986712556863925, "ret_signed_p1w": -0.15986712556863925, "alpha_spy_p1w": -0.14066869522898584, "alpha_c_p1w": -0.10522626247583089, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.03779927511900849, "ret_signed_p1m": -0.03779927511900849, "alpha_spy_p1m": -0.0508957951985366, "alpha_c_p1m": -0.060757078839578926, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.6631113086000686, "ret_signed_p3m": 0.6631113086000686, "alpha_spy_p3m": 0.6302586493383382, "alpha_c_p3m": 0.4570396480444323, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.224833073262897, "ret_signed_p6m": 2.224833073262897, "alpha_spy_p6m": 2.118304837756841, "alpha_c_p6m": 1.549074459549816, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.5058, -0.5254, -0.5312, -0.5235, -0.5286, -0.5283, -0.5232, -0.506, -0.5181, -0.5181, -0.5203, -0.5193, -0.4971, -0.4681, -0.4677, -0.4524, -0.4332, -0.3904, -0.3634, -0.3612, -0.357, -0.3522, -0.3162, -0.3411, -0.3335, -0.3262, -0.3453, -0.365, -0.3632, -0.3364, -0.3225, -0.2625, -0.256, -0.239, -0.2264, -0.2477, -0.2037, -0.2208, -0.2643, -0.219, -0.2422, -0.2224, -0.1795, -0.1801, -0.1623, -0.1804, -0.1959, -0.1625, -0.1127, -0.1083, -0.101, -0.0818, -0.0925, -0.0934, -0.0491, -0.1167, -0.0378, -0.0344, -0.0361, 0.0262, -0.0232, -0.0078, -0.04, 0.0, -0.0198, -0.0743, -0.0847, -0.1842, -0.1599, -0.0928, -0.0903, -0.0671, -0.0671, -0.0419, -0.0258, -0.0297, -0.0513, -0.0818, -0.0389, 0.0004, 0.0226, 0.0684, 0.0471, -0.0231, -0.0378, -0.058, -0.0863, 0.007, 0.0773, 0.1206, 0.1193, 0.1614, 0.1614, 0.1538, 0.1931, 0.186, 0.1568, 0.1568, 0.2784, 0.2651, 0.3919, 0.3762, 0.3254, 0.3987, 0.4018, 0.3704, 0.3511, 0.3644, 0.4702, 0.4702, 0.4793, 0.5771, 0.6114, 0.6198, 0.577, 0.6627, 0.7642, 0.7663, 0.6815, 0.7744, 0.7, 0.5377, 0.5519, 0.5997, 0.5543, 0.5128, 0.6631, 0.6778, 0.6685, 0.6685, 0.6203, 0.7061, 0.6915, 0.7354, 0.7062, 0.6942, 0.7387, 0.6843, 0.6713, 0.6726, 0.5388, 0.6243, 0.6092, 0.5008, 0.5779, 0.6338, 0.697, 0.6429, 0.7271, 0.7906, 0.8712, 0.8714, 0.8006, 0.714, 0.6388, 0.6031, 0.5486, 0.4407, 0.4478, 0.3048, 0.3698, 0.4915, 0.485, 0.485, 0.5317, 0.531, 0.6492, 0.7091, 0.7054, 0.7296, 0.8881, 0.8499, 0.8539, 0.8452, 0.8182, 0.8221, 0.9766, 0.9532, 1.0141, 1.1269, 1.0448, 1.1022, 1.0969, 1.1985, 1.3373, 1.5958, 1.7028, 1.6219, 2.0281, 2.2248]}
{"unit_id": "orig:1989252062845247953", "ticker_idx": 4, "ts": "2025-11-14T08:40:24+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "In the U.S., we prefer Micron [memory] and KLA [SPE]", "tweet_type": "analysis", "timeline": "quarters", "summary": "JPMorgan prefers SKH, Samsung, Tokyo Electron, Micron, KLA into 2026 on sustained memory up-cycle, HBM demand raised +15-21%, DRAM price upside risk, and disciplined CAPEX.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "251114_Global Memory Market: Key considerations heading into 2026 and picturing bull/bear case scenarios - JPMorgan\n\n[Key Takeaways]\n\n(1) Whether the cycle can be sustained: We remain positive that the memory up-cycle will stay strong, supported by Token Economics\n■ Every +1% increase in ASP adds an additional KRW 0.17–0.23 tn in operating profit\n\n(2) Valuation: Given high margins and a long up-cycle, peak P/B of [2.6x] or higher is possible\n■ However, a full transition into a P/E-driven regime will take some time\n\n(3) HBM demand revision: We raise our 2026–2027 demand outlook by +15–21%\n■ Rubin Ultra Pro, scheduled for launch in 2H27, will feature 4x the HBM content versus existing products\n\n(4) SKH vs. SEC: SEC in the near term, SKH in the mid- to long-term\n■ SKH is expected to maintain its NVIDIA-centric Tier 1 status with 50%+ wallet share \n\n[Contents]\n\n(1) We remain positive on a stronger and longer memory up-cycle driven by AI inference-based Token Economics in an environment where memory supply remains constrained\n\n(2) Within Asian semiconductors, we prefer SK hynix / Samsung Electronics [memory] and Tokyo Electron [SPE]. In the U.S., we prefer Micron [memory] and KLA [SPE]\n\n(3) In the near term, DRAM price momentum leans more towards upside risk\n\n(4) D5 spot prices have surged +177% over the past month, with the spread widening to as much as 263%\n\n(5) This implies further upside risk to our near-term DRAM contract price forecasts of +20% in 4Q25(F) and +10% in 1Q26(F)\n\n(6) Increasing SSD-based purchasing demand and supply disruptions [yield deterioration related to QLC migration, longer lead times until utilization levels fully recover] are also making the NAND price outlook stronger\n\n(7) We estimate that for every +1% ASP increase, the impact on SK hynix’s / Samsung’s earnings is roughly as follows:\n■ DRAM OP: +KRW 0.17–0.26 tn\n■ NAND OP: +KRW 0.07–0.09 tn\n\n(8) Margins are returning to historical peak levels, and narrowing the demand-supply gap will be the key topic going forward\n\n(9) For legacy DRAM in particular, the likelihood of reaching 70% OPM by 4Q25(F) is gradually increasing\n\n(10) If additional margin expansion continues into 1Q26(F), the memory industry will enter a new margin phase\n\n(11) Investor reactions are likely to be very mixed\n\n(12) Some will positively view high margins as evidence of a strong up-cycle, while others will question the sustainability of margins and the slowing pace of EPS upgrades\n\n(13) SK Group Chairman Chey Tae-won mentioned at the SK AI Summit that the top priority for memory makers going forward is to resolve supply-demand imbalances without causing oversupply\n\n(14) We judge that the sustainability of the up-cycle itself is more important for share prices than the exact peak level of margins or absolute earnings size\n\n(15) While memory makers’ CAPEX is increasing, we confirm that “discipline” is being maintained from several angles\n\n(16) Versus the combined SKH+SEC CAPEX growth of +50% in 2017, we expect 2026 CAPEX growth to be around +20%\n\n(17) We also believe most of this will be focused on advanced node shrink and on building new fabs scheduled to start operations in 2027\n■ SEC: Pyeongtaek P4-2 / P4-4 → ±200k wpm on a 1znm basis, 120k wpm on a 1cnm basis\n■ SKH: First phase of Yongin → Equivalent to six M15X-class fabs, 360k+ wpm on a 1cnm basis\n\n(18) Meanwhile, under Token Economics, the role of memory continues to be highlighted\n\n(19) As LLM parameter counts keep expanding from billions to trillions, higher-bandwidth HBM is required to load the models into systems\n\n(20) The AI inference process itself demands high-speed read/write operations and higher memory usage in the pre-fill and decode stages\n\n(21) Since GPUs are extremely expensive resources, they cannot be allowed to process only a single user query at a time\n\n(22) Therefore, improving efficiency — especially by off-loading some operations to other types of memory — is essential, and this has become even more important now that AI has entered an inference-centric era\n\n(23) As a result, hardware architects are allocating more compute resources to test-time compute, and the share of conventional DRAM consumption versus HBM is rising significantly\n\n(24) In particular, the Vera Rubin architecture has 2.4–6.0x higher dependence on conventional DRAM compared with Grace Blackwell\n\n(25) Most CSPs simultaneously run various AI inference and AI agent workloads and operate their businesses based on token pricing models\n\n(26) In this context, resolving data transfer bottlenecks — which are key to token generation — is of utmost importance\n\n(27) Even though memory TCO has more than doubled, CSPs continue to invest in these computing architectures, which is the fundamental reason server memory demand remains structurally strong\n\n(28) We also revise up our HBM demand outlook\n\n(29) Since our CoWoS outlook update, we have raised our 2026–2027 HBM bit demand forecast by +15–21%\n\n(30) This is because we believe HBM supply shortages will deepen over the next two years\n\n(31) This is inevitable given the increase in HBM content on GPUs such as Rubin and the significant upward revision in ASIC shipment forecasts\n■ Rubin Ultra Pro, scheduled for launch in mid-2027 → 4x HBM content versus existing products\n\n(32) Accordingly, we also see room for HBM3E discounts to narrow\n\n(33) GPU vs. ASIC: ASIC mix rising on stronger TPU volumes\n■ NVIDIA GPU shipments: ’26 6.3 mn → 7.2 mn / ’27 6.3 mn → 7.3 mn\n■ ASIC cnfgkfid: ’26 6.0 mn → 6.5 mn / ’27 6.8 mn → 8.7 mn\n\n(34) As TPU volumes are revised up significantly, ASIC bit-demand mix is projected to rise from 22% in 25(F) to 32–33% in 26–27(F)\n\n(35) However, since most ASICs adopt HBM3E rather than HBM4/HBM4E, we expect the ASIC share to fall again in 2028\n\n(36) Meanwhile, Samsung Electronics is aggressively expanding 1cnm HBM production capacity at P4-1/P4-2 based on HBM4 demand from AMD and some NVIDIA products. As a result, Samsung’s market share is projected to rise from 25% this year to 30% next year\n\n(37) Nevertheless, SK hynix is expected to maintain its Tier-1 vendor status in HBM with 50%+ wallet share, based on its strong partnership with NVIDIA\n\n(38) The following is the recent P/E vs. P/B valuation framework\n\n(39) The top three memory makers are currently trading at 2.6x P/B, near historical peak levels, with AI proliferation and expectations for an extended up-cycle raising the risk premium\n\n(40) Considering high margins and a long cycle, applying a valuation above peak P/B can be justified\n\n(41) However, for the valuation framework to shift firmly to a P/E-driven regime, industry-wide structural changes are needed, such as longer-term supply contracts, expansion of customized memory, and more stable margin structures\n\n(42) There is still a gap between share-price expectations and fundamentals, so we judge it is premature to move to a P/E-centric valuation logic\n\n(43) Lastly, on SEC vs. SKH, we prefer SKH from a mid- to long-term perspective\n\n(44) After risk-off phases, SKH has shown stronger share-price elasticity, driven by the expansion of high-margin products built on its technological competitiveness\n\n(45) From a short-term perspective into early 2026, however, SEC could outperform SKH on expectations of HBM share gains and narrowing margin gaps between the two\n\n(46) From a mid-term perspective, SK hynix is better positioned. This is because it has robust technological leadership in HBM and is the furthest ahead in MR-MUF-based 16-high products", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.024155468059003038, "ret_signed_m1d": 0.024155468059003038, "alpha_spy_m1d": 0.023991855194628764, "alpha_c_m1d": 0.02461929703061072, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0011302358516258604, "ret_signed_p1d": 0.0011302358516258604, "alpha_spy_p1d": 0.010446759782588022, "alpha_c_p1d": 0.014696368084145539, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.03117215984202204, "ret_signed_p1w": -0.03117215984202204, "alpha_spy_p1w": -0.011973729502368635, "alpha_c_p1w": 0.023468703250786316, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": 0.08185131724676364, "ret_signed_p1m": 0.08185131724676364, "alpha_spy_p1m": 0.06875479716723554, "alpha_c_p1m": 0.05889351352619321, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.3064940684772919, "ret_signed_p3m": 0.3064940684772919, "alpha_spy_p3m": 0.27364140921556146, "alpha_c_p3m": 0.10042240792165558, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.6315392217129785, "ret_signed_p6m": 0.6315392217129785, "alpha_spy_p6m": 0.5250109862069223, "alpha_c_p6m": -0.04421939200010239, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.2277, -0.2256, -0.2309, -0.2328, -0.2246, -0.2171, -0.2164, -0.2119, -0.2313, -0.2313, -0.2539, -0.256, -0.2301, -0.2021, -0.1987, -0.1909, -0.1778, -0.1543, -0.1501, -0.1282, -0.1267, -0.1273, -0.0773, -0.0789, -0.0556, -0.0556, -0.0579, -0.0663, -0.0617, -0.0619, -0.0491, -0.0048, 0.0044, -0.0289, 0.0048, -0.0437, -0.0632, -0.0713, -0.1336, -0.0964, -0.0957, -0.0417, -0.0313, -0.0244, 0.0164, 0.0116, -0.0176, 0.0218, 0.0428, 0.0712, 0.0632, 0.089, 0.0706, 0.0656, 0.0748, 0.0522, 0.0818, 0.0635, 0.0521, 0.0737, 0.0499, 0.057, 0.0242, 0.0, 0.0011, -0.0082, 0.0309, -0.0265, -0.0312, 0.0038, 0.0119, 0.0235, 0.0235, 0.038, 0.0219, 0.0507, 0.0701, 0.0668, 0.0724, 0.0814, 0.0823, 0.094, 0.1005, 0.0543, 0.0819, 0.0803, 0.035, 0.0794, 0.1, 0.1177, 0.1204, 0.1277, 0.1277, 0.13, 0.113, 0.0982, 0.073, 0.073, 0.1254, 0.1943, 0.2319, 0.2007, 0.1697, 0.2363, 0.2612, 0.2732, 0.2668, 0.3643, 0.3845, 0.3845, 0.3124, 0.3423, 0.3246, 0.3359, 0.3626, 0.4273, 0.4369, 0.4877, 0.261, 0.2455, 0.197, 0.1544, 0.1754, 0.2742, 0.2718, 0.2635, 0.3065, 0.2812, 0.2929, 0.2929, 0.3, 0.3089, 0.2997, 0.3228, 0.3154, 0.3322, 0.3676, 0.3478, 0.348, 0.3572, 0.2745, 0.305, 0.2639, 0.1889, 0.2636, 0.2847, 0.2954, 0.2464, 0.2544, 0.2717, 0.3098, 0.3107, 0.3365, 0.3251, 0.3364, 0.3848, 0.3651, 0.2831, 0.2761, 0.2225, 0.3019, 0.3439, 0.3412, 0.3412, 0.3617, 0.3695, 0.4787, 0.5273, 0.5361, 0.564, 0.588, 0.5457, 0.534, 0.584, 0.5963, 0.5786, 0.6022, 0.6052, 0.711, 0.68, 0.5995, 0.6059, 0.5477, 0.5264, 0.5149, 0.5323, 0.606, 0.5591, 0.6528, 0.6315]}
{"unit_id": "orig:1975206586818982075", "ticker_idx": 0, "ts": "2025-10-06T14:28:42+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The final tranche will only be issued once AMD's stock reaches $600. Based on projected quarterly earnings, this deal adds approximately $4.5 billion per quarter", "tweet_type": "news_reaction", "timeline": "years", "summary": "Barclays raises AMD PT to $300; deal with OpenAI adds ~$4.5B/quarter by end-2026, warrant structure targets $600 stock price, +$1.30 EPS/quarter from 6GW deployment by 2030.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Barclays) Raises AMD Target Price to $300\n\nThis agreement is structured to be mutually beneficial for both OpenAI and AMD, but more directly, it aims to boost AMD’s share price. The unique aspect of this deal is that warrants (stock options) are issued in tranches tied both to stock price milestones and 1GW (gigawatt) deployment intervals. The final tranche will only be issued once AMD’s stock reaches $600.\n\nBased on projected quarterly earnings, this deal adds approximately $4.5 billion per quarter to the previously expected $3 billion by the end of 2026. Assuming 6GW is gradually deployed by the end of 2030 (when the warrants expire), the incremental impact on EPS (Earnings Per Share) is about +$1.30 per quarter, bringing AMD’s total quarterly EPS to around $3.00.\n\nWe assume an annual deployment rate of about 1.2GW (6GW over 5 years), translating to roughly $18 billion in annual revenue (each GW corresponds to low double-digit billions in revenue, assumed at $15 billion per GW). We also estimate an additional 27 million diluted shares per year due to warrant issuance.\n\nFinally, we expect 1GW to be deployed in H2 2026, potentially adding around $15 billion in incremental revenue next year.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.19164502114890114, "ret_signed_m1d": -0.19164502114890114, "alpha_spy_m1d": -0.1880716073185521, "alpha_c_m1d": -0.1721330726044944, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.019511948544406743, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03828966440477655, "ret_signed_p1d": 0.03828966440477655, "alpha_spy_p1d": 0.041997156655738443, "alpha_c_p1d": 0.056724049510276986, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.01843438510550044, "ret_p1w": 0.06239257295264178, "ret_signed_p1w": 0.06239257295264178, "alpha_spy_p1w": 0.07515296339046496, "alpha_c_p1w": 0.07296396329119392, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.010571390338552145, "ret_p1m": 0.22748021604545032, "ret_signed_p1m": 0.22748021604545032, "alpha_spy_p1m": 0.22207521970038968, "alpha_c_p1m": 0.20016348808328077, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.02731672796216955, "ret_p3m": 0.05129839774106104, "ret_signed_p3m": 0.05129839774106104, "alpha_spy_p3m": 0.03294692802404775, "alpha_c_p3m": -0.0007447871716601551, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.052043184912721197, "ret_p6m": -0.0013745718367151882, "ret_signed_p6m": -0.0013745718367151882, "alpha_spy_p6m": 0.02478187053542913, "alpha_c_p6m": -0.12139607853018763, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.12002150669347245, "price_path": [-0.3206, -0.2923, -0.2812, -0.2821, -0.2361, -0.2142, -0.2126, -0.2293, -0.2293, -0.2405, -0.2212, -0.2042, -0.1828, -0.1475, -0.129, -0.1188, -0.1345, -0.1571, -0.1322, -0.1443, -0.1993, -0.1537, -0.1519, -0.1543, -0.1412, -0.0947, -0.1117, -0.1286, -0.1353, -0.1824, -0.189, -0.1964, -0.1765, -0.1981, -0.1821, -0.1796, -0.1725, -0.2017, -0.2017, -0.2032, -0.2041, -0.2058, -0.2581, -0.2567, -0.2351, -0.2168, -0.2358, -0.2216, -0.2089, -0.2123, -0.2187, -0.2248, -0.2274, -0.2156, -0.2102, -0.2102, -0.2083, -0.2172, -0.2079, -0.2058, -0.1949, -0.1668, -0.1916, 0.0, 0.0383, 0.1563, 0.1432, 0.0549, 0.0624, 0.0706, 0.1713, 0.1514, 0.1442, 0.1809, 0.1685, 0.1302, 0.1536, 0.2416, 0.2747, 0.2666, 0.2976, 0.251, 0.2573, 0.2746, 0.2275, 0.2583, 0.1669, 0.1464, 0.1977, 0.166, 0.2709, 0.2172, 0.2116, 0.1807, 0.1305, 0.0974, 0.0113, 0.0003, 0.0557, 0.0119, 0.0517, 0.0517, 0.0678, 0.0788, 0.0566, 0.0682, 0.0602, 0.07, 0.0854, 0.0879, 0.0869, 0.087, 0.0347, 0.019, 0.0268, -0.0275, -0.013, 0.0477, 0.0552, 0.0549, 0.0556, 0.0556, 0.0554, 0.0584, 0.0571, 0.0513, 0.0513, 0.097, 0.0853, 0.0522, 0.031, 0.0048, -0.0027, 0.0195, 0.0847, 0.0976, 0.1188, 0.138, 0.138, 0.1385, 0.2263, 0.2455, 0.2748, 0.2337, 0.2372, 0.2407, 0.2379, 0.1621, 0.2089, 0.1885, -0.0173, -0.055, 0.0232, 0.0603, 0.0484, 0.0485, 0.0109, 0.0177, 0.0177, -0.0031, -0.0176, -0.0017, -0.0175, -0.0349, 0.0497, 0.0351, -0.0001, -0.0172, -0.025, -0.0626, -0.0081, -0.0209, -0.0554, -0.0051, -0.0024, 0.0055, -0.0293, -0.0507, -0.035, -0.0363, -0.0209, 0.0077, -0.0117, -0.0051, 0.0081, 0.0813, 0.0003, -0.0084, -0.0377, -0.0014]}
{"unit_id": "thread:1997897553044726179", "ticker_idx": 0, "ts": "2025-12-08T05:14:30+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's early capacity shift allows it to earn more from DDR5 without the yield loss associated with HBM production, benefiting overall operations", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung benefits from DDR5 shift; SK Hynix and Micron benefit from Samsung's HBM retreat; Winbond benefits from structural DRAM shortages and rising prices.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung shifts focus from HBM to DDR5 modules for higher profits\n\nSamsung Electronics has shifted its internal strategy in response to intensified HBM market competition, reallocating capacity toward DDR5 RDIMM modules and freeing up around 80,000 DRAM wafers monthly to yield stronger profits.\n\nSoaring DDR5 prices outpace HBM margins\nThe official price of a 64GB RDIMM has risen from about US$265 in the third quarter of 2025 to US$450 in the fourth, nearly a 70% jump. DDR5 is now contributing more profit to Samsung than HBM3E, and future quarterly price increases could push modules toward US$500.\n\nSamsung has been struggling to catch up to SK Hynix and Micron in HBM3E. Although its HBM3E finally passed Nvidia and US CSP certifications in the second half of 2025, market sources say Samsung has been slashing prices to expand HBM share. Its HBM3E ASP is roughly 30% lower than SK Hynix's, and further price cuts of up to 30% are expected starting in 2026.\n\nProcess migration frees capacity\nAdditionally, Samsung has reportedly decided to shift 30–40% of its DRAM capacity from the 1a process to next-generation 1b or 1c nodes, with 1c set to become the primary focus. Through this process transition, Samsung is expected to free up roughly 80,000 wafers per month and redirect them into general-purpose DRAM, including DDR5, LPDDR5X, LPDDR6, and GDDR7.\n\nAccording to sources, Samsung's HBM3/HBM3E margins have fallen to around 30%, while prices for general DRAM have surged far beyond HBM profitability. Based on Samsung's fourth quarter 2025 price of US$450 for a 64GB RDIMM, gross margins exceed 75%. Prices are expected to rise to around US$480 in the first quarter of 2026. Although the pace of increase is slowing, spot market prices have already reached US$780, meaning Samsung could soon surpass US$500 per module.\n\nHowever, Samsung has not secured a dominant position in HBM3E, leading to insufficient capacity utilization. This indicates that its cost structure and competitiveness still need improvement. While major DRAM makers are preparing for competition in 2026 regarding sixth-generation HBM4, HBM3E will remain the mainstream in terms of annual shipments, with HBM4 volume ramping only in the second half of 2026. Samsung's early capacity shift allows it to earn more from DDR5 without the yield loss associated with HBM production, benefiting overall operations.\n\nThis also means Samsung will concentrate resources on HBM4 and DDR5 RDIMMs in 2026, slowing its push to expand HBM market share. SK Hynix and Micron are expected to benefit, and other server customers seeking HBM may have to pay premiums to secure supply from those two vendors.\n\nSK Hynix expands capacity as tight supply persists\nOn the other hand, SK Hynix has announced plans to expand general DRAM capacity. Unlike Samsung, which uses 1c for HBM3E, SK Hynix plans to use its 1c node mainly for DDR5, LPDDR5X, and GDDR7. Though the company has not revealed the scale of the expansion, market estimates indicate that capacity will grow from 20,000 wafers to 190,000 wafers by the end of 2026.\n\nEven with new DRAM capacity, most of it will be absorbed immediately by the AI sector. Only server DDR5 supply may see minor relief, while PC and smartphone DRAM will remain tight through 2026. Electronics system makers without a server business will be effectively excluded from priority supplier lists, leading to rising PC and smartphone costs and likely weakening end-market demand.\n\nTaiwanese memory suppliers see strong revenue growth\nTaiwanese suppliers continue to benefit from structural shortages and price increases. Winbond's November 2025 consolidated revenue reached NT$8.629 billion (US$276.3 million). This was the highest in 41 months, up 4.9% month-on-month and 38.7% year-on-year. Cumulative revenue for the first eleven months of 2025 reached NT$79.635 billion, up about 5.85% from the same period in 2024.\n\nFurthermore, memory module maker Team Group posted November 2025 revenue of NT$1.617 billion, up 46.76% from the previous month and 78.63% year-on-year. The company said its long-term cooperation with upstream suppliers has ensured sufficient future supply and safety stock, enabling timely deliveries. It will continue adjusting strategies and turnover efficiency based on market price trends.\n\n---\n\nhttps://t.co/7KlAPJ8Zo4", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010045630607961487, "ret_signed_m1d": -0.010045630607961487, "alpha_spy_m1d": -0.013058961003387548, "alpha_c_m1d": -0.0030687442129728737, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.006976886394988613, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010045630607961487, "ret_signed_p1d": -0.010045630607961487, "alpha_spy_p1d": -0.009182485640999882, "alpha_c_p1d": -0.012687511204079183, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.002641880596117696, "ret_p1w": -0.0429223704926619, "ret_signed_p1w": -0.0429223704926619, "alpha_spy_p1w": -0.0386802729201976, "alpha_c_p1w": -0.006818672792621605, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.036103697700040294, "ret_p1m": 0.2746594766938337, "ret_signed_p1m": 0.2746594766938337, "alpha_spy_p1m": 0.25970371455694075, "alpha_c_p1m": 0.27979098513613987, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": -0.005131508442306143, "ret_p3m": 0.758277682526922, "ret_signed_p3m": 0.758277682526922, "alpha_spy_p3m": 0.7587264496698403, "alpha_c_p3m": 0.8073014097032605, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": -0.049023727176338516, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3399, -0.3327, -0.3145, -0.3045, -0.2781, -0.289, -0.2699, -0.2699, -0.2408, -0.2299, -0.2236, -0.2172, -0.2426, -0.2311, -0.2338, -0.2146, -0.1804, -0.1804, -0.1804, -0.1804, -0.1804, -0.1804, -0.1379, -0.1479, -0.1635, -0.1324, -0.1078, -0.1059, -0.1041, -0.1096, -0.0995, -0.1187, -0.0977, -0.0685, -0.0913, -0.0822, -0.0493, -0.0183, 0.0146, -0.042, -0.0813, -0.0941, -0.1059, -0.0813, -0.0548, -0.0584, -0.0612, -0.1123, -0.0813, -0.1068, -0.1187, -0.0813, -0.1342, -0.1169, -0.0932, -0.0612, -0.0548, -0.0822, -0.0795, -0.0557, -0.0457, -0.0402, -0.01, 0.0, -0.01, -0.0137, -0.0201, -0.0055, -0.0429, -0.0612, -0.0146, -0.0174, -0.0292, 0.0091, 0.0183, 0.0146, 0.0146, 0.0685, 0.0966, 0.1003, 0.1003, 0.1003, 0.1792, 0.2673, 0.2747, 0.2939, 0.2737, 0.2756, 0.2737, 0.2627, 0.2875, 0.3205, 0.3664, 0.3701, 0.3325, 0.3719, 0.3976, 0.3958, 0.3958, 0.4637, 0.4903, 0.4747, 0.4729, 0.3802, 0.5371, 0.5518, 0.4619, 0.4554, 0.527, 0.5215, 0.5399, 0.639, 0.6628, 0.6628, 0.6628, 0.6628, 0.7436, 0.7445, 0.7711, 0.8354, 0.8675, 1.0005, 0.9868, 0.9868, 0.7904, 0.5802, 0.7583, 0.7271, 0.5922, 0.7243, 0.7436, 0.7243, 0.6839, 0.7317, 0.7794, 0.9134, 0.84, 0.8299, 0.7096, 0.7408, 0.7344, 0.6527, 0.6527, 0.6212, 0.5375, 0.744, 0.6405, 0.7123, 0.7757, 0.807, 0.9357, 0.8759, 0.8943, 0.8484, 0.8989, 0.9403, 1.0001, 0.9863, 0.9725, 1.0139, 1.0001, 1.0645, 1.0185, 1.0645, 1.0415, 1.0783, 1.0277, 1.0277, 1.138, 1.138, 1.4461, 1.4967, 1.4691, 1.6254, 1.5656, 1.6116, 1.722, 1.4875, 1.584, 1.5334, 1.538, 1.7541, 1.6898, 1.6898, 1.7495, 1.8231, 1.7541, 1.9151, null, null]}
{"unit_id": "thread:1997897553044726179", "ticker_idx": 1, "ts": "2025-12-08T05:14:30+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron are expected to benefit, and other server customers seeking HBM may have to pay premiums to secure supply from those two vendors", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung benefits from DDR5 shift; SK Hynix and Micron benefit from Samsung's HBM retreat; Winbond benefits from structural DRAM shortages and rising prices.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung shifts focus from HBM to DDR5 modules for higher profits\n\nSamsung Electronics has shifted its internal strategy in response to intensified HBM market competition, reallocating capacity toward DDR5 RDIMM modules and freeing up around 80,000 DRAM wafers monthly to yield stronger profits.\n\nSoaring DDR5 prices outpace HBM margins\nThe official price of a 64GB RDIMM has risen from about US$265 in the third quarter of 2025 to US$450 in the fourth, nearly a 70% jump. DDR5 is now contributing more profit to Samsung than HBM3E, and future quarterly price increases could push modules toward US$500.\n\nSamsung has been struggling to catch up to SK Hynix and Micron in HBM3E. Although its HBM3E finally passed Nvidia and US CSP certifications in the second half of 2025, market sources say Samsung has been slashing prices to expand HBM share. Its HBM3E ASP is roughly 30% lower than SK Hynix's, and further price cuts of up to 30% are expected starting in 2026.\n\nProcess migration frees capacity\nAdditionally, Samsung has reportedly decided to shift 30–40% of its DRAM capacity from the 1a process to next-generation 1b or 1c nodes, with 1c set to become the primary focus. Through this process transition, Samsung is expected to free up roughly 80,000 wafers per month and redirect them into general-purpose DRAM, including DDR5, LPDDR5X, LPDDR6, and GDDR7.\n\nAccording to sources, Samsung's HBM3/HBM3E margins have fallen to around 30%, while prices for general DRAM have surged far beyond HBM profitability. Based on Samsung's fourth quarter 2025 price of US$450 for a 64GB RDIMM, gross margins exceed 75%. Prices are expected to rise to around US$480 in the first quarter of 2026. Although the pace of increase is slowing, spot market prices have already reached US$780, meaning Samsung could soon surpass US$500 per module.\n\nHowever, Samsung has not secured a dominant position in HBM3E, leading to insufficient capacity utilization. This indicates that its cost structure and competitiveness still need improvement. While major DRAM makers are preparing for competition in 2026 regarding sixth-generation HBM4, HBM3E will remain the mainstream in terms of annual shipments, with HBM4 volume ramping only in the second half of 2026. Samsung's early capacity shift allows it to earn more from DDR5 without the yield loss associated with HBM production, benefiting overall operations.\n\nThis also means Samsung will concentrate resources on HBM4 and DDR5 RDIMMs in 2026, slowing its push to expand HBM market share. SK Hynix and Micron are expected to benefit, and other server customers seeking HBM may have to pay premiums to secure supply from those two vendors.\n\nSK Hynix expands capacity as tight supply persists\nOn the other hand, SK Hynix has announced plans to expand general DRAM capacity. Unlike Samsung, which uses 1c for HBM3E, SK Hynix plans to use its 1c node mainly for DDR5, LPDDR5X, and GDDR7. Though the company has not revealed the scale of the expansion, market estimates indicate that capacity will grow from 20,000 wafers to 190,000 wafers by the end of 2026.\n\nEven with new DRAM capacity, most of it will be absorbed immediately by the AI sector. Only server DDR5 supply may see minor relief, while PC and smartphone DRAM will remain tight through 2026. Electronics system makers without a server business will be effectively excluded from priority supplier lists, leading to rising PC and smartphone costs and likely weakening end-market demand.\n\nTaiwanese memory suppliers see strong revenue growth\nTaiwanese suppliers continue to benefit from structural shortages and price increases. Winbond's November 2025 consolidated revenue reached NT$8.629 billion (US$276.3 million). This was the highest in 41 months, up 4.9% month-on-month and 38.7% year-on-year. Cumulative revenue for the first eleven months of 2025 reached NT$79.635 billion, up about 5.85% from the same period in 2024.\n\nFurthermore, memory module maker Team Group posted November 2025 revenue of NT$1.617 billion, up 46.76% from the previous month and 78.63% year-on-year. The company said its long-term cooperation with upstream suppliers has ensured sufficient future supply and safety stock, enabling timely deliveries. It will continue adjusting strategies and turnover efficiency based on market price trends.\n\n---\n\nhttps://t.co/7KlAPJ8Zo4", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05719233447170469, "ret_signed_m1d": -0.05719233447170469, "alpha_spy_m1d": -0.06020566486713075, "alpha_c_m1d": -0.04598634336848628, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01906407531159593, "ret_signed_p1d": -0.01906407531159593, "alpha_spy_p1d": -0.018200930344634325, "alpha_c_p1d": -0.020285088195773793, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.03986126811277635, "ret_signed_p1w": -0.03986126811277635, "alpha_spy_p1w": -0.03561917054031205, "alpha_c_p1w": 0.002602403639090589, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.25823222667283874, "ret_signed_p1m": 0.25823222667283874, "alpha_spy_p1m": 0.24327646453594576, "alpha_c_p1m": 0.2032162932853212, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6338586339218024, "ret_signed_p3m": 0.6338586339218024, "alpha_spy_p3m": 0.6343074010647207, "alpha_c_p3m": 0.5578033195462861, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4735, -0.4683, -0.4311, -0.4268, -0.3973, -0.4224, -0.3843, -0.3843, -0.3921, -0.3748, -0.3809, -0.3826, -0.4172, -0.3956, -0.3982, -0.3765, -0.315, -0.315, -0.315, -0.315, -0.315, -0.315, -0.2588, -0.2813, -0.2873, -0.2683, -0.2163, -0.1938, -0.1592, -0.1704, -0.1661, -0.1713, -0.1168, -0.0735, -0.0977, -0.0336, -0.0163, -0.0319, 0.0738, 0.0149, 0.0027, 0.027, 0.0045, 0.0495, 0.072, 0.0686, 0.0599, -0.0302, 0.0495, -0.0128, -0.0267, -0.0111, -0.0977, -0.0994, -0.1012, -0.0925, -0.0572, -0.0815, -0.0676, -0.0329, -0.0433, -0.0607, -0.0572, 0.0, -0.0191, 0.0173, -0.0208, -0.0104, -0.0399, -0.0815, -0.0451, -0.0433, -0.052, 0.0052, 0.0121, 0.0191, 0.0191, 0.0381, 0.1092, 0.1282, 0.1282, 0.1282, 0.1733, 0.2062, 0.2582, 0.286, 0.3102, 0.2894, 0.2981, 0.279, 0.286, 0.2981, 0.3102, 0.3241, 0.2877, 0.2825, 0.3085, 0.3293, 0.2756, 0.3865, 0.4575, 0.4922, 0.5754, 0.4385, 0.5719, 0.5598, 0.4593, 0.4541, 0.5373, 0.5182, 0.4905, 0.539, 0.5251, 0.5251, 0.5251, 0.5251, 0.5494, 0.6447, 0.6482, 0.7418, 0.7643, 0.9082, 0.8422, 0.8422, 0.6304, 0.4741, 0.6339, 0.6043, 0.4515, 0.6286, 0.6582, 0.6148, 0.58, 0.6912, 0.6842, 0.8335, 0.7589, 0.7485, 0.62, 0.712, 0.7276, 0.62, 0.62, 0.5158, 0.4012, 0.5575, 0.4411, 0.521, 0.5384, 0.5905, 0.7936, 0.7328, 0.7832, 0.8058, 0.9151, 0.9724, 1.0054, 0.9585, 1.0245, 1.1252, 1.1235, 1.127, 1.1218, 1.2433, 1.2572, 1.245, 1.2329, 1.2329, 1.5124, 1.5124, 1.7798, 1.8718, 1.9274, 2.2642, 2.1861, 2.4309, 2.4205, 2.1583, 2.1948, 2.0298, 2.0298, 2.3684, 2.3702, 2.3702, 2.5629, 2.8945, 2.9751, 3.0515, null, null]}
{"unit_id": "thread:1997897553044726179", "ticker_idx": 2, "ts": "2025-12-08T05:14:30+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron are expected to benefit, and other server customers seeking HBM may have to pay premiums to secure supply from those two vendors", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung benefits from DDR5 shift; SK Hynix and Micron benefit from Samsung's HBM retreat; Winbond benefits from structural DRAM shortages and rising prices.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung shifts focus from HBM to DDR5 modules for higher profits\n\nSamsung Electronics has shifted its internal strategy in response to intensified HBM market competition, reallocating capacity toward DDR5 RDIMM modules and freeing up around 80,000 DRAM wafers monthly to yield stronger profits.\n\nSoaring DDR5 prices outpace HBM margins\nThe official price of a 64GB RDIMM has risen from about US$265 in the third quarter of 2025 to US$450 in the fourth, nearly a 70% jump. DDR5 is now contributing more profit to Samsung than HBM3E, and future quarterly price increases could push modules toward US$500.\n\nSamsung has been struggling to catch up to SK Hynix and Micron in HBM3E. Although its HBM3E finally passed Nvidia and US CSP certifications in the second half of 2025, market sources say Samsung has been slashing prices to expand HBM share. Its HBM3E ASP is roughly 30% lower than SK Hynix's, and further price cuts of up to 30% are expected starting in 2026.\n\nProcess migration frees capacity\nAdditionally, Samsung has reportedly decided to shift 30–40% of its DRAM capacity from the 1a process to next-generation 1b or 1c nodes, with 1c set to become the primary focus. Through this process transition, Samsung is expected to free up roughly 80,000 wafers per month and redirect them into general-purpose DRAM, including DDR5, LPDDR5X, LPDDR6, and GDDR7.\n\nAccording to sources, Samsung's HBM3/HBM3E margins have fallen to around 30%, while prices for general DRAM have surged far beyond HBM profitability. Based on Samsung's fourth quarter 2025 price of US$450 for a 64GB RDIMM, gross margins exceed 75%. Prices are expected to rise to around US$480 in the first quarter of 2026. Although the pace of increase is slowing, spot market prices have already reached US$780, meaning Samsung could soon surpass US$500 per module.\n\nHowever, Samsung has not secured a dominant position in HBM3E, leading to insufficient capacity utilization. This indicates that its cost structure and competitiveness still need improvement. While major DRAM makers are preparing for competition in 2026 regarding sixth-generation HBM4, HBM3E will remain the mainstream in terms of annual shipments, with HBM4 volume ramping only in the second half of 2026. Samsung's early capacity shift allows it to earn more from DDR5 without the yield loss associated with HBM production, benefiting overall operations.\n\nThis also means Samsung will concentrate resources on HBM4 and DDR5 RDIMMs in 2026, slowing its push to expand HBM market share. SK Hynix and Micron are expected to benefit, and other server customers seeking HBM may have to pay premiums to secure supply from those two vendors.\n\nSK Hynix expands capacity as tight supply persists\nOn the other hand, SK Hynix has announced plans to expand general DRAM capacity. Unlike Samsung, which uses 1c for HBM3E, SK Hynix plans to use its 1c node mainly for DDR5, LPDDR5X, and GDDR7. Though the company has not revealed the scale of the expansion, market estimates indicate that capacity will grow from 20,000 wafers to 190,000 wafers by the end of 2026.\n\nEven with new DRAM capacity, most of it will be absorbed immediately by the AI sector. Only server DDR5 supply may see minor relief, while PC and smartphone DRAM will remain tight through 2026. Electronics system makers without a server business will be effectively excluded from priority supplier lists, leading to rising PC and smartphone costs and likely weakening end-market demand.\n\nTaiwanese memory suppliers see strong revenue growth\nTaiwanese suppliers continue to benefit from structural shortages and price increases. Winbond's November 2025 consolidated revenue reached NT$8.629 billion (US$276.3 million). This was the highest in 41 months, up 4.9% month-on-month and 38.7% year-on-year. Cumulative revenue for the first eleven months of 2025 reached NT$79.635 billion, up about 5.85% from the same period in 2024.\n\nFurthermore, memory module maker Team Group posted November 2025 revenue of NT$1.617 billion, up 46.76% from the previous month and 78.63% year-on-year. The company said its long-term cooperation with upstream suppliers has ensured sufficient future supply and safety stock, enabling timely deliveries. It will continue adjusting strategies and turnover efficiency based on market price trends.\n\n---\n\nhttps://t.co/7KlAPJ8Zo4", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03928399978675012, "ret_signed_m1d": -0.03928399978675012, "alpha_spy_m1d": -0.042297330182176185, "alpha_c_m1d": -0.02807800868353172, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022274411827899465, "ret_signed_p1d": 0.022274411827899465, "alpha_spy_p1d": 0.02313755679486107, "alpha_c_p1d": 0.021053398943721602, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.03814996540873539, "ret_signed_p1w": -0.03814996540873539, "alpha_spy_p1w": -0.03390786783627109, "alpha_c_p1w": 0.0043137063431315514, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.3914171360164642, "ret_signed_p1m": 0.3914171360164642, "alpha_spy_p1m": 0.37646137387957124, "alpha_c_p1m": 0.3364012026289467, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6086602969591723, "ret_signed_p3m": 0.6086602969591723, "alpha_spy_p3m": 0.6091090641020906, "alpha_c_p3m": 0.5326049825836561, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4334, -0.3906, -0.3636, -0.3614, -0.3572, -0.3525, -0.3164, -0.3414, -0.3337, -0.3265, -0.3455, -0.3653, -0.3635, -0.3366, -0.3228, -0.2628, -0.2563, -0.2393, -0.2266, -0.248, -0.204, -0.2211, -0.2645, -0.2193, -0.2424, -0.2227, -0.1798, -0.1804, -0.1626, -0.1807, -0.1962, -0.1628, -0.113, -0.1086, -0.1013, -0.0822, -0.0928, -0.0938, -0.0495, -0.117, -0.0381, -0.0348, -0.0364, 0.0258, -0.0235, -0.0082, -0.0404, -0.0004, -0.0201, -0.0746, -0.085, -0.1845, -0.1602, -0.0931, -0.0907, -0.0675, -0.0675, -0.0423, -0.0262, -0.0301, -0.0517, -0.0821, -0.0393, 0.0, 0.0223, 0.068, 0.0467, -0.0234, -0.0381, -0.0584, -0.0867, 0.0066, 0.0769, 0.1202, 0.1189, 0.161, 0.161, 0.1534, 0.1926, 0.1856, 0.1563, 0.1563, 0.2779, 0.2647, 0.3914, 0.3757, 0.3249, 0.3981, 0.4013, 0.3699, 0.3506, 0.3639, 0.4697, 0.4697, 0.4788, 0.5765, 0.6108, 0.6192, 0.5764, 0.6621, 0.7636, 0.7656, 0.6809, 0.7738, 0.6994, 0.5372, 0.5513, 0.5991, 0.5538, 0.5122, 0.6625, 0.6772, 0.6679, 0.6679, 0.6197, 0.7055, 0.6909, 0.7347, 0.7056, 0.6936, 0.7381, 0.6837, 0.6707, 0.6719, 0.5383, 0.6237, 0.6087, 0.5003, 0.5773, 0.6332, 0.6963, 0.6423, 0.7265, 0.79, 0.8706, 0.8707, 0.8, 0.7134, 0.6382, 0.6025, 0.548, 0.4402, 0.4473, 0.3043, 0.3693, 0.491, 0.4845, 0.4845, 0.5311, 0.5304, 0.6486, 0.7085, 0.7047, 0.7289, 0.8874, 0.8492, 0.8533, 0.8445, 0.8176, 0.8214, 0.9759, 0.9525, 1.0133, 1.1262, 1.044, 1.1014, 1.0962, 1.1977, 1.3365, 1.5949, 1.7018, 1.6209, 2.027, 2.2237, 2.1071, 2.2573, 2.1453, 1.9372, 1.7624, 1.8322, 1.9669, 2.089, 2.044, 2.044, 2.6312, 2.7631, 2.7432, 2.9357, null, null]}
{"unit_id": "thread:1997897553044726179", "ticker_idx": 3, "ts": "2025-12-08T05:14:30+00:00", "symbol": "Winbond", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwanese suppliers continue to benefit from structural shortages and price increases", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung benefits from DDR5 shift; SK Hynix and Micron benefit from Samsung's HBM retreat; Winbond benefits from structural DRAM shortages and rising prices.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung shifts focus from HBM to DDR5 modules for higher profits\n\nSamsung Electronics has shifted its internal strategy in response to intensified HBM market competition, reallocating capacity toward DDR5 RDIMM modules and freeing up around 80,000 DRAM wafers monthly to yield stronger profits.\n\nSoaring DDR5 prices outpace HBM margins\nThe official price of a 64GB RDIMM has risen from about US$265 in the third quarter of 2025 to US$450 in the fourth, nearly a 70% jump. DDR5 is now contributing more profit to Samsung than HBM3E, and future quarterly price increases could push modules toward US$500.\n\nSamsung has been struggling to catch up to SK Hynix and Micron in HBM3E. Although its HBM3E finally passed Nvidia and US CSP certifications in the second half of 2025, market sources say Samsung has been slashing prices to expand HBM share. Its HBM3E ASP is roughly 30% lower than SK Hynix's, and further price cuts of up to 30% are expected starting in 2026.\n\nProcess migration frees capacity\nAdditionally, Samsung has reportedly decided to shift 30–40% of its DRAM capacity from the 1a process to next-generation 1b or 1c nodes, with 1c set to become the primary focus. Through this process transition, Samsung is expected to free up roughly 80,000 wafers per month and redirect them into general-purpose DRAM, including DDR5, LPDDR5X, LPDDR6, and GDDR7.\n\nAccording to sources, Samsung's HBM3/HBM3E margins have fallen to around 30%, while prices for general DRAM have surged far beyond HBM profitability. Based on Samsung's fourth quarter 2025 price of US$450 for a 64GB RDIMM, gross margins exceed 75%. Prices are expected to rise to around US$480 in the first quarter of 2026. Although the pace of increase is slowing, spot market prices have already reached US$780, meaning Samsung could soon surpass US$500 per module.\n\nHowever, Samsung has not secured a dominant position in HBM3E, leading to insufficient capacity utilization. This indicates that its cost structure and competitiveness still need improvement. While major DRAM makers are preparing for competition in 2026 regarding sixth-generation HBM4, HBM3E will remain the mainstream in terms of annual shipments, with HBM4 volume ramping only in the second half of 2026. Samsung's early capacity shift allows it to earn more from DDR5 without the yield loss associated with HBM production, benefiting overall operations.\n\nThis also means Samsung will concentrate resources on HBM4 and DDR5 RDIMMs in 2026, slowing its push to expand HBM market share. SK Hynix and Micron are expected to benefit, and other server customers seeking HBM may have to pay premiums to secure supply from those two vendors.\n\nSK Hynix expands capacity as tight supply persists\nOn the other hand, SK Hynix has announced plans to expand general DRAM capacity. Unlike Samsung, which uses 1c for HBM3E, SK Hynix plans to use its 1c node mainly for DDR5, LPDDR5X, and GDDR7. Though the company has not revealed the scale of the expansion, market estimates indicate that capacity will grow from 20,000 wafers to 190,000 wafers by the end of 2026.\n\nEven with new DRAM capacity, most of it will be absorbed immediately by the AI sector. Only server DDR5 supply may see minor relief, while PC and smartphone DRAM will remain tight through 2026. Electronics system makers without a server business will be effectively excluded from priority supplier lists, leading to rising PC and smartphone costs and likely weakening end-market demand.\n\nTaiwanese memory suppliers see strong revenue growth\nTaiwanese suppliers continue to benefit from structural shortages and price increases. Winbond's November 2025 consolidated revenue reached NT$8.629 billion (US$276.3 million). This was the highest in 41 months, up 4.9% month-on-month and 38.7% year-on-year. Cumulative revenue for the first eleven months of 2025 reached NT$79.635 billion, up about 5.85% from the same period in 2024.\n\nFurthermore, memory module maker Team Group posted November 2025 revenue of NT$1.617 billion, up 46.76% from the previous month and 78.63% year-on-year. The company said its long-term cooperation with upstream suppliers has ensured sufficient future supply and safety stock, enabling timely deliveries. It will continue adjusting strategies and turnover efficiency based on market price trends.\n\n---\n\nhttps://t.co/7KlAPJ8Zo4", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0898380018785927, "ret_signed_m1d": -0.0898380018785927, "alpha_spy_m1d": -0.09285133227401876, "alpha_c_m1d": -0.0786320107753743, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05007366631441168, "ret_signed_p1d": 0.05007366631441168, "alpha_spy_p1d": 0.050936811281373284, "alpha_c_p1d": 0.04885265343023382, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": 0.04712814324291714, "ret_signed_p1w": 0.04712814324291714, "alpha_spy_p1w": 0.05137024081538144, "alpha_c_p1w": 0.08959181499478408, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.5316642035928199, "ret_signed_p1m": 0.5316642035928199, "alpha_spy_p1m": 0.5167084414559269, "alpha_c_p1m": 0.4766482702053023, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6494845617001128, "ret_signed_p3m": 0.6494845617001128, "alpha_spy_p3m": 0.6499333288430311, "alpha_c_p3m": 0.5734292473245965, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6325, -0.6318, -0.6252, -0.6053, -0.5884, -0.567, -0.5243, -0.5066, -0.4956, -0.4632, -0.4816, -0.5081, -0.5376, -0.5376, -0.4985, -0.5022, -0.4698, -0.4286, -0.4286, -0.3719, -0.3925, -0.3601, -0.3601, -0.3667, -0.4006, -0.4109, -0.352, -0.3527, -0.3174, -0.3432, -0.3306, -0.3181, -0.3181, -0.2504, -0.2047, -0.2047, -0.187, -0.2018, -0.1664, -0.215, -0.1753, -0.1311, -0.1443, -0.0589, -0.0501, -0.0457, -0.0545, -0.1119, -0.025, -0.0957, -0.1252, -0.1458, -0.2312, -0.1959, -0.1532, -0.2106, -0.162, -0.1458, -0.1517, -0.1664, -0.1782, -0.1487, -0.0898, 0.0, 0.0501, -0.0059, 0.0074, 0.0987, 0.0471, 0.0398, 0.0884, 0.0972, 0.0501, 0.0766, 0.0751, 0.1325, 0.1325, 0.1267, 0.1325, 0.2327, 0.2165, 0.2165, 0.3373, 0.4065, 0.5317, 0.5685, 0.5979, 0.4404, 0.5022, 0.4698, 0.4698, 0.461, 0.5685, 0.7158, 0.6642, 0.5538, 0.539, 0.5169, 0.6642, 0.701, 0.8704, 0.9146, 0.8925, 0.7084, 0.5538, 0.5979, 0.5096, 0.4404, 0.5832, 0.5243, 0.5538, 0.5538, 0.5538, 0.5538, 0.5538, 0.5538, 0.5538, 0.5538, 0.6937, 0.7968, 0.7378, 0.8041, 0.8041, 0.8336, 0.6568, 0.5317, 0.6495, 0.5685, 0.4875, 0.5538, 0.6789, 0.6053, 0.6053, 0.7231, 0.8189, 0.8851, 0.7968, 0.62, 0.5832, 0.4654, 0.4462, 0.4138, 0.3679, 0.3768, 0.3354, 0.3783, 0.3591, 0.3591, 0.3591, 0.3235, 0.4153, 0.3472, 0.3842, 0.3976, 0.3842, 0.3354, 0.3265, 0.2998, 0.2688, 0.3487, 0.3413, 0.2998, 0.3058, 0.3902, 0.4138, 0.3724, 0.3295, 0.3295, 0.4124, 0.4627, 0.6063, 0.6877, 0.5841, 0.7395, 0.7988, 0.8062, 0.9838, 0.9172, 0.932, 0.7395, 0.7099, 0.6877, 0.8506, 0.9024, 1.0875, 1.2947, 1.1319, 1.3391, null, null]}
{"unit_id": "orig:1974712749168234614", "ticker_idx": 0, "ts": "2025-10-05T05:46:22+00:00", "symbol": "South Korea battery and LCD equipment", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I expect Korea to benefit from this trend. After all, we're already the world's second-largest battery producer, and we were manufacturing LCDs until just a few years ago.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Korean battery and LCD equipment sector as China restricts tech exports, driving Indian/Vietnamese buyers to Korean suppliers.", "resolved_tickers": ["006400.KS", "051910.KS", "373220.KS"], "resolution_method": "basket", "resolution_reason": "Korea battery/LCD equipment basket: Samsung SDI, LG Chem, LG Energy", "bench_c": "XLI", "bench_c_reason": "industry='Electrical Equipment & Parts'", "bench_c_industry": "Electrical Equipment & Parts", "unit_kind": "original", "lang": "en", "tweet_text": "Here’s one trend you probably don’t know about:\n\nChina is now heavily restricting the export of its own companies’ battery manufacturing and LCD display equipment to countries like Vietnam and India.\n\nBeijing is deeply worried that its advanced industrial technologies could leak abroad.\n\nAs a result, many Indian and Vietnamese companies have recently been visiting South Korea.\n\nIn fact, not long ago, Tata Group, one of India’s largest conglomerates, came to Korea to conduct an on-site inspection to purchase Korean battery manufacturing equipment — reportedly worth several billion dollars.\n\nI expect Korea to benefit from this trend. After all, we’re already the world’s second-largest battery producer, and we were manufacturing LCDs until just a few years ago.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.003573413830349037, "alpha_c_m1d": 0.004641259253716612, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.004641259253716612, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.003707492250961897, "alpha_c_p1d": 0.005672683190828365, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.005672683190828365, "ret_p1w": -0.02071926129463275, "ret_signed_p1w": -0.02071926129463275, "alpha_spy_p1w": -0.007958870856809575, "alpha_c_p1w": 0.0029382643885259574, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.02365752568315871, "ret_p1m": 0.41442319630505436, "ret_signed_p1m": 0.41442319630505436, "alpha_spy_p1m": 0.4090181999599937, "alpha_c_p1m": 0.4306676531808246, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": -0.016244456875770252, "ret_p3m": 0.1428811841121248, "ret_signed_p3m": 0.1428811841121248, "alpha_spy_p3m": 0.12452971439511151, "alpha_c_p3m": 0.13952086609188863, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.003360318020236175, "ret_p6m": 0.34963843207688733, "ret_signed_p6m": 0.34963843207688733, "alpha_spy_p6m": 0.37579487444903165, "alpha_c_p6m": 0.30058329202058204, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.04905514005630529, "price_path": [-0.1583, -0.1303, -0.1151, -0.116, -0.1184, -0.1394, -0.1225, -0.1087, -0.0788, -0.0806, -0.0713, -0.019, -0.0212, 0.0078, 0.0059, 0.0284, 0.0062, -0.0333, -0.0259, 0.0255, 0.0343, 0.0245, 0.0002, 0.0215, 0.0122, 0.0187, 0.0301, 0.0301, 0.0058, 0.0109, 0.007, 0.011, 0.001, 0.0264, 0.0228, 0.0113, -0.0091, -0.0385, -0.0546, -0.057, -0.0593, -0.046, -0.0617, -0.0594, -0.0329, -0.051, -0.0392, -0.0254, -0.0252, -0.0328, -0.0276, -0.0167, -0.0167, -0.0191, -0.0042, -0.034, -0.0069, -0.0422, -0.0375, -0.0448, -0.0423, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0304, -0.0207, 0.0329, 0.0336, 0.107, 0.1822, 0.1992, 0.1988, 0.2731, 0.2582, 0.3657, 0.3589, 0.4264, 0.4646, 0.4183, 0.3924, 0.3972, 0.4144, 0.3747, 0.3755, 0.34, 0.3742, 0.3948, 0.3889, 0.4486, 0.3848, 0.3814, 0.3228, 0.3132, 0.3226, 0.262, 0.2284, 0.2407, 0.3322, 0.3301, 0.2736, 0.2745, 0.2853, 0.298, 0.2825, 0.3105, 0.3483, 0.3268, 0.333, 0.3088, 0.3156, 0.302, 0.2482, 0.2605, 0.1633, 0.1744, 0.197, 0.1903, 0.1938, 0.1938, 0.1663, 0.172, 0.1429, 0.1429, 0.1429, 0.1127, 0.1507, 0.1575, 0.13, 0.1216, 0.1104, 0.1516, 0.2248, 0.1832, 0.2016, 0.1938, 0.2452, 0.2701, 0.2502, 0.3901, 0.3763, 0.3921, 0.38, 0.4253, 0.3863, 0.335, 0.2483, 0.2993, 0.3502, 0.327, 0.2841, 0.3159, 0.3268, 0.3234, 0.369, 0.3243, 0.3243, 0.3243, 0.3243, 0.3975, 0.3891, 0.3622, 0.446, 0.5203, 0.55, 0.6104, 0.6104, 0.4206, 0.2257, 0.3337, 0.3702, 0.2886, 0.3315, 0.3333, 0.3341, 0.2919, 0.2748, 0.3152, 0.3585, 0.3242, 0.3326, 0.2535, 0.3467, 0.3633, 0.3384, 0.3456, 0.3977, 0.3496]}
{"unit_id": "thread:1977302176071590083", "ticker_idx": 0, "ts": "2025-10-12T09:15:49+00:00", "symbol": "Fujikura", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still think it's an attractive stock", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Fujikura: 50%+ share in optical fiber fusion splicers makes it a major beneficiary of surging optical fiber demand; still attractive despite 255% run.", "resolved_tickers": ["5803.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Fujikura Ltd 5803.T Japan cables/connectivity", "bench_c": "XLI", "bench_c_reason": "industry='Conglomerates'", "bench_c_industry": "Conglomerates", "unit_kind": "self_thread", "lang": "en", "tweet_text": "I’ve been personally researching this company lately.\nGiven the surge in optical fiber demand, I believe Fujikura, which holds over 50% market share in the optical fiber fusion splicer market, will also benefit significantly.\n\n---\n\nIt’s gone up 255% over the past six months, but I still think it’s an attractive stock. https://t.co/oQswwlw3fN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.015112192089016707, "alpha_c_m1d": 0.010299879851799698, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.010299879851799698, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.049214923518699716, "ret_signed_p1d": -0.049214923518699716, "alpha_spy_p1d": -0.04799326871524301, "alpha_c_p1d": -0.06096718547178315, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": 0.011752261953083432, "ret_p1w": 0.03834534477686624, "ret_signed_p1w": 0.03834534477686624, "alpha_spy_p1w": 0.02588753935992516, "alpha_c_p1w": 0.024744430628695913, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.013600914148170329, "ret_p1m": 0.16455313252937098, "ret_signed_p1m": 0.16455313252937098, "alpha_spy_p1m": 0.13444937704252458, "alpha_c_p1m": 0.1461984978622155, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.01835463466715548, "ret_p3m": 0.061896128342474155, "ret_signed_p3m": 0.061896128342474155, "alpha_spy_p3m": 0.018901026256162412, "alpha_c_p3m": 0.0007672340181485993, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.061128894324325556, "ret_p6m": 0.5908181557212777, "ret_signed_p6m": 0.5908181557212777, "alpha_spy_p6m": 0.5909183247495696, "alpha_c_p6m": 0.49940232494898384, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.09141583077229387, "price_path": [-0.5229, -0.5192, -0.5151, -0.5151, -0.4936, -0.4986, -0.4779, -0.4688, -0.4713, -0.4638, -0.416, -0.3777, -0.3828, -0.384, -0.3597, -0.3444, -0.3096, -0.3345, -0.3345, -0.3024, -0.2802, -0.2859, -0.2638, -0.2563, -0.2796, -0.3204, -0.3135, -0.3111, -0.3105, -0.3087, -0.2997, -0.2611, -0.2341, -0.2554, -0.2548, -0.2503, -0.2119, -0.1999, -0.187, -0.226, -0.1825, -0.1609, -0.1543, -0.1543, -0.1621, -0.193, -0.184, -0.1618, -0.1453, -0.1453, -0.1261, -0.1273, -0.1543, -0.1135, -0.1271, -0.1368, -0.1452, -0.1464, -0.1244, -0.0767, -0.0359, 0.0136, 0.0, 0.0, -0.0492, -0.0242, 0.0145, -0.0069, 0.0383, 0.0069, 0.0226, 0.0414, 0.0824, 0.1688, 0.1591, 0.2144, 0.2563, 0.2784, 0.2784, 0.2506, 0.1845, 0.2998, 0.2328, 0.2385, 0.1646, 0.2123, 0.2503, 0.1703, 0.1981, 0.0794, 0.0773, 0.1407, 0.0462, 0.0462, 0.0495, 0.0891, 0.093, 0.0839, -0.013, 0.0024, 0.0284, 0.0236, 0.0417, 0.1144, 0.1153, 0.1147, 0.1002, 0.1171, 0.0682, -0.0036, -0.0021, -0.0362, 0.0042, 0.074, 0.0773, 0.0912, 0.064, 0.0362, 0.0752, 0.0531, 0.0531, 0.0531, 0.0531, 0.1138, 0.1171, 0.1018, 0.0619, 0.0317, 0.0317, 0.0365, 0.0444, 0.0447, 0.0694, 0.0441, 0.0193, 0.0827, 0.0969, 0.0812, 0.0845, 0.1093, 0.212, 0.1757, 0.1818, 0.1993, 0.3152, 0.3774, 0.3077, 0.3524, 0.3258, 0.3925, 0.3925, 0.3919, 0.3161, 0.295, 0.2935, 0.349, 0.3859, 0.3829, 0.3829, 0.5211, 0.6199, 0.6585, 0.6162, 0.7165, 0.6516, 0.5323, 0.5894, 0.5157, 0.3656, 0.4532, 0.5489, 0.5519, 0.6014, 0.5477, 0.4858, 0.5522, 0.5296, 0.5296, 0.4692, 0.5459, 0.6591, 0.6978, 0.6685, 0.64, 0.4889, 0.6185, 0.5628, 0.6804, 0.6745, 0.5908]}
{"unit_id": "thread:1990306908536991962", "ticker_idx": 0, "ts": "2025-11-17T06:31:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix will enjoy a supercycle in earnings that lasts at least through 2027", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM shortage intensifying; Samsung and SK Hynix expected to enjoy a supercycle in earnings through at least 2027 as long-term contracts lock in rising prices.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Shortage Worsens… Companies Shift from Monthly Contracts to “Bulk Purchases for Six Months”\n\nAs the global boom in artificial intelligence (AI) investment deepens the shortage of DRAM semiconductors, pricing negotiations that used to be conducted on a monthly or quarterly basis are being reorganized around long-term supply contracts of six months or more. With the severe DRAM supply shortage expected to drive price increases again next year, demand-side companies are actively agreeing to these semiannual contracts. As it has become difficult to secure even next year’s volumes, the market is now seeing discussions extend to supply contracts running through 2027.\n\nAn industry insider said on the 17th, “The DRAM market has shifted into a long-term contract-driven market,” adding, “We’re seeing buying demand that is even stronger than during the 2017 supercycle.” DRAM temporarily stores data so that central processing units (CPUs) or graphics processing units (GPUs) can process information quickly.\n\nWith the rise of large language model (LLM) AIs such as ChatGPT, the role of GPUs has grown, and to support them, memory semiconductors such as high bandwidth memory (HBM), which stacks DRAM in multiple layers, are facing chronic supply shortages. A representative example is the practice of global AI semiconductor leader Nvidia and other U.S. big tech companies pre-purchasing HBM from SK Hynix and Samsung Electronics through annual long-term contracts.\n\nRecently, however, the shortage has spread beyond HBM to general-purpose DRAM as well. OpenAI, Meta, and others have announced AI infrastructure investment plans measured in hundreds of trillions of won, and as major corporations and governments worldwide build data centers for their own AI development, demand for general-purpose DRAM such as double data rate (DDR), graphics (G) DDR, and low-power (LP) DDR has surged. Although general-purpose DRAM delivers lower performance than HBM, it is nonetheless indispensable for AI inference and computation.\n\nAccording to the industry, DRAM demand is rising sharply, particularly among companies in the United States and China. In particular, Samsung Electronics and SK Hynix are understood to be signing supply contracts for next year with major customers on a semiannual basis.\n\nTypically, DRAM is supplied under monthly contracts in which the product is delivered at a fixed monthly price, and then adjusted later to reflect market prices. However, as DRAM demand has surged since the second half of this year, contract periods have been extended beyond quarterly to semiannual or longer-term supply agreements. This is because demand-side companies are not only offering prices above the prevailing market level but also seeking contracts that guarantee supply for more than half a year.\n\nIn reality, DRAM inventories in the market are diminishing rapidly. As of the end of the third quarter, finished goods inventories (“products and merchandise”) at Samsung Electronics’ Device Solutions (DS) division stood at 3.4043 trillion won, down 14.6% (580.4 billion won) from the previous quarter. SK Hynix’s inventories are also falling. In the third quarter, its “products and merchandise” inventories totaled 2.1522 trillion won, down 368.9 billion won compared with the end of last year.\n\nAmid the DRAM shortage, the market is now seeing discussions around supply contracts extending out to 2027. SK Hynix has already completed its DRAM supply contracts for next year and is now negotiating supply for 2027. As SK Hynix’s DRAM has effectively sold out, global big tech companies and other customers are flocking to Samsung Electronics. However, most of the DRAM that Samsung plans to produce next year is also already under contract. With demand soaring, Samsung is even discussing plans to raise DRAM supply prices by more than 40%.\n\nThe industry expects that as the DRAM market shifts toward long-term supply contracts, Samsung Electronics and SK Hynix will enjoy a supercycle in earnings that lasts at least through 2027. When the proportion of long-term supply contracts increases, manufacturers find it easier to plan production, including unit costs and distribution, which in turn improves profitability. An industry official said, “As semiconductor contracts that used to be monthly or quarterly are renewed on a semiannual basis, prices are rising again,” adding, “Because long-term supply contracts through 2027 are being signed at prices higher than today, profitability will improve even further.”\n\n---\n\nhttps://t.co/hcNZI3UCUB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03379720954848586, "ret_signed_m1d": -0.03379720954848586, "alpha_spy_m1d": -0.04320134735436654, "alpha_c_m1d": -0.049697687662226486, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.015900478113740624, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02783296389954426, "ret_signed_p1d": -0.02783296389954426, "alpha_spy_p1d": -0.0194355033413377, "alpha_c_p1d": -0.011580009741389619, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.016252954158154642, "ret_p1w": -0.0387674012229241, "ret_signed_p1w": -0.0387674012229241, "alpha_spy_p1w": -0.043364256354718855, "alpha_c_p1w": -0.024911758881995483, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.01385564234092862, "ret_p1m": 0.021868796448681227, "ret_signed_p1m": 0.021868796448681227, "alpha_spy_p1m": 0.0020390896861928187, "alpha_c_p1m": 0.016650921047398626, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.005217875401282601, "ret_p3m": 0.7839781208032068, "ret_signed_p3m": 0.7839781208032068, "alpha_spy_p3m": 0.7575188399158186, "alpha_c_p3m": 0.8008927632320656, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": -0.01691464242885876, "ret_p6m": 1.7926097811095025, "ret_signed_p6m": 1.7926097811095025, "alpha_spy_p6m": 1.677367689847022, "alpha_c_p6m": 1.5538288929281312, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.23878088818137133, "price_path": [-0.3023, -0.3013, -0.2934, -0.2924, -0.3043, -0.3013, -0.3112, -0.3102, -0.331, -0.3162, -0.3092, -0.3063, -0.3122, -0.3063, -0.2924, -0.2815, -0.2736, -0.2538, -0.2429, -0.2142, -0.2261, -0.2053, -0.2053, -0.1737, -0.1618, -0.1549, -0.1479, -0.1756, -0.163, -0.166, -0.1451, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.0616, -0.0726, -0.0895, -0.0557, -0.0288, -0.0268, -0.0249, -0.0308, -0.0199, -0.0408, -0.0179, 0.0139, -0.0109, -0.001, 0.0348, 0.0686, 0.1044, 0.0427, 0.0, -0.0139, -0.0268, 0.0, 0.0288, 0.0249, 0.0219, -0.0338, 0.0, -0.0278, -0.0408, 0.0, -0.0577, -0.0388, -0.0129, 0.0219, 0.0288, -0.001, 0.002, 0.0278, 0.0388, 0.0447, 0.0775, 0.0885, 0.0775, 0.0736, 0.0666, 0.0825, 0.0417, 0.0219, 0.0726, 0.0696, 0.0567, 0.0984, 0.1083, 0.1044, 0.1044, 0.163, 0.1936, 0.1976, 0.1976, 0.1976, 0.2835, 0.3794, 0.3874, 0.4084, 0.3864, 0.3884, 0.3864, 0.3744, 0.4014, 0.4374, 0.4873, 0.4913, 0.4504, 0.4933, 0.5213, 0.5193, 0.5193, 0.5932, 0.6222, 0.6052, 0.6032, 0.5023, 0.6731, 0.6891, 0.5912, 0.5842, 0.6621, 0.6561, 0.6761, 0.784, 0.8099, 0.8099, 0.8099, 0.8099, 0.8978, 0.8988, 0.9278, 0.9977, 1.0327, 1.1775, 1.1625, 1.1625, 0.9488, 0.7201, 0.9138, 0.8799, 0.733, 0.8769, 0.8978, 0.8769, 0.8329, 0.8849, 0.9368, 1.0826, 1.0027, 0.9917, 0.8609, 0.8949, 0.8879, 0.799, 0.799, 0.7646, 0.6736, 0.8983, 0.7857, 0.8637, 0.9328, 0.9668, 1.107, 1.0419, 1.0619, 1.0119, 1.0669, 1.112, 1.177, 1.162, 1.147, 1.192, 1.177, 1.2471, 1.1971, 1.2471, 1.2221, 1.2621, 1.2071, 1.2071, 1.3272, 1.3272, 1.6625, 1.7175, 1.6875, 1.8577, 1.7926]}
{"unit_id": "thread:1990306908536991962", "ticker_idx": 1, "ts": "2025-11-17T06:31:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix will enjoy a supercycle in earnings that lasts at least through 2027", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM shortage intensifying; Samsung and SK Hynix expected to enjoy a supercycle in earnings through at least 2027 as long-term contracts lock in rising prices.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Shortage Worsens… Companies Shift from Monthly Contracts to “Bulk Purchases for Six Months”\n\nAs the global boom in artificial intelligence (AI) investment deepens the shortage of DRAM semiconductors, pricing negotiations that used to be conducted on a monthly or quarterly basis are being reorganized around long-term supply contracts of six months or more. With the severe DRAM supply shortage expected to drive price increases again next year, demand-side companies are actively agreeing to these semiannual contracts. As it has become difficult to secure even next year’s volumes, the market is now seeing discussions extend to supply contracts running through 2027.\n\nAn industry insider said on the 17th, “The DRAM market has shifted into a long-term contract-driven market,” adding, “We’re seeing buying demand that is even stronger than during the 2017 supercycle.” DRAM temporarily stores data so that central processing units (CPUs) or graphics processing units (GPUs) can process information quickly.\n\nWith the rise of large language model (LLM) AIs such as ChatGPT, the role of GPUs has grown, and to support them, memory semiconductors such as high bandwidth memory (HBM), which stacks DRAM in multiple layers, are facing chronic supply shortages. A representative example is the practice of global AI semiconductor leader Nvidia and other U.S. big tech companies pre-purchasing HBM from SK Hynix and Samsung Electronics through annual long-term contracts.\n\nRecently, however, the shortage has spread beyond HBM to general-purpose DRAM as well. OpenAI, Meta, and others have announced AI infrastructure investment plans measured in hundreds of trillions of won, and as major corporations and governments worldwide build data centers for their own AI development, demand for general-purpose DRAM such as double data rate (DDR), graphics (G) DDR, and low-power (LP) DDR has surged. Although general-purpose DRAM delivers lower performance than HBM, it is nonetheless indispensable for AI inference and computation.\n\nAccording to the industry, DRAM demand is rising sharply, particularly among companies in the United States and China. In particular, Samsung Electronics and SK Hynix are understood to be signing supply contracts for next year with major customers on a semiannual basis.\n\nTypically, DRAM is supplied under monthly contracts in which the product is delivered at a fixed monthly price, and then adjusted later to reflect market prices. However, as DRAM demand has surged since the second half of this year, contract periods have been extended beyond quarterly to semiannual or longer-term supply agreements. This is because demand-side companies are not only offering prices above the prevailing market level but also seeking contracts that guarantee supply for more than half a year.\n\nIn reality, DRAM inventories in the market are diminishing rapidly. As of the end of the third quarter, finished goods inventories (“products and merchandise”) at Samsung Electronics’ Device Solutions (DS) division stood at 3.4043 trillion won, down 14.6% (580.4 billion won) from the previous quarter. SK Hynix’s inventories are also falling. In the third quarter, its “products and merchandise” inventories totaled 2.1522 trillion won, down 368.9 billion won compared with the end of last year.\n\nAmid the DRAM shortage, the market is now seeing discussions around supply contracts extending out to 2027. SK Hynix has already completed its DRAM supply contracts for next year and is now negotiating supply for 2027. As SK Hynix’s DRAM has effectively sold out, global big tech companies and other customers are flocking to Samsung Electronics. However, most of the DRAM that Samsung plans to produce next year is also already under contract. With demand soaring, Samsung is even discussing plans to raise DRAM supply prices by more than 40%.\n\nThe industry expects that as the DRAM market shifts toward long-term supply contracts, Samsung Electronics and SK Hynix will enjoy a supercycle in earnings that lasts at least through 2027. When the proportion of long-term supply contracts increases, manufacturers find it easier to plan production, including unit costs and distribution, which in turn improves profitability. An industry official said, “As semiconductor contracts that used to be monthly or quarterly are renewed on a semiannual basis, prices are rising again,” adding, “Because long-term supply contracts through 2027 are being signed at prices higher than today, profitability will improve even further.”\n\n---\n\nhttps://t.co/hcNZI3UCUB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07590760236361427, "ret_signed_m1d": -0.07590760236361427, "alpha_spy_m1d": -0.08531174016949494, "alpha_c_m1d": -0.08966030558661164, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.013752703222997376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05940598564697308, "ret_signed_p1d": -0.05940598564697308, "alpha_spy_p1d": -0.05100852508876652, "alpha_c_p1d": -0.03880661678882169, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.02059936885815139, "ret_p1w": -0.1419141726470945, "ret_signed_p1w": -0.1419141726470945, "alpha_spy_p1w": -0.14651102777888925, "alpha_c_p1w": -0.1384466706950167, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.0034675019520777894, "ret_p1m": -0.12478615967282203, "ret_signed_p1m": -0.12478615967282203, "alpha_spy_p1m": -0.14461586643531044, "alpha_c_p1m": -0.15899133114981623, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.0342051714769942, "ret_p3m": 0.4663960177041486, "ret_signed_p3m": 0.4663960177041486, "alpha_spy_p3m": 0.4399367368167604, "alpha_c_p3m": 0.2692944484854476, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.19710156921870103, "ret_p6m": 2.0358130710294264, "ret_signed_p6m": 2.0358130710294264, "alpha_spy_p6m": 1.920570979766946, "alpha_c_p6m": 1.3814010152845604, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.654412055744866, "price_path": [-0.579, -0.5963, -0.5864, -0.5724, -0.5691, -0.5716, -0.5569, -0.5561, -0.5776, -0.5701, -0.5668, -0.5619, -0.5487, -0.5429, -0.5248, -0.4983, -0.4934, -0.4579, -0.4538, -0.4257, -0.4497, -0.4134, -0.4134, -0.4208, -0.4043, -0.4101, -0.4117, -0.4447, -0.4241, -0.4266, -0.4059, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.2937, -0.3152, -0.321, -0.3028, -0.2533, -0.2318, -0.1988, -0.2096, -0.2054, -0.2104, -0.1584, -0.1172, -0.1403, -0.0792, -0.0627, -0.0776, 0.0231, -0.033, -0.0446, -0.0215, -0.0429, 0.0, 0.0215, 0.0182, 0.0099, -0.0759, 0.0, -0.0594, -0.0726, -0.0578, -0.1403, -0.1419, -0.1436, -0.1353, -0.1017, -0.1248, -0.1116, -0.0785, -0.0885, -0.105, -0.1017, -0.0472, -0.0653, -0.0307, -0.067, -0.0571, -0.0852, -0.1248, -0.0901, -0.0885, -0.0967, -0.0422, -0.0356, -0.029, -0.029, -0.0108, 0.0569, 0.075, 0.075, 0.075, 0.118, 0.1493, 0.1989, 0.2253, 0.2484, 0.2286, 0.2369, 0.2187, 0.2253, 0.2369, 0.2484, 0.2616, 0.227, 0.222, 0.2468, 0.2666, 0.2154, 0.3211, 0.3888, 0.4218, 0.5011, 0.3706, 0.4978, 0.4862, 0.3904, 0.3855, 0.4647, 0.4466, 0.4202, 0.4664, 0.4532, 0.4532, 0.4532, 0.4532, 0.4763, 0.5671, 0.5704, 0.6596, 0.6811, 0.8182, 0.7553, 0.7553, 0.5535, 0.4046, 0.5568, 0.5287, 0.3831, 0.5518, 0.5799, 0.5386, 0.5055, 0.6114, 0.6048, 0.747, 0.6759, 0.666, 0.5435, 0.6312, 0.6461, 0.5435, 0.5435, 0.4443, 0.3351, 0.484, 0.3731, 0.4492, 0.4658, 0.5154, 0.709, 0.6511, 0.6991, 0.7206, 0.8248, 0.8794, 0.9108, 0.8662, 0.929, 1.025, 1.0233, 1.0266, 1.0217, 1.1375, 1.1507, 1.1391, 1.1276, 1.1276, 1.3939, 1.3939, 1.6487, 1.7364, 1.7893, 2.1103, 2.0358]}
{"unit_id": "reply:1990294256863609129", "ticker_idx": 0, "ts": "2025-11-17T05:41:43+00:00", "symbol": "VRT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I believe their market share will decline, or their margins will come under pressure", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bearish Vertiv: market share loss or margin compression ahead despite staying in NVIDIA supply chain.", "resolved_tickers": ["VRT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLI", "bench_c_reason": "industry='Electrical Equipment & Parts'", "bench_c_industry": "Electrical Equipment & Parts", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@kermankohli No, Vertiv will definitely remain in NVIDIA’s supply chain.\nBut I believe their market share will decline, or their margins will come under pressure.", "reply_text": null, "parent_text": "@Jukanlosreve i guess it's good for a while but with this update does it mean that vertiv won't be able to cool nvidia's next gen gpus?", "quoted_text": null, "parent_author": "kermankohli", "ret_m1d": 0.025922724729915192, "ret_signed_m1d": -0.025922724729915192, "alpha_spy_m1d": -0.016518586924034517, "alpha_c_m1d": -0.016023774768157528, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.009898949961757664, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010741003446480768, "ret_signed_p1d": 0.010741003446480768, "alpha_spy_p1d": 0.002343542888274208, "alpha_c_p1d": 0.006289831989469863, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.004451171457010905, "ret_p1w": 0.013561459369951345, "ret_signed_p1w": -0.013561459369951345, "alpha_spy_p1w": -0.008964604238156593, "alpha_c_p1w": -0.01701612578608913, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.003454666416137786, "ret_p1m": -0.03558373680313387, "ret_signed_p1m": 0.03558373680313387, "alpha_spy_p1m": 0.05541344356562228, "alpha_c_p1m": 0.07238944494395472, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.03680570814082085, "ret_p3m": 0.4197315743192198, "ret_signed_p3m": -0.4197315743192198, "alpha_spy_p3m": -0.39327229343183157, "alpha_c_p3m": -0.26811275601159945, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.15161881830762036, "ret_p6m": 1.2043470477334566, "ret_signed_p6m": -1.2043470477334566, "alpha_spy_p6m": -1.089104956470976, "alpha_c_p6m": -1.0387960301231958, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.1655510176102608, "price_path": [-0.2349, -0.2407, -0.2443, -0.25, -0.2326, -0.2243, -0.1948, -0.2348, -0.2348, -0.2561, -0.2466, -0.2459, -0.2561, -0.2692, -0.2467, -0.1797, -0.1859, -0.1911, -0.1704, -0.18, -0.1789, -0.1482, -0.1383, -0.0881, -0.1443, -0.1502, -0.1613, -0.1682, -0.1401, -0.0947, -0.0304, -0.0301, -0.0387, -0.0231, -0.0467, 0.0044, 0.0134, 0.0142, 0.0741, 0.051, 0.0832, 0.067, 0.0441, 0.0545, 0.0489, 0.0296, 0.0993, 0.1165, 0.1575, 0.1435, 0.1957, 0.1627, 0.1573, 0.1485, 0.085, 0.1444, 0.0982, 0.0789, 0.1272, 0.0744, 0.0403, -0.0181, 0.0259, 0.0, -0.0107, 0.024, -0.0422, -0.0409, 0.0136, 0.0179, 0.0326, 0.0326, 0.0789, 0.0758, 0.086, 0.0738, 0.0958, 0.1347, 0.1142, 0.0708, 0.0914, 0.0725, -0.0319, -0.0291, -0.0356, -0.1006, -0.0732, -0.0406, -0.002, -0.002, 0.0017, 0.0017, 0.006, -0.0058, -0.0135, -0.0275, -0.0275, 0.0542, 0.0442, 0.0502, 0.0297, -0.0349, -0.0181, 0.0207, 0.0368, 0.0256, 0.0357, 0.0621, 0.0621, 0.0516, 0.0893, 0.0872, 0.0955, 0.0879, 0.1358, 0.1631, 0.1712, 0.1176, 0.1406, 0.1414, 0.0959, 0.067, 0.174, 0.2126, 0.1983, 0.4918, 0.4197, 0.4078, 0.4078, 0.4619, 0.46, 0.4591, 0.4632, 0.4732, 0.5196, 0.5739, 0.5561, 0.5301, 0.5472, 0.4673, 0.5084, 0.4992, 0.4514, 0.5869, 0.6211, 0.6103, 0.593, 0.554, 0.5892, 0.6116, 0.5894, 0.6162, 0.5364, 0.5371, 0.6265, 0.6581, 0.5155, 0.5075, 0.4063, 0.5045, 0.5573, 0.5689, 0.5689, 0.5535, 0.5749, 0.6874, 0.7271, 0.7719, 0.801, 0.8644, 0.8082, 0.766, 0.8454, 0.8878, 0.876, 0.8321, 0.9319, 0.9421, 0.936, 0.8315, 0.8384, 0.9723, 0.9713, 0.9872, 1.0476, 1.1551, 1.0415, 1.0413, 1.2091, 1.2043]}
{"unit_id": "orig:1977935993799131312", "ticker_idx": 0, "ts": "2025-10-14T03:14:23+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks: NVDA > AVGO > CRDO > AMD. Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares BofA bullish thesis: NVDA>AVGO>CRDO>AMD top picks; overweight AI compute, networking, optical, semicap; AI cycle structural not bubble.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Resilient Semiconductor Growth Cycle Amid ‘AI Infrastructure Bubble’ Debate (BofA)\n\n⸻\n\n1. Industry and Market Diagnosis\n    • While the AI infrastructure buildout carries overinvestment risks, the current momentum remains fundamentally positive.\n    • BofA identifies four indicators for a potential “AI bubble”:\n① AI server utilization,\n② Cloud CapEx plans,\n③ Valuations of AI leaders,\n④ Interest rate environment.\n    • Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO, as well as semiconductor equipment, memory, optical communication, and foundry-related names.\n\n⸻\n\n2. Four Key Differences from the Dot-Com Bubble\n    \n1. No “Dark Fiber” effect – Unlike the IT bubble era’s overinvestment leading to low utilization, AI GPUs are running at full capacity, and OpenAI is projected to reach 1 billion users within three years.\n    \n2. Healthy CapEx funding – Based on operating cash flow, not debt-financed (CapEx/OCF ≈ 25% vs. >30% debt ratio in the past).\n    \n3. Interest rate trajectory – Unlike in 2000 or 2008, the U.S. Fed is more likely to enter a rate-cutting cycle.\n    \n4. Valuation levels – NVDA’s 2026E PER of 29× is reasonable compared with Cisco/Nortel’s 100×+ during the bubble era.\n\n    • AI infrastructure expansion is naturally constrained by physical limits such as power, space, and cooling water, reducing overinvestment risks.\n    • Of NVDA’s projected $3–4 trillion data center TAM by 2030, only 10–20% is expected to translate into actual revenue.\n    • OpenAI’s 250 GW target by 2033 is viewed as unrealistically large, with power, space, and labor competition acting as growth limiters.\n\n⸻\n\n3. Geopolitical Risks\n    • Despite renewed U.S.–China trade tensions, perceived risks are easing.\n    • Rare earth supply: Taiwan’s non-China sourcing structure limits potential impact.\n    • The threat of 100% tariffs is considered unlikely to materialize.\n\n⸻\n\n4. Conclusion and Implications\n    • AI infrastructure investment remains structural expansion over speculative overheating.\n    • Supply constraints (power, space) serve as natural stabilizers.\n    • Valuations are reasonable, and interest rate / cash flow conditions are favorable.\n    • Top Picks: NVDA > AVGO > CRDO > AMD.\n    • Strategic Focus: Maintain overweight in AI Compute + Networking + Optical + Semicap-linked sectors.\n\n$NVDA $AVGO $CRDO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.046047986166150245, "ret_signed_m1d": 0.046047986166150245, "alpha_spy_m1d": 0.044824837096757664, "alpha_c_m1d": 0.027307423501326866, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0011108615216387507, "ret_signed_p1d": -0.0011108615216387507, "alpha_spy_p1d": -0.0055504773883180425, "alpha_c_p1d": -0.025938407846619937, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.0062767702368793366, "ret_signed_p1w": 0.0062767702368793366, "alpha_spy_p1w": -0.0074042206955409196, "alpha_c_p1w": -0.029105445853656597, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.07648737195000477, "ret_signed_p1m": 0.07648737195000477, "alpha_spy_p1m": 0.044549796673410524, "alpha_c_p1m": 0.010430477612360045, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.02688606994062126, "ret_signed_p3m": 0.02688606994062126, "alpha_spy_p3m": -0.024290891255076552, "alpha_c_p3m": -0.14382181854713783, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 0.01149811607040685, "ret_signed_p6m": 0.01149811607040685, "alpha_spy_p6m": -0.015122894768310413, "alpha_c_p6m": -0.2605737054351056, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.0391, -0.0424, -0.0481, -0.0723, -0.0514, -0.035, -0.0363, -0.0183, -0.0252, -0.0043, -0.0121, -0.0351, -0.0002, -0.0099, -0.0034, 0.0041, 0.0148, 0.0112, 0.0173, 0.0086, 0.011, 0.0023, 0.0109, -0.0244, -0.0258, -0.0281, -0.0114, -0.0013, 0.0096, 0.0087, 0.0007, -0.0325, -0.0325, -0.0514, -0.0523, -0.0465, -0.0723, -0.0652, -0.0515, -0.0151, -0.0159, -0.0123, -0.0127, -0.0286, -0.0541, -0.0211, -0.0187, 0.0199, -0.0089, -0.017, -0.013, -0.0102, 0.0101, 0.0364, 0.04, 0.0492, 0.0422, 0.0306, 0.0278, 0.0504, 0.0697, 0.0174, 0.046, 0.0, -0.0011, 0.0099, 0.0177, 0.0145, 0.0063, 0.0014, 0.0118, 0.0346, 0.0637, 0.1166, 0.15, 0.127, 0.1248, 0.1491, 0.1036, 0.0843, 0.0447, 0.0451, 0.1056, 0.0729, 0.0765, 0.0379, 0.0563, 0.0365, 0.0074, 0.036, 0.0034, -0.0064, 0.014, -0.0123, 0.0013, 0.0013, -0.0168, -0.0006, 0.0079, -0.0024, 0.0187, 0.0133, 0.0307, 0.0275, 0.0209, 0.0051, -0.0278, -0.0207, -0.0128, -0.0504, -0.0327, 0.0054, 0.0204, 0.0511, 0.0477, 0.0477, 0.0584, 0.0456, 0.0418, 0.036, 0.036, 0.0491, 0.045, 0.0401, 0.0505, 0.0279, 0.0269, 0.0273, 0.0322, 0.0173, 0.0391, 0.0345, 0.0345, -0.0108, 0.0183, 0.0268, 0.0425, 0.0358, 0.0472, 0.0639, 0.0694, 0.0617, 0.0311, 0.0018, -0.0324, -0.0452, 0.0299, 0.0557, 0.0473, 0.0557, 0.0384, 0.0155, 0.0155, 0.0275, 0.0442, 0.0438, 0.0544, 0.064, 0.0713, 0.0863, 0.0271, -0.0157, 0.0137, 0.0002, 0.0168, 0.0184, -0.0122, 0.0146, 0.0264, 0.0334, 0.0174, 0.0013, 0.0178, 0.0107, 0.0022, -0.0081, -0.0406, -0.0243, -0.0267, -0.0074, -0.0487, -0.0694, -0.0824, -0.0312, -0.0237, -0.0146, -0.0146, -0.0132, -0.0106, 0.0115]}
{"unit_id": "orig:1977935993799131312", "ticker_idx": 1, "ts": "2025-10-14T03:14:23+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks: NVDA > AVGO > CRDO > AMD. Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares BofA bullish thesis: NVDA>AVGO>CRDO>AMD top picks; overweight AI compute, networking, optical, semicap; AI cycle structural not bubble.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Resilient Semiconductor Growth Cycle Amid ‘AI Infrastructure Bubble’ Debate (BofA)\n\n⸻\n\n1. Industry and Market Diagnosis\n    • While the AI infrastructure buildout carries overinvestment risks, the current momentum remains fundamentally positive.\n    • BofA identifies four indicators for a potential “AI bubble”:\n① AI server utilization,\n② Cloud CapEx plans,\n③ Valuations of AI leaders,\n④ Interest rate environment.\n    • Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO, as well as semiconductor equipment, memory, optical communication, and foundry-related names.\n\n⸻\n\n2. Four Key Differences from the Dot-Com Bubble\n    \n1. No “Dark Fiber” effect – Unlike the IT bubble era’s overinvestment leading to low utilization, AI GPUs are running at full capacity, and OpenAI is projected to reach 1 billion users within three years.\n    \n2. Healthy CapEx funding – Based on operating cash flow, not debt-financed (CapEx/OCF ≈ 25% vs. >30% debt ratio in the past).\n    \n3. Interest rate trajectory – Unlike in 2000 or 2008, the U.S. Fed is more likely to enter a rate-cutting cycle.\n    \n4. Valuation levels – NVDA’s 2026E PER of 29× is reasonable compared with Cisco/Nortel’s 100×+ during the bubble era.\n\n    • AI infrastructure expansion is naturally constrained by physical limits such as power, space, and cooling water, reducing overinvestment risks.\n    • Of NVDA’s projected $3–4 trillion data center TAM by 2030, only 10–20% is expected to translate into actual revenue.\n    • OpenAI’s 250 GW target by 2033 is viewed as unrealistically large, with power, space, and labor competition acting as growth limiters.\n\n⸻\n\n3. Geopolitical Risks\n    • Despite renewed U.S.–China trade tensions, perceived risks are easing.\n    • Rare earth supply: Taiwan’s non-China sourcing structure limits potential impact.\n    • The threat of 100% tariffs is considered unlikely to materialize.\n\n⸻\n\n4. Conclusion and Implications\n    • AI infrastructure investment remains structural expansion over speculative overheating.\n    • Supply constraints (power, space) serve as natural stabilizers.\n    • Valuations are reasonable, and interest rate / cash flow conditions are favorable.\n    • Top Picks: NVDA > AVGO > CRDO > AMD.\n    • Strategic Focus: Maintain overweight in AI Compute + Networking + Optical + Semicap-linked sectors.\n\n$NVDA $AVGO $CRDO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03652684176987697, "ret_signed_m1d": 0.03652684176987697, "alpha_spy_m1d": 0.03530369270048439, "alpha_c_m1d": 0.017786279105053593, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0209222269401641, "ret_signed_p1d": 0.0209222269401641, "alpha_spy_p1d": 0.01648261107348481, "alpha_c_p1d": -0.0039053193848170853, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": -0.004271612058988072, "ret_signed_p1w": -0.004271612058988072, "alpha_spy_p1w": -0.01795260299140833, "alpha_c_p1w": -0.039653828149524006, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.032226203832715905, "ret_signed_p1m": 0.032226203832715905, "alpha_spy_p1m": 0.0002886285561216617, "alpha_c_p1m": -0.03383069050492882, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.004358956802815506, "ret_signed_p3m": 0.004358956802815506, "alpha_spy_p3m": -0.046818004392882306, "alpha_c_p3m": -0.16634893168494358, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 0.02297914668726131, "ret_signed_p6m": 0.02297914668726131, "alpha_spy_p6m": -0.0036418641514559535, "alpha_c_p6m": -0.24909267481825115, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.169, -0.1781, -0.1639, -0.1918, -0.177, -0.1625, -0.1582, -0.1463, -0.1372, -0.1221, -0.148, -0.1627, -0.1363, -0.1502, -0.1249, -0.1188, -0.1153, -0.1184, -0.0925, -0.1034, -0.0972, -0.1113, -0.113, -0.1445, -0.1553, -0.1599, -0.1471, -0.1465, -0.1355, -0.129, -0.1046, -0.1373, -0.1373, -0.1348, -0.1228, -0.112, -0.0285, 0.0027, -0.0234, 0.0721, 0.0433, 0.0439, 0.0562, 0.0443, 0.0042, 0.0018, 0.0006, -0.0155, -0.0151, -0.014, -0.0233, -0.0279, -0.0472, -0.0413, -0.0312, -0.0173, -0.0167, -0.0251, -0.0224, 0.004, 0.0026, -0.0567, 0.0365, 0.0, 0.0209, 0.0291, 0.0151, 0.0148, -0.0043, -0.0111, 0.0005, 0.0291, 0.0521, 0.0838, 0.1216, 0.094, 0.0741, 0.0535, 0.0227, 0.0432, 0.0333, 0.0154, 0.0414, 0.0228, 0.0322, -0.0121, -0.0049, -0.0043, -0.0105, 0.0299, 0.0078, -0.0114, 0.0983, 0.1189, 0.1553, 0.1553, 0.171, 0.1219, 0.1088, 0.106, 0.1072, 0.134, 0.1655, 0.1806, 0.2, 0.1809, 0.0459, -0.0126, -0.0082, -0.0526, -0.0414, -0.011, -0.0059, 0.017, 0.0196, 0.0196, 0.0252, 0.0172, 0.0186, 0.0076, 0.0076, 0.0121, -0.0002, 0.0009, 0.0001, -0.032, 0.0044, 0.0254, 0.0324, -0.0104, -0.0013, 0.024, 0.024, -0.0317, -0.0427, -0.0524, -0.0682, -0.0542, -0.0311, -0.0298, -0.0371, -0.0354, -0.036, -0.0674, -0.1031, -0.096, -0.0307, 0.0014, -0.0088, -0.0021, -0.0358, -0.0533, -0.0533, -0.0318, -0.029, -0.0276, -0.0315, -0.0382, -0.0524, -0.0325, -0.0634, -0.0696, -0.0718, -0.0863, -0.0755, -0.0312, -0.0378, 0.0066, -0.0026, -0.0055, -0.0218, -0.0621, -0.054, -0.0645, -0.0802, -0.0688, -0.096, -0.0591, -0.0714, -0.0699, -0.0973, -0.1228, -0.144, -0.097, -0.0854, -0.0823, -0.0823, -0.0826, -0.0256, 0.023]}
{"unit_id": "orig:1977935993799131312", "ticker_idx": 2, "ts": "2025-10-14T03:14:23+00:00", "symbol": "CRDO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks: NVDA > AVGO > CRDO > AMD. Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares BofA bullish thesis: NVDA>AVGO>CRDO>AMD top picks; overweight AI compute, networking, optical, semicap; AI cycle structural not bubble.", "resolved_tickers": ["CRDO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Resilient Semiconductor Growth Cycle Amid ‘AI Infrastructure Bubble’ Debate (BofA)\n\n⸻\n\n1. Industry and Market Diagnosis\n    • While the AI infrastructure buildout carries overinvestment risks, the current momentum remains fundamentally positive.\n    • BofA identifies four indicators for a potential “AI bubble”:\n① AI server utilization,\n② Cloud CapEx plans,\n③ Valuations of AI leaders,\n④ Interest rate environment.\n    • Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO, as well as semiconductor equipment, memory, optical communication, and foundry-related names.\n\n⸻\n\n2. Four Key Differences from the Dot-Com Bubble\n    \n1. No “Dark Fiber” effect – Unlike the IT bubble era’s overinvestment leading to low utilization, AI GPUs are running at full capacity, and OpenAI is projected to reach 1 billion users within three years.\n    \n2. Healthy CapEx funding – Based on operating cash flow, not debt-financed (CapEx/OCF ≈ 25% vs. >30% debt ratio in the past).\n    \n3. Interest rate trajectory – Unlike in 2000 or 2008, the U.S. Fed is more likely to enter a rate-cutting cycle.\n    \n4. Valuation levels – NVDA’s 2026E PER of 29× is reasonable compared with Cisco/Nortel’s 100×+ during the bubble era.\n\n    • AI infrastructure expansion is naturally constrained by physical limits such as power, space, and cooling water, reducing overinvestment risks.\n    • Of NVDA’s projected $3–4 trillion data center TAM by 2030, only 10–20% is expected to translate into actual revenue.\n    • OpenAI’s 250 GW target by 2033 is viewed as unrealistically large, with power, space, and labor competition acting as growth limiters.\n\n⸻\n\n3. Geopolitical Risks\n    • Despite renewed U.S.–China trade tensions, perceived risks are easing.\n    • Rare earth supply: Taiwan’s non-China sourcing structure limits potential impact.\n    • The threat of 100% tariffs is considered unlikely to materialize.\n\n⸻\n\n4. Conclusion and Implications\n    • AI infrastructure investment remains structural expansion over speculative overheating.\n    • Supply constraints (power, space) serve as natural stabilizers.\n    • Valuations are reasonable, and interest rate / cash flow conditions are favorable.\n    • Top Picks: NVDA > AVGO > CRDO > AMD.\n    • Strategic Focus: Maintain overweight in AI Compute + Networking + Optical + Semicap-linked sectors.\n\n$NVDA $AVGO $CRDO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.15529860421182562, "ret_signed_m1d": 0.15529860421182562, "alpha_spy_m1d": 0.15407545514243304, "alpha_c_m1d": 0.13655804154700224, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012793862521074173, "ret_signed_p1d": 0.012793862521074173, "alpha_spy_p1d": 0.008354246654394881, "alpha_c_p1d": -0.012033683803907014, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.11113678742925104, "ret_signed_p1w": 0.11113678742925104, "alpha_spy_p1w": 0.09745579649683078, "alpha_c_p1w": 0.0757545713387151, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.23576105077372356, "ret_signed_p1m": 0.23576105077372356, "alpha_spy_p1m": 0.20382347549712931, "alpha_c_p1m": 0.16970415643607883, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.1593063442693281, "ret_signed_p3m": 0.1593063442693281, "alpha_spy_p3m": 0.10812938307363029, "alpha_c_p3m": -0.011401544218430981, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": -0.15059730956090434, "ret_signed_p6m": -0.15059730956090434, "alpha_spy_p6m": -0.1772183203996216, "alpha_c_p6m": -0.4226691310664168, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.2412, -0.2796, -0.2621, -0.2838, -0.2415, -0.2203, -0.2199, -0.168, -0.157, -0.1059, -0.1403, -0.171, -0.116, -0.15, -0.0956, -0.0768, -0.072, -0.0862, -0.0337, -0.0664, -0.0957, -0.1003, -0.0849, -0.1807, -0.1831, -0.1456, -0.1211, -0.1105, -0.0744, -0.0541, 0.016, -0.0516, -0.0516, -0.0422, -0.0384, 0.0328, 0.0853, 0.137, 0.1486, 0.2637, 0.2279, 0.2485, 0.2638, 0.2672, 0.2674, 0.328, 0.3068, 0.2647, 0.2506, 0.1513, 0.1171, 0.1016, 0.1253, 0.1222, 0.1168, 0.1532, 0.1088, 0.1362, 0.0574, 0.1474, 0.1649, 0.07, 0.1553, 0.0, 0.0128, 0.0523, 0.1068, 0.1689, 0.1111, 0.0574, 0.1635, 0.1988, 0.1943, 0.2499, 0.3219, 0.2842, 0.446, 0.3922, 0.2657, 0.3346, 0.2543, 0.261, 0.3114, 0.2216, 0.2358, 0.1017, 0.1215, 0.122, 0.0756, 0.0935, 0.0384, 0.0288, 0.1626, 0.1883, 0.264, 0.264, 0.3688, 0.3189, 0.4523, 0.4581, 0.3944, 0.3568, 0.3791, 0.3124, 0.2176, 0.1905, 0.1091, 0.0946, 0.0816, 0.0338, 0.068, 0.1571, 0.1556, 0.1392, 0.1575, 0.1575, 0.1162, 0.1152, 0.1169, 0.109, 0.109, 0.1038, 0.0808, 0.0247, 0.0867, 0.0913, 0.1593, 0.2029, 0.2438, 0.2088, 0.1493, 0.1635, 0.1635, 0.1809, 0.0725, 0.0412, 0.0263, -0.0133, -0.0014, -0.0007, -0.0022, -0.0345, -0.0755, -0.1421, -0.2528, -0.2442, -0.1414, -0.0489, 0.0383, -0.0104, -0.0614, -0.064, -0.064, -0.0382, -0.0142, 0.007, -0.0439, -0.0439, -0.0687, -0.0485, -0.1177, -0.1347, -0.1197, -0.2501, -0.2097, -0.1157, -0.1535, -0.1061, -0.1343, -0.1067, -0.1401, -0.0929, -0.0992, -0.198, -0.216, -0.1746, -0.2031, -0.19, -0.227, -0.1992, -0.2567, -0.266, -0.3232, -0.2765, -0.2607, -0.2181, -0.2181, -0.2103, -0.177, -0.1506]}
{"unit_id": "orig:1977935993799131312", "ticker_idx": 3, "ts": "2025-10-14T03:14:23+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Top Picks: NVDA > AVGO > CRDO > AMD. Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares BofA bullish thesis: NVDA>AVGO>CRDO>AMD top picks; overweight AI compute, networking, optical, semicap; AI cycle structural not bubble.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Resilient Semiconductor Growth Cycle Amid ‘AI Infrastructure Bubble’ Debate (BofA)\n\n⸻\n\n1. Industry and Market Diagnosis\n    • While the AI infrastructure buildout carries overinvestment risks, the current momentum remains fundamentally positive.\n    • BofA identifies four indicators for a potential “AI bubble”:\n① AI server utilization,\n② Cloud CapEx plans,\n③ Valuations of AI leaders,\n④ Interest rate environment.\n    • Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO, as well as semiconductor equipment, memory, optical communication, and foundry-related names.\n\n⸻\n\n2. Four Key Differences from the Dot-Com Bubble\n    \n1. No “Dark Fiber” effect – Unlike the IT bubble era’s overinvestment leading to low utilization, AI GPUs are running at full capacity, and OpenAI is projected to reach 1 billion users within three years.\n    \n2. Healthy CapEx funding – Based on operating cash flow, not debt-financed (CapEx/OCF ≈ 25% vs. >30% debt ratio in the past).\n    \n3. Interest rate trajectory – Unlike in 2000 or 2008, the U.S. Fed is more likely to enter a rate-cutting cycle.\n    \n4. Valuation levels – NVDA’s 2026E PER of 29× is reasonable compared with Cisco/Nortel’s 100×+ during the bubble era.\n\n    • AI infrastructure expansion is naturally constrained by physical limits such as power, space, and cooling water, reducing overinvestment risks.\n    • Of NVDA’s projected $3–4 trillion data center TAM by 2030, only 10–20% is expected to translate into actual revenue.\n    • OpenAI’s 250 GW target by 2033 is viewed as unrealistically large, with power, space, and labor competition acting as growth limiters.\n\n⸻\n\n3. Geopolitical Risks\n    • Despite renewed U.S.–China trade tensions, perceived risks are easing.\n    • Rare earth supply: Taiwan’s non-China sourcing structure limits potential impact.\n    • The threat of 100% tariffs is considered unlikely to materialize.\n\n⸻\n\n4. Conclusion and Implications\n    • AI infrastructure investment remains structural expansion over speculative overheating.\n    • Supply constraints (power, space) serve as natural stabilizers.\n    • Valuations are reasonable, and interest rate / cash flow conditions are favorable.\n    • Top Picks: NVDA > AVGO > CRDO > AMD.\n    • Strategic Focus: Maintain overweight in AI Compute + Networking + Optical + Semicap-linked sectors.\n\n$NVDA $AVGO $CRDO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0076573808839812685, "ret_signed_m1d": -0.0076573808839812685, "alpha_spy_m1d": -0.00888052995337385, "alpha_c_m1d": -0.026397943548804648, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09404378976580152, "ret_signed_p1d": 0.09404378976580152, "alpha_spy_p1d": 0.08960417389912223, "alpha_c_p1d": 0.06921624344082034, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.09143015624849138, "ret_signed_p1w": 0.09143015624849138, "alpha_spy_p1w": 0.07774916531607112, "alpha_c_p1w": 0.056047940157955445, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.18707881606515642, "ret_signed_p1m": 0.18707881606515642, "alpha_spy_p1m": 0.15514124078856217, "alpha_c_p1m": 0.1210219217275117, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": -0.06841211618816989, "ret_signed_p3m": -0.06841211618816989, "alpha_spy_p3m": -0.1195890773838677, "alpha_c_p3m": -0.23912000467592898, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 0.06295571193946925, "ret_signed_p6m": 0.06295571193946925, "alpha_spy_p6m": 0.03633470110075199, "alpha_c_p6m": -0.2091161095660432, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.2645, -0.2802, -0.2801, -0.2906, -0.2725, -0.2566, -0.2367, -0.2037, -0.1864, -0.1769, -0.1916, -0.2127, -0.1894, -0.2007, -0.2521, -0.2095, -0.2078, -0.2101, -0.1978, -0.1544, -0.1703, -0.1861, -0.1924, -0.2363, -0.2425, -0.2493, -0.2308, -0.251, -0.236, -0.2337, -0.227, -0.2543, -0.2543, -0.2557, -0.2566, -0.2582, -0.307, -0.3057, -0.2855, -0.2685, -0.2862, -0.2729, -0.261, -0.2642, -0.2702, -0.2759, -0.2783, -0.2673, -0.2622, -0.2623, -0.2605, -0.2688, -0.2601, -0.2582, -0.248, -0.2217, -0.2449, -0.0659, -0.0302, 0.0801, 0.0679, -0.0146, -0.0077, 0.0, 0.094, 0.0755, 0.0687, 0.103, 0.0914, 0.0557, 0.0775, 0.1597, 0.1907, 0.183, 0.212, 0.1685, 0.1744, 0.1906, 0.1465, 0.1753, 0.0899, 0.0708, 0.1187, 0.0891, 0.1871, 0.137, 0.1317, 0.1028, 0.0559, 0.025, -0.0553, -0.0656, -0.0139, -0.0548, -0.0177, -0.0177, -0.0026, 0.0077, -0.0131, -0.0022, -0.0097, -0.0006, 0.0138, 0.0162, 0.0153, 0.0153, -0.0335, -0.0482, -0.0409, -0.0916, -0.0781, -0.0214, -0.0144, -0.0146, -0.014, -0.014, -0.0142, -0.0114, -0.0126, -0.018, -0.018, 0.0247, 0.0137, -0.0171, -0.037, -0.0615, -0.0684, -0.0477, 0.0132, 0.0253, 0.0451, 0.063, 0.063, 0.0634, 0.1454, 0.1634, 0.1907, 0.1523, 0.1556, 0.1589, 0.1563, 0.0855, 0.1292, 0.1101, -0.0821, -0.1173, -0.0442, -0.0096, -0.0207, -0.0207, -0.0557, -0.0494, -0.0494, -0.0688, -0.0824, -0.0675, -0.0823, -0.0985, -0.0195, -0.0332, -0.0661, -0.082, -0.0893, -0.1244, -0.0735, -0.0855, -0.1177, -0.0707, -0.0681, -0.0608, -0.0933, -0.1133, -0.0986, -0.0999, -0.0854, -0.0588, -0.0768, -0.0707, -0.0583, 0.01, -0.0657, -0.0738, -0.1011, -0.0672, -0.0361, -0.0027, -0.0027, 0.0096, 0.0158, 0.063]}
{"unit_id": "orig:1977935993799131312", "ticker_idx": 4, "ts": "2025-10-14T03:14:23+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": "equipment", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Strategic Focus: Maintain overweight in AI Compute + Networking + Optical + Semicap-linked sectors", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares BofA bullish thesis: NVDA>AVGO>CRDO>AMD top picks; overweight AI compute, networking, optical, semicap; AI cycle structural not bubble.", "resolved_tickers": ["AMAT", "LRCX", "KLAC"], "resolution_method": "basket", "resolution_reason": "US semiconductor equipment sub-sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Resilient Semiconductor Growth Cycle Amid ‘AI Infrastructure Bubble’ Debate (BofA)\n\n⸻\n\n1. Industry and Market Diagnosis\n    • While the AI infrastructure buildout carries overinvestment risks, the current momentum remains fundamentally positive.\n    • BofA identifies four indicators for a potential “AI bubble”:\n① AI server utilization,\n② Cloud CapEx plans,\n③ Valuations of AI leaders,\n④ Interest rate environment.\n    • Maintains an optimistic outlook on NVDA, AVGO, AMD, CRDO, as well as semiconductor equipment, memory, optical communication, and foundry-related names.\n\n⸻\n\n2. Four Key Differences from the Dot-Com Bubble\n    \n1. No “Dark Fiber” effect – Unlike the IT bubble era’s overinvestment leading to low utilization, AI GPUs are running at full capacity, and OpenAI is projected to reach 1 billion users within three years.\n    \n2. Healthy CapEx funding – Based on operating cash flow, not debt-financed (CapEx/OCF ≈ 25% vs. >30% debt ratio in the past).\n    \n3. Interest rate trajectory – Unlike in 2000 or 2008, the U.S. Fed is more likely to enter a rate-cutting cycle.\n    \n4. Valuation levels – NVDA’s 2026E PER of 29× is reasonable compared with Cisco/Nortel’s 100×+ during the bubble era.\n\n    • AI infrastructure expansion is naturally constrained by physical limits such as power, space, and cooling water, reducing overinvestment risks.\n    • Of NVDA’s projected $3–4 trillion data center TAM by 2030, only 10–20% is expected to translate into actual revenue.\n    • OpenAI’s 250 GW target by 2033 is viewed as unrealistically large, with power, space, and labor competition acting as growth limiters.\n\n⸻\n\n3. Geopolitical Risks\n    • Despite renewed U.S.–China trade tensions, perceived risks are easing.\n    • Rare earth supply: Taiwan’s non-China sourcing structure limits potential impact.\n    • The threat of 100% tariffs is considered unlikely to materialize.\n\n⸻\n\n4. Conclusion and Implications\n    • AI infrastructure investment remains structural expansion over speculative overheating.\n    • Supply constraints (power, space) serve as natural stabilizers.\n    • Valuations are reasonable, and interest rate / cash flow conditions are favorable.\n    • Top Picks: NVDA > AVGO > CRDO > AMD.\n    • Strategic Focus: Maintain overweight in AI Compute + Networking + Optical + Semicap-linked sectors.\n\n$NVDA $AVGO $CRDO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.000535040689595867, "ret_signed_m1d": 0.000535040689595867, "alpha_spy_m1d": -0.0006881083797967147, "alpha_c_m1d": -0.018205521975227512, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04985949935892342, "ret_signed_p1d": 0.04985949935892342, "alpha_spy_p1d": 0.04541988349224413, "alpha_c_p1d": 0.025031953033942234, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.06770746973301711, "ret_signed_p1w": 0.06770746973301711, "alpha_spy_p1w": 0.054026478800596855, "alpha_c_p1w": 0.03232525364248118, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.13115947767046898, "ret_signed_p1m": 0.13115947767046898, "alpha_spy_p1m": 0.09922190239387474, "alpha_c_p1m": 0.06510258333282426, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.4438774430860615, "ret_signed_p3m": 0.4438774430860615, "alpha_spy_p3m": 0.39270048189036366, "alpha_c_p3m": 0.2731695545983024, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 0.7319999684126121, "ret_signed_p6m": 0.7319999684126121, "alpha_spy_p6m": 0.7053789575738948, "alpha_c_p6m": 0.4599281469070996, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.1604, -0.1658, -0.1578, -0.1906, -0.1907, -0.185, -0.1914, -0.1736, -0.1779, -0.1732, -0.2125, -0.2065, -0.1879, -0.2081, -0.2098, -0.1868, -0.177, -0.1786, -0.1563, -0.1459, -0.1451, -0.2305, -0.226, -0.2269, -0.2308, -0.2356, -0.2282, -0.2231, -0.2108, -0.2106, -0.2066, -0.2301, -0.2301, -0.2508, -0.252, -0.2329, -0.2096, -0.2043, -0.198, -0.189, -0.1503, -0.149, -0.1308, -0.1232, -0.113, -0.0661, -0.0648, -0.0275, -0.0277, -0.0357, -0.0421, -0.0333, -0.0251, -0.014, 0.0436, 0.0661, 0.0417, 0.0719, 0.014, 0.0211, 0.0187, -0.0433, 0.0005, 0.0, 0.0499, 0.0481, 0.0444, 0.0703, 0.0677, 0.0395, 0.0813, 0.0994, 0.1264, 0.1148, 0.1488, 0.138, 0.1284, 0.1479, 0.115, 0.1646, 0.1397, 0.1233, 0.1559, 0.12, 0.1312, 0.0881, 0.0712, 0.0738, 0.0547, 0.0979, 0.0324, 0.0438, 0.0859, 0.1103, 0.1338, 0.1338, 0.1447, 0.1397, 0.1747, 0.1913, 0.1849, 0.1887, 0.202, 0.2081, 0.2307, 0.2264, 0.1729, 0.1954, 0.1886, 0.1357, 0.1835, 0.2138, 0.2316, 0.2342, 0.2429, 0.2429, 0.2473, 0.2375, 0.2223, 0.2021, 0.2021, 0.2732, 0.3456, 0.4068, 0.3801, 0.3474, 0.4439, 0.4672, 0.4535, 0.4331, 0.5163, 0.5491, 0.5491, 0.5078, 0.544, 0.509, 0.512, 0.5293, 0.6111, 0.6235, 0.67, 0.5218, 0.5351, 0.4845, 0.3875, 0.4135, 0.521, 0.5283, 0.5165, 0.5695, 0.5333, 0.5884, 0.5884, 0.5976, 0.6274, 0.619, 0.6542, 0.6423, 0.6599, 0.7117, 0.6498, 0.6321, 0.6284, 0.5327, 0.5683, 0.5159, 0.4177, 0.4956, 0.5239, 0.5446, 0.4827, 0.4987, 0.5296, 0.5704, 0.5619, 0.6057, 0.5876, 0.611, 0.6609, 0.6335, 0.5033, 0.4981, 0.4291, 0.5202, 0.5742, 0.5564, 0.5564, 0.5757, 0.5901, 0.732]}
{"unit_id": "orig:1992209835249819925", "ticker_idx": 0, "ts": "2025-11-22T12:33:32+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The 2026 NVIDIA Rubin GPU unit forecast has been revised up to 3 million units (NVIDIA's total 2026 GPU procurement volume has been revised up to 7.5 million units)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises NVDA/AMD GPU forecasts, names SK Hynix biggest HBM beneficiary at 60-70% share with 13% ASP growth; Micron 24% and Samsung 15% NVIDIA HBM share.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s latest semiconductor supply chain update:\n\n• The 2026 NVIDIA Rubin GPU unit forecast has been revised up to 3 million units (NVIDIA’s total 2026 GPU procurement volume has been revised up to 7.5 million units).\n    \n• The AMD MI450 unit forecast has been revised up.\n    \n• The AWS ASIC unit forecast has been slightly revised down.\n    \n• There is a possibility that Rubin’s initial ramp could be pulled forward.\n   \n • The OpenAI ASIC forecast has been revised down (from 700,000 units to 500,000 units).\n    \n• Instead, the assumption for HBM3E has been revised from 8-Hi to 12-Hi.\n    \n• SK Hynix is expected to be the biggest beneficiary.\n    \n• Hynix has effectively passed qualification for NVIDIA Rubin’s HBMA4 and is expected to begin full-scale shipments from 2Q26.\n    \n• SK Hynix’s HBM4 share for NVIDIA in 2026 is projected to be around 70%, with its overall HBM market share expected to be around 60–70%.\n    \n• There is a possibility that HBM4 prices will carry about a 30% premium to HBM3E 12-Hi, and based on this, SK Hynix’s HBM ASP (average selling price) in 2026 is expected to grow about 13% YoY.\n    \n• Micron’s and Samsung’s 2026 HBM share for NVIDIA are projected at 24% and 15%, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020104103865114742, "ret_signed_m1d": -0.020104103865114742, "alpha_spy_m1d": -0.0055990889671471145, "alpha_c_m1d": 0.018200836080618288, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02591077088622873, "ret_signed_p1d": -0.02591077088622873, "alpha_spy_p1d": -0.03531667053833565, "alpha_c_p1d": -0.028535295905334435, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": -0.014407031299373929, "ret_signed_p1w": -0.014407031299373929, "alpha_spy_p1w": -0.03166372528793093, "alpha_c_p1w": -0.055218602968658215, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.036540847884902306, "ret_signed_p1m": 0.036540847884902306, "alpha_spy_p1m": 0.00474490134740746, "alpha_c_p1m": -0.037680744931153365, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.029364291969660616, "ret_signed_p3m": 0.029364291969660616, "alpha_spy_p3m": 0.002787705511193117, "alpha_c_p3m": -0.1840009080598759, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 0.20862345351439582, "ret_signed_p6m": 0.20862345351439582, "alpha_spy_p6m": 0.10517682354106728, "alpha_c_p6m": -0.4004017566940443, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.0053, -0.0131, -0.0459, -0.0459, -0.0645, -0.0654, -0.0597, -0.0851, -0.0781, -0.0646, -0.0286, -0.0295, -0.0259, -0.0263, -0.042, -0.0672, -0.0346, -0.0322, 0.0058, -0.0226, -0.0306, -0.0266, -0.0239, -0.0038, 0.0221, 0.0257, 0.0347, 0.0278, 0.0164, 0.0136, 0.0359, 0.0549, 0.0033, 0.0316, -0.0138, -0.0149, -0.0041, 0.0037, 0.0005, -0.0076, -0.0124, -0.0021, 0.0203, 0.049, 0.1012, 0.1342, 0.1114, 0.1092, 0.1333, 0.0884, 0.0694, 0.0303, 0.0307, 0.0904, 0.0581, 0.0616, 0.0236, 0.0417, 0.0222, -0.0065, 0.0217, -0.0105, -0.0201, 0.0, -0.0259, -0.0125, -0.0125, -0.0304, -0.0144, -0.006, -0.0162, 0.0046, -0.0007, 0.0165, 0.0133, 0.0068, -0.0088, -0.0412, -0.0342, -0.0264, -0.0635, -0.046, -0.0085, 0.0063, 0.0365, 0.0333, 0.0333, 0.0438, 0.0311, 0.0274, 0.0217, 0.0217, 0.0346, 0.0306, 0.0257, 0.036, 0.0137, 0.0127, 0.0131, 0.0179, 0.0033, 0.0247, 0.0202, 0.0202, -0.0245, 0.0043, 0.0126, 0.0281, 0.0215, 0.0328, 0.0492, 0.0546, 0.0471, 0.0168, -0.0121, -0.0457, -0.0584, 0.0157, 0.0411, 0.0329, 0.0411, 0.0241, 0.0015, 0.0015, 0.0133, 0.0298, 0.0294, 0.0399, 0.0494, 0.0565, 0.0713, 0.0129, -0.0293, -0.0003, -0.0136, 0.0027, 0.0044, -0.0259, 0.0006, 0.0122, 0.0192, 0.0033, -0.0125, 0.0038, -0.0033, -0.0117, -0.0217, -0.0539, -0.0377, -0.0402, -0.0211, -0.0619, -0.0822, -0.0951, -0.0445, -0.0371, -0.0282, -0.0282, -0.0268, -0.0243, -0.0025, 0.0076, 0.0334, 0.0371, 0.0766, 0.0895, 0.0867, 0.1049, 0.107, 0.0951, 0.1094, 0.0937, 0.141, 0.1867, 0.1679, 0.1464, 0.0934, 0.0872, 0.0874, 0.0765, 0.1386, 0.1587, 0.179, 0.2022, 0.2096, 0.2372, 0.2915, 0.2344, 0.218, 0.2086]}
{"unit_id": "orig:1992209835249819925", "ticker_idx": 1, "ts": "2025-11-22T12:33:32+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The AMD MI450 unit forecast has been revised up", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises NVDA/AMD GPU forecasts, names SK Hynix biggest HBM beneficiary at 60-70% share with 13% ASP growth; Micron 24% and Samsung 15% NVIDIA HBM share.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s latest semiconductor supply chain update:\n\n• The 2026 NVIDIA Rubin GPU unit forecast has been revised up to 3 million units (NVIDIA’s total 2026 GPU procurement volume has been revised up to 7.5 million units).\n    \n• The AMD MI450 unit forecast has been revised up.\n    \n• The AWS ASIC unit forecast has been slightly revised down.\n    \n• There is a possibility that Rubin’s initial ramp could be pulled forward.\n   \n • The OpenAI ASIC forecast has been revised down (from 700,000 units to 500,000 units).\n    \n• Instead, the assumption for HBM3E has been revised from 8-Hi to 12-Hi.\n    \n• SK Hynix is expected to be the biggest beneficiary.\n    \n• Hynix has effectively passed qualification for NVIDIA Rubin’s HBMA4 and is expected to begin full-scale shipments from 2Q26.\n    \n• SK Hynix’s HBM4 share for NVIDIA in 2026 is projected to be around 70%, with its overall HBM market share expected to be around 60–70%.\n    \n• There is a possibility that HBM4 prices will carry about a 30% premium to HBM3E 12-Hi, and based on this, SK Hynix’s HBM ASP (average selling price) in 2026 is expected to grow about 13% YoY.\n    \n• Micron’s and Samsung’s 2026 HBM share for NVIDIA are projected at 24% and 15%, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.052406436235894804, "ret_signed_m1d": -0.052406436235894804, "alpha_spy_m1d": -0.03790142133792718, "alpha_c_m1d": -0.014101496290161775, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0414787167745283, "ret_signed_p1d": -0.0414787167745283, "alpha_spy_p1d": -0.05088461642663522, "alpha_c_p1d": -0.04410324179363401, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.02190184323756794, "ret_signed_p1w": 0.02190184323756794, "alpha_spy_p1w": 0.004645149249010938, "alpha_c_p1w": -0.018909728431716344, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": -0.000697554769331199, "ret_signed_p1m": -0.000697554769331199, "alpha_spy_p1m": -0.032493501306826045, "alpha_c_p1m": -0.07491914758538687, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": -0.05431298660227957, "ret_signed_p3m": -0.05431298660227957, "alpha_spy_p3m": -0.08088957306074707, "alpha_c_p3m": -0.2676781866318161, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 0.9253661098219841, "ret_signed_p6m": 0.9253661098219841, "alpha_spy_p6m": 0.8219194798486555, "alpha_c_p6m": 0.31634089961354395, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.2228, -0.2161, -0.2438, -0.2438, -0.2452, -0.2461, -0.2477, -0.2972, -0.2959, -0.2754, -0.2581, -0.2761, -0.2626, -0.2506, -0.2538, -0.2599, -0.2657, -0.2681, -0.257, -0.2518, -0.2519, -0.2501, -0.2585, -0.2497, -0.2477, -0.2373, -0.2107, -0.2343, -0.0527, -0.0165, 0.0954, 0.083, -0.0007, 0.0064, 0.0141, 0.1095, 0.0907, 0.0838, 0.1186, 0.1069, 0.0706, 0.0927, 0.1761, 0.2075, 0.1998, 0.2292, 0.185, 0.191, 0.2074, 0.1628, 0.192, 0.1053, 0.086, 0.1345, 0.1045, 0.2039, 0.153, 0.1477, 0.1184, 0.0709, 0.0395, -0.042, -0.0524, 0.0, -0.0415, -0.0038, -0.0038, 0.0115, 0.0219, 0.0009, 0.0119, 0.0043, 0.0136, 0.0282, 0.0306, 0.0296, 0.0297, -0.0199, -0.0347, -0.0273, -0.0788, -0.0651, -0.0075, -0.0005, -0.0007, -0.0, -0.0, -0.0003, 0.0026, 0.0013, -0.0041, -0.0041, 0.0392, 0.028, -0.0033, -0.0234, -0.0482, -0.0552, -0.0342, 0.0275, 0.0398, 0.0598, 0.078, 0.078, 0.0784, 0.1616, 0.1799, 0.2075, 0.1686, 0.172, 0.1753, 0.1727, 0.1008, 0.1452, 0.1258, -0.0691, -0.1049, -0.0307, 0.0044, -0.0069, -0.0068, -0.0424, -0.0359, -0.0359, -0.0557, -0.0694, -0.0543, -0.0693, -0.0858, -0.0056, -0.0195, -0.0529, -0.069, -0.0764, -0.1121, -0.0604, -0.0725, -0.1052, -0.0575, -0.055, -0.0475, -0.0805, -0.1007, -0.0859, -0.0871, -0.0725, -0.0455, -0.0638, -0.0575, -0.045, 0.0243, -0.0525, -0.0607, -0.0884, -0.054, -0.0225, 0.0114, 0.0114, 0.0239, 0.0301, 0.078, 0.1004, 0.1395, 0.1478, 0.1861, 0.2003, 0.2939, 0.2945, 0.2785, 0.3229, 0.4111, 0.4198, 0.6173, 0.5561, 0.503, 0.5676, 0.6484, 0.6765, 0.5882, 0.652, 0.9595, 0.8994, 1.1167, 1.1334, 1.0846, 1.0716, 1.0911, 0.9721, 0.9576, 0.9254]}
{"unit_id": "orig:1992209835249819925", "ticker_idx": 2, "ts": "2025-11-22T12:33:32+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is expected to be the biggest beneficiary... SK Hynix's HBM4 share for NVIDIA in 2026 is projected to be around 70%, with its overall HBM market share expected to be around 60–70%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises NVDA/AMD GPU forecasts, names SK Hynix biggest HBM beneficiary at 60-70% share with 13% ASP growth; Micron 24% and Samsung 15% NVIDIA HBM share.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s latest semiconductor supply chain update:\n\n• The 2026 NVIDIA Rubin GPU unit forecast has been revised up to 3 million units (NVIDIA’s total 2026 GPU procurement volume has been revised up to 7.5 million units).\n    \n• The AMD MI450 unit forecast has been revised up.\n    \n• The AWS ASIC unit forecast has been slightly revised down.\n    \n• There is a possibility that Rubin’s initial ramp could be pulled forward.\n   \n • The OpenAI ASIC forecast has been revised down (from 700,000 units to 500,000 units).\n    \n• Instead, the assumption for HBM3E has been revised from 8-Hi to 12-Hi.\n    \n• SK Hynix is expected to be the biggest beneficiary.\n    \n• Hynix has effectively passed qualification for NVIDIA Rubin’s HBMA4 and is expected to begin full-scale shipments from 2Q26.\n    \n• SK Hynix’s HBM4 share for NVIDIA in 2026 is projected to be around 70%, with its overall HBM market share expected to be around 60–70%.\n    \n• There is a possibility that HBM4 prices will carry about a 30% premium to HBM3E 12-Hi, and based on this, SK Hynix’s HBM ASP (average selling price) in 2026 is expected to grow about 13% YoY.\n    \n• Micron’s and Samsung’s 2026 HBM share for NVIDIA are projected at 24% and 15%, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001923085035032246, "ret_signed_m1d": 0.001923085035032246, "alpha_spy_m1d": 0.016428099932999873, "alpha_c_m1d": 0.040228024980765276, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001923085035032246, "ret_signed_p1d": -0.001923085035032246, "alpha_spy_p1d": -0.011328984687139165, "alpha_c_p1d": -0.004547610054137952, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.03535624949908667, "ret_signed_p1w": 0.03535624949908667, "alpha_spy_p1w": 0.018099555510529663, "alpha_c_p1w": -0.0054553221701976184, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.12388131384195789, "ret_signed_p1m": 0.12388131384195789, "alpha_spy_p1m": 0.09208536730446304, "alpha_c_p1m": 0.049659721025902215, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.720462027311146, "ret_signed_p3m": 0.720462027311146, "alpha_spy_p3m": 0.6938854408526784, "alpha_c_p3m": 0.5070968272816094, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.3643692048480567, "ret_signed_p6m": 2.3643692048480567, "alpha_spy_p6m": 2.260922574874728, "alpha_c_p6m": 1.7553439946396165, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.5007, -0.4837, -0.4827, -0.5077, -0.499, -0.4952, -0.4894, -0.474, -0.4673, -0.4462, -0.4154, -0.4096, -0.3683, -0.3635, -0.3308, -0.3587, -0.3163, -0.3163, -0.325, -0.3058, -0.3125, -0.3144, -0.3529, -0.3288, -0.3317, -0.3077, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.1769, -0.2019, -0.2087, -0.1875, -0.1298, -0.1048, -0.0663, -0.0788, -0.074, -0.0798, -0.0192, 0.0288, 0.0019, 0.0731, 0.0923, 0.075, 0.1923, 0.1269, 0.1135, 0.1404, 0.1154, 0.1654, 0.1904, 0.1865, 0.1769, 0.0769, 0.1654, 0.0962, 0.0808, 0.0981, 0.0019, 0.0, -0.0019, 0.0077, 0.0469, 0.02, 0.0354, 0.0738, 0.0623, 0.0431, 0.0469, 0.1104, 0.0892, 0.1297, 0.0873, 0.0989, 0.0661, 0.02, 0.0604, 0.0623, 0.0527, 0.1162, 0.1239, 0.1316, 0.1316, 0.1527, 0.2317, 0.2528, 0.2528, 0.2528, 0.3029, 0.3394, 0.3972, 0.4279, 0.4549, 0.4318, 0.4414, 0.4202, 0.4279, 0.4414, 0.4549, 0.4703, 0.4299, 0.4241, 0.453, 0.4761, 0.4164, 0.5396, 0.6185, 0.657, 0.7493, 0.5973, 0.7455, 0.732, 0.6204, 0.6146, 0.707, 0.6858, 0.655, 0.7089, 0.6935, 0.6935, 0.6935, 0.6935, 0.7205, 0.8263, 0.8302, 0.9341, 0.9591, 1.1189, 1.0456, 1.0456, 0.8104, 0.6369, 0.8143, 0.7815, 0.6118, 0.8085, 0.8412, 0.793, 0.7545, 0.8779, 0.8702, 1.036, 0.9531, 0.9415, 0.7988, 0.901, 0.9184, 0.7988, 0.7988, 0.6831, 0.5559, 0.7294, 0.6002, 0.6889, 0.7082, 0.7661, 0.9916, 0.9241, 0.9801, 1.0051, 1.1266, 1.1902, 1.2268, 1.1748, 1.2481, 1.3599, 1.358, 1.3618, 1.356, 1.491, 1.5064, 1.4929, 1.4794, 1.4794, 1.7898, 1.7898, 2.0867, 2.1889, 2.2506, 2.6246, 2.5379, 2.8097, 2.7982, 2.507, 2.5475, 2.3644]}
{"unit_id": "orig:1992209835249819925", "ticker_idx": 3, "ts": "2025-11-22T12:33:32+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's and Samsung's 2026 HBM share for NVIDIA are projected at 24% and 15%, respectively", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises NVDA/AMD GPU forecasts, names SK Hynix biggest HBM beneficiary at 60-70% share with 13% ASP growth; Micron 24% and Samsung 15% NVIDIA HBM share.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s latest semiconductor supply chain update:\n\n• The 2026 NVIDIA Rubin GPU unit forecast has been revised up to 3 million units (NVIDIA’s total 2026 GPU procurement volume has been revised up to 7.5 million units).\n    \n• The AMD MI450 unit forecast has been revised up.\n    \n• The AWS ASIC unit forecast has been slightly revised down.\n    \n• There is a possibility that Rubin’s initial ramp could be pulled forward.\n   \n • The OpenAI ASIC forecast has been revised down (from 700,000 units to 500,000 units).\n    \n• Instead, the assumption for HBM3E has been revised from 8-Hi to 12-Hi.\n    \n• SK Hynix is expected to be the biggest beneficiary.\n    \n• Hynix has effectively passed qualification for NVIDIA Rubin’s HBMA4 and is expected to begin full-scale shipments from 2Q26.\n    \n• SK Hynix’s HBM4 share for NVIDIA in 2026 is projected to be around 70%, with its overall HBM market share expected to be around 60–70%.\n    \n• There is a possibility that HBM4 prices will carry about a 30% premium to HBM3E 12-Hi, and based on this, SK Hynix’s HBM ASP (average selling price) in 2026 is expected to grow about 13% YoY.\n    \n• Micron’s and Samsung’s 2026 HBM share for NVIDIA are projected at 24% and 15%, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07395165783628477, "ret_signed_m1d": -0.07395165783628477, "alpha_spy_m1d": -0.05944664293831714, "alpha_c_m1d": -0.03564671789055174, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002679462252589637, "ret_signed_p1d": 0.002679462252589637, "alpha_spy_p1d": -0.006726437399517282, "alpha_c_p1d": 5.4937233483931536e-05, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.07381771882218224, "ret_signed_p1w": 0.07381771882218224, "alpha_spy_p1w": 0.05656102483362524, "alpha_c_p1w": 0.03300614715289796, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.23373367277280366, "ret_signed_p1m": 0.23373367277280366, "alpha_spy_p1m": 0.2019377262353088, "alpha_c_p1m": 0.159512079956748, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.8645050019129275, "ret_signed_p3m": 0.8645050019129275, "alpha_spy_p3m": 0.83792841545446, "alpha_c_p3m": 0.6511398018833909, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.1229220528131627, "ret_signed_p6m": 2.1229220528131627, "alpha_spy_p6m": 2.019475422839834, "alpha_c_p6m": 1.5138968426047226, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.4745, -0.4555, -0.4689, -0.4689, -0.4712, -0.4702, -0.4457, -0.4137, -0.4133, -0.3964, -0.3752, -0.328, -0.2983, -0.2959, -0.2912, -0.286, -0.2463, -0.2738, -0.2653, -0.2573, -0.2783, -0.3001, -0.2981, -0.2685, -0.2533, -0.1871, -0.1799, -0.1612, -0.1472, -0.1708, -0.1223, -0.1411, -0.189, -0.1392, -0.1646, -0.1429, -0.0956, -0.0962, -0.0766, -0.0966, -0.1137, -0.0769, -0.0219, -0.0171, -0.009, 0.0121, 0.0004, -0.0007, 0.0481, -0.0263, 0.0606, 0.0643, 0.0625, 0.1312, 0.0767, 0.0936, 0.0581, 0.1023, 0.0805, 0.0204, 0.0089, -0.1007, -0.074, 0.0, 0.0027, 0.0283, 0.0283, 0.056, 0.0738, 0.0695, 0.0457, 0.0121, 0.0593, 0.1027, 0.1272, 0.1776, 0.1542, 0.0769, 0.0606, 0.0383, 0.0071, 0.1099, 0.1875, 0.2352, 0.2337, 0.2802, 0.2802, 0.2718, 0.3151, 0.3073, 0.2751, 0.2751, 0.4091, 0.3945, 0.5343, 0.5169, 0.461, 0.5417, 0.5452, 0.5106, 0.4892, 0.5039, 0.6206, 0.6206, 0.6306, 0.7383, 0.7762, 0.7854, 0.7383, 0.8327, 0.9446, 0.9469, 0.8535, 0.9559, 0.8738, 0.695, 0.7106, 0.7633, 0.7133, 0.6675, 0.8332, 0.8494, 0.8391, 0.8391, 0.786, 0.8806, 0.8645, 0.9128, 0.8807, 0.8675, 0.9166, 0.8565, 0.8423, 0.8436, 0.6962, 0.7904, 0.7738, 0.6543, 0.7393, 0.8009, 0.8705, 0.8109, 0.9037, 0.9737, 1.0626, 1.0628, 0.9848, 0.8893, 0.8064, 0.767, 0.707, 0.588, 0.5959, 0.4382, 0.5099, 0.6441, 0.6369, 0.6369, 0.6883, 0.6875, 0.8178, 0.8839, 0.8798, 0.9065, 1.0812, 1.0391, 1.0435, 1.0339, 1.0042, 1.0084, 1.1787, 1.153, 1.22, 1.3444, 1.2539, 1.3172, 1.3114, 1.4233, 1.5764, 1.8613, 1.9792, 1.89, 2.3378, 2.5546, 2.4261, 2.5917, 2.4683, 2.2388, 2.046, 2.1229]}
{"unit_id": "orig:1992209835249819925", "ticker_idx": 4, "ts": "2025-11-22T12:33:32+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's 2026 HBM share for NVIDIA are projected at... 15%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises NVDA/AMD GPU forecasts, names SK Hynix biggest HBM beneficiary at 60-70% share with 13% ASP growth; Micron 24% and Samsung 15% NVIDIA HBM share.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s latest semiconductor supply chain update:\n\n• The 2026 NVIDIA Rubin GPU unit forecast has been revised up to 3 million units (NVIDIA’s total 2026 GPU procurement volume has been revised up to 7.5 million units).\n    \n• The AMD MI450 unit forecast has been revised up.\n    \n• The AWS ASIC unit forecast has been slightly revised down.\n    \n• There is a possibility that Rubin’s initial ramp could be pulled forward.\n   \n • The OpenAI ASIC forecast has been revised down (from 700,000 units to 500,000 units).\n    \n• Instead, the assumption for HBM3E has been revised from 8-Hi to 12-Hi.\n    \n• SK Hynix is expected to be the biggest beneficiary.\n    \n• Hynix has effectively passed qualification for NVIDIA Rubin’s HBMA4 and is expected to begin full-scale shipments from 2Q26.\n    \n• SK Hynix’s HBM4 share for NVIDIA in 2026 is projected to be around 70%, with its overall HBM market share expected to be around 60–70%.\n    \n• There is a possibility that HBM4 prices will carry about a 30% premium to HBM3E 12-Hi, and based on this, SK Hynix’s HBM ASP (average selling price) in 2026 is expected to grow about 13% YoY.\n    \n• Micron’s and Samsung’s 2026 HBM share for NVIDIA are projected at 24% and 15%, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01964843134458627, "ret_signed_m1d": -0.01964843134458627, "alpha_spy_m1d": -0.005143416446618643, "alpha_c_m1d": 0.0036255541335565855, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.023273985478142856, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026887284286339463, "ret_signed_p1d": 0.026887284286339463, "alpha_spy_p1d": 0.017481384634232544, "alpha_c_p1d": 0.024205859810904284, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.002681424475435179, "ret_p1w": 0.04239921661394108, "ret_signed_p1w": 0.04239921661394108, "alpha_spy_p1w": 0.02514252262538408, "alpha_c_p1w": 0.018660256880602244, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.023738959733338838, "ret_p1m": 0.1530506263321878, "ret_signed_p1m": 0.1530506263321878, "alpha_spy_p1m": 0.12125467979469295, "alpha_c_p1m": 0.10788685638219397, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.045163769949993826, "ret_p3m": 0.9743910791481385, "ret_signed_p3m": 0.9743910791481385, "alpha_spy_p3m": 0.947814492689671, "alpha_c_p3m": 0.9703320180088377, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.004059061139300768, "ret_p6m": 1.8687926787213338, "ret_signed_p6m": 1.8687926787213338, "alpha_spy_p6m": 1.7653460487480053, "alpha_c_p6m": 1.6266596875507027, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.24213299117063114, "price_path": [-0.2731, -0.2834, -0.2824, -0.304, -0.2886, -0.2814, -0.2783, -0.2845, -0.2783, -0.2639, -0.2526, -0.2443, -0.2237, -0.2124, -0.1826, -0.1949, -0.1733, -0.1733, -0.1403, -0.128, -0.1208, -0.1136, -0.1424, -0.1293, -0.1324, -0.1107, -0.0719, -0.0719, -0.0719, -0.0719, -0.0719, -0.0719, -0.0238, -0.0352, -0.0527, -0.0176, 0.0103, 0.0124, 0.0145, 0.0083, 0.0196, -0.0021, 0.0217, 0.0548, 0.029, 0.0393, 0.0765, 0.1117, 0.1489, 0.0848, 0.0403, 0.0259, 0.0124, 0.0403, 0.0703, 0.0662, 0.0631, 0.0052, 0.0403, 0.0114, -0.0021, 0.0403, -0.0196, 0.0, 0.0269, 0.0631, 0.0703, 0.0393, 0.0424, 0.0693, 0.0807, 0.0869, 0.121, 0.1324, 0.121, 0.1169, 0.1096, 0.1262, 0.0838, 0.0631, 0.1158, 0.1127, 0.0993, 0.1427, 0.1531, 0.1489, 0.1489, 0.2099, 0.2418, 0.2459, 0.2459, 0.2459, 0.3353, 0.4351, 0.4434, 0.4652, 0.4423, 0.4444, 0.4423, 0.4299, 0.4579, 0.4953, 0.5473, 0.5515, 0.5089, 0.5535, 0.5826, 0.5806, 0.5806, 0.6574, 0.6876, 0.6699, 0.6678, 0.5629, 0.7406, 0.7572, 0.6554, 0.6481, 0.7292, 0.7229, 0.7437, 0.8559, 0.8829, 0.8829, 0.8829, 0.8829, 0.9744, 0.9754, 1.0056, 1.0783, 1.1147, 1.2654, 1.2498, 1.2498, 1.0274, 0.7894, 0.991, 0.9557, 0.8029, 0.9526, 0.9744, 0.9526, 0.9068, 0.9609, 1.0149, 1.1666, 1.0835, 1.0721, 0.9359, 0.9713, 0.964, 0.8715, 0.8715, 0.8358, 0.7411, 0.9748, 0.8577, 0.9389, 1.0108, 1.0462, 1.1919, 1.1243, 1.1451, 1.093, 1.1503, 1.1972, 1.2648, 1.2492, 1.2336, 1.2805, 1.2648, 1.3377, 1.2857, 1.3377, 1.3117, 1.3533, 1.2961, 1.2961, 1.421, 1.421, 1.7699, 1.8271, 1.7959, 1.9729, 1.9052, 1.9573, 2.0823, 1.8167, 1.9261, 1.8688]}
{"unit_id": "quote:1975344183482458216", "ticker_idx": 0, "ts": "2025-10-06T23:35:28+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley previously believed that the DRAM price rebound through summer 2025 was merely temporary… We now observe double-digit QoQ DRAM price increases occurring without meaningful demand recovery… we now see Micron's quarterly EPS reaching as high as $5 in the near term… raising our price target accordingly", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Morgan Stanley upgrade of MU to Overweight, PT $220, on structural DRAM supply tightness, revised HBM view, DDR5-HBM pricing linkage, and AI-premium re-rating.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Why We Changed Our View\n\n⸻\n\n1. Revising Our Underestimation of the DRAM Cycle\n\n(1) Previous View\n\nMorgan Stanley previously believed that the DRAM price rebound through summer 2025 was merely temporary and not a sign of structural strength, given weak demand across PCs, mobile, and servers.\nAccordingly, we maintained a cautiously neutral (Equal-weight) stance on Micron, expecting only a limited recovery in industry conditions.\n\n(2) What Changed\n\nWe now observe double-digit QoQ DRAM price increases occurring without meaningful demand recovery, driven purely by supply tightness.\nThis suggests a structural supply constraint, not a simple “inventory normalization” phase.\n\nIn short, the current price strength cannot be explained by traditional demand models — we recognize that we underestimated the impact of coordinated shipment control among suppliers.\n\n⸻\n\n2. Withdrawing Our Bearish View on HBM (Especially HBM4)\n\n(1) Previous View\n\nWe believed Micron was technologically behind in meeting NVIDIA’s HBM4 speed requirement (>9.6Gbps), implying inevitable market share loss and margin pressure in the HBM segment.\n\n(2) What Changed\n\nWe now correct that — Micron’s disadvantage has been overstated.\n\nAlthough its HBM shipments are roughly one quarter behind SK hynix, technical maturity remains strong, and Micron is expected to maintain a low-20% market share in CY26.\n\nWhile some yield loss may occur as Micron aligns with NVIDIA’s spec demands, this could actually tighten overall HBM supply, helping support industry-wide ASPs.\n\nMoreover, the transition to TSMC base dies for HBM4E may serve as a structural advantage for Micron rather than a risk.\n\n⸻\n\n3. Newly Recognizing the Pricing Leverage Between DDR5 and HBM\n\nIn the past, we treated DRAM and HBM as separate markets.\nNow, we emphasize that DDR5 price increases serve as direct leverage in HBM price negotiations.\n\nThis interlinkage — previously excluded from our models — limits potential downside in HBM pricing and supports gross margin stability.\n\nIn other words, the old narrative that “HBM cannibalizes DRAM” has been replaced with a new one: “DRAM sustains HBM.”\n\n⸻\n\n4. Acknowledging Upward Pressure on Earnings Estimates — “The Start of an EPS Revision Cycle”\n\nDriven by recent sharp DDR5/DRAM price hikes and inventory drawdowns, we now see Micron’s quarterly EPS reaching as high as $5 in the near term.\n\nMorgan Stanley noted, “We thought just last week…”, implying that even their own estimates from one week ago were overly conservative.\nIn short, we are correcting an underestimation of both the speed and magnitude of the price uptrend.\n\n⸻\n\n5. Reassessing Valuation — From “Conservative 10x” to “AI Premium 22x”\n\nWe have abandoned the previous framework that applied a 10x PER, based on the 2021 peak.\nInstead, we now reflect structural growth drivers tied to AI, applying a through-cycle EPS of $10 × 22x multiple, raising our price target accordingly.\n\nThis is not a simple re-rating — it represents a fundamental paradigm shift, acknowledging that “AI memory deserves a valuation premium.”\n\n⸻\n\n6. Key Misjudgments We Corrected\n    1. Viewed DRAM strength as temporary → Underestimated supply tightness and structural inventory drawdown.\n    2. Overstated HBM4 competitiveness concerns → Misjudged Micron’s technological maturity and yield improvement pace.\n    3. Ignored DRAM–HBM price interdependence → In reality, rising DDR5 prices strengthen HBM pricing power.\n    4. Underestimated EPS elasticity → Micron’s profit ramp-up potential is much higher, even in the near term.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": "Micron: Rating Upgraded from Equal-weight → Overweight, Price Target Raised from $160 → $220 (Morgan Stanley)\n\n    • DRAM and NAND prices are improving faster than expected, with buyers increasing inventories amid concerns over supply shortages through CY26.\n    • Concerns", "parent_author": null, "ret_m1d": -0.016390887466603887, "ret_signed_m1d": -0.016390887466603887, "alpha_spy_m1d": -0.01281747363625485, "alpha_c_m1d": 0.003121061077802856, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.019511948544406743, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027597379748390383, "ret_signed_p1d": -0.027597379748390383, "alpha_spy_p1d": -0.023889887497428486, "alpha_c_p1d": -0.009162994642889943, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.01843438510550044, "ret_p1w": 0.009478387210694939, "ret_signed_p1w": 0.009478387210694939, "alpha_spy_p1w": 0.022238777648518115, "alpha_c_p1w": 0.020049777549247083, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.010571390338552145, "ret_p1m": 0.14175739704783452, "ret_signed_p1m": 0.14175739704783452, "alpha_spy_p1m": 0.13635240070277388, "alpha_c_p1m": 0.11444066908566497, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.02731672796216955, "ret_p3m": 0.4952100244648945, "ret_signed_p3m": 0.4952100244648945, "alpha_spy_p3m": 0.4768585547478812, "alpha_c_p3m": 0.4431668395521733, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.052043184912721197, "ret_p6m": 0.7706245009775721, "ret_signed_p6m": 0.7706245009775721, "alpha_spy_p6m": 0.7967809433497164, "alpha_c_p6m": 0.6506029942840996, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.12002150669347245, "price_path": [-0.3603, -0.3557, -0.3483, -0.3793, -0.3714, -0.3907, -0.4073, -0.4013, -0.4074, -0.4284, -0.4252, -0.4153, -0.4177, -0.4178, -0.4141, -0.3995, -0.4288, -0.4511, -0.436, -0.4292, -0.4307, -0.4145, -0.3778, -0.3525, -0.3314, -0.3496, -0.3443, -0.3674, -0.3534, -0.3613, -0.3866, -0.394, -0.3841, -0.3907, -0.3903, -0.3838, -0.3615, -0.3772, -0.3772, -0.3799, -0.3787, -0.35, -0.3125, -0.312, -0.2922, -0.2673, -0.212, -0.1771, -0.1743, -0.1688, -0.1627, -0.1161, -0.1484, -0.1385, -0.1291, -0.1537, -0.1792, -0.1769, -0.1422, -0.1243, -0.0467, -0.0384, -0.0164, 0.0, -0.0276, 0.0292, 0.0072, -0.049, 0.0095, -0.0204, 0.0051, 0.0606, 0.0598, 0.0828, 0.0593, 0.0393, 0.0825, 0.1469, 0.1526, 0.1621, 0.1868, 0.1731, 0.1718, 0.2291, 0.1418, 0.2437, 0.2481, 0.2459, 0.3265, 0.2626, 0.2825, 0.2408, 0.2926, 0.267, 0.1966, 0.1831, 0.0545, 0.0859, 0.1727, 0.1758, 0.2058, 0.2058, 0.2384, 0.2592, 0.2541, 0.2262, 0.1869, 0.2422, 0.293, 0.3218, 0.381, 0.3535, 0.2628, 0.2437, 0.2176, 0.181, 0.3016, 0.3925, 0.4484, 0.4467, 0.5013, 0.5013, 0.4914, 0.5421, 0.533, 0.4952, 0.4952, 0.6524, 0.6353, 0.7992, 0.7788, 0.7132, 0.8079, 0.8119, 0.7714, 0.7464, 0.7635, 0.9004, 0.9004, 0.9122, 1.0385, 1.0828, 1.0937, 1.0384, 1.1492, 1.2804, 1.283, 1.1735, 1.2936, 1.1974, 0.9876, 1.0059, 1.0677, 1.0091, 0.9554, 1.1497, 1.1687, 1.1566, 1.1566, 1.0944, 1.2053, 1.1864, 1.2431, 1.2054, 1.1899, 1.2475, 1.177, 1.1603, 1.1619, 0.9891, 1.0996, 1.0801, 0.9399, 1.0396, 1.1118, 1.1934, 1.1236, 1.2324, 1.3145, 1.4187, 1.4189, 1.3274, 1.2155, 1.1183, 1.0721, 1.0017, 0.8622, 0.8714, 0.6866, 0.7706]}
{"unit_id": "thread:2003812123558941035", "ticker_idx": 1, "ts": "2025-12-24T12:56:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' MX Business Division has reportedly set sales targets for the Galaxy S26 and Z Fold/Flip 8 series 10% higher than their predecessors", "tweet_type": "analysis", "timeline": "months", "summary": "Mobile DRAM super-cycle driving smartphone price hikes across Xiaomi, Samsung, and Apple; bullish on mobile DRAM prices and Samsung's higher ASP trajectory into 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Xiaomi to Hike Prices for New Smartphones; Samsung and Apple Weigh Options\n\nChina's Xiaomi is set to implement a price increase for its new smartphone models. This decision comes as the company can no longer withstand the pressure of soaring memory prices. With Samsung Electronics and Apple also preparing for next year's product launches, both giants appear to be agonizing over the possibility of raising their own prices.\n\nAccording to industry sources on the 24th, Xiaomi will unveil its new flagship, the \"Xiaomi 17 Ultra,\" on the 25th at 7:00 PM (local time in China). The official domestic starting price is expected to be at least 6,999 yuan (approximately 1.45 million KRW). This represents a roughly 10% increase compared to the previous \"Xiaomi 15 Ultra\" (6,499 yuan). Global launch prices are anticipated to be set about 10% higher than the Chinese domestic prices.\n\nIndustry experts agree that Xiaomi's price hike is a direct consequence of the supply shortage in mobile DRAM. Facing a \"semiconductor super cycle,\" the surge in mobile DRAM prices has led to a significant increase in component costs. Since DRAM accounts for about 20% of a smartphone's total bill of materials (BOM), a rise in DRAM prices inevitably leads to higher product pricing.\n\nDuring the third-quarter earnings call, Xiaomi President Lu Weibing foreshadowed the price hike, stating, \"We have reached a point where price increases alone cannot offset the burden of rising costs.\" To prepare for even sharper memory price hikes expected next year, the company also moved up the release date. The new flagship is being revealed just nine months after the launch of the Xiaomi 15 Ultra last March.\n\nFollowing Xiaomi's move, changes in pricing strategies are expected across the entire smartphone market. Samsung Electronics' MX (Mobile eXperience) Business Division has reportedly set sales targets for the Galaxy S26 and Z Fold/Flip 8 series 10% higher than their predecessors. This is interpreted as a countermeasure to offset declining profit margins caused by rising manufacturing costs.\n\nThe race to secure memory is also intensifying. Roh Tae-moon, President of the DX (Device eXperience) Division overseeing smartphones and home appliances, is scheduled to hold a high-level meeting with Micron in the U.S. during CES 2026 this coming January. This move comes as memory supply alarms go off; despite requesting long-term DRAM contracts from Samsung's own DS (Device Solutions) division, they were only able to secure a 3-month short-term contract.\n\nApple is also not immune to \"Phone-flation\" (smartphone + inflation). While the global launch prices for the iPhone 17 series released last September remained unchanged from the previous generation, the iPhone 18 series faces increased cost pressures due to DRAM upgrades and other factors. Furthermore, with the upcoming launch of Apple's first foldable phone, the use of more advanced technology and high-performance components is inevitable compared to the standard series.\n\nCounterpoint Research, a market research firm, projected that \"manufacturing costs for smartphones will increase by 25%, 15%, and 10% for low-end, mid-range, and high-end models, respectively, next year,\" and added that \"consequently, the average selling price (ASP) is expected to rise by 6.9%.\"\n\n---\n\nhttps://t.co/AxJqaf394x", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0036003552560108787, "ret_signed_m1d": 0.0036003552560108787, "alpha_spy_m1d": 0.007105645282170503, "alpha_c_m1d": 0.005992791939719799, "bench_spy_m1d": -0.0035052900261596243, "bench_c_m1d": -0.0023924366837089206, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.08445416491506297, "ret_signed_p1w": 0.08445416491506297, "alpha_spy_p1w": 0.09670827446226882, "alpha_c_p1w": 0.10038042063887109, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.015926255723808125, "ret_p1m": 0.3775010049372032, "ret_signed_p1m": 0.3775010049372032, "alpha_spy_p1m": 0.37952886233454775, "alpha_c_p1m": 0.3872071341386003, "bench_spy_p1m": -0.002027857397344568, "bench_c_p1m": -0.009706129201397129, "ret_p3m": 0.6850192239158699, "ret_signed_p3m": 0.6850192239158699, "alpha_spy_p3m": 0.7331233700620233, "alpha_c_p3m": 0.747765183585298, "bench_spy_p3m": -0.04810414614615344, "bench_c_p3m": -0.06274595966942809, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2536, -0.2421, -0.2448, -0.2259, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1503, -0.1602, -0.1755, -0.1449, -0.1206, -0.1188, -0.117, -0.1224, -0.1125, -0.1314, -0.1107, -0.0819, -0.1044, -0.0954, -0.063, -0.0324, 0.0, -0.0558, -0.0945, -0.1071, -0.1188, -0.0945, -0.0684, -0.072, -0.0747, -0.1251, -0.0945, -0.1197, -0.1314, -0.0945, -0.1467, -0.1296, -0.1062, -0.0747, -0.0684, -0.0954, -0.0927, -0.0693, -0.0594, -0.054, -0.0243, -0.0144, -0.0243, -0.0279, -0.0342, -0.0198, -0.0567, -0.0747, -0.0288, -0.0315, -0.0432, -0.0054, 0.0036, 0.0, 0.0, 0.0531, 0.0808, 0.0845, 0.0845, 0.0845, 0.1622, 0.2491, 0.2563, 0.2753, 0.2554, 0.2572, 0.2554, 0.2445, 0.269, 0.3015, 0.3467, 0.3504, 0.3133, 0.3522, 0.3775, 0.3757, 0.3757, 0.4426, 0.4689, 0.4535, 0.4517, 0.3603, 0.515, 0.5295, 0.4408, 0.4345, 0.505, 0.4996, 0.5177, 0.6154, 0.6389, 0.6389, 0.6389, 0.6389, 0.7185, 0.7194, 0.7456, 0.8089, 0.8406, 0.9717, 0.9582, 0.9582, 0.7646, 0.5575, 0.733, 0.7022, 0.5692, 0.6995, 0.7185, 0.6995, 0.6597, 0.7067, 0.7538, 0.8858, 0.8135, 0.8035, 0.685, 0.7158, 0.7094, 0.6289, 0.6289, 0.5979, 0.5154, 0.7189, 0.6169, 0.6876, 0.7501, 0.781, 0.9078, 0.8489, 0.8671, 0.8217, 0.8716, 0.9124, 0.9713, 0.9577, 0.9441, 0.9849, 0.9713, 1.0347, 0.9894, 1.0347, 1.0121, 1.0483, 0.9985, 0.9985, 1.1072, 1.1072, 1.4109, 1.4607, 1.4335, 1.5876, 1.5287, 1.574, 1.6828, 1.4516, 1.5468, 1.497, 1.5015, 1.7145, 1.651, 1.651, 1.7099, 1.7825, 1.7145, 1.8731, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2003658302836871398", "ticker_idx": 0, "ts": "2025-12-24T02:45:40+00:00", "symbol": "0981.HK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The move is widely viewed as a potential inflection point for supply-demand dynamics in China's mature-node foundry segment, signaling that pricing is gradually emerging from the cycle trough", "tweet_type": "analysis", "timeline": "quarters", "summary": "SMIC's ~10% price hike signals mature-node pricing inflection; multiple structural drivers (AI demand, rising input costs, tightening supply) support sustained price recovery.", "resolved_tickers": ["0981.HK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SMIC signals mature-node price turnaround with reported 10% hike\n\nChina's leading foundry, SMIC, is reported to have issued price increase notices to selected customers, marking the start of a pricing adjustment. The company has yet to formally confirm the move.\n\nIndustry sources in China say that from December 23, parts of SMIC's mature-process capacity notified customers of price increases of around 10%, with multiple downstream clients already receiving the notices.\n\nThe move is widely viewed as a potential inflection point for supply-demand dynamics in China's mature-node foundry segment, signaling that pricing is gradually emerging from the cycle trough.\n\nAnalysts say the adjustment reflects multiple structural drivers rather than a single trigger.\n\nFirst, AI server demand has surged over the past year, becoming a key engine of semiconductor growth. Beyond advanced logic, AI workloads are also lifting shipments of mature-node products such as memory, power management ICs, and peripheral control chips, gradually tightening previously loose mature-process capacity.\n\nAnother factor is rising input costs across the supply chain. Since the second half of 2025, prices for key materials, including metals, chemicals, and energy, have continued to climb, pushing up manufacturing costs. For mature processes built on large-scale capacity, the ability to absorb these pressures is nearing its limit, making price adjustments increasingly unavoidable.\n\nThird, structural shifts in the global foundry landscape have also strengthened pricing power. As leading players exit mature nodes, the global supply of 8-inch wafers and certain mature processes has tightened, improving pricing leverage for major suppliers such as SMIC.\n\nSMIC's management has long taken a cautious approach to pricing. Co-CEO Zhao Haijun previously stated that the company would not be the first to raise prices, but would not rule out following peers. With players such as TSMC adjusting capacity allocation and mature-node supply tightening, the move is seen as a response to more supportive conditions.\n\nOperationally, SMIC's 8-inch equivalent monthly capacity reached 1.023 million wafers by the end of the third quarter of 2025. Revenue from 12-inch and 8-inch wafers accounted for 77% and 23%, with utilization at 95.8%. Third-quarter shipments totaled about 2.499 million 8-inch equivalent wafers, up 4.6% quarter on quarter, while average selling prices rose 3.8%, driven by a higher mix of complex, higher value-added mature-node products.\n\nOn market trends, Zhao said AI demand has driven rapid increases in memory prices, with cost pressures spreading across the semiconductor supply chain. As memory customers seek foundry capacity, some displacement of other chips may occur, creating additional room for price adjustments.\n\nAgainst this backdrop, the moderate rise in SMIC's average wafer prices reflects its effort to balance supply, demand, and costs. As mature-node conditions continue to improve, the move may serve as an early indicator of the next phase of price recovery.\n\n---\n\nhttps://t.co/UJKUZ8ZMlj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028893523756720496, "ret_signed_m1d": -0.028893523756720496, "alpha_spy_m1d": -0.025388233730560872, "alpha_c_m1d": -0.026120067051380258, "bench_spy_m1d": -0.0035052900261596243, "bench_c_m1d": -0.002773456705340238, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.005637794828528442, "ret_signed_p1w": 0.005637794828528442, "alpha_spy_p1w": 0.01789190437573429, "alpha_c_p1w": 0.016731537849529743, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.011093743021001301, "ret_p1m": 0.09725161131313986, "ret_signed_p1m": 0.09725161131313986, "alpha_spy_p1m": 0.09927946871048443, "alpha_c_p1m": -0.008880119667928454, "bench_spy_p1m": -0.002027857397344568, "bench_c_p1m": 0.10613173098106832, "ret_p3m": -0.23678641501026176, "ret_signed_p3m": -0.23678641501026176, "alpha_spy_p3m": -0.18868226886410833, "alpha_c_p3m": -0.31144926730802147, "bench_spy_p3m": -0.04810414614615344, "bench_c_p3m": 0.07466285229775971, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0282, 0.0782, 0.1212, 0.1212, 0.2636, 0.2812, 0.2833, 0.2833, 0.2615, 0.1769, 0.093, 0.1297, 0.0338, 0.0712, 0.0416, -0.0261, 0.012, 0.0437, 0.055, 0.0437, 0.1276, 0.167, 0.129, 0.129, 0.1163, 0.0571, 0.0268, 0.0127, 0.0106, 0.0846, 0.0634, 0.0529, 0.024, 0.0324, 0.0655, 0.0359, 0.0282, 0.043, 0.0289, 0.0359, -0.0303, -0.0409, -0.0423, -0.0296, -0.0366, -0.0303, -0.0211, -0.0324, -0.0529, -0.0162, -0.0169, 0.012, -0.0296, -0.0317, -0.0536, -0.0458, -0.0881, -0.1057, -0.0874, -0.0881, -0.0832, -0.0289, -0.0289, 0.0, 0.0, 0.0, -0.0197, 0.0218, 0.0056, 0.0056, 0.0585, 0.0782, 0.0655, 0.0529, 0.0564, 0.0486, 0.0613, 0.0493, 0.0705, 0.0902, 0.1163, 0.0853, 0.05, 0.0888, 0.0973, 0.1008, 0.0648, 0.0796, 0.1177, 0.0888, 0.0627, 0.0176, -0.007, -0.031, -0.0493, -0.0472, -0.0085, 0.0085, -0.0134, -0.0162, -0.0085, -0.0078, -0.0078, -0.0078, -0.0078, -0.0458, 0.0021, -0.0218, -0.0155, -0.0472, -0.0416, -0.0895, -0.1184, -0.1367, -0.1332, -0.129, -0.1424, -0.0951, -0.1015, -0.1106, -0.1233, -0.1191, -0.1283, -0.1304, -0.1579, -0.198, -0.2368, -0.222, -0.2037, -0.2509, -0.26, -0.2755, -0.2854, -0.2551, -0.2812, -0.2812, -0.2812, -0.2812, -0.2086, -0.2149, -0.179, -0.1945, -0.1832, -0.1642, -0.155, -0.1635, -0.1572, -0.1529, -0.1642, -0.1762, -0.0937, -0.0381, -0.0691, -0.0726, -0.0007, -0.0007, 0.0169, -0.0021, 0.055, 0.0817, 0.0338, 0.0796, 0.0796, 0.0451, 0.0078, 0.0028, -0.0317, -0.0345, 0.0592, 0.0458, 0.1254, 0.1254, 0.1896, 0.2008, 0.2438, 0.1501, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1993201908249710881", "ticker_idx": 0, "ts": "2025-11-25T06:15:41+00:00", "symbol": "Baidu", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JP Morgan stated that Baidu's investment narrative is undergoing a fundamental shift... sales of Kunlun chips and demand for GPU computing are expected to significantly drive revenue growth rates", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM Buy-thesis on Baidu: transitioning to AI infrastructure provider with GPU compute revenue set to double in 2026, cloud valued at $34B.", "resolved_tickers": ["BIDU"], "resolution_method": "exchange_ticker", "resolution_reason": "Baidu Inc. BIDU US ADR", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Kunlun Chip Sales Increase, Baidu's Cloud Business Valued at $34 Billion\n\nIn a report dated November 23, JP Morgan stated that Baidu's investment narrative is undergoing a fundamental shift. The company is transitioning from a traditional search advertising company to an AI infrastructure provider, and sales of Kunlun chips and demand for GPU computing are expected to significantly drive revenue growth rates.\n\nThis forecast is based on a structural shift where major Chinese cloud service providers are accelerating the adoption of domestic AI accelerators. Starting in 2025, China's major hyperscale cloud companies—Baidu, Alibaba, Tencent, and ByteDance—are accelerating the purchase and deployment of domestic AI accelerators, moving from pilot applications to large-scale multi-vendor procurement.\n\nJP Morgan expects Baidu's GPU computing revenue (subscription-based AI cloud infrastructure revenue) to double in 2026. Third-quarter GPU rental subscription revenue increased by 128% year-over-year, accelerating further compared to the approximately 50% increase in the previous quarter.\n\nAnalysts stated that starting from 2025, Baidu's cloud business has clearly shifted to an 'AI-first' and 'GPU-intensive' growth phase. AI cloud revenue increased by 26% year-over-year in the fourth quarter, accelerated to 42% in the first quarter, and maintained robust growth of 27% and 21% in the second and third quarters, respectively. During this period, AI cloud has grown to account for approximately one-quarter of Baidu's core revenue, becoming a major growth engine.\n\nIn the Q1 Large Language Model (LLM) bidding market, Baidu Smart Cloud ranked first by winning 19 projects and recording a bid amount of approximately 450 million RMB, leading other major model suppliers. These contracts typically involve multi-year projects to deploy generative AI applications on Baidu AI Cloud, locking in future revenue streams for GPU and accelerator usage.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011282676128634872, "ret_signed_m1d": 0.011282676128634872, "alpha_spy_m1d": 0.020600929227164055, "alpha_c_m1d": 0.02256882193154186, "bench_spy_m1d": -0.009318253098529183, "bench_c_m1d": -0.011286145802906988, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01306414108762366, "ret_signed_p1d": -0.01306414108762366, "alpha_spy_p1d": -0.019967716100643185, "alpha_c_p1d": -0.01446387522143977, "bench_spy_p1d": 0.006903575013019525, "bench_c_p1d": 0.0013997341338161107, "ret_p1w": 0.009416361013016461, "ret_signed_p1w": 0.009416361013016461, "alpha_spy_p1w": -0.00022789272332812693, "alpha_c_p1w": 0.0012799529043934221, "bench_spy_p1w": 0.009644253736344588, "bench_c_p1w": 0.008136408108623039, "ret_p1m": 0.04648798341208571, "ret_signed_p1m": 0.04648798341208571, "alpha_spy_p1m": 0.02071092656403084, "alpha_c_p1m": 0.009984991678551713, "bench_spy_p1m": 0.02577705684805487, "bench_c_p1m": 0.036502991733533996, "ret_p3m": 0.15252802660217446, "ret_signed_p3m": 0.15252802660217446, "alpha_spy_p3m": 0.12816245703450213, "alpha_c_p3m": 0.1273475914941875, "bench_spy_p3m": 0.02436556956767233, "bench_c_p3m": 0.025180435107986954, "ret_p6m": 0.14709878156857847, "ret_signed_p6m": 0.14709878156857847, "alpha_spy_p6m": 0.04273047304994093, "alpha_c_p6m": 0.12481058842945236, "bench_spy_p6m": 0.10436830851863754, "bench_c_p6m": 0.02228819313912611, "price_path": [-0.2283, -0.1916, -0.1916, -0.1832, -0.1822, -0.1669, -0.1351, -0.0783, -0.0798, -0.0872, -0.0514, -0.0263, -0.026, 0.0501, 0.1692, 0.1474, 0.1482, 0.1591, 0.0652, 0.1276, 0.1486, 0.1142, 0.144, 0.1178, 0.1659, 0.1896, 0.2046, 0.2293, 0.1787, 0.1698, 0.1232, 0.0323, 0.0665, 0.0175, 0.0314, 0.0255, 0.0181, 0.036, 0.0107, -0.0072, 0.0221, 0.0414, 0.0914, 0.0758, 0.0825, 0.0333, 0.0254, 0.0284, 0.0603, 0.0462, 0.0777, 0.0685, 0.1225, 0.1194, 0.0938, 0.0235, -0.0159, -0.032, -0.0063, -0.0203, -0.063, -0.0588, 0.0113, 0.0, -0.0131, -0.0131, -0.0084, 0.0155, 0.0094, -0.0051, 0.0071, 0.066, 0.1033, 0.0514, 0.0695, 0.0885, 0.0605, 0.0081, 0.0121, 0.0137, 0.0221, 0.0504, 0.0554, 0.0505, 0.0465, 0.0465, 0.0587, 0.0758, 0.123, 0.1084, 0.1084, 0.275, 0.2698, 0.2421, 0.2454, 0.1983, 0.2177, 0.2917, 0.267, 0.2767, 0.2674, 0.2685, 0.2685, 0.2727, 0.3767, 0.3787, 0.3695, 0.3241, 0.3373, 0.3375, 0.3362, 0.2995, 0.2603, 0.2289, 0.1703, 0.1788, 0.238, 0.2468, 0.2526, 0.2311, 0.1739, 0.1623, 0.1623, 0.165, 0.1696, 0.1631, 0.1525, 0.1362, 0.1336, 0.1253, 0.0617, 0.0556, 0.0482, 0.007, 0.0093, 0.0004, 0.0099, 0.039, 0.0646, 0.0617, 0.0448, 0.0525, 0.0333, 0.0238, 0.0338, 0.0096, -0.0307, -0.0284, -0.0454, -0.0193, -0.0676, -0.0828, -0.0957, -0.0548, -0.0507, -0.0587, -0.0587, -0.056, -0.0617, -0.0371, -0.0819, -0.0803, -0.0498, 0.003, 0.026, 0.0607, 0.07, 0.0837, 0.0476, 0.0459, 0.031, 0.0919, 0.0859, 0.0668, 0.0266, 0.0734, 0.0679, 0.0814, 0.0709, 0.1927, 0.1865, 0.1966, 0.2367, 0.1871, 0.2767, 0.2156, 0.148, 0.1682, 0.168, 0.1471]}
{"unit_id": "orig:1998271269524172914", "ticker_idx": 0, "ts": "2025-12-09T05:59:31+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "memory prices will peak around late 2026", "tweet_type": "prediction", "timeline": "years", "summary": "Memory prices will peak ~late 2026, not sustain through end-2027; bearish on memory pricing outlook beyond 2026.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I’d like to point out that expecting memory prices to keep rising all the way through the end of 2027 is far too optimistic.\n\nIn my view, memory prices will peak around late 2026. https://t.co/tf381E5M1a", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0025976912871312763, "ret_signed_m1d": -0.0025976912871312763, "alpha_spy_m1d": -0.00173380065732065, "alpha_c_m1d": -0.0038172151170002833, "bench_spy_m1d": 0.0008638906298106264, "bench_c_m1d": -0.001219523829869007, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026046470450656556, "ret_signed_p1d": -0.026046470450656556, "alpha_spy_p1d": -0.019414202441641826, "alpha_c_p1d": -0.012225283339613147, "bench_spy_p1d": 0.00663226800901473, "bench_c_p1d": 0.01382118711104341, "ret_p1w": -0.06471374278808169, "ret_signed_p1w": 0.06471374278808169, "alpha_spy_p1w": 0.058608813821585296, "alpha_c_p1w": 0.01848068182889713, "bench_spy_p1w": -0.006104928966496392, "bench_c_p1w": -0.04623306095918456, "ret_p1m": 0.3212457793389265, "ret_signed_p1m": -0.3212457793389265, "alpha_spy_p1m": -0.3086875972531047, "alpha_c_p1m": -0.274584474705607, "bench_spy_p1m": 0.012558182085821823, "bench_c_p1m": 0.04666130463331952, "ret_p3m": 0.6159051131338579, "ret_signed_p3m": -0.6159051131338579, "alpha_spy_p3m": -0.6286029484016963, "alpha_c_p3m": -0.5813681153111642, "bench_spy_p3m": -0.012697835267838409, "bench_c_p3m": 0.034536997822693705, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3959, -0.3683, -0.3628, -0.3425, -0.3532, -0.3221, -0.3302, -0.3206, -0.3086, -0.3148, -0.3196, -0.3394, -0.3194, -0.3167, -0.2833, -0.2488, -0.2432, -0.2391, -0.246, -0.2317, -0.2373, -0.218, -0.2143, -0.2291, -0.2058, -0.1658, -0.1578, -0.1396, -0.1512, -0.1513, -0.1487, -0.1068, -0.0808, -0.0944, -0.0633, -0.0498, -0.045, 0.0164, -0.0446, -0.0363, -0.0312, -0.0434, 0.0005, 0.0009, 0.0035, -0.0108, -0.0456, -0.0145, -0.0621, -0.0742, -0.0887, -0.128, -0.1009, -0.0927, -0.0714, -0.0573, -0.0665, -0.0557, -0.0372, -0.0444, -0.0583, -0.033, 0.0026, 0.0, 0.026, 0.004, -0.0104, -0.0378, -0.0647, -0.0459, -0.0158, 0.0002, 0.0466, 0.0516, 0.0665, 0.0665, 0.0886, 0.1351, 0.1405, 0.1309, 0.1309, 0.2125, 0.249, 0.3105, 0.3212, 0.3061, 0.3236, 0.3269, 0.3065, 0.3109, 0.3305, 0.3846, 0.3905, 0.3684, 0.4118, 0.4405, 0.4497, 0.4175, 0.506, 0.5721, 0.5793, 0.5794, 0.5319, 0.6058, 0.5538, 0.4939, 0.5056, 0.5432, 0.5213, 0.5671, 0.6217, 0.622, 0.622, 0.6063, 0.6343, 0.665, 0.7119, 0.7126, 0.7621, 0.7951, 0.871, 0.8398, 0.8402, 0.6585, 0.5625, 0.6718, 0.6159, 0.5437, 0.6666, 0.7037, 0.6648, 0.6669, 0.7414, 0.7814, 0.8773, 0.8041, 0.769, 0.6603, 0.6905, 0.6758, 0.5766, 0.5789, 0.4863, 0.4404, 0.6026, 0.5261, 0.5774, 0.6199, 0.6479, 0.7988, 0.7776, 0.7997, 0.7997, 0.9056, 0.9266, 0.9592, 0.9358, 0.9448, 0.9942, 1.0393, 1.0546, 1.0571, 1.1507, 1.1209, 1.1479, 1.125, 1.1581, 1.3355, 1.4198, 1.6492, 1.6712, 1.8132, 2.0444, 1.9597, 2.1073, 2.1045, 1.8685, 1.8565, 1.8061, 1.8516, 2.0792, 2.0435, 2.0435, 2.3206, 2.501, 2.4987, 2.6416, null, null, null]}
{"unit_id": "thread:1985682952202961060", "ticker_idx": 0, "ts": "2025-11-04T12:18:02+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the share-price gains of Samsung Electronics, SK Hynix, and others that have recently led the KOSPI rally will continue for some time", "tweet_type": "news_reaction", "timeline": "months", "summary": "Author shares BofA Korea head's bullish thesis: Samsung and SK Hynix share-price gains to continue on HBM supercycle, stable long-term contracts, and AI sovereignty demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "BofA Research Center Head: This memory supercycle is different from the past; Samsung and SK Hynix will go further\n\n“The current boom for semiconductor companies is different from past cycles. Customers are trying to secure high-quality memory on a stable basis even if it is expensive. The memory from Korean semiconductor companies, which used to be off-the-shelf, has turned into a luxury product.”\n\nSimon Woo, head of Korea Research at Bank of America (BofA) (photo), on the 4th forecast that the share-price gains of Samsung Electronics, SK Hynix, and others that have recently led the KOSPI rally will continue for some time. He said that the shortage of high-bandwidth memory (HBM) semiconductors for artificial intelligence (AI) has changed the standing of memory companies. Woo has analyzed the Korean market at domestic and overseas securities firms for more than 30 years. BofA recently raised its target price for SK Hynix to 700,000 won.\n\nAs reasons for anticipating further share-price increases, Woo pointed to the fact that “in the past, the profits of Korean semiconductor companies fluctuated with the cycle, but now they are securing stable results through long-term contracts.” He added, “Unlike before, they are not engaging in excessive capital spending to expand supply; instead, they are responding to the boom by raising memory prices, which is also a positive factor.”\n\nFrom October through the 4th of this month, Samsung Electronics’ share price has risen 25.03%, and SK Hynix’s 68.63%. Target prices have also continued to be raised, with Samsung Electronics at 170,000 won (SK Securities) and SK Hynix even appearing at 1,000,000 won (SK Securities).\n\nWoo assessed that the global AI industry is growing on the back of earnings, making it unlikely that a bubble will burst as it did during the dot-com crash.\n\nHe said, “Already, the profits of AI big-tech companies amount to as much as several tens of percent of sales, so the situation is different from the dot-com bubble, when investment rose without substance,” and emphasized that “so far the United States has accounted for most of the demand for AI semiconductors, but as countries invest to secure AI sovereignty, the results of Korean semiconductor firms will increase further.”\n\nWoo went on to stress, “If a company makes luxury chips, there is no need to be overly bound by the current valuation.” He added, “Memory chips play a key role in the U.S. AI ecosystem, and those chips do not come out of the United States, Taiwan, or China.”\n\n---\n\nhttps://t.co/D0c3X1mz2x", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.059103865090061714, "ret_signed_m1d": 0.059103865090061714, "alpha_spy_m1d": 0.04710813326124241, "alpha_c_m1d": 0.03198944821004046, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.027114416880021253, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04099142054336058, "ret_signed_p1d": -0.04099142054336058, "alpha_spy_p1d": -0.04445680974035682, "alpha_c_p1d": -0.04490374141301423, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.0039123208696536516, "ret_p1w": -0.013346063081999038, "ret_signed_p1w": -0.013346063081999038, "alpha_spy_p1w": -0.024838181255579572, "alpha_c_p1w": -0.01018218348890787, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.003163879593091168, "ret_p1m": -0.0038131501673385193, "ret_signed_p1m": -0.0038131501673385193, "alpha_spy_p1m": -0.016623390723764397, "alpha_c_p1m": 0.00962495731003643, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": -0.01343810747737495, "ret_p3m": 0.537466317061033, "ret_signed_p3m": 0.537466317061033, "alpha_spy_p3m": 0.5096619344222866, "alpha_c_p3m": 0.5570060584703556, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": -0.019539741409322575, "ret_p6m": 1.1693867492825456, "ret_signed_p6m": 1.1693867492825456, "alpha_spy_p6m": 1.1095684465441964, "alpha_c_p6m": 1.0837946967315362, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.08559205255100943, "price_path": [-0.3309, -0.3186, -0.3262, -0.3252, -0.3176, -0.3205, -0.3205, -0.3357, -0.3357, -0.3309, -0.33, -0.3224, -0.3214, -0.3328, -0.33, -0.3395, -0.3385, -0.3584, -0.3442, -0.3376, -0.3347, -0.3404, -0.3347, -0.3214, -0.311, -0.3034, -0.2844, -0.274, -0.2465, -0.2578, -0.2379, -0.2379, -0.2075, -0.1962, -0.1895, -0.1829, -0.2094, -0.1973, -0.2002, -0.1802, -0.1444, -0.1444, -0.1444, -0.1444, -0.1444, -0.1444, -0.1001, -0.1106, -0.1268, -0.0944, -0.0686, -0.0667, -0.0648, -0.0705, -0.0601, -0.0801, -0.0582, -0.0276, -0.0515, -0.0419, -0.0076, 0.0248, 0.0591, 0.0, -0.041, -0.0543, -0.0667, -0.041, -0.0133, -0.0172, -0.02, -0.0734, -0.041, -0.0677, -0.0801, -0.041, -0.0963, -0.0782, -0.0534, -0.02, -0.0133, -0.0419, -0.0391, -0.0143, -0.0038, 0.0019, 0.0334, 0.0439, 0.0334, 0.0296, 0.0229, 0.0381, -0.001, -0.02, 0.0286, 0.0257, 0.0133, 0.0534, 0.0629, 0.0591, 0.0591, 0.1153, 0.1447, 0.1485, 0.1485, 0.1485, 0.2309, 0.3229, 0.3306, 0.3507, 0.3296, 0.3315, 0.3296, 0.3181, 0.344, 0.3785, 0.4263, 0.4302, 0.3909, 0.4321, 0.4589, 0.457, 0.457, 0.5279, 0.5557, 0.5394, 0.5375, 0.4407, 0.6045, 0.6198, 0.526, 0.5193, 0.594, 0.5882, 0.6074, 0.7109, 0.7358, 0.7358, 0.7358, 0.7358, 0.8201, 0.821, 0.8488, 0.9158, 0.9494, 1.0883, 1.0739, 1.0739, 0.8689, 0.6495, 0.8354, 0.8028, 0.662, 0.7999, 0.8201, 0.7999, 0.7578, 0.8076, 0.8574, 0.9973, 0.9206, 0.9101, 0.7846, 0.8172, 0.8105, 0.7252, 0.7252, 0.6923, 0.605, 0.8205, 0.7125, 0.7873, 0.8536, 0.8862, 1.0206, 0.9582, 0.9774, 0.9294, 0.9822, 1.0254, 1.0878, 1.0734, 1.059, 1.1022, 1.0878, 1.155, 1.107, 1.155, 1.131, 1.1694]}
{"unit_id": "thread:1985682952202961060", "ticker_idx": 1, "ts": "2025-11-04T12:18:02+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BofA recently raised its target price for SK Hynix to 700,000 won", "tweet_type": "news_reaction", "timeline": "months", "summary": "Author shares BofA Korea head's bullish thesis: Samsung and SK Hynix share-price gains to continue on HBM supercycle, stable long-term contracts, and AI sovereignty demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "BofA Research Center Head: This memory supercycle is different from the past; Samsung and SK Hynix will go further\n\n“The current boom for semiconductor companies is different from past cycles. Customers are trying to secure high-quality memory on a stable basis even if it is expensive. The memory from Korean semiconductor companies, which used to be off-the-shelf, has turned into a luxury product.”\n\nSimon Woo, head of Korea Research at Bank of America (BofA) (photo), on the 4th forecast that the share-price gains of Samsung Electronics, SK Hynix, and others that have recently led the KOSPI rally will continue for some time. He said that the shortage of high-bandwidth memory (HBM) semiconductors for artificial intelligence (AI) has changed the standing of memory companies. Woo has analyzed the Korean market at domestic and overseas securities firms for more than 30 years. BofA recently raised its target price for SK Hynix to 700,000 won.\n\nAs reasons for anticipating further share-price increases, Woo pointed to the fact that “in the past, the profits of Korean semiconductor companies fluctuated with the cycle, but now they are securing stable results through long-term contracts.” He added, “Unlike before, they are not engaging in excessive capital spending to expand supply; instead, they are responding to the boom by raising memory prices, which is also a positive factor.”\n\nFrom October through the 4th of this month, Samsung Electronics’ share price has risen 25.03%, and SK Hynix’s 68.63%. Target prices have also continued to be raised, with Samsung Electronics at 170,000 won (SK Securities) and SK Hynix even appearing at 1,000,000 won (SK Securities).\n\nWoo assessed that the global AI industry is growing on the back of earnings, making it unlikely that a bubble will burst as it did during the dot-com crash.\n\nHe said, “Already, the profits of AI big-tech companies amount to as much as several tens of percent of sales, so the situation is different from the dot-com bubble, when investment rose without substance,” and emphasized that “so far the United States has accounted for most of the demand for AI semiconductors, but as countries invest to secure AI sovereignty, the results of Korean semiconductor firms will increase further.”\n\nWoo went on to stress, “If a company makes luxury chips, there is no need to be overly bound by the current valuation.” He added, “Memory chips play a key role in the U.S. AI ecosystem, and those chips do not come out of the United States, Taiwan, or China.”\n\n---\n\nhttps://t.co/D0c3X1mz2x", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.058020619072690405, "ret_signed_m1d": 0.058020619072690405, "alpha_spy_m1d": 0.046024887243871104, "alpha_c_m1d": 0.0202045574019174, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.037816061670773005, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01194539011893414, "ret_signed_p1d": -0.01194539011893414, "alpha_spy_p1d": -0.015410779315930379, "alpha_c_p1d": -0.031051843570454984, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.019106453451520844, "ret_p1w": 0.05631396671118849, "ret_signed_p1w": 0.05631396671118849, "alpha_spy_p1w": 0.044821848537607956, "alpha_c_p1w": 0.061161501081300584, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.004847534370112094, "ret_p1m": -0.05734578661241485, "ret_signed_p1m": -0.05734578661241485, "alpha_spy_p1m": -0.07015602716884073, "alpha_c_p1m": -0.0902576465313415, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": 0.032911859918926645, "ret_p3m": 0.5523055581490084, "ret_signed_p3m": 0.5523055581490084, "alpha_spy_p3m": 0.524501175510262, "alpha_c_p3m": 0.40502304113838017, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": 0.14728251701062822, "ret_p6m": 1.2121395194187587, "ret_signed_p6m": 1.2121395194187587, "alpha_spy_p6m": 1.1523212166804095, "alpha_c_p6m": 0.7915293063077491, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.42061021311100966, "price_path": [-0.5535, -0.5629, -0.545, -0.5416, -0.5263, -0.5288, -0.5288, -0.5442, -0.5518, -0.5646, -0.5825, -0.5723, -0.5578, -0.5544, -0.557, -0.5418, -0.541, -0.5631, -0.5555, -0.552, -0.5469, -0.5333, -0.5273, -0.5085, -0.4812, -0.4761, -0.4394, -0.4352, -0.4061, -0.4309, -0.3933, -0.3933, -0.401, -0.384, -0.3899, -0.3916, -0.4258, -0.4044, -0.407, -0.3857, -0.3251, -0.3251, -0.3251, -0.3251, -0.3251, -0.3251, -0.2696, -0.2918, -0.2978, -0.279, -0.2278, -0.2056, -0.1715, -0.1826, -0.1783, -0.1834, -0.1297, -0.087, -0.1109, -0.0478, -0.0307, -0.0461, 0.058, 0.0, -0.0119, 0.0119, -0.0102, 0.0341, 0.0563, 0.0529, 0.0444, -0.0444, 0.0341, -0.0273, -0.041, -0.0256, -0.1109, -0.1126, -0.1143, -0.1058, -0.071, -0.0949, -0.0813, -0.0471, -0.0573, -0.0744, -0.071, -0.0147, -0.0334, 0.0024, -0.0351, -0.0249, -0.0539, -0.0949, -0.0591, -0.0573, -0.0659, -0.0095, -0.0027, 0.0041, 0.0041, 0.0229, 0.0929, 0.1117, 0.1117, 0.1117, 0.1561, 0.1886, 0.2398, 0.2671, 0.291, 0.2705, 0.2791, 0.2603, 0.2671, 0.2791, 0.291, 0.3047, 0.2688, 0.2637, 0.2893, 0.3098, 0.2569, 0.3662, 0.4362, 0.4703, 0.5523, 0.4174, 0.5489, 0.5369, 0.4379, 0.4328, 0.5147, 0.496, 0.4686, 0.5164, 0.5028, 0.5028, 0.5028, 0.5028, 0.5267, 0.6206, 0.624, 0.7162, 0.7384, 0.8802, 0.8152, 0.8152, 0.6065, 0.4525, 0.6099, 0.5808, 0.4303, 0.6048, 0.6339, 0.5911, 0.5569, 0.6664, 0.6595, 0.8067, 0.7331, 0.7228, 0.5962, 0.6869, 0.7023, 0.5962, 0.5962, 0.4936, 0.3807, 0.5346, 0.42, 0.4987, 0.5158, 0.5671, 0.7673, 0.7074, 0.7571, 0.7793, 0.8871, 0.9435, 0.976, 0.9298, 0.9949, 1.0941, 1.0924, 1.0958, 1.0907, 1.2104, 1.2241, 1.2121]}
{"unit_id": "reply:1997651954290270599", "ticker_idx": 0, "ts": "2025-12-07T12:58:35+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think SK Hynix can reach 800,000 KRW", "tweet_type": "prediction", "timeline": "months", "summary": "SK Hynix to reach ₩800,000, surpassing all-time high of ₩646,000; bullish on continued undervaluation.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@JEONwh7859 I think SK Hynix can reach 800,000 KRW.", "reply_text": null, "parent_text": "@jukan05 2. SK Hynix Price Outlook\n​\"Do you think SK Hynix will break its all-time high(￦646,000) later this year or early next year? What is your estimate for the peak price in this current bull market? I’m asking because I believe the stock is still significantly undervalued.\"", "quoted_text": null, "parent_author": "JEONwh7859", "ret_m1d": -0.05719233447170469, "ret_signed_m1d": -0.05719233447170469, "alpha_spy_m1d": -0.06020566486713075, "alpha_c_m1d": -0.04598634336848628, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01906407531159593, "ret_signed_p1d": -0.01906407531159593, "alpha_spy_p1d": -0.018200930344634325, "alpha_c_p1d": -0.020285088195773793, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.03986126811277635, "ret_signed_p1w": -0.03986126811277635, "alpha_spy_p1w": -0.03561917054031205, "alpha_c_p1w": 0.002602403639090589, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.25823222667283874, "ret_signed_p1m": 0.25823222667283874, "alpha_spy_p1m": 0.24327646453594576, "alpha_c_p1m": 0.2032162932853212, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6338586339218024, "ret_signed_p3m": 0.6338586339218024, "alpha_spy_p3m": 0.6343074010647207, "alpha_c_p3m": 0.5578033195462861, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4735, -0.4683, -0.4311, -0.4268, -0.3973, -0.4224, -0.3843, -0.3843, -0.3921, -0.3748, -0.3809, -0.3826, -0.4172, -0.3956, -0.3982, -0.3765, -0.315, -0.315, -0.315, -0.315, -0.315, -0.315, -0.2588, -0.2813, -0.2873, -0.2683, -0.2163, -0.1938, -0.1592, -0.1704, -0.1661, -0.1713, -0.1168, -0.0735, -0.0977, -0.0336, -0.0163, -0.0319, 0.0738, 0.0149, 0.0027, 0.027, 0.0045, 0.0495, 0.072, 0.0686, 0.0599, -0.0302, 0.0495, -0.0128, -0.0267, -0.0111, -0.0977, -0.0994, -0.1012, -0.0925, -0.0572, -0.0815, -0.0676, -0.0329, -0.0433, -0.0607, -0.0572, 0.0, -0.0191, 0.0173, -0.0208, -0.0104, -0.0399, -0.0815, -0.0451, -0.0433, -0.052, 0.0052, 0.0121, 0.0191, 0.0191, 0.0381, 0.1092, 0.1282, 0.1282, 0.1282, 0.1733, 0.2062, 0.2582, 0.286, 0.3102, 0.2894, 0.2981, 0.279, 0.286, 0.2981, 0.3102, 0.3241, 0.2877, 0.2825, 0.3085, 0.3293, 0.2756, 0.3865, 0.4575, 0.4922, 0.5754, 0.4385, 0.5719, 0.5598, 0.4593, 0.4541, 0.5373, 0.5182, 0.4905, 0.539, 0.5251, 0.5251, 0.5251, 0.5251, 0.5494, 0.6447, 0.6482, 0.7418, 0.7643, 0.9082, 0.8422, 0.8422, 0.6304, 0.4741, 0.6339, 0.6043, 0.4515, 0.6286, 0.6582, 0.6148, 0.58, 0.6912, 0.6842, 0.8335, 0.7589, 0.7485, 0.62, 0.712, 0.7276, 0.62, 0.62, 0.5158, 0.4012, 0.5575, 0.4411, 0.521, 0.5384, 0.5905, 0.7936, 0.7328, 0.7832, 0.8058, 0.9151, 0.9724, 1.0054, 0.9585, 1.0245, 1.1252, 1.1235, 1.127, 1.1218, 1.2433, 1.2572, 1.245, 1.2329, 1.2329, 1.5124, 1.5124, 1.7798, 1.8718, 1.9274, 2.2642, 2.1861, 2.4309, 2.4205, 2.1583, 2.1948, 2.0298, 2.0298, 2.3684, 2.3702, 2.3702, 2.5629, 2.8945, 2.9751, 3.0515, null, null]}
{"unit_id": "thread:1991067128938836430", "ticker_idx": 0, "ts": "2025-11-19T08:52:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "earnings forecasts for Samsung Electronics and SK Hynix are being revised upward... Samsung Electronics' annual operating profit for next year has increased by about 28%", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND supply crunch driving prices sharply higher; Samsung, SK Hynix, and Kioxia earnings estimates revised up ~28% on surging AI-driven NAND demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Competition to secure NAND flash heats up after DRAM… “Negotiations for 2027 supply also begin”\n\nAs growth in the artificial intelligence (AI) industry leads to a shortage in memory semiconductor supply centered on DRAM, competition to secure NAND flash is also reportedly intensifying as demand surges. With production capacity at memory semiconductor companies concentrated on high-bandwidth memory (HBM) and commodity DRAM, NAND supply is expected to be constrained, driving NAND prices sharply higher. As next year’s NAND volumes are effectively sold out, some big tech customers are said to have already begun negotiations for supply in 2027.\n\nAccording to industry sources on the 19th, memory semiconductor companies such as Samsung Electronics, SK Hynix, and Kioxia are in talks with some big tech companies over supply volumes for 2027. With NAND supply limited and next year’s volumes effectively sold out, pushing prices higher, companies are taking the unusual step of preemptively negotiating NAND supply volumes two years in advance. A semiconductor industry official said, “Demand for NAND is increasing, centered on eSSDs installed in AI data centers, but it is true that production capacity is limited,” adding, “Some big tech companies that are expanding their data centers have already begun supply negotiations to secure volumes.”\n\nNAND demand is rising in step with the growth of the AI market. Kim Rok-ho, an analyst at Hana Securities, analyzed, “NAND demand next year is expected to increase by 18% year-on-year, but demand will exceed supply by 5%.” Kioxia, the world’s third-largest NAND company, stated in its earnings announcement that the NAND market will grow at an average annual rate of 20% from this year through 2029, and that demand is increasing mainly in the AI sector. The company also projected that AI-related products will account for about half of NAND demand by 2029.\n\nDespite rising demand, most NAND manufacturers are focusing on maintenance and process migration to advanced nodes rather than aggressive investment, leading to a steep upward trend in prices. New production lines being added by Samsung Electronics and SK Hynix, the world’s No. 1 and No. 2 NAND players, are centered on HBM and commodity DRAM. For NAND, instead of additional capacity expansion, investments are mainly going into converting to advanced processes, leaving supply unable to keep up with demand. Last month, the average fixed transaction price of general-purpose NAND flash products for memory cards and USBs (128Gb 16Gx8 MLC) was 4.35 dollars, up 14.9% from the previous month. This is the highest level of the year and marks ten consecutive months of increases since January.\n\nFor these reasons, competition to secure NAND volumes is said to be becoming even fiercer. A semiconductor industry official said, “Some suppliers have completed next year’s contracts and are changing the unit of price and volume contracts from one quarter to two–three quarters. Many have also begun negotiations for 2027 supply volumes,” adding, “Compared to solid demand, NAND companies’ stance on capacity expansion remains conservative, and as a result, supply continues to lag demand, which in turn is expected to keep the upward trend in prices intact.”\n\nWith profitability expected to improve significantly not only in DRAM but also in the NAND business, earnings forecasts for Samsung Electronics and SK Hynix are being revised upward. According to financial information provider FnGuide, the forecast for Samsung Electronics’ annual operating profit for next year has increased by about 28% over the past month, from 59.6927 trillion won to 76.5275 trillion won. Over the same period, SK Hynix’s operating profit forecast was also revised up by about 28%, from 55.0499 trillion won to 70.5481 trillion won.\n\n---\n\nhttps://t.co/HCQhI01eRe", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013471504026785652, "ret_signed_m1d": 0.013471504026785652, "alpha_spy_m1d": 0.017319836931528054, "alpha_c_m1d": 0.020376174805940117, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.006904670779154465, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04248700726575749, "ret_signed_p1d": 0.04248700726575749, "alpha_spy_p1d": 0.057729343948522915, "alpha_c_p1d": 0.07387829053141459, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.0313912832656571, "ret_p1w": 0.06528494342804714, "ret_signed_p1w": 0.06528494342804714, "alpha_spy_p1w": 0.03955420204476301, "alpha_c_p1w": 0.05528392616130562, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.010001017266741519, "ret_p1m": 0.11502580612342728, "ret_signed_p1m": 0.11502580612342728, "alpha_spy_p1m": 0.09413950104918234, "alpha_c_p1m": 0.10751622105280778, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.0075095850706194955, "ret_p3m": 0.886847942129589, "ret_signed_p3m": 0.886847942129589, "alpha_spy_p3m": 0.8549531798567576, "alpha_c_p3m": 0.8919255626315872, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": -0.005077620501998226, "ret_p6m": 2.0886479795637936, "ret_signed_p6m": 2.0886479795637936, "alpha_spy_p6m": 1.9531274305425002, "alpha_c_p6m": 1.8074024791175685, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.2812455004462251, "price_path": [-0.2634, -0.2624, -0.2747, -0.2716, -0.282, -0.2809, -0.3026, -0.2871, -0.2799, -0.2768, -0.283, -0.2768, -0.2624, -0.251, -0.2428, -0.2221, -0.2108, -0.1809, -0.1932, -0.1716, -0.1716, -0.1386, -0.1262, -0.119, -0.1117, -0.1406, -0.1275, -0.1306, -0.1088, -0.0699, -0.0699, -0.0699, -0.0699, -0.0699, -0.0699, -0.0218, -0.0332, -0.0508, -0.0155, 0.0124, 0.0145, 0.0166, 0.0104, 0.0218, 0.0, 0.0238, 0.057, 0.0311, 0.0415, 0.0788, 0.114, 0.1513, 0.087, 0.0425, 0.028, 0.0145, 0.0425, 0.0725, 0.0684, 0.0653, 0.0073, 0.0425, 0.0135, 0.0, 0.0425, -0.0176, 0.0021, 0.029, 0.0653, 0.0725, 0.0415, 0.0446, 0.0715, 0.0829, 0.0891, 0.1233, 0.1347, 0.1233, 0.1192, 0.1119, 0.1285, 0.086, 0.0653, 0.1181, 0.115, 0.1016, 0.1451, 0.1554, 0.1513, 0.1513, 0.2124, 0.2444, 0.2485, 0.2485, 0.2485, 0.3381, 0.438, 0.4464, 0.4682, 0.4453, 0.4474, 0.4453, 0.4328, 0.461, 0.4984, 0.5505, 0.5547, 0.512, 0.5568, 0.5859, 0.5838, 0.5838, 0.6609, 0.6911, 0.6734, 0.6713, 0.5661, 0.7442, 0.7608, 0.6588, 0.6515, 0.7327, 0.7265, 0.7473, 0.8598, 0.8868, 0.8868, 0.8868, 0.8868, 0.9785, 0.9795, 1.0097, 1.0826, 1.1191, 1.27, 1.2544, 1.2544, 1.0316, 0.7931, 0.9951, 0.9597, 0.8067, 0.9566, 0.9785, 0.9566, 0.9108, 0.9649, 1.0191, 1.1711, 1.0878, 1.0764, 0.94, 0.9754, 0.9681, 0.8754, 0.8754, 0.8396, 0.7447, 0.9789, 0.8615, 0.9429, 1.0149, 1.0504, 1.1965, 1.1287, 1.1495, 1.0974, 1.1547, 1.2017, 1.2695, 1.2539, 1.2382, 1.2852, 1.2695, 1.3426, 1.2904, 1.3426, 1.3165, 1.3582, 1.3008, 1.3008, 1.426, 1.426, 1.7756, 1.833, 1.8017, 1.9791, 1.9113, 1.9634, 2.0886]}
{"unit_id": "thread:1991067128938836430", "ticker_idx": 1, "ts": "2025-11-19T08:52:50+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's operating profit forecast was also revised up by about 28%, from 55.0499 trillion won to 70.5481 trillion won", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND supply crunch driving prices sharply higher; Samsung, SK Hynix, and Kioxia earnings estimates revised up ~28% on surging AI-driven NAND demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Competition to secure NAND flash heats up after DRAM… “Negotiations for 2027 supply also begin”\n\nAs growth in the artificial intelligence (AI) industry leads to a shortage in memory semiconductor supply centered on DRAM, competition to secure NAND flash is also reportedly intensifying as demand surges. With production capacity at memory semiconductor companies concentrated on high-bandwidth memory (HBM) and commodity DRAM, NAND supply is expected to be constrained, driving NAND prices sharply higher. As next year’s NAND volumes are effectively sold out, some big tech customers are said to have already begun negotiations for supply in 2027.\n\nAccording to industry sources on the 19th, memory semiconductor companies such as Samsung Electronics, SK Hynix, and Kioxia are in talks with some big tech companies over supply volumes for 2027. With NAND supply limited and next year’s volumes effectively sold out, pushing prices higher, companies are taking the unusual step of preemptively negotiating NAND supply volumes two years in advance. A semiconductor industry official said, “Demand for NAND is increasing, centered on eSSDs installed in AI data centers, but it is true that production capacity is limited,” adding, “Some big tech companies that are expanding their data centers have already begun supply negotiations to secure volumes.”\n\nNAND demand is rising in step with the growth of the AI market. Kim Rok-ho, an analyst at Hana Securities, analyzed, “NAND demand next year is expected to increase by 18% year-on-year, but demand will exceed supply by 5%.” Kioxia, the world’s third-largest NAND company, stated in its earnings announcement that the NAND market will grow at an average annual rate of 20% from this year through 2029, and that demand is increasing mainly in the AI sector. The company also projected that AI-related products will account for about half of NAND demand by 2029.\n\nDespite rising demand, most NAND manufacturers are focusing on maintenance and process migration to advanced nodes rather than aggressive investment, leading to a steep upward trend in prices. New production lines being added by Samsung Electronics and SK Hynix, the world’s No. 1 and No. 2 NAND players, are centered on HBM and commodity DRAM. For NAND, instead of additional capacity expansion, investments are mainly going into converting to advanced processes, leaving supply unable to keep up with demand. Last month, the average fixed transaction price of general-purpose NAND flash products for memory cards and USBs (128Gb 16Gx8 MLC) was 4.35 dollars, up 14.9% from the previous month. This is the highest level of the year and marks ten consecutive months of increases since January.\n\nFor these reasons, competition to secure NAND volumes is said to be becoming even fiercer. A semiconductor industry official said, “Some suppliers have completed next year’s contracts and are changing the unit of price and volume contracts from one quarter to two–three quarters. Many have also begun negotiations for 2027 supply volumes,” adding, “Compared to solid demand, NAND companies’ stance on capacity expansion remains conservative, and as a result, supply continues to lag demand, which in turn is expected to keep the upward trend in prices intact.”\n\nWith profitability expected to improve significantly not only in DRAM but also in the NAND business, earnings forecasts for Samsung Electronics and SK Hynix are being revised upward. According to financial information provider FnGuide, the forecast for Samsung Electronics’ annual operating profit for next year has increased by about 28% over the past month, from 59.6927 trillion won to 76.5275 trillion won. Over the same period, SK Hynix’s operating profit forecast was also revised up by about 28%, from 55.0499 trillion won to 70.5481 trillion won.\n\n---\n\nhttps://t.co/HCQhI01eRe", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014234824251757594, "ret_signed_m1d": 0.014234824251757594, "alpha_spy_m1d": 0.018083157156499996, "alpha_c_m1d": 0.03235289282657916, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.018118068574821566, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016014246979235702, "ret_signed_p1d": 0.016014246979235702, "alpha_spy_p1d": 0.031256583662001125, "alpha_c_p1d": 0.05826033987670176, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.04224609289746606, "ret_p1w": -0.0676156103446719, "ret_signed_p1w": -0.0676156103446719, "alpha_spy_p1w": -0.09334635172795602, "alpha_c_p1w": -0.0919791989592329, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.024363588614561005, "ret_p1m": -0.0170901303203842, "ret_signed_p1m": -0.0170901303203842, "alpha_spy_p1m": -0.03797643539462914, "alpha_c_p1m": -0.04009859295855278, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.02300846263816858, "ret_p3m": 0.56695773578299, "ret_signed_p3m": 0.56695773578299, "alpha_spy_p3m": 0.5350629735101586, "alpha_c_p3m": 0.3620654117518609, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": 0.2048923240311291, "ret_p6m": 2.5143214145724873, "ret_signed_p6m": 2.5143214145724873, "alpha_spy_p6m": 2.378800865551194, "alpha_c_p6m": 1.8052135539038958, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.7091078606685914, "price_path": [-0.554, -0.5389, -0.5354, -0.538, -0.5222, -0.5214, -0.5445, -0.5365, -0.5329, -0.5276, -0.5133, -0.5071, -0.4875, -0.4591, -0.4537, -0.4155, -0.411, -0.3808, -0.4066, -0.3674, -0.3674, -0.3754, -0.3577, -0.3639, -0.3657, -0.4012, -0.379, -0.3817, -0.3594, -0.2963, -0.2963, -0.2963, -0.2963, -0.2963, -0.2963, -0.2384, -0.2616, -0.2678, -0.2482, -0.1948, -0.1717, -0.1361, -0.1477, -0.1432, -0.1486, -0.0925, -0.048, -0.073, -0.0071, 0.0107, -0.0053, 0.1032, 0.0427, 0.0302, 0.0552, 0.032, 0.0783, 0.1014, 0.0979, 0.089, -0.0036, 0.0783, 0.0142, 0.0, 0.016, -0.073, -0.0747, -0.0765, -0.0676, -0.0313, -0.0563, -0.042, -0.0064, -0.0171, -0.0349, -0.0313, 0.0274, 0.0078, 0.0452, 0.0061, 0.0167, -0.0135, -0.0563, -0.0189, -0.0171, -0.026, 0.0328, 0.0399, 0.047, 0.047, 0.0666, 0.1396, 0.1592, 0.1592, 0.1592, 0.2055, 0.2393, 0.2927, 0.3212, 0.3462, 0.3248, 0.3337, 0.3141, 0.3212, 0.3337, 0.3462, 0.3604, 0.323, 0.3177, 0.3444, 0.3657, 0.3105, 0.4245, 0.4975, 0.5331, 0.6186, 0.4779, 0.615, 0.6026, 0.4993, 0.494, 0.5794, 0.5598, 0.5313, 0.5812, 0.567, 0.567, 0.567, 0.567, 0.5919, 0.6898, 0.6934, 0.7895, 0.8127, 0.9605, 0.8927, 0.8927, 0.6751, 0.5145, 0.6787, 0.6483, 0.4914, 0.6733, 0.7036, 0.659, 0.6234, 0.7375, 0.7304, 0.8838, 0.8071, 0.7964, 0.6644, 0.7589, 0.775, 0.6644, 0.6644, 0.5574, 0.4396, 0.6002, 0.4807, 0.5627, 0.5806, 0.6341, 0.8428, 0.7804, 0.8321, 0.8553, 0.9677, 1.0265, 1.0604, 1.0123, 1.0801, 1.1835, 1.1817, 1.1853, 1.1799, 1.3048, 1.3191, 1.3066, 1.2941, 1.2941, 1.5813, 1.5813, 1.8561, 1.9506, 2.0077, 2.3538, 2.2735, 2.525, 2.5143]}
{"unit_id": "thread:1991067128938836430", "ticker_idx": 2, "ts": "2025-11-19T08:52:50+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Kioxia, the world's third-largest NAND company, stated in its earnings announcement that the NAND market will grow at an average annual rate of 20% from this year through 2029", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND supply crunch driving prices sharply higher; Samsung, SK Hynix, and Kioxia earnings estimates revised up ~28% on surging AI-driven NAND demand.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Competition to secure NAND flash heats up after DRAM… “Negotiations for 2027 supply also begin”\n\nAs growth in the artificial intelligence (AI) industry leads to a shortage in memory semiconductor supply centered on DRAM, competition to secure NAND flash is also reportedly intensifying as demand surges. With production capacity at memory semiconductor companies concentrated on high-bandwidth memory (HBM) and commodity DRAM, NAND supply is expected to be constrained, driving NAND prices sharply higher. As next year’s NAND volumes are effectively sold out, some big tech customers are said to have already begun negotiations for supply in 2027.\n\nAccording to industry sources on the 19th, memory semiconductor companies such as Samsung Electronics, SK Hynix, and Kioxia are in talks with some big tech companies over supply volumes for 2027. With NAND supply limited and next year’s volumes effectively sold out, pushing prices higher, companies are taking the unusual step of preemptively negotiating NAND supply volumes two years in advance. A semiconductor industry official said, “Demand for NAND is increasing, centered on eSSDs installed in AI data centers, but it is true that production capacity is limited,” adding, “Some big tech companies that are expanding their data centers have already begun supply negotiations to secure volumes.”\n\nNAND demand is rising in step with the growth of the AI market. Kim Rok-ho, an analyst at Hana Securities, analyzed, “NAND demand next year is expected to increase by 18% year-on-year, but demand will exceed supply by 5%.” Kioxia, the world’s third-largest NAND company, stated in its earnings announcement that the NAND market will grow at an average annual rate of 20% from this year through 2029, and that demand is increasing mainly in the AI sector. The company also projected that AI-related products will account for about half of NAND demand by 2029.\n\nDespite rising demand, most NAND manufacturers are focusing on maintenance and process migration to advanced nodes rather than aggressive investment, leading to a steep upward trend in prices. New production lines being added by Samsung Electronics and SK Hynix, the world’s No. 1 and No. 2 NAND players, are centered on HBM and commodity DRAM. For NAND, instead of additional capacity expansion, investments are mainly going into converting to advanced processes, leaving supply unable to keep up with demand. Last month, the average fixed transaction price of general-purpose NAND flash products for memory cards and USBs (128Gb 16Gx8 MLC) was 4.35 dollars, up 14.9% from the previous month. This is the highest level of the year and marks ten consecutive months of increases since January.\n\nFor these reasons, competition to secure NAND volumes is said to be becoming even fiercer. A semiconductor industry official said, “Some suppliers have completed next year’s contracts and are changing the unit of price and volume contracts from one quarter to two–three quarters. Many have also begun negotiations for 2027 supply volumes,” adding, “Compared to solid demand, NAND companies’ stance on capacity expansion remains conservative, and as a result, supply continues to lag demand, which in turn is expected to keep the upward trend in prices intact.”\n\nWith profitability expected to improve significantly not only in DRAM but also in the NAND business, earnings forecasts for Samsung Electronics and SK Hynix are being revised upward. According to financial information provider FnGuide, the forecast for Samsung Electronics’ annual operating profit for next year has increased by about 28% over the past month, from 59.6927 trillion won to 76.5275 trillion won. Over the same period, SK Hynix’s operating profit forecast was also revised up by about 28%, from 55.0499 trillion won to 70.5481 trillion won.\n\n---\n\nhttps://t.co/HCQhI01eRe", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.053068758652515036, "ret_signed_m1d": -0.053068758652515036, "alpha_spy_m1d": -0.049220425747772634, "alpha_c_m1d": -0.03495069007769347, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.018118068574821566, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.046146746654360804, "ret_signed_p1d": 0.046146746654360804, "alpha_spy_p1d": 0.06138908333712623, "alpha_c_p1d": 0.08839283955182686, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.04224609289746606, "ret_p1w": -0.22602676511305952, "ret_signed_p1w": -0.22602676511305952, "alpha_spy_p1w": -0.25175750649634365, "alpha_c_p1w": -0.25039035372762053, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.024363588614561005, "ret_p1m": -0.12228887863405635, "ret_signed_p1m": -0.12228887863405635, "alpha_spy_p1m": -0.1431751837083013, "alpha_c_p1m": -0.14529734127222493, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.02300846263816858, "ret_p3m": 1.0581449007844945, "ret_signed_p3m": 1.0581449007844945, "alpha_spy_p3m": 1.0262501385116631, "alpha_c_p3m": 0.8532525767533654, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": 0.2048923240311291, "ret_p6m": 3.472542685740655, "ret_signed_p6m": 3.472542685740655, "alpha_spy_p6m": 3.3370221367193618, "alpha_c_p6m": 2.7634348250720637, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.7091078606685914, "price_path": [-0.7782, -0.7792, -0.7754, -0.7771, -0.7719, -0.7604, -0.753, -0.7586, -0.7579, -0.7576, -0.7171, -0.7093, -0.707, -0.6784, -0.6304, -0.5902, -0.5902, -0.5658, -0.5902, -0.5828, -0.5819, -0.5551, -0.5551, -0.5431, -0.5487, -0.5944, -0.5685, -0.5501, -0.5658, -0.5007, -0.4296, -0.4167, -0.4555, -0.4573, -0.4259, -0.4315, -0.4315, -0.4462, -0.4167, -0.365, -0.3946, -0.3401, -0.3503, -0.3253, -0.3198, -0.1897, -0.0946, -0.1343, -0.0697, 0.006, -0.0009, -0.0009, -0.0069, -0.0268, 0.0614, 0.1112, 0.2293, 0.2178, 0.222, 0.2021, -0.0748, 0.0305, -0.0531, 0.0, 0.0461, -0.0743, -0.0743, -0.0906, -0.226, -0.1648, -0.1319, -0.1802, -0.1493, -0.1683, -0.1629, -0.1297, -0.0632, -0.0887, -0.126, -0.1131, -0.0872, -0.1509, -0.1982, -0.1417, -0.1223, -0.138, -0.0655, -0.0815, -0.0212, -0.0037, 0.0535, -0.0148, -0.0369, -0.0369, -0.0369, -0.0369, 0.0475, 0.0706, 0.1712, 0.1998, 0.1712, 0.1712, 0.263, 0.2257, 0.258, 0.3613, 0.4038, 0.4033, 0.5228, 0.653, 0.5999, 0.6105, 0.7028, 0.7485, 0.7762, 0.9714, 0.6945, 0.9183, 0.9289, 0.8025, 0.7623, 0.7443, 0.7393, 0.7393, 0.9543, 1.1084, 1.0581, 1.0646, 0.9769, 0.9612, 0.8976, 0.8976, 1.0554, 0.9811, 0.9603, 0.9575, 0.9857, 0.8643, 0.7762, 0.8736, 0.8436, 0.6641, 0.8062, 0.9732, 0.9608, 0.9442, 1.0872, 0.9945, 1.1587, 1.0637, 1.0637, 0.9811, 0.9474, 1.0715, 0.9534, 0.8708, 0.8477, 0.761, 1.012, 0.9469, 1.0166, 1.1043, 1.1477, 1.5473, 1.5565, 1.7817, 1.886, 2.2303, 1.9912, 2.126, 1.8177, 1.8159, 2.0217, 2.2118, 2.2653, 2.1915, 2.2737, 2.3521, 2.3521, 2.4665, 2.3604, 2.3604, 2.3604, 2.3604, 3.0065, 3.1061, 3.24, 3.2547, 3.6608, 3.4725]}
{"unit_id": "quote:1989479324337250395", "ticker_idx": 0, "ts": "2025-11-14T23:43:28+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA: Full-Scale Ramp of Blackwell, Demand Accelerating, Outlook Raised, Target Price Increased", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley bullish note on NVDA: Blackwell ramp, accelerating demand, raised target.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Morgan Stanley vs. Street Consensus on NVIDIA https://t.co/Lnue1GSNhk", "reply_text": null, "parent_text": null, "quoted_text": "NVIDIA: Full-Scale Ramp of Blackwell, Demand Accelerating, Outlook Raised, Target Price Increased\nMorgan Stanley (J. Moore, 25/11/14)\n\nMorgan Stanley stated that over the past 45 days, industry conditions have clearly inflected to the upside, Blackwell cabinets have entered a https://t.co/gjURAPLWM5", "parent_author": null, "ret_m1d": -0.017405531691603637, "ret_signed_m1d": -0.017405531691603637, "alpha_spy_m1d": -0.01756914455597791, "alpha_c_m1d": -0.016941702719995955, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018772690180637097, "ret_signed_p1d": -0.018772690180637097, "alpha_spy_p1d": -0.009456166249674935, "alpha_c_p1d": -0.0052065579481174185, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.05936791649684514, "ret_signed_p1w": -0.05936791649684514, "alpha_spy_p1w": -0.040169486157191736, "alpha_c_p1w": -0.004727053404036785, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.07293574274855108, "ret_signed_p1m": -0.07293574274855108, "alpha_spy_p1m": -0.08603226282807919, "alpha_c_p1m": -0.09589354646912152, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.0005753669288237617, "ret_signed_p3m": -0.0005753669288237617, "alpha_spy_p3m": -0.03342802619055418, "alpha_c_p3m": -0.20664702748446007, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.15404161563996577, "ret_signed_p6m": 0.15404161563996577, "alpha_spy_p6m": 0.0475133801339096, "alpha_c_p6m": -0.5217169980731151, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.0765, -0.0777, -0.0799, -0.0641, -0.0545, -0.0442, -0.0451, -0.0526, -0.0841, -0.0841, -0.102, -0.1029, -0.0974, -0.1218, -0.115, -0.1021, -0.0676, -0.0684, -0.0649, -0.0653, -0.0804, -0.1045, -0.0733, -0.071, -0.0345, -0.0617, -0.0694, -0.0656, -0.063, -0.0438, -0.0189, -0.0154, -0.0067, -0.0134, -0.0243, -0.027, -0.0056, 0.0126, -0.0369, -0.0097, -0.0533, -0.0544, -0.044, -0.0365, -0.0396, -0.0474, -0.052, -0.0421, -0.0206, 0.0069, 0.0571, 0.0887, 0.0669, 0.0648, 0.0879, 0.0448, 0.0265, -0.011, -0.0106, 0.0467, 0.0157, 0.0191, -0.0174, 0.0, -0.0188, -0.0463, -0.0192, -0.0501, -0.0594, -0.0401, -0.0649, -0.0521, -0.0521, -0.0693, -0.0539, -0.0458, -0.0556, -0.0357, -0.0408, -0.0242, -0.0273, -0.0335, -0.0485, -0.0796, -0.0729, -0.0654, -0.1011, -0.0842, -0.0482, -0.034, -0.005, -0.0081, -0.0081, 0.0019, -0.0102, -0.0138, -0.0192, -0.0192, -0.0069, -0.0107, -0.0154, -0.0055, -0.0269, -0.0279, -0.0274, -0.0229, -0.0369, -0.0164, -0.0207, -0.0207, -0.0636, -0.036, -0.028, -0.0131, -0.0194, -0.0086, 0.0072, 0.0124, 0.0051, -0.0239, -0.0516, -0.084, -0.0961, -0.025, -0.0006, -0.0085, -0.0006, -0.0169, -0.0386, -0.0386, -0.0273, -0.0115, -0.0119, -0.0018, 0.0073, 0.0141, 0.0284, -0.0277, -0.0682, -0.0404, -0.0532, -0.0374, -0.0359, -0.0649, -0.0395, -0.0283, -0.0217, -0.0369, -0.0521, -0.0364, -0.0432, -0.0513, -0.0609, -0.0918, -0.0763, -0.0786, -0.0603, -0.0994, -0.119, -0.1314, -0.0828, -0.0757, -0.0671, -0.0671, -0.0658, -0.0634, -0.0424, -0.0328, -0.008, -0.0044, 0.0335, 0.0459, 0.0431, 0.0606, 0.0626, 0.0512, 0.065, 0.0499, 0.0953, 0.1392, 0.1211, 0.1005, 0.0495, 0.0437, 0.0438, 0.0334, 0.093, 0.1123, 0.1317, 0.154]}
{"unit_id": "orig:1984167171405312282", "ticker_idx": 0, "ts": "2025-10-31T07:54:52+00:00", "symbol": "LITE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Lumentum's EML (External Modulated Laser) demand is rising rapidly, with price increases expected in 2026. Its OCS (Optical Communication Systems) business shipments are projected at 15 thousand / 30 thousand units in 2025 / 2026, representing about 30 % market share in 2026.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Lumentum, Fabrinet, and Chinese optical-module makers on surging CoWoS/TPU-driven 1.6T optical-module demand through 2025-2026.", "resolved_tickers": ["LITE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK: CoWoS and Optical Module Demand in 2025/2026\n\nNVIDIA’s CoWoS demand in 2025/2026 is projected to reach 377,000 / 630,000 wafers, respectively, of which 50,000 wafers will be for the Vera CPU, with the remainder driven by Rubin’s mid-2026 packaging shipments. From a CoWoS capacity perspective, Blackwell/Rubin-related product shipments in 2026 are estimated at 5 million / 2 million units, corresponding to a 1.6 T optical-module demand of 20 million units.\n\nFor Google, due to its own cluster upgrades and the announcement of 1 million TPU demand from companies such as Meta and HPE, its CoWoS demand may rise to 220–230 thousand wafers. TPU shipments are expected to reach 2.7 million / 4 million units in 2025/2026, driving Google’s 800 G / 1.6 T optical-module demand to 6 million / 10 million units, respectively. Consequently, Google’s 1.6 T optical-module forecast will also be revised upward.\n\nThe NVIDIA Rubin GPU is equipped with two CX9 NIC (Network Interface Card) chips, doubling bandwidth compared with Blackwell.\nRubin Ultra adopts four CX9 chips per GPU, further doubling bandwidth.\n\nGoogle plans to complete its large-scale cluster migration in 2026. In the vertical-expansion layer, it will deploy optical interconnects, where the TPU-to-1.6 T optical-module ratio is about 1 : 1.5; in the horizontal-expansion layer, about 1 : 2.5; overall, the TPU-to-optical-module ratio will be roughly 1 : 4 (1.6 T basis).\n\nIn Meta’s Minera architecture, each ASIC chip is paired with two NICs, and the ASIC-to-optical-module ratio will expand as bandwidth increases.\n\nLumentum’s EML (External Modulated Laser) demand is rising rapidly, with price increases expected in 2026.\nIts OCS (Optical Communication Systems) business shipments are projected at 15 thousand / 30 thousand units in 2025 / 2026, representing about 30 % market share in 2026.\n\nIn the CPO/OIO (Co-Packaged Optics / On-Interposer Optics) field, NVIDIA’s CPO switch shipments are expected to reach 2 k / 20 k / 35 k units in 2025 / 2026 / 2027, respectively.\nStarting in 2027, NVIDIA and ASIC vendors will launch OIO solutions, driving incremental demand for CW (Continuous-Wave) lasers, FAUs, and optical engines.\n\nBeneficiaries include Lumentum, FN, and leading Chinese optical-module manufacturers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007094625387488174, "ret_signed_m1d": -0.007094625387488174, "alpha_spy_m1d": -0.003825069836452366, "alpha_c_m1d": -0.003825069836452366, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": -0.0032695555510358076, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009823356576314057, "ret_signed_p1d": -0.009823356576314057, "alpha_spy_p1d": -0.01170004366427857, "alpha_c_p1d": -0.01170004366427857, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.0018766870879645126, "ret_p1w": 0.19125820360536205, "ret_signed_p1w": 0.19125820360536205, "alpha_spy_p1w": 0.20751787525306864, "alpha_c_p1w": 0.20751787525306864, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.016259671647706586, "ret_p1m": 0.5773466785410066, "ret_signed_p1m": 0.5773466785410066, "alpha_spy_p1m": 0.5799710525961609, "alpha_c_p1m": 0.5799710525961609, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.002624374055154255, "ret_p3m": 0.9101012336938534, "ret_signed_p3m": 0.9101012336938534, "alpha_spy_p3m": 0.8875008021809176, "alpha_c_p3m": 0.8875008021809176, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.02260043151293578, "ret_p6m": 3.2651319859530714, "ret_signed_p6m": 3.2651319859530714, "alpha_spy_p6m": 3.210617524401983, "alpha_c_p6m": 3.210617524401983, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.05451446155108863, "price_path": [-0.4634, -0.4542, -0.4487, -0.4231, -0.4293, -0.4063, -0.4035, -0.4313, -0.4252, -0.4097, -0.4148, -0.425, -0.4174, -0.4079, -0.3877, -0.3817, -0.3757, -0.3275, -0.3411, -0.3411, -0.3424, -0.3346, -0.2959, -0.2585, -0.2588, -0.2472, -0.182, -0.182, -0.1912, -0.1627, -0.1521, -0.1896, -0.1491, -0.1629, -0.1828, -0.1864, -0.2422, -0.2106, -0.2025, -0.1934, -0.1927, -0.1496, -0.1578, -0.1873, -0.2032, -0.2203, -0.1909, -0.2074, -0.2577, -0.2034, -0.2222, -0.2232, -0.1902, -0.1825, -0.2012, -0.1916, -0.2158, -0.164, -0.1104, -0.0385, -0.0157, 0.0631, -0.0071, 0.0, -0.0098, -0.0655, 0.1547, 0.1891, 0.1913, 0.2894, 0.2526, 0.2592, 0.1255, 0.1518, 0.201, 0.2276, 0.3342, 0.1572, 0.2681, 0.4852, 0.4451, 0.5295, 0.5295, 0.6132, 0.5773, 0.5023, 0.5032, 0.6266, 0.6442, 0.6995, 0.7877, 0.8158, 0.8461, 0.6092, 0.6605, 0.5685, 0.5889, 0.6726, 0.8428, 0.9343, 0.9221, 0.9643, 0.9643, 0.9387, 0.8486, 0.8415, 0.8287, 0.8287, 0.9156, 0.7714, 0.9717, 0.9492, 0.7278, 0.7435, 0.6862, 0.7927, 0.6453, 0.7031, 0.6087, 0.6087, 0.7703, 0.7982, 0.7587, 0.6828, 0.6494, 0.839, 0.9101, 0.8924, 0.944, 1.1007, 1.1587, 1.3097, 1.5026, 1.7386, 1.8634, 1.7839, 1.8483, 1.8947, 1.7919, 1.7919, 1.9789, 1.9483, 2.1536, 2.313, 2.3475, 2.4147, 2.589, 2.3588, 2.4774, 2.8859, 2.4453, 2.3777, 2.2289, 1.7706, 2.1787, 2.334, 2.334, 2.0566, 2.0884, 2.1, 2.2227, 2.4769, 2.8308, 2.5044, 2.6165, 2.9789, 2.8558, 2.4173, 2.4865, 2.2486, 2.4866, 2.7937, 3.1024, 3.1024, 2.8315, 3.0472, 3.4454, 3.436, 3.4518, 3.3222, 3.2309, 3.0882, 3.4216, 3.4358, 3.4409, 3.1522, 3.3342, 3.2017, 3.3741, 3.2651]}
{"unit_id": "orig:1984167171405312282", "ticker_idx": 1, "ts": "2025-10-31T07:54:52+00:00", "symbol": "Fabrinet", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Beneficiaries include Lumentum, FN, and leading Chinese optical-module manufacturers.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Lumentum, Fabrinet, and Chinese optical-module makers on surging CoWoS/TPU-driven 1.6T optical-module demand through 2025-2026.", "resolved_tickers": ["FN"], "resolution_method": "exchange_ticker", "resolution_reason": "Fabrinet FN US optical manufacturing", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK: CoWoS and Optical Module Demand in 2025/2026\n\nNVIDIA’s CoWoS demand in 2025/2026 is projected to reach 377,000 / 630,000 wafers, respectively, of which 50,000 wafers will be for the Vera CPU, with the remainder driven by Rubin’s mid-2026 packaging shipments. From a CoWoS capacity perspective, Blackwell/Rubin-related product shipments in 2026 are estimated at 5 million / 2 million units, corresponding to a 1.6 T optical-module demand of 20 million units.\n\nFor Google, due to its own cluster upgrades and the announcement of 1 million TPU demand from companies such as Meta and HPE, its CoWoS demand may rise to 220–230 thousand wafers. TPU shipments are expected to reach 2.7 million / 4 million units in 2025/2026, driving Google’s 800 G / 1.6 T optical-module demand to 6 million / 10 million units, respectively. Consequently, Google’s 1.6 T optical-module forecast will also be revised upward.\n\nThe NVIDIA Rubin GPU is equipped with two CX9 NIC (Network Interface Card) chips, doubling bandwidth compared with Blackwell.\nRubin Ultra adopts four CX9 chips per GPU, further doubling bandwidth.\n\nGoogle plans to complete its large-scale cluster migration in 2026. In the vertical-expansion layer, it will deploy optical interconnects, where the TPU-to-1.6 T optical-module ratio is about 1 : 1.5; in the horizontal-expansion layer, about 1 : 2.5; overall, the TPU-to-optical-module ratio will be roughly 1 : 4 (1.6 T basis).\n\nIn Meta’s Minera architecture, each ASIC chip is paired with two NICs, and the ASIC-to-optical-module ratio will expand as bandwidth increases.\n\nLumentum’s EML (External Modulated Laser) demand is rising rapidly, with price increases expected in 2026.\nIts OCS (Optical Communication Systems) business shipments are projected at 15 thousand / 30 thousand units in 2025 / 2026, representing about 30 % market share in 2026.\n\nIn the CPO/OIO (Co-Packaged Optics / On-Interposer Optics) field, NVIDIA’s CPO switch shipments are expected to reach 2 k / 20 k / 35 k units in 2025 / 2026 / 2027, respectively.\nStarting in 2027, NVIDIA and ASIC vendors will launch OIO solutions, driving incremental demand for CW (Continuous-Wave) lasers, FAUs, and optical engines.\n\nBeneficiaries include Lumentum, FN, and leading Chinese optical-module manufacturers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005833368775663095, "ret_signed_m1d": -0.005833368775663095, "alpha_spy_m1d": -0.002563813224627287, "alpha_c_m1d": -0.004868943803366443, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": -0.0009644249722966514, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0055155653707334995, "ret_signed_p1d": 0.0055155653707334995, "alpha_spy_p1d": 0.003638878282768987, "alpha_c_p1d": 0.0014247995057594665, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.004090765864974033, "ret_p1w": 0.01942936971715903, "ret_signed_p1w": 0.01942936971715903, "alpha_spy_p1w": 0.035689041364865615, "alpha_c_p1w": 0.06106838736340847, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04163901764624944, "ret_p1m": 0.02047345226347641, "ret_signed_p1m": 0.02047345226347641, "alpha_spy_p1m": 0.023097826318630665, "alpha_c_p1m": 0.06813208571929319, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.04765863345581678, "ret_p3m": 0.13716320146893324, "ret_signed_p3m": 0.13716320146893324, "alpha_spy_p3m": 0.11456276995599746, "alpha_c_p3m": 0.14304094218232277, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": -0.005877740713389534, "ret_p6m": 0.5542592513831859, "ret_signed_p6m": 0.5542592513831859, "alpha_spy_p6m": 0.49974478983209725, "alpha_c_p6m": 0.4832634876880664, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.07099576369511951, "price_path": [-0.2553, -0.2454, -0.2394, -0.2192, -0.2455, -0.2094, -0.2344, -0.2689, -0.2514, -0.2575, -0.3526, -0.3637, -0.3712, -0.3319, -0.2912, -0.2495, -0.2457, -0.1941, -0.248, -0.248, -0.238, -0.2339, -0.178, -0.1601, -0.1587, -0.1622, -0.1543, -0.178, -0.1874, -0.1851, -0.193, -0.1869, -0.142, -0.1304, -0.116, -0.137, -0.1678, -0.1807, -0.1854, -0.1742, -0.1724, -0.1572, -0.1612, -0.1693, -0.1595, -0.1598, -0.1297, -0.1324, -0.1785, -0.1316, -0.1483, -0.0959, -0.0638, -0.0696, -0.0749, -0.0558, -0.0915, -0.0581, -0.0444, -0.0385, -0.0068, 0.0076, -0.0058, 0.0, 0.0055, 0.0432, 0.1052, 0.0469, 0.0194, 0.0605, 0.0198, 0.0157, -0.0865, -0.0619, -0.0543, -0.0711, -0.0542, -0.1211, -0.1137, -0.055, -0.0247, 0.0193, 0.0193, 0.0428, 0.0205, 0.0023, 0.0153, 0.0708, 0.086, 0.1145, 0.1314, 0.1726, 0.1988, 0.0688, 0.0668, 0.024, -0.0319, 0.027, 0.0792, 0.0976, 0.1052, 0.0934, 0.0934, 0.0857, 0.0715, 0.0485, 0.0334, 0.0334, 0.0882, 0.0371, 0.0876, 0.0679, 0.0054, 0.0151, 0.0508, 0.1099, 0.0828, 0.1294, 0.1223, 0.1223, 0.0953, 0.1085, 0.0677, 0.0598, 0.0762, 0.1317, 0.1372, 0.1264, 0.1109, 0.134, 0.0181, -0.0398, 0.0048, 0.1439, 0.1392, 0.0584, 0.0567, 0.0488, 0.1248, 0.1248, 0.1276, 0.148, 0.1663, 0.2396, 0.309, 0.3298, 0.3982, 0.2766, 0.2385, 0.3196, 0.2553, 0.2799, 0.2358, 0.1108, 0.1974, 0.2325, 0.1901, 0.1639, 0.1398, 0.164, 0.1349, 0.1359, 0.2274, 0.1491, 0.2404, 0.3639, 0.3892, 0.2477, 0.2529, 0.1165, 0.1837, 0.2143, 0.2665, 0.2665, 0.2583, 0.2659, 0.385, 0.4033, 0.5029, 0.5659, 0.5473, 0.5566, 0.5267, 0.5659, 0.5889, 0.5787, 0.5663, 0.5649, 0.6347, 0.5543]}
{"unit_id": "orig:1984167171405312282", "ticker_idx": 2, "ts": "2025-10-31T07:54:52+00:00", "symbol": "Chinese optical-module manufacturers", "kind": "sector", "geography": "China", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Beneficiaries include Lumentum, FN, and leading Chinese optical-module manufacturers.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Lumentum, Fabrinet, and Chinese optical-module makers on surging CoWoS/TPU-driven 1.6T optical-module demand through 2025-2026.", "resolved_tickers": ["002281.SZ", "300308.SZ"], "resolution_method": "basket", "resolution_reason": "Chinese optical module basket: Eoptolink+Innolight proxies", "bench_c": "MCHI", "bench_c_reason": "country_fallback (suffix='SZ')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK: CoWoS and Optical Module Demand in 2025/2026\n\nNVIDIA’s CoWoS demand in 2025/2026 is projected to reach 377,000 / 630,000 wafers, respectively, of which 50,000 wafers will be for the Vera CPU, with the remainder driven by Rubin’s mid-2026 packaging shipments. From a CoWoS capacity perspective, Blackwell/Rubin-related product shipments in 2026 are estimated at 5 million / 2 million units, corresponding to a 1.6 T optical-module demand of 20 million units.\n\nFor Google, due to its own cluster upgrades and the announcement of 1 million TPU demand from companies such as Meta and HPE, its CoWoS demand may rise to 220–230 thousand wafers. TPU shipments are expected to reach 2.7 million / 4 million units in 2025/2026, driving Google’s 800 G / 1.6 T optical-module demand to 6 million / 10 million units, respectively. Consequently, Google’s 1.6 T optical-module forecast will also be revised upward.\n\nThe NVIDIA Rubin GPU is equipped with two CX9 NIC (Network Interface Card) chips, doubling bandwidth compared with Blackwell.\nRubin Ultra adopts four CX9 chips per GPU, further doubling bandwidth.\n\nGoogle plans to complete its large-scale cluster migration in 2026. In the vertical-expansion layer, it will deploy optical interconnects, where the TPU-to-1.6 T optical-module ratio is about 1 : 1.5; in the horizontal-expansion layer, about 1 : 2.5; overall, the TPU-to-optical-module ratio will be roughly 1 : 4 (1.6 T basis).\n\nIn Meta’s Minera architecture, each ASIC chip is paired with two NICs, and the ASIC-to-optical-module ratio will expand as bandwidth increases.\n\nLumentum’s EML (External Modulated Laser) demand is rising rapidly, with price increases expected in 2026.\nIts OCS (Optical Communication Systems) business shipments are projected at 15 thousand / 30 thousand units in 2025 / 2026, representing about 30 % market share in 2026.\n\nIn the CPO/OIO (Co-Packaged Optics / On-Interposer Optics) field, NVIDIA’s CPO switch shipments are expected to reach 2 k / 20 k / 35 k units in 2025 / 2026 / 2027, respectively.\nStarting in 2027, NVIDIA and ASIC vendors will launch OIO solutions, driving incremental demand for CW (Continuous-Wave) lasers, FAUs, and optical engines.\n\nBeneficiaries include Lumentum, FN, and leading Chinese optical-module manufacturers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06634034784185239, "ret_signed_m1d": 0.06634034784185239, "alpha_spy_m1d": 0.0696099033928882, "alpha_c_m1d": 0.05359853078385812, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": 0.012741817057994265, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0001532516440783116, "ret_signed_p1d": -0.0001532516440783116, "alpha_spy_p1d": -0.002029938732042824, "alpha_c_p1d": 0.0001613881189044264, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": -0.000314639762982738, "ret_p1w": 0.006103738293654182, "ret_signed_p1w": 0.006103738293654182, "alpha_spy_p1w": 0.022363409941360768, "alpha_c_p1w": 0.005474519508955833, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": 0.0006292187846983488, "ret_p1m": 0.06067992704839276, "ret_signed_p1m": 0.06067992704839276, "alpha_spy_p1m": 0.06330430110354701, "alpha_c_p1m": 0.07499482143876668, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.014314894390373922, "ret_p3m": 0.2370315587214361, "ret_signed_p3m": 0.2370315587214361, "alpha_spy_p3m": 0.21443112720850033, "alpha_c_p3m": 0.21939527111105106, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.01763628761038505, "ret_p6m": 0.9738903949355047, "ret_signed_p6m": 0.9738903949355047, "alpha_spy_p6m": 0.919375933384416, "alpha_c_p6m": 1.0643710114253495, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": -0.09048061648984473, "price_path": [-0.3716, -0.3695, -0.3804, -0.3819, -0.3695, -0.3489, -0.3131, -0.335, -0.3336, -0.2963, -0.2724, -0.2369, -0.2622, -0.2298, -0.1671, -0.1619, -0.1343, -0.0601, -0.0797, 0.0178, -0.0644, -0.0303, -0.1425, -0.0793, -0.1284, -0.155, -0.1216, -0.0279, -0.0371, -0.0571, -0.0296, -0.0364, -0.0186, -0.0093, 0.0245, 0.0131, -0.0066, 0.0276, -0.0259, -0.0124, -0.045, -0.045, -0.045, -0.045, -0.045, -0.045, -0.045, -0.0513, -0.0873, -0.1078, -0.168, -0.1499, -0.1337, -0.1399, -0.1027, -0.0457, -0.038, -0.0485, 0.0242, 0.0685, 0.0722, 0.0847, 0.0663, 0.0, -0.0002, -0.0122, -0.0217, 0.0174, 0.0061, 0.0012, -0.038, -0.0219, -0.0249, -0.0613, -0.0375, -0.0456, -0.0296, -0.0288, -0.0776, -0.065, -0.0254, 0.0528, 0.0379, 0.0252, 0.0607, 0.0584, 0.0337, 0.0482, 0.0656, 0.1221, 0.1677, 0.1954, 0.147, 0.149, 0.1181, 0.0783, 0.1682, 0.1437, 0.148, 0.2104, 0.218, 0.2352, 0.2374, 0.2153, 0.2153, 0.2274, 0.2001, 0.2001, 0.2001, 0.2214, 0.1914, 0.2201, 0.1921, 0.1896, 0.2237, 0.1758, 0.2391, 0.2839, 0.2595, 0.2185, 0.1695, 0.1943, 0.23, 0.184, 0.1882, 0.222, 0.237, 0.2045, 0.2609, 0.1735, 0.1927, 0.1453, 0.1207, 0.1008, 0.1544, 0.1502, 0.1174, 0.1325, 0.1172, 0.1172, 0.1172, 0.1172, 0.1172, 0.1172, 0.1172, 0.1679, 0.1721, 0.2029, 0.1672, 0.2146, 0.2026, 0.1878, 0.2153, 0.2126, 0.1779, 0.2622, 0.3127, 0.3499, 0.279, 0.2567, 0.2034, 0.2768, 0.2782, 0.3065, 0.2611, 0.2717, 0.3506, 0.3241, 0.303, 0.3221, 0.2712, 0.3267, 0.2813, 0.3587, 0.3587, 0.3881, 0.5075, 0.5907, 0.6326, 0.683, 0.7015, 0.6716, 0.699, 0.8271, 0.8516, 0.8811, 0.9214, 0.9936, 0.9594, 0.9739]}
{"unit_id": "orig:2000447662836056163", "ticker_idx": 0, "ts": "2025-12-15T06:07:43+00:00", "symbol": "Philadelphia Semiconductor Index", "kind": "index", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we view the current situation as a correction", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities analyst calls the SOX/Samsung Electronics/SK Hynix selloff a temporary correction, not a peak-out, with rebound catalysts in 1H 2026 earnings and CES.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "SOX/Philadelphia Semiconductor Index best proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Reasons Why I Believe This is a Temporary Correction\n\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\nLast Friday (Dec 12), the Philadelphia Semiconductor Index fell by 5.2%, and in the aftermath, as of 9:20 AM today (Dec 15), the stock prices of Samsung Electronics (-3.3%) and SK Hynix (-3.4%) are undergoing a correction. Media reports stating that Oracle has delayed the completion of some OpenAI data centers from 2027 to 2028 due to labor and material shortages had an impact, although Oracle has officially denied these facts.\n\n1. Recurring Highlight of AI Risks\n\nI believe the first correction of this cycle was Sequoia Capital’s \"AI Bubble\" theory in July 2024. Since then, risks have been highlighted in the stock market on a quarterly basis, continuing with the interest rate hike risk in October 2024, the Deepseek shock in January 2025, tariff issues in April, the Sam Altman issue in July, and Michael Burry’s bubble theory in November.\n\n2. The Common Denominator is the ROI Time Lag Issue\n\nThere are causes commonly found in stock price corrections:\n\n- Massive Capex scale\n\n- Weak links to profit (OpenAI, Oracle, NeoCloud)\n\n- Burden of short-term stock price surges\n\nLooking at it more fundamentally, it is the \"time lag issue between investment and return.\" It is the fear that while AI seems likely to generate money, it is unknown how much it can make. Since higher valuations mean future earnings have been priced in (pre-reflected), this fear naturally grows larger.\n\n3. How to Distinguish Between a Cycle Correction and a Peak-out\n\nA period where there is no way to find out the unknown is a correction.\nWhen the unknown parts become certain, it is a peak-out.\nIn a period where future profits are unknown, only investor sentiment wavers, and concerns can turn into cheers at any time.\nWe view the events where future AI profits become concrete as:\n\n1. Full-scale operation of AI DTs (2028)\n\n2. Operation of TSMC Ph2, Hynix Yongin, and Micron ID fabs (2H 2027)\n\n3. OpenAI IPO (2H 2027)\n\nBefore these events, no conclusion can be reached on whether things will get worse or better, and thus the cycle is hard to break. Therefore, we view the current situation as a correction.\n\n4. Rebound Scenario\n\nThe correction phases until recently have been resolved by customer-side profits and services.\n\n1. Service Expansion: Increase in AI traffic (token usage, etc.), expansion of service areas (humanoids, etc.), competitive performance improvements (GPT-5.2, etc.)\n\n2. Earnings: Highlighting the solid growth stories of CSPs like Google and MS, or profitability improvement in weak links like OpenAI or Oracle.\nUltimately, proving the increase in user utility due to AI is the yardstick that drives a return to the rally. From this perspective, we look forward to the earnings announcements in January 2026 and CES.\n\n5. Micron\n\nI believe the guidance figures are the most important part of Micron's earnings. We can check the magnitude of future memory price increases and the profit upside of the supply chain in the interim.\nGood results are never bad. However, the clue to resolving the AI sector correction must be found in the \"end market.\" This is due to the causal relationship of added value. With Micron alone, we cannot distinguish whether the cause of good results is due to the boom of AI customers or their sacrifice.\nThank you.\n\n(Published Data 2025/12/15)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003457073963756052, "ret_signed_m1d": 0.003457073963756052, "alpha_spy_m1d": 0.0019439901167286333, "alpha_c_m1d": 0.0019439901167286333, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.0015130838470274188, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0027203432490249257, "ret_signed_p1d": -0.0027203432490249257, "alpha_spy_p1d": 1.195434625378411e-05, "alpha_c_p1d": 1.195434625378411e-05, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.00273229759527871, "ret_p1w": 0.022440489672194364, "ret_signed_p1w": 0.022440489672194364, "alpha_spy_p1w": 0.013444764359491934, "alpha_c_p1w": 0.013444764359491934, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.00899572531270243, "ret_p1m": 0.11291662520947998, "ret_signed_p1m": 0.11291662520947998, "alpha_spy_p1m": 0.09074913530376905, "alpha_c_p1m": 0.09074913530376905, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.022167489905710935, "ret_p3m": 0.10325220678737779, "ret_signed_p3m": 0.10325220678737779, "alpha_spy_p3m": 0.12191138005206992, "alpha_c_p3m": 0.12191138005206992, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.01865917326469213, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.135, -0.1018, -0.1054, -0.0871, -0.0884, -0.0901, -0.0902, -0.0881, -0.0857, -0.0752, -0.0544, -0.0415, -0.046, -0.027, -0.0449, -0.0194, -0.0218, -0.0781, -0.0373, -0.055, -0.0315, -0.0272, -0.0285, -0.0161, -0.0215, -0.0405, -0.023, -0.005, 0.0199, 0.0289, 0.0445, 0.0307, 0.0287, 0.0374, -0.0004, 0.0187, -0.0051, -0.0135, 0.0167, -0.0052, 0.0075, -0.0229, -0.0224, -0.0357, -0.0556, -0.0381, -0.0788, -0.0759, -0.039, -0.0365, -0.0147, -0.0147, -0.0018, 0.0002, 0.0185, 0.0325, 0.0247, 0.0326, 0.0443, 0.0456, 0.0601, 0.051, 0.0035, 0.0, -0.0027, -0.0387, -0.016, 0.0094, 0.0224, 0.0323, 0.0351, 0.0351, 0.04, 0.0353, 0.0327, 0.0237, 0.0237, 0.0611, 0.0733, 0.1018, 0.0944, 0.0772, 0.1064, 0.1104, 0.1129, 0.1039, 0.1268, 0.1381, 0.1381, 0.1097, 0.1425, 0.145, 0.1372, 0.1336, 0.1576, 0.1843, 0.1868, 0.1468, 0.1596, 0.1304, 0.0859, 0.0832, 0.1417, 0.1559, 0.1505, 0.179, 0.1544, 0.1589, 0.1589, 0.1583, 0.1727, 0.166, 0.1797, 0.1736, 0.1915, 0.2114, 0.1711, 0.1551, 0.1552, 0.1116, 0.1344, 0.1238, 0.0817, 0.121, 0.1294, 0.1399, 0.1033, 0.101, 0.1197, 0.1281, 0.119, 0.1226, 0.0936, 0.1124, 0.1216, 0.1342, 0.0825, 0.0638, 0.0305, 0.0898, 0.1142, 0.1152, 0.1152, 0.1256, 0.1367, 0.2021, 0.2231, 0.2418, 0.2602, 0.2848, 0.2876, 0.2928, 0.3194, 0.3188, 0.3208, 0.3554, 0.3696, 0.4395, 0.439, 0.3963, 0.42, 0.4403, 0.4492, 0.4405, 0.4857, 0.5627, 0.5352, 0.6104, 0.6382, 0.5953, 0.6272, 0.6439, 0.5814, 0.5524, 0.5462, 0.605, 0.6142, 0.6382, 0.6382, 0.7116, 0.6927, 0.705, 0.7024, null, null, null, null, null, null, null]}
{"unit_id": "orig:2000447662836056163", "ticker_idx": 1, "ts": "2025-12-15T06:07:43+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reasons Why I Believe This is a Temporary Correction", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities analyst calls the SOX/Samsung Electronics/SK Hynix selloff a temporary correction, not a peak-out, with rebound catalysts in 1H 2026 earnings and CES.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Reasons Why I Believe This is a Temporary Correction\n\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\nLast Friday (Dec 12), the Philadelphia Semiconductor Index fell by 5.2%, and in the aftermath, as of 9:20 AM today (Dec 15), the stock prices of Samsung Electronics (-3.3%) and SK Hynix (-3.4%) are undergoing a correction. Media reports stating that Oracle has delayed the completion of some OpenAI data centers from 2027 to 2028 due to labor and material shortages had an impact, although Oracle has officially denied these facts.\n\n1. Recurring Highlight of AI Risks\n\nI believe the first correction of this cycle was Sequoia Capital’s \"AI Bubble\" theory in July 2024. Since then, risks have been highlighted in the stock market on a quarterly basis, continuing with the interest rate hike risk in October 2024, the Deepseek shock in January 2025, tariff issues in April, the Sam Altman issue in July, and Michael Burry’s bubble theory in November.\n\n2. The Common Denominator is the ROI Time Lag Issue\n\nThere are causes commonly found in stock price corrections:\n\n- Massive Capex scale\n\n- Weak links to profit (OpenAI, Oracle, NeoCloud)\n\n- Burden of short-term stock price surges\n\nLooking at it more fundamentally, it is the \"time lag issue between investment and return.\" It is the fear that while AI seems likely to generate money, it is unknown how much it can make. Since higher valuations mean future earnings have been priced in (pre-reflected), this fear naturally grows larger.\n\n3. How to Distinguish Between a Cycle Correction and a Peak-out\n\nA period where there is no way to find out the unknown is a correction.\nWhen the unknown parts become certain, it is a peak-out.\nIn a period where future profits are unknown, only investor sentiment wavers, and concerns can turn into cheers at any time.\nWe view the events where future AI profits become concrete as:\n\n1. Full-scale operation of AI DTs (2028)\n\n2. Operation of TSMC Ph2, Hynix Yongin, and Micron ID fabs (2H 2027)\n\n3. OpenAI IPO (2H 2027)\n\nBefore these events, no conclusion can be reached on whether things will get worse or better, and thus the cycle is hard to break. Therefore, we view the current situation as a correction.\n\n4. Rebound Scenario\n\nThe correction phases until recently have been resolved by customer-side profits and services.\n\n1. Service Expansion: Increase in AI traffic (token usage, etc.), expansion of service areas (humanoids, etc.), competitive performance improvements (GPT-5.2, etc.)\n\n2. Earnings: Highlighting the solid growth stories of CSPs like Google and MS, or profitability improvement in weak links like OpenAI or Oracle.\nUltimately, proving the increase in user utility due to AI is the yardstick that drives a return to the rally. From this perspective, we look forward to the earnings announcements in January 2026 and CES.\n\n5. Micron\n\nI believe the guidance figures are the most important part of Micron's earnings. We can check the magnitude of future memory price increases and the profit upside of the supply chain in the interim.\nGood results are never bad. However, the clue to resolving the AI sector correction must be found in the \"end market.\" This is due to the causal relationship of added value. With Micron alone, we cannot distinguish whether the cause of good results is due to the boom of AI customers or their sacrifice.\nThank you.\n\n(Published Data 2025/12/15)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03912210331054089, "ret_signed_m1d": 0.03912210331054089, "alpha_spy_m1d": 0.03760901946351347, "alpha_c_m1d": 0.029354021596027735, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.009768081714513155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019083943680241022, "ret_signed_p1d": -0.019083943680241022, "alpha_spy_p1d": -0.016351646084962312, "alpha_c_p1d": -0.02091102355885721, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": 0.001827079878616189, "ret_p1w": 0.05438933331892093, "ret_signed_p1w": 0.05438933331892093, "alpha_spy_p1w": 0.0453936080062185, "alpha_c_p1w": 0.032743727030979164, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.021645606287941765, "ret_p1m": 0.319359684466187, "ret_signed_p1m": 0.319359684466187, "alpha_spy_p1m": 0.2971921945604761, "alpha_c_p1m": 0.2884238415073155, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.03093584295887153, "ret_p3m": 0.8016547449724258, "ret_signed_p3m": 0.8016547449724258, "alpha_spy_p3m": 0.820313918237118, "alpha_c_p3m": 0.831527802060285, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.029873057087859145, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2571, -0.2372, -0.2372, -0.2068, -0.1954, -0.1887, -0.1821, -0.2087, -0.1966, -0.1994, -0.1794, -0.1436, -0.1436, -0.1436, -0.1436, -0.1436, -0.1436, -0.0992, -0.1097, -0.126, -0.0935, -0.0677, -0.0658, -0.0639, -0.0697, -0.0592, -0.0792, -0.0573, -0.0267, -0.0506, -0.041, -0.0067, 0.0258, 0.0601, 0.001, -0.0401, -0.0534, -0.0658, -0.0401, -0.0124, -0.0162, -0.0191, -0.0725, -0.0401, -0.0668, -0.0792, -0.0401, -0.0954, -0.0773, -0.0525, -0.0191, -0.0124, -0.041, -0.0382, -0.0134, -0.0029, 0.0029, 0.0344, 0.0448, 0.0344, 0.0305, 0.0239, 0.0391, 0.0, -0.0191, 0.0296, 0.0267, 0.0143, 0.0544, 0.0639, 0.0601, 0.0601, 0.1164, 0.1458, 0.1496, 0.1496, 0.1496, 0.2321, 0.3242, 0.3318, 0.352, 0.3309, 0.3328, 0.3309, 0.3194, 0.3452, 0.3798, 0.4277, 0.4315, 0.3922, 0.4335, 0.4603, 0.4584, 0.4584, 0.5293, 0.5572, 0.5409, 0.5389, 0.4421, 0.6061, 0.6214, 0.5274, 0.5207, 0.5955, 0.5898, 0.6089, 0.7125, 0.7374, 0.7374, 0.7374, 0.7374, 0.8218, 0.8227, 0.8506, 0.9177, 0.9512, 1.0903, 1.0759, 1.0759, 0.8707, 0.6511, 0.8371, 0.8045, 0.6636, 0.8017, 0.8218, 0.8017, 0.7595, 0.8093, 0.8592, 0.9992, 0.9225, 0.9119, 0.7863, 0.8189, 0.8122, 0.7269, 0.7269, 0.6939, 0.6065, 0.8222, 0.7141, 0.789, 0.8553, 0.888, 1.0225, 0.9601, 0.9793, 0.9313, 0.9841, 1.0273, 1.0898, 1.0754, 1.061, 1.1042, 1.0898, 1.157, 1.109, 1.157, 1.133, 1.1715, 1.1186, 1.1186, 1.2339, 1.2339, 1.5558, 1.6086, 1.5798, 1.7431, 1.6807, 1.7287, 1.844, 1.599, 1.6999, 1.6471, 1.6519, 1.8777, 1.8104, 1.8104, 1.8729, 1.9497, 1.8777, 2.0458, null, null, null, null, null, null, null]}
{"unit_id": "orig:2000447662836056163", "ticker_idx": 2, "ts": "2025-12-15T06:07:43+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reasons Why I Believe This is a Temporary Correction", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities analyst calls the SOX/Samsung Electronics/SK Hynix selloff a temporary correction, not a peak-out, with rebound catalysts in 1H 2026 earnings and CES.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Reasons Why I Believe This is a Temporary Correction\n\n[Samsung Securities Semiconductor, IT / Lee Jong-wook]\n\nLast Friday (Dec 12), the Philadelphia Semiconductor Index fell by 5.2%, and in the aftermath, as of 9:20 AM today (Dec 15), the stock prices of Samsung Electronics (-3.3%) and SK Hynix (-3.4%) are undergoing a correction. Media reports stating that Oracle has delayed the completion of some OpenAI data centers from 2027 to 2028 due to labor and material shortages had an impact, although Oracle has officially denied these facts.\n\n1. Recurring Highlight of AI Risks\n\nI believe the first correction of this cycle was Sequoia Capital’s \"AI Bubble\" theory in July 2024. Since then, risks have been highlighted in the stock market on a quarterly basis, continuing with the interest rate hike risk in October 2024, the Deepseek shock in January 2025, tariff issues in April, the Sam Altman issue in July, and Michael Burry’s bubble theory in November.\n\n2. The Common Denominator is the ROI Time Lag Issue\n\nThere are causes commonly found in stock price corrections:\n\n- Massive Capex scale\n\n- Weak links to profit (OpenAI, Oracle, NeoCloud)\n\n- Burden of short-term stock price surges\n\nLooking at it more fundamentally, it is the \"time lag issue between investment and return.\" It is the fear that while AI seems likely to generate money, it is unknown how much it can make. Since higher valuations mean future earnings have been priced in (pre-reflected), this fear naturally grows larger.\n\n3. How to Distinguish Between a Cycle Correction and a Peak-out\n\nA period where there is no way to find out the unknown is a correction.\nWhen the unknown parts become certain, it is a peak-out.\nIn a period where future profits are unknown, only investor sentiment wavers, and concerns can turn into cheers at any time.\nWe view the events where future AI profits become concrete as:\n\n1. Full-scale operation of AI DTs (2028)\n\n2. Operation of TSMC Ph2, Hynix Yongin, and Micron ID fabs (2H 2027)\n\n3. OpenAI IPO (2H 2027)\n\nBefore these events, no conclusion can be reached on whether things will get worse or better, and thus the cycle is hard to break. Therefore, we view the current situation as a correction.\n\n4. Rebound Scenario\n\nThe correction phases until recently have been resolved by customer-side profits and services.\n\n1. Service Expansion: Increase in AI traffic (token usage, etc.), expansion of service areas (humanoids, etc.), competitive performance improvements (GPT-5.2, etc.)\n\n2. Earnings: Highlighting the solid growth stories of CSPs like Google and MS, or profitability improvement in weak links like OpenAI or Oracle.\nUltimately, proving the increase in user utility due to AI is the yardstick that drives a return to the rally. From this perspective, we look forward to the earnings announcements in January 2026 and CES.\n\n5. Micron\n\nI believe the guidance figures are the most important part of Micron's earnings. We can check the magnitude of future memory price increases and the profit upside of the supply chain in the interim.\nGood results are never bad. However, the clue to resolving the AI sector correction must be found in the \"end market.\" This is due to the causal relationship of added value. With Micron alone, we cannot distinguish whether the cause of good results is due to the boom of AI customers or their sacrifice.\nThank you.\n\n(Published Data 2025/12/15)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030685848548443806, "ret_signed_m1d": 0.030685848548443806, "alpha_spy_m1d": 0.029172764701416387, "alpha_c_m1d": 0.027228774584687754, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.043321337592038156, "ret_signed_p1d": -0.043321337592038156, "alpha_spy_p1d": -0.040589039996759446, "alpha_c_p1d": -0.04060099434301323, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.046931364275827114, "ret_signed_p1w": 0.046931364275827114, "alpha_spy_p1w": 0.037935638963124685, "alpha_c_p1w": 0.02449087460363275, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3321299157434354, "ret_signed_p1m": 0.3321299157434354, "alpha_spy_p1m": 0.3099624258377245, "alpha_c_p1m": 0.21921329053395544, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.6817978803316003, "ret_signed_p3m": 0.6817978803316003, "alpha_spy_p3m": 0.7004570535962924, "alpha_c_p3m": 0.5785456735442225, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3984, -0.3588, -0.3588, -0.3669, -0.3488, -0.3552, -0.357, -0.393, -0.3705, -0.3732, -0.3506, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.228, -0.2514, -0.2578, -0.2379, -0.1838, -0.1603, -0.1243, -0.136, -0.1315, -0.1369, -0.0801, -0.035, -0.0602, 0.0065, 0.0245, 0.0083, 0.1183, 0.057, 0.0444, 0.0696, 0.0462, 0.0931, 0.1165, 0.1129, 0.1039, 0.0101, 0.0931, 0.0281, 0.0137, 0.0299, -0.0602, -0.062, -0.0638, -0.0548, -0.0181, -0.0433, -0.0289, 0.0072, -0.0036, -0.0217, -0.0181, 0.0415, 0.0217, 0.0596, 0.0199, 0.0307, 0.0, -0.0433, -0.0054, -0.0036, -0.0126, 0.0469, 0.0542, 0.0614, 0.0614, 0.0812, 0.1552, 0.1751, 0.1751, 0.1751, 0.222, 0.2563, 0.3105, 0.3394, 0.3646, 0.343, 0.352, 0.3321, 0.3394, 0.352, 0.3646, 0.3791, 0.3412, 0.3357, 0.3628, 0.3845, 0.3285, 0.444, 0.5181, 0.5542, 0.6408, 0.4982, 0.6372, 0.6245, 0.5199, 0.5144, 0.6011, 0.5812, 0.5523, 0.6029, 0.5884, 0.5884, 0.5884, 0.5884, 0.6137, 0.713, 0.7166, 0.8141, 0.8375, 0.9874, 0.9187, 0.9187, 0.6981, 0.5353, 0.7017, 0.6709, 0.5118, 0.6963, 0.727, 0.6818, 0.6456, 0.7614, 0.7541, 0.9097, 0.8319, 0.821, 0.6872, 0.7831, 0.7993, 0.6872, 0.6872, 0.5787, 0.4594, 0.6221, 0.501, 0.5841, 0.6022, 0.6565, 0.8681, 0.8048, 0.8572, 0.8807, 0.9946, 1.0543, 1.0887, 1.0399, 1.1086, 1.2135, 1.2117, 1.2153, 1.2098, 1.3364, 1.3509, 1.3382, 1.3256, 1.3256, 1.6167, 1.6167, 1.8952, 1.9911, 2.0489, 2.3998, 2.3184, 2.5734, 2.5625, 2.2895, 2.3274, 2.1556, 2.1556, 2.5083, 2.5101, 2.5101, 2.7108, 3.0562, 3.1401, 3.2197, null, null, null, null, null, null, null]}
{"unit_id": "orig:2001547972786458646", "ticker_idx": 0, "ts": "2025-12-18T06:59:58+00:00", "symbol": "semiconductors", "kind": "sector", "geography": null, "sub_sector": "NAND", "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "we are turning more cautious on short-term memory cycle, especially for NAND", "tweet_type": "analysis", "timeline": "quarters", "summary": "Turning cautious near-term on NAND/DRAM pricing into 2Q26 on consumer weakness; long-term still constructive on NAND with SNDK as an explicit name.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND semiconductor basket: Micron, SK Hynix, Samsung, SanDisk", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Memory Supercycle: Time for a Brief Pause\n\nAfter a strong run in stock prices, we are turning more cautious on short-term memory cycle, especially for NAND. The key concern is weakening consumer electronics demand due to rising BOM costs. We are seeing major local Android OEMs and Samsung planning at least a 10% decline in units in 2026. As consumer electronics still represent 37% of total DRAM demand and 56% of total NAND demand, this slowdown is likely to materially narrow the demand–supply gap in 2026. Additionally, our checks suggest that channel inventory will like support Android OEMs moving toward more normalized inventory levels in 1Q26, reducing the urgency for aggressive restocking. As a result, we expect price increases in both DRAM and NAND to moderate significantly in 2Q26.\n\n(....)\n\nNAND – More exposed to weakness; LT supply-demand healthy\n\nWe expect NAND contract price increase to slightly fall behind DRAM and we maintain our 30- 35%/25-30% QoQ forecast. We are turning cautious on NAND pricing in 2Q26, as our surveys indicate higher NAND inventory accumulation among local smartphone OEMs than in DRAM, while non-CSP demand still accounts for ~56% of total NAND demand in 2026, leaving NAND more exposed to consumer electronics weakness. As a matter of fact, we already see Kioxia’s smartphone price hike in 1Q26 already down to 10-15% vs 25%+ in 4Q25. In a longer run, we still like NAND cycle thanks to the inflection point in 2027 for token-driven eSSD demand and refrained NAND expansion. NAND capacity additions remaining disciplined, with meaningful expansions limited to Kioxia/SNDK’s K2 and YMTC. Samsung’s P4L has shifted away from NAND and foundry foward DRAM and Micron/SK has no plans for NAND expansion for their ID1/M15X fab.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03730782653202666, "ret_signed_m1d": 0.03730782653202666, "alpha_spy_m1d": 0.029813089208523297, "alpha_c_m1d": 0.014240828582912285, "bench_spy_m1d": -0.007494737323503364, "bench_c_m1d": -0.023066997949114376, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032862103860807734, "ret_signed_p1d": -0.032862103860807734, "alpha_spy_p1d": -0.02379865838135592, "alpha_c_p1d": -0.00700176155755311, "bench_spy_p1d": 0.009063445479451815, "bench_c_p1d": 0.025860342303254624, "ret_p1w": 0.0976698353014081, "ret_signed_p1w": -0.0976698353014081, "alpha_spy_p1w": -0.07409148387321501, "alpha_c_p1w": -0.04568088878523385, "bench_spy_p1w": 0.02357835142819309, "bench_c_p1w": 0.05198894651617425, "ret_p1m": 0.5262231626113187, "ret_signed_p1m": -0.5262231626113187, "alpha_spy_p1m": -0.5007470808283414, "alpha_c_p1m": -0.36960440166221287, "bench_spy_p1m": 0.025476081782977333, "bench_c_p1m": 0.15661876094910587, "ret_p3m": 1.1777830300371313, "ret_signed_p3m": -1.1777830300371313, "alpha_spy_p3m": -1.1832494625590428, "alpha_c_p3m": -1.0313037405970766, "bench_spy_p3m": -0.005466432521911591, "bench_c_p3m": 0.14647928944005462, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3653, -0.3522, -0.3644, -0.3744, -0.3863, -0.3524, -0.3518, -0.3162, -0.2864, -0.2773, -0.2824, -0.2879, -0.2646, -0.2713, -0.2711, -0.2482, -0.2678, -0.2306, -0.2002, -0.1986, -0.1757, -0.1831, -0.186, -0.161, -0.1073, -0.0984, -0.1099, -0.0532, -0.0527, -0.0451, 0.0106, -0.0501, -0.0187, -0.0248, 0.0021, 0.068, 0.0725, 0.0876, 0.0316, 0.0171, 0.0542, -0.006, -0.014, -0.0821, -0.1072, -0.0562, -0.0574, -0.0474, -0.0366, -0.0342, -0.0409, -0.0322, -0.0502, -0.0394, -0.0029, 0.0209, 0.0121, 0.0473, 0.0404, -0.011, -0.0368, -0.0488, -0.0373, 0.0, 0.0329, 0.0722, 0.0804, 0.0977, 0.0977, 0.1144, 0.1433, 0.1429, 0.1324, 0.1324, 0.2376, 0.264, 0.397, 0.4097, 0.3766, 0.4387, 0.4547, 0.4398, 0.4408, 0.4801, 0.5262, 0.5308, 0.559, 0.6468, 0.6711, 0.6444, 0.6163, 0.6959, 0.7991, 0.818, 0.8603, 0.9254, 1.0162, 0.8501, 0.7949, 0.8286, 0.8407, 0.7764, 0.8769, 0.9537, 0.9496, 0.9496, 0.8967, 0.9291, 0.976, 1.0449, 1.0641, 1.0701, 1.0882, 1.1684, 1.1256, 1.1074, 0.9077, 0.873, 0.9181, 0.832, 0.8468, 0.9749, 1.0449, 0.9735, 1.0238, 1.1286, 1.1778, 1.2891, 1.2543, 1.1565, 1.0654, 1.0885, 1.0494, 0.8886, 0.9048, 0.7847, 0.8212, 1.0101, 0.9619, 1.001, 1.0595, 1.0652, 1.2602, 1.3245, 1.3416, 1.4566, 1.5282, 1.4842, 1.5406, 1.5244, 1.5224, 1.5493, 1.6698, 1.6282, 1.6957, 1.8586, 1.7587, 1.8496, 1.8691, 1.9974, 2.2114, 2.447, 2.626, 2.563, 2.9247, 3.0845, 2.9111, 3.0186, 2.9427, 2.7912, 2.6969, 2.7155, 2.7607, 3.105, 3.0055, 3.0055, 3.3432, 3.4818, 3.5392, 3.7087, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2001547972786458646", "ticker_idx": 1, "ts": "2025-12-18T06:59:58+00:00", "symbol": "semiconductors", "kind": "sector", "geography": null, "sub_sector": "DRAM", "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "we expect price increases in both DRAM and NAND to moderate significantly in 2Q26", "tweet_type": "analysis", "timeline": "quarters", "summary": "Turning cautious near-term on NAND/DRAM pricing into 2Q26 on consumer weakness; long-term still constructive on NAND with SNDK as an explicit name.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "DRAM semiconductor basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Memory Supercycle: Time for a Brief Pause\n\nAfter a strong run in stock prices, we are turning more cautious on short-term memory cycle, especially for NAND. The key concern is weakening consumer electronics demand due to rising BOM costs. We are seeing major local Android OEMs and Samsung planning at least a 10% decline in units in 2026. As consumer electronics still represent 37% of total DRAM demand and 56% of total NAND demand, this slowdown is likely to materially narrow the demand–supply gap in 2026. Additionally, our checks suggest that channel inventory will like support Android OEMs moving toward more normalized inventory levels in 1Q26, reducing the urgency for aggressive restocking. As a result, we expect price increases in both DRAM and NAND to moderate significantly in 2Q26.\n\n(....)\n\nNAND – More exposed to weakness; LT supply-demand healthy\n\nWe expect NAND contract price increase to slightly fall behind DRAM and we maintain our 30- 35%/25-30% QoQ forecast. We are turning cautious on NAND pricing in 2Q26, as our surveys indicate higher NAND inventory accumulation among local smartphone OEMs than in DRAM, while non-CSP demand still accounts for ~56% of total NAND demand in 2026, leaving NAND more exposed to consumer electronics weakness. As a matter of fact, we already see Kioxia’s smartphone price hike in 1Q26 already down to 10-15% vs 25%+ in 4Q25. In a longer run, we still like NAND cycle thanks to the inflection point in 2027 for token-driven eSSD demand and refrained NAND expansion. NAND capacity additions remaining disciplined, with meaningful expansions limited to Kioxia/SNDK’s K2 and YMTC. Samsung’s P4L has shifted away from NAND and foundry foward DRAM and Micron/SK has no plans for NAND expansion for their ID1/M15X fab.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030560311592769136, "ret_signed_m1d": 0.030560311592769136, "alpha_spy_m1d": 0.02306557426926577, "alpha_c_m1d": 0.007493313643654759, "bench_spy_m1d": -0.007494737323503364, "bench_c_m1d": -0.023066997949114376, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01624848250545859, "ret_signed_p1d": -0.01624848250545859, "alpha_spy_p1d": -0.007185037026006775, "alpha_c_p1d": 0.009611859797796034, "bench_spy_p1d": 0.009063445479451815, "bench_c_p1d": 0.025860342303254624, "ret_p1w": 0.08371836027909103, "ret_signed_p1w": -0.08371836027909103, "alpha_spy_p1w": -0.060140008850897944, "alpha_c_p1w": -0.03172941376291678, "bench_spy_p1w": 0.02357835142819309, "bench_c_p1w": 0.05198894651617425, "ret_p1m": 0.4067252325454476, "ret_signed_p1m": -0.4067252325454476, "alpha_spy_p1m": -0.3812491507624703, "alpha_c_p1m": -0.25010647159634175, "bench_spy_p1m": 0.025476081782977333, "bench_c_p1m": 0.15661876094910587, "ret_p3m": 0.8098591910592067, "ret_signed_p3m": -0.8098591910592067, "alpha_spy_p3m": -0.8153256235811183, "alpha_c_p3m": -0.6633799016191521, "bench_spy_p3m": -0.005466432521911591, "bench_c_p3m": 0.14647928944005462, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.31, -0.2979, -0.3042, -0.3091, -0.3293, -0.3089, -0.3061, -0.2722, -0.237, -0.2314, -0.2272, -0.2343, -0.2197, -0.2254, -0.2057, -0.202, -0.2171, -0.1933, -0.1527, -0.1444, -0.1258, -0.1376, -0.1378, -0.1351, -0.0924, -0.066, -0.0798, -0.048, -0.0343, -0.0296, 0.0331, -0.029, -0.0205, -0.0152, -0.0276, 0.017, 0.0175, 0.0201, 0.0055, -0.0299, 0.0018, -0.0466, -0.0589, -0.0737, -0.1138, -0.0863, -0.0781, -0.0565, -0.0421, -0.0515, -0.0404, -0.0215, -0.0289, -0.0432, -0.0175, 0.0188, 0.0161, 0.0427, 0.0202, 0.0056, -0.0223, -0.0497, -0.0306, 0.0, 0.0162, 0.0635, 0.0686, 0.0837, 0.0837, 0.1061, 0.1534, 0.159, 0.1493, 0.1493, 0.232, 0.269, 0.3316, 0.3426, 0.3273, 0.345, 0.3484, 0.3276, 0.3321, 0.3519, 0.4067, 0.4128, 0.3904, 0.4343, 0.4634, 0.4728, 0.4399, 0.53, 0.5974, 0.6049, 0.6052, 0.5568, 0.6319, 0.5789, 0.5181, 0.5299, 0.5681, 0.5459, 0.5922, 0.6476, 0.6478, 0.6478, 0.6318, 0.6602, 0.6913, 0.7393, 0.7399, 0.7903, 0.8238, 0.901, 0.8691, 0.8695, 0.6848, 0.5874, 0.6984, 0.6417, 0.5682, 0.6932, 0.731, 0.6914, 0.6935, 0.7694, 0.8099, 0.9074, 0.833, 0.7973, 0.6869, 0.7177, 0.7029, 0.602, 0.6044, 0.51, 0.4632, 0.628, 0.5502, 0.6024, 0.6454, 0.6739, 0.8275, 0.8059, 0.8284, 0.8287, 0.9365, 0.9578, 0.9909, 0.967, 0.9764, 1.0268, 1.0727, 1.088, 1.0907, 1.186, 1.1558, 1.1831, 1.16, 1.1936, 1.3744, 1.46, 1.693, 1.7155, 1.8599, 2.0955, 2.0094, 2.16, 2.1567, 1.9169, 1.9045, 1.8529, 1.8991, 2.1308, 2.0947, 2.0947, 2.3766, 2.5608, 2.5589, 2.7038, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2001547972786458646", "ticker_idx": 2, "ts": "2025-12-18T06:59:58+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "In a longer run, we still like NAND cycle thanks to the inflection point in 2027 for token-driven eSSD demand and refrained NAND expansion", "tweet_type": "analysis", "timeline": "quarters", "summary": "Turning cautious near-term on NAND/DRAM pricing into 2Q26 on consumer weakness; long-term still constructive on NAND with SNDK as an explicit name.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Memory Supercycle: Time for a Brief Pause\n\nAfter a strong run in stock prices, we are turning more cautious on short-term memory cycle, especially for NAND. The key concern is weakening consumer electronics demand due to rising BOM costs. We are seeing major local Android OEMs and Samsung planning at least a 10% decline in units in 2026. As consumer electronics still represent 37% of total DRAM demand and 56% of total NAND demand, this slowdown is likely to materially narrow the demand–supply gap in 2026. Additionally, our checks suggest that channel inventory will like support Android OEMs moving toward more normalized inventory levels in 1Q26, reducing the urgency for aggressive restocking. As a result, we expect price increases in both DRAM and NAND to moderate significantly in 2Q26.\n\n(....)\n\nNAND – More exposed to weakness; LT supply-demand healthy\n\nWe expect NAND contract price increase to slightly fall behind DRAM and we maintain our 30- 35%/25-30% QoQ forecast. We are turning cautious on NAND pricing in 2Q26, as our surveys indicate higher NAND inventory accumulation among local smartphone OEMs than in DRAM, while non-CSP demand still accounts for ~56% of total NAND demand in 2026, leaving NAND more exposed to consumer electronics weakness. As a matter of fact, we already see Kioxia’s smartphone price hike in 1Q26 already down to 10-15% vs 25%+ in 4Q25. In a longer run, we still like NAND cycle thanks to the inflection point in 2027 for token-driven eSSD demand and refrained NAND expansion. NAND capacity additions remaining disciplined, with meaningful expansions limited to Kioxia/SNDK’s K2 and YMTC. Samsung’s P4L has shifted away from NAND and foundry foward DRAM and Micron/SK has no plans for NAND expansion for their ID1/M15X fab.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.057550371349799234, "ret_signed_m1d": -0.057550371349799234, "alpha_spy_m1d": -0.05005563402629587, "alpha_c_m1d": -0.042360350145345116, "bench_spy_m1d": -0.007494737323503364, "bench_c_m1d": -0.015190021204454118, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08270296792685516, "ret_signed_p1d": 0.08270296792685516, "alpha_spy_p1d": 0.07363952244740335, "alpha_c_p1d": 0.06108350544868535, "bench_spy_p1d": 0.009063445479451815, "bench_c_p1d": 0.021619462478169815, "ret_p1w": 0.1395242603683593, "ret_signed_p1w": 0.1395242603683593, "alpha_spy_p1w": 0.11594590894016621, "alpha_c_p1w": 0.10432627195378541, "bench_spy_p1w": 0.02357835142819309, "bench_c_p1w": 0.03519798841457389, "ret_p1m": 0.8847169528089318, "ret_signed_p1m": 0.8847169528089318, "alpha_spy_p1m": 0.8592408710259545, "alpha_c_p1m": 0.8543306404110556, "bench_spy_p1m": 0.025476081782977333, "bench_c_p1m": 0.03038631239787626, "ret_p3m": 2.2815545469709053, "ret_signed_p3m": 2.2815545469709053, "alpha_spy_p3m": 2.2870209794928167, "alpha_c_p3m": 2.294189439567616, "bench_spy_p3m": -0.005466432521911591, "bench_c_p3m": -0.012634892596710579, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.531, -0.5152, -0.5451, -0.5704, -0.5575, -0.4828, -0.4887, -0.4481, -0.4343, -0.4149, -0.4479, -0.4489, -0.3991, -0.4091, -0.4673, -0.3866, -0.42, -0.3425, -0.3426, -0.3613, -0.3254, -0.3197, -0.3304, -0.2388, -0.1517, -0.1958, -0.2002, -0.0688, -0.1077, -0.0917, -0.0567, -0.1134, -0.0135, -0.0536, 0.0912, 0.221, 0.2375, 0.29, 0.1099, 0.1581, 0.2115, 0.1161, 0.1208, -0.1071, -0.0874, 0.0342, 0.0047, -0.0201, -0.0201, 0.0174, -0.0423, -0.0643, -0.1143, -0.028, 0.0411, 0.0274, 0.0, 0.0611, 0.1009, -0.0605, -0.0802, -0.0462, -0.0576, 0.0, 0.0827, 0.0984, 0.1159, 0.1395, 0.1395, 0.1394, 0.113, 0.0946, 0.0817, 0.0817, 0.2542, 0.2489, 0.5931, 0.611, 0.5244, 0.7197, 0.7738, 0.7762, 0.7671, 0.8648, 0.8847, 0.8847, 1.0647, 1.2842, 1.294, 1.1591, 1.1453, 1.1937, 1.4042, 1.4574, 1.6258, 2.0313, 2.1692, 1.6636, 1.6255, 1.7246, 1.6583, 1.4681, 1.731, 1.872, 1.855, 1.855, 1.6911, 1.7358, 1.8301, 1.9617, 2.037, 1.9095, 1.8815, 1.9705, 1.8951, 1.8209, 1.5764, 1.7297, 1.5772, 1.4029, 1.6826, 1.8201, 1.9866, 1.8197, 2.0148, 2.2062, 2.2816, 2.4343, 2.5181, 2.2339, 2.201, 2.2009, 2.0888, 1.7484, 1.8061, 1.6087, 1.895, 2.1565, 2.1969, 2.1969, 2.3019, 2.2389, 2.5583, 2.8803, 2.8812, 3.3402, 3.3036, 3.0632, 3.1897, 3.1966, 3.1603, 3.1169, 3.4613, 3.2487, 3.5106, 3.8765, 3.5673, 3.8492, 3.9964, 4.4087, 4.7225, 5.4081, 5.4248, 5.1057, 6.119, 6.0517, 5.6163, 5.5945, 5.3006, 5.414, 5.074, 5.3032, 5.3454, 6.0274, 5.7379, 5.7379, 6.243, 6.2448, 6.4804, 6.7234, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:1980790963833590087", "ticker_idx": 0, "ts": "2025-10-22T00:19:01+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA still holds five strategic lines of defense that will allow it to maintain its absolute dominance in the AI market", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish NVDA: five strategic defenses (speed, NVLink, OpenAI investment, US alignment, power moat) ensure continued dominance despite ASIC market share growth capped at 10–20%.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Concerns Over ASIC Market Share Limits… NVIDIA’s “Five Strategies” to Block the Counterattack\n\nGPU vs. ASIC: Not a Substitution, But a Coexistence\n\nGlobal cloud service providers (CSPs) such as Google, AWS, Meta, and Microsoft continue to purchase NVIDIA’s AI GPUs in large quantities while simultaneously accelerating the development and production of their own ASICs (application-specific integrated circuits) to reduce reliance on NVIDIA.\n\nHowever, supply-chain insiders note that “NVIDIA still holds five strategic lines of defense that will allow it to maintain its absolute dominance in the AI market.”\n\nThe prevailing industry view is that GPUs and ASICs complement each other rather than substitute for one another. AI GPUs are highly versatile and flexible, while ASICs are optimized for specific purposes.\nIn short, “GPUs specialize in training, while ASICs specialize in inference,” and the two technologies are evolving in parallel.\n\nTherefore, even if CSPs develop in-house ASICs, these efforts are more likely to serve as cost-saving supplements rather than genuine threats to NVIDIA’s market dominance.\n\nASIC’s 20% Market Share Ceiling… NVIDIA’s Profitability Remains Overwhelming\n\nIndustry participants cite a lack of flexibility as ASIC’s biggest limitation.\n\nWhen AI model architectures or algorithms change, ASICs must be redesigned — an expensive and risky process.\nAs a result, many believe ASICs will ultimately capture only around 10–20% of the total market.\nCurrently, the adoption ratio between GPUs and ASICs in the AI accelerator market is estimated to cap out at roughly 6:4, even under favorable assumptions for ASICs.\n\nIn that case, NVIDIA would still claim the overwhelming majority of total AI semiconductor profits.\nAccording to supply-chain data, NVIDIA occupies more than half of TSMC’s CoWoS (2.5D packaging) production capacity and, together with Apple, has emerged as one of TSMC’s largest customers.\n\nIn other words, even if ASIC volumes grow, NVIDIA’s dominance in allocation and scale remains unchanged.\n\nNVIDIA’s “Five Strategic Defenses”\n\n① Rapid Generation Turnover — “Winning With Speed”\n\nIndustry observers note that NVIDIA is already executing an ultra-fast product generation-cycle strategy to counter the ASIC camp’s challenge.\n\nAI semiconductors now generate hundreds of billions of dollars annually, and NVIDIA has shifted to releasing a new AI GPU every year.\n\nFor example, the transition from B200 → B300 is already underway, with VR200 currently in testing and scheduled for mass production in Q3 2026.\n\nThis accelerated cadence leaves competitors no time to catch up and forces the entire supply chain to align its production rhythm with NVIDIA’s cycle.\nIn short, “speed itself becomes the entry barrier.”\n\n② NVLink Fusion — The Start of Platform Expansion\nThis year, NVIDIA announced its “NVLink Fusion” strategy, introducing a new standard that seamlessly connects its GPUs with other CPU architectures.\n\nIn particular, it is collaborating with Intel to extend NVLink to the x86 CPU ecosystem — a move expected to become a key bridge for the future of heterogeneous computing.\n\n③ Investment in OpenAI — Building an “AI Perpetual Motion Machine”\n\nNVIDIA has made large-scale investments in OpenAI, aiming to create an integrated hardware-software “AI perpetual motion” ecosystem.\nBy anchoring OpenAI’s model-training infrastructure on NVIDIA GPUs, the company further entrenches its platform as the core of the entire AI ecosystem.\n\n④ Aligning With “America First” Amid U.S.–China Tech Tensions\n\nAlthough U.S. export controls and AI regulations have repeatedly posed challenges for NVIDIA, CEO Jensen Huang has turned these into political opportunities.\n\nAt the upcoming GTC DC (Data Center Conference), NVIDIA is expected not only to showcase its AI technological prowess but also to deliver a message of goodwill toward the Trump administration’s “America First” agenda.\nIn effect, this is a strategic gesture to realign with the U.S. government and minimize political risk.\n\n⑤ Electricity — The New Bottleneck\n\nFinally, the industry is recognizing that “the real bottleneck for AI is now electricity.”\n\nA single AI server (equipped with eight GPUs) consumes over 10 kW of power.\n\nLarge-scale data centers require hundreds of megawatts (MW) — comparable to a small nuclear power plant.\nThus, the future growth limit of the AI industry may come not from computational capacity, but from power scarcity.\n\nIn other words, the sustainability of AI hinges on securing sufficient electrical infrastructure — signaling a massive structural transformation across global power grids, renewable energy systems, and cooling technologies.\n\n$NVDA\n\n---\n\nhttps://t.co/52wxn7tYeo", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004881266256250294, "ret_signed_m1d": 0.004881266256250294, "alpha_spy_m1d": -0.0003447186403731095, "alpha_c_m1d": -0.014906040806114351, "bench_spy_m1d": 0.0052259848966234035, "bench_c_m1d": 0.019787307062364645, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010428209748937878, "ret_signed_p1d": 0.010428209748937878, "alpha_spy_p1d": 0.004498318558438186, "alpha_c_p1d": -0.007882444751792006, "bench_spy_p1d": 0.005929891190499692, "bench_c_p1d": 0.018310654500729884, "ret_p1w": 0.1484357812349617, "ret_signed_p1w": 0.1484357812349617, "alpha_spy_p1w": 0.11910068406087237, "alpha_c_p1w": 0.059806159655094504, "bench_spy_p1w": 0.02933509717408933, "bench_c_p1w": 0.0886296215798672, "ret_p1m": 0.0019968624119472977, "ret_signed_p1m": 0.0019968624119472977, "alpha_spy_p1m": 0.0248629706658966, "alpha_c_p1m": 0.041866933621806046, "bench_spy_p1m": -0.022866108253949302, "bench_c_p1m": -0.03987007120985875, "ret_p3m": 0.033061659031214496, "ret_signed_p3m": 0.033061659031214496, "alpha_spy_p3m": -0.0057280081979582675, "alpha_c_p3m": -0.15310451225361166, "bench_spy_p3m": 0.038789667229172764, "bench_c_p3m": 0.18616617128482615, "ret_p6m": 0.10035380609678746, "ret_signed_p6m": 0.10035380609678746, "alpha_spy_p6m": 0.04366738109235935, "alpha_c_p6m": -0.2470033718649347, "bench_spy_p6m": 0.05668642500442811, "bench_c_p6m": 0.34735717796172216, "price_path": [-0.0377, -0.0196, -0.0265, -0.0057, -0.0134, -0.0364, -0.0016, -0.0113, -0.0048, 0.0027, 0.0134, 0.0098, 0.0159, 0.0072, 0.0096, 0.0009, 0.0095, -0.0258, -0.0271, -0.0295, -0.0128, -0.0027, 0.0082, 0.0073, -0.0007, -0.0339, -0.0339, -0.0527, -0.0536, -0.0479, -0.0736, -0.0664, -0.0529, -0.0164, -0.0173, -0.0136, -0.014, -0.03, -0.0554, -0.0224, -0.02, 0.0185, -0.0103, -0.0184, -0.0144, -0.0116, 0.0087, 0.0349, 0.0386, 0.0478, 0.0407, 0.0292, 0.0264, 0.049, 0.0682, 0.016, 0.0446, -0.0014, -0.0025, 0.0085, 0.0163, 0.0131, 0.0049, 0.0, 0.0104, 0.0332, 0.0622, 0.1151, 0.1484, 0.1254, 0.1232, 0.1475, 0.1021, 0.0828, 0.0433, 0.0437, 0.1041, 0.0714, 0.075, 0.0365, 0.0549, 0.0351, 0.006, 0.0346, 0.002, -0.0078, 0.0126, -0.0136, -0.0001, -0.0001, -0.0182, -0.002, 0.0065, -0.0038, 0.0173, 0.0119, 0.0293, 0.0261, 0.0195, 0.0037, -0.0291, -0.0221, -0.0141, -0.0518, -0.034, 0.004, 0.019, 0.0496, 0.0463, 0.0463, 0.0569, 0.0441, 0.0403, 0.0346, 0.0346, 0.0476, 0.0435, 0.0387, 0.049, 0.0265, 0.0255, 0.0259, 0.0307, 0.0159, 0.0376, 0.0331, 0.0331, -0.0122, 0.0169, 0.0254, 0.041, 0.0344, 0.0458, 0.0624, 0.0679, 0.0602, 0.0296, 0.0004, -0.0337, -0.0465, 0.0285, 0.0542, 0.0459, 0.0543, 0.037, 0.0141, 0.0141, 0.0261, 0.0428, 0.0423, 0.053, 0.0626, 0.0698, 0.0848, 0.0256, -0.0171, 0.0123, -0.0012, 0.0154, 0.017, -0.0136, 0.0132, 0.025, 0.032, 0.016, -0.0001, 0.0164, 0.0093, 0.0008, -0.0094, -0.0419, -0.0256, -0.0281, -0.0088, -0.05, -0.0707, -0.0837, -0.0325, -0.025, -0.0159, -0.0159, -0.0145, -0.012, 0.0101, 0.0202, 0.0464, 0.0502, 0.0901, 0.1032, 0.1004]}
{"unit_id": "orig:1988018501504954812", "ticker_idx": 0, "ts": "2025-11-10T22:58:41+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "At the current share price, HBM growth momentum is not fully reflected; upside remains", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs NDR note on Samsung: HBM growth not priced in, upside remains; DRAM/HBM earnings turnaround expected in 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electronics: Top 5 Points from the U.S. NDR – Outlook for Tight Memory Supply–Demand in 2026 (Goldman Sachs)\n    \n1. Memory market outlook\n\n• Both DRAM and NAND demand are expected to exceed supply in 2026\n\n• Given today’s low inventory levels, Samsung judges that a meaningful supply increase in the short term is difficult\n\n• Goldman Sachs estimates DRAM demand +23% YoY (supply +20%), NAND demand +19% YoY (supply +17%)\n\n• In 2026, DRAM is expected to face a supply shortfall of about 3.3%, and NAND about 2.5%\n    \n2. 2026 investment strategy and production plans\n\n• Despite strong memory demand, Samsung maintains a conservative approach\n\n• No specific annual guidance due to uncertainty in the second half of 2026\n\n• P4 cleanroom to be used mainly for DRAM production, preparing for HBM4 on the 1Cnm process\n\n• DRAM demand remains solid across all applications, while NAND demand is more limited and server-centric\n\n• Given low margins in NAND, additional CapEx is currently limited\n    \n3. DRAM competitiveness and flexibility\n\n• Continued investment even during downturns has secured cleanroom headroom versus competitors\n\n• Focus on the new P4 fab for 1Cnm HBM4; 1Bnm will concentrate on node transitions to serve general DRAM\n\n• HBM3E is based on 1A–1Bnm, while HBM4 will expand centered on 1Cnm\n\n• DRAM supply flexibility is expected to act as a competitive advantage\n    \n4. HBM technology progress\n\n• Supplying HBM3E 12-Hi to all customers, with implied potential for Nvidia qualification\n\n• HBM4 samples achieve 11 Gbps speed and low power that exceed JEDEC standards\n\n• HBM4 core die (1Cnm DRAM) and base die (4nm foundry) are each produced in-house on Samsung’s own processes\n\n• Introduced a dual-track strategy for Custom HBM (customers can choose the foundry)\n    \n5. NAND segment strategy\n\n• No intention for large-scale expansion given the low margin structure\n\n• 2026 NAND bit demand is expected to be in the mid-teens, but Samsung’s production growth will be below that level\n\n• Despite server-led demand recovery, limited cleanroom space leads to DRAM-first allocation, keeping NAND supply tight\n    \n6. Overall implications\n\n• Expected to enter an earnings turnaround phase centered on DRAM/HBM in 2026\n\n• Structural earnings improvement possible on broad memory undersupply and strengthened HBM competitiveness\n\n• Conservative CapEx and limited NAND investment act positively for price stability\n\n• At the current share price, HBM growth momentum is not fully reflected; upside remains", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0268389099250409, "ret_signed_m1d": -0.0268389099250409, "alpha_spy_m1d": -0.01147423224448918, "alpha_c_m1d": -0.0019006588509518085, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02493825107408909, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02882701787404751, "ret_signed_p1d": 0.02882701787404751, "alpha_spy_p1d": 0.026537868171601753, "alpha_c_p1d": 0.03735397683432329, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.008526958960275777, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": 0.02314227390131529, "alpha_c_p1w": 0.04023276598428149, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.04023276598428149, "ret_p1m": 0.0775348024458482, "ret_signed_p1m": 0.0775348024458482, "alpha_spy_p1m": 0.07518701202974065, "alpha_c_p1m": 0.0758089493486993, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.0017258530971488995, "ret_p3m": 0.5911965541547499, "ret_signed_p3m": 0.5911965541547499, "alpha_spy_p3m": 0.5938640310421116, "alpha_c_p3m": 0.6719278967011242, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": -0.08073134254637426, "ret_p6m": 1.3271748175912519, "ret_signed_p6m": 1.3271748175912519, "alpha_spy_p6m": 1.2590087261920881, "alpha_c_p6m": 1.2031773412902789, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.12399747630097302, "price_path": [-0.2885, -0.2914, -0.2914, -0.3073, -0.3073, -0.3023, -0.3013, -0.2934, -0.2924, -0.3043, -0.3013, -0.3112, -0.3102, -0.331, -0.3162, -0.3092, -0.3063, -0.3122, -0.3063, -0.2924, -0.2815, -0.2736, -0.2538, -0.2429, -0.2142, -0.2261, -0.2053, -0.2053, -0.1737, -0.1618, -0.1549, -0.1479, -0.1756, -0.163, -0.166, -0.1451, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.0616, -0.0726, -0.0895, -0.0557, -0.0288, -0.0268, -0.0249, -0.0308, -0.0199, -0.0408, -0.0179, 0.0139, -0.0109, -0.001, 0.0348, 0.0686, 0.1044, 0.0427, 0.0, -0.0139, -0.0268, 0.0, 0.0288, 0.0249, 0.0219, -0.0338, 0.0, -0.0278, -0.0408, 0.0, -0.0577, -0.0388, -0.0129, 0.0219, 0.0288, -0.001, 0.002, 0.0278, 0.0388, 0.0447, 0.0775, 0.0885, 0.0775, 0.0736, 0.0666, 0.0825, 0.0417, 0.0219, 0.0726, 0.0696, 0.0567, 0.0984, 0.1083, 0.1044, 0.1044, 0.163, 0.1936, 0.1976, 0.1976, 0.1976, 0.2835, 0.3794, 0.3874, 0.4084, 0.3864, 0.3884, 0.3864, 0.3744, 0.4014, 0.4374, 0.4873, 0.4913, 0.4504, 0.4933, 0.5213, 0.5193, 0.5193, 0.5932, 0.6222, 0.6052, 0.6032, 0.5023, 0.6731, 0.6891, 0.5912, 0.5842, 0.6621, 0.6561, 0.6761, 0.784, 0.8099, 0.8099, 0.8099, 0.8099, 0.8978, 0.8988, 0.9278, 0.9977, 1.0327, 1.1775, 1.1625, 1.1625, 0.9488, 0.7201, 0.9138, 0.8799, 0.733, 0.8769, 0.8978, 0.8769, 0.8329, 0.8849, 0.9368, 1.0826, 1.0027, 0.9917, 0.8609, 0.8949, 0.8879, 0.799, 0.799, 0.7646, 0.6736, 0.8983, 0.7857, 0.8637, 0.9328, 0.9668, 1.107, 1.0419, 1.0619, 1.0119, 1.0669, 1.112, 1.177, 1.162, 1.147, 1.192, 1.177, 1.2471, 1.1971, 1.2471, 1.2221, 1.2621, 1.2071, 1.2071, 1.3272, 1.3272]}
{"unit_id": "orig:1981492896965751066", "ticker_idx": 0, "ts": "2025-10-23T22:48:15+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM raises the target price from KRW 460,000 → KRW 650,000", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM raise on SK Hynix to KRW 650K on HBM-led structural upcycle, record 56% OP margin forecast, and 2026 EPS ~20% above consensus.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Hynix: Transitioning from Cyclical Volatility to Structural Stability, Target Price Raised to KRW 650,000 (JPM)\n\n⸻\n\n1. Key Investment Points\n    •    The company is seen entering the early stage of a memory upcycle, with greater structural stability than in previous cycles.\n    •    Its AI solution execution capabilities and high-margin, HBM-centered portfolio have significantly strengthened profit resilience compared to the past.\n    •    DRAM/NAND ASPs are projected to rise +24% / +18% YoY, with the 2026 operating margin forecasted to reach a record 56%.\n    •    Applying an average P/B of 2.4x for 2026–2027, JPM raises the target price from KRW 460,000 → KRW 650,000.\n    •    This reflects the visibility of a long-term upcycle through 2027 and a sustained high-margin structure.\n\n⸻\n\n2. Demand and Industry Outlook\n    •    Continuous memory procurement of DDR5/LPDDR5 by CSPs, with AI inference demand surging and reinforcing Token Economics.\n    •    HBM revenue share within DRAM expected to rise from 43% in 2025 → 50% in 2027.\n    •    Launch of 1TB HBM stack solutions to drive high-capacity product sales expansion.\n    •    Broader collaboration with OpenAI and major CSPs is expected to support mid- to long-term demand.\n    •    Supply expansion remains limited due to cleanroom constraints and the HBM transition, which raises HBM’s share of total DRAM capacity from 21% in 2025 → 30% in 2027.\n\n⸻\n\n3. Profitability and Valuation\n    •    Operating margins: 2025E 46.2% → 2026E 55.5% → 2027E 51.7%.\n    •    2026 JPM-estimated EPS is about 20% above consensus.\n    •    ROE: 2025E 42.8% → 2026E 40.7% → 2027E 33.3%.\n    •    EV/EBITDA: 2025E 6.1x → 2026E 3.5x → 2027E 2.6x.\n    •    Free Cash Flow: 2025E KRW 1.86 trillion → 2027E KRW 3.98 trillion.\n    •    JPM allows for the upper range of historical peak valuations, reflecting a sustained expansion cycle lasting over four years.\n\n⸻\n\n4. Major Growth Drivers\n    •    Transition from HBM3E → HBM4 → HBM4E to reinforce technological leadership.\n    •    Continued ASP premium driven by growing demand for AI accelerators (ASIC/HBM).\n    •    HBM OP mix projected to account for ~42% of total operating profit by 2027.\n    •    CapEx growth restrained at around +23% YoY through 2026, maintaining supply discipline.\n    •    Potential long-term upside if the “Stargate Project” materializes into new AI infrastructure memory partnerships.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006269553061375888, "ret_signed_m1d": 0.006269553061375888, "alpha_spy_m1d": 0.012164487929532286, "alpha_c_m1d": 0.024250956299965765, "bench_spy_m1d": -0.0058949348681563984, "bench_c_m1d": -0.017981403238589877, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06583063457728788, "ret_signed_p1d": 0.06583063457728788, "alpha_spy_p1d": 0.05765806347265712, "alpha_c_p1d": 0.047443047002391214, "bench_spy_p1d": 0.008172571104630766, "bench_c_p1d": 0.01838758757489667, "ret_p1w": 0.18704267045842626, "ret_signed_p1w": 0.18704267045842626, "alpha_spy_p1w": 0.17502947796828439, "alpha_c_p1w": 0.13214119746089503, "bench_spy_p1w": 0.012013192490141877, "bench_c_p1w": 0.05490147299753123, "ret_p1m": 0.0888191704331811, "ret_signed_p1m": 0.0888191704331811, "alpha_spy_p1m": 0.10776934880811884, "alpha_c_p1m": 0.14296642792537329, "bench_spy_p1m": -0.01895017837493773, "bench_c_p1m": -0.05414725749219218, "ret_p3m": 0.5538810057977224, "ret_signed_p3m": 0.5538810057977224, "alpha_spy_p3m": 0.542236697905456, "alpha_c_p3m": 0.41813633140707207, "bench_spy_p3m": 0.011644307892266337, "bench_c_p3m": 0.1357446743906503, "ret_p6m": 1.3634078869534187, "ret_signed_p6m": 1.3634078869534187, "alpha_spy_p6m": 1.3002550630993321, "alpha_c_p6m": 1.0130473950255667, "bench_spy_p6m": 0.0631528238540866, "bench_c_p6m": 0.350360491927852, "price_path": [-0.4532, -0.4522, -0.4501, -0.4292, -0.4616, -0.4616, -0.4501, -0.4605, -0.4532, -0.4647, -0.4428, -0.4386, -0.4199, -0.423, -0.423, -0.4418, -0.4512, -0.4668, -0.4887, -0.4762, -0.4585, -0.4543, -0.4574, -0.4389, -0.4378, -0.465, -0.4556, -0.4514, -0.4451, -0.4284, -0.4211, -0.3981, -0.3647, -0.3584, -0.3135, -0.3083, -0.2727, -0.303, -0.2571, -0.2571, -0.2665, -0.2456, -0.2529, -0.255, -0.2968, -0.2706, -0.2738, -0.2476, -0.1735, -0.1735, -0.1735, -0.1735, -0.1735, -0.1735, -0.1055, -0.1327, -0.14, -0.117, -0.0543, -0.0272, 0.0146, 0.001, 0.0063, 0.0, 0.0658, 0.1181, 0.0888, 0.1661, 0.187, 0.1682, 0.2957, 0.2247, 0.21, 0.2393, 0.2121, 0.2665, 0.2936, 0.2894, 0.279, 0.1703, 0.2665, 0.1912, 0.1745, 0.1933, 0.0888, 0.0867, 0.0846, 0.0951, 0.1377, 0.1084, 0.1252, 0.167, 0.1544, 0.1335, 0.1377, 0.2067, 0.1837, 0.2276, 0.1816, 0.1942, 0.1586, 0.1084, 0.1523, 0.1544, 0.144, 0.213, 0.2214, 0.2297, 0.2297, 0.2527, 0.3385, 0.3615, 0.3615, 0.3615, 0.4159, 0.4556, 0.5183, 0.5518, 0.5811, 0.556, 0.5664, 0.5434, 0.5518, 0.5664, 0.5811, 0.5978, 0.5539, 0.5476, 0.579, 0.6041, 0.5392, 0.6731, 0.7588, 0.8007, 0.901, 0.7358, 0.8969, 0.8822, 0.7609, 0.7547, 0.855, 0.832, 0.7986, 0.8571, 0.8404, 0.8404, 0.8404, 0.8404, 0.8697, 0.9847, 0.9889, 1.1018, 1.129, 1.3026, 1.223, 1.223, 0.9674, 0.7788, 0.9716, 0.936, 0.7516, 0.9653, 1.0009, 0.9486, 0.9067, 1.0407, 1.0324, 1.2126, 1.1225, 1.1099, 0.9548, 1.0659, 1.0847, 0.9548, 0.9548, 0.8291, 0.6908, 0.8794, 0.739, 0.8354, 0.8564, 0.9192, 1.1644, 1.091, 1.1518, 1.179, 1.311, 1.3802, 1.42, 1.3634]}
{"unit_id": "orig:2000459627721568722", "ticker_idx": 0, "ts": "2025-12-15T06:55:16+00:00", "symbol": "memory semiconductors", "kind": "sector", "geography": "global", "sub_sector": "DRAM, NAND, HBM", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "J.P. Morgan's answer is very clear: Stay Long. This implies that top players still have over 50% market upside from current levels.", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares J.P. Morgan's bullish 2026-27 memory outlook: Stay Long, 50%+ upside, longest/strongest up-cycle in history, DRAM shortage persists through 2027.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T"], "resolution_method": "basket", "resolution_reason": "Global memory semicon (DRAM/NAND/HBM) basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan 2026 Memory Market Outlook: Memory Giants' Market Cap Approaching $1 Trillion This Year, to Reach $1.5 Trillion in 2027\n\nIn its latest research report, J.P. Morgan noted that the total market capitalization of top memory chip manufacturers is nearing $1 trillion. Based on historical valuation medians, this figure is projected to surge to $1.5 trillion by 2027, implying that top players still have over 50% upside potential.\n\nOn December 14, J.P. Morgan stated in its latest report that the current cycle will be the longest and strongest memory up-cycle in history.\n\nInvestors are concerned that new capacity in 2027 will lead to DRAM oversupply. However, J.P. Morgan's data model shows that such concerns are unnecessary. Due to HBM capacity crowding out and structural demand from AI inference (inference consumes 3x the memory of training), DRAM bit supply growth will still lag behind demand growth over the next two years.\n\nThe market is experiencing \"Dual-track\" pricing. B2B (Enterprise/AI) demand strongly supports high prices, whereas B2C (Consumer) faces cyclical pressure. However, JPM argues that overall, the upside in server demand will fully offset the downside risk in the consumer sector.\n\nValuation Reshaping: Towards $1.5 Trillion\n\nJ.P. Morgan addressed investor concerns directly in the report: With memory stocks having risen sharply over the past three months and approaching the $1 trillion market cap barrier, what is the next step?\nJ.P. Morgan's answer is very clear: Stay Long.\n\nBased on the \"Market Cap/TAM (Total Addressable Market)\" valuation framework, J.P. Morgan predicts the memory market size will reach approximately $420 billion in 2027. Applying the median Price-to-Sales (P/S) ratio of 3.5x from the 2018 and 2021 cycles, the combined market cap of top storage and memory makers is projected to approach $1.5 trillion in 2027. This implies that top players still have over 50% market upside from current levels.\n\nThe Truth of Supply and Demand: Still Short in 2027\n\nThe biggest short thesis currently in the market is that new fab ramp-ups and accelerated tech migration in 2027 will trigger DRAM oversupply. J.P. Morgan refutes this view through a bottom-up \"Capacity-Bit\" analysis model.\n\n- Shortage Persists: Although the supply-demand gap may narrow from 5% in 2026 to 3% in 2027, it will remain in a shortage state.\n\n- Capacity Crowding Out: Strong CSP (Cloud Service Provider) demand forces manufacturers to allocate more capacity to HBM. HBM's share of total DRAM capacity will rise significantly from 19% in 2025 to 28% in 2027.\n\n- Supply Constraints: Regular DRAM capacity in 2026 is actually expected to decline year-over-year. Even with new capacity coming online in 2027, such as Samsung P4 and Hynix M15X, DRAM bit shipment growth will be limited to under 20% due to cleanroom space constraints and natural capacity loss from increased process steps.\n\nB2B Feast and B2C Cycle\n\nCSPs and specific tier-2/3 consumer electronics brands aggressively securing resources have triggered the recent sharp rise in memory prices. J.P. Morgan predicts a distinct price polarization in the market from 2H 2026 to 1H 2027.\n\n- B2B (AI Driven): Prices will remain firm supported by AI inference demand.\n\n- B2C (Consumer): Will face cyclical price declines due to customer resistance to high prices.\n\nKey Data Forecasts:\n\n- FY26E: J.P. Morgan models show DRAM Average Selling Price (ASP) skyrocketing 53% and NAND ASP rising about 30%.\n\n- FY27E: Even against the backdrop of an extended duration of high prices, DRAM ASP is expected to rise only slightly by 1%, while NAND ASP may correct slightly by 6%.\n\nAI Growth Drivers: Structural Opportunities in HBM and eSSD\n\nHBM (High Bandwidth Memory): Beneficiary of GPU vs. ASIC Competition. \nJPM points out that positive feedback on Google Gemini 3.0 has triggered a route competition between GPUs and ASICs, but this serves as a double tailwind for HBM demand.\n\n- Spec Upgrade: Google's next-gen 2nm TPU is likely to adopt HBM4, and the 4x capacity increase brought by Rubin Ultra GPUs will continue to pressure the supply chain.\n\n- Continued Shortage: J.P. Morgan expects the HBM supply shortage (gap of about 8%-12%) to persist through 2027 and possibly extend into 2028.\n\nSSD (Solid State Drive): Key to Inference Applications. The rise of AI inference is driving the expansion of the enterprise SSD (eSSD) market. AI servers carry 3x the SSD volume of regular servers.\nSince HDD (Hard Disk Drive) manufacturers' capital expenditure guidance for 2026 is conservative, J.P. Morgan expects eSSD to enjoy massive demand tailwinds over the next 6 months, driving NAND prices up 27% in FY26.\n\nCapital Expenditure (Capex): Increasing but Disciplined\n\nAlthough memory makers have announced a series of capacity expansion plans, J.P. Morgan believes the actual bit supply growth generated will be offset by physical migration challenges.\n\n- Equipment Spending Leads: Growth in Memory Wafer Fab Equipment (WFE) is expected to significantly outpace overall capex growth (DRAM WFE growing 19%/26% in 2026/2027 respectively).\n\n- Intensity Control: Implied capital intensity for both DRAM (below 30%) and NAND (below 20%) will be lower than the 5-year average. This implies that significant discipline is still being maintained on the supply side.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04003355747144954, "ret_signed_m1d": 0.04003355747144954, "alpha_spy_m1d": 0.038520473624422125, "alpha_c_m1d": 0.03657648350769349, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0347670061926503, "ret_signed_p1d": -0.0347670061926503, "alpha_spy_p1d": -0.03203470859737159, "alpha_c_p1d": -0.032046662943625376, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.09161340383096189, "ret_signed_p1w": 0.09161340383096189, "alpha_spy_p1w": 0.08261767851825946, "alpha_c_p1w": 0.06917291415876753, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3908174816664404, "ret_signed_p1m": 0.3908174816664404, "alpha_spy_p1m": 0.3686499917607295, "alpha_c_p1m": 0.27790085645696044, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.8750294965104504, "ret_signed_p3m": 0.8750294965104504, "alpha_spy_p3m": 0.8936886697751425, "alpha_c_p3m": 0.7717772897230726, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3749, -0.3485, -0.3547, -0.3393, -0.33, -0.3313, -0.3369, -0.3656, -0.3423, -0.3347, -0.313, -0.2672, -0.2419, -0.2348, -0.2518, -0.2409, -0.2361, -0.2233, -0.22, -0.236, -0.2091, -0.1627, -0.1653, -0.1351, -0.1472, -0.1401, -0.1362, -0.0652, -0.0172, -0.0392, 0.0038, 0.0365, 0.0382, 0.0858, 0.0364, 0.0376, 0.0674, 0.0727, 0.1418, 0.1384, 0.1417, 0.1246, 0.0166, 0.0713, 0.0097, 0.0159, 0.0175, -0.0481, -0.0266, -0.025, -0.0482, -0.0193, -0.0166, -0.0223, 0.001, -0.0103, -0.0197, 0.01, 0.0573, 0.048, 0.0575, 0.0441, 0.04, 0.0, -0.0348, -0.0039, 0.0258, 0.0341, 0.0916, 0.0908, 0.1203, 0.1255, 0.1594, 0.1753, 0.1729, 0.1653, 0.1653, 0.2292, 0.2823, 0.3374, 0.3752, 0.3715, 0.3772, 0.3798, 0.3908, 0.3831, 0.4078, 0.4809, 0.498, 0.4809, 0.5504, 0.6111, 0.6026, 0.5806, 0.6767, 0.742, 0.7556, 0.8123, 0.695, 0.8173, 0.779, 0.6957, 0.6933, 0.7166, 0.6979, 0.7345, 0.8402, 0.8858, 0.8709, 0.8603, 0.8568, 0.8758, 0.8935, 0.8938, 0.9783, 0.9822, 1.0342, 1.0093, 1.0179, 0.8409, 0.7416, 0.8545, 0.8016, 0.6938, 0.8308, 0.9091, 0.875, 0.8724, 0.9725, 0.9767, 1.099, 1.014, 0.9862, 0.8775, 0.8904, 0.9152, 0.803, 0.7805, 0.7012, 0.6409, 0.841, 0.7628, 0.8229, 0.8819, 0.9162, 1.1511, 1.138, 1.2212, 1.2521, 1.4363, 1.3818, 1.4467, 1.3378, 1.3439, 1.4425, 1.5346, 1.562, 1.5427, 1.6399, 1.6392, 1.6606, 1.6765, 1.6717, 1.8094, 1.8765, 2.0544, 2.2608, 2.4029, 2.622, 2.5601, 2.7944, 2.736, 2.4434, 2.6229, 2.5392, 2.6168, 2.9032, 2.9311, 3.1498, 3.287, 3.3749, 3.3926, 3.6287, null, null, null, null, null, null, null]}
{"unit_id": "orig:1991451583767519527", "ticker_idx": 0, "ts": "2025-11-20T10:20:31+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it will be difficult to achieve a production growth rate of more than 20% year-on-year in 2026… forecast that production in 2026 will be limited to growth of around 9% versus 2025", "tweet_type": "analysis", "timeline": "quarters", "summary": "Relays bearish analyst note on MU: DRAM CAPA growth capped ~20% in 2026 due to EUV/fab constraints; NAND growth ~9% as company retreats from China.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "If you’re a Micron investor, you really need to see this.\n\nThere’s a comment about Micron in a report written by Hyung Geun Ryu at Daishin Securities, whose work I personally like to follow.\n\nTo sum it up… it’s basically saying that Micron’s CAPA increase next year won’t be that big.\n\nI’ll translate the full text.\n\nDRAM: It is estimated that the number of EUV layers within DRAM 1c was reduced in the process of transferring to mass production.\n\nAs a result, the process line is likely to have become longer than before, and constraints on fab space are also likely to have increased.\n\nUntil Idaho Fab 1 (1Q27 wafer in, 2Q27 wafer out) begins operation, available fab space will be limited, and given that the yield base is high, it will be difficult to achieve a production growth rate of more than 20% year-on-year in 2026.\n\nWe forecast DRAM production growth of up to 20%.\n-------------------------------------------------------\nNAND: We estimate that the company has made efficiency its overarching business principle.\n\nAmid geopolitical risk and the rise of YMTC, we believe the company will drastically reduce its dependence on the Chinese market, where medium- to long-term competitiveness deterioration is inevitable, and will restructure its business with a focus on responding to the server eSSD market, where demand growth is strong.\n\nWe estimate that it is even considering reducing medium- to long-term capacity, and we forecast that production in 2026 will be limited to growth of around 9% versus 2025.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.12191482313669932, "ret_signed_m1d": -0.12191482313669932, "alpha_spy_m1d": -0.10643656158002313, "alpha_c_m1d": -0.07780527413441751, "bench_spy_m1d": 0.015478261556676198, "bench_c_m1d": 0.04410954900228181, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029795885308456338, "ret_signed_p1d": -0.029795885308456338, "alpha_spy_p1d": -0.019834662331823516, "alpha_c_p1d": -0.02662756776821662, "bench_spy_p1d": 0.009961222976632822, "bench_c_p1d": 0.003168317540239718, "ret_p1w": 0.14346718927696367, "ret_signed_p1w": -0.14346718927696367, "alpha_spy_p1w": -0.10185991919182591, "alpha_c_p1w": -0.07391938475425563, "bench_spy_p1w": 0.04160727008513776, "bench_c_p1w": 0.06954780452270803, "ret_p1m": 0.3205541343637506, "ret_signed_p1m": -0.3205541343637506, "alpha_spy_p1m": -0.2744703202293177, "alpha_c_p1m": -0.2247989472760452, "bench_spy_p1m": 0.046083814134432854, "bench_c_p1m": 0.09575518708770536, "ret_p3m": 0.9861026552473828, "ret_signed_p3m": -0.9861026552473828, "alpha_spy_p3m": -0.9365452581600149, "alpha_c_p3m": -0.7287110128937184, "bench_spy_p3m": 0.04955739708736795, "bench_c_p3m": 0.2573916423536644, "ret_p6m": 2.601615214063034, "ret_signed_p6m": -2.601615214063034, "alpha_spy_p6m": -2.462389783740798, "alpha_c_p6m": -1.8850014295888542, "bench_spy_p6m": 0.13922543032223622, "bench_c_p6m": 0.71661378447418, "price_path": [-0.4222, -0.4218, -0.4156, -0.3945, -0.4094, -0.4094, -0.412, -0.4108, -0.3836, -0.348, -0.3476, -0.3288, -0.3052, -0.2527, -0.2197, -0.217, -0.2118, -0.206, -0.1618, -0.1924, -0.183, -0.1741, -0.1975, -0.2217, -0.2195, -0.1866, -0.1696, -0.096, -0.0881, -0.0672, -0.0517, -0.0779, -0.024, -0.0449, -0.0982, -0.0427, -0.0711, -0.0468, 0.0058, 0.005, 0.0268, 0.0046, -0.0144, 0.0265, 0.0876, 0.093, 0.102, 0.1254, 0.1124, 0.1112, 0.1655, 0.0827, 0.1794, 0.1835, 0.1815, 0.2579, 0.1973, 0.2162, 0.1767, 0.2258, 0.2015, 0.1347, 0.1219, 0.0, 0.0298, 0.112, 0.115, 0.1435, 0.1435, 0.1744, 0.1941, 0.1893, 0.1628, 0.1255, 0.178, 0.2262, 0.2535, 0.3096, 0.2835, 0.1975, 0.1794, 0.1546, 0.1199, 0.2343, 0.3206, 0.3735, 0.372, 0.4236, 0.4236, 0.4143, 0.4624, 0.4538, 0.4179, 0.4179, 0.567, 0.5508, 0.7062, 0.6869, 0.6246, 0.7144, 0.7183, 0.6798, 0.6561, 0.6724, 0.8021, 0.8021, 0.8133, 0.9331, 0.9752, 0.9855, 0.933, 1.0381, 1.1625, 1.165, 1.0611, 1.175, 1.0838, 0.8849, 0.9022, 0.9608, 0.9052, 0.8543, 1.0386, 1.0566, 1.0451, 1.0451, 0.9861, 1.0913, 1.0734, 1.1271, 1.0914, 1.0767, 1.1313, 1.0645, 1.0486, 1.0501, 0.8862, 0.991, 0.9725, 0.8396, 0.9341, 1.0026, 1.08, 1.0138, 1.117, 1.1949, 1.2937, 1.2939, 1.2071, 1.101, 1.0088, 0.965, 0.8982, 0.7659, 0.7747, 0.5994, 0.6791, 0.8282, 0.8202, 0.8202, 0.8775, 0.8766, 1.0215, 1.0949, 1.0904, 1.12, 1.3144, 1.2675, 1.2725, 1.2617, 1.2287, 1.2335, 1.4228, 1.3942, 1.4687, 1.6071, 1.5064, 1.5768, 1.5703, 1.6948, 1.865, 2.1818, 2.313, 2.2138, 2.7117, 2.9529, 2.81, 2.9941, 2.8568, 2.6016]}
{"unit_id": "quote:1990210382208843888", "ticker_idx": 0, "ts": "2025-11-17T00:08:26+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley has cut PC OEMs' price targets across the board… DELL OW → UW, TP $144 → $110", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses Morgan Stanley downgrade of DELL to Underweight, PT cut to $110, on memory-cost headwinds to general-purpose server demand.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "The part I’m especially worried about is this.\n\nTraditional, general-purpose servers are already suffering from weak demand, and memory now accounts for about 53% of their BOM cost.\n\nThis means that when memory prices rise, general-purpose server prices inevitably rise as well.\nAnd if that happens… demand for general-purpose servers will deteriorate even further.", "reply_text": null, "parent_text": null, "quoted_text": "This is a very important signal.\n\nMorgan Stanley has cut PC OEMs’ price targets across the board due to the sharp rise in memory prices.\n\nThey revised their targets as follows, projecting that OEMs’ EPS could face a headwind of up to -16%:\n\n• DELL\no OW → UW, TP $144 → $110.\no https://t.co/8bu5guIU8A", "parent_author": null, "ret_m1d": 0.09209663654308153, "ret_signed_m1d": -0.09209663654308153, "alpha_spy_m1d": -0.08269249873720086, "alpha_c_m1d": -0.07619615842934091, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.015900478113740624, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0017145849039588157, "ret_signed_p1d": -0.0017145849039588157, "alpha_spy_p1d": -0.010112045462165375, "alpha_c_p1d": -0.017967539062113458, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.016252954158154642, "ret_p1w": 0.038700128633091024, "ret_signed_p1w": -0.038700128633091024, "alpha_spy_p1w": -0.03410327350129627, "alpha_c_p1w": -0.052555770974019644, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.01385564234092862, "ret_p1m": 0.09201492688585988, "ret_signed_p1m": -0.09201492688585988, "alpha_spy_p1m": -0.07218522012337147, "alpha_c_p1m": -0.08679705148457728, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.005217875401282601, "ret_p3m": -0.07484027236803437, "ret_signed_p3m": 0.07484027236803437, "alpha_spy_p3m": 0.10129955325542261, "alpha_c_p3m": 0.057925629939175605, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": -0.01691464242885876, "ret_p6m": 0.9654463540642615, "ret_signed_p6m": -0.9654463540642615, "alpha_spy_p6m": -0.850204262801781, "alpha_c_p6m": -0.7266654658828902, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.23878088818137133, "price_path": [0.0453, 0.04, 0.0645, 0.0658, 0.0657, 0.078, 0.0906, -0.0062, -0.0062, -0.0159, 0.009, 0.0305, 0.0156, 0.0007, -0.0132, 0.0125, 0.02, 0.0173, 0.0316, 0.0388, 0.0604, 0.0748, 0.0734, 0.1039, 0.0929, 0.0746, 0.0654, 0.0638, 0.0894, 0.1534, 0.2177, 0.1989, 0.145, 0.1858, 0.2274, 0.3386, 0.2687, 0.225, 0.248, 0.2103, 0.2504, 0.231, 0.217, 0.203, 0.22, 0.2258, 0.2592, 0.2952, 0.3242, 0.3462, 0.3357, 0.3146, 0.3227, 0.3072, 0.2626, 0.2444, 0.218, 0.1977, 0.165, 0.1329, 0.1488, 0.0936, 0.0921, 0.0, 0.0017, -0.0253, -0.0415, 0.0002, 0.0387, 0.0281, 0.088, 0.088, 0.0887, 0.0785, 0.11, 0.091, 0.1348, 0.1341, 0.1464, 0.1285, 0.1482, 0.1316, 0.0612, 0.0656, 0.092, 0.0442, 0.0038, 0.0322, 0.0337, 0.042, 0.0482, 0.0482, 0.0552, 0.0407, 0.0444, 0.0278, 0.0278, 0.0434, 0.0125, 0.0118, -0.0197, -0.0325, -0.0152, -0.0164, -0.023, -0.0309, -0.023, -0.0159, -0.0159, -0.0892, -0.0712, -0.0392, -0.0534, -0.0493, -0.0598, -0.0379, -0.0283, -0.0616, -0.0229, -0.0393, 0.0008, -0.0538, -0.0074, -0.0085, 0.0333, 0.0182, -0.0748, -0.0365, -0.0365, -0.048, -0.0424, -0.0237, 0.0027, -0.023, -0.0178, 0.0126, -0.0041, 0.2143, 0.2592, 0.1905, 0.2063, 0.2015, 0.2012, 0.2014, 0.1792, 0.2085, 0.2293, 0.2433, 0.2837, 0.2547, 0.2236, 0.2855, 0.2929, 0.3497, 0.4507, 0.5089, 0.4418, 0.4089, 0.3503, 0.3459, 0.389, 0.4299, 0.4299, 0.4201, 0.4571, 0.5209, 0.488, 0.458, 0.5563, 0.513, 0.4538, 0.5834, 0.6118, 0.6748, 0.7468, 0.7656, 0.745, 0.7775, 0.7765, 0.6939, 0.6917, 0.7188, 0.7288, 0.7409, 0.7794, 0.9643, 0.8941, 1.1425, 1.0321, 0.9654]}
{"unit_id": "orig:1993239813152293249", "ticker_idx": 0, "ts": "2025-11-25T08:46:18+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price: USD 535.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares HSBC Buy-equivalent coverage on Broadcom with $535 PT, citing CoWoS capacity ramp and ASIC wins at Google, Meta, and Anthropic.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "HSBC states the following in its latest Broadcom coverage report:\n\n- Broadcom’s CoWoS capacity for 2026 has increased significantly to around 250,000 wafers.\n\n- Accordingly, Google’s TPU shipments in FY2026 are expected to rise from 1.7 million units to 3.1 million units.\n\n- Meta’s ASIC shipments are also revised up from 750,000 units to 900,000 units, driven by the ramp of its Athena and Iris chips.\n\n- The announced fourth customer is Anthropic, which is expected to contribute around USD 5 billion in ASIC revenue in FY2026.\n\n- Target price: USD 535.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018362260248894424, "ret_signed_m1d": -0.018362260248894424, "alpha_spy_m1d": -0.00904400715036524, "alpha_c_m1d": -0.01574460533061306, "bench_spy_m1d": -0.009318253098529183, "bench_c_m1d": -0.0026176549182813647, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032568851910707464, "ret_signed_p1d": 0.032568851910707464, "alpha_spy_p1d": 0.02566527689768794, "alpha_c_p1d": 0.009922544502187547, "bench_spy_p1d": 0.006903575013019525, "bench_c_p1d": 0.022646307408519917, "ret_p1w": -0.008986321968701971, "ret_signed_p1w": -0.008986321968701971, "alpha_spy_p1w": -0.01863057570504656, "alpha_c_p1w": -0.06610223058324827, "bench_spy_p1w": 0.009644253736344588, "bench_c_p1w": 0.057115908614546296, "ret_p1m": -0.0886681894276945, "ret_signed_p1m": -0.0886681894276945, "alpha_spy_p1m": -0.11444524627574937, "alpha_c_p1m": -0.16305761335994273, "bench_spy_p1m": 0.02577705684805487, "bench_c_p1m": 0.07438942393224823, "ret_p3m": -0.13438834257809973, "ret_signed_p3m": -0.13438834257809973, "alpha_spy_p3m": -0.15875391214577206, "alpha_c_p3m": -0.35882699722195677, "bench_spy_p3m": 0.02436556956767233, "bench_c_p3m": 0.22443865464385704, "ret_p6m": 0.08936275053200204, "ret_signed_p6m": 0.08936275053200204, "alpha_spy_p6m": -0.015005557986635498, "alpha_c_p6m": -0.5765204383504283, "bench_spy_p6m": 0.10436830851863754, "bench_c_p6m": 0.6658831888824304, "price_path": [-0.1997, -0.2289, -0.2289, -0.2267, -0.216, -0.2064, -0.1317, -0.1038, -0.1271, -0.0418, -0.0676, -0.0669, -0.056, -0.0666, -0.1025, -0.1046, -0.1057, -0.1201, -0.1197, -0.1187, -0.1271, -0.1312, -0.1484, -0.1432, -0.1341, -0.1217, -0.1212, -0.1287, -0.1263, -0.1027, -0.1039, -0.1569, -0.0736, -0.1062, -0.0875, -0.0802, -0.0927, -0.093, -0.11, -0.1162, -0.1058, -0.0803, -0.0597, -0.0313, 0.0025, -0.0222, -0.04, -0.0584, -0.0859, -0.0677, -0.0765, -0.0925, -0.0692, -0.0859, -0.0774, -0.117, -0.1106, -0.1101, -0.1157, -0.0795, -0.0992, -0.1164, -0.0184, 0.0, 0.0326, 0.0326, 0.0466, 0.0027, -0.009, -0.0115, -0.0104, 0.0135, 0.0417, 0.0552, 0.0726, 0.0554, -0.0652, -0.1174, -0.1136, -0.1533, -0.1432, -0.116, -0.1115, -0.091, -0.0887, -0.0887, -0.0837, -0.0908, -0.0896, -0.0994, -0.0994, -0.0954, -0.1064, -0.1055, -0.1062, -0.1348, -0.1023, -0.0835, -0.0772, -0.1155, -0.1074, -0.0848, -0.0848, -0.1345, -0.1444, -0.153, -0.1672, -0.1547, -0.134, -0.1329, -0.1394, -0.1379, -0.1384, -0.1664, -0.1984, -0.192, -0.1337, -0.105, -0.1141, -0.1081, -0.1382, -0.1539, -0.1539, -0.1347, -0.1322, -0.1309, -0.1344, -0.1404, -0.153, -0.1353, -0.1629, -0.1685, -0.1704, -0.1833, -0.1737, -0.1341, -0.14, -0.1003, -0.1085, -0.1112, -0.1257, -0.1617, -0.1545, -0.1639, -0.1779, -0.1677, -0.192, -0.159, -0.17, -0.1687, -0.1931, -0.2159, -0.2349, -0.1929, -0.1825, -0.1798, -0.1798, -0.1801, -0.1291, -0.0857, -0.0745, -0.0311, -0.0098, -0.0071, 0.0345, 0.0391, 0.0601, 0.0421, 0.0487, 0.1021, 0.095, 0.1024, 0.0905, 0.0426, 0.0573, 0.0885, 0.0985, 0.0861, 0.1144, 0.1094, 0.0758, 0.1213, 0.1172, 0.0934, 0.0868, 0.1468, 0.1087, 0.0971, 0.0719, 0.0894]}
{"unit_id": "orig:1983845390983295192", "ticker_idx": 0, "ts": "2025-10-30T10:36:13+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix is the biggest beneficiary of the sharp rebound in prices across commodity memory, underpinned by its HBM leadership, and the stock still has meaningful upside potential", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley Overweight on SK Hynix, PT raised to 630K KRW (+15% upside); 2026 DRAM ASP forecast lifted to +30% on AI-driven inventory tightness through 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: “Memory supply tighter, price upcycle to last through 2026”\n\nAccording to Morgan Stanley analysts Shawn Kim and Duan Liu in a report dated October 29,\nSK hynix’s “sold-out” signal suggests that supply will become even tighter and the upward price trend will continue through 2026.\n\nAccordingly, the Morgan Stanley analysts raised their 2026 DRAM price increase forecast from the previous 26% to 30%,\nand re-evaluated SK hynix’s earnings outlook to a level comparable to the 2017–2018 super-cycle.\n\nWith this adjustment, Morgan Stanley raised its 2025–2027 earnings per share (EPS) forecasts for SK hynix by 5%, 14%, and 15%, respectively.\nIn addition, based on the Residual Income Model, it raised its target price from 570,000 KRW to 630,000 KRW.\nThis implies a 2026E P/B of 2.6x and about 15% upside from the current level,\nand the firm maintained its “Overweight” rating and “Attractive” industry view.\n\nAI-driven memory tightness… DRAM inventory down to “two weeks”\n\nThe explosive growth in AI inference demand is the key driver of this memory market boom.\nMorgan Stanley assessed that the spread of AI servers is significantly stimulating demand for commodity memory (DRAM and NAND),\nrapidly drawing down industry inventories and supporting the ongoing price uptrend.\n\nSK hynix management stated that inventories are extremely tight.\n    • DRAM (DDR5) inventory is down to about two weeks, effectively at a “produce-and-ship” level.\n    • NAND inventory has also normalized to about 4–5 weeks.\n\nThis reflects the surge in demand from data center servers and eSSDs for storing AI-generated content.\n\nOn this basis, Morgan Stanley expects DRAM contract prices in 4Q25 to post a high-teen-digit QoQ increase,\nand NAND prices to rise by 10–15% QoQ.\n\nThe analysts said, “market participants are already prepared to reward SK hynix,” and\nthat the AI memory (HBM) business, which currently accounts for about 20% of total revenue,\nis expanding into DRAM and NAND as a whole, thereby securing strong pricing power.\n\n⸻\n\nHBM leadership intact… CAPEX seen at 35 trillion KRW in 2026\n\nAccording to Morgan Stanley, SK hynix’s 2026 capital expenditure (CAPEX)\nis expected to expand to around 35 trillion KRW, up 30% from about 27 trillion KRW in 2025.\n    • The share of WFE (wafer fab equipment) and infrastructure will rise from 55% to 60%,\n    • DRAM is expected to account for about 90% of total WFE spending.\n\nThe M15X fab has started up earlier than scheduled and equipment installation is under way.\nThis fab will play a core role in HBM production in 2026.\n\nBoth DRAM and NAND will be expanded through technology migration.\n    • 1c DDR5,\n    • Migration to 321-layer QLC (from 238-layer TLC).\n\nBased on this, Morgan Stanley estimates that total CAPEX in 2026 will increase by about 42% to 35 trillion KRW.\n\nThe report also stressed that “surging AI inference demand is lifting commodity memory demand, drawing down inventories, and sustaining price momentum.”\nOn the back of its solid leadership in HBM,\nSK hynix is expected to be the biggest beneficiary of the structural increase in commodity memory demand and the rise in prices across the board.\n\nMorgan Stanley concluded that\nthe recent unexpected spike in commodity DRAM and NAND demand has laid the groundwork for a severe supply shortage in 2026,\nand that the stronger market conditions could last through 2026.\n\n⸻\n\nTarget price raised, “Overweight” maintained\n\nThe Morgan Stanley analyst team revised its key assumptions as follows:\n    • 2026 DRAM blended ASP growth: +20% → +30%\n    • EPS revisions:\n    • +5% for 2025\n    • +14% for 2026\n    • +15% for 2027\n\nUnder the new valuation model, the target price is 630,000 KRW,\nthe 2026 P/B is 2.6x,\nand the investment rating is maintained at Overweight.\n\nThe report concluded,\n“SK hynix is the biggest beneficiary of the sharp rebound in prices across commodity memory, underpinned by its HBM leadership,\nand the stock still has meaningful upside potential.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017605599225796742, "ret_signed_m1d": -0.017605599225796742, "alpha_spy_m1d": -0.028726048457201725, "alpha_c_m1d": -0.031022075305803143, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.013416476080006401, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015844973101821203, "ret_signed_p1d": -0.015844973101821203, "alpha_spy_p1d": -0.019125253712536705, "alpha_c_p1d": -0.013892901123101598, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": -0.001952071978719605, "ret_p1w": 0.04401421873581146, "ret_signed_p1w": 0.04401421873581146, "alpha_spy_p1w": 0.05801766713870393, "alpha_c_p1w": 0.07876529126445075, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.03475107252863929, "ret_p1m": -0.0662330622349393, "ret_signed_p1m": -0.0662330622349393, "alpha_spy_p1m": -0.07146968306014345, "alpha_c_p1m": -0.03475365022319088, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.03147941201174842, "ret_p3m": 0.4094594953931001, "ret_signed_p3m": 0.4094594953931001, "alpha_spy_p3m": 0.38340135433796396, "alpha_c_p3m": 0.28632645774051513, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": 0.12313303765258499, "ret_p6m": 1.1569222441434661, "ret_signed_p6m": 1.1569222441434661, "alpha_spy_p6m": 1.1007682282420959, "alpha_c_p6m": 0.7602383931163348, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.39668385102713133, "price_path": [-0.5464, -0.5368, -0.5456, -0.5394, -0.5491, -0.5306, -0.5271, -0.5113, -0.5139, -0.5139, -0.5297, -0.5376, -0.5508, -0.5693, -0.5587, -0.5438, -0.5403, -0.5429, -0.5273, -0.5264, -0.5493, -0.5414, -0.5379, -0.5326, -0.5185, -0.5123, -0.493, -0.4648, -0.4595, -0.4217, -0.4173, -0.3873, -0.4129, -0.3741, -0.3741, -0.382, -0.3644, -0.3706, -0.3724, -0.4076, -0.3856, -0.3882, -0.3662, -0.3037, -0.3037, -0.3037, -0.3037, -0.3037, -0.3037, -0.2465, -0.2694, -0.2755, -0.2562, -0.2033, -0.1805, -0.1452, -0.1567, -0.1523, -0.1576, -0.1021, -0.0581, -0.0827, -0.0176, 0.0, -0.0158, 0.0915, 0.0317, 0.0194, 0.044, 0.0211, 0.0669, 0.0898, 0.0863, 0.0775, -0.0141, 0.0669, 0.0035, -0.0106, 0.0053, -0.0827, -0.0845, -0.0863, -0.0775, -0.0416, -0.0662, -0.0521, -0.0169, -0.0275, -0.0451, -0.0416, 0.0166, -0.0028, 0.0342, -0.0046, 0.006, -0.0239, -0.0662, -0.0292, -0.0275, -0.0363, 0.0219, 0.0289, 0.036, 0.036, 0.0553, 0.1276, 0.1469, 0.1469, 0.1469, 0.1928, 0.2262, 0.2791, 0.3073, 0.3319, 0.3108, 0.3196, 0.3002, 0.3073, 0.3196, 0.3319, 0.346, 0.309, 0.3038, 0.3302, 0.3513, 0.2967, 0.4095, 0.4817, 0.5169, 0.6015, 0.4623, 0.598, 0.5856, 0.4835, 0.4782, 0.5627, 0.5434, 0.5152, 0.5645, 0.5504, 0.5504, 0.5504, 0.5504, 0.5751, 0.672, 0.6755, 0.7706, 0.7935, 0.9398, 0.8727, 0.8727, 0.6574, 0.4985, 0.6609, 0.6309, 0.4756, 0.6556, 0.6856, 0.6415, 0.6062, 0.7192, 0.7121, 0.8639, 0.788, 0.7774, 0.6468, 0.7404, 0.7562, 0.6468, 0.6468, 0.5409, 0.4244, 0.5833, 0.465, 0.5462, 0.5639, 0.6168, 0.8233, 0.7615, 0.8127, 0.8357, 0.9469, 1.0051, 1.0387, 0.991, 1.0581, 1.1605, 1.1587, 1.1622, 1.1569]}
{"unit_id": "quote:1980791576646549831", "ticker_idx": 0, "ts": "2025-10-22T00:21:27+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Even if the adoption rate of ASICs increases, NVIDIA is still expected to capture the majority of overall profits in the AI semiconductor market.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish NVDA: even with rising ASIC adoption, NVIDIA expected to capture majority of AI semiconductor profits.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Even if the adoption rate of ASICs increases, NVIDIA is still expected to capture the majority of overall profits in the AI semiconductor market.\"\n\n👀 https://t.co/yBil2j400I", "reply_text": null, "parent_text": null, "quoted_text": "Concerns Over ASIC Market Share Limits… NVIDIA’s “Five Strategies” to Block the Counterattack\n\nGPU vs. ASIC: Not a Substitution, But a Coexistence\n\nGlobal cloud service providers (CSPs) such as Google, AWS, Meta, and Microsoft continue to purchase NVIDIA’s AI GPUs in large https://t.co/XfQEdfy8uk", "parent_author": null, "ret_m1d": 0.004881266256250294, "ret_signed_m1d": 0.004881266256250294, "alpha_spy_m1d": -0.0003447186403731095, "alpha_c_m1d": -0.014906040806114351, "bench_spy_m1d": 0.0052259848966234035, "bench_c_m1d": 0.019787307062364645, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010428209748937878, "ret_signed_p1d": 0.010428209748937878, "alpha_spy_p1d": 0.004498318558438186, "alpha_c_p1d": -0.007882444751792006, "bench_spy_p1d": 0.005929891190499692, "bench_c_p1d": 0.018310654500729884, "ret_p1w": 0.1484357812349617, "ret_signed_p1w": 0.1484357812349617, "alpha_spy_p1w": 0.11910068406087237, "alpha_c_p1w": 0.059806159655094504, "bench_spy_p1w": 0.02933509717408933, "bench_c_p1w": 0.0886296215798672, "ret_p1m": 0.0019968624119472977, "ret_signed_p1m": 0.0019968624119472977, "alpha_spy_p1m": 0.0248629706658966, "alpha_c_p1m": 0.041866933621806046, "bench_spy_p1m": -0.022866108253949302, "bench_c_p1m": -0.03987007120985875, "ret_p3m": 0.033061659031214496, "ret_signed_p3m": 0.033061659031214496, "alpha_spy_p3m": -0.0057280081979582675, "alpha_c_p3m": -0.15310451225361166, "bench_spy_p3m": 0.038789667229172764, "bench_c_p3m": 0.18616617128482615, "ret_p6m": 0.10035380609678746, "ret_signed_p6m": 0.10035380609678746, "alpha_spy_p6m": 0.04366738109235935, "alpha_c_p6m": -0.2470033718649347, "bench_spy_p6m": 0.05668642500442811, "bench_c_p6m": 0.34735717796172216, "price_path": [-0.0377, -0.0196, -0.0265, -0.0057, -0.0134, -0.0364, -0.0016, -0.0113, -0.0048, 0.0027, 0.0134, 0.0098, 0.0159, 0.0072, 0.0096, 0.0009, 0.0095, -0.0258, -0.0271, -0.0295, -0.0128, -0.0027, 0.0082, 0.0073, -0.0007, -0.0339, -0.0339, -0.0527, -0.0536, -0.0479, -0.0736, -0.0664, -0.0529, -0.0164, -0.0173, -0.0136, -0.014, -0.03, -0.0554, -0.0224, -0.02, 0.0185, -0.0103, -0.0184, -0.0144, -0.0116, 0.0087, 0.0349, 0.0386, 0.0478, 0.0407, 0.0292, 0.0264, 0.049, 0.0682, 0.016, 0.0446, -0.0014, -0.0025, 0.0085, 0.0163, 0.0131, 0.0049, 0.0, 0.0104, 0.0332, 0.0622, 0.1151, 0.1484, 0.1254, 0.1232, 0.1475, 0.1021, 0.0828, 0.0433, 0.0437, 0.1041, 0.0714, 0.075, 0.0365, 0.0549, 0.0351, 0.006, 0.0346, 0.002, -0.0078, 0.0126, -0.0136, -0.0001, -0.0001, -0.0182, -0.002, 0.0065, -0.0038, 0.0173, 0.0119, 0.0293, 0.0261, 0.0195, 0.0037, -0.0291, -0.0221, -0.0141, -0.0518, -0.034, 0.004, 0.019, 0.0496, 0.0463, 0.0463, 0.0569, 0.0441, 0.0403, 0.0346, 0.0346, 0.0476, 0.0435, 0.0387, 0.049, 0.0265, 0.0255, 0.0259, 0.0307, 0.0159, 0.0376, 0.0331, 0.0331, -0.0122, 0.0169, 0.0254, 0.041, 0.0344, 0.0458, 0.0624, 0.0679, 0.0602, 0.0296, 0.0004, -0.0337, -0.0465, 0.0285, 0.0542, 0.0459, 0.0543, 0.037, 0.0141, 0.0141, 0.0261, 0.0428, 0.0423, 0.053, 0.0626, 0.0698, 0.0848, 0.0256, -0.0171, 0.0123, -0.0012, 0.0154, 0.017, -0.0136, 0.0132, 0.025, 0.032, 0.016, -0.0001, 0.0164, 0.0093, 0.0008, -0.0094, -0.0419, -0.0256, -0.0281, -0.0088, -0.05, -0.0707, -0.0837, -0.0325, -0.025, -0.0159, -0.0159, -0.0145, -0.012, 0.0101, 0.0202, 0.0464, 0.0502, 0.0901, 0.1032, 0.1004]}
{"unit_id": "orig:1988120981907583379", "ticker_idx": 0, "ts": "2025-11-11T05:45:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi: Maintains Buy on NVIDIA, raises target price from $210 to $220", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates Buy on NVDA, raises PT to $220; expects beat-and-raise on Oct-quarter $57B vs $55B consensus, Jan-quarter guide $62B vs $61B, on Blackwell ramp and CoWoS supply constraints through 2026.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi: Maintains Buy on NVIDIA, raises target price from $210 to $220\n\nCiti presented a “30-day short-term bullish” view on NVIDIA, betting that the earnings scheduled to be announced on November 19 will deliver a “beat and raise” (revenue surprise and guidance increase).\n\nAnalysts Atif Malik and Papa Sylla pointed out in the report that, despite questions about the funding sources for AI investment, a more fundamental fact is that due to advanced packaging (CoWoS) capacity constraints, AI chip supply will continue to fall short of demand until before 2026.\n\nCiti indicated that NVIDIA’s stock remains attractive at current levels. The report stated that NVIDIA’s current price-to-earnings ratio (PER) is about 28x, which is more appealing in valuation compared with AI peers Broadcom at 38x and AMD at 37x.\n\nEarnings surprise imminent?\nCiti expects NVIDIA’s upcoming results to easily surpass Wall Street consensus. The report forecasts that the company’s October-quarter sales will reach $57 billion, exceeding the market average expectation of about $55 billion.\n\nFor the outlook, Citi expects NVIDIA to guide January-quarter sales to $62 billion, also above the market expectation of about $61 billion.\n\nThe report cited the strong shipment momentum of NVIDIA’s Blackwell-architecture GPUs as the backdrop for this optimistic view. Citi’s analysts believe that NVIDIA’s disclosure at the GTC Washington conference that it has already shipped 6 million GPUs is a strong signal that October- and January-quarter results may beat expectations.\n\nIn the report, Citi stated that its model assumes NVIDIA’s data-center sales will grow 24% and 12% quarter-over-quarter in the October and January quarters, respectively—more aggressive than the market’s expectations (19% and 15%, respectively).\n\nRebuttal to the AI bubble narrative\nIn response to the spreading “AI bubble” argument, Citi presents the opposite view. The core thesis of the report is that the main contradiction in today’s AI chip market is “undersupply,” not “insufficient demand.”\n\nCiti explained as follows:\n“While there are concerns about debt and circular financing structures surrounding an AI CapEx bubble, we fundamentally see that due to CoWoS capacity constraints, AI supply will fall short of demand until before 2026, and may only catch up at some point in 2027.”\n\nSo the report wrote. This judgment means that robust demand is real, not a product of speculative froth.\n\nTo further substantiate the strength on the demand side, Citi pointed out that hyperscalers’ cloud revenue is reaching an inflection point starting in 2025 and is expected to accelerate further in 2026, with the main driver being the proliferation of enterprise AI applications.\n\nTarget raised to $220\nBased on strong confidence in NVIDIA’s growth prospects, Citi not only raised its target price but also broadly lifted its forward financial forecasts. The new $220 target price is derived by applying a 30x PER to an EPS estimate of $7.24 for the 2026 calendar year.\n\nAccording to the report, Citi raised its EPS forecasts for NVIDIA’s FY2026, FY2027, and FY2028 by 2%, 7%, and 8%, respectively, to better reflect the bank’s revised global AI CapEx model. Notably, Citi continues to assume zero sales to Chinese data centers in its model, implying additional upside could materialize if there is any policy easing.\n\nAI chip market not at the peak yet?\nCiti’s bullish stance extends beyond NVIDIA itself to the broader AI semiconductor market. The report sharply raised its forecast for future market size, arguing that the AI wave has not yet peaked.\n\nAccording to Citi’s latest model, by 2028 the total addressable market (TAM) for global data-center semiconductors is expected to reach $654 billion, 16% higher than the previous projection of $563 billion. The report explains that the main reason for the upward revision is “higher-than-expected demand from key AI players such as OpenAI.”\n\nRisks cannot be overlooked\nAlthough the overall tone is very optimistic, the original report also clearly identifies the risks and market focal points that investors should watch—factors that should not be ignored in investment decisions. The main risks listed by Citi are as follows:\n\nIntensifying competition: In the gaming and AI GPU markets, competition from rivals such as AMD could lead to NVIDIA losing market share.\n\nSlow adoption of new platforms: If the adoption pace of new platforms in data centers and gaming falls short of expectations, sales will be affected.\n\nMarket volatility: Demand in the automotive and data-center markets may experience cyclical fluctuations.\n\nCryptocurrency impact: Changes in cryptocurrency mining demand may interfere with sales of gaming graphics cards.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030492882347475936, "ret_signed_m1d": 0.030492882347475936, "alpha_spy_m1d": 0.03277680381178116, "alpha_c_m1d": 0.00844470525187968, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003313360000960719, "ret_signed_p1d": 0.003313360000960719, "alpha_spy_p1d": 0.002756961394518642, "alpha_c_p1d": -0.00944833118887245, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.061089260286197655, "ret_signed_p1w": -0.061089260286197655, "alpha_spy_p1w": -0.02753153112227391, "alpha_c_p1w": -0.010498217267669863, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": -0.04850783058703967, "ret_signed_p1m": -0.04850783058703967, "alpha_spy_p1m": -0.055198993411596176, "alpha_c_p1m": -0.11416803762039751, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": -0.04006875267429444, "ret_signed_p3m": -0.04006875267429444, "alpha_spy_p3m": -0.05421328419987903, "alpha_c_p3m": -0.1877678549390902, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.07606554636507679, "ret_signed_p6m": 0.07606554636507679, "alpha_spy_p6m": -0.004473938298911406, "alpha_c_p6m": -0.4948521164189732, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.0577, -0.0659, -0.0578, -0.0908, -0.092, -0.0942, -0.0786, -0.0692, -0.059, -0.0599, -0.0673, -0.0983, -0.0983, -0.1159, -0.1167, -0.1114, -0.1354, -0.1287, -0.116, -0.082, -0.0828, -0.0794, -0.0798, -0.0946, -0.1184, -0.0876, -0.0854, -0.0494, -0.0763, -0.0838, -0.0801, -0.0775, -0.0586, -0.0341, -0.0306, -0.0221, -0.0287, -0.0394, -0.042, -0.021, -0.0031, -0.0518, -0.0251, -0.068, -0.069, -0.0588, -0.0515, -0.0545, -0.0621, -0.0667, -0.0569, -0.0357, -0.0086, 0.0407, 0.0719, 0.0504, 0.0483, 0.071, 0.0286, 0.0106, -0.0263, -0.0259, 0.0305, 0.0, 0.0033, -0.0326, -0.0155, -0.034, -0.0611, -0.0344, -0.0648, -0.0739, -0.0549, -0.0794, -0.0668, -0.0668, -0.0837, -0.0685, -0.0606, -0.0703, -0.0506, -0.0556, -0.0393, -0.0423, -0.0485, -0.0633, -0.0939, -0.0873, -0.0799, -0.115, -0.0984, -0.063, -0.049, -0.0204, -0.0235, -0.0235, -0.0136, -0.0255, -0.029, -0.0344, -0.0344, -0.0223, -0.026, -0.0306, -0.0209, -0.042, -0.0429, -0.0425, -0.038, -0.0518, -0.0316, -0.0358, -0.0358, -0.0781, -0.0509, -0.043, -0.0284, -0.0346, -0.024, -0.0084, -0.0033, -0.0105, -0.039, -0.0663, -0.0982, -0.1101, -0.0401, -0.0161, -0.0239, -0.016, -0.0321, -0.0535, -0.0535, -0.0423, -0.0268, -0.0272, -0.0172, -0.0083, -0.0015, 0.0125, -0.0428, -0.0826, -0.0552, -0.0678, -0.0523, -0.0508, -0.0794, -0.0544, -0.0434, -0.0368, -0.0518, -0.0667, -0.0514, -0.058, -0.066, -0.0755, -0.1058, -0.0906, -0.0929, -0.0749, -0.1134, -0.1326, -0.1448, -0.097, -0.09, -0.0815, -0.0815, -0.0802, -0.0779, -0.0573, -0.0478, -0.0233, -0.0198, 0.0175, 0.0297, 0.027, 0.0442, 0.0462, 0.0349, 0.0485, 0.0337, 0.0783, 0.1215, 0.1037, 0.0834, 0.0333, 0.0275, 0.0277, 0.0174, 0.0761]}
{"unit_id": "quote:1975917764180344979", "ticker_idx": 0, "ts": "2025-10-08T13:34:40+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS Raises $MU PT to $225 from $195 - Buy", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares UBS Buy reiteration on MU with PT raised to $225, driven by HBM demand expansion and higher Rubin unit volumes.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Higher NVIDIA (Rubin) unit volumes into next year, now expected at 7.4 million units versus 7.0 million prior.\"\n\n👀", "reply_text": null, "parent_text": null, "quoted_text": "UBS Raises $MU PT to $225 from $195 - Buy\n\nAnalyst comments: \"Following recent announcements across the compute space and updated CoWoS supply checks, we have increased confidence that high-bandwidth memory (HBM) demand will expand significantly again in calendar 2027.\n\nIn the", "parent_author": null, "ret_m1d": -0.055204972681472175, "ret_signed_m1d": -0.055204972681472175, "alpha_spy_m1d": -0.04927726748545125, "alpha_c_m1d": -0.029140840538495638, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.026064132142976537, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021420551013971667, "ret_signed_p1d": -0.021420551013971667, "alpha_spy_p1d": -0.018523542634734214, "alpha_c_p1d": -0.018935422070154084, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0024851289438175828, "ret_p1w": -0.02340488073003566, "ret_signed_p1w": -0.02340488073003566, "alpha_spy_p1w": -0.011608927951418102, "alpha_c_p1w": -0.011008515489854598, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.012396365240181062, "ret_p1m": 0.21262856709063205, "ret_signed_p1m": 0.21262856709063205, "alpha_spy_p1m": 0.21678832021386663, "alpha_c_p1m": 0.1981228409622089, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": 0.014505726128423158, "ret_p3m": 0.5888679864399216, "ret_signed_p3m": 0.5888679864399216, "alpha_spy_p3m": 0.5641437416405248, "alpha_c_p3m": 0.49439319217912403, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": 0.09447479426079752, "ret_p6m": 0.8649734258483615, "ret_signed_p6m": 0.8649734258483615, "alpha_spy_p6m": 0.885097456983359, "alpha_c_p6m": 0.7278025170186437, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": 0.13717090882971772, "price_path": [-0.3668, -0.3969, -0.3893, -0.408, -0.4241, -0.4183, -0.4242, -0.4446, -0.4415, -0.4319, -0.4343, -0.4343, -0.4307, -0.4166, -0.445, -0.4667, -0.452, -0.4454, -0.4469, -0.4312, -0.3955, -0.3709, -0.3504, -0.3681, -0.3629, -0.3854, -0.3718, -0.3794, -0.404, -0.4112, -0.4016, -0.408, -0.4076, -0.4013, -0.3796, -0.3949, -0.3949, -0.3975, -0.3963, -0.3684, -0.332, -0.3315, -0.3123, -0.2881, -0.2344, -0.2005, -0.1978, -0.1924, -0.1865, -0.1412, -0.1725, -0.1629, -0.1538, -0.1777, -0.2025, -0.2003, -0.1666, -0.1492, -0.0738, -0.0657, -0.0443, -0.0284, -0.0552, 0.0, -0.0214, -0.076, -0.0192, -0.0482, -0.0234, 0.0305, 0.0297, 0.0521, 0.0293, 0.0098, 0.0517, 0.1144, 0.1199, 0.1291, 0.1531, 0.1398, 0.1385, 0.1942, 0.1093, 0.2084, 0.2126, 0.2105, 0.2888, 0.2268, 0.2461, 0.2056, 0.2559, 0.231, 0.1626, 0.1495, 0.0246, 0.0551, 0.1394, 0.1424, 0.1716, 0.1716, 0.2032, 0.2235, 0.2185, 0.1914, 0.1532, 0.207, 0.2563, 0.2843, 0.3418, 0.3151, 0.2269, 0.2084, 0.183, 0.1475, 0.2646, 0.353, 0.4073, 0.4057, 0.4586, 0.4586, 0.449, 0.4984, 0.4895, 0.4528, 0.4528, 0.6055, 0.5889, 0.7481, 0.7283, 0.6646, 0.7565, 0.7605, 0.7211, 0.6968, 0.7135, 0.8464, 0.8464, 0.8579, 0.9806, 1.0237, 1.0342, 0.9805, 1.0882, 1.2156, 1.2182, 1.1118, 1.2284, 1.135, 0.9312, 0.9489, 1.009, 0.952, 0.8999, 1.0887, 1.1071, 1.0954, 1.0954, 1.0349, 1.1427, 1.1243, 1.1794, 1.1428, 1.1277, 1.1836, 1.1152, 1.099, 1.1005, 0.9326, 1.04, 1.021, 0.8849, 0.9817, 1.0519, 1.1312, 1.0633, 1.169, 1.2488, 1.35, 1.3502, 1.2614, 1.1526, 1.0582, 1.0133, 0.9449, 0.8093, 0.8183, 0.6387, 0.7204, 0.8732, 0.865]}
{"unit_id": "quote:1985294373559599266", "ticker_idx": 0, "ts": "2025-11-03T10:33:58+00:00", "symbol": "Vertiv", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "The bottlenecks in power consumption and cooling are driving an explosive increase in the value of power and liquid cooling solutions... power and cooling challenges brought by AI hardware upgrades are transforming into massive business opportunities for power and cooling suppliers", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish power/cooling suppliers as AI hardware upgrades drive explosive value growth; per-cabinet power solution value to 10x by 2027.", "resolved_tickers": ["VRT"], "resolution_method": "exchange_ticker", "resolution_reason": "Vertiv Holdings listed on NYSE as VRT", "bench_c": "XLI", "bench_c_reason": "industry='Electrical Equipment & Parts'", "bench_c_industry": "Electrical Equipment & Parts", "unit_kind": "quote", "lang": "en", "tweet_text": "The bottlenecks in power consumption and cooling are driving an explosive increase in the value of power and liquid cooling solutions.\nOne of the most notable insights from the report is that the power and cooling challenges brought by AI hardware upgrades are transforming into massive business opportunities for power and cooling suppliers. This segment is viewed as one of the fastest-growing value drivers in the current upgrade cycle.\n\nIn the power domain, as the total power consumption per server cabinet rises sharply, traditional power architectures are becoming unsustainable.\nThe report predicts that power systems will transition to 800V high-voltage direct current (HVDC) architectures, a shift expected to significantly enhance the value content of power solutions.\n\nMorgan Stanley forecasts that by 2027, the power solution for Rubin Ultra cabinets adopting the Kyber architecture will have a per-cabinet value more than ten times that of the current GB200 server cabinets.\nIt also projects that by 2027, the value of power solutions per watt of consumption within AI server cabinets will double compared with current levels.", "reply_text": null, "parent_text": null, "quoted_text": "Morgan Stanley stated in its latest report that the total value of cooling (thermal) components in a single GB300 (NVL72) server cabinet amounts to approximately $49,860.\n\nFor the next-generation Vera Rubin (NVL144) platform, as the cooling demand for both the computing trays and https://t.co/668Mb01WE5", "parent_author": null, "ret_m1d": 0.0076280756899749225, "ret_signed_m1d": 0.0076280756899749225, "alpha_spy_m1d": 0.009501247420738657, "alpha_c_m1d": 0.0033533980683488895, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": 0.004274677621626033, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.055276805616369606, "ret_signed_p1d": -0.055276805616369606, "alpha_spy_p1d": -0.043423265673922073, "alpha_c_p1d": -0.04368358475825784, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.011593220858111764, "ret_p1w": -0.018599741367351652, "ret_signed_p1w": -0.018599741367351652, "alpha_spy_p1w": -0.01581931243231438, "alpha_c_p1w": -0.017304390539191306, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.0012953508281603465, "ret_p1m": -0.05445389080520335, "ret_signed_p1m": -0.05445389080520335, "alpha_spy_p1m": -0.05180514890595467, "alpha_c_p1m": -0.043832133346469626, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.010621757458733727, "ret_p3m": 0.019711745290156735, "ret_signed_p3m": 0.019711745290156735, "alpha_spy_p3m": 0.0010522492490767643, "alpha_c_p3m": -0.058190390984608875, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.07790213627476561, "ret_p6m": 0.5946523938756185, "ret_signed_p6m": 0.5946523938756185, "alpha_spy_p6m": 0.5472348206628814, "alpha_c_p6m": 0.4803538428566858, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.11429855101893271, "price_path": [-0.2701, -0.2719, -0.2691, -0.2696, -0.2493, -0.2823, -0.3078, -0.305, -0.2913, -0.3259, -0.3338, -0.3388, -0.342, -0.347, -0.3318, -0.3246, -0.2989, -0.3338, -0.3338, -0.3523, -0.344, -0.3434, -0.3523, -0.3637, -0.3441, -0.2858, -0.2912, -0.2957, -0.2776, -0.2861, -0.2851, -0.2583, -0.2497, -0.2061, -0.2549, -0.2601, -0.2697, -0.2758, -0.2513, -0.2118, -0.1557, -0.1555, -0.163, -0.1494, -0.17, -0.1255, -0.1177, -0.117, -0.0648, -0.0849, -0.0569, -0.071, -0.0909, -0.0819, -0.0867, -0.1035, -0.0428, -0.0279, 0.0078, -0.0043, 0.0411, 0.0123, 0.0076, 0.0, -0.0553, -0.0036, -0.0438, -0.0606, -0.0186, -0.0645, -0.0942, -0.145, -0.1067, -0.1293, -0.1387, -0.1084, -0.1661, -0.1649, -0.1175, -0.1137, -0.1009, -0.1009, -0.0606, -0.0633, -0.0545, -0.0651, -0.0459, -0.0121, -0.0299, -0.0677, -0.0497, -0.0662, -0.1571, -0.1546, -0.1603, -0.2169, -0.1931, -0.1647, -0.1311, -0.131, -0.1278, -0.1278, -0.1241, -0.1344, -0.1411, -0.1532, -0.1532, -0.0822, -0.0908, -0.0856, -0.1034, -0.1597, -0.145, -0.1113, -0.0973, -0.107, -0.0982, -0.0753, -0.0753, -0.0844, -0.0515, -0.0534, -0.0462, -0.0528, -0.0111, 0.0127, 0.0197, -0.0269, -0.0069, -0.0062, -0.0458, -0.071, 0.0222, 0.0558, 0.0433, 0.2989, 0.2361, 0.2258, 0.2258, 0.2728, 0.2712, 0.2704, 0.274, 0.2827, 0.3231, 0.3704, 0.3549, 0.3322, 0.3472, 0.2776, 0.3133, 0.3053, 0.2637, 0.3817, 0.4115, 0.4021, 0.387, 0.3531, 0.3837, 0.4032, 0.3839, 0.4072, 0.3377, 0.3383, 0.4162, 0.4437, 0.3195, 0.3126, 0.2245, 0.31, 0.3559, 0.366, 0.366, 0.3526, 0.3713, 0.4692, 0.5037, 0.5428, 0.5681, 0.6233, 0.5744, 0.5377, 0.6067, 0.6437, 0.6334, 0.5952, 0.6821, 0.691, 0.6856, 0.5947]}
{"unit_id": "thread:1982943867382182149", "ticker_idx": 0, "ts": "2025-10-27T22:53:53+00:00", "symbol": "034220.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "LG Display's Q3 operating-profit consensus stood at ₩440.4 billion—almost double the ₩223.4 billion projected six months ago. Analysts expect LG Display to post an 'earnings surprise' for Q3 and return to annual profitability for the first time in four years", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Korean display makers dominate iPhone 17 OLED supply at 98%; LG Display expected earnings surprise and first annual profit in 4 years; Samsung Display operating margins above 15% in H2.", "resolved_tickers": ["034220.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Korean Display Makers Dominate iPhone 17 OLED Panel Supply with 98% Market Share as China’s BOE Struggles with Quality Issues\n\nAs Apple’s iPhone 17 series continues to exceed sales expectations, Korean display manufacturers Samsung Display and LG Display have virtually dominated the OLED panel supply chain for the devices. China’s largest display panel maker, BOE, has reportedly faced difficulties delivering panels due to quality problems. With Samsung and LG responsible for more than 98% of iPhone 17 panel shipments, their earnings outlooks are improving significantly.\n\nAccording to display market research firm UBI Research on the 28th, the total number of OLED panels supplied for the iPhone 17 series so far this year is estimated at approximately 88.9 million units. Apple’s panel procurement network consists of three players: Samsung Display, LG Display, and BOE. Of these, Samsung Display shipped about 57.3 million units, accounting for 64.5% of the total, followed by LG Display with 30.3 million units (34.1%). In contrast, BOE supplied only around 1.3 million units (1.4%).\n\nIndustry insiders predict that the strong sales momentum of the iPhone 17 will further boost panel shipments from Korean display companies. Apple enhanced key specifications such as the camera, display, and battery in the iPhone 17 series while keeping prices unchanged. The base model, in particular, has earned positive reviews for its improved chipset performance, increased brightness, and upgraded front-camera quality, offering strong value for money. In China—one of Apple’s largest markets—sales of the base model are estimated to have nearly doubled compared with the previous generation, according to Counterpoint Research. Reflecting this trend, Japan’s Mizuho Securities recently raised its annual iPhone 17 shipment forecast by 6.8%, from 88 million to 94 million units.\n\nApple restructured its panel supply chain this year to align with the four iPhone 17 models (Air, Standard, Pro, and Pro Max). Samsung Display continues to supply panels for all models, maintaining its position as the largest supplier by volume. However, roughly 60% of its shipments are for the lower-priced Air and Standard models. LG Display, on the other hand, provides panels for the Air, Standard, and Pro Max models—of which the high-end Pro Max accounts for about 70% of LG’s total supply. Panel pricing varies considerably: about $40 for the Standard model and $60–70 for the Pro Max, roughly 1.5 times higher. Industry observers note that “Apple had initially planned to assign BOE to produce panels for the Pro models and reduce LG Display’s share, but BOE’s quality issues have disrupted those plans.”\n\nDespite Apple’s strategic “support” for BOE, the Chinese manufacturer is still struggling to overcome technological limitations. According to Kim Jun-ho, a researcher at UBI Research, “BOE has faced production difficulties for the Pro models, making even its year-end certification target for supplying base-model panels uncertain. Its total shipments this year are now expected to fall more than 1 million units below the already-revised forecast of 3 million.” Consequently, total iPhone 17 panel supplies are expected to reach about 83–88 million units from Samsung Display, 46 million from LG Display, and 2 million from BOE.\n\nBuoyed by the iPhone 17 boom, both Samsung Display and LG Display are seeing improving financial results. As of October 26, LG Display’s Q3 operating-profit consensus stood at ₩440.4 billion—almost double the ₩223.4 billion projected six months ago. Analysts expect LG Display to post an “earnings surprise” for Q3 and return to annual profitability for the first time in four years. Daishin Securities’ Park Kang-ho noted, “LG Display’s share of iPhone panel supply has steadily risen over the past two to three years, and this structural shift is translating into profit growth,” projecting an annual profit of about ₩859 billion for 2024.\n\nSamsung Display is also expected to continue its strong earnings trajectory. Brokerages estimate its Q3 operating profit at ₩1.2–1.4 trillion, more than double the previous quarter’s ₩500 billion. Eugene Investment & Securities’ Lim So-jeong explained, “Driven by robust sales from key clients such as Apple, Samsung Display’s Q3 operating margin will improve sharply from 7.8% in the prior quarter.” KB Securities forecasts Samsung Display’s H2 operating profit at ₩2.5 trillion, slightly above last year’s ₩2.4 trillion, while Hyundai Motor Securities projects ₩2.7 trillion. Both expect Samsung Display to maintain a stable operating margin above 15% in the second half of the year.\n\n---\n\nhttps://t.co/yK6dccAZ3q", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0330056179775281, "ret_signed_m1d": -0.0330056179775281, "alpha_spy_m1d": -0.021345469033764664, "alpha_c_m1d": -0.014878564952881312, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.01812705302464679, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004915730337078594, "ret_signed_p1d": 0.004915730337078594, "alpha_spy_p1d": 0.0022597313804875174, "alpha_c_p1d": -0.004983923165837512, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.009899653502916106, "ret_p1w": 0.0196629213483146, "ret_signed_p1w": 0.0196629213483146, "alpha_spy_p1w": 0.022435640933588186, "alpha_c_p1w": 0.009930401130867406, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.009732520217447194, "ret_p1m": -0.1411516853932584, "ret_signed_p1m": -0.1411516853932584, "alpha_spy_p1m": -0.1262371539168491, "alpha_c_p1m": -0.07914503178953014, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.06200665360372826, "ret_p3m": -0.1109550561797753, "ret_signed_p3m": -0.1109550561797753, "alpha_spy_p3m": -0.1193840233912219, "alpha_c_p3m": -0.0815219941864268, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": -0.029433061993348497, "ret_p6m": 0.03721910112359561, "ret_signed_p6m": 0.03721910112359561, "alpha_spy_p6m": 0.0038745346614543763, "alpha_c_p6m": -0.0003545698060103408, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.03757367092960595, "price_path": [-0.2612, -0.2388, -0.2584, -0.2633, -0.243, -0.231, -0.2275, -0.2324, -0.224, -0.2381, -0.0667, -0.1657, -0.1657, -0.1615, -0.1875, -0.2008, -0.1973, -0.1994, -0.1812, -0.1742, -0.1699, -0.1594, -0.1573, -0.1812, -0.1756, -0.1573, -0.1531, -0.158, -0.1636, -0.1678, -0.1826, -0.1749, -0.1503, -0.1538, -0.0653, -0.0955, -0.0787, -0.0787, -0.0632, -0.0646, -0.0414, 0.0042, -0.0147, 0.0274, 0.0119, 0.0133, 0.0295, 0.0295, 0.0295, 0.0295, 0.0295, 0.0295, 0.0969, 0.0667, -0.0147, 0.0176, 0.0098, 0.0091, 0.0204, -0.0105, -0.0035, -0.0063, -0.033, 0.0, 0.0049, 0.0183, -0.0204, 0.0323, 0.0197, 0.0232, -0.0162, -0.014, -0.0583, -0.0358, -0.0527, -0.0449, -0.0548, -0.0737, -0.0934, -0.1482, -0.1306, -0.1222, -0.1468, -0.1419, -0.1412, -0.1138, -0.1081, -0.1194, -0.1341, -0.1215, -0.1018, -0.0976, -0.0681, -0.0913, -0.0983, -0.1096, -0.1236, -0.1208, -0.1362, -0.1538, -0.1461, -0.165, -0.1594, -0.1735, -0.1629, -0.1538, -0.1538, -0.1735, -0.1573, -0.1706, -0.1706, -0.1706, -0.1692, -0.1559, -0.1355, -0.1194, -0.1412, -0.1728, -0.1587, -0.1482, -0.1538, -0.1404, -0.1208, -0.125, -0.1243, -0.144, -0.111, -0.1201, -0.1264, -0.1369, -0.1482, -0.1706, -0.1805, -0.2247, -0.2015, -0.1889, -0.1959, -0.2135, -0.1945, -0.1805, -0.1636, -0.1636, -0.1854, -0.1854, -0.1854, -0.1854, -0.1376, -0.0653, -0.0197, -0.0126, 0.0407, 0.1131, 0.007, 0.007, -0.0969, -0.2275, -0.1833, -0.1601, -0.2044, -0.1587, -0.1538, -0.1671, -0.2022, -0.1973, -0.1868, -0.1468, -0.177, -0.1552, -0.2142, -0.1959, -0.1749, -0.1903, -0.1903, -0.2177, -0.2381, -0.2015, -0.2149, -0.2374, -0.231, -0.2086, -0.1496, -0.1713, -0.111, -0.1306, -0.0962, -0.0449, -0.0288, 0.0119, 0.0162, 0.0372]}
{"unit_id": "thread:1982943867382182149", "ticker_idx": 1, "ts": "2025-10-27T22:53:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Display is also expected to continue its strong earnings trajectory. Brokerages estimate its Q3 operating profit at ₩1.2–1.4 trillion, more than double the previous quarter's ₩500 billion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Korean display makers dominate iPhone 17 OLED supply at 98%; LG Display expected earnings surprise and first annual profit in 4 years; Samsung Display operating margins above 15% in H2.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Korean Display Makers Dominate iPhone 17 OLED Panel Supply with 98% Market Share as China’s BOE Struggles with Quality Issues\n\nAs Apple’s iPhone 17 series continues to exceed sales expectations, Korean display manufacturers Samsung Display and LG Display have virtually dominated the OLED panel supply chain for the devices. China’s largest display panel maker, BOE, has reportedly faced difficulties delivering panels due to quality problems. With Samsung and LG responsible for more than 98% of iPhone 17 panel shipments, their earnings outlooks are improving significantly.\n\nAccording to display market research firm UBI Research on the 28th, the total number of OLED panels supplied for the iPhone 17 series so far this year is estimated at approximately 88.9 million units. Apple’s panel procurement network consists of three players: Samsung Display, LG Display, and BOE. Of these, Samsung Display shipped about 57.3 million units, accounting for 64.5% of the total, followed by LG Display with 30.3 million units (34.1%). In contrast, BOE supplied only around 1.3 million units (1.4%).\n\nIndustry insiders predict that the strong sales momentum of the iPhone 17 will further boost panel shipments from Korean display companies. Apple enhanced key specifications such as the camera, display, and battery in the iPhone 17 series while keeping prices unchanged. The base model, in particular, has earned positive reviews for its improved chipset performance, increased brightness, and upgraded front-camera quality, offering strong value for money. In China—one of Apple’s largest markets—sales of the base model are estimated to have nearly doubled compared with the previous generation, according to Counterpoint Research. Reflecting this trend, Japan’s Mizuho Securities recently raised its annual iPhone 17 shipment forecast by 6.8%, from 88 million to 94 million units.\n\nApple restructured its panel supply chain this year to align with the four iPhone 17 models (Air, Standard, Pro, and Pro Max). Samsung Display continues to supply panels for all models, maintaining its position as the largest supplier by volume. However, roughly 60% of its shipments are for the lower-priced Air and Standard models. LG Display, on the other hand, provides panels for the Air, Standard, and Pro Max models—of which the high-end Pro Max accounts for about 70% of LG’s total supply. Panel pricing varies considerably: about $40 for the Standard model and $60–70 for the Pro Max, roughly 1.5 times higher. Industry observers note that “Apple had initially planned to assign BOE to produce panels for the Pro models and reduce LG Display’s share, but BOE’s quality issues have disrupted those plans.”\n\nDespite Apple’s strategic “support” for BOE, the Chinese manufacturer is still struggling to overcome technological limitations. According to Kim Jun-ho, a researcher at UBI Research, “BOE has faced production difficulties for the Pro models, making even its year-end certification target for supplying base-model panels uncertain. Its total shipments this year are now expected to fall more than 1 million units below the already-revised forecast of 3 million.” Consequently, total iPhone 17 panel supplies are expected to reach about 83–88 million units from Samsung Display, 46 million from LG Display, and 2 million from BOE.\n\nBuoyed by the iPhone 17 boom, both Samsung Display and LG Display are seeing improving financial results. As of October 26, LG Display’s Q3 operating-profit consensus stood at ₩440.4 billion—almost double the ₩223.4 billion projected six months ago. Analysts expect LG Display to post an “earnings surprise” for Q3 and return to annual profitability for the first time in four years. Daishin Securities’ Park Kang-ho noted, “LG Display’s share of iPhone panel supply has steadily risen over the past two to three years, and this structural shift is translating into profit growth,” projecting an annual profit of about ₩859 billion for 2024.\n\nSamsung Display is also expected to continue its strong earnings trajectory. Brokerages estimate its Q3 operating profit at ₩1.2–1.4 trillion, more than double the previous quarter’s ₩500 billion. Eugene Investment & Securities’ Lim So-jeong explained, “Driven by robust sales from key clients such as Apple, Samsung Display’s Q3 operating margin will improve sharply from 7.8% in the prior quarter.” KB Securities forecasts Samsung Display’s H2 operating profit at ₩2.5 trillion, slightly above last year’s ₩2.4 trillion, while Hyundai Motor Securities projects ₩2.7 trillion. Both expect Samsung Display to maintain a stable operating margin above 15% in the second half of the year.\n\n---\n\nhttps://t.co/yK6dccAZ3q", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03137258259156661, "ret_signed_m1d": -0.03137258259156661, "alpha_spy_m1d": -0.019712433647803174, "alpha_c_m1d": -0.013245529566919823, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.01812705302464679, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024509789177127317, "ret_signed_p1d": -0.024509789177127317, "alpha_spy_p1d": -0.027165788133718394, "alpha_c_p1d": -0.03440944268004342, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.009899653502916106, "ret_p1w": 0.08921562026478003, "ret_signed_p1w": 0.08921562026478003, "alpha_spy_p1w": 0.09198833985005361, "alpha_c_p1w": 0.07948310004733283, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.009732520217447194, "ret_p1m": -0.026470609331187056, "ret_signed_p1m": -0.026470609331187056, "alpha_spy_p1m": -0.011556077854777747, "alpha_c_p1m": 0.0355360442725412, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.06200665360372826, "ret_p3m": 0.5003956115080337, "ret_signed_p3m": 0.5003956115080337, "alpha_spy_p3m": 0.4919666442965871, "alpha_c_p3m": 0.5298286735013822, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": -0.029433061993348497, "ret_p6m": 1.1619615543818709, "ret_signed_p6m": 1.1619615543818709, "alpha_spy_p6m": 1.1286169879197296, "alpha_c_p6m": 1.124387883452265, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.03757367092960595, "price_path": [-0.2914, -0.3031, -0.3275, -0.3197, -0.3177, -0.3285, -0.3119, -0.2992, -0.307, -0.306, -0.2982, -0.3012, -0.3012, -0.3168, -0.3168, -0.3119, -0.3109, -0.3031, -0.3021, -0.3138, -0.3109, -0.3207, -0.3197, -0.3402, -0.3256, -0.3187, -0.3158, -0.3217, -0.3158, -0.3021, -0.2914, -0.2836, -0.2641, -0.2533, -0.225, -0.2367, -0.2162, -0.2162, -0.185, -0.1733, -0.1665, -0.1596, -0.187, -0.1745, -0.1775, -0.1569, -0.1201, -0.1201, -0.1201, -0.1201, -0.1201, -0.1201, -0.0745, -0.0853, -0.102, -0.0686, -0.0422, -0.0402, -0.0382, -0.0441, -0.0333, -0.0539, -0.0314, 0.0, -0.0245, -0.0147, 0.0206, 0.0539, 0.0892, 0.0284, -0.0137, -0.0275, -0.0402, -0.0137, 0.0147, 0.0108, 0.0078, -0.0471, -0.0137, -0.0412, -0.0539, -0.0137, -0.0706, -0.052, -0.0265, 0.0078, 0.0147, -0.0147, -0.0118, 0.0137, 0.0245, 0.0304, 0.0627, 0.0735, 0.0627, 0.0588, 0.052, 0.0676, 0.0275, 0.0078, 0.0578, 0.0549, 0.0422, 0.0833, 0.0931, 0.0892, 0.0892, 0.1471, 0.1773, 0.1812, 0.1812, 0.1812, 0.2659, 0.3605, 0.3684, 0.3891, 0.3674, 0.3694, 0.3674, 0.3556, 0.3822, 0.4176, 0.4669, 0.4708, 0.4304, 0.4728, 0.5004, 0.4984, 0.4984, 0.5713, 0.5999, 0.5831, 0.5812, 0.4817, 0.6501, 0.6659, 0.5694, 0.5625, 0.6393, 0.6334, 0.6531, 0.7595, 0.7851, 0.7851, 0.7851, 0.7851, 0.8718, 0.8728, 0.9014, 0.9703, 1.0048, 1.1476, 1.1329, 1.1329, 0.922, 0.6964, 0.8876, 0.8541, 0.7092, 0.8511, 0.8718, 0.8511, 0.8078, 0.859, 0.9102, 1.0541, 0.9752, 0.9644, 0.8353, 0.8688, 0.8619, 0.7743, 0.7743, 0.7404, 0.6506, 0.8722, 0.7612, 0.8382, 0.9063, 0.9398, 1.078, 1.0139, 1.0336, 0.9843, 1.0386, 1.083, 1.1472, 1.1323, 1.1175, 1.162]}
{"unit_id": "thread:2000043564039033285", "ticker_idx": 0, "ts": "2025-12-14T03:21:59+00:00", "symbol": "Doosan", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Doosan has established a plan to expand its CCL production capacity by 50% compared to current levels by early 2027… Supply to new GPU clients, rather than existing ones, is expected to gain momentum", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Doosan Electro-Materials on accelerating CCL capex (+23%), 100%+ utilization, NVIDIA supply wins, AWS/Google qualification pipeline, and 50% capacity expansion by 2027.", "resolved_tickers": ["006280.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Electronics CCL supplier context 006280.KS", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Doosan Expands Facility Investment for CCL… Accelerating Push into AI Chip Market\n\nDoosan’s Electro-Materials BG is rapidly expanding its Copper Clad Laminate (CCL) production capacity. It has been revealed that the company recently increased this year's facility investment scale compared to its original plan. This is interpreted as a strategy to address the rapidly growing demand for high-value-added CCLs amidst the global competition in AI semiconductor development.\n\nAccording to Doosan on the 14th, the company plans to invest approximately 107 billion KRW in CCL-related facility upgrades this year. This marks an increase of about 23% from the originally projected investment size of 86.5 billion KRW.\n\nCCL is a core material for semiconductor PCBs (Printed Circuit Boards), manufactured by laminating copper foil onto an insulating layer composed of resin, glass fiber, fillers, and other chemical substances. Currently, it is used across various industries, including semiconductor packaging, electronic devices, and telecommunications.\n\nRecently, demand for CCL has been surging, driven by the AI semiconductor market. This is a result of not only global fabless companies like NVIDIA and AMD but also Cloud Service Providers (CSPs) such as Google, AWS, and Meta entering the development of their own in-house AI accelerators. Since AI semiconductors must process massive amounts of data, they require CCL products equipped with superior high-frequency, high-speed, and low-loss properties.\n\nAs a result, some of Doosan’s CCL production lines in Jeungpyeong and Gimcheon are currently exceeding a 100% utilization rate, leading to tight supply availability. Consequently, Doosan has established a plan to expand its CCL production capacity by 50% compared to current levels by early 2027.\n\nRecently, the company has also been expanding the scale of its investment. According to the company's quarterly report, Doosan plans to invest 107 billion KRW in facility investments for CCL this year. This is an increase of approximately 20 billion KRW compared to the 86.5 billion KRW figure presented mid-year. This is interpreted as a strategy to accelerate investment as CCL demand proves stronger than anticipated.\n\nDoosan Electro-Materials BG began supplying NVIDIA in 2023 and achieved significant milestones last year by establishing itself as a key part of the supply chain for major AI semiconductors, such as the 'B100'. Furthermore, it is projected that the company will secure a substantial supply share for the 'Rubin' chip, which NVIDIA plans to launch in the second half of next year.\n\nIn particular, Doosan is leading the Compute Tray CCL sector, and its market position is evaluated to be even more solid. CCLs are categorized into Compute Trays and Switch Trays depending on the application, with Compute Tray CCLs known to have higher technical entry barriers.\n\nCustomer expansion is another key expectation. Currently, Doosan is in discussions regarding CCL supply with Amazon Web Services (AWS) and Google, in addition to NVIDIA. Full-scale supply is expected to begin next year.\n\nYang Seung-soo, an analyst at Meritz Securities, stated in a recent report, \"Supply to new GPU clients, rather than existing ones, is expected to gain momentum.\" He added, \"Supply of switching CCLs to North American CSP 'Company A' has newly commenced, and regarding ASIC CCLs for 'Company G', joint qualification (quality testing) with a domestic substrate manufacturer has been completed, with full-scale mass production expected in the first half of next year.\"\n\n---\n\nhttps://t.co/xHiXzwrdip", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029358825987370274, "ret_signed_m1d": -0.029358825987370274, "alpha_spy_m1d": -0.030871909834397693, "alpha_c_m1d": -0.0287117495622774, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": -0.000647076425092874, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012582299961297339, "ret_signed_p1d": -0.012582299961297339, "alpha_spy_p1d": -0.009850002366018629, "alpha_c_p1d": 0.0003514891969421674, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.012933789158239506, "ret_p1w": -0.01677643146781549, "ret_signed_p1w": -0.01677643146781549, "alpha_spy_p1w": -0.02577215678051792, "alpha_c_p1w": -0.02843112584660401, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.011654694378788522, "ret_p1m": -0.03535050996686584, "ret_signed_p1m": -0.03535050996686584, "alpha_spy_p1m": -0.057517999872576775, "alpha_c_p1m": -0.22110905775642942, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.18575854778956358, "ret_p3m": -0.10605152990059752, "ret_signed_p3m": -0.10605152990059752, "alpha_spy_p3m": -0.08739235663590539, "alpha_c_p3m": -0.46778836647586064, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.3617368365752631, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2109, -0.2133, -0.2133, -0.1642, -0.154, -0.1738, -0.1797, -0.2109, -0.2163, -0.2259, -0.2253, -0.2073, -0.2073, -0.2073, -0.2073, -0.2073, -0.2073, -0.2199, -0.2241, -0.2289, -0.2175, -0.2247, -0.2319, -0.2169, -0.2169, -0.2079, -0.2283, -0.2283, -0.2049, -0.2013, -0.2145, -0.2271, -0.2229, -0.2169, -0.2133, -0.2115, -0.2049, -0.2307, -0.2097, -0.2121, -0.1797, -0.1654, -0.1648, -0.1702, -0.1971, -0.2067, -0.1977, -0.2175, -0.2181, -0.2229, -0.1953, -0.1953, -0.1875, -0.178, -0.1762, -0.1732, -0.178, -0.1887, -0.2025, -0.1989, -0.1905, -0.0947, -0.0294, 0.0, -0.0126, -0.0096, -0.0024, 0.0006, -0.0168, -0.0431, -0.0605, -0.0605, -0.0815, -0.1055, -0.0473, -0.0473, -0.0473, -0.0449, -0.0354, -0.0354, -0.0359, -0.0515, -0.0629, -0.0413, -0.0354, -0.024, 0.0216, -0.0395, -0.0503, -0.0503, -0.0959, -0.0851, -0.0324, -0.0228, 0.0336, 0.0192, 0.0144, -0.012, -0.0611, -0.0216, -0.0186, 0.0186, -0.0156, 0.015, 0.0132, 0.0473, 0.0803, 0.0443, 0.0443, 0.0443, 0.0443, 0.0635, 0.0342, 0.0371, 0.0599, 0.0102, -0.0024, 0.0132, 0.0132, -0.0288, -0.169, -0.1037, -0.0887, -0.136, -0.0923, -0.0929, -0.1061, -0.0881, -0.1043, -0.1114, -0.0959, -0.13, -0.1198, -0.1444, -0.1342, -0.1072, -0.1114, -0.0953, -0.1225, -0.1352, -0.1013, -0.1461, -0.1491, -0.1528, -0.1624, -0.137, -0.1267, -0.1267, -0.1388, -0.1286, -0.1043, -0.0947, -0.085, -0.1062, -0.1286, -0.1388, -0.1394, -0.1352, -0.1431, -0.1522, -0.1576, -0.1703, -0.1703, -0.1679, -0.1679, -0.1782, -0.1733, -0.1643, -0.1879, -0.1691, -0.1637, -0.1237, -0.1606, -0.1873, -0.209, -0.2157, -0.2018, -0.1588, -0.1588, -0.1334, -0.091, -0.1364, -0.1443, null, null, null, null, null, null, null]}
{"unit_id": "orig:1980511112010559948", "ticker_idx": 0, "ts": "2025-10-21T05:46:59+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the memory (storage) sector is in the early to middle stages of a strong up-cycle, and that even though related share prices have hit record highs, the 'best' part of the rally may still lie ahead—so investors should not leave the market too early due to 'peak fears'", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley says memory upcycle is early-to-mid stage, peak fears are misplaced, stay long; SK Hynix and Samsung as key names with strong earnings momentum ahead.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Surging demand, depleted inventories, memory is already a “seller’s market”… Morgan Stanley: investors shouldn’t exit on “peak fears”\n\nThe demand wave triggered by AI is reshaping the global memory market.\n\nOn October 19, Morgan Stanley noted in its latest report that the memory (storage) sector is in the early to middle stages of a strong up-cycle, and that even though related share prices have hit record highs, the “best” part of the rally may still lie ahead—so investors should not leave the market too early due to “peak fears.”\n\nWorsening supply shortage, shifting to a seller’s market\nThrough channel checks, Morgan Stanley found that, aided by a surge in orders from U.S. cloud service customers, memory chip makers are quoting price hikes of up to 25% for DRAM and NAND flash for the fourth quarter of 2025. This shows the market is rapidly tilting toward the seller and that the momentum for price increases is strong.\n\nThe report emphasized that the supply-demand imbalance in commodity memory is more severe than expected. DRAM vendors’ inventories have fallen sharply to less than two weeks, and NAND flash inventories have also returned to below their long-term average levels.\n\nThe firm expects it will still take 4–6 quarters for new capacity to catch up with demand, and that supply delays will persist. Such an environment typically prompts customers to build buffer inventory, further solidifying the seller’s-market structure.\n\n“We believe the current rally is still in the early to middle stages… A return of prices to historical peaks would imply ‘double’ the potential from here. We have seen no evidence that seasonal demand softness will affect memory pricing for the first quarter of 2026… Such conditions typically lead customers to build buffer inventories, thereby strengthening the seller’s market.”\n\nHow much further can prices rise? Morgan Stanley: still a “long way” from the peak\nMorgan Stanley believes current prices still have considerable upside before reaching historical highs.\n\nNAND flash: Given strong AI-led demand, prices are expected to “double” from current levels and return to the peak of about $0.13/GB seen in the second quarter of 2022.\n\nDRAM memory: Server memory module prices peaked at $10/GB in the first quarter of 2018, while current negotiated prices are about $5.4/GB. Considering that the potential total addressable market (TAM) for AI memory is expected to surpass that of the traditional commodity DRAM market by 2027, the report sees “sufficient grounds” for the price peak in this cycle to exceed the past.\n\nMove beyond “peak fears,” the key is “earnings momentum”\n\nRegarding the widely prevalent “peak fears” among investors, Morgan Stanley asserts this is a classic cognitive bias.\n\n“New highs usually call for higher highs, and this time there are sufficient earnings to support them; ‘staying in the market’ is more important than timing the market. The best phase of this rally is likely still ahead.”\n\nThe report stresses that what drives stock prices is not merely the AI narrative but “earnings momentum.” Taking SK hynix and Samsung Electronics as examples, SK hynix’s share price has risen about 140% over the past 12 months, underpinned by a 62% upward revision in earnings forecasts; by contrast, Samsung Electronics’ share price rose only about 64%, with earnings forecasts raised by just 14%. This difference clearly shows that “stronger earnings upgrades deliver stronger share-price returns.”\n\nAnalysts Shawn Kim and Duan Liu wrote in the report that it is almost impossible to time the market precisely, and that historical cycles show that during upswings, “time in the market” matters more than “timing the market.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013956793581783552, "ret_signed_m1d": 0.013956793581783552, "alpha_spy_m1d": 0.013941797436890132, "alpha_c_m1d": 0.00839635806634583, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0007941345459554124, "ret_signed_p1d": -0.0007941345459554124, "alpha_spy_p1d": 0.004404681416965299, "alpha_c_p1d": 0.01860923214171654, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.06839506472916757, "ret_signed_p1w": 0.06839506472916757, "alpha_spy_p1w": 0.04490287223426259, "alpha_c_p1w": 0.016816706029302297, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.0932779612689435, "ret_signed_p1m": 0.0932779612689435, "alpha_spy_p1m": 0.10617830347365476, "alpha_c_p1m": 0.11024863741861146, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.6359880167447961, "ret_signed_p3m": 0.6359880167447961, "alpha_spy_p3m": 0.6025988258197316, "alpha_c_p3m": 0.47283746263392146, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 1.27133163964156, "ret_signed_p6m": 1.27133163964156, "alpha_spy_p6m": 1.222715509495505, "alpha_c_p6m": 0.9553467705511627, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.4041, -0.4076, -0.3951, -0.3929, -0.3808, -0.3872, -0.4135, -0.406, -0.3994, -0.4071, -0.3938, -0.3816, -0.3691, -0.3607, -0.3575, -0.3578, -0.3651, -0.3724, -0.378, -0.3895, -0.3988, -0.3888, -0.3846, -0.3872, -0.3851, -0.3754, -0.3796, -0.3958, -0.3884, -0.3843, -0.3721, -0.3568, -0.3522, -0.3335, -0.3108, -0.2886, -0.2558, -0.2495, -0.226, -0.2383, -0.2012, -0.2113, -0.2004, -0.1865, -0.1942, -0.2006, -0.2233, -0.1994, -0.1958, -0.1555, -0.1153, -0.1084, -0.1033, -0.112, -0.0941, -0.101, -0.0802, -0.0746, -0.0922, -0.0649, -0.0174, -0.0079, 0.014, 0.0, -0.0008, 0.0035, 0.0536, 0.0837, 0.0684, 0.1053, 0.1203, 0.1253, 0.198, 0.1257, 0.1382, 0.1445, 0.1304, 0.183, 0.1819, 0.1854, 0.1678, 0.1287, 0.1643, 0.1075, 0.0933, 0.0731, 0.0284, 0.0615, 0.0706, 0.0955, 0.1121, 0.1023, 0.1155, 0.1367, 0.1275, 0.1102, 0.1403, 0.1831, 0.1807, 0.2126, 0.1862, 0.1673, 0.1354, 0.1037, 0.1242, 0.1618, 0.1825, 0.2375, 0.2431, 0.2617, 0.2617, 0.2864, 0.3415, 0.3476, 0.3357, 0.3357, 0.4329, 0.4737, 0.5489, 0.5609, 0.5424, 0.5645, 0.5686, 0.544, 0.5482, 0.5709, 0.636, 0.6429, 0.6179, 0.6704, 0.7044, 0.7155, 0.6765, 0.7813, 0.8611, 0.8701, 0.8683, 0.8164, 0.8985, 0.8331, 0.7648, 0.7798, 0.8215, 0.7949, 0.8518, 0.9144, 0.9139, 0.9139, 0.8943, 0.9292, 0.9633, 1.0198, 1.0193, 1.076, 1.1152, 1.2007, 1.1638, 1.1643, 0.9513, 0.8446, 0.9693, 0.9016, 0.8211, 0.9645, 1.0092, 0.9626, 0.9677, 1.0561, 1.104, 1.2142, 1.1279, 1.0847, 0.9575, 0.9916, 0.9733, 0.8556, 0.8585, 0.7467, 0.6958, 0.8853, 0.7972, 0.8562, 0.9059, 0.9382, 1.1161, 1.0937, 1.1193, 1.121, 1.2484, 1.2713]}
{"unit_id": "orig:1980511112010559948", "ticker_idx": 1, "ts": "2025-10-21T05:46:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix's share price has risen about 140% over the past 12 months, underpinned by a 62% upward revision in earnings forecasts", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley says memory upcycle is early-to-mid stage, peak fears are misplaced, stay long; SK Hynix and Samsung as key names with strong earnings momentum ahead.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Surging demand, depleted inventories, memory is already a “seller’s market”… Morgan Stanley: investors shouldn’t exit on “peak fears”\n\nThe demand wave triggered by AI is reshaping the global memory market.\n\nOn October 19, Morgan Stanley noted in its latest report that the memory (storage) sector is in the early to middle stages of a strong up-cycle, and that even though related share prices have hit record highs, the “best” part of the rally may still lie ahead—so investors should not leave the market too early due to “peak fears.”\n\nWorsening supply shortage, shifting to a seller’s market\nThrough channel checks, Morgan Stanley found that, aided by a surge in orders from U.S. cloud service customers, memory chip makers are quoting price hikes of up to 25% for DRAM and NAND flash for the fourth quarter of 2025. This shows the market is rapidly tilting toward the seller and that the momentum for price increases is strong.\n\nThe report emphasized that the supply-demand imbalance in commodity memory is more severe than expected. DRAM vendors’ inventories have fallen sharply to less than two weeks, and NAND flash inventories have also returned to below their long-term average levels.\n\nThe firm expects it will still take 4–6 quarters for new capacity to catch up with demand, and that supply delays will persist. Such an environment typically prompts customers to build buffer inventory, further solidifying the seller’s-market structure.\n\n“We believe the current rally is still in the early to middle stages… A return of prices to historical peaks would imply ‘double’ the potential from here. We have seen no evidence that seasonal demand softness will affect memory pricing for the first quarter of 2026… Such conditions typically lead customers to build buffer inventories, thereby strengthening the seller’s market.”\n\nHow much further can prices rise? Morgan Stanley: still a “long way” from the peak\nMorgan Stanley believes current prices still have considerable upside before reaching historical highs.\n\nNAND flash: Given strong AI-led demand, prices are expected to “double” from current levels and return to the peak of about $0.13/GB seen in the second quarter of 2022.\n\nDRAM memory: Server memory module prices peaked at $10/GB in the first quarter of 2018, while current negotiated prices are about $5.4/GB. Considering that the potential total addressable market (TAM) for AI memory is expected to surpass that of the traditional commodity DRAM market by 2027, the report sees “sufficient grounds” for the price peak in this cycle to exceed the past.\n\nMove beyond “peak fears,” the key is “earnings momentum”\n\nRegarding the widely prevalent “peak fears” among investors, Morgan Stanley asserts this is a classic cognitive bias.\n\n“New highs usually call for higher highs, and this time there are sufficient earnings to support them; ‘staying in the market’ is more important than timing the market. The best phase of this rally is likely still ahead.”\n\nThe report stresses that what drives stock prices is not merely the AI narrative but “earnings momentum.” Taking SK hynix and Samsung Electronics as examples, SK hynix’s share price has risen about 140% over the past 12 months, underpinned by a 62% upward revision in earnings forecasts; by contrast, Samsung Electronics’ share price rose only about 64%, with earnings forecasts raised by just 14%. This difference clearly shows that “stronger earnings upgrades deliver stronger share-price returns.”\n\nAnalysts Shawn Kim and Duan Liu wrote in the report that it is almost impossible to time the market precisely, and that historical cycles show that during upswings, “time in the market” matters more than “timing the market.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013569962362283139, "ret_signed_m1d": 0.013569962362283139, "alpha_spy_m1d": 0.013554966217389719, "alpha_c_m1d": 0.008009526846845416, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005219261582720458, "ret_signed_p1d": 0.005219261582720458, "alpha_spy_p1d": 0.010418077545641169, "alpha_c_p1d": 0.02462262827039241, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.08768271798560034, "ret_signed_p1w": 0.08768271798560034, "alpha_spy_p1w": 0.06419052549069537, "alpha_c_p1w": 0.03610435928573508, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.1732776790035393, "ret_signed_p1m": 0.1732776790035393, "alpha_spy_p1m": 0.18617802120825055, "alpha_c_p1m": 0.19024835515320726, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.5794184346702942, "ret_signed_p3m": 0.5794184346702942, "alpha_spy_p3m": 0.5460292437452297, "alpha_c_p3m": 0.41626788055941955, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 1.3776855322772827, "ret_signed_p6m": 1.3776855322772827, "alpha_spy_p6m": 1.3290694021312277, "alpha_c_p6m": 1.0617006631868855, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.4382, -0.4455, -0.4538, -0.4528, -0.4507, -0.4298, -0.4622, -0.4622, -0.4507, -0.4611, -0.4538, -0.4653, -0.4434, -0.4392, -0.4205, -0.4236, -0.4236, -0.4424, -0.4517, -0.4674, -0.4893, -0.4767, -0.459, -0.4549, -0.458, -0.4395, -0.4384, -0.4656, -0.4562, -0.452, -0.4457, -0.429, -0.4217, -0.3987, -0.3653, -0.3591, -0.3142, -0.309, -0.2735, -0.3038, -0.2578, -0.2578, -0.2672, -0.2463, -0.2537, -0.2557, -0.2975, -0.2714, -0.2745, -0.2484, -0.1743, -0.1743, -0.1743, -0.1743, -0.1743, -0.1743, -0.1065, -0.1336, -0.1409, -0.118, -0.0553, -0.0282, 0.0136, 0.0, 0.0052, -0.001, 0.0647, 0.1169, 0.0877, 0.1649, 0.1858, 0.167, 0.2944, 0.2234, 0.2088, 0.238, 0.2109, 0.2651, 0.2923, 0.2881, 0.2777, 0.1691, 0.2651, 0.19, 0.1733, 0.1921, 0.0877, 0.0856, 0.0835, 0.0939, 0.1365, 0.1073, 0.124, 0.1658, 0.1532, 0.1323, 0.1365, 0.2055, 0.1825, 0.2263, 0.1804, 0.1929, 0.1574, 0.1073, 0.1511, 0.1532, 0.1428, 0.2117, 0.2201, 0.2284, 0.2284, 0.2514, 0.3371, 0.3601, 0.3601, 0.3601, 0.4144, 0.4541, 0.5167, 0.5502, 0.5794, 0.5543, 0.5648, 0.5418, 0.5502, 0.5648, 0.5794, 0.5961, 0.5523, 0.546, 0.5773, 0.6024, 0.5376, 0.6713, 0.757, 0.7988, 0.8991, 0.734, 0.8949, 0.8803, 0.7591, 0.7528, 0.8531, 0.8301, 0.7967, 0.8552, 0.8385, 0.8385, 0.8385, 0.8385, 0.8677, 0.9826, 0.9868, 1.0996, 1.1268, 1.3002, 1.2207, 1.2207, 0.9654, 0.777, 0.9695, 0.934, 0.7498, 0.9633, 0.9988, 0.9465, 0.9047, 1.0386, 1.0302, 1.2102, 1.1202, 1.1077, 0.9528, 1.0637, 1.0826, 0.9528, 0.9528, 0.8272, 0.6891, 0.8775, 0.7372, 0.8335, 0.8544, 0.9172, 1.1621, 1.0888, 1.1495, 1.1768, 1.3086, 1.3777]}
{"unit_id": "orig:1980511112010559948", "ticker_idx": 2, "ts": "2025-10-21T05:46:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' share price rose only about 64%, with earnings forecasts raised by just 14%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley says memory upcycle is early-to-mid stage, peak fears are misplaced, stay long; SK Hynix and Samsung as key names with strong earnings momentum ahead.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Surging demand, depleted inventories, memory is already a “seller’s market”… Morgan Stanley: investors shouldn’t exit on “peak fears”\n\nThe demand wave triggered by AI is reshaping the global memory market.\n\nOn October 19, Morgan Stanley noted in its latest report that the memory (storage) sector is in the early to middle stages of a strong up-cycle, and that even though related share prices have hit record highs, the “best” part of the rally may still lie ahead—so investors should not leave the market too early due to “peak fears.”\n\nWorsening supply shortage, shifting to a seller’s market\nThrough channel checks, Morgan Stanley found that, aided by a surge in orders from U.S. cloud service customers, memory chip makers are quoting price hikes of up to 25% for DRAM and NAND flash for the fourth quarter of 2025. This shows the market is rapidly tilting toward the seller and that the momentum for price increases is strong.\n\nThe report emphasized that the supply-demand imbalance in commodity memory is more severe than expected. DRAM vendors’ inventories have fallen sharply to less than two weeks, and NAND flash inventories have also returned to below their long-term average levels.\n\nThe firm expects it will still take 4–6 quarters for new capacity to catch up with demand, and that supply delays will persist. Such an environment typically prompts customers to build buffer inventory, further solidifying the seller’s-market structure.\n\n“We believe the current rally is still in the early to middle stages… A return of prices to historical peaks would imply ‘double’ the potential from here. We have seen no evidence that seasonal demand softness will affect memory pricing for the first quarter of 2026… Such conditions typically lead customers to build buffer inventories, thereby strengthening the seller’s market.”\n\nHow much further can prices rise? Morgan Stanley: still a “long way” from the peak\nMorgan Stanley believes current prices still have considerable upside before reaching historical highs.\n\nNAND flash: Given strong AI-led demand, prices are expected to “double” from current levels and return to the peak of about $0.13/GB seen in the second quarter of 2022.\n\nDRAM memory: Server memory module prices peaked at $10/GB in the first quarter of 2018, while current negotiated prices are about $5.4/GB. Considering that the potential total addressable market (TAM) for AI memory is expected to surpass that of the traditional commodity DRAM market by 2027, the report sees “sufficient grounds” for the price peak in this cycle to exceed the past.\n\nMove beyond “peak fears,” the key is “earnings momentum”\n\nRegarding the widely prevalent “peak fears” among investors, Morgan Stanley asserts this is a classic cognitive bias.\n\n“New highs usually call for higher highs, and this time there are sufficient earnings to support them; ‘staying in the market’ is more important than timing the market. The best phase of this rally is likely still ahead.”\n\nThe report stresses that what drives stock prices is not merely the AI narrative but “earnings momentum.” Taking SK hynix and Samsung Electronics as examples, SK hynix’s share price has risen about 140% over the past 12 months, underpinned by a 62% upward revision in earnings forecasts; by contrast, Samsung Electronics’ share price rose only about 64%, with earnings forecasts raised by just 14%. This difference clearly shows that “stronger earnings upgrades deliver stronger share-price returns.”\n\nAnalysts Shawn Kim and Duan Liu wrote in the report that it is almost impossible to time the market precisely, and that historical cycles show that during upswings, “time in the market” matters more than “timing the market.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006153878179337546, "ret_signed_m1d": 0.006153878179337546, "alpha_spy_m1d": 0.006138882034444126, "alpha_c_m1d": 0.006674027187541509, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": -0.0005201490082039628, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011282015863678163, "ret_signed_p1d": 0.011282015863678163, "alpha_spy_p1d": 0.016480831826598874, "alpha_c_p1d": 0.021166120593945736, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.009884104730267573, "ret_p1w": 0.020512873474873627, "ret_signed_p1w": 0.020512873474873627, "alpha_spy_p1w": -0.0029793190200313457, "alpha_c_p1w": -0.02672304556824745, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.04723591904312108, "ret_p1m": -0.010256356053058968, "ret_signed_p1m": -0.010256356053058968, "alpha_spy_p1m": 0.002643986151652289, "alpha_c_p1m": 0.015303645220515416, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.025560001273574384, "ret_p3m": 0.5346033108188957, "ret_signed_p3m": 0.5346033108188957, "alpha_spy_p3m": 0.5012141198938311, "alpha_c_p3m": 0.5230136777144521, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.011589633104443609, "ret_p6m": 1.1791236756768795, "ret_signed_p6m": 1.1791236756768795, "alpha_spy_p6m": 1.1305075455308244, "alpha_c_p6m": 1.1337248876951764, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.04539878798170305, "price_path": [-0.3261, -0.3271, -0.2812, -0.2791, -0.2587, -0.2709, -0.2965, -0.2883, -0.2863, -0.2975, -0.2801, -0.2669, -0.275, -0.274, -0.2658, -0.2689, -0.2689, -0.2852, -0.2852, -0.2801, -0.2791, -0.2709, -0.2699, -0.2822, -0.2791, -0.2893, -0.2883, -0.3097, -0.2944, -0.2873, -0.2842, -0.2903, -0.2842, -0.2699, -0.2587, -0.2505, -0.2301, -0.2189, -0.1893, -0.2015, -0.1801, -0.1801, -0.1474, -0.1351, -0.128, -0.1208, -0.1494, -0.1364, -0.1395, -0.1179, -0.0795, -0.0795, -0.0795, -0.0795, -0.0795, -0.0795, -0.0318, -0.0431, -0.0605, -0.0256, 0.0021, 0.0041, 0.0062, 0.0, 0.0113, -0.0103, 0.0133, 0.0462, 0.0205, 0.0308, 0.0677, 0.1026, 0.1395, 0.0759, 0.0318, 0.0174, 0.0041, 0.0318, 0.0615, 0.0574, 0.0544, -0.0031, 0.0318, 0.0031, -0.0103, 0.0318, -0.0277, -0.0082, 0.0185, 0.0544, 0.0615, 0.0308, 0.0338, 0.0605, 0.0718, 0.0779, 0.1118, 0.1231, 0.1118, 0.1077, 0.1005, 0.1169, 0.0749, 0.0544, 0.1067, 0.1036, 0.0903, 0.1333, 0.1436, 0.1395, 0.1395, 0.2, 0.2316, 0.2357, 0.2357, 0.2357, 0.3244, 0.4233, 0.4315, 0.4532, 0.4305, 0.4326, 0.4305, 0.4181, 0.446, 0.4831, 0.5346, 0.5387, 0.4965, 0.5408, 0.5696, 0.5676, 0.5676, 0.6438, 0.6737, 0.6562, 0.6542, 0.5501, 0.7263, 0.7428, 0.6418, 0.6346, 0.715, 0.7088, 0.7294, 0.8407, 0.8675, 0.8675, 0.8675, 0.8675, 0.9582, 0.9592, 0.9891, 1.0613, 1.0973, 1.2468, 1.2313, 1.2313, 1.0108, 0.7747, 0.9747, 0.9396, 0.7881, 0.9365, 0.9582, 0.9365, 0.8912, 0.9448, 0.9984, 1.1489, 1.0664, 1.0551, 0.9201, 0.9551, 0.9479, 0.8562, 0.8562, 0.8208, 0.7268, 0.9586, 0.8424, 0.923, 0.9943, 1.0294, 1.174, 1.1068, 1.1275, 1.0758, 1.1326, 1.1791]}
{"unit_id": "reply:1997647175002530063", "ticker_idx": 0, "ts": "2025-12-07T12:39:35+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Nvidia is cheap right now.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Nvidia is cheap; implied bullish.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@ryanTesling Nvidia is cheap right now.\n\nI don’t have a view on Meta.", "reply_text": null, "parent_text": "@jukan05 top underrated AI stock right now ?\nwhat do you think of Meta?", "quoted_text": null, "parent_author": "ryanTesling", "ret_m1d": -0.016922669898804643, "ret_signed_m1d": -0.016922669898804643, "alpha_spy_m1d": -0.019936000294230705, "alpha_c_m1d": -0.005716678795586239, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0031258565800189597, "ret_signed_p1d": -0.0031258565800189597, "alpha_spy_p1d": -0.0022627116130573555, "alpha_c_p1d": -0.004346869464196823, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.04990572428397211, "ret_signed_p1w": -0.04990572428397211, "alpha_spy_p1w": -0.04566362671150781, "alpha_c_p1w": -0.00744205253210517, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.00910806956871002, "ret_signed_p1m": 0.00910806956871002, "alpha_spy_p1m": -0.005847692568182961, "alpha_c_p1m": -0.045907863818807515, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": -0.011910558839075214, "ret_signed_p3m": -0.011910558839075214, "alpha_spy_p3m": -0.011461791696156909, "alpha_c_p3m": -0.08796587321459148, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0444, -0.0452, -0.0417, -0.0421, -0.0576, -0.0823, -0.0502, -0.0479, -0.0105, -0.0384, -0.0463, -0.0424, -0.0397, -0.02, 0.0055, 0.0091, 0.0179, 0.0111, -0.0001, -0.0028, 0.0191, 0.0378, -0.0129, 0.0149, -0.0298, -0.0309, -0.0202, -0.0126, -0.0157, -0.0237, -0.0285, -0.0183, 0.0038, 0.032, 0.0834, 0.1158, 0.0934, 0.0912, 0.1149, 0.0708, 0.052, 0.0136, 0.014, 0.0727, 0.041, 0.0444, 0.007, 0.0248, 0.0056, -0.0226, 0.0052, -0.0265, -0.036, -0.0162, -0.0417, -0.0286, -0.0286, -0.0461, -0.0304, -0.0221, -0.0322, -0.0117, -0.0169, 0.0, -0.0031, -0.0095, -0.0249, -0.0568, -0.0499, -0.0422, -0.0787, -0.0615, -0.0246, -0.01, 0.0197, 0.0165, 0.0165, 0.0268, 0.0144, 0.0107, 0.0051, 0.0051, 0.0178, 0.0139, 0.0091, 0.0192, -0.0027, -0.0037, -0.0033, 0.0014, -0.013, 0.0081, 0.0037, 0.0037, -0.0403, -0.012, -0.0038, 0.0114, 0.005, 0.016, 0.0322, 0.0375, 0.0301, 0.0003, -0.0281, -0.0612, -0.0737, -0.0008, 0.0242, 0.0161, 0.0243, 0.0075, -0.0148, -0.0148, -0.0031, 0.0131, 0.0127, 0.023, 0.0323, 0.0393, 0.0539, -0.0036, -0.0451, -0.0165, -0.0296, -0.0135, -0.0119, -0.0417, -0.0156, -0.0042, 0.0026, -0.0129, -0.0285, -0.0125, -0.0195, -0.0277, -0.0376, -0.0692, -0.0534, -0.0557, -0.037, -0.0771, -0.0971, -0.1098, -0.06, -0.0528, -0.0439, -0.0439, -0.0426, -0.0401, -0.0186, -0.0088, 0.0167, 0.0203, 0.0591, 0.0718, 0.069, 0.087, 0.089, 0.0773, 0.0914, 0.076, 0.1225, 0.1675, 0.1489, 0.1278, 0.0756, 0.0696, 0.0697, 0.0591, 0.1201, 0.1399, 0.1599, 0.1827, 0.1899, 0.2172, 0.2706, 0.2144, 0.1982, 0.189, 0.2044, 0.1831, 0.1606, 0.1606, 0.158, 0.1458, 0.1547, 0.138, null, null]}
{"unit_id": "orig:1978992034322223376", "ticker_idx": 0, "ts": "2025-10-17T01:10:43+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we raise our 2025–2027 EPS estimates by 5–9%, lift our 12-month target price (TP) from NT$1,600 to NT$1,720, and maintain our Buy rating (on Conviction List)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Goldman raises TSMC TP to $344/NT$1,720, maintains Conviction Buy on sustained AI demand, N2 ramp, CoWoS growth, and upward revenue guidance revision.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) TSMC: Strong 3Q GM Performance; Sustained AI Growth to Keep Multi-Year Trajectory; Raises TP to $344, Maintains Buy Rating\n\n1. AI Demand Expansion\nTSMC held its 3Q25 earnings call on October 16, 2025. The company reported a strong gross margin (GM) for the quarter, and its 4Q GM guidance indicated a further 50bps (0.5ppt) improvement quarter-over-quarter. We believe one of the most important takeaways from the call was TSMC’s progressively more positive stance on AI. Management acknowledged that AI token generation is growing exponentially every three months, confirming that demand for advanced nodes is real. While TSMC maintained its long-term AI revenue CAGR guidance in the mid-40% range, management did not rule out a potential upward revision amid continued AI strength.\n    \n2. Advanced Packaging Continues to Gain Attention\nCoWoS remains another key focus area among investors. TSMC stated that it is working very hard to expand 2026 capacity and narrow the demand-supply gap. Overall, with CoWoS adoption expanding beyond AI into non-AI sectors, we continue to expect more sustained and diversified growth in CoWoS demand. We maintain our projection that CoWoS capacity and shipment will grow at CAGRs of 61% and 54% respectively from 2025–2027, with annual capacity expected to reach 675K / 1.2M / 1.74M wafers and shipments at 664K / 1.08M / 1.566M wafers.\n    \n3. Upward Revision to Annual Guidance\nAlthough management provided limited detail on the 2026 outlook, for 2025, TSMC raised its full-year revenue growth guidance (in USD terms) from 30% to the mid-30% range. Capex guidance was also slightly lifted from $38–42B to $40–42B. Furthermore, TSMC lowered its gross margin dilution guidance from overseas fabs from 2–3ppt to 1–2ppt. Overall, we raise our 2025–2027 EPS estimates by 5–9%, lift our 12-month target price (TP) from NT$1,600 to NT$1,720, and maintain our Buy rating (on Conviction List).\n    \n4. Stronger AI Demand Drives 2025 Guidance Upward\nCiting robust and sustained AI demand, TSMC revised its 2025 full-year USD revenue growth guidance upward from 30% YoY to the mid-30% range. Management emphasized that AI-related demand remains even higher than it was three months ago. For non-AI applications, TSMC noted that markets have bottomed and are gradually recovering. Overall, we now forecast TSMC’s USD revenue growth of 35.4% in 2025 and 21.7% in 2026 (previously 34.1% and 18.9%, respectively), supported by strong AI-driven demand for leading-edge nodes.\n    \n5. Solid N2 Demand Led by Smartphones and AI/HPC\nManagement stated that the N2 process remains on track for mass production in 4Q25, with expansion expected to accelerate in 2026 driven by smartphone and AI/HPC demand. We remain constructive on N2 development. At our Communacopia + Technology Conference 2025, management reiterated that smartphone and HPC customers will drive the initial adoption, with continued customer migration sustaining momentum thereafter. We expect AI to form the second adoption wave by late 2026/2027.\nWe project N2 capacity to reach 660K/1.23M wafers in 2026/2027, with shipments at 405K/990K wafers.\n\nIn conclusion, we expect N2’s revenue contribution during its ramp-up phase to be significantly higher than N3’s, driving a faster top-line increase. We now estimate that N2 will contribute 9.0% and 19.8% of wafer revenue in 2026/2027, compared with N3’s 5.1% and 17.8% in its first two years. As a result, we view 2027 as a year of reacceleration for TSMC, following already strong 21.7% YoY USD revenue growth in 2026. We now model TSMC’s USD revenue growth at 22.1% in 2027 (previously 20.4%).\n\n6. Potential Upward Revision to Long-Term AI Revenue Outlook\nDuring the call, management faced numerous investor questions about whether it plans to revise its five-year CAGR guidance for AI accelerator revenue, initially presented in the mid-40% range starting from 2024. Although management did not formally update this outlook, it did not rule out an upward revision and noted that further details will be provided in early 2026.\nThis commentary aligns with our previous view (see prior note) that TSMC’s AI-related revenue growth will likely match or exceed the overall AI infrastructure market CAGR of 42.3% over the next five years. Leveraging its leadership in both advanced nodes and advanced packaging, we expect TSMC to continue gaining share in the AI infrastructure market, with AI revenue growth potentially outpacing total AI infrastructure TAM expansion.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.016131256361758384, "ret_signed_m1d": 0.016131256361758384, "alpha_spy_m1d": 0.02177551819519552, "alpha_c_m1d": 0.014760331120714731, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008879111718613775, "ret_signed_p1d": 0.008879111718613775, "alpha_spy_p1d": -0.001521363681007859, "alpha_c_p1d": -0.003926131889613549, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": -0.00040661177788969827, "ret_signed_p1w": -0.00040661177788969827, "alpha_spy_p1w": -0.019762682546799915, "alpha_c_p1w": -0.024646173753300693, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": -0.0442930217099492, "ret_signed_p1m": -0.0442930217099492, "alpha_spy_p1m": -0.04621952841706023, "alpha_c_p1m": -0.03691325961320602, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.11153953826372809, "ret_signed_p3m": 0.11153953826372809, "alpha_spy_p3m": 0.06938071251261957, "alpha_c_p3m": -0.024762469253724717, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.25935987035011876, "ret_signed_p6m": 0.25935987035011876, "alpha_spy_p6m": 0.22080331279919552, "alpha_c_p6m": -0.037841416978417364, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.2075, -0.1881, -0.1838, -0.1703, -0.1799, -0.1847, -0.1794, -0.1837, -0.2054, -0.1926, -0.2147, -0.2184, -0.1804, -0.183, -0.1822, -0.1747, -0.1844, -0.1858, -0.193, -0.1845, -0.2139, -0.2277, -0.232, -0.2129, -0.2041, -0.1935, -0.1916, -0.1951, -0.2201, -0.2201, -0.2284, -0.2183, -0.2054, -0.1777, -0.1649, -0.1523, -0.1202, -0.1253, -0.1239, -0.117, -0.1119, -0.1094, -0.0896, -0.1024, -0.0761, -0.0419, -0.0487, -0.0624, -0.0736, -0.074, -0.0535, -0.0224, -0.0236, -0.0098, 0.0248, -0.0036, 0.032, 0.0163, -0.0489, 0.0265, 0.0029, 0.0326, 0.0161, 0.0, 0.0089, -0.0019, -0.021, -0.0147, -0.0004, 0.0107, 0.0219, 0.0339, 0.0276, 0.0181, 0.0331, -0.0035, -0.0049, -0.0198, -0.0291, 0.0006, -0.0133, -0.0151, -0.0436, -0.0348, -0.0443, -0.0582, -0.0431, -0.0596, -0.0678, -0.0354, -0.0352, -0.0174, -0.0174, -0.0121, -0.0251, -0.0101, 0.0013, -0.0073, -0.0012, 0.023, 0.0282, 0.051, 0.0359, -0.0076, -0.0222, -0.0252, -0.0589, -0.0326, -0.0181, -0.0034, 0.0091, 0.0153, 0.0153, 0.0291, 0.0225, 0.018, 0.0326, 0.0326, 0.0861, 0.095, 0.1126, 0.0829, 0.0806, 0.0997, 0.1274, 0.1255, 0.1115, 0.1609, 0.1635, 0.1635, 0.1117, 0.1082, 0.1124, 0.1379, 0.1306, 0.1497, 0.1632, 0.1538, 0.1233, 0.16, 0.1409, 0.1069, 0.1238, 0.1854, 0.2077, 0.2298, 0.2712, 0.2508, 0.2449, 0.2449, 0.2376, 0.231, 0.2246, 0.2591, 0.2574, 0.3108, 0.3175, 0.2804, 0.2728, 0.2543, 0.2, 0.2146, 0.2024, 0.1516, 0.1849, 0.1794, 0.2048, 0.1442, 0.1496, 0.1561, 0.179, 0.1571, 0.1545, 0.1219, 0.1533, 0.1697, 0.185, 0.1113, 0.1134, 0.0785, 0.1516, 0.1637, 0.1553, 0.1553, 0.1646, 0.1767, 0.2469, 0.2455, 0.2629, 0.2594]}
{"unit_id": "thread:2004120319456694625", "ticker_idx": 0, "ts": "2025-12-25T09:21:33+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix and Samsung Electronics, which are vying for the industry lead, has already been fully contracted out through next year's production", "tweet_type": "analysis", "timeline": "quarters", "summary": "Big-tech AI demand has fully booked SK Hynix and Samsung HBM/DRAM through next year; supply crunch structurally bullish for all three memory makers including Micron.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Microsoft and Google stationed in Korea pleading for supply… Get fired if you fail to secure Samsung and SK volume\n\nEarlier this month, purchasing executives from Microsoft (MS) who visited Korea held long-term agreement (LTA) and price negotiations with SK hynix. At this meeting, SK hynix expressed the position that “it is difficult to supply under the conditions MS wants.” An industry source explained, “Upon hearing this, an MS executive could not contain their anger and stormed out of the meeting.”\n\nAccording to the industry on the 25th, as the global shortage of artificial intelligence (AI) semiconductors intensifies, procurement executives from global big tech firms such as MS and Google are flocking to Korea to secure volume. An official in the semiconductor industry said, “To sign memory supply contracts with Samsung Electronics (005930) and SK hynix, procurement executives from the headquarters of major big tech companies including MS, Google, and Meta are practically stationed in Korea.” They are waging an all-out war to preempt memory semiconductors essential for AI chips and data center build-outs, including high-bandwidth memory (HBM), DRAM, and enterprise solid-state drives (eSSD).\n\nThe reason big-tech procurement executives are entreating SK hynix and Samsung Electronics is that without high-performance, low-power memory semiconductors they cannot secure a competitive edge in the AI industry. AI’s training and inference capabilities are determined by the performance of graphics processing units (GPUs), tensor processing units (TPUs), and data centers. GPUs and TPUs require HBM, and AI data centers need LPDDR in order to deliver high performance. There are only three companies in the world capable of producing high-performance HBM and LPDDR: SK hynix, Samsung Electronics, and Micron. In particular, it is known that the HBM and DRAM output of SK hynix and Samsung Electronics, which are vying for the industry lead, has already been fully contracted out through next year’s production. As memory semiconductors have effectively dried up in the market, big-tech procurement executives are staying in Korea, pleading for supply.\n\nGoogle, which is pushing to sell its in-house AI accelerator TPU to external customers, is understood to have fired the executives in charge after running into a memory supply crunch. Currently, about 60% of the HBM mounted on TPUs is supplied to Google by Samsung Electronics. As recent TPU demand exceeded expectations, Google sounded out SK hynix and Micron on whether additional volume could be secured. The answer was “impossible.” Google’s management is said to have dismissed the procurement personnel responsible, holding them accountable for creating supply-chain risk by failing to sign LTAs in advance. It is a disciplinary personnel measure against the working-level staff who did not secure volume in time.\n\nBig tech firms are even changing their hiring practices in order to put everything on the line in managing Asian supply chains, going beyond simply placing volume orders. Whereas in the past memory procurement managers worked at headquarters in Silicon Valley or Seattle, recently they are increasingly being hired to work in Asian regions such as Korea, Taiwan, and Singapore. This is seen as an attempt to closely manage the supply chain on the ground in Asia, where key semiconductor manufacturers like Samsung Electronics, SK hynix, and TSMC are located.\n\nIn fact, Google recently posted a job opening for a Global Memory Commodity Manager, seeking an expert to establish sourcing strategies for data-center memory such as DRAM and NAND flash. Meta is also recruiting talent under the title Memory Silicon Global Sourcing Manager, looking for someone capable of collaborating on technology roadmaps. The strategy is to place experts with engineering knowledge in the field to simultaneously conduct technical coordination and secure volume, going beyond simple purchasing. One industry insider said, “Big techs are currently placing open-ended orders with the three memory companies, effectively saying ‘Give us all the volume you can, regardless of price,’” adding, “Since both Samsung and SK’s advanced HBM and other cutting-edge process lines are already running at full capacity, it is physically difficult to meet all of these demands.”\n\n---\n\nhttps://t.co/gs8riSJdby", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018707539869095013, "ret_signed_p1d": 0.018707539869095013, "alpha_spy_p1d": 0.01880895490133816, "alpha_c_p1d": 0.014066927348801661, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.004640612520293352, "ret_p1w": 0.10714283432004468, "ret_signed_p1w": 0.10714283432004468, "alpha_spy_p1w": 0.11939694386725053, "alpha_c_p1w": 0.11823657734104598, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.011093743021001301, "ret_p1m": 0.30442176472254756, "ret_signed_p1m": 0.30442176472254756, "alpha_spy_p1m": 0.3060875775467443, "alpha_c_p1m": 0.20578656252107863, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": 0.09863520220146893, "ret_p3m": 0.6799649210329166, "ret_signed_p3m": 0.6799649210329166, "alpha_spy_p3m": 0.7312644385402628, "alpha_c_p3m": 0.5964325548618816, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": 0.08353236617103499, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4069, -0.4094, -0.3882, -0.3279, -0.3279, -0.3279, -0.3279, -0.3279, -0.3279, -0.2726, -0.2947, -0.3007, -0.282, -0.231, -0.2089, -0.1749, -0.186, -0.1817, -0.1868, -0.1333, -0.0908, -0.1146, -0.0517, -0.0347, -0.05, 0.0537, -0.0041, -0.016, 0.0078, -0.0143, 0.0299, 0.052, 0.0486, 0.0401, -0.0483, 0.0299, -0.0313, -0.0449, -0.0296, -0.1146, -0.1163, -0.118, -0.1095, -0.0748, -0.0986, -0.085, -0.051, -0.0612, -0.0782, -0.0748, -0.0187, -0.0374, -0.0017, -0.0391, -0.0289, -0.0578, -0.0986, -0.0629, -0.0612, -0.0697, -0.0136, -0.0068, 0.0, 0.0, 0.0187, 0.0884, 0.1071, 0.1071, 0.1071, 0.1514, 0.1837, 0.2347, 0.2619, 0.2857, 0.2653, 0.2738, 0.2551, 0.2619, 0.2738, 0.2857, 0.2993, 0.2636, 0.2585, 0.284, 0.3044, 0.2517, 0.3605, 0.4303, 0.4643, 0.5459, 0.4116, 0.5425, 0.5306, 0.432, 0.4269, 0.5085, 0.4898, 0.4626, 0.5102, 0.4966, 0.4966, 0.4966, 0.4966, 0.5204, 0.6139, 0.6173, 0.7092, 0.7313, 0.8725, 0.8078, 0.8078, 0.5999, 0.4465, 0.6033, 0.5743, 0.4244, 0.5982, 0.6271, 0.5846, 0.5505, 0.6595, 0.6527, 0.7992, 0.726, 0.7157, 0.5897, 0.68, 0.6953, 0.5897, 0.5897, 0.4874, 0.375, 0.5283, 0.4142, 0.4925, 0.5096, 0.5607, 0.76, 0.7004, 0.7498, 0.772, 0.8793, 0.9355, 0.9679, 0.9219, 0.9867, 1.0855, 1.0838, 1.0872, 1.0821, 1.2013, 1.215, 1.203, 1.1911, 1.1911, 1.4654, 1.4654, 1.7278, 1.8181, 1.8726, 2.2032, 2.1265, 2.3667, 2.3565, 2.0992, 2.135, 1.9732, 1.9732, 2.3054, 2.3071, 2.3071, 2.4962, 2.8217, 2.9007, 2.9757, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2004120319456694625", "ticker_idx": 1, "ts": "2025-12-25T09:21:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "To sign memory supply contracts with Samsung Electronics (005930) and SK hynix, procurement executives from the headquarters of major big tech companies including MS, Google, and Meta are practically stationed in Korea", "tweet_type": "analysis", "timeline": "quarters", "summary": "Big-tech AI demand has fully booked SK Hynix and Samsung HBM/DRAM through next year; supply crunch structurally bullish for all three memory makers including Micron.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Microsoft and Google stationed in Korea pleading for supply… Get fired if you fail to secure Samsung and SK volume\n\nEarlier this month, purchasing executives from Microsoft (MS) who visited Korea held long-term agreement (LTA) and price negotiations with SK hynix. At this meeting, SK hynix expressed the position that “it is difficult to supply under the conditions MS wants.” An industry source explained, “Upon hearing this, an MS executive could not contain their anger and stormed out of the meeting.”\n\nAccording to the industry on the 25th, as the global shortage of artificial intelligence (AI) semiconductors intensifies, procurement executives from global big tech firms such as MS and Google are flocking to Korea to secure volume. An official in the semiconductor industry said, “To sign memory supply contracts with Samsung Electronics (005930) and SK hynix, procurement executives from the headquarters of major big tech companies including MS, Google, and Meta are practically stationed in Korea.” They are waging an all-out war to preempt memory semiconductors essential for AI chips and data center build-outs, including high-bandwidth memory (HBM), DRAM, and enterprise solid-state drives (eSSD).\n\nThe reason big-tech procurement executives are entreating SK hynix and Samsung Electronics is that without high-performance, low-power memory semiconductors they cannot secure a competitive edge in the AI industry. AI’s training and inference capabilities are determined by the performance of graphics processing units (GPUs), tensor processing units (TPUs), and data centers. GPUs and TPUs require HBM, and AI data centers need LPDDR in order to deliver high performance. There are only three companies in the world capable of producing high-performance HBM and LPDDR: SK hynix, Samsung Electronics, and Micron. In particular, it is known that the HBM and DRAM output of SK hynix and Samsung Electronics, which are vying for the industry lead, has already been fully contracted out through next year’s production. As memory semiconductors have effectively dried up in the market, big-tech procurement executives are staying in Korea, pleading for supply.\n\nGoogle, which is pushing to sell its in-house AI accelerator TPU to external customers, is understood to have fired the executives in charge after running into a memory supply crunch. Currently, about 60% of the HBM mounted on TPUs is supplied to Google by Samsung Electronics. As recent TPU demand exceeded expectations, Google sounded out SK hynix and Micron on whether additional volume could be secured. The answer was “impossible.” Google’s management is said to have dismissed the procurement personnel responsible, holding them accountable for creating supply-chain risk by failing to sign LTAs in advance. It is a disciplinary personnel measure against the working-level staff who did not secure volume in time.\n\nBig tech firms are even changing their hiring practices in order to put everything on the line in managing Asian supply chains, going beyond simply placing volume orders. Whereas in the past memory procurement managers worked at headquarters in Silicon Valley or Seattle, recently they are increasingly being hired to work in Asian regions such as Korea, Taiwan, and Singapore. This is seen as an attempt to closely manage the supply chain on the ground in Asia, where key semiconductor manufacturers like Samsung Electronics, SK hynix, and TSMC are located.\n\nIn fact, Google recently posted a job opening for a Global Memory Commodity Manager, seeking an expert to establish sourcing strategies for data-center memory such as DRAM and NAND flash. Meta is also recruiting talent under the title Memory Silicon Global Sourcing Manager, looking for someone capable of collaborating on technology roadmaps. The strategy is to place experts with engineering knowledge in the field to simultaneously conduct technical coordination and secure volume, going beyond simple purchasing. One industry insider said, “Big techs are currently placing open-ended orders with the three memory companies, effectively saying ‘Give us all the volume you can, regardless of price,’” adding, “Since both Samsung and SK’s advanced HBM and other cutting-edge process lines are already running at full capacity, it is physically difficult to meet all of these demands.”\n\n---\n\nhttps://t.co/gs8riSJdby", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05310536393950405, "ret_signed_p1d": 0.05310536393950405, "alpha_spy_p1d": 0.05320677897174719, "alpha_c_p1d": 0.05153331060748512, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.00157205333201893, "ret_p1w": 0.08445416491506297, "ret_signed_p1w": 0.08445416491506297, "alpha_spy_p1w": 0.09670827446226882, "alpha_c_p1w": 0.10038042063887109, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.015926255723808125, "ret_p1m": 0.3756920117600133, "ret_signed_p1m": 0.3756920117600133, "alpha_spy_p1m": 0.3773578245842101, "alpha_c_p1m": 0.3839627833788012, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": -0.008270771618787864, "ret_p3m": 0.7157711166213605, "ret_signed_p3m": 0.7157711166213605, "alpha_spy_p3m": 0.7670706341287067, "alpha_c_p3m": 0.7839921117726819, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": -0.0682209951513214, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2421, -0.2448, -0.2259, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1503, -0.1602, -0.1755, -0.1449, -0.1206, -0.1188, -0.117, -0.1224, -0.1125, -0.1314, -0.1107, -0.0819, -0.1044, -0.0954, -0.063, -0.0324, 0.0, -0.0558, -0.0945, -0.1071, -0.1188, -0.0945, -0.0684, -0.072, -0.0747, -0.1251, -0.0945, -0.1197, -0.1314, -0.0945, -0.1467, -0.1296, -0.1062, -0.0747, -0.0684, -0.0954, -0.0927, -0.0693, -0.0594, -0.054, -0.0243, -0.0144, -0.0243, -0.0279, -0.0342, -0.0198, -0.0567, -0.0747, -0.0288, -0.0315, -0.0432, -0.0054, 0.0036, 0.0, 0.0, 0.0531, 0.0808, 0.0845, 0.0845, 0.0845, 0.1622, 0.2491, 0.2563, 0.2753, 0.2554, 0.2572, 0.2554, 0.2445, 0.269, 0.3015, 0.3467, 0.3504, 0.3133, 0.3522, 0.3775, 0.3757, 0.3757, 0.4426, 0.4689, 0.4535, 0.4517, 0.3603, 0.515, 0.5295, 0.4408, 0.4345, 0.505, 0.4996, 0.5177, 0.6154, 0.6389, 0.6389, 0.6389, 0.6389, 0.7185, 0.7194, 0.7456, 0.8089, 0.8406, 0.9717, 0.9582, 0.9582, 0.7646, 0.5575, 0.733, 0.7022, 0.5692, 0.6995, 0.7185, 0.6995, 0.6597, 0.7067, 0.7538, 0.8858, 0.8135, 0.8035, 0.685, 0.7158, 0.7094, 0.6289, 0.6289, 0.5979, 0.5154, 0.7189, 0.6169, 0.6876, 0.7501, 0.781, 0.9078, 0.8489, 0.8671, 0.8217, 0.8716, 0.9124, 0.9713, 0.9577, 0.9441, 0.9849, 0.9713, 1.0347, 0.9894, 1.0347, 1.0121, 1.0483, 0.9985, 0.9985, 1.1072, 1.1072, 1.4109, 1.4607, 1.4335, 1.5876, 1.5287, 1.574, 1.6828, 1.4516, 1.5468, 1.497, 1.5015, 1.7145, 1.651, 1.651, 1.7099, 1.7825, 1.7145, 1.8731, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2004120319456694625", "ticker_idx": 2, "ts": "2025-12-25T09:21:33+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "There are only three companies in the world capable of producing high-performance HBM and LPDDR: SK hynix, Samsung Electronics, and Micron", "tweet_type": "analysis", "timeline": "quarters", "summary": "Big-tech AI demand has fully booked SK Hynix and Samsung HBM/DRAM through next year; supply crunch structurally bullish for all three memory makers including Micron.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Microsoft and Google stationed in Korea pleading for supply… Get fired if you fail to secure Samsung and SK volume\n\nEarlier this month, purchasing executives from Microsoft (MS) who visited Korea held long-term agreement (LTA) and price negotiations with SK hynix. At this meeting, SK hynix expressed the position that “it is difficult to supply under the conditions MS wants.” An industry source explained, “Upon hearing this, an MS executive could not contain their anger and stormed out of the meeting.”\n\nAccording to the industry on the 25th, as the global shortage of artificial intelligence (AI) semiconductors intensifies, procurement executives from global big tech firms such as MS and Google are flocking to Korea to secure volume. An official in the semiconductor industry said, “To sign memory supply contracts with Samsung Electronics (005930) and SK hynix, procurement executives from the headquarters of major big tech companies including MS, Google, and Meta are practically stationed in Korea.” They are waging an all-out war to preempt memory semiconductors essential for AI chips and data center build-outs, including high-bandwidth memory (HBM), DRAM, and enterprise solid-state drives (eSSD).\n\nThe reason big-tech procurement executives are entreating SK hynix and Samsung Electronics is that without high-performance, low-power memory semiconductors they cannot secure a competitive edge in the AI industry. AI’s training and inference capabilities are determined by the performance of graphics processing units (GPUs), tensor processing units (TPUs), and data centers. GPUs and TPUs require HBM, and AI data centers need LPDDR in order to deliver high performance. There are only three companies in the world capable of producing high-performance HBM and LPDDR: SK hynix, Samsung Electronics, and Micron. In particular, it is known that the HBM and DRAM output of SK hynix and Samsung Electronics, which are vying for the industry lead, has already been fully contracted out through next year’s production. As memory semiconductors have effectively dried up in the market, big-tech procurement executives are staying in Korea, pleading for supply.\n\nGoogle, which is pushing to sell its in-house AI accelerator TPU to external customers, is understood to have fired the executives in charge after running into a memory supply crunch. Currently, about 60% of the HBM mounted on TPUs is supplied to Google by Samsung Electronics. As recent TPU demand exceeded expectations, Google sounded out SK hynix and Micron on whether additional volume could be secured. The answer was “impossible.” Google’s management is said to have dismissed the procurement personnel responsible, holding them accountable for creating supply-chain risk by failing to sign LTAs in advance. It is a disciplinary personnel measure against the working-level staff who did not secure volume in time.\n\nBig tech firms are even changing their hiring practices in order to put everything on the line in managing Asian supply chains, going beyond simply placing volume orders. Whereas in the past memory procurement managers worked at headquarters in Silicon Valley or Seattle, recently they are increasingly being hired to work in Asian regions such as Korea, Taiwan, and Singapore. This is seen as an attempt to closely manage the supply chain on the ground in Asia, where key semiconductor manufacturers like Samsung Electronics, SK hynix, and TSMC are located.\n\nIn fact, Google recently posted a job opening for a Global Memory Commodity Manager, seeking an expert to establish sourcing strategies for data-center memory such as DRAM and NAND flash. Meta is also recruiting talent under the title Memory Silicon Global Sourcing Manager, looking for someone capable of collaborating on technology roadmaps. The strategy is to place experts with engineering knowledge in the field to simultaneously conduct technical coordination and secure volume, going beyond simple purchasing. One industry insider said, “Big techs are currently placing open-ended orders with the three memory companies, effectively saying ‘Give us all the volume you can, regardless of price,’” adding, “Since both Samsung and SK’s advanced HBM and other cutting-edge process lines are already running at full capacity, it is physically difficult to meet all of these demands.”\n\n---\n\nhttps://t.co/gs8riSJdby", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006592661519674525, "ret_signed_p1d": -0.006592661519674525, "alpha_spy_p1d": -0.006491246487431379, "alpha_c_p1d": -0.011233274039967878, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.004640612520293352, "ret_p1w": -0.0040278306627797855, "ret_signed_p1w": -0.0040278306627797855, "alpha_spy_p1w": 0.008226278884426064, "alpha_c_p1w": 0.0070659123582215155, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.011093743021001301, "ret_p1m": 0.39462619246949293, "ret_signed_p1m": 0.39462619246949293, "alpha_spy_p1m": 0.3962920052936897, "alpha_c_p1m": 0.295990990268024, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": 0.09863520220146893, "ret_p3m": 0.38024901253905474, "ret_signed_p3m": 0.38024901253905474, "alpha_spy_p3m": 0.4315485300464009, "alpha_c_p3m": 0.29671664636801975, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": 0.08353236617103499, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4286, -0.4167, -0.365, -0.3594, -0.3448, -0.3339, -0.3523, -0.3144, -0.3291, -0.3665, -0.3276, -0.3475, -0.3305, -0.2935, -0.2941, -0.2787, -0.2944, -0.3077, -0.279, -0.236, -0.2322, -0.2259, -0.2095, -0.2186, -0.2194, -0.1813, -0.2395, -0.1716, -0.1687, -0.1701, -0.1164, -0.159, -0.1457, -0.1735, -0.139, -0.156, -0.2029, -0.2119, -0.2976, -0.2766, -0.2189, -0.2168, -0.1968, -0.1968, -0.1751, -0.1612, -0.1646, -0.1832, -0.2094, -0.1725, -0.1387, -0.1195, -0.0801, -0.0984, -0.1589, -0.1716, -0.189, -0.2133, -0.133, -0.0724, -0.0352, -0.0363, 0.0, 0.0, -0.0066, 0.0272, 0.0212, -0.004, -0.004, 0.1007, 0.0893, 0.1984, 0.1849, 0.1412, 0.2042, 0.207, 0.1799, 0.1633, 0.1747, 0.2659, 0.2659, 0.2737, 0.3578, 0.3874, 0.3946, 0.3578, 0.4316, 0.519, 0.5207, 0.4478, 0.5278, 0.4637, 0.324, 0.3361, 0.3773, 0.3383, 0.3025, 0.4319, 0.4446, 0.4365, 0.4365, 0.3951, 0.469, 0.4564, 0.4942, 0.469, 0.4587, 0.497, 0.4501, 0.439, 0.4401, 0.3249, 0.3985, 0.3856, 0.2922, 0.3586, 0.4067, 0.4611, 0.4145, 0.487, 0.5417, 0.6111, 0.6113, 0.5503, 0.4758, 0.411, 0.3802, 0.3333, 0.2404, 0.2466, 0.1234, 0.1794, 0.2842, 0.2786, 0.2786, 0.3188, 0.3182, 0.4199, 0.4715, 0.4683, 0.4892, 0.6257, 0.5927, 0.5962, 0.5887, 0.5655, 0.5688, 0.7018, 0.6817, 0.7341, 0.8313, 0.7605, 0.81, 0.8054, 0.8929, 1.0124, 1.235, 1.3271, 1.2574, 1.6072, 1.7766, 1.6762, 1.8055, 1.7091, 1.5298, 1.3793, 1.4394, 1.5554, 1.6606, 1.6218, 1.6218, 2.1276, 2.2412, 2.2241, 2.3898, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000092734280048732", "ticker_idx": 0, "ts": "2025-12-14T06:37:22+00:00", "symbol": "Lotte Energy Materials", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Lotte Energy Materials is the only company in Korea possessing the relevant capabilities", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish on Lotte Energy Materials and Doosan as they enter Nvidia's supply chain via HVLP copper foil/CCL collaboration, targeting 20–30% AI circuit foil share by 2028.", "resolved_tickers": ["020150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Lotte Energy Materials 020150.KS Korea copper foil", "bench_c": "XLI", "bench_c_reason": "industry='Electrical Equipment & Parts'", "bench_c_industry": "Electrical Equipment & Parts", "unit_kind": "self_thread", "lang": "en", "tweet_text": "A Potential Korean Competitor to Mitsui Emerges: Lotte Energy Materials Enters Nvidia Supply Chain in Collaboration with Doosan\n\nRemember the company I covered on Substack last month? Korean media has reported that Lotte Energy Materials is aiming to enter Nvidia’s supply chain through a collaboration with Doosan.\n\nThe structure is that Doosan supplies CCL to Nvidia, and Lotte enters the supply chain by providing a portion of the HVLP copper foil used in those laminates.\n\nLotte Energy Materials has stated its goal to capture a 20–30% market share in AI circuit foil by 2028.\n\nIn reality, market entry is difficult as Mitsui Mining & Smelting currently dominates the circuit foil market with a 98% share. However, analysts suggest that an opportunity has opened up for Lotte Energy Materials due to client demands for vendor diversification. The industry assessment is that Lotte Energy Materials is the only company in Korea possessing the relevant capabilities. It is understood that Doosan BG's CCL, utilizing Lotte Energy Materials' HVLP4, is currently awaiting re-approval from Nvidia.\n\n---\n\nRelated: https://t.co/QI8HWsVxHc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030042918454935563, "ret_signed_m1d": 0.030042918454935563, "alpha_spy_m1d": 0.028529834607908144, "alpha_c_m1d": 0.03163534475729568, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": -0.001592426302360117, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03719599427753939, "ret_signed_p1d": -0.03719599427753939, "alpha_spy_p1d": -0.03446369668226068, "alpha_c_p1d": -0.03127197673913784, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.005924017538401549, "ret_p1w": -0.07296137339055797, "ret_signed_p1w": -0.07296137339055797, "alpha_spy_p1w": -0.0819570987032604, "alpha_c_p1w": -0.07676981708383301, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.003808443693275043, "ret_p1m": -0.12017167381974247, "ret_signed_p1m": -0.12017167381974247, "alpha_spy_p1m": -0.1423391637254534, "alpha_c_p1m": -0.16846594778455048, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.04829427396480801, "ret_p3m": 0.10157367668097272, "ret_signed_p3m": 0.10157367668097272, "alpha_spy_p3m": 0.12023284994566485, "alpha_c_p3m": 0.04541751914441572, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.056156157536557005, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3062, -0.3076, -0.3076, -0.3176, -0.3133, -0.3233, -0.3176, -0.3391, -0.3505, -0.3534, -0.3591, -0.3577, -0.3577, -0.3577, -0.3577, -0.3577, -0.3577, -0.3605, -0.3419, -0.3133, -0.3076, -0.2575, -0.2246, -0.2332, -0.2318, -0.2303, -0.2561, -0.196, -0.2132, -0.2017, -0.1931, -0.2489, -0.2561, -0.2461, -0.2461, -0.2704, -0.2647, -0.2947, -0.0844, 0.0057, 0.1502, 0.1502, 0.0572, 0.0744, -0.0014, -0.0472, 0.1159, 0.1788, 0.1731, 0.2232, 0.2246, 0.1202, 0.1187, 0.2046, 0.2089, 0.2089, 0.1574, 0.1488, 0.0873, 0.0501, 0.0515, 0.0515, 0.03, 0.0, -0.0372, -0.02, -0.0801, -0.0959, -0.073, -0.0758, -0.0672, -0.0672, -0.0844, -0.0944, -0.1259, -0.1259, -0.1259, -0.1416, -0.123, -0.1316, -0.1531, -0.1702, -0.1774, -0.1445, -0.1202, -0.0272, -0.0272, -0.0343, -0.0143, -0.02, -0.0429, 0.2117, 0.113, 0.166, 0.1545, 0.2246, 0.1702, 0.1116, 0.0529, 0.2003, 0.3662, 0.3977, 0.3262, 0.2947, 0.2961, 0.216, 0.2532, 0.2146, 0.2146, 0.2146, 0.2146, 0.2618, 0.2661, 0.2618, 0.2775, 0.3076, 0.3019, 0.279, 0.279, 0.1488, -0.0486, 0.0658, 0.1731, 0.0744, 0.083, 0.073, 0.1016, 0.0658, 0.0415, 0.0529, 0.1187, 0.0744, 0.1602, 0.1116, 0.133, 0.1688, 0.1602, 0.2046, 0.2031, 0.1845, 0.2117, 0.123, 0.1574, 0.3004, 0.3019, 0.2904, 0.2761, 0.4063, 0.3548, 0.3162, 0.3219, 0.3977, 0.5136, 0.5021, 0.6223, 0.7768, 0.814, 0.9485, 0.9199, 0.9285, 0.9542, 0.9084, 0.9084, 1.2461, 1.2461, 1.2461, 1.1173, 1.0429, 1.1574, 1.0572, 0.9342, 0.9142, 0.7511, 0.711, 0.6137, 0.5393, 0.6509, 0.7196, 0.7196, 0.6624, 0.6652, 0.6395, 0.6052, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000092734280048732", "ticker_idx": 1, "ts": "2025-12-14T06:37:22+00:00", "symbol": "Doosan", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Doosan supplies CCL to Nvidia, and Lotte enters the supply chain by providing a portion of the HVLP copper foil used in those laminates", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish on Lotte Energy Materials and Doosan as they enter Nvidia's supply chain via HVLP copper foil/CCL collaboration, targeting 20–30% AI circuit foil share by 2028.", "resolved_tickers": ["006280.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Electronics CCL supplier context 006280.KS", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "self_thread", "lang": "en", "tweet_text": "A Potential Korean Competitor to Mitsui Emerges: Lotte Energy Materials Enters Nvidia Supply Chain in Collaboration with Doosan\n\nRemember the company I covered on Substack last month? Korean media has reported that Lotte Energy Materials is aiming to enter Nvidia’s supply chain through a collaboration with Doosan.\n\nThe structure is that Doosan supplies CCL to Nvidia, and Lotte enters the supply chain by providing a portion of the HVLP copper foil used in those laminates.\n\nLotte Energy Materials has stated its goal to capture a 20–30% market share in AI circuit foil by 2028.\n\nIn reality, market entry is difficult as Mitsui Mining & Smelting currently dominates the circuit foil market with a 98% share. However, analysts suggest that an opportunity has opened up for Lotte Energy Materials due to client demands for vendor diversification. The industry assessment is that Lotte Energy Materials is the only company in Korea possessing the relevant capabilities. It is understood that Doosan BG's CCL, utilizing Lotte Energy Materials' HVLP4, is currently awaiting re-approval from Nvidia.\n\n---\n\nRelated: https://t.co/QI8HWsVxHc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029358825987370274, "ret_signed_m1d": -0.029358825987370274, "alpha_spy_m1d": -0.030871909834397693, "alpha_c_m1d": -0.0287117495622774, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": -0.000647076425092874, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012582299961297339, "ret_signed_p1d": -0.012582299961297339, "alpha_spy_p1d": -0.009850002366018629, "alpha_c_p1d": 0.0003514891969421674, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.012933789158239506, "ret_p1w": -0.01677643146781549, "ret_signed_p1w": -0.01677643146781549, "alpha_spy_p1w": -0.02577215678051792, "alpha_c_p1w": -0.02843112584660401, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.011654694378788522, "ret_p1m": -0.03535050996686584, "ret_signed_p1m": -0.03535050996686584, "alpha_spy_p1m": -0.057517999872576775, "alpha_c_p1m": -0.22110905775642942, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.18575854778956358, "ret_p3m": -0.10605152990059752, "ret_signed_p3m": -0.10605152990059752, "alpha_spy_p3m": -0.08739235663590539, "alpha_c_p3m": -0.46778836647586064, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.3617368365752631, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2109, -0.2133, -0.2133, -0.1642, -0.154, -0.1738, -0.1797, -0.2109, -0.2163, -0.2259, -0.2253, -0.2073, -0.2073, -0.2073, -0.2073, -0.2073, -0.2073, -0.2199, -0.2241, -0.2289, -0.2175, -0.2247, -0.2319, -0.2169, -0.2169, -0.2079, -0.2283, -0.2283, -0.2049, -0.2013, -0.2145, -0.2271, -0.2229, -0.2169, -0.2133, -0.2115, -0.2049, -0.2307, -0.2097, -0.2121, -0.1797, -0.1654, -0.1648, -0.1702, -0.1971, -0.2067, -0.1977, -0.2175, -0.2181, -0.2229, -0.1953, -0.1953, -0.1875, -0.178, -0.1762, -0.1732, -0.178, -0.1887, -0.2025, -0.1989, -0.1905, -0.0947, -0.0294, 0.0, -0.0126, -0.0096, -0.0024, 0.0006, -0.0168, -0.0431, -0.0605, -0.0605, -0.0815, -0.1055, -0.0473, -0.0473, -0.0473, -0.0449, -0.0354, -0.0354, -0.0359, -0.0515, -0.0629, -0.0413, -0.0354, -0.024, 0.0216, -0.0395, -0.0503, -0.0503, -0.0959, -0.0851, -0.0324, -0.0228, 0.0336, 0.0192, 0.0144, -0.012, -0.0611, -0.0216, -0.0186, 0.0186, -0.0156, 0.015, 0.0132, 0.0473, 0.0803, 0.0443, 0.0443, 0.0443, 0.0443, 0.0635, 0.0342, 0.0371, 0.0599, 0.0102, -0.0024, 0.0132, 0.0132, -0.0288, -0.169, -0.1037, -0.0887, -0.136, -0.0923, -0.0929, -0.1061, -0.0881, -0.1043, -0.1114, -0.0959, -0.13, -0.1198, -0.1444, -0.1342, -0.1072, -0.1114, -0.0953, -0.1225, -0.1352, -0.1013, -0.1461, -0.1491, -0.1528, -0.1624, -0.137, -0.1267, -0.1267, -0.1388, -0.1286, -0.1043, -0.0947, -0.085, -0.1062, -0.1286, -0.1388, -0.1394, -0.1352, -0.1431, -0.1522, -0.1576, -0.1703, -0.1703, -0.1679, -0.1679, -0.1782, -0.1733, -0.1643, -0.1879, -0.1691, -0.1637, -0.1237, -0.1606, -0.1873, -0.209, -0.2157, -0.2018, -0.1588, -0.1588, -0.1334, -0.091, -0.1364, -0.1443, null, null, null, null, null, null, null]}
{"unit_id": "quote:1994237376101798217", "ticker_idx": 0, "ts": "2025-11-28T02:50:16+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's unsustainable for SK Hynix to maintain the kind of overwhelming HBM technology lead it had in 2025", "tweet_type": "analysis", "timeline": "years", "summary": "Author endorses thesis that SK Hynix's HBM dominance fades by ~2028 as Samsung and Micron close the gap, with Samsung best positioned on capacity.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I share that outlook as well.\nThe view is that it’s unsustainable for SK Hynix to maintain the kind of overwhelming HBM technology lead it had in 2025.\n\n'A semiconductor industry official said, \"By around 2028, there is a high possibility that Samsung Electronics and Micron will both overcome HBM technological barriers and share profits similarly to SK Hynix,\" adding, \"They have already caught up significantly, and assuming that the technology of all major manufacturers becomes sufficient to meet Nvidia's requirements soon, Samsung Electronics, possessing the largest capacity among memory companies, will be able to occupy an advantageous position.\"'", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics Fired Up on HBM... SK Hynix's 'Iron Fortress' Shaken\n\nInterpretations are emerging that SK Hynix's stronghold is shaking due to Samsung Electronics' aggressive HBM investments. While SK Hynix has effectively monopolized the initial volume of the latest HBM https://t.co/SWHmmPLKMf", "parent_author": null, "ret_m1d": 0.026415088988895885, "ret_signed_m1d": -0.026415088988895885, "alpha_spy_m1d": -0.03184386287372698, "alpha_c_m1d": -0.03938778971182033, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015094235283469093, "ret_signed_p1d": -0.015094235283469093, "alpha_spy_p1d": -0.019659650886020286, "alpha_c_p1d": -0.01316396318053692, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.026415088988895885, "ret_signed_p1w": -0.026415088988895885, "alpha_spy_p1w": -0.023049500678759882, "alpha_c_p1w": 0.008046172788431782, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.20754721217124117, "ret_signed_p1m": -0.20754721217124117, "alpha_spy_p1m": -0.19804676805190846, "alpha_c_p1m": -0.17046518434014635, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.9207547330333123, "ret_signed_p3m": -0.9207547330333123, "alpha_spy_p3m": -0.9034758008526564, "alpha_c_p3m": -0.7072710528830408, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 2.6690219480971566, "ret_signed_p6m": -2.6690219480971566, "alpha_spy_p6m": -2.5717192600906063, "alpha_c_p6m": -2.02796009944342, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.5088, -0.5051, -0.4994, -0.4843, -0.4777, -0.457, -0.4268, -0.4212, -0.3806, -0.3759, -0.3439, -0.3712, -0.3297, -0.3297, -0.3382, -0.3194, -0.326, -0.3278, -0.3655, -0.342, -0.3448, -0.3212, -0.2543, -0.2543, -0.2543, -0.2543, -0.2543, -0.2543, -0.193, -0.2175, -0.2241, -0.2034, -0.1468, -0.1223, -0.0846, -0.0969, -0.0922, -0.0978, -0.0384, 0.0087, -0.0177, 0.0521, 0.0709, 0.054, 0.169, 0.1049, 0.0917, 0.1181, 0.0936, 0.1426, 0.1671, 0.1633, 0.1539, 0.0558, 0.1426, 0.0747, 0.0596, 0.0766, -0.0177, -0.0196, -0.0215, -0.012, 0.0264, 0.0, 0.0151, 0.0528, 0.0415, 0.0226, 0.0264, 0.0887, 0.0679, 0.1075, 0.066, 0.0774, 0.0453, 0.0, 0.0396, 0.0415, 0.0321, 0.0943, 0.1019, 0.1094, 0.1094, 0.1302, 0.2075, 0.2283, 0.2283, 0.2283, 0.2774, 0.3132, 0.3698, 0.4, 0.4264, 0.4038, 0.4132, 0.3925, 0.4, 0.4132, 0.4264, 0.4415, 0.4019, 0.3962, 0.4245, 0.4472, 0.3887, 0.5094, 0.5868, 0.6245, 0.7151, 0.566, 0.7113, 0.6981, 0.5887, 0.583, 0.6736, 0.6528, 0.6226, 0.6755, 0.6604, 0.6604, 0.6604, 0.6604, 0.6868, 0.7906, 0.7943, 0.8962, 0.9208, 1.0774, 1.0056, 1.0056, 0.775, 0.6048, 0.7787, 0.7466, 0.5803, 0.7731, 0.8052, 0.758, 0.7201, 0.8411, 0.8336, 0.9961, 0.9148, 0.9035, 0.7636, 0.8638, 0.8808, 0.7636, 0.7636, 0.6502, 0.5255, 0.6956, 0.5689, 0.6559, 0.6748, 0.7315, 0.9527, 0.8865, 0.9413, 0.9659, 1.085, 1.1474, 1.1833, 1.1322, 1.2041, 1.3137, 1.3118, 1.3156, 1.3099, 1.4422, 1.4574, 1.4441, 1.4309, 1.4309, 1.7352, 1.7352, 2.0263, 2.1265, 2.187, 2.5537, 2.4687, 2.7352, 2.7238, 2.4384, 2.4781, 2.2985, 2.2985, 2.6671, 2.669, 2.669]}
{"unit_id": "quote:1994237376101798217", "ticker_idx": 1, "ts": "2025-11-28T02:50:16+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics, possessing the largest capacity among memory companies, will be able to occupy an advantageous position", "tweet_type": "analysis", "timeline": "years", "summary": "Author endorses thesis that SK Hynix's HBM dominance fades by ~2028 as Samsung and Micron close the gap, with Samsung best positioned on capacity.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "I share that outlook as well.\nThe view is that it’s unsustainable for SK Hynix to maintain the kind of overwhelming HBM technology lead it had in 2025.\n\n'A semiconductor industry official said, \"By around 2028, there is a high possibility that Samsung Electronics and Micron will both overcome HBM technological barriers and share profits similarly to SK Hynix,\" adding, \"They have already caught up significantly, and assuming that the technology of all major manufacturers becomes sufficient to meet Nvidia's requirements soon, Samsung Electronics, possessing the largest capacity among memory companies, will be able to occupy an advantageous position.\"'", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics Fired Up on HBM... SK Hynix's 'Iron Fortress' Shaken\n\nInterpretations are emerging that SK Hynix's stronghold is shaking due to Samsung Electronics' aggressive HBM investments. While SK Hynix has effectively monopolized the initial volume of the latest HBM https://t.co/SWHmmPLKMf", "parent_author": null, "ret_m1d": 0.029850745100359788, "ret_signed_m1d": 0.029850745100359788, "alpha_spy_m1d": 0.03527951898519088, "alpha_c_m1d": 0.03837558247112216, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.008524837370762373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002985129303158196, "ret_signed_p1d": 0.002985129303158196, "alpha_spy_p1d": 0.007550544905709389, "alpha_c_p1d": 0.0025308376800112775, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0004542916231469185, "ret_p1w": 0.07860699601716625, "ret_signed_p1w": 0.07860699601716625, "alpha_spy_p1w": 0.07524140770703025, "alpha_c_p1w": 0.054220121313661274, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.02438687470350498, "ret_p1m": 0.19483494980871718, "ret_signed_p1m": 0.19483494980871718, "alpha_spy_p1m": 0.18533450568938448, "alpha_c_p1m": 0.17400296853378605, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.020831981274931133, "ret_p3m": 1.0347188009135735, "ret_signed_p3m": 1.0347188009135735, "alpha_spy_p3m": 1.0174398687329176, "alpha_c_p3m": 1.0339018457964952, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.0008169551170782441, "ret_p6m": 1.9306492844096503, "ret_signed_p6m": 1.9306492844096503, "alpha_spy_p6m": 1.8333465964031, "alpha_c_p6m": 1.6666687510569578, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.26398053335269256, "price_path": [-0.3155, -0.3086, -0.3056, -0.3115, -0.3056, -0.2917, -0.2808, -0.2729, -0.2531, -0.2422, -0.2135, -0.2253, -0.2045, -0.2045, -0.1728, -0.161, -0.154, -0.1471, -0.1748, -0.1622, -0.1652, -0.1443, -0.107, -0.107, -0.107, -0.107, -0.107, -0.107, -0.0607, -0.0716, -0.0886, -0.0547, -0.0279, -0.0259, -0.0239, -0.0299, -0.0189, -0.0398, -0.0169, 0.0149, -0.01, 0.0, 0.0358, 0.0697, 0.1055, 0.0438, 0.001, -0.0129, -0.0259, 0.001, 0.0299, 0.0259, 0.0229, -0.0328, 0.001, -0.0269, -0.0398, 0.001, -0.0567, -0.0378, -0.0119, 0.0229, 0.0299, 0.0, 0.003, 0.0289, 0.0398, 0.0458, 0.0786, 0.0896, 0.0786, 0.0746, 0.0677, 0.0836, 0.0428, 0.0229, 0.0736, 0.0706, 0.0577, 0.0995, 0.1095, 0.1055, 0.1055, 0.1642, 0.1948, 0.1988, 0.1988, 0.1988, 0.2848, 0.3808, 0.3888, 0.4098, 0.3878, 0.3898, 0.3878, 0.3758, 0.4028, 0.4388, 0.4888, 0.4928, 0.4518, 0.4948, 0.5228, 0.5208, 0.5208, 0.5948, 0.6238, 0.6068, 0.6048, 0.5038, 0.6748, 0.6908, 0.5928, 0.5858, 0.6638, 0.6578, 0.6778, 0.7858, 0.8117, 0.8117, 0.8117, 0.8117, 0.8997, 0.9007, 0.9297, 0.9997, 1.0347, 1.1797, 1.1647, 1.1647, 0.9507, 0.7218, 0.9157, 0.8817, 0.7348, 0.8787, 0.8997, 0.8787, 0.8347, 0.8867, 0.9387, 1.0847, 1.0047, 0.9937, 0.8627, 0.8967, 0.8897, 0.8008, 0.8008, 0.7664, 0.6752, 0.9002, 0.7874, 0.8656, 0.9347, 0.9688, 1.1091, 1.0439, 1.064, 1.0139, 1.069, 1.1141, 1.1792, 1.1642, 1.1491, 1.1942, 1.1792, 1.2493, 1.1992, 1.2493, 1.2243, 1.2644, 1.2093, 1.2093, 1.3295, 1.3295, 1.6651, 1.7202, 1.6902, 1.8605, 1.7954, 1.8455, 1.9657, 1.7102, 1.8154, 1.7603, 1.7653, 2.0008, 1.9306, 1.9306]}
{"unit_id": "quote:1994237376101798217", "ticker_idx": 2, "ts": "2025-11-28T02:50:16+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and Micron will both overcome HBM technological barriers and share profits similarly to SK Hynix", "tweet_type": "analysis", "timeline": "years", "summary": "Author endorses thesis that SK Hynix's HBM dominance fades by ~2028 as Samsung and Micron close the gap, with Samsung best positioned on capacity.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I share that outlook as well.\nThe view is that it’s unsustainable for SK Hynix to maintain the kind of overwhelming HBM technology lead it had in 2025.\n\n'A semiconductor industry official said, \"By around 2028, there is a high possibility that Samsung Electronics and Micron will both overcome HBM technological barriers and share profits similarly to SK Hynix,\" adding, \"They have already caught up significantly, and assuming that the technology of all major manufacturers becomes sufficient to meet Nvidia's requirements soon, Samsung Electronics, possessing the largest capacity among memory companies, will be able to occupy an advantageous position.\"'", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics Fired Up on HBM... SK Hynix's 'Iron Fortress' Shaken\n\nInterpretations are emerging that SK Hynix's stronghold is shaking due to Samsung Electronics' aggressive HBM investments. While SK Hynix has effectively monopolized the initial volume of the latest HBM https://t.co/SWHmmPLKMf", "parent_author": null, "ret_m1d": -0.026302486440269668, "ret_signed_m1d": -0.026302486440269668, "alpha_spy_m1d": -0.020873712555438573, "alpha_c_m1d": -0.013329785717345222, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016830145448651246, "ret_signed_p1d": 0.016830145448651246, "alpha_spy_p1d": 0.02139556105120244, "alpha_c_p1d": 0.014899873345719072, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.003129183432686311, "ret_signed_p1w": 0.003129183432686311, "alpha_spy_p1w": -0.00023640487744969185, "alpha_c_p1w": -0.031332078344641356, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.24530160753597463, "ret_signed_p1m": 0.24530160753597463, "alpha_spy_p1m": 0.23580116341664192, "alpha_c_p1m": 0.2082195797048798, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.8148397275073795, "ret_signed_p3m": 0.8148397275073795, "alpha_spy_p3m": 0.7975607953267236, "alpha_c_p3m": 0.601356047357108, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 2.178361776639486, "ret_signed_p6m": 2.178361776639486, "alpha_spy_p6m": 2.0810590886329354, "alpha_c_p6m": 1.537299927985749, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.4993, -0.4983, -0.4751, -0.4448, -0.4444, -0.4285, -0.4084, -0.3637, -0.3355, -0.3333, -0.3288, -0.3239, -0.2863, -0.3123, -0.3043, -0.2967, -0.3166, -0.3372, -0.3354, -0.3074, -0.2929, -0.2302, -0.2235, -0.2057, -0.1925, -0.2148, -0.1689, -0.1867, -0.2321, -0.1848, -0.209, -0.1883, -0.1436, -0.1442, -0.1256, -0.1446, -0.1607, -0.1259, -0.0738, -0.0693, -0.0616, -0.0417, -0.0527, -0.0537, -0.0075, -0.078, 0.0043, 0.0078, 0.0061, 0.0711, 0.0196, 0.0356, 0.002, 0.0438, 0.0231, -0.0337, -0.0447, -0.1485, -0.1231, -0.0531, -0.0505, -0.0263, -0.0263, 0.0, 0.0168, 0.0127, -0.0098, -0.0416, 0.0031, 0.0441, 0.0674, 0.1151, 0.0929, 0.0197, 0.0043, -0.0168, -0.0463, 0.051, 0.1245, 0.1696, 0.1683, 0.2123, 0.2123, 0.2043, 0.2453, 0.2379, 0.2074, 0.2074, 0.3344, 0.3205, 0.4528, 0.4364, 0.3834, 0.4599, 0.4632, 0.4304, 0.4102, 0.4241, 0.5346, 0.5346, 0.5441, 0.6461, 0.6819, 0.6907, 0.646, 0.7355, 0.8414, 0.8436, 0.7551, 0.8521, 0.7744, 0.605, 0.6198, 0.6697, 0.6224, 0.579, 0.7359, 0.7513, 0.7415, 0.7415, 0.6912, 0.7808, 0.7656, 0.8113, 0.7809, 0.7683, 0.8148, 0.758, 0.7445, 0.7458, 0.6062, 0.6954, 0.6797, 0.5665, 0.647, 0.7053, 0.7712, 0.7148, 0.8027, 0.869, 0.9531, 0.9533, 0.8794, 0.789, 0.7106, 0.6732, 0.6164, 0.5037, 0.5112, 0.3619, 0.4298, 0.5568, 0.55, 0.55, 0.5987, 0.598, 0.7214, 0.7839, 0.78, 0.8053, 0.9708, 0.9308, 0.9351, 0.9259, 0.8978, 0.9019, 1.0631, 1.0387, 1.1022, 1.22, 1.1342, 1.1942, 1.1887, 1.2947, 1.4396, 1.7094, 1.8211, 1.7366, 2.1606, 2.366, 2.2443, 2.4011, 2.2842, 2.0669, 1.8844, 1.9572, 2.0979, 2.2253, 2.1784, 2.1784]}
{"unit_id": "thread:1988504045162840284", "ticker_idx": 0, "ts": "2025-11-12T07:08:03+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics… pushing to raise NAND prices… internally considering price hikes of 20–30% or more", "tweet_type": "analysis", "timeline": "months", "summary": "NAND supply cuts by all four major makers driving 40-50%+ price surge; Samsung, Hynix, and Micron positioned to benefit from rising ASPs and QLC shift.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics, SK hynix, Kioxia and other NAND makers sharply cut NAND flash supply… A full-fledged price surge amid ‘short supply’\n\nSamsung Electronics, SK hynix, Kioxia, and Micron—the four global NAND flash manufacturers—are collectively reducing NAND flash supply in the second half of this year. While steering price increases through supply control, analysts note that production losses are inevitable as production equipment is being replaced with the quad-level cell (QLC) process, for which demand is surging, centered on AI data centers.\n\nAt the same time, Samsung Electronics, SK hynix, and Kioxia are pushing to raise NAND prices, which had stayed at cost levels throughout last year. In particular, Samsung is said to be discussing next year’s NAND supply volumes with major overseas customers while internally considering price hikes of 20–30% or more.\n\nAccording to annual NAND flash production data from market researcher Omdia obtained by ChosunBiz on the 12th, Samsung Electronics has lowered its target for NAND wafer output this year to around 4.72 million sheets, about 7% down from the previous year’s 5.07 million. Kioxia also adjusted its output from 4.80 million last year to 4.69 million this year. Omdia expects the production-cut stance at Samsung and Kioxia to continue into next year.\n\nSK hynix and Micron are likewise keeping output conservatively constrained in a bid to benefit from higher prices. SK hynix’s NAND output fell about 10%, from 2.01 million sheets last year to around 1.80 million this year. Micron’s situation is similar: it is maintaining production at Fab 7 in Singapore—its largest NAND production base—in the low 300,000-sheet range, keeping a conservative supply posture.\n\nAs major suppliers move in unison to control production, average selling prices (ASPs) for NAND products are rising sharply. NAND prices, which climbed 15% in just the last quarter, could surge a further 40–50% or more going forward, according to overseas market research firms. TrendForce reports that the spot wafer price of the most widely used 512-Gb triple-level cell (TLC) NAND chip rose 14.2% from the previous week to $5.51. Spot prices refer to prices for immediate transactions in the distribution market; when this price rises, it means the product has become harder to obtain.\n\nThat TLC-based NAND is in short supply also implies that major NAND suppliers are focusing on QLC, which offers higher profitability, instead of TLC. QLC and TLC refer to the number of bits stored in a single cell, the basic unit of NAND flash. TLC stores three bits per cell, whereas QLC stores four. For the same area, QLC can secure 30% more capacity than TLC, making it advantageous for producing high-capacity SSDs essential for AI data centers.\n\nA semiconductor industry source explained, “As existing TLC-based NAND production lines are converted to QLC, which is essential for AI data-center SSDs, a natural production-cut effect—where some TLC NAND production equipment goes idle—is occurring at all four major suppliers. As is customary, the ‘losses’ in output that occur during equipment and process transitions tend to lead to sharp rises in market prices.”\n\nMeanwhile, Samsung Electronics and SK hynix, which had long struggled with oversupply, are also moving to maximize profitability amid signs of NAND price increases. Earlier, U.S.-based SanDisk was said to have raised contract prices for NAND products by up to around 50% starting this month. With major North American big-tech firms, concerned about soaring NAND prices, embarking on “stockpiling,” some analyses suggest that next year’s NAND supply volumes are already sold out.\n\n---\n\nhttps://t.co/pOeySfy90b", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0038797231644660535, "ret_signed_m1d": 0.0038797231644660535, "alpha_spy_m1d": 0.004435812363652647, "alpha_c_m1d": 0.0069758299120047385, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.003096106747538685, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0029097732978722313, "ret_signed_p1d": -0.0029097732978722313, "alpha_spy_p1d": 0.01368424817286562, "alpha_c_p1d": 0.022029102618702723, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.024938875916574954, "ret_p1w": -0.06401547036464372, "ret_signed_p1w": -0.06401547036464372, "alpha_spy_p1w": -0.03365179647166183, "alpha_c_p1w": -0.019955842120262357, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.044059628244381366, "ret_p1m": 0.040737131377847735, "ret_signed_p1m": 0.040737131377847735, "alpha_spy_p1m": 0.03226456316552673, "alpha_c_p1m": 0.033864517788774195, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.00687261358907354, "ret_p3m": 0.6218127929611794, "ret_signed_p3m": 0.6218127929611794, "alpha_spy_p3m": 0.6033449008922456, "alpha_c_p3m": 0.6448977738396983, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": -0.023084980878518846, "ret_p6m": 1.651643939888746, "ret_signed_p6m": 1.651643939888746, "alpha_spy_p6m": 1.5750165368980622, "alpha_c_p6m": 1.493786849312865, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.15785709057588093, "price_path": [-0.3086, -0.3241, -0.3241, -0.3192, -0.3183, -0.3105, -0.3096, -0.3212, -0.3183, -0.3279, -0.327, -0.3472, -0.3328, -0.326, -0.3231, -0.3289, -0.3231, -0.3096, -0.299, -0.2912, -0.2719, -0.2613, -0.2333, -0.2449, -0.2246, -0.2246, -0.1937, -0.1821, -0.1754, -0.1686, -0.1956, -0.1833, -0.1862, -0.1659, -0.1295, -0.1295, -0.1295, -0.1295, -0.1295, -0.1295, -0.0844, -0.0951, -0.1115, -0.0786, -0.0524, -0.0504, -0.0485, -0.0543, -0.0436, -0.064, -0.0417, -0.0107, -0.0349, -0.0252, 0.0097, 0.0427, 0.0776, 0.0175, -0.0242, -0.0378, -0.0504, -0.0242, 0.0039, 0.0, -0.0029, -0.0572, -0.0242, -0.0514, -0.064, -0.0242, -0.0805, -0.0621, -0.0369, -0.0029, 0.0039, -0.0252, -0.0223, 0.0029, 0.0136, 0.0194, 0.0514, 0.0621, 0.0514, 0.0475, 0.0407, 0.0563, 0.0165, -0.0029, 0.0466, 0.0436, 0.031, 0.0718, 0.0815, 0.0776, 0.0776, 0.1348, 0.1647, 0.1686, 0.1686, 0.1686, 0.2524, 0.346, 0.3538, 0.3743, 0.3528, 0.3548, 0.3528, 0.3411, 0.3674, 0.4025, 0.4512, 0.4551, 0.4152, 0.4571, 0.4844, 0.4824, 0.4824, 0.5546, 0.5828, 0.5663, 0.5643, 0.4659, 0.6325, 0.6481, 0.5526, 0.5458, 0.6218, 0.616, 0.6355, 0.7407, 0.7661, 0.7661, 0.7661, 0.7661, 0.8518, 0.8528, 0.8811, 0.9493, 0.9834, 1.1247, 1.1101, 1.1101, 0.9015, 0.6783, 0.8674, 0.8343, 0.691, 0.8314, 0.8518, 0.8314, 0.7885, 0.8392, 0.8898, 1.0321, 0.9542, 0.9434, 0.8158, 0.8489, 0.8421, 0.7553, 0.7553, 0.7219, 0.633, 0.8522, 0.7424, 0.8185, 0.8859, 0.9191, 1.0559, 0.9924, 1.0119, 0.9631, 1.0168, 1.0608, 1.1242, 1.1096, 1.0949, 1.1389, 1.1242, 1.1926, 1.1438, 1.1926, 1.1682, 1.2073, 1.1535, 1.1535, 1.2707, 1.2707, 1.5979, 1.6516]}
{"unit_id": "thread:1988504045162840284", "ticker_idx": 1, "ts": "2025-11-12T07:08:03+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix… moving to maximize profitability amid signs of NAND price increases", "tweet_type": "analysis", "timeline": "months", "summary": "NAND supply cuts by all four major makers driving 40-50%+ price surge; Samsung, Hynix, and Micron positioned to benefit from rising ASPs and QLC shift.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics, SK hynix, Kioxia and other NAND makers sharply cut NAND flash supply… A full-fledged price surge amid ‘short supply’\n\nSamsung Electronics, SK hynix, Kioxia, and Micron—the four global NAND flash manufacturers—are collectively reducing NAND flash supply in the second half of this year. While steering price increases through supply control, analysts note that production losses are inevitable as production equipment is being replaced with the quad-level cell (QLC) process, for which demand is surging, centered on AI data centers.\n\nAt the same time, Samsung Electronics, SK hynix, and Kioxia are pushing to raise NAND prices, which had stayed at cost levels throughout last year. In particular, Samsung is said to be discussing next year’s NAND supply volumes with major overseas customers while internally considering price hikes of 20–30% or more.\n\nAccording to annual NAND flash production data from market researcher Omdia obtained by ChosunBiz on the 12th, Samsung Electronics has lowered its target for NAND wafer output this year to around 4.72 million sheets, about 7% down from the previous year’s 5.07 million. Kioxia also adjusted its output from 4.80 million last year to 4.69 million this year. Omdia expects the production-cut stance at Samsung and Kioxia to continue into next year.\n\nSK hynix and Micron are likewise keeping output conservatively constrained in a bid to benefit from higher prices. SK hynix’s NAND output fell about 10%, from 2.01 million sheets last year to around 1.80 million this year. Micron’s situation is similar: it is maintaining production at Fab 7 in Singapore—its largest NAND production base—in the low 300,000-sheet range, keeping a conservative supply posture.\n\nAs major suppliers move in unison to control production, average selling prices (ASPs) for NAND products are rising sharply. NAND prices, which climbed 15% in just the last quarter, could surge a further 40–50% or more going forward, according to overseas market research firms. TrendForce reports that the spot wafer price of the most widely used 512-Gb triple-level cell (TLC) NAND chip rose 14.2% from the previous week to $5.51. Spot prices refer to prices for immediate transactions in the distribution market; when this price rises, it means the product has become harder to obtain.\n\nThat TLC-based NAND is in short supply also implies that major NAND suppliers are focusing on QLC, which offers higher profitability, instead of TLC. QLC and TLC refer to the number of bits stored in a single cell, the basic unit of NAND flash. TLC stores three bits per cell, whereas QLC stores four. For the same area, QLC can secure 30% more capacity than TLC, making it advantageous for producing high-capacity SSDs essential for AI data centers.\n\nA semiconductor industry source explained, “As existing TLC-based NAND production lines are converted to QLC, which is essential for AI data-center SSDs, a natural production-cut effect—where some TLC NAND production equipment goes idle—is occurring at all four major suppliers. As is customary, the ‘losses’ in output that occur during equipment and process transitions tend to lead to sharp rises in market prices.”\n\nMeanwhile, Samsung Electronics and SK hynix, which had long struggled with oversupply, are also moving to maximize profitability amid signs of NAND price increases. Earlier, U.S.-based SanDisk was said to have raised contract prices for NAND products by up to around 50% starting this month. With major North American big-tech firms, concerned about soaring NAND prices, embarking on “stockpiling,” some analyses suggest that next year’s NAND supply volumes are already sold out.\n\n---\n\nhttps://t.co/pOeySfy90b", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0032414031765386486, "ret_signed_m1d": 0.0032414031765386486, "alpha_spy_m1d": 0.003797492375725242, "alpha_c_m1d": 0.01584228579369229, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008103711087697363, "ret_signed_p1d": -0.008103711087697363, "alpha_spy_p1d": 0.008490310383040489, "alpha_c_p1d": 0.022020272573561428, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.08914102511102207, "ret_signed_p1w": -0.08914102511102207, "alpha_spy_p1w": -0.058777351218040175, "alpha_c_p1w": -0.04388470269168021, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": -0.08362296077883868, "ret_signed_p1m": -0.08362296077883868, "alpha_spy_p1m": -0.09209552899115969, "alpha_c_p1m": -0.12682613450156632, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.4386309601742753, "ret_signed_p3m": 0.4386309601742753, "alpha_spy_p3m": 0.42016306810534143, "alpha_c_p3m": 0.29128660834802345, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6875831566891168, "ret_signed_p6m": 1.6875831566891168, "alpha_spy_p6m": 1.6109557536984331, "alpha_c_p6m": 1.1637157874256956, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.5525, -0.5671, -0.5744, -0.5865, -0.6035, -0.5938, -0.58, -0.5768, -0.5792, -0.5648, -0.564, -0.5851, -0.5778, -0.5746, -0.5697, -0.5567, -0.5511, -0.5332, -0.5073, -0.5024, -0.4676, -0.4635, -0.436, -0.4595, -0.4238, -0.4238, -0.4311, -0.4149, -0.4206, -0.4222, -0.4546, -0.4344, -0.4368, -0.4165, -0.359, -0.359, -0.359, -0.359, -0.359, -0.359, -0.3063, -0.3274, -0.3331, -0.3152, -0.2666, -0.2455, -0.2131, -0.2237, -0.2196, -0.2245, -0.1734, -0.1329, -0.1556, -0.0956, -0.0794, -0.094, 0.0049, -0.0502, -0.0616, -0.0389, -0.06, -0.0178, 0.0032, 0.0, -0.0081, -0.0924, -0.0178, -0.0762, -0.0891, -0.0746, -0.1556, -0.1572, -0.1588, -0.1507, -0.1177, -0.1404, -0.1274, -0.095, -0.1047, -0.1209, -0.1177, -0.0642, -0.082, -0.0479, -0.0836, -0.0739, -0.1015, -0.1404, -0.1063, -0.1047, -0.1128, -0.0593, -0.0528, -0.0463, -0.0463, -0.0285, 0.038, 0.0559, 0.0559, 0.0559, 0.098, 0.1288, 0.1775, 0.2035, 0.2262, 0.2067, 0.2148, 0.197, 0.2035, 0.2148, 0.2262, 0.2391, 0.2051, 0.2002, 0.2245, 0.244, 0.1937, 0.2975, 0.364, 0.3965, 0.4743, 0.3462, 0.4711, 0.4597, 0.3656, 0.3608, 0.4386, 0.4208, 0.3948, 0.4403, 0.4273, 0.4273, 0.4273, 0.4273, 0.45, 0.5392, 0.5424, 0.63, 0.6511, 0.7858, 0.724, 0.724, 0.5258, 0.3795, 0.529, 0.5014, 0.3584, 0.5242, 0.5518, 0.5112, 0.4787, 0.5827, 0.5762, 0.7159, 0.646, 0.6363, 0.516, 0.6022, 0.6168, 0.516, 0.516, 0.4185, 0.3113, 0.4575, 0.3487, 0.4234, 0.4397, 0.4884, 0.6785, 0.6216, 0.6688, 0.6899, 0.7923, 0.8459, 0.8768, 0.8329, 0.8946, 0.9889, 0.9873, 0.9905, 0.9856, 1.0994, 1.1124, 1.101, 1.0896, 1.0896, 1.3512, 1.3512, 1.6015, 1.6876]}
{"unit_id": "thread:1988504045162840284", "ticker_idx": 2, "ts": "2025-11-12T07:08:03+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron… maintaining production… keeping a conservative supply posture", "tweet_type": "analysis", "timeline": "months", "summary": "NAND supply cuts by all four major makers driving 40-50%+ price surge; Samsung, Hynix, and Micron positioned to benefit from rising ASPs and QLC shift.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics, SK hynix, Kioxia and other NAND makers sharply cut NAND flash supply… A full-fledged price surge amid ‘short supply’\n\nSamsung Electronics, SK hynix, Kioxia, and Micron—the four global NAND flash manufacturers—are collectively reducing NAND flash supply in the second half of this year. While steering price increases through supply control, analysts note that production losses are inevitable as production equipment is being replaced with the quad-level cell (QLC) process, for which demand is surging, centered on AI data centers.\n\nAt the same time, Samsung Electronics, SK hynix, and Kioxia are pushing to raise NAND prices, which had stayed at cost levels throughout last year. In particular, Samsung is said to be discussing next year’s NAND supply volumes with major overseas customers while internally considering price hikes of 20–30% or more.\n\nAccording to annual NAND flash production data from market researcher Omdia obtained by ChosunBiz on the 12th, Samsung Electronics has lowered its target for NAND wafer output this year to around 4.72 million sheets, about 7% down from the previous year’s 5.07 million. Kioxia also adjusted its output from 4.80 million last year to 4.69 million this year. Omdia expects the production-cut stance at Samsung and Kioxia to continue into next year.\n\nSK hynix and Micron are likewise keeping output conservatively constrained in a bid to benefit from higher prices. SK hynix’s NAND output fell about 10%, from 2.01 million sheets last year to around 1.80 million this year. Micron’s situation is similar: it is maintaining production at Fab 7 in Singapore—its largest NAND production base—in the low 300,000-sheet range, keeping a conservative supply posture.\n\nAs major suppliers move in unison to control production, average selling prices (ASPs) for NAND products are rising sharply. NAND prices, which climbed 15% in just the last quarter, could surge a further 40–50% or more going forward, according to overseas market research firms. TrendForce reports that the spot wafer price of the most widely used 512-Gb triple-level cell (TLC) NAND chip rose 14.2% from the previous week to $5.51. Spot prices refer to prices for immediate transactions in the distribution market; when this price rises, it means the product has become harder to obtain.\n\nThat TLC-based NAND is in short supply also implies that major NAND suppliers are focusing on QLC, which offers higher profitability, instead of TLC. QLC and TLC refer to the number of bits stored in a single cell, the basic unit of NAND flash. TLC stores three bits per cell, whereas QLC stores four. For the same area, QLC can secure 30% more capacity than TLC, making it advantageous for producing high-capacity SSDs essential for AI data centers.\n\nA semiconductor industry source explained, “As existing TLC-based NAND production lines are converted to QLC, which is essential for AI data-center SSDs, a natural production-cut effect—where some TLC NAND production equipment goes idle—is occurring at all four major suppliers. As is customary, the ‘losses’ in output that occur during equipment and process transitions tend to lead to sharp rises in market prices.”\n\nMeanwhile, Samsung Electronics and SK hynix, which had long struggled with oversupply, are also moving to maximize profitability amid signs of NAND price increases. Earlier, U.S.-based SanDisk was said to have raised contract prices for NAND products by up to around 50% starting this month. With major North American big-tech firms, concerned about soaring NAND prices, embarking on “stockpiling,” some analyses suggest that next year’s NAND supply volumes are already sold out.\n\n---\n\nhttps://t.co/pOeySfy90b", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015475712104625061, "ret_signed_m1d": -0.015475712104625061, "alpha_spy_m1d": -0.014919622905438468, "alpha_c_m1d": -0.0028748294874714198, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03246222316673142, "ret_signed_p1d": -0.03246222316673142, "alpha_spy_p1d": -0.015868201695993567, "alpha_c_p1d": -0.002338239505472628, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.07750103077046255, "ret_signed_p1w": -0.07750103077046255, "alpha_spy_p1w": -0.04713735687748066, "alpha_c_p1w": -0.0322447083511207, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": 0.055369522168985386, "ret_signed_p1m": 0.055369522168985386, "alpha_spy_p1m": 0.04689695395666438, "alpha_c_p1m": 0.012166348446257746, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.5665779193876388, "ret_signed_p3m": 0.5665779193876388, "alpha_spy_p3m": 0.548110027318705, "alpha_c_p3m": 0.41923356756138697, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6425602317905326, "ret_signed_p6m": 1.6425602317905326, "alpha_spy_p6m": 1.5659328287998489, "alpha_c_p6m": 1.1186928625271113, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.5068, -0.4958, -0.5019, -0.5217, -0.5275, -0.5198, -0.5249, -0.5246, -0.5195, -0.5021, -0.5144, -0.5144, -0.5165, -0.5155, -0.4931, -0.4639, -0.4635, -0.4481, -0.4287, -0.3856, -0.3584, -0.3562, -0.3519, -0.3471, -0.3108, -0.3359, -0.3282, -0.3209, -0.3401, -0.36, -0.3582, -0.3312, -0.3172, -0.2567, -0.2502, -0.233, -0.2203, -0.2418, -0.1975, -0.2147, -0.2585, -0.2129, -0.2362, -0.2163, -0.173, -0.1736, -0.1557, -0.174, -0.1896, -0.1559, -0.1057, -0.1013, -0.0939, -0.0746, -0.0853, -0.0863, -0.0416, -0.1097, -0.0302, -0.0268, -0.0285, 0.0343, -0.0155, 0.0, -0.0325, 0.0079, -0.012, -0.067, -0.0775, -0.1777, -0.1532, -0.0856, -0.0832, -0.0598, -0.0598, -0.0344, -0.0181, -0.0221, -0.0439, -0.0745, -0.0314, 0.0082, 0.0307, 0.0768, 0.0554, -0.0154, -0.0302, -0.0506, -0.0791, 0.0149, 0.0858, 0.1294, 0.1281, 0.1706, 0.1706, 0.1629, 0.2025, 0.1954, 0.1659, 0.1659, 0.2885, 0.2751, 0.4029, 0.387, 0.3359, 0.4097, 0.4129, 0.3812, 0.3617, 0.3751, 0.4818, 0.4818, 0.491, 0.5895, 0.6241, 0.6325, 0.5894, 0.6758, 0.7781, 0.7802, 0.6948, 0.7884, 0.7134, 0.5498, 0.5641, 0.6123, 0.5666, 0.5247, 0.6762, 0.691, 0.6816, 0.6816, 0.6331, 0.7196, 0.7049, 0.7491, 0.7196, 0.7075, 0.7524, 0.6975, 0.6845, 0.6857, 0.551, 0.6371, 0.6219, 0.5127, 0.5904, 0.6467, 0.7103, 0.6558, 0.7407, 0.8047, 0.886, 0.8861, 0.8148, 0.7275, 0.6517, 0.6157, 0.5608, 0.452, 0.4592, 0.3151, 0.3806, 0.5033, 0.4967, 0.4967, 0.5438, 0.543, 0.6622, 0.7226, 0.7188, 0.7432, 0.903, 0.8645, 0.8685, 0.8597, 0.8325, 0.8365, 0.9922, 0.9686, 1.0299, 1.1437, 1.0609, 1.1188, 1.1135, 1.2158, 1.3558, 1.6163, 1.7241, 1.6426]}
{"unit_id": "orig:1995360938330046836", "ticker_idx": 0, "ts": "2025-12-01T05:14:54+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Amid the accelerated ramp-up of TSMC's CoWoS, the total volume of GPUs and ASICs that can be shipped in 2026 is being revised upward", "tweet_type": "analysis", "timeline": "quarters", "summary": "Daishin note: TSMC CoWoS ramp drives upward revision to 2026 GPU/ASIC/HBM volumes; NVDA, AVGO, AMD all raising HBM orders.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities December 1 Note – Key Points\n\n1. Amid the accelerated ramp-up of TSMC’s CoWoS, the total volume of GPUs and ASICs that can be shipped in 2026 is being revised upward. The same applies to HBM.\n    \n2. It is estimated that Nvidia has increased its orders for 12-high HBM3e in order to maximize GB300 shipments, and Broadcom is also believed to have raised its HBM3e orders to prepare for strong volume growth from major customers such as Google and OpenAI.\n    \n3. AMD is likely to place more HBM4 orders than expected. In the AMD market, attention should be paid to the larger-than-expected HBM4 purchasing TAM.\n\n$TSM $NVDA $AVGO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013313420589721403, "ret_signed_m1d": 0.013313420589721403, "alpha_spy_m1d": 0.008727066374075276, "alpha_c_m1d": 0.015239973920504246, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015329548546699323, "ret_signed_p1d": 0.015329548546699323, "alpha_spy_p1d": 0.013477339272587319, "alpha_c_p1d": -0.003001112463262956, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.049325747546560716, "ret_signed_p1w": 0.049325747546560716, "alpha_spy_p1w": 0.044386582970814326, "alpha_c_p1w": 0.005156479025373528, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.044176679839072364, "ret_signed_p1m": 0.044176679839072364, "alpha_spy_p1m": 0.031284664442825294, "alpha_c_p1m": 0.011650426614526088, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 0.3133595757437866, "ret_signed_p3m": 0.3133595757437866, "alpha_spy_p3m": 0.2970914088300174, "alpha_c_p3m": 0.14242798967946357, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 0.4411780988407237, "ret_signed_p6m": 0.4411780988407237, "alpha_spy_p6m": 0.3315248193872169, "alpha_c_p6m": -0.2701025634571923, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.1982, -0.185, -0.1565, -0.1434, -0.1305, -0.0975, -0.1028, -0.1014, -0.0943, -0.0891, -0.0865, -0.0662, -0.0793, -0.0523, -0.0173, -0.0242, -0.0383, -0.0498, -0.0502, -0.0292, 0.0027, 0.0015, 0.0157, 0.0512, 0.0221, 0.0585, 0.0424, -0.0244, 0.0529, 0.0287, 0.0592, 0.0423, 0.0257, 0.0348, 0.0237, 0.0042, 0.0106, 0.0253, 0.0367, 0.0481, 0.0605, 0.054, 0.0443, 0.0597, 0.0221, 0.0207, 0.0054, -0.0041, 0.0264, 0.0121, 0.0102, -0.019, -0.0099, -0.0197, -0.034, -0.0185, -0.0354, -0.0439, -0.0106, -0.0104, 0.0079, 0.0079, 0.0133, 0.0, 0.0153, 0.027, 0.0182, 0.0245, 0.0493, 0.0547, 0.0781, 0.0625, 0.0179, 0.0029, -0.0001, -0.0347, -0.0078, 0.0071, 0.0222, 0.035, 0.0415, 0.0415, 0.0555, 0.0488, 0.0442, 0.0592, 0.0592, 0.114, 0.1232, 0.1412, 0.1107, 0.1084, 0.128, 0.1564, 0.1544, 0.1401, 0.1908, 0.1934, 0.1934, 0.1403, 0.1367, 0.141, 0.1672, 0.1597, 0.1793, 0.1931, 0.1835, 0.1522, 0.1898, 0.1702, 0.1354, 0.1527, 0.2159, 0.2388, 0.2614, 0.3039, 0.283, 0.2769, 0.2769, 0.2694, 0.2626, 0.2561, 0.2915, 0.2898, 0.3445, 0.3514, 0.3134, 0.3056, 0.2865, 0.2308, 0.2458, 0.2334, 0.1812, 0.2154, 0.2098, 0.2358, 0.1736, 0.1792, 0.1859, 0.2093, 0.1869, 0.1842, 0.1508, 0.183, 0.1998, 0.2155, 0.1398, 0.1421, 0.1063, 0.1812, 0.1936, 0.185, 0.185, 0.1946, 0.207, 0.2789, 0.2775, 0.2954, 0.2918, 0.3278, 0.3111, 0.27, 0.295, 0.2801, 0.2865, 0.3542, 0.3375, 0.4067, 0.4155, 0.3713, 0.3766, 0.3843, 0.39, 0.4037, 0.3786, 0.4663, 0.4476, 0.4389, 0.414, 0.3886, 0.3974, 0.4601, 0.4133, 0.384, 0.3723, 0.4038, 0.4231, 0.4139, 0.4139, 0.4412]}
{"unit_id": "orig:1995360938330046836", "ticker_idx": 1, "ts": "2025-12-01T05:14:54+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia has increased its orders for 12-high HBM3e in order to maximize GB300 shipments", "tweet_type": "analysis", "timeline": "quarters", "summary": "Daishin note: TSMC CoWoS ramp drives upward revision to 2026 GPU/ASIC/HBM volumes; NVDA, AVGO, AMD all raising HBM orders.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities December 1 Note – Key Points\n\n1. Amid the accelerated ramp-up of TSMC’s CoWoS, the total volume of GPUs and ASICs that can be shipped in 2026 is being revised upward. The same applies to HBM.\n    \n2. It is estimated that Nvidia has increased its orders for 12-high HBM3e in order to maximize GB300 shipments, and Broadcom is also believed to have raised its HBM3e orders to prepare for strong volume growth from major customers such as Google and OpenAI.\n    \n3. AMD is likely to place more HBM4 orders than expected. In the AMD market, attention should be paid to the larger-than-expected HBM4 purchasing TAM.\n\n$TSM $NVDA $AVGO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0162295101850789, "ret_signed_m1d": -0.0162295101850789, "alpha_spy_m1d": -0.020815864400725026, "alpha_c_m1d": -0.014302956854296056, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008559328912786102, "ret_signed_p1d": 0.008559328912786102, "alpha_spy_p1d": 0.006707119638674097, "alpha_c_p1d": -0.009771332097176177, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.03134909133996655, "ret_signed_p1w": 0.03134909133996655, "alpha_spy_p1w": 0.02640992676422016, "alpha_c_p1w": -0.012820177181220638, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.042410128700367755, "ret_signed_p1m": 0.042410128700367755, "alpha_spy_p1m": 0.029518113304120686, "alpha_c_p1m": 0.00988387547582148, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 0.027680540949180266, "ret_signed_p3m": 0.027680540949180266, "alpha_spy_p3m": 0.011412374035411066, "alpha_c_p3m": -0.14325104511514275, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 0.19432850523535183, "ret_signed_p6m": 0.19432850523535183, "alpha_spy_p6m": 0.08467522578184505, "alpha_c_p6m": -0.5169521570625641, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.0517, -0.046, -0.0718, -0.0646, -0.051, -0.0145, -0.0153, -0.0117, -0.0121, -0.028, -0.0535, -0.0205, -0.0181, 0.0205, -0.0083, -0.0164, -0.0124, -0.0096, 0.0107, 0.037, 0.0407, 0.0499, 0.0428, 0.0312, 0.0285, 0.0511, 0.0703, 0.018, 0.0467, 0.0006, -0.0005, 0.0105, 0.0183, 0.0151, 0.0069, 0.002, 0.0124, 0.0352, 0.0643, 0.1173, 0.1507, 0.1277, 0.1254, 0.1498, 0.1043, 0.085, 0.0454, 0.0457, 0.1063, 0.0736, 0.0771, 0.0386, 0.057, 0.0371, 0.008, 0.0367, 0.004, -0.0058, 0.0146, -0.0117, 0.0019, 0.0019, -0.0162, 0.0, 0.0086, -0.0018, 0.0193, 0.0139, 0.0313, 0.0281, 0.0215, 0.0057, -0.0272, -0.0201, -0.0122, -0.0499, -0.0321, 0.006, 0.021, 0.0517, 0.0484, 0.0484, 0.059, 0.0462, 0.0424, 0.0366, 0.0366, 0.0497, 0.0456, 0.0407, 0.0511, 0.0285, 0.0275, 0.028, 0.0328, 0.018, 0.0397, 0.0351, 0.0351, -0.0102, 0.019, 0.0274, 0.0431, 0.0365, 0.0479, 0.0645, 0.07, 0.0624, 0.0317, 0.0024, -0.0318, -0.0446, 0.0306, 0.0563, 0.048, 0.0564, 0.0391, 0.0161, 0.0161, 0.0281, 0.0449, 0.0444, 0.0551, 0.0647, 0.0719, 0.087, 0.0277, -0.0151, 0.0143, 0.0008, 0.0174, 0.0191, -0.0116, 0.0152, 0.027, 0.0341, 0.018, 0.0019, 0.0185, 0.0113, 0.0028, -0.0074, -0.04, -0.0237, -0.0261, -0.0068, -0.0481, -0.0688, -0.0819, -0.0306, -0.0231, -0.014, -0.014, -0.0126, -0.01, 0.0121, 0.0223, 0.0485, 0.0523, 0.0923, 0.1054, 0.1026, 0.1211, 0.1232, 0.1111, 0.1256, 0.1097, 0.1577, 0.2041, 0.1849, 0.1631, 0.1093, 0.1031, 0.1033, 0.0923, 0.1553, 0.1757, 0.1962, 0.2198, 0.2272, 0.2553, 0.3104, 0.2525, 0.2358, 0.2263, 0.2422, 0.2202, 0.1969, 0.1969, 0.1943]}
{"unit_id": "orig:1995360938330046836", "ticker_idx": 2, "ts": "2025-12-01T05:14:54+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom is also believed to have raised its HBM3e orders to prepare for strong volume growth from major customers such as Google and OpenAI", "tweet_type": "analysis", "timeline": "quarters", "summary": "Daishin note: TSMC CoWoS ramp drives upward revision to 2026 GPU/ASIC/HBM volumes; NVDA, AVGO, AMD all raising HBM orders.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities December 1 Note – Key Points\n\n1. Amid the accelerated ramp-up of TSMC’s CoWoS, the total volume of GPUs and ASICs that can be shipped in 2026 is being revised upward. The same applies to HBM.\n    \n2. It is estimated that Nvidia has increased its orders for 12-high HBM3e in order to maximize GB300 shipments, and Broadcom is also believed to have raised its HBM3e orders to prepare for strong volume growth from major customers such as Google and OpenAI.\n    \n3. AMD is likely to place more HBM4 orders than expected. In the AMD market, attention should be paid to the larger-than-expected HBM4 purchasing TAM.\n\n$TSM $NVDA $AVGO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.043721519431367595, "ret_signed_m1d": 0.043721519431367595, "alpha_spy_m1d": 0.03913516521572147, "alpha_c_m1d": 0.04564807276215044, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011681462535108822, "ret_signed_p1d": -0.011681462535108822, "alpha_spy_p1d": -0.013533671809220826, "alpha_c_p1d": -0.0300121235450711, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.038903924033987014, "ret_signed_p1w": 0.038903924033987014, "alpha_spy_p1w": 0.03396475945824062, "alpha_c_p1w": -0.005265344487200174, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": -0.09210683055887492, "ret_signed_p1m": -0.09210683055887492, "alpha_spy_p1m": -0.10499884595512199, "alpha_c_p1m": -0.1246330837834212, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": -0.16515866819962455, "ret_signed_p3m": -0.16515866819962455, "alpha_spy_p3m": -0.18142683511339375, "alpha_c_p3m": -0.33609025426394756, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 0.09745242184085567, "ret_signed_p6m": 0.09745242184085567, "alpha_spy_p6m": -0.012200857612651106, "alpha_c_p6m": -0.6138282404570603, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.2181, -0.2085, -0.1341, -0.1063, -0.1295, -0.0444, -0.0701, -0.0695, -0.0586, -0.0691, -0.1049, -0.107, -0.1081, -0.1225, -0.1221, -0.1211, -0.1295, -0.1335, -0.1507, -0.1455, -0.1365, -0.1241, -0.1236, -0.131, -0.1287, -0.1051, -0.1064, -0.1592, -0.0761, -0.1087, -0.09, -0.0827, -0.0952, -0.0954, -0.1125, -0.1186, -0.1082, -0.0828, -0.0622, -0.034, -0.0003, -0.0249, -0.0426, -0.0609, -0.0884, -0.0702, -0.079, -0.0949, -0.0717, -0.0884, -0.0799, -0.1194, -0.113, -0.1125, -0.1181, -0.082, -0.1017, -0.1188, -0.021, -0.0027, 0.0298, 0.0298, 0.0437, 0.0, -0.0117, -0.0142, -0.0131, 0.0108, 0.0389, 0.0523, 0.0696, 0.0526, -0.0677, -0.1198, -0.116, -0.1556, -0.1456, -0.1184, -0.1139, -0.0935, -0.0911, -0.0911, -0.0862, -0.0933, -0.0921, -0.1018, -0.1018, -0.0979, -0.1088, -0.1079, -0.1086, -0.1372, -0.1048, -0.086, -0.0798, -0.118, -0.1098, -0.0873, -0.0873, -0.1369, -0.1467, -0.1553, -0.1694, -0.157, -0.1364, -0.1352, -0.1417, -0.1402, -0.1407, -0.1687, -0.2006, -0.1942, -0.136, -0.1074, -0.1165, -0.1105, -0.1406, -0.1562, -0.1562, -0.137, -0.1345, -0.1333, -0.1367, -0.1427, -0.1553, -0.1376, -0.1652, -0.1707, -0.1726, -0.1856, -0.176, -0.1364, -0.1424, -0.1027, -0.111, -0.1136, -0.1281, -0.164, -0.1568, -0.1662, -0.1801, -0.17, -0.1942, -0.1613, -0.1723, -0.1709, -0.1953, -0.2181, -0.237, -0.1951, -0.1848, -0.182, -0.182, -0.1823, -0.1315, -0.0882, -0.077, -0.0338, -0.0124, -0.0098, 0.0317, 0.0362, 0.0572, 0.0393, 0.0459, 0.0991, 0.0921, 0.0994, 0.0875, 0.0398, 0.0544, 0.0855, 0.0956, 0.0831, 0.1114, 0.1064, 0.0729, 0.1182, 0.1141, 0.0904, 0.0839, 0.1437, 0.1057, 0.0941, 0.069, 0.0864, 0.0781, 0.077, 0.077, 0.0975]}
{"unit_id": "orig:1995360938330046836", "ticker_idx": 3, "ts": "2025-12-01T05:14:54+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD is likely to place more HBM4 orders than expected. In the AMD market, attention should be paid to the larger-than-expected HBM4 purchasing TAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Daishin note: TSMC CoWoS ramp drives upward revision to 2026 GPU/ASIC/HBM volumes; NVDA, AVGO, AMD all raising HBM orders.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities December 1 Note – Key Points\n\n1. Amid the accelerated ramp-up of TSMC’s CoWoS, the total volume of GPUs and ASICs that can be shipped in 2026 is being revised upward. The same applies to HBM.\n    \n2. It is estimated that Nvidia has increased its orders for 12-high HBM3e in order to maximize GB300 shipments, and Broadcom is also believed to have raised its HBM3e orders to prepare for strong volume growth from major customers such as Google and OpenAI.\n    \n3. AMD is likely to place more HBM4 orders than expected. In the AMD market, attention should be paid to the larger-than-expected HBM4 purchasing TAM.\n\n$TSM $NVDA $AVGO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010147414376048713, "ret_signed_m1d": -0.010147414376048713, "alpha_spy_m1d": -0.01473376859169484, "alpha_c_m1d": -0.00822086104526587, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020567842767812206, "ret_signed_p1d": -0.020567842767812206, "alpha_spy_p1d": -0.02242005204192421, "alpha_c_p1d": -0.038898503777774485, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.0061430930891002156, "ret_signed_p1w": 0.0061430930891002156, "alpha_spy_p1w": 0.0012039285133538247, "alpha_c_p1w": -0.03802617543208697, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": -0.020112842552914345, "ret_signed_p1m": -0.020112842552914345, "alpha_spy_p1m": -0.033004857949161415, "alpha_c_p1m": -0.05263909577746062, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": -0.07317074186837269, "ret_signed_p3m": -0.07317074186837269, "alpha_spy_p3m": -0.08943890878214189, "alpha_c_p3m": -0.2441023279326957, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 1.2929105717317593, "ret_signed_p6m": 1.2929105717317593, "alpha_spy_p6m": 1.1832572922782525, "alpha_c_p6m": 0.5816299094338433, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.2622, -0.2638, -0.3122, -0.311, -0.291, -0.274, -0.2916, -0.2784, -0.2667, -0.2698, -0.2758, -0.2814, -0.2838, -0.2729, -0.2678, -0.2679, -0.2662, -0.2744, -0.2657, -0.2638, -0.2537, -0.2277, -0.2507, -0.073, -0.0375, 0.0719, 0.0597, -0.0221, -0.0152, -0.0076, 0.0857, 0.0673, 0.0606, 0.0946, 0.0831, 0.0476, 0.0693, 0.1509, 0.1816, 0.1741, 0.2028, 0.1596, 0.1655, 0.1815, 0.1378, 0.1664, 0.0816, 0.0627, 0.1102, 0.0808, 0.1781, 0.1283, 0.1231, 0.0945, 0.0479, 0.0172, -0.0625, -0.0727, -0.0214, -0.062, -0.0251, -0.0251, -0.0101, 0.0, -0.0206, -0.0098, -0.0172, -0.0081, 0.0061, 0.0085, 0.0076, 0.0076, -0.0409, -0.0554, -0.0482, -0.0985, -0.0851, -0.0288, -0.0219, -0.0221, -0.0215, -0.0215, -0.0217, -0.0189, -0.0201, -0.0255, -0.0255, 0.0169, 0.006, -0.0246, -0.0443, -0.0686, -0.0755, -0.0549, 0.0055, 0.0175, 0.0371, 0.0549, 0.0549, 0.0553, 0.1367, 0.1546, 0.1817, 0.1436, 0.1468, 0.1501, 0.1475, 0.0772, 0.1206, 0.1017, -0.0891, -0.124, -0.0515, -0.0171, -0.0282, -0.0281, -0.0629, -0.0566, -0.0566, -0.0759, -0.0894, -0.0746, -0.0892, -0.1054, -0.0269, -0.0405, -0.0732, -0.089, -0.0962, -0.1311, -0.0805, -0.0924, -0.1244, -0.0777, -0.0752, -0.0679, -0.1002, -0.12, -0.1055, -0.1067, -0.0924, -0.0659, -0.0839, -0.0777, -0.0655, 0.0023, -0.0728, -0.0809, -0.1079, -0.0743, -0.0435, -0.0103, -0.0103, 0.0019, 0.0081, 0.0549, 0.0768, 0.115, 0.1232, 0.1607, 0.1746, 0.2662, 0.2668, 0.2511, 0.2945, 0.3809, 0.3894, 0.5827, 0.5227, 0.4707, 0.534, 0.6131, 0.6406, 0.5542, 0.6166, 0.9175, 0.8587, 1.0713, 1.0877, 1.0399, 1.0272, 1.0463, 0.9298, 0.9157, 0.8841, 1.0367, 1.0458, 1.1274, 1.1274, 1.2929]}
{"unit_id": "orig:1995360938330046836", "ticker_idx": 4, "ts": "2025-12-01T05:14:54+00:00", "symbol": "HBM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The same applies to HBM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Daishin note: TSMC CoWoS ramp drives upward revision to 2026 GPU/ASIC/HBM volumes; NVDA, AVGO, AMD all raising HBM orders.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities December 1 Note – Key Points\n\n1. Amid the accelerated ramp-up of TSMC’s CoWoS, the total volume of GPUs and ASICs that can be shipped in 2026 is being revised upward. The same applies to HBM.\n    \n2. It is estimated that Nvidia has increased its orders for 12-high HBM3e in order to maximize GB300 shipments, and Broadcom is also believed to have raised its HBM3e orders to prepare for strong volume growth from major customers such as Google and OpenAI.\n    \n3. AMD is likely to place more HBM4 orders than expected. In the AMD market, attention should be paid to the larger-than-expected HBM4 purchasing TAM.\n\n$TSM $NVDA $AVGO $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0114658706649311, "ret_signed_m1d": -0.0114658706649311, "alpha_spy_m1d": -0.016052224880577226, "alpha_c_m1d": -0.009539317334148256, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019644831246580636, "ret_signed_p1d": 0.019644831246580636, "alpha_spy_p1d": 0.01779262197246863, "alpha_c_p1d": 0.0013141702366183568, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.061888469616404373, "ret_signed_p1w": 0.061888469616404373, "alpha_spy_p1w": 0.05694930504065798, "alpha_c_p1w": 0.017719201095217185, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.20758480185732853, "ret_signed_p1m": 0.20758480185732853, "alpha_spy_p1m": 0.19469278646108146, "alpha_c_p1m": 0.17505854863278225, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 0.9828726284367323, "ret_signed_p3m": 0.9828726284367323, "alpha_spy_p3m": 0.9666044615229631, "alpha_c_p3m": 0.8119410423724093, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.5122622366497445, "ret_signed_p6m": 2.5122622366497445, "alpha_spy_p6m": 2.4026089571962377, "alpha_c_p6m": 1.8009815743518285, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.4432, -0.4328, -0.4199, -0.4156, -0.3989, -0.3788, -0.3597, -0.3306, -0.3246, -0.3031, -0.3144, -0.2816, -0.2901, -0.2797, -0.2671, -0.2735, -0.2786, -0.2995, -0.2784, -0.2756, -0.2404, -0.2038, -0.198, -0.1936, -0.2009, -0.1859, -0.1917, -0.1711, -0.1673, -0.183, -0.1582, -0.116, -0.1075, -0.0884, -0.1006, -0.1007, -0.0981, -0.0539, -0.0264, -0.0408, -0.008, 0.0064, 0.0118, 0.0766, 0.0119, 0.0204, 0.0256, 0.0127, 0.059, 0.0597, 0.0624, 0.0473, 0.0103, 0.0433, -0.0069, -0.0198, -0.0347, -0.0765, -0.0479, -0.039, -0.0164, -0.0015, -0.0115, 0.0, 0.0196, 0.0122, -0.0024, 0.0244, 0.0619, 0.0591, 0.0864, 0.0632, 0.0482, 0.019, -0.0094, 0.0108, 0.0424, 0.0591, 0.1082, 0.1135, 0.1291, 0.1291, 0.1528, 0.2019, 0.2076, 0.1976, 0.1976, 0.2839, 0.323, 0.3876, 0.3991, 0.3831, 0.4014, 0.4049, 0.3834, 0.3882, 0.4091, 0.4662, 0.4725, 0.449, 0.4949, 0.5252, 0.5349, 0.501, 0.5946, 0.6644, 0.6718, 0.6719, 0.6212, 0.7002, 0.6457, 0.582, 0.5942, 0.6343, 0.6113, 0.6595, 0.7178, 0.7182, 0.7182, 0.7018, 0.7311, 0.764, 0.8135, 0.8143, 0.867, 0.9019, 0.9829, 0.9499, 0.9503, 0.7577, 0.655, 0.7714, 0.7125, 0.6354, 0.7656, 0.8048, 0.7638, 0.7656, 0.8443, 0.8867, 0.9886, 0.9112, 0.8741, 0.7589, 0.7909, 0.7755, 0.6705, 0.673, 0.5754, 0.5264, 0.6986, 0.6174, 0.6719, 0.717, 0.7467, 0.9064, 0.8836, 0.9069, 0.9066, 1.0183, 1.0407, 1.0755, 1.0508, 1.0601, 1.1125, 1.1597, 1.1763, 1.1786, 1.2773, 1.2458, 1.2744, 1.25, 1.2847, 1.4721, 1.5606, 1.8043, 1.8278, 1.9767, 2.221, 2.1316, 2.2872, 2.2851, 2.0352, 2.0234, 1.9699, 2.0177, 2.2588, 2.2207, 2.2207, 2.5123]}
{"unit_id": "reply:1997644001155993703", "ticker_idx": 0, "ts": "2025-12-07T12:26:58+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If the stock even looks like it might go up a little, Masa will probably dump it", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bearish ARM: SoftBank's Masa Son overhang means any rally will be sold into.", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@xfan0209 Masa is holding way too much.\nIf the stock even looks like it might go up a little, Masa will probably dump it.", "reply_text": null, "parent_text": "@jukan05 What's your take on ARM? The stock has been flatlining for 2 years", "quoted_text": null, "parent_author": "xfan0209", "ret_m1d": 0.010945763289872623, "ret_signed_m1d": -0.010945763289872623, "alpha_spy_m1d": -0.007932432894446562, "alpha_c_m1d": -0.022151754393091028, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015381269961799449, "ret_signed_p1d": -0.015381269961799449, "alpha_spy_p1d": -0.016244414928761053, "alpha_c_p1d": -0.014160257077621585, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.1102446427764403, "ret_signed_p1w": 0.1102446427764403, "alpha_spy_p1w": 0.106002545203976, "alpha_c_p1w": 0.06778097102457337, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": -0.17348690951334966, "ret_signed_p1m": 0.17348690951334966, "alpha_spy_p1m": 0.18844267165024264, "alpha_c_p1m": 0.2285028429008672, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": -0.13707251538159748, "ret_signed_p3m": 0.13707251538159748, "alpha_spy_p3m": 0.13662374823867918, "alpha_c_p3m": 0.21312782975711375, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1027, 0.1067, 0.0777, 0.1007, 0.1007, 0.0972, 0.0484, 0.0224, 0.0336, 0.0087, 0.0323, 0.0062, -0.0011, 0.0001, 0.0122, 0.0758, 0.0885, 0.092, 0.1176, 0.14, 0.1931, 0.2209, 0.1075, 0.2301, 0.203, 0.221, 0.2247, 0.1848, 0.2269, 0.2118, 0.1855, 0.1919, 0.2211, 0.2779, 0.2383, 0.219, 0.1836, 0.2149, 0.2068, 0.1499, 0.146, 0.1321, 0.0901, 0.1077, 0.0713, 0.0642, 0.0038, -0.0001, 0.0034, -0.0268, -0.02, -0.0519, -0.0587, -0.0363, -0.0597, -0.0513, -0.0513, -0.0302, -0.0341, -0.0236, -0.0042, 0.0051, 0.0109, 0.0, 0.0154, 0.0124, -0.026, -0.0636, -0.1102, -0.1336, -0.1803, -0.1879, -0.1842, -0.1895, -0.1986, -0.202, -0.202, -0.2111, -0.2094, -0.2069, -0.218, -0.218, -0.1792, -0.1693, -0.1735, -0.1724, -0.191, -0.2002, -0.2049, -0.2285, -0.2489, -0.248, -0.2432, -0.2432, -0.2333, -0.185, -0.1472, -0.1696, -0.1792, -0.1781, -0.2133, -0.2243, -0.2462, -0.235, -0.252, -0.2495, -0.2068, -0.115, -0.1085, -0.0989, -0.1037, -0.1258, -0.1037, -0.1037, -0.0922, -0.0897, -0.0919, -0.1016, -0.1145, -0.0833, -0.0575, -0.0753, -0.0882, -0.1102, -0.1292, -0.1121, -0.1371, -0.1817, -0.1585, -0.1376, -0.1408, -0.1764, -0.1719, -0.1293, -0.0892, -0.0817, -0.0713, -0.0532, -0.0207, -0.0345, 0.1237, 0.1075, 0.0311, -0.0202, 0.0823, 0.1094, 0.0667, 0.0667, 0.0643, 0.0292, 0.0653, 0.0716, 0.0655, 0.1273, 0.1534, 0.1399, 0.1613, 0.1928, 0.2527, 0.2555, 0.4063, 0.4638, 0.6799, 0.5444, 0.4212, 0.4429, 0.5047, 0.5108, 0.4541, 0.4941, 0.6977, 0.526, 0.5258, 0.5213, 0.4875, 0.5826, 0.6347, 0.4964, 0.539, 0.5964, 0.8367, 1.1336, 1.1928, 1.1928, 1.298, 1.1656, 1.3986, 1.5275, null, null]}
{"unit_id": "orig:1996080008838996469", "ticker_idx": 0, "ts": "2025-12-03T04:52:14+00:00", "symbol": "ASE Technology Holding", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASE is expected to be a major beneficiary of expanded outsourcing from TSMC", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan: ASE, Amkor, Advantest, Chroma, and GPTC are key beneficiaries of AI advanced packaging demand surge and CoWoS/test market growth into 2026-27.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ticker ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251203_ Asian Semis: Mapping out the Advanced Back-end market for AI - JPMorgan\n\n[Key Takeaways]\n\n(1) CoWoS demand is expected to exceed supply until 2027.\n\n- Robust demand for GPUs/ASICs, 2) N3 yield bottlenecks (due to EUV usage, etc.), 3) Preparation for CoPoS.\n\n(2) The trickle-down effect to OSATs is expected to continue, with high growth in the Test market fully materializing in 2026.\n\n(3) Meanwhile, TSMC's CoPoS pilot line is scheduled for completion in early 2026. A follow-up on the related value chain in 2H26 is necessary ahead of the scheduled full-scale ramp-up in 2027.\n\n[Contents]\n\n(1) Due to the recent surge in AI chip demand and limited expansion of packaging CAPA, there are numerous inquiries regarding AI and 2.5D/3D chip advanced packaging.\n\n(2) First, ASE is expected to be a major beneficiary of expanded outsourcing from TSMC, specifically in 1) On-Substrate & Wafer Probing, and 2) Full-process 2.5D packaging.\n\n(3) Amkor is also expected to benefit from increased outsourcing by Broadcom.\n\n(4) In the test equipment sector, Chroma and Advantest are expected to benefit, primarily due to \"increased test intensity.\"\n\n(5) Among packaging equipment vendors, we view GPTC positively due to the potential for CoWoS CAPA expansion and increased orders for CoPoS starting from 2H26.\n\n(6) CoWoS demand in '26–'27 is projected to exceed supply by +20–30%.\n\n- CoWoS Capacity: 105K (End of '26) → 125K (End of '27).\n\n(7) Even with outsourcing to external vendors, the excess demand situation is expected to persist for customers like NVIDIA, Broadcom, AMD, and Amazon.\n(8) TSMC appears to be proceeding cautiously with CAPA expansion, centering on CoWoS-L, and it will likely remain difficult to satisfy all overall customer demand.\n\n(9) Furthermore, as the N3 Front-end process is likely to become a bottleneck, it is judged that aggressive CoWoS expansion in '26 will be difficult.\n\n(10) AP07 and AP08A are already filled with capacity expansions for WMCM/SoIC and CoWoS-L, and some processes have been offloaded to OSATs.\n\n(11) Additionally, due to the possibility of technological replacement by CoPoS in the future, TSMC continues to maintain a cautious stance on CoWoS CAPA expansion beyond '27.\n\n(12) Meanwhile, ASE is already receiving outsourcing orders for On-Substrate and Wafer Probing, which is projected to account for more than 80% of its LEAP (Leading Edge Advanced Packaging) revenue in '26.\n\n(13) ASE's 2.5D Full Processing packaging CAPA is expected to reach 15–20K by the end of '26.\n\n- Major customer: AMD (Venice CPU, legacy MI series, etc.).\n\n(14) There is also a probability of securing orders for NVIDIA's Vera CPU and a portion of Broadcom's Tomahawk Switch IC.\n- '27 Outlook: CoWoS revenue ~$1bn, LEAP total revenue ~$5bn.\n\n(15) ASE has executed CAPEX 3–4 times higher than Amkor over the past two years and is projected to capture approximately 70–80% of TSMC's outsourcing demand.\n\n(16) In Amkor's case, while it struggled due to the blocking of NVIDIA H20, it is entering a recovery phase with increased CoWoS-S outsourcing from Broadcom in '26.\n\n(17) Its 2.5D packaging capacity is expected to grow from the existing 6–8K to 12–15K by the end of '26.\n\n- Major customers: Broadcom (Tomahawk 6, potential for TPUs), NVIDIA (GB10, ARM CPU, potential for Vera CPU).\n\n(18) Limited CAPA within Taiwan is a bottleneck for securing TSMC outsourcing, and expansion at the Korea K5 plant is slow. Meaningful growth is difficult until the Arizona Peoria facility comes online in '28.\n\n(19) TSMC's CoPoS (310x310mm) pilot line is expected to be completed in early '26.\n\n(20) If yield issues are resolved, there is a possibility that CoPoS will be adopted in NVIDIA's Feynman Project in '28.\n\n(21) Orders for CoPoS CAPA are expected to start appearing as early as 2H26, with ramp-up anticipated in '27.\n\n(22) Lastly, the Test market is projected to increase very strongly in '26.\n\n(23) Factors such as 1) N3 process transition, 2) expanded adoption of Chiplets, and 3) introduction of HBM4 will drive longer test times, strongly boosting demand for new testers.\n\n(24) For NVIDIA GPUs, KYEC is expected to maintain a high market share for Blackwell and Rubin, while ASE may capture a small share of Rubin by the end of '26.\n\n(25) According to our channel checks, certification for Advantest's Burn-in Oven is currently not proceeding smoothly, which will be a limiting factor for ASE's Final Test share expansion.\n\n(26) In ASICs, KYEC is expected to continue its good positioning in Broadcom testing (Alphabet TPU + Meta MTIA), while ASE/Amkor are projected to benefit from the ramp-up of Amazon Trainium 3.\n\n(27) For AMD, Chroma's SLT (System Level Test) is expected to be qualified for the MI450, and it is anticipated to sustain a 100% share in NVIDIA Rubin SLT thereafter.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00844150415498568, "ret_signed_m1d": -0.00844150415498568, "alpha_spy_m1d": -0.004990685744577283, "alpha_c_m1d": 0.005116160363138, "bench_spy_m1d": -0.003450818410408396, "bench_c_m1d": -0.01355766451812368, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015584401107996193, "ret_signed_p1d": -0.015584401107996193, "alpha_spy_p1d": -0.01631545726199457, "alpha_c_p1d": -0.00798222209560906, "bench_spy_p1d": 0.0007310561539983773, "bench_c_p1d": -0.0076021790123871336, "ret_p1w": 0.06688316312723441, "ret_signed_p1w": 0.06688316312723441, "alpha_spy_p1w": 0.061502159008816726, "alpha_c_p1w": 0.040179507467649644, "bench_spy_p1w": 0.005381004118417687, "bench_c_p1w": 0.02670365565958477, "ret_p1m": 0.04545459612201341, "ret_signed_p1m": 0.04545459612201341, "alpha_spy_p1m": 0.04538880555591396, "alpha_c_p1m": 0.054015722057100035, "bench_spy_p1m": 6.57905660994551e-05, "bench_c_p1m": -0.008561125935086622, "ret_p3m": 0.5714286598955789, "ret_signed_p3m": 0.5714286598955789, "alpha_spy_p3m": 0.5648220484373396, "alpha_c_p3m": 0.45263585426590347, "bench_spy_p3m": 0.006606611458239264, "bench_c_p3m": 0.11879280562967542, "ret_p6m": 1.6363636025853245, "ret_signed_p6m": 1.6363636025853245, "alpha_spy_p6m": 1.5266871386891319, "alpha_c_p6m": 0.9850299488748873, "bench_spy_p6m": 0.10967646389619268, "bench_c_p6m": 0.6513336537104373, "price_path": [-0.3201, -0.2526, -0.2734, -0.2662, -0.2786, -0.2825, -0.2727, -0.2662, -0.2662, -0.2552, -0.2558, -0.2396, -0.2403, -0.2506, -0.2558, -0.2747, -0.2773, -0.2799, -0.2773, -0.276, -0.2805, -0.2604, -0.261, -0.2299, -0.2396, -0.2805, -0.2383, -0.2649, -0.2253, -0.1831, -0.174, -0.1565, -0.1766, -0.1909, -0.1675, -0.1571, -0.1409, -0.1325, -0.063, -0.0234, 0.0396, 0.0513, -0.0019, 0.0279, -0.0019, -0.0117, -0.0013, -0.0273, -0.0266, -0.05, -0.0513, -0.0714, -0.0799, -0.0864, -0.1071, -0.1026, -0.0805, -0.0792, -0.0649, -0.0649, -0.0292, -0.037, -0.0084, 0.0, -0.0156, -0.0123, 0.0221, 0.037, 0.0669, 0.0649, 0.0234, 0.0117, -0.0065, -0.0422, -0.0344, -0.0156, -0.0032, 0.0084, 0.0097, 0.0097, 0.0208, 0.0403, 0.0409, 0.0455, 0.0455, 0.0948, 0.1013, 0.1364, 0.1442, 0.1253, 0.1455, 0.1864, 0.2143, 0.226, 0.2461, 0.2604, 0.2604, 0.2273, 0.2325, 0.2377, 0.2591, 0.2877, 0.3032, 0.3156, 0.2844, 0.2325, 0.2604, 0.2591, 0.2331, 0.3156, 0.3565, 0.4383, 0.4604, 0.5416, 0.5234, 0.5195, 0.5195, 0.524, 0.5156, 0.5052, 0.5558, 0.5247, 0.6117, 0.6305, 0.5734, 0.5773, 0.5714, 0.4487, 0.4351, 0.4325, 0.3714, 0.4065, 0.4097, 0.4221, 0.363, 0.3961, 0.3974, 0.4175, 0.4065, 0.4221, 0.3838, 0.4078, 0.3981, 0.4545, 0.3916, 0.3961, 0.3565, 0.4078, 0.4597, 0.4468, 0.4468, 0.4643, 0.4409, 0.5675, 0.5916, 0.613, 0.724, 0.7448, 0.7682, 0.7922, 0.8565, 0.9013, 0.9221, 0.9448, 0.9357, 1.0779, 1.0032, 0.9532, 0.9864, 1.0396, 1.0513, 1.0961, 1.1714, 1.2182, 1.1649, 1.2227, 1.2922, 1.226, 1.3032, 1.289, 1.1955, 1.0552, 1.0058, 1.0578, 1.1195, 1.2604, 1.2604, 1.5292, 1.537, 1.6364]}
{"unit_id": "orig:1996080008838996469", "ticker_idx": 1, "ts": "2025-12-03T04:52:14+00:00", "symbol": "Amkor Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Amkor is also expected to benefit from increased outsourcing by Broadcom", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan: ASE, Amkor, Advantest, Chroma, and GPTC are key beneficiaries of AI advanced packaging demand surge and CoWoS/test market growth into 2026-27.", "resolved_tickers": ["AMKR"], "resolution_method": "exchange_ticker", "resolution_reason": "Amkor Technology AMKR US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "251203_ Asian Semis: Mapping out the Advanced Back-end market for AI - JPMorgan\n\n[Key Takeaways]\n\n(1) CoWoS demand is expected to exceed supply until 2027.\n\n- Robust demand for GPUs/ASICs, 2) N3 yield bottlenecks (due to EUV usage, etc.), 3) Preparation for CoPoS.\n\n(2) The trickle-down effect to OSATs is expected to continue, with high growth in the Test market fully materializing in 2026.\n\n(3) Meanwhile, TSMC's CoPoS pilot line is scheduled for completion in early 2026. A follow-up on the related value chain in 2H26 is necessary ahead of the scheduled full-scale ramp-up in 2027.\n\n[Contents]\n\n(1) Due to the recent surge in AI chip demand and limited expansion of packaging CAPA, there are numerous inquiries regarding AI and 2.5D/3D chip advanced packaging.\n\n(2) First, ASE is expected to be a major beneficiary of expanded outsourcing from TSMC, specifically in 1) On-Substrate & Wafer Probing, and 2) Full-process 2.5D packaging.\n\n(3) Amkor is also expected to benefit from increased outsourcing by Broadcom.\n\n(4) In the test equipment sector, Chroma and Advantest are expected to benefit, primarily due to \"increased test intensity.\"\n\n(5) Among packaging equipment vendors, we view GPTC positively due to the potential for CoWoS CAPA expansion and increased orders for CoPoS starting from 2H26.\n\n(6) CoWoS demand in '26–'27 is projected to exceed supply by +20–30%.\n\n- CoWoS Capacity: 105K (End of '26) → 125K (End of '27).\n\n(7) Even with outsourcing to external vendors, the excess demand situation is expected to persist for customers like NVIDIA, Broadcom, AMD, and Amazon.\n(8) TSMC appears to be proceeding cautiously with CAPA expansion, centering on CoWoS-L, and it will likely remain difficult to satisfy all overall customer demand.\n\n(9) Furthermore, as the N3 Front-end process is likely to become a bottleneck, it is judged that aggressive CoWoS expansion in '26 will be difficult.\n\n(10) AP07 and AP08A are already filled with capacity expansions for WMCM/SoIC and CoWoS-L, and some processes have been offloaded to OSATs.\n\n(11) Additionally, due to the possibility of technological replacement by CoPoS in the future, TSMC continues to maintain a cautious stance on CoWoS CAPA expansion beyond '27.\n\n(12) Meanwhile, ASE is already receiving outsourcing orders for On-Substrate and Wafer Probing, which is projected to account for more than 80% of its LEAP (Leading Edge Advanced Packaging) revenue in '26.\n\n(13) ASE's 2.5D Full Processing packaging CAPA is expected to reach 15–20K by the end of '26.\n\n- Major customer: AMD (Venice CPU, legacy MI series, etc.).\n\n(14) There is also a probability of securing orders for NVIDIA's Vera CPU and a portion of Broadcom's Tomahawk Switch IC.\n- '27 Outlook: CoWoS revenue ~$1bn, LEAP total revenue ~$5bn.\n\n(15) ASE has executed CAPEX 3–4 times higher than Amkor over the past two years and is projected to capture approximately 70–80% of TSMC's outsourcing demand.\n\n(16) In Amkor's case, while it struggled due to the blocking of NVIDIA H20, it is entering a recovery phase with increased CoWoS-S outsourcing from Broadcom in '26.\n\n(17) Its 2.5D packaging capacity is expected to grow from the existing 6–8K to 12–15K by the end of '26.\n\n- Major customers: Broadcom (Tomahawk 6, potential for TPUs), NVIDIA (GB10, ARM CPU, potential for Vera CPU).\n\n(18) Limited CAPA within Taiwan is a bottleneck for securing TSMC outsourcing, and expansion at the Korea K5 plant is slow. Meaningful growth is difficult until the Arizona Peoria facility comes online in '28.\n\n(19) TSMC's CoPoS (310x310mm) pilot line is expected to be completed in early '26.\n\n(20) If yield issues are resolved, there is a possibility that CoPoS will be adopted in NVIDIA's Feynman Project in '28.\n\n(21) Orders for CoPoS CAPA are expected to start appearing as early as 2H26, with ramp-up anticipated in '27.\n\n(22) Lastly, the Test market is projected to increase very strongly in '26.\n\n(23) Factors such as 1) N3 process transition, 2) expanded adoption of Chiplets, and 3) introduction of HBM4 will drive longer test times, strongly boosting demand for new testers.\n\n(24) For NVIDIA GPUs, KYEC is expected to maintain a high market share for Blackwell and Rubin, while ASE may capture a small share of Rubin by the end of '26.\n\n(25) According to our channel checks, certification for Advantest's Burn-in Oven is currently not proceeding smoothly, which will be a limiting factor for ASE's Final Test share expansion.\n\n(26) In ASICs, KYEC is expected to continue its good positioning in Broadcom testing (Alphabet TPU + Meta MTIA), while ASE/Amkor are projected to benefit from the ramp-up of Amazon Trainium 3.\n\n(27) For AMD, Chroma's SLT (System Level Test) is expected to be qualified for the MI450, and it is anticipated to sustain a 100% share in NVIDIA Rubin SLT thereafter.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07657638650418253, "ret_signed_m1d": -0.07657638650418253, "alpha_spy_m1d": -0.07312556809377413, "alpha_c_m1d": -0.06301872198605885, "bench_spy_m1d": -0.003450818410408396, "bench_c_m1d": -0.01355766451812368, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017072617524976796, "ret_signed_p1d": -0.017072617524976796, "alpha_spy_p1d": -0.017803673678975174, "alpha_c_p1d": -0.009470438512589663, "bench_spy_p1d": 0.0007310561539983773, "bench_c_p1d": -0.0076021790123871336, "ret_p1w": 0.06646933985594705, "ret_signed_p1w": 0.06646933985594705, "alpha_spy_p1w": 0.061088335737529365, "alpha_c_p1w": 0.03976568419636228, "bench_spy_p1w": 0.005381004118417687, "bench_c_p1w": 0.02670365565958477, "ret_p1m": -0.1012975422484903, "ret_signed_p1m": -0.1012975422484903, "alpha_spy_p1m": -0.10136333281458976, "alpha_c_p1m": -0.09273641631340368, "bench_spy_p1m": 6.57905660994551e-05, "bench_c_p1m": -0.008561125935086622, "ret_p3m": 0.08832229376800571, "ret_signed_p3m": 0.08832229376800571, "alpha_spy_p3m": 0.08171568230976645, "alpha_c_p3m": -0.030470511861669713, "bench_spy_p3m": 0.006606611458239264, "bench_c_p3m": 0.11879280562967542, "ret_p6m": 0.60972566376378, "ret_signed_p6m": 0.60972566376378, "alpha_spy_p6m": 0.5000491998675873, "alpha_c_p6m": -0.041607989946657264, "bench_spy_p6m": 0.10967646389619268, "bench_c_p6m": 0.6513336537104373, "price_path": [-0.4403, -0.436, -0.435, -0.4369, -0.4235, -0.4216, -0.4166, -0.3857, -0.3953, -0.3376, -0.3569, -0.3292, -0.328, -0.333, -0.3376, -0.3405, -0.3521, -0.3549, -0.3305, -0.3267, -0.3326, -0.3056, -0.3242, -0.3046, -0.3042, -0.358, -0.3081, -0.3206, -0.2894, -0.2844, -0.2892, -0.2667, -0.2697, -0.301, -0.2735, -0.2556, -0.2463, -0.2694, -0.2574, -0.2749, -0.2667, -0.1404, -0.1992, -0.1697, -0.2015, -0.2099, -0.2002, -0.2217, -0.217, -0.2742, -0.2803, -0.2983, -0.3028, -0.286, -0.3142, -0.2719, -0.2472, -0.2322, -0.2067, -0.2067, -0.1733, -0.1427, -0.0766, 0.0, -0.0171, -0.0159, 0.0178, 0.0155, 0.0665, 0.0708, 0.0087, -0.0005, -0.0831, -0.1152, -0.1054, -0.0806, -0.0715, -0.0699, -0.0722, -0.0722, -0.076, -0.0874, -0.0804, -0.1013, -0.1013, -0.023, 0.0956, 0.1675, 0.2001, 0.1609, 0.1903, 0.1741, 0.1826, 0.1079, 0.12, 0.0926, 0.0926, 0.1165, 0.2169, 0.1939, 0.1316, 0.1471, 0.155, 0.16, 0.1382, 0.1002, 0.0972, 0.0526, -0.0014, 0.0077, 0.1236, 0.1955, 0.2169, 0.2786, 0.1744, 0.0808, 0.0808, 0.0662, 0.064, 0.1004, 0.0913, 0.0715, 0.1047, 0.1596, 0.1045, 0.0885, 0.0883, 0.015, 0.0419, 0.0143, -0.0528, -0.0162, 0.0002, -0.0002, -0.0588, -0.0195, 0.0179, 0.0569, 0.069, 0.0977, 0.0448, 0.0519, 0.1463, 0.1388, 0.0295, 0.0138, -0.0592, 0.027, 0.0603, 0.0651, 0.0651, 0.0726, 0.0861, 0.1955, 0.2576, 0.3219, 0.3785, 0.3985, 0.362, 0.4346, 0.5365, 0.5837, 0.6013, 0.6478, 0.6629, 0.7812, 0.7247, 0.6275, 0.6104, 0.5908, 0.6214, 0.6184, 0.7495, 0.7612, 0.6483, 0.7473, 0.7491, 0.6752, 0.7016, 0.6442, 0.6045, 0.5062, 0.4948, 0.5621, 0.503, 0.4996, 0.4996, 0.6754, 0.6464, 0.6097]}
{"unit_id": "orig:1996080008838996469", "ticker_idx": 2, "ts": "2025-12-03T04:52:14+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Chroma and Advantest are expected to benefit, primarily due to 'increased test intensity'", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan: ASE, Amkor, Advantest, Chroma, and GPTC are key beneficiaries of AI advanced packaging demand surge and CoWoS/test market growth into 2026-27.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "251203_ Asian Semis: Mapping out the Advanced Back-end market for AI - JPMorgan\n\n[Key Takeaways]\n\n(1) CoWoS demand is expected to exceed supply until 2027.\n\n- Robust demand for GPUs/ASICs, 2) N3 yield bottlenecks (due to EUV usage, etc.), 3) Preparation for CoPoS.\n\n(2) The trickle-down effect to OSATs is expected to continue, with high growth in the Test market fully materializing in 2026.\n\n(3) Meanwhile, TSMC's CoPoS pilot line is scheduled for completion in early 2026. A follow-up on the related value chain in 2H26 is necessary ahead of the scheduled full-scale ramp-up in 2027.\n\n[Contents]\n\n(1) Due to the recent surge in AI chip demand and limited expansion of packaging CAPA, there are numerous inquiries regarding AI and 2.5D/3D chip advanced packaging.\n\n(2) First, ASE is expected to be a major beneficiary of expanded outsourcing from TSMC, specifically in 1) On-Substrate & Wafer Probing, and 2) Full-process 2.5D packaging.\n\n(3) Amkor is also expected to benefit from increased outsourcing by Broadcom.\n\n(4) In the test equipment sector, Chroma and Advantest are expected to benefit, primarily due to \"increased test intensity.\"\n\n(5) Among packaging equipment vendors, we view GPTC positively due to the potential for CoWoS CAPA expansion and increased orders for CoPoS starting from 2H26.\n\n(6) CoWoS demand in '26–'27 is projected to exceed supply by +20–30%.\n\n- CoWoS Capacity: 105K (End of '26) → 125K (End of '27).\n\n(7) Even with outsourcing to external vendors, the excess demand situation is expected to persist for customers like NVIDIA, Broadcom, AMD, and Amazon.\n(8) TSMC appears to be proceeding cautiously with CAPA expansion, centering on CoWoS-L, and it will likely remain difficult to satisfy all overall customer demand.\n\n(9) Furthermore, as the N3 Front-end process is likely to become a bottleneck, it is judged that aggressive CoWoS expansion in '26 will be difficult.\n\n(10) AP07 and AP08A are already filled with capacity expansions for WMCM/SoIC and CoWoS-L, and some processes have been offloaded to OSATs.\n\n(11) Additionally, due to the possibility of technological replacement by CoPoS in the future, TSMC continues to maintain a cautious stance on CoWoS CAPA expansion beyond '27.\n\n(12) Meanwhile, ASE is already receiving outsourcing orders for On-Substrate and Wafer Probing, which is projected to account for more than 80% of its LEAP (Leading Edge Advanced Packaging) revenue in '26.\n\n(13) ASE's 2.5D Full Processing packaging CAPA is expected to reach 15–20K by the end of '26.\n\n- Major customer: AMD (Venice CPU, legacy MI series, etc.).\n\n(14) There is also a probability of securing orders for NVIDIA's Vera CPU and a portion of Broadcom's Tomahawk Switch IC.\n- '27 Outlook: CoWoS revenue ~$1bn, LEAP total revenue ~$5bn.\n\n(15) ASE has executed CAPEX 3–4 times higher than Amkor over the past two years and is projected to capture approximately 70–80% of TSMC's outsourcing demand.\n\n(16) In Amkor's case, while it struggled due to the blocking of NVIDIA H20, it is entering a recovery phase with increased CoWoS-S outsourcing from Broadcom in '26.\n\n(17) Its 2.5D packaging capacity is expected to grow from the existing 6–8K to 12–15K by the end of '26.\n\n- Major customers: Broadcom (Tomahawk 6, potential for TPUs), NVIDIA (GB10, ARM CPU, potential for Vera CPU).\n\n(18) Limited CAPA within Taiwan is a bottleneck for securing TSMC outsourcing, and expansion at the Korea K5 plant is slow. Meaningful growth is difficult until the Arizona Peoria facility comes online in '28.\n\n(19) TSMC's CoPoS (310x310mm) pilot line is expected to be completed in early '26.\n\n(20) If yield issues are resolved, there is a possibility that CoPoS will be adopted in NVIDIA's Feynman Project in '28.\n\n(21) Orders for CoPoS CAPA are expected to start appearing as early as 2H26, with ramp-up anticipated in '27.\n\n(22) Lastly, the Test market is projected to increase very strongly in '26.\n\n(23) Factors such as 1) N3 process transition, 2) expanded adoption of Chiplets, and 3) introduction of HBM4 will drive longer test times, strongly boosting demand for new testers.\n\n(24) For NVIDIA GPUs, KYEC is expected to maintain a high market share for Blackwell and Rubin, while ASE may capture a small share of Rubin by the end of '26.\n\n(25) According to our channel checks, certification for Advantest's Burn-in Oven is currently not proceeding smoothly, which will be a limiting factor for ASE's Final Test share expansion.\n\n(26) In ASICs, KYEC is expected to continue its good positioning in Broadcom testing (Alphabet TPU + Meta MTIA), while ASE/Amkor are projected to benefit from the ramp-up of Amazon Trainium 3.\n\n(27) For AMD, Chroma's SLT (System Level Test) is expected to be qualified for the MI450, and it is anticipated to sustain a 100% share in NVIDIA Rubin SLT thereafter.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.050335549075446906, "ret_signed_m1d": -0.050335549075446906, "alpha_spy_m1d": -0.04688473066503851, "alpha_c_m1d": -0.036777884557323226, "bench_spy_m1d": -0.003450818410408396, "bench_c_m1d": -0.01355766451812368, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0076701217592450854, "ret_signed_p1d": -0.0076701217592450854, "alpha_spy_p1d": -0.008401177913243463, "alpha_c_p1d": -6.794274685795187e-05, "bench_spy_p1d": 0.0007310561539983773, "bench_c_p1d": -0.0076021790123871336, "ret_p1w": -0.03403635282185058, "ret_signed_p1w": -0.03403635282185058, "alpha_spy_p1w": -0.03941735694026827, "alpha_c_p1w": -0.06074000848143535, "bench_spy_p1w": 0.005381004118417687, "bench_c_p1w": 0.02670365565958477, "ret_p1m": -0.05872479162842137, "ret_signed_p1m": -0.05872479162842137, "alpha_spy_p1m": -0.058790582194520824, "alpha_c_p1m": -0.05016366569333475, "bench_spy_p1m": 6.57905660994551e-05, "bench_c_p1m": -0.008561125935086622, "ret_p3m": 0.2368168844686973, "ret_signed_p3m": 0.2368168844686973, "alpha_spy_p3m": 0.23021027301045804, "alpha_c_p3m": 0.11802407883902188, "bench_spy_p3m": 0.006606611458239264, "bench_c_p3m": 0.11879280562967542, "ret_p6m": 0.2644224972486833, "ret_signed_p6m": 0.2644224972486833, "alpha_spy_p6m": 0.1547460333524906, "alpha_c_p6m": -0.386911156461754, "bench_spy_p6m": 0.10967646389619268, "bench_c_p6m": 0.6513336537104373, "price_path": [-0.4527, -0.4288, -0.3919, -0.3721, -0.3446, -0.3326, -0.3326, -0.324, -0.3376, -0.3048, -0.2814, -0.2584, -0.2584, -0.263, -0.2809, -0.3077, -0.274, -0.2977, -0.2965, -0.2788, -0.2478, -0.1424, -0.1369, -0.1388, -0.1369, -0.145, -0.145, -0.185, -0.1675, -0.1616, -0.1913, -0.1606, -0.1659, -0.1802, -0.2107, -0.1812, -0.1278, -0.1314, 0.0604, 0.0676, 0.1091, 0.1091, 0.0441, -0.018, 0.0129, -0.0431, -0.0067, -0.0472, -0.0494, -0.0091, -0.0638, -0.0412, -0.0767, -0.082, -0.0012, -0.122, -0.122, -0.0853, -0.0671, -0.0216, -0.0137, -0.0556, -0.0503, 0.0, -0.0077, -0.0316, -0.0292, -0.029, -0.034, 0.0086, -0.0036, -0.0676, -0.0808, -0.0676, -0.0985, -0.0798, -0.0388, -0.0568, -0.0336, -0.0511, -0.0295, -0.0518, -0.0587, -0.0587, -0.0587, -0.0587, 0.0151, 0.0333, -0.0122, -0.0362, -0.029, -0.029, 0.0539, 0.1055, 0.0781, 0.093, 0.0623, 0.03, 0.0436, 0.0954, 0.128, 0.1302, 0.1963, 0.2244, 0.2876, 0.2227, 0.1651, 0.2478, 0.2212, 0.1625, 0.1759, 0.3114, 0.3267, 0.3267, 0.2852, 0.3006, 0.2991, 0.2845, 0.2924, 0.2462, 0.2212, 0.2212, 0.2761, 0.3718, 0.3483, 0.2872, 0.2368, 0.2332, 0.1745, 0.2241, 0.2325, 0.0966, 0.1541, 0.193, 0.1745, 0.1337, 0.1572, 0.1287, 0.2047, 0.1496, 0.1496, 0.0896, 0.0712, 0.123, 0.1009, 0.058, 0.0045, -0.0241, 0.0801, 0.0141, 0.0347, 0.0527, 0.0657, 0.2107, 0.1905, 0.1996, 0.1943, 0.2961, 0.3242, 0.3741, 0.3379, 0.3009, 0.3057, 0.3393, 0.3393, 0.4132, 0.5121, 0.4281, 0.4281, 0.3566, 0.3352, 0.3352, 0.3352, 0.3352, 0.4264, 0.4346, 0.3816, 0.378, 0.358, 0.3736, 0.2654, 0.2553, 0.214, 0.2301, 0.2839, 0.2887, 0.3323, 0.2517, 0.3023, 0.2644]}
{"unit_id": "orig:1996080008838996469", "ticker_idx": 3, "ts": "2025-12-03T04:52:14+00:00", "symbol": "Chroma ATE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Chroma and Advantest are expected to benefit, primarily due to 'increased test intensity'", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPMorgan: ASE, Amkor, Advantest, Chroma, and GPTC are key beneficiaries of AI advanced packaging demand surge and CoWoS/test market growth into 2026-27.", "resolved_tickers": ["2360.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Chroma ATE Inc. 2360.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Scientific & Technical Instruments'", "bench_c_industry": "Scientific & Technical Instruments", "unit_kind": "original", "lang": "en", "tweet_text": "251203_ Asian Semis: Mapping out the Advanced Back-end market for AI - JPMorgan\n\n[Key Takeaways]\n\n(1) CoWoS demand is expected to exceed supply until 2027.\n\n- Robust demand for GPUs/ASICs, 2) N3 yield bottlenecks (due to EUV usage, etc.), 3) Preparation for CoPoS.\n\n(2) The trickle-down effect to OSATs is expected to continue, with high growth in the Test market fully materializing in 2026.\n\n(3) Meanwhile, TSMC's CoPoS pilot line is scheduled for completion in early 2026. A follow-up on the related value chain in 2H26 is necessary ahead of the scheduled full-scale ramp-up in 2027.\n\n[Contents]\n\n(1) Due to the recent surge in AI chip demand and limited expansion of packaging CAPA, there are numerous inquiries regarding AI and 2.5D/3D chip advanced packaging.\n\n(2) First, ASE is expected to be a major beneficiary of expanded outsourcing from TSMC, specifically in 1) On-Substrate & Wafer Probing, and 2) Full-process 2.5D packaging.\n\n(3) Amkor is also expected to benefit from increased outsourcing by Broadcom.\n\n(4) In the test equipment sector, Chroma and Advantest are expected to benefit, primarily due to \"increased test intensity.\"\n\n(5) Among packaging equipment vendors, we view GPTC positively due to the potential for CoWoS CAPA expansion and increased orders for CoPoS starting from 2H26.\n\n(6) CoWoS demand in '26–'27 is projected to exceed supply by +20–30%.\n\n- CoWoS Capacity: 105K (End of '26) → 125K (End of '27).\n\n(7) Even with outsourcing to external vendors, the excess demand situation is expected to persist for customers like NVIDIA, Broadcom, AMD, and Amazon.\n(8) TSMC appears to be proceeding cautiously with CAPA expansion, centering on CoWoS-L, and it will likely remain difficult to satisfy all overall customer demand.\n\n(9) Furthermore, as the N3 Front-end process is likely to become a bottleneck, it is judged that aggressive CoWoS expansion in '26 will be difficult.\n\n(10) AP07 and AP08A are already filled with capacity expansions for WMCM/SoIC and CoWoS-L, and some processes have been offloaded to OSATs.\n\n(11) Additionally, due to the possibility of technological replacement by CoPoS in the future, TSMC continues to maintain a cautious stance on CoWoS CAPA expansion beyond '27.\n\n(12) Meanwhile, ASE is already receiving outsourcing orders for On-Substrate and Wafer Probing, which is projected to account for more than 80% of its LEAP (Leading Edge Advanced Packaging) revenue in '26.\n\n(13) ASE's 2.5D Full Processing packaging CAPA is expected to reach 15–20K by the end of '26.\n\n- Major customer: AMD (Venice CPU, legacy MI series, etc.).\n\n(14) There is also a probability of securing orders for NVIDIA's Vera CPU and a portion of Broadcom's Tomahawk Switch IC.\n- '27 Outlook: CoWoS revenue ~$1bn, LEAP total revenue ~$5bn.\n\n(15) ASE has executed CAPEX 3–4 times higher than Amkor over the past two years and is projected to capture approximately 70–80% of TSMC's outsourcing demand.\n\n(16) In Amkor's case, while it struggled due to the blocking of NVIDIA H20, it is entering a recovery phase with increased CoWoS-S outsourcing from Broadcom in '26.\n\n(17) Its 2.5D packaging capacity is expected to grow from the existing 6–8K to 12–15K by the end of '26.\n\n- Major customers: Broadcom (Tomahawk 6, potential for TPUs), NVIDIA (GB10, ARM CPU, potential for Vera CPU).\n\n(18) Limited CAPA within Taiwan is a bottleneck for securing TSMC outsourcing, and expansion at the Korea K5 plant is slow. Meaningful growth is difficult until the Arizona Peoria facility comes online in '28.\n\n(19) TSMC's CoPoS (310x310mm) pilot line is expected to be completed in early '26.\n\n(20) If yield issues are resolved, there is a possibility that CoPoS will be adopted in NVIDIA's Feynman Project in '28.\n\n(21) Orders for CoPoS CAPA are expected to start appearing as early as 2H26, with ramp-up anticipated in '27.\n\n(22) Lastly, the Test market is projected to increase very strongly in '26.\n\n(23) Factors such as 1) N3 process transition, 2) expanded adoption of Chiplets, and 3) introduction of HBM4 will drive longer test times, strongly boosting demand for new testers.\n\n(24) For NVIDIA GPUs, KYEC is expected to maintain a high market share for Blackwell and Rubin, while ASE may capture a small share of Rubin by the end of '26.\n\n(25) According to our channel checks, certification for Advantest's Burn-in Oven is currently not proceeding smoothly, which will be a limiting factor for ASE's Final Test share expansion.\n\n(26) In ASICs, KYEC is expected to continue its good positioning in Broadcom testing (Alphabet TPU + Meta MTIA), while ASE/Amkor are projected to benefit from the ramp-up of Amazon Trainium 3.\n\n(27) For AMD, Chroma's SLT (System Level Test) is expected to be qualified for the MI450, and it is anticipated to sustain a 100% share in NVIDIA Rubin SLT thereafter.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03354037267080745, "ret_signed_m1d": -0.03354037267080745, "alpha_spy_m1d": -0.030089554260399054, "alpha_c_m1d": -0.03116094264708169, "bench_spy_m1d": -0.003450818410408396, "bench_c_m1d": -0.0023794300237257593, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03602484472049694, "ret_signed_p1d": -0.03602484472049694, "alpha_spy_p1d": -0.03675590087449532, "alpha_c_p1d": -0.03974911491729172, "bench_spy_p1d": 0.0007310561539983773, "bench_c_p1d": 0.0037242701967947767, "ret_p1w": -0.006211180124223614, "ret_signed_p1w": -0.006211180124223614, "alpha_spy_p1w": -0.011592184242641301, "alpha_c_p1w": -0.031970610124733345, "bench_spy_p1w": 0.005381004118417687, "bench_c_p1w": 0.02575943000050973, "ret_p1m": -0.037267080745341574, "ret_signed_p1m": -0.037267080745341574, "alpha_spy_p1m": -0.03733287131144103, "alpha_c_p1m": -0.0317038914606701, "bench_spy_p1m": 6.57905660994551e-05, "bench_c_p1m": -0.005563189284671477, "ret_p3m": 0.8819875776397517, "ret_signed_p3m": 0.8819875776397517, "alpha_spy_p3m": 0.8753809661815124, "alpha_c_p3m": 0.9181498908999243, "bench_spy_p3m": 0.006606611458239264, "bench_c_p3m": -0.03616231326017261, "ret_p6m": 2.1925465838509317, "ret_signed_p6m": 2.1925465838509317, "alpha_spy_p6m": 2.0828701199547393, "alpha_c_p6m": 1.9003221217527047, "bench_spy_p6m": 0.10967646389619268, "bench_c_p6m": 0.29222446209822706, "price_path": [-0.2894, -0.2919, -0.2845, -0.2795, -0.287, -0.2907, -0.3143, -0.3093, -0.2969, -0.2373, -0.2571, -0.2609, -0.2373, -0.2509, -0.2596, -0.2994, -0.2994, -0.2807, -0.2472, -0.2124, -0.2062, -0.2062, -0.195, -0.2149, -0.2273, -0.2273, -0.2, -0.1764, -0.1466, -0.1379, -0.1776, -0.1851, -0.159, -0.1391, -0.1404, -0.1404, -0.0832, -0.0783, -0.0211, -0.0012, 0.0174, 0.0075, -0.036, -0.0534, -0.0646, -0.0795, -0.0758, -0.0559, -0.0087, 0.0025, -0.0248, -0.0385, -0.0733, -0.0435, 0.0037, -0.0385, 0.0062, 0.0298, -0.0062, 0.0075, 0.0174, -0.0248, -0.0335, 0.0, -0.036, -0.0112, 0.0, 0.0037, -0.0062, -0.0112, 0.0, -0.0447, -0.0733, -0.0807, -0.0957, -0.0795, -0.0609, -0.0522, -0.0422, -0.0422, -0.0199, -0.0298, -0.0186, -0.0373, -0.0373, -0.0112, 0.0075, 0.0634, 0.1453, 0.1491, 0.1466, 0.1851, 0.2224, 0.2286, 0.2, 0.2286, 0.1801, 0.2795, 0.236, 0.2547, 0.2733, 0.2609, 0.2857, 0.2857, 0.2609, 0.2174, 0.2062, 0.2919, 0.2795, 0.2099, 0.2012, 0.2795, 0.2919, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.3043, 0.4348, 0.5217, 0.559, 0.7143, 0.7143, 0.882, 0.7329, 0.6646, 0.7453, 0.7081, 0.5404, 0.6894, 0.8571, 0.8571, 0.9006, 0.8012, 0.8261, 0.9441, 0.9814, 0.9689, 0.8882, 0.8385, 1.0186, 1.0373, 1.0062, 1.0186, 0.8199, 1.0, 0.9068, 0.9068, 0.9068, 0.9503, 1.1429, 1.2857, 1.2174, 1.3354, 1.2857, 1.5093, 1.5466, 1.8012, 1.5466, 1.5155, 1.5217, 1.4099, 1.3851, 1.3913, 1.5652, 1.5528, 1.6335, 1.6335, 1.8696, 1.7764, 1.913, 1.8882, 1.7702, 1.9317, 2.0311, 1.9876, 1.8075, 1.7826, 1.7143, 1.5528, 1.5652, 1.8199, 1.8199, 2.0559, 2.1056, 2.2609, 2.1925]}
{"unit_id": "orig:1988853087067410788", "ticker_idx": 0, "ts": "2025-11-13T06:15:01+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raised its price target for Micron to $325", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Morgan Stanley raises MU PT to $325; author shares/adopts the bullish stance.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley raised its price target for Micron to $325.\n\n$MU https://t.co/Z7Zhu6NPxr", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.033551375402601336, "ret_signed_m1d": 0.033551375402601336, "alpha_spy_m1d": 0.016677345924404507, "alpha_c_m1d": 0.0024917521582035196, "bench_spy_m1d": 0.016874029478196828, "bench_c_m1d": 0.031059623244397816, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04169659820084681, "ret_signed_p1d": 0.04169659820084681, "alpha_spy_p1d": 0.041860184300430814, "alpha_c_p1d": 0.041232553992091026, "bench_spy_p1d": -0.00016358609958400105, "bench_c_p1d": 0.0004640442087557872, "ret_p1w": -0.1501582305607011, "ret_signed_p1w": -0.1501582305607011, "alpha_spy_p1w": -0.12112731516266395, "alpha_c_p1w": -0.09296893221738423, "bench_spy_p1w": -0.02903091539803715, "bench_c_p1w": -0.057189298343316874, "ret_p1m": 0.01768307561385485, "ret_signed_p1m": 0.01768307561385485, "alpha_spy_p1m": 0.003219634823665496, "alpha_c_p1m": -0.00928750760546504, "bench_spy_p1m": 0.014463440790189352, "bench_c_p1m": 0.026970583219319888, "ret_p3m": 0.5758632108371466, "ret_signed_p3m": 0.5758632108371466, "alpha_spy_p3m": 0.5429408460964087, "alpha_c_p3m": 0.3983809050194953, "bench_spy_p3m": 0.03292236474073795, "bench_c_p3m": 0.17748230581765134, "ret_p6m": 2.154359838117717, "ret_signed_p6m": 2.154359838117717, "alpha_spy_p6m": 2.0505266897283185, "alpha_c_p6m": 1.5062454405224726, "bench_spy_p6m": 0.10383314838939839, "bench_c_p6m": 0.6481143975952444, "price_path": [-0.4789, -0.4852, -0.5056, -0.5116, -0.5037, -0.509, -0.5086, -0.5034, -0.4854, -0.4981, -0.4981, -0.5003, -0.4993, -0.4761, -0.4459, -0.4455, -0.4296, -0.4095, -0.3649, -0.3369, -0.3346, -0.3302, -0.3252, -0.2877, -0.3137, -0.3057, -0.2981, -0.318, -0.3385, -0.3367, -0.3087, -0.2943, -0.2318, -0.225, -0.2073, -0.1941, -0.2163, -0.1705, -0.1883, -0.2336, -0.1865, -0.2106, -0.19, -0.1453, -0.1459, -0.1274, -0.1463, -0.1624, -0.1276, -0.0757, -0.0711, -0.0635, -0.0436, -0.0546, -0.0556, -0.0095, -0.0798, 0.0023, 0.0058, 0.0041, 0.069, 0.0176, 0.0336, 0.0, 0.0417, 0.0211, -0.0357, -0.0465, -0.1502, -0.1248, -0.0549, -0.0524, -0.0282, -0.0282, -0.002, 0.0148, 0.0107, -0.0118, -0.0435, 0.0011, 0.0421, 0.0653, 0.1129, 0.0908, 0.0177, 0.0023, -0.0187, -0.0482, 0.049, 0.1223, 0.1673, 0.1659, 0.2099, 0.2099, 0.2019, 0.2428, 0.2355, 0.205, 0.205, 0.3317, 0.3179, 0.45, 0.4336, 0.3807, 0.457, 0.4603, 0.4276, 0.4074, 0.4213, 0.5315, 0.5315, 0.541, 0.6428, 0.6786, 0.6873, 0.6427, 0.732, 0.8378, 0.8399, 0.7516, 0.8484, 0.7709, 0.6018, 0.6166, 0.6664, 0.6191, 0.5759, 0.7325, 0.7478, 0.738, 0.738, 0.6879, 0.7773, 0.7621, 0.8077, 0.7773, 0.7648, 0.8112, 0.7545, 0.741, 0.7423, 0.603, 0.6921, 0.6763, 0.5634, 0.6437, 0.7019, 0.7677, 0.7114, 0.7991, 0.8653, 0.9493, 0.9494, 0.8757, 0.7855, 0.7072, 0.6699, 0.6132, 0.5008, 0.5082, 0.3592, 0.427, 0.5537, 0.5469, 0.5469, 0.5956, 0.5948, 0.7179, 0.7804, 0.7765, 0.8017, 0.9668, 0.927, 0.9312, 0.9221, 0.894, 0.8981, 1.059, 1.0347, 1.098, 1.2156, 1.13, 1.1899, 1.1844, 1.2902, 1.4348, 1.7041, 1.8155, 1.7312, 2.1544]}
{"unit_id": "quote:1989608182759518283", "ticker_idx": 0, "ts": "2025-11-15T08:15:30+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I had been holding Broadcom for a long time but recently decided to sell it… it is impossible for Broadcom to permanently and consistently earn profit margins from its ASIC projects that are on par with what Nvidia makes… the profits Broadcom earns from ASIC projects can only trend downward over time", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Sold AVGO; bearish on structural margin erosion as ASIC know-how transfers to customers; implicitly favors NVDA's durable margin model.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "This is also why I had been holding Broadcom for a long time but recently decided to sell it.\n\n“AVGO only handles the back-end design in Google’s TPU project… Google carries out the remaining 100% entirely in-house.”\n\nExactly. What I keep emphasizing is that it is impossible for Broadcom to permanently and consistently earn profit margins from its ASIC projects that are on par with what Nvidia makes.\n\nBy its very nature, what Broadcom does inevitably transfers its know-how to its customers, and that means the profits Broadcom earns from ASIC projects can only trend downward over time.\n\n$AVGO $NVDA", "reply_text": null, "parent_text": null, "quoted_text": "A very informative interview with an $AMD employee on the ASIC vs $NVDA battle and $GOOGL's TPUs:\n\n1. In the case of the $GOOGL / $AVGO TPU relationship, he thinks that $AVGO doesn't know everything about the chip anymore. In the first two generations, they collaborated on the https://t.co/9nLOpfJttX", "parent_author": null, "ret_m1d": -0.0005545891131348846, "ret_signed_m1d": 0.0005545891131348846, "alpha_spy_m1d": 0.00995872691901556, "alpha_c_m1d": 0.01430729233613226, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.013752703222997376, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0062746723748271505, "ret_signed_p1d": 0.0062746723748271505, "alpha_spy_p1d": -0.0021227881833794093, "alpha_c_p1d": -0.01432469648332424, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.02059936885815139, "ret_p1w": 0.10304975724878007, "ret_signed_p1w": -0.10304975724878007, "alpha_spy_p1w": -0.09845290211698532, "alpha_c_p1w": -0.10651725920085786, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.0034675019520777894, "ret_p1m": -0.003939889765353866, "ret_signed_p1m": 0.003939889765353866, "alpha_spy_p1m": 0.023769596527842274, "alpha_c_p1m": 0.038145061242348066, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.0342051714769942, "ret_p3m": -0.0316542332711619, "ret_signed_p3m": 0.0316542332711619, "alpha_spy_p3m": 0.05811351415855015, "alpha_c_p3m": 0.22875580248986294, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.19710156921870103, "ret_p6m": 0.22861089645100519, "ret_signed_p6m": -0.22861089645100519, "alpha_spy_p6m": -0.11336880518852466, "alpha_c_p6m": 0.42580115929386086, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.654412055744866, "price_path": [-0.1517, -0.1563, -0.1434, -0.1428, -0.1318, -0.1252, -0.1008, -0.1336, -0.1336, -0.1311, -0.119, -0.1082, -0.0243, 0.007, -0.0191, 0.0767, 0.0478, 0.0485, 0.0608, 0.0488, 0.0085, 0.0062, 0.005, -0.0113, -0.0108, -0.0097, -0.0191, -0.0237, -0.043, -0.0372, -0.027, -0.013, -0.0125, -0.0209, -0.0182, 0.0083, 0.0069, -0.0526, 0.041, 0.0043, 0.0253, 0.0336, 0.0195, 0.0192, 0.0, -0.0069, 0.0048, 0.0335, 0.0566, 0.0885, 0.1265, 0.0987, 0.0787, 0.0581, 0.0271, 0.0477, 0.0378, 0.0198, 0.0459, 0.0272, 0.0367, -0.0078, -0.0006, 0.0, -0.0063, 0.0343, 0.0122, -0.0072, 0.103, 0.1237, 0.1603, 0.1603, 0.176, 0.1267, 0.1136, 0.1108, 0.112, 0.1389, 0.1706, 0.1857, 0.2052, 0.186, 0.0504, -0.0083, -0.0039, -0.0485, -0.0373, -0.0067, -0.0016, 0.0214, 0.024, 0.024, 0.0296, 0.0216, 0.023, 0.012, 0.012, 0.0164, 0.0042, 0.0052, 0.0044, -0.0278, 0.0087, 0.0299, 0.0369, -0.0062, 0.003, 0.0284, 0.0284, -0.0275, -0.0386, -0.0483, -0.0642, -0.0501, -0.0269, -0.0256, -0.0329, -0.0313, -0.0318, -0.0634, -0.0993, -0.0921, -0.0265, 0.0057, -0.0045, 0.0022, -0.0317, -0.0492, -0.0492, -0.0276, -0.0248, -0.0234, -0.0273, -0.0341, -0.0483, -0.0283, -0.0593, -0.0656, -0.0678, -0.0823, -0.0715, -0.027, -0.0337, 0.011, 0.0017, -0.0012, -0.0176, -0.058, -0.0499, -0.0605, -0.0762, -0.0648, -0.0921, -0.055, -0.0674, -0.0658, -0.0934, -0.119, -0.1403, -0.0931, -0.0814, -0.0783, -0.0783, -0.0787, -0.0214, 0.0274, 0.0399, 0.0887, 0.1127, 0.1157, 0.1624, 0.1676, 0.1912, 0.171, 0.1784, 0.2384, 0.2305, 0.2387, 0.2254, 0.1716, 0.188, 0.2231, 0.2344, 0.2204, 0.2522, 0.2466, 0.2089, 0.26, 0.2554, 0.2286]}
{"unit_id": "quote:1989608182759518283", "ticker_idx": 1, "ts": "2025-11-15T08:15:30+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is impossible for Broadcom to permanently and consistently earn profit margins from its ASIC projects that are on par with what Nvidia makes", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Sold AVGO; bearish on structural margin erosion as ASIC know-how transfers to customers; implicitly favors NVDA's durable margin model.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "This is also why I had been holding Broadcom for a long time but recently decided to sell it.\n\n“AVGO only handles the back-end design in Google’s TPU project… Google carries out the remaining 100% entirely in-house.”\n\nExactly. What I keep emphasizing is that it is impossible for Broadcom to permanently and consistently earn profit margins from its ASIC projects that are on par with what Nvidia makes.\n\nBy its very nature, what Broadcom does inevitably transfers its know-how to its customers, and that means the profits Broadcom earns from ASIC projects can only trend downward over time.\n\n$AVGO $NVDA", "reply_text": null, "parent_text": null, "quoted_text": "A very informative interview with an $AMD employee on the ASIC vs $NVDA battle and $GOOGL's TPUs:\n\n1. In the case of the $GOOGL / $AVGO TPU relationship, he thinks that $AVGO doesn't know everything about the chip anymore. In the first two generations, they collaborated on the https://t.co/9nLOpfJttX", "parent_author": null, "ret_m1d": 0.019131846405796793, "ret_signed_m1d": 0.019131846405796793, "alpha_spy_m1d": 0.009727708599916118, "alpha_c_m1d": 0.005379143182799417, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.013752703222997376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028081462747415187, "ret_signed_p1d": -0.028081462747415187, "alpha_spy_p1d": -0.019684002189208627, "alpha_c_p1d": -0.007482093889263797, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.02059936885815139, "ret_p1w": -0.021704126091018905, "ret_signed_p1w": -0.021704126091018905, "alpha_spy_p1w": -0.026300981222813657, "alpha_c_p1w": -0.018236624138941115, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.0034675019520777894, "ret_p1m": -0.04753536547217685, "ret_signed_p1m": -0.04753536547217685, "alpha_spy_p1m": -0.06736507223466526, "alpha_c_p1m": -0.08174053694917105, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.0342051714769942, "ret_p3m": 0.0018778713042340023, "ret_signed_p3m": 0.0018778713042340023, "alpha_spy_p3m": -0.024581409583154246, "alpha_c_p3m": -0.19522369791446703, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.19710156921870103, "ret_p6m": 0.18330246755822355, "ret_signed_p6m": 0.18330246755822355, "alpha_spy_p6m": 0.06806037629574302, "alpha_c_p6m": -0.4711095881866425, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.654412055744866, "price_path": [-0.0601, -0.0623, -0.0462, -0.0364, -0.0259, -0.0269, -0.0345, -0.0666, -0.0666, -0.0848, -0.0857, -0.0801, -0.105, -0.0981, -0.0849, -0.0497, -0.0505, -0.0471, -0.0474, -0.0628, -0.0874, -0.0555, -0.0532, -0.016, -0.0438, -0.0516, -0.0477, -0.0451, -0.0255, -0.0001, 0.0034, 0.0123, 0.0055, -0.0057, -0.0084, 0.0135, 0.032, -0.0184, 0.0092, -0.0352, -0.0363, -0.0257, -0.0181, -0.0212, -0.0292, -0.0339, -0.0238, -0.0018, 0.0262, 0.0773, 0.1095, 0.0873, 0.0852, 0.1087, 0.0648, 0.0461, 0.0079, 0.0083, 0.0667, 0.0352, 0.0386, 0.0014, 0.0191, 0.0, -0.0281, -0.0004, -0.0319, -0.0414, -0.0217, -0.0471, -0.034, -0.034, -0.0514, -0.0358, -0.0275, -0.0376, -0.0172, -0.0224, -0.0056, -0.0087, -0.0151, -0.0303, -0.062, -0.0552, -0.0475, -0.0839, -0.0667, -0.03, -0.0155, 0.014, 0.0108, 0.0108, 0.0211, 0.0087, 0.0051, -0.0005, -0.0005, 0.0121, 0.0082, 0.0035, 0.0135, -0.0083, -0.0093, -0.0088, -0.0042, -0.0185, 0.0025, -0.0019, -0.0019, -0.0457, -0.0175, -0.0094, 0.0058, -0.0006, 0.0103, 0.0264, 0.0317, 0.0243, -0.0053, -0.0335, -0.0665, -0.0788, -0.0063, 0.0185, 0.0105, 0.0185, 0.0019, -0.0203, -0.0203, -0.0087, 0.0075, 0.007, 0.0173, 0.0266, 0.0336, 0.0481, -0.0091, -0.0504, -0.022, -0.035, -0.019, -0.0174, -0.047, -0.0211, -0.0098, -0.0029, -0.0184, -0.0339, -0.018, -0.0249, -0.0331, -0.043, -0.0744, -0.0586, -0.061, -0.0423, -0.0822, -0.1022, -0.1147, -0.0653, -0.058, -0.0493, -0.0493, -0.0479, -0.0454, -0.0241, -0.0143, 0.011, 0.0146, 0.0532, 0.0659, 0.0631, 0.0809, 0.083, 0.0713, 0.0853, 0.07, 0.1163, 0.161, 0.1425, 0.1215, 0.0696, 0.0636, 0.0638, 0.0532, 0.1139, 0.1336, 0.1534, 0.1761, 0.1833]}
{"unit_id": "orig:1980129896954519889", "ticker_idx": 0, "ts": "2025-10-20T04:32:11+00:00", "symbol": "Asia Tech", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "JPMorgan expects strong AI demand from the top four CSPs and incremental demand from players such as OpenAI and Oracle to lift Asia Tech EPS by an additional 20–25% over the next two to three quarters", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM relays bullish AI demand thesis: Asia Tech EPS +20–25% over 2–3 quarters, no bubble signs, TSMC CoWoS +60%, CSP capex upgrades into 2026–27.", "resolved_tickers": ["AAXJ", "SMH", "EWY", "EWT"], "resolution_method": "basket", "resolution_reason": "Asia tech sector; AAXJ broad + semiconductor ETFs", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Surge in AI Demand and Price Increases Likely to Drive EPS Upgrades Through 1H26, No Signs of AI Bubble Burst\n\n⸻\n\n1. Continued EPS Upgrades Driven by AI Demand\n    \n• JPMorgan expects strong AI demand from the top four CSPs (cloud service providers) and incremental demand from players such as OpenAI and Oracle to lift Asia Tech EPS by an additional 20–25% over the next two to three quarters.\n\nThis improvement is likely to materialize mainly in 4Q25 and 1Q26 as the market re-rates growth expectations for the AI value chain in 2026.\n    \n• The firm currently forecasts capex for the top four CSPs to grow by 20% in 2026, with potential for further upside.\nThe hardware component of spending is expected to increase its share — for instance, TSMC’s CoWoS wafer shipments are projected to rise by around 60%.\n\n⸻\n\n2. No Signs of a Semiconductor Bubble — Capacity Expansion Still Absent\n    \n• Typical semiconductor/tech bubbles occur when a new demand frontier emerges rapidly (as with generative AI since 2023).\nThe supply chain races to expand production capacity to meet this demand, leading to a surge in wafer fab equipment (WFE) spending.\nHowever, because it takes 1.5–2 years for new capacity to come online, by the time it does, demand has already shifted, resulting in oversupply and a cyclical downturn.\n    \n• Applying this logic to the AI-driven tech cycle, JPMorgan notes that WFE spending has only rebounded 8% from the 2023 trough, and excluding China, the rebound is just 12% — far below the ~80% rebound typically seen in previous up-cycles.\nThus, there are no imminent signs of overheating or a bubble.\n    \n• To illustrate, CSP capex totals about USD 379 billion, while NVIDIA + AMD + ASIC revenue amounts to USD 287 billion, and downstream AI-server revenue is only USD 96 billion.\nThis disparity indicates persistent bottlenecks across the supply chain, from chips to racks and power infrastructure.\n\n⸻\n\n3. Potential for Further CSP Capex Upgrades into 2026 — Early Signs of Debt Utilization\n    \n• In addition to the top four CSPs’ strong demand, JPMorgan expects Oracle, OpenAI, and other AI labs to further lift CSP capex in 2026.\nSo far, most of the top four CSPs’ capex has been financed through operating cash flows, with limited leverage, implying no immediate funding risk.\n    \n• Given the competitive nature of generative-AI investments, the top four CSPs are likely to continue aggressive capex spending into 2026 and 2027.\nHowever, for smaller players such as Oracle and CoreWeave, debt financing is becoming an increasingly important factor in funding expansion.\n    \n• Still, these smaller CSPs account for only ~11% of the top six CSPs’ total capex, so any structural shift in the composition of capital spending should be monitored beginning in 2026, when debt discussions begin in earnest.\nFor now, given ongoing constraints in chip supply, rack capacity, and data-center power availability, JPMorgan sees limited near-term downside risk to the AI capex cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01594587868363348, "ret_signed_m1d": -0.01594587868363348, "alpha_spy_m1d": -0.005652459734569487, "alpha_c_m1d": -0.005652459734569487, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.010293418949063993, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01011388687401274, "ret_signed_p1d": -0.01011388687401274, "alpha_spy_p1d": -0.010098890954000211, "alpha_c_p1d": -0.010098890954000211, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -1.4995920012528252e-05, "ret_p1w": 0.027059776814196412, "ret_signed_p1w": 0.027059776814196412, "alpha_spy_p1w": 0.006294085138358552, "alpha_c_p1w": 0.006294085138358552, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.02076569167583786, "ret_p1m": -0.025275639945918027, "ret_signed_p1m": -0.025275639945918027, "alpha_spy_p1m": -0.0085618641451633, "alpha_c_p1m": -0.0085618641451633, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.016713775800754727, "ret_p3m": 0.1327700967211805, "ret_signed_p3m": 0.1327700967211805, "alpha_spy_p3m": 0.09852983940582322, "alpha_c_p3m": 0.09852983940582322, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.03424025731535729, "ret_p6m": 0.3417554677413321, "ret_signed_p6m": 0.3417554677413321, "alpha_spy_p6m": 0.3013647801345091, "alpha_c_p6m": 0.3013647801345091, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.04039068760682296, "price_path": [-0.1351, -0.1382, -0.1392, -0.1403, -0.1388, -0.1387, -0.1481, -0.1596, -0.1463, -0.146, -0.1441, -0.1336, -0.1322, -0.1342, -0.1202, -0.1147, -0.1218, -0.1278, -0.1247, -0.1381, -0.1452, -0.1477, -0.1285, -0.1329, -0.1313, -0.1313, -0.1258, -0.1411, -0.1411, -0.1477, -0.1414, -0.1374, -0.127, -0.1179, -0.1109, -0.0988, -0.0891, -0.0866, -0.0807, -0.0736, -0.0768, -0.0647, -0.0701, -0.0578, -0.0582, -0.0661, -0.0693, -0.0741, -0.0661, -0.0611, -0.0463, -0.0372, -0.0345, -0.0243, -0.0353, -0.0218, -0.03, -0.0735, -0.0369, -0.0496, -0.0279, -0.0172, -0.0159, 0.0, -0.0101, -0.0164, -0.0055, 0.0092, 0.0271, 0.0299, 0.0423, 0.0286, 0.0329, 0.0488, 0.0149, 0.0264, 0.0055, -0.0037, 0.0234, 0.0154, 0.0205, -0.0006, 0.0041, -0.0127, -0.0253, -0.0253, -0.0477, -0.0447, -0.0269, -0.026, -0.0108, -0.0108, -0.0054, -0.0075, 0.0031, 0.01, 0.0031, 0.0176, 0.0231, 0.0245, 0.0381, 0.0258, -0.0028, -0.0041, -0.0102, -0.028, -0.012, 0.0037, 0.0109, 0.0193, 0.0261, 0.0261, 0.0348, 0.0426, 0.044, 0.0362, 0.0362, 0.0726, 0.0884, 0.1073, 0.101, 0.0948, 0.1117, 0.1193, 0.1119, 0.1179, 0.1328, 0.1379, 0.1379, 0.1207, 0.1496, 0.156, 0.167, 0.1651, 0.197, 0.2131, 0.2042, 0.1759, 0.1792, 0.1815, 0.1526, 0.1495, 0.1928, 0.2048, 0.2023, 0.2395, 0.2275, 0.2404, 0.2404, 0.2297, 0.2413, 0.2445, 0.2796, 0.2627, 0.2967, 0.3226, 0.3103, 0.3069, 0.2891, 0.2064, 0.2215, 0.1839, 0.1712, 0.2112, 0.2026, 0.2196, 0.1673, 0.1693, 0.215, 0.2261, 0.205, 0.2179, 0.1662, 0.2102, 0.1882, 0.1978, 0.1452, 0.1395, 0.1134, 0.1637, 0.1845, 0.1686, 0.1686, 0.1845, 0.1942, 0.2794, 0.2817, 0.2875, 0.3066, 0.3418]}
{"unit_id": "orig:1980129896954519889", "ticker_idx": 1, "ts": "2025-10-20T04:32:11+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's CoWoS wafer shipments are projected to rise by around 60%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM relays bullish AI demand thesis: Asia Tech EPS +20–25% over 2–3 quarters, no bubble signs, TSMC CoWoS +60%, CSP capex upgrades into 2026–27.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Surge in AI Demand and Price Increases Likely to Drive EPS Upgrades Through 1H26, No Signs of AI Bubble Burst\n\n⸻\n\n1. Continued EPS Upgrades Driven by AI Demand\n    \n• JPMorgan expects strong AI demand from the top four CSPs (cloud service providers) and incremental demand from players such as OpenAI and Oracle to lift Asia Tech EPS by an additional 20–25% over the next two to three quarters.\n\nThis improvement is likely to materialize mainly in 4Q25 and 1Q26 as the market re-rates growth expectations for the AI value chain in 2026.\n    \n• The firm currently forecasts capex for the top four CSPs to grow by 20% in 2026, with potential for further upside.\nThe hardware component of spending is expected to increase its share — for instance, TSMC’s CoWoS wafer shipments are projected to rise by around 60%.\n\n⸻\n\n2. No Signs of a Semiconductor Bubble — Capacity Expansion Still Absent\n    \n• Typical semiconductor/tech bubbles occur when a new demand frontier emerges rapidly (as with generative AI since 2023).\nThe supply chain races to expand production capacity to meet this demand, leading to a surge in wafer fab equipment (WFE) spending.\nHowever, because it takes 1.5–2 years for new capacity to come online, by the time it does, demand has already shifted, resulting in oversupply and a cyclical downturn.\n    \n• Applying this logic to the AI-driven tech cycle, JPMorgan notes that WFE spending has only rebounded 8% from the 2023 trough, and excluding China, the rebound is just 12% — far below the ~80% rebound typically seen in previous up-cycles.\nThus, there are no imminent signs of overheating or a bubble.\n    \n• To illustrate, CSP capex totals about USD 379 billion, while NVIDIA + AMD + ASIC revenue amounts to USD 287 billion, and downstream AI-server revenue is only USD 96 billion.\nThis disparity indicates persistent bottlenecks across the supply chain, from chips to racks and power infrastructure.\n\n⸻\n\n3. Potential for Further CSP Capex Upgrades into 2026 — Early Signs of Debt Utilization\n    \n• In addition to the top four CSPs’ strong demand, JPMorgan expects Oracle, OpenAI, and other AI labs to further lift CSP capex in 2026.\nSo far, most of the top four CSPs’ capex has been financed through operating cash flows, with limited leverage, implying no immediate funding risk.\n    \n• Given the competitive nature of generative-AI investments, the top four CSPs are likely to continue aggressive capex spending into 2026 and 2027.\nHowever, for smaller players such as Oracle and CoreWeave, debt financing is becoming an increasingly important factor in funding expansion.\n    \n• Still, these smaller CSPs account for only ~11% of the top six CSPs’ total capex, so any structural shift in the composition of capital spending should be monitored beginning in 2026, when debt discussions begin in earnest.\nFor now, given ongoing constraints in chip supply, rack capacity, and data-center power availability, JPMorgan sees limited near-term downside risk to the AI capex cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020270355539053875, "ret_signed_m1d": -0.020270355539053875, "alpha_spy_m1d": -0.009976936589989882, "alpha_c_m1d": -0.007627013011908823, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 1.4995920012528252e-05, "alpha_c_p1d": 0.005529688041661651, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.02076569167583786, "alpha_c_p1w": -0.03663383061733261, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": -0.050675722797898226, "ret_signed_p1m": -0.050675722797898226, "alpha_spy_p1m": -0.0339619469971435, "alpha_c_p1m": -0.010557095472422784, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.14569809966039515, "ret_signed_p3m": 0.14569809966039515, "alpha_spy_p3m": 0.11145784234503786, "alpha_c_p3m": 0.00047757690033312095, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 0.39768712642960513, "ret_signed_p6m": 0.39768712642960513, "alpha_spy_p6m": 0.35729643882278217, "alpha_c_p6m": 0.09186822300765218, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.2294, -0.2294, -0.2294, -0.2294, -0.2362, -0.2227, -0.2193, -0.2193, -0.2362, -0.2261, -0.2429, -0.2059, -0.2092, -0.2059, -0.2059, -0.1924, -0.2092, -0.2059, -0.2059, -0.2025, -0.2362, -0.2261, -0.2362, -0.2126, -0.2092, -0.1991, -0.2193, -0.2193, -0.216, -0.2193, -0.2193, -0.2193, -0.2059, -0.2059, -0.1924, -0.1756, -0.1655, -0.152, -0.1554, -0.1351, -0.1453, -0.1318, -0.1453, -0.125, -0.0946, -0.0946, -0.1081, -0.1216, -0.1216, -0.1182, -0.1047, -0.0777, -0.0541, -0.0541, -0.0304, -0.0439, -0.027, -0.027, -0.0439, -0.0372, -0.0101, 0.0034, -0.0203, 0.0, 0.0, -0.0135, -0.0203, -0.0203, 0.0, -0.0034, 0.0169, 0.0169, 0.0135, 0.0203, 0.0169, -0.0135, -0.0101, -0.0135, -0.0034, -0.0101, -0.0034, -0.0135, -0.0338, -0.0236, -0.0507, -0.0574, -0.0169, -0.0642, -0.0709, -0.0439, -0.027, -0.0304, -0.027, -0.0473, -0.0338, -0.0203, -0.0236, -0.0135, 0.0101, 0.0, 0.0169, -0.0034, 0.0033, -0.017, -0.0272, -0.0306, -0.0306, -0.0306, -0.0068, 0.0101, 0.0135, 0.0135, 0.0237, 0.0372, 0.0305, 0.0508, 0.0508, 0.0745, 0.1321, 0.1559, 0.1355, 0.1423, 0.1389, 0.1457, 0.1593, 0.1593, 0.1457, 0.1796, 0.1932, 0.2033, 0.1796, 0.1932, 0.1999, 0.1898, 0.2067, 0.2338, 0.2237, 0.2033, 0.1965, 0.2203, 0.2101, 0.1965, 0.2067, 0.2304, 0.2745, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2881, 0.3321, 0.366, 0.3525, 0.3525, 0.3389, 0.3118, 0.2643, 0.2881, 0.2813, 0.227, 0.2542, 0.3152, 0.2779, 0.2643, 0.2508, 0.2719, 0.2957, 0.2583, 0.2515, 0.2311, 0.2311, 0.2549, 0.2515, 0.2379, 0.2106, 0.197, 0.2617, 0.2311, 0.2311, 0.2311, 0.2651, 0.3263, 0.3297, 0.3603, 0.3535, 0.3977]}
{"unit_id": "orig:1980129896954519889", "ticker_idx": 2, "ts": "2025-10-20T04:32:11+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CSP capex totals about USD 379 billion, while NVIDIA + AMD + ASIC revenue amounts to USD 287 billion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM relays bullish AI demand thesis: Asia Tech EPS +20–25% over 2–3 quarters, no bubble signs, TSMC CoWoS +60%, CSP capex upgrades into 2026–27.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Surge in AI Demand and Price Increases Likely to Drive EPS Upgrades Through 1H26, No Signs of AI Bubble Burst\n\n⸻\n\n1. Continued EPS Upgrades Driven by AI Demand\n    \n• JPMorgan expects strong AI demand from the top four CSPs (cloud service providers) and incremental demand from players such as OpenAI and Oracle to lift Asia Tech EPS by an additional 20–25% over the next two to three quarters.\n\nThis improvement is likely to materialize mainly in 4Q25 and 1Q26 as the market re-rates growth expectations for the AI value chain in 2026.\n    \n• The firm currently forecasts capex for the top four CSPs to grow by 20% in 2026, with potential for further upside.\nThe hardware component of spending is expected to increase its share — for instance, TSMC’s CoWoS wafer shipments are projected to rise by around 60%.\n\n⸻\n\n2. No Signs of a Semiconductor Bubble — Capacity Expansion Still Absent\n    \n• Typical semiconductor/tech bubbles occur when a new demand frontier emerges rapidly (as with generative AI since 2023).\nThe supply chain races to expand production capacity to meet this demand, leading to a surge in wafer fab equipment (WFE) spending.\nHowever, because it takes 1.5–2 years for new capacity to come online, by the time it does, demand has already shifted, resulting in oversupply and a cyclical downturn.\n    \n• Applying this logic to the AI-driven tech cycle, JPMorgan notes that WFE spending has only rebounded 8% from the 2023 trough, and excluding China, the rebound is just 12% — far below the ~80% rebound typically seen in previous up-cycles.\nThus, there are no imminent signs of overheating or a bubble.\n    \n• To illustrate, CSP capex totals about USD 379 billion, while NVIDIA + AMD + ASIC revenue amounts to USD 287 billion, and downstream AI-server revenue is only USD 96 billion.\nThis disparity indicates persistent bottlenecks across the supply chain, from chips to racks and power infrastructure.\n\n⸻\n\n3. Potential for Further CSP Capex Upgrades into 2026 — Early Signs of Debt Utilization\n    \n• In addition to the top four CSPs’ strong demand, JPMorgan expects Oracle, OpenAI, and other AI labs to further lift CSP capex in 2026.\nSo far, most of the top four CSPs’ capex has been financed through operating cash flows, with limited leverage, implying no immediate funding risk.\n    \n• Given the competitive nature of generative-AI investments, the top four CSPs are likely to continue aggressive capex spending into 2026 and 2027.\nHowever, for smaller players such as Oracle and CoreWeave, debt financing is becoming an increasingly important factor in funding expansion.\n    \n• Still, these smaller CSPs account for only ~11% of the top six CSPs’ total capex, so any structural shift in the composition of capital spending should be monitored beginning in 2026, when debt discussions begin in earnest.\nFor now, given ongoing constraints in chip supply, rack capacity, and data-center power availability, JPMorgan sees limited near-term downside risk to the AI capex cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003175670555116028, "ret_signed_m1d": 0.003175670555116028, "alpha_spy_m1d": 0.01346908950418002, "alpha_c_m1d": 0.01581901308226108, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008103403516373953, "ret_signed_p1d": -0.008103403516373953, "alpha_spy_p1d": -0.008088407596361424, "alpha_c_p1d": -0.0025737154747123014, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.04845598508244486, "ret_signed_p1w": 0.04845598508244486, "alpha_spy_p1w": 0.027690293406607003, "alpha_c_p1w": 0.011822154465112256, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": -0.007008333179453041, "ret_signed_p1m": -0.007008333179453041, "alpha_spy_p1m": 0.009705442621301685, "alpha_c_p1m": 0.0331102941460224, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.024202926075399533, "ret_signed_p3m": 0.024202926075399533, "alpha_spy_p3m": -0.010037331239957759, "alpha_c_p3m": -0.1210175966846625, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 0.07605984568582302, "ret_signed_p6m": 0.07605984568582302, "alpha_spy_p6m": 0.03566915807900006, "alpha_c_p6m": -0.22975905773612992, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.065, -0.0488, -0.0501, -0.0323, -0.0391, -0.0185, -0.0262, -0.0489, -0.0145, -0.024, -0.0177, -0.0103, 0.0003, -0.0032, 0.0028, -0.0058, -0.0035, -0.012, -0.0035, -0.0384, -0.0397, -0.042, -0.0255, -0.0156, -0.0048, -0.0058, -0.0136, -0.0464, -0.0464, -0.065, -0.0659, -0.0602, -0.0856, -0.0785, -0.0651, -0.0291, -0.0299, -0.0264, -0.0268, -0.0425, -0.0676, -0.035, -0.0327, 0.0053, -0.0231, -0.031, -0.0271, -0.0244, -0.0043, 0.0216, 0.0252, 0.0342, 0.0273, 0.0159, 0.0131, 0.0354, 0.0544, 0.0028, 0.0311, -0.0143, -0.0154, -0.0045, 0.0032, 0.0, -0.0081, -0.0129, -0.0026, 0.0198, 0.0485, 0.1007, 0.1336, 0.1109, 0.1087, 0.1327, 0.0879, 0.0688, 0.0298, 0.0302, 0.0898, 0.0576, 0.0611, 0.0231, 0.0412, 0.0217, -0.007, 0.0212, -0.011, -0.0206, -0.0005, -0.0264, -0.013, -0.013, -0.0309, -0.0149, -0.0065, -0.0167, 0.0041, -0.0012, 0.016, 0.0128, 0.0063, -0.0093, -0.0417, -0.0347, -0.0269, -0.064, -0.0465, -0.009, 0.0058, 0.036, 0.0327, 0.0327, 0.0433, 0.0306, 0.0269, 0.0212, 0.0212, 0.0341, 0.0301, 0.0252, 0.0355, 0.0132, 0.0122, 0.0126, 0.0174, 0.0028, 0.0242, 0.0197, 0.0197, -0.025, 0.0038, 0.0121, 0.0276, 0.021, 0.0323, 0.0487, 0.0541, 0.0465, 0.0163, -0.0125, -0.0462, -0.0589, 0.0152, 0.0406, 0.0324, 0.0406, 0.0236, 0.001, 0.001, 0.0128, 0.0293, 0.0289, 0.0394, 0.0488, 0.056, 0.0708, 0.0124, -0.0298, -0.0008, -0.0141, 0.0022, 0.0039, -0.0263, 0.0001, 0.0117, 0.0187, 0.0028, -0.013, 0.0033, -0.0038, -0.0122, -0.0222, -0.0543, -0.0382, -0.0406, -0.0216, -0.0623, -0.0827, -0.0956, -0.045, -0.0376, -0.0286, -0.0286, -0.0273, -0.0248, -0.003, 0.0071, 0.0329, 0.0366, 0.0761]}
{"unit_id": "orig:1980129896954519889", "ticker_idx": 3, "ts": "2025-10-20T04:32:11+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA + AMD + ASIC revenue amounts to USD 287 billion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM relays bullish AI demand thesis: Asia Tech EPS +20–25% over 2–3 quarters, no bubble signs, TSMC CoWoS +60%, CSP capex upgrades into 2026–27.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JP Morgan: Surge in AI Demand and Price Increases Likely to Drive EPS Upgrades Through 1H26, No Signs of AI Bubble Burst\n\n⸻\n\n1. Continued EPS Upgrades Driven by AI Demand\n    \n• JPMorgan expects strong AI demand from the top four CSPs (cloud service providers) and incremental demand from players such as OpenAI and Oracle to lift Asia Tech EPS by an additional 20–25% over the next two to three quarters.\n\nThis improvement is likely to materialize mainly in 4Q25 and 1Q26 as the market re-rates growth expectations for the AI value chain in 2026.\n    \n• The firm currently forecasts capex for the top four CSPs to grow by 20% in 2026, with potential for further upside.\nThe hardware component of spending is expected to increase its share — for instance, TSMC’s CoWoS wafer shipments are projected to rise by around 60%.\n\n⸻\n\n2. No Signs of a Semiconductor Bubble — Capacity Expansion Still Absent\n    \n• Typical semiconductor/tech bubbles occur when a new demand frontier emerges rapidly (as with generative AI since 2023).\nThe supply chain races to expand production capacity to meet this demand, leading to a surge in wafer fab equipment (WFE) spending.\nHowever, because it takes 1.5–2 years for new capacity to come online, by the time it does, demand has already shifted, resulting in oversupply and a cyclical downturn.\n    \n• Applying this logic to the AI-driven tech cycle, JPMorgan notes that WFE spending has only rebounded 8% from the 2023 trough, and excluding China, the rebound is just 12% — far below the ~80% rebound typically seen in previous up-cycles.\nThus, there are no imminent signs of overheating or a bubble.\n    \n• To illustrate, CSP capex totals about USD 379 billion, while NVIDIA + AMD + ASIC revenue amounts to USD 287 billion, and downstream AI-server revenue is only USD 96 billion.\nThis disparity indicates persistent bottlenecks across the supply chain, from chips to racks and power infrastructure.\n\n⸻\n\n3. Potential for Further CSP Capex Upgrades into 2026 — Early Signs of Debt Utilization\n    \n• In addition to the top four CSPs’ strong demand, JPMorgan expects Oracle, OpenAI, and other AI labs to further lift CSP capex in 2026.\nSo far, most of the top four CSPs’ capex has been financed through operating cash flows, with limited leverage, implying no immediate funding risk.\n    \n• Given the competitive nature of generative-AI investments, the top four CSPs are likely to continue aggressive capex spending into 2026 and 2027.\nHowever, for smaller players such as Oracle and CoreWeave, debt financing is becoming an increasingly important factor in funding expansion.\n    \n• Still, these smaller CSPs account for only ~11% of the top six CSPs’ total capex, so any structural shift in the composition of capital spending should be monitored beginning in 2026, when debt discussions begin in earnest.\nFor now, given ongoing constraints in chip supply, rack capacity, and data-center power availability, JPMorgan sees limited near-term downside risk to the AI capex cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03109409628970894, "ret_signed_m1d": -0.03109409628970894, "alpha_spy_m1d": -0.020800677340644946, "alpha_c_m1d": -0.018450753762563887, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01051712173667041, "ret_signed_p1d": -0.01051712173667041, "alpha_spy_p1d": -0.010502125816657881, "alpha_c_p1d": -0.004987433695008758, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.079439707611761, "ret_signed_p1w": 0.079439707611761, "alpha_spy_p1w": 0.05867401593592314, "alpha_c_p1w": 0.04280587699442839, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": -0.042692070072703814, "ret_signed_p1m": -0.042692070072703814, "alpha_spy_p1m": -0.025978294271949087, "alpha_c_p1m": -0.0025734427472283716, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": -0.052544061847064505, "ret_signed_p3m": -0.052544061847064505, "alpha_spy_p3m": -0.0867843191624218, "alpha_c_p3m": -0.19776458460712654, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 0.06031763349220509, "ret_signed_p6m": 0.06031763349220509, "alpha_spy_p6m": 0.01992694588538213, "alpha_c_p6m": -0.24550126992974786, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.3405, -0.3261, -0.308, -0.2781, -0.2624, -0.2538, -0.2671, -0.2862, -0.2651, -0.2754, -0.3219, -0.2833, -0.2818, -0.2838, -0.2727, -0.2334, -0.2478, -0.2621, -0.2678, -0.3077, -0.3133, -0.3195, -0.3026, -0.3209, -0.3074, -0.3052, -0.2992, -0.324, -0.324, -0.3252, -0.326, -0.3274, -0.3717, -0.3706, -0.3523, -0.3368, -0.3529, -0.3408, -0.3301, -0.333, -0.3384, -0.3435, -0.3457, -0.3358, -0.3311, -0.3312, -0.3296, -0.3371, -0.3292, -0.3274, -0.3182, -0.2944, -0.3155, -0.1532, -0.1208, -0.0208, -0.0319, -0.1067, -0.1003, -0.0934, -0.0081, -0.0249, -0.0311, 0.0, -0.0105, -0.0429, -0.0232, 0.0514, 0.0794, 0.0725, 0.0988, 0.0594, 0.0647, 0.0794, 0.0394, 0.0656, -0.0119, -0.0292, 0.0142, -0.0126, 0.0762, 0.0308, 0.026, -0.0002, -0.0427, -0.0707, -0.1436, -0.1529, -0.106, -0.1431, -0.1094, -0.1094, -0.0957, -0.0865, -0.1053, -0.0954, -0.1022, -0.0939, -0.0809, -0.0787, -0.0796, -0.0795, -0.1238, -0.1371, -0.1305, -0.1765, -0.1642, -0.1128, -0.1065, -0.1067, -0.1061, -0.1061, -0.1063, -0.1037, -0.1048, -0.1097, -0.1097, -0.071, -0.081, -0.109, -0.127, -0.1492, -0.1554, -0.1366, -0.0814, -0.0705, -0.0525, -0.0363, -0.0363, -0.0359, 0.0384, 0.0547, 0.0795, 0.0447, 0.0477, 0.0506, 0.0483, -0.0159, 0.0237, 0.0064, -0.1678, -0.1998, -0.1335, -0.1021, -0.1122, -0.1122, -0.1439, -0.1382, -0.1382, -0.1558, -0.1681, -0.1546, -0.168, -0.1827, -0.1111, -0.1235, -0.1533, -0.1677, -0.1743, -0.2062, -0.16, -0.1709, -0.2001, -0.1575, -0.1552, -0.1485, -0.178, -0.1961, -0.1828, -0.1839, -0.1709, -0.1467, -0.1631, -0.1575, -0.1463, -0.0843, -0.1529, -0.1603, -0.1851, -0.1543, -0.1262, -0.0959, -0.0959, -0.0847, -0.0791, -0.0363, -0.0163, 0.0186, 0.0261, 0.0603]}
{"unit_id": "thread:1994274202585866583", "ticker_idx": 0, "ts": "2025-11-28T05:16:36+00:00", "symbol": "GBM", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwanese PCB firms such as PSA's GBM and Dynamic have passed certification for compute boards used in Google TPU v7", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan PCB makers GBM, Dynamic, Unimicron, GCE named as primary beneficiaries of Google TPU v7 ASIC supply chain ramp through 2026-2027.", "resolved_tickers": ["8462.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "GBM (Gold Bridge Machinery) PCB Taiwan 8462.TW", "bench_c": "XLY", "bench_c_reason": "industry='Leisure'", "bench_c_industry": "Leisure", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Meta's Google TPU shift opens new AI supply chain for Taiwan PCB makers\n\nRecent market rumors indicate that Meta is negotiating to deploy Google's seventh-generation TPU \"Ironwood\" in its data centers by 2027, with procurement potentially reaching tens of billions of US dollars. This move signals confidence that Google will become a leading ASIC chip supplier, challenging Nvidia's dominance and securing the second spot in the AI accelerator market.\n\nShifting market dynamics\nIndustry experts analyze that if confirmed, this development will shift the AI computing power market shares, presenting Nvidia with coexistence threats from Google-led ASIC forces. It will also accelerate the emergence of an alternative server PCB supply chain diverging from Nvidia's established path.\n\nBecause the ASIC server supply chain structure differs somewhat from the GPU ecosystem—and ASIC chips demand higher quantities and specifications for PCBs—Taiwanese manufacturers entering the Google TPU supply chain stand to benefit first from the PCB upgrade wave.\n\nTaiwan's strategic advantage\nCurrently, Nvidia holds a greater market share in PCB motherboards and upstream copper-clad laminate (CCL) materials through Chinese and South Korean suppliers. In contrast, ASIC vendors like Google, AWS, and Meta rely more heavily on Taiwanese suppliers.\n\nReportedly, Taiwanese PCB firms such as PSA's GBM and Dynamic have passed certification for compute boards used in Google TPU v7. These boards utilize M7-grade CCL materials supplied mainly by Japan's Panasonic and EMC.\n\nNotably, GBM leveraged its acquisition of Japanese PCB giant Lincstech to successfully enter Google's TPU supply chain. Together, they shipped 220,000 units of AI server compute and switch boards in the first three quarters of 2025, accounting for 45% of revenue, with full-year shipments expected to reach 320,000 units.\n\nExpanding portfolios\nDynamic, originally focused on automotive PCBs, has intensified efforts in networking and server markets, with plans to join Google TPU supplies starting in the second half of 2026. This expansion is projected to balance GPU and ASIC customer revenues evenly, pushing AI-related income from GPU, ASIC, and switch products above 20% of total sales.\n\nAdditionally, IC substrate leader Unimicron secured ABF substrate orders for Google TPU v7 advanced packaging via CoWoS technology. Meanwhile, GCE primarily supplies universal baseboards (UBB) for networking and server PCBs.\n\nGrowth projections\nDIGITIMES Research notes that, driven by increased external computing demands from Anthropic and Apple, alongside delays in AWS's next-generation Trainium3 production, Google TPU stands out as the most significant ASIC product in 2026. Shipments are forecasted at 3.325 million units, nearly one million more than in 2025, maintaining Google's lead in global ASIC accelerators.\n\nOfficial announcements describe TPU v7 as Google's custom-designed ASIC chip tailored for large language model (LLM) training and AI inference workloads. Its peak performance improves tenfold over TPU v5p, with a formal launch planned within weeks.\n\nGoogle TPU v7p reportedly entered mass production in the third quarter of 2025, and related PCB supply chains are expected to ramp up ASIC customer shipments starting in the first quarter of 2026, contributing to steadily growing AI application revenues quarter by quarter.\n\n---\n\nhttps://t.co/djDczHGrCN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012698412698412653, "ret_signed_m1d": -0.012698412698412653, "alpha_spy_m1d": -0.007269638813581558, "alpha_c_m1d": -0.006523998286893318, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.0061744144115193356, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012698412698412653, "ret_signed_p1d": -0.012698412698412653, "alpha_spy_p1d": -0.00813299709586146, "alpha_c_p1d": -0.013290509121435035, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0005920964230223813, "ret_p1w": -0.01904761904761909, "ret_signed_p1w": -0.01904761904761909, "alpha_spy_p1w": -0.022413207357755094, "alpha_c_p1w": -0.03173477017299098, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.012687151125371887, "ret_p1m": -0.050793650793650835, "ret_signed_p1m": -0.050793650793650835, "alpha_spy_p1m": -0.06029409491298354, "alpha_c_p1m": -0.07497954898623749, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.024185898192586652, "ret_p3m": -0.0888888888888889, "ret_signed_p3m": -0.0888888888888889, "alpha_spy_p3m": -0.10616782106954481, "alpha_c_p3m": -0.08120932248047108, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": -0.007679566408417826, "ret_p6m": -0.12380952380952381, "ret_signed_p6m": -0.12380952380952381, "alpha_spy_p6m": -0.2211122118160742, "alpha_c_p6m": -0.13587140711800227, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.01206188330847846, "price_path": [0.0254, 0.0317, 0.0127, 0.0159, 0.0413, 0.0635, 0.0698, 0.0286, 0.019, 0.0127, -0.0032, -0.0159, -0.0222, -0.019, -0.0032, -0.0127, -0.0032, 0.0254, 0.0349, 0.0349, 0.0413, 0.054, 0.0571, 0.019, 0.019, 0.0222, 0.0063, 0.0032, 0.0032, 0.0032, 0.019, 0.0413, 0.0317, 0.0317, 0.0349, 0.0349, 0.054, 0.0254, 0.0254, 0.0317, 0.0159, 0.019, 0.0349, 0.0413, 0.0476, 0.0667, 0.0571, 0.073, 0.0349, 0.0254, 0.0127, 0.0, -0.0095, -0.0159, -0.0413, -0.0317, -0.0095, 0.0032, 0.0, 0.0032, -0.0095, -0.0032, -0.0127, 0.0, -0.0127, -0.0222, -0.0254, -0.019, -0.019, -0.0159, -0.0222, -0.0254, -0.0476, -0.0317, -0.0349, -0.0508, -0.0603, -0.0635, -0.0603, -0.073, -0.073, -0.073, -0.073, -0.0603, -0.0508, -0.054, -0.0413, -0.0413, -0.0635, -0.0603, -0.0444, -0.019, 0.0032, 0.0063, 0.0095, 0.0063, -0.0095, -0.0127, -0.0254, -0.0413, -0.0508, -0.0476, -0.0698, -0.0571, -0.0635, -0.0698, -0.0825, -0.1048, -0.0984, -0.1206, -0.1206, -0.1111, -0.0952, -0.0921, -0.1048, -0.1016, -0.0984, -0.0984, -0.0984, -0.0984, -0.0984, -0.0984, -0.0984, -0.0984, -0.073, -0.0952, -0.0889, -0.0921, -0.0921, -0.1016, -0.1048, -0.127, -0.1143, -0.1111, -0.146, -0.127, -0.1206, -0.0889, -0.0762, -0.0508, -0.0667, -0.0635, -0.0921, -0.0921, -0.1016, -0.1048, -0.0984, -0.0762, -0.0825, -0.1143, -0.1302, -0.1048, -0.1206, -0.1206, -0.1206, -0.1175, -0.0984, -0.1175, -0.1175, -0.1079, -0.1111, -0.1079, -0.0984, -0.0984, -0.0952, -0.0857, -0.0857, -0.0857, -0.0889, -0.0889, -0.0508, -0.0508, -0.054, -0.054, -0.0635, -0.0603, -0.0635, -0.0635, -0.0603, -0.0635, -0.0698, -0.073, -0.0857, -0.0984, -0.1143, -0.1143, -0.1238, -0.1143, -0.1206, -0.1238]}
{"unit_id": "thread:1994274202585866583", "ticker_idx": 1, "ts": "2025-11-28T05:16:36+00:00", "symbol": "Dynamic", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Dynamic, originally focused on automotive PCBs, has intensified efforts in networking and server markets, with plans to join Google TPU supplies starting in the second half of 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan PCB makers GBM, Dynamic, Unimicron, GCE named as primary beneficiaries of Google TPU v7 ASIC supply chain ramp through 2026-2027.", "resolved_tickers": ["6214.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Dynamic Electronics 6214.TW Taiwan PCB for Google TPU", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Meta's Google TPU shift opens new AI supply chain for Taiwan PCB makers\n\nRecent market rumors indicate that Meta is negotiating to deploy Google's seventh-generation TPU \"Ironwood\" in its data centers by 2027, with procurement potentially reaching tens of billions of US dollars. This move signals confidence that Google will become a leading ASIC chip supplier, challenging Nvidia's dominance and securing the second spot in the AI accelerator market.\n\nShifting market dynamics\nIndustry experts analyze that if confirmed, this development will shift the AI computing power market shares, presenting Nvidia with coexistence threats from Google-led ASIC forces. It will also accelerate the emergence of an alternative server PCB supply chain diverging from Nvidia's established path.\n\nBecause the ASIC server supply chain structure differs somewhat from the GPU ecosystem—and ASIC chips demand higher quantities and specifications for PCBs—Taiwanese manufacturers entering the Google TPU supply chain stand to benefit first from the PCB upgrade wave.\n\nTaiwan's strategic advantage\nCurrently, Nvidia holds a greater market share in PCB motherboards and upstream copper-clad laminate (CCL) materials through Chinese and South Korean suppliers. In contrast, ASIC vendors like Google, AWS, and Meta rely more heavily on Taiwanese suppliers.\n\nReportedly, Taiwanese PCB firms such as PSA's GBM and Dynamic have passed certification for compute boards used in Google TPU v7. These boards utilize M7-grade CCL materials supplied mainly by Japan's Panasonic and EMC.\n\nNotably, GBM leveraged its acquisition of Japanese PCB giant Lincstech to successfully enter Google's TPU supply chain. Together, they shipped 220,000 units of AI server compute and switch boards in the first three quarters of 2025, accounting for 45% of revenue, with full-year shipments expected to reach 320,000 units.\n\nExpanding portfolios\nDynamic, originally focused on automotive PCBs, has intensified efforts in networking and server markets, with plans to join Google TPU supplies starting in the second half of 2026. This expansion is projected to balance GPU and ASIC customer revenues evenly, pushing AI-related income from GPU, ASIC, and switch products above 20% of total sales.\n\nAdditionally, IC substrate leader Unimicron secured ABF substrate orders for Google TPU v7 advanced packaging via CoWoS technology. Meanwhile, GCE primarily supplies universal baseboards (UBB) for networking and server PCBs.\n\nGrowth projections\nDIGITIMES Research notes that, driven by increased external computing demands from Anthropic and Apple, alongside delays in AWS's next-generation Trainium3 production, Google TPU stands out as the most significant ASIC product in 2026. Shipments are forecasted at 3.325 million units, nearly one million more than in 2025, maintaining Google's lead in global ASIC accelerators.\n\nOfficial announcements describe TPU v7 as Google's custom-designed ASIC chip tailored for large language model (LLM) training and AI inference workloads. Its peak performance improves tenfold over TPU v5p, with a formal launch planned within weeks.\n\nGoogle TPU v7p reportedly entered mass production in the third quarter of 2025, and related PCB supply chains are expected to ramp up ASIC customer shipments starting in the first quarter of 2026, contributing to steadily growing AI application revenues quarter by quarter.\n\n---\n\nhttps://t.co/djDczHGrCN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012448132780082943, "ret_signed_m1d": 0.012448132780082943, "alpha_spy_m1d": 0.01787690666491404, "alpha_c_m1d": 0.020972970150845316, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.008524837370762373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008298755186721962, "ret_signed_p1d": -0.008298755186721962, "alpha_spy_p1d": -0.0037333395841707695, "alpha_c_p1d": -0.00875304680986888, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0004542916231469185, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.003365588310136003, "alpha_c_p1w": -0.02438687470350498, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.02438687470350498, "ret_p1m": 0.012448132780082943, "ret_signed_p1m": 0.012448132780082943, "alpha_spy_p1m": 0.002947688660750236, "alpha_c_p1m": -0.00838384849484819, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.020831981274931133, "ret_p3m": -0.08713692946058094, "ret_signed_p3m": -0.08713692946058094, "alpha_spy_p3m": -0.10441586164123684, "alpha_c_p3m": -0.08795388457765918, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.0008169551170782441, "ret_p6m": 0.0705394190871369, "ret_signed_p6m": 0.0705394190871369, "alpha_spy_p6m": -0.026763268919413497, "alpha_c_p6m": -0.19344111426555566, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.26398053335269256, "price_path": [-0.0415, -0.0249, -0.0166, -0.0083, 0.0124, 0.0166, 0.0332, 0.029, 0.0332, 0.0124, 0.0166, 0.0083, 0.0124, 0.0166, 0.0166, 0.0166, 0.0, 0.0083, -0.0124, -0.0124, 0.0041, -0.0083, -0.0124, -0.0041, -0.0041, 0.0124, 0.0249, 0.1245, 0.1245, 0.1245, 0.0581, 0.112, 0.0871, 0.0622, 0.0913, 0.1079, 0.0788, 0.083, 0.083, 0.1203, 0.0954, 0.0788, 0.0664, 0.0747, 0.0664, 0.0622, 0.0498, 0.0622, 0.0664, 0.0415, 0.0581, 0.0622, 0.0456, 0.0166, -0.0083, -0.0083, -0.0332, 0.0041, -0.0207, 0.0, 0.0, 0.0124, 0.0124, 0.0, -0.0083, 0.0041, 0.029, 0.0166, 0.0, 0.0124, 0.0166, 0.0041, 0.0, 0.0083, -0.0124, -0.0207, -0.0207, -0.0207, 0.0124, 0.0747, 0.0581, 0.0622, 0.0622, 0.0207, 0.0124, -0.0041, -0.0207, -0.0207, -0.0207, -0.0332, -0.0415, -0.0415, -0.0373, -0.0332, -0.0166, -0.029, -0.0166, -0.0166, -0.0249, -0.0207, -0.029, -0.0373, -0.0332, -0.0249, 0.0041, 0.0124, 0.0041, -0.0166, -0.0373, -0.0581, -0.0498, -0.0581, -0.0871, -0.083, -0.0954, -0.0705, -0.0747, -0.0747, -0.0747, -0.0747, -0.0747, -0.0747, -0.0747, -0.0747, -0.0913, -0.0954, -0.0871, -0.0456, -0.0456, -0.0581, -0.0705, -0.0996, -0.0788, -0.0705, -0.0996, -0.0913, -0.0664, -0.0788, -0.0788, -0.0622, -0.0747, -0.0581, -0.0705, -0.0539, -0.0913, -0.0954, -0.0871, -0.0954, -0.0996, -0.1162, -0.1286, -0.1079, -0.1079, -0.1079, -0.1079, -0.0871, -0.0788, -0.0456, -0.0498, -0.0373, -0.0166, -0.0083, 0.0249, 0.0456, 0.0498, 0.0581, 0.0456, 0.0041, -0.0041, -0.0124, 0.0415, 0.0083, -0.0041, -0.0041, 0.0083, 0.0373, 0.0415, 0.0373, 0.0415, 0.0539, 0.029, 0.0124, 0.0124, 0.0124, 0.0249, 0.0249, 0.029, 0.0581, 0.0788, 0.0705]}
{"unit_id": "thread:1994274202585866583", "ticker_idx": 2, "ts": "2025-11-28T05:16:36+00:00", "symbol": "Unimicron", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "IC substrate leader Unimicron secured ABF substrate orders for Google TPU v7 advanced packaging via CoWoS technology", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan PCB makers GBM, Dynamic, Unimicron, GCE named as primary beneficiaries of Google TPU v7 ASIC supply chain ramp through 2026-2027.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Meta's Google TPU shift opens new AI supply chain for Taiwan PCB makers\n\nRecent market rumors indicate that Meta is negotiating to deploy Google's seventh-generation TPU \"Ironwood\" in its data centers by 2027, with procurement potentially reaching tens of billions of US dollars. This move signals confidence that Google will become a leading ASIC chip supplier, challenging Nvidia's dominance and securing the second spot in the AI accelerator market.\n\nShifting market dynamics\nIndustry experts analyze that if confirmed, this development will shift the AI computing power market shares, presenting Nvidia with coexistence threats from Google-led ASIC forces. It will also accelerate the emergence of an alternative server PCB supply chain diverging from Nvidia's established path.\n\nBecause the ASIC server supply chain structure differs somewhat from the GPU ecosystem—and ASIC chips demand higher quantities and specifications for PCBs—Taiwanese manufacturers entering the Google TPU supply chain stand to benefit first from the PCB upgrade wave.\n\nTaiwan's strategic advantage\nCurrently, Nvidia holds a greater market share in PCB motherboards and upstream copper-clad laminate (CCL) materials through Chinese and South Korean suppliers. In contrast, ASIC vendors like Google, AWS, and Meta rely more heavily on Taiwanese suppliers.\n\nReportedly, Taiwanese PCB firms such as PSA's GBM and Dynamic have passed certification for compute boards used in Google TPU v7. These boards utilize M7-grade CCL materials supplied mainly by Japan's Panasonic and EMC.\n\nNotably, GBM leveraged its acquisition of Japanese PCB giant Lincstech to successfully enter Google's TPU supply chain. Together, they shipped 220,000 units of AI server compute and switch boards in the first three quarters of 2025, accounting for 45% of revenue, with full-year shipments expected to reach 320,000 units.\n\nExpanding portfolios\nDynamic, originally focused on automotive PCBs, has intensified efforts in networking and server markets, with plans to join Google TPU supplies starting in the second half of 2026. This expansion is projected to balance GPU and ASIC customer revenues evenly, pushing AI-related income from GPU, ASIC, and switch products above 20% of total sales.\n\nAdditionally, IC substrate leader Unimicron secured ABF substrate orders for Google TPU v7 advanced packaging via CoWoS technology. Meanwhile, GCE primarily supplies universal baseboards (UBB) for networking and server PCBs.\n\nGrowth projections\nDIGITIMES Research notes that, driven by increased external computing demands from Anthropic and Apple, alongside delays in AWS's next-generation Trainium3 production, Google TPU stands out as the most significant ASIC product in 2026. Shipments are forecasted at 3.325 million units, nearly one million more than in 2025, maintaining Google's lead in global ASIC accelerators.\n\nOfficial announcements describe TPU v7 as Google's custom-designed ASIC chip tailored for large language model (LLM) training and AI inference workloads. Its peak performance improves tenfold over TPU v5p, with a formal launch planned within weeks.\n\nGoogle TPU v7p reportedly entered mass production in the third quarter of 2025, and related PCB supply chains are expected to ramp up ASIC customer shipments starting in the first quarter of 2026, contributing to steadily growing AI application revenues quarter by quarter.\n\n---\n\nhttps://t.co/djDczHGrCN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002680965147453085, "ret_signed_m1d": 0.002680965147453085, "alpha_spy_m1d": 0.00810973903228418, "alpha_c_m1d": 0.011205802518215457, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.008524837370762373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05630026809651478, "ret_signed_p1d": 0.05630026809651478, "alpha_spy_p1d": 0.060865683699065976, "alpha_c_p1d": 0.055845976473367864, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0004542916231469185, "ret_p1w": 0.15281501340482584, "ret_signed_p1w": 0.15281501340482584, "alpha_spy_p1w": 0.14944942509468984, "alpha_c_p1w": 0.12842813870132086, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.02438687470350498, "ret_p1m": 0.18230563002680955, "ret_signed_p1m": 0.18230563002680955, "alpha_spy_p1m": 0.17280518590747684, "alpha_c_p1m": 0.16147364875187842, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.020831981274931133, "ret_p3m": 1.4584450402144773, "ret_signed_p3m": 1.4584450402144773, "alpha_spy_p3m": 1.4411661080338214, "alpha_c_p3m": 1.457628085097399, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.0008169551170782441, "ret_p6m": 4.3083109919571045, "ret_signed_p6m": 4.3083109919571045, "alpha_spy_p6m": 4.211008303950554, "alpha_c_p6m": 4.044330458604412, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.26398053335269256, "price_path": [-0.2976, -0.2895, -0.2842, -0.2815, -0.2842, -0.2761, -0.2413, -0.2359, -0.2252, -0.2386, -0.185, -0.2091, -0.2038, -0.1555, -0.1501, -0.1475, -0.1582, -0.1582, -0.2011, -0.2011, -0.1877, -0.1689, -0.1796, -0.1635, -0.1635, -0.1528, -0.1421, -0.1367, -0.1367, -0.1769, -0.2172, -0.1501, -0.1126, -0.1394, -0.1421, -0.1421, -0.1475, -0.1635, -0.1635, -0.1233, -0.1287, -0.1099, -0.1421, -0.1233, -0.0429, -0.0965, -0.1274, -0.1195, -0.1407, -0.1248, -0.1141, -0.1088, -0.0503, -0.0804, -0.1019, -0.0885, -0.1394, -0.0804, -0.1421, -0.1206, -0.0563, -0.0429, 0.0027, 0.0, 0.0563, 0.0509, 0.0536, 0.0643, 0.1528, 0.1823, 0.2118, 0.2198, 0.2064, 0.2064, 0.1877, 0.1475, 0.1367, 0.1099, 0.1528, 0.1823, 0.1555, 0.1635, 0.1635, 0.1823, 0.1823, 0.1769, 0.1796, 0.1796, 0.1743, 0.2332, 0.2359, 0.2118, 0.2198, 0.2091, 0.3056, 0.3619, 0.3405, 0.3378, 0.4692, 0.6139, 0.6515, 0.7212, 0.7534, 0.8177, 0.7855, 0.8981, 0.9705, 0.8472, 1.0295, 0.9571, 1.0643, 1.0804, 0.9598, 0.8231, 0.8633, 0.992, 0.9786, 0.9786, 0.9786, 0.9786, 0.9786, 0.9786, 0.9786, 0.9786, 1.1743, 1.3914, 1.4584, 1.5818, 1.5818, 1.5791, 1.4826, 1.2359, 1.4129, 1.3164, 1.0858, 1.2735, 1.4987, 1.5737, 1.7239, 1.9437, 1.9115, 2.1099, 2.1153, 1.9705, 1.6756, 1.4665, 1.7131, 1.6917, 1.7078, 1.6461, 1.3834, 1.6193, 1.7828, 1.7828, 1.7828, 2.0241, 2.3244, 2.3512, 2.4209, 2.3566, 2.2011, 2.3137, 2.4584, 2.4477, 2.6354, 2.807, 2.8391, 2.9035, 3.2359, 3.4987, 3.4236, 3.3056, 3.7346, 3.7346, 3.8847, 3.8418, 3.6005, 3.8043, 3.3861, 3.6166, 3.6917, 3.7828, 3.7239, 3.4021, 3.3753, 3.3861, 3.4129, 3.8525, 4.2011, 4.3083]}
{"unit_id": "thread:1994274202585866583", "ticker_idx": 3, "ts": "2025-11-28T05:16:36+00:00", "symbol": "GCE", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GCE primarily supplies universal baseboards (UBB) for networking and server PCBs", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan PCB makers GBM, Dynamic, Unimicron, GCE named as primary beneficiaries of Google TPU v7 ASIC supply chain ramp through 2026-2027.", "resolved_tickers": ["6691.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "GCE Taiwan PCB/substrate 6691.TW", "bench_c": "XLI", "bench_c_reason": "industry='Engineering & Construction'", "bench_c_industry": "Engineering & Construction", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Meta's Google TPU shift opens new AI supply chain for Taiwan PCB makers\n\nRecent market rumors indicate that Meta is negotiating to deploy Google's seventh-generation TPU \"Ironwood\" in its data centers by 2027, with procurement potentially reaching tens of billions of US dollars. This move signals confidence that Google will become a leading ASIC chip supplier, challenging Nvidia's dominance and securing the second spot in the AI accelerator market.\n\nShifting market dynamics\nIndustry experts analyze that if confirmed, this development will shift the AI computing power market shares, presenting Nvidia with coexistence threats from Google-led ASIC forces. It will also accelerate the emergence of an alternative server PCB supply chain diverging from Nvidia's established path.\n\nBecause the ASIC server supply chain structure differs somewhat from the GPU ecosystem—and ASIC chips demand higher quantities and specifications for PCBs—Taiwanese manufacturers entering the Google TPU supply chain stand to benefit first from the PCB upgrade wave.\n\nTaiwan's strategic advantage\nCurrently, Nvidia holds a greater market share in PCB motherboards and upstream copper-clad laminate (CCL) materials through Chinese and South Korean suppliers. In contrast, ASIC vendors like Google, AWS, and Meta rely more heavily on Taiwanese suppliers.\n\nReportedly, Taiwanese PCB firms such as PSA's GBM and Dynamic have passed certification for compute boards used in Google TPU v7. These boards utilize M7-grade CCL materials supplied mainly by Japan's Panasonic and EMC.\n\nNotably, GBM leveraged its acquisition of Japanese PCB giant Lincstech to successfully enter Google's TPU supply chain. Together, they shipped 220,000 units of AI server compute and switch boards in the first three quarters of 2025, accounting for 45% of revenue, with full-year shipments expected to reach 320,000 units.\n\nExpanding portfolios\nDynamic, originally focused on automotive PCBs, has intensified efforts in networking and server markets, with plans to join Google TPU supplies starting in the second half of 2026. This expansion is projected to balance GPU and ASIC customer revenues evenly, pushing AI-related income from GPU, ASIC, and switch products above 20% of total sales.\n\nAdditionally, IC substrate leader Unimicron secured ABF substrate orders for Google TPU v7 advanced packaging via CoWoS technology. Meanwhile, GCE primarily supplies universal baseboards (UBB) for networking and server PCBs.\n\nGrowth projections\nDIGITIMES Research notes that, driven by increased external computing demands from Anthropic and Apple, alongside delays in AWS's next-generation Trainium3 production, Google TPU stands out as the most significant ASIC product in 2026. Shipments are forecasted at 3.325 million units, nearly one million more than in 2025, maintaining Google's lead in global ASIC accelerators.\n\nOfficial announcements describe TPU v7 as Google's custom-designed ASIC chip tailored for large language model (LLM) training and AI inference workloads. Its peak performance improves tenfold over TPU v5p, with a formal launch planned within weeks.\n\nGoogle TPU v7p reportedly entered mass production in the third quarter of 2025, and related PCB supply chains are expected to ramp up ASIC customer shipments starting in the first quarter of 2026, contributing to steadily growing AI application revenues quarter by quarter.\n\n---\n\nhttps://t.co/djDczHGrCN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03379721669980118, "ret_signed_m1d": -0.03379721669980118, "alpha_spy_m1d": -0.028368442814970085, "alpha_c_m1d": -0.028006346583728026, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.005790870116073155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03876739562624254, "ret_signed_p1d": -0.03876739562624254, "alpha_spy_p1d": -0.03420198002369135, "alpha_c_p1d": -0.024127628407932145, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": -0.014639767218310396, "ret_p1w": -0.016898608349900646, "ret_signed_p1w": -0.016898608349900646, "alpha_spy_p1w": -0.02026419666003665, "alpha_c_p1w": -0.0226894784659738, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.005790870116073155, "ret_p1m": 0.09940357852882697, "ret_signed_p1m": 0.09940357852882697, "alpha_spy_p1m": 0.08990313440949427, "alpha_c_p1m": 0.07521669301891976, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.02418688550990722, "ret_p3m": 0.23459244532803192, "ret_signed_p3m": 0.23459244532803192, "alpha_spy_p3m": 0.21731351314737601, "alpha_c_p3m": 0.0881194463776116, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.14647299895042032, "ret_p6m": 0.2842942345924453, "ret_signed_p6m": 0.2842942345924453, "alpha_spy_p6m": 0.1869915465858949, "alpha_c_p6m": 0.15967568874272042, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.12461854584972487, "price_path": [-0.006, -0.0278, -0.0487, -0.0209, -0.0308, -0.0328, -0.0408, -0.0487, -0.0636, -0.0835, -0.0775, -0.1153, -0.1123, -0.1262, -0.1173, -0.1074, -0.1034, -0.1133, -0.1352, -0.1352, -0.1203, -0.1213, -0.1272, -0.1332, -0.1332, -0.1064, -0.1024, -0.1004, -0.1004, -0.1034, -0.1133, -0.0825, -0.0805, -0.0905, -0.1083, -0.1113, -0.1213, -0.1243, -0.1243, -0.1133, -0.1113, -0.0954, -0.0875, -0.0855, -0.1014, -0.0984, -0.0974, -0.0855, -0.0865, -0.1024, -0.1282, -0.1421, -0.1153, -0.1312, -0.1143, -0.1421, -0.1183, -0.0666, -0.1054, -0.0934, -0.0755, -0.0328, -0.0338, 0.0, -0.0388, -0.0586, -0.0408, -0.0378, -0.0169, 0.0795, 0.1014, 0.1332, 0.1233, 0.2187, 0.2545, 0.2366, 0.1968, 0.1789, 0.1988, 0.161, 0.1571, 0.1233, 0.1233, 0.1233, 0.0994, 0.0954, 0.1093, 0.1093, 0.1014, 0.1352, 0.1451, 0.1272, 0.163, 0.159, 0.1789, 0.2008, 0.2227, 0.2406, 0.1789, 0.171, 0.1928, 0.173, 0.173, 0.1988, 0.1928, 0.1968, 0.2028, 0.2187, 0.1789, 0.167, 0.2008, 0.1948, 0.167, 0.173, 0.175, 0.1849, 0.1909, 0.1909, 0.1909, 0.1909, 0.1909, 0.1909, 0.1909, 0.1909, 0.1988, 0.2286, 0.2346, 0.3579, 0.3579, 0.3459, 0.3042, 0.2247, 0.2922, 0.2624, 0.1968, 0.2068, 0.2445, 0.2644, 0.2286, 0.2227, 0.2087, 0.2286, 0.2227, 0.2147, 0.2187, 0.2028, 0.2227, 0.173, 0.1769, 0.1511, 0.1034, 0.1412, 0.1153, 0.1153, 0.1153, 0.1173, 0.1571, 0.1928, 0.2525, 0.2406, 0.2604, 0.2684, 0.3082, 0.3101, 0.3221, 0.3459, 0.3598, 0.3439, 0.3419, 0.2664, 0.2684, 0.2644, 0.2545, 0.2545, 0.2744, 0.2704, 0.3459, 0.334, 0.3002, 0.2783, 0.2584, 0.2565, 0.2664, 0.2485, 0.2386, 0.2326, 0.2187, 0.2346, 0.2485, 0.2843]}
{"unit_id": "reply:1987015238877581548", "ticker_idx": 0, "ts": "2025-11-08T04:32:04+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "in 2026, shouldn't the sector be allowed to trade at a somewhat higher valuation?", "tweet_type": "prediction", "timeline": "quarters", "summary": "Bullish DRAM sector into 2026 on constrained cleanroom capacity / no new supply until 2027; conservative (bearish) on NAND.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "I think of it this way.\n\nRight now, the three major memory companies don’t have cleanroom capacity. Even if they want to expand, they can’t.\n\nAmong the fabs currently under construction, the earliest one won’t come online until 2027.\n\nSo in 2026, shouldn’t the sector be allowed to trade at a somewhat higher valuation?\n\nOf course, this is assuming DRAM only. I’m still conservative on NAND.", "reply_text": null, "parent_text": "@Jukanlosreve @BenBajarin Tsmc does that because tsmc has no competition. Also tsmc is not building standardized products like ddr and nand which are close to pure commodities like oil and gas.\n\nAll it takes is one player to overbuild for the market to get loose. And even if they wait for LTAs to build,", "quoted_text": null, "parent_author": "evrgn11112231", "ret_m1d": -0.04348723941581648, "ret_signed_m1d": -0.04348723941581648, "alpha_spy_m1d": -0.028122561735264763, "alpha_c_m1d": -0.013748396878358972, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0007181129709183942, "ret_signed_p1d": 0.0007181129709183942, "alpha_spy_p1d": -0.0015710367315273643, "alpha_c_p1d": 0.02229065484291859, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.014936176078392297, "ret_signed_p1w": -0.014936176078392297, "alpha_spy_p1w": 0.008206097822922994, "alpha_c_p1w": 0.036598238984896746, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": 0.0029076674749224374, "ret_signed_p1m": 0.0029076674749224374, "alpha_spy_p1m": 0.0005598770588148838, "alpha_c_p1m": -0.02554904742859332, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.49794942706365525, "ret_signed_p3m": 0.49794942706365525, "alpha_spy_p3m": 0.5006169039510169, "alpha_c_p3m": 0.4325470103123638, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 1.4168680594793785, "ret_signed_p6m": 1.4168680594793785, "alpha_spy_p6m": 1.3487019680802148, "alpha_c_p6m": 0.9555218996935289, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.4467, -0.4472, -0.453, -0.4597, -0.4641, -0.473, -0.4803, -0.4718, -0.4685, -0.4713, -0.4694, -0.4623, -0.4656, -0.4797, -0.473, -0.4692, -0.4594, -0.4475, -0.4435, -0.4279, -0.4092, -0.391, -0.3638, -0.3581, -0.3378, -0.3482, -0.3175, -0.3256, -0.315, -0.3032, -0.309, -0.3136, -0.3333, -0.3135, -0.3108, -0.2775, -0.2434, -0.2379, -0.2338, -0.2407, -0.2264, -0.232, -0.2128, -0.2089, -0.224, -0.2002, -0.1609, -0.1532, -0.1358, -0.1473, -0.1473, -0.145, -0.1039, -0.0781, -0.0917, -0.0618, -0.0478, -0.0419, 0.018, -0.0432, -0.0356, -0.0315, -0.0435, 0.0, 0.0007, 0.0033, -0.0109, -0.0451, -0.0149, -0.0617, -0.0738, -0.0876, -0.1264, -0.0989, -0.09, -0.0681, -0.0546, -0.0641, -0.0534, -0.0351, -0.0418, -0.0551, -0.0292, 0.0054, 0.0029, 0.028, 0.0067, -0.0075, -0.0353, -0.0617, -0.0424, -0.0125, 0.0033, 0.0494, 0.0545, 0.069, 0.069, 0.0922, 0.1377, 0.1428, 0.1333, 0.1333, 0.2158, 0.2539, 0.3142, 0.3249, 0.3088, 0.3267, 0.3298, 0.3095, 0.3144, 0.3346, 0.3895, 0.3952, 0.373, 0.4174, 0.4461, 0.4548, 0.4238, 0.5115, 0.5767, 0.5827, 0.5809, 0.534, 0.6091, 0.5579, 0.4979, 0.5095, 0.5472, 0.5256, 0.5723, 0.6284, 0.6297, 0.6297, 0.614, 0.6419, 0.6742, 0.719, 0.7203, 0.7694, 0.8027, 0.879, 0.8488, 0.8492, 0.6673, 0.5692, 0.6796, 0.6237, 0.5512, 0.6736, 0.7105, 0.6721, 0.6738, 0.747, 0.7883, 0.8844, 0.8111, 0.776, 0.6671, 0.6961, 0.681, 0.5821, 0.5844, 0.4935, 0.4478, 0.6119, 0.5353, 0.5867, 0.6304, 0.658, 0.8077, 0.7861, 0.8076, 0.8059, 0.9105, 0.9313, 0.9648, 0.9421, 0.9493, 0.9975, 1.0422, 1.059, 1.0604, 1.1524, 1.1218, 1.1499, 1.126, 1.159, 1.3329, 1.4169]}
{"unit_id": "reply:1987015238877581548", "ticker_idx": 1, "ts": "2025-11-08T04:32:04+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm still conservative on NAND", "tweet_type": "prediction", "timeline": "quarters", "summary": "Bullish DRAM sector into 2026 on constrained cleanroom capacity / no new supply until 2027; conservative (bearish) on NAND.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "I think of it this way.\n\nRight now, the three major memory companies don’t have cleanroom capacity. Even if they want to expand, they can’t.\n\nAmong the fabs currently under construction, the earliest one won’t come online until 2027.\n\nSo in 2026, shouldn’t the sector be allowed to trade at a somewhat higher valuation?\n\nOf course, this is assuming DRAM only. I’m still conservative on NAND.", "reply_text": null, "parent_text": "@Jukanlosreve @BenBajarin Tsmc does that because tsmc has no competition. Also tsmc is not building standardized products like ddr and nand which are close to pure commodities like oil and gas.\n\nAll it takes is one player to overbuild for the market to get loose. And even if they wait for LTAs to build,", "quoted_text": null, "parent_author": "evrgn11112231", "ret_m1d": -0.06656180745350493, "ret_signed_m1d": 0.06656180745350493, "alpha_spy_m1d": 0.05119712977295321, "alpha_c_m1d": 0.03682296491604742, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0012634323685315695, "ret_signed_p1d": -0.0012634323685315695, "alpha_spy_p1d": 0.001025717333914189, "alpha_c_p1d": -0.022835974240531763, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.04286413277854104, "ret_signed_p1w": 0.04286413277854104, "alpha_spy_p1w": 0.019721858877225752, "alpha_c_p1w": -0.008670282284747997, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": -0.08619046335725403, "ret_signed_p1m": 0.08619046335725403, "alpha_spy_p1m": 0.08853825377336158, "alpha_c_p1m": 0.11464717826076978, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.6220931278501748, "ret_signed_p3m": -0.6220931278501748, "alpha_spy_p3m": -0.6247606047375365, "alpha_c_p3m": -0.5566907110988834, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 2.046505819170048, "ret_signed_p6m": -2.046505819170048, "alpha_spy_p6m": -1.9783397277708843, "alpha_c_p6m": -1.5851596593841983, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.5942, -0.5951, -0.6001, -0.6052, -0.6081, -0.6152, -0.6196, -0.6124, -0.6103, -0.6109, -0.6092, -0.6023, -0.6012, -0.6085, -0.6064, -0.6026, -0.5895, -0.5713, -0.5662, -0.5564, -0.538, -0.5115, -0.4873, -0.4809, -0.4637, -0.4721, -0.4488, -0.451, -0.4398, -0.4301, -0.4365, -0.4444, -0.4615, -0.4332, -0.4295, -0.4054, -0.3722, -0.3541, -0.3549, -0.3656, -0.3491, -0.349, -0.3479, -0.3324, -0.3493, -0.3175, -0.2855, -0.2888, -0.2636, -0.2712, -0.2689, -0.2517, -0.1915, -0.1678, -0.1832, -0.1332, -0.1189, -0.1138, -0.0721, -0.1191, -0.1015, -0.0912, -0.0666, 0.0, 0.0013, 0.0121, -0.0292, -0.0868, -0.0429, -0.1002, -0.098, -0.1361, -0.1758, -0.1393, -0.1415, -0.1545, -0.1364, -0.1305, -0.1418, -0.1294, -0.1447, -0.1377, -0.1054, -0.0761, -0.0862, -0.0672, -0.0714, -0.1021, -0.1323, -0.1503, -0.1314, -0.1009, -0.0805, -0.0384, -0.0351, -0.0127, -0.0098, 0.0133, 0.0252, 0.0217, 0.0138, 0.0138, 0.0916, 0.1273, 0.2237, 0.2494, 0.2302, 0.2682, 0.279, 0.2822, 0.2775, 0.3109, 0.3639, 0.3742, 0.3903, 0.4723, 0.5123, 0.4868, 0.4677, 0.5433, 0.6243, 0.6411, 0.6994, 0.6926, 0.7967, 0.6849, 0.6221, 0.6387, 0.6475, 0.6026, 0.6737, 0.7655, 0.7885, 0.7803, 0.7451, 0.7549, 0.7871, 0.8253, 0.8384, 0.8726, 0.8759, 0.9329, 0.902, 0.8947, 0.7257, 0.6776, 0.7347, 0.6677, 0.6409, 0.7599, 0.8365, 0.7842, 0.8144, 0.913, 0.935, 1.0444, 0.9987, 0.9311, 0.8469, 0.8588, 0.8516, 0.7173, 0.7147, 0.624, 0.6294, 0.8115, 0.7616, 0.8038, 0.8614, 0.8748, 1.0819, 1.1232, 1.1729, 1.264, 1.3768, 1.311, 1.3737, 1.3111, 1.3092, 1.3645, 1.4786, 1.4626, 1.4944, 1.6228, 1.5666, 1.6296, 1.658, 1.7281, 1.8838, 2.0465]}
{"unit_id": "quote:1980563978725343320", "ticker_idx": 0, "ts": "2025-10-21T09:17:04+00:00", "symbol": "3661.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The stock remains undervalued, trading at below 20x 2026 P/E, making it relatively inexpensive within the AI sector.", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish Alchip (3661): confirmed AWS Trn3 orders worth $3B+, potential $6B+ total, above-consensus EPS of $180/$240/$300 for 2026-2028, stock undervalued at <20x 2026 P/E.", "resolved_tickers": ["3661.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Update on Alchip (3661) — Trn3 and new customer developments\n\n    1. Reconfirmation of Trn3 schedule\nIn a recent closed-door meeting, the company reaffirmed that the Trn3 timeline remains unchanged.\n    • Trn3-1: Tape-out in Q1 2025, mass production in Q1 2026\n    • Trn3-2: Tape-out in Q3 2025, mass production in Q2 2026\n    2. AWS Trn3 2026 shipment confirmed\nAlchip currently has a confirmed order volume of at least 1 million units for AWS’s Trn3 project in 2026.\n    • Unit price: $3,000 per chip\n    • Estimated revenue contribution of at least $3 billion in 2026\n    3. 2026 Trn3 volume exceeds market expectations\nTSMC has agreed to provide capacity support for at least 3.5 million units,\nand Alchip may revise upward the total revenue scale of the Trn3 project to over $6 billion.\n    4.Actively seeking additional orders from Annapurna\nAlchip is actively negotiating with Annapurna Labs for more wafer input.\nThe main reason for the potential upward revision is that the CoWoS design capability and yield improvement for this project have exceeded expectations.\n    5. New customer engagements\n    • Intel Jaguar Shores (rack-level GPU): Alchip is in discussions with Intel, and the project may be finalized in 1H 2026.\n    • Meta Arke project: Alchip also has the opportunity to undertake the main chip backend design work.\n    6. EPS forecasts and valuation\n    • 2026/2027/2028 EPS: $180 / $240 / $300\n    • These figures are above market expectations and may be revised upward.\n    • The stock remains undervalued, trading at below 20x 2026 P/E, making it relatively inexpensive within the AI sector.", "reply_text": null, "parent_text": null, "quoted_text": "#市場小作文 #RUMOR\n轉 \n世芯3661 Trn3及新客戶近況更新\n1. 近期閉門會議公司重申Trn3時間點，確認無延期，維持Trn3-1 Q1 25 tape out，Q1 26量產。Trn3-2 Q3 25 tape out，Q2 26量產計畫。\n2. AWS Trn3 2026目前世芯有確定的量至少1mn，一顆$3000，營收貢獻2026年至少$3B\n3.", "parent_author": null, "ret_m1d": -0.03576751117734722, "ret_signed_m1d": -0.03576751117734722, "alpha_spy_m1d": -0.03578250732224064, "alpha_c_m1d": -0.041327946692784945, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04023845007451565, "ret_signed_p1d": -0.04023845007451565, "alpha_spy_p1d": -0.03503963411159494, "alpha_c_p1d": -0.0208350833868437, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": -0.08643815201192251, "ret_signed_p1w": -0.08643815201192251, "alpha_spy_p1w": -0.10993034450682748, "alpha_c_p1w": -0.13801651071178778, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": -0.010432190760059634, "ret_signed_p1m": -0.010432190760059634, "alpha_spy_p1m": 0.002468151444651623, "alpha_c_p1m": 0.006538485389608328, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": -0.0044709388971684305, "ret_signed_p3m": -0.0044709388971684305, "alpha_spy_p3m": -0.03786012982223297, "alpha_c_p3m": -0.16762149300804308, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": -0.0432190760059612, "ret_signed_p6m": -0.0432190760059612, "alpha_spy_p6m": -0.09183520615201624, "alpha_c_p6m": -0.35920394509635845, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [0.1737, 0.1471, 0.1456, 0.1264, 0.1072, 0.1471, 0.1471, 0.1087, 0.1072, 0.0954, 0.1117, 0.1338, 0.1545, 0.2032, 0.2165, 0.2549, 0.2711, 0.249, 0.1914, 0.0807, 0.1176, 0.0895, 0.1501, 0.1752, 0.2047, 0.1811, 0.1973, 0.156, 0.1461, 0.1863, 0.1475, 0.1833, 0.1848, 0.1669, 0.1744, 0.1699, 0.1297, 0.1103, 0.0969, 0.1207, 0.1088, 0.1177, 0.1207, 0.1028, 0.0164, 0.0045, -0.0238, -0.0238, 0.0328, 0.0119, -0.0209, -0.0134, -0.0134, 0.0313, 0.0268, 0.0477, 0.0477, 0.0045, -0.0835, -0.082, -0.0775, -0.0805, -0.0358, 0.0, -0.0402, -0.0611, -0.0611, -0.0864, -0.0864, -0.0313, 0.0, 0.0417, 0.0283, 0.0268, -0.0045, 0.0939, 0.0626, 0.1073, 0.0835, 0.0492, 0.0507, 0.0462, 0.076, -0.0313, -0.0104, -0.0835, -0.1088, -0.1148, -0.1207, -0.073, -0.0209, -0.0149, -0.0194, 0.0194, -0.0492, -0.0551, -0.0596, -0.07, -0.0283, 0.0045, -0.0179, -0.0402, -0.0551, -0.1013, -0.0462, -0.0551, -0.0417, -0.0387, -0.0402, -0.0522, -0.0522, -0.0417, 0.0104, 0.0387, 0.0462, 0.0462, 0.0835, 0.0879, 0.1133, 0.0969, 0.073, 0.0641, 0.082, 0.0954, 0.0313, -0.0104, -0.0045, 0.0238, 0.0119, -0.0343, -0.0238, 0.0224, 0.0492, 0.0224, 0.0045, -0.0253, -0.0686, -0.0641, -0.0104, 0.0089, -0.0328, -0.0715, -0.0209, -0.0164, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, 0.0075, -0.0075, 0.0224, 0.0373, 0.0402, 0.0402, 0.0253, 0.0, -0.07, -0.0253, -0.0134, -0.1073, -0.0879, 0.003, -0.0119, -0.0492, -0.0686, -0.0492, -0.0164, -0.0402, 0.006, -0.0492, -0.073, -0.0611, -0.0864, -0.0939, -0.1803, -0.2593, -0.1863, -0.1937, -0.1937, -0.1937, -0.1937, -0.1475, -0.1371, -0.0984, -0.079, -0.0462, -0.0432]}
{"unit_id": "orig:1980637566903939537", "ticker_idx": 0, "ts": "2025-10-21T14:09:29+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Mizuho) Broadcom raises target price to $435", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Mizuho raises AVGO PT to $435 on Anthropic/OAI AI-rack demand ramp; flags potential headwind for MRVL.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Mizuho) Broadcom raises target price to $435\n\nAVGO — As OAI has become the “fifth” customer and is expected to contribute an additional 2GW of demand in 2025, Anthropic is likely to become the “fourth” customer. We now view Anthropic as AVGO’s fourth customer, with its revenue from the “AI rack” expected to increase to $10 billion by the second half of 2026 (this could be potentially negative for MRVL/Trainium).\n\nThe 10GW OAI partnership (link) represents incremental demand: as OAI targets approximately 2GW of capacity expansion in 2026 (about double year-over-year), it will become AVGO’s fifth customer, with demand ramping from the fourth quarter of 2026 (expected). We believe OAI will increase its capacity across GPUs/ASICs to approximately 2GW/3GW/4GW in 2026/27/28 (expected). We also expect other major CSPs (cloud service providers) to expand their capacity to 2–4GW over the coming years, which remains below many widely reported GW figures.\n\nAVGO also launched the third-generation CPO (Co-Packaged Optics) Tomahawk 6-Davisson, featuring more than 70% lower power requirements and 1 million “flap-free” hours (we believe over 5 million “flap-free” hours may be required for reliability). AVGO additionally introduced the new Thor Ultra 800G NIC (Network Interface Card) with telemetry, load balancing RoCE, and open architecture, and announced the expansion of its Jericho product line to support “Scale Across.”\n\nWith Anthropic adding incremental tailwinds, we raise our estimates and target price to $435.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019202600716187623, "ret_signed_m1d": 0.019202600716187623, "alpha_spy_m1d": 0.019187604571294203, "alpha_c_m1d": 0.0136421652007499, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0068873625963382645, "ret_signed_p1d": -0.0068873625963382645, "alpha_spy_p1d": -0.0016885466334175536, "alpha_c_p1d": 0.012516004091333688, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.08845491359008517, "ret_signed_p1w": 0.08845491359008517, "alpha_spy_p1w": 0.0649627210951802, "alpha_c_p1w": 0.036876554890219904, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.034319674881862605, "ret_signed_p1m": 0.034319674881862605, "alpha_spy_p1m": 0.04722001708657386, "alpha_c_p1m": 0.05129035103153057, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.02837487160635943, "ret_signed_p3m": 0.02837487160635943, "alpha_spy_p3m": -0.005014319318705107, "alpha_c_p3m": -0.13477568250451522, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.16241421087236496, "ret_signed_p6m": 0.16241421087236496, "alpha_spy_p6m": 0.11379808072630992, "alpha_c_p6m": -0.15357065821803229, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.1589, -0.1546, -0.1426, -0.1335, -0.1184, -0.1443, -0.1591, -0.1326, -0.1466, -0.1211, -0.115, -0.1115, -0.1146, -0.0886, -0.0995, -0.0933, -0.1075, -0.1092, -0.1408, -0.1517, -0.1563, -0.1435, -0.1428, -0.1318, -0.1253, -0.1008, -0.1336, -0.1336, -0.1311, -0.119, -0.1082, -0.0243, 0.007, -0.0192, 0.0767, 0.0477, 0.0484, 0.0607, 0.0488, 0.0085, 0.0061, 0.0049, -0.0113, -0.0109, -0.0098, -0.0191, -0.0237, -0.0431, -0.0372, -0.0271, -0.0131, -0.0125, -0.0209, -0.0182, 0.0083, 0.0069, -0.0526, 0.041, 0.0043, 0.0253, 0.0335, 0.0195, 0.0192, 0.0, -0.0069, 0.0048, 0.0335, 0.0566, 0.0885, 0.1264, 0.0987, 0.0787, 0.058, 0.0271, 0.0476, 0.0377, 0.0198, 0.0459, 0.0271, 0.0367, -0.0078, -0.0006, -0.0, -0.0063, 0.0343, 0.0121, -0.0072, 0.103, 0.1237, 0.1602, 0.1602, 0.176, 0.1267, 0.1136, 0.1108, 0.112, 0.1389, 0.1705, 0.1857, 0.2052, 0.1859, 0.0504, -0.0083, -0.004, -0.0486, -0.0373, -0.0067, -0.0016, 0.0214, 0.024, 0.024, 0.0296, 0.0216, 0.0229, 0.012, 0.012, 0.0164, 0.0041, 0.0052, 0.0044, -0.0279, 0.0087, 0.0298, 0.0369, -0.0062, 0.003, 0.0284, 0.0284, -0.0275, -0.0386, -0.0483, -0.0642, -0.0502, -0.0269, -0.0256, -0.033, -0.0313, -0.0319, -0.0634, -0.0993, -0.0921, -0.0266, 0.0057, -0.0046, 0.0022, -0.0317, -0.0492, -0.0492, -0.0277, -0.0248, -0.0234, -0.0274, -0.0341, -0.0483, -0.0283, -0.0594, -0.0657, -0.0678, -0.0824, -0.0716, -0.027, -0.0337, 0.0109, 0.0017, -0.0013, -0.0176, -0.058, -0.05, -0.0605, -0.0762, -0.0648, -0.0921, -0.055, -0.0674, -0.0659, -0.0934, -0.119, -0.1403, -0.0931, -0.0815, -0.0783, -0.0783, -0.0787, -0.0214, 0.0274, 0.0399, 0.0887, 0.1127, 0.1157, 0.1624]}
{"unit_id": "orig:1980637566903939537", "ticker_idx": 1, "ts": "2025-10-21T14:09:29+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this could be potentially negative for MRVL/Trainium", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Mizuho raises AVGO PT to $435 on Anthropic/OAI AI-rack demand ramp; flags potential headwind for MRVL.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Mizuho) Broadcom raises target price to $435\n\nAVGO — As OAI has become the “fifth” customer and is expected to contribute an additional 2GW of demand in 2025, Anthropic is likely to become the “fourth” customer. We now view Anthropic as AVGO’s fourth customer, with its revenue from the “AI rack” expected to increase to $10 billion by the second half of 2026 (this could be potentially negative for MRVL/Trainium).\n\nThe 10GW OAI partnership (link) represents incremental demand: as OAI targets approximately 2GW of capacity expansion in 2026 (about double year-over-year), it will become AVGO’s fifth customer, with demand ramping from the fourth quarter of 2026 (expected). We believe OAI will increase its capacity across GPUs/ASICs to approximately 2GW/3GW/4GW in 2026/27/28 (expected). We also expect other major CSPs (cloud service providers) to expand their capacity to 2–4GW over the coming years, which remains below many widely reported GW figures.\n\nAVGO also launched the third-generation CPO (Co-Packaged Optics) Tomahawk 6-Davisson, featuring more than 70% lower power requirements and 1 million “flap-free” hours (we believe over 5 million “flap-free” hours may be required for reliability). AVGO additionally introduced the new Thor Ultra 800G NIC (Network Interface Card) with telemetry, load balancing RoCE, and open architecture, and announced the expansion of its Jericho product line to support “Scale Across.”\n\nWith Anthropic adding incremental tailwinds, we raise our estimates and target price to $435.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018751455034801134, "ret_signed_m1d": -0.018751455034801134, "alpha_spy_m1d": -0.018736458889907714, "alpha_c_m1d": -0.013191019519363412, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.038096256036807885, "ret_signed_p1d": 0.038096256036807885, "alpha_spy_p1d": 0.032897440073887174, "alpha_c_p1d": 0.018692889349135933, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.04984568472033968, "ret_signed_p1w": -0.04984568472033968, "alpha_spy_p1w": -0.026353492225434705, "alpha_c_p1w": 0.0017326739795255897, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": -0.034892006501149875, "ret_signed_p1m": 0.034892006501149875, "alpha_spy_p1m": 0.021991664296438618, "alpha_c_p1m": 0.017921330351481912, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": -0.0444113790430023, "ret_signed_p3m": 0.0444113790430023, "alpha_spy_p3m": 0.07780056996806684, "alpha_c_p3m": 0.20756193315387694, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.5993862200861162, "ret_signed_p6m": -0.5993862200861162, "alpha_spy_p6m": -0.5507700899400612, "alpha_c_p6m": -0.28340135099571895, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.1219, -0.1199, -0.0997, -0.0946, -0.0305, -0.0468, -0.117, -0.0923, -0.0912, -0.1067, -0.1004, -0.0827, -0.0834, -0.0772, -0.0593, -0.0626, -0.0964, -0.0899, -0.1452, -0.1553, -0.1554, -0.1342, -0.1348, -0.1193, -0.113, -0.084, -0.2544, -0.2544, -0.2338, -0.261, -0.2398, -0.2489, -0.2172, -0.2073, -0.2042, -0.2102, -0.2012, -0.2003, -0.1833, -0.1582, -0.1196, -0.1193, -0.1042, -0.115, -0.0501, -0.006, -0.0136, -0.0228, -0.0029, -0.005, 0.0223, 0.0226, 0.0546, 0.0315, 0.0971, 0.0755, 0.016, 0.0609, 0.0233, 0.0549, 0.0471, 0.0438, 0.0188, 0.0, -0.0381, -0.0177, -0.0015, 0.0528, 0.0498, 0.0699, 0.0512, 0.1125, 0.0725, 0.0395, 0.1025, 0.1076, 0.079, 0.1065, 0.0602, 0.0602, 0.0387, 0.026, -0.0096, -0.0662, -0.0349, -0.09, -0.0808, -0.0056, -0.0099, 0.0411, 0.0411, 0.061, 0.0812, 0.1024, 0.1892, 0.1653, 0.1739, 0.0919, 0.0551, 0.0974, 0.0614, 0.002, 0.0, -0.0023, -0.0304, 0.0025, -0.002, 0.0064, 0.0406, 0.0265, 0.0265, 0.0247, 0.0178, 0.0297, 0.0085, 0.0085, 0.0609, 0.0709, 0.0471, 0.0045, -0.0096, -0.0116, -0.0156, -0.0137, -0.0355, -0.0454, -0.0444, -0.0444, -0.0522, -0.0196, -0.0131, -0.0471, -0.0289, -0.0151, -0.0069, -0.034, -0.0627, -0.0658, -0.1028, -0.1243, -0.1186, -0.0465, -0.022, -0.026, -0.034, -0.0709, -0.0664, -0.0664, -0.0616, -0.0607, -0.0545, -0.0561, -0.0761, -0.069, -0.0389, -0.0583, -0.0298, -0.0397, -0.0794, -0.0726, -0.1012, 0.0638, 0.1004, 0.1081, 0.0741, 0.0412, 0.0435, 0.0877, 0.0783, 0.0406, 0.0633, 0.0441, 0.0708, 0.0969, 0.1692, 0.1601, 0.1268, 0.0429, 0.1764, 0.2673, 0.2721, 0.2721, 0.3006, 0.2991, 0.3593, 0.4244, 0.5268, 0.5602, 0.5902, 0.5994]}
{"unit_id": "orig:1987660455909953589", "ticker_idx": 0, "ts": "2025-11-09T23:15:56+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "The latest outlook is higher than our current estimates for SEC/SKH Legacy DRAM price increases [SEC: +22% / SKH: +20%]", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman relays TrendForce's sharply upgraded 4Q25 DRAM/NAND price outlook, beating GS estimates for Samsung and SK Hynix; bullish memory cycle into 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "251110_South Korea Technology: Oct. 2025 memory pricing - Goldman Sachs\n\n(1) TrendForce again significantly raised its outlook for 4Q25 DRAM price increases.\n\n(2) This comes on top of its upgrade in mid/late October, when it had already raised the 4Q25 Legacy DRAM price increase outlook from QoQ +8–13% to +18–23%.\n\n(3) The latest outlook is higher than our current estimates for SEC/SKH Legacy DRAM price increases [SEC: +22% / SKH: +20%].\n\n(4) TrendForce noted that server customers are now requesting allocations out to ’27.\n\n(5) It also assessed that strong server customer demand is increasing supply uncertainty for non-server applications, thereby strengthening memory vendors’ bargaining power.\n\n(6) PC DRAM price outlook raised: 4Q25 QoQ +25–30%\n■ In October, D4 8GB rose MoM +17% to $31.5, and D5 8GB rose MoM +25% to $33.5.\n■ D5 is contracting at a +6% premium to D4 [vs. a 1% discount in September].\n■ As supply tightens further in 4Q25, TrendForce raised its PC DRAM price increase outlook from +18–23% to +25–30%.\n\n(7) Server DRAM price outlook raised: 4Q25 QoQ +28–33%\n■ In October, D4 64GB rose MoM +28% to $300, and D5 64GB rose MoM +24% to $338.\n■ The D5 premium fell by 3pp MoM to about 13%.\n■ U.S. hyperscalers are now discussing allocations not only for ’25 and ’26 but out to ’27, providing memory suppliers with medium- to long-term visibility.\n\n(8) LPDDR [handset] price outlook sharply raised: 4Q25 +38–43%\n■ For both LPD4/5, on intensifying tightness, the price increase outlooks were raised from +18–23% and +20–25% to +38–43%.\n■ The situation is so tight that some vendors are temporarily refraining from offering quotes at all.\n\n(9) NAND: Price upcycle expected to continue throughout ’26\n■ TrendForce expects the NAND price upcycle to persist throughout 2026.\n■ 1Q26 +3–8% / 2Q26 +5–10% / 3Q26 +0–5% / 4Q26 +0–5%\n■ Strong server demand and tight supply offset concerns over weak consumer demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0268389099250409, "ret_signed_m1d": -0.0268389099250409, "alpha_spy_m1d": -0.01147423224448918, "alpha_c_m1d": -0.0019006588509518085, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02493825107408909, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02882701787404751, "ret_signed_p1d": 0.02882701787404751, "alpha_spy_p1d": 0.026537868171601753, "alpha_c_p1d": 0.03735397683432329, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.008526958960275777, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": 0.02314227390131529, "alpha_c_p1w": 0.04023276598428149, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.04023276598428149, "ret_p1m": 0.0775348024458482, "ret_signed_p1m": 0.0775348024458482, "alpha_spy_p1m": 0.07518701202974065, "alpha_c_p1m": 0.0758089493486993, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.0017258530971488995, "ret_p3m": 0.5911965541547499, "ret_signed_p3m": 0.5911965541547499, "alpha_spy_p3m": 0.5938640310421116, "alpha_c_p3m": 0.6719278967011242, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": -0.08073134254637426, "ret_p6m": 1.3271748175912519, "ret_signed_p6m": 1.3271748175912519, "alpha_spy_p6m": 1.2590087261920881, "alpha_c_p6m": 1.2031773412902789, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.12399747630097302, "price_path": [-0.2885, -0.2914, -0.2914, -0.3073, -0.3073, -0.3023, -0.3013, -0.2934, -0.2924, -0.3043, -0.3013, -0.3112, -0.3102, -0.331, -0.3162, -0.3092, -0.3063, -0.3122, -0.3063, -0.2924, -0.2815, -0.2736, -0.2538, -0.2429, -0.2142, -0.2261, -0.2053, -0.2053, -0.1737, -0.1618, -0.1549, -0.1479, -0.1756, -0.163, -0.166, -0.1451, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.0616, -0.0726, -0.0895, -0.0557, -0.0288, -0.0268, -0.0249, -0.0308, -0.0199, -0.0408, -0.0179, 0.0139, -0.0109, -0.001, 0.0348, 0.0686, 0.1044, 0.0427, 0.0, -0.0139, -0.0268, 0.0, 0.0288, 0.0249, 0.0219, -0.0338, 0.0, -0.0278, -0.0408, 0.0, -0.0577, -0.0388, -0.0129, 0.0219, 0.0288, -0.001, 0.002, 0.0278, 0.0388, 0.0447, 0.0775, 0.0885, 0.0775, 0.0736, 0.0666, 0.0825, 0.0417, 0.0219, 0.0726, 0.0696, 0.0567, 0.0984, 0.1083, 0.1044, 0.1044, 0.163, 0.1936, 0.1976, 0.1976, 0.1976, 0.2835, 0.3794, 0.3874, 0.4084, 0.3864, 0.3884, 0.3864, 0.3744, 0.4014, 0.4374, 0.4873, 0.4913, 0.4504, 0.4933, 0.5213, 0.5193, 0.5193, 0.5932, 0.6222, 0.6052, 0.6032, 0.5023, 0.6731, 0.6891, 0.5912, 0.5842, 0.6621, 0.6561, 0.6761, 0.784, 0.8099, 0.8099, 0.8099, 0.8099, 0.8978, 0.8988, 0.9278, 0.9977, 1.0327, 1.1775, 1.1625, 1.1625, 0.9488, 0.7201, 0.9138, 0.8799, 0.733, 0.8769, 0.8978, 0.8769, 0.8329, 0.8849, 0.9368, 1.0826, 1.0027, 0.9917, 0.8609, 0.8949, 0.8879, 0.799, 0.799, 0.7646, 0.6736, 0.8983, 0.7857, 0.8637, 0.9328, 0.9668, 1.107, 1.0419, 1.0619, 1.0119, 1.0669, 1.112, 1.177, 1.162, 1.147, 1.192, 1.177, 1.2471, 1.1971, 1.2471, 1.2221, 1.2621, 1.2071, 1.2071, 1.3272, 1.3272]}
{"unit_id": "orig:1987660455909953589", "ticker_idx": 1, "ts": "2025-11-09T23:15:56+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "The latest outlook is higher than our current estimates for SEC/SKH Legacy DRAM price increases [SEC: +22% / SKH: +20%]", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman relays TrendForce's sharply upgraded 4Q25 DRAM/NAND price outlook, beating GS estimates for Samsung and SK Hynix; bullish memory cycle into 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251110_South Korea Technology: Oct. 2025 memory pricing - Goldman Sachs\n\n(1) TrendForce again significantly raised its outlook for 4Q25 DRAM price increases.\n\n(2) This comes on top of its upgrade in mid/late October, when it had already raised the 4Q25 Legacy DRAM price increase outlook from QoQ +8–13% to +18–23%.\n\n(3) The latest outlook is higher than our current estimates for SEC/SKH Legacy DRAM price increases [SEC: +22% / SKH: +20%].\n\n(4) TrendForce noted that server customers are now requesting allocations out to ’27.\n\n(5) It also assessed that strong server customer demand is increasing supply uncertainty for non-server applications, thereby strengthening memory vendors’ bargaining power.\n\n(6) PC DRAM price outlook raised: 4Q25 QoQ +25–30%\n■ In October, D4 8GB rose MoM +17% to $31.5, and D5 8GB rose MoM +25% to $33.5.\n■ D5 is contracting at a +6% premium to D4 [vs. a 1% discount in September].\n■ As supply tightens further in 4Q25, TrendForce raised its PC DRAM price increase outlook from +18–23% to +25–30%.\n\n(7) Server DRAM price outlook raised: 4Q25 QoQ +28–33%\n■ In October, D4 64GB rose MoM +28% to $300, and D5 64GB rose MoM +24% to $338.\n■ The D5 premium fell by 3pp MoM to about 13%.\n■ U.S. hyperscalers are now discussing allocations not only for ’25 and ’26 but out to ’27, providing memory suppliers with medium- to long-term visibility.\n\n(8) LPDDR [handset] price outlook sharply raised: 4Q25 +38–43%\n■ For both LPD4/5, on intensifying tightness, the price increase outlooks were raised from +18–23% and +20–25% to +38–43%.\n■ The situation is so tight that some vendors are temporarily refraining from offering quotes at all.\n\n(9) NAND: Price upcycle expected to continue throughout ’26\n■ TrendForce expects the NAND price upcycle to persist throughout 2026.\n■ 1Q26 +3–8% / 2Q26 +5–10% / 3Q26 +0–5% / 4Q26 +0–5%\n■ Strong server demand and tight supply offset concerns over weak consumer demand.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04290426551341642, "ret_signed_m1d": -0.04290426551341642, "alpha_spy_m1d": -0.027539587832864698, "alpha_c_m1d": -0.013165422975958907, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02145208104825036, "ret_signed_p1d": 0.02145208104825036, "alpha_spy_p1d": 0.0191629313458046, "alpha_c_p1d": 0.043024622920250555, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": 0.02314227390131529, "alpha_c_p1w": 0.05153441506328904, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": -0.06533766967354793, "ret_signed_p1m": -0.06533766967354793, "alpha_spy_p1m": -0.06768546008965548, "alpha_c_p1m": -0.09379438457706368, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": 0.3904341214598579, "ret_signed_p3m": 0.3904341214598579, "alpha_spy_p3m": 0.3931015983472196, "alpha_c_p3m": 0.32503170470856646, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": 1.3939082059047334, "ret_signed_p6m": 1.3939082059047334, "alpha_spy_p6m": 1.3257421145055697, "alpha_c_p6m": 0.9325620461188837, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.5419, -0.5444, -0.5444, -0.5592, -0.5666, -0.579, -0.5963, -0.5864, -0.5724, -0.5691, -0.5716, -0.5569, -0.5561, -0.5776, -0.5701, -0.5668, -0.5619, -0.5487, -0.5429, -0.5248, -0.4983, -0.4934, -0.4579, -0.4538, -0.4257, -0.4497, -0.4134, -0.4134, -0.4208, -0.4043, -0.4101, -0.4117, -0.4447, -0.4241, -0.4266, -0.4059, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.2937, -0.3152, -0.321, -0.3028, -0.2533, -0.2318, -0.1988, -0.2096, -0.2054, -0.2104, -0.1584, -0.1172, -0.1403, -0.0792, -0.0627, -0.0776, 0.0231, -0.033, -0.0446, -0.0215, -0.0429, 0.0, 0.0215, 0.0182, 0.0099, -0.0759, 0.0, -0.0594, -0.0726, -0.0578, -0.1403, -0.1419, -0.1436, -0.1353, -0.1017, -0.1248, -0.1116, -0.0785, -0.0885, -0.105, -0.1017, -0.0472, -0.0653, -0.0307, -0.067, -0.0571, -0.0852, -0.1248, -0.0901, -0.0885, -0.0967, -0.0422, -0.0356, -0.029, -0.029, -0.0108, 0.0569, 0.075, 0.075, 0.075, 0.118, 0.1493, 0.1989, 0.2253, 0.2484, 0.2286, 0.2369, 0.2187, 0.2253, 0.2369, 0.2484, 0.2616, 0.227, 0.222, 0.2468, 0.2666, 0.2154, 0.3211, 0.3888, 0.4218, 0.5011, 0.3706, 0.4978, 0.4862, 0.3904, 0.3855, 0.4647, 0.4466, 0.4202, 0.4664, 0.4532, 0.4532, 0.4532, 0.4532, 0.4763, 0.5671, 0.5704, 0.6596, 0.6811, 0.8182, 0.7553, 0.7553, 0.5535, 0.4046, 0.5568, 0.5287, 0.3831, 0.5518, 0.5799, 0.5386, 0.5055, 0.6114, 0.6048, 0.747, 0.6759, 0.666, 0.5435, 0.6312, 0.6461, 0.5435, 0.5435, 0.4443, 0.3351, 0.484, 0.3731, 0.4492, 0.4658, 0.5154, 0.709, 0.6511, 0.6991, 0.7206, 0.8248, 0.8794, 0.9108, 0.8662, 0.929, 1.025, 1.0233, 1.0266, 1.0217, 1.1375, 1.1507, 1.1391, 1.1276, 1.1276, 1.3939, 1.3939]}
{"unit_id": "orig:1978054480487993424", "ticker_idx": 0, "ts": "2025-10-14T11:05:13+00:00", "symbol": "ANET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Derivative concerns about ANET (Arista Networks) are excessive — we maintain that Arista remains well positioned", "tweet_type": "news_reaction", "timeline": "months", "summary": "Wells Fargo maintains ANET as well-positioned despite Spectrum-X fears; NVDA networking momentum continues with Meta/Oracle wins.", "resolved_tickers": ["ANET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Wells Fargo) NVIDIA: Meta and Oracle to Use NVIDIA Spectrum-X Ethernet; Concerns About ANET Are Overblown\n\nOur View\n\nThis morning, Meta and Oracle announced they will each use NVIDIA’s Spectrum-X Ethernet for the Facebook Open Switching System (FBOSS) OS and for giga-scale AI factories, respectively.\n\nDerivative concerns about ANET (Arista Networks) are excessive — we maintain that Arista remains well positioned.\n\nInitial Analysis\n\nWhat’s new:\n\nThis morning, NVIDIA announced that Meta will integrate NVIDIA’s Spectrum-X switches with the FBOSS (Facebook Open Switching System) on its next-generation Minipack3N switch. We view this as a natural extension of Meta’s Minipack line.\n\nFor reference, the Minipack3 introduced last year transitioned to Broadcom’s Tomahawk 5 (51.2T) silicon, integrated with Cisco’s G200-based 8501 switch and Celestica-manufactured hardware.\nNVIDIA also announced that Oracle will adopt its Spectrum-X Ethernet solution to build giga-scale AI factories, aimed at interconnecting millions of GPUs into a single massive computer for next-generation frontier AI model training.\nAccording to NVIDIA’s press release, this deployment is aligned with Oracle Cloud Infrastructure’s (OCI) next-generation Vera-Rubin architecture, targeting rollout in 2026.\n\nNetworking Context\nFor reference, NVIDIA’s F2Q26 networking revenue reached $7.252 billion (+98% YoY, +46% QoQ).\n\nSpectrum-X Ethernet revenue alone exceeded an $8 billion annualized run rate in Q1 and surpassed $10 billion by the latest quarter.\n\nAt Hot Chips 2025 (Aug 26), NVIDIA highlighted its new Spectrum-XGS Ethernet designed for giga-scale AI superfactories with 1,000 km+ reach, focusing on inter-data center connectivity.\n\nThis architecture emphasizes AI-optimized Ethernet over generic off-the-shelf (OTS) Ethernet, claiming a 1.9× NCCL performance advantage in multi-data-center communication.\n\nANET Analysis – Concerns Are Overblown\n\nWhile this announcement may weigh on ANET’s share price and reinforces the view that Spectrum-X momentum is a key competitive dynamic, we believe these fears are overstated.\nWe continue to see Arista Networks as well positioned in Oracle’s (OCI) and OpenAI’s Stargate data-center build-outs.\n\nWe also believe Arista maintains a solid position within Meta’s data-center leaf-spine architecture, particularly in distributed Ethernet switch (DES) networks supporting AI clusters.\n\nArista also holds strong potential to play a core role in Meta’s future scale-up architectures.\n\nArista and Broadcom are scheduled to co-host a session at the Open Compute Project (OCP) titled “Scale-Up Ethernet (SUE) Framework for AI/ML Accelerators” on Wednesday at 10:05 a.m. PT.\n\nWe see Arista’s participation in scale-up networking as a potential additional growth driver beyond 2027.\nAt its recent Analyst Day, Arista management mentioned that it is already involved in “many projects” within the scale-up segment.\n\n$ANET", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06239645565985219, "ret_signed_m1d": 0.06239645565985219, "alpha_spy_m1d": 0.06117330659045961, "alpha_c_m1d": 0.04939520075611803, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01300125490373416, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03307163209682429, "ret_signed_p1d": 0.03307163209682429, "alpha_spy_p1d": 0.028632016230144997, "alpha_c_p1d": 0.023160892886344042, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.009910739210480246, "ret_p1w": 0.05151674833309339, "ret_signed_p1w": 0.05151674833309339, "alpha_spy_p1w": 0.03783575740067313, "alpha_c_p1w": 0.02725484131424194, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.024261907018851447, "ret_p1m": -0.027451529237694894, "ret_signed_p1m": -0.027451529237694894, "alpha_spy_p1m": -0.05938910451428914, "alpha_c_p1m": -0.07153524079594087, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.044083711558245975, "ret_p3m": -0.11456152940150777, "ret_signed_p3m": -0.11456152940150777, "alpha_spy_p3m": -0.16573849059720558, "alpha_c_p3m": -0.1544653809906612, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.03990385158915344, "ret_p6m": 0.04524832006540214, "ret_signed_p6m": 0.04524832006540214, "alpha_spy_p6m": 0.018627309226684874, "alpha_c_p6m": 0.03582523661995651, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.009423083445445624, "price_path": [-0.1932, -0.1946, -0.1958, -0.209, -0.1855, -0.1783, -0.1766, -0.153, -0.1453, -0.1203, -0.1122, -0.1529, -0.1329, -0.1489, -0.0001, 0.0035, 0.0028, -0.0082, 0.0177, -0.0056, -0.0166, -0.0107, -0.0054, -0.0433, -0.0527, -0.0487, -0.0399, -0.0414, -0.0326, -0.0398, -0.0184, -0.0161, -0.0161, -0.021, -0.0102, 0.0171, 0.0293, 0.0088, 0.0225, 0.086, 0.1027, 0.0043, 0.0478, 0.0243, 0.0292, 0.0567, 0.078, 0.0476, 0.0382, 0.0277, 0.0308, 0.0267, 0.033, 0.0499, 0.0755, 0.0409, 0.0483, 0.0772, 0.0468, 0.1338, 0.1401, 0.1103, 0.0624, 0.0, 0.0331, 0.052, 0.0311, 0.0554, 0.0515, 0.0562, 0.1007, 0.1083, 0.1298, 0.1295, 0.1674, 0.1416, 0.1362, 0.1355, 0.1063, 0.0117, -0.0344, -0.0298, -0.011, -0.0278, -0.0275, -0.0612, -0.0535, -0.0831, -0.1105, -0.1007, -0.1383, -0.1539, -0.1197, -0.0991, -0.0803, -0.0803, -0.0584, -0.077, -0.0834, -0.0792, -0.0738, -0.0735, -0.0697, -0.063, -0.0463, -0.0317, -0.1011, -0.0929, -0.0912, -0.1184, -0.1021, -0.0553, -0.0581, -0.0538, -0.0578, -0.0578, -0.0501, -0.0334, -0.0458, -0.0559, -0.0559, -0.0374, -0.0115, -0.0447, -0.0628, -0.1086, -0.1146, -0.1107, -0.0638, -0.0987, -0.0591, -0.0646, -0.0646, -0.0812, -0.0829, -0.0027, -0.0177, 0.0355, 0.0569, 0.0819, 0.0674, 0.0213, -0.003, 0.0043, -0.0613, -0.0729, -0.0094, 0.0213, 0.0336, 0.0135, -0.0264, 0.0202, 0.0202, 0.0273, 0.0054, -0.0112, -0.0432, -0.0819, -0.0722, -0.0425, -0.0615, -0.0381, -0.0684, -0.1022, -0.0285, 0.0044, -0.0425, -0.0117, 0.006, -0.004, -0.0343, -0.0376, -0.0248, -0.0412, -0.0196, -0.0182, -0.0545, -0.021, -0.0576, -0.0272, -0.117, -0.1298, -0.1633, -0.1154, -0.1004, -0.0873, -0.0873, -0.0904, -0.0371, 0.0452]}
{"unit_id": "orig:1978054480487993424", "ticker_idx": 1, "ts": "2025-10-14T11:05:13+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Spectrum-X Ethernet revenue alone exceeded an $8 billion annualized run rate in Q1 and surpassed $10 billion by the latest quarter", "tweet_type": "news_reaction", "timeline": "months", "summary": "Wells Fargo maintains ANET as well-positioned despite Spectrum-X fears; NVDA networking momentum continues with Meta/Oracle wins.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Wells Fargo) NVIDIA: Meta and Oracle to Use NVIDIA Spectrum-X Ethernet; Concerns About ANET Are Overblown\n\nOur View\n\nThis morning, Meta and Oracle announced they will each use NVIDIA’s Spectrum-X Ethernet for the Facebook Open Switching System (FBOSS) OS and for giga-scale AI factories, respectively.\n\nDerivative concerns about ANET (Arista Networks) are excessive — we maintain that Arista remains well positioned.\n\nInitial Analysis\n\nWhat’s new:\n\nThis morning, NVIDIA announced that Meta will integrate NVIDIA’s Spectrum-X switches with the FBOSS (Facebook Open Switching System) on its next-generation Minipack3N switch. We view this as a natural extension of Meta’s Minipack line.\n\nFor reference, the Minipack3 introduced last year transitioned to Broadcom’s Tomahawk 5 (51.2T) silicon, integrated with Cisco’s G200-based 8501 switch and Celestica-manufactured hardware.\nNVIDIA also announced that Oracle will adopt its Spectrum-X Ethernet solution to build giga-scale AI factories, aimed at interconnecting millions of GPUs into a single massive computer for next-generation frontier AI model training.\nAccording to NVIDIA’s press release, this deployment is aligned with Oracle Cloud Infrastructure’s (OCI) next-generation Vera-Rubin architecture, targeting rollout in 2026.\n\nNetworking Context\nFor reference, NVIDIA’s F2Q26 networking revenue reached $7.252 billion (+98% YoY, +46% QoQ).\n\nSpectrum-X Ethernet revenue alone exceeded an $8 billion annualized run rate in Q1 and surpassed $10 billion by the latest quarter.\n\nAt Hot Chips 2025 (Aug 26), NVIDIA highlighted its new Spectrum-XGS Ethernet designed for giga-scale AI superfactories with 1,000 km+ reach, focusing on inter-data center connectivity.\n\nThis architecture emphasizes AI-optimized Ethernet over generic off-the-shelf (OTS) Ethernet, claiming a 1.9× NCCL performance advantage in multi-data-center communication.\n\nANET Analysis – Concerns Are Overblown\n\nWhile this announcement may weigh on ANET’s share price and reinforces the view that Spectrum-X momentum is a key competitive dynamic, we believe these fears are overstated.\nWe continue to see Arista Networks as well positioned in Oracle’s (OCI) and OpenAI’s Stargate data-center build-outs.\n\nWe also believe Arista maintains a solid position within Meta’s data-center leaf-spine architecture, particularly in distributed Ethernet switch (DES) networks supporting AI clusters.\n\nArista also holds strong potential to play a core role in Meta’s future scale-up architectures.\n\nArista and Broadcom are scheduled to co-host a session at the Open Compute Project (OCP) titled “Scale-Up Ethernet (SUE) Framework for AI/ML Accelerators” on Wednesday at 10:05 a.m. PT.\n\nWe see Arista’s participation in scale-up networking as a potential additional growth driver beyond 2027.\nAt its recent Analyst Day, Arista management mentioned that it is already involved in “many projects” within the scale-up segment.\n\n$ANET", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.046047986166150245, "ret_signed_m1d": 0.046047986166150245, "alpha_spy_m1d": 0.044824837096757664, "alpha_c_m1d": 0.027307423501326866, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0011108615216387507, "ret_signed_p1d": -0.0011108615216387507, "alpha_spy_p1d": -0.0055504773883180425, "alpha_c_p1d": -0.025938407846619937, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.0062767702368793366, "ret_signed_p1w": 0.0062767702368793366, "alpha_spy_p1w": -0.0074042206955409196, "alpha_c_p1w": -0.029105445853656597, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.07648737195000477, "ret_signed_p1m": 0.07648737195000477, "alpha_spy_p1m": 0.044549796673410524, "alpha_c_p1m": 0.010430477612360045, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.02688606994062126, "ret_signed_p3m": 0.02688606994062126, "alpha_spy_p3m": -0.024290891255076552, "alpha_c_p3m": -0.14382181854713783, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 0.01149811607040685, "ret_signed_p6m": 0.01149811607040685, "alpha_spy_p6m": -0.015122894768310413, "alpha_c_p6m": -0.2605737054351056, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.0391, -0.0424, -0.0481, -0.0723, -0.0514, -0.035, -0.0363, -0.0183, -0.0252, -0.0043, -0.0121, -0.0351, -0.0002, -0.0099, -0.0034, 0.0041, 0.0148, 0.0112, 0.0173, 0.0086, 0.011, 0.0023, 0.0109, -0.0244, -0.0258, -0.0281, -0.0114, -0.0013, 0.0096, 0.0087, 0.0007, -0.0325, -0.0325, -0.0514, -0.0523, -0.0465, -0.0723, -0.0652, -0.0515, -0.0151, -0.0159, -0.0123, -0.0127, -0.0286, -0.0541, -0.0211, -0.0187, 0.0199, -0.0089, -0.017, -0.013, -0.0102, 0.0101, 0.0364, 0.04, 0.0492, 0.0422, 0.0306, 0.0278, 0.0504, 0.0697, 0.0174, 0.046, 0.0, -0.0011, 0.0099, 0.0177, 0.0145, 0.0063, 0.0014, 0.0118, 0.0346, 0.0637, 0.1166, 0.15, 0.127, 0.1248, 0.1491, 0.1036, 0.0843, 0.0447, 0.0451, 0.1056, 0.0729, 0.0765, 0.0379, 0.0563, 0.0365, 0.0074, 0.036, 0.0034, -0.0064, 0.014, -0.0123, 0.0013, 0.0013, -0.0168, -0.0006, 0.0079, -0.0024, 0.0187, 0.0133, 0.0307, 0.0275, 0.0209, 0.0051, -0.0278, -0.0207, -0.0128, -0.0504, -0.0327, 0.0054, 0.0204, 0.0511, 0.0477, 0.0477, 0.0584, 0.0456, 0.0418, 0.036, 0.036, 0.0491, 0.045, 0.0401, 0.0505, 0.0279, 0.0269, 0.0273, 0.0322, 0.0173, 0.0391, 0.0345, 0.0345, -0.0108, 0.0183, 0.0268, 0.0425, 0.0358, 0.0472, 0.0639, 0.0694, 0.0617, 0.0311, 0.0018, -0.0324, -0.0452, 0.0299, 0.0557, 0.0473, 0.0557, 0.0384, 0.0155, 0.0155, 0.0275, 0.0442, 0.0438, 0.0544, 0.064, 0.0713, 0.0863, 0.0271, -0.0157, 0.0137, 0.0002, 0.0168, 0.0184, -0.0122, 0.0146, 0.0264, 0.0334, 0.0174, 0.0013, 0.0178, 0.0107, 0.0022, -0.0081, -0.0406, -0.0243, -0.0267, -0.0074, -0.0487, -0.0694, -0.0824, -0.0312, -0.0237, -0.0146, -0.0146, -0.0132, -0.0106, 0.0115]}
{"unit_id": "orig:1990013494243299340", "ticker_idx": 0, "ts": "2025-11-16T11:06:04+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our positive investment opinion on the front-end equipment sector following this investment announcement", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz maintains positive view on Korean front-end equipment sector on Samsung P5 investment; top picks Wonik IPS and TES, also positive on Eugene Tech, GST, PSK, EOTechnics.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities IT Kim Sun-woo / Kim Dong-kwan]\n\nHello. This is Kim Sun-woo / Kim Dong-kwan from Meritz. We are sharing our firm's comments on the media reports regarding Samsung Electronics' P5 investment.\n\n▶ Implications of Samsung Electronics' P5 Investment Announcement\n\n[Samsung Electronics' P5 investment announcement. Assessed as a long-term investment plan aimed at securing mid-to-long-term competitiveness]\n\nAccording to media reports, Samsung Electronics held a temporary management committee meeting and announced plans to proceed with the framework construction for Pyeongtaek P5.\n\nThe company currently operates a total of 4 production fabs up to P4 at its Pyeongtaek campus. In the case of P4, where investment recently began, around 60k of DRAM Capa is expected to be added between 4Q25 and 2Q26.\n\nThe company had originally planned to start the P4 investment in early 2025, but it has shifted direction to execute the investment at a time when visibility for DRAM price increases is secured. This investment announcement is unrelated to oversupply and is assessed as a sound investment execution plan aimed at securing mid-to-long-term competitiveness.\n\nThe company's P5 investment has the following two implications:\n\n1. Production start at the earliest 1H28, unrelated to short-term supply increase: P5 was prepared with the intention of securing cleanrooms in advance during a downcycle, following Samsung Electronics' past 'Shell-First' strategy, but it had been delayed until recently due to the HBM-focused market development in 2024-2025.\n\nNow, Samsung Electronics has decided to partially pull forward the P5 investment to lead the market, based on establishing price negotiation power for some commodity products and securing business visibility. It is assessed that this was partially influenced by the disclosure of the operation schedule for competitor SK hynix's Y1 fab.\n\nHowever, Y1 also begins operation in 1H27 at the earliest, and P5 will only start operations in 2028, so concerns about a short-term increase in memory supply are non-existent.\n\nInvestors should note that, unlike in the past, memory makers' production expansion efforts are focused on a 'mid-to-long-term' perspective. The immediate front-end equipment investment that used to occur during past upcycles is proceeding only in a limited manner, and an unusual cycle is continuing where all producers are focusing on profitability rather than current volume.\n\n2. 3-story structure for synergy creation upon foundry recovery: P5 is presumed to be the latest fab with a wider site area and a 3-story structure (vs. the existing 2-story structure), enabling greater vertical cleanroom space utilization.\n\nAssuming the recovery of foundry competitiveness, Samsung Electronics will be able to configure efficient production lines through sequential investments in a total of Phase 6 starting from 2028.\n\nIt is judged that it will be possible to reduce production time and secure cost competitiveness by configuring continuous and sequential production lines for foundry processes (ASIC, etc.) and memory processes in the future. However, expanding foundry orders and securing process competitiveness after SF2P are essential prerequisites.\n\n[Impact on Value Chain and Investment Opinion – Maintain Positive View]\n\nWe maintain our positive investment opinion on the front-end equipment sector following this investment announcement. Considering the expected mass production timing (2028E), new orders for P5 from front-end equipment makers are expected to occur starting from 2H27E.\n\nThe impact on 2025-26 earnings is limited, but this period is already a time when earnings growth is expected due to new P4 DRAM investment and conversion investments in existing DRAM/NAND lines.\n\nTypically, the stock prices of the front-end equipment sector move in tandem with expectations for the memory industry outlook and front-end customer CAPEX. The stock price rise since September is the result of expectations for 2026E CAPEX expansion being reflected as a multiple-centric expansion.\n\nThis investment announcement is positive in that it secures visibility for the long-term profit growth of equipment makers. 2027 is when SK hynix's Yongin Cluster construction and equipment investment will be underway. In addition, the P5 equipment order momentum in 2H27 and equipment revenue recognition in 2028E are expected to unfold sequentially.\n\nAs profit growth beyond 2026 and further acceleration of investment unfold, it is expected to have a positive impact on equipment sector stock prices.\n\nWe suggest Wonik IPS and TES as beneficiaries within our coverage. As of 1H25 cumulative, the sales exposure to Samsung Electronics is estimated at 58% for Wonik IPS and 35% for TES. Wonik IPS can benefit from investments across all of Samsung Electronics' product lines—DRAM, NAND, and Foundry—through its general-purpose CVD/ALD equipment.\n\nIn addition to the companies mentioned above, it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics.\n\nThe Samsung Electronics exposure we estimate for major front-end equipment/facility companies is as follows: Wonik IPS (58%), TES (35%), Eugene Technology (40%), GST (50%), UNISEM (50%), PSK (25%), EOTechnics (25%).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03379720954848586, "ret_signed_m1d": -0.03379720954848586, "alpha_spy_m1d": -0.04320134735436654, "alpha_c_m1d": -0.049697687662226486, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.015900478113740624, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02783296389954426, "ret_signed_p1d": -0.02783296389954426, "alpha_spy_p1d": -0.0194355033413377, "alpha_c_p1d": -0.011580009741389619, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.016252954158154642, "ret_p1w": -0.0387674012229241, "ret_signed_p1w": -0.0387674012229241, "alpha_spy_p1w": -0.043364256354718855, "alpha_c_p1w": -0.024911758881995483, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.01385564234092862, "ret_p1m": 0.021868796448681227, "ret_signed_p1m": 0.021868796448681227, "alpha_spy_p1m": 0.0020390896861928187, "alpha_c_p1m": 0.016650921047398626, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.005217875401282601, "ret_p3m": 0.7839781208032068, "ret_signed_p3m": 0.7839781208032068, "alpha_spy_p3m": 0.7575188399158186, "alpha_c_p3m": 0.8008927632320656, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": -0.01691464242885876, "ret_p6m": 1.7926097811095025, "ret_signed_p6m": 1.7926097811095025, "alpha_spy_p6m": 1.677367689847022, "alpha_c_p6m": 1.5538288929281312, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.23878088818137133, "price_path": [-0.3023, -0.3013, -0.2934, -0.2924, -0.3043, -0.3013, -0.3112, -0.3102, -0.331, -0.3162, -0.3092, -0.3063, -0.3122, -0.3063, -0.2924, -0.2815, -0.2736, -0.2538, -0.2429, -0.2142, -0.2261, -0.2053, -0.2053, -0.1737, -0.1618, -0.1549, -0.1479, -0.1756, -0.163, -0.166, -0.1451, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.0616, -0.0726, -0.0895, -0.0557, -0.0288, -0.0268, -0.0249, -0.0308, -0.0199, -0.0408, -0.0179, 0.0139, -0.0109, -0.001, 0.0348, 0.0686, 0.1044, 0.0427, 0.0, -0.0139, -0.0268, 0.0, 0.0288, 0.0249, 0.0219, -0.0338, 0.0, -0.0278, -0.0408, 0.0, -0.0577, -0.0388, -0.0129, 0.0219, 0.0288, -0.001, 0.002, 0.0278, 0.0388, 0.0447, 0.0775, 0.0885, 0.0775, 0.0736, 0.0666, 0.0825, 0.0417, 0.0219, 0.0726, 0.0696, 0.0567, 0.0984, 0.1083, 0.1044, 0.1044, 0.163, 0.1936, 0.1976, 0.1976, 0.1976, 0.2835, 0.3794, 0.3874, 0.4084, 0.3864, 0.3884, 0.3864, 0.3744, 0.4014, 0.4374, 0.4873, 0.4913, 0.4504, 0.4933, 0.5213, 0.5193, 0.5193, 0.5932, 0.6222, 0.6052, 0.6032, 0.5023, 0.6731, 0.6891, 0.5912, 0.5842, 0.6621, 0.6561, 0.6761, 0.784, 0.8099, 0.8099, 0.8099, 0.8099, 0.8978, 0.8988, 0.9278, 0.9977, 1.0327, 1.1775, 1.1625, 1.1625, 0.9488, 0.7201, 0.9138, 0.8799, 0.733, 0.8769, 0.8978, 0.8769, 0.8329, 0.8849, 0.9368, 1.0826, 1.0027, 0.9917, 0.8609, 0.8949, 0.8879, 0.799, 0.799, 0.7646, 0.6736, 0.8983, 0.7857, 0.8637, 0.9328, 0.9668, 1.107, 1.0419, 1.0619, 1.0119, 1.0669, 1.112, 1.177, 1.162, 1.147, 1.192, 1.177, 1.2471, 1.1971, 1.2471, 1.2221, 1.2621, 1.2071, 1.2071, 1.3272, 1.3272, 1.6625, 1.7175, 1.6875, 1.8577, 1.7926]}
{"unit_id": "orig:1990013494243299340", "ticker_idx": 2, "ts": "2025-11-16T11:06:04+00:00", "symbol": "TES", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We suggest Wonik IPS and TES as beneficiaries within our coverage", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz maintains positive view on Korean front-end equipment sector on Samsung P5 investment; top picks Wonik IPS and TES, also positive on Eugene Tech, GST, PSK, EOTechnics.", "resolved_tickers": ["095610.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "TES (Korea semiconductor equipment, Wonik IPS context) listed on KOSDAQ 095610.KQ", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities IT Kim Sun-woo / Kim Dong-kwan]\n\nHello. This is Kim Sun-woo / Kim Dong-kwan from Meritz. We are sharing our firm's comments on the media reports regarding Samsung Electronics' P5 investment.\n\n▶ Implications of Samsung Electronics' P5 Investment Announcement\n\n[Samsung Electronics' P5 investment announcement. Assessed as a long-term investment plan aimed at securing mid-to-long-term competitiveness]\n\nAccording to media reports, Samsung Electronics held a temporary management committee meeting and announced plans to proceed with the framework construction for Pyeongtaek P5.\n\nThe company currently operates a total of 4 production fabs up to P4 at its Pyeongtaek campus. In the case of P4, where investment recently began, around 60k of DRAM Capa is expected to be added between 4Q25 and 2Q26.\n\nThe company had originally planned to start the P4 investment in early 2025, but it has shifted direction to execute the investment at a time when visibility for DRAM price increases is secured. This investment announcement is unrelated to oversupply and is assessed as a sound investment execution plan aimed at securing mid-to-long-term competitiveness.\n\nThe company's P5 investment has the following two implications:\n\n1. Production start at the earliest 1H28, unrelated to short-term supply increase: P5 was prepared with the intention of securing cleanrooms in advance during a downcycle, following Samsung Electronics' past 'Shell-First' strategy, but it had been delayed until recently due to the HBM-focused market development in 2024-2025.\n\nNow, Samsung Electronics has decided to partially pull forward the P5 investment to lead the market, based on establishing price negotiation power for some commodity products and securing business visibility. It is assessed that this was partially influenced by the disclosure of the operation schedule for competitor SK hynix's Y1 fab.\n\nHowever, Y1 also begins operation in 1H27 at the earliest, and P5 will only start operations in 2028, so concerns about a short-term increase in memory supply are non-existent.\n\nInvestors should note that, unlike in the past, memory makers' production expansion efforts are focused on a 'mid-to-long-term' perspective. The immediate front-end equipment investment that used to occur during past upcycles is proceeding only in a limited manner, and an unusual cycle is continuing where all producers are focusing on profitability rather than current volume.\n\n2. 3-story structure for synergy creation upon foundry recovery: P5 is presumed to be the latest fab with a wider site area and a 3-story structure (vs. the existing 2-story structure), enabling greater vertical cleanroom space utilization.\n\nAssuming the recovery of foundry competitiveness, Samsung Electronics will be able to configure efficient production lines through sequential investments in a total of Phase 6 starting from 2028.\n\nIt is judged that it will be possible to reduce production time and secure cost competitiveness by configuring continuous and sequential production lines for foundry processes (ASIC, etc.) and memory processes in the future. However, expanding foundry orders and securing process competitiveness after SF2P are essential prerequisites.\n\n[Impact on Value Chain and Investment Opinion – Maintain Positive View]\n\nWe maintain our positive investment opinion on the front-end equipment sector following this investment announcement. Considering the expected mass production timing (2028E), new orders for P5 from front-end equipment makers are expected to occur starting from 2H27E.\n\nThe impact on 2025-26 earnings is limited, but this period is already a time when earnings growth is expected due to new P4 DRAM investment and conversion investments in existing DRAM/NAND lines.\n\nTypically, the stock prices of the front-end equipment sector move in tandem with expectations for the memory industry outlook and front-end customer CAPEX. The stock price rise since September is the result of expectations for 2026E CAPEX expansion being reflected as a multiple-centric expansion.\n\nThis investment announcement is positive in that it secures visibility for the long-term profit growth of equipment makers. 2027 is when SK hynix's Yongin Cluster construction and equipment investment will be underway. In addition, the P5 equipment order momentum in 2H27 and equipment revenue recognition in 2028E are expected to unfold sequentially.\n\nAs profit growth beyond 2026 and further acceleration of investment unfold, it is expected to have a positive impact on equipment sector stock prices.\n\nWe suggest Wonik IPS and TES as beneficiaries within our coverage. As of 1H25 cumulative, the sales exposure to Samsung Electronics is estimated at 58% for Wonik IPS and 35% for TES. Wonik IPS can benefit from investments across all of Samsung Electronics' product lines—DRAM, NAND, and Foundry—through its general-purpose CVD/ALD equipment.\n\nIn addition to the companies mentioned above, it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics.\n\nThe Samsung Electronics exposure we estimate for major front-end equipment/facility companies is as follows: Wonik IPS (58%), TES (35%), Eugene Technology (40%), GST (50%), UNISEM (50%), PSK (25%), EOTechnics (25%).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10515014122854816, "ret_signed_m1d": -0.10515014122854816, "alpha_spy_m1d": -0.11455427903442883, "alpha_c_m1d": -0.11708306447020023, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.011932923241652071, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05579399435085808, "ret_signed_p1d": -0.05579399435085808, "alpha_spy_p1d": -0.04739653379265152, "alpha_c_p1d": -0.04042100726747133, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.015372987083386747, "ret_p1w": -0.15879827189042495, "ret_signed_p1w": -0.15879827189042495, "alpha_spy_p1w": -0.1633951270222197, "alpha_c_p1w": -0.12955735186098638, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.029240920029438566, "ret_p1m": -0.2167381299302643, "ret_signed_p1m": -0.2167381299302643, "alpha_spy_p1m": -0.23656783669275272, "alpha_c_p1m": -0.20072712803232118, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": -0.016011001897943133, "ret_p3m": 0.44845585334384475, "ret_signed_p3m": 0.44845585334384475, "alpha_spy_p3m": 0.4219965724564565, "alpha_c_p3m": 0.010723851768599024, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.4377320015752457, "ret_p6m": 1.3761677594130144, "ret_signed_p6m": 1.3761677594130144, "alpha_spy_p6m": 1.2609256681505339, "alpha_c_p6m": 0.414125280104888, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.9620424793081264, "price_path": [-0.367, -0.368, -0.368, -0.3573, -0.3562, -0.3562, -0.3616, -0.3552, -0.3927, -0.3927, -0.3745, -0.3777, -0.382, -0.3841, -0.3348, -0.3036, -0.2897, -0.1921, -0.1105, -0.1084, -0.1094, -0.0665, -0.0665, -0.059, -0.0494, -0.0215, -0.059, -0.0987, -0.0676, -0.073, -0.0644, -0.0172, -0.0172, -0.0172, -0.0172, -0.0172, -0.0172, 0.0118, -0.0215, -0.0215, 0.0043, -0.0118, 0.0043, 0.0612, -0.0118, -0.0032, 0.0461, 0.0687, 0.0773, -0.0354, -0.0139, -0.0461, -0.0494, -0.0976, -0.0515, -0.1105, -0.0708, -0.1406, -0.073, -0.0343, -0.03, -0.044, -0.1052, 0.0, -0.0558, -0.0579, -0.088, -0.1803, -0.1588, -0.1781, -0.1631, -0.1642, -0.1534, -0.1137, -0.1245, -0.118, -0.1685, -0.1663, -0.162, -0.1803, -0.1567, -0.1727, -0.1953, -0.2028, -0.2167, -0.1631, -0.176, -0.1727, -0.1298, -0.1084, -0.1148, -0.1148, -0.0515, -0.0307, -0.0252, -0.0252, -0.0252, 0.164, 0.1946, 0.2909, 0.2669, 0.1925, 0.1443, 0.1203, 0.1334, 0.1334, 0.0984, 0.1093, 0.1071, 0.0645, 0.0349, 0.0371, 0.0437, 0.1859, 0.2187, 0.4616, 0.5754, 0.5797, 0.4725, 0.5382, 0.4922, 0.4047, 0.4747, 0.5404, 0.455, 0.4069, 0.4485, 0.4463, 0.4463, 0.4463, 0.4463, 0.5644, 0.6147, 0.5207, 0.5754, 0.5294, 0.6541, 0.6104, 0.6104, 0.5382, 0.3237, 0.5732, 0.6869, 0.4835, 0.5053, 0.455, 0.4134, 0.4288, 0.4463, 0.4266, 0.4944, 0.4463, 0.4397, 0.315, 0.3784, 0.4966, 0.3872, 0.3872, 0.2384, 0.2122, 0.3566, 0.2844, 0.3106, 0.3566, 0.3653, 0.5119, 0.6344, 0.8051, 0.8095, 0.8335, 0.9714, 0.9648, 0.9254, 0.9758, 0.9386, 1.0414, 0.9604, 0.9626, 0.9364, 0.8948, 0.8948, 0.8379, 0.8379, 0.897, 0.897, 0.9626, 0.9517, 1.0327, 1.3018, 1.3762]}
{"unit_id": "orig:1990013494243299340", "ticker_idx": 3, "ts": "2025-11-16T11:06:04+00:00", "symbol": "Eugene Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz maintains positive view on Korean front-end equipment sector on Samsung P5 investment; top picks Wonik IPS and TES, also positive on Eugene Tech, GST, PSK, EOTechnics.", "resolved_tickers": ["084370.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "Eugene Technology 084370.KQ Korea KOSDAQ LPCVD", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KQ')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities IT Kim Sun-woo / Kim Dong-kwan]\n\nHello. This is Kim Sun-woo / Kim Dong-kwan from Meritz. We are sharing our firm's comments on the media reports regarding Samsung Electronics' P5 investment.\n\n▶ Implications of Samsung Electronics' P5 Investment Announcement\n\n[Samsung Electronics' P5 investment announcement. Assessed as a long-term investment plan aimed at securing mid-to-long-term competitiveness]\n\nAccording to media reports, Samsung Electronics held a temporary management committee meeting and announced plans to proceed with the framework construction for Pyeongtaek P5.\n\nThe company currently operates a total of 4 production fabs up to P4 at its Pyeongtaek campus. In the case of P4, where investment recently began, around 60k of DRAM Capa is expected to be added between 4Q25 and 2Q26.\n\nThe company had originally planned to start the P4 investment in early 2025, but it has shifted direction to execute the investment at a time when visibility for DRAM price increases is secured. This investment announcement is unrelated to oversupply and is assessed as a sound investment execution plan aimed at securing mid-to-long-term competitiveness.\n\nThe company's P5 investment has the following two implications:\n\n1. Production start at the earliest 1H28, unrelated to short-term supply increase: P5 was prepared with the intention of securing cleanrooms in advance during a downcycle, following Samsung Electronics' past 'Shell-First' strategy, but it had been delayed until recently due to the HBM-focused market development in 2024-2025.\n\nNow, Samsung Electronics has decided to partially pull forward the P5 investment to lead the market, based on establishing price negotiation power for some commodity products and securing business visibility. It is assessed that this was partially influenced by the disclosure of the operation schedule for competitor SK hynix's Y1 fab.\n\nHowever, Y1 also begins operation in 1H27 at the earliest, and P5 will only start operations in 2028, so concerns about a short-term increase in memory supply are non-existent.\n\nInvestors should note that, unlike in the past, memory makers' production expansion efforts are focused on a 'mid-to-long-term' perspective. The immediate front-end equipment investment that used to occur during past upcycles is proceeding only in a limited manner, and an unusual cycle is continuing where all producers are focusing on profitability rather than current volume.\n\n2. 3-story structure for synergy creation upon foundry recovery: P5 is presumed to be the latest fab with a wider site area and a 3-story structure (vs. the existing 2-story structure), enabling greater vertical cleanroom space utilization.\n\nAssuming the recovery of foundry competitiveness, Samsung Electronics will be able to configure efficient production lines through sequential investments in a total of Phase 6 starting from 2028.\n\nIt is judged that it will be possible to reduce production time and secure cost competitiveness by configuring continuous and sequential production lines for foundry processes (ASIC, etc.) and memory processes in the future. However, expanding foundry orders and securing process competitiveness after SF2P are essential prerequisites.\n\n[Impact on Value Chain and Investment Opinion – Maintain Positive View]\n\nWe maintain our positive investment opinion on the front-end equipment sector following this investment announcement. Considering the expected mass production timing (2028E), new orders for P5 from front-end equipment makers are expected to occur starting from 2H27E.\n\nThe impact on 2025-26 earnings is limited, but this period is already a time when earnings growth is expected due to new P4 DRAM investment and conversion investments in existing DRAM/NAND lines.\n\nTypically, the stock prices of the front-end equipment sector move in tandem with expectations for the memory industry outlook and front-end customer CAPEX. The stock price rise since September is the result of expectations for 2026E CAPEX expansion being reflected as a multiple-centric expansion.\n\nThis investment announcement is positive in that it secures visibility for the long-term profit growth of equipment makers. 2027 is when SK hynix's Yongin Cluster construction and equipment investment will be underway. In addition, the P5 equipment order momentum in 2H27 and equipment revenue recognition in 2028E are expected to unfold sequentially.\n\nAs profit growth beyond 2026 and further acceleration of investment unfold, it is expected to have a positive impact on equipment sector stock prices.\n\nWe suggest Wonik IPS and TES as beneficiaries within our coverage. As of 1H25 cumulative, the sales exposure to Samsung Electronics is estimated at 58% for Wonik IPS and 35% for TES. Wonik IPS can benefit from investments across all of Samsung Electronics' product lines—DRAM, NAND, and Foundry—through its general-purpose CVD/ALD equipment.\n\nIn addition to the companies mentioned above, it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics.\n\nThe Samsung Electronics exposure we estimate for major front-end equipment/facility companies is as follows: Wonik IPS (58%), TES (35%), Eugene Technology (40%), GST (50%), UNISEM (50%), PSK (25%), EOTechnics (25%).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01112343827868445, "ret_signed_m1d": -0.01112343827868445, "alpha_spy_m1d": -0.020527576084565125, "alpha_c_m1d": -0.02305636152033652, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.011932923241652071, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05450496072982247, "ret_signed_p1d": -0.05450496072982247, "alpha_spy_p1d": -0.04610750017161591, "alpha_c_p1d": -0.039131973646435725, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.015372987083386747, "ret_p1w": -0.1423804103945734, "ret_signed_p1w": -0.1423804103945734, "alpha_spy_p1w": -0.14697726552636814, "alpha_c_p1w": -0.11313949036513482, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.029240920029438566, "ret_p1m": -0.2013348160754196, "ret_signed_p1m": -0.2013348160754196, "alpha_spy_p1m": -0.221164522837908, "alpha_c_p1m": -0.18532381417747645, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": -0.016011001897943133, "ret_p3m": 0.125695209451828, "ret_signed_p3m": 0.125695209451828, "alpha_spy_p3m": 0.09923592856443975, "alpha_c_p3m": -0.3120367921234177, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.4377320015752457, "ret_p6m": 0.4830438576828746, "ret_signed_p6m": 0.4830438576828746, "alpha_spy_p6m": 0.3678017664203941, "alpha_c_p6m": -0.47899862162525175, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.9620424793081264, "price_path": [-0.4522, -0.4505, -0.4661, -0.4549, -0.4611, -0.4627, -0.4689, -0.4666, -0.4922, -0.4944, -0.485, -0.4889, -0.4944, -0.4828, -0.4538, -0.4461, -0.4316, -0.3915, -0.3571, -0.3548, -0.3126, -0.2848, -0.2848, -0.2469, -0.2125, -0.2258, -0.2336, -0.2581, -0.228, -0.2225, -0.1935, -0.1346, -0.1346, -0.1346, -0.1346, -0.1346, -0.1346, -0.1146, -0.1268, -0.1457, -0.1179, -0.1146, -0.0968, -0.0211, -0.0033, -0.0078, 0.0122, 0.0222, 0.0289, -0.01, 0.0923, 0.0879, 0.0556, 0.0467, 0.1324, 0.1613, 0.1758, 0.1201, 0.1179, 0.0923, 0.0845, 0.0879, -0.0111, 0.0, -0.0545, -0.0334, -0.0745, -0.1402, -0.1424, -0.1413, -0.1201, -0.1157, -0.1123, -0.1068, -0.1079, -0.1023, -0.1457, -0.1435, -0.1335, -0.1546, -0.1368, -0.1424, -0.1657, -0.1902, -0.2013, -0.1769, -0.2013, -0.1835, -0.1479, -0.1479, -0.178, -0.178, -0.1591, -0.1657, -0.1657, -0.1657, -0.1657, -0.0478, -0.0044, 0.0189, -0.0111, -0.0367, -0.0845, -0.0845, -0.0634, -0.0645, -0.0845, -0.0879, -0.0845, -0.1168, -0.1335, -0.1546, -0.1602, -0.0701, 0.0211, 0.1368, 0.1958, 0.2347, 0.1057, 0.1591, 0.1691, 0.0734, 0.1324, 0.1835, 0.0979, 0.0868, 0.1257, 0.2458, 0.2458, 0.2458, 0.2458, 0.4794, 0.5072, 0.4171, 0.3971, 0.3637, 0.5317, 0.4905, 0.4905, 0.4427, 0.3115, 0.6118, 0.6463, 0.4894, 0.4383, 0.3815, 0.4171, 0.3515, 0.3548, 0.3504, 0.4383, 0.4672, 0.4394, 0.3915, 0.4294, 0.5117, 0.3828, 0.3014, 0.2401, 0.2335, 0.3672, 0.3984, 0.3349, 0.3193, 0.3382, 0.4886, 0.4151, 0.4084, 0.3805, 0.4151, 0.4362, 0.5031, 0.4775, 0.4407, 0.4307, 0.4998, 0.4908, 0.5098, 0.5154, 0.5822, 0.5599, 0.4485, 0.4485, 0.4708, 0.4708, 0.5454, 0.6368, 0.6557, 0.6747, 0.483]}
{"unit_id": "orig:1990013494243299340", "ticker_idx": 4, "ts": "2025-11-16T11:06:04+00:00", "symbol": "GST", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz maintains positive view on Korean front-end equipment sector on Samsung P5 investment; top picks Wonik IPS and TES, also positive on Eugene Tech, GST, PSK, EOTechnics.", "resolved_tickers": ["036830.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "GST Co. 036830.KQ Korea semiconductor equipment", "bench_c": "XLB", "bench_c_reason": "industry='Chemicals'", "bench_c_industry": "Chemicals", "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities IT Kim Sun-woo / Kim Dong-kwan]\n\nHello. This is Kim Sun-woo / Kim Dong-kwan from Meritz. We are sharing our firm's comments on the media reports regarding Samsung Electronics' P5 investment.\n\n▶ Implications of Samsung Electronics' P5 Investment Announcement\n\n[Samsung Electronics' P5 investment announcement. Assessed as a long-term investment plan aimed at securing mid-to-long-term competitiveness]\n\nAccording to media reports, Samsung Electronics held a temporary management committee meeting and announced plans to proceed with the framework construction for Pyeongtaek P5.\n\nThe company currently operates a total of 4 production fabs up to P4 at its Pyeongtaek campus. In the case of P4, where investment recently began, around 60k of DRAM Capa is expected to be added between 4Q25 and 2Q26.\n\nThe company had originally planned to start the P4 investment in early 2025, but it has shifted direction to execute the investment at a time when visibility for DRAM price increases is secured. This investment announcement is unrelated to oversupply and is assessed as a sound investment execution plan aimed at securing mid-to-long-term competitiveness.\n\nThe company's P5 investment has the following two implications:\n\n1. Production start at the earliest 1H28, unrelated to short-term supply increase: P5 was prepared with the intention of securing cleanrooms in advance during a downcycle, following Samsung Electronics' past 'Shell-First' strategy, but it had been delayed until recently due to the HBM-focused market development in 2024-2025.\n\nNow, Samsung Electronics has decided to partially pull forward the P5 investment to lead the market, based on establishing price negotiation power for some commodity products and securing business visibility. It is assessed that this was partially influenced by the disclosure of the operation schedule for competitor SK hynix's Y1 fab.\n\nHowever, Y1 also begins operation in 1H27 at the earliest, and P5 will only start operations in 2028, so concerns about a short-term increase in memory supply are non-existent.\n\nInvestors should note that, unlike in the past, memory makers' production expansion efforts are focused on a 'mid-to-long-term' perspective. The immediate front-end equipment investment that used to occur during past upcycles is proceeding only in a limited manner, and an unusual cycle is continuing where all producers are focusing on profitability rather than current volume.\n\n2. 3-story structure for synergy creation upon foundry recovery: P5 is presumed to be the latest fab with a wider site area and a 3-story structure (vs. the existing 2-story structure), enabling greater vertical cleanroom space utilization.\n\nAssuming the recovery of foundry competitiveness, Samsung Electronics will be able to configure efficient production lines through sequential investments in a total of Phase 6 starting from 2028.\n\nIt is judged that it will be possible to reduce production time and secure cost competitiveness by configuring continuous and sequential production lines for foundry processes (ASIC, etc.) and memory processes in the future. However, expanding foundry orders and securing process competitiveness after SF2P are essential prerequisites.\n\n[Impact on Value Chain and Investment Opinion – Maintain Positive View]\n\nWe maintain our positive investment opinion on the front-end equipment sector following this investment announcement. Considering the expected mass production timing (2028E), new orders for P5 from front-end equipment makers are expected to occur starting from 2H27E.\n\nThe impact on 2025-26 earnings is limited, but this period is already a time when earnings growth is expected due to new P4 DRAM investment and conversion investments in existing DRAM/NAND lines.\n\nTypically, the stock prices of the front-end equipment sector move in tandem with expectations for the memory industry outlook and front-end customer CAPEX. The stock price rise since September is the result of expectations for 2026E CAPEX expansion being reflected as a multiple-centric expansion.\n\nThis investment announcement is positive in that it secures visibility for the long-term profit growth of equipment makers. 2027 is when SK hynix's Yongin Cluster construction and equipment investment will be underway. In addition, the P5 equipment order momentum in 2H27 and equipment revenue recognition in 2028E are expected to unfold sequentially.\n\nAs profit growth beyond 2026 and further acceleration of investment unfold, it is expected to have a positive impact on equipment sector stock prices.\n\nWe suggest Wonik IPS and TES as beneficiaries within our coverage. As of 1H25 cumulative, the sales exposure to Samsung Electronics is estimated at 58% for Wonik IPS and 35% for TES. Wonik IPS can benefit from investments across all of Samsung Electronics' product lines—DRAM, NAND, and Foundry—through its general-purpose CVD/ALD equipment.\n\nIn addition to the companies mentioned above, it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics.\n\nThe Samsung Electronics exposure we estimate for major front-end equipment/facility companies is as follows: Wonik IPS (58%), TES (35%), Eugene Technology (40%), GST (50%), UNISEM (50%), PSK (25%), EOTechnics (25%).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019300382096225754, "ret_signed_m1d": -0.019300382096225754, "alpha_spy_m1d": -0.02870451990210643, "alpha_c_m1d": -0.03474806424528032, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.015447682149054565, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05186977688360683, "ret_signed_p1d": -0.05186977688360683, "alpha_spy_p1d": -0.04347231632540027, "alpha_c_p1d": -0.05268901380696123, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": 0.0008192369233543939, "ret_p1w": -0.11338964960071751, "ret_signed_p1w": -0.11338964960071751, "alpha_spy_p1w": -0.11798650473251227, "alpha_c_p1w": -0.12591158861768725, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": 0.012521939016969741, "ret_p1m": -0.057901146288677374, "ret_signed_p1m": -0.057901146288677374, "alpha_spy_p1m": -0.07773085305116578, "alpha_c_p1m": -0.11114875181456019, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.05324760552588281, "ret_p3m": 0.24004821667798137, "ret_signed_p3m": 0.24004821667798137, "alpha_spy_p3m": 0.21358893579059313, "alpha_c_p3m": -0.002847418054851669, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.24289563473283304, "ret_p6m": 0.4868713336424635, "ret_signed_p6m": 0.4868713336424635, "alpha_spy_p6m": 0.37162924237998296, "alpha_c_p6m": 0.2547537846825103, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.23211754895995318, "price_path": [0.0326, 0.0241, 0.0157, 0.0386, 0.0193, 0.0193, -0.041, -0.0507, -0.0736, -0.0724, -0.0736, -0.0796, -0.0651, 0.029, 0.0097, 0.0181, 0.0181, 0.0326, 0.0615, 0.0832, 0.0567, 0.0688, 0.0688, 0.0579, 0.0784, 0.0495, 0.0555, 0.0121, 0.0157, 0.0253, 0.0386, 0.0458, 0.0458, 0.0458, 0.0458, 0.0458, 0.0458, 0.0543, 0.0615, 0.0555, 0.0531, 0.0832, 0.0953, 0.0808, 0.082, 0.082, 0.0374, 0.0856, 0.1267, 0.1315, 0.1158, 0.0651, 0.035, 0.0277, 0.035, -0.0109, 0.0097, -0.0434, -0.0109, 0.0302, 0.0205, 0.041, -0.0193, 0.0, -0.0519, -0.082, -0.0615, -0.1086, -0.1134, -0.1074, -0.0724, -0.0434, -0.0084, 0.0338, 0.0338, 0.0338, 0.0133, 0.0109, 0.0277, 0.0326, 0.0314, 0.0, -0.0133, -0.0302, -0.0579, -0.0555, -0.0893, -0.0905, -0.0615, -0.0639, -0.0639, -0.0639, -0.0712, -0.0832, -0.1001, -0.1001, -0.1001, -0.0796, -0.0543, -0.047, -0.0724, -0.0808, -0.1025, -0.082, -0.082, -0.082, -0.0796, -0.1037, -0.0772, -0.076, -0.0712, -0.0398, -0.0169, 0.0302, 0.0615, 0.3534, 0.3486, 0.421, 0.2183, 0.2497, 0.2835, 0.2328, 0.1725, 0.339, 0.2762, 0.2304, 0.24, 0.1942, 0.1942, 0.1942, 0.1942, 0.3317, 0.3004, 0.2473, 0.2762, 0.2955, 0.3148, 0.3438, 0.3438, 0.2545, 0.0591, 0.24, 0.3221, 0.2039, 0.3848, 0.5054, 0.4982, 0.3752, 0.3148, 0.2979, 0.3269, 0.2473, 0.2618, 0.1339, 0.2063, 0.2376, 0.1713, 0.1785, 0.1517, 0.1066, 0.2017, 0.1481, 0.1651, 0.1749, 0.2553, 0.2651, 0.2236, 0.3162, 0.3382, 0.3041, 0.3041, 0.3162, 0.3284, 0.4869, 0.6453, 0.5746, 0.482, 0.5381, 0.7306, 0.8744, 0.9402, 0.7794, 0.7794, 0.7526, 0.7526, 0.7355, 0.6843, 0.6185, 0.5405, 0.4869]}
{"unit_id": "orig:1990013494243299340", "ticker_idx": 6, "ts": "2025-11-16T11:06:04+00:00", "symbol": "EOTechnics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics", "tweet_type": "analysis", "timeline": "quarters", "summary": "Meritz maintains positive view on Korean front-end equipment sector on Samsung P5 investment; top picks Wonik IPS and TES, also positive on Eugene Tech, GST, PSK, EOTechnics.", "resolved_tickers": ["039030.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "EO Technics 039030.KQ Korea KOSDAQ", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "[Meritz Securities IT Kim Sun-woo / Kim Dong-kwan]\n\nHello. This is Kim Sun-woo / Kim Dong-kwan from Meritz. We are sharing our firm's comments on the media reports regarding Samsung Electronics' P5 investment.\n\n▶ Implications of Samsung Electronics' P5 Investment Announcement\n\n[Samsung Electronics' P5 investment announcement. Assessed as a long-term investment plan aimed at securing mid-to-long-term competitiveness]\n\nAccording to media reports, Samsung Electronics held a temporary management committee meeting and announced plans to proceed with the framework construction for Pyeongtaek P5.\n\nThe company currently operates a total of 4 production fabs up to P4 at its Pyeongtaek campus. In the case of P4, where investment recently began, around 60k of DRAM Capa is expected to be added between 4Q25 and 2Q26.\n\nThe company had originally planned to start the P4 investment in early 2025, but it has shifted direction to execute the investment at a time when visibility for DRAM price increases is secured. This investment announcement is unrelated to oversupply and is assessed as a sound investment execution plan aimed at securing mid-to-long-term competitiveness.\n\nThe company's P5 investment has the following two implications:\n\n1. Production start at the earliest 1H28, unrelated to short-term supply increase: P5 was prepared with the intention of securing cleanrooms in advance during a downcycle, following Samsung Electronics' past 'Shell-First' strategy, but it had been delayed until recently due to the HBM-focused market development in 2024-2025.\n\nNow, Samsung Electronics has decided to partially pull forward the P5 investment to lead the market, based on establishing price negotiation power for some commodity products and securing business visibility. It is assessed that this was partially influenced by the disclosure of the operation schedule for competitor SK hynix's Y1 fab.\n\nHowever, Y1 also begins operation in 1H27 at the earliest, and P5 will only start operations in 2028, so concerns about a short-term increase in memory supply are non-existent.\n\nInvestors should note that, unlike in the past, memory makers' production expansion efforts are focused on a 'mid-to-long-term' perspective. The immediate front-end equipment investment that used to occur during past upcycles is proceeding only in a limited manner, and an unusual cycle is continuing where all producers are focusing on profitability rather than current volume.\n\n2. 3-story structure for synergy creation upon foundry recovery: P5 is presumed to be the latest fab with a wider site area and a 3-story structure (vs. the existing 2-story structure), enabling greater vertical cleanroom space utilization.\n\nAssuming the recovery of foundry competitiveness, Samsung Electronics will be able to configure efficient production lines through sequential investments in a total of Phase 6 starting from 2028.\n\nIt is judged that it will be possible to reduce production time and secure cost competitiveness by configuring continuous and sequential production lines for foundry processes (ASIC, etc.) and memory processes in the future. However, expanding foundry orders and securing process competitiveness after SF2P are essential prerequisites.\n\n[Impact on Value Chain and Investment Opinion – Maintain Positive View]\n\nWe maintain our positive investment opinion on the front-end equipment sector following this investment announcement. Considering the expected mass production timing (2028E), new orders for P5 from front-end equipment makers are expected to occur starting from 2H27E.\n\nThe impact on 2025-26 earnings is limited, but this period is already a time when earnings growth is expected due to new P4 DRAM investment and conversion investments in existing DRAM/NAND lines.\n\nTypically, the stock prices of the front-end equipment sector move in tandem with expectations for the memory industry outlook and front-end customer CAPEX. The stock price rise since September is the result of expectations for 2026E CAPEX expansion being reflected as a multiple-centric expansion.\n\nThis investment announcement is positive in that it secures visibility for the long-term profit growth of equipment makers. 2027 is when SK hynix's Yongin Cluster construction and equipment investment will be underway. In addition, the P5 equipment order momentum in 2H27 and equipment revenue recognition in 2028E are expected to unfold sequentially.\n\nAs profit growth beyond 2026 and further acceleration of investment unfold, it is expected to have a positive impact on equipment sector stock prices.\n\nWe suggest Wonik IPS and TES as beneficiaries within our coverage. As of 1H25 cumulative, the sales exposure to Samsung Electronics is estimated at 58% for Wonik IPS and 35% for TES. Wonik IPS can benefit from investments across all of Samsung Electronics' product lines—DRAM, NAND, and Foundry—through its general-purpose CVD/ALD equipment.\n\nIn addition to the companies mentioned above, it is expected to lead to benefits for all front-end equipment companies with high sales exposure to Samsung Electronics.\n\nThe Samsung Electronics exposure we estimate for major front-end equipment/facility companies is as follows: Wonik IPS (58%), TES (35%), Eugene Technology (40%), GST (50%), UNISEM (50%), PSK (25%), EOTechnics (25%).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01423496053125084, "ret_signed_m1d": -0.01423496053125084, "alpha_spy_m1d": -0.023639098337131514, "alpha_c_m1d": -0.027987663754248215, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.013752703222997376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021352329071261833, "ret_signed_p1d": -0.021352329071261833, "alpha_spy_p1d": -0.012954868513055273, "alpha_c_p1d": -0.0007529602131104429, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.02059936885815139, "ret_p1w": -0.1263345764297611, "ret_signed_p1w": -0.1263345764297611, "alpha_spy_p1w": -0.13093143156155584, "alpha_c_p1w": -0.1228670744776833, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.0034675019520777894, "ret_p1m": -0.15658367203884382, "ret_signed_p1m": -0.15658367203884382, "alpha_spy_p1m": -0.17641337880133223, "alpha_c_p1m": -0.19078884351583802, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.0342051714769942, "ret_p3m": 0.3532206416647832, "ret_signed_p3m": 0.3532206416647832, "alpha_spy_p3m": 0.3267613607773949, "alpha_c_p3m": 0.15611907244608214, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.19710156921870103, "ret_p6m": 0.5301940148811815, "ret_signed_p6m": 0.5301940148811815, "alpha_spy_p6m": 0.4149519236187009, "alpha_c_p6m": -0.1242180408636846, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.654412055744866, "price_path": [-0.2456, -0.2491, -0.2562, -0.2473, -0.2562, -0.2438, -0.2562, -0.2651, -0.3004, -0.3078, -0.2993, -0.2865, -0.289, -0.258, -0.2473, -0.2456, -0.2491, -0.2278, -0.2206, -0.1904, -0.2135, -0.1068, -0.1068, -0.1246, -0.1352, -0.1477, -0.1637, -0.1779, -0.1601, -0.1708, -0.169, -0.153, -0.153, -0.153, -0.153, -0.153, -0.153, -0.0854, -0.0765, -0.1103, -0.0872, -0.073, -0.1157, -0.1423, -0.1851, -0.1868, -0.1762, -0.1655, -0.1601, -0.1815, -0.1637, -0.1673, -0.1601, -0.1335, 0.0445, -0.0285, -0.0214, -0.0676, -0.0053, 0.0374, 0.016, 0.073, -0.0142, 0.0, -0.0214, -0.0231, -0.0214, -0.1174, -0.1263, -0.121, -0.0872, -0.0854, -0.0854, -0.0285, -0.0178, -0.0053, -0.0107, 0.0142, -0.0125, -0.032, -0.0356, -0.0356, -0.0391, -0.1032, -0.1566, -0.1032, -0.0996, -0.1174, -0.0854, -0.0819, -0.0925, -0.0925, -0.0747, -0.0454, -0.0275, -0.0275, -0.0275, 0.0332, 0.1333, 0.1459, 0.0994, 0.1101, 0.1065, 0.1101, 0.1012, 0.1226, 0.0779, 0.0726, 0.069, 0.035, 0.0458, 0.0708, 0.0708, 0.1548, 0.1959, 0.2513, 0.2889, 0.3443, 0.2674, 0.4158, 0.3979, 0.3711, 0.414, 0.4944, 0.3979, 0.3425, 0.3532, 0.305, 0.305, 0.305, 0.305, 0.3514, 0.3318, 0.3157, 0.3818, 0.3371, 0.5052, 0.4962, 0.4962, 0.507, 0.2924, 0.5945, 0.6196, 0.4766, 0.498, 0.4247, 0.4676, 0.4676, 0.4658, 0.4104, 0.4676, 0.4748, 0.5356, 0.498, 0.5141, 0.5963, 0.4587, 0.4909, 0.38, 0.3747, 0.5499, 0.4211, 0.4158, 0.4122, 0.473, 0.5802, 0.5248, 0.6017, 0.7519, 0.7215, 0.6875, 0.6893, 0.6821, 0.6804, 0.7268, 0.7268, 0.7322, 0.8091, 0.8055, 0.8341, 0.8448, 0.6732, 0.6732, 0.6982, 0.6982, 0.7125, 0.6571, 0.7107, 0.7912, 0.5302]}
{"unit_id": "thread:1997813691853722054", "ticker_idx": 0, "ts": "2025-12-07T23:41:16+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "If every CEO were like Hock Tan, it would be wonderful. Broadcom investors must be very reassured.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Broadcom on Hock Tan's cost-cutting discipline; investors should feel reassured.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "If every CEO were like Hock Tan, it would be wonderful. Broadcom investors must be very reassured. Imagine having a CEO who is so focused on extreme cost-cutting. https://t.co/xS13HUgKJ9\n\n---\n\nhttps://t.co/dwfAoZz7tb", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02707558848142777, "ret_signed_m1d": -0.02707558848142777, "alpha_spy_m1d": -0.030088918876853832, "alpha_c_m1d": -0.015869597378209366, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0129393534410156, "ret_signed_p1d": 0.0129393534410156, "alpha_spy_p1d": 0.013802498407977204, "alpha_c_p1d": 0.011718340556837736, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.1528048379439273, "ret_signed_p1w": -0.1528048379439273, "alpha_spy_p1w": -0.148562740371463, "alpha_c_p1w": -0.11034116619206036, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": -0.1412920714233059, "ret_signed_p1m": -0.1412920714233059, "alpha_spy_p1m": -0.1562478335601989, "alpha_c_p1m": -0.19630800481082344, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": -0.16876916677775078, "ret_signed_p3m": -0.16876916677775078, "alpha_spy_p3m": -0.16832039963483247, "alpha_c_p3m": -0.24482448115326705, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0802, -0.1049, -0.1043, -0.0938, -0.104, -0.1384, -0.1405, -0.1415, -0.1553, -0.155, -0.1541, -0.1621, -0.166, -0.1825, -0.1775, -0.1688, -0.1569, -0.1564, -0.1636, -0.1613, -0.1386, -0.1398, -0.1907, -0.1107, -0.142, -0.1241, -0.1171, -0.1291, -0.1293, -0.1457, -0.1516, -0.1416, -0.1171, -0.0974, -0.0701, -0.0377, -0.0614, -0.0785, -0.0961, -0.1226, -0.105, -0.1135, -0.1288, -0.1065, -0.1225, -0.1144, -0.1524, -0.1462, -0.1457, -0.1511, -0.1164, -0.1353, -0.1518, -0.0577, -0.0401, -0.0088, -0.0088, 0.0046, -0.0374, -0.0487, -0.0511, -0.05, -0.0271, 0.0, 0.0129, 0.0296, 0.0131, -0.1026, -0.1528, -0.1491, -0.1872, -0.1776, -0.1514, -0.1471, -0.1274, -0.1252, -0.1252, -0.1204, -0.1273, -0.1261, -0.1355, -0.1355, -0.1317, -0.1422, -0.1413, -0.142, -0.1695, -0.1383, -0.1202, -0.1142, -0.151, -0.1432, -0.1215, -0.1215, -0.1692, -0.1787, -0.187, -0.2005, -0.1886, -0.1687, -0.1676, -0.1739, -0.1724, -0.1729, -0.1998, -0.2305, -0.2244, -0.1684, -0.1409, -0.1496, -0.1438, -0.1728, -0.1878, -0.1878, -0.1693, -0.1669, -0.1657, -0.1691, -0.1748, -0.187, -0.1699, -0.1964, -0.2018, -0.2036, -0.2161, -0.2068, -0.1688, -0.1745, -0.1363, -0.1443, -0.1468, -0.1608, -0.1953, -0.1884, -0.1974, -0.2108, -0.2011, -0.2244, -0.1927, -0.2033, -0.202, -0.2255, -0.2474, -0.2655, -0.2252, -0.2153, -0.2126, -0.2126, -0.2129, -0.164, -0.1223, -0.1116, -0.07, -0.0494, -0.0468, -0.0069, -0.0026, 0.0176, 0.0003, 0.0067, 0.058, 0.0512, 0.0582, 0.0468, 0.0008, 0.0149, 0.0449, 0.0545, 0.0426, 0.0697, 0.0649, 0.0327, 0.0764, 0.0724, 0.0496, 0.0433, 0.1009, 0.0643, 0.0531, 0.029, 0.0457, 0.0377, 0.0367, 0.0367, 0.0564, 0.056, 0.0678, 0.1183, null, null]}
{"unit_id": "thread:1991703663396651183", "ticker_idx": 0, "ts": "2025-11-21T03:02:12+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' DRAM average selling price (ASP) will rise by the high teens percent compared with the previous quarter", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "DRAM super cycle driving high-teens ASP increase for Samsung and high-single-digit for SK Hynix in Q4; profitability revisions higher expected, with prices rising into Q1 next year.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics Eyes High-Teens Increase in DRAM ASP in Q4\n\nSamsung Electronics and SK hynix are fully enjoying the benefits of the DRAM “super cycle” triggered by the AI industry. As customer orders for high value-added and legacy DRAM are surging, there are projections that in the fourth quarter Samsung Electronics’ DRAM average selling price (ASP) will rise by the high teens percent compared with the previous quarter, while SK hynix’s will increase by the high single digits.\n\nAccording to industry sources on the 21st, the DRAM ASP of major domestic memory suppliers in the fourth quarter of this year is expected to increase significantly compared with earlier expectations.\n\nRecently, demand in the memory market has been surging as global big tech companies aggressively expand their AI infrastructure. In particular, orders for high value-added products such as server DRAM and HBM (high-bandwidth memory) have been brisk.\n\nCommodity DRAM used in PCs and smartphones is also suffering from a severe supply-demand imbalance. This is the result of major memory suppliers allocating most of their DRAM production capacity to HBM and focusing on conversion investments instead of building new mass-production lines. In addition, as memory suppliers sharply reduce the share of legacy products such as DDR4, the prices of these DRAMs are also skyrocketing.\n\nIn response, major IT companies are pursuing strategies that prioritize securing memory supply even at the cost of higher expenses. Chinese companies such as Xiaomi and Alibaba have accepted price increases of more than 50% compared with the previous quarter, while Lenovo has stated that it has signed long-term supply contracts for next year’s memory to stabilize its supply chain.\n\nA semiconductor industry official said, “Price negotiations for the fourth quarter have progressed to a considerable extent, but since the timing of contract signings differs by customer, Samsung Electronics and SK hynix are expected to continue signing memory supply contracts through the end of this year,” adding, “In particular, Samsung Electronics is known to be pursuing a more aggressive price hike strategy than its competitors in commodity DRAM.”\n\nThere is also a very high likelihood that the increase in DRAM ASP for the domestic memory industry in the fourth quarter of this year will exceed earlier expectations. Previously, Samsung Electronics disclosed that its DRAM ASP in the third quarter of this year rose by the mid-teens percent compared with the previous quarter, while SK hynix’s increased by the mid-single digits.\n\nIn the fourth quarter, Samsung Electronics is expected to see an ASP increase in the high teens percent. At the time of last month’s third-quarter conference call, the increase was estimated to be in the low- to mid-teens percent, but reflecting contracts recently concluded, the prevailing view in the industry is that an upward revision of the increase is inevitable.\n\nSK hynix is expected to see an increase in the high single digits. Because the sales proportion of commodity DRAM is relatively small, its price uptrend is structurally bound to be more moderate than Samsung Electronics’.\n\nAnother industry official said, “Although the contracts have not yet been finalized, given the current mood, the profitability of the two companies’ DRAM businesses will need to be revised higher,” adding, “Since prices are expected to continue rising in the first quarter of next year as well, the key will be how much the two companies choose to moderate the pace.”\n\n---\n\nhttps://t.co/4oHiuNRlMA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06118147784101491, "ret_signed_m1d": 0.06118147784101491, "alpha_spy_m1d": 0.07104445351771893, "alpha_c_m1d": 0.06502485721903084, "bench_spy_m1d": -0.009862975676704022, "bench_c_m1d": -0.003843379378015932, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020042229719216653, "ret_signed_p1d": 0.020042229719216653, "alpha_spy_p1d": 0.005323722657036178, "alpha_c_p1d": -0.003786341581340169, "bench_spy_p1d": 0.014718507062180475, "bench_c_p1d": 0.023828571300556822, "ret_p1w": 0.06012660617529764, "ret_signed_p1w": 0.06012660617529764, "alpha_spy_p1w": 0.02316324363761635, "alpha_c_p1w": 0.01246935156149509, "bench_spy_p1w": 0.03696336253768129, "bench_c_p1w": 0.04765725461380255, "ret_p1m": 0.16561186415290985, "ret_signed_p1m": 0.16561186415290985, "alpha_spy_p1m": 0.1233928366656154, "alpha_c_p1m": 0.10133539891492771, "bench_spy_p1m": 0.04221902748729445, "bench_c_p1m": 0.06427646523798214, "ret_p3m": 0.9206839737427239, "ret_signed_p3m": 0.9206839737427239, "alpha_spy_p3m": 0.8762430686547593, "alpha_c_p3m": 0.8875674222997765, "bench_spy_p3m": 0.04444090508796461, "bench_c_p3m": 0.03311655144294745, "ret_p6m": 1.9847092568059015, "ret_signed_p6m": 1.9847092568059015, "alpha_spy_p6m": 1.8575134470314096, "alpha_c_p6m": 1.7047562797326585, "bench_spy_p6m": 0.12719580977449185, "bench_c_p6m": 0.27995297707324296, "price_path": [-0.2617, -0.2586, -0.2691, -0.268, -0.2901, -0.2743, -0.267, -0.2638, -0.2701, -0.2638, -0.2491, -0.2376, -0.2292, -0.2082, -0.1966, -0.1662, -0.1788, -0.1567, -0.1567, -0.1231, -0.1105, -0.1032, -0.0958, -0.1252, -0.1118, -0.115, -0.0928, -0.0533, -0.0533, -0.0533, -0.0533, -0.0533, -0.0533, -0.0042, -0.0158, -0.0338, 0.0021, 0.0306, 0.0327, 0.0348, 0.0285, 0.0401, 0.0179, 0.0422, 0.0759, 0.0496, 0.0601, 0.0981, 0.134, 0.1719, 0.1065, 0.0612, 0.0464, 0.0327, 0.0612, 0.0918, 0.0876, 0.0844, 0.0253, 0.0612, 0.0316, 0.0179, 0.0612, 0.0, 0.02, 0.0475, 0.0844, 0.0918, 0.0601, 0.0633, 0.0907, 0.1023, 0.1086, 0.1435, 0.1551, 0.1435, 0.1392, 0.1319, 0.1487, 0.1055, 0.0844, 0.1382, 0.135, 0.1213, 0.1656, 0.1762, 0.1719, 0.1719, 0.2342, 0.2667, 0.2709, 0.2709, 0.2709, 0.3621, 0.4638, 0.4723, 0.4946, 0.4713, 0.4734, 0.4713, 0.4585, 0.4872, 0.5253, 0.5783, 0.5826, 0.5391, 0.5847, 0.6143, 0.6122, 0.6122, 0.6907, 0.7214, 0.7034, 0.7013, 0.5942, 0.7755, 0.7924, 0.6885, 0.6811, 0.7638, 0.7574, 0.7786, 0.8931, 0.9207, 0.9207, 0.9207, 0.9207, 1.014, 1.015, 1.0458, 1.12, 1.1571, 1.3108, 1.2949, 1.2949, 1.068, 0.8253, 1.0309, 0.9949, 0.8391, 0.9917, 1.014, 0.9917, 0.9451, 1.0002, 1.0553, 1.2101, 1.1253, 1.1136, 0.9747, 1.0108, 1.0034, 0.909, 0.909, 0.8726, 0.776, 1.0144, 0.8949, 0.9778, 1.0511, 1.0872, 1.2359, 1.1668, 1.1881, 1.135, 1.1934, 1.2412, 1.3102, 1.2943, 1.2784, 1.3262, 1.3102, 1.3846, 1.3315, 1.3846, 1.358, 1.4005, 1.3421, 1.3421, 1.4696, 1.4696, 1.8254, 1.8838, 1.8519, 2.0325, 1.9635, 2.0166, 2.144, 1.8732, 1.9847]}
{"unit_id": "thread:1991703663396651183", "ticker_idx": 1, "ts": "2025-11-21T03:02:12+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix's will increase by the high single digits", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "DRAM super cycle driving high-teens ASP increase for Samsung and high-single-digit for SK Hynix in Q4; profitability revisions higher expected, with prices rising into Q1 next year.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics Eyes High-Teens Increase in DRAM ASP in Q4\n\nSamsung Electronics and SK hynix are fully enjoying the benefits of the DRAM “super cycle” triggered by the AI industry. As customer orders for high value-added and legacy DRAM are surging, there are projections that in the fourth quarter Samsung Electronics’ DRAM average selling price (ASP) will rise by the high teens percent compared with the previous quarter, while SK hynix’s will increase by the high single digits.\n\nAccording to industry sources on the 21st, the DRAM ASP of major domestic memory suppliers in the fourth quarter of this year is expected to increase significantly compared with earlier expectations.\n\nRecently, demand in the memory market has been surging as global big tech companies aggressively expand their AI infrastructure. In particular, orders for high value-added products such as server DRAM and HBM (high-bandwidth memory) have been brisk.\n\nCommodity DRAM used in PCs and smartphones is also suffering from a severe supply-demand imbalance. This is the result of major memory suppliers allocating most of their DRAM production capacity to HBM and focusing on conversion investments instead of building new mass-production lines. In addition, as memory suppliers sharply reduce the share of legacy products such as DDR4, the prices of these DRAMs are also skyrocketing.\n\nIn response, major IT companies are pursuing strategies that prioritize securing memory supply even at the cost of higher expenses. Chinese companies such as Xiaomi and Alibaba have accepted price increases of more than 50% compared with the previous quarter, while Lenovo has stated that it has signed long-term supply contracts for next year’s memory to stabilize its supply chain.\n\nA semiconductor industry official said, “Price negotiations for the fourth quarter have progressed to a considerable extent, but since the timing of contract signings differs by customer, Samsung Electronics and SK hynix are expected to continue signing memory supply contracts through the end of this year,” adding, “In particular, Samsung Electronics is known to be pursuing a more aggressive price hike strategy than its competitors in commodity DRAM.”\n\nThere is also a very high likelihood that the increase in DRAM ASP for the domestic memory industry in the fourth quarter of this year will exceed earlier expectations. Previously, Samsung Electronics disclosed that its DRAM ASP in the third quarter of this year rose by the mid-teens percent compared with the previous quarter, while SK hynix’s increased by the mid-single digits.\n\nIn the fourth quarter, Samsung Electronics is expected to see an ASP increase in the high teens percent. At the time of last month’s third-quarter conference call, the increase was estimated to be in the low- to mid-teens percent, but reflecting contracts recently concluded, the prevailing view in the industry is that an upward revision of the increase is inevitable.\n\nSK hynix is expected to see an increase in the high single digits. Because the sales proportion of commodity DRAM is relatively small, its price uptrend is structurally bound to be more moderate than Samsung Electronics’.\n\nAnother industry official said, “Although the contracts have not yet been finalized, given the current mood, the profitability of the two companies’ DRAM businesses will need to be revised higher,” adding, “Since prices are expected to continue rising in the first quarter of next year as well, the key will be how much the two companies choose to moderate the pace.”\n\n---\n\nhttps://t.co/4oHiuNRlMA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09596927285748036, "ret_signed_m1d": 0.09596927285748036, "alpha_spy_m1d": 0.10583224853418438, "alpha_c_m1d": 0.09912758386555576, "bench_spy_m1d": -0.009862975676704022, "bench_c_m1d": -0.0031583110080753984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001919393877390374, "ret_signed_p1d": -0.001919393877390374, "alpha_spy_p1d": -0.01663790093957085, "alpha_c_p1d": -0.04175004450335018, "bench_spy_p1d": 0.014718507062180475, "bench_c_p1d": 0.03983065062595981, "ret_p1w": 0.018003016009944917, "ret_signed_p1w": 0.018003016009944917, "alpha_spy_p1w": -0.01896034652773637, "alpha_c_p1w": -0.06217971115364174, "bench_spy_p1w": 0.03696336253768129, "bench_c_p1w": 0.08018272716358665, "ret_p1m": 0.11404103411863509, "ret_signed_p1m": 0.11404103411863509, "alpha_spy_p1m": 0.07182200663134064, "alpha_c_p1m": 0.007674747575056262, "bench_spy_p1m": 0.04221902748729445, "bench_c_p1m": 0.10636628654357883, "ret_p3m": 0.6902691427707752, "ret_signed_p3m": 0.6902691427707752, "alpha_spy_p3m": 0.6458282376828106, "alpha_c_p3m": 0.4213467560086668, "bench_spy_p3m": 0.04444090508796461, "bench_c_p3m": 0.26892238676210845, "ret_p6m": 2.5407206150288184, "ret_signed_p6m": 2.5407206150288184, "alpha_spy_p6m": 2.4135248052543266, "alpha_c_p6m": 1.8608402679609732, "bench_spy_p6m": 0.12719580977449185, "bench_c_p6m": 0.6798803470678452, "price_path": [-0.4988, -0.5017, -0.4846, -0.4837, -0.5086, -0.5, -0.4962, -0.4904, -0.475, -0.4683, -0.4472, -0.4165, -0.4107, -0.3695, -0.3647, -0.3321, -0.3599, -0.3177, -0.3177, -0.3263, -0.3071, -0.3138, -0.3157, -0.3541, -0.3301, -0.333, -0.309, -0.2409, -0.2409, -0.2409, -0.2409, -0.2409, -0.2409, -0.1785, -0.2035, -0.2102, -0.1891, -0.1315, -0.1065, -0.0681, -0.0806, -0.0758, -0.0816, -0.0211, 0.0269, 0.0, 0.071, 0.0902, 0.0729, 0.19, 0.1248, 0.1113, 0.1382, 0.1132, 0.1631, 0.1881, 0.1843, 0.1747, 0.0749, 0.1631, 0.094, 0.0787, 0.096, 0.0, -0.0019, -0.0038, 0.0058, 0.0449, 0.018, 0.0334, 0.0718, 0.0603, 0.0411, 0.0449, 0.1083, 0.0872, 0.1275, 0.0852, 0.0968, 0.0641, 0.018, 0.0583, 0.0603, 0.0507, 0.114, 0.1217, 0.1294, 0.1294, 0.1505, 0.2293, 0.2504, 0.2504, 0.2504, 0.3004, 0.3368, 0.3945, 0.4252, 0.4521, 0.429, 0.4386, 0.4175, 0.4252, 0.4386, 0.4521, 0.4675, 0.4271, 0.4214, 0.4502, 0.4732, 0.4137, 0.5366, 0.6154, 0.6538, 0.746, 0.5942, 0.7421, 0.7287, 0.6173, 0.6115, 0.7037, 0.6826, 0.6519, 0.7056, 0.6903, 0.6903, 0.6903, 0.6903, 0.7172, 0.8228, 0.8266, 0.9304, 0.9553, 1.1148, 1.0417, 1.0417, 0.8069, 0.6337, 0.8108, 0.7781, 0.6087, 0.805, 0.8377, 0.7896, 0.7511, 0.8743, 0.8666, 1.0321, 0.9493, 0.9378, 0.7954, 0.8974, 0.9147, 0.7954, 0.7954, 0.6799, 0.5529, 0.7261, 0.5972, 0.6857, 0.7049, 0.7627, 0.9878, 0.9205, 0.9763, 1.0013, 1.1225, 1.186, 1.2226, 1.1706, 1.2437, 1.3553, 1.3534, 1.3573, 1.3515, 1.4862, 1.5016, 1.4881, 1.4747, 1.4747, 1.7845, 1.7845, 2.0808, 2.1828, 2.2444, 2.6177, 2.5311, 2.8024, 2.7909, 2.5003, 2.5407]}
{"unit_id": "thread:1991703663396651183", "ticker_idx": 2, "ts": "2025-11-21T03:02:12+00:00", "symbol": "DRAM", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the profitability of the two companies' DRAM businesses will need to be revised higher", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "DRAM super cycle driving high-teens ASP increase for Samsung and high-single-digit for SK Hynix in Q4; profitability revisions higher expected, with prices rising into Q1 next year.", "resolved_tickers": ["005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: Samsung Electronics + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics Eyes High-Teens Increase in DRAM ASP in Q4\n\nSamsung Electronics and SK hynix are fully enjoying the benefits of the DRAM “super cycle” triggered by the AI industry. As customer orders for high value-added and legacy DRAM are surging, there are projections that in the fourth quarter Samsung Electronics’ DRAM average selling price (ASP) will rise by the high teens percent compared with the previous quarter, while SK hynix’s will increase by the high single digits.\n\nAccording to industry sources on the 21st, the DRAM ASP of major domestic memory suppliers in the fourth quarter of this year is expected to increase significantly compared with earlier expectations.\n\nRecently, demand in the memory market has been surging as global big tech companies aggressively expand their AI infrastructure. In particular, orders for high value-added products such as server DRAM and HBM (high-bandwidth memory) have been brisk.\n\nCommodity DRAM used in PCs and smartphones is also suffering from a severe supply-demand imbalance. This is the result of major memory suppliers allocating most of their DRAM production capacity to HBM and focusing on conversion investments instead of building new mass-production lines. In addition, as memory suppliers sharply reduce the share of legacy products such as DDR4, the prices of these DRAMs are also skyrocketing.\n\nIn response, major IT companies are pursuing strategies that prioritize securing memory supply even at the cost of higher expenses. Chinese companies such as Xiaomi and Alibaba have accepted price increases of more than 50% compared with the previous quarter, while Lenovo has stated that it has signed long-term supply contracts for next year’s memory to stabilize its supply chain.\n\nA semiconductor industry official said, “Price negotiations for the fourth quarter have progressed to a considerable extent, but since the timing of contract signings differs by customer, Samsung Electronics and SK hynix are expected to continue signing memory supply contracts through the end of this year,” adding, “In particular, Samsung Electronics is known to be pursuing a more aggressive price hike strategy than its competitors in commodity DRAM.”\n\nThere is also a very high likelihood that the increase in DRAM ASP for the domestic memory industry in the fourth quarter of this year will exceed earlier expectations. Previously, Samsung Electronics disclosed that its DRAM ASP in the third quarter of this year rose by the mid-teens percent compared with the previous quarter, while SK hynix’s increased by the mid-single digits.\n\nIn the fourth quarter, Samsung Electronics is expected to see an ASP increase in the high teens percent. At the time of last month’s third-quarter conference call, the increase was estimated to be in the low- to mid-teens percent, but reflecting contracts recently concluded, the prevailing view in the industry is that an upward revision of the increase is inevitable.\n\nSK hynix is expected to see an increase in the high single digits. Because the sales proportion of commodity DRAM is relatively small, its price uptrend is structurally bound to be more moderate than Samsung Electronics’.\n\nAnother industry official said, “Although the contracts have not yet been finalized, given the current mood, the profitability of the two companies’ DRAM businesses will need to be revised higher,” adding, “Since prices are expected to continue rising in the first quarter of next year as well, the key will be how much the two companies choose to moderate the pace.”\n\n---\n\nhttps://t.co/4oHiuNRlMA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07857537534924763, "ret_signed_m1d": 0.07857537534924763, "alpha_spy_m1d": 0.08843835102595166, "alpha_c_m1d": 0.08241875472726357, "bench_spy_m1d": -0.009862975676704022, "bench_c_m1d": -0.003843379378015932, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00906141792091314, "ret_signed_p1d": 0.00906141792091314, "alpha_spy_p1d": -0.005657089141267335, "alpha_c_p1d": -0.014767153379643683, "bench_spy_p1d": 0.014718507062180475, "bench_c_p1d": 0.023828571300556822, "ret_p1w": 0.03906481109262128, "ret_signed_p1w": 0.03906481109262128, "alpha_spy_p1w": 0.0021014485549399886, "alpha_c_p1w": -0.00859244352118127, "bench_spy_p1w": 0.03696336253768129, "bench_c_p1w": 0.04765725461380255, "ret_p1m": 0.13982644913577247, "ret_signed_p1m": 0.13982644913577247, "alpha_spy_p1m": 0.09760742164847802, "alpha_c_p1m": 0.07554998389779033, "bench_spy_p1m": 0.04221902748729445, "bench_c_p1m": 0.06427646523798214, "ret_p3m": 0.8054765582567496, "ret_signed_p3m": 0.8054765582567496, "alpha_spy_p3m": 0.761035653168785, "alpha_c_p3m": 0.7723600068138021, "bench_spy_p3m": 0.04444090508796461, "bench_c_p3m": 0.03311655144294745, "ret_p6m": 2.26271493591736, "ret_signed_p6m": 2.26271493591736, "alpha_spy_p6m": 2.135519126142868, "alpha_c_p6m": 1.982761958844117, "bench_spy_p6m": 0.12719580977449185, "bench_c_p6m": 0.27995297707324296, "price_path": [-0.3803, -0.3801, -0.3769, -0.3759, -0.3994, -0.3872, -0.3816, -0.3771, -0.3726, -0.3661, -0.3482, -0.327, -0.32, -0.2888, -0.2807, -0.2491, -0.2693, -0.2372, -0.2372, -0.2247, -0.2088, -0.2085, -0.2058, -0.2397, -0.221, -0.224, -0.2009, -0.1471, -0.1471, -0.1471, -0.1471, -0.1471, -0.1471, -0.0914, -0.1096, -0.122, -0.0935, -0.0504, -0.0369, -0.0167, -0.0261, -0.0179, -0.0318, 0.0105, 0.0514, 0.0248, 0.0656, 0.0942, 0.1035, 0.181, 0.1157, 0.0863, 0.0923, 0.073, 0.1122, 0.1399, 0.1359, 0.1295, 0.0501, 0.1122, 0.0628, 0.0483, 0.0786, 0.0, 0.0091, 0.0218, 0.0451, 0.0683, 0.0391, 0.0483, 0.0813, 0.0813, 0.0749, 0.0942, 0.1317, 0.1153, 0.1334, 0.1085, 0.1227, 0.0848, 0.0512, 0.0983, 0.0976, 0.086, 0.1398, 0.1489, 0.1507, 0.1507, 0.1924, 0.248, 0.2607, 0.2607, 0.2607, 0.3312, 0.4003, 0.4334, 0.4599, 0.4617, 0.4512, 0.455, 0.438, 0.4562, 0.482, 0.5152, 0.525, 0.4831, 0.503, 0.5323, 0.5427, 0.513, 0.6136, 0.6684, 0.6786, 0.7236, 0.5942, 0.7588, 0.7606, 0.6529, 0.6463, 0.7338, 0.72, 0.7153, 0.7994, 0.8055, 0.8055, 0.8055, 0.8055, 0.8656, 0.9189, 0.9362, 1.0252, 1.0562, 1.2128, 1.1683, 1.1683, 0.9375, 0.7295, 0.9208, 0.8865, 0.7239, 0.8984, 0.9258, 0.8907, 0.8481, 0.9372, 0.9609, 1.1211, 1.0373, 1.0257, 0.8851, 0.9541, 0.959, 0.8522, 0.8522, 0.7763, 0.6644, 0.8703, 0.746, 0.8317, 0.878, 0.9249, 1.1118, 1.0436, 1.0822, 1.0681, 1.1579, 1.2136, 1.2664, 1.2325, 1.2611, 1.3408, 1.3318, 1.3709, 1.3415, 1.4354, 1.4298, 1.4443, 1.4084, 1.4084, 1.627, 1.627, 1.9531, 2.0333, 2.0482, 2.3251, 2.2473, 2.4095, 2.4675, 2.1867, 2.2627]}
{"unit_id": "orig:1989278385353162823", "ticker_idx": 0, "ts": "2025-11-14T10:25:00+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Overweight and Buy recommendations on the semiconductor sector", "tweet_type": "analysis", "timeline": "quarters", "summary": "KIS maintains Buy/Overweight on Korean semis: DRAM supply shortfall in 2026, HBM margins 65–70%, WFE-driven equipment earnings rising through 2026–2027.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "[KIS Chae Min-sook / Hwang Joon-tae] Semiconductor Industry Brief: Comments on Semiconductor Stock Price Declines\n\n● Fundamentals remain solid; we need to wait for the storm to pass\n    \n• Following Kioxia’s earnings shock the previous day, share price pressure has spread to memory-related names that had surged in a short period, and semiconductor stocks are falling on macro issues such as FX as well.\n    \n• Compared to the previous day: Samsung Electronics -5.4%, SK hynix -8.5%, PSK -7.1%, Eugene Technology -9.1%, with share prices across semiconductor names declining.\n    \n• In 4Q, server DRAM prices are up about 30–40% QoQ. For some Chinese customers, contract prices have been agreed at levels 70% higher QoQ.\n    \n• Customers are betting on memory price increases even in 2026. Large customers are negotiating prices for six-month periods instead of on a quarterly basis. This reflects moves to lengthen the contract price application period in anticipation of price hikes every quarter.\n    \n• DRAM demand bit growth in 2026 is forecast at 24.6%, while the three DRAM players’ supply bit growth is only 21%; as such, ASPs are expected to continue rising through supply shortages during 2026.\n    \n• For NAND as well, the supply–demand ratio will remain below 100% throughout 2026, with supply shortages persisting.\n    \n• NAND has lower profitability than DRAM, and bit growth from node migrations is larger. As a result, NAND suppliers are expected to maintain a conservative investment stance on capacity expansion, concerned that the cycle could turn down if they add too much capacity.\n    \n• As of 4Q25, server DRAM operating margins will reach about 60–65%. For SK hynix, HBM operating margins are around 65–70%. Solid earnings are a factor that continues to ease valuation burdens.\n    \n• Memory capex is expected to increase, centered on DRAM. With growth in WFE, the earnings of materials, components, and equipment companies are expected to keep rising from 2H25 through 2026.\n    \n• In the second half of 2027, large-scale expansion investments such as Samsung Electronics’ P5 and SK hynix’s Yongin Y1 are scheduled. Front-end equipment companies’ earnings are expected to remain on an uptrend in 2027 following 2026.\n    \n• Investor sentiment is expected to recover starting with NVIDIA’s earnings on 11/19.\n    \n• We maintain Overweight and Buy recommendations on the semiconductor sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05761317038957614, "ret_signed_m1d": 0.05761317038957614, "alpha_spy_m1d": 0.057449557525201866, "alpha_c_m1d": 0.06302709370298476, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03497941620794953, "ret_signed_p1d": 0.03497941620794953, "alpha_spy_p1d": 0.044295940138911694, "alpha_c_p1d": 0.05063102623894755, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.024691404986047938, "ret_signed_p1w": -0.024691404986047938, "alpha_spy_p1w": -0.005492974646394533, "alpha_c_p1w": 0.027191265659056163, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.07818927376678131, "ret_signed_p1m": 0.07818927376678131, "alpha_spy_p1m": 0.0650927536872532, "alpha_c_p1m": 0.09050924685115647, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": 0.7347294453035076, "ret_signed_p3m": 0.7347294453035076, "alpha_spy_p3m": 0.7018767860417772, "alpha_c_p3m": 0.7408938922948921, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 1.957630231150651, "ret_signed_p6m": 1.957630231150651, "alpha_spy_p6m": 1.8511019956445949, "alpha_c_p6m": 1.7195854032114168, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.283, -0.2779, -0.2769, -0.2687, -0.2677, -0.28, -0.2769, -0.2871, -0.2861, -0.3076, -0.2923, -0.2851, -0.282, -0.2882, -0.282, -0.2677, -0.2564, -0.2482, -0.2277, -0.2165, -0.1868, -0.199, -0.1775, -0.1775, -0.1448, -0.1325, -0.1253, -0.1181, -0.1468, -0.1337, -0.1368, -0.1152, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0288, -0.0401, -0.0576, -0.0226, 0.0051, 0.0072, 0.0093, 0.0031, 0.0144, -0.0072, 0.0165, 0.0494, 0.0237, 0.034, 0.071, 0.106, 0.143, 0.0792, 0.035, 0.0206, 0.0072, 0.035, 0.0648, 0.0607, 0.0576, 0.0, 0.035, 0.0062, -0.0072, 0.035, -0.0247, -0.0051, 0.0216, 0.0576, 0.0648, 0.034, 0.037, 0.0638, 0.0751, 0.0813, 0.1152, 0.1265, 0.1152, 0.1111, 0.1039, 0.1204, 0.0782, 0.0576, 0.1101, 0.107, 0.0936, 0.1368, 0.1471, 0.143, 0.143, 0.2037, 0.2354, 0.2395, 0.2395, 0.2395, 0.3284, 0.4277, 0.436, 0.4577, 0.4349, 0.437, 0.4349, 0.4225, 0.4504, 0.4876, 0.5393, 0.5435, 0.5011, 0.5455, 0.5745, 0.5724, 0.5724, 0.6489, 0.6789, 0.6613, 0.6593, 0.5548, 0.7316, 0.7482, 0.6469, 0.6396, 0.7203, 0.7141, 0.7347, 0.8464, 0.8733, 0.8733, 0.8733, 0.8733, 0.9642, 0.9653, 0.9952, 1.0676, 1.1038, 1.2537, 1.2382, 1.2382, 1.017, 0.7802, 0.9808, 0.9456, 0.7937, 0.9425, 0.9642, 0.9425, 0.897, 0.9508, 1.0046, 1.1555, 1.0728, 1.0614, 0.926, 0.9611, 0.9539, 0.8619, 0.8619, 0.8264, 0.7321, 0.9647, 0.8481, 0.9289, 1.0004, 1.0356, 1.1807, 1.1133, 1.1341, 1.0823, 1.1392, 1.1858, 1.2532, 1.2376, 1.2221, 1.2687, 1.2532, 1.3257, 1.2739, 1.3257, 1.2998, 1.3412, 1.2843, 1.2843, 1.4086, 1.4086, 1.7556, 1.8126, 1.7815, 1.9576]}
{"unit_id": "orig:1989278385353162823", "ticker_idx": 1, "ts": "2025-11-14T10:25:00+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain Overweight and Buy recommendations on the semiconductor sector", "tweet_type": "analysis", "timeline": "quarters", "summary": "KIS maintains Buy/Overweight on Korean semis: DRAM supply shortfall in 2026, HBM margins 65–70%, WFE-driven equipment earnings rising through 2026–2027.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[KIS Chae Min-sook / Hwang Joon-tae] Semiconductor Industry Brief: Comments on Semiconductor Stock Price Declines\n\n● Fundamentals remain solid; we need to wait for the storm to pass\n    \n• Following Kioxia’s earnings shock the previous day, share price pressure has spread to memory-related names that had surged in a short period, and semiconductor stocks are falling on macro issues such as FX as well.\n    \n• Compared to the previous day: Samsung Electronics -5.4%, SK hynix -8.5%, PSK -7.1%, Eugene Technology -9.1%, with share prices across semiconductor names declining.\n    \n• In 4Q, server DRAM prices are up about 30–40% QoQ. For some Chinese customers, contract prices have been agreed at levels 70% higher QoQ.\n    \n• Customers are betting on memory price increases even in 2026. Large customers are negotiating prices for six-month periods instead of on a quarterly basis. This reflects moves to lengthen the contract price application period in anticipation of price hikes every quarter.\n    \n• DRAM demand bit growth in 2026 is forecast at 24.6%, while the three DRAM players’ supply bit growth is only 21%; as such, ASPs are expected to continue rising through supply shortages during 2026.\n    \n• For NAND as well, the supply–demand ratio will remain below 100% throughout 2026, with supply shortages persisting.\n    \n• NAND has lower profitability than DRAM, and bit growth from node migrations is larger. As a result, NAND suppliers are expected to maintain a conservative investment stance on capacity expansion, concerned that the cycle could turn down if they add too much capacity.\n    \n• As of 4Q25, server DRAM operating margins will reach about 60–65%. For SK hynix, HBM operating margins are around 65–70%. Solid earnings are a factor that continues to ease valuation burdens.\n    \n• Memory capex is expected to increase, centered on DRAM. With growth in WFE, the earnings of materials, components, and equipment companies are expected to keep rising from 2H25 through 2026.\n    \n• In the second half of 2027, large-scale expansion investments such as Samsung Electronics’ P5 and SK hynix’s Yongin Y1 are scheduled. Front-end equipment companies’ earnings are expected to remain on an uptrend in 2027 following 2026.\n    \n• Investor sentiment is expected to recover starting with NVIDIA’s earnings on 11/19.\n    \n• We maintain Overweight and Buy recommendations on the semiconductor sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09285720201056402, "ret_signed_m1d": 0.09285720201056402, "alpha_spy_m1d": 0.09269358914618975, "alpha_c_m1d": 0.0933210309821717, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08214287073215654, "ret_signed_p1d": 0.08214287073215654, "alpha_spy_p1d": 0.0914593946631187, "alpha_c_p1d": 0.09570900296467622, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.06964281757401458, "ret_signed_p1w": -0.06964281757401458, "alpha_spy_p1w": -0.05044438723436118, "alpha_c_p1w": -0.015001954481206226, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.010005705267535414, "ret_signed_p1m": -0.010005705267535414, "alpha_spy_p1m": -0.02310222534706352, "alpha_c_p1m": -0.03296350898810585, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.536814195618829, "ret_signed_p3m": 0.536814195618829, "alpha_spy_p3m": 0.5039615363570986, "alpha_c_p3m": 0.33074253506319273, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.3657463705299318, "ret_signed_p6m": 2.3657463705299318, "alpha_spy_p6m": 2.2592181350238754, "alpha_c_p6m": 1.6899877568168509, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.531, -0.5444, -0.5631, -0.5524, -0.5373, -0.5337, -0.5364, -0.5205, -0.5196, -0.5429, -0.5348, -0.5312, -0.5259, -0.5116, -0.5054, -0.4857, -0.4571, -0.4518, -0.4134, -0.4089, -0.3786, -0.4045, -0.3652, -0.3652, -0.3732, -0.3554, -0.3616, -0.3634, -0.3991, -0.3768, -0.3795, -0.3571, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2357, -0.2589, -0.2652, -0.2455, -0.192, -0.1687, -0.133, -0.1446, -0.1402, -0.1455, -0.0893, -0.0446, -0.0696, -0.0036, 0.0143, -0.0018, 0.1071, 0.0464, 0.0339, 0.0589, 0.0357, 0.0821, 0.1054, 0.1018, 0.0929, 0.0, 0.0821, 0.0179, 0.0036, 0.0196, -0.0696, -0.0714, -0.0732, -0.0643, -0.0279, -0.0529, -0.0386, -0.0029, -0.0136, -0.0314, -0.0279, 0.0311, 0.0114, 0.049, 0.0097, 0.0204, -0.01, -0.0529, -0.0154, -0.0136, -0.0225, 0.0365, 0.0436, 0.0508, 0.0508, 0.0704, 0.1437, 0.1633, 0.1633, 0.1633, 0.2098, 0.2437, 0.2974, 0.3259, 0.351, 0.3295, 0.3385, 0.3188, 0.3259, 0.3385, 0.351, 0.3653, 0.3277, 0.3224, 0.3492, 0.3706, 0.3152, 0.4296, 0.5029, 0.5386, 0.6244, 0.4832, 0.6208, 0.6083, 0.5046, 0.4993, 0.5851, 0.5654, 0.5368, 0.5868, 0.5726, 0.5726, 0.5726, 0.5726, 0.5976, 0.6959, 0.6994, 0.7959, 0.8192, 0.9675, 0.8995, 0.8995, 0.6811, 0.52, 0.6847, 0.6542, 0.4967, 0.6793, 0.7097, 0.665, 0.6292, 0.7437, 0.7366, 0.8905, 0.8136, 0.8028, 0.6703, 0.7652, 0.7813, 0.6703, 0.6703, 0.5629, 0.4448, 0.6059, 0.4859, 0.5683, 0.5862, 0.6399, 0.8494, 0.7867, 0.8386, 0.8619, 0.9747, 1.0338, 1.0678, 1.0194, 1.0875, 1.1913, 1.1895, 1.1931, 1.1877, 1.3131, 1.3274, 1.3148, 1.3023, 1.3023, 1.5906, 1.5906, 1.8663, 1.9611, 2.0184, 2.3657]}
{"unit_id": "orig:1989278385353162823", "ticker_idx": 2, "ts": "2025-11-14T10:25:00+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "We maintain Overweight and Buy recommendations on the semiconductor sector. Fundamentals remain solid; we need to wait for the storm to pass", "tweet_type": "analysis", "timeline": "quarters", "summary": "KIS maintains Buy/Overweight on Korean semis: DRAM supply shortfall in 2026, HBM margins 65–70%, WFE-driven equipment earnings rising through 2026–2027.", "resolved_tickers": ["EWY"], "resolution_method": "etf", "resolution_reason": "Korea semiconductor sector proxied by EWY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "[KIS Chae Min-sook / Hwang Joon-tae] Semiconductor Industry Brief: Comments on Semiconductor Stock Price Declines\n\n● Fundamentals remain solid; we need to wait for the storm to pass\n    \n• Following Kioxia’s earnings shock the previous day, share price pressure has spread to memory-related names that had surged in a short period, and semiconductor stocks are falling on macro issues such as FX as well.\n    \n• Compared to the previous day: Samsung Electronics -5.4%, SK hynix -8.5%, PSK -7.1%, Eugene Technology -9.1%, with share prices across semiconductor names declining.\n    \n• In 4Q, server DRAM prices are up about 30–40% QoQ. For some Chinese customers, contract prices have been agreed at levels 70% higher QoQ.\n    \n• Customers are betting on memory price increases even in 2026. Large customers are negotiating prices for six-month periods instead of on a quarterly basis. This reflects moves to lengthen the contract price application period in anticipation of price hikes every quarter.\n    \n• DRAM demand bit growth in 2026 is forecast at 24.6%, while the three DRAM players’ supply bit growth is only 21%; as such, ASPs are expected to continue rising through supply shortages during 2026.\n    \n• For NAND as well, the supply–demand ratio will remain below 100% throughout 2026, with supply shortages persisting.\n    \n• NAND has lower profitability than DRAM, and bit growth from node migrations is larger. As a result, NAND suppliers are expected to maintain a conservative investment stance on capacity expansion, concerned that the cycle could turn down if they add too much capacity.\n    \n• As of 4Q25, server DRAM operating margins will reach about 60–65%. For SK hynix, HBM operating margins are around 65–70%. Solid earnings are a factor that continues to ease valuation burdens.\n    \n• Memory capex is expected to increase, centered on DRAM. With growth in WFE, the earnings of materials, components, and equipment companies are expected to keep rising from 2H25 through 2026.\n    \n• In the second half of 2027, large-scale expansion investments such as Samsung Electronics’ P5 and SK hynix’s Yongin Y1 are scheduled. Front-end equipment companies’ earnings are expected to remain on an uptrend in 2027 following 2026.\n    \n• Investor sentiment is expected to recover starting with NVIDIA’s earnings on 11/19.\n    \n• We maintain Overweight and Buy recommendations on the semiconductor sector.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0006373695986220396, "ret_signed_m1d": -0.0006373695986220396, "alpha_spy_m1d": -0.0008009824629963136, "alpha_c_m1d": -0.0008009824629963136, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": 0.00016361286437427403, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011792207731937121, "ret_signed_p1d": -0.011792207731937121, "alpha_spy_p1d": -0.0024756838009749593, "alpha_c_p1d": -0.0024756838009749593, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.009316523930962162, "ret_p1w": -0.05216179425467371, "ret_signed_p1w": -0.05216179425467371, "alpha_spy_p1w": -0.032963363915020305, "alpha_c_p1w": -0.032963363915020305, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.019198430339653405, "ret_p1m": -0.014872979390859964, "ret_signed_p1m": -0.014872979390859964, "alpha_spy_p1m": -0.02796949947038807, "alpha_c_p1m": -0.02796949947038807, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.013096520079528107, "ret_p3m": 0.41860555697451063, "ret_signed_p3m": 0.41860555697451063, "alpha_spy_p3m": 0.3857528977127802, "alpha_c_p3m": 0.3857528977127802, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.03285265926173042, "ret_p6m": 1.0947785416194642, "ret_signed_p6m": 1.0947785416194642, "alpha_spy_p6m": 0.988250306113408, "alpha_c_p6m": 0.988250306113408, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.10652823550605617, "price_path": [-0.241, -0.2419, -0.2434, -0.2191, -0.2273, -0.231, -0.2301, -0.2177, -0.2326, -0.2326, -0.236, -0.2254, -0.2237, -0.2178, -0.2118, -0.2041, -0.1857, -0.1728, -0.1643, -0.1587, -0.1428, -0.1519, -0.1421, -0.1496, -0.1335, -0.1347, -0.1497, -0.1513, -0.1621, -0.1476, -0.149, -0.13, -0.1126, -0.1092, -0.0993, -0.1145, -0.1056, -0.1137, -0.1425, -0.1117, -0.1215, -0.0952, -0.0664, -0.0596, -0.0393, -0.0583, -0.0497, -0.0431, -0.0221, 0.0019, 0.0034, 0.0294, 0.012, 0.0305, 0.0706, 0.0173, 0.0265, -0.0066, -0.0248, 0.0164, 0.0123, 0.019, -0.0006, 0.0, -0.0118, -0.027, -0.0379, -0.0595, -0.0522, -0.0407, -0.0461, -0.0274, -0.0274, -0.0346, -0.0373, -0.0219, -0.012, -0.024, 0.0018, 0.01, 0.0153, 0.0278, 0.0088, -0.0155, -0.0149, -0.0276, -0.0378, -0.0216, -0.0078, -0.0034, 0.0069, 0.0276, 0.0276, 0.041, 0.0723, 0.0712, 0.056, 0.056, 0.1103, 0.1387, 0.1689, 0.1626, 0.1578, 0.1828, 0.1832, 0.1681, 0.1909, 0.2062, 0.219, 0.219, 0.2054, 0.2563, 0.2599, 0.2814, 0.2744, 0.3328, 0.3639, 0.3523, 0.3296, 0.3136, 0.3504, 0.3063, 0.3057, 0.3546, 0.3692, 0.3518, 0.4186, 0.4208, 0.4552, 0.4552, 0.4195, 0.4439, 0.4687, 0.5411, 0.5131, 0.5701, 0.6173, 0.6338, 0.6442, 0.6026, 0.4375, 0.4596, 0.3658, 0.3766, 0.4543, 0.4153, 0.443, 0.3415, 0.3485, 0.4458, 0.4602, 0.4325, 0.4644, 0.3662, 0.4535, 0.3985, 0.3875, 0.3034, 0.3141, 0.2647, 0.3362, 0.371, 0.3346, 0.3346, 0.368, 0.3816, 0.5215, 0.513, 0.5069, 0.5341, 0.5933, 0.5742, 0.6019, 0.6546, 0.6303, 0.5945, 0.6923, 0.6357, 0.679, 0.7024, 0.6768, 0.6723, 0.7462, 0.7597, 0.7769, 0.8843, 0.9767, 0.9199, 1.066, 1.0948]}
{"unit_id": "orig:1988096168480698719", "ticker_idx": 0, "ts": "2025-11-11T04:07:18+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC gross margin expected to 'rise significantly'… TSMC's gross margin could reach the low-to-mid 60% range in the first half of 2026, exceeding current market expectations", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM thesis: TSMC N3 capacity extremely tight, hot-run premiums + 6-10% price hikes push gross margin to low-to-mid 60% in H1 2026, above consensus.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM: Production capacity extremely tight… customers place “urgent capacity expansion” orders; TSMC gross margin expected to “rise significantly”\n\nIn research published on November 10, J.P. Morgan pointed out that despite reports that Nvidia has asked TSMC to expand N3 capacity to 160,000 wafers per month (wfpm), TSMC’s actual capacity may only reach 140,000–145,000 wafers per month by the end of 2026. This means the shortage of the world’s most advanced chips will persist over the next two years.\n\nThis supply–demand imbalance foreshadows a dual impact. On the one hand, chipmakers that rely on leading-edge nodes may face growth bottlenecks; on the other, TSMC, which holds pricing power, could see its profitability improve markedly. J.P. Morgan believes supply tensions have already triggered “hot-runs,” which could push TSMC’s gross margin into the low-to-mid 60% range in the first half of 2026.\n\nProduction capacity “extremely tight”\nConfronted with urgent customer demand, TSMC has not opted to build new N3 fabs. According to J.P. Morgan’s supply-chain checks, TSMC tends to prioritize its new fabs in Kaohsiung (Fab 22) and Hsinchu (Fab 20) for next-generation process nodes such as N2 and A16. Analyst Gokul Hariharan notes that TSMC’s strategy is to encourage customers to move to leading nodes more quickly rather than inject excessive new capacity late in N3’s lifecycle.\n\nAccordingly, the main increment of N3 capacity in 2026 is expected to come from line conversions at the Tainan Fab 18. J.P. Morgan expects that as some accelerator demand shifts to N3, utilization of the N4 process will drop below 100% in the first half of 2026, at which point around 30,000 wafers per month of N4 lines could be converted into roughly 25,000 wafers per month of N3 lines.\n\nHowever, there are variables to this conversion. If Nvidia is permitted to ship China-bound GPUs such as the B30A to China, N4 demand could tighten, slowing the pace of N4-to-N3 conversion.\n\nMaximizing supply through multiple avenues\nBeyond line conversions, TSMC is trying other ways to “squeeze out” more capacity.\n\nThe report states that through “cross-fab collaboration,” TSMC is using idle N6/N7 capacity at Fab 14 to handle parts of the back end of line (BEOL) for N3. J.P. Morgan judges that while this approach is not highly efficient, given that utilization at the N6/N7 nodes could remain at a low level of around 65% in 2026, there is room to add an extra 5,000–10,000 wafers per month in the second half of 2026.\n\nAnother is “overseas support.” The much-watched U.S. Arizona fab (Fab 21) is expected to begin tool move-in for its second phase in the second quarter of 2026, but it will not be until early 2027 that it can supply about 10,000 wafers per month of N3 chips, meaning it will not ease the supply tightness in 2026. In aggregate, J.P. Morgan estimates TSMC’s total N3 capacity at around 140,000–145,000 wafers per month by the end of 2026, still leaving a gap with the reported 160,000-wafer demand.\n\nCustomers’ “urgent expansion orders”\nWith capacity extremely tight, major tech companies have already moved preemptively and jumped into a “capacity grab.” The demand-side lineup is star-studded, encompassing virtually all Big Tech.\n\nIn high-performance computing (HPC), beyond Nvidia’s Rubin and Rubin CPX, there are in-house designed ASICs such as Broadcom’s TPU v7, Amazon’s Trainium 3, Meta’s Iris, and Microsoft’s Maia 300.\n\nIn consumer electronics, Apple’s C2 baseband chip and the iPhone 17, as well as Qualcomm’s and MediaTek’s Android flagship SoCs, are slated to move to N3 at scale. With capacity already pre-booked by major customers, the report specifies that demand from customers such as cryptocurrency miners in 2026 will be “essentially unable to be met.”\n\nTSMC gross margin expected to “rise significantly”\nThe scarcity of capacity is feeding directly into TSMC’s profitability. According to J.P. Morgan’s supply-chain checks, as capacity becomes extremely tight, many customers are executing or planning “hot-runs” or “super hot-runs”—paying 50% to as much as 100% more per wafer in exchange for earlier delivery.\n\nThe report analyzes that although the total of such high-priced orders usually does not exceed 10% of total capacity, it is sufficient to have a clear positive impact on TSMC’s financial performance.\n\nJ.P. Morgan forecasts that if “hot-runs” persist and exchange rates remain stable, TSMC’s gross margin could reach the low-to-mid 60% range in the first half of 2026, exceeding current market expectations. In addition, TSMC’s plan to raise advanced-node pricing by 6%–10% starting in the first quarter of 2026 will further support its profit level.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006825996105350063, "ret_signed_m1d": 0.006825996105350063, "alpha_spy_m1d": 0.00910991756965529, "alpha_c_m1d": -0.015222180990246192, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006825996105350063, "ret_signed_p1d": 0.006825996105350063, "alpha_spy_p1d": 0.006269597498907986, "alpha_c_p1d": -0.005935695084483106, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.04095555725728539, "ret_signed_p1w": -0.04095555725728539, "alpha_spy_p1w": -0.007397828093361647, "alpha_c_p1w": 0.009635485761242402, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": 0.02730381667147408, "ret_signed_p1m": 0.02730381667147408, "alpha_spy_p1m": 0.020612653846917572, "alpha_c_p1m": -0.03835639036188376, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.2190672282151349, "ret_signed_p3m": 0.2190672282151349, "alpha_spy_p3m": 0.20492269668955032, "alpha_c_p3m": 0.07136812595033915, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.5459833160956304, "ret_signed_p6m": 0.5459833160956304, "alpha_spy_p6m": 0.46544383143164225, "alpha_c_p6m": -0.02493434668841954, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.2011, -0.1977, -0.1977, -0.1943, -0.2283, -0.2181, -0.2283, -0.2045, -0.2011, -0.1909, -0.2113, -0.2113, -0.2079, -0.2113, -0.2113, -0.2113, -0.1977, -0.1977, -0.1842, -0.1672, -0.157, -0.1434, -0.1468, -0.1263, -0.1365, -0.1229, -0.1365, -0.116, -0.0853, -0.0853, -0.099, -0.1126, -0.1126, -0.1092, -0.0956, -0.0683, -0.0444, -0.0444, -0.0205, -0.0341, -0.0171, -0.0171, -0.0341, -0.0273, 0.0, 0.0137, -0.0102, 0.0102, 0.0102, -0.0034, -0.0102, -0.0102, 0.0102, 0.0068, 0.0273, 0.0273, 0.0239, 0.0307, 0.0273, -0.0034, 0.0, -0.0034, 0.0068, 0.0, 0.0068, -0.0034, -0.0239, -0.0137, -0.041, -0.0478, -0.0068, -0.0546, -0.0614, -0.0341, -0.0171, -0.0205, -0.0171, -0.0375, -0.0239, -0.0102, -0.0137, -0.0034, 0.0205, 0.0102, 0.0273, 0.0068, 0.0136, -0.0069, -0.0172, -0.0206, -0.0206, -0.0206, 0.0033, 0.0205, 0.0239, 0.0239, 0.0342, 0.0478, 0.041, 0.0615, 0.0615, 0.0855, 0.1437, 0.1677, 0.1472, 0.154, 0.1506, 0.1574, 0.1711, 0.1711, 0.1574, 0.1917, 0.2054, 0.2156, 0.1917, 0.2054, 0.2122, 0.2019, 0.2191, 0.2465, 0.2362, 0.2156, 0.2088, 0.2328, 0.2225, 0.2088, 0.2191, 0.243, 0.2876, 0.3115, 0.3115, 0.3115, 0.3115, 0.3115, 0.3115, 0.3115, 0.3115, 0.3013, 0.3458, 0.38, 0.3663, 0.3663, 0.3526, 0.3252, 0.2773, 0.3013, 0.2944, 0.2396, 0.267, 0.3286, 0.291, 0.2773, 0.2636, 0.2849, 0.3089, 0.2711, 0.2643, 0.2437, 0.2437, 0.2677, 0.2643, 0.2505, 0.223, 0.2093, 0.2746, 0.2437, 0.2437, 0.2437, 0.278, 0.3399, 0.3433, 0.3742, 0.3673, 0.412, 0.4292, 0.4326, 0.3948, 0.3914, 0.4086, 0.4086, 0.4292, 0.5013, 0.5563, 0.5219, 0.4979, 0.467, 0.467, 0.5632, 0.546, 0.546]}
{"unit_id": "quote:1980928527038865813", "ticker_idx": 0, "ts": "2025-10-22T09:25:39+00:00", "symbol": "FN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "as $FN expands its market share in component supply", "tweet_type": "news_reaction", "timeline": "months", "summary": "Fabrinet gaining share in AWS Trainium3 accelerator modules at the expense of Taiwanese suppliers; bullish for FN into 2026.", "resolved_tickers": ["FN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "quote", "lang": "en", "tweet_text": "Rumor:\n\nAccording to reports from the supply chain, as $FN expands its market share in component supply, the share of Taiwanese companies in the accelerator module segment for AWS’s next-generation ASIC server, Trainium3 (T3), is expected to drop from 90% for the previous T2 generation to 60%.\n\nThis adjustment in supply allocation is projected to affect revenue growth in 2026.", "reply_text": null, "parent_text": null, "quoted_text": "#市場小作文 #RUMOR\n轉傳 \n供應鏈傳出消息，因Fabrinet(FN US)擴大供貨市佔率，AWS下一代ASIC Server Trainium3 (T3)的加速器模組比重，台廠將由原有T2的90%降至60%，供貨比重修正將影響2026年營收成長率。\n#零級分", "parent_author": null, "ret_m1d": 0.03932645696582182, "ret_signed_m1d": 0.03932645696582182, "alpha_spy_m1d": 0.03410047206919842, "alpha_c_m1d": 0.029343681436725833, "bench_spy_m1d": 0.0052259848966234035, "bench_c_m1d": 0.009982775529095989, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.036852889166228975, "ret_signed_p1d": 0.036852889166228975, "alpha_spy_p1d": 0.030922997975729283, "alpha_c_p1d": 0.02424297717942725, "bench_spy_p1d": 0.005929891190499692, "bench_c_p1d": 0.012609911986801725, "ret_p1w": 0.10913457989987707, "ret_signed_p1w": 0.10913457989987707, "alpha_spy_p1w": 0.07979948272578774, "alpha_c_p1w": 0.043843234014729004, "bench_spy_p1w": 0.02933509717408933, "bench_c_p1w": 0.06529134588514807, "ret_p1m": -0.03253045505534702, "ret_signed_p1m": -0.03253045505534702, "alpha_spy_p1m": -0.009664346801397716, "alpha_c_p1m": 0.01419621481203348, "bench_spy_p1m": -0.022866108253949302, "bench_c_p1m": -0.0467266698673805, "ret_p3m": 0.23538383038008814, "ret_signed_p3m": 0.23538383038008814, "alpha_spy_p3m": 0.19659416315091538, "alpha_c_p3m": 0.21369572504080225, "bench_spy_p3m": 0.038789667229172764, "bench_c_p3m": 0.02168810533928589, "ret_p6m": 0.6805917226175946, "ret_signed_p6m": 0.6805917226175946, "alpha_spy_p6m": 0.6239052976131665, "alpha_c_p6m": 0.6126742049908906, "bench_spy_p6m": 0.05668642500442811, "bench_c_p6m": 0.06791751762670395, "price_path": [-0.2282, -0.2089, -0.1985, -0.1988, -0.1912, -0.2125, -0.1766, -0.1802, -0.1694, -0.1627, -0.1405, -0.1695, -0.1297, -0.1572, -0.1953, -0.176, -0.1827, -0.2874, -0.2996, -0.3078, -0.2646, -0.2198, -0.1738, -0.1696, -0.1129, -0.1723, -0.1723, -0.1613, -0.1567, -0.0951, -0.0755, -0.0739, -0.0778, -0.0691, -0.0951, -0.1055, -0.103, -0.1117, -0.105, -0.0555, -0.0427, -0.0269, -0.05, -0.0839, -0.0982, -0.1033, -0.091, -0.089, -0.0723, -0.0767, -0.0855, -0.0748, -0.0751, -0.042, -0.045, -0.0957, -0.0441, -0.0625, -0.0047, 0.0305, 0.0241, 0.0183, 0.0393, 0.0, 0.0369, 0.0519, 0.0584, 0.0932, 0.1091, 0.0943, 0.1008, 0.1068, 0.1483, 0.2165, 0.1524, 0.1222, 0.1674, 0.1226, 0.1181, 0.0056, 0.0327, 0.041, 0.0225, 0.0411, -0.0325, -0.0244, 0.0403, 0.0736, 0.122, 0.122, 0.1478, 0.1233, 0.1033, 0.1176, 0.1787, 0.1954, 0.2268, 0.2454, 0.2908, 0.3196, 0.1765, 0.1743, 0.1272, 0.0657, 0.1305, 0.188, 0.2082, 0.2166, 0.2036, 0.2036, 0.1951, 0.1795, 0.1541, 0.1375, 0.1375, 0.1978, 0.1416, 0.1972, 0.1755, 0.1067, 0.1174, 0.1567, 0.2218, 0.1919, 0.2432, 0.2354, 0.2354, 0.2057, 0.2202, 0.1753, 0.1666, 0.1846, 0.2457, 0.2517, 0.2399, 0.2229, 0.2483, 0.1207, 0.0569, 0.1061, 0.2592, 0.254, 0.165, 0.1631, 0.1544, 0.2381, 0.2381, 0.2412, 0.2637, 0.2838, 0.3645, 0.4409, 0.4638, 0.5391, 0.4053, 0.3633, 0.4526, 0.3817, 0.4088, 0.3603, 0.2227, 0.318, 0.3567, 0.31, 0.2812, 0.2546, 0.2813, 0.2493, 0.2503, 0.3511, 0.2649, 0.3654, 0.5014, 0.5292, 0.3734, 0.3791, 0.229, 0.303, 0.3366, 0.3941, 0.3941, 0.385, 0.3935, 0.5245, 0.5447, 0.6543, 0.7237, 0.7032, 0.7135, 0.6806]}
{"unit_id": "orig:1990331445777252469", "ticker_idx": 0, "ts": "2025-11-17T08:09:29+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "We maintain our Overweight view on the semiconductor sector. The intensifying benefit to AI-oriented memory and the synergy from supply shortages are only just in their early stages.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "SK Securities reiterates Overweight on Korean semis; PT KRW 170K Samsung, KRW 1M SK Hynix on prolonged DRAM supply shortage and structural memory supercycle.", "resolved_tickers": ["EWY"], "resolution_method": "etf", "resolution_reason": "Korea semiconductor sector proxied by EWY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "SK Securities semiconductor comment:\n\nImprovement in the supply shortage will be difficult even at the end of 2026\nIt will be difficult for the current memory supply shortage to improve significantly even by the end of 2026. DRAM demand in 2026 is forecast to grow by 20–25%, but we judge that production will find it difficult to outpace this. This implies that the current three-week inventory level of suppliers will remain at a similar level, or lower, by the end of 2026. DRAM investment in 2025 has been focused on conversions and HBM, and this is due to space constraints for new investments in 2026 and the impact of upward revisions to HBM demand (HBM3e 12H, 4 12H). The production contribution from Samsung Electronics’ P4 investment in 2026 will begin gradually from the second half of 2026, and SK hynix’s M15X, also focused on HBM, will begin contributing from mid-2026. For subsequent new fabs, we expect SK hynix’s Yongin fab to begin production in 2027 and Samsung Electronics’ P5 to start production in 2028. The prolongation of a supplier-favorable supercycle is the driving force that allows the industry to overcome the macro economy. This time is different.\n\nFull-scale expansion of long-term supply contracts. Driving structural improvement in earnings stability for memory\nThe share of long-term supply contracts within the memory industry will begin to increase in earnest. The longer the supply shortage persists, the larger the potential magnitude of memory price hikes becomes, and such violent price increases act as a powerful incentive for long-term supply contracts. In addition, as the AI cycle shifts from scale-up to scale-out/across, memory demand expands structurally. Long-term supply contracts aimed at securing stable volumes will spread not only to HBM but also to server DRAM and other products. The resulting improvement in demand visibility will translate into a structural rise in earnings stability within the memory industry and provide justification for capacity expansions.\n\nGoing the distance\nWe maintain our Overweight view on the semiconductor sector. The intensifying benefit to AI-oriented memory and the synergy from supply shortages are only just in their early stages. The fundamental structural changes in the memory industry will continue to provide justification for valuation expansion. Although concerns remain about a correction following the steep share price rally, earnings that exceed expectations at both the top and bottom ends will validate the move. We maintain our target prices of KRW 170,000 for Samsung Electronics and KRW 1,000,000 for SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011932923241652071, "ret_signed_m1d": 0.011932923241652071, "alpha_spy_m1d": 0.002528785435771397, "alpha_c_m1d": 0.002528785435771397, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.009404137805880675, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015372987083386747, "ret_signed_p1d": -0.015372987083386747, "alpha_spy_p1d": -0.006975526525180187, "alpha_c_p1d": -0.006975526525180187, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.00839746055820656, "ret_p1w": -0.029240920029438566, "ret_signed_p1w": -0.029240920029438566, "alpha_spy_p1w": -0.03383777516123332, "alpha_c_p1w": -0.03383777516123332, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": 0.004596855131794753, "ret_p1m": -0.016011001897943133, "ret_signed_p1m": -0.016011001897943133, "alpha_spy_p1m": -0.03584070866043154, "alpha_c_p1m": -0.03584070866043154, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.019829706762488408, "ret_p3m": 0.4377320015752457, "ret_signed_p3m": 0.4377320015752457, "alpha_spy_p3m": 0.4112727206878575, "alpha_c_p3m": 0.4112727206878575, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.026459280887388248, "ret_p6m": 0.9620424793081264, "ret_signed_p6m": 0.9620424793081264, "alpha_spy_p6m": 0.8468003880456458, "alpha_c_p6m": 0.8468003880456458, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.11524209126248053, "price_path": [-0.2329, -0.2344, -0.2097, -0.2181, -0.2218, -0.2209, -0.2083, -0.2234, -0.2234, -0.2268, -0.2162, -0.2145, -0.2084, -0.2024, -0.1946, -0.176, -0.163, -0.1544, -0.1487, -0.1326, -0.1418, -0.1319, -0.1394, -0.1232, -0.1244, -0.1395, -0.1412, -0.1521, -0.1374, -0.1389, -0.1197, -0.102, -0.0986, -0.0886, -0.104, -0.0949, -0.1031, -0.1322, -0.1011, -0.1111, -0.0844, -0.0553, -0.0484, -0.0278, -0.0471, -0.0384, -0.0317, -0.0104, 0.0139, 0.0154, 0.0417, 0.0241, 0.0428, 0.0834, 0.0295, 0.0387, 0.0053, -0.0131, 0.0285, 0.0244, 0.0312, 0.0113, 0.0119, 0.0, -0.0154, -0.0264, -0.0483, -0.0409, -0.0292, -0.0347, -0.0158, -0.0158, -0.0231, -0.0258, -0.0102, -0.0002, -0.0124, 0.0138, 0.022, 0.0274, 0.0401, 0.0209, -0.0038, -0.0031, -0.016, -0.0263, -0.01, 0.004, 0.0085, 0.0189, 0.0398, 0.0398, 0.0535, 0.0851, 0.084, 0.0686, 0.0686, 0.1236, 0.1523, 0.1828, 0.1765, 0.1716, 0.1969, 0.1973, 0.1821, 0.2051, 0.2206, 0.2335, 0.2335, 0.2198, 0.2713, 0.2749, 0.2967, 0.2896, 0.3487, 0.3801, 0.3685, 0.3455, 0.3292, 0.3665, 0.3219, 0.3213, 0.3708, 0.3855, 0.3679, 0.4355, 0.4377, 0.4726, 0.4726, 0.4364, 0.4611, 0.4862, 0.5595, 0.5312, 0.5889, 0.6366, 0.6533, 0.6638, 0.6217, 0.4547, 0.477, 0.3821, 0.393, 0.4717, 0.4322, 0.4603, 0.3575, 0.3646, 0.463, 0.4776, 0.4496, 0.4819, 0.3826, 0.4708, 0.4152, 0.4041, 0.3189, 0.3298, 0.2798, 0.3521, 0.3874, 0.3506, 0.3506, 0.3843, 0.3981, 0.5396, 0.5311, 0.5249, 0.5524, 0.6123, 0.5929, 0.621, 0.6744, 0.6498, 0.6135, 0.7125, 0.6553, 0.699, 0.7228, 0.6968, 0.6923, 0.767, 0.7807, 0.7982, 0.9068, 1.0003, 0.9428, 1.0906, 1.1198, 0.962]}
{"unit_id": "orig:1990331445777252469", "ticker_idx": 1, "ts": "2025-11-17T08:09:29+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our target prices of KRW 170,000 for Samsung Electronics", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "SK Securities reiterates Overweight on Korean semis; PT KRW 170K Samsung, KRW 1M SK Hynix on prolonged DRAM supply shortage and structural memory supercycle.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "SK Securities semiconductor comment:\n\nImprovement in the supply shortage will be difficult even at the end of 2026\nIt will be difficult for the current memory supply shortage to improve significantly even by the end of 2026. DRAM demand in 2026 is forecast to grow by 20–25%, but we judge that production will find it difficult to outpace this. This implies that the current three-week inventory level of suppliers will remain at a similar level, or lower, by the end of 2026. DRAM investment in 2025 has been focused on conversions and HBM, and this is due to space constraints for new investments in 2026 and the impact of upward revisions to HBM demand (HBM3e 12H, 4 12H). The production contribution from Samsung Electronics’ P4 investment in 2026 will begin gradually from the second half of 2026, and SK hynix’s M15X, also focused on HBM, will begin contributing from mid-2026. For subsequent new fabs, we expect SK hynix’s Yongin fab to begin production in 2027 and Samsung Electronics’ P5 to start production in 2028. The prolongation of a supplier-favorable supercycle is the driving force that allows the industry to overcome the macro economy. This time is different.\n\nFull-scale expansion of long-term supply contracts. Driving structural improvement in earnings stability for memory\nThe share of long-term supply contracts within the memory industry will begin to increase in earnest. The longer the supply shortage persists, the larger the potential magnitude of memory price hikes becomes, and such violent price increases act as a powerful incentive for long-term supply contracts. In addition, as the AI cycle shifts from scale-up to scale-out/across, memory demand expands structurally. Long-term supply contracts aimed at securing stable volumes will spread not only to HBM but also to server DRAM and other products. The resulting improvement in demand visibility will translate into a structural rise in earnings stability within the memory industry and provide justification for capacity expansions.\n\nGoing the distance\nWe maintain our Overweight view on the semiconductor sector. The intensifying benefit to AI-oriented memory and the synergy from supply shortages are only just in their early stages. The fundamental structural changes in the memory industry will continue to provide justification for valuation expansion. Although concerns remain about a correction following the steep share price rally, earnings that exceed expectations at both the top and bottom ends will validate the move. We maintain our target prices of KRW 170,000 for Samsung Electronics and KRW 1,000,000 for SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03379720954848586, "ret_signed_m1d": -0.03379720954848586, "alpha_spy_m1d": -0.04320134735436654, "alpha_c_m1d": -0.049697687662226486, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.015900478113740624, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02783296389954426, "ret_signed_p1d": -0.02783296389954426, "alpha_spy_p1d": -0.0194355033413377, "alpha_c_p1d": -0.011580009741389619, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.016252954158154642, "ret_p1w": -0.0387674012229241, "ret_signed_p1w": -0.0387674012229241, "alpha_spy_p1w": -0.043364256354718855, "alpha_c_p1w": -0.024911758881995483, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.01385564234092862, "ret_p1m": 0.021868796448681227, "ret_signed_p1m": 0.021868796448681227, "alpha_spy_p1m": 0.0020390896861928187, "alpha_c_p1m": 0.016650921047398626, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.005217875401282601, "ret_p3m": 0.7839781208032068, "ret_signed_p3m": 0.7839781208032068, "alpha_spy_p3m": 0.7575188399158186, "alpha_c_p3m": 0.8008927632320656, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": -0.01691464242885876, "ret_p6m": 1.7926097811095025, "ret_signed_p6m": 1.7926097811095025, "alpha_spy_p6m": 1.677367689847022, "alpha_c_p6m": 1.5538288929281312, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.23878088818137133, "price_path": [-0.3023, -0.3013, -0.2934, -0.2924, -0.3043, -0.3013, -0.3112, -0.3102, -0.331, -0.3162, -0.3092, -0.3063, -0.3122, -0.3063, -0.2924, -0.2815, -0.2736, -0.2538, -0.2429, -0.2142, -0.2261, -0.2053, -0.2053, -0.1737, -0.1618, -0.1549, -0.1479, -0.1756, -0.163, -0.166, -0.1451, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.1079, -0.0616, -0.0726, -0.0895, -0.0557, -0.0288, -0.0268, -0.0249, -0.0308, -0.0199, -0.0408, -0.0179, 0.0139, -0.0109, -0.001, 0.0348, 0.0686, 0.1044, 0.0427, 0.0, -0.0139, -0.0268, 0.0, 0.0288, 0.0249, 0.0219, -0.0338, 0.0, -0.0278, -0.0408, 0.0, -0.0577, -0.0388, -0.0129, 0.0219, 0.0288, -0.001, 0.002, 0.0278, 0.0388, 0.0447, 0.0775, 0.0885, 0.0775, 0.0736, 0.0666, 0.0825, 0.0417, 0.0219, 0.0726, 0.0696, 0.0567, 0.0984, 0.1083, 0.1044, 0.1044, 0.163, 0.1936, 0.1976, 0.1976, 0.1976, 0.2835, 0.3794, 0.3874, 0.4084, 0.3864, 0.3884, 0.3864, 0.3744, 0.4014, 0.4374, 0.4873, 0.4913, 0.4504, 0.4933, 0.5213, 0.5193, 0.5193, 0.5932, 0.6222, 0.6052, 0.6032, 0.5023, 0.6731, 0.6891, 0.5912, 0.5842, 0.6621, 0.6561, 0.6761, 0.784, 0.8099, 0.8099, 0.8099, 0.8099, 0.8978, 0.8988, 0.9278, 0.9977, 1.0327, 1.1775, 1.1625, 1.1625, 0.9488, 0.7201, 0.9138, 0.8799, 0.733, 0.8769, 0.8978, 0.8769, 0.8329, 0.8849, 0.9368, 1.0826, 1.0027, 0.9917, 0.8609, 0.8949, 0.8879, 0.799, 0.799, 0.7646, 0.6736, 0.8983, 0.7857, 0.8637, 0.9328, 0.9668, 1.107, 1.0419, 1.0619, 1.0119, 1.0669, 1.112, 1.177, 1.162, 1.147, 1.192, 1.177, 1.2471, 1.1971, 1.2471, 1.2221, 1.2621, 1.2071, 1.2071, 1.3272, 1.3272, 1.6625, 1.7175, 1.6875, 1.8577, 1.7926]}
{"unit_id": "orig:1990331445777252469", "ticker_idx": 2, "ts": "2025-11-17T08:09:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KRW 1,000,000 for SK hynix", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "SK Securities reiterates Overweight on Korean semis; PT KRW 170K Samsung, KRW 1M SK Hynix on prolonged DRAM supply shortage and structural memory supercycle.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Securities semiconductor comment:\n\nImprovement in the supply shortage will be difficult even at the end of 2026\nIt will be difficult for the current memory supply shortage to improve significantly even by the end of 2026. DRAM demand in 2026 is forecast to grow by 20–25%, but we judge that production will find it difficult to outpace this. This implies that the current three-week inventory level of suppliers will remain at a similar level, or lower, by the end of 2026. DRAM investment in 2025 has been focused on conversions and HBM, and this is due to space constraints for new investments in 2026 and the impact of upward revisions to HBM demand (HBM3e 12H, 4 12H). The production contribution from Samsung Electronics’ P4 investment in 2026 will begin gradually from the second half of 2026, and SK hynix’s M15X, also focused on HBM, will begin contributing from mid-2026. For subsequent new fabs, we expect SK hynix’s Yongin fab to begin production in 2027 and Samsung Electronics’ P5 to start production in 2028. The prolongation of a supplier-favorable supercycle is the driving force that allows the industry to overcome the macro economy. This time is different.\n\nFull-scale expansion of long-term supply contracts. Driving structural improvement in earnings stability for memory\nThe share of long-term supply contracts within the memory industry will begin to increase in earnest. The longer the supply shortage persists, the larger the potential magnitude of memory price hikes becomes, and such violent price increases act as a powerful incentive for long-term supply contracts. In addition, as the AI cycle shifts from scale-up to scale-out/across, memory demand expands structurally. Long-term supply contracts aimed at securing stable volumes will spread not only to HBM but also to server DRAM and other products. The resulting improvement in demand visibility will translate into a structural rise in earnings stability within the memory industry and provide justification for capacity expansions.\n\nGoing the distance\nWe maintain our Overweight view on the semiconductor sector. The intensifying benefit to AI-oriented memory and the synergy from supply shortages are only just in their early stages. The fundamental structural changes in the memory industry will continue to provide justification for valuation expansion. Although concerns remain about a correction following the steep share price rally, earnings that exceed expectations at both the top and bottom ends will validate the move. We maintain our target prices of KRW 170,000 for Samsung Electronics and KRW 1,000,000 for SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07590760236361427, "ret_signed_m1d": -0.07590760236361427, "alpha_spy_m1d": -0.08531174016949494, "alpha_c_m1d": -0.08966030558661164, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.013752703222997376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05940598564697308, "ret_signed_p1d": -0.05940598564697308, "alpha_spy_p1d": -0.05100852508876652, "alpha_c_p1d": -0.03880661678882169, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.02059936885815139, "ret_p1w": -0.1419141726470945, "ret_signed_p1w": -0.1419141726470945, "alpha_spy_p1w": -0.14651102777888925, "alpha_c_p1w": -0.1384466706950167, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.0034675019520777894, "ret_p1m": -0.12478615967282203, "ret_signed_p1m": -0.12478615967282203, "alpha_spy_p1m": -0.14461586643531044, "alpha_c_p1m": -0.15899133114981623, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.0342051714769942, "ret_p3m": 0.4663960177041486, "ret_signed_p3m": 0.4663960177041486, "alpha_spy_p3m": 0.4399367368167604, "alpha_c_p3m": 0.2692944484854476, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.19710156921870103, "ret_p6m": 2.0358130710294264, "ret_signed_p6m": 2.0358130710294264, "alpha_spy_p6m": 1.920570979766946, "alpha_c_p6m": 1.3814010152845604, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.654412055744866, "price_path": [-0.579, -0.5963, -0.5864, -0.5724, -0.5691, -0.5716, -0.5569, -0.5561, -0.5776, -0.5701, -0.5668, -0.5619, -0.5487, -0.5429, -0.5248, -0.4983, -0.4934, -0.4579, -0.4538, -0.4257, -0.4497, -0.4134, -0.4134, -0.4208, -0.4043, -0.4101, -0.4117, -0.4447, -0.4241, -0.4266, -0.4059, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.3474, -0.2937, -0.3152, -0.321, -0.3028, -0.2533, -0.2318, -0.1988, -0.2096, -0.2054, -0.2104, -0.1584, -0.1172, -0.1403, -0.0792, -0.0627, -0.0776, 0.0231, -0.033, -0.0446, -0.0215, -0.0429, 0.0, 0.0215, 0.0182, 0.0099, -0.0759, 0.0, -0.0594, -0.0726, -0.0578, -0.1403, -0.1419, -0.1436, -0.1353, -0.1017, -0.1248, -0.1116, -0.0785, -0.0885, -0.105, -0.1017, -0.0472, -0.0653, -0.0307, -0.067, -0.0571, -0.0852, -0.1248, -0.0901, -0.0885, -0.0967, -0.0422, -0.0356, -0.029, -0.029, -0.0108, 0.0569, 0.075, 0.075, 0.075, 0.118, 0.1493, 0.1989, 0.2253, 0.2484, 0.2286, 0.2369, 0.2187, 0.2253, 0.2369, 0.2484, 0.2616, 0.227, 0.222, 0.2468, 0.2666, 0.2154, 0.3211, 0.3888, 0.4218, 0.5011, 0.3706, 0.4978, 0.4862, 0.3904, 0.3855, 0.4647, 0.4466, 0.4202, 0.4664, 0.4532, 0.4532, 0.4532, 0.4532, 0.4763, 0.5671, 0.5704, 0.6596, 0.6811, 0.8182, 0.7553, 0.7553, 0.5535, 0.4046, 0.5568, 0.5287, 0.3831, 0.5518, 0.5799, 0.5386, 0.5055, 0.6114, 0.6048, 0.747, 0.6759, 0.666, 0.5435, 0.6312, 0.6461, 0.5435, 0.5435, 0.4443, 0.3351, 0.484, 0.3731, 0.4492, 0.4658, 0.5154, 0.709, 0.6511, 0.6991, 0.7206, 0.8248, 0.8794, 0.9108, 0.8662, 0.929, 1.025, 1.0233, 1.0266, 1.0217, 1.1375, 1.1507, 1.1391, 1.1276, 1.1276, 1.3939, 1.3939, 1.6487, 1.7364, 1.7893, 2.1103, 2.0358]}
{"unit_id": "thread:1984853391449919540", "ticker_idx": 0, "ts": "2025-11-02T05:21:39+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this price adjustment is expected to have a positive impact on TSMC's revenue and profit growth", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "TSMC's 2026 price hikes (3–10%) mark fourth consecutive year of increases; expected to positively impact revenue and profit growth.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "• As production capacity for advanced process nodes continues to lag behind demand, TSMC has moved up its annual price negotiations. According to industry sources, TSMC began sequential discussions with customers in September regarding plans to raise prices for advanced process products in 2026. This move is aimed at reflecting higher production costs. However, the extent of the increase will vary depending on each customer’s purchasing volume and partnership status.\n\n• According to previous forecasts from research institutions and securities firms, the unit price of advanced process products in 2026 is expected to rise by approximately 3% to 10%. While the degree of increase differs by process, it will generally be higher than this year’s level, marking TSMC’s fourth consecutive year of price hikes. Although the upward trend is clear, the overall rate of increase is relatively moderate, which will help maintain stable cooperation with clients. Nevertheless, this price adjustment is expected to have a positive impact on TSMC’s revenue and profit growth. (Taiwan Economic Daily)\n\n---\n\nhttps://t.co/5PCaFb8ycO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00662257152508261, "ret_signed_m1d": -0.00662257152508261, "alpha_spy_m1d": -0.004749399794318876, "alpha_c_m1d": 0.001790482476316968, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0033113264502250273, "ret_signed_p1d": -0.0033113264502250273, "alpha_spy_p1d": 0.008542213492222506, "alpha_c_p1d": 0.033126789192245654, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": -0.02317887827473808, "ret_signed_p1w": -0.02317887827473808, "alpha_spy_p1w": -0.020398449339700808, "alpha_c_p1w": -0.0032116615328835474, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": -0.05298016532382388, "ret_signed_p1m": -0.05298016532382388, "alpha_spy_p1m": -0.050331423424575195, "alpha_c_p1m": -0.034761065693607596, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": 0.19934873665962738, "ret_signed_p3m": 0.19934873665962738, "alpha_spy_p3m": 0.1806892406185474, "alpha_c_p3m": 0.05534673250431643, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 0.47657882656882955, "ret_signed_p6m": 0.47657882656882955, "alpha_spy_p6m": 0.42916125335609245, "alpha_c_p6m": 0.1306667038624807, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.2579, -0.2217, -0.225, -0.2217, -0.2217, -0.2085, -0.225, -0.2217, -0.2217, -0.2184, -0.2513, -0.2414, -0.2513, -0.2283, -0.225, -0.2151, -0.2348, -0.2348, -0.2316, -0.2348, -0.2348, -0.2348, -0.2217, -0.2217, -0.2085, -0.192, -0.1821, -0.1689, -0.1722, -0.1523, -0.1623, -0.149, -0.1623, -0.1424, -0.1126, -0.1126, -0.1258, -0.1391, -0.1391, -0.1358, -0.1225, -0.096, -0.0728, -0.0728, -0.0497, -0.0629, -0.0464, -0.0464, -0.0629, -0.0563, -0.0298, -0.0166, -0.0397, -0.0199, -0.0199, -0.0331, -0.0397, -0.0397, -0.0199, -0.0232, -0.0033, -0.0033, -0.0066, 0.0, -0.0033, -0.0331, -0.0298, -0.0331, -0.0232, -0.0298, -0.0232, -0.0331, -0.053, -0.043, -0.0695, -0.0762, -0.0364, -0.0828, -0.0894, -0.0629, -0.0464, -0.0497, -0.0464, -0.0662, -0.053, -0.0397, -0.043, -0.0331, -0.0099, -0.0199, -0.0033, -0.0232, -0.0166, -0.0365, -0.0465, -0.0498, -0.0498, -0.0498, -0.0266, -0.01, -0.0066, -0.0066, 0.0033, 0.0166, 0.01, 0.0299, 0.0299, 0.0532, 0.1096, 0.1329, 0.113, 0.1196, 0.1163, 0.1229, 0.1362, 0.1362, 0.1229, 0.1562, 0.1694, 0.1794, 0.1562, 0.1694, 0.1761, 0.1661, 0.1827, 0.2093, 0.1993, 0.1794, 0.1728, 0.196, 0.1861, 0.1728, 0.1827, 0.206, 0.2492, 0.2724, 0.2724, 0.2724, 0.2724, 0.2724, 0.2724, 0.2724, 0.2724, 0.2625, 0.3057, 0.3389, 0.3256, 0.3256, 0.3123, 0.2857, 0.2392, 0.2625, 0.2558, 0.2027, 0.2292, 0.2891, 0.2525, 0.2392, 0.2259, 0.2466, 0.2699, 0.2333, 0.2266, 0.2066, 0.2066, 0.2299, 0.2266, 0.2133, 0.1866, 0.1733, 0.2366, 0.2066, 0.2066, 0.2066, 0.2399, 0.2999, 0.3033, 0.3333, 0.3266, 0.3699, 0.3866, 0.3899, 0.3533, 0.3499, 0.3666, 0.3666, 0.3866, 0.4566, 0.5099, 0.4766]}
{"unit_id": "reply:1974267660402627001", "ticker_idx": 0, "ts": "2025-10-04T00:17:45+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I think it's fine to stay long until 2027", "tweet_type": "prediction", "timeline": "years", "summary": "Stay long memory through 2027; China supply-flood risk overstated given domestic demand absorption and effective US sanctions.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@TShirtnJeans2 China is currently struggling just to meet its own domestic demand, and as I’ve discussed in my previous writings, I believe the U.S. sanctions are proving effective.\n\nI think it’s fine to stay long until 2027.", "reply_text": null, "parent_text": "@Jukanlosreve I'm just not confident in my ability to trade the memory cycle. If I'm lucky I can trade the first 3 innings, to use a baseball analogy, but I don't know what to do about the rest.\n\nPlus this time, I'm very worried about China flooding the market if prices keep going up.", "quoted_text": null, "parent_author": "TShirtnJeans2", "ret_m1d": -0.005463629155534629, "ret_signed_m1d": -0.005463629155534629, "alpha_spy_m1d": -0.0018902153251855918, "alpha_c_m1d": 0.014048319388872115, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.019511948544406743, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009199126582796794, "ret_signed_p1d": -0.009199126582796794, "alpha_spy_p1d": -0.005491634331834897, "alpha_c_p1d": 0.009235258522703646, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.01843438510550044, "ret_p1w": 0.03277911118485314, "ret_signed_p1w": 0.03277911118485314, "alpha_spy_p1w": 0.045539501622676316, "alpha_c_p1w": 0.043350501523405284, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.010571390338552145, "ret_p1m": 0.2640760827968611, "ret_signed_p1m": 0.2640760827968611, "alpha_spy_p1m": 0.2586710864518005, "alpha_c_p1m": 0.23675935483469157, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.02731672796216955, "ret_p3m": 0.49494460438006355, "ret_signed_p3m": 0.49494460438006355, "alpha_spy_p3m": 0.47659313466305026, "alpha_c_p3m": 0.44290141946734235, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.052043184912721197, "ret_p6m": 0.8973973671972154, "ret_signed_p6m": 0.8973973671972154, "alpha_spy_p6m": 0.9235538095693597, "alpha_c_p6m": 0.777375860503743, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.12002150669347245, "price_path": [-0.3269, -0.3097, -0.3034, -0.3095, -0.3037, -0.3085, -0.329, -0.326, -0.3224, -0.3395, -0.3365, -0.3342, -0.3384, -0.3251, -0.3227, -0.3096, -0.3154, -0.3451, -0.3371, -0.3295, -0.3383, -0.3237, -0.3112, -0.2969, -0.2879, -0.2834, -0.284, -0.2917, -0.3005, -0.3069, -0.3198, -0.3308, -0.3195, -0.3141, -0.3168, -0.3147, -0.3035, -0.308, -0.3267, -0.3183, -0.3136, -0.3004, -0.2833, -0.278, -0.257, -0.2311, -0.2072, -0.1701, -0.1629, -0.1361, -0.1507, -0.1088, -0.1196, -0.1083, -0.0923, -0.1008, -0.1076, -0.134, -0.1072, -0.1036, -0.0594, -0.0128, -0.0055, 0.0, -0.0092, 0.0097, 0.0024, 0.0283, 0.0328, 0.0135, 0.044, 0.0978, 0.1092, 0.1345, 0.1189, 0.1185, 0.1225, 0.1791, 0.2139, 0.196, 0.2391, 0.2564, 0.261, 0.3449, 0.2641, 0.2762, 0.2842, 0.2677, 0.3265, 0.327, 0.3304, 0.3112, 0.2638, 0.3067, 0.2425, 0.2264, 0.2064, 0.1532, 0.1883, 0.1982, 0.2254, 0.2452, 0.2331, 0.2479, 0.2727, 0.2624, 0.2431, 0.2755, 0.3244, 0.3206, 0.3565, 0.3262, 0.307, 0.2711, 0.2347, 0.2591, 0.2991, 0.3203, 0.3824, 0.3889, 0.409, 0.409, 0.4369, 0.4998, 0.5076, 0.4949, 0.4949, 0.6014, 0.6475, 0.7304, 0.745, 0.7267, 0.7489, 0.7537, 0.7264, 0.7315, 0.7566, 0.8268, 0.835, 0.8059, 0.8616, 0.8995, 0.9124, 0.8679, 0.9864, 1.0755, 1.0869, 1.0902, 1.0259, 1.1226, 1.0527, 0.9733, 0.9888, 1.0388, 1.0094, 1.0681, 1.1384, 1.1373, 1.1373, 1.1166, 1.1536, 1.1919, 1.2576, 1.2575, 1.324, 1.3672, 1.4679, 1.4246, 1.4251, 1.1846, 1.0599, 1.2035, 1.1298, 1.0338, 1.1978, 1.2472, 1.1949, 1.1979, 1.2988, 1.3495, 1.4767, 1.3801, 1.3336, 1.1898, 1.2318, 1.2133, 1.0812, 1.0843, 0.9592, 0.8974]}
{"unit_id": "thread:2000395081304658182", "ticker_idx": 0, "ts": "2025-12-15T02:38:47+00:00", "symbol": "ASE Technology Holding", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The main beneficiaries of TSMC's advanced packaging order releases include ASE and its subsidiary Silicon Precision Industries (SPIL), Amkor, and Powertech Technology", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC CoW order spillover to OSATs from 2026-2027; ASE, Amkor, and Powertech named as primary beneficiaries with meaningful revenue expected from 2H26.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ticker ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "TSMC to expand CoW Orders in 2H26 as OSAT CoWoS-like tech rises\n\nAs major US CSPs are ramping up ASIC development, the capacity gap in TSMC's CoWoS advanced packaging continues to widen. Thus, the rise of \"CoWoS-like\" capacity by OSAT providers has emerged. To address the strong demand for advanced packaging capacity from AI GPU and ASIC vendors, and as IC design houses increasingly introduce second suppliers to enhance supply chain resilience, TSMC is expected to begin expanding the release of Chip-on-Wafer (CoW) packaging process orders to OSAT providers starting in 2026-2027.\n\nThe main beneficiaries of TSMC's advanced packaging order releases include ASE and its subsidiary Silicon Precision Industries (SPIL), Amkor, and Powertech Technology, the three leading players in the industry. These orders are expected to further support OSAT operations, drive utilization rate recovery, and optimize product mix.\n\nTech gaps limit near-term revenue contribution\nHowever, since most OSAT providers' CoWoS-like technologies still lack sufficient mastery of TSMC's CoW packaging process, there have previously been reports of suspended technology transfers in mass production yields. As a result, current progress remains at the small-volume shipment stage. Meaningful revenue contributions are not expected until the second half of 2026 at the earliest, which would then help ease supply tightness in the market.\n\nIt is estimated that by the end of 2026, TSMC's CoWoS monthly capacity will reach 125,000 wafers, representing annual growth of more than 70%. Meanwhile, the combined average monthly CoWoS-like capacity of ASE, SPIL, and Amkor is expected to reach around 40,000 wafers. This marks multiple-fold growth compared with 2024.\n\nTSMC shifts focus to higher-value advanced packaging\nAlthough TSMC will continue to dominate the global CoWoS advanced packaging market in terms of scale, it is expected to focus future expansion on the CoW segment of CoWoS-L, as well as new technologies such as WMCM, SoIC, and CoPoS. These efforts will concentrate on higher value-added orders to meet demand from customers, including Apple, AMD, Nvidia, and Broadcom.\n\nAside from on-substrate (oS) packaging and wafer testing (CP) orders that have already been increasingly outsourced to OSAT providers in 2025, lower-margin CoWoS-R and CoWoS-S processes are also expected to see a gradual expansion in outsourced packaging volumes. Orders for these two processes are expected to flow to ASE, SPIL, and Amkor.\n\nASE stated that its advanced packaging and testing revenue target of US$1.6 billion for 2025 is on track. Testing revenue growth is expected to be twice that of packaging, with demand from AI and HPC applications serving as the biggest growth drivers going forward. As customer order visibility continues to improve steadily, related revenue is projected to increase by another US$1 billion in 2026.\n\nUtilization rates and revenue trends\nWith spillover CoWoS orders from TSMC, ASE Group has recently seen that advanced packaging, testing, and traditional packaging orders all remain robust. Consolidated revenue for November reached NT$58.82 billion (approx. US$1.9 billion), down 2.3% month-over-month but up 11.1% year-over-year, marking the second-highest level for the same period on record. Cumulative revenue for the first 11 months totaled NT$586.523 billion, representing year-on-year growth of 8.1%.\n\nASE stated that its capacity utilization for advanced packaging processes such as flip-chip and wafer bumping has already reached full load, while utilization for traditional wire-bond packaging lines was still gradually recovering. Overall, average utilization across the packaging business stands at the upper end of the 70–80% range.\n\n---\n\nhttps://t.co/A7OaFXwkLU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011553293071709048, "ret_signed_m1d": 0.011553293071709048, "alpha_spy_m1d": 0.010040209224681629, "alpha_c_m1d": 0.008096219107952995, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017971741613756476, "ret_signed_p1d": -0.017971741613756476, "alpha_spy_p1d": -0.015239444018477766, "alpha_c_p1d": -0.01525139836473155, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": -0.014762486737010083, "ret_signed_p1w": -0.014762486737010083, "alpha_spy_p1w": -0.023758212049712513, "alpha_c_p1w": -0.03720297640920445, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.20025679425621168, "ret_signed_p1m": 0.20025679425621168, "alpha_spy_p1m": 0.17808930435050074, "alpha_c_p1m": 0.0873401690467317, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.3472400432544558, "ret_signed_p3m": 0.3472400432544558, "alpha_spy_p3m": 0.3658992165191479, "alpha_c_p3m": 0.243987836467078, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2747, -0.2638, -0.2644, -0.2484, -0.249, -0.2593, -0.2644, -0.2831, -0.2856, -0.2882, -0.2856, -0.2843, -0.2888, -0.2689, -0.2696, -0.2388, -0.2484, -0.2888, -0.2471, -0.2734, -0.2343, -0.1926, -0.1836, -0.1662, -0.1861, -0.2003, -0.1772, -0.1669, -0.1508, -0.1425, -0.0738, -0.0347, 0.0276, 0.0392, -0.0135, 0.016, -0.0135, -0.0231, -0.0128, -0.0385, -0.0379, -0.061, -0.0623, -0.0822, -0.0905, -0.0969, -0.1175, -0.113, -0.0911, -0.0899, -0.0757, -0.0757, -0.0404, -0.0481, -0.0199, -0.0116, -0.027, -0.0237, 0.0103, 0.025, 0.0546, 0.0526, 0.0116, 0.0, -0.018, -0.0533, -0.0456, -0.027, -0.0148, -0.0032, -0.0019, -0.0019, 0.009, 0.0282, 0.0289, 0.0334, 0.0334, 0.0822, 0.0886, 0.1232, 0.1309, 0.1123, 0.1322, 0.1727, 0.2003, 0.2118, 0.2317, 0.2458, 0.2458, 0.2131, 0.2182, 0.2234, 0.2445, 0.2728, 0.2882, 0.3004, 0.2696, 0.2182, 0.2458, 0.2445, 0.2189, 0.3004, 0.3408, 0.4217, 0.4435, 0.5237, 0.5058, 0.5019, 0.5019, 0.5064, 0.4981, 0.4878, 0.5379, 0.5071, 0.5931, 0.6117, 0.5552, 0.5591, 0.5533, 0.432, 0.4185, 0.4159, 0.3556, 0.3902, 0.3935, 0.4056, 0.3472, 0.38, 0.3813, 0.4012, 0.3902, 0.4056, 0.3678, 0.3915, 0.3819, 0.4377, 0.3755, 0.38, 0.3408, 0.3915, 0.4429, 0.43, 0.43, 0.4474, 0.4243, 0.5494, 0.5732, 0.5944, 0.7041, 0.7246, 0.7478, 0.7715, 0.835, 0.8793, 0.8999, 0.9223, 0.9134, 1.0539, 0.9801, 0.9307, 0.9634, 1.016, 1.0276, 1.0719, 1.1463, 1.1926, 1.1399, 1.197, 1.2657, 1.2003, 1.2766, 1.2625, 1.1701, 1.0315, 0.9827, 1.034, 1.095, 1.2343, 1.2343, 1.5, 1.5077, 1.6059, 1.4615, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000395081304658182", "ticker_idx": 1, "ts": "2025-12-15T02:38:47+00:00", "symbol": "Amkor Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The main beneficiaries of TSMC's advanced packaging order releases include ASE and its subsidiary Silicon Precision Industries (SPIL), Amkor, and Powertech Technology", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC CoW order spillover to OSATs from 2026-2027; ASE, Amkor, and Powertech named as primary beneficiaries with meaningful revenue expected from 2H26.", "resolved_tickers": ["AMKR"], "resolution_method": "exchange_ticker", "resolution_reason": "Amkor Technology AMKR US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "TSMC to expand CoW Orders in 2H26 as OSAT CoWoS-like tech rises\n\nAs major US CSPs are ramping up ASIC development, the capacity gap in TSMC's CoWoS advanced packaging continues to widen. Thus, the rise of \"CoWoS-like\" capacity by OSAT providers has emerged. To address the strong demand for advanced packaging capacity from AI GPU and ASIC vendors, and as IC design houses increasingly introduce second suppliers to enhance supply chain resilience, TSMC is expected to begin expanding the release of Chip-on-Wafer (CoW) packaging process orders to OSAT providers starting in 2026-2027.\n\nThe main beneficiaries of TSMC's advanced packaging order releases include ASE and its subsidiary Silicon Precision Industries (SPIL), Amkor, and Powertech Technology, the three leading players in the industry. These orders are expected to further support OSAT operations, drive utilization rate recovery, and optimize product mix.\n\nTech gaps limit near-term revenue contribution\nHowever, since most OSAT providers' CoWoS-like technologies still lack sufficient mastery of TSMC's CoW packaging process, there have previously been reports of suspended technology transfers in mass production yields. As a result, current progress remains at the small-volume shipment stage. Meaningful revenue contributions are not expected until the second half of 2026 at the earliest, which would then help ease supply tightness in the market.\n\nIt is estimated that by the end of 2026, TSMC's CoWoS monthly capacity will reach 125,000 wafers, representing annual growth of more than 70%. Meanwhile, the combined average monthly CoWoS-like capacity of ASE, SPIL, and Amkor is expected to reach around 40,000 wafers. This marks multiple-fold growth compared with 2024.\n\nTSMC shifts focus to higher-value advanced packaging\nAlthough TSMC will continue to dominate the global CoWoS advanced packaging market in terms of scale, it is expected to focus future expansion on the CoW segment of CoWoS-L, as well as new technologies such as WMCM, SoIC, and CoPoS. These efforts will concentrate on higher value-added orders to meet demand from customers, including Apple, AMD, Nvidia, and Broadcom.\n\nAside from on-substrate (oS) packaging and wafer testing (CP) orders that have already been increasingly outsourced to OSAT providers in 2025, lower-margin CoWoS-R and CoWoS-S processes are also expected to see a gradual expansion in outsourced packaging volumes. Orders for these two processes are expected to flow to ASE, SPIL, and Amkor.\n\nASE stated that its advanced packaging and testing revenue target of US$1.6 billion for 2025 is on track. Testing revenue growth is expected to be twice that of packaging, with demand from AI and HPC applications serving as the biggest growth drivers going forward. As customer order visibility continues to improve steadily, related revenue is projected to increase by another US$1 billion in 2026.\n\nUtilization rates and revenue trends\nWith spillover CoWoS orders from TSMC, ASE Group has recently seen that advanced packaging, testing, and traditional packaging orders all remain robust. Consolidated revenue for November reached NT$58.82 billion (approx. US$1.9 billion), down 2.3% month-over-month but up 11.1% year-over-year, marking the second-highest level for the same period on record. Cumulative revenue for the first 11 months totaled NT$586.523 billion, representing year-on-year growth of 8.1%.\n\nASE stated that its capacity utilization for advanced packaging processes such as flip-chip and wafer bumping has already reached full load, while utilization for traditional wire-bond packaging lines was still gradually recovering. Overall, average utilization across the packaging business stands at the upper end of the 70–80% range.\n\n---\n\nhttps://t.co/A7OaFXwkLU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009109537626793784, "ret_signed_m1d": 0.009109537626793784, "alpha_spy_m1d": 0.007596453779766366, "alpha_c_m1d": 0.005652463663037732, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08266911706848656, "ret_signed_p1d": -0.08266911706848656, "alpha_spy_p1d": -0.07993681947320785, "alpha_c_p1d": -0.07994877381946164, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": -0.07105442830572717, "ret_signed_p1w": -0.07105442830572717, "alpha_spy_p1w": -0.0800501536184296, "alpha_c_p1w": -0.09349491797792153, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.18310181945294546, "ret_signed_p1m": 0.18310181945294546, "alpha_spy_p1m": 0.16093432954723452, "alpha_c_p1m": 0.07018519424346548, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": -0.05832194826093895, "ret_signed_p3m": -0.05832194826093895, "alpha_spy_p3m": -0.03966277499624682, "alpha_c_p3m": -0.16157415504831674, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.395, -0.3373, -0.3566, -0.3289, -0.3277, -0.3327, -0.3373, -0.3402, -0.3518, -0.3546, -0.3302, -0.3264, -0.3323, -0.3052, -0.3239, -0.3043, -0.3039, -0.3577, -0.3077, -0.3202, -0.2891, -0.2841, -0.2889, -0.2664, -0.2693, -0.3007, -0.2732, -0.2552, -0.2459, -0.2691, -0.2571, -0.2746, -0.2664, -0.14, -0.1989, -0.1693, -0.2012, -0.2096, -0.1998, -0.2214, -0.2166, -0.2739, -0.28, -0.298, -0.3025, -0.2857, -0.3139, -0.2716, -0.2468, -0.2318, -0.2064, -0.2064, -0.173, -0.1423, -0.0762, 0.0005, -0.0166, -0.0155, 0.0182, 0.0159, 0.067, 0.0713, 0.0091, 0.0, -0.0827, -0.1148, -0.105, -0.0802, -0.0711, -0.0695, -0.0717, -0.0717, -0.0756, -0.087, -0.0799, -0.1009, -0.1009, -0.0225, 0.0961, 0.1681, 0.2006, 0.1615, 0.1908, 0.1747, 0.1831, 0.1084, 0.1205, 0.0931, 0.0931, 0.1171, 0.2175, 0.1945, 0.1321, 0.1476, 0.1555, 0.1606, 0.1387, 0.1007, 0.0977, 0.0531, -0.0009, 0.0082, 0.1241, 0.1961, 0.2175, 0.2792, 0.1749, 0.0813, 0.0813, 0.0667, 0.0645, 0.1009, 0.0918, 0.072, 0.1052, 0.1601, 0.105, 0.089, 0.0888, 0.0155, 0.0424, 0.0148, -0.0524, -0.0157, 0.0007, 0.0002, -0.0583, -0.0191, 0.0183, 0.0574, 0.0695, 0.0982, 0.0453, 0.0523, 0.1468, 0.1393, 0.03, 0.0142, -0.0588, 0.0275, 0.0608, 0.0656, 0.0656, 0.0731, 0.0866, 0.1961, 0.2582, 0.3225, 0.3791, 0.3992, 0.3627, 0.4352, 0.5372, 0.5844, 0.602, 0.6486, 0.6636, 0.782, 0.7255, 0.6283, 0.6111, 0.5915, 0.6221, 0.6191, 0.7503, 0.762, 0.649, 0.748, 0.7499, 0.6759, 0.7024, 0.6449, 0.6052, 0.5069, 0.4955, 0.5628, 0.5037, 0.5003, 0.5003, 0.6762, 0.6472, 0.6105, 0.5872, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000395081304658182", "ticker_idx": 2, "ts": "2025-12-15T02:38:47+00:00", "symbol": "Powertech Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The main beneficiaries of TSMC's advanced packaging order releases include ASE and its subsidiary Silicon Precision Industries (SPIL), Amkor, and Powertech Technology", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC CoW order spillover to OSATs from 2026-2027; ASE, Amkor, and Powertech named as primary beneficiaries with meaningful revenue expected from 2H26.", "resolved_tickers": ["6239.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Powertech Technology listed on TWSE as 6239.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "TSMC to expand CoW Orders in 2H26 as OSAT CoWoS-like tech rises\n\nAs major US CSPs are ramping up ASIC development, the capacity gap in TSMC's CoWoS advanced packaging continues to widen. Thus, the rise of \"CoWoS-like\" capacity by OSAT providers has emerged. To address the strong demand for advanced packaging capacity from AI GPU and ASIC vendors, and as IC design houses increasingly introduce second suppliers to enhance supply chain resilience, TSMC is expected to begin expanding the release of Chip-on-Wafer (CoW) packaging process orders to OSAT providers starting in 2026-2027.\n\nThe main beneficiaries of TSMC's advanced packaging order releases include ASE and its subsidiary Silicon Precision Industries (SPIL), Amkor, and Powertech Technology, the three leading players in the industry. These orders are expected to further support OSAT operations, drive utilization rate recovery, and optimize product mix.\n\nTech gaps limit near-term revenue contribution\nHowever, since most OSAT providers' CoWoS-like technologies still lack sufficient mastery of TSMC's CoW packaging process, there have previously been reports of suspended technology transfers in mass production yields. As a result, current progress remains at the small-volume shipment stage. Meaningful revenue contributions are not expected until the second half of 2026 at the earliest, which would then help ease supply tightness in the market.\n\nIt is estimated that by the end of 2026, TSMC's CoWoS monthly capacity will reach 125,000 wafers, representing annual growth of more than 70%. Meanwhile, the combined average monthly CoWoS-like capacity of ASE, SPIL, and Amkor is expected to reach around 40,000 wafers. This marks multiple-fold growth compared with 2024.\n\nTSMC shifts focus to higher-value advanced packaging\nAlthough TSMC will continue to dominate the global CoWoS advanced packaging market in terms of scale, it is expected to focus future expansion on the CoW segment of CoWoS-L, as well as new technologies such as WMCM, SoIC, and CoPoS. These efforts will concentrate on higher value-added orders to meet demand from customers, including Apple, AMD, Nvidia, and Broadcom.\n\nAside from on-substrate (oS) packaging and wafer testing (CP) orders that have already been increasingly outsourced to OSAT providers in 2025, lower-margin CoWoS-R and CoWoS-S processes are also expected to see a gradual expansion in outsourced packaging volumes. Orders for these two processes are expected to flow to ASE, SPIL, and Amkor.\n\nASE stated that its advanced packaging and testing revenue target of US$1.6 billion for 2025 is on track. Testing revenue growth is expected to be twice that of packaging, with demand from AI and HPC applications serving as the biggest growth drivers going forward. As customer order visibility continues to improve steadily, related revenue is projected to increase by another US$1 billion in 2026.\n\nUtilization rates and revenue trends\nWith spillover CoWoS orders from TSMC, ASE Group has recently seen that advanced packaging, testing, and traditional packaging orders all remain robust. Consolidated revenue for November reached NT$58.82 billion (approx. US$1.9 billion), down 2.3% month-over-month but up 11.1% year-over-year, marking the second-highest level for the same period on record. Cumulative revenue for the first 11 months totaled NT$586.523 billion, representing year-on-year growth of 8.1%.\n\nASE stated that its capacity utilization for advanced packaging processes such as flip-chip and wafer bumping has already reached full load, while utilization for traditional wire-bond packaging lines was still gradually recovering. Overall, average utilization across the packaging business stands at the upper end of the 70–80% range.\n\n---\n\nhttps://t.co/A7OaFXwkLU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018867924528301883, "ret_signed_m1d": 0.018867924528301883, "alpha_spy_m1d": 0.017354840681274464, "alpha_c_m1d": 0.01541085056454583, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.022012578616352196, "ret_signed_p1d": -0.022012578616352196, "alpha_spy_p1d": -0.019280281021073487, "alpha_c_p1d": -0.01929223536732727, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.08176100628930816, "ret_signed_p1w": 0.08176100628930816, "alpha_spy_p1w": 0.07276528097660573, "alpha_c_p1w": 0.059320516617113794, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.5220125786163523, "ret_signed_p1m": 0.5220125786163523, "alpha_spy_p1m": 0.49984508871064137, "alpha_c_p1m": 0.4090959534068723, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.28616352201257866, "ret_signed_p3m": 0.28616352201257866, "alpha_spy_p3m": 0.3048226952772708, "alpha_c_p3m": 0.18291131522520088, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0881, -0.0597, -0.0535, -0.0503, -0.0535, -0.0503, -0.0535, -0.1195, -0.1195, -0.0818, -0.0755, -0.0503, -0.044, -0.044, -0.0063, -0.0063, 0.0, 0.0, -0.0346, -0.0723, -0.0566, -0.0189, -0.0346, -0.0189, -0.044, -0.0472, -0.0566, -0.0566, 0.0094, 0.1101, 0.1447, 0.1132, 0.0881, 0.0881, 0.0943, 0.0849, 0.0692, 0.0692, 0.0818, 0.066, 0.0566, 0.0692, 0.0094, 0.0346, 0.0094, -0.0314, -0.0031, -0.0503, -0.0472, -0.044, -0.0283, -0.022, -0.0126, -0.0346, -0.022, -0.0031, -0.0189, -0.0126, 0.0063, 0.0094, 0.0, 0.0189, 0.0189, 0.0, -0.022, 0.0189, 0.0157, 0.0, 0.0818, 0.0314, 0.0346, 0.0346, 0.1006, 0.1101, 0.1006, 0.0881, 0.0881, 0.1541, 0.2673, 0.283, 0.2925, 0.2767, 0.2579, 0.3836, 0.522, 0.5818, 0.6195, 0.6541, 0.6509, 0.5818, 0.5063, 0.6195, 0.5503, 0.6195, 0.6572, 0.6447, 0.6069, 0.5975, 0.5377, 0.6164, 0.5881, 0.5346, 0.5063, 0.5283, 0.5094, 0.5283, 0.5283, 0.5283, 0.5283, 0.5283, 0.5283, 0.5283, 0.5283, 0.566, 0.6195, 0.6226, 0.6226, 0.6226, 0.5912, 0.4528, 0.3648, 0.4119, 0.4151, 0.2736, 0.3491, 0.3553, 0.2862, 0.2642, 0.283, 0.2736, 0.327, 0.2925, 0.2673, 0.2075, 0.2075, 0.2799, 0.2484, 0.2264, 0.2327, 0.1761, 0.2264, 0.2704, 0.2704, 0.2704, 0.2075, 0.327, 0.3176, 0.3491, 0.3742, 0.3679, 0.3208, 0.3365, 0.305, 0.3082, 0.3396, 0.3742, 0.3019, 0.3208, 0.3648, 0.3742, 0.2862, 0.2767, 0.2767, 0.4025, 0.4308, 0.4748, 0.4497, 0.3962, 0.4182, 0.4245, 0.4591, 0.4403, 0.5597, 0.5723, 0.4277, 0.4748, 0.6195, 0.7799, 0.956, 1.1509, 1.1415, 1.2484, 1.3836, null, null, null, null, null, null, null]}
{"unit_id": "thread:1978988409910890917", "ticker_idx": 0, "ts": "2025-10-17T00:56:19+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": "foundry", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC one of the biggest beneficiaries of this cycle", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long TSMC: near-monopoly on sub-2nm, 2-3yr order visibility, AI demand accelerating, CoWoS tight, record profits with growth into 2026.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "TSMC’s Hidden Strategy to Dominate Global AI — Why C.C. Wei’s Tone Has Changed\n\nAnother “spoiler” from TSMC’s earnings call.\n\nIn the third quarter of 2025, TSMC posted a net profit of NT$452.3 billion, hitting a new record and meeting expectations.\nCumulative profit for the first three quarters reached NT$1.21 trillion, already surpassing the NT$1.17 trillion achieved in all of 2024.\nWhat caught attention was Chairman C.C. Wei’s noticeably different attitude.\nGone were the bold declarations of having “no competitors.”\nFor the first time, he described TSMC’s expansion plans as guided by “discipline.”\nHis tone—optimistic about AI but not exuberant—earned widespread approval.\n\nAmid Surging AI Infrastructure Investment, TSMC Remains Cautious\n\nWhile companies such as NVIDIA and OpenAI are making aggressive moves in AI infrastructure expansion,\nTSMC has adopted a decidedly prudent and conservative stance.\nWei reiterated that the company maintains technological leadership in 2-nanometer and A16 processes,\nand that nearly all major chipmakers are working with TSMC—giving it near-complete control over sub-2-nanometer orders.\n\nCAPEX Upgrades as Expected — Even Stronger Order Visibility\n\nTSMC continues to see strong long-term AI chip demand,\nexpects record-high revenue this year,\nand anticipates continued growth momentum into 2026.\nWei delivered several notable remarks—most prominently, repeated mentions of “customers’ customers.”\n\nThis phrase was widely interpreted as a signal that TSMC now holds a deeper grasp of the entire AI ecosystem,\nspecifically referencing the escalating capital-spending competition among OpenAI, NVIDIA, Oracle, AMD, Broadcom, and Google,\nas well as major cloud service providers (CSPs).\n\nAccording to supply-chain insiders,\nthanks to robust AI demand, full utilization rates, and favorable exchange rates,\nTSMC’s Q3 and full-year results, upward CAPEX revision, and U.S.-dollar revenue growth were all within expectations.\nThe company’s discussion no longer centered on smartphone clients such as Apple, Qualcomm, and MediaTek—\ninstead, AI contributions took the spotlight.\n\nWei’s New Tone — “The AI Trend Is Extremely Positive”\n\nWei’s remarks were far more measured than in the past.\nHe stated:\n\n“Recent AI developments are very positive.\nThe number of tokens processed by large language models (LLMs) is growing explosively,\nshowing that user adoption of AI models is rapidly expanding.\nThis translates into higher computing demand,\nwhich in turn drives advanced semiconductor needs.”\n\nHe added that enterprise AI demand remains strong and that sovereign AI is rising globally.\nWei emphasized that TSMC’s clients maintain a solid outlook and that\n\n“very strong signals are coming directly from our customers’ customers,\nasking us to secure more capacity to support their business.”\n\nWho Are the “Customers’ Customers”?\n\nTSMC’s direct clients include Apple, NVIDIA, AMD, Qualcomm, and Bitmain.\nBut Wei’s “customers’ customers” refer to the end-users driving AI adoption.\n\nSupply-chain analysts noted that TSMC already collaborates with major chipmakers such as NVIDIA and Broadcom.\nRecent AI-related wafer starts include Google and Amazon’s Annapurna Labs.\nThe recently unfolding alliances—OpenAI with NVIDIA, Oracle, AMD, and Broadcom—\nhave sparked what the market calls an “AI perpetual-motion phenomenon,”\ndriving an unprecedented surge in AI compute capacity.\nOpenAI and Oracle, in particular, are believed to be the “customers’ customers” Wei referred to—\nmaking TSMC one of the biggest beneficiaries of this cycle.\n\nOrders Placed 2–3 Years in Advance — Switching Foundries Is Nearly Impossible\n\nWei explained that to address structural long-term demand growth,\nTSMC has established a highly rigorous and sophisticated capacity-planning system.\nThe company now collaborates not only with direct clients but also their end customers,\nand as manufacturing processes grow more complex,\nproject lead time has extended—\ntoday, engagements must start at least 2 to 3 years ahead.\n\nThis gives TSMC the deepest and broadest visibility in the entire industry.\nSupply-chain sources noted that TSMC’s visibility now extends beyond NVIDIA and Broadcom\nto OpenAI and leading CSPs, allowing the company to assess true market supply-demand conditions\nand minimize the risk of overbooking.\n\nEarly collaboration also means chip firms must commit wafers early to secure capacity.\nAs a result, TSMC has near-complete control over long-term order flows—\nshifting foundries has become extremely difficult.\nFor Samsung Electronics and Intel, this is clearly bad news.\n\nAccelerating the U.S. Arizona Fab — Moving to Sub-2 nm\n\nRegarding the Arizona plant,\nWei said TSMC is “speeding up construction” and for the first time confirmed\nthat it will upgrade to sub-2-nanometer process technology,\nwhile also securing a larger land parcel.\n\nAccording to the supply chain,\nthe previously 3-nanometer-focused second fab will now include 2-nanometer production,\nmeaning TSMC’s U.S. operations will enter the 2 nm era ahead of schedule.\nWith this acceleration and expansion,\nTSMC’s goal for the U.S. fab to account for 30% of sub-2 nm capacity may be raised.\nIf so, the U.S. government’s ambition for a 50-50 capacity balance between Taiwan and the U.S.\ncould soon be realized.\n\nBeyond AI: Other Markets Recovering Gradually\n\nWei also addressed concerns about non-AI markets and the geopolitical environment.\nHe said demand for smartphones and PCs has bottomed out and is slowly recovering,\nwith inventory levels “healthy.”\nHe emphasized that U.S. restrictions on Chinese AI chips have no impact on TSMC’s performance,\nand forecast that the AI accelerator market’s CAGR for 2024–2029\nwill likely exceed the previous estimate of 44–46%.\n\nCoWoS Capacity Remains Tight — Next Update in January 2026\n\nFinally, on CoWoS advanced packaging, Wei reiterated that both front-end and back-end capacity\nremain extremely tight.\nTSMC is working to narrow the supply gap and boost output,\nand Wei stated that a new capacity target will be announced in January 2026.\n\n---\n\nhttps://t.co/8e9zuDX0QL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02413800642620334, "ret_signed_m1d": 0.02413800642620334, "alpha_spy_m1d": 0.029782268259640476, "alpha_c_m1d": 0.022767081185159688, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020689744006068933, "ret_signed_p1d": 0.020689744006068933, "alpha_spy_p1d": 0.0102892686064473, "alpha_c_p1d": 0.00788450039784161, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.019356070768910216, "alpha_c_p1w": -0.024239561975410995, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": -0.003448262420134296, "ret_signed_p1m": -0.003448262420134296, "alpha_spy_p1m": -0.005374769127245327, "alpha_c_p1m": 0.003931499676608885, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.18324144972270506, "ret_signed_p3m": 0.18324144972270506, "alpha_spy_p3m": 0.14108262397159654, "alpha_c_p3m": 0.04693944220525226, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.38148115883175837, "ret_signed_p6m": 0.38148115883175837, "alpha_spy_p6m": 0.34292460128083513, "alpha_c_p6m": 0.08427987150322225, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.2238, -0.2135, -0.2135, -0.2135, -0.2135, -0.2204, -0.2066, -0.2032, -0.2032, -0.2204, -0.2101, -0.2272, -0.1894, -0.1929, -0.1894, -0.1894, -0.1757, -0.1929, -0.1894, -0.1894, -0.186, -0.2204, -0.2101, -0.2204, -0.1963, -0.1929, -0.1826, -0.2032, -0.2032, -0.1998, -0.2032, -0.2032, -0.2032, -0.1894, -0.1894, -0.1757, -0.1585, -0.1482, -0.1345, -0.1379, -0.1172, -0.1276, -0.1138, -0.1276, -0.1069, -0.0759, -0.0759, -0.0897, -0.1034, -0.1034, -0.1, -0.0862, -0.0586, -0.0345, -0.0345, -0.0103, -0.0241, -0.0069, -0.0069, -0.0241, -0.0172, 0.0103, 0.0241, 0.0, 0.0207, 0.0207, 0.0069, 0.0, 0.0, 0.0207, 0.0172, 0.0379, 0.0379, 0.0345, 0.0414, 0.0379, 0.0069, 0.0103, 0.0069, 0.0172, 0.0103, 0.0172, 0.0069, -0.0138, -0.0034, -0.031, -0.0379, 0.0034, -0.0448, -0.0517, -0.0241, -0.0069, -0.0103, -0.0069, -0.0276, -0.0138, 0.0, -0.0034, 0.0069, 0.031, 0.0207, 0.0379, 0.0172, 0.0241, 0.0033, -0.007, -0.0105, -0.0105, -0.0105, 0.0137, 0.031, 0.0345, 0.0345, 0.0449, 0.0587, 0.0518, 0.0725, 0.0725, 0.0967, 0.1556, 0.1798, 0.159, 0.1659, 0.1625, 0.1694, 0.1832, 0.1832, 0.1694, 0.204, 0.2178, 0.2282, 0.204, 0.2178, 0.2248, 0.2144, 0.2317, 0.2594, 0.249, 0.2282, 0.2213, 0.2455, 0.2351, 0.2213, 0.2317, 0.2559, 0.3009, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3147, 0.3597, 0.3943, 0.3804, 0.3804, 0.3666, 0.3389, 0.2905, 0.3147, 0.3078, 0.2524, 0.2801, 0.3424, 0.3043, 0.2905, 0.2767, 0.2982, 0.3225, 0.2843, 0.2773, 0.2565, 0.2565, 0.2808, 0.2773, 0.2635, 0.2357, 0.2218, 0.2878, 0.2565, 0.2565, 0.2565, 0.2912, 0.3537, 0.3572, 0.3884, 0.3815]}
{"unit_id": "orig:1975343583558574106", "ticker_idx": 0, "ts": "2025-10-06T23:33:05+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rating Upgraded from Equal-weight → Overweight, Price Target Raised from $160 → $220 (Morgan Stanley)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Morgan Stanley upgrade of MU to Overweight, PT raised to $220 on faster-than-expected DRAM/NAND price recovery and EPS upward revision cycle.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron: Rating Upgraded from Equal-weight → Overweight, Price Target Raised from $160 → $220 (Morgan Stanley)\n\n    • DRAM and NAND prices are improving faster than expected, with buyers increasing inventories amid concerns over supply shortages through CY26.\n    • Concerns over HBM are easing, and the strength in the DRAM market is positively influencing HBM3e price negotiations.\n    • There is a high likelihood of continued EPS upward revisions over the coming quarters.\n    • While the current valuation appears stretched, Morgan Stanley is betting on short-term upside given the strong industry sentiment and confidence in ongoing supply tightness.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016390887466603887, "ret_signed_m1d": -0.016390887466603887, "alpha_spy_m1d": -0.01281747363625485, "alpha_c_m1d": 0.003121061077802856, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.019511948544406743, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027597379748390383, "ret_signed_p1d": -0.027597379748390383, "alpha_spy_p1d": -0.023889887497428486, "alpha_c_p1d": -0.009162994642889943, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.01843438510550044, "ret_p1w": 0.009478387210694939, "ret_signed_p1w": 0.009478387210694939, "alpha_spy_p1w": 0.022238777648518115, "alpha_c_p1w": 0.020049777549247083, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.010571390338552145, "ret_p1m": 0.14175739704783452, "ret_signed_p1m": 0.14175739704783452, "alpha_spy_p1m": 0.13635240070277388, "alpha_c_p1m": 0.11444066908566497, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.02731672796216955, "ret_p3m": 0.4952100244648945, "ret_signed_p3m": 0.4952100244648945, "alpha_spy_p3m": 0.4768585547478812, "alpha_c_p3m": 0.4431668395521733, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.052043184912721197, "ret_p6m": 0.7706245009775721, "ret_signed_p6m": 0.7706245009775721, "alpha_spy_p6m": 0.7967809433497164, "alpha_c_p6m": 0.6506029942840996, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.12002150669347245, "price_path": [-0.3603, -0.3557, -0.3483, -0.3793, -0.3714, -0.3907, -0.4073, -0.4013, -0.4074, -0.4284, -0.4252, -0.4153, -0.4177, -0.4178, -0.4141, -0.3995, -0.4288, -0.4511, -0.436, -0.4292, -0.4307, -0.4145, -0.3778, -0.3525, -0.3314, -0.3496, -0.3443, -0.3674, -0.3534, -0.3613, -0.3866, -0.394, -0.3841, -0.3907, -0.3903, -0.3838, -0.3615, -0.3772, -0.3772, -0.3799, -0.3787, -0.35, -0.3125, -0.312, -0.2922, -0.2673, -0.212, -0.1771, -0.1743, -0.1688, -0.1627, -0.1161, -0.1484, -0.1385, -0.1291, -0.1537, -0.1792, -0.1769, -0.1422, -0.1243, -0.0467, -0.0384, -0.0164, 0.0, -0.0276, 0.0292, 0.0072, -0.049, 0.0095, -0.0204, 0.0051, 0.0606, 0.0598, 0.0828, 0.0593, 0.0393, 0.0825, 0.1469, 0.1526, 0.1621, 0.1868, 0.1731, 0.1718, 0.2291, 0.1418, 0.2437, 0.2481, 0.2459, 0.3265, 0.2626, 0.2825, 0.2408, 0.2926, 0.267, 0.1966, 0.1831, 0.0545, 0.0859, 0.1727, 0.1758, 0.2058, 0.2058, 0.2384, 0.2592, 0.2541, 0.2262, 0.1869, 0.2422, 0.293, 0.3218, 0.381, 0.3535, 0.2628, 0.2437, 0.2176, 0.181, 0.3016, 0.3925, 0.4484, 0.4467, 0.5013, 0.5013, 0.4914, 0.5421, 0.533, 0.4952, 0.4952, 0.6524, 0.6353, 0.7992, 0.7788, 0.7132, 0.8079, 0.8119, 0.7714, 0.7464, 0.7635, 0.9004, 0.9004, 0.9122, 1.0385, 1.0828, 1.0937, 1.0384, 1.1492, 1.2804, 1.283, 1.1735, 1.2936, 1.1974, 0.9876, 1.0059, 1.0677, 1.0091, 0.9554, 1.1497, 1.1687, 1.1566, 1.1566, 1.0944, 1.2053, 1.1864, 1.2431, 1.2054, 1.1899, 1.2475, 1.177, 1.1603, 1.1619, 0.9891, 1.0996, 1.0801, 0.9399, 1.0396, 1.1118, 1.1934, 1.1236, 1.2324, 1.3145, 1.4187, 1.4189, 1.3274, 1.2155, 1.1183, 1.0721, 1.0017, 0.8622, 0.8714, 0.6866, 0.7706]}
{"unit_id": "thread:2002957111677083993", "ticker_idx": 0, "ts": "2025-12-22T04:19:23+00:00", "symbol": "Kinsus", "kind": "equity", "geography": "Taiwan", "sub_sector": "BT substrate", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Kinsus is the largest supplier of BT substrates for memory, accounting for 36% of related revenue", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory supercycle tightening BT substrate supply; Kinsus and Nanya PCB are primary beneficiaries via surging orders and quarterly price hikes through Q1 2026; Nittobo expanding fiberglass capacity.", "resolved_tickers": ["3189.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Kinsus 3189.TW Taiwan BT/ABF substrate", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Memory supercycle exposes T-Glass constraint, tightening BT substrate supply\n\n• As the global memory industry enters a supercycle driven by structural supply shortages, the supply shortage of BT substrates for memory chips is intensifying.\n\n• Taiwan IC substrate manufacturers Kinsus and Nanya PCB have secured urgent orders as customers seek to sign long-term supply agreements amid material shortages.\n\n• This has significantly improved order outlooks from the second half of 2025 and has acted as a major driver behind the increase in Taiwan PCB production in Q3.\n\n• According to the supply chain, shipments of ABF substrates for advanced AI packaging have been constrained by upstream shortages of T-Glass fiberglass fabric, and BT substrate lead times have been significantly extended due to material production capacity saturation.\n\n• These supply pressures are expected to ease only in the second half of 2026, when memory customers’ urgent order requirements can be handled more smoothly.\n\n• Currently, AI and memory customers view the situation as a short-term bottleneck caused by material shortages, and this is driving growth in long-term supply agreement signings to secure core ABF and BT substrate capacity.\n\n• BT substrates are mainly used in consumer electronics and have remained in oversupply in recent years due to sluggish smartphone demand.\n\n• However, the shortage of upstream T-Glass fiberglass and the surge in AI-related ABF substrate orders have accelerated BT substrate shortages.\n\n• Supply chain analysis indicates that Japan’s leading fiberglass fabric manufacturer, Nittobo, has confirmed capacity expansion, and end customers are reviewing the introduction of alternative suppliers.\n\n• BT substrate lead-time pressure is expected to begin easing in Q2 2026, and once new material production facilities are fully operational in 2027, they are expected to meet AI and memory customer demand.\n\n• Taiwan IC substrate manufacturers have been raising BT substrate prices on a quarterly basis since the end of Q2, with increases of about 3–5% or 5–10% per quarter.\n\n• Based on current order conditions, this price increase is expected to continue into Q1 2026, offsetting traditional off-season operating weakness resulting from declining consumer demand.\n\n• Industry analysis shows that among Taiwan’s three major IC substrate manufacturers, Kinsus is the largest supplier of BT substrates for memory, accounting for 36% of related revenue.\n\n• Nanya PCB follows, with BT substrates accounting for 30–35% of revenue, while Unimicron generates around 13% of its revenue from BT substrates but is focused on smartphones and game consoles, limiting its exposure to the memory market boom.\n\n---\n\nhttps://t.co/4B7SnWHenn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0625, "ret_signed_m1d": -0.0625, "alpha_spy_m1d": -0.05630867898466585, "alpha_c_m1d": -0.049781034051235906, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.012718965948764094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01041666666666663, "ret_signed_p1d": -0.01041666666666663, "alpha_spy_p1d": -0.014987121485099508, "alpha_c_p1d": -0.020035769634855316, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.009619102968188686, "ret_p1w": 0.04513888888888884, "ret_signed_p1w": 0.04513888888888884, "alpha_spy_p1w": 0.04072911761360132, "alpha_c_p1w": 0.03260072070708997, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.01253816818179887, "ret_p1m": 0.3863766988118489, "ret_signed_p1m": 0.3863766988118489, "alpha_spy_p1m": 0.3969633134973596, "alpha_c_p1m": 0.3010555940025943, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": 0.08532110480925459, "ret_p3m": 1.6631944444444446, "ret_signed_p3m": 1.6631944444444446, "alpha_spy_p3m": 1.69974369475473, "alpha_c_p3m": 1.5652795152660945, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": 0.0979149291783501, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2326, -0.2292, -0.2708, -0.2708, -0.2535, -0.2639, -0.2431, -0.2222, -0.2222, -0.2014, -0.1319, -0.0451, -0.0451, -0.0729, -0.1493, -0.066, 0.0243, 0.0486, 0.0486, 0.0417, 0.0174, 0.0069, 0.0069, 0.0312, -0.0451, -0.0347, -0.0451, -0.0521, -0.0347, -0.066, -0.0799, -0.0972, -0.1111, -0.1111, -0.1111, -0.0868, 0.0035, -0.0486, -0.0174, 0.0104, -0.066, -0.0312, -0.1042, -0.0833, -0.0451, -0.0417, -0.0208, -0.0104, -0.0174, -0.0208, 0.0278, 0.0278, 0.0903, 0.0833, 0.1076, 0.059, 0.0521, 0.0521, 0.0139, -0.0382, -0.0556, -0.0833, -0.0625, 0.0, -0.0104, 0.0, 0.0, 0.0625, 0.0451, 0.0278, 0.1042, 0.1042, 0.0833, 0.1667, 0.1771, 0.2153, 0.1528, 0.1701, 0.2569, 0.2257, 0.1944, 0.2153, 0.3368, 0.4199, 0.3864, 0.3427, 0.3797, 0.4367, 0.4444, 0.5868, 0.691, 0.5938, 0.75, 0.6632, 0.7083, 0.8785, 0.8299, 0.684, 0.7708, 0.7604, 0.8438, 0.8438, 0.8438, 0.8438, 0.8438, 0.8438, 0.8438, 0.8438, 1.0278, 1.2292, 1.2604, 1.184, 1.184, 1.1146, 1.1458, 0.9757, 1.066, 1.0243, 0.8368, 0.9028, 1.0174, 1.0069, 1.2049, 1.3403, 1.3715, 1.5382, 1.6632, 1.6389, 1.4028, 1.2257, 1.441, 1.441, 1.4444, 1.4062, 1.1736, 1.3056, 1.3333, 1.3333, 1.3333, 1.4896, 1.7361, 1.6875, 1.6458, 1.5451, 1.5764, 1.6632, 1.7326, 1.7188, 1.9896, 2.2882, 2.4132, 2.3264, 2.6528, 2.5764, 2.5278, 2.5278, 2.6667, 2.6667, 2.6875, 2.6597, 2.3507, 2.4201, 2.1806, 2.4931, 2.4306, 2.4549, 2.625, 2.375, 2.5069, 2.5625, 2.5, 2.8472, 3.2292, 3.6458, 3.8333, 3.875, 3.7292, 4.0625, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2002957111677083993", "ticker_idx": 1, "ts": "2025-12-22T04:19:23+00:00", "symbol": "Nanya PCB", "kind": "equity", "geography": "Taiwan", "sub_sector": "BT substrate", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nanya PCB follows, with BT substrates accounting for 30–35% of revenue", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory supercycle tightening BT substrate supply; Kinsus and Nanya PCB are primary beneficiaries via surging orders and quarterly price hikes through Q1 2026; Nittobo expanding fiberglass capacity.", "resolved_tickers": ["8046.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya PCB = Nan Ya PCB, TWSE 8046.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Memory supercycle exposes T-Glass constraint, tightening BT substrate supply\n\n• As the global memory industry enters a supercycle driven by structural supply shortages, the supply shortage of BT substrates for memory chips is intensifying.\n\n• Taiwan IC substrate manufacturers Kinsus and Nanya PCB have secured urgent orders as customers seek to sign long-term supply agreements amid material shortages.\n\n• This has significantly improved order outlooks from the second half of 2025 and has acted as a major driver behind the increase in Taiwan PCB production in Q3.\n\n• According to the supply chain, shipments of ABF substrates for advanced AI packaging have been constrained by upstream shortages of T-Glass fiberglass fabric, and BT substrate lead times have been significantly extended due to material production capacity saturation.\n\n• These supply pressures are expected to ease only in the second half of 2026, when memory customers’ urgent order requirements can be handled more smoothly.\n\n• Currently, AI and memory customers view the situation as a short-term bottleneck caused by material shortages, and this is driving growth in long-term supply agreement signings to secure core ABF and BT substrate capacity.\n\n• BT substrates are mainly used in consumer electronics and have remained in oversupply in recent years due to sluggish smartphone demand.\n\n• However, the shortage of upstream T-Glass fiberglass and the surge in AI-related ABF substrate orders have accelerated BT substrate shortages.\n\n• Supply chain analysis indicates that Japan’s leading fiberglass fabric manufacturer, Nittobo, has confirmed capacity expansion, and end customers are reviewing the introduction of alternative suppliers.\n\n• BT substrate lead-time pressure is expected to begin easing in Q2 2026, and once new material production facilities are fully operational in 2027, they are expected to meet AI and memory customer demand.\n\n• Taiwan IC substrate manufacturers have been raising BT substrate prices on a quarterly basis since the end of Q2, with increases of about 3–5% or 5–10% per quarter.\n\n• Based on current order conditions, this price increase is expected to continue into Q1 2026, offsetting traditional off-season operating weakness resulting from declining consumer demand.\n\n• Industry analysis shows that among Taiwan’s three major IC substrate manufacturers, Kinsus is the largest supplier of BT substrates for memory, accounting for 36% of related revenue.\n\n• Nanya PCB follows, with BT substrates accounting for 30–35% of revenue, while Unimicron generates around 13% of its revenue from BT substrates but is focused on smartphones and game consoles, limiting its exposure to the memory market boom.\n\n---\n\nhttps://t.co/4B7SnWHenn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018442622950819665, "ret_signed_m1d": -0.018442622950819665, "alpha_spy_m1d": -0.012251301935485515, "alpha_c_m1d": -0.013076251370739644, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.0053663715800800205, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020491803278688492, "ret_signed_p1d": -0.020491803278688492, "alpha_spy_p1d": -0.02506225809712137, "alpha_c_p1d": -0.025934058476663635, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.005442255197975143, "ret_p1w": -0.014344262295082011, "ret_signed_p1w": -0.014344262295082011, "alpha_spy_p1w": -0.018754033570369533, "alpha_c_p1w": -0.01923544034562452, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.004891178050542511, "ret_p1m": 0.39959016393442615, "ret_signed_p1m": 0.39959016393442615, "alpha_spy_p1m": 0.41017677861993684, "alpha_c_p1m": 0.42246144426892873, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": -0.022871280334502586, "ret_p3m": 1.2991803278688523, "ret_signed_p3m": 1.2991803278688523, "alpha_spy_p3m": 1.3357295781791376, "alpha_c_p3m": 1.3455430082003237, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": -0.046362680331471395, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0963, -0.0615, -0.1025, -0.1025, -0.0738, -0.0656, -0.0533, 0.041, 0.041, 0.0369, 0.0676, 0.0861, 0.0861, 0.0963, 0.0492, 0.1537, 0.1537, 0.1496, 0.1373, 0.1209, 0.1291, 0.1168, 0.1168, 0.1414, 0.1311, 0.1906, 0.1557, 0.1906, 0.168, 0.1475, 0.1148, 0.1189, 0.1189, 0.1906, 0.1844, 0.166, 0.1967, 0.1578, 0.1168, 0.0881, 0.0287, 0.1025, 0.0246, 0.0123, 0.0246, 0.0533, 0.0676, 0.0758, 0.0922, 0.0738, 0.1025, 0.0594, 0.1066, 0.1066, 0.1639, 0.1475, 0.1393, 0.1127, 0.0902, 0.0205, -0.002, -0.0389, -0.0184, 0.0, -0.0205, -0.0205, -0.0205, 0.0102, -0.0143, -0.0143, -0.0123, -0.0123, -0.0287, -0.0102, 0.002, -0.0102, -0.0287, -0.0143, 0.043, 0.0984, 0.0717, 0.0799, 0.1865, 0.3033, 0.3996, 0.4795, 0.5184, 0.5656, 0.5164, 0.6189, 0.623, 0.5328, 0.5922, 0.5082, 0.6168, 0.6803, 0.6291, 0.5533, 0.707, 0.6783, 0.7049, 0.7049, 0.7049, 0.7049, 0.7049, 0.7049, 0.7049, 0.7049, 0.875, 1.0615, 1.2664, 1.2746, 1.2746, 1.2172, 1.0492, 0.8443, 0.8893, 0.9242, 0.7336, 0.7541, 0.9283, 0.9201, 1.1107, 1.3197, 1.2295, 1.3238, 1.2992, 1.2172, 1.1885, 1.0123, 1.2131, 1.332, 1.2992, 1.377, 1.1393, 1.3525, 1.3279, 1.3279, 1.3279, 1.5574, 1.6107, 1.7418, 1.7336, 1.6844, 1.6475, 1.6885, 1.9549, 1.9672, 2.2623, 2.2336, 2.291, 2.2582, 2.582, 2.7418, 2.8197, 2.7541, 3.1189, 3.1189, 3.041, 3.0861, 2.8525, 2.8811, 2.5, 2.7213, 2.6885, 2.6066, 2.4713, 2.2213, 2.2418, 2.2705, 2.3975, 2.5738, 2.8279, 2.623, 2.7787, 2.709, 2.4836, 2.4754, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2002957111677083993", "ticker_idx": 2, "ts": "2025-12-22T04:19:23+00:00", "symbol": "Nittobo", "kind": "equity", "geography": "Japan", "sub_sector": "T-Glass fiberglass", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Japan's leading fiberglass fabric manufacturer, Nittobo, has confirmed capacity expansion", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory supercycle tightening BT substrate supply; Kinsus and Nanya PCB are primary beneficiaries via surging orders and quarterly price hikes through Q1 2026; Nittobo expanding fiberglass capacity.", "resolved_tickers": ["3110.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Nittobo = Nitto Boseki, TSE 3110.T", "bench_c": "EWJ", "bench_c_reason": "country_fallback (suffix='T')", "bench_c_industry": "Textile Manufacturing", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Memory supercycle exposes T-Glass constraint, tightening BT substrate supply\n\n• As the global memory industry enters a supercycle driven by structural supply shortages, the supply shortage of BT substrates for memory chips is intensifying.\n\n• Taiwan IC substrate manufacturers Kinsus and Nanya PCB have secured urgent orders as customers seek to sign long-term supply agreements amid material shortages.\n\n• This has significantly improved order outlooks from the second half of 2025 and has acted as a major driver behind the increase in Taiwan PCB production in Q3.\n\n• According to the supply chain, shipments of ABF substrates for advanced AI packaging have been constrained by upstream shortages of T-Glass fiberglass fabric, and BT substrate lead times have been significantly extended due to material production capacity saturation.\n\n• These supply pressures are expected to ease only in the second half of 2026, when memory customers’ urgent order requirements can be handled more smoothly.\n\n• Currently, AI and memory customers view the situation as a short-term bottleneck caused by material shortages, and this is driving growth in long-term supply agreement signings to secure core ABF and BT substrate capacity.\n\n• BT substrates are mainly used in consumer electronics and have remained in oversupply in recent years due to sluggish smartphone demand.\n\n• However, the shortage of upstream T-Glass fiberglass and the surge in AI-related ABF substrate orders have accelerated BT substrate shortages.\n\n• Supply chain analysis indicates that Japan’s leading fiberglass fabric manufacturer, Nittobo, has confirmed capacity expansion, and end customers are reviewing the introduction of alternative suppliers.\n\n• BT substrate lead-time pressure is expected to begin easing in Q2 2026, and once new material production facilities are fully operational in 2027, they are expected to meet AI and memory customer demand.\n\n• Taiwan IC substrate manufacturers have been raising BT substrate prices on a quarterly basis since the end of Q2, with increases of about 3–5% or 5–10% per quarter.\n\n• Based on current order conditions, this price increase is expected to continue into Q1 2026, offsetting traditional off-season operating weakness resulting from declining consumer demand.\n\n• Industry analysis shows that among Taiwan’s three major IC substrate manufacturers, Kinsus is the largest supplier of BT substrates for memory, accounting for 36% of related revenue.\n\n• Nanya PCB follows, with BT substrates accounting for 30–35% of revenue, while Unimicron generates around 13% of its revenue from BT substrates but is focused on smartphones and game consoles, limiting its exposure to the memory market boom.\n\n---\n\nhttps://t.co/4B7SnWHenn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05457909141701034, "ret_signed_m1d": -0.05457909141701034, "alpha_spy_m1d": -0.04838777040167619, "alpha_c_m1d": -0.05607054988019078, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": 0.0014914584631804395, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0074005655364046685, "ret_signed_p1d": -0.0074005655364046685, "alpha_spy_p1d": -0.011971020354837547, "alpha_c_p1d": -0.015851973850017353, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.008451408313612685, "ret_p1w": -0.041628169797422676, "ret_signed_p1w": -0.041628169797422676, "alpha_spy_p1w": -0.0460379410727102, "alpha_c_p1w": -0.04945811337915884, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.007829943581736165, "ret_p1m": 0.4153561394534795, "ret_signed_p1m": 0.4153561394534795, "alpha_spy_p1m": 0.4259427541389902, "alpha_c_p1m": 0.3748390853480241, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": 0.04051705410545536, "ret_p3m": 0.9962997626112113, "ret_signed_p3m": 0.9962997626112113, "alpha_spy_p3m": 1.032849012921497, "alpha_c_p3m": 0.9513084279384196, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": 0.044991334672791705, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3718, -0.3958, -0.4133, -0.3922, -0.3895, -0.4061, -0.3247, -0.3284, -0.3071, -0.3145, -0.2784, -0.2646, -0.2979, -0.2979, -0.3293, -0.3191, -0.2895, -0.3043, -0.2692, -0.2942, -0.3173, -0.3228, -0.2636, -0.2488, -0.2618, -0.2368, -0.2118, -0.1906, -0.1906, -0.2017, -0.2424, -0.234, -0.0953, 0.0435, 0.1341, 0.2488, 0.2599, 0.2377, 0.3182, 0.2655, 0.2507, 0.4024, 0.3256, 0.3256, 0.3099, 0.2285, 0.2525, 0.2636, 0.259, 0.1573, 0.0962, 0.0749, 0.1369, 0.0971, 0.0564, 0.0879, 0.0611, 0.074, 0.025, -0.0296, -0.0185, -0.0398, -0.0546, 0.0, -0.0074, -0.0139, -0.0231, -0.0564, -0.0416, -0.0564, -0.0564, -0.0564, -0.0564, -0.0315, -0.0241, -0.0111, 0.099, 0.16, 0.16, 0.1656, 0.2081, 0.2322, 0.3293, 0.4366, 0.4154, 0.4884, 0.6263, 0.5791, 0.4533, 0.4283, 0.4154, 0.371, 0.3719, 0.3256, 0.4061, 0.5162, 0.5893, 0.5634, 0.6475, 0.7382, 0.7382, 0.8927, 0.8677, 0.9158, 0.975, 0.9565, 0.9658, 0.9611, 0.9611, 1.4237, 1.5606, 1.3562, 1.3312, 1.3247, 1.1582, 1.0111, 1.1462, 1.2026, 0.913, 0.9584, 1.0703, 1.0074, 1.0481, 1.0241, 0.9241, 1.0629, 0.9963, 0.9963, 0.8612, 0.8594, 1.0685, 1.0176, 0.9889, 0.8364, 0.7416, 0.9396, 0.8225, 0.9034, 0.9294, 1.013, 1.3095, 1.3504, 1.5493, 1.6915, 1.7361, 1.4628, 1.513, 1.5567, 1.4833, 1.4591, 1.5521, 1.3745, 1.5446, 1.5418, 1.6218, 1.6218, 1.6524, 1.6171, 1.6171, 1.6171, 1.6171, 1.9693, 1.8253, 1.7035, 1.6952, 1.3968, 1.434, 1.2212, 1.1125, 1.0056, 0.96, 1.0465, 1.3522, 1.5613, 1.4061, 1.1654, 1.1106, 1.1069, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1980414975652118795", "ticker_idx": 0, "ts": "2025-10-20T23:24:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price raised from ₩480,000 → ₩640,000, applying FY26(F) P/B 2.8x target", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi raises SK Hynix TP to ₩640K (from ₩480K) and FY26 OP estimate to ₩81.5T vs ₩55T consensus on AI inference memory demand surge.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251022_SK Hynix: Opening Another Upside Catalyst Watch - CITI\n\n⸻\n\n[Key Takeaways]\n\n(1) FY26(F) operating profit estimate revised upward from ₩64.1T → ₩81.5T vs. ₩55.0T consensus\n■ Target price raised from ₩480,000 → ₩640,000, applying FY26(F) P/B 2.8x target\n\n(2) 4Q25(F) operating profit estimate revised upward from ₩12.2T → ₩15.0T vs. ₩12.5T consensus\n\n(3) The revision reflects:\n    1. a surge in tokenized data generated by AI Agents,\n    2. worsening supply shortage in legacy products, and\n    3. rapidly rising AI inference memory demand in China.\n\n⸻\n\n[Contents]\n\n(1) We expect AI inference memory demand to surge not only in the U.S. but also in China.\n\n(2) Chinese hyperscalers, in particular, are projected to make concentrated investments in AI inference infrastructure.\n\n(3) Accordingly, we have raised our 4Q25(F) server DRAM and eSSD price forecasts versus previous estimates.\n\n(4) Server DRAM [64GB DDR5 RDIMM] price forecast raised from $261 → $360.\n\n(5) The upward revision for eSSD is due to the explosive increase in the amount of tokenized data generated from AI Agent operations.\n\n(6) Reflecting these trends, we have raised SK hynix’s 4Q25(F) operating profit from ₩12.2T → ₩15.0T.\n\n(7) FY26(F) DRAM ASP: YoY +37% vs. prior +25%\n■ Driven by a sharp rise in AI inference memory demand in China and deepening legacy memory supply shortages.\n\n(8) FY26(F) DRAM QoQ growth:\n■ 1Q26(F): +10% vs. prior +3%\n■ 2Q26(F): +7% vs. prior +8%\n■ 3Q26(F): +3% vs. prior +5%\n■ 4Q26(F): +1% vs. prior +2%\n\n(9) FY26(F) NAND ASP: YoY +39% vs. prior +23%\n■ 1Q26(F): +11% vs. prior +3%\n■ 2Q26(F): +9% vs. prior +9%\n■ 3Q26(F): +5% vs. prior +7%\n■ 4Q26(F): +3% vs. prior +5%\n\n(10) As a result, FY26(F) operating profit estimate is raised from ₩64.1T → ₩81.5T, and target price from ₩480,000 → ₩640,000, applying FY26(F) P/B 2.8x target.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04119462522047679, "ret_signed_m1d": -0.04119462522047679, "alpha_spy_m1d": -0.030901206271412796, "alpha_c_m1d": -0.028551282693331737, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013388283854285032, "ret_signed_p1d": -0.013388283854285032, "alpha_spy_p1d": -0.013373287934272504, "alpha_c_p1d": -0.007858595812623381, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.1019566925800015, "ret_signed_p1w": 0.1019566925800015, "alpha_spy_p1w": 0.08119100090416365, "alpha_c_p1w": 0.0653228619626689, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": 0.17404730285143066, "ret_signed_p1m": 0.17404730285143066, "alpha_spy_p1m": 0.1907610786521854, "alpha_c_p1m": 0.2141659301769061, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.5438442835073052, "ret_signed_p3m": 0.5438442835073052, "alpha_spy_p3m": 0.5096040261919479, "alpha_c_p3m": 0.39862376074724315, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 1.2777070106448454, "ret_signed_p6m": 1.2777070106448454, "alpha_spy_p6m": 1.2373163230380224, "alpha_c_p6m": 0.9718881072228924, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.4467, -0.4457, -0.4529, -0.4611, -0.4601, -0.458, -0.4375, -0.4694, -0.4694, -0.458, -0.4683, -0.4611, -0.4724, -0.4508, -0.4467, -0.4282, -0.4313, -0.4313, -0.4498, -0.4591, -0.4745, -0.4961, -0.4838, -0.4663, -0.4622, -0.4652, -0.447, -0.4459, -0.4727, -0.4634, -0.4593, -0.4531, -0.4367, -0.4295, -0.4068, -0.3738, -0.3677, -0.3234, -0.3182, -0.2832, -0.3131, -0.2678, -0.2678, -0.277, -0.2564, -0.2636, -0.2657, -0.3069, -0.2812, -0.2842, -0.2585, -0.1854, -0.1854, -0.1854, -0.1854, -0.1854, -0.1854, -0.1184, -0.1452, -0.1524, -0.1298, -0.068, -0.0412, 0.0, -0.0134, -0.0082, -0.0144, 0.0505, 0.102, 0.0731, 0.1493, 0.1699, 0.1514, 0.277, 0.207, 0.1926, 0.2214, 0.1946, 0.2482, 0.275, 0.2709, 0.2606, 0.1535, 0.2482, 0.174, 0.1576, 0.1761, 0.0731, 0.0711, 0.069, 0.0793, 0.1213, 0.0924, 0.1089, 0.1502, 0.1378, 0.1172, 0.1213, 0.1893, 0.1666, 0.2099, 0.1646, 0.1769, 0.1419, 0.0924, 0.1357, 0.1378, 0.1275, 0.1955, 0.2037, 0.212, 0.212, 0.2347, 0.3192, 0.3418, 0.3418, 0.3418, 0.3954, 0.4346, 0.4964, 0.5294, 0.5583, 0.5335, 0.5438, 0.5212, 0.5294, 0.5438, 0.5583, 0.5748, 0.5315, 0.5253, 0.5562, 0.5809, 0.517, 0.649, 0.7335, 0.7747, 0.8736, 0.7108, 0.8695, 0.8551, 0.7355, 0.7294, 0.8283, 0.8056, 0.7726, 0.8304, 0.8139, 0.8139, 0.8139, 0.8139, 0.8427, 0.9561, 0.9602, 1.0715, 1.0983, 1.2694, 1.191, 1.191, 0.939, 0.7532, 0.9432, 0.9081, 0.7263, 0.937, 0.9721, 0.9205, 0.8792, 1.0113, 1.0031, 1.1807, 1.0919, 1.0795, 0.9267, 1.0361, 1.0547, 0.9267, 0.9267, 0.8028, 0.6665, 0.8523, 0.714, 0.8089, 0.8296, 0.8915, 1.1332, 1.0609, 1.1208, 1.1476, 1.2777]}
{"unit_id": "orig:1977539642707853711", "ticker_idx": 0, "ts": "2025-10-13T00:59:26+00:00", "symbol": "memory semiconductors", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory companies should remain Overweight during this period of simultaneous earnings upgrades and rate cuts", "tweet_type": "news_reaction", "timeline": "years", "summary": "Samsung Securities reiterates Overweight on memory semiconductor sector: dual profit engines, AI-driven structural demand, and rate cuts make this cycle different from prior ones.", "resolved_tickers": ["000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Korea memory semicon basket: SK Hynix, Samsung Electronics", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Why This Memory Semiconductor Cycle Is Different (Samsung Securities, Korea)\n\nThis memory cycle differs from previous ones in three key aspects.\nFirst, there are two profit growth engines. Although demand and supply for AI-focused HBM and general DRAM are connected, they do not move in perfect synchrony. The recent rally has been driven by a mismatch between AI demand and general DRAM demand.\nSecond, the main demand source — AI infrastructure investment — is larger and more prolonged than in any prior cycle. It is the first time in 30 years, since the Internet infrastructure boom, that we have seen a trend of continuous growth for three consecutive years.\nThird, memory demand is no longer correlated with the consumer economy. The prolonged weakness in smartphones and PCs repeatedly leads to underestimation of memory price hikes during this cycle.\n\nRelentless Demand Outlook\nSo far, memory companies’ share prices have followed the typical pattern of rising profits alongside declining valuations. A rapid profit increase usually signals the approach of a cycle peak.\nHowever, the current rally — where both profit upgrades and valuation re-rating are occurring simultaneously — evokes the early phase of a new cycle. This phenomenon seems driven by the emergence of massive, seemingly “out of nowhere” demand — sometimes worth $100 billion — whenever capital is available.\nTraditionally, memory demand has been judged based on customers’ ROI. But in this cycle, demand is being dictated less by profitability and more by the preemptive competition and fundraising capacity of Silicon Valley’s mega-tech companies.\n\nB2B Demand and Interest Rates\nThis kind of demand surge is not entirely new. Previous B2B infrastructure cycles have shown similar patterns — the Internet infrastructure boom in 1999 and Amazon’s data-center expansion in 2017 both seemed to have no visible end to demand.\nYet, those cycles were preceded by economic expansion and rising interest rates. This time is different: as the cycle enters its third year of growth, interest rates are falling. The conclusion is that memory companies should remain Overweight during this period of simultaneous earnings upgrades and rate cuts.\n\nSeptember Inflection Point Confirming Long-Term AI Growth\nThe key feature of the AI theme in September was the reinforcement of its long-term growth narrative. Until recently, investors had compared cloud providers’ AI investments for this year and next with projected AI revenue growth — repeatedly worrying about bubbles.\nNow, however, Big Tech companies are publicly discussing multi-hundred-trillion-won data-center clusters and multi-year semiconductor procurement contracts.\nInvestor conviction in 5-year-plus structural growth for semiconductors is strengthening.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.021557585193363815, "ret_signed_m1d": 0.021557585193363815, "alpha_spy_m1d": 0.03666977728238052, "alpha_c_m1d": 0.0639711091680939, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.04241352397473008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013327261050920414, "ret_signed_p1d": -0.013327261050920414, "alpha_spy_p1d": -0.012105606247463707, "alpha_c_p1d": 0.005068553694888789, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.018395814745809203, "ret_p1w": 0.11066323911870957, "ret_signed_p1w": 0.11066323911870957, "alpha_spy_p1w": 0.09820543370176849, "alpha_c_p1w": 0.08867645436408644, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.021986784754623123, "ret_p1m": 0.3004454447095112, "ret_signed_p1m": 0.3004454447095112, "alpha_spy_p1m": 0.27034168922266477, "alpha_c_p1m": 0.2671856775756357, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.03325976713387546, "ret_p3m": 0.6589486243314387, "ret_signed_p3m": 0.6589486243314387, "alpha_spy_p3m": 0.615953522245127, "alpha_c_p3m": 0.5400104248045522, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.11893819952688656, "ret_p6m": 1.1668067087103218, "ret_signed_p6m": 1.1668067087103218, "alpha_spy_p6m": 1.1669068777386138, "alpha_c_p6m": 0.9861022752580584, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.1807044334522634, "price_path": [-0.2987, -0.3199, -0.3184, -0.3104, -0.3248, -0.3221, -0.3236, -0.3284, -0.3092, -0.3075, -0.2956, -0.29, -0.322, -0.3177, -0.3101, -0.3219, -0.3087, -0.3083, -0.3, -0.297, -0.2819, -0.2853, -0.2853, -0.3047, -0.3101, -0.3165, -0.3286, -0.3171, -0.3063, -0.3103, -0.3105, -0.3052, -0.304, -0.3309, -0.3175, -0.3113, -0.3061, -0.2997, -0.2923, -0.2715, -0.2464, -0.2385, -0.2019, -0.1931, -0.1571, -0.181, -0.1433, -0.1433, -0.1316, -0.1132, -0.1136, -0.1111, -0.1502, -0.1283, -0.1317, -0.1054, -0.0425, -0.0425, -0.0425, -0.0425, -0.0425, -0.0425, 0.0216, 0.0, -0.0133, 0.0181, 0.0688, 0.0855, 0.1107, 0.0996, 0.1085, 0.0937, 0.1439, 0.1912, 0.1609, 0.2109, 0.2422, 0.2496, 0.3424, 0.2682, 0.2367, 0.2461, 0.2234, 0.2692, 0.3004, 0.2959, 0.2883, 0.1956, 0.2692, 0.2109, 0.1943, 0.2271, 0.1357, 0.1447, 0.1575, 0.1822, 0.2106, 0.1776, 0.1888, 0.2269, 0.2256, 0.2167, 0.2368, 0.2825, 0.2633, 0.2865, 0.2562, 0.272, 0.2296, 0.1899, 0.2426, 0.2422, 0.2292, 0.2915, 0.3017, 0.3043, 0.3043, 0.3492, 0.4152, 0.4306, 0.4306, 0.4306, 0.5082, 0.5828, 0.6233, 0.6539, 0.6589, 0.6456, 0.6505, 0.6308, 0.6501, 0.678, 0.7133, 0.7251, 0.6777, 0.6973, 0.7304, 0.7438, 0.7065, 0.8235, 0.8885, 0.9035, 0.9603, 0.8106, 0.9956, 0.9957, 0.873, 0.8656, 0.9655, 0.949, 0.9405, 1.0324, 1.0368, 1.0368, 1.0368, 1.0368, 1.1011, 1.1679, 1.1859, 1.2887, 1.3233, 1.5014, 1.4475, 1.4475, 1.1849, 0.9528, 1.1684, 1.1296, 0.9441, 1.1449, 1.1767, 1.1352, 1.0874, 1.1927, 1.2158, 1.3983, 1.3033, 1.2902, 1.1302, 1.2126, 1.2197, 1.0968, 1.0968, 1.0059, 0.877, 1.1069, 0.9653, 1.0629, 1.1122, 1.1668]}
{"unit_id": "orig:1975771979597885443", "ticker_idx": 0, "ts": "2025-10-08T03:55:22+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "earnings estimates and target price raised… We raise our 12-month target price from $150 to $210", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Sachs raising AMD PT $150→$210 on OpenAI partnership; AVGO seen neutral-to-mildly positive from spending uplift.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) AMD: The strategic partnership with OpenAI is a strong long-term catalyst for AMD’s GPU business momentum; earnings estimates and target price raised\n\nOur View and Potential Industry Impact\nWe view this announcement as a significant positive development for AMD’s competitive positioning in the accelerator market for both training and inference workloads. The expectation that AMD will participate in part of OpenAI’s training stack is encouraging, as it slightly undermines Nvidia’s (NVDA) dominant share in OpenAI’s training workloads to date. For Broadcom (AVGO), the impact is seen as neutral to mildly positive, considering the expected increase in spending levels and the potential boost to its networking business. Our assumptions for OpenAI’s contribution are detailed further in Exhibit 1. Based on our analysis, AMD’s contribution to OpenAI (via equity) could reach approximately $75 billion, while generating around $135 billion in revenue. Accordingly, we are revising our estimates and target price upward. The key assumptions used in our analysis are as follows:\n    \n1. We assume that for each gigawatt (GW) deployed, the same number of shares (i.e., 160 million shares per GW tranche) is issued evenly.\n    \n2. Out of the total 6 GW planned, we assume deployments of 0.25 / 1 / 1.25 / 1.5 / 2 GW from 2026 through 2030, respectively.\n    \n3. We also assume AMD’s stock price will be $225 in 2026, $300 in 2027, and will increase linearly to $600 by 2030. Under these assumptions, all else being equal, this equates to approximately $75 billion worth of AMD equity granted to OpenAI.\n\nBased on industry benchmarks and our average selling price (ASP) assumptions, we estimate a revenue opportunity of $22.5 billion per GW for AMD.\n\nEarnings Estimate Revisions\nReflecting the incremental contribution from the strategic partnership with OpenAI, we raise our 2026 and 2027 non-GAAP EPS (excluding stock-based compensation) estimates by 21% and 62%, respectively.\n\nInvestment Thesis, Target Price, and Risks\nWe believe this announcement strengthens AMD’s competitive position in the accelerator market, representing a net positive for the company. However, two key risks remain:\n\n(1) High customer concentration, with OpenAI expected to account for over 40% of AMD’s total revenue in 2027 according to our estimates; and\n(2) Uncertainty over OpenAI’s additional funding requirements, which, as previously discussed, could pose potential execution challenges and may lead to a discount being applied to this revenue stream. As confidence in the revenue flow and execution timeline for this contract improves over the next few quarters, our outlook could turn more constructive.\n\nWe raise our 12-month target price from $150 to $210, reflecting higher earnings estimates. This is based on increasing our normalized EPS estimate from $6.00 to $7.00 and raising the applied multiple from 25x to 30x, factoring in a higher mix of data center revenue.\n\nKey Upside Risks:\n(1) Stronger-than-expected demand for AMD GPUs,\n(2) Better-than-expected share trends in server x86 architecture,\n(3) Greater operating expense leverage.\n\nKey Downside Risks:\n(1) Weaker-than-expected demand for AMD GPUs,\n(2) Sharper-than-expected share loss in the server CPU market.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10209714425631866, "ret_signed_m1d": -0.10209714425631866, "alpha_spy_m1d": -0.09616943906029773, "alpha_c_m1d": -0.07603301211334212, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.026064132142976537, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011334684142544904, "ret_signed_p1d": -0.011334684142544904, "alpha_spy_p1d": -0.00843767576330745, "alpha_c_p1d": -0.008849555198727321, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0024851289438175828, "ret_p1w": 0.012905453287609614, "ret_signed_p1w": 0.012905453287609614, "alpha_spy_p1w": 0.02470140606622717, "alpha_c_p1w": 0.025301818527790676, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.012396365240181062, "ret_p1m": 0.009084731753387532, "ret_signed_p1m": 0.009084731753387532, "alpha_spy_p1m": 0.013244484876622109, "alpha_c_p1m": -0.005420994375035626, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": 0.014505726128423158, "ret_p3m": -0.0614705207913635, "ret_signed_p3m": -0.0614705207913635, "alpha_spy_p3m": -0.08619476559076023, "alpha_c_p3m": -0.15594531505216103, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": 0.09447479426079752, "ret_p6m": -0.07666835517818116, "ret_signed_p6m": -0.07666835517818116, "alpha_spy_p6m": -0.05654432404318366, "alpha_c_p6m": -0.21383926400789888, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": 0.13717090882971772, "price_path": [-0.3784, -0.3792, -0.3394, -0.3204, -0.319, -0.3335, -0.3335, -0.3432, -0.3265, -0.3118, -0.2933, -0.2628, -0.2467, -0.2379, -0.2515, -0.2711, -0.2495, -0.26, -0.3075, -0.2681, -0.2666, -0.2686, -0.2573, -0.2171, -0.2318, -0.2464, -0.2522, -0.293, -0.2987, -0.305, -0.2878, -0.3065, -0.2927, -0.2905, -0.2843, -0.3096, -0.3096, -0.3109, -0.3117, -0.3132, -0.3584, -0.3572, -0.3385, -0.3227, -0.3391, -0.3268, -0.3158, -0.3188, -0.3243, -0.3296, -0.3318, -0.3217, -0.3169, -0.317, -0.3154, -0.3231, -0.315, -0.3132, -0.3037, -0.2795, -0.3009, -0.1352, -0.1021, 0.0, -0.0113, -0.0877, -0.0813, -0.0742, 0.0129, -0.0042, -0.0105, 0.0212, 0.0105, -0.0226, -0.0024, 0.0737, 0.1024, 0.0953, 0.1221, 0.0818, 0.0873, 0.1023, 0.0615, 0.0882, 0.0091, -0.0086, 0.0357, 0.0083, 0.099, 0.0526, 0.0478, 0.0211, -0.0224, -0.051, -0.1254, -0.1349, -0.0871, -0.1249, -0.0905, -0.0905, -0.0765, -0.0671, -0.0863, -0.0762, -0.0831, -0.0747, -0.0613, -0.0592, -0.06, -0.06, -0.1052, -0.1188, -0.112, -0.159, -0.1465, -0.0939, -0.0875, -0.0877, -0.0871, -0.0871, -0.0873, -0.0847, -0.0858, -0.0908, -0.0908, -0.0513, -0.0615, -0.09, -0.1084, -0.1311, -0.1375, -0.1183, -0.0619, -0.0508, -0.0324, -0.0158, -0.0158, -0.0155, 0.0605, 0.0771, 0.1024, 0.0669, 0.0699, 0.0729, 0.0706, 0.005, 0.0455, 0.0278, -0.1502, -0.1828, -0.1151, -0.083, -0.0934, -0.0933, -0.1257, -0.1199, -0.1199, -0.1379, -0.1505, -0.1367, -0.1503, -0.1654, -0.0922, -0.1049, -0.1353, -0.1501, -0.1568, -0.1894, -0.1422, -0.1533, -0.1831, -0.1396, -0.1372, -0.1305, -0.1606, -0.179, -0.1655, -0.1666, -0.1533, -0.1286, -0.1453, -0.1396, -0.1282, -0.0649, -0.135, -0.1425, -0.1678, -0.1364, -0.1076, -0.0767]}
{"unit_id": "orig:1975771979597885443", "ticker_idx": 1, "ts": "2025-10-08T03:55:22+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "For Broadcom (AVGO), the impact is seen as neutral to mildly positive, considering the expected increase in spending levels and the potential boost to its networking business", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Goldman Sachs raising AMD PT $150→$210 on OpenAI partnership; AVGO seen neutral-to-mildly positive from spending uplift.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs) AMD: The strategic partnership with OpenAI is a strong long-term catalyst for AMD’s GPU business momentum; earnings estimates and target price raised\n\nOur View and Potential Industry Impact\nWe view this announcement as a significant positive development for AMD’s competitive positioning in the accelerator market for both training and inference workloads. The expectation that AMD will participate in part of OpenAI’s training stack is encouraging, as it slightly undermines Nvidia’s (NVDA) dominant share in OpenAI’s training workloads to date. For Broadcom (AVGO), the impact is seen as neutral to mildly positive, considering the expected increase in spending levels and the potential boost to its networking business. Our assumptions for OpenAI’s contribution are detailed further in Exhibit 1. Based on our analysis, AMD’s contribution to OpenAI (via equity) could reach approximately $75 billion, while generating around $135 billion in revenue. Accordingly, we are revising our estimates and target price upward. The key assumptions used in our analysis are as follows:\n    \n1. We assume that for each gigawatt (GW) deployed, the same number of shares (i.e., 160 million shares per GW tranche) is issued evenly.\n    \n2. Out of the total 6 GW planned, we assume deployments of 0.25 / 1 / 1.25 / 1.5 / 2 GW from 2026 through 2030, respectively.\n    \n3. We also assume AMD’s stock price will be $225 in 2026, $300 in 2027, and will increase linearly to $600 by 2030. Under these assumptions, all else being equal, this equates to approximately $75 billion worth of AMD equity granted to OpenAI.\n\nBased on industry benchmarks and our average selling price (ASP) assumptions, we estimate a revenue opportunity of $22.5 billion per GW for AMD.\n\nEarnings Estimate Revisions\nReflecting the incremental contribution from the strategic partnership with OpenAI, we raise our 2026 and 2027 non-GAAP EPS (excluding stock-based compensation) estimates by 21% and 62%, respectively.\n\nInvestment Thesis, Target Price, and Risks\nWe believe this announcement strengthens AMD’s competitive position in the accelerator market, representing a net positive for the company. However, two key risks remain:\n\n(1) High customer concentration, with OpenAI expected to account for over 40% of AMD’s total revenue in 2027 according to our estimates; and\n(2) Uncertainty over OpenAI’s additional funding requirements, which, as previously discussed, could pose potential execution challenges and may lead to a discount being applied to this revenue stream. As confidence in the revenue flow and execution timeline for this contract improves over the next few quarters, our outlook could turn more constructive.\n\nWe raise our 12-month target price from $150 to $210, reflecting higher earnings estimates. This is based on increasing our normalized EPS estimate from $6.00 to $7.00 and raising the applied multiple from 25x to 30x, factoring in a higher mix of data center revenue.\n\nKey Upside Risks:\n(1) Stronger-than-expected demand for AMD GPUs,\n(2) Better-than-expected share trends in server x86 architecture,\n(3) Greater operating expense leverage.\n\nKey Downside Risks:\n(1) Weaker-than-expected demand for AMD GPUs,\n(2) Sharper-than-expected share loss in the server CPU market.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02630970065897964, "ret_signed_m1d": -0.02630970065897964, "alpha_spy_m1d": -0.020381995462958713, "alpha_c_m1d": -0.0002455685160031029, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.026064132142976537, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0013893139532934473, "ret_signed_p1d": -0.0013893139532934473, "alpha_spy_p1d": 0.0015076944259440062, "alpha_c_p1d": 0.0010958149905241354, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0024851289438175828, "ret_p1w": 0.016874054247096204, "ret_signed_p1w": 0.016874054247096204, "alpha_spy_p1w": 0.028670007025713762, "alpha_c_p1w": 0.029270419487277266, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.012396365240181062, "ret_p1m": 0.029204060386647246, "ret_signed_p1m": 0.029204060386647246, "alpha_spy_p1m": 0.03336381350988182, "alpha_c_p1m": 0.014698334258224088, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": 0.014505726128423158, "ret_p3m": -0.004118279162711791, "ret_signed_p3m": -0.004118279162711791, "alpha_spy_p3m": -0.028842523962108513, "alpha_c_p3m": -0.09859307342350931, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": 0.09447479426079752, "ret_p6m": -0.08592489102363599, "ret_signed_p6m": -0.08592489102363599, "alpha_spy_p6m": -0.06580085988863849, "alpha_c_p6m": -0.2230957998533537, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": 0.13717090882971772, "price_path": [-0.2072, -0.2037, -0.1883, -0.1886, -0.1723, -0.1813, -0.1672, -0.195, -0.1803, -0.1658, -0.1616, -0.1496, -0.1406, -0.1256, -0.1514, -0.166, -0.1398, -0.1536, -0.1284, -0.1223, -0.1188, -0.1219, -0.0961, -0.1069, -0.1007, -0.1149, -0.1165, -0.1479, -0.1587, -0.1632, -0.1505, -0.1499, -0.1389, -0.1325, -0.1082, -0.1407, -0.1407, -0.1383, -0.1263, -0.1156, -0.0324, -0.0013, -0.0272, 0.0678, 0.0391, 0.0398, 0.052, 0.0402, 0.0002, -0.0021, -0.0033, -0.0194, -0.019, -0.0179, -0.0272, -0.0318, -0.0509, -0.0451, -0.0351, -0.0212, -0.0206, -0.029, -0.0263, 0.0, -0.0014, -0.0604, 0.0324, -0.004, 0.0169, 0.025, 0.0111, 0.0108, -0.0082, -0.0151, -0.0035, 0.025, 0.0479, 0.0795, 0.1172, 0.0896, 0.0698, 0.0493, 0.0186, 0.039, 0.0292, 0.0114, 0.0373, 0.0187, 0.0281, -0.016, -0.0088, -0.0082, -0.0145, 0.0258, 0.0038, -0.0153, 0.094, 0.1144, 0.1507, 0.1507, 0.1663, 0.1175, 0.1044, 0.1016, 0.1028, 0.1295, 0.1609, 0.1759, 0.1953, 0.1762, 0.0418, -0.0165, -0.0122, -0.0564, -0.0452, -0.0149, -0.0098, 0.013, 0.0156, 0.0156, 0.0211, 0.0132, 0.0145, 0.0037, 0.0037, 0.0081, -0.0041, -0.0031, -0.0039, -0.0358, 0.0004, 0.0214, 0.0283, -0.0144, -0.0053, 0.0199, 0.0199, -0.0355, -0.0465, -0.0561, -0.0719, -0.058, -0.0349, -0.0336, -0.0409, -0.0393, -0.0398, -0.0711, -0.1067, -0.0996, -0.0346, -0.0026, -0.0128, -0.006, -0.0396, -0.057, -0.057, -0.0357, -0.0329, -0.0315, -0.0354, -0.042, -0.0561, -0.0363, -0.0671, -0.0733, -0.0755, -0.0899, -0.0792, -0.035, -0.0416, 0.0026, -0.0066, -0.0095, -0.0257, -0.0658, -0.0578, -0.0682, -0.0838, -0.0725, -0.0996, -0.0628, -0.0751, -0.0735, -0.1008, -0.1262, -0.1474, -0.1006, -0.089, -0.0859]}
{"unit_id": "thread:1991469905099501758", "ticker_idx": 0, "ts": "2025-11-20T11:33:19+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Super boom continues… SK Hynix to aggressively invest in commodity DRAM as well", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SK Hynix: 8x+ 1c DRAM capacity ramp next year, AI inference driving commodity DRAM super-cycle alongside HBM growth.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Super boom continues”… SK Hynix to aggressively invest in commodity DRAM as well\n\nSK Hynix will increase production capacity for its cutting-edge commodity DRAM — 6th-generation 10-nanometer (nm; 1 nm = one billionth of a meter) DRAM (1c DRAM) — by more than eightfold next year. The company judges that demand for state-of-the-art commodity DRAM with superior cost-effectiveness to high-bandwidth memory (HBM) is growing, as the center of gravity in artificial intelligence (AI) shifts from “training,” which requires high-performance memory, to “inference” (service). SK Hynix plans to ramp up production of GDDR7, its latest graphics DRAM based on 1c DRAM, and SOCAMM2, a low-power DRAM module, in order to respond to orders from big tech companies such as Nvidia.\n\n◇ 140,000-wafer 1c DRAM conversion investment next year\nAccording to the semiconductor industry on the 20th, SK Hynix will, through process conversion, add production capacity for 1c DRAM equivalent to 140,000 additional 300 mm (approximately 12-inch) wafers per month at its Icheon campus in Gyeonggi Province next year. An industry official said, “The 140,000-wafer figure is the minimum increase,” adding, “I understand SK Hynix is also reviewing plans to boost it by 160,000–170,000 wafers per month.”\n\nSK Hynix first plans to secure around 20,000 wafers per month of 1c DRAM production capacity by the end of this year. From next year, it will accelerate process conversion and aims to allocate an additional 140,000–170,000 wafers per month to 1c DRAM production by year-end. If this goes as planned, SK Hynix’s 1c DRAM production capacity will expand from 20,000 wafers per month at the end of this year to 160,000–190,000 wafers per month next year. Considering that SK Hynix currently processes around 500,000 DRAM wafers per month on average, this means it will devote more than 32% of its total DRAM production capacity to state-of-the-art 1c DRAM.\n\nAt the same time, the company will also expand production of 10 nm 5th-generation DRAM (1b DRAM), which is used in 6th-generation HBM (HBM4) that it will begin supplying in earnest to Nvidia next year. It will install a 1b DRAM line capable of processing 60,000 wafers per month in its new “M15X” fab, which opens at its Cheongju campus in North Chungcheong Province at the end of this year. Given SK Hynix’s aggressive DRAM investment pace, the industry expects its capital expenditures, projected at around 25 trillion won this year, to easily exceed 30 trillion won next year.\n\n◇ Riding the “AI memory super-cycle”\nSK Hynix’s 1c DRAM is used not in HBM, but to manufacture the latest commodity DRAM products such as double data rate 5 (DDR5), low-power DRAM (LPDDR), and GDDR7. The company has raised the yield (the ratio of good chips among total output) of 1c DRAM to above 80%, but it has focused its production capabilities on HBM, judging that demand growth for HBM outstrips that for commodity DRAM.\n\nThe reason SK Hynix is now turning its attention back to commodity DRAM is its assessment that the landscape of the AI memory industry is changing. As AI models, once concentrated on training, expand into inference, the company expects demand for commodity DRAM to grow as much as for HBM. In AI inference, cutting-edge commodity DRAM, which offers higher power efficiency and lower cost than HBM, is mainly used.\n\nNvidia’s recently introduced AI accelerator “Rubin CPX” is a case in point. This AI accelerator uses GDDR next to the processor instead of HBM. Big tech companies such as Google, OpenAI, and Amazon Web Services (AWS), which are developing their own AI chips, are also working on custom AI accelerators that are loaded with large quantities of commodity DRAM.\n\n1c DRAM is also used in the SOCAMM2 memory modules that SK Hynix supplies to Nvidia. SOCAMM is a memory module standard for AI servers and PCs led by Nvidia, and while its bandwidth is lower than that of HBM, it has the advantage of higher power efficiency. Nvidia plans to place SOCAMM2 next to “Vera,” its self-developed central processing unit (CPU). If things go according to plan, the industry expects SK Hynix to secure a certain portion of this demand.\n\n---\n\nhttps://t.co/ArGQG8IUxg", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015761833091266553, "ret_signed_m1d": -0.015761833091266553, "alpha_spy_m1d": -0.03124009464794275, "alpha_c_m1d": -0.059871382093548364, "bench_spy_m1d": 0.015478261556676198, "bench_c_m1d": 0.04410954900228181, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08756566012773614, "ret_signed_p1d": -0.08756566012773614, "alpha_spy_p1d": -0.09752688310436897, "alpha_c_p1d": -0.09073397766797586, "bench_spy_p1d": 0.009961222976632822, "bench_c_p1d": 0.003168317540239718, "ret_p1w": -0.04660314650565034, "ret_signed_p1w": -0.04660314650565034, "alpha_spy_p1w": -0.0882104165907881, "alpha_c_p1w": -0.11615095102835837, "bench_spy_p1w": 0.04160727008513776, "bench_c_p1w": 0.06954780452270803, "ret_p1m": -0.04134550702476525, "ret_signed_p1m": -0.04134550702476525, "alpha_spy_p1m": -0.0874293211591981, "alpha_c_p1m": -0.1371006941124706, "bench_spy_p1m": 0.046083814134432854, "bench_c_p1m": 0.09575518708770536, "ret_p3m": 0.5422596094905097, "ret_signed_p3m": 0.5422596094905097, "alpha_spy_p3m": 0.4927022124031417, "alpha_c_p3m": 0.28486796713684526, "bench_spy_p3m": 0.04955739708736795, "bench_c_p3m": 0.2573916423536644, "ret_p6m": 2.193803222720185, "ret_signed_p6m": 2.193803222720185, "alpha_spy_p6m": 2.0545777923979487, "alpha_c_p6m": 1.477189438246005, "bench_spy_p6m": 0.13922543032223622, "bench_c_p6m": 0.71661378447418, "price_path": [-0.5462, -0.5427, -0.5453, -0.5298, -0.5289, -0.5517, -0.5438, -0.5403, -0.535, -0.521, -0.5149, -0.4956, -0.4676, -0.4623, -0.4247, -0.4203, -0.3905, -0.4159, -0.3774, -0.3774, -0.3853, -0.3678, -0.3739, -0.3757, -0.4107, -0.3888, -0.3914, -0.3695, -0.3074, -0.3074, -0.3074, -0.3074, -0.3074, -0.3074, -0.2504, -0.2732, -0.2793, -0.2601, -0.2075, -0.1848, -0.1497, -0.1611, -0.1567, -0.162, -0.1068, -0.063, -0.0876, -0.0228, -0.0053, -0.021, 0.0858, 0.0263, 0.014, 0.0385, 0.0158, 0.0613, 0.0841, 0.0806, 0.0718, -0.0193, 0.0613, -0.0018, -0.0158, 0.0, -0.0876, -0.0893, -0.0911, -0.0823, -0.0466, -0.0711, -0.0571, -0.0221, -0.0326, -0.0501, -0.0466, 0.0112, -0.008, 0.0288, -0.0098, 0.0007, -0.0291, -0.0711, -0.0343, -0.0326, -0.0413, 0.0165, 0.0235, 0.0305, 0.0305, 0.0498, 0.1216, 0.1409, 0.1409, 0.1409, 0.1865, 0.2198, 0.2724, 0.3004, 0.3249, 0.3039, 0.3127, 0.2934, 0.3004, 0.3127, 0.3249, 0.339, 0.3022, 0.2969, 0.3232, 0.3442, 0.2899, 0.4021, 0.4739, 0.509, 0.5931, 0.4546, 0.5896, 0.5773, 0.4757, 0.4704, 0.5545, 0.5352, 0.5072, 0.5563, 0.5423, 0.5423, 0.5423, 0.5423, 0.5668, 0.6632, 0.6667, 0.7613, 0.7841, 0.9296, 0.8629, 0.8629, 0.6487, 0.4907, 0.6522, 0.6224, 0.4679, 0.6469, 0.6768, 0.6329, 0.5978, 0.7102, 0.7031, 0.8541, 0.7786, 0.7681, 0.6382, 0.7312, 0.747, 0.6382, 0.6382, 0.5328, 0.4169, 0.575, 0.4573, 0.5381, 0.5556, 0.6083, 0.8137, 0.7523, 0.8032, 0.826, 0.9366, 0.9946, 1.028, 0.9805, 1.0473, 1.1491, 1.1473, 1.1509, 1.1456, 1.2685, 1.2825, 1.2703, 1.258, 1.258, 1.5406, 1.5406, 1.811, 1.9041, 1.9603, 2.3009, 2.2219, 2.4695, 2.4589, 2.1938]}
{"unit_id": "orig:1999410969244741989", "ticker_idx": 0, "ts": "2025-12-12T09:28:16+00:00", "symbol": "Samsung Foundry", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I expect a model to emerge over the next year to two years where Samsung Foundry wins the front-end business", "tweet_type": "analysis", "timeline": "years", "summary": "Long Samsung Foundry: TSMC front-end bottleneck + price dissatisfaction driving fabless clients to adopt Samsung front-end + Intel packaging model over 1–2 years.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Foundry is a division of Samsung Electronics, 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "A few days ago at work, someone asked me what everyone is ignoring right now but will end up realizing is important next year.\n\nWithout even thinking, I immediately answered: Samsung Foundry.\n\nHere’s why.\n\nFirst, not only TSMC’s back-end (i.e., CoWoS), but also its front-end is becoming a severe bottleneck.\n\nIn other words, a scenario where CoWoS is simply substituted with Intel packaging—i.e., front-end at TSMC and back-end at Intel—is, for now, very difficult to execute. The front-end itself is the bottleneck, so there simply aren’t enough chips to package on Intel’s EMIB in the first place.\n\nSo I recently did some supply-chain checks, and it seems like a number of fabless companies are increasingly dissatisfied with TSMC. They’re saying the PPA improvement is not that large, yet the magnitude of the price increases is far too big.\n\nBecause of this, I expect a model to emerge over the next year to two years where Samsung Foundry wins the front-end business, and Intel handles the packaging on the back-end.\n\nTesla has already decided to do exactly that: front-end at Samsung Foundry, and packaging at Intel IFS. It’s the project that tightly integrates AI6 into a single Dojo system.\n\nNow that Tesla has made the first move, I think Nvidia and AMD may be difficult, but within the ASIC camp, I expect there will gradually be more players that adopt the “Samsung front-end + Intel back-end” model.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014692358959118579, "ret_signed_m1d": -0.014692358959118579, "alpha_spy_m1d": -0.025561280226020777, "alpha_c_m1d": -0.04447875924109268, "bench_spy_m1d": 0.010868921266902198, "bench_c_m1d": 0.0297864002819741, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03764918789226179, "ret_signed_p1d": -0.03764918789226179, "alpha_spy_p1d": -0.03613839000910757, "alpha_c_p1d": -0.02797559858854004, "bench_spy_p1d": -0.0015107978831542246, "bench_c_p1d": -0.009673589303721752, "ret_p1w": -0.02387511942760856, "ret_signed_p1w": -0.02387511942760856, "alpha_spy_p1w": -0.025108877675097396, "alpha_c_p1w": -0.030208251502919614, "bench_spy_p1w": 0.0012337582474888364, "bench_c_p1w": 0.006333132075311054, "ret_p1m": 0.28075966144611675, "ret_signed_p1m": 0.28075966144611675, "alpha_spy_p1m": 0.25809166081154555, "alpha_c_p1m": 0.25763596656179755, "bench_spy_p1m": 0.022668000634571195, "bench_c_p1m": 0.0231236948843192, "ret_p3m": 0.7532013389472918, "ret_signed_p3m": 0.7532013389472918, "alpha_spy_p3m": 0.7582346008103414, "alpha_c_p3m": 0.7744067999612702, "bench_spy_p3m": -0.005033261863049643, "bench_c_p3m": -0.021205461013978355, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2741, -0.2851, -0.2659, -0.2659, -0.2366, -0.2257, -0.2193, -0.2129, -0.2385, -0.2268, -0.2296, -0.2103, -0.1758, -0.1758, -0.1758, -0.1758, -0.1758, -0.1758, -0.1331, -0.1433, -0.1589, -0.1276, -0.1028, -0.101, -0.0992, -0.1047, -0.0946, -0.1139, -0.0927, -0.0634, -0.0863, -0.0771, -0.0441, -0.0129, 0.0202, -0.0367, -0.0762, -0.0891, -0.101, -0.0762, -0.0496, -0.0533, -0.056, -0.1074, -0.0762, -0.1019, -0.1139, -0.0762, -0.1295, -0.112, -0.0882, -0.056, -0.0496, -0.0771, -0.0744, -0.0505, -0.0404, -0.0349, -0.0046, 0.0055, -0.0046, -0.0083, -0.0147, 0.0, -0.0376, -0.056, -0.0092, -0.0119, -0.0239, 0.0147, 0.0239, 0.0202, 0.0202, 0.0744, 0.1027, 0.1064, 0.1064, 0.1064, 0.1857, 0.2743, 0.2817, 0.3011, 0.2808, 0.2826, 0.2808, 0.2697, 0.2946, 0.3278, 0.374, 0.3776, 0.3398, 0.3795, 0.4053, 0.4035, 0.4035, 0.4718, 0.4985, 0.4828, 0.481, 0.3878, 0.5456, 0.5603, 0.4699, 0.4635, 0.5354, 0.5299, 0.5484, 0.648, 0.672, 0.672, 0.672, 0.672, 0.7532, 0.7541, 0.7809, 0.8455, 0.8778, 1.0116, 0.9977, 0.9977, 0.8003, 0.589, 0.768, 0.7366, 0.6009, 0.7338, 0.7532, 0.7338, 0.6932, 0.7412, 0.7892, 0.9239, 0.8501, 0.8399, 0.7191, 0.7504, 0.744, 0.6619, 0.6619, 0.6302, 0.546, 0.7536, 0.6496, 0.7217, 0.7855, 0.8169, 0.9464, 0.8863, 0.9048, 0.8585, 0.9094, 0.951, 1.0111, 0.9972, 0.9834, 1.025, 1.0111, 1.0758, 1.0296, 1.0758, 1.0527, 1.0897, 1.0388, 1.0388, 1.1498, 1.1498, 1.4596, 1.5104, 1.4827, 1.6399, 1.5798, 1.626, 1.737, 1.5012, 1.5983, 1.5474, 1.552, 1.7693, 1.7046, 1.7046, 1.7647, 1.8387, 1.7693, 1.9311, null, null, null, null, null, null]}
{"unit_id": "orig:1980089157881123249", "ticker_idx": 0, "ts": "2025-10-20T01:50:18+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley maintains its 2026 earnings growth forecast at +10% (higher than the market's +7.2%), believing that as concerns over RPO and DRAM exposure fade, market expectations will converge toward Morgan Stanley's view", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Buy-equivalent view on KLAC: 2026 EPS +10% above consensus +7.2%, RPO/DRAM concerns to ease, structural demand robust.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "KLA 2026 Outlook Remains Solid; RPO and DRAM Concerns Expected to Ease\nMorgan Stanley (J. Moore, Oct 16, 2025)\n\nMorgan Stanley expects KLA’s December-quarter revenue to remain flat quarter-on-quarter at around $3.2 billion, slightly above the market consensus of $3.16 billion, and in line with buy-side expectations. The firm maintains its 2026 earnings growth forecast at +10% (higher than the market’s +7.2%), believing that as concerns over RPO (Remaining Performance Obligations) and DRAM exposure fade, market expectations will converge toward Morgan Stanley’s view.\n\nAddressing the market’s three main concerns, the firm notes:\n    \n1. RPO decline impact is limited — Current RPO covers 78% of the next 12 months’ shipments. Only if the book-to-bill ratio (BB) remains below 1 for the next four quarters would it pose a risk for 2027.\n    \n2. DRAM inspection intensity rising — New DRAM capacity additions will drive higher inspection intensity. Morgan Stanley expects DRAM-related revenue to grow 15% YoY in 2026, roughly matching the overall DRAM WFE growth of +18%.\n    \n3. Advanced packaging momentum — Driven by higher sampling rates for CoWoS and HBM, advanced packaging revenue is projected to reach $925 million in 2025.\n\nOverall, Morgan Stanley forecasts KLA’s system shipments to grow by about 9.4% in 2026, slightly below the global WFE growth of 10%, but notes that structural demand remains robust, implying further upward revisions in market expectations.\n\n$KLAC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04009914065902653, "ret_signed_m1d": -0.04009914065902653, "alpha_spy_m1d": -0.02980572170996254, "alpha_c_m1d": -0.02745579813188148, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0047358602596907495, "ret_signed_p1d": -0.0047358602596907495, "alpha_spy_p1d": -0.004720864339678221, "alpha_c_p1d": 0.0007938277819709016, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.05398612736824915, "ret_signed_p1w": 0.05398612736824915, "alpha_spy_p1w": 0.03322043569241129, "alpha_c_p1w": 0.01735229675091654, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": -0.024213624748203433, "ret_signed_p1m": -0.024213624748203433, "alpha_spy_p1m": -0.007499848947448706, "alpha_c_p1m": 0.01590500257727201, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.342324286803136, "ret_signed_p3m": 0.342324286803136, "alpha_spy_p3m": 0.3080840294877787, "alpha_c_p3m": 0.19710376404307395, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 0.5623873933267094, "ret_signed_p6m": 0.5623873933267094, "alpha_spy_p6m": 0.5219967057198864, "alpha_c_p6m": 0.25656848990475645, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.2236, -0.2174, -0.2192, -0.201, -0.2071, -0.1994, -0.2392, -0.2326, -0.2075, -0.2354, -0.2312, -0.2106, -0.2082, -0.2122, -0.1903, -0.1782, -0.1731, -0.2427, -0.2338, -0.2401, -0.2381, -0.2433, -0.2451, -0.2371, -0.2297, -0.229, -0.2246, -0.2436, -0.2436, -0.2659, -0.268, -0.2425, -0.2149, -0.2116, -0.204, -0.1911, -0.1679, -0.1638, -0.1422, -0.1408, -0.1414, -0.0921, -0.0937, -0.0708, -0.0709, -0.0731, -0.0814, -0.0769, -0.077, -0.0644, -0.0208, -0.0118, -0.0445, -0.0114, -0.0591, -0.0783, -0.0862, -0.1476, -0.1109, -0.1103, -0.0571, -0.0469, -0.0401, 0.0, -0.0047, -0.0335, 0.0053, 0.026, 0.054, 0.0461, 0.0715, 0.0534, 0.0484, 0.0575, 0.0352, 0.0644, 0.0464, 0.0351, 0.0564, 0.033, 0.04, 0.0077, -0.0161, -0.015, -0.0242, 0.0143, -0.0421, -0.0468, -0.0124, -0.0044, 0.007, 0.007, 0.0213, 0.0054, 0.0338, 0.0528, 0.0496, 0.0552, 0.064, 0.0649, 0.0764, 0.0827, 0.0373, 0.0644, 0.0629, 0.0183, 0.0621, 0.0823, 0.0997, 0.1023, 0.1095, 0.1095, 0.1118, 0.0951, 0.0805, 0.0557, 0.0557, 0.1073, 0.1751, 0.212, 0.1814, 0.1509, 0.2164, 0.2409, 0.2527, 0.2464, 0.3423, 0.3622, 0.3622, 0.2913, 0.3206, 0.3033, 0.3144, 0.3406, 0.4043, 0.4138, 0.4637, 0.2407, 0.2255, 0.1777, 0.1358, 0.1565, 0.2537, 0.2513, 0.2432, 0.2854, 0.2606, 0.2721, 0.2721, 0.279, 0.2878, 0.2788, 0.3015, 0.2942, 0.3107, 0.3456, 0.3261, 0.3263, 0.3354, 0.2539, 0.284, 0.2435, 0.1697, 0.2433, 0.264, 0.2745, 0.2263, 0.2342, 0.2512, 0.2887, 0.2896, 0.315, 0.3038, 0.3149, 0.3625, 0.3431, 0.2624, 0.2555, 0.2028, 0.281, 0.3222, 0.3196, 0.3196, 0.3398, 0.3475, 0.4549, 0.5027, 0.5114, 0.5388, 0.5624]}
{"unit_id": "orig:1981657106601345201", "ticker_idx": 0, "ts": "2025-10-24T09:40:46+00:00", "symbol": "BABA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs reiterated its \"Buy\" ratings on Alibaba and Tencent", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Buy reiterations on Alibaba and Tencent; valuations still attractive vs global peers, AI capex/inference demand drive revenue growth.", "resolved_tickers": ["BABA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs sharply raises Alibaba’s capital expenditure outlook to RMB 460 billion(USD 64.5 billion): Explosive growth in inference demand, AI efficiency improvements drive stronger revenue\n\nGoldman Sachs believes that explosive demand growth will continue to push up capital expenditures (Capex) for cloud service providers. The strategic paths of China’s internet giants in AI are increasingly diverging: Alibaba, leveraging its full-stack capabilities, is betting on the enterprise-oriented (b2b) AI cloud market, while ByteDance is going all-in on consumer-facing (b2c) applications.\n\nIn a report on the 23rd, Goldman Sachs raised its capex forecasts for China’s leading cloud companies, estimating that Alibaba’s cumulative capital expenditures for fiscal years 2026–2028 will reach RMB 460 billion(USD 64.5 billion), far exceeding the company’s previous target of RMB 380 billion. The core logic supporting this view is the surge in AI inference demand, and higher compute efficiency could actually increase the conversion rate of capex into cloud revenue, thereby accelerating revenue growth.\n\nAt the same time, the strategic differences among the big tech companies are becoming more pronounced. Goldman Sachs pointed out that Alibaba holds a leading position in external AI cloud revenue scale and enterprise-oriented (b2b) services, with a clearer commercialization path. In contrast, ByteDance, led by its chatbot “Doubao (豆包),” holds the largest share in the b2c space and in average daily token consumption, demonstrating its commitment to exploring consumer-oriented AI applications.\n\nThe report views the current valuations of major Chinese tech stocks as still attractive. The market has not yet entered an AI bubble, and the valuations of Tencent and Alibaba remain at a discount relative to their earnings growth outlooks and to global peers (such as Google and Amazon). Based on this, Goldman Sachs reiterated its “Buy” ratings on Alibaba and Tencent.\n\nEven with AI efficiency improvements, it is hard to curb capex expansion\nRecently, Chinese companies have achieved multiple breakthroughs in AI infrastructure and compute efficiency. For example, Alibaba Cloud announced a new GPU pooling system, “Aegaeon,” which it claims can save 82% of GPU resources; DeepSeek’s new OCR model can reduce token consumption for text input by up to 90%.\n\nHowever, Goldman Sachs believes these efficiency gains do not imply a corresponding reduction in capex. According to data cited in the report, AI inference demand and token consumption are growing exponentially. ByteDance recently announced that its average daily token consumption surpassed 30 trillion in September, doubling from April–May, a level approaching Google’s 43 trillion. Alibaba also revealed at the Yunqi Conference (Cloud Conference) that its AI inference demand doubles every 23 months.\n\nGoldman Sachs forecasts that Chinese cloud service providers’ (CSPs) capital expenditures in the third quarter of 2025 will increase by 50% year-on-year. The strong inference demand is the key driver of continued investment by cloud vendors, and improvements in technical efficiency help convert these investments more effectively into revenue growth, forming a positive cycle.\n\nDiverging paths of the giants: Alibaba b2b, ByteDance b2c\nGoldman Sachs’ report clearly outlines the different strategic layouts of China’s two tech giants in the AI field.\n\nAlibaba is focusing on the enterprise AI market and, leveraging its “unique full-stack AI capabilities,” occupies a leading position in external AI cloud revenue. On October 23, Alibaba officially launched the new Quark AI chatbot assistant service, which uses its most advanced closed-source Tongyi Qianwen (Qwen) model to compete directly with ByteDance’s “Doubao” and Tencent’s “Yuanbao.” In addition, Quark’s “Zaodian” application has also made progress in multimodal video and image editing functions based on the closed-source video model Wan2.5.\n\nBy contrast, ByteDance is more focused on exploring consumer-oriented AI applications. Its “Doubao” chatbot has secured a leadership position in China’s b2c market and in average daily token consumption. ByteDance is accelerating the commercialization of “Doubao,” for example by seamlessly integrating Douyin e-commerce services into chats to allow users to complete transactions directly within the app, and by adding new features such as an AI keyboard.\n\nMultimodality and commercialization accelerate\nThe report points out that China’s multimodal large models are making progress in global markets and are forming differentiated competitive advantages through strategies such as open-source, low prices, and high speed. For example, Tencent’s “Hunyuan Image 3.0 (混元图像3.0)” model ranks among the top on LMArena’s text-to-image model leaderboard. In terms of pricing, the report cites comparison data showing that Alibaba’s Qwen3 Max model’s output price is 40% cheaper than GPT-5/Gemini 2.5 Pro.\n\nGlobal adoption of Chinese AI models is also increasing. The report cites a statement by Airbnb’s CEO that the company is making extensive use of Alibaba’s Qwen model to support its customer service agents, highlighting that Chinese open-source AI models are gaining recognition in global markets.\n\nIn terms of commercialization, China’s b2c applications are following the path of ChatGPT. In addition to ByteDance’s rapid integration of e-commerce features into “Doubao,” Alibaba’s Quark has launched a one-stop AI image and video creation platform, “Zaodian.” Goldman Sachs believes that the commercialization path of China’s b2c chatbots is still evolving and may ultimately be more driven by advertising revenue.\n\nFrom a valuation perspective, Goldman Sachs judges that an AI bubble has not yet formed, and its U.S. strategists expect the U.S. AI capex boom to continue until 2026. According to the report, Tencent and Alibaba’s expected 2026 P/E ratios (GAAP) are 21x and 23x, respectively, which remain “not demanding” compared with Google’s 24x and Amazon’s and Microsoft’s 28–30x.\n\n$BABA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016027493421120975, "ret_signed_m1d": -0.016027493421120975, "alpha_spy_m1d": -0.007921171806284488, "alpha_c_m1d": -0.01901486227331428, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": 0.002987368852193306, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02718946813380474, "ret_signed_p1d": 0.02718946813380474, "alpha_spy_p1d": 0.015391756110655797, "alpha_c_p1d": 0.012799186545412011, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.014390281588392728, "ret_p1w": -0.0244419252836392, "ret_signed_p1w": -0.0244419252836392, "alpha_spy_p1w": -0.03154419015757792, "alpha_c_p1w": -0.03399326752720966, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.00955134224357046, "ret_p1m": -0.07996566379358305, "ret_signed_p1m": -0.07996566379358305, "alpha_spy_p1m": -0.06738531682209936, "alpha_c_p1m": -0.04146525798095124, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.03850040581263181, "ret_p3m": -0.03451630729612065, "ret_signed_p3m": -0.03451630729612065, "alpha_spy_p3m": -0.049540723059358416, "alpha_c_p3m": -0.05816843791396342, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": 0.02365213061784277, "ret_p6m": -0.19765311609894853, "ret_signed_p6m": -0.19765311609894853, "alpha_spy_p6m": -0.2500789626838278, "alpha_c_p6m": -0.21043465082811053, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.012781534729161992, "price_path": [-0.3168, -0.3281, -0.3095, -0.3299, -0.3274, -0.3301, -0.3082, -0.3076, -0.311, -0.3209, -0.2993, -0.2738, -0.3001, -0.3059, -0.3051, -0.3132, -0.316, -0.324, -0.2963, -0.2882, -0.2891, -0.3003, -0.3156, -0.2272, -0.2272, -0.2069, -0.2189, -0.2506, -0.2239, -0.1918, -0.158, -0.1761, -0.1102, -0.1124, -0.0954, -0.0715, -0.0488, -0.0699, -0.0681, -0.0598, -0.0665, 0.01, 0.0044, -0.016, 0.0298, 0.0231, 0.0463, 0.0838, 0.0763, 0.0717, 0.038, 0.0367, -0.0058, -0.0898, -0.0452, -0.0678, -0.0503, -0.055, -0.0438, -0.007, -0.046, -0.0506, -0.016, 0.0, 0.0272, 0.0116, 0.0302, -0.0044, -0.0244, -0.0401, -0.0595, -0.0566, -0.0406, -0.0479, -0.0504, -0.0796, -0.0961, -0.0851, -0.1196, -0.0973, -0.0857, -0.0905, -0.1226, -0.1246, -0.08, -0.1013, -0.0979, -0.0979, -0.0996, -0.0598, -0.0777, -0.0951, -0.0988, -0.0938, -0.0948, -0.1073, -0.0909, -0.1019, -0.1089, -0.1409, -0.1454, -0.158, -0.1567, -0.1426, -0.1359, -0.1343, -0.141, -0.141, -0.1286, -0.15, -0.1565, -0.161, -0.161, -0.1085, -0.1056, -0.1362, -0.16, -0.1158, -0.1359, -0.048, -0.044, -0.0275, -0.0216, -0.0532, -0.0532, -0.0705, -0.0345, 0.0142, -0.0084, -0.0191, -0.0113, 0.0055, -0.0026, -0.0294, -0.0361, -0.0633, -0.0891, -0.097, -0.0698, -0.067, -0.0469, -0.0594, -0.0914, -0.1086, -0.1086, -0.1103, -0.1084, -0.1169, -0.1159, -0.1255, -0.1236, -0.1283, -0.1525, -0.1751, -0.184, -0.2239, -0.2371, -0.2539, -0.2513, -0.2408, -0.2167, -0.2199, -0.2318, -0.226, -0.2175, -0.2183, -0.2305, -0.2851, -0.2993, -0.2784, -0.2817, -0.2566, -0.2821, -0.2977, -0.3018, -0.2819, -0.2918, -0.3014, -0.3014, -0.2999, -0.3147, -0.2827, -0.2691, -0.2712, -0.2673, -0.2481, -0.2371, -0.2067, -0.1928, -0.1977]}
{"unit_id": "orig:1981657106601345201", "ticker_idx": 1, "ts": "2025-10-24T09:40:46+00:00", "symbol": "Tencent", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs reiterated its \"Buy\" ratings on Alibaba and Tencent", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Buy reiterations on Alibaba and Tencent; valuations still attractive vs global peers, AI capex/inference demand drive revenue growth.", "resolved_tickers": ["0700.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "Tencent Holdings listed on HKEX as 0700.HK", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs sharply raises Alibaba’s capital expenditure outlook to RMB 460 billion(USD 64.5 billion): Explosive growth in inference demand, AI efficiency improvements drive stronger revenue\n\nGoldman Sachs believes that explosive demand growth will continue to push up capital expenditures (Capex) for cloud service providers. The strategic paths of China’s internet giants in AI are increasingly diverging: Alibaba, leveraging its full-stack capabilities, is betting on the enterprise-oriented (b2b) AI cloud market, while ByteDance is going all-in on consumer-facing (b2c) applications.\n\nIn a report on the 23rd, Goldman Sachs raised its capex forecasts for China’s leading cloud companies, estimating that Alibaba’s cumulative capital expenditures for fiscal years 2026–2028 will reach RMB 460 billion(USD 64.5 billion), far exceeding the company’s previous target of RMB 380 billion. The core logic supporting this view is the surge in AI inference demand, and higher compute efficiency could actually increase the conversion rate of capex into cloud revenue, thereby accelerating revenue growth.\n\nAt the same time, the strategic differences among the big tech companies are becoming more pronounced. Goldman Sachs pointed out that Alibaba holds a leading position in external AI cloud revenue scale and enterprise-oriented (b2b) services, with a clearer commercialization path. In contrast, ByteDance, led by its chatbot “Doubao (豆包),” holds the largest share in the b2c space and in average daily token consumption, demonstrating its commitment to exploring consumer-oriented AI applications.\n\nThe report views the current valuations of major Chinese tech stocks as still attractive. The market has not yet entered an AI bubble, and the valuations of Tencent and Alibaba remain at a discount relative to their earnings growth outlooks and to global peers (such as Google and Amazon). Based on this, Goldman Sachs reiterated its “Buy” ratings on Alibaba and Tencent.\n\nEven with AI efficiency improvements, it is hard to curb capex expansion\nRecently, Chinese companies have achieved multiple breakthroughs in AI infrastructure and compute efficiency. For example, Alibaba Cloud announced a new GPU pooling system, “Aegaeon,” which it claims can save 82% of GPU resources; DeepSeek’s new OCR model can reduce token consumption for text input by up to 90%.\n\nHowever, Goldman Sachs believes these efficiency gains do not imply a corresponding reduction in capex. According to data cited in the report, AI inference demand and token consumption are growing exponentially. ByteDance recently announced that its average daily token consumption surpassed 30 trillion in September, doubling from April–May, a level approaching Google’s 43 trillion. Alibaba also revealed at the Yunqi Conference (Cloud Conference) that its AI inference demand doubles every 23 months.\n\nGoldman Sachs forecasts that Chinese cloud service providers’ (CSPs) capital expenditures in the third quarter of 2025 will increase by 50% year-on-year. The strong inference demand is the key driver of continued investment by cloud vendors, and improvements in technical efficiency help convert these investments more effectively into revenue growth, forming a positive cycle.\n\nDiverging paths of the giants: Alibaba b2b, ByteDance b2c\nGoldman Sachs’ report clearly outlines the different strategic layouts of China’s two tech giants in the AI field.\n\nAlibaba is focusing on the enterprise AI market and, leveraging its “unique full-stack AI capabilities,” occupies a leading position in external AI cloud revenue. On October 23, Alibaba officially launched the new Quark AI chatbot assistant service, which uses its most advanced closed-source Tongyi Qianwen (Qwen) model to compete directly with ByteDance’s “Doubao” and Tencent’s “Yuanbao.” In addition, Quark’s “Zaodian” application has also made progress in multimodal video and image editing functions based on the closed-source video model Wan2.5.\n\nBy contrast, ByteDance is more focused on exploring consumer-oriented AI applications. Its “Doubao” chatbot has secured a leadership position in China’s b2c market and in average daily token consumption. ByteDance is accelerating the commercialization of “Doubao,” for example by seamlessly integrating Douyin e-commerce services into chats to allow users to complete transactions directly within the app, and by adding new features such as an AI keyboard.\n\nMultimodality and commercialization accelerate\nThe report points out that China’s multimodal large models are making progress in global markets and are forming differentiated competitive advantages through strategies such as open-source, low prices, and high speed. For example, Tencent’s “Hunyuan Image 3.0 (混元图像3.0)” model ranks among the top on LMArena’s text-to-image model leaderboard. In terms of pricing, the report cites comparison data showing that Alibaba’s Qwen3 Max model’s output price is 40% cheaper than GPT-5/Gemini 2.5 Pro.\n\nGlobal adoption of Chinese AI models is also increasing. The report cites a statement by Airbnb’s CEO that the company is making extensive use of Alibaba’s Qwen model to support its customer service agents, highlighting that Chinese open-source AI models are gaining recognition in global markets.\n\nIn terms of commercialization, China’s b2c applications are following the path of ChatGPT. In addition to ByteDance’s rapid integration of e-commerce features into “Doubao,” Alibaba’s Quark has launched a one-stop AI image and video creation platform, “Zaodian.” Goldman Sachs believes that the commercialization path of China’s b2c chatbots is still evolving and may ultimately be more driven by advertising revenue.\n\nFrom a valuation perspective, Goldman Sachs judges that an AI bubble has not yet formed, and its U.S. strategists expect the U.S. AI capex boom to continue until 2026. According to the report, Tencent and Alibaba’s expected 2026 P/E ratios (GAAP) are 21x and 23x, respectively, which remain “not demanding” compared with Google’s 24x and Amazon’s and Microsoft’s 28–30x.\n\n$BABA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007058833784847951, "ret_signed_m1d": -0.007058833784847951, "alpha_spy_m1d": 0.001047487829988536, "alpha_c_m1d": -0.004462557417973523, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.002596276366874428, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029019585433110118, "ret_signed_p1d": 0.029019585433110118, "alpha_spy_p1d": 0.017221873409961175, "alpha_c_p1d": 0.015086213446912078, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.01393337198619804, "ret_p1w": -0.013333320419080419, "ret_signed_p1w": -0.013333320419080419, "alpha_spy_p1w": -0.020435585293019143, "alpha_c_p1w": -0.006929149896818143, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": -0.006404170522262276, "ret_p1m": -0.02039215420392837, "ret_signed_p1m": -0.02039215420392837, "alpha_spy_p1m": -0.0078118072324446786, "alpha_c_p1m": 0.001589674519580142, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.021981828723508512, "ret_p3m": -0.05490197597755242, "ret_signed_p3m": -0.05490197597755242, "alpha_spy_p3m": -0.06992639174079018, "alpha_c_p3m": -0.05015159429435856, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": -0.0047503816831938606, "ret_p6m": -0.18039219294668707, "ret_signed_p6m": -0.18039219294668707, "alpha_spy_p6m": -0.23281803953156632, "alpha_c_p6m": -0.21470297859107956, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.034310785644392494, "price_path": [-0.1294, -0.1388, -0.1373, -0.1608, -0.1373, -0.1231, -0.1082, -0.1106, -0.12, -0.12, -0.1224, -0.0808, -0.0745, -0.0714, -0.0792, -0.0706, -0.0737, -0.0698, -0.0588, -0.0361, -0.0439, -0.0604, -0.0682, -0.0643, -0.051, -0.058, -0.0612, -0.0706, -0.0502, -0.0314, -0.0165, -0.0063, -0.0125, 0.0094, 0.0094, 0.0118, 0.0376, 0.0071, 0.0078, 0.0055, -0.0031, 0.0173, 0.0196, 0.0102, 0.0353, 0.04, 0.04, 0.0612, 0.0565, 0.0627, 0.0627, 0.0588, 0.0596, 0.022, 0.0024, -0.0259, -0.0165, -0.0275, -0.0463, -0.0157, -0.011, -0.022, -0.0071, 0.0, 0.029, 0.0118, 0.0118, 0.0212, -0.0133, -0.0149, -0.0133, -0.0133, 0.0102, -0.0055, 0.0188, 0.0196, 0.0306, 0.029, 0.0055, -0.0016, -0.022, -0.0235, -0.0259, -0.0431, -0.0204, -0.0196, -0.0282, -0.0408, -0.0408, -0.0282, -0.0322, -0.0416, -0.04, -0.0431, -0.051, -0.0549, -0.0541, -0.0565, -0.0337, -0.0541, -0.0643, -0.051, -0.051, -0.0369, -0.0361, -0.0557, -0.0549, -0.0549, -0.0549, -0.0643, -0.0588, -0.0604, -0.0604, -0.0227, -0.0204, -0.0078, -0.0204, -0.0337, -0.0416, -0.0227, -0.0157, -0.0071, -0.0243, -0.0314, -0.0431, -0.0573, -0.0549, -0.0627, -0.0667, -0.0596, -0.0478, -0.0259, -0.0243, -0.0494, -0.0612, -0.0886, -0.1247, -0.1239, -0.1412, -0.1216, -0.1357, -0.1404, -0.16, -0.1655, -0.1616, -0.1616, -0.1616, -0.1616, -0.1812, -0.1561, -0.1843, -0.1804, -0.1969, -0.1875, -0.1937, -0.1992, -0.2063, -0.2125, -0.1859, -0.1906, -0.1318, -0.1341, -0.1427, -0.1412, -0.1239, -0.1373, -0.1365, -0.1953, -0.2031, -0.2182, -0.1937, -0.2071, -0.2226, -0.226, -0.2445, -0.2408, -0.221, -0.2326, -0.2326, -0.2326, -0.2326, -0.2031, -0.2024, -0.2086, -0.2314, -0.2264, -0.2173, -0.189, -0.1992, -0.1804]}
{"unit_id": "orig:1977884592540696959", "ticker_idx": 0, "ts": "2025-10-13T23:50:08+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We view this announcement as a positive catalyst for Broadcom, reaffirming the company's technological leadership in the ASIC domain. a 1GW-scale AI data center buildout could contribute an estimated $10–15 billion in additional revenue and $0.96–1.44 EPS upside potential", "tweet_type": "news_reaction", "timeline": "years", "summary": "Goldman endorses AVGO as beneficiary of OpenAI partnership; $10–15B revenue and $0.96–1.44 EPS upside from 1GW-scale AI data center buildout.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251014_Broadcom: Partnership with OpenAI reinforces Broadcom’s technology advantage in custom silicon - Goldman Sachs\n\n(1) This collaboration includes the joint development of Scale-up and Scale-out network systems utilizing Broadcom’s AI accelerators and Ethernet solutions.\n\n(2) Deployment is expected to begin in 2H26 and be completed by the end of 2029.\n\n(3) We view this announcement as a positive catalyst for Broadcom, reaffirming the company’s technological leadership in the ASIC domain.\n\n(4) While it has not been disclosed whether this agreement involves any equity investment or financing, we believe it strengthens Broadcom’s technological positioning versus competitors.\n\n(5) This can be compared to AMD’s recent announcement of a partnership with OpenAI for a 6GW-scale deployment, which included an equity investment.\n\n(6) For Broadcom, a 1GW-scale AI data center buildout could contribute an estimated $10–15 billion in additional revenue and $0.96–1.44 EPS upside potential.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0899074637745334, "ret_signed_m1d": -0.0899074637745334, "alpha_spy_m1d": -0.0747952716855167, "alpha_c_m1d": -0.04749393979980332, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.04241352397473008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.035239648697864, "ret_signed_p1d": -0.035239648697864, "alpha_spy_p1d": -0.034017993894407295, "alpha_c_p1d": -0.0168438339520548, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.018395814745809203, "ret_p1w": -0.02091395832687304, "ret_signed_p1w": -0.02091395832687304, "alpha_spy_p1w": -0.03337176374381412, "alpha_c_p1w": -0.04290074308149616, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.021986784754623123, "ret_p1m": -0.01328846207863299, "ret_signed_p1m": -0.01328846207863299, "alpha_spy_p1m": -0.043392217565479396, "alpha_c_p1m": -0.04654822921250845, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.03325976713387546, "ret_p3m": -0.0661166679537174, "ret_signed_p3m": -0.0661166679537174, "alpha_spy_p3m": -0.10911177004002914, "alpha_c_p3m": -0.18505486748060396, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.11893819952688656, "ret_p6m": -0.059963735234743853, "ret_signed_p6m": -0.059963735234743853, "alpha_spy_p6m": -0.059863566206451946, "alpha_c_p6m": -0.24066816868700724, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.1807044334522634, "price_path": [-0.2141, -0.1983, -0.207, -0.1934, -0.2203, -0.206, -0.192, -0.1879, -0.1763, -0.1676, -0.1531, -0.178, -0.1922, -0.1668, -0.1802, -0.1557, -0.1499, -0.1465, -0.1495, -0.1245, -0.135, -0.129, -0.1427, -0.1443, -0.1746, -0.1851, -0.1895, -0.1772, -0.1765, -0.166, -0.1597, -0.1362, -0.1677, -0.1677, -0.1653, -0.1537, -0.1433, -0.0627, -0.0326, -0.0578, 0.0343, 0.0065, 0.0072, 0.019, 0.0075, -0.0312, -0.0335, -0.0346, -0.0502, -0.0498, -0.0488, -0.0578, -0.0622, -0.0807, -0.0751, -0.0653, -0.0519, -0.0514, -0.0595, -0.0569, -0.0314, -0.0327, -0.0899, 0.0, -0.0352, -0.0151, -0.0071, -0.0207, -0.0209, -0.0394, -0.046, -0.0348, -0.0072, 0.015, 0.0456, 0.0821, 0.0554, 0.0362, 0.0164, -0.0133, 0.0064, -0.0031, -0.0204, 0.0047, -0.0133, -0.0041, -0.0469, -0.0399, -0.0394, -0.0454, -0.0064, -0.0277, -0.0463, 0.0596, 0.0794, 0.1146, 0.1146, 0.1297, 0.0824, 0.0697, 0.067, 0.0682, 0.094, 0.1245, 0.139, 0.1578, 0.1392, 0.0091, -0.0474, -0.0432, -0.086, -0.0752, -0.0458, -0.0409, -0.0188, -0.0163, -0.0163, -0.0109, -0.0186, -0.0173, -0.0279, -0.0279, -0.0236, -0.0354, -0.0344, -0.0352, -0.0661, -0.031, -0.0107, -0.004, -0.0453, -0.0365, -0.0121, -0.0121, -0.0658, -0.0765, -0.0858, -0.101, -0.0875, -0.0652, -0.064, -0.071, -0.0694, -0.07, -0.1002, -0.1347, -0.1278, -0.0649, -0.0339, -0.0438, -0.0372, -0.0698, -0.0866, -0.0866, -0.0659, -0.0632, -0.0619, -0.0656, -0.0721, -0.0858, -0.0666, -0.0964, -0.1024, -0.1045, -0.1185, -0.1081, -0.0653, -0.0717, -0.0288, -0.0377, -0.0406, -0.0563, -0.0951, -0.0874, -0.0975, -0.1126, -0.1016, -0.1278, -0.0922, -0.1041, -0.1026, -0.1291, -0.1537, -0.1741, -0.1288, -0.1176, -0.1146, -0.1146, -0.115, -0.06]}
{"unit_id": "orig:1988063459448418377", "ticker_idx": 0, "ts": "2025-11-11T01:57:19+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix, leveraging its technology lead, will reach a 55% HBM market share in 2025, and plans to expand monthly HBM capacity from 145,000 wafers to 195,000 wafers in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman HBM report: SK Hynix dominant at 55% share, Samsung reclaiming 35%; both benefit from 20-30% YoY HBM revenue growth despite price decline.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Insights from Goldman Sachs’ latest memory report:\n\nCurrently, Samsung’s HBM3E price is 5% lower than SK hynix’s, but Samsung may further cut HBM3E prices to regain market share.\n\nHBM pricing trend: HBM3E prices are expected to decline 28% YoY in 2026 (from $15/GB to $10/GB) under competitive pressure, but HBM4’s premium (about 40% above HBM3E) and shipment growth will partially offset the price decline, so leading vendors’ HBM revenue will still rise 20–30% YoY.\n\nSK hynix, leveraging its technology lead, will reach a 55% HBM market share in 2025, and plans to expand monthly HBM capacity from 145,000 wafers to 195,000 wafers in 2026.\n\nSamsung is aiming to reclaim a 35% market share, supported by HBM3E capacity expansion (to 160,000 wafers/month in 2026).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021001554009499435, "ret_signed_m1d": -0.021001554009499435, "alpha_spy_m1d": -0.018717632545194207, "alpha_c_m1d": -0.04304973110509569, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0032309304283848483, "ret_signed_p1d": -0.0032309304283848483, "alpha_spy_p1d": -0.003787329034826925, "alpha_c_p1d": -0.015992621618218017, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.07915992164042007, "ret_signed_p1w": -0.07915992164042007, "alpha_spy_p1w": -0.04560219247649633, "alpha_c_p1w": -0.02856887862189228, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": -0.05101705158159375, "ret_signed_p1m": -0.05101705158159375, "alpha_spy_p1m": -0.05770821440615026, "alpha_c_p1m": -0.1166772586149516, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.35638280369880393, "ret_signed_p3m": 0.35638280369880393, "alpha_spy_p3m": 0.34223827217321934, "alpha_c_p3m": 0.20868370143400816, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 1.593058379587955, "ret_signed_p6m": 1.593058379587955, "alpha_spy_p6m": 1.5125188949239667, "alpha_c_p6m": 1.022140716803905, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.554, -0.554, -0.5685, -0.5757, -0.5878, -0.6048, -0.5951, -0.5814, -0.5782, -0.5806, -0.5662, -0.5654, -0.5864, -0.5792, -0.5759, -0.5711, -0.5582, -0.5525, -0.5347, -0.5089, -0.504, -0.4693, -0.4653, -0.4378, -0.4612, -0.4257, -0.4257, -0.433, -0.4168, -0.4225, -0.4241, -0.4564, -0.4362, -0.4386, -0.4184, -0.3611, -0.3611, -0.3611, -0.3611, -0.3611, -0.3611, -0.3086, -0.3296, -0.3352, -0.3174, -0.269, -0.248, -0.2157, -0.2262, -0.2221, -0.227, -0.1761, -0.1357, -0.1583, -0.0985, -0.0824, -0.0969, 0.0016, -0.0533, -0.0646, -0.042, -0.063, -0.021, 0.0, -0.0032, -0.0113, -0.0953, -0.021, -0.0792, -0.0921, -0.0775, -0.1583, -0.1599, -0.1616, -0.1535, -0.1205, -0.1432, -0.1302, -0.0979, -0.1076, -0.1238, -0.1205, -0.0672, -0.085, -0.051, -0.0866, -0.0769, -0.1044, -0.1432, -0.1092, -0.1076, -0.1157, -0.0623, -0.0559, -0.0494, -0.0494, -0.0316, 0.0347, 0.0524, 0.0524, 0.0524, 0.0945, 0.1252, 0.1737, 0.1996, 0.2222, 0.2028, 0.2109, 0.1931, 0.1996, 0.2109, 0.2222, 0.2351, 0.2012, 0.1963, 0.2206, 0.24, 0.1899, 0.2933, 0.3596, 0.3919, 0.4695, 0.3418, 0.4663, 0.455, 0.3612, 0.3564, 0.434, 0.4162, 0.3903, 0.4356, 0.4227, 0.4227, 0.4227, 0.4227, 0.4453, 0.5342, 0.5374, 0.6247, 0.6458, 0.78, 0.7184, 0.7184, 0.5209, 0.3751, 0.5241, 0.4966, 0.354, 0.5192, 0.5468, 0.5063, 0.4739, 0.5775, 0.5711, 0.7103, 0.6407, 0.631, 0.5111, 0.597, 0.6116, 0.5111, 0.5111, 0.414, 0.3071, 0.4528, 0.3443, 0.4188, 0.435, 0.4836, 0.6731, 0.6164, 0.6634, 0.6844, 0.7865, 0.8399, 0.8707, 0.827, 0.8885, 0.9825, 0.9808, 0.9841, 0.9792, 1.0926, 1.1055, 1.0942, 1.0829, 1.0829, 1.3436, 1.3436, 1.5931]}
{"unit_id": "orig:1988063459448418377", "ticker_idx": 1, "ts": "2025-11-11T01:57:19+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "leading vendors' HBM revenue will still rise 20–30% YoY", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman HBM report: SK Hynix dominant at 55% share, Samsung reclaiming 35%; both benefit from 20-30% YoY HBM revenue growth despite price decline.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Insights from Goldman Sachs’ latest memory report:\n\nCurrently, Samsung’s HBM3E price is 5% lower than SK hynix’s, but Samsung may further cut HBM3E prices to regain market share.\n\nHBM pricing trend: HBM3E prices are expected to decline 28% YoY in 2026 (from $15/GB to $10/GB) under competitive pressure, but HBM4’s premium (about 40% above HBM3E) and shipment growth will partially offset the price decline, so leading vendors’ HBM revenue will still rise 20–30% YoY.\n\nSK hynix, leveraging its technology lead, will reach a 55% HBM market share in 2025, and plans to expand monthly HBM capacity from 145,000 wafers to 195,000 wafers in 2026.\n\nSamsung is aiming to reclaim a 35% market share, supported by HBM3E capacity expansion (to 160,000 wafers/month in 2026).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028019304871692907, "ret_signed_m1d": -0.028019304871692907, "alpha_spy_m1d": -0.02573538340738768, "alpha_c_m1d": -0.036619598180055335, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.008600293308362428, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0038647290855086114, "ret_signed_p1d": -0.0038647290855086114, "alpha_spy_p1d": -0.004421127691950688, "alpha_c_p1d": -0.006970451481112305, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.003105722395603694, "ret_p1w": -0.055072408470252965, "ret_signed_p1w": -0.055072408470252965, "alpha_spy_p1w": -0.02151467930632922, "alpha_c_p1w": -0.007360712919403656, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.04771169555084931, "ret_p1m": 0.04347829722074992, "ret_signed_p1m": 0.04347829722074992, "alpha_spy_p1m": 0.03678713439619341, "alpha_c_p1m": 0.028291202955803563, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.015187094264946355, "ret_p3m": 0.5398162408613907, "ret_signed_p3m": 0.5398162408613907, "alpha_spy_p3m": 0.5256717093358061, "alpha_c_p3m": 0.57504332298627, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": -0.03522708212487935, "ret_p6m": 1.5878871104174541, "ret_signed_p6m": 1.5878871104174541, "alpha_spy_p6m": 1.507347625753466, "alpha_c_p6m": 1.4241069124708634, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.1637801979465907, "price_path": [-0.3113, -0.3113, -0.3267, -0.3267, -0.3219, -0.3209, -0.3132, -0.3122, -0.3238, -0.3209, -0.3305, -0.3296, -0.3498, -0.3353, -0.3286, -0.3257, -0.3315, -0.3257, -0.3122, -0.3017, -0.294, -0.2747, -0.2642, -0.2363, -0.2478, -0.2276, -0.2276, -0.1968, -0.1853, -0.1785, -0.1718, -0.1987, -0.1865, -0.1894, -0.1691, -0.1329, -0.1329, -0.1329, -0.1329, -0.1329, -0.1329, -0.0879, -0.0986, -0.115, -0.0821, -0.056, -0.0541, -0.0522, -0.058, -0.0473, -0.0676, -0.0454, -0.0145, -0.0386, -0.029, 0.0058, 0.0386, 0.0734, 0.0135, -0.028, -0.0415, -0.0541, -0.028, 0.0, -0.0039, -0.0068, -0.0609, -0.028, -0.0551, -0.0676, -0.028, -0.0841, -0.0657, -0.0406, -0.0068, 0.0, -0.029, -0.0261, -0.001, 0.0097, 0.0155, 0.0473, 0.058, 0.0473, 0.0435, 0.0367, 0.0522, 0.0126, -0.0068, 0.0425, 0.0396, 0.0271, 0.0676, 0.0773, 0.0734, 0.0734, 0.1304, 0.1602, 0.1641, 0.1641, 0.1641, 0.2476, 0.3408, 0.3486, 0.3689, 0.3476, 0.3495, 0.3476, 0.3359, 0.3621, 0.3971, 0.4456, 0.4495, 0.4097, 0.4515, 0.4787, 0.4767, 0.4767, 0.5486, 0.5767, 0.5602, 0.5583, 0.4602, 0.6262, 0.6418, 0.5466, 0.5398, 0.6155, 0.6097, 0.6291, 0.734, 0.7592, 0.7592, 0.7592, 0.7592, 0.8447, 0.8456, 0.8738, 0.9418, 0.9757, 1.1165, 1.102, 1.102, 0.8942, 0.6719, 0.8602, 0.8272, 0.6845, 0.8243, 0.8447, 0.8243, 0.7816, 0.8321, 0.8825, 1.0243, 0.9466, 0.9359, 0.8088, 0.8418, 0.835, 0.7486, 0.7486, 0.7152, 0.6267, 0.8451, 0.7356, 0.8115, 0.8787, 0.9117, 1.0479, 0.9847, 1.0042, 0.9555, 1.009, 1.0528, 1.116, 1.1014, 1.0868, 1.1306, 1.116, 1.1841, 1.1355, 1.1841, 1.1598, 1.1987, 1.1452, 1.1452, 1.262, 1.262, 1.5879]}
{"unit_id": "orig:1977545340988584147", "ticker_idx": 0, "ts": "2025-10-13T01:22:04+00:00", "symbol": "general-purpose DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "general-purpose DRAM will continue to suffer from persistent supply shortages", "tweet_type": "analysis", "timeline": "quarters", "summary": "General-purpose DRAM faces persistent supply shortages in 2026 as all incremental Samsung/Hynix capacity goes to HBM; bullish on DRAM pricing/suppliers.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "2408.TW"], "resolution_method": "basket", "resolution_reason": "General-purpose DRAM basket: Micron, SK Hynix, Samsung, Nanya", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "This is a highly important point.\n\nSamsung Electronics and SK Hynix’s total DRAM capacity in 2026 is expected to increase only slightly (+single-digit).\nHowever, all of that incremental capacity will go toward HBM, while general DRAM capacity will actually decline.\n\nIn other words, general-purpose DRAM will continue to suffer from persistent supply shortages.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003876307870674528, "ret_signed_m1d": 0.003876307870674528, "alpha_spy_m1d": 0.018988499959691235, "alpha_c_m1d": 0.04628983184540461, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.04241352397473008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.031066749247729564, "ret_signed_p1d": -0.031066749247729564, "alpha_spy_p1d": -0.029845094444272857, "alpha_c_p1d": -0.012670934501920361, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.018395814745809203, "ret_p1w": 0.10329970445084724, "ret_signed_p1w": 0.10329970445084724, "alpha_spy_p1w": 0.09084189903390616, "alpha_c_p1w": 0.08131291969622412, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.021986784754623123, "ret_p1m": 0.3930918416720066, "ret_signed_p1m": 0.3930918416720066, "alpha_spy_p1m": 0.3629880861851602, "alpha_c_p1m": 0.35983207453813115, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.03325976713387546, "ret_p3m": 0.8852897442079669, "ret_signed_p3m": 0.8852897442079669, "alpha_spy_p3m": 0.8422946421216552, "alpha_c_p3m": 0.7663515446810804, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.11893819952688656, "ret_p6m": 1.1121867166365935, "ret_signed_p6m": 1.1121867166365935, "alpha_spy_p6m": 1.1122868856648855, "alpha_c_p6m": 0.9314822831843301, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.1807044334522634, "price_path": [-0.3902, -0.4005, -0.4004, -0.3988, -0.4148, -0.4061, -0.4046, -0.4069, -0.3972, -0.3983, -0.3817, -0.3871, -0.4086, -0.4048, -0.3976, -0.4066, -0.399, -0.3851, -0.3751, -0.3589, -0.3595, -0.3591, -0.3605, -0.3684, -0.3772, -0.3898, -0.3927, -0.3874, -0.3795, -0.3801, -0.3777, -0.3725, -0.3749, -0.389, -0.3858, -0.3806, -0.366, -0.341, -0.3358, -0.3195, -0.2995, -0.2764, -0.2471, -0.227, -0.1917, -0.1914, -0.1435, -0.1536, -0.1412, -0.1291, -0.1485, -0.152, -0.1856, -0.166, -0.1581, -0.1173, -0.0747, -0.0606, -0.0566, -0.0419, -0.0289, -0.0142, 0.0039, 0.0, -0.0311, -0.0163, 0.0452, 0.0772, 0.1033, 0.088, 0.0993, 0.1013, 0.1424, 0.1949, 0.2082, 0.2472, 0.2633, 0.2615, 0.3338, 0.2607, 0.2872, 0.3192, 0.3047, 0.3855, 0.3931, 0.3918, 0.379, 0.3324, 0.3838, 0.3215, 0.3085, 0.2813, 0.2029, 0.2367, 0.2452, 0.2467, 0.2857, 0.2773, 0.2972, 0.315, 0.3113, 0.2972, 0.3262, 0.389, 0.3827, 0.3932, 0.3713, 0.3724, 0.3477, 0.3097, 0.3439, 0.388, 0.4158, 0.4752, 0.4707, 0.5182, 0.5182, 0.5382, 0.5824, 0.6036, 0.5903, 0.5903, 0.7047, 0.7403, 0.8548, 0.8977, 0.8853, 0.8393, 0.899, 0.8555, 0.876, 0.8968, 0.9811, 1.0523, 1.0237, 1.0099, 1.078, 1.1004, 1.1059, 1.1684, 1.2857, 1.2938, 1.3722, 1.2422, 1.2676, 1.2615, 1.1524, 1.1353, 1.1982, 1.1675, 1.2296, 1.2803, 1.2795, 1.2795, 1.2641, 1.2915, 1.319, 1.3664, 1.3949, 1.466, 1.4714, 1.5365, 1.5053, 1.5044, 1.2558, 1.1096, 1.2727, 1.1545, 1.0263, 1.1995, 1.2958, 1.2198, 1.2189, 1.3534, 1.4169, 1.52, 1.408, 1.3057, 1.1729, 1.1856, 1.1979, 1.0993, 1.0859, 0.9972, 0.8961, 1.0827, 0.9823, 1.0311, 1.0707, 1.1122]}
{"unit_id": "thread:2000470737883472320", "ticker_idx": 0, "ts": "2025-12-15T07:39:25+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron will supply less than 10% of the volume, falling behind SK Hynix and Samsung... Micron failed to secure an advantage in volume discussions as it struggled to enhance performance", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "HBM4 share: SK Hynix ~70%, Samsung >30%, Micron <10% — bullish Hynix/Samsung, bearish Micron on next-gen HBM supply split.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"NVIDIA Supply Talks in Final Stages\"... Samsung to Supply More HBM4 to NVIDIA than Micron\n\nIt has been reported that Samsung Electronics' negotiations with NVIDIA regarding the supply of next year's 6th generation High Bandwidth Memory (HBM4) have reached the final stages. Following SK Hynix’s announcement that its HBM4 volume for next year is \"sold out,\" it appears that Samsung Electronics will supply the second-largest amount of HBM4, trailing only SK Hynix. Although Samsung suffered a blow to its reputation this year by only supplying a small quantity of 5th generation HBM (HBM3E) to NVIDIA, the company seems to be setting out to restore its pride starting with HBM4.\n\nAccording to industry sources on the 15th, it is reported that Samsung Electronics will supply over 30% of the HBM4 volume NVIDIA has requested from the memory semiconductor industry for next year. It is known that Samsung is currently finalizing quality certification by mounting its HBM4 samples on the 'Rubin Platform,' the next-generation AI accelerator NVIDIA plans to release to the market next year.\n\nAn official familiar with Samsung Electronics' situation stated, \"Unlike in the HBM3E market, Samsung Electronics' share within NVIDIA's supply chain for HBM4 is expected to exceed 30%,\" adding, \"It appears they will supply a larger volume than Micron while meeting the performance requirements demanded by NVIDIA.\"\n\nUntil now, Samsung Electronics has faced difficulties in supplying HBM to NVIDIA. While competitors SK Hynix and Micron stably supplied 8-layer and 12-layer HBM3E products, Samsung suffered the humiliation of failing quality certification multiple times. To recover competitiveness in the HBM4 market, Samsung has made a strategic move by utilizing its internal foundry (contract manufacturing) 4nm process and incorporating DRAM that is one generation ahead of its competitors. Samsung is also pushing to expand HBM4 production capacity at its Pyeongtaek Campus to respond to HBM4 demand.\n\nSK Hynix is forecasted to solidify its status as the top supplier by providing close to 70% of the total HBM4 volume requested by NVIDIA next year. It has been confirmed that SK Hynix agreed to supply the maximum amount of HBM4 volume it can handle within its scheduled production capacity for next year. SK Hynix is also maximizing its production capacity to meet HBM demand. A semiconductor industry official noted, \"The contract has been finalized for SK Hynix to supply the maximum volume the company can handle out of the HBM4 quantity requested by NVIDIA,\" adding, \"Additional discussions may proceed once plans for capacity expansion become visible.\"\n\nIt is reported that Micron will supply less than 10% of the volume, falling behind SK Hynix and Samsung. It is known that negotiations were somewhat delayed as Micron experienced technical issues, leading to a partial redesign of its HBM4. Although Micron supplied Customer Samples (CS) that met the HBM4 product standards requested by NVIDIA, it was identified that they showed inferior performance compared to competitors.\n\nUnlike SK Hynix and Samsung, Micron utilizes its own DRAM process rather than a foundry process for the logic die, which acts as the brain of the HBM. While SK Hynix and Samsung entered supply negotiations meeting the product performance requested by NVIDIA by utilizing advanced foundry processes, it is analyzed that Micron failed to secure an advantage in volume discussions as it struggled to enhance performance.\n\n$MU\n\n---\n\nhttps://t.co/8mM2d5Y3lt", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015326278026813522, "ret_signed_m1d": -0.015326278026813522, "alpha_spy_m1d": -0.013813194179786104, "alpha_c_m1d": -0.01186920406305747, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021010569585278582, "ret_signed_p1d": 0.021010569585278582, "alpha_spy_p1d": 0.018278271989999872, "alpha_c_p1d": 0.018290226336253657, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.1645894394682299, "ret_signed_p1w": -0.1645894394682299, "alpha_spy_p1w": -0.15559371415552747, "alpha_c_p1w": -0.14214894979603554, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4242803264561392, "ret_signed_p1m": -0.4242803264561392, "alpha_spy_p1m": -0.4021128365504283, "alpha_c_p1m": -0.31136370124665924, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.7074262303029928, "ret_signed_p3m": -0.7074262303029928, "alpha_spy_p3m": -0.7260854035676849, "alpha_c_p3m": -0.604174023515615, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3268, -0.2893, -0.3152, -0.3073, -0.2998, -0.3195, -0.3401, -0.3382, -0.3103, -0.2959, -0.2335, -0.2268, -0.2091, -0.196, -0.2181, -0.1725, -0.1902, -0.2354, -0.1883, -0.2124, -0.1918, -0.1472, -0.1479, -0.1294, -0.1483, -0.1643, -0.1296, -0.0778, -0.0733, -0.0656, -0.0458, -0.0568, -0.0578, -0.0118, -0.082, 0.0, 0.0035, 0.0018, 0.0665, 0.0152, 0.0312, -0.0023, 0.0393, 0.0187, -0.0379, -0.0488, -0.1521, -0.1269, -0.0571, -0.0546, -0.0305, -0.0305, -0.0043, 0.0125, 0.0084, -0.0141, -0.0457, -0.0012, 0.0397, 0.0628, 0.1104, 0.0883, 0.0153, 0.0, -0.021, -0.0504, 0.0465, 0.1197, 0.1646, 0.1632, 0.2071, 0.2071, 0.1991, 0.24, 0.2326, 0.2022, 0.2022, 0.3286, 0.3148, 0.4466, 0.4303, 0.3775, 0.4536, 0.4569, 0.4243, 0.4041, 0.418, 0.528, 0.528, 0.5375, 0.639, 0.6747, 0.6834, 0.6389, 0.728, 0.8335, 0.8356, 0.7476, 0.8441, 0.7668, 0.5981, 0.6128, 0.6625, 0.6154, 0.5722, 0.7284, 0.7437, 0.734, 0.734, 0.684, 0.7731, 0.758, 0.8035, 0.7732, 0.7608, 0.807, 0.7504, 0.737, 0.7383, 0.5993, 0.6881, 0.6725, 0.5598, 0.6399, 0.698, 0.7636, 0.7074, 0.795, 0.861, 0.9447, 0.9449, 0.8714, 0.7814, 0.7032, 0.6661, 0.6094, 0.4973, 0.5047, 0.3561, 0.4237, 0.5501, 0.5433, 0.5433, 0.5919, 0.5911, 0.714, 0.7762, 0.7724, 0.7975, 0.9623, 0.9226, 0.9268, 0.9177, 0.8896, 0.8937, 1.0542, 1.03, 1.0932, 1.2105, 1.1251, 1.1848, 1.1793, 1.2849, 1.4292, 1.6978, 1.809, 1.7249, 2.1471, 2.3515, 2.2304, 2.3865, 2.2701, 2.0537, 1.872, 1.9445, 2.0846, 2.2115, 2.1647, 2.1647, 2.7752, 2.9123, 2.8917, 3.0918, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000470737883472320", "ticker_idx": 1, "ts": "2025-12-15T07:39:25+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is forecasted to solidify its status as the top supplier by providing close to 70% of the total HBM4 volume requested by NVIDIA next year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "HBM4 share: SK Hynix ~70%, Samsung >30%, Micron <10% — bullish Hynix/Samsung, bearish Micron on next-gen HBM supply split.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"NVIDIA Supply Talks in Final Stages\"... Samsung to Supply More HBM4 to NVIDIA than Micron\n\nIt has been reported that Samsung Electronics' negotiations with NVIDIA regarding the supply of next year's 6th generation High Bandwidth Memory (HBM4) have reached the final stages. Following SK Hynix’s announcement that its HBM4 volume for next year is \"sold out,\" it appears that Samsung Electronics will supply the second-largest amount of HBM4, trailing only SK Hynix. Although Samsung suffered a blow to its reputation this year by only supplying a small quantity of 5th generation HBM (HBM3E) to NVIDIA, the company seems to be setting out to restore its pride starting with HBM4.\n\nAccording to industry sources on the 15th, it is reported that Samsung Electronics will supply over 30% of the HBM4 volume NVIDIA has requested from the memory semiconductor industry for next year. It is known that Samsung is currently finalizing quality certification by mounting its HBM4 samples on the 'Rubin Platform,' the next-generation AI accelerator NVIDIA plans to release to the market next year.\n\nAn official familiar with Samsung Electronics' situation stated, \"Unlike in the HBM3E market, Samsung Electronics' share within NVIDIA's supply chain for HBM4 is expected to exceed 30%,\" adding, \"It appears they will supply a larger volume than Micron while meeting the performance requirements demanded by NVIDIA.\"\n\nUntil now, Samsung Electronics has faced difficulties in supplying HBM to NVIDIA. While competitors SK Hynix and Micron stably supplied 8-layer and 12-layer HBM3E products, Samsung suffered the humiliation of failing quality certification multiple times. To recover competitiveness in the HBM4 market, Samsung has made a strategic move by utilizing its internal foundry (contract manufacturing) 4nm process and incorporating DRAM that is one generation ahead of its competitors. Samsung is also pushing to expand HBM4 production capacity at its Pyeongtaek Campus to respond to HBM4 demand.\n\nSK Hynix is forecasted to solidify its status as the top supplier by providing close to 70% of the total HBM4 volume requested by NVIDIA next year. It has been confirmed that SK Hynix agreed to supply the maximum amount of HBM4 volume it can handle within its scheduled production capacity for next year. SK Hynix is also maximizing its production capacity to meet HBM demand. A semiconductor industry official noted, \"The contract has been finalized for SK Hynix to supply the maximum volume the company can handle out of the HBM4 quantity requested by NVIDIA,\" adding, \"Additional discussions may proceed once plans for capacity expansion become visible.\"\n\nIt is reported that Micron will supply less than 10% of the volume, falling behind SK Hynix and Samsung. It is known that negotiations were somewhat delayed as Micron experienced technical issues, leading to a partial redesign of its HBM4. Although Micron supplied Customer Samples (CS) that met the HBM4 product standards requested by NVIDIA, it was identified that they showed inferior performance compared to competitors.\n\nUnlike SK Hynix and Samsung, Micron utilizes its own DRAM process rather than a foundry process for the logic die, which acts as the brain of the HBM. While SK Hynix and Samsung entered supply negotiations meeting the product performance requested by NVIDIA by utilizing advanced foundry processes, it is analyzed that Micron failed to secure an advantage in volume discussions as it struggled to enhance performance.\n\n$MU\n\n---\n\nhttps://t.co/8mM2d5Y3lt", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030685848548443806, "ret_signed_m1d": 0.030685848548443806, "alpha_spy_m1d": 0.029172764701416387, "alpha_c_m1d": 0.027228774584687754, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.043321337592038156, "ret_signed_p1d": -0.043321337592038156, "alpha_spy_p1d": -0.040589039996759446, "alpha_c_p1d": -0.04060099434301323, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.046931364275827114, "ret_signed_p1w": 0.046931364275827114, "alpha_spy_p1w": 0.037935638963124685, "alpha_c_p1w": 0.02449087460363275, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3321299157434354, "ret_signed_p1m": 0.3321299157434354, "alpha_spy_p1m": 0.3099624258377245, "alpha_c_p1m": 0.21921329053395544, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.6817978803316003, "ret_signed_p3m": 0.6817978803316003, "alpha_spy_p3m": 0.7004570535962924, "alpha_c_p3m": 0.5785456735442225, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3984, -0.3588, -0.3588, -0.3669, -0.3488, -0.3552, -0.357, -0.393, -0.3705, -0.3732, -0.3506, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.228, -0.2514, -0.2578, -0.2379, -0.1838, -0.1603, -0.1243, -0.136, -0.1315, -0.1369, -0.0801, -0.035, -0.0602, 0.0065, 0.0245, 0.0083, 0.1183, 0.057, 0.0444, 0.0696, 0.0462, 0.0931, 0.1165, 0.1129, 0.1039, 0.0101, 0.0931, 0.0281, 0.0137, 0.0299, -0.0602, -0.062, -0.0638, -0.0548, -0.0181, -0.0433, -0.0289, 0.0072, -0.0036, -0.0217, -0.0181, 0.0415, 0.0217, 0.0596, 0.0199, 0.0307, 0.0, -0.0433, -0.0054, -0.0036, -0.0126, 0.0469, 0.0542, 0.0614, 0.0614, 0.0812, 0.1552, 0.1751, 0.1751, 0.1751, 0.222, 0.2563, 0.3105, 0.3394, 0.3646, 0.343, 0.352, 0.3321, 0.3394, 0.352, 0.3646, 0.3791, 0.3412, 0.3357, 0.3628, 0.3845, 0.3285, 0.444, 0.5181, 0.5542, 0.6408, 0.4982, 0.6372, 0.6245, 0.5199, 0.5144, 0.6011, 0.5812, 0.5523, 0.6029, 0.5884, 0.5884, 0.5884, 0.5884, 0.6137, 0.713, 0.7166, 0.8141, 0.8375, 0.9874, 0.9187, 0.9187, 0.6981, 0.5353, 0.7017, 0.6709, 0.5118, 0.6963, 0.727, 0.6818, 0.6456, 0.7614, 0.7541, 0.9097, 0.8319, 0.821, 0.6872, 0.7831, 0.7993, 0.6872, 0.6872, 0.5787, 0.4594, 0.6221, 0.501, 0.5841, 0.6022, 0.6565, 0.8681, 0.8048, 0.8572, 0.8807, 0.9946, 1.0543, 1.0887, 1.0399, 1.1086, 1.2135, 1.2117, 1.2153, 1.2098, 1.3364, 1.3509, 1.3382, 1.3256, 1.3256, 1.6167, 1.6167, 1.8952, 1.9911, 2.0489, 2.3998, 2.3184, 2.5734, 2.5625, 2.2895, 2.3274, 2.1556, 2.1556, 2.5083, 2.5101, 2.5101, 2.7108, 3.0562, 3.1401, 3.2197, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000470737883472320", "ticker_idx": 2, "ts": "2025-12-15T07:39:25+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics will supply over 30% of the HBM4 volume NVIDIA has requested from the memory semiconductor industry for next year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "HBM4 share: SK Hynix ~70%, Samsung >30%, Micron <10% — bullish Hynix/Samsung, bearish Micron on next-gen HBM supply split.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"NVIDIA Supply Talks in Final Stages\"... Samsung to Supply More HBM4 to NVIDIA than Micron\n\nIt has been reported that Samsung Electronics' negotiations with NVIDIA regarding the supply of next year's 6th generation High Bandwidth Memory (HBM4) have reached the final stages. Following SK Hynix’s announcement that its HBM4 volume for next year is \"sold out,\" it appears that Samsung Electronics will supply the second-largest amount of HBM4, trailing only SK Hynix. Although Samsung suffered a blow to its reputation this year by only supplying a small quantity of 5th generation HBM (HBM3E) to NVIDIA, the company seems to be setting out to restore its pride starting with HBM4.\n\nAccording to industry sources on the 15th, it is reported that Samsung Electronics will supply over 30% of the HBM4 volume NVIDIA has requested from the memory semiconductor industry for next year. It is known that Samsung is currently finalizing quality certification by mounting its HBM4 samples on the 'Rubin Platform,' the next-generation AI accelerator NVIDIA plans to release to the market next year.\n\nAn official familiar with Samsung Electronics' situation stated, \"Unlike in the HBM3E market, Samsung Electronics' share within NVIDIA's supply chain for HBM4 is expected to exceed 30%,\" adding, \"It appears they will supply a larger volume than Micron while meeting the performance requirements demanded by NVIDIA.\"\n\nUntil now, Samsung Electronics has faced difficulties in supplying HBM to NVIDIA. While competitors SK Hynix and Micron stably supplied 8-layer and 12-layer HBM3E products, Samsung suffered the humiliation of failing quality certification multiple times. To recover competitiveness in the HBM4 market, Samsung has made a strategic move by utilizing its internal foundry (contract manufacturing) 4nm process and incorporating DRAM that is one generation ahead of its competitors. Samsung is also pushing to expand HBM4 production capacity at its Pyeongtaek Campus to respond to HBM4 demand.\n\nSK Hynix is forecasted to solidify its status as the top supplier by providing close to 70% of the total HBM4 volume requested by NVIDIA next year. It has been confirmed that SK Hynix agreed to supply the maximum amount of HBM4 volume it can handle within its scheduled production capacity for next year. SK Hynix is also maximizing its production capacity to meet HBM demand. A semiconductor industry official noted, \"The contract has been finalized for SK Hynix to supply the maximum volume the company can handle out of the HBM4 quantity requested by NVIDIA,\" adding, \"Additional discussions may proceed once plans for capacity expansion become visible.\"\n\nIt is reported that Micron will supply less than 10% of the volume, falling behind SK Hynix and Samsung. It is known that negotiations were somewhat delayed as Micron experienced technical issues, leading to a partial redesign of its HBM4. Although Micron supplied Customer Samples (CS) that met the HBM4 product standards requested by NVIDIA, it was identified that they showed inferior performance compared to competitors.\n\nUnlike SK Hynix and Samsung, Micron utilizes its own DRAM process rather than a foundry process for the logic die, which acts as the brain of the HBM. While SK Hynix and Samsung entered supply negotiations meeting the product performance requested by NVIDIA by utilizing advanced foundry processes, it is analyzed that Micron failed to secure an advantage in volume discussions as it struggled to enhance performance.\n\n$MU\n\n---\n\nhttps://t.co/8mM2d5Y3lt", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03912210331054089, "ret_signed_m1d": 0.03912210331054089, "alpha_spy_m1d": 0.03760901946351347, "alpha_c_m1d": 0.029354021596027735, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.009768081714513155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019083943680241022, "ret_signed_p1d": -0.019083943680241022, "alpha_spy_p1d": -0.016351646084962312, "alpha_c_p1d": -0.02091102355885721, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": 0.001827079878616189, "ret_p1w": 0.05438933331892093, "ret_signed_p1w": 0.05438933331892093, "alpha_spy_p1w": 0.0453936080062185, "alpha_c_p1w": 0.032743727030979164, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.021645606287941765, "ret_p1m": 0.319359684466187, "ret_signed_p1m": 0.319359684466187, "alpha_spy_p1m": 0.2971921945604761, "alpha_c_p1m": 0.2884238415073155, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.03093584295887153, "ret_p3m": 0.8016547449724258, "ret_signed_p3m": 0.8016547449724258, "alpha_spy_p3m": 0.820313918237118, "alpha_c_p3m": 0.831527802060285, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.029873057087859145, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2571, -0.2372, -0.2372, -0.2068, -0.1954, -0.1887, -0.1821, -0.2087, -0.1966, -0.1994, -0.1794, -0.1436, -0.1436, -0.1436, -0.1436, -0.1436, -0.1436, -0.0992, -0.1097, -0.126, -0.0935, -0.0677, -0.0658, -0.0639, -0.0697, -0.0592, -0.0792, -0.0573, -0.0267, -0.0506, -0.041, -0.0067, 0.0258, 0.0601, 0.001, -0.0401, -0.0534, -0.0658, -0.0401, -0.0124, -0.0162, -0.0191, -0.0725, -0.0401, -0.0668, -0.0792, -0.0401, -0.0954, -0.0773, -0.0525, -0.0191, -0.0124, -0.041, -0.0382, -0.0134, -0.0029, 0.0029, 0.0344, 0.0448, 0.0344, 0.0305, 0.0239, 0.0391, 0.0, -0.0191, 0.0296, 0.0267, 0.0143, 0.0544, 0.0639, 0.0601, 0.0601, 0.1164, 0.1458, 0.1496, 0.1496, 0.1496, 0.2321, 0.3242, 0.3318, 0.352, 0.3309, 0.3328, 0.3309, 0.3194, 0.3452, 0.3798, 0.4277, 0.4315, 0.3922, 0.4335, 0.4603, 0.4584, 0.4584, 0.5293, 0.5572, 0.5409, 0.5389, 0.4421, 0.6061, 0.6214, 0.5274, 0.5207, 0.5955, 0.5898, 0.6089, 0.7125, 0.7374, 0.7374, 0.7374, 0.7374, 0.8218, 0.8227, 0.8506, 0.9177, 0.9512, 1.0903, 1.0759, 1.0759, 0.8707, 0.6511, 0.8371, 0.8045, 0.6636, 0.8017, 0.8218, 0.8017, 0.7595, 0.8093, 0.8592, 0.9992, 0.9225, 0.9119, 0.7863, 0.8189, 0.8122, 0.7269, 0.7269, 0.6939, 0.6065, 0.8222, 0.7141, 0.789, 0.8553, 0.888, 1.0225, 0.9601, 0.9793, 0.9313, 0.9841, 1.0273, 1.0898, 1.0754, 1.061, 1.1042, 1.0898, 1.157, 1.109, 1.157, 1.133, 1.1715, 1.1186, 1.1186, 1.2339, 1.2339, 1.5558, 1.6086, 1.5798, 1.7431, 1.6807, 1.7287, 1.844, 1.599, 1.6999, 1.6471, 1.6519, 1.8777, 1.8104, 1.8104, 1.8729, 1.9497, 1.8777, 2.0458, null, null, null, null, null, null, null]}
{"unit_id": "orig:1988905916578816234", "ticker_idx": 0, "ts": "2025-11-13T09:44:57+00:00", "symbol": "Delta Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we remain positive on Delta, and we expect the company's server power-supply business revenues to post a CAGR of 42% in 2025–2027, with room for further upside", "tweet_type": "analysis", "timeline": "years", "summary": "JPM bullish on Delta (power/cooling CAGR 42%/36%) and Jentech (sole supplier of lids for both Rubin GPU variants, ASP up 2-9x) into 2026-2028 AI server ramp.", "resolved_tickers": ["2308.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Delta Electronics Taiwan 2308.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Outlook for Power, Liquid Cooling, and ODMs in Nvidia Racks – JPM\n\n2025 order backlog + 2026 AI capital expenditure increase = strong AI server shipments next year\n\nWe estimate that this year the demand and shipments for NVL72 cabinets will be around 38,000 units and 27,000 units, respectively, which implies an order backlog of about 10,000–11,000 units. According to the initial capex outlook, demand for NVL72 cabinets next year will reach 50,000–60,000 units, most of which will be based on GB300. Taking into account the backlog at the end of 2025, we estimate that the number of cabinets assembled by ODMs in 2026 will be in the range of 50,000–70,000 units along the yield-improvement curve. Growth that is close to a doubling year over year implies that revenues for AI server ODMs and downstream component suppliers will all show strong growth next year.\n\nGB300/VR200 schedule and spec update\n\nWe believe the mass-production ramp of GB300 servers is somewhat behind our original expectations, and we expect to see a more meaningful ramp in 1Q26. Even so, Nvidia is still maintaining its plan to start volume production of VR200 modules in 2Q26 and to begin downstream AI ODM cabinet production in early 3Q26. However, since detailed specifications are still under discussion, there is a possibility of some schedule slippage, and we expect full-scale shipments of VR200 systems to begin from 4Q26.\n\nThe key spec changes for VR200 are as follows. The CPU LPDDR modules adopt an SOCAMM design, and a midplane structure is introduced to replace the current flying-cable-based PCIe interconnect. The TDP of the Rubin chip increases from 1.4 kW for Blackwell Ultra to 1.8/2.3 kW, and as a result, the power consumption of the NVL72 system rises from 140 kW to 220–250 kW. Two cooling solutions are offered: a combination of microchannel cold plate (MCP or MCCP) plus a gold-plated lid for the 1.8 kW Rubin GPU, and a microchannel lid (MCL) configuration for the 2.3 kW Rubin GPU. The rated capacity of the power shelf increases from 33 kW for GB200/300 to 110 kW, the NIC-to-GPU ratio doubles from 1:1 to 2:1, and NIC specs are upgraded from 400G to 800G. In addition, the level of standardization for the compute tray – including the Bianca board, midplane, HMC modules, and so on – is further enhanced.\n\nOptional manufacturing partners for the VR200 compute tray\n\nAt the GB200/300 stage, CSP customers could, with the support of the ODM assigned to them, highly customize most parts of the compute tray other than the Bianca board. According to our research, at the VR200 stage Nvidia is attempting to improve manufacturing yield and reduce lead time by providing a standardized reference for the L10 compute tray. Foxconn (Hon Hai), Wistron, and Quanta are highly likely to become Nvidia’s manufacturing partners, and CSPs will co-design the remaining components of the compute tray together with their designated ODM. We believe CSPs are still taking a conservative stance toward this proposal and that no final decision has yet been made. If Nvidia aggressively pushes standardized L10 trays, competition in the NVL72 L10 server ODM market will become more concentrated and could trigger share shifts from server brands to server ODMs.\n\nExponential increase in AI server power consumption favors power-supply vendors\n\nWe expect the installed power capacity of global data centers to increase from 117 GW in 2025 to 242 GW in 2028, implying a CAGR of 27% over this period. The strong growth in power demand is mainly driven by the expansion of AI demand and the rapid rise in AI chip TDP. At the same time, deeper vertical integration of components will lead the server power supply unit (PSU) market to record a CAGR of 64% over the same period. We therefore remain positive on Delta, and we expect the company’s server power-supply business revenues to post a CAGR of 42% in 2025–2027, with room for further upside. The rising share of high-margin AI server business also implies stronger growth momentum on the profitability side.\n\nOpportunities created for PSU vendors by the new data-center power architecture\n\nThe power-supply and power-infrastructure markets have historically been separate domains, but we believe that in future AI data-center power designs, opportunities for convergence between the two areas will increase significantly. CSPs tend to prefer more simplified power architectures in order to reduce power-conversion losses. For example, solid-state transformers (SSTs) integrate AC/DC conversion and medium-voltage step-down functions into a single piece of equipment, while high-voltage DC/DC conversion modules integrate the existing power shelf into a single onboard module. We believe Delta, with its leading position in power technologies and its complete product lineup that spans “from grid to chip,” will be the biggest beneficiary as the market expands.\n\nAdoption of immersion/liquid cooling for AI chips drives high growth in the Sidecar/CDU market\n\nWe expect the Sidecar/CDU market to grow from around USD 6 billion in 2025 to around USD 21 billion in 2028, implying a CAGR of 54% over this period. The growth drivers are the continued increase in AI server power consumption and the widespread adoption of liquid cooling technologies. According to our research, not only Nvidia’s NVL72 servers but also AMD’s GPUs and ASICs have already exceeded 1 kW in AI chip TDP, and thus are expected to start transitioning to liquid-cooling systems from next year. Delta is the leading supplier of customized Sidecar/CDUs for major US CSPs (e.g., Microsoft and Meta) and is also continuing to expand to new customers such as Amazon and Google. We expect Delta’s liquid-cooling-system business to post a CAGR of 36% in 2025–2027.\n\nDual-cooling design for Rubin GPUs: a trade-off between on-time delivery and rising TDP\n\nWe still believe that the chip designer will provide cooling solutions that allow two TDP versions of Rubin GPUs – 1.8 kW and 2.3 kW – to coexist. In the early phase of VR200 (3Q26), we expect the 1.8 kW Rubin using a microchannel cold plate plus gold-plated lid (MCP + gold-plated lid) to be the mainstream configuration. This solution is structurally very similar to the GB series, with only the cold plate, lid, and TIM (thermal interface material) upgraded. Once the MCL-related supply chain enters a mature stage from 4Q26 onward, the center of gravity in supply will gradually converge toward the higher-spec 2.3 kW version. Notably, we estimate that the two Rubin GPU versions are in fact likely to use the same physical chip, with different TDPs achieved through firmware and hardware settings.\n\nPotential adoption of a consignment model by Nvidia for VR200 (1.8 kW Rubin) MCPs, intensifying cold-plate competition\n\nAt the VR200 (1.8 kW Rubin) stage, Nvidia appears to be considering sourcing microchannel cold plates (MCPs) under a consignment model. This implies that there will be a small number of qualified suppliers (Group A) of MCPs going forward, and that their direct customer will no longer be the CSPs or enterprise customers but Nvidia itself. In the future, Nvidia will bundle its Bianca mainboard together with the MCP and sell them to end customers as a package. As a result, competition among major cold-plate suppliers is expected to become even more intense, since MCPs will no longer be customizable for individual customers and all vendors will have to actively compete to obtain Nvidia certification. This could affect both the ASP of MCPs and the allocation of volumes.\n\nAmong the potential Group A MCP suppliers for VR200, we still see AVC and Cooler Master as the most likely to receive Nvidia certification, and we also believe that Jentech, which has strong microchannel-machining capabilities, is very likely to be included as a candidate. It is understood that the company has already been supplying microchannel-version cold plates to US customers for more than a year.\n\nJentech, the lid (heat-spreader) supplier, as the biggest beneficiary: gold-plated lid for 1.8 kW Rubin, MCL for 2.3 kW Rubin\n\nRegardless of which Rubin GPU version is used, the lid (heat-spreader) structure will be significantly upgraded, and as a result the ASP will rise sharply. For the 1.8 kW Rubin, gold plating is required on the lid surface, which demands much higher machining precision, and we believe the ASP will be at least twice that of the existing lid solution used for Blackwell. In contrast, for the high-end 2.3 kW Rubin that transitions to MCL, we expect the ASP to increase by about nine times. We see Jentech as the main beneficiary of both Rubin solutions and believe it will establish itself as the sole supplier of gold-plated lids and the current MCL version.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009394572025052117, "ret_signed_m1d": 0.009394572025052117, "alpha_spy_m1d": -0.007479457453144711, "alpha_c_m1d": -0.016182158808123015, "bench_spy_m1d": 0.016874029478196828, "bench_c_m1d": 0.025576730833175132, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0375782881002088, "ret_signed_p1d": -0.0375782881002088, "alpha_spy_p1d": -0.0374147020006248, "alpha_c_p1d": -0.043021681528201294, "bench_spy_p1d": -0.00016358609958400105, "bench_c_p1d": 0.005443393427992493, "ret_p1w": -0.0041753653444676075, "ret_signed_p1w": -0.0041753653444676075, "alpha_spy_p1w": 0.02485555005356954, "alpha_c_p1w": 0.04621013983998612, "bench_spy_p1w": -0.02903091539803715, "bench_c_p1w": -0.05038550518445373, "ret_p1m": -0.02087682672233826, "ret_signed_p1m": -0.02087682672233826, "alpha_spy_p1m": -0.03534026751252761, "alpha_c_p1m": -0.023633440256354055, "bench_spy_p1m": 0.014463440790189352, "bench_c_p1m": 0.002756613534015795, "ret_p3m": 0.23695198329853873, "ret_signed_p3m": 0.23695198329853873, "alpha_spy_p3m": 0.20402961855780077, "alpha_c_p3m": 0.24064206117692866, "bench_spy_p3m": 0.03292236474073795, "bench_c_p3m": -0.00369007787838993, "ret_p6m": 1.2964509394572024, "ret_signed_p6m": 1.2964509394572024, "alpha_spy_p6m": 1.192617791067804, "alpha_c_p6m": 1.0681819616316255, "bench_spy_p6m": 0.10383314838939839, "bench_c_p6m": 0.22826897782557687, "price_path": [-0.2818, -0.31, -0.3476, -0.3236, -0.3069, -0.2944, -0.285, -0.2484, -0.2599, -0.2578, -0.2818, -0.285, -0.2902, -0.2787, -0.2422, -0.2422, -0.1816, -0.1263, -0.1326, -0.1263, -0.1273, -0.1221, -0.1273, -0.0689, -0.0731, -0.0532, -0.0678, -0.0699, -0.0772, -0.1148, -0.1148, -0.1086, -0.0908, -0.0668, -0.0167, -0.0167, 0.0303, 0.024, 0.0428, 0.0428, 0.0752, 0.0355, 0.0595, 0.0699, 0.0428, 0.0595, 0.0647, 0.0647, 0.0543, 0.0543, 0.1169, 0.1221, 0.1065, 0.0543, 0.0386, 0.0261, 0.0177, 0.0094, 0.0209, 0.0073, 0.0251, 0.0146, 0.0094, 0.0, -0.0376, -0.0438, -0.0626, -0.0668, -0.0042, -0.0658, -0.0678, -0.0501, -0.0136, -0.0167, -0.0271, 0.0157, 0.0324, 0.0407, 0.0428, 0.0219, 0.0219, 0.0136, 0.0104, -0.0136, -0.0209, -0.0376, -0.0595, -0.071, -0.0699, -0.0491, -0.0052, -0.0157, -0.0031, -0.0031, -0.0021, 0.0042, 0.0031, 0.0052, 0.0052, 0.0386, 0.0543, 0.096, 0.0752, 0.0543, 0.0595, 0.1013, 0.0752, 0.1065, 0.0908, 0.1743, 0.1691, 0.1795, 0.1848, 0.2996, 0.3152, 0.2839, 0.3048, 0.3361, 0.31, 0.2735, 0.2265, 0.2683, 0.263, 0.2004, 0.2109, 0.2317, 0.237, 0.3152, 0.3152, 0.3152, 0.3152, 0.3152, 0.3152, 0.3152, 0.3152, 0.3622, 0.4457, 0.4979, 0.4927, 0.4927, 0.4875, 0.3987, 0.31, 0.3935, 0.3779, 0.2735, 0.3466, 0.4614, 0.4301, 0.4457, 0.4196, 0.5031, 0.5292, 0.5188, 0.5397, 0.4718, 0.477, 0.618, 0.5814, 0.5762, 0.5501, 0.4405, 0.5397, 0.4927, 0.4927, 0.4927, 0.5814, 0.738, 0.7171, 0.8111, 0.8163, 0.8267, 0.8633, 0.9259, 0.9207, 0.9833, 1.1033, 1.1033, 1.0668, 1.166, 1.1086, 1.2182, 1.2599, 1.2599, 1.2599, 1.3173, 1.2599, 1.3069, 1.38, 1.2965]}
{"unit_id": "orig:1978655881572331754", "ticker_idx": 0, "ts": "2025-10-16T02:54:58+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HSBC: NVIDIA Target Price Raised Sharply to $320", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC Buy on NVDA with PT $320; CoWoS allocation +140% to 700K wafers, FY2027 DC revenue $351B vs $258B consensus, plus $251–400B Stargate revenue opportunity.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "HSBC: NVIDIA Target Price Raised Sharply to $320\n\nFrank Lee, Global Head of Technology, Hardware, and Semiconductor Research at HSBC, analyzed as follows in the report:\n\n1. Significant increase in TSMC’s CoWoS advanced packaging capacity allocation —\nNVIDIA’s fiscal 2027 CoWoS wafer allocation has increased from 480,000 to 700,000 units, a 140% rise.\n    \n2. Circular (“closed-loop”) transaction structure with OpenAI and the “Stargate” project —\nThrough this, NVIDIA is estimated to gain a GPU revenue opportunity worth between $251 billion and $400 billion.\n\n⸻\n\n■ 140% Surge in TSMC CoWoS Allocation: A Strong Signal from the Supply Chain\n\nAccording to HSBC Research, CoWoS capacity allocation, which had shown some adjustment up to the first half of 2025, was sharply revised upward starting in fiscal 2027 (beginning October).\nIn this update, HSBC raised NVIDIA’s 2027 CoWoS wafer allocation from 480,000 to 700,000 units (+140% YoY).\nThis is interpreted as the first sign of a renewed capacity expansion momentum since Q4 2025.\n\nThis expansion means more than just increased supply. HSBC emphasized:\n\n“The capacity increase implies that NVIDIA’s data center revenue could reach $351 billion in FY2027, which is 36% higher than the market consensus of $258 billion.\nIt also reaffirms that the AI GPU TAM (Total Addressable Market) is expanding beyond traditional CSP (Cloud Service Provider) CapEx dependence.”\n\nHSBC also noted that under an optimistic scenario, capacity could expand further to 800,000 wafers, leading to $390 billion in revenue and $9.68 EPS, representing an additional 11% upside from the base case.\n\nFurthermore, HSBC stated that “historically, CoWoS capacity expansion has served as a leading indicator for NVIDIA’s earnings upgrades.”\nJust as the consecutive capacity expansions in 2023 and 2024 led to unprecedented profit surges, the latest increase signals the start of a new growth cycle.\n\n⸻\n\n■ Explosive Scale of the OpenAI “Closed-Loop Transaction Structure”\n\nThe report also highlighted the triangular structure among NVIDIA, OpenAI, and Oracle:\n    • NVIDIA invests up to $100 billion in OpenAI.\n    • OpenAI signs a $300 billion, 5-year “Stargate” project contract with Oracle.\n    • Oracle uses those funds to purchase NVIDIA GPUs in bulk.\n\nThis structure functions as a “closed-loop transaction”, where the funds cycle back into NVIDIA’s revenue stream.\n\nBased on internal modeling, HSBC estimated that the 18.3 GW GPU deployment plan for OpenAI’s Stargate project alone would provide NVIDIA with a $251–400 billion revenue opportunity — roughly double the market’s current 2027 earnings forecast.\n    • If entirely using Blackwell B300 chips (1,820W each):\n→ ~10 million chips needed × $40,000 = $400 billion total revenue\n    • If entirely using Rubin chips (2,340W each):\n→ ~8 million chips needed × $42,000 = $336 billion total revenue\n    • If entirely using Rubin Ultra chips (4,176W each):\n→ ~4.4 million chips needed × $57,000 = $251 billion total revenue\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010890487260479254, "ret_signed_m1d": -0.010890487260479254, "alpha_spy_m1d": -0.017747445457740096, "alpha_c_m1d": -0.006462949315599875, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.004427537944879378, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007755381326113531, "ret_signed_p1d": 0.007755381326113531, "alpha_spy_p1d": 0.002079080967205549, "alpha_c_p1d": 0.009124429704179593, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": -0.0013690483780660623, "ret_p1w": 0.0019250797730898661, "ret_signed_p1w": 0.0019250797730898661, "alpha_spy_p1w": -0.014907029511580294, "alpha_c_p1w": -0.0024443184243461147, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.004369398197435981, "ret_p1m": 0.04598208530080106, "ret_signed_p1m": 0.04598208530080106, "alpha_spy_p1m": 0.028892604314530335, "alpha_c_p1m": 0.04108827014828087, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.004893815152520187, "ret_p3m": 0.02205788048003532, "ret_signed_p3m": 0.02205788048003532, "alpha_spy_p3m": -0.031193532462732376, "alpha_c_p3m": -0.12198031198375148, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.1440381924637868, "ret_p6m": 0.03762563590165069, "ret_signed_p6m": 0.03762563590165069, "alpha_spy_p6m": 0.0032819374572907734, "alpha_c_p6m": -0.23892383072412793, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.2765494666257786, "price_path": [-0.0574, -0.0813, -0.0607, -0.0444, -0.0458, -0.0279, -0.0347, -0.014, -0.0217, -0.0446, -0.01, -0.0196, -0.0132, -0.0058, 0.0048, 0.0013, 0.0074, -0.0013, 0.0011, -0.0075, 0.001, -0.034, -0.0353, -0.0376, -0.0211, -0.0111, -0.0003, -0.0012, -0.0091, -0.042, -0.042, -0.0607, -0.0616, -0.0559, -0.0814, -0.0743, -0.0608, -0.0247, -0.0255, -0.0219, -0.0223, -0.0381, -0.0634, -0.0306, -0.0283, 0.0099, -0.0186, -0.0266, -0.0227, -0.0199, 0.0002, 0.0262, 0.0299, 0.0389, 0.032, 0.0205, 0.0178, 0.0402, 0.0592, 0.0074, 0.0358, -0.0098, -0.0109, 0.0, 0.0078, 0.0046, -0.0036, -0.0084, 0.0019, 0.0245, 0.0532, 0.1057, 0.1388, 0.1159, 0.1137, 0.1379, 0.0928, 0.0737, 0.0345, 0.0349, 0.0948, 0.0624, 0.0659, 0.0278, 0.046, 0.0263, -0.0025, 0.0259, -0.0064, -0.0161, 0.0041, -0.0219, -0.0085, -0.0085, -0.0265, -0.0104, -0.0019, -0.0122, 0.0087, 0.0034, 0.0206, 0.0174, 0.0109, -0.0048, -0.0373, -0.0303, -0.0224, -0.0597, -0.0421, -0.0045, 0.0104, 0.0408, 0.0375, 0.0375, 0.048, 0.0353, 0.0316, 0.0259, 0.0259, 0.0388, 0.0348, 0.0299, 0.0402, 0.0178, 0.0168, 0.0173, 0.0221, 0.0074, 0.0289, 0.0244, 0.0244, -0.0205, 0.0084, 0.0167, 0.0323, 0.0257, 0.037, 0.0535, 0.0589, 0.0513, 0.021, -0.008, -0.0419, -0.0546, 0.0199, 0.0453, 0.0371, 0.0454, 0.0283, 0.0056, 0.0056, 0.0174, 0.034, 0.0336, 0.0441, 0.0536, 0.0608, 0.0757, 0.017, -0.0254, 0.0037, -0.0096, 0.0068, 0.0085, -0.0219, 0.0047, 0.0163, 0.0233, 0.0074, -0.0085, 0.0079, 0.0008, -0.0076, -0.0178, -0.05, -0.0338, -0.0363, -0.0171, -0.058, -0.0785, -0.0914, -0.0407, -0.0332, -0.0242, -0.0242, -0.0228, -0.0203, 0.0016, 0.0117, 0.0376]}
{"unit_id": "orig:1978655881572331754", "ticker_idx": 1, "ts": "2025-10-16T02:54:58+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "140% Surge in TSMC CoWoS Allocation: A Strong Signal from the Supply Chain", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares HSBC Buy on NVDA with PT $320; CoWoS allocation +140% to 700K wafers, FY2027 DC revenue $351B vs $258B consensus, plus $251–400B Stargate revenue opportunity.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "HSBC: NVIDIA Target Price Raised Sharply to $320\n\nFrank Lee, Global Head of Technology, Hardware, and Semiconductor Research at HSBC, analyzed as follows in the report:\n\n1. Significant increase in TSMC’s CoWoS advanced packaging capacity allocation —\nNVIDIA’s fiscal 2027 CoWoS wafer allocation has increased from 480,000 to 700,000 units, a 140% rise.\n    \n2. Circular (“closed-loop”) transaction structure with OpenAI and the “Stargate” project —\nThrough this, NVIDIA is estimated to gain a GPU revenue opportunity worth between $251 billion and $400 billion.\n\n⸻\n\n■ 140% Surge in TSMC CoWoS Allocation: A Strong Signal from the Supply Chain\n\nAccording to HSBC Research, CoWoS capacity allocation, which had shown some adjustment up to the first half of 2025, was sharply revised upward starting in fiscal 2027 (beginning October).\nIn this update, HSBC raised NVIDIA’s 2027 CoWoS wafer allocation from 480,000 to 700,000 units (+140% YoY).\nThis is interpreted as the first sign of a renewed capacity expansion momentum since Q4 2025.\n\nThis expansion means more than just increased supply. HSBC emphasized:\n\n“The capacity increase implies that NVIDIA’s data center revenue could reach $351 billion in FY2027, which is 36% higher than the market consensus of $258 billion.\nIt also reaffirms that the AI GPU TAM (Total Addressable Market) is expanding beyond traditional CSP (Cloud Service Provider) CapEx dependence.”\n\nHSBC also noted that under an optimistic scenario, capacity could expand further to 800,000 wafers, leading to $390 billion in revenue and $9.68 EPS, representing an additional 11% upside from the base case.\n\nFurthermore, HSBC stated that “historically, CoWoS capacity expansion has served as a leading indicator for NVIDIA’s earnings upgrades.”\nJust as the consecutive capacity expansions in 2023 and 2024 led to unprecedented profit surges, the latest increase signals the start of a new growth cycle.\n\n⸻\n\n■ Explosive Scale of the OpenAI “Closed-Loop Transaction Structure”\n\nThe report also highlighted the triangular structure among NVIDIA, OpenAI, and Oracle:\n    • NVIDIA invests up to $100 billion in OpenAI.\n    • OpenAI signs a $300 billion, 5-year “Stargate” project contract with Oracle.\n    • Oracle uses those funds to purchase NVIDIA GPUs in bulk.\n\nThis structure functions as a “closed-loop transaction”, where the funds cycle back into NVIDIA’s revenue stream.\n\nBased on internal modeling, HSBC estimated that the 18.3 GW GPU deployment plan for OpenAI’s Stargate project alone would provide NVIDIA with a $251–400 billion revenue opportunity — roughly double the market’s current 2027 earnings forecast.\n    • If entirely using Blackwell B300 chips (1,820W each):\n→ ~10 million chips needed × $40,000 = $400 billion total revenue\n    • If entirely using Rubin chips (2,340W each):\n→ ~8 million chips needed × $42,000 = $336 billion total revenue\n    • If entirely using Rubin Ultra chips (4,176W each):\n→ ~4.4 million chips needed × $57,000 = $251 billion total revenue\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01346812546673537, "ret_signed_m1d": -0.01346812546673537, "alpha_spy_m1d": -0.020325083663996213, "alpha_c_m1d": -0.009040587521855992, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.004427537944879378, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02356909544880026, "ret_signed_p1d": -0.02356909544880026, "alpha_spy_p1d": -0.02924539580770824, "alpha_c_p1d": -0.022200047070734197, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": -0.0013690483780660623, "ret_p1w": -0.02356909544880026, "ret_signed_p1w": -0.02356909544880026, "alpha_spy_p1w": -0.04040120473347042, "alpha_c_p1w": -0.02793849364623624, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.004369398197435981, "ret_p1m": -0.03703705542488689, "ret_signed_p1m": -0.03703705542488689, "alpha_spy_p1m": -0.05412653641115761, "alpha_c_p1m": -0.041930870577407076, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.004893815152520187, "ret_p3m": 0.155353519055214, "ret_signed_p3m": 0.155353519055214, "alpha_spy_p3m": 0.1021021061124463, "alpha_c_p3m": 0.0113153265914272, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.1440381924637868, "ret_p6m": 0.3556993945110889, "ret_signed_p6m": 0.3556993945110889, "alpha_spy_p6m": 0.321355696066729, "alpha_c_p6m": 0.07914992788531028, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.2765494666257786, "price_path": [-0.2287, -0.2421, -0.232, -0.232, -0.232, -0.232, -0.2387, -0.2253, -0.222, -0.222, -0.2387, -0.2287, -0.2454, -0.2086, -0.2119, -0.2086, -0.2086, -0.1951, -0.2119, -0.2086, -0.2086, -0.2052, -0.2387, -0.2287, -0.2387, -0.2153, -0.2119, -0.2018, -0.222, -0.222, -0.2186, -0.222, -0.222, -0.222, -0.2086, -0.2086, -0.1951, -0.1784, -0.1683, -0.1549, -0.1582, -0.138, -0.1481, -0.1347, -0.1481, -0.1279, -0.0976, -0.0976, -0.1111, -0.1246, -0.1246, -0.1212, -0.1077, -0.0808, -0.0572, -0.0572, -0.0337, -0.0471, -0.0303, -0.0303, -0.0471, -0.0404, -0.0135, 0.0, -0.0236, -0.0034, -0.0034, -0.0168, -0.0236, -0.0236, -0.0034, -0.0067, 0.0135, 0.0135, 0.0101, 0.0168, 0.0135, -0.0168, -0.0135, -0.0168, -0.0067, -0.0135, -0.0067, -0.0168, -0.037, -0.0269, -0.0539, -0.0606, -0.0202, -0.0673, -0.0741, -0.0471, -0.0303, -0.0337, -0.0303, -0.0505, -0.037, -0.0236, -0.0269, -0.0168, 0.0067, -0.0034, 0.0135, -0.0068, -0.0, -0.0203, -0.0304, -0.0338, -0.0338, -0.0338, -0.0102, 0.0067, 0.0101, 0.0101, 0.0202, 0.0337, 0.027, 0.0473, 0.0473, 0.0709, 0.1283, 0.152, 0.1317, 0.1385, 0.1351, 0.1418, 0.1554, 0.1554, 0.1418, 0.1756, 0.1891, 0.1993, 0.1756, 0.1891, 0.1959, 0.1858, 0.2026, 0.2297, 0.2195, 0.1993, 0.1925, 0.2162, 0.206, 0.1925, 0.2026, 0.2263, 0.2702, 0.2939, 0.2939, 0.2939, 0.2939, 0.2939, 0.2939, 0.2939, 0.2939, 0.2837, 0.3276, 0.3614, 0.3479, 0.3479, 0.3344, 0.3074, 0.2601, 0.2837, 0.277, 0.2229, 0.2499, 0.3108, 0.2736, 0.2601, 0.2466, 0.2676, 0.2913, 0.254, 0.2472, 0.2269, 0.2269, 0.2506, 0.2472, 0.2337, 0.2066, 0.193, 0.2574, 0.2269, 0.2269, 0.2269, 0.2608, 0.3218, 0.3252, 0.3557]}
{"unit_id": "orig:1985154251392418235", "ticker_idx": 0, "ts": "2025-11-03T01:17:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Securities just raised its target price for SK Hynix to 1 million won… They're basically saying there's a 78.9% upside potential", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares SK Securities PT raise on SK Hynix to ₩1M, implying 78.9% upside; author adopts the bullish stance by sharing it.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Crazy lol\n\nKorea’s SK Securities just raised its target price for SK Hynix to 1 million won…\n\nThey’re basically saying there’s a 78.9% upside potential. https://t.co/aoUSxDwcpR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09838715073877069, "ret_signed_m1d": -0.09838715073877069, "alpha_spy_m1d": -0.09651397900800696, "alpha_c_m1d": -0.08997409673737111, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.054838835866490854, "ret_signed_p1d": -0.054838835866490854, "alpha_spy_p1d": -0.04298529592404332, "alpha_c_p1d": -0.018400720224020173, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": -0.022580738743366635, "ret_signed_p1w": -0.022580738743366635, "alpha_spy_p1w": -0.019800309808329364, "alpha_c_p1w": -0.002613522001512103, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": -0.09935551328319481, "ret_signed_p1m": -0.09935551328319481, "alpha_spy_p1m": -0.09670677138394612, "alpha_c_p1m": -0.08113641365297852, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": 0.38970416322104473, "ret_signed_p3m": 0.38970416322104473, "alpha_spy_p3m": 0.37104466717996476, "alpha_c_p3m": 0.24570215906573378, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 1.102147490510704, "ret_signed_p6m": 1.102147490510704, "alpha_spy_p6m": 1.054729917297967, "alpha_c_p6m": 0.7562353678043552, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.5837, -0.578, -0.5869, -0.57, -0.5668, -0.5523, -0.5547, -0.5547, -0.5692, -0.5764, -0.5885, -0.6054, -0.5957, -0.5821, -0.5788, -0.5813, -0.5669, -0.5661, -0.5871, -0.5798, -0.5766, -0.5718, -0.5589, -0.5532, -0.5355, -0.5097, -0.5048, -0.4702, -0.4661, -0.4387, -0.4621, -0.4266, -0.4266, -0.4339, -0.4177, -0.4234, -0.425, -0.4573, -0.4371, -0.4395, -0.4194, -0.3621, -0.3621, -0.3621, -0.3621, -0.3621, -0.3621, -0.3097, -0.3306, -0.3363, -0.3185, -0.2702, -0.2492, -0.2169, -0.2274, -0.2234, -0.2282, -0.1774, -0.1371, -0.1597, -0.1, -0.0839, -0.0984, 0.0, -0.0548, -0.0661, -0.0435, -0.0645, -0.0226, -0.0016, -0.0048, -0.0129, -0.0968, -0.0226, -0.0806, -0.0935, -0.079, -0.1597, -0.1613, -0.1629, -0.1548, -0.122, -0.1445, -0.1316, -0.0994, -0.109, -0.1252, -0.122, -0.0687, -0.0864, -0.0525, -0.0881, -0.0784, -0.1058, -0.1445, -0.1107, -0.109, -0.1171, -0.0638, -0.0574, -0.0509, -0.0509, -0.0332, 0.033, 0.0508, 0.0508, 0.0508, 0.0927, 0.1234, 0.1718, 0.1976, 0.2202, 0.2009, 0.2089, 0.1912, 0.1976, 0.2089, 0.2202, 0.2331, 0.1992, 0.1944, 0.2186, 0.238, 0.1879, 0.2912, 0.3574, 0.3897, 0.4672, 0.3397, 0.464, 0.4527, 0.359, 0.3542, 0.4317, 0.4139, 0.3881, 0.4333, 0.4204, 0.4204, 0.4204, 0.4204, 0.443, 0.5317, 0.535, 0.6221, 0.6431, 0.7771, 0.7157, 0.7157, 0.5184, 0.3729, 0.5216, 0.4941, 0.3518, 0.5168, 0.5443, 0.5038, 0.4715, 0.575, 0.5685, 0.7076, 0.6381, 0.6284, 0.5087, 0.5944, 0.609, 0.5087, 0.5087, 0.4117, 0.3049, 0.4505, 0.3421, 0.4165, 0.4327, 0.4812, 0.6704, 0.6138, 0.6607, 0.6817, 0.7836, 0.837, 0.8677, 0.824, 0.8855, 0.9793, 0.9776, 0.9809, 0.976, 1.0892, 1.1021]}
{"unit_id": "thread:1996104860459225119", "ticker_idx": 0, "ts": "2025-12-03T06:30:59+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Market forecasts suggest another 20% price hike in the first quarter of 2026. While this quarter-over-quarter increase may moderate slightly from fourth quarter levels, under current \"supply-first\" conditions, hikes could reach 30-40%", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM/NAND spot prices up sharply; another 20-40% QoQ hike forecast for Q1 2026 under tight supply; Samsung RDIMM releases limited, market stays seller-driven into 2026.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory spot prices surged as Samsung's RDIMM limitedly ease supply in late 4Q25\n\nGlobal memory capacity remained tight with strong demand pushing DRAM and NAND flash spot prices sharply higher in November. Samsung Electronics planned to moderately release RDIMM modules mainly for server customers in the fourth quarter of 2025, slightly easing shortages but leaving many smaller buyers struggling amid soaring costs.\n\nDRAM and NAND prices surge sharply in November\nNovember spot market prices for both DRAM and NAND Flash showed significant increases. Notably, DDR5 prices rose substantially for two consecutive months, widening the price gap with DDR4 of the same specifications. This indicates persistent tight supply-demand dynamics, pushing DDR4 16Gb chip prices to a record US$42.5 per unit, a monthly increase exceeding 60%.\n\nAfter doubling in October, DDR5 16Gb spot prices remained robust in November with a 55% rise. Mainstream DDR4 8Gb also sustained a 56% jump. Meanwhile, Flash wafer 512Gb TLC pricing climbed to US$9, up 80% from October and tripling since February-March 2025.\n\nSamsung's tail-end RDIMM release eases shortages, but benefits are uneven\nAs memory price pressures intensify, competition for supply has grown fiercer. Industry sources report that Samsung began small-scale shipments from late November through early December, releasing batches of DDR5 RDIMM and DDR4 modules. Although these volumes cannot fully meet demand, they offer slight relief compared to near-zero availability in October, helping ease shortages for electronic system vendors heading into year-end.\n\nRecently released DDR5 RDIMM server module prices surged again, with official fourth quarter prices rising approximately 65-70% or more versus third quarter 2025, higher than earlier estimates of 50-60%. The memory market remains firmly seller-driven, as manufacturers continue raising contract prices and controlling capacity expansion, driving prices upward.\n\nMarket forecasts suggest another 20% price hike in the first quarter of 2026. While this quarter-over-quarter increase may moderate slightly from fourth quarter levels, under current \"supply-first\" conditions, hikes could reach 30-40%. Supply chain participants warn that if the first quarter of 2026 rises exceed 20%, many customers will struggle to absorb costs.\n\nSmaller buyers face growing challenges amid prioritized CSP supply\nMeanwhile, smaller firms face bleak prospects as upstream suppliers prioritize large CSP customers. Despite tight purchasing conditions, LPDDR5 products sold to servers command 50-60% higher prices than those for other applications. Consequently, consumer application orders are deferred behind server demands.\n\nSome industry insiders reveal that after Samsung halted pricing quotes in October, it resumed DRAM chip quotations mid-November with average contract price increases of 30-40%. Yet actual DDR5 availability remains scarce, forcing inventory consumption while spot prices jumped 3-5x, escalating procurement cost pressures.\n\nNAND Flash prices followed suit. Phison CEO Khein-Seng Pua noted nearly doubled prices in the fourth quarter. Sources indicate US-based NAND giants raised prices repeatedly, with November quotes 100-150% above October. Some NAND makers held prices steady in the fourth quarter but faced severe supply shortages.\n\nExpectations point to even steeper hikes in the first quarter of 2026.\n\nBenefiting from soaring memory average selling prices (ASPs), multiple memory module manufacturers across Taiwan and China have already met their annual sales targets ahead of schedule. With scarce resources at original makers and in the spot market, the industry favors a slower shipment pace, adopting selective or conservative delivery approaches. This trend is expected to keep overall memory prices elevated throughout 2026.\n\n---\n\nhttps://t.co/88lknakVB9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007701796225511888, "ret_signed_m1d": 0.007701796225511888, "alpha_spy_m1d": 0.011152614635920283, "alpha_c_m1d": 0.021259460743635567, "bench_spy_m1d": -0.003450818410408396, "bench_c_m1d": -0.01355766451812368, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014815480714484353, "ret_signed_p1d": -0.014815480714484353, "alpha_spy_p1d": -0.01554653686848273, "alpha_c_p1d": -0.007213301702097219, "bench_spy_p1d": 0.0007310561539983773, "bench_c_p1d": -0.0076021790123871336, "ret_p1w": 0.07436475079044642, "ret_signed_p1w": 0.07436475079044642, "alpha_spy_p1w": 0.06898374667202874, "alpha_c_p1w": 0.047661095130861655, "bench_spy_p1w": 0.005381004118417687, "bench_c_p1w": 0.02670365565958477, "ret_p1m": 0.1838845306993213, "ret_signed_p1m": 0.1838845306993213, "alpha_spy_p1m": 0.18381874013322186, "alpha_c_p1m": 0.19244565663440794, "bench_spy_p1m": 6.57905660994551e-05, "bench_c_p1m": -0.008561125935086622, "ret_p3m": 0.9235145591379187, "ret_signed_p3m": 0.9235145591379187, "alpha_spy_p3m": 0.9169079476796794, "alpha_c_p3m": 0.8047217535082433, "bench_spy_p3m": 0.006606611458239264, "bench_c_p3m": 0.11879280562967542, "ret_p6m": 2.662742678069237, "ret_signed_p6m": 2.662742678069237, "alpha_spy_p6m": 2.553066214173044, "alpha_c_p6m": 2.0114090243588, "bench_spy_p6m": 0.10967646389619268, "bench_c_p6m": 0.6513336537104373, "price_path": [-0.4274, -0.4232, -0.4068, -0.3868, -0.3674, -0.3386, -0.3329, -0.3119, -0.3228, -0.2902, -0.299, -0.2888, -0.2764, -0.283, -0.2883, -0.3087, -0.2877, -0.2846, -0.2493, -0.2136, -0.2077, -0.2032, -0.2107, -0.1953, -0.2013, -0.1821, -0.1776, -0.1932, -0.1688, -0.127, -0.1187, -0.0998, -0.112, -0.1124, -0.1092, -0.0653, -0.0385, -0.0523, -0.0201, -0.0063, -0.0012, 0.0626, -0.0014, 0.0084, 0.0135, 0.001, 0.0472, 0.0469, 0.0498, 0.0345, -0.0007, 0.031, -0.0188, -0.0315, -0.0479, -0.088, -0.059, -0.0504, -0.0281, -0.0136, -0.0227, -0.0113, 0.0077, 0.0, -0.0148, 0.012, 0.0492, 0.0469, 0.0744, 0.0514, 0.0354, 0.0069, -0.0211, -0.0021, 0.0304, 0.0479, 0.0964, 0.1016, 0.1176, 0.1176, 0.1403, 0.1887, 0.1942, 0.1839, 0.1839, 0.2699, 0.3075, 0.3727, 0.3836, 0.3671, 0.3863, 0.3897, 0.3682, 0.3725, 0.3929, 0.4504, 0.4565, 0.4339, 0.4802, 0.5103, 0.5198, 0.4861, 0.5786, 0.6483, 0.6556, 0.6542, 0.6068, 0.6819, 0.6258, 0.5643, 0.5771, 0.6151, 0.592, 0.6415, 0.6982, 0.6984, 0.6984, 0.6815, 0.7117, 0.7432, 0.7922, 0.7924, 0.8432, 0.878, 0.9554, 0.9231, 0.9235, 0.7341, 0.6363, 0.7489, 0.6896, 0.6163, 0.7438, 0.783, 0.7422, 0.7456, 0.8233, 0.8658, 0.9647, 0.8882, 0.8506, 0.7374, 0.7678, 0.7519, 0.6479, 0.6504, 0.5529, 0.5066, 0.6759, 0.5969, 0.6498, 0.6944, 0.7232, 0.8805, 0.8595, 0.8822, 0.8825, 0.994, 1.015, 1.0488, 1.0245, 1.0332, 1.0841, 1.133, 1.1485, 1.152, 1.2501, 1.218, 1.2468, 1.223, 1.2587, 1.4434, 1.5343, 1.7726, 1.7939, 1.9464, 2.1875, 2.0984, 2.2526, 2.2481, 2.0017, 1.9867, 1.9361, 1.985, 2.2214, 2.1837, 2.1837, 2.4781, 2.6657, 2.6627]}
{"unit_id": "thread:1996104860459225119", "ticker_idx": 1, "ts": "2025-12-03T06:30:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Benefiting from soaring memory average selling prices (ASPs), multiple memory module manufacturers across Taiwan and China have already met their annual sales targets ahead of schedule", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM/NAND spot prices up sharply; another 20-40% QoQ hike forecast for Q1 2026 under tight supply; Samsung RDIMM releases limited, market stays seller-driven into 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory spot prices surged as Samsung's RDIMM limitedly ease supply in late 4Q25\n\nGlobal memory capacity remained tight with strong demand pushing DRAM and NAND flash spot prices sharply higher in November. Samsung Electronics planned to moderately release RDIMM modules mainly for server customers in the fourth quarter of 2025, slightly easing shortages but leaving many smaller buyers struggling amid soaring costs.\n\nDRAM and NAND prices surge sharply in November\nNovember spot market prices for both DRAM and NAND Flash showed significant increases. Notably, DDR5 prices rose substantially for two consecutive months, widening the price gap with DDR4 of the same specifications. This indicates persistent tight supply-demand dynamics, pushing DDR4 16Gb chip prices to a record US$42.5 per unit, a monthly increase exceeding 60%.\n\nAfter doubling in October, DDR5 16Gb spot prices remained robust in November with a 55% rise. Mainstream DDR4 8Gb also sustained a 56% jump. Meanwhile, Flash wafer 512Gb TLC pricing climbed to US$9, up 80% from October and tripling since February-March 2025.\n\nSamsung's tail-end RDIMM release eases shortages, but benefits are uneven\nAs memory price pressures intensify, competition for supply has grown fiercer. Industry sources report that Samsung began small-scale shipments from late November through early December, releasing batches of DDR5 RDIMM and DDR4 modules. Although these volumes cannot fully meet demand, they offer slight relief compared to near-zero availability in October, helping ease shortages for electronic system vendors heading into year-end.\n\nRecently released DDR5 RDIMM server module prices surged again, with official fourth quarter prices rising approximately 65-70% or more versus third quarter 2025, higher than earlier estimates of 50-60%. The memory market remains firmly seller-driven, as manufacturers continue raising contract prices and controlling capacity expansion, driving prices upward.\n\nMarket forecasts suggest another 20% price hike in the first quarter of 2026. While this quarter-over-quarter increase may moderate slightly from fourth quarter levels, under current \"supply-first\" conditions, hikes could reach 30-40%. Supply chain participants warn that if the first quarter of 2026 rises exceed 20%, many customers will struggle to absorb costs.\n\nSmaller buyers face growing challenges amid prioritized CSP supply\nMeanwhile, smaller firms face bleak prospects as upstream suppliers prioritize large CSP customers. Despite tight purchasing conditions, LPDDR5 products sold to servers command 50-60% higher prices than those for other applications. Consequently, consumer application orders are deferred behind server demands.\n\nSome industry insiders reveal that after Samsung halted pricing quotes in October, it resumed DRAM chip quotations mid-November with average contract price increases of 30-40%. Yet actual DDR5 availability remains scarce, forcing inventory consumption while spot prices jumped 3-5x, escalating procurement cost pressures.\n\nNAND Flash prices followed suit. Phison CEO Khein-Seng Pua noted nearly doubled prices in the fourth quarter. Sources indicate US-based NAND giants raised prices repeatedly, with November quotes 100-150% above October. Some NAND makers held prices steady in the fourth quarter but faced severe supply shortages.\n\nExpectations point to even steeper hikes in the first quarter of 2026.\n\nBenefiting from soaring memory average selling prices (ASPs), multiple memory module manufacturers across Taiwan and China have already met their annual sales targets ahead of schedule. With scarce resources at original makers and in the spot market, the industry favors a slower shipment pace, adopting selective or conservative delivery approaches. This trend is expected to keep overall memory prices elevated throughout 2026.\n\n---\n\nhttps://t.co/88lknakVB9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010526357787611462, "ret_signed_m1d": -0.010526357787611462, "alpha_spy_m1d": -0.007075539377203066, "alpha_c_m1d": -0.008146927763885703, "bench_spy_m1d": -0.003450818410408396, "bench_c_m1d": -0.0023794300237257593, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00574158126626001, "ret_signed_p1d": 0.00574158126626001, "alpha_spy_p1d": 0.005010525112261632, "alpha_c_p1d": 0.002017311069465233, "bench_spy_p1d": 0.0007310561539983773, "bench_c_p1d": 0.0037242701967947767, "ret_p1w": 0.033492833412013256, "ret_signed_p1w": 0.033492833412013256, "alpha_spy_p1w": 0.02811182929359557, "alpha_c_p1w": 0.0077334034115035255, "bench_spy_p1w": 0.005381004118417687, "bench_c_p1w": 0.02575943000050973, "ret_p1m": 0.15294595363057706, "ret_signed_p1m": 0.15294595363057706, "alpha_spy_p1m": 0.1528801630644776, "alpha_c_p1m": 0.15850914291524854, "bench_spy_p1m": 6.57905660994551e-05, "bench_c_p1m": -0.005563189284671477, "ret_p3m": 1.0818415096225875, "ret_signed_p3m": 1.0818415096225875, "alpha_spy_p3m": 1.0752348981643483, "alpha_c_p3m": 1.11800382288276, "bench_spy_p3m": 0.006606611458239264, "bench_c_p3m": -0.03616231326017261, "ret_p6m": 1.8859218490014564, "ret_signed_p6m": 1.8859218490014564, "alpha_spy_p6m": 1.7762453851052638, "alpha_c_p6m": 1.5936973869032294, "bench_spy_p6m": 0.10967646389619268, "bench_c_p6m": 0.29222446209822706, "price_path": [-0.3379, -0.3322, -0.3188, -0.3083, -0.3007, -0.2817, -0.2712, -0.2436, -0.255, -0.235, -0.235, -0.2045, -0.1931, -0.1864, -0.1797, -0.2064, -0.1943, -0.1971, -0.177, -0.1411, -0.1411, -0.1411, -0.1411, -0.1411, -0.1411, -0.0967, -0.1072, -0.1234, -0.0909, -0.0651, -0.0632, -0.0612, -0.067, -0.0565, -0.0766, -0.0545, -0.0239, -0.0478, -0.0383, -0.0038, 0.0287, 0.0632, 0.0038, -0.0373, -0.0507, -0.0632, -0.0373, -0.0096, -0.0134, -0.0163, -0.0699, -0.0373, -0.0641, -0.0766, -0.0373, -0.0928, -0.0746, -0.0498, -0.0163, -0.0096, -0.0383, -0.0354, -0.0105, 0.0, 0.0057, 0.0373, 0.0478, 0.0373, 0.0335, 0.0268, 0.0421, 0.0029, -0.0163, 0.0325, 0.0297, 0.0172, 0.0574, 0.067, 0.0632, 0.0632, 0.1196, 0.1491, 0.1529, 0.1529, 0.1529, 0.2356, 0.328, 0.3356, 0.3558, 0.3347, 0.3366, 0.3347, 0.3231, 0.3491, 0.3837, 0.4318, 0.4357, 0.3962, 0.4376, 0.4645, 0.4626, 0.4626, 0.5337, 0.5616, 0.5453, 0.5434, 0.4462, 0.6107, 0.626, 0.5318, 0.5251, 0.6001, 0.5943, 0.6135, 0.7174, 0.7424, 0.7424, 0.7424, 0.7424, 0.827, 0.828, 0.8559, 0.9232, 0.9568, 1.0963, 1.0818, 1.0818, 0.8761, 0.6559, 0.8424, 0.8097, 0.6684, 0.8068, 0.827, 0.8068, 0.7645, 0.8145, 0.8645, 1.0049, 0.928, 0.9174, 0.7914, 0.8241, 0.8174, 0.7318, 0.7318, 0.6988, 0.6111, 0.8274, 0.719, 0.7942, 0.8607, 0.8934, 1.0283, 0.9657, 0.985, 0.9368, 0.9898, 1.0332, 1.0958, 1.0813, 1.0669, 1.1102, 1.0958, 1.1632, 1.1151, 1.1632, 1.1391, 1.1777, 1.1247, 1.1247, 1.2403, 1.2403, 1.5631, 1.6161, 1.5872, 1.751, 1.6884, 1.7366, 1.8522, 1.6065, 1.7077, 1.6547, 1.6595, 1.8859, 1.8185, 1.8185, 1.8811, 1.9582, 1.8859]}
{"unit_id": "orig:2004058112211444085", "ticker_idx": 0, "ts": "2025-12-25T05:14:22+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "revise up earnings forecasts for Samsung Electronics (SEC; 005930 KS, Buy)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura reiterates Buy on Samsung & SK Hynix, raising earnings estimates on stronger commodity DRAM/NAND prices; memory supercycle to persist through 2027F.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura Global memory: Stronger memory prices drive up earnings\n\nRevising up earnings for 4Q25F and beyond, reflecting a stronger commodity memory price outlook\nWe remain the most bullish in our earnings outlook for memory companies on the Street – since our report on memory supercycle published in September – given: (1) the surprisingly strong 2026F demand for commodity DRAM from AI servers, in addition to the strong HBM demand, and (2) the sharply increased 2026F commodity DRAM demand from conventional servers and SSD demand from both AI and non-AI servers, leading to memory customers commencing proactive purchases, with suppliers raising prices more than our previous forecasts. Data center expansion plans by global AI companies have exceeded expectations despite recent concerns about AI capex capability, and we expect additional government-led AI investments to continue globally, along with the US investment for the Genesis Mission for AI. Still, we expect meaningful memory supply increase to happen in 2028F at the earliest. We therefore reiterate that the memory supercycle should continue until at least 2027F, and revise up earnings forecasts for Samsung Electronics (SEC; 005930 KS, Buy) and SK Hynix (000660 KS, Buy) to reflect the stronger-than-expected memory prices.\n\nStrong commodity prices likely to drive memory players’ 4Q25F OP surprise\nWe forecast memory players’ 4Q25F OP to strongly beat the Bloomberg consensus based on greater-than-expected rise in commodity memory prices. DRAM prices for consumer products, including PC and mobile, have increased 30-40% q-q, and DRAM prices for servers, for which demand is tangibly growing, have increased 40-60% q-q, depending on the category, by our estimates.\n\nRegarding NAND, while price increases for certain mobile applications have been relatively slow, we estimate that enterprise SSD prices will increase by 30-40% q-q in 4Q25F. Hence, our previous forecast of commodity DRAM’s average price growth of +25% q-q in 4Q25F appears conservative, and thus we revise up 4Q25F OP for SEC/Hynix by 22%/8% to KRW21.5tn/17.5tn, reflecting the stronger-than-expected memory price increase. SEC/Hynix’s commodity memory profitability has surpassed/is approaching their respective HBM profitability, in our view. We see room for an additional 20-30% price increase for commodity memory prices in 1H26F, and thus revise up our 2026F OP for SEC/Hynix by 21.5%/9.7% to KRW133tn/109tn (vs. Bloomberg consensus of KRW93tn/79tn).\n\nSupply shortage likely to last until 2027F; suppliers can be flexible in product mix depending on each product’s profitability\nInvestors have already been monitoring memory players’ supply expansion efforts, but we think memory supply shortage will continue until at least 2027F, considering greenfield, brownfield, and upgrade plans until 2028. Memory companies have allocated capacity primarily to HBM so far, but we expect them to more flexibly produce commodities than before, as commodity profitability rapidly improves. This should also underpin their bargaining power for HBM price negotiations with ASIC customers in the future.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05310536393950405, "ret_signed_p1d": 0.05310536393950405, "alpha_spy_p1d": 0.05320677897174719, "alpha_c_p1d": 0.05153331060748512, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.00157205333201893, "ret_p1w": 0.08445416491506297, "ret_signed_p1w": 0.08445416491506297, "alpha_spy_p1w": 0.09670827446226882, "alpha_c_p1w": 0.10038042063887109, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.015926255723808125, "ret_p1m": 0.3756920117600133, "ret_signed_p1m": 0.3756920117600133, "alpha_spy_p1m": 0.3773578245842101, "alpha_c_p1m": 0.3839627833788012, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": -0.008270771618787864, "ret_p3m": 0.7157711166213605, "ret_signed_p3m": 0.7157711166213605, "alpha_spy_p3m": 0.7670706341287067, "alpha_c_p3m": 0.7839921117726819, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": -0.0682209951513214, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2421, -0.2448, -0.2259, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1503, -0.1602, -0.1755, -0.1449, -0.1206, -0.1188, -0.117, -0.1224, -0.1125, -0.1314, -0.1107, -0.0819, -0.1044, -0.0954, -0.063, -0.0324, 0.0, -0.0558, -0.0945, -0.1071, -0.1188, -0.0945, -0.0684, -0.072, -0.0747, -0.1251, -0.0945, -0.1197, -0.1314, -0.0945, -0.1467, -0.1296, -0.1062, -0.0747, -0.0684, -0.0954, -0.0927, -0.0693, -0.0594, -0.054, -0.0243, -0.0144, -0.0243, -0.0279, -0.0342, -0.0198, -0.0567, -0.0747, -0.0288, -0.0315, -0.0432, -0.0054, 0.0036, 0.0, 0.0, 0.0531, 0.0808, 0.0845, 0.0845, 0.0845, 0.1622, 0.2491, 0.2563, 0.2753, 0.2554, 0.2572, 0.2554, 0.2445, 0.269, 0.3015, 0.3467, 0.3504, 0.3133, 0.3522, 0.3775, 0.3757, 0.3757, 0.4426, 0.4689, 0.4535, 0.4517, 0.3603, 0.515, 0.5295, 0.4408, 0.4345, 0.505, 0.4996, 0.5177, 0.6154, 0.6389, 0.6389, 0.6389, 0.6389, 0.7185, 0.7194, 0.7456, 0.8089, 0.8406, 0.9717, 0.9582, 0.9582, 0.7646, 0.5575, 0.733, 0.7022, 0.5692, 0.6995, 0.7185, 0.6995, 0.6597, 0.7067, 0.7538, 0.8858, 0.8135, 0.8035, 0.685, 0.7158, 0.7094, 0.6289, 0.6289, 0.5979, 0.5154, 0.7189, 0.6169, 0.6876, 0.7501, 0.781, 0.9078, 0.8489, 0.8671, 0.8217, 0.8716, 0.9124, 0.9713, 0.9577, 0.9441, 0.9849, 0.9713, 1.0347, 0.9894, 1.0347, 1.0121, 1.0483, 0.9985, 0.9985, 1.1072, 1.1072, 1.4109, 1.4607, 1.4335, 1.5876, 1.5287, 1.574, 1.6828, 1.4516, 1.5468, 1.497, 1.5015, 1.7145, 1.651, 1.651, 1.7099, 1.7825, 1.7145, 1.8731, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2004058112211444085", "ticker_idx": 1, "ts": "2025-12-25T05:14:22+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix (000660 KS, Buy)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura reiterates Buy on Samsung & SK Hynix, raising earnings estimates on stronger commodity DRAM/NAND prices; memory supercycle to persist through 2027F.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura Global memory: Stronger memory prices drive up earnings\n\nRevising up earnings for 4Q25F and beyond, reflecting a stronger commodity memory price outlook\nWe remain the most bullish in our earnings outlook for memory companies on the Street – since our report on memory supercycle published in September – given: (1) the surprisingly strong 2026F demand for commodity DRAM from AI servers, in addition to the strong HBM demand, and (2) the sharply increased 2026F commodity DRAM demand from conventional servers and SSD demand from both AI and non-AI servers, leading to memory customers commencing proactive purchases, with suppliers raising prices more than our previous forecasts. Data center expansion plans by global AI companies have exceeded expectations despite recent concerns about AI capex capability, and we expect additional government-led AI investments to continue globally, along with the US investment for the Genesis Mission for AI. Still, we expect meaningful memory supply increase to happen in 2028F at the earliest. We therefore reiterate that the memory supercycle should continue until at least 2027F, and revise up earnings forecasts for Samsung Electronics (SEC; 005930 KS, Buy) and SK Hynix (000660 KS, Buy) to reflect the stronger-than-expected memory prices.\n\nStrong commodity prices likely to drive memory players’ 4Q25F OP surprise\nWe forecast memory players’ 4Q25F OP to strongly beat the Bloomberg consensus based on greater-than-expected rise in commodity memory prices. DRAM prices for consumer products, including PC and mobile, have increased 30-40% q-q, and DRAM prices for servers, for which demand is tangibly growing, have increased 40-60% q-q, depending on the category, by our estimates.\n\nRegarding NAND, while price increases for certain mobile applications have been relatively slow, we estimate that enterprise SSD prices will increase by 30-40% q-q in 4Q25F. Hence, our previous forecast of commodity DRAM’s average price growth of +25% q-q in 4Q25F appears conservative, and thus we revise up 4Q25F OP for SEC/Hynix by 22%/8% to KRW21.5tn/17.5tn, reflecting the stronger-than-expected memory price increase. SEC/Hynix’s commodity memory profitability has surpassed/is approaching their respective HBM profitability, in our view. We see room for an additional 20-30% price increase for commodity memory prices in 1H26F, and thus revise up our 2026F OP for SEC/Hynix by 21.5%/9.7% to KRW133tn/109tn (vs. Bloomberg consensus of KRW93tn/79tn).\n\nSupply shortage likely to last until 2027F; suppliers can be flexible in product mix depending on each product’s profitability\nInvestors have already been monitoring memory players’ supply expansion efforts, but we think memory supply shortage will continue until at least 2027F, considering greenfield, brownfield, and upgrade plans until 2028. Memory companies have allocated capacity primarily to HBM so far, but we expect them to more flexibly produce commodities than before, as commodity profitability rapidly improves. This should also underpin their bargaining power for HBM price negotiations with ASIC customers in the future.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018707539869095013, "ret_signed_p1d": 0.018707539869095013, "alpha_spy_p1d": 0.01880895490133816, "alpha_c_p1d": 0.014066927348801661, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.004640612520293352, "ret_p1w": 0.10714283432004468, "ret_signed_p1w": 0.10714283432004468, "alpha_spy_p1w": 0.11939694386725053, "alpha_c_p1w": 0.11823657734104598, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.011093743021001301, "ret_p1m": 0.30442176472254756, "ret_signed_p1m": 0.30442176472254756, "alpha_spy_p1m": 0.3060875775467443, "alpha_c_p1m": 0.20578656252107863, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": 0.09863520220146893, "ret_p3m": 0.6799649210329166, "ret_signed_p3m": 0.6799649210329166, "alpha_spy_p3m": 0.7312644385402628, "alpha_c_p3m": 0.5964325548618816, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": 0.08353236617103499, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4069, -0.4094, -0.3882, -0.3279, -0.3279, -0.3279, -0.3279, -0.3279, -0.3279, -0.2726, -0.2947, -0.3007, -0.282, -0.231, -0.2089, -0.1749, -0.186, -0.1817, -0.1868, -0.1333, -0.0908, -0.1146, -0.0517, -0.0347, -0.05, 0.0537, -0.0041, -0.016, 0.0078, -0.0143, 0.0299, 0.052, 0.0486, 0.0401, -0.0483, 0.0299, -0.0313, -0.0449, -0.0296, -0.1146, -0.1163, -0.118, -0.1095, -0.0748, -0.0986, -0.085, -0.051, -0.0612, -0.0782, -0.0748, -0.0187, -0.0374, -0.0017, -0.0391, -0.0289, -0.0578, -0.0986, -0.0629, -0.0612, -0.0697, -0.0136, -0.0068, 0.0, 0.0, 0.0187, 0.0884, 0.1071, 0.1071, 0.1071, 0.1514, 0.1837, 0.2347, 0.2619, 0.2857, 0.2653, 0.2738, 0.2551, 0.2619, 0.2738, 0.2857, 0.2993, 0.2636, 0.2585, 0.284, 0.3044, 0.2517, 0.3605, 0.4303, 0.4643, 0.5459, 0.4116, 0.5425, 0.5306, 0.432, 0.4269, 0.5085, 0.4898, 0.4626, 0.5102, 0.4966, 0.4966, 0.4966, 0.4966, 0.5204, 0.6139, 0.6173, 0.7092, 0.7313, 0.8725, 0.8078, 0.8078, 0.5999, 0.4465, 0.6033, 0.5743, 0.4244, 0.5982, 0.6271, 0.5846, 0.5505, 0.6595, 0.6527, 0.7992, 0.726, 0.7157, 0.5897, 0.68, 0.6953, 0.5897, 0.5897, 0.4874, 0.375, 0.5283, 0.4142, 0.4925, 0.5096, 0.5607, 0.76, 0.7004, 0.7498, 0.772, 0.8793, 0.9355, 0.9679, 0.9219, 0.9867, 1.0855, 1.0838, 1.0872, 1.0821, 1.2013, 1.215, 1.203, 1.1911, 1.1911, 1.4654, 1.4654, 1.7278, 1.8181, 1.8726, 2.2032, 2.1265, 2.3667, 2.3565, 2.0992, 2.135, 1.9732, 1.9732, 2.3054, 2.3071, 2.3071, 2.4962, 2.8217, 2.9007, 2.9757, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1990945292699775428", "ticker_idx": 0, "ts": "2025-11-19T00:48:42+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's revenue growth potential is expected to far exceed the market's previous expectations", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares UBS analysis that TSMC earns $1–2B per GW of AI server buildout; $34.4B revenue opportunity from OpenAI deals alone, far exceeding prior market expectations.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Why is TSMC the one sweeping up the money in AI infrastructure? UBS sees it this way: TSMC will earn USD 1–2 billion per GW\n\nAmid the global investment boom in cloud AI servers, TSMC, the leading large foundry, is facing an unprecedented growth opportunity.\n\nAccording to the latest report from UBS, every 1GW server project brings TSMC a revenue opportunity of USD 1–2 billion, which corresponds to 1.0–1.5% of its expected 2025 sales. As OpenAI and multiple hyperscale cloud service providers have announced server build-out plans on the order of tens of gigawatts, TSMC’s revenue growth potential is expected to far exceed the market’s previous expectations.\n\nThe demand differences between different AI platforms are very pronounced\n\nAccording to the report’s analysis, the total revenue that TSMC will actually generate from NVIDIA’s next-generation AI GPU platforms will gradually increase over time. For each 1GW server build-out, TSMC earns about USD 1.1 billion from the Blackwell Ultra/Rubin platform, and when the transition is made to the Rubin Ultra/Feynman platform, this figure rises to USD 1.4–1.9 billion.\n\nThis growth is the result of multiple factors acting in combination. The key drivers cited include migration to more advanced process nodes, an increase in the number of GPUs installed per rack, the application of advanced packaging technologies such as CoWoS, and the potential transition to panel-level packaging (PLP) in 2028.\n\nFrom a capacity consumption perspective, NVIDIA’s Blackwell Ultra requires about 3,500 wafers per month, of which xPU chips consume 2,300 wafers per month, CPUs 500 wafers per month, and networking chips 700 wafers per month. CoWoS packaging capacity demand comes to about 2,300 wafers per month.\n\nBroadcom’s ASIC solutions show a striking advantage in terms of efficiency. Thanks to higher efficiency tailored to specific workloads, ASIC chips can drive higher chip demand than GPUs. The TPU v7p that Broadcom designed for Google requires about 4,900 wafers per month of N3 capacity per 1GW, a level far above NVIDIA’s 2,000–4,000 wafers per month.\n\nOn the basis of revenue share relative to rack value, ASIC solutions generate greater actual total revenue for TSMC, at as high as 10–11%. In contrast, GPU solutions are at the 4–6% level. Specifically, in the case of Google’s TPU v7p project, TSMC’s revenue opportunity per 1GW comes to USD 1.895 billion.\n\nAMD’s AI GPU capacity demand is the lowest among the major players. The upcoming MI455 adopts a chiplet structure, so the N2 capacity required per 1GW is only about 1,000 wafers per month, resulting in a TSMC revenue opportunity of roughly USD 745 million.\n\nThe massive business opportunity created by a 1GW server build-out\n\nThe UBS report breaks down AI server investments in detail. To support a 1GW server build-out, TSMC needs to provide 2,000–5,000 wafers per month (kwpm) of capacity on advanced nodes such as N3 or N2, along with an additional 3,000–6,000 wafers per month of advanced CoWoS packaging capacity.\n\nFrom a capital expenditure (capex) perspective, each 1GW project entails wafer fab equipment (WFE) investments of USD 1–2 billion for logic chip production, including both front-end capacity and CoWoS packaging capacity. This means that every time AI infrastructure is expanded, it directly translates into increased demand for TSMC capacity and higher capex.\n\nThe USD 34.4 billion revenue potential hidden in the OpenAI deal\n\nUBS conducted a detailed estimate of the deals that OpenAI has already announced. The total scale of the deals OpenAI has signed with NVIDIA, AMD, and Google is 26GW, consisting of 10GW with NVIDIA, 6GW with AMD, and 10GW with Google.\n\nBased on these deals, TSMC’s potential revenue amounts to USD 34.4 billion. Of this, USD 11.0 billion comes from NVIDIA, USD 4.5 billion from AMD, and USD 18.9 billion from Google. Assuming an operating margin of 40–45%, these projects could generate around USD 14.6 billion in operating profit for TSMC, clearly illustrating how the AI infrastructure build-out boom can powerfully boost TSMC’s earnings.\n\n$TSM $AVGO $NVDA $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015794850171595765, "ret_signed_m1d": -0.015794850171595765, "alpha_spy_m1d": -0.011946517266853363, "alpha_c_m1d": 0.002323218403225802, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.018118068574821566, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017246947899663745, "ret_signed_p1d": -0.017246947899663745, "alpha_spy_p1d": -0.002004611216898322, "alpha_c_p1d": 0.024999144997802314, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.04224609289746606, "ret_p1w": 0.026879565653030957, "ret_signed_p1w": 0.026879565653030957, "alpha_spy_p1w": 0.0011488242697468287, "alpha_c_p1w": 0.002515977038469952, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.024363588614561005, "ret_p1m": 0.010902361044397457, "ret_signed_p1m": 0.010902361044397457, "alpha_spy_p1m": -0.00998394402984748, "alpha_c_p1m": -0.012106101593771124, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.02300846263816858, "ret_p3m": 0.3009491559315183, "ret_signed_p3m": 0.3009491559315183, "alpha_spy_p3m": 0.2690543936586869, "alpha_c_p3m": 0.09605683190038916, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": 0.2048923240311291, "ret_p6m": 0.487508938410278, "ret_signed_p6m": 0.487508938410278, "alpha_spy_p6m": 0.3519883893889846, "alpha_c_p6m": -0.22159892225831346, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.7091078606685914, "price_path": [-0.1775, -0.1683, -0.1572, -0.1552, -0.1588, -0.185, -0.185, -0.1937, -0.1831, -0.1696, -0.1407, -0.1273, -0.1142, -0.0806, -0.086, -0.0845, -0.0772, -0.0719, -0.0693, -0.0486, -0.062, -0.0345, 0.0012, -0.0059, -0.0202, -0.0319, -0.0324, -0.0109, 0.0216, 0.0203, 0.0348, 0.0709, 0.0413, 0.0784, 0.062, -0.0061, 0.0727, 0.0481, 0.0791, 0.0619, 0.045, 0.0543, 0.043, 0.0231, 0.0296, 0.0446, 0.0562, 0.0679, 0.0805, 0.0738, 0.064, 0.0796, 0.0414, 0.0399, 0.0243, 0.0146, 0.0457, 0.0312, 0.0292, -0.0006, 0.0087, -0.0013, -0.0158, 0.0, -0.0172, -0.0259, 0.008, 0.0082, 0.0269, 0.0269, 0.0324, 0.0188, 0.0344, 0.0463, 0.0374, 0.0437, 0.0691, 0.0745, 0.0983, 0.0825, 0.037, 0.0218, 0.0187, -0.0165, 0.0109, 0.0261, 0.0414, 0.0545, 0.061, 0.061, 0.0754, 0.0686, 0.0638, 0.0791, 0.0791, 0.1349, 0.1443, 0.1627, 0.1316, 0.1293, 0.1492, 0.1781, 0.1761, 0.1616, 0.2132, 0.2159, 0.2159, 0.1617, 0.1581, 0.1625, 0.1891, 0.1815, 0.2014, 0.2155, 0.2057, 0.1738, 0.2122, 0.1923, 0.1567, 0.1744, 0.2388, 0.2621, 0.2851, 0.3284, 0.3071, 0.3009, 0.3009, 0.2933, 0.2864, 0.2797, 0.3158, 0.314, 0.3698, 0.3768, 0.3381, 0.3301, 0.3107, 0.254, 0.2693, 0.2566, 0.2034, 0.2382, 0.2325, 0.259, 0.1957, 0.2013, 0.2082, 0.232, 0.2092, 0.2064, 0.1724, 0.2052, 0.2223, 0.2383, 0.1613, 0.1635, 0.1271, 0.2034, 0.2161, 0.2073, 0.2073, 0.217, 0.2297, 0.303, 0.3015, 0.3197, 0.316, 0.3528, 0.3357, 0.2939, 0.3194, 0.3042, 0.3107, 0.3797, 0.3627, 0.4332, 0.4421, 0.3971, 0.4024, 0.4104, 0.4161, 0.4301, 0.4045, 0.4938, 0.4748, 0.466, 0.4406, 0.4147, 0.4237, 0.4875]}
{"unit_id": "orig:1976535639446389242", "ticker_idx": 0, "ts": "2025-10-10T06:29:53+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS raised NVIDIA's total GPU production at TSMC to 6.9 million units in 2025 and 7.4 million units in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for NVDA and AVGO; favors TSMC and ASE as key beneficiaries of AI packaging buildout.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS Raises CoWoS Demand Forecasts for NVIDIA and Broadcom, With Rubin and CPX as Key Growth Drivers\n\nUBS (T. Arcuri, Oct 8, 2025)\n\nUBS has raised its CoWoS demand forecasts for NVIDIA and Broadcom in its latest report. The firm had previously revised TSMC’s CoWoS capacity forecast for the end of 2026 from 100,000 wafers per month to 110,000 wafers per month.\n\nFor NVIDIA, UBS increased its 2025/2026 CoWoS demand estimates by 5% and 26%, respectively, based on three key factors:\n    \n1. Strong Blackwell shipment growth, expected to rise by about 30% QoQ in Q3 (previously 17%) and remain at 1.6–1.7 million units in Q4.\n    \n2. Clearer mass production trajectory for Rubin, starting in mid-2026.\n    \n3. The newly introduced CPX variant within the Rubin family, which will bring additional CoWoS packaging demand.\n\nAs a result, UBS raised NVIDIA’s total GPU production at TSMC to 6.9 million units in 2025 and 7.4 million units in 2026 (previously 6.5 million and 6.7 million).\n\nFor Broadcom, UBS also raised its 2026 CoWoS demand forecast for AI accelerators, reflecting stronger growth momentum from Google’s TPU v7p and OpenAI projects.\n\nUBS noted that NVIDIA’s Rubin project on TSMC’s N3 process is progressing smoothly, with small-scale shipments to begin in Q2 2026 and mass production starting in Q3. Rubin’s trial production will be completed this month, and samples are expected to be distributed to supply-chain partners this quarter in preparation for mass production in the second half of the year. UBS lifted its 2026 production estimate for Rubin at TSMC from 1.3 million to 2.3 million units.\n\nIn addition, NVIDIA’s newly launched Rubin CPX version, which uses a single-GPU package with GDDR7 memory, is also believed to employ TSMC’s CoWoS packaging. The key advantage of this design lies in the use of embedded deep trench capacitors (eDTC), which enhance voltage stability by embedding capacitors into the interposer—thereby preventing GPU performance degradation. UBS estimates that with Rubin ramping up and CPX contributing, NVIDIA’s CoWoS demand will rise from 444,000 wafers in 2025 to 678,000 wafers in 2026.\n\nOn the investment side, UBS continues to favor TSMC as the leading cloud and edge AI foundry, benefiting from its moat in advanced process and packaging technologies. The report also highlights ASE (Advanced Semiconductor Engineering) as a key beneficiary in advanced packaging and testing. As cloud AI demand continues to expand, UBS expects significant opportunities for the advanced packaging and CoWoS equipment supply chain.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05137584134182771, "ret_signed_m1d": 0.05137584134182771, "alpha_spy_m1d": 0.023597217580441754, "alpha_c_m1d": -0.009698681846317392, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02817212030804872, "ret_signed_p1d": 0.02817212030804872, "alpha_spy_p1d": 0.012828045614839967, "alpha_c_p1d": -0.016119988067143876, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.0003276050769120964, "ret_signed_p1w": 0.0003276050769120964, "alpha_spy_p1w": -0.017083865558432887, "alpha_c_p1w": -0.05343148722476987, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.0867547734359273, "ret_signed_p1m": 0.0867547734359273, "alpha_spy_p1m": 0.04323380469821747, "alpha_c_p1m": -0.016060782030970833, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.03254271530285635, "ret_signed_p3m": 0.03254271530285635, "alpha_spy_p3m": -0.026563715450941272, "alpha_c_p3m": -0.15457850936422024, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": -0.030031104152419386, "ret_signed_p6m": -0.030031104152419386, "alpha_spy_p6m": -0.04482688746932173, "alpha_c_p6m": -0.25094505407484824, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.0681, -0.0644, -0.0555, -0.0587, -0.0644, -0.0881, -0.0676, -0.0515, -0.0528, -0.0351, -0.0418, -0.0213, -0.0289, -0.0516, -0.0173, -0.0268, -0.0205, -0.0131, -0.0026, -0.0061, -0.0001, -0.0086, -0.0063, -0.0149, -0.0063, -0.0411, -0.0424, -0.0447, -0.0283, -0.0183, -0.0076, -0.0086, -0.0164, -0.0491, -0.0491, -0.0676, -0.0685, -0.0628, -0.0882, -0.0811, -0.0678, -0.0319, -0.0327, -0.0292, -0.0295, -0.0452, -0.0703, -0.0378, -0.0354, 0.0025, -0.0258, -0.0338, -0.0299, -0.0271, -0.0072, 0.0187, 0.0223, 0.0313, 0.0244, 0.013, 0.0103, 0.0325, 0.0514, 0.0, 0.0282, -0.0171, -0.0182, -0.0074, 0.0003, -0.0028, -0.0109, -0.0157, -0.0055, 0.0169, 0.0455, 0.0976, 0.1304, 0.1077, 0.1055, 0.1295, 0.0848, 0.0658, 0.0269, 0.0272, 0.0868, 0.0546, 0.0581, 0.0202, 0.0383, 0.0188, -0.0098, 0.0183, -0.0138, -0.0234, -0.0033, -0.0292, -0.0158, -0.0158, -0.0336, -0.0177, -0.0093, -0.0195, 0.0013, -0.004, 0.0131, 0.0099, 0.0034, -0.0121, -0.0444, -0.0375, -0.0296, -0.0667, -0.0492, -0.0118, 0.0029, 0.0331, 0.0298, 0.0298, 0.0403, 0.0277, 0.024, 0.0183, 0.0183, 0.0311, 0.0271, 0.0223, 0.0325, 0.0103, 0.0093, 0.0098, 0.0145, -0.0001, 0.0213, 0.0168, 0.0168, -0.0277, 0.0009, 0.0092, 0.0247, 0.0181, 0.0293, 0.0457, 0.0511, 0.0436, 0.0134, -0.0153, -0.0489, -0.0615, 0.0123, 0.0376, 0.0294, 0.0377, 0.0207, -0.0019, -0.0019, 0.0099, 0.0264, 0.0259, 0.0364, 0.0459, 0.053, 0.0678, 0.0095, -0.0325, -0.0037, -0.0169, -0.0006, 0.001, -0.0291, -0.0027, 0.0088, 0.0158, 0.0, -0.0158, 0.0004, -0.0066, -0.015, -0.025, -0.057, -0.041, -0.0434, -0.0244, -0.065, -0.0853, -0.0981, -0.0477, -0.0404, -0.0314, -0.0314, -0.03]}
{"unit_id": "orig:1976535639446389242", "ticker_idx": 1, "ts": "2025-10-10T06:29:53+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS also raised its 2026 CoWoS demand forecast for AI accelerators, reflecting stronger growth momentum from Google's TPU v7p and OpenAI projects", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for NVDA and AVGO; favors TSMC and ASE as key beneficiaries of AI packaging buildout.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS Raises CoWoS Demand Forecasts for NVIDIA and Broadcom, With Rubin and CPX as Key Growth Drivers\n\nUBS (T. Arcuri, Oct 8, 2025)\n\nUBS has raised its CoWoS demand forecasts for NVIDIA and Broadcom in its latest report. The firm had previously revised TSMC’s CoWoS capacity forecast for the end of 2026 from 100,000 wafers per month to 110,000 wafers per month.\n\nFor NVIDIA, UBS increased its 2025/2026 CoWoS demand estimates by 5% and 26%, respectively, based on three key factors:\n    \n1. Strong Blackwell shipment growth, expected to rise by about 30% QoQ in Q3 (previously 17%) and remain at 1.6–1.7 million units in Q4.\n    \n2. Clearer mass production trajectory for Rubin, starting in mid-2026.\n    \n3. The newly introduced CPX variant within the Rubin family, which will bring additional CoWoS packaging demand.\n\nAs a result, UBS raised NVIDIA’s total GPU production at TSMC to 6.9 million units in 2025 and 7.4 million units in 2026 (previously 6.5 million and 6.7 million).\n\nFor Broadcom, UBS also raised its 2026 CoWoS demand forecast for AI accelerators, reflecting stronger growth momentum from Google’s TPU v7p and OpenAI projects.\n\nUBS noted that NVIDIA’s Rubin project on TSMC’s N3 process is progressing smoothly, with small-scale shipments to begin in Q2 2026 and mass production starting in Q3. Rubin’s trial production will be completed this month, and samples are expected to be distributed to supply-chain partners this quarter in preparation for mass production in the second half of the year. UBS lifted its 2026 production estimate for Rubin at TSMC from 1.3 million to 2.3 million units.\n\nIn addition, NVIDIA’s newly launched Rubin CPX version, which uses a single-GPU package with GDDR7 memory, is also believed to employ TSMC’s CoWoS packaging. The key advantage of this design lies in the use of embedded deep trench capacitors (eDTC), which enhance voltage stability by embedding capacitors into the interposer—thereby preventing GPU performance degradation. UBS estimates that with Rubin ramping up and CPX contributing, NVIDIA’s CoWoS demand will rise from 444,000 wafers in 2025 to 678,000 wafers in 2026.\n\nOn the investment side, UBS continues to favor TSMC as the leading cloud and edge AI foundry, benefiting from its moat in advanced process and packaging technologies. The report also highlights ASE (Advanced Semiconductor Engineering) as a key beneficiary in advanced packaging and testing. As cloud AI demand continues to expand, UBS expects significant opportunities for the advanced packaging and CoWoS equipment supply chain.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06280991220711485, "ret_signed_m1d": 0.06280991220711485, "alpha_spy_m1d": 0.03503128844572889, "alpha_c_m1d": 0.0017353890189697463, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09878936503250224, "ret_signed_p1d": 0.09878936503250224, "alpha_spy_p1d": 0.08344529033929349, "alpha_c_p1d": 0.054497256657309645, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.07608663233076629, "ret_signed_p1w": 0.07608663233076629, "alpha_spy_p1w": 0.05867516169542131, "alpha_c_p1w": 0.022327540029084325, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.10399543706540548, "ret_signed_p1m": 0.10399543706540548, "alpha_spy_p1m": 0.06047446832769565, "alpha_c_p1m": 0.0011798815985073485, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.06015232182620078, "ret_signed_p3m": 0.06015232182620078, "alpha_spy_p3m": 0.00104589107240316, "alpha_c_p3m": -0.1269689028408758, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": -0.02753145464965523, "ret_signed_p6m": -0.02753145464965523, "alpha_spy_p6m": -0.042327237966557574, "alpha_c_p6m": -0.24844540457208408, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.1361, -0.1365, -0.1191, -0.1287, -0.1137, -0.1433, -0.1276, -0.1122, -0.1076, -0.095, -0.0854, -0.0694, -0.0968, -0.1124, -0.0845, -0.0992, -0.0723, -0.0659, -0.0622, -0.0655, -0.038, -0.0495, -0.0429, -0.058, -0.0597, -0.0931, -0.1046, -0.1094, -0.0959, -0.0952, -0.0836, -0.0767, -0.0509, -0.0855, -0.0855, -0.0829, -0.0701, -0.0587, 0.0298, 0.0629, 0.0353, 0.1365, 0.1059, 0.1067, 0.1196, 0.1071, 0.0645, 0.062, 0.0607, 0.0436, 0.0441, 0.0452, 0.0353, 0.0305, 0.0101, 0.0163, 0.027, 0.0417, 0.0423, 0.0335, 0.0363, 0.0643, 0.0628, 0.0, 0.0988, 0.0601, 0.0822, 0.0909, 0.0761, 0.0758, 0.0555, 0.0483, 0.0606, 0.0909, 0.1153, 0.1489, 0.189, 0.1597, 0.1386, 0.1168, 0.0841, 0.1058, 0.0954, 0.0764, 0.104, 0.0842, 0.0942, 0.0473, 0.0549, 0.0555, 0.0489, 0.0918, 0.0684, 0.048, 0.1643, 0.1861, 0.2247, 0.2247, 0.2413, 0.1893, 0.1754, 0.1724, 0.1737, 0.2021, 0.2356, 0.2515, 0.2721, 0.2518, 0.1087, 0.0468, 0.0514, 0.0043, 0.0162, 0.0485, 0.0538, 0.0781, 0.0809, 0.0809, 0.0868, 0.0783, 0.0798, 0.0682, 0.0682, 0.0729, 0.0599, 0.061, 0.0602, 0.0261, 0.0647, 0.087, 0.0944, 0.049, 0.0587, 0.0855, 0.0855, 0.0265, 0.0148, 0.0046, -0.0122, 0.0026, 0.0271, 0.0285, 0.0207, 0.0225, 0.0219, -0.0114, -0.0493, -0.0417, 0.0275, 0.0615, 0.0507, 0.0579, 0.0221, 0.0036, 0.0036, 0.0263, 0.0293, 0.0308, 0.0267, 0.0195, 0.0046, 0.0256, -0.0071, -0.0138, -0.016, -0.0314, -0.02, 0.027, 0.02, 0.0671, 0.0573, 0.0542, 0.0369, -0.0057, 0.0028, -0.0083, -0.0249, -0.0129, -0.0417, -0.0025, -0.0156, -0.014, -0.043, -0.0701, -0.0925, -0.0427, -0.0304, -0.0272, -0.0272, -0.0275]}
{"unit_id": "orig:1976535639446389242", "ticker_idx": 2, "ts": "2025-10-10T06:29:53+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "UBS continues to favor TSMC as the leading cloud and edge AI foundry, benefiting from its moat in advanced process and packaging technologies", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for NVDA and AVGO; favors TSMC and ASE as key beneficiaries of AI packaging buildout.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS Raises CoWoS Demand Forecasts for NVIDIA and Broadcom, With Rubin and CPX as Key Growth Drivers\n\nUBS (T. Arcuri, Oct 8, 2025)\n\nUBS has raised its CoWoS demand forecasts for NVIDIA and Broadcom in its latest report. The firm had previously revised TSMC’s CoWoS capacity forecast for the end of 2026 from 100,000 wafers per month to 110,000 wafers per month.\n\nFor NVIDIA, UBS increased its 2025/2026 CoWoS demand estimates by 5% and 26%, respectively, based on three key factors:\n    \n1. Strong Blackwell shipment growth, expected to rise by about 30% QoQ in Q3 (previously 17%) and remain at 1.6–1.7 million units in Q4.\n    \n2. Clearer mass production trajectory for Rubin, starting in mid-2026.\n    \n3. The newly introduced CPX variant within the Rubin family, which will bring additional CoWoS packaging demand.\n\nAs a result, UBS raised NVIDIA’s total GPU production at TSMC to 6.9 million units in 2025 and 7.4 million units in 2026 (previously 6.5 million and 6.7 million).\n\nFor Broadcom, UBS also raised its 2026 CoWoS demand forecast for AI accelerators, reflecting stronger growth momentum from Google’s TPU v7p and OpenAI projects.\n\nUBS noted that NVIDIA’s Rubin project on TSMC’s N3 process is progressing smoothly, with small-scale shipments to begin in Q2 2026 and mass production starting in Q3. Rubin’s trial production will be completed this month, and samples are expected to be distributed to supply-chain partners this quarter in preparation for mass production in the second half of the year. UBS lifted its 2026 production estimate for Rubin at TSMC from 1.3 million to 2.3 million units.\n\nIn addition, NVIDIA’s newly launched Rubin CPX version, which uses a single-GPU package with GDDR7 memory, is also believed to employ TSMC’s CoWoS packaging. The key advantage of this design lies in the use of embedded deep trench capacitors (eDTC), which enhance voltage stability by embedding capacitors into the interposer—thereby preventing GPU performance degradation. UBS estimates that with Rubin ramping up and CPX contributing, NVIDIA’s CoWoS demand will rise from 444,000 wafers in 2025 to 678,000 wafers in 2026.\n\nOn the investment side, UBS continues to favor TSMC as the leading cloud and edge AI foundry, benefiting from its moat in advanced process and packaging technologies. The report also highlights ASE (Advanced Semiconductor Engineering) as a key beneficiary in advanced packaging and testing. As cloud AI demand continues to expand, UBS expects significant opportunities for the advanced packaging and CoWoS equipment supply chain.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.027778623761385957, "alpha_c_m1d": -0.0610745231881451, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017361126814406158, "ret_signed_p1d": -0.017361126814406158, "alpha_spy_p1d": -0.03270520150761491, "alpha_c_p1d": -0.061653235189598754, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.006944331262205061, "ret_signed_p1w": 0.006944331262205061, "alpha_spy_p1w": -0.010467139373139922, "alpha_c_p1w": -0.046814761039476904, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.02430545807661111, "ret_signed_p1m": 0.02430545807661111, "alpha_spy_p1m": -0.019215510661098723, "alpha_c_p1m": -0.07851009739028703, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.16707166234820137, "ret_signed_p3m": 0.16707166234820137, "alpha_spy_p3m": 0.10796523159440374, "alpha_c_p3m": -0.020049562318875225, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": 0.2652487764585114, "ret_signed_p6m": 0.2652487764585114, "alpha_spy_p6m": 0.25045299314160907, "alpha_c_p6m": 0.04433482653608256, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.2322, -0.2184, -0.2184, -0.2011, -0.2046, -0.2184, -0.208, -0.208, -0.208, -0.208, -0.2149, -0.2011, -0.1977, -0.1977, -0.2149, -0.2046, -0.2219, -0.1838, -0.1873, -0.1838, -0.1838, -0.17, -0.1873, -0.1838, -0.1838, -0.1804, -0.2149, -0.2046, -0.2149, -0.1907, -0.1873, -0.1769, -0.1977, -0.1977, -0.1942, -0.1977, -0.1977, -0.1977, -0.1838, -0.1838, -0.17, -0.1527, -0.1423, -0.1285, -0.1319, -0.1111, -0.1215, -0.1076, -0.1215, -0.1007, -0.0694, -0.0694, -0.0833, -0.0972, -0.0972, -0.0938, -0.0799, -0.0521, -0.0278, -0.0278, -0.0035, -0.0174, 0.0, 0.0, -0.0174, -0.0104, 0.0174, 0.0312, 0.0069, 0.0278, 0.0278, 0.0139, 0.0069, 0.0069, 0.0278, 0.0243, 0.0451, 0.0451, 0.0417, 0.0486, 0.0451, 0.0139, 0.0174, 0.0139, 0.0243, 0.0174, 0.0243, 0.0139, -0.0069, 0.0035, -0.0243, -0.0313, 0.0104, -0.0382, -0.0451, -0.0174, 0.0, -0.0035, 0.0, -0.0208, -0.0069, 0.0069, 0.0035, 0.0139, 0.0382, 0.0278, 0.0451, 0.0242, 0.0312, 0.0103, -0.0002, -0.0036, -0.0036, -0.0036, 0.0208, 0.0382, 0.0417, 0.0417, 0.0521, 0.066, 0.0591, 0.08, 0.08, 0.1044, 0.1636, 0.188, 0.1671, 0.174, 0.1706, 0.1775, 0.1915, 0.1915, 0.1775, 0.2124, 0.2263, 0.2367, 0.2124, 0.2263, 0.2333, 0.2228, 0.2402, 0.2681, 0.2577, 0.2367, 0.2298, 0.2542, 0.2437, 0.2298, 0.2402, 0.2646, 0.3099, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3238, 0.3691, 0.404, 0.39, 0.39, 0.3761, 0.3482, 0.2995, 0.3238, 0.3169, 0.2611, 0.289, 0.3517, 0.3134, 0.2995, 0.2855, 0.3072, 0.3317, 0.2932, 0.2862, 0.2652, 0.2652, 0.2897, 0.2862, 0.2722, 0.2443, 0.2303, 0.2967, 0.2652, 0.2652, 0.2652]}
{"unit_id": "orig:1976535639446389242", "ticker_idx": 3, "ts": "2025-10-10T06:29:53+00:00", "symbol": "ASE Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASE (Advanced Semiconductor Engineering) as a key beneficiary in advanced packaging and testing", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises CoWoS demand forecasts for NVDA and AVGO; favors TSMC and ASE as key beneficiaries of AI packaging buildout.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS Raises CoWoS Demand Forecasts for NVIDIA and Broadcom, With Rubin and CPX as Key Growth Drivers\n\nUBS (T. Arcuri, Oct 8, 2025)\n\nUBS has raised its CoWoS demand forecasts for NVIDIA and Broadcom in its latest report. The firm had previously revised TSMC’s CoWoS capacity forecast for the end of 2026 from 100,000 wafers per month to 110,000 wafers per month.\n\nFor NVIDIA, UBS increased its 2025/2026 CoWoS demand estimates by 5% and 26%, respectively, based on three key factors:\n    \n1. Strong Blackwell shipment growth, expected to rise by about 30% QoQ in Q3 (previously 17%) and remain at 1.6–1.7 million units in Q4.\n    \n2. Clearer mass production trajectory for Rubin, starting in mid-2026.\n    \n3. The newly introduced CPX variant within the Rubin family, which will bring additional CoWoS packaging demand.\n\nAs a result, UBS raised NVIDIA’s total GPU production at TSMC to 6.9 million units in 2025 and 7.4 million units in 2026 (previously 6.5 million and 6.7 million).\n\nFor Broadcom, UBS also raised its 2026 CoWoS demand forecast for AI accelerators, reflecting stronger growth momentum from Google’s TPU v7p and OpenAI projects.\n\nUBS noted that NVIDIA’s Rubin project on TSMC’s N3 process is progressing smoothly, with small-scale shipments to begin in Q2 2026 and mass production starting in Q3. Rubin’s trial production will be completed this month, and samples are expected to be distributed to supply-chain partners this quarter in preparation for mass production in the second half of the year. UBS lifted its 2026 production estimate for Rubin at TSMC from 1.3 million to 2.3 million units.\n\nIn addition, NVIDIA’s newly launched Rubin CPX version, which uses a single-GPU package with GDDR7 memory, is also believed to employ TSMC’s CoWoS packaging. The key advantage of this design lies in the use of embedded deep trench capacitors (eDTC), which enhance voltage stability by embedding capacitors into the interposer—thereby preventing GPU performance degradation. UBS estimates that with Rubin ramping up and CPX contributing, NVIDIA’s CoWoS demand will rise from 444,000 wafers in 2025 to 678,000 wafers in 2026.\n\nOn the investment side, UBS continues to favor TSMC as the leading cloud and edge AI foundry, benefiting from its moat in advanced process and packaging technologies. The report also highlights ASE (Advanced Semiconductor Engineering) as a key beneficiary in advanced packaging and testing. As cloud AI demand continues to expand, UBS expects significant opportunities for the advanced packaging and CoWoS equipment supply chain.\n\n$NVDA $AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05685921649629355, "ret_signed_m1d": 0.05685921649629355, "alpha_spy_m1d": 0.029080592734907595, "alpha_c_m1d": -0.00421530669185155, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05866422590306852, "ret_signed_p1d": 0.05866422590306852, "alpha_spy_p1d": 0.04332015120985977, "alpha_c_p1d": 0.014372117527875927, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.14801447243820953, "ret_signed_p1w": 0.14801447243820953, "alpha_spy_p1w": 0.13060300180286455, "alpha_c_p1w": 0.09425538013652757, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.38808666248587786, "ret_signed_p1m": 0.38808666248587786, "alpha_spy_p1m": 0.34456569374816803, "alpha_c_p1m": 0.28527110701897973, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.5902527942743734, "ret_signed_p3m": 0.5902527942743734, "alpha_spy_p3m": 0.5311463635205758, "alpha_c_p3m": 0.4031315696072968, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": 1.035198501113165, "ret_signed_p6m": 1.035198501113165, "alpha_spy_p6m": 1.0204027177962627, "alpha_c_p6m": 0.8142845511907362, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.0722, -0.0596, -0.0451, -0.0505, -0.0542, -0.0713, -0.0451, -0.046, -0.0523, -0.046, -0.0542, -0.0677, -0.1426, -0.0966, -0.0993, -0.1218, -0.1227, -0.1056, -0.1002, -0.0966, -0.0857, -0.0948, -0.1029, -0.1128, -0.0884, -0.1128, -0.1255, -0.1309, -0.1101, -0.1083, -0.0939, -0.0957, -0.1002, -0.1056, -0.1056, -0.0722, -0.0677, -0.0695, -0.0551, 0.0388, 0.0099, 0.0199, 0.0027, -0.0027, 0.0108, 0.0199, 0.0199, 0.0352, 0.0343, 0.0569, 0.056, 0.0415, 0.0343, 0.0081, 0.0045, 0.0009, 0.0045, 0.0063, 0.0, 0.028, 0.0271, 0.0704, 0.0569, 0.0, 0.0587, 0.0217, 0.0767, 0.1354, 0.148, 0.1724, 0.1444, 0.1245, 0.157, 0.1715, 0.194, 0.2058, 0.3023, 0.3574, 0.4449, 0.4612, 0.3872, 0.4287, 0.3872, 0.3736, 0.3881, 0.352, 0.3529, 0.3204, 0.3186, 0.2906, 0.2789, 0.2699, 0.241, 0.2473, 0.278, 0.2798, 0.2996, 0.2996, 0.3493, 0.3384, 0.3782, 0.3899, 0.3682, 0.3727, 0.4206, 0.4413, 0.4829, 0.4801, 0.4224, 0.4061, 0.3809, 0.3312, 0.3421, 0.3682, 0.3854, 0.4016, 0.4034, 0.4034, 0.4188, 0.4458, 0.4468, 0.4531, 0.4531, 0.5217, 0.5307, 0.5794, 0.5903, 0.5641, 0.5921, 0.6489, 0.6877, 0.704, 0.7319, 0.7518, 0.7518, 0.7058, 0.713, 0.7202, 0.75, 0.7897, 0.8114, 0.8285, 0.7852, 0.713, 0.7518, 0.75, 0.7139, 0.8285, 0.8854, 0.9991, 1.0298, 1.1426, 1.1173, 1.1119, 1.1119, 1.1182, 1.1065, 1.0921, 1.1625, 1.1191, 1.2401, 1.2662, 1.1868, 1.1922, 1.1841, 1.0135, 0.9946, 0.991, 0.9061, 0.9549, 0.9594, 0.9765, 0.8944, 0.9404, 0.9422, 0.9702, 0.9549, 0.9765, 0.9233, 0.9567, 0.9431, 1.0217, 0.9341, 0.9404, 0.8854, 0.9567, 1.0289, 1.0108, 1.0108, 1.0352]}
{"unit_id": "thread:1994187350004649994", "ticker_idx": 0, "ts": "2025-11-27T23:31:29+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "contract prices remain far lower than spot prices. It is expected that there could still be room for a 50% price hike in the next 6 to 9 months", "tweet_type": "analysis", "timeline": "months", "summary": "Global memory shortage is structural with 50% contract price upside in 6-9 months; SK Hynix best positioned via HBM/DRAM focus; Samsung and Micron supply-constrained near-term.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "(Exclusive) 2026 Capacity Sold Out! CSPs Trigger Early \"Two-Year LTA\" War for Memory\n\nThe explosive growth of AI demand is accelerating the global memory market into an unprecedented shortage frenzy. Industry sources point out that with major memory manufacturers' 2026 production capacity almost fully booked, a full-year shortage is irreversible. Multiple hyperscale Cloud Service Providers (CSPs) are proposing Long Term Agreement (LTA) negotiations for one year or longer, extending the battlefront to 2027-2028 to lock in future capacity requirements for the next two years ahead of schedule.\n\nThe memory shortage presents an unprecedented level of \"madness.\" With AI being the primary driver making servers the largest application outlet, and CSP giants wielding massive procurement funds, they fell into a panic buying spree in the second half of 2025. Consequently, companies are inclined to build higher inventories for 2026-2027, resulting in purchasing volumes far exceeding actual consumption.\n\nIndustry insiders revealed that memory manufacturers' expansion plans for 2026 are too slow to meet immediate needs, resulting in the entire year's capacity supply being nearly fully booked. This has prompted major CSPs to seek discussions with memory players early regarding 2027 supply capacity. Major CSPs from the US and China are accelerating their pace, conducting \"bundled negotiations\" that combine 2026 procurement volumes and pricing with subsequent purchasing targets.\n\nThis implies that the bargaining power of memory sellers has significantly increased, ensuring a continuous price hike in 2026. There will be no chance for price drops even in the second half of the year, reshaping the pricing model of the memory industry.\n\nTo avoid falling behind in the AI arms race, server operators are willing to offer higher premiums to secure capacity. As the main source of profit for original manufacturers, they are bound to become the priority customer group for supply. However, to qualify for long-term agreements, some CSP giants may propose more favorable conditions to persuade manufacturers—such as advance payments or funding equipment procurement—thereby locking in capacity for 2027-2028 ahead of time. It is reported that supply LTAs for 2027 could be finalized as early as the first quarter of 2026.\n\nIt is not just DRAM that is in severe shortage. Phison CEO K.S. Pua recently pointed out that NAND Flash prices are skyrocketing rapidly; there are even cases where suppliers have proposed price hikes but ultimately still have no stock to sell. He believes that the situation of \"clients holding cash and asking manufacturers to start production\" will happen very soon.\n\nServer OEM operators also privately noted that while AI GPU supply or production yields are currently smooth with stable output, memory shortages have become the shipment bottleneck for the entire supply chain. There are even rumors in the industry of buyers frantically searching for stock with cash in hand. This will cause the memory industry to shift from the past business cycle model to a structural transformation of \"securing orders first, then expanding production.\"\n\nADATA Technology Chairman Chen Li-bai also confirmed: \"Currently, every major CSP is taking the lead in wanting to sign contracts exceeding one year.\"\n\nHowever, only one or two major CSPs are likely to sign relatively long supply contracts. Other CSP operators will at most secure one-year contracts, while buyers outside of CSPs basically cannot sign any contracts.\n\nIt is understood that the few major CSPs currently signing long-term agreements mostly depend on faster decision-making in the second half of 2025. They realized early on that memory supply could not meet demand and responded quickly by requesting to sign contracts for one year or longer, yet only about 30% of CSP operators managed to secure LTA qualifications.\n\nAfter all, memory suppliers cannot overcommit capacity. Chen Li-bai admitted that an LTA is just a dream; 90% of manufacturers cannot sign long-term agreements. Some operators with better relationships might negotiate once a quarter, but the majority can only negotiate monthly, or cannot even confirm supply status for one month later.\n\nFrom the perspective of memory manufacturers, recent spot prices are already at an \"abnormally high\" level. While a correction in the future cannot be ruled out, contract prices remain far lower than spot prices. It is expected that there could still be room for a 50% price hike in the next 6 to 9 months.\n\nAccording to current planning by memory manufacturers, SK Hynix has no plans to expand NAND wafer capacity in 2026, concentrating firepower on capital investment for High Bandwidth Memory (HBM) and DRAM. This includes the new M15X plant in Cheongju, which is planned as a dedicated facility for HBM production, with production expected to start as early as the beginning of 2026.\n\nHowever, Samsung Electronics and Micron are unlikely to have significant capacity contributions in the short term. New plant construction may not be completed until the second half of 2027. Coupled with equipment installation and the wafer production cycle, there will not only be a lack of new capacity in 2026, but output contribution in 2027 will also be relatively limited.\n\n---\n\nhttps://t.co/hIvtpNrG36", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014739094448437203, "ret_signed_m1d": -0.014739094448437203, "alpha_spy_m1d": -0.014739094448437203, "alpha_c_m1d": -0.014739094448437203, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009235933266170004, "ret_signed_p1d": -0.009235933266170004, "alpha_spy_p1d": -0.014694339604784945, "alpha_c_p1d": -0.02237913683554038, "bench_spy_p1d": 0.005458406338614941, "bench_c_p1d": 0.013143203569370376, "ret_p1w": -0.0012984644520754503, "ret_signed_p1w": -0.0012984644520754503, "alpha_spy_p1w": -0.008228129851947505, "alpha_c_p1w": -0.04124557680522175, "bench_spy_p1w": 0.006929665399872054, "bench_c_p1w": 0.0399471123531463, "ret_p1m": 0.15611907630982916, "ret_signed_p1m": 0.15611907630982916, "alpha_spy_p1m": 0.13747831160552462, "alpha_c_p1m": 0.1006463863630426, "bench_spy_p1m": 0.018640764704304535, "bench_c_p1m": 0.05547268994678656, "ret_p3m": 0.868434420109299, "ret_signed_p3m": 0.868434420109299, "alpha_spy_p3m": 0.8541615545209296, "alpha_c_p3m": 0.6591960384966484, "bench_spy_p3m": 0.014272865588369488, "bench_c_p3m": 0.20923838161265063, "ret_p6m": 2.2281734626617755, "ret_signed_p6m": 2.2281734626617755, "alpha_spy_p6m": 2.124881250707631, "alpha_c_p6m": 1.5655428040612556, "bench_spy_p6m": 0.10329221195414462, "bench_c_p6m": 0.66263065860052, "price_path": [-0.4543, -0.4475, -0.4437, -0.433, -0.4196, -0.4154, -0.3987, -0.3785, -0.3588, -0.3296, -0.3238, -0.3026, -0.3136, -0.2805, -0.2894, -0.2792, -0.2666, -0.2733, -0.2788, -0.2993, -0.278, -0.2749, -0.2391, -0.2029, -0.1969, -0.1923, -0.2, -0.1843, -0.1904, -0.171, -0.1663, -0.1822, -0.1575, -0.1151, -0.1067, -0.0875, -0.0999, -0.1003, -0.097, -0.0525, -0.0253, -0.0393, -0.0066, 0.0073, 0.0124, 0.0772, 0.0123, 0.0223, 0.0276, 0.0149, 0.0617, 0.0614, 0.0644, 0.0488, 0.0133, 0.0453, -0.0052, -0.018, -0.0349, -0.0755, -0.046, -0.0374, -0.0147, 0.0, -0.0092, 0.0024, 0.0216, 0.0138, -0.0013, 0.0259, 0.0637, 0.0613, 0.0893, 0.0659, 0.0497, 0.0208, -0.0076, 0.0116, 0.0446, 0.0625, 0.1117, 0.1169, 0.1331, 0.1331, 0.1561, 0.2052, 0.2107, 0.2003, 0.2003, 0.2875, 0.3255, 0.3917, 0.4027, 0.386, 0.4055, 0.409, 0.3872, 0.3915, 0.4122, 0.4705, 0.4767, 0.4538, 0.5008, 0.5313, 0.541, 0.5067, 0.6005, 0.6713, 0.6788, 0.6772, 0.6293, 0.7053, 0.6482, 0.586, 0.599, 0.6374, 0.6139, 0.6643, 0.7216, 0.7218, 0.7218, 0.7046, 0.7353, 0.7671, 0.8168, 0.817, 0.8684, 0.9036, 0.982, 0.9492, 0.9496, 0.7577, 0.6589, 0.7727, 0.7126, 0.6385, 0.7677, 0.8075, 0.766, 0.7696, 0.8484, 0.8916, 0.9917, 0.9141, 0.8759, 0.7613, 0.792, 0.7758, 0.6704, 0.6729, 0.5739, 0.5271, 0.6986, 0.6187, 0.6722, 0.7174, 0.7466, 0.9061, 0.8849, 0.9079, 0.9083, 1.0214, 1.0426, 1.0768, 1.0523, 1.0611, 1.1127, 1.1624, 1.178, 1.1817, 1.2812, 1.2486, 1.2778, 1.2538, 1.2901, 1.4774, 1.5698, 1.8112, 1.8327, 1.9877, 2.2323, 2.1419, 2.2983, 2.2936, 2.0438, 2.0282, 1.977, 2.0268, 2.2663, 2.2282]}
{"unit_id": "thread:1994187350004649994", "ticker_idx": 1, "ts": "2025-11-27T23:31:29+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND Flash prices are skyrocketing rapidly; there are even cases where suppliers have proposed price hikes but ultimately still have no stock to sell", "tweet_type": "analysis", "timeline": "months", "summary": "Global memory shortage is structural with 50% contract price upside in 6-9 months; SK Hynix best positioned via HBM/DRAM focus; Samsung and Micron supply-constrained near-term.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "Global NAND memory basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "(Exclusive) 2026 Capacity Sold Out! CSPs Trigger Early \"Two-Year LTA\" War for Memory\n\nThe explosive growth of AI demand is accelerating the global memory market into an unprecedented shortage frenzy. Industry sources point out that with major memory manufacturers' 2026 production capacity almost fully booked, a full-year shortage is irreversible. Multiple hyperscale Cloud Service Providers (CSPs) are proposing Long Term Agreement (LTA) negotiations for one year or longer, extending the battlefront to 2027-2028 to lock in future capacity requirements for the next two years ahead of schedule.\n\nThe memory shortage presents an unprecedented level of \"madness.\" With AI being the primary driver making servers the largest application outlet, and CSP giants wielding massive procurement funds, they fell into a panic buying spree in the second half of 2025. Consequently, companies are inclined to build higher inventories for 2026-2027, resulting in purchasing volumes far exceeding actual consumption.\n\nIndustry insiders revealed that memory manufacturers' expansion plans for 2026 are too slow to meet immediate needs, resulting in the entire year's capacity supply being nearly fully booked. This has prompted major CSPs to seek discussions with memory players early regarding 2027 supply capacity. Major CSPs from the US and China are accelerating their pace, conducting \"bundled negotiations\" that combine 2026 procurement volumes and pricing with subsequent purchasing targets.\n\nThis implies that the bargaining power of memory sellers has significantly increased, ensuring a continuous price hike in 2026. There will be no chance for price drops even in the second half of the year, reshaping the pricing model of the memory industry.\n\nTo avoid falling behind in the AI arms race, server operators are willing to offer higher premiums to secure capacity. As the main source of profit for original manufacturers, they are bound to become the priority customer group for supply. However, to qualify for long-term agreements, some CSP giants may propose more favorable conditions to persuade manufacturers—such as advance payments or funding equipment procurement—thereby locking in capacity for 2027-2028 ahead of time. It is reported that supply LTAs for 2027 could be finalized as early as the first quarter of 2026.\n\nIt is not just DRAM that is in severe shortage. Phison CEO K.S. Pua recently pointed out that NAND Flash prices are skyrocketing rapidly; there are even cases where suppliers have proposed price hikes but ultimately still have no stock to sell. He believes that the situation of \"clients holding cash and asking manufacturers to start production\" will happen very soon.\n\nServer OEM operators also privately noted that while AI GPU supply or production yields are currently smooth with stable output, memory shortages have become the shipment bottleneck for the entire supply chain. There are even rumors in the industry of buyers frantically searching for stock with cash in hand. This will cause the memory industry to shift from the past business cycle model to a structural transformation of \"securing orders first, then expanding production.\"\n\nADATA Technology Chairman Chen Li-bai also confirmed: \"Currently, every major CSP is taking the lead in wanting to sign contracts exceeding one year.\"\n\nHowever, only one or two major CSPs are likely to sign relatively long supply contracts. Other CSP operators will at most secure one-year contracts, while buyers outside of CSPs basically cannot sign any contracts.\n\nIt is understood that the few major CSPs currently signing long-term agreements mostly depend on faster decision-making in the second half of 2025. They realized early on that memory supply could not meet demand and responded quickly by requesting to sign contracts for one year or longer, yet only about 30% of CSP operators managed to secure LTA qualifications.\n\nAfter all, memory suppliers cannot overcommit capacity. Chen Li-bai admitted that an LTA is just a dream; 90% of manufacturers cannot sign long-term agreements. Some operators with better relationships might negotiate once a quarter, but the majority can only negotiate monthly, or cannot even confirm supply status for one month later.\n\nFrom the perspective of memory manufacturers, recent spot prices are already at an \"abnormally high\" level. While a correction in the future cannot be ruled out, contract prices remain far lower than spot prices. It is expected that there could still be room for a 50% price hike in the next 6 to 9 months.\n\nAccording to current planning by memory manufacturers, SK Hynix has no plans to expand NAND wafer capacity in 2026, concentrating firepower on capital investment for High Bandwidth Memory (HBM) and DRAM. This includes the new M15X plant in Cheongju, which is planned as a dedicated facility for HBM production, with production expected to start as early as the beginning of 2026.\n\nHowever, Samsung Electronics and Micron are unlikely to have significant capacity contributions in the short term. New plant construction may not be completed until the second half of 2027. Coupled with equipment installation and the wafer production cycle, there will not only be a lack of new capacity in 2026, but output contribution in 2027 will also be relatively limited.\n\n---\n\nhttps://t.co/hIvtpNrG36", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023497009547833448, "ret_signed_m1d": -0.023497009547833448, "alpha_spy_m1d": -0.023497009547833448, "alpha_c_m1d": -0.023497009547833448, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010012510491760551, "ret_signed_p1d": 0.010012510491760551, "alpha_spy_p1d": 0.004554104153145611, "alpha_c_p1d": -0.003130693077609825, "bench_spy_p1d": 0.005458406338614941, "bench_c_p1d": 0.013143203569370376, "ret_p1w": -0.0019239380400851269, "ret_signed_p1w": -0.0019239380400851269, "alpha_spy_p1w": -0.00885360343995718, "alpha_c_p1w": -0.041871050393231427, "bench_spy_p1w": 0.006929665399872054, "bench_c_p1w": 0.0399471123531463, "ret_p1m": 0.17852591087127184, "ret_signed_p1m": 0.17852591087127184, "alpha_spy_p1m": 0.1598851461669673, "alpha_c_p1m": 0.12305322092448528, "bench_spy_p1m": 0.018640764704304535, "bench_c_p1m": 0.05547268994678656, "ret_p3m": 1.2071313797915728, "ret_signed_p3m": 1.2071313797915728, "alpha_spy_p3m": 1.1928585142032033, "alpha_c_p3m": 0.9978929981789222, "bench_spy_p3m": 0.014272865588369488, "bench_c_p3m": 0.20923838161265063, "ret_p6m": 3.5808222527453895, "ret_signed_p6m": 3.5808222527453895, "alpha_spy_p6m": 3.477530040791245, "alpha_c_p6m": 2.9181915941448695, "bench_spy_p6m": 0.10329221195414462, "bench_c_p6m": 0.66263065860052, "price_path": [-0.5647, -0.5632, -0.559, -0.5436, -0.5203, -0.5141, -0.5035, -0.4813, -0.4484, -0.4195, -0.4124, -0.3924, -0.4026, -0.3765, -0.3785, -0.3653, -0.3545, -0.3617, -0.3715, -0.3921, -0.3579, -0.3529, -0.3268, -0.2867, -0.2621, -0.263, -0.2771, -0.2579, -0.2561, -0.2577, -0.2385, -0.2583, -0.2206, -0.1828, -0.1887, -0.1568, -0.1655, -0.1619, -0.1399, -0.0643, -0.0342, -0.053, 0.0089, 0.0274, 0.0321, 0.0781, 0.0261, 0.0478, 0.0639, 0.0978, 0.1806, 0.1811, 0.1945, 0.1437, 0.0659, 0.1212, 0.0514, 0.0574, 0.0119, -0.0373, 0.0052, 0.0004, -0.0235, 0.0, 0.01, -0.0068, 0.0077, -0.0118, -0.0019, 0.0364, 0.0722, 0.0591, 0.0794, 0.0767, 0.0402, 0.0036, -0.0179, 0.0049, 0.0411, 0.0649, 0.115, 0.1179, 0.1468, 0.151, 0.1785, 0.1862, 0.1805, 0.1716, 0.1716, 0.2591, 0.301, 0.4166, 0.4509, 0.4301, 0.4748, 0.4879, 0.4973, 0.4891, 0.5292, 0.593, 0.6069, 0.6298, 0.7314, 0.7828, 0.7484, 0.7276, 0.8158, 0.9122, 0.9342, 1.0144, 1.0021, 1.1294, 0.9945, 0.9191, 0.9375, 0.9428, 0.8886, 0.9725, 1.0872, 1.1207, 1.1087, 1.0665, 1.073, 1.1076, 1.149, 1.1645, 1.2071, 1.2048, 1.2649, 1.2292, 1.2211, 1.0269, 0.9778, 1.0383, 0.9595, 0.9292, 1.0688, 1.1666, 1.1047, 1.1427, 1.2633, 1.2824, 1.413, 1.3608, 1.2798, 1.1845, 1.1949, 1.192, 1.031, 1.0245, 0.9193, 0.9289, 1.1453, 1.09, 1.1388, 1.2083, 1.2234, 1.4799, 1.5352, 1.6031, 1.722, 1.8648, 1.7713, 1.8498, 1.7627, 1.7601, 1.8315, 1.9772, 1.956, 1.9939, 2.148, 2.0842, 2.1592, 2.2022, 2.2828, 2.4592, 2.6546, 2.8028, 2.9053, 3.229, 3.394, 3.2545, 3.4412, 3.3332, 3.1179, 3.1939, 3.1728, 3.2449, 3.6173, 3.5808]}
{"unit_id": "thread:1994187350004649994", "ticker_idx": 2, "ts": "2025-11-27T23:31:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix has no plans to expand NAND wafer capacity in 2026, concentrating firepower on capital investment for High Bandwidth Memory (HBM) and DRAM. This includes the new M15X plant in Cheongju, which is planned as a dedicated facility for HBM production, with production expected to start as early as the beginning of 2026", "tweet_type": "analysis", "timeline": "months", "summary": "Global memory shortage is structural with 50% contract price upside in 6-9 months; SK Hynix best positioned via HBM/DRAM focus; Samsung and Micron supply-constrained near-term.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "(Exclusive) 2026 Capacity Sold Out! CSPs Trigger Early \"Two-Year LTA\" War for Memory\n\nThe explosive growth of AI demand is accelerating the global memory market into an unprecedented shortage frenzy. Industry sources point out that with major memory manufacturers' 2026 production capacity almost fully booked, a full-year shortage is irreversible. Multiple hyperscale Cloud Service Providers (CSPs) are proposing Long Term Agreement (LTA) negotiations for one year or longer, extending the battlefront to 2027-2028 to lock in future capacity requirements for the next two years ahead of schedule.\n\nThe memory shortage presents an unprecedented level of \"madness.\" With AI being the primary driver making servers the largest application outlet, and CSP giants wielding massive procurement funds, they fell into a panic buying spree in the second half of 2025. Consequently, companies are inclined to build higher inventories for 2026-2027, resulting in purchasing volumes far exceeding actual consumption.\n\nIndustry insiders revealed that memory manufacturers' expansion plans for 2026 are too slow to meet immediate needs, resulting in the entire year's capacity supply being nearly fully booked. This has prompted major CSPs to seek discussions with memory players early regarding 2027 supply capacity. Major CSPs from the US and China are accelerating their pace, conducting \"bundled negotiations\" that combine 2026 procurement volumes and pricing with subsequent purchasing targets.\n\nThis implies that the bargaining power of memory sellers has significantly increased, ensuring a continuous price hike in 2026. There will be no chance for price drops even in the second half of the year, reshaping the pricing model of the memory industry.\n\nTo avoid falling behind in the AI arms race, server operators are willing to offer higher premiums to secure capacity. As the main source of profit for original manufacturers, they are bound to become the priority customer group for supply. However, to qualify for long-term agreements, some CSP giants may propose more favorable conditions to persuade manufacturers—such as advance payments or funding equipment procurement—thereby locking in capacity for 2027-2028 ahead of time. It is reported that supply LTAs for 2027 could be finalized as early as the first quarter of 2026.\n\nIt is not just DRAM that is in severe shortage. Phison CEO K.S. Pua recently pointed out that NAND Flash prices are skyrocketing rapidly; there are even cases where suppliers have proposed price hikes but ultimately still have no stock to sell. He believes that the situation of \"clients holding cash and asking manufacturers to start production\" will happen very soon.\n\nServer OEM operators also privately noted that while AI GPU supply or production yields are currently smooth with stable output, memory shortages have become the shipment bottleneck for the entire supply chain. There are even rumors in the industry of buyers frantically searching for stock with cash in hand. This will cause the memory industry to shift from the past business cycle model to a structural transformation of \"securing orders first, then expanding production.\"\n\nADATA Technology Chairman Chen Li-bai also confirmed: \"Currently, every major CSP is taking the lead in wanting to sign contracts exceeding one year.\"\n\nHowever, only one or two major CSPs are likely to sign relatively long supply contracts. Other CSP operators will at most secure one-year contracts, while buyers outside of CSPs basically cannot sign any contracts.\n\nIt is understood that the few major CSPs currently signing long-term agreements mostly depend on faster decision-making in the second half of 2025. They realized early on that memory supply could not meet demand and responded quickly by requesting to sign contracts for one year or longer, yet only about 30% of CSP operators managed to secure LTA qualifications.\n\nAfter all, memory suppliers cannot overcommit capacity. Chen Li-bai admitted that an LTA is just a dream; 90% of manufacturers cannot sign long-term agreements. Some operators with better relationships might negotiate once a quarter, but the majority can only negotiate monthly, or cannot even confirm supply status for one month later.\n\nFrom the perspective of memory manufacturers, recent spot prices are already at an \"abnormally high\" level. While a correction in the future cannot be ruled out, contract prices remain far lower than spot prices. It is expected that there could still be room for a 50% price hike in the next 6 to 9 months.\n\nAccording to current planning by memory manufacturers, SK Hynix has no plans to expand NAND wafer capacity in 2026, concentrating firepower on capital investment for High Bandwidth Memory (HBM) and DRAM. This includes the new M15X plant in Cheongju, which is planned as a dedicated facility for HBM production, with production expected to start as early as the beginning of 2026.\n\nHowever, Samsung Electronics and Micron are unlikely to have significant capacity contributions in the short term. New plant construction may not be completed until the second half of 2027. Coupled with equipment installation and the wafer production cycle, there will not only be a lack of new capacity in 2026, but output contribution in 2027 will also be relatively limited.\n\n---\n\nhttps://t.co/hIvtpNrG36", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.037454026447427236, "ret_signed_m1d": -0.037454026447427236, "alpha_spy_m1d": -0.037454026447427236, "alpha_c_m1d": -0.037454026447427236, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025735289038781528, "ret_signed_p1d": -0.025735289038781528, "alpha_spy_p1d": -0.03119369537739647, "alpha_c_p1d": -0.038878492608151904, "bench_spy_p1d": 0.005458406338614941, "bench_c_p1d": 0.013143203569370376, "ret_p1w": -0.003676387633430922, "ret_signed_p1w": -0.003676387633430922, "alpha_spy_p1w": -0.010606053033302976, "alpha_c_p1w": -0.04362349998657722, "bench_spy_p1w": 0.006929665399872054, "bench_c_p1w": 0.0399471123531463, "ret_p1m": 0.10110301989888693, "ret_signed_p1m": 0.10110301989888693, "alpha_spy_p1m": 0.08246225519458239, "alpha_c_p1m": 0.04563032995210037, "bench_spy_p1m": 0.018640764704304535, "bench_c_p1m": 0.05547268994678656, "ret_p3m": 0.847426517144503, "ret_signed_p3m": 0.847426517144503, "alpha_spy_p3m": 0.8331536515561335, "alpha_c_p3m": 0.6381881355318524, "bench_spy_p3m": 0.014272865588369488, "bench_c_p3m": 0.20923838161265063, "ret_p6m": 2.5745986077732432, "ret_signed_p6m": 2.5745986077732432, "alpha_spy_p6m": 2.4713063958190986, "alpha_c_p6m": 1.9119679491727233, "bench_spy_p6m": 0.10329221195414462, "bench_c_p6m": 0.66263065860052, "price_path": [-0.5297, -0.5215, -0.5178, -0.5123, -0.4976, -0.4912, -0.471, -0.4416, -0.4361, -0.3966, -0.392, -0.3608, -0.3874, -0.347, -0.347, -0.3552, -0.3369, -0.3433, -0.3451, -0.3819, -0.3589, -0.3617, -0.3387, -0.2735, -0.2735, -0.2735, -0.2735, -0.2735, -0.2735, -0.2138, -0.2377, -0.2441, -0.2239, -0.1688, -0.1449, -0.1082, -0.1201, -0.1155, -0.121, -0.0632, -0.0172, -0.043, 0.025, 0.0434, 0.0268, 0.1389, 0.0764, 0.0636, 0.0893, 0.0654, 0.1132, 0.1371, 0.1334, 0.1242, 0.0287, 0.1132, 0.047, 0.0323, 0.0489, -0.043, -0.0448, -0.0466, -0.0375, 0.0, -0.0257, -0.011, 0.0257, 0.0147, -0.0037, 0.0, 0.0607, 0.0404, 0.079, 0.0386, 0.0496, 0.0184, -0.0257, 0.0129, 0.0147, 0.0055, 0.0662, 0.0735, 0.0809, 0.0809, 0.1011, 0.1765, 0.1967, 0.1967, 0.1967, 0.2445, 0.2794, 0.3346, 0.364, 0.3897, 0.3676, 0.3768, 0.3566, 0.364, 0.3768, 0.3897, 0.4044, 0.3658, 0.3603, 0.3879, 0.4099, 0.3529, 0.4706, 0.546, 0.5827, 0.671, 0.5257, 0.6673, 0.6544, 0.5478, 0.5423, 0.6305, 0.6103, 0.5809, 0.6324, 0.6176, 0.6176, 0.6176, 0.6176, 0.6434, 0.7445, 0.7482, 0.8474, 0.8713, 1.0239, 0.954, 0.954, 0.7293, 0.5635, 0.733, 0.7017, 0.5396, 0.7274, 0.7588, 0.7127, 0.6759, 0.7937, 0.7864, 0.9448, 0.8656, 0.8545, 0.7182, 0.8158, 0.8324, 0.7182, 0.7182, 0.6077, 0.4862, 0.6519, 0.5286, 0.6133, 0.6317, 0.6869, 0.9024, 0.8379, 0.8914, 0.9153, 1.0313, 1.0921, 1.1271, 1.0774, 1.1473, 1.2542, 1.2523, 1.256, 1.2505, 1.3794, 1.3941, 1.3812, 1.3683, 1.3683, 1.6648, 1.6648, 1.9484, 2.0461, 2.105, 2.4623, 2.3794, 2.6391, 2.628, 2.3499, 2.3886, 2.2136, 2.2136, 2.5728, 2.5746]}
{"unit_id": "thread:1994187350004649994", "ticker_idx": 3, "ts": "2025-11-27T23:31:29+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and Micron are unlikely to have significant capacity contributions in the short term. New plant construction may not be completed until the second half of 2027", "tweet_type": "analysis", "timeline": "months", "summary": "Global memory shortage is structural with 50% contract price upside in 6-9 months; SK Hynix best positioned via HBM/DRAM focus; Samsung and Micron supply-constrained near-term.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "(Exclusive) 2026 Capacity Sold Out! CSPs Trigger Early \"Two-Year LTA\" War for Memory\n\nThe explosive growth of AI demand is accelerating the global memory market into an unprecedented shortage frenzy. Industry sources point out that with major memory manufacturers' 2026 production capacity almost fully booked, a full-year shortage is irreversible. Multiple hyperscale Cloud Service Providers (CSPs) are proposing Long Term Agreement (LTA) negotiations for one year or longer, extending the battlefront to 2027-2028 to lock in future capacity requirements for the next two years ahead of schedule.\n\nThe memory shortage presents an unprecedented level of \"madness.\" With AI being the primary driver making servers the largest application outlet, and CSP giants wielding massive procurement funds, they fell into a panic buying spree in the second half of 2025. Consequently, companies are inclined to build higher inventories for 2026-2027, resulting in purchasing volumes far exceeding actual consumption.\n\nIndustry insiders revealed that memory manufacturers' expansion plans for 2026 are too slow to meet immediate needs, resulting in the entire year's capacity supply being nearly fully booked. This has prompted major CSPs to seek discussions with memory players early regarding 2027 supply capacity. Major CSPs from the US and China are accelerating their pace, conducting \"bundled negotiations\" that combine 2026 procurement volumes and pricing with subsequent purchasing targets.\n\nThis implies that the bargaining power of memory sellers has significantly increased, ensuring a continuous price hike in 2026. There will be no chance for price drops even in the second half of the year, reshaping the pricing model of the memory industry.\n\nTo avoid falling behind in the AI arms race, server operators are willing to offer higher premiums to secure capacity. As the main source of profit for original manufacturers, they are bound to become the priority customer group for supply. However, to qualify for long-term agreements, some CSP giants may propose more favorable conditions to persuade manufacturers—such as advance payments or funding equipment procurement—thereby locking in capacity for 2027-2028 ahead of time. It is reported that supply LTAs for 2027 could be finalized as early as the first quarter of 2026.\n\nIt is not just DRAM that is in severe shortage. Phison CEO K.S. Pua recently pointed out that NAND Flash prices are skyrocketing rapidly; there are even cases where suppliers have proposed price hikes but ultimately still have no stock to sell. He believes that the situation of \"clients holding cash and asking manufacturers to start production\" will happen very soon.\n\nServer OEM operators also privately noted that while AI GPU supply or production yields are currently smooth with stable output, memory shortages have become the shipment bottleneck for the entire supply chain. There are even rumors in the industry of buyers frantically searching for stock with cash in hand. This will cause the memory industry to shift from the past business cycle model to a structural transformation of \"securing orders first, then expanding production.\"\n\nADATA Technology Chairman Chen Li-bai also confirmed: \"Currently, every major CSP is taking the lead in wanting to sign contracts exceeding one year.\"\n\nHowever, only one or two major CSPs are likely to sign relatively long supply contracts. Other CSP operators will at most secure one-year contracts, while buyers outside of CSPs basically cannot sign any contracts.\n\nIt is understood that the few major CSPs currently signing long-term agreements mostly depend on faster decision-making in the second half of 2025. They realized early on that memory supply could not meet demand and responded quickly by requesting to sign contracts for one year or longer, yet only about 30% of CSP operators managed to secure LTA qualifications.\n\nAfter all, memory suppliers cannot overcommit capacity. Chen Li-bai admitted that an LTA is just a dream; 90% of manufacturers cannot sign long-term agreements. Some operators with better relationships might negotiate once a quarter, but the majority can only negotiate monthly, or cannot even confirm supply status for one month later.\n\nFrom the perspective of memory manufacturers, recent spot prices are already at an \"abnormally high\" level. While a correction in the future cannot be ruled out, contract prices remain far lower than spot prices. It is expected that there could still be room for a 50% price hike in the next 6 to 9 months.\n\nAccording to current planning by memory manufacturers, SK Hynix has no plans to expand NAND wafer capacity in 2026, concentrating firepower on capital investment for High Bandwidth Memory (HBM) and DRAM. This includes the new M15X plant in Cheongju, which is planned as a dedicated facility for HBM production, with production expected to start as early as the beginning of 2026.\n\nHowever, Samsung Electronics and Micron are unlikely to have significant capacity contributions in the short term. New plant construction may not be completed until the second half of 2027. Coupled with equipment installation and the wafer production cycle, there will not only be a lack of new capacity in 2026, but output contribution in 2027 will also be relatively limited.\n\n---\n\nhttps://t.co/hIvtpNrG36", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02701299538509727, "ret_signed_p1d": -0.02701299538509727, "alpha_spy_p1d": -0.02155458904648233, "alpha_c_p1d": -0.013869791815726895, "bench_spy_p1d": 0.005458406338614941, "bench_c_p1d": 0.013143203569370376, "ret_p1w": -0.015677922064829874, "ret_signed_p1w": 0.015677922064829874, "alpha_spy_p1w": 0.022607587464701928, "alpha_c_p1w": 0.05562503441797617, "bench_spy_p1w": 0.006929665399872054, "bench_c_p1w": 0.0399471123531463, "ret_p1m": 0.23681939337537328, "ret_signed_p1m": -0.23681939337537328, "alpha_spy_p1m": -0.21817862867106874, "alpha_c_p1m": -0.18134670342858672, "bench_spy_p1m": 0.018640764704304535, "bench_c_p1m": 0.05547268994678656, "ret_p3m": 0.8161160602633353, "ret_signed_p3m": -0.8161160602633353, "alpha_spy_p3m": -0.8018431946749658, "alpha_c_p3m": -0.6068776786506846, "bench_spy_p3m": 0.014272865588369488, "bench_c_p3m": 0.20923838161265063, "ret_p6m": 2.264218848644018, "ret_signed_p6m": -2.264218848644018, "alpha_spy_p6m": -2.160926636689873, "alpha_c_p6m": -1.6015881900434978, "bench_spy_p6m": 0.10329221195414462, "bench_c_p6m": 0.66263065860052, "price_path": [-0.4835, -0.4858, -0.4847, -0.4609, -0.4298, -0.4294, -0.413, -0.3924, -0.3465, -0.3176, -0.3152, -0.3107, -0.3056, -0.267, -0.2937, -0.2855, -0.2777, -0.2981, -0.3193, -0.3174, -0.2886, -0.2738, -0.2094, -0.2025, -0.1843, -0.1707, -0.1936, -0.1464, -0.1647, -0.2113, -0.1628, -0.1876, -0.1664, -0.1204, -0.1211, -0.102, -0.1215, -0.1381, -0.1023, -0.0488, -0.0441, -0.0363, -0.0158, -0.0271, -0.0282, 0.0193, -0.0531, 0.0314, 0.035, 0.0333, 0.1001, 0.0471, 0.0636, 0.0291, 0.072, 0.0508, -0.0076, -0.0188, -0.1255, -0.0994, -0.0275, -0.0249, 0.0, 0.0, 0.027, 0.0443, 0.0401, 0.0169, -0.0157, 0.0302, 0.0724, 0.0962, 0.1453, 0.1225, 0.0473, 0.0314, 0.0098, -0.0206, 0.0794, 0.1549, 0.2012, 0.1998, 0.245, 0.245, 0.2368, 0.2789, 0.2714, 0.24, 0.24, 0.3704, 0.3562, 0.4921, 0.4752, 0.4208, 0.4993, 0.5027, 0.4691, 0.4483, 0.4625, 0.576, 0.576, 0.5858, 0.6906, 0.7274, 0.7363, 0.6905, 0.7824, 0.8911, 0.8934, 0.8025, 0.9021, 0.8223, 0.6484, 0.6635, 0.7148, 0.6662, 0.6216, 0.7828, 0.7986, 0.7885, 0.7885, 0.7369, 0.8289, 0.8132, 0.8603, 0.829, 0.8161, 0.8639, 0.8055, 0.7916, 0.7929, 0.6496, 0.7412, 0.7251, 0.6088, 0.6915, 0.7514, 0.8191, 0.7611, 0.8514, 0.9195, 1.0059, 1.0061, 0.9302, 0.8374, 0.7568, 0.7184, 0.6601, 0.5444, 0.552, 0.3987, 0.4684, 0.5989, 0.5919, 0.5919, 0.6419, 0.6412, 0.7679, 0.8321, 0.8281, 0.854, 1.024, 0.983, 0.9873, 0.978, 0.9491, 0.9532, 1.1188, 1.0938, 1.159, 1.28, 1.1919, 1.2535, 1.2478, 1.3567, 1.5055, 1.7826, 1.8973, 1.8106, 2.246, 2.4569, 2.3319, 2.493, 2.3729, 2.1497, 1.9623, 2.0371, 2.1816, 2.3125, 2.2642]}
{"unit_id": "orig:1975558930785771798", "ticker_idx": 0, "ts": "2025-10-07T13:48:47+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Harlan estimates that the planned 6 GW deployment of AMD GPUs will drive roughly $35 – 38 billion in revenue on an annualized basis between 2027 and 2030, translating to an additional $30 – 35 billion per year of OpenAI-related revenue, equivalent to about $4.00–4.50 of incremental EPS—a 50–60% earnings boost over the current 2027 market consensus of around $7.60", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares J.P. Morgan's bullish AMD thesis: OpenAI deal adds $4–4.50 EPS, a 50–60% boost above 2027 consensus, positioning AMD to capture far larger AI GPU TAM share.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "(J.P. Morgan) Investor Commentary on the AMD–OpenAI Deal\n\n– Investors broadly welcomed the deal, calling it a major step for AMD to meaningfully expand its installed base in AI computing. In fact, the scale of the deal is impressive (about 60% the size of OpenAI’s NVDA contract) and effectively serves as a powerful “stamp of approval” for AMD’s Instinct GPUs by associating them directly with OpenAI.\nHarlan estimates that the planned 6 GW deployment of AMD GPUs will drive roughly $35 – 38 billion in revenue on an annualized basis between 2027 and 2030, translating to an additional $30 – 35 billion per year of OpenAI-related revenue, equivalent to about $4.00–4.50 of incremental EPS—a 50–60% earnings boost over the current 2027 market consensus of around $7.60.\n\nInvestors I spoke with emphasized that this partnership is crucial for facilitating the Helios ramp-up and improving the competitiveness of ROCm (AMD’s software platform), while more importantly validating the technology and paving the way for broader adoption. Harlan added that this deal vividly demonstrates how enormous the demand for AI computing will be over the next several years, and that AMD is now positioned to capture a far larger share of the overall data-center GPU TAM than its current mid-single-digit percentage.\n\nOpinions were divided on what this means for ASIC demand. Harlan noted that the ASIC roadmap remains separate, though some investors mentioned that ASIC programs could act as a pressure card in today’s tight GPU-pricing environment.\n\nEarly in the session, there was strong short covering, followed by some profit-taking from real-money investors as the stock rose (a pattern seen broadly across AI names). Most activity occurred through block trades between institutions (“through the pipes”). Overall investor sentiment, however, remained quite positive. Many said this was probably the best possible outcome AMD could hope for—they “can’t write a $100 billion check,” but they’d gladly accept a 10% dilution if it means a $600 share price.\n\nStill, skeptics pointed out potential shareholder dilution and the possibility that part of the 6 GW deployment could overlap with revenues from the Stargate project, leading to some double counting concerns.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03687763224218066, "ret_signed_m1d": -0.03687763224218066, "alpha_spy_m1d": -0.04059892114290564, "alpha_c_m1d": -0.0556582260463776, "bench_spy_m1d": 0.003721288900724984, "bench_c_m1d": 0.018780593804196943, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11370622512583206, "ret_signed_p1d": 0.11370622512583206, "alpha_spy_p1d": 0.10774317271303979, "alpha_c_p1d": 0.08694457376543374, "bench_spy_p1d": 0.005963052412792269, "bench_c_p1d": 0.02676165136039832, "ret_p1w": 0.031109649671150708, "ret_signed_p1w": 0.031109649671150708, "alpha_spy_p1w": 0.04140679043879947, "alpha_c_p1w": 0.04164216067572424, "bench_spy_p1w": -0.01029714076764876, "bench_c_p1w": -0.01053251100457353, "ret_p1m": 0.2119048424635135, "ret_signed_p1m": 0.2119048424635135, "alpha_spy_p1m": 0.19926135863623595, "alpha_c_p1m": 0.14529748446485868, "bench_spy_p1m": 0.012643483827277535, "bench_c_p1m": 0.0666073579986548, "ret_p3m": 0.05654582300828648, "ret_signed_p3m": 0.05654582300828648, "alpha_spy_p3m": 0.03253114811261604, "alpha_c_p3m": -0.05445121201243741, "bench_spy_p3m": 0.02401467489567044, "bench_c_p3m": 0.11099703502072389, "ret_p6m": -0.006146223945586393, "ret_signed_p6m": -0.006146223945586393, "alpha_spy_p6m": 0.009021571645064674, "alpha_c_p6m": -0.17270803325050788, "bench_spy_p6m": -0.015167795590651068, "bench_c_p6m": 0.16656180930492148, "price_path": [-0.3184, -0.3077, -0.3086, -0.2643, -0.2432, -0.2416, -0.2578, -0.2577, -0.2685, -0.2499, -0.2335, -0.2129, -0.179, -0.1611, -0.1513, -0.1664, -0.1882, -0.1642, -0.1759, -0.2288, -0.1849, -0.1832, -0.1855, -0.1729, -0.1281, -0.1445, -0.1607, -0.1672, -0.2126, -0.2189, -0.226, -0.2068, -0.2276, -0.2122, -0.2098, -0.203, -0.2311, -0.2311, -0.2326, -0.2335, -0.2351, -0.2854, -0.2841, -0.2633, -0.2457, -0.264, -0.2503, -0.2381, -0.2414, -0.2475, -0.2534, -0.2559, -0.2445, -0.2393, -0.2394, -0.2375, -0.2461, -0.2371, -0.2351, -0.2246, -0.1975, -0.2215, -0.0369, 0.0, 0.1137, 0.1011, 0.016, 0.0232, 0.0311, 0.1281, 0.109, 0.102, 0.1373, 0.1254, 0.0885, 0.111, 0.1958, 0.2277, 0.2198, 0.2497, 0.2049, 0.2109, 0.2276, 0.1822, 0.2119, 0.1238, 0.1042, 0.1535, 0.123, 0.224, 0.1723, 0.1669, 0.1372, 0.0888, 0.0569, -0.026, -0.0365, 0.0167, -0.0254, 0.0129, 0.0129, 0.0285, 0.039, 0.0176, 0.0288, 0.0211, 0.0305, 0.0454, 0.0478, 0.0469, 0.0469, -0.0035, -0.0186, -0.0111, -0.0634, -0.0494, 0.0091, 0.0163, 0.016, 0.0167, 0.0167, 0.0165, 0.0194, 0.0181, 0.0125, 0.0125, 0.0565, 0.0452, 0.0134, -0.007, -0.0323, -0.0394, -0.0181, 0.0447, 0.0572, 0.0776, 0.0961, 0.0961, 0.0965, 0.181, 0.1996, 0.2277, 0.1882, 0.1916, 0.1949, 0.1923, 0.1192, 0.1643, 0.1447, -0.0535, -0.0899, -0.0145, 0.0212, 0.0097, 0.0098, -0.0263, -0.0198, -0.0198, -0.0399, -0.0539, -0.0385, -0.0537, -0.0705, 0.011, -0.0031, -0.037, -0.0534, -0.0609, -0.0972, -0.0446, -0.057, -0.0902, -0.0417, -0.0391, -0.0316, -0.0651, -0.0857, -0.0706, -0.0719, -0.057, -0.0295, -0.0481, -0.0417, -0.029, 0.0414, -0.0366, -0.045, -0.0731, -0.0382, -0.0061]}
{"unit_id": "orig:1987834602363539719", "ticker_idx": 0, "ts": "2025-11-10T10:47:56+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our forecast for total GB200/300 rack shipments in 2025 at 28,000 units... We forecast GB200/300 rack shipments in 2026 to reach 60,000–70,000 units", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley's bullish GB200/300 rack shipment outlook: 28K units in 2025, 60K–70K in 2026, no yield issues; implies sustained NVDA demand.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: GB200/300 rack shipment outlook maintained (28,000 units)\n\nBased on industry supply chain channel checks, we maintain our forecast for total GB200/300 rack shipments in 2025 at 28,000 units.\nFourth-quarter shipments are expected to increase significantly to 14,000–15,000 units, representing a substantial rise compared with the 8,000–8,500 units shipped in the third quarter.\n\nRack production in October increased by approximately 3% month-over-month, which enhances confidence in achieving the fourth-quarter shipment target.\n\nGB200 rack production is expected to continue through the first quarter of 2026, and GB300 production has already begun, with a stable ramp-up anticipated over the coming months.\n\nDuring discussions with multiple supply chain vendors at the San Jose OCP Summit in mid-October, no specific yield issues related to GB300 were reported.\n\nWe forecast GB200/300 rack shipments in 2026 to reach 60,000–70,000 units.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.054760110948671836, "ret_signed_m1d": -0.054760110948671836, "alpha_spy_m1d": -0.03939543326812012, "alpha_c_m1d": -0.025021268411214326, "bench_spy_m1d": -0.01536467768055172, "bench_c_m1d": -0.02973884253745751, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02959058026486583, "ret_signed_p1d": -0.02959058026486583, "alpha_spy_p1d": -0.03187972996731159, "alpha_c_p1d": -0.008018038392865634, "bench_spy_p1d": 0.0022891497024457585, "bench_c_p1d": -0.021572541872000195, "ret_p1w": -0.06254712593060652, "ret_signed_p1w": -0.06254712593060652, "alpha_spy_p1w": -0.039404852029291226, "alpha_c_p1w": -0.011012710867317477, "bench_spy_p1w": -0.02314227390131529, "bench_c_p1w": -0.05153441506328904, "ret_p1m": -0.07068427675405442, "ret_signed_p1m": -0.07068427675405442, "alpha_spy_p1m": -0.07303206717016197, "alpha_c_p1m": -0.09914099165757018, "bench_spy_p1m": 0.0023477904161075536, "bench_c_p1m": 0.028456714903515756, "ret_p3m": -0.1364503217654084, "ret_signed_p3m": -0.1364503217654084, "alpha_spy_p3m": -0.1337828448780467, "alpha_c_p3m": -0.20185273851669983, "bench_spy_p3m": -0.002667476887361686, "bench_c_p3m": 0.06540241675129144, "ret_p6m": -0.012702486731139606, "ret_signed_p6m": -0.012702486731139606, "alpha_spy_p6m": -0.08086857813030335, "alpha_c_p6m": -0.4740486465169893, "bench_spy_p6m": 0.06816609139916374, "bench_c_p6m": 0.4613461597858497, "price_path": [-0.0878, -0.0856, -0.0935, -0.0857, -0.1177, -0.1189, -0.121, -0.1059, -0.0967, -0.0869, -0.0877, -0.0949, -0.125, -0.125, -0.1421, -0.1429, -0.1377, -0.161, -0.1545, -0.1422, -0.1092, -0.1099, -0.1067, -0.107, -0.1214, -0.1445, -0.1146, -0.1124, -0.0776, -0.1036, -0.1109, -0.1073, -0.1048, -0.0864, -0.0626, -0.0593, -0.051, -0.0574, -0.0679, -0.0704, -0.0499, -0.0326, -0.0798, -0.0539, -0.0956, -0.0966, -0.0866, -0.0795, -0.0824, -0.0899, -0.0943, -0.0849, -0.0643, -0.038, 0.0099, 0.0401, 0.0193, 0.0173, 0.0393, -0.0018, -0.0193, -0.0551, -0.0548, 0.0, -0.0296, -0.0264, -0.0612, -0.0446, -0.0625, -0.0889, -0.0629, -0.0925, -0.1013, -0.0829, -0.1067, -0.0944, -0.0944, -0.1108, -0.0961, -0.0884, -0.0978, -0.0787, -0.0835, -0.0678, -0.0707, -0.0767, -0.091, -0.1207, -0.1143, -0.1071, -0.1412, -0.1251, -0.0907, -0.0771, -0.0494, -0.0524, -0.0524, -0.0428, -0.0544, -0.0578, -0.063, -0.063, -0.0512, -0.0549, -0.0593, -0.0499, -0.0703, -0.0712, -0.0708, -0.0665, -0.0799, -0.0602, -0.0644, -0.0644, -0.1054, -0.079, -0.0713, -0.0571, -0.0631, -0.0528, -0.0378, -0.0328, -0.0397, -0.0675, -0.0939, -0.1248, -0.1365, -0.0685, -0.0452, -0.0527, -0.0452, -0.0608, -0.0815, -0.0815, -0.0707, -0.0556, -0.056, -0.0463, -0.0376, -0.0311, -0.0175, -0.0711, -0.1098, -0.0832, -0.0954, -0.0804, -0.0789, -0.1066, -0.0823, -0.0717, -0.0653, -0.0798, -0.0943, -0.0794, -0.0859, -0.0936, -0.1028, -0.1323, -0.1175, -0.1197, -0.1022, -0.1396, -0.1583, -0.1701, -0.1237, -0.117, -0.1087, -0.1087, -0.1075, -0.1052, -0.0852, -0.076, -0.0522, -0.0488, -0.0127, -0.0008, -0.0034, 0.0133, 0.0152, 0.0043, 0.0174, 0.0031, 0.0464, 0.0883, 0.0711, 0.0514, 0.0027, -0.0029, -0.0028, -0.0127]}
{"unit_id": "orig:1977687040746733995", "ticker_idx": 0, "ts": "2025-10-13T10:45:08+00:00", "symbol": "BABA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman raised its forecast for Alibaba Cloud's external revenue growth from 30% / 25% / 17% → 33% / 29% / 19%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish revision on Alibaba: raised cloud revenue growth forecasts, RMB 460B CapEx base case, AI cloud triple-digit growth for 8 consecutive quarters.", "resolved_tickers": ["BABA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Major Upward Revision of Alibaba’s Capital Expenditure Outlook\n\nIn its October 13 report, Goldman Sachs stated that\nAlibaba’s AI-related cloud revenue has achieved triple-digit growth for eight consecutive quarters,\nreaching 20% of total cloud revenue in the June quarter.\n\nCiting full-stack AI capabilities and the rapid surge in multimodal AI demand,\nGoldman raised its forecast for Alibaba Cloud’s external revenue growth from\n30% / 25% / 17% → 33% / 29% / 19%.\n\nIt also significantly increased its three-year CapEx forecast to RMB 460 billion($64.4B),\ncalling it one of the most aggressive projections on Wall Street.\n\nAnalysts believe this investment scale will strengthen Alibaba Cloud’s leadership in China’s AI-cloud market\nwhile driving rapid international expansion.\nBy FY2028, international business is expected to contribute 25% of Alibaba Cloud’s external revenue.\n\nInterestingly, Goldman’s analysts compared Amazon AWS and Google Cloud’s historical data,\nconcluding that Alibaba’s growth trajectory lags U.S. cloud giants by roughly two years —\na gap that coincides with the technological breakthroughs of ChatGPT (late 2022) and DeepSeek (January 2025).\n\nGoldman estimates Alibaba’s current data-center capacity at 3–4 GW,\nwith plans to expand to 20 GW by 2032, implying about 2 GW of new capacity per year.\nAssuming 1 GW ≈ 1 million GPUs, this expansion indicates massive capital investment over the next three years.\n\nRegarding capital expenditure (CapEx), Goldman outlined three scenarios:\n    • Base case: FY2026–2028 total CapEx RMB 460 billion, CapEx-to-Revenue conversion ratio 0.2–0.3\n    • Optimistic case: Total CapEx RMB 550 billion, conversion ratio > 0.3\n    • Pessimistic case: Total CapEx RMB 380 billion, conversion ratio < 0.2\n\n$BABA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04675980556272108, "ret_signed_m1d": -0.04675980556272108, "alpha_spy_m1d": -0.03164761347370437, "alpha_c_m1d": -0.024909096197653136, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.021850709365067944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023679617565216415, "ret_signed_p1d": -0.023679617565216415, "alpha_spy_p1d": -0.02245796276175971, "alpha_c_p1d": -0.025774896450209694, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": 0.002095278884993279, "ret_p1w": 0.03992568646714334, "ret_signed_p1w": 0.03992568646714334, "alpha_spy_p1w": 0.02746788105020226, "alpha_c_p1w": 0.029577622551343774, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.010348063915799566, "ret_p1m": -0.03602898264371035, "ret_signed_p1m": -0.03602898264371035, "alpha_spy_p1m": -0.06613273813055676, "alpha_c_p1m": -0.060359765896755846, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.024330783253045496, "ret_p3m": -0.07397635944186187, "ret_signed_p3m": -0.07397635944186187, "alpha_spy_p3m": -0.11697146152817361, "alpha_c_p3m": -0.12728684439820492, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.05331048495634305, "ret_p6m": -0.28229720656491086, "ret_signed_p6m": -0.28229720656491086, "alpha_spy_p6m": -0.28219703753661896, "alpha_c_p6m": -0.20764134228316966, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": -0.0746558642817412, "price_path": [-0.3062, -0.2968, -0.2792, -0.279, -0.2764, -0.2652, -0.2737, -0.2804, -0.2677, -0.2845, -0.2963, -0.2768, -0.2982, -0.2956, -0.2984, -0.2755, -0.2749, -0.2785, -0.2888, -0.2661, -0.2395, -0.267, -0.2731, -0.2722, -0.2807, -0.2837, -0.2921, -0.263, -0.2545, -0.2555, -0.2673, -0.2832, -0.1907, -0.1907, -0.1694, -0.182, -0.2152, -0.1872, -0.1535, -0.1182, -0.1372, -0.0682, -0.0704, -0.0526, -0.0276, -0.0038, -0.026, -0.024, -0.0153, -0.0224, 0.0577, 0.0519, 0.0306, 0.0785, 0.0715, 0.0957, 0.1351, 0.1272, 0.1224, 0.087, 0.0858, 0.0412, -0.0468, 0.0, -0.0237, -0.0054, -0.0103, 0.0014, 0.0399, -0.0008, -0.0057, 0.0305, 0.0473, 0.0758, 0.0594, 0.0789, 0.0427, 0.0217, 0.0053, -0.015, -0.0119, 0.0048, -0.0028, -0.0055, -0.036, -0.0534, -0.0418, -0.078, -0.0546, -0.0425, -0.0475, -0.0811, -0.0832, -0.0364, -0.0587, -0.0552, -0.0552, -0.057, -0.0153, -0.0341, -0.0523, -0.0562, -0.0509, -0.052, -0.065, -0.0479, -0.0594, -0.0667, -0.1002, -0.105, -0.1182, -0.1168, -0.102, -0.095, -0.0934, -0.1004, -0.1004, -0.0873, -0.1098, -0.1166, -0.1213, -0.1213, -0.0664, -0.0632, -0.0954, -0.1203, -0.074, -0.095, -0.003, 0.0012, 0.0185, 0.0247, -0.0085, -0.0085, -0.0265, 0.0112, 0.0622, 0.0385, 0.0273, 0.0354, 0.0531, 0.0446, 0.0165, 0.0095, -0.0189, -0.046, -0.0543, -0.0258, -0.0228, -0.0018, -0.0149, -0.0484, -0.0664, -0.0664, -0.0682, -0.0662, -0.0752, -0.0741, -0.0841, -0.0821, -0.0871, -0.1125, -0.1361, -0.1454, -0.1872, -0.2011, -0.2186, -0.2159, -0.2048, -0.1796, -0.183, -0.1955, -0.1894, -0.1804, -0.1813, -0.1941, -0.2512, -0.2662, -0.2443, -0.2478, -0.2214, -0.2482, -0.2645, -0.2687, -0.2479, -0.2583, -0.2683, -0.2683, -0.2668, -0.2823]}
{"unit_id": "reply:1997643638575079810", "ticker_idx": 0, "ts": "2025-12-07T12:25:32+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is currently my largest holding, and I plan to keep holding it through 2026", "tweet_type": "position_disclosure", "timeline": "years", "summary": "Long SK Hynix as largest holding, committed to holding through 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@aric96748244 SK Hynix is currently my largest holding, and I plan to keep holding it through 2026.", "reply_text": null, "parent_text": "@jukan05 Are you seeing $mu micron as a good pick as compare to sk hynix for 2026, not only considering the technical advancement and market share, but also market accessibility and fund inflow, and valuation expansion", "quoted_text": null, "parent_author": "aric96748244", "ret_m1d": -0.05719233447170469, "ret_signed_m1d": -0.05719233447170469, "alpha_spy_m1d": -0.06020566486713075, "alpha_c_m1d": -0.04598634336848628, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01906407531159593, "ret_signed_p1d": -0.01906407531159593, "alpha_spy_p1d": -0.018200930344634325, "alpha_c_p1d": -0.020285088195773793, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.03986126811277635, "ret_signed_p1w": -0.03986126811277635, "alpha_spy_p1w": -0.03561917054031205, "alpha_c_p1w": 0.002602403639090589, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.25823222667283874, "ret_signed_p1m": 0.25823222667283874, "alpha_spy_p1m": 0.24327646453594576, "alpha_c_p1m": 0.2032162932853212, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6338586339218024, "ret_signed_p3m": 0.6338586339218024, "alpha_spy_p3m": 0.6343074010647207, "alpha_c_p3m": 0.5578033195462861, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4735, -0.4683, -0.4311, -0.4268, -0.3973, -0.4224, -0.3843, -0.3843, -0.3921, -0.3748, -0.3809, -0.3826, -0.4172, -0.3956, -0.3982, -0.3765, -0.315, -0.315, -0.315, -0.315, -0.315, -0.315, -0.2588, -0.2813, -0.2873, -0.2683, -0.2163, -0.1938, -0.1592, -0.1704, -0.1661, -0.1713, -0.1168, -0.0735, -0.0977, -0.0336, -0.0163, -0.0319, 0.0738, 0.0149, 0.0027, 0.027, 0.0045, 0.0495, 0.072, 0.0686, 0.0599, -0.0302, 0.0495, -0.0128, -0.0267, -0.0111, -0.0977, -0.0994, -0.1012, -0.0925, -0.0572, -0.0815, -0.0676, -0.0329, -0.0433, -0.0607, -0.0572, 0.0, -0.0191, 0.0173, -0.0208, -0.0104, -0.0399, -0.0815, -0.0451, -0.0433, -0.052, 0.0052, 0.0121, 0.0191, 0.0191, 0.0381, 0.1092, 0.1282, 0.1282, 0.1282, 0.1733, 0.2062, 0.2582, 0.286, 0.3102, 0.2894, 0.2981, 0.279, 0.286, 0.2981, 0.3102, 0.3241, 0.2877, 0.2825, 0.3085, 0.3293, 0.2756, 0.3865, 0.4575, 0.4922, 0.5754, 0.4385, 0.5719, 0.5598, 0.4593, 0.4541, 0.5373, 0.5182, 0.4905, 0.539, 0.5251, 0.5251, 0.5251, 0.5251, 0.5494, 0.6447, 0.6482, 0.7418, 0.7643, 0.9082, 0.8422, 0.8422, 0.6304, 0.4741, 0.6339, 0.6043, 0.4515, 0.6286, 0.6582, 0.6148, 0.58, 0.6912, 0.6842, 0.8335, 0.7589, 0.7485, 0.62, 0.712, 0.7276, 0.62, 0.62, 0.5158, 0.4012, 0.5575, 0.4411, 0.521, 0.5384, 0.5905, 0.7936, 0.7328, 0.7832, 0.8058, 0.9151, 0.9724, 1.0054, 0.9585, 1.0245, 1.1252, 1.1235, 1.127, 1.1218, 1.2433, 1.2572, 1.245, 1.2329, 1.2329, 1.5124, 1.5124, 1.7798, 1.8718, 1.9274, 2.2642, 2.1861, 2.4309, 2.4205, 2.1583, 2.1948, 2.0298, 2.0298, 2.3684, 2.3702, 2.3702, 2.5629, 2.8945, 2.9751, 3.0515, null, null]}
{"unit_id": "orig:1990259788245324163", "ticker_idx": 0, "ts": "2025-11-17T03:24:45+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "I'm genuinely starting to think Google is going to eat Microsoft 365 alive", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Google will take significant share from Microsoft 365; Copilot unlikely to fend off Google's enterprise push.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "I’m genuinely starting to think Google is going to eat Microsoft 365 alive.\n\nCan Copilot really fend off Google’s invasion? https://t.co/a1T4D18pAU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030208331546380474, "ret_signed_m1d": -0.030208331546380474, "alpha_spy_m1d": -0.03961246935226115, "alpha_c_m1d": -0.03155756301071244, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.001349231464331968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0025962638674692506, "ret_signed_p1d": -0.0025962638674692506, "alpha_spy_p1d": 0.005801196690737309, "alpha_c_p1d": -0.008262815013328373, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": 0.005666551145859122, "ret_p1w": 0.11774611487848441, "ret_signed_p1w": 0.11774611487848441, "alpha_spy_p1w": 0.11314925974668966, "alpha_c_p1w": 0.10128626452880729, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": 0.01645985034967712, "ret_p1m": 0.07631235965271932, "ret_signed_p1m": 0.07631235965271932, "alpha_spy_p1m": 0.05648265289023091, "alpha_c_p1m": 0.027112789008890914, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.049199570643828405, "ret_p3m": 0.08484363018480812, "ret_signed_p3m": 0.08484363018480812, "alpha_spy_p3m": 0.05838434929741987, "alpha_c_p3m": 0.05038477242084993, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.034458857763958184, "ret_p6m": 0.3608738156790965, "ret_signed_p6m": 0.3608738156790965, "alpha_spy_p6m": 0.24563172441661596, "alpha_c_p6m": 0.31206998092238014, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.04880383475671635, "price_path": [-0.3013, -0.2998, -0.2776, -0.2692, -0.2739, -0.2727, -0.2581, -0.2537, -0.2537, -0.2591, -0.1914, -0.1857, -0.1762, -0.1789, -0.1593, -0.1609, -0.1567, -0.1551, -0.1172, -0.1188, -0.1245, -0.1157, -0.1063, -0.114, -0.117, -0.1329, -0.1376, -0.135, -0.1437, -0.1471, -0.1408, -0.138, -0.1392, -0.1214, -0.1377, -0.1417, -0.1526, -0.17, -0.1434, -0.1388, -0.1193, -0.1177, -0.1113, -0.0999, -0.1213, -0.1169, -0.1121, -0.0881, -0.0553, -0.0616, -0.0367, -0.0124, -0.0134, -0.0046, -0.0262, -0.0025, -0.0009, -0.0217, 0.0178, 0.0221, 0.0059, -0.0226, -0.0302, 0.0, -0.0026, 0.0273, 0.0155, 0.0514, 0.1177, 0.1348, 0.1226, 0.1226, 0.1234, 0.1048, 0.108, 0.1214, 0.1144, 0.1272, 0.1014, 0.1132, 0.1242, 0.0969, 0.0859, 0.0821, 0.0763, 0.0417, 0.0619, 0.0784, 0.0876, 0.1036, 0.1027, 0.1027, 0.1007, 0.1009, 0.1019, 0.0989, 0.0989, 0.1064, 0.1113, 0.1036, 0.1304, 0.1426, 0.1536, 0.1651, 0.1795, 0.1791, 0.1683, 0.1586, 0.1586, 0.1305, 0.1529, 0.1605, 0.1513, 0.17, 0.1745, 0.1797, 0.1875, 0.1867, 0.2066, 0.1927, 0.1692, 0.163, 0.1335, 0.1386, 0.1185, 0.0917, 0.0848, 0.0733, 0.0733, 0.0603, 0.0649, 0.0633, 0.1058, 0.0936, 0.0915, 0.0985, 0.0792, 0.0945, 0.0761, 0.0658, 0.0642, 0.0563, 0.0481, 0.0763, 0.0787, 0.0846, 0.0665, 0.062, 0.0735, 0.0924, 0.081, 0.079, 0.0575, 0.0612, 0.0204, 0.0221, -0.013, -0.0362, -0.0391, 0.0103, 0.0448, 0.0391, 0.0391, 0.054, 0.0732, 0.1148, 0.1189, 0.1146, 0.1289, 0.1696, 0.1844, 0.1805, 0.2004, 0.1855, 0.1674, 0.1921, 0.1906, 0.21, 0.2308, 0.2289, 0.2294, 0.3519, 0.355, 0.3465, 0.3647, 0.3984, 0.3983, 0.4081, 0.3654, 0.3609]}
{"unit_id": "orig:1990259788245324163", "ticker_idx": 1, "ts": "2025-11-17T03:24:45+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Google is going to eat Microsoft 365 alive", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Google will take significant share from Microsoft 365; Copilot unlikely to fend off Google's enterprise push.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "I’m genuinely starting to think Google is going to eat Microsoft 365 alive.\n\nCan Copilot really fend off Google’s invasion? https://t.co/a1T4D18pAU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005300646417195187, "ret_signed_m1d": -0.005300646417195187, "alpha_spy_m1d": 0.004103491388685487, "alpha_c_m1d": 0.004103491388685487, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.009404137805880675, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026995583662036693, "ret_signed_p1d": 0.026995583662036693, "alpha_spy_p1d": 0.018598123103830133, "alpha_c_p1d": 0.018598123103830133, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.00839746055820656, "ret_p1w": -0.06424328751909936, "ret_signed_p1w": 0.06424328751909936, "alpha_spy_p1w": 0.06884014265089411, "alpha_c_p1w": 0.06884014265089411, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": 0.004596855131794753, "ret_p1m": -0.059525039278896186, "ret_signed_p1m": 0.059525039278896186, "alpha_spy_p1m": 0.0793547460413846, "alpha_c_p1m": 0.0793547460413846, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.019829706762488408, "ret_p3m": -0.20669949131537446, "ret_signed_p3m": 0.20669949131537446, "alpha_spy_p3m": 0.2331587722027627, "alpha_c_p3m": 0.2331587722027627, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": 0.026459280887388248, "ret_p6m": -0.1931552843345481, "ret_signed_p6m": 0.1931552843345481, "alpha_spy_p6m": 0.3083973755970286, "alpha_c_p6m": 0.3083973755970286, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.11524209126248053, "price_path": [-0.0051, -0.0064, -0.0005, -0.0064, -0.0107, -0.0015, 0.0042, -0.0016, -0.0016, -0.0047, -0.0042, 0.0009, -0.0246, -0.0183, -0.0179, -0.014, -0.0128, 0.0047, 0.0155, 0.0031, 0.005, 0.0019, 0.0206, 0.0137, 0.0034, 0.0052, -0.0009, 0.0078, 0.014, 0.0206, 0.0241, 0.0163, 0.0194, 0.0415, 0.0325, 0.0342, 0.0294, 0.0068, 0.0129, 0.012, 0.0117, 0.0081, 0.012, 0.0183, 0.02, 0.0257, 0.0258, 0.0318, 0.0474, 0.0681, 0.0671, 0.036, 0.0203, 0.0188, 0.0135, -0.0007, -0.0205, -0.021, -0.0029, 0.0023, 0.0072, -0.0083, 0.0053, 0.0, -0.027, -0.0401, -0.0555, -0.068, -0.0642, -0.0583, -0.0415, -0.0415, -0.0287, -0.0391, -0.0327, -0.0569, -0.0507, -0.0462, -0.0306, -0.0287, -0.0552, -0.0455, -0.0553, -0.0626, -0.0595, -0.0601, -0.0445, -0.0407, -0.0427, -0.0389, -0.0366, -0.0366, -0.0372, -0.0384, -0.0376, -0.0453, -0.0453, -0.0663, -0.0665, -0.0553, -0.0455, -0.0561, -0.0538, -0.058, -0.0708, -0.0931, -0.0985, -0.0922, -0.0922, -0.1027, -0.1233, -0.1094, -0.0801, -0.0716, -0.0513, -0.0492, -0.1442, -0.1505, -0.1642, -0.1882, -0.1823, -0.2228, -0.2081, -0.1835, -0.1841, -0.2017, -0.2067, -0.2077, -0.2077, -0.2165, -0.2111, -0.2116, -0.214, -0.2393, -0.2303, -0.2073, -0.2051, -0.2229, -0.2114, -0.2008, -0.1982, -0.1874, -0.1908, -0.1899, -0.1971, -0.1989, -0.2048, -0.2173, -0.2086, -0.2097, -0.2248, -0.2303, -0.2444, -0.2422, -0.2625, -0.2658, -0.2759, -0.2941, -0.2897, -0.2676, -0.2691, -0.261, -0.261, -0.2622, -0.2634, -0.2593, -0.2618, -0.2662, -0.2395, -0.2222, -0.1863, -0.1684, -0.1634, -0.1728, -0.1607, -0.1434, -0.1774, -0.1598, -0.1594, -0.1507, -0.1601, -0.1931, -0.18, -0.1816, -0.186, -0.1809, -0.1674, -0.1786, -0.1835, -0.1932]}
{"unit_id": "orig:1976280758173171771", "ticker_idx": 0, "ts": "2025-10-09T13:37:04+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Equipment suppliers with large memory exposure—such as Applied Materials and Lam Research—are expected to benefit from the ongoing supply tightness", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman shares constructive WFE outlook for 2026; AMAT and LRCX benefit from memory tightness, TER draws interest with Titan, TSMC CapEx rising 10% YoY.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Key Takeaways from SEMICON West 2025\n\nWe attended the SEMICON West conference held in Phoenix, Arizona, on October 7–8, meeting with various industry players including AMAT, TER, ENTG, and MKSI. Overall, we came away from the event with a gradually more positive view on 2026 WFE (Wafer Fab Equipment) growth, consistent with our recent WFE outlook update. Below we summarize the key industry insights and company-level highlights.\n\n⸻\n\nIndustry Takeaways\n\n1. 2026 WFE Outlook\nWhile most companies avoided providing specific figures for their 2026 WFE outlook, many indicated expectations for improvement in the memory segment, even though there has been no notable change in current order trends. We estimate that the WFE market will grow 9% YoY to about USD 120 billion in 2026, driven by strength in leading-edge logic, HBM, and NAND.\n\n2. Leading-Edge Logic\nAs AI data center build-outs continue to expand, the leading-edge logic segment is expected to maintain growth momentum into next year. The transition to the N2 (2nm) node is likely to support this investment trend. Our colleague Bruce Lu projects TSMC’s 2026 CapEx at USD 44 billion, up about 10% YoY.\n\n3. DRAM\nThe DRAM investment cycle should remain active as competition in the HBM market intensifies and demand grows for solutions such as 4F² that require higher capital intensity. Equipment suppliers with large memory exposure—such as Applied Materials and Lam Research—are expected to benefit from the ongoing supply tightness.\n\n4. NAND\nAlthough companies did not highlight additional NAND equipment orders, our channel checks suggest utilization rates at suppliers are ramping rapidly. This implies more equipment upgrades through 2026.\n\n5. Testing\nAs chip complexity rises, innovative solutions are needed to shorten test times and reduce customers’ time-to-market. Teradyne recently introduced its new AI-oriented product, Titan, which features excellent thermal management capabilities and is drawing strong customer interest.\n\n6. Materials and Wafer Consumables\nMost industry participants maintained a relatively positive outlook for the consumables business in 2026, particularly in leading-edge logic and DRAM. However, this optimism is partially offset by the weak utilization in legacy logic and analog semiconductor segments.\n\n⸻\n\nCompany-Specific Highlights\n\nApplied Materials (AMAT)\nManagement maintained a constructive stance on its competitive position across logic, DRAM, and advanced packaging. While avoiding specific commentary on 2026 WFE figures, they emphasized that there is typically a time lag between market demand signals and actual equipment orders.\n\nThey expect their market share to expand through the transition to 4F² DRAM, as 3D structures become more critical—requiring additional material engineering steps such as deposition and etch.\n\nRegarding China, management noted that despite comprehensive export controls, AMAT’s China business remains focused on ICAPS (IoT, communications, automotive, power, and sensors)—areas still permitted for trade—and expressed confidence in competing effectively within those boundaries.\n\nThey also highlighted progress in hybrid bonding technology, stating that customer qualification is underway and adoption is expected to begin in the logic market first, followed later by the DRAM segment.\n\n$AMAT $LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012664470869275024, "ret_signed_m1d": -0.012664470869275024, "alpha_spy_m1d": -0.015569896290302476, "alpha_c_m1d": -0.015155791065019408, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.046981330527432474, "ret_signed_p1d": -0.046981330527432474, "alpha_spy_p1d": -0.019953502628342834, "alpha_c_p1d": 0.010577796426853636, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": 0.03368138544787702, "ret_signed_p1w": 0.03368138544787702, "alpha_spy_p1w": 0.04935566538151903, "alpha_c_p1w": 0.039214272931790806, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": 0.044348711289748266, "ret_signed_p1m": 0.044348711289748266, "alpha_spy_p1m": 0.04463186405679376, "alpha_c_p1m": 0.03591897904060426, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": 0.34630171959293476, "ret_signed_p3m": 0.34630171959293476, "alpha_spy_p3m": 0.31248828562208386, "alpha_c_p3m": 0.2199550552342613, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": 0.5868748346786083, "ret_signed_p6m": 0.5868748346786083, "alpha_spy_p6m": 0.6041519092642116, "alpha_c_p6m": 0.4468708689977101, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.1079, -0.098, -0.1182, -0.1286, -0.138, -0.1282, -0.1529, -0.1535, -0.1485, -0.1595, -0.1388, -0.1472, -0.1428, -0.185, -0.1853, -0.1725, -0.1891, -0.1937, -0.171, -0.1632, -0.1654, -0.147, -0.1399, -0.148, -0.2678, -0.2598, -0.2657, -0.2714, -0.2744, -0.2624, -0.2647, -0.2532, -0.2538, -0.2498, -0.2703, -0.2703, -0.2847, -0.2907, -0.2817, -0.2612, -0.2644, -0.2578, -0.2582, -0.2276, -0.2383, -0.2241, -0.2123, -0.1914, -0.1386, -0.1371, -0.0898, -0.0882, -0.0856, -0.094, -0.0744, -0.0697, -0.0706, -0.0116, 0.0149, -0.0126, 0.0164, -0.0397, -0.0127, 0.0, -0.047, -0.0037, -0.0096, 0.033, 0.0337, 0.0213, 0.0355, 0.0259, 0.0012, 0.0371, 0.0384, 0.0501, 0.0333, 0.0701, 0.0556, 0.0581, 0.079, 0.0449, 0.0935, 0.0601, 0.0443, 0.0671, 0.038, 0.0473, 0.0133, 0.0259, 0.0382, 0.0219, 0.0673, 0.0016, 0.0188, 0.0502, 0.1027, 0.1369, 0.1369, 0.1473, 0.1586, 0.2068, 0.2218, 0.2255, 0.2189, 0.2196, 0.215, 0.2514, 0.2285, 0.1789, 0.1883, 0.1772, 0.1292, 0.153, 0.1662, 0.178, 0.1836, 0.1861, 0.1861, 0.1912, 0.1964, 0.1824, 0.1688, 0.1688, 0.2229, 0.2931, 0.3463, 0.329, 0.2809, 0.3698, 0.3974, 0.3866, 0.373, 0.4512, 0.4873, 0.4873, 0.4474, 0.4792, 0.4499, 0.4662, 0.453, 0.5132, 0.5316, 0.5525, 0.466, 0.4936, 0.4494, 0.3535, 0.3826, 0.4668, 0.5035, 0.4967, 0.5458, 0.4936, 0.6142, 0.6142, 0.6334, 0.6796, 0.6841, 0.7094, 0.7011, 0.721, 0.7985, 0.711, 0.6954, 0.6948, 0.5999, 0.6292, 0.578, 0.4788, 0.5435, 0.5751, 0.5987, 0.5359, 0.5553, 0.5764, 0.605, 0.5914, 0.6267, 0.626, 0.6475, 0.7031, 0.6819, 0.5417, 0.5354, 0.4714, 0.5565, 0.6111, 0.5869, 0.5869]}
{"unit_id": "orig:1976280758173171771", "ticker_idx": 1, "ts": "2025-10-09T13:37:04+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Equipment suppliers with large memory exposure—such as Applied Materials and Lam Research—are expected to benefit from the ongoing supply tightness", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman shares constructive WFE outlook for 2026; AMAT and LRCX benefit from memory tightness, TER draws interest with Titan, TSMC CapEx rising 10% YoY.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Key Takeaways from SEMICON West 2025\n\nWe attended the SEMICON West conference held in Phoenix, Arizona, on October 7–8, meeting with various industry players including AMAT, TER, ENTG, and MKSI. Overall, we came away from the event with a gradually more positive view on 2026 WFE (Wafer Fab Equipment) growth, consistent with our recent WFE outlook update. Below we summarize the key industry insights and company-level highlights.\n\n⸻\n\nIndustry Takeaways\n\n1. 2026 WFE Outlook\nWhile most companies avoided providing specific figures for their 2026 WFE outlook, many indicated expectations for improvement in the memory segment, even though there has been no notable change in current order trends. We estimate that the WFE market will grow 9% YoY to about USD 120 billion in 2026, driven by strength in leading-edge logic, HBM, and NAND.\n\n2. Leading-Edge Logic\nAs AI data center build-outs continue to expand, the leading-edge logic segment is expected to maintain growth momentum into next year. The transition to the N2 (2nm) node is likely to support this investment trend. Our colleague Bruce Lu projects TSMC’s 2026 CapEx at USD 44 billion, up about 10% YoY.\n\n3. DRAM\nThe DRAM investment cycle should remain active as competition in the HBM market intensifies and demand grows for solutions such as 4F² that require higher capital intensity. Equipment suppliers with large memory exposure—such as Applied Materials and Lam Research—are expected to benefit from the ongoing supply tightness.\n\n4. NAND\nAlthough companies did not highlight additional NAND equipment orders, our channel checks suggest utilization rates at suppliers are ramping rapidly. This implies more equipment upgrades through 2026.\n\n5. Testing\nAs chip complexity rises, innovative solutions are needed to shorten test times and reduce customers’ time-to-market. Teradyne recently introduced its new AI-oriented product, Titan, which features excellent thermal management capabilities and is drawing strong customer interest.\n\n6. Materials and Wafer Consumables\nMost industry participants maintained a relatively positive outlook for the consumables business in 2026, particularly in leading-edge logic and DRAM. However, this optimism is partially offset by the weak utilization in legacy logic and analog semiconductor segments.\n\n⸻\n\nCompany-Specific Highlights\n\nApplied Materials (AMAT)\nManagement maintained a constructive stance on its competitive position across logic, DRAM, and advanced packaging. While avoiding specific commentary on 2026 WFE figures, they emphasized that there is typically a time lag between market demand signals and actual equipment orders.\n\nThey expect their market share to expand through the transition to 4F² DRAM, as 3D structures become more critical—requiring additional material engineering steps such as deposition and etch.\n\nRegarding China, management noted that despite comprehensive export controls, AMAT’s China business remains focused on ICAPS (IoT, communications, automotive, power, and sensors)—areas still permitted for trade—and expressed confidence in competing effectively within those boundaries.\n\nThey also highlighted progress in hybrid bonding technology, stating that customer qualification is underway and adoption is expected to begin in the logic market first, followed later by the DRAM segment.\n\n$AMAT $LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010921988186250342, "ret_signed_m1d": 0.010921988186250342, "alpha_spy_m1d": 0.00801656276522289, "alpha_c_m1d": 0.008430667990505958, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06829783290147062, "ret_signed_p1d": -0.06829783290147062, "alpha_spy_p1d": -0.04127000500238098, "alpha_c_p1d": -0.010738705947184513, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": 0.009716368428889455, "ret_signed_p1w": 0.009716368428889455, "alpha_spy_p1w": 0.025390648362531465, "alpha_c_p1w": 0.015249255912803239, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": 0.13014192648248146, "ret_signed_p1m": 0.13014192648248146, "alpha_spy_p1m": 0.13042507924952695, "alpha_c_p1m": 0.12171219423333746, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": 0.470217968759157, "ret_signed_p3m": 0.470217968759157, "alpha_spy_p3m": 0.4364045347883061, "alpha_c_p3m": 0.3438713044004835, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": 0.5536296219765089, "ret_signed_p6m": 0.5536296219765089, "alpha_spy_p6m": 0.5709066965621121, "alpha_c_p6m": 0.41362565629561066, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.2949, -0.2846, -0.2896, -0.2866, -0.2875, -0.2799, -0.3085, -0.3127, -0.3079, -0.3137, -0.3019, -0.2997, -0.2986, -0.3287, -0.3179, -0.3034, -0.3157, -0.3209, -0.2982, -0.2798, -0.278, -0.2548, -0.2445, -0.2399, -0.2956, -0.3001, -0.2898, -0.2982, -0.3034, -0.2916, -0.2831, -0.2665, -0.2662, -0.2632, -0.2911, -0.2911, -0.3132, -0.3082, -0.2892, -0.2713, -0.2563, -0.2528, -0.2401, -0.1819, -0.1721, -0.1562, -0.1473, -0.1372, -0.1059, -0.1016, -0.0643, -0.0662, -0.0899, -0.0913, -0.0899, -0.0703, -0.0504, 0.0127, 0.0425, 0.0341, 0.0578, -0.0046, 0.0109, 0.0, -0.0683, -0.0226, -0.0191, 0.0268, 0.0097, 0.0036, 0.0216, 0.0287, 0.0018, 0.0464, 0.0757, 0.1128, 0.1037, 0.1395, 0.1419, 0.1167, 0.1435, 0.1048, 0.1706, 0.1503, 0.1301, 0.1799, 0.1289, 0.1448, 0.0874, 0.0515, 0.0458, 0.0159, 0.0553, -0.01, 0.0117, 0.0665, 0.0775, 0.1003, 0.1003, 0.1064, 0.0978, 0.1219, 0.1348, 0.1159, 0.1274, 0.1561, 0.1779, 0.1953, 0.1985, 0.1403, 0.1672, 0.1598, 0.101, 0.17, 0.2238, 0.245, 0.2443, 0.2597, 0.2597, 0.265, 0.2494, 0.2345, 0.216, 0.216, 0.3146, 0.3836, 0.4702, 0.4427, 0.4276, 0.5512, 0.5657, 0.5229, 0.4832, 0.5449, 0.5839, 0.5839, 0.58, 0.6225, 0.5678, 0.5482, 0.5832, 0.694, 0.7019, 0.763, 0.6585, 0.6872, 0.6346, 0.4903, 0.5153, 0.6411, 0.6288, 0.6098, 0.6703, 0.6431, 0.6732, 0.6732, 0.6735, 0.7056, 0.6864, 0.7399, 0.7211, 0.7351, 0.7723, 0.6983, 0.6615, 0.641, 0.5435, 0.586, 0.5269, 0.4177, 0.5018, 0.5308, 0.5567, 0.49, 0.5092, 0.5605, 0.6107, 0.5982, 0.6642, 0.6242, 0.6594, 0.6987, 0.6604, 0.5051, 0.5036, 0.422, 0.5196, 0.579, 0.5536, 0.5536]}
{"unit_id": "orig:1976280758173171771", "ticker_idx": 2, "ts": "2025-10-09T13:37:04+00:00", "symbol": "TER", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Teradyne recently introduced its new AI-oriented product, Titan, which features excellent thermal management capabilities and is drawing strong customer interest", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman shares constructive WFE outlook for 2026; AMAT and LRCX benefit from memory tightness, TER draws interest with Titan, TSMC CapEx rising 10% YoY.", "resolved_tickers": ["TER"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Key Takeaways from SEMICON West 2025\n\nWe attended the SEMICON West conference held in Phoenix, Arizona, on October 7–8, meeting with various industry players including AMAT, TER, ENTG, and MKSI. Overall, we came away from the event with a gradually more positive view on 2026 WFE (Wafer Fab Equipment) growth, consistent with our recent WFE outlook update. Below we summarize the key industry insights and company-level highlights.\n\n⸻\n\nIndustry Takeaways\n\n1. 2026 WFE Outlook\nWhile most companies avoided providing specific figures for their 2026 WFE outlook, many indicated expectations for improvement in the memory segment, even though there has been no notable change in current order trends. We estimate that the WFE market will grow 9% YoY to about USD 120 billion in 2026, driven by strength in leading-edge logic, HBM, and NAND.\n\n2. Leading-Edge Logic\nAs AI data center build-outs continue to expand, the leading-edge logic segment is expected to maintain growth momentum into next year. The transition to the N2 (2nm) node is likely to support this investment trend. Our colleague Bruce Lu projects TSMC’s 2026 CapEx at USD 44 billion, up about 10% YoY.\n\n3. DRAM\nThe DRAM investment cycle should remain active as competition in the HBM market intensifies and demand grows for solutions such as 4F² that require higher capital intensity. Equipment suppliers with large memory exposure—such as Applied Materials and Lam Research—are expected to benefit from the ongoing supply tightness.\n\n4. NAND\nAlthough companies did not highlight additional NAND equipment orders, our channel checks suggest utilization rates at suppliers are ramping rapidly. This implies more equipment upgrades through 2026.\n\n5. Testing\nAs chip complexity rises, innovative solutions are needed to shorten test times and reduce customers’ time-to-market. Teradyne recently introduced its new AI-oriented product, Titan, which features excellent thermal management capabilities and is drawing strong customer interest.\n\n6. Materials and Wafer Consumables\nMost industry participants maintained a relatively positive outlook for the consumables business in 2026, particularly in leading-edge logic and DRAM. However, this optimism is partially offset by the weak utilization in legacy logic and analog semiconductor segments.\n\n⸻\n\nCompany-Specific Highlights\n\nApplied Materials (AMAT)\nManagement maintained a constructive stance on its competitive position across logic, DRAM, and advanced packaging. While avoiding specific commentary on 2026 WFE figures, they emphasized that there is typically a time lag between market demand signals and actual equipment orders.\n\nThey expect their market share to expand through the transition to 4F² DRAM, as 3D structures become more critical—requiring additional material engineering steps such as deposition and etch.\n\nRegarding China, management noted that despite comprehensive export controls, AMAT’s China business remains focused on ICAPS (IoT, communications, automotive, power, and sensors)—areas still permitted for trade—and expressed confidence in competing effectively within those boundaries.\n\nThey also highlighted progress in hybrid bonding technology, stating that customer qualification is underway and adoption is expected to begin in the logic market first, followed later by the DRAM segment.\n\n$AMAT $LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004063712939157971, "ret_signed_m1d": -0.004063712939157971, "alpha_spy_m1d": -0.006969138360185423, "alpha_c_m1d": -0.006555033134902355, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09029553941007906, "ret_signed_p1d": -0.09029553941007906, "alpha_spy_p1d": -0.06326771151098942, "alpha_c_p1d": -0.03273641245579295, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": -0.040567658815263474, "ret_signed_p1w": -0.040567658815263474, "alpha_spy_p1w": -0.024893378881621464, "alpha_c_p1w": -0.03503477133134969, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": 0.25545825581487147, "ret_signed_p1m": 0.25545825581487147, "alpha_spy_p1m": 0.25574140858191696, "alpha_c_p1m": 0.24702852356572746, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": 0.5773325930155757, "ret_signed_p3m": 0.5773325930155757, "alpha_spy_p3m": 0.5435191590447248, "alpha_c_p3m": 0.45098592865690224, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": 1.1349508048324153, "ret_signed_p6m": 1.1349508048324153, "alpha_spy_p6m": 1.1522278794180185, "alpha_c_p6m": 0.994946839151517, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.3448, -0.3599, -0.3671, -0.3615, -0.3538, -0.3476, -0.3553, -0.3599, -0.3716, -0.3797, -0.3729, -0.377, -0.2593, -0.2608, -0.2833, -0.2726, -0.2643, -0.2784, -0.2675, -0.2605, -0.2797, -0.2319, -0.2155, -0.2306, -0.2471, -0.2403, -0.2389, -0.2464, -0.248, -0.2067, -0.1945, -0.1839, -0.195, -0.1888, -0.1864, -0.1864, -0.1701, -0.1769, -0.1785, -0.1721, -0.1811, -0.2075, -0.1885, -0.2041, -0.2269, -0.2143, -0.2153, -0.2126, -0.1816, -0.1745, -0.0689, -0.0724, -0.0825, -0.0849, -0.068, -0.0748, -0.052, -0.028, -0.0045, 0.0031, 0.0185, -0.0348, -0.0041, 0.0, -0.0903, -0.0417, -0.0566, -0.0287, -0.0406, -0.0484, -0.0413, -0.0128, -0.0437, -0.001, -0.0063, 0.0159, -0.0056, 0.198, 0.2183, 0.2519, 0.2609, 0.2098, 0.292, 0.2743, 0.2555, 0.2679, 0.2207, 0.2347, 0.1681, 0.1707, 0.1532, 0.1276, 0.1587, 0.0738, 0.0945, 0.144, 0.1557, 0.2364, 0.2364, 0.2537, 0.2381, 0.3092, 0.3446, 0.3691, 0.3839, 0.3992, 0.3783, 0.4062, 0.4059, 0.3328, 0.342, 0.3254, 0.2766, 0.3127, 0.3453, 0.3594, 0.3682, 0.3684, 0.3684, 0.371, 0.3603, 0.3556, 0.3342, 0.3342, 0.4307, 0.513, 0.5773, 0.5335, 0.491, 0.4975, 0.5465, 0.5803, 0.5866, 0.5695, 0.5726, 0.5726, 0.5438, 0.5984, 0.5794, 0.5797, 0.5974, 0.6469, 0.7265, 0.7361, 0.6615, 0.7199, 0.9505, 0.8546, 0.8688, 1.0686, 1.1368, 1.1015, 1.2157, 1.1438, 1.1698, 1.1698, 1.1068, 1.1709, 1.1783, 1.24, 1.1963, 1.2693, 1.364, 1.2942, 1.2068, 1.2468, 1.0978, 1.1045, 1.1072, 0.8828, 1.0441, 1.074, 1.0824, 0.9764, 0.975, 1.0568, 1.0645, 1.0692, 1.0852, 1.0055, 1.0957, 1.2076, 1.2298, 1.0503, 1.0384, 0.9056, 1.0443, 1.1528, 1.135, 1.135]}
{"unit_id": "orig:1976280758173171771", "ticker_idx": 3, "ts": "2025-10-09T13:37:04+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Bruce Lu projects TSMC's 2026 CapEx at USD 44 billion, up about 10% YoY", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman shares constructive WFE outlook for 2026; AMAT and LRCX benefit from memory tightness, TER draws interest with Titan, TSMC CapEx rising 10% YoY.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Key Takeaways from SEMICON West 2025\n\nWe attended the SEMICON West conference held in Phoenix, Arizona, on October 7–8, meeting with various industry players including AMAT, TER, ENTG, and MKSI. Overall, we came away from the event with a gradually more positive view on 2026 WFE (Wafer Fab Equipment) growth, consistent with our recent WFE outlook update. Below we summarize the key industry insights and company-level highlights.\n\n⸻\n\nIndustry Takeaways\n\n1. 2026 WFE Outlook\nWhile most companies avoided providing specific figures for their 2026 WFE outlook, many indicated expectations for improvement in the memory segment, even though there has been no notable change in current order trends. We estimate that the WFE market will grow 9% YoY to about USD 120 billion in 2026, driven by strength in leading-edge logic, HBM, and NAND.\n\n2. Leading-Edge Logic\nAs AI data center build-outs continue to expand, the leading-edge logic segment is expected to maintain growth momentum into next year. The transition to the N2 (2nm) node is likely to support this investment trend. Our colleague Bruce Lu projects TSMC’s 2026 CapEx at USD 44 billion, up about 10% YoY.\n\n3. DRAM\nThe DRAM investment cycle should remain active as competition in the HBM market intensifies and demand grows for solutions such as 4F² that require higher capital intensity. Equipment suppliers with large memory exposure—such as Applied Materials and Lam Research—are expected to benefit from the ongoing supply tightness.\n\n4. NAND\nAlthough companies did not highlight additional NAND equipment orders, our channel checks suggest utilization rates at suppliers are ramping rapidly. This implies more equipment upgrades through 2026.\n\n5. Testing\nAs chip complexity rises, innovative solutions are needed to shorten test times and reduce customers’ time-to-market. Teradyne recently introduced its new AI-oriented product, Titan, which features excellent thermal management capabilities and is drawing strong customer interest.\n\n6. Materials and Wafer Consumables\nMost industry participants maintained a relatively positive outlook for the consumables business in 2026, particularly in leading-edge logic and DRAM. However, this optimism is partially offset by the weak utilization in legacy logic and analog semiconductor segments.\n\n⸻\n\nCompany-Specific Highlights\n\nApplied Materials (AMAT)\nManagement maintained a constructive stance on its competitive position across logic, DRAM, and advanced packaging. While avoiding specific commentary on 2026 WFE figures, they emphasized that there is typically a time lag between market demand signals and actual equipment orders.\n\nThey expect their market share to expand through the transition to 4F² DRAM, as 3D structures become more critical—requiring additional material engineering steps such as deposition and etch.\n\nRegarding China, management noted that despite comprehensive export controls, AMAT’s China business remains focused on ICAPS (IoT, communications, automotive, power, and sensors)—areas still permitted for trade—and expressed confidence in competing effectively within those boundaries.\n\nThey also highlighted progress in hybrid bonding technology, stating that customer qualification is underway and adoption is expected to begin in the logic market first, followed later by the DRAM segment.\n\n$AMAT $LRCX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017361126814406158, "ret_signed_m1d": -0.017361126814406158, "alpha_spy_m1d": -0.02026655223543361, "alpha_c_m1d": -0.019852447010150542, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.02702782789908964, "alpha_c_p1d": 0.05755912695428611, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": 0.031249960001040966, "ret_signed_p1w": 0.031249960001040966, "alpha_spy_p1w": 0.046924239934682976, "alpha_c_p1w": 0.03678284748495475, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": 0.013888833186634919, "ret_signed_p1m": 0.013888833186634919, "alpha_spy_p1m": 0.01417198595368041, "alpha_c_p1m": 0.005459100937490913, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": 0.18797445698479964, "ret_signed_p3m": 0.18797445698479964, "alpha_spy_p3m": 0.15416102301394874, "alpha_c_p3m": 0.061627792626126165, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": 0.2652487764585114, "ret_signed_p6m": 0.2652487764585114, "alpha_spy_p6m": 0.28252585104411465, "alpha_c_p6m": 0.12524481077761318, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.2426, -0.2322, -0.2184, -0.2184, -0.2011, -0.2046, -0.2184, -0.208, -0.208, -0.208, -0.208, -0.2149, -0.2011, -0.1977, -0.1977, -0.2149, -0.2046, -0.2219, -0.1838, -0.1873, -0.1838, -0.1838, -0.17, -0.1873, -0.1838, -0.1838, -0.1804, -0.2149, -0.2046, -0.2149, -0.1907, -0.1873, -0.1769, -0.1977, -0.1977, -0.1942, -0.1977, -0.1977, -0.1977, -0.1838, -0.1838, -0.17, -0.1527, -0.1423, -0.1285, -0.1319, -0.1111, -0.1215, -0.1076, -0.1215, -0.1007, -0.0694, -0.0694, -0.0833, -0.0972, -0.0972, -0.0938, -0.0799, -0.0521, -0.0278, -0.0278, -0.0035, -0.0174, 0.0, 0.0, -0.0174, -0.0104, 0.0174, 0.0312, 0.0069, 0.0278, 0.0278, 0.0139, 0.0069, 0.0069, 0.0278, 0.0243, 0.0451, 0.0451, 0.0417, 0.0486, 0.0451, 0.0139, 0.0174, 0.0139, 0.0243, 0.0174, 0.0243, 0.0139, -0.0069, 0.0035, -0.0243, -0.0313, 0.0104, -0.0382, -0.0451, -0.0174, 0.0, -0.0035, 0.0, -0.0208, -0.0069, 0.0069, 0.0035, 0.0139, 0.0382, 0.0278, 0.0451, 0.0242, 0.0312, 0.0103, -0.0002, -0.0036, -0.0036, -0.0036, 0.0208, 0.0382, 0.0417, 0.0417, 0.0521, 0.066, 0.0591, 0.08, 0.08, 0.1044, 0.1636, 0.188, 0.1671, 0.174, 0.1706, 0.1775, 0.1915, 0.1915, 0.1775, 0.2124, 0.2263, 0.2367, 0.2124, 0.2263, 0.2333, 0.2228, 0.2402, 0.2681, 0.2577, 0.2367, 0.2298, 0.2542, 0.2437, 0.2298, 0.2402, 0.2646, 0.3099, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3343, 0.3238, 0.3691, 0.404, 0.39, 0.39, 0.3761, 0.3482, 0.2995, 0.3238, 0.3169, 0.2611, 0.289, 0.3517, 0.3134, 0.2995, 0.2855, 0.3072, 0.3317, 0.2932, 0.2862, 0.2652, 0.2652, 0.2897, 0.2862, 0.2722, 0.2443, 0.2303, 0.2967, 0.2652, 0.2652]}
{"unit_id": "quote:1978979134744465559", "ticker_idx": 0, "ts": "2025-10-17T00:19:28+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "U.S. Brokerage Firms Raise Target Prices ... major U.S. brokers revised their EPS estimates for 2025–2027 upward to $63.6 / $76 / $102.9, and simultaneously increased their target prices based on a 2026E PER of 22x", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses major U.S. brokers raising TSMC EPS estimates and target prices post Q3 beat, with CoWoS expansion and N2 margin upside.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "■ 2330 TSMC: U.S. Brokerage Firms Raise Target Prices\n\nTSMC’s Q3 gross profit margin (GPM) and EPS exceeded expectations from major U.S. securities firms, and its Q4 revenue and GPM outlook also surpassed market forecasts.\nAs a result, the company raised both its full-year revenue guidance and capital expenditure (CAPEX).\n\nFollowing the earnings release, major U.S. brokers revised their EPS estimates for 2025–2027 upward to $63.6 / $76 / $102.9, and simultaneously increased their target prices based on a 2026E PER of 22x.\n\nManagement stated that they plan to significantly expand CoWoS production capacity next year, noting that demand is broadening beyond AI-related applications.\n\nThe company emphasized its core strategy of “Moore’s Law 2.0,” highlighting that continuous progress in both front-end process technology and back-end packaging is essential.\nCurrently, advanced packaging accounts for about 10% of total revenue.\n\nBrokerages also project that wafer prices will rise by 3–5% in 2025, and assess that the profit margin structure of the N2 process will be superior to that of N3.", "reply_text": null, "parent_text": null, "quoted_text": "早安！10/17 外電綜合整理\n\n- 2330台積電: 美系升目標\n3Q 毛利率/EPS優於美系大行預估，4Q營收/毛利率展望亦優於預期，同時上調今年營收目標及資本支出。美系大行會後上調25-27年EPS至$63.6/$76/$102.9，以明年22xPE評價，同步升目標。\n\n管理層表示全力擴充明年的 CoWoS 產能，以縮小供給缺口。CoWoS", "parent_author": null, "ret_m1d": 0.02413800642620334, "ret_signed_m1d": 0.02413800642620334, "alpha_spy_m1d": 0.029782268259640476, "alpha_c_m1d": 0.022767081185159688, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020689744006068933, "ret_signed_p1d": 0.020689744006068933, "alpha_spy_p1d": 0.0102892686064473, "alpha_c_p1d": 0.00788450039784161, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.019356070768910216, "alpha_c_p1w": -0.024239561975410995, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": -0.003448262420134296, "ret_signed_p1m": -0.003448262420134296, "alpha_spy_p1m": -0.005374769127245327, "alpha_c_p1m": 0.003931499676608885, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.18324144972270506, "ret_signed_p3m": 0.18324144972270506, "alpha_spy_p3m": 0.14108262397159654, "alpha_c_p3m": 0.04693944220525226, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.38148115883175837, "ret_signed_p6m": 0.38148115883175837, "alpha_spy_p6m": 0.34292460128083513, "alpha_c_p6m": 0.08427987150322225, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.2238, -0.2135, -0.2135, -0.2135, -0.2135, -0.2204, -0.2066, -0.2032, -0.2032, -0.2204, -0.2101, -0.2272, -0.1894, -0.1929, -0.1894, -0.1894, -0.1757, -0.1929, -0.1894, -0.1894, -0.186, -0.2204, -0.2101, -0.2204, -0.1963, -0.1929, -0.1826, -0.2032, -0.2032, -0.1998, -0.2032, -0.2032, -0.2032, -0.1894, -0.1894, -0.1757, -0.1585, -0.1482, -0.1345, -0.1379, -0.1172, -0.1276, -0.1138, -0.1276, -0.1069, -0.0759, -0.0759, -0.0897, -0.1034, -0.1034, -0.1, -0.0862, -0.0586, -0.0345, -0.0345, -0.0103, -0.0241, -0.0069, -0.0069, -0.0241, -0.0172, 0.0103, 0.0241, 0.0, 0.0207, 0.0207, 0.0069, 0.0, 0.0, 0.0207, 0.0172, 0.0379, 0.0379, 0.0345, 0.0414, 0.0379, 0.0069, 0.0103, 0.0069, 0.0172, 0.0103, 0.0172, 0.0069, -0.0138, -0.0034, -0.031, -0.0379, 0.0034, -0.0448, -0.0517, -0.0241, -0.0069, -0.0103, -0.0069, -0.0276, -0.0138, 0.0, -0.0034, 0.0069, 0.031, 0.0207, 0.0379, 0.0172, 0.0241, 0.0033, -0.007, -0.0105, -0.0105, -0.0105, 0.0137, 0.031, 0.0345, 0.0345, 0.0449, 0.0587, 0.0518, 0.0725, 0.0725, 0.0967, 0.1556, 0.1798, 0.159, 0.1659, 0.1625, 0.1694, 0.1832, 0.1832, 0.1694, 0.204, 0.2178, 0.2282, 0.204, 0.2178, 0.2248, 0.2144, 0.2317, 0.2594, 0.249, 0.2282, 0.2213, 0.2455, 0.2351, 0.2213, 0.2317, 0.2559, 0.3009, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3251, 0.3147, 0.3597, 0.3943, 0.3804, 0.3804, 0.3666, 0.3389, 0.2905, 0.3147, 0.3078, 0.2524, 0.2801, 0.3424, 0.3043, 0.2905, 0.2767, 0.2982, 0.3225, 0.2843, 0.2773, 0.2565, 0.2565, 0.2808, 0.2773, 0.2635, 0.2357, 0.2218, 0.2878, 0.2565, 0.2565, 0.2565, 0.2912, 0.3537, 0.3572, 0.3884, 0.3815]}
{"unit_id": "quote:1988450320268272098", "ticker_idx": 0, "ts": "2025-11-12T03:34:34+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "It's still a long way before the cycle reaches its peak.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Author endorses Morgan Stanley's AI-driven memory super-cycle thesis; cycle still far from peak.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "“It’s still a long way before the cycle reaches its peak.” https://t.co/KGnSmxmaav", "reply_text": null, "parent_text": null, "quoted_text": "Morgan Stanley: This memory “super cycle” will far surpass the historical peak\n\nMorgan Stanley pointed out that a new memory “super cycle” driven by AI has already arrived, and its intensity and logic are completely different from any previous cycle.\n\nMorgan Stanley stated that, https://t.co/rstNpd3NdK", "parent_author": null, "ret_m1d": -0.0027848619212067862, "ret_signed_m1d": -0.0027848619212067862, "alpha_spy_m1d": -0.0022287727220201927, "alpha_c_m1d": 0.009816020695946856, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014491902517433672, "ret_signed_p1d": -0.014491902517433672, "alpha_spy_p1d": 0.0021021189533041804, "alpha_c_p1d": 0.015632081143825118, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.07688584208204279, "ret_signed_p1w": -0.07688584208204279, "alpha_spy_p1w": -0.046522168189060895, "alpha_c_p1w": -0.03162951966270093, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": 0.004161230922664812, "ret_signed_p1m": 0.004161230922664812, "alpha_spy_p1m": -0.004311337289656196, "alpha_c_p1m": -0.039041942800062825, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.5423405575076978, "ret_signed_p3m": 0.5423405575076978, "alpha_spy_p3m": 0.523872665438764, "alpha_c_p3m": 0.394996205681446, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6605957761227985, "ret_signed_p6m": 1.6605957761227985, "alpha_spy_p6m": 1.5839683731321148, "alpha_c_p6m": 1.1367284068593773, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.456, -0.4623, -0.4668, -0.4758, -0.4831, -0.4747, -0.4715, -0.4742, -0.4723, -0.465, -0.4684, -0.4822, -0.4757, -0.472, -0.462, -0.4498, -0.4459, -0.4303, -0.4116, -0.3931, -0.366, -0.3603, -0.3404, -0.3505, -0.3197, -0.3281, -0.3177, -0.306, -0.312, -0.3169, -0.3362, -0.3163, -0.3134, -0.2797, -0.2462, -0.2405, -0.2362, -0.2434, -0.2286, -0.2344, -0.2164, -0.2118, -0.2269, -0.2034, -0.164, -0.1565, -0.1391, -0.1507, -0.1509, -0.1481, -0.1069, -0.0816, -0.0948, -0.0651, -0.0517, -0.0459, 0.0136, -0.0475, -0.0387, -0.0345, -0.0463, -0.0026, -0.0028, 0.0, -0.0145, -0.0472, -0.018, -0.0648, -0.0769, -0.0922, -0.1298, -0.1016, -0.093, -0.0711, -0.0579, -0.0667, -0.056, -0.0381, -0.045, -0.0587, -0.0325, 0.0021, 0.0, 0.0255, 0.0042, -0.011, -0.0384, -0.0646, -0.0463, -0.0154, 0.0014, 0.0473, 0.0523, 0.0673, 0.0673, 0.0897, 0.1351, 0.1399, 0.1301, 0.1301, 0.213, 0.25, 0.3114, 0.3216, 0.3049, 0.3237, 0.3268, 0.3064, 0.3109, 0.3308, 0.3864, 0.392, 0.3704, 0.4156, 0.4443, 0.453, 0.4219, 0.5093, 0.575, 0.581, 0.5778, 0.5335, 0.6057, 0.5526, 0.4941, 0.5063, 0.5423, 0.5205, 0.5688, 0.624, 0.625, 0.625, 0.6088, 0.6376, 0.6689, 0.7137, 0.7144, 0.7623, 0.7957, 0.8693, 0.8395, 0.84, 0.6594, 0.565, 0.6728, 0.6161, 0.5466, 0.6674, 0.7046, 0.6661, 0.6693, 0.7422, 0.784, 0.8781, 0.805, 0.7691, 0.6612, 0.6889, 0.6732, 0.5745, 0.5769, 0.4852, 0.4416, 0.6044, 0.5292, 0.5796, 0.6231, 0.6502, 0.7989, 0.7789, 0.7998, 0.7987, 0.904, 0.9237, 0.9565, 0.9341, 0.9407, 0.9881, 1.0346, 1.0506, 1.0531, 1.1452, 1.1138, 1.1423, 1.1189, 1.153, 1.3259, 1.4128, 1.6412, 1.6606]}
{"unit_id": "thread:1977942315210768641", "ticker_idx": 0, "ts": "2025-10-14T03:39:30+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the earnings of Samsung Electronics (005930) and SK Hynix (000660) are projected to more than double from current levels", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares analyst call: Samsung and SK Hynix earnings to 2x+ by 2028 AI semi peak; HBM drives structural upcycle, no downturn in earnings or share prices until then.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“The peak of the AI semiconductor cycle will come in 2028… Samsung and SK Hynix’s earnings will double over the next three years”\n\nThe artificial intelligence (AI) semiconductor cycle is expected to reach its peak in 2028, and by then, the earnings of Samsung Electronics (005930) and SK Hynix (000660) are projected to more than double from current levels.\n\nAt a press briefing hosted by the Korea Exchange on the 14th, Noh Geun-chang, head of research at Hyundai Motor Securities, said, “Considering NVIDIA CEO Jensen Huang’s comment that data center investments will reach $1 trillion (about KRW 1,427 trillion) by 2028, it’s highly likely that AI semiconductor investments will peak around that time.”\n\nNoh continued, “As data center investments concentrate during this period, the earnings of domestic semiconductor firms have room to at least double over the next three years,” adding that transistor computing performance and HBM (High Bandwidth Memory) capacity will expand by about 2.5 times compared to today. He also stated, “Given the strong earnings visibility through 2028, current concerns over high valuations are not particularly worrisome.”\n\nHe added, “Semiconductor stocks are unlikely to see a downturn in either earnings or share prices until 2028,” and noted that “small and mid-cap semiconductor stocks will only see a rotation once the sector enters a breather phase.”\n\nWhile the semiconductor upturn this year has been driven by commodity DRAM, he forecast that from next year, HBM and advanced foundries will take center stage. He explained, “HBM is not a simple memory product but a made-to-order item, signaling the semiconductor industry’s shift from a ‘commodity’ business to a structurally growing industry,” emphasizing that “both Samsung and SK Hynix are positioned at the core of this cycle and will continue to achieve record-high earnings.”\n\nNoh further pointed out that collaborations among global players such as OpenAI, Broadcom, SoftBank, and Oracle are broadening HBM distribution channels, a development favorable to Samsung Electronics. “The AI ecosystem has so far been led by NVIDIA, TSMC, and SK Hynix,” he said, “but as OpenAI partners with SoftBank and Oracle, a new current is forming — one that opens fresh growth opportunities for Samsung.”\n\nFinally, he projected, “By next year, memory for AI servers will account for 70% of the entire DRAM market,” adding that “despite weakening consumer demand, structural growth centered on high-performance computing (HPC) will persist, and the expansion of inference services will drive a simultaneous recovery in demand for solid-state drives (SSDs).”\n\n---\n\nhttps://t.co/M4xma9ohfb", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018558947933519976, "ret_signed_m1d": 0.018558947933519976, "alpha_spy_m1d": 0.017335798864127394, "alpha_c_m1d": 0.005557693029785815, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01300125490373416, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03711789586703995, "ret_signed_p1d": 0.03711789586703995, "alpha_spy_p1d": 0.03267828000036066, "alpha_c_p1d": 0.027207156656559706, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.009910739210480246, "ret_p1w": 0.0644104562538288, "ret_signed_p1w": 0.0644104562538288, "alpha_spy_p1w": 0.05072946532140854, "alpha_c_p1w": 0.04014854923497735, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.024261907018851447, "ret_p1m": 0.1255458293080054, "ret_signed_p1m": 0.1255458293080054, "alpha_spy_p1m": 0.09360825403141115, "alpha_c_p1m": 0.08146211774975942, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.044083711558245975, "ret_p3m": 0.5248438699520155, "ret_signed_p3m": 0.5248438699520155, "alpha_spy_p3m": 0.4736669087563177, "alpha_c_p3m": 0.48494001836286205, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.03990385158915344, "ret_p6m": 1.313985622955864, "ret_signed_p6m": 1.313985622955864, "alpha_spy_p6m": 1.2873646121171467, "alpha_c_p6m": 1.3045625395104183, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.009423083445445624, "price_path": [-0.2751, -0.2707, -0.2631, -0.2827, -0.2783, -0.2827, -0.2838, -0.2349, -0.2327, -0.2109, -0.224, -0.2512, -0.2425, -0.2403, -0.2522, -0.2338, -0.2196, -0.2283, -0.2272, -0.2186, -0.2218, -0.2218, -0.2392, -0.2392, -0.2338, -0.2327, -0.224, -0.2229, -0.2359, -0.2327, -0.2435, -0.2425, -0.2653, -0.249, -0.2414, -0.2381, -0.2446, -0.2381, -0.2229, -0.2109, -0.2022, -0.1805, -0.1686, -0.137, -0.1501, -0.1273, -0.1273, -0.0925, -0.0794, -0.0718, -0.0642, -0.0947, -0.0808, -0.0841, -0.0611, -0.0202, -0.0202, -0.0202, -0.0202, -0.0202, -0.0202, 0.0306, 0.0186, 0.0, 0.0371, 0.0666, 0.0688, 0.071, 0.0644, 0.0764, 0.0535, 0.0786, 0.1135, 0.0862, 0.0972, 0.1365, 0.1736, 0.2129, 0.1452, 0.0983, 0.083, 0.0688, 0.0983, 0.1299, 0.1255, 0.1223, 0.0611, 0.0983, 0.0677, 0.0535, 0.0983, 0.0349, 0.0557, 0.0841, 0.1223, 0.1299, 0.0972, 0.1004, 0.1288, 0.1408, 0.1474, 0.1834, 0.1954, 0.1834, 0.179, 0.1714, 0.1889, 0.1441, 0.1223, 0.1779, 0.1747, 0.1605, 0.2063, 0.2172, 0.2129, 0.2129, 0.2773, 0.3109, 0.3153, 0.3153, 0.3153, 0.4097, 0.515, 0.5237, 0.5468, 0.5226, 0.5248, 0.5226, 0.5095, 0.5391, 0.5786, 0.6334, 0.6378, 0.5929, 0.64, 0.6707, 0.6686, 0.6686, 0.7497, 0.7815, 0.7629, 0.7607, 0.6499, 0.8375, 0.855, 0.7475, 0.7399, 0.8254, 0.8188, 0.8408, 0.9593, 0.9878, 0.9878, 0.9878, 0.9878, 1.0843, 1.0854, 1.1172, 1.194, 1.2324, 1.3915, 1.375, 1.375, 1.1403, 0.8891, 1.1019, 1.0646, 0.9033, 1.0613, 1.0843, 1.0613, 1.013, 1.0701, 1.1271, 1.2873, 1.1995, 1.1874, 1.0437, 1.081, 1.0733, 0.9757, 0.9757, 0.938, 0.838, 1.0848, 0.9611, 1.0469, 1.1227, 1.1601, 1.314]}
{"unit_id": "thread:1977942315210768641", "ticker_idx": 1, "ts": "2025-10-14T03:39:30+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the earnings of Samsung Electronics (005930) and SK Hynix (000660) are projected to more than double from current levels", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares analyst call: Samsung and SK Hynix earnings to 2x+ by 2028 AI semi peak; HBM drives structural upcycle, no downturn in earnings or share prices until then.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“The peak of the AI semiconductor cycle will come in 2028… Samsung and SK Hynix’s earnings will double over the next three years”\n\nThe artificial intelligence (AI) semiconductor cycle is expected to reach its peak in 2028, and by then, the earnings of Samsung Electronics (005930) and SK Hynix (000660) are projected to more than double from current levels.\n\nAt a press briefing hosted by the Korea Exchange on the 14th, Noh Geun-chang, head of research at Hyundai Motor Securities, said, “Considering NVIDIA CEO Jensen Huang’s comment that data center investments will reach $1 trillion (about KRW 1,427 trillion) by 2028, it’s highly likely that AI semiconductor investments will peak around that time.”\n\nNoh continued, “As data center investments concentrate during this period, the earnings of domestic semiconductor firms have room to at least double over the next three years,” adding that transistor computing performance and HBM (High Bandwidth Memory) capacity will expand by about 2.5 times compared to today. He also stated, “Given the strong earnings visibility through 2028, current concerns over high valuations are not particularly worrisome.”\n\nHe added, “Semiconductor stocks are unlikely to see a downturn in either earnings or share prices until 2028,” and noted that “small and mid-cap semiconductor stocks will only see a rotation once the sector enters a breather phase.”\n\nWhile the semiconductor upturn this year has been driven by commodity DRAM, he forecast that from next year, HBM and advanced foundries will take center stage. He explained, “HBM is not a simple memory product but a made-to-order item, signaling the semiconductor industry’s shift from a ‘commodity’ business to a structurally growing industry,” emphasizing that “both Samsung and SK Hynix are positioned at the core of this cycle and will continue to achieve record-high earnings.”\n\nNoh further pointed out that collaborations among global players such as OpenAI, Broadcom, SoftBank, and Oracle are broadening HBM distribution channels, a development favorable to Samsung Electronics. “The AI ecosystem has so far been led by NVIDIA, TSMC, and SK Hynix,” he said, “but as OpenAI partners with SoftBank and Oracle, a new current is forming — one that opens fresh growth opportunities for Samsung.”\n\nFinally, he projected, “By next year, memory for AI servers will account for 70% of the entire DRAM market,” adding that “despite weakening consumer demand, structural growth centered on high-performance computing (HPC) will persist, and the expansion of inference services will drive a simultaneous recovery in demand for solid-state drives (SSDs).”\n\n---\n\nhttps://t.co/M4xma9ohfb", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008505465672627333, "ret_signed_m1d": 0.008505465672627333, "alpha_spy_m1d": 0.007282316603234751, "alpha_c_m1d": -0.010235096992196047, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026731507056273474, "ret_signed_p1d": 0.026731507056273474, "alpha_spy_p1d": 0.022291891189594182, "alpha_c_p1d": 0.0019039607312922868, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.1640339917076754, "ret_signed_p1w": 0.1640339917076754, "alpha_spy_p1w": 0.15035300077525515, "alpha_c_p1w": 0.12865177561713947, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.49939248305531825, "ret_signed_p1m": 0.49939248305531825, "alpha_spy_p1m": 0.467454907778724, "alpha_c_p1m": 0.4333355887176735, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.8093143054173029, "ret_signed_p3m": 0.8093143054173029, "alpha_spy_p3m": 0.7581373442216051, "alpha_c_p3m": 0.6386064169295438, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 1.516761467129815, "ret_signed_p6m": 1.516761467129815, "alpha_spy_p6m": 1.4901404562910978, "alpha_c_p6m": 1.2446896456243026, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.346, -0.3472, -0.3387, -0.3485, -0.3472, -0.346, -0.3545, -0.3642, -0.363, -0.3606, -0.3363, -0.3739, -0.3739, -0.3606, -0.3727, -0.3642, -0.3776, -0.3521, -0.3472, -0.3254, -0.329, -0.329, -0.3509, -0.3618, -0.38, -0.4055, -0.3909, -0.3703, -0.3654, -0.3691, -0.3475, -0.3463, -0.3779, -0.367, -0.3621, -0.3548, -0.3354, -0.3269, -0.3001, -0.2612, -0.2539, -0.2017, -0.1956, -0.1543, -0.1896, -0.1361, -0.1361, -0.147, -0.1227, -0.1312, -0.1337, -0.1823, -0.1519, -0.1555, -0.1252, -0.0389, -0.0389, -0.0389, -0.0389, -0.0389, -0.0389, 0.0401, 0.0085, 0.0, 0.0267, 0.0996, 0.1312, 0.1798, 0.164, 0.1701, 0.1628, 0.2394, 0.3001, 0.2661, 0.356, 0.3803, 0.3584, 0.5067, 0.4241, 0.407, 0.4411, 0.4095, 0.4727, 0.5043, 0.4994, 0.4872, 0.3609, 0.4727, 0.3852, 0.3657, 0.3876, 0.2661, 0.2637, 0.2612, 0.2734, 0.3229, 0.2889, 0.3083, 0.357, 0.3424, 0.3181, 0.3229, 0.4032, 0.3764, 0.4275, 0.374, 0.3886, 0.3473, 0.2889, 0.34, 0.3424, 0.3302, 0.4105, 0.4202, 0.4299, 0.4299, 0.4567, 0.5564, 0.5832, 0.5832, 0.5832, 0.6464, 0.6926, 0.7655, 0.8045, 0.8385, 0.8093, 0.8215, 0.7947, 0.8045, 0.8215, 0.8385, 0.858, 0.8069, 0.7996, 0.8361, 0.8652, 0.7899, 0.9455, 1.0452, 1.0938, 1.2106, 1.0185, 1.2057, 1.1887, 1.0476, 1.0403, 1.1571, 1.1303, 1.0914, 1.1595, 1.14, 1.14, 1.14, 1.14, 1.1741, 1.3078, 1.3127, 1.444, 1.4756, 1.6776, 1.585, 1.585, 1.2877, 1.0685, 1.2926, 1.2512, 1.0368, 1.2853, 1.3267, 1.2658, 1.2171, 1.373, 1.3633, 1.5728, 1.468, 1.4534, 1.2731, 1.4023, 1.4242, 1.2731, 1.2731, 1.1269, 0.9661, 1.1854, 1.0222, 1.1343, 1.1586, 1.2317, 1.5168]}
{"unit_id": "orig:1988211005940507039", "ticker_idx": 0, "ts": "2025-11-11T11:43:37+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA (undervalued, EPS ~$8 in 2026 at 24x P/E), AVGO, AMD, LRCX, KLAC, AMAT; AI skepticism seen as contrarian positive, not fundamental risk.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: The “AI bubble” narrative is actually a positive for semiconductor stocks, helps avoid overcrowding, and Nvidia is undervalued\n\nOn the recent “AI bubble” narrative, Bank of America presented a clear contrarian view: it is a “reverse positive” for semiconductor stocks, helping to prevent excessive crowding. There is a clear undervaluation in Nvidia, with the share price failing to reflect its true growth potential.\n\nIn its latest report, Bank of America stated that the widespread skepticism about AI capital expenditures at this point is not a risk but rather a “reverse positive” that ensures this track can deliver sustained returns. Such skepticism helps prevent excessive crowding in the sector.\n\nThe bank believes that the recent volatility in AI semiconductor stocks stems from fixable macro factors rather than a deterioration in the fundamentals of the AI spending cycle. In fact, given the strong performance of AI-related complementary industries and Nvidia’s optimistic outlook, underlying demand remains solid.\n\nBank of America added that doubts about OpenAI’s grand plans are a narrow and fragmentary argument. What truly drives AI investment is the defensive strategy of large tech companies. In particular, it pointed out that the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued.\n\nAI skepticism: a healthy “contrarian positive” signal\nAlthough large AI semiconductor stocks fell by an average of 7–8% last week, Bank of America does not see this as a sign of problems in the AI spending cycle. Rather, the price moves were triggered by macro factors—concerns about a U.S. government shutdown, weak employment data, tariff turmoil, and misunderstandings of comments related to OpenAI—all of which are fixable short-term noise, not fundamental negatives.\n\nThe real fundamental signals are quite the opposite. The report highlighted the strength of AI-related complementary areas—for example, memory and small-/mid-cap optical module stocks rose 14% last week—indicating that AI infrastructure build-out is rolling out across the board.\n\nIt also assessed that Nvidia’s recently disclosed outlook at GTC for its data center business in fiscal years 2025/26—exceeding $500 billion—further proves that AI demand is strong and persistent.\n\nTherefore, Bank of America believes that widespread skepticism actually helps keep the market cool-headed, and it reaffirmed its “Buy” ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers, and Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment.\n\nRebutting the “AI spending is unsustainable” critique: defensive investment by Big Tech is key\nA common bearish narrative in the market is: “If OpenAI can’t justify its $1.4 trillion long-term commitment, then AI stocks must be overvalued.” Bank of America dismisses this as a “lazy and out-of-context argument.”\n\nFirst, OpenAI’s grand plan has not yet materialized and will be constrained by real-world factors such as power and data center space. More importantly, most current AI spending is not coming from startups but from profitable, publicly listed hyperscalers.\n\nFor these tech giants, Bank of America views upgrading from traditional CPU computing to accelerated computing as not only “mission-critical” but also a “defensive” investment.\n\nThe report cites Google as an example: its $92 billion in capital expenditures is “defending” the leadership of a search business that generates over $200 billion, preventing users from shifting to emerging competitors such as ChatGPT and Perplexity.\n\nAt the same time, private AI companies such as OpenAI (with over 1 million commercial customers) and Anthropic (with over 300,000 commercial customers) are rapidly attracting enterprise users, which in turn continues to pressure publicly listed software and IaaS providers to increase AI investment. Therefore, the core drivers of AI spending are solid and powerful.\n\nNvidia: not a bubble, but undervalued\nThe report emphasizes that Nvidia’s stock is highly attractive, and the current valuation reflects only a “very cautious AI deployment.”\n\nBased on approximately $500 billion in data center orders for fiscal years 2025/26 recently disclosed by Nvidia, Bank of America estimates the company’s EPS could reach about $8 in 2026.\n\nThis implies that a company with revenue/EPS growing 50%/70% annually is trading at a P/E of just 24x—roughly in line with the market average—meaning the valuation is “not high.”\n\nTo further illustrate its potential, the report presents a scenario assumption:\n\nEven if by 2030 AI capital expenditures reach only 50% of Nvidia’s projected $3–4 trillion range, the company’s EPS at that time could exceed $40 per share. This would imply a forward P/E of under 5x at the current share price.\n\nBank of America regards this number as “beyond common sense,” using it to reaffirm the following view: despite media headlines being filled with bubble narratives, Nvidia’s stock is in fact undervalued.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030492882347475936, "ret_signed_m1d": 0.030492882347475936, "alpha_spy_m1d": 0.03277680381178116, "alpha_c_m1d": 0.00844470525187968, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003313360000960719, "ret_signed_p1d": 0.003313360000960719, "alpha_spy_p1d": 0.002756961394518642, "alpha_c_p1d": -0.00944833118887245, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.061089260286197655, "ret_signed_p1w": -0.061089260286197655, "alpha_spy_p1w": -0.02753153112227391, "alpha_c_p1w": -0.010498217267669863, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": -0.04850783058703967, "ret_signed_p1m": -0.04850783058703967, "alpha_spy_p1m": -0.055198993411596176, "alpha_c_p1m": -0.11416803762039751, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": -0.04006875267429444, "ret_signed_p3m": -0.04006875267429444, "alpha_spy_p3m": -0.05421328419987903, "alpha_c_p3m": -0.1877678549390902, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.07606554636507679, "ret_signed_p6m": 0.07606554636507679, "alpha_spy_p6m": -0.004473938298911406, "alpha_c_p6m": -0.4948521164189732, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.0577, -0.0659, -0.0578, -0.0908, -0.092, -0.0942, -0.0786, -0.0692, -0.059, -0.0599, -0.0673, -0.0983, -0.0983, -0.1159, -0.1167, -0.1114, -0.1354, -0.1287, -0.116, -0.082, -0.0828, -0.0794, -0.0798, -0.0946, -0.1184, -0.0876, -0.0854, -0.0494, -0.0763, -0.0838, -0.0801, -0.0775, -0.0586, -0.0341, -0.0306, -0.0221, -0.0287, -0.0394, -0.042, -0.021, -0.0031, -0.0518, -0.0251, -0.068, -0.069, -0.0588, -0.0515, -0.0545, -0.0621, -0.0667, -0.0569, -0.0357, -0.0086, 0.0407, 0.0719, 0.0504, 0.0483, 0.071, 0.0286, 0.0106, -0.0263, -0.0259, 0.0305, 0.0, 0.0033, -0.0326, -0.0155, -0.034, -0.0611, -0.0344, -0.0648, -0.0739, -0.0549, -0.0794, -0.0668, -0.0668, -0.0837, -0.0685, -0.0606, -0.0703, -0.0506, -0.0556, -0.0393, -0.0423, -0.0485, -0.0633, -0.0939, -0.0873, -0.0799, -0.115, -0.0984, -0.063, -0.049, -0.0204, -0.0235, -0.0235, -0.0136, -0.0255, -0.029, -0.0344, -0.0344, -0.0223, -0.026, -0.0306, -0.0209, -0.042, -0.0429, -0.0425, -0.038, -0.0518, -0.0316, -0.0358, -0.0358, -0.0781, -0.0509, -0.043, -0.0284, -0.0346, -0.024, -0.0084, -0.0033, -0.0105, -0.039, -0.0663, -0.0982, -0.1101, -0.0401, -0.0161, -0.0239, -0.016, -0.0321, -0.0535, -0.0535, -0.0423, -0.0268, -0.0272, -0.0172, -0.0083, -0.0015, 0.0125, -0.0428, -0.0826, -0.0552, -0.0678, -0.0523, -0.0508, -0.0794, -0.0544, -0.0434, -0.0368, -0.0518, -0.0667, -0.0514, -0.058, -0.066, -0.0755, -0.1058, -0.0906, -0.0929, -0.0749, -0.1134, -0.1326, -0.1448, -0.097, -0.09, -0.0815, -0.0815, -0.0802, -0.0779, -0.0573, -0.0478, -0.0233, -0.0198, 0.0175, 0.0297, 0.027, 0.0442, 0.0462, 0.0349, 0.0485, 0.0337, 0.0783, 0.1215, 0.1037, 0.0834, 0.0333, 0.0275, 0.0277, 0.0174, 0.0761]}
{"unit_id": "orig:1988211005940507039", "ticker_idx": 1, "ts": "2025-11-11T11:43:37+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reaffirmed its \"Buy\" ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA (undervalued, EPS ~$8 in 2026 at 24x P/E), AVGO, AMD, LRCX, KLAC, AMAT; AI skepticism seen as contrarian positive, not fundamental risk.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: The “AI bubble” narrative is actually a positive for semiconductor stocks, helps avoid overcrowding, and Nvidia is undervalued\n\nOn the recent “AI bubble” narrative, Bank of America presented a clear contrarian view: it is a “reverse positive” for semiconductor stocks, helping to prevent excessive crowding. There is a clear undervaluation in Nvidia, with the share price failing to reflect its true growth potential.\n\nIn its latest report, Bank of America stated that the widespread skepticism about AI capital expenditures at this point is not a risk but rather a “reverse positive” that ensures this track can deliver sustained returns. Such skepticism helps prevent excessive crowding in the sector.\n\nThe bank believes that the recent volatility in AI semiconductor stocks stems from fixable macro factors rather than a deterioration in the fundamentals of the AI spending cycle. In fact, given the strong performance of AI-related complementary industries and Nvidia’s optimistic outlook, underlying demand remains solid.\n\nBank of America added that doubts about OpenAI’s grand plans are a narrow and fragmentary argument. What truly drives AI investment is the defensive strategy of large tech companies. In particular, it pointed out that the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued.\n\nAI skepticism: a healthy “contrarian positive” signal\nAlthough large AI semiconductor stocks fell by an average of 7–8% last week, Bank of America does not see this as a sign of problems in the AI spending cycle. Rather, the price moves were triggered by macro factors—concerns about a U.S. government shutdown, weak employment data, tariff turmoil, and misunderstandings of comments related to OpenAI—all of which are fixable short-term noise, not fundamental negatives.\n\nThe real fundamental signals are quite the opposite. The report highlighted the strength of AI-related complementary areas—for example, memory and small-/mid-cap optical module stocks rose 14% last week—indicating that AI infrastructure build-out is rolling out across the board.\n\nIt also assessed that Nvidia’s recently disclosed outlook at GTC for its data center business in fiscal years 2025/26—exceeding $500 billion—further proves that AI demand is strong and persistent.\n\nTherefore, Bank of America believes that widespread skepticism actually helps keep the market cool-headed, and it reaffirmed its “Buy” ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers, and Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment.\n\nRebutting the “AI spending is unsustainable” critique: defensive investment by Big Tech is key\nA common bearish narrative in the market is: “If OpenAI can’t justify its $1.4 trillion long-term commitment, then AI stocks must be overvalued.” Bank of America dismisses this as a “lazy and out-of-context argument.”\n\nFirst, OpenAI’s grand plan has not yet materialized and will be constrained by real-world factors such as power and data center space. More importantly, most current AI spending is not coming from startups but from profitable, publicly listed hyperscalers.\n\nFor these tech giants, Bank of America views upgrading from traditional CPU computing to accelerated computing as not only “mission-critical” but also a “defensive” investment.\n\nThe report cites Google as an example: its $92 billion in capital expenditures is “defending” the leadership of a search business that generates over $200 billion, preventing users from shifting to emerging competitors such as ChatGPT and Perplexity.\n\nAt the same time, private AI companies such as OpenAI (with over 1 million commercial customers) and Anthropic (with over 300,000 commercial customers) are rapidly attracting enterprise users, which in turn continues to pressure publicly listed software and IaaS providers to increase AI investment. Therefore, the core drivers of AI spending are solid and powerful.\n\nNvidia: not a bubble, but undervalued\nThe report emphasizes that Nvidia’s stock is highly attractive, and the current valuation reflects only a “very cautious AI deployment.”\n\nBased on approximately $500 billion in data center orders for fiscal years 2025/26 recently disclosed by Nvidia, Bank of America estimates the company’s EPS could reach about $8 in 2026.\n\nThis implies that a company with revenue/EPS growing 50%/70% annually is trading at a P/E of just 24x—roughly in line with the market average—meaning the valuation is “not high.”\n\nTo further illustrate its potential, the report presents a scenario assumption:\n\nEven if by 2030 AI capital expenditures reach only 50% of Nvidia’s projected $3–4 trillion range, the company’s EPS at that time could exceed $40 per share. This would imply a forward P/E of under 5x at the current share price.\n\nBank of America regards this number as “beyond common sense,” using it to reaffirm the following view: despite media headlines being filled with bubble narratives, Nvidia’s stock is in fact undervalued.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01826923947479675, "ret_signed_m1d": 0.01826923947479675, "alpha_spy_m1d": 0.020553160939101978, "alpha_c_m1d": -0.003778937620799505, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009262460972947917, "ret_signed_p1d": 0.009262460972947917, "alpha_spy_p1d": 0.00870606236650584, "alpha_c_p1d": -0.003499230216885252, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.03256055765765953, "ret_signed_p1w": -0.03256055765765953, "alpha_spy_p1w": 0.0009971715062642161, "alpha_c_p1w": 0.018030485360868265, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": 0.17334357456108274, "ret_signed_p1m": 0.17334357456108274, "alpha_spy_p1m": 0.16665241173652623, "alpha_c_p1m": 0.1076833675277249, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": -0.05228710886762877, "ret_signed_p3m": -0.05228710886762877, "alpha_spy_p3m": -0.06643164039321336, "alpha_c_p3m": -0.19998621113242454, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.21362710091422366, "ret_signed_p6m": 0.21362710091422366, "alpha_spy_p6m": 0.13308761625023546, "alpha_c_p6m": -0.35729056186982633, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.1172, -0.1311, -0.1328, -0.1635, -0.1741, -0.1786, -0.1661, -0.1655, -0.1547, -0.1484, -0.1246, -0.1565, -0.1565, -0.1541, -0.1423, -0.1318, -0.0501, -0.0196, -0.0451, 0.0482, 0.02, 0.0207, 0.0327, 0.0211, -0.0181, -0.0205, -0.0216, -0.0374, -0.037, -0.0359, -0.0451, -0.0495, -0.0684, -0.0626, -0.0528, -0.0392, -0.0386, -0.0468, -0.0442, -0.0184, -0.0197, -0.0777, 0.0135, -0.0222, -0.0018, 0.0062, -0.0075, -0.0077, -0.0264, -0.0331, -0.0218, 0.0062, 0.0287, 0.0597, 0.0967, 0.0696, 0.0502, 0.0301, -0.0001, 0.0199, 0.0103, -0.0072, 0.0183, 0.0, 0.0093, -0.034, -0.027, -0.0265, -0.0326, 0.007, -0.0146, -0.0334, 0.0739, 0.094, 0.1296, 0.1296, 0.1449, 0.0969, 0.0841, 0.0814, 0.0826, 0.1088, 0.1396, 0.1544, 0.1733, 0.1546, 0.0226, -0.0345, -0.0303, -0.0737, -0.0627, -0.033, -0.028, -0.0056, -0.003, -0.003, 0.0024, -0.0054, -0.0041, -0.0148, -0.0148, -0.0104, -0.0224, -0.0214, -0.0222, -0.0535, -0.018, 0.0026, 0.0095, -0.0324, -0.0235, 0.0012, 0.0012, -0.0532, -0.064, -0.0734, -0.0889, -0.0753, -0.0527, -0.0514, -0.0585, -0.0569, -0.0574, -0.0881, -0.1231, -0.1161, -0.0523, -0.0209, -0.0309, -0.0243, -0.0573, -0.0743, -0.0743, -0.0534, -0.0506, -0.0492, -0.0531, -0.0596, -0.0734, -0.054, -0.0842, -0.0903, -0.0924, -0.1066, -0.0961, -0.0527, -0.0592, -0.0158, -0.0248, -0.0277, -0.0436, -0.0829, -0.0751, -0.0853, -0.1007, -0.0895, -0.1161, -0.08, -0.092, -0.0905, -0.1173, -0.1423, -0.163, -0.1171, -0.1057, -0.1027, -0.1027, -0.103, -0.0473, 0.0002, 0.0124, 0.0599, 0.0833, 0.0862, 0.1317, 0.1367, 0.1597, 0.14, 0.1472, 0.2057, 0.1979, 0.206, 0.193, 0.1406, 0.1566, 0.1908, 0.2018, 0.1881, 0.2191, 0.2136]}
{"unit_id": "orig:1988211005940507039", "ticker_idx": 2, "ts": "2025-11-11T11:43:37+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reaffirmed its \"Buy\" ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA (undervalued, EPS ~$8 in 2026 at 24x P/E), AVGO, AMD, LRCX, KLAC, AMAT; AI skepticism seen as contrarian positive, not fundamental risk.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: The “AI bubble” narrative is actually a positive for semiconductor stocks, helps avoid overcrowding, and Nvidia is undervalued\n\nOn the recent “AI bubble” narrative, Bank of America presented a clear contrarian view: it is a “reverse positive” for semiconductor stocks, helping to prevent excessive crowding. There is a clear undervaluation in Nvidia, with the share price failing to reflect its true growth potential.\n\nIn its latest report, Bank of America stated that the widespread skepticism about AI capital expenditures at this point is not a risk but rather a “reverse positive” that ensures this track can deliver sustained returns. Such skepticism helps prevent excessive crowding in the sector.\n\nThe bank believes that the recent volatility in AI semiconductor stocks stems from fixable macro factors rather than a deterioration in the fundamentals of the AI spending cycle. In fact, given the strong performance of AI-related complementary industries and Nvidia’s optimistic outlook, underlying demand remains solid.\n\nBank of America added that doubts about OpenAI’s grand plans are a narrow and fragmentary argument. What truly drives AI investment is the defensive strategy of large tech companies. In particular, it pointed out that the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued.\n\nAI skepticism: a healthy “contrarian positive” signal\nAlthough large AI semiconductor stocks fell by an average of 7–8% last week, Bank of America does not see this as a sign of problems in the AI spending cycle. Rather, the price moves were triggered by macro factors—concerns about a U.S. government shutdown, weak employment data, tariff turmoil, and misunderstandings of comments related to OpenAI—all of which are fixable short-term noise, not fundamental negatives.\n\nThe real fundamental signals are quite the opposite. The report highlighted the strength of AI-related complementary areas—for example, memory and small-/mid-cap optical module stocks rose 14% last week—indicating that AI infrastructure build-out is rolling out across the board.\n\nIt also assessed that Nvidia’s recently disclosed outlook at GTC for its data center business in fiscal years 2025/26—exceeding $500 billion—further proves that AI demand is strong and persistent.\n\nTherefore, Bank of America believes that widespread skepticism actually helps keep the market cool-headed, and it reaffirmed its “Buy” ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers, and Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment.\n\nRebutting the “AI spending is unsustainable” critique: defensive investment by Big Tech is key\nA common bearish narrative in the market is: “If OpenAI can’t justify its $1.4 trillion long-term commitment, then AI stocks must be overvalued.” Bank of America dismisses this as a “lazy and out-of-context argument.”\n\nFirst, OpenAI’s grand plan has not yet materialized and will be constrained by real-world factors such as power and data center space. More importantly, most current AI spending is not coming from startups but from profitable, publicly listed hyperscalers.\n\nFor these tech giants, Bank of America views upgrading from traditional CPU computing to accelerated computing as not only “mission-critical” but also a “defensive” investment.\n\nThe report cites Google as an example: its $92 billion in capital expenditures is “defending” the leadership of a search business that generates over $200 billion, preventing users from shifting to emerging competitors such as ChatGPT and Perplexity.\n\nAt the same time, private AI companies such as OpenAI (with over 1 million commercial customers) and Anthropic (with over 300,000 commercial customers) are rapidly attracting enterprise users, which in turn continues to pressure publicly listed software and IaaS providers to increase AI investment. Therefore, the core drivers of AI spending are solid and powerful.\n\nNvidia: not a bubble, but undervalued\nThe report emphasizes that Nvidia’s stock is highly attractive, and the current valuation reflects only a “very cautious AI deployment.”\n\nBased on approximately $500 billion in data center orders for fiscal years 2025/26 recently disclosed by Nvidia, Bank of America estimates the company’s EPS could reach about $8 in 2026.\n\nThis implies that a company with revenue/EPS growing 50%/70% annually is trading at a P/E of just 24x—roughly in line with the market average—meaning the valuation is “not high.”\n\nTo further illustrate its potential, the report presents a scenario assumption:\n\nEven if by 2030 AI capital expenditures reach only 50% of Nvidia’s projected $3–4 trillion range, the company’s EPS at that time could exceed $40 per share. This would imply a forward P/E of under 5x at the current share price.\n\nBank of America regards this number as “beyond common sense,” using it to reaffirm the following view: despite media headlines being filled with bubble narratives, Nvidia’s stock is in fact undervalued.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02719767320173938, "ret_signed_m1d": 0.02719767320173938, "alpha_spy_m1d": 0.029481594666044608, "alpha_c_m1d": 0.0051494961061431255, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08997141289818633, "ret_signed_p1d": 0.08997141289818633, "alpha_spy_p1d": 0.08941501429174425, "alpha_c_p1d": 0.07720972170835316, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.030439587640098464, "ret_signed_p1w": -0.030439587640098464, "alpha_spy_p1w": 0.0031181415238252796, "alpha_c_p1w": 0.020151455378429328, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": -0.06778379005520385, "ret_signed_p1m": -0.06778379005520385, "alpha_spy_p1m": -0.07447495287976036, "alpha_c_p1m": -0.1334439970885617, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": -0.1224318007240216, "ret_signed_p3m": -0.1224318007240216, "alpha_spy_p3m": -0.1365763322496062, "alpha_c_p3m": -0.27013090298881737, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.7741243140306528, "ret_signed_p6m": 0.7741243140306528, "alpha_spy_p6m": 0.6935848293666647, "alpha_c_p6m": 0.20320665124660287, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.2382, -0.2527, -0.2584, -0.2988, -0.3045, -0.3108, -0.2937, -0.3122, -0.2985, -0.2964, -0.2902, -0.3153, -0.3153, -0.3166, -0.3174, -0.3188, -0.3637, -0.3625, -0.344, -0.3283, -0.3446, -0.3324, -0.3215, -0.3244, -0.3299, -0.3351, -0.3374, -0.3273, -0.3226, -0.3227, -0.321, -0.3286, -0.3206, -0.3188, -0.3095, -0.2854, -0.3067, -0.1423, -0.1095, -0.0083, -0.0195, -0.0952, -0.0888, -0.0818, 0.0045, -0.0125, -0.0187, 0.0128, 0.0021, -0.0307, -0.0107, 0.0648, 0.0933, 0.0863, 0.1129, 0.0729, 0.0783, 0.0932, 0.0528, 0.0792, 0.0008, -0.0168, 0.0272, 0.0, 0.09, 0.044, 0.0391, 0.0126, -0.0304, -0.0588, -0.1326, -0.1421, -0.0946, -0.1322, -0.098, -0.098, -0.0842, -0.0748, -0.0938, -0.0839, -0.0907, -0.0823, -0.0691, -0.0669, -0.0678, -0.0677, -0.1126, -0.1261, -0.1194, -0.1659, -0.1535, -0.1014, -0.095, -0.0952, -0.0946, -0.0946, -0.0949, -0.0922, -0.0934, -0.0983, -0.0983, -0.0592, -0.0692, -0.0975, -0.1158, -0.1383, -0.1446, -0.1256, -0.0697, -0.0586, -0.0404, -0.024, -0.024, -0.0236, 0.0517, 0.0682, 0.0933, 0.0581, 0.0611, 0.0641, 0.0617, -0.0033, 0.0368, 0.0193, -0.1572, -0.1895, -0.1224, -0.0906, -0.1008, -0.1008, -0.133, -0.1271, -0.1271, -0.145, -0.1575, -0.1438, -0.1573, -0.1723, -0.0997, -0.1122, -0.1425, -0.1571, -0.1638, -0.1961, -0.1493, -0.1603, -0.1898, -0.1467, -0.1444, -0.1376, -0.1675, -0.1858, -0.1724, -0.1735, -0.1602, -0.1358, -0.1524, -0.1467, -0.1354, -0.0726, -0.1421, -0.1496, -0.1746, -0.1435, -0.115, -0.0843, -0.0843, -0.073, -0.0673, -0.024, -0.0037, 0.0317, 0.0392, 0.0739, 0.0867, 0.1715, 0.1721, 0.1576, 0.1978, 0.2776, 0.2855, 0.4643, 0.4088, 0.3608, 0.4193, 0.4925, 0.5179, 0.4379, 0.4957, 0.7741]}
{"unit_id": "orig:1988211005940507039", "ticker_idx": 3, "ts": "2025-11-11T11:43:37+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reaffirmed its \"Buy\" ratings on the core beneficiaries of AI build-out. These include … Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA (undervalued, EPS ~$8 in 2026 at 24x P/E), AVGO, AMD, LRCX, KLAC, AMAT; AI skepticism seen as contrarian positive, not fundamental risk.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: The “AI bubble” narrative is actually a positive for semiconductor stocks, helps avoid overcrowding, and Nvidia is undervalued\n\nOn the recent “AI bubble” narrative, Bank of America presented a clear contrarian view: it is a “reverse positive” for semiconductor stocks, helping to prevent excessive crowding. There is a clear undervaluation in Nvidia, with the share price failing to reflect its true growth potential.\n\nIn its latest report, Bank of America stated that the widespread skepticism about AI capital expenditures at this point is not a risk but rather a “reverse positive” that ensures this track can deliver sustained returns. Such skepticism helps prevent excessive crowding in the sector.\n\nThe bank believes that the recent volatility in AI semiconductor stocks stems from fixable macro factors rather than a deterioration in the fundamentals of the AI spending cycle. In fact, given the strong performance of AI-related complementary industries and Nvidia’s optimistic outlook, underlying demand remains solid.\n\nBank of America added that doubts about OpenAI’s grand plans are a narrow and fragmentary argument. What truly drives AI investment is the defensive strategy of large tech companies. In particular, it pointed out that the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued.\n\nAI skepticism: a healthy “contrarian positive” signal\nAlthough large AI semiconductor stocks fell by an average of 7–8% last week, Bank of America does not see this as a sign of problems in the AI spending cycle. Rather, the price moves were triggered by macro factors—concerns about a U.S. government shutdown, weak employment data, tariff turmoil, and misunderstandings of comments related to OpenAI—all of which are fixable short-term noise, not fundamental negatives.\n\nThe real fundamental signals are quite the opposite. The report highlighted the strength of AI-related complementary areas—for example, memory and small-/mid-cap optical module stocks rose 14% last week—indicating that AI infrastructure build-out is rolling out across the board.\n\nIt also assessed that Nvidia’s recently disclosed outlook at GTC for its data center business in fiscal years 2025/26—exceeding $500 billion—further proves that AI demand is strong and persistent.\n\nTherefore, Bank of America believes that widespread skepticism actually helps keep the market cool-headed, and it reaffirmed its “Buy” ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers, and Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment.\n\nRebutting the “AI spending is unsustainable” critique: defensive investment by Big Tech is key\nA common bearish narrative in the market is: “If OpenAI can’t justify its $1.4 trillion long-term commitment, then AI stocks must be overvalued.” Bank of America dismisses this as a “lazy and out-of-context argument.”\n\nFirst, OpenAI’s grand plan has not yet materialized and will be constrained by real-world factors such as power and data center space. More importantly, most current AI spending is not coming from startups but from profitable, publicly listed hyperscalers.\n\nFor these tech giants, Bank of America views upgrading from traditional CPU computing to accelerated computing as not only “mission-critical” but also a “defensive” investment.\n\nThe report cites Google as an example: its $92 billion in capital expenditures is “defending” the leadership of a search business that generates over $200 billion, preventing users from shifting to emerging competitors such as ChatGPT and Perplexity.\n\nAt the same time, private AI companies such as OpenAI (with over 1 million commercial customers) and Anthropic (with over 300,000 commercial customers) are rapidly attracting enterprise users, which in turn continues to pressure publicly listed software and IaaS providers to increase AI investment. Therefore, the core drivers of AI spending are solid and powerful.\n\nNvidia: not a bubble, but undervalued\nThe report emphasizes that Nvidia’s stock is highly attractive, and the current valuation reflects only a “very cautious AI deployment.”\n\nBased on approximately $500 billion in data center orders for fiscal years 2025/26 recently disclosed by Nvidia, Bank of America estimates the company’s EPS could reach about $8 in 2026.\n\nThis implies that a company with revenue/EPS growing 50%/70% annually is trading at a P/E of just 24x—roughly in line with the market average—meaning the valuation is “not high.”\n\nTo further illustrate its potential, the report presents a scenario assumption:\n\nEven if by 2030 AI capital expenditures reach only 50% of Nvidia’s projected $3–4 trillion range, the company’s EPS at that time could exceed $40 per share. This would imply a forward P/E of under 5x at the current share price.\n\nBank of America regards this number as “beyond common sense,” using it to reaffirm the following view: despite media headlines being filled with bubble narratives, Nvidia’s stock is in fact undervalued.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.045168951697832416, "ret_signed_m1d": 0.045168951697832416, "alpha_spy_m1d": 0.04745287316213764, "alpha_c_m1d": 0.02312077460223616, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01407212231856625, "ret_signed_p1d": 0.01407212231856625, "alpha_spy_p1d": 0.013515723712124172, "alpha_c_p1d": 0.0013104311287330805, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.100138131438952, "ret_signed_p1w": -0.100138131438952, "alpha_spy_p1w": -0.06658040227502826, "alpha_c_p1w": -0.04954708842042421, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": 0.0587825264739108, "ret_signed_p1m": 0.0587825264739108, "alpha_spy_p1m": 0.052091363649354294, "alpha_c_p1m": -0.006877680559447041, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.4536393534152803, "ret_signed_p3m": 0.4536393534152803, "alpha_spy_p3m": 0.4394948218896957, "alpha_c_p3m": 0.3059402511504845, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.8721942441997341, "ret_signed_p6m": 0.8721942441997341, "alpha_spy_p6m": 0.7916547595357459, "alpha_c_p6m": 0.30127658141568414, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.3267, -0.3761, -0.38, -0.3709, -0.3783, -0.383, -0.3725, -0.365, -0.3503, -0.35, -0.3474, -0.3721, -0.3721, -0.3916, -0.3872, -0.3704, -0.3545, -0.3412, -0.3381, -0.3269, -0.2753, -0.2667, -0.2526, -0.2447, -0.2357, -0.208, -0.2042, -0.1711, -0.1728, -0.1939, -0.1951, -0.1938, -0.1765, -0.1588, -0.103, -0.0766, -0.084, -0.063, -0.1183, -0.1045, -0.1142, -0.1747, -0.1343, -0.1311, -0.0905, -0.1056, -0.111, -0.095, -0.0888, -0.1126, -0.0731, -0.0471, -0.0143, -0.0224, 0.0094, 0.0115, -0.0108, 0.0129, -0.0214, 0.0369, 0.0189, 0.0011, 0.0452, 0.0, 0.0141, -0.0368, -0.0686, -0.0736, -0.1001, -0.0652, -0.1231, -0.1038, -0.0553, -0.0455, -0.0254, -0.0254, -0.02, -0.0276, -0.0062, 0.0052, -0.0115, -0.0014, 0.024, 0.0434, 0.0588, 0.0616, 0.0101, 0.0339, 0.0273, -0.0248, 0.0364, 0.084, 0.1028, 0.1022, 0.1159, 0.1159, 0.1205, 0.1067, 0.0935, 0.0772, 0.0772, 0.1645, 0.2255, 0.3023, 0.2779, 0.2645, 0.374, 0.3869, 0.349, 0.3138, 0.3684, 0.403, 0.403, 0.3995, 0.4372, 0.3888, 0.3714, 0.4024, 0.5005, 0.5076, 0.5616, 0.4691, 0.4945, 0.4479, 0.32, 0.3423, 0.4536, 0.4428, 0.426, 0.4795, 0.4554, 0.4821, 0.4821, 0.4824, 0.5108, 0.4938, 0.5412, 0.5245, 0.537, 0.5699, 0.5044, 0.4718, 0.4536, 0.3672, 0.4049, 0.3525, 0.2558, 0.3303, 0.356, 0.3789, 0.3198, 0.3369, 0.3822, 0.4268, 0.4157, 0.4742, 0.4387, 0.4699, 0.5047, 0.4708, 0.3332, 0.3319, 0.2596, 0.3461, 0.3987, 0.3762, 0.3762, 0.3901, 0.4134, 0.5529, 0.6302, 0.6611, 0.6841, 0.7162, 0.6705, 0.6441, 0.6859, 0.6579, 0.6278, 0.673, 0.6289, 0.687, 0.6347, 0.5828, 0.5671, 0.6245, 0.6174, 0.629, 0.7376, 0.8722]}
{"unit_id": "orig:1988211005940507039", "ticker_idx": 4, "ts": "2025-11-11T11:43:37+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reaffirmed its \"Buy\" ratings on the core beneficiaries of AI build-out. These include … KLA (KLAC)", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA (undervalued, EPS ~$8 in 2026 at 24x P/E), AVGO, AMD, LRCX, KLAC, AMAT; AI skepticism seen as contrarian positive, not fundamental risk.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: The “AI bubble” narrative is actually a positive for semiconductor stocks, helps avoid overcrowding, and Nvidia is undervalued\n\nOn the recent “AI bubble” narrative, Bank of America presented a clear contrarian view: it is a “reverse positive” for semiconductor stocks, helping to prevent excessive crowding. There is a clear undervaluation in Nvidia, with the share price failing to reflect its true growth potential.\n\nIn its latest report, Bank of America stated that the widespread skepticism about AI capital expenditures at this point is not a risk but rather a “reverse positive” that ensures this track can deliver sustained returns. Such skepticism helps prevent excessive crowding in the sector.\n\nThe bank believes that the recent volatility in AI semiconductor stocks stems from fixable macro factors rather than a deterioration in the fundamentals of the AI spending cycle. In fact, given the strong performance of AI-related complementary industries and Nvidia’s optimistic outlook, underlying demand remains solid.\n\nBank of America added that doubts about OpenAI’s grand plans are a narrow and fragmentary argument. What truly drives AI investment is the defensive strategy of large tech companies. In particular, it pointed out that the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued.\n\nAI skepticism: a healthy “contrarian positive” signal\nAlthough large AI semiconductor stocks fell by an average of 7–8% last week, Bank of America does not see this as a sign of problems in the AI spending cycle. Rather, the price moves were triggered by macro factors—concerns about a U.S. government shutdown, weak employment data, tariff turmoil, and misunderstandings of comments related to OpenAI—all of which are fixable short-term noise, not fundamental negatives.\n\nThe real fundamental signals are quite the opposite. The report highlighted the strength of AI-related complementary areas—for example, memory and small-/mid-cap optical module stocks rose 14% last week—indicating that AI infrastructure build-out is rolling out across the board.\n\nIt also assessed that Nvidia’s recently disclosed outlook at GTC for its data center business in fiscal years 2025/26—exceeding $500 billion—further proves that AI demand is strong and persistent.\n\nTherefore, Bank of America believes that widespread skepticism actually helps keep the market cool-headed, and it reaffirmed its “Buy” ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers, and Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment.\n\nRebutting the “AI spending is unsustainable” critique: defensive investment by Big Tech is key\nA common bearish narrative in the market is: “If OpenAI can’t justify its $1.4 trillion long-term commitment, then AI stocks must be overvalued.” Bank of America dismisses this as a “lazy and out-of-context argument.”\n\nFirst, OpenAI’s grand plan has not yet materialized and will be constrained by real-world factors such as power and data center space. More importantly, most current AI spending is not coming from startups but from profitable, publicly listed hyperscalers.\n\nFor these tech giants, Bank of America views upgrading from traditional CPU computing to accelerated computing as not only “mission-critical” but also a “defensive” investment.\n\nThe report cites Google as an example: its $92 billion in capital expenditures is “defending” the leadership of a search business that generates over $200 billion, preventing users from shifting to emerging competitors such as ChatGPT and Perplexity.\n\nAt the same time, private AI companies such as OpenAI (with over 1 million commercial customers) and Anthropic (with over 300,000 commercial customers) are rapidly attracting enterprise users, which in turn continues to pressure publicly listed software and IaaS providers to increase AI investment. Therefore, the core drivers of AI spending are solid and powerful.\n\nNvidia: not a bubble, but undervalued\nThe report emphasizes that Nvidia’s stock is highly attractive, and the current valuation reflects only a “very cautious AI deployment.”\n\nBased on approximately $500 billion in data center orders for fiscal years 2025/26 recently disclosed by Nvidia, Bank of America estimates the company’s EPS could reach about $8 in 2026.\n\nThis implies that a company with revenue/EPS growing 50%/70% annually is trading at a P/E of just 24x—roughly in line with the market average—meaning the valuation is “not high.”\n\nTo further illustrate its potential, the report presents a scenario assumption:\n\nEven if by 2030 AI capital expenditures reach only 50% of Nvidia’s projected $3–4 trillion range, the company’s EPS at that time could exceed $40 per share. This would imply a forward P/E of under 5x at the current share price.\n\nBank of America regards this number as “beyond common sense,” using it to reaffirm the following view: despite media headlines being filled with bubble narratives, Nvidia’s stock is in fact undervalued.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02271394210329869, "ret_signed_m1d": 0.02271394210329869, "alpha_spy_m1d": 0.024997863567603917, "alpha_c_m1d": 0.0006657650077024346, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006776280424447112, "ret_signed_p1d": 0.006776280424447112, "alpha_spy_p1d": 0.006219881818005035, "alpha_c_p1d": -0.005985410765386057, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.05535791232752285, "ret_signed_p1w": -0.05535791232752285, "alpha_spy_p1w": -0.021800183163599107, "alpha_c_p1w": -0.004766869308995059, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": 0.042059508317177086, "ret_signed_p1m": 0.042059508317177086, "alpha_spy_p1m": 0.03536834549262058, "alpha_c_p1m": -0.023600698716180757, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.2136795054472327, "ret_signed_p3m": 0.2136795054472327, "alpha_spy_p3m": 0.1995349739216481, "alpha_c_p3m": 0.06598040318243692, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.5296846746411592, "ret_signed_p6m": 0.5296846746411592, "alpha_spy_p6m": 0.449145189977171, "alpha_c_p6m": -0.04123298814289078, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.1995, -0.2669, -0.2582, -0.2644, -0.2624, -0.2675, -0.2692, -0.2614, -0.2543, -0.2536, -0.2493, -0.2678, -0.2678, -0.2893, -0.2914, -0.2667, -0.24, -0.2367, -0.2294, -0.2169, -0.1945, -0.1905, -0.1696, -0.1682, -0.1688, -0.1211, -0.1227, -0.1005, -0.1005, -0.1026, -0.1107, -0.1063, -0.1064, -0.0943, -0.0521, -0.0434, -0.075, -0.043, -0.0891, -0.1077, -0.1154, -0.1748, -0.1393, -0.1387, -0.0872, -0.0773, -0.0707, -0.0319, -0.0365, -0.0643, -0.0268, -0.0068, 0.0203, 0.0127, 0.0373, 0.0197, 0.015, 0.0237, 0.0022, 0.0304, 0.013, 0.0021, 0.0227, 0.0, 0.0068, -0.0245, -0.0475, -0.0464, -0.0554, -0.018, -0.0727, -0.0772, -0.0439, -0.0362, -0.0251, -0.0251, -0.0113, -0.0267, 0.0008, 0.0192, 0.0161, 0.0215, 0.03, 0.0309, 0.0421, 0.0482, 0.0042, 0.0305, 0.029, -0.0142, 0.0282, 0.0477, 0.0646, 0.0672, 0.0741, 0.0741, 0.0763, 0.0601, 0.046, 0.022, 0.022, 0.072, 0.1376, 0.1733, 0.1436, 0.1141, 0.1776, 0.2012, 0.2127, 0.2066, 0.2995, 0.3187, 0.3187, 0.25, 0.2785, 0.2617, 0.2724, 0.2979, 0.3595, 0.3687, 0.417, 0.2011, 0.1863, 0.1402, 0.0995, 0.1195, 0.2137, 0.2113, 0.2035, 0.2444, 0.2203, 0.2315, 0.2315, 0.2382, 0.2467, 0.238, 0.2599, 0.2529, 0.2689, 0.3026, 0.2838, 0.284, 0.2927, 0.2139, 0.243, 0.2038, 0.1324, 0.2036, 0.2237, 0.2338, 0.1871, 0.1948, 0.2113, 0.2476, 0.2484, 0.273, 0.2622, 0.2729, 0.319, 0.3002, 0.2221, 0.2155, 0.1644, 0.2401, 0.28, 0.2775, 0.2775, 0.297, 0.3044, 0.4084, 0.4547, 0.4631, 0.4897, 0.5125, 0.4723, 0.4611, 0.5088, 0.5204, 0.5036, 0.5261, 0.529, 0.6297, 0.6002, 0.5235, 0.5296, 0.4741, 0.4539, 0.443, 0.4595, 0.5297]}
{"unit_id": "orig:1988211005940507039", "ticker_idx": 5, "ts": "2025-11-11T11:43:37+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "reaffirmed its \"Buy\" ratings on the core beneficiaries of AI build-out. These include … Applied Materials (AMAT) in semiconductor equipment", "tweet_type": "news_reaction", "timeline": "months", "summary": "BofA reiterates Buy on NVDA (undervalued, EPS ~$8 in 2026 at 24x P/E), AVGO, AMD, LRCX, KLAC, AMAT; AI skepticism seen as contrarian positive, not fundamental risk.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: The “AI bubble” narrative is actually a positive for semiconductor stocks, helps avoid overcrowding, and Nvidia is undervalued\n\nOn the recent “AI bubble” narrative, Bank of America presented a clear contrarian view: it is a “reverse positive” for semiconductor stocks, helping to prevent excessive crowding. There is a clear undervaluation in Nvidia, with the share price failing to reflect its true growth potential.\n\nIn its latest report, Bank of America stated that the widespread skepticism about AI capital expenditures at this point is not a risk but rather a “reverse positive” that ensures this track can deliver sustained returns. Such skepticism helps prevent excessive crowding in the sector.\n\nThe bank believes that the recent volatility in AI semiconductor stocks stems from fixable macro factors rather than a deterioration in the fundamentals of the AI spending cycle. In fact, given the strong performance of AI-related complementary industries and Nvidia’s optimistic outlook, underlying demand remains solid.\n\nBank of America added that doubts about OpenAI’s grand plans are a narrow and fragmentary argument. What truly drives AI investment is the defensive strategy of large tech companies. In particular, it pointed out that the current share price of industry leader Nvidia does not reflect its true growth potential and is clearly undervalued.\n\nAI skepticism: a healthy “contrarian positive” signal\nAlthough large AI semiconductor stocks fell by an average of 7–8% last week, Bank of America does not see this as a sign of problems in the AI spending cycle. Rather, the price moves were triggered by macro factors—concerns about a U.S. government shutdown, weak employment data, tariff turmoil, and misunderstandings of comments related to OpenAI—all of which are fixable short-term noise, not fundamental negatives.\n\nThe real fundamental signals are quite the opposite. The report highlighted the strength of AI-related complementary areas—for example, memory and small-/mid-cap optical module stocks rose 14% last week—indicating that AI infrastructure build-out is rolling out across the board.\n\nIt also assessed that Nvidia’s recently disclosed outlook at GTC for its data center business in fiscal years 2025/26—exceeding $500 billion—further proves that AI demand is strong and persistent.\n\nTherefore, Bank of America believes that widespread skepticism actually helps keep the market cool-headed, and it reaffirmed its “Buy” ratings on the core beneficiaries of AI build-out. These include Nvidia, Broadcom, and AMD in data centers, and Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) in semiconductor equipment.\n\nRebutting the “AI spending is unsustainable” critique: defensive investment by Big Tech is key\nA common bearish narrative in the market is: “If OpenAI can’t justify its $1.4 trillion long-term commitment, then AI stocks must be overvalued.” Bank of America dismisses this as a “lazy and out-of-context argument.”\n\nFirst, OpenAI’s grand plan has not yet materialized and will be constrained by real-world factors such as power and data center space. More importantly, most current AI spending is not coming from startups but from profitable, publicly listed hyperscalers.\n\nFor these tech giants, Bank of America views upgrading from traditional CPU computing to accelerated computing as not only “mission-critical” but also a “defensive” investment.\n\nThe report cites Google as an example: its $92 billion in capital expenditures is “defending” the leadership of a search business that generates over $200 billion, preventing users from shifting to emerging competitors such as ChatGPT and Perplexity.\n\nAt the same time, private AI companies such as OpenAI (with over 1 million commercial customers) and Anthropic (with over 300,000 commercial customers) are rapidly attracting enterprise users, which in turn continues to pressure publicly listed software and IaaS providers to increase AI investment. Therefore, the core drivers of AI spending are solid and powerful.\n\nNvidia: not a bubble, but undervalued\nThe report emphasizes that Nvidia’s stock is highly attractive, and the current valuation reflects only a “very cautious AI deployment.”\n\nBased on approximately $500 billion in data center orders for fiscal years 2025/26 recently disclosed by Nvidia, Bank of America estimates the company’s EPS could reach about $8 in 2026.\n\nThis implies that a company with revenue/EPS growing 50%/70% annually is trading at a P/E of just 24x—roughly in line with the market average—meaning the valuation is “not high.”\n\nTo further illustrate its potential, the report presents a scenario assumption:\n\nEven if by 2030 AI capital expenditures reach only 50% of Nvidia’s projected $3–4 trillion range, the company’s EPS at that time could exceed $40 per share. This would imply a forward P/E of under 5x at the current share price.\n\nBank of America regards this number as “beyond common sense,” using it to reaffirm the following view: despite media headlines being filled with bubble narratives, Nvidia’s stock is in fact undervalued.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02803167721217581, "ret_signed_m1d": 0.02803167721217581, "alpha_spy_m1d": 0.030315598676481037, "alpha_c_m1d": 0.005983500116579554, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009008606854021384, "ret_signed_p1d": 0.009008606854021384, "alpha_spy_p1d": 0.008452208247579307, "alpha_c_p1d": -0.003753084335811785, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.015524590068340172, "ret_signed_p1w": -0.015524590068340172, "alpha_spy_p1w": 0.018033139095583572, "alpha_c_p1w": 0.03506645295018762, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": 0.2056208398323256, "ret_signed_p1m": 0.2056208398323256, "alpha_spy_p1m": 0.1989296770077691, "alpha_c_p1m": 0.13996063279896775, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.41313763935424364, "ret_signed_p3m": 0.41313763935424364, "alpha_spy_p3m": 0.39899310782865904, "alpha_c_p3m": 0.26543853708944787, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 0.8804205549189219, "ret_signed_p6m": 0.8804205549189219, "alpha_spy_p6m": 0.7998810702549337, "alpha_c_p6m": 0.3095028921348719, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.1792, -0.2946, -0.2869, -0.2926, -0.2981, -0.301, -0.2894, -0.2916, -0.2806, -0.2811, -0.2773, -0.297, -0.297, -0.3109, -0.3167, -0.308, -0.2883, -0.2913, -0.285, -0.2853, -0.2559, -0.2662, -0.2525, -0.2411, -0.221, -0.1702, -0.1687, -0.1231, -0.1216, -0.1191, -0.1271, -0.1082, -0.1037, -0.1046, -0.0478, -0.0222, -0.0487, -0.0208, -0.0748, -0.0488, -0.0366, -0.0819, -0.0402, -0.0458, -0.0048, -0.0042, -0.0161, -0.0024, -0.0117, -0.0355, -0.0009, 0.0003, 0.0116, -0.0045, 0.031, 0.017, 0.0194, 0.0395, 0.0066, 0.0534, 0.0213, 0.0061, 0.028, 0.0, 0.009, -0.0238, -0.0116, 0.0002, -0.0155, 0.0283, -0.035, -0.0185, 0.0118, 0.0624, 0.0953, 0.0953, 0.1053, 0.1162, 0.1626, 0.1771, 0.1806, 0.1743, 0.175, 0.1705, 0.2056, 0.1835, 0.1358, 0.1448, 0.1342, 0.0878, 0.1108, 0.1235, 0.1349, 0.1402, 0.1427, 0.1427, 0.1476, 0.1526, 0.1391, 0.126, 0.126, 0.1781, 0.2458, 0.297, 0.2803, 0.2341, 0.3197, 0.3462, 0.3358, 0.3228, 0.3981, 0.4329, 0.4329, 0.3944, 0.4251, 0.3968, 0.4126, 0.3998, 0.4578, 0.4755, 0.4956, 0.4123, 0.4389, 0.3963, 0.304, 0.332, 0.4131, 0.4485, 0.4419, 0.4892, 0.4389, 0.5551, 0.5551, 0.5736, 0.6182, 0.6225, 0.6468, 0.6388, 0.658, 0.7327, 0.6483, 0.6333, 0.6328, 0.5413, 0.5695, 0.5203, 0.4247, 0.487, 0.5174, 0.5402, 0.4797, 0.4983, 0.5187, 0.5463, 0.5332, 0.5671, 0.5665, 0.5872, 0.6408, 0.6204, 0.4853, 0.4792, 0.4176, 0.4995, 0.5522, 0.5288, 0.5288, 0.547, 0.5544, 0.6922, 0.7453, 0.7526, 0.7361, 0.7357, 0.7297, 0.7106, 0.7414, 0.7181, 0.73, 0.7701, 0.772, 0.8296, 0.7762, 0.672, 0.6785, 0.7307, 0.707, 0.717, 0.8023, 0.8804]}
{"unit_id": "quote:1983795196384178252", "ticker_idx": 0, "ts": "2025-10-30T07:16:46+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA accounts for only 30% of global CAPEX in calendar year 2025. There's still plenty of room for NVIDIA to expand.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Bullish NVDA: only 30% of global CAPEX share in 2025 leaves significant room for further expansion.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "NVIDIA accounts for only 30% of global CAPEX in calendar year 2025.\nThere’s still plenty of room for NVIDIA to expand. https://t.co/ruYayp0bP6", "reply_text": null, "parent_text": null, "quoted_text": "UBS’s Dissection of NVIDIA’s GTC\n\nAside from several new partnerships (detailed below), NVIDIA’s (NVDA) latest demand guidance indicates that the market’s current earnings expectations remain significantly underestimated.\nThis view is also supported by recent findings from supply https://t.co/8CBVwwpPGL", "parent_author": null, "ret_m1d": 0.020454468358698685, "ret_signed_m1d": 0.020454468358698685, "alpha_spy_m1d": 0.009334019127293702, "alpha_c_m1d": 0.007037992278692284, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.013416476080006401, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0019714724306379994, "ret_signed_p1d": -0.0019714724306379994, "alpha_spy_p1d": -0.005251753041353502, "alpha_c_p1d": -1.9400451918394346e-05, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": -0.001952071978719605, "ret_p1w": -0.07299517515386633, "ret_signed_p1w": -0.07299517515386633, "alpha_spy_p1w": -0.05899172675097386, "alpha_c_p1w": -0.03824410262522704, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.03475107252863929, "ret_p1m": -0.1276061122790818, "ret_signed_p1m": -0.1276061122790818, "alpha_spy_p1m": -0.13284273310428596, "alpha_c_p1m": -0.09612670026733339, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.03147941201174842, "ret_p3m": -0.07077481476500547, "ret_signed_p3m": -0.07077481476500547, "alpha_spy_p3m": -0.09683295582014162, "alpha_c_p3m": -0.19390785241759045, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": 0.12313303765258499, "ret_p6m": 0.026629582436106913, "ret_signed_p6m": 0.026629582436106913, "alpha_spy_p6m": -0.029524433465263344, "alpha_c_p6m": -0.3700542685910244, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.39668385102713133, "price_path": [-0.1129, -0.1214, -0.1157, -0.1091, -0.0996, -0.1027, -0.0973, -0.105, -0.1029, -0.1107, -0.103, -0.1344, -0.1355, -0.1376, -0.1228, -0.1138, -0.1041, -0.105, -0.112, -0.1416, -0.1416, -0.1583, -0.1591, -0.154, -0.1768, -0.1705, -0.1584, -0.126, -0.1268, -0.1236, -0.1239, -0.1381, -0.1607, -0.1314, -0.1292, -0.095, -0.1206, -0.1278, -0.1242, -0.1217, -0.1037, -0.0804, -0.0771, -0.069, -0.0753, -0.0855, -0.088, -0.0679, -0.0509, -0.0972, -0.0718, -0.1127, -0.1137, -0.1039, -0.0969, -0.0998, -0.1071, -0.1114, -0.1022, -0.082, -0.0562, -0.0092, 0.0205, 0.0, -0.002, 0.0197, -0.0207, -0.0379, -0.073, -0.0727, -0.0189, -0.048, -0.0448, -0.079, -0.0627, -0.0803, -0.1061, -0.0807, -0.1097, -0.1183, -0.1003, -0.1236, -0.1115, -0.1115, -0.1276, -0.1132, -0.1056, -0.1148, -0.0961, -0.1009, -0.0854, -0.0883, -0.0941, -0.1082, -0.1373, -0.1311, -0.124, -0.1574, -0.1417, -0.1079, -0.0946, -0.0674, -0.0703, -0.0703, -0.0609, -0.0723, -0.0756, -0.0807, -0.0807, -0.0691, -0.0727, -0.0771, -0.0679, -0.0879, -0.0888, -0.0884, -0.0841, -0.0973, -0.078, -0.0821, -0.0821, -0.1223, -0.0964, -0.0889, -0.075, -0.0809, -0.0708, -0.056, -0.0511, -0.0579, -0.0851, -0.1111, -0.1414, -0.1528, -0.0861, -0.0633, -0.0707, -0.0632, -0.0786, -0.0989, -0.0989, -0.0883, -0.0734, -0.0738, -0.0644, -0.0558, -0.0494, -0.0361, -0.0887, -0.1266, -0.1005, -0.1125, -0.0978, -0.0963, -0.1235, -0.0997, -0.0893, -0.083, -0.0972, -0.1115, -0.0968, -0.1032, -0.1108, -0.1198, -0.1487, -0.1342, -0.1364, -0.1192, -0.1559, -0.1742, -0.1858, -0.1403, -0.1337, -0.1256, -0.1256, -0.1244, -0.1221, -0.1025, -0.0934, -0.0702, -0.0668, -0.0313, -0.0197, -0.0223, -0.0059, -0.004, -0.0147, -0.0018, -0.0159, 0.0266]}
{"unit_id": "thread:1989149542735110164", "ticker_idx": 0, "ts": "2025-11-14T01:53:02+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia AI servers drive strong electronics demand, breaking seasonal downturn in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan AI server supply chain (Quanta, Wistron, Foxconn, Pegatron) set for unusually strong 2025 Q4 and 2026 driven by Nvidia GB300 demand, breaking seasonal patterns.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia AI servers drive strong electronics demand, breaking seasonal downturn in 2026\n\nWith demand for AI servers accelerating, Taiwan's electronics supply chain is preparing for an unusually strong fourth quarter in 2025 and a 2026 first half that shows little sign of the traditional slowdown.\n\nQuanta Computer says shipments of its next-generation AI servers will begin ramping sharply from the fourth quarter of 2025 and gain momentum through the first quarter of 2026. Wistron, another major supplier, similarly expects its first-quarter 2026 performance to match what is typically the year's high point: the fourth quarter.\n\nIn past years, without the tailwind from AI servers, downstream electronics makers would see customers pull back orders after November, with weakness often lasting into the second quarter. But with AI infrastructure demand running hot, those seasonal patterns appear to be fading.\n\nSupply-chain executives say much of the strength stems from Nvidia's GB300 server cabinets, which will not reach meaningful production levels until late 2025. As cloud customers transition rapidly to higher-end systems, expanded chip output is expected to lift demand throughout the ecosystem. The first quarter of 2026, they say, could become the strongest slow season in history.\n\nQuanta expects volume shipments of new AI server platforms to surge in early 2026, followed by another refresh cycle in the second half of the year. Those newer systems carry higher average selling prices, the company noted, suggesting even stronger revenue momentum later in 2026. Taken together, Quanta anticipates a rare scenario: a first half that is not soft, and an even stronger second half.\n\nWistron expects its revenue in the first quarter of 2026 to remain on par with the fourth quarter of 2025, with sequential growth anticipated across the remainder of the year. According to President and CEO Jeff Lin, the company recorded triple-digit growth in its AI server business in both 2024 and 2025. Lin noted that “no one would believe” forecasts of another year of triple-digit expansion in 2026, given the elevated base. Still, he added that double-digit growth remains well within reach.\n\nHon Hai Technology Group (Foxconn) is similarly upbeat. Chairman Young Liu said AI's importance will rise sharply in 2026, with deeper collaborations expected with major North American cloud providers and a broader range of solution offerings. He identified AI servers and AI cabinets as Foxconn's most critical growth engines for the year.\n\nNorth America's five largest cloud service providers are projected to spend close to $600 billion on capital expenditures in 2026 — much of it earmarked for AI data-center expansion. Foxconn, citing that figure, believes the market for AI server cabinets could at least double.\n\nThe company expects its AI cabinet shipments to jump threefold in the third quarter of 2025 and continue growing at a high double-digit pace in the fourth quarter, as earlier testing and yield bottlenecks are resolved. Growth should extend into the first quarter of 2026, although the year-over-year slope will flatten due to the high base effect.\n\nTo meet soaring demand, Foxconn is investing in sovereign AI projects across the United States, Taiwan, and Japan, while continuing to develop platforms spanning GPU and ASIC architectures. The company expects its market share — roughly 40 percent today — to rise further in 2026.\n\nFoxconn also plans to extend its value chain downstream, moving from server components and full-system assembly into turnkey data-center system integration (SI/EPC). It is also pushing modular data-center (MDC) deployments, which it says carry higher-than-average margins. Benefits from those initiatives should begin to appear in 2026.\n\nPegatron, despite posting strong third-quarter 2025 results, remains under pressure at the operating-profit level. But President & CEO Kuang-Chih Cheng said the company's 2026 product mix will undergo a meaningful shift as its AI-related server business expands. With AI server demand robust, Pegatron expects a resilient fourth quarter and stronger growth in 2026.\n\nPegatron has started shipping Nvidia's GB200 and GB300 systems in the second half of this year, boosting revenue. Its current customer base mainly includes mid-size and smaller enterprises, but the company is also working with regional sovereign AI players and emerging data-center operators, and is in discussions with top-tier cloud providers.\n\nTo support that growth, Pegatron is expanding capacity worldwide. Beyond its existing Taoyuan facility, it has acquired two additional factory buildings, converting most of the space into server production lines. Its Mexico plant offers L1–L11 manufacturing capability, while a newly built Texas facility, focused on later-stage assembly, will support L6–L11 production beginning in the first quarter of 2026. Pegatron expects that additional capacity to help it capture orders from major cloud providers next year.\n\n$NVDA\n\n---\n\nhttps://t.co/57tuJDcUW8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017405531691603637, "ret_signed_m1d": -0.017405531691603637, "alpha_spy_m1d": -0.01756914455597791, "alpha_c_m1d": -0.016941702719995955, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018772690180637097, "ret_signed_p1d": -0.018772690180637097, "alpha_spy_p1d": -0.009456166249674935, "alpha_c_p1d": -0.0052065579481174185, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.05936791649684514, "ret_signed_p1w": -0.05936791649684514, "alpha_spy_p1w": -0.040169486157191736, "alpha_c_p1w": -0.004727053404036785, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.07293574274855108, "ret_signed_p1m": -0.07293574274855108, "alpha_spy_p1m": -0.08603226282807919, "alpha_c_p1m": -0.09589354646912152, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.0005753669288237617, "ret_signed_p3m": -0.0005753669288237617, "alpha_spy_p3m": -0.03342802619055418, "alpha_c_p3m": -0.20664702748446007, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.15404161563996577, "ret_signed_p6m": 0.15404161563996577, "alpha_spy_p6m": 0.0475133801339096, "alpha_c_p6m": -0.5217169980731151, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.0765, -0.0777, -0.0799, -0.0641, -0.0545, -0.0442, -0.0451, -0.0526, -0.0841, -0.0841, -0.102, -0.1029, -0.0974, -0.1218, -0.115, -0.1021, -0.0676, -0.0684, -0.0649, -0.0653, -0.0804, -0.1045, -0.0733, -0.071, -0.0345, -0.0617, -0.0694, -0.0656, -0.063, -0.0438, -0.0189, -0.0154, -0.0067, -0.0134, -0.0243, -0.027, -0.0056, 0.0126, -0.0369, -0.0097, -0.0533, -0.0544, -0.044, -0.0365, -0.0396, -0.0474, -0.052, -0.0421, -0.0206, 0.0069, 0.0571, 0.0887, 0.0669, 0.0648, 0.0879, 0.0448, 0.0265, -0.011, -0.0106, 0.0467, 0.0157, 0.0191, -0.0174, 0.0, -0.0188, -0.0463, -0.0192, -0.0501, -0.0594, -0.0401, -0.0649, -0.0521, -0.0521, -0.0693, -0.0539, -0.0458, -0.0556, -0.0357, -0.0408, -0.0242, -0.0273, -0.0335, -0.0485, -0.0796, -0.0729, -0.0654, -0.1011, -0.0842, -0.0482, -0.034, -0.005, -0.0081, -0.0081, 0.0019, -0.0102, -0.0138, -0.0192, -0.0192, -0.0069, -0.0107, -0.0154, -0.0055, -0.0269, -0.0279, -0.0274, -0.0229, -0.0369, -0.0164, -0.0207, -0.0207, -0.0636, -0.036, -0.028, -0.0131, -0.0194, -0.0086, 0.0072, 0.0124, 0.0051, -0.0239, -0.0516, -0.084, -0.0961, -0.025, -0.0006, -0.0085, -0.0006, -0.0169, -0.0386, -0.0386, -0.0273, -0.0115, -0.0119, -0.0018, 0.0073, 0.0141, 0.0284, -0.0277, -0.0682, -0.0404, -0.0532, -0.0374, -0.0359, -0.0649, -0.0395, -0.0283, -0.0217, -0.0369, -0.0521, -0.0364, -0.0432, -0.0513, -0.0609, -0.0918, -0.0763, -0.0786, -0.0603, -0.0994, -0.119, -0.1314, -0.0828, -0.0757, -0.0671, -0.0671, -0.0658, -0.0634, -0.0424, -0.0328, -0.008, -0.0044, 0.0335, 0.0459, 0.0431, 0.0606, 0.0626, 0.0512, 0.065, 0.0499, 0.0953, 0.1392, 0.1211, 0.1005, 0.0495, 0.0437, 0.0438, 0.0334, 0.093, 0.1123, 0.1317, 0.154]}
{"unit_id": "thread:1989149542735110164", "ticker_idx": 1, "ts": "2025-11-14T01:53:02+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Quanta expects volume shipments of new AI server platforms to surge in early 2026, followed by another refresh cycle in the second half of the year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan AI server supply chain (Quanta, Wistron, Foxconn, Pegatron) set for unusually strong 2025 Q4 and 2026 driven by Nvidia GB300 demand, breaking seasonal patterns.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia AI servers drive strong electronics demand, breaking seasonal downturn in 2026\n\nWith demand for AI servers accelerating, Taiwan's electronics supply chain is preparing for an unusually strong fourth quarter in 2025 and a 2026 first half that shows little sign of the traditional slowdown.\n\nQuanta Computer says shipments of its next-generation AI servers will begin ramping sharply from the fourth quarter of 2025 and gain momentum through the first quarter of 2026. Wistron, another major supplier, similarly expects its first-quarter 2026 performance to match what is typically the year's high point: the fourth quarter.\n\nIn past years, without the tailwind from AI servers, downstream electronics makers would see customers pull back orders after November, with weakness often lasting into the second quarter. But with AI infrastructure demand running hot, those seasonal patterns appear to be fading.\n\nSupply-chain executives say much of the strength stems from Nvidia's GB300 server cabinets, which will not reach meaningful production levels until late 2025. As cloud customers transition rapidly to higher-end systems, expanded chip output is expected to lift demand throughout the ecosystem. The first quarter of 2026, they say, could become the strongest slow season in history.\n\nQuanta expects volume shipments of new AI server platforms to surge in early 2026, followed by another refresh cycle in the second half of the year. Those newer systems carry higher average selling prices, the company noted, suggesting even stronger revenue momentum later in 2026. Taken together, Quanta anticipates a rare scenario: a first half that is not soft, and an even stronger second half.\n\nWistron expects its revenue in the first quarter of 2026 to remain on par with the fourth quarter of 2025, with sequential growth anticipated across the remainder of the year. According to President and CEO Jeff Lin, the company recorded triple-digit growth in its AI server business in both 2024 and 2025. Lin noted that “no one would believe” forecasts of another year of triple-digit expansion in 2026, given the elevated base. Still, he added that double-digit growth remains well within reach.\n\nHon Hai Technology Group (Foxconn) is similarly upbeat. Chairman Young Liu said AI's importance will rise sharply in 2026, with deeper collaborations expected with major North American cloud providers and a broader range of solution offerings. He identified AI servers and AI cabinets as Foxconn's most critical growth engines for the year.\n\nNorth America's five largest cloud service providers are projected to spend close to $600 billion on capital expenditures in 2026 — much of it earmarked for AI data-center expansion. Foxconn, citing that figure, believes the market for AI server cabinets could at least double.\n\nThe company expects its AI cabinet shipments to jump threefold in the third quarter of 2025 and continue growing at a high double-digit pace in the fourth quarter, as earlier testing and yield bottlenecks are resolved. Growth should extend into the first quarter of 2026, although the year-over-year slope will flatten due to the high base effect.\n\nTo meet soaring demand, Foxconn is investing in sovereign AI projects across the United States, Taiwan, and Japan, while continuing to develop platforms spanning GPU and ASIC architectures. The company expects its market share — roughly 40 percent today — to rise further in 2026.\n\nFoxconn also plans to extend its value chain downstream, moving from server components and full-system assembly into turnkey data-center system integration (SI/EPC). It is also pushing modular data-center (MDC) deployments, which it says carry higher-than-average margins. Benefits from those initiatives should begin to appear in 2026.\n\nPegatron, despite posting strong third-quarter 2025 results, remains under pressure at the operating-profit level. But President & CEO Kuang-Chih Cheng said the company's 2026 product mix will undergo a meaningful shift as its AI-related server business expands. With AI server demand robust, Pegatron expects a resilient fourth quarter and stronger growth in 2026.\n\nPegatron has started shipping Nvidia's GB200 and GB300 systems in the second half of this year, boosting revenue. Its current customer base mainly includes mid-size and smaller enterprises, but the company is also working with regional sovereign AI players and emerging data-center operators, and is in discussions with top-tier cloud providers.\n\nTo support that growth, Pegatron is expanding capacity worldwide. Beyond its existing Taoyuan facility, it has acquired two additional factory buildings, converting most of the space into server production lines. Its Mexico plant offers L1–L11 manufacturing capability, while a newly built Texas facility, focused on later-stage assembly, will support L6–L11 production beginning in the first quarter of 2026. Pegatron expects that additional capacity to help it capture orders from major cloud providers next year.\n\n$NVDA\n\n---\n\nhttps://t.co/57tuJDcUW8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019677996422182487, "ret_signed_m1d": 0.019677996422182487, "alpha_spy_m1d": 0.019514383557808213, "alpha_c_m1d": 0.025091919735591106, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012522361359570633, "ret_signed_p1d": -0.012522361359570633, "alpha_spy_p1d": -0.003205837428608471, "alpha_c_p1d": 0.003129248671427387, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.03398926654740608, "ret_signed_p1w": -0.03398926654740608, "alpha_spy_p1w": -0.014790836207752678, "alpha_c_p1w": 0.017893404097698018, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.019677996422182487, "ret_signed_p1m": 0.019677996422182487, "alpha_spy_p1m": 0.00658147634265438, "alpha_c_p1m": 0.031997969506557644, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": 0.023255813953488413, "ret_signed_p3m": 0.023255813953488413, "alpha_spy_p3m": -0.009596845308242008, "alpha_c_p3m": 0.029420260944872934, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 0.22898032200357776, "ret_signed_p6m": 0.22898032200357776, "alpha_spy_p6m": 0.1224520864975216, "alpha_c_p6m": -0.009064505935656442, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.0411, -0.0716, -0.0733, -0.0751, -0.0411, -0.0519, -0.0376, -0.0483, -0.0608, -0.0841, -0.093, -0.093, -0.0805, -0.0751, -0.0698, -0.0608, -0.0233, -0.034, -0.0197, -0.0125, -0.0215, -0.0233, 0.0, 0.0054, 0.0054, 0.025, 0.0143, 0.0197, 0.0197, 0.0197, 0.0376, 0.025, 0.0519, 0.0733, 0.0733, 0.1038, 0.0948, 0.0662, 0.0662, 0.0537, 0.0161, 0.0233, 0.0429, 0.0322, 0.0608, 0.0411, 0.0358, 0.0501, 0.0501, 0.1073, 0.0877, 0.102, 0.0948, 0.0751, 0.0644, 0.0286, 0.0358, 0.0483, 0.0268, 0.0429, 0.0429, 0.0304, 0.0197, 0.0, -0.0125, -0.034, -0.0447, -0.0197, -0.034, -0.0268, -0.0161, 0.0, 0.0072, 0.0089, 0.0268, 0.0322, 0.0465, 0.0537, 0.0644, 0.0644, 0.059, 0.0465, 0.0304, 0.025, 0.0197, -0.0233, -0.0519, -0.0698, -0.0519, -0.0447, -0.0537, -0.0555, -0.0555, -0.059, -0.059, -0.059, -0.0268, -0.0268, -0.0089, -0.0054, 0.0089, 0.0089, -0.0233, 0.0125, 0.0036, -0.0018, 0.0233, 0.0125, 0.0358, 0.0089, -0.0072, -0.0161, 0.0125, 0.0018, 0.0268, 0.0286, 0.0268, 0.0125, 0.0018, -0.0233, -0.0107, -0.0054, -0.0233, 0.0, 0.0161, 0.0197, 0.0233, 0.0233, 0.0233, 0.0233, 0.0233, 0.0233, 0.0233, 0.0233, 0.0143, 0.0161, 0.0644, 0.0429, 0.0429, 0.0769, 0.0537, 0.0054, 0.0358, 0.034, 0.0089, 0.0107, 0.034, 0.0322, 0.034, 0.034, 0.034, 0.0483, 0.025, 0.0143, 0.0018, 0.0072, 0.0197, 0.0268, 0.034, 0.0054, -0.0036, 0.0394, 0.0179, 0.0179, 0.0179, 0.034, 0.1002, 0.1199, 0.1503, 0.1413, 0.1556, 0.1038, 0.1395, 0.1556, 0.1771, 0.2165, 0.1986, 0.1521, 0.1556, 0.1646, 0.1485, 0.1521, 0.1181, 0.1181, 0.1377, 0.1485, 0.2397, 0.2308, 0.2182, 0.229]}
{"unit_id": "thread:1989149542735110164", "ticker_idx": 2, "ts": "2025-11-14T01:53:02+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Wistron expects its revenue in the first quarter of 2026 to remain on par with the fourth quarter of 2025, with sequential growth anticipated across the remainder of the year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan AI server supply chain (Quanta, Wistron, Foxconn, Pegatron) set for unusually strong 2025 Q4 and 2026 driven by Nvidia GB300 demand, breaking seasonal patterns.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia AI servers drive strong electronics demand, breaking seasonal downturn in 2026\n\nWith demand for AI servers accelerating, Taiwan's electronics supply chain is preparing for an unusually strong fourth quarter in 2025 and a 2026 first half that shows little sign of the traditional slowdown.\n\nQuanta Computer says shipments of its next-generation AI servers will begin ramping sharply from the fourth quarter of 2025 and gain momentum through the first quarter of 2026. Wistron, another major supplier, similarly expects its first-quarter 2026 performance to match what is typically the year's high point: the fourth quarter.\n\nIn past years, without the tailwind from AI servers, downstream electronics makers would see customers pull back orders after November, with weakness often lasting into the second quarter. But with AI infrastructure demand running hot, those seasonal patterns appear to be fading.\n\nSupply-chain executives say much of the strength stems from Nvidia's GB300 server cabinets, which will not reach meaningful production levels until late 2025. As cloud customers transition rapidly to higher-end systems, expanded chip output is expected to lift demand throughout the ecosystem. The first quarter of 2026, they say, could become the strongest slow season in history.\n\nQuanta expects volume shipments of new AI server platforms to surge in early 2026, followed by another refresh cycle in the second half of the year. Those newer systems carry higher average selling prices, the company noted, suggesting even stronger revenue momentum later in 2026. Taken together, Quanta anticipates a rare scenario: a first half that is not soft, and an even stronger second half.\n\nWistron expects its revenue in the first quarter of 2026 to remain on par with the fourth quarter of 2025, with sequential growth anticipated across the remainder of the year. According to President and CEO Jeff Lin, the company recorded triple-digit growth in its AI server business in both 2024 and 2025. Lin noted that “no one would believe” forecasts of another year of triple-digit expansion in 2026, given the elevated base. Still, he added that double-digit growth remains well within reach.\n\nHon Hai Technology Group (Foxconn) is similarly upbeat. Chairman Young Liu said AI's importance will rise sharply in 2026, with deeper collaborations expected with major North American cloud providers and a broader range of solution offerings. He identified AI servers and AI cabinets as Foxconn's most critical growth engines for the year.\n\nNorth America's five largest cloud service providers are projected to spend close to $600 billion on capital expenditures in 2026 — much of it earmarked for AI data-center expansion. Foxconn, citing that figure, believes the market for AI server cabinets could at least double.\n\nThe company expects its AI cabinet shipments to jump threefold in the third quarter of 2025 and continue growing at a high double-digit pace in the fourth quarter, as earlier testing and yield bottlenecks are resolved. Growth should extend into the first quarter of 2026, although the year-over-year slope will flatten due to the high base effect.\n\nTo meet soaring demand, Foxconn is investing in sovereign AI projects across the United States, Taiwan, and Japan, while continuing to develop platforms spanning GPU and ASIC architectures. The company expects its market share — roughly 40 percent today — to rise further in 2026.\n\nFoxconn also plans to extend its value chain downstream, moving from server components and full-system assembly into turnkey data-center system integration (SI/EPC). It is also pushing modular data-center (MDC) deployments, which it says carry higher-than-average margins. Benefits from those initiatives should begin to appear in 2026.\n\nPegatron, despite posting strong third-quarter 2025 results, remains under pressure at the operating-profit level. But President & CEO Kuang-Chih Cheng said the company's 2026 product mix will undergo a meaningful shift as its AI-related server business expands. With AI server demand robust, Pegatron expects a resilient fourth quarter and stronger growth in 2026.\n\nPegatron has started shipping Nvidia's GB200 and GB300 systems in the second half of this year, boosting revenue. Its current customer base mainly includes mid-size and smaller enterprises, but the company is also working with regional sovereign AI players and emerging data-center operators, and is in discussions with top-tier cloud providers.\n\nTo support that growth, Pegatron is expanding capacity worldwide. Beyond its existing Taoyuan facility, it has acquired two additional factory buildings, converting most of the space into server production lines. Its Mexico plant offers L1–L11 manufacturing capability, while a newly built Texas facility, focused on later-stage assembly, will support L6–L11 production beginning in the first quarter of 2026. Pegatron expects that additional capacity to help it capture orders from major cloud providers next year.\n\n$NVDA\n\n---\n\nhttps://t.co/57tuJDcUW8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02491103202846978, "ret_signed_m1d": 0.02491103202846978, "alpha_spy_m1d": 0.024747419164095508, "alpha_c_m1d": 0.0303249553418784, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017793594306049876, "ret_signed_p1d": -0.017793594306049876, "alpha_spy_p1d": -0.008477070375087714, "alpha_c_p1d": -0.002141984275051856, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.010676156583629859, "ret_signed_p1w": -0.010676156583629859, "alpha_spy_p1w": 0.008522273756023546, "alpha_c_p1w": 0.04120651406147424, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.007117437722419906, "ret_signed_p1m": 0.007117437722419906, "alpha_spy_p1m": -0.005979082357108201, "alpha_c_p1m": 0.019437410806795064, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": -0.06405693950177938, "ret_signed_p3m": -0.06405693950177938, "alpha_spy_p3m": -0.0969095987635098, "alpha_c_p3m": -0.057892492510394855, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": -0.010676156583629859, "ret_signed_p6m": -0.010676156583629859, "alpha_spy_p6m": -0.11720439208968603, "alpha_c_p6m": -0.24872098452286406, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.1779, -0.1922, -0.2064, -0.2028, -0.1922, -0.1957, -0.1815, -0.1815, -0.1957, -0.2206, -0.2171, -0.2171, -0.1957, -0.1779, -0.1851, -0.1744, -0.1459, -0.153, -0.1459, -0.1495, -0.1601, -0.153, -0.1459, -0.1388, -0.1352, -0.1281, -0.1352, -0.1068, -0.089, -0.089, 0.0, 0.0391, 0.0641, 0.0925, 0.0925, 0.1103, 0.0712, 0.0747, 0.0747, 0.0356, -0.0178, -0.0178, -0.0071, -0.0285, 0.0107, 0.0142, 0.0214, 0.0178, 0.0178, 0.0427, 0.0534, 0.0854, 0.1032, 0.0712, 0.0214, -0.0107, -0.0214, 0.0107, -0.0214, -0.0142, 0.0285, -0.0107, 0.0249, 0.0, -0.0178, -0.0249, -0.0249, 0.0178, -0.0107, -0.0107, -0.0071, 0.0142, 0.0356, 0.0285, 0.0036, 0.0178, 0.032, 0.0391, 0.0747, 0.0605, 0.0641, 0.0498, 0.0249, 0.0214, 0.0071, -0.0107, -0.0178, -0.0071, 0.0249, 0.0427, 0.032, 0.0356, 0.0356, 0.0427, 0.0391, 0.0463, 0.0712, 0.0712, 0.1032, 0.1032, 0.0996, 0.1281, 0.0747, 0.0605, 0.0391, 0.0214, 0.0391, 0.0214, 0.032, -0.0142, -0.032, -0.0498, -0.0356, -0.0285, -0.0285, -0.0427, -0.0427, -0.0569, -0.0712, -0.089, -0.0854, -0.0819, -0.0925, -0.0819, -0.0676, -0.0783, -0.0641, -0.0641, -0.0641, -0.0641, -0.0641, -0.0641, -0.0641, -0.0641, -0.0783, -0.0676, -0.0178, -0.032, -0.032, -0.0391, -0.0641, -0.1103, -0.0854, -0.0712, -0.0925, -0.0819, -0.0534, -0.0569, -0.0463, -0.0391, -0.0463, -0.0712, -0.0925, -0.0747, -0.1068, -0.0819, -0.089, -0.0961, -0.0961, -0.1068, -0.1281, -0.0961, -0.1103, -0.1103, -0.1103, -0.121, -0.0676, -0.0569, -0.0391, -0.0427, -0.0285, -0.0534, -0.0463, -0.0285, 0.0, 0.0285, 0.0249, -0.0071, 0.0071, 0.0142, 0.0, 0.0, -0.0249, -0.0249, 0.0, 0.0107, 0.0427, 0.0391, 0.0427, -0.0107]}
{"unit_id": "thread:1989149542735110164", "ticker_idx": 3, "ts": "2025-11-14T01:53:02+00:00", "symbol": "Hon Hai Technology Group", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn believes the market for AI server cabinets could at least double", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan AI server supply chain (Quanta, Wistron, Foxconn, Pegatron) set for unusually strong 2025 Q4 and 2026 driven by Nvidia GB300 demand, breaking seasonal patterns.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Technology Group 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia AI servers drive strong electronics demand, breaking seasonal downturn in 2026\n\nWith demand for AI servers accelerating, Taiwan's electronics supply chain is preparing for an unusually strong fourth quarter in 2025 and a 2026 first half that shows little sign of the traditional slowdown.\n\nQuanta Computer says shipments of its next-generation AI servers will begin ramping sharply from the fourth quarter of 2025 and gain momentum through the first quarter of 2026. Wistron, another major supplier, similarly expects its first-quarter 2026 performance to match what is typically the year's high point: the fourth quarter.\n\nIn past years, without the tailwind from AI servers, downstream electronics makers would see customers pull back orders after November, with weakness often lasting into the second quarter. But with AI infrastructure demand running hot, those seasonal patterns appear to be fading.\n\nSupply-chain executives say much of the strength stems from Nvidia's GB300 server cabinets, which will not reach meaningful production levels until late 2025. As cloud customers transition rapidly to higher-end systems, expanded chip output is expected to lift demand throughout the ecosystem. The first quarter of 2026, they say, could become the strongest slow season in history.\n\nQuanta expects volume shipments of new AI server platforms to surge in early 2026, followed by another refresh cycle in the second half of the year. Those newer systems carry higher average selling prices, the company noted, suggesting even stronger revenue momentum later in 2026. Taken together, Quanta anticipates a rare scenario: a first half that is not soft, and an even stronger second half.\n\nWistron expects its revenue in the first quarter of 2026 to remain on par with the fourth quarter of 2025, with sequential growth anticipated across the remainder of the year. According to President and CEO Jeff Lin, the company recorded triple-digit growth in its AI server business in both 2024 and 2025. Lin noted that “no one would believe” forecasts of another year of triple-digit expansion in 2026, given the elevated base. Still, he added that double-digit growth remains well within reach.\n\nHon Hai Technology Group (Foxconn) is similarly upbeat. Chairman Young Liu said AI's importance will rise sharply in 2026, with deeper collaborations expected with major North American cloud providers and a broader range of solution offerings. He identified AI servers and AI cabinets as Foxconn's most critical growth engines for the year.\n\nNorth America's five largest cloud service providers are projected to spend close to $600 billion on capital expenditures in 2026 — much of it earmarked for AI data-center expansion. Foxconn, citing that figure, believes the market for AI server cabinets could at least double.\n\nThe company expects its AI cabinet shipments to jump threefold in the third quarter of 2025 and continue growing at a high double-digit pace in the fourth quarter, as earlier testing and yield bottlenecks are resolved. Growth should extend into the first quarter of 2026, although the year-over-year slope will flatten due to the high base effect.\n\nTo meet soaring demand, Foxconn is investing in sovereign AI projects across the United States, Taiwan, and Japan, while continuing to develop platforms spanning GPU and ASIC architectures. The company expects its market share — roughly 40 percent today — to rise further in 2026.\n\nFoxconn also plans to extend its value chain downstream, moving from server components and full-system assembly into turnkey data-center system integration (SI/EPC). It is also pushing modular data-center (MDC) deployments, which it says carry higher-than-average margins. Benefits from those initiatives should begin to appear in 2026.\n\nPegatron, despite posting strong third-quarter 2025 results, remains under pressure at the operating-profit level. But President & CEO Kuang-Chih Cheng said the company's 2026 product mix will undergo a meaningful shift as its AI-related server business expands. With AI server demand robust, Pegatron expects a resilient fourth quarter and stronger growth in 2026.\n\nPegatron has started shipping Nvidia's GB200 and GB300 systems in the second half of this year, boosting revenue. Its current customer base mainly includes mid-size and smaller enterprises, but the company is also working with regional sovereign AI players and emerging data-center operators, and is in discussions with top-tier cloud providers.\n\nTo support that growth, Pegatron is expanding capacity worldwide. Beyond its existing Taoyuan facility, it has acquired two additional factory buildings, converting most of the space into server production lines. Its Mexico plant offers L1–L11 manufacturing capability, while a newly built Texas facility, focused on later-stage assembly, will support L6–L11 production beginning in the first quarter of 2026. Pegatron expects that additional capacity to help it capture orders from major cloud providers next year.\n\n$NVDA\n\n---\n\nhttps://t.co/57tuJDcUW8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.045643153526971014, "ret_signed_m1d": 0.045643153526971014, "alpha_spy_m1d": 0.04547954066259674, "alpha_c_m1d": 0.051057076840379634, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020746887966805017, "ret_signed_p1d": -0.020746887966805017, "alpha_spy_p1d": -0.011430364035842855, "alpha_c_p1d": -0.005095277935806997, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.06639004149377592, "ret_signed_p1w": -0.06639004149377592, "alpha_spy_p1w": -0.047191611154122515, "alpha_c_p1w": -0.014507370848671819, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": -0.08091286307053946, "ret_signed_p1m": -0.08091286307053946, "alpha_spy_p1m": -0.09400938315006757, "alpha_c_p1m": -0.06859288998616431, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": -0.05809128630705396, "ret_signed_p3m": -0.05809128630705396, "alpha_spy_p3m": -0.09094394556878438, "alpha_c_p3m": -0.05192683931566944, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 0.045643153526971014, "ret_signed_p6m": 0.045643153526971014, "alpha_spy_p6m": -0.060885081979085154, "alpha_c_p6m": -0.1924016744122632, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.1369, -0.168, -0.1452, -0.1598, -0.139, -0.1328, -0.1369, -0.1452, -0.1556, -0.1763, -0.1763, -0.168, -0.1577, -0.1494, -0.1556, -0.139, -0.1286, -0.11, -0.0975, -0.1037, -0.1058, -0.1203, -0.1079, -0.112, -0.1037, -0.083, -0.0747, -0.0456, -0.0892, -0.0892, -0.1037, -0.0913, -0.0643, -0.0602, -0.0602, -0.0519, -0.0664, -0.0809, -0.0809, -0.1162, -0.1452, -0.1432, -0.0747, -0.0602, -0.0104, -0.0083, 0.0062, -0.0083, -0.0083, 0.0228, 0.0373, 0.0788, 0.0871, 0.0685, 0.0436, 0.0124, 0.0249, 0.029, 0.0124, 0.0353, 0.0228, 0.0415, 0.0456, 0.0, -0.0207, -0.0456, -0.0498, -0.0187, -0.0664, -0.0871, -0.0913, -0.0581, -0.0415, -0.0643, -0.0788, -0.0788, -0.0581, -0.0519, -0.0415, -0.0373, -0.0249, -0.0311, -0.0622, -0.0581, -0.0809, -0.0954, -0.1017, -0.1037, -0.0809, -0.0705, -0.0705, -0.0705, -0.0705, -0.0643, -0.0415, -0.0539, -0.0436, -0.0436, -0.0373, -0.027, -0.0207, -0.0104, -0.0477, -0.0436, -0.0519, -0.0602, -0.027, -0.0311, -0.027, -0.0477, -0.0726, -0.0913, -0.0726, -0.0809, -0.0705, -0.0643, -0.0643, -0.0705, -0.0851, -0.11, -0.1017, -0.0913, -0.1058, -0.1079, -0.0934, -0.083, -0.0581, -0.0581, -0.0581, -0.0581, -0.0581, -0.0581, -0.0581, -0.0581, -0.0519, -0.0373, 0.0228, 0.0083, 0.0083, -0.0083, -0.0498, -0.0996, -0.0705, -0.0747, -0.1266, -0.1307, -0.0913, -0.11, -0.11, -0.1017, -0.1203, -0.1286, -0.1494, -0.1577, -0.1867, -0.1909, -0.1701, -0.168, -0.1722, -0.195, -0.222, -0.1826, -0.1992, -0.1992, -0.1992, -0.2033, -0.1639, -0.1701, -0.168, -0.1701, -0.139, -0.139, -0.1411, -0.1452, -0.139, -0.1245, -0.083, -0.0664, -0.0809, -0.0539, -0.0643, -0.0664, -0.0892, -0.0892, -0.056, -0.0062, 0.0456, 0.0519, 0.0373, 0.0456]}
{"unit_id": "thread:1989149542735110164", "ticker_idx": 4, "ts": "2025-11-14T01:53:02+00:00", "symbol": "Pegatron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Pegatron expects a resilient fourth quarter and stronger growth in 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan AI server supply chain (Quanta, Wistron, Foxconn, Pegatron) set for unusually strong 2025 Q4 and 2026 driven by Nvidia GB300 demand, breaking seasonal patterns.", "resolved_tickers": ["4938.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Pegatron Corp listed on TWSE as 4938.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia AI servers drive strong electronics demand, breaking seasonal downturn in 2026\n\nWith demand for AI servers accelerating, Taiwan's electronics supply chain is preparing for an unusually strong fourth quarter in 2025 and a 2026 first half that shows little sign of the traditional slowdown.\n\nQuanta Computer says shipments of its next-generation AI servers will begin ramping sharply from the fourth quarter of 2025 and gain momentum through the first quarter of 2026. Wistron, another major supplier, similarly expects its first-quarter 2026 performance to match what is typically the year's high point: the fourth quarter.\n\nIn past years, without the tailwind from AI servers, downstream electronics makers would see customers pull back orders after November, with weakness often lasting into the second quarter. But with AI infrastructure demand running hot, those seasonal patterns appear to be fading.\n\nSupply-chain executives say much of the strength stems from Nvidia's GB300 server cabinets, which will not reach meaningful production levels until late 2025. As cloud customers transition rapidly to higher-end systems, expanded chip output is expected to lift demand throughout the ecosystem. The first quarter of 2026, they say, could become the strongest slow season in history.\n\nQuanta expects volume shipments of new AI server platforms to surge in early 2026, followed by another refresh cycle in the second half of the year. Those newer systems carry higher average selling prices, the company noted, suggesting even stronger revenue momentum later in 2026. Taken together, Quanta anticipates a rare scenario: a first half that is not soft, and an even stronger second half.\n\nWistron expects its revenue in the first quarter of 2026 to remain on par with the fourth quarter of 2025, with sequential growth anticipated across the remainder of the year. According to President and CEO Jeff Lin, the company recorded triple-digit growth in its AI server business in both 2024 and 2025. Lin noted that “no one would believe” forecasts of another year of triple-digit expansion in 2026, given the elevated base. Still, he added that double-digit growth remains well within reach.\n\nHon Hai Technology Group (Foxconn) is similarly upbeat. Chairman Young Liu said AI's importance will rise sharply in 2026, with deeper collaborations expected with major North American cloud providers and a broader range of solution offerings. He identified AI servers and AI cabinets as Foxconn's most critical growth engines for the year.\n\nNorth America's five largest cloud service providers are projected to spend close to $600 billion on capital expenditures in 2026 — much of it earmarked for AI data-center expansion. Foxconn, citing that figure, believes the market for AI server cabinets could at least double.\n\nThe company expects its AI cabinet shipments to jump threefold in the third quarter of 2025 and continue growing at a high double-digit pace in the fourth quarter, as earlier testing and yield bottlenecks are resolved. Growth should extend into the first quarter of 2026, although the year-over-year slope will flatten due to the high base effect.\n\nTo meet soaring demand, Foxconn is investing in sovereign AI projects across the United States, Taiwan, and Japan, while continuing to develop platforms spanning GPU and ASIC architectures. The company expects its market share — roughly 40 percent today — to rise further in 2026.\n\nFoxconn also plans to extend its value chain downstream, moving from server components and full-system assembly into turnkey data-center system integration (SI/EPC). It is also pushing modular data-center (MDC) deployments, which it says carry higher-than-average margins. Benefits from those initiatives should begin to appear in 2026.\n\nPegatron, despite posting strong third-quarter 2025 results, remains under pressure at the operating-profit level. But President & CEO Kuang-Chih Cheng said the company's 2026 product mix will undergo a meaningful shift as its AI-related server business expands. With AI server demand robust, Pegatron expects a resilient fourth quarter and stronger growth in 2026.\n\nPegatron has started shipping Nvidia's GB200 and GB300 systems in the second half of this year, boosting revenue. Its current customer base mainly includes mid-size and smaller enterprises, but the company is also working with regional sovereign AI players and emerging data-center operators, and is in discussions with top-tier cloud providers.\n\nTo support that growth, Pegatron is expanding capacity worldwide. Beyond its existing Taoyuan facility, it has acquired two additional factory buildings, converting most of the space into server production lines. Its Mexico plant offers L1–L11 manufacturing capability, while a newly built Texas facility, focused on later-stage assembly, will support L6–L11 production beginning in the first quarter of 2026. Pegatron expects that additional capacity to help it capture orders from major cloud providers next year.\n\n$NVDA\n\n---\n\nhttps://t.co/57tuJDcUW8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.00016361286437427403, "alpha_c_m1d": 0.0054139233134086195, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0163264890917304, "ret_signed_p1d": -0.0163264890917304, "alpha_spy_p1d": -0.007009965160768239, "alpha_c_p1d": -0.0006748790607323807, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.036734652357036546, "ret_signed_p1w": -0.036734652357036546, "alpha_spy_p1w": -0.01753622201738314, "alpha_c_p1w": 0.015148018288067555, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": -0.04081632653061229, "ret_signed_p1m": -0.04081632653061229, "alpha_spy_p1m": -0.0539128466101404, "alpha_c_p1m": -0.028496353446237133, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": -0.04081632653061229, "ret_signed_p3m": -0.04081632653061229, "alpha_spy_p3m": -0.07366898579234271, "alpha_c_p3m": -0.03465187953922777, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 0.13469389831127754, "ret_signed_p6m": 0.13469389831127754, "alpha_spy_p6m": 0.02816566280522137, "alpha_c_p6m": -0.10335092962795667, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.0449, -0.0612, -0.0544, -0.0544, -0.0422, -0.0435, -0.0408, -0.0395, -0.0449, -0.0612, -0.0599, -0.0599, -0.049, -0.0449, -0.0367, -0.0381, -0.0381, -0.0612, -0.0435, -0.049, -0.0476, -0.0517, -0.0408, -0.0476, -0.0381, -0.0327, -0.0204, -0.0272, -0.0422, -0.0422, -0.0367, -0.0422, -0.0272, -0.0272, -0.0272, -0.0177, -0.0163, -0.019, -0.019, -0.0231, -0.0367, -0.0272, -0.0218, 0.0136, 0.0136, 0.0218, 0.0435, 0.0327, 0.0327, 0.0408, 0.034, 0.0463, 0.0354, 0.0095, 0.0136, -0.0082, -0.0054, 0.0068, -0.0122, -0.0109, -0.0109, -0.0177, 0.0, 0.0, -0.0163, -0.0395, -0.049, -0.0259, -0.0367, -0.0381, -0.0463, -0.0367, -0.0272, -0.019, -0.0218, -0.0231, -0.0109, -0.0204, -0.0313, -0.0395, -0.0476, -0.0408, -0.0544, -0.0626, -0.0408, -0.0408, -0.0585, -0.0707, -0.0422, -0.0463, -0.0599, -0.0558, -0.0558, -0.0653, -0.0653, -0.068, -0.0667, -0.0667, -0.0626, -0.0653, -0.0626, -0.0531, -0.0408, -0.0558, -0.0367, -0.0449, -0.0286, -0.0299, -0.0299, -0.0367, -0.0544, -0.068, -0.0544, -0.0449, -0.0354, -0.0435, -0.0395, -0.0463, -0.0517, -0.049, -0.0571, -0.0531, -0.0544, -0.0435, -0.0476, -0.0381, -0.0408, -0.0408, -0.0408, -0.0408, -0.0408, -0.0408, -0.0408, -0.0408, -0.0354, -0.0218, -0.0054, 0.0, 0.0, -0.0218, -0.0367, -0.068, -0.0558, -0.034, -0.0544, -0.0435, -0.0463, 0.049, 0.0313, 0.0476, 0.0639, 0.0639, 0.0517, 0.0476, 0.0299, 0.0395, 0.0531, 0.098, 0.1007, 0.0694, 0.0367, 0.0639, 0.0612, 0.0612, 0.0612, 0.0599, 0.083, 0.0884, 0.098, 0.0612, 0.068, 0.0721, 0.0898, 0.1293, 0.1252, 0.1429, 0.1524, 0.1429, 0.1469, 0.1306, 0.1252, 0.1156, 0.1197, 0.1197, 0.1061, 0.1088, 0.1374, 0.1469, 0.1483, 0.1347]}
{"unit_id": "thread:2002667062226022877", "ticker_idx": 0, "ts": "2025-12-21T09:06:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung had achieved the best results in the memory industry in terms of operating speed and power efficiency… a green light has been turned on for Samsung Electronics' supply of NVIDIA HBM4 in the first half of next year… the volume of HBM4 Samsung is required to supply next year by NVIDIA far exceeded internal expectations", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung HBM4 rated best among Big Three in NVIDIA tests, green-lit for H1 supply with volumes exceeding internal expectations; bullish for Samsung earnings.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung’s HBM4 Rated ‘Best’ Among the Big Three Memory Makers in NVIDIA Test…Green Light for Supply Next Year\n\nSamsung Electronics’ HBM4 is said to have received the highest score in tests for NVIDIA’s next-generation artificial intelligence (AI) accelerator “Vera Rubin,” set to be released next year.\n\nAccording to the semiconductor industry on the 21st, an NVIDIA-related team visited Samsung Electronics last week to receive an update on the progress of HBM4 SiP (System in Package) testing. At the meeting, it was shared that Samsung had achieved the best results in the memory industry in terms of operating speed and power efficiency.\n\nAccordingly, a green light has been turned on for Samsung Electronics’ supply of NVIDIA HBM4 in the first half of next year. It is reported that the volume of HBM4 Samsung is required to supply next year by NVIDIA far exceeded internal expectations. This is also expected to significantly help improve Samsung’s earnings.\n\nSamsung is said to be planning to officially sign a supply contract in the first quarter of next year, taking into account the expansion pace and production capacity of its Pyeongtaek P4 production line. It is then expected to begin full-scale supply from the second quarter.\n\nSamsung says it cannot officially confirm matters related to NVIDIA’s tests. However, a Samsung official said, “Unlike with HBM3E, the overall mood within the development team is that we are taking the lead in HBM4 development,” adding, “We expect to achieve meaningful results next year.”\n\nOf course, SK hynix completed preparations for mass production of HBM4 at the end of September, putting it about three months ahead of Samsung. In particular, SK hynix is known to be supplying paid samples to NVIDIA, and is thus effectively considered to have entered mass production. However, considering that there was a one-year gap during the HBM3E mass-production phase, the gap has narrowed rapidly.\n\nIn particular, since full-scale mass production is expected to take place in the second quarter of next year, Samsung—which scored highly in SiP—may increase its supply volume even further.\n\nSiP refers to a technology that integrates multiple semiconductor chips into a single package. In the case of HBM4, it means that the graphics processing unit (GPU), memory semiconductors, interposer, power chips, and the like are all included in one package. In the semiconductor industry, the SiP test is interpreted as the final gate for verifying mass production—checking whether factors such as speed, heat generation, and power signals are within controllable levels.\n\nSamsung recently regained the No. 2 spot in the HBM market from Micron in just three quarters. According to market research firm Counterpoint Research on the 21st, based on third-quarter revenue, Samsung’s HBM market share was 22%, placing it second after SK hynix (57%). Micron followed in third with 21%, behind Samsung.\n\n---\n\nhttps://t.co/0BrkSE4B9o", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03800910447247974, "ret_signed_m1d": -0.03800910447247974, "alpha_spy_m1d": -0.03181778345714559, "alpha_c_m1d": -0.03264273289239972, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.0053663715800800205, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009049725757364735, "ret_signed_p1d": 0.009049725757364735, "alpha_spy_p1d": 0.004479270938931856, "alpha_c_p1d": 0.0036074705593895917, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.005442255197975143, "ret_p1w": 0.08670506815824552, "ret_signed_p1w": 0.08670506815824552, "alpha_spy_p1w": 0.082295296882958, "alpha_c_p1w": 0.08181389010770301, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.004891178050542511, "ret_p1m": 0.32041476800311086, "ret_signed_p1m": 0.32041476800311086, "alpha_spy_p1m": 0.33100138268862156, "alpha_c_p1m": 0.34328604833761345, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": -0.022871280334502586, "ret_p3m": 0.8233000768945837, "ret_signed_p3m": 0.8233000768945837, "alpha_spy_p3m": 0.8598493272048692, "alpha_c_p3m": 0.8696627572260551, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": -0.046362680331471395, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2306, -0.2243, -0.2495, -0.238, -0.2407, -0.2217, -0.1878, -0.1878, -0.1878, -0.1878, -0.1878, -0.1878, -0.1457, -0.1557, -0.171, -0.1403, -0.1158, -0.114, -0.1122, -0.1176, -0.1077, -0.1267, -0.1059, -0.0769, -0.0995, -0.0905, -0.0579, -0.0271, 0.0054, -0.0507, -0.0896, -0.1023, -0.114, -0.0896, -0.0633, -0.067, -0.0697, -0.1204, -0.0896, -0.1149, -0.1267, -0.0896, -0.1421, -0.1249, -0.1014, -0.0697, -0.0633, -0.0905, -0.0878, -0.0643, -0.0543, -0.0489, -0.019, -0.009, -0.019, -0.0226, -0.029, -0.0145, -0.0516, -0.0697, -0.0235, -0.0262, -0.038, 0.0, 0.009, 0.0054, 0.0054, 0.0588, 0.0867, 0.0903, 0.0903, 0.0903, 0.1685, 0.2558, 0.2631, 0.2822, 0.2622, 0.264, 0.2622, 0.2513, 0.2759, 0.3086, 0.3541, 0.3577, 0.3204, 0.3595, 0.385, 0.3832, 0.3832, 0.4505, 0.4768, 0.4614, 0.4595, 0.3677, 0.5232, 0.5378, 0.4486, 0.4423, 0.5132, 0.5077, 0.5259, 0.6241, 0.6478, 0.6478, 0.6478, 0.6478, 0.7278, 0.7287, 0.7551, 0.8188, 0.8506, 0.9824, 0.9688, 0.9688, 0.7742, 0.5659, 0.7424, 0.7114, 0.5778, 0.7087, 0.7278, 0.7087, 0.6687, 0.716, 0.7633, 0.8961, 0.8233, 0.8133, 0.6942, 0.7251, 0.7187, 0.6378, 0.6378, 0.6065, 0.5236, 0.7282, 0.6257, 0.6968, 0.7596, 0.7906, 0.9182, 0.859, 0.8772, 0.8316, 0.8817, 0.9228, 0.982, 0.9683, 0.9547, 0.9957, 0.982, 1.0458, 1.0002, 1.0458, 1.023, 1.0594, 1.0093, 1.0093, 1.1187, 1.1187, 1.4239, 1.4741, 1.4467, 1.6016, 1.5424, 1.588, 1.6973, 1.465, 1.5606, 1.5105, 1.5151, 1.7292, 1.6654, 1.6654, 1.7247, 1.7976, 1.7292, 1.8887, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:1999717102924747092", "ticker_idx": 0, "ts": "2025-12-13T05:44:44+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I project that Apple will raise the prices of its products, including the iPhone, within the first half of next year... this will lead to reduced shareholder returns", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish AAPL: expiring LTAs expose Apple to memory price surge, cash buffer will reduce shareholder returns, product price hikes coming H1 next year.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "What I find hard to understand is that the sell-side and the market are significantly overestimating Apple's supply chain management capabilities.\n\nIn my view, Apple is also likely to take a significant hit from this memory price surge. The LTAs (Long-Term Agreements) that Apple previously secured are already expiring.\n\nAccording to my supply chain channel checks, Samsung and SK Hynix are planning to raise memory prices for Apple starting next January.\n\nFolks, if you are planning to buy electronics, buy them right now. This is the cheapest they will be.\n\n$AAPL\n\n---\n\nOf course, Apple sits on a massive pile of cash, which will serve as a strong buffer to withstand this memory price surge.\n\nHowever, while this may be beneficial for expanding market share (based on the logic that Apple uses its cash to offset product cost increases), I believe everyone is overlooking the inevitable fact that this will lead to reduced shareholder returns.\n\nFinally, I project that Apple will raise the prices of its products, including the iPhone, within the first half of next year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01521296534145411, "ret_signed_m1d": -0.01521296534145411, "alpha_spy_m1d": -0.013699881494426691, "alpha_c_m1d": -0.005444883626940955, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.009768081714513155, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0018241281266524467, "ret_signed_p1d": -0.0018241281266524467, "alpha_spy_p1d": -0.004556425721931157, "alpha_c_p1d": 2.9517519637423106e-06, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": 0.001827079878616189, "ret_p1w": -0.011455176630011943, "ret_signed_p1w": 0.011455176630011943, "alpha_spy_p1w": 0.020450901942714372, "alpha_c_p1w": 0.03310078291795371, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.021645606287941765, "ret_p1m": -0.04764505772706318, "ret_signed_p1m": 0.04764505772706318, "alpha_spy_p1m": 0.06981254763277411, "alpha_c_p1m": 0.07858090068593471, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.03093584295887153, "ret_p3m": -0.06607082645362794, "ret_signed_p3m": 0.06607082645362794, "alpha_spy_p3m": 0.04741165318893581, "alpha_c_p3m": 0.036197769365768795, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.029873057087859145, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.129, -0.133, -0.1052, -0.0667, -0.0727, -0.0804, -0.0638, -0.0689, -0.0727, -0.072, -0.069, -0.0629, -0.0596, -0.0645, -0.0652, -0.0595, -0.0741, -0.1061, -0.0974, -0.097, -0.0912, -0.0981, -0.0805, -0.0442, -0.0423, -0.058, -0.0539, -0.0421, -0.0203, -0.0196, -0.017, -0.0108, -0.0146, -0.0194, -0.0158, -0.0154, -0.0168, -0.0215, -0.0171, 0.0042, -0.0023, -0.0042, -0.0062, -0.0243, -0.0243, -0.0202, -0.0287, -0.0096, 0.0066, 0.0104, 0.0125, 0.0125, 0.0173, 0.0328, 0.0441, 0.0366, 0.024, 0.017, 0.0138, 0.0112, 0.017, 0.0143, 0.0152, 0.0, 0.0018, -0.0083, -0.007, -0.0016, -0.0115, -0.0064, -0.0011, -0.0011, -0.0026, -0.0013, -0.0038, -0.0082, -0.0082, -0.0113, -0.025, -0.0429, -0.0503, -0.055, -0.0538, -0.0506, -0.0476, -0.0516, -0.058, -0.0678, -0.0678, -0.1, -0.0965, -0.094, -0.0951, -0.0682, -0.0578, -0.0645, -0.0578, -0.0534, -0.015, -0.0169, 0.0087, 0.0066, 0.0146, 0.0028, -0.0006, 0.006, -0.0443, -0.066, -0.066, -0.0364, -0.0347, -0.0485, -0.0339, -0.028, -0.0063, 0.0014, -0.0033, -0.0353, -0.0334, -0.0369, -0.0414, -0.0495, -0.0599, -0.051, -0.0476, -0.0476, -0.0661, -0.0867, -0.0768, -0.0717, -0.0873, -0.0909, -0.0944, -0.0817, -0.0811, -0.0775, -0.0766, -0.0915, -0.0994, -0.0733, -0.0665, -0.0655, -0.0655, -0.0548, -0.0743, -0.0546, -0.0488, -0.0488, -0.0535, -0.0549, -0.0271, -0.0382, -0.0132, -0.0029, -0.0281, -0.0025, -0.0015, -0.0102, -0.0228, -0.0115, -0.0135, -0.0091, 0.023, 0.0109, 0.0377, 0.0499, 0.0496, 0.0711, 0.0697, 0.0775, 0.0924, 0.0899, 0.0973, 0.0886, 0.0927, 0.1047, 0.1147, 0.1287, 0.1287, 0.1269, 0.1361, 0.1422, 0.1406, null, null, null, null, null, null, null]}
{"unit_id": "reply:1997652696958980385", "ticker_idx": 0, "ts": "2025-12-07T13:01:32+00:00", "symbol": "BTC-USD", "kind": "crypto", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bitcoin is worth holding for the long term", "tweet_type": "prediction", "timeline": "years", "summary": "Long BTC for the long term.", "resolved_tickers": ["BTC-USD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": null, "bench_c_reason": "no_country_fallback (all FX/crypto/index)", "bench_c_industry": null, "unit_kind": "reply_other", "lang": "en", "tweet_text": "@Tientientien121 Japan’s rate hikes won’t hurt equities, and I think Bitcoin is worth holding for the long term.", "reply_text": null, "parent_text": "@jukan05 What are your thoughts on ETH and BTC?\nAlso, how do you think Japan’s interest rate hike on December 19 will affect the U.S. stock market?", "quoted_text": null, "parent_author": "Tientientien121", "ret_m1d": -0.013817768157169508, "ret_signed_m1d": -0.013817768157169508, "alpha_spy_m1d": -0.01683109855259557, "alpha_c_m1d": null, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": null, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": null, "bench_spy_t0": 0.0, "bench_c_t0": null, "ret_p1d": 0.022633530616329445, "ret_signed_p1d": 0.022633530616329445, "alpha_spy_p1d": 0.02349667558329105, "alpha_c_p1d": null, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": null, "ret_p1w": -0.04656236117630608, "ret_signed_p1w": -0.04656236117630608, "alpha_spy_p1w": -0.04232026360384178, "alpha_c_p1w": null, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": null, "ret_p1m": 0.03407790382751297, "ret_signed_p1m": 0.03407790382751297, "alpha_spy_p1m": 0.019122141690619987, "alpha_c_p1m": null, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": null, "ret_p3m": -0.21843594169460878, "ret_signed_p3m": -0.21843594169460878, "alpha_spy_p3m": -0.21798717455169048, "alpha_c_p3m": null, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.2572, 0.2744, 0.2809, 0.2737, 0.2891, 0.285, 0.2923, 0.2764, 0.2439, 0.2358, 0.2503, 0.2031, 0.2104, 0.2621, 0.2583, 0.309, 0.3314, 0.3489, 0.3763, 0.3399, 0.3609, 0.3427, 0.2491, 0.2717, 0.248, 0.2222, 0.1936, 0.1746, 0.2201, 0.1968, 0.1881, 0.2144, 0.225, 0.259, 0.2462, 0.2142, 0.1949, 0.2087, 0.1755, 0.1208, 0.1462, 0.1176, 0.1405, 0.1694, 0.1363, 0.1216, 0.0999, 0.0415, 0.016, 0.0255, 0.0091, -0.0442, -0.0612, -0.0261, -0.0364, -0.0013, 0.0071, 0.0031, -0.0476, 0.0078, 0.0319, 0.0166, -0.0138, 0.0, 0.0226, 0.0152, 0.0206, -0.0041, -0.0466, -0.0308, -0.0496, -0.0571, -0.028, -0.0237, -0.0356, -0.0334, -0.0376, -0.0368, -0.0386, -0.0244, -0.0345, -0.0211, -0.0077, 0.0358, 0.0341, 0.0074, 0.0043, -0.0014, 0.0061, 0.0517, 0.0694, 0.0542, 0.0539, 0.0211, -0.0257, -0.0139, -0.013, -0.0125, -0.0262, -0.017, -0.0161, -0.0671, -0.0718, -0.1319, -0.1656, -0.1944, -0.3082, -0.2216, -0.2264, -0.241, -0.2609, -0.2694, -0.2403, -0.2405, -0.2554, -0.2672, -0.2613, -0.2497, -0.2871, -0.293, -0.2502, -0.2558, -0.2732, -0.2412, -0.2465, -0.1978, -0.2184, -0.2483, -0.2453, -0.2285, -0.2255, -0.2223, -0.217, -0.1741, -0.1844, -0.214, -0.2287, -0.222, -0.2176, -0.222, -0.2133, -0.241, -0.2681, -0.2642, -0.2472, -0.2489, -0.262, -0.2616, -0.2403, -0.2063, -0.2153, -0.2082, -0.1948, -0.1782, -0.1816, -0.1747, -0.1709, -0.1491, -0.1629, -0.1576, -0.1372, -0.1365, -0.1455, -0.1464, -0.1577, -0.164, -0.1582, -0.1375, -0.1193, -0.1072, -0.1016, -0.1173, -0.1153, -0.0983, -0.1121, -0.1254, -0.1058, -0.1277, -0.151, -0.1532, -0.1454, -0.1445, -0.1672, -0.1474, -0.1634, -0.1798, -0.1887, -0.1905, null, null]}
{"unit_id": "thread:1990709450756530306", "ticker_idx": 0, "ts": "2025-11-18T09:11:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics has now secured HBM4 supply to Nvidia. It plans to begin supplying initial volumes from the fourth quarter of this year.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung finalizes HBM4 supply to Nvidia, ramps 1c DRAM to 200k wafers/month; author shares bullish thesis on Samsung's DRAM recovery and HBM market recapture.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung finalizes HBM4 supply to Nvidia… begins expanding 1c DRAM\n\nThe 10-nanometer (nm)-class 6th-generation DRAM (hereinafter 1c DRAM) production capacity that Samsung Electronics aims to secure by 2026—200,000 wafers per month on an input basis—is a considerable scale, approaching one-third of the company’s total DRAM output. It is no exaggeration to say that the center of gravity of Samsung’s DRAM will shift to 1c next year.\n\nSamsung Electronics’ moves are aimed at the artificial intelligence (AI) market. First of all, its confidence in recapturing the high-bandwidth memory (HBM) market is seen as having led to this production capacity investment.\n\nSamsung Electronics has struggled to supply HBM to Nvidia. Nvidia is the world’s No.1 in AI semiconductor chips built by combining graphics processing units (GPUs) with HBM.\n\nHowever, because Samsung failed to deliver HBM in a timely manner, its leadership in the DRAM market also weakened. As HBM is manufactured by vertically stacking DRAM, an increase in HBM sales must be accompanied by an increase in DRAM shipments.\n\nFrom the first to the third quarter of this year, Samsung Electronics ceded the No.1 spot in DRAM market share to SK hynix. Demand for HBM has surged due to AI infrastructure investment, but Samsung’s response lagged behind that of its competitors.\n\nSamsung traced the primary cause back to the underlying DRAM. Judging that DRAM quality directly translates into HBM competitiveness, it embarked on an extensive design overhaul.\n\n1c DRAM is the fruit of these efforts. It is now being recognized internally for its productivity. Yields are around 70%, and stable production (yields of 80–90%) is imminent. The yield of HBM4 produced with 1c DRAM is also understood to have exceeded 50%.\n\nAfter this period of hard reflection, Samsung Electronics has now secured HBM4 supply to Nvidia. It plans to begin supplying initial volumes from the fourth quarter of this year. With an explosive surge in HBM4 demand virtually assured, the company has jumped into securing 1c DRAM production capacity.\n\nDemand for server, PC, and mobile DRAM has also driven Samsung’s investments. Server DRAM demand is skyrocketing due to increased AI data center investment. As data center operators competitively invest in infrastructure to implement AI services, server DRAM prices are also surging.\n\nThe PC and mobile markets, whose growth had slowed, are also stirring again—thanks to on-device AI. As the trend of implementing AI directly on laptops and smartphones without connecting to the network spreads, it is pulling up DRAM demand.\n\nTo run AI computation on the device itself, large-capacity memory is essential. That means far more DRAM is required. PCs and mobiles are also expected to become one of the pillars of DRAM demand, contributing to a supply shortage.\n\nFundamentally, it is this strong demand for high-performance memory that has driven Samsung to increase the share of 1c DRAM production. On top of that, the fact that 1c DRAM is a differentiated weapon unique to Samsung has further accelerated the pace.\n\nSamsung has decided to build HBM with 1c DRAM while its competitors are using 1b DRAM. Having fallen behind in response, it is aiming to turn the tables with a generation-ahead product. It appears the company is massively expanding production capacity to upend the market with 1c DRAM.\n\nAn industry insider said, “Samsung Electronics has led the market by leveraging its formidable production capacity compared with competitors,” adding, “There is a high likelihood that this strategy will be replayed in 1c DRAM as well.”\n\nPreparation for 1c production is proceeding through both capacity expansion and process conversion. Capacity expansion is centered on Pyeongtaek Plant 4 (P4), where equipment move-in and installation (setup) are currently in full swing.\n\nProcess conversion will be done by changing 10-nanometer-class 1st to 3rd-generation (1x, 1y, 1z) DRAM lines into 1c manufacturing lines. This is a tech migration. In this case, existing line equipment is utilized as much as possible and only essential equipment is newly brought in. This enables the rapid establishment of DRAM manufacturing lines and allows the company to increase its speed in responding to the market.\n\nThe effect of cost reduction relative to investment is also significant. It can also be read as a determination to realize a “super gap” by reducing production of older-generation DRAM—where China’s chase has been fierce—and concentrating on next-generation DRAM. Tech migration is said to be proceeding simultaneously across each line.\n\n---\n\nhttps://t.co/vOGCVt4gw2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.028629816550031828, "ret_signed_m1d": 0.028629816550031828, "alpha_spy_m1d": 0.020161241466577717, "alpha_c_m1d": 0.012108339584125272, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.016521476965906556, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013292434936019282, "ret_signed_p1d": -0.013292434936019282, "alpha_spy_p1d": -0.01715563471960646, "alpha_c_p1d": -0.020245111658979686, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.006952676722960405, "ret_p1w": 0.015337381614012546, "ret_signed_p1w": 0.015337381614012546, "alpha_spy_p1w": -0.007296135352197863, "alpha_c_p1w": 0.010212503775267834, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.0051248778387447125, "ret_p1m": 0.10327185837897668, "ret_signed_p1m": 0.10327185837897668, "alpha_spy_p1m": 0.08612249864833599, "alpha_c_p1m": 0.10416788123293341, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": -0.000896022853956735, "ret_p3m": 0.8617671386246695, "ret_signed_p3m": 0.8617671386246695, "alpha_spy_p3m": 0.825885960729541, "alpha_c_p3m": 0.8599273854575797, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.0018397531670897305, "ret_p6m": 1.924041286961046, "ret_signed_p6m": 1.924041286961046, "alpha_spy_p6m": 1.7930622336816595, "alpha_c_p6m": 1.6529344925524478, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.2711067944085981, "price_path": [-0.2813, -0.2732, -0.2722, -0.2844, -0.2813, -0.2915, -0.2905, -0.3119, -0.2966, -0.2895, -0.2864, -0.2925, -0.2864, -0.2722, -0.261, -0.2528, -0.2325, -0.2213, -0.1917, -0.204, -0.1826, -0.1826, -0.15, -0.1378, -0.1307, -0.1235, -0.152, -0.1391, -0.1421, -0.1207, -0.0823, -0.0823, -0.0823, -0.0823, -0.0823, -0.0823, -0.0348, -0.046, -0.0634, -0.0286, -0.001, 0.001, 0.0031, -0.0031, 0.0082, -0.0133, 0.0102, 0.0429, 0.0174, 0.0276, 0.0644, 0.0992, 0.136, 0.0726, 0.0286, 0.0143, 0.001, 0.0286, 0.0583, 0.0542, 0.0511, -0.0061, 0.0286, 0.0, -0.0133, 0.0286, -0.0307, -0.0112, 0.0153, 0.0511, 0.0583, 0.0276, 0.0307, 0.0573, 0.0685, 0.0746, 0.1084, 0.1196, 0.1084, 0.1043, 0.0971, 0.1135, 0.0716, 0.0511, 0.1033, 0.1002, 0.0869, 0.1299, 0.1401, 0.136, 0.136, 0.1963, 0.2278, 0.2319, 0.2319, 0.2319, 0.3203, 0.4189, 0.4271, 0.4487, 0.4261, 0.4282, 0.4261, 0.4138, 0.4415, 0.4785, 0.5299, 0.534, 0.4919, 0.5361, 0.5648, 0.5628, 0.5628, 0.6388, 0.6686, 0.6511, 0.6491, 0.5453, 0.721, 0.7374, 0.6368, 0.6296, 0.7097, 0.7035, 0.7241, 0.8351, 0.8618, 0.8618, 0.8618, 0.8618, 0.9522, 0.9532, 0.983, 1.0549, 1.0909, 1.2399, 1.2245, 1.2245, 1.0046, 0.7693, 0.9686, 0.9337, 0.7827, 0.9306, 0.9522, 0.9306, 0.8854, 0.9388, 0.9923, 1.1423, 1.0601, 1.0488, 0.9142, 0.9491, 0.9419, 0.8505, 0.8505, 0.8152, 0.7215, 0.9526, 0.8368, 0.9171, 0.9881, 1.0231, 1.1673, 1.1004, 1.121, 1.0695, 1.1261, 1.1724, 1.2394, 1.2239, 1.2085, 1.2548, 1.2394, 1.3114, 1.26, 1.3114, 1.2857, 1.3269, 1.2703, 1.2703, 1.3938, 1.3938, 1.7387, 1.7953, 1.7645, 1.9395, 1.8726, 1.924]}
{"unit_id": "thread:1982338857195725066", "ticker_idx": 0, "ts": "2025-10-26T06:49:48+00:00", "symbol": "KLAC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Success Could Impact KLA", "tweet_type": "analysis", "timeline": "months", "summary": "Intekplus entering TSMC CoWoS inspection equipment qualification threatens KLA's exclusive supply position; success would erode KLAC's market share.", "resolved_tickers": ["KLAC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Korean Local Semiconductor Equipment Company Enters TSMC CoWoS Equipment Qualification Testing; Success Could Impact KLA\n\nIntekplus is attempting to enter TSMC’s advanced packaging ecosystem with its semiconductor visual inspection equipment.\n\nAccording to industry sources on the 9th, Intekplus began the quality evaluation of TSMC’s “Chip on Wafer on Substrate (CoWoS)” visual inspection equipment last month through a Taiwanese outsourced semiconductor assembly and test (OSAT) company.\n\nThe customer is presumed to be ASE, the world’s largest OSAT company. ASE handles most of TSMC’s advanced packaging volume.\n\nFollowing its supply of the six-sided visual inspection equipment “iPIS-HX Series” to Samsung Electronics’ advanced packaging sector, the company has now turned its focus to TSMC. The equipment currently under delivery and evaluation is a customized model optimized for CoWoS.\n\nCoWoS is a packaging technology that places processors such as GPUs on a substrate and connects them to high-bandwidth memory (HBM) through an interposer. NVIDIA’s AI accelerators adopt this CoWoS approach.\n\nAs global foundry and advanced packaging demand centered on AI semiconductors continues to concentrate on TSMC, and investments to expand CoWoS production capacity accelerate, Intekplus is expected to benefit if it enters the supply chain.\n\nThe company expects to begin mass production supply next year. Currently, CoWoS inspection equipment is exclusively supplied by U.S.-based KLA.\n\nPreviously, Intekplus was recognized for its technical capabilities after directly supplying TSMC with its “ISIS-NTV” automatic optical inspection (AOI) equipment for flip-chip ball grid array (FC-BGA) substrates last year.\n\nA company official stated, “The equipment quality evaluation is expected to take at least six months,” adding, “We cannot confirm specific customer information.”\n\n$KLAC\n\n---\n\nhttps://t.co/BYvr8zK2vL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026589760176527255, "ret_signed_m1d": 0.026589760176527255, "alpha_spy_m1d": 0.014929611232763818, "alpha_c_m1d": 0.002141319941138864, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0074806368443333104, "ret_signed_p1d": 0.0074806368443333104, "alpha_spy_p1d": 0.010136635800924387, "alpha_c_p1d": 0.01628763072965056, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.00329999907184364, "ret_signed_p1w": -0.00329999907184364, "alpha_spy_p1w": -0.0060727186571172265, "alpha_c_p1w": 0.013813985637026738, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": -0.05539932712706175, "ret_signed_p1m": 0.05539932712706175, "alpha_spy_p1m": 0.04048479565065244, "alpha_c_p1m": 2.902718986197428e-05, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.23650692828297948, "ret_signed_p3m": -0.23650692828297948, "alpha_spy_p3m": -0.2280779610715329, "alpha_c_p3m": -0.11389365832589182, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 0.47366064024219234, "ret_signed_p6m": -0.47366064024219234, "alpha_spy_p6m": -0.4403160737800511, "alpha_c_p6m": -0.17870637619158747, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.2404, -0.2782, -0.2719, -0.2481, -0.2746, -0.2706, -0.251, -0.2488, -0.2526, -0.2318, -0.2203, -0.2154, -0.2815, -0.273, -0.279, -0.2771, -0.2821, -0.2838, -0.2762, -0.2691, -0.2685, -0.2643, -0.2824, -0.2824, -0.3035, -0.3055, -0.2813, -0.2551, -0.2519, -0.2447, -0.2325, -0.2106, -0.2067, -0.1862, -0.1848, -0.1854, -0.1386, -0.1402, -0.1184, -0.1184, -0.1205, -0.1284, -0.1241, -0.1243, -0.1124, -0.071, -0.0624, -0.0935, -0.0621, -0.1073, -0.1255, -0.133, -0.1912, -0.1565, -0.1559, -0.1054, -0.0957, -0.0893, -0.0512, -0.0557, -0.083, -0.0462, -0.0266, 0.0, -0.0075, 0.0166, -0.0006, -0.0053, 0.0033, -0.0178, 0.0099, -0.0072, -0.0179, 0.0023, -0.0199, -0.0133, -0.044, -0.0665, -0.0654, -0.0742, -0.0376, -0.0912, -0.0956, -0.063, -0.0554, -0.0445, -0.0445, -0.031, -0.0461, -0.0192, -0.0011, -0.0041, 0.0011, 0.0095, 0.0103, 0.0213, 0.0273, -0.0158, 0.0099, 0.0085, -0.0339, 0.0077, 0.0269, 0.0433, 0.0459, 0.0527, 0.0527, 0.0548, 0.039, 0.0252, 0.0016, 0.0016, 0.0506, 0.1149, 0.15, 0.1208, 0.0919, 0.1541, 0.1773, 0.1885, 0.1825, 0.2736, 0.2924, 0.2924, 0.2251, 0.253, 0.2365, 0.247, 0.272, 0.3324, 0.3414, 0.3888, 0.1771, 0.1627, 0.1174, 0.0776, 0.0972, 0.1895, 0.1872, 0.1795, 0.2196, 0.196, 0.2069, 0.2069, 0.2135, 0.2219, 0.2133, 0.2348, 0.2279, 0.2436, 0.2766, 0.2582, 0.2584, 0.267, 0.1897, 0.2182, 0.1798, 0.1098, 0.1796, 0.1993, 0.2092, 0.1635, 0.171, 0.1871, 0.2227, 0.2236, 0.2476, 0.237, 0.2475, 0.2927, 0.2743, 0.1978, 0.1912, 0.1412, 0.2153, 0.2545, 0.252, 0.252, 0.2712, 0.2784, 0.3804, 0.4257, 0.434, 0.46, 0.4824, 0.4429, 0.432, 0.4787, 0.4901, 0.4737]}
{"unit_id": "orig:1980421807011467367", "ticker_idx": 0, "ts": "2025-10-20T23:52:07+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi raises its target price to ₩145,000 (from ₩133,000) and reiterates its Buy rating", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Citi Buy reiteration on Samsung Electronics with PT raised to ₩145,000 on AI inference memory demand and sharply higher 2026 OP estimate of ₩80.1T.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electronics (005930.KS)\nRegaining Market Position with OP Upside from AI Inference Memory Demand\n\n⸻\n\nCITI’s View\n\nCiti raises its 2026 operating profit (OP) estimate for Samsung Electronics to ₩80.1 trillion.\nThis revision reflects expectations that, following the aggressive AI investments by U.S. hyperscalers, Chinese hyperscalers will also accelerate their AI inference infrastructure build-out.\nIn addition, the rapid surge in tokenized data generated by AI agents is expected to deepen the shortage of commodity memory.\nGiven Samsung’s solid market position in server DDR5, GDDR7, and eSSD, the company is expected to benefit directly from the expansion of AI inference memory demand.\nThe target price is raised from ₩133,000 to ₩145,000, with a Buy rating maintained.\n\n⸻\n\nMemory Business: Core Beneficiary of AI Inference Demand\n\nWith the memory market strengthening on the back of aggressive AI investments by U.S. hyperscalers,\nCiti projects that Chinese hyperscalers will also initiate large-scale investments to prepare for AI inference.\n\nAs a result, demand for high-performance memory products such as server DDR5, GDDR7, SOCAMM2, and eSSD is expected to rise,\ndriving Samsung Electronics’ earnings recovery in 4Q25 and 2026.\n\n⸻\n\n2026 Global Memory ASP Outlook\n    • DRAM: +37% YoY (from +25%)\n    • NAND: +39% YoY (from +23%)\n\nCiti revises up its 2026 DRAM average selling price (ASP) growth forecast from +25% to +37%,\nand NAND from +23% to +39%.\n\nThe upward revision is based on:\n    1. A sharp increase in demand for AI inference memory solutions in China, and\n    2. AI agents’ token generation outpacing bit supply growth.\n\nAccordingly, the shortage of commodity memory is expected to intensify further,\nlikely extending beyond DRAM to NAND as well.\nThe acceleration of AI inference and token generation is projected to trigger additional NAND demand,\nleading both DRAM and NAND ASPs to rise +37% and +39% YoY, respectively, in 2026.\n\n⸻\n\n2026 Operating Profit Forecast: ₩80.1 Trillion\n\nSamsung Electronics’ 2026 operating profit has been revised up from ₩61.5 trillion to ₩80.1 trillion,\nreflecting expectations of stronger-than-anticipated memory price increases driven by\ntight commodity memory supply and surging AI inference memory demand.\n    • 2026 Semiconductor OP: ₩63.0 trillion (vs. ₩21.6 trillion in 2025, a sharp increase)\n    • Smartphone shipments: 229 million in 2025 → 232 million in 2026 (+1%)\n→ Supported by successful launches of new foldable and flagship models\n    •    2026E OP by division:\n    •    Semiconductor: ₩63.0 trillion\n    •    Mobile: ₩12.7 trillion\n    •    Display: ₩3.1 trillion\n    •    CE (Consumer Electronics): ₩0.9 trillion\n\n⸻\n\nTarget Price Raised to ₩145,000 / Rating: Buy Maintained\n\nCiti has raised its 2026E/2027E operating profit estimates by 30% and 25%, respectively,\nto reflect earnings upside from incremental AI inference demand in China\nand the explosive growth of tokenized data generated by AI agents.\n\nBased on 2026E EV/EBITDA SOTP analysis,\nCiti raises its target price to ₩145,000 (from ₩133,000)\nand reiterates its Buy rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002038693068074693, "ret_signed_m1d": -0.002038693068074693, "alpha_spy_m1d": 0.0082547258809893, "alpha_c_m1d": 0.00899567002942181, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.011034363097496502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0061162395860097085, "ret_signed_p1d": -0.0061162395860097085, "alpha_spy_p1d": -0.00610124366599718, "alpha_c_p1d": -0.0066366592900064525, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": 0.000520419703996744, "ret_p1w": 0.03975535683183096, "ret_signed_p1w": 0.03975535683183096, "alpha_spy_p1w": 0.018989665155993096, "alpha_c_p1w": 0.0022454244184331174, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03750993241339784, "ret_p1m": -0.0030580796975584468, "ret_signed_p1m": -0.0030580796975584468, "alpha_spy_p1m": 0.01365569610319628, "alpha_c_p1m": 0.028726492666849124, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.03178457236440757, "ret_p3m": 0.47400115025816647, "ret_signed_p3m": 0.47400115025816647, "alpha_spy_p3m": 0.4397608929428092, "alpha_c_p3m": 0.46299705029383453, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.011004099964331937, "ret_p6m": 1.119605678926571, "ret_signed_p6m": 1.119605678926571, "alpha_spy_p6m": 1.079214991319748, "alpha_c_p6m": 1.090086169346484, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.029519509580086867, "price_path": [-0.3261, -0.3302, -0.3312, -0.2856, -0.2835, -0.2632, -0.2754, -0.3008, -0.2927, -0.2906, -0.3018, -0.2845, -0.2713, -0.2795, -0.2784, -0.2703, -0.2734, -0.2734, -0.2896, -0.2896, -0.2845, -0.2835, -0.2754, -0.2744, -0.2866, -0.2835, -0.2937, -0.2927, -0.314, -0.2987, -0.2916, -0.2886, -0.2947, -0.2886, -0.2744, -0.2632, -0.2551, -0.2348, -0.2236, -0.1942, -0.2064, -0.1851, -0.1851, -0.1526, -0.1404, -0.1333, -0.1262, -0.1546, -0.1417, -0.1448, -0.1233, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0377, -0.0489, -0.0663, -0.0316, -0.0041, -0.002, 0.0, -0.0061, 0.0051, -0.0163, 0.0071, 0.0398, 0.0143, 0.0245, 0.0612, 0.0958, 0.1325, 0.0693, 0.0255, 0.0112, -0.002, 0.0255, 0.055, 0.051, 0.0479, -0.0092, 0.0255, -0.0031, -0.0163, 0.0255, -0.0336, -0.0143, 0.0122, 0.0479, 0.055, 0.0245, 0.0275, 0.054, 0.0652, 0.0714, 0.105, 0.1162, 0.105, 0.1009, 0.0938, 0.1101, 0.0683, 0.0479, 0.0999, 0.0968, 0.0836, 0.1264, 0.1366, 0.1325, 0.1325, 0.1927, 0.2241, 0.2282, 0.2282, 0.2282, 0.3163, 0.4146, 0.4228, 0.4443, 0.4218, 0.4238, 0.4218, 0.4095, 0.4371, 0.474, 0.5252, 0.5293, 0.4873, 0.5314, 0.56, 0.558, 0.558, 0.6338, 0.6635, 0.6461, 0.644, 0.5406, 0.7157, 0.7321, 0.6317, 0.6246, 0.7045, 0.6983, 0.7188, 0.8294, 0.8561, 0.8561, 0.8561, 0.8561, 0.9462, 0.9472, 0.9769, 1.0486, 1.0845, 1.233, 1.2177, 1.2177, 0.9985, 0.7639, 0.9626, 0.9278, 0.7772, 0.9247, 0.9462, 0.9247, 0.8796, 0.9329, 0.9862, 1.1357, 1.0538, 1.0425, 0.9083, 0.9431, 0.936, 0.8448, 0.8448, 0.8096, 0.7162, 0.9466, 0.8312, 0.9112, 0.9821, 1.017, 1.1607, 1.0939, 1.1145, 1.0632, 1.1196]}
{"unit_id": "quote:1975796782488801763", "ticker_idx": 0, "ts": "2025-10-08T05:33:56+00:00", "symbol": "Kinsus Interconnect Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ABF price hike expected at the end of this month — Kinsus / Nandem (Nan Ya PCB)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares rumor of ABF substrate price hikes (25% special-spec, 10–20% others) with supply shortages to 2027H1; Kinsus and Nan Ya PCB are direct beneficiaries.", "resolved_tickers": ["3189.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Kinsus Interconnect Technology 3189.TW Taiwan substrate", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "(Rumor)\n\nABF price hike expected at the end of this month — Kinsus / Nandem (Nan Ya PCB)\nSpecial-spec ABF products will see a 25% price increase, while other categories will rise by 10–20%.\nThe companies expect supply shortages to continue until the first half of 2027, and the upcoming price hike will be implemented in one go.\n\nCurrently, all ABF production capacity at Nandem’s four factories has been fully reserved by NVIDIA,\nmeaning there is no available capacity left until the new plant is completed by the end of next year.\n\nAs a result, the market is expected to enter a price hike and supply shortage cycle,\nsimilar to what Nanya Technology experienced in the past.", "reply_text": null, "parent_text": null, "quoted_text": "傳\nABF月底調漲 景碩/南電\nABF特殊規格產品調整25%、其餘10-20%不等\n公司預期缺貨缺到2027年H1，漲價會一次調整到位\n南電目前四個廠ABF產能全被輝達包下來\n目前全部無產能可用，要等到明年底廠房蓋好\n將進入漲價缺貨循環，比照南亞科辦理\n\n #零級分", "parent_author": null, "ret_m1d": -0.07999999999999996, "ret_signed_m1d": -0.07999999999999996, "alpha_spy_m1d": -0.07407229480397903, "alpha_c_m1d": -0.05393586785702342, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.026064132142976537, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10000000000000009, "ret_signed_p1d": 0.10000000000000009, "alpha_spy_p1d": 0.10289700837923754, "alpha_c_p1d": 0.10248512894381767, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0024851289438175828, "ret_p1w": 0.07600000000000007, "ret_signed_p1w": 0.07600000000000007, "alpha_spy_p1w": 0.08779595277861763, "alpha_c_p1w": 0.08839636524018113, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.012396365240181062, "ret_p1m": 0.040000000000000036, "ret_signed_p1m": 0.040000000000000036, "alpha_spy_p1m": 0.04415975312323461, "alpha_c_p1m": 0.025494273871576878, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": 0.014505726128423158, "ret_p3m": 0.3440000000000001, "ret_signed_p3m": 0.3440000000000001, "alpha_spy_p3m": 0.31927575520060336, "alpha_c_p3m": 0.24952520573920256, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": 0.09447479426079752, "ret_p6m": 1.6880000000000002, "ret_signed_p6m": 1.6880000000000002, "alpha_spy_p6m": 1.7081240311349977, "alpha_c_p6m": 1.5508290911702824, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": 0.13717090882971772, "price_path": [-0.2776, -0.2784, -0.2064, -0.18, -0.188, -0.192, -0.164, -0.196, -0.176, -0.176, -0.2, -0.2, -0.152, -0.16, -0.176, -0.176, -0.188, -0.188, -0.192, -0.188, -0.188, -0.164, -0.172, -0.176, -0.176, -0.184, -0.108, -0.132, -0.18, -0.16, -0.16, -0.14, -0.088, -0.084, -0.092, -0.092, -0.156, -0.168, -0.168, -0.176, -0.176, -0.172, -0.168, -0.152, -0.164, -0.156, -0.18, -0.116, -0.136, -0.12, -0.124, -0.092, -0.108, -0.116, -0.112, -0.16, -0.16, -0.14, -0.152, -0.128, -0.104, -0.104, -0.08, 0.0, 0.1, 0.1, 0.068, -0.02, 0.076, 0.18, 0.208, 0.208, 0.2, 0.172, 0.16, 0.16, 0.188, 0.1, 0.112, 0.1, 0.092, 0.112, 0.076, 0.06, 0.04, 0.024, 0.024, 0.024, 0.052, 0.156, 0.096, 0.132, 0.164, 0.076, 0.116, 0.032, 0.056, 0.1, 0.104, 0.128, 0.14, 0.132, 0.128, 0.184, 0.184, 0.256, 0.248, 0.276, 0.22, 0.212, 0.212, 0.168, 0.108, 0.088, 0.056, 0.08, 0.152, 0.14, 0.152, 0.152, 0.224, 0.204, 0.184, 0.272, 0.272, 0.248, 0.344, 0.356, 0.4, 0.328, 0.348, 0.448, 0.412, 0.376, 0.4, 0.54, 0.6358, 0.5971, 0.5468, 0.5894, 0.6551, 0.664, 0.828, 0.948, 0.836, 1.016, 0.916, 0.968, 1.164, 1.108, 0.94, 1.04, 1.028, 1.124, 1.124, 1.124, 1.124, 1.124, 1.124, 1.124, 1.124, 1.336, 1.568, 1.604, 1.516, 1.516, 1.436, 1.472, 1.276, 1.38, 1.332, 1.116, 1.192, 1.324, 1.312, 1.54, 1.696, 1.732, 1.924, 2.068, 2.04, 1.768, 1.564, 1.812, 1.812, 1.816, 1.772, 1.504, 1.656, 1.688]}
{"unit_id": "quote:1975796782488801763", "ticker_idx": 1, "ts": "2025-10-08T05:33:56+00:00", "symbol": "Nan Ya PCB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "all ABF production capacity at Nandem's four factories has been fully reserved by NVIDIA, meaning there is no available capacity left until the new plant is completed by the end of next year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares rumor of ABF substrate price hikes (25% special-spec, 10–20% others) with supply shortages to 2027H1; Kinsus and Nan Ya PCB are direct beneficiaries.", "resolved_tickers": ["8046.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nan Ya PCB (NYPCB) listed on TWSE as 8046.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "quote", "lang": "en", "tweet_text": "(Rumor)\n\nABF price hike expected at the end of this month — Kinsus / Nandem (Nan Ya PCB)\nSpecial-spec ABF products will see a 25% price increase, while other categories will rise by 10–20%.\nThe companies expect supply shortages to continue until the first half of 2027, and the upcoming price hike will be implemented in one go.\n\nCurrently, all ABF production capacity at Nandem’s four factories has been fully reserved by NVIDIA,\nmeaning there is no available capacity left until the new plant is completed by the end of next year.\n\nAs a result, the market is expected to enter a price hike and supply shortage cycle,\nsimilar to what Nanya Technology experienced in the past.", "reply_text": null, "parent_text": null, "quoted_text": "傳\nABF月底調漲 景碩/南電\nABF特殊規格產品調整25%、其餘10-20%不等\n公司預期缺貨缺到2027年H1，漲價會一次調整到位\n南電目前四個廠ABF產能全被輝達包下來\n目前全部無產能可用，要等到明年底廠房蓋好\n將進入漲價缺貨循環，比照南亞科辦理\n\n #零級分", "parent_author": null, "ret_m1d": -0.028790786948176605, "ret_signed_m1d": -0.028790786948176605, "alpha_spy_m1d": -0.02286308175215568, "alpha_c_m1d": -0.01155587831005711, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.017234908638119495, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017274472168905985, "ret_signed_p1d": 0.017274472168905985, "alpha_spy_p1d": 0.02017148054814344, "alpha_c_p1d": 0.018960106786863773, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0016856346179577875, "ret_p1w": 0.08061420345489445, "ret_signed_p1w": 0.08061420345489445, "alpha_spy_p1w": 0.09241015623351201, "alpha_c_p1w": 0.10259640744005383, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.021982203985159376, "ret_p1m": 0.04798464491362764, "ret_signed_p1m": 0.04798464491362764, "alpha_spy_p1m": 0.052144398036862216, "alpha_c_p1m": 0.05324788425658389, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": -0.00526323934295625, "ret_p3m": -0.07293666026871404, "ret_signed_p3m": -0.07293666026871404, "alpha_spy_p3m": -0.09766090506811076, "alpha_c_p3m": -0.06945769654028233, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": -0.0034789637284317054, "ret_p6m": 1.18042226487524, "ret_signed_p6m": 1.18042226487524, "alpha_spy_p6m": 1.2005462960102375, "alpha_c_p6m": 1.2422020579565374, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": -0.06177979308129733, "price_path": [-0.4453, -0.4491, -0.3954, -0.3359, -0.3474, -0.3436, -0.3551, -0.3743, -0.3417, -0.3282, -0.3071, -0.2917, -0.3205, -0.309, -0.3186, -0.3186, -0.3225, -0.3186, -0.3321, -0.3378, -0.3532, -0.3052, -0.3205, -0.3378, -0.3436, -0.3493, -0.2975, -0.309, -0.3589, -0.3512, -0.3378, -0.286, -0.2821, -0.2111, -0.1324, -0.1555, -0.2284, -0.2438, -0.2687, -0.2879, -0.2802, -0.286, -0.2937, -0.2898, -0.3109, -0.2937, -0.3129, -0.2457, -0.2572, -0.2322, -0.1555, -0.096, -0.0998, -0.1536, -0.1209, -0.1593, -0.1593, -0.1324, -0.1248, -0.1132, -0.025, -0.025, -0.0288, 0.0, 0.0173, 0.0173, 0.0269, -0.0173, 0.0806, 0.0806, 0.0768, 0.0653, 0.0499, 0.0576, 0.0461, 0.0461, 0.0691, 0.0595, 0.1152, 0.0825, 0.1152, 0.094, 0.0749, 0.0441, 0.048, 0.048, 0.1152, 0.1094, 0.0921, 0.1209, 0.0845, 0.0461, 0.0192, -0.0365, 0.0326, -0.0403, -0.0518, -0.0403, -0.0134, 0.0, 0.0077, 0.023, 0.0058, 0.0326, -0.0077, 0.0365, 0.0365, 0.0902, 0.0749, 0.0672, 0.0422, 0.0211, -0.0441, -0.0653, -0.0998, -0.0806, -0.0633, -0.0825, -0.0825, -0.0825, -0.0537, -0.0768, -0.0768, -0.0749, -0.0749, -0.0902, -0.0729, -0.0614, -0.0729, -0.0902, -0.0768, -0.023, 0.0288, 0.0038, 0.0115, 0.1113, 0.2207, 0.3109, 0.3858, 0.4223, 0.4664, 0.4203, 0.5163, 0.5202, 0.4357, 0.4914, 0.4127, 0.5144, 0.5739, 0.5259, 0.4549, 0.5988, 0.572, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.5969, 0.7562, 0.9309, 1.1228, 1.1305, 1.1305, 1.0768, 0.9194, 0.7274, 0.7697, 0.8023, 0.6238, 0.643, 0.8061, 0.7985, 0.977, 1.1727, 1.0883, 1.1766, 1.1536, 1.0768, 1.0499, 0.8848, 1.0729, 1.1843, 1.1536, 1.2265, 1.0038, 1.2035, 1.1804]}
{"unit_id": "thread:1979007496267403352", "ticker_idx": 0, "ts": "2025-10-17T02:12:09+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung has staked its competitiveness on the HBM market and is beginning to see tangible results. The yield of its HBM4-grade 1c DRAM has reportedly reached around 50%", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Samsung's aggressive EUV expansion and 50% HBM4 yield progress bullish for Samsung; ASML tagged as the EUV equipment supplier beneficiary.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Electronics Bets Big on EUV in Memory—Counters SK hynix with Facility Investment\n\nFollowing its adoption of the latest EUV (Extreme Ultraviolet) lithography equipment for its foundry (contract manufacturing) business, Samsung Electronics is now expanding its production capacity in memory as well, having secured five additional EUV lithography machines dedicated to its memory division.\n\nCoinciding with the arrival of the “memory supercycle,” the competition between Samsung and SK hynix is heating up as both companies ramp up investments in semiconductor manufacturing equipment to seize leadership in the next-generation DRAM and high-bandwidth memory (HBM) markets—core enablers of the AI era.\n\nAccording to semiconductor industry sources on the 17th, Samsung Electronics has decided to introduce two of the latest “High-NA (High Numerical Aperture) EUV” machines for its foundry business, and will additionally bring in five regular EUV systems exclusively for its memory division.\n\nThe strategy aims to establish memory-dedicated production lines to maximize efficiency and specialization.\n\nEUV equipment is essential for sub-10-nanometer ultra-fine manufacturing processes. The High-NA EUV machine is a next-generation system used for sub-2-nanometer advanced nodes. Semiconductor manufacturers are competing to draw ever-finer circuit patterns to produce chips that are both higher-performance and lower-power. EUV lithography uses much shorter wavelengths than conventional technologies, allowing for extremely fine patterning and dramatically increasing chip density several-fold.\n\nEach regular EUV machine costs about 300 billion KRW, while a High-NA EUV system is estimated at around 550 billion KRW. Samsung plans to invest roughly 2.6 trillion KRW solely in EUV equipment as part of its preparations for the next upcycle.\n\nFive EUV Systems Dedicated to Memory — A Preemptive Investment for the Supercycle\n\nAn industry insider explained, “Until now, the Pyeongtaek campus shared EUV equipment between the foundry and memory divisions, but recently the direction has changed—five new units will be introduced solely for memory production.”\n\nSamsung has also secured two next-generation High-NA EUV systems, which will be used in its foundry for 2-nanometer processing.\n\n“One unit will be installed at the Hwaseong campus, and the other’s location is yet to be decided,” said the source. “If additional orders arise from major North American clients, the remaining unit may be sent to the Taylor, Texas plant,” indicating a flexible deployment strategy based on customer demand.\n\nSamsung’s aggressive investment move is closely tied to the industry’s entry into a “memory supercycle.”\n\nDuring the previous downturn, memory manufacturers drastically reduced capital expenditures (CAPEX), but as the AI market has exploded, demand for high-performance memory has surged. This has caused severe supply shortages and rapid price increases.\n\nIn fact, Samsung’s production lines are operating at maximum capacity. Another insider said, “The wafer input volume for DRAM at the Pyeongtaek campus has reached its highest level this year,” adding, “The P2D line, which produces 1b (fifth-generation 10-nm) DRAM, is processing about 150,000 wafers per month, while the P1D line, which produces 1a (fourth-generation 10-nm) DRAM, is handling roughly 140,000 per month.”\n\nThe total DRAM production capacity of Samsung’s Pyeongtaek plant is estimated at around 300,000 wafers per month. The recent increase in the contract amount and acceleration of construction timelines for Samsung Electronics’ Pyeongtaek Fab 4 (P4) Phase 4 project—handled by Samsung C&T and Samsung E&A—also reflects this production expansion trend.\n\nTo Be Used in HBM — 50% Yield Achieved, Intensifying Rivalry with SK hynix\n\nRival SK hynix is also significantly increasing the number of EUV machines and planning to expand its DRAM production facilities next year, signaling an even fiercer “equipment war” between the two giants.\n\nThe new EUV equipment introduced by Samsung will play a critical role not only in general DRAM but also in HBM manufacturing.\n\nSamsung has staked its competitiveness on the HBM market and is beginning to see tangible results. The yield of its HBM4-grade 1c (6th-generation 10-nm) DRAM has reportedly reached around 50%. Each new DRAM generation delivers higher performance, and the 1c node is a full generation ahead of the currently commercialized 1b (5th-generation 10-nm) DRAM.\n\n“The P3 line in Pyeongtaek continues to process wafers for HBM4-grade 1c DRAM,” said another insider. “The yield is around 50%, though still volatile depending on process conditions.” While yield stabilization is still in progress, industry observers see this as a major step toward full-scale mass production.\n\nRegarding the additional EUV equipment, Samsung Electronics stated, “We cannot confirm details of that matter.”\n\n$ASML\n\n---\n\nhttps://t.co/Wx0H3VHel8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0020429381828794835, "ret_signed_m1d": -0.0020429381828794835, "alpha_spy_m1d": 0.003601323650557653, "alpha_c_m1d": -0.0002886000640831554, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": -0.001754338118796328, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002042857828167932, "ret_signed_p1d": 0.002042857828167932, "alpha_spy_p1d": -0.008357617571453702, "alpha_c_p1d": -0.009114620941323759, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.011157478769491691, "ret_p1w": 0.009192980758823355, "ret_signed_p1w": 0.009192980758823355, "alpha_spy_p1w": -0.010163090010086862, "alpha_c_p1w": -0.02087611046418192, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.030069091223005273, "ret_p1m": 0.027579102985892723, "ret_signed_p1m": 0.027579102985892723, "alpha_spy_p1m": 0.025652596278781692, "alpha_c_p1m": 0.03238600446336348, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.00480690147747076, "ret_p3m": 0.4400613717143862, "ret_signed_p3m": 0.4400613717143862, "alpha_spy_p3m": 0.3979025459632777, "alpha_c_p3m": 0.4231183029336807, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.016943068780705506, "ret_p6m": 1.067366015147988, "ret_signed_p6m": 1.067366015147988, "alpha_spy_p6m": 1.0288094575970648, "alpha_c_p6m": 1.0427551391320953, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.024610876015892735, "price_path": [-0.3288, -0.3248, -0.3288, -0.3299, -0.2841, -0.2821, -0.2617, -0.2739, -0.2993, -0.2912, -0.2892, -0.3004, -0.2831, -0.2699, -0.278, -0.277, -0.2688, -0.2719, -0.2719, -0.2882, -0.2882, -0.2831, -0.2821, -0.2739, -0.2729, -0.2851, -0.2821, -0.2922, -0.2912, -0.3126, -0.2973, -0.2902, -0.2871, -0.2932, -0.2871, -0.2729, -0.2617, -0.2536, -0.2332, -0.2221, -0.1926, -0.2048, -0.1834, -0.1834, -0.1509, -0.1387, -0.1316, -0.1244, -0.1529, -0.1399, -0.143, -0.1216, -0.0832, -0.0832, -0.0832, -0.0832, -0.0832, -0.0832, -0.0358, -0.047, -0.0644, -0.0296, -0.002, 0.0, 0.002, -0.0041, 0.0072, -0.0143, 0.0092, 0.0419, 0.0163, 0.0266, 0.0633, 0.0981, 0.1348, 0.0715, 0.0276, 0.0133, 0.0, 0.0276, 0.0572, 0.0531, 0.0501, -0.0072, 0.0276, -0.001, -0.0143, 0.0276, -0.0317, -0.0123, 0.0143, 0.0501, 0.0572, 0.0266, 0.0296, 0.0562, 0.0674, 0.0735, 0.1073, 0.1185, 0.1073, 0.1032, 0.096, 0.1124, 0.0705, 0.0501, 0.1021, 0.0991, 0.0858, 0.1287, 0.1389, 0.1348, 0.1348, 0.1951, 0.2266, 0.2307, 0.2307, 0.2307, 0.3189, 0.4175, 0.4257, 0.4472, 0.4247, 0.4267, 0.4247, 0.4123, 0.4401, 0.477, 0.5283, 0.5324, 0.4904, 0.5345, 0.5632, 0.5612, 0.5612, 0.6371, 0.6669, 0.6495, 0.6474, 0.5437, 0.7192, 0.7357, 0.6351, 0.6279, 0.708, 0.7018, 0.7223, 0.8332, 0.8599, 0.8599, 0.8599, 0.8599, 0.9502, 0.9512, 0.981, 1.0528, 1.0888, 1.2376, 1.2222, 1.2222, 1.0025, 0.7675, 0.9666, 0.9317, 0.7808, 0.9286, 0.9502, 0.9286, 0.8835, 0.9368, 0.9902, 1.1401, 1.058, 1.0467, 0.9122, 0.9471, 0.9399, 0.8486, 0.8486, 0.8133, 0.7197, 0.9506, 0.8349, 0.9151, 0.9861, 1.0211, 1.1651, 1.0982, 1.1188, 1.0674]}
{"unit_id": "thread:1979007496267403352", "ticker_idx": 1, "ts": "2025-10-17T02:12:09+00:00", "symbol": "ASML", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$ASML", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Samsung's aggressive EUV expansion and 50% HBM4 yield progress bullish for Samsung; ASML tagged as the EUV equipment supplier beneficiary.", "resolved_tickers": ["ASML"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Electronics Bets Big on EUV in Memory—Counters SK hynix with Facility Investment\n\nFollowing its adoption of the latest EUV (Extreme Ultraviolet) lithography equipment for its foundry (contract manufacturing) business, Samsung Electronics is now expanding its production capacity in memory as well, having secured five additional EUV lithography machines dedicated to its memory division.\n\nCoinciding with the arrival of the “memory supercycle,” the competition between Samsung and SK hynix is heating up as both companies ramp up investments in semiconductor manufacturing equipment to seize leadership in the next-generation DRAM and high-bandwidth memory (HBM) markets—core enablers of the AI era.\n\nAccording to semiconductor industry sources on the 17th, Samsung Electronics has decided to introduce two of the latest “High-NA (High Numerical Aperture) EUV” machines for its foundry business, and will additionally bring in five regular EUV systems exclusively for its memory division.\n\nThe strategy aims to establish memory-dedicated production lines to maximize efficiency and specialization.\n\nEUV equipment is essential for sub-10-nanometer ultra-fine manufacturing processes. The High-NA EUV machine is a next-generation system used for sub-2-nanometer advanced nodes. Semiconductor manufacturers are competing to draw ever-finer circuit patterns to produce chips that are both higher-performance and lower-power. EUV lithography uses much shorter wavelengths than conventional technologies, allowing for extremely fine patterning and dramatically increasing chip density several-fold.\n\nEach regular EUV machine costs about 300 billion KRW, while a High-NA EUV system is estimated at around 550 billion KRW. Samsung plans to invest roughly 2.6 trillion KRW solely in EUV equipment as part of its preparations for the next upcycle.\n\nFive EUV Systems Dedicated to Memory — A Preemptive Investment for the Supercycle\n\nAn industry insider explained, “Until now, the Pyeongtaek campus shared EUV equipment between the foundry and memory divisions, but recently the direction has changed—five new units will be introduced solely for memory production.”\n\nSamsung has also secured two next-generation High-NA EUV systems, which will be used in its foundry for 2-nanometer processing.\n\n“One unit will be installed at the Hwaseong campus, and the other’s location is yet to be decided,” said the source. “If additional orders arise from major North American clients, the remaining unit may be sent to the Taylor, Texas plant,” indicating a flexible deployment strategy based on customer demand.\n\nSamsung’s aggressive investment move is closely tied to the industry’s entry into a “memory supercycle.”\n\nDuring the previous downturn, memory manufacturers drastically reduced capital expenditures (CAPEX), but as the AI market has exploded, demand for high-performance memory has surged. This has caused severe supply shortages and rapid price increases.\n\nIn fact, Samsung’s production lines are operating at maximum capacity. Another insider said, “The wafer input volume for DRAM at the Pyeongtaek campus has reached its highest level this year,” adding, “The P2D line, which produces 1b (fifth-generation 10-nm) DRAM, is processing about 150,000 wafers per month, while the P1D line, which produces 1a (fourth-generation 10-nm) DRAM, is handling roughly 140,000 per month.”\n\nThe total DRAM production capacity of Samsung’s Pyeongtaek plant is estimated at around 300,000 wafers per month. The recent increase in the contract amount and acceleration of construction timelines for Samsung Electronics’ Pyeongtaek Fab 4 (P4) Phase 4 project—handled by Samsung C&T and Samsung E&A—also reflects this production expansion trend.\n\nTo Be Used in HBM — 50% Yield Achieved, Intensifying Rivalry with SK hynix\n\nRival SK hynix is also significantly increasing the number of EUV machines and planning to expand its DRAM production facilities next year, signaling an even fiercer “equipment war” between the two giants.\n\nThe new EUV equipment introduced by Samsung will play a critical role not only in general DRAM but also in HBM manufacturing.\n\nSamsung has staked its competitiveness on the HBM market and is beginning to see tangible results. The yield of its HBM4-grade 1c (6th-generation 10-nm) DRAM has reportedly reached around 50%. Each new DRAM generation delivers higher performance, and the 1c node is a full generation ahead of the currently commercialized 1b (5th-generation 10-nm) DRAM.\n\n“The P3 line in Pyeongtaek continues to process wafers for HBM4-grade 1c DRAM,” said another insider. “The yield is around 50%, though still volatile depending on process conditions.” While yield stabilization is still in progress, industry observers see this as a major step toward full-scale mass production.\n\nRegarding the additional EUV equipment, Samsung Electronics stated, “We cannot confirm details of that matter.”\n\n$ASML\n\n---\n\nhttps://t.co/Wx0H3VHel8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009404717053117517, "ret_signed_m1d": -0.009404717053117517, "alpha_spy_m1d": -0.003760455219680381, "alpha_c_m1d": -0.01077564229416117, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012513710977366088, "ret_signed_p1d": 0.012513710977366088, "alpha_spy_p1d": 0.0021132355777444545, "alpha_c_p1d": -0.0002915326308612354, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": 0.003721040031284817, "ret_signed_p1w": 0.003721040031284817, "alpha_spy_p1w": -0.0156350307376254, "alpha_c_p1w": -0.020518521944126178, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": -0.00724821876520132, "ret_signed_p1m": -0.00724821876520132, "alpha_spy_p1m": -0.00917472547231235, "alpha_c_p1m": 0.00013154333154186126, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.22996098928874598, "ret_signed_p3m": 0.22996098928874598, "alpha_spy_p3m": 0.18780216353763746, "alpha_c_p3m": 0.09365898177129317, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.46204956458704927, "ret_signed_p6m": 0.46204956458704927, "alpha_spy_p6m": 0.423493007036126, "alpha_c_p6m": 0.16484827725851314, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.3163, -0.3052, -0.2973, -0.3107, -0.2926, -0.3019, -0.2991, -0.325, -0.3298, -0.3205, -0.33, -0.3287, -0.3072, -0.2982, -0.2992, -0.2793, -0.2659, -0.2663, -0.2789, -0.2737, -0.2775, -0.2718, -0.2855, -0.2666, -0.267, -0.2585, -0.2519, -0.2583, -0.2785, -0.2785, -0.2948, -0.2841, -0.268, -0.2405, -0.2264, -0.2178, -0.2294, -0.2187, -0.2093, -0.1574, -0.1466, -0.1525, -0.0986, -0.0944, -0.0694, -0.0639, -0.08, -0.0775, -0.0755, -0.0648, -0.0594, -0.0253, 0.0009, 0.0029, 0.0136, -0.0262, -0.0403, -0.0473, -0.0904, -0.0433, -0.0448, -0.0189, -0.0094, 0.0, 0.0125, -0.0041, -0.0172, 0.0069, 0.0037, 0.0298, 0.0225, 0.0422, 0.0467, 0.0309, 0.0383, 0.0026, 0.0159, 0.0017, -0.0102, 0.011, -0.0049, 0.0096, -0.0074, -0.0199, -0.0072, -0.0228, 0.0116, -0.0452, -0.0593, -0.0386, -0.0236, 0.0132, 0.0132, 0.0317, 0.0589, 0.0792, 0.1104, 0.0804, 0.0701, 0.0898, 0.0817, 0.0894, 0.0928, 0.052, 0.0588, 0.0473, -0.0117, 0.0086, 0.0278, 0.0287, 0.0335, 0.0371, 0.0371, 0.0441, 0.0375, 0.0435, 0.0413, 0.0413, 0.1327, 0.1954, 0.209, 0.1957, 0.1624, 0.2398, 0.247, 0.2362, 0.23, 0.296, 0.3223, 0.3223, 0.2906, 0.3238, 0.3577, 0.3519, 0.3756, 0.4157, 0.3849, 0.4163, 0.385, 0.4029, 0.3586, 0.3034, 0.3141, 0.3753, 0.3913, 0.3777, 0.3991, 0.3711, 0.3708, 0.3708, 0.3837, 0.4314, 0.4218, 0.4322, 0.4482, 0.4597, 0.4877, 0.4266, 0.4137, 0.3873, 0.3263, 0.3638, 0.3336, 0.2599, 0.3229, 0.3482, 0.3514, 0.3172, 0.3115, 0.3406, 0.3538, 0.3207, 0.3316, 0.2838, 0.3348, 0.3638, 0.3584, 0.2957, 0.2693, 0.2221, 0.2872, 0.3252, 0.2837, 0.2837, 0.2708, 0.2732, 0.3849, 0.4118, 0.4407, 0.462]}
{"unit_id": "quote:1988064615478263837", "ticker_idx": 0, "ts": "2025-11-11T02:01:55+00:00", "symbol": "NAND memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs has raised its NAND price outlook, projecting a 2.5% supply shortfall in the global NAND market in 2026, with prices up 33% year over year and the uptrend extending through 4Q26", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish NAND outlook: 2.5% supply shortfall in 2026, prices +33% YoY with uptrend through 4Q26.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND memory sector basket: Micron, SK Hynix, Samsung, Kioxia, SanDisk", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Outlook for the 2026 NAND Market — Goldman Sachs\n\nThe supply side faces multiple constraints. First, SK hynix and Micron, due to wafer capacity limitations, will prioritize high-margin DRAM in 2026, resulting in limited NAND expansion. Second, U.S. export controls on manufacturing equipment for Korean memory plants in China have blocked Samsung and SK hynix from upgrading their capacity in China. Third, leading companies are focusing on 3D NAND technology migration—such as moving to higher layer counts—rather than adding new capacity, so global NAND capacity growth in 2026 is expected to be only 5%. Amid this supply–demand mismatch, Goldman Sachs has raised its NAND price outlook, projecting a 2.5% supply shortfall in the global NAND market in 2026, with prices up 33% year over year and the uptrend extending through 4Q26 (previously, the peak was expected in 3Q25).", "reply_text": null, "parent_text": null, "quoted_text": "Insights from Goldman Sachs’ latest memory report:\n\nCurrently, Samsung’s HBM3E price is 5% lower than SK hynix’s, but Samsung may further cut HBM3E prices to regain market share.\n\nHBM pricing trend: HBM3E prices are expected to decline 28% YoY in 2026 (from $15/GB to $10/GB) https://t.co/ZqDB4EHmFn", "parent_author": null, "ret_m1d": -0.00047117259070623697, "ret_signed_m1d": -0.00047117259070623697, "alpha_spy_m1d": 0.0018127488735989906, "alpha_c_m1d": -0.02251934968630249, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010890442377459198, "ret_signed_p1d": 0.010890442377459198, "alpha_spy_p1d": 0.010334043771017121, "alpha_c_p1d": -0.0018712488123739709, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.1014188551768996, "ret_signed_p1w": -0.1014188551768996, "alpha_spy_p1w": -0.06786112601297585, "alpha_c_p1w": -0.0508278121583718, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": -0.06773652016821544, "ret_signed_p1m": -0.06773652016821544, "alpha_spy_p1m": -0.07442768299277194, "alpha_c_p1m": -0.13339672720157328, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.6365431270064692, "ret_signed_p3m": 0.6365431270064692, "alpha_spy_p3m": 0.6223985954808846, "alpha_c_p3m": 0.48884402474167343, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 2.179808322758362, "ret_signed_p6m": 2.179808322758362, "alpha_spy_p6m": 2.099268838094374, "alpha_c_p6m": 1.6088906599743118, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.5961, -0.6012, -0.6061, -0.609, -0.6163, -0.6207, -0.6135, -0.6115, -0.6121, -0.6104, -0.6033, -0.6024, -0.6094, -0.6074, -0.6037, -0.5906, -0.5721, -0.567, -0.5572, -0.5388, -0.512, -0.4877, -0.4815, -0.4644, -0.4727, -0.4493, -0.4518, -0.4406, -0.4311, -0.4376, -0.4456, -0.4625, -0.4342, -0.4304, -0.406, -0.373, -0.3547, -0.3554, -0.3663, -0.3495, -0.3495, -0.3492, -0.3332, -0.3502, -0.3186, -0.2865, -0.2899, -0.2647, -0.2724, -0.2702, -0.2527, -0.1924, -0.1688, -0.1839, -0.1342, -0.1201, -0.1152, -0.0738, -0.1207, -0.1023, -0.0918, -0.0672, -0.0005, 0.0, 0.0109, -0.0303, -0.0874, -0.0441, -0.1014, -0.0992, -0.138, -0.177, -0.1402, -0.1425, -0.1556, -0.1376, -0.1313, -0.1424, -0.1302, -0.1456, -0.1391, -0.1067, -0.0771, -0.0869, -0.0677, -0.072, -0.1032, -0.1332, -0.1513, -0.133, -0.1017, -0.0806, -0.0385, -0.0354, -0.0125, -0.0096, 0.0132, 0.0249, 0.0212, 0.0131, 0.0131, 0.091, 0.1261, 0.2227, 0.2482, 0.2287, 0.2671, 0.2777, 0.2809, 0.2759, 0.3089, 0.3628, 0.3731, 0.3892, 0.4717, 0.512, 0.4867, 0.4674, 0.543, 0.6242, 0.6409, 0.698, 0.6919, 0.7938, 0.6815, 0.6197, 0.6365, 0.6443, 0.5994, 0.6714, 0.7626, 0.7857, 0.7774, 0.7421, 0.7525, 0.7838, 0.8216, 0.834, 0.868, 0.8714, 0.9262, 0.8957, 0.8885, 0.7206, 0.6748, 0.7306, 0.6631, 0.6376, 0.7556, 0.8325, 0.7802, 0.8112, 0.9095, 0.9317, 1.0397, 0.9936, 0.9259, 0.8422, 0.8531, 0.8458, 0.7119, 0.7091, 0.618, 0.6243, 0.8055, 0.7564, 0.798, 0.8556, 0.8687, 1.0753, 1.1171, 1.1668, 1.2575, 1.3717, 1.3052, 1.3674, 1.3045, 1.3022, 1.3573, 1.4726, 1.4565, 1.4884, 1.6166, 1.5604, 1.6232, 1.6518, 1.7218, 1.8764, 2.0401, 2.1798]}
{"unit_id": "orig:1988401761187754440", "ticker_idx": 0, "ts": "2025-11-12T00:21:37+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "expectations seem to have run even further ahead... To keep enjoying a valuation premium over Nvidia as it does now, I think the company needs to show more", "tweet_type": "analysis", "timeline": "unspecified", "summary": "AMD Analyst Day guidance failed to exceed already-elevated AI expectations; needs more catalysts (e.g. new mega-scale customers) to justify current premium valuation — implied cautious/bearish lean.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AMD: Why didn’t Analyst Day get the dopamine flowing?\n\nHello, this is Moon Junho from Samsung Securities.”\n\nAMD held its Financial Analyst Day overnight and presented its mid- to long-term outlook.\n\nWith bubble concerns resurfacing across the sector, the stock fell during regular trading and was up about 4% after hours; in the end, that amounts to only a 1–2% gain versus yesterday’s close. In other words, it didn’t get the dopamine flowing.\n\nCompared with consensus:\n\n■ AMD’s medium- to long-term guidance for the next 3–5 years\n    1.    Revenue CAGR +35%\n1-1) Data center revenue CAGR +60%\n1-2) AI revenue CAGR +80%\n    2.    Gross margin 55–58%\n    3.    Operating margin 35% or higher\n    4.    EPS above $20\n\nThey offered a somewhat ambiguous “3 to 5 years (?)” window, but since there isn’t any compiled consensus for 2030 anyway, let’s assume it refers to 2028, three years out.\n\nThe 2025–2028E overall revenue CAGR consensus (FactSet) is around 29%, which is above expectations. The 2028 EPS consensus is only $10, which likewise makes this a surprise.\n\nHowever, we’ve been saying the stock’s narrative is AI. The 2025–2028E AI GPU revenue CAGR consensus is already 82%, so it’s hard to say this meaningfully exceeds market expectations. For reference, the 2028 gross margin consensus has already climbed to 60%.\n\nOf course, in absolute terms, this is undeniably an impressive growth outlook, but expectations seem to have run even further ahead.\n\nTo keep enjoying a valuation premium over Nvidia as it does now, I think the company needs to show more, such as securing new mega-scale customers.\n\nThank you.\n\n(Published material on 2025/11/12)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0825447455167253, "ret_signed_m1d": 0.0825447455167253, "alpha_spy_m1d": 0.08198865631753871, "alpha_c_m1d": 0.06994386289957166, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04221873118363728, "ret_signed_p1d": 0.04221873118363728, "alpha_spy_p1d": 0.02562470971289943, "alpha_c_p1d": 0.01209474752237849, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.13650588897632854, "ret_signed_p1w": 0.13650588897632854, "alpha_spy_p1w": 0.10614221508334665, "alpha_c_p1w": 0.09124956655698668, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": -0.1446947346483235, "ret_signed_p1m": 0.1446947346483235, "alpha_spy_p1m": 0.1531673028606445, "alpha_c_p1m": 0.18789790837105114, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": -0.16566886407990844, "ret_signed_p3m": 0.16566886407990844, "alpha_spy_p3m": 0.18413675614884228, "alpha_c_p3m": 0.31301321590616027, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 0.5777355955954917, "ret_signed_p6m": -0.5777355955954917, "alpha_spy_p6m": -0.501108192604808, "alpha_c_p6m": -0.053868226332070446, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.3143, -0.3196, -0.3567, -0.3619, -0.3676, -0.352, -0.369, -0.3564, -0.3544, -0.3488, -0.3718, -0.3718, -0.373, -0.3737, -0.3751, -0.4162, -0.4152, -0.3981, -0.3838, -0.3987, -0.3875, -0.3775, -0.3802, -0.3852, -0.39, -0.3921, -0.3828, -0.3785, -0.3786, -0.3771, -0.3841, -0.3767, -0.3751, -0.3665, -0.3444, -0.3639, -0.2131, -0.183, -0.0901, -0.1004, -0.1699, -0.164, -0.1576, -0.0784, -0.094, -0.0997, -0.0708, -0.0806, -0.1107, -0.0923, -0.0231, 0.003, -0.0034, 0.021, -0.0156, -0.0107, 0.0029, -0.0341, -0.0099, -0.0818, -0.0979, -0.0576, -0.0825, 0.0, -0.0422, -0.0467, -0.071, -0.1105, -0.1365, -0.2042, -0.2129, -0.1693, -0.2038, -0.1725, -0.1725, -0.1598, -0.1511, -0.1686, -0.1595, -0.1657, -0.1581, -0.1459, -0.144, -0.1447, -0.1447, -0.1858, -0.1982, -0.1921, -0.2348, -0.2234, -0.1756, -0.1697, -0.1699, -0.1694, -0.1694, -0.1696, -0.1672, -0.1682, -0.1728, -0.1728, -0.1368, -0.146, -0.172, -0.1888, -0.2094, -0.2152, -0.1978, -0.1465, -0.1363, -0.1196, -0.1045, -0.1045, -0.1042, -0.0351, -0.0199, 0.0031, -0.0293, -0.0265, -0.0238, -0.0259, -0.0856, -0.0487, -0.0648, -0.2267, -0.2564, -0.1949, -0.1657, -0.1751, -0.175, -0.2045, -0.1992, -0.1992, -0.2156, -0.227, -0.2145, -0.2269, -0.2406, -0.174, -0.1855, -0.2133, -0.2267, -0.2328, -0.2624, -0.2195, -0.2296, -0.2567, -0.2171, -0.215, -0.2088, -0.2362, -0.253, -0.2407, -0.2417, -0.2296, -0.2071, -0.2223, -0.2171, -0.2067, -0.1492, -0.2129, -0.2198, -0.2428, -0.2142, -0.188, -0.1599, -0.1599, -0.1495, -0.1443, -0.1046, -0.0859, -0.0535, -0.0466, -0.0148, -0.003, 0.0748, 0.0753, 0.062, 0.0989, 0.1722, 0.1794, 0.3435, 0.2926, 0.2484, 0.3021, 0.3693, 0.3926, 0.3192, 0.3722, 0.6277, 0.5777]}
{"unit_id": "orig:1983815741125337538", "ticker_idx": 0, "ts": "2025-10-30T08:38:24+00:00", "symbol": "NAND memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still haven't bought any pure NAND companies, and I probably never will", "tweet_type": "analysis", "timeline": "years", "summary": "Avoids all pure NAND companies permanently; NAND boom structurally unsustainable due to fragmented supply base, Kioxia/SanDisk CAPEX overhang, and no HBM-like capacity absorber until HBF ~2027–28.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND memory sector basket: Micron, SK Hynix, Samsung, Kioxia, SanDisk", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "A Random Thought — How Long Will the NAND Boom Last?\n\nEveryone probably knows that HDD lead times have been pushed out to more than 52 weeks.\n\nBecause of that, hyperscalers are gradually shifting from HDDs to SSDs — specifically, to QLC SSDs.\n\nIt’s gotten to the point where there are even rumors that 1TB QLC SSDs are being completely swept up.\n\nHowever, what I want to talk about today is this:\n\nUnlike DRAM, I believe the NAND boom cannot last for several years.\nLet me explain why.\n\n• There are simply too many players\n\nIn the DRAM cycle, we’ve seen companies reduce investment during downturns and over-invest during booms — and that’s what created the cycle.\n\nAs I posted recently, the management teams of Samsung, SK hynix, and Micron have finally learned their lesson:\n\nIt’s more profitable in total to invest less during booms and thereby extend the cycle.\n\nThe three players are now indirectly aligned in this behavior.\nEven if they increase CAPEX, there’s the “magical product” called HBM, which absorbs most of the incremental capacity.\n\n(Meaning: since a large portion of DRAM wafers are consumed in producing HBM, the supply of commodity DRAM has actually decreased.)\nHowever, in NAND, the situation is different — there are not only the three memory giants but also SanDisk, Kioxia, and YMTC, which are pure NAND players.\nThese companies, even when they make money, have nowhere else to invest but NAND.\n\nIn other words, unlike the three memory makers who can also allocate CAPEX to DRAM, these companies have no alternative investment targets.\n\nFor Kioxia or SanDisk, increasing NAND CAPEX is inevitable.\nSince NAND is their only business, not investing in it would raise questions about the company’s very purpose of existence.\n\nIn fact, Kioxia has publicly stated that it plans to double its CAPEX within the next five years.\nIf that happens, do you think Samsung, SK hynix, and Micron will just sit back and watch their market share erode?\n\nI don’t think so.\n\nEventually, the three memory makers that also produce DRAM will have to increase NAND investment as well —\nand that means the traditional memory-cycle dynamics will inevitably reappear in NAND, at least until HBF arrives.\n\nIf HBF is commercialized, this story could change.\nIf HBF can, like HBM, consume a large share of NAND wafers and thereby reduce the supply of commodity NAND,\n\nthen the NAND industry could be re-rated.\n\nHowever, the commercialization of HBF is expected around 2027–2028.\nIf the AI bubble bursts before then, NAND will likely fall back into another state of oversupply.\nThat’s why I still haven’t bought any pure NAND companies, and I probably never will.\n\nThis is not investment advice.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0144219522076243, "ret_signed_m1d": 0.0144219522076243, "alpha_spy_m1d": 0.025542401439029284, "alpha_c_m1d": 0.027838428287630702, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.013416476080006401, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0053576764997617324, "ret_signed_p1d": -0.0053576764997617324, "alpha_spy_p1d": -0.00207739588904623, "alpha_c_p1d": -0.0073097484784813375, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": -0.001952071978719605, "ret_p1w": 0.03530650884273987, "ret_signed_p1w": -0.03530650884273987, "alpha_spy_p1w": -0.04930995724563234, "alpha_c_p1w": -0.07005758137137916, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.03475107252863929, "ret_p1m": -0.00839630037830299, "ret_signed_p1m": 0.00839630037830299, "alpha_spy_p1m": 0.013632921203507143, "alpha_c_p1m": -0.02308311163344543, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.03147941201174842, "ret_p3m": 0.7864737100416463, "ret_signed_p3m": -0.7864737100416463, "alpha_spy_p3m": -0.7604155689865102, "alpha_c_p3m": -0.6633406723890614, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": 0.12313303765258499, "ret_p6m": 1.9453931063882386, "ret_signed_p6m": -1.9453931063882386, "alpha_spy_p6m": -1.8892390904868683, "alpha_c_p6m": -1.5487092553611073, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.39668385102713133, "price_path": [-0.5927, -0.59, -0.5947, -0.5895, -0.5777, -0.5723, -0.5596, -0.5585, -0.5595, -0.5655, -0.5705, -0.5738, -0.5821, -0.5867, -0.5787, -0.5765, -0.5766, -0.5747, -0.5662, -0.5646, -0.5717, -0.5702, -0.5659, -0.5496, -0.5273, -0.5213, -0.5109, -0.4897, -0.4576, -0.4304, -0.4229, -0.4041, -0.4128, -0.3866, -0.3885, -0.3763, -0.3653, -0.3737, -0.3838, -0.402, -0.3674, -0.3631, -0.3355, -0.2983, -0.2761, -0.2781, -0.2907, -0.2702, -0.27, -0.2734, -0.252, -0.272, -0.234, -0.1986, -0.2038, -0.1736, -0.1818, -0.1796, -0.1557, -0.0838, -0.0589, -0.0759, -0.0144, 0.0, 0.0054, 0.0514, -0.0013, 0.0238, 0.0353, 0.0702, 0.1509, 0.1516, 0.1661, 0.1123, 0.0478, 0.0991, 0.031, 0.035, -0.0178, -0.0607, -0.0153, -0.02, -0.0389, -0.0182, -0.0084, -0.0245, -0.0121, -0.0318, -0.0205, 0.0181, 0.0507, 0.0384, 0.0614, 0.0591, 0.0178, -0.0164, -0.035, -0.0153, 0.0218, 0.0484, 0.0956, 0.0994, 0.1266, 0.1301, 0.155, 0.1645, 0.1591, 0.1497, 0.1497, 0.2409, 0.2789, 0.4007, 0.4309, 0.4067, 0.457, 0.4712, 0.4769, 0.4698, 0.5105, 0.5709, 0.583, 0.6099, 0.7117, 0.758, 0.7229, 0.7015, 0.7865, 0.8852, 0.9068, 0.9812, 0.9902, 1.1093, 0.963, 0.8931, 0.9154, 0.919, 0.8611, 0.9514, 1.0597, 1.0867, 1.0767, 1.0306, 1.0421, 1.0788, 1.1249, 1.1417, 1.1744, 1.1745, 1.2355, 1.1989, 1.1882, 0.9954, 0.9551, 1.0069, 0.9254, 0.9099, 1.0451, 1.1396, 1.075, 1.1184, 1.2363, 1.2612, 1.3867, 1.3403, 1.2533, 1.1615, 1.1722, 1.1616, 0.9989, 0.997, 0.8889, 0.9092, 1.1198, 1.069, 1.1142, 1.1823, 1.1938, 1.4393, 1.5016, 1.5599, 1.6837, 1.8118, 1.7223, 1.7976, 1.7235, 1.7196, 1.7808, 1.9266, 1.8987, 1.9454]}
{"unit_id": "orig:1983815741125337538", "ticker_idx": 1, "ts": "2025-10-30T08:38:24+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still haven't bought any pure NAND companies, and I probably never will", "tweet_type": "analysis", "timeline": "years", "summary": "Avoids all pure NAND companies permanently; NAND boom structurally unsustainable due to fragmented supply base, Kioxia/SanDisk CAPEX overhang, and no HBM-like capacity absorber until HBF ~2027–28.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "A Random Thought — How Long Will the NAND Boom Last?\n\nEveryone probably knows that HDD lead times have been pushed out to more than 52 weeks.\n\nBecause of that, hyperscalers are gradually shifting from HDDs to SSDs — specifically, to QLC SSDs.\n\nIt’s gotten to the point where there are even rumors that 1TB QLC SSDs are being completely swept up.\n\nHowever, what I want to talk about today is this:\n\nUnlike DRAM, I believe the NAND boom cannot last for several years.\nLet me explain why.\n\n• There are simply too many players\n\nIn the DRAM cycle, we’ve seen companies reduce investment during downturns and over-invest during booms — and that’s what created the cycle.\n\nAs I posted recently, the management teams of Samsung, SK hynix, and Micron have finally learned their lesson:\n\nIt’s more profitable in total to invest less during booms and thereby extend the cycle.\n\nThe three players are now indirectly aligned in this behavior.\nEven if they increase CAPEX, there’s the “magical product” called HBM, which absorbs most of the incremental capacity.\n\n(Meaning: since a large portion of DRAM wafers are consumed in producing HBM, the supply of commodity DRAM has actually decreased.)\nHowever, in NAND, the situation is different — there are not only the three memory giants but also SanDisk, Kioxia, and YMTC, which are pure NAND players.\nThese companies, even when they make money, have nowhere else to invest but NAND.\n\nIn other words, unlike the three memory makers who can also allocate CAPEX to DRAM, these companies have no alternative investment targets.\n\nFor Kioxia or SanDisk, increasing NAND CAPEX is inevitable.\nSince NAND is their only business, not investing in it would raise questions about the company’s very purpose of existence.\n\nIn fact, Kioxia has publicly stated that it plans to double its CAPEX within the next five years.\nIf that happens, do you think Samsung, SK hynix, and Micron will just sit back and watch their market share erode?\n\nI don’t think so.\n\nEventually, the three memory makers that also produce DRAM will have to increase NAND investment as well —\nand that means the traditional memory-cycle dynamics will inevitably reappear in NAND, at least until HBF arrives.\n\nIf HBF is commercialized, this story could change.\nIf HBF can, like HBM, consume a large share of NAND wafers and thereby reduce the supply of commodity NAND,\n\nthen the NAND industry could be re-rated.\n\nHowever, the commercialization of HBF is expected around 2027–2028.\nIf the AI bubble bursts before then, NAND will likely fall back into another state of oversupply.\nThat’s why I still haven’t bought any pure NAND companies, and I probably never will.\n\nThis is not investment advice.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07522935779816509, "ret_signed_m1d": 0.07522935779816509, "alpha_spy_m1d": 0.08634980702957007, "alpha_c_m1d": 0.08864583387817149, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.013416476080006401, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006880733944954143, "ret_signed_p1d": 0.006880733944954143, "alpha_spy_p1d": 0.010161014555669645, "alpha_c_p1d": 0.004928661966234538, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": -0.001952071978719605, "ret_p1w": 0.05504587155963292, "ret_signed_p1w": -0.05504587155963292, "alpha_spy_p1w": -0.06904931996252539, "alpha_c_p1w": -0.08979694408827221, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.03475107252863929, "ret_p1m": -0.13706422018348619, "ret_signed_p1m": 0.13706422018348619, "alpha_spy_p1m": 0.14230084100869034, "alpha_c_p1m": 0.10558480817173777, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.03147941201174842, "ret_p3m": 0.6926605504587156, "ret_signed_p3m": -0.6926605504587156, "alpha_spy_p3m": -0.6666024094035794, "alpha_c_p3m": -0.5695275128061306, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": 0.12313303765258499, "ret_p6m": 2.1724770642201836, "ret_signed_p6m": -2.1724770642201836, "alpha_spy_p6m": -2.1163230483188133, "alpha_c_p6m": -1.7757932131930523, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.39668385102713133, "price_path": [-0.7815, -0.7826, -0.786, -0.789, -0.7831, -0.7831, -0.7598, -0.7635, -0.7656, -0.7651, -0.776, -0.7731, -0.7837, -0.7885, -0.7795, -0.7806, -0.7768, -0.7784, -0.7733, -0.7618, -0.7545, -0.76, -0.7594, -0.7591, -0.7188, -0.711, -0.7087, -0.6803, -0.6326, -0.5927, -0.5927, -0.5683, -0.5927, -0.5853, -0.5844, -0.5578, -0.5578, -0.5459, -0.5514, -0.5968, -0.5711, -0.5528, -0.5683, -0.5037, -0.433, -0.4202, -0.4587, -0.4606, -0.4294, -0.4349, -0.4349, -0.4495, -0.4202, -0.3688, -0.3982, -0.344, -0.3541, -0.3294, -0.3239, -0.1945, -0.1, -0.1394, -0.0752, 0.0, -0.0069, -0.0069, -0.0128, -0.0326, 0.055, 0.1046, 0.222, 0.2106, 0.2147, 0.195, -0.0803, 0.0243, -0.0587, -0.006, 0.0399, -0.0798, -0.0798, -0.0961, -0.2306, -0.1698, -0.1371, -0.1851, -0.1544, -0.1733, -0.1679, -0.1349, -0.0688, -0.0941, -0.1312, -0.1183, -0.0927, -0.156, -0.2029, -0.1468, -0.1275, -0.1431, -0.0711, -0.087, -0.0271, -0.0096, 0.0472, -0.0206, -0.0427, -0.0427, -0.0427, -0.0427, 0.0413, 0.0642, 0.1642, 0.1927, 0.1642, 0.1642, 0.2555, 0.2183, 0.2505, 0.3532, 0.3954, 0.395, 0.5138, 0.6431, 0.5904, 0.6009, 0.6927, 0.7381, 0.7656, 0.9596, 0.6844, 0.9069, 0.9174, 0.7917, 0.7518, 0.7339, 0.7289, 0.7289, 0.9427, 1.0959, 1.0459, 1.0523, 0.9651, 0.9495, 0.8862, 0.8862, 1.0431, 0.9693, 0.9486, 0.9459, 0.9739, 0.8532, 0.7656, 0.8624, 0.8326, 0.6541, 0.7954, 0.9615, 0.9491, 0.9326, 1.0748, 0.9826, 1.1459, 1.0514, 1.0514, 0.9693, 0.9358, 1.0592, 0.9417, 0.8596, 0.8367, 0.7505, 1.0, 0.9353, 1.0046, 1.0917, 1.1349, 1.5321, 1.5413, 1.7651, 1.8688, 2.211, 1.9734, 2.1073, 1.8009, 1.7991, 2.0037, 2.1927, 2.2459, 2.1725]}
{"unit_id": "orig:1983815741125337538", "ticker_idx": 2, "ts": "2025-10-30T08:38:24+00:00", "symbol": "SanDisk", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still haven't bought any pure NAND companies, and I probably never will", "tweet_type": "analysis", "timeline": "years", "summary": "Avoids all pure NAND companies permanently; NAND boom structurally unsustainable due to fragmented supply base, Kioxia/SanDisk CAPEX overhang, and no HBM-like capacity absorber until HBF ~2027–28.", "resolved_tickers": ["SNDK"], "resolution_method": "exchange_ticker", "resolution_reason": "SanDisk (spun off from WD) listed on Nasdaq as SNDK", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "A Random Thought — How Long Will the NAND Boom Last?\n\nEveryone probably knows that HDD lead times have been pushed out to more than 52 weeks.\n\nBecause of that, hyperscalers are gradually shifting from HDDs to SSDs — specifically, to QLC SSDs.\n\nIt’s gotten to the point where there are even rumors that 1TB QLC SSDs are being completely swept up.\n\nHowever, what I want to talk about today is this:\n\nUnlike DRAM, I believe the NAND boom cannot last for several years.\nLet me explain why.\n\n• There are simply too many players\n\nIn the DRAM cycle, we’ve seen companies reduce investment during downturns and over-invest during booms — and that’s what created the cycle.\n\nAs I posted recently, the management teams of Samsung, SK hynix, and Micron have finally learned their lesson:\n\nIt’s more profitable in total to invest less during booms and thereby extend the cycle.\n\nThe three players are now indirectly aligned in this behavior.\nEven if they increase CAPEX, there’s the “magical product” called HBM, which absorbs most of the incremental capacity.\n\n(Meaning: since a large portion of DRAM wafers are consumed in producing HBM, the supply of commodity DRAM has actually decreased.)\nHowever, in NAND, the situation is different — there are not only the three memory giants but also SanDisk, Kioxia, and YMTC, which are pure NAND players.\nThese companies, even when they make money, have nowhere else to invest but NAND.\n\nIn other words, unlike the three memory makers who can also allocate CAPEX to DRAM, these companies have no alternative investment targets.\n\nFor Kioxia or SanDisk, increasing NAND CAPEX is inevitable.\nSince NAND is their only business, not investing in it would raise questions about the company’s very purpose of existence.\n\nIn fact, Kioxia has publicly stated that it plans to double its CAPEX within the next five years.\nIf that happens, do you think Samsung, SK hynix, and Micron will just sit back and watch their market share erode?\n\nI don’t think so.\n\nEventually, the three memory makers that also produce DRAM will have to increase NAND investment as well —\nand that means the traditional memory-cycle dynamics will inevitably reappear in NAND, at least until HBF arrives.\n\nIf HBF is commercialized, this story could change.\nIf HBF can, like HBM, consume a large share of NAND wafers and thereby reduce the supply of commodity NAND,\n\nthen the NAND industry could be re-rated.\n\nHowever, the commercialization of HBF is expected around 2027–2028.\nIf the AI bubble bursts before then, NAND will likely fall back into another state of oversupply.\nThat’s why I still haven’t bought any pure NAND companies, and I probably never will.\n\nThis is not investment advice.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0436114440134463, "ret_signed_m1d": -0.0436114440134463, "alpha_spy_m1d": -0.032490994782041316, "alpha_c_m1d": -0.03116093939184017, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.012450504621606129, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017924595932756082, "ret_signed_p1d": -0.017924595932756082, "alpha_spy_p1d": -0.01464431532204058, "alpha_c_p1d": -0.016959239947039517, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": 0.000965355985716565, "ret_p1w": 0.06061686586261006, "ret_signed_p1w": -0.06061686586261006, "alpha_spy_p1w": -0.07462031426550253, "alpha_c_p1w": -0.09800159058130598, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.037384724718695916, "ret_p1m": 0.1402307753446903, "ret_signed_p1m": -0.1402307753446903, "alpha_spy_p1m": -0.13499415451948615, "alpha_c_p1m": -0.18740292207391018, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.04717214672921988, "ret_p3m": 1.4585332175924122, "ret_signed_p3m": -1.4585332175924122, "alpha_spy_p3m": -1.432475076537276, "alpha_c_p3m": -1.471319600082835, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": -0.012786382490422854, "ret_p6m": 4.055152626846179, "ret_signed_p6m": -4.055152626846179, "alpha_spy_p6m": -3.998998610944809, "alpha_c_p6m": -3.9854597518579453, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.06969287498823351, "price_path": [-0.7829, -0.7859, -0.785, -0.7922, -0.7736, -0.7785, -0.7609, -0.7599, -0.7616, -0.7725, -0.7675, -0.7723, -0.7733, -0.7676, -0.7632, -0.7611, -0.7582, -0.7526, -0.7402, -0.732, -0.732, -0.7392, -0.7293, -0.6808, -0.6499, -0.64, -0.6399, -0.6225, -0.5695, -0.5602, -0.5399, -0.5325, -0.5201, -0.4951, -0.478, -0.4744, -0.4566, -0.4902, -0.5185, -0.504, -0.4204, -0.427, -0.3815, -0.3661, -0.3442, -0.3812, -0.3823, -0.3265, -0.3378, -0.403, -0.3126, -0.35, -0.2631, -0.2633, -0.2842, -0.244, -0.2376, -0.2496, -0.1469, -0.0493, -0.0987, -0.1036, 0.0436, 0.0, 0.0179, 0.0571, -0.0064, 0.1056, 0.0606, 0.223, 0.3683, 0.3869, 0.4457, 0.2438, 0.2979, 0.3578, 0.2508, 0.2561, 0.0007, 0.0227, 0.159, 0.126, 0.0982, 0.0982, 0.1402, 0.0733, 0.0487, -0.0074, 0.0893, 0.1667, 0.1514, 0.1207, 0.1892, 0.2338, 0.0529, 0.0309, 0.0689, 0.0562, 0.1207, 0.2134, 0.231, 0.2506, 0.2771, 0.2771, 0.2769, 0.2473, 0.2267, 0.2122, 0.2122, 0.4056, 0.3997, 0.7855, 0.8055, 0.7084, 0.9273, 0.9879, 0.9907, 0.9804, 1.0899, 1.1122, 1.1122, 1.314, 1.56, 1.5709, 1.4197, 1.4042, 1.4585, 1.6945, 1.7541, 1.9428, 2.3972, 2.5518, 1.9851, 1.9425, 2.0536, 1.9793, 1.766, 2.0607, 2.2187, 2.1997, 2.1997, 2.016, 2.0661, 2.1717, 2.3192, 2.4036, 2.2607, 2.2294, 2.3291, 2.2446, 2.1615, 1.8874, 2.0592, 1.8883, 1.6929, 2.0065, 2.1605, 2.3471, 2.1601, 2.3787, 2.5932, 2.6777, 2.8489, 2.9429, 2.6243, 2.5874, 2.5874, 2.4616, 2.0802, 2.1449, 1.9236, 2.2445, 2.5376, 2.5828, 2.5828, 2.7005, 2.6299, 2.9878, 3.3487, 3.3498, 3.8642, 3.8231, 3.5538, 3.6955, 3.7032, 3.6625, 3.6139, 3.9998, 3.7617, 4.0552]}
{"unit_id": "orig:1982442576104886277", "ticker_idx": 0, "ts": "2025-10-26T13:41:56+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm considering selling my AMAT holdings", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Considering selling AMAT holdings; disappointed by the small WFE opportunity ($8B per $100B AI capex) implied by Lam Research's presentation.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "One of the most striking parts of Lam Research’s recent presentation was their statement that every $100 billion increase in AI investment creates about an $8 billion market opportunity in WFE.\n\nHonestly, is that considered small, or is it a satisfactory figure?\n\nNo matter how much higher margins the companies selling shovels (like fabless firms such as NVIDIA) enjoy compared to those making shovels (semiconductor equipment makers), this number still feels somewhat disappointing to me.\n\nI’m considering selling my AMAT holdings.\n\nWhat do you all think?", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011152988761674698, "ret_signed_m1d": 0.011152988761674698, "alpha_spy_m1d": -0.0005071601820887395, "alpha_c_m1d": -0.013295451473713693, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01595131680056161, "ret_signed_p1d": 0.01595131680056161, "alpha_spy_p1d": 0.018607315757152687, "alpha_c_p1d": 0.02475831068587886, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.02757961976718737, "ret_signed_p1w": -0.02757961976718737, "alpha_spy_p1w": -0.030352339352460955, "alpha_c_p1w": -0.010465635058316991, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": 0.05016751281075793, "ret_signed_p1m": -0.05016751281075793, "alpha_spy_p1m": -0.06508204428716724, "alpha_c_p1m": -0.1055378127479577, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.3807758394622629, "ret_signed_p3m": -0.3807758394622629, "alpha_spy_p3m": -0.3723468722508163, "alpha_c_p3m": -0.25816256950517524, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 0.7100923901095952, "ret_signed_p6m": -0.7100923901095952, "alpha_spy_p6m": -0.676747823647454, "alpha_c_p6m": -0.41513812605899036, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.1836, -0.2239, -0.2242, -0.212, -0.2278, -0.2321, -0.2105, -0.2031, -0.2052, -0.1877, -0.1809, -0.1886, -0.3027, -0.2951, -0.3008, -0.3062, -0.309, -0.2976, -0.2997, -0.2889, -0.2894, -0.2856, -0.3051, -0.3051, -0.3189, -0.3246, -0.316, -0.2965, -0.2995, -0.2932, -0.2936, -0.2645, -0.2746, -0.2611, -0.2498, -0.23, -0.1797, -0.1782, -0.1332, -0.1317, -0.1292, -0.1372, -0.1185, -0.114, -0.1149, -0.0587, -0.0335, -0.0597, -0.0321, -0.0855, -0.0597, -0.0477, -0.0924, -0.0512, -0.0568, -0.0162, -0.0156, -0.0274, -0.0138, -0.023, -0.0466, -0.0124, -0.0112, 0.0, -0.016, 0.0191, 0.0053, 0.0077, 0.0276, -0.0049, 0.0413, 0.0095, -0.0054, 0.0162, -0.0115, -0.0026, -0.035, -0.023, -0.0113, -0.0268, 0.0164, -0.0461, -0.0297, 0.0001, 0.0502, 0.0827, 0.0827, 0.0926, 0.1034, 0.1492, 0.1635, 0.167, 0.1608, 0.1615, 0.1571, 0.1918, 0.1699, 0.1227, 0.1316, 0.1211, 0.0753, 0.098, 0.1106, 0.1219, 0.1271, 0.1295, 0.1295, 0.1344, 0.1393, 0.126, 0.1131, 0.1131, 0.1646, 0.2315, 0.2821, 0.2656, 0.2199, 0.3045, 0.3307, 0.3205, 0.3076, 0.382, 0.4164, 0.4164, 0.3784, 0.4087, 0.3808, 0.3963, 0.3837, 0.4411, 0.4586, 0.4784, 0.3961, 0.4224, 0.3803, 0.289, 0.3167, 0.3969, 0.4318, 0.4253, 0.4721, 0.4224, 0.5372, 0.5372, 0.5555, 0.5995, 0.6038, 0.6279, 0.62, 0.639, 0.7128, 0.6294, 0.6146, 0.614, 0.5236, 0.5515, 0.5028, 0.4083, 0.4699, 0.5, 0.5225, 0.4626, 0.4811, 0.5013, 0.5285, 0.5155, 0.5491, 0.5485, 0.569, 0.6219, 0.6017, 0.4682, 0.4622, 0.4013, 0.4822, 0.5343, 0.5112, 0.5112, 0.5292, 0.5365, 0.6728, 0.7252, 0.7325, 0.7162, 0.7158, 0.7098, 0.6909, 0.7214, 0.6983, 0.7101]}
{"unit_id": "orig:1983794079109652612", "ticker_idx": 0, "ts": "2025-10-30T07:12:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the price target is raised again—from USD 205 to USD 235—and the report notes that the stock's setup ahead of earnings in three weeks looks structurally attractive", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses UBS Buy on NVDA, PT raised $205→$235 on EPS upgrades, CoWoS capacity, and Blackwell ramp; MU cited as HBM4 supply partner beneficiary.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s Dissection of NVIDIA’s GTC\n\nAside from several new partnerships (detailed below), NVIDIA’s (NVDA) latest demand guidance indicates that the market’s current earnings expectations remain significantly underestimated.\nThis view is also supported by recent findings from supply chain checks.\nAs a result, the report raises its 2026 EPS forecast to approximately USD 7.72 and 2027 EPS forecast to USD 9.42, projecting data-center revenue to reach about USD 377 billion in 2027 (versus roughly USD 190 billion this year).\n\nHowever, current production capacity and HBM support still appear insufficient to sustain the shipment volume and revenue scale implied by NVIDIA (details below).\nOne possible reason is that NVIDIA’s recognized GPU shipments for this year are expected to be roughly 15% lower than actual production, mainly to allow channel inventories to gradually clear before GB300 cabinet volumes ramp up.\nAfter adjusting for this factor, performance could align more closely with NVIDIA’s implied expectations and potentially lift 2026 EPS further to around USD 8.75.\nNVIDIA also suggested that if customers place orders sooner—allowing delivery cycles to accommodate additional shipments before the end of 2026—there remains further upside potential.\n\nThe upward forecast revision is based on several drivers:\nNVIDIA’s 2026 CoWoS capacity allocation has increased by about 10% compared with earlier plans, while HBM4 supply agreements with SK Hynix and Micron (MU) are still under negotiation.\nThe new model assumes GPU shipments of about 6 million units in 2025 and 8 million units in 2026, consistent with UBS Global Semiconductor Team’s capacity estimates, but not yet fully reflecting several billion USD of inventory on NVIDIA’s balance sheet, which could provide additional upside.\n\nIn addition, the model does not yet include the China market factor.\nIf a future Trump administration allows export to China of GPUs with roughly 80% of the B200’s performance—representing a USD 50 billion annual market of which Chinese domestic suppliers can meet only USD 10–20 billion—then the China market could become a major upside “wild card.”\n\nBeyond growth assumptions, the key investor concern now is whether NVIDIA’s high profit margins can remain sustainable amid the rapid development of custom ASICs.\nThe report concludes that such margins are likely to remain sustainable for the foreseeable future, owing to the company’s strong lead in cabinet-level ecosystem integration and explosive market demand growth.\nBased on higher earnings expectations, the price target is raised again—from USD 205 to USD 235—and the report notes that the stock’s setup ahead of earnings in three weeks looks structurally attractive.\n\nView on Upcoming Earnings\n\nNVIDIA’s third fiscal-quarter (ending October) results are approaching, and the company is expected to report on November 19.\n\nAlthough the full consensus has yet to converge, fundamentals heading into the report appear very optimistic.\nThe company is expected to guide revenue around USD 63 billion, with gross margin in the mid-70% range.\nConsidering current production conditions and the large volume of chips and boards already on its balance sheet (ready to ship), analysts believe quarterly revenue could reach USD 66 billion, well above the USD 61 billion market consensus.\n\nFrom a qualitative standpoint, while much of NVIDIA’s demand outlook has already been partially priced in, analysts still expect the tone of the earnings call to be highly confident and constructive.\n\nGTC Summary and Demand Clarification\n\nAt the recent GTC event, NVIDIA (NVDA) stated that, based on products shipped through 2025 and expected shipments through the end of 2026, the combined lifetime revenue of Blackwell and Rubin series could reach USD 500 billion.\nThis figure includes networking revenue and is roughly five times the lifetime revenue of the Hopper series.\nThe CEO’s reference to “five quarters” was clarified internally—not to mean that USD 500 billion would be achieved within the next five quarters.\n\nAccording to a bottom-up GPU model, analysts estimate Hopper’s lifetime revenue at around USD 130 billion (including last quarter’s stronger-than-expected tail contribution).\nThus, the “five-times revenue” remark broadly validates the model’s reasonableness.\nAs for forward potential, the model currently assumes Blackwell + Rubin (excluding Rubin Ultra and China sales) will generate USD 395 billion in total revenue from about 8 million GPU units, slightly below NVIDIA’s implied cumulative shipment figure of 10 million.\n\nNVIDIA is also accelerating construction of seven new U.S. supercomputers, including two systems at Argonne National Laboratory:\n    • Solstice (equipped with 100 K Blackwell GPUs)\n    • Equinox (equipped with 10 K Blackwell GPUs)\n\nThe company additionally announced these collaborations:\n    \n1. Nokia – launch of an AI-native 5G/6G base-station platform bringing AI to edge computing; T-Mobile US expects to begin joint pilots with Nokia/NVIDIA in 2026.\n    \n2. Uber – support for its Hyperion 10 autonomous-driving platform, targeting expansion to a 100 K-vehicle fleet by 2027.\n    \n3. NVQLink – an open system architecture connecting quantum processors (QPUs) to GPUs, already supported by 17 QPU vendors.\n    \n4. CrowdStrike & Palantir – partnerships expanding AI applications in cybersecurity and systems optimization.\n    \n5. Omniverse updates – introduction of DSX (for AI factories and data centers) and Mega (for manufacturing AI applications).\n\n⸻\n\nAnalysis and Key Takeaways\n\nFollowing last month’s UBS Asia Tech Tour and UBS Taiwan Conference, analysts raised TSMC’s CoWoS monthly capacity forecast from 100 K to 110 K wafers by end-2026, identifying NVIDIA as the primary beneficiary of the expansion.\n\nGPU shipment forecasts for 2025 and 2026 were lifted by 23%.\nFrom a product-mix standpoint, Blackwell series is expected to account for 73% of total revenue in 2025, peaking at 83% in 2026.\nThe Rubin series should represent about 13% of 2026 shipments, spanning R200, VR200, and CPX models.\n\nRubin CPX revenue contribution is projected to begin in Q1 2027.\nA Vera Rubin CPX cabinet contains 18 trays, each housing 2 VR200 and 8 CPX chips (a 1:2 ratio).\nBy end-2026, the CPX cabinet adoption rate is expected to reach roughly one-third.\n\n⸻\n\nValuation: Target Price USD 235 (previously USD 205)\n\nApplying a 28× P/E multiple (in line with large-cap tech peers) and raising average 2026–2027 EPS estimates from USD 7.01 to USD 8.57,\nthe report lifts its price target from USD 205 to USD 235.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020454468358698685, "ret_signed_m1d": 0.020454468358698685, "alpha_spy_m1d": 0.009334019127293702, "alpha_c_m1d": 0.007037992278692284, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.013416476080006401, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0019714724306379994, "ret_signed_p1d": -0.0019714724306379994, "alpha_spy_p1d": -0.005251753041353502, "alpha_c_p1d": -1.9400451918394346e-05, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": -0.001952071978719605, "ret_p1w": -0.07299517515386633, "ret_signed_p1w": -0.07299517515386633, "alpha_spy_p1w": -0.05899172675097386, "alpha_c_p1w": -0.03824410262522704, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.03475107252863929, "ret_p1m": -0.1276061122790818, "ret_signed_p1m": -0.1276061122790818, "alpha_spy_p1m": -0.13284273310428596, "alpha_c_p1m": -0.09612670026733339, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.03147941201174842, "ret_p3m": -0.07077481476500547, "ret_signed_p3m": -0.07077481476500547, "alpha_spy_p3m": -0.09683295582014162, "alpha_c_p3m": -0.19390785241759045, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": 0.12313303765258499, "ret_p6m": 0.026629582436106913, "ret_signed_p6m": 0.026629582436106913, "alpha_spy_p6m": -0.029524433465263344, "alpha_c_p6m": -0.3700542685910244, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.39668385102713133, "price_path": [-0.1129, -0.1214, -0.1157, -0.1091, -0.0996, -0.1027, -0.0973, -0.105, -0.1029, -0.1107, -0.103, -0.1344, -0.1355, -0.1376, -0.1228, -0.1138, -0.1041, -0.105, -0.112, -0.1416, -0.1416, -0.1583, -0.1591, -0.154, -0.1768, -0.1705, -0.1584, -0.126, -0.1268, -0.1236, -0.1239, -0.1381, -0.1607, -0.1314, -0.1292, -0.095, -0.1206, -0.1278, -0.1242, -0.1217, -0.1037, -0.0804, -0.0771, -0.069, -0.0753, -0.0855, -0.088, -0.0679, -0.0509, -0.0972, -0.0718, -0.1127, -0.1137, -0.1039, -0.0969, -0.0998, -0.1071, -0.1114, -0.1022, -0.082, -0.0562, -0.0092, 0.0205, 0.0, -0.002, 0.0197, -0.0207, -0.0379, -0.073, -0.0727, -0.0189, -0.048, -0.0448, -0.079, -0.0627, -0.0803, -0.1061, -0.0807, -0.1097, -0.1183, -0.1003, -0.1236, -0.1115, -0.1115, -0.1276, -0.1132, -0.1056, -0.1148, -0.0961, -0.1009, -0.0854, -0.0883, -0.0941, -0.1082, -0.1373, -0.1311, -0.124, -0.1574, -0.1417, -0.1079, -0.0946, -0.0674, -0.0703, -0.0703, -0.0609, -0.0723, -0.0756, -0.0807, -0.0807, -0.0691, -0.0727, -0.0771, -0.0679, -0.0879, -0.0888, -0.0884, -0.0841, -0.0973, -0.078, -0.0821, -0.0821, -0.1223, -0.0964, -0.0889, -0.075, -0.0809, -0.0708, -0.056, -0.0511, -0.0579, -0.0851, -0.1111, -0.1414, -0.1528, -0.0861, -0.0633, -0.0707, -0.0632, -0.0786, -0.0989, -0.0989, -0.0883, -0.0734, -0.0738, -0.0644, -0.0558, -0.0494, -0.0361, -0.0887, -0.1266, -0.1005, -0.1125, -0.0978, -0.0963, -0.1235, -0.0997, -0.0893, -0.083, -0.0972, -0.1115, -0.0968, -0.1032, -0.1108, -0.1198, -0.1487, -0.1342, -0.1364, -0.1192, -0.1559, -0.1742, -0.1858, -0.1403, -0.1337, -0.1256, -0.1256, -0.1244, -0.1221, -0.1025, -0.0934, -0.0702, -0.0668, -0.0313, -0.0197, -0.0223, -0.0059, -0.004, -0.0147, -0.0018, -0.0159, 0.0266]}
{"unit_id": "orig:1983794079109652612", "ticker_idx": 1, "ts": "2025-10-30T07:12:20+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM4 supply agreements with SK Hynix and Micron (MU) are still under negotiation", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses UBS Buy on NVDA, PT raised $205→$235 on EPS upgrades, CoWoS capacity, and Blackwell ramp; MU cited as HBM4 supply partner beneficiary.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS’s Dissection of NVIDIA’s GTC\n\nAside from several new partnerships (detailed below), NVIDIA’s (NVDA) latest demand guidance indicates that the market’s current earnings expectations remain significantly underestimated.\nThis view is also supported by recent findings from supply chain checks.\nAs a result, the report raises its 2026 EPS forecast to approximately USD 7.72 and 2027 EPS forecast to USD 9.42, projecting data-center revenue to reach about USD 377 billion in 2027 (versus roughly USD 190 billion this year).\n\nHowever, current production capacity and HBM support still appear insufficient to sustain the shipment volume and revenue scale implied by NVIDIA (details below).\nOne possible reason is that NVIDIA’s recognized GPU shipments for this year are expected to be roughly 15% lower than actual production, mainly to allow channel inventories to gradually clear before GB300 cabinet volumes ramp up.\nAfter adjusting for this factor, performance could align more closely with NVIDIA’s implied expectations and potentially lift 2026 EPS further to around USD 8.75.\nNVIDIA also suggested that if customers place orders sooner—allowing delivery cycles to accommodate additional shipments before the end of 2026—there remains further upside potential.\n\nThe upward forecast revision is based on several drivers:\nNVIDIA’s 2026 CoWoS capacity allocation has increased by about 10% compared with earlier plans, while HBM4 supply agreements with SK Hynix and Micron (MU) are still under negotiation.\nThe new model assumes GPU shipments of about 6 million units in 2025 and 8 million units in 2026, consistent with UBS Global Semiconductor Team’s capacity estimates, but not yet fully reflecting several billion USD of inventory on NVIDIA’s balance sheet, which could provide additional upside.\n\nIn addition, the model does not yet include the China market factor.\nIf a future Trump administration allows export to China of GPUs with roughly 80% of the B200’s performance—representing a USD 50 billion annual market of which Chinese domestic suppliers can meet only USD 10–20 billion—then the China market could become a major upside “wild card.”\n\nBeyond growth assumptions, the key investor concern now is whether NVIDIA’s high profit margins can remain sustainable amid the rapid development of custom ASICs.\nThe report concludes that such margins are likely to remain sustainable for the foreseeable future, owing to the company’s strong lead in cabinet-level ecosystem integration and explosive market demand growth.\nBased on higher earnings expectations, the price target is raised again—from USD 205 to USD 235—and the report notes that the stock’s setup ahead of earnings in three weeks looks structurally attractive.\n\nView on Upcoming Earnings\n\nNVIDIA’s third fiscal-quarter (ending October) results are approaching, and the company is expected to report on November 19.\n\nAlthough the full consensus has yet to converge, fundamentals heading into the report appear very optimistic.\nThe company is expected to guide revenue around USD 63 billion, with gross margin in the mid-70% range.\nConsidering current production conditions and the large volume of chips and boards already on its balance sheet (ready to ship), analysts believe quarterly revenue could reach USD 66 billion, well above the USD 61 billion market consensus.\n\nFrom a qualitative standpoint, while much of NVIDIA’s demand outlook has already been partially priced in, analysts still expect the tone of the earnings call to be highly confident and constructive.\n\nGTC Summary and Demand Clarification\n\nAt the recent GTC event, NVIDIA (NVDA) stated that, based on products shipped through 2025 and expected shipments through the end of 2026, the combined lifetime revenue of Blackwell and Rubin series could reach USD 500 billion.\nThis figure includes networking revenue and is roughly five times the lifetime revenue of the Hopper series.\nThe CEO’s reference to “five quarters” was clarified internally—not to mean that USD 500 billion would be achieved within the next five quarters.\n\nAccording to a bottom-up GPU model, analysts estimate Hopper’s lifetime revenue at around USD 130 billion (including last quarter’s stronger-than-expected tail contribution).\nThus, the “five-times revenue” remark broadly validates the model’s reasonableness.\nAs for forward potential, the model currently assumes Blackwell + Rubin (excluding Rubin Ultra and China sales) will generate USD 395 billion in total revenue from about 8 million GPU units, slightly below NVIDIA’s implied cumulative shipment figure of 10 million.\n\nNVIDIA is also accelerating construction of seven new U.S. supercomputers, including two systems at Argonne National Laboratory:\n    • Solstice (equipped with 100 K Blackwell GPUs)\n    • Equinox (equipped with 10 K Blackwell GPUs)\n\nThe company additionally announced these collaborations:\n    \n1. Nokia – launch of an AI-native 5G/6G base-station platform bringing AI to edge computing; T-Mobile US expects to begin joint pilots with Nokia/NVIDIA in 2026.\n    \n2. Uber – support for its Hyperion 10 autonomous-driving platform, targeting expansion to a 100 K-vehicle fleet by 2027.\n    \n3. NVQLink – an open system architecture connecting quantum processors (QPUs) to GPUs, already supported by 17 QPU vendors.\n    \n4. CrowdStrike & Palantir – partnerships expanding AI applications in cybersecurity and systems optimization.\n    \n5. Omniverse updates – introduction of DSX (for AI factories and data centers) and Mega (for manufacturing AI applications).\n\n⸻\n\nAnalysis and Key Takeaways\n\nFollowing last month’s UBS Asia Tech Tour and UBS Taiwan Conference, analysts raised TSMC’s CoWoS monthly capacity forecast from 100 K to 110 K wafers by end-2026, identifying NVIDIA as the primary beneficiary of the expansion.\n\nGPU shipment forecasts for 2025 and 2026 were lifted by 23%.\nFrom a product-mix standpoint, Blackwell series is expected to account for 73% of total revenue in 2025, peaking at 83% in 2026.\nThe Rubin series should represent about 13% of 2026 shipments, spanning R200, VR200, and CPX models.\n\nRubin CPX revenue contribution is projected to begin in Q1 2027.\nA Vera Rubin CPX cabinet contains 18 trays, each housing 2 VR200 and 8 CPX chips (a 1:2 ratio).\nBy end-2026, the CPX cabinet adoption rate is expected to reach roughly one-third.\n\n⸻\n\nValuation: Target Price USD 235 (previously USD 205)\n\nApplying a 28× P/E multiple (in line with large-cap tech peers) and raising average 2026–2027 EPS estimates from USD 7.01 to USD 8.57,\nthe report lifts its price target from USD 205 to USD 235.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011695912412809761, "ret_signed_m1d": 0.011695912412809761, "alpha_spy_m1d": 0.0005754631814047784, "alpha_c_m1d": -0.0017205636671966396, "bench_spy_m1d": 0.011120449231404983, "bench_c_m1d": 0.013416476080006401, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0010714021058136902, "ret_signed_p1d": -0.0010714021058136902, "alpha_spy_p1d": -0.004351682716529193, "alpha_c_p1d": 0.0008806698729059148, "bench_spy_p1d": 0.0032802806107155025, "bench_c_p1d": -0.001952071978719605, "ret_p1w": 0.0639257424807731, "ret_signed_p1w": 0.0639257424807731, "alpha_spy_p1w": 0.07792919088366557, "alpha_c_p1w": 0.09867681500941239, "bench_spy_p1w": -0.014003448402892471, "bench_c_p1w": -0.03475107252863929, "ret_p1m": 0.05566716562263596, "ret_signed_p1m": 0.05566716562263596, "alpha_spy_p1m": 0.05043054479743181, "alpha_c_p1m": 0.08714657763438438, "bench_spy_p1m": 0.005236620825204152, "bench_c_p1m": -0.03147941201174842, "ret_p3m": 0.8320865569645262, "ret_signed_p3m": 0.8320865569645262, "alpha_spy_p3m": 0.80602841590939, "alpha_c_p3m": 0.7089535193119412, "bench_spy_p3m": 0.026058141055136153, "bench_c_p3m": 0.12313303765258499, "ret_p6m": 1.2192286682006888, "ret_signed_p6m": 1.2192286682006888, "alpha_spy_p6m": 1.1630746522993185, "alpha_c_p6m": 0.8225448171735574, "bench_spy_p6m": 0.05615401590137026, "bench_c_p6m": 0.39668385102713133, "price_path": [-0.5192, -0.5135, -0.5147, -0.5009, -0.4696, -0.448, -0.4301, -0.4456, -0.441, -0.4608, -0.4488, -0.4555, -0.4771, -0.4834, -0.475, -0.4806, -0.4803, -0.4747, -0.4557, -0.4691, -0.4691, -0.4714, -0.4704, -0.4459, -0.4139, -0.4135, -0.3967, -0.3754, -0.3283, -0.2986, -0.2961, -0.2915, -0.2862, -0.2465, -0.274, -0.2656, -0.2576, -0.2786, -0.3003, -0.2984, -0.2688, -0.2535, -0.1874, -0.1802, -0.1615, -0.1475, -0.1711, -0.1226, -0.1414, -0.1893, -0.1395, -0.1649, -0.1432, -0.0959, -0.0966, -0.077, -0.097, -0.114, -0.0772, -0.0223, -0.0175, -0.0094, 0.0117, 0.0, -0.0011, 0.0477, -0.0267, 0.0602, 0.0639, 0.0621, 0.1308, 0.0763, 0.0933, 0.0578, 0.1019, 0.0801, 0.02, 0.0085, -0.1011, -0.0743, -0.0004, 0.0023, 0.0279, 0.0279, 0.0557, 0.0734, 0.0691, 0.0453, 0.0118, 0.059, 0.1023, 0.1268, 0.1772, 0.1538, 0.0765, 0.0602, 0.0379, 0.0067, 0.1095, 0.1871, 0.2347, 0.2333, 0.2798, 0.2798, 0.2713, 0.3146, 0.3069, 0.2746, 0.2746, 0.4086, 0.394, 0.5337, 0.5164, 0.4604, 0.5411, 0.5446, 0.5101, 0.4887, 0.5034, 0.62, 0.62, 0.63, 0.7377, 0.7755, 0.7848, 0.7376, 0.8321, 0.9439, 0.9462, 0.8528, 0.9552, 0.8732, 0.6944, 0.7099, 0.7626, 0.7127, 0.6669, 0.8325, 0.8487, 0.8384, 0.8384, 0.7854, 0.8799, 0.8638, 0.9122, 0.88, 0.8668, 0.9159, 0.8558, 0.8416, 0.8429, 0.6956, 0.7898, 0.7732, 0.6537, 0.7387, 0.8002, 0.8698, 0.8102, 0.903, 0.973, 1.0619, 1.062, 0.9841, 0.8886, 0.8058, 0.7664, 0.7064, 0.5874, 0.5953, 0.4377, 0.5094, 0.6435, 0.6363, 0.6363, 0.6877, 0.6869, 0.8172, 0.8832, 0.8791, 0.9058, 1.0805, 1.0383, 1.0428, 1.0331, 1.0034, 1.0077, 1.1779, 1.1522, 1.2192]}
{"unit_id": "thread:1996409704990568476", "ticker_idx": 0, "ts": "2025-12-04T02:42:19+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Savvy foreign investors have started buying Samsung Electronics again. foreigners have resumed net buying this month, recording a net purchase of 500 billion KRW as of the 1st to the 4th", "tweet_type": "news_reaction", "timeline": "months", "summary": "Bullish Samsung on HBM4 qual pass for Google TPU, Nvidia GPU qual imminent, and resumed foreign net buying.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung HBM4, Final Approval for Google TPU... \"Next Year's Supply Contract Completed\"\n\nSamsung Electronics has finally passed the qualification test (Qual Test) for HBM4 (High Bandwidth Memory), which will be mounted on Google's next-generation AI chip, the Tensor Processing Unit (TPU), and has completed the supply volume contract for next year. It is reported that Samsung has secured orders for more than three times the volume compared to this year, including the HBM3E previously supplied.\n\nAccording to related industries, Hock Tan, Chairman and CEO of Broadcom, visited South Korea earlier this week and met with Vice Chairman Jun Young-hyun (Head of DS Division), who oversees Samsung Electronics' semiconductor division, to sign the HBM4 supply contract. Broadcom serves as the design and production hub for Google's TPU and acts as a practical gateway responsible for the procurement, integration, and certification of core components like HBM.\n\nAn industry insider stated, \"CEO Hock Tan brought relevant executives and finished discussions with Samsung Electronics,\" adding, \"They requested an HBM supply contract through 2028, but the Samsung side gave a firm commitment only for next year's volume, with future volumes to be discussed later.\" Google's 7th generation TPU is scheduled to use the 5th generation HBM3E, while next year's 8th generation model will use the 6th generation HBM4.\n\nThe supply contract Samsung Electronics signed with Broadcom this time amounts to half of Samsung's annual HBM production capacity (CAPA). Broadcom requested a larger volume, but it is known that the Samsung side expressed reluctance due to supply commitments for Nvidia's Graphics Processing Units (GPUs).\n\nSince the release of Google's AI model 'Gemini 3', the AI ecosystem, previously centered on Nvidia GPUs, is rapidly diversifying toward Google TPUs. Wall Street estimates Google TPU volumes to grow from approximately 1.5-2 million units this year to 8-9 million units by 2028. This forecasts a growth of over five times in three years.\n\nThere are also observations that Samsung Electronics will obtain HBM4 qualification approval for Nvidia GPUs within this month. Attention is focused on whether Samsung Electronics, which has been making determined efforts after losing HBM leadership to SK Hynix due to delayed Nvidia qualification for HBM3E, can regain the market leader position through the passing of TPU and GPU HBM4 qualification tests.\n\nSavvy foreign investors have started buying Samsung Electronics again. Foreigners, who net bought over 7 trillion KRW in October pushing Samsung to '100,000 KRW Electronics', turned to net selling over 2 trillion KRW in November to realize profits. Samsung Electronics' stock price has also fallen about 8% from its peak after hitting 110,000 KRW. However, foreigners have resumed net buying this month, recording a net purchase of 500 billion KRW as of the 1st to the 4th.\n\n---\n\nhttps://t.co/oMLbnhwSHY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0057088037058496965, "ret_signed_m1d": -0.0057088037058496965, "alpha_spy_m1d": -0.004978281604529067, "alpha_c_m1d": -0.001998352232892353, "bench_spy_m1d": -0.0007305221013206298, "bench_c_m1d": -0.0037104514729573435, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.031398719781869255, "ret_signed_p1d": 0.031398719781869255, "alpha_spy_p1d": 0.029499182939159763, "alpha_c_p1d": 0.024080955022936612, "bench_spy_p1d": 0.0018995368427094927, "bench_c_p1d": 0.007317764758932643, "ret_p1w": 0.020932454904604825, "ret_signed_p1w": 0.020932454904604825, "alpha_spy_p1w": 0.01394814341607975, "alpha_c_p1w": 0.004201043880636135, "bench_spy_p1w": 0.0069843114885250746, "bench_c_p1w": 0.01673141102396869, "ret_p1m": 0.22858859914400131, "ret_signed_p1m": 0.22858859914400131, "alpha_spy_p1m": 0.22742155757464788, "alpha_c_p1m": 0.2355706654003511, "bench_spy_p1m": 0.001167041569353433, "bench_c_p1m": -0.006982066256349784, "ret_p3m": 0.8653512740617244, "ret_signed_p3m": 0.8653512740617244, "alpha_spy_p3m": 0.8683461007634059, "alpha_c_p3m": 0.9191283237860006, "bench_spy_p3m": -0.0029948267016814967, "bench_c_p3m": -0.053777049724276216, "ret_p6m": 2.03711051748468, "ret_signed_p6m": 2.03711051748468, "alpha_spy_p6m": 1.9254820765417193, "alpha_c_p6m": 1.7209487635845901, "bench_spy_p6m": 0.11162844094296087, "bench_c_p6m": 0.31616175390009005, "price_path": [-0.336, -0.3227, -0.3123, -0.3047, -0.2858, -0.2754, -0.2479, -0.2593, -0.2394, -0.2394, -0.209, -0.1977, -0.191, -0.1844, -0.2109, -0.1989, -0.2017, -0.1817, -0.1461, -0.1461, -0.1461, -0.1461, -0.1461, -0.1461, -0.1018, -0.1123, -0.1284, -0.0961, -0.0704, -0.0685, -0.0666, -0.0723, -0.0618, -0.0818, -0.0599, -0.0295, -0.0533, -0.0438, -0.0095, 0.0228, 0.0571, -0.0019, -0.0428, -0.0561, -0.0685, -0.0428, -0.0152, -0.019, -0.0219, -0.0752, -0.0428, -0.0695, -0.0818, -0.0428, -0.098, -0.0799, -0.0552, -0.0219, -0.0152, -0.0438, -0.0409, -0.0162, -0.0057, 0.0, 0.0314, 0.0419, 0.0314, 0.0276, 0.0209, 0.0362, -0.0029, -0.0219, 0.0266, 0.0238, 0.0114, 0.0514, 0.0609, 0.0571, 0.0571, 0.1132, 0.1425, 0.1464, 0.1464, 0.1464, 0.2286, 0.3204, 0.328, 0.3481, 0.3271, 0.329, 0.3271, 0.3156, 0.3414, 0.3758, 0.4236, 0.4275, 0.3883, 0.4294, 0.4561, 0.4542, 0.4542, 0.525, 0.5527, 0.5365, 0.5345, 0.438, 0.6015, 0.6168, 0.5231, 0.5164, 0.591, 0.5852, 0.6043, 0.7076, 0.7325, 0.7325, 0.7325, 0.7325, 0.8166, 0.8175, 0.8453, 0.9122, 0.9457, 1.0843, 1.07, 1.07, 0.8654, 0.6464, 0.8319, 0.7994, 0.6588, 0.7965, 0.8166, 0.7965, 0.7544, 0.8042, 0.8539, 0.9935, 0.917, 0.9065, 0.7812, 0.8137, 0.807, 0.7219, 0.7219, 0.6891, 0.6019, 0.817, 0.7092, 0.7839, 0.8501, 0.8826, 1.0168, 0.9545, 0.9736, 0.9257, 0.9784, 1.0215, 1.0838, 1.0695, 1.0551, 1.0982, 1.0838, 1.1509, 1.103, 1.1509, 1.1269, 1.1653, 1.1126, 1.1126, 1.2275, 1.2275, 1.5485, 1.6012, 1.5724, 1.7353, 1.673, 1.7209, 1.8359, 1.5916, 1.6922, 1.6395, 1.6443, 1.8694, 1.8024, 1.8024, 1.8647, 1.9413, 1.8694, 2.0371]}
{"unit_id": "orig:1985171470742602044", "ticker_idx": 0, "ts": "2025-11-03T02:25:35+00:00", "symbol": "Amazon", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "AWS currently has a $200 billion backlog... cloud revenues for CSPs will start to rebound significantly", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Citi's bullish AI adoption note: CSP cloud revenues to rebound late 2025, accelerate in 2026; $4.4T US hyperscaler CAPEX over 5 years.", "resolved_tickers": ["AMZN"], "resolution_method": "exchange_ticker", "resolution_reason": "Amazon.com Inc. AMZN US", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "251103_Updating AI Industry Model on Accelerating Adoption - CITI \n\n(1) Citi raised its estimates for AI service revenues for 2025–2030.\n\n(2) *AI service revenues are projected to grow from $43 billion in 2025 to $975 billion in 2030 [CAGR +86%].\n*AI services include: ChatGPT, Gemini, Claude, Microsoft AI Foundry, Grok, etc.\n\n(3) Citi also estimates that over the next five years, U.S. hyperscalers’ total CAPEX will reach $4.4 trillion, while global CAPEX will total approximately $7.75 trillion.\n\n(4) The firm believes that the market still underestimates the impact of AI on\n    \n1. corporate productivity,\n\n2. consumer behavior, and\n\n3. the scale of infrastructure investment required to realize it.\n\n(5) AI adoption within enterprises is transitioning from the Proof of Concept (PoC) stage to full-scale production, while at the same time, capacity constraints are being alleviated through expanded computing infrastructure.\n\n(6) Considering that AWS currently has a $200 billion backlog and Azure AI’s sales are being constrained by capacity shortages (i.e., demand exceeding supply), Citi expects that by late 2025,\n    \n1. cloud revenues for CSPs will start to rebound significantly, and\n    \n2. growth momentum will accelerate in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03850393220195625, "ret_signed_m1d": -0.03850393220195625, "alpha_spy_m1d": -0.036630760471192514, "alpha_c_m1d": -0.0292132783959953, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.009290653805960947, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018425168014886806, "ret_signed_p1d": -0.018425168014886806, "alpha_spy_p1d": -0.006571628072439273, "alpha_c_p1d": -0.0018259761948624664, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.01659919182002434, "ret_p1w": -0.022047268124077313, "ret_signed_p1w": -0.022047268124077313, "alpha_spy_p1w": -0.01926683918904004, "alpha_c_p1w": -0.008751455022655064, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.01329581310142225, "ret_p1m": -0.07708662138210509, "ret_signed_p1m": -0.07708662138210509, "alpha_spy_p1m": -0.0744378794828564, "alpha_c_p1m": -0.05338530132107411, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.023701320061030984, "ret_p3m": -0.048307103434885534, "ret_signed_p3m": -0.048307103434885534, "alpha_spy_p3m": -0.0669665994759655, "alpha_c_p3m": -0.05111637650305967, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.002809273068174134, "ret_p6m": 0.022440992941067828, "ret_signed_p6m": 0.022440992941067828, "alpha_spy_p6m": -0.02497658027166927, "alpha_c_p6m": 0.05227489095452398, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": -0.02983389801345615, "price_path": [-0.1248, -0.1215, -0.1233, -0.1287, -0.1281, -0.1159, -0.0906, -0.0904, -0.0886, -0.1023, -0.1189, -0.1262, -0.0991, -0.1026, -0.0996, -0.098, -0.0882, -0.0984, -0.0984, -0.1128, -0.1103, -0.0721, -0.0853, -0.0715, -0.062, -0.0932, -0.0947, -0.1018, -0.0889, -0.0785, -0.0881, -0.0896, -0.0887, -0.1038, -0.1311, -0.133, -0.1411, -0.1347, -0.1253, -0.1356, -0.1314, -0.1244, -0.1358, -0.1303, -0.1269, -0.1133, -0.1034, -0.1481, -0.1336, -0.1481, -0.1513, -0.1556, -0.1613, -0.1477, -0.1259, -0.1419, -0.1296, -0.1173, -0.1064, -0.0974, -0.0933, -0.1226, -0.0385, 0.0, -0.0184, -0.015, -0.0431, -0.0378, -0.022, -0.0193, -0.0386, -0.0646, -0.076, -0.0832, -0.1238, -0.1233, -0.1451, -0.1311, -0.1091, -0.0958, -0.0978, -0.0978, -0.0818, -0.0792, -0.0771, -0.0851, -0.098, -0.0963, -0.1067, -0.1027, -0.0875, -0.0934, -0.1095, -0.1239, -0.1238, -0.1289, -0.1072, -0.1049, -0.1007, -0.0861, -0.0851, -0.0851, -0.0846, -0.0863, -0.0845, -0.0913, -0.0913, -0.1083, -0.0824, -0.0515, -0.049, -0.0304, -0.0261, -0.0296, -0.0449, -0.0683, -0.0623, -0.0586, -0.0586, -0.0906, -0.0893, -0.0774, -0.0584, -0.0613, -0.0367, -0.0433, -0.0483, -0.0579, -0.0435, -0.0606, -0.0827, -0.1233, -0.172, -0.1783, -0.1852, -0.1965, -0.2142, -0.2174, -0.2174, -0.2081, -0.1937, -0.1935, -0.1728, -0.1919, -0.1789, -0.1707, -0.1814, -0.1732, -0.1796, -0.1782, -0.1464, -0.138, -0.1606, -0.1595, -0.1562, -0.1628, -0.1751, -0.1824, -0.1664, -0.1528, -0.1737, -0.1781, -0.1915, -0.1727, -0.1841, -0.1665, -0.1829, -0.2152, -0.2089, -0.18, -0.171, -0.1741, -0.1741, -0.1622, -0.1584, -0.1289, -0.0801, -0.0615, -0.0556, -0.0196, -0.0217, -0.0169, -0.0135, -0.0225, -0.0161, 0.0054, 0.0043, 0.0393, 0.028, 0.0224]}
{"unit_id": "orig:1985171470742602044", "ticker_idx": 1, "ts": "2025-11-03T02:25:35+00:00", "symbol": "Microsoft", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Azure AI's sales are being constrained by capacity shortages... growth momentum will accelerate in 2026", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Citi's bullish AI adoption note: CSP cloud revenues to rebound late 2025, accelerate in 2026; $4.4T US hyperscaler CAPEX over 5 years.", "resolved_tickers": ["MSFT"], "resolution_method": "exchange_ticker", "resolution_reason": "Microsoft Corporation MSFT US", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "251103_Updating AI Industry Model on Accelerating Adoption - CITI \n\n(1) Citi raised its estimates for AI service revenues for 2025–2030.\n\n(2) *AI service revenues are projected to grow from $43 billion in 2025 to $975 billion in 2030 [CAGR +86%].\n*AI services include: ChatGPT, Gemini, Claude, Microsoft AI Foundry, Grok, etc.\n\n(3) Citi also estimates that over the next five years, U.S. hyperscalers’ total CAPEX will reach $4.4 trillion, while global CAPEX will total approximately $7.75 trillion.\n\n(4) The firm believes that the market still underestimates the impact of AI on\n    \n1. corporate productivity,\n\n2. consumer behavior, and\n\n3. the scale of infrastructure investment required to realize it.\n\n(5) AI adoption within enterprises is transitioning from the Proof of Concept (PoC) stage to full-scale production, while at the same time, capacity constraints are being alleviated through expanded computing infrastructure.\n\n(6) Considering that AWS currently has a $200 billion backlog and Azure AI’s sales are being constrained by capacity shortages (i.e., demand exceeding supply), Citi expects that by late 2025,\n    \n1. cloud revenues for CSPs will start to rebound significantly, and\n    \n2. growth momentum will accelerate in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001508605430361909, "ret_signed_m1d": 0.001508605430361909, "alpha_spy_m1d": 0.003381777161125643, "alpha_c_m1d": 0.003381777161125643, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.0018731717307637341, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005222027183646616, "ret_signed_p1d": -0.005222027183646616, "alpha_spy_p1d": 0.006631512758800917, "alpha_c_p1d": 0.006631512758800917, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.011853539942447533, "ret_p1w": -0.02133336057748736, "ret_signed_p1w": -0.02133336057748736, "alpha_spy_p1w": -0.018552931642450088, "alpha_c_p1w": -0.018552931642450088, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.0027804289350372713, "ret_p1m": -0.050505553293129224, "ret_signed_p1m": -0.050505553293129224, "alpha_spy_p1m": -0.04785681139388054, "alpha_c_p1m": -0.04785681139388054, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.0026487418992486855, "ret_p3m": -0.15998808701846157, "ret_signed_p3m": -0.15998808701846157, "alpha_spy_p3m": -0.17864758305954154, "alpha_c_p3m": -0.17864758305954154, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.01865949604107997, "ret_p6m": -0.16632502576710906, "ret_signed_p6m": -0.16632502576710906, "alpha_spy_p6m": -0.21374259897984615, "alpha_c_p6m": -0.21374259897984615, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.047417573212737096, "price_path": [0.0136, 0.0057, 0.008, 0.0075, 0.0219, 0.0052, 0.0089, 0.0044, -0.0015, -0.0157, -0.0235, -0.0247, -0.019, -0.0247, -0.029, -0.0199, -0.0143, -0.02, -0.02, -0.023, -0.0226, -0.0175, -0.0426, -0.0364, -0.036, -0.0322, -0.031, -0.0138, -0.0032, -0.0155, -0.0136, -0.0166, 0.0017, -0.005, -0.0151, -0.0133, -0.0193, -0.0108, -0.0047, 0.0018, 0.0052, -0.0025, 0.0006, 0.0223, 0.0134, 0.0151, 0.0104, -0.0117, -0.0058, -0.0067, -0.007, -0.0105, -0.0067, -0.0005, 0.0012, 0.0068, 0.0068, 0.0127, 0.028, 0.0484, 0.0474, 0.0169, 0.0015, 0.0, -0.0052, -0.0191, -0.0385, -0.0391, -0.0213, -0.0161, -0.0114, -0.0266, -0.0132, -0.0185, -0.0449, -0.0578, -0.0729, -0.0852, -0.0815, -0.0757, -0.0592, -0.0592, -0.0466, -0.0568, -0.0505, -0.0743, -0.0683, -0.0638, -0.0485, -0.0466, -0.0727, -0.0632, -0.0727, -0.0799, -0.0769, -0.0774, -0.0622, -0.0584, -0.0603, -0.0566, -0.0543, -0.0543, -0.0549, -0.0561, -0.0554, -0.0629, -0.0629, -0.0836, -0.0837, -0.0728, -0.0632, -0.0735, -0.0713, -0.0753, -0.088, -0.1098, -0.1151, -0.1089, -0.1089, -0.1193, -0.1394, -0.1258, -0.0971, -0.0887, -0.0688, -0.0667, -0.16, -0.1662, -0.1796, -0.2032, -0.1974, -0.2372, -0.2227, -0.1985, -0.1992, -0.2164, -0.2213, -0.2223, -0.2223, -0.231, -0.2257, -0.2261, -0.2285, -0.2533, -0.2445, -0.222, -0.2198, -0.2372, -0.2259, -0.2155, -0.213, -0.2024, -0.2057, -0.2049, -0.2119, -0.2137, -0.2195, -0.2318, -0.2232, -0.2243, -0.2391, -0.2445, -0.2583, -0.2562, -0.2761, -0.2794, -0.2892, -0.3071, -0.3028, -0.2811, -0.2826, -0.2747, -0.2747, -0.2758, -0.277, -0.273, -0.2754, -0.2797, -0.2535, -0.2365, -0.2013, -0.1838, -0.1789, -0.188, -0.1762, -0.1592, -0.1925, -0.1753, -0.1749, -0.1663]}
{"unit_id": "orig:1985171470742602044", "ticker_idx": 2, "ts": "2025-11-03T02:25:35+00:00", "symbol": "cloud computing", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "cloud revenues for CSPs will start to rebound significantly, and growth momentum will accelerate in 2026", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Citi's bullish AI adoption note: CSP cloud revenues to rebound late 2025, accelerate in 2026; $4.4T US hyperscaler CAPEX over 5 years.", "resolved_tickers": ["CLOU"], "resolution_method": "etf", "resolution_reason": "US cloud computing sector proxied by CLOU ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "251103_Updating AI Industry Model on Accelerating Adoption - CITI \n\n(1) Citi raised its estimates for AI service revenues for 2025–2030.\n\n(2) *AI service revenues are projected to grow from $43 billion in 2025 to $975 billion in 2030 [CAGR +86%].\n*AI services include: ChatGPT, Gemini, Claude, Microsoft AI Foundry, Grok, etc.\n\n(3) Citi also estimates that over the next five years, U.S. hyperscalers’ total CAPEX will reach $4.4 trillion, while global CAPEX will total approximately $7.75 trillion.\n\n(4) The firm believes that the market still underestimates the impact of AI on\n    \n1. corporate productivity,\n\n2. consumer behavior, and\n\n3. the scale of infrastructure investment required to realize it.\n\n(5) AI adoption within enterprises is transitioning from the Proof of Concept (PoC) stage to full-scale production, while at the same time, capacity constraints are being alleviated through expanded computing infrastructure.\n\n(6) Considering that AWS currently has a $200 billion backlog and Azure AI’s sales are being constrained by capacity shortages (i.e., demand exceeding supply), Citi expects that by late 2025,\n    \n1. cloud revenues for CSPs will start to rebound significantly, and\n    \n2. growth momentum will accelerate in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0024834216192600733, "ret_signed_m1d": -0.0024834216192600733, "alpha_spy_m1d": -0.0006102498884963392, "alpha_c_m1d": -0.0006102498884963392, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.0018731717307637341, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03394038493390361, "ret_signed_p1d": -0.03394038493390361, "alpha_spy_p1d": -0.02208684499145608, "alpha_c_p1d": -0.02208684499145608, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.011853539942447533, "ret_p1w": -0.007450343804330073, "ret_signed_p1w": -0.007450343804330073, "alpha_spy_p1w": -0.004669914869292802, "alpha_c_p1w": -0.004669914869292802, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.0027804289350372713, "ret_p1m": -0.061258259585413755, "ret_signed_p1m": -0.061258259585413755, "alpha_spy_p1m": -0.05860951768616507, "alpha_c_p1m": -0.05860951768616507, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.0026487418992486855, "ret_p3m": -0.1324503635409675, "ret_signed_p3m": -0.1324503635409675, "alpha_spy_p3m": -0.15110985958204748, "alpha_c_p3m": -0.15110985958204748, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.01865949604107997, "ret_p6m": -0.1812913571369813, "ret_signed_p6m": -0.1812913571369813, "alpha_spy_p6m": -0.2287089303497184, "alpha_c_p6m": -0.2287089303497184, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.047417573212737096, "price_path": [-0.0567, -0.072, -0.0981, -0.1213, -0.1151, -0.0886, -0.1064, -0.0952, -0.0857, -0.0911, -0.0969, -0.1002, -0.0687, -0.0807, -0.0898, -0.0716, -0.0522, -0.0513, -0.0513, -0.0716, -0.065, -0.0687, -0.0546, -0.0468, -0.0418, -0.0505, -0.0315, -0.043, -0.0302, -0.0339, -0.0261, -0.0132, -0.0174, -0.0075, -0.0236, -0.0228, -0.0319, -0.0257, -0.0195, -0.0373, -0.0464, -0.0294, -0.031, -0.0182, -0.0393, -0.0224, -0.0215, -0.0567, -0.0385, -0.0443, -0.0393, -0.0476, -0.0476, -0.0257, -0.0145, -0.0269, -0.0157, -0.0079, -0.0021, 0.0, -0.0203, -0.0207, -0.0025, 0.0, -0.0339, -0.0236, -0.0327, -0.0219, -0.0075, 0.0012, -0.0116, -0.0377, -0.0381, -0.0629, -0.0654, -0.0795, -0.094, -0.0811, -0.0708, -0.0505, -0.0662, -0.0662, -0.0617, -0.0675, -0.0613, -0.0447, -0.048, -0.0459, -0.0418, -0.0368, -0.0368, -0.0397, -0.0488, -0.072, -0.0642, -0.0642, -0.0571, -0.0509, -0.043, -0.0534, -0.0531, -0.0531, -0.0513, -0.0538, -0.0565, -0.0637, -0.0637, -0.0857, -0.0741, -0.0592, -0.0517, -0.0629, -0.0654, -0.0579, -0.0828, -0.094, -0.1047, -0.1167, -0.1167, -0.1432, -0.1382, -0.1151, -0.1134, -0.0915, -0.0927, -0.0948, -0.1325, -0.1465, -0.1453, -0.1933, -0.1916, -0.2173, -0.2016, -0.1954, -0.1892, -0.2156, -0.2082, -0.1879, -0.1879, -0.2041, -0.1974, -0.2003, -0.2165, -0.2504, -0.2318, -0.2219, -0.2132, -0.2235, -0.2169, -0.2065, -0.1945, -0.1817, -0.1784, -0.1772, -0.1842, -0.1796, -0.19, -0.1867, -0.19, -0.178, -0.1734, -0.1751, -0.1863, -0.1689, -0.2012, -0.1908, -0.1937, -0.2219, -0.2198, -0.1929, -0.1829, -0.1767, -0.1767, -0.1821, -0.1896, -0.1925, -0.2256, -0.2715, -0.2397, -0.2475, -0.2103, -0.1883, -0.1846, -0.1701, -0.1651, -0.156, -0.1933, -0.1838, -0.1772, -0.1813]}
{"unit_id": "orig:1980783536102924372", "ticker_idx": 0, "ts": "2025-10-21T23:49:30+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We believe demand for the iPhone 17 series could surpass that of the iPhone 16. Given the continued strength in iPhone demand, we have raised our EPS forecasts by an average of about 3% for fiscal 2025 through fiscal 2027.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish AAPL call: iPhone 17 demand poised to surpass iPhone 16, EPS raised ~3% FY25-27; supply-chain rally including TSMC and Foxconn seen continuing through year-end.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: “Has the iPhone Entered a Super Cycle?”\n\nApple shares rose 4% on Monday to a record high of $262.24. With Counterpoint Research data indicating strong demand for the iPhone 17 series, speculation is spreading among investors that an iPhone super cycle has arrived.\n\nGoldman Sachs’ latest report highlights that delivery (shipping) lead times for the newest iPhone models have been somewhat extended, emphasizing this as another sign of strong demand. Buoyed by this optimistic report, the share prices of Apple’s major supply-chain partners in Asia also rose.\n\nAccording to the latest data from Counterpoint Research, sales of the iPhone 17 series in the first 10 days after launch in the U.S. and China increased by 14% compared to the same period for the iPhone 16 series. Counterpoint Senior Analyst Mengmeng Zhang wrote in the report:\n\n“The base model of the iPhone 17 is highly attractive to consumers and offers strong value for money.”\n\nIn the report, the team led by Goldman Sachs analyst Michael Ng pointed out that demand for the iPhone 17 series could surpass that of the iPhone 16, citing as supporting evidence longer delivery lead times, higher production capacity planning, and active feedback from carriers.\n\n“How does the demand trend for the iPhone 17 compare to previous generations? Are we entering an iPhone super cycle?” On this question, Michael Ng stated in the report:\n\n“We believe demand for the iPhone 17 series could surpass that of the iPhone 16.”\n\nThe report cited the following evidence:\n\n(1) Noticeably extended delivery lead times versus the 16 series; (2) An increase in total production planning; (3) Positive recent feedback from carrier channels.\n\nGlobally, during the 33 days after pre-orders began, the iPhone 17 series had an average delivery lead time of 24 days, exceeding the iPhone 16 series’ 20 days in the same period in 2024.\n\nBase model: Lead times are distinctly longer than for the base models of the iPhone 15 and 16\nPro and Pro Max: Somewhat longer than the equivalent iPhone 16 models\niPhone Air: Initially shorter than the iPhone 16 Plus, but has now flipped\n\nBy region, delivery lead times for the iPhone 17 have already surpassed those of the iPhone 16 series during the same period in the U.S., U.K., India, Japan, and Hong Kong. In mainland China, there are delays for the iPhone Air model, but lead times for the other iPhone 17 models still robustly exceed those of the prior year’s period. Goldman Sachs noted that there may be some distortion in the mainland China data. Consumers who would originally have chosen the Air model may have shifted to other iPhone 17 models, potentially leading to an overestimation of demand.\n\nBefore Apple reports earnings on October 30, Goldman Sachs analyst Michael Ng projected that iPhone sales would rise 10% year over year and Mac sales 12% year over year, leading Apple’s fiscal Q4 2025 revenue and EPS to beat expectations. The projected gross margin is 46.5%, reflecting roughly $1.1 billion in tariff costs. The report stated:\n\n“Given the continued strength in iPhone demand, we have raised our EPS forecasts by an average of about 3% for fiscal 2025 through fiscal 2027.”\n\nSince the iPhone 17 was announced on September 9, the strong momentum of the iPhone 17 series has triggered concurrent share-price gains among Apple’s supply-chain companies in Asia. Key suppliers that rely on Apple for more than half of their revenue—such as Luxshare Precision, LG Display, and LG Innotek—have all seen their share prices rise more than 20%. Although valuation doubts remain, the record high in Apple’s share price could keep the rally in supply-chain companies going through year-end.\n\nAccording to Bloomberg, Apple’s recovery is also having a positive impact on Taiwan’s stock market. Even amid the artificial intelligence boom, TSMC and Hon Hai (Foxconn) still rely significantly on Apple-related revenue. These two giants account for as much as 47% of the Taiwan Weighted Index, making the iPhone 17’s success effectively an invisible stimulus for the Taiwan market.\n\nThat said, as the Goldman Sachs analyst emphasized, whether this marks the start of an “iPhone super cycle” remains to be seen.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0020169677792949425, "ret_signed_m1d": -0.0020169677792949425, "alpha_spy_m1d": -0.0020319639241883625, "alpha_c_m1d": -0.0014968187710909797, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": -0.0005201490082039628, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01644018580608131, "ret_signed_p1d": -0.01644018580608131, "alpha_spy_p1d": -0.011241369843160598, "alpha_c_p1d": -0.0065560810758137356, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.009884104730267573, "ret_p1w": 0.023708979897406834, "ret_signed_p1w": 0.023708979897406834, "alpha_spy_p1w": 0.00021678740250186124, "alpha_c_p1w": -0.023526939145714243, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.04723591904312108, "ret_p1m": 0.02302520475984826, "ret_signed_p1m": 0.02302520475984826, "alpha_spy_p1m": 0.03592554696455952, "alpha_c_p1m": 0.048585206033422645, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.025560001273574384, "ret_p3m": -0.02660987058702402, "ret_signed_p3m": -0.02660987058702402, "alpha_spy_p3m": -0.05999906151208856, "alpha_c_p3m": -0.03819950369146763, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.011589633104443609, "ret_p6m": 0.015861114100419238, "ret_signed_p6m": 0.015861114100419238, "alpha_spy_p6m": -0.0327550160456358, "alpha_c_p6m": -0.02953767388128381, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.04539878798170305, "price_path": [-0.1874, -0.187, -0.1863, -0.1969, -0.2053, -0.211, -0.2307, -0.227, -0.2286, -0.1894, -0.1636, -0.1282, -0.1354, -0.126, -0.112, -0.1141, -0.1187, -0.1213, -0.1226, -0.1399, -0.1441, -0.1332, -0.1355, -0.1273, -0.1228, -0.115, -0.1166, -0.1166, -0.1258, -0.0925, -0.0875, -0.0878, -0.0947, -0.1082, -0.1369, -0.1246, -0.1092, -0.0992, -0.0937, -0.0905, -0.0947, -0.0657, -0.0255, -0.0317, -0.0398, -0.0225, -0.0278, -0.0317, -0.031, -0.0279, -0.0215, -0.0181, -0.0231, -0.0239, -0.0179, -0.0332, -0.0666, -0.0575, -0.0571, -0.0511, -0.0583, -0.0399, -0.002, 0.0, -0.0164, -0.0121, 0.0002, 0.023, 0.0237, 0.0264, 0.0328, 0.0289, 0.0239, 0.0277, 0.028, 0.0266, 0.0217, 0.0263, 0.0485, 0.0417, 0.0397, 0.0377, 0.0188, 0.0188, 0.023, 0.0142, 0.0342, 0.0511, 0.0551, 0.0573, 0.0573, 0.0622, 0.0784, 0.0902, 0.0824, 0.0693, 0.062, 0.0586, 0.0559, 0.062, 0.0591, 0.0601, 0.0442, 0.0461, 0.0355, 0.0369, 0.0425, 0.0322, 0.0375, 0.043, 0.043, 0.0415, 0.0428, 0.0402, 0.0356, 0.0356, 0.0324, 0.0181, -0.0006, -0.0083, -0.0132, -0.012, -0.0086, -0.0056, -0.0097, -0.0164, -0.0266, -0.0266, -0.0602, -0.0566, -0.054, -0.0551, -0.0271, -0.0162, -0.0231, -0.0161, -0.0116, 0.0285, 0.0265, 0.0532, 0.051, 0.0594, 0.0471, 0.0435, 0.0504, -0.0021, -0.0247, -0.0247, 0.0061, 0.0079, -0.0064, 0.0088, 0.0149, 0.0376, 0.0456, 0.0407, 0.0073, 0.0093, 0.0056, 0.001, -0.0075, -0.0183, -0.0091, -0.0055, -0.0056, -0.0248, -0.0463, -0.036, -0.0307, -0.047, -0.0507, -0.0544, -0.0411, -0.0405, -0.0368, -0.0358, -0.0514, -0.0596, -0.0323, -0.0253, -0.0242, -0.0242, -0.013, -0.0334, -0.0128, -0.0068, -0.0068, -0.0117, -0.0131, 0.0159]}
{"unit_id": "orig:1980783536102924372", "ticker_idx": 1, "ts": "2025-10-21T23:49:30+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC and Hon Hai (Foxconn) still rely significantly on Apple-related revenue. These two giants account for as much as 47% of the Taiwan Weighted Index, making the iPhone 17's success effectively an invisible stimulus for the Taiwan market.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish AAPL call: iPhone 17 demand poised to surpass iPhone 16, EPS raised ~3% FY25-27; supply-chain rally including TSMC and Foxconn seen continuing through year-end.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: “Has the iPhone Entered a Super Cycle?”\n\nApple shares rose 4% on Monday to a record high of $262.24. With Counterpoint Research data indicating strong demand for the iPhone 17 series, speculation is spreading among investors that an iPhone super cycle has arrived.\n\nGoldman Sachs’ latest report highlights that delivery (shipping) lead times for the newest iPhone models have been somewhat extended, emphasizing this as another sign of strong demand. Buoyed by this optimistic report, the share prices of Apple’s major supply-chain partners in Asia also rose.\n\nAccording to the latest data from Counterpoint Research, sales of the iPhone 17 series in the first 10 days after launch in the U.S. and China increased by 14% compared to the same period for the iPhone 16 series. Counterpoint Senior Analyst Mengmeng Zhang wrote in the report:\n\n“The base model of the iPhone 17 is highly attractive to consumers and offers strong value for money.”\n\nIn the report, the team led by Goldman Sachs analyst Michael Ng pointed out that demand for the iPhone 17 series could surpass that of the iPhone 16, citing as supporting evidence longer delivery lead times, higher production capacity planning, and active feedback from carriers.\n\n“How does the demand trend for the iPhone 17 compare to previous generations? Are we entering an iPhone super cycle?” On this question, Michael Ng stated in the report:\n\n“We believe demand for the iPhone 17 series could surpass that of the iPhone 16.”\n\nThe report cited the following evidence:\n\n(1) Noticeably extended delivery lead times versus the 16 series; (2) An increase in total production planning; (3) Positive recent feedback from carrier channels.\n\nGlobally, during the 33 days after pre-orders began, the iPhone 17 series had an average delivery lead time of 24 days, exceeding the iPhone 16 series’ 20 days in the same period in 2024.\n\nBase model: Lead times are distinctly longer than for the base models of the iPhone 15 and 16\nPro and Pro Max: Somewhat longer than the equivalent iPhone 16 models\niPhone Air: Initially shorter than the iPhone 16 Plus, but has now flipped\n\nBy region, delivery lead times for the iPhone 17 have already surpassed those of the iPhone 16 series during the same period in the U.S., U.K., India, Japan, and Hong Kong. In mainland China, there are delays for the iPhone Air model, but lead times for the other iPhone 17 models still robustly exceed those of the prior year’s period. Goldman Sachs noted that there may be some distortion in the mainland China data. Consumers who would originally have chosen the Air model may have shifted to other iPhone 17 models, potentially leading to an overestimation of demand.\n\nBefore Apple reports earnings on October 30, Goldman Sachs analyst Michael Ng projected that iPhone sales would rise 10% year over year and Mac sales 12% year over year, leading Apple’s fiscal Q4 2025 revenue and EPS to beat expectations. The projected gross margin is 46.5%, reflecting roughly $1.1 billion in tariff costs. The report stated:\n\n“Given the continued strength in iPhone demand, we have raised our EPS forecasts by an average of about 3% for fiscal 2025 through fiscal 2027.”\n\nSince the iPhone 17 was announced on September 9, the strong momentum of the iPhone 17 series has triggered concurrent share-price gains among Apple’s supply-chain companies in Asia. Key suppliers that rely on Apple for more than half of their revenue—such as Luxshare Precision, LG Display, and LG Innotek—have all seen their share prices rise more than 20%. Although valuation doubts remain, the record high in Apple’s share price could keep the rally in supply-chain companies going through year-end.\n\nAccording to Bloomberg, Apple’s recovery is also having a positive impact on Taiwan’s stock market. Even amid the artificial intelligence boom, TSMC and Hon Hai (Foxconn) still rely significantly on Apple-related revenue. These two giants account for as much as 47% of the Taiwan Weighted Index, making the iPhone 17’s success effectively an invisible stimulus for the Taiwan market.\n\nThat said, as the Goldman Sachs analyst emphasized, whether this marks the start of an “iPhone super cycle” remains to be seen.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -1.4996144893419938e-05, "alpha_c_m1d": -0.005560435515437723, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013513542684413182, "ret_signed_p1d": -0.013513542684413182, "alpha_spy_p1d": -0.008314726721492471, "alpha_c_p1d": 0.00588982400325877, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": -0.003378447939754503, "ret_signed_p1w": -0.003378447939754503, "alpha_spy_p1w": -0.026870640434659476, "alpha_c_p1w": -0.05495680663961977, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": -0.05743253565253892, "ret_signed_p1m": -0.05743253565253892, "alpha_spy_p1m": -0.04453219344782766, "alpha_c_p1m": -0.040461859502870956, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.17959449658037174, "ret_signed_p3m": 0.17959449658037174, "alpha_spy_p3m": 0.1462053056553072, "alpha_c_p3m": 0.016443942469497097, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.41469061945186314, "ret_signed_p6m": 0.41469061945186314, "alpha_spy_p6m": 0.3660744893058081, "alpha_c_p6m": 0.0987057503614659, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.2294, -0.2294, -0.2294, -0.2362, -0.2227, -0.2193, -0.2193, -0.2362, -0.2261, -0.2429, -0.2059, -0.2092, -0.2059, -0.2059, -0.1924, -0.2092, -0.2059, -0.2059, -0.2025, -0.2362, -0.2261, -0.2362, -0.2126, -0.2092, -0.1991, -0.2193, -0.2193, -0.216, -0.2193, -0.2193, -0.2193, -0.2059, -0.2059, -0.1924, -0.1756, -0.1655, -0.152, -0.1554, -0.1351, -0.1453, -0.1318, -0.1453, -0.125, -0.0946, -0.0946, -0.1081, -0.1216, -0.1216, -0.1182, -0.1047, -0.0777, -0.0541, -0.0541, -0.0304, -0.0439, -0.027, -0.027, -0.0439, -0.0372, -0.0101, 0.0034, -0.0203, 0.0, 0.0, -0.0135, -0.0203, -0.0203, 0.0, -0.0034, 0.0169, 0.0169, 0.0135, 0.0203, 0.0169, -0.0135, -0.0101, -0.0135, -0.0034, -0.0101, -0.0034, -0.0135, -0.0338, -0.0236, -0.0507, -0.0574, -0.0169, -0.0642, -0.0709, -0.0439, -0.027, -0.0304, -0.027, -0.0473, -0.0338, -0.0203, -0.0236, -0.0135, 0.0101, 0.0, 0.0169, -0.0034, 0.0033, -0.017, -0.0272, -0.0306, -0.0306, -0.0306, -0.0068, 0.0101, 0.0135, 0.0135, 0.0237, 0.0372, 0.0305, 0.0508, 0.0508, 0.0745, 0.1321, 0.1559, 0.1355, 0.1423, 0.1389, 0.1457, 0.1593, 0.1593, 0.1457, 0.1796, 0.1932, 0.2033, 0.1796, 0.1932, 0.1999, 0.1898, 0.2067, 0.2338, 0.2237, 0.2033, 0.1965, 0.2203, 0.2101, 0.1965, 0.2067, 0.2304, 0.2745, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2982, 0.2881, 0.3321, 0.366, 0.3525, 0.3525, 0.3389, 0.3118, 0.2643, 0.2881, 0.2813, 0.227, 0.2542, 0.3152, 0.2779, 0.2643, 0.2508, 0.2719, 0.2957, 0.2583, 0.2515, 0.2311, 0.2311, 0.2549, 0.2515, 0.2379, 0.2106, 0.197, 0.2617, 0.2311, 0.2311, 0.2311, 0.2651, 0.3263, 0.3297, 0.3603, 0.3535, 0.3977, 0.4147]}
{"unit_id": "orig:1980783536102924372", "ticker_idx": 2, "ts": "2025-10-21T23:49:30+00:00", "symbol": "Hon Hai Precision", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC and Hon Hai (Foxconn) still rely significantly on Apple-related revenue. These two giants account for as much as 47% of the Taiwan Weighted Index, making the iPhone 17's success effectively an invisible stimulus for the Taiwan market.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman Sachs bullish AAPL call: iPhone 17 demand poised to surpass iPhone 16, EPS raised ~3% FY25-27; supply-chain rally including TSMC and Foxconn seen continuing through year-end.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan (Foxconn)", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: “Has the iPhone Entered a Super Cycle?”\n\nApple shares rose 4% on Monday to a record high of $262.24. With Counterpoint Research data indicating strong demand for the iPhone 17 series, speculation is spreading among investors that an iPhone super cycle has arrived.\n\nGoldman Sachs’ latest report highlights that delivery (shipping) lead times for the newest iPhone models have been somewhat extended, emphasizing this as another sign of strong demand. Buoyed by this optimistic report, the share prices of Apple’s major supply-chain partners in Asia also rose.\n\nAccording to the latest data from Counterpoint Research, sales of the iPhone 17 series in the first 10 days after launch in the U.S. and China increased by 14% compared to the same period for the iPhone 16 series. Counterpoint Senior Analyst Mengmeng Zhang wrote in the report:\n\n“The base model of the iPhone 17 is highly attractive to consumers and offers strong value for money.”\n\nIn the report, the team led by Goldman Sachs analyst Michael Ng pointed out that demand for the iPhone 17 series could surpass that of the iPhone 16, citing as supporting evidence longer delivery lead times, higher production capacity planning, and active feedback from carriers.\n\n“How does the demand trend for the iPhone 17 compare to previous generations? Are we entering an iPhone super cycle?” On this question, Michael Ng stated in the report:\n\n“We believe demand for the iPhone 17 series could surpass that of the iPhone 16.”\n\nThe report cited the following evidence:\n\n(1) Noticeably extended delivery lead times versus the 16 series; (2) An increase in total production planning; (3) Positive recent feedback from carrier channels.\n\nGlobally, during the 33 days after pre-orders began, the iPhone 17 series had an average delivery lead time of 24 days, exceeding the iPhone 16 series’ 20 days in the same period in 2024.\n\nBase model: Lead times are distinctly longer than for the base models of the iPhone 15 and 16\nPro and Pro Max: Somewhat longer than the equivalent iPhone 16 models\niPhone Air: Initially shorter than the iPhone 16 Plus, but has now flipped\n\nBy region, delivery lead times for the iPhone 17 have already surpassed those of the iPhone 16 series during the same period in the U.S., U.K., India, Japan, and Hong Kong. In mainland China, there are delays for the iPhone Air model, but lead times for the other iPhone 17 models still robustly exceed those of the prior year’s period. Goldman Sachs noted that there may be some distortion in the mainland China data. Consumers who would originally have chosen the Air model may have shifted to other iPhone 17 models, potentially leading to an overestimation of demand.\n\nBefore Apple reports earnings on October 30, Goldman Sachs analyst Michael Ng projected that iPhone sales would rise 10% year over year and Mac sales 12% year over year, leading Apple’s fiscal Q4 2025 revenue and EPS to beat expectations. The projected gross margin is 46.5%, reflecting roughly $1.1 billion in tariff costs. The report stated:\n\n“Given the continued strength in iPhone demand, we have raised our EPS forecasts by an average of about 3% for fiscal 2025 through fiscal 2027.”\n\nSince the iPhone 17 was announced on September 9, the strong momentum of the iPhone 17 series has triggered concurrent share-price gains among Apple’s supply-chain companies in Asia. Key suppliers that rely on Apple for more than half of their revenue—such as Luxshare Precision, LG Display, and LG Innotek—have all seen their share prices rise more than 20%. Although valuation doubts remain, the record high in Apple’s share price could keep the rally in supply-chain companies going through year-end.\n\nAccording to Bloomberg, Apple’s recovery is also having a positive impact on Taiwan’s stock market. Even amid the artificial intelligence boom, TSMC and Hon Hai (Foxconn) still rely significantly on Apple-related revenue. These two giants account for as much as 47% of the Taiwan Weighted Index, making the iPhone 17’s success effectively an invisible stimulus for the Taiwan market.\n\nThat said, as the Goldman Sachs analyst emphasized, whether this marks the start of an “iPhone super cycle” remains to be seen.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0020920502092049986, "ret_signed_m1d": -0.0020920502092049986, "alpha_spy_m1d": -0.0021070463540984186, "alpha_c_m1d": -0.0015719012010010358, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": -0.0005201490082039628, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014644351464435212, "ret_signed_p1d": 0.014644351464435212, "alpha_spy_p1d": 0.019843167427355923, "alpha_c_p1d": 0.024528456194702786, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.009884104730267573, "ret_p1w": 0.046025104602510414, "ret_signed_p1w": 0.046025104602510414, "alpha_spy_p1w": 0.02253291210760544, "alpha_c_p1w": -0.0012108144406106636, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.04723591904312108, "ret_p1m": -0.04184100418410042, "ret_signed_p1m": -0.04184100418410042, "alpha_spy_p1m": -0.02894066197938916, "alpha_c_p1m": -0.016281002910526032, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.025560001273574384, "ret_p3m": -0.01882845188284521, "ret_signed_p3m": -0.01882845188284521, "alpha_spy_p3m": -0.05221764280790975, "alpha_c_p3m": -0.030418084987288818, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.011589633104443609, "ret_p6m": -0.13179916317991636, "ret_signed_p6m": -0.13179916317991636, "alpha_spy_p6m": -0.1804152933259714, "alpha_c_p6m": -0.1771979511616194, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.04539878798170305, "price_path": [-0.272, -0.2699, -0.2636, -0.2824, -0.2824, -0.2552, -0.2552, -0.2469, -0.228, -0.2218, -0.1862, -0.1862, -0.1715, -0.1757, -0.1695, -0.1653, -0.1339, -0.1213, -0.1297, -0.1611, -0.1381, -0.1527, -0.1318, -0.1255, -0.1297, -0.1381, -0.1485, -0.1695, -0.1695, -0.1611, -0.1506, -0.1423, -0.1485, -0.1318, -0.1213, -0.1025, -0.09, -0.0962, -0.0983, -0.113, -0.1004, -0.1046, -0.0962, -0.0753, -0.0669, -0.0377, -0.0816, -0.0816, -0.0962, -0.0837, -0.0565, -0.0523, -0.0523, -0.0439, -0.0586, -0.0732, -0.0732, -0.1088, -0.1381, -0.136, -0.0669, -0.0523, -0.0021, 0.0, 0.0146, 0.0, 0.0, 0.0314, 0.046, 0.0879, 0.0962, 0.0774, 0.0523, 0.0209, 0.0335, 0.0377, 0.0209, 0.0439, 0.0314, 0.0502, 0.0544, 0.0084, -0.0126, -0.0377, -0.0418, -0.0105, -0.0586, -0.0795, -0.0837, -0.0502, -0.0335, -0.0565, -0.0711, -0.0711, -0.0502, -0.0439, -0.0335, -0.0293, -0.0167, -0.023, -0.0544, -0.0502, -0.0732, -0.0879, -0.0941, -0.0962, -0.0732, -0.0628, -0.0628, -0.0628, -0.0628, -0.0565, -0.0335, -0.046, -0.0356, -0.0356, -0.0293, -0.0188, -0.0126, -0.0021, -0.0397, -0.0356, -0.0439, -0.0523, -0.0188, -0.023, -0.0188, -0.0397, -0.0649, -0.0837, -0.0649, -0.0732, -0.0628, -0.0565, -0.0565, -0.0628, -0.0774, -0.1025, -0.0941, -0.0837, -0.0983, -0.1004, -0.0858, -0.0753, -0.0502, -0.0502, -0.0502, -0.0502, -0.0502, -0.0502, -0.0502, -0.0502, -0.0439, -0.0293, 0.0314, 0.0167, 0.0167, 0.0, -0.0418, -0.0921, -0.0628, -0.0669, -0.1192, -0.1234, -0.0837, -0.1025, -0.1025, -0.0941, -0.113, -0.1213, -0.1423, -0.1506, -0.1799, -0.1841, -0.1632, -0.1611, -0.1653, -0.1883, -0.2155, -0.1757, -0.1925, -0.1925, -0.1925, -0.1967, -0.1569, -0.1632, -0.1611, -0.1632, -0.1318, -0.1318]}
{"unit_id": "reply:1994182609774006401", "ticker_idx": 0, "ts": "2025-11-27T23:12:38+00:00", "symbol": "GOOGL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "I'm going to buy Google the moment the market opens", "tweet_type": "prediction", "timeline": "intraday", "summary": "Buying Google at market open.", "resolved_tickers": ["GOOGL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@zephyr_z9 Then try the exact same thing on Gemini if you don’t believe me lol.\n\nI’m going to buy Google the moment the market opens.", "reply_text": null, "parent_text": "@jukan05 u probably edited that in lol", "quoted_text": null, "parent_author": "zephyr_z9", "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0007188225980021556, "ret_signed_p1d": 0.0007188225980021556, "alpha_spy_p1d": -0.004739583740612785, "alpha_c_p1d": -0.007406346790963303, "bench_spy_p1d": 0.005458406338614941, "bench_c_p1d": 0.008125169388965459, "ret_p1w": -0.007282496230383906, "ret_signed_p1w": -0.007282496230383906, "alpha_spy_p1w": -0.01421216163025596, "alpha_c_p1w": -0.01890231091550154, "bench_spy_p1w": 0.006929665399872054, "bench_c_p1w": 0.011619814685117635, "ret_p1m": -0.019487294003504685, "ret_signed_p1m": -0.019487294003504685, "alpha_spy_p1m": -0.03812805870780922, "alpha_c_p1m": -0.053489730521216305, "bench_spy_p1m": 0.018640764704304535, "bench_c_p1m": 0.03400243651771162, "ret_p3m": -0.027650123463464182, "ret_signed_p3m": -0.027650123463464182, "alpha_spy_p3m": -0.04192298905183367, "alpha_c_p3m": -0.04578802553584593, "bench_spy_p3m": 0.014272865588369488, "bench_c_p3m": 0.01813790207238175, "ret_p6m": 0.19859427488268988, "ret_signed_p6m": 0.19859427488268988, "alpha_spy_p6m": 0.09530206292854526, "alpha_c_p6m": 0.1833624836933312, "bench_spy_p6m": 0.10329221195414462, "bench_c_p6m": 0.01523179118935869, "price_path": [-0.3351, -0.34, -0.2797, -0.2746, -0.2662, -0.2685, -0.251, -0.2525, -0.2487, -0.2474, -0.2136, -0.215, -0.2201, -0.2123, -0.2039, -0.2107, -0.2134, -0.2276, -0.2318, -0.2294, -0.2372, -0.2402, -0.2346, -0.2321, -0.2332, -0.2173, -0.2319, -0.2354, -0.2451, -0.2606, -0.2369, -0.2328, -0.2154, -0.2141, -0.2083, -0.1982, -0.2172, -0.2133, -0.209, -0.1876, -0.1584, -0.164, -0.1418, -0.1202, -0.1211, -0.1132, -0.1326, -0.1114, -0.11, -0.1285, -0.0933, -0.0895, -0.1039, -0.1293, -0.1361, -0.1092, -0.1115, -0.0848, -0.0953, -0.0634, -0.0043, 0.0109, 0.0, 0.0, 0.0007, -0.0158, -0.0129, -0.001, -0.0073, 0.0041, -0.0188, -0.0083, 0.0015, -0.0229, -0.0327, -0.036, -0.0412, -0.072, -0.054, -0.0393, -0.0312, -0.0169, -0.0177, -0.0177, -0.0195, -0.0193, -0.0184, -0.0211, -0.0211, -0.0144, -0.01, -0.0169, 0.007, 0.0178, 0.0276, 0.0379, 0.0508, 0.0504, 0.0408, 0.0321, 0.0321, 0.0071, 0.027, 0.0338, 0.0256, 0.0423, 0.0463, 0.0509, 0.0579, 0.0571, 0.0749, 0.0625, 0.0416, 0.036, 0.0098, 0.0143, -0.0036, -0.0275, -0.0336, -0.0439, -0.0439, -0.0554, -0.0513, -0.0528, -0.0149, -0.0258, -0.0277, -0.0214, -0.0387, -0.025, -0.0413, -0.0505, -0.052, -0.059, -0.0664, -0.0412, -0.039, -0.0339, -0.05, -0.0539, -0.0437, -0.0269, -0.037, -0.0388, -0.058, -0.0546, -0.091, -0.0895, -0.1208, -0.1414, -0.144, -0.1, -0.0692, -0.0743, -0.0743, -0.0611, -0.044, -0.0069, -0.0032, -0.0071, 0.0056, 0.0419, 0.0551, 0.0517, 0.0694, 0.056, 0.04, 0.062, 0.0606, 0.0779, 0.0965, 0.0947, 0.0952, 0.2043, 0.2071, 0.1995, 0.2157, 0.2458, 0.2456, 0.2544, 0.2163, 0.2123, 0.2601, 0.2552, 0.2418, 0.2423, 0.2133, 0.2172, 0.2133, 0.1986]}
{"unit_id": "quote:1980511393200894180", "ticker_idx": 0, "ts": "2025-10-21T05:48:06+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The party's not even halfway over yet.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Endorses Morgan Stanley bullish memory call; author says the cycle upcycle is far from over, don't exit on peak fears.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"The party’s not even halfway over yet.\"", "reply_text": null, "parent_text": null, "quoted_text": "Surging demand, depleted inventories, memory is already a “seller’s market”… Morgan Stanley: investors shouldn’t exit on “peak fears”\n\nThe demand wave triggered by AI is reshaping the global memory market.\n\nOn October 19, Morgan Stanley noted in its latest report that the memory https://t.co/kAnPASNxR9", "parent_author": null, "ret_m1d": 0.013956793581783552, "ret_signed_m1d": 0.013956793581783552, "alpha_spy_m1d": 0.013941797436890132, "alpha_c_m1d": 0.00839635806634583, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0007941345459554124, "ret_signed_p1d": -0.0007941345459554124, "alpha_spy_p1d": 0.004404681416965299, "alpha_c_p1d": 0.01860923214171654, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.06839506472916757, "ret_signed_p1w": 0.06839506472916757, "alpha_spy_p1w": 0.04490287223426259, "alpha_c_p1w": 0.016816706029302297, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.0932779612689435, "ret_signed_p1m": 0.0932779612689435, "alpha_spy_p1m": 0.10617830347365476, "alpha_c_p1m": 0.11024863741861146, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.6359880167447961, "ret_signed_p3m": 0.6359880167447961, "alpha_spy_p3m": 0.6025988258197316, "alpha_c_p3m": 0.47283746263392146, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 1.27133163964156, "ret_signed_p6m": 1.27133163964156, "alpha_spy_p6m": 1.222715509495505, "alpha_c_p6m": 0.9553467705511627, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.4041, -0.4076, -0.3951, -0.3929, -0.3808, -0.3872, -0.4135, -0.406, -0.3994, -0.4071, -0.3938, -0.3816, -0.3691, -0.3607, -0.3575, -0.3578, -0.3651, -0.3724, -0.378, -0.3895, -0.3988, -0.3888, -0.3846, -0.3872, -0.3851, -0.3754, -0.3796, -0.3958, -0.3884, -0.3843, -0.3721, -0.3568, -0.3522, -0.3335, -0.3108, -0.2886, -0.2558, -0.2495, -0.226, -0.2383, -0.2012, -0.2113, -0.2004, -0.1865, -0.1942, -0.2006, -0.2233, -0.1994, -0.1958, -0.1555, -0.1153, -0.1084, -0.1033, -0.112, -0.0941, -0.101, -0.0802, -0.0746, -0.0922, -0.0649, -0.0174, -0.0079, 0.014, 0.0, -0.0008, 0.0035, 0.0536, 0.0837, 0.0684, 0.1053, 0.1203, 0.1253, 0.198, 0.1257, 0.1382, 0.1445, 0.1304, 0.183, 0.1819, 0.1854, 0.1678, 0.1287, 0.1643, 0.1075, 0.0933, 0.0731, 0.0284, 0.0615, 0.0706, 0.0955, 0.1121, 0.1023, 0.1155, 0.1367, 0.1275, 0.1102, 0.1403, 0.1831, 0.1807, 0.2126, 0.1862, 0.1673, 0.1354, 0.1037, 0.1242, 0.1618, 0.1825, 0.2375, 0.2431, 0.2617, 0.2617, 0.2864, 0.3415, 0.3476, 0.3357, 0.3357, 0.4329, 0.4737, 0.5489, 0.5609, 0.5424, 0.5645, 0.5686, 0.544, 0.5482, 0.5709, 0.636, 0.6429, 0.6179, 0.6704, 0.7044, 0.7155, 0.6765, 0.7813, 0.8611, 0.8701, 0.8683, 0.8164, 0.8985, 0.8331, 0.7648, 0.7798, 0.8215, 0.7949, 0.8518, 0.9144, 0.9139, 0.9139, 0.8943, 0.9292, 0.9633, 1.0198, 1.0193, 1.076, 1.1152, 1.2007, 1.1638, 1.1643, 0.9513, 0.8446, 0.9693, 0.9016, 0.8211, 0.9645, 1.0092, 0.9626, 0.9677, 1.0561, 1.104, 1.2142, 1.1279, 1.0847, 0.9575, 0.9916, 0.9733, 0.8556, 0.8585, 0.7467, 0.6958, 0.8853, 0.7972, 0.8562, 0.9059, 0.9382, 1.1161, 1.0937, 1.1193, 1.121, 1.2484, 1.2713]}
{"unit_id": "reply:1997651797926654146", "ticker_idx": 0, "ts": "2025-12-07T12:57:57+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "now that Rubin volumes for 2026 have been reduced, I rate Hynix more highly", "tweet_type": "prediction", "timeline": "quarters", "summary": "Rates SK Hynix more highly than Samsung Electronics following reduced Rubin volumes for 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@JEONwh7859 It’s difficult to predict which one will outperform.\nHowever, now that Rubin volumes for 2026 have been reduced, I rate Hynix more highly.", "reply_text": null, "parent_text": "@jukan05 1. Samsung Electronics vs. SK Hynix\n​\"Do you know why people are suggesting that Samsung Electronics' stock performance will likely outperform SK Hynix's moving forward? Do you consider this rationale to be reasonable?\"", "quoted_text": null, "parent_author": "JEONwh7859", "ret_m1d": -0.05719233447170469, "ret_signed_m1d": -0.05719233447170469, "alpha_spy_m1d": -0.06020566486713075, "alpha_c_m1d": -0.04598634336848628, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01906407531159593, "ret_signed_p1d": -0.01906407531159593, "alpha_spy_p1d": -0.018200930344634325, "alpha_c_p1d": -0.020285088195773793, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.03986126811277635, "ret_signed_p1w": -0.03986126811277635, "alpha_spy_p1w": -0.03561917054031205, "alpha_c_p1w": 0.002602403639090589, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.25823222667283874, "ret_signed_p1m": 0.25823222667283874, "alpha_spy_p1m": 0.24327646453594576, "alpha_c_p1m": 0.2032162932853212, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6338586339218024, "ret_signed_p3m": 0.6338586339218024, "alpha_spy_p3m": 0.6343074010647207, "alpha_c_p3m": 0.5578033195462861, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4735, -0.4683, -0.4311, -0.4268, -0.3973, -0.4224, -0.3843, -0.3843, -0.3921, -0.3748, -0.3809, -0.3826, -0.4172, -0.3956, -0.3982, -0.3765, -0.315, -0.315, -0.315, -0.315, -0.315, -0.315, -0.2588, -0.2813, -0.2873, -0.2683, -0.2163, -0.1938, -0.1592, -0.1704, -0.1661, -0.1713, -0.1168, -0.0735, -0.0977, -0.0336, -0.0163, -0.0319, 0.0738, 0.0149, 0.0027, 0.027, 0.0045, 0.0495, 0.072, 0.0686, 0.0599, -0.0302, 0.0495, -0.0128, -0.0267, -0.0111, -0.0977, -0.0994, -0.1012, -0.0925, -0.0572, -0.0815, -0.0676, -0.0329, -0.0433, -0.0607, -0.0572, 0.0, -0.0191, 0.0173, -0.0208, -0.0104, -0.0399, -0.0815, -0.0451, -0.0433, -0.052, 0.0052, 0.0121, 0.0191, 0.0191, 0.0381, 0.1092, 0.1282, 0.1282, 0.1282, 0.1733, 0.2062, 0.2582, 0.286, 0.3102, 0.2894, 0.2981, 0.279, 0.286, 0.2981, 0.3102, 0.3241, 0.2877, 0.2825, 0.3085, 0.3293, 0.2756, 0.3865, 0.4575, 0.4922, 0.5754, 0.4385, 0.5719, 0.5598, 0.4593, 0.4541, 0.5373, 0.5182, 0.4905, 0.539, 0.5251, 0.5251, 0.5251, 0.5251, 0.5494, 0.6447, 0.6482, 0.7418, 0.7643, 0.9082, 0.8422, 0.8422, 0.6304, 0.4741, 0.6339, 0.6043, 0.4515, 0.6286, 0.6582, 0.6148, 0.58, 0.6912, 0.6842, 0.8335, 0.7589, 0.7485, 0.62, 0.712, 0.7276, 0.62, 0.62, 0.5158, 0.4012, 0.5575, 0.4411, 0.521, 0.5384, 0.5905, 0.7936, 0.7328, 0.7832, 0.8058, 0.9151, 0.9724, 1.0054, 0.9585, 1.0245, 1.1252, 1.1235, 1.127, 1.1218, 1.2433, 1.2572, 1.245, 1.2329, 1.2329, 1.5124, 1.5124, 1.7798, 1.8718, 1.9274, 2.2642, 2.1861, 2.4309, 2.4205, 2.1583, 2.1948, 2.0298, 2.0298, 2.3684, 2.3702, 2.3702, 2.5629, 2.8945, 2.9751, 3.0515, null, null]}
{"unit_id": "orig:1989187356747030646", "ticker_idx": 0, "ts": "2025-11-14T04:23:17+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "our view is that the surprise could come from NAND", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish NAND into 2026 on supply discipline + QLC server SSD penetration acceleration; Kioxia and Samsung seen as key beneficiaries.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: FY2Q25 (September 2025 quarter) Earnings Release Comment\n[Daishin Securities / Semiconductor / Hyungkeun Ryu]\n\n■ Kioxia FY2Q25 (September 2025 quarter) Review\n    \n• Both the results and guidance fell short of the market’s elevated expectations. As a result, memory semiconductor share prices are showing weakness.\n    \n• One notable point in the results is the blended ASP. The blended ASP declined by a low single-digit percentage QoQ, but Kioxia’s business characteristics need to be taken into account.\n    \n• In the September quarter, it is the peak season for mobile products (centered on Apple), which have lower pricing compared to servers. On the server side as well, sales were focused on discrete components with relatively lower unit prices, so we believe there was clearly an impact from the change in product mix.\n\n■ In 2026, a surprise could come from NAND\n\nGiven Kioxia’s results, which came in below expectations, and its guidance for the next quarter, we believe some may be concerned about the NAND cycle. We see such concerns as excessive. On the contrary, our view is that the surprise could come from NAND.\n    \n1. Strengthening supply discipline\n\nEveryone is exercising restraint. Production lines for low value-added products with weak demand are being permanently shut down (withdrawn), and capex is being executed at a restrained level. Next year as well, it is highly likely that production growth will fall short of demand growth.\n    \n2. Upward revision of demand\n\n• The penetration of server QLC is set to begin in earnest. In Kioxia’s case, it is on the verge of obtaining certification for high-capacity eSSD products (122TB QLC) for North American hyperscalers, and Samsung Electronics is also planning to actively respond to high-capacity eSSD demand with its V9P process.\n    \n• Even if the products are not the 2Tb devices where customer demand is strongest, customers are buying any product as long as it is QLC. This shows that SSD penetration within servers is accelerating. Given the nature of SSDs, which have relatively higher pricing, we believe there is now an open opportunity for further profitability improvement driven by price hikes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07360560139148209, "ret_signed_m1d": 0.07360560139148209, "alpha_spy_m1d": 0.07344198852710782, "alpha_c_m1d": 0.07406943036308977, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05143599017271903, "ret_signed_p1d": 0.05143599017271903, "alpha_spy_p1d": 0.06075251410368119, "alpha_c_p1d": 0.0650021224052387, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.09314687611054955, "ret_signed_p1w": -0.09314687611054955, "alpha_spy_p1w": -0.07394844577089614, "alpha_c_p1w": -0.03850601301774119, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.05152930910482356, "ret_signed_p1m": -0.05152930910482356, "alpha_spy_p1m": -0.06462582918435167, "alpha_c_p1m": -0.074487112825394, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.8345152777411698, "ret_signed_p3m": 0.8345152777411698, "alpha_spy_p3m": 0.8016626184794394, "alpha_c_p3m": 0.6284436171855335, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 3.043934764754222, "ret_signed_p6m": 3.043934764754222, "alpha_spy_p6m": 2.9374065292481655, "alpha_c_p6m": 2.368176151041141, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.5796, -0.5876, -0.5925, -0.5845, -0.5822, -0.5826, -0.581, -0.5734, -0.5717, -0.579, -0.5771, -0.5731, -0.5594, -0.5385, -0.5327, -0.5223, -0.5016, -0.4718, -0.4445, -0.4378, -0.4186, -0.4286, -0.4038, -0.406, -0.3931, -0.3829, -0.3891, -0.3976, -0.4177, -0.3868, -0.3822, -0.3578, -0.3197, -0.2975, -0.2979, -0.3107, -0.2937, -0.2921, -0.2909, -0.2748, -0.2931, -0.2585, -0.2225, -0.2272, -0.1981, -0.2065, -0.2029, -0.1846, -0.1154, -0.0861, -0.1041, -0.0484, -0.0299, -0.025, 0.0191, -0.0304, -0.013, 0.0019, 0.03, 0.1053, 0.1063, 0.1178, 0.0736, 0.0, 0.0514, -0.0126, -0.008, -0.0456, -0.0931, -0.0552, -0.0583, -0.0782, -0.0563, -0.0488, -0.0629, -0.0483, -0.0652, -0.0576, -0.0224, 0.0115, -0.0005, 0.0179, 0.014, -0.0169, -0.0515, -0.0726, -0.05, -0.0175, 0.003, 0.0504, 0.0533, 0.0794, 0.0832, 0.1101, 0.1196, 0.115, 0.1069, 0.1069, 0.1881, 0.2294, 0.3316, 0.3633, 0.3449, 0.3832, 0.3945, 0.4021, 0.3956, 0.432, 0.4919, 0.5047, 0.5215, 0.6126, 0.6605, 0.6313, 0.6115, 0.6952, 0.7823, 0.8016, 0.8726, 0.8523, 0.9724, 0.8558, 0.7837, 0.7992, 0.8081, 0.7606, 0.8345, 0.9406, 0.9717, 0.9608, 0.9242, 0.9302, 0.9633, 1.0009, 1.0148, 1.0583, 1.0582, 1.1178, 1.0849, 1.0784, 0.8953, 0.8402, 0.905, 0.8336, 0.7966, 0.9286, 1.0165, 0.9609, 0.9915, 1.1019, 1.1203, 1.2432, 1.191, 1.1202, 1.0281, 1.0396, 1.038, 0.8915, 0.885, 0.7887, 0.7899, 0.9924, 0.9367, 0.9844, 1.0487, 1.0649, 1.301, 1.3446, 1.4072, 1.5081, 1.6419, 1.5622, 1.6342, 1.5543, 1.5527, 1.6206, 1.7486, 1.734, 1.764, 1.9029, 1.8477, 1.9137, 1.9489, 2.0175, 2.1819, 2.352, 2.5008, 2.5996, 2.8826, 3.0439]}
{"unit_id": "orig:1989187356747030646", "ticker_idx": 1, "ts": "2025-11-14T04:23:17+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe there is now an open opportunity for further profitability improvement driven by price hikes", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish NAND into 2026 on supply discipline + QLC server SSD penetration acceleration; Kioxia and Samsung seen as key beneficiaries.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: FY2Q25 (September 2025 quarter) Earnings Release Comment\n[Daishin Securities / Semiconductor / Hyungkeun Ryu]\n\n■ Kioxia FY2Q25 (September 2025 quarter) Review\n    \n• Both the results and guidance fell short of the market’s elevated expectations. As a result, memory semiconductor share prices are showing weakness.\n    \n• One notable point in the results is the blended ASP. The blended ASP declined by a low single-digit percentage QoQ, but Kioxia’s business characteristics need to be taken into account.\n    \n• In the September quarter, it is the peak season for mobile products (centered on Apple), which have lower pricing compared to servers. On the server side as well, sales were focused on discrete components with relatively lower unit prices, so we believe there was clearly an impact from the change in product mix.\n\n■ In 2026, a surprise could come from NAND\n\nGiven Kioxia’s results, which came in below expectations, and its guidance for the next quarter, we believe some may be concerned about the NAND cycle. We see such concerns as excessive. On the contrary, our view is that the surprise could come from NAND.\n    \n1. Strengthening supply discipline\n\nEveryone is exercising restraint. Production lines for low value-added products with weak demand are being permanently shut down (withdrawn), and capex is being executed at a restrained level. Next year as well, it is highly likely that production growth will fall short of demand growth.\n    \n2. Upward revision of demand\n\n• The penetration of server QLC is set to begin in earnest. In Kioxia’s case, it is on the verge of obtaining certification for high-capacity eSSD products (122TB QLC) for North American hyperscalers, and Samsung Electronics is also planning to actively respond to high-capacity eSSD demand with its V9P process.\n    \n• Even if the products are not the 2Tb devices where customer demand is strongest, customers are buying any product as long as it is QLC. This shows that SSD penetration within servers is accelerating. Given the nature of SSDs, which have relatively higher pricing, we believe there is now an open opportunity for further profitability improvement driven by price hikes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.29925187032418954, "ret_signed_m1d": 0.29925187032418954, "alpha_spy_m1d": 0.29908825745981527, "alpha_c_m1d": 0.2997156992957972, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11371571072319209, "ret_signed_p1d": 0.11371571072319209, "alpha_spy_p1d": 0.12303223465415425, "alpha_c_p1d": 0.12728184295571177, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": 0.0004987531172069293, "ret_signed_p1w": 0.0004987531172069293, "alpha_spy_p1w": 0.019697183456860334, "alpha_c_p1w": 0.055139616210015285, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.0822942643391521, "ret_signed_p1m": -0.0822942643391521, "alpha_spy_p1m": -0.09539078441868021, "alpha_c_p1m": -0.10525206805972254, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.8798004987531172, "ret_signed_p3m": 0.8798004987531172, "alpha_spy_p3m": 0.8469478394913867, "alpha_c_p3m": 0.6737288381974809, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 3.5825436408977556, "ret_signed_p6m": 3.5825436408977556, "alpha_spy_p6m": 3.476015405391699, "alpha_c_p6m": 2.9067850271846747, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.7533, -0.7648, -0.7701, -0.7603, -0.7614, -0.7573, -0.7591, -0.7535, -0.741, -0.7331, -0.7391, -0.7384, -0.7381, -0.6943, -0.6858, -0.6833, -0.6524, -0.6005, -0.5571, -0.5571, -0.5307, -0.5571, -0.5491, -0.5481, -0.5192, -0.5192, -0.5062, -0.5122, -0.5616, -0.5337, -0.5137, -0.5307, -0.4603, -0.3835, -0.3696, -0.4115, -0.4135, -0.3796, -0.3855, -0.3855, -0.4015, -0.3696, -0.3137, -0.3456, -0.2868, -0.2978, -0.2708, -0.2648, -0.1242, -0.0214, -0.0643, 0.0055, 0.0873, 0.0798, 0.0798, 0.0733, 0.0519, 0.1471, 0.201, 0.3287, 0.3162, 0.3207, 0.2993, 0.0, 0.1137, 0.0234, 0.0808, 0.1307, 0.0005, 0.0005, -0.0172, -0.1635, -0.0974, -0.0617, -0.114, -0.0806, -0.1011, -0.0953, -0.0594, 0.0125, -0.0151, -0.0554, -0.0414, -0.0135, -0.0823, -0.1334, -0.0723, -0.0514, -0.0683, 0.01, -0.0073, 0.0579, 0.0768, 0.1387, 0.0648, 0.0409, 0.0409, 0.0409, 0.0409, 0.1322, 0.1571, 0.2658, 0.2968, 0.2658, 0.2658, 0.3651, 0.3247, 0.3596, 0.4713, 0.5172, 0.5167, 0.6459, 0.7865, 0.7292, 0.7406, 0.8404, 0.8898, 0.9197, 1.1307, 0.8314, 1.0733, 1.0848, 0.9481, 0.9047, 0.8853, 0.8798, 0.8798, 1.1122, 1.2788, 1.2244, 1.2314, 1.1367, 1.1197, 1.0509, 1.0509, 1.2214, 1.1411, 1.1187, 1.1157, 1.1461, 1.015, 0.9197, 1.0249, 0.9925, 0.7985, 0.9521, 1.1327, 1.1192, 1.1012, 1.2559, 1.1556, 1.3332, 1.2304, 1.2304, 1.1411, 1.1047, 1.2389, 1.1112, 1.0219, 0.997, 0.9032, 1.1746, 1.1042, 1.1796, 1.2743, 1.3212, 1.7531, 1.7631, 2.0065, 2.1192, 2.4913, 2.2329, 2.3786, 2.0454, 2.0434, 2.2658, 2.4713, 2.5292, 2.4494, 2.5382, 2.6229, 2.6229, 2.7466, 2.6319, 2.6319, 2.6319, 2.6319, 3.3302, 3.4379, 3.5825]}
{"unit_id": "orig:1989187356747030646", "ticker_idx": 2, "ts": "2025-11-14T04:23:17+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics is also planning to actively respond to high-capacity eSSD demand with its V9P process", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish NAND into 2026 on supply discipline + QLC server SSD penetration acceleration; Kioxia and Samsung seen as key beneficiaries.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: FY2Q25 (September 2025 quarter) Earnings Release Comment\n[Daishin Securities / Semiconductor / Hyungkeun Ryu]\n\n■ Kioxia FY2Q25 (September 2025 quarter) Review\n    \n• Both the results and guidance fell short of the market’s elevated expectations. As a result, memory semiconductor share prices are showing weakness.\n    \n• One notable point in the results is the blended ASP. The blended ASP declined by a low single-digit percentage QoQ, but Kioxia’s business characteristics need to be taken into account.\n    \n• In the September quarter, it is the peak season for mobile products (centered on Apple), which have lower pricing compared to servers. On the server side as well, sales were focused on discrete components with relatively lower unit prices, so we believe there was clearly an impact from the change in product mix.\n\n■ In 2026, a surprise could come from NAND\n\nGiven Kioxia’s results, which came in below expectations, and its guidance for the next quarter, we believe some may be concerned about the NAND cycle. We see such concerns as excessive. On the contrary, our view is that the surprise could come from NAND.\n    \n1. Strengthening supply discipline\n\nEveryone is exercising restraint. Production lines for low value-added products with weak demand are being permanently shut down (withdrawn), and capex is being executed at a restrained level. Next year as well, it is highly likely that production growth will fall short of demand growth.\n    \n2. Upward revision of demand\n\n• The penetration of server QLC is set to begin in earnest. In Kioxia’s case, it is on the verge of obtaining certification for high-capacity eSSD products (122TB QLC) for North American hyperscalers, and Samsung Electronics is also planning to actively respond to high-capacity eSSD demand with its V9P process.\n    \n• Even if the products are not the 2Tb devices where customer demand is strongest, customers are buying any product as long as it is QLC. This shows that SSD penetration within servers is accelerating. Given the nature of SSDs, which have relatively higher pricing, we believe there is now an open opportunity for further profitability improvement driven by price hikes.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05761317038957614, "ret_signed_m1d": 0.05761317038957614, "alpha_spy_m1d": 0.057449557525201866, "alpha_c_m1d": 0.06302709370298476, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03497941620794953, "ret_signed_p1d": 0.03497941620794953, "alpha_spy_p1d": 0.044295940138911694, "alpha_c_p1d": 0.05063102623894755, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.024691404986047938, "ret_signed_p1w": -0.024691404986047938, "alpha_spy_p1w": -0.005492974646394533, "alpha_c_p1w": 0.027191265659056163, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.07818927376678131, "ret_signed_p1m": 0.07818927376678131, "alpha_spy_p1m": 0.0650927536872532, "alpha_c_p1m": 0.09050924685115647, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": 0.7347294453035076, "ret_signed_p3m": 0.7347294453035076, "alpha_spy_p3m": 0.7018767860417772, "alpha_c_p3m": 0.7408938922948921, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 1.957630231150651, "ret_signed_p6m": 1.957630231150651, "alpha_spy_p6m": 1.8511019956445949, "alpha_c_p6m": 1.7195854032114168, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.283, -0.2779, -0.2769, -0.2687, -0.2677, -0.28, -0.2769, -0.2871, -0.2861, -0.3076, -0.2923, -0.2851, -0.282, -0.2882, -0.282, -0.2677, -0.2564, -0.2482, -0.2277, -0.2165, -0.1868, -0.199, -0.1775, -0.1775, -0.1448, -0.1325, -0.1253, -0.1181, -0.1468, -0.1337, -0.1368, -0.1152, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0288, -0.0401, -0.0576, -0.0226, 0.0051, 0.0072, 0.0093, 0.0031, 0.0144, -0.0072, 0.0165, 0.0494, 0.0237, 0.034, 0.071, 0.106, 0.143, 0.0792, 0.035, 0.0206, 0.0072, 0.035, 0.0648, 0.0607, 0.0576, 0.0, 0.035, 0.0062, -0.0072, 0.035, -0.0247, -0.0051, 0.0216, 0.0576, 0.0648, 0.034, 0.037, 0.0638, 0.0751, 0.0813, 0.1152, 0.1265, 0.1152, 0.1111, 0.1039, 0.1204, 0.0782, 0.0576, 0.1101, 0.107, 0.0936, 0.1368, 0.1471, 0.143, 0.143, 0.2037, 0.2354, 0.2395, 0.2395, 0.2395, 0.3284, 0.4277, 0.436, 0.4577, 0.4349, 0.437, 0.4349, 0.4225, 0.4504, 0.4876, 0.5393, 0.5435, 0.5011, 0.5455, 0.5745, 0.5724, 0.5724, 0.6489, 0.6789, 0.6613, 0.6593, 0.5548, 0.7316, 0.7482, 0.6469, 0.6396, 0.7203, 0.7141, 0.7347, 0.8464, 0.8733, 0.8733, 0.8733, 0.8733, 0.9642, 0.9653, 0.9952, 1.0676, 1.1038, 1.2537, 1.2382, 1.2382, 1.017, 0.7802, 0.9808, 0.9456, 0.7937, 0.9425, 0.9642, 0.9425, 0.897, 0.9508, 1.0046, 1.1555, 1.0728, 1.0614, 0.926, 0.9611, 0.9539, 0.8619, 0.8619, 0.8264, 0.7321, 0.9647, 0.8481, 0.9289, 1.0004, 1.0356, 1.1807, 1.1133, 1.1341, 1.0823, 1.1392, 1.1858, 1.2532, 1.2376, 1.2221, 1.2687, 1.2532, 1.3257, 1.2739, 1.3257, 1.2998, 1.3412, 1.2843, 1.2843, 1.4086, 1.4086, 1.7556, 1.8126, 1.7815, 1.9576]}
{"unit_id": "orig:1992825312649052667", "ticker_idx": 0, "ts": "2025-11-24T05:19:14+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Memory conditions continue to improve, as we reports of intensifying shortages across the board... we are seeing a generational tightness across all areas... current spending levels do not seem likely to catch up to demand in the next few quarters", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rebuttal: memory shortages intensifying across DRAM/NAND, supply can't catch demand for several quarters; bullish MU, Samsung, Hynix and NVDA on margin resilience.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s rebuttal to the recent concerns about the memory semiconductor sector:\n\nMemory conditions continue to improve, as we reports of intensifying shortages across the board. It seems most pronounced in DDR5 DRAM, where there is an active sense of crisis that product is not available at any price, whereas in other areas, such as consumer grade NAND, there is more of a sense that there will be a shortage, but less of a near term crisis. Nevertheless, we are seeing a generational tightness across all areas.\n\nStocks have sold off lately due to concerns about higher capital spending, or the view that shortages may be easing, but neither impacts our view. In terms of spending we saw multiple articles last week on increased investment activity from both Samsung and Hynix with Micron also speaking to \"upside pressure\" to their capital investment guidance from last earnings at a competitor conference. We recognize that's likely to invite additional skepticism on the longevity of the cycle, but context here is key. All three suppliers have spoken already to sold out market conditions in the face of accelerating cloud demand, and with the HBM trade ratio any attempts to catch-up to that demand needs to be met with higher capital spending. Samsung/Hynix will both see greenfield in 2026, MU closer to 2027 something that hasn't changed over the last few weeks.\n\nWe have argued that DRAM spending is high relative to longer term run rates, but the \"correct\" level of capital spending is driven by a somewhat steady state view of\ndemand, whereas currently we are clearly seeing strong demand on the back of the AI build; current spending levels do not seem likely to catch up to demand in the\nnext few quarters. China's memory efforts have not particularly inflected, though their spending is part of what we would point to as elevated vs. the longer term\ntrend line. That's more of a concern for investors in NAND than DRAM, as YMTC has had an outsized impact on that market in the past. But we would emphasize that all suppliers outside of China have taken a much more measured approach to serving domestic Chinese demand with MU specifically exiting large segments of that market in its entirety. YMTC is significant in scale, but the supply-demand imbalance in NAND is far more significant than the few percentage points impact YMTC is likely to have even in an aggressive capital expansion scenario. We also heard concerns that a pc OEM said that \"memory is not a problem\". But we are visiting with several DRAM customers in Asia this week, and while we will form conclusions at the end of that trip, it is very clear that the industry collectively does not have enough inventory, even if one employee from one OEM does feel comfortable. Generally speaking, the focus should not be on companies securing supply, it's what they have to do to get it. And we continue to hear about agreements that embed multiple quarters of significant pice increases, with large penalties if customers don't take the supply.\n\nWe also heard some concerns about NVIDIA's statement that they are aspiring to continue gross margins in the mid 70s even as input costs increase. While it is\ntrue that HBM price increases are sticky, and thus will likely lag DDR5 pricing in the short run, we think that NVIDIA has a very good opportunity to pass through\nmemory increases. We believe that HBM costs for NVIDIA are about one third of cost of sales, or about 8% of revenues (given 75% gross margins), while those costs for competitors tend to run above 50% of cost of sales for companies with 50% gross margin - that is, HBM is closer to 25% of revenues. The most likely path for NVIDIA to maintain 75% gross margins is to raise prices slightly, given that peers would have to raise them by a higher percentage.\n\nThe risk in our view is demand destruction in consumer markets, but that likely seems to be a case for moderation of rate of change - which we already expect over the course of 2026 - rather than a reversal. In the most marginal areas of the market we're seeing this begin to play out (China’s SMIC Expects Memory Shortage to Hit Cars, Phones in 2026; PC hardware vendor PNY suspends Black Friday storage deals — cites rising NAND costs). As DRAM/NAND increasingly gets diverted to markets where buyers have the highest willingness to pay, and a situation where bit shipment growth is limited by production capacity - a good\noutcome for memory earnings.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07395165783628477, "ret_signed_m1d": -0.07395165783628477, "alpha_spy_m1d": -0.05944664293831714, "alpha_c_m1d": -0.03564671789055174, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002679462252589637, "ret_signed_p1d": 0.002679462252589637, "alpha_spy_p1d": -0.006726437399517282, "alpha_c_p1d": 5.4937233483931536e-05, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.07381771882218224, "ret_signed_p1w": 0.07381771882218224, "alpha_spy_p1w": 0.05656102483362524, "alpha_c_p1w": 0.03300614715289796, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.23373367277280366, "ret_signed_p1m": 0.23373367277280366, "alpha_spy_p1m": 0.2019377262353088, "alpha_c_p1m": 0.159512079956748, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.8645050019129275, "ret_signed_p3m": 0.8645050019129275, "alpha_spy_p3m": 0.83792841545446, "alpha_c_p3m": 0.6511398018833909, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.1229220528131627, "ret_signed_p6m": 2.1229220528131627, "alpha_spy_p6m": 2.019475422839834, "alpha_c_p6m": 1.5138968426047226, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.4745, -0.4555, -0.4689, -0.4689, -0.4712, -0.4702, -0.4457, -0.4137, -0.4133, -0.3964, -0.3752, -0.328, -0.2983, -0.2959, -0.2912, -0.286, -0.2463, -0.2738, -0.2653, -0.2573, -0.2783, -0.3001, -0.2981, -0.2685, -0.2533, -0.1871, -0.1799, -0.1612, -0.1472, -0.1708, -0.1223, -0.1411, -0.189, -0.1392, -0.1646, -0.1429, -0.0956, -0.0962, -0.0766, -0.0966, -0.1137, -0.0769, -0.0219, -0.0171, -0.009, 0.0121, 0.0004, -0.0007, 0.0481, -0.0263, 0.0606, 0.0643, 0.0625, 0.1312, 0.0767, 0.0936, 0.0581, 0.1023, 0.0805, 0.0204, 0.0089, -0.1007, -0.074, 0.0, 0.0027, 0.0283, 0.0283, 0.056, 0.0738, 0.0695, 0.0457, 0.0121, 0.0593, 0.1027, 0.1272, 0.1776, 0.1542, 0.0769, 0.0606, 0.0383, 0.0071, 0.1099, 0.1875, 0.2352, 0.2337, 0.2802, 0.2802, 0.2718, 0.3151, 0.3073, 0.2751, 0.2751, 0.4091, 0.3945, 0.5343, 0.5169, 0.461, 0.5417, 0.5452, 0.5106, 0.4892, 0.5039, 0.6206, 0.6206, 0.6306, 0.7383, 0.7762, 0.7854, 0.7383, 0.8327, 0.9446, 0.9469, 0.8535, 0.9559, 0.8738, 0.695, 0.7106, 0.7633, 0.7133, 0.6675, 0.8332, 0.8494, 0.8391, 0.8391, 0.786, 0.8806, 0.8645, 0.9128, 0.8807, 0.8675, 0.9166, 0.8565, 0.8423, 0.8436, 0.6962, 0.7904, 0.7738, 0.6543, 0.7393, 0.8009, 0.8705, 0.8109, 0.9037, 0.9737, 1.0626, 1.0628, 0.9848, 0.8893, 0.8064, 0.767, 0.707, 0.588, 0.5959, 0.4382, 0.5099, 0.6441, 0.6369, 0.6369, 0.6883, 0.6875, 0.8178, 0.8839, 0.8798, 0.9065, 1.0812, 1.0391, 1.0435, 1.0339, 1.0042, 1.0084, 1.1787, 1.153, 1.22, 1.3444, 1.2539, 1.3172, 1.3114, 1.4233, 1.5764, 1.8613, 1.9792, 1.89, 2.3378, 2.5546, 2.4261, 2.5917, 2.4683, 2.2388, 2.046, 2.1229]}
{"unit_id": "orig:1992825312649052667", "ticker_idx": 1, "ts": "2025-11-24T05:19:14+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung/Hynix will both see greenfield in 2026... all three suppliers have spoken already to sold out market conditions in the face of accelerating cloud demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rebuttal: memory shortages intensifying across DRAM/NAND, supply can't catch demand for several quarters; bullish MU, Samsung, Hynix and NVDA on margin resilience.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s rebuttal to the recent concerns about the memory semiconductor sector:\n\nMemory conditions continue to improve, as we reports of intensifying shortages across the board. It seems most pronounced in DDR5 DRAM, where there is an active sense of crisis that product is not available at any price, whereas in other areas, such as consumer grade NAND, there is more of a sense that there will be a shortage, but less of a near term crisis. Nevertheless, we are seeing a generational tightness across all areas.\n\nStocks have sold off lately due to concerns about higher capital spending, or the view that shortages may be easing, but neither impacts our view. In terms of spending we saw multiple articles last week on increased investment activity from both Samsung and Hynix with Micron also speaking to \"upside pressure\" to their capital investment guidance from last earnings at a competitor conference. We recognize that's likely to invite additional skepticism on the longevity of the cycle, but context here is key. All three suppliers have spoken already to sold out market conditions in the face of accelerating cloud demand, and with the HBM trade ratio any attempts to catch-up to that demand needs to be met with higher capital spending. Samsung/Hynix will both see greenfield in 2026, MU closer to 2027 something that hasn't changed over the last few weeks.\n\nWe have argued that DRAM spending is high relative to longer term run rates, but the \"correct\" level of capital spending is driven by a somewhat steady state view of\ndemand, whereas currently we are clearly seeing strong demand on the back of the AI build; current spending levels do not seem likely to catch up to demand in the\nnext few quarters. China's memory efforts have not particularly inflected, though their spending is part of what we would point to as elevated vs. the longer term\ntrend line. That's more of a concern for investors in NAND than DRAM, as YMTC has had an outsized impact on that market in the past. But we would emphasize that all suppliers outside of China have taken a much more measured approach to serving domestic Chinese demand with MU specifically exiting large segments of that market in its entirety. YMTC is significant in scale, but the supply-demand imbalance in NAND is far more significant than the few percentage points impact YMTC is likely to have even in an aggressive capital expansion scenario. We also heard concerns that a pc OEM said that \"memory is not a problem\". But we are visiting with several DRAM customers in Asia this week, and while we will form conclusions at the end of that trip, it is very clear that the industry collectively does not have enough inventory, even if one employee from one OEM does feel comfortable. Generally speaking, the focus should not be on companies securing supply, it's what they have to do to get it. And we continue to hear about agreements that embed multiple quarters of significant pice increases, with large penalties if customers don't take the supply.\n\nWe also heard some concerns about NVIDIA's statement that they are aspiring to continue gross margins in the mid 70s even as input costs increase. While it is\ntrue that HBM price increases are sticky, and thus will likely lag DDR5 pricing in the short run, we think that NVIDIA has a very good opportunity to pass through\nmemory increases. We believe that HBM costs for NVIDIA are about one third of cost of sales, or about 8% of revenues (given 75% gross margins), while those costs for competitors tend to run above 50% of cost of sales for companies with 50% gross margin - that is, HBM is closer to 25% of revenues. The most likely path for NVIDIA to maintain 75% gross margins is to raise prices slightly, given that peers would have to raise them by a higher percentage.\n\nThe risk in our view is demand destruction in consumer markets, but that likely seems to be a case for moderation of rate of change - which we already expect over the course of 2026 - rather than a reversal. In the most marginal areas of the market we're seeing this begin to play out (China’s SMIC Expects Memory Shortage to Hit Cars, Phones in 2026; PC hardware vendor PNY suspends Black Friday storage deals — cites rising NAND costs). As DRAM/NAND increasingly gets diverted to markets where buyers have the highest willingness to pay, and a situation where bit shipment growth is limited by production capacity - a good\noutcome for memory earnings.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01964843134458627, "ret_signed_m1d": -0.01964843134458627, "alpha_spy_m1d": -0.005143416446618643, "alpha_c_m1d": 0.0036255541335565855, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.023273985478142856, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026887284286339463, "ret_signed_p1d": 0.026887284286339463, "alpha_spy_p1d": 0.017481384634232544, "alpha_c_p1d": 0.024205859810904284, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.002681424475435179, "ret_p1w": 0.04239921661394108, "ret_signed_p1w": 0.04239921661394108, "alpha_spy_p1w": 0.02514252262538408, "alpha_c_p1w": 0.018660256880602244, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.023738959733338838, "ret_p1m": 0.1530506263321878, "ret_signed_p1m": 0.1530506263321878, "alpha_spy_p1m": 0.12125467979469295, "alpha_c_p1m": 0.10788685638219397, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.045163769949993826, "ret_p3m": 0.9743910791481385, "ret_signed_p3m": 0.9743910791481385, "alpha_spy_p3m": 0.947814492689671, "alpha_c_p3m": 0.9703320180088377, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.004059061139300768, "ret_p6m": 1.8687926787213338, "ret_signed_p6m": 1.8687926787213338, "alpha_spy_p6m": 1.7653460487480053, "alpha_c_p6m": 1.6266596875507027, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.24213299117063114, "price_path": [-0.2731, -0.2834, -0.2824, -0.304, -0.2886, -0.2814, -0.2783, -0.2845, -0.2783, -0.2639, -0.2526, -0.2443, -0.2237, -0.2124, -0.1826, -0.1949, -0.1733, -0.1733, -0.1403, -0.128, -0.1208, -0.1136, -0.1424, -0.1293, -0.1324, -0.1107, -0.0719, -0.0719, -0.0719, -0.0719, -0.0719, -0.0719, -0.0238, -0.0352, -0.0527, -0.0176, 0.0103, 0.0124, 0.0145, 0.0083, 0.0196, -0.0021, 0.0217, 0.0548, 0.029, 0.0393, 0.0765, 0.1117, 0.1489, 0.0848, 0.0403, 0.0259, 0.0124, 0.0403, 0.0703, 0.0662, 0.0631, 0.0052, 0.0403, 0.0114, -0.0021, 0.0403, -0.0196, 0.0, 0.0269, 0.0631, 0.0703, 0.0393, 0.0424, 0.0693, 0.0807, 0.0869, 0.121, 0.1324, 0.121, 0.1169, 0.1096, 0.1262, 0.0838, 0.0631, 0.1158, 0.1127, 0.0993, 0.1427, 0.1531, 0.1489, 0.1489, 0.2099, 0.2418, 0.2459, 0.2459, 0.2459, 0.3353, 0.4351, 0.4434, 0.4652, 0.4423, 0.4444, 0.4423, 0.4299, 0.4579, 0.4953, 0.5473, 0.5515, 0.5089, 0.5535, 0.5826, 0.5806, 0.5806, 0.6574, 0.6876, 0.6699, 0.6678, 0.5629, 0.7406, 0.7572, 0.6554, 0.6481, 0.7292, 0.7229, 0.7437, 0.8559, 0.8829, 0.8829, 0.8829, 0.8829, 0.9744, 0.9754, 1.0056, 1.0783, 1.1147, 1.2654, 1.2498, 1.2498, 1.0274, 0.7894, 0.991, 0.9557, 0.8029, 0.9526, 0.9744, 0.9526, 0.9068, 0.9609, 1.0149, 1.1666, 1.0835, 1.0721, 0.9359, 0.9713, 0.964, 0.8715, 0.8715, 0.8358, 0.7411, 0.9748, 0.8577, 0.9389, 1.0108, 1.0462, 1.1919, 1.1243, 1.1451, 1.093, 1.1503, 1.1972, 1.2648, 1.2492, 1.2336, 1.2805, 1.2648, 1.3377, 1.2857, 1.3377, 1.3117, 1.3533, 1.2961, 1.2961, 1.421, 1.421, 1.7699, 1.8271, 1.7959, 1.9729, 1.9052, 1.9573, 2.0823, 1.8167, 1.9261, 1.8688]}
{"unit_id": "orig:1992825312649052667", "ticker_idx": 2, "ts": "2025-11-24T05:19:14+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung/Hynix will both see greenfield in 2026... generational tightness across all areas", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rebuttal: memory shortages intensifying across DRAM/NAND, supply can't catch demand for several quarters; bullish MU, Samsung, Hynix and NVDA on margin resilience.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s rebuttal to the recent concerns about the memory semiconductor sector:\n\nMemory conditions continue to improve, as we reports of intensifying shortages across the board. It seems most pronounced in DDR5 DRAM, where there is an active sense of crisis that product is not available at any price, whereas in other areas, such as consumer grade NAND, there is more of a sense that there will be a shortage, but less of a near term crisis. Nevertheless, we are seeing a generational tightness across all areas.\n\nStocks have sold off lately due to concerns about higher capital spending, or the view that shortages may be easing, but neither impacts our view. In terms of spending we saw multiple articles last week on increased investment activity from both Samsung and Hynix with Micron also speaking to \"upside pressure\" to their capital investment guidance from last earnings at a competitor conference. We recognize that's likely to invite additional skepticism on the longevity of the cycle, but context here is key. All three suppliers have spoken already to sold out market conditions in the face of accelerating cloud demand, and with the HBM trade ratio any attempts to catch-up to that demand needs to be met with higher capital spending. Samsung/Hynix will both see greenfield in 2026, MU closer to 2027 something that hasn't changed over the last few weeks.\n\nWe have argued that DRAM spending is high relative to longer term run rates, but the \"correct\" level of capital spending is driven by a somewhat steady state view of\ndemand, whereas currently we are clearly seeing strong demand on the back of the AI build; current spending levels do not seem likely to catch up to demand in the\nnext few quarters. China's memory efforts have not particularly inflected, though their spending is part of what we would point to as elevated vs. the longer term\ntrend line. That's more of a concern for investors in NAND than DRAM, as YMTC has had an outsized impact on that market in the past. But we would emphasize that all suppliers outside of China have taken a much more measured approach to serving domestic Chinese demand with MU specifically exiting large segments of that market in its entirety. YMTC is significant in scale, but the supply-demand imbalance in NAND is far more significant than the few percentage points impact YMTC is likely to have even in an aggressive capital expansion scenario. We also heard concerns that a pc OEM said that \"memory is not a problem\". But we are visiting with several DRAM customers in Asia this week, and while we will form conclusions at the end of that trip, it is very clear that the industry collectively does not have enough inventory, even if one employee from one OEM does feel comfortable. Generally speaking, the focus should not be on companies securing supply, it's what they have to do to get it. And we continue to hear about agreements that embed multiple quarters of significant pice increases, with large penalties if customers don't take the supply.\n\nWe also heard some concerns about NVIDIA's statement that they are aspiring to continue gross margins in the mid 70s even as input costs increase. While it is\ntrue that HBM price increases are sticky, and thus will likely lag DDR5 pricing in the short run, we think that NVIDIA has a very good opportunity to pass through\nmemory increases. We believe that HBM costs for NVIDIA are about one third of cost of sales, or about 8% of revenues (given 75% gross margins), while those costs for competitors tend to run above 50% of cost of sales for companies with 50% gross margin - that is, HBM is closer to 25% of revenues. The most likely path for NVIDIA to maintain 75% gross margins is to raise prices slightly, given that peers would have to raise them by a higher percentage.\n\nThe risk in our view is demand destruction in consumer markets, but that likely seems to be a case for moderation of rate of change - which we already expect over the course of 2026 - rather than a reversal. In the most marginal areas of the market we're seeing this begin to play out (China’s SMIC Expects Memory Shortage to Hit Cars, Phones in 2026; PC hardware vendor PNY suspends Black Friday storage deals — cites rising NAND costs). As DRAM/NAND increasingly gets diverted to markets where buyers have the highest willingness to pay, and a situation where bit shipment growth is limited by production capacity - a good\noutcome for memory earnings.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001923085035032246, "ret_signed_m1d": 0.001923085035032246, "alpha_spy_m1d": 0.016428099932999873, "alpha_c_m1d": 0.040228024980765276, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001923085035032246, "ret_signed_p1d": -0.001923085035032246, "alpha_spy_p1d": -0.011328984687139165, "alpha_c_p1d": -0.004547610054137952, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.03535624949908667, "ret_signed_p1w": 0.03535624949908667, "alpha_spy_p1w": 0.018099555510529663, "alpha_c_p1w": -0.0054553221701976184, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.12388131384195789, "ret_signed_p1m": 0.12388131384195789, "alpha_spy_p1m": 0.09208536730446304, "alpha_c_p1m": 0.049659721025902215, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.720462027311146, "ret_signed_p3m": 0.720462027311146, "alpha_spy_p3m": 0.6938854408526784, "alpha_c_p3m": 0.5070968272816094, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.3643692048480567, "ret_signed_p6m": 2.3643692048480567, "alpha_spy_p6m": 2.260922574874728, "alpha_c_p6m": 1.7553439946396165, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.5007, -0.4837, -0.4827, -0.5077, -0.499, -0.4952, -0.4894, -0.474, -0.4673, -0.4462, -0.4154, -0.4096, -0.3683, -0.3635, -0.3308, -0.3587, -0.3163, -0.3163, -0.325, -0.3058, -0.3125, -0.3144, -0.3529, -0.3288, -0.3317, -0.3077, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.1769, -0.2019, -0.2087, -0.1875, -0.1298, -0.1048, -0.0663, -0.0788, -0.074, -0.0798, -0.0192, 0.0288, 0.0019, 0.0731, 0.0923, 0.075, 0.1923, 0.1269, 0.1135, 0.1404, 0.1154, 0.1654, 0.1904, 0.1865, 0.1769, 0.0769, 0.1654, 0.0962, 0.0808, 0.0981, 0.0019, 0.0, -0.0019, 0.0077, 0.0469, 0.02, 0.0354, 0.0738, 0.0623, 0.0431, 0.0469, 0.1104, 0.0892, 0.1297, 0.0873, 0.0989, 0.0661, 0.02, 0.0604, 0.0623, 0.0527, 0.1162, 0.1239, 0.1316, 0.1316, 0.1527, 0.2317, 0.2528, 0.2528, 0.2528, 0.3029, 0.3394, 0.3972, 0.4279, 0.4549, 0.4318, 0.4414, 0.4202, 0.4279, 0.4414, 0.4549, 0.4703, 0.4299, 0.4241, 0.453, 0.4761, 0.4164, 0.5396, 0.6185, 0.657, 0.7493, 0.5973, 0.7455, 0.732, 0.6204, 0.6146, 0.707, 0.6858, 0.655, 0.7089, 0.6935, 0.6935, 0.6935, 0.6935, 0.7205, 0.8263, 0.8302, 0.9341, 0.9591, 1.1189, 1.0456, 1.0456, 0.8104, 0.6369, 0.8143, 0.7815, 0.6118, 0.8085, 0.8412, 0.793, 0.7545, 0.8779, 0.8702, 1.036, 0.9531, 0.9415, 0.7988, 0.901, 0.9184, 0.7988, 0.7988, 0.6831, 0.5559, 0.7294, 0.6002, 0.6889, 0.7082, 0.7661, 0.9916, 0.9241, 0.9801, 1.0051, 1.1266, 1.1902, 1.2268, 1.1748, 1.2481, 1.3599, 1.358, 1.3618, 1.356, 1.491, 1.5064, 1.4929, 1.4794, 1.4794, 1.7898, 1.7898, 2.0867, 2.1889, 2.2506, 2.6246, 2.5379, 2.8097, 2.7982, 2.507, 2.5475, 2.3644]}
{"unit_id": "orig:1992825312649052667", "ticker_idx": 3, "ts": "2025-11-24T05:19:14+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA has a very good opportunity to pass through memory increases... The most likely path for NVIDIA to maintain 75% gross margins is to raise prices slightly", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley rebuttal: memory shortages intensifying across DRAM/NAND, supply can't catch demand for several quarters; bullish MU, Samsung, Hynix and NVDA on margin resilience.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s rebuttal to the recent concerns about the memory semiconductor sector:\n\nMemory conditions continue to improve, as we reports of intensifying shortages across the board. It seems most pronounced in DDR5 DRAM, where there is an active sense of crisis that product is not available at any price, whereas in other areas, such as consumer grade NAND, there is more of a sense that there will be a shortage, but less of a near term crisis. Nevertheless, we are seeing a generational tightness across all areas.\n\nStocks have sold off lately due to concerns about higher capital spending, or the view that shortages may be easing, but neither impacts our view. In terms of spending we saw multiple articles last week on increased investment activity from both Samsung and Hynix with Micron also speaking to \"upside pressure\" to their capital investment guidance from last earnings at a competitor conference. We recognize that's likely to invite additional skepticism on the longevity of the cycle, but context here is key. All three suppliers have spoken already to sold out market conditions in the face of accelerating cloud demand, and with the HBM trade ratio any attempts to catch-up to that demand needs to be met with higher capital spending. Samsung/Hynix will both see greenfield in 2026, MU closer to 2027 something that hasn't changed over the last few weeks.\n\nWe have argued that DRAM spending is high relative to longer term run rates, but the \"correct\" level of capital spending is driven by a somewhat steady state view of\ndemand, whereas currently we are clearly seeing strong demand on the back of the AI build; current spending levels do not seem likely to catch up to demand in the\nnext few quarters. China's memory efforts have not particularly inflected, though their spending is part of what we would point to as elevated vs. the longer term\ntrend line. That's more of a concern for investors in NAND than DRAM, as YMTC has had an outsized impact on that market in the past. But we would emphasize that all suppliers outside of China have taken a much more measured approach to serving domestic Chinese demand with MU specifically exiting large segments of that market in its entirety. YMTC is significant in scale, but the supply-demand imbalance in NAND is far more significant than the few percentage points impact YMTC is likely to have even in an aggressive capital expansion scenario. We also heard concerns that a pc OEM said that \"memory is not a problem\". But we are visiting with several DRAM customers in Asia this week, and while we will form conclusions at the end of that trip, it is very clear that the industry collectively does not have enough inventory, even if one employee from one OEM does feel comfortable. Generally speaking, the focus should not be on companies securing supply, it's what they have to do to get it. And we continue to hear about agreements that embed multiple quarters of significant pice increases, with large penalties if customers don't take the supply.\n\nWe also heard some concerns about NVIDIA's statement that they are aspiring to continue gross margins in the mid 70s even as input costs increase. While it is\ntrue that HBM price increases are sticky, and thus will likely lag DDR5 pricing in the short run, we think that NVIDIA has a very good opportunity to pass through\nmemory increases. We believe that HBM costs for NVIDIA are about one third of cost of sales, or about 8% of revenues (given 75% gross margins), while those costs for competitors tend to run above 50% of cost of sales for companies with 50% gross margin - that is, HBM is closer to 25% of revenues. The most likely path for NVIDIA to maintain 75% gross margins is to raise prices slightly, given that peers would have to raise them by a higher percentage.\n\nThe risk in our view is demand destruction in consumer markets, but that likely seems to be a case for moderation of rate of change - which we already expect over the course of 2026 - rather than a reversal. In the most marginal areas of the market we're seeing this begin to play out (China’s SMIC Expects Memory Shortage to Hit Cars, Phones in 2026; PC hardware vendor PNY suspends Black Friday storage deals — cites rising NAND costs). As DRAM/NAND increasingly gets diverted to markets where buyers have the highest willingness to pay, and a situation where bit shipment growth is limited by production capacity - a good\noutcome for memory earnings.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020104103865114742, "ret_signed_m1d": -0.020104103865114742, "alpha_spy_m1d": -0.0055990889671471145, "alpha_c_m1d": 0.018200836080618288, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02591077088622873, "ret_signed_p1d": -0.02591077088622873, "alpha_spy_p1d": -0.03531667053833565, "alpha_c_p1d": -0.028535295905334435, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": -0.014407031299373929, "ret_signed_p1w": -0.014407031299373929, "alpha_spy_p1w": -0.03166372528793093, "alpha_c_p1w": -0.055218602968658215, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.036540847884902306, "ret_signed_p1m": 0.036540847884902306, "alpha_spy_p1m": 0.00474490134740746, "alpha_c_p1m": -0.037680744931153365, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.029364291969660616, "ret_signed_p3m": 0.029364291969660616, "alpha_spy_p3m": 0.002787705511193117, "alpha_c_p3m": -0.1840009080598759, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 0.20862345351439582, "ret_signed_p6m": 0.20862345351439582, "alpha_spy_p6m": 0.10517682354106728, "alpha_c_p6m": -0.4004017566940443, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.0053, -0.0131, -0.0459, -0.0459, -0.0645, -0.0654, -0.0597, -0.0851, -0.0781, -0.0646, -0.0286, -0.0295, -0.0259, -0.0263, -0.042, -0.0672, -0.0346, -0.0322, 0.0058, -0.0226, -0.0306, -0.0266, -0.0239, -0.0038, 0.0221, 0.0257, 0.0347, 0.0278, 0.0164, 0.0136, 0.0359, 0.0549, 0.0033, 0.0316, -0.0138, -0.0149, -0.0041, 0.0037, 0.0005, -0.0076, -0.0124, -0.0021, 0.0203, 0.049, 0.1012, 0.1342, 0.1114, 0.1092, 0.1333, 0.0884, 0.0694, 0.0303, 0.0307, 0.0904, 0.0581, 0.0616, 0.0236, 0.0417, 0.0222, -0.0065, 0.0217, -0.0105, -0.0201, 0.0, -0.0259, -0.0125, -0.0125, -0.0304, -0.0144, -0.006, -0.0162, 0.0046, -0.0007, 0.0165, 0.0133, 0.0068, -0.0088, -0.0412, -0.0342, -0.0264, -0.0635, -0.046, -0.0085, 0.0063, 0.0365, 0.0333, 0.0333, 0.0438, 0.0311, 0.0274, 0.0217, 0.0217, 0.0346, 0.0306, 0.0257, 0.036, 0.0137, 0.0127, 0.0131, 0.0179, 0.0033, 0.0247, 0.0202, 0.0202, -0.0245, 0.0043, 0.0126, 0.0281, 0.0215, 0.0328, 0.0492, 0.0546, 0.0471, 0.0168, -0.0121, -0.0457, -0.0584, 0.0157, 0.0411, 0.0329, 0.0411, 0.0241, 0.0015, 0.0015, 0.0133, 0.0298, 0.0294, 0.0399, 0.0494, 0.0565, 0.0713, 0.0129, -0.0293, -0.0003, -0.0136, 0.0027, 0.0044, -0.0259, 0.0006, 0.0122, 0.0192, 0.0033, -0.0125, 0.0038, -0.0033, -0.0117, -0.0217, -0.0539, -0.0377, -0.0402, -0.0211, -0.0619, -0.0822, -0.0951, -0.0445, -0.0371, -0.0282, -0.0282, -0.0268, -0.0243, -0.0025, 0.0076, 0.0334, 0.0371, 0.0766, 0.0895, 0.0867, 0.1049, 0.107, 0.0951, 0.1094, 0.0937, 0.141, 0.1867, 0.1679, 0.1464, 0.0934, 0.0872, 0.0874, 0.0765, 0.1386, 0.1587, 0.179, 0.2022, 0.2096, 0.2372, 0.2915, 0.2344, 0.218, 0.2086]}
{"unit_id": "reply:1978990067931623606", "ticker_idx": 0, "ts": "2025-10-17T01:02:54+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Buy TSMC stock", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Terse buy instruction on TSMC.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "tl", "tweet_text": "@edwardtoung Buy TSMC stock", "reply_text": null, "parent_text": "@Jukanlosreve What’s the summary", "quoted_text": null, "parent_author": "edwardtoung", "ret_m1d": 0.016131256361758384, "ret_signed_m1d": 0.016131256361758384, "alpha_spy_m1d": 0.02177551819519552, "alpha_c_m1d": 0.014760331120714731, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008879111718613775, "ret_signed_p1d": 0.008879111718613775, "alpha_spy_p1d": -0.001521363681007859, "alpha_c_p1d": -0.003926131889613549, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": -0.00040661177788969827, "ret_signed_p1w": -0.00040661177788969827, "alpha_spy_p1w": -0.019762682546799915, "alpha_c_p1w": -0.024646173753300693, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": -0.0442930217099492, "ret_signed_p1m": -0.0442930217099492, "alpha_spy_p1m": -0.04621952841706023, "alpha_c_p1m": -0.03691325961320602, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.11153953826372809, "ret_signed_p3m": 0.11153953826372809, "alpha_spy_p3m": 0.06938071251261957, "alpha_c_p3m": -0.024762469253724717, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.25935987035011876, "ret_signed_p6m": 0.25935987035011876, "alpha_spy_p6m": 0.22080331279919552, "alpha_c_p6m": -0.037841416978417364, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.2075, -0.1881, -0.1838, -0.1703, -0.1799, -0.1847, -0.1794, -0.1837, -0.2054, -0.1926, -0.2147, -0.2184, -0.1804, -0.183, -0.1822, -0.1747, -0.1844, -0.1858, -0.193, -0.1845, -0.2139, -0.2277, -0.232, -0.2129, -0.2041, -0.1935, -0.1916, -0.1951, -0.2201, -0.2201, -0.2284, -0.2183, -0.2054, -0.1777, -0.1649, -0.1523, -0.1202, -0.1253, -0.1239, -0.117, -0.1119, -0.1094, -0.0896, -0.1024, -0.0761, -0.0419, -0.0487, -0.0624, -0.0736, -0.074, -0.0535, -0.0224, -0.0236, -0.0098, 0.0248, -0.0036, 0.032, 0.0163, -0.0489, 0.0265, 0.0029, 0.0326, 0.0161, 0.0, 0.0089, -0.0019, -0.021, -0.0147, -0.0004, 0.0107, 0.0219, 0.0339, 0.0276, 0.0181, 0.0331, -0.0035, -0.0049, -0.0198, -0.0291, 0.0006, -0.0133, -0.0151, -0.0436, -0.0348, -0.0443, -0.0582, -0.0431, -0.0596, -0.0678, -0.0354, -0.0352, -0.0174, -0.0174, -0.0121, -0.0251, -0.0101, 0.0013, -0.0073, -0.0012, 0.023, 0.0282, 0.051, 0.0359, -0.0076, -0.0222, -0.0252, -0.0589, -0.0326, -0.0181, -0.0034, 0.0091, 0.0153, 0.0153, 0.0291, 0.0225, 0.018, 0.0326, 0.0326, 0.0861, 0.095, 0.1126, 0.0829, 0.0806, 0.0997, 0.1274, 0.1255, 0.1115, 0.1609, 0.1635, 0.1635, 0.1117, 0.1082, 0.1124, 0.1379, 0.1306, 0.1497, 0.1632, 0.1538, 0.1233, 0.16, 0.1409, 0.1069, 0.1238, 0.1854, 0.2077, 0.2298, 0.2712, 0.2508, 0.2449, 0.2449, 0.2376, 0.231, 0.2246, 0.2591, 0.2574, 0.3108, 0.3175, 0.2804, 0.2728, 0.2543, 0.2, 0.2146, 0.2024, 0.1516, 0.1849, 0.1794, 0.2048, 0.1442, 0.1496, 0.1561, 0.179, 0.1571, 0.1545, 0.1219, 0.1533, 0.1697, 0.185, 0.1113, 0.1134, 0.0785, 0.1516, 0.1637, 0.1553, 0.1553, 0.1646, 0.1767, 0.2469, 0.2455, 0.2629, 0.2594]}
{"unit_id": "orig:1978802728420995324", "ticker_idx": 0, "ts": "2025-10-16T12:38:29+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We are raising both our above-consensus estimates and target price, while maintaining our Buy rating on Micron (MU)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Citi reiterates Buy on MU, raises estimates and PT; expects DRAM long-term AI contracts to drive gross margins to 60% and peak EPS >$23.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi) Micron – Following the Footsteps of AI: DRAM Expected to Face Unprecedented Demand Next; Margins to Return to Peak Levels, Estimates and Target Price Raised\n\nWe believe that DRAM will be the next semiconductor to sign long-term contracts with AI-related companies — similar to what has happened with NVIDIA (NVDA), AMD, and Broadcom (AVGO) — due to its critical role in the AI ecosystem and the emerging supply shortage. This will benefit Micron through higher and more sustainable DRAM prices, driving its gross margins back to the peak 60% range and boosting peak EPS to over $23.00, nearly double its previous record of $12.26.\n\nWe are raising both our above-consensus estimates and target price, while maintaining our Buy rating on Micron (MU).\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05228857475168858, "ret_signed_m1d": -0.05228857475168858, "alpha_spy_m1d": -0.05914553294894942, "alpha_c_m1d": -0.0478610368068092, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.004427537944879378, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0007406080873467724, "ret_signed_p1d": -0.0007406080873467724, "alpha_spy_p1d": -0.006416908446254754, "alpha_c_p1d": 0.0006284402907192899, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": -0.0013690483780660623, "ret_p1w": 0.02063892716637472, "ret_signed_p1w": 0.02063892716637472, "alpha_spy_p1w": 0.0038068178817045606, "alpha_c_p1w": 0.01626952896893874, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.004369398197435981, "ret_p1m": 0.2187330413981221, "ret_signed_p1m": 0.2187330413981221, "alpha_spy_p1m": 0.20164356041185139, "alpha_c_p1m": 0.21383922624560192, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.004893815152520187, "ret_p3m": 0.6702048510508347, "ret_signed_p3m": 0.6702048510508347, "alpha_spy_p3m": 0.616953438108067, "alpha_c_p3m": 0.5261666585870479, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.1440381924637868, "ret_p6m": 1.078391654397973, "ret_signed_p6m": 1.078391654397973, "alpha_spy_p6m": 1.044047955953613, "alpha_c_p6m": 0.8018421877721944, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.2765494666257786, "price_path": [-0.4413, -0.4611, -0.458, -0.4487, -0.451, -0.451, -0.4475, -0.4338, -0.4615, -0.4825, -0.4682, -0.4618, -0.4632, -0.448, -0.4133, -0.3895, -0.3696, -0.3868, -0.3818, -0.4036, -0.3903, -0.3978, -0.4216, -0.4286, -0.4193, -0.4255, -0.4251, -0.419, -0.398, -0.4128, -0.4128, -0.4154, -0.4142, -0.3871, -0.3518, -0.3513, -0.3327, -0.3092, -0.257, -0.2242, -0.2215, -0.2163, -0.2105, -0.1666, -0.197, -0.1877, -0.1789, -0.2021, -0.2261, -0.224, -0.1912, -0.1744, -0.1012, -0.0933, -0.0726, -0.0571, -0.0831, -0.0296, -0.0504, -0.1033, -0.0482, -0.0764, -0.0523, 0.0, -0.0007, 0.0209, -0.0012, -0.02, 0.0206, 0.0814, 0.0868, 0.0957, 0.119, 0.1061, 0.1049, 0.1588, 0.0765, 0.1727, 0.1768, 0.1747, 0.2507, 0.1905, 0.2092, 0.17, 0.2187, 0.1946, 0.1282, 0.1155, -0.0057, 0.0239, 0.1057, 0.1086, 0.1369, 0.1369, 0.1676, 0.1873, 0.1825, 0.1562, 0.1191, 0.1713, 0.2192, 0.2463, 0.3021, 0.2762, 0.1906, 0.1727, 0.148, 0.1135, 0.2272, 0.313, 0.3657, 0.3641, 0.4155, 0.4155, 0.4062, 0.4541, 0.4455, 0.4098, 0.4098, 0.558, 0.5419, 0.6964, 0.6772, 0.6153, 0.7046, 0.7084, 0.6702, 0.6466, 0.6628, 0.7918, 0.7918, 0.8029, 0.922, 0.9639, 0.9741, 0.9219, 1.0264, 1.1501, 1.1526, 1.0493, 1.1625, 1.0718, 0.8741, 0.8913, 0.9496, 0.8943, 0.8437, 1.0269, 1.0448, 1.0334, 1.0334, 0.9747, 1.0793, 1.0615, 1.115, 1.0794, 1.0648, 1.1191, 1.0527, 1.0369, 1.0384, 0.8754, 0.9796, 0.9612, 0.8291, 0.9231, 0.9912, 1.0681, 1.0022, 1.1049, 1.1823, 1.2805, 1.2807, 1.1945, 1.0889, 0.9973, 0.9537, 0.8873, 0.7558, 0.7645, 0.5902, 0.6695, 0.8178, 0.8098, 0.8098, 0.8667, 0.8659, 1.0099, 1.0829, 1.0784]}
{"unit_id": "quote:1989133785045369200", "ticker_idx": 0, "ts": "2025-11-14T00:50:25+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "U.S. broker raises CoWoS capacity forecast", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author shares/endorses U.S. broker upgrades to CoWoS capacity and wafer demand forecasts for TSMC, NVDA, AVGO, and AMD into 2026-2027.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• 2330 TSMC: U.S. broker raises CoWoS capacity forecast\n\nThe U.S. securities house has revised up its CoWoS capacity outlook. It expects capacity to reach 105,000 wafers per month in 4Q26 (previously 95,000 wafers), and to reach 141,000 wafers per month by the end of 2027. The updates to wafer usage by major customers are as follows:\n\n• Nvidia: Revised up\n• The outlook for next year’s CoWoS wafer usage has been revised up to around 700,000 wafers (previously 580,000 wafers).\n• Rubin is expected to start ramping from 2Q26.\n• However, next year is still expected to be constrained by N3 and CoWoS-L capacity.\n• CVP is expected to adopt CoWoS-R, as it is introducing SoCAMM memory for the first time.\n• Rubin CPX is not expected to use CoWoS-S, and is ultimately likely to be manufactured with an FCBGA process.\n    \n• Broadcom (博通): Revised up\n• The current expected wafer usage is 20,000 wafers, but actual customer demand may be higher than this.\n• Demand for Google TPU (v6p / v7p) is strong, and Tomahawk 6 is scheduled to adopt CoWoS-S (Tomahawk 5 is mostly flip-chip BGA).\n• Anthropic and internal token usage are both on an uptrend.\n    \n• AMD: Revised up\n• Next year’s CoWoS demand has been revised up from 80,000 wafers to 90,000 wafers, reflecting strong demand for the MI450 series and Venice CPU, both of which are likely to use CoWoS.\n• However, the broker still sees uncertainty in AMD’s outlook. Since the production schedule is mostly concentrated in the second half of next year, any issues with OpenAI’s custom designs or Helios rack assembly could cause delays.\n    \n• AWS: Unchanged (stable)\n• Trainium 3 remains at around 80,000 wafers, with first wafer shipments expected to occur in 2Q26.\n• For next year, around 700,000 additional Trainium 2 / 2.5 chips are also expected.\n    \n• Outsourcing (OSAT): Significant increase\n• Some of the substrate packaging steps for CoWoS-L have already been outsourced, and by next year about 70% is expected to be outsourced to 3711 ASE (日月光).\n    \n• 3D-SoIC\n• TSMC is making early capital investments to expand N3-SoIC capacity.\n• It is observed that the goal is to increase N3-SoIC capacity to 22,000 wafers per month / 30,000 wafers per month next year / the year after.\n    \n• CoPoS (panel-level packaging)\n• It appears likely to become a mainstream process route going forward.\n• TSMC and the OSAT ecosystem are actively adopting 310×310mm square panels as the next-generation interposer material after 2028.\n• However, since it remains quite difficult to secure good yield at very fine L/S (line width/spacing), CoWoP is likely to take the position of a lower-cost but not cutting-edge packaging solution going forward.\n    \n• WMCM (wafer-level module packaging)\n• There is no change in the capacity outlook.\n• Non-Apple customers (e.g., Qualcomm) are also seen to be interested in applying WMCM-type solutions to PC or flagship smartphone processors.\n\n$NVDA $AMD $AMZN $AVGO $TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/14 外電綜合整理\n\n- 2330台積電：美系上修CoWoS 產能\n美系券商上修CoWoS產能在4Q26可達到每月 10.5 萬片（先前為 9.5 萬片），預估 在2027 年底可達每月 14.1 萬片。主要客戶用量更新如下：\n\n*Nvidia: 上修。預估明年 CoWoS 晶圓用量至約 70 萬片（原先為 58 萬片）。預期 Rubin", "parent_author": null, "ret_m1d": 0.02097902398349083, "ret_signed_m1d": 0.02097902398349083, "alpha_spy_m1d": 0.020815411119116556, "alpha_c_m1d": 0.021442852955098513, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010489469027826104, "ret_signed_p1d": 0.010489469027826104, "alpha_spy_p1d": 0.019805992958788265, "alpha_c_p1d": 0.024055601260345782, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.031468493011317045, "ret_signed_p1w": -0.031468493011317045, "alpha_spy_p1w": -0.01227006267166364, "alpha_c_p1w": 0.02317237008149131, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": 0.01736593023071964, "ret_signed_p1m": 0.01736593023071964, "alpha_spy_p1m": 0.004269410151191533, "alpha_c_p1m": -0.005591873489850796, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.3436246527656046, "ret_signed_p3m": 0.3436246527656046, "alpha_spy_p3m": 0.3107719935038742, "alpha_c_p3m": 0.1375529922099683, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.5732631062294933, "ret_signed_p6m": 0.5732631062294933, "alpha_spy_p6m": 0.4667348707234371, "alpha_c_p6m": -0.1024955074835876, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.1746, -0.2095, -0.199, -0.2095, -0.1851, -0.1816, -0.1711, -0.192, -0.192, -0.1886, -0.192, -0.192, -0.192, -0.1781, -0.1781, -0.1642, -0.1468, -0.1363, -0.1224, -0.1259, -0.1049, -0.1154, -0.1014, -0.1154, -0.0944, -0.0629, -0.0629, -0.0769, -0.0909, -0.0909, -0.0874, -0.0734, -0.0455, -0.021, -0.021, 0.0035, -0.0105, 0.007, 0.007, -0.0105, -0.0035, 0.0245, 0.0385, 0.014, 0.035, 0.035, 0.021, 0.014, 0.014, 0.035, 0.0315, 0.0524, 0.0524, 0.049, 0.0559, 0.0524, 0.021, 0.0245, 0.021, 0.0315, 0.0245, 0.0315, 0.021, 0.0, 0.0105, -0.0175, -0.0245, 0.0175, -0.0315, -0.0385, -0.0105, 0.007, 0.0035, 0.007, -0.014, 0.0, 0.014, 0.0105, 0.021, 0.0455, 0.035, 0.0524, 0.0314, 0.0384, 0.0174, 0.0068, 0.0033, 0.0033, 0.0033, 0.0279, 0.0454, 0.0489, 0.0489, 0.0595, 0.0735, 0.0665, 0.0875, 0.0875, 0.1121, 0.1717, 0.1963, 0.1752, 0.1822, 0.1787, 0.1858, 0.1998, 0.1998, 0.1858, 0.2208, 0.2349, 0.2454, 0.2208, 0.2349, 0.2419, 0.2314, 0.2489, 0.277, 0.2664, 0.2454, 0.2384, 0.2629, 0.2524, 0.2384, 0.2489, 0.2735, 0.3191, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3331, 0.3787, 0.4138, 0.3998, 0.3998, 0.3857, 0.3577, 0.3085, 0.3331, 0.3261, 0.27, 0.298, 0.3612, 0.3226, 0.3085, 0.2945, 0.3163, 0.341, 0.3023, 0.2952, 0.2741, 0.2741, 0.2987, 0.2952, 0.2811, 0.253, 0.2389, 0.3058, 0.2741, 0.2741, 0.2741, 0.3093, 0.3726, 0.3762, 0.4078, 0.4008, 0.4466, 0.4642, 0.4677, 0.429, 0.4254, 0.443, 0.443, 0.4642, 0.5381, 0.5944, 0.5592, 0.5345, 0.5029, 0.5029, 0.6014, 0.5838, 0.5838, 0.6261, 0.612, 0.5733]}
{"unit_id": "quote:1989133785045369200", "ticker_idx": 1, "ts": "2025-11-14T00:50:25+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The outlook for next year's CoWoS wafer usage has been revised up to around 700,000 wafers (previously 580,000 wafers)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author shares/endorses U.S. broker upgrades to CoWoS capacity and wafer demand forecasts for TSMC, NVDA, AVGO, and AMD into 2026-2027.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• 2330 TSMC: U.S. broker raises CoWoS capacity forecast\n\nThe U.S. securities house has revised up its CoWoS capacity outlook. It expects capacity to reach 105,000 wafers per month in 4Q26 (previously 95,000 wafers), and to reach 141,000 wafers per month by the end of 2027. The updates to wafer usage by major customers are as follows:\n\n• Nvidia: Revised up\n• The outlook for next year’s CoWoS wafer usage has been revised up to around 700,000 wafers (previously 580,000 wafers).\n• Rubin is expected to start ramping from 2Q26.\n• However, next year is still expected to be constrained by N3 and CoWoS-L capacity.\n• CVP is expected to adopt CoWoS-R, as it is introducing SoCAMM memory for the first time.\n• Rubin CPX is not expected to use CoWoS-S, and is ultimately likely to be manufactured with an FCBGA process.\n    \n• Broadcom (博通): Revised up\n• The current expected wafer usage is 20,000 wafers, but actual customer demand may be higher than this.\n• Demand for Google TPU (v6p / v7p) is strong, and Tomahawk 6 is scheduled to adopt CoWoS-S (Tomahawk 5 is mostly flip-chip BGA).\n• Anthropic and internal token usage are both on an uptrend.\n    \n• AMD: Revised up\n• Next year’s CoWoS demand has been revised up from 80,000 wafers to 90,000 wafers, reflecting strong demand for the MI450 series and Venice CPU, both of which are likely to use CoWoS.\n• However, the broker still sees uncertainty in AMD’s outlook. Since the production schedule is mostly concentrated in the second half of next year, any issues with OpenAI’s custom designs or Helios rack assembly could cause delays.\n    \n• AWS: Unchanged (stable)\n• Trainium 3 remains at around 80,000 wafers, with first wafer shipments expected to occur in 2Q26.\n• For next year, around 700,000 additional Trainium 2 / 2.5 chips are also expected.\n    \n• Outsourcing (OSAT): Significant increase\n• Some of the substrate packaging steps for CoWoS-L have already been outsourced, and by next year about 70% is expected to be outsourced to 3711 ASE (日月光).\n    \n• 3D-SoIC\n• TSMC is making early capital investments to expand N3-SoIC capacity.\n• It is observed that the goal is to increase N3-SoIC capacity to 22,000 wafers per month / 30,000 wafers per month next year / the year after.\n    \n• CoPoS (panel-level packaging)\n• It appears likely to become a mainstream process route going forward.\n• TSMC and the OSAT ecosystem are actively adopting 310×310mm square panels as the next-generation interposer material after 2028.\n• However, since it remains quite difficult to secure good yield at very fine L/S (line width/spacing), CoWoP is likely to take the position of a lower-cost but not cutting-edge packaging solution going forward.\n    \n• WMCM (wafer-level module packaging)\n• There is no change in the capacity outlook.\n• Non-Apple customers (e.g., Qualcomm) are also seen to be interested in applying WMCM-type solutions to PC or flagship smartphone processors.\n\n$NVDA $AMD $AMZN $AVGO $TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/14 外電綜合整理\n\n- 2330台積電：美系上修CoWoS 產能\n美系券商上修CoWoS產能在4Q26可達到每月 10.5 萬片（先前為 9.5 萬片），預估 在2027 年底可達每月 14.1 萬片。主要客戶用量更新如下：\n\n*Nvidia: 上修。預估明年 CoWoS 晶圓用量至約 70 萬片（原先為 58 萬片）。預期 Rubin", "parent_author": null, "ret_m1d": -0.017405531691603637, "ret_signed_m1d": -0.017405531691603637, "alpha_spy_m1d": -0.01756914455597791, "alpha_c_m1d": -0.016941702719995955, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018772690180637097, "ret_signed_p1d": -0.018772690180637097, "alpha_spy_p1d": -0.009456166249674935, "alpha_c_p1d": -0.0052065579481174185, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.05936791649684514, "ret_signed_p1w": -0.05936791649684514, "alpha_spy_p1w": -0.040169486157191736, "alpha_c_p1w": -0.004727053404036785, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.07293574274855108, "ret_signed_p1m": -0.07293574274855108, "alpha_spy_p1m": -0.08603226282807919, "alpha_c_p1m": -0.09589354646912152, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.0005753669288237617, "ret_signed_p3m": -0.0005753669288237617, "alpha_spy_p3m": -0.03342802619055418, "alpha_c_p3m": -0.20664702748446007, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.15404161563996577, "ret_signed_p6m": 0.15404161563996577, "alpha_spy_p6m": 0.0475133801339096, "alpha_c_p6m": -0.5217169980731151, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.0765, -0.0777, -0.0799, -0.0641, -0.0545, -0.0442, -0.0451, -0.0526, -0.0841, -0.0841, -0.102, -0.1029, -0.0974, -0.1218, -0.115, -0.1021, -0.0676, -0.0684, -0.0649, -0.0653, -0.0804, -0.1045, -0.0733, -0.071, -0.0345, -0.0617, -0.0694, -0.0656, -0.063, -0.0438, -0.0189, -0.0154, -0.0067, -0.0134, -0.0243, -0.027, -0.0056, 0.0126, -0.0369, -0.0097, -0.0533, -0.0544, -0.044, -0.0365, -0.0396, -0.0474, -0.052, -0.0421, -0.0206, 0.0069, 0.0571, 0.0887, 0.0669, 0.0648, 0.0879, 0.0448, 0.0265, -0.011, -0.0106, 0.0467, 0.0157, 0.0191, -0.0174, 0.0, -0.0188, -0.0463, -0.0192, -0.0501, -0.0594, -0.0401, -0.0649, -0.0521, -0.0521, -0.0693, -0.0539, -0.0458, -0.0556, -0.0357, -0.0408, -0.0242, -0.0273, -0.0335, -0.0485, -0.0796, -0.0729, -0.0654, -0.1011, -0.0842, -0.0482, -0.034, -0.005, -0.0081, -0.0081, 0.0019, -0.0102, -0.0138, -0.0192, -0.0192, -0.0069, -0.0107, -0.0154, -0.0055, -0.0269, -0.0279, -0.0274, -0.0229, -0.0369, -0.0164, -0.0207, -0.0207, -0.0636, -0.036, -0.028, -0.0131, -0.0194, -0.0086, 0.0072, 0.0124, 0.0051, -0.0239, -0.0516, -0.084, -0.0961, -0.025, -0.0006, -0.0085, -0.0006, -0.0169, -0.0386, -0.0386, -0.0273, -0.0115, -0.0119, -0.0018, 0.0073, 0.0141, 0.0284, -0.0277, -0.0682, -0.0404, -0.0532, -0.0374, -0.0359, -0.0649, -0.0395, -0.0283, -0.0217, -0.0369, -0.0521, -0.0364, -0.0432, -0.0513, -0.0609, -0.0918, -0.0763, -0.0786, -0.0603, -0.0994, -0.119, -0.1314, -0.0828, -0.0757, -0.0671, -0.0671, -0.0658, -0.0634, -0.0424, -0.0328, -0.008, -0.0044, 0.0335, 0.0459, 0.0431, 0.0606, 0.0626, 0.0512, 0.065, 0.0499, 0.0953, 0.1392, 0.1211, 0.1005, 0.0495, 0.0437, 0.0438, 0.0334, 0.093, 0.1123, 0.1317, 0.154]}
{"unit_id": "quote:1989133785045369200", "ticker_idx": 2, "ts": "2025-11-14T00:50:25+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom (博通): Revised up. The current expected wafer usage is 20,000 wafers, but actual customer demand may be higher than this", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author shares/endorses U.S. broker upgrades to CoWoS capacity and wafer demand forecasts for TSMC, NVDA, AVGO, and AMD into 2026-2027.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• 2330 TSMC: U.S. broker raises CoWoS capacity forecast\n\nThe U.S. securities house has revised up its CoWoS capacity outlook. It expects capacity to reach 105,000 wafers per month in 4Q26 (previously 95,000 wafers), and to reach 141,000 wafers per month by the end of 2027. The updates to wafer usage by major customers are as follows:\n\n• Nvidia: Revised up\n• The outlook for next year’s CoWoS wafer usage has been revised up to around 700,000 wafers (previously 580,000 wafers).\n• Rubin is expected to start ramping from 2Q26.\n• However, next year is still expected to be constrained by N3 and CoWoS-L capacity.\n• CVP is expected to adopt CoWoS-R, as it is introducing SoCAMM memory for the first time.\n• Rubin CPX is not expected to use CoWoS-S, and is ultimately likely to be manufactured with an FCBGA process.\n    \n• Broadcom (博通): Revised up\n• The current expected wafer usage is 20,000 wafers, but actual customer demand may be higher than this.\n• Demand for Google TPU (v6p / v7p) is strong, and Tomahawk 6 is scheduled to adopt CoWoS-S (Tomahawk 5 is mostly flip-chip BGA).\n• Anthropic and internal token usage are both on an uptrend.\n    \n• AMD: Revised up\n• Next year’s CoWoS demand has been revised up from 80,000 wafers to 90,000 wafers, reflecting strong demand for the MI450 series and Venice CPU, both of which are likely to use CoWoS.\n• However, the broker still sees uncertainty in AMD’s outlook. Since the production schedule is mostly concentrated in the second half of next year, any issues with OpenAI’s custom designs or Helios rack assembly could cause delays.\n    \n• AWS: Unchanged (stable)\n• Trainium 3 remains at around 80,000 wafers, with first wafer shipments expected to occur in 2Q26.\n• For next year, around 700,000 additional Trainium 2 / 2.5 chips are also expected.\n    \n• Outsourcing (OSAT): Significant increase\n• Some of the substrate packaging steps for CoWoS-L have already been outsourced, and by next year about 70% is expected to be outsourced to 3711 ASE (日月光).\n    \n• 3D-SoIC\n• TSMC is making early capital investments to expand N3-SoIC capacity.\n• It is observed that the goal is to increase N3-SoIC capacity to 22,000 wafers per month / 30,000 wafers per month next year / the year after.\n    \n• CoPoS (panel-level packaging)\n• It appears likely to become a mainstream process route going forward.\n• TSMC and the OSAT ecosystem are actively adopting 310×310mm square panels as the next-generation interposer material after 2028.\n• However, since it remains quite difficult to secure good yield at very fine L/S (line width/spacing), CoWoP is likely to take the position of a lower-cost but not cutting-edge packaging solution going forward.\n    \n• WMCM (wafer-level module packaging)\n• There is no change in the capacity outlook.\n• Non-Apple customers (e.g., Qualcomm) are also seen to be interested in applying WMCM-type solutions to PC or flagship smartphone processors.\n\n$NVDA $AMD $AMZN $AVGO $TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/14 外電綜合整理\n\n- 2330台積電：美系上修CoWoS 產能\n美系券商上修CoWoS產能在4Q26可達到每月 10.5 萬片（先前為 9.5 萬片），預估 在2027 年底可達每月 14.1 萬片。主要客戶用量更新如下：\n\n*Nvidia: 上修。預估明年 CoWoS 晶圓用量至約 70 萬片（原先為 58 萬片）。預期 Rubin", "parent_author": null, "ret_m1d": -0.007241573924393063, "ret_signed_m1d": -0.007241573924393063, "alpha_spy_m1d": -0.007405186788767337, "alpha_c_m1d": -0.006777744952785381, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0005548968528883425, "ret_signed_p1d": 0.0005548968528883425, "alpha_spy_p1d": 0.009871420783850504, "alpha_c_p1d": 0.014121029085408021, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.006599174794477025, "ret_signed_p1w": -0.006599174794477025, "alpha_spy_p1w": 0.01259925554517638, "alpha_c_p1w": 0.04804168829833133, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.007738046455298853, "ret_signed_p1m": -0.007738046455298853, "alpha_spy_p1m": -0.02083456653482696, "alpha_c_p1m": -0.030695850175869288, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.0027912152724218764, "ret_signed_p3m": 0.0027912152724218764, "alpha_spy_p3m": -0.030061443989308545, "alpha_c_p3m": -0.20328044528321443, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.25605975636073564, "ret_signed_p6m": 0.25605975636073564, "alpha_spy_p6m": 0.14953152085467947, "alpha_c_p6m": -0.41969885735234524, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.1403, -0.1512, -0.1558, -0.143, -0.1423, -0.1313, -0.1248, -0.1003, -0.1331, -0.1331, -0.1306, -0.1185, -0.1077, -0.0238, 0.0076, -0.0186, 0.0773, 0.0483, 0.049, 0.0613, 0.0494, 0.0091, 0.0067, 0.0055, -0.0107, -0.0103, -0.0092, -0.0186, -0.0232, -0.0425, -0.0366, -0.0265, -0.0125, -0.0119, -0.0204, -0.0177, 0.0089, 0.0075, -0.0521, 0.0416, 0.0049, 0.0259, 0.0341, 0.0201, 0.0198, 0.0006, -0.0063, 0.0053, 0.0341, 0.0572, 0.0891, 0.1271, 0.0993, 0.0793, 0.0587, 0.0277, 0.0482, 0.0383, 0.0204, 0.0465, 0.0277, 0.0373, -0.0072, 0.0, 0.0006, -0.0057, 0.0349, 0.0127, -0.0066, 0.1037, 0.1243, 0.1609, 0.1609, 0.1767, 0.1274, 0.1142, 0.1114, 0.1126, 0.1395, 0.1712, 0.1864, 0.2059, 0.1866, 0.051, -0.0077, -0.0034, -0.048, -0.0367, -0.0061, -0.001, 0.022, 0.0246, 0.0246, 0.0302, 0.0222, 0.0235, 0.0126, 0.0126, 0.017, 0.0047, 0.0057, 0.005, -0.0273, 0.0093, 0.0304, 0.0375, -0.0056, 0.0036, 0.029, 0.029, -0.0269, -0.0381, -0.0477, -0.0637, -0.0496, -0.0264, -0.0251, -0.0324, -0.0307, -0.0313, -0.0628, -0.0988, -0.0916, -0.026, 0.0062, -0.004, 0.0028, -0.0311, -0.0487, -0.0487, -0.0271, -0.0243, -0.0229, -0.0268, -0.0335, -0.0477, -0.0278, -0.0588, -0.0651, -0.0672, -0.0818, -0.071, -0.0264, -0.0331, 0.0115, 0.0023, -0.0007, -0.0171, -0.0575, -0.0494, -0.06, -0.0757, -0.0643, -0.0916, -0.0545, -0.0668, -0.0653, -0.0929, -0.1185, -0.1398, -0.0926, -0.0809, -0.0778, -0.0778, -0.0782, -0.0209, 0.028, 0.0405, 0.0893, 0.1133, 0.1164, 0.1631, 0.1682, 0.1919, 0.1716, 0.1791, 0.2391, 0.2312, 0.2394, 0.2261, 0.1722, 0.1887, 0.2238, 0.2351, 0.2211, 0.2529, 0.2473, 0.2095, 0.2607, 0.2561]}
{"unit_id": "quote:1989133785045369200", "ticker_idx": 3, "ts": "2025-11-14T00:50:25+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD: Revised up. Next year's CoWoS demand has been revised up from 80,000 wafers to 90,000 wafers", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author shares/endorses U.S. broker upgrades to CoWoS capacity and wafer demand forecasts for TSMC, NVDA, AVGO, and AMD into 2026-2027.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "• 2330 TSMC: U.S. broker raises CoWoS capacity forecast\n\nThe U.S. securities house has revised up its CoWoS capacity outlook. It expects capacity to reach 105,000 wafers per month in 4Q26 (previously 95,000 wafers), and to reach 141,000 wafers per month by the end of 2027. The updates to wafer usage by major customers are as follows:\n\n• Nvidia: Revised up\n• The outlook for next year’s CoWoS wafer usage has been revised up to around 700,000 wafers (previously 580,000 wafers).\n• Rubin is expected to start ramping from 2Q26.\n• However, next year is still expected to be constrained by N3 and CoWoS-L capacity.\n• CVP is expected to adopt CoWoS-R, as it is introducing SoCAMM memory for the first time.\n• Rubin CPX is not expected to use CoWoS-S, and is ultimately likely to be manufactured with an FCBGA process.\n    \n• Broadcom (博通): Revised up\n• The current expected wafer usage is 20,000 wafers, but actual customer demand may be higher than this.\n• Demand for Google TPU (v6p / v7p) is strong, and Tomahawk 6 is scheduled to adopt CoWoS-S (Tomahawk 5 is mostly flip-chip BGA).\n• Anthropic and internal token usage are both on an uptrend.\n    \n• AMD: Revised up\n• Next year’s CoWoS demand has been revised up from 80,000 wafers to 90,000 wafers, reflecting strong demand for the MI450 series and Venice CPU, both of which are likely to use CoWoS.\n• However, the broker still sees uncertainty in AMD’s outlook. Since the production schedule is mostly concentrated in the second half of next year, any issues with OpenAI’s custom designs or Helios rack assembly could cause delays.\n    \n• AWS: Unchanged (stable)\n• Trainium 3 remains at around 80,000 wafers, with first wafer shipments expected to occur in 2Q26.\n• For next year, around 700,000 additional Trainium 2 / 2.5 chips are also expected.\n    \n• Outsourcing (OSAT): Significant increase\n• Some of the substrate packaging steps for CoWoS-L have already been outsourced, and by next year about 70% is expected to be outsourced to 3711 ASE (日月光).\n    \n• 3D-SoIC\n• TSMC is making early capital investments to expand N3-SoIC capacity.\n• It is observed that the goal is to increase N3-SoIC capacity to 22,000 wafers per month / 30,000 wafers per month next year / the year after.\n    \n• CoPoS (panel-level packaging)\n• It appears likely to become a mainstream process route going forward.\n• TSMC and the OSAT ecosystem are actively adopting 310×310mm square panels as the next-generation interposer material after 2028.\n• However, since it remains quite difficult to secure good yield at very fine L/S (line width/spacing), CoWoP is likely to take the position of a lower-cost but not cutting-edge packaging solution going forward.\n    \n• WMCM (wafer-level module packaging)\n• There is no change in the capacity outlook.\n• Non-Apple customers (e.g., Qualcomm) are also seen to be interested in applying WMCM-type solutions to PC or flagship smartphone processors.\n\n$NVDA $AMD $AMZN $AVGO $TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/14 外電綜合整理\n\n- 2330台積電：美系上修CoWoS 產能\n美系券商上修CoWoS產能在4Q26可達到每月 10.5 萬片（先前為 9.5 萬片），預估 在2027 年底可達每月 14.1 萬片。主要客戶用量更新如下：\n\n*Nvidia: 上修。預估明年 CoWoS 晶圓用量至約 70 萬片（原先為 58 萬片）。預期 Rubin", "parent_author": null, "ret_m1d": 0.0046594917817315284, "ret_signed_m1d": 0.0046594917817315284, "alpha_spy_m1d": 0.004495878917357254, "alpha_c_m1d": 0.005123320753339211, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02548516408716195, "ret_signed_p1d": -0.02548516408716195, "alpha_spy_p1d": -0.016168640156199787, "alpha_c_p1d": -0.01191903185464227, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.1743446343541305, "ret_signed_p1w": -0.1743446343541305, "alpha_spy_p1w": -0.1551462040144771, "alpha_c_p1w": -0.11970377126132214, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.1589481630225522, "ret_signed_p1m": -0.1589481630225522, "alpha_spy_p1m": -0.17204468310208032, "alpha_c_p1m": -0.18190596674312265, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.13463796465396471, "ret_signed_p3m": -0.13463796465396471, "alpha_spy_p3m": -0.16749062391569514, "alpha_c_p3m": -0.340709625209601, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.8588793528754974, "ret_signed_p6m": 0.8588793528754974, "alpha_spy_p6m": 0.7523511173694413, "alpha_c_p6m": 0.18312073916241656, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.3252, -0.3307, -0.3367, -0.3203, -0.3381, -0.3249, -0.3228, -0.317, -0.3411, -0.3411, -0.3423, -0.3431, -0.3445, -0.3876, -0.3865, -0.3687, -0.3536, -0.3693, -0.3575, -0.347, -0.3499, -0.3551, -0.3602, -0.3623, -0.3526, -0.3481, -0.3482, -0.3466, -0.3539, -0.3462, -0.3445, -0.3355, -0.3123, -0.3328, -0.1746, -0.143, -0.0456, -0.0564, -0.1293, -0.1231, -0.1164, -0.0333, -0.0496, -0.0556, -0.0253, -0.0356, -0.0672, -0.0479, 0.0248, 0.0521, 0.0454, 0.071, 0.0325, 0.0377, 0.052, 0.0131, 0.0386, -0.0369, -0.0538, -0.0115, -0.0376, 0.0489, 0.0047, 0.0, -0.0255, -0.0669, -0.0942, -0.1653, -0.1743, -0.1287, -0.1648, -0.132, -0.132, -0.1186, -0.1096, -0.1279, -0.1184, -0.1249, -0.1169, -0.1041, -0.1021, -0.1029, -0.1028, -0.146, -0.1589, -0.1525, -0.1973, -0.1854, -0.1352, -0.1291, -0.1293, -0.1287, -0.1287, -0.1289, -0.1264, -0.1275, -0.1323, -0.1323, -0.0946, -0.1043, -0.1315, -0.1491, -0.1707, -0.1768, -0.1585, -0.1047, -0.094, -0.0765, -0.0607, -0.0607, -0.0603, 0.0121, 0.028, 0.0521, 0.0182, 0.0211, 0.024, 0.0218, -0.0408, -0.0022, -0.019, -0.1889, -0.22, -0.1555, -0.1248, -0.1347, -0.1346, -0.1656, -0.16, -0.16, -0.1772, -0.1892, -0.176, -0.1891, -0.2034, -0.1336, -0.1457, -0.1747, -0.1888, -0.1953, -0.2263, -0.1813, -0.1919, -0.2203, -0.1788, -0.1766, -0.1701, -0.1988, -0.2164, -0.2035, -0.2046, -0.1918, -0.1683, -0.1843, -0.1788, -0.1679, -0.1075, -0.1744, -0.1816, -0.2057, -0.1758, -0.1483, -0.1188, -0.1188, -0.1079, -0.1024, -0.0607, -0.0412, -0.0072, 0.0001, 0.0335, 0.0458, 0.1274, 0.128, 0.114, 0.1527, 0.2295, 0.2371, 0.4092, 0.3558, 0.3095, 0.3659, 0.4363, 0.4608, 0.3838, 0.4394, 0.7073, 0.655, 0.8443, 0.8589]}
{"unit_id": "orig:2001410765912236262", "ticker_idx": 0, "ts": "2025-12-17T21:54:45+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This is ridiculously fantastic to the point it's hard to believe lol. Revenue: midpoint $18.7B (vs. consensus $14.38B) … EPS: midpoint $8.42 (vs. consensus $4.71)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "MU FY2Q26 guidance massively above consensus on revenue, GPM, and EPS; author calls it ridiculously fantastic — durable bullish read.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "FY2Q26 guidance (Non-GAAP basis)\n• Revenue: midpoint $18.7B (vs. consensus $14.38B)\n• GPM: midpoint 68% (vs. consensus 55.7%)\n• EPS: midpoint $8.42 (vs. consensus $4.71)\n\nThis is ridiculously fantastic to the point it’s hard to believe lol.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030995008162176596, "ret_signed_m1d": 0.030995008162176596, "alpha_spy_m1d": 0.019868961266861662, "alpha_c_m1d": -0.006441642094904054, "bench_spy_m1d": 0.011126046895314934, "bench_c_m1d": 0.03743665025708065, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10211953852468114, "ret_signed_p1d": 0.10211953852468114, "alpha_spy_p1d": 0.09456820594706628, "alpha_c_p1d": 0.0785078907446497, "bench_spy_p1d": 0.0075513325776148665, "bench_c_p1d": 0.02361164778003144, "ret_p1w": 0.2711954698921084, "ret_signed_p1w": 0.2711954698921084, "alpha_spy_p1w": 0.2398877379130342, "alpha_c_p1w": 0.19436733090230773, "bench_spy_p1w": 0.031307731979074216, "bench_c_p1w": 0.07682813898980068, "ret_p1m": 0.49328657843195867, "ret_signed_p1m": 0.49328657843195867, "alpha_spy_p1m": 0.45920035175136986, "alpha_c_p1m": 0.32112678827637464, "bench_spy_p1m": 0.034086226680588805, "bench_c_p1m": 0.17215979015558402, "ret_p3m": 0.9598193520409368, "ret_signed_p3m": 0.9598193520409368, "alpha_spy_p3m": 0.9604047457948046, "alpha_c_p3m": 0.7950223085457582, "bench_spy_p3m": -0.0005853937538677911, "bench_c_p3m": 0.16479704349517865, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2789, -0.2705, -0.2626, -0.2834, -0.305, -0.3031, -0.2737, -0.2585, -0.1928, -0.1857, -0.1671, -0.1532, -0.1766, -0.1285, -0.1472, -0.1947, -0.1452, -0.1705, -0.1489, -0.1019, -0.1026, -0.0831, -0.103, -0.1199, -0.0834, -0.0288, -0.024, -0.016, 0.0049, -0.0067, -0.0078, 0.0407, -0.0332, 0.0531, 0.0568, 0.055, 0.1232, 0.0691, 0.0859, 0.0507, 0.0945, 0.0729, 0.0132, 0.0018, -0.1071, -0.0805, -0.0071, -0.0044, 0.021, 0.021, 0.0486, 0.0662, 0.0619, 0.0383, 0.005, 0.0519, 0.0949, 0.1193, 0.1693, 0.1461, 0.0693, 0.0531, 0.031, 0.0, 0.1021, 0.1791, 0.2265, 0.225, 0.2712, 0.2712, 0.2628, 0.3058, 0.2981, 0.2661, 0.2661, 0.3992, 0.3847, 0.5235, 0.5062, 0.4507, 0.5308, 0.5343, 0.4999, 0.4787, 0.4933, 0.6092, 0.6092, 0.6191, 0.7261, 0.7637, 0.7728, 0.726, 0.8198, 0.9309, 0.9332, 0.8404, 0.9421, 0.8606, 0.683, 0.6985, 0.7508, 0.7012, 0.6557, 0.8203, 0.8364, 0.8261, 0.8261, 0.7734, 0.8673, 0.8514, 0.8994, 0.8674, 0.8543, 0.903, 0.8434, 0.8293, 0.8306, 0.6843, 0.7778, 0.7613, 0.6426, 0.727, 0.7882, 0.8573, 0.7981, 0.8903, 0.9598, 1.0481, 1.0482, 0.9708, 0.876, 0.7937, 0.7546, 0.6949, 0.5768, 0.5846, 0.4281, 0.4993, 0.6325, 0.6253, 0.6253, 0.6764, 0.6756, 0.805, 0.8706, 0.8665, 0.893, 1.0665, 1.0247, 1.0291, 1.0195, 0.99, 0.9943, 1.1634, 1.1378, 1.2044, 1.3279, 1.238, 1.3008, 1.2951, 1.4062, 1.5582, 1.8411, 1.9582, 1.8696, 2.3142, 2.5296, 2.402, 2.5664, 2.4438, 2.2159, 2.0246, 2.1009, 2.2485, 2.3821, 2.3328, 2.3328, 2.9758, 3.1201, 3.0984, 3.3092, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:1979187241537933575", "ticker_idx": 0, "ts": "2025-10-17T14:06:24+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If this project goes well, we expect the next few generations of the Maia chip to also be manufactured (fabbed) at Intel.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Bullish INTC: IFS wins Microsoft Maia 2 order with potential for multi-gen foundry relationship; 18A process proceeding smoothly.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"If this project goes well, we expect the next few generations of the Maia chip to also be manufactured (fabbed) at Intel. The last sentence is not an evaluation of Intel’s 18A or 18AP process, but rather a general statement about the project as a whole. From a process standpoint, as SemiAccurate has been saying for years, everything is proceeding smoothly without any issues.\"\n\nfrom Semiaccurate", "reply_text": null, "parent_text": null, "quoted_text": "Just in: IFS wins Microsoft Maia 2 chip order – SemiAccurate\n\n$INTC", "parent_author": null, "ret_m1d": -0.0045933038815190175, "ret_signed_m1d": -0.0045933038815190175, "alpha_spy_m1d": 0.001050957951918119, "alpha_c_m1d": -0.00596422912256267, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029451505053263105, "ret_signed_p1d": 0.029451505053263105, "alpha_spy_p1d": 0.01905102965364147, "alpha_c_p1d": 0.01664626144503578, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": 0.03431506391138517, "ret_signed_p1w": 0.03431506391138517, "alpha_spy_p1w": 0.014958993142474952, "alpha_c_p1w": 0.010075501935974174, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": -0.06214534832124963, "ret_signed_p1m": -0.06214534832124963, "alpha_spy_p1m": -0.06407185502836066, "alpha_c_p1m": -0.054765586224506446, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.31640106539418, "ret_signed_p3m": 0.31640106539418, "alpha_spy_p3m": 0.2742422396430715, "alpha_c_p3m": 0.1800990578767272, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.7611457244312456, "ret_signed_p6m": 0.7611457244312456, "alpha_spy_p6m": 0.7225891668803224, "alpha_c_p6m": 0.4639444371027095, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.3721, -0.3653, -0.3885, -0.4407, -0.4412, -0.4485, -0.4504, -0.465, -0.4782, -0.4731, -0.4545, -0.4485, -0.4658, -0.461, -0.442, -0.4107, -0.3996, -0.3553, -0.3364, -0.3607, -0.3161, -0.364, -0.365, -0.3299, -0.3367, -0.3421, -0.3286, -0.3264, -0.3421, -0.3421, -0.3459, -0.3515, -0.335, -0.3383, -0.3386, -0.3396, -0.3307, -0.335, -0.3494, -0.3307, -0.3172, -0.3272, -0.174, -0.2008, -0.2229, -0.2072, -0.1564, -0.0816, -0.0408, -0.0684, -0.0935, -0.0289, 0.0078, -0.0049, -0.0113, 0.0043, 0.0113, 0.0213, -0.0173, 0.0057, -0.0373, 0.0038, -0.0046, 0.0, 0.0295, 0.03, -0.0024, 0.0311, 0.0343, 0.0684, 0.1221, 0.117, 0.0851, 0.0805, 0.0673, 0.0005, 0.037, 0.0062, 0.0303, 0.0389, 0.0235, 0.0238, -0.0297, -0.0403, -0.0621, -0.0724, -0.0513, -0.0916, -0.0678, -0.033, -0.0319, -0.0054, -0.0054, 0.0959, 0.0811, 0.1745, 0.1824, 0.0943, 0.1189, 0.0889, 0.0943, 0.1019, 0.0675, 0.0216, 0.0135, 0.0081, -0.0259, -0.0197, -0.0051, -0.0173, -0.0178, -0.023, -0.023, -0.0219, -0.0089, 0.0078, -0.003, -0.003, 0.064, 0.0638, 0.0819, 0.1519, 0.1108, 0.2307, 0.1905, 0.2778, 0.3164, 0.3056, 0.2688, 0.2688, 0.3121, 0.4658, 0.4677, 0.2178, 0.1481, 0.187, 0.318, 0.3148, 0.2556, 0.3188, 0.3307, 0.3132, 0.3034, 0.3669, 0.3575, 0.2734, 0.3048, 0.2559, 0.2643, 0.2643, 0.2478, 0.2283, 0.2056, 0.1918, 0.1789, 0.2461, 0.2667, 0.2283, 0.2324, 0.2294, 0.1646, 0.2316, 0.2416, 0.1732, 0.2316, 0.264, 0.2964, 0.2226, 0.2367, 0.2364, 0.1905, 0.2167, 0.2478, 0.1854, 0.1891, 0.1905, 0.2748, 0.1916, 0.1654, 0.1129, 0.1924, 0.2978, 0.3613, 0.3613, 0.3721, 0.4296, 0.5928, 0.6677, 0.6855, 0.7611]}
{"unit_id": "thread:1989299736915173748", "ticker_idx": 0, "ts": "2025-11-14T11:49:51+00:00", "symbol": "memory chips", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "There is a high probability that memory chip prices will see another sharp rise in the first half of next year.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply-side cuts + AI-driven demand surge point to sharp memory price rise in H1 next year; bullish Samsung, Hynix, Micron and memory sector broadly.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory chips basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Many smartphone makers temporarily halt memory chip purchases… some have less than three weeks of inventory left\n\nOn November 14, Jiemian News (界面新闻) learned from exclusive sources that as upstream memory semiconductor (storage chip) prices have soared, multiple smartphone makers have already temporarily suspended their memory chip purchases for this quarter. Smartphone makers such as Xiaomi, OPPO and vivo generally have less than two months of inventory, and some manufacturers have less than three weeks of DRAM (dynamic random-access memory) inventory left. They are hesitating over whether to accept the quotes from original manufacturers (Micron, Samsung and SK hynix), which include price hikes of as much as nearly 50%.\n\nAn employee at one original manufacturer, Jackey, told Jiemian News, “We are not in a situation right now where we’re worried about our products not selling. If the smartphone makers don’t want them, we can allocate our capacity to server customers. They’re also short of supply at the moment, and we can sell to them at even higher prices.”\n\nThe surge in data-center demand for memory chips driven by the boom in large AI models is one of the factors behind this round of sharp memory price increases. Server makers are generally willing to pay higher prices in order to secure memory chips. According to information obtained by Jiemian News, for the same memory product, the prices that distributors quote to server customers are typically more than 30% higher than those offered to smartphone makers.\n\nAccording to a report by the Sci-Tech Innovation Board Daily (科创板日报), Samsung Electronics earlier became the first to suspend offering contract prices for DDR5 DRAM in October, and the issuance of new quotes is expected to be postponed until mid-November. This move drove spot prices for DDR5 up by 25% in just one week. At the same time, the four major memory makers—Samsung, SK hynix, Kioxia and Micron—have all cut their NAND flash supply in the second half of this year in a concerted effort to push prices higher.\n\nSamsung’s NAND wafer production target for this year has been lowered to 4.72 million wafers, about 7% down from last year’s level of 5.07 million. Kioxia has likewise reduced its annual output from 4.8 million wafers last year to 4.69 million. SK hynix’s NAND production is falling from 2.01 million wafers last year to around 1.8 million, a cut of 10%. Micron’s largest NAND manufacturing base, its Fab 7 plant in Singapore, is also keeping output at just over 300,000 wafers, maintaining a conservative stance on supply.\n\nThe report also pointed out that as more production lines are converted to 4-bit cell (QLC) processes in response to rising AI demand, an effective decline in NAND flash output will be unavoidable. North American tech giants, worried that NAND prices could spike, have already begun “panic stockpiling (panic buying),” and some suppliers’ NAND allocations for next year are already fully booked.\n\nA week ago, flash memory leader SanDisk sharply raised its November contract prices for NAND flash by 50% in one go, which in turn spurred further price-hike moves by memory giants such as Micron.\n\nJackey explained to Jiemian News that “the typical stockpiling (inventory) cycle for smartphone makers is about three months, and at present, most companies have just over one month of DRAM inventory and around two months of NAND inventory left.” He added, “Right now we’re in a situation of ‘undersupply,’ so if smartphone makers want to build finished sets (final products) properly, in the end they have no choice but to accept the prices that the original manufacturers are currently quoting—it’s only a matter of time. That said, from the perspective of the original manufacturers, they cannot ignore their relationships with smartphone customers, so they can’t simply rely on the (server) market alone.”\n\nMemory chips are known as “the second-largest component in a handset,” with their cost contribution ranking second only to the AP (application processor, i.e., CPU). Depending on the model and configuration, memory chips generally account for around 10–30% of total cost, and in high-end models with high-capacity configurations such as 12GB + 512GB, the share of memory can exceed 20%. Due to the sharp rise in memory chip prices in the third quarter of 2025, prices of mid- to high-end smartphones have generally increased by about 100–500 yuan, and for example, the price gap between the 12GB + 256GB and 16GB + 512GB configurations of the OPPO Find X9 series has reached as much as 900 yuan.\n\nMeanwhile, this surge in memory chip prices is already affecting the broader smartphone industry. A representative of a domestic smartphone brand told Jiemian News reporters, “Right now, every company is closely watching rivals’ product strategies for next year. Under the backdrop of rising memory prices, it’s still undecided whether to lower specifications or to raise consumer prices, and quite a few projects have been put on hold because memory price hikes have pushed costs up too much.”\n\nOn November 13, Zhao Haijun (赵��军), co-CEO of SMIC (Semiconductor Manufacturing International Corporation, 中芯国际), said at the company’s earnings briefing for the third quarter of 2025 that “because of memory shortages and price increases, industries such as smartphones and network communications equipment have become somewhat more cautious about securing volumes.” He added that in order to handle urgent rush orders related to memory and analog, SMIC postponed shipments of some smartphone-related products in the third quarter.\n\nAccording to supply-chain sources, original manufacturers typically hold about 6–8 weeks of inventory and adjust contract prices each quarter based on supply-demand conditions at the time. As inventories at several major original manufacturers have now begun to run low, there is practically no room left to make large concessions on price.\n\n“We are currently in a phase where all parties are engaged in a tug-of-war,” one source explained. “To maintain customer relationships, the chip original manufacturers are trying to allocate inventory to smartphone makers first, while the smartphone makers are waiting to see who moves first and sets a reference price. Only then can they start further negotiations from that baseline.” The source added, “But in this kind of situation, the only company that truly has the clout to lead price negotiations is Apple.”\n\nHe also said, “Even if the original manufacturers decided to expand capacity now, it would still take at least one to two years to bring production capacity in line with current market demand. We cannot rule out the possibility that there is some froth in today’s hot AI demand, and we will only know the true scale of demand next year. At this point in time, it’s unlikely that original manufacturers represented by Samsung and SK hynix will blindly embark on capacity expansion; instead, they are more likely to opt for production cuts in order to secure higher profits.” Therefore, he predicted, “There is a high probability that memory chip prices will see another sharp rise in the first half of next year.”\n\n---\n\nhttps://t.co/YGRzcik5nO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03681426276361636, "ret_signed_m1d": 0.03681426276361636, "alpha_spy_m1d": 0.036650649899242084, "alpha_c_m1d": 0.03727809173522404, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03245051702635008, "ret_signed_p1d": 0.03245051702635008, "alpha_spy_p1d": 0.04176704095731224, "alpha_c_p1d": 0.04601664925886976, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.08473378270956726, "ret_signed_p1w": -0.08473378270956726, "alpha_spy_p1w": -0.06553535236991385, "alpha_c_p1w": -0.030092919616758904, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": 0.010128097793412469, "ret_signed_p1m": 0.010128097793412469, "alpha_spy_p1m": -0.002968422286115638, "alpha_c_p1m": -0.012829705927157967, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.6448849831741351, "ret_signed_p3m": 0.6448849831741351, "alpha_spy_p3m": 0.6120323239124047, "alpha_c_p3m": 0.4388133226184988, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.182736558314493, "ret_signed_p6m": 2.182736558314493, "alpha_spy_p6m": 2.076208322808437, "alpha_c_p6m": 1.5069779446014122, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.44, -0.4493, -0.4571, -0.4482, -0.4445, -0.4473, -0.4455, -0.4379, -0.4413, -0.4562, -0.4491, -0.4452, -0.435, -0.4226, -0.4184, -0.4019, -0.3822, -0.3635, -0.3348, -0.3289, -0.3074, -0.3186, -0.2863, -0.2946, -0.2838, -0.2714, -0.2774, -0.2822, -0.3031, -0.2823, -0.2796, -0.245, -0.2088, -0.2031, -0.1989, -0.206, -0.1914, -0.1971, -0.1763, -0.1727, -0.1883, -0.1635, -0.1221, -0.1139, -0.0954, -0.1073, -0.1072, -0.1051, -0.0618, -0.0345, -0.049, -0.0172, -0.0024, 0.0036, 0.067, 0.003, 0.0104, 0.015, 0.0023, 0.0478, 0.049, 0.0516, 0.0368, 0.0, 0.0325, -0.0167, -0.0294, -0.0432, -0.0847, -0.0564, -0.0473, -0.0246, -0.0101, -0.0203, -0.0091, 0.0104, 0.0034, -0.0106, 0.0161, 0.0527, 0.0498, 0.0762, 0.0536, 0.0392, 0.0101, -0.0178, 0.0028, 0.0335, 0.0495, 0.098, 0.1033, 0.1184, 0.1184, 0.1426, 0.1907, 0.1963, 0.1865, 0.1865, 0.2722, 0.3122, 0.3751, 0.3866, 0.3704, 0.3884, 0.3917, 0.3706, 0.3758, 0.3968, 0.4535, 0.4597, 0.4361, 0.4817, 0.5117, 0.5209, 0.4882, 0.5804, 0.6487, 0.6554, 0.6551, 0.6042, 0.6841, 0.6314, 0.5678, 0.5795, 0.6199, 0.5974, 0.6449, 0.7037, 0.7048, 0.7048, 0.6887, 0.7173, 0.7511, 0.7988, 0.8003, 0.8526, 0.8872, 0.9685, 0.9363, 0.9367, 0.7456, 0.6415, 0.7582, 0.7002, 0.6228, 0.7519, 0.7903, 0.7501, 0.7511, 0.8284, 0.8708, 0.9725, 0.8957, 0.8594, 0.7451, 0.7765, 0.7613, 0.6576, 0.66, 0.5647, 0.5156, 0.6874, 0.6064, 0.6607, 0.7061, 0.7355, 0.8931, 0.8697, 0.8927, 0.8912, 1.0007, 1.0232, 1.0583, 1.0341, 1.0426, 1.094, 1.1398, 1.1573, 1.1586, 1.2552, 1.224, 1.2528, 1.2278, 1.2617, 1.4455, 1.5317, 1.7749, 1.7985, 1.9427, 2.1827]}
{"unit_id": "thread:1989299736915173748", "ticker_idx": 1, "ts": "2025-11-14T11:49:51+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the four major memory makers—Samsung, SK hynix, Kioxia and Micron—have all cut their NAND flash supply in the second half of this year in a concerted effort to push prices higher", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply-side cuts + AI-driven demand surge point to sharp memory price rise in H1 next year; bullish Samsung, Hynix, Micron and memory sector broadly.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Many smartphone makers temporarily halt memory chip purchases… some have less than three weeks of inventory left\n\nOn November 14, Jiemian News (界面新闻) learned from exclusive sources that as upstream memory semiconductor (storage chip) prices have soared, multiple smartphone makers have already temporarily suspended their memory chip purchases for this quarter. Smartphone makers such as Xiaomi, OPPO and vivo generally have less than two months of inventory, and some manufacturers have less than three weeks of DRAM (dynamic random-access memory) inventory left. They are hesitating over whether to accept the quotes from original manufacturers (Micron, Samsung and SK hynix), which include price hikes of as much as nearly 50%.\n\nAn employee at one original manufacturer, Jackey, told Jiemian News, “We are not in a situation right now where we’re worried about our products not selling. If the smartphone makers don’t want them, we can allocate our capacity to server customers. They’re also short of supply at the moment, and we can sell to them at even higher prices.”\n\nThe surge in data-center demand for memory chips driven by the boom in large AI models is one of the factors behind this round of sharp memory price increases. Server makers are generally willing to pay higher prices in order to secure memory chips. According to information obtained by Jiemian News, for the same memory product, the prices that distributors quote to server customers are typically more than 30% higher than those offered to smartphone makers.\n\nAccording to a report by the Sci-Tech Innovation Board Daily (科创板日报), Samsung Electronics earlier became the first to suspend offering contract prices for DDR5 DRAM in October, and the issuance of new quotes is expected to be postponed until mid-November. This move drove spot prices for DDR5 up by 25% in just one week. At the same time, the four major memory makers—Samsung, SK hynix, Kioxia and Micron—have all cut their NAND flash supply in the second half of this year in a concerted effort to push prices higher.\n\nSamsung’s NAND wafer production target for this year has been lowered to 4.72 million wafers, about 7% down from last year’s level of 5.07 million. Kioxia has likewise reduced its annual output from 4.8 million wafers last year to 4.69 million. SK hynix’s NAND production is falling from 2.01 million wafers last year to around 1.8 million, a cut of 10%. Micron’s largest NAND manufacturing base, its Fab 7 plant in Singapore, is also keeping output at just over 300,000 wafers, maintaining a conservative stance on supply.\n\nThe report also pointed out that as more production lines are converted to 4-bit cell (QLC) processes in response to rising AI demand, an effective decline in NAND flash output will be unavoidable. North American tech giants, worried that NAND prices could spike, have already begun “panic stockpiling (panic buying),” and some suppliers’ NAND allocations for next year are already fully booked.\n\nA week ago, flash memory leader SanDisk sharply raised its November contract prices for NAND flash by 50% in one go, which in turn spurred further price-hike moves by memory giants such as Micron.\n\nJackey explained to Jiemian News that “the typical stockpiling (inventory) cycle for smartphone makers is about three months, and at present, most companies have just over one month of DRAM inventory and around two months of NAND inventory left.” He added, “Right now we’re in a situation of ‘undersupply,’ so if smartphone makers want to build finished sets (final products) properly, in the end they have no choice but to accept the prices that the original manufacturers are currently quoting—it’s only a matter of time. That said, from the perspective of the original manufacturers, they cannot ignore their relationships with smartphone customers, so they can’t simply rely on the (server) market alone.”\n\nMemory chips are known as “the second-largest component in a handset,” with their cost contribution ranking second only to the AP (application processor, i.e., CPU). Depending on the model and configuration, memory chips generally account for around 10–30% of total cost, and in high-end models with high-capacity configurations such as 12GB + 512GB, the share of memory can exceed 20%. Due to the sharp rise in memory chip prices in the third quarter of 2025, prices of mid- to high-end smartphones have generally increased by about 100–500 yuan, and for example, the price gap between the 12GB + 256GB and 16GB + 512GB configurations of the OPPO Find X9 series has reached as much as 900 yuan.\n\nMeanwhile, this surge in memory chip prices is already affecting the broader smartphone industry. A representative of a domestic smartphone brand told Jiemian News reporters, “Right now, every company is closely watching rivals’ product strategies for next year. Under the backdrop of rising memory prices, it’s still undecided whether to lower specifications or to raise consumer prices, and quite a few projects have been put on hold because memory price hikes have pushed costs up too much.”\n\nOn November 13, Zhao Haijun (赵��军), co-CEO of SMIC (Semiconductor Manufacturing International Corporation, 中芯国际), said at the company’s earnings briefing for the third quarter of 2025 that “because of memory shortages and price increases, industries such as smartphones and network communications equipment have become somewhat more cautious about securing volumes.” He added that in order to handle urgent rush orders related to memory and analog, SMIC postponed shipments of some smartphone-related products in the third quarter.\n\nAccording to supply-chain sources, original manufacturers typically hold about 6–8 weeks of inventory and adjust contract prices each quarter based on supply-demand conditions at the time. As inventories at several major original manufacturers have now begun to run low, there is practically no room left to make large concessions on price.\n\n“We are currently in a phase where all parties are engaged in a tug-of-war,” one source explained. “To maintain customer relationships, the chip original manufacturers are trying to allocate inventory to smartphone makers first, while the smartphone makers are waiting to see who moves first and sets a reference price. Only then can they start further negotiations from that baseline.” The source added, “But in this kind of situation, the only company that truly has the clout to lead price negotiations is Apple.”\n\nHe also said, “Even if the original manufacturers decided to expand capacity now, it would still take at least one to two years to bring production capacity in line with current market demand. We cannot rule out the possibility that there is some froth in today’s hot AI demand, and we will only know the true scale of demand next year. At this point in time, it’s unlikely that original manufacturers represented by Samsung and SK hynix will blindly embark on capacity expansion; instead, they are more likely to opt for production cuts in order to secure higher profits.” Therefore, he predicted, “There is a high probability that memory chip prices will see another sharp rise in the first half of next year.”\n\n---\n\nhttps://t.co/YGRzcik5nO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04002758410929108, "ret_signed_m1d": -0.04002758410929108, "alpha_spy_m1d": -0.040191196973665355, "alpha_c_m1d": -0.0395637551376834, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01977073586105582, "ret_signed_p1d": -0.01977073586105582, "alpha_spy_p1d": -0.01045421193009366, "alpha_c_p1d": -0.006204603628536143, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.15986712556863925, "ret_signed_p1w": -0.15986712556863925, "alpha_spy_p1w": -0.14066869522898584, "alpha_c_p1w": -0.10522626247583089, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.03779927511900849, "ret_signed_p1m": -0.03779927511900849, "alpha_spy_p1m": -0.0508957951985366, "alpha_c_p1m": -0.060757078839578926, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.6631113086000686, "ret_signed_p3m": 0.6631113086000686, "alpha_spy_p3m": 0.6302586493383382, "alpha_c_p3m": 0.4570396480444323, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.224833073262897, "ret_signed_p6m": 2.224833073262897, "alpha_spy_p6m": 2.118304837756841, "alpha_c_p6m": 1.549074459549816, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.5058, -0.5254, -0.5312, -0.5235, -0.5286, -0.5283, -0.5232, -0.506, -0.5181, -0.5181, -0.5203, -0.5193, -0.4971, -0.4681, -0.4677, -0.4524, -0.4332, -0.3904, -0.3634, -0.3612, -0.357, -0.3522, -0.3162, -0.3411, -0.3335, -0.3262, -0.3453, -0.365, -0.3632, -0.3364, -0.3225, -0.2625, -0.256, -0.239, -0.2264, -0.2477, -0.2037, -0.2208, -0.2643, -0.219, -0.2422, -0.2224, -0.1795, -0.1801, -0.1623, -0.1804, -0.1959, -0.1625, -0.1127, -0.1083, -0.101, -0.0818, -0.0925, -0.0934, -0.0491, -0.1167, -0.0378, -0.0344, -0.0361, 0.0262, -0.0232, -0.0078, -0.04, 0.0, -0.0198, -0.0743, -0.0847, -0.1842, -0.1599, -0.0928, -0.0903, -0.0671, -0.0671, -0.0419, -0.0258, -0.0297, -0.0513, -0.0818, -0.0389, 0.0004, 0.0226, 0.0684, 0.0471, -0.0231, -0.0378, -0.058, -0.0863, 0.007, 0.0773, 0.1206, 0.1193, 0.1614, 0.1614, 0.1538, 0.1931, 0.186, 0.1568, 0.1568, 0.2784, 0.2651, 0.3919, 0.3762, 0.3254, 0.3987, 0.4018, 0.3704, 0.3511, 0.3644, 0.4702, 0.4702, 0.4793, 0.5771, 0.6114, 0.6198, 0.577, 0.6627, 0.7642, 0.7663, 0.6815, 0.7744, 0.7, 0.5377, 0.5519, 0.5997, 0.5543, 0.5128, 0.6631, 0.6778, 0.6685, 0.6685, 0.6203, 0.7061, 0.6915, 0.7354, 0.7062, 0.6942, 0.7387, 0.6843, 0.6713, 0.6726, 0.5388, 0.6243, 0.6092, 0.5008, 0.5779, 0.6338, 0.697, 0.6429, 0.7271, 0.7906, 0.8712, 0.8714, 0.8006, 0.714, 0.6388, 0.6031, 0.5486, 0.4407, 0.4478, 0.3048, 0.3698, 0.4915, 0.485, 0.485, 0.5317, 0.531, 0.6492, 0.7091, 0.7054, 0.7296, 0.8881, 0.8499, 0.8539, 0.8452, 0.8182, 0.8221, 0.9766, 0.9532, 1.0141, 1.1269, 1.0448, 1.1022, 1.0969, 1.1985, 1.3373, 1.5958, 1.7028, 1.6219, 2.0281, 2.2248]}
{"unit_id": "thread:1989299736915173748", "ticker_idx": 2, "ts": "2025-11-14T11:49:51+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics earlier became the first to suspend offering contract prices for DDR5 DRAM in October", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply-side cuts + AI-driven demand surge point to sharp memory price rise in H1 next year; bullish Samsung, Hynix, Micron and memory sector broadly.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Many smartphone makers temporarily halt memory chip purchases… some have less than three weeks of inventory left\n\nOn November 14, Jiemian News (界面新闻) learned from exclusive sources that as upstream memory semiconductor (storage chip) prices have soared, multiple smartphone makers have already temporarily suspended their memory chip purchases for this quarter. Smartphone makers such as Xiaomi, OPPO and vivo generally have less than two months of inventory, and some manufacturers have less than three weeks of DRAM (dynamic random-access memory) inventory left. They are hesitating over whether to accept the quotes from original manufacturers (Micron, Samsung and SK hynix), which include price hikes of as much as nearly 50%.\n\nAn employee at one original manufacturer, Jackey, told Jiemian News, “We are not in a situation right now where we’re worried about our products not selling. If the smartphone makers don’t want them, we can allocate our capacity to server customers. They’re also short of supply at the moment, and we can sell to them at even higher prices.”\n\nThe surge in data-center demand for memory chips driven by the boom in large AI models is one of the factors behind this round of sharp memory price increases. Server makers are generally willing to pay higher prices in order to secure memory chips. According to information obtained by Jiemian News, for the same memory product, the prices that distributors quote to server customers are typically more than 30% higher than those offered to smartphone makers.\n\nAccording to a report by the Sci-Tech Innovation Board Daily (科创板日报), Samsung Electronics earlier became the first to suspend offering contract prices for DDR5 DRAM in October, and the issuance of new quotes is expected to be postponed until mid-November. This move drove spot prices for DDR5 up by 25% in just one week. At the same time, the four major memory makers—Samsung, SK hynix, Kioxia and Micron—have all cut their NAND flash supply in the second half of this year in a concerted effort to push prices higher.\n\nSamsung’s NAND wafer production target for this year has been lowered to 4.72 million wafers, about 7% down from last year’s level of 5.07 million. Kioxia has likewise reduced its annual output from 4.8 million wafers last year to 4.69 million. SK hynix’s NAND production is falling from 2.01 million wafers last year to around 1.8 million, a cut of 10%. Micron’s largest NAND manufacturing base, its Fab 7 plant in Singapore, is also keeping output at just over 300,000 wafers, maintaining a conservative stance on supply.\n\nThe report also pointed out that as more production lines are converted to 4-bit cell (QLC) processes in response to rising AI demand, an effective decline in NAND flash output will be unavoidable. North American tech giants, worried that NAND prices could spike, have already begun “panic stockpiling (panic buying),” and some suppliers’ NAND allocations for next year are already fully booked.\n\nA week ago, flash memory leader SanDisk sharply raised its November contract prices for NAND flash by 50% in one go, which in turn spurred further price-hike moves by memory giants such as Micron.\n\nJackey explained to Jiemian News that “the typical stockpiling (inventory) cycle for smartphone makers is about three months, and at present, most companies have just over one month of DRAM inventory and around two months of NAND inventory left.” He added, “Right now we’re in a situation of ‘undersupply,’ so if smartphone makers want to build finished sets (final products) properly, in the end they have no choice but to accept the prices that the original manufacturers are currently quoting—it’s only a matter of time. That said, from the perspective of the original manufacturers, they cannot ignore their relationships with smartphone customers, so they can’t simply rely on the (server) market alone.”\n\nMemory chips are known as “the second-largest component in a handset,” with their cost contribution ranking second only to the AP (application processor, i.e., CPU). Depending on the model and configuration, memory chips generally account for around 10–30% of total cost, and in high-end models with high-capacity configurations such as 12GB + 512GB, the share of memory can exceed 20%. Due to the sharp rise in memory chip prices in the third quarter of 2025, prices of mid- to high-end smartphones have generally increased by about 100–500 yuan, and for example, the price gap between the 12GB + 256GB and 16GB + 512GB configurations of the OPPO Find X9 series has reached as much as 900 yuan.\n\nMeanwhile, this surge in memory chip prices is already affecting the broader smartphone industry. A representative of a domestic smartphone brand told Jiemian News reporters, “Right now, every company is closely watching rivals’ product strategies for next year. Under the backdrop of rising memory prices, it’s still undecided whether to lower specifications or to raise consumer prices, and quite a few projects have been put on hold because memory price hikes have pushed costs up too much.”\n\nOn November 13, Zhao Haijun (赵��军), co-CEO of SMIC (Semiconductor Manufacturing International Corporation, 中芯国际), said at the company’s earnings briefing for the third quarter of 2025 that “because of memory shortages and price increases, industries such as smartphones and network communications equipment have become somewhat more cautious about securing volumes.” He added that in order to handle urgent rush orders related to memory and analog, SMIC postponed shipments of some smartphone-related products in the third quarter.\n\nAccording to supply-chain sources, original manufacturers typically hold about 6–8 weeks of inventory and adjust contract prices each quarter based on supply-demand conditions at the time. As inventories at several major original manufacturers have now begun to run low, there is practically no room left to make large concessions on price.\n\n“We are currently in a phase where all parties are engaged in a tug-of-war,” one source explained. “To maintain customer relationships, the chip original manufacturers are trying to allocate inventory to smartphone makers first, while the smartphone makers are waiting to see who moves first and sets a reference price. Only then can they start further negotiations from that baseline.” The source added, “But in this kind of situation, the only company that truly has the clout to lead price negotiations is Apple.”\n\nHe also said, “Even if the original manufacturers decided to expand capacity now, it would still take at least one to two years to bring production capacity in line with current market demand. We cannot rule out the possibility that there is some froth in today’s hot AI demand, and we will only know the true scale of demand next year. At this point in time, it’s unlikely that original manufacturers represented by Samsung and SK hynix will blindly embark on capacity expansion; instead, they are more likely to opt for production cuts in order to secure higher profits.” Therefore, he predicted, “There is a high probability that memory chip prices will see another sharp rise in the first half of next year.”\n\n---\n\nhttps://t.co/YGRzcik5nO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05761317038957614, "ret_signed_m1d": 0.05761317038957614, "alpha_spy_m1d": 0.057449557525201866, "alpha_c_m1d": 0.06302709370298476, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03497941620794953, "ret_signed_p1d": 0.03497941620794953, "alpha_spy_p1d": 0.044295940138911694, "alpha_c_p1d": 0.05063102623894755, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.024691404986047938, "ret_signed_p1w": -0.024691404986047938, "alpha_spy_p1w": -0.005492974646394533, "alpha_c_p1w": 0.027191265659056163, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.07818927376678131, "ret_signed_p1m": 0.07818927376678131, "alpha_spy_p1m": 0.0650927536872532, "alpha_c_p1m": 0.09050924685115647, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": 0.7347294453035076, "ret_signed_p3m": 0.7347294453035076, "alpha_spy_p3m": 0.7018767860417772, "alpha_c_p3m": 0.7408938922948921, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 1.957630231150651, "ret_signed_p6m": 1.957630231150651, "alpha_spy_p6m": 1.8511019956445949, "alpha_c_p6m": 1.7195854032114168, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.283, -0.2779, -0.2769, -0.2687, -0.2677, -0.28, -0.2769, -0.2871, -0.2861, -0.3076, -0.2923, -0.2851, -0.282, -0.2882, -0.282, -0.2677, -0.2564, -0.2482, -0.2277, -0.2165, -0.1868, -0.199, -0.1775, -0.1775, -0.1448, -0.1325, -0.1253, -0.1181, -0.1468, -0.1337, -0.1368, -0.1152, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0288, -0.0401, -0.0576, -0.0226, 0.0051, 0.0072, 0.0093, 0.0031, 0.0144, -0.0072, 0.0165, 0.0494, 0.0237, 0.034, 0.071, 0.106, 0.143, 0.0792, 0.035, 0.0206, 0.0072, 0.035, 0.0648, 0.0607, 0.0576, 0.0, 0.035, 0.0062, -0.0072, 0.035, -0.0247, -0.0051, 0.0216, 0.0576, 0.0648, 0.034, 0.037, 0.0638, 0.0751, 0.0813, 0.1152, 0.1265, 0.1152, 0.1111, 0.1039, 0.1204, 0.0782, 0.0576, 0.1101, 0.107, 0.0936, 0.1368, 0.1471, 0.143, 0.143, 0.2037, 0.2354, 0.2395, 0.2395, 0.2395, 0.3284, 0.4277, 0.436, 0.4577, 0.4349, 0.437, 0.4349, 0.4225, 0.4504, 0.4876, 0.5393, 0.5435, 0.5011, 0.5455, 0.5745, 0.5724, 0.5724, 0.6489, 0.6789, 0.6613, 0.6593, 0.5548, 0.7316, 0.7482, 0.6469, 0.6396, 0.7203, 0.7141, 0.7347, 0.8464, 0.8733, 0.8733, 0.8733, 0.8733, 0.9642, 0.9653, 0.9952, 1.0676, 1.1038, 1.2537, 1.2382, 1.2382, 1.017, 0.7802, 0.9808, 0.9456, 0.7937, 0.9425, 0.9642, 0.9425, 0.897, 0.9508, 1.0046, 1.1555, 1.0728, 1.0614, 0.926, 0.9611, 0.9539, 0.8619, 0.8619, 0.8264, 0.7321, 0.9647, 0.8481, 0.9289, 1.0004, 1.0356, 1.1807, 1.1133, 1.1341, 1.0823, 1.1392, 1.1858, 1.2532, 1.2376, 1.2221, 1.2687, 1.2532, 1.3257, 1.2739, 1.3257, 1.2998, 1.3412, 1.2843, 1.2843, 1.4086, 1.4086, 1.7556, 1.8126, 1.7815, 1.9576]}
{"unit_id": "thread:1989299736915173748", "ticker_idx": 3, "ts": "2025-11-14T11:49:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix's NAND production is falling from 2.01 million wafers last year to around 1.8 million, a cut of 10%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply-side cuts + AI-driven demand surge point to sharp memory price rise in H1 next year; bullish Samsung, Hynix, Micron and memory sector broadly.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Many smartphone makers temporarily halt memory chip purchases… some have less than three weeks of inventory left\n\nOn November 14, Jiemian News (界面新闻) learned from exclusive sources that as upstream memory semiconductor (storage chip) prices have soared, multiple smartphone makers have already temporarily suspended their memory chip purchases for this quarter. Smartphone makers such as Xiaomi, OPPO and vivo generally have less than two months of inventory, and some manufacturers have less than three weeks of DRAM (dynamic random-access memory) inventory left. They are hesitating over whether to accept the quotes from original manufacturers (Micron, Samsung and SK hynix), which include price hikes of as much as nearly 50%.\n\nAn employee at one original manufacturer, Jackey, told Jiemian News, “We are not in a situation right now where we’re worried about our products not selling. If the smartphone makers don’t want them, we can allocate our capacity to server customers. They’re also short of supply at the moment, and we can sell to them at even higher prices.”\n\nThe surge in data-center demand for memory chips driven by the boom in large AI models is one of the factors behind this round of sharp memory price increases. Server makers are generally willing to pay higher prices in order to secure memory chips. According to information obtained by Jiemian News, for the same memory product, the prices that distributors quote to server customers are typically more than 30% higher than those offered to smartphone makers.\n\nAccording to a report by the Sci-Tech Innovation Board Daily (科创板日报), Samsung Electronics earlier became the first to suspend offering contract prices for DDR5 DRAM in October, and the issuance of new quotes is expected to be postponed until mid-November. This move drove spot prices for DDR5 up by 25% in just one week. At the same time, the four major memory makers—Samsung, SK hynix, Kioxia and Micron—have all cut their NAND flash supply in the second half of this year in a concerted effort to push prices higher.\n\nSamsung’s NAND wafer production target for this year has been lowered to 4.72 million wafers, about 7% down from last year’s level of 5.07 million. Kioxia has likewise reduced its annual output from 4.8 million wafers last year to 4.69 million. SK hynix’s NAND production is falling from 2.01 million wafers last year to around 1.8 million, a cut of 10%. Micron’s largest NAND manufacturing base, its Fab 7 plant in Singapore, is also keeping output at just over 300,000 wafers, maintaining a conservative stance on supply.\n\nThe report also pointed out that as more production lines are converted to 4-bit cell (QLC) processes in response to rising AI demand, an effective decline in NAND flash output will be unavoidable. North American tech giants, worried that NAND prices could spike, have already begun “panic stockpiling (panic buying),” and some suppliers’ NAND allocations for next year are already fully booked.\n\nA week ago, flash memory leader SanDisk sharply raised its November contract prices for NAND flash by 50% in one go, which in turn spurred further price-hike moves by memory giants such as Micron.\n\nJackey explained to Jiemian News that “the typical stockpiling (inventory) cycle for smartphone makers is about three months, and at present, most companies have just over one month of DRAM inventory and around two months of NAND inventory left.” He added, “Right now we’re in a situation of ‘undersupply,’ so if smartphone makers want to build finished sets (final products) properly, in the end they have no choice but to accept the prices that the original manufacturers are currently quoting—it’s only a matter of time. That said, from the perspective of the original manufacturers, they cannot ignore their relationships with smartphone customers, so they can’t simply rely on the (server) market alone.”\n\nMemory chips are known as “the second-largest component in a handset,” with their cost contribution ranking second only to the AP (application processor, i.e., CPU). Depending on the model and configuration, memory chips generally account for around 10–30% of total cost, and in high-end models with high-capacity configurations such as 12GB + 512GB, the share of memory can exceed 20%. Due to the sharp rise in memory chip prices in the third quarter of 2025, prices of mid- to high-end smartphones have generally increased by about 100–500 yuan, and for example, the price gap between the 12GB + 256GB and 16GB + 512GB configurations of the OPPO Find X9 series has reached as much as 900 yuan.\n\nMeanwhile, this surge in memory chip prices is already affecting the broader smartphone industry. A representative of a domestic smartphone brand told Jiemian News reporters, “Right now, every company is closely watching rivals’ product strategies for next year. Under the backdrop of rising memory prices, it’s still undecided whether to lower specifications or to raise consumer prices, and quite a few projects have been put on hold because memory price hikes have pushed costs up too much.”\n\nOn November 13, Zhao Haijun (赵��军), co-CEO of SMIC (Semiconductor Manufacturing International Corporation, 中芯国际), said at the company’s earnings briefing for the third quarter of 2025 that “because of memory shortages and price increases, industries such as smartphones and network communications equipment have become somewhat more cautious about securing volumes.” He added that in order to handle urgent rush orders related to memory and analog, SMIC postponed shipments of some smartphone-related products in the third quarter.\n\nAccording to supply-chain sources, original manufacturers typically hold about 6–8 weeks of inventory and adjust contract prices each quarter based on supply-demand conditions at the time. As inventories at several major original manufacturers have now begun to run low, there is practically no room left to make large concessions on price.\n\n“We are currently in a phase where all parties are engaged in a tug-of-war,” one source explained. “To maintain customer relationships, the chip original manufacturers are trying to allocate inventory to smartphone makers first, while the smartphone makers are waiting to see who moves first and sets a reference price. Only then can they start further negotiations from that baseline.” The source added, “But in this kind of situation, the only company that truly has the clout to lead price negotiations is Apple.”\n\nHe also said, “Even if the original manufacturers decided to expand capacity now, it would still take at least one to two years to bring production capacity in line with current market demand. We cannot rule out the possibility that there is some froth in today’s hot AI demand, and we will only know the true scale of demand next year. At this point in time, it’s unlikely that original manufacturers represented by Samsung and SK hynix will blindly embark on capacity expansion; instead, they are more likely to opt for production cuts in order to secure higher profits.” Therefore, he predicted, “There is a high probability that memory chip prices will see another sharp rise in the first half of next year.”\n\n---\n\nhttps://t.co/YGRzcik5nO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09285720201056402, "ret_signed_m1d": 0.09285720201056402, "alpha_spy_m1d": 0.09269358914618975, "alpha_c_m1d": 0.0933210309821717, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08214287073215654, "ret_signed_p1d": 0.08214287073215654, "alpha_spy_p1d": 0.0914593946631187, "alpha_c_p1d": 0.09570900296467622, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.06964281757401458, "ret_signed_p1w": -0.06964281757401458, "alpha_spy_p1w": -0.05044438723436118, "alpha_c_p1w": -0.015001954481206226, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.010005705267535414, "ret_signed_p1m": -0.010005705267535414, "alpha_spy_p1m": -0.02310222534706352, "alpha_c_p1m": -0.03296350898810585, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.536814195618829, "ret_signed_p3m": 0.536814195618829, "alpha_spy_p3m": 0.5039615363570986, "alpha_c_p3m": 0.33074253506319273, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.3657463705299318, "ret_signed_p6m": 2.3657463705299318, "alpha_spy_p6m": 2.2592181350238754, "alpha_c_p6m": 1.6899877568168509, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.531, -0.5444, -0.5631, -0.5524, -0.5373, -0.5337, -0.5364, -0.5205, -0.5196, -0.5429, -0.5348, -0.5312, -0.5259, -0.5116, -0.5054, -0.4857, -0.4571, -0.4518, -0.4134, -0.4089, -0.3786, -0.4045, -0.3652, -0.3652, -0.3732, -0.3554, -0.3616, -0.3634, -0.3991, -0.3768, -0.3795, -0.3571, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2357, -0.2589, -0.2652, -0.2455, -0.192, -0.1687, -0.133, -0.1446, -0.1402, -0.1455, -0.0893, -0.0446, -0.0696, -0.0036, 0.0143, -0.0018, 0.1071, 0.0464, 0.0339, 0.0589, 0.0357, 0.0821, 0.1054, 0.1018, 0.0929, 0.0, 0.0821, 0.0179, 0.0036, 0.0196, -0.0696, -0.0714, -0.0732, -0.0643, -0.0279, -0.0529, -0.0386, -0.0029, -0.0136, -0.0314, -0.0279, 0.0311, 0.0114, 0.049, 0.0097, 0.0204, -0.01, -0.0529, -0.0154, -0.0136, -0.0225, 0.0365, 0.0436, 0.0508, 0.0508, 0.0704, 0.1437, 0.1633, 0.1633, 0.1633, 0.2098, 0.2437, 0.2974, 0.3259, 0.351, 0.3295, 0.3385, 0.3188, 0.3259, 0.3385, 0.351, 0.3653, 0.3277, 0.3224, 0.3492, 0.3706, 0.3152, 0.4296, 0.5029, 0.5386, 0.6244, 0.4832, 0.6208, 0.6083, 0.5046, 0.4993, 0.5851, 0.5654, 0.5368, 0.5868, 0.5726, 0.5726, 0.5726, 0.5726, 0.5976, 0.6959, 0.6994, 0.7959, 0.8192, 0.9675, 0.8995, 0.8995, 0.6811, 0.52, 0.6847, 0.6542, 0.4967, 0.6793, 0.7097, 0.665, 0.6292, 0.7437, 0.7366, 0.8905, 0.8136, 0.8028, 0.6703, 0.7652, 0.7813, 0.6703, 0.6703, 0.5629, 0.4448, 0.6059, 0.4859, 0.5683, 0.5862, 0.6399, 0.8494, 0.7867, 0.8386, 0.8619, 0.9747, 1.0338, 1.0678, 1.0194, 1.0875, 1.1913, 1.1895, 1.1931, 1.1877, 1.3131, 1.3274, 1.3148, 1.3023, 1.3023, 1.5906, 1.5906, 1.8663, 1.9611, 2.0184, 2.3657]}
{"unit_id": "thread:1985629145733226869", "ticker_idx": 1, "ts": "2025-11-04T08:44:14+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "there is a high possibility that the timing of HBM4 mass shipments will be postponed to 2027", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Hanmi Semi faces Sell/Underperform from UBS/Macquarie on Micron HBM4 delays and near-zero Samsung supply chance; Micron HBM4 ramp pushed to 2027.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Despite Hanmi Semiconductor’s courting… “Samsung Electronics is unlikely to accept delivery of TC bonders.”\n\nU.S. Micron, the world’s No. 3 player in the high-bandwidth memory (HBM) market after SK hynix and Samsung Electronics, is increasingly expected to face difficulty supplying its next-generation HBM4 to “major HBM buyer” Nvidia. Because it has begun to redesign HBM4 to meet Nvidia’s requirements, the view is that it will start deliveries only in 2027—about a year later than competitors.\n\nAs Micron’s HBM4 development falls behind, Hanmi Semiconductor, which supplies “thermo-compression (TC) bonders” (DRAM bonding equipment) needed for HBM manufacturing, is likely to take a hit as well. Micron accounts for a high share of its revenue, and the chances of breaking into Samsung Electronics as a new customer are almost nil. SK hynix began awarding work to Hanwha Semitec this year. Taking these factors into account, global investment banks (IBs) have downgraded Hanmi Semiconductor’s rating to “Sell.”\n\nAccording to the semiconductor industry on the 4th, Hong Kong’s GF Securities said in an analysis report on the 3rd that “Micron is having difficulty meeting the data-processing speed of ‘10 gigabits per second (Gb)’ presented by Nvidia,” adding, “Because there are also yield issues, there is a high possibility that the timing of HBM4 mass shipments will be postponed to 2027.” Although Micron dismissed market concerns at the earnings briefing held in September, the aftershocks continue. By contrast, competitors SK hynix and Samsung Electronics have set up mass-production systems for HBM4.\n\nIt is also a negative for Micron that Nvidia, in a press release on the 31st of last month, officially referred to Samsung Electronics as a “core partner of the HBM4 supply chain.” If Samsung Electronics’ share of HBM4 deliveries increases, Micron’s position is more likely to narrow than SK hynix’s. SK hynix, which has firmly established itself as Nvidia’s No. 1 supplier for HBM3E, is expected by the industry to occupy the same position in HBM4.\n\nSparks fly to Hanmi… Global IBs recommend ‘Sell’\n\nThe spillover from Micron’s HBM4 slump is hitting Hanmi Semiconductor, a producer of TC bonders. This year Hanmi Semiconductor has increased Micron’s share of sales in place of SK hynix. In the first half, about 82% (269.9 billion won) of Hanmi Semiconductor’s sales (327.4 billion won) came from exports. As recently as last year, domestic sales (58.7%) were larger than overseas (41.3%), but with Micron exports increasing, the balance flipped in just one year.\n\nIn this situation, as HBM4 equipment orders from customers such as Micron were delayed, third-quarter results registered an “earnings shock.” Operating profit was 67.822 billion won, 38.7% below the consensus (the average of securities-firm forecasts) of 110.6 billion won.\n\nGlobal IBs are rushing to issue negative opinions on Hanmi Semiconductor. On the 3rd, UBS said “the TC bonder outlook needs to be reset,” and downgraded its rating from Neutral to “Sell.” Macquarie also lowered its recommendation to “Underperform,” saying “the current share price already reflects rosy expectations (of supplying Samsung Electronics).” J.P. Morgan cut its 2026 EPS forecast by 15.8%, from 5,331 won to 4,488 won.\n\nLow likelihood of a Samsung Electronics supply deal\n\nTo diversify its customer base, Hanmi Semiconductor is pursuing a plan to supply TC bonders to Samsung Electronics. The fact that the share price has risen 45.7% compared with late September is analyzed as reflecting expectations of a deal.\n\nHanmi Semiconductor Chairman Kwak Dong-shin even personally welcomed Song Jae-hyuk, President and Chief Technology Officer (CTO) of Samsung Electronics’ Device Solutions (DS) Division, during a booth tour at the “Semiconductor Grand Expo” held at COEX in Samseong-dong, Seoul, on the 22nd of last month, underscoring the effort being made.\n\nHowever, the semiconductor industry assesses the likelihood of a deal as low. This is because Samsung Electronics is placing emphasis on strengthening cooperation with companies already in its TC bonder supply chain, such as SEMES.\n\nAn industry source well-versed in Samsung Electronics said, “You can assume there is almost no possibility that Samsung Electronics will order Hanmi Semiconductor’s TC bonder,” adding, “There will be no change in the existing TC bonder procurement policy.” Global IB J.P. Morgan likewise wrote in an analysis report that “it is uncertain whether Hanmi Semiconductor will receive TC bonder orders from Samsung Electronics,” cautioning against optimism because “Samsung Electronics prefers SEMES.”\n\n---\n\nhttps://t.co/MSybSwG8bp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07645741282895635, "ret_signed_m1d": -0.07645741282895635, "alpha_spy_m1d": -0.06446168100013705, "alpha_c_m1d": -0.038641351158183346, "bench_spy_m1d": 0.0119957318288193, "bench_c_m1d": 0.037816061670773005, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08929970527516695, "ret_signed_p1d": -0.08929970527516695, "alpha_spy_p1d": -0.08583431607817071, "alpha_c_p1d": -0.07019325182364611, "bench_spy_p1d": 0.003465389196996238, "bench_c_p1d": 0.019106453451520844, "ret_p1w": 0.10585698189947701, "ret_signed_p1w": -0.10585698189947701, "alpha_spy_p1w": -0.09436486372589648, "alpha_c_p1w": -0.1107045162695891, "bench_spy_p1w": 0.011492118173580534, "bench_c_p1w": -0.004847534370112094, "ret_p1m": 0.073980710015223, "ret_signed_p1m": -0.073980710015223, "alpha_spy_p1m": -0.061170469458797117, "alpha_c_p1m": -0.04106885009629635, "bench_spy_p1m": 0.012810240556425878, "bench_c_p1m": 0.032911859918926645, "ret_p3m": 0.9036261422706069, "ret_signed_p3m": -0.9036261422706069, "alpha_spy_p3m": -0.8758217596318605, "alpha_c_p3m": -0.7563436252599787, "bench_spy_p3m": 0.027804382638746405, "bench_c_p3m": 0.14728251701062822, "ret_p6m": 1.379889789526489, "ret_signed_p6m": -1.379889789526489, "alpha_spy_p6m": -1.3200714867881398, "alpha_c_p6m": -0.9592795764154793, "bench_spy_p6m": 0.059818302738349205, "bench_c_p6m": 0.42061021311100966, "price_path": [-0.4872, -0.455, -0.4329, -0.4144, -0.4304, -0.4257, -0.446, -0.4337, -0.4406, -0.4627, -0.4693, -0.4606, -0.4664, -0.466, -0.4603, -0.4408, -0.4545, -0.4545, -0.4569, -0.4558, -0.4307, -0.3978, -0.3974, -0.3801, -0.3583, -0.3098, -0.2793, -0.2768, -0.272, -0.2667, -0.2259, -0.2541, -0.2454, -0.2372, -0.2588, -0.2811, -0.2791, -0.2487, -0.2331, -0.1651, -0.1578, -0.1385, -0.1242, -0.1483, -0.0986, -0.1179, -0.1671, -0.1159, -0.142, -0.1197, -0.0711, -0.0718, -0.0516, -0.0722, -0.0897, -0.0519, 0.0045, 0.0095, 0.0178, 0.0394, 0.0274, 0.0263, 0.0765, 0.0, 0.0893, 0.0931, 0.0912, 0.1618, 0.1059, 0.1232, 0.0868, 0.1321, 0.1097, 0.048, 0.0362, -0.0764, -0.0489, 0.0271, 0.0298, 0.0561, 0.0561, 0.0846, 0.1029, 0.0984, 0.074, 0.0395, 0.088, 0.1325, 0.1577, 0.2095, 0.1854, 0.106, 0.0893, 0.0664, 0.0344, 0.14, 0.2196, 0.2686, 0.2671, 0.3149, 0.3149, 0.3062, 0.3507, 0.3427, 0.3096, 0.3096, 0.4473, 0.4323, 0.5758, 0.558, 0.5005, 0.5834, 0.587, 0.5515, 0.5295, 0.5446, 0.6644, 0.6644, 0.6748, 0.7854, 0.8242, 0.8337, 0.7853, 0.8823, 0.9972, 0.9996, 0.9036, 1.0088, 0.9245, 0.7408, 0.7568, 0.811, 0.7596, 0.7126, 0.8828, 0.8995, 0.8889, 0.8889, 0.8343, 0.9315, 0.915, 0.9646, 0.9316, 0.918, 0.9684, 0.9067, 0.8921, 0.8935, 0.7421, 0.8389, 0.8218, 0.6991, 0.7863, 0.8496, 0.9211, 0.8599, 0.9552, 1.0271, 1.1184, 1.1186, 1.0385, 0.9404, 0.8553, 0.8148, 0.7532, 0.631, 0.6391, 0.4772, 0.5508, 0.6885, 0.6812, 0.6812, 0.734, 0.7332, 0.867, 0.9349, 0.9306, 0.958, 1.1375, 1.0942, 1.0988, 1.0889, 1.0584, 1.0628, 1.2377, 1.2112, 1.2801, 1.4079, 1.3148, 1.3799]}
{"unit_id": "orig:1984863737178964446", "ticker_idx": 0, "ts": "2025-11-02T06:02:46+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley projects AWS's revenue growth over the next two years at 23% and 25%, respectively, and expects the explosion in AI demand to generate an additional $6 billion in revenue in 2026", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley bullish AWS/Amazon thesis: 23%/25% revenue growth, $6B AI upside in 2026, capacity expansion laying groundwork for 2026–2027 acceleration.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: AWS is suffering from compute bottlenecks, and AWS is now entering an “AI growth acceleration cycle.”\n\nMorgan Stanley’s analyst summarized four key growth drivers.\n① Rapid expansion of capacity:\nBy year-end, it will secure an additional 1 GW of compute capacity, and the number of Trainium2 chips will double.\n\n② Structural expansion cycle:\nOver the next 24 months, AWS plans to add 10 GW of data center capacity.\n\n③ Surge in AI orders:\nIn October alone, new order volume exceeded the entire third quarter, and the new business signed in a single month amounts to about $18 billion.\n\n④ Acceleration of innovation:\nThere is a high likelihood that Trainium3 will be unveiled by year-end, and the Bedrock software platform is also continuing to expand.\n\nThe report notes that AWS remains “capacity constrained” at present, and that “the supply shortage itself is becoming the core of the growth logic.”\n\nAccording to Morgan Stanley’s analysis, AWS appears to have signed approximately $18 billion in new business in October alone.\nThis exceeds the total new business contract volume for the entire third quarter.\n\n“If there were no compute bottlenecks, AWS’s growth would have been faster.”\n“These expansion plans are laying the groundwork for another acceleration in 2026–2027,” the analyst stated.\n\nMorgan Stanley projects AWS’s revenue growth over the next two years at 23% and 25%, respectively, and expects the explosion in AI demand to generate an additional $6 billion in revenue in 2026.\nThis is expected to lift the overall growth rate by about 4 percentage points.\n\nAdditionally, Morgan Stanley raised Amazon’s 2026/2027 capital expenditure (Capex) outlook by 13%/19%, projecting total capex of $169 billion/$202 billion. This implies Amazon would invest $140 billion/$170 billion in technology and infrastructure, respectively—figures higher than those of other big tech companies such as Microsoft, Meta, and Google.\n\n$AMZN $META $GOOG $MSFT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03850393220195625, "ret_signed_m1d": -0.03850393220195625, "alpha_spy_m1d": -0.036630760471192514, "alpha_c_m1d": -0.0292132783959953, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.009290653805960947, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018425168014886806, "ret_signed_p1d": -0.018425168014886806, "alpha_spy_p1d": -0.006571628072439273, "alpha_c_p1d": -0.0018259761948624664, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.01659919182002434, "ret_p1w": -0.022047268124077313, "ret_signed_p1w": -0.022047268124077313, "alpha_spy_p1w": -0.01926683918904004, "alpha_c_p1w": -0.008751455022655064, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.01329581310142225, "ret_p1m": -0.07708662138210509, "ret_signed_p1m": -0.07708662138210509, "alpha_spy_p1m": -0.0744378794828564, "alpha_c_p1m": -0.05338530132107411, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.023701320061030984, "ret_p3m": -0.048307103434885534, "ret_signed_p3m": -0.048307103434885534, "alpha_spy_p3m": -0.0669665994759655, "alpha_c_p3m": -0.05111637650305967, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.002809273068174134, "ret_p6m": 0.022440992941067828, "ret_signed_p6m": 0.022440992941067828, "alpha_spy_p6m": -0.02497658027166927, "alpha_c_p6m": 0.05227489095452398, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": -0.02983389801345615, "price_path": [-0.1248, -0.1215, -0.1233, -0.1287, -0.1281, -0.1159, -0.0906, -0.0904, -0.0886, -0.1023, -0.1189, -0.1262, -0.0991, -0.1026, -0.0996, -0.098, -0.0882, -0.0984, -0.0984, -0.1128, -0.1103, -0.0721, -0.0853, -0.0715, -0.062, -0.0932, -0.0947, -0.1018, -0.0889, -0.0785, -0.0881, -0.0896, -0.0887, -0.1038, -0.1311, -0.133, -0.1411, -0.1347, -0.1253, -0.1356, -0.1314, -0.1244, -0.1358, -0.1303, -0.1269, -0.1133, -0.1034, -0.1481, -0.1336, -0.1481, -0.1513, -0.1556, -0.1613, -0.1477, -0.1259, -0.1419, -0.1296, -0.1173, -0.1064, -0.0974, -0.0933, -0.1226, -0.0385, 0.0, -0.0184, -0.015, -0.0431, -0.0378, -0.022, -0.0193, -0.0386, -0.0646, -0.076, -0.0832, -0.1238, -0.1233, -0.1451, -0.1311, -0.1091, -0.0958, -0.0978, -0.0978, -0.0818, -0.0792, -0.0771, -0.0851, -0.098, -0.0963, -0.1067, -0.1027, -0.0875, -0.0934, -0.1095, -0.1239, -0.1238, -0.1289, -0.1072, -0.1049, -0.1007, -0.0861, -0.0851, -0.0851, -0.0846, -0.0863, -0.0845, -0.0913, -0.0913, -0.1083, -0.0824, -0.0515, -0.049, -0.0304, -0.0261, -0.0296, -0.0449, -0.0683, -0.0623, -0.0586, -0.0586, -0.0906, -0.0893, -0.0774, -0.0584, -0.0613, -0.0367, -0.0433, -0.0483, -0.0579, -0.0435, -0.0606, -0.0827, -0.1233, -0.172, -0.1783, -0.1852, -0.1965, -0.2142, -0.2174, -0.2174, -0.2081, -0.1937, -0.1935, -0.1728, -0.1919, -0.1789, -0.1707, -0.1814, -0.1732, -0.1796, -0.1782, -0.1464, -0.138, -0.1606, -0.1595, -0.1562, -0.1628, -0.1751, -0.1824, -0.1664, -0.1528, -0.1737, -0.1781, -0.1915, -0.1727, -0.1841, -0.1665, -0.1829, -0.2152, -0.2089, -0.18, -0.171, -0.1741, -0.1741, -0.1622, -0.1584, -0.1289, -0.0801, -0.0615, -0.0556, -0.0196, -0.0217, -0.0169, -0.0135, -0.0225, -0.0161, 0.0054, 0.0043, 0.0393, 0.028, 0.0224]}
{"unit_id": "orig:2003320154105741402", "ticker_idx": 0, "ts": "2025-12-23T04:21:59+00:00", "symbol": "Microchip Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices, with Microchip as our top pick.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi bullish on analog semis into 2026 upturn; top pick Microchip, also positive on TXN, NXPI, ADI on low inventories and constrained supply.", "resolved_tickers": ["MCHP"], "resolution_method": "exchange_ticker", "resolution_reason": "Microchip Technology MCHP US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi noted in its latest report that the biggest surprise within semiconductors in 2026 could come from analog semiconductors.\n\n“Given currently very low inventories, constrained supply growth, and depressed margins, we expect an analog upturn in ’26. We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices, with Microchip as our top pick.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01361896590183509, "ret_signed_m1d": 0.01361896590183509, "alpha_spy_m1d": 0.018168626701145607, "alpha_c_m1d": 0.02314642327651162, "bench_spy_m1d": -0.004549660799310518, "bench_c_m1d": -0.009527457374676529, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0001529584356694258, "ret_signed_p1d": 0.0001529584356694258, "alpha_spy_p1d": -0.0033646618698630615, "alpha_c_p1d": -0.0026282117247007353, "bench_spy_p1d": 0.0035176203055324873, "bench_c_p1d": 0.002781170160370161, "ret_p1w": -0.010252523047035944, "ret_signed_p1w": -0.010252523047035944, "alpha_spy_p1w": -0.00887166514302562, "alpha_c_p1w": -0.010665547320006019, "bench_spy_p1w": -0.0013808579040103242, "bench_c_p1w": 0.00041302427297007505, "ret_p1m": 0.16602906446324694, "ret_signed_p1m": 0.16602906446324694, "alpha_spy_p1m": 0.16975022831241904, "alpha_c_p1m": 0.05927174366052945, "bench_spy_p1m": -0.0037211638491720933, "bench_c_p1m": 0.10675732080271749, "ret_p3m": -0.030745648769192058, "ret_signed_p3m": -0.030745648769192058, "alpha_spy_p3m": 0.02393595059199438, "alpha_c_p3m": -0.09016845512080784, "bench_spy_p3m": -0.05468159936118644, "bench_c_p3m": 0.05942280635161579, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0167, -0.023, -0.0284, -0.0261, -0.0277, 0.0029, 0.0091, 0.0099, -0.0149, 0.0149, -0.0012, -0.0839, -0.0235, -0.0203, -0.0111, -0.0089, -0.0121, 0.0171, 0.024, -0.0218, -0.0129, -0.042, -0.0211, -0.0349, -0.0516, -0.0587, -0.0534, -0.0535, -0.0977, -0.0779, -0.0999, -0.1465, -0.1597, -0.1703, -0.1563, -0.1688, -0.189, -0.2159, -0.2285, -0.2296, -0.2566, -0.2281, -0.2158, -0.2069, -0.1956, -0.1956, -0.1801, -0.1824, -0.1322, -0.0266, -0.0096, 0.007, 0.0306, 0.023, 0.039, 0.0572, 0.028, 0.028, 0.0084, -0.0208, -0.0197, -0.0067, 0.0136, 0.0, 0.0002, 0.0002, -0.0063, -0.0107, -0.0103, -0.0249, -0.0249, -0.0049, 0.0262, 0.1457, 0.1314, 0.1252, 0.151, 0.123, 0.1334, 0.1428, 0.1393, 0.1431, 0.1431, 0.1197, 0.166, 0.1549, 0.1432, 0.1445, 0.1501, 0.2285, 0.2144, 0.1617, 0.1948, 0.1731, 0.1971, 0.1942, 0.1631, 0.1386, 0.1761, 0.2357, 0.2077, 0.2021, 0.2021, 0.208, 0.2106, 0.1807, 0.1894, 0.1791, 0.1687, 0.1617, 0.154, 0.1489, 0.1438, 0.0989, 0.0759, 0.0438, -0.003, 0.0005, 0.0056, 0.0127, -0.0344, -0.0466, -0.0175, -0.0058, -0.004, -0.0258, -0.0307, -0.0097, 0.0102, 0.003, -0.0118, -0.0457, -0.0755, -0.0055, 0.0063, 0.0097, 0.0097, 0.0347, 0.0391, 0.0887, 0.0962, 0.1015, 0.1321, 0.1467, 0.1466, 0.1832, 0.2123, 0.2374, 0.2457, 0.2696, 0.3952, 0.3767, 0.3367, 0.297, 0.3879, 0.4301, 0.4461, 0.4669, 0.5158, 0.5842, 0.5636, 0.5252, 0.5243, 0.5038, 0.4886, 0.4937, 0.4446, 0.4278, 0.4132, 0.4472, 0.4024, 0.4453, 0.4453, 0.5168, 0.4982, 0.4857, 0.4642, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2003320154105741402", "ticker_idx": 1, "ts": "2025-12-23T04:21:59+00:00", "symbol": "TXN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi bullish on analog semis into 2026 upturn; top pick Microchip, also positive on TXN, NXPI, ADI on low inventories and constrained supply.", "resolved_tickers": ["TXN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi noted in its latest report that the biggest surprise within semiconductors in 2026 could come from analog semiconductors.\n\n“Given currently very low inventories, constrained supply growth, and depressed margins, we expect an analog upturn in ’26. We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices, with Microchip as our top pick.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009826031643971334, "ret_signed_m1d": 0.009826031643971334, "alpha_spy_m1d": 0.014375692443281851, "alpha_c_m1d": 0.019353489018647863, "bench_spy_m1d": -0.004549660799310518, "bench_c_m1d": -0.009527457374676529, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00028235021730682597, "ret_signed_p1d": 0.00028235021730682597, "alpha_spy_p1d": -0.0032352700882256613, "alpha_c_p1d": -0.002498819943063335, "bench_spy_p1d": 0.0035176203055324873, "bench_c_p1d": 0.002781170160370161, "ret_p1w": -0.009374306165093937, "ret_signed_p1w": -0.009374306165093937, "alpha_spy_p1w": -0.007993448261083613, "alpha_c_p1w": -0.009787330438064012, "bench_spy_p1w": -0.0013808579040103242, "bench_c_p1w": 0.00041302427297007505, "ret_p1m": 0.09786533995480484, "ret_signed_p1m": 0.09786533995480484, "alpha_spy_p1m": 0.10158650380397694, "alpha_c_p1m": -0.00889198084791265, "bench_spy_p1m": -0.0037211638491720933, "bench_c_p1m": 0.10675732080271749, "ret_p3m": 0.06399272845763293, "ret_signed_p3m": 0.06399272845763293, "alpha_spy_p3m": 0.11867432781881937, "alpha_c_p3m": 0.0045699221060171435, "bench_spy_p3m": -0.05468159936118644, "bench_c_p3m": 0.05942280635161579, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0189, 0.033, 0.0256, 0.0284, 0.0097, 0.0205, 0.0093, 0.0176, -0.009, 0.0165, 0.0017, -0.039, -0.0199, -0.0264, -0.019, -0.0178, -0.0116, 0.0052, 0.0122, -0.0445, -0.0362, -0.0533, -0.0518, -0.0658, -0.103, -0.1016, -0.0882, -0.0882, -0.1001, -0.0763, -0.0887, -0.0933, -0.0932, -0.098, -0.079, -0.0839, -0.1002, -0.1247, -0.1116, -0.1129, -0.1341, -0.0998, -0.0893, -0.0865, -0.0662, -0.0662, -0.0498, -0.0504, -0.0103, 0.0312, 0.0172, 0.0308, 0.0218, 0.0138, 0.0259, 0.0259, 0.0132, 0.005, 0.0027, -0.0146, -0.005, -0.0045, 0.0098, 0.0, 0.0003, 0.0003, -0.0011, -0.0078, -0.0094, -0.0203, -0.0203, 0.0025, 0.0005, 0.0848, 0.0487, 0.0642, 0.0747, 0.0677, 0.0647, 0.0924, 0.068, 0.0819, 0.0819, 0.0706, 0.0979, 0.1011, 0.0917, 0.1102, 0.1104, 0.2207, 0.2366, 0.2252, 0.279, 0.2801, 0.2671, 0.2731, 0.2587, 0.2435, 0.2557, 0.2878, 0.2675, 0.2855, 0.2855, 0.2828, 0.2694, 0.2394, 0.249, 0.2497, 0.2127, 0.2158, 0.2086, 0.2056, 0.1926, 0.152, 0.1504, 0.1253, 0.0983, 0.1152, 0.1224, 0.1292, 0.0802, 0.0844, 0.1034, 0.1053, 0.0844, 0.0702, 0.064, 0.0722, 0.1063, 0.1184, 0.0993, 0.0818, 0.0596, 0.1035, 0.1158, 0.1076, 0.1076, 0.1335, 0.1353, 0.1874, 0.222, 0.2205, 0.2318, 0.2441, 0.2294, 0.2681, 0.3063, 0.3284, 0.3252, 0.3432, 0.6042, 0.5753, 0.5318, 0.5063, 0.5303, 0.5977, 0.5973, 0.5966, 0.6053, 0.6535, 0.6295, 0.6442, 0.7011, 0.6863, 0.7501, 0.7605, 0.7295, 0.7173, 0.7271, 0.7418, 0.7047, 0.7665, 0.7665, 0.8561, 0.8136, 0.805, 0.7463, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2003320154105741402", "ticker_idx": 2, "ts": "2025-12-23T04:21:59+00:00", "symbol": "NXPI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi bullish on analog semis into 2026 upturn; top pick Microchip, also positive on TXN, NXPI, ADI on low inventories and constrained supply.", "resolved_tickers": ["NXPI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi noted in its latest report that the biggest surprise within semiconductors in 2026 could come from analog semiconductors.\n\n“Given currently very low inventories, constrained supply growth, and depressed margins, we expect an analog upturn in ’26. We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices, with Microchip as our top pick.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013098503036950815, "ret_signed_m1d": 0.013098503036950815, "alpha_spy_m1d": 0.017648163836261332, "alpha_c_m1d": 0.022625960411627344, "bench_spy_m1d": -0.004549660799310518, "bench_c_m1d": -0.009527457374676529, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003186170054996995, "ret_signed_p1d": -0.003186170054996995, "alpha_spy_p1d": -0.006703790360529482, "alpha_c_p1d": -0.005967340215367156, "bench_spy_p1d": 0.0035176203055324873, "bench_c_p1d": 0.002781170160370161, "ret_p1w": -0.026551032510364392, "ret_signed_p1w": -0.026551032510364392, "alpha_spy_p1w": -0.025170174606354068, "alpha_c_p1w": -0.026964056783334467, "bench_spy_p1w": -0.0013808579040103242, "bench_c_p1w": 0.00041302427297007505, "ret_p1m": 0.034250836010760066, "ret_signed_p1m": 0.034250836010760066, "alpha_spy_p1m": 0.03797199985993216, "alpha_c_p1m": -0.07250648479195743, "bench_spy_p1m": -0.0037211638491720933, "bench_c_p1m": 0.10675732080271749, "ret_p3m": -0.1531551893481481, "ret_signed_p3m": -0.1531551893481481, "alpha_spy_p3m": -0.09847358998696165, "alpha_c_p3m": -0.21257799569976388, "bench_spy_p3m": -0.05468159936118644, "bench_c_p3m": 0.05942280635161579, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0008, -0.0042, -0.0039, 0.0033, -0.0092, 0.0032, 0.0084, 0.0195, -0.0326, -0.0059, -0.0245, -0.0952, -0.0453, -0.0479, -0.043, -0.0422, -0.0557, -0.0316, -0.0205, -0.0433, -0.0276, -0.0345, -0.0239, -0.0618, -0.0982, -0.0908, -0.0787, -0.0731, -0.0994, -0.0729, -0.0905, -0.0988, -0.0963, -0.1063, -0.1009, -0.1135, -0.1317, -0.1607, -0.1692, -0.1627, -0.1886, -0.157, -0.1561, -0.1585, -0.1464, -0.1464, -0.1412, -0.1211, -0.0513, 0.0025, -0.0037, 0.0042, 0.0089, 0.0047, 0.0212, 0.0259, 0.0096, 0.0259, 0.0167, -0.0122, -0.0173, 0.0013, 0.0131, 0.0, -0.0032, -0.0032, -0.0138, -0.0244, -0.0266, -0.0395, -0.0395, -0.0208, -0.0093, 0.0884, 0.0591, 0.0527, 0.0671, 0.0547, 0.058, 0.0656, 0.0558, 0.0493, 0.0493, 0.0209, 0.0343, 0.0477, 0.0288, 0.0224, 0.0152, 0.0622, 0.0333, 0.0007, 0.0226, -0.0235, 0.0039, -0.017, -0.0073, 0.013, 0.0471, 0.1052, 0.0717, 0.0816, 0.0816, 0.0846, 0.0502, 0.0271, 0.0278, 0.0229, 0.0383, 0.0402, 0.0277, 0.0046, -0.0054, -0.0475, -0.0425, -0.0681, -0.1073, -0.0917, -0.1016, -0.1155, -0.1538, -0.1554, -0.1509, -0.1414, -0.1473, -0.1488, -0.1532, -0.1442, -0.1309, -0.121, -0.1241, -0.1475, -0.1665, -0.1243, -0.13, -0.1346, -0.1346, -0.1234, -0.1321, -0.0914, -0.0852, -0.0909, -0.0748, -0.0664, -0.0686, -0.0493, -0.0391, -0.0154, -0.0014, 0.0042, 0.0727, 0.0855, 0.0536, 0.0248, 0.2866, 0.3059, 0.3133, 0.2933, 0.3004, 0.3502, 0.2909, 0.3111, 0.3611, 0.3088, 0.3274, 0.3085, 0.2966, 0.2974, 0.309, 0.3796, 0.3317, 0.4077, 0.4077, 0.4798, 0.4645, 0.4691, 0.4294, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2003320154105741402", "ticker_idx": 3, "ts": "2025-12-23T04:21:59+00:00", "symbol": "ADI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi bullish on analog semis into 2026 upturn; top pick Microchip, also positive on TXN, NXPI, ADI on low inventories and constrained supply.", "resolved_tickers": ["ADI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi noted in its latest report that the biggest surprise within semiconductors in 2026 could come from analog semiconductors.\n\n“Given currently very low inventories, constrained supply growth, and depressed margins, we expect an analog upturn in ’26. We maintain a positive view on Microchip, Texas Instruments, NXP, and Analog Devices, with Microchip as our top pick.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0032884252982852136, "ret_signed_m1d": -0.0032884252982852136, "alpha_spy_m1d": 0.001261235501025304, "alpha_c_m1d": 0.006239032076391315, "bench_spy_m1d": -0.004549660799310518, "bench_c_m1d": -0.009527457374676529, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002999252731181734, "ret_signed_p1d": 0.002999252731181734, "alpha_spy_p1d": -0.0005183675743507532, "alpha_c_p1d": 0.000218082570811573, "bench_spy_p1d": 0.0035176203055324873, "bench_c_p1d": 0.002781170160370161, "ret_p1w": -0.0069020867661513385, "ret_signed_p1w": -0.0069020867661513385, "alpha_spy_p1w": -0.005521228862141014, "alpha_c_p1w": -0.007315111039121414, "bench_spy_p1w": -0.0013808579040103242, "bench_c_p1w": 0.00041302427297007505, "ret_p1m": 0.10204882193153786, "ret_signed_p1m": 0.10204882193153786, "alpha_spy_p1m": 0.10576998578070995, "alpha_c_p1m": -0.004708498871179634, "bench_spy_p1m": -0.0037211638491720933, "bench_c_p1m": 0.10675732080271749, "ret_p3m": 0.12166675793191839, "ret_signed_p3m": 0.12166675793191839, "alpha_spy_p3m": 0.17634835729310483, "alpha_c_p3m": 0.0622439515803026, "bench_spy_p3m": -0.05468159936118644, "bench_c_p3m": 0.05942280635161579, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1087, -0.1086, -0.1185, -0.1153, -0.1384, -0.1298, -0.1286, -0.1268, -0.1583, -0.1432, -0.1434, -0.1886, -0.155, -0.1523, -0.1424, -0.13, -0.1254, -0.1134, -0.1128, -0.1345, -0.1239, -0.1429, -0.1249, -0.1381, -0.1536, -0.1613, -0.1569, -0.1588, -0.174, -0.1502, -0.1614, -0.1773, -0.1646, -0.1595, -0.1306, -0.1447, -0.1542, -0.172, -0.1713, -0.1639, -0.1891, -0.1634, -0.1379, -0.0925, -0.0713, -0.0713, -0.0445, -0.0403, -0.0171, 0.0019, -0.0016, 0.0129, 0.0087, -0.0018, 0.0175, 0.0241, 0.0094, 0.0134, 0.006, -0.0206, -0.0065, -0.0083, -0.0033, 0.0, 0.003, 0.003, 0.0004, -0.004, -0.0069, -0.02, -0.02, -0.0108, 0.002, 0.0586, 0.0584, 0.0811, 0.0874, 0.0619, 0.0704, 0.0768, 0.0917, 0.085, 0.085, 0.0684, 0.102, 0.1149, 0.1043, 0.0986, 0.0979, 0.1478, 0.1517, 0.1234, 0.145, 0.1249, 0.158, 0.164, 0.158, 0.1671, 0.175, 0.2178, 0.1974, 0.2182, 0.2182, 0.2196, 0.2517, 0.2478, 0.2829, 0.2834, 0.2868, 0.3038, 0.2805, 0.2857, 0.2735, 0.2288, 0.238, 0.1952, 0.1448, 0.1589, 0.1557, 0.1572, 0.1138, 0.1095, 0.1271, 0.137, 0.1186, 0.1253, 0.1217, 0.1317, 0.1666, 0.1673, 0.1361, 0.1145, 0.0987, 0.1532, 0.1621, 0.154, 0.154, 0.1867, 0.1868, 0.255, 0.2737, 0.2692, 0.2688, 0.2637, 0.2613, 0.2825, 0.3465, 0.3813, 0.3603, 0.3826, 0.464, 0.4484, 0.4231, 0.3893, 0.4112, 0.4582, 0.4416, 0.4392, 0.4673, 0.5066, 0.4809, 0.5099, 0.5324, 0.5212, 0.5674, 0.5471, 0.5134, 0.5173, 0.5019, 0.4429, 0.3927, 0.4394, 0.4394, 0.5223, 0.5112, 0.5189, 0.5002, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2001537710662783477", "ticker_idx": 0, "ts": "2025-12-18T06:19:11+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy rating; raise target price to US$323", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF Securities reiterates Buy on MU, raises PT to $323 on strong FY1Q26, blowout FY2Q26 guidance, HBM4 upside, and DRAM upcycle through 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics & Telecom》\n☄️ Micron FY1Q26 Earnings Review: DRAM Remains Strong\n\n☀️ Maintain Buy rating; raise target price to US$323: Micron reported strong FY1Q26 results, with revenue of US$13.6bn and EPS of US$4.78. FY2Q26 guidance far exceeded expectations, with revenue midpoint of US$18.7bn and EPS midpoint of US$8.42 (vs. recent buy-side expectations of ~US$6.5 and our prior forecast of US$7.34). We see Micron’s gross margin on a clear upward trend: 57% in FY1Q26, expected to exceed 68% in FY2Q. We remain bullish on the long-term AI-driven upcycle in DRAM. Micron’s HBM4 development has also exceeded expectations, with volume shipments starting in FY2Q26. Maintaining 4x FY2026E P/B, we raise our target price to US$323.\n\n☀️ DRAM cycle remains strong throughout 2026: After ~45% QoQ increases in DRAM contract prices, we expect prices in CY1Q26 to stay strong, with QoQ gains of over 30%. Although Samsung and SK are accelerating DRAM capacity expansion, we believe supply pressure will not emerge in the near term, as wafers from new capacity will only come online in 2027. Management stated that DRAM supply will remain tight for the foreseeable future. Meanwhile, we see CSPs working to raise inventory to 3–4 months, versus current inventory of less than 2 months. CSPs have already begun discussions with memory makers on 2027 LTAs, and our channel checks indicate that Meta and Google are even negotiating with Samsung for 5-year LTAs. Despite weak demand in consumer electronics, we expect DRAM pricing to remain healthy in CY2Q26, supported by strong CSP demand, though the pace of increases should slow to 5–10%.\n\n☀️ HBM4 development exceeds expectations: Management stated that all HBM capacity for 2026 has been fully booked, including its industry-leading HBM4 products. We believe Micron’s HBM4 development has exceeded expectations and is planned for volume shipments starting in FY2Q26. Our channel checks show that Micron’s 11 GB/s HBM4 has passed NVIDIA certification, and we estimate FY2Q26 HBM4 shipments at ~300k units. Looking ahead, Micron raised its HBM TAM forecast to US$100bn by 2028, implying a 40% CAGR, reaching the US$100bn milestone two years earlier than previously expected.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09265740689197877, "ret_signed_m1d": -0.09265740689197877, "alpha_spy_m1d": -0.0851626695684754, "alpha_c_m1d": -0.06959040894286439, "bench_spy_m1d": -0.007494737323503364, "bench_c_m1d": -0.023066997949114376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06988528442465691, "ret_signed_p1d": 0.06988528442465691, "alpha_spy_p1d": 0.0608218389452051, "alpha_c_p1d": 0.04402494212140229, "bench_spy_p1d": 0.009063445479451815, "bench_c_p1d": 0.025860342303254624, "ret_p1w": 0.15340979399907528, "ret_signed_p1w": 0.15340979399907528, "alpha_spy_p1w": 0.1298314425708822, "alpha_c_p1w": 0.10142084748290103, "bench_spy_p1w": 0.02357835142819309, "bench_c_p1w": 0.05198894651617425, "ret_p1m": 0.46005451856089974, "ret_signed_p1m": 0.46005451856089974, "alpha_spy_p1m": 0.4345784367779224, "alpha_c_p1m": 0.30343575761179387, "bench_spy_p1m": 0.025476081782977333, "bench_c_p1m": 0.15661876094910587, "ret_p3m": 0.858284060061105, "ret_signed_p3m": 0.858284060061105, "alpha_spy_p3m": 0.8637504925830166, "alpha_c_p3m": 0.7118047706210504, "bench_spy_p3m": -0.005466432521911591, "bench_c_p3m": 0.14647928944005462, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3381, -0.3309, -0.3498, -0.3694, -0.3676, -0.341, -0.3272, -0.2676, -0.2612, -0.2443, -0.2317, -0.2529, -0.2093, -0.2262, -0.2694, -0.2244, -0.2474, -0.2278, -0.1852, -0.1858, -0.1681, -0.1861, -0.2015, -0.1683, -0.1188, -0.1145, -0.1072, -0.0882, -0.0987, -0.0997, -0.0557, -0.1228, -0.0445, -0.0411, -0.0428, 0.0191, -0.0299, -0.0147, -0.0467, -0.0069, -0.0266, -0.0807, -0.091, -0.1898, -0.1657, -0.0991, -0.0966, -0.0736, -0.0736, -0.0486, -0.0325, -0.0365, -0.0579, -0.0881, -0.0456, -0.0066, 0.0156, 0.061, 0.0399, -0.0298, -0.0445, -0.0645, -0.0927, 0.0, 0.0699, 0.1128, 0.1115, 0.1534, 0.1534, 0.1458, 0.1848, 0.1778, 0.1488, 0.1488, 0.2696, 0.2564, 0.3823, 0.3667, 0.3162, 0.389, 0.3921, 0.361, 0.3417, 0.3549, 0.4601, 0.4601, 0.4691, 0.5662, 0.6002, 0.6086, 0.5661, 0.6512, 0.752, 0.754, 0.6699, 0.7621, 0.6882, 0.5271, 0.5411, 0.5886, 0.5436, 0.5023, 0.6516, 0.6662, 0.6569, 0.6569, 0.6091, 0.6943, 0.6798, 0.7234, 0.6944, 0.6825, 0.7267, 0.6726, 0.6598, 0.661, 0.5282, 0.6131, 0.5981, 0.4904, 0.567, 0.6225, 0.6852, 0.6315, 0.7152, 0.7782, 0.8583, 0.8584, 0.7882, 0.7022, 0.6275, 0.592, 0.5379, 0.4307, 0.4378, 0.2958, 0.3604, 0.4812, 0.4747, 0.4747, 0.5211, 0.5204, 0.6378, 0.6973, 0.6936, 0.7176, 0.8751, 0.8371, 0.8411, 0.8324, 0.8056, 0.8095, 0.9629, 0.9397, 1.0001, 1.1122, 1.0306, 1.0877, 1.0824, 1.1833, 1.3212, 1.5779, 1.6841, 1.6038, 2.0071, 2.2025, 2.0868, 2.2359, 2.1247, 1.918, 1.7443, 1.8136, 1.9475, 2.0687, 2.024, 2.024, 2.6074, 2.7384, 2.7187, 2.9099, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1976080143195185392", "ticker_idx": 0, "ts": "2025-10-09T00:19:54+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price raised from ₩97,000 → ₩111,000, applying 2026E P/B of 1.62x", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises PTs on Samsung (₩111k) and SK Hynix (₩480k); bullish DRAM cycle through 2026 on AI-driven demand, supply shortfall, and HBM expansion.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Korea Memory Revival – DRAM-Centric Bullish Cycle Resumes (Morgan Stanley)\n\n⸻\n\n1. Market and Cycle Outlook\n    • Memory price uptrend in 4Q is steeper than expected; strong cycle expected to persist through 2026.\n    • DRAM-driven supply-demand imbalance is deepening, with supply unable to keep up with demand for the next 4–6 quarters.\n    • DRAM prices in 2025 could rise by more than 50% YoY; NAND also accelerating recovery, led by enterprise SSD demand.\n    • Rapid AI token growth (50% MoM) is driving memory demand, mirroring past DRAM price upcycles.\n    • Morgan Stanley sees DRAM as still undervalued relative to the semiconductor sector, leaving room for further valuation re-rating.\n\n⸻\n\n2. Samsung Electronics\n    • Target price raised from ₩97,000 → ₩111,000, applying 2026E P/B of 1.62x.\n    • EPS estimates revised up by +6.7% (2025), +9.3% (2026), +18.1% (2027).\n    • DRAM price increase assumption revised from +26% → +35%, NAND from +10% → +15%.\n    • Expanding DRAM capacity at Pyeongtaek P4 fab and transitioning to HBM4 production, slated for commercialization in 2026.\n    • Over the next two quarters, HBM4 qualification and customer wins are seen as key share price catalysts.\n\n⸻\n\n3. SK Hynix\n    • Target price raised from ₩410,000 → ₩480,000.\n    • EPS estimates revised up by +6% (2025), +16% (2026), +26% (2027).\n    • DRAM ASP assumption raised (4Q25 +15%) and NAND margin revised from 10% → 15%.\n    • Expected surge in DRAM/HBM demand due to expanding AI infrastructure investments (including OpenAI’s Stargate project).\n    • Reflecting higher HBM/DRAM shipment forecasts for 2026–27, 2027E EPS: ₩77,739.\n    • Valuation based on 2026E P/B of 2.4x, similar to historical cycle peaks.\n    • Additional profit leverage expected from full depreciation of Solidigm’s Dalian SSD fab assets.\n\n⸻\n\n4. Cycle and Valuation Insights\n    • DRAM stocks remain undervalued relative to the broader semiconductor sector, but AI-driven memory TAM expansion (HBM expected to surpass DRAM by 2028) supports re-rating potential.\n    • SK Hynix’s 2027E EPS estimated at ₩65,000–₩77,000, with bull-case target price of ₩580,000.\n    • Both DRAM and NAND inventories have fallen below normal levels, with customer restocking worsening supply shortages.\n    • While short-term overheating risks exist, valuations remain below past cycle peaks (Samsung P/B 1.3x, SK Hynix P/B 2.1x).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.002905425421027452, "alpha_c_m1d": -0.0016884807796115897, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0016884807796115897, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05181056991828514, "ret_signed_p1d": 0.05181056991828514, "alpha_spy_p1d": 0.07883839781737478, "alpha_c_p1d": 0.09250652094938994, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.04069595103110479, "ret_p1w": 0.08857943215993314, "ret_signed_p1w": 0.08857943215993314, "alpha_spy_p1w": 0.10425371209357515, "alpha_c_p1w": 0.10818656195245524, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.019607129792522104, "ret_p1m": 0.09080788523887362, "ret_signed_p1m": 0.09080788523887362, "alpha_spy_p1m": 0.09109103800591911, "alpha_c_p1m": 0.09783756261904153, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": -0.007029677380167909, "ret_p3m": 0.5551555723028456, "ret_signed_p3m": 0.5551555723028456, "alpha_spy_p3m": 0.5213421383319947, "alpha_c_p3m": 0.5429403439398273, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.01221522836301836, "ret_p6m": 1.0890520493327482, "ret_signed_p6m": 1.0890520493327482, "alpha_spy_p6m": 1.1063291239183515, "alpha_c_p6m": 1.14924767562762, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": -0.06019562629487185, "price_path": [-0.3067, -0.2934, -0.2823, -0.2601, -0.2557, -0.2479, -0.2679, -0.2635, -0.2679, -0.269, -0.2191, -0.2169, -0.1947, -0.208, -0.2357, -0.2268, -0.2246, -0.2368, -0.218, -0.2036, -0.2124, -0.2113, -0.2024, -0.2058, -0.2058, -0.2235, -0.2235, -0.218, -0.2169, -0.208, -0.2069, -0.2202, -0.2169, -0.228, -0.2268, -0.2501, -0.2335, -0.2257, -0.2224, -0.2291, -0.2224, -0.2069, -0.1947, -0.1858, -0.1636, -0.1514, -0.1192, -0.1326, -0.1093, -0.1093, -0.0738, -0.0605, -0.0527, -0.0449, -0.076, -0.0618, -0.0652, -0.0418, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0518, 0.0396, 0.0206, 0.0585, 0.0886, 0.0908, 0.093, 0.0864, 0.0986, 0.0752, 0.1008, 0.1365, 0.1086, 0.1198, 0.1599, 0.1978, 0.2379, 0.1688, 0.1209, 0.1053, 0.0908, 0.1209, 0.1532, 0.1487, 0.1454, 0.083, 0.1209, 0.0897, 0.0752, 0.1209, 0.0563, 0.0774, 0.1064, 0.1454, 0.1532, 0.1198, 0.1231, 0.1521, 0.1643, 0.171, 0.2078, 0.2201, 0.2078, 0.2033, 0.1955, 0.2134, 0.1677, 0.1454, 0.2022, 0.1989, 0.1844, 0.2312, 0.2423, 0.2379, 0.2379, 0.3036, 0.3379, 0.3424, 0.3424, 0.3424, 0.4387, 0.5462, 0.5552, 0.5787, 0.554, 0.5563, 0.554, 0.5406, 0.5708, 0.6111, 0.6671, 0.6716, 0.6257, 0.6738, 0.7052, 0.7029, 0.7029, 0.7858, 0.8183, 0.7992, 0.797, 0.6839, 0.8754, 0.8933, 0.7836, 0.7757, 0.8631, 0.8563, 0.8787, 0.9996, 1.0288, 1.0288, 1.0288, 1.0288, 1.1273, 1.1284, 1.1609, 1.2392, 1.2784, 1.4408, 1.424, 1.424, 1.1844, 0.928, 1.1452, 1.1071, 0.9425, 1.1038, 1.1273, 1.1038, 1.0545, 1.1127, 1.1709, 1.3344, 1.2448, 1.2325, 1.0859, 1.1239, 1.1161, 1.0164, 1.0164, 0.978, 0.8759, 1.1278, 1.0015, 1.0891]}
{"unit_id": "orig:1976080143195185392", "ticker_idx": 1, "ts": "2025-10-09T00:19:54+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price raised from ₩410,000 → ₩480,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises PTs on Samsung (₩111k) and SK Hynix (₩480k); bullish DRAM cycle through 2026 on AI-driven demand, supply shortfall, and HBM expansion.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Korea Memory Revival – DRAM-Centric Bullish Cycle Resumes (Morgan Stanley)\n\n⸻\n\n1. Market and Cycle Outlook\n    • Memory price uptrend in 4Q is steeper than expected; strong cycle expected to persist through 2026.\n    • DRAM-driven supply-demand imbalance is deepening, with supply unable to keep up with demand for the next 4–6 quarters.\n    • DRAM prices in 2025 could rise by more than 50% YoY; NAND also accelerating recovery, led by enterprise SSD demand.\n    • Rapid AI token growth (50% MoM) is driving memory demand, mirroring past DRAM price upcycles.\n    • Morgan Stanley sees DRAM as still undervalued relative to the semiconductor sector, leaving room for further valuation re-rating.\n\n⸻\n\n2. Samsung Electronics\n    • Target price raised from ₩97,000 → ₩111,000, applying 2026E P/B of 1.62x.\n    • EPS estimates revised up by +6.7% (2025), +9.3% (2026), +18.1% (2027).\n    • DRAM price increase assumption revised from +26% → +35%, NAND from +10% → +15%.\n    • Expanding DRAM capacity at Pyeongtaek P4 fab and transitioning to HBM4 production, slated for commercialization in 2026.\n    • Over the next two quarters, HBM4 qualification and customer wins are seen as key share price catalysts.\n\n⸻\n\n3. SK Hynix\n    • Target price raised from ₩410,000 → ₩480,000.\n    • EPS estimates revised up by +6% (2025), +16% (2026), +26% (2027).\n    • DRAM ASP assumption raised (4Q25 +15%) and NAND margin revised from 10% → 15%.\n    • Expected surge in DRAM/HBM demand due to expanding AI infrastructure investments (including OpenAI’s Stargate project).\n    • Reflecting higher HBM/DRAM shipment forecasts for 2026–27, 2027E EPS: ₩77,739.\n    • Valuation based on 2026E P/B of 2.4x, similar to historical cycle peaks.\n    • Additional profit leverage expected from full depreciation of Solidigm’s Dalian SSD fab assets.\n\n⸻\n\n4. Cycle and Valuation Insights\n    • DRAM stocks remain undervalued relative to the broader semiconductor sector, but AI-driven memory TAM expansion (HBM expected to surpass DRAM by 2028) supports re-rating potential.\n    • SK Hynix’s 2027E EPS estimated at ₩65,000–₩77,000, with bull-case target price of ₩580,000.\n    • Both DRAM and NAND inventories have fallen below normal levels, with customer restocking worsening supply shortages.\n    • While short-term overheating risks exist, valuations remain below past cycle peaks (Samsung P/B 1.3x, SK Hynix P/B 2.1x).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.002905425421027452, "alpha_c_m1d": -0.0024913201957443842, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08217436875170292, "ret_signed_p1d": 0.08217436875170292, "alpha_spy_p1d": 0.10920219665079256, "alpha_c_p1d": 0.13973349570598903, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": 0.14412133100408764, "ret_signed_p1w": 0.14412133100408764, "alpha_spy_p1w": 0.15979561093772965, "alpha_c_p1w": 0.14965421848800142, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": 0.46649802373545723, "ret_signed_p1m": 0.46649802373545723, "alpha_spy_p1m": 0.4667811765025027, "alpha_c_p1m": 0.45806829148631323, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": 0.8369655237764586, "ret_signed_p3m": 0.8369655237764586, "alpha_spy_p3m": 0.8031520898056077, "alpha_c_p3m": 0.7106188594177851, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": 1.2205940178927466, "ret_signed_p6m": 1.2205940178927466, "alpha_spy_p6m": 1.2378710924783498, "alpha_c_p6m": 1.0805900522118483, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.2426, -0.2463, -0.2527, -0.3196, -0.3208, -0.312, -0.3221, -0.3208, -0.3196, -0.3284, -0.3385, -0.3372, -0.3347, -0.3095, -0.3486, -0.3486, -0.3347, -0.3473, -0.3385, -0.3524, -0.3259, -0.3208, -0.2981, -0.3019, -0.3019, -0.3246, -0.336, -0.3549, -0.3814, -0.3663, -0.3448, -0.3398, -0.3436, -0.3211, -0.3198, -0.3527, -0.3413, -0.3363, -0.3287, -0.3085, -0.2996, -0.2718, -0.2314, -0.2238, -0.1694, -0.1631, -0.1201, -0.1568, -0.1011, -0.1011, -0.1125, -0.0872, -0.0961, -0.0986, -0.1492, -0.1176, -0.1214, -0.0898, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0822, 0.0493, 0.0405, 0.0683, 0.1441, 0.177, 0.2276, 0.2111, 0.2174, 0.2099, 0.2895, 0.3527, 0.3173, 0.4109, 0.4362, 0.4134, 0.5676, 0.4817, 0.464, 0.4994, 0.4665, 0.5322, 0.5651, 0.5601, 0.5474, 0.4159, 0.5322, 0.4412, 0.421, 0.4437, 0.3173, 0.3148, 0.3123, 0.3249, 0.3765, 0.341, 0.3613, 0.4119, 0.3967, 0.3714, 0.3765, 0.46, 0.4321, 0.4853, 0.4296, 0.4448, 0.4018, 0.341, 0.3942, 0.3967, 0.384, 0.4675, 0.4777, 0.4878, 0.4878, 0.5156, 0.6194, 0.6472, 0.6472, 0.6472, 0.713, 0.7611, 0.837, 0.8774, 0.9129, 0.8825, 0.8952, 0.8673, 0.8774, 0.8952, 0.9129, 0.9331, 0.88, 0.8724, 0.9103, 0.9407, 0.8623, 1.0242, 1.1279, 1.1786, 1.3, 1.1001, 1.2949, 1.2772, 1.1305, 1.1229, 1.2443, 1.2165, 1.176, 1.2469, 1.2266, 1.2266, 1.2266, 1.2266, 1.262, 1.4012, 1.4063, 1.5429, 1.5758, 1.7859, 1.6896, 1.6896, 1.3803, 1.1522, 1.3854, 1.3423, 1.1192, 1.3778, 1.4209, 1.3575, 1.3068, 1.469, 1.4589, 1.6769, 1.5679, 1.5527, 1.3651, 1.4994, 1.5222, 1.3651, 1.3651, 1.213, 1.0457, 1.2738, 1.104, 1.2206]}
{"unit_id": "orig:1976080143195185392", "ticker_idx": 2, "ts": "2025-10-09T00:19:54+00:00", "symbol": "DRAM", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM prices in 2025 could rise by more than 50% YoY; DRAM stocks remain undervalued relative to the broader semiconductor sector", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises PTs on Samsung (₩111k) and SK Hynix (₩480k); bullish DRAM cycle through 2026 on AI-driven demand, supply shortfall, and HBM expansion.", "resolved_tickers": ["005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: Samsung Electronics + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Korea Memory Revival – DRAM-Centric Bullish Cycle Resumes (Morgan Stanley)\n\n⸻\n\n1. Market and Cycle Outlook\n    • Memory price uptrend in 4Q is steeper than expected; strong cycle expected to persist through 2026.\n    • DRAM-driven supply-demand imbalance is deepening, with supply unable to keep up with demand for the next 4–6 quarters.\n    • DRAM prices in 2025 could rise by more than 50% YoY; NAND also accelerating recovery, led by enterprise SSD demand.\n    • Rapid AI token growth (50% MoM) is driving memory demand, mirroring past DRAM price upcycles.\n    • Morgan Stanley sees DRAM as still undervalued relative to the semiconductor sector, leaving room for further valuation re-rating.\n\n⸻\n\n2. Samsung Electronics\n    • Target price raised from ₩97,000 → ₩111,000, applying 2026E P/B of 1.62x.\n    • EPS estimates revised up by +6.7% (2025), +9.3% (2026), +18.1% (2027).\n    • DRAM price increase assumption revised from +26% → +35%, NAND from +10% → +15%.\n    • Expanding DRAM capacity at Pyeongtaek P4 fab and transitioning to HBM4 production, slated for commercialization in 2026.\n    • Over the next two quarters, HBM4 qualification and customer wins are seen as key share price catalysts.\n\n⸻\n\n3. SK Hynix\n    • Target price raised from ₩410,000 → ₩480,000.\n    • EPS estimates revised up by +6% (2025), +16% (2026), +26% (2027).\n    • DRAM ASP assumption raised (4Q25 +15%) and NAND margin revised from 10% → 15%.\n    • Expected surge in DRAM/HBM demand due to expanding AI infrastructure investments (including OpenAI’s Stargate project).\n    • Reflecting higher HBM/DRAM shipment forecasts for 2026–27, 2027E EPS: ₩77,739.\n    • Valuation based on 2026E P/B of 2.4x, similar to historical cycle peaks.\n    • Additional profit leverage expected from full depreciation of Solidigm’s Dalian SSD fab assets.\n\n⸻\n\n4. Cycle and Valuation Insights\n    • DRAM stocks remain undervalued relative to the broader semiconductor sector, but AI-driven memory TAM expansion (HBM expected to surpass DRAM by 2028) supports re-rating potential.\n    • SK Hynix’s 2027E EPS estimated at ₩65,000–₩77,000, with bull-case target price of ₩580,000.\n    • Both DRAM and NAND inventories have fallen below normal levels, with customer restocking worsening supply shortages.\n    • While short-term overheating risks exist, valuations remain below past cycle peaks (Samsung P/B 1.3x, SK Hynix P/B 2.1x).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.002905425421027452, "alpha_c_m1d": -0.0016884807796115897, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0016884807796115897, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06699246933499403, "ret_signed_p1d": 0.06699246933499403, "alpha_spy_p1d": 0.09402029723408367, "alpha_c_p1d": 0.10768842036609882, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.04069595103110479, "ret_p1w": 0.11635038158201039, "ret_signed_p1w": 0.11635038158201039, "alpha_spy_p1w": 0.1320246615156524, "alpha_c_p1w": 0.1359575113745325, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.019607129792522104, "ret_p1m": 0.2786529544871654, "ret_signed_p1m": 0.2786529544871654, "alpha_spy_p1m": 0.2789361072542109, "alpha_c_p1m": 0.28568263186733334, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": -0.007029677380167909, "ret_p3m": 0.6960605480396521, "ret_signed_p3m": 0.6960605480396521, "alpha_spy_p3m": 0.6622471140688012, "alpha_c_p3m": 0.6838453196766338, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.01221522836301836, "ret_p6m": 1.1548230336127474, "ret_signed_p6m": 1.1548230336127474, "alpha_spy_p6m": 1.1721001081983506, "alpha_c_p6m": 1.2150186599076194, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": -0.06019562629487185, "price_path": [-0.2746, -0.2699, -0.2675, -0.2898, -0.2883, -0.28, -0.295, -0.2921, -0.2937, -0.2987, -0.2788, -0.2771, -0.2647, -0.2587, -0.2922, -0.2877, -0.2797, -0.2921, -0.2782, -0.278, -0.2692, -0.2661, -0.2503, -0.2538, -0.2538, -0.2741, -0.2797, -0.2864, -0.2991, -0.2871, -0.2758, -0.28, -0.2802, -0.2745, -0.2733, -0.3014, -0.2874, -0.281, -0.2756, -0.2688, -0.261, -0.2393, -0.213, -0.2048, -0.1665, -0.1573, -0.1197, -0.1447, -0.1052, -0.1052, -0.0931, -0.0738, -0.0744, -0.0718, -0.1126, -0.0897, -0.0933, -0.0658, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.067, 0.0444, 0.0305, 0.0634, 0.1164, 0.1339, 0.1603, 0.1487, 0.158, 0.1425, 0.1952, 0.2446, 0.213, 0.2653, 0.298, 0.3056, 0.4028, 0.3252, 0.2924, 0.3023, 0.2787, 0.3266, 0.3592, 0.3544, 0.3464, 0.2495, 0.3266, 0.2655, 0.2481, 0.2823, 0.1868, 0.1961, 0.2093, 0.2352, 0.2648, 0.2304, 0.2422, 0.282, 0.2805, 0.2712, 0.2921, 0.34, 0.32, 0.3443, 0.3126, 0.3291, 0.2847, 0.2432, 0.2982, 0.2978, 0.2842, 0.3494, 0.36, 0.3628, 0.3628, 0.4096, 0.4787, 0.4948, 0.4948, 0.4948, 0.5758, 0.6536, 0.6961, 0.7281, 0.7335, 0.7194, 0.7246, 0.704, 0.7241, 0.7531, 0.79, 0.8024, 0.7528, 0.7731, 0.8078, 0.8218, 0.7826, 0.905, 0.9731, 0.9889, 1.0485, 0.892, 1.0852, 1.0853, 0.957, 0.9493, 1.0537, 1.0364, 1.0274, 1.1233, 1.1277, 1.1277, 1.1277, 1.1277, 1.1947, 1.2648, 1.2836, 1.3911, 1.4271, 1.6133, 1.5568, 1.5568, 1.2823, 1.0401, 1.2653, 1.2247, 1.0309, 1.2408, 1.2741, 1.2306, 1.1806, 1.2909, 1.3149, 1.5056, 1.4064, 1.3926, 1.2255, 1.3117, 1.3192, 1.1908, 1.1908, 1.0955, 0.9608, 1.2008, 1.0528, 1.1548]}
{"unit_id": "thread:1982737417657524660", "ticker_idx": 0, "ts": "2025-10-27T09:13:32+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Finally, memory companies are colluding to limit supply. This is a welcome development.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish on memory sector; supply discipline by Samsung and SK Hynix limiting capex signals favorable pricing environment ahead.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Finally, memory companies are colluding to limit supply. This is a welcome development.\n\nAn executive at Samsung Semiconductor stated, “Korean companies, too, have learned their own lessons about how to navigate a super-boom cycle,” adding, “In the past, many companies repeated the vicious cycle of overinvesting in equipment during a boom → oversupply → price collapse. In fact, since Samsung Electronics has the largest ‘scale’ in the memory industry, the cost of maintaining its invested facilities is enormous, so the company is taking an extremely conservative approach.”\n\nSK Hynix, despite holding a dominant position in the high-bandwidth memory (HBM) market, is also proceeding with facility investments as conservatively as possible. In particular, it has been cautious in placing orders for essential equipment such as TC bonders. Some analysts attribute this to Nvidia’s move to diversify its HBM suppliers from the sixth generation (HBM4) onward — adding Samsung Electronics and Micron alongside SK Hynix — but others believe the company is also mindful of the potential risk of HBM oversupply.\"\n\n---\n\nhttps://t.co/3hvmj6XgzP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027669446585673607, "ret_signed_m1d": -0.027669446585673607, "alpha_spy_m1d": -0.01600929764191017, "alpha_c_m1d": -0.003221006350285216, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014151459320133028, "ret_signed_p1d": -0.014151459320133028, "alpha_spy_p1d": -0.016807458276724106, "alpha_c_p1d": -0.02295845320545028, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.10480926356812585, "ret_signed_p1w": 0.10480926356812585, "alpha_spy_p1w": 0.10758198315339944, "alpha_c_p1w": 0.08769527885925547, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": -0.012083281799602311, "ret_signed_p1m": -0.012083281799602311, "alpha_spy_p1m": 0.0028312496768069986, "alpha_c_p1m": 0.043287018137597465, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.5732372537610889, "ret_signed_p3m": 0.5732372537610889, "alpha_spy_p3m": 0.5648082865496423, "alpha_c_p3m": 0.45062398380400126, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 1.1663555257714706, "ret_signed_p6m": 1.1663555257714706, "alpha_spy_p6m": 1.1330109593093294, "alpha_c_p6m": 0.8714012617208657, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.4262, -0.4324, -0.4566, -0.4496, -0.4436, -0.4507, -0.4383, -0.4269, -0.4156, -0.408, -0.405, -0.4054, -0.4121, -0.4188, -0.4239, -0.4343, -0.4426, -0.4334, -0.4297, -0.4323, -0.4303, -0.4216, -0.4255, -0.4404, -0.4336, -0.4297, -0.4185, -0.4046, -0.4004, -0.3833, -0.3625, -0.342, -0.312, -0.3061, -0.2845, -0.2956, -0.2616, -0.271, -0.2605, -0.2476, -0.2547, -0.2604, -0.2813, -0.2593, -0.2561, -0.219, -0.1822, -0.1758, -0.1711, -0.1791, -0.1626, -0.169, -0.1498, -0.1446, -0.161, -0.1356, -0.0921, -0.0835, -0.0638, -0.0766, -0.0772, -0.0735, -0.0277, 0.0, -0.0142, 0.0193, 0.0333, 0.0385, 0.1048, 0.0381, 0.0492, 0.0546, 0.0416, 0.0899, 0.0891, 0.0922, 0.0761, 0.0404, 0.0728, 0.0208, 0.0077, -0.0105, -0.0515, -0.0209, -0.0121, 0.0111, 0.0261, 0.017, 0.029, 0.0485, 0.0403, 0.0247, 0.0527, 0.0916, 0.0894, 0.1183, 0.0944, 0.0771, 0.0476, 0.0185, 0.0377, 0.0722, 0.0912, 0.1416, 0.1469, 0.1639, 0.1639, 0.1871, 0.2375, 0.243, 0.2321, 0.2321, 0.322, 0.3604, 0.4291, 0.4401, 0.4226, 0.4432, 0.4468, 0.4243, 0.4284, 0.4496, 0.5099, 0.5162, 0.4931, 0.5419, 0.5732, 0.5832, 0.5479, 0.6441, 0.7171, 0.7248, 0.7224, 0.6747, 0.751, 0.6913, 0.6282, 0.6419, 0.6805, 0.6562, 0.7089, 0.7674, 0.7674, 0.7674, 0.7494, 0.7815, 0.8137, 0.8647, 0.8645, 0.9167, 0.953, 1.032, 0.9985, 0.9989, 0.8025, 0.703, 0.8185, 0.7562, 0.6818, 0.8137, 0.8548, 0.8121, 0.8166, 0.8974, 0.9421, 1.0439, 0.9643, 0.9245, 0.8072, 0.8381, 0.8211, 0.7128, 0.7154, 0.6132, 0.5664, 0.7419, 0.6606, 0.715, 0.7614, 0.7911, 0.9544, 0.9336, 0.9569, 0.9576, 1.0743, 1.0954, 1.1302, 1.1051, 1.1139, 1.1664]}
{"unit_id": "thread:1995431391430098981", "ticker_idx": 0, "ts": "2025-12-01T09:54:51+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the forecast that the 'memory super cycle' will continue beyond 2028 is gaining traction", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM supply shortage persists until H1 2027+; Samsung and SK Hynix adopt conservative CAPEX/profitability strategies, supporting memory supercycle thesis beyond 2028.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"DRAM Supply Shortage Unlikely to Ease Until H1 2027\"\n\n“We will minimize oversupply risks.” (Samsung Electronics)\n\n“It is difficult to resolve the supply shortage until the first half of 2027.” (SK Hynix)\n\nSamsung Electronics and SK Hynix, which dominate approximately 70% of the global DRAM market, recently expressed a cautious stance regarding the possibility of aggressive supply expansion during investor relations meetings for global investment banks (IB). With semiconductor supply shortages spreading from server applications to PC and smartphone products, and considering both companies have announced production strategies centered on ‘profitability,’ the forecast that the ‘memory super cycle’ will continue beyond 2028 is gaining traction.\n\nAccording to industry sources on the 1st, Samsung Electronics unveiled a conservative semiconductor supply strategy at an Investor Relations (IR) event hosted by Morgan Stanley on the 23rd of last month. Samsung explained, “We will employ a strategy of maintaining long-term profitability rather than rapidly increasing facilities,” adding, “We will minimize oversupply risks through a Capital Expenditure (CAPEX) strategy that finds a balance between customer demand and price.”\n\nIt is reported that Samsung Electronics’ Memory Business Division is currently able to supply only about 70% of its DRAM orders. As the supply shortage intensifies, the division has reportedly refused requests for long-term mobile DRAM supply contracts from major clients. Samsung stated, “While clients want multi-year long-term contracts, Samsung does not want to tie up volumes with specific clients during a phase where prices are surging.”\n\nSK Hynix is in a similar position. At a recent IR event hosted by Bank of America Merrill Lynch, SK Hynix stated it would convert about half of its general-purpose DRAM production capacity to the latest 10-nanometer (nm) 6th generation (1c) DRAM. However, the company explained that this is insufficient to meet client demand. A company official explained, “We plan to spend about 30% of our revenue on CAPEX in 2026 and accelerate the transition to 1c DRAM, but it is difficult to resolve the supply shortage situation.”\n\nSK Hynix also anticipates that the upward trend in DRAM prices will continue for several years and is responding with ‘short-term contracts.’ An SK Hynix official explained, “We will lower the burden on clients through a strategy of phased contract price increases.”\n\n---\n\nhttps://t.co/nuIS5boUkR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0114658706649311, "ret_signed_m1d": -0.0114658706649311, "alpha_spy_m1d": -0.016052224880577226, "alpha_c_m1d": -0.009539317334148256, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019644831246580636, "ret_signed_p1d": 0.019644831246580636, "alpha_spy_p1d": 0.01779262197246863, "alpha_c_p1d": 0.0013141702366183568, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.061888469616404373, "ret_signed_p1w": 0.061888469616404373, "alpha_spy_p1w": 0.05694930504065798, "alpha_c_p1w": 0.017719201095217185, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.20758480185732853, "ret_signed_p1m": 0.20758480185732853, "alpha_spy_p1m": 0.19469278646108146, "alpha_c_p1m": 0.17505854863278225, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 0.9828726284367323, "ret_signed_p3m": 0.9828726284367323, "alpha_spy_p3m": 0.9666044615229631, "alpha_c_p3m": 0.8119410423724093, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.512262236649745, "ret_signed_p6m": 2.512262236649745, "alpha_spy_p6m": 2.402608957196238, "alpha_c_p6m": 1.800981574351829, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.4432, -0.4328, -0.4199, -0.4156, -0.3989, -0.3788, -0.3597, -0.3306, -0.3246, -0.3031, -0.3144, -0.2816, -0.2901, -0.2797, -0.2671, -0.2735, -0.2786, -0.2995, -0.2784, -0.2756, -0.2404, -0.2038, -0.198, -0.1936, -0.2009, -0.1859, -0.1917, -0.1711, -0.1673, -0.183, -0.1582, -0.116, -0.1075, -0.0884, -0.1006, -0.1007, -0.0981, -0.0539, -0.0264, -0.0408, -0.008, 0.0064, 0.0118, 0.0766, 0.0119, 0.0204, 0.0256, 0.0127, 0.059, 0.0597, 0.0624, 0.0473, 0.0103, 0.0433, -0.0069, -0.0198, -0.0347, -0.0765, -0.0479, -0.039, -0.0164, -0.0015, -0.0115, 0.0, 0.0196, 0.0122, -0.0024, 0.0244, 0.0619, 0.0591, 0.0864, 0.0632, 0.0482, 0.019, -0.0094, 0.0108, 0.0424, 0.0591, 0.1082, 0.1135, 0.1291, 0.1291, 0.1528, 0.2019, 0.2076, 0.1976, 0.1976, 0.2839, 0.323, 0.3876, 0.3991, 0.3831, 0.4014, 0.4049, 0.3834, 0.3882, 0.4091, 0.4662, 0.4725, 0.449, 0.4949, 0.5252, 0.5349, 0.501, 0.5946, 0.6644, 0.6718, 0.6719, 0.6212, 0.7002, 0.6457, 0.582, 0.5942, 0.6343, 0.6113, 0.6595, 0.7178, 0.7182, 0.7182, 0.7018, 0.7311, 0.764, 0.8135, 0.8143, 0.867, 0.9019, 0.9829, 0.9499, 0.9503, 0.7577, 0.655, 0.7714, 0.7125, 0.6354, 0.7656, 0.8048, 0.7638, 0.7656, 0.8443, 0.8867, 0.9886, 0.9112, 0.8741, 0.7589, 0.7909, 0.7755, 0.6705, 0.673, 0.5754, 0.5264, 0.6986, 0.6174, 0.6719, 0.717, 0.7467, 0.9064, 0.8836, 0.9069, 0.9066, 1.0183, 1.0407, 1.0755, 1.0508, 1.0601, 1.1125, 1.1597, 1.1763, 1.1786, 1.2773, 1.2458, 1.2744, 1.25, 1.2847, 1.4721, 1.5606, 1.8043, 1.8278, 1.9767, 2.221, 2.1316, 2.2872, 2.2851, 2.0352, 2.0234, 1.9699, 2.0177, 2.2588, 2.2207, 2.2207, 2.5123]}
{"unit_id": "thread:1995431391430098981", "ticker_idx": 1, "ts": "2025-12-01T09:54:51+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix, which dominate approximately 70% of the global DRAM market, recently expressed a cautious stance regarding the possibility of aggressive supply expansion", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM supply shortage persists until H1 2027+; Samsung and SK Hynix adopt conservative CAPEX/profitability strategies, supporting memory supercycle thesis beyond 2028.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"DRAM Supply Shortage Unlikely to Ease Until H1 2027\"\n\n“We will minimize oversupply risks.” (Samsung Electronics)\n\n“It is difficult to resolve the supply shortage until the first half of 2027.” (SK Hynix)\n\nSamsung Electronics and SK Hynix, which dominate approximately 70% of the global DRAM market, recently expressed a cautious stance regarding the possibility of aggressive supply expansion during investor relations meetings for global investment banks (IB). With semiconductor supply shortages spreading from server applications to PC and smartphone products, and considering both companies have announced production strategies centered on ‘profitability,’ the forecast that the ‘memory super cycle’ will continue beyond 2028 is gaining traction.\n\nAccording to industry sources on the 1st, Samsung Electronics unveiled a conservative semiconductor supply strategy at an Investor Relations (IR) event hosted by Morgan Stanley on the 23rd of last month. Samsung explained, “We will employ a strategy of maintaining long-term profitability rather than rapidly increasing facilities,” adding, “We will minimize oversupply risks through a Capital Expenditure (CAPEX) strategy that finds a balance between customer demand and price.”\n\nIt is reported that Samsung Electronics’ Memory Business Division is currently able to supply only about 70% of its DRAM orders. As the supply shortage intensifies, the division has reportedly refused requests for long-term mobile DRAM supply contracts from major clients. Samsung stated, “While clients want multi-year long-term contracts, Samsung does not want to tie up volumes with specific clients during a phase where prices are surging.”\n\nSK Hynix is in a similar position. At a recent IR event hosted by Bank of America Merrill Lynch, SK Hynix stated it would convert about half of its general-purpose DRAM production capacity to the latest 10-nanometer (nm) 6th generation (1c) DRAM. However, the company explained that this is insufficient to meet client demand. A company official explained, “We plan to spend about 30% of our revenue on CAPEX in 2026 and accelerate the transition to 1c DRAM, but it is difficult to resolve the supply shortage situation.”\n\nSK Hynix also anticipates that the upward trend in DRAM prices will continue for several years and is responding with ‘short-term contracts.’ An SK Hynix official explained, “We will lower the burden on clients through a strategy of phased contract price increases.”\n\n---\n\nhttps://t.co/nuIS5boUkR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002976244827510355, "ret_signed_m1d": -0.002976244827510355, "alpha_spy_m1d": -0.007562599043156482, "alpha_c_m1d": -0.0025221594915277867, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.00045408533598256806, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02579357553797723, "ret_signed_p1d": 0.02579357553797723, "alpha_spy_p1d": 0.023941366263865227, "alpha_c_p1d": 0.015491534515764682, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01030204102221255, "ret_p1w": 0.08630946109646254, "ret_signed_p1w": 0.08630946109646254, "alpha_spy_p1w": 0.08137029652071615, "alpha_c_p1w": 0.05519376840312207, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.03111569269334047, "ret_p1m": 0.1952663534648169, "ret_signed_p1m": 0.1952663534648169, "alpha_spy_p1m": 0.18237433806856984, "alpha_c_p1m": 0.1781156283389469, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.017150725125870014, "ret_p3m": 1.1732115801334566, "ret_signed_p3m": 1.1732115801334566, "alpha_spy_p3m": 1.1569434132196874, "alpha_c_p3m": 1.1868391630443735, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": -0.013627582910916813, "ret_p6m": 1.9868586647386945, "ret_signed_p6m": 1.9868586647386945, "alpha_spy_p6m": 1.8772053852851878, "alpha_c_p6m": 1.6901842707486512, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.29667439399004336, "price_path": [-0.3106, -0.3077, -0.3136, -0.3077, -0.2938, -0.283, -0.2751, -0.2553, -0.2444, -0.2158, -0.2277, -0.2069, -0.2069, -0.1753, -0.1635, -0.1565, -0.1496, -0.1773, -0.1647, -0.1677, -0.1468, -0.1096, -0.1096, -0.1096, -0.1096, -0.1096, -0.1096, -0.0635, -0.0744, -0.0913, -0.0575, -0.0308, -0.0288, -0.0268, -0.0327, -0.0218, -0.0427, -0.0198, 0.0119, -0.0129, -0.003, 0.0327, 0.0665, 0.1022, 0.0407, -0.002, -0.0159, -0.0288, -0.002, 0.0268, 0.0228, 0.0198, -0.0357, -0.002, -0.0298, -0.0427, -0.002, -0.0595, -0.0407, -0.0149, 0.0198, 0.0268, -0.003, 0.0, 0.0258, 0.0367, 0.0427, 0.0754, 0.0863, 0.0754, 0.0714, 0.0645, 0.0804, 0.0397, 0.0198, 0.0704, 0.0675, 0.0546, 0.0962, 0.1062, 0.1022, 0.1022, 0.1607, 0.1913, 0.1953, 0.1953, 0.1953, 0.281, 0.3767, 0.3847, 0.4056, 0.3837, 0.3857, 0.3837, 0.3717, 0.3986, 0.4345, 0.4844, 0.4884, 0.4475, 0.4903, 0.5183, 0.5163, 0.5163, 0.59, 0.6189, 0.602, 0.6, 0.4993, 0.6698, 0.6857, 0.588, 0.5811, 0.6588, 0.6528, 0.6728, 0.7804, 0.8064, 0.8064, 0.8064, 0.8064, 0.8941, 0.8951, 0.924, 0.9938, 1.0287, 1.1732, 1.1583, 1.1583, 0.9449, 0.7166, 0.91, 0.8761, 0.7296, 0.8731, 0.8941, 0.8731, 0.8293, 0.8811, 0.933, 1.0785, 0.9988, 0.9878, 0.8572, 0.8911, 0.8841, 0.7954, 0.7954, 0.7611, 0.6702, 0.8945, 0.7821, 0.86, 0.929, 0.9629, 1.1028, 1.0379, 1.0578, 1.0079, 1.0628, 1.1078, 1.1727, 1.1577, 1.1427, 1.1877, 1.1727, 1.2426, 1.1927, 1.2426, 1.2177, 1.2576, 1.2027, 1.2027, 1.3226, 1.3226, 1.6572, 1.7121, 1.6822, 1.852, 1.7871, 1.837, 1.9569, 1.7022, 1.807, 1.7521, 1.7571, 1.9919, 1.9219, 1.9219, 1.9869]}
{"unit_id": "thread:1995431391430098981", "ticker_idx": 2, "ts": "2025-12-01T09:54:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it is difficult to resolve the supply shortage until the first half of 2027", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM supply shortage persists until H1 2027+; Samsung and SK Hynix adopt conservative CAPEX/profitability strategies, supporting memory supercycle thesis beyond 2028.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"DRAM Supply Shortage Unlikely to Ease Until H1 2027\"\n\n“We will minimize oversupply risks.” (Samsung Electronics)\n\n“It is difficult to resolve the supply shortage until the first half of 2027.” (SK Hynix)\n\nSamsung Electronics and SK Hynix, which dominate approximately 70% of the global DRAM market, recently expressed a cautious stance regarding the possibility of aggressive supply expansion during investor relations meetings for global investment banks (IB). With semiconductor supply shortages spreading from server applications to PC and smartphone products, and considering both companies have announced production strategies centered on ‘profitability,’ the forecast that the ‘memory super cycle’ will continue beyond 2028 is gaining traction.\n\nAccording to industry sources on the 1st, Samsung Electronics unveiled a conservative semiconductor supply strategy at an Investor Relations (IR) event hosted by Morgan Stanley on the 23rd of last month. Samsung explained, “We will employ a strategy of maintaining long-term profitability rather than rapidly increasing facilities,” adding, “We will minimize oversupply risks through a Capital Expenditure (CAPEX) strategy that finds a balance between customer demand and price.”\n\nIt is reported that Samsung Electronics’ Memory Business Division is currently able to supply only about 70% of its DRAM orders. As the supply shortage intensifies, the division has reportedly refused requests for long-term mobile DRAM supply contracts from major clients. Samsung stated, “While clients want multi-year long-term contracts, Samsung does not want to tie up volumes with specific clients during a phase where prices are surging.”\n\nSK Hynix is in a similar position. At a recent IR event hosted by Bank of America Merrill Lynch, SK Hynix stated it would convert about half of its general-purpose DRAM production capacity to the latest 10-nanometer (nm) 6th generation (1c) DRAM. However, the company explained that this is insufficient to meet client demand. A company official explained, “We plan to spend about 30% of our revenue on CAPEX in 2026 and accelerate the transition to 1c DRAM, but it is difficult to resolve the supply shortage situation.”\n\nSK Hynix also anticipates that the upward trend in DRAM prices will continue for several years and is responding with ‘short-term contracts.’ An SK Hynix official explained, “We will lower the burden on clients through a strategy of phased contract price increases.”\n\n---\n\nhttps://t.co/nuIS5boUkR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01486978721660659, "ret_signed_m1d": -0.01486978721660659, "alpha_spy_m1d": -0.019456141432252716, "alpha_c_m1d": -0.012943233885823746, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.037174817256039994, "ret_signed_p1d": 0.037174817256039994, "alpha_spy_p1d": 0.03532260798192799, "alpha_c_p1d": 0.018844156246077715, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.07249076560395262, "ret_signed_p1w": 0.07249076560395262, "alpha_spy_p1w": 0.06755160102820623, "alpha_c_p1w": 0.028321497082765434, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.21003728679871325, "ret_signed_p1m": 0.21003728679871325, "alpha_spy_p1m": 0.19714527140246618, "alpha_c_p1m": 0.17751103357416698, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 1.0465204963316177, "ret_signed_p3m": 1.0465204963316177, "alpha_spy_p3m": 1.0302523294178485, "alpha_c_p3m": 0.8755889102672947, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.821164681653712, "ret_signed_p6m": 2.821164681653712, "alpha_spy_p6m": 2.7115114022002054, "alpha_c_p6m": 2.109884019355796, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.5124, -0.5069, -0.492, -0.4855, -0.4651, -0.4353, -0.4298, -0.3898, -0.3852, -0.3536, -0.3806, -0.3397, -0.3397, -0.3481, -0.3295, -0.336, -0.3378, -0.375, -0.3518, -0.3546, -0.3313, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.205, -0.2292, -0.2357, -0.2152, -0.1595, -0.1354, -0.0982, -0.1103, -0.1057, -0.1112, -0.0527, -0.0063, -0.0323, 0.0364, 0.055, 0.0383, 0.1516, 0.0884, 0.0754, 0.1014, 0.0773, 0.1256, 0.1497, 0.146, 0.1367, 0.0401, 0.1256, 0.0587, 0.0439, 0.0606, -0.0323, -0.0341, -0.036, -0.0267, 0.0112, -0.0149, 0.0, 0.0372, 0.026, 0.0074, 0.0112, 0.0725, 0.052, 0.0911, 0.0502, 0.0613, 0.0297, -0.0149, 0.0242, 0.026, 0.0167, 0.0781, 0.0855, 0.0929, 0.0929, 0.1134, 0.1896, 0.21, 0.21, 0.21, 0.2584, 0.2937, 0.3494, 0.3792, 0.4052, 0.3829, 0.3922, 0.3717, 0.3792, 0.3922, 0.4052, 0.4201, 0.381, 0.3755, 0.4033, 0.4257, 0.368, 0.487, 0.5632, 0.6004, 0.6896, 0.5428, 0.6859, 0.6729, 0.5651, 0.5595, 0.6487, 0.6283, 0.5985, 0.6506, 0.6357, 0.6357, 0.6357, 0.6357, 0.6617, 0.7639, 0.7677, 0.868, 0.8922, 1.0465, 0.9758, 0.9758, 0.7486, 0.581, 0.7523, 0.7206, 0.5568, 0.7467, 0.7784, 0.7318, 0.6946, 0.8137, 0.8063, 0.9664, 0.8864, 0.8752, 0.7374, 0.8361, 0.8529, 0.7374, 0.7374, 0.6257, 0.5028, 0.6704, 0.5456, 0.6313, 0.6499, 0.7057, 0.9236, 0.8584, 0.9124, 0.9367, 1.054, 1.1154, 1.1508, 1.1005, 1.1713, 1.2793, 1.2774, 1.2812, 1.2756, 1.4059, 1.4208, 1.4078, 1.3947, 1.3947, 1.6946, 1.6946, 1.9813, 2.08, 2.1396, 2.5009, 2.4171, 2.6796, 2.6685, 2.3873, 2.4264, 2.2495, 2.2495, 2.6126, 2.6145, 2.6145, 2.8212]}
{"unit_id": "thread:1991076801028984912", "ticker_idx": 0, "ts": "2025-11-19T09:31:16+00:00", "symbol": "Doosan", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Doosan may become the sole supplier of CCL for Rubin, Nvidia's new chip scheduled for release next year... sales from Nvidia to exceed 1 trillion won next year... operating profit at Doosan's Electronic BG is projected at 500.9 billion won this year, roughly four times last year's", "tweet_type": "analysis", "timeline": "quarters", "summary": "Doosan locks in 1T KRW Nvidia CCL sales as sole GB300/Rubin supplier after EMC fails GB300 qualification; EMC structurally displaced from Nvidia AI chip supply chain.", "resolved_tickers": ["006280.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Electronics CCL supplier context 006280.KS", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Doosan semiconductor materials, 1 trillion won in Nvidia sales ‘locked in’, EMC fails GB300 qualification\n\nDoosan’s Electronic Business Group (BG), which has been enjoying the AI semiconductor boom thanks to its high-quality copper-clad laminates (CCL), is expected to generate more than 1 trillion won in sales from Nvidia alone next year. This is because Taiwanese EMC, a competitor and the world’s No. 1 CCL maker, recently failed to clear Nvidia’s quality verification for the Blackwell GB300, raising the possibility that Doosan could become the sole supplier of CCL for Rubin, Nvidia’s next-generation AI chip, thereby solidifying its lead.\n\nAccording to industry sources on the 19th, Taiwanese CCL company EMC recently failed the quality verification for CCL used in the computing tray (GPU connection substrate) for Nvidia’s next-generation Blackwell model GB300. The CCL for the GB300 computing tray is a product from Doosan that has already passed Nvidia’s quality verification.\n\nCCL is a product in which both sides of an insulating layer composed of resin and reinforcement materials are bonded to thin copper foil that serves as a conductor. PCB (printed circuit board) manufacturers drill holes into the CCL and, through processes such as electrolytic and electroless plating, add copper layers to the circuitry to create substrates through which electrical signals can travel. In this sense, CCL is effectively the most basic material for semiconductors.\n\nThe reason EMC failed to pass GB300 computing tray quality verification is that there is a very high technological entry barrier for high-end CCL used in AI semiconductors. EMC holds a 60% share of the global CCL market and is the No. 1 player. It supplied CCL for Nvidia’s Hopper series launched in 2022 alongside Doosan, but in the supply chain for the next-generation AI semiconductor Blackwell it has been completely pushed out.\n\nThere is also talk that Doosan may become the sole supplier of CCL for Rubin, Nvidia’s new chip scheduled for release next year. In effect, the CCL market is split between EMC and Doosan, and EMC has failed to meet even the GB300 quality requirements.\n\nAn industry insider said, “At this point, the likelihood that Doosan will supply CCL exclusively in the early stages of Rubin is very high,” adding, “Starting from the third quarter of next year, Doosan is expected to begin exclusive supply, and in the fourth quarter, as chip production scales up, a sharp increase in sales can be expected.”\n\nThe industry expects Doosan’s sales from Nvidia to exceed 1 trillion won next year. Doosan is already projected to have generated 660 billion won in sales from Nvidia this year, up from 100 billion won in the fourth quarter of last year, and there is analysis that this could grow a further 70% to reach 1.15 trillion won next year.\n\nDoosan is enjoying an unprecedented boom as it expands supply of high-end CCL. Operating profit at Doosan’s Electronic BG is projected at 500.9 billion won this year, roughly four times last year’s 122.6 billion won. In particular, its operating margin has soared vertically from 12.2% to 27.4%. Next year as well, Doosan Electronic BG is expected to continue its growth, recording 717.5 billion won in operating profit with its profit margin approaching 30%. There are also expectations that Doosan will be able to supply CCL not only for Nvidia but also for custom ASICs ordered by global big tech companies such as Amazon and Google.\n\nDoosan’s CCL plants are running in “full operation” mode, fueling the company’s sales growth. According to Doosan’s quarterly report, the utilization rates at the Electronic BG’s Jeungpyeong and Gimcheon plants are 130.4% and 125.2%, respectively, exceeding 100%. At both plants, utilization has continued to rise every quarter this year. As of the third quarter, Doosan has invested 60.3 billion won in expanding its CCL facilities, more than a 50% increase from last year.\n\n---\n\nhttps://t.co/kvuZR8vvFk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0120846184333252, "ret_signed_m1d": 0.0120846184333252, "alpha_spy_m1d": 0.015932951338067602, "alpha_c_m1d": 0.0007109018408557599, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": 0.01137371659246944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011329381930227589, "ret_signed_p1d": 0.011329381930227589, "alpha_spy_p1d": 0.02657171861299301, "alpha_c_p1d": 0.03374543811495245, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.022416056184724864, "ret_p1w": 0.01435044714029865, "ret_signed_p1w": 0.01435044714029865, "alpha_spy_p1w": -0.011380294242985478, "alpha_c_p1w": 0.0034183655569521854, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.010932081583346465, "ret_p1m": 0.2575529312199585, "ret_signed_p1m": 0.2575529312199585, "alpha_spy_p1m": 0.23666662614571354, "alpha_c_p1m": 0.24062479515054958, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.016928136069408906, "ret_p3m": 0.3164653119850287, "ret_signed_p3m": 0.3164653119850287, "alpha_spy_p3m": 0.2845705497121973, "alpha_c_p3m": -0.19611228424221783, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": 0.5125775962272465, "ret_p6m": 0.10462874392464316, "ret_signed_p6m": 0.10462874392464316, "alpha_spy_p6m": -0.030891805096650238, "alpha_c_p6m": -1.0464221441088655, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 1.1510508880335086, "price_path": [-0.0136, 0.0136, 0.0053, 0.0053, -0.0045, -0.0204, -0.0325, -0.0264, -0.0098, -0.0038, -0.0106, 0.0106, 0.0015, 0.0045, 0.0068, 0.0121, 0.0174, 0.006, -0.0053, -0.0083, -0.0083, 0.0536, 0.0665, 0.0415, 0.034, -0.0053, -0.0121, -0.0242, -0.0234, -0.0008, -0.0008, -0.0008, -0.0008, -0.0008, -0.0008, -0.0166, -0.0219, -0.0279, -0.0136, -0.0227, -0.0317, -0.0128, -0.0128, -0.0015, -0.0272, -0.0272, 0.0023, 0.0068, -0.0098, -0.0257, -0.0204, -0.0128, -0.0083, -0.006, 0.0023, -0.0302, -0.0038, -0.0068, 0.034, 0.0521, 0.0529, 0.0461, 0.0121, 0.0, 0.0113, -0.0136, -0.0144, -0.0204, 0.0144, 0.0144, 0.0242, 0.0363, 0.0385, 0.0423, 0.0363, 0.0227, 0.0053, 0.0098, 0.0204, 0.1412, 0.2236, 0.2606, 0.2447, 0.2485, 0.2576, 0.2613, 0.2394, 0.2062, 0.1843, 0.1843, 0.1579, 0.1276, 0.2009, 0.2009, 0.2009, 0.2039, 0.216, 0.216, 0.2153, 0.1956, 0.1813, 0.2085, 0.216, 0.2304, 0.2878, 0.2107, 0.1971, 0.1971, 0.1397, 0.1533, 0.2198, 0.2319, 0.3029, 0.2847, 0.2787, 0.2455, 0.1835, 0.2334, 0.2372, 0.284, 0.2409, 0.2795, 0.2772, 0.3202, 0.3618, 0.3165, 0.3165, 0.3165, 0.3165, 0.3406, 0.3036, 0.3074, 0.3361, 0.2734, 0.2576, 0.2772, 0.2772, 0.2243, 0.0476, 0.1299, 0.1488, 0.0891, 0.1443, 0.1435, 0.1269, 0.1495, 0.1292, 0.1201, 0.1397, 0.0967, 0.1095, 0.0785, 0.0914, 0.1254, 0.1201, 0.1405, 0.1062, 0.0901, 0.1329, 0.0764, 0.0726, 0.068, 0.0558, 0.0878, 0.1008, 0.1008, 0.0856, 0.0985, 0.129, 0.1412, 0.1535, 0.1268, 0.0985, 0.0856, 0.0848, 0.0901, 0.0802, 0.0688, 0.0619, 0.0459, 0.0459, 0.0489, 0.0489, 0.036, 0.0421, 0.0535, 0.0238, 0.0474, 0.0543, 0.1046]}
{"unit_id": "thread:1991076801028984912", "ticker_idx": 1, "ts": "2025-11-19T09:31:16+00:00", "symbol": "EMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwanese EMC, a competitor and the world's No. 1 CCL maker, recently failed to clear Nvidia's quality verification for the Blackwell GB300... in the supply chain for the next-generation AI semiconductor Blackwell it has been completely pushed out", "tweet_type": "analysis", "timeline": "quarters", "summary": "Doosan locks in 1T KRW Nvidia CCL sales as sole GB300/Rubin supplier after EMC fails GB300 qualification; EMC structurally displaced from Nvidia AI chip supply chain.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "CCL context: EMC = Elite Material Co. 2383.TW not US EMC", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Doosan semiconductor materials, 1 trillion won in Nvidia sales ‘locked in’, EMC fails GB300 qualification\n\nDoosan’s Electronic Business Group (BG), which has been enjoying the AI semiconductor boom thanks to its high-quality copper-clad laminates (CCL), is expected to generate more than 1 trillion won in sales from Nvidia alone next year. This is because Taiwanese EMC, a competitor and the world’s No. 1 CCL maker, recently failed to clear Nvidia’s quality verification for the Blackwell GB300, raising the possibility that Doosan could become the sole supplier of CCL for Rubin, Nvidia’s next-generation AI chip, thereby solidifying its lead.\n\nAccording to industry sources on the 19th, Taiwanese CCL company EMC recently failed the quality verification for CCL used in the computing tray (GPU connection substrate) for Nvidia’s next-generation Blackwell model GB300. The CCL for the GB300 computing tray is a product from Doosan that has already passed Nvidia’s quality verification.\n\nCCL is a product in which both sides of an insulating layer composed of resin and reinforcement materials are bonded to thin copper foil that serves as a conductor. PCB (printed circuit board) manufacturers drill holes into the CCL and, through processes such as electrolytic and electroless plating, add copper layers to the circuitry to create substrates through which electrical signals can travel. In this sense, CCL is effectively the most basic material for semiconductors.\n\nThe reason EMC failed to pass GB300 computing tray quality verification is that there is a very high technological entry barrier for high-end CCL used in AI semiconductors. EMC holds a 60% share of the global CCL market and is the No. 1 player. It supplied CCL for Nvidia’s Hopper series launched in 2022 alongside Doosan, but in the supply chain for the next-generation AI semiconductor Blackwell it has been completely pushed out.\n\nThere is also talk that Doosan may become the sole supplier of CCL for Rubin, Nvidia’s new chip scheduled for release next year. In effect, the CCL market is split between EMC and Doosan, and EMC has failed to meet even the GB300 quality requirements.\n\nAn industry insider said, “At this point, the likelihood that Doosan will supply CCL exclusively in the early stages of Rubin is very high,” adding, “Starting from the third quarter of next year, Doosan is expected to begin exclusive supply, and in the fourth quarter, as chip production scales up, a sharp increase in sales can be expected.”\n\nThe industry expects Doosan’s sales from Nvidia to exceed 1 trillion won next year. Doosan is already projected to have generated 660 billion won in sales from Nvidia this year, up from 100 billion won in the fourth quarter of last year, and there is analysis that this could grow a further 70% to reach 1.15 trillion won next year.\n\nDoosan is enjoying an unprecedented boom as it expands supply of high-end CCL. Operating profit at Doosan’s Electronic BG is projected at 500.9 billion won this year, roughly four times last year’s 122.6 billion won. In particular, its operating margin has soared vertically from 12.2% to 27.4%. Next year as well, Doosan Electronic BG is expected to continue its growth, recording 717.5 billion won in operating profit with its profit margin approaching 30%. There are also expectations that Doosan will be able to supply CCL not only for Nvidia but also for custom ASICs ordered by global big tech companies such as Amazon and Google.\n\nDoosan’s CCL plants are running in “full operation” mode, fueling the company’s sales growth. According to Doosan’s quarterly report, the utilization rates at the Electronic BG’s Jeungpyeong and Gimcheon plants are 130.4% and 125.2%, respectively, exceeding 100%. At both plants, utilization has continued to rise every quarter this year. As of the third quarter, Doosan has invested 60.3 billion won in expanding its CCL facilities, more than a 50% increase from last year.\n\n---\n\nhttps://t.co/kvuZR8vvFk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015037593984962516, "ret_signed_m1d": -0.015037593984962516, "alpha_spy_m1d": -0.018885926889704918, "alpha_c_m1d": -0.02194226476411698, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.006904670779154465, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07518796992481214, "ret_signed_p1d": -0.07518796992481214, "alpha_spy_p1d": -0.09043030660757756, "alpha_c_p1d": -0.10657925319046924, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.0313912832656571, "ret_p1w": 0.10526315789473695, "ret_signed_p1w": -0.10526315789473695, "alpha_spy_p1w": -0.07953241651145282, "alpha_c_p1w": -0.09526214062799543, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.010001017266741519, "ret_p1m": 0.1278195488721805, "ret_signed_p1m": -0.1278195488721805, "alpha_spy_p1m": -0.10693324379793556, "alpha_c_p1m": -0.120309963801561, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.0075095850706194955, "ret_p3m": 0.650375939849624, "ret_signed_p3m": -0.650375939849624, "alpha_spy_p3m": -0.6184811775767927, "alpha_c_p3m": -0.6554535603516223, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": -0.005077620501998226, "ret_p6m": 2.699248120300752, "ret_signed_p6m": -2.699248120300752, "alpha_spy_p6m": -2.5637275712794585, "alpha_c_p6m": -2.418002619854527, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.2812455004462251, "price_path": [-0.1057, -0.0723, -0.0723, -0.0501, -0.0602, -0.0827, -0.1316, -0.1353, -0.1241, -0.1617, -0.1241, -0.1391, -0.0789, -0.0564, -0.0038, 0.0113, -0.0113, -0.0338, -0.0113, 0.0038, -0.0301, -0.0226, 0.0113, -0.0263, -0.0564, -0.0789, -0.0789, -0.0789, -0.0752, -0.0263, -0.0188, -0.0188, -0.0714, -0.1015, -0.0714, -0.0714, -0.0977, -0.1203, -0.1429, -0.1429, -0.1692, -0.1541, -0.1504, -0.1579, -0.1617, -0.1617, -0.1241, -0.0564, -0.0602, -0.0414, 0.0226, 0.0602, 0.1053, 0.0827, 0.0902, 0.0301, 0.0226, -0.0075, 0.0113, 0.0338, 0.0376, 0.0902, 0.015, 0.0, 0.0752, -0.015, -0.0075, 0.0902, 0.1053, 0.1241, 0.1466, 0.109, 0.0827, 0.0752, 0.0714, 0.1241, 0.1165, 0.1278, 0.203, 0.1992, 0.2068, 0.2068, 0.1241, 0.1241, 0.1278, 0.1654, 0.203, 0.2143, 0.2707, 0.2707, 0.2331, 0.2594, 0.2368, 0.2368, 0.2368, 0.1955, 0.218, 0.2744, 0.188, 0.1617, 0.1767, 0.2105, 0.2293, 0.2556, 0.1805, 0.2218, 0.188, 0.2143, 0.1729, 0.2481, 0.3534, 0.3083, 0.3684, 0.4286, 0.3459, 0.312, 0.3421, 0.3947, 0.5338, 0.4662, 0.4474, 0.5639, 0.5827, 0.6504, 0.6504, 0.6504, 0.6504, 0.6504, 0.6504, 0.6504, 0.6504, 0.6353, 0.7368, 0.8195, 0.8346, 0.8346, 0.7895, 0.7444, 0.6466, 0.7782, 0.7744, 0.6353, 0.7857, 0.8421, 0.9549, 1.0338, 0.9436, 1.0301, 1.0301, 1.1729, 1.1128, 1.1015, 1.0188, 1.1391, 1.2143, 1.188, 1.094, 0.9549, 1.1504, 1.094, 1.094, 1.094, 1.1692, 1.3835, 1.4211, 1.5714, 1.5, 1.7105, 1.8647, 1.8835, 1.8571, 1.8947, 1.9699, 2.0564, 2.0827, 2.3647, 2.5564, 2.2895, 2.3383, 2.4173, 2.4173, 2.5752, 2.5564, 2.7218, 2.7293, 2.5714, 2.6466, 2.6842, 2.7068, 2.6992]}
{"unit_id": "orig:1980150867488821474", "ticker_idx": 0, "ts": "2025-10-20T05:55:30+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The shortage in the DRAM market will deepen through 2026, and the supply-demand imbalance will remain extremely tight", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Micron CBO flags deepening DRAM shortage through 2026, HBM supply tight, 40–47% revenue growth guided; relayed bullish thesis adopted by sharing.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron Technology Chief Business Officer (CBO) Sumit Sadana: The shortage in the DRAM market will deepen through 2026, and the supply-demand imbalance will remain extremely tight.\n\nAccording to Nikkei Asia, Sadana noted that high-bandwidth memory (HBM) used in AI applications requires about three times the wafer input compared to standard DRAM.\n\nDespite Micron allocating significant production capacity to HBM products, clear supply constraints have emerged as new fabs take a long time to become operational and costs continue to rise. Micron’s HBM sales reached approximately USD 2 billion in the second quarter of 2025 (June–August), and the company expects revenue for the third quarter of 2025 (September–November) to grow by 40–47% year over year, driven mainly by HBM demand.\n\nHBM4 sampling begins — preparation for 2026 commercialization\n\nSadana said that HBM supply contracts are expected to be finalized within a few months ahead of 2026. Micron has already started sampling its next-generation HBM4 product, and while shipment volumes will likely remain limited in 2026, he mentioned that it offers lower power consumption than competitors’ products.\n\nRegarding the recent increase in DDR4 DRAM prices, Micron is struggling to fully meet demand due to limited production capacity. Sadana stated that while the company plans to gradually phase out DDR4 production, it will extend the production period to support key long-term customers through its U.S. fabs. However, he added that overall capacity may still fall short of market demand.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021231312112164225, "ret_signed_m1d": -0.021231312112164225, "alpha_spy_m1d": -0.010937893163100232, "alpha_c_m1d": -0.008587969585019173, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021666697809607394, "ret_signed_p1d": -0.021666697809607394, "alpha_spy_p1d": -0.021651701889594865, "alpha_c_p1d": -0.016137009767945742, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.06446773378383996, "ret_signed_p1w": 0.06446773378383996, "alpha_spy_p1w": 0.043702042108002104, "alpha_c_p1w": 0.027833903166507357, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": 0.10509257983862152, "ret_signed_p1m": 0.10509257983862152, "alpha_spy_p1m": 0.12180635563937625, "alpha_c_p1m": 0.14521120716409697, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.628698498100293, "ret_signed_p3m": 0.628698498100293, "alpha_spy_p3m": 0.5944582407849357, "alpha_c_p3m": 0.48347797534023096, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 1.2539237335723143, "ret_signed_p6m": 1.2539237335723143, "alpha_spy_p6m": 1.2135330459654914, "alpha_c_p6m": 0.9481048301503614, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.4692, -0.46, -0.4623, -0.4623, -0.4589, -0.4454, -0.4725, -0.4931, -0.4791, -0.4729, -0.4742, -0.4593, -0.4254, -0.402, -0.3826, -0.3994, -0.3944, -0.4158, -0.4029, -0.4101, -0.4335, -0.4404, -0.4312, -0.4373, -0.4369, -0.4309, -0.4103, -0.4248, -0.4248, -0.4274, -0.4262, -0.3997, -0.3651, -0.3646, -0.3463, -0.3233, -0.2723, -0.2401, -0.2375, -0.2324, -0.2267, -0.1837, -0.2135, -0.2043, -0.1957, -0.2184, -0.242, -0.2399, -0.2078, -0.1913, -0.1196, -0.1119, -0.0916, -0.0765, -0.1019, -0.0495, -0.0698, -0.1217, -0.0677, -0.0953, -0.0717, -0.0205, -0.0212, 0.0, -0.0217, -0.0401, -0.0003, 0.0592, 0.0645, 0.0732, 0.096, 0.0834, 0.0822, 0.1351, 0.0545, 0.1486, 0.1526, 0.1507, 0.225, 0.1661, 0.1844, 0.146, 0.1937, 0.1701, 0.1051, 0.0926, -0.0261, 0.0029, 0.083, 0.0859, 0.1136, 0.1136, 0.1437, 0.1629, 0.1582, 0.1325, 0.0961, 0.1473, 0.1942, 0.2208, 0.2754, 0.25, 0.1662, 0.1486, 0.1245, 0.0907, 0.2021, 0.2861, 0.3377, 0.3361, 0.3865, 0.3865, 0.3773, 0.4242, 0.4158, 0.3809, 0.3809, 0.5261, 0.5103, 0.6616, 0.6428, 0.5822, 0.6696, 0.6734, 0.636, 0.6128, 0.6287, 0.7551, 0.7551, 0.766, 0.8826, 0.9236, 0.9336, 0.8825, 0.9848, 1.106, 1.1085, 1.0073, 1.1182, 1.0294, 0.8356, 0.8525, 0.9096, 0.8555, 0.8059, 0.9853, 1.0029, 0.9917, 0.9917, 0.9342, 1.0367, 1.0192, 1.0716, 1.0368, 1.0224, 1.0756, 1.0106, 0.9951, 0.9966, 0.837, 0.939, 0.921, 0.7916, 0.8836, 0.9503, 1.0257, 0.9612, 1.0617, 1.1375, 1.2338, 1.234, 1.1495, 1.0461, 0.9563, 0.9137, 0.8486, 0.7198, 0.7283, 0.5576, 0.6352, 0.7805, 0.7727, 0.7727, 0.8285, 0.8276, 0.9687, 1.0402, 1.0358, 1.0647, 1.2539]}
{"unit_id": "orig:1988450104777576548", "ticker_idx": 0, "ts": "2025-11-12T03:33:43+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The firm maintains Overweight ratings on SK hynix and Samsung Electronics, expecting rising memory prices to push share prices to new highs and for memory makers' profits to significantly beat expectations.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Overweight on SK Hynix and Samsung; AI-driven memory super cycle to push prices and profits well above historical peaks and consensus.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: This memory “super cycle” will far surpass the historical peak\n\nMorgan Stanley pointed out that a new memory “super cycle” driven by AI has already arrived, and its intensity and logic are completely different from any previous cycle.\n\nMorgan Stanley stated that, unlike in the past, this cycle is led by AI data centers and cloud service providers, whose customers have low price sensitivity, and inference workloads have become the main driver of general-purpose memory demand. Meanwhile, the latest channel checks indicate that in Q4, server DRAM contract prices have surged by nearly 70%, NAND contract prices are up 20–30%, and suppliers hold unprecedented pricing power.\n\nThe firm maintains Overweight ratings on SK hynix and Samsung Electronics, expecting rising memory prices to push share prices to new highs and for memory makers’ profits to significantly beat expectations.\n\nMorgan Stanley noted that the core drivers of this cycle have changed in quality. The buyers are no longer traditional customers sensitive to price, but AI data centers and cloud-service giants engaged in an arms race to build compute infrastructure.\n\nFor them, securing memory is a strategic “must-have,” and price sensitivity has fallen to the lowest level. At the same time, HBM (high-bandwidth memory) production continues to structurally cannibalize traditional DRAM capacity. As Morgan Stanley emphasized in the report:\n\n“The peculiar point is as follows: today’s memory demand has evolved into a competition led by AI data centers (compute-intensive platforms) and cloud service providers, and they are not as price-sensitive as traditional customers… The exponential upward trend in inference demand provides a solid foundation for this, and this is what makes this cycle fundamentally different from any previous one.”\n\nAccording to Morgan Stanley’s latest channel checks, the DRAM price outlook turned sharply bullish in just two weeks. Q4 server RDIMM contract prices jumped about 70%, far surpassing the previous 30% expectation. The DDR5 (16Gb) spot price soared 336%, from $7.50 in September to $20.90 now. DDR4 prices also saw a 50% increase in offer levels. Most contract transactions will not be finalized until the end of this month, but customer acceptance seems inevitable—because they fear further price hikes and supply constraints.\n\nFurthermore, NAND has fallen into a severe shortage:\n\nNAND has become a key component of AI compute infrastructure and video storage. Prices of 3D NAND wafers (TLC and QLC) are expected to rise 65–70% quarter-over-quarter to respond to capacity constraints. Nearline storage specifications are shifting from 128TB to 256TB QLC SSDs. TrendForce forecasts that enterprise SSD bit-based server demand in 2026 will increase by about 50% year-over-year. Samsung’s bit production in the first half of 2025 was constrained by the transition from V6 176-layer to 321-layer V8 NAND, and in the second half it will ramp only gradually, resulting in this year’s bit shipment growth rate being limited to 10%.\n\nThere is still ample room for price increases, and the cycle is not at its peak yet\nThe market is often constrained by “peak fear,” assuming that once stock prices hit new highs a reversal will soon follow. However, Morgan Stanley emphasizes that in this AI-led market, the ultimate determinant is earnings growth, not historical valuation:\n\nCurrently, server DRAM prices are $1/Gb, whereas the peak of the cloud super cycle in Q1 2018 was $1.25/Gb. Given the scale of AI infrastructure investment and the dynamic nature of hyperscale customers, the peak price of this cycle is highly likely to surpass the previous high. Memory cycles typically last 4–6 quarters, and profits are in the process of being realized. The key, however, is comparison with market expectations—the market is showing noticeably greater enthusiasm for general-purpose memory prices. Valuation is not a predictor of future returns; it reflects supply and demand, not historical precedents.\n\nBased on the strong long-term drivers of AI, memory price increases have already entered “no-man’s-land,” and earnings prospects far exceed the market’s general expectations. This implies there is still enormous upside potential in stock prices.\n\n“As AI-related capital expenditures continue to expand, the share of memory in total spending is likely to keep rising—this will support P/B to far exceed historical peaks. In our view, this is a story where multiple expansion is layered on top of cyclical earnings recovery…\n\nWe believe analysts’ data revisions are always lagging—for SK hynix and Samsung, our 2026 and 2027 earnings forecasts are 31~48% and 38~51% higher than market consensus, respectively.”\n\nIn sum, the drivers of this memory “super cycle” are more durable, the magnitude of price increases has surpassed historical records, and earnings prospects are higher still. Coupled with a strong cycle, this creates a rare investment opportunity for memory manufacturers that hold pricing power.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0032414031765386486, "ret_signed_m1d": 0.0032414031765386486, "alpha_spy_m1d": 0.003797492375725242, "alpha_c_m1d": 0.01584228579369229, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008103711087697363, "ret_signed_p1d": -0.008103711087697363, "alpha_spy_p1d": 0.008490310383040489, "alpha_c_p1d": 0.022020272573561428, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.08914102511102207, "ret_signed_p1w": -0.08914102511102207, "alpha_spy_p1w": -0.058777351218040175, "alpha_c_p1w": -0.04388470269168021, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": -0.08362296077883868, "ret_signed_p1m": -0.08362296077883868, "alpha_spy_p1m": -0.09209552899115969, "alpha_c_p1m": -0.12682613450156632, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.4386309601742753, "ret_signed_p3m": 0.4386309601742753, "alpha_spy_p3m": 0.42016306810534143, "alpha_c_p3m": 0.29128660834802345, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 1.6875831566891168, "ret_signed_p6m": 1.6875831566891168, "alpha_spy_p6m": 1.6109557536984331, "alpha_c_p6m": 1.1637157874256956, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.5525, -0.5671, -0.5744, -0.5865, -0.6035, -0.5938, -0.58, -0.5768, -0.5792, -0.5648, -0.564, -0.5851, -0.5778, -0.5746, -0.5697, -0.5567, -0.5511, -0.5332, -0.5073, -0.5024, -0.4676, -0.4635, -0.436, -0.4595, -0.4238, -0.4238, -0.4311, -0.4149, -0.4206, -0.4222, -0.4546, -0.4344, -0.4368, -0.4165, -0.359, -0.359, -0.359, -0.359, -0.359, -0.359, -0.3063, -0.3274, -0.3331, -0.3152, -0.2666, -0.2455, -0.2131, -0.2237, -0.2196, -0.2245, -0.1734, -0.1329, -0.1556, -0.0956, -0.0794, -0.094, 0.0049, -0.0502, -0.0616, -0.0389, -0.06, -0.0178, 0.0032, 0.0, -0.0081, -0.0924, -0.0178, -0.0762, -0.0891, -0.0746, -0.1556, -0.1572, -0.1588, -0.1507, -0.1177, -0.1404, -0.1274, -0.095, -0.1047, -0.1209, -0.1177, -0.0642, -0.082, -0.0479, -0.0836, -0.0739, -0.1015, -0.1404, -0.1063, -0.1047, -0.1128, -0.0593, -0.0528, -0.0463, -0.0463, -0.0285, 0.038, 0.0559, 0.0559, 0.0559, 0.098, 0.1288, 0.1775, 0.2035, 0.2262, 0.2067, 0.2148, 0.197, 0.2035, 0.2148, 0.2262, 0.2391, 0.2051, 0.2002, 0.2245, 0.244, 0.1937, 0.2975, 0.364, 0.3965, 0.4743, 0.3462, 0.4711, 0.4597, 0.3656, 0.3608, 0.4386, 0.4208, 0.3948, 0.4403, 0.4273, 0.4273, 0.4273, 0.4273, 0.45, 0.5392, 0.5424, 0.63, 0.6511, 0.7858, 0.724, 0.724, 0.5258, 0.3795, 0.529, 0.5014, 0.3584, 0.5242, 0.5518, 0.5112, 0.4787, 0.5827, 0.5762, 0.7159, 0.646, 0.6363, 0.516, 0.6022, 0.6168, 0.516, 0.516, 0.4185, 0.3113, 0.4575, 0.3487, 0.4234, 0.4397, 0.4884, 0.6785, 0.6216, 0.6688, 0.6899, 0.7923, 0.8459, 0.8768, 0.8329, 0.8946, 0.9889, 0.9873, 0.9905, 0.9856, 1.0994, 1.1124, 1.101, 1.0896, 1.0896, 1.3512, 1.3512, 1.6015, 1.6876]}
{"unit_id": "orig:1988450104777576548", "ticker_idx": 1, "ts": "2025-11-12T03:33:43+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The firm maintains Overweight ratings on SK hynix and Samsung Electronics, expecting rising memory prices to push share prices to new highs and for memory makers' profits to significantly beat expectations.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley Overweight on SK Hynix and Samsung; AI-driven memory super cycle to push prices and profits well above historical peaks and consensus.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: This memory “super cycle” will far surpass the historical peak\n\nMorgan Stanley pointed out that a new memory “super cycle” driven by AI has already arrived, and its intensity and logic are completely different from any previous cycle.\n\nMorgan Stanley stated that, unlike in the past, this cycle is led by AI data centers and cloud service providers, whose customers have low price sensitivity, and inference workloads have become the main driver of general-purpose memory demand. Meanwhile, the latest channel checks indicate that in Q4, server DRAM contract prices have surged by nearly 70%, NAND contract prices are up 20–30%, and suppliers hold unprecedented pricing power.\n\nThe firm maintains Overweight ratings on SK hynix and Samsung Electronics, expecting rising memory prices to push share prices to new highs and for memory makers’ profits to significantly beat expectations.\n\nMorgan Stanley noted that the core drivers of this cycle have changed in quality. The buyers are no longer traditional customers sensitive to price, but AI data centers and cloud-service giants engaged in an arms race to build compute infrastructure.\n\nFor them, securing memory is a strategic “must-have,” and price sensitivity has fallen to the lowest level. At the same time, HBM (high-bandwidth memory) production continues to structurally cannibalize traditional DRAM capacity. As Morgan Stanley emphasized in the report:\n\n“The peculiar point is as follows: today’s memory demand has evolved into a competition led by AI data centers (compute-intensive platforms) and cloud service providers, and they are not as price-sensitive as traditional customers… The exponential upward trend in inference demand provides a solid foundation for this, and this is what makes this cycle fundamentally different from any previous one.”\n\nAccording to Morgan Stanley’s latest channel checks, the DRAM price outlook turned sharply bullish in just two weeks. Q4 server RDIMM contract prices jumped about 70%, far surpassing the previous 30% expectation. The DDR5 (16Gb) spot price soared 336%, from $7.50 in September to $20.90 now. DDR4 prices also saw a 50% increase in offer levels. Most contract transactions will not be finalized until the end of this month, but customer acceptance seems inevitable—because they fear further price hikes and supply constraints.\n\nFurthermore, NAND has fallen into a severe shortage:\n\nNAND has become a key component of AI compute infrastructure and video storage. Prices of 3D NAND wafers (TLC and QLC) are expected to rise 65–70% quarter-over-quarter to respond to capacity constraints. Nearline storage specifications are shifting from 128TB to 256TB QLC SSDs. TrendForce forecasts that enterprise SSD bit-based server demand in 2026 will increase by about 50% year-over-year. Samsung’s bit production in the first half of 2025 was constrained by the transition from V6 176-layer to 321-layer V8 NAND, and in the second half it will ramp only gradually, resulting in this year’s bit shipment growth rate being limited to 10%.\n\nThere is still ample room for price increases, and the cycle is not at its peak yet\nThe market is often constrained by “peak fear,” assuming that once stock prices hit new highs a reversal will soon follow. However, Morgan Stanley emphasizes that in this AI-led market, the ultimate determinant is earnings growth, not historical valuation:\n\nCurrently, server DRAM prices are $1/Gb, whereas the peak of the cloud super cycle in Q1 2018 was $1.25/Gb. Given the scale of AI infrastructure investment and the dynamic nature of hyperscale customers, the peak price of this cycle is highly likely to surpass the previous high. Memory cycles typically last 4–6 quarters, and profits are in the process of being realized. The key, however, is comparison with market expectations—the market is showing noticeably greater enthusiasm for general-purpose memory prices. Valuation is not a predictor of future returns; it reflects supply and demand, not historical precedents.\n\nBased on the strong long-term drivers of AI, memory price increases have already entered “no-man’s-land,” and earnings prospects far exceed the market’s general expectations. This implies there is still enormous upside potential in stock prices.\n\n“As AI-related capital expenditures continue to expand, the share of memory in total spending is likely to keep rising—this will support P/B to far exceed historical peaks. In our view, this is a story where multiple expansion is layered on top of cyclical earnings recovery…\n\nWe believe analysts’ data revisions are always lagging—for SK hynix and Samsung, our 2026 and 2027 earnings forecasts are 31~48% and 38~51% higher than market consensus, respectively.”\n\nIn sum, the drivers of this memory “super cycle” are more durable, the magnitude of price increases has surpassed historical records, and earnings prospects are higher still. Coupled with a strong cycle, this creates a rare investment opportunity for memory manufacturers that hold pricing power.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0038797231644660535, "ret_signed_m1d": 0.0038797231644660535, "alpha_spy_m1d": 0.004435812363652647, "alpha_c_m1d": 0.0069758299120047385, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.003096106747538685, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0029097732978722313, "ret_signed_p1d": -0.0029097732978722313, "alpha_spy_p1d": 0.01368424817286562, "alpha_c_p1d": 0.022029102618702723, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.024938875916574954, "ret_p1w": -0.06401547036464372, "ret_signed_p1w": -0.06401547036464372, "alpha_spy_p1w": -0.03365179647166183, "alpha_c_p1w": -0.019955842120262357, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.044059628244381366, "ret_p1m": 0.040737131377847735, "ret_signed_p1m": 0.040737131377847735, "alpha_spy_p1m": 0.03226456316552673, "alpha_c_p1m": 0.033864517788774195, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.00687261358907354, "ret_p3m": 0.6218127929611794, "ret_signed_p3m": 0.6218127929611794, "alpha_spy_p3m": 0.6033449008922456, "alpha_c_p3m": 0.6448977738396983, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": -0.023084980878518846, "ret_p6m": 1.651643939888746, "ret_signed_p6m": 1.651643939888746, "alpha_spy_p6m": 1.5750165368980622, "alpha_c_p6m": 1.493786849312865, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.15785709057588093, "price_path": [-0.3086, -0.3241, -0.3241, -0.3192, -0.3183, -0.3105, -0.3096, -0.3212, -0.3183, -0.3279, -0.327, -0.3472, -0.3328, -0.326, -0.3231, -0.3289, -0.3231, -0.3096, -0.299, -0.2912, -0.2719, -0.2613, -0.2333, -0.2449, -0.2246, -0.2246, -0.1937, -0.1821, -0.1754, -0.1686, -0.1956, -0.1833, -0.1862, -0.1659, -0.1295, -0.1295, -0.1295, -0.1295, -0.1295, -0.1295, -0.0844, -0.0951, -0.1115, -0.0786, -0.0524, -0.0504, -0.0485, -0.0543, -0.0436, -0.064, -0.0417, -0.0107, -0.0349, -0.0252, 0.0097, 0.0427, 0.0776, 0.0175, -0.0242, -0.0378, -0.0504, -0.0242, 0.0039, 0.0, -0.0029, -0.0572, -0.0242, -0.0514, -0.064, -0.0242, -0.0805, -0.0621, -0.0369, -0.0029, 0.0039, -0.0252, -0.0223, 0.0029, 0.0136, 0.0194, 0.0514, 0.0621, 0.0514, 0.0475, 0.0407, 0.0563, 0.0165, -0.0029, 0.0466, 0.0436, 0.031, 0.0718, 0.0815, 0.0776, 0.0776, 0.1348, 0.1647, 0.1686, 0.1686, 0.1686, 0.2524, 0.346, 0.3538, 0.3743, 0.3528, 0.3548, 0.3528, 0.3411, 0.3674, 0.4025, 0.4512, 0.4551, 0.4152, 0.4571, 0.4844, 0.4824, 0.4824, 0.5546, 0.5828, 0.5663, 0.5643, 0.4659, 0.6325, 0.6481, 0.5526, 0.5458, 0.6218, 0.616, 0.6355, 0.7407, 0.7661, 0.7661, 0.7661, 0.7661, 0.8518, 0.8528, 0.8811, 0.9493, 0.9834, 1.1247, 1.1101, 1.1101, 0.9015, 0.6783, 0.8674, 0.8343, 0.691, 0.8314, 0.8518, 0.8314, 0.7885, 0.8392, 0.8898, 1.0321, 0.9542, 0.9434, 0.8158, 0.8489, 0.8421, 0.7553, 0.7553, 0.7219, 0.633, 0.8522, 0.7424, 0.8185, 0.8859, 0.9191, 1.0559, 0.9924, 1.0119, 0.9631, 1.0168, 1.0608, 1.1242, 1.1096, 1.0949, 1.1389, 1.1242, 1.1926, 1.1438, 1.1926, 1.1682, 1.2073, 1.1535, 1.1535, 1.2707, 1.2707, 1.5979, 1.6516]}
{"unit_id": "reply:2001491736510841239", "ticker_idx": 0, "ts": "2025-12-18T03:16:30+00:00", "symbol": "Western semiconductor equipment companies", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I believe valuation discounts and downside share price pressure could begin to apply to Western semiconductor-related companies", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Western semis/equipment: China's domestic EUV timeline is now concrete, warranting valuation discounts and share price pressure.", "resolved_tickers": ["AMAT", "LRCX", "KLAC", "ASML", "ONTO"], "resolution_method": "basket", "resolution_reason": "Western semicon equipment basket: AMAT, Lam, KLA, ASML, Onto", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "I want to argue that we now have a visible, concrete timeline for China’s EUV.\n\nOf course, Western companies won’t collapse overnight.\n\nBut we’re playing a massive chess game here, right? I think it’s time we start treating China’s domestic EUV as a serious risk factor.\n\nIn other words, I believe valuation discounts and downside share price pressure could begin to apply to Western semiconductor-related companies.", "reply_text": null, "parent_text": "@jukan05 Don’t be confused. They haven’t figured it out.", "quoted_text": null, "parent_author": "evrgn11112231", "ret_m1d": -0.033723401362047746, "ret_signed_m1d": 0.033723401362047746, "alpha_spy_m1d": 0.02622866403854438, "alpha_c_m1d": 0.01065640341293337, "bench_spy_m1d": -0.007494737323503364, "bench_c_m1d": -0.023066997949114376, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025204105482670425, "ret_signed_p1d": -0.025204105482670425, "alpha_spy_p1d": -0.01614066000321861, "alpha_c_p1d": 0.0006562368205841988, "bench_spy_p1d": 0.009063445479451815, "bench_c_p1d": 0.025860342303254624, "ret_p1w": 0.04900386774478327, "ret_signed_p1w": -0.04900386774478327, "alpha_spy_p1w": -0.02542551631659018, "alpha_c_p1w": 0.002985078771390981, "bench_spy_p1w": 0.02357835142819309, "bench_c_p1w": 0.05198894651617425, "ret_p1m": 0.3377852812175156, "ret_signed_p1m": -0.3377852812175156, "alpha_spy_p1m": -0.3123091994345383, "alpha_c_p1m": -0.18116652026840974, "bench_spy_p1m": 0.025476081782977333, "bench_c_p1m": 0.15661876094910587, "ret_p3m": 0.3240534738521664, "ret_signed_p3m": -0.3240534738521664, "alpha_spy_p3m": -0.32951990637407796, "alpha_c_p3m": -0.17757418441211176, "bench_spy_p3m": -0.005466432521911591, "bench_c_p3m": 0.14647928944005462, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1511, -0.1414, -0.1562, -0.1629, -0.1585, -0.1485, -0.1434, -0.0969, -0.0769, -0.0837, -0.0642, -0.1149, -0.1087, -0.1146, -0.1746, -0.1324, -0.1317, -0.0978, -0.0934, -0.0956, -0.0778, -0.0843, -0.1128, -0.0784, -0.0662, -0.0479, -0.0601, -0.0305, -0.0351, -0.0458, -0.0387, -0.066, -0.0294, -0.0501, -0.0511, -0.0278, -0.0592, -0.048, -0.0808, -0.0941, -0.091, -0.108, -0.0703, -0.1222, -0.1166, -0.0846, -0.0629, -0.039, -0.039, -0.0263, -0.0183, 0.0128, 0.0328, 0.0254, 0.0225, 0.0372, 0.0415, 0.0621, 0.05, 0.0095, 0.0217, 0.0136, -0.0337, 0.0, 0.0252, 0.0382, 0.0429, 0.049, 0.049, 0.0509, 0.0427, 0.0374, 0.024, 0.024, 0.0886, 0.154, 0.1997, 0.1792, 0.1512, 0.2277, 0.2428, 0.2421, 0.2368, 0.3127, 0.3378, 0.3378, 0.3052, 0.3306, 0.3133, 0.3139, 0.3204, 0.372, 0.3791, 0.4081, 0.3124, 0.3233, 0.2856, 0.2105, 0.2262, 0.3186, 0.3329, 0.3291, 0.3663, 0.3322, 0.3637, 0.3637, 0.375, 0.3992, 0.4002, 0.4083, 0.4012, 0.4269, 0.4602, 0.4069, 0.3928, 0.3885, 0.314, 0.3415, 0.2929, 0.2059, 0.2735, 0.2957, 0.3055, 0.2565, 0.2701, 0.2982, 0.3241, 0.3201, 0.3544, 0.3236, 0.3583, 0.4021, 0.3892, 0.2893, 0.2807, 0.2166, 0.2965, 0.3402, 0.328, 0.328, 0.3426, 0.3477, 0.47, 0.5173, 0.5454, 0.5503, 0.5762, 0.5478, 0.5343, 0.5977, 0.5994, 0.5825, 0.6094, 0.5901, 0.6595, 0.6171, 0.5432, 0.547, 0.5795, 0.5647, 0.5675, 0.6252, 0.6842, 0.6167, 0.6913, 0.6949, 0.6499, 0.6785, 0.6942, 0.6357, 0.5734, 0.549, 0.6394, 0.6606, 0.6847, 0.6847, 0.7595, 0.7246, 0.7106, 0.7108, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1977266539377443085", "ticker_idx": 0, "ts": "2025-10-12T06:54:13+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley maintains NVIDIA as its top pick in the AI semiconductor sector. Reiterated an Overweight (OW) rating.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Morgan Stanley reiterates NVDA Overweight as top AI semiconductor pick; durable competitive moat, above-consensus growth expected for several years.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA NDR Commentary: Broad Prospects for AI Infrastructure, Competitive Dynamics Remain Strong\n\nMorgan Stanley (J. Moore, Oct 11, 2025)\n\nMorgan Stanley stated that at its investor meeting in Boston, NVIDIA emphasized that its AI growth is still in the early stages, particularly that agentic AI applications in enterprise and industrial scenarios remain in the “embryonic phase.” Management presented strong arguments supporting the durability of NVIDIA’s competitive position and reiterated an Overweight (OW) rating.\n\nManagement noted that current AI demand growth is mainly driven by the migration of cloud computing CapEx from CPUs to GPUs, and there remains substantial room for further expansion. CEO Jensen Huang stated that most of the current growth stems from increasing NVIDIA’s revenue share within existing cloud CapEx (which is growing about 30–40%), while AI applications that could fundamentally reshape the labor market (such as enterprise- and industry-level AI) have yet to fully arrive.\n\nIn terms of AI and LLMs, hyperscalers are simultaneously supporting both third-party and self-developed model training. These foundational models will form the core of many future business models. Management highlighted that AI models are evolving from the “pre-training” phase to the “post-training” phase, shifting focus from knowledge accumulation to reasoning and action capabilities. This evolution has two key effects: (1) inference models’ token-intensive nature significantly increases compute demand, and (2) higher-performance models will generate more high-value applications. NVIDIA believes this evolution will trigger a structural transformation of the global labor market—AI will gradually replace traditional human labor. Following programming, healthcare, law, and industrial automation will become the next key application areas, and robotics will drive AI into the physical world.\n\nThe company projects that by 2030, the AI infrastructure market will reach $3–5 trillion, higher than most market assumptions. On the supply side, NVIDIA believes that wafer capacity is no longer the main bottleneck—manufacturing and packaging have already seen substantial expansion. The current constraints lie in data center space, power availability, and supporting infrastructure.\n\nOn financing and partnerships, the company emphasized that its role is that of an “accelerator,” not a “financier.” NVIDIA invests in companies such as CoreWeave and in projects with the UK and London sovereign funds to accelerate data center construction and innovation. This model will expand to include more partners in the future. Regarding OpenAI’s recent 10GW data center build-out target, NVIDIA views it as incremental upside but cautions against interpreting that figure too literally, advising investors to remain prudent when evaluating similar numbers announced by competitors.\n\nOn architectural strategy, NVIDIA reiterated that scale-up remains based on copper interconnects, scale-out will evolve toward co-packaged optics, and scale-across/DCI will continue to use optical modules. The company emphasized that its full-stack architecture and GPU flexibility constitute its long-term moat.\n\nRegarding AMD competition, management noted that AMD’s collaboration with OpenAI is primarily advanced through cloud service providers, whereas NVIDIA directly signs agreements with non-cloud data centers and invests equity on a per-GW basis. AMD provides warrants to OpenAI as an incentive, but this does not directly drive CSP investment. NVIDIA argued that it remains uncertain whether CSPs would commit tens of billions of dollars to chips that are not yet mass-produced or performance-verified.\n\nMorgan Stanley pointed out that the market’s three major concerns about NVIDIA are:\n\n1. Whether long-term projections are becoming bubble-like,\n\n2. Circular financing risks, and\n    \n3. Rising competitor confidence.\n\nThe company responded that although some three-year agreements with customers may include aggressive assumptions, even partial realization would be sufficient to sustain above-consensus growth for the next several years. Current spending is primarily driven by workload migration from CPUs to GPUs, not by speculative expectations. As for financing, most co-investment projects will materialize gradually after 2026, likely driving a positive inflection in 2027 performance.\n\nOn the competitive landscape, Morgan Stanley noted that whenever replacement competitors reach peak market hype, it usually takes another 12 months before any real productization. Historically, AMD, Intel, and ASIC vendors have all gone through similar cycles. NVIDIA continues to maintain a strong market share, and the company expects its XPU market share in 2026 to remain at the high-80% range, while still being undervalued relative to peers.\n\nMorgan Stanley maintains NVIDIA as its top pick in the AI semiconductor sector.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027400198616169602, "ret_signed_m1d": -0.027400198616169602, "alpha_spy_m1d": -0.012288006527152895, "alpha_c_m1d": 0.015013325358560481, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.04241352397473008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.044020911827305276, "ret_signed_p1d": -0.044020911827305276, "alpha_spy_p1d": -0.04279925702384857, "alpha_c_p1d": -0.025625097081496073, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.018395814745809203, "ret_p1w": -0.0301614576854774, "ret_signed_p1w": -0.0301614576854774, "alpha_spy_p1w": -0.04261926310241848, "alpha_c_p1w": -0.05214824244010052, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.021986784754623123, "ret_p1m": 0.02570089993140412, "ret_signed_p1m": 0.02570089993140412, "alpha_spy_p1m": -0.004402855555442287, "alpha_c_p1m": -0.007558867202471342, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.03325976713387546, "ret_p3m": -0.01736250530271133, "ret_signed_p3m": -0.01736250530271133, "alpha_spy_p3m": -0.06035760738902307, "alpha_c_p3m": -0.13630070482959789, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.11893819952688656, "ret_p6m": -0.05416548997458348, "ret_signed_p6m": -0.05416548997458348, "alpha_spy_p6m": -0.05406532094629157, "alpha_c_p6m": -0.23486992342684687, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.1807044334522634, "price_path": [-0.0901, -0.0814, -0.0845, -0.09, -0.1131, -0.0932, -0.0775, -0.0787, -0.0615, -0.0681, -0.0481, -0.0555, -0.0776, -0.0442, -0.0535, -0.0473, -0.0401, -0.0299, -0.0333, -0.0275, -0.0358, -0.0335, -0.0418, -0.0336, -0.0674, -0.0687, -0.0709, -0.0549, -0.0452, -0.0348, -0.0357, -0.0433, -0.0751, -0.0751, -0.0932, -0.094, -0.0885, -0.1132, -0.1063, -0.0933, -0.0584, -0.0592, -0.0558, -0.0561, -0.0714, -0.0957, -0.0641, -0.0619, -0.025, -0.0525, -0.0603, -0.0564, -0.0538, -0.0344, -0.0092, -0.0057, 0.003, -0.0037, -0.0148, -0.0174, 0.0042, 0.0226, -0.0274, 0.0, -0.044, -0.0451, -0.0346, -0.0271, -0.0302, -0.038, -0.0427, -0.0327, -0.0109, 0.0168, 0.0675, 0.0994, 0.0774, 0.0752, 0.0986, 0.0551, 0.0366, -0.0013, -0.0009, 0.057, 0.0257, 0.0291, -0.0078, 0.0098, -0.0091, -0.037, -0.0096, -0.0408, -0.0501, -0.0306, -0.0558, -0.0428, -0.0428, -0.0601, -0.0446, -0.0364, -0.0464, -0.0262, -0.0313, -0.0147, -0.0177, -0.0241, -0.0392, -0.0706, -0.0638, -0.0562, -0.0922, -0.0752, -0.0389, -0.0245, 0.0048, 0.0016, 0.0016, 0.0118, -0.0005, -0.0041, -0.0096, -0.0096, 0.0029, -0.001, -0.0057, 0.0043, -0.0174, -0.0183, -0.0179, -0.0133, -0.0275, -0.0067, -0.011, -0.011, -0.0544, -0.0265, -0.0184, -0.0034, -0.0098, 0.0011, 0.017, 0.0223, 0.015, -0.0143, -0.0423, -0.075, -0.0872, -0.0154, 0.0092, 0.0012, 0.0092, -0.0073, -0.0292, -0.0292, -0.0177, -0.0017, -0.0022, 0.008, 0.0172, 0.0241, 0.0385, -0.0182, -0.059, -0.031, -0.0439, -0.028, -0.0264, -0.0557, -0.0301, -0.0188, -0.0121, -0.0274, -0.0427, -0.027, -0.0338, -0.042, -0.0517, -0.0828, -0.0672, -0.0696, -0.0511, -0.0906, -0.1104, -0.1228, -0.0738, -0.0666, -0.0579, -0.0579, -0.0566, -0.0542]}
{"unit_id": "orig:1978246136763924769", "ticker_idx": 0, "ts": "2025-10-14T23:46:47+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung will catch up faster than the market expects. Samsung is still cheap.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung on cheap valuation + faster-than-expected share recovery; SK Hynix maintains leadership but faces market-share pessimism headwind.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "I believe SK hynix will continue to maintain its leadership, but Samsung will catch up faster than the market expects.\n\nWhile the leader suffers from pessimism about losing market share, optimism and expectations of gaining share tend to surround the second place.\n\nSamsung is still cheap.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018558947933519976, "ret_signed_m1d": 0.018558947933519976, "alpha_spy_m1d": 0.017335798864127394, "alpha_c_m1d": 0.005557693029785815, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01300125490373416, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03711789586703995, "ret_signed_p1d": 0.03711789586703995, "alpha_spy_p1d": 0.03267828000036066, "alpha_c_p1d": 0.027207156656559706, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.009910739210480246, "ret_p1w": 0.0644104562538288, "ret_signed_p1w": 0.0644104562538288, "alpha_spy_p1w": 0.05072946532140854, "alpha_c_p1w": 0.04014854923497735, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.024261907018851447, "ret_p1m": 0.1255458293080054, "ret_signed_p1m": 0.1255458293080054, "alpha_spy_p1m": 0.09360825403141115, "alpha_c_p1m": 0.08146211774975942, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.044083711558245975, "ret_p3m": 0.5248438699520155, "ret_signed_p3m": 0.5248438699520155, "alpha_spy_p3m": 0.4736669087563177, "alpha_c_p3m": 0.48494001836286205, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.03990385158915344, "ret_p6m": 1.313985622955864, "ret_signed_p6m": 1.313985622955864, "alpha_spy_p6m": 1.2873646121171467, "alpha_c_p6m": 1.3045625395104183, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.009423083445445624, "price_path": [-0.2751, -0.2707, -0.2631, -0.2827, -0.2783, -0.2827, -0.2838, -0.2349, -0.2327, -0.2109, -0.224, -0.2512, -0.2425, -0.2403, -0.2522, -0.2338, -0.2196, -0.2283, -0.2272, -0.2186, -0.2218, -0.2218, -0.2392, -0.2392, -0.2338, -0.2327, -0.224, -0.2229, -0.2359, -0.2327, -0.2435, -0.2425, -0.2653, -0.249, -0.2414, -0.2381, -0.2446, -0.2381, -0.2229, -0.2109, -0.2022, -0.1805, -0.1686, -0.137, -0.1501, -0.1273, -0.1273, -0.0925, -0.0794, -0.0718, -0.0642, -0.0947, -0.0808, -0.0841, -0.0611, -0.0202, -0.0202, -0.0202, -0.0202, -0.0202, -0.0202, 0.0306, 0.0186, 0.0, 0.0371, 0.0666, 0.0688, 0.071, 0.0644, 0.0764, 0.0535, 0.0786, 0.1135, 0.0862, 0.0972, 0.1365, 0.1736, 0.2129, 0.1452, 0.0983, 0.083, 0.0688, 0.0983, 0.1299, 0.1255, 0.1223, 0.0611, 0.0983, 0.0677, 0.0535, 0.0983, 0.0349, 0.0557, 0.0841, 0.1223, 0.1299, 0.0972, 0.1004, 0.1288, 0.1408, 0.1474, 0.1834, 0.1954, 0.1834, 0.179, 0.1714, 0.1889, 0.1441, 0.1223, 0.1779, 0.1747, 0.1605, 0.2063, 0.2172, 0.2129, 0.2129, 0.2773, 0.3109, 0.3153, 0.3153, 0.3153, 0.4097, 0.515, 0.5237, 0.5468, 0.5226, 0.5248, 0.5226, 0.5095, 0.5391, 0.5786, 0.6334, 0.6378, 0.5929, 0.64, 0.6707, 0.6686, 0.6686, 0.7497, 0.7815, 0.7629, 0.7607, 0.6499, 0.8375, 0.855, 0.7475, 0.7399, 0.8254, 0.8188, 0.8408, 0.9593, 0.9878, 0.9878, 0.9878, 0.9878, 1.0843, 1.0854, 1.1172, 1.194, 1.2324, 1.3915, 1.375, 1.375, 1.1403, 0.8891, 1.1019, 1.0646, 0.9033, 1.0613, 1.0843, 1.0613, 1.013, 1.0701, 1.1271, 1.2873, 1.1995, 1.1874, 1.0437, 1.081, 1.0733, 0.9757, 0.9757, 0.938, 0.838, 1.0848, 0.9611, 1.0469, 1.1227, 1.1601, 1.314]}
{"unit_id": "orig:1978246136763924769", "ticker_idx": 1, "ts": "2025-10-14T23:46:47+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix will continue to maintain its leadership", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung on cheap valuation + faster-than-expected share recovery; SK Hynix maintains leadership but faces market-share pessimism headwind.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I believe SK hynix will continue to maintain its leadership, but Samsung will catch up faster than the market expects.\n\nWhile the leader suffers from pessimism about losing market share, optimism and expectations of gaining share tend to surround the second place.\n\nSamsung is still cheap.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008505465672627333, "ret_signed_m1d": 0.008505465672627333, "alpha_spy_m1d": 0.007282316603234751, "alpha_c_m1d": -0.010235096992196047, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01874056266482338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026731507056273474, "ret_signed_p1d": 0.026731507056273474, "alpha_spy_p1d": 0.022291891189594182, "alpha_c_p1d": 0.0019039607312922868, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.024827546324981187, "ret_p1w": 0.1640339917076754, "ret_signed_p1w": 0.1640339917076754, "alpha_spy_p1w": 0.15035300077525515, "alpha_c_p1w": 0.12865177561713947, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.035382216090535934, "ret_p1m": 0.49939248305531825, "ret_signed_p1m": 0.49939248305531825, "alpha_spy_p1m": 0.467454907778724, "alpha_c_p1m": 0.4333355887176735, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.06605689433764472, "ret_p3m": 0.8093143054173029, "ret_signed_p3m": 0.8093143054173029, "alpha_spy_p3m": 0.7581373442216051, "alpha_c_p3m": 0.6386064169295438, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.17070788848775909, "ret_p6m": 1.516761467129815, "ret_signed_p6m": 1.516761467129815, "alpha_spy_p6m": 1.4901404562910978, "alpha_c_p6m": 1.2446896456243026, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.27207182150551246, "price_path": [-0.346, -0.3472, -0.3387, -0.3485, -0.3472, -0.346, -0.3545, -0.3642, -0.363, -0.3606, -0.3363, -0.3739, -0.3739, -0.3606, -0.3727, -0.3642, -0.3776, -0.3521, -0.3472, -0.3254, -0.329, -0.329, -0.3509, -0.3618, -0.38, -0.4055, -0.3909, -0.3703, -0.3654, -0.3691, -0.3475, -0.3463, -0.3779, -0.367, -0.3621, -0.3548, -0.3354, -0.3269, -0.3001, -0.2612, -0.2539, -0.2017, -0.1956, -0.1543, -0.1896, -0.1361, -0.1361, -0.147, -0.1227, -0.1312, -0.1337, -0.1823, -0.1519, -0.1555, -0.1252, -0.0389, -0.0389, -0.0389, -0.0389, -0.0389, -0.0389, 0.0401, 0.0085, 0.0, 0.0267, 0.0996, 0.1312, 0.1798, 0.164, 0.1701, 0.1628, 0.2394, 0.3001, 0.2661, 0.356, 0.3803, 0.3584, 0.5067, 0.4241, 0.407, 0.4411, 0.4095, 0.4727, 0.5043, 0.4994, 0.4872, 0.3609, 0.4727, 0.3852, 0.3657, 0.3876, 0.2661, 0.2637, 0.2612, 0.2734, 0.3229, 0.2889, 0.3083, 0.357, 0.3424, 0.3181, 0.3229, 0.4032, 0.3764, 0.4275, 0.374, 0.3886, 0.3473, 0.2889, 0.34, 0.3424, 0.3302, 0.4105, 0.4202, 0.4299, 0.4299, 0.4567, 0.5564, 0.5832, 0.5832, 0.5832, 0.6464, 0.6926, 0.7655, 0.8045, 0.8385, 0.8093, 0.8215, 0.7947, 0.8045, 0.8215, 0.8385, 0.858, 0.8069, 0.7996, 0.8361, 0.8652, 0.7899, 0.9455, 1.0452, 1.0938, 1.2106, 1.0185, 1.2057, 1.1887, 1.0476, 1.0403, 1.1571, 1.1303, 1.0914, 1.1595, 1.14, 1.14, 1.14, 1.14, 1.1741, 1.3078, 1.3127, 1.444, 1.4756, 1.6776, 1.585, 1.585, 1.2877, 1.0685, 1.2926, 1.2512, 1.0368, 1.2853, 1.3267, 1.2658, 1.2171, 1.373, 1.3633, 1.5728, 1.468, 1.4534, 1.2731, 1.4023, 1.4242, 1.2731, 1.2731, 1.1269, 0.9661, 1.1854, 1.0222, 1.1343, 1.1586, 1.2317, 1.5168]}
{"unit_id": "reply:1997640497381134376", "ticker_idx": 0, "ts": "2025-12-07T12:13:03+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still think TSMC is a stock worth buying.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish TSMC despite geopolitical discount; calls it worth buying at current valuation.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@0kotatsu Naturally, it’s because Taiwan continues to suffer from a geopolitical discount due to the risk of a Chinese invasion.\nWhenever China conducts even a minor military drill near Taiwan, foreign investors immediately panic.\n\nI still think TSMC is a stock worth buying.", "reply_text": null, "parent_text": "@jukan05 In your opinion why does TSMC trade at such a low valuation compared to ASML?\n\nAnd what is your thoughts on TSMC current valuation? (Sub 24PE on TW exchange)", "quoted_text": null, "parent_author": "0kotatsu", "ret_m1d": -0.023411361066560832, "ret_signed_m1d": -0.023411361066560832, "alpha_spy_m1d": -0.026424691461986893, "alpha_c_m1d": -0.012205369963342427, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010033405232020565, "ret_signed_p1d": -0.010033405232020565, "alpha_spy_p1d": -0.00917026026505896, "alpha_c_p1d": -0.011254418116198428, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.02686736381242605, "ret_signed_p1w": -0.02686736381242605, "alpha_spy_p1w": -0.02262526623996175, "alpha_c_p1w": 0.01559630793944089, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.14426978585818606, "ret_signed_p1m": 0.14426978585818606, "alpha_spy_p1m": 0.12931402372129308, "alpha_c_p1m": 0.08925385247066853, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.275139341891532, "ret_signed_p3m": 0.275139341891532, "alpha_spy_p3m": 0.2755881090344503, "alpha_c_p3m": 0.19908402751601573, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1839, -0.1739, -0.1605, -0.1639, -0.1438, -0.1538, -0.1405, -0.1538, -0.1338, -0.1037, -0.1037, -0.1171, -0.1304, -0.1304, -0.1271, -0.1137, -0.087, -0.0635, -0.0635, -0.0401, -0.0535, -0.0368, -0.0368, -0.0535, -0.0468, -0.0201, -0.0067, -0.0301, -0.01, -0.01, -0.0234, -0.0301, -0.0301, -0.01, -0.0134, 0.0067, 0.0067, 0.0033, 0.01, 0.0067, -0.0234, -0.0201, -0.0234, -0.0134, -0.0201, -0.0134, -0.0234, -0.0435, -0.0334, -0.0602, -0.0669, -0.0268, -0.0736, -0.0803, -0.0535, -0.0368, -0.0401, -0.0368, -0.0569, -0.0435, -0.0301, -0.0334, -0.0234, 0.0, -0.01, 0.0067, -0.0134, -0.0067, -0.0269, -0.0369, -0.0403, -0.0403, -0.0403, -0.0168, -0.0, 0.0033, 0.0033, 0.0134, 0.0268, 0.0201, 0.0402, 0.0402, 0.0637, 0.1208, 0.1443, 0.1241, 0.1308, 0.1275, 0.1342, 0.1476, 0.1476, 0.1342, 0.1678, 0.1812, 0.1912, 0.1678, 0.1812, 0.1879, 0.1778, 0.1946, 0.2214, 0.2114, 0.1912, 0.1845, 0.208, 0.198, 0.1845, 0.1946, 0.2181, 0.2617, 0.2852, 0.2852, 0.2852, 0.2852, 0.2852, 0.2852, 0.2852, 0.2852, 0.2751, 0.3188, 0.3523, 0.3389, 0.3389, 0.3255, 0.2986, 0.2516, 0.2751, 0.2684, 0.2147, 0.2416, 0.302, 0.2651, 0.2516, 0.2382, 0.2591, 0.2827, 0.2456, 0.2389, 0.2187, 0.2187, 0.2423, 0.2389, 0.2254, 0.1985, 0.185, 0.249, 0.2187, 0.2187, 0.2187, 0.2524, 0.313, 0.3163, 0.3466, 0.3399, 0.3837, 0.4005, 0.4039, 0.3668, 0.3635, 0.3803, 0.3803, 0.4005, 0.4712, 0.5251, 0.4914, 0.4678, 0.4375, 0.4375, 0.5318, 0.515, 0.515, 0.5554, 0.5419, 0.5049, 0.5183, 0.4948, 0.5284, 0.5251, 0.5082, 0.4847, 0.4712, 0.5015, 0.5183, 0.5554, 0.5284, 0.5486, 0.5453, 0.5857, null, null]}
{"unit_id": "orig:1993571622213837112", "ticker_idx": 0, "ts": "2025-11-26T06:44:48+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "has not slowed NVDA down in the slightest given the state of compute demand with the company recently indicating hundreds of billions of dollars of revenue visibility going forward (with numbers that likely continue to have a solid upward bias) and a roadmap that suggests even more to come", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bernstein bullish NVDA (demand scarcity, upward revenue bias) and AVGO (Street AI estimates far too low); incrementally bearish AMD as TPU viability threatens its GPU second-source thesis.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein’s comments on Google’s TPUs:\n\nThe success of Google’s TPUs is not new. \nIn fact they are the only really successful ASIC program in our opinion (certainly the only one that has truly reached scale). This should not be surprising; Google (partnered with AVGO) has been making these chips across numerous generations for well over 10 years. And they have been trying to push TPUs to more external customers for a while, most notably with Fluidstack earlier this year (though they also offered a multi-billion dollar backstop for the effort too). Meta would clearly be a greater coup though. \n\nGPUs are clearly not going anywhere though…\nRight now the overarching theme is of compute scarcity, and if anything this feels like an effort to secure more. We note others like OpenAI are taking a similar diversified approach, something that has not slowed NVDA down in the slightest given the state of compute demand with the company recently indicating hundreds of billions of dollars of revenue visibility going forward (with numbers that likely continue to have a solid upward bias) and a roadmap that suggests even more to come. \n\nTo that end, we still think the question of “ASIC or GPU” kind of misses the point. \nRight now the real question should really be “is the opportunity in front of us still big, or is it not?” as (hopefully!) we are not yet in a mature, saturated market for AI hardware. In other words, it’s still the size of the pie that matters; if it’s big both GPU and ASIC should thrive (and if it’s not, they’re both in trouble). \n\nAVGO would be a clear winner here as they are (of course) Google’s ASIC partner of choice, though we note they have also been working with Meta on their own ASICs (so we don’t know how much of any TPU usage might be incremental). AVGO’s stock has of course been on fire lately as the ASIC narrative has taken off, and while the stock is growing more expensive perhaps it is justified as we continue to believe that Street estimates look far too low ($40B in AI revenue in FY26, ~$63B in 27, numbers that have barely moved even after the OpenAI announcement). \n\nIf anything the news feels incrementally negative to the AMD narrative as the company has built their story around becoming the viable second source to NVDA, and hitched their wagon to OpenAI’s horse. In a world where Gemini hypothetically dominates over ChatGPT, and TPUs prove to be viable second sources for a broader swath of AI customers, will investors still be as eager to underwrite AMD’s targets out to 2030?", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01353597993671296, "ret_signed_m1d": -0.01353597993671296, "alpha_spy_m1d": -0.006679737507610195, "alpha_c_m1d": 0.008608829314011968, "bench_spy_m1d": -0.006856242429102766, "bench_c_m1d": -0.02214480925072493, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.003716828214031187, "ret_signed_p1w": -0.003716828214031187, "alpha_spy_p1w": -0.009910909238853405, "alpha_c_p1w": -0.05163036687760214, "bench_spy_p1w": 0.006194081024822218, "bench_c_p1w": 0.047913538663570954, "ret_p1m": 0.04638026613451718, "ret_signed_p1m": 0.04638026613451718, "alpha_spy_p1m": 0.027636185466424212, "alpha_c_p1m": -0.004217008943755207, "bench_spy_p1m": 0.018744080668092966, "bench_c_p1m": 0.050597275078272386, "ret_p3m": 0.06269099920404475, "ret_signed_p3m": 0.06269099920404475, "alpha_spy_p3m": 0.05573713041405437, "alpha_c_p3m": -0.12843015868959862, "bench_spy_p3m": 0.006953868789990381, "bench_c_p3m": 0.19112115789364337, "ret_p6m": 0.21787475717947058, "ret_signed_p6m": 0.21787475717947058, "alpha_spy_p6m": 0.11890321160606754, "alpha_c_p6m": -0.4204072641372303, "bench_spy_p6m": 0.09897154557340304, "bench_c_p6m": 0.6382820213167009, "price_path": [-0.0338, -0.0338, -0.0526, -0.0535, -0.0478, -0.0735, -0.0663, -0.0528, -0.0163, -0.0171, -0.0135, -0.0139, -0.0298, -0.0553, -0.0223, -0.0199, 0.0186, -0.0102, -0.0183, -0.0143, -0.0115, 0.0088, 0.0351, 0.0387, 0.0479, 0.0408, 0.0293, 0.0265, 0.0491, 0.0683, 0.0161, 0.0447, -0.0013, -0.0024, 0.0086, 0.0164, 0.0132, 0.005, 0.0001, 0.0105, 0.0333, 0.0623, 0.1152, 0.1486, 0.1255, 0.1233, 0.1477, 0.1022, 0.0829, 0.0434, 0.0438, 0.1042, 0.0716, 0.0751, 0.0366, 0.055, 0.0352, 0.0061, 0.0347, 0.0021, -0.0077, 0.0127, -0.0135, 0.0, 0.0, -0.0181, -0.0019, 0.0067, -0.0037, 0.0174, 0.012, 0.0294, 0.0262, 0.0196, 0.0038, -0.029, -0.022, -0.014, -0.0517, -0.0339, 0.0041, 0.0191, 0.0497, 0.0464, 0.0464, 0.057, 0.0442, 0.0404, 0.0347, 0.0347, 0.0477, 0.0437, 0.0388, 0.0492, 0.0266, 0.0256, 0.026, 0.0308, 0.016, 0.0377, 0.0332, 0.0332, -0.0121, 0.017, 0.0255, 0.0412, 0.0345, 0.0459, 0.0625, 0.068, 0.0604, 0.0297, 0.0005, -0.0336, -0.0464, 0.0286, 0.0543, 0.046, 0.0544, 0.0371, 0.0142, 0.0142, 0.0262, 0.0429, 0.0424, 0.0531, 0.0627, 0.0699, 0.0849, 0.0257, -0.017, 0.0124, -0.0011, 0.0155, 0.0171, -0.0135, 0.0133, 0.0251, 0.0321, 0.0161, 0.0001, 0.0165, 0.0094, 0.0009, -0.0093, -0.0418, -0.0255, -0.028, -0.0087, -0.0499, -0.0706, -0.0836, -0.0324, -0.0249, -0.0158, -0.0158, -0.0144, -0.0119, 0.0102, 0.0204, 0.0465, 0.0503, 0.0903, 0.1034, 0.1005, 0.119, 0.1211, 0.109, 0.1235, 0.1076, 0.1555, 0.2018, 0.1827, 0.161, 0.1072, 0.101, 0.1012, 0.0902, 0.1531, 0.1734, 0.194, 0.2175, 0.2249, 0.2529, 0.3079, 0.2501, 0.2335, 0.224, 0.2398, 0.2179]}
{"unit_id": "orig:1993571622213837112", "ticker_idx": 1, "ts": "2025-11-26T06:44:48+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we continue to believe that Street estimates look far too low ($40B in AI revenue in FY26, ~$63B in 27, numbers that have barely moved even after the OpenAI announcement)", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bernstein bullish NVDA (demand scarcity, upward revenue bias) and AVGO (Street AI estimates far too low); incrementally bearish AMD as TPU viability threatens its GPU second-source thesis.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein’s comments on Google’s TPUs:\n\nThe success of Google’s TPUs is not new. \nIn fact they are the only really successful ASIC program in our opinion (certainly the only one that has truly reached scale). This should not be surprising; Google (partnered with AVGO) has been making these chips across numerous generations for well over 10 years. And they have been trying to push TPUs to more external customers for a while, most notably with Fluidstack earlier this year (though they also offered a multi-billion dollar backstop for the effort too). Meta would clearly be a greater coup though. \n\nGPUs are clearly not going anywhere though…\nRight now the overarching theme is of compute scarcity, and if anything this feels like an effort to secure more. We note others like OpenAI are taking a similar diversified approach, something that has not slowed NVDA down in the slightest given the state of compute demand with the company recently indicating hundreds of billions of dollars of revenue visibility going forward (with numbers that likely continue to have a solid upward bias) and a roadmap that suggests even more to come. \n\nTo that end, we still think the question of “ASIC or GPU” kind of misses the point. \nRight now the real question should really be “is the opportunity in front of us still big, or is it not?” as (hopefully!) we are not yet in a mature, saturated market for AI hardware. In other words, it’s still the size of the pie that matters; if it’s big both GPU and ASIC should thrive (and if it’s not, they’re both in trouble). \n\nAVGO would be a clear winner here as they are (of course) Google’s ASIC partner of choice, though we note they have also been working with Meta on their own ASICs (so we don’t know how much of any TPU usage might be incremental). AVGO’s stock has of course been on fire lately as the ASIC narrative has taken off, and while the stock is growing more expensive perhaps it is justified as we continue to believe that Street estimates look far too low ($40B in AI revenue in FY26, ~$63B in 27, numbers that have barely moved even after the OpenAI announcement). \n\nIf anything the news feels incrementally negative to the AMD narrative as the company has built their story around becoming the viable second source to NVDA, and hitched their wagon to OpenAI’s horse. In a world where Gemini hypothetically dominates over ChatGPT, and TPUs prove to be viable second sources for a broader swath of AI customers, will investors still be as eager to underwrite AMD’s targets out to 2030?", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031541578898531375, "ret_signed_m1d": -0.031541578898531375, "alpha_spy_m1d": -0.02468533646942861, "alpha_c_m1d": -0.009396769647806447, "bench_spy_m1d": -0.006856242429102766, "bench_c_m1d": -0.02214480925072493, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.04265924425431489, "ret_signed_p1w": -0.04265924425431489, "alpha_spy_p1w": -0.048853325279137105, "alpha_c_p1w": -0.09057278291788584, "bench_spy_p1w": 0.006194081024822218, "bench_c_p1w": 0.047913538663570954, "ret_p1m": -0.11741303363360234, "ret_signed_p1m": -0.11741303363360234, "alpha_spy_p1m": -0.1361571143016953, "alpha_c_p1m": -0.16801030871187472, "bench_spy_p1m": 0.018744080668092966, "bench_c_p1m": 0.050597275078272386, "ret_p3m": -0.16751246736535907, "ret_signed_p3m": -0.16751246736535907, "alpha_spy_p3m": -0.17446633615534946, "alpha_c_p3m": -0.35863362525900244, "bench_spy_p3m": 0.006953868789990381, "bench_c_p3m": 0.19112115789364337, "ret_p6m": 0.0469465628852459, "ret_signed_p6m": 0.0469465628852459, "alpha_spy_p6m": -0.05202498268815714, "alpha_c_p6m": -0.591335458431455, "bench_spy_p6m": 0.09897154557340304, "bench_c_p6m": 0.6382820213167009, "price_path": [-0.2533, -0.2533, -0.2511, -0.2407, -0.2314, -0.1591, -0.1321, -0.1546, -0.072, -0.097, -0.0964, -0.0858, -0.096, -0.1308, -0.1328, -0.1339, -0.1478, -0.1475, -0.1465, -0.1546, -0.1586, -0.1752, -0.1702, -0.1614, -0.1494, -0.1489, -0.1561, -0.1538, -0.131, -0.1322, -0.1835, -0.1028, -0.1344, -0.1163, -0.1092, -0.1213, -0.1216, -0.1381, -0.1441, -0.134, -0.1093, -0.0893, -0.0619, -0.0292, -0.0531, -0.0703, -0.0881, -0.1148, -0.0971, -0.1056, -0.1211, -0.0985, -0.1147, -0.1065, -0.1449, -0.1386, -0.1381, -0.1435, -0.1085, -0.1277, -0.1443, -0.0493, -0.0315, 0.0, 0.0, 0.0136, -0.0289, -0.0402, -0.0427, -0.0416, -0.0184, 0.0089, 0.0219, 0.0387, 0.0221, -0.0947, -0.1453, -0.1415, -0.18, -0.1703, -0.1439, -0.1395, -0.1197, -0.1174, -0.1174, -0.1126, -0.1195, -0.1183, -0.1278, -0.1278, -0.124, -0.1345, -0.1337, -0.1343, -0.1621, -0.1306, -0.1124, -0.1063, -0.1434, -0.1356, -0.1137, -0.1137, -0.1618, -0.1714, -0.1797, -0.1934, -0.1813, -0.1613, -0.1602, -0.1665, -0.1651, -0.1656, -0.1927, -0.2237, -0.2175, -0.161, -0.1332, -0.1421, -0.1362, -0.1654, -0.1805, -0.1805, -0.162, -0.1595, -0.1583, -0.1617, -0.1675, -0.1797, -0.1625, -0.1893, -0.1947, -0.1965, -0.2091, -0.1998, -0.1614, -0.1672, -0.1287, -0.1367, -0.1392, -0.1533, -0.1881, -0.1812, -0.1903, -0.2038, -0.194, -0.2175, -0.1855, -0.1962, -0.1949, -0.2186, -0.2407, -0.259, -0.2184, -0.2083, -0.2056, -0.2056, -0.2059, -0.1566, -0.1145, -0.1037, -0.0617, -0.041, -0.0384, 0.0019, 0.0063, 0.0267, 0.0092, 0.0156, 0.0674, 0.0605, 0.0676, 0.0561, 0.0097, 0.0239, 0.0542, 0.0639, 0.0518, 0.0792, 0.0744, 0.0419, 0.0859, 0.0819, 0.0589, 0.0526, 0.1106, 0.0738, 0.0625, 0.0381, 0.055, 0.0469]}
{"unit_id": "orig:1993571622213837112", "ticker_idx": 2, "ts": "2025-11-26T06:44:48+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the news feels incrementally negative to the AMD narrative as the company has built their story around becoming the viable second source to NVDA, and hitched their wagon to OpenAI's horse", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bernstein bullish NVDA (demand scarcity, upward revenue bias) and AVGO (Street AI estimates far too low); incrementally bearish AMD as TPU viability threatens its GPU second-source thesis.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein’s comments on Google’s TPUs:\n\nThe success of Google’s TPUs is not new. \nIn fact they are the only really successful ASIC program in our opinion (certainly the only one that has truly reached scale). This should not be surprising; Google (partnered with AVGO) has been making these chips across numerous generations for well over 10 years. And they have been trying to push TPUs to more external customers for a while, most notably with Fluidstack earlier this year (though they also offered a multi-billion dollar backstop for the effort too). Meta would clearly be a greater coup though. \n\nGPUs are clearly not going anywhere though…\nRight now the overarching theme is of compute scarcity, and if anything this feels like an effort to secure more. We note others like OpenAI are taking a similar diversified approach, something that has not slowed NVDA down in the slightest given the state of compute demand with the company recently indicating hundreds of billions of dollars of revenue visibility going forward (with numbers that likely continue to have a solid upward bias) and a roadmap that suggests even more to come. \n\nTo that end, we still think the question of “ASIC or GPU” kind of misses the point. \nRight now the real question should really be “is the opportunity in front of us still big, or is it not?” as (hopefully!) we are not yet in a mature, saturated market for AI hardware. In other words, it’s still the size of the pie that matters; if it’s big both GPU and ASIC should thrive (and if it’s not, they’re both in trouble). \n\nAVGO would be a clear winner here as they are (of course) Google’s ASIC partner of choice, though we note they have also been working with Meta on their own ASICs (so we don’t know how much of any TPU usage might be incremental). AVGO’s stock has of course been on fire lately as the ASIC narrative has taken off, and while the stock is growing more expensive perhaps it is justified as we continue to believe that Street estimates look far too low ($40B in AI revenue in FY26, ~$63B in 27, numbers that have barely moved even after the OpenAI announcement). \n\nIf anything the news feels incrementally negative to the AMD narrative as the company has built their story around becoming the viable second source to NVDA, and hitched their wagon to OpenAI’s horse. In a world where Gemini hypothetically dominates over ChatGPT, and TPUs prove to be viable second sources for a broader swath of AI customers, will investors still be as eager to underwrite AMD’s targets out to 2030?", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03785474422334412, "ret_signed_m1d": 0.03785474422334412, "alpha_spy_m1d": 0.03099850179424135, "alpha_c_m1d": 0.01570993497261919, "bench_spy_m1d": -0.006856242429102766, "bench_c_m1d": -0.02214480925072493, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.015683348227223393, "ret_signed_p1w": -0.015683348227223393, "alpha_spy_p1w": -0.009489267202401175, "alpha_c_p1w": 0.03223019043634756, "bench_spy_p1w": 0.006194081024822218, "bench_c_p1w": 0.047913538663570954, "ret_p1m": 0.003734072873678551, "ret_signed_p1m": -0.003734072873678551, "alpha_spy_p1m": 0.015010007794414415, "alpha_c_p1m": 0.046863202204593835, "bench_spy_p1m": 0.018744080668092966, "bench_c_p1m": 0.050597275078272386, "ret_p3m": -0.08233756038720463, "ret_signed_p3m": 0.08233756038720463, "alpha_spy_p3m": 0.08929142917719501, "alpha_c_p3m": 0.273458718280848, "bench_spy_p3m": 0.006953868789990381, "bench_c_p3m": 0.19112115789364337, "ret_p6m": 1.0985342830951152, "ret_signed_p6m": -1.0985342830951152, "alpha_spy_p6m": -0.9995627375217122, "alpha_c_p6m": -0.4602522617784144, "bench_spy_p6m": 0.09897154557340304, "bench_c_p6m": 0.6382820213167009, "price_path": [-0.2409, -0.2409, -0.2423, -0.2432, -0.2448, -0.2945, -0.2933, -0.2727, -0.2553, -0.2734, -0.2598, -0.2478, -0.251, -0.2571, -0.2629, -0.2654, -0.2542, -0.249, -0.2491, -0.2472, -0.2557, -0.2468, -0.2448, -0.2345, -0.2078, -0.2314, -0.0492, -0.0127, 0.0995, 0.0871, 0.0031, 0.0102, 0.018, 0.1137, 0.0948, 0.0879, 0.1229, 0.111, 0.0746, 0.0969, 0.1805, 0.2121, 0.2043, 0.2338, 0.1895, 0.1955, 0.212, 0.1671, 0.1965, 0.1095, 0.0901, 0.1388, 0.1087, 0.2084, 0.1574, 0.152, 0.1227, 0.0749, 0.0435, -0.0384, -0.0488, 0.0038, -0.0379, 0.0, 0.0, 0.0154, 0.0258, 0.0047, 0.0157, 0.0081, 0.0174, 0.0321, 0.0344, 0.0335, 0.0336, -0.0162, -0.0311, -0.0237, -0.0753, -0.0615, -0.0038, 0.0033, 0.0031, 0.0037, 0.0037, 0.0035, 0.0064, 0.0051, -0.0004, -0.0004, 0.0431, 0.0319, 0.0005, -0.0197, -0.0446, -0.0517, -0.0306, 0.0314, 0.0437, 0.0639, 0.0821, 0.0821, 0.0825, 0.166, 0.1843, 0.2121, 0.173, 0.1764, 0.1797, 0.1771, 0.105, 0.1495, 0.1301, -0.0656, -0.1015, -0.0271, 0.0082, -0.0031, -0.0031, -0.0387, -0.0323, -0.0323, -0.0521, -0.0659, -0.0507, -0.0658, -0.0823, -0.0019, -0.0158, -0.0493, -0.0655, -0.0729, -0.1087, -0.0568, -0.069, -0.1018, -0.054, -0.0514, -0.0439, -0.077, -0.0973, -0.0824, -0.0837, -0.069, -0.0419, -0.0603, -0.054, -0.0414, 0.0281, -0.0489, -0.0572, -0.085, -0.0505, -0.0188, 0.0152, 0.0152, 0.0277, 0.034, 0.0821, 0.1046, 0.1438, 0.1521, 0.1906, 0.2048, 0.2988, 0.2994, 0.2834, 0.3279, 0.4164, 0.4252, 0.6235, 0.5619, 0.5086, 0.5735, 0.6546, 0.6829, 0.5942, 0.6582, 0.9669, 0.9066, 1.1247, 1.1415, 1.0925, 1.0794, 1.099, 0.9796, 0.965, 0.9326, 1.0892, 1.0985]}
{"unit_id": "reply:1997650161795805454", "ticker_idx": 0, "ts": "2025-12-07T12:51:27+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think it'll keep rising until the second quarter of 2026.", "tweet_type": "prediction", "timeline": "quarters", "summary": "Long SNDK; expects it to keep rising until Q2 2026.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@Tientientien121 I take a long-term view.\nI think it’ll keep rising until the second quarter of 2026.", "reply_text": null, "parent_text": "@jukan05 One more question,  Could you share your views on the stock $SNDK?\nWhat price levels do you expect in the short term and the long term?\nIn other words, how high do you think it could go?\nThanks a lot.", "quoted_text": null, "parent_author": "Tientientien121", "ret_m1d": 0.01330553946759161, "ret_signed_m1d": 0.01330553946759161, "alpha_spy_m1d": 0.010292209072165548, "alpha_c_m1d": 0.020282425862580222, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.006976886394988613, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02665540637023822, "ret_signed_p1d": -0.02665540637023822, "alpha_spy_p1d": -0.025792261403276617, "alpha_c_p1d": -0.029297286966355918, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.002641880596117696, "ret_p1w": -0.10467027088191028, "ret_signed_p1w": -0.10467027088191028, "alpha_spy_p1w": -0.10042817330944598, "alpha_c_p1w": -0.06856657318186998, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.036103697700040294, "ret_p1m": 0.5506719430075062, "ret_signed_p1m": 0.5506719430075062, "alpha_spy_p1m": 0.5357161808706132, "alpha_c_p1m": 0.5558034514498124, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": -0.005131508442306143, "ret_p3m": 1.5084934749338843, "ret_signed_p3m": 1.5084934749338843, "alpha_spy_p3m": 1.5089422420768026, "alpha_c_p3m": 1.5575172021102228, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": -0.049023727176338516, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6722, -0.6261, -0.618, -0.6004, -0.594, -0.5832, -0.5615, -0.5467, -0.5435, -0.5281, -0.5572, -0.5818, -0.5693, -0.4966, -0.5024, -0.4628, -0.4494, -0.4305, -0.4626, -0.4636, -0.4151, -0.4248, -0.4815, -0.403, -0.4354, -0.36, -0.3601, -0.3784, -0.3434, -0.3379, -0.3483, -0.2591, -0.1743, -0.2172, -0.2215, -0.0936, -0.1315, -0.1159, -0.0819, -0.137, -0.0398, -0.0789, 0.0621, 0.1884, 0.2045, 0.2556, 0.0803, 0.1272, 0.1792, 0.0863, 0.0909, -0.1309, -0.1118, 0.0066, -0.022, -0.0463, -0.0463, -0.0097, -0.0679, -0.0892, -0.1379, -0.0539, 0.0133, 0.0, -0.0267, 0.0328, 0.0716, -0.0856, -0.1047, -0.0717, -0.0827, -0.0267, 0.0538, 0.0691, 0.0862, 0.1091, 0.1091, 0.109, 0.0833, 0.0654, 0.0528, 0.0528, 0.2207, 0.2156, 0.5507, 0.5681, 0.4837, 0.6739, 0.7265, 0.7289, 0.72, 0.8151, 0.8345, 0.8345, 1.0097, 1.2233, 1.2328, 1.1015, 1.0881, 1.1352, 1.3401, 1.3919, 1.5558, 1.9505, 2.0847, 1.5926, 1.5556, 1.652, 1.5875, 1.4023, 1.6582, 1.7954, 1.7789, 1.7789, 1.6194, 1.6629, 1.7546, 1.8827, 1.956, 1.832, 1.8047, 1.8913, 1.8179, 1.7457, 1.5077, 1.6569, 1.5085, 1.3388, 1.6111, 1.7449, 1.9069, 1.7446, 1.9344, 2.1207, 2.1941, 2.3428, 2.4244, 2.1477, 2.1157, 2.1156, 2.0064, 1.6752, 1.7313, 1.5391, 1.8178, 2.0724, 2.1117, 2.1117, 2.2139, 2.1525, 2.4634, 2.7769, 2.7778, 3.2245, 3.1889, 2.9549, 3.078, 3.0848, 3.0494, 3.0071, 3.3424, 3.1355, 3.3904, 3.7465, 3.4456, 3.72, 3.8632, 4.2646, 4.57, 5.2373, 5.2535, 4.943, 5.9293, 5.8637, 5.44, 5.4187, 5.1326, 5.243, 4.9121, 5.1351, 5.1763, 5.8401, 5.5583, 5.5583, 6.0499, 6.0517, 6.281, 6.5175, null, null]}
{"unit_id": "quote:1980998521970184464", "ticker_idx": 0, "ts": "2025-10-22T14:03:47+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's fab capacity has become so constrained that the costs are expected to rise at least twofold for those who need to pay for wafer production", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish TSMC: severe fab capacity constraints expected to drive wafer costs up at least 2x, boosting pricing power.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "It’s highly interesting — TSMC’s fab capacity has become so constrained that the costs are expected to rise at least twofold for those who need to pay for wafer production. https://t.co/vZBfrNKQrY", "reply_text": null, "parent_text": null, "quoted_text": "A semiconductor consultant working with hyperscalers shares some interesting insights on the current $NVDA/ASIC buildout:\n\n1. He thinks that some smaller data center deals might start hitting some headwinds because there is backlash starting in some areas from residents because https://t.co/2Tul6thyzs", "parent_author": null, "ret_m1d": 0.013698660112563488, "ret_signed_m1d": 0.013698660112563488, "alpha_spy_m1d": 0.008472675215940084, "alpha_c_m1d": -0.0060886469498011575, "bench_spy_m1d": 0.0052259848966234035, "bench_c_m1d": 0.019787307062364645, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006849372137380683, "ret_signed_p1d": -0.006849372137380683, "alpha_spy_p1d": -0.012779263327880375, "alpha_c_p1d": -0.025160026638110566, "bench_spy_p1d": 0.005929891190499692, "bench_c_p1d": 0.018310654500729884, "ret_p1w": 0.03082188005052089, "ret_signed_p1w": 0.03082188005052089, "alpha_spy_p1w": 0.0014867828764315583, "alpha_c_p1w": -0.05780774152934631, "bench_spy_p1w": 0.02933509717408933, "bench_c_p1w": 0.0886296215798672, "ret_p1m": -0.003424643987591458, "ret_signed_p1m": -0.003424643987591458, "alpha_spy_p1m": 0.019441464266357844, "alpha_c_p1m": 0.03644542722226729, "bench_spy_p1m": -0.022866108253949302, "bench_c_p1m": -0.03987007120985875, "ret_p3m": 0.2094976366834067, "ret_signed_p3m": 0.2094976366834067, "alpha_spy_p3m": 0.17070796945423394, "alpha_c_p3m": 0.023331465398580553, "bench_spy_p3m": 0.038789667229172764, "bench_c_p3m": 0.18616617128482615, "ret_p6m": 0.43751726903094545, "ret_signed_p6m": 0.43751726903094545, "alpha_spy_p6m": 0.38083084402651735, "alpha_c_p6m": 0.09016009106922329, "bench_spy_p6m": 0.05668642500442811, "bench_c_p6m": 0.34735717796172216, "price_path": [-0.2189, -0.2189, -0.2257, -0.2121, -0.2086, -0.2086, -0.2257, -0.2155, -0.2325, -0.195, -0.1984, -0.195, -0.195, -0.1814, -0.1984, -0.195, -0.195, -0.1916, -0.2257, -0.2155, -0.2257, -0.2018, -0.1984, -0.1882, -0.2086, -0.2086, -0.2052, -0.2086, -0.2086, -0.2086, -0.195, -0.195, -0.1814, -0.1643, -0.1541, -0.1404, -0.1438, -0.1233, -0.1336, -0.1199, -0.1336, -0.113, -0.0822, -0.0822, -0.0959, -0.1096, -0.1096, -0.1062, -0.0925, -0.0651, -0.0411, -0.0411, -0.0171, -0.0308, -0.0137, -0.0137, -0.0308, -0.024, 0.0034, 0.0171, -0.0068, 0.0137, 0.0137, 0.0, -0.0068, -0.0068, 0.0137, 0.0103, 0.0308, 0.0308, 0.0274, 0.0342, 0.0308, 0.0, 0.0034, 0.0, 0.0103, 0.0034, 0.0103, 0.0, -0.0205, -0.0103, -0.0377, -0.0445, -0.0034, -0.0514, -0.0582, -0.0308, -0.0137, -0.0171, -0.0137, -0.0342, -0.0205, -0.0068, -0.0103, 0.0, 0.024, 0.0137, 0.0308, 0.0102, 0.0171, -0.0035, -0.0138, -0.0173, -0.0173, -0.0173, 0.0068, 0.0239, 0.0274, 0.0274, 0.0377, 0.0514, 0.0446, 0.0652, 0.0652, 0.0892, 0.1476, 0.1717, 0.1511, 0.158, 0.1545, 0.1614, 0.1751, 0.1751, 0.1614, 0.1958, 0.2095, 0.2198, 0.1958, 0.2095, 0.2164, 0.2061, 0.2232, 0.2507, 0.2404, 0.2198, 0.2129, 0.237, 0.2267, 0.2129, 0.2232, 0.2473, 0.292, 0.316, 0.316, 0.316, 0.316, 0.316, 0.316, 0.316, 0.316, 0.3057, 0.3504, 0.3847, 0.371, 0.371, 0.3572, 0.3298, 0.2817, 0.3057, 0.2988, 0.2439, 0.2713, 0.3332, 0.2954, 0.2817, 0.2679, 0.2893, 0.3134, 0.2755, 0.2686, 0.2479, 0.2479, 0.272, 0.2686, 0.2548, 0.2272, 0.2134, 0.2789, 0.2479, 0.2479, 0.2479, 0.2824, 0.3444, 0.3479, 0.3789, 0.372, 0.4168, 0.4341, 0.4375]}
{"unit_id": "thread:1980183944483193254", "ticker_idx": 0, "ts": "2025-10-20T08:06:57+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's turnaround momentum is expected to accelerate... Industry watchers expect further improvement in the fourth quarter as Exynos 2600 shipments ramp up", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung: Exynos 2600 return to flagship lineup improves Foundry/System LSI losses, reduces AP procurement costs, with momentum accelerating into Q4 and beyond.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Galaxy S26 confirmed to feature ‘Exynos 2600’—six times faster than Apple’s chip\n\nThe upcoming Galaxy S26 series, set for release early next year, will be equipped with Samsung’s in-house application processor (AP), the Exynos 2600. It is reported that not only the base and Plus models but also the flagship ‘Ultra’ model will likely adopt the chip. This marks the first time in four years—since 2022—that Samsung has used its own Exynos AP in a flagship model.\n\nAccording to industry sources on the 20th, Samsung Electronics has decided to use the Exynos 2600 in the Galaxy S26 and has begun mass production. The company has traditionally employed a dual-sourcing strategy, using both Qualcomm and Exynos chipsets. The Exynos share in the S26 lineup is reportedly around 50%, with Qualcomm chips used in the U.S., Japan, and China, while Exynos chips power the Korean and European models.\n\nFrom the Galaxy S1 in 2010 through the S6 in 2015, Samsung equipped most of its phones with Exynos chips. However, starting in 2016, it sharply increased Qualcomm’s share, switching to a mixed approach. The S23 (2023) and S25 (2025) series—except for lower-end and foldable models—did not use Exynos at all.\n\nThe AP serves as the “brain” of a smartphone. Analysts say Exynos’s return to the Galaxy S lineup reflects improvements in both performance and foundry yield (the ratio of good chips). While Apple and Qualcomm rely on TSMC’s foundries, Exynos chips are manufactured by Samsung Foundry.\n\nThe expansion of Exynos adoption is expected to improve the financial performance of Samsung’s Foundry and System LSI divisions, which have been struggling with losses, while also reducing semiconductor procurement costs for its smartphone division. Samsung spent ₩7.7899 trillion in the first half of this year and ₩10.9326 trillion last year on mobile AP purchases.\n\n“Its performance no longer lags behind Apple’s latest A19 Pro chipset.”\n\nSamsung’s decision to reintroduce its in-house AP into the Galaxy S series stems from major performance enhancements, which reportedly rival or exceed competitors.\n\nAccording to internal testing results revealed by industry sources on the 20th, the Exynos 2600’s NPU (neural processing unit) performance is more than six times higher than that of Apple’s A19 Pro chipset. Its CPU multi-core performance is 14% higher, and GPU performance up to 75% better.\n\nWhen compared to Qualcomm’s latest Snapdragon 8 Gen 5 Elite chipset, Exynos 2600 shows about 30% higher NPU performance and up to 29% higher GPU performance. The AP, which functions as the smartphone’s central brain, contains CPU, GPU, and NPU units responsible for AI computation, video processing, and overall system control.\n\nThe Exynos 2600 is produced using Samsung Foundry’s latest 2-nanometer (nm) process technology. To achieve the performance boost, the System LSI division reportedly took the drastic step of redesigning the chip’s architecture. For the North American market, Samsung is expected to continue using Qualcomm chips to ensure smooth local regulatory approvals.\n\nSamsung is said to have focused heavily on AI performance during Exynos 2600’s development—positioning AI as its weapon to surpass Apple. While Samsung launched the world’s first “AI phone” (Galaxy S24) in January 2024, Apple has struggled to fully implement AI capabilities in the iPhone.\n\nWith its AI semiconductors—including HBM (High Bandwidth Memory), DRAM, and now system semiconductors—entering a recovery phase, Samsung’s turnaround momentum is expected to accelerate. Although the Foundry and System LSI divisions each recorded losses of around ₩2 trillion in the first half of this year, those losses reportedly narrowed to the ₩1 trillion range in Q3.\n\nIndustry watchers expect further improvement in the fourth quarter as Exynos 2600 shipments ramp up. The smartphone division’s profitability will also benefit, given reduced spending on Qualcomm chipsets—Samsung spent ₩10.9326 trillion last year on AP purchases.\n\nThe company’s business outlook beyond next year appears even brighter. As early as 2026, Samsung will begin producing Tesla’s autonomous driving chip (AI6) and Apple’s image sensors. In July, Samsung signed a $16.5 billion (₩22.8 trillion) semiconductor supply contract with Tesla, and in August, it agreed to supply image sensors for the iPhone.\n\n---\n\nhttps://t.co/9rJY61p47O", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002038693068074693, "ret_signed_m1d": -0.002038693068074693, "alpha_spy_m1d": 0.0082547258809893, "alpha_c_m1d": 0.00899567002942181, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.011034363097496502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0061162395860097085, "ret_signed_p1d": -0.0061162395860097085, "alpha_spy_p1d": -0.00610124366599718, "alpha_c_p1d": -0.0066366592900064525, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": 0.000520419703996744, "ret_p1w": 0.03975535683183096, "ret_signed_p1w": 0.03975535683183096, "alpha_spy_p1w": 0.018989665155993096, "alpha_c_p1w": 0.0022454244184331174, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03750993241339784, "ret_p1m": -0.0030580796975584468, "ret_signed_p1m": -0.0030580796975584468, "alpha_spy_p1m": 0.01365569610319628, "alpha_c_p1m": 0.028726492666849124, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.03178457236440757, "ret_p3m": 0.47400115025816647, "ret_signed_p3m": 0.47400115025816647, "alpha_spy_p3m": 0.4397608929428092, "alpha_c_p3m": 0.46299705029383453, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.011004099964331937, "ret_p6m": 1.119605678926571, "ret_signed_p6m": 1.119605678926571, "alpha_spy_p6m": 1.079214991319748, "alpha_c_p6m": 1.090086169346484, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.029519509580086867, "price_path": [-0.3261, -0.3302, -0.3312, -0.2856, -0.2835, -0.2632, -0.2754, -0.3008, -0.2927, -0.2906, -0.3018, -0.2845, -0.2713, -0.2795, -0.2784, -0.2703, -0.2734, -0.2734, -0.2896, -0.2896, -0.2845, -0.2835, -0.2754, -0.2744, -0.2866, -0.2835, -0.2937, -0.2927, -0.314, -0.2987, -0.2916, -0.2886, -0.2947, -0.2886, -0.2744, -0.2632, -0.2551, -0.2348, -0.2236, -0.1942, -0.2064, -0.1851, -0.1851, -0.1526, -0.1404, -0.1333, -0.1262, -0.1546, -0.1417, -0.1448, -0.1233, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0377, -0.0489, -0.0663, -0.0316, -0.0041, -0.002, 0.0, -0.0061, 0.0051, -0.0163, 0.0071, 0.0398, 0.0143, 0.0245, 0.0612, 0.0958, 0.1325, 0.0693, 0.0255, 0.0112, -0.002, 0.0255, 0.055, 0.051, 0.0479, -0.0092, 0.0255, -0.0031, -0.0163, 0.0255, -0.0336, -0.0143, 0.0122, 0.0479, 0.055, 0.0245, 0.0275, 0.054, 0.0652, 0.0714, 0.105, 0.1162, 0.105, 0.1009, 0.0938, 0.1101, 0.0683, 0.0479, 0.0999, 0.0968, 0.0836, 0.1264, 0.1366, 0.1325, 0.1325, 0.1927, 0.2241, 0.2282, 0.2282, 0.2282, 0.3163, 0.4146, 0.4228, 0.4443, 0.4218, 0.4238, 0.4218, 0.4095, 0.4371, 0.474, 0.5252, 0.5293, 0.4873, 0.5314, 0.56, 0.558, 0.558, 0.6338, 0.6635, 0.6461, 0.644, 0.5406, 0.7157, 0.7321, 0.6317, 0.6246, 0.7045, 0.6983, 0.7188, 0.8294, 0.8561, 0.8561, 0.8561, 0.8561, 0.9462, 0.9472, 0.9769, 1.0486, 1.0845, 1.233, 1.2177, 1.2177, 0.9985, 0.7639, 0.9626, 0.9278, 0.7772, 0.9247, 0.9462, 0.9247, 0.8796, 0.9329, 0.9862, 1.1357, 1.0538, 1.0425, 0.9083, 0.9431, 0.936, 0.8448, 0.8448, 0.8096, 0.7162, 0.9466, 0.8312, 0.9112, 0.9821, 1.017, 1.1607, 1.0939, 1.1145, 1.0632, 1.1196]}
{"unit_id": "reply:1994383349260128598", "ticker_idx": 0, "ts": "2025-11-28T12:30:18+00:00", "symbol": "Infineon", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "My pick: Infineon", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Picks Infineon and MediaTek as buys for semi exposure to industrial and mobile end markets.", "resolved_tickers": ["IFX.DE"], "resolution_method": "exchange_ticker", "resolution_reason": "Infineon Technologies IFX.DE Germany", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@OmerCheeema My pick:\n\nInfineon\n\nMTK", "reply_text": null, "parent_text": "Guys, it would be a good time to buy semi stocks with industrial and mobile semiconductors exposure. Source: A semiconductor industry leader who talks to customers on daily basis.", "quoted_text": null, "parent_author": "OmerCheeema", "ret_m1d": -0.01706576511527713, "ret_signed_m1d": -0.01706576511527713, "alpha_spy_m1d": -0.011636991230446037, "alpha_c_m1d": -0.0040930643923526855, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025598647672915642, "ret_signed_p1d": -0.025598647672915642, "alpha_spy_p1d": -0.02103323207036445, "alpha_c_p1d": -0.027528919775847815, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.03234240132862576, "ret_signed_p1w": 0.03234240132862576, "alpha_spy_p1w": 0.028976813018489755, "alpha_c_p1w": -0.002118860448701909, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.011285510648221475, "ret_signed_p1m": 0.011285510648221475, "alpha_spy_p1m": 0.001785066528888768, "alpha_c_p1m": -0.025796517182873346, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.305731751498894, "ret_signed_p3m": 0.305731751498894, "alpha_spy_p3m": 0.2884528193182381, "alpha_c_p3m": 0.09224807134862245, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 1.1282804047162154, "ret_signed_p6m": 1.1282804047162154, "alpha_spy_p6m": 1.030977716709665, "alpha_c_p6m": 0.48721855606247866, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.0885, -0.1325, -0.1391, -0.1309, -0.1171, -0.1168, -0.1239, -0.1214, -0.1259, -0.1013, -0.113, -0.1013, -0.0728, -0.0919, -0.0919, -0.0696, -0.0778, -0.083, -0.1009, -0.0906, -0.0906, -0.0765, -0.0577, -0.0603, -0.0619, -0.0768, -0.09, -0.0867, -0.1186, -0.1031, -0.1157, -0.1051, -0.0845, -0.1065, -0.0614, -0.03, -0.0584, -0.0801, -0.078, -0.0562, -0.0567, -0.0538, -0.0524, -0.0553, -0.0473, -0.0589, -0.0573, -0.0663, -0.0663, -0.0822, -0.0674, -0.0029, -0.0029, -0.0285, -0.0537, -0.0921, -0.0979, -0.0941, -0.1276, -0.0954, -0.0765, -0.0424, -0.0171, 0.0, -0.0256, -0.0151, 0.0114, 0.0047, 0.0323, 0.0352, 0.0323, 0.0105, 0.0066, -0.0011, -0.0011, -0.0091, -0.0091, -0.0091, -0.0143, 0.0094, 0.0107, 0.0107, 0.0107, 0.0107, 0.0113, 0.0385, 0.0385, 0.0385, 0.0533, 0.0985, 0.1513, 0.1598, 0.1164, 0.1437, 0.1468, 0.1708, 0.1396, 0.1452, 0.1541, 0.1185, 0.1141, 0.1307, 0.1722, 0.1642, 0.1521, 0.1649, 0.1952, 0.1506, 0.1453, 0.1528, 0.1306, 0.1094, 0.1397, 0.157, 0.1558, 0.1799, 0.1976, 0.1767, 0.1976, 0.1895, 0.2271, 0.2686, 0.2616, 0.2537, 0.2729, 0.2948, 0.3057, 0.2973, 0.2711, 0.232, 0.1561, 0.2189, 0.1733, 0.0934, 0.0804, 0.146, 0.1525, 0.117, 0.1049, 0.0938, 0.0994, 0.1087, 0.0289, 0.0442, 0.0676, 0.0628, 0.1001, 0.0786, 0.0382, 0.0352, 0.054, 0.1137, 0.0806, 0.0806, 0.0806, 0.0614, 0.1869, 0.1758, 0.1878, 0.1908, 0.2331, 0.2403, 0.2761, 0.3564, 0.3312, 0.324, 0.3699, 0.4798, 0.502, 0.4867, 0.4645, 0.545, 0.5846, 0.5846, 0.5758, 0.6781, 0.6429, 0.6504, 0.7103, 0.7139, 0.6126, 0.7852, 0.8881, 0.8082, 0.8398, 0.7929, 0.8845, 0.8861, 1.0365, 1.1283]}
{"unit_id": "reply:1994383349260128598", "ticker_idx": 1, "ts": "2025-11-28T12:30:18+00:00", "symbol": "MTK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "My pick: Infineon MTK", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Picks Infineon and MediaTek as buys for semi exposure to industrial and mobile end markets.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MTK = MediaTek 2454.TW Taiwan; context mentions Infineon MTK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@OmerCheeema My pick:\n\nInfineon\n\nMTK", "reply_text": null, "parent_text": "Guys, it would be a good time to buy semi stocks with industrial and mobile semiconductors exposure. Source: A semiconductor industry leader who talks to customers on daily basis.", "quoted_text": null, "parent_author": "OmerCheeema", "ret_m1d": -0.0394265245763713, "ret_signed_m1d": -0.0394265245763713, "alpha_spy_m1d": -0.03399775069154021, "alpha_c_m1d": -0.026453823853446856, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03584230317869164, "ret_signed_p1d": 0.03584230317869164, "alpha_spy_p1d": 0.04040771878124283, "alpha_c_p1d": 0.03391203107575946, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.021505328386077327, "ret_signed_p1w": 0.021505328386077327, "alpha_spy_p1w": 0.018139740075941324, "alpha_c_p1w": -0.01295593339125034, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": 0.017921106988397772, "ret_signed_p1m": 0.017921106988397772, "alpha_spy_p1m": 0.008420662869065065, "alpha_c_p1m": -0.01916092084269705, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.36648741842932586, "ret_signed_p3m": 0.36648741842932586, "alpha_spy_p3m": 0.34920848624866996, "alpha_c_p3m": 0.15300373827905434, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 2.1019995140280687, "ret_signed_p6m": 2.1019995140280687, "alpha_spy_p6m": 2.0046968260215183, "alpha_c_p6m": 1.460937665374332, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.0251, 0.0, -0.0072, 0.0287, 0.0251, 0.0824, 0.0681, 0.0609, 0.0645, 0.0645, 0.1004, 0.0824, 0.0824, 0.0323, 0.0072, -0.0143, -0.0466, -0.0358, -0.0609, -0.0609, -0.0573, -0.0789, -0.0824, -0.0538, -0.0538, -0.0502, -0.043, -0.0358, -0.0358, -0.0573, -0.0645, -0.0573, -0.0466, -0.0466, -0.0394, -0.0394, -0.0466, -0.0717, -0.0717, -0.043, -0.0573, -0.0681, -0.0609, -0.0609, -0.0681, -0.0573, -0.0466, -0.0753, -0.0968, -0.1039, -0.1111, -0.1075, -0.1075, -0.1183, -0.1183, -0.1613, -0.1685, -0.1505, -0.1792, -0.1756, -0.1505, -0.0681, -0.0394, 0.0, 0.0358, 0.0143, 0.0036, 0.0072, 0.0215, 0.0323, 0.0179, 0.0466, 0.0, 0.0072, 0.0179, 0.0179, 0.0215, 0.0179, 0.0108, 0.0036, -0.0036, -0.0108, -0.0108, -0.0072, 0.0179, 0.0179, 0.0251, 0.0251, 0.0538, 0.0932, 0.0852, 0.0888, 0.0559, 0.0377, 0.0559, 0.0852, 0.0961, 0.0888, 0.0998, 0.0852, 0.0852, 0.0705, 0.0852, 0.1911, 0.2934, 0.2971, 0.3007, 0.3007, 0.2861, 0.2459, 0.3117, 0.3153, 0.2934, 0.2496, 0.3373, 0.3482, 0.3555, 0.3555, 0.3555, 0.3555, 0.3555, 0.3555, 0.3555, 0.3555, 0.3117, 0.3336, 0.3665, 0.4213, 0.4213, 0.3884, 0.3263, 0.2569, 0.2971, 0.2898, 0.2167, 0.2459, 0.2898, 0.3044, 0.2569, 0.2496, 0.2642, 0.2642, 0.2276, 0.2423, 0.1875, 0.1838, 0.1838, 0.1619, 0.1582, 0.1034, 0.0888, 0.0705, 0.0705, 0.0705, 0.0705, 0.0742, 0.1546, 0.1509, 0.1509, 0.1838, 0.2569, 0.308, 0.3848, 0.4067, 0.3884, 0.5273, 0.6771, 0.6186, 0.7794, 0.7794, 0.9109, 0.8817, 0.9072, 0.9072, 1.0972, 1.3055, 1.5064, 1.4991, 1.6526, 1.8353, 1.7037, 1.5539, 1.4882, 1.3822, 1.4845, 1.3055, 1.3603, 1.5941, 1.8207, 2.102]}
{"unit_id": "thread:1992514112346062942", "ticker_idx": 0, "ts": "2025-11-23T08:42:38+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The supply-demand imbalance in DRAM could continue for at least the next two years", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM shortage deepening with inventory at 2.7 weeks vs 8-week healthy level; supply-demand imbalance projected to persist 2+ years, bullish for SK Hynix and Samsung.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory Shortage to Worsen … DRAM Inventory Period Shortens from 3.3 Weeks to 2.7 Weeks\n\nThe semiconductor super-boom driven by investments in artificial intelligence (AI) is deepening the memory shortage.\n\nAccording to market research firm TrendForce on the 23rd, the average global DRAM inventory at the beginning of the fourth quarter stood at 2.7 weeks. This means that once produced, products are sold out in roughly 19 days, a further decrease compared to the third quarter (3.3 weeks).\n\nBy company, SK Hynix’s average inventory was only 2 weeks, and Samsung Electronics’ was 4 weeks. In terms of the decline, Samsung’s inventory fell by as much as 2 weeks from an average of 6 weeks.\n\nGiven that the industry views a healthy inventory level as around 8 weeks, this means products are selling so briskly that inventory does not even reach half of normal levels. This appears to be due to a combination of growing AI server replacement demand from global big tech companies and disrupted product supply caused by sanctions against China.\n\nA semiconductor industry official said, “Purchasing companies are practically offering higher prices almost every day,” adding, “Meanwhile, semiconductor manufacturers are conservatively managing their production capacity, so the shortage could intensify further.”\n\nKim Dong-won, a researcher at KB Securities, projected, “The supply-demand imbalance in DRAM could continue for at least the next two years.”\n\n---\n\nhttps://t.co/Vg40LiarSC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030559001381946265, "ret_signed_m1d": -0.030559001381946265, "alpha_spy_m1d": -0.016053986483978638, "alpha_c_m1d": 0.007745938563786765, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009214553834632285, "ret_signed_p1d": 0.009214553834632285, "alpha_spy_p1d": -0.00019134581747463388, "alpha_c_p1d": 0.00659002881552658, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.05052439497840333, "ret_signed_p1w": 0.05052439497840333, "alpha_spy_p1w": 0.03326770098984633, "alpha_c_p1w": 0.009712823309119045, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.17022187098231645, "ret_signed_p1m": 0.17022187098231645, "alpha_spy_p1m": 0.1384259244448216, "alpha_c_p1m": 0.09600027816626078, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.8531193694574041, "ret_signed_p3m": 0.8531193694574041, "alpha_spy_p3m": 0.8265427829989366, "alpha_c_p3m": 0.6397541694278676, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.118694645460851, "ret_signed_p6m": 2.118694645460851, "alpha_spy_p6m": 2.0152480154875225, "alpha_c_p6m": 1.509669435252411, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.4161, -0.4075, -0.4113, -0.4269, -0.4196, -0.4156, -0.4045, -0.3907, -0.3863, -0.3688, -0.3477, -0.3273, -0.2968, -0.2906, -0.2682, -0.2798, -0.2453, -0.2545, -0.2436, -0.2304, -0.2372, -0.2427, -0.2645, -0.2422, -0.2391, -0.2018, -0.1637, -0.1575, -0.1528, -0.1607, -0.1445, -0.1508, -0.1299, -0.1254, -0.142, -0.116, -0.0717, -0.0629, -0.0428, -0.0557, -0.056, -0.0529, -0.0065, 0.0222, 0.0073, 0.0415, 0.0564, 0.062, 0.1298, 0.0618, 0.0715, 0.0768, 0.0634, 0.1123, 0.1125, 0.1155, 0.0994, 0.0615, 0.0954, 0.0426, 0.0292, 0.0126, -0.0306, 0.0, 0.0092, 0.033, 0.0485, 0.0384, 0.0505, 0.0709, 0.0629, 0.0474, 0.0757, 0.1151, 0.1125, 0.1414, 0.117, 0.1006, 0.0702, 0.0405, 0.0611, 0.095, 0.1132, 0.1647, 0.1702, 0.1869, 0.1869, 0.2115, 0.2628, 0.2687, 0.2579, 0.2579, 0.3491, 0.3897, 0.4583, 0.47, 0.4527, 0.4726, 0.4763, 0.4536, 0.4584, 0.4802, 0.5409, 0.5474, 0.5231, 0.572, 0.6039, 0.614, 0.5784, 0.6766, 0.7502, 0.7579, 0.7569, 0.7053, 0.7866, 0.7281, 0.6621, 0.6753, 0.7165, 0.6921, 0.744, 0.8047, 0.8052, 0.8052, 0.7875, 0.819, 0.8531, 0.9049, 0.9055, 0.9599, 0.9968, 1.0802, 1.0459, 1.0463, 0.8447, 0.7389, 0.8597, 0.7972, 0.718, 0.854, 0.8954, 0.8522, 0.855, 0.9375, 0.9826, 1.0885, 1.0071, 0.9676, 0.8471, 0.8798, 0.8631, 0.7528, 0.7554, 0.6524, 0.6023, 0.7828, 0.6983, 0.7549, 0.8024, 0.8333, 1.0005, 0.9774, 1.0016, 1.0015, 1.1194, 1.1421, 1.1784, 1.1526, 1.1619, 1.2163, 1.2672, 1.2842, 1.2872, 1.3911, 1.3573, 1.3878, 1.3623, 1.3996, 1.5957, 1.6907, 1.9453, 1.9687, 2.1281, 2.3841, 2.2898, 2.4529, 2.4496, 2.1875, 2.1732, 2.1187]}
{"unit_id": "thread:1992514112346062942", "ticker_idx": 1, "ts": "2025-11-23T08:42:38+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's average inventory was only 2 weeks", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM shortage deepening with inventory at 2.7 weeks vs 8-week healthy level; supply-demand imbalance projected to persist 2+ years, bullish for SK Hynix and Samsung.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory Shortage to Worsen … DRAM Inventory Period Shortens from 3.3 Weeks to 2.7 Weeks\n\nThe semiconductor super-boom driven by investments in artificial intelligence (AI) is deepening the memory shortage.\n\nAccording to market research firm TrendForce on the 23rd, the average global DRAM inventory at the beginning of the fourth quarter stood at 2.7 weeks. This means that once produced, products are sold out in roughly 19 days, a further decrease compared to the third quarter (3.3 weeks).\n\nBy company, SK Hynix’s average inventory was only 2 weeks, and Samsung Electronics’ was 4 weeks. In terms of the decline, Samsung’s inventory fell by as much as 2 weeks from an average of 6 weeks.\n\nGiven that the industry views a healthy inventory level as around 8 weeks, this means products are selling so briskly that inventory does not even reach half of normal levels. This appears to be due to a combination of growing AI server replacement demand from global big tech companies and disrupted product supply caused by sanctions against China.\n\nA semiconductor industry official said, “Purchasing companies are practically offering higher prices almost every day,” adding, “Meanwhile, semiconductor manufacturers are conservatively managing their production capacity, so the shortage could intensify further.”\n\nKim Dong-won, a researcher at KB Securities, projected, “The supply-demand imbalance in DRAM could continue for at least the next two years.”\n\n---\n\nhttps://t.co/Vg40LiarSC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001923085035032246, "ret_signed_m1d": 0.001923085035032246, "alpha_spy_m1d": 0.016428099932999873, "alpha_c_m1d": 0.040228024980765276, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001923085035032246, "ret_signed_p1d": -0.001923085035032246, "alpha_spy_p1d": -0.011328984687139165, "alpha_c_p1d": -0.004547610054137952, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.03535624949908667, "ret_signed_p1w": 0.03535624949908667, "alpha_spy_p1w": 0.018099555510529663, "alpha_c_p1w": -0.0054553221701976184, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.12388131384195789, "ret_signed_p1m": 0.12388131384195789, "alpha_spy_p1m": 0.09208536730446304, "alpha_c_p1m": 0.049659721025902215, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.720462027311146, "ret_signed_p3m": 0.720462027311146, "alpha_spy_p3m": 0.6938854408526784, "alpha_c_p3m": 0.5070968272816094, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.3643692048480567, "ret_signed_p6m": 2.3643692048480567, "alpha_spy_p6m": 2.260922574874728, "alpha_c_p6m": 1.7553439946396165, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.5007, -0.4837, -0.4827, -0.5077, -0.499, -0.4952, -0.4894, -0.474, -0.4673, -0.4462, -0.4154, -0.4096, -0.3683, -0.3635, -0.3308, -0.3587, -0.3163, -0.3163, -0.325, -0.3058, -0.3125, -0.3144, -0.3529, -0.3288, -0.3317, -0.3077, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.1769, -0.2019, -0.2087, -0.1875, -0.1298, -0.1048, -0.0663, -0.0788, -0.074, -0.0798, -0.0192, 0.0288, 0.0019, 0.0731, 0.0923, 0.075, 0.1923, 0.1269, 0.1135, 0.1404, 0.1154, 0.1654, 0.1904, 0.1865, 0.1769, 0.0769, 0.1654, 0.0962, 0.0808, 0.0981, 0.0019, 0.0, -0.0019, 0.0077, 0.0469, 0.02, 0.0354, 0.0738, 0.0623, 0.0431, 0.0469, 0.1104, 0.0892, 0.1297, 0.0873, 0.0989, 0.0661, 0.02, 0.0604, 0.0623, 0.0527, 0.1162, 0.1239, 0.1316, 0.1316, 0.1527, 0.2317, 0.2528, 0.2528, 0.2528, 0.3029, 0.3394, 0.3972, 0.4279, 0.4549, 0.4318, 0.4414, 0.4202, 0.4279, 0.4414, 0.4549, 0.4703, 0.4299, 0.4241, 0.453, 0.4761, 0.4164, 0.5396, 0.6185, 0.657, 0.7493, 0.5973, 0.7455, 0.732, 0.6204, 0.6146, 0.707, 0.6858, 0.655, 0.7089, 0.6935, 0.6935, 0.6935, 0.6935, 0.7205, 0.8263, 0.8302, 0.9341, 0.9591, 1.1189, 1.0456, 1.0456, 0.8104, 0.6369, 0.8143, 0.7815, 0.6118, 0.8085, 0.8412, 0.793, 0.7545, 0.8779, 0.8702, 1.036, 0.9531, 0.9415, 0.7988, 0.901, 0.9184, 0.7988, 0.7988, 0.6831, 0.5559, 0.7294, 0.6002, 0.6889, 0.7082, 0.7661, 0.9916, 0.9241, 0.9801, 1.0051, 1.1266, 1.1902, 1.2268, 1.1748, 1.2481, 1.3599, 1.358, 1.3618, 1.356, 1.491, 1.5064, 1.4929, 1.4794, 1.4794, 1.7898, 1.7898, 2.0867, 2.1889, 2.2506, 2.6246, 2.5379, 2.8097, 2.7982, 2.507, 2.5475, 2.3644]}
{"unit_id": "thread:1992514112346062942", "ticker_idx": 2, "ts": "2025-11-23T08:42:38+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' was 4 weeks ... Samsung's inventory fell by as much as 2 weeks from an average of 6 weeks", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM shortage deepening with inventory at 2.7 weeks vs 8-week healthy level; supply-demand imbalance projected to persist 2+ years, bullish for SK Hynix and Samsung.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory Shortage to Worsen … DRAM Inventory Period Shortens from 3.3 Weeks to 2.7 Weeks\n\nThe semiconductor super-boom driven by investments in artificial intelligence (AI) is deepening the memory shortage.\n\nAccording to market research firm TrendForce on the 23rd, the average global DRAM inventory at the beginning of the fourth quarter stood at 2.7 weeks. This means that once produced, products are sold out in roughly 19 days, a further decrease compared to the third quarter (3.3 weeks).\n\nBy company, SK Hynix’s average inventory was only 2 weeks, and Samsung Electronics’ was 4 weeks. In terms of the decline, Samsung’s inventory fell by as much as 2 weeks from an average of 6 weeks.\n\nGiven that the industry views a healthy inventory level as around 8 weeks, this means products are selling so briskly that inventory does not even reach half of normal levels. This appears to be due to a combination of growing AI server replacement demand from global big tech companies and disrupted product supply caused by sanctions against China.\n\nA semiconductor industry official said, “Purchasing companies are practically offering higher prices almost every day,” adding, “Meanwhile, semiconductor manufacturers are conservatively managing their production capacity, so the shortage could intensify further.”\n\nKim Dong-won, a researcher at KB Securities, projected, “The supply-demand imbalance in DRAM could continue for at least the next two years.”\n\n---\n\nhttps://t.co/Vg40LiarSC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01964843134458627, "ret_signed_m1d": -0.01964843134458627, "alpha_spy_m1d": -0.005143416446618643, "alpha_c_m1d": 0.0036255541335565855, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.023273985478142856, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026887284286339463, "ret_signed_p1d": 0.026887284286339463, "alpha_spy_p1d": 0.017481384634232544, "alpha_c_p1d": 0.024205859810904284, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.002681424475435179, "ret_p1w": 0.04239921661394108, "ret_signed_p1w": 0.04239921661394108, "alpha_spy_p1w": 0.02514252262538408, "alpha_c_p1w": 0.018660256880602244, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.023738959733338838, "ret_p1m": 0.1530506263321878, "ret_signed_p1m": 0.1530506263321878, "alpha_spy_p1m": 0.12125467979469295, "alpha_c_p1m": 0.10788685638219397, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.045163769949993826, "ret_p3m": 0.9743910791481385, "ret_signed_p3m": 0.9743910791481385, "alpha_spy_p3m": 0.947814492689671, "alpha_c_p3m": 0.9703320180088377, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.004059061139300768, "ret_p6m": 1.8687926787213338, "ret_signed_p6m": 1.8687926787213338, "alpha_spy_p6m": 1.7653460487480053, "alpha_c_p6m": 1.6266596875507027, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.24213299117063114, "price_path": [-0.2731, -0.2834, -0.2824, -0.304, -0.2886, -0.2814, -0.2783, -0.2845, -0.2783, -0.2639, -0.2526, -0.2443, -0.2237, -0.2124, -0.1826, -0.1949, -0.1733, -0.1733, -0.1403, -0.128, -0.1208, -0.1136, -0.1424, -0.1293, -0.1324, -0.1107, -0.0719, -0.0719, -0.0719, -0.0719, -0.0719, -0.0719, -0.0238, -0.0352, -0.0527, -0.0176, 0.0103, 0.0124, 0.0145, 0.0083, 0.0196, -0.0021, 0.0217, 0.0548, 0.029, 0.0393, 0.0765, 0.1117, 0.1489, 0.0848, 0.0403, 0.0259, 0.0124, 0.0403, 0.0703, 0.0662, 0.0631, 0.0052, 0.0403, 0.0114, -0.0021, 0.0403, -0.0196, 0.0, 0.0269, 0.0631, 0.0703, 0.0393, 0.0424, 0.0693, 0.0807, 0.0869, 0.121, 0.1324, 0.121, 0.1169, 0.1096, 0.1262, 0.0838, 0.0631, 0.1158, 0.1127, 0.0993, 0.1427, 0.1531, 0.1489, 0.1489, 0.2099, 0.2418, 0.2459, 0.2459, 0.2459, 0.3353, 0.4351, 0.4434, 0.4652, 0.4423, 0.4444, 0.4423, 0.4299, 0.4579, 0.4953, 0.5473, 0.5515, 0.5089, 0.5535, 0.5826, 0.5806, 0.5806, 0.6574, 0.6876, 0.6699, 0.6678, 0.5629, 0.7406, 0.7572, 0.6554, 0.6481, 0.7292, 0.7229, 0.7437, 0.8559, 0.8829, 0.8829, 0.8829, 0.8829, 0.9744, 0.9754, 1.0056, 1.0783, 1.1147, 1.2654, 1.2498, 1.2498, 1.0274, 0.7894, 0.991, 0.9557, 0.8029, 0.9526, 0.9744, 0.9526, 0.9068, 0.9609, 1.0149, 1.1666, 1.0835, 1.0721, 0.9359, 0.9713, 0.964, 0.8715, 0.8715, 0.8358, 0.7411, 0.9748, 0.8577, 0.9389, 1.0108, 1.0462, 1.1919, 1.1243, 1.1451, 1.093, 1.1503, 1.1972, 1.2648, 1.2492, 1.2336, 1.2805, 1.2648, 1.3377, 1.2857, 1.3377, 1.3117, 1.3533, 1.2961, 1.2961, 1.421, 1.421, 1.7699, 1.8271, 1.7959, 1.9729, 1.9052, 1.9573, 2.0823, 1.8167, 1.9261, 1.8688]}
{"unit_id": "thread:1984120562873168364", "ticker_idx": 0, "ts": "2025-10-31T04:49:40+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung has already sold out all the HBM volumes it planned to produce", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish Samsung: 2026 HBM production already fully sold out with additional demand still coming in.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "@rwang07 Bro, what do you mean Samsung didn’t declare that next year’s HBM is sold out?\n\nListen to the conference call again — they clearly said everything’s already sold.\n\nhttps://t.co/Kh6EfVkz13\n\n---\n\nRegarding next year’s HBM sales outlook, Samsung stated:\n내년 HBM 판매 전망 관련, 2026년 HBM 비트 생산계획은 올해 대비 매우 대폭 확대 수립했으나, 해당 계획분에 대한 고객 수요를 이미 확보한 상태다. 다만 추가적인 고객 수요가 지속 접수되고 있어, 당사는 HBM 증산 가능성을 내부 검토 중이다.\n\n“The 2026 HBM bit production plan has been set at a significantly higher level compared to this year, and we have already secured customer demand for that planned volume. However, as additional customer requests continue to come in, we are internally reviewing the possibility of further HBM capacity expansion.”\n\nThis means that Samsung has already sold out all the HBM volumes it planned to produce.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031627900169409084, "ret_signed_m1d": -0.031627900169409084, "alpha_spy_m1d": -0.028358344618373277, "alpha_c_m1d": -0.030663475197112433, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": -0.0009644249722966514, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03348832614351105, "ret_signed_p1d": 0.03348832614351105, "alpha_spy_p1d": 0.031611639055546537, "alpha_c_p1d": 0.029397560278537016, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.004090765864974033, "ret_p1w": -0.08930227622821552, "ret_signed_p1w": -0.08930227622821552, "alpha_spy_p1w": -0.07304260458050893, "alpha_c_p1w": -0.04766325858196607, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04163901764624944, "ret_p1m": -0.0623255507623407, "ret_signed_p1m": -0.0623255507623407, "alpha_spy_p1m": -0.05970117670718644, "alpha_c_p1m": -0.014666917306523919, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.04765863345581678, "ret_p3m": 0.5180414411843266, "ret_signed_p3m": 0.5180414411843266, "alpha_spy_p3m": 0.4954410096713908, "alpha_c_p3m": 0.5239191818977161, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": -0.005877740713389534, "ret_p6m": 1.102867513345993, "ret_signed_p6m": 1.102867513345993, "alpha_spy_p6m": 1.0483530517949045, "alpha_c_p6m": 1.0318717496508736, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.07099576369511951, "price_path": [-0.3527, -0.3628, -0.3471, -0.3351, -0.3425, -0.3415, -0.3341, -0.3369, -0.3369, -0.3517, -0.3517, -0.3471, -0.3462, -0.3388, -0.3378, -0.349, -0.3462, -0.3554, -0.3545, -0.374, -0.3601, -0.3536, -0.3508, -0.3564, -0.3508, -0.3378, -0.3277, -0.3202, -0.3017, -0.2915, -0.2647, -0.2758, -0.2563, -0.2563, -0.2267, -0.2156, -0.2091, -0.2026, -0.2286, -0.2167, -0.2195, -0.2, -0.1651, -0.1651, -0.1651, -0.1651, -0.1651, -0.1651, -0.1219, -0.1321, -0.1479, -0.1163, -0.0912, -0.0893, -0.0874, -0.093, -0.0828, -0.1023, -0.0809, -0.0512, -0.0744, -0.0651, -0.0316, 0.0, 0.0335, -0.0242, -0.0642, -0.0772, -0.0893, -0.0642, -0.0372, -0.0409, -0.0437, -0.0958, -0.0642, -0.0902, -0.1023, -0.0642, -0.1181, -0.1005, -0.0763, -0.0437, -0.0372, -0.0651, -0.0623, -0.0381, -0.0279, -0.0223, 0.0084, 0.0186, 0.0084, 0.0047, -0.0019, 0.013, -0.0251, -0.0437, 0.0037, 0.0009, -0.0112, 0.0279, 0.0372, 0.0335, 0.0335, 0.0884, 0.117, 0.1208, 0.1208, 0.1208, 0.2012, 0.2909, 0.2984, 0.318, 0.2974, 0.2993, 0.2974, 0.2862, 0.3115, 0.3451, 0.3918, 0.3956, 0.3573, 0.3975, 0.4236, 0.4218, 0.4218, 0.4909, 0.518, 0.5022, 0.5003, 0.4059, 0.5657, 0.5807, 0.4891, 0.4825, 0.5554, 0.5498, 0.5685, 0.6695, 0.6938, 0.6938, 0.6938, 0.6938, 0.776, 0.777, 0.8041, 0.8695, 0.9022, 1.0378, 1.0237, 1.0237, 0.8237, 0.6096, 0.791, 0.7592, 0.6218, 0.7564, 0.776, 0.7564, 0.7153, 0.7639, 0.8125, 0.949, 0.8742, 0.8639, 0.7414, 0.7732, 0.7667, 0.6835, 0.6835, 0.6514, 0.5661, 0.7764, 0.6711, 0.7441, 0.8087, 0.8406, 0.9717, 0.9108, 0.9296, 0.8827, 0.9343, 0.9764, 1.0373, 1.0232, 1.0092, 1.0513, 1.0373, 1.1029, 1.056, 1.1029]}
{"unit_id": "orig:1976976204122280203", "ticker_idx": 0, "ts": "2025-10-11T11:40:32+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CFO Conference Call Suggests Strong AI Supply Expansion Visibility, Maintains 'Buy' Rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA reiterates Buy on AMD after CFO call; OpenAI 6GW deal ($15–20B/GW) incremental to existing outlook, with operating margin expansion to mid-30s.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA) AMD CFO Conference Call: Positive on OpenAI Deal, Opens Opportunity for Market Share Expansion\n\nCFO Conference Call Suggests Strong AI Supply Expansion Visibility, Maintains ‘Buy’ Rating\n\nThis morning, we hosted a conference call with AMD’s Chief Financial Officer (CFO) Jean Hu and Vice President of Financial Strategy and Investor Relations (IR) Matthew Ramsay, attended by many investors. The key points are as follows:\n\n1. OpenAI’s first 1GW buildout will be shipped in the second half of 2026 using the MI450 series racks, with most deliveries concentrated in Q4.\n    \n2. AMD’s supply contribution will initially include GPU, CPU, and DPU components, valued at approximately $15–20 billion per GW (including <5% impact from warrant-related revenue deductions).\n    \n3. The total value of AMD’s contribution may increase over time, supported by higher average selling prices (ASP) and potential inclusion of networking components.\n    \n4. AMD will supply products to cloud service providers (CSPs) for the buildout and invoice them accordingly; this deal could lead to further AMD-based system deployments by CSPs.\n    \n5. OpenAI is financially incentivized to complete the full 6GW buildout before the warrant expiry in October 2030.\n    \n6. This transaction and opportunity are incremental to what AMD has already announced.\n    \n7. The choice and combination of accelerators (NVIDIA vs. AMD vs. Broadcom) are determined by the total cost of ownership (TCO) for each workload.\n    \n8. UALink is based on what AMD has developed over five to six generations with Infinity Fabric, representing a proven concept.\n\n⸻\n\nInitial Supply per GW: $15–20 Billion, with Potential for Expansion\n\nAccording to AMD, its supply contribution to OpenAI’s rack-level solution will initially include AMD-based GPUs, CPUs, and DPUs. The net value (excluding warrant-related deductions) amounts to $15–20 billion per GW, with the warrant effect being less than 5%. As next-generation GPUs deliver significantly higher token generation per dollar with only minimal increases in power consumption, the total value could increase due to ASP expansion. This aligns with our previous assumption of $17.5 billion per GW, based on a GPU ASP of over $40,000 (see Figure 1).\nAMD also noted that NVIDIA’s $35 billion per GW opportunity appears higher primarily because it includes roughly 30% additional elements—such as networking, cabling, power supply units, and OEM margins.\n\n⸻\n\nGPU Margins Slightly Dilutive to GM but Accretive to OpM\n\nAMD indicated that its data center GPU products still slightly lag the company-wide gross margin average, as the firm continues prioritizing market share expansion during the early phase of AI infrastructure buildouts. However, AMD’s overall data center portfolio—covering CPU, DPU, and GPU silicon—significantly contributes to overall gross margin.\nThat said, since GPUs include high-cost HBM (High Bandwidth Memory) components in their bill of materials (BOM), there remains limited room to expand margins on that portion, making it a continued drag on gross margin.\nAMD expects limited incremental operating expenses (opex) associated with GPU supply expansion, which should help it reach its long-term operating margin target in the mid-30% range (currently mid-20%). Separately, AMD confirmed that no separate HBM purchases are expected during the 6GW project period.\n\nWarrant Conditions = OpenAI Buildout + AMD Stock Price\n\nThe two fundamental milestones for warrant vesting are:\n    \n1. OpenAI’s capacity purchases and actual buildout per GW, and\n    \n2. AMD’s share price performance.\n\nTo satisfy the first condition, AMD and OpenAI will work together to deploy AMD-based rack solutions across multiple CSPs (selected by OpenAI), potentially unlocking further opportunities for both first-party and third-party AMD cluster buildouts in the future.\nFor the second condition, each warrant tranche requires AMD’s stock to reach progressively higher price targets, with the final tranche set at $600 per share. The warrant expires on October 5, 2030, thereby incentivizing OpenAI to complete the full 6GW buildout before that date in order to receive the entire 10% warrant allocation.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007023400264860813, "ret_signed_m1d": -0.007023400264860813, "alpha_spy_m1d": 0.008088791824155894, "alpha_c_m1d": 0.03539012370986927, "bench_spy_m1d": -0.015112192089016707, "bench_c_m1d": -0.04241352397473008, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0077164688248525515, "ret_signed_p1d": 0.0077164688248525515, "alpha_spy_p1d": 0.008938123628309258, "alpha_c_p1d": 0.026112283570661754, "bench_spy_p1d": -0.0012216548034567065, "bench_c_p1d": -0.018395814745809203, "ret_p1w": 0.11154236943853912, "ret_signed_p1w": 0.11154236943853912, "alpha_spy_p1w": 0.09908456402159804, "alpha_c_p1w": 0.089555584683916, "bench_spy_p1w": 0.012457805416941081, "bench_c_p1w": 0.021986784754623123, "ret_p1m": 0.09749563941426609, "ret_signed_p1m": 0.09749563941426609, "alpha_spy_p1m": 0.06739188392741968, "alpha_c_p1m": 0.06423587228039063, "bench_spy_p1m": 0.030103755486846406, "bench_c_p1m": 0.03325976713387546, "ret_p3m": -0.054246398634563264, "ret_signed_p3m": -0.054246398634563264, "alpha_spy_p3m": -0.09724150072087501, "alpha_c_p3m": -0.17318459816144982, "bench_spy_p3m": 0.04299510208631174, "bench_c_p3m": 0.11893819952688656, "ret_p6m": 0.02361149918485128, "ret_signed_p6m": 0.02361149918485128, "alpha_spy_p6m": 0.02371166821314319, "alpha_c_p6m": -0.1570929342674121, "bench_spy_p6m": -0.00010016902829190766, "bench_c_p6m": 0.1807044334522634, "price_path": [-0.2603, -0.2588, -0.2746, -0.2746, -0.2851, -0.2669, -0.2509, -0.2308, -0.1976, -0.1801, -0.1705, -0.1853, -0.2066, -0.1832, -0.1946, -0.2463, -0.2034, -0.2017, -0.204, -0.1916, -0.1479, -0.1639, -0.1798, -0.1861, -0.2304, -0.2367, -0.2436, -0.2248, -0.2452, -0.2301, -0.2278, -0.2211, -0.2485, -0.2485, -0.25, -0.2509, -0.2524, -0.3016, -0.3004, -0.28, -0.2628, -0.2807, -0.2673, -0.2553, -0.2586, -0.2646, -0.2703, -0.2728, -0.2617, -0.2565, -0.2566, -0.2548, -0.2632, -0.2544, -0.2524, -0.2422, -0.2157, -0.2391, -0.0587, -0.0227, 0.0884, 0.0761, -0.007, 0.0, 0.0077, 0.1025, 0.0838, 0.077, 0.1115, 0.0999, 0.0638, 0.0858, 0.1687, 0.1998, 0.1922, 0.2214, 0.1775, 0.1834, 0.1998, 0.1554, 0.1844, 0.0983, 0.0791, 0.1273, 0.0975, 0.1962, 0.1457, 0.1404, 0.1114, 0.0641, 0.0329, -0.0481, -0.0584, -0.0063, -0.0475, -0.0101, -0.0101, 0.0051, 0.0154, -0.0055, 0.0055, -0.002, 0.0072, 0.0217, 0.024, 0.0231, 0.0231, -0.0261, -0.0408, -0.0335, -0.0846, -0.071, -0.0138, -0.0068, -0.007, -0.0064, -0.0064, -0.0066, -0.0037, -0.005, -0.0104, -0.0104, 0.0326, 0.0215, -0.0096, -0.0296, -0.0542, -0.0612, -0.0403, 0.021, 0.0332, 0.0531, 0.0712, 0.0712, 0.0716, 0.1542, 0.1724, 0.1999, 0.1612, 0.1645, 0.1678, 0.1652, 0.0938, 0.1379, 0.1187, -0.075, -0.1105, -0.0369, -0.0019, -0.0132, -0.0131, -0.0484, -0.042, -0.042, -0.0616, -0.0753, -0.0603, -0.0752, -0.0916, -0.0119, -0.0257, -0.0589, -0.0749, -0.0822, -0.1177, -0.0663, -0.0784, -0.1108, -0.0635, -0.0609, -0.0536, -0.0863, -0.1064, -0.0917, -0.0929, -0.0784, -0.0515, -0.0697, -0.0635, -0.0511, 0.0178, -0.0585, -0.0667, -0.0942, -0.06, -0.0287, 0.005, 0.005, 0.0174, 0.0236]}
{"unit_id": "thread:1985300102685368760", "ticker_idx": 0, "ts": "2025-11-03T10:56:44+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's HBM4 delayed to 2027 due to redesign", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Bearish MU: HBM4 delayed to 2027 due to redesign, author reacts with mockery implying negative view.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "LMAO\n\nMicron’s HBM4 delayed to 2027 due to redesign.\n\n$MU https://t.co/ZztYk9QiA7\n\n---\n\n@insane_analyst", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04657011762660468, "ret_signed_m1d": 0.04657011762660468, "alpha_spy_m1d": 0.044696945895840945, "alpha_c_m1d": 0.0381570636252051, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07102688124746559, "ret_signed_p1d": 0.07102688124746559, "alpha_spy_p1d": 0.05917334130501806, "alpha_c_p1d": 0.03458876560499491, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": 0.07925005947178643, "ret_signed_p1w": -0.07925005947178643, "alpha_spy_p1w": -0.0820304884068237, "alpha_c_p1w": -0.09921727621364096, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": 0.020409038358131015, "ret_signed_p1m": -0.020409038358131015, "alpha_spy_p1m": -0.0230577802573797, "alpha_c_p1m": -0.0386281379883473, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": 0.857546071764878, "ret_signed_p3m": -0.857546071764878, "alpha_spy_p3m": -0.838886575723798, "alpha_c_p3m": -0.713544067609567, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 1.1504288735653243, "ret_signed_p6m": -1.1504288735653243, "alpha_spy_p6m": -1.1030113003525872, "alpha_c_p6m": -0.8045167508589754, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.5368, -0.5236, -0.4938, -0.4732, -0.456, -0.4708, -0.4665, -0.4853, -0.4739, -0.4803, -0.5009, -0.507, -0.4989, -0.5043, -0.5039, -0.4986, -0.4805, -0.4932, -0.4932, -0.4955, -0.4945, -0.4711, -0.4406, -0.4402, -0.4241, -0.4039, -0.3589, -0.3305, -0.3282, -0.3237, -0.3187, -0.2809, -0.3071, -0.299, -0.2914, -0.3114, -0.3322, -0.3303, -0.3021, -0.2875, -0.2244, -0.2176, -0.1997, -0.1864, -0.2088, -0.1626, -0.1805, -0.2262, -0.1787, -0.203, -0.1822, -0.1371, -0.1377, -0.119, -0.1381, -0.1544, -0.1193, -0.0668, -0.0622, -0.0545, -0.0344, -0.0455, -0.0466, 0.0, -0.071, 0.0119, 0.0155, 0.0137, 0.0793, 0.0273, 0.0435, 0.0096, 0.0517, 0.0309, -0.0264, -0.0374, -0.142, -0.1164, -0.0459, -0.0433, -0.0189, -0.0189, 0.0076, 0.0245, 0.0204, -0.0023, -0.0343, 0.0107, 0.0521, 0.0755, 0.1236, 0.1012, 0.0274, 0.0119, -0.0093, -0.0391, 0.059, 0.133, 0.1785, 0.1771, 0.2215, 0.2215, 0.2134, 0.2547, 0.2473, 0.2166, 0.2166, 0.3445, 0.3305, 0.4639, 0.4473, 0.3939, 0.4709, 0.4743, 0.4413, 0.4209, 0.4349, 0.5462, 0.5462, 0.5558, 0.6586, 0.6947, 0.7035, 0.6585, 0.7486, 0.8554, 0.8575, 0.7684, 0.8661, 0.7879, 0.6172, 0.6321, 0.6824, 0.6347, 0.591, 0.7491, 0.7645, 0.7547, 0.7547, 0.7041, 0.7943, 0.7789, 0.8251, 0.7944, 0.7818, 0.8286, 0.7713, 0.7577, 0.759, 0.6184, 0.7083, 0.6924, 0.5784, 0.6595, 0.7182, 0.7847, 0.7278, 0.8164, 0.8832, 0.9679, 0.9681, 0.8937, 0.8026, 0.7235, 0.6859, 0.6287, 0.5151, 0.5226, 0.3722, 0.4406, 0.5686, 0.5617, 0.5617, 0.6109, 0.6101, 0.7344, 0.7974, 0.7935, 0.819, 0.9857, 0.9455, 0.9498, 0.9405, 0.9122, 0.9163, 1.0787, 1.0542, 1.1181, 1.2369, 1.1504]}
{"unit_id": "thread:1977977365159633237", "ticker_idx": 0, "ts": "2025-10-14T05:58:47+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung is projected to maintain its No. 1 position in the memory market in the fourth quarter as well", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung: projected to hold memory market No.1 in Q4 on commodity DRAM/NAND momentum and upcoming HBM3E/HBM4 ramp.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung regains the throne in the memory market, overtaking SK Hynix to reclaim the No. 1 spot\n\nIn the third quarter of 2025, Samsung Electronics recorded $19.4 billion in memory sales, surpassing SK Hynix (which posted $17.5 billion) to reclaim the top position in the global memory market.\n\nSamsung’s recovery was driven by strong demand for commodity DRAM and NAND. Although its HBM business underperformed earlier this year, it is expected to enter a full-fledged growth phase starting next year, supported by HBM3E and HBM4.\n\nWith continued momentum in both commodity DRAM and HBM, Samsung is projected to maintain its No. 1 position in the memory market in the fourth quarter as well.\n\n---\n\nhttps://t.co/LAiSLP1C6v", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018558947933519976, "ret_signed_m1d": 0.018558947933519976, "alpha_spy_m1d": 0.017335798864127394, "alpha_c_m1d": 0.005557693029785815, "bench_spy_m1d": 0.0012231490693925817, "bench_c_m1d": 0.01300125490373416, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03711789586703995, "ret_signed_p1d": 0.03711789586703995, "alpha_spy_p1d": 0.03267828000036066, "alpha_c_p1d": 0.027207156656559706, "bench_spy_p1d": 0.004439615866679292, "bench_c_p1d": 0.009910739210480246, "ret_p1w": 0.0644104562538288, "ret_signed_p1w": 0.0644104562538288, "alpha_spy_p1w": 0.05072946532140854, "alpha_c_p1w": 0.04014854923497735, "bench_spy_p1w": 0.013680990932420256, "bench_c_p1w": 0.024261907018851447, "ret_p1m": 0.1255458293080054, "ret_signed_p1m": 0.1255458293080054, "alpha_spy_p1m": 0.09360825403141115, "alpha_c_p1m": 0.08146211774975942, "bench_spy_p1m": 0.03193757527659424, "bench_c_p1m": 0.044083711558245975, "ret_p3m": 0.5248438699520155, "ret_signed_p3m": 0.5248438699520155, "alpha_spy_p3m": 0.4736669087563177, "alpha_c_p3m": 0.48494001836286205, "bench_spy_p3m": 0.05117696119569781, "bench_c_p3m": 0.03990385158915344, "ret_p6m": 1.313985622955864, "ret_signed_p6m": 1.313985622955864, "alpha_spy_p6m": 1.2873646121171467, "alpha_c_p6m": 1.3045625395104183, "bench_spy_p6m": 0.026621010838717263, "bench_c_p6m": 0.009423083445445624, "price_path": [-0.2751, -0.2707, -0.2631, -0.2827, -0.2783, -0.2827, -0.2838, -0.2349, -0.2327, -0.2109, -0.224, -0.2512, -0.2425, -0.2403, -0.2522, -0.2338, -0.2196, -0.2283, -0.2272, -0.2186, -0.2218, -0.2218, -0.2392, -0.2392, -0.2338, -0.2327, -0.224, -0.2229, -0.2359, -0.2327, -0.2435, -0.2425, -0.2653, -0.249, -0.2414, -0.2381, -0.2446, -0.2381, -0.2229, -0.2109, -0.2022, -0.1805, -0.1686, -0.137, -0.1501, -0.1273, -0.1273, -0.0925, -0.0794, -0.0718, -0.0642, -0.0947, -0.0808, -0.0841, -0.0611, -0.0202, -0.0202, -0.0202, -0.0202, -0.0202, -0.0202, 0.0306, 0.0186, 0.0, 0.0371, 0.0666, 0.0688, 0.071, 0.0644, 0.0764, 0.0535, 0.0786, 0.1135, 0.0862, 0.0972, 0.1365, 0.1736, 0.2129, 0.1452, 0.0983, 0.083, 0.0688, 0.0983, 0.1299, 0.1255, 0.1223, 0.0611, 0.0983, 0.0677, 0.0535, 0.0983, 0.0349, 0.0557, 0.0841, 0.1223, 0.1299, 0.0972, 0.1004, 0.1288, 0.1408, 0.1474, 0.1834, 0.1954, 0.1834, 0.179, 0.1714, 0.1889, 0.1441, 0.1223, 0.1779, 0.1747, 0.1605, 0.2063, 0.2172, 0.2129, 0.2129, 0.2773, 0.3109, 0.3153, 0.3153, 0.3153, 0.4097, 0.515, 0.5237, 0.5468, 0.5226, 0.5248, 0.5226, 0.5095, 0.5391, 0.5786, 0.6334, 0.6378, 0.5929, 0.64, 0.6707, 0.6686, 0.6686, 0.7497, 0.7815, 0.7629, 0.7607, 0.6499, 0.8375, 0.855, 0.7475, 0.7399, 0.8254, 0.8188, 0.8408, 0.9593, 0.9878, 0.9878, 0.9878, 0.9878, 1.0843, 1.0854, 1.1172, 1.194, 1.2324, 1.3915, 1.375, 1.375, 1.1403, 0.8891, 1.1019, 1.0646, 0.9033, 1.0613, 1.0843, 1.0613, 1.013, 1.0701, 1.1271, 1.2873, 1.1995, 1.1874, 1.0437, 1.081, 1.0733, 0.9757, 0.9757, 0.938, 0.838, 1.0848, 0.9611, 1.0469, 1.1227, 1.1601, 1.314]}
{"unit_id": "quote:1976243342024507862", "ticker_idx": 0, "ts": "2025-10-09T11:08:24+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Jensen Huang is doing whatever it takes to eliminate his competitors.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Endorses bullish NVDA thesis; Jensen's strategy to close cost gap vs ASICs strengthens competitive moat.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Why didn’t I think of this sooner…\n\n“The core of this vision is to eliminate the margin stacking by server ODMs (original design manufacturers) and CSPs (cloud service providers), narrowing the cost gap between NVIDIA chips and ASICs to around 15% on average.”\n\nJensen Huang is doing whatever it takes to eliminate his competitors.", "reply_text": null, "parent_text": null, "quoted_text": "Cantor Fitzgerald) NVIDIA: Raises Target Price to $300 Amid Accelerating AI Infrastructure Growth\n\n“Last week, we hosted an investor meeting in New York with NVIDIA’s senior management, including CEO Jensen Huang, CFO Colette Kress, IR head Toshiya Hari, and other executives.", "parent_author": null, "ret_m1d": -0.017967518834077323, "ret_signed_m1d": -0.017967518834077323, "alpha_spy_m1d": -0.020872944255104775, "alpha_c_m1d": -0.020458839029821707, "bench_spy_m1d": 0.002905425421027452, "bench_c_m1d": 0.0024913201957443842, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04886534322137259, "ret_signed_p1d": -0.04886534322137259, "alpha_spy_p1d": -0.021837515322282952, "alpha_c_p1d": 0.008693783732913518, "bench_spy_p1d": -0.02702782789908964, "bench_c_p1d": -0.05755912695428611, "ret_p1w": -0.05587578994716014, "ret_signed_p1w": -0.05587578994716014, "alpha_spy_p1w": -0.04020151001351813, "alpha_c_p1w": -0.05034290246324635, "bench_spy_p1w": -0.01567427993364201, "bench_c_p1w": -0.005532887483913784, "ret_p1m": -0.02295266728066714, "ret_signed_p1m": -0.02295266728066714, "alpha_spy_p1m": -0.022669514513621647, "alpha_c_p1m": -0.031382399529811145, "bench_spy_p1m": -0.0002831527670454914, "bench_c_p1m": 0.008429732249144006, "ret_p3m": -0.027624079534145696, "ret_signed_p3m": -0.027624079534145696, "alpha_spy_p3m": -0.0614375135049966, "alpha_c_p3m": -0.15397074389281917, "bench_spy_p3m": 0.033813433970850904, "bench_c_p3m": 0.12634666435867348, "ret_p6m": -0.07872733891956518, "ret_signed_p6m": -0.07872733891956518, "alpha_spy_p6m": -0.061450264333961946, "alpha_c_p6m": -0.21873130460046342, "bench_spy_p6m": -0.017277074585603236, "bench_c_p6m": 0.14000396568089823, "price_path": [-0.148, -0.1136, -0.1101, -0.1017, -0.1047, -0.1101, -0.1327, -0.1132, -0.0978, -0.0991, -0.0822, -0.0886, -0.0691, -0.0764, -0.0979, -0.0653, -0.0744, -0.0683, -0.0613, -0.0513, -0.0546, -0.0489, -0.0571, -0.0548, -0.063, -0.0549, -0.088, -0.0892, -0.0914, -0.0758, -0.0663, -0.0561, -0.057, -0.0644, -0.0955, -0.0955, -0.1132, -0.114, -0.1086, -0.1327, -0.126, -0.1133, -0.0792, -0.08, -0.0766, -0.077, -0.0919, -0.1157, -0.0848, -0.0826, -0.0465, -0.0734, -0.081, -0.0773, -0.0747, -0.0557, -0.0311, -0.0277, -0.0191, -0.0257, -0.0365, -0.0391, -0.018, 0.0, -0.0489, -0.0221, -0.0651, -0.0662, -0.0559, -0.0486, -0.0516, -0.0593, -0.0638, -0.0541, -0.0328, -0.0056, 0.0439, 0.0751, 0.0536, 0.0515, 0.0743, 0.0318, 0.0137, -0.0233, -0.023, 0.0337, 0.0031, 0.0064, -0.0297, -0.0125, -0.031, -0.0582, -0.0314, -0.062, -0.0711, -0.052, -0.0766, -0.0639, -0.0639, -0.0809, -0.0657, -0.0577, -0.0674, -0.0477, -0.0527, -0.0364, -0.0394, -0.0456, -0.0604, -0.0911, -0.0845, -0.0771, -0.1123, -0.0957, -0.0601, -0.0461, -0.0174, -0.0205, -0.0205, -0.0105, -0.0225, -0.0261, -0.0315, -0.0315, -0.0193, -0.0231, -0.0276, -0.0179, -0.039, -0.04, -0.0396, -0.0351, -0.0489, -0.0286, -0.0329, -0.0329, -0.0752, -0.048, -0.0401, -0.0254, -0.0316, -0.021, -0.0054, -0.0003, -0.0074, -0.0361, -0.0635, -0.0954, -0.1074, -0.0371, -0.0131, -0.0209, -0.013, -0.0292, -0.0506, -0.0506, -0.0394, -0.0238, -0.0242, -0.0142, -0.0052, 0.0015, 0.0156, -0.0398, -0.0798, -0.0523, -0.065, -0.0494, -0.0479, -0.0765, -0.0515, -0.0405, -0.0339, -0.0489, -0.0639, -0.0484, -0.0551, -0.0631, -0.0727, -0.1031, -0.0878, -0.0901, -0.072, -0.1107, -0.13, -0.1422, -0.0943, -0.0872, -0.0787, -0.0787]}
{"unit_id": "orig:1992822415504925071", "ticker_idx": 0, "ts": "2025-11-24T05:07:43+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU $338 PT", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses $338 price target on MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "und", "tweet_text": "$MU $338 PT https://t.co/SUHJr2RL57", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07395165783628477, "ret_signed_m1d": -0.07395165783628477, "alpha_spy_m1d": -0.05944664293831714, "alpha_c_m1d": -0.03564671789055174, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002679462252589637, "ret_signed_p1d": 0.002679462252589637, "alpha_spy_p1d": -0.006726437399517282, "alpha_c_p1d": 5.4937233483931536e-05, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.07381771882218224, "ret_signed_p1w": 0.07381771882218224, "alpha_spy_p1w": 0.05656102483362524, "alpha_c_p1w": 0.03300614715289796, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.23373367277280366, "ret_signed_p1m": 0.23373367277280366, "alpha_spy_p1m": 0.2019377262353088, "alpha_c_p1m": 0.159512079956748, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.8645050019129275, "ret_signed_p3m": 0.8645050019129275, "alpha_spy_p3m": 0.83792841545446, "alpha_c_p3m": 0.6511398018833909, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.1229220528131627, "ret_signed_p6m": 2.1229220528131627, "alpha_spy_p6m": 2.019475422839834, "alpha_c_p6m": 1.5138968426047226, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.4745, -0.4555, -0.4689, -0.4689, -0.4712, -0.4702, -0.4457, -0.4137, -0.4133, -0.3964, -0.3752, -0.328, -0.2983, -0.2959, -0.2912, -0.286, -0.2463, -0.2738, -0.2653, -0.2573, -0.2783, -0.3001, -0.2981, -0.2685, -0.2533, -0.1871, -0.1799, -0.1612, -0.1472, -0.1708, -0.1223, -0.1411, -0.189, -0.1392, -0.1646, -0.1429, -0.0956, -0.0962, -0.0766, -0.0966, -0.1137, -0.0769, -0.0219, -0.0171, -0.009, 0.0121, 0.0004, -0.0007, 0.0481, -0.0263, 0.0606, 0.0643, 0.0625, 0.1312, 0.0767, 0.0936, 0.0581, 0.1023, 0.0805, 0.0204, 0.0089, -0.1007, -0.074, 0.0, 0.0027, 0.0283, 0.0283, 0.056, 0.0738, 0.0695, 0.0457, 0.0121, 0.0593, 0.1027, 0.1272, 0.1776, 0.1542, 0.0769, 0.0606, 0.0383, 0.0071, 0.1099, 0.1875, 0.2352, 0.2337, 0.2802, 0.2802, 0.2718, 0.3151, 0.3073, 0.2751, 0.2751, 0.4091, 0.3945, 0.5343, 0.5169, 0.461, 0.5417, 0.5452, 0.5106, 0.4892, 0.5039, 0.6206, 0.6206, 0.6306, 0.7383, 0.7762, 0.7854, 0.7383, 0.8327, 0.9446, 0.9469, 0.8535, 0.9559, 0.8738, 0.695, 0.7106, 0.7633, 0.7133, 0.6675, 0.8332, 0.8494, 0.8391, 0.8391, 0.786, 0.8806, 0.8645, 0.9128, 0.8807, 0.8675, 0.9166, 0.8565, 0.8423, 0.8436, 0.6962, 0.7904, 0.7738, 0.6543, 0.7393, 0.8009, 0.8705, 0.8109, 0.9037, 0.9737, 1.0626, 1.0628, 0.9848, 0.8893, 0.8064, 0.767, 0.707, 0.588, 0.5959, 0.4382, 0.5099, 0.6441, 0.6369, 0.6369, 0.6883, 0.6875, 0.8178, 0.8839, 0.8798, 0.9065, 1.0812, 1.0391, 1.0435, 1.0339, 1.0042, 1.0084, 1.1787, 1.153, 1.22, 1.3444, 1.2539, 1.3172, 1.3114, 1.4233, 1.5764, 1.8613, 1.9792, 1.89, 2.3378, 2.5546, 2.4261, 2.5917, 2.4683, 2.2388, 2.046, 2.1229]}
{"unit_id": "orig:1981634554201788474", "ticker_idx": 0, "ts": "2025-10-24T08:11:09+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I've been waiting for SanDisk to pull back. I think it's a bit too expensive to buy right now. But… the correction just isn't coming.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Wants to buy SNDK on NAND shortage thesis but finds it too expensive; implicitly bullish, waiting for a pullback that isn't arriving.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "NAND supply is expected to enter a shortage starting in Q4 2025.\n\nI’ve been waiting for SanDisk to pull back.\n\nI think it’s a bit too expensive to buy right now.\n\nBut… the correction just isn’t coming.🥲\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10265363254417026, "ret_signed_m1d": -0.10265363254417026, "alpha_spy_m1d": -0.09454731092933377, "alpha_c_m1d": -0.08735968241193537, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.01529395013223489, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.051944552958308354, "ret_signed_p1d": -0.051944552958308354, "alpha_spy_p1d": -0.0637422649814573, "alpha_c_p1d": -0.07040626236835668, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.01846170941004832, "ret_p1w": 0.07074558395932118, "ret_signed_p1w": 0.07074558395932118, "alpha_spy_p1w": 0.06364331908538245, "alpha_c_p1w": 0.04656137318064246, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.024184210778678716, "ret_p1m": 0.21916632063357744, "ret_signed_p1m": 0.21916632063357744, "alpha_spy_p1m": 0.23174666760506113, "alpha_c_p1m": 0.266410755738396, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.04724443510481857, "ret_p3m": 1.69279114032889, "ret_signed_p3m": 1.69279114032889, "alpha_spy_p3m": 1.6777667245656522, "alpha_c_p3m": 1.7116063170542088, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": -0.01881517672531874, "ret_p6m": 3.904490769072133, "ret_signed_p6m": 3.904490769072133, "alpha_spy_p6m": 3.852064922487254, "alpha_c_p6m": 3.848649812894575, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.05584095617755791, "price_path": [-0.7694, -0.7669, -0.7694, -0.778, -0.7716, -0.7748, -0.7739, -0.7814, -0.7618, -0.767, -0.7484, -0.7475, -0.7492, -0.7607, -0.7555, -0.7605, -0.7615, -0.7556, -0.7509, -0.7487, -0.7456, -0.7398, -0.7267, -0.7181, -0.7181, -0.7257, -0.7152, -0.6643, -0.6318, -0.6213, -0.6212, -0.6029, -0.5472, -0.5373, -0.5161, -0.5082, -0.4952, -0.4689, -0.451, -0.4471, -0.4284, -0.4637, -0.4935, -0.4783, -0.3903, -0.3973, -0.3494, -0.3332, -0.3102, -0.3491, -0.3503, -0.2916, -0.3034, -0.372, -0.2769, -0.3162, -0.2249, -0.225, -0.2471, -0.2048, -0.1981, -0.2106, -0.1027, 0.0, -0.0519, -0.0571, 0.0978, 0.0519, 0.0707, 0.112, 0.0452, 0.163, 0.1157, 0.2864, 0.4394, 0.4589, 0.5207, 0.3084, 0.3653, 0.4282, 0.3157, 0.3212, 0.0526, 0.0758, 0.2192, 0.1845, 0.1551, 0.1551, 0.1994, 0.129, 0.1031, 0.0442, 0.1458, 0.2273, 0.2112, 0.1789, 0.2509, 0.2979, 0.1075, 0.0844, 0.1244, 0.111, 0.1789, 0.2764, 0.2949, 0.3155, 0.3434, 0.3434, 0.3432, 0.312, 0.2904, 0.2751, 0.2751, 0.4785, 0.4723, 0.8781, 0.8992, 0.7971, 1.0273, 1.0911, 1.094, 1.0832, 1.1983, 1.2219, 1.2219, 1.434, 1.6928, 1.7043, 1.5453, 1.529, 1.5861, 1.8343, 1.897, 2.0955, 2.5735, 2.7361, 2.14, 2.0952, 2.212, 2.1339, 1.9095, 2.2195, 2.3857, 2.3657, 2.3657, 2.1725, 2.2252, 2.3363, 2.4915, 2.5802, 2.43, 2.397, 2.5018, 2.413, 2.3255, 2.0372, 2.218, 2.0382, 1.8327, 2.1625, 2.3245, 2.5208, 2.3241, 2.554, 2.7797, 2.8686, 3.0486, 3.1475, 2.8124, 2.7736, 2.7735, 2.6413, 2.2401, 2.3081, 2.0753, 2.4129, 2.7212, 2.7687, 2.7687, 2.8925, 2.8182, 3.1948, 3.5744, 3.5755, 4.1166, 4.0734, 3.7901, 3.9391, 3.9473, 3.9045]}
{"unit_id": "reply:1988242654249763290", "ticker_idx": 0, "ts": "2025-11-11T13:49:23+00:00", "symbol": "ASE Technology Holding", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I have a positive view on ASE.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Positive/bullish on ASE for OSAT demand tailwinds; prefers ASE over AMKR on advanced packaging exposure.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ticker ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@MrThomG_ AMKR is still far from being able to handle CoWoS-level packaging.\nRight now, AMKR is only handling H20 and GB10-class products.\nI have a positive view on ASE.", "reply_text": null, "parent_text": "@Jukanlosreve What abt OSAT demand tailwinds from Stargate or chap fab in the U.S. altogether?", "quoted_text": null, "parent_author": "MrThomG_", "ret_m1d": 0.02670230870680479, "ret_signed_m1d": 0.02670230870680479, "alpha_spy_m1d": 0.028986230171110017, "alpha_c_m1d": 0.004654131611208534, "bench_spy_m1d": -0.0022839214643052275, "bench_c_m1d": 0.022048177095596255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.000667572041884279, "ret_signed_p1d": 0.000667572041884279, "alpha_spy_p1d": 0.00011117343544220226, "alpha_c_p1d": -0.01209411914794889, "bench_spy_p1d": 0.0005563986064420767, "bench_c_p1d": 0.012761691189833169, "ret_p1w": -0.05407206212914606, "ret_signed_p1w": -0.05407206212914606, "alpha_spy_p1w": -0.020514332965222315, "alpha_c_p1w": -0.0034810191106182664, "bench_spy_p1w": -0.033557729163923744, "bench_c_p1w": -0.05059104301852779, "ret_p1m": 0.09679578152530355, "ret_signed_p1m": 0.09679578152530355, "alpha_spy_p1m": 0.09010461870074704, "alpha_c_p1m": 0.031135574491945706, "bench_spy_p1m": 0.006691162824556507, "bench_c_p1m": 0.06566020703335784, "ret_p3m": 0.39452603658289687, "ret_signed_p3m": 0.39452603658289687, "alpha_spy_p3m": 0.3803815050573123, "alpha_c_p3m": 0.2468269343181011, "bench_spy_p3m": 0.014144531525584592, "bench_c_p3m": 0.14769910226479577, "ret_p6m": 1.2803738912739941, "ret_signed_p6m": 1.2803738912739941, "alpha_spy_p6m": 1.199834406610006, "alpha_c_p6m": 0.7094562284899442, "bench_spy_p6m": 0.08053948466398819, "bench_c_p6m": 0.57091766278405, "price_path": [-0.3364, -0.3438, -0.3258, -0.3438, -0.3531, -0.3571, -0.3418, -0.3405, -0.3298, -0.3311, -0.3344, -0.3385, -0.3385, -0.3138, -0.3104, -0.3117, -0.3011, -0.2316, -0.253, -0.2457, -0.2583, -0.2623, -0.2523, -0.2457, -0.2457, -0.2343, -0.235, -0.2183, -0.219, -0.2296, -0.235, -0.2543, -0.257, -0.2597, -0.257, -0.2557, -0.2603, -0.2397, -0.2403, -0.2083, -0.2183, -0.2603, -0.217, -0.2443, -0.2036, -0.1602, -0.1509, -0.1328, -0.1535, -0.1682, -0.1442, -0.1335, -0.1168, -0.1081, -0.0367, 0.004, 0.0688, 0.0808, 0.026, 0.0567, 0.026, 0.016, 0.0267, 0.0, 0.0007, -0.0234, -0.0247, -0.0454, -0.0541, -0.0607, -0.0821, -0.0774, -0.0547, -0.0534, -0.0387, -0.0387, -0.002, -0.01, 0.0194, 0.028, 0.012, 0.0154, 0.0507, 0.0661, 0.0968, 0.0948, 0.0521, 0.0401, 0.0214, -0.0154, -0.0073, 0.012, 0.0247, 0.0367, 0.0381, 0.0381, 0.0494, 0.0694, 0.0701, 0.0748, 0.0748, 0.1255, 0.1322, 0.1682, 0.1762, 0.1569, 0.1776, 0.2196, 0.2483, 0.2603, 0.281, 0.2957, 0.2957, 0.2617, 0.267, 0.2724, 0.2944, 0.3238, 0.3398, 0.3525, 0.3204, 0.267, 0.2957, 0.2944, 0.2677, 0.3525, 0.3945, 0.4786, 0.5013, 0.5848, 0.5661, 0.5621, 0.5621, 0.5668, 0.5581, 0.5474, 0.5995, 0.5674, 0.6569, 0.6762, 0.6175, 0.6215, 0.6155, 0.4893, 0.4753, 0.4726, 0.4099, 0.4459, 0.4493, 0.4619, 0.4012, 0.4352, 0.4366, 0.4573, 0.4459, 0.4619, 0.4226, 0.4473, 0.4372, 0.4953, 0.4306, 0.4352, 0.3945, 0.4473, 0.5007, 0.4873, 0.4873, 0.5053, 0.4813, 0.6115, 0.6362, 0.6582, 0.7724, 0.7937, 0.8178, 0.8425, 0.9085, 0.9546, 0.976, 0.9993, 0.99, 1.1362, 1.0594, 1.008, 1.0421, 1.0968, 1.1088, 1.1549, 1.2323, 1.2804]}
{"unit_id": "thread:2003607816024785388", "ticker_idx": 0, "ts": "2025-12-23T23:25:03+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' projected annual operating profit for next year was revised up from 76.6544 trillion won a month ago to 85.4387 trillion won", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares bullish HBM3E supply/price dynamics driving upward earnings revisions for Samsung and SK Hynix into 2025.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Next year’s HBM3E market will get even bigger”… Samsung Electronics and SK hynix raise prices in succession\n\nAs NVIDIA is reportedly set to ramp up exports to China of its AI chip H200, which is equipped with 5th-generation high-bandwidth memory (HBM) ‘HBM3E,’ memory semiconductor companies such as Samsung Electronics and SK hynix are said to have raised their HBM3E supply prices for next year. This is because order volumes are surging not only from NVIDIA but also from companies such as Google and Amazon that design their own AI accelerators. It is also understood that the price hike was influenced by the fact that memory chipmakers are focusing their efforts on expanding production capacity for 6th-generation HBM (HBM4), for which demand is expected to spike next year, leaving them with insufficient room to respond to HBM3E.\n\nAccording to the industry on the 24th, memory semiconductor companies such as Samsung Electronics and SK hynix are reportedly raising HBM3E supply prices by close to 20%. Analysts say this is unusual given that prices typically fall ahead of the launch of next-generation HBM. The prevailing view is that the impact stems from upward adjustments in next year’s HBM3E order volumes by NVIDIA—formerly the largest customer for HBM3E—as well as by Google and Amazon, which design their own AI accelerators.\n\nAn industry source said, “This year, HBM3E established itself as the core product in the HBM market, but it had been expected that prices would ease somewhat next year as the HBM4 market takes off,” adding, “However, global big-tech companies other than NVIDIA are set to launch AI accelerators equipped with HBM3E next year, so demand for HBM3E is continuing to grow. From the perspective of memory chipmakers, they cannot forgo expanding HBM4 production capacity, so supply has failed to keep up with demand, resulting in a price premium of around 20%.”\n\nHBM3E demand is believed to have expanded more than expected as exports to China of NVIDIA’s AI chip H200 were permitted. Each H200 is equipped with six HBM3E units. Reuters reported on the 22nd (local time) that “NVIDIA plans to fulfill initial orders with existing inventory, and shipments are expected to total 5,000 to 10,000 chip modules (about 40,000 to 80,000 H200 units),” adding that “NVIDIA informed Chinese customers of plans to expand new production capacity for the chip and will begin taking related new orders from the second quarter of next year.”\n\nGlobal big-tech companies such as Google and Amazon are also in need of HBM3E. This is because Google’s TPU (Tensor Processing Unit) and Amazon’s Trainium, both equipped with HBM3E, are set to begin shipping next year. For both products, the amount of HBM installed increases by nearly 20% to 30% versus the previous generation. Google’s 7th-generation TPU is said to include eight HBM3E units per device, while Amazon’s Trainium3 is said to include four HBM3E units.\n\nAs Samsung Electronics and SK hynix are expected to focus on producing HBM4, the next-generation product, demand for HBM3E is exceeding supply. KB Securities analyst Kim Dong-won said, “Next year, the revenue mix in the HBM market is projected at 55% for HBM4 and 45% for HBM3E, and from the third quarter of next year HBM4 is expected to rapidly absorb HBM3E demand.”\n\nAmid the surge in HBM demand and rising prices for key products such as DRAM, expectations for next year’s earnings for Samsung Electronics and SK hynix are also rising. According to the financial information provider FnGuide, Samsung Electronics’ projected annual operating profit for next year was revised up from 76.6544 trillion won a month ago to 85.4387 trillion won, while SK hynix’s operating profit over the same period was revised up from 71.4037 trillion won to 76.1434 trillion won.\n\n---\n\nhttps://t.co/ZL63h7zT5Z", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008968562724272267, "ret_signed_m1d": -0.008968562724272267, "alpha_spy_m1d": -0.004418901924961749, "alpha_c_m1d": -0.003555765350937712, "bench_spy_m1d": -0.004549660799310518, "bench_c_m1d": -0.005412797373334555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0035874392004301425, "ret_signed_p1d": -0.0035874392004301425, "alpha_spy_p1d": -0.00710505950596263, "alpha_c_p1d": -0.005985613363981845, "bench_spy_p1d": 0.0035176203055324873, "bench_c_p1d": 0.002398174163551703, "ret_p1w": 0.08056375153277706, "ret_signed_p1w": 0.08056375153277706, "alpha_spy_p1w": 0.08194460943678739, "alpha_c_p1w": 0.08426359726873234, "bench_spy_p1w": -0.0013808579040103242, "bench_c_p1w": -0.003699845735955276, "ret_p1m": 0.34732510113859405, "ret_signed_p1m": 0.34732510113859405, "alpha_spy_p1m": 0.35104626498776614, "alpha_c_p1m": 0.3619876698688943, "bench_spy_p1m": -0.0037211638491720933, "bench_c_p1m": -0.014662568730300252, "ret_p3m": 0.7970342086213684, "ret_signed_p3m": 0.7970342086213684, "alpha_spy_p3m": 0.8517158079825549, "alpha_c_p3m": 0.8700729004555697, "bench_spy_p3m": -0.05468159936118644, "bench_c_p3m": -0.07303869183420131, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2312, -0.2562, -0.2448, -0.2475, -0.2287, -0.1951, -0.1951, -0.1951, -0.1951, -0.1951, -0.1951, -0.1534, -0.1632, -0.1785, -0.148, -0.1238, -0.122, -0.1202, -0.1256, -0.1157, -0.1345, -0.1139, -0.0852, -0.1076, -0.0987, -0.0664, -0.0359, -0.0036, -0.0592, -0.0978, -0.1103, -0.122, -0.0978, -0.0717, -0.0753, -0.078, -0.1283, -0.0978, -0.1229, -0.1345, -0.0978, -0.1498, -0.1327, -0.1094, -0.078, -0.0717, -0.0987, -0.096, -0.0726, -0.0628, -0.0574, -0.0278, -0.0179, -0.0278, -0.0314, -0.0377, -0.0233, -0.0601, -0.078, -0.0323, -0.035, -0.0466, -0.009, 0.0, -0.0036, -0.0036, 0.0493, 0.077, 0.0806, 0.0806, 0.0806, 0.1581, 0.2446, 0.2518, 0.2707, 0.2509, 0.2527, 0.2509, 0.2401, 0.2644, 0.2969, 0.3419, 0.3455, 0.3086, 0.3473, 0.3726, 0.3708, 0.3708, 0.4374, 0.4636, 0.4483, 0.4465, 0.3554, 0.5095, 0.524, 0.4356, 0.4293, 0.4996, 0.4942, 0.5122, 0.6096, 0.633, 0.633, 0.633, 0.633, 0.7123, 0.7132, 0.7394, 0.8024, 0.834, 0.9647, 0.9511, 0.9511, 0.7583, 0.5519, 0.7267, 0.6961, 0.5636, 0.6934, 0.7123, 0.6934, 0.6537, 0.7006, 0.7475, 0.879, 0.8069, 0.797, 0.679, 0.7096, 0.7033, 0.6231, 0.6231, 0.5921, 0.51, 0.7127, 0.6111, 0.6815, 0.7439, 0.7746, 0.901, 0.8423, 0.8604, 0.8152, 0.8649, 0.9055, 0.9642, 0.9507, 0.9371, 0.9778, 0.9642, 1.0274, 0.9823, 1.0274, 1.0049, 1.041, 0.9913, 0.9913, 1.0997, 1.0997, 1.4022, 1.4519, 1.4248, 1.5783, 1.5196, 1.5648, 1.6731, 1.4428, 1.5377, 1.488, 1.4925, 1.7047, 1.6415, 1.6415, 1.7002, 1.7725, 1.7047, 1.8628, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2003607816024785388", "ticker_idx": 1, "ts": "2025-12-23T23:25:03+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix's operating profit over the same period was revised up from 71.4037 trillion won to 76.1434 trillion won", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares bullish HBM3E supply/price dynamics driving upward earnings revisions for Samsung and SK Hynix into 2025.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Next year’s HBM3E market will get even bigger”… Samsung Electronics and SK hynix raise prices in succession\n\nAs NVIDIA is reportedly set to ramp up exports to China of its AI chip H200, which is equipped with 5th-generation high-bandwidth memory (HBM) ‘HBM3E,’ memory semiconductor companies such as Samsung Electronics and SK hynix are said to have raised their HBM3E supply prices for next year. This is because order volumes are surging not only from NVIDIA but also from companies such as Google and Amazon that design their own AI accelerators. It is also understood that the price hike was influenced by the fact that memory chipmakers are focusing their efforts on expanding production capacity for 6th-generation HBM (HBM4), for which demand is expected to spike next year, leaving them with insufficient room to respond to HBM3E.\n\nAccording to the industry on the 24th, memory semiconductor companies such as Samsung Electronics and SK hynix are reportedly raising HBM3E supply prices by close to 20%. Analysts say this is unusual given that prices typically fall ahead of the launch of next-generation HBM. The prevailing view is that the impact stems from upward adjustments in next year’s HBM3E order volumes by NVIDIA—formerly the largest customer for HBM3E—as well as by Google and Amazon, which design their own AI accelerators.\n\nAn industry source said, “This year, HBM3E established itself as the core product in the HBM market, but it had been expected that prices would ease somewhat next year as the HBM4 market takes off,” adding, “However, global big-tech companies other than NVIDIA are set to launch AI accelerators equipped with HBM3E next year, so demand for HBM3E is continuing to grow. From the perspective of memory chipmakers, they cannot forgo expanding HBM4 production capacity, so supply has failed to keep up with demand, resulting in a price premium of around 20%.”\n\nHBM3E demand is believed to have expanded more than expected as exports to China of NVIDIA’s AI chip H200 were permitted. Each H200 is equipped with six HBM3E units. Reuters reported on the 22nd (local time) that “NVIDIA plans to fulfill initial orders with existing inventory, and shipments are expected to total 5,000 to 10,000 chip modules (about 40,000 to 80,000 H200 units),” adding that “NVIDIA informed Chinese customers of plans to expand new production capacity for the chip and will begin taking related new orders from the second quarter of next year.”\n\nGlobal big-tech companies such as Google and Amazon are also in need of HBM3E. This is because Google’s TPU (Tensor Processing Unit) and Amazon’s Trainium, both equipped with HBM3E, are set to begin shipping next year. For both products, the amount of HBM installed increases by nearly 20% to 30% versus the previous generation. Google’s 7th-generation TPU is said to include eight HBM3E units per device, while Amazon’s Trainium3 is said to include four HBM3E units.\n\nAs Samsung Electronics and SK hynix are expected to focus on producing HBM4, the next-generation product, demand for HBM3E is exceeding supply. KB Securities analyst Kim Dong-won said, “Next year, the revenue mix in the HBM market is projected at 55% for HBM4 and 45% for HBM3E, and from the third quarter of next year HBM4 is expected to rapidly absorb HBM3E demand.”\n\nAmid the surge in HBM demand and rising prices for key products such as DRAM, expectations for next year’s earnings for Samsung Electronics and SK hynix are also rising. According to the financial information provider FnGuide, Samsung Electronics’ projected annual operating profit for next year was revised up from 76.6544 trillion won a month ago to 85.4387 trillion won, while SK hynix’s operating profit over the same period was revised up from 71.4037 trillion won to 76.1434 trillion won.\n\n---\n\nhttps://t.co/ZL63h7zT5Z", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006849374562410304, "ret_signed_m1d": -0.006849374562410304, "alpha_spy_m1d": -0.0022997137630997866, "alpha_c_m1d": 0.002678082812266225, "bench_spy_m1d": -0.004549660799310518, "bench_c_m1d": -0.009527457374676529, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006849267326460851, "ret_signed_p1d": 0.006849267326460851, "alpha_spy_p1d": 0.0033316470209283633, "alpha_c_p1d": 0.0040680971660906895, "bench_spy_p1d": 0.0035176203055324873, "bench_c_p1d": 0.002781170160370161, "ret_p1w": 0.11472595156087828, "ret_signed_p1w": 0.11472595156087828, "alpha_spy_p1w": 0.1161068094648886, "alpha_c_p1w": 0.1143129272879082, "bench_spy_p1w": -0.0013808579040103242, "bench_c_p1w": 0.00041302427297007505, "ret_p1m": 0.2671232487431143, "ret_signed_p1m": 0.2671232487431143, "alpha_spy_p1m": 0.2708444125922864, "alpha_c_p1m": 0.16036592794039684, "bench_spy_p1m": -0.0037211638491720933, "bench_c_p1m": 0.10675732080271749, "ret_p3m": 0.7274967236736549, "ret_signed_p3m": 0.7274967236736549, "alpha_spy_p3m": 0.7821783230348414, "alpha_c_p3m": 0.6680739173220391, "bench_spy_p3m": -0.05468159936118644, "bench_c_p3m": 0.05942280635161579, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.39, -0.4242, -0.4028, -0.4054, -0.384, -0.3233, -0.3233, -0.3233, -0.3233, -0.3233, -0.3233, -0.2676, -0.2899, -0.2959, -0.2771, -0.2257, -0.2035, -0.1693, -0.1804, -0.1761, -0.1812, -0.1273, -0.0846, -0.1085, -0.0452, -0.0281, -0.0435, 0.0609, 0.0027, -0.0093, 0.0147, -0.0076, 0.0369, 0.0592, 0.0558, 0.0472, -0.0418, 0.0369, -0.0247, -0.0384, -0.023, -0.1085, -0.1102, -0.1119, -0.1034, -0.0685, -0.0925, -0.0788, -0.0445, -0.0548, -0.0719, -0.0685, -0.012, -0.0308, 0.0051, -0.0325, -0.0223, -0.0514, -0.0925, -0.0565, -0.0548, -0.0634, -0.0068, 0.0, 0.0068, 0.0068, 0.0257, 0.0959, 0.1147, 0.1147, 0.1147, 0.1592, 0.1918, 0.2432, 0.2705, 0.2945, 0.274, 0.2825, 0.2637, 0.2705, 0.2825, 0.2945, 0.3082, 0.2723, 0.2671, 0.2928, 0.3134, 0.2603, 0.3699, 0.4401, 0.4743, 0.5565, 0.4212, 0.5531, 0.5411, 0.4418, 0.4366, 0.5188, 0.5, 0.4726, 0.5205, 0.5068, 0.5068, 0.5068, 0.5068, 0.5308, 0.625, 0.6284, 0.7209, 0.7432, 0.8853, 0.8201, 0.8201, 0.6108, 0.4564, 0.6143, 0.5851, 0.4341, 0.6091, 0.6383, 0.5954, 0.5611, 0.6709, 0.664, 0.8116, 0.7378, 0.7275, 0.6006, 0.6915, 0.7069, 0.6006, 0.6006, 0.4976, 0.3844, 0.5388, 0.4239, 0.5028, 0.5199, 0.5714, 0.7721, 0.7121, 0.7618, 0.7841, 0.8922, 0.9488, 0.9814, 0.9351, 1.0003, 1.0998, 1.098, 1.1015, 1.0963, 1.2164, 1.2301, 1.2181, 1.2061, 1.2061, 1.4823, 1.4823, 1.7465, 1.8374, 1.8923, 2.2251, 2.1479, 2.3898, 2.3795, 2.1205, 2.1565, 1.9935, 1.9935, 2.328, 2.3298, 2.3298, 2.5202, 2.8478, 2.9274, 3.0029, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:1997133046794027367", "ticker_idx": 0, "ts": "2025-12-06T02:36:37+00:00", "symbol": "memory stocks", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I still recommend buying all memory stocks, Micron included", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Reiterates buy recommendation on all memory stocks including Micron.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory stocks basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I was Googling my X header “jukan05,” and I saw someone saying I’m an MU hater.\n\nThat’s a bit disappointing — I still recommend buying all memory stocks, Micron included.\n\nLet me state once again that I am not a Micron hater. https://t.co/24rdsr1CEK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.035507321622138766, "ret_signed_m1d": -0.035507321622138766, "alpha_spy_m1d": -0.03852065201756483, "alpha_c_m1d": -0.02430133051892036, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0022784313638859834, "ret_signed_p1d": -0.0022784313638859834, "alpha_spy_p1d": -0.0014152863969243792, "alpha_c_p1d": -0.003499444248063847, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.04031120133805788, "ret_signed_p1w": -0.04031120133805788, "alpha_spy_p1w": -0.03606910376559358, "alpha_c_p1w": 0.00215247041380906, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.3081029464610456, "ret_signed_p1m": 0.3081029464610456, "alpha_spy_p1m": 0.2931471843241526, "alpha_c_p1m": 0.25308701307352804, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6669322044692989, "ret_signed_p3m": 0.6669322044692989, "alpha_spy_p3m": 0.6673809716122172, "alpha_c_p3m": 0.5908768900937826, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4156, -0.3972, -0.3697, -0.3642, -0.3442, -0.3546, -0.3236, -0.3319, -0.3222, -0.3104, -0.3166, -0.3217, -0.3411, -0.3211, -0.3183, -0.2846, -0.2506, -0.2449, -0.2407, -0.2478, -0.2331, -0.2388, -0.2204, -0.2162, -0.2311, -0.2078, -0.168, -0.16, -0.142, -0.1536, -0.154, -0.151, -0.1092, -0.0835, -0.0968, -0.066, -0.0528, -0.048, 0.013, -0.048, -0.0389, -0.034, -0.046, -0.002, -0.0021, 0.0006, -0.0139, -0.0476, -0.0173, -0.0648, -0.0768, -0.0923, -0.1307, -0.1031, -0.095, -0.0737, -0.0598, -0.0686, -0.0577, -0.0396, -0.0469, -0.061, -0.0355, 0.0, -0.0023, 0.0239, 0.0019, -0.0131, -0.0403, -0.067, -0.0488, -0.018, -0.0014, 0.0448, 0.0498, 0.0649, 0.0649, 0.0867, 0.1328, 0.138, 0.1283, 0.1283, 0.2102, 0.2461, 0.3081, 0.3185, 0.303, 0.321, 0.3244, 0.3039, 0.308, 0.3275, 0.3821, 0.388, 0.3663, 0.4103, 0.439, 0.4481, 0.4159, 0.5041, 0.5705, 0.5775, 0.5764, 0.5308, 0.6028, 0.5496, 0.4908, 0.5029, 0.5393, 0.5173, 0.5643, 0.6184, 0.6186, 0.6186, 0.6026, 0.6312, 0.6613, 0.708, 0.7083, 0.7569, 0.79, 0.8641, 0.8332, 0.8336, 0.653, 0.5594, 0.6669, 0.6106, 0.5404, 0.6621, 0.6994, 0.6605, 0.6635, 0.7376, 0.778, 0.8725, 0.7996, 0.7639, 0.6559, 0.6851, 0.67, 0.571, 0.5733, 0.4804, 0.436, 0.5975, 0.522, 0.5726, 0.6151, 0.6426, 0.7926, 0.7724, 0.7941, 0.7944, 0.9005, 0.9207, 0.9529, 0.9298, 0.9382, 0.9869, 1.0332, 1.048, 1.0512, 1.1446, 1.1142, 1.1416, 1.1189, 1.1528, 1.329, 1.4151, 1.6426, 1.6631, 1.8078, 2.0378, 1.953, 2.0999, 2.0959, 1.861, 1.8471, 1.7985, 1.8449, 2.0705, 2.0346, 2.0346, 2.3145, 2.4936, 2.4908, 2.6341, null, null]}
{"unit_id": "orig:1997133046794027367", "ticker_idx": 1, "ts": "2025-12-06T02:36:37+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still recommend buying all memory stocks, Micron included", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Reiterates buy recommendation on all memory stocks including Micron.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I was Googling my X header “jukan05,” and I saw someone saying I’m an MU hater.\n\nThat’s a bit disappointing — I still recommend buying all memory stocks, Micron included.\n\nLet me state once again that I am not a Micron hater. https://t.co/24rdsr1CEK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03928399978675012, "ret_signed_m1d": -0.03928399978675012, "alpha_spy_m1d": -0.042297330182176185, "alpha_c_m1d": -0.02807800868353172, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.011205991103218405, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022274411827899465, "ret_signed_p1d": 0.022274411827899465, "alpha_spy_p1d": 0.02313755679486107, "alpha_c_p1d": 0.021053398943721602, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.0012210128841778634, "ret_p1w": -0.03814996540873539, "ret_signed_p1w": -0.03814996540873539, "alpha_spy_p1w": -0.03390786783627109, "alpha_c_p1w": 0.0043137063431315514, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.04246367175186694, "ret_p1m": 0.3914171360164642, "ret_signed_p1m": 0.3914171360164642, "alpha_spy_p1m": 0.37646137387957124, "alpha_c_p1m": 0.3364012026289467, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.055015933387517535, "ret_p3m": 0.6086602969591723, "ret_signed_p3m": 0.6086602969591723, "alpha_spy_p3m": 0.6091090641020906, "alpha_c_p3m": 0.5326049825836561, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": 0.07605531437551627, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4334, -0.3906, -0.3636, -0.3614, -0.3572, -0.3525, -0.3164, -0.3414, -0.3337, -0.3265, -0.3455, -0.3653, -0.3635, -0.3366, -0.3228, -0.2628, -0.2563, -0.2393, -0.2266, -0.248, -0.204, -0.2211, -0.2645, -0.2193, -0.2424, -0.2227, -0.1798, -0.1804, -0.1626, -0.1807, -0.1962, -0.1628, -0.113, -0.1086, -0.1013, -0.0822, -0.0928, -0.0938, -0.0495, -0.117, -0.0381, -0.0348, -0.0364, 0.0258, -0.0235, -0.0082, -0.0404, -0.0004, -0.0201, -0.0746, -0.085, -0.1845, -0.1602, -0.0931, -0.0907, -0.0675, -0.0675, -0.0423, -0.0262, -0.0301, -0.0517, -0.0821, -0.0393, 0.0, 0.0223, 0.068, 0.0467, -0.0234, -0.0381, -0.0584, -0.0867, 0.0066, 0.0769, 0.1202, 0.1189, 0.161, 0.161, 0.1534, 0.1926, 0.1856, 0.1563, 0.1563, 0.2779, 0.2647, 0.3914, 0.3757, 0.3249, 0.3981, 0.4013, 0.3699, 0.3506, 0.3639, 0.4697, 0.4697, 0.4788, 0.5765, 0.6108, 0.6192, 0.5764, 0.6621, 0.7636, 0.7656, 0.6809, 0.7738, 0.6994, 0.5372, 0.5513, 0.5991, 0.5538, 0.5122, 0.6625, 0.6772, 0.6679, 0.6679, 0.6197, 0.7055, 0.6909, 0.7347, 0.7056, 0.6936, 0.7381, 0.6837, 0.6707, 0.6719, 0.5383, 0.6237, 0.6087, 0.5003, 0.5773, 0.6332, 0.6963, 0.6423, 0.7265, 0.79, 0.8706, 0.8707, 0.8, 0.7134, 0.6382, 0.6025, 0.548, 0.4402, 0.4473, 0.3043, 0.3693, 0.491, 0.4845, 0.4845, 0.5311, 0.5304, 0.6486, 0.7085, 0.7047, 0.7289, 0.8874, 0.8492, 0.8533, 0.8445, 0.8176, 0.8214, 0.9759, 0.9525, 1.0133, 1.1262, 1.044, 1.1014, 1.0962, 1.1977, 1.3365, 1.5949, 1.7018, 1.6209, 2.027, 2.2237, 2.1071, 2.2573, 2.1453, 1.9372, 1.7624, 1.8322, 1.9669, 2.089, 2.044, 2.044, 2.6312, 2.7631, 2.7432, 2.9357, null, null]}
{"unit_id": "orig:1985213606779392180", "ticker_idx": 0, "ts": "2025-11-03T05:13:01+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bernstein expects NVIDIA's AI server power-component TAM to exceed USD 8 billion in 2027, doubling from USD 4 billion in 2025", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Bernstein analysis projecting NVDA AI server power-component TAM doubling to $8B+ by 2027, implying sustained growth.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: NVIDIA power component value\n\nAccording to Bernstein, for Rubin AI cabinets, the power-component TAM needs to be an integer multiple of that of the GB200/GB300 series; from the GB200 series to the Vera Rubin series (including GB300, VR200, and VR Ultra), the power components will undergo multiple rounds of upgrades as products iterate, and the upgrade paths and key drivers differ across models.\n\nSpecifically, within the GB300 series, represented by the GB300 NV72, the share of power components increases by about 15% versus GB200, driven by higher BBU adoption; at the same time, because the GB300 has higher load dynamics, adoption of components such as capacitors will increase to mitigate AC grid interference, while the PSU share remains unchanged, as the GB200 cabinet itself has ample power redundancy (single-cabinet design supply capacity of 198 kW, total rated power of 130 kW).\n\nThe power-component upgrade is more significant for the VR200 series, where the share of power components in the power rack/power cabinet is expected to reach 2–3× that of the GB200 NV72 series. The core reason is that GPU TDP (thermal design power) rises from 1.2 kW for B200 to 1.8–2.3 kW for R200, driving the single-cabinet design supply capacity from 198 kW to 288 kW/330 kW; the PSU and BBU shares increase accordingly, and this series of power-cabinet form-factor products will adopt BBUs across the board.\n\nThe VR Ultra series (i.e., the Kyber cabinet equipped with the NV576 platform) relies on 800 V HVDC technology, with the power-component share expected to reach 7–8× that of GB200. On one hand, its GPU-cabinet density increases from 72 to 144 (the number of cores rises from 144 to 576), and GPU TDP is expected to reach 3.6 kW; traditional architectures cannot meet such high-power demand, hence the adoption of 800 V HVDC power cabinets. On the other hand, in addition to PSU and BBU increasing with the rise in single-cabinet design supply capacity, the shares of other power/mechanical components—such as power capacitors, e-Fuse modules, HVDC cooling systems, and enclosures—will also increase substantially.\n\nBernstein expects NVIDIA’s AI server power-component TAM to exceed USD 8 billion in 2027, doubling from USD 4 billion in 2025.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02121999772951899, "ret_signed_m1d": -0.02121999772951899, "alpha_spy_m1d": -0.019346825998755257, "alpha_c_m1d": -0.012806943728119413, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.039588173995557474, "ret_signed_p1d": -0.039588173995557474, "alpha_spy_p1d": -0.02773463405310994, "alpha_c_p1d": -0.003150058353086793, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": -0.037847988274935296, "ret_signed_p1w": -0.037847988274935296, "alpha_spy_p1w": -0.035067559339898025, "alpha_c_p1w": -0.017880771533080764, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": -0.12287314834642526, "ret_signed_p1m": -0.12287314834642526, "alpha_spy_p1m": -0.12022440644717658, "alpha_c_p1m": -0.10465404871620898, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": -0.06940878018151897, "ret_signed_p3m": -0.06940878018151897, "alpha_spy_p3m": -0.08806827622259894, "alpha_c_p3m": -0.21341078433682992, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 0.030517241086794655, "ret_signed_p6m": 0.030517241086794655, "alpha_spy_p6m": -0.01690033212594244, "alpha_c_p6m": -0.3153948816195542, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.1328, -0.1263, -0.1169, -0.12, -0.1147, -0.1223, -0.1202, -0.1278, -0.1203, -0.1511, -0.1522, -0.1542, -0.1397, -0.1309, -0.1214, -0.1222, -0.1292, -0.1581, -0.1581, -0.1745, -0.1753, -0.1703, -0.1927, -0.1865, -0.1746, -0.1429, -0.1436, -0.1405, -0.1408, -0.1547, -0.1769, -0.1481, -0.146, -0.1125, -0.1375, -0.1446, -0.1411, -0.1387, -0.121, -0.0981, -0.0949, -0.087, -0.0931, -0.1032, -0.1056, -0.0859, -0.0692, -0.1147, -0.0897, -0.1298, -0.1308, -0.1212, -0.1144, -0.1172, -0.1243, -0.1286, -0.1195, -0.0997, -0.0744, -0.0283, 0.0008, -0.0193, -0.0212, 0.0, -0.0396, -0.0564, -0.0909, -0.0905, -0.0378, -0.0663, -0.0632, -0.0968, -0.0808, -0.098, -0.1234, -0.0984, -0.1268, -0.1353, -0.1176, -0.1405, -0.1287, -0.1287, -0.1444, -0.1303, -0.1229, -0.1319, -0.1135, -0.1182, -0.1031, -0.1059, -0.1116, -0.1254, -0.154, -0.1478, -0.1409, -0.1737, -0.1582, -0.1251, -0.112, -0.0854, -0.0883, -0.0883, -0.079, -0.0901, -0.0934, -0.0985, -0.0985, -0.0871, -0.0906, -0.0949, -0.0858, -0.1055, -0.1064, -0.106, -0.1018, -0.1147, -0.0958, -0.0998, -0.0998, -0.1392, -0.1138, -0.1065, -0.0928, -0.0986, -0.0887, -0.0742, -0.0694, -0.0761, -0.1028, -0.1282, -0.158, -0.1691, -0.1037, -0.0813, -0.0886, -0.0813, -0.0963, -0.1163, -0.1163, -0.1059, -0.0913, -0.0917, -0.0824, -0.074, -0.0678, -0.0547, -0.1062, -0.1435, -0.1179, -0.1296, -0.1152, -0.1137, -0.1404, -0.1171, -0.1068, -0.1007, -0.1147, -0.1286, -0.1143, -0.1205, -0.1279, -0.1368, -0.1651, -0.1509, -0.153, -0.1362, -0.1722, -0.1902, -0.2015, -0.1569, -0.1504, -0.1425, -0.1425, -0.1412, -0.139, -0.1198, -0.1109, -0.0881, -0.0848, -0.05, -0.0386, -0.0411, -0.025, -0.0232, -0.0337, -0.0211, -0.0349, 0.0068, 0.0471, 0.0305]}
{"unit_id": "orig:1991100251332366618", "ticker_idx": 0, "ts": "2025-11-19T11:04:27+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I have no doubt that NVIDIA will deliver an astonishing earnings report", "tweet_type": "prediction", "timeline": "weeks", "summary": "Confident NVDA will beat earnings, but doubts it can reverse the broader market's bearish trend.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Jensen Huang, please save this bear market.\nYou’re our only hope.\n\n(I have no doubt that NVIDIA will deliver an astonishing earnings report, but I’m not sure if that alone can overturn the market’s bearish outlook.) https://t.co/XaJD3CkTp8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02766454335765578, "ret_signed_m1d": -0.02766454335765578, "alpha_spy_m1d": -0.023816210452913378, "alpha_c_m1d": -0.009546474782834213, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.018118068574821566, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03152474073899536, "ret_signed_p1d": -0.03152474073899536, "alpha_spy_p1d": -0.016282404056229938, "alpha_c_p1d": 0.010721352158470698, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.04224609289746606, "ret_p1w": -0.03356206031663744, "ret_signed_p1w": -0.03356206031663744, "alpha_spy_p1w": -0.05929280169992157, "alpha_c_p1w": -0.05792564893119845, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.024363588614561005, "ret_p1m": -0.06632157986233833, "ret_signed_p1m": -0.06632157986233833, "alpha_spy_p1m": -0.08720788493658327, "alpha_c_p1m": -0.08933004250050691, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.02300846263816858, "ret_p3m": -0.01983598681506271, "ret_signed_p3m": -0.01983598681506271, "alpha_spy_p3m": -0.0517307490878941, "alpha_c_p3m": -0.22472831084619183, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": 0.2048923240311291, "ret_p6m": 0.2640247865563907, "ret_signed_p6m": 0.2640247865563907, "alpha_spy_p6m": 0.12850423753509732, "alpha_c_p6m": -0.4450830741122007, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.7091078606685914, "price_path": [-0.0458, -0.036, -0.0255, -0.0264, -0.0341, -0.0662, -0.0662, -0.0844, -0.0853, -0.0797, -0.1046, -0.0977, -0.0845, -0.0493, -0.0501, -0.0466, -0.047, -0.0624, -0.087, -0.0551, -0.0528, -0.0156, -0.0434, -0.0512, -0.0473, -0.0447, -0.025, 0.0003, 0.0039, 0.0127, 0.0059, -0.0053, -0.0079, 0.0139, 0.0324, -0.018, 0.0097, -0.0348, -0.0359, -0.0253, -0.0177, -0.0208, -0.0287, -0.0335, -0.0234, -0.0014, 0.0266, 0.0778, 0.11, 0.0878, 0.0856, 0.1092, 0.0652, 0.0466, 0.0084, 0.0087, 0.0672, 0.0356, 0.039, 0.0018, 0.0196, 0.0004, -0.0277, 0.0, -0.0315, -0.041, -0.0213, -0.0466, -0.0336, -0.0336, -0.051, -0.0354, -0.0271, -0.0372, -0.0168, -0.022, -0.0051, -0.0083, -0.0146, -0.0299, -0.0616, -0.0548, -0.0471, -0.0835, -0.0663, -0.0296, -0.0151, 0.0145, 0.0113, 0.0113, 0.0216, 0.0092, 0.0055, -0.0001, -0.0001, 0.0125, 0.0086, 0.0039, 0.0139, -0.0079, -0.0088, -0.0084, -0.0038, -0.0181, 0.0029, -0.0015, -0.0015, -0.0453, -0.0171, -0.009, 0.0062, -0.0002, 0.0108, 0.0269, 0.0322, 0.0248, -0.0048, -0.0331, -0.0661, -0.0784, -0.0059, 0.0189, 0.0109, 0.019, 0.0023, -0.0198, -0.0198, -0.0083, 0.0079, 0.0075, 0.0177, 0.027, 0.034, 0.0485, -0.0087, -0.05, -0.0216, -0.0346, -0.0186, -0.017, -0.0466, -0.0207, -0.0093, -0.0025, -0.018, -0.0335, -0.0176, -0.0245, -0.0327, -0.0426, -0.074, -0.0582, -0.0606, -0.0419, -0.0818, -0.1018, -0.1144, -0.0649, -0.0576, -0.0488, -0.0488, -0.0475, -0.045, -0.0237, -0.0139, 0.0114, 0.0151, 0.0537, 0.0663, 0.0635, 0.0814, 0.0834, 0.0717, 0.0858, 0.0705, 0.1167, 0.1615, 0.143, 0.122, 0.0701, 0.0641, 0.0642, 0.0536, 0.1144, 0.1341, 0.1539, 0.1766, 0.1838, 0.2109, 0.264]}
{"unit_id": "orig:1991100251332366618", "ticker_idx": 1, "ts": "2025-11-19T11:04:27+00:00", "symbol": "S&P500", "kind": "index", "geography": "US", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "not sure if that alone can overturn the market's bearish outlook", "tweet_type": "prediction", "timeline": "weeks", "summary": "Confident NVDA will beat earnings, but doubts it can reverse the broader market's bearish trend.", "resolved_tickers": ["SPY"], "resolution_method": "etf", "resolution_reason": "S&P 500 best proxied by SPY ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Jensen Huang, please save this bear market.\nYou’re our only hope.\n\n(I have no doubt that NVIDIA will deliver an astonishing earnings report, but I’m not sure if that alone can overturn the market’s bearish outlook.) https://t.co/XaJD3CkTp8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003848332904742402, "ret_signed_m1d": 0.003848332904742402, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": -0.003848332904742402, "bench_c_m1d": -0.003848332904742402, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015242336682765423, "ret_signed_p1d": 0.015242336682765423, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": -0.015242336682765423, "bench_c_p1d": -0.015242336682765423, "ret_p1w": 0.025730741383284128, "ret_signed_p1w": -0.025730741383284128, "alpha_spy_p1w": 0.0, "alpha_c_p1w": 0.0, "bench_spy_p1w": 0.025730741383284128, "bench_c_p1w": 0.025730741383284128, "ret_p1m": 0.020886305074244937, "ret_signed_p1m": -0.020886305074244937, "alpha_spy_p1m": 0.0, "alpha_c_p1m": 0.0, "bench_spy_p1m": 0.020886305074244937, "bench_c_p1m": 0.020886305074244937, "ret_p3m": 0.03189476227283139, "ret_signed_p3m": -0.03189476227283139, "alpha_spy_p3m": 0.0, "alpha_c_p3m": 0.0, "bench_spy_p3m": 0.03189476227283139, "bench_c_p3m": 0.03189476227283139, "ret_p6m": 0.1355205490212934, "ret_signed_p6m": -0.1355205490212934, "alpha_spy_p6m": 0.0, "alpha_c_p6m": 0.0, "bench_spy_p6m": 0.1355205490212934, "bench_c_p6m": 0.1355205490212934, "price_path": [-0.0288, -0.0331, -0.0291, -0.0268, -0.0234, -0.0292, -0.0292, -0.0364, -0.0312, -0.0231, -0.0259, -0.0235, -0.0213, -0.0184, -0.0103, -0.0106, -0.0054, -0.0067, -0.008, -0.0033, 0.0016, 0.0064, 0.0009, -0.0023, -0.0069, -0.0012, 0.0016, 0.0054, 0.0088, 0.0099, 0.0099, 0.0136, 0.0098, 0.0158, 0.0129, -0.0145, 0.0006, -0.0006, 0.0038, -0.003, 0.0027, 0.0131, 0.0131, 0.0078, 0.0138, 0.0221, 0.0341, 0.0369, 0.0374, 0.026, 0.0293, 0.0313, 0.019, 0.0226, 0.0116, 0.0126, 0.0284, 0.0307, 0.0313, 0.0142, 0.014, 0.0046, -0.0038, 0.0, -0.0152, -0.0054, 0.0092, 0.0187, 0.0257, 0.0257, 0.0313, 0.0266, 0.0285, 0.0321, 0.0328, 0.0348, 0.0317, 0.0308, 0.0376, 0.0401, 0.0289, 0.0273, 0.0245, 0.0132, 0.0209, 0.0301, 0.0366, 0.0413, 0.045, 0.045, 0.0449, 0.0411, 0.0399, 0.0322, 0.0322, 0.034, 0.0409, 0.0471, 0.0437, 0.0436, 0.0505, 0.0522, 0.0501, 0.0449, 0.0478, 0.0469, 0.0469, 0.0256, 0.0374, 0.0428, 0.0432, 0.0485, 0.0527, 0.0526, 0.0505, 0.0474, 0.0526, 0.0437, 0.0386, 0.0256, 0.0453, 0.0504, 0.0476, 0.0473, 0.0312, 0.0319, 0.0319, 0.0336, 0.0388, 0.036, 0.0435, 0.0329, 0.0404, 0.0491, 0.0433, 0.0383, 0.0389, 0.0297, 0.037, 0.0312, 0.0177, 0.0266, 0.025, 0.0237, 0.0081, 0.0024, 0.0126, 0.0153, 0.0011, -0.0013, -0.0156, -0.0053, -0.0086, -0.0031, -0.0209, -0.0376, -0.0408, -0.013, -0.0055, -0.0046, -0.0046, 0.0001, 0.0005, 0.026, 0.0319, 0.0312, 0.0413, 0.054, 0.0623, 0.0649, 0.0778, 0.0756, 0.0686, 0.0794, 0.0752, 0.0836, 0.0854, 0.0802, 0.08, 0.0907, 0.0938, 0.0897, 0.0985, 0.1138, 0.1103, 0.1195, 0.1221, 0.1204, 0.1266, 0.1355]}
{"unit_id": "quote:1984785416307032492", "ticker_idx": 0, "ts": "2025-11-02T00:51:33+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "semiconductor stocks are trading 35% above their 200-day moving average already reflects an extremely overheated market condition", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Rebuttal to UBS: AI is a bubble; semis 35% above 200-DMA is overheated; hyperscaler capex concentration in NVIDIA mirrors bubble-era mega-deals.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "US semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "My rebuttal to UBS’s claim that “this is not yet a bubble”:\n\nThe fallacy of the “reasonable” TAM assumption:\n\nUBS claims that it is “not unreasonable” to assume semiconductor spending will reach 1.3% of global GDP, but that figure is nearly double the current level (0.7%).\nThis assumption presupposes that AI infrastructure investment represents a structural, permanent upshift, not a one-off boom.\n\nIf AI-related capex eventually peaks and slows, doesn’t that undermine the 1.3% premise itself?\n\nRebuttal to “there is no Mega M&A in this bubble”:\n\nIsn’t the hyperscalers’ collective $100 billion annual capex commitment effectively the “deal of the century” in this cycle?\n\nInstead of acquiring a single company, they are, in effect, buying out an entire supply chain (NVIDIA) — the difference is semantic.\n\nPrice momentum:\n\nThe fact that semiconductor stocks are trading 35% above their 200-day moving average already reflects an extremely overheated market condition.\n\nIt’s only “lower” compared to the 70% seen in 2000, but that doesn’t make it normal by any means.", "reply_text": null, "parent_text": null, "quoted_text": "How big is the current AI bubble? UBS: This bubble is more “reasonable” than the TMT era, and the three peak signals have not yet appeared\n\nUBS believes that although the current U.S. stock market already meets the seven preconditions of a bubble cycle, the “rationality” of the https://t.co/XMG6D49bgJ", "parent_author": null, "ret_m1d": -0.008413054001399578, "ret_signed_m1d": 0.008413054001399578, "alpha_spy_m1d": 0.006539882270635844, "alpha_c_m1d": 0.006539882270635844, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.0018731717307637341, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03643811564247068, "ret_signed_p1d": 0.03643811564247068, "alpha_spy_p1d": 0.024584575700023148, "alpha_c_p1d": 0.024584575700023148, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.011853539942447533, "ret_p1w": -0.019967216741854532, "ret_signed_p1w": 0.019967216741854532, "alpha_spy_p1w": 0.01718678780681726, "alpha_c_p1w": 0.01718678780681726, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.0027804289350372713, "ret_p1m": -0.018219099630216284, "ret_signed_p1m": 0.018219099630216284, "alpha_spy_p1m": 0.015570357730967599, "alpha_c_p1m": 0.015570357730967599, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.0026487418992486855, "ret_p3m": 0.14400200415531095, "ret_signed_p3m": -0.14400200415531095, "alpha_spy_p3m": -0.12534250811423098, "alpha_c_p3m": -0.12534250811423098, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.01865949604107997, "ret_p6m": 0.34591212270634886, "ret_signed_p6m": -0.34591212270634886, "alpha_spy_p6m": -0.29849454949361176, "alpha_c_p6m": -0.29849454949361176, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.047417573212737096, "price_path": [-0.2171, -0.2048, -0.1982, -0.1986, -0.1803, -0.1771, -0.1753, -0.1924, -0.1891, -0.2055, -0.2107, -0.2148, -0.1979, -0.1972, -0.1894, -0.1869, -0.1836, -0.2071, -0.2071, -0.2175, -0.2176, -0.2082, -0.1989, -0.1901, -0.1865, -0.1783, -0.1698, -0.1688, -0.161, -0.1609, -0.1662, -0.1342, -0.1376, -0.12, -0.1213, -0.1229, -0.123, -0.121, -0.1187, -0.1085, -0.0885, -0.0761, -0.0804, -0.0621, -0.0793, -0.0547, -0.0571, -0.1113, -0.072, -0.089, -0.0664, -0.0623, -0.0636, -0.0516, -0.0568, -0.0751, -0.0582, -0.0409, -0.0168, -0.0082, 0.0069, -0.0065, -0.0084, 0.0, -0.0364, -0.018, -0.041, -0.0491, -0.02, -0.0411, -0.0289, -0.0581, -0.0577, -0.0705, -0.0896, -0.0728, -0.112, -0.1092, -0.0737, -0.0713, -0.0502, -0.0502, -0.0377, -0.0359, -0.0182, -0.0047, -0.0123, -0.0046, 0.0067, 0.0079, 0.0219, 0.0131, -0.0327, -0.0361, -0.0387, -0.0734, -0.0515, -0.027, -0.0144, -0.0049, -0.0022, -0.0022, 0.0025, -0.0021, -0.0045, -0.0132, -0.0132, 0.0228, 0.0346, 0.0621, 0.055, 0.0384, 0.0665, 0.0704, 0.0728, 0.0641, 0.0862, 0.0971, 0.0971, 0.0697, 0.1013, 0.1037, 0.0962, 0.0928, 0.1159, 0.1416, 0.144, 0.1055, 0.1178, 0.0896, 0.0467, 0.0441, 0.1005, 0.1142, 0.109, 0.1365, 0.1127, 0.1172, 0.1172, 0.1166, 0.1304, 0.1239, 0.1372, 0.1313, 0.1485, 0.1677, 0.1289, 0.1135, 0.1135, 0.0715, 0.0935, 0.0833, 0.0427, 0.0806, 0.0887, 0.0988, 0.0635, 0.0613, 0.0793, 0.0874, 0.0787, 0.0821, 0.0542, 0.0723, 0.0812, 0.0933, 0.0435, 0.0254, -0.0067, 0.0505, 0.074, 0.075, 0.075, 0.085, 0.0957, 0.1588, 0.179, 0.197, 0.2147, 0.2385, 0.2412, 0.2461, 0.2718, 0.2712, 0.2732, 0.3065, 0.3203, 0.3876, 0.3871, 0.3459]}
{"unit_id": "quote:1984785416307032492", "ticker_idx": 1, "ts": "2025-11-02T00:51:33+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "they are, in effect, buying out an entire supply chain (NVIDIA) — the difference is semantic", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Rebuttal to UBS: AI is a bubble; semis 35% above 200-DMA is overheated; hyperscaler capex concentration in NVIDIA mirrors bubble-era mega-deals.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "My rebuttal to UBS’s claim that “this is not yet a bubble”:\n\nThe fallacy of the “reasonable” TAM assumption:\n\nUBS claims that it is “not unreasonable” to assume semiconductor spending will reach 1.3% of global GDP, but that figure is nearly double the current level (0.7%).\nThis assumption presupposes that AI infrastructure investment represents a structural, permanent upshift, not a one-off boom.\n\nIf AI-related capex eventually peaks and slows, doesn’t that undermine the 1.3% premise itself?\n\nRebuttal to “there is no Mega M&A in this bubble”:\n\nIsn’t the hyperscalers’ collective $100 billion annual capex commitment effectively the “deal of the century” in this cycle?\n\nInstead of acquiring a single company, they are, in effect, buying out an entire supply chain (NVIDIA) — the difference is semantic.\n\nPrice momentum:\n\nThe fact that semiconductor stocks are trading 35% above their 200-day moving average already reflects an extremely overheated market condition.\n\nIt’s only “lower” compared to the 70% seen in 2000, but that doesn’t make it normal by any means.", "reply_text": null, "parent_text": null, "quoted_text": "How big is the current AI bubble? UBS: This bubble is more “reasonable” than the TMT era, and the three peak signals have not yet appeared\n\nUBS believes that although the current U.S. stock market already meets the seven preconditions of a bubble cycle, the “rationality” of the https://t.co/XMG6D49bgJ", "parent_author": null, "ret_m1d": -0.02121999772951899, "ret_signed_m1d": 0.02121999772951899, "alpha_spy_m1d": 0.019346825998755257, "alpha_c_m1d": 0.012806943728119413, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.039588173995557474, "ret_signed_p1d": 0.039588173995557474, "alpha_spy_p1d": 0.02773463405310994, "alpha_c_p1d": 0.003150058353086793, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": -0.037847988274935296, "ret_signed_p1w": 0.037847988274935296, "alpha_spy_p1w": 0.035067559339898025, "alpha_c_p1w": 0.017880771533080764, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": -0.12287314834642526, "ret_signed_p1m": 0.12287314834642526, "alpha_spy_p1m": 0.12022440644717658, "alpha_c_p1m": 0.10465404871620898, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": -0.06940878018151897, "ret_signed_p3m": 0.06940878018151897, "alpha_spy_p3m": 0.08806827622259894, "alpha_c_p3m": 0.21341078433682992, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 0.030517241086794655, "ret_signed_p6m": -0.030517241086794655, "alpha_spy_p6m": 0.01690033212594244, "alpha_c_p6m": 0.3153948816195542, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.1328, -0.1263, -0.1169, -0.12, -0.1147, -0.1223, -0.1202, -0.1278, -0.1203, -0.1511, -0.1522, -0.1542, -0.1397, -0.1309, -0.1214, -0.1222, -0.1292, -0.1581, -0.1581, -0.1745, -0.1753, -0.1703, -0.1927, -0.1865, -0.1746, -0.1429, -0.1436, -0.1405, -0.1408, -0.1547, -0.1769, -0.1481, -0.146, -0.1125, -0.1375, -0.1446, -0.1411, -0.1387, -0.121, -0.0981, -0.0949, -0.087, -0.0931, -0.1032, -0.1056, -0.0859, -0.0692, -0.1147, -0.0897, -0.1298, -0.1308, -0.1212, -0.1144, -0.1172, -0.1243, -0.1286, -0.1195, -0.0997, -0.0744, -0.0283, 0.0008, -0.0193, -0.0212, 0.0, -0.0396, -0.0564, -0.0909, -0.0905, -0.0378, -0.0663, -0.0632, -0.0968, -0.0808, -0.098, -0.1234, -0.0984, -0.1268, -0.1353, -0.1176, -0.1405, -0.1287, -0.1287, -0.1444, -0.1303, -0.1229, -0.1319, -0.1135, -0.1182, -0.1031, -0.1059, -0.1116, -0.1254, -0.154, -0.1478, -0.1409, -0.1737, -0.1582, -0.1251, -0.112, -0.0854, -0.0883, -0.0883, -0.079, -0.0901, -0.0934, -0.0985, -0.0985, -0.0871, -0.0906, -0.0949, -0.0858, -0.1055, -0.1064, -0.106, -0.1018, -0.1147, -0.0958, -0.0998, -0.0998, -0.1392, -0.1138, -0.1065, -0.0928, -0.0986, -0.0887, -0.0742, -0.0694, -0.0761, -0.1028, -0.1282, -0.158, -0.1691, -0.1037, -0.0813, -0.0886, -0.0813, -0.0963, -0.1163, -0.1163, -0.1059, -0.0913, -0.0917, -0.0824, -0.074, -0.0678, -0.0547, -0.1062, -0.1435, -0.1179, -0.1296, -0.1152, -0.1137, -0.1404, -0.1171, -0.1068, -0.1007, -0.1147, -0.1286, -0.1143, -0.1205, -0.1279, -0.1368, -0.1651, -0.1509, -0.153, -0.1362, -0.1722, -0.1902, -0.2015, -0.1569, -0.1504, -0.1425, -0.1425, -0.1412, -0.139, -0.1198, -0.1109, -0.0881, -0.0848, -0.05, -0.0386, -0.0411, -0.025, -0.0232, -0.0337, -0.0211, -0.0349, 0.0068, 0.0471, 0.0305]}
{"unit_id": "thread:1983397632211853568", "ticker_idx": 0, "ts": "2025-10-29T04:56:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "foundry", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's foundry business could return to profitability as early as 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung on foundry turnaround: 3nm yield stabilization, Tesla AI5/AI6 wins, and foundry profitability expected by 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung secures Tesla AI5 chip order as 3nm yield gains boost profit outlook\n\nSamsung Electronics has secured the foundry order for Tesla's AI5 chip, following its earlier win of the next-generation AI6 chip contract. Some industry observers believe this reflects the stability of Samsung's 3nm process yield, and note that the company is significantly ramping up the utilization rate of its 3nm production lines in an effort to attract more major customers.\n\nSamsung and TSMC to jointly produce AI5 chip\nAccording to South Korean media outlet Business Post, Tesla CEO Elon Musk recently confirmed during the company's third-quarter 2025 earnings call that both Samsung and Taiwan Semiconductor Manufacturing Company (TSMC) will jointly produce the AI5 chip. The AI5 was originally expected to be manufactured solely on TSMC's 3nm node.\n\nSamsung currently produces Tesla's AI4 chips using its 4nm to 7nm processes. In July 2025, Samsung announced it had secured a US$16.5 billion wafer foundry order, which Musk later confirmed was for the next-generation AI6 chip built on Samsung's 2nm technology. Industry analysts believe that adding Samsung to the AI5 chip supply chain reflects Tesla's strategy to reduce costs and manage supply chain risks.\n\nSamsung's 3nm process sees significant improvement\nNotably, this collaboration signals significant improvements in Samsung's 3nm process. Between 2024 and early 2025, Samsung struggled to effectively improve the yield of its 3nm process, hindering efforts to secure large customers and putting the company in a challenging position. However, from the second half of 2025 onward, 3nm yields have gradually stabilized. This enabled the successful deployment of its Exynos 2500 mobile application processor (AP) in the Galaxy Z Flip7 foldable phone, marking a key turning point.\n\nWith the new 3nm order from Tesla, the market expects Samsung to secure more major customers in the 3nm segment. Qualcomm, Google, and Baidu, which all currently manufacture products on Samsung's 4nm to 7nm nodes, are considered potential candidates for its 3nm process.\n\nSamsung foundry aims for profitability in 2026\nSamsung's 3nm process is already ready for mass production, giving it the advantage of contributing to revenue starting in 2026. Samsung's foundry business recorded an operating loss of approximately KRW5 trillion (US$3.48 billion) in 2024 and is expected to see losses of around KRW1 trillion in the third quarter of 2025.\n\nReports indicate that Samsung plans to begin full-scale mass production of the Exynos 2600 based on its 2nm process in November 2025. Industry observers believe Samsung's foundry business could return to profitability as early as 2026.\n\nWith utilization rates rising significantly across 2nm and 3nm lines in 2026, the market predicts that Samsung's foundry business may generate operating profits worth several hundred billion Korean won. Jinman Han, the head of the foundry business, is expected to accelerate the previously projected timeline for returning to profitability from 2027 to 2026, based on the continued influx of orders.\n\n---\n\nhttps://t.co/Lp0aQcgGGM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009950196182860638, "ret_signed_m1d": -0.009950196182860638, "alpha_spy_m1d": -0.009470129059189647, "alpha_c_m1d": -0.0028149596435107505, "bench_spy_m1d": -0.0004800671236709908, "bench_c_m1d": -0.007135236539349887, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03582092543078752, "ret_signed_p1d": 0.03582092543078752, "alpha_spy_p1d": 0.0468190703499789, "alpha_c_p1d": 0.04811832126868021, "bench_spy_p1d": -0.010998144919191377, "bench_c_p1d": -0.012297395837892688, "ret_p1w": 0.000995043101052806, "ret_signed_p1w": 0.000995043101052806, "alpha_spy_p1w": 0.015266413544455903, "alpha_c_p1w": 0.030719292249522412, "bench_spy_p1w": -0.014271370443403097, "bench_c_p1w": -0.029724249148469606, "ret_p1m": 0.029850745100359788, "ret_signed_p1m": 0.029850745100359788, "alpha_spy_p1m": 0.041067045522365175, "alpha_c_p1m": 0.09676300751027511, "bench_spy_p1m": -0.011216300422005387, "bench_c_p1m": -0.06691226240991532, "ret_p3m": 0.520789841292504, "ret_signed_p3m": 0.520789841292504, "alpha_spy_p3m": 0.5100435172975597, "alpha_c_p3m": 0.5586909555996736, "bench_spy_p3m": 0.010746323994944351, "bench_c_p3m": -0.03790111430716958, "ret_p6m": 1.249335946495612, "ret_signed_p6m": 1.249335946495612, "alpha_spy_p6m": 1.2128298600195921, "alpha_c_p6m": 1.2216805657824645, "bench_spy_p6m": 0.03650608647601983, "bench_c_p6m": 0.027655380713147437, "price_path": [-0.3175, -0.3095, -0.3076, -0.3185, -0.3016, -0.2887, -0.2967, -0.2957, -0.2878, -0.2907, -0.2907, -0.3066, -0.3066, -0.3016, -0.3006, -0.2927, -0.2917, -0.3036, -0.3006, -0.3105, -0.3095, -0.3304, -0.3155, -0.3086, -0.3056, -0.3115, -0.3056, -0.2917, -0.2808, -0.2729, -0.2531, -0.2422, -0.2135, -0.2253, -0.2045, -0.2045, -0.1728, -0.161, -0.154, -0.1471, -0.1748, -0.1622, -0.1652, -0.1443, -0.107, -0.107, -0.107, -0.107, -0.107, -0.107, -0.0607, -0.0716, -0.0886, -0.0547, -0.0279, -0.0259, -0.0239, -0.0299, -0.0189, -0.0398, -0.0169, 0.0149, -0.01, 0.0, 0.0358, 0.0697, 0.1055, 0.0438, 0.001, -0.0129, -0.0259, 0.001, 0.0299, 0.0259, 0.0229, -0.0328, 0.001, -0.0269, -0.0398, 0.001, -0.0567, -0.0378, -0.0119, 0.0229, 0.0299, 0.0, 0.003, 0.0289, 0.0398, 0.0458, 0.0786, 0.0896, 0.0786, 0.0746, 0.0677, 0.0836, 0.0428, 0.0229, 0.0736, 0.0706, 0.0577, 0.0995, 0.1095, 0.1055, 0.1055, 0.1642, 0.1948, 0.1988, 0.1988, 0.1988, 0.2848, 0.3808, 0.3888, 0.4098, 0.3878, 0.3898, 0.3878, 0.3758, 0.4028, 0.4388, 0.4888, 0.4928, 0.4518, 0.4948, 0.5228, 0.5208, 0.5208, 0.5948, 0.6238, 0.6068, 0.6048, 0.5038, 0.6748, 0.6908, 0.5928, 0.5858, 0.6638, 0.6578, 0.6778, 0.7858, 0.8117, 0.8117, 0.8117, 0.8117, 0.8997, 0.9007, 0.9297, 0.9997, 1.0347, 1.1797, 1.1647, 1.1647, 0.9507, 0.7218, 0.9157, 0.8817, 0.7348, 0.8787, 0.8997, 0.8787, 0.8347, 0.8867, 0.9387, 1.0847, 1.0047, 0.9937, 0.8627, 0.8967, 0.8897, 0.8008, 0.8008, 0.7664, 0.6752, 0.9002, 0.7874, 0.8656, 0.9347, 0.9688, 1.1091, 1.0439, 1.064, 1.0139, 1.069, 1.1141, 1.1792, 1.1642, 1.1491, 1.1942, 1.1792, 1.2493]}
{"unit_id": "orig:1983679575940030550", "ticker_idx": 0, "ts": "2025-10-29T23:37:20+00:00", "symbol": "ASE Technology Holding", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASE: Target price raised to NT$228… Both companies maintain an 'Overweight' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises TPs and reiterates Overweight on ASE and KYEC on OSAT price hike cycle driven by AI demand and capacity constraints.", "resolved_tickers": ["ASX"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding US ADR ticker ASX", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Chip Inflation Wave Expands — Morgan Stanley Expects OSAT Companies to Raise Prices in 2026, the First Upcycle Since the Pandemic\n\nMorgan Stanley stated that the semiconductor inflation wave, which began with wafer foundries and memory, is now spreading downstream to back-end processes. Chip back-end (OSAT) companies are preparing for the first price hike cycle since the COVID-19 pandemic.\n\nMorgan Stanley warned that the strong demand for AI semiconductors is severely straining packaging and testing capacity, giving back-end companies stronger bargaining power. Taiwan’s leading firms, ASE and KYEC, are expected to lead this price increase.\n\nAnalysts predict that due to three key factors, the price of advanced packaging will rise by 5–10% in 2026, marking the first price upcycle since the semiconductor shortage during the pandemic:\n    \n1. Surging AI demand — TSMC’s CoWoS capacity shortage is quickly spilling over to ASE’s CoWoS capacity and KYEC’s testing lines.\n    \n2. Capacity constraints — ASE and KYEC are rejecting lower-margin products or reallocating capacity from non-AI wire-bonding to AI flip-chip packaging.\n    \n3. Material cost inflation — Rising prices of gold, copper, and BT substrates.\n\nCapacity Tightening Driven by AI\n\nMorgan Stanley identified capacity shortage as the core driver of this price hike cycle. As TSMC’s CoWoS capacity remains undersupplied, orders are rapidly flowing to ASE and other OSAT firms.\n\nCurrently, ASE’s capacity utilization rate (UTR) in Q3 2025 has reached 90%, effectively indicating a shortage. This situation gives ASE strong leverage in 2026 price negotiations. To meet AI-related demand, ASE is shifting part of its wire-bonding capacity toward higher-margin flip-chip packaging.\n\nReflecting this positive outlook, Morgan Stanley raised its earnings forecasts and target prices for related companies:\n    •    ASE: Target price raised to NT$228\n    •    KYEC: Target price raised to NT$218\nBoth companies maintain an “Overweight” rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0741511175238676, "ret_signed_m1d": -0.0741511175238676, "alpha_spy_m1d": -0.07367105040019661, "alpha_c_m1d": -0.0592301943060175, "bench_spy_m1d": -0.0004800671236709908, "bench_c_m1d": -0.014920923217850102, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.042273017586732076, "ret_signed_p1d": 0.042273017586732076, "alpha_spy_p1d": 0.05327116250592345, "alpha_c_p1d": 0.055511874854873966, "bench_spy_p1d": -0.010998144919191377, "bench_c_p1d": -0.01323885726814189, "ret_p1w": 0.09702006853236989, "ret_signed_p1w": 0.09702006853236989, "alpha_spy_p1w": 0.11129143897577298, "alpha_c_p1w": 0.12173449560720373, "bench_spy_p1w": -0.014271370443403097, "bench_c_p1w": -0.024714427074833845, "ret_p1m": -0.0020790496196141373, "ret_signed_p1m": -0.0020790496196141373, "alpha_spy_p1m": 0.00913725080239125, "alpha_c_p1m": 0.054620458608216405, "bench_spy_p1m": -0.011216300422005387, "bench_c_p1m": -0.05669950822783054, "ret_p3m": 0.3742203398702211, "ret_signed_p3m": 0.3742203398702211, "alpha_spy_p3m": 0.3634740158752767, "alpha_c_p3m": 0.2888971417856536, "bench_spy_p3m": 0.010746323994944351, "bench_c_p3m": 0.08532319808456745, "ret_p6m": 1.0658349851351057, "ret_signed_p6m": 1.0658349851351057, "alpha_spy_p6m": 1.0293288986590858, "alpha_c_p6m": 0.7545592717374179, "bench_spy_p6m": 0.03650608647601983, "bench_c_p6m": 0.3112757133976878, "price_path": [-0.3063, -0.3084, -0.3257, -0.3264, -0.3132, -0.3091, -0.3063, -0.298, -0.3049, -0.3112, -0.3188, -0.3001, -0.3188, -0.3285, -0.3326, -0.3167, -0.3153, -0.3042, -0.3056, -0.3091, -0.3132, -0.3132, -0.2876, -0.2841, -0.2855, -0.2744, -0.2024, -0.2245, -0.2169, -0.2301, -0.2342, -0.2238, -0.2169, -0.2169, -0.2051, -0.2058, -0.1885, -0.1892, -0.2003, -0.2058, -0.2259, -0.2287, -0.2315, -0.2287, -0.2273, -0.2322, -0.2107, -0.2114, -0.1781, -0.1885, -0.2322, -0.1871, -0.2155, -0.1733, -0.1282, -0.1185, -0.0998, -0.1213, -0.1365, -0.1116, -0.1005, -0.0832, -0.0742, 0.0, 0.0423, 0.1095, 0.122, 0.0651, 0.097, 0.0651, 0.0547, 0.0658, 0.0381, 0.0388, 0.0139, 0.0125, -0.009, -0.018, -0.0249, -0.0471, -0.0423, -0.0187, -0.0173, -0.0021, -0.0021, 0.036, 0.0277, 0.0582, 0.0672, 0.0506, 0.0541, 0.0908, 0.1067, 0.1386, 0.1365, 0.0922, 0.0797, 0.0603, 0.0222, 0.0305, 0.0506, 0.0638, 0.0762, 0.0776, 0.0776, 0.0894, 0.1102, 0.1109, 0.1157, 0.1157, 0.1684, 0.1753, 0.2128, 0.2211, 0.201, 0.2225, 0.2661, 0.2959, 0.3084, 0.3299, 0.3451, 0.3451, 0.3098, 0.3153, 0.3209, 0.3437, 0.3742, 0.3909, 0.404, 0.3708, 0.3153, 0.3451, 0.3437, 0.316, 0.404, 0.4477, 0.535, 0.5586, 0.6452, 0.6258, 0.6216, 0.6216, 0.6265, 0.6175, 0.6064, 0.6604, 0.6272, 0.72, 0.7401, 0.6791, 0.6833, 0.6771, 0.5461, 0.5315, 0.5288, 0.4636, 0.501, 0.5045, 0.5177, 0.4546, 0.49, 0.4913, 0.5128, 0.501, 0.5177, 0.4768, 0.5024, 0.492, 0.5523, 0.4851, 0.49, 0.4477, 0.5024, 0.5579, 0.544, 0.544, 0.5627, 0.5378, 0.6729, 0.6985, 0.7214, 0.8399, 0.8621, 0.887, 0.9127, 0.9813, 1.0291, 1.0513, 1.0755, 1.0658]}
{"unit_id": "orig:1983679575940030550", "ticker_idx": 1, "ts": "2025-10-29T23:37:20+00:00", "symbol": "KYEC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KYEC: Target price raised to NT$218… Both companies maintain an 'Overweight' rating", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises TPs and reiterates Overweight on ASE and KYEC on OSAT price hike cycle driven by AI demand and capacity constraints.", "resolved_tickers": ["2449.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "King Yuan Electronics 2449.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Chip Inflation Wave Expands — Morgan Stanley Expects OSAT Companies to Raise Prices in 2026, the First Upcycle Since the Pandemic\n\nMorgan Stanley stated that the semiconductor inflation wave, which began with wafer foundries and memory, is now spreading downstream to back-end processes. Chip back-end (OSAT) companies are preparing for the first price hike cycle since the COVID-19 pandemic.\n\nMorgan Stanley warned that the strong demand for AI semiconductors is severely straining packaging and testing capacity, giving back-end companies stronger bargaining power. Taiwan’s leading firms, ASE and KYEC, are expected to lead this price increase.\n\nAnalysts predict that due to three key factors, the price of advanced packaging will rise by 5–10% in 2026, marking the first price upcycle since the semiconductor shortage during the pandemic:\n    \n1. Surging AI demand — TSMC’s CoWoS capacity shortage is quickly spilling over to ASE’s CoWoS capacity and KYEC’s testing lines.\n    \n2. Capacity constraints — ASE and KYEC are rejecting lower-margin products or reallocating capacity from non-AI wire-bonding to AI flip-chip packaging.\n    \n3. Material cost inflation — Rising prices of gold, copper, and BT substrates.\n\nCapacity Tightening Driven by AI\n\nMorgan Stanley identified capacity shortage as the core driver of this price hike cycle. As TSMC’s CoWoS capacity remains undersupplied, orders are rapidly flowing to ASE and other OSAT firms.\n\nCurrently, ASE’s capacity utilization rate (UTR) in Q3 2025 has reached 90%, effectively indicating a shortage. This situation gives ASE strong leverage in 2026 price negotiations. To meet AI-related demand, ASE is shifting part of its wire-bonding capacity toward higher-margin flip-chip packaging.\n\nReflecting this positive outlook, Morgan Stanley raised its earnings forecasts and target prices for related companies:\n    •    ASE: Target price raised to NT$228\n    •    KYEC: Target price raised to NT$218\nBoth companies maintain an “Overweight” rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08937198067632846, "ret_signed_m1d": -0.08937198067632846, "alpha_spy_m1d": -0.08889191355265746, "alpha_c_m1d": -0.07445105745847835, "bench_spy_m1d": -0.0004800671236709908, "bench_c_m1d": -0.014920923217850102, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019323671497584516, "ret_signed_p1d": -0.019323671497584516, "alpha_spy_p1d": -0.008325526578393139, "alpha_c_p1d": -0.0060848142294426255, "bench_spy_p1d": -0.010998144919191377, "bench_c_p1d": -0.01323885726814189, "ret_p1w": 0.0652173913043479, "ret_signed_p1w": 0.0652173913043479, "alpha_spy_p1w": 0.07948876174775099, "alpha_c_p1w": 0.08993181837918174, "bench_spy_p1w": -0.014271370443403097, "bench_c_p1w": -0.024714427074833845, "ret_p1m": 0.07246376811594213, "ret_signed_p1m": 0.07246376811594213, "alpha_spy_p1m": 0.08368006853794752, "alpha_c_p1m": 0.12916327634377267, "bench_spy_p1m": -0.011216300422005387, "bench_c_p1m": -0.05669950822783054, "ret_p3m": 0.4371980676328502, "ret_signed_p3m": 0.4371980676328502, "alpha_spy_p3m": 0.42645174363790583, "alpha_c_p3m": 0.35187486954828273, "bench_spy_p3m": 0.010746323994944351, "bench_c_p3m": 0.08532319808456745, "ret_p6m": 0.3405797101449275, "ret_signed_p6m": 0.3405797101449275, "alpha_spy_p6m": 0.30407362366890767, "alpha_c_p6m": 0.029303996747239713, "bench_spy_p6m": 0.03650608647601983, "bench_c_p6m": 0.3112757133976878, "price_path": [-0.43, -0.4324, -0.4275, -0.4396, -0.413, -0.3551, -0.3237, -0.3188, -0.3237, -0.3261, -0.3309, -0.3188, -0.3382, -0.3841, -0.3575, -0.3647, -0.3454, -0.3261, -0.3285, -0.314, -0.2464, -0.2343, -0.2657, -0.2585, -0.2705, -0.2271, -0.1836, -0.1763, -0.1667, -0.1981, -0.1981, -0.1957, -0.1981, -0.2029, -0.1908, -0.1812, -0.1739, -0.1932, -0.1981, -0.1981, -0.2488, -0.2488, -0.2077, -0.2029, -0.1908, -0.1836, -0.1836, -0.1667, -0.1594, -0.1498, -0.1498, -0.1787, -0.2198, -0.1981, -0.128, -0.1546, -0.1401, -0.1111, -0.1473, -0.1546, -0.1546, -0.0918, -0.0894, 0.0, -0.0193, 0.0483, 0.0821, 0.0338, 0.0652, 0.0507, 0.0314, 0.0483, 0.029, 0.0169, 0.0411, 0.0072, 0.0314, -0.0435, -0.0507, 0.0217, -0.0121, -0.0169, -0.0097, 0.0242, 0.0725, 0.1063, 0.0507, 0.0942, 0.0918, 0.0652, 0.07, 0.1014, 0.1039, 0.1063, 0.07, 0.1111, 0.07, 0.0435, 0.0386, 0.0072, 0.0411, 0.0676, 0.0628, 0.058, 0.058, 0.0773, 0.1836, 0.1763, 0.1957, 0.1957, 0.2923, 0.2705, 0.2899, 0.2874, 0.2705, 0.2729, 0.3043, 0.3068, 0.3068, 0.2923, 0.3213, 0.3116, 0.3261, 0.2899, 0.3261, 0.3841, 0.4372, 0.4493, 0.5024, 0.4734, 0.43, 0.3768, 0.4179, 0.4348, 0.3551, 0.3696, 0.4082, 0.4589, 0.5145, 0.5145, 0.5145, 0.5145, 0.5145, 0.5145, 0.5145, 0.5145, 0.529, 0.5749, 0.5459, 0.5797, 0.5797, 0.5266, 0.4734, 0.343, 0.4251, 0.3744, 0.314, 0.4155, 0.4783, 0.4493, 0.4275, 0.3913, 0.4203, 0.5604, 0.5266, 0.4879, 0.3816, 0.3454, 0.3937, 0.3913, 0.3454, 0.314, 0.2609, 0.3551, 0.2512, 0.2512, 0.2512, 0.2947, 0.3527, 0.3285, 0.3406, 0.4251, 0.4541, 0.4058, 0.3696, 0.3285, 0.3092, 0.3575, 0.3889, 0.3406]}
{"unit_id": "orig:2001623883536019784", "ticker_idx": 0, "ts": "2025-12-18T12:01:36+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We see a high likelihood that NVIDIA and AMD's server SKUs will use Intel 14A", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses GFHK bullish thesis: NVDA and AMD server SKUs likely to adopt Intel 14A, with PTL yields targeting 70% by end-2025.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK: We see a high likelihood that NVIDIA and AMD’s server SKUs will use Intel 14A.\n\nPTL yields reached 60–65% as of November, and the target is 70% by the end of 2025.\n\n$INTC https://t.co/K6pcx7gakr", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006339568632168113, "ret_signed_m1d": -0.006339568632168113, "alpha_spy_m1d": 0.001155168691335251, "alpha_c_m1d": 0.016727429316946263, "bench_spy_m1d": -0.007494737323503364, "bench_c_m1d": -0.023066997949114376, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014884259473445516, "ret_signed_p1d": 0.014884259473445516, "alpha_spy_p1d": 0.005820813993993701, "alpha_c_p1d": -0.010976082829809108, "bench_spy_p1d": 0.009063445479451815, "bench_c_p1d": 0.025860342303254624, "ret_p1w": -0.003307578167649816, "ret_signed_p1w": -0.003307578167649816, "alpha_spy_p1w": -0.026885929595842906, "alpha_c_p1w": -0.055296524683824066, "bench_spy_p1w": 0.02357835142819309, "bench_c_p1w": 0.05198894651617425, "ret_p1m": 0.2943770855711909, "ret_signed_p1m": 0.2943770855711909, "alpha_spy_p1m": 0.2689010037882136, "alpha_c_p1m": 0.13775832462208504, "bench_spy_p1m": 0.025476081782977333, "bench_c_p1m": 0.15661876094910587, "ret_p3m": 0.21444329811923213, "ret_signed_p3m": 0.21444329811923213, "alpha_spy_p3m": 0.21990973064114372, "alpha_c_p3m": 0.06796400867917751, "bench_spy_p3m": -0.005466432521911591, "bench_c_p3m": 0.14647928944005462, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2073, -0.1913, -0.1395, -0.0631, -0.0215, -0.0496, -0.0752, -0.0094, 0.0281, 0.0152, 0.0085, 0.0245, 0.0317, 0.0419, 0.0025, 0.0259, -0.0179, 0.024, 0.0154, 0.0201, 0.0502, 0.0507, 0.0176, 0.0518, 0.0551, 0.0899, 0.1447, 0.1395, 0.1069, 0.1023, 0.0888, 0.0207, 0.0579, 0.0265, 0.051, 0.0598, 0.0441, 0.0444, -0.0102, -0.0209, -0.0433, -0.0537, -0.0322, -0.0733, -0.0491, -0.0135, -0.0124, 0.0146, 0.0146, 0.118, 0.1028, 0.1982, 0.2062, 0.1163, 0.1414, 0.1108, 0.1163, 0.124, 0.089, 0.0422, 0.0339, 0.0284, -0.0063, 0.0, 0.0149, 0.0025, 0.0019, -0.0033, -0.0033, -0.0022, 0.011, 0.0281, 0.0171, 0.0171, 0.0854, 0.0852, 0.1036, 0.175, 0.1331, 0.2555, 0.2144, 0.3035, 0.3429, 0.3319, 0.2944, 0.2944, 0.3385, 0.4953, 0.4972, 0.2423, 0.1712, 0.2109, 0.3445, 0.3412, 0.2809, 0.3454, 0.3575, 0.3396, 0.3297, 0.3944, 0.3848, 0.2991, 0.331, 0.2811, 0.2897, 0.2897, 0.2729, 0.253, 0.2299, 0.2158, 0.2026, 0.2712, 0.2922, 0.253, 0.2572, 0.2541, 0.188, 0.2563, 0.2665, 0.1968, 0.2563, 0.2894, 0.3225, 0.2472, 0.2616, 0.2613, 0.2144, 0.2412, 0.2729, 0.2092, 0.2131, 0.2144, 0.3004, 0.2155, 0.1888, 0.1353, 0.2164, 0.3239, 0.3886, 0.3886, 0.3997, 0.4584, 0.6249, 0.7012, 0.7194, 0.7966, 0.7588, 0.79, 0.8881, 0.8881, 0.8109, 0.8264, 0.7991, 0.8407, 1.2751, 1.3426, 1.3297, 1.6116, 1.6042, 1.7459, 1.64, 1.981, 2.1149, 2.0215, 2.4432, 2.5678, 2.3244, 2.3156, 2.1954, 1.9981, 1.9815, 2.054, 2.2789, 2.2663, 2.3032, 2.3032, 2.4046, 2.3564, 2.3321, 2.161, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000424742302753189", "ticker_idx": 0, "ts": "2025-12-15T04:36:39+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's decision is consequently interpreted as meaning the memory industry will no longer sell products at bargain prices", "tweet_type": "analysis", "timeline": "months", "summary": "NAND majors (Micron, Samsung, SK Hynix) pivoting to high-margin server SSD/AI supply, cutting consumer exposure; persistent price hikes support profitability uplift.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Name Your Price\"... NAND Industry Focuses on 'Maximizing Profitability'... Samsung and SK Hynix Concentrate on Server SSD Production Instead of PC SSDs\n\nThe NAND flash industry is showing a changed stance by boldly phasing out low-profit production and aggressively raising prices.\n\nAlthough NAND, used for data storage, was previously evaluated as facing inevitable market consolidation due to the large number of suppliers, the prevailing sentiment recently is that the players have become \"ruthless.\" This is because they are executing a strategy to maximize profits without increasing production, despite concerns about prolonged supply shortages.\n\nThis is a scene not seen in past memory cycles.\n\nAccording to the industry on the 15th, US memory manufacturer Micron has decided to operate 'Crucial,' its consumer memory brand for Solid State Drives (SSDs), only until the end of next February.\n\nThe industry views this suspension of the Crucial brand operations as a symbolic event marking the shift of the memory industry's center of gravity from consumers to enterprises.\n\nCrucial is a brand launched by Micron in 1996 to create an environment where general consumers could upgrade computer performance themselves like professionals. Through this, Micron became the first company to sell memory directly to consumers.\n\nHowever, Micron has now decided to be the first to withdraw from the consumer market.\n\nSumit Sadana, Executive Vice President and Chief Business Officer, stated the reason was \"to improve supply and support for key strategic customers such as rapidly growing AI data centers.\" This implies that core customers are moving from PCs and smartphones to servers.\n\nMicron's decision is consequently interpreted as meaning the memory industry will no longer sell products at bargain prices.\n\nRecently, big tech companies like Google, Amazon, Meta, and Microsoft are sweeping up memory to build artificial intelligence (AI) infrastructure. As memory is a 'strategic asset' for the AI war, they appear willing to accept price hikes.\n\nIn this situation, the concerns of Samsung Electronics and SK Hynix are also growing. Since cleanrooms (manufacturing facilities) are limited, they must maximize product production efficiency.\n\nThere is high concern in the industry that Samsung Electronics and SK Hynix might also reduce the production of consumer products.\n\nA tech YouTuber active overseas recently claimed that Samsung Electronics would soon end production of 'SATA SSDs,' which are relatively low-value-added products. However, Samsung Electronics denied this.\n\nHowever, with the memory industry operating with a focus on profitability, the observation is that Samsung Electronics could change its supply priorities at any time. SK Hynix is also reportedly concentrating on NAND production for AI data centers instead of PC SSDs, which have high concerns about slowing growth.\n\nAs a result, consumer NAND product prices may face continuous upward pressure.\n\nAccording to the foreign media outlet Business Insider, global PC maker Dell plans to raise the prices of all enterprise laptops by 10-30% starting on the 17th.\n\nAn industry insider said, \"Due to the unprecedented rise in memory prices, demand for smartphones and PCs may slow down starting from the second half of next year,\" adding, \"Production strategies focused on server customers will likely intensify.\"\n\n---\n\nhttps://t.co/1FOcgd92c2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015326278026813522, "ret_signed_m1d": 0.015326278026813522, "alpha_spy_m1d": 0.013813194179786104, "alpha_c_m1d": 0.01186920406305747, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021010569585278582, "ret_signed_p1d": -0.021010569585278582, "alpha_spy_p1d": -0.018278271989999872, "alpha_c_p1d": -0.018290226336253657, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.1645894394682299, "ret_signed_p1w": 0.1645894394682299, "alpha_spy_p1w": 0.15559371415552747, "alpha_c_p1w": 0.14214894979603554, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4242803264561392, "ret_signed_p1m": 0.4242803264561392, "alpha_spy_p1m": 0.4021128365504283, "alpha_c_p1m": 0.31136370124665924, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.7074262303029928, "ret_signed_p3m": 0.7074262303029928, "alpha_spy_p3m": 0.7260854035676849, "alpha_c_p3m": 0.604174023515615, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3268, -0.2893, -0.3152, -0.3073, -0.2998, -0.3195, -0.3401, -0.3382, -0.3103, -0.2959, -0.2335, -0.2268, -0.2091, -0.196, -0.2181, -0.1725, -0.1902, -0.2354, -0.1883, -0.2124, -0.1918, -0.1472, -0.1479, -0.1294, -0.1483, -0.1643, -0.1296, -0.0778, -0.0733, -0.0656, -0.0458, -0.0568, -0.0578, -0.0118, -0.082, 0.0, 0.0035, 0.0018, 0.0665, 0.0152, 0.0312, -0.0023, 0.0393, 0.0187, -0.0379, -0.0488, -0.1521, -0.1269, -0.0571, -0.0546, -0.0305, -0.0305, -0.0043, 0.0125, 0.0084, -0.0141, -0.0457, -0.0012, 0.0397, 0.0628, 0.1104, 0.0883, 0.0153, 0.0, -0.021, -0.0504, 0.0465, 0.1197, 0.1646, 0.1632, 0.2071, 0.2071, 0.1991, 0.24, 0.2326, 0.2022, 0.2022, 0.3286, 0.3148, 0.4466, 0.4303, 0.3775, 0.4536, 0.4569, 0.4243, 0.4041, 0.418, 0.528, 0.528, 0.5375, 0.639, 0.6747, 0.6834, 0.6389, 0.728, 0.8335, 0.8356, 0.7476, 0.8441, 0.7668, 0.5981, 0.6128, 0.6625, 0.6154, 0.5722, 0.7284, 0.7437, 0.734, 0.734, 0.684, 0.7731, 0.758, 0.8035, 0.7732, 0.7608, 0.807, 0.7504, 0.737, 0.7383, 0.5993, 0.6881, 0.6725, 0.5598, 0.6399, 0.698, 0.7636, 0.7074, 0.795, 0.861, 0.9447, 0.9449, 0.8714, 0.7814, 0.7032, 0.6661, 0.6094, 0.4973, 0.5047, 0.3561, 0.4237, 0.5501, 0.5433, 0.5433, 0.5919, 0.5911, 0.714, 0.7762, 0.7724, 0.7975, 0.9623, 0.9226, 0.9268, 0.9177, 0.8896, 0.8937, 1.0542, 1.03, 1.0932, 1.2105, 1.1251, 1.1848, 1.1793, 1.2849, 1.4292, 1.6978, 1.809, 1.7249, 2.1471, 2.3515, 2.2304, 2.3865, 2.2701, 2.0537, 1.872, 1.9445, 2.0846, 2.2115, 2.1647, 2.1647, 2.7752, 2.9123, 2.8917, 3.0918, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000424742302753189", "ticker_idx": 1, "ts": "2025-12-15T04:36:39+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "consumer NAND product prices may face continuous upward pressure", "tweet_type": "analysis", "timeline": "months", "summary": "NAND majors (Micron, Samsung, SK Hynix) pivoting to high-margin server SSD/AI supply, cutting consumer exposure; persistent price hikes support profitability uplift.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Name Your Price\"... NAND Industry Focuses on 'Maximizing Profitability'... Samsung and SK Hynix Concentrate on Server SSD Production Instead of PC SSDs\n\nThe NAND flash industry is showing a changed stance by boldly phasing out low-profit production and aggressively raising prices.\n\nAlthough NAND, used for data storage, was previously evaluated as facing inevitable market consolidation due to the large number of suppliers, the prevailing sentiment recently is that the players have become \"ruthless.\" This is because they are executing a strategy to maximize profits without increasing production, despite concerns about prolonged supply shortages.\n\nThis is a scene not seen in past memory cycles.\n\nAccording to the industry on the 15th, US memory manufacturer Micron has decided to operate 'Crucial,' its consumer memory brand for Solid State Drives (SSDs), only until the end of next February.\n\nThe industry views this suspension of the Crucial brand operations as a symbolic event marking the shift of the memory industry's center of gravity from consumers to enterprises.\n\nCrucial is a brand launched by Micron in 1996 to create an environment where general consumers could upgrade computer performance themselves like professionals. Through this, Micron became the first company to sell memory directly to consumers.\n\nHowever, Micron has now decided to be the first to withdraw from the consumer market.\n\nSumit Sadana, Executive Vice President and Chief Business Officer, stated the reason was \"to improve supply and support for key strategic customers such as rapidly growing AI data centers.\" This implies that core customers are moving from PCs and smartphones to servers.\n\nMicron's decision is consequently interpreted as meaning the memory industry will no longer sell products at bargain prices.\n\nRecently, big tech companies like Google, Amazon, Meta, and Microsoft are sweeping up memory to build artificial intelligence (AI) infrastructure. As memory is a 'strategic asset' for the AI war, they appear willing to accept price hikes.\n\nIn this situation, the concerns of Samsung Electronics and SK Hynix are also growing. Since cleanrooms (manufacturing facilities) are limited, they must maximize product production efficiency.\n\nThere is high concern in the industry that Samsung Electronics and SK Hynix might also reduce the production of consumer products.\n\nA tech YouTuber active overseas recently claimed that Samsung Electronics would soon end production of 'SATA SSDs,' which are relatively low-value-added products. However, Samsung Electronics denied this.\n\nHowever, with the memory industry operating with a focus on profitability, the observation is that Samsung Electronics could change its supply priorities at any time. SK Hynix is also reportedly concentrating on NAND production for AI data centers instead of PC SSDs, which have high concerns about slowing growth.\n\nAs a result, consumer NAND product prices may face continuous upward pressure.\n\nAccording to the foreign media outlet Business Insider, global PC maker Dell plans to raise the prices of all enterprise laptops by 10-30% starting on the 17th.\n\nAn industry insider said, \"Due to the unprecedented rise in memory prices, demand for smartphones and PCs may slow down starting from the second half of next year,\" adding, \"Production strategies focused on server customers will likely intensify.\"\n\n---\n\nhttps://t.co/1FOcgd92c2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03912210331054089, "ret_signed_m1d": 0.03912210331054089, "alpha_spy_m1d": 0.03760901946351347, "alpha_c_m1d": 0.029354021596027735, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.009768081714513155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019083943680241022, "ret_signed_p1d": -0.019083943680241022, "alpha_spy_p1d": -0.016351646084962312, "alpha_c_p1d": -0.02091102355885721, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": 0.001827079878616189, "ret_p1w": 0.05438933331892093, "ret_signed_p1w": 0.05438933331892093, "alpha_spy_p1w": 0.0453936080062185, "alpha_c_p1w": 0.032743727030979164, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.021645606287941765, "ret_p1m": 0.319359684466187, "ret_signed_p1m": 0.319359684466187, "alpha_spy_p1m": 0.2971921945604761, "alpha_c_p1m": 0.2884238415073155, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.03093584295887153, "ret_p3m": 0.8016547449724258, "ret_signed_p3m": 0.8016547449724258, "alpha_spy_p3m": 0.820313918237118, "alpha_c_p3m": 0.831527802060285, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.029873057087859145, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2571, -0.2372, -0.2372, -0.2068, -0.1954, -0.1887, -0.1821, -0.2087, -0.1966, -0.1994, -0.1794, -0.1436, -0.1436, -0.1436, -0.1436, -0.1436, -0.1436, -0.0992, -0.1097, -0.126, -0.0935, -0.0677, -0.0658, -0.0639, -0.0697, -0.0592, -0.0792, -0.0573, -0.0267, -0.0506, -0.041, -0.0067, 0.0258, 0.0601, 0.001, -0.0401, -0.0534, -0.0658, -0.0401, -0.0124, -0.0162, -0.0191, -0.0725, -0.0401, -0.0668, -0.0792, -0.0401, -0.0954, -0.0773, -0.0525, -0.0191, -0.0124, -0.041, -0.0382, -0.0134, -0.0029, 0.0029, 0.0344, 0.0448, 0.0344, 0.0305, 0.0239, 0.0391, 0.0, -0.0191, 0.0296, 0.0267, 0.0143, 0.0544, 0.0639, 0.0601, 0.0601, 0.1164, 0.1458, 0.1496, 0.1496, 0.1496, 0.2321, 0.3242, 0.3318, 0.352, 0.3309, 0.3328, 0.3309, 0.3194, 0.3452, 0.3798, 0.4277, 0.4315, 0.3922, 0.4335, 0.4603, 0.4584, 0.4584, 0.5293, 0.5572, 0.5409, 0.5389, 0.4421, 0.6061, 0.6214, 0.5274, 0.5207, 0.5955, 0.5898, 0.6089, 0.7125, 0.7374, 0.7374, 0.7374, 0.7374, 0.8218, 0.8227, 0.8506, 0.9177, 0.9512, 1.0903, 1.0759, 1.0759, 0.8707, 0.6511, 0.8371, 0.8045, 0.6636, 0.8017, 0.8218, 0.8017, 0.7595, 0.8093, 0.8592, 0.9992, 0.9225, 0.9119, 0.7863, 0.8189, 0.8122, 0.7269, 0.7269, 0.6939, 0.6065, 0.8222, 0.7141, 0.789, 0.8553, 0.888, 1.0225, 0.9601, 0.9793, 0.9313, 0.9841, 1.0273, 1.0898, 1.0754, 1.061, 1.1042, 1.0898, 1.157, 1.109, 1.157, 1.133, 1.1715, 1.1186, 1.1186, 1.2339, 1.2339, 1.5558, 1.6086, 1.5798, 1.7431, 1.6807, 1.7287, 1.844, 1.599, 1.6999, 1.6471, 1.6519, 1.8777, 1.8104, 1.8104, 1.8729, 1.9497, 1.8777, 2.0458, null, null, null, null, null, null, null]}
{"unit_id": "thread:2000424742302753189", "ticker_idx": 2, "ts": "2025-12-15T04:36:39+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is also reportedly concentrating on NAND production for AI data centers instead of PC SSDs", "tweet_type": "analysis", "timeline": "months", "summary": "NAND majors (Micron, Samsung, SK Hynix) pivoting to high-margin server SSD/AI supply, cutting consumer exposure; persistent price hikes support profitability uplift.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Name Your Price\"... NAND Industry Focuses on 'Maximizing Profitability'... Samsung and SK Hynix Concentrate on Server SSD Production Instead of PC SSDs\n\nThe NAND flash industry is showing a changed stance by boldly phasing out low-profit production and aggressively raising prices.\n\nAlthough NAND, used for data storage, was previously evaluated as facing inevitable market consolidation due to the large number of suppliers, the prevailing sentiment recently is that the players have become \"ruthless.\" This is because they are executing a strategy to maximize profits without increasing production, despite concerns about prolonged supply shortages.\n\nThis is a scene not seen in past memory cycles.\n\nAccording to the industry on the 15th, US memory manufacturer Micron has decided to operate 'Crucial,' its consumer memory brand for Solid State Drives (SSDs), only until the end of next February.\n\nThe industry views this suspension of the Crucial brand operations as a symbolic event marking the shift of the memory industry's center of gravity from consumers to enterprises.\n\nCrucial is a brand launched by Micron in 1996 to create an environment where general consumers could upgrade computer performance themselves like professionals. Through this, Micron became the first company to sell memory directly to consumers.\n\nHowever, Micron has now decided to be the first to withdraw from the consumer market.\n\nSumit Sadana, Executive Vice President and Chief Business Officer, stated the reason was \"to improve supply and support for key strategic customers such as rapidly growing AI data centers.\" This implies that core customers are moving from PCs and smartphones to servers.\n\nMicron's decision is consequently interpreted as meaning the memory industry will no longer sell products at bargain prices.\n\nRecently, big tech companies like Google, Amazon, Meta, and Microsoft are sweeping up memory to build artificial intelligence (AI) infrastructure. As memory is a 'strategic asset' for the AI war, they appear willing to accept price hikes.\n\nIn this situation, the concerns of Samsung Electronics and SK Hynix are also growing. Since cleanrooms (manufacturing facilities) are limited, they must maximize product production efficiency.\n\nThere is high concern in the industry that Samsung Electronics and SK Hynix might also reduce the production of consumer products.\n\nA tech YouTuber active overseas recently claimed that Samsung Electronics would soon end production of 'SATA SSDs,' which are relatively low-value-added products. However, Samsung Electronics denied this.\n\nHowever, with the memory industry operating with a focus on profitability, the observation is that Samsung Electronics could change its supply priorities at any time. SK Hynix is also reportedly concentrating on NAND production for AI data centers instead of PC SSDs, which have high concerns about slowing growth.\n\nAs a result, consumer NAND product prices may face continuous upward pressure.\n\nAccording to the foreign media outlet Business Insider, global PC maker Dell plans to raise the prices of all enterprise laptops by 10-30% starting on the 17th.\n\nAn industry insider said, \"Due to the unprecedented rise in memory prices, demand for smartphones and PCs may slow down starting from the second half of next year,\" adding, \"Production strategies focused on server customers will likely intensify.\"\n\n---\n\nhttps://t.co/1FOcgd92c2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030685848548443806, "ret_signed_m1d": 0.030685848548443806, "alpha_spy_m1d": 0.029172764701416387, "alpha_c_m1d": 0.027228774584687754, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.043321337592038156, "ret_signed_p1d": -0.043321337592038156, "alpha_spy_p1d": -0.040589039996759446, "alpha_c_p1d": -0.04060099434301323, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.046931364275827114, "ret_signed_p1w": 0.046931364275827114, "alpha_spy_p1w": 0.037935638963124685, "alpha_c_p1w": 0.02449087460363275, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3321299157434354, "ret_signed_p1m": 0.3321299157434354, "alpha_spy_p1m": 0.3099624258377245, "alpha_c_p1m": 0.21921329053395544, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.6817978803316003, "ret_signed_p3m": 0.6817978803316003, "alpha_spy_p3m": 0.7004570535962924, "alpha_c_p3m": 0.5785456735442225, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3984, -0.3588, -0.3588, -0.3669, -0.3488, -0.3552, -0.357, -0.393, -0.3705, -0.3732, -0.3506, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.228, -0.2514, -0.2578, -0.2379, -0.1838, -0.1603, -0.1243, -0.136, -0.1315, -0.1369, -0.0801, -0.035, -0.0602, 0.0065, 0.0245, 0.0083, 0.1183, 0.057, 0.0444, 0.0696, 0.0462, 0.0931, 0.1165, 0.1129, 0.1039, 0.0101, 0.0931, 0.0281, 0.0137, 0.0299, -0.0602, -0.062, -0.0638, -0.0548, -0.0181, -0.0433, -0.0289, 0.0072, -0.0036, -0.0217, -0.0181, 0.0415, 0.0217, 0.0596, 0.0199, 0.0307, 0.0, -0.0433, -0.0054, -0.0036, -0.0126, 0.0469, 0.0542, 0.0614, 0.0614, 0.0812, 0.1552, 0.1751, 0.1751, 0.1751, 0.222, 0.2563, 0.3105, 0.3394, 0.3646, 0.343, 0.352, 0.3321, 0.3394, 0.352, 0.3646, 0.3791, 0.3412, 0.3357, 0.3628, 0.3845, 0.3285, 0.444, 0.5181, 0.5542, 0.6408, 0.4982, 0.6372, 0.6245, 0.5199, 0.5144, 0.6011, 0.5812, 0.5523, 0.6029, 0.5884, 0.5884, 0.5884, 0.5884, 0.6137, 0.713, 0.7166, 0.8141, 0.8375, 0.9874, 0.9187, 0.9187, 0.6981, 0.5353, 0.7017, 0.6709, 0.5118, 0.6963, 0.727, 0.6818, 0.6456, 0.7614, 0.7541, 0.9097, 0.8319, 0.821, 0.6872, 0.7831, 0.7993, 0.6872, 0.6872, 0.5787, 0.4594, 0.6221, 0.501, 0.5841, 0.6022, 0.6565, 0.8681, 0.8048, 0.8572, 0.8807, 0.9946, 1.0543, 1.0887, 1.0399, 1.1086, 1.2135, 1.2117, 1.2153, 1.2098, 1.3364, 1.3509, 1.3382, 1.3256, 1.3256, 1.6167, 1.6167, 1.8952, 1.9911, 2.0489, 2.3998, 2.3184, 2.5734, 2.5625, 2.2895, 2.3274, 2.1556, 2.1556, 2.5083, 2.5101, 2.5101, 2.7108, 3.0562, 3.1401, 3.2197, null, null, null, null, null, null, null]}
{"unit_id": "quote:1990220906447454581", "ticker_idx": 0, "ts": "2025-11-17T00:50:15+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "OW → UW, TP $144 → $110", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Morgan Stanley downgrade of DELL from OW to UW, PT cut to $110 on memory-price headwinds to OEM EPS.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "Downgraded Dell to UW due to the sharp surge in memory prices, says Morgan Stanley https://t.co/QYhpPJOgei", "reply_text": null, "parent_text": null, "quoted_text": "This is a very important signal.\n\nMorgan Stanley has cut PC OEMs’ price targets across the board due to the sharp rise in memory prices.\n\nThey revised their targets as follows, projecting that OEMs’ EPS could face a headwind of up to -16%:\n\n• DELL\no OW → UW, TP $144 → $110.\no https://t.co/8bu5guIU8A", "parent_author": null, "ret_m1d": 0.09209663654308153, "ret_signed_m1d": -0.09209663654308153, "alpha_spy_m1d": -0.08269249873720086, "alpha_c_m1d": -0.07619615842934091, "bench_spy_m1d": 0.009404137805880675, "bench_c_m1d": 0.015900478113740624, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0017145849039588157, "ret_signed_p1d": -0.0017145849039588157, "alpha_spy_p1d": -0.010112045462165375, "alpha_c_p1d": -0.017967539062113458, "bench_spy_p1d": -0.00839746055820656, "bench_c_p1d": -0.016252954158154642, "ret_p1w": 0.038700128633091024, "ret_signed_p1w": -0.038700128633091024, "alpha_spy_p1w": -0.03410327350129627, "alpha_c_p1w": -0.052555770974019644, "bench_spy_p1w": 0.004596855131794753, "bench_c_p1w": -0.01385564234092862, "ret_p1m": 0.09201492688585988, "ret_signed_p1m": -0.09201492688585988, "alpha_spy_p1m": -0.07218522012337147, "alpha_c_p1m": -0.08679705148457728, "bench_spy_p1m": 0.019829706762488408, "bench_c_p1m": 0.005217875401282601, "ret_p3m": -0.07484027236803437, "ret_signed_p3m": 0.07484027236803437, "alpha_spy_p3m": 0.10129955325542261, "alpha_c_p3m": 0.057925629939175605, "bench_spy_p3m": 0.026459280887388248, "bench_c_p3m": -0.01691464242885876, "ret_p6m": 0.9654463540642615, "ret_signed_p6m": -0.9654463540642615, "alpha_spy_p6m": -0.850204262801781, "alpha_c_p6m": -0.7266654658828902, "bench_spy_p6m": 0.11524209126248053, "bench_c_p6m": 0.23878088818137133, "price_path": [0.0453, 0.04, 0.0645, 0.0658, 0.0657, 0.078, 0.0906, -0.0062, -0.0062, -0.0159, 0.009, 0.0305, 0.0156, 0.0007, -0.0132, 0.0125, 0.02, 0.0173, 0.0316, 0.0388, 0.0604, 0.0748, 0.0734, 0.1039, 0.0929, 0.0746, 0.0654, 0.0638, 0.0894, 0.1534, 0.2177, 0.1989, 0.145, 0.1858, 0.2274, 0.3386, 0.2687, 0.225, 0.248, 0.2103, 0.2504, 0.231, 0.217, 0.203, 0.22, 0.2258, 0.2592, 0.2952, 0.3242, 0.3462, 0.3357, 0.3146, 0.3227, 0.3072, 0.2626, 0.2444, 0.218, 0.1977, 0.165, 0.1329, 0.1488, 0.0936, 0.0921, 0.0, 0.0017, -0.0253, -0.0415, 0.0002, 0.0387, 0.0281, 0.088, 0.088, 0.0887, 0.0785, 0.11, 0.091, 0.1348, 0.1341, 0.1464, 0.1285, 0.1482, 0.1316, 0.0612, 0.0656, 0.092, 0.0442, 0.0038, 0.0322, 0.0337, 0.042, 0.0482, 0.0482, 0.0552, 0.0407, 0.0444, 0.0278, 0.0278, 0.0434, 0.0125, 0.0118, -0.0197, -0.0325, -0.0152, -0.0164, -0.023, -0.0309, -0.023, -0.0159, -0.0159, -0.0892, -0.0712, -0.0392, -0.0534, -0.0493, -0.0598, -0.0379, -0.0283, -0.0616, -0.0229, -0.0393, 0.0008, -0.0538, -0.0074, -0.0085, 0.0333, 0.0182, -0.0748, -0.0365, -0.0365, -0.048, -0.0424, -0.0237, 0.0027, -0.023, -0.0178, 0.0126, -0.0041, 0.2143, 0.2592, 0.1905, 0.2063, 0.2015, 0.2012, 0.2014, 0.1792, 0.2085, 0.2293, 0.2433, 0.2837, 0.2547, 0.2236, 0.2855, 0.2929, 0.3497, 0.4507, 0.5089, 0.4418, 0.4089, 0.3503, 0.3459, 0.389, 0.4299, 0.4299, 0.4201, 0.4571, 0.5209, 0.488, 0.458, 0.5563, 0.513, 0.4538, 0.5834, 0.6118, 0.6748, 0.7468, 0.7656, 0.745, 0.7775, 0.7765, 0.6939, 0.6917, 0.7188, 0.7288, 0.7409, 0.7794, 0.9643, 0.8941, 1.1425, 1.0321, 0.9654]}
{"unit_id": "quote:1984583938002862095", "ticker_idx": 0, "ts": "2025-11-01T11:30:57+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix target price raised to 840,000 KRW.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses bullish SK Hynix PT raise to 840,000 KRW on limited 2026 supply expansion and super-cycle OPM projections.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "“Supply expansion in 2026 will be limited.”\n\n“SK Hynix target price raised to 840,000 KRW.”\n\nThat’s insane. https://t.co/zVOjeP4hvy", "reply_text": null, "parent_text": null, "quoted_text": "\"Consequently, we project OPM for commodity DRAM\nand NAND to rise to levels seen during past super-cycle peaks, reaching 60-70% and 30-40%, respectively.\" https://t.co/I3o3l2ZKG6", "parent_author": null, "ret_m1d": -0.09838715073877069, "ret_signed_m1d": -0.09838715073877069, "alpha_spy_m1d": -0.09651397900800696, "alpha_c_m1d": -0.08997409673737111, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.054838835866490854, "ret_signed_p1d": -0.054838835866490854, "alpha_spy_p1d": -0.04298529592404332, "alpha_c_p1d": -0.018400720224020173, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": -0.022580738743366635, "ret_signed_p1w": -0.022580738743366635, "alpha_spy_p1w": -0.019800309808329364, "alpha_c_p1w": -0.002613522001512103, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": -0.09935551328319481, "ret_signed_p1m": -0.09935551328319481, "alpha_spy_p1m": -0.09670677138394612, "alpha_c_p1m": -0.08113641365297852, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": 0.38970416322104473, "ret_signed_p3m": 0.38970416322104473, "alpha_spy_p3m": 0.37104466717996476, "alpha_c_p3m": 0.24570215906573378, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 1.102147490510704, "ret_signed_p6m": 1.102147490510704, "alpha_spy_p6m": 1.054729917297967, "alpha_c_p6m": 0.7562353678043552, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.5837, -0.578, -0.5869, -0.57, -0.5668, -0.5523, -0.5547, -0.5547, -0.5692, -0.5764, -0.5885, -0.6054, -0.5957, -0.5821, -0.5788, -0.5813, -0.5669, -0.5661, -0.5871, -0.5798, -0.5766, -0.5718, -0.5589, -0.5532, -0.5355, -0.5097, -0.5048, -0.4702, -0.4661, -0.4387, -0.4621, -0.4266, -0.4266, -0.4339, -0.4177, -0.4234, -0.425, -0.4573, -0.4371, -0.4395, -0.4194, -0.3621, -0.3621, -0.3621, -0.3621, -0.3621, -0.3621, -0.3097, -0.3306, -0.3363, -0.3185, -0.2702, -0.2492, -0.2169, -0.2274, -0.2234, -0.2282, -0.1774, -0.1371, -0.1597, -0.1, -0.0839, -0.0984, 0.0, -0.0548, -0.0661, -0.0435, -0.0645, -0.0226, -0.0016, -0.0048, -0.0129, -0.0968, -0.0226, -0.0806, -0.0935, -0.079, -0.1597, -0.1613, -0.1629, -0.1548, -0.122, -0.1445, -0.1316, -0.0994, -0.109, -0.1252, -0.122, -0.0687, -0.0864, -0.0525, -0.0881, -0.0784, -0.1058, -0.1445, -0.1107, -0.109, -0.1171, -0.0638, -0.0574, -0.0509, -0.0509, -0.0332, 0.033, 0.0508, 0.0508, 0.0508, 0.0927, 0.1234, 0.1718, 0.1976, 0.2202, 0.2009, 0.2089, 0.1912, 0.1976, 0.2089, 0.2202, 0.2331, 0.1992, 0.1944, 0.2186, 0.238, 0.1879, 0.2912, 0.3574, 0.3897, 0.4672, 0.3397, 0.464, 0.4527, 0.359, 0.3542, 0.4317, 0.4139, 0.3881, 0.4333, 0.4204, 0.4204, 0.4204, 0.4204, 0.443, 0.5317, 0.535, 0.6221, 0.6431, 0.7771, 0.7157, 0.7157, 0.5184, 0.3729, 0.5216, 0.4941, 0.3518, 0.5168, 0.5443, 0.5038, 0.4715, 0.575, 0.5685, 0.7076, 0.6381, 0.6284, 0.5087, 0.5944, 0.609, 0.5087, 0.5087, 0.4117, 0.3049, 0.4505, 0.3421, 0.4165, 0.4327, 0.4812, 0.6704, 0.6138, 0.6607, 0.6817, 0.7836, 0.837, 0.8677, 0.824, 0.8855, 0.9793, 0.9776, 0.9809, 0.976, 1.0892, 1.1021]}
{"unit_id": "thread:1985187152570077257", "ticker_idx": 0, "ts": "2025-11-03T03:27:54+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics led the move by halting DDR5 DRAM contract pricing in October, prompting other manufacturers to follow suit", "tweet_type": "analysis", "timeline": "quarters", "summary": "DDR5 spot prices tripling and top-3 DRAM vendors restricting contract supply signal sustained pricing power for Samsung, SK Hynix, and Micron through 1H 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Samsung delays DDR5 contract pricing to mid-November as spot prices triple\n\nMemory shortages are intensifying as major cloud service providers (CSPs) continue expanding capital expenditures, with AI demand increasingly squeezing supply. Industry sources reveal that Samsung Electronics led the move by halting DDR5 DRAM contract pricing in October, prompting other manufacturers to follow suit. This disruption is expected to delay contract price announcements until mid-November, exacerbating supply constraints. Meanwhile, DRAM spot prices have surged sharply, with DDR5 prices more than doubling since late last month, driving procurement costs to soar.\n\nSamsung stalls contract pricing; stricter supply strategies from top three vendors\nIndustry insiders note that although a memory contract price increase in the fourth quarter of 2025 was widely anticipated, Samsung's unexpected refusal to provide contract quotes before the end of October has left downstream customers without available stock. This caused DDR5 spot prices to jump 25% within just one week.\n\nMarket reports indicate Samsung plans to resume contract price confirmations only by mid-November. SK Hynix and Micron have also withheld contract pricing, signaling a tightening supply strategy among the top three suppliers, who will now offer quotes exclusively to long-term partners.\n\nThis shift suggests non-quote periods may become standard, forcing urgent buyers to compete aggressively in the spot market.\n\nDRAM spot prices surge dramatically; DDR5 triples in value\nWhile fourth-quarter contract prices remain unconfirmed, projections suggest DDR5 prices will experience \"three consecutive jumps\" from the fourth quarter of 2025 through the first half of 2026, with quarterly increases ranging between 30% and 50%. Under this trajectory, DDR5 16Gb chip prices could reach historic highs of near US$30 by early 2026.\n\nOver the past month, DRAM spot prices have skyrocketed across the board. DDR5 16Gb prices soared to US$15.5, up from US$7.68 at September's end—a staggering 102% monthly increase—making it the fastest-rising product line.\n\nDDR4 prices also climbed steadily: DDR4 16Gb modules exceeded US$25, with Taiwanese manufacturers' scarce DDR4 16Gb (1Gx16) units reaching US$27 due to limited availability.\n\nThe DDR4 16Gb (2Gx8) variant saw nearly 90% monthly growth, hitting around US$25, second only to the same-spec DDR5 16Gb (2Gx8).\n\nMainstream DDR4 8Gb modules rose 60% over the month, while DDR3 4Gb prices increased 40%, marking the highest single-month rise since 2025 began.\n\nFlash wafer prices climb amid memory seller's market\nNot only have DDR chip prices surged, but flash wafer quotations have also risen significantly. Since September, 512Gb TLC wafers climbed steadily, reaching US$5 by late October. Prices for 1Tb TLC/QLC wafers hit US$6.5-7.2.\n\nMemory industry experts attribute these hikes to extremely tight circulating inventory. Despite restocking needs, suppliers hesitate to buy heavily due to high prices and limited supply, relying mostly on existing stock to maintain production.\n\nWith upstream manufacturers restricting fourth-quarter contract pricing mainly to tech giants and leading CSPs, DDR5 supplies are virtually unavailable to general customers, leaving many frustrated as the memory market fully shifts into a seller's market.\n\nMajor US CSPs recently raised their capex budgets: Amazon increased its capex from US$118 billion to US$125 billion, with further growth expected in 2026. Google, Meta, and Microsoft have also revised their spending plans upward, boosting investments in AI computing and high-end memory chips. This strong demand is causing structural capacity shifts in the overall supply chain.\n\nAccording to market data, DRAM contract prices in the fourth quarter of 2025 have already surged 1.7x year-over-year. To maximize profits, manufacturers are shifting capacity toward higher-margin DDR5 and high-bandwidth memory (HBM) products. However, DDR5 supply shortages have exceeded expectations. With Samsung suspending DDR5 contract pricing first, module makers worry about worsening shortages. Yet if they refrain from purchasing, prices risk climbing even higher, creating a dilemma for procurement costs over the next six months.\n\n---\n\nhttps://t.co/zuAvPuRgTT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032403197304098796, "ret_signed_m1d": -0.032403197304098796, "alpha_spy_m1d": -0.030530025573335062, "alpha_c_m1d": -0.0283290976270143, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.004074099677084497, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05580554187198228, "ret_signed_p1d": -0.05580554187198228, "alpha_spy_p1d": -0.043952001929534745, "alpha_c_p1d": -0.029406908541155907, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.02639863333082637, "ret_p1w": -0.09450941397981827, "ret_signed_p1w": -0.09450941397981827, "alpha_spy_p1w": -0.091728985044781, "alpha_c_p1w": -0.07337718849400054, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.021132225485817724, "ret_p1m": -0.06930692718774134, "ret_signed_p1m": -0.06930692718774134, "alpha_spy_p1m": -0.06665818528849266, "alpha_c_p1m": -0.02753944866763469, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.04176747852010665, "ret_p3m": 0.45347602768946293, "ret_signed_p3m": 0.45347602768946293, "alpha_spy_p3m": 0.43481653164838296, "alpha_c_p3m": 0.4790614425869597, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": -0.025585414897496772, "ret_p6m": 1.0120694427362107, "ret_signed_p6m": 1.0120694427362107, "alpha_spy_p6m": 0.9646518695234736, "alpha_c_p6m": 0.963505412351459, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.048564030384751655, "price_path": [-0.3835, -0.3683, -0.3566, -0.3638, -0.3629, -0.3557, -0.3584, -0.3584, -0.3727, -0.3727, -0.3683, -0.3674, -0.3602, -0.3593, -0.37, -0.3674, -0.3763, -0.3754, -0.3942, -0.3808, -0.3745, -0.3718, -0.3772, -0.3718, -0.3593, -0.3494, -0.3423, -0.3243, -0.3145, -0.2885, -0.2993, -0.2804, -0.2804, -0.2518, -0.241, -0.2347, -0.2285, -0.2536, -0.2421, -0.2448, -0.2259, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1503, -0.1602, -0.1755, -0.1449, -0.1206, -0.1188, -0.117, -0.1224, -0.1125, -0.1314, -0.1107, -0.0819, -0.1044, -0.0954, -0.063, -0.0324, 0.0, -0.0558, -0.0945, -0.1071, -0.1188, -0.0945, -0.0684, -0.072, -0.0747, -0.1251, -0.0945, -0.1197, -0.1314, -0.0945, -0.1467, -0.1296, -0.1062, -0.0747, -0.0684, -0.0954, -0.0927, -0.0693, -0.0594, -0.054, -0.0243, -0.0144, -0.0243, -0.0279, -0.0342, -0.0198, -0.0567, -0.0747, -0.0288, -0.0315, -0.0432, -0.0054, 0.0036, 0.0, 0.0, 0.0531, 0.0808, 0.0845, 0.0845, 0.0845, 0.1622, 0.2491, 0.2563, 0.2753, 0.2554, 0.2572, 0.2554, 0.2445, 0.269, 0.3015, 0.3467, 0.3504, 0.3133, 0.3522, 0.3775, 0.3757, 0.3757, 0.4426, 0.4689, 0.4535, 0.4517, 0.3603, 0.515, 0.5295, 0.4408, 0.4345, 0.505, 0.4996, 0.5177, 0.6154, 0.6389, 0.6389, 0.6389, 0.6389, 0.7185, 0.7194, 0.7456, 0.8089, 0.8406, 0.9717, 0.9582, 0.9582, 0.7646, 0.5575, 0.733, 0.7022, 0.5692, 0.6995, 0.7185, 0.6995, 0.6597, 0.7067, 0.7538, 0.8858, 0.8135, 0.8035, 0.685, 0.7158, 0.7094, 0.6289, 0.6289, 0.5979, 0.5154, 0.7189, 0.6169, 0.6876, 0.7501, 0.781, 0.9078, 0.8489, 0.8671, 0.8217, 0.8716, 0.9124, 0.9713, 0.9577, 0.9441, 0.9849, 0.9713, 1.0347, 0.9894, 1.0347, 1.0121]}
{"unit_id": "thread:1985187152570077257", "ticker_idx": 1, "ts": "2025-11-03T03:27:54+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron have also withheld contract pricing, signaling a tightening supply strategy among the top three suppliers", "tweet_type": "analysis", "timeline": "quarters", "summary": "DDR5 spot prices tripling and top-3 DRAM vendors restricting contract supply signal sustained pricing power for Samsung, SK Hynix, and Micron through 1H 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Samsung delays DDR5 contract pricing to mid-November as spot prices triple\n\nMemory shortages are intensifying as major cloud service providers (CSPs) continue expanding capital expenditures, with AI demand increasingly squeezing supply. Industry sources reveal that Samsung Electronics led the move by halting DDR5 DRAM contract pricing in October, prompting other manufacturers to follow suit. This disruption is expected to delay contract price announcements until mid-November, exacerbating supply constraints. Meanwhile, DRAM spot prices have surged sharply, with DDR5 prices more than doubling since late last month, driving procurement costs to soar.\n\nSamsung stalls contract pricing; stricter supply strategies from top three vendors\nIndustry insiders note that although a memory contract price increase in the fourth quarter of 2025 was widely anticipated, Samsung's unexpected refusal to provide contract quotes before the end of October has left downstream customers without available stock. This caused DDR5 spot prices to jump 25% within just one week.\n\nMarket reports indicate Samsung plans to resume contract price confirmations only by mid-November. SK Hynix and Micron have also withheld contract pricing, signaling a tightening supply strategy among the top three suppliers, who will now offer quotes exclusively to long-term partners.\n\nThis shift suggests non-quote periods may become standard, forcing urgent buyers to compete aggressively in the spot market.\n\nDRAM spot prices surge dramatically; DDR5 triples in value\nWhile fourth-quarter contract prices remain unconfirmed, projections suggest DDR5 prices will experience \"three consecutive jumps\" from the fourth quarter of 2025 through the first half of 2026, with quarterly increases ranging between 30% and 50%. Under this trajectory, DDR5 16Gb chip prices could reach historic highs of near US$30 by early 2026.\n\nOver the past month, DRAM spot prices have skyrocketed across the board. DDR5 16Gb prices soared to US$15.5, up from US$7.68 at September's end—a staggering 102% monthly increase—making it the fastest-rising product line.\n\nDDR4 prices also climbed steadily: DDR4 16Gb modules exceeded US$25, with Taiwanese manufacturers' scarce DDR4 16Gb (1Gx16) units reaching US$27 due to limited availability.\n\nThe DDR4 16Gb (2Gx8) variant saw nearly 90% monthly growth, hitting around US$25, second only to the same-spec DDR5 16Gb (2Gx8).\n\nMainstream DDR4 8Gb modules rose 60% over the month, while DDR3 4Gb prices increased 40%, marking the highest single-month rise since 2025 began.\n\nFlash wafer prices climb amid memory seller's market\nNot only have DDR chip prices surged, but flash wafer quotations have also risen significantly. Since September, 512Gb TLC wafers climbed steadily, reaching US$5 by late October. Prices for 1Tb TLC/QLC wafers hit US$6.5-7.2.\n\nMemory industry experts attribute these hikes to extremely tight circulating inventory. Despite restocking needs, suppliers hesitate to buy heavily due to high prices and limited supply, relying mostly on existing stock to maintain production.\n\nWith upstream manufacturers restricting fourth-quarter contract pricing mainly to tech giants and leading CSPs, DDR5 supplies are virtually unavailable to general customers, leaving many frustrated as the memory market fully shifts into a seller's market.\n\nMajor US CSPs recently raised their capex budgets: Amazon increased its capex from US$118 billion to US$125 billion, with further growth expected in 2026. Google, Meta, and Microsoft have also revised their spending plans upward, boosting investments in AI computing and high-end memory chips. This strong demand is causing structural capacity shifts in the overall supply chain.\n\nAccording to market data, DRAM contract prices in the fourth quarter of 2025 have already surged 1.7x year-over-year. To maximize profits, manufacturers are shifting capacity toward higher-margin DDR5 and high-bandwidth memory (HBM) products. However, DDR5 supply shortages have exceeded expectations. With Samsung suspending DDR5 contract pricing first, module makers worry about worsening shortages. Yet if they refrain from purchasing, prices risk climbing even higher, creating a dilemma for procurement costs over the next six months.\n\n---\n\nhttps://t.co/zuAvPuRgTT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09838715073877069, "ret_signed_m1d": -0.09838715073877069, "alpha_spy_m1d": -0.09651397900800696, "alpha_c_m1d": -0.08997409673737111, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.054838835866490854, "ret_signed_p1d": -0.054838835866490854, "alpha_spy_p1d": -0.04298529592404332, "alpha_c_p1d": -0.018400720224020173, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": -0.022580738743366635, "ret_signed_p1w": -0.022580738743366635, "alpha_spy_p1w": -0.019800309808329364, "alpha_c_p1w": -0.002613522001512103, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": -0.09935551328319481, "ret_signed_p1m": -0.09935551328319481, "alpha_spy_p1m": -0.09670677138394612, "alpha_c_p1m": -0.08113641365297852, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": 0.38970416322104473, "ret_signed_p3m": 0.38970416322104473, "alpha_spy_p3m": 0.37104466717996476, "alpha_c_p3m": 0.24570215906573378, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 1.102147490510704, "ret_signed_p6m": 1.102147490510704, "alpha_spy_p6m": 1.054729917297967, "alpha_c_p6m": 0.7562353678043552, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.5837, -0.578, -0.5869, -0.57, -0.5668, -0.5523, -0.5547, -0.5547, -0.5692, -0.5764, -0.5885, -0.6054, -0.5957, -0.5821, -0.5788, -0.5813, -0.5669, -0.5661, -0.5871, -0.5798, -0.5766, -0.5718, -0.5589, -0.5532, -0.5355, -0.5097, -0.5048, -0.4702, -0.4661, -0.4387, -0.4621, -0.4266, -0.4266, -0.4339, -0.4177, -0.4234, -0.425, -0.4573, -0.4371, -0.4395, -0.4194, -0.3621, -0.3621, -0.3621, -0.3621, -0.3621, -0.3621, -0.3097, -0.3306, -0.3363, -0.3185, -0.2702, -0.2492, -0.2169, -0.2274, -0.2234, -0.2282, -0.1774, -0.1371, -0.1597, -0.1, -0.0839, -0.0984, 0.0, -0.0548, -0.0661, -0.0435, -0.0645, -0.0226, -0.0016, -0.0048, -0.0129, -0.0968, -0.0226, -0.0806, -0.0935, -0.079, -0.1597, -0.1613, -0.1629, -0.1548, -0.122, -0.1445, -0.1316, -0.0994, -0.109, -0.1252, -0.122, -0.0687, -0.0864, -0.0525, -0.0881, -0.0784, -0.1058, -0.1445, -0.1107, -0.109, -0.1171, -0.0638, -0.0574, -0.0509, -0.0509, -0.0332, 0.033, 0.0508, 0.0508, 0.0508, 0.0927, 0.1234, 0.1718, 0.1976, 0.2202, 0.2009, 0.2089, 0.1912, 0.1976, 0.2089, 0.2202, 0.2331, 0.1992, 0.1944, 0.2186, 0.238, 0.1879, 0.2912, 0.3574, 0.3897, 0.4672, 0.3397, 0.464, 0.4527, 0.359, 0.3542, 0.4317, 0.4139, 0.3881, 0.4333, 0.4204, 0.4204, 0.4204, 0.4204, 0.443, 0.5317, 0.535, 0.6221, 0.6431, 0.7771, 0.7157, 0.7157, 0.5184, 0.3729, 0.5216, 0.4941, 0.3518, 0.5168, 0.5443, 0.5038, 0.4715, 0.575, 0.5685, 0.7076, 0.6381, 0.6284, 0.5087, 0.5944, 0.609, 0.5087, 0.5087, 0.4117, 0.3049, 0.4505, 0.3421, 0.4165, 0.4327, 0.4812, 0.6704, 0.6138, 0.6607, 0.6817, 0.7836, 0.837, 0.8677, 0.824, 0.8855, 0.9793, 0.9776, 0.9809, 0.976, 1.0892, 1.1021]}
{"unit_id": "thread:1985187152570077257", "ticker_idx": 2, "ts": "2025-11-03T03:27:54+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron have also withheld contract pricing, signaling a tightening supply strategy among the top three suppliers", "tweet_type": "analysis", "timeline": "quarters", "summary": "DDR5 spot prices tripling and top-3 DRAM vendors restricting contract supply signal sustained pricing power for Samsung, SK Hynix, and Micron through 1H 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Samsung delays DDR5 contract pricing to mid-November as spot prices triple\n\nMemory shortages are intensifying as major cloud service providers (CSPs) continue expanding capital expenditures, with AI demand increasingly squeezing supply. Industry sources reveal that Samsung Electronics led the move by halting DDR5 DRAM contract pricing in October, prompting other manufacturers to follow suit. This disruption is expected to delay contract price announcements until mid-November, exacerbating supply constraints. Meanwhile, DRAM spot prices have surged sharply, with DDR5 prices more than doubling since late last month, driving procurement costs to soar.\n\nSamsung stalls contract pricing; stricter supply strategies from top three vendors\nIndustry insiders note that although a memory contract price increase in the fourth quarter of 2025 was widely anticipated, Samsung's unexpected refusal to provide contract quotes before the end of October has left downstream customers without available stock. This caused DDR5 spot prices to jump 25% within just one week.\n\nMarket reports indicate Samsung plans to resume contract price confirmations only by mid-November. SK Hynix and Micron have also withheld contract pricing, signaling a tightening supply strategy among the top three suppliers, who will now offer quotes exclusively to long-term partners.\n\nThis shift suggests non-quote periods may become standard, forcing urgent buyers to compete aggressively in the spot market.\n\nDRAM spot prices surge dramatically; DDR5 triples in value\nWhile fourth-quarter contract prices remain unconfirmed, projections suggest DDR5 prices will experience \"three consecutive jumps\" from the fourth quarter of 2025 through the first half of 2026, with quarterly increases ranging between 30% and 50%. Under this trajectory, DDR5 16Gb chip prices could reach historic highs of near US$30 by early 2026.\n\nOver the past month, DRAM spot prices have skyrocketed across the board. DDR5 16Gb prices soared to US$15.5, up from US$7.68 at September's end—a staggering 102% monthly increase—making it the fastest-rising product line.\n\nDDR4 prices also climbed steadily: DDR4 16Gb modules exceeded US$25, with Taiwanese manufacturers' scarce DDR4 16Gb (1Gx16) units reaching US$27 due to limited availability.\n\nThe DDR4 16Gb (2Gx8) variant saw nearly 90% monthly growth, hitting around US$25, second only to the same-spec DDR5 16Gb (2Gx8).\n\nMainstream DDR4 8Gb modules rose 60% over the month, while DDR3 4Gb prices increased 40%, marking the highest single-month rise since 2025 began.\n\nFlash wafer prices climb amid memory seller's market\nNot only have DDR chip prices surged, but flash wafer quotations have also risen significantly. Since September, 512Gb TLC wafers climbed steadily, reaching US$5 by late October. Prices for 1Tb TLC/QLC wafers hit US$6.5-7.2.\n\nMemory industry experts attribute these hikes to extremely tight circulating inventory. Despite restocking needs, suppliers hesitate to buy heavily due to high prices and limited supply, relying mostly on existing stock to maintain production.\n\nWith upstream manufacturers restricting fourth-quarter contract pricing mainly to tech giants and leading CSPs, DDR5 supplies are virtually unavailable to general customers, leaving many frustrated as the memory market fully shifts into a seller's market.\n\nMajor US CSPs recently raised their capex budgets: Amazon increased its capex from US$118 billion to US$125 billion, with further growth expected in 2026. Google, Meta, and Microsoft have also revised their spending plans upward, boosting investments in AI computing and high-end memory chips. This strong demand is causing structural capacity shifts in the overall supply chain.\n\nAccording to market data, DRAM contract prices in the fourth quarter of 2025 have already surged 1.7x year-over-year. To maximize profits, manufacturers are shifting capacity toward higher-margin DDR5 and high-bandwidth memory (HBM) products. However, DDR5 supply shortages have exceeded expectations. With Samsung suspending DDR5 contract pricing first, module makers worry about worsening shortages. Yet if they refrain from purchasing, prices risk climbing even higher, creating a dilemma for procurement costs over the next six months.\n\n---\n\nhttps://t.co/zuAvPuRgTT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04657011762660468, "ret_signed_m1d": -0.04657011762660468, "alpha_spy_m1d": -0.044696945895840945, "alpha_c_m1d": -0.0381570636252051, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.008413054001399578, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07102688124746559, "ret_signed_p1d": -0.07102688124746559, "alpha_spy_p1d": -0.05917334130501806, "alpha_c_p1d": -0.03458876560499491, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.03643811564247068, "ret_p1w": 0.07925005947178643, "ret_signed_p1w": 0.07925005947178643, "alpha_spy_p1w": 0.0820304884068237, "alpha_c_p1w": 0.09921727621364096, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.019967216741854532, "ret_p1m": 0.020409038358131015, "ret_signed_p1m": 0.020409038358131015, "alpha_spy_p1m": 0.0230577802573797, "alpha_c_p1m": 0.0386281379883473, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.018219099630216284, "ret_p3m": 0.857546071764878, "ret_signed_p3m": 0.857546071764878, "alpha_spy_p3m": 0.838886575723798, "alpha_c_p3m": 0.713544067609567, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.14400200415531095, "ret_p6m": 1.1504288735653243, "ret_signed_p6m": 1.1504288735653243, "alpha_spy_p6m": 1.1030113003525872, "alpha_c_p6m": 0.8045167508589754, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.34591212270634886, "price_path": [-0.5368, -0.5236, -0.4938, -0.4732, -0.456, -0.4708, -0.4665, -0.4853, -0.4739, -0.4803, -0.5009, -0.507, -0.4989, -0.5043, -0.5039, -0.4986, -0.4805, -0.4932, -0.4932, -0.4955, -0.4945, -0.4711, -0.4406, -0.4402, -0.4241, -0.4039, -0.3589, -0.3305, -0.3282, -0.3237, -0.3187, -0.2809, -0.3071, -0.299, -0.2914, -0.3114, -0.3322, -0.3303, -0.3021, -0.2875, -0.2244, -0.2176, -0.1997, -0.1864, -0.2088, -0.1626, -0.1805, -0.2262, -0.1787, -0.203, -0.1822, -0.1371, -0.1377, -0.119, -0.1381, -0.1544, -0.1193, -0.0668, -0.0622, -0.0545, -0.0344, -0.0455, -0.0466, 0.0, -0.071, 0.0119, 0.0155, 0.0137, 0.0793, 0.0273, 0.0435, 0.0096, 0.0517, 0.0309, -0.0264, -0.0374, -0.142, -0.1164, -0.0459, -0.0433, -0.0189, -0.0189, 0.0076, 0.0245, 0.0204, -0.0023, -0.0343, 0.0107, 0.0521, 0.0755, 0.1236, 0.1012, 0.0274, 0.0119, -0.0093, -0.0391, 0.059, 0.133, 0.1785, 0.1771, 0.2215, 0.2215, 0.2134, 0.2547, 0.2473, 0.2166, 0.2166, 0.3445, 0.3305, 0.4639, 0.4473, 0.3939, 0.4709, 0.4743, 0.4413, 0.4209, 0.4349, 0.5462, 0.5462, 0.5558, 0.6586, 0.6947, 0.7035, 0.6585, 0.7486, 0.8554, 0.8575, 0.7684, 0.8661, 0.7879, 0.6172, 0.6321, 0.6824, 0.6347, 0.591, 0.7491, 0.7645, 0.7547, 0.7547, 0.7041, 0.7943, 0.7789, 0.8251, 0.7944, 0.7818, 0.8286, 0.7713, 0.7577, 0.759, 0.6184, 0.7083, 0.6924, 0.5784, 0.6595, 0.7182, 0.7847, 0.7278, 0.8164, 0.8832, 0.9679, 0.9681, 0.8937, 0.8026, 0.7235, 0.6859, 0.6287, 0.5151, 0.5226, 0.3722, 0.4406, 0.5686, 0.5617, 0.5617, 0.6109, 0.6101, 0.7344, 0.7974, 0.7935, 0.819, 0.9857, 0.9455, 0.9498, 0.9405, 0.9122, 0.9163, 1.0787, 1.0542, 1.1181, 1.2369, 1.1504]}
{"unit_id": "orig:1982953003650056295", "ticker_idx": 0, "ts": "2025-10-27T23:30:12+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we slightly raise our earnings estimates and target price following this deal", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Citi raising QCOM estimates and target price on HUMAIN contract; modest bullish revision with caveats on competitive positioning.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251028: QCOM Joins AI Race - CITI\n\n1. Today, Qualcomm announced two new AI-related products [AI200/AI250] and signed a chip supply contract with Saudi Arabia’s HUMAIN worth up to 200MW.\n\n(2) We estimate that this contract will contribute approximately $1 billion in additional revenue and a positive EPS impact of about $0.25 for Qualcomm.\n\n(3) For reference, Broadcom’s revenue per 1 GW is roughly $10 billion, whereas this deal is only 200 MW in scale and involves lower-priced chips.\n\n(4) In terms of competitiveness, Qualcomm remains several years behind NVIDIA/AMD/Broadcom and, since it does not use HBM, is disadvantaged in high-performance compute.\n\n(5) The current market competition prioritizes maximum compute performance rather than cost efficiency.\n\n(6) Nonetheless, we slightly raise our earnings estimates and target price following this deal:\n ■ FY26(F) Revenue: $43 bn → $44 bn / EPS: $9.03 → $9.24\n ■ FY27(F) Revenue: $48 bn → $49 bn / EPS: $10.75 → $10.97\n\n⸻\n\n[Implication]\n\n(1) AI200 is a rack-level AI inference-only solution designed to drastically improve TCO, while AI250 adopts a near-memory computing architecture delivering over 10× memory bandwidth with lower power consumption.\n ■ AI200 to be commercialized in 2026, AI250 in 2027\n ■ Equipped with Direct Liquid Cooling system\n ■ Supports PCIe (scale-up) and Ethernet (scale-out) configurations\n ■ Includes Confidential Computing for secure AI workloads\n ■ Rack-level power draw: 160 kW\n\n(2) The AI accelerator market continues to multiply, with multiple vendors driving diverse product innovations.\n\n(3) Although the HUMAIN contract is not large in scale, securing a meaningful customer for a rack-level solution at this stage is impressive.\n\n(4) Moreover, Qualcomm is the first to commercialize an architecture that reduces data-movement bottlenecks by placing compute units close to memory.\n\n(5) This design minimizes data movement by positioning compute cores directly adjacent to memory; however, it is limited to inference, not training.\n\n(6) There is no need for premature conclusions regarding HBM; it is worth tracking whether Qualcomm secures additional meaningful contracts going forward.\n\n(7) As the overall AI market expands, diverse and fragmented demand will emerge, giving rise to differentiated products targeting the low-cost + high-efficiency inference segment.\n\n(8) The memory-centric computing paradigm will persist.\n\n$QCOM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0998508043415024, "ret_signed_m1d": -0.0998508043415024, "alpha_spy_m1d": -0.08819065539773896, "alpha_c_m1d": -0.075402364106114, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.035432673577811546, "ret_signed_p1d": -0.035432673577811546, "alpha_spy_p1d": -0.03808867253440262, "alpha_c_p1d": -0.044239667463128796, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": -0.0370843446045509, "ret_signed_p1w": -0.0370843446045509, "alpha_spy_p1w": -0.034311625019277314, "alpha_c_p1w": -0.05419832931342128, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": -0.12990195655125691, "ret_signed_p1m": -0.12990195655125691, "alpha_spy_p1m": -0.1149874250748476, "alpha_c_p1m": -0.07453165661405714, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": -0.15491100114226253, "ret_signed_p3m": -0.15491100114226253, "alpha_spy_p3m": -0.16333996835370912, "alpha_c_p3m": -0.2775242710993502, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": -0.2693549518590901, "ret_signed_p6m": -0.2693549518590901, "alpha_spy_p6m": -0.3026995183212313, "alpha_c_p6m": -0.564309215909695, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.1573, -0.2225, -0.2149, -0.2185, -0.2227, -0.2273, -0.227, -0.2182, -0.216, -0.1855, -0.1704, -0.1624, -0.1637, -0.1581, -0.1722, -0.1765, -0.1834, -0.1629, -0.1713, -0.1567, -0.1535, -0.1481, -0.1484, -0.1484, -0.1588, -0.1667, -0.149, -0.1483, -0.1462, -0.1546, -0.1531, -0.1394, -0.1377, -0.141, -0.1254, -0.1195, -0.1042, -0.111, -0.0957, -0.0967, -0.0753, -0.0959, -0.0985, -0.1192, -0.1136, -0.1129, -0.1003, -0.0986, -0.1016, -0.1184, -0.1061, -0.1173, -0.1816, -0.138, -0.1382, -0.1317, -0.1257, -0.1291, -0.11, -0.1004, -0.0981, -0.094, -0.0999, 0.0, -0.0354, -0.048, -0.0555, -0.0361, -0.0371, -0.0791, -0.0424, -0.0772, -0.0895, -0.0858, -0.073, -0.0587, -0.0702, -0.073, -0.1115, -0.1205, -0.1149, -0.1497, -0.1299, -0.1205, -0.1299, -0.1201, -0.1201, -0.1044, -0.1046, -0.0905, -0.0672, -0.0663, -0.0638, -0.0611, -0.0574, -0.0242, -0.0292, -0.0452, -0.04, -0.0568, -0.077, -0.0671, -0.0615, -0.067, -0.0641, -0.064, -0.064, -0.0638, -0.0712, -0.07, -0.084, -0.084, -0.0736, -0.0558, -0.0229, -0.035, -0.026, -0.0479, -0.0935, -0.1148, -0.1188, -0.1357, -0.1462, -0.1462, -0.1749, -0.1626, -0.1549, -0.1655, -0.1725, -0.1804, -0.1822, -0.1848, -0.1882, -0.1827, -0.2118, -0.2026, -0.2701, -0.2645, -0.256, -0.2498, -0.2447, -0.2584, -0.2465, -0.2465, -0.2362, -0.2329, -0.2434, -0.2348, -0.248, -0.2246, -0.2191, -0.2203, -0.2376, -0.2447, -0.2603, -0.2529, -0.2616, -0.2687, -0.2556, -0.2713, -0.2771, -0.2931, -0.3003, -0.3026, -0.2908, -0.2968, -0.2924, -0.2999, -0.3082, -0.3065, -0.2974, -0.2964, -0.3149, -0.3151, -0.3059, -0.314, -0.3166, -0.3166, -0.3223, -0.3313, -0.3127, -0.3114, -0.3098, -0.2926, -0.284, -0.2829, -0.2752, -0.2659, -0.2588, -0.2694]}
{"unit_id": "orig:2000552356380070073", "ticker_idx": 0, "ts": "2025-12-15T13:03:44+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AVGO is confident it can deliver AI revenues above current Street expectations for F2026, indicating it will beat any estimates seen on the buy-side or sell-side", "tweet_type": "analysis", "timeline": "quarters", "summary": "AVGO guiding AI revenues above all Street estimates for F2026; $73B backlog ships within 12 months.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "- The $73B AI backlog is expected to be shipped within 12 months.\n\n- Anthropic is currently the only rack customer, while OpenAI is expected to be a traditional ASIC customer.\n\nUltimately, AVGO is confident it can deliver AI revenues above current Street expectations for F2026, indicating it will beat any estimates seen on the buy-side or sell-side.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05920950302357042, "ret_signed_m1d": 0.05920950302357042, "alpha_spy_m1d": 0.057696419176543, "alpha_c_m1d": 0.05575242905981437, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004384797073856417, "ret_signed_p1d": 0.004384797073856417, "alpha_spy_p1d": 0.007117094669135127, "alpha_c_p1d": 0.0071051403228813426, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.006748921248308726, "ret_signed_p1w": 0.006748921248308726, "alpha_spy_p1w": -0.0022468040643937037, "alpha_c_p1w": -0.015691568423885638, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.04555043950311832, "ret_signed_p1m": 0.04555043950311832, "alpha_spy_p1m": 0.023382949597407388, "alpha_c_p1m": -0.06736618570636166, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": -0.009408617299791255, "ret_signed_p3m": -0.009408617299791255, "alpha_spy_p3m": 0.009250555964900875, "alpha_c_p3m": -0.11266082408716904, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.017, 0.0146, 0.0134, -0.003, -0.0026, -0.0015, -0.0109, -0.0155, -0.035, -0.0291, -0.0189, -0.0048, -0.0042, -0.0127, -0.01, 0.0167, 0.0153, -0.0447, 0.0497, 0.0127, 0.0339, 0.0422, 0.028, 0.0278, 0.0084, 0.0014, 0.0132, 0.0421, 0.0654, 0.0976, 0.1359, 0.1079, 0.0878, 0.0669, 0.0357, 0.0564, 0.0464, 0.0283, 0.0547, 0.0358, 0.0453, 0.0005, 0.0078, 0.0084, 0.002, 0.043, 0.0206, 0.0011, 0.1123, 0.1331, 0.17, 0.17, 0.1858, 0.1362, 0.1229, 0.1201, 0.1213, 0.1484, 0.1804, 0.1956, 0.2153, 0.1959, 0.0592, 0.0, 0.0044, -0.0406, -0.0292, 0.0016, 0.0067, 0.03, 0.0326, 0.0326, 0.0382, 0.0302, 0.0315, 0.0205, 0.0205, 0.0249, 0.0126, 0.0136, 0.0128, -0.0197, 0.0171, 0.0385, 0.0456, 0.0021, 0.0114, 0.037, 0.037, -0.0193, -0.0305, -0.0403, -0.0563, -0.0422, -0.0188, -0.0175, -0.0249, -0.0232, -0.0237, -0.0555, -0.0917, -0.0845, -0.0184, 0.0141, 0.0038, 0.0106, -0.0236, -0.0413, -0.0413, -0.0195, -0.0167, -0.0152, -0.0192, -0.026, -0.0403, -0.0202, -0.0515, -0.0578, -0.06, -0.0747, -0.0638, -0.0188, -0.0256, 0.0194, 0.0101, 0.0071, -0.0094, -0.0501, -0.042, -0.0526, -0.0685, -0.057, -0.0845, -0.0471, -0.0596, -0.058, -0.0858, -0.1116, -0.1331, -0.0855, -0.0738, -0.0706, -0.0706, -0.071, -0.0132, 0.036, 0.0486, 0.0978, 0.122, 0.1251, 0.1722, 0.1773, 0.2012, 0.1808, 0.1883, 0.2488, 0.2408, 0.2491, 0.2356, 0.1814, 0.198, 0.2334, 0.2447, 0.2306, 0.2627, 0.257, 0.219, 0.2705, 0.2659, 0.2389, 0.2315, 0.2994, 0.2563, 0.243, 0.2146, 0.2343, 0.2249, 0.2236, 0.2236, 0.2469, 0.2464, 0.2604, 0.32, null, null, null, null, null, null, null]}
{"unit_id": "orig:1992005994294194187", "ticker_idx": 0, "ts": "2025-11-21T23:03:33+00:00", "symbol": "AMKR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Expected beneficiary companies • Amkor, which was packaging the H20", "tweet_type": "analysis", "timeline": "unspecified", "summary": "H200 export-to-China scenario benefits Amkor (packaging), Samsung (legacy HBM), and TSMC if White House allows H200 exports.", "resolved_tickers": ["AMKR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts on Reuters’ report that the White House is considering allowing exports of the H200 to China:\n\n1. Has the idea that the H20 is just too weak finally spread within the White House?\n\nThe H20 was from the start a low-performance, neutered chip. As Rutnik said, it wasn’t a first-tier or second-tier product, but more like a fourth-tier chip.\n\nBecause its performance wasn’t that good, China could confidently block imports of the H20. With that level of performance, they can come up with something comparable using Chinese chips.\n\nIn particular, while the U.S. political establishment was arguing back and forth over exports of the H20 to China, its relative performance was only getting worse (in other words, AI chips with better performance than the H20 kept coming out).\n\nBut the H200 is a different story. It is more powerful than any existing Chinese chip and has a level of performance China would clearly covet.\n\nEven if there is opposition within U.S. political circles to exporting it to China, the Chinese leadership will welcome imports of the H200.\n    \n2. Perceptions within the Trump administration\n\nTrump has said before that Blackwell would be monopolized by the United States.\n\nThere was also a media report saying that for the B30A, exports to China would be allowed around the time Rubin ramps up.\n\nIn light of this, here’s how I interpret this latest Reuters report about allowing exports of the H200.\n\nNow that Blackwell has begun to be deployed in the United States, members of the Trump administration seem to be thinking: since Blackwell is rolling out domestically, it should be fine to export Hopper to China as well.\n\nThe idea is that although the H200 is powerful, the most powerful chips are still owned by the United States, so the U.S. continues to hold the upper hand in computing power.\n\nExpected beneficiary companies\n    \n• Amkor, which was packaging the H20\n    \n• Samsung Electronics, which is strong in legacy HBM\n    \n• TSMC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05803424643006194, "ret_signed_m1d": -0.05803424643006194, "alpha_spy_m1d": -0.04817127075335792, "alpha_c_m1d": -0.054875935421986544, "bench_spy_m1d": -0.009862975676704022, "bench_c_m1d": -0.0031583110080753984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03400940511638084, "ret_signed_p1d": 0.03400940511638084, "alpha_spy_p1d": 0.019290898054200367, "alpha_c_p1d": -0.005821245509578965, "bench_spy_p1d": 0.014718507062180475, "bench_c_p1d": 0.03983065062595981, "ret_p1w": 0.13541342091764563, "ret_signed_p1w": 0.13541342091764563, "alpha_spy_p1w": 0.09845005837996434, "alpha_c_p1w": 0.05523069375405898, "bench_spy_p1w": 0.03696336253768129, "bench_c_p1w": 0.08018272716358665, "ret_p1m": 0.2753343787445923, "ret_signed_p1m": 0.2753343787445923, "alpha_spy_p1m": 0.23311535125729788, "alpha_c_p1m": 0.1689680922010135, "bench_spy_p1m": 0.04221902748729445, "bench_c_p1m": 0.10636628654357883, "ret_p3m": 0.46136627822737375, "ret_signed_p3m": 0.46136627822737375, "alpha_spy_p3m": 0.41692537313940914, "alpha_c_p3m": 0.1924438914652653, "bench_spy_p3m": 0.04444090508796461, "bench_c_p3m": 0.26892238676210845, "ret_p6m": 1.0687539639837995, "ret_signed_p6m": 1.0687539639837995, "alpha_spy_p6m": 0.9415581542093077, "alpha_c_p6m": 0.38887361691595435, "bench_spy_p6m": 0.12719580977449185, "bench_c_p6m": 0.6798803470678452, "price_path": [-0.2426, -0.2357, -0.2332, -0.2479, -0.2479, -0.2569, -0.2574, -0.2421, -0.2312, -0.2253, -0.224, -0.2265, -0.2081, -0.2056, -0.1988, -0.1563, -0.1694, -0.0902, -0.1167, -0.0786, -0.0771, -0.0839, -0.0902, -0.0942, -0.1101, -0.1139, -0.0805, -0.0752, -0.0833, -0.0462, -0.0718, -0.0449, -0.0443, -0.1183, -0.0496, -0.0668, -0.024, -0.0172, -0.0237, 0.0072, 0.0031, -0.0399, -0.0022, 0.0225, 0.0353, 0.0034, 0.02, -0.0041, 0.0072, 0.1807, 0.0998, 0.1404, 0.0967, 0.0852, 0.0986, 0.069, 0.0755, -0.0031, -0.0115, -0.0362, -0.0424, -0.0193, -0.058, 0.0, 0.034, 0.0546, 0.0895, 0.0895, 0.1354, 0.1775, 0.2683, 0.3735, 0.3501, 0.3516, 0.3979, 0.3948, 0.4648, 0.4707, 0.3854, 0.3729, 0.2594, 0.2153, 0.2287, 0.2628, 0.2753, 0.2775, 0.2744, 0.2744, 0.2691, 0.2534, 0.2631, 0.2344, 0.2344, 0.3419, 0.5048, 0.6036, 0.6483, 0.5946, 0.6349, 0.6127, 0.6243, 0.5217, 0.5383, 0.5008, 0.5008, 0.5336, 0.6715, 0.6399, 0.5542, 0.5755, 0.5864, 0.5933, 0.5633, 0.5111, 0.507, 0.4457, 0.3716, 0.3841, 0.5433, 0.6421, 0.6715, 0.7562, 0.613, 0.4845, 0.4845, 0.4645, 0.4614, 0.5114, 0.4989, 0.4717, 0.5173, 0.5927, 0.517, 0.4951, 0.4948, 0.3941, 0.431, 0.3932, 0.301, 0.3513, 0.3738, 0.3732, 0.2928, 0.3467, 0.3981, 0.4516, 0.4682, 0.5077, 0.435, 0.4447, 0.5744, 0.5641, 0.414, 0.3924, 0.2922, 0.4106, 0.4563, 0.4629, 0.4629, 0.4733, 0.4917, 0.6421, 0.7273, 0.8156, 0.8933, 0.9209, 0.8708, 0.9704, 1.1104, 1.1753, 1.1994, 1.2633, 1.284, 1.4465, 1.3689, 1.2354, 1.2119, 1.185, 1.2269, 1.2229, 1.403, 1.419, 1.2639, 1.3999, 1.4024, 1.3009, 1.3372, 1.2583, 1.2038, 1.0688]}
{"unit_id": "orig:1992005994294194187", "ticker_idx": 1, "ts": "2025-11-21T23:03:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics, which is strong in legacy HBM", "tweet_type": "analysis", "timeline": "unspecified", "summary": "H200 export-to-China scenario benefits Amkor (packaging), Samsung (legacy HBM), and TSMC if White House allows H200 exports.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts on Reuters’ report that the White House is considering allowing exports of the H200 to China:\n\n1. Has the idea that the H20 is just too weak finally spread within the White House?\n\nThe H20 was from the start a low-performance, neutered chip. As Rutnik said, it wasn’t a first-tier or second-tier product, but more like a fourth-tier chip.\n\nBecause its performance wasn’t that good, China could confidently block imports of the H20. With that level of performance, they can come up with something comparable using Chinese chips.\n\nIn particular, while the U.S. political establishment was arguing back and forth over exports of the H20 to China, its relative performance was only getting worse (in other words, AI chips with better performance than the H20 kept coming out).\n\nBut the H200 is a different story. It is more powerful than any existing Chinese chip and has a level of performance China would clearly covet.\n\nEven if there is opposition within U.S. political circles to exporting it to China, the Chinese leadership will welcome imports of the H200.\n    \n2. Perceptions within the Trump administration\n\nTrump has said before that Blackwell would be monopolized by the United States.\n\nThere was also a media report saying that for the B30A, exports to China would be allowed around the time Rubin ramps up.\n\nIn light of this, here’s how I interpret this latest Reuters report about allowing exports of the H200.\n\nNow that Blackwell has begun to be deployed in the United States, members of the Trump administration seem to be thinking: since Blackwell is rolling out domestically, it should be fine to export Hopper to China as well.\n\nThe idea is that although the H200 is powerful, the most powerful chips are still owned by the United States, so the U.S. continues to hold the upper hand in computing power.\n\nExpected beneficiary companies\n    \n• Amkor, which was packaging the H20\n    \n• Samsung Electronics, which is strong in legacy HBM\n    \n• TSMC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06118147784101491, "ret_signed_m1d": 0.06118147784101491, "alpha_spy_m1d": 0.07104445351771893, "alpha_c_m1d": 0.06502485721903084, "bench_spy_m1d": -0.009862975676704022, "bench_c_m1d": -0.003843379378015932, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020042229719216653, "ret_signed_p1d": 0.020042229719216653, "alpha_spy_p1d": 0.005323722657036178, "alpha_c_p1d": -0.003786341581340169, "bench_spy_p1d": 0.014718507062180475, "bench_c_p1d": 0.023828571300556822, "ret_p1w": 0.06012660617529764, "ret_signed_p1w": 0.06012660617529764, "alpha_spy_p1w": 0.02316324363761635, "alpha_c_p1w": 0.01246935156149509, "bench_spy_p1w": 0.03696336253768129, "bench_c_p1w": 0.04765725461380255, "ret_p1m": 0.16561186415290985, "ret_signed_p1m": 0.16561186415290985, "alpha_spy_p1m": 0.1233928366656154, "alpha_c_p1m": 0.10133539891492771, "bench_spy_p1m": 0.04221902748729445, "bench_c_p1m": 0.06427646523798214, "ret_p3m": 0.9206839737427239, "ret_signed_p3m": 0.9206839737427239, "alpha_spy_p3m": 0.8762430686547593, "alpha_c_p3m": 0.8875674222997765, "bench_spy_p3m": 0.04444090508796461, "bench_c_p3m": 0.03311655144294745, "ret_p6m": 1.9847092568059015, "ret_signed_p6m": 1.9847092568059015, "alpha_spy_p6m": 1.8575134470314096, "alpha_c_p6m": 1.7047562797326585, "bench_spy_p6m": 0.12719580977449185, "bench_c_p6m": 0.27995297707324296, "price_path": [-0.2617, -0.2586, -0.2691, -0.268, -0.2901, -0.2743, -0.267, -0.2638, -0.2701, -0.2638, -0.2491, -0.2376, -0.2292, -0.2082, -0.1966, -0.1662, -0.1788, -0.1567, -0.1567, -0.1231, -0.1105, -0.1032, -0.0958, -0.1252, -0.1118, -0.115, -0.0928, -0.0533, -0.0533, -0.0533, -0.0533, -0.0533, -0.0533, -0.0042, -0.0158, -0.0338, 0.0021, 0.0306, 0.0327, 0.0348, 0.0285, 0.0401, 0.0179, 0.0422, 0.0759, 0.0496, 0.0601, 0.0981, 0.134, 0.1719, 0.1065, 0.0612, 0.0464, 0.0327, 0.0612, 0.0918, 0.0876, 0.0844, 0.0253, 0.0612, 0.0316, 0.0179, 0.0612, 0.0, 0.02, 0.0475, 0.0844, 0.0918, 0.0601, 0.0633, 0.0907, 0.1023, 0.1086, 0.1435, 0.1551, 0.1435, 0.1392, 0.1319, 0.1487, 0.1055, 0.0844, 0.1382, 0.135, 0.1213, 0.1656, 0.1762, 0.1719, 0.1719, 0.2342, 0.2667, 0.2709, 0.2709, 0.2709, 0.3621, 0.4638, 0.4723, 0.4946, 0.4713, 0.4734, 0.4713, 0.4585, 0.4872, 0.5253, 0.5783, 0.5826, 0.5391, 0.5847, 0.6143, 0.6122, 0.6122, 0.6907, 0.7214, 0.7034, 0.7013, 0.5942, 0.7755, 0.7924, 0.6885, 0.6811, 0.7638, 0.7574, 0.7786, 0.8931, 0.9207, 0.9207, 0.9207, 0.9207, 1.014, 1.015, 1.0458, 1.12, 1.1571, 1.3108, 1.2949, 1.2949, 1.068, 0.8253, 1.0309, 0.9949, 0.8391, 0.9917, 1.014, 0.9917, 0.9451, 1.0002, 1.0553, 1.2101, 1.1253, 1.1136, 0.9747, 1.0108, 1.0034, 0.909, 0.909, 0.8726, 0.776, 1.0144, 0.8949, 0.9778, 1.0511, 1.0872, 1.2359, 1.1668, 1.1881, 1.135, 1.1934, 1.2412, 1.3102, 1.2943, 1.2784, 1.3262, 1.3102, 1.3846, 1.3315, 1.3846, 1.358, 1.4005, 1.3421, 1.3421, 1.4696, 1.4696, 1.8254, 1.8838, 1.8519, 2.0325, 1.9635, 2.0166, 2.144, 1.8732, 1.9847]}
{"unit_id": "orig:1992005994294194187", "ticker_idx": 2, "ts": "2025-11-21T23:03:33+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC", "tweet_type": "analysis", "timeline": "unspecified", "summary": "H200 export-to-China scenario benefits Amkor (packaging), Samsung (legacy HBM), and TSMC if White House allows H200 exports.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts on Reuters’ report that the White House is considering allowing exports of the H200 to China:\n\n1. Has the idea that the H20 is just too weak finally spread within the White House?\n\nThe H20 was from the start a low-performance, neutered chip. As Rutnik said, it wasn’t a first-tier or second-tier product, but more like a fourth-tier chip.\n\nBecause its performance wasn’t that good, China could confidently block imports of the H20. With that level of performance, they can come up with something comparable using Chinese chips.\n\nIn particular, while the U.S. political establishment was arguing back and forth over exports of the H20 to China, its relative performance was only getting worse (in other words, AI chips with better performance than the H20 kept coming out).\n\nBut the H200 is a different story. It is more powerful than any existing Chinese chip and has a level of performance China would clearly covet.\n\nEven if there is opposition within U.S. political circles to exporting it to China, the Chinese leadership will welcome imports of the H200.\n    \n2. Perceptions within the Trump administration\n\nTrump has said before that Blackwell would be monopolized by the United States.\n\nThere was also a media report saying that for the B30A, exports to China would be allowed around the time Rubin ramps up.\n\nIn light of this, here’s how I interpret this latest Reuters report about allowing exports of the H200.\n\nNow that Blackwell has begun to be deployed in the United States, members of the Trump administration seem to be thinking: since Blackwell is rolling out domestically, it should be fine to export Hopper to China as well.\n\nThe idea is that although the H200 is powerful, the most powerful chips are still owned by the United States, so the U.S. continues to hold the upper hand in computing power.\n\nExpected beneficiary companies\n    \n• Amkor, which was packaging the H20\n    \n• Samsung Electronics, which is strong in legacy HBM\n    \n• TSMC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05054149190361623, "ret_signed_m1d": 0.05054149190361623, "alpha_spy_m1d": 0.060404467580320254, "alpha_c_m1d": 0.05369980291169163, "bench_spy_m1d": -0.009862975676704022, "bench_c_m1d": -0.0031583110080753984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007220276500312428, "ret_signed_p1d": -0.007220276500312428, "alpha_spy_p1d": -0.021938783562492903, "alpha_c_p1d": -0.047050927126272235, "bench_spy_p1d": 0.014718507062180475, "bench_c_p1d": 0.03983065062595981, "ret_p1w": 0.03971121023271884, "ret_signed_p1w": 0.03971121023271884, "alpha_spy_p1w": 0.002747847695037553, "alpha_c_p1w": -0.04047151693086781, "bench_spy_p1w": 0.03696336253768129, "bench_c_p1w": 0.08018272716358665, "ret_p1m": 0.06128748984602872, "ret_signed_p1m": 0.06128748984602872, "alpha_spy_p1m": 0.019068462358734273, "alpha_c_p1m": -0.045078796697550105, "bench_spy_p1m": 0.04221902748729445, "bench_c_p1m": 0.10636628654357883, "ret_p3m": 0.38728027231983964, "ret_signed_p3m": 0.38728027231983964, "alpha_spy_p3m": 0.342839367231875, "alpha_c_p3m": 0.11835788555773119, "bench_spy_p3m": 0.04444090508796461, "bench_c_p3m": 0.26892238676210845, "ret_p6m": 0.6280138530510533, "ret_signed_p6m": 0.6280138530510533, "alpha_spy_p6m": 0.5008180432765614, "alpha_c_p6m": -0.051866494016791886, "bench_spy_p6m": 0.12719580977449185, "bench_c_p6m": 0.6798803470678452, "price_path": [-0.155, -0.1442, -0.1658, -0.1658, -0.1622, -0.1658, -0.1658, -0.1658, -0.1514, -0.1514, -0.137, -0.119, -0.1083, -0.0939, -0.0975, -0.0758, -0.0866, -0.0722, -0.0866, -0.065, -0.0325, -0.0325, -0.0469, -0.0614, -0.0614, -0.0578, -0.0433, -0.0144, 0.0108, 0.0108, 0.0361, 0.0217, 0.0397, 0.0397, 0.0217, 0.0289, 0.0578, 0.0722, 0.0469, 0.0686, 0.0686, 0.0542, 0.0469, 0.0469, 0.0686, 0.065, 0.0866, 0.0866, 0.083, 0.0903, 0.0866, 0.0542, 0.0578, 0.0542, 0.065, 0.0578, 0.065, 0.0542, 0.0325, 0.0433, 0.0144, 0.0072, 0.0505, 0.0, -0.0072, 0.0217, 0.0397, 0.0361, 0.0397, 0.0181, 0.0325, 0.0469, 0.0433, 0.0542, 0.0794, 0.0686, 0.0866, 0.0649, 0.0722, 0.0504, 0.0396, 0.0359, 0.0359, 0.0359, 0.0613, 0.0794, 0.083, 0.083, 0.0939, 0.1084, 0.1011, 0.1229, 0.1229, 0.1482, 0.2098, 0.2352, 0.2134, 0.2207, 0.217, 0.2243, 0.2388, 0.2388, 0.2243, 0.2605, 0.275, 0.2859, 0.2605, 0.275, 0.2822, 0.2714, 0.2895, 0.3185, 0.3076, 0.2859, 0.2786, 0.304, 0.2931, 0.2786, 0.2895, 0.3148, 0.3619, 0.3873, 0.3873, 0.3873, 0.3873, 0.3873, 0.3873, 0.3873, 0.3873, 0.3764, 0.4235, 0.4597, 0.4452, 0.4452, 0.4307, 0.4018, 0.3511, 0.3764, 0.3692, 0.3112, 0.3402, 0.4054, 0.3655, 0.3511, 0.3366, 0.3591, 0.3845, 0.3446, 0.3373, 0.3155, 0.3155, 0.3409, 0.3373, 0.3228, 0.2937, 0.2792, 0.3482, 0.3155, 0.3155, 0.3155, 0.3518, 0.4172, 0.4209, 0.4536, 0.4463, 0.4936, 0.5117, 0.5154, 0.4754, 0.4718, 0.4899, 0.4899, 0.5117, 0.588, 0.6462, 0.6098, 0.5844, 0.5517, 0.5517, 0.6535, 0.6353, 0.6353, 0.6789, 0.6644, 0.6244, 0.6389, 0.6135, 0.6498, 0.6462, 0.628]}
{"unit_id": "thread:1974815502531977338", "ticker_idx": 0, "ts": "2025-10-05T12:34:40+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "Citi's analyst emphasized that 'in an industry where unit growth typically averages around 50% from peak to peak, the current 11% shortfall in shipments gives us confidence that there is still substantial upside left in this cycle.'", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's bullish semi cycle call: 16% revenue growth to $731B in 2025, shipments 11% below peak, ample upside; NVDA named as key driver.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "Global semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "self_thread", "lang": "en", "tweet_text": "Citi: The Semiconductor Cycle Still Has Significant Upside Potential 🧵/1\n\nCiti analyst Christopher Danely expects global semiconductor sales to rise 16% in 2025 to reach an all-time high of $731 billion. However, he emphasized that this revenue growth is entirely driven by pricing, while shipment volumes remain well below previous peaks.\nThis indicates that inventory levels are low and the industry still has substantial room for further growth.\n\n---\n\nAccording to Citi’s data, the current semiconductor industry’s revenue growth has been primarily driven by the surge in logic chip prices.\n\nCiti noted that the average selling price (ASP) of logic computing chips (including AI accelerators) has risen 24% over the past three years, far exceeding the 2% growth rate of the previous decade. The share of logic computing chips in total semiconductor sales has also climbed from 27% in 2020 to 39% in 2025. 🧵/2\n\n---\n\nWithin this segment, logic computing revenue is expanding rapidly at a compound annual growth rate (CAGR) of 53%, rising from about $29.6 billion in 2022 to approximately $106.4 billion in 2025, with its share of total semiconductor sales jumping from 5% to 15%. 🧵/3\n\nThe average selling price of logic computing chips is expected to soar from $7.80–$8.50 during 2018–2022 to $26.40 in 2025, representing a 47% compound annual growth rate.\n\nCiti identified NVIDIA’s rapid data center expansion as the key driver behind this transformation. The company’s data center business share of logic chip sales has surged from less than 10% in 2021 to 66% in 2025, while its share of total semiconductor sales has also increased from less than 3% in 2021 to 24% in 2025.\n\n---\n\nShipments Still Below Peak Levels, Inventories Remain Low 🧵/4\n\nWhile semiconductor revenue has reached a record high, Citi’s analysis indicates that shipment volumes still have significant room for growth.\n\nCurrently, total semiconductor shipments are about 11% below their previous peak, and in the current upcycle have increased only 18%, far short of the historical average growth rate of 60%.\n\n---\n\nBy product category: 🧵/N\n\n- Microcontroller (MCU) shipments are 27% below their 2021 peak,\n- Logic computing chips are 15% lower,\n- Microprocessors are 12% lower.\n\nCiti interprets this gap between current and peak shipment levels as a sign that supply chain inventories remain lean, leaving ample room for further growth in the future.\n\nCiti’s analyst emphasized that “in an industry where unit growth typically averages around 50% from peak to peak, the current 11% shortfall in shipments gives us confidence that there is still substantial upside left in this cycle.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019511948544406743, "ret_signed_m1d": -0.019511948544406743, "alpha_spy_m1d": -0.015938534714057706, "alpha_c_m1d": -0.015938534714057706, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.003573413830349037, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01843438510550044, "ret_signed_p1d": -0.01843438510550044, "alpha_spy_p1d": -0.014726892854538542, "alpha_c_p1d": -0.014726892854538542, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.003707492250961897, "ret_p1w": -0.010571390338552145, "ret_signed_p1w": -0.010571390338552145, "alpha_spy_p1w": 0.0021890000992710323, "alpha_c_p1w": 0.0021890000992710323, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.012760390437823177, "ret_p1m": 0.02731672796216955, "ret_signed_p1m": 0.02731672796216955, "alpha_spy_p1m": 0.02191173161710891, "alpha_c_p1m": 0.02191173161710891, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.005404996345060642, "ret_p3m": 0.052043184912721197, "ret_signed_p3m": 0.052043184912721197, "alpha_spy_p3m": 0.033691715195707905, "alpha_c_p3m": 0.033691715195707905, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.01835146971701329, "ret_p6m": 0.12002150669347245, "ret_signed_p6m": 0.12002150669347245, "alpha_spy_p6m": 0.14617794906561676, "alpha_c_p6m": 0.14617794906561676, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": -0.026156442372144317, "price_path": [-0.1688, -0.1627, -0.1628, -0.1691, -0.1531, -0.1576, -0.1505, -0.1546, -0.154, -0.1689, -0.1653, -0.1619, -0.1628, -0.1513, -0.147, -0.1372, -0.159, -0.1731, -0.1549, -0.1639, -0.1653, -0.1522, -0.1452, -0.1456, -0.126, -0.1226, -0.1207, -0.139, -0.1355, -0.1529, -0.1585, -0.1629, -0.1449, -0.144, -0.1358, -0.1331, -0.1296, -0.1546, -0.1546, -0.1657, -0.1659, -0.1559, -0.1459, -0.1365, -0.1327, -0.1239, -0.1149, -0.1138, -0.1055, -0.1053, -0.111, -0.0769, -0.0806, -0.0618, -0.0631, -0.0649, -0.065, -0.0628, -0.0603, -0.0496, -0.0282, -0.015, -0.0195, 0.0, -0.0184, 0.0078, 0.0053, -0.0525, -0.0106, -0.0288, -0.0047, -0.0002, -0.0016, 0.0112, 0.0056, -0.0139, 0.0041, 0.0226, 0.0482, 0.0575, 0.0735, 0.0593, 0.0572, 0.0662, 0.0273, 0.0469, 0.0225, 0.0138, 0.0449, 0.0223, 0.0354, 0.0042, 0.0047, -0.009, -0.0294, -0.0115, -0.0532, -0.0502, -0.0124, -0.0098, 0.0126, 0.0126, 0.0259, 0.0279, 0.0467, 0.0611, 0.0531, 0.0613, 0.0733, 0.0746, 0.0895, 0.0801, 0.0313, 0.0277, 0.0249, -0.0121, 0.0113, 0.0374, 0.0508, 0.0609, 0.0638, 0.0638, 0.0688, 0.064, 0.0613, 0.052, 0.052, 0.0905, 0.103, 0.1323, 0.1248, 0.1071, 0.137, 0.1412, 0.1438, 0.1345, 0.158, 0.1697, 0.1697, 0.1404, 0.1742, 0.1768, 0.1688, 0.1651, 0.1897, 0.2171, 0.2197, 0.1786, 0.1918, 0.1617, 0.116, 0.1132, 0.1733, 0.1879, 0.1824, 0.2117, 0.1864, 0.1911, 0.1911, 0.1905, 0.2052, 0.1983, 0.2124, 0.2061, 0.2245, 0.2449, 0.2036, 0.1871, 0.1872, 0.1424, 0.1659, 0.1549, 0.1117, 0.1521, 0.1607, 0.1715, 0.1338, 0.1315, 0.1508, 0.1594, 0.15, 0.1537, 0.1239, 0.1433, 0.1527, 0.1657, 0.1125, 0.0933, 0.0591, 0.12]}
{"unit_id": "thread:1974815502531977338", "ticker_idx": 1, "ts": "2025-10-05T12:34:40+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi identified NVIDIA's rapid data center expansion as the key driver behind this transformation", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's bullish semi cycle call: 16% revenue growth to $731B in 2025, shipments 11% below peak, ample upside; NVDA named as key driver.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Citi: The Semiconductor Cycle Still Has Significant Upside Potential 🧵/1\n\nCiti analyst Christopher Danely expects global semiconductor sales to rise 16% in 2025 to reach an all-time high of $731 billion. However, he emphasized that this revenue growth is entirely driven by pricing, while shipment volumes remain well below previous peaks.\nThis indicates that inventory levels are low and the industry still has substantial room for further growth.\n\n---\n\nAccording to Citi’s data, the current semiconductor industry’s revenue growth has been primarily driven by the surge in logic chip prices.\n\nCiti noted that the average selling price (ASP) of logic computing chips (including AI accelerators) has risen 24% over the past three years, far exceeding the 2% growth rate of the previous decade. The share of logic computing chips in total semiconductor sales has also climbed from 27% in 2020 to 39% in 2025. 🧵/2\n\n---\n\nWithin this segment, logic computing revenue is expanding rapidly at a compound annual growth rate (CAGR) of 53%, rising from about $29.6 billion in 2022 to approximately $106.4 billion in 2025, with its share of total semiconductor sales jumping from 5% to 15%. 🧵/3\n\nThe average selling price of logic computing chips is expected to soar from $7.80–$8.50 during 2018–2022 to $26.40 in 2025, representing a 47% compound annual growth rate.\n\nCiti identified NVIDIA’s rapid data center expansion as the key driver behind this transformation. The company’s data center business share of logic chip sales has surged from less than 10% in 2021 to 66% in 2025, while its share of total semiconductor sales has also increased from less than 3% in 2021 to 24% in 2025.\n\n---\n\nShipments Still Below Peak Levels, Inventories Remain Low 🧵/4\n\nWhile semiconductor revenue has reached a record high, Citi’s analysis indicates that shipment volumes still have significant room for growth.\n\nCurrently, total semiconductor shipments are about 11% below their previous peak, and in the current upcycle have increased only 18%, far short of the historical average growth rate of 60%.\n\n---\n\nBy product category: 🧵/N\n\n- Microcontroller (MCU) shipments are 27% below their 2021 peak,\n- Logic computing chips are 15% lower,\n- Microprocessors are 12% lower.\n\nCiti interprets this gap between current and peak shipment levels as a sign that supply chain inventories remain lean, leaving ample room for further growth in the future.\n\nCiti’s analyst emphasized that “in an industry where unit growth typically averages around 50% from peak to peak, the current 11% shortfall in shipments gives us confidence that there is still substantial upside left in this cycle.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011210528125265684, "ret_signed_m1d": 0.011210528125265684, "alpha_spy_m1d": 0.014783941955614721, "alpha_c_m1d": 0.030722476669672427, "bench_spy_m1d": -0.003573413830349037, "bench_c_m1d": -0.019511948544406743, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002694803694029413, "ret_signed_p1d": -0.002694803694029413, "alpha_spy_p1d": 0.001012688556932484, "alpha_c_p1d": 0.015739581411471026, "bench_spy_p1d": -0.003707492250961897, "bench_c_p1d": -0.01843438510550044, "ret_p1w": 0.014983368445398382, "ret_signed_p1w": 0.014983368445398382, "alpha_spy_p1w": 0.02774375888322156, "alpha_c_p1w": 0.025554758783950526, "bench_spy_p1w": -0.012760390437823177, "bench_c_p1w": -0.010571390338552145, "ret_p1m": 0.07087422544133548, "ret_signed_p1m": 0.07087422544133548, "alpha_spy_p1m": 0.06546922909627484, "alpha_c_p1m": 0.04355749747916593, "bench_spy_p1m": 0.005404996345060642, "bench_c_p1m": 0.02731672796216955, "ret_p3m": 0.005230126727494788, "ret_signed_p3m": 0.005230126727494788, "alpha_spy_p3m": -0.013121342989518503, "alpha_c_p3m": -0.04681305818522641, "bench_spy_p3m": 0.01835146971701329, "bench_c_p3m": 0.052043184912721197, "ret_p6m": -0.05993775072555885, "ret_signed_p6m": -0.05993775072555885, "alpha_spy_p6m": -0.03378130835341453, "alpha_c_p6m": -0.1799592574190313, "bench_spy_p6m": -0.026156442372144317, "bench_c_p6m": 0.12002150669347245, "price_path": [-0.1222, -0.1156, -0.1112, -0.1158, -0.08, -0.0764, -0.0676, -0.0708, -0.0764, -0.0998, -0.0796, -0.0637, -0.0649, -0.0474, -0.0541, -0.0338, -0.0414, -0.0638, -0.0299, -0.0393, -0.033, -0.0258, -0.0154, -0.0188, -0.0129, -0.0213, -0.019, -0.0275, -0.0191, -0.0534, -0.0547, -0.057, -0.0407, -0.0309, -0.0204, -0.0213, -0.029, -0.0613, -0.0613, -0.0796, -0.0805, -0.0749, -0.0999, -0.0929, -0.0797, -0.0443, -0.0451, -0.0416, -0.042, -0.0575, -0.0822, -0.0501, -0.0478, -0.0104, -0.0383, -0.0462, -0.0423, -0.0396, -0.0199, 0.0056, 0.0092, 0.0181, 0.0112, 0.0, -0.0027, 0.0192, 0.0379, -0.0128, 0.015, -0.0297, -0.0308, -0.0201, -0.0125, -0.0156, -0.0236, -0.0283, -0.0182, 0.0039, 0.0321, 0.0835, 0.1159, 0.0935, 0.0914, 0.115, 0.0709, 0.0521, 0.0137, 0.0141, 0.0728, 0.0411, 0.0445, 0.0071, 0.025, 0.0057, -0.0225, 0.0053, -0.0264, -0.0359, -0.0161, -0.0416, -0.0285, -0.0285, -0.046, -0.0303, -0.022, -0.0321, -0.0116, -0.0168, 0.0001, -0.003, -0.0094, -0.0248, -0.0566, -0.0498, -0.0421, -0.0786, -0.0614, -0.0245, -0.0099, 0.0198, 0.0166, 0.0166, 0.027, 0.0145, 0.0108, 0.0052, 0.0052, 0.0179, 0.014, 0.0092, 0.0193, -0.0026, -0.0036, -0.0032, 0.0015, -0.0129, 0.0082, 0.0038, 0.0038, -0.0402, -0.0119, -0.0037, 0.0115, 0.0051, 0.0161, 0.0323, 0.0376, 0.0302, 0.0004, -0.028, -0.0611, -0.0736, -0.0006, 0.0243, 0.0162, 0.0244, 0.0076, -0.0147, -0.0147, -0.003, 0.0132, 0.0128, 0.0231, 0.0324, 0.0395, 0.0541, -0.0034, -0.045, -0.0164, -0.0295, -0.0134, -0.0118, -0.0416, -0.0155, -0.0041, 0.0028, -0.0128, -0.0284, -0.0124, -0.0193, -0.0276, -0.0375, -0.0691, -0.0533, -0.0556, -0.0369, -0.077, -0.097, -0.1097, -0.0599]}
{"unit_id": "reply:1981622965780480356", "ticker_idx": 0, "ts": "2025-10-24T07:25:06+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I've been telling people to buy Samsung, SK hynix, and Micron all together", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Buy all three memory players — Samsung, SK Hynix, Micron — as every company benefits in a memory supercycle.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@MMehri54 Why did you sell? It’s true that Micron is lagging behind in HBM4,\nbut that doesn’t mean Micron won’t make money.\n\nIn a memory supercycle, every company benefits.\n\nI’ve been telling people to buy Samsung, SK hynix, and Micron all together.", "reply_text": null, "parent_text": "@Jukanlosreve Hi, are you 100% convinced that Micron is behind in HBM4? I've sold 70% of my MU holding at open then prices surged 5% yesterday 😒", "quoted_text": null, "parent_author": "MMehri54", "ret_m1d": -0.02327930081041618, "ret_signed_m1d": -0.02327930081041618, "alpha_spy_m1d": -0.015172979195579694, "alpha_c_m1d": -0.007985350678181291, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.01529395013223489, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03238869974949088, "ret_signed_p1d": 0.03238869974949088, "alpha_spy_p1d": 0.020590987726341936, "alpha_c_p1d": 0.013926990339442558, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.01846170941004832, "ret_p1w": 0.08805670030939794, "ret_signed_p1w": 0.08805670030939794, "alpha_spy_p1w": 0.08095443543545922, "alpha_c_p1w": 0.06387248953071922, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.024184210778678716, "ret_p1m": -0.021255051863864893, "ret_signed_p1m": -0.021255051863864893, "alpha_spy_p1m": -0.008674704892381202, "alpha_c_p1m": 0.02598938324095368, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.04724443510481857, "ret_p3m": 0.5205136048989005, "ret_signed_p3m": 0.5205136048989005, "alpha_spy_p3m": 0.5054891891356628, "alpha_c_p3m": 0.5393287816242193, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": -0.01881517672531874, "ret_p6m": 1.1861219791729205, "ret_signed_p6m": 1.1861219791729205, "alpha_spy_p6m": 1.1336961325880412, "alpha_c_p6m": 1.1302810229953626, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.05584095617755791, "price_path": [-0.2886, -0.2684, -0.2805, -0.3057, -0.2977, -0.2957, -0.3067, -0.2896, -0.2765, -0.2846, -0.2836, -0.2755, -0.2785, -0.2785, -0.2946, -0.2946, -0.2896, -0.2886, -0.2805, -0.2795, -0.2916, -0.2886, -0.2987, -0.2977, -0.3188, -0.3037, -0.2967, -0.2936, -0.2997, -0.2936, -0.2795, -0.2684, -0.2604, -0.2402, -0.2291, -0.1999, -0.212, -0.1909, -0.1909, -0.1586, -0.1465, -0.1395, -0.1324, -0.1606, -0.1478, -0.1508, -0.1296, -0.0916, -0.0916, -0.0916, -0.0916, -0.0916, -0.0916, -0.0445, -0.0557, -0.0729, -0.0385, -0.0111, -0.0091, -0.0071, -0.0132, -0.002, -0.0233, 0.0, 0.0324, 0.0071, 0.0172, 0.0536, 0.0881, 0.1245, 0.0617, 0.0182, 0.004, -0.0091, 0.0182, 0.0476, 0.0435, 0.0405, -0.0162, 0.0182, -0.0101, -0.0233, 0.0182, -0.0405, -0.0213, 0.0051, 0.0405, 0.0476, 0.0172, 0.0202, 0.0466, 0.0577, 0.0638, 0.0972, 0.1083, 0.0972, 0.0931, 0.086, 0.1022, 0.0607, 0.0405, 0.0921, 0.0891, 0.0759, 0.1184, 0.1285, 0.1245, 0.1245, 0.1842, 0.2154, 0.2195, 0.2195, 0.2195, 0.3069, 0.4046, 0.4127, 0.4341, 0.4117, 0.4137, 0.4117, 0.3995, 0.4269, 0.4636, 0.5144, 0.5185, 0.4768, 0.5205, 0.549, 0.547, 0.547, 0.6222, 0.6517, 0.6344, 0.6324, 0.5297, 0.7036, 0.7199, 0.6202, 0.6131, 0.6924, 0.6863, 0.7066, 0.8165, 0.8429, 0.8429, 0.8429, 0.8429, 0.9324, 0.9334, 0.9629, 1.0341, 1.0697, 1.2172, 1.2019, 1.2019, 0.9843, 0.7514, 0.9487, 0.9141, 0.7646, 0.9111, 0.9324, 0.9111, 0.8663, 0.9192, 0.9721, 1.1206, 1.0392, 1.028, 0.8948, 0.9294, 0.9223, 0.8317, 0.8317, 0.7968, 0.7041, 0.9329, 0.8182, 0.8977, 0.968, 1.0027, 1.1454, 1.0791, 1.0995, 1.0485, 1.1046, 1.1505, 1.2167, 1.2014, 1.1861]}
{"unit_id": "reply:1981622965780480356", "ticker_idx": 1, "ts": "2025-10-24T07:25:06+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I've been telling people to buy Samsung, SK hynix, and Micron all together", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Buy all three memory players — Samsung, SK Hynix, Micron — as every company benefits in a memory supercycle.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@MMehri54 Why did you sell? It’s true that Micron is lagging behind in HBM4,\nbut that doesn’t mean Micron won’t make money.\n\nIn a memory supercycle, every company benefits.\n\nI’ve been telling people to buy Samsung, SK hynix, and Micron all together.", "reply_text": null, "parent_text": "@Jukanlosreve Hi, are you 100% convinced that Micron is behind in HBM4? I've sold 70% of my MU holding at open then prices surged 5% yesterday 😒", "quoted_text": null, "parent_author": "MMehri54", "ret_m1d": -0.06176462980292963, "ret_signed_m1d": -0.06176462980292963, "alpha_spy_m1d": -0.05365830818809314, "alpha_c_m1d": -0.04370904094929551, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.01805558885363412, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0490195698937812, "ret_signed_p1d": 0.0490195698937812, "alpha_spy_p1d": 0.03722185787063226, "alpha_c_p1d": 0.02395842372403134, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.02506114616974986, "ret_p1w": 0.0960784700447499, "ret_signed_p1w": 0.0960784700447499, "alpha_spy_p1w": 0.08897620517081117, "alpha_c_p1w": 0.06224592707460874, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.033832542970141155, "ret_p1m": 0.01960786482260124, "ret_signed_p1m": 0.01960786482260124, "alpha_spy_p1m": 0.03218821179408493, "alpha_c_p1m": 0.05383934541209778, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.034231480589496543, "ret_p3m": 0.45201962623592506, "ret_signed_p3m": 0.45201962623592506, "alpha_spy_p3m": 0.4369952104726873, "alpha_c_p3m": 0.30381491090834234, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": 0.14820471532758273, "ret_p6m": 1.292133586055805, "ret_signed_p6m": 1.292133586055805, "alpha_spy_p6m": 1.2397077394709257, "alpha_c_p6m": 0.9667260284720454, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.3254075575837596, "price_path": [-0.486, -0.4841, -0.4645, -0.4948, -0.4948, -0.4841, -0.4939, -0.487, -0.4978, -0.4772, -0.4733, -0.4557, -0.4586, -0.4586, -0.4762, -0.4851, -0.4997, -0.5203, -0.5086, -0.4919, -0.488, -0.4909, -0.4735, -0.4725, -0.498, -0.4892, -0.4853, -0.4794, -0.4637, -0.4569, -0.4353, -0.4039, -0.398, -0.3559, -0.351, -0.3176, -0.3461, -0.3029, -0.3029, -0.3118, -0.2922, -0.299, -0.301, -0.3402, -0.3157, -0.3186, -0.2941, -0.2245, -0.2245, -0.2245, -0.2245, -0.2245, -0.2245, -0.1608, -0.1863, -0.1931, -0.1716, -0.1127, -0.0873, -0.048, -0.0608, -0.0559, -0.0618, 0.0, 0.049, 0.0216, 0.0941, 0.1137, 0.0961, 0.2157, 0.149, 0.1353, 0.1627, 0.1373, 0.1882, 0.2137, 0.2098, 0.2, 0.098, 0.1882, 0.1176, 0.102, 0.1196, 0.0216, 0.0196, 0.0176, 0.0275, 0.0674, 0.04, 0.0557, 0.0949, 0.0831, 0.0635, 0.0674, 0.1322, 0.1106, 0.1518, 0.1086, 0.1204, 0.0871, 0.04, 0.0812, 0.0831, 0.0733, 0.1381, 0.1459, 0.1538, 0.1538, 0.1754, 0.2558, 0.2774, 0.2774, 0.2774, 0.3284, 0.3657, 0.4245, 0.4559, 0.4834, 0.4599, 0.4697, 0.4481, 0.4559, 0.4697, 0.4834, 0.4991, 0.4579, 0.452, 0.4815, 0.505, 0.4442, 0.5698, 0.6502, 0.6894, 0.7836, 0.6286, 0.7797, 0.766, 0.6522, 0.6463, 0.7405, 0.7189, 0.6875, 0.7424, 0.7267, 0.7267, 0.7267, 0.7267, 0.7542, 0.8621, 0.866, 0.972, 0.9975, 1.1604, 1.0857, 1.0857, 0.8459, 0.669, 0.8498, 0.8164, 0.6434, 0.8439, 0.8773, 0.8282, 0.7889, 0.9147, 0.9068, 1.0759, 0.9914, 0.9796, 0.8341, 0.9383, 0.956, 0.8341, 0.8341, 0.7162, 0.5864, 0.7633, 0.6316, 0.722, 0.7417, 0.8007, 1.0307, 0.9619, 1.0189, 1.0444, 1.1683, 1.2332, 1.2705, 1.2174, 1.2921]}
{"unit_id": "reply:1981622965780480356", "ticker_idx": 2, "ts": "2025-10-24T07:25:06+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I've been telling people to buy Samsung, SK hynix, and Micron all together", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Buy all three memory players — Samsung, SK Hynix, Micron — as every company benefits in a memory supercycle.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@MMehri54 Why did you sell? It’s true that Micron is lagging behind in HBM4,\nbut that doesn’t mean Micron won’t make money.\n\nIn a memory supercycle, every company benefits.\n\nI’ve been telling people to buy Samsung, SK hynix, and Micron all together.", "reply_text": null, "parent_text": "@Jukanlosreve Hi, are you 100% convinced that Micron is behind in HBM4? I've sold 70% of my MU holding at open then prices surged 5% yesterday 😒", "quoted_text": null, "parent_author": "MMehri54", "ret_m1d": -0.05620492830992663, "ret_signed_m1d": -0.05620492830992663, "alpha_spy_m1d": -0.04809860669509014, "alpha_c_m1d": -0.03814933945629251, "bench_spy_m1d": -0.008106321614836487, "bench_c_m1d": -0.01805558885363412, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004931015291307128, "ret_signed_p1d": 0.004931015291307128, "alpha_spy_p1d": -0.006866696731841815, "alpha_c_p1d": -0.02013013087844273, "bench_spy_p1d": 0.011797712023148943, "bench_c_p1d": 0.02506114616974986, "ret_p1w": 0.021687472641301886, "ret_signed_p1w": 0.021687472641301886, "alpha_spy_p1w": 0.014585207767363162, "alpha_c_p1w": -0.01214507032883927, "bench_spy_p1w": 0.007102264873938724, "bench_c_p1w": 0.033832542970141155, "ret_p1m": 0.02241799084275886, "ret_signed_p1m": 0.02241799084275886, "alpha_spy_p1m": 0.03499833781424255, "alpha_c_p1m": 0.056649471432255405, "bench_spy_p1m": -0.012580346971483691, "bench_c_p1m": -0.034231480589496543, "ret_p3m": 0.7773133413387958, "ret_signed_p3m": 0.7773133413387958, "alpha_spy_p3m": 0.762288925575558, "alpha_c_p3m": 0.6291086260112131, "bench_spy_p3m": 0.015024415763237764, "bench_c_p3m": 0.14820471532758273, "ret_p6m": 1.0490804758150851, "ret_signed_p6m": 1.0490804758150851, "alpha_spy_p6m": 0.9966546292302059, "alpha_c_p6m": 0.7236729182313255, "bench_spy_p6m": 0.052425846584879254, "bench_c_p6m": 0.3254075575837596, "price_path": [-0.4891, -0.4764, -0.502, -0.5214, -0.5083, -0.5024, -0.5036, -0.4895, -0.4575, -0.4355, -0.4171, -0.433, -0.4283, -0.4485, -0.4362, -0.4431, -0.4652, -0.4717, -0.463, -0.4688, -0.4684, -0.4627, -0.4433, -0.457, -0.457, -0.4594, -0.4583, -0.4332, -0.4006, -0.4002, -0.3829, -0.3612, -0.313, -0.2826, -0.2801, -0.2753, -0.27, -0.2294, -0.2575, -0.2488, -0.2407, -0.2621, -0.2844, -0.2824, -0.2521, -0.2365, -0.1689, -0.1616, -0.1424, -0.1281, -0.1522, -0.1026, -0.1219, -0.1709, -0.1199, -0.1459, -0.1236, -0.0753, -0.076, -0.0559, -0.0764, -0.0938, -0.0562, 0.0, 0.0049, 0.0132, 0.0347, 0.0228, 0.0217, 0.0716, -0.0045, 0.0844, 0.0882, 0.0863, 0.1565, 0.1009, 0.1182, 0.0819, 0.127, 0.1047, 0.0433, 0.0315, -0.0806, -0.0532, 0.0224, 0.0252, 0.0513, 0.0513, 0.0797, 0.0979, 0.0935, 0.0691, 0.0348, 0.0831, 0.1274, 0.1525, 0.204, 0.1801, 0.101, 0.0844, 0.0616, 0.0297, 0.1348, 0.2141, 0.2629, 0.2614, 0.3089, 0.3089, 0.3003, 0.3446, 0.3366, 0.3036, 0.3036, 0.4407, 0.4258, 0.5687, 0.5509, 0.4937, 0.5762, 0.5798, 0.5445, 0.5226, 0.5376, 0.6569, 0.6569, 0.6672, 0.7773, 0.816, 0.8255, 0.7772, 0.8738, 0.9882, 0.9905, 0.895, 0.9997, 0.9158, 0.733, 0.7489, 0.8028, 0.7517, 0.7049, 0.8743, 0.8909, 0.8803, 0.8803, 0.8261, 0.9227, 0.9063, 0.9557, 0.9228, 0.9093, 0.9595, 0.8981, 0.8836, 0.8849, 0.7342, 0.8306, 0.8136, 0.6914, 0.7783, 0.8413, 0.9124, 0.8515, 0.9464, 1.018, 1.1088, 1.109, 1.0293, 0.9317, 0.8469, 0.8066, 0.7452, 0.6236, 0.6317, 0.4705, 0.5438, 0.6809, 0.6736, 0.6736, 0.7262, 0.7254, 0.8586, 0.9261, 0.9219, 0.9492, 1.1279, 1.0848, 1.0893, 1.0795, 1.0491]}
{"unit_id": "reply:1982437728483729447", "ticker_idx": 0, "ts": "2025-10-26T13:22:41+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Please put a rocket under my Samsung stocks, Trump! 🚀", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Samsung; hoping US export restriction lift on Korean HBM to China catalyzes the stock.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@stevehou I really hope Trump lifts the restrictions on exporting Korean HBM to China.\nPlease put a rocket under my Samsung stocks, Trump! 🚀", "reply_text": null, "parent_text": "I think the US-China trade deal increasingly looks like a win-win. Everybody won. Congrats to all!", "quoted_text": null, "parent_author": "stevehou", "ret_m1d": -0.03137258259156661, "ret_signed_m1d": -0.03137258259156661, "alpha_spy_m1d": -0.019712433647803174, "alpha_c_m1d": -0.013245529566919823, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.01812705302464679, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024509789177127317, "ret_signed_p1d": -0.024509789177127317, "alpha_spy_p1d": -0.027165788133718394, "alpha_c_p1d": -0.03440944268004342, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.009899653502916106, "ret_p1w": 0.08921562026478003, "ret_signed_p1w": 0.08921562026478003, "alpha_spy_p1w": 0.09198833985005361, "alpha_c_p1w": 0.07948310004733283, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.009732520217447194, "ret_p1m": -0.026470609331187056, "ret_signed_p1m": -0.026470609331187056, "alpha_spy_p1m": -0.011556077854777747, "alpha_c_p1m": 0.0355360442725412, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.06200665360372826, "ret_p3m": 0.5003956115080337, "ret_signed_p3m": 0.5003956115080337, "alpha_spy_p3m": 0.4919666442965871, "alpha_c_p3m": 0.5298286735013822, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": -0.029433061993348497, "ret_p6m": 1.1619615543818709, "ret_signed_p6m": 1.1619615543818709, "alpha_spy_p6m": 1.1286169879197296, "alpha_c_p6m": 1.124387883452265, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.03757367092960595, "price_path": [-0.2914, -0.3031, -0.3275, -0.3197, -0.3177, -0.3285, -0.3119, -0.2992, -0.307, -0.306, -0.2982, -0.3012, -0.3012, -0.3168, -0.3168, -0.3119, -0.3109, -0.3031, -0.3021, -0.3138, -0.3109, -0.3207, -0.3197, -0.3402, -0.3256, -0.3187, -0.3158, -0.3217, -0.3158, -0.3021, -0.2914, -0.2836, -0.2641, -0.2533, -0.225, -0.2367, -0.2162, -0.2162, -0.185, -0.1733, -0.1665, -0.1596, -0.187, -0.1745, -0.1775, -0.1569, -0.1201, -0.1201, -0.1201, -0.1201, -0.1201, -0.1201, -0.0745, -0.0853, -0.102, -0.0686, -0.0422, -0.0402, -0.0382, -0.0441, -0.0333, -0.0539, -0.0314, 0.0, -0.0245, -0.0147, 0.0206, 0.0539, 0.0892, 0.0284, -0.0137, -0.0275, -0.0402, -0.0137, 0.0147, 0.0108, 0.0078, -0.0471, -0.0137, -0.0412, -0.0539, -0.0137, -0.0706, -0.052, -0.0265, 0.0078, 0.0147, -0.0147, -0.0118, 0.0137, 0.0245, 0.0304, 0.0627, 0.0735, 0.0627, 0.0588, 0.052, 0.0676, 0.0275, 0.0078, 0.0578, 0.0549, 0.0422, 0.0833, 0.0931, 0.0892, 0.0892, 0.1471, 0.1773, 0.1812, 0.1812, 0.1812, 0.2659, 0.3605, 0.3684, 0.3891, 0.3674, 0.3694, 0.3674, 0.3556, 0.3822, 0.4176, 0.4669, 0.4708, 0.4304, 0.4728, 0.5004, 0.4984, 0.4984, 0.5713, 0.5999, 0.5831, 0.5812, 0.4817, 0.6501, 0.6659, 0.5694, 0.5625, 0.6393, 0.6334, 0.6531, 0.7595, 0.7851, 0.7851, 0.7851, 0.7851, 0.8718, 0.8728, 0.9014, 0.9703, 1.0048, 1.1476, 1.1329, 1.1329, 0.922, 0.6964, 0.8876, 0.8541, 0.7092, 0.8511, 0.8718, 0.8511, 0.8078, 0.859, 0.9102, 1.0541, 0.9752, 0.9644, 0.8353, 0.8688, 0.8619, 0.7743, 0.7743, 0.7404, 0.6506, 0.8722, 0.7612, 0.8382, 0.9063, 0.9398, 1.078, 1.0139, 1.0336, 0.9843, 1.0386, 1.083, 1.1472, 1.1323, 1.1175, 1.162]}
{"unit_id": "thread:1981132250608763103", "ticker_idx": 0, "ts": "2025-10-22T22:55:10+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The combined operating profit of Samsung Electronics and SK Hynix next year will climb 64 percent year-on-year to ₩128 trillion. With demand from AI data centers and general-server replacement cycles occurring simultaneously, a prolonged supply shortage and sustained price increases are unavoidable", "tweet_type": "news_reaction", "timeline": "years", "summary": "Relays bullish analyst/bank forecasts: DRAM shortage inevitable over 1-2 years, Samsung+Hynix combined OP up 64% YoY; memory market entering once-in-a-decade boom.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Prolonged DRAM Shortage Forecast… Surge in ‘Pre-Order’ Requests to Samsung and SK Hynix\n\nDemand for memory semiconductors required for artificial-intelligence (AI) infrastructure is soaring, and as data generated by AI agents such as OpenAI and Gemini explodes, demand is surging not only for high-bandwidth memory (HBM) but also for commodity DRAM. Consequently, some large overseas electronics and server companies are rushing to hoard memory inventory—requesting long-term supply contracts with Samsung Electronics and SK Hynix.\n\nAccording to industry sources on the 23rd, amid growing concerns about a DRAM shortage, a number of electronics, IT, and data-center companies in the U.S. and China are signing medium- to long-term DRAM supply contracts with Samsung Electronics and SK Hynix, typically spanning two to three years. Ordinarily, electronics manufacturers and server vendors sign DRAM supply agreements lasting only a quarter to a year in order to maintain inventory flexibility, but with forecasts pointing to a prolonged shortage of commodity DRAM, buyers are now moving to secure inventory well in excess of normal levels.\n\nSamsung Electronics and SK Hynix—together controlling more than 70 percent of the global commodity-DRAM market—are currently focusing their fabrication investments on HBM. On the supply side, this inevitably reduces production capacity for standard DRAM. An industry insider explained, “If we assume one DRAM wafer yields 100 conventional DRAM chips, HBM would yield roughly half that amount. Because HBM now occupies a greater share of DRAM production lines, the shipment volume of general-purpose DRAM is likely to remain constrained.”\n\nWith both companies prioritizing higher-margin HBM, analysts say expansion of commodity-DRAM output will remain difficult for the time being.  Ryu Hyung-keun, analyst at Daishin Securities, noted, “Concerns over DRAM supply shortages are spurring pre-emptive buying sentiment among customers, and some server clients are already negotiating supply volumes extending beyond 2027.”  Citigroup likewise assessed that “given these dynamics, DRAM supply growth over the next one to two years will be limited, making a severe shortage inevitable.”\n\nAccording to market-research firm DRAMeXchange, spot prices for 8 Gb DDR4 DRAM—the benchmark commodity product—stood at US $7.30 yesterday.  It is the first time DRAM spot prices have surpassed $7 since October 2018 ($7.042), when the previous memory-super-cycle ended.  Considering prices were just $2 in April, this represents a 265 percent surge in barely six months.\n\nForeign investment banks are also saying the memory-semiconductor market has entered a “once-in-a-decade boom.”  UBS recently raised its forecast for fourth-quarter DRAM contract-price growth from 5 percent quarter-on-quarter to 17 percent, citing the sharp increase in AI-server demand.  “The memory market is set to experience a ten-year-high boom by 2026,” UBS predicted.\n\nAs rising memory prices feed directly into stronger earnings, Samsung Electronics and SK Hynix are expected to post soaring profits.  Kim Dong-won, analyst at KB Securities, estimated, “The combined operating profit of Samsung Electronics and SK Hynix next year will climb 64 percent year-on-year to ₩128 trillion.  With demand from AI data centers and general-server replacement cycles occurring simultaneously, a prolonged supply shortage and sustained price increases are unavoidable, while any near-term capacity expansion in memory remains virtually impossible.”\n\n---\n\nhttps://t.co/tCz0OycLYn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01115615198006159, "ret_signed_m1d": -0.01115615198006159, "alpha_spy_m1d": -0.016382136876684994, "alpha_c_m1d": -0.02113892750915758, "bench_spy_m1d": 0.0052259848966234035, "bench_c_m1d": 0.009982775529095989, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02129808656623111, "ret_signed_p1d": -0.02129808656623111, "alpha_spy_p1d": -0.027227977756730803, "alpha_c_p1d": -0.033907998553032836, "bench_spy_p1d": 0.005929891190499692, "bench_c_p1d": 0.012609911986801725, "ret_p1w": 0.01926981134694805, "ret_signed_p1w": 0.01926981134694805, "alpha_spy_p1w": -0.01006528582714128, "alpha_c_p1w": -0.046021534538200015, "bench_spy_p1w": 0.02933509717408933, "bench_c_p1w": 0.06529134588514807, "ret_p1m": 0.02028402874084012, "ret_signed_p1m": 0.02028402874084012, "alpha_spy_p1m": 0.04315013699478942, "alpha_c_p1m": 0.06701069860822062, "bench_spy_p1m": -0.022866108253949302, "bench_c_p1m": -0.0467266698673805, "ret_p3m": 0.5215595395555155, "ret_signed_p3m": 0.5215595395555155, "alpha_spy_p3m": 0.4827698723263427, "alpha_c_p3m": 0.4998714342162296, "bench_spy_p3m": 0.038789667229172764, "bench_c_p3m": 0.02168810533928589, "ret_p6m": 1.2211935372842175, "ret_signed_p6m": 1.2211935372842175, "alpha_spy_p6m": 1.1645071122797894, "alpha_c_p6m": 1.1532760196575136, "bench_spy_p6m": 0.05668642500442811, "bench_c_p6m": 0.06791751762670395, "price_path": [-0.3346, -0.2892, -0.2872, -0.267, -0.2791, -0.3043, -0.2962, -0.2942, -0.3053, -0.2882, -0.275, -0.2831, -0.2821, -0.274, -0.2771, -0.2771, -0.2932, -0.2932, -0.2882, -0.2872, -0.2791, -0.2781, -0.2902, -0.2872, -0.2973, -0.2962, -0.3174, -0.3023, -0.2952, -0.2922, -0.2983, -0.2922, -0.2781, -0.267, -0.2589, -0.2387, -0.2276, -0.1983, -0.2104, -0.1892, -0.1892, -0.1569, -0.1448, -0.1377, -0.1307, -0.1589, -0.146, -0.1491, -0.1278, -0.0898, -0.0898, -0.0898, -0.0898, -0.0898, -0.0898, -0.0426, -0.0538, -0.071, -0.0365, -0.0091, -0.0071, -0.0051, -0.0112, 0.0, -0.0213, 0.002, 0.0345, 0.0091, 0.0193, 0.0558, 0.0903, 0.1268, 0.0639, 0.0203, 0.0061, -0.0071, 0.0203, 0.0497, 0.0456, 0.0426, -0.0142, 0.0203, -0.0081, -0.0213, 0.0203, -0.0385, -0.0193, 0.0071, 0.0426, 0.0497, 0.0193, 0.0223, 0.0487, 0.0598, 0.0659, 0.0994, 0.1105, 0.0994, 0.0953, 0.0882, 0.1045, 0.0629, 0.0426, 0.0943, 0.0913, 0.0781, 0.1207, 0.1308, 0.1268, 0.1268, 0.1866, 0.2179, 0.2219, 0.2219, 0.2219, 0.3096, 0.4074, 0.4156, 0.437, 0.4146, 0.4166, 0.4146, 0.4023, 0.4298, 0.4665, 0.5175, 0.5216, 0.4798, 0.5236, 0.5521, 0.5501, 0.5501, 0.6255, 0.6551, 0.6377, 0.6357, 0.5328, 0.707, 0.7233, 0.6235, 0.6163, 0.6958, 0.6897, 0.7101, 0.8202, 0.8467, 0.8467, 0.8467, 0.8467, 0.9363, 0.9374, 0.9669, 1.0383, 1.0739, 1.2217, 1.2064, 1.2064, 0.9883, 0.7549, 0.9527, 0.918, 0.7682, 0.9149, 0.9363, 0.9149, 0.8701, 0.9231, 0.9761, 1.1249, 1.0434, 1.0321, 0.8986, 0.9333, 0.9262, 0.8355, 0.8355, 0.8004, 0.7075, 0.9368, 0.8219, 0.9015, 0.972, 1.0067, 1.1497, 1.0833, 1.1038, 1.0527, 1.1089, 1.1548, 1.2212]}
{"unit_id": "thread:1981132250608763103", "ticker_idx": 1, "ts": "2025-10-22T22:55:10+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The combined operating profit of Samsung Electronics and SK Hynix next year will climb 64 percent year-on-year to ₩128 trillion. With demand from AI data centers and general-server replacement cycles occurring simultaneously, a prolonged supply shortage and sustained price increases are unavoidable", "tweet_type": "news_reaction", "timeline": "years", "summary": "Relays bullish analyst/bank forecasts: DRAM shortage inevitable over 1-2 years, Samsung+Hynix combined OP up 64% YoY; memory market entering once-in-a-decade boom.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Prolonged DRAM Shortage Forecast… Surge in ‘Pre-Order’ Requests to Samsung and SK Hynix\n\nDemand for memory semiconductors required for artificial-intelligence (AI) infrastructure is soaring, and as data generated by AI agents such as OpenAI and Gemini explodes, demand is surging not only for high-bandwidth memory (HBM) but also for commodity DRAM. Consequently, some large overseas electronics and server companies are rushing to hoard memory inventory—requesting long-term supply contracts with Samsung Electronics and SK Hynix.\n\nAccording to industry sources on the 23rd, amid growing concerns about a DRAM shortage, a number of electronics, IT, and data-center companies in the U.S. and China are signing medium- to long-term DRAM supply contracts with Samsung Electronics and SK Hynix, typically spanning two to three years. Ordinarily, electronics manufacturers and server vendors sign DRAM supply agreements lasting only a quarter to a year in order to maintain inventory flexibility, but with forecasts pointing to a prolonged shortage of commodity DRAM, buyers are now moving to secure inventory well in excess of normal levels.\n\nSamsung Electronics and SK Hynix—together controlling more than 70 percent of the global commodity-DRAM market—are currently focusing their fabrication investments on HBM. On the supply side, this inevitably reduces production capacity for standard DRAM. An industry insider explained, “If we assume one DRAM wafer yields 100 conventional DRAM chips, HBM would yield roughly half that amount. Because HBM now occupies a greater share of DRAM production lines, the shipment volume of general-purpose DRAM is likely to remain constrained.”\n\nWith both companies prioritizing higher-margin HBM, analysts say expansion of commodity-DRAM output will remain difficult for the time being.  Ryu Hyung-keun, analyst at Daishin Securities, noted, “Concerns over DRAM supply shortages are spurring pre-emptive buying sentiment among customers, and some server clients are already negotiating supply volumes extending beyond 2027.”  Citigroup likewise assessed that “given these dynamics, DRAM supply growth over the next one to two years will be limited, making a severe shortage inevitable.”\n\nAccording to market-research firm DRAMeXchange, spot prices for 8 Gb DDR4 DRAM—the benchmark commodity product—stood at US $7.30 yesterday.  It is the first time DRAM spot prices have surpassed $7 since October 2018 ($7.042), when the previous memory-super-cycle ended.  Considering prices were just $2 in April, this represents a 265 percent surge in barely six months.\n\nForeign investment banks are also saying the memory-semiconductor market has entered a “once-in-a-decade boom.”  UBS recently raised its forecast for fourth-quarter DRAM contract-price growth from 5 percent quarter-on-quarter to 17 percent, citing the sharp increase in AI-server demand.  “The memory market is set to experience a ten-year-high boom by 2026,” UBS predicted.\n\nAs rising memory prices feed directly into stronger earnings, Samsung Electronics and SK Hynix are expected to post soaring profits.  Kim Dong-won, analyst at KB Securities, estimated, “The combined operating profit of Samsung Electronics and SK Hynix next year will climb 64 percent year-on-year to ₩128 trillion.  With demand from AI data centers and general-server replacement cycles occurring simultaneously, a prolonged supply shortage and sustained price increases are unavoidable, while any near-term capacity expansion in memory remains virtually impossible.”\n\n---\n\nhttps://t.co/tCz0OycLYn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005192162329343719, "ret_signed_m1d": -0.005192162329343719, "alpha_spy_m1d": -0.010418147225967123, "alpha_c_m1d": -0.024979469391708364, "bench_spy_m1d": 0.0052259848966234035, "bench_c_m1d": 0.019787307062364645, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0062304906695249684, "ret_signed_p1d": -0.0062304906695249684, "alpha_spy_p1d": -0.01216038186002466, "alpha_c_p1d": -0.024541145170254852, "bench_spy_p1d": 0.005929891190499692, "bench_c_p1d": 0.018310654500729884, "ret_p1w": 0.15887842317265122, "ret_signed_p1w": 0.15887842317265122, "alpha_spy_p1w": 0.1295433259985619, "alpha_c_p1w": 0.07024880159278402, "bench_spy_p1w": 0.02933509717408933, "bench_c_p1w": 0.0886296215798672, "ret_p1m": 0.18587743300244153, "ret_signed_p1m": 0.18587743300244153, "alpha_spy_p1m": 0.20874354125639083, "alpha_c_p1m": 0.22574750421230028, "bench_spy_p1m": -0.022866108253949302, "bench_c_p1m": -0.03987007120985875, "ret_p3m": 0.5878444973966952, "ret_signed_p3m": 0.5878444973966952, "alpha_spy_p3m": 0.5490548301675224, "alpha_c_p3m": 0.40167832611186904, "bench_spy_p3m": 0.038789667229172764, "bench_c_p3m": 0.18616617128482615, "ret_p6m": 1.4049011961112958, "ret_signed_p6m": 1.4049011961112958, "alpha_spy_p6m": 1.3482147711068677, "alpha_c_p6m": 1.0575440181495737, "bench_spy_p6m": 0.05668642500442811, "bench_c_p6m": 0.34735717796172216, "price_path": [-0.4484, -0.4567, -0.4556, -0.4535, -0.4328, -0.4649, -0.4649, -0.4535, -0.4639, -0.4567, -0.4681, -0.4463, -0.4421, -0.4235, -0.4266, -0.4266, -0.4452, -0.4546, -0.4701, -0.4919, -0.4795, -0.4618, -0.4577, -0.4608, -0.4424, -0.4413, -0.4683, -0.459, -0.4548, -0.4486, -0.432, -0.4247, -0.4019, -0.3686, -0.3624, -0.3178, -0.3126, -0.2773, -0.3074, -0.2617, -0.2617, -0.271, -0.2503, -0.2575, -0.2596, -0.3011, -0.2752, -0.2783, -0.2523, -0.1786, -0.1786, -0.1786, -0.1786, -0.1786, -0.1786, -0.1111, -0.1381, -0.1454, -0.1225, -0.0602, -0.0332, 0.0083, -0.0052, 0.0, -0.0062, 0.0592, 0.1111, 0.082, 0.1589, 0.1796, 0.161, 0.2876, 0.217, 0.2025, 0.2316, 0.2046, 0.2586, 0.2856, 0.2814, 0.271, 0.163, 0.2586, 0.1838, 0.1672, 0.1859, 0.082, 0.08, 0.0779, 0.0883, 0.1306, 0.1015, 0.1181, 0.1597, 0.1472, 0.1265, 0.1306, 0.1992, 0.1763, 0.22, 0.1743, 0.1867, 0.1514, 0.1015, 0.1452, 0.1472, 0.1368, 0.2054, 0.2137, 0.2221, 0.2221, 0.2449, 0.3301, 0.353, 0.353, 0.353, 0.407, 0.4465, 0.5089, 0.5421, 0.5712, 0.5463, 0.5567, 0.5338, 0.5421, 0.5567, 0.5712, 0.5878, 0.5442, 0.538, 0.5691, 0.5941, 0.5297, 0.6627, 0.7479, 0.7894, 0.8892, 0.725, 0.885, 0.8705, 0.75, 0.7437, 0.8435, 0.8206, 0.7874, 0.8456, 0.8289, 0.8289, 0.8289, 0.8289, 0.858, 0.9723, 0.9765, 1.0887, 1.1157, 1.2883, 1.2092, 1.2092, 0.9552, 0.7678, 0.9593, 0.9239, 0.7407, 0.9531, 0.9885, 0.9364, 0.8948, 1.028, 1.0197, 1.1988, 1.1092, 1.0967, 0.9427, 1.053, 1.0718, 0.9427, 0.9427, 0.8177, 0.6803, 0.8677, 0.7282, 0.824, 0.8448, 0.9073, 1.1509, 1.078, 1.1384, 1.1655, 1.2966, 1.3653, 1.4049]}
{"unit_id": "orig:1975918922450046983", "ticker_idx": 0, "ts": "2025-10-08T13:39:16+00:00", "symbol": "memory", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The memory boom is coming.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Bullish Korean memory sector; endorses Morgan's raised price targets as confirmation of an incoming memory boom.", "resolved_tickers": ["000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Korea memory sector: SK Hynix, Samsung Electronics", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan, who once warned that “winter is coming” for memory, has finally raised the target prices of all the Korean memory makers!!\n\nThe memory boom is coming. https://t.co/iAGhIYelnl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.005927705196020927, "alpha_c_m1d": 0.026064132142976537, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.026064132142976537, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0028970083792374535, "alpha_c_p1d": 0.0024851289438175828, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0024851289438175828, "ret_p1w": 0.06338191186015119, "ret_signed_p1w": 0.06338191186015119, "alpha_spy_p1w": 0.07517786463876874, "alpha_c_p1w": 0.07577827710033225, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.012396365240181062, "ret_p1m": 0.30233017297874143, "ret_signed_p1m": 0.30233017297874143, "alpha_spy_p1m": 0.306489926101976, "alpha_c_p1m": 0.28782444685031827, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": 0.014505726128423158, "ret_p3m": 0.6536283086689896, "ret_signed_p3m": 0.6536283086689896, "alpha_spy_p3m": 0.6289040638695929, "alpha_c_p3m": 0.559153514408192, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": 0.09447479426079752, "ret_p6m": 1.0527641081377057, "ret_signed_p6m": 1.0527641081377057, "alpha_spy_p6m": 1.0728881392727032, "alpha_c_p6m": 0.915593199307988, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": 0.13717090882971772, "price_path": [-0.281, -0.2746, -0.2699, -0.2675, -0.2898, -0.2883, -0.28, -0.295, -0.2921, -0.2937, -0.2987, -0.2788, -0.2771, -0.2647, -0.2587, -0.2922, -0.2877, -0.2797, -0.2921, -0.2782, -0.278, -0.2692, -0.2661, -0.2503, -0.2538, -0.2538, -0.2741, -0.2797, -0.2864, -0.2991, -0.2871, -0.2758, -0.28, -0.2802, -0.2745, -0.2733, -0.3014, -0.2874, -0.281, -0.2756, -0.2688, -0.261, -0.2393, -0.213, -0.2048, -0.1665, -0.1573, -0.1197, -0.1447, -0.1052, -0.1052, -0.0931, -0.0738, -0.0744, -0.0718, -0.1126, -0.0897, -0.0933, -0.0658, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.067, 0.0444, 0.0305, 0.0634, 0.1164, 0.1339, 0.1603, 0.1487, 0.158, 0.1425, 0.1952, 0.2446, 0.213, 0.2653, 0.298, 0.3056, 0.4028, 0.3252, 0.2924, 0.3023, 0.2787, 0.3266, 0.3592, 0.3544, 0.3464, 0.2495, 0.3266, 0.2655, 0.2481, 0.2823, 0.1868, 0.1961, 0.2093, 0.2352, 0.2648, 0.2304, 0.2422, 0.282, 0.2805, 0.2712, 0.2921, 0.34, 0.32, 0.3443, 0.3126, 0.3291, 0.2847, 0.2432, 0.2982, 0.2978, 0.2842, 0.3494, 0.36, 0.3628, 0.3628, 0.4096, 0.4787, 0.4948, 0.4948, 0.4948, 0.5758, 0.6536, 0.6961, 0.7281, 0.7335, 0.7194, 0.7246, 0.704, 0.7241, 0.7531, 0.79, 0.8024, 0.7528, 0.7731, 0.8078, 0.8218, 0.7826, 0.905, 0.9731, 0.9889, 1.0485, 0.892, 1.0852, 1.0853, 0.957, 0.9493, 1.0537, 1.0364, 1.0274, 1.1233, 1.1277, 1.1277, 1.1277, 1.1277, 1.1947, 1.2648, 1.2836, 1.3911, 1.4271, 1.6133, 1.5568, 1.5568, 1.2823, 1.0401, 1.2653, 1.2247, 1.0309, 1.2408, 1.2741, 1.2306, 1.1806, 1.2909, 1.3149, 1.5056, 1.4064, 1.3926, 1.2255, 1.3117, 1.3192, 1.1908, 1.1908, 1.0955, 0.9608, 1.2008, 1.0528]}
{"unit_id": "thread:1992750734979617103", "ticker_idx": 0, "ts": "2025-11-24T00:22:53+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "starting by offering back-end advanced packaging services is not a bad way to break into the AI chip supply chain... If Intel fails to seize this opportunity to win some projects and gain experience, the business for this type of advanced-packaging alternative will end up being carved up directly by traditional OSAT providers", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish INTC: EMIB attracting Marvell, MediaTek, Apple, Qualcomm as CoWoS alternative, opening a new revenue stream in AI chip back-end packaging.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Intel’s EMIB Becomes an Alternative to TSMC’s CoWoS? Rumors Say Marvell and MediaTek Are Eager to Try\n\nRecently, as TSMC’s CoWoS advanced packaging capacity has become extremely tight, only a handful of leading AI chip makers have the ability to “lock in capacity” on a large scale, while other specialty chip (ASIC) and second-tier AI chip vendors are finding it difficult to secure sufficient CoWoS capacity support. Notably, there have been a steady stream of market rumors that Intel’s EMIB advanced packaging process has become one of the options under consideration for chip companies. Within the semiconductor industry, it is widely rumored that networking chip giant Marvell, and even MediaTek, are actively looking to try it, and that there could even be a new business model where “front-end wafer fabrication is done at TSMC, and back-end is handed over to Intel.”\n\nIntel’s Advanced Packaging Capabilities Are Solid, EMIB Is Attractive\n\nAccording to industry sources, Intel’s EMIB technology itself is relatively affordable and also performs well in terms of heat dissipation. For products whose technical specifications are not so demanding, it can indeed provide sufficient support. It is rumored that some ASIC vendors, such as Marvell and MediaTek, have already tried adopting EMIB as a process option in order to offer lower-cost solutions for their customers to consider. There is also word that EMIB already has confirmed adopters that are “non-Intel” customers.\n\nIn fact, the tightness in CoWoS capacity is not a recent development; it has persisted for some time. However, as investment in and demand for cloud AI chips continue to surge across the market, both TSMC and the test and packaging supply chain that works with its CoWoS process are operating at full capacity and rushing to fulfill orders.\n\nThis is evident from the fact that TSMC’s high-performance computing (HPC)-related revenue showed virtually no quarter-on-quarter growth last quarter, indicating that existing capacity is already fully utilized. What troubles the market even more is that an increasing number of U.S. customers now have requirements for domestic U.S. production. Although TSMC’s front-end wafer foundry processes in the U.S. are gradually coming online, all back-end processing still has to be shipped back to Taiwan, which makes it impossible to fully satisfy customers’ local-manufacturing requirements.\n\nBoth the shortage of capacity and the gap in U.S. domestic manufacturing have turned Intel’s EMIB into an alternative option worth considering.\n\nApple and Qualcomm Are Also Moving; Intel’s 2.5D/3D Advanced Packaging Are Both on the Table\n\nBoth Apple and Qualcomm have recently, and rather quietly, added “familiarity with EMIB technology” as a requirement in their job postings.\n\nAccording to players in the semiconductor supply chain, given Apple’s and Qualcomm’s technical requirements for cloud AI chips, one is likely to focus on its own cloud ASIC, while the other is targeting Tier-2 AI accelerator card products. In these cases, CoWoS is not necessarily required, and the more cost-effective EMIB may even deliver better overall benefits. For other ASIC projects aimed at inference applications with relatively lower compute requirements, EMIB can indeed provide adequate support as well.\n\nAlthough some in the semiconductor industry view EMIB as a way for Intel to mask its shortcomings in wafer-level circuit design by shifting part of the compute burden onto the substrate, EMIB is in fact a relatively mature technology with a solid track record. It is indeed capable of supporting certain urgent Tier-2 projects that need to go from design to mass production within a short timeframe.\n\nIn today’s environment, where ASICs and second-tier AI compute chips place increasing emphasis on price-performance, EMIB does offer certain advantages. People familiar with IC design say that the number of companies working with Intel to conduct cooperative testing on EMIB is on the rise. While shipment volumes may not be large in the short term, over the longer run, if the results of these collaborations are good, Intel’s 2.5D EMIB, or even its 3D Foveros packaging process, could have opportunities to secure a steady stream of orders.\n\nFront-End Wafers at TSMC, Back-End with Intel: Not a Mission Impossible\n\nSupply-chain sources stress that the approach currently being discussed in the market involves integrating TSMC’s front-end wafer foundry processes with Intel’s EMIB back-end process. Bringing the two together into a seamless flow will indeed take some careful effort.\n\nHowever, for Intel, given that its wafer foundry services are not yet able to convince these leading companies, starting by offering back-end advanced packaging services is not a bad way to break into the AI chip supply chain.\n\nAfter all, the current window in which CoWoS capacity is in short supply may only be temporary. If Intel fails to seize this opportunity to win some projects and gain experience, the business for this type of advanced-packaging alternative will end up being carved up directly by traditional OSAT providers such as Amkor, ASE Technology Holding, and SPIL.\n\nMediaTek’s spokesperson system indicated that it will not make any public comments regarding the technical details of its ASIC-related business.\n\n$INTC $TSM\n\n---\n\nhttps://t.co/h2ERFnrmOl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.036043612252820045, "ret_signed_m1d": -0.036043612252820045, "alpha_spy_m1d": -0.021538597354852418, "alpha_c_m1d": 0.0022613276929129844, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0011176561750236846, "ret_signed_p1d": 0.0011176561750236846, "alpha_spy_p1d": -0.008288243477083235, "alpha_c_p1d": -0.001506868844082021, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.11790995524046033, "ret_signed_p1w": 0.11790995524046033, "alpha_spy_p1w": 0.10065326125190333, "alpha_c_p1w": 0.07709838357117604, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.015646760108094737, "ret_signed_p1m": 0.015646760108094737, "alpha_spy_p1m": -0.01614918642940011, "alpha_c_p1m": -0.05857483270796093, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.24671689830906263, "ret_signed_p3m": 0.24671689830906263, "alpha_spy_p3m": 0.22014031185059513, "alpha_c_p3m": 0.03335169827952611, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.095836832004318, "ret_signed_p6m": 2.095836832004318, "alpha_spy_p6m": 1.9923902020309896, "alpha_c_p6m": 1.486811621795878, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.3057, -0.3034, -0.3196, -0.3196, -0.3236, -0.3294, -0.3124, -0.3157, -0.316, -0.3171, -0.3079, -0.3124, -0.3272, -0.3079, -0.2939, -0.3043, -0.1459, -0.1735, -0.1964, -0.1802, -0.1277, -0.0503, -0.0081, -0.0366, -0.0626, 0.0042, 0.0422, 0.0291, 0.0224, 0.0386, 0.0458, 0.0562, 0.0162, 0.04, -0.0045, 0.038, 0.0293, 0.0341, 0.0645, 0.0651, 0.0316, 0.0662, 0.0696, 0.1048, 0.1604, 0.1551, 0.1221, 0.1174, 0.1037, 0.0346, 0.0724, 0.0405, 0.0654, 0.0743, 0.0584, 0.0587, 0.0034, -0.0075, -0.0302, -0.0408, -0.019, -0.0606, -0.036, 0.0, 0.0011, 0.0285, 0.0285, 0.1333, 0.1179, 0.2146, 0.2227, 0.1316, 0.157, 0.126, 0.1316, 0.1394, 0.1039, 0.0564, 0.0481, 0.0425, 0.0073, 0.0137, 0.0288, 0.0162, 0.0156, 0.0103, 0.0103, 0.0115, 0.0249, 0.0422, 0.031, 0.031, 0.1003, 0.1, 0.1187, 0.1911, 0.1486, 0.2727, 0.2311, 0.3213, 0.3613, 0.3501, 0.3121, 0.3121, 0.3568, 0.5158, 0.5177, 0.2593, 0.1872, 0.2274, 0.363, 0.3596, 0.2984, 0.3638, 0.3761, 0.3579, 0.3479, 0.4135, 0.4037, 0.3168, 0.3493, 0.2987, 0.3073, 0.3073, 0.2903, 0.2702, 0.2467, 0.2325, 0.2191, 0.2886, 0.3099, 0.2702, 0.2744, 0.2713, 0.2042, 0.2735, 0.2839, 0.2132, 0.2735, 0.3071, 0.3406, 0.2643, 0.2788, 0.2786, 0.2311, 0.2582, 0.2903, 0.2258, 0.2297, 0.2311, 0.3182, 0.2322, 0.2051, 0.1509, 0.233, 0.342, 0.4077, 0.4077, 0.4188, 0.4783, 0.6471, 0.7245, 0.7429, 0.8212, 0.7829, 0.8145, 0.9139, 0.9139, 0.8357, 0.8514, 0.8237, 0.8659, 1.3062, 1.3747, 1.3616, 1.6474, 1.6398, 1.7835, 1.6762, 2.0218, 2.1576, 2.0629, 2.4904, 2.6167, 2.3699, 2.361, 2.2392, 2.0391, 2.0224, 2.0958]}
{"unit_id": "thread:1986202602686849294", "ticker_idx": 0, "ts": "2025-11-05T22:42:56+00:00", "symbol": "Winbond Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the shortage of DDR5, DDR4, and DDR3 is expected to persist until 2027. This quarter, the company's DRAM revenue, shipments, and prices are all expected to grow significantly, with active customer demand and long-term contract inquiries. Stable growth is also anticipated to continue through 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Winbond bullish: DRAM shortage persists to 2027, revenue/shipments/prices to grow significantly this quarter and through 2026.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "• Winbond Electronics: The memory semiconductor industry is undergoing a structural transformation, and the shortage of DDR5, DDR4, and DDR3 is expected to persist until 2027. This quarter, the company’s DRAM revenue, shipments, and prices are all expected to grow significantly, with active customer demand and long-term contract inquiries. Stable growth is also anticipated to continue through 2026.\n\n---\n\nhttps://t.co/B9S6QG7795", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04821432459900654, "ret_signed_m1d": -0.04821432459900654, "alpha_spy_m1d": -0.04476090285242362, "alpha_c_m1d": -0.02946608354252922, "bench_spy_m1d": -0.0034534217465829187, "bench_c_m1d": -0.01874824105647732, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05357142612584864, "ret_signed_p1d": 0.05357142612584864, "alpha_spy_p1d": 0.06430075933851465, "alpha_c_p1d": 0.07696502372044411, "bench_spy_p1d": -0.010729333212666003, "bench_c_p1d": -0.023393597594595472, "ret_p1w": 0.15714283170882593, "ret_signed_p1w": 0.15714283170882593, "alpha_spy_p1w": 0.14858297316857572, "alpha_c_p1w": 0.16818599538456125, "bench_spy_p1w": 0.00855985854025021, "bench_c_p1w": -0.011043163675735324, "ret_p1m": 0.03214281458976642, "ret_signed_p1m": 0.03214281458976642, "alpha_spy_p1m": 0.02209237077058157, "alpha_c_p1m": 0.026301397744526556, "bench_spy_p1m": 0.010050443819184851, "bench_c_p1m": 0.005841416845239866, "ret_p3m": 1.0714285225169728, "ret_signed_p3m": 1.0714285225169728, "alpha_spy_p3m": 1.0420816396601802, "alpha_c_p3m": 0.933071270390168, "bench_spy_p3m": 0.029346882856792655, "bench_c_p3m": 0.13835725212680483, "ret_p6m": 0.6119670716944843, "ret_signed_p6m": 0.6119670716944843, "alpha_spy_p6m": 0.5453003869096058, "alpha_c_p6m": 0.1980680436744342, "bench_spy_p6m": 0.06666668478487847, "bench_c_p6m": 0.4138990280200501, "price_path": [-0.6884, -0.6893, -0.6705, -0.6768, -0.6786, -0.6714, -0.6589, -0.6679, -0.6786, -0.6714, -0.675, -0.6714, -0.6652, -0.6464, -0.65, -0.6482, -0.6375, -0.6438, -0.6411, -0.6339, -0.5973, -0.5607, -0.5616, -0.5545, -0.5536, -0.5455, -0.5214, -0.5009, -0.475, -0.4232, -0.4018, -0.3884, -0.3491, -0.3714, -0.4036, -0.4393, -0.4393, -0.392, -0.3964, -0.3571, -0.3071, -0.3071, -0.2384, -0.2634, -0.2241, -0.2241, -0.2321, -0.2732, -0.2857, -0.2143, -0.2152, -0.1723, -0.2036, -0.1884, -0.1732, -0.1732, -0.0911, -0.0357, -0.0357, -0.0143, -0.0321, 0.0107, -0.0482, 0.0, 0.0536, 0.0375, 0.1411, 0.1518, 0.1571, 0.1464, 0.0768, 0.1821, 0.0964, 0.0607, 0.0357, -0.0679, -0.025, 0.0268, -0.0429, 0.0161, 0.0357, 0.0286, 0.0107, -0.0036, 0.0321, 0.1036, 0.2125, 0.2732, 0.2054, 0.2214, 0.3321, 0.2696, 0.2607, 0.3196, 0.3304, 0.2732, 0.3054, 0.3036, 0.3732, 0.3732, 0.3661, 0.3732, 0.4946, 0.475, 0.475, 0.6214, 0.7054, 0.8571, 0.9018, 0.9375, 0.7464, 0.8214, 0.7821, 0.7821, 0.7714, 0.9018, 1.0804, 1.0179, 0.8839, 0.8661, 0.8393, 1.0179, 1.0625, 1.2679, 1.3214, 1.2946, 1.0714, 0.8839, 0.9375, 0.8304, 0.7464, 0.9196, 0.8482, 0.8839, 0.8839, 0.8839, 0.8839, 0.8839, 0.8839, 0.8839, 0.8839, 1.0536, 1.1786, 1.1071, 1.1875, 1.1875, 1.2232, 1.0089, 0.8571, 1.0, 0.9018, 0.8036, 0.8839, 1.0357, 0.9464, 0.9464, 1.0893, 1.2054, 1.2857, 1.1786, 0.9643, 0.9196, 0.7768, 0.7536, 0.7143, 0.6586, 0.6694, 0.6191, 0.6712, 0.6479, 0.6479, 0.6479, 0.6048, 0.7161, 0.6335, 0.6784, 0.6945, 0.6784, 0.6191, 0.6084, 0.5761, 0.5384, 0.6353, 0.6263, 0.5761, 0.5832, 0.6856, 0.7143, 0.664, 0.612]}
{"unit_id": "quote:1983340631281053832", "ticker_idx": 0, "ts": "2025-10-29T01:10:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix: Even under conservative assumptions, the HBM market is expected to grow at an average annual rate of more than 30% over the next five years. Bullish 🔥", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish SK Hynix on 30%+ annual HBM market growth over next five years under conservative assumptions.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "SK Hynix: Even under conservative assumptions, the HBM market is expected to grow at an average annual rate of more than 30% over the next five years.\n\nBullish 🔥", "reply_text": null, "parent_text": null, "quoted_text": "SK hynix 3Q25 Earnings Call – Key Q&amp;A (5)\n\n⸻\n\n9. Possibility of early shipments or preparations related to M15X and the Yongin fab (May 2027)\n    \n•    Regarding new fab preparation plans: To meet the substantial wafer capacity requirements driven by HBM demand, we began", "parent_author": null, "ret_m1d": -0.06630818978515529, "ret_signed_m1d": -0.06630818978515529, "alpha_spy_m1d": -0.0658281226614843, "alpha_c_m1d": -0.05138726656730519, "bench_spy_m1d": -0.0004800671236709908, "bench_c_m1d": -0.014920923217850102, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017921111125961398, "ret_signed_p1d": 0.017921111125961398, "alpha_spy_p1d": 0.028919256045152775, "alpha_c_p1d": 0.031159968394103288, "bench_spy_p1d": -0.010998144919191377, "bench_c_p1d": -0.01323885726814189, "ret_p1w": 0.037634400752317276, "ret_signed_p1w": 0.037634400752317276, "alpha_spy_p1w": 0.051905771195720374, "alpha_c_p1w": 0.06234882782715112, "bench_spy_p1w": -0.014271370443403097, "bench_c_p1w": -0.024714427074833845, "ret_p1m": -0.024391350596365213, "ret_signed_p1m": -0.024391350596365213, "alpha_spy_p1m": -0.013175050174359826, "alpha_c_p1m": 0.03230815763146533, "bench_spy_p1m": -0.011216300422005387, "bench_c_p1m": -0.05669950822783054, "ret_p3m": 0.31994109407909455, "ret_signed_p3m": 0.31994109407909455, "alpha_spy_p3m": 0.3091947700841502, "alpha_c_p3m": 0.2346178959945271, "bench_spy_p3m": 0.010746323994944351, "bench_c_p3m": 0.08532319808456745, "ret_p6m": 1.2009668127446949, "ret_signed_p6m": 1.2009668127446949, "alpha_spy_p6m": 1.164460726268675, "alpha_c_p6m": 0.8896910993470071, "bench_spy_p6m": 0.03650608647601983, "bench_c_p6m": 0.3112757133976878, "price_path": [-0.5383, -0.5383, -0.5285, -0.5374, -0.5311, -0.541, -0.5222, -0.5186, -0.5025, -0.5052, -0.5052, -0.5213, -0.5294, -0.5428, -0.5616, -0.5508, -0.5356, -0.532, -0.5347, -0.5188, -0.5179, -0.5412, -0.5332, -0.5296, -0.5242, -0.5099, -0.5036, -0.4839, -0.4552, -0.4498, -0.4113, -0.4068, -0.3763, -0.4023, -0.3629, -0.3629, -0.371, -0.353, -0.3593, -0.3611, -0.397, -0.3746, -0.3772, -0.3548, -0.2912, -0.2912, -0.2912, -0.2912, -0.2912, -0.2912, -0.233, -0.2563, -0.2625, -0.2428, -0.1891, -0.1658, -0.1299, -0.1416, -0.1371, -0.1425, -0.086, -0.0412, -0.0663, 0.0, 0.0179, 0.0018, 0.1111, 0.0502, 0.0376, 0.0627, 0.0394, 0.086, 0.1093, 0.1057, 0.0968, 0.0036, 0.086, 0.0215, 0.0072, 0.0233, -0.0663, -0.0681, -0.0699, -0.0609, -0.0244, -0.0495, -0.0352, 0.0007, -0.01, -0.028, -0.0244, 0.0348, 0.0151, 0.0527, 0.0133, 0.024, -0.0065, -0.0495, -0.0118, -0.01, -0.019, 0.0402, 0.0473, 0.0545, 0.0545, 0.0742, 0.1478, 0.1675, 0.1675, 0.1675, 0.2141, 0.2482, 0.302, 0.3307, 0.3558, 0.3343, 0.3433, 0.3235, 0.3307, 0.3433, 0.3558, 0.3702, 0.3325, 0.3271, 0.354, 0.3755, 0.3199, 0.4347, 0.5082, 0.5441, 0.6302, 0.4885, 0.6266, 0.6141, 0.51, 0.5047, 0.5907, 0.571, 0.5423, 0.5925, 0.5782, 0.5782, 0.5782, 0.5782, 0.6033, 0.7019, 0.7055, 0.8024, 0.8257, 0.9746, 0.9063, 0.9063, 0.6871, 0.5254, 0.6907, 0.6602, 0.502, 0.6853, 0.7159, 0.6709, 0.635, 0.75, 0.7428, 0.8973, 0.8201, 0.8093, 0.6763, 0.7716, 0.7877, 0.6763, 0.6763, 0.5685, 0.4499, 0.6116, 0.4913, 0.5739, 0.5919, 0.6458, 0.856, 0.7931, 0.8452, 0.8686, 0.9818, 1.0411, 1.0752, 1.0267, 1.095, 1.1992, 1.1974, 1.201]}
{"unit_id": "thread:1984404886214389895", "ticker_idx": 0, "ts": "2025-10-31T23:39:27+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this move is essentially equivalent to Intel poaching TSMC's N2 process know-how, and expectations within Intel are said to be very high", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish INTC: hiring TSMC's N2 strategy SVP seen as poaching key process know-how with high internal expectations.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Sources say that Wei-Jen Lo, former Senior Vice President of Corporate Strategy Development at TSMC, has joined Intel.\n\nAccording to the sources, this move is essentially equivalent to Intel poaching TSMC’s N2 process know-how, and expectations within Intel are said to be very high.\n\nSource: Own\n\n$INTC\n\n---\n\nRelated: https://t.co/wousOJttLV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004251016799452945, "ret_signed_m1d": 0.004251016799452945, "alpha_spy_m1d": 0.007520572350488752, "alpha_c_m1d": 0.0022951267826381017, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": 0.001955890016814843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012253104723690122, "ret_signed_p1d": -0.012253104723690122, "alpha_spy_p1d": -0.014129791811654635, "alpha_c_p1d": -0.02073753872652684, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.008484434002836716, "ret_p1w": -0.04651164121738727, "ret_signed_p1w": -0.04651164121738727, "alpha_spy_p1w": -0.03025196956968068, "alpha_c_p1w": -0.0054671281491680546, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04104451306821921, "ret_p1m": 0.0005000410659441457, "ret_signed_p1m": 0.0005000410659441457, "alpha_spy_p1m": 0.0031244151210984006, "alpha_c_p1m": 0.02821196850389185, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.027711927437947703, "ret_p3m": 0.2198048695147512, "ret_signed_p3m": 0.2198048695147512, "alpha_spy_p3m": 0.19720443800181542, "alpha_c_p3m": 0.06855589620100422, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.15124897331374698, "ret_p6m": 1.125281177708318, "ret_signed_p6m": 1.125281177708318, "alpha_spy_p6m": 1.0707667161572294, "alpha_c_p6m": 0.7263629297621121, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.3989182479462059, "price_path": [-0.4951, -0.4896, -0.5056, -0.5011, -0.4836, -0.4546, -0.4444, -0.4034, -0.3858, -0.4084, -0.3671, -0.4114, -0.4124, -0.3798, -0.3861, -0.3911, -0.3786, -0.3766, -0.3911, -0.3911, -0.3946, -0.3998, -0.3846, -0.3876, -0.3878, -0.3888, -0.3806, -0.3846, -0.3978, -0.3806, -0.3681, -0.3773, -0.2356, -0.2603, -0.2808, -0.2663, -0.2193, -0.15, -0.1123, -0.1378, -0.161, -0.1013, -0.0673, -0.079, -0.085, -0.0705, -0.064, -0.0548, -0.0905, -0.0693, -0.109, -0.071, -0.0788, -0.0745, -0.0473, -0.0468, -0.0768, -0.0458, -0.0428, -0.0113, 0.0385, 0.0338, 0.0043, 0.0, -0.0123, -0.074, -0.0403, -0.0688, -0.0465, -0.0385, -0.0528, -0.0525, -0.102, -0.1118, -0.132, -0.1415, -0.122, -0.1593, -0.1373, -0.105, -0.104, -0.0795, -0.0795, 0.0143, 0.0005, 0.087, 0.0943, 0.0128, 0.0355, 0.0078, 0.0128, 0.0198, -0.012, -0.0545, -0.062, -0.067, -0.0985, -0.0928, -0.0793, -0.0905, -0.091, -0.0958, -0.0958, -0.0948, -0.0828, -0.0673, -0.0773, -0.0773, -0.0153, -0.0155, 0.0013, 0.066, 0.028, 0.139, 0.1018, 0.1825, 0.2183, 0.2083, 0.1743, 0.1743, 0.2143, 0.3566, 0.3583, 0.127, 0.0625, 0.0985, 0.2198, 0.2168, 0.162, 0.2206, 0.2316, 0.2153, 0.2063, 0.2651, 0.2563, 0.1785, 0.2076, 0.1623, 0.17, 0.17, 0.1548, 0.1368, 0.1158, 0.103, 0.091, 0.1533, 0.1723, 0.1368, 0.1405, 0.1378, 0.0778, 0.1398, 0.149, 0.0858, 0.1398, 0.1698, 0.1998, 0.1315, 0.1445, 0.1443, 0.1018, 0.126, 0.1548, 0.097, 0.1005, 0.1018, 0.1798, 0.1028, 0.0785, 0.03, 0.1035, 0.2011, 0.2598, 0.2598, 0.2698, 0.3231, 0.4741, 0.5434, 0.5599, 0.6299, 0.5956, 0.6239, 0.7129, 0.7129, 0.6429, 0.6569, 0.6322, 0.6699, 1.064, 1.1253]}
{"unit_id": "orig:1980569181776146720", "ticker_idx": 0, "ts": "2025-10-21T09:37:44+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Current market consensus estimates $60 billion in 2027 AI revenue, suggesting 50–70% upside potential. Including Google's TPU revenue (~$20 billion), total AI revenue in 2027 could approach $100 billion.", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM analysis: AVGO, NVDA, AMD all have significant 2026–2028 earnings upside on AI accelerator demand; AVGO 50–70% consensus upside, NVDA and AMD estimates likely revised higher.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI Accelerator Market Revenue per GW – Growth Potential of AVGO, NVDA, and AMD (JPM)\n\n⸻\n\n1. Overall Market Outlook\n    • With the recent expansion of large-scale AI compute capacity contracts, interest in the AI revenue potential of NVIDIA (NVDA), AMD, and Broadcom (AVGO) has surged.\n    • JP Morgan estimates GPU/XPU-based data center revenue on a per-gigawatt (GW) basis to project the upside potential of AI-related revenues over the next 3–4 years.\n    • The AI accelerator market is expected to grow at a CAGR of 40–50%, gradually diversifying from GPU-centric to XPU (custom AI SoC)-based architectures.\n    • Although conservative risks remain (e.g., funding, competition, and execution), the ramp-up of large customer contracts (e.g., OpenAI, Google) could drive further upward revisions.\n\n⸻\n\n2. Broadcom\n    • A 10GW custom AI silicon and rack system supply contract with OAI will significantly accelerate Broadcom’s AI revenue structure.\n    • The price of XPU systems is about 30% lower than GPUs, but efficiency is 30% higher, leading to roughly $27 billion per GW of revenue.\n    • Based on Ironwood TPU v6p (3nm, Google reference), the result is consistent at approximately $28 billion per GW.\n    • Including networking (30%) and rack components (30%), total revenue per GW is around $28 billion.\n    • If the 10GW contract is deployed over the next 3–4 years, it implies $70–90 billion in annual revenue potential between 2027–2029.\n    • Current market consensus estimates $60 billion in 2027 AI revenue, suggesting 50–70% upside potential.\n    • Including Google’s TPU revenue (~$20 billion), total AI revenue in 2027 could approach $100 billion.\n\n⸻\n\n3. NVIDIA\n    • Management has indicated that 1GW of AI data center capacity represents a $35–40 billion revenue opportunity, roughly 60% of total data center CapEx.\n    • Blackwell (BW) and Blackwell Ultra (BWU) are estimated at $27–29 billion per GW, while Rubin (R) and Rubin Ultra (RU) are projected at $30–35 billion per GW.\n    • Rack power density is expected to jump from 120kW (BW) to 1.1MW (RU) — about 9x higher. Although GPU count per rack decreases, ASP per GPU triples.\n    • GPU revenue per unit rises from approximately $52,000 (BW) to $300,000 (RU).\n    • The market currently assumes 10GW of AI data center capacity by 2027, but actual deployments could exceed that due to multiple large-scale projects.\n    • Consequently, AI revenue estimates for the next 2–3 years are likely to be revised upward.\n\n4. AMD\n    • The OpenAI contract is explicitly described as being worth several billion dollars, implying at least $20 billion per GW.\n    • The Helios platform is assumed to be priced about 20% lower than NVIDIA’s Rubin, with networking contributing roughly 5% of total revenue.\n    • Accordingly, the MI450 (Helios) system is estimated at $20 billion per GW, and the next-generation MI500 could exceed $25 billion per GW.\n    • If OpenAI’s 6GW deployment plan is fully realized, it represents $30–35 billion in annual revenue potential.\n    • The market currently assumes $31 billion in 2027 data center GPU revenue, suggesting room for upward revision.\n\n5. Comparative Summary\n\nAVGO: approximately $28 billion per GW\nNVDA: approximately $35 billion per GW (Rubin Ultra basis)\nAMD: approximately $20 billion per GW (Helios basis)\n\nBroadcom benefits from a broader scope including networking and rack components. NVIDIA’s growth is driven by rising GPU profitability per unit. AMD targets market share gains through aggressive pricing and major customer wins.\n\n6. Investment View\n    • All three companies have significant earnings upside potential for 2026–2028, driven by accelerating AI demand.\n    • NVIDIA leads in high-performance platforms, Broadcom in XPU and networking integration, and AMD in Helios-based large customer expansion.\n    • If long-term AI infrastructure CapEx continues to grow, all three are positioned for sustained re-rating and valuation expansion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019202600716187623, "ret_signed_m1d": 0.019202600716187623, "alpha_spy_m1d": 0.019187604571294203, "alpha_c_m1d": 0.0136421652007499, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0068873625963382645, "ret_signed_p1d": -0.0068873625963382645, "alpha_spy_p1d": -0.0016885466334175536, "alpha_c_p1d": 0.012516004091333688, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.08845491359008517, "ret_signed_p1w": 0.08845491359008517, "alpha_spy_p1w": 0.0649627210951802, "alpha_c_p1w": 0.036876554890219904, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.034319674881862605, "ret_signed_p1m": 0.034319674881862605, "alpha_spy_p1m": 0.04722001708657386, "alpha_c_p1m": 0.05129035103153057, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.02837487160635943, "ret_signed_p3m": 0.02837487160635943, "alpha_spy_p3m": -0.005014319318705107, "alpha_c_p3m": -0.13477568250451522, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.16241421087236496, "ret_signed_p6m": 0.16241421087236496, "alpha_spy_p6m": 0.11379808072630992, "alpha_c_p6m": -0.15357065821803229, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.1589, -0.1546, -0.1426, -0.1335, -0.1184, -0.1443, -0.1591, -0.1326, -0.1466, -0.1211, -0.115, -0.1115, -0.1146, -0.0886, -0.0995, -0.0933, -0.1075, -0.1092, -0.1408, -0.1517, -0.1563, -0.1435, -0.1428, -0.1318, -0.1253, -0.1008, -0.1336, -0.1336, -0.1311, -0.119, -0.1082, -0.0243, 0.007, -0.0192, 0.0767, 0.0477, 0.0484, 0.0607, 0.0488, 0.0085, 0.0061, 0.0049, -0.0113, -0.0109, -0.0098, -0.0191, -0.0237, -0.0431, -0.0372, -0.0271, -0.0131, -0.0125, -0.0209, -0.0182, 0.0083, 0.0069, -0.0526, 0.041, 0.0043, 0.0253, 0.0335, 0.0195, 0.0192, 0.0, -0.0069, 0.0048, 0.0335, 0.0566, 0.0885, 0.1264, 0.0987, 0.0787, 0.058, 0.0271, 0.0476, 0.0377, 0.0198, 0.0459, 0.0271, 0.0367, -0.0078, -0.0006, -0.0, -0.0063, 0.0343, 0.0121, -0.0072, 0.103, 0.1237, 0.1602, 0.1602, 0.176, 0.1267, 0.1136, 0.1108, 0.112, 0.1389, 0.1705, 0.1857, 0.2052, 0.1859, 0.0504, -0.0083, -0.004, -0.0486, -0.0373, -0.0067, -0.0016, 0.0214, 0.024, 0.024, 0.0296, 0.0216, 0.0229, 0.012, 0.012, 0.0164, 0.0041, 0.0052, 0.0044, -0.0279, 0.0087, 0.0298, 0.0369, -0.0062, 0.003, 0.0284, 0.0284, -0.0275, -0.0386, -0.0483, -0.0642, -0.0502, -0.0269, -0.0256, -0.033, -0.0313, -0.0319, -0.0634, -0.0993, -0.0921, -0.0266, 0.0057, -0.0046, 0.0022, -0.0317, -0.0492, -0.0492, -0.0277, -0.0248, -0.0234, -0.0274, -0.0341, -0.0483, -0.0283, -0.0594, -0.0657, -0.0678, -0.0824, -0.0716, -0.027, -0.0337, 0.0109, 0.0017, -0.0013, -0.0176, -0.058, -0.05, -0.0605, -0.0762, -0.0648, -0.0921, -0.055, -0.0674, -0.0659, -0.0934, -0.119, -0.1403, -0.0931, -0.0815, -0.0783, -0.0783, -0.0787, -0.0214, 0.0274, 0.0399, 0.0887, 0.1127, 0.1157, 0.1624]}
{"unit_id": "orig:1980569181776146720", "ticker_idx": 1, "ts": "2025-10-21T09:37:44+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI revenue estimates for the next 2–3 years are likely to be revised upward", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM analysis: AVGO, NVDA, AMD all have significant 2026–2028 earnings upside on AI accelerator demand; AVGO 50–70% consensus upside, NVDA and AMD estimates likely revised higher.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI Accelerator Market Revenue per GW – Growth Potential of AVGO, NVDA, and AMD (JPM)\n\n⸻\n\n1. Overall Market Outlook\n    • With the recent expansion of large-scale AI compute capacity contracts, interest in the AI revenue potential of NVIDIA (NVDA), AMD, and Broadcom (AVGO) has surged.\n    • JP Morgan estimates GPU/XPU-based data center revenue on a per-gigawatt (GW) basis to project the upside potential of AI-related revenues over the next 3–4 years.\n    • The AI accelerator market is expected to grow at a CAGR of 40–50%, gradually diversifying from GPU-centric to XPU (custom AI SoC)-based architectures.\n    • Although conservative risks remain (e.g., funding, competition, and execution), the ramp-up of large customer contracts (e.g., OpenAI, Google) could drive further upward revisions.\n\n⸻\n\n2. Broadcom\n    • A 10GW custom AI silicon and rack system supply contract with OAI will significantly accelerate Broadcom’s AI revenue structure.\n    • The price of XPU systems is about 30% lower than GPUs, but efficiency is 30% higher, leading to roughly $27 billion per GW of revenue.\n    • Based on Ironwood TPU v6p (3nm, Google reference), the result is consistent at approximately $28 billion per GW.\n    • Including networking (30%) and rack components (30%), total revenue per GW is around $28 billion.\n    • If the 10GW contract is deployed over the next 3–4 years, it implies $70–90 billion in annual revenue potential between 2027–2029.\n    • Current market consensus estimates $60 billion in 2027 AI revenue, suggesting 50–70% upside potential.\n    • Including Google’s TPU revenue (~$20 billion), total AI revenue in 2027 could approach $100 billion.\n\n⸻\n\n3. NVIDIA\n    • Management has indicated that 1GW of AI data center capacity represents a $35–40 billion revenue opportunity, roughly 60% of total data center CapEx.\n    • Blackwell (BW) and Blackwell Ultra (BWU) are estimated at $27–29 billion per GW, while Rubin (R) and Rubin Ultra (RU) are projected at $30–35 billion per GW.\n    • Rack power density is expected to jump from 120kW (BW) to 1.1MW (RU) — about 9x higher. Although GPU count per rack decreases, ASP per GPU triples.\n    • GPU revenue per unit rises from approximately $52,000 (BW) to $300,000 (RU).\n    • The market currently assumes 10GW of AI data center capacity by 2027, but actual deployments could exceed that due to multiple large-scale projects.\n    • Consequently, AI revenue estimates for the next 2–3 years are likely to be revised upward.\n\n4. AMD\n    • The OpenAI contract is explicitly described as being worth several billion dollars, implying at least $20 billion per GW.\n    • The Helios platform is assumed to be priced about 20% lower than NVIDIA’s Rubin, with networking contributing roughly 5% of total revenue.\n    • Accordingly, the MI450 (Helios) system is estimated at $20 billion per GW, and the next-generation MI500 could exceed $25 billion per GW.\n    • If OpenAI’s 6GW deployment plan is fully realized, it represents $30–35 billion in annual revenue potential.\n    • The market currently assumes $31 billion in 2027 data center GPU revenue, suggesting room for upward revision.\n\n5. Comparative Summary\n\nAVGO: approximately $28 billion per GW\nNVDA: approximately $35 billion per GW (Rubin Ultra basis)\nAMD: approximately $20 billion per GW (Helios basis)\n\nBroadcom benefits from a broader scope including networking and rack components. NVIDIA’s growth is driven by rising GPU profitability per unit. AMD targets market share gains through aggressive pricing and major customer wins.\n\n6. Investment View\n    • All three companies have significant earnings upside potential for 2026–2028, driven by accelerating AI demand.\n    • NVIDIA leads in high-performance platforms, Broadcom in XPU and networking integration, and AMD in Helios-based large customer expansion.\n    • If long-term AI infrastructure CapEx continues to grow, all three are positioned for sustained re-rating and valuation expansion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008169605123257151, "ret_signed_m1d": 0.008169605123257151, "alpha_spy_m1d": 0.008154608978363731, "alpha_c_m1d": 0.002609169607819428, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004857555235790012, "ret_signed_p1d": -0.004857555235790012, "alpha_spy_p1d": 0.0003412607271306989, "alpha_c_p1d": 0.01454581145188194, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.10968210622045049, "ret_signed_p1w": 0.10968210622045049, "alpha_spy_p1w": 0.08618991372554552, "alpha_c_p1w": 0.058103747520585225, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": 0.029587072846396723, "ret_signed_p1m": 0.029587072846396723, "alpha_spy_p1m": 0.04248741505110798, "alpha_c_p1m": 0.046557748996064685, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": 0.02804350496049346, "ret_signed_p3m": 0.02804350496049346, "alpha_spy_p3m": -0.005345685964571079, "alpha_c_p3m": -0.1351070491503812, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.097879422110557, "ret_signed_p6m": 0.097879422110557, "alpha_spy_p6m": 0.049263291964501965, "alpha_c_p6m": -0.21810544697984025, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.041, -0.0423, -0.0244, -0.0312, -0.0105, -0.0182, -0.0411, -0.0065, -0.0161, -0.0097, -0.0022, 0.0084, 0.0049, 0.011, 0.0023, 0.0047, -0.004, 0.0046, -0.0305, -0.0318, -0.0342, -0.0176, -0.0075, 0.0033, 0.0024, -0.0055, -0.0386, -0.0386, -0.0574, -0.0582, -0.0525, -0.0781, -0.071, -0.0575, -0.0212, -0.022, -0.0184, -0.0188, -0.0347, -0.06, -0.0272, -0.0248, 0.0135, -0.0151, -0.0231, -0.0192, -0.0164, 0.0038, 0.0299, 0.0336, 0.0427, 0.0357, 0.0242, 0.0214, 0.0439, 0.063, 0.011, 0.0395, -0.0062, -0.0073, 0.0036, 0.0114, 0.0082, 0.0, -0.0049, 0.0055, 0.0282, 0.057, 0.1097, 0.1429, 0.1199, 0.1177, 0.142, 0.0968, 0.0776, 0.0382, 0.0386, 0.0988, 0.0662, 0.0698, 0.0315, 0.0497, 0.03, 0.0011, 0.0296, -0.0029, -0.0126, 0.0077, -0.0184, -0.005, -0.005, -0.023, -0.0068, 0.0017, -0.0087, 0.0123, 0.007, 0.0243, 0.0211, 0.0145, -0.0012, -0.0338, -0.0268, -0.0189, -0.0564, -0.0387, -0.0009, 0.014, 0.0445, 0.0412, 0.0412, 0.0518, 0.039, 0.0353, 0.0295, 0.0295, 0.0425, 0.0385, 0.0336, 0.0439, 0.0215, 0.0205, 0.0209, 0.0257, 0.011, 0.0326, 0.028, 0.028, -0.017, 0.012, 0.0204, 0.036, 0.0294, 0.0407, 0.0572, 0.0627, 0.0551, 0.0246, -0.0045, -0.0384, -0.0512, 0.0235, 0.0491, 0.0408, 0.0491, 0.032, 0.0092, 0.0092, 0.0211, 0.0377, 0.0373, 0.0479, 0.0574, 0.0646, 0.0795, 0.0206, -0.0219, 0.0073, -0.0061, 0.0104, 0.0121, -0.0184, 0.0083, 0.02, 0.027, 0.011, -0.0049, 0.0115, 0.0044, -0.0041, -0.0142, -0.0466, -0.0304, -0.0328, -0.0136, -0.0547, -0.0752, -0.0882, -0.0372, -0.0298, -0.0207, -0.0207, -0.0193, -0.0168, 0.0052, 0.0153, 0.0413, 0.0451, 0.0849, 0.0979]}
{"unit_id": "orig:1980569181776146720", "ticker_idx": 2, "ts": "2025-10-21T09:37:44+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The market currently assumes $31 billion in 2027 data center GPU revenue, suggesting room for upward revision", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM analysis: AVGO, NVDA, AMD all have significant 2026–2028 earnings upside on AI accelerator demand; AVGO 50–70% consensus upside, NVDA and AMD estimates likely revised higher.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "AI Accelerator Market Revenue per GW – Growth Potential of AVGO, NVDA, and AMD (JPM)\n\n⸻\n\n1. Overall Market Outlook\n    • With the recent expansion of large-scale AI compute capacity contracts, interest in the AI revenue potential of NVIDIA (NVDA), AMD, and Broadcom (AVGO) has surged.\n    • JP Morgan estimates GPU/XPU-based data center revenue on a per-gigawatt (GW) basis to project the upside potential of AI-related revenues over the next 3–4 years.\n    • The AI accelerator market is expected to grow at a CAGR of 40–50%, gradually diversifying from GPU-centric to XPU (custom AI SoC)-based architectures.\n    • Although conservative risks remain (e.g., funding, competition, and execution), the ramp-up of large customer contracts (e.g., OpenAI, Google) could drive further upward revisions.\n\n⸻\n\n2. Broadcom\n    • A 10GW custom AI silicon and rack system supply contract with OAI will significantly accelerate Broadcom’s AI revenue structure.\n    • The price of XPU systems is about 30% lower than GPUs, but efficiency is 30% higher, leading to roughly $27 billion per GW of revenue.\n    • Based on Ironwood TPU v6p (3nm, Google reference), the result is consistent at approximately $28 billion per GW.\n    • Including networking (30%) and rack components (30%), total revenue per GW is around $28 billion.\n    • If the 10GW contract is deployed over the next 3–4 years, it implies $70–90 billion in annual revenue potential between 2027–2029.\n    • Current market consensus estimates $60 billion in 2027 AI revenue, suggesting 50–70% upside potential.\n    • Including Google’s TPU revenue (~$20 billion), total AI revenue in 2027 could approach $100 billion.\n\n⸻\n\n3. NVIDIA\n    • Management has indicated that 1GW of AI data center capacity represents a $35–40 billion revenue opportunity, roughly 60% of total data center CapEx.\n    • Blackwell (BW) and Blackwell Ultra (BWU) are estimated at $27–29 billion per GW, while Rubin (R) and Rubin Ultra (RU) are projected at $30–35 billion per GW.\n    • Rack power density is expected to jump from 120kW (BW) to 1.1MW (RU) — about 9x higher. Although GPU count per rack decreases, ASP per GPU triples.\n    • GPU revenue per unit rises from approximately $52,000 (BW) to $300,000 (RU).\n    • The market currently assumes 10GW of AI data center capacity by 2027, but actual deployments could exceed that due to multiple large-scale projects.\n    • Consequently, AI revenue estimates for the next 2–3 years are likely to be revised upward.\n\n4. AMD\n    • The OpenAI contract is explicitly described as being worth several billion dollars, implying at least $20 billion per GW.\n    • The Helios platform is assumed to be priced about 20% lower than NVIDIA’s Rubin, with networking contributing roughly 5% of total revenue.\n    • Accordingly, the MI450 (Helios) system is estimated at $20 billion per GW, and the next-generation MI500 could exceed $25 billion per GW.\n    • If OpenAI’s 6GW deployment plan is fully realized, it represents $30–35 billion in annual revenue potential.\n    • The market currently assumes $31 billion in 2027 data center GPU revenue, suggesting room for upward revision.\n\n5. Comparative Summary\n\nAVGO: approximately $28 billion per GW\nNVDA: approximately $35 billion per GW (Rubin Ultra basis)\nAMD: approximately $20 billion per GW (Helios basis)\n\nBroadcom benefits from a broader scope including networking and rack components. NVIDIA’s growth is driven by rising GPU profitability per unit. AMD targets market share gains through aggressive pricing and major customer wins.\n\n6. Investment View\n    • All three companies have significant earnings upside potential for 2026–2028, driven by accelerating AI demand.\n    • NVIDIA leads in high-performance platforms, Broadcom in XPU and networking integration, and AMD in Helios-based large customer expansion.\n    • If long-term AI infrastructure CapEx continues to grow, all three are positioned for sustained re-rating and valuation expansion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010628907248126707, "ret_signed_m1d": 0.010628907248126707, "alpha_spy_m1d": 0.010613911103233287, "alpha_c_m1d": 0.005068471732688984, "bench_spy_m1d": 1.4996144893419938e-05, "bench_c_m1d": 0.005560435515437723, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03276899168911074, "ret_signed_p1d": -0.03276899168911074, "alpha_spy_p1d": -0.027570175726190027, "alpha_c_p1d": -0.013365625001438786, "bench_spy_p1d": -0.005198815962920711, "bench_c_p1d": -0.019403366687671952, "ret_p1w": 0.08393904587149836, "ret_signed_p1w": 0.08393904587149836, "alpha_spy_p1w": 0.06044685337659339, "alpha_c_p1w": 0.032360687171633096, "bench_spy_p1w": 0.023492192494904973, "bench_c_p1w": 0.05157835869986527, "ret_p1m": -0.06083265051379094, "ret_signed_p1m": -0.06083265051379094, "alpha_spy_p1m": -0.047932308309079685, "alpha_c_p1m": -0.04386197436412298, "bench_spy_p1m": -0.012900342204711257, "bench_c_p1m": -0.016970676149667963, "ret_p3m": -0.026047124228198104, "ret_signed_p3m": -0.026047124228198104, "alpha_spy_p3m": -0.05943631515326264, "alpha_c_p3m": -0.18919767833907275, "bench_spy_p3m": 0.03338919092506454, "bench_c_p3m": 0.16315055411087465, "ret_p6m": 0.08440111095626324, "ret_signed_p6m": 0.08440111095626324, "alpha_spy_p6m": 0.0357849808102082, "alpha_c_p6m": -0.231583758134134, "bench_spy_p6m": 0.04861613014605504, "bench_c_p6m": 0.31598486909039725, "price_path": [-0.3189, -0.3006, -0.2704, -0.2545, -0.2459, -0.2593, -0.2787, -0.2573, -0.2677, -0.3147, -0.2757, -0.2742, -0.2762, -0.265, -0.2252, -0.2398, -0.2543, -0.26, -0.3003, -0.306, -0.3122, -0.2952, -0.3137, -0.3, -0.2979, -0.2918, -0.3168, -0.3168, -0.3181, -0.3189, -0.3203, -0.365, -0.3639, -0.3454, -0.3297, -0.346, -0.3338, -0.3229, -0.3259, -0.3313, -0.3366, -0.3388, -0.3287, -0.324, -0.3241, -0.3225, -0.3301, -0.3221, -0.3203, -0.311, -0.2869, -0.3082, -0.1442, -0.1114, -0.0104, -0.0216, -0.0972, -0.0908, -0.0838, 0.0024, -0.0146, -0.0208, 0.0106, 0.0, -0.0328, -0.0128, 0.0626, 0.0909, 0.0839, 0.1105, 0.0706, 0.076, 0.0908, 0.0505, 0.0769, -0.0014, -0.0189, 0.025, -0.0021, 0.0876, 0.0417, 0.0369, 0.0105, -0.0325, -0.0608, -0.1345, -0.1439, -0.0965, -0.134, -0.0999, -0.0999, -0.0861, -0.0768, -0.0957, -0.0858, -0.0926, -0.0843, -0.0711, -0.0689, -0.0698, -0.0697, -0.1145, -0.1279, -0.1212, -0.1677, -0.1553, -0.1033, -0.097, -0.0972, -0.0966, -0.0966, -0.0968, -0.0942, -0.0953, -0.1003, -0.1003, -0.0612, -0.0712, -0.0995, -0.1177, -0.1401, -0.1465, -0.1275, -0.0717, -0.0606, -0.0425, -0.026, -0.026, -0.0257, 0.0494, 0.066, 0.091, 0.0558, 0.0588, 0.0618, 0.0594, -0.0055, 0.0346, 0.0171, -0.159, -0.1913, -0.1243, -0.0926, -0.1028, -0.1027, -0.1348, -0.129, -0.129, -0.1468, -0.1593, -0.1456, -0.1591, -0.1741, -0.1016, -0.1141, -0.1443, -0.1589, -0.1656, -0.1978, -0.1511, -0.1621, -0.1916, -0.1485, -0.1462, -0.1395, -0.1693, -0.1875, -0.1741, -0.1753, -0.162, -0.1376, -0.1542, -0.1485, -0.1372, -0.0746, -0.1439, -0.1514, -0.1764, -0.1454, -0.1169, -0.0862, -0.0862, -0.075, -0.0693, -0.0261, -0.0058, 0.0295, 0.037, 0.0716, 0.0844]}
{"unit_id": "thread:1995261108203266269", "ticker_idx": 0, "ts": "2025-11-30T22:38:13+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has resolved the gap with competitors, and its market share and margins will improve next year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung seen gaining HBM market share and margins in 2026 as Google's primary HBM supplier; SK Hynix growth constrained by Nvidia concentration and limited Google volume upside.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Succeeded in Turnaround... Supplied 'Over 60%' of HBM for Google\n\nIt has been confirmed that Samsung Electronics supplied \"more than 60%\" of the total High Bandwidth Memory (HBM) volume for Google through Broadcom this year. This is analyzed to be the result of boosting HBM3E (5th generation HBM) performance through the redesign of DRAM used in HBM and increasing supply volumes to key customers. Google develops its own AI accelerator called the TPU (Tensor Processing Unit) through Broadcom, which specializes in custom semiconductor design for clients.\n\nRegarding \"HBM4\" (6th generation HBM), the next-generation HBM for which the market opens next year, Samsung Electronics is reportedly expecting quality certification results from major customers as early as the beginning of this month. In the semiconductor industry, there is an analysis that \"Samsung has resolved the gap with competitors, and its market share and margins will improve next year.\"\n\nGoogle HBM 'Primary Supplier' is Samsung \nAccording to the semiconductor industry on the 1st, it was confirmed that Samsung Electronics accounted for more than 60% of the HBM volume supplied to Google through Broadcom so far this year. Although SK Hynix had a higher share of Broadcom/Google HBM in the first half of the year, it is known that Samsung Electronics succeeded in a \"reversal\" on an annual basis by increasing supply in the second half. Google, recently in the spotlight for its TPUs, develops semiconductors through Broadcom, a company specializing in custom semiconductors (ASICs). Broadcom also handles HBM procurement for TPUs.\n\nSamsung Electronics' expansion of HBM supply to major customers is also confirmed in export data. According to alternative data platform Aicel, the cumulative export amount of complex structure chip integrated circuits (including HBM) from Asan, Chungnam, where Samsung Electronics' HBM packaging line is located, from July to October this year was $8.2 billion (approximately 12 trillion won), a 19.2% increase compared to the same period last year ($6.9 billion).\n\nThe reason Samsung was able to increase supply to Broadcom/Google in the second half of this year is thanks to resolving heat issues through the redesign of 1a DRAM. The turning point was initiated by Vice Chairman Jun Young-hyun, who took office as the head of the Samsung Electronics Device Solutions (DS) Division in May last year. He ordered the redesign of 1a DRAM for HBM3E early this year, and the HBM development team, working with a \"do-or-die\" determination, solved the heat generation problem. A semiconductor industry official predicted, \"Although they passed Nvidia's HBM3E quality test in late September, Samsung Electronics' strategic customer was Broadcom,\" adding, \"Next year as well, Samsung Electronics will play the role of Google's 'primary supplier' through HBM delivery to Broadcom.\"\n\nSamsung Electronics, HBM4 Quality Certification Early This Month\nCompetition between Samsung Electronics and SK Hynix for HBM4 delivery is also igniting.\n\nSamsung Electronics was the first to achieve an operating speed of over 11 gigabits (Gb) per second through a winning move of developing and producing logic dies utilizing 10-nanometer (nm) 6th generation (1c) DRAM as the base die and a 4nm foundry process. At the \"International Solid-State Circuits Conference (ISSCC) 2026\" held on the 26th of last month, it unveiled HBM4 equipped with 36 gigabytes (GB) of capacity and a bandwidth (data processing capability) of 3.3 terabytes (TB) per second.\n\nSamsung Electronics is currently sending HBM4 samples to major customers such as Nvidia and undergoing quality tests. Observations inside and outside Samsung suggest that they will receive \"positive results\" early this month. It is known that Samsung Electronics has established a \"mass production system\" to increase supply volume as soon as it receives HBM4 quality certification. A semiconductor industry official explained, \"Samsung has virtually resolved the technology gap in HBM, and we expect improvements in market share and margins next year,\" adding, \"The proportion of HBM in total DRAM (based on bits) is about 15%, so there is significant room for expansion.\"\n\nSK Hynix, Dismisses HBM4 Redesign Rumors\nSK Hynix, the No. 1 HBM company, recently faced difficulties due to rumors that \"Nvidia requested a redesign of HBM4.\" SK Hynix plans to produce HBM4 through 10nm 5th generation (1b) DRAM and TSMC's 12nm foundry process. At a recent IR event held by a foreign securities firm, SK Hynix stated regarding the HBM4 redesign rumors, \"As we mentioned in the third-quarter conference call, both price and volume for next year have been finalized,\" and clarified, \"There are no issues regarding yield, redesign, or certification delays.\"\n\nThey also expressed confidence regarding the supply of large-scale HBM4 volumes to major customers. SK Hynix emphasized, \"We will begin full-scale wafer input in the fourth quarter of this year,\" and \"Meaningful production volume will emerge starting from the end of the second quarter of 2026.\"\n\nSK Hynix's main target is Nvidia, the world's No. 1 AI accelerator company. The prevailing analysis is that SK Hynix is highly likely to play the role of \"primary supplier\" for HBM4 as well, based on its long-standing cooperative relationship with Nvidia.\n\nRegarding HBM delivery for Google, they indicated that \"it is not easy to increase volume in the short term.\" It is known that SK Hynix's sales proportion for customers other than Nvidia, such as Google, is around 30%. SK Hynix stated, \"The major source of demand for HBM4 in 2026 is the main customer,\" and \"There is no significant change in the proportion of customers in 2026.\"\n\n---\n\nhttps://t.co/BcCTujAeCf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002976244827510355, "ret_signed_m1d": -0.002976244827510355, "alpha_spy_m1d": -0.007562599043156482, "alpha_c_m1d": -0.0025221594915277867, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.00045408533598256806, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02579357553797723, "ret_signed_p1d": 0.02579357553797723, "alpha_spy_p1d": 0.023941366263865227, "alpha_c_p1d": 0.015491534515764682, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01030204102221255, "ret_p1w": 0.08630946109646254, "ret_signed_p1w": 0.08630946109646254, "alpha_spy_p1w": 0.08137029652071615, "alpha_c_p1w": 0.05519376840312207, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.03111569269334047, "ret_p1m": 0.1952663534648169, "ret_signed_p1m": 0.1952663534648169, "alpha_spy_p1m": 0.18237433806856984, "alpha_c_p1m": 0.1781156283389469, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.017150725125870014, "ret_p3m": 1.1732115801334566, "ret_signed_p3m": 1.1732115801334566, "alpha_spy_p3m": 1.1569434132196874, "alpha_c_p3m": 1.1868391630443735, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": -0.013627582910916813, "ret_p6m": 1.9868586647386945, "ret_signed_p6m": 1.9868586647386945, "alpha_spy_p6m": 1.8772053852851878, "alpha_c_p6m": 1.6901842707486512, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.29667439399004336, "price_path": [-0.3106, -0.3077, -0.3136, -0.3077, -0.2938, -0.283, -0.2751, -0.2553, -0.2444, -0.2158, -0.2277, -0.2069, -0.2069, -0.1753, -0.1635, -0.1565, -0.1496, -0.1773, -0.1647, -0.1677, -0.1468, -0.1096, -0.1096, -0.1096, -0.1096, -0.1096, -0.1096, -0.0635, -0.0744, -0.0913, -0.0575, -0.0308, -0.0288, -0.0268, -0.0327, -0.0218, -0.0427, -0.0198, 0.0119, -0.0129, -0.003, 0.0327, 0.0665, 0.1022, 0.0407, -0.002, -0.0159, -0.0288, -0.002, 0.0268, 0.0228, 0.0198, -0.0357, -0.002, -0.0298, -0.0427, -0.002, -0.0595, -0.0407, -0.0149, 0.0198, 0.0268, -0.003, 0.0, 0.0258, 0.0367, 0.0427, 0.0754, 0.0863, 0.0754, 0.0714, 0.0645, 0.0804, 0.0397, 0.0198, 0.0704, 0.0675, 0.0546, 0.0962, 0.1062, 0.1022, 0.1022, 0.1607, 0.1913, 0.1953, 0.1953, 0.1953, 0.281, 0.3767, 0.3847, 0.4056, 0.3837, 0.3857, 0.3837, 0.3717, 0.3986, 0.4345, 0.4844, 0.4884, 0.4475, 0.4903, 0.5183, 0.5163, 0.5163, 0.59, 0.6189, 0.602, 0.6, 0.4993, 0.6698, 0.6857, 0.588, 0.5811, 0.6588, 0.6528, 0.6728, 0.7804, 0.8064, 0.8064, 0.8064, 0.8064, 0.8941, 0.8951, 0.924, 0.9938, 1.0287, 1.1732, 1.1583, 1.1583, 0.9449, 0.7166, 0.91, 0.8761, 0.7296, 0.8731, 0.8941, 0.8731, 0.8293, 0.8811, 0.933, 1.0785, 0.9988, 0.9878, 0.8572, 0.8911, 0.8841, 0.7954, 0.7954, 0.7611, 0.6702, 0.8945, 0.7821, 0.86, 0.929, 0.9629, 1.1028, 1.0379, 1.0578, 1.0079, 1.0628, 1.1078, 1.1727, 1.1577, 1.1427, 1.1877, 1.1727, 1.2426, 1.1927, 1.2426, 1.2177, 1.2576, 1.2027, 1.2027, 1.3226, 1.3226, 1.6572, 1.7121, 1.6822, 1.852, 1.7871, 1.837, 1.9569, 1.7022, 1.807, 1.7521, 1.7571, 1.9919, 1.9219, 1.9219, 1.9869]}
{"unit_id": "thread:1995261108203266269", "ticker_idx": 1, "ts": "2025-11-30T22:38:13+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's sales proportion for customers other than Nvidia, such as Google, is around 30%... There is no significant change in the proportion of customers in 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung seen gaining HBM market share and margins in 2026 as Google's primary HBM supplier; SK Hynix growth constrained by Nvidia concentration and limited Google volume upside.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Succeeded in Turnaround... Supplied 'Over 60%' of HBM for Google\n\nIt has been confirmed that Samsung Electronics supplied \"more than 60%\" of the total High Bandwidth Memory (HBM) volume for Google through Broadcom this year. This is analyzed to be the result of boosting HBM3E (5th generation HBM) performance through the redesign of DRAM used in HBM and increasing supply volumes to key customers. Google develops its own AI accelerator called the TPU (Tensor Processing Unit) through Broadcom, which specializes in custom semiconductor design for clients.\n\nRegarding \"HBM4\" (6th generation HBM), the next-generation HBM for which the market opens next year, Samsung Electronics is reportedly expecting quality certification results from major customers as early as the beginning of this month. In the semiconductor industry, there is an analysis that \"Samsung has resolved the gap with competitors, and its market share and margins will improve next year.\"\n\nGoogle HBM 'Primary Supplier' is Samsung \nAccording to the semiconductor industry on the 1st, it was confirmed that Samsung Electronics accounted for more than 60% of the HBM volume supplied to Google through Broadcom so far this year. Although SK Hynix had a higher share of Broadcom/Google HBM in the first half of the year, it is known that Samsung Electronics succeeded in a \"reversal\" on an annual basis by increasing supply in the second half. Google, recently in the spotlight for its TPUs, develops semiconductors through Broadcom, a company specializing in custom semiconductors (ASICs). Broadcom also handles HBM procurement for TPUs.\n\nSamsung Electronics' expansion of HBM supply to major customers is also confirmed in export data. According to alternative data platform Aicel, the cumulative export amount of complex structure chip integrated circuits (including HBM) from Asan, Chungnam, where Samsung Electronics' HBM packaging line is located, from July to October this year was $8.2 billion (approximately 12 trillion won), a 19.2% increase compared to the same period last year ($6.9 billion).\n\nThe reason Samsung was able to increase supply to Broadcom/Google in the second half of this year is thanks to resolving heat issues through the redesign of 1a DRAM. The turning point was initiated by Vice Chairman Jun Young-hyun, who took office as the head of the Samsung Electronics Device Solutions (DS) Division in May last year. He ordered the redesign of 1a DRAM for HBM3E early this year, and the HBM development team, working with a \"do-or-die\" determination, solved the heat generation problem. A semiconductor industry official predicted, \"Although they passed Nvidia's HBM3E quality test in late September, Samsung Electronics' strategic customer was Broadcom,\" adding, \"Next year as well, Samsung Electronics will play the role of Google's 'primary supplier' through HBM delivery to Broadcom.\"\n\nSamsung Electronics, HBM4 Quality Certification Early This Month\nCompetition between Samsung Electronics and SK Hynix for HBM4 delivery is also igniting.\n\nSamsung Electronics was the first to achieve an operating speed of over 11 gigabits (Gb) per second through a winning move of developing and producing logic dies utilizing 10-nanometer (nm) 6th generation (1c) DRAM as the base die and a 4nm foundry process. At the \"International Solid-State Circuits Conference (ISSCC) 2026\" held on the 26th of last month, it unveiled HBM4 equipped with 36 gigabytes (GB) of capacity and a bandwidth (data processing capability) of 3.3 terabytes (TB) per second.\n\nSamsung Electronics is currently sending HBM4 samples to major customers such as Nvidia and undergoing quality tests. Observations inside and outside Samsung suggest that they will receive \"positive results\" early this month. It is known that Samsung Electronics has established a \"mass production system\" to increase supply volume as soon as it receives HBM4 quality certification. A semiconductor industry official explained, \"Samsung has virtually resolved the technology gap in HBM, and we expect improvements in market share and margins next year,\" adding, \"The proportion of HBM in total DRAM (based on bits) is about 15%, so there is significant room for expansion.\"\n\nSK Hynix, Dismisses HBM4 Redesign Rumors\nSK Hynix, the No. 1 HBM company, recently faced difficulties due to rumors that \"Nvidia requested a redesign of HBM4.\" SK Hynix plans to produce HBM4 through 10nm 5th generation (1b) DRAM and TSMC's 12nm foundry process. At a recent IR event held by a foreign securities firm, SK Hynix stated regarding the HBM4 redesign rumors, \"As we mentioned in the third-quarter conference call, both price and volume for next year have been finalized,\" and clarified, \"There are no issues regarding yield, redesign, or certification delays.\"\n\nThey also expressed confidence regarding the supply of large-scale HBM4 volumes to major customers. SK Hynix emphasized, \"We will begin full-scale wafer input in the fourth quarter of this year,\" and \"Meaningful production volume will emerge starting from the end of the second quarter of 2026.\"\n\nSK Hynix's main target is Nvidia, the world's No. 1 AI accelerator company. The prevailing analysis is that SK Hynix is highly likely to play the role of \"primary supplier\" for HBM4 as well, based on its long-standing cooperative relationship with Nvidia.\n\nRegarding HBM delivery for Google, they indicated that \"it is not easy to increase volume in the short term.\" It is known that SK Hynix's sales proportion for customers other than Nvidia, such as Google, is around 30%. SK Hynix stated, \"The major source of demand for HBM4 in 2026 is the main customer,\" and \"There is no significant change in the proportion of customers in 2026.\"\n\n---\n\nhttps://t.co/BcCTujAeCf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01486978721660659, "ret_signed_m1d": 0.01486978721660659, "alpha_spy_m1d": 0.019456141432252716, "alpha_c_m1d": 0.012943233885823746, "bench_spy_m1d": 0.004586354215646127, "bench_c_m1d": -0.001926553330782843, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.037174817256039994, "ret_signed_p1d": -0.037174817256039994, "alpha_spy_p1d": -0.03532260798192799, "alpha_c_p1d": -0.018844156246077715, "bench_spy_p1d": 0.0018522092741120044, "bench_c_p1d": 0.01833066100996228, "ret_p1w": 0.07249076560395262, "ret_signed_p1w": -0.07249076560395262, "alpha_spy_p1w": -0.06755160102820623, "alpha_c_p1w": -0.028321497082765434, "bench_spy_p1w": 0.004939164575746391, "bench_c_p1w": 0.04416926852118719, "ret_p1m": 0.21003728679871325, "ret_signed_p1m": -0.21003728679871325, "alpha_spy_p1m": -0.19714527140246618, "alpha_c_p1m": -0.17751103357416698, "bench_spy_p1m": 0.01289201539624707, "bench_c_p1m": 0.032526253224546275, "ret_p3m": 1.0465204963316177, "ret_signed_p3m": -1.0465204963316177, "alpha_spy_p3m": -1.0302523294178485, "alpha_c_p3m": -0.8755889102672947, "bench_spy_p3m": 0.0162681669137692, "bench_c_p3m": 0.17093158606432302, "ret_p6m": 2.821164681653712, "ret_signed_p6m": -2.821164681653712, "alpha_spy_p6m": -2.7115114022002054, "alpha_c_p6m": -2.109884019355796, "bench_spy_p6m": 0.10965327945350678, "bench_c_p6m": 0.711280662297916, "price_path": [-0.5124, -0.5069, -0.492, -0.4855, -0.4651, -0.4353, -0.4298, -0.3898, -0.3852, -0.3536, -0.3806, -0.3397, -0.3397, -0.3481, -0.3295, -0.336, -0.3378, -0.375, -0.3518, -0.3546, -0.3313, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.2654, -0.205, -0.2292, -0.2357, -0.2152, -0.1595, -0.1354, -0.0982, -0.1103, -0.1057, -0.1112, -0.0527, -0.0063, -0.0323, 0.0364, 0.055, 0.0383, 0.1516, 0.0884, 0.0754, 0.1014, 0.0773, 0.1256, 0.1497, 0.146, 0.1367, 0.0401, 0.1256, 0.0587, 0.0439, 0.0606, -0.0323, -0.0341, -0.036, -0.0267, 0.0112, -0.0149, 0.0, 0.0372, 0.026, 0.0074, 0.0112, 0.0725, 0.052, 0.0911, 0.0502, 0.0613, 0.0297, -0.0149, 0.0242, 0.026, 0.0167, 0.0781, 0.0855, 0.0929, 0.0929, 0.1134, 0.1896, 0.21, 0.21, 0.21, 0.2584, 0.2937, 0.3494, 0.3792, 0.4052, 0.3829, 0.3922, 0.3717, 0.3792, 0.3922, 0.4052, 0.4201, 0.381, 0.3755, 0.4033, 0.4257, 0.368, 0.487, 0.5632, 0.6004, 0.6896, 0.5428, 0.6859, 0.6729, 0.5651, 0.5595, 0.6487, 0.6283, 0.5985, 0.6506, 0.6357, 0.6357, 0.6357, 0.6357, 0.6617, 0.7639, 0.7677, 0.868, 0.8922, 1.0465, 0.9758, 0.9758, 0.7486, 0.581, 0.7523, 0.7206, 0.5568, 0.7467, 0.7784, 0.7318, 0.6946, 0.8137, 0.8063, 0.9664, 0.8864, 0.8752, 0.7374, 0.8361, 0.8529, 0.7374, 0.7374, 0.6257, 0.5028, 0.6704, 0.5456, 0.6313, 0.6499, 0.7057, 0.9236, 0.8584, 0.9124, 0.9367, 1.054, 1.1154, 1.1508, 1.1005, 1.1713, 1.2793, 1.2774, 1.2812, 1.2756, 1.4059, 1.4208, 1.4078, 1.3947, 1.3947, 1.6946, 1.6946, 1.9813, 2.08, 2.1396, 2.5009, 2.4171, 2.6796, 2.6685, 2.3873, 2.4264, 2.2495, 2.2495, 2.6126, 2.6145, 2.6145, 2.8212]}
{"unit_id": "orig:1992749276611641671", "ticker_idx": 0, "ts": "2025-11-24T00:17:05+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TP raised from 770,000 won → 830,000 won", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi raises SK Hynix TP to W830k on multi-year demand visibility, structural memory upcycle, and sharply raised DRAM/NAND ASP forecasts for 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251124_Multi-year Demand Visibility Supports Structural Growth Outlook; Raise TP to W830k – CITI\n\n[Key Takeaways]\n\n(1) Multi-year demand visibility supports SK Hynix’s structural growth. TP raised from 770,000 won → 830,000 won.\n\n(2) Despite extremely strong demand for server DRAM and eSSD for AI inference, the memory market’s supply shortage is expected to deepen as suppliers focus CAPA expansion almost entirely on HBM.\n\n(3) Global DRAM/NAND ASP growth forecasts are raised from YoY +37%/+39% to +53%/+44%. As a result, SK Hynix’s 2026 OP is revised up by +12%.\n\n⸻\n\n[Contents]\n\n(1) We expect that in 1H26, global commodity memory supply will fall short of total demand based on *FPO, and will be able to meet only about 60% of total demand.\n*Firm Purchase Order\n\n(2) The remaining 40% of customers will not receive supply, and we believe they will move to secure memory chips—paying higher premiums if necessary—to avoid supply disruptions.\n\n(3) Moreover, since next year’s capacity additions are almost entirely focused on HBM, global DRAM ASP in 2026 is expected to increase YoY +53% [previous +37%], and NAND ASP to increase YoY +44% [previous +39%].\n\n(4) As highlighted at Citi Korea Corporate Day, we believe SK Hynix’s demand visibility has expanded to as much as 2–3 years, compared to the 2017–18 memory upcycle.\n\n(5) Even for commodity DRAM, major customers have already been allocated CAPA based on FPOs, and some customers have even communicated purchase intentions for 2028 volumes in advance.\n\n(6) With server DRAM/HBM accounting for 80% as of end-3Q25, the volatility that used to arise from consumer-facing memory has also been significantly reduced.\n\n(7) We believe the memory market has entered the early stage of a structural upcycle driven by the explosive increase in data/tokens generated by AI inference demand.\n\n(8) Reflecting our DRAM/NAND outlook, we revise up SK Hynix’s OP forecast by +12%, and apply 3.3x to 2026(F) BVPS, raising our TP to 830,000 won.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.001923085035032246, "ret_signed_m1d": 0.001923085035032246, "alpha_spy_m1d": 0.016428099932999873, "alpha_c_m1d": 0.040228024980765276, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001923085035032246, "ret_signed_p1d": -0.001923085035032246, "alpha_spy_p1d": -0.011328984687139165, "alpha_c_p1d": -0.004547610054137952, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.03535624949908667, "ret_signed_p1w": 0.03535624949908667, "alpha_spy_p1w": 0.018099555510529663, "alpha_c_p1w": -0.0054553221701976184, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.12388131384195789, "ret_signed_p1m": 0.12388131384195789, "alpha_spy_p1m": 0.09208536730446304, "alpha_c_p1m": 0.049659721025902215, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.720462027311146, "ret_signed_p3m": 0.720462027311146, "alpha_spy_p3m": 0.6938854408526784, "alpha_c_p3m": 0.5070968272816094, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.3643692048480567, "ret_signed_p6m": 2.3643692048480567, "alpha_spy_p6m": 2.260922574874728, "alpha_c_p6m": 1.7553439946396165, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.5007, -0.4837, -0.4827, -0.5077, -0.499, -0.4952, -0.4894, -0.474, -0.4673, -0.4462, -0.4154, -0.4096, -0.3683, -0.3635, -0.3308, -0.3587, -0.3163, -0.3163, -0.325, -0.3058, -0.3125, -0.3144, -0.3529, -0.3288, -0.3317, -0.3077, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.2394, -0.1769, -0.2019, -0.2087, -0.1875, -0.1298, -0.1048, -0.0663, -0.0788, -0.074, -0.0798, -0.0192, 0.0288, 0.0019, 0.0731, 0.0923, 0.075, 0.1923, 0.1269, 0.1135, 0.1404, 0.1154, 0.1654, 0.1904, 0.1865, 0.1769, 0.0769, 0.1654, 0.0962, 0.0808, 0.0981, 0.0019, 0.0, -0.0019, 0.0077, 0.0469, 0.02, 0.0354, 0.0738, 0.0623, 0.0431, 0.0469, 0.1104, 0.0892, 0.1297, 0.0873, 0.0989, 0.0661, 0.02, 0.0604, 0.0623, 0.0527, 0.1162, 0.1239, 0.1316, 0.1316, 0.1527, 0.2317, 0.2528, 0.2528, 0.2528, 0.3029, 0.3394, 0.3972, 0.4279, 0.4549, 0.4318, 0.4414, 0.4202, 0.4279, 0.4414, 0.4549, 0.4703, 0.4299, 0.4241, 0.453, 0.4761, 0.4164, 0.5396, 0.6185, 0.657, 0.7493, 0.5973, 0.7455, 0.732, 0.6204, 0.6146, 0.707, 0.6858, 0.655, 0.7089, 0.6935, 0.6935, 0.6935, 0.6935, 0.7205, 0.8263, 0.8302, 0.9341, 0.9591, 1.1189, 1.0456, 1.0456, 0.8104, 0.6369, 0.8143, 0.7815, 0.6118, 0.8085, 0.8412, 0.793, 0.7545, 0.8779, 0.8702, 1.036, 0.9531, 0.9415, 0.7988, 0.901, 0.9184, 0.7988, 0.7988, 0.6831, 0.5559, 0.7294, 0.6002, 0.6889, 0.7082, 0.7661, 0.9916, 0.9241, 0.9801, 1.0051, 1.1266, 1.1902, 1.2268, 1.1748, 1.2481, 1.3599, 1.358, 1.3618, 1.356, 1.491, 1.5064, 1.4929, 1.4794, 1.4794, 1.7898, 1.7898, 2.0867, 2.1889, 2.2506, 2.6246, 2.5379, 2.8097, 2.7982, 2.507, 2.5475, 2.3644]}
{"unit_id": "orig:1992749276611641671", "ticker_idx": 1, "ts": "2025-11-24T00:17:05+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "global DRAM ASP in 2026 is expected to increase YoY +53%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi raises SK Hynix TP to W830k on multi-year demand visibility, structural memory upcycle, and sharply raised DRAM/NAND ASP forecasts for 2026.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251124_Multi-year Demand Visibility Supports Structural Growth Outlook; Raise TP to W830k – CITI\n\n[Key Takeaways]\n\n(1) Multi-year demand visibility supports SK Hynix’s structural growth. TP raised from 770,000 won → 830,000 won.\n\n(2) Despite extremely strong demand for server DRAM and eSSD for AI inference, the memory market’s supply shortage is expected to deepen as suppliers focus CAPA expansion almost entirely on HBM.\n\n(3) Global DRAM/NAND ASP growth forecasts are raised from YoY +37%/+39% to +53%/+44%. As a result, SK Hynix’s 2026 OP is revised up by +12%.\n\n⸻\n\n[Contents]\n\n(1) We expect that in 1H26, global commodity memory supply will fall short of total demand based on *FPO, and will be able to meet only about 60% of total demand.\n*Firm Purchase Order\n\n(2) The remaining 40% of customers will not receive supply, and we believe they will move to secure memory chips—paying higher premiums if necessary—to avoid supply disruptions.\n\n(3) Moreover, since next year’s capacity additions are almost entirely focused on HBM, global DRAM ASP in 2026 is expected to increase YoY +53% [previous +37%], and NAND ASP to increase YoY +44% [previous +39%].\n\n(4) As highlighted at Citi Korea Corporate Day, we believe SK Hynix’s demand visibility has expanded to as much as 2–3 years, compared to the 2017–18 memory upcycle.\n\n(5) Even for commodity DRAM, major customers have already been allocated CAPA based on FPOs, and some customers have even communicated purchase intentions for 2028 volumes in advance.\n\n(6) With server DRAM/HBM accounting for 80% as of end-3Q25, the volatility that used to arise from consumer-facing memory has also been significantly reduced.\n\n(7) We believe the memory market has entered the early stage of a structural upcycle driven by the explosive increase in data/tokens generated by AI inference demand.\n\n(8) Reflecting our DRAM/NAND outlook, we revise up SK Hynix’s OP forecast by +12%, and apply 3.3x to 2026(F) BVPS, raising our TP to 830,000 won.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030559001381946265, "ret_signed_m1d": -0.030559001381946265, "alpha_spy_m1d": -0.016053986483978638, "alpha_c_m1d": 0.007745938563786765, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009214553834632285, "ret_signed_p1d": 0.009214553834632285, "alpha_spy_p1d": -0.00019134581747463388, "alpha_c_p1d": 0.00659002881552658, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": 0.05052439497840333, "ret_signed_p1w": 0.05052439497840333, "alpha_spy_p1w": 0.03326770098984633, "alpha_c_p1w": 0.009712823309119045, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.17022187098231645, "ret_signed_p1m": 0.17022187098231645, "alpha_spy_p1m": 0.1384259244448216, "alpha_c_p1m": 0.09600027816626078, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 0.8531193694574041, "ret_signed_p3m": 0.8531193694574041, "alpha_spy_p3m": 0.8265427829989366, "alpha_c_p3m": 0.6397541694278676, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 2.118694645460851, "ret_signed_p6m": 2.118694645460851, "alpha_spy_p6m": 2.0152480154875225, "alpha_c_p6m": 1.509669435252411, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.4161, -0.4075, -0.4113, -0.4269, -0.4196, -0.4156, -0.4045, -0.3907, -0.3863, -0.3688, -0.3477, -0.3273, -0.2968, -0.2906, -0.2682, -0.2798, -0.2453, -0.2545, -0.2436, -0.2304, -0.2372, -0.2427, -0.2645, -0.2422, -0.2391, -0.2018, -0.1637, -0.1575, -0.1528, -0.1607, -0.1445, -0.1508, -0.1299, -0.1254, -0.142, -0.116, -0.0717, -0.0629, -0.0428, -0.0557, -0.056, -0.0529, -0.0065, 0.0222, 0.0073, 0.0415, 0.0564, 0.062, 0.1298, 0.0618, 0.0715, 0.0768, 0.0634, 0.1123, 0.1125, 0.1155, 0.0994, 0.0615, 0.0954, 0.0426, 0.0292, 0.0126, -0.0306, 0.0, 0.0092, 0.033, 0.0485, 0.0384, 0.0505, 0.0709, 0.0629, 0.0474, 0.0757, 0.1151, 0.1125, 0.1414, 0.117, 0.1006, 0.0702, 0.0405, 0.0611, 0.095, 0.1132, 0.1647, 0.1702, 0.1869, 0.1869, 0.2115, 0.2628, 0.2687, 0.2579, 0.2579, 0.3491, 0.3897, 0.4583, 0.47, 0.4527, 0.4726, 0.4763, 0.4536, 0.4584, 0.4802, 0.5409, 0.5474, 0.5231, 0.572, 0.6039, 0.614, 0.5784, 0.6766, 0.7502, 0.7579, 0.7569, 0.7053, 0.7866, 0.7281, 0.6621, 0.6753, 0.7165, 0.6921, 0.744, 0.8047, 0.8052, 0.8052, 0.7875, 0.819, 0.8531, 0.9049, 0.9055, 0.9599, 0.9968, 1.0802, 1.0459, 1.0463, 0.8447, 0.7389, 0.8597, 0.7972, 0.718, 0.854, 0.8954, 0.8522, 0.855, 0.9375, 0.9826, 1.0885, 1.0071, 0.9676, 0.8471, 0.8798, 0.8631, 0.7528, 0.7554, 0.6524, 0.6023, 0.7828, 0.6983, 0.7549, 0.8024, 0.8333, 1.0005, 0.9774, 1.0016, 1.0015, 1.1194, 1.1421, 1.1784, 1.1526, 1.1619, 1.2163, 1.2672, 1.2842, 1.2872, 1.3911, 1.3573, 1.3878, 1.3623, 1.3996, 1.5957, 1.6907, 1.9453, 1.9687, 2.1281, 2.3841, 2.2898, 2.4529, 2.4496, 2.1875, 2.1732, 2.1187]}
{"unit_id": "orig:1992749276611641671", "ticker_idx": 2, "ts": "2025-11-24T00:17:05+00:00", "symbol": "NAND", "kind": "sector", "geography": "global", "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND ASP to increase YoY +44%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Citi raises SK Hynix TP to W830k on multi-year demand visibility, structural memory upcycle, and sharply raised DRAM/NAND ASP forecasts for 2026.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "Global NAND sector basket: dominant NAND producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "251124_Multi-year Demand Visibility Supports Structural Growth Outlook; Raise TP to W830k – CITI\n\n[Key Takeaways]\n\n(1) Multi-year demand visibility supports SK Hynix’s structural growth. TP raised from 770,000 won → 830,000 won.\n\n(2) Despite extremely strong demand for server DRAM and eSSD for AI inference, the memory market’s supply shortage is expected to deepen as suppliers focus CAPA expansion almost entirely on HBM.\n\n(3) Global DRAM/NAND ASP growth forecasts are raised from YoY +37%/+39% to +53%/+44%. As a result, SK Hynix’s 2026 OP is revised up by +12%.\n\n⸻\n\n[Contents]\n\n(1) We expect that in 1H26, global commodity memory supply will fall short of total demand based on *FPO, and will be able to meet only about 60% of total demand.\n*Firm Purchase Order\n\n(2) The remaining 40% of customers will not receive supply, and we believe they will move to secure memory chips—paying higher premiums if necessary—to avoid supply disruptions.\n\n(3) Moreover, since next year’s capacity additions are almost entirely focused on HBM, global DRAM ASP in 2026 is expected to increase YoY +53% [previous +37%], and NAND ASP to increase YoY +44% [previous +39%].\n\n(4) As highlighted at Citi Korea Corporate Day, we believe SK Hynix’s demand visibility has expanded to as much as 2–3 years, compared to the 2017–18 memory upcycle.\n\n(5) Even for commodity DRAM, major customers have already been allocated CAPA based on FPOs, and some customers have even communicated purchase intentions for 2028 volumes in advance.\n\n(6) With server DRAM/HBM accounting for 80% as of end-3Q25, the volatility that used to arise from consumer-facing memory has also been significantly reduced.\n\n(7) We believe the memory market has entered the early stage of a structural upcycle driven by the explosive increase in data/tokens generated by AI inference demand.\n\n(8) Reflecting our DRAM/NAND outlook, we revise up SK Hynix’s OP forecast by +12%, and apply 3.3x to 2026(F) BVPS, raising our TP to 830,000 won.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04185496520339238, "ret_signed_m1d": -0.04185496520339238, "alpha_spy_m1d": -0.027349950305424753, "alpha_c_m1d": -0.0035500252576593505, "bench_spy_m1d": -0.014505014897967627, "bench_c_m1d": -0.03830493994573303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0036933182749673056, "ret_signed_p1d": -0.0036933182749673056, "alpha_spy_p1d": -0.013099217927074224, "alpha_c_p1d": -0.006317843294073011, "bench_spy_p1d": 0.00940589965210692, "bench_c_p1d": 0.0026245250191057057, "ret_p1w": -0.007372254815862034, "ret_signed_p1w": -0.007372254815862034, "alpha_spy_p1w": -0.024628948804419036, "alpha_c_p1w": -0.04818382648514632, "bench_spy_p1w": 0.017256693988557004, "bench_c_p1w": 0.040811571669284286, "ret_p1m": 0.11638672994838206, "ret_signed_p1m": 0.11638672994838206, "alpha_spy_p1m": 0.08459078341088722, "alpha_c_p1m": 0.04216513713232639, "bench_spy_p1m": 0.031795946537494846, "bench_c_p1m": 0.07422159281605567, "ret_p3m": 1.0829127444358775, "ret_signed_p3m": 1.0829127444358775, "alpha_spy_p3m": 1.05633615797741, "alpha_c_p3m": 0.869547544406341, "bench_spy_p3m": 0.0265765864584675, "bench_c_p3m": 0.21336520002953652, "ret_p6m": 3.082612023612646, "ret_signed_p6m": 3.082612023612646, "alpha_spy_p6m": 2.9791653936393176, "alpha_c_p6m": 2.4735868134042063, "bench_spy_p6m": 0.10344662997332854, "bench_c_p6m": 0.6090252102084401, "price_path": [-0.5588, -0.5504, -0.5488, -0.5565, -0.5546, -0.5503, -0.5352, -0.5129, -0.5069, -0.4959, -0.474, -0.4422, -0.4136, -0.4064, -0.3864, -0.3966, -0.3699, -0.3723, -0.3593, -0.3483, -0.3556, -0.365, -0.3855, -0.3521, -0.3474, -0.3205, -0.281, -0.2581, -0.2589, -0.2722, -0.2533, -0.2522, -0.2521, -0.2338, -0.2534, -0.2164, -0.1787, -0.1834, -0.1527, -0.1615, -0.1584, -0.1376, -0.0648, -0.0356, -0.0539, 0.006, 0.0237, 0.0287, 0.0761, 0.0231, 0.0439, 0.0584, 0.0892, 0.1691, 0.1699, 0.1828, 0.134, 0.0607, 0.1142, 0.046, 0.0503, 0.0062, -0.0419, 0.0, -0.0037, -0.0235, -0.001, 0.0074, -0.0074, 0.0073, -0.0113, -0.0028, 0.0348, 0.0702, 0.0578, 0.0789, 0.0748, 0.0393, 0.0034, -0.018, 0.0044, 0.04, 0.0635, 0.1131, 0.1164, 0.144, 0.1478, 0.1749, 0.1858, 0.181, 0.172, 0.172, 0.2601, 0.3016, 0.4144, 0.4466, 0.4257, 0.4692, 0.4819, 0.4885, 0.4816, 0.5205, 0.5832, 0.5962, 0.6164, 0.714, 0.7631, 0.7316, 0.7099, 0.7981, 0.8929, 0.9137, 0.9878, 0.9755, 1.0993, 0.9687, 0.8944, 0.9129, 0.9209, 0.8683, 0.9503, 1.0605, 1.0908, 1.0799, 1.039, 1.0476, 1.0829, 1.1256, 1.1406, 1.1827, 1.1833, 1.2461, 1.2103, 1.2023, 1.0078, 0.955, 1.019, 0.9413, 0.9091, 1.048, 1.1411, 1.0802, 1.1161, 1.2335, 1.2551, 1.3836, 1.3305, 1.2518, 1.1553, 1.1676, 1.1628, 1.0052, 1.0001, 0.8951, 0.9017, 1.1148, 1.0578, 1.1069, 1.1747, 1.1903, 1.4388, 1.4892, 1.5526, 1.6638, 1.8018, 1.7173, 1.7927, 1.7119, 1.7101, 1.7788, 1.917, 1.8977, 1.9342, 2.085, 2.0219, 2.0947, 2.1326, 2.2118, 2.3901, 2.5797, 2.7357, 2.8276, 3.1407, 3.3102, 3.1726, 3.354, 3.2545, 3.0393, 3.1045, 3.0826]}
{"unit_id": "orig:2000701627007996303", "ticker_idx": 0, "ts": "2025-12-15T22:56:53+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "some suppliers see potential for conventional DRAM prices to rise by double-digit percentages (%) quarter-over-quarter (QoQ) throughout every quarter of 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman sees DRAM/HBM demand significantly outpacing supply through 2026, with conventional DRAM prices potentially rising double-digits QoQ every quarter and HBM margins expanding.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs DRAM Commentary: Supply Growth Modest, Demand Continues to Significantly Outpace Supply\n\nDemand for both High Bandwidth Memory (HBM) and conventional DRAM continues to outstrip supply. Overall, full-year 2026 HBM volume and pricing negotiations for most DRAM suppliers appear to be effectively concluded. HBM4 shipments are projected to commence in 2Q26, with a meaningful ramp-up expected in the second half of the year. Due to anticipated price declines in HBM3E, the blended HBM ASP (inclusive of HBM3 and HBM4) is expected to dip in 1H26 before rebounding in 2H26. Most suppliers anticipate continued margin expansion throughout 2026.\n\nBroadly, industry-wide capacity expansion remains constrained by a shortage of available cleanroom space. Suppliers are focused on bringing in Wafer Fab Equipment (WFE) as quickly as possible to fabs (such as SK Hynix’s M15X) to boost HBM production. At this stage, there appear to be no plans among vendors to convert HBM production lines to conventional DRAM or to repurpose NAND lines for DRAM.\n\nConventional DRAM supply-demand dynamics appear even tighter, particularly in the near term. Volume forecasts for conventional DRAM continue to be provided by customers, while pricing negotiations remain on a quarterly basis. Given current demand levels and persistent supply-demand imbalances, some suppliers see potential for conventional DRAM prices to rise by double-digit percentages (%) quarter-over-quarter (QoQ) throughout every quarter of 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.028378076628599407, "ret_signed_m1d": 0.028378076628599407, "alpha_spy_m1d": 0.02686499278157199, "alpha_c_m1d": 0.024921002664843355, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02780528361918592, "ret_signed_p1d": -0.02780528361918592, "alpha_spy_p1d": -0.02507298602390721, "alpha_c_p1d": -0.025084940370160996, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.08863671235432598, "ret_signed_p1w": 0.08863671235432598, "alpha_spy_p1w": 0.07964098704162355, "alpha_c_p1w": 0.06619622268213161, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3585899755552539, "ret_signed_p1m": 0.3585899755552539, "alpha_spy_p1m": 0.33642248564954297, "alpha_c_p1m": 0.24567335034577392, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.7302929518690062, "ret_signed_p3m": 0.7302929518690062, "alpha_spy_p3m": 0.7489521251336984, "alpha_c_p3m": 0.6270407450816284, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3275, -0.2951, -0.3037, -0.2937, -0.2813, -0.2878, -0.293, -0.3133, -0.2925, -0.2895, -0.2545, -0.219, -0.2131, -0.2087, -0.2161, -0.2009, -0.2068, -0.1875, -0.1832, -0.1987, -0.1744, -0.1329, -0.1247, -0.1059, -0.118, -0.1183, -0.1152, -0.0717, -0.045, -0.0588, -0.0268, -0.013, -0.0079, 0.0556, -0.008, 0.0014, 0.0066, -0.006, 0.0398, 0.0398, 0.0426, 0.0275, -0.0077, 0.0239, -0.0255, -0.0381, -0.0541, -0.0942, -0.0655, -0.057, -0.0348, -0.0203, -0.0295, -0.0182, 0.0007, -0.0068, -0.0215, 0.005, 0.042, 0.0396, 0.0668, 0.044, 0.0284, 0.0, -0.0278, -0.0088, 0.0232, 0.0404, 0.0886, 0.0938, 0.1095, 0.1095, 0.1323, 0.1803, 0.1858, 0.1756, 0.1756, 0.2609, 0.2984, 0.363, 0.3739, 0.3577, 0.3764, 0.3799, 0.3586, 0.3629, 0.3832, 0.4401, 0.4462, 0.4236, 0.4694, 0.4993, 0.5088, 0.4753, 0.5671, 0.6362, 0.6435, 0.6424, 0.5948, 0.67, 0.6147, 0.5534, 0.5659, 0.604, 0.5811, 0.6299, 0.6864, 0.6866, 0.6866, 0.6699, 0.6997, 0.7312, 0.7798, 0.7801, 0.8308, 0.8653, 0.9427, 0.9105, 0.9109, 0.7227, 0.6249, 0.7371, 0.6784, 0.6051, 0.732, 0.7708, 0.7303, 0.7334, 0.8106, 0.8527, 0.9512, 0.8752, 0.8381, 0.7256, 0.756, 0.7403, 0.6371, 0.6396, 0.5429, 0.4965, 0.6648, 0.5861, 0.6388, 0.6832, 0.7119, 0.8682, 0.847, 0.8696, 0.8698, 0.9803, 1.0014, 1.0351, 1.011, 1.0197, 1.0704, 1.1186, 1.1341, 1.1373, 1.2347, 1.203, 1.2315, 1.2078, 1.243, 1.4266, 1.5161, 1.7533, 1.7749, 1.9253, 2.1648, 2.0765, 2.2295, 2.2256, 1.9807, 1.9664, 1.9157, 1.964, 2.1991, 2.1617, 2.1617, 2.453, 2.6394, 2.6365, 2.7858, null, null, null, null, null, null, null]}
{"unit_id": "orig:2000701627007996303", "ticker_idx": 1, "ts": "2025-12-15T22:56:53+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Suppliers are focused on bringing in Wafer Fab Equipment (WFE) as quickly as possible to fabs (such as SK Hynix's M15X) to boost HBM production", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman sees DRAM/HBM demand significantly outpacing supply through 2026, with conventional DRAM prices potentially rising double-digits QoQ every quarter and HBM margins expanding.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs DRAM Commentary: Supply Growth Modest, Demand Continues to Significantly Outpace Supply\n\nDemand for both High Bandwidth Memory (HBM) and conventional DRAM continues to outstrip supply. Overall, full-year 2026 HBM volume and pricing negotiations for most DRAM suppliers appear to be effectively concluded. HBM4 shipments are projected to commence in 2Q26, with a meaningful ramp-up expected in the second half of the year. Due to anticipated price declines in HBM3E, the blended HBM ASP (inclusive of HBM3 and HBM4) is expected to dip in 1H26 before rebounding in 2H26. Most suppliers anticipate continued margin expansion throughout 2026.\n\nBroadly, industry-wide capacity expansion remains constrained by a shortage of available cleanroom space. Suppliers are focused on bringing in Wafer Fab Equipment (WFE) as quickly as possible to fabs (such as SK Hynix’s M15X) to boost HBM production. At this stage, there appear to be no plans among vendors to convert HBM production lines to conventional DRAM or to repurpose NAND lines for DRAM.\n\nConventional DRAM supply-demand dynamics appear even tighter, particularly in the near term. Volume forecasts for conventional DRAM continue to be provided by customers, while pricing negotiations remain on a quarterly basis. Given current demand levels and persistent supply-demand imbalances, some suppliers see potential for conventional DRAM prices to rise by double-digit percentages (%) quarter-over-quarter (QoQ) throughout every quarter of 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.030685848548443806, "ret_signed_m1d": 0.030685848548443806, "alpha_spy_m1d": 0.029172764701416387, "alpha_c_m1d": 0.027228774584687754, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.043321337592038156, "ret_signed_p1d": -0.043321337592038156, "alpha_spy_p1d": -0.040589039996759446, "alpha_c_p1d": -0.04060099434301323, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.046931364275827114, "ret_signed_p1w": 0.046931364275827114, "alpha_spy_p1w": 0.037935638963124685, "alpha_c_p1w": 0.02449087460363275, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3321299157434354, "ret_signed_p1m": 0.3321299157434354, "alpha_spy_p1m": 0.3099624258377245, "alpha_c_p1m": 0.21921329053395544, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.6817978803316003, "ret_signed_p3m": 0.6817978803316003, "alpha_spy_p3m": 0.7004570535962924, "alpha_c_p3m": 0.5785456735442225, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3984, -0.3588, -0.3588, -0.3669, -0.3488, -0.3552, -0.357, -0.393, -0.3705, -0.3732, -0.3506, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.2866, -0.228, -0.2514, -0.2578, -0.2379, -0.1838, -0.1603, -0.1243, -0.136, -0.1315, -0.1369, -0.0801, -0.035, -0.0602, 0.0065, 0.0245, 0.0083, 0.1183, 0.057, 0.0444, 0.0696, 0.0462, 0.0931, 0.1165, 0.1129, 0.1039, 0.0101, 0.0931, 0.0281, 0.0137, 0.0299, -0.0602, -0.062, -0.0638, -0.0548, -0.0181, -0.0433, -0.0289, 0.0072, -0.0036, -0.0217, -0.0181, 0.0415, 0.0217, 0.0596, 0.0199, 0.0307, 0.0, -0.0433, -0.0054, -0.0036, -0.0126, 0.0469, 0.0542, 0.0614, 0.0614, 0.0812, 0.1552, 0.1751, 0.1751, 0.1751, 0.222, 0.2563, 0.3105, 0.3394, 0.3646, 0.343, 0.352, 0.3321, 0.3394, 0.352, 0.3646, 0.3791, 0.3412, 0.3357, 0.3628, 0.3845, 0.3285, 0.444, 0.5181, 0.5542, 0.6408, 0.4982, 0.6372, 0.6245, 0.5199, 0.5144, 0.6011, 0.5812, 0.5523, 0.6029, 0.5884, 0.5884, 0.5884, 0.5884, 0.6137, 0.713, 0.7166, 0.8141, 0.8375, 0.9874, 0.9187, 0.9187, 0.6981, 0.5353, 0.7017, 0.6709, 0.5118, 0.6963, 0.727, 0.6818, 0.6456, 0.7614, 0.7541, 0.9097, 0.8319, 0.821, 0.6872, 0.7831, 0.7993, 0.6872, 0.6872, 0.5787, 0.4594, 0.6221, 0.501, 0.5841, 0.6022, 0.6565, 0.8681, 0.8048, 0.8572, 0.8807, 0.9946, 1.0543, 1.0887, 1.0399, 1.1086, 1.2135, 1.2117, 1.2153, 1.2098, 1.3364, 1.3509, 1.3382, 1.3256, 1.3256, 1.6167, 1.6167, 1.8952, 1.9911, 2.0489, 2.3998, 2.3184, 2.5734, 2.5625, 2.2895, 2.3274, 2.1556, 2.1556, 2.5083, 2.5101, 2.5101, 2.7108, 3.0562, 3.1401, 3.2197, null, null, null, null, null, null, null]}
{"unit_id": "orig:2000701627007996303", "ticker_idx": 2, "ts": "2025-12-15T22:56:53+00:00", "symbol": "HBM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Most suppliers anticipate continued margin expansion throughout 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman sees DRAM/HBM demand significantly outpacing supply through 2026, with conventional DRAM prices potentially rising double-digits QoQ every quarter and HBM margins expanding.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs DRAM Commentary: Supply Growth Modest, Demand Continues to Significantly Outpace Supply\n\nDemand for both High Bandwidth Memory (HBM) and conventional DRAM continues to outstrip supply. Overall, full-year 2026 HBM volume and pricing negotiations for most DRAM suppliers appear to be effectively concluded. HBM4 shipments are projected to commence in 2Q26, with a meaningful ramp-up expected in the second half of the year. Due to anticipated price declines in HBM3E, the blended HBM ASP (inclusive of HBM3 and HBM4) is expected to dip in 1H26 before rebounding in 2H26. Most suppliers anticipate continued margin expansion throughout 2026.\n\nBroadly, industry-wide capacity expansion remains constrained by a shortage of available cleanroom space. Suppliers are focused on bringing in Wafer Fab Equipment (WFE) as quickly as possible to fabs (such as SK Hynix’s M15X) to boost HBM production. At this stage, there appear to be no plans among vendors to convert HBM production lines to conventional DRAM or to repurpose NAND lines for DRAM.\n\nConventional DRAM supply-demand dynamics appear even tighter, particularly in the near term. Volume forecasts for conventional DRAM continue to be provided by customers, while pricing negotiations remain on a quarterly basis. Given current demand levels and persistent supply-demand imbalances, some suppliers see potential for conventional DRAM prices to rise by double-digit percentages (%) quarter-over-quarter (QoQ) throughout every quarter of 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.028378076628599407, "ret_signed_m1d": 0.028378076628599407, "alpha_spy_m1d": 0.02686499278157199, "alpha_c_m1d": 0.024921002664843355, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02780528361918592, "ret_signed_p1d": -0.02780528361918592, "alpha_spy_p1d": -0.02507298602390721, "alpha_c_p1d": -0.025084940370160996, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.08863671235432598, "ret_signed_p1w": 0.08863671235432598, "alpha_spy_p1w": 0.07964098704162355, "alpha_c_p1w": 0.06619622268213161, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.3585899755552539, "ret_signed_p1m": 0.3585899755552539, "alpha_spy_p1m": 0.33642248564954297, "alpha_c_p1m": 0.24567335034577392, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.7302929518690062, "ret_signed_p3m": 0.7302929518690062, "alpha_spy_p3m": 0.7489521251336984, "alpha_c_p3m": 0.6270407450816284, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3275, -0.2951, -0.3037, -0.2937, -0.2813, -0.2878, -0.293, -0.3133, -0.2925, -0.2895, -0.2545, -0.219, -0.2131, -0.2087, -0.2161, -0.2009, -0.2068, -0.1875, -0.1832, -0.1987, -0.1744, -0.1329, -0.1247, -0.1059, -0.118, -0.1183, -0.1152, -0.0717, -0.045, -0.0588, -0.0268, -0.013, -0.0079, 0.0556, -0.008, 0.0014, 0.0066, -0.006, 0.0398, 0.0398, 0.0426, 0.0275, -0.0077, 0.0239, -0.0255, -0.0381, -0.0541, -0.0942, -0.0655, -0.057, -0.0348, -0.0203, -0.0295, -0.0182, 0.0007, -0.0068, -0.0215, 0.005, 0.042, 0.0396, 0.0668, 0.044, 0.0284, 0.0, -0.0278, -0.0088, 0.0232, 0.0404, 0.0886, 0.0938, 0.1095, 0.1095, 0.1323, 0.1803, 0.1858, 0.1756, 0.1756, 0.2609, 0.2984, 0.363, 0.3739, 0.3577, 0.3764, 0.3799, 0.3586, 0.3629, 0.3832, 0.4401, 0.4462, 0.4236, 0.4694, 0.4993, 0.5088, 0.4753, 0.5671, 0.6362, 0.6435, 0.6424, 0.5948, 0.67, 0.6147, 0.5534, 0.5659, 0.604, 0.5811, 0.6299, 0.6864, 0.6866, 0.6866, 0.6699, 0.6997, 0.7312, 0.7798, 0.7801, 0.8308, 0.8653, 0.9427, 0.9105, 0.9109, 0.7227, 0.6249, 0.7371, 0.6784, 0.6051, 0.732, 0.7708, 0.7303, 0.7334, 0.8106, 0.8527, 0.9512, 0.8752, 0.8381, 0.7256, 0.756, 0.7403, 0.6371, 0.6396, 0.5429, 0.4965, 0.6648, 0.5861, 0.6388, 0.6832, 0.7119, 0.8682, 0.847, 0.8696, 0.8698, 0.9803, 1.0014, 1.0351, 1.011, 1.0197, 1.0704, 1.1186, 1.1341, 1.1373, 1.2347, 1.203, 1.2315, 1.2078, 1.243, 1.4266, 1.5161, 1.7533, 1.7749, 1.9253, 2.1648, 2.0765, 2.2295, 2.2256, 1.9807, 1.9664, 1.9157, 1.964, 2.1991, 2.1617, 2.1617, 2.453, 2.6394, 2.6365, 2.7858, null, null, null, null, null, null, null]}
{"unit_id": "quote:2004058658066534636", "ticker_idx": 0, "ts": "2025-12-25T05:16:32+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect the outperformance versus our prior forecasts to continue from 4Q25F onward, and we therefore raise our 2026F operating profit (OP) forecast to KRW133.4tn", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Nomura bullish call on Samsung and SK Hynix; raises 2026F OP forecasts on stronger DRAM/NAND price outlook.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Nomura: For Samsung Electronics, we expect the outperformance versus our prior forecasts to continue from 4Q25F onward, and we therefore raise our 2026F operating profit (OP) forecast to KRW133.4tn. In 2026F, we forecast Samsung’s DRAM and NAND prices to rise 68% y-y (commodity DRAM +81% y-y) and 73% y-y, respectively.\n\nNomura: For SK hynix, as we expect commodity prices to beat our previous estimates from 4Q25F onward, we also raise our 2026F OP forecast by 9.7% to KRW109tn. In 2026F, we expect SK hynix’s average DRAM/NAND prices to increase 46.7%/70.6% y-y (commodity DRAM +81% y-y).", "reply_text": null, "parent_text": null, "quoted_text": "Nomura Global memory: Stronger memory prices drive up earnings\n\nRevising up earnings for 4Q25F and beyond, reflecting a stronger commodity memory price outlook\nWe remain the most bullish in our earnings outlook for memory companies on the Street – since our report on memory", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05310536393950405, "ret_signed_p1d": 0.05310536393950405, "alpha_spy_p1d": 0.05320677897174719, "alpha_c_p1d": 0.05153331060748512, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.00157205333201893, "ret_p1w": 0.08445416491506297, "ret_signed_p1w": 0.08445416491506297, "alpha_spy_p1w": 0.09670827446226882, "alpha_c_p1w": 0.10038042063887109, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.015926255723808125, "ret_p1m": 0.3756920117600133, "ret_signed_p1m": 0.3756920117600133, "alpha_spy_p1m": 0.3773578245842101, "alpha_c_p1m": 0.3839627833788012, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": -0.008270771618787864, "ret_p3m": 0.7157711166213605, "ret_signed_p3m": 0.7157711166213605, "alpha_spy_p3m": 0.7670706341287067, "alpha_c_p3m": 0.7839921117726819, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": -0.0682209951513214, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2421, -0.2448, -0.2259, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1922, -0.1503, -0.1602, -0.1755, -0.1449, -0.1206, -0.1188, -0.117, -0.1224, -0.1125, -0.1314, -0.1107, -0.0819, -0.1044, -0.0954, -0.063, -0.0324, 0.0, -0.0558, -0.0945, -0.1071, -0.1188, -0.0945, -0.0684, -0.072, -0.0747, -0.1251, -0.0945, -0.1197, -0.1314, -0.0945, -0.1467, -0.1296, -0.1062, -0.0747, -0.0684, -0.0954, -0.0927, -0.0693, -0.0594, -0.054, -0.0243, -0.0144, -0.0243, -0.0279, -0.0342, -0.0198, -0.0567, -0.0747, -0.0288, -0.0315, -0.0432, -0.0054, 0.0036, 0.0, 0.0, 0.0531, 0.0808, 0.0845, 0.0845, 0.0845, 0.1622, 0.2491, 0.2563, 0.2753, 0.2554, 0.2572, 0.2554, 0.2445, 0.269, 0.3015, 0.3467, 0.3504, 0.3133, 0.3522, 0.3775, 0.3757, 0.3757, 0.4426, 0.4689, 0.4535, 0.4517, 0.3603, 0.515, 0.5295, 0.4408, 0.4345, 0.505, 0.4996, 0.5177, 0.6154, 0.6389, 0.6389, 0.6389, 0.6389, 0.7185, 0.7194, 0.7456, 0.8089, 0.8406, 0.9717, 0.9582, 0.9582, 0.7646, 0.5575, 0.733, 0.7022, 0.5692, 0.6995, 0.7185, 0.6995, 0.6597, 0.7067, 0.7538, 0.8858, 0.8135, 0.8035, 0.685, 0.7158, 0.7094, 0.6289, 0.6289, 0.5979, 0.5154, 0.7189, 0.6169, 0.6876, 0.7501, 0.781, 0.9078, 0.8489, 0.8671, 0.8217, 0.8716, 0.9124, 0.9713, 0.9577, 0.9441, 0.9849, 0.9713, 1.0347, 0.9894, 1.0347, 1.0121, 1.0483, 0.9985, 0.9985, 1.1072, 1.1072, 1.4109, 1.4607, 1.4335, 1.5876, 1.5287, 1.574, 1.6828, 1.4516, 1.5468, 1.497, 1.5015, 1.7145, 1.651, 1.651, 1.7099, 1.7825, 1.7145, 1.8731, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2004058658066534636", "ticker_idx": 1, "ts": "2025-12-25T05:16:32+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we also raise our 2026F OP forecast by 9.7% to KRW109tn", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Nomura bullish call on Samsung and SK Hynix; raises 2026F OP forecasts on stronger DRAM/NAND price outlook.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Nomura: For Samsung Electronics, we expect the outperformance versus our prior forecasts to continue from 4Q25F onward, and we therefore raise our 2026F operating profit (OP) forecast to KRW133.4tn. In 2026F, we forecast Samsung’s DRAM and NAND prices to rise 68% y-y (commodity DRAM +81% y-y) and 73% y-y, respectively.\n\nNomura: For SK hynix, as we expect commodity prices to beat our previous estimates from 4Q25F onward, we also raise our 2026F OP forecast by 9.7% to KRW109tn. In 2026F, we expect SK hynix’s average DRAM/NAND prices to increase 46.7%/70.6% y-y (commodity DRAM +81% y-y).", "reply_text": null, "parent_text": null, "quoted_text": "Nomura Global memory: Stronger memory prices drive up earnings\n\nRevising up earnings for 4Q25F and beyond, reflecting a stronger commodity memory price outlook\nWe remain the most bullish in our earnings outlook for memory companies on the Street – since our report on memory", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018707539869095013, "ret_signed_p1d": 0.018707539869095013, "alpha_spy_p1d": 0.01880895490133816, "alpha_c_p1d": 0.014066927348801661, "bench_spy_p1d": -0.00010141503224314619, "bench_c_p1d": 0.004640612520293352, "ret_p1w": 0.10714283432004468, "ret_signed_p1w": 0.10714283432004468, "alpha_spy_p1w": 0.11939694386725053, "alpha_c_p1w": 0.11823657734104598, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.011093743021001301, "ret_p1m": 0.30442176472254756, "ret_signed_p1m": 0.30442176472254756, "alpha_spy_p1m": 0.3060875775467443, "alpha_c_p1m": 0.20578656252107863, "bench_spy_p1m": -0.0016658128241967551, "bench_c_p1m": 0.09863520220146893, "ret_p3m": 0.6799649210329166, "ret_signed_p3m": 0.6799649210329166, "alpha_spy_p3m": 0.7312644385402628, "alpha_c_p3m": 0.5964325548618816, "bench_spy_p3m": -0.05129951750734618, "bench_c_p3m": 0.08353236617103499, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4069, -0.4094, -0.3882, -0.3279, -0.3279, -0.3279, -0.3279, -0.3279, -0.3279, -0.2726, -0.2947, -0.3007, -0.282, -0.231, -0.2089, -0.1749, -0.186, -0.1817, -0.1868, -0.1333, -0.0908, -0.1146, -0.0517, -0.0347, -0.05, 0.0537, -0.0041, -0.016, 0.0078, -0.0143, 0.0299, 0.052, 0.0486, 0.0401, -0.0483, 0.0299, -0.0313, -0.0449, -0.0296, -0.1146, -0.1163, -0.118, -0.1095, -0.0748, -0.0986, -0.085, -0.051, -0.0612, -0.0782, -0.0748, -0.0187, -0.0374, -0.0017, -0.0391, -0.0289, -0.0578, -0.0986, -0.0629, -0.0612, -0.0697, -0.0136, -0.0068, 0.0, 0.0, 0.0187, 0.0884, 0.1071, 0.1071, 0.1071, 0.1514, 0.1837, 0.2347, 0.2619, 0.2857, 0.2653, 0.2738, 0.2551, 0.2619, 0.2738, 0.2857, 0.2993, 0.2636, 0.2585, 0.284, 0.3044, 0.2517, 0.3605, 0.4303, 0.4643, 0.5459, 0.4116, 0.5425, 0.5306, 0.432, 0.4269, 0.5085, 0.4898, 0.4626, 0.5102, 0.4966, 0.4966, 0.4966, 0.4966, 0.5204, 0.6139, 0.6173, 0.7092, 0.7313, 0.8725, 0.8078, 0.8078, 0.5999, 0.4465, 0.6033, 0.5743, 0.4244, 0.5982, 0.6271, 0.5846, 0.5505, 0.6595, 0.6527, 0.7992, 0.726, 0.7157, 0.5897, 0.68, 0.6953, 0.5897, 0.5897, 0.4874, 0.375, 0.5283, 0.4142, 0.4925, 0.5096, 0.5607, 0.76, 0.7004, 0.7498, 0.772, 0.8793, 0.9355, 0.9679, 0.9219, 0.9867, 1.0855, 1.0838, 1.0872, 1.0821, 1.2013, 1.215, 1.203, 1.1911, 1.1911, 1.4654, 1.4654, 1.7278, 1.8181, 1.8726, 2.2032, 2.1265, 2.3667, 2.3565, 2.0992, 2.135, 1.9732, 1.9732, 2.3054, 2.3071, 2.3071, 2.4962, 2.8217, 2.9007, 2.9757, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:1990571404492976544", "ticker_idx": 0, "ts": "2025-11-18T00:03:00+00:00", "symbol": "King Yuan Electronics", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish CoWoS capacity upgrade thesis; positive view on KYEC, ASE, Marketech unchanged; $TSM tagged as beneficiary of CoWoS ramp.", "resolved_tickers": ["2449.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "King Yuan Electronics (KYEC) 2449.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "CoWoS: U.S. brokers raise capacity forecast\n\nA major U.S. securities house has revised up its forecast for next year’s CoWoS production capacity growth by 20–30%. In other words, it has raised its estimate from 100,000 wafers per month to at least 120,000–130,000 wafers per month. The new capacity will be added to the AP8 fab’s P1 and P2 lines, and the exact timing will depend on the progress of plant construction and equipment installation.\n\nThe positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/18 外電綜合整理\n\n- 記憶體產業: 美系再升目標\n*升目標：南亞科、華邦電、力積電、旺宏\n美系大行觀察DDR4 的重要性仍被低估，因eSSD 和部分 HDD 也需要 DDR4 做為快取記憶體。DDR4 16Gb 在通路端已幾乎沒有庫存情況下，券商預期明年合約價與現貨價都有更大上行空間，尤其在1Q26。", "parent_author": null, "ret_m1d": 0.07828282828282829, "ret_signed_m1d": 0.07828282828282829, "alpha_spy_m1d": 0.06981425319937418, "alpha_c_m1d": 0.057250200568329745, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.021032627714498542, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007575757575757569, "ret_signed_p1d": -0.007575757575757569, "alpha_spy_p1d": -0.011438957359344748, "alpha_c_p1d": -0.026028147822437786, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.018452390246680217, "ret_p1w": 0.03535353535353525, "ret_signed_p1w": 0.03535353535353525, "alpha_spy_p1w": 0.012720018387324838, "alpha_c_p1w": 0.015190906546027172, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.020162628807508076, "ret_p1m": 0.08585858585858586, "ret_signed_p1m": 0.08585858585858586, "alpha_spy_p1m": 0.06870922612794517, "alpha_c_p1m": 0.06800634055049026, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": 0.017852245308095593, "ret_p3m": 0.5833333333333333, "ret_signed_p3m": 0.5833333333333333, "alpha_spy_p3m": 0.5474521554382048, "alpha_c_p3m": 0.35620786593395204, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.22712546739938122, "ret_p6m": 0.5404040404040404, "ret_signed_p6m": 0.5404040404040404, "alpha_spy_p6m": 0.409424987124654, "alpha_c_p6m": -0.18254374221012615, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.7229477826141666, "price_path": [-0.3283, -0.3359, -0.3157, -0.2955, -0.298, -0.2828, -0.2121, -0.1995, -0.2323, -0.2247, -0.2374, -0.1919, -0.1465, -0.1389, -0.1288, -0.1616, -0.1616, -0.1591, -0.1616, -0.1667, -0.154, -0.1439, -0.1364, -0.1566, -0.1616, -0.1616, -0.2146, -0.2146, -0.1717, -0.1667, -0.154, -0.1465, -0.1465, -0.1288, -0.1212, -0.1111, -0.1111, -0.1414, -0.1843, -0.1616, -0.0884, -0.1162, -0.101, -0.0707, -0.1086, -0.1162, -0.1162, -0.0505, -0.048, 0.0455, 0.0253, 0.096, 0.1313, 0.0808, 0.1136, 0.0985, 0.0783, 0.096, 0.0758, 0.0631, 0.0884, 0.053, 0.0783, 0.0, -0.0076, 0.0682, 0.0328, 0.0278, 0.0354, 0.0707, 0.1212, 0.1566, 0.0985, 0.1439, 0.1414, 0.1136, 0.1187, 0.1515, 0.154, 0.1566, 0.1187, 0.1616, 0.1187, 0.0909, 0.0859, 0.053, 0.0884, 0.1162, 0.1111, 0.1061, 0.1061, 0.1263, 0.2374, 0.2298, 0.25, 0.25, 0.351, 0.3283, 0.3485, 0.346, 0.3283, 0.3308, 0.3636, 0.3662, 0.3662, 0.351, 0.3813, 0.3712, 0.3864, 0.3485, 0.3864, 0.447, 0.5025, 0.5152, 0.5707, 0.5404, 0.4949, 0.4394, 0.4823, 0.5, 0.4167, 0.4318, 0.4722, 0.5253, 0.5833, 0.5833, 0.5833, 0.5833, 0.5833, 0.5833, 0.5833, 0.5833, 0.5985, 0.6465, 0.6162, 0.6515, 0.6515, 0.596, 0.5404, 0.404, 0.4899, 0.4369, 0.3737, 0.4798, 0.5455, 0.5152, 0.4924, 0.4545, 0.4848, 0.6313, 0.596, 0.5556, 0.4444, 0.4066, 0.4571, 0.4545, 0.4066, 0.3737, 0.3182, 0.4167, 0.3081, 0.3081, 0.3081, 0.3535, 0.4141, 0.3889, 0.4015, 0.4899, 0.5202, 0.4697, 0.4318, 0.3889, 0.3687, 0.4192, 0.452, 0.4015, 0.4444, 0.4318, 0.4318, 0.4318, 0.5278, 0.5278, 0.6793, 0.7879, 0.6591, 0.7121, 0.5707, 0.5126, 0.5202, 0.5404]}
{"unit_id": "quote:1990571404492976544", "ticker_idx": 1, "ts": "2025-11-18T00:03:00+00:00", "symbol": "ASE Technology", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish CoWoS capacity upgrade thesis; positive view on KYEC, ASE, Marketech unchanged; $TSM tagged as beneficiary of CoWoS ramp.", "resolved_tickers": ["3711.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "ASE Technology Holding Taiwan listing 3711.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "CoWoS: U.S. brokers raise capacity forecast\n\nA major U.S. securities house has revised up its forecast for next year’s CoWoS production capacity growth by 20–30%. In other words, it has raised its estimate from 100,000 wafers per month to at least 120,000–130,000 wafers per month. The new capacity will be added to the AP8 fab’s P1 and P2 lines, and the exact timing will depend on the progress of plant construction and equipment installation.\n\nThe positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/18 外電綜合整理\n\n- 記憶體產業: 美系再升目標\n*升目標：南亞科、華邦電、力積電、旺宏\n美系大行觀察DDR4 的重要性仍被低估，因eSSD 和部分 HDD 也需要 DDR4 做為快取記憶體。DDR4 16Gb 在通路端已幾乎沒有庫存情況下，券商預期明年合約價與現貨價都有更大上行空間，尤其在1Q26。", "parent_author": null, "ret_m1d": 0.023255813953488413, "ret_signed_m1d": 0.023255813953488413, "alpha_spy_m1d": 0.014787238870034303, "alpha_c_m1d": 0.0022231862389898716, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.021032627714498542, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025581395348837188, "ret_signed_p1d": -0.025581395348837188, "alpha_spy_p1d": -0.029444595132424367, "alpha_c_p1d": -0.044033785595517405, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.018452390246680217, "ret_p1w": -0.013953488372092981, "ret_signed_p1w": -0.013953488372092981, "alpha_spy_p1w": -0.03658700533830339, "alpha_c_p1w": -0.03411611717960106, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.020162628807508076, "ret_p1m": 0.06279069767441858, "ret_signed_p1m": 0.06279069767441858, "alpha_spy_p1m": 0.04564133794377789, "alpha_c_p1m": 0.04493845236632299, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": 0.017852245308095593, "ret_p3m": 0.6488372093023256, "ret_signed_p3m": 0.6488372093023256, "alpha_spy_p3m": 0.6129560314071971, "alpha_c_p3m": 0.42171174190294436, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.22712546739938122, "ret_p6m": 1.5488372093023255, "ret_signed_p6m": 1.5488372093023255, "alpha_spy_p6m": 1.417858156022939, "alpha_c_p6m": 0.8258894266881589, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.7229477826141666, "price_path": [-0.3302, -0.3326, -0.3093, -0.307, -0.3, -0.314, -0.2977, -0.2744, -0.2698, -0.2698, -0.2837, -0.2744, -0.2023, -0.2233, -0.214, -0.2256, -0.2233, -0.2256, -0.2116, -0.214, -0.2093, -0.2093, -0.1977, -0.1884, -0.2, -0.2, -0.2302, -0.2302, -0.2372, -0.2419, -0.2349, -0.2302, -0.2302, -0.193, -0.1884, -0.1674, -0.1674, -0.1953, -0.2116, -0.1721, -0.1093, -0.0884, -0.0698, -0.093, -0.1023, -0.0884, -0.0884, -0.0581, -0.0558, 0.0279, 0.0465, 0.1512, 0.1442, 0.1116, 0.093, 0.0884, 0.0628, 0.0628, 0.0605, 0.0558, 0.0767, 0.0349, 0.0233, 0.0, -0.0256, 0.0349, -0.0279, -0.0209, -0.014, 0.0116, 0.0721, 0.0674, 0.0442, 0.0628, 0.0651, 0.0605, 0.0744, 0.1047, 0.1163, 0.1372, 0.1372, 0.1326, 0.0884, 0.0651, 0.0628, 0.0326, 0.0674, 0.0791, 0.0907, 0.0907, 0.0907, 0.1186, 0.1674, 0.1581, 0.1651, 0.1651, 0.2, 0.2256, 0.2628, 0.2791, 0.2721, 0.2651, 0.3163, 0.3465, 0.3698, 0.3884, 0.4116, 0.3907, 0.3767, 0.3535, 0.386, 0.4302, 0.4512, 0.4256, 0.4605, 0.4419, 0.3814, 0.3581, 0.4047, 0.407, 0.3209, 0.4163, 0.5209, 0.6, 0.6488, 0.6488, 0.6488, 0.6488, 0.6488, 0.6488, 0.6488, 0.6488, 0.6233, 0.7744, 0.8093, 0.8023, 0.8023, 0.7233, 0.6721, 0.5233, 0.6442, 0.5907, 0.4884, 0.5605, 0.607, 0.5581, 0.5884, 0.5698, 0.5907, 0.6628, 0.614, 0.6, 0.5326, 0.5465, 0.6372, 0.6698, 0.6442, 0.6442, 0.5279, 0.6791, 0.6791, 0.6791, 0.6791, 0.6372, 0.7814, 0.8233, 0.8279, 0.9419, 0.9721, 1.0767, 1.0953, 1.0558, 1.1465, 1.1953, 1.1628, 1.1605, 1.307, 1.3047, 1.3047, 1.2721, 1.2233, 1.2233, 1.4419, 1.4186, 1.4372, 1.5116, 1.4, 1.4977, 1.5814, 1.5488]}
{"unit_id": "quote:1990571404492976544", "ticker_idx": 2, "ts": "2025-11-18T00:03:00+00:00", "symbol": "Marketech", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish CoWoS capacity upgrade thesis; positive view on KYEC, ASE, Marketech unchanged; $TSM tagged as beneficiary of CoWoS ramp.", "resolved_tickers": ["6196.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Marketech International 6196.TW Taiwan semiconductor equip", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "quote", "lang": "en", "tweet_text": "CoWoS: U.S. brokers raise capacity forecast\n\nA major U.S. securities house has revised up its forecast for next year’s CoWoS production capacity growth by 20–30%. In other words, it has raised its estimate from 100,000 wafers per month to at least 120,000–130,000 wafers per month. The new capacity will be added to the AP8 fab’s P1 and P2 lines, and the exact timing will depend on the progress of plant construction and equipment installation.\n\nThe positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/18 外電綜合整理\n\n- 記憶體產業: 美系再升目標\n*升目標：南亞科、華邦電、力積電、旺宏\n美系大行觀察DDR4 的重要性仍被低估，因eSSD 和部分 HDD 也需要 DDR4 做為快取記憶體。DDR4 16Gb 在通路端已幾乎沒有庫存情況下，券商預期明年合約價與現貨價都有更大上行空間，尤其在1Q26。", "parent_author": null, "ret_m1d": 0.06349206349206349, "ret_signed_m1d": 0.06349206349206349, "alpha_spy_m1d": 0.05502348840860938, "alpha_c_m1d": 0.04697058652615693, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.016521476965906556, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.003863199783587179, "alpha_c_p1d": -0.006952676722960405, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.006952676722960405, "ret_p1w": 0.02947845804988658, "ret_signed_p1w": 0.02947845804988658, "alpha_spy_p1w": 0.00684494108367617, "alpha_c_p1w": 0.024353580211141868, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.0051248778387447125, "ret_p1m": 0.18140589569161003, "ret_signed_p1m": 0.18140589569161003, "alpha_spy_p1m": 0.16425653596096934, "alpha_c_p1m": 0.18230191854556677, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": -0.000896022853956735, "ret_p3m": 0.19274376417233552, "ret_signed_p3m": 0.19274376417233552, "alpha_spy_p3m": 0.15686258627720706, "alpha_c_p3m": 0.1909040110052458, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.0018397531670897305, "ret_p6m": 0.9024943310657596, "ret_signed_p6m": 0.9024943310657596, "alpha_spy_p6m": 0.7715152777863732, "alpha_c_p6m": 0.6313875366571615, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.2711067944085981, "price_path": [0.0975, 0.0658, 0.0794, 0.1202, 0.1905, 0.1973, 0.1973, 0.2177, 0.2018, 0.2177, 0.1927, 0.2018, 0.2245, 0.2358, 0.2426, 0.2426, 0.2313, 0.1927, 0.2063, 0.1565, 0.1587, 0.1429, 0.1655, 0.1565, 0.1655, 0.1383, 0.0952, 0.0952, 0.1088, 0.1043, 0.1043, 0.1134, 0.1134, 0.1202, 0.1134, 0.1361, 0.1361, 0.1043, 0.0907, 0.1497, 0.1474, 0.0839, 0.068, 0.0703, 0.034, 0.0227, 0.0227, 0.0748, 0.0544, 0.0703, 0.0794, 0.1156, 0.1451, 0.1247, 0.1247, 0.1043, 0.0998, 0.034, 0.0544, 0.0748, 0.1111, 0.0703, 0.0635, 0.0, 0.0, 0.0204, 0.0, 0.0045, 0.0295, 0.0635, 0.0499, 0.0658, 0.0522, 0.0363, 0.0431, 0.0295, 0.0295, 0.0476, 0.0748, 0.0816, 0.1111, 0.1655, 0.1519, 0.161, 0.1814, 0.1678, 0.2336, 0.2449, 0.2313, 0.2358, 0.2358, 0.2449, 0.2313, 0.2404, 0.2472, 0.2472, 0.2517, 0.2676, 0.3424, 0.381, 0.3016, 0.2517, 0.2358, 0.3265, 0.3469, 0.3855, 0.3515, 0.3265, 0.3968, 0.3152, 0.3424, 0.3243, 0.3107, 0.3288, 0.3401, 0.3175, 0.288, 0.2585, 0.2744, 0.254, 0.2041, 0.1701, 0.1905, 0.2018, 0.1927, 0.1927, 0.1927, 0.1927, 0.1927, 0.1927, 0.1927, 0.1927, 0.2653, 0.3016, 0.3243, 0.3832, 0.3832, 0.3515, 0.3129, 0.1927, 0.288, 0.2721, 0.2063, 0.229, 0.2971, 0.2857, 0.2517, 0.2698, 0.2766, 0.2948, 0.2721, 0.3061, 0.2404, 0.2721, 0.3605, 0.3152, 0.3288, 0.2766, 0.2381, 0.2698, 0.2789, 0.2789, 0.2789, 0.3243, 0.3696, 0.4444, 0.4762, 0.4535, 0.5011, 0.6009, 0.542, 0.6281, 0.7891, 0.8027, 0.771, 0.7052, 0.7506, 0.7256, 0.7302, 0.7075, 0.7143, 0.7143, 0.737, 0.6961, 0.7937, 0.7823, 0.8594, 0.9274, 0.9025, 0.9025]}
{"unit_id": "quote:1990571404492976544", "ticker_idx": 3, "ts": "2025-11-18T00:03:00+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "A major U.S. securities house has revised up its forecast for next year's CoWoS production capacity growth by 20–30%", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish CoWoS capacity upgrade thesis; positive view on KYEC, ASE, Marketech unchanged; $TSM tagged as beneficiary of CoWoS ramp.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "CoWoS: U.S. brokers raise capacity forecast\n\nA major U.S. securities house has revised up its forecast for next year’s CoWoS production capacity growth by 20–30%. In other words, it has raised its estimate from 100,000 wafers per month to at least 120,000–130,000 wafers per month. The new capacity will be added to the AP8 fab’s P1 and P2 lines, and the exact timing will depend on the progress of plant construction and equipment installation.\n\nThe positive view on the Taiwanese companies under its coverage — King Yuan Electronics (KYEC), ASE, and Marketech — remains unchanged.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": "早安！11/18 外電綜合整理\n\n- 記憶體產業: 美系再升目標\n*升目標：南亞科、華邦電、力積電、旺宏\n美系大行觀察DDR4 的重要性仍被低估，因eSSD 和部分 HDD 也需要 DDR4 做為快取記憶體。DDR4 16Gb 在通路端已幾乎沒有庫存情況下，券商預期明年合約價與現貨價都有更大上行空間，尤其在1Q26。", "parent_author": null, "ret_m1d": 0.01475291055569028, "ret_signed_m1d": 0.01475291055569028, "alpha_spy_m1d": 0.006284335472236169, "alpha_c_m1d": -0.006279717158808262, "bench_spy_m1d": 0.00846857508345411, "bench_c_m1d": 0.021032627714498542, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016048331157736406, "ret_signed_p1d": 0.016048331157736406, "alpha_spy_p1d": 0.012185131374149227, "alpha_c_p1d": -0.0024040590889438107, "bench_spy_p1d": 0.003863199783587179, "bench_c_p1d": 0.018452390246680217, "ret_p1w": 0.02436024529515657, "ret_signed_p1w": 0.02436024529515657, "alpha_spy_p1w": 0.0017267283289461588, "alpha_c_p1w": 0.004197616487648492, "bench_spy_p1w": 0.02263351696621041, "bench_c_p1w": 0.020162628807508076, "ret_p1m": -0.0007280943956606833, "ret_signed_p1m": -0.0007280943956606833, "alpha_spy_p1m": -0.017877454126301373, "alpha_c_p1m": -0.018580339703756277, "bench_spy_p1m": 0.01714935973064069, "bench_c_p1m": 0.017852245308095593, "ret_p3m": 0.32182721880528486, "ret_signed_p3m": 0.32182721880528486, "alpha_spy_p3m": 0.2859460409101564, "alpha_c_p3m": 0.09470175140590364, "bench_spy_p3m": 0.03588117789512846, "bench_c_p3m": 0.22712546739938122, "ret_p6m": 0.4465433624710955, "ret_signed_p6m": 0.4465433624710955, "alpha_spy_p6m": 0.3155643091917091, "alpha_c_p6m": -0.2764044201430711, "bench_spy_p6m": 0.13097905327938641, "bench_c_p6m": 0.7229477826141666, "price_path": [-0.1846, -0.1643, -0.1549, -0.1437, -0.1417, -0.1453, -0.1719, -0.1719, -0.1808, -0.17, -0.1563, -0.1269, -0.1133, -0.1, -0.0658, -0.0713, -0.0698, -0.0624, -0.057, -0.0544, -0.0334, -0.0469, -0.019, 0.0173, 0.0101, -0.0045, -0.0164, -0.0168, 0.005, 0.038, 0.0367, 0.0514, 0.0881, 0.058, 0.0958, 0.0791, 0.0099, 0.0899, 0.0649, 0.0964, 0.0789, 0.0618, 0.0712, 0.0597, 0.0395, 0.0461, 0.0614, 0.0732, 0.085, 0.0978, 0.0911, 0.081, 0.097, 0.0581, 0.0566, 0.0408, 0.0309, 0.0625, 0.0477, 0.0457, 0.0154, 0.0249, 0.0148, 0.0, 0.016, -0.0015, -0.0103, 0.0242, 0.0244, 0.0434, 0.0434, 0.0489, 0.0352, 0.051, 0.0631, 0.054, 0.0605, 0.0862, 0.0918, 0.116, 0.0999, 0.0537, 0.0382, 0.035, -0.0007, 0.0271, 0.0425, 0.0582, 0.0714, 0.0781, 0.0781, 0.0926, 0.0857, 0.0809, 0.0964, 0.0964, 0.1532, 0.1627, 0.1814, 0.1498, 0.1474, 0.1677, 0.197, 0.195, 0.1802, 0.2326, 0.2354, 0.2354, 0.1804, 0.1766, 0.1812, 0.2082, 0.2004, 0.2207, 0.235, 0.2251, 0.1927, 0.2316, 0.2114, 0.1753, 0.1933, 0.2587, 0.2823, 0.3058, 0.3497, 0.3281, 0.3218, 0.3218, 0.314, 0.307, 0.3003, 0.3369, 0.3351, 0.3918, 0.3989, 0.3595, 0.3515, 0.3317, 0.2741, 0.2896, 0.2767, 0.2227, 0.2581, 0.2523, 0.2793, 0.2149, 0.2206, 0.2276, 0.2518, 0.2286, 0.2258, 0.1912, 0.2246, 0.2419, 0.2582, 0.1799, 0.1822, 0.1452, 0.2228, 0.2356, 0.2267, 0.2267, 0.2365, 0.2494, 0.3239, 0.3224, 0.3409, 0.3372, 0.3745, 0.3572, 0.3147, 0.3405, 0.3251, 0.3318, 0.4018, 0.3845, 0.4562, 0.4653, 0.4196, 0.4249, 0.433, 0.4388, 0.4531, 0.427, 0.5178, 0.4985, 0.4895, 0.4637, 0.4374, 0.4465]}
{"unit_id": "thread:1998698470270710042", "ticker_idx": 0, "ts": "2025-12-10T10:17:03+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics has received positive signals during the 6th generation High Bandwidth Memory (HBM4) verification process currently underway with Nvidia... the signing of a supply contract is imminent", "tweet_type": "news_reaction", "timeline": "months", "summary": "Samsung received positive HBM4 qualification feedback from Nvidia with supply contract imminent, signaling recovery in HBM competitive position.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung HBM4 Receives \"Good News\" Rating from Nvidia... Supply Deal Imminent\n\nIt has been confirmed that Samsung Electronics has received positive signals during the 6th generation High Bandwidth Memory (HBM4) verification process currently underway with Nvidia. The likelihood of Samsung and Nvidia proceeding with an HBM4 supply contract in the near future has also increased.\nAccording to industry sources on the 10th, Samsung Electronics is currently conducting HBM4 qualification tests (qual tests) with Nvidia. It is reported that Nvidia recently provided feedback regarding Samsung's HBM4 performance, stating that \"there will be good news.\"\n\nThe industry interprets this as a positive assessment of the qualification test results and forecasts that the signing of a supply contract is imminent.\nTypically, Nvidia is known not to notify vendors of pass/fail results based on specific performance indicators or evaluation items. Instead, once a certain level of performance is secured, they deliver positive feedback and naturally transition to the contract stage. This is why the mention of \"good news\" is being interpreted as a sign that the process has entered the steps toward a formal contract.\n\nSamsung recently initiated internal Production Readiness Approval (PRA) procedures and has already completed preparations for mass production. Industry observers are lending weight to the view that, given PRA is typically the stage immediately preceding actual mass production, virtually only the supply contract remains for Samsung's HBM4.\nSamsung was evaluated as having a weakened presence after falling behind SK Hynix in the HBM3E competition last year. However, it is understood that by improving completeness using 1c-based DRAM and utilizing the 4-nanometer (nm) foundry process to enhance the performance of the logic die (the component acting as the brain), they have boosted operating speeds to over 11 gigabits per second (Gbps).\n\n---\n\nhttps://t.co/Fop4lMWyU5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0037036985779217435, "ret_signed_m1d": 0.0037036985779217435, "alpha_spy_m1d": 0.010292269419319933, "alpha_c_m1d": 0.008477306363096737, "bench_spy_m1d": -0.006588570841398189, "bench_c_m1d": -0.004773607785174994, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006481527141408594, "ret_signed_p1d": -0.006481527141408594, "alpha_spy_p1d": -0.008808475271935845, "alpha_c_p1d": -0.001371716467460704, "bench_spy_p1d": 0.002326948130527251, "bench_c_p1d": -0.00510981067394789, "ret_p1w": -0.0009260156945565257, "ret_signed_p1w": -0.0009260156945565257, "alpha_spy_p1w": 0.022591649604687736, "alpha_c_p1w": 0.06187232193235925, "bench_spy_p1w": -0.02351766529924426, "bench_c_p1w": -0.06279833762691578, "ret_p1m": 0.2914325738858621, "ret_signed_p1m": 0.2914325738858621, "alpha_spy_p1m": 0.2856478335055139, "alpha_c_p1m": 0.3201504526472734, "bench_spy_p1m": 0.005784740380348152, "bench_c_p1m": -0.028717878761411297, "ret_p3m": 0.6142907537773581, "ret_signed_p3m": 0.6142907537773581, "alpha_spy_p3m": 0.6249016908174533, "alpha_c_p3m": 0.6731760519069158, "bench_spy_p3m": -0.010610937040095192, "bench_c_p3m": -0.05888529812955767, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.305, -0.2948, -0.2681, -0.2791, -0.2598, -0.2598, -0.2303, -0.2192, -0.2128, -0.2063, -0.2321, -0.2204, -0.2231, -0.2037, -0.169, -0.169, -0.169, -0.169, -0.169, -0.169, -0.1259, -0.1361, -0.1519, -0.1204, -0.0954, -0.0935, -0.0917, -0.0972, -0.087, -0.1065, -0.0852, -0.0556, -0.0787, -0.0694, -0.0361, -0.0046, 0.0287, -0.0287, -0.0685, -0.0815, -0.0935, -0.0685, -0.0417, -0.0454, -0.0481, -0.1, -0.0685, -0.0944, -0.1065, -0.0685, -0.1222, -0.1046, -0.0806, -0.0481, -0.0417, -0.0694, -0.0667, -0.0426, -0.0324, -0.0269, 0.0037, 0.0139, 0.0037, 0.0, -0.0065, 0.0083, -0.0296, -0.0481, -0.0009, -0.0037, -0.0157, 0.0231, 0.0324, 0.0287, 0.0287, 0.0833, 0.1119, 0.1156, 0.1156, 0.1156, 0.1956, 0.2849, 0.2924, 0.3119, 0.2914, 0.2933, 0.2914, 0.2803, 0.3054, 0.3389, 0.3854, 0.3891, 0.351, 0.391, 0.417, 0.4152, 0.4152, 0.484, 0.511, 0.4952, 0.4933, 0.3994, 0.5585, 0.5734, 0.4822, 0.4757, 0.5482, 0.5426, 0.5613, 0.6617, 0.6859, 0.6859, 0.6859, 0.6859, 0.7678, 0.7687, 0.7957, 0.8609, 0.8934, 1.0283, 1.0144, 1.0144, 0.8153, 0.6022, 0.7827, 0.7511, 0.6143, 0.7483, 0.7678, 0.7483, 0.7073, 0.7557, 0.8041, 0.9399, 0.8655, 0.8553, 0.7334, 0.765, 0.7585, 0.6757, 0.6757, 0.6437, 0.5589, 0.7682, 0.6633, 0.736, 0.8004, 0.8321, 0.9626, 0.902, 0.9206, 0.874, 0.9253, 0.9673, 1.0279, 1.0139, 0.9999, 1.0419, 1.0279, 1.0931, 1.0465, 1.0931, 1.0698, 1.1071, 1.0558, 1.0558, 1.1677, 1.1677, 1.4801, 1.5313, 1.5034, 1.6619, 1.6013, 1.6479, 1.7598, 1.522, 1.6199, 1.5686, 1.5733, 1.7924, 1.7271, 1.7271, 1.7877, 1.8623, 1.7924, 1.9556, null, null, null, null]}
{"unit_id": "orig:1991310071775846873", "ticker_idx": 0, "ts": "2025-11-20T00:58:12+00:00", "symbol": "009150.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electro-Mechanics has entered a phase where valuation re-rating driven by ABF substrate growth is set to accelerate in earnest, and we maintain our existing view that its investment attractiveness is high within our large-cap IT hardware coverage", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SEMCO as biggest beneficiary of T-glass shortage; secured inventory early, added 4 new customers, full utilization through 2027, valuation re-rating accelerating.", "resolved_tickers": ["009150.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "▶ Acceleration of T-Glass Shortage and Its Future Impact (ft. Samsung Electro-Mechanics)\n\n[Yang Seung-soo, Meritz Securities / Electronics & IT Components]\n\n[T-Glass Shortage Accelerating]\n    \n• T-glass is a key material that lowers the coefficient of thermal expansion (CTE) and prevents warpage of substrates.\n    \n• Recently, the T-glass shortage has intensified, and the impact of this supply shortage is spreading across the entire ABF/BT substrate industry.\n    \n• There are two main reasons why the T-glass shortage is accelerating.\n    \n1. Japan’s Nittobo continued to expand capacity conservatively without fully anticipating the surge in AI demand.\n    \n• Although Nittobo has recently announced aggressive T-glass capacity expansion plans, it takes time to expand furnaces, so the actual increase in supply is expected to materialize only from 2027.\n    \n• Competitors are also successively launching low-CTE glass fiber products, but there are differences in quality and capacity. Recently, as Nittobo has reallocated internal production capacity toward T-glass, tight supply-demand conditions have been spreading to other glass fiber products as well.\n    \n•  2. Nvidia has secured volumes in advance by entering into a mid- to long-term binding contract with Nittobo.\n    \n• As a result, it has become even more difficult for other customers to secure Nittobo’s T-glass, and this tight supply-demand situation has recently begun to spread even to general IT set customers such as Apple.\n\n[Future Price Increases and Vendor Diversification Expected]\n    \n• Substrate makers are now actively moving to improve profitability in specific product groups by taking advantage of the current T-glass shortage phase.\n    \n• So far, ABF/BT price hikes have been applied mainly to relatively smaller customers (i.e., non-Nvidia/Apple) or memory customers, largely reflecting the rise in raw material (copper, gold) and BT CCL prices.\n    \n• However, as the supply shortage is expected to deepen further going forward, the likelihood of price hikes being passed on to major customers is also gradually increasing.\n    \n• In fact, Nittobo hinted at the possibility of raising prices even for existing customers (Nvidia and Apple) in its recent earnings announcement.\n    \n• In particular for ABF, as the production difficulty of AI ABF substrates rises and volumes expand, continuous price hikes that reflect both the T-glass shortage and higher raw material costs are expected to persist over the next several quarters.\n    \n• Based on the tight T-glass inventory situation, moves to diversify the supply chain in AI ABF substrates are also expanding.\n    \n• As production bottlenecks are worsening for AI ABF substrates due to higher manufacturing complexity, some players that secured T-glass in advance have succeeded in winning multiple new customers and are actively leveraging their supplier advantage.\n\n[Implications for Domestic ABF Substrate Makers]\n    \n• According to our channel checks, Samsung Electro-Mechanics (SEMCO) is emerging as the biggest beneficiary of the T-glass shortage.\n    \n• The key factor is that among companies with the technological capability to supply AI ABF substrates, Samsung Electro-Mechanics was one of the first to successfully secure T-glass inventory.\n    \n• On this basis, it has added four new customers, including North American GPU and CPU customers, in addition to its existing ASIC customers, and is expected to operate at effectively full utilization through 2027.\n    \n• In the past, the ABF substrate business was centered on mass-producing identical products according to the roadmap of the leading player, Intel. With AI, however, it is shifting to a customized business that tailors specifications to each customer.\n\n• As a result, separate design and verification processes are required for each product, which has a capacity-diluting effect. In addition, since these substrates are fundamentally used in CoWoS packaging, larger panel sizes and higher layer counts are required, causing yield declines and making supply even tighter.\n    \n• Consequently, the supply-demand imbalance for AI ABF substrates is worsening, and the degree of benefit accruing to the small group of players participating in this market, including Samsung Electro-Mechanics, is expected to increase further.\n    \n• We believe Samsung Electro-Mechanics has entered a phase where valuation re-rating driven by ABF substrate growth is set to accelerate in earnest, and we maintain our existing view that its investment attractiveness is high within our large-cap IT hardware coverage.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01601833801227648, "ret_signed_m1d": -0.01601833801227648, "alpha_spy_m1d": -0.03149659956895268, "alpha_c_m1d": -0.04842696981923833, "bench_spy_m1d": 0.015478261556676198, "bench_c_m1d": 0.03240863180696185, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004576637075617063, "ret_signed_p1d": -0.004576637075617063, "alpha_spy_p1d": -0.014537860052249885, "alpha_c_p1d": -0.00843484501044589, "bench_spy_p1d": 0.009961222976632822, "bench_c_p1d": 0.003858207934828828, "ret_p1w": 0.1670480109933392, "ret_signed_p1w": 0.1670480109933392, "alpha_spy_p1w": 0.12544074090820145, "alpha_c_p1w": 0.12431424263334301, "bench_spy_p1w": 0.04160727008513776, "bench_c_p1w": 0.042733768359996205, "ret_p1m": 0.1578947368421053, "ret_signed_p1m": 0.1578947368421053, "alpha_spy_p1m": 0.11181092270767246, "alpha_c_p1m": 0.09524541007195331, "bench_spy_p1m": 0.046083814134432854, "bench_c_p1m": 0.062649326770152, "ret_p3m": 0.4294458631296927, "ret_signed_p3m": 0.4294458631296927, "alpha_spy_p3m": 0.37988846604232474, "alpha_c_p3m": 0.4028681718295064, "bench_spy_p3m": 0.04955739708736795, "bench_c_p3m": 0.02657769130018628, "ret_p6m": 3.6647506357382538, "ret_signed_p6m": 3.6647506357382538, "alpha_spy_p6m": 3.5255252054160175, "alpha_c_p6m": 3.36585792522349, "bench_spy_p6m": 0.13922543032223622, "bench_c_p6m": 0.2988927105147636, "price_path": [-0.2787, -0.27, -0.2618, -0.259, -0.2654, -0.2577, -0.2224, -0.2027, -0.1835, -0.1844, -0.1822, -0.1748, -0.1739, -0.1396, -0.1286, -0.135, -0.1172, -0.1208, -0.1076, -0.1076, -0.0526, -0.0664, -0.0778, -0.0847, -0.1227, -0.0957, -0.1144, -0.1153, -0.1007, -0.1007, -0.1007, -0.1007, -0.1007, -0.1007, -0.0366, -0.0686, -0.0929, -0.0847, -0.0778, -0.0526, -0.0275, -0.0114, -0.0297, 0.016, 0.016, 0.0458, 0.0618, 0.0618, 0.0297, 0.1213, 0.1121, 0.0938, 0.0503, 0.0618, 0.0183, 0.0252, 0.0412, 0.0366, 0.0137, -0.0252, -0.0137, -0.0664, -0.016, 0.0, -0.0046, 0.0343, 0.1053, 0.1762, 0.167, 0.1487, 0.1442, 0.1533, 0.1945, 0.2128, 0.2037, 0.1991, 0.2334, 0.2288, 0.2334, 0.2449, 0.2059, 0.167, 0.1831, 0.1625, 0.1579, 0.1648, 0.151, 0.1602, 0.1602, 0.1854, 0.1893, 0.1777, 0.1777, 0.1777, 0.247, 0.247, 0.2378, 0.2355, 0.2562, 0.2285, 0.2886, 0.3348, 0.314, 0.3371, 0.3371, 0.3486, 0.3024, 0.2932, 0.3071, 0.2562, 0.2701, 0.2793, 0.2932, 0.2909, 0.2886, 0.3071, 0.4202, 0.4087, 0.3394, 0.3117, 0.404, 0.4479, 0.4294, 0.4756, 0.4294, 0.4294, 0.4294, 0.4294, 0.6534, 0.7412, 0.9698, 1.0599, 1.1476, 1.1546, 1.0714, 1.0714, 0.889, 0.6742, 0.8544, 0.8682, 0.7643, 0.8567, 0.8751, 0.8451, 0.8497, 0.9075, 1.0622, 1.143, 1.2146, 1.1453, 0.9744, 1.0183, 1.0991, 1.1107, 1.1107, 0.9814, 0.8821, 1.0506, 0.9259, 1.1061, 1.1338, 1.1107, 1.3739, 1.3832, 1.6095, 1.6233, 1.6972, 1.8358, 1.9513, 2.136, 2.1406, 2.5655, 2.7503, 2.5748, 2.6394, 2.6671, 2.875, 2.8196, 2.8426, 2.8426, 3.2398, 3.2398, 3.2121, 3.2352, 3.2214, 3.1567, 3.4246, 3.7525, 3.7294, 3.6648]}
{"unit_id": "orig:1995707727516193158", "ticker_idx": 0, "ts": "2025-12-02T04:12:55+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory", "tweet_type": "news_reaction", "timeline": "years", "summary": "Raises INTC TP $50→$54 on 14A customer wins (Apple, NVDA, AMD), EMIB AI accelerator pipeline, and IFS GM turnaround; endorses Buy on AAPL, NVDA, AMD, MediaTek, TSMC.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Intel-related update (18A yield, 14A external customer wins, and EMIB developments)\n\nMore to come, raise TP to $54: On Nov 28, Intel share price was up 10% owing to news of order win for Apple’s M-series in 2027. We’ve highlighted the potential order win of M-series in our earlier Intel report (Belief in the Story Takes Hold on Oct 24) and indeed, the progress is earlier than expected. We expect more to come and now expect the upcoming win in Apple (AAPL Buy) non-pro smartphone SoC based on Intel 14A for 2028; coupled with the likely engagement with Nvidia (NVDA Buy) and AMD (Buy)’s server SKUs, we believe its 14A will be a strong node. For back-end, we expect Intel to benefit from the CoWoS undersupply and potential “Made in USA” trend. As highlighted in our MediaTek (2454 TT Buy) Upgrade report on Nov 28, we expect TPU v8e to adopt EMIB, and we also expect Apple to be an important customer. Near term, despite memory concerns, server CPU sales are expected to accelerate due to increasing general purpose compute. In all, we believe the recent developments bode well for IFS in the longer term. We raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory.\n\n18A the key to build customers’ confidence: According to Intel, Panther Lake is on track to enter HVM and debut on Jan 5 at CES 2026. We believe the yield rate has reached 60-65% in November and targets 70% by end-2025. We also expect the entry-level Wildcat Lake to start HVM in 1Q26, with much higher yield rates due to smaller die area.\n\nRobust IFS customer pipeline with top-tier U.S. fabless: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple M-series on 18A-P in 2027, we expect 14A to be driven by the potential wins for: 1) Apple’s non-pro smartphone SoC in 2028 (we expect firm visibility by end of 2025); 2) Nvidia & AMD server projects in 2028; 3) Tier-1 Ethernet switch in 2028. With thoughtful cost/scale management, we expect its IFS GM turnaround by end-2026, followed by a strong ramp of 14A in 2027.\n\nBackend EMIB to win AI accelerators on “TSMC + Intel” model: We believe that EMIB emerges as an alternative to CoWoS, given the capacity constraints at TSMC (2330 TT Buy), higher design flexibility and larger reticle size (vs. OSAT’s). As said, we expect TPU v8e to adopt EMIB-T in 2027, as Google provides a very aggressive volume in 2027 and 2028. We also see high chance for Intel to win more ASICs, likely Broadcom and Apple’s. Capacity wise, we expect Intel to add equipment order in 1Q26. Aside from current Malaysia factory, it would be making sense for Intel to accelerate in New Mexico Fab to benefit from TSMC’s Arizona. In sum, we estimate its advanced packaging business will contribute >$1bn revenue in 2027E, with growth in the following years through more project wins. \n\nRisks: 1) execution in product & foundry; 2) weak macro; and 3) peer competition.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07959518753181105, "ret_signed_m1d": -0.07959518753181105, "alpha_spy_m1d": -0.0777464025943061, "alpha_c_m1d": -0.061594491184534084, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.018000696347276968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006671200659915444, "ret_signed_p1d": 0.006671200659915444, "alpha_spy_p1d": 0.0032084328666550643, "alpha_c_p1d": -0.00707280041390379, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.013744001073819234, "ret_p1w": -0.06832300752935394, "ret_signed_p1w": -0.06832300752935394, "alpha_spy_p1w": -0.07053845118369928, "alpha_c_p1w": -0.09494849636071345, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02662548883135951, "ret_p1m": -0.15113870509152816, "ret_signed_p1m": -0.15113870509152816, "alpha_spy_p1m": -0.15466749126834134, "alpha_c_p1m": -0.15620391610621576, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": 0.005065211014687598, "ret_p3m": 0.049229338153991886, "ret_signed_p3m": 0.049229338153991886, "alpha_spy_p3m": 0.03971107240739147, "alpha_c_p3m": -0.08488428568970208, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": 0.13411362384369396, "ret_p6m": 1.8012420801375995, "ret_signed_p6m": 1.8012420801375995, "alpha_spy_p6m": 1.6938321515491355, "alpha_c_p6m": 1.1392968796703244, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.661945200467275, "price_path": [-0.4339, -0.4366, -0.4369, -0.4378, -0.4302, -0.4339, -0.4461, -0.4302, -0.4187, -0.4272, -0.2968, -0.3195, -0.3384, -0.3251, -0.2818, -0.2181, -0.1833, -0.2068, -0.2282, -0.1732, -0.1419, -0.1527, -0.1583, -0.1449, -0.1389, -0.1304, -0.1633, -0.1438, -0.1804, -0.1454, -0.1525, -0.1486, -0.1235, -0.1231, -0.1507, -0.1222, -0.1194, -0.0904, -0.0446, -0.049, -0.0761, -0.0801, -0.0913, -0.1481, -0.1171, -0.1433, -0.1228, -0.1155, -0.1286, -0.1284, -0.1739, -0.1829, -0.2015, -0.2103, -0.1923, -0.2266, -0.2063, -0.1767, -0.1758, -0.1532, -0.1532, -0.0669, -0.0796, 0.0, 0.0067, -0.0683, -0.0474, -0.0729, -0.0683, -0.0619, -0.0911, -0.1302, -0.1371, -0.1417, -0.1707, -0.1654, -0.153, -0.1633, -0.1638, -0.1682, -0.1682, -0.1672, -0.1562, -0.1419, -0.1511, -0.1511, -0.0941, -0.0943, -0.0789, -0.0193, -0.0543, 0.0478, 0.0136, 0.0879, 0.1208, 0.1116, 0.0803, 0.0803, 0.1171, 0.248, 0.2496, 0.0368, -0.0225, 0.0106, 0.1222, 0.1194, 0.069, 0.1228, 0.133, 0.118, 0.1097, 0.1638, 0.1557, 0.0842, 0.1109, 0.0692, 0.0764, 0.0764, 0.0623, 0.0458, 0.0265, 0.0147, 0.0037, 0.061, 0.0784, 0.0458, 0.0492, 0.0467, -0.0085, 0.0485, 0.0571, -0.0012, 0.0485, 0.0761, 0.1037, 0.0409, 0.0529, 0.0527, 0.0136, 0.0359, 0.0623, 0.0092, 0.0124, 0.0136, 0.0853, 0.0145, -0.0078, -0.0525, 0.0152, 0.1049, 0.159, 0.159, 0.1682, 0.2172, 0.3561, 0.4198, 0.435, 0.4994, 0.4679, 0.4939, 0.5758, 0.5758, 0.5114, 0.5243, 0.5015, 0.5362, 0.8988, 0.9551, 0.9443, 1.1797, 1.1735, 1.2917, 1.2034, 1.4879, 1.5997, 1.5217, 1.8737, 1.9777, 1.7746, 1.7672, 1.6669, 1.5022, 1.4884, 1.5489, 1.7366, 1.726, 1.7568, 1.7568, 1.8415, 1.8012]}
{"unit_id": "orig:1995707727516193158", "ticker_idx": 1, "ts": "2025-12-02T04:12:55+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Apple (AAPL Buy)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Raises INTC TP $50→$54 on 14A customer wins (Apple, NVDA, AMD), EMIB AI accelerator pipeline, and IFS GM turnaround; endorses Buy on AAPL, NVDA, AMD, MediaTek, TSMC.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Intel-related update (18A yield, 14A external customer wins, and EMIB developments)\n\nMore to come, raise TP to $54: On Nov 28, Intel share price was up 10% owing to news of order win for Apple’s M-series in 2027. We’ve highlighted the potential order win of M-series in our earlier Intel report (Belief in the Story Takes Hold on Oct 24) and indeed, the progress is earlier than expected. We expect more to come and now expect the upcoming win in Apple (AAPL Buy) non-pro smartphone SoC based on Intel 14A for 2028; coupled with the likely engagement with Nvidia (NVDA Buy) and AMD (Buy)’s server SKUs, we believe its 14A will be a strong node. For back-end, we expect Intel to benefit from the CoWoS undersupply and potential “Made in USA” trend. As highlighted in our MediaTek (2454 TT Buy) Upgrade report on Nov 28, we expect TPU v8e to adopt EMIB, and we also expect Apple to be an important customer. Near term, despite memory concerns, server CPU sales are expected to accelerate due to increasing general purpose compute. In all, we believe the recent developments bode well for IFS in the longer term. We raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory.\n\n18A the key to build customers’ confidence: According to Intel, Panther Lake is on track to enter HVM and debut on Jan 5 at CES 2026. We believe the yield rate has reached 60-65% in November and targets 70% by end-2025. We also expect the entry-level Wildcat Lake to start HVM in 1Q26, with much higher yield rates due to smaller die area.\n\nRobust IFS customer pipeline with top-tier U.S. fabless: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple M-series on 18A-P in 2027, we expect 14A to be driven by the potential wins for: 1) Apple’s non-pro smartphone SoC in 2028 (we expect firm visibility by end of 2025); 2) Nvidia & AMD server projects in 2028; 3) Tier-1 Ethernet switch in 2028. With thoughtful cost/scale management, we expect its IFS GM turnaround by end-2026, followed by a strong ramp of 14A in 2027.\n\nBackend EMIB to win AI accelerators on “TSMC + Intel” model: We believe that EMIB emerges as an alternative to CoWoS, given the capacity constraints at TSMC (2330 TT Buy), higher design flexibility and larger reticle size (vs. OSAT’s). As said, we expect TPU v8e to adopt EMIB-T in 2027, as Google provides a very aggressive volume in 2027 and 2028. We also see high chance for Intel to win more ASICs, likely Broadcom and Apple’s. Capacity wise, we expect Intel to add equipment order in 1Q26. Aside from current Malaysia factory, it would be making sense for Intel to accelerate in New Mexico Fab to benefit from TSMC’s Arizona. In sum, we estimate its advanced packaging business will contribute >$1bn revenue in 2027E, with growth in the following years through more project wins. \n\nRisks: 1) execution in product & foundry; 2) weak macro; and 3) peer competition.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010796985582902097, "ret_signed_m1d": -0.010796985582902097, "alpha_spy_m1d": -0.008948200645397142, "alpha_c_m1d": -0.0005999943823409293, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.010196991200561167, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007128157809805358, "ret_signed_p1d": -0.007128157809805358, "alpha_spy_p1d": -0.010590925603065737, "alpha_c_p1d": -0.009513263024488627, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.0023851052146832696, "ret_p1w": -0.0314826702852834, "ret_signed_p1w": -0.0314826702852834, "alpha_spy_p1w": -0.03369811393962874, "alpha_c_p1w": -0.05478039240825561, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02329772212297221, "ret_p1m": -0.05007166831027843, "ret_signed_p1m": -0.05007166831027843, "alpha_spy_p1m": -0.05360045448709161, "alpha_c_p1m": -0.04688031544851712, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": -0.0031913528617613096, "ret_p3m": -0.0760431913609586, "ret_signed_p3m": -0.0760431913609586, "alpha_spy_p3m": -0.08556145710755902, "alpha_c_p3m": -0.036779184817891886, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": -0.03926400654306672, "ret_p6m": 0.08818455201510433, "ret_signed_p6m": 0.08818455201510433, "alpha_spy_p6m": -0.019225376573359654, "alpha_c_p6m": -0.19034566127452668, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.278530213289631, "price_path": [-0.163, -0.1633, -0.1696, -0.1819, -0.2083, -0.197, -0.1829, -0.1737, -0.1687, -0.1657, -0.1696, -0.143, -0.1061, -0.1118, -0.1192, -0.1033, -0.1082, -0.1118, -0.1111, -0.1083, -0.1024, -0.0993, -0.1039, -0.1047, -0.0992, -0.1132, -0.1438, -0.1355, -0.1351, -0.1296, -0.1362, -0.1193, -0.0846, -0.0827, -0.0978, -0.0939, -0.0825, -0.0616, -0.061, -0.0585, -0.0526, -0.0562, -0.0608, -0.0573, -0.057, -0.0583, -0.0628, -0.0586, -0.0382, -0.0444, -0.0463, -0.0481, -0.0654, -0.0655, -0.0616, -0.0697, -0.0514, -0.0359, -0.0322, -0.0302, -0.0302, -0.0256, -0.0108, 0.0, -0.0071, -0.0192, -0.0259, -0.029, -0.0315, -0.0259, -0.0285, -0.0276, -0.0422, -0.0405, -0.0501, -0.0489, -0.0437, -0.0532, -0.0483, -0.0433, -0.0433, -0.0447, -0.0434, -0.0458, -0.0501, -0.0501, -0.053, -0.0661, -0.0833, -0.0904, -0.0949, -0.0937, -0.0906, -0.0878, -0.0917, -0.0978, -0.1071, -0.1071, -0.138, -0.1347, -0.1322, -0.1333, -0.1076, -0.0976, -0.104, -0.0975, -0.0933, -0.0565, -0.0584, -0.0339, -0.0359, -0.0282, -0.0395, -0.0428, -0.0365, -0.0846, -0.1054, -0.1054, -0.0771, -0.0754, -0.0886, -0.0746, -0.069, -0.0482, -0.0409, -0.0454, -0.076, -0.0742, -0.0775, -0.0818, -0.0896, -0.0995, -0.0911, -0.0878, -0.0878, -0.1055, -0.1252, -0.1158, -0.1108, -0.1258, -0.1293, -0.1327, -0.1204, -0.1199, -0.1165, -0.1155, -0.1298, -0.1374, -0.1124, -0.1059, -0.1049, -0.1049, -0.0946, -0.1134, -0.0945, -0.0889, -0.089, -0.0935, -0.0948, -0.0682, -0.0788, -0.0549, -0.045, -0.0691, -0.0446, -0.0437, -0.052, -0.064, -0.0532, -0.0551, -0.051, -0.0202, -0.0318, -0.0061, 0.0056, 0.0053, 0.0259, 0.0246, 0.032, 0.0462, 0.0439, 0.051, 0.0426, 0.0466, 0.0581, 0.0677, 0.0811, 0.0811, 0.0794, 0.0882]}
{"unit_id": "orig:1995707727516193158", "ticker_idx": 2, "ts": "2025-12-02T04:12:55+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia (NVDA Buy)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Raises INTC TP $50→$54 on 14A customer wins (Apple, NVDA, AMD), EMIB AI accelerator pipeline, and IFS GM turnaround; endorses Buy on AAPL, NVDA, AMD, MediaTek, TSMC.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Intel-related update (18A yield, 14A external customer wins, and EMIB developments)\n\nMore to come, raise TP to $54: On Nov 28, Intel share price was up 10% owing to news of order win for Apple’s M-series in 2027. We’ve highlighted the potential order win of M-series in our earlier Intel report (Belief in the Story Takes Hold on Oct 24) and indeed, the progress is earlier than expected. We expect more to come and now expect the upcoming win in Apple (AAPL Buy) non-pro smartphone SoC based on Intel 14A for 2028; coupled with the likely engagement with Nvidia (NVDA Buy) and AMD (Buy)’s server SKUs, we believe its 14A will be a strong node. For back-end, we expect Intel to benefit from the CoWoS undersupply and potential “Made in USA” trend. As highlighted in our MediaTek (2454 TT Buy) Upgrade report on Nov 28, we expect TPU v8e to adopt EMIB, and we also expect Apple to be an important customer. Near term, despite memory concerns, server CPU sales are expected to accelerate due to increasing general purpose compute. In all, we believe the recent developments bode well for IFS in the longer term. We raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory.\n\n18A the key to build customers’ confidence: According to Intel, Panther Lake is on track to enter HVM and debut on Jan 5 at CES 2026. We believe the yield rate has reached 60-65% in November and targets 70% by end-2025. We also expect the entry-level Wildcat Lake to start HVM in 1Q26, with much higher yield rates due to smaller die area.\n\nRobust IFS customer pipeline with top-tier U.S. fabless: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple M-series on 18A-P in 2027, we expect 14A to be driven by the potential wins for: 1) Apple’s non-pro smartphone SoC in 2028 (we expect firm visibility by end of 2025); 2) Nvidia & AMD server projects in 2028; 3) Tier-1 Ethernet switch in 2028. With thoughtful cost/scale management, we expect its IFS GM turnaround by end-2026, followed by a strong ramp of 14A in 2027.\n\nBackend EMIB to win AI accelerators on “TSMC + Intel” model: We believe that EMIB emerges as an alternative to CoWoS, given the capacity constraints at TSMC (2330 TT Buy), higher design flexibility and larger reticle size (vs. OSAT’s). As said, we expect TPU v8e to adopt EMIB-T in 2027, as Google provides a very aggressive volume in 2027 and 2028. We also see high chance for Intel to win more ASICs, likely Broadcom and Apple’s. Capacity wise, we expect Intel to add equipment order in 1Q26. Aside from current Malaysia factory, it would be making sense for Intel to accelerate in New Mexico Fab to benefit from TSMC’s Arizona. In sum, we estimate its advanced packaging business will contribute >$1bn revenue in 2027E, with growth in the following years through more project wins. \n\nRisks: 1) execution in product & foundry; 2) weak macro; and 3) peer competition.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00848668855407142, "ret_signed_m1d": -0.00848668855407142, "alpha_spy_m1d": -0.006637903616566465, "alpha_c_m1d": 0.009514007793205548, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.018000696347276968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010305312728931093, "ret_signed_p1d": -0.010305312728931093, "alpha_spy_p1d": -0.013768080522191473, "alpha_c_p1d": -0.024049313802750327, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.013744001073819234, "ret_p1w": 0.019399863273101392, "ret_signed_p1w": 0.019399863273101392, "alpha_spy_p1w": 0.017184419618756053, "alpha_c_p1w": -0.00722562555825812, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02662548883135951, "ret_p1m": 0.027831972077370093, "ret_signed_p1m": 0.027831972077370093, "alpha_spy_p1m": 0.024303185900556912, "alpha_c_p1m": 0.022766761062682495, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": 0.005065211014687598, "ret_p3m": -0.02347694535389666, "ret_signed_p3m": -0.02347694535389666, "alpha_spy_p3m": -0.03299521110049708, "alpha_c_p3m": -0.15759056919759062, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": 0.13411362384369396, "ret_p6m": 0.1717367385714923, "ret_signed_p6m": 0.1717367385714923, "alpha_spy_p6m": 0.06432680998302831, "alpha_c_p6m": -0.4902084618957827, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.661945200467275, "price_path": [-0.0541, -0.0796, -0.0725, -0.059, -0.0228, -0.0236, -0.0201, -0.0204, -0.0363, -0.0616, -0.0288, -0.0264, 0.0118, -0.0167, -0.0247, -0.0208, -0.018, 0.0021, 0.0282, 0.0319, 0.0409, 0.0339, 0.0225, 0.0197, 0.0422, 0.0612, 0.0094, 0.0378, -0.0079, -0.009, 0.0019, 0.0097, 0.0065, -0.0017, -0.0065, 0.0039, 0.0265, 0.0553, 0.1078, 0.141, 0.1181, 0.1159, 0.1401, 0.095, 0.0758, 0.0365, 0.0369, 0.0969, 0.0645, 0.068, 0.0298, 0.048, 0.0283, -0.0006, 0.0279, -0.0045, -0.0142, 0.006, -0.0201, -0.0066, -0.0066, -0.0246, -0.0085, 0.0, -0.0103, 0.0106, 0.0053, 0.0226, 0.0194, 0.0128, -0.0029, -0.0354, -0.0284, -0.0206, -0.0579, -0.0403, -0.0025, 0.0123, 0.0428, 0.0395, 0.0395, 0.05, 0.0373, 0.0336, 0.0278, 0.0278, 0.0408, 0.0368, 0.0319, 0.0422, 0.0198, 0.0188, 0.0192, 0.024, 0.0093, 0.0309, 0.0263, 0.0263, -0.0186, 0.0103, 0.0187, 0.0343, 0.0277, 0.039, 0.0555, 0.061, 0.0533, 0.0229, -0.0061, -0.04, -0.0527, 0.0218, 0.0473, 0.0391, 0.0474, 0.0303, 0.0075, 0.0075, 0.0194, 0.036, 0.0355, 0.0461, 0.0557, 0.0628, 0.0778, 0.019, -0.0235, 0.0057, -0.0077, 0.0088, 0.0104, -0.02, 0.0066, 0.0183, 0.0253, 0.0094, -0.0066, 0.0098, 0.0027, -0.0057, -0.0159, -0.0482, -0.032, -0.0344, -0.0152, -0.0562, -0.0767, -0.0897, -0.0388, -0.0314, -0.0223, -0.0223, -0.0209, -0.0184, 0.0035, 0.0136, 0.0396, 0.0434, 0.0831, 0.0961, 0.0932, 0.1116, 0.1136, 0.1016, 0.1161, 0.1003, 0.1479, 0.1938, 0.1749, 0.1533, 0.0999, 0.0937, 0.0939, 0.083, 0.1454, 0.1657, 0.1861, 0.2094, 0.2168, 0.2447, 0.2993, 0.2418, 0.2253, 0.2159, 0.2316, 0.2098, 0.1868, 0.1868, 0.1842, 0.1717]}
{"unit_id": "orig:1995707727516193158", "ticker_idx": 3, "ts": "2025-12-02T04:12:55+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD (Buy)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Raises INTC TP $50→$54 on 14A customer wins (Apple, NVDA, AMD), EMIB AI accelerator pipeline, and IFS GM turnaround; endorses Buy on AAPL, NVDA, AMD, MediaTek, TSMC.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Intel-related update (18A yield, 14A external customer wins, and EMIB developments)\n\nMore to come, raise TP to $54: On Nov 28, Intel share price was up 10% owing to news of order win for Apple’s M-series in 2027. We’ve highlighted the potential order win of M-series in our earlier Intel report (Belief in the Story Takes Hold on Oct 24) and indeed, the progress is earlier than expected. We expect more to come and now expect the upcoming win in Apple (AAPL Buy) non-pro smartphone SoC based on Intel 14A for 2028; coupled with the likely engagement with Nvidia (NVDA Buy) and AMD (Buy)’s server SKUs, we believe its 14A will be a strong node. For back-end, we expect Intel to benefit from the CoWoS undersupply and potential “Made in USA” trend. As highlighted in our MediaTek (2454 TT Buy) Upgrade report on Nov 28, we expect TPU v8e to adopt EMIB, and we also expect Apple to be an important customer. Near term, despite memory concerns, server CPU sales are expected to accelerate due to increasing general purpose compute. In all, we believe the recent developments bode well for IFS in the longer term. We raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory.\n\n18A the key to build customers’ confidence: According to Intel, Panther Lake is on track to enter HVM and debut on Jan 5 at CES 2026. We believe the yield rate has reached 60-65% in November and targets 70% by end-2025. We also expect the entry-level Wildcat Lake to start HVM in 1Q26, with much higher yield rates due to smaller die area.\n\nRobust IFS customer pipeline with top-tier U.S. fabless: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple M-series on 18A-P in 2027, we expect 14A to be driven by the potential wins for: 1) Apple’s non-pro smartphone SoC in 2028 (we expect firm visibility by end of 2025); 2) Nvidia & AMD server projects in 2028; 3) Tier-1 Ethernet switch in 2028. With thoughtful cost/scale management, we expect its IFS GM turnaround by end-2026, followed by a strong ramp of 14A in 2027.\n\nBackend EMIB to win AI accelerators on “TSMC + Intel” model: We believe that EMIB emerges as an alternative to CoWoS, given the capacity constraints at TSMC (2330 TT Buy), higher design flexibility and larger reticle size (vs. OSAT’s). As said, we expect TPU v8e to adopt EMIB-T in 2027, as Google provides a very aggressive volume in 2027 and 2028. We also see high chance for Intel to win more ASICs, likely Broadcom and Apple’s. Capacity wise, we expect Intel to add equipment order in 1Q26. Aside from current Malaysia factory, it would be making sense for Intel to accelerate in New Mexico Fab to benefit from TSMC’s Arizona. In sum, we estimate its advanced packaging business will contribute >$1bn revenue in 2027E, with growth in the following years through more project wins. \n\nRisks: 1) execution in product & foundry; 2) weak macro; and 3) peer competition.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020999762582776027, "ret_signed_m1d": 0.020999762582776027, "alpha_spy_m1d": 0.022848547520280982, "alpha_c_m1d": 0.039000458930052995, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.018000696347276968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010964507294748849, "ret_signed_p1d": 0.010964507294748849, "alpha_spy_p1d": 0.007501739501488469, "alpha_c_p1d": -0.0027794937790703855, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.013744001073819234, "ret_p1w": 0.02964128164467117, "ret_signed_p1w": 0.02964128164467117, "alpha_spy_p1w": 0.02742583799032583, "alpha_c_p1w": 0.0030157928133116574, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02662548883135951, "ret_p1m": -0.005017663090001068, "ret_signed_p1m": -0.005017663090001068, "alpha_spy_p1m": -0.008546449266814249, "alpha_c_p1m": -0.010082874104688666, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": 0.005065211014687598, "ret_p3m": -0.06982902060822616, "ret_signed_p3m": -0.06982902060822616, "alpha_spy_p3m": -0.07934728635482657, "alpha_c_p3m": -0.20394264445192012, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": 0.13411362384369396, "ret_p6m": 1.3022672175162162, "ret_signed_p6m": 1.3022672175162162, "alpha_spy_p6m": 1.1948572889277522, "alpha_c_p6m": 0.6403220170489412, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.661945200467275, "price_path": [-0.2483, -0.2978, -0.2966, -0.2761, -0.2588, -0.2768, -0.2633, -0.2513, -0.2545, -0.2605, -0.2663, -0.2688, -0.2576, -0.2525, -0.2526, -0.2507, -0.2592, -0.2503, -0.2483, -0.238, -0.2114, -0.2349, -0.0536, -0.0173, 0.0944, 0.082, -0.0016, 0.0055, 0.0132, 0.1085, 0.0898, 0.0829, 0.1176, 0.1059, 0.0696, 0.0918, 0.1751, 0.2064, 0.1987, 0.2281, 0.184, 0.1899, 0.2063, 0.1617, 0.1909, 0.1043, 0.085, 0.1335, 0.1035, 0.2028, 0.152, 0.1467, 0.1175, 0.0699, 0.0386, -0.0428, -0.0532, -0.0009, -0.0423, -0.0046, -0.0046, 0.0106, 0.021, 0.0, 0.011, 0.0034, 0.0127, 0.0273, 0.0296, 0.0287, 0.0288, -0.0207, -0.0356, -0.0282, -0.0796, -0.0659, -0.0084, -0.0013, -0.0016, -0.0009, -0.0009, -0.0012, 0.0017, 0.0005, -0.005, -0.005, 0.0382, 0.0271, -0.0041, -0.0243, -0.0491, -0.0561, -0.0351, 0.0266, 0.0388, 0.0589, 0.0771, 0.0771, 0.0775, 0.1606, 0.1788, 0.2065, 0.1676, 0.1709, 0.1742, 0.1716, 0.0998, 0.1442, 0.1248, -0.0699, -0.1056, -0.0316, 0.0035, -0.0078, -0.0077, -0.0432, -0.0368, -0.0368, -0.0565, -0.0702, -0.0551, -0.0701, -0.0866, -0.0065, -0.0203, -0.0537, -0.0698, -0.0772, -0.1129, -0.0612, -0.0734, -0.106, -0.0584, -0.0558, -0.0484, -0.0813, -0.1015, -0.0867, -0.0879, -0.0733, -0.0463, -0.0646, -0.0584, -0.0459, 0.0234, -0.0533, -0.0616, -0.0892, -0.0549, -0.0234, 0.0105, 0.0105, 0.023, 0.0292, 0.077, 0.0994, 0.1385, 0.1468, 0.185, 0.1992, 0.2928, 0.2934, 0.2774, 0.3217, 0.4099, 0.4186, 0.6159, 0.5547, 0.5016, 0.5662, 0.647, 0.6751, 0.5868, 0.6505, 0.9578, 0.8977, 1.1148, 1.1315, 1.0827, 1.0698, 1.0893, 0.9704, 0.9559, 0.9237, 1.0794, 1.0888, 1.172, 1.172, 1.3411, 1.3023]}
{"unit_id": "orig:1995707727516193158", "ticker_idx": 4, "ts": "2025-12-02T04:12:55+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek (2454 TT Buy)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Raises INTC TP $50→$54 on 14A customer wins (Apple, NVDA, AMD), EMIB AI accelerator pipeline, and IFS GM turnaround; endorses Buy on AAPL, NVDA, AMD, MediaTek, TSMC.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Intel-related update (18A yield, 14A external customer wins, and EMIB developments)\n\nMore to come, raise TP to $54: On Nov 28, Intel share price was up 10% owing to news of order win for Apple’s M-series in 2027. We’ve highlighted the potential order win of M-series in our earlier Intel report (Belief in the Story Takes Hold on Oct 24) and indeed, the progress is earlier than expected. We expect more to come and now expect the upcoming win in Apple (AAPL Buy) non-pro smartphone SoC based on Intel 14A for 2028; coupled with the likely engagement with Nvidia (NVDA Buy) and AMD (Buy)’s server SKUs, we believe its 14A will be a strong node. For back-end, we expect Intel to benefit from the CoWoS undersupply and potential “Made in USA” trend. As highlighted in our MediaTek (2454 TT Buy) Upgrade report on Nov 28, we expect TPU v8e to adopt EMIB, and we also expect Apple to be an important customer. Near term, despite memory concerns, server CPU sales are expected to accelerate due to increasing general purpose compute. In all, we believe the recent developments bode well for IFS in the longer term. We raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory.\n\n18A the key to build customers’ confidence: According to Intel, Panther Lake is on track to enter HVM and debut on Jan 5 at CES 2026. We believe the yield rate has reached 60-65% in November and targets 70% by end-2025. We also expect the entry-level Wildcat Lake to start HVM in 1Q26, with much higher yield rates due to smaller die area.\n\nRobust IFS customer pipeline with top-tier U.S. fabless: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple M-series on 18A-P in 2027, we expect 14A to be driven by the potential wins for: 1) Apple’s non-pro smartphone SoC in 2028 (we expect firm visibility by end of 2025); 2) Nvidia & AMD server projects in 2028; 3) Tier-1 Ethernet switch in 2028. With thoughtful cost/scale management, we expect its IFS GM turnaround by end-2026, followed by a strong ramp of 14A in 2027.\n\nBackend EMIB to win AI accelerators on “TSMC + Intel” model: We believe that EMIB emerges as an alternative to CoWoS, given the capacity constraints at TSMC (2330 TT Buy), higher design flexibility and larger reticle size (vs. OSAT’s). As said, we expect TPU v8e to adopt EMIB-T in 2027, as Google provides a very aggressive volume in 2027 and 2028. We also see high chance for Intel to win more ASICs, likely Broadcom and Apple’s. Capacity wise, we expect Intel to add equipment order in 1Q26. Aside from current Malaysia factory, it would be making sense for Intel to accelerate in New Mexico Fab to benefit from TSMC’s Arizona. In sum, we estimate its advanced packaging business will contribute >$1bn revenue in 2027E, with growth in the following years through more project wins. \n\nRisks: 1) execution in product & foundry; 2) weak macro; and 3) peer competition.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.021201454756768934, "ret_signed_m1d": 0.021201454756768934, "alpha_spy_m1d": 0.02305023969427389, "alpha_c_m1d": 0.0392021511040459, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.018000696347276968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010600683407837042, "ret_signed_p1d": -0.010600683407837042, "alpha_spy_p1d": -0.014063451201097421, "alpha_c_p1d": -0.024344684481656276, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.013744001073819234, "ret_p1w": 0.0035335611359457175, "ret_signed_p1w": 0.0035335611359457175, "alpha_spy_p1w": 0.0013181174816003782, "alpha_c_p1w": -0.023091927695413794, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02662548883135951, "ret_p1m": 0.01060068340783693, "ret_signed_p1m": 0.01060068340783693, "alpha_spy_p1m": 0.007071897231023749, "alpha_c_p1m": 0.005535472393149332, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": 0.005065211014687598, "ret_p3m": 0.40120415982484947, "ret_signed_p3m": 0.40120415982484947, "alpha_spy_p3m": 0.39168589407824905, "alpha_c_p3m": 0.2670905359811555, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": 0.13411362384369396, "ret_p6m": 2.342718406985759, "ret_signed_p6m": 2.342718406985759, "alpha_spy_p6m": 2.235308478397295, "alpha_c_p6m": 1.680773206518484, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.661945200467275, "price_path": [-0.0212, 0.0141, 0.0106, 0.0671, 0.053, 0.0459, 0.0495, 0.0495, 0.0848, 0.0671, 0.0671, 0.0177, -0.0071, -0.0283, -0.0601, -0.0495, -0.0742, -0.0742, -0.0707, -0.0919, -0.0954, -0.0671, -0.0671, -0.0636, -0.0565, -0.0495, -0.0495, -0.0707, -0.0777, -0.0707, -0.0601, -0.0601, -0.053, -0.053, -0.0601, -0.0848, -0.0848, -0.0565, -0.0707, -0.0813, -0.0742, -0.0742, -0.0813, -0.0707, -0.0601, -0.0883, -0.1095, -0.1166, -0.1237, -0.1201, -0.1201, -0.1307, -0.1307, -0.1731, -0.1802, -0.1625, -0.1908, -0.1873, -0.1625, -0.0813, -0.053, -0.0141, 0.0212, 0.0, -0.0106, -0.0071, 0.0071, 0.0177, 0.0035, 0.0318, -0.0141, -0.0071, 0.0035, 0.0035, 0.0071, 0.0035, -0.0035, -0.0106, -0.0177, -0.0247, -0.0247, -0.0212, 0.0035, 0.0035, 0.0106, 0.0106, 0.0389, 0.0777, 0.0698, 0.0734, 0.041, 0.023, 0.041, 0.0698, 0.0806, 0.0734, 0.0842, 0.0698, 0.0698, 0.0554, 0.0698, 0.1743, 0.2751, 0.2787, 0.2823, 0.2823, 0.2679, 0.2283, 0.2931, 0.2967, 0.2751, 0.2319, 0.3184, 0.3292, 0.3364, 0.3364, 0.3364, 0.3364, 0.3364, 0.3364, 0.3364, 0.3364, 0.2931, 0.3148, 0.3472, 0.4012, 0.4012, 0.3688, 0.3076, 0.2391, 0.2787, 0.2715, 0.1995, 0.2283, 0.2715, 0.2859, 0.2391, 0.2319, 0.2463, 0.2463, 0.2103, 0.2247, 0.1707, 0.1671, 0.1671, 0.1455, 0.1419, 0.0878, 0.0734, 0.0554, 0.0554, 0.0554, 0.0554, 0.059, 0.1383, 0.1347, 0.1347, 0.1671, 0.2391, 0.2895, 0.3652, 0.3868, 0.3688, 0.5057, 0.6533, 0.5957, 0.7542, 0.7542, 0.8839, 0.8551, 0.8803, 0.8803, 1.0676, 1.2729, 1.471, 1.4638, 1.6151, 1.7952, 1.6655, 1.5178, 1.453, 1.3485, 1.4494, 1.2729, 1.3269, 1.5575, 1.7808, 2.0582, 2.0726, 2.3427]}
{"unit_id": "orig:1995707727516193158", "ticker_idx": 5, "ts": "2025-12-02T04:12:55+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC (2330 TT Buy)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Raises INTC TP $50→$54 on 14A customer wins (Apple, NVDA, AMD), EMIB AI accelerator pipeline, and IFS GM turnaround; endorses Buy on AAPL, NVDA, AMD, MediaTek, TSMC.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GFHK) Intel-related update (18A yield, 14A external customer wins, and EMIB developments)\n\nMore to come, raise TP to $54: On Nov 28, Intel share price was up 10% owing to news of order win for Apple’s M-series in 2027. We’ve highlighted the potential order win of M-series in our earlier Intel report (Belief in the Story Takes Hold on Oct 24) and indeed, the progress is earlier than expected. We expect more to come and now expect the upcoming win in Apple (AAPL Buy) non-pro smartphone SoC based on Intel 14A for 2028; coupled with the likely engagement with Nvidia (NVDA Buy) and AMD (Buy)’s server SKUs, we believe its 14A will be a strong node. For back-end, we expect Intel to benefit from the CoWoS undersupply and potential “Made in USA” trend. As highlighted in our MediaTek (2454 TT Buy) Upgrade report on Nov 28, we expect TPU v8e to adopt EMIB, and we also expect Apple to be an important customer. Near term, despite memory concerns, server CPU sales are expected to accelerate due to increasing general purpose compute. In all, we believe the recent developments bode well for IFS in the longer term. We raise TP from $50 to $54, now based on 2x FY27E BVPS, reflecting the better visibility of earnings improvement trajectory.\n\n18A the key to build customers’ confidence: According to Intel, Panther Lake is on track to enter HVM and debut on Jan 5 at CES 2026. We believe the yield rate has reached 60-65% in November and targets 70% by end-2025. We also expect the entry-level Wildcat Lake to start HVM in 1Q26, with much higher yield rates due to smaller die area.\n\nRobust IFS customer pipeline with top-tier U.S. fabless: We still expect Intel to ramp 18A and 18A-P at small scale in 2026 to ensure yield and quality. That said, the external customer is expected to start by end-2026. In addition to the aforementioned Apple M-series on 18A-P in 2027, we expect 14A to be driven by the potential wins for: 1) Apple’s non-pro smartphone SoC in 2028 (we expect firm visibility by end of 2025); 2) Nvidia & AMD server projects in 2028; 3) Tier-1 Ethernet switch in 2028. With thoughtful cost/scale management, we expect its IFS GM turnaround by end-2026, followed by a strong ramp of 14A in 2027.\n\nBackend EMIB to win AI accelerators on “TSMC + Intel” model: We believe that EMIB emerges as an alternative to CoWoS, given the capacity constraints at TSMC (2330 TT Buy), higher design flexibility and larger reticle size (vs. OSAT’s). As said, we expect TPU v8e to adopt EMIB-T in 2027, as Google provides a very aggressive volume in 2027 and 2028. We also see high chance for Intel to win more ASICs, likely Broadcom and Apple’s. Capacity wise, we expect Intel to add equipment order in 1Q26. Aside from current Malaysia factory, it would be making sense for Intel to accelerate in New Mexico Fab to benefit from TSMC’s Arizona. In sum, we estimate its advanced packaging business will contribute >$1bn revenue in 2027E, with growth in the following years through more project wins. \n\nRisks: 1) execution in product & foundry; 2) weak macro; and 3) peer competition.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013986044631606687, "ret_signed_m1d": -0.013986044631606687, "alpha_spy_m1d": -0.012137259694101732, "alpha_c_m1d": 0.004014651715670281, "bench_spy_m1d": -0.0018487849375049548, "bench_c_m1d": -0.018000696347276968, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013985958703768286, "ret_signed_p1d": 0.013985958703768286, "alpha_spy_p1d": 0.010523190910507907, "alpha_c_p1d": 0.00024195762994905223, "bench_spy_p1d": 0.0034627677932603795, "bench_c_p1d": 0.013744001073819234, "ret_p1w": 0.034965068615097517, "ret_signed_p1w": 0.034965068615097517, "alpha_spy_p1w": 0.03274962496075218, "alpha_c_p1w": 0.008339579783738005, "bench_spy_p1w": 0.0022154436543453393, "bench_c_p1w": 0.02662548883135951, "ret_p1m": 0.08752910375889611, "ret_signed_p1m": 0.08752910375889611, "alpha_spy_p1m": 0.08400031758208293, "alpha_c_p1m": 0.08246389274420851, "bench_spy_p1m": 0.003528786176813181, "bench_c_p1m": 0.005065211014687598, "ret_p3m": 0.3997552087737135, "ret_signed_p3m": 0.3997552087737135, "alpha_spy_p3m": 0.3902369430271131, "alpha_c_p3m": 0.26564158493001955, "bench_spy_p3m": 0.009518265746600418, "bench_c_p3m": 0.13411362384369396, "ret_p6m": 0.619017961667935, "ret_signed_p6m": 0.619017961667935, "alpha_spy_p6m": 0.511608033079471, "alpha_c_p6m": -0.04292723879934002, "bench_spy_p6m": 0.10740992858846399, "bench_c_p6m": 0.661945200467275, "price_path": [-0.192, -0.1781, -0.1781, -0.1642, -0.1468, -0.1363, -0.1224, -0.1259, -0.1049, -0.1154, -0.1014, -0.1154, -0.0944, -0.0629, -0.0629, -0.0769, -0.0909, -0.0909, -0.0874, -0.0734, -0.0455, -0.021, -0.021, 0.0035, -0.0105, 0.007, 0.007, -0.0105, -0.0035, 0.0245, 0.0385, 0.014, 0.035, 0.035, 0.021, 0.014, 0.014, 0.035, 0.0315, 0.0524, 0.0524, 0.049, 0.0559, 0.0524, 0.021, 0.0245, 0.021, 0.0315, 0.0245, 0.0315, 0.021, 0.0, 0.0105, -0.0175, -0.0245, 0.0175, -0.0315, -0.0385, -0.0105, 0.007, 0.0035, 0.007, -0.014, 0.0, 0.014, 0.0105, 0.021, 0.0455, 0.035, 0.0524, 0.0314, 0.0384, 0.0174, 0.0068, 0.0033, 0.0033, 0.0033, 0.0279, 0.0454, 0.0489, 0.0489, 0.0595, 0.0735, 0.0665, 0.0875, 0.0875, 0.1121, 0.1717, 0.1963, 0.1752, 0.1822, 0.1787, 0.1858, 0.1998, 0.1998, 0.1858, 0.2208, 0.2349, 0.2454, 0.2208, 0.2349, 0.2419, 0.2314, 0.2489, 0.277, 0.2664, 0.2454, 0.2384, 0.2629, 0.2524, 0.2384, 0.2489, 0.2735, 0.3191, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3436, 0.3331, 0.3787, 0.4138, 0.3998, 0.3998, 0.3857, 0.3577, 0.3085, 0.3331, 0.3261, 0.27, 0.298, 0.3612, 0.3226, 0.3085, 0.2945, 0.3163, 0.341, 0.3023, 0.2952, 0.2741, 0.2741, 0.2987, 0.2952, 0.2811, 0.253, 0.2389, 0.3058, 0.2741, 0.2741, 0.2741, 0.3093, 0.3726, 0.3762, 0.4078, 0.4008, 0.4466, 0.4642, 0.4677, 0.429, 0.4254, 0.443, 0.443, 0.4642, 0.5381, 0.5944, 0.5592, 0.5345, 0.5029, 0.5029, 0.6014, 0.5838, 0.5838, 0.6261, 0.612, 0.5733, 0.5873, 0.5627, 0.5979, 0.5944, 0.5768, 0.5521, 0.5381, 0.5697, 0.5873, 0.6261, 0.5979, 0.619]}
{"unit_id": "quote:2000702268266700870", "ticker_idx": 0, "ts": "2025-12-15T22:59:26+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect supply shortages to persist for at least the next several quarters, and potentially through 2027. We also anticipate this will lead to sequential (QoQ) pricing increases over the coming quarters, with companies having higher exposure to channel and retail markets relative to fixed contract pricing (including SanDisk) likely realizing greater upside", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman NAND bull thesis: supply shortfalls into 2027, QoQ price increases ahead; SanDisk flagged as top upside pick on Google eSSD contract win.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "Goldman Sachs NAND Commentary: Supply/Demand Significantly Tightening; Incremental SSD Data Points\n\nNAND supply and demand has tightened significantly over the last two quarters, driven by a combination of production discipline from the supply side—including Samsung, SK Hynix, Micron, and YMTC—and robust demand (from legacy markets such as smartphones and the AI inference sector). We have not heard reports of plans to add new NAND wafer capacity; instead, typical technology migration via upgrades appears to be the consensus plan across suppliers.\n\nSuppliers generally project 2026 NAND bit supply growth to be in the mid-10% range. Bit demand growth has the potential to reach the high-teens to 20% range, primarily driven by data center applications, with recent surges in demand for AI inference and HDD replacement. Consequently, we expect supply shortages to persist for at least the next several quarters, and potentially through 2027. We also anticipate this will lead to sequential (QoQ) pricing increases over the coming quarters, with companies having higher exposure to channel and retail markets relative to fixed contract pricing (including SanDisk) likely realizing greater upside.\n\nRegarding the Enterprise SSD (eSSD) market, hyperscalers and the enterprise sector continue to drive demand. Our partner checks indicate that SanDisk has secured a second hyperscaler contract for eSSD from Google. Volume shipments are scheduled to ramp up in 1H 2026, with the potential to surpass the existing customer (Meta) in terms of volume.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": "Goldman Sachs DRAM Commentary: Supply Growth Modest, Demand Continues to Significantly Outpace Supply\n\nDemand for both High Bandwidth Memory (HBM) and conventional DRAM continues to outstrip supply. Overall, full-year 2026 HBM volume and pricing negotiations for most DRAM https://t.co/crvJQ3GMIO", "parent_author": null, "ret_m1d": 0.021350362425539116, "ret_signed_m1d": 0.021350362425539116, "alpha_spy_m1d": 0.019837278578511697, "alpha_c_m1d": 0.01158228071102596, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.009768081714513155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03685541497678857, "ret_signed_p1d": 0.03685541497678857, "alpha_spy_p1d": 0.03958771257206728, "alpha_c_p1d": 0.03502833509817238, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": 0.001827079878616189, "ret_p1w": 0.19408534642221564, "ret_signed_p1w": 0.19408534642221564, "alpha_spy_p1w": 0.1850896211095132, "alpha_c_p1w": 0.17243974013427388, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.021645606287941765, "ret_p1m": 0.930995229540206, "ret_signed_p1m": 0.930995229540206, "alpha_spy_p1m": 0.908827739634495, "alpha_c_p1m": 0.9000593865813344, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.03093584295887153, "ret_p3m": 2.0654382636953437, "ret_signed_p3m": 2.0654382636953437, "alpha_spy_p3m": 2.084097436960036, "alpha_c_p3m": 2.095311320783203, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": -0.029873057087859145, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5345, -0.5102, -0.4937, -0.4902, -0.4729, -0.5055, -0.5329, -0.5189, -0.4378, -0.4442, -0.4, -0.385, -0.3639, -0.3998, -0.4009, -0.3467, -0.3576, -0.4209, -0.3332, -0.3694, -0.2852, -0.2853, -0.3057, -0.2667, -0.2605, -0.2721, -0.1725, -0.0778, -0.1257, -0.1305, 0.0123, -0.03, -0.0126, 0.0255, -0.0362, 0.0725, 0.0288, 0.1863, 0.3273, 0.3453, 0.4024, 0.2066, 0.259, 0.3171, 0.2133, 0.2184, -0.0293, -0.0079, 0.1243, 0.0923, 0.0652, 0.0652, 0.1061, 0.0411, 0.0172, -0.0371, 0.0567, 0.1318, 0.1169, 0.0871, 0.1535, 0.1969, 0.0214, 0.0, 0.0369, 0.0246, 0.0871, 0.177, 0.1941, 0.2132, 0.2388, 0.2388, 0.2387, 0.2099, 0.19, 0.1759, 0.1759, 0.3635, 0.3577, 0.732, 0.7514, 0.6572, 0.8696, 0.9283, 0.931, 0.9211, 1.0272, 1.0489, 1.0489, 1.2446, 1.4832, 1.4939, 1.3472, 1.3322, 1.3849, 1.6137, 1.6715, 1.8546, 2.2954, 2.4453, 1.8957, 1.8543, 1.9621, 1.89, 1.6831, 1.9689, 2.1223, 2.1038, 2.1038, 1.9256, 1.9742, 2.0767, 2.2197, 2.3016, 2.163, 2.1326, 2.2293, 2.1474, 2.0667, 1.8009, 1.9676, 1.8018, 1.6122, 1.9164, 2.0658, 2.2468, 2.0654, 2.2775, 2.4856, 2.5675, 2.7335, 2.8247, 2.5157, 2.4799, 2.4799, 2.3579, 1.9879, 2.0506, 1.836, 2.1473, 2.4316, 2.4755, 2.4755, 2.5896, 2.5211, 2.8683, 3.2184, 3.2194, 3.7184, 3.6786, 3.4173, 3.5548, 3.5623, 3.5228, 3.4756, 3.85, 3.619, 3.9037, 4.3014, 3.9653, 4.2718, 4.4318, 4.88, 5.2211, 5.9665, 5.9846, 5.6377, 6.7393, 6.6661, 6.1928, 6.1691, 5.8496, 5.9729, 5.6033, 5.8524, 5.8983, 6.6398, 6.325, 6.325, 6.8741, 6.8761, 7.1322, 7.3964, null, null, null, null, null, null, null]}
{"unit_id": "quote:2000702268266700870", "ticker_idx": 1, "ts": "2025-12-15T22:59:26+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "supply shortages to persist for at least the next several quarters, and potentially through 2027. We also anticipate this will lead to sequential (QoQ) pricing increases over the coming quarters", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman NAND bull thesis: supply shortfalls into 2027, QoQ price increases ahead; SanDisk flagged as top upside pick on Google eSSD contract win.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Goldman Sachs NAND Commentary: Supply/Demand Significantly Tightening; Incremental SSD Data Points\n\nNAND supply and demand has tightened significantly over the last two quarters, driven by a combination of production discipline from the supply side—including Samsung, SK Hynix, Micron, and YMTC—and robust demand (from legacy markets such as smartphones and the AI inference sector). We have not heard reports of plans to add new NAND wafer capacity; instead, typical technology migration via upgrades appears to be the consensus plan across suppliers.\n\nSuppliers generally project 2026 NAND bit supply growth to be in the mid-10% range. Bit demand growth has the potential to reach the high-teens to 20% range, primarily driven by data center applications, with recent surges in demand for AI inference and HDD replacement. Consequently, we expect supply shortages to persist for at least the next several quarters, and potentially through 2027. We also anticipate this will lead to sequential (QoQ) pricing increases over the coming quarters, with companies having higher exposure to channel and retail markets relative to fixed contract pricing (including SanDisk) likely realizing greater upside.\n\nRegarding the Enterprise SSD (eSSD) market, hyperscalers and the enterprise sector continue to drive demand. Our partner checks indicate that SanDisk has secured a second hyperscaler contract for eSSD from Google. Volume shipments are scheduled to ramp up in 1H 2026, with the potential to surpass the existing customer (Meta) in terms of volume.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": "Goldman Sachs DRAM Commentary: Supply Growth Modest, Demand Continues to Significantly Outpace Supply\n\nDemand for both High Bandwidth Memory (HBM) and conventional DRAM continues to outstrip supply. Overall, full-year 2026 HBM volume and pricing negotiations for most DRAM https://t.co/crvJQ3GMIO", "parent_author": null, "ret_m1d": 0.03629691846226746, "ret_signed_m1d": 0.03629691846226746, "alpha_spy_m1d": 0.03478383461524004, "alpha_c_m1d": 0.032839844498511406, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020442521958762526, "ret_signed_p1d": -0.020442521958762526, "alpha_spy_p1d": -0.017710224363483816, "alpha_c_p1d": -0.0177221787097376, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.11210779234921264, "ret_signed_p1w": 0.11210779234921264, "alpha_spy_p1w": 0.10311206703651021, "alpha_c_p1w": 0.08966730267701828, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4988530312411935, "ret_signed_p1m": 0.4988530312411935, "alpha_spy_p1m": 0.4766855413354826, "alpha_c_p1m": 0.38593640603171353, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 1.1131112499474292, "ret_signed_p3m": 1.1131112499474292, "alpha_spy_p3m": 1.1317704232121213, "alpha_c_p3m": 1.0098590431600514, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4068, -0.3808, -0.3825, -0.3694, -0.3586, -0.3662, -0.3761, -0.3962, -0.3614, -0.3566, -0.3304, -0.2908, -0.2663, -0.2678, -0.2816, -0.2621, -0.2604, -0.2628, -0.2426, -0.2627, -0.2243, -0.1873, -0.1933, -0.1614, -0.1698, -0.1665, -0.1434, -0.0677, -0.0389, -0.0575, 0.0055, 0.0232, 0.0281, 0.0738, 0.0219, 0.0446, 0.0597, 0.0954, 0.1789, 0.1798, 0.1939, 0.141, 0.0651, 0.1205, 0.0504, 0.0564, 0.0081, -0.04, 0.0036, -0.0015, -0.0255, -0.0024, 0.008, -0.0096, 0.0043, -0.0156, -0.0044, 0.0344, 0.0692, 0.0558, 0.0767, 0.0747, 0.0363, 0.0, -0.0204, 0.0018, 0.0381, 0.0627, 0.1121, 0.1152, 0.144, 0.1481, 0.1752, 0.1823, 0.1763, 0.1674, 0.1674, 0.2561, 0.2973, 0.4163, 0.4505, 0.4286, 0.4757, 0.4895, 0.4989, 0.4907, 0.5317, 0.5945, 0.6082, 0.6336, 0.737, 0.7877, 0.7515, 0.731, 0.8183, 0.9163, 0.9388, 1.0207, 1.0151, 1.1429, 1.0023, 0.9274, 0.9471, 0.9513, 0.8949, 0.9814, 1.0966, 1.1294, 1.1175, 1.0734, 1.0803, 1.116, 1.1588, 1.1753, 1.2152, 1.2123, 1.2732, 1.2369, 1.2276, 1.0329, 0.9868, 1.0439, 0.9637, 0.9383, 1.0778, 1.1766, 1.1131, 1.1534, 1.2751, 1.2949, 1.4259, 1.3762, 1.2921, 1.198, 1.2083, 1.2037, 1.04, 1.0345, 0.9282, 0.9421, 1.1591, 1.1054, 1.1534, 1.2235, 1.2372, 1.4946, 1.5541, 1.6209, 1.7454, 1.8848, 1.7889, 1.8683, 1.7827, 1.7797, 1.8491, 1.9977, 1.9734, 2.0149, 2.1722, 2.1044, 2.1828, 2.2276, 2.3133, 2.4917, 2.6945, 2.8404, 2.9362, 3.2702, 3.4308, 3.2866, 3.4694, 3.3587, 3.1493, 3.219, 3.2019, 3.2731, 3.6505, 3.6099, 3.7849, 4.0044, 4.0752, 4.1405, 4.3823, null, null, null, null, null, null, null]}
{"unit_id": "quote:1983085352282882479", "ticker_idx": 0, "ts": "2025-10-28T08:16:06+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nvidia and AMD, armed with huge cash reserves, will continue to push partners to invest in AI capex... The biggest beneficiaries of all this will be the semiconductor supply chain", "tweet_type": "analysis", "timeline": "months", "summary": "Long NVDA and AMD as the primary beneficiaries of GPU capex supercycle as hyperscalers slash headcount to fund AI infrastructure buildout.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Amazon laid off 30,000 employees today, an even larger cut than during the industry contraction in 2022. The reason is simple:\n\nThey don’t have enough capex left to buy GPUs.\nAs a result, AWS growth has slowed, the market has punished them harshly, and now they must cut salaries to save money for GPU purchases—so the financials look better and they can tell a story of “AWS growth bottoming out.”\n\nEvery internet company software engineer (SDE) should buy Nvidia/AMD stock as a hedge—compensating for the risk of being squeezed out of the value chain by GPUs.\n\nEntering 2024–2025, the main factor behind weak employment for American SDEs is no longer the massive over-expansion of 2021, nor the competition from lower-wage overseas engineering centers, and not yet the reduced demand caused by AI efficiency gains.\n\nA new boss has arrived: GPU capex.\nGPU capex is creating a strange “prosperous depression” inside internet companies:\n\nThe company’s revenue growth looks strong, stock prices keep rising—but wage expenses have become immovable constraints for management. Everyone worries about their jobs. Continuous layoffs increase the workload for those who remain. Morale collapses. It feels like the Great Depression all over again.\n\nThis isn’t a traditional recession. It’s capital’s radical redistribution between manpower and compute power.\n\nAmazon’s 30,000-person layoff has been rumored for two months. The mid-year performance review, usually in July, was delayed to mid- or late August. The return-to-office (RTO) policy also served as a major excuse for the cuts. The AGI group will remain untouched; PXT, Devices & Services, and Operations will be the hardest hit. By convention, AWS will likely announce its cuts later—probably after AWS re:Invent—to squeeze every bit of output first.\n\nMeanwhile, AWS’s Q2 backlog reached $195 billion, up 25% YoY.\n\nThis shows customers want to buy but AWS can’t deliver—demand remains scorching hot, and they simply can’t buy GPUs fast enough.\n\nIn an era where AI server supply can’t keep up with explosive demand, shifting opex (wages) to capex directly boosts company performance. Capital will ruthlessly punish any CSP or hyperscaler that fails to fully embrace this path.\nMeta has quietly entered a “5% layoff every six months” rhythm. Recently, it cut 600 people in its AI org and also removed many directors across departments—same logic again: AI datacenter capacity is insufficient. In the past year, Meta revised up its 18-month capacity plan three times. Each time they thought they had overestimated demand—only to painfully realize a few months later that they had underestimated it.\n\nDoes this mean internet companies no longer need people?\n\nOf course not. But after the hiring budget is slashed, they’re forced to squeeze productivity internally to compensate.\n\nBig tech companies are now pulling every possible lever: building internal tools with agent functions, encouraging “one-click deployment vibes,” setting KPIs for AI usage rates, requiring departments to report AI adoption progress and use cases, and mandating periodic peer learning sessions.\n\nAll these frantic moves aim for just one modest goal: ~20% efficiency improvement.\nWhat happens next?\n\nTo maintain growth and competitiveness—once efficiency gains plateau and layoffs hit the bone, when opex yields no more savings—the next step will be to sacrifice cash flow. Some, like Oracle, may even take on debt to sustain growth.\n\nNvidia and AMD, armed with huge cash reserves, will continue to push partners to invest in AI capex—just like OpenAI has done.\n\nThe biggest beneficiaries of all this will be the semiconductor supply chain.\n\nA new normal may emerge: semiconductor companies’ profit margins surpass those of internet firms.\n\nBut they also bear the greatest risk:\n\nOnce VCs or hyperscalers notice token demand slowing—or even just growth decelerating—they’ll ruthlessly cut orders.\n\nThe transmission of that shock will be far faster than semiconductor production cycles.\nWhen might that point arrive?\n\nOne reference indicator: when enterprise adoption reaches ~50%.\nDuring the March 2000 internet bubble burst, U.S. internet penetration was around 52% (some data say 43%).\nCurrently, major internet companies’ GenAI daily-user penetration is rising from 50% toward 90%, while overall enterprise AI adoption is still under 10%.\n\nThat means growth is safe for now. Historically, the fastest phase of any technological revolution is when corporate adoption moves from 10% to 50%.\nThe Cisco bubble won’t repeat itself so simply.\n\nThis time, information flow is vastly richer. There will always be enough skeptics warning about bubbles—ensuring that when it does burst, it won’t be nearly as catastrophic.", "reply_text": null, "parent_text": null, "quoted_text": "今天Amazon裁员3万人，比2022年行业收缩期裁的还厉害，原因很简单\n\n买GPU的钱capex不够，导致AWS增速下降，被市场严厉的惩罚了，只能砍工资来省钱买GPU，让财报好看一点，营造一个AWS增速触底的故事\n\n每一个互联网公司的SDE打工人，都应该买入Nvidia/amd作为风险对冲，弥补自己被GPU挤出价值链的风险 https://t.co/nAzA2nsAfi", "parent_author": null, "ret_m1d": -0.04745561765443562, "ret_signed_m1d": -0.04745561765443562, "alpha_spy_m1d": -0.04480665434163944, "alpha_c_m1d": -0.03872550977582356, "bench_spy_m1d": -0.002648963312796182, "bench_c_m1d": -0.008730107878612059, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029896025705414164, "ret_signed_p1d": 0.029896025705414164, "alpha_spy_p1d": 0.0294157280066083, "alpha_c_p1d": 0.014749096317175914, "bench_spy_p1d": 0.0004802976988058649, "bench_c_p1d": 0.01514692938823825, "ret_p1w": -0.011640098696935719, "ret_signed_p1w": -0.011640098696935719, "alpha_spy_p1w": 0.005563600238600919, "alpha_c_p1w": 0.016863595614398097, "bench_spy_p1w": -0.017203698935536638, "bench_c_p1w": -0.028503694311333816, "ret_p1m": -0.10331796482206068, "ret_signed_p1m": -0.10331796482206068, "alpha_spy_p1m": -0.09257657493557925, "alpha_c_p1m": -0.0609065625349936, "bench_spy_p1m": -0.010741389886481434, "bench_c_p1m": -0.04241140228706708, "ret_p3m": -0.06640583011519707, "ret_signed_p3m": -0.06640583011519707, "alpha_spy_p3m": -0.07252837499971387, "alpha_c_p3m": -0.17167675752054623, "bench_spy_p3m": 0.006122544884516801, "bench_c_p3m": 0.10527092740534916, "ret_p6m": 0.00742291806976203, "ret_signed_p6m": 0.00742291806976203, "alpha_spy_p6m": -0.03362100397549139, "alpha_c_p6m": -0.3098465120519429, "bench_spy_p6m": 0.04104392204525342, "bench_c_p6m": 0.31726943012170494, "price_path": [-0.1153, -0.1359, -0.1047, -0.1133, -0.1075, -0.1008, -0.0912, -0.0944, -0.0889, -0.0968, -0.0946, -0.1024, -0.0947, -0.1263, -0.1275, -0.1296, -0.1147, -0.1056, -0.0959, -0.0967, -0.1038, -0.1336, -0.1336, -0.1505, -0.1513, -0.1461, -0.1692, -0.1628, -0.1506, -0.1179, -0.1187, -0.1155, -0.1158, -0.1301, -0.1529, -0.1233, -0.1212, -0.0867, -0.1124, -0.1197, -0.1161, -0.1136, -0.0954, -0.0719, -0.0686, -0.0604, -0.0667, -0.0771, -0.0795, -0.0593, -0.0421, -0.0889, -0.0632, -0.1045, -0.1055, -0.0956, -0.0886, -0.0915, -0.0988, -0.1032, -0.0939, -0.0735, -0.0475, 0.0, 0.0299, 0.0093, 0.0073, 0.0291, -0.0116, -0.029, -0.0644, -0.0641, -0.0098, -0.0391, -0.036, -0.0705, -0.054, -0.0718, -0.0978, -0.0722, -0.1014, -0.1102, -0.0919, -0.1155, -0.1033, -0.1033, -0.1195, -0.105, -0.0973, -0.1067, -0.0877, -0.0926, -0.077, -0.0798, -0.0858, -0.0999, -0.1293, -0.123, -0.1159, -0.1496, -0.1337, -0.0996, -0.0862, -0.0587, -0.0617, -0.0617, -0.0522, -0.0637, -0.0671, -0.0722, -0.0722, -0.0605, -0.0642, -0.0685, -0.0592, -0.0795, -0.0804, -0.08, -0.0757, -0.0889, -0.0695, -0.0736, -0.0736, -0.1142, -0.088, -0.0805, -0.0664, -0.0724, -0.0622, -0.0473, -0.0423, -0.0492, -0.0767, -0.1029, -0.1335, -0.145, -0.0776, -0.0546, -0.0621, -0.0546, -0.07, -0.0906, -0.0906, -0.0798, -0.0649, -0.0653, -0.0557, -0.0471, -0.0406, -0.0272, -0.0802, -0.1185, -0.0922, -0.1043, -0.0894, -0.0879, -0.1154, -0.0914, -0.0808, -0.0745, -0.0889, -0.1033, -0.0885, -0.0949, -0.1025, -0.1117, -0.1408, -0.1262, -0.1284, -0.1111, -0.1481, -0.1666, -0.1783, -0.1324, -0.1257, -0.1175, -0.1175, -0.1163, -0.114, -0.0942, -0.0851, -0.0616, -0.0582, -0.0224, -0.0106, -0.0132, 0.0033, 0.0052, -0.0056, 0.0074]}
{"unit_id": "quote:1983085352282882479", "ticker_idx": 1, "ts": "2025-10-28T08:16:06+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Every internet company software engineer (SDE) should buy Nvidia/AMD stock as a hedge", "tweet_type": "analysis", "timeline": "months", "summary": "Long NVDA and AMD as the primary beneficiaries of GPU capex supercycle as hyperscalers slash headcount to fund AI infrastructure buildout.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Amazon laid off 30,000 employees today, an even larger cut than during the industry contraction in 2022. The reason is simple:\n\nThey don’t have enough capex left to buy GPUs.\nAs a result, AWS growth has slowed, the market has punished them harshly, and now they must cut salaries to save money for GPU purchases—so the financials look better and they can tell a story of “AWS growth bottoming out.”\n\nEvery internet company software engineer (SDE) should buy Nvidia/AMD stock as a hedge—compensating for the risk of being squeezed out of the value chain by GPUs.\n\nEntering 2024–2025, the main factor behind weak employment for American SDEs is no longer the massive over-expansion of 2021, nor the competition from lower-wage overseas engineering centers, and not yet the reduced demand caused by AI efficiency gains.\n\nA new boss has arrived: GPU capex.\nGPU capex is creating a strange “prosperous depression” inside internet companies:\n\nThe company’s revenue growth looks strong, stock prices keep rising—but wage expenses have become immovable constraints for management. Everyone worries about their jobs. Continuous layoffs increase the workload for those who remain. Morale collapses. It feels like the Great Depression all over again.\n\nThis isn’t a traditional recession. It’s capital’s radical redistribution between manpower and compute power.\n\nAmazon’s 30,000-person layoff has been rumored for two months. The mid-year performance review, usually in July, was delayed to mid- or late August. The return-to-office (RTO) policy also served as a major excuse for the cuts. The AGI group will remain untouched; PXT, Devices & Services, and Operations will be the hardest hit. By convention, AWS will likely announce its cuts later—probably after AWS re:Invent—to squeeze every bit of output first.\n\nMeanwhile, AWS’s Q2 backlog reached $195 billion, up 25% YoY.\n\nThis shows customers want to buy but AWS can’t deliver—demand remains scorching hot, and they simply can’t buy GPUs fast enough.\n\nIn an era where AI server supply can’t keep up with explosive demand, shifting opex (wages) to capex directly boosts company performance. Capital will ruthlessly punish any CSP or hyperscaler that fails to fully embrace this path.\nMeta has quietly entered a “5% layoff every six months” rhythm. Recently, it cut 600 people in its AI org and also removed many directors across departments—same logic again: AI datacenter capacity is insufficient. In the past year, Meta revised up its 18-month capacity plan three times. Each time they thought they had overestimated demand—only to painfully realize a few months later that they had underestimated it.\n\nDoes this mean internet companies no longer need people?\n\nOf course not. But after the hiring budget is slashed, they’re forced to squeeze productivity internally to compensate.\n\nBig tech companies are now pulling every possible lever: building internal tools with agent functions, encouraging “one-click deployment vibes,” setting KPIs for AI usage rates, requiring departments to report AI adoption progress and use cases, and mandating periodic peer learning sessions.\n\nAll these frantic moves aim for just one modest goal: ~20% efficiency improvement.\nWhat happens next?\n\nTo maintain growth and competitiveness—once efficiency gains plateau and layoffs hit the bone, when opex yields no more savings—the next step will be to sacrifice cash flow. Some, like Oracle, may even take on debt to sustain growth.\n\nNvidia and AMD, armed with huge cash reserves, will continue to push partners to invest in AI capex—just like OpenAI has done.\n\nThe biggest beneficiaries of all this will be the semiconductor supply chain.\n\nA new normal may emerge: semiconductor companies’ profit margins surpass those of internet firms.\n\nBut they also bear the greatest risk:\n\nOnce VCs or hyperscalers notice token demand slowing—or even just growth decelerating—they’ll ruthlessly cut orders.\n\nThe transmission of that shock will be far faster than semiconductor production cycles.\nWhen might that point arrive?\n\nOne reference indicator: when enterprise adoption reaches ~50%.\nDuring the March 2000 internet bubble burst, U.S. internet penetration was around 52% (some data say 43%).\nCurrently, major internet companies’ GenAI daily-user penetration is rising from 50% toward 90%, while overall enterprise AI adoption is still under 10%.\n\nThat means growth is safe for now. Historically, the fastest phase of any technological revolution is when corporate adoption moves from 10% to 50%.\nThe Cisco bubble won’t repeat itself so simply.\n\nThis time, information flow is vastly richer. There will always be enough skeptics warning about bubbles—ensuring that when it does burst, it won’t be nearly as catastrophic.", "reply_text": null, "parent_text": null, "quoted_text": "今天Amazon裁员3万人，比2022年行业收缩期裁的还厉害，原因很简单\n\n买GPU的钱capex不够，导致AWS增速下降，被市场严厉的惩罚了，只能砍工资来省钱买GPU，让财报好看一点，营造一个AWS增速触底的故事\n\n每一个互联网公司的SDE打工人，都应该买入Nvidia/amd作为风险对冲，弥补自己被GPU挤出价值链的风险 https://t.co/nAzA2nsAfi", "parent_author": null, "ret_m1d": 0.006433873102897447, "ret_signed_m1d": 0.006433873102897447, "alpha_spy_m1d": 0.00908283641569363, "alpha_c_m1d": 0.015163980981509506, "bench_spy_m1d": -0.002648963312796182, "bench_c_m1d": -0.008730107878612059, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02449508378524401, "ret_signed_p1d": 0.02449508378524401, "alpha_spy_p1d": 0.024014786086438145, "alpha_c_p1d": 0.009348154397005759, "bench_spy_p1d": 0.0004802976988058649, "bench_c_p1d": 0.01514692938823825, "ret_p1w": -0.030851542237055107, "ret_signed_p1w": -0.030851542237055107, "alpha_spy_p1w": -0.01364784330151847, "alpha_c_p1w": -0.0023478479257212914, "bench_spy_p1w": -0.017203698935536638, "bench_c_p1w": -0.028503694311333816, "ret_p1m": -0.16964459755736372, "ret_signed_p1m": -0.16964459755736372, "alpha_spy_p1m": -0.15890320767088228, "alpha_c_p1m": -0.12723319527029664, "bench_spy_p1m": -0.010741389886481434, "bench_c_p1m": -0.04241140228706708, "ret_p3m": 0.006472550858291237, "ret_signed_p3m": 0.006472550858291237, "alpha_spy_p3m": 0.00035000597377443654, "alpha_c_p3m": -0.09879837654705792, "bench_spy_p3m": 0.006122544884516801, "bench_c_p3m": 0.10527092740534916, "ret_p6m": 0.17615588531134763, "ret_signed_p6m": 0.17615588531134763, "alpha_spy_p6m": 0.1351119632660942, "alpha_c_p6m": -0.14111354481035732, "bench_spy_p6m": 0.04104392204525342, "bench_c_p6m": 0.31726943012170494, "price_path": [-0.3167, -0.3345, -0.3148, -0.3244, -0.3678, -0.3318, -0.3304, -0.3323, -0.3219, -0.2852, -0.2987, -0.312, -0.3173, -0.3545, -0.3597, -0.3655, -0.3498, -0.3668, -0.3542, -0.3522, -0.3466, -0.3697, -0.3697, -0.3709, -0.3716, -0.3729, -0.4142, -0.4132, -0.3961, -0.3817, -0.3967, -0.3854, -0.3754, -0.3781, -0.3831, -0.3879, -0.39, -0.3807, -0.3764, -0.3765, -0.3749, -0.382, -0.3746, -0.3729, -0.3643, -0.3422, -0.3618, -0.2105, -0.1802, -0.087, -0.0974, -0.1671, -0.1612, -0.1547, -0.0752, -0.0909, -0.0966, -0.0676, -0.0774, -0.1077, -0.0892, -0.0197, 0.0064, 0.0, 0.0245, -0.0123, -0.0073, 0.0064, -0.0309, -0.0065, -0.0787, -0.0948, -0.0544, -0.0794, 0.0034, -0.039, -0.0434, -0.0678, -0.1074, -0.1336, -0.2015, -0.2102, -0.1665, -0.2011, -0.1696, -0.1696, -0.1569, -0.1483, -0.1658, -0.1566, -0.1629, -0.1552, -0.143, -0.141, -0.1418, -0.1418, -0.1831, -0.1955, -0.1893, -0.2322, -0.2207, -0.1728, -0.1669, -0.1671, -0.1665, -0.1665, -0.1667, -0.1643, -0.1654, -0.17, -0.17, -0.1339, -0.1431, -0.1692, -0.186, -0.2067, -0.2125, -0.195, -0.1436, -0.1334, -0.1166, -0.1015, -0.1015, -0.1011, -0.0318, -0.0166, 0.0065, -0.026, -0.0232, -0.0204, -0.0226, -0.0825, -0.0455, -0.0616, -0.2241, -0.2539, -0.1921, -0.1628, -0.1722, -0.1722, -0.2018, -0.1965, -0.1965, -0.2129, -0.2244, -0.2118, -0.2243, -0.238, -0.1712, -0.1827, -0.2106, -0.224, -0.2302, -0.2599, -0.2168, -0.227, -0.2542, -0.2144, -0.2123, -0.2061, -0.2336, -0.2505, -0.2381, -0.2391, -0.2269, -0.2044, -0.2197, -0.2144, -0.204, -0.1463, -0.2102, -0.2171, -0.2402, -0.2115, -0.1853, -0.157, -0.157, -0.1466, -0.1414, -0.1015, -0.0828, -0.0503, -0.0433, -0.0114, 0.0004, 0.0785, 0.079, 0.0657, 0.1026, 0.1762]}
{"unit_id": "orig:1976555832558092419", "ticker_idx": 0, "ts": "2025-10-10T07:50:07+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "UBS maintains SK Hynix as its \"Key Call Buy\", raising the target price from ₩434,000 to ₩516,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises HBM forecasts; Key Call Buy SK Hynix (PT ₩516K), Buy Micron, Neutral Samsung; NVDA GPU volume raised to 7.4M units for 2026.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Raising NVIDIA Purchase Forecasts and Including OpenAI ASIC Model for the First Time — Significant Expansion in HBM Market Size\n\nUBS (N. Gaudois, 251008)\n\nBased on the latest industry research and bottom-up model updates, UBS has raised its forecast for global High Bandwidth Memory (HBM) end-market consumption. HBM shipments are projected to reach 17.1 billion Gb in 2025, up 99% YoY, and further rise to 27.2 billion Gb in 2026, up 59% YoY. Corresponding revenue is expected to reach $33.2 billion in 2025 (+103%) and $54.5 billion in 2026 (+64%), at which point HBM will account for 9% of total DRAM shipments and 29% of DRAM revenue.\n\nAt the company level, UBS raised NVIDIA’s 2026 AI GPU purchase volume from 7.0 million to 7.4 million units, reflecting stronger expectations for the Rubin project, and for the first time included OpenAI’s ASIC in its model, estimating initial shipments of 700,000 units in 2026. UBS expects HBM bit demand to maintain around 35% YoY growth in 2027.\n\nUBS noted that OpenAI’s ASIC is expected to enter mass production in the second half of 2026, with a significant ramp-up in 2027, adopting an HBM3E 12-Hi packaging configuration. Preliminary estimates suggest that OpenAI could account for about 10% of global HBM bit demand in 2027, comparable to Google. Samsung is expected to be the primary supplier, with SK Hynix as the secondary. As a result, UBS has raised its 2027 DRAM CAPEX forecasts for both Samsung and SK Hynix.\n\nIn terms of market share, UBS projects that in the 2026 HBM bit market, SK Hynix will maintain around 51% share, followed by Micron and Samsung at 25% and 24%, respectively.\n\nBy customer structure:\n\n    • SK Hynix’s share with NVIDIA, Google, and Amazon is expected to exceed 60%, 67%, and 84%, respectively.\n    • Micron’s share with NVIDIA, Meta, and AMD is projected at 24%, 63%, and 34%, respectively.\n    • Samsung’s share with NVIDIA, AMD, Google, Amazon, and Meta is estimated at 15%, 66%, 33%, 16%, and 37%, respectively.\n\nIn terms of investment recommendations, UBS maintains SK Hynix as its “Key Call Buy”, raising the target price from ₩434,000 to ₩516,000; Micron is also rated Buy; Samsung remains Neutral, with a target price raised from ₩85,000 to ₩93,000.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07593449921244366, "ret_signed_m1d": -0.07593449921244366, "alpha_spy_m1d": -0.10371312297382962, "alpha_c_m1d": -0.13700902240058876, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03037381432765296, "ret_signed_p1d": -0.03037381432765296, "alpha_spy_p1d": -0.04571788902086171, "alpha_c_p1d": -0.07466592270284556, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.08761686502367061, "ret_signed_p1w": 0.08761686502367061, "alpha_spy_p1w": 0.07020539438832563, "alpha_c_p1w": 0.033857772721988644, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.41588785936237294, "ret_signed_p1m": 0.41588785936237294, "alpha_spy_p1m": 0.3723668906246631, "alpha_c_p1m": 0.3130723038954748, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.7348865990672908, "ret_signed_p3m": 0.7348865990672908, "alpha_spy_p3m": 0.6757801683134932, "alpha_c_p3m": 0.5477653744002142, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": 1.0753986505017386, "ret_signed_p6m": 1.0753986505017386, "alpha_spy_p6m": 1.0606028671848362, "alpha_c_p6m": 0.8544847005793097, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.3036, -0.3094, -0.3712, -0.3724, -0.3642, -0.3736, -0.3724, -0.3712, -0.3794, -0.3887, -0.3876, -0.3852, -0.3619, -0.3981, -0.3981, -0.3852, -0.3969, -0.3887, -0.4016, -0.3771, -0.3724, -0.3514, -0.3549, -0.3549, -0.3759, -0.3864, -0.4039, -0.4284, -0.4144, -0.3946, -0.3899, -0.3934, -0.3727, -0.3715, -0.4019, -0.3914, -0.3867, -0.3797, -0.361, -0.3528, -0.3271, -0.2897, -0.2827, -0.2325, -0.2266, -0.1869, -0.2208, -0.1694, -0.1694, -0.1799, -0.1565, -0.1647, -0.1671, -0.2138, -0.1846, -0.1881, -0.1589, -0.0759, -0.0759, -0.0759, -0.0759, -0.0759, -0.0759, 0.0, -0.0304, -0.0386, -0.0129, 0.0572, 0.0876, 0.1343, 0.1192, 0.125, 0.118, 0.1916, 0.25, 0.2173, 0.3037, 0.3271, 0.3061, 0.4486, 0.3692, 0.3528, 0.3855, 0.3551, 0.4159, 0.4463, 0.4416, 0.4299, 0.3084, 0.4159, 0.3318, 0.3131, 0.3341, 0.2173, 0.215, 0.2126, 0.2243, 0.2719, 0.2392, 0.2579, 0.3047, 0.2906, 0.2673, 0.2719, 0.3491, 0.3234, 0.3725, 0.321, 0.3351, 0.2953, 0.2392, 0.2883, 0.2906, 0.279, 0.3561, 0.3655, 0.3748, 0.3748, 0.4005, 0.4964, 0.5221, 0.5221, 0.5221, 0.5829, 0.6273, 0.6975, 0.7349, 0.7676, 0.7396, 0.7513, 0.7255, 0.7349, 0.7513, 0.7676, 0.7863, 0.7372, 0.7302, 0.7653, 0.7933, 0.7209, 0.8705, 0.9664, 1.0131, 1.1254, 0.9406, 1.1207, 1.1043, 0.9687, 0.9617, 1.0739, 1.0482, 1.0108, 1.0763, 1.0575, 1.0575, 1.0575, 1.0575, 1.0903, 1.2189, 1.2236, 1.3498, 1.3802, 1.5743, 1.4853, 1.4853, 1.1995, 0.9887, 1.2042, 1.1644, 0.9583, 1.1972, 1.237, 1.1785, 1.1316, 1.2815, 1.2722, 1.4736, 1.3729, 1.3588, 1.1855, 1.3096, 1.3307, 1.1855, 1.1855, 1.0449, 0.8903, 1.1012, 0.9442, 1.052, 1.0754]}
{"unit_id": "orig:1976555832558092419", "ticker_idx": 1, "ts": "2025-10-10T07:50:07+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is also rated Buy", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises HBM forecasts; Key Call Buy SK Hynix (PT ₩516K), Buy Micron, Neutral Samsung; NVDA GPU volume raised to 7.4M units for 2026.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Raising NVIDIA Purchase Forecasts and Including OpenAI ASIC Model for the First Time — Significant Expansion in HBM Market Size\n\nUBS (N. Gaudois, 251008)\n\nBased on the latest industry research and bottom-up model updates, UBS has raised its forecast for global High Bandwidth Memory (HBM) end-market consumption. HBM shipments are projected to reach 17.1 billion Gb in 2025, up 99% YoY, and further rise to 27.2 billion Gb in 2026, up 59% YoY. Corresponding revenue is expected to reach $33.2 billion in 2025 (+103%) and $54.5 billion in 2026 (+64%), at which point HBM will account for 9% of total DRAM shipments and 29% of DRAM revenue.\n\nAt the company level, UBS raised NVIDIA’s 2026 AI GPU purchase volume from 7.0 million to 7.4 million units, reflecting stronger expectations for the Rubin project, and for the first time included OpenAI’s ASIC in its model, estimating initial shipments of 700,000 units in 2026. UBS expects HBM bit demand to maintain around 35% YoY growth in 2027.\n\nUBS noted that OpenAI’s ASIC is expected to enter mass production in the second half of 2026, with a significant ramp-up in 2027, adopting an HBM3E 12-Hi packaging configuration. Preliminary estimates suggest that OpenAI could account for about 10% of global HBM bit demand in 2027, comparable to Google. Samsung is expected to be the primary supplier, with SK Hynix as the secondary. As a result, UBS has raised its 2027 DRAM CAPEX forecasts for both Samsung and SK Hynix.\n\nIn terms of market share, UBS projects that in the 2026 HBM bit market, SK Hynix will maintain around 51% share, followed by Micron and Samsung at 25% and 24%, respectively.\n\nBy customer structure:\n\n    • SK Hynix’s share with NVIDIA, Google, and Amazon is expected to exceed 60%, 67%, and 84%, respectively.\n    • Micron’s share with NVIDIA, Meta, and AMD is projected at 24%, 63%, and 34%, respectively.\n    • Samsung’s share with NVIDIA, AMD, Google, Amazon, and Meta is estimated at 15%, 66%, 33%, 16%, and 37%, respectively.\n\nIn terms of investment recommendations, UBS maintains SK Hynix as its “Key Call Buy”, raising the target price from ₩434,000 to ₩516,000; Micron is also rated Buy; Samsung remains Neutral, with a target price raised from ₩85,000 to ₩93,000.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.059085891006847246, "ret_signed_m1d": 0.059085891006847246, "alpha_spy_m1d": 0.03130726724546129, "alpha_c_m1d": -0.001988632181297856, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06150879079035709, "ret_signed_p1d": 0.06150879079035709, "alpha_spy_p1d": 0.04616471609714834, "alpha_c_p1d": 0.017216682415164497, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.11442735404408655, "ret_signed_p1w": 0.11442735404408655, "alpha_spy_p1w": 0.09701588340874157, "alpha_c_p1w": 0.06066826174240458, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.394823763725729, "ret_signed_p1m": 0.394823763725729, "alpha_spy_p1m": 0.3513027949880192, "alpha_c_p1m": 0.2920082082588309, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.8705240433195509, "ret_signed_p3m": 0.8705240433195509, "alpha_spy_p3m": 0.8114176125657533, "alpha_c_p3m": 0.6834028186524743, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": 1.0818908507698248, "ret_signed_p6m": 1.0818908507698248, "alpha_spy_p6m": 1.0670950674529225, "alpha_c_p6m": 0.860976900847396, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.339, -0.3593, -0.3767, -0.3705, -0.3769, -0.3989, -0.3956, -0.3851, -0.3877, -0.3878, -0.3839, -0.3686, -0.3994, -0.4228, -0.4069, -0.3998, -0.4014, -0.3844, -0.3457, -0.3191, -0.297, -0.3161, -0.3105, -0.3348, -0.3201, -0.3283, -0.355, -0.3628, -0.3524, -0.3593, -0.3589, -0.352, -0.3286, -0.3451, -0.3451, -0.348, -0.3467, -0.3165, -0.277, -0.2766, -0.2558, -0.2296, -0.1714, -0.1347, -0.1318, -0.126, -0.1195, -0.0706, -0.1045, -0.0941, -0.0842, -0.1101, -0.1369, -0.1345, -0.098, -0.0792, 0.0024, 0.0112, 0.0343, 0.0515, 0.0225, 0.0823, 0.0591, 0.0, 0.0615, 0.0301, 0.0569, 0.1153, 0.1144, 0.1386, 0.1139, 0.0929, 0.1383, 0.2061, 0.212, 0.222, 0.248, 0.2335, 0.2322, 0.2924, 0.2006, 0.3078, 0.3124, 0.3101, 0.3948, 0.3277, 0.3486, 0.3048, 0.3592, 0.3323, 0.2583, 0.2441, 0.1089, 0.1419, 0.2331, 0.2364, 0.268, 0.268, 0.3022, 0.3241, 0.3188, 0.2894, 0.2481, 0.3063, 0.3597, 0.39, 0.4521, 0.4232, 0.3279, 0.3078, 0.2803, 0.2419, 0.3687, 0.4643, 0.5231, 0.5213, 0.5786, 0.5786, 0.5682, 0.6216, 0.612, 0.5723, 0.5723, 0.7376, 0.7196, 0.8919, 0.8705, 0.8015, 0.901, 0.9053, 0.8627, 0.8364, 0.8544, 0.9983, 0.9983, 1.0107, 1.1435, 1.1902, 1.2016, 1.1434, 1.2599, 1.3979, 1.4007, 1.2855, 1.4118, 1.3106, 1.0901, 1.1093, 1.1743, 1.1126, 1.0562, 1.2605, 1.2805, 1.2678, 1.2678, 1.2023, 1.3189, 1.2991, 1.3587, 1.3191, 1.3027, 1.3633, 1.2893, 1.2717, 1.2733, 1.0916, 1.2078, 1.1873, 1.0399, 1.1447, 1.2207, 1.3065, 1.233, 1.3475, 1.4338, 1.5434, 1.5436, 1.4474, 1.3297, 1.2275, 1.1789, 1.1049, 0.9582, 0.9679, 0.7735, 0.8619, 1.0273, 1.0184, 1.0184, 1.0819]}
{"unit_id": "orig:1976555832558092419", "ticker_idx": 2, "ts": "2025-10-10T07:50:07+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung remains Neutral, with a target price raised from ₩85,000 to ₩93,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises HBM forecasts; Key Call Buy SK Hynix (PT ₩516K), Buy Micron, Neutral Samsung; NVDA GPU volume raised to 7.4M units for 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Raising NVIDIA Purchase Forecasts and Including OpenAI ASIC Model for the First Time — Significant Expansion in HBM Market Size\n\nUBS (N. Gaudois, 251008)\n\nBased on the latest industry research and bottom-up model updates, UBS has raised its forecast for global High Bandwidth Memory (HBM) end-market consumption. HBM shipments are projected to reach 17.1 billion Gb in 2025, up 99% YoY, and further rise to 27.2 billion Gb in 2026, up 59% YoY. Corresponding revenue is expected to reach $33.2 billion in 2025 (+103%) and $54.5 billion in 2026 (+64%), at which point HBM will account for 9% of total DRAM shipments and 29% of DRAM revenue.\n\nAt the company level, UBS raised NVIDIA’s 2026 AI GPU purchase volume from 7.0 million to 7.4 million units, reflecting stronger expectations for the Rubin project, and for the first time included OpenAI’s ASIC in its model, estimating initial shipments of 700,000 units in 2026. UBS expects HBM bit demand to maintain around 35% YoY growth in 2027.\n\nUBS noted that OpenAI’s ASIC is expected to enter mass production in the second half of 2026, with a significant ramp-up in 2027, adopting an HBM3E 12-Hi packaging configuration. Preliminary estimates suggest that OpenAI could account for about 10% of global HBM bit demand in 2027, comparable to Google. Samsung is expected to be the primary supplier, with SK Hynix as the secondary. As a result, UBS has raised its 2027 DRAM CAPEX forecasts for both Samsung and SK Hynix.\n\nIn terms of market share, UBS projects that in the 2026 HBM bit market, SK Hynix will maintain around 51% share, followed by Micron and Samsung at 25% and 24%, respectively.\n\nBy customer structure:\n\n    • SK Hynix’s share with NVIDIA, Google, and Amazon is expected to exceed 60%, 67%, and 84%, respectively.\n    • Micron’s share with NVIDIA, Meta, and AMD is projected at 24%, 63%, and 34%, respectively.\n    • Samsung’s share with NVIDIA, AMD, Google, Amazon, and Meta is estimated at 15%, 66%, 33%, 16%, and 37%, respectively.\n\nIn terms of investment recommendations, UBS maintains SK Hynix as its “Key Call Buy”, raising the target price from ₩434,000 to ₩516,000; Micron is also rated Buy; Samsung remains Neutral, with a target price raised from ₩85,000 to ₩93,000.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04925846098153419, "ret_signed_m1d": 0.04925846098153419, "alpha_spy_m1d": 0.07703708474292015, "alpha_c_m1d": 0.091680830693042, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.04242236971150781, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01165250600254597, "ret_signed_p1d": 0.01165250600254597, "alpha_spy_p1d": 0.026996580695754724, "alpha_c_p1d": 0.03600670541295525, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.024354199410409283, "ret_p1w": 0.037076367585484826, "ret_signed_p1w": -0.037076367585484826, "alpha_spy_p1w": -0.019664896950139843, "alpha_c_p1w": -0.013296853148118792, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.023779514437366034, "ret_p1m": 0.06567800353136044, "ret_signed_p1m": -0.06567800353136044, "alpha_spy_p1m": -0.02215703479365061, "alpha_c_p1m": -0.004109874232078203, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.06156812929928224, "ret_p3m": 0.5009050069383865, "ret_signed_p3m": -0.5009050069383865, "alpha_spy_p3m": -0.44179857618458884, "alpha_c_p3m": -0.4466125564829806, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.05429245045540587, "ret_p6m": 1.0597491247432895, "ret_signed_p6m": -1.0597491247432895, "alpha_spy_p6m": -1.0449533414263872, "alpha_c_p6m": -1.0743849044010685, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": -0.014635779657778847, "price_path": [-0.3282, -0.3177, -0.2966, -0.2924, -0.285, -0.304, -0.2997, -0.304, -0.305, -0.2575, -0.2554, -0.2343, -0.247, -0.2734, -0.2649, -0.2628, -0.2744, -0.2565, -0.2428, -0.2512, -0.2502, -0.2417, -0.2449, -0.2449, -0.2618, -0.2618, -0.2565, -0.2554, -0.247, -0.2459, -0.2586, -0.2554, -0.266, -0.2649, -0.2871, -0.2713, -0.2639, -0.2607, -0.267, -0.2607, -0.2459, -0.2343, -0.2259, -0.2048, -0.1932, -0.1626, -0.1753, -0.1531, -0.1531, -0.1194, -0.1067, -0.0994, -0.092, -0.1215, -0.1081, -0.1112, -0.089, -0.0493, -0.0493, -0.0493, -0.0493, -0.0493, -0.0493, 0.0, -0.0117, -0.0297, 0.0064, 0.035, 0.0371, 0.0392, 0.0328, 0.0445, 0.0222, 0.0466, 0.0805, 0.054, 0.0646, 0.1028, 0.1388, 0.1769, 0.1112, 0.0657, 0.0508, 0.0371, 0.0657, 0.0964, 0.0922, 0.089, 0.0297, 0.0657, 0.036, 0.0222, 0.0657, 0.0042, 0.0244, 0.0519, 0.089, 0.0964, 0.0646, 0.0678, 0.0953, 0.107, 0.1133, 0.1483, 0.16, 0.1483, 0.1441, 0.1367, 0.1536, 0.1102, 0.089, 0.143, 0.1398, 0.1261, 0.1706, 0.1811, 0.1769, 0.1769, 0.2394, 0.272, 0.2763, 0.2763, 0.2763, 0.3678, 0.47, 0.4786, 0.5009, 0.4775, 0.4796, 0.4775, 0.4647, 0.4935, 0.5318, 0.585, 0.5893, 0.5456, 0.5914, 0.6212, 0.6191, 0.6191, 0.6978, 0.7287, 0.7106, 0.7085, 0.601, 0.783, 0.8, 0.6957, 0.6883, 0.7713, 0.7649, 0.7862, 0.9011, 0.9288, 0.9288, 0.9288, 0.9288, 1.0225, 1.0236, 1.0544, 1.1289, 1.1662, 1.3205, 1.3046, 1.3046, 1.0768, 0.833, 1.0395, 1.0033, 0.8469, 1.0001, 1.0225, 1.0001, 0.9533, 1.0087, 1.064, 1.2194, 1.1343, 1.1226, 0.9831, 1.0193, 1.0119, 0.9171, 0.9171, 0.8805, 0.7835, 1.0229, 0.9029, 0.9861, 1.0597]}
{"unit_id": "orig:1976555832558092419", "ticker_idx": 3, "ts": "2025-10-10T07:50:07+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS raised NVIDIA's 2026 AI GPU purchase volume from 7.0 million to 7.4 million units", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS raises HBM forecasts; Key Call Buy SK Hynix (PT ₩516K), Buy Micron, Neutral Samsung; NVDA GPU volume raised to 7.4M units for 2026.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Raising NVIDIA Purchase Forecasts and Including OpenAI ASIC Model for the First Time — Significant Expansion in HBM Market Size\n\nUBS (N. Gaudois, 251008)\n\nBased on the latest industry research and bottom-up model updates, UBS has raised its forecast for global High Bandwidth Memory (HBM) end-market consumption. HBM shipments are projected to reach 17.1 billion Gb in 2025, up 99% YoY, and further rise to 27.2 billion Gb in 2026, up 59% YoY. Corresponding revenue is expected to reach $33.2 billion in 2025 (+103%) and $54.5 billion in 2026 (+64%), at which point HBM will account for 9% of total DRAM shipments and 29% of DRAM revenue.\n\nAt the company level, UBS raised NVIDIA’s 2026 AI GPU purchase volume from 7.0 million to 7.4 million units, reflecting stronger expectations for the Rubin project, and for the first time included OpenAI’s ASIC in its model, estimating initial shipments of 700,000 units in 2026. UBS expects HBM bit demand to maintain around 35% YoY growth in 2027.\n\nUBS noted that OpenAI’s ASIC is expected to enter mass production in the second half of 2026, with a significant ramp-up in 2027, adopting an HBM3E 12-Hi packaging configuration. Preliminary estimates suggest that OpenAI could account for about 10% of global HBM bit demand in 2027, comparable to Google. Samsung is expected to be the primary supplier, with SK Hynix as the secondary. As a result, UBS has raised its 2027 DRAM CAPEX forecasts for both Samsung and SK Hynix.\n\nIn terms of market share, UBS projects that in the 2026 HBM bit market, SK Hynix will maintain around 51% share, followed by Micron and Samsung at 25% and 24%, respectively.\n\nBy customer structure:\n\n    • SK Hynix’s share with NVIDIA, Google, and Amazon is expected to exceed 60%, 67%, and 84%, respectively.\n    • Micron’s share with NVIDIA, Meta, and AMD is projected at 24%, 63%, and 34%, respectively.\n    • Samsung’s share with NVIDIA, AMD, Google, Amazon, and Meta is estimated at 15%, 66%, 33%, 16%, and 37%, respectively.\n\nIn terms of investment recommendations, UBS maintains SK Hynix as its “Key Call Buy”, raising the target price from ₩434,000 to ₩516,000; Micron is also rated Buy; Samsung remains Neutral, with a target price raised from ₩85,000 to ₩93,000.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05137584134182771, "ret_signed_m1d": 0.05137584134182771, "alpha_spy_m1d": 0.023597217580441754, "alpha_c_m1d": -0.009698681846317392, "bench_spy_m1d": 0.027778623761385957, "bench_c_m1d": 0.0610745231881451, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02817212030804872, "ret_signed_p1d": 0.02817212030804872, "alpha_spy_p1d": 0.012828045614839967, "alpha_c_p1d": -0.016119988067143876, "bench_spy_p1d": 0.015344074693208754, "bench_c_p1d": 0.044292108375192596, "ret_p1w": 0.0003276050769120964, "ret_signed_p1w": 0.0003276050769120964, "alpha_spy_p1w": -0.017083865558432887, "alpha_c_p1w": -0.05343148722476987, "bench_spy_p1w": 0.017411470635344983, "bench_c_p1w": 0.053759092301681966, "ret_p1m": 0.0867547734359273, "ret_signed_p1m": 0.0867547734359273, "alpha_spy_p1m": 0.04323380469821747, "alpha_c_p1m": -0.016060782030970833, "bench_spy_p1m": 0.04352096873770983, "bench_c_p1m": 0.10281555546689813, "ret_p3m": 0.03254271530285635, "ret_signed_p3m": 0.03254271530285635, "alpha_spy_p3m": -0.026563715450941272, "alpha_c_p3m": -0.15457850936422024, "bench_spy_p3m": 0.05910643075379762, "bench_c_p3m": 0.1871212246670766, "ret_p6m": -0.030031104152419386, "ret_signed_p6m": -0.030031104152419386, "alpha_spy_p6m": -0.04482688746932173, "alpha_c_p6m": -0.25094505407484824, "bench_spy_p6m": 0.014795783316902344, "bench_c_p6m": 0.22091394992242885, "price_path": [-0.0681, -0.0644, -0.0555, -0.0587, -0.0644, -0.0881, -0.0676, -0.0515, -0.0528, -0.0351, -0.0418, -0.0213, -0.0289, -0.0516, -0.0173, -0.0268, -0.0205, -0.0131, -0.0026, -0.0061, -0.0001, -0.0086, -0.0063, -0.0149, -0.0063, -0.0411, -0.0424, -0.0447, -0.0283, -0.0183, -0.0076, -0.0086, -0.0164, -0.0491, -0.0491, -0.0676, -0.0685, -0.0628, -0.0882, -0.0811, -0.0678, -0.0319, -0.0327, -0.0292, -0.0295, -0.0452, -0.0703, -0.0378, -0.0354, 0.0025, -0.0258, -0.0338, -0.0299, -0.0271, -0.0072, 0.0187, 0.0223, 0.0313, 0.0244, 0.013, 0.0103, 0.0325, 0.0514, 0.0, 0.0282, -0.0171, -0.0182, -0.0074, 0.0003, -0.0028, -0.0109, -0.0157, -0.0055, 0.0169, 0.0455, 0.0976, 0.1304, 0.1077, 0.1055, 0.1295, 0.0848, 0.0658, 0.0269, 0.0272, 0.0868, 0.0546, 0.0581, 0.0202, 0.0383, 0.0188, -0.0098, 0.0183, -0.0138, -0.0234, -0.0033, -0.0292, -0.0158, -0.0158, -0.0336, -0.0177, -0.0093, -0.0195, 0.0013, -0.004, 0.0131, 0.0099, 0.0034, -0.0121, -0.0444, -0.0375, -0.0296, -0.0667, -0.0492, -0.0118, 0.0029, 0.0331, 0.0298, 0.0298, 0.0403, 0.0277, 0.024, 0.0183, 0.0183, 0.0311, 0.0271, 0.0223, 0.0325, 0.0103, 0.0093, 0.0098, 0.0145, -0.0001, 0.0213, 0.0168, 0.0168, -0.0277, 0.0009, 0.0092, 0.0247, 0.0181, 0.0293, 0.0457, 0.0511, 0.0436, 0.0134, -0.0153, -0.0489, -0.0615, 0.0123, 0.0376, 0.0294, 0.0377, 0.0207, -0.0019, -0.0019, 0.0099, 0.0264, 0.0259, 0.0364, 0.0459, 0.053, 0.0678, 0.0095, -0.0325, -0.0037, -0.0169, -0.0006, 0.001, -0.0291, -0.0027, 0.0088, 0.0158, 0.0, -0.0158, 0.0004, -0.0066, -0.015, -0.025, -0.057, -0.041, -0.0434, -0.0244, -0.065, -0.0853, -0.0981, -0.0477, -0.0404, -0.0314, -0.0314, -0.03]}
{"unit_id": "quote:1994327901634777224", "ticker_idx": 0, "ts": "2025-11-28T08:49:59+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom has adopted Intel's EMIB and entered the back-end process of AWS Trainium 3. Also, Broadcom is said to be very active on Trainium 4.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Broadcom wins Trainium 3 back-end with Intel EMIB, taking share from Marvell/Alchip; active on Trainium 4.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Rumor: Broadcom has adopted Intel’s EMIB and entered the back-end process of AWS Trainium 3.\n\nAlso, Broadcom is said to be very active on Trainium 4.\n\n$AVGO $INTC", "reply_text": null, "parent_text": null, "quoted_text": "Just in: I'm not sure if I can share this, but it seems  that $AVGO did break into Trainium 3 back-end by proposing an Intel-EMIB solution and will gain share from Alchip and $MRVL (should be proportionally decreased?).\n\nOne more thing: $AVGO is also very \"active\" in Trainium 4.", "parent_author": null, "ret_m1d": -0.01337600157671559, "ret_signed_m1d": -0.01337600157671559, "alpha_spy_m1d": -0.007947227691884495, "alpha_c_m1d": -0.0004033008537911442, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04189002393587471, "ret_signed_p1d": -0.04189002393587471, "alpha_spy_p1d": -0.03732460833332352, "alpha_c_p1d": -0.04382029603880688, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": -0.031566399564335046, "ret_signed_p1w": -0.031566399564335046, "alpha_spy_p1w": -0.03493198787447105, "alpha_c_p1w": -0.06602766134166271, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": -0.1312821786352757, "ret_signed_p1m": -0.1312821786352757, "alpha_spy_p1m": -0.1407826227546084, "alpha_c_p1m": -0.16836420646637051, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": -0.1737496320730837, "ret_signed_p3m": -0.1737496320730837, "alpha_spy_p3m": -0.1910285642537396, "alpha_c_p3m": -0.3872333122233552, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 0.03187123428559513, "ret_signed_p6m": 0.03187123428559513, "alpha_spy_p6m": -0.06543145372095527, "alpha_c_p6m": -0.6091906143681416, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.2611, -0.2509, -0.2417, -0.1703, -0.1437, -0.1659, -0.0844, -0.1091, -0.1085, -0.098, -0.1081, -0.1424, -0.1444, -0.1454, -0.1592, -0.1589, -0.158, -0.1659, -0.1698, -0.1863, -0.1813, -0.1726, -0.1608, -0.1603, -0.1674, -0.1652, -0.1426, -0.1438, -0.1944, -0.1148, -0.146, -0.1281, -0.1211, -0.1331, -0.1333, -0.1496, -0.1555, -0.1456, -0.1212, -0.1015, -0.0744, -0.0421, -0.0657, -0.0827, -0.1003, -0.1266, -0.1091, -0.1176, -0.1328, -0.1106, -0.1266, -0.1185, -0.1563, -0.1501, -0.1497, -0.155, -0.1205, -0.1393, -0.1557, -0.062, -0.0445, -0.0134, -0.0134, 0.0, -0.0419, -0.0531, -0.0555, -0.0544, -0.0316, -0.0046, 0.0083, 0.0248, 0.0085, -0.1068, -0.1567, -0.153, -0.1909, -0.1814, -0.1554, -0.151, -0.1315, -0.1292, -0.1292, -0.1245, -0.1313, -0.1301, -0.1395, -0.1395, -0.1357, -0.1461, -0.1453, -0.1459, -0.1733, -0.1423, -0.1243, -0.1183, -0.1549, -0.1471, -0.1255, -0.1255, -0.173, -0.1825, -0.1907, -0.2042, -0.1923, -0.1726, -0.1714, -0.1777, -0.1763, -0.1767, -0.2035, -0.2341, -0.228, -0.1722, -0.1448, -0.1535, -0.1478, -0.1766, -0.1915, -0.1915, -0.1732, -0.1708, -0.1696, -0.1729, -0.1786, -0.1907, -0.1737, -0.2001, -0.2055, -0.2073, -0.2197, -0.2105, -0.1726, -0.1783, -0.1403, -0.1482, -0.1507, -0.1646, -0.199, -0.1921, -0.2011, -0.2145, -0.2048, -0.228, -0.1964, -0.2069, -0.2057, -0.229, -0.2508, -0.2689, -0.2288, -0.2189, -0.2163, -0.2163, -0.2166, -0.1679, -0.1264, -0.1157, -0.0742, -0.0538, -0.0512, -0.0115, -0.0072, 0.0129, -0.0043, 0.002, 0.0531, 0.0463, 0.0533, 0.042, -0.0038, 0.0102, 0.0401, 0.0497, 0.0378, 0.0648, 0.06, 0.0279, 0.0714, 0.0675, 0.0447, 0.0385, 0.0958, 0.0594, 0.0482, 0.0242, 0.0409, 0.0329, 0.0319, 0.0319]}
{"unit_id": "quote:1994327901634777224", "ticker_idx": 1, "ts": "2025-11-28T08:49:59+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom has adopted Intel's EMIB and entered the back-end process of AWS Trainium 3.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Broadcom wins Trainium 3 back-end with Intel EMIB, taking share from Marvell/Alchip; active on Trainium 4.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Rumor: Broadcom has adopted Intel’s EMIB and entered the back-end process of AWS Trainium 3.\n\nAlso, Broadcom is said to be very active on Trainium 4.\n\n$AVGO $INTC", "reply_text": null, "parent_text": null, "quoted_text": "Just in: I'm not sure if I can share this, but it seems  that $AVGO did break into Trainium 3 back-end by proposing an Intel-EMIB solution and will gain share from Alchip and $MRVL (should be proportionally decreased?).\n\nOne more thing: $AVGO is also very \"active\" in Trainium 4.", "parent_author": null, "ret_m1d": -0.09245561817138881, "ret_signed_m1d": -0.09245561817138881, "alpha_spy_m1d": -0.08702684428655771, "alpha_c_m1d": -0.07948291744846436, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013560232572378372, "ret_signed_p1d": -0.013560232572378372, "alpha_spy_p1d": -0.00899481696982718, "alpha_c_p1d": -0.015490504675310546, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.020956569165227368, "ret_signed_p1w": 0.020956569165227368, "alpha_spy_p1w": 0.017590980855091365, "alpha_c_p1w": -0.0135046926121003, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": -0.09566077260219807, "ret_signed_p1m": -0.09566077260219807, "alpha_spy_p1m": -0.10516121672153078, "alpha_c_p1m": -0.1327428004332929, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": 0.15581852763412307, "ret_signed_p3m": 0.15581852763412307, "alpha_spy_p3m": 0.13853959545346717, "alpha_c_p3m": -0.057665152516148455, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 1.954634918153773, "ret_signed_p6m": 1.954634918153773, "alpha_spy_p6m": 1.8573322301472226, "alpha_c_p6m": 1.3135730695000363, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.4031, -0.4083, -0.3932, -0.3962, -0.3964, -0.3974, -0.3893, -0.3932, -0.4063, -0.3893, -0.377, -0.3861, -0.2463, -0.2707, -0.2909, -0.2766, -0.2303, -0.162, -0.1248, -0.1499, -0.1728, -0.1139, -0.0804, -0.092, -0.0979, -0.0836, -0.0772, -0.068, -0.1033, -0.0823, -0.1215, -0.0841, -0.0917, -0.0875, -0.0607, -0.0602, -0.0897, -0.0592, -0.0562, -0.0251, 0.0239, 0.0192, -0.0099, -0.0141, -0.0261, -0.087, -0.0537, -0.0819, -0.0599, -0.052, -0.0661, -0.0658, -0.1146, -0.1243, -0.1442, -0.1536, -0.1344, -0.1711, -0.1494, -0.1176, -0.1166, -0.0925, -0.0925, 0.0, -0.0136, 0.0717, 0.0789, -0.0015, 0.021, -0.0064, -0.0015, 0.0054, -0.0259, -0.0678, -0.0752, -0.0801, -0.1112, -0.1055, -0.0922, -0.1033, -0.1038, -0.1085, -0.1085, -0.1075, -0.0957, -0.0804, -0.0902, -0.0902, -0.0291, -0.0293, -0.0128, 0.051, 0.0136, 0.123, 0.0863, 0.1659, 0.2012, 0.1913, 0.1578, 0.1578, 0.1972, 0.3375, 0.3393, 0.1112, 0.0476, 0.0831, 0.2027, 0.1997, 0.1457, 0.2034, 0.2143, 0.1982, 0.1893, 0.2473, 0.2387, 0.162, 0.1906, 0.146, 0.1536, 0.1536, 0.1386, 0.1208, 0.1001, 0.0875, 0.0757, 0.1371, 0.1558, 0.1208, 0.1245, 0.1218, 0.0626, 0.1238, 0.1329, 0.0705, 0.1238, 0.1534, 0.1829, 0.1156, 0.1285, 0.1282, 0.0863, 0.1102, 0.1386, 0.0816, 0.0851, 0.0863, 0.1632, 0.0873, 0.0634, 0.0155, 0.088, 0.1842, 0.2421, 0.2421, 0.252, 0.3045, 0.4534, 0.5217, 0.538, 0.607, 0.5732, 0.6011, 0.6889, 0.6889, 0.6198, 0.6336, 0.6092, 0.6464, 1.035, 1.0954, 1.0838, 1.336, 1.3294, 1.4561, 1.3614, 1.6664, 1.7862, 1.7027, 2.0799, 2.1913, 1.9736, 1.9657, 1.8582, 1.6817, 1.6669, 1.7318, 1.9329, 1.9216, 1.9546, 1.9546]}
{"unit_id": "quote:1994327901634777224", "ticker_idx": 2, "ts": "2025-11-28T08:49:59+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "will gain share from Alchip and $MRVL (should be proportionally decreased?)", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Broadcom wins Trainium 3 back-end with Intel EMIB, taking share from Marvell/Alchip; active on Trainium 4.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Rumor: Broadcom has adopted Intel’s EMIB and entered the back-end process of AWS Trainium 3.\n\nAlso, Broadcom is said to be very active on Trainium 4.\n\n$AVGO $INTC", "reply_text": null, "parent_text": null, "quoted_text": "Just in: I'm not sure if I can share this, but it seems  that $AVGO did break into Trainium 3 back-end by proposing an Intel-EMIB solution and will gain share from Alchip and $MRVL (should be proportionally decreased?).\n\nOne more thing: $AVGO is also very \"active\" in Trainium 4.", "parent_author": null, "ret_m1d": -0.018791984041770515, "ret_signed_m1d": 0.018791984041770515, "alpha_spy_m1d": 0.01336321015693942, "alpha_c_m1d": 0.005819283318846069, "bench_spy_m1d": -0.005428773884831095, "bench_c_m1d": -0.012972700722924446, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019015591332342208, "ret_signed_p1d": -0.019015591332342208, "alpha_spy_p1d": -0.0235810069348934, "alpha_c_p1d": -0.017085319229410034, "bench_spy_p1d": -0.004565415602551193, "bench_c_p1d": 0.0019302721029321734, "ret_p1w": 0.10637580687028914, "ret_signed_p1w": -0.10637580687028914, "alpha_spy_p1w": -0.10301021856015313, "alpha_c_p1w": -0.07191454509296147, "bench_spy_p1w": 0.003365588310136003, "bench_c_p1w": 0.03446126177732767, "ret_p1m": -0.04071592270177071, "ret_signed_p1m": 0.04071592270177071, "alpha_spy_p1m": 0.05021636682110342, "alpha_c_p1m": 0.07779795053286553, "bench_spy_p1m": 0.009500444119332707, "bench_c_p1m": 0.03708202783109482, "ret_p3m": -0.09420343598192482, "ret_signed_p3m": 0.09420343598192482, "alpha_spy_p3m": 0.11148236816258073, "alpha_c_p3m": 0.30768711613219635, "bench_spy_p3m": 0.017278932180655904, "bench_c_p3m": 0.21348368015027153, "ret_p6m": 1.198765093064893, "ret_signed_p6m": -1.198765093064893, "alpha_spy_p6m": -1.1014624050583426, "alpha_c_p6m": -0.5577032444111563, "bench_spy_p6m": 0.0973026880065504, "bench_c_p6m": 0.6410618486537367, "price_path": [-0.2779, -0.3035, -0.2835, -0.2921, -0.2622, -0.2528, -0.2499, -0.2556, -0.2471, -0.2462, -0.2303, -0.2066, -0.1702, -0.1699, -0.1557, -0.1659, -0.1047, -0.0631, -0.0703, -0.079, -0.0602, -0.0623, -0.0364, -0.0362, -0.006, -0.0278, 0.034, 0.0136, -0.0424, -0.0001, -0.0356, -0.0057, -0.0131, -0.0162, -0.0398, -0.0575, -0.0934, -0.0742, -0.0589, -0.0077, -0.0105, 0.0084, -0.0093, 0.0485, 0.0109, -0.0202, 0.0391, 0.044, 0.017, 0.0428, -0.0008, -0.0008, -0.021, -0.033, -0.0666, -0.1199, -0.0904, -0.1423, -0.1337, -0.0628, -0.0668, -0.0188, -0.0188, 0.0, 0.019, 0.039, 0.1208, 0.0983, 0.1064, 0.0291, -0.0056, 0.0343, 0.0003, -0.0556, -0.0575, -0.0596, -0.0861, -0.0551, -0.0594, -0.0515, -0.0192, -0.0326, -0.0326, -0.0342, -0.0407, -0.0295, -0.0494, -0.0494, -0.0001, 0.0093, -0.0131, -0.0532, -0.0666, -0.0685, -0.0722, -0.0704, -0.091, -0.1002, -0.0994, -0.0994, -0.1067, -0.076, -0.0698, -0.1019, -0.0847, -0.0717, -0.064, -0.0895, -0.1166, -0.1195, -0.1544, -0.1747, -0.1693, -0.1014, -0.0782, -0.082, -0.0895, -0.1243, -0.1201, -0.1201, -0.1156, -0.1147, -0.1089, -0.1103, -0.1292, -0.1225, -0.0942, -0.1124, -0.0856, -0.0949, -0.1324, -0.1259, -0.1529, 0.0026, 0.0371, 0.0444, 0.0124, -0.0186, -0.0165, 0.0251, 0.0163, -0.0192, 0.0022, -0.016, 0.0092, 0.0339, 0.102, 0.0934, 0.0621, -0.0171, 0.1087, 0.1945, 0.199, 0.199, 0.2258, 0.2244, 0.2811, 0.3425, 0.439, 0.4705, 0.4988, 0.5074, 0.4937, 0.5644, 0.6557, 0.6946, 0.7619, 0.8542, 0.8402, 0.7718, 0.7161, 0.7535, 0.8496, 0.8473, 0.8329, 0.8899, 0.928, 0.792, 0.9053, 0.9133, 0.8423, 0.9929, 1.0448, 0.9811, 0.8919, 0.9741, 1.092, 1.1356, 1.1988, 1.1988]}
{"unit_id": "thread:1982948031524941889", "ticker_idx": 0, "ts": "2025-10-27T23:10:26+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the contract price of server memory for the fourth quarter has skyrocketed by 40–50%… DDR5 prices have risen sharply since late September… shortages and price increases are expected to continue", "tweet_type": "analysis", "timeline": "quarters", "summary": "Severe DRAM shortage with server memory prices up 40–50%; shortage expected to persist through end-2026, bullish for memory makers including Samsung and large Korean DRAM producers.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe Memory Shortage… Major Customers’ Fulfillment Rate Plunges to 70%, Server DRAM Prices Surge 50%!\n\nThe memory shortage storm has entered a “grim” phase. Recently, upstream international original manufacturers have successively finalized contract prices for the fourth quarter of 2025. It is reported that due to fierce competition for supply, the order fulfillment rate of many first-tier major customers in the U.S. and China is only about 70%. Meanwhile, the contract price of server memory for the fourth quarter has skyrocketed by 40–50%, and the capacity allocation effect has also impacted other non-priority customers — their fulfillment rate over the next two quarters may fall below 40%.\n\nThe explosion of AI demand has triggered a widespread memory shortage and a rapid price surge across the entire industry. According to the supply chain, the shortage began to take a “turning point” in the third quarter, and now it’s too late to secure supply. Many memory module manufacturers may face an “out of stock” crisis in October, while large Korean manufacturers have “stopped giving quotes altogether.” DDR5 DRAM dies are completely sold out — buyers can only queue again for new orders in November.\n\nIndustry insiders privately admit that judging from the attitude of the original manufacturers, it’s not just October — the entire fourth quarter will likely be difficult to secure supply. “If you really have an urgent order,” one said, “just buy at the spot price.”\n\nLooking at the spot market, DDR5 prices have risen sharply since late September, becoming the fastest-rising product line recently. Compared with the $7–8 range for 16Gb DDR5 in September, prices have now jumped to around $13, and shortages and price increases are expected to continue.\n\nIt is learned that Samsung Electronics recently announced its fourth-quarter contract pricing. With strong server demand and production capacity prioritized for advanced nodes, server applications have become the focal point of this wave of price hikes. Among them, enterprise SSDs rose by 15–35%, while server memory (RDIMM) prices soared by 40–50%, far exceeding the previously expected 30% increase.\n\nAlthough memory makers have significantly raised prices for the fourth quarter, the supply chain estimates that the pressure of the supply scramble has already spread across the industry. Major U.S. and Chinese cloud service providers (CSPs) cannot fully meet demand, with order fulfillment around 70%, based on the average delivery volume from the first three quarters of 2025.\n\nAs the four major U.S. hyperscalers account for about 50% of global data center share, and with OpenAI and Oracle expected to ramp up data center shipments in the second half of 2026, strong server memory demand means the DRAM supply shortage will likely persist until the end of 2026.\n\nMeanwhile, the stocking pressure on small and medium-sized firms, distributors, and module makers continues to intensify. Since they cannot compete with first-tier manufacturers in delivery prioritization, their average fulfillment rate is only 35–40%, depending on company size and shipping capacity. The memory industry believes that at least until the first quarter of 2026, it will be difficult to resolve the “no stock available” dilemma.\n\nNanya Technology previously pointed out that DDR4 currently accounts for about 20% of the global DRAM market, of which more than 10% is DDR4 and the rest LPDDR4. Although PC makers are rapidly switching to DDR5, TV, networking, consumer, and IPC applications are less willing to transition. The industry also believes that data center switches still largely use DDR4. While the supply gap will gradually narrow over time, the long-tail effect will not disappear within the next one to two years, leaving more than 10% capacity shortfall.\n\nIndustry experts note that overall, the memory shortage not only drives price increases but may also trigger supply chain disruptions. As PC makers shift to DDR5 as the mainstream specification, DRAM contract prices continue to rise, OEM inventory levels are falling, and PC DRAM shortages are tightening further.\n\nAlthough the supply chain pulled in orders early in the first half of 2025 due to tariff concerns, even if tariffs ease and the economy recovers in 2026, PC OEM customers may not be able to secure sufficient volumes even at high prices. When PC demand rebounds, it could resemble the “supply chain collapse and shortage” seen during the pandemic.\n\nRecently, Xiaomi executives also revealed that the rise in memory chip costs has far exceeded expectations. This cost pressure has already been reflected in new product pricing and could intensify further in the future — showing that the memory price surge and shortage are gradually spreading to the PC and smartphone industries.\n\n---\n\nhttps://t.co/f8jjUqVUYH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027669446585673607, "ret_signed_m1d": -0.027669446585673607, "alpha_spy_m1d": -0.01600929764191017, "alpha_c_m1d": -0.003221006350285216, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014151459320133028, "ret_signed_p1d": -0.014151459320133028, "alpha_spy_p1d": -0.016807458276724106, "alpha_c_p1d": -0.02295845320545028, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.10480926356812585, "ret_signed_p1w": 0.10480926356812585, "alpha_spy_p1w": 0.10758198315339944, "alpha_c_p1w": 0.08769527885925547, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": -0.012083281799602311, "ret_signed_p1m": -0.012083281799602311, "alpha_spy_p1m": 0.0028312496768069986, "alpha_c_p1m": 0.043287018137597465, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.5732372537610889, "ret_signed_p3m": 0.5732372537610889, "alpha_spy_p3m": 0.5648082865496423, "alpha_c_p3m": 0.45062398380400126, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 1.1663555257714706, "ret_signed_p6m": 1.1663555257714706, "alpha_spy_p6m": 1.1330109593093294, "alpha_c_p6m": 0.8714012617208657, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.4262, -0.4324, -0.4566, -0.4496, -0.4436, -0.4507, -0.4383, -0.4269, -0.4156, -0.408, -0.405, -0.4054, -0.4121, -0.4188, -0.4239, -0.4343, -0.4426, -0.4334, -0.4297, -0.4323, -0.4303, -0.4216, -0.4255, -0.4404, -0.4336, -0.4297, -0.4185, -0.4046, -0.4004, -0.3833, -0.3625, -0.342, -0.312, -0.3061, -0.2845, -0.2956, -0.2616, -0.271, -0.2605, -0.2476, -0.2547, -0.2604, -0.2813, -0.2593, -0.2561, -0.219, -0.1822, -0.1758, -0.1711, -0.1791, -0.1626, -0.169, -0.1498, -0.1446, -0.161, -0.1356, -0.0921, -0.0835, -0.0638, -0.0766, -0.0772, -0.0735, -0.0277, 0.0, -0.0142, 0.0193, 0.0333, 0.0385, 0.1048, 0.0381, 0.0492, 0.0546, 0.0416, 0.0899, 0.0891, 0.0922, 0.0761, 0.0404, 0.0728, 0.0208, 0.0077, -0.0105, -0.0515, -0.0209, -0.0121, 0.0111, 0.0261, 0.017, 0.029, 0.0485, 0.0403, 0.0247, 0.0527, 0.0916, 0.0894, 0.1183, 0.0944, 0.0771, 0.0476, 0.0185, 0.0377, 0.0722, 0.0912, 0.1416, 0.1469, 0.1639, 0.1639, 0.1871, 0.2375, 0.243, 0.2321, 0.2321, 0.322, 0.3604, 0.4291, 0.4401, 0.4226, 0.4432, 0.4468, 0.4243, 0.4284, 0.4496, 0.5099, 0.5162, 0.4931, 0.5419, 0.5732, 0.5832, 0.5479, 0.6441, 0.7171, 0.7248, 0.7224, 0.6747, 0.751, 0.6913, 0.6282, 0.6419, 0.6805, 0.6562, 0.7089, 0.7674, 0.7674, 0.7674, 0.7494, 0.7815, 0.8137, 0.8647, 0.8645, 0.9167, 0.953, 1.032, 0.9985, 0.9989, 0.8025, 0.703, 0.8185, 0.7562, 0.6818, 0.8137, 0.8548, 0.8121, 0.8166, 0.8974, 0.9421, 1.0439, 0.9643, 0.9245, 0.8072, 0.8381, 0.8211, 0.7128, 0.7154, 0.6132, 0.5664, 0.7419, 0.6606, 0.715, 0.7614, 0.7911, 0.9544, 0.9336, 0.9569, 0.9576, 1.0743, 1.0954, 1.1302, 1.1051, 1.1139, 1.1664]}
{"unit_id": "thread:1982948031524941889", "ticker_idx": 1, "ts": "2025-10-27T23:10:26+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics recently announced its fourth-quarter contract pricing… server memory (RDIMM) prices soared by 40–50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Severe DRAM shortage with server memory prices up 40–50%; shortage expected to persist through end-2026, bullish for memory makers including Samsung and large Korean DRAM producers.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe Memory Shortage… Major Customers’ Fulfillment Rate Plunges to 70%, Server DRAM Prices Surge 50%!\n\nThe memory shortage storm has entered a “grim” phase. Recently, upstream international original manufacturers have successively finalized contract prices for the fourth quarter of 2025. It is reported that due to fierce competition for supply, the order fulfillment rate of many first-tier major customers in the U.S. and China is only about 70%. Meanwhile, the contract price of server memory for the fourth quarter has skyrocketed by 40–50%, and the capacity allocation effect has also impacted other non-priority customers — their fulfillment rate over the next two quarters may fall below 40%.\n\nThe explosion of AI demand has triggered a widespread memory shortage and a rapid price surge across the entire industry. According to the supply chain, the shortage began to take a “turning point” in the third quarter, and now it’s too late to secure supply. Many memory module manufacturers may face an “out of stock” crisis in October, while large Korean manufacturers have “stopped giving quotes altogether.” DDR5 DRAM dies are completely sold out — buyers can only queue again for new orders in November.\n\nIndustry insiders privately admit that judging from the attitude of the original manufacturers, it’s not just October — the entire fourth quarter will likely be difficult to secure supply. “If you really have an urgent order,” one said, “just buy at the spot price.”\n\nLooking at the spot market, DDR5 prices have risen sharply since late September, becoming the fastest-rising product line recently. Compared with the $7–8 range for 16Gb DDR5 in September, prices have now jumped to around $13, and shortages and price increases are expected to continue.\n\nIt is learned that Samsung Electronics recently announced its fourth-quarter contract pricing. With strong server demand and production capacity prioritized for advanced nodes, server applications have become the focal point of this wave of price hikes. Among them, enterprise SSDs rose by 15–35%, while server memory (RDIMM) prices soared by 40–50%, far exceeding the previously expected 30% increase.\n\nAlthough memory makers have significantly raised prices for the fourth quarter, the supply chain estimates that the pressure of the supply scramble has already spread across the industry. Major U.S. and Chinese cloud service providers (CSPs) cannot fully meet demand, with order fulfillment around 70%, based on the average delivery volume from the first three quarters of 2025.\n\nAs the four major U.S. hyperscalers account for about 50% of global data center share, and with OpenAI and Oracle expected to ramp up data center shipments in the second half of 2026, strong server memory demand means the DRAM supply shortage will likely persist until the end of 2026.\n\nMeanwhile, the stocking pressure on small and medium-sized firms, distributors, and module makers continues to intensify. Since they cannot compete with first-tier manufacturers in delivery prioritization, their average fulfillment rate is only 35–40%, depending on company size and shipping capacity. The memory industry believes that at least until the first quarter of 2026, it will be difficult to resolve the “no stock available” dilemma.\n\nNanya Technology previously pointed out that DDR4 currently accounts for about 20% of the global DRAM market, of which more than 10% is DDR4 and the rest LPDDR4. Although PC makers are rapidly switching to DDR5, TV, networking, consumer, and IPC applications are less willing to transition. The industry also believes that data center switches still largely use DDR4. While the supply gap will gradually narrow over time, the long-tail effect will not disappear within the next one to two years, leaving more than 10% capacity shortfall.\n\nIndustry experts note that overall, the memory shortage not only drives price increases but may also trigger supply chain disruptions. As PC makers shift to DDR5 as the mainstream specification, DRAM contract prices continue to rise, OEM inventory levels are falling, and PC DRAM shortages are tightening further.\n\nAlthough the supply chain pulled in orders early in the first half of 2025 due to tariff concerns, even if tariffs ease and the economy recovers in 2026, PC OEM customers may not be able to secure sufficient volumes even at high prices. When PC demand rebounds, it could resemble the “supply chain collapse and shortage” seen during the pandemic.\n\nRecently, Xiaomi executives also revealed that the rise in memory chip costs has far exceeded expectations. This cost pressure has already been reflected in new product pricing and could intensify further in the future — showing that the memory price surge and shortage are gradually spreading to the PC and smartphone industries.\n\n---\n\nhttps://t.co/f8jjUqVUYH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03137258259156661, "ret_signed_m1d": -0.03137258259156661, "alpha_spy_m1d": -0.019712433647803174, "alpha_c_m1d": -0.013245529566919823, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.01812705302464679, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024509789177127317, "ret_signed_p1d": -0.024509789177127317, "alpha_spy_p1d": -0.027165788133718394, "alpha_c_p1d": -0.03440944268004342, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.009899653502916106, "ret_p1w": 0.08921562026478003, "ret_signed_p1w": 0.08921562026478003, "alpha_spy_p1w": 0.09198833985005361, "alpha_c_p1w": 0.07948310004733283, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.009732520217447194, "ret_p1m": -0.026470609331187056, "ret_signed_p1m": -0.026470609331187056, "alpha_spy_p1m": -0.011556077854777747, "alpha_c_p1m": 0.0355360442725412, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.06200665360372826, "ret_p3m": 0.5003956115080337, "ret_signed_p3m": 0.5003956115080337, "alpha_spy_p3m": 0.4919666442965871, "alpha_c_p3m": 0.5298286735013822, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": -0.029433061993348497, "ret_p6m": 1.1619615543818709, "ret_signed_p6m": 1.1619615543818709, "alpha_spy_p6m": 1.1286169879197296, "alpha_c_p6m": 1.124387883452265, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.03757367092960595, "price_path": [-0.2914, -0.3031, -0.3275, -0.3197, -0.3177, -0.3285, -0.3119, -0.2992, -0.307, -0.306, -0.2982, -0.3012, -0.3012, -0.3168, -0.3168, -0.3119, -0.3109, -0.3031, -0.3021, -0.3138, -0.3109, -0.3207, -0.3197, -0.3402, -0.3256, -0.3187, -0.3158, -0.3217, -0.3158, -0.3021, -0.2914, -0.2836, -0.2641, -0.2533, -0.225, -0.2367, -0.2162, -0.2162, -0.185, -0.1733, -0.1665, -0.1596, -0.187, -0.1745, -0.1775, -0.1569, -0.1201, -0.1201, -0.1201, -0.1201, -0.1201, -0.1201, -0.0745, -0.0853, -0.102, -0.0686, -0.0422, -0.0402, -0.0382, -0.0441, -0.0333, -0.0539, -0.0314, 0.0, -0.0245, -0.0147, 0.0206, 0.0539, 0.0892, 0.0284, -0.0137, -0.0275, -0.0402, -0.0137, 0.0147, 0.0108, 0.0078, -0.0471, -0.0137, -0.0412, -0.0539, -0.0137, -0.0706, -0.052, -0.0265, 0.0078, 0.0147, -0.0147, -0.0118, 0.0137, 0.0245, 0.0304, 0.0627, 0.0735, 0.0627, 0.0588, 0.052, 0.0676, 0.0275, 0.0078, 0.0578, 0.0549, 0.0422, 0.0833, 0.0931, 0.0892, 0.0892, 0.1471, 0.1773, 0.1812, 0.1812, 0.1812, 0.2659, 0.3605, 0.3684, 0.3891, 0.3674, 0.3694, 0.3674, 0.3556, 0.3822, 0.4176, 0.4669, 0.4708, 0.4304, 0.4728, 0.5004, 0.4984, 0.4984, 0.5713, 0.5999, 0.5831, 0.5812, 0.4817, 0.6501, 0.6659, 0.5694, 0.5625, 0.6393, 0.6334, 0.6531, 0.7595, 0.7851, 0.7851, 0.7851, 0.7851, 0.8718, 0.8728, 0.9014, 0.9703, 1.0048, 1.1476, 1.1329, 1.1329, 0.922, 0.6964, 0.8876, 0.8541, 0.7092, 0.8511, 0.8718, 0.8511, 0.8078, 0.859, 0.9102, 1.0541, 0.9752, 0.9644, 0.8353, 0.8688, 0.8619, 0.7743, 0.7743, 0.7404, 0.6506, 0.8722, 0.7612, 0.8382, 0.9063, 0.9398, 1.078, 1.0139, 1.0336, 0.9843, 1.0386, 1.083, 1.1472, 1.1323, 1.1175, 1.162]}
{"unit_id": "thread:1982948031524941889", "ticker_idx": 2, "ts": "2025-10-27T23:10:26+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "the DRAM supply shortage will likely persist until the end of 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Severe DRAM shortage with server memory prices up 40–50%; shortage expected to persist through end-2026, bullish for memory makers including Samsung and large Korean DRAM producers.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe Memory Shortage… Major Customers’ Fulfillment Rate Plunges to 70%, Server DRAM Prices Surge 50%!\n\nThe memory shortage storm has entered a “grim” phase. Recently, upstream international original manufacturers have successively finalized contract prices for the fourth quarter of 2025. It is reported that due to fierce competition for supply, the order fulfillment rate of many first-tier major customers in the U.S. and China is only about 70%. Meanwhile, the contract price of server memory for the fourth quarter has skyrocketed by 40–50%, and the capacity allocation effect has also impacted other non-priority customers — their fulfillment rate over the next two quarters may fall below 40%.\n\nThe explosion of AI demand has triggered a widespread memory shortage and a rapid price surge across the entire industry. According to the supply chain, the shortage began to take a “turning point” in the third quarter, and now it’s too late to secure supply. Many memory module manufacturers may face an “out of stock” crisis in October, while large Korean manufacturers have “stopped giving quotes altogether.” DDR5 DRAM dies are completely sold out — buyers can only queue again for new orders in November.\n\nIndustry insiders privately admit that judging from the attitude of the original manufacturers, it’s not just October — the entire fourth quarter will likely be difficult to secure supply. “If you really have an urgent order,” one said, “just buy at the spot price.”\n\nLooking at the spot market, DDR5 prices have risen sharply since late September, becoming the fastest-rising product line recently. Compared with the $7–8 range for 16Gb DDR5 in September, prices have now jumped to around $13, and shortages and price increases are expected to continue.\n\nIt is learned that Samsung Electronics recently announced its fourth-quarter contract pricing. With strong server demand and production capacity prioritized for advanced nodes, server applications have become the focal point of this wave of price hikes. Among them, enterprise SSDs rose by 15–35%, while server memory (RDIMM) prices soared by 40–50%, far exceeding the previously expected 30% increase.\n\nAlthough memory makers have significantly raised prices for the fourth quarter, the supply chain estimates that the pressure of the supply scramble has already spread across the industry. Major U.S. and Chinese cloud service providers (CSPs) cannot fully meet demand, with order fulfillment around 70%, based on the average delivery volume from the first three quarters of 2025.\n\nAs the four major U.S. hyperscalers account for about 50% of global data center share, and with OpenAI and Oracle expected to ramp up data center shipments in the second half of 2026, strong server memory demand means the DRAM supply shortage will likely persist until the end of 2026.\n\nMeanwhile, the stocking pressure on small and medium-sized firms, distributors, and module makers continues to intensify. Since they cannot compete with first-tier manufacturers in delivery prioritization, their average fulfillment rate is only 35–40%, depending on company size and shipping capacity. The memory industry believes that at least until the first quarter of 2026, it will be difficult to resolve the “no stock available” dilemma.\n\nNanya Technology previously pointed out that DDR4 currently accounts for about 20% of the global DRAM market, of which more than 10% is DDR4 and the rest LPDDR4. Although PC makers are rapidly switching to DDR5, TV, networking, consumer, and IPC applications are less willing to transition. The industry also believes that data center switches still largely use DDR4. While the supply gap will gradually narrow over time, the long-tail effect will not disappear within the next one to two years, leaving more than 10% capacity shortfall.\n\nIndustry experts note that overall, the memory shortage not only drives price increases but may also trigger supply chain disruptions. As PC makers shift to DDR5 as the mainstream specification, DRAM contract prices continue to rise, OEM inventory levels are falling, and PC DRAM shortages are tightening further.\n\nAlthough the supply chain pulled in orders early in the first half of 2025 due to tariff concerns, even if tariffs ease and the economy recovers in 2026, PC OEM customers may not be able to secure sufficient volumes even at high prices. When PC demand rebounds, it could resemble the “supply chain collapse and shortage” seen during the pandemic.\n\nRecently, Xiaomi executives also revealed that the rise in memory chip costs has far exceeded expectations. This cost pressure has already been reflected in new product pricing and could intensify further in the future — showing that the memory price surge and shortage are gradually spreading to the PC and smartphone industries.\n\n---\n\nhttps://t.co/f8jjUqVUYH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004906819688392017, "ret_signed_m1d": -0.004906819688392017, "alpha_spy_m1d": 0.006753329255371421, "alpha_c_m1d": 0.019541620546996374, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008223571690148201, "ret_signed_p1d": 0.008223571690148201, "alpha_spy_p1d": 0.0055675727335571246, "alpha_c_p1d": -0.0005834221951690477, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.0663335135923484, "ret_signed_p1w": 0.0663335135923484, "alpha_spy_p1w": 0.06910623317762199, "alpha_c_p1w": 0.049219528883478025, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": 0.020127258146590332, "ret_signed_p1m": 0.020127258146590332, "alpha_spy_p1m": 0.03504178962299964, "alpha_c_p1m": 0.0754975580837901, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.8070903822032314, "ret_signed_p3m": 0.8070903822032314, "alpha_spy_p3m": 0.7986614149917848, "alpha_c_p3m": 0.6844771122461437, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 1.0433912174310884, "ret_signed_p6m": 1.0433912174310884, "alpha_spy_p6m": 1.0100466509689472, "alpha_c_p6m": 0.7484369533804835, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.479, -0.5044, -0.5238, -0.5107, -0.5048, -0.5061, -0.492, -0.4602, -0.4382, -0.4199, -0.4357, -0.4311, -0.4512, -0.439, -0.4458, -0.4678, -0.4743, -0.4657, -0.4714, -0.471, -0.4654, -0.4461, -0.4596, -0.4596, -0.462, -0.4609, -0.436, -0.4035, -0.4031, -0.3859, -0.3643, -0.3163, -0.2861, -0.2836, -0.2789, -0.2736, -0.2331, -0.2611, -0.2525, -0.2444, -0.2657, -0.2879, -0.2859, -0.2558, -0.2403, -0.1729, -0.1657, -0.1466, -0.1324, -0.1563, -0.107, -0.1262, -0.1749, -0.1242, -0.1501, -0.1279, -0.0798, -0.0805, -0.0606, -0.0809, -0.0983, -0.0608, -0.0049, 0.0, 0.0082, 0.0297, 0.0178, 0.0167, 0.0663, -0.0094, 0.0791, 0.0828, 0.081, 0.1508, 0.0955, 0.1127, 0.0766, 0.1214, 0.0993, 0.0382, 0.0264, -0.0851, -0.0578, 0.0174, 0.0201, 0.0462, 0.0462, 0.0744, 0.0925, 0.0881, 0.0639, 0.0298, 0.0778, 0.1219, 0.1468, 0.1981, 0.1743, 0.0956, 0.0791, 0.0564, 0.0246, 0.1293, 0.2082, 0.2567, 0.2552, 0.3025, 0.3025, 0.2939, 0.338, 0.3301, 0.2973, 0.2973, 0.4337, 0.4188, 0.561, 0.5433, 0.4864, 0.5685, 0.5721, 0.5369, 0.5152, 0.5301, 0.6488, 0.6488, 0.659, 0.7686, 0.8071, 0.8165, 0.7685, 0.8646, 0.9784, 0.9808, 0.8857, 0.9899, 0.9064, 0.7245, 0.7403, 0.794, 0.7431, 0.6965, 0.8651, 0.8816, 0.8711, 0.8711, 0.8171, 0.9133, 0.8969, 0.9461, 0.9134, 0.8999, 0.9499, 0.8888, 0.8743, 0.8757, 0.7257, 0.8216, 0.8047, 0.6831, 0.7695, 0.8322, 0.903, 0.8424, 0.9369, 1.0081, 1.0985, 1.0987, 1.0193, 0.9222, 0.8379, 0.7978, 0.7367, 0.6156, 0.6236, 0.4633, 0.5362, 0.6727, 0.6653, 0.6653, 0.7177, 0.7169, 0.8495, 0.9167, 0.9125, 0.9396, 1.1174, 1.0745, 1.0791, 1.0693, 1.039, 1.0434]}
{"unit_id": "thread:1982948031524941889", "ticker_idx": 3, "ts": "2025-10-27T23:10:26+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "large Korean manufacturers have 'stopped giving quotes altogether'", "tweet_type": "analysis", "timeline": "quarters", "summary": "Severe DRAM shortage with server memory prices up 40–50%; shortage expected to persist through end-2026, bullish for memory makers including Samsung and large Korean DRAM producers.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe Memory Shortage… Major Customers’ Fulfillment Rate Plunges to 70%, Server DRAM Prices Surge 50%!\n\nThe memory shortage storm has entered a “grim” phase. Recently, upstream international original manufacturers have successively finalized contract prices for the fourth quarter of 2025. It is reported that due to fierce competition for supply, the order fulfillment rate of many first-tier major customers in the U.S. and China is only about 70%. Meanwhile, the contract price of server memory for the fourth quarter has skyrocketed by 40–50%, and the capacity allocation effect has also impacted other non-priority customers — their fulfillment rate over the next two quarters may fall below 40%.\n\nThe explosion of AI demand has triggered a widespread memory shortage and a rapid price surge across the entire industry. According to the supply chain, the shortage began to take a “turning point” in the third quarter, and now it’s too late to secure supply. Many memory module manufacturers may face an “out of stock” crisis in October, while large Korean manufacturers have “stopped giving quotes altogether.” DDR5 DRAM dies are completely sold out — buyers can only queue again for new orders in November.\n\nIndustry insiders privately admit that judging from the attitude of the original manufacturers, it’s not just October — the entire fourth quarter will likely be difficult to secure supply. “If you really have an urgent order,” one said, “just buy at the spot price.”\n\nLooking at the spot market, DDR5 prices have risen sharply since late September, becoming the fastest-rising product line recently. Compared with the $7–8 range for 16Gb DDR5 in September, prices have now jumped to around $13, and shortages and price increases are expected to continue.\n\nIt is learned that Samsung Electronics recently announced its fourth-quarter contract pricing. With strong server demand and production capacity prioritized for advanced nodes, server applications have become the focal point of this wave of price hikes. Among them, enterprise SSDs rose by 15–35%, while server memory (RDIMM) prices soared by 40–50%, far exceeding the previously expected 30% increase.\n\nAlthough memory makers have significantly raised prices for the fourth quarter, the supply chain estimates that the pressure of the supply scramble has already spread across the industry. Major U.S. and Chinese cloud service providers (CSPs) cannot fully meet demand, with order fulfillment around 70%, based on the average delivery volume from the first three quarters of 2025.\n\nAs the four major U.S. hyperscalers account for about 50% of global data center share, and with OpenAI and Oracle expected to ramp up data center shipments in the second half of 2026, strong server memory demand means the DRAM supply shortage will likely persist until the end of 2026.\n\nMeanwhile, the stocking pressure on small and medium-sized firms, distributors, and module makers continues to intensify. Since they cannot compete with first-tier manufacturers in delivery prioritization, their average fulfillment rate is only 35–40%, depending on company size and shipping capacity. The memory industry believes that at least until the first quarter of 2026, it will be difficult to resolve the “no stock available” dilemma.\n\nNanya Technology previously pointed out that DDR4 currently accounts for about 20% of the global DRAM market, of which more than 10% is DDR4 and the rest LPDDR4. Although PC makers are rapidly switching to DDR5, TV, networking, consumer, and IPC applications are less willing to transition. The industry also believes that data center switches still largely use DDR4. While the supply gap will gradually narrow over time, the long-tail effect will not disappear within the next one to two years, leaving more than 10% capacity shortfall.\n\nIndustry experts note that overall, the memory shortage not only drives price increases but may also trigger supply chain disruptions. As PC makers shift to DDR5 as the mainstream specification, DRAM contract prices continue to rise, OEM inventory levels are falling, and PC DRAM shortages are tightening further.\n\nAlthough the supply chain pulled in orders early in the first half of 2025 due to tariff concerns, even if tariffs ease and the economy recovers in 2026, PC OEM customers may not be able to secure sufficient volumes even at high prices. When PC demand rebounds, it could resemble the “supply chain collapse and shortage” seen during the pandemic.\n\nRecently, Xiaomi executives also revealed that the rise in memory chip costs has far exceeded expectations. This cost pressure has already been reflected in new product pricing and could intensify further in the future — showing that the memory price surge and shortage are gradually spreading to the PC and smartphone industries.\n\n---\n\nhttps://t.co/f8jjUqVUYH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.046728937477062193, "ret_signed_m1d": -0.046728937477062193, "alpha_spy_m1d": -0.035068788533298756, "alpha_c_m1d": -0.022280497241673802, "bench_spy_m1d": -0.011660148943763438, "bench_c_m1d": -0.02444844023538839, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026168160473419966, "ret_signed_p1d": -0.026168160473419966, "alpha_spy_p1d": -0.028824159430011043, "alpha_c_p1d": -0.034975154358737215, "bench_spy_p1d": 0.002655998956591077, "bench_c_p1d": 0.00880699388531725, "ret_p1w": 0.1588786568472491, "ret_signed_p1w": 0.1588786568472491, "alpha_spy_p1w": 0.1616513764325227, "alpha_c_p1w": 0.14176467213837873, "bench_spy_p1w": -0.0027727195852735864, "bench_c_p1w": 0.017113984708870378, "ret_p1m": -0.02990649421421021, "ret_signed_p1m": -0.02990649421421021, "alpha_spy_p1m": -0.014991962737800901, "alpha_c_p1m": 0.025463805722989563, "bench_spy_p1m": -0.01491453147640931, "bench_c_p1m": -0.055370299937199774, "ret_p3m": 0.4122257675720018, "ret_signed_p3m": 0.4122257675720018, "alpha_spy_p3m": 0.4037968003605552, "alpha_c_p3m": 0.28961249761491414, "bench_spy_p3m": 0.008428967211446592, "bench_c_p3m": 0.12261326995708766, "ret_p6m": 1.2937138055014525, "ret_signed_p6m": 1.2937138055014525, "alpha_spy_p6m": 1.2603692390393113, "alpha_c_p6m": 0.9987595414508477, "bench_spy_p6m": 0.033344566462141234, "bench_c_p6m": 0.2949542640506049, "price_path": [-0.5082, -0.4895, -0.5185, -0.5185, -0.5082, -0.5175, -0.511, -0.5213, -0.5017, -0.4979, -0.4811, -0.4839, -0.4839, -0.5007, -0.5091, -0.5231, -0.5427, -0.5315, -0.5157, -0.5119, -0.5147, -0.4981, -0.4972, -0.5215, -0.5131, -0.5093, -0.5037, -0.4888, -0.4822, -0.4617, -0.4318, -0.4262, -0.386, -0.3813, -0.3495, -0.3766, -0.3355, -0.3355, -0.3439, -0.3252, -0.3318, -0.3336, -0.371, -0.3477, -0.3505, -0.3271, -0.2607, -0.2607, -0.2607, -0.2607, -0.2607, -0.2607, -0.2, -0.2243, -0.2308, -0.2103, -0.1542, -0.1299, -0.0925, -0.1047, -0.1, -0.1056, -0.0467, 0.0, -0.0262, 0.043, 0.0617, 0.0449, 0.1589, 0.0953, 0.0822, 0.1084, 0.0841, 0.1327, 0.157, 0.1533, 0.1439, 0.0467, 0.1327, 0.0654, 0.0505, 0.0673, -0.0262, -0.028, -0.0299, -0.0206, 0.0176, -0.0086, 0.0063, 0.0437, 0.0325, 0.0138, 0.0176, 0.0793, 0.0587, 0.098, 0.0568, 0.0681, 0.0363, -0.0086, 0.0306, 0.0325, 0.0232, 0.0849, 0.0924, 0.0999, 0.0999, 0.1204, 0.1971, 0.2177, 0.2177, 0.2177, 0.2663, 0.3019, 0.358, 0.3879, 0.4141, 0.3917, 0.401, 0.3804, 0.3879, 0.401, 0.4141, 0.4291, 0.3898, 0.3842, 0.4122, 0.4347, 0.3767, 0.4964, 0.5731, 0.6105, 0.7003, 0.5525, 0.6965, 0.6834, 0.575, 0.5693, 0.6591, 0.6386, 0.6086, 0.661, 0.646, 0.646, 0.646, 0.646, 0.6722, 0.7751, 0.7788, 0.8799, 0.9042, 1.0595, 0.9883, 0.9883, 0.7596, 0.591, 0.7634, 0.7315, 0.5666, 0.7578, 0.7896, 0.7428, 0.7053, 0.8252, 0.8177, 0.9789, 0.8983, 0.8871, 0.7484, 0.8477, 0.8646, 0.7484, 0.7484, 0.636, 0.5123, 0.6809, 0.5554, 0.6416, 0.6603, 0.7165, 0.9358, 0.8702, 0.9245, 0.9489, 1.067, 1.1288, 1.1644, 1.1138, 1.185, 1.2937]}
{"unit_id": "orig:1985479798395977784", "ticker_idx": 0, "ts": "2025-11-03T22:50:47+00:00", "symbol": "DLR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "positive signals are strengthening for the data center value chain, including companies such as Digital Realty and Equinix. These companies are expected to benefit over the medium term due to the tight supply in the data center market", "tweet_type": "news_reaction", "timeline": "months", "summary": "Sharing Goldman read-through: raised hyperscaler CAPEX forecasts drive medium-term bullish view on DLR and EQIX on tight supply and accelerating pricing power.", "resolved_tickers": ["DLR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "REIT - Specialty", "unit_kind": "original", "lang": "en", "tweet_text": "251104_Americas Data Centers: Read-through to our coverage from 3Q25 hyperscaler earnings - Goldman Sachs\n\n(1) Following 3Q25 results, we have raised our 2025(F)–2026(F) CAPEX forecasts for the major hyperscalers.\n\n(2) CAPEX estimates\n■ 2025(F): $377.9bn → $403.9bn [+7%]\n■ 2026(F): $459.7bn → $551.7bn [+20%]\n\n(3) This is because GPU-centric infrastructure investment is accelerating amid a surge in AI and cloud demand.\n\n(4) Meanwhile, positive signals are strengthening for the data center value chain, including companies such as Digital Realty and Equinix.\n\n(5) These companies are expected to benefit over the medium term due to the tight supply in the data center market.\n\n(6) In particular, Digital Realty noted that when renewing contracts of 1MW or larger, the current renewal spread is about +20%.\n■ Renewal spread = [new rent − prior rent] / prior rent = rate of increase.\n\n(7) Demand for data center space remains very strong, further accelerating these companies’ pricing power.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009013668280761555, "ret_signed_m1d": -0.009013668280761555, "alpha_spy_m1d": -0.007140496549997821, "alpha_c_m1d": -0.007140496549997821, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.0018731717307637341, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021458477663362396, "ret_signed_p1d": -0.021458477663362396, "alpha_spy_p1d": -0.009604937720914863, "alpha_c_p1d": -0.009604937720914863, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.011853539942447533, "ret_p1w": -0.01599202987553927, "ret_signed_p1w": -0.01599202987553927, "alpha_spy_p1w": -0.013211600940502, "alpha_c_p1w": -0.013211600940502, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.0027804289350372713, "ret_p1m": -0.08990461407925043, "ret_signed_p1m": -0.08990461407925043, "alpha_spy_p1m": -0.08725587218000175, "alpha_c_p1m": -0.08725587218000175, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.0026487418992486855, "ret_p3m": -0.018490843641420396, "ret_signed_p3m": -0.018490843641420396, "alpha_spy_p3m": -0.037150339682500366, "alpha_c_p3m": -0.037150339682500366, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.01865949604107997, "ret_p6m": 0.14836603173010787, "ret_signed_p6m": 0.14836603173010787, "alpha_spy_p6m": 0.10094845851737078, "alpha_c_p6m": 0.10094845851737078, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.047417573212737096, "price_path": [-0.0213, -0.0112, -0.0189, -0.0257, -0.0226, -0.0265, -0.0402, -0.0363, -0.0471, -0.0382, -0.0453, -0.0543, -0.0353, -0.0423, -0.0304, -0.0199, -0.0266, -0.032, -0.032, -0.0591, -0.0674, -0.0597, -0.0574, -0.0691, -0.0531, 0.0041, 0.0086, 0.0019, 0.0179, 0.011, -0.0047, -0.0035, 0.0007, 0.0159, 0.0126, 0.0002, -0.0035, -0.0029, -0.0108, 0.0054, -0.003, 0.0104, 0.0255, 0.0291, 0.0185, 0.013, 0.0189, -0.0202, -0.0023, -0.0005, 0.0188, 0.0116, 0.0045, 0.0056, -0.0019, 0.0026, 0.0199, 0.0426, 0.0359, 0.0122, -0.028, -0.007, -0.009, 0.0, -0.0215, -0.0316, -0.021, -0.012, -0.016, -0.0197, -0.0469, -0.0805, -0.0801, -0.0829, -0.0718, -0.0736, -0.0836, -0.086, -0.0753, -0.0848, -0.0759, -0.0759, -0.0689, -0.0905, -0.0899, -0.0714, -0.0636, -0.042, -0.0476, -0.0535, -0.0724, -0.0764, -0.1066, -0.1038, -0.1191, -0.1302, -0.1329, -0.1208, -0.1001, -0.098, -0.0903, -0.0903, -0.0914, -0.0876, -0.0833, -0.0931, -0.0931, -0.0912, -0.0869, -0.081, -0.0922, -0.1035, -0.0706, -0.0746, -0.0579, -0.0664, -0.0588, -0.041, -0.041, -0.0643, -0.068, -0.0697, -0.067, -0.0529, -0.0436, -0.0433, -0.0185, -0.0272, -0.0321, -0.036, -0.0223, -0.0348, 0.006, 0.009, 0.0113, 0.0233, 0.057, 0.0608, 0.0608, 0.0545, 0.0318, 0.037, 0.0298, 0.0322, 0.0448, 0.0527, 0.0339, 0.0387, 0.0475, 0.0442, 0.0705, 0.0537, 0.0328, 0.0531, 0.0602, 0.0586, 0.0575, 0.06, 0.0697, 0.0646, 0.0561, 0.0614, 0.0228, 0.0415, 0.0311, 0.0412, 0.0313, 0.0353, 0.0339, 0.0636, 0.065, 0.0723, 0.0723, 0.0741, 0.0766, 0.0955, 0.1104, 0.1147, 0.1323, 0.1555, 0.1549, 0.1747, 0.2017, 0.2034, 0.1878, 0.1854, 0.1804, 0.1804, 0.1588, 0.1484]}
{"unit_id": "orig:1985479798395977784", "ticker_idx": 1, "ts": "2025-11-03T22:50:47+00:00", "symbol": "EQIX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "positive signals are strengthening for the data center value chain, including companies such as Digital Realty and Equinix. These companies are expected to benefit over the medium term due to the tight supply in the data center market", "tweet_type": "news_reaction", "timeline": "months", "summary": "Sharing Goldman read-through: raised hyperscaler CAPEX forecasts drive medium-term bullish view on DLR and EQIX on tight supply and accelerating pricing power.", "resolved_tickers": ["EQIX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "REIT - Specialty", "unit_kind": "original", "lang": "en", "tweet_text": "251104_Americas Data Centers: Read-through to our coverage from 3Q25 hyperscaler earnings - Goldman Sachs\n\n(1) Following 3Q25 results, we have raised our 2025(F)–2026(F) CAPEX forecasts for the major hyperscalers.\n\n(2) CAPEX estimates\n■ 2025(F): $377.9bn → $403.9bn [+7%]\n■ 2026(F): $459.7bn → $551.7bn [+20%]\n\n(3) This is because GPU-centric infrastructure investment is accelerating amid a surge in AI and cloud demand.\n\n(4) Meanwhile, positive signals are strengthening for the data center value chain, including companies such as Digital Realty and Equinix.\n\n(5) These companies are expected to benefit over the medium term due to the tight supply in the data center market.\n\n(6) In particular, Digital Realty noted that when renewing contracts of 1MW or larger, the current renewal spread is about +20%.\n■ Renewal spread = [new rent − prior rent] / prior rent = rate of increase.\n\n(7) Demand for data center space remains very strong, further accelerating these companies’ pricing power.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015813363188103224, "ret_signed_m1d": 0.015813363188103224, "alpha_spy_m1d": 0.01768653491886696, "alpha_c_m1d": 0.01768653491886696, "bench_spy_m1d": -0.0018731717307637341, "bench_c_m1d": -0.0018731717307637341, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0035181615358002016, "ret_signed_p1d": -0.0035181615358002016, "alpha_spy_p1d": 0.008335378406647331, "alpha_c_p1d": 0.008335378406647331, "bench_spy_p1d": -0.011853539942447533, "bench_c_p1d": -0.011853539942447533, "ret_p1w": -0.005655281704725623, "ret_signed_p1w": -0.005655281704725623, "alpha_spy_p1w": -0.002874852769688352, "alpha_c_p1w": -0.002874852769688352, "bench_spy_p1w": -0.0027804289350372713, "bench_c_p1w": -0.0027804289350372713, "ret_p1m": -0.12133809483935432, "ret_signed_p1m": -0.12133809483935432, "alpha_spy_p1m": -0.11868935294010563, "alpha_c_p1m": -0.11868935294010563, "bench_spy_p1m": -0.0026487418992486855, "bench_c_p1m": -0.0026487418992486855, "ret_p3m": -0.0021465068995449865, "ret_signed_p3m": -0.0021465068995449865, "alpha_spy_p3m": -0.020806002940624957, "alpha_c_p3m": -0.020806002940624957, "bench_spy_p3m": 0.01865949604107997, "bench_c_p3m": 0.01865949604107997, "ret_p6m": 0.30735866042116333, "ret_signed_p6m": 0.30735866042116333, "alpha_spy_p6m": 0.25994108720842624, "alpha_c_p6m": 0.25994108720842624, "bench_spy_p6m": 0.047417573212737096, "bench_c_p6m": 0.047417573212737096, "price_path": [-0.0775, -0.0699, -0.0737, -0.073, -0.0604, -0.0644, -0.0747, -0.0676, -0.0815, -0.0769, -0.0679, -0.0728, -0.0557, -0.0606, -0.0613, -0.0508, -0.0558, -0.056, -0.056, -0.0736, -0.0848, -0.0779, -0.071, -0.0808, -0.064, -0.0627, -0.0425, -0.0531, -0.0494, -0.0643, -0.0649, -0.0571, -0.0499, -0.0417, -0.0355, -0.0517, -0.06, -0.051, -0.0614, -0.0596, -0.0707, -0.0723, -0.065, -0.0516, -0.0549, -0.0351, -0.0294, -0.0387, -0.0319, -0.0185, -0.0127, -0.0241, -0.0227, -0.0092, -0.013, -0.0046, -0.0011, 0.008, 0.0119, -0.0201, -0.0422, 0.0004, 0.0158, 0.0, -0.0035, -0.0129, -0.0162, -0.0097, -0.0057, -0.0142, -0.0235, -0.0556, -0.0568, -0.0672, -0.0644, -0.0758, -0.0906, -0.0884, -0.0812, -0.0962, -0.0935, -0.0935, -0.09, -0.1142, -0.1213, -0.1149, -0.1229, -0.1042, -0.1062, -0.1053, -0.1129, -0.089, -0.0936, -0.0803, -0.0842, -0.0966, -0.1012, -0.0837, -0.0844, -0.0818, -0.0835, -0.0835, -0.0779, -0.0759, -0.0702, -0.0745, -0.0745, -0.077, -0.0674, -0.048, -0.0619, -0.0555, -0.0332, -0.0531, -0.032, -0.0348, -0.0314, -0.0315, -0.0315, -0.0527, -0.0391, -0.0424, -0.0442, -0.0259, -0.0221, -0.0141, -0.0021, -0.0083, -0.0211, -0.0314, -0.031, -0.0245, 0.0245, 0.0361, 0.0349, 0.0479, 0.1571, 0.1551, 0.1551, 0.15, 0.1165, 0.109, 0.1211, 0.1423, 0.1499, 0.1666, 0.1514, 0.1833, 0.1734, 0.1806, 0.177, 0.1575, 0.1383, 0.1477, 0.1612, 0.1823, 0.1799, 0.178, 0.1957, 0.1865, 0.1825, 0.1839, 0.165, 0.1744, 0.1715, 0.1732, 0.1701, 0.1696, 0.1709, 0.1906, 0.2097, 0.215, 0.215, 0.2341, 0.2234, 0.236, 0.2529, 0.2513, 0.2836, 0.2842, 0.2789, 0.3007, 0.3222, 0.3408, 0.3291, 0.3388, 0.3546, 0.3467, 0.3237, 0.3074]}
{"unit_id": "thread:1985997719597703622", "ticker_idx": 0, "ts": "2025-11-05T09:08:49+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix successfully secured a steep price increase, reaffirming its dominant leadership in the HBM market and setting the stage for strong earnings growth", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "SK Hynix locked in HBM4 supply to NVIDIA at ~$560/device (+50% vs HBM3E), above $500 market expectations, setting stage for record earnings next year.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SK hynix Raises HBM4 Supply Price by 50%\nAgrees with NVIDIA on ~$560 Per Device\nConfirms Dominant Position in 6th-Gen HBM\n\nSK hynix has reportedly raised the price of its 6th-generation High Bandwidth Memory (HBM), HBM4, by more than 50% in its supply contract with NVIDIA, the world’s largest AI semiconductor company. Despite a rival’s market entry, SK hynix successfully secured a steep price increase, reaffirming its dominant leadership in the HBM market and setting the stage for strong earnings growth.\n\nAccording to SK hynix on the 5th, the company has completed negotiations with NVIDIA regarding both pricing and volumes for HBM4 supply next year. A source familiar with SK hynix stated, “HBM4 will be supplied at a price more than 50% higher than HBM3E.”\n\nThe supply price agreed upon with NVIDIA for HBM4 is confirmed to be around $560 per device (approximately 800,000 KRW). The market had expected SK hynix to supply HBM4 at about $500, but the actual contract price exceeds those expectations by more than 10%. Compared to the currently-supplied HBM3E (approximately $370), the HBM4 price is over 50% higher. It is also known that SK hynix recently shared guidance on HBM4 pricing and profitability during investor relations (IR) meetings with institutional investors.\n\nWith SK hynix concluding HBM4 price and supply negotiations with NVIDIA, expectations are building for record-breaking earnings next year. The global expansion of AI infrastructure is driving price increases not only for HBM but also for general-purpose DRAM. SK hynix recorded 11.3834 trillion KRW in operating profit in the third quarter of this year, and market projections suggest that quarterly operating profit may exceed 15 trillion KRW next year.\n\n---\n\nOriginal article:\nhttps://t.co/PqtsY7MVie", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.012089807587023937, "ret_signed_m1d": 0.012089807587023937, "alpha_spy_m1d": 0.015543229333606856, "alpha_c_m1d": 0.03083804864350126, "bench_spy_m1d": -0.0034534217465829187, "bench_c_m1d": -0.01874824105647732, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02417972341351793, "ret_signed_p1d": 0.02417972341351793, "alpha_spy_p1d": 0.03490905662618393, "alpha_c_p1d": 0.0475733210081134, "bench_spy_p1d": -0.010729333212666003, "bench_c_p1d": -0.023393597594595472, "ret_p1w": 0.06563046135775163, "ret_signed_p1w": 0.06563046135775163, "alpha_spy_p1w": 0.057070602817501426, "alpha_c_p1w": 0.07667362503348696, "bench_spy_p1w": 0.00855985854025021, "bench_c_p1w": -0.011043163675735324, "ret_p1m": -0.06323274062121231, "ret_signed_p1m": -0.06323274062121231, "alpha_spy_p1m": -0.07328318444039716, "alpha_c_p1m": -0.06907415746645218, "bench_spy_p1m": 0.010050443819184851, "bench_c_p1m": 0.005841416845239866, "ret_p3m": 0.4345327637628109, "ret_signed_p3m": 0.4345327637628109, "alpha_spy_p3m": 0.40518588090601826, "alpha_c_p3m": 0.2961755116360061, "bench_spy_p3m": 0.029346882856792655, "bench_c_p3m": 0.13835725212680483, "ret_p6m": 1.2267629882493067, "ret_signed_p6m": 1.2267629882493067, "alpha_spy_p6m": 1.1600963034644283, "alpha_c_p6m": 0.8128639602292567, "bench_spy_p6m": 0.06666668478487847, "bench_c_p6m": 0.4138990280200501, "price_path": [-0.5576, -0.5395, -0.5361, -0.5206, -0.5231, -0.5231, -0.5387, -0.5464, -0.5594, -0.5775, -0.5671, -0.5525, -0.549, -0.5516, -0.5363, -0.5354, -0.5579, -0.5501, -0.5466, -0.5415, -0.5276, -0.5216, -0.5026, -0.475, -0.4698, -0.4326, -0.4283, -0.399, -0.424, -0.386, -0.386, -0.3938, -0.3765, -0.3826, -0.3843, -0.4188, -0.3972, -0.3998, -0.3782, -0.3169, -0.3169, -0.3169, -0.3169, -0.3169, -0.3169, -0.2608, -0.2832, -0.2893, -0.2703, -0.2185, -0.196, -0.1615, -0.1727, -0.1684, -0.1736, -0.1192, -0.076, -0.1002, -0.0363, -0.019, -0.0345, 0.0708, 0.0121, 0.0, 0.0242, 0.0017, 0.0466, 0.0691, 0.0656, 0.057, -0.0328, 0.0466, -0.0155, -0.0294, -0.0138, -0.1002, -0.1019, -0.1036, -0.095, -0.0598, -0.084, -0.0701, -0.0356, -0.0459, -0.0632, -0.0598, -0.0027, -0.0218, 0.0145, -0.0235, -0.0131, -0.0425, -0.084, -0.0477, -0.0459, -0.0546, 0.0024, 0.0094, 0.0163, 0.0163, 0.0353, 0.1061, 0.1252, 0.1252, 0.1252, 0.1701, 0.2029, 0.2548, 0.2824, 0.3066, 0.2859, 0.2945, 0.2755, 0.2824, 0.2945, 0.3066, 0.3205, 0.2842, 0.279, 0.3049, 0.3256, 0.2721, 0.3827, 0.4535, 0.4881, 0.5711, 0.4345, 0.5676, 0.5555, 0.4553, 0.4501, 0.533, 0.514, 0.4864, 0.5348, 0.521, 0.521, 0.521, 0.521, 0.5451, 0.6402, 0.6437, 0.737, 0.7595, 0.903, 0.8372, 0.8372, 0.6259, 0.4701, 0.6294, 0.5999, 0.4476, 0.6242, 0.6536, 0.6103, 0.5757, 0.6865, 0.6796, 0.8285, 0.7541, 0.7437, 0.6155, 0.7073, 0.7229, 0.6155, 0.6155, 0.5116, 0.3974, 0.5532, 0.4372, 0.5168, 0.5341, 0.5861, 0.7887, 0.7281, 0.7783, 0.8008, 0.9099, 0.967, 0.9999, 0.9532, 1.019, 1.1194, 1.1177, 1.1211, 1.1159, 1.2372, 1.251, 1.2389, 1.2268]}
{"unit_id": "orig:1984241741709574405", "ticker_idx": 0, "ts": "2025-10-31T12:51:11+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I hold both Samsung and SK hynix", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Holds both Samsung and SK Hynix long, with a larger position in SK Hynix.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "@momqoqyoy I hold both Samsung and SK hynix, but I hold more of SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031627900169409084, "ret_signed_m1d": -0.031627900169409084, "alpha_spy_m1d": -0.028358344618373277, "alpha_c_m1d": -0.030663475197112433, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": -0.0009644249722966514, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03348832614351105, "ret_signed_p1d": 0.03348832614351105, "alpha_spy_p1d": 0.031611639055546537, "alpha_c_p1d": 0.029397560278537016, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.004090765864974033, "ret_p1w": -0.08930227622821552, "ret_signed_p1w": -0.08930227622821552, "alpha_spy_p1w": -0.07304260458050893, "alpha_c_p1w": -0.04766325858196607, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04163901764624944, "ret_p1m": -0.0623255507623407, "ret_signed_p1m": -0.0623255507623407, "alpha_spy_p1m": -0.05970117670718644, "alpha_c_p1m": -0.014666917306523919, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.04765863345581678, "ret_p3m": 0.5180414411843266, "ret_signed_p3m": 0.5180414411843266, "alpha_spy_p3m": 0.4954410096713908, "alpha_c_p3m": 0.5239191818977161, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": -0.005877740713389534, "ret_p6m": 1.102867513345993, "ret_signed_p6m": 1.102867513345993, "alpha_spy_p6m": 1.0483530517949045, "alpha_c_p6m": 1.0318717496508736, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.07099576369511951, "price_path": [-0.3527, -0.3628, -0.3471, -0.3351, -0.3425, -0.3415, -0.3341, -0.3369, -0.3369, -0.3517, -0.3517, -0.3471, -0.3462, -0.3388, -0.3378, -0.349, -0.3462, -0.3554, -0.3545, -0.374, -0.3601, -0.3536, -0.3508, -0.3564, -0.3508, -0.3378, -0.3277, -0.3202, -0.3017, -0.2915, -0.2647, -0.2758, -0.2563, -0.2563, -0.2267, -0.2156, -0.2091, -0.2026, -0.2286, -0.2167, -0.2195, -0.2, -0.1651, -0.1651, -0.1651, -0.1651, -0.1651, -0.1651, -0.1219, -0.1321, -0.1479, -0.1163, -0.0912, -0.0893, -0.0874, -0.093, -0.0828, -0.1023, -0.0809, -0.0512, -0.0744, -0.0651, -0.0316, 0.0, 0.0335, -0.0242, -0.0642, -0.0772, -0.0893, -0.0642, -0.0372, -0.0409, -0.0437, -0.0958, -0.0642, -0.0902, -0.1023, -0.0642, -0.1181, -0.1005, -0.0763, -0.0437, -0.0372, -0.0651, -0.0623, -0.0381, -0.0279, -0.0223, 0.0084, 0.0186, 0.0084, 0.0047, -0.0019, 0.013, -0.0251, -0.0437, 0.0037, 0.0009, -0.0112, 0.0279, 0.0372, 0.0335, 0.0335, 0.0884, 0.117, 0.1208, 0.1208, 0.1208, 0.2012, 0.2909, 0.2984, 0.318, 0.2974, 0.2993, 0.2974, 0.2862, 0.3115, 0.3451, 0.3918, 0.3956, 0.3573, 0.3975, 0.4236, 0.4218, 0.4218, 0.4909, 0.518, 0.5022, 0.5003, 0.4059, 0.5657, 0.5807, 0.4891, 0.4825, 0.5554, 0.5498, 0.5685, 0.6695, 0.6938, 0.6938, 0.6938, 0.6938, 0.776, 0.777, 0.8041, 0.8695, 0.9022, 1.0378, 1.0237, 1.0237, 0.8237, 0.6096, 0.791, 0.7592, 0.6218, 0.7564, 0.776, 0.7564, 0.7153, 0.7639, 0.8125, 0.949, 0.8742, 0.8639, 0.7414, 0.7732, 0.7667, 0.6835, 0.6835, 0.6514, 0.5661, 0.7764, 0.6711, 0.7441, 0.8087, 0.8406, 0.9717, 0.9108, 0.9296, 0.8827, 0.9343, 0.9764, 1.0373, 1.0232, 1.0092, 1.0513, 1.0373, 1.1029, 1.056, 1.1029]}
{"unit_id": "orig:1984241741709574405", "ticker_idx": 1, "ts": "2025-10-31T12:51:11+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I hold more of SK hynix", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Holds both Samsung and SK Hynix long, with a larger position in SK Hynix.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "@momqoqyoy I hold both Samsung and SK hynix, but I hold more of SK hynix.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01610007841118355, "ret_signed_m1d": 0.01610007841118355, "alpha_spy_m1d": 0.019369633962219357, "alpha_c_m1d": 0.014144188394368706, "bench_spy_m1d": -0.0032695555510358076, "bench_c_m1d": 0.001955890016814843, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10912350108961721, "ret_signed_p1d": 0.10912350108961721, "alpha_spy_p1d": 0.1072468140016527, "alpha_c_p1d": 0.1006390670867805, "bench_spy_p1d": 0.0018766870879645126, "bench_c_p1d": 0.008484434002836716, "ret_p1w": 0.03756707385024027, "ret_signed_p1w": 0.03756707385024027, "alpha_spy_p1w": 0.05382674549794686, "alpha_c_p1w": 0.07861158691845949, "bench_spy_p1w": -0.016259671647706586, "bench_c_p1w": -0.04104451306821921, "ret_p1m": -0.03687792093991171, "ret_signed_p1m": -0.03687792093991171, "alpha_spy_p1m": -0.03425354688475746, "alpha_c_p1m": -0.00916599350196401, "bench_spy_p1m": -0.002624374055154255, "bench_c_p1m": -0.027711927437947703, "ret_p3m": 0.5055496597062261, "ret_signed_p3m": 0.5055496597062261, "alpha_spy_p3m": 0.48294922819329034, "alpha_c_p3m": 0.35430068639247914, "bench_spy_p3m": 0.02260043151293578, "bench_c_p3m": 0.15124897331374698, "ret_p6m": 1.31719330565361, "ret_signed_p6m": 1.31719330565361, "alpha_spy_p6m": 1.2626788441025214, "alpha_c_p6m": 0.9182750577074041, "bench_spy_p6m": 0.05451446155108863, "bench_c_p6m": 0.3989182479462059, "price_path": [-0.5293, -0.5382, -0.532, -0.5418, -0.5231, -0.5195, -0.5034, -0.5061, -0.5061, -0.5222, -0.5302, -0.5436, -0.5623, -0.5516, -0.5364, -0.5329, -0.5356, -0.5197, -0.5188, -0.542, -0.534, -0.5304, -0.525, -0.5107, -0.5045, -0.4848, -0.4562, -0.4508, -0.4123, -0.4079, -0.3775, -0.4034, -0.364, -0.364, -0.3721, -0.3542, -0.3605, -0.3623, -0.398, -0.3757, -0.3784, -0.356, -0.2925, -0.2925, -0.2925, -0.2925, -0.2925, -0.2925, -0.2343, -0.2576, -0.2639, -0.2442, -0.1905, -0.1673, -0.1315, -0.1431, -0.1386, -0.144, -0.0877, -0.0429, -0.068, -0.0018, 0.0161, 0.0, 0.1091, 0.0483, 0.0358, 0.0608, 0.0376, 0.0841, 0.1073, 0.1038, 0.0948, 0.0018, 0.0841, 0.0197, 0.0054, 0.0215, -0.068, -0.0698, -0.0716, -0.0626, -0.0261, -0.0512, -0.0369, -0.0011, -0.0118, -0.0297, -0.0261, 0.0329, 0.0132, 0.0508, 0.0115, 0.0222, -0.0082, -0.0512, -0.0136, -0.0118, -0.0208, 0.0383, 0.0455, 0.0526, 0.0526, 0.0723, 0.1457, 0.1654, 0.1654, 0.1654, 0.212, 0.246, 0.2997, 0.3283, 0.3534, 0.3319, 0.3409, 0.3212, 0.3283, 0.3409, 0.3534, 0.3677, 0.3301, 0.3247, 0.3516, 0.3731, 0.3176, 0.4322, 0.5055, 0.5414, 0.6273, 0.4859, 0.6237, 0.6112, 0.5073, 0.502, 0.5879, 0.5682, 0.5396, 0.5897, 0.5754, 0.5754, 0.5754, 0.5754, 0.6004, 0.6989, 0.7025, 0.7991, 0.8224, 0.971, 0.9029, 0.9029, 0.6841, 0.5227, 0.6877, 0.6572, 0.4994, 0.6823, 0.7128, 0.6679, 0.6321, 0.7469, 0.7397, 0.8939, 0.8168, 0.806, 0.6733, 0.7684, 0.7845, 0.6733, 0.6733, 0.5657, 0.4473, 0.6088, 0.4886, 0.5711, 0.589, 0.6428, 0.8527, 0.7899, 0.8419, 0.8652, 0.9782, 1.0374, 1.0715, 1.0231, 1.0912, 1.1952, 1.1934, 1.197, 1.1916, 1.3172]}
{"unit_id": "quote:1981306608157376790", "ticker_idx": 0, "ts": "2025-10-23T10:28:00+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I agree with Jefferies' note — Micron needs to redesign its HBM4. to achieve commercially viable yield and profitability.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish MU: author agrees Micron must redesign HBM4 to reach viable yield/profitability, implying near-term competitive disadvantage.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"He said he heard from his engineering sources that Micron is currently redesigning its HBM4 chip.\"\n\nI agree with Jefferies’ note — Micron needs to redesign its HBM4.\n\nto achieve commercially viable yield and profitability.", "reply_text": null, "parent_text": null, "quoted_text": "This Jefferies note on $MU reads really intg imo. Would love to hear what @Jukanlosreve or other specialist in the space have to say: \n\n\"Notes from Call with Jeff Kim on Samsung TOP PICK. He thinks $MU Micron needs to do a redesign to be able to deliver HBM4 in scale and may be https://t.co/Tn4e80ueJ0", "parent_author": null, "ret_m1d": -0.039862616071910106, "ret_signed_m1d": 0.039862616071910106, "alpha_spy_m1d": 0.03396768120375371, "alpha_c_m1d": 0.02188121283332023, "bench_spy_m1d": -0.0058949348681563984, "bench_c_m1d": -0.017981403238589877, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05955204683287785, "ret_signed_p1d": -0.05955204683287785, "alpha_spy_p1d": -0.05137947572824708, "alpha_c_p1d": -0.04116445925798118, "bench_spy_p1d": 0.008172571104630766, "bench_c_p1d": 0.01838758757489667, "ret_p1w": 0.08369212288311201, "ret_signed_p1w": -0.08369212288311201, "alpha_spy_p1w": -0.07167893039297013, "alpha_c_p1w": -0.028790649885580777, "bench_spy_p1w": 0.012013192490141877, "bench_c_p1w": 0.05490147299753123, "ret_p1m": 0.0031928686836897313, "ret_signed_p1m": -0.0031928686836897313, "alpha_spy_p1m": -0.022143047058627463, "alpha_c_p1m": -0.05734012617588191, "bench_spy_p1m": -0.01895017837493773, "bench_c_p1m": -0.05414725749219218, "ret_p3m": 0.7664721453273138, "ret_signed_p3m": -0.7664721453273138, "alpha_spy_p3m": -0.7548278374350474, "alpha_c_p3m": -0.6307274709366635, "bench_spy_p3m": 0.011644307892266337, "bench_c_p3m": 0.1357446743906503, "ret_p6m": 1.2033045725443436, "ret_signed_p6m": -1.2033045725443436, "alpha_spy_p6m": -1.140151748690257, "alpha_c_p6m": -0.8529440806164916, "bench_spy_p6m": 0.0631528238540866, "bench_c_p6m": 0.350360491927852, "price_path": [-0.4621, -0.4587, -0.4453, -0.4723, -0.4929, -0.479, -0.4727, -0.4741, -0.4591, -0.4252, -0.4019, -0.3824, -0.3992, -0.3943, -0.4156, -0.4027, -0.4099, -0.4333, -0.4402, -0.4311, -0.4371, -0.4368, -0.4307, -0.4102, -0.4246, -0.4246, -0.4272, -0.426, -0.3995, -0.3649, -0.3644, -0.3462, -0.3231, -0.272, -0.2398, -0.2372, -0.2322, -0.2265, -0.1835, -0.2133, -0.2041, -0.1955, -0.2182, -0.2418, -0.2397, -0.2076, -0.1911, -0.1194, -0.1116, -0.0913, -0.0762, -0.1017, -0.0492, -0.0696, -0.1215, -0.0674, -0.0951, -0.0715, -0.0202, -0.0209, 0.0003, -0.0214, -0.0399, 0.0, 0.0596, 0.0648, 0.0735, 0.0964, 0.0837, 0.0825, 0.1354, 0.0548, 0.149, 0.153, 0.151, 0.2254, 0.1664, 0.1848, 0.1463, 0.1941, 0.1705, 0.1054, 0.0929, -0.0258, 0.0032, 0.0833, 0.0862, 0.1139, 0.1139, 0.144, 0.1633, 0.1586, 0.1328, 0.0965, 0.1476, 0.1945, 0.2211, 0.2757, 0.2504, 0.1666, 0.149, 0.1248, 0.091, 0.2024, 0.2864, 0.3381, 0.3365, 0.3869, 0.3869, 0.3777, 0.4246, 0.4162, 0.3813, 0.3813, 0.5265, 0.5107, 0.6621, 0.6433, 0.5827, 0.6701, 0.6739, 0.6364, 0.6133, 0.6292, 0.7556, 0.7556, 0.7665, 0.8832, 0.9241, 0.9342, 0.8831, 0.9854, 1.1066, 1.1091, 1.0079, 1.1188, 1.0299, 0.8362, 0.8531, 0.9102, 0.856, 0.8064, 0.9859, 1.0035, 0.9923, 0.9923, 0.9348, 1.0373, 1.0198, 1.0722, 1.0373, 1.023, 1.0762, 1.0112, 0.9957, 0.9972, 0.8375, 0.9396, 0.9216, 0.7921, 0.8842, 0.9509, 1.0263, 0.9618, 1.0623, 1.1382, 1.2344, 1.2346, 1.1501, 1.0467, 0.9569, 0.9142, 0.8492, 0.7203, 0.7288, 0.5581, 0.6357, 0.781, 0.7732, 0.7732, 0.829, 0.8281, 0.9693, 1.0408, 1.0364, 1.0653, 1.2546, 1.2089, 1.2138, 1.2033]}
{"unit_id": "orig:2000021897355149617", "ticker_idx": 0, "ts": "2025-12-14T01:55:53+00:00", "symbol": "commodity DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "there's a real possibility that the shortage of commodity DRAM could worsen as CXMT reallocates capacity toward HBM", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Commodity DRAM shortage likely to worsen as CXMT shifts capacity to HBM, bullish for DRAM pricing/supply tightness.", "resolved_tickers": ["MU", "2408.TW", "2344.TW"], "resolution_method": "basket", "resolution_reason": "Commodity DRAM basket: Micron, Nanya, Winbond", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I think there’s a real possibility that the shortage of commodity DRAM could worsen as CXMT reallocates capacity toward HBM. https://t.co/lwfVqCdP2j", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020491907515382397, "ret_signed_m1d": 0.020491907515382397, "alpha_spy_m1d": 0.018978823668354978, "alpha_c_m1d": 0.017034833551626345, "bench_spy_m1d": 0.0015130838470274188, "bench_c_m1d": 0.003457073963756052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01857842057810229, "ret_signed_p1d": -0.01857842057810229, "alpha_spy_p1d": -0.01584612298282358, "alpha_c_p1d": -0.015858077329077364, "bench_spy_p1d": -0.00273229759527871, "bench_c_p1d": -0.0027203432490249257, "ret_p1w": 0.10013702308452059, "ret_signed_p1w": 0.10013702308452059, "alpha_spy_p1w": 0.09114129777181816, "alpha_c_p1w": 0.07769653341232623, "bench_spy_p1w": 0.00899572531270243, "bench_c_p1w": 0.022440489672194364, "ret_p1m": 0.4144405867282879, "ret_signed_p1m": 0.4144405867282879, "alpha_spy_p1m": 0.39227309682257694, "alpha_c_p1m": 0.3015239615188079, "bench_spy_p1m": 0.022167489905710935, "bench_c_p1m": 0.11291662520947998, "ret_p3m": 0.5714414764317799, "ret_signed_p3m": 0.5714414764317799, "alpha_spy_p3m": 0.590100649696472, "alpha_c_p3m": 0.4681892696444021, "bench_spy_p3m": -0.01865917326469213, "bench_c_p3m": 0.10325220678737779, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4874, -0.4476, -0.4522, -0.4428, -0.4295, -0.4523, -0.4663, -0.4866, -0.4773, -0.4559, -0.4297, -0.41, -0.3842, -0.3798, -0.3524, -0.3445, -0.3243, -0.3394, -0.3317, -0.3639, -0.3661, -0.3148, -0.2966, -0.273, -0.2906, -0.2827, -0.2682, -0.2509, -0.2063, -0.1687, -0.1559, -0.1509, -0.16, -0.1242, -0.1744, -0.121, -0.0853, -0.0922, -0.0136, -0.0218, -0.0181, -0.0311, -0.0457, -0.0085, -0.0622, -0.0762, -0.1265, -0.177, -0.1364, -0.1209, -0.1455, -0.1106, -0.0967, -0.0858, -0.0919, -0.1, -0.1012, -0.0635, 0.0003, 0.0209, 0.0066, 0.0034, 0.0205, 0.0, -0.0186, 0.0004, 0.0468, 0.0654, 0.1001, 0.092, 0.1506, 0.1506, 0.146, 0.1605, 0.2023, 0.1839, 0.1839, 0.2932, 0.3127, 0.4385, 0.4704, 0.4632, 0.3892, 0.4541, 0.4144, 0.4211, 0.4249, 0.5214, 0.6196, 0.6002, 0.5558, 0.5948, 0.6009, 0.6627, 0.6857, 0.8158, 0.8306, 0.8547, 0.7616, 0.6528, 0.6465, 0.5823, 0.5542, 0.6055, 0.5652, 0.641, 0.6461, 0.6429, 0.6429, 0.6262, 0.6559, 0.6509, 0.666, 0.723, 0.7702, 0.7463, 0.7434, 0.7389, 0.7477, 0.5867, 0.5313, 0.6108, 0.4972, 0.4509, 0.5345, 0.6434, 0.5714, 0.5975, 0.7053, 0.7842, 0.8146, 0.7291, 0.5895, 0.5292, 0.4659, 0.4615, 0.4117, 0.3873, 0.3426, 0.3068, 0.3883, 0.3584, 0.3584, 0.3745, 0.3743, 0.4793, 0.4415, 0.4725, 0.5066, 0.546, 0.4998, 0.5025, 0.4838, 0.4584, 0.5118, 0.561, 0.5171, 0.536, 0.644, 0.6456, 0.6472, 0.5917, 0.6269, 0.7455, 0.891, 1.0261, 1.0343, 1.1153, 1.2884, 1.3079, 1.3623, 1.421, 1.2672, 1.198, 1.0983, 1.1366, 1.2088, 1.2809, 1.2677, 1.5455, 1.6746, 1.6405, 1.8203, null, null, null, null, null, null, null]}
{"unit_id": "orig:1988563383675932964", "ticker_idx": 0, "ts": "2025-11-12T11:03:51+00:00", "symbol": "Asia tech", "kind": "sector", "geography": "global", "sub_sector": "AI semiconductors", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "J.P. Morgan's latest research argues that...the AI-driven semiconductor upcycle has not yet peaked and will extend to 2027", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM thesis: AI/semi upcycle extends to 2027, Asia tech has further upside, TSMC top pick Overweight.", "resolved_tickers": ["SMH", "EWY", "EWT", "SOXX"], "resolution_method": "basket", "resolution_reason": "AI semiconductor-focused Asia tech basket", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan: The AI- and semiconductor-led upcycle extends to 2027; Asia tech still has upside next year\n\nJ.P. Morgan’s latest research argues that despite persistent market concerns about an AI bubble, the AI-driven semiconductor upcycle has not yet peaked and will extend to 2027, transcending the usual rules of the cycle.\n\nIn a report published on November 11, 2025, the bank raised its outlook, forecasting global semiconductor revenue growth of 18% in 2026 and 11% in 2027. This view is based mainly on the still-steep early adoption curve of generative AI, the strong capital-expenditure commitment of leading cloud service providers (CSPs), and conservative expansion strategies at key points in the supply chain.\n\nJ.P. Morgan said Asia’s tech market will be in a “subtle” equilibrium in 2026. On one hand, investors will continue to fret about a cycle peak; on the other, they will keep seeing sustained upward revisions to EPS powered by AI infrastructure build-out and price increases in some components. The bank believes this powerful earnings-upgrade trend will ultimately drive further gains in Asia tech stocks over the next few quarters.\n\nThis view directly challenges the growing market narrative that an “AI bubble is about to burst,” offering investors a new lens for evaluating the tech sector. J.P. Morgan argues that given the unique character of the current cycle, past rules of thumb cannot be applied wholesale; instead, more attention should be paid to its embedded structural drivers.\n\nCoexistence of divided opinions and earnings upgrades\nIn the report, J.P. Morgan notes that by classical gauges—such as cycle duration and the magnitude of earnings growth—the current upcycle may look late-stage, but that does not mean the trend is about to reverse. In the first half of 2026, Asia tech still has room to rise on the back of strong profit growth at leading AI companies.\n\nOver the past three months, EPS-upgrade momentum for Asia tech has visibly accelerated, driven by memory chips and other general tech products. The report points out that strong AI demand is “crowding out” supply across the industry, creating shortages from advanced packaging (CoWoS), leading-edge foundry and high-bandwidth memory (HBM), to mainstream OSAT, MLCC, and commodity DRAM/NAND. Because capacity additions by suppliers in this cycle are far slower than normal, the resulting price increases should further propel earnings upgrades.\n\nFour reasons supporting an extra-long cycle\nJ.P. Morgan’s analysts believe several unique features of this cycle distinguish it from the broad, GDP-led recoveries of the past and are sufficient to extend the upcycle to 2027.\n\nFirst, this is a “K-shaped recovery.” Since mid-2023, the market has seen a split between AI and non-AI demand. AI-related demand is strong, while other areas of tech are in a corrective phase.\n\nSecond, generative AI adoption is rising along a steep S-curve, with a trajectory more akin to the early phases of smartphones or public cloud. The report notes that it took smartphones roughly seven years and public cloud about nine years before annual growth fell below 20%. By contrast, generative AI is only in year four and is expected to maintain 50–60% year-over-year growth in 2026, underscoring enormous growth potential.\n\nThird, leading CSPs have both the capacity and the will to keep investing. While there are concerns about free cash flow (FCF) at smaller CSPs, J.P. Morgan’s analysis suggests capex at the top six CSPs will surge 67% in 2025 and can grow another ~32% in 2026. Even assuming flat free cash flow in 2027, these companies would still have room to lift capex by a further 13%, indicating balance sheets can support continued expansion of AI infrastructure.\n\nLastly, the supply response in core areas is sluggish. Semiconductor equipment spending has been notably conservative in this cycle, and supply is highly concentrated among a handful of firms such as TSMC and SK hynix. The report expects TSMC’s capex to increase 16% in 2026, while DRAM industry capex rises only 11%. This implies that shortages in key stages—3nm, CoWoS, and HBM—will persist through 2026, prolonging chip tightness.\n\nBased on these judgments, J.P. Morgan proposes a “barbell” portfolio strategy for 2026: on one end, front-line AI enablers; on the other, companies poised to benefit from price increases and margin expansion with clear earnings-upgrade potential. The report maintains an Overweight on TSMC and names it a top pick.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008639477009589225, "ret_signed_m1d": -0.008639477009589225, "alpha_spy_m1d": -0.008083387810402631, "alpha_c_m1d": -0.008083387810402631, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.0005560891991865935, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026364478350001413, "ret_signed_p1d": -0.026364478350001413, "alpha_spy_p1d": -0.009770456879263562, "alpha_c_p1d": -0.009770456879263562, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.016594021470737852, "ret_p1w": -0.051472818122369424, "ret_signed_p1w": -0.051472818122369424, "alpha_spy_p1w": -0.021109144229387533, "alpha_c_p1w": -0.021109144229387533, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.03036367389298189, "ret_p1m": 0.02263654599863449, "ret_signed_p1m": 0.02263654599863449, "alpha_spy_p1m": 0.01416397778631348, "alpha_c_p1m": 0.01416397778631348, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.008472568212321008, "ret_p3m": 0.20034689029711428, "ret_signed_p3m": 0.20034689029711428, "alpha_spy_p3m": 0.18187899822818043, "alpha_c_p3m": 0.18187899822818043, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.018467892068933844, "ret_p6m": 0.6406084891022021, "ret_signed_p6m": 0.6406084891022021, "alpha_spy_p6m": 0.5639810861115184, "alpha_c_p6m": 0.5639810861115184, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.07662740299068371, "price_path": [-0.1632, -0.1609, -0.1755, -0.1832, -0.1866, -0.1665, -0.1704, -0.1664, -0.1641, -0.1587, -0.1786, -0.1786, -0.1871, -0.1829, -0.1758, -0.1661, -0.1579, -0.1525, -0.1386, -0.1311, -0.128, -0.1221, -0.1159, -0.12, -0.0997, -0.1053, -0.0911, -0.0909, -0.0981, -0.1012, -0.1041, -0.0985, -0.0924, -0.0762, -0.0648, -0.0641, -0.0485, -0.0629, -0.0431, -0.0499, -0.0984, -0.0587, -0.0737, -0.0493, -0.0394, -0.039, -0.0228, -0.0317, -0.0432, -0.0288, -0.0123, 0.0086, 0.0119, 0.0266, 0.0132, 0.0186, 0.0331, -0.0051, 0.0118, -0.013, -0.0234, 0.0049, -0.0086, 0.0, -0.0264, -0.0223, -0.0403, -0.0563, -0.0515, -0.0813, -0.076, -0.0514, -0.0503, -0.0305, -0.0305, -0.0213, -0.0228, -0.0076, 0.0045, -0.0042, 0.0099, 0.0185, 0.0202, 0.0344, 0.0226, -0.0146, -0.0165, -0.0222, -0.0471, -0.0285, -0.0092, -0.0002, 0.0078, 0.0147, 0.0147, 0.0211, 0.0282, 0.0285, 0.0187, 0.0187, 0.0567, 0.0725, 0.0977, 0.0904, 0.08, 0.1029, 0.1082, 0.1055, 0.1093, 0.1263, 0.1368, 0.1368, 0.1189, 0.1522, 0.1574, 0.1633, 0.1602, 0.1927, 0.2142, 0.2072, 0.1722, 0.1803, 0.1757, 0.1384, 0.1368, 0.1869, 0.2003, 0.1963, 0.2348, 0.2188, 0.2329, 0.2329, 0.2229, 0.236, 0.2375, 0.2685, 0.2531, 0.2861, 0.3129, 0.2939, 0.2869, 0.2741, 0.1932, 0.2112, 0.1776, 0.1544, 0.1989, 0.1922, 0.2102, 0.1576, 0.1607, 0.2028, 0.2145, 0.1971, 0.2106, 0.1633, 0.2025, 0.1893, 0.1999, 0.1435, 0.1346, 0.1006, 0.1565, 0.1826, 0.171, 0.171, 0.1876, 0.1989, 0.2865, 0.2967, 0.3086, 0.3308, 0.3658, 0.3644, 0.3784, 0.4136, 0.4105, 0.4046, 0.457, 0.4466, 0.5072, 0.5047, 0.4734, 0.4874, 0.5309, 0.5403, 0.5411, 0.6054, 0.6756, 0.6406]}
{"unit_id": "orig:1988563383675932964", "ticker_idx": 1, "ts": "2025-11-12T11:03:51+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The report maintains an Overweight on TSMC and names it a top pick", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares JPM thesis: AI/semi upcycle extends to 2027, Asia tech has further upside, TSMC top pick Overweight.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan: The AI- and semiconductor-led upcycle extends to 2027; Asia tech still has upside next year\n\nJ.P. Morgan’s latest research argues that despite persistent market concerns about an AI bubble, the AI-driven semiconductor upcycle has not yet peaked and will extend to 2027, transcending the usual rules of the cycle.\n\nIn a report published on November 11, 2025, the bank raised its outlook, forecasting global semiconductor revenue growth of 18% in 2026 and 11% in 2027. This view is based mainly on the still-steep early adoption curve of generative AI, the strong capital-expenditure commitment of leading cloud service providers (CSPs), and conservative expansion strategies at key points in the supply chain.\n\nJ.P. Morgan said Asia’s tech market will be in a “subtle” equilibrium in 2026. On one hand, investors will continue to fret about a cycle peak; on the other, they will keep seeing sustained upward revisions to EPS powered by AI infrastructure build-out and price increases in some components. The bank believes this powerful earnings-upgrade trend will ultimately drive further gains in Asia tech stocks over the next few quarters.\n\nThis view directly challenges the growing market narrative that an “AI bubble is about to burst,” offering investors a new lens for evaluating the tech sector. J.P. Morgan argues that given the unique character of the current cycle, past rules of thumb cannot be applied wholesale; instead, more attention should be paid to its embedded structural drivers.\n\nCoexistence of divided opinions and earnings upgrades\nIn the report, J.P. Morgan notes that by classical gauges—such as cycle duration and the magnitude of earnings growth—the current upcycle may look late-stage, but that does not mean the trend is about to reverse. In the first half of 2026, Asia tech still has room to rise on the back of strong profit growth at leading AI companies.\n\nOver the past three months, EPS-upgrade momentum for Asia tech has visibly accelerated, driven by memory chips and other general tech products. The report points out that strong AI demand is “crowding out” supply across the industry, creating shortages from advanced packaging (CoWoS), leading-edge foundry and high-bandwidth memory (HBM), to mainstream OSAT, MLCC, and commodity DRAM/NAND. Because capacity additions by suppliers in this cycle are far slower than normal, the resulting price increases should further propel earnings upgrades.\n\nFour reasons supporting an extra-long cycle\nJ.P. Morgan’s analysts believe several unique features of this cycle distinguish it from the broad, GDP-led recoveries of the past and are sufficient to extend the upcycle to 2027.\n\nFirst, this is a “K-shaped recovery.” Since mid-2023, the market has seen a split between AI and non-AI demand. AI-related demand is strong, while other areas of tech are in a corrective phase.\n\nSecond, generative AI adoption is rising along a steep S-curve, with a trajectory more akin to the early phases of smartphones or public cloud. The report notes that it took smartphones roughly seven years and public cloud about nine years before annual growth fell below 20%. By contrast, generative AI is only in year four and is expected to maintain 50–60% year-over-year growth in 2026, underscoring enormous growth potential.\n\nThird, leading CSPs have both the capacity and the will to keep investing. While there are concerns about free cash flow (FCF) at smaller CSPs, J.P. Morgan’s analysis suggests capex at the top six CSPs will surge 67% in 2025 and can grow another ~32% in 2026. Even assuming flat free cash flow in 2027, these companies would still have room to lift capex by a further 13%, indicating balance sheets can support continued expansion of AI infrastructure.\n\nLastly, the supply response in core areas is sluggish. Semiconductor equipment spending has been notably conservative in this cycle, and supply is highly concentrated among a handful of firms such as TSMC and SK hynix. The report expects TSMC’s capex to increase 16% in 2026, while DRAM industry capex rises only 11%. This implies that shortages in key stages—3nm, CoWoS, and HBM—will persist through 2026, prolonging chip tightness.\n\nBased on these judgments, J.P. Morgan proposes a “barbell” portfolio strategy for 2026: on one end, front-line AI enablers; on the other, companies poised to benefit from price increases and margin expansion with clear earnings-upgrade potential. The report maintains an Overweight on TSMC and names it a top pick.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0067797177782006335, "ret_signed_m1d": -0.0067797177782006335, "alpha_spy_m1d": -0.00622362857901404, "alpha_c_m1d": 0.005821164838953008, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.012600882617153641, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010169451707829458, "ret_signed_p1d": -0.010169451707829458, "alpha_spy_p1d": 0.006424569762908394, "alpha_c_p1d": 0.019954531953429333, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.03012398366125879, "ret_p1w": -0.05423732569403328, "ret_signed_p1w": -0.05423732569403328, "alpha_spy_p1w": -0.023873651801051388, "alpha_c_p1w": -0.008981003274691424, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.045256322419341855, "ret_p1m": -6.781133985545029e-05, "ret_signed_p1m": -6.781133985545029e-05, "alpha_spy_p1m": -0.008540379552176458, "alpha_c_p1m": -0.04327098506258309, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.04320317372272764, "ret_p3m": 0.23461015810538788, "ret_signed_p3m": 0.23461015810538788, "alpha_spy_p3m": 0.21614226603645403, "alpha_c_p3m": 0.08726580627913605, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.14734435182625183, "ret_p6m": 0.576448705136634, "ret_signed_p6m": 0.576448705136634, "alpha_spy_p6m": 0.49982130214595033, "alpha_c_p6m": 0.05258133587321279, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.5238673692634213, "price_path": [-0.2032, -0.2032, -0.1998, -0.2336, -0.2234, -0.2336, -0.2099, -0.2066, -0.1964, -0.2167, -0.2167, -0.2133, -0.2167, -0.2167, -0.2167, -0.2032, -0.2032, -0.1897, -0.1728, -0.1627, -0.1492, -0.1525, -0.1322, -0.1424, -0.1288, -0.1424, -0.122, -0.0915, -0.0915, -0.1051, -0.1186, -0.1186, -0.1153, -0.1017, -0.0746, -0.0508, -0.0508, -0.0271, -0.0407, -0.0237, -0.0237, -0.0407, -0.0339, -0.0068, 0.0068, -0.0169, 0.0034, 0.0034, -0.0102, -0.0169, -0.0169, 0.0034, 0.0, 0.0203, 0.0203, 0.0169, 0.0237, 0.0203, -0.0102, -0.0068, -0.0102, 0.0, -0.0068, 0.0, -0.0102, -0.0305, -0.0203, -0.0475, -0.0542, -0.0136, -0.061, -0.0678, -0.0407, -0.0237, -0.0271, -0.0237, -0.0441, -0.0305, -0.0169, -0.0203, -0.0102, 0.0136, 0.0034, 0.0203, -0.0001, 0.0067, -0.0137, -0.0239, -0.0273, -0.0273, -0.0273, -0.0035, 0.0135, 0.0169, 0.0169, 0.0271, 0.0407, 0.0339, 0.0544, 0.0544, 0.0782, 0.136, 0.1598, 0.1394, 0.1462, 0.1428, 0.1496, 0.1632, 0.1632, 0.1496, 0.1836, 0.1972, 0.2074, 0.1836, 0.1972, 0.204, 0.1938, 0.2108, 0.238, 0.2278, 0.2074, 0.2006, 0.2244, 0.2142, 0.2006, 0.2108, 0.2346, 0.2788, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.2924, 0.3366, 0.3707, 0.3571, 0.3571, 0.3434, 0.3162, 0.2686, 0.2924, 0.2856, 0.2312, 0.2584, 0.3196, 0.2822, 0.2686, 0.255, 0.2762, 0.3001, 0.2625, 0.2557, 0.2352, 0.2352, 0.2591, 0.2557, 0.2421, 0.2148, 0.2011, 0.2659, 0.2352, 0.2352, 0.2352, 0.2693, 0.3308, 0.3342, 0.3649, 0.3581, 0.4024, 0.4195, 0.4229, 0.3854, 0.382, 0.399, 0.399, 0.4195, 0.4911, 0.5457, 0.5116, 0.4877, 0.457, 0.457, 0.5526, 0.5355, 0.5355, 0.5764]}
{"unit_id": "thread:1989155380988113291", "ticker_idx": 0, "ts": "2025-11-14T02:16:14+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The DRAM inventory days of both companies have fallen to the lowest level in the past three years, providing a basis for additional price hikes", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung and SK Hynix: DDR5 supply tightening from HBM capacity shift drives above-expected Q4 price hikes persisting through H1 next year.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Server DDR5 prices, additional year-end hikes becoming a reality … A signal that the memory market is overheating again\n\nServer DDR5 prices are turning upward again ahead of year-end, making signs of overheating in the memory market increasingly clear. As factories and manpower have been heavily shifted toward expanding HBM (high bandwidth memory) production, the production capacity for commodity memory has declined, while DDR5 demand is rapidly increasing on the back of AI server proliferation. Against this backdrop, the view that the magnitude of the fourth-quarter contract price hikes is turning out larger than expected is gaining traction.\n\nAccording to the semiconductor industry on the 14th, major memory suppliers are in talks with customers over plans to raise fourth-quarter DDR5 contract prices once more compared with the previous quarter. The biggest reason is that, with production capacity (CAPA) concentrated on HBM not returning easily, the available volume of server DRAM has decreased.\n\nIn particular, in the second half of this year, the ramp-up preparations for HBM3E and early HBM4 supply overlapped, creating severe bottlenecks at leading-edge process nodes, and DDR5 production has effectively been pushed down the priority list. Within the industry, the consensus is that as production lines are being reconfigured around HBM, supply pressure on commodity memory is intensifying and the extent of DDR5 contract price hikes in the fourth quarter will be visibly greater than in the first half.\n\nThe strategies put in place by the three global cloud giants to lock in next year’s volumes are also cited as another factor fueling price increases. Major cloud service providers (CSPs) such as Amazon, Google, and Microsoft are accelerating early contracts by prioritizing next year’s DDR5 volumes in order to speed up AI infrastructure build-outs, and in the process, DDR5’s premium pricing structure has become entrenched.\n\nIn the DDR4 era, prices were adjusted through quarterly negotiations, but DDR5 is creating a new market order in which demand outstrips supply. The growing prevalence of advance purchases and early-allocation contracts, rarely seen in the traditional commodity memory market, is part of the same trend.\n\nThe inventory days of Samsung Electronics and SK hynix also lend support to the outlook for further price increases. The DRAM inventory days of both companies have fallen to the lowest level in the past three years, providing a basis for additional price hikes. As DDR5 is already in a structurally “tight supply-demand” state in the market, analysis is emerging that its strength is likely to continue into next year without any short-term correction.\n\nThe fact that surging AI server demand is simultaneously tightening both commodity memory and HBM is another key variable. While the share of HBM has increased as AI training GPUs require high-performance memory, the baseline configuration of AI servers has also expanded, pushing DDR5 installed capacity significantly higher compared with the past. AI companies and CSPs are competing to scale up their AI models, and the dominant assessment is that the likelihood of server DRAM demand settling at a lower level next year is not high.\n\nManufacturers project that the upward pressure on prices will persist at least through the first half of next year. In the initial phase of HBM4 mass production, it will take time to stabilize processes and secure yields, and during this period, production capacity is unlikely to stretch far enough to cover DDR5. Within the semiconductor industry, the prevailing view is that as HBM4 preparations move into full swing, tight conditions across overall DRAM will be maintained at least through the first half, and in the near term, DDR5 prices are more likely to see further increases than any meaningful correction.\n\nSemiconductor industry experts and market analysts judge that the current DDR5 price hikes should be seen not as a mere seasonal factor but as a “medium- to long-term inflection point” created by changes in AI infrastructure and process transitions. With yet another round of price increases signaled for year-end and upward pressure likely to persist across memory into the first half of next year, how cloud and server companies respond is expected to become another key point of interest for the industry.\n\n---\n\nhttps://t.co/Saw141jWTd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05761317038957614, "ret_signed_m1d": 0.05761317038957614, "alpha_spy_m1d": 0.057449557525201866, "alpha_c_m1d": 0.06302709370298476, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.0054139233134086195, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03497941620794953, "ret_signed_p1d": 0.03497941620794953, "alpha_spy_p1d": 0.044295940138911694, "alpha_c_p1d": 0.05063102623894755, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.01565161003099802, "ret_p1w": -0.024691404986047938, "ret_signed_p1w": -0.024691404986047938, "alpha_spy_p1w": -0.005492974646394533, "alpha_c_p1w": 0.027191265659056163, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.0518826706451041, "ret_p1m": 0.07818927376678131, "ret_signed_p1m": 0.07818927376678131, "alpha_spy_p1m": 0.0650927536872532, "alpha_c_p1m": 0.09050924685115647, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": -0.012319973084375158, "ret_p3m": 0.7347294453035076, "ret_signed_p3m": 0.7347294453035076, "alpha_spy_p3m": 0.7018767860417772, "alpha_c_p3m": 0.7408938922948921, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": -0.00616444699138452, "ret_p6m": 1.957630231150651, "ret_signed_p6m": 1.957630231150651, "alpha_spy_p6m": 1.8511019956445949, "alpha_c_p6m": 1.7195854032114168, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.2380448279392342, "price_path": [-0.283, -0.2779, -0.2769, -0.2687, -0.2677, -0.28, -0.2769, -0.2871, -0.2861, -0.3076, -0.2923, -0.2851, -0.282, -0.2882, -0.282, -0.2677, -0.2564, -0.2482, -0.2277, -0.2165, -0.1868, -0.199, -0.1775, -0.1775, -0.1448, -0.1325, -0.1253, -0.1181, -0.1468, -0.1337, -0.1368, -0.1152, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0766, -0.0288, -0.0401, -0.0576, -0.0226, 0.0051, 0.0072, 0.0093, 0.0031, 0.0144, -0.0072, 0.0165, 0.0494, 0.0237, 0.034, 0.071, 0.106, 0.143, 0.0792, 0.035, 0.0206, 0.0072, 0.035, 0.0648, 0.0607, 0.0576, 0.0, 0.035, 0.0062, -0.0072, 0.035, -0.0247, -0.0051, 0.0216, 0.0576, 0.0648, 0.034, 0.037, 0.0638, 0.0751, 0.0813, 0.1152, 0.1265, 0.1152, 0.1111, 0.1039, 0.1204, 0.0782, 0.0576, 0.1101, 0.107, 0.0936, 0.1368, 0.1471, 0.143, 0.143, 0.2037, 0.2354, 0.2395, 0.2395, 0.2395, 0.3284, 0.4277, 0.436, 0.4577, 0.4349, 0.437, 0.4349, 0.4225, 0.4504, 0.4876, 0.5393, 0.5435, 0.5011, 0.5455, 0.5745, 0.5724, 0.5724, 0.6489, 0.6789, 0.6613, 0.6593, 0.5548, 0.7316, 0.7482, 0.6469, 0.6396, 0.7203, 0.7141, 0.7347, 0.8464, 0.8733, 0.8733, 0.8733, 0.8733, 0.9642, 0.9653, 0.9952, 1.0676, 1.1038, 1.2537, 1.2382, 1.2382, 1.017, 0.7802, 0.9808, 0.9456, 0.7937, 0.9425, 0.9642, 0.9425, 0.897, 0.9508, 1.0046, 1.1555, 1.0728, 1.0614, 0.926, 0.9611, 0.9539, 0.8619, 0.8619, 0.8264, 0.7321, 0.9647, 0.8481, 0.9289, 1.0004, 1.0356, 1.1807, 1.1133, 1.1341, 1.0823, 1.1392, 1.1858, 1.2532, 1.2376, 1.2221, 1.2687, 1.2532, 1.3257, 1.2739, 1.3257, 1.2998, 1.3412, 1.2843, 1.2843, 1.4086, 1.4086, 1.7556, 1.8126, 1.7815, 1.9576]}
{"unit_id": "thread:1989155380988113291", "ticker_idx": 1, "ts": "2025-11-14T02:16:14+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The DRAM inventory days of both companies have fallen to the lowest level in the past three years, providing a basis for additional price hikes", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung and SK Hynix: DDR5 supply tightening from HBM capacity shift drives above-expected Q4 price hikes persisting through H1 next year.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Server DDR5 prices, additional year-end hikes becoming a reality … A signal that the memory market is overheating again\n\nServer DDR5 prices are turning upward again ahead of year-end, making signs of overheating in the memory market increasingly clear. As factories and manpower have been heavily shifted toward expanding HBM (high bandwidth memory) production, the production capacity for commodity memory has declined, while DDR5 demand is rapidly increasing on the back of AI server proliferation. Against this backdrop, the view that the magnitude of the fourth-quarter contract price hikes is turning out larger than expected is gaining traction.\n\nAccording to the semiconductor industry on the 14th, major memory suppliers are in talks with customers over plans to raise fourth-quarter DDR5 contract prices once more compared with the previous quarter. The biggest reason is that, with production capacity (CAPA) concentrated on HBM not returning easily, the available volume of server DRAM has decreased.\n\nIn particular, in the second half of this year, the ramp-up preparations for HBM3E and early HBM4 supply overlapped, creating severe bottlenecks at leading-edge process nodes, and DDR5 production has effectively been pushed down the priority list. Within the industry, the consensus is that as production lines are being reconfigured around HBM, supply pressure on commodity memory is intensifying and the extent of DDR5 contract price hikes in the fourth quarter will be visibly greater than in the first half.\n\nThe strategies put in place by the three global cloud giants to lock in next year’s volumes are also cited as another factor fueling price increases. Major cloud service providers (CSPs) such as Amazon, Google, and Microsoft are accelerating early contracts by prioritizing next year’s DDR5 volumes in order to speed up AI infrastructure build-outs, and in the process, DDR5’s premium pricing structure has become entrenched.\n\nIn the DDR4 era, prices were adjusted through quarterly negotiations, but DDR5 is creating a new market order in which demand outstrips supply. The growing prevalence of advance purchases and early-allocation contracts, rarely seen in the traditional commodity memory market, is part of the same trend.\n\nThe inventory days of Samsung Electronics and SK hynix also lend support to the outlook for further price increases. The DRAM inventory days of both companies have fallen to the lowest level in the past three years, providing a basis for additional price hikes. As DDR5 is already in a structurally “tight supply-demand” state in the market, analysis is emerging that its strength is likely to continue into next year without any short-term correction.\n\nThe fact that surging AI server demand is simultaneously tightening both commodity memory and HBM is another key variable. While the share of HBM has increased as AI training GPUs require high-performance memory, the baseline configuration of AI servers has also expanded, pushing DDR5 installed capacity significantly higher compared with the past. AI companies and CSPs are competing to scale up their AI models, and the dominant assessment is that the likelihood of server DRAM demand settling at a lower level next year is not high.\n\nManufacturers project that the upward pressure on prices will persist at least through the first half of next year. In the initial phase of HBM4 mass production, it will take time to stabilize processes and secure yields, and during this period, production capacity is unlikely to stretch far enough to cover DDR5. Within the semiconductor industry, the prevailing view is that as HBM4 preparations move into full swing, tight conditions across overall DRAM will be maintained at least through the first half, and in the near term, DDR5 prices are more likely to see further increases than any meaningful correction.\n\nSemiconductor industry experts and market analysts judge that the current DDR5 price hikes should be seen not as a mere seasonal factor but as a “medium- to long-term inflection point” created by changes in AI infrastructure and process transitions. With yet another round of price increases signaled for year-end and upward pressure likely to persist across memory into the first half of next year, how cloud and server companies respond is expected to become another key point of interest for the industry.\n\n---\n\nhttps://t.co/Saw141jWTd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09285720201056402, "ret_signed_m1d": 0.09285720201056402, "alpha_spy_m1d": 0.09269358914618975, "alpha_c_m1d": 0.0933210309821717, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08214287073215654, "ret_signed_p1d": 0.08214287073215654, "alpha_spy_p1d": 0.0914593946631187, "alpha_c_p1d": 0.09570900296467622, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.06964281757401458, "ret_signed_p1w": -0.06964281757401458, "alpha_spy_p1w": -0.05044438723436118, "alpha_c_p1w": -0.015001954481206226, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.010005705267535414, "ret_signed_p1m": -0.010005705267535414, "alpha_spy_p1m": -0.02310222534706352, "alpha_c_p1m": -0.03296350898810585, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": 0.536814195618829, "ret_signed_p3m": 0.536814195618829, "alpha_spy_p3m": 0.5039615363570986, "alpha_c_p3m": 0.33074253506319273, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 2.3657463705299318, "ret_signed_p6m": 2.3657463705299318, "alpha_spy_p6m": 2.2592181350238754, "alpha_c_p6m": 1.6899877568168509, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.531, -0.5444, -0.5631, -0.5524, -0.5373, -0.5337, -0.5364, -0.5205, -0.5196, -0.5429, -0.5348, -0.5312, -0.5259, -0.5116, -0.5054, -0.4857, -0.4571, -0.4518, -0.4134, -0.4089, -0.3786, -0.4045, -0.3652, -0.3652, -0.3732, -0.3554, -0.3616, -0.3634, -0.3991, -0.3768, -0.3795, -0.3571, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2937, -0.2357, -0.2589, -0.2652, -0.2455, -0.192, -0.1687, -0.133, -0.1446, -0.1402, -0.1455, -0.0893, -0.0446, -0.0696, -0.0036, 0.0143, -0.0018, 0.1071, 0.0464, 0.0339, 0.0589, 0.0357, 0.0821, 0.1054, 0.1018, 0.0929, 0.0, 0.0821, 0.0179, 0.0036, 0.0196, -0.0696, -0.0714, -0.0732, -0.0643, -0.0279, -0.0529, -0.0386, -0.0029, -0.0136, -0.0314, -0.0279, 0.0311, 0.0114, 0.049, 0.0097, 0.0204, -0.01, -0.0529, -0.0154, -0.0136, -0.0225, 0.0365, 0.0436, 0.0508, 0.0508, 0.0704, 0.1437, 0.1633, 0.1633, 0.1633, 0.2098, 0.2437, 0.2974, 0.3259, 0.351, 0.3295, 0.3385, 0.3188, 0.3259, 0.3385, 0.351, 0.3653, 0.3277, 0.3224, 0.3492, 0.3706, 0.3152, 0.4296, 0.5029, 0.5386, 0.6244, 0.4832, 0.6208, 0.6083, 0.5046, 0.4993, 0.5851, 0.5654, 0.5368, 0.5868, 0.5726, 0.5726, 0.5726, 0.5726, 0.5976, 0.6959, 0.6994, 0.7959, 0.8192, 0.9675, 0.8995, 0.8995, 0.6811, 0.52, 0.6847, 0.6542, 0.4967, 0.6793, 0.7097, 0.665, 0.6292, 0.7437, 0.7366, 0.8905, 0.8136, 0.8028, 0.6703, 0.7652, 0.7813, 0.6703, 0.6703, 0.5629, 0.4448, 0.6059, 0.4859, 0.5683, 0.5862, 0.6399, 0.8494, 0.7867, 0.8386, 0.8619, 0.9747, 1.0338, 1.0678, 1.0194, 1.0875, 1.1913, 1.1895, 1.1931, 1.1877, 1.3131, 1.3274, 1.3148, 1.3023, 1.3023, 1.5906, 1.5906, 1.8663, 1.9611, 2.0184, 2.3657]}
{"unit_id": "reply:1978992163611427037", "ticker_idx": 0, "ts": "2025-10-17T01:11:14+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I hold a long position in AMAT.", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Holds a live long position in AMAT.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@ArikCR23 I hold a long position in AMAT.", "reply_text": null, "parent_text": "@Jukanlosreve How do you think this will affect AMAT KLAC CAMT NVMI?", "quoted_text": null, "parent_author": "ArikCR23", "ret_m1d": 0.01213384443999832, "ret_signed_m1d": 0.01213384443999832, "alpha_spy_m1d": 0.017778106273435457, "alpha_c_m1d": 0.010762919198954668, "bench_spy_m1d": -0.005644261833437136, "bench_c_m1d": 0.0013709252410436523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013956179354378095, "ret_signed_p1d": 0.013956179354378095, "alpha_spy_p1d": 0.0035557039547564617, "alpha_c_p1d": 0.0011509357461507719, "bench_spy_p1d": 0.010400475399621634, "bench_c_p1d": 0.012805243608227324, "ret_p1w": 0.016711787354885432, "ret_signed_p1w": 0.016711787354885432, "alpha_spy_p1w": -0.002644283414024784, "alpha_c_p1w": -0.0075277746205255625, "bench_spy_p1w": 0.019356070768910216, "bench_c_p1w": 0.024239561975410995, "ret_p1m": 0.016534056735782787, "ret_signed_p1m": 0.016534056735782787, "alpha_spy_p1m": 0.014607550028671756, "alpha_c_p1m": 0.023913818832525968, "bench_spy_p1m": 0.001926506707111031, "bench_c_p1m": -0.007379762096743181, "ret_p3m": 0.344423206237207, "ret_signed_p3m": 0.344423206237207, "alpha_spy_p3m": 0.3022643804860985, "alpha_c_p3m": 0.2081211987197542, "bench_spy_p3m": 0.04215882575110852, "bench_c_p3m": 0.1363020075174528, "ret_p6m": 0.7645238920975543, "ret_signed_p6m": 0.7645238920975543, "alpha_spy_p6m": 0.7259673345466311, "alpha_c_p6m": 0.4673226047690182, "bench_spy_p6m": 0.03855655755092324, "bench_c_p6m": 0.2972012873285361, "price_path": [-0.1706, -0.1712, -0.1663, -0.177, -0.1567, -0.165, -0.1606, -0.202, -0.2023, -0.1898, -0.206, -0.2105, -0.1883, -0.1807, -0.1828, -0.1648, -0.1578, -0.1657, -0.2831, -0.2752, -0.2811, -0.2866, -0.2896, -0.2778, -0.28, -0.2688, -0.2693, -0.2654, -0.2855, -0.2855, -0.2997, -0.3055, -0.2967, -0.2766, -0.2797, -0.2733, -0.2737, -0.2437, -0.2542, -0.2403, -0.2287, -0.2083, -0.1566, -0.1551, -0.1088, -0.1072, -0.1047, -0.1128, -0.0936, -0.0891, -0.09, -0.0322, -0.0062, -0.0332, -0.0048, -0.0597, -0.0332, -0.0208, -0.0668, -0.0245, -0.0302, 0.0115, 0.0121, 0.0, 0.014, 0.0045, -0.0197, 0.0155, 0.0167, 0.0282, 0.0118, 0.0478, 0.0336, 0.036, 0.0565, 0.0231, 0.0707, 0.038, 0.0226, 0.0448, 0.0164, 0.0255, -0.0078, 0.0045, 0.0165, 0.0006, 0.0451, -0.0192, -0.0024, 0.0283, 0.0798, 0.1132, 0.1132, 0.1234, 0.1345, 0.1816, 0.1963, 0.1999, 0.1935, 0.1942, 0.1897, 0.2253, 0.2029, 0.1544, 0.1635, 0.1527, 0.1056, 0.1289, 0.1419, 0.1535, 0.1589, 0.1613, 0.1613, 0.1663, 0.1715, 0.1577, 0.1445, 0.1445, 0.1974, 0.2662, 0.3182, 0.3013, 0.2542, 0.3413, 0.3682, 0.3577, 0.3444, 0.421, 0.4563, 0.4563, 0.4172, 0.4484, 0.4197, 0.4357, 0.4227, 0.4817, 0.4997, 0.5201, 0.4354, 0.4625, 0.4192, 0.3253, 0.3538, 0.4363, 0.4721, 0.4655, 0.5136, 0.4624, 0.5805, 0.5805, 0.5993, 0.6446, 0.649, 0.6738, 0.6656, 0.6852, 0.761, 0.6753, 0.6601, 0.6595, 0.5665, 0.5952, 0.5451, 0.448, 0.5113, 0.5422, 0.5654, 0.5039, 0.5229, 0.5436, 0.5716, 0.5583, 0.5928, 0.5921, 0.6132, 0.6676, 0.6469, 0.5096, 0.5034, 0.4408, 0.524, 0.5776, 0.5538, 0.5538, 0.5723, 0.5798, 0.7199, 0.7738, 0.7813, 0.7645]}
{"unit_id": "orig:1989479175758242204", "ticker_idx": 0, "ts": "2025-11-14T23:42:52+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raised its forecast for NVIDIA's FY2027 revenue/EPS to 298.5 billion U.S. dollars / 7.11 U.S. dollars and increased its target price from 210 to 220 U.S. dollars. It maintains an 'Overweight' rating.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/endorses Morgan Stanley Overweight on NVDA, PT raised to $220 on Blackwell full-ramp, accelerating demand, and raised FY2027 estimates.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "NVIDIA: Full-Scale Ramp of Blackwell, Demand Accelerating, Outlook Raised, Target Price Increased\nMorgan Stanley (J. Moore, 25/11/14)\n\nMorgan Stanley stated that over the past 45 days, industry conditions have clearly inflected to the upside, Blackwell cabinets have entered a phase of fully accelerated shipments, and this quarter is expected to be the strongest in several quarters. While the stock’s performance this year has not been weak, it has already clearly lagged AI peers, and the firm believes this trend will reverse.\n\nAccording to supply-chain research, cabinet-related issues have been completely resolved and demand on NVIDIA’s side is heating up sharply. Bottlenecks are now more visible in related infrastructure such as memory, servers, power and data-center space, but not to a degree that restricts GPU shipments. Customers and suppliers broadly report that demand is accelerating, whereas market expectations for NVIDIA’s growth remain low, creating a clear mismatch with the signals coming from the broader ecosystem. CEO Jensen Huang noted at GTC that revenue for the next five quarters needs to be raised by 70–80 billion U.S. dollars, yet the share price is currently about 10% lower than it was at that time.\n\nCloud providers’ 2026 Capex was revised up by 142 billion dollars in the third quarter, and the ROIs of major hyperscale customers remain very solid, supporting continued expansion. On the ODM side, Quanta expects its AI server revenue to double in 2026 and plans to double capacity as well; Wistron forecasts that AI servers will grow quarter-on-quarter throughout 2026; and Hon Hai (Foxconn) is expected to achieve “strong double-digit growth” in cabinet shipments in the fourth quarter. Downstream demand for key components such as optical modules, DRAM and NAND is likewise very strong.\n\nMorgan Stanley raised its forecast for NVIDIA’s FY2027 revenue/EPS to 298.5 billion U.S. dollars / 7.11 U.S. dollars and increased its target price from 210 to 220 U.S. dollars (applying a 26x multiple to 2027 MW EPS of 8.43 U.S. dollars). The firm believes that although a higher ASIC mix next year could put some pressure on NVIDIA’s market share, enthusiasm for alternative solutions will decline markedly after Rubin is launched, and product leadership will remain intact. It maintains an “Overweight” rating.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017405531691603637, "ret_signed_m1d": -0.017405531691603637, "alpha_spy_m1d": -0.01756914455597791, "alpha_c_m1d": -0.016941702719995955, "bench_spy_m1d": 0.00016361286437427403, "bench_c_m1d": -0.00046382897160768266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018772690180637097, "ret_signed_p1d": -0.018772690180637097, "alpha_spy_p1d": -0.009456166249674935, "alpha_c_p1d": -0.0052065579481174185, "bench_spy_p1d": -0.009316523930962162, "bench_c_p1d": -0.013566132232519679, "ret_p1w": -0.05936791649684514, "ret_signed_p1w": -0.05936791649684514, "alpha_spy_p1w": -0.040169486157191736, "alpha_c_p1w": -0.004727053404036785, "bench_spy_p1w": -0.019198430339653405, "bench_c_p1w": -0.054640863092808356, "ret_p1m": -0.07293574274855108, "ret_signed_p1m": -0.07293574274855108, "alpha_spy_p1m": -0.08603226282807919, "alpha_c_p1m": -0.09589354646912152, "bench_spy_p1m": 0.013096520079528107, "bench_c_p1m": 0.022957803720570436, "ret_p3m": -0.0005753669288237617, "ret_signed_p3m": -0.0005753669288237617, "alpha_spy_p3m": -0.03342802619055418, "alpha_c_p3m": -0.20664702748446007, "bench_spy_p3m": 0.03285265926173042, "bench_c_p3m": 0.2060716605556363, "ret_p6m": 0.15404161563996577, "ret_signed_p6m": 0.15404161563996577, "alpha_spy_p6m": 0.0475133801339096, "alpha_c_p6m": -0.5217169980731151, "bench_spy_p6m": 0.10652823550605617, "bench_c_p6m": 0.6757586137130809, "price_path": [-0.0765, -0.0777, -0.0799, -0.0641, -0.0545, -0.0442, -0.0451, -0.0526, -0.0841, -0.0841, -0.102, -0.1029, -0.0974, -0.1218, -0.115, -0.1021, -0.0676, -0.0684, -0.0649, -0.0653, -0.0804, -0.1045, -0.0733, -0.071, -0.0345, -0.0617, -0.0694, -0.0656, -0.063, -0.0438, -0.0189, -0.0154, -0.0067, -0.0134, -0.0243, -0.027, -0.0056, 0.0126, -0.0369, -0.0097, -0.0533, -0.0544, -0.044, -0.0365, -0.0396, -0.0474, -0.052, -0.0421, -0.0206, 0.0069, 0.0571, 0.0887, 0.0669, 0.0648, 0.0879, 0.0448, 0.0265, -0.011, -0.0106, 0.0467, 0.0157, 0.0191, -0.0174, 0.0, -0.0188, -0.0463, -0.0192, -0.0501, -0.0594, -0.0401, -0.0649, -0.0521, -0.0521, -0.0693, -0.0539, -0.0458, -0.0556, -0.0357, -0.0408, -0.0242, -0.0273, -0.0335, -0.0485, -0.0796, -0.0729, -0.0654, -0.1011, -0.0842, -0.0482, -0.034, -0.005, -0.0081, -0.0081, 0.0019, -0.0102, -0.0138, -0.0192, -0.0192, -0.0069, -0.0107, -0.0154, -0.0055, -0.0269, -0.0279, -0.0274, -0.0229, -0.0369, -0.0164, -0.0207, -0.0207, -0.0636, -0.036, -0.028, -0.0131, -0.0194, -0.0086, 0.0072, 0.0124, 0.0051, -0.0239, -0.0516, -0.084, -0.0961, -0.025, -0.0006, -0.0085, -0.0006, -0.0169, -0.0386, -0.0386, -0.0273, -0.0115, -0.0119, -0.0018, 0.0073, 0.0141, 0.0284, -0.0277, -0.0682, -0.0404, -0.0532, -0.0374, -0.0359, -0.0649, -0.0395, -0.0283, -0.0217, -0.0369, -0.0521, -0.0364, -0.0432, -0.0513, -0.0609, -0.0918, -0.0763, -0.0786, -0.0603, -0.0994, -0.119, -0.1314, -0.0828, -0.0757, -0.0671, -0.0671, -0.0658, -0.0634, -0.0424, -0.0328, -0.008, -0.0044, 0.0335, 0.0459, 0.0431, 0.0606, 0.0626, 0.0512, 0.065, 0.0499, 0.0953, 0.1392, 0.1211, 0.1005, 0.0495, 0.0437, 0.0438, 0.0334, 0.093, 0.1123, 0.1317, 0.154]}
{"unit_id": "thread:1976063938195521727", "ticker_idx": 0, "ts": "2025-10-08T23:15:31+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has finally secured a place in NVIDIA's supply chain — a development that is expected to reshape the global HBM market landscape", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung HBM3E confirmed for NVIDIA GB300; re-entry into NVIDIA supply chain after 19 months seen as highly meaningful with HBM4 qualification potentially accelerated.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung’s HBM Confirmed for NVIDIA GB300 — Jensen Huang Sends Letter to Lee Jae-yong\n\nNVIDIA has officially decided to equip its latest AI accelerator, the GB300, with Samsung Electronics’ (005930) fifth-generation high-bandwidth memory (HBM3E). After years of setbacks, Samsung has finally secured a place in NVIDIA’s supply chain — a development that is expected to reshape the global HBM market landscape.\n\nAccording to industry sources cited by News1 on the 9th, NVIDIA CEO Jensen Huang recently sent a letter to Samsung Chairman Lee Jae-yong, informing him of NVIDIA’s decision to adopt Samsung’s HBM3E 12-high stack for the GB300 system.\n\nThis letter effectively served as Huang’s official notification that Samsung had passed NVIDIA’s HBM3E qualification tests and that NVIDIA intended to place an order. The two companies are now said to be finalizing the details of supply volume, pricing, and schedule.\n\nThis marks Samsung’s re-entry into NVIDIA’s supply chain after about 19 months. During NVIDIA’s GTC 2024 in March last year, Huang personally signed “Jensen Approved” on Samsung’s 12-high HBM3E module, fueling speculation about a potential supply deal.\n\nHowever, Samsung repeatedly failed to secure the contract. Matters worsened when Huang remarked at CES 2025 in January that “HBM needs to be redesigned,” embarrassing the company publicly. Around that time, Chairman Lee’s call for a “tougher Samsung” pushed the firm to deliver a redesigned sample in the first half of this year.\n\nAlthough the industry is already transitioning to sixth-generation HBM4, the supply volume of Samsung’s HBM3E 12-high chips for GB300 is expected to be limited. Still, the fact that Samsung has overcome its long-standing technical bottlenecks is considered highly meaningful. The successful qualification of Samsung’s upcoming HBM4 could also be accelerated.\n\nIndustry observers attribute the breakthrough to Lee Jae-yong’s personal sales diplomacy. Following his acquittal by the Supreme Court in August, Lee spent much of that month in the United States holding business meetings — reportedly including a meeting with Jensen Huang. Upon returning to Korea after his first U.S. trip on August 15, Lee told reporters, “I’ve been preparing for next year’s business.”\n\nMeanwhile, sources say that in their letter exchange, Lee also expressed Samsung’s intention to purchase 50,000 NVIDIA GPUs. Although the specific purpose remains unclear, the GPUs are believed to support Samsung’s AX (AI transformation) initiative and expansion into AI-related businesses.\n\nAt a press conference during IFA 2025 in Berlin last month, Roh Tae-moon, acting head of Samsung’s DX (Device Experience) division, stated that the company aims to apply AI to 90% of its business operations by 2030. Some industry insiders believe the purchased GPUs could be used in the Pohang AI data center jointly built by Samsung and OpenAI.\n\n---\n\nhttps://t.co/lMJ2i3U5CG", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.005927705196020927, "alpha_c_m1d": 0.017234908638119495, "bench_spy_m1d": -0.005927705196020927, "bench_c_m1d": -0.017234908638119495, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0028970083792374535, "alpha_c_p1d": 0.0016856346179577875, "bench_spy_p1d": -0.0028970083792374535, "bench_c_p1d": -0.0016856346179577875, "ret_p1w": 0.05849584150355369, "ret_signed_p1w": 0.05849584150355369, "alpha_spy_p1w": 0.07029179428217125, "alpha_c_p1w": 0.08047804548871307, "bench_spy_p1w": -0.011795952778617558, "bench_c_p1w": -0.021982203985159376, "ret_p1m": 0.10529247964577548, "ret_signed_p1m": 0.10529247964577548, "alpha_spy_p1m": 0.10945223276901006, "alpha_c_p1m": 0.11055571898873173, "bench_spy_p1m": -0.0041597531232345775, "bench_c_p1m": -0.00526323934295625, "ret_p3m": 0.5461986354230954, "ret_signed_p3m": 0.5461986354230954, "alpha_spy_p3m": 0.5214743906236987, "alpha_c_p3m": 0.5496775991515271, "bench_spy_p3m": 0.024724244799396722, "bench_c_p3m": -0.0034789637284317054, "ret_p6m": 1.0015407389955011, "ret_signed_p6m": 1.0015407389955011, "alpha_spy_p6m": 1.0216647701304986, "alpha_c_p6m": 1.0633205320767984, "bench_spy_p6m": -0.0201240311349975, "bench_c_p6m": -0.06177979308129733, "price_path": [-0.3056, -0.3067, -0.2934, -0.2823, -0.2601, -0.2557, -0.2479, -0.2679, -0.2635, -0.2679, -0.269, -0.2191, -0.2169, -0.1947, -0.208, -0.2357, -0.2268, -0.2246, -0.2368, -0.218, -0.2036, -0.2124, -0.2113, -0.2024, -0.2058, -0.2058, -0.2235, -0.2235, -0.218, -0.2169, -0.208, -0.2069, -0.2202, -0.2169, -0.228, -0.2268, -0.2501, -0.2335, -0.2257, -0.2224, -0.2291, -0.2224, -0.2069, -0.1947, -0.1858, -0.1636, -0.1514, -0.1192, -0.1326, -0.1093, -0.1093, -0.0738, -0.0605, -0.0527, -0.0449, -0.076, -0.0618, -0.0652, -0.0418, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0518, 0.0396, 0.0206, 0.0585, 0.0886, 0.0908, 0.093, 0.0864, 0.0986, 0.0752, 0.1008, 0.1365, 0.1086, 0.1198, 0.1599, 0.1978, 0.2379, 0.1688, 0.1209, 0.1053, 0.0908, 0.1209, 0.1532, 0.1487, 0.1454, 0.083, 0.1209, 0.0897, 0.0752, 0.1209, 0.0563, 0.0774, 0.1064, 0.1454, 0.1532, 0.1198, 0.1231, 0.1521, 0.1643, 0.171, 0.2078, 0.2201, 0.2078, 0.2033, 0.1955, 0.2134, 0.1677, 0.1454, 0.2022, 0.1989, 0.1844, 0.2312, 0.2423, 0.2379, 0.2379, 0.3036, 0.3379, 0.3424, 0.3424, 0.3424, 0.4387, 0.5462, 0.5552, 0.5787, 0.554, 0.5563, 0.554, 0.5406, 0.5708, 0.6111, 0.6671, 0.6716, 0.6257, 0.6738, 0.7052, 0.7029, 0.7029, 0.7858, 0.8183, 0.7992, 0.797, 0.6839, 0.8754, 0.8933, 0.7836, 0.7757, 0.8631, 0.8563, 0.8787, 0.9996, 1.0288, 1.0288, 1.0288, 1.0288, 1.1273, 1.1284, 1.1609, 1.2392, 1.2784, 1.4408, 1.424, 1.424, 1.1844, 0.928, 1.1452, 1.1071, 0.9425, 1.1038, 1.1273, 1.1038, 1.0545, 1.1127, 1.1709, 1.3344, 1.2448, 1.2325, 1.0859, 1.1239, 1.1161, 1.0164, 1.0164, 0.978, 0.8759, 1.1278, 1.0015]}
{"unit_id": "quote:1978803285923704859", "ticker_idx": 0, "ts": "2025-10-16T12:40:42+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi) Micron – Following the Footsteps of AI: DRAM Expected to Face Unprecedented Demand Next; Margins to Return to Peak Levels, Estimates and Target Price Raised", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Citi bullish note on Micron: DRAM next AI contract wave, margins to peak, estimates/PT raised.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Will OpenAI sign a contract with Micron?\n\n$MU https://t.co/zyya6QpivJ", "reply_text": null, "parent_text": null, "quoted_text": "Citi) Micron – Following the Footsteps of AI: DRAM Expected to Face Unprecedented Demand Next; Margins to Return to Peak Levels, Estimates and Target Price Raised\n\nWe believe that DRAM will be the next semiconductor to sign long-term contracts with AI-related companies — similar", "parent_author": null, "ret_m1d": -0.05228857475168858, "ret_signed_m1d": -0.05228857475168858, "alpha_spy_m1d": -0.05914553294894942, "alpha_c_m1d": -0.0478610368068092, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.004427537944879378, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0007406080873467724, "ret_signed_p1d": -0.0007406080873467724, "alpha_spy_p1d": -0.006416908446254754, "alpha_c_p1d": 0.0006284402907192899, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": -0.0013690483780660623, "ret_p1w": 0.02063892716637472, "ret_signed_p1w": 0.02063892716637472, "alpha_spy_p1w": 0.0038068178817045606, "alpha_c_p1w": 0.01626952896893874, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.004369398197435981, "ret_p1m": 0.2187330413981221, "ret_signed_p1m": 0.2187330413981221, "alpha_spy_p1m": 0.20164356041185139, "alpha_c_p1m": 0.21383922624560192, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.004893815152520187, "ret_p3m": 0.6702048510508347, "ret_signed_p3m": 0.6702048510508347, "alpha_spy_p3m": 0.616953438108067, "alpha_c_p3m": 0.5261666585870479, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.1440381924637868, "ret_p6m": 1.078391654397973, "ret_signed_p6m": 1.078391654397973, "alpha_spy_p6m": 1.044047955953613, "alpha_c_p6m": 0.8018421877721944, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.2765494666257786, "price_path": [-0.4413, -0.4611, -0.458, -0.4487, -0.451, -0.451, -0.4475, -0.4338, -0.4615, -0.4825, -0.4682, -0.4618, -0.4632, -0.448, -0.4133, -0.3895, -0.3696, -0.3868, -0.3818, -0.4036, -0.3903, -0.3978, -0.4216, -0.4286, -0.4193, -0.4255, -0.4251, -0.419, -0.398, -0.4128, -0.4128, -0.4154, -0.4142, -0.3871, -0.3518, -0.3513, -0.3327, -0.3092, -0.257, -0.2242, -0.2215, -0.2163, -0.2105, -0.1666, -0.197, -0.1877, -0.1789, -0.2021, -0.2261, -0.224, -0.1912, -0.1744, -0.1012, -0.0933, -0.0726, -0.0571, -0.0831, -0.0296, -0.0504, -0.1033, -0.0482, -0.0764, -0.0523, 0.0, -0.0007, 0.0209, -0.0012, -0.02, 0.0206, 0.0814, 0.0868, 0.0957, 0.119, 0.1061, 0.1049, 0.1588, 0.0765, 0.1727, 0.1768, 0.1747, 0.2507, 0.1905, 0.2092, 0.17, 0.2187, 0.1946, 0.1282, 0.1155, -0.0057, 0.0239, 0.1057, 0.1086, 0.1369, 0.1369, 0.1676, 0.1873, 0.1825, 0.1562, 0.1191, 0.1713, 0.2192, 0.2463, 0.3021, 0.2762, 0.1906, 0.1727, 0.148, 0.1135, 0.2272, 0.313, 0.3657, 0.3641, 0.4155, 0.4155, 0.4062, 0.4541, 0.4455, 0.4098, 0.4098, 0.558, 0.5419, 0.6964, 0.6772, 0.6153, 0.7046, 0.7084, 0.6702, 0.6466, 0.6628, 0.7918, 0.7918, 0.8029, 0.922, 0.9639, 0.9741, 0.9219, 1.0264, 1.1501, 1.1526, 1.0493, 1.1625, 1.0718, 0.8741, 0.8913, 0.9496, 0.8943, 0.8437, 1.0269, 1.0448, 1.0334, 1.0334, 0.9747, 1.0793, 1.0615, 1.115, 1.0794, 1.0648, 1.1191, 1.0527, 1.0369, 1.0384, 0.8754, 0.9796, 0.9612, 0.8291, 0.9231, 0.9912, 1.0681, 1.0022, 1.1049, 1.1823, 1.2805, 1.2807, 1.1945, 1.0889, 0.9973, 0.9537, 0.8873, 0.7558, 0.7645, 0.5902, 0.6695, 0.8178, 0.8098, 0.8098, 0.8667, 0.8659, 1.0099, 1.0829, 1.0784]}
{"unit_id": "quote:1988564824209359042", "ticker_idx": 0, "ts": "2025-11-12T11:09:34+00:00", "symbol": "Asia tech", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The AI- and semiconductor-led upcycle extends to 2027; Asia tech still has upside next year", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares J.P. Morgan view that AI/semi upcycle extends to 2027 with Asia tech upside; capex at top CSPs surging 67% in 2025 and growing further through 2027.", "resolved_tickers": ["AAXJ", "SMH", "EWY", "EWT"], "resolution_method": "basket", "resolution_reason": "Asia tech broad sector proxy basket", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "\"While there are concerns about FCF at smaller CSPs, J.P. Morgan’s analysis suggests capex at the top six CSPs will surge 67% in 2025 and can grow another ~32% in 2026. Even assuming flat free cash flow in 2027, these companies would still have room to lift capex by a further 13%, indicating balance sheets can support continued expansion of AI infrastructure.\"", "reply_text": null, "parent_text": null, "quoted_text": "J.P. Morgan: The AI- and semiconductor-led upcycle extends to 2027; Asia tech still has upside next year\n\nJ.P. Morgan’s latest research argues that despite persistent market concerns about an AI bubble, the AI-driven semiconductor upcycle has not yet peaked and will extend to https://t.co/EveRsVnFAt", "parent_author": null, "ret_m1d": -0.004947565268211662, "ret_signed_m1d": -0.004947565268211662, "alpha_spy_m1d": -0.004391476069025069, "alpha_c_m1d": -0.004391476069025069, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.0005560891991865935, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02072676059877432, "ret_signed_p1d": -0.02072676059877432, "alpha_spy_p1d": -0.004132739128036467, "alpha_c_p1d": -0.004132739128036467, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.016594021470737852, "ret_p1w": -0.04474656220275103, "ret_signed_p1w": -0.04474656220275103, "alpha_spy_p1w": -0.014382888309769137, "alpha_c_p1w": -0.014382888309769137, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.03036367389298189, "ret_p1m": 0.0052753108574379315, "ret_signed_p1m": 0.0052753108574379315, "alpha_spy_p1m": -0.0031972573548830763, "alpha_c_p1m": -0.0031972573548830763, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.008472568212321008, "ret_p3m": 0.17832958979315494, "ret_signed_p3m": 0.17832958979315494, "alpha_spy_p3m": 0.1598616977242211, "alpha_c_p3m": 0.1598616977242211, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.018467892068933844, "ret_p6m": 0.5355167818257083, "ret_signed_p6m": 0.5355167818257083, "alpha_spy_p6m": 0.45888937883502456, "alpha_c_p6m": 0.45888937883502456, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.07662740299068371, "price_path": [-0.1438, -0.1406, -0.1537, -0.1608, -0.1632, -0.1445, -0.1488, -0.1472, -0.1472, -0.1419, -0.1569, -0.1569, -0.1634, -0.1572, -0.1533, -0.143, -0.134, -0.1271, -0.1153, -0.1059, -0.1035, -0.0977, -0.0909, -0.094, -0.0821, -0.0874, -0.0755, -0.0759, -0.0835, -0.0867, -0.0914, -0.0836, -0.0786, -0.0642, -0.0555, -0.0528, -0.0428, -0.0535, -0.0403, -0.0484, -0.0912, -0.0552, -0.0677, -0.0464, -0.0362, -0.035, -0.0195, -0.0293, -0.0356, -0.0249, -0.0106, 0.0068, 0.0095, 0.0214, 0.0081, 0.0122, 0.0274, -0.0055, 0.0057, -0.0146, -0.0235, 0.0029, -0.0049, 0.0, -0.0207, -0.0159, -0.0325, -0.0448, -0.0447, -0.0666, -0.0637, -0.0463, -0.0454, -0.0305, -0.0305, -0.0251, -0.0271, -0.0169, -0.0102, -0.0168, -0.0028, 0.0025, 0.0039, 0.0172, 0.0053, -0.0227, -0.0241, -0.03, -0.0474, -0.0318, -0.0164, -0.0093, -0.0012, 0.0053, 0.0053, 0.0138, 0.0211, 0.0226, 0.0151, 0.0151, 0.0504, 0.0657, 0.0841, 0.0779, 0.0719, 0.0883, 0.0958, 0.0886, 0.0943, 0.1088, 0.1137, 0.1137, 0.0969, 0.1249, 0.1312, 0.142, 0.1402, 0.1711, 0.1866, 0.1778, 0.15, 0.1535, 0.1554, 0.1274, 0.1244, 0.1666, 0.1783, 0.1762, 0.2122, 0.2004, 0.2127, 0.2127, 0.2026, 0.2137, 0.2166, 0.2506, 0.2341, 0.2671, 0.2923, 0.2799, 0.2765, 0.2594, 0.1793, 0.1939, 0.1578, 0.1451, 0.1838, 0.1757, 0.1922, 0.1417, 0.1436, 0.1878, 0.1986, 0.1779, 0.1904, 0.1403, 0.1829, 0.1617, 0.1714, 0.1203, 0.1146, 0.0894, 0.1382, 0.1584, 0.1431, 0.1431, 0.1584, 0.1679, 0.2506, 0.253, 0.2588, 0.2774, 0.3116, 0.3095, 0.3217, 0.3539, 0.3478, 0.3356, 0.3849, 0.3617, 0.4118, 0.4143, 0.3938, 0.3977, 0.4378, 0.4444, 0.4489, 0.5017, 0.5619, 0.5355]}
{"unit_id": "quote:1988564824209359042", "ticker_idx": 1, "ts": "2025-11-12T11:09:34+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "the AI-driven semiconductor upcycle has not yet peaked and will extend to", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares J.P. Morgan view that AI/semi upcycle extends to 2027 with Asia tech upside; capex at top CSPs surging 67% in 2025 and growing further through 2027.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "Global semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "\"While there are concerns about FCF at smaller CSPs, J.P. Morgan’s analysis suggests capex at the top six CSPs will surge 67% in 2025 and can grow another ~32% in 2026. Even assuming flat free cash flow in 2027, these companies would still have room to lift capex by a further 13%, indicating balance sheets can support continued expansion of AI infrastructure.\"", "reply_text": null, "parent_text": null, "quoted_text": "J.P. Morgan: The AI- and semiconductor-led upcycle extends to 2027; Asia tech still has upside next year\n\nJ.P. Morgan’s latest research argues that despite persistent market concerns about an AI bubble, the AI-driven semiconductor upcycle has not yet peaked and will extend to https://t.co/EveRsVnFAt", "parent_author": null, "ret_m1d": -0.012600882617153641, "ret_signed_m1d": -0.012600882617153641, "alpha_spy_m1d": -0.012044793417967048, "alpha_c_m1d": -0.012044793417967048, "bench_spy_m1d": -0.0005560891991865935, "bench_c_m1d": -0.0005560891991865935, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03012398366125879, "ret_signed_p1d": -0.03012398366125879, "alpha_spy_p1d": -0.013529962190520939, "alpha_c_p1d": -0.013529962190520939, "bench_spy_p1d": -0.016594021470737852, "bench_c_p1d": -0.016594021470737852, "ret_p1w": -0.045256322419341855, "ret_signed_p1w": -0.045256322419341855, "alpha_spy_p1w": -0.014892648526359964, "alpha_c_p1w": -0.014892648526359964, "bench_spy_p1w": -0.03036367389298189, "bench_c_p1w": -0.03036367389298189, "ret_p1m": 0.04320317372272764, "ret_signed_p1m": 0.04320317372272764, "alpha_spy_p1m": 0.03473060551040663, "alpha_c_p1m": 0.03473060551040663, "bench_spy_p1m": 0.008472568212321008, "bench_c_p1m": 0.008472568212321008, "ret_p3m": 0.14734435182625183, "ret_signed_p3m": 0.14734435182625183, "alpha_spy_p3m": 0.12887645975731798, "alpha_c_p3m": 0.12887645975731798, "bench_spy_p3m": 0.018467892068933844, "bench_c_p3m": 0.018467892068933844, "ret_p6m": 0.5238673692634213, "ret_signed_p6m": 0.5238673692634213, "alpha_spy_p6m": 0.44723996627273754, "alpha_c_p6m": 0.44723996627273754, "bench_spy_p6m": 0.07662740299068371, "bench_c_p6m": 0.07662740299068371, "price_path": [-0.1684, -0.165, -0.1819, -0.1872, -0.1915, -0.1741, -0.1733, -0.1653, -0.1628, -0.1593, -0.1835, -0.1835, -0.1942, -0.1944, -0.1847, -0.1751, -0.166, -0.1623, -0.1539, -0.1451, -0.144, -0.136, -0.1359, -0.1414, -0.1085, -0.112, -0.0938, -0.0952, -0.0968, -0.097, -0.0948, -0.0925, -0.082, -0.0614, -0.0486, -0.053, -0.0342, -0.052, -0.0266, -0.029, -0.0849, -0.0444, -0.062, -0.0387, -0.0344, -0.0357, -0.0234, -0.0288, -0.0476, -0.0302, -0.0123, 0.0124, 0.0213, 0.0368, 0.0231, 0.0211, 0.0297, -0.0078, 0.0112, -0.0125, -0.0208, 0.0092, -0.0126, 0.0, -0.0301, -0.0297, -0.0428, -0.0626, -0.0453, -0.0856, -0.0827, -0.0462, -0.0437, -0.022, -0.022, -0.0091, -0.0072, 0.011, 0.0249, 0.0171, 0.025, 0.0366, 0.0379, 0.0522, 0.0432, -0.004, -0.0074, -0.0101, -0.0458, -0.0233, 0.002, 0.0149, 0.0246, 0.0275, 0.0275, 0.0323, 0.0276, 0.0251, 0.0161, 0.0161, 0.0532, 0.0654, 0.0937, 0.0863, 0.0693, 0.0982, 0.1022, 0.1047, 0.0957, 0.1185, 0.1297, 0.1297, 0.1015, 0.134, 0.1365, 0.1288, 0.1253, 0.149, 0.1755, 0.178, 0.1383, 0.1511, 0.122, 0.0779, 0.0752, 0.1332, 0.1473, 0.142, 0.1703, 0.1458, 0.1504, 0.1504, 0.1498, 0.164, 0.1574, 0.171, 0.1649, 0.1826, 0.2024, 0.1625, 0.1466, 0.1466, 0.1034, 0.126, 0.1155, 0.0737, 0.1127, 0.121, 0.1315, 0.0951, 0.0928, 0.1114, 0.1198, 0.1107, 0.1142, 0.0855, 0.1042, 0.1133, 0.1258, 0.0745, 0.0559, 0.0229, 0.0817, 0.1059, 0.1069, 0.1069, 0.1172, 0.1283, 0.1932, 0.2141, 0.2326, 0.2509, 0.2753, 0.2781, 0.2832, 0.3096, 0.309, 0.311, 0.3454, 0.3595, 0.4289, 0.4284, 0.3859, 0.4095, 0.4297, 0.4384, 0.4299, 0.4747, 0.5511, 0.5239]}
{"unit_id": "thread:2002966788523643238", "ticker_idx": 0, "ts": "2025-12-22T04:57:50+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "revenue guidance for the second quarter is projected at up to $19.1 billion, far exceeding the market expectation of $14.4 billion", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: record revenue beat, Q2 guide $19.1B vs $14.4B consensus, rising DRAM prices, 1γ ramp driving further growth.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Micron’s Performance Soars Despite Dwindling China Revenue, Fueled by Rising DRAM Prices\n\nMicron Technology recently delivered an \"earnings surprise\" despite a decline in mobile DRAM sales to the Chinese market. A steep upward trend in commodity DRAM prices bolstered the overall performance. As a result, observers suggest that while Micron may lag behind competitors in the next-generation HBM4 market due to technical hurdles, its broader profitability remains on a high-growth trajectory.\n\nAccording to Micron’s results for the first quarter of fiscal year 2026 (September–November 2025), revenue reached $13.643 billion (approximately KRW 20.146 trillion), a 55.6% year-over-year increase. This figure surpassed the market consensus of $13 billion. Earnings per share (EPS) also came in at $4.78, significantly exceeding the average market estimate of $3.93.\n\nDuring the conference call, Micron stated, \"The company achieved record-high revenue overall, as well as in DRAM, NAND Flash, and HBM segments.\" The company added, \"We plan to increase DRAM and NAND bit shipments by approximately 20% next year.\" Despite the planned supply expansion, Micron explained that market demand is expected to remain robust for the foreseeable future.\n\nThis earnings surprise is particularly noteworthy because it comes at a time when Micron's supply of commodity DRAM to China has significantly decreased. According to industry sources, Micron’s revenue from mobile DRAM products such as LPDDR4 and LPDDR5 supplied to China has seen a substantial drop.\n\nAn industry official noted, \"Micron's mobile DRAM revenue from China fell by about 30% compared to last year. While demand from Chinese clients for commodity DRAM remains high due to supply shortages, geopolitical tensions between the U.S. and China have restricted Micron’s sales, leading to a decline in actual shipments.\"\n\nAs a U.S.-based company, Micron is more susceptible to geopolitical risks than its competitors. The official added, \"In the Chinese market, Samsung Electronics and SK Hynix only go as far as slightly lowering supply prices compared to local firms like CXMT, but Micron faces direct pressure on shipment volumes themselves.\"\n\nHowever, the overall impact on revenue was limited, as Micron diversified its mobile DRAM supply to various clients, including Apple and Samsung Electronics. Furthermore, the rising prices of commodity DRAM are generating significant profits. Market research firm TrendForce predicts that the price gap between HBM3E and server-grade DDR5 will narrow from the current 4–5x range to 1–2x by the end of next year.\n\nTo maximize profitability, Micron is currently using quarterly contract structures for all customers instead of Long-Term Agreements (LTA). Another industry source explained, \"Micron previously utilized LTAs or MOUs to secure volumes for key customers and negotiate prices. However, with the recent volatility in price increases, they have reportedly withdrawn these practices without exception.\"\n\nThere is also speculation that Micron may allocate more capacity to commodity DRAM. Nvidia, the largest consumer of HBM, prioritizes speed for the next-generation HBM4. However, Micron's design focuses on power efficiency, failing to meet these specific requirements. Compounded by redesign issues, Micron is evaluated as a latecomer in the Nvidia HBM4 supply chain. Consequently, Micron may pivot a significant portion of its capacity toward commodity DRAM such as GDDR7 and DDR5.\n\nMeanwhile, Micron provided an optimistic outlook for the next quarter. The revenue guidance for the second quarter (December 2025–February 2026) is projected at up to $19.1 billion, far exceeding the market expectation of $14.4 billion. Micron stated, \"The 1γ (gamma, equivalent to 1c) DRAM process is ramping up smoothly. It will be a key driver for DRAM bit growth next year and is expected to account for more than half of total bit production by the second half of the year.\"\n\n$MU\n\n---\n\nhttps://t.co/0Oqh5FDZ9c", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.038576997355407316, "ret_signed_m1d": -0.038576997355407316, "alpha_spy_m1d": -0.032385676340073166, "alpha_c_m1d": -0.025858031406643223, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.012718965948764094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0011570437939640854, "ret_signed_p1d": -0.0011570437939640854, "alpha_spy_p1d": -0.005727498612396964, "alpha_c_p1d": -0.010776146762152772, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.009619102968188686, "ret_p1w": 0.06471284763628149, "ret_signed_p1w": 0.06471284763628149, "alpha_spy_p1w": 0.06030307636099397, "alpha_c_p1w": 0.05217467945448262, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.01253816818179887, "ret_p1m": 0.3201759263113946, "ret_signed_p1m": 0.3201759263113946, "alpha_spy_p1m": 0.3307625409969053, "alpha_c_p1m": 0.23485482150214, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": 0.08532110480925459, "ret_p3m": 0.6068891257649625, "ret_signed_p3m": 0.6068891257649625, "alpha_spy_p3m": 0.6434383760752479, "alpha_c_p3m": 0.5089741965866124, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": 0.0979149291783501, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4157, -0.4333, -0.4318, -0.4078, -0.3954, -0.3419, -0.3361, -0.3209, -0.3096, -0.3286, -0.2894, -0.3046, -0.3434, -0.303, -0.3237, -0.306, -0.2678, -0.2683, -0.2524, -0.2686, -0.2824, -0.2526, -0.2081, -0.2042, -0.1977, -0.1806, -0.1901, -0.191, -0.1515, -0.2117, -0.1413, -0.1383, -0.1398, -0.0842, -0.1283, -0.1146, -0.1433, -0.1076, -0.1252, -0.1739, -0.1832, -0.272, -0.2503, -0.1904, -0.1882, -0.1675, -0.1675, -0.145, -0.1306, -0.1341, -0.1534, -0.1806, -0.1423, -0.1073, -0.0874, -0.0466, -0.0655, -0.1282, -0.1413, -0.1594, -0.1846, -0.1014, -0.0386, 0.0, -0.0012, 0.0365, 0.0365, 0.0296, 0.0647, 0.0584, 0.0323, 0.0323, 0.1408, 0.129, 0.2422, 0.2281, 0.1828, 0.2482, 0.251, 0.223, 0.2057, 0.2176, 0.312, 0.312, 0.3202, 0.4074, 0.438, 0.4455, 0.4073, 0.4838, 0.5744, 0.5762, 0.5006, 0.5835, 0.5171, 0.3723, 0.3849, 0.4276, 0.3871, 0.35, 0.4842, 0.4973, 0.4889, 0.4889, 0.446, 0.5225, 0.5095, 0.5487, 0.5226, 0.5119, 0.5517, 0.503, 0.4915, 0.4926, 0.3733, 0.4496, 0.4361, 0.3393, 0.4081, 0.458, 0.5144, 0.4661, 0.5413, 0.598, 0.6699, 0.67, 0.6069, 0.5296, 0.4625, 0.4306, 0.382, 0.2857, 0.292, 0.1644, 0.2225, 0.331, 0.3252, 0.3252, 0.3669, 0.3663, 0.4717, 0.5252, 0.5219, 0.5435, 0.685, 0.6508, 0.6545, 0.6466, 0.6226, 0.6261, 0.7639, 0.7431, 0.7974, 0.8981, 0.8247, 0.876, 0.8713, 0.962, 1.0858, 1.3165, 1.412, 1.3398, 1.7023, 1.8779, 1.7738, 1.9079, 1.8079, 1.6221, 1.4661, 1.5283, 1.6487, 1.7576, 1.7174, 1.7174, 2.2417, 2.3594, 2.3417, 2.5135, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:1978636564948525277", "ticker_idx": 0, "ts": "2025-10-16T01:38:13+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix KRW 440,000", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Price targets: Samsung at 97,000 KRW and SK Hynix at 440,000 KRW implying upside from current levels.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "cy", "tweet_text": "SK Hynix KRW 440,000 https://t.co/IEFJDrh27x", "reply_text": null, "parent_text": null, "quoted_text": "Samsung -&gt; 97,000 KRW https://t.co/Qrphdd2tfX", "parent_author": null, "ret_m1d": -0.06629834330662243, "ret_signed_m1d": -0.06629834330662243, "alpha_spy_m1d": -0.07315530150388327, "alpha_c_m1d": -0.06187080536174305, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.004427537944879378, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028729263633047175, "ret_signed_p1d": 0.028729263633047175, "alpha_spy_p1d": 0.023052963274139193, "alpha_c_p1d": 0.030098312011113237, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": -0.0013690483780660623, "ret_p1w": 0.05745859651542884, "ret_signed_p1w": 0.05745859651542884, "alpha_spy_p1w": 0.04062648723075868, "alpha_c_p1w": 0.05308919831799286, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.004369398197435981, "ret_p1m": 0.2375689827245071, "ret_signed_p1m": 0.2375689827245071, "alpha_spy_p1m": 0.22047950173823638, "alpha_c_p1m": 0.23267516757198692, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.004893815152520187, "ret_p3m": 0.6321072323173902, "ret_signed_p3m": 0.6321072323173902, "alpha_spy_p3m": 0.5788558193746225, "alpha_c_p3m": 0.4880690398536034, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.1440381924637868, "ret_p6m": 1.2754296713454285, "ret_signed_p6m": 1.2754296713454285, "alpha_spy_p6m": 1.2410859729010686, "alpha_c_p6m": 0.9988802047196499, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.2765494666257786, "price_path": [-0.3987, -0.4075, -0.4064, -0.4053, -0.413, -0.4218, -0.4207, -0.4185, -0.3965, -0.4307, -0.4307, -0.4185, -0.4296, -0.4218, -0.434, -0.4108, -0.4064, -0.3865, -0.3898, -0.3898, -0.4097, -0.4196, -0.4362, -0.4593, -0.4461, -0.4273, -0.4229, -0.4262, -0.4066, -0.4055, -0.4343, -0.4243, -0.4199, -0.4133, -0.3956, -0.3878, -0.3635, -0.3282, -0.3215, -0.274, -0.2685, -0.2309, -0.263, -0.2144, -0.2144, -0.2243, -0.2022, -0.2099, -0.2122, -0.2564, -0.2287, -0.232, -0.2044, -0.126, -0.126, -0.126, -0.126, -0.126, -0.126, -0.0541, -0.0829, -0.0906, -0.0663, 0.0, 0.0287, 0.0729, 0.0586, 0.0641, 0.0575, 0.1271, 0.1823, 0.1514, 0.2331, 0.2552, 0.2354, 0.3702, 0.295, 0.2796, 0.3105, 0.2818, 0.3392, 0.368, 0.3635, 0.3525, 0.2376, 0.3392, 0.2597, 0.242, 0.2619, 0.1514, 0.1492, 0.147, 0.158, 0.2031, 0.1721, 0.1898, 0.234, 0.2208, 0.1986, 0.2031, 0.2761, 0.2517, 0.2982, 0.2495, 0.2628, 0.2252, 0.1721, 0.2186, 0.2208, 0.2097, 0.2827, 0.2915, 0.3004, 0.3004, 0.3247, 0.4154, 0.4397, 0.4397, 0.4397, 0.4972, 0.5392, 0.6056, 0.641, 0.6719, 0.6454, 0.6564, 0.6321, 0.641, 0.6564, 0.6719, 0.6896, 0.6432, 0.6365, 0.6697, 0.6962, 0.6277, 0.7692, 0.8599, 0.9041, 1.0103, 0.8356, 1.0059, 0.9904, 0.8621, 0.8555, 0.9616, 0.9373, 0.9019, 0.9638, 0.9461, 0.9461, 0.9461, 0.9461, 0.9771, 1.0987, 1.1032, 1.2226, 1.2513, 1.435, 1.3508, 1.3508, 1.0805, 0.8811, 1.0849, 1.0472, 0.8522, 1.0782, 1.1159, 1.0605, 1.0162, 1.158, 1.1491, 1.3397, 1.2444, 1.2311, 1.0672, 1.1846, 1.2045, 1.0672, 1.0672, 0.9342, 0.788, 0.9874, 0.839, 0.9409, 0.963, 1.0295, 1.2887, 1.2112, 1.2754]}
{"unit_id": "quote:1978636564948525277", "ticker_idx": 1, "ts": "2025-10-16T01:38:13+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung -> 97,000 KRW", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Price targets: Samsung at 97,000 KRW and SK Hynix at 440,000 KRW implying upside from current levels.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "cy", "tweet_text": "SK Hynix KRW 440,000 https://t.co/IEFJDrh27x", "reply_text": null, "parent_text": null, "quoted_text": "Samsung -&gt; 97,000 KRW https://t.co/Qrphdd2tfX", "parent_author": null, "ret_m1d": -0.027635641247316545, "ret_signed_m1d": -0.027635641247316545, "alpha_spy_m1d": -0.03449259944457739, "alpha_c_m1d": -0.026897457643640377, "bench_spy_m1d": 0.006856958197260843, "bench_c_m1d": -0.000738183603676168, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0020471203231524626, "ret_signed_p1d": 0.0020471203231524626, "alpha_spy_p1d": -0.0036291800357555193, "alpha_c_p1d": 0.00028969909330189836, "bench_spy_p1d": 0.005676300358907982, "bench_c_p1d": 0.0017574212298505643, "ret_p1w": -0.01228248038129387, "ret_signed_p1w": -0.01228248038129387, "alpha_spy_p1w": -0.02911458966596403, "alpha_c_p1w": -0.028380325472448287, "bench_spy_p1w": 0.01683210928467016, "bench_c_p1w": 0.016097845091154417, "ret_p1m": -0.005117720288674188, "ret_signed_p1m": -0.005117720288674188, "alpha_spy_p1m": -0.02220720127494491, "alpha_c_p1m": -0.017911647886877824, "bench_spy_p1m": 0.017089480986270722, "bench_c_p1m": 0.012793927598203636, "ret_p3m": 0.41523940333012943, "ret_signed_p3m": 0.41523940333012943, "alpha_spy_p3m": 0.36198799038736174, "alpha_c_p3m": 0.38397744065233974, "bench_spy_p3m": 0.053251412942767695, "bench_c_p3m": 0.03126196267778969, "ret_p6m": 1.1231304547233654, "ret_signed_p6m": 1.1231304547233654, "alpha_spy_p6m": 1.0887867562790055, "alpha_c_p6m": 1.1177955911711674, "bench_spy_p6m": 0.034343698444359916, "bench_c_p6m": 0.0053348635521979215, "price_path": [-0.3091, -0.3275, -0.3234, -0.3275, -0.3285, -0.2826, -0.2806, -0.2602, -0.2724, -0.2979, -0.2898, -0.2877, -0.2989, -0.2816, -0.2684, -0.2765, -0.2755, -0.2673, -0.2704, -0.2704, -0.2867, -0.2867, -0.2816, -0.2806, -0.2724, -0.2714, -0.2836, -0.2806, -0.2908, -0.2898, -0.3112, -0.2959, -0.2887, -0.2857, -0.2918, -0.2857, -0.2714, -0.2602, -0.2521, -0.2317, -0.2205, -0.1909, -0.2031, -0.1817, -0.1817, -0.1491, -0.1369, -0.1298, -0.1226, -0.1512, -0.1382, -0.1412, -0.1198, -0.0814, -0.0814, -0.0814, -0.0814, -0.0814, -0.0814, -0.0338, -0.045, -0.0624, -0.0276, 0.0, 0.002, 0.0041, -0.002, 0.0092, -0.0123, 0.0113, 0.044, 0.0184, 0.0287, 0.0655, 0.1003, 0.1372, 0.0737, 0.0297, 0.0154, 0.002, 0.0297, 0.0594, 0.0553, 0.0522, -0.0051, 0.0297, 0.001, -0.0123, 0.0297, -0.0297, -0.0102, 0.0164, 0.0522, 0.0594, 0.0287, 0.0317, 0.0583, 0.0696, 0.0757, 0.1095, 0.1208, 0.1095, 0.1054, 0.0983, 0.1146, 0.0727, 0.0522, 0.1044, 0.1013, 0.088, 0.131, 0.1412, 0.1372, 0.1372, 0.1975, 0.2291, 0.2332, 0.2332, 0.2332, 0.3216, 0.4204, 0.4286, 0.4502, 0.4276, 0.4296, 0.4276, 0.4152, 0.443, 0.48, 0.5315, 0.5356, 0.4934, 0.5376, 0.5664, 0.5644, 0.5644, 0.6405, 0.6703, 0.6528, 0.6508, 0.5469, 0.7228, 0.7392, 0.6384, 0.6312, 0.7115, 0.7053, 0.7259, 0.8369, 0.8637, 0.8637, 0.8637, 0.8637, 0.9542, 0.9552, 0.985, 1.057, 1.093, 1.2422, 1.2267, 1.2267, 1.0066, 0.7711, 0.9706, 0.9357, 0.7845, 0.9326, 0.9542, 0.9326, 0.8873, 0.9408, 0.9943, 1.1445, 1.0622, 1.0509, 0.9161, 0.9511, 0.9439, 0.8524, 0.8524, 0.817, 0.7232, 0.9546, 0.8387, 0.9191, 0.9902, 1.0252, 1.1695, 1.1025, 1.1231]}
{"unit_id": "reply:1997645169521238066", "ticker_idx": 0, "ts": "2025-12-07T12:31:37+00:00", "symbol": "CSCO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Cisco looks good.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish on Cisco.", "resolved_tickers": ["CSCO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@blueskyalpha I still don’t have conviction about the MI450.\n\nCisco looks good.\n\nI don’t have a view on IBM.", "reply_text": null, "parent_text": "@jukan05 What's your view on AMD 2026 releases onwards? And on IBM and Cisco?", "quoted_text": null, "parent_author": "blueskyalpha", "ret_m1d": -0.01128574265901594, "ret_signed_m1d": -0.01128574265901594, "alpha_spy_m1d": -0.014299073054442002, "alpha_c_m1d": -0.014299073054442002, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": 0.0030133303954260615, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00824259080682399, "ret_signed_p1d": 0.00824259080682399, "alpha_spy_p1d": 0.009105735773785595, "alpha_c_p1d": 0.009105735773785595, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": -0.0008631449669616043, "ret_p1w": -0.007735154354462925, "ret_signed_p1w": -0.007735154354462925, "alpha_spy_p1w": -0.003493056781998627, "alpha_c_p1w": -0.003493056781998627, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.004242097572464298, "ret_p1m": -0.04092613362271591, "ret_signed_p1m": -0.04092613362271591, "alpha_spy_p1m": -0.05588189575960889, "alpha_c_p1m": -0.05588189575960889, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": 0.01495576213689298, "ret_p3m": 0.020011908963416536, "ret_signed_p3m": 0.020011908963416536, "alpha_spy_p3m": 0.02046067610633484, "alpha_c_p3m": 0.02046067610633484, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": -0.0004487671429183049, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1412, -0.1468, -0.1614, -0.1552, -0.1564, -0.1464, -0.1343, -0.1402, -0.146, -0.1487, -0.1515, -0.1448, -0.1527, -0.1464, -0.1376, -0.1334, -0.139, -0.1387, -0.1262, -0.1252, -0.1082, -0.1129, -0.1385, -0.1446, -0.1293, -0.1184, -0.1239, -0.1107, -0.1041, -0.1032, -0.104, -0.1089, -0.1044, -0.0947, -0.0791, -0.0955, -0.0755, -0.0729, -0.0559, -0.0829, -0.0857, -0.0992, -0.0988, -0.0858, -0.0907, -0.0621, -0.0188, -0.0109, -0.0137, -0.0189, -0.006, -0.0434, -0.035, -0.0332, -0.0322, -0.0354, -0.0354, -0.0243, -0.0358, -0.0252, -0.0141, -0.0139, -0.0113, 0.0, 0.0082, 0.0176, 0.0052, -0.0134, -0.0077, -0.0166, -0.0363, -0.0242, -0.0056, -0.0095, -0.0107, -0.0107, -0.0107, -0.0089, -0.0136, -0.0184, -0.0232, -0.0232, -0.0306, -0.0365, -0.0409, -0.0474, -0.0571, -0.0581, -0.0565, -0.0379, -0.0514, -0.0407, -0.0414, -0.0414, -0.0649, -0.0606, -0.0524, -0.0491, -0.0182, 0.0031, 0.0066, -0.0001, -0.0015, 0.028, 0.0595, 0.0347, 0.05, 0.0813, 0.1063, 0.1001, 0.0905, -0.0439, -0.0203, -0.0203, -0.0203, -0.0033, 0.0015, 0.0097, -0.0089, -0.0038, 0.0087, -0.0043, 0.013, 0.0125, 0.0066, 0.031, 0.02, 0.0025, -0.0284, -0.0094, -0.0043, -0.0089, -0.0014, 0.0059, 0.0106, -0.0107, 0.0009, -0.0101, 0.0048, 0.0308, 0.0432, 0.0474, 0.0189, -0.0179, -0.0108, -0.0065, 0.0128, 0.0128, 0.0311, 0.0341, 0.0728, 0.066, 0.0539, 0.0555, 0.0589, 0.0557, 0.0831, 0.1055, 0.1242, 0.1497, 0.151, 0.1355, 0.1409, 0.1313, 0.1133, 0.1481, 0.1728, 0.1773, 0.1873, 0.2087, 0.1746, 0.1813, 0.2378, 0.2654, 0.2727, 0.3057, 0.4808, 0.5152, 0.5238, 0.4789, 0.4657, 0.515, 0.5434, 0.5434, 0.5167, 0.5339, 0.5207, 0.5435, null, null]}
{"unit_id": "thread:1980411935284949049", "ticker_idx": 0, "ts": "2025-10-20T23:12:54+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung is expected to secure the No. 2 position in NVIDIA's HBM4 supply chain next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung set to capture ~30% of NVIDIA HBM4 in H2 2026; bearish on SK Hynix margins and Micron's HBM4 competitiveness as Samsung disrupts the supply chain.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Expected to Rank Second in NVIDIA-Bound HBM4 Market Share… “Positive Shift in Samsung DS Division’s Morale”\n\nSamsung Electronics, which had long struggled in the high-bandwidth memory (HBM) market, is seeking a turnaround. Although it did not achieve notable success with the previous-generation HBM3E, industry observers say the atmosphere is shifting with the next-generation HBM4. In particular, visible collaboration with NVIDIA — the largest HBM customer — is positioning Samsung as a potential “game changer” capable of reshaping the market landscape.\n\nAccording to industry sources, Samsung is expected to secure the No. 2 position in NVIDIA’s HBM4 supply chain next year. While SK hynix is likely to maintain exclusivity in the first half, the competitive landscape is projected to be reorganized in the second half. Semiconductor industry insiders anticipate that SK hynix’s share of NVIDIA’s HBM4 orders will fall from 100% in the first half of next year to around 60% in the second half, while Samsung could capture up to 30% and Micron between 10–15%.\n\nGiven that Samsung’s share of NVIDIA’s HBM supply was previously at 0%, this represents remarkable growth. If these projections materialize, Samsung will not only break SK hynix’s near-monopoly but also widen the gap with competitor Micron.\n\nAn industry official stated, “NVIDIA has largely completed negotiations with SK hynix for first-half HBM4 supply,” adding, “Since SK hynix was the first to submit customer (CS) samples, it’s likely to complete qualification testing by year-end, making a monopoly structure inevitable for the first half.” The official continued, “However, from the second half onward, both Samsung and Micron will enter the supply chain, which will completely reshape the landscape.”\n\nThe fact that NVIDIA only finalized supply for the first half with SK hynix — the No. 1 HBM supplier — already signals a shift in its procurement strategy. Normally, NVIDIA signs one-year HBM supply contracts with vendors, but in the case of HBM4, the company is cautiously adjusting volumes and prices, anticipating new entrants.\n\nMicron, meanwhile, has found itself at a disadvantage after failing to meet NVIDIA’s requirements for higher HBM4 speeds. The company’s strategy of producing base dies in-house prioritized power efficiency, but that approach has limited performance improvement. Although Micron officially announced in its latest earnings release that the issue had been “resolved,” skepticism remains within the industry.\n\n“The base die issue Micron faces isn’t something that can be fixed in a month or two,” another industry source said. “It’s likely they were referring to samples that happened to perform better than average.”\n\nSamsung’s HBM4 yield still requires improvement but has shown progress compared to before. Another industry insider noted, “Samsung’s current 1c DRAM sample yield exceeds 70% for DDR5 and has reached about 50% for HBM4. That’s definitely an improvement over the past,” adding, “However, it’s still insufficient for full-scale mass production. To begin mass production of HBM, yields need to surpass at least 70%.”\n\nEven if Samsung is waging an “aggressive pricing offensive” for HBM in the second half, ramping up wafer input at this stage would sharply erode profitability. The key task is to improve yields before shifting to volume production. If the company can also resolve heat dissipation issues, its competitiveness could rise significantly. “Samsung’s HBM4 design is structurally advantageous for bandwidth expansion,” the insider said. “If it manages thermal control as well, it could become a very competitive chip.”\n\nWith these positive signals around HBM4, morale has reportedly improved within Samsung’s semiconductor (DS) division. Earlier this year, the team had been demoralized by falling behind competitors — to the extent that, during executive gatherings, conversation was often minimal.\n\nOne company insider shared, “Honestly, up until the first half, the department’s mood was terrible. Many engineers felt a sense of relative deprivation watching SK hynix perform so well,” adding, “But recently, there’s been a renewed sense of energy and determination — people are motivated again.”\n\nMeanwhile, Samsung’s faster-than-expected emergence as a “catfish” (disruptor) in NVIDIA’s HBM supply chain is expected to put pressure on SK hynix’s profitability. The previous-generation HBM3E 12-layer product is already projected to see prices fall by around 30% within a year. Currently, SK hynix’s HBM3E 12-layer chips are priced roughly 30% higher than Samsung’s.\n\nSamsung’s “HBM volume offensive” is thus likely to drive SK hynix’s product prices downward, bringing both companies’ prices to similar levels by the end of next year. “HBM4 will likely follow a similar trend,” one industry figure explained. “SK hynix will be able to maintain a price premium as the ‘sole vendor’ in the first half of next year, but from the second half onward, it’ll be difficult to sustain that margin.”\n\n$MU\n\n---\n\nhttps://t.co/y0o4tv4MIm", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002038693068074693, "ret_signed_m1d": -0.002038693068074693, "alpha_spy_m1d": 0.0082547258809893, "alpha_c_m1d": 0.00899567002942181, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.011034363097496502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0061162395860097085, "ret_signed_p1d": -0.0061162395860097085, "alpha_spy_p1d": -0.00610124366599718, "alpha_c_p1d": -0.0066366592900064525, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": 0.000520419703996744, "ret_p1w": 0.03975535683183096, "ret_signed_p1w": 0.03975535683183096, "alpha_spy_p1w": 0.018989665155993096, "alpha_c_p1w": 0.0022454244184331174, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03750993241339784, "ret_p1m": -0.0030580796975584468, "ret_signed_p1m": -0.0030580796975584468, "alpha_spy_p1m": 0.01365569610319628, "alpha_c_p1m": 0.028726492666849124, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.03178457236440757, "ret_p3m": 0.47400115025816647, "ret_signed_p3m": 0.47400115025816647, "alpha_spy_p3m": 0.4397608929428092, "alpha_c_p3m": 0.46299705029383453, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.011004099964331937, "ret_p6m": 1.119605678926571, "ret_signed_p6m": 1.119605678926571, "alpha_spy_p6m": 1.079214991319748, "alpha_c_p6m": 1.090086169346484, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.029519509580086867, "price_path": [-0.3261, -0.3302, -0.3312, -0.2856, -0.2835, -0.2632, -0.2754, -0.3008, -0.2927, -0.2906, -0.3018, -0.2845, -0.2713, -0.2795, -0.2784, -0.2703, -0.2734, -0.2734, -0.2896, -0.2896, -0.2845, -0.2835, -0.2754, -0.2744, -0.2866, -0.2835, -0.2937, -0.2927, -0.314, -0.2987, -0.2916, -0.2886, -0.2947, -0.2886, -0.2744, -0.2632, -0.2551, -0.2348, -0.2236, -0.1942, -0.2064, -0.1851, -0.1851, -0.1526, -0.1404, -0.1333, -0.1262, -0.1546, -0.1417, -0.1448, -0.1233, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0851, -0.0377, -0.0489, -0.0663, -0.0316, -0.0041, -0.002, 0.0, -0.0061, 0.0051, -0.0163, 0.0071, 0.0398, 0.0143, 0.0245, 0.0612, 0.0958, 0.1325, 0.0693, 0.0255, 0.0112, -0.002, 0.0255, 0.055, 0.051, 0.0479, -0.0092, 0.0255, -0.0031, -0.0163, 0.0255, -0.0336, -0.0143, 0.0122, 0.0479, 0.055, 0.0245, 0.0275, 0.054, 0.0652, 0.0714, 0.105, 0.1162, 0.105, 0.1009, 0.0938, 0.1101, 0.0683, 0.0479, 0.0999, 0.0968, 0.0836, 0.1264, 0.1366, 0.1325, 0.1325, 0.1927, 0.2241, 0.2282, 0.2282, 0.2282, 0.3163, 0.4146, 0.4228, 0.4443, 0.4218, 0.4238, 0.4218, 0.4095, 0.4371, 0.474, 0.5252, 0.5293, 0.4873, 0.5314, 0.56, 0.558, 0.558, 0.6338, 0.6635, 0.6461, 0.644, 0.5406, 0.7157, 0.7321, 0.6317, 0.6246, 0.7045, 0.6983, 0.7188, 0.8294, 0.8561, 0.8561, 0.8561, 0.8561, 0.9462, 0.9472, 0.9769, 1.0486, 1.0845, 1.233, 1.2177, 1.2177, 0.9985, 0.7639, 0.9626, 0.9278, 0.7772, 0.9247, 0.9462, 0.9247, 0.8796, 0.9329, 0.9862, 1.1357, 1.0538, 1.0425, 0.9083, 0.9431, 0.936, 0.8448, 0.8448, 0.8096, 0.7162, 0.9466, 0.8312, 0.9112, 0.9821, 1.017, 1.1607, 1.0939, 1.1145, 1.0632, 1.1196]}
{"unit_id": "thread:1980411935284949049", "ticker_idx": 1, "ts": "2025-10-20T23:12:54+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung's 'HBM volume offensive' is thus likely to drive SK hynix's product prices downward… SK hynix will be able to maintain a price premium as the 'sole vendor' in the first half of next year, but from the second half onward, it'll be difficult to sustain that margin", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung set to capture ~30% of NVIDIA HBM4 in H2 2026; bearish on SK Hynix margins and Micron's HBM4 competitiveness as Samsung disrupts the supply chain.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Expected to Rank Second in NVIDIA-Bound HBM4 Market Share… “Positive Shift in Samsung DS Division’s Morale”\n\nSamsung Electronics, which had long struggled in the high-bandwidth memory (HBM) market, is seeking a turnaround. Although it did not achieve notable success with the previous-generation HBM3E, industry observers say the atmosphere is shifting with the next-generation HBM4. In particular, visible collaboration with NVIDIA — the largest HBM customer — is positioning Samsung as a potential “game changer” capable of reshaping the market landscape.\n\nAccording to industry sources, Samsung is expected to secure the No. 2 position in NVIDIA’s HBM4 supply chain next year. While SK hynix is likely to maintain exclusivity in the first half, the competitive landscape is projected to be reorganized in the second half. Semiconductor industry insiders anticipate that SK hynix’s share of NVIDIA’s HBM4 orders will fall from 100% in the first half of next year to around 60% in the second half, while Samsung could capture up to 30% and Micron between 10–15%.\n\nGiven that Samsung’s share of NVIDIA’s HBM supply was previously at 0%, this represents remarkable growth. If these projections materialize, Samsung will not only break SK hynix’s near-monopoly but also widen the gap with competitor Micron.\n\nAn industry official stated, “NVIDIA has largely completed negotiations with SK hynix for first-half HBM4 supply,” adding, “Since SK hynix was the first to submit customer (CS) samples, it’s likely to complete qualification testing by year-end, making a monopoly structure inevitable for the first half.” The official continued, “However, from the second half onward, both Samsung and Micron will enter the supply chain, which will completely reshape the landscape.”\n\nThe fact that NVIDIA only finalized supply for the first half with SK hynix — the No. 1 HBM supplier — already signals a shift in its procurement strategy. Normally, NVIDIA signs one-year HBM supply contracts with vendors, but in the case of HBM4, the company is cautiously adjusting volumes and prices, anticipating new entrants.\n\nMicron, meanwhile, has found itself at a disadvantage after failing to meet NVIDIA’s requirements for higher HBM4 speeds. The company’s strategy of producing base dies in-house prioritized power efficiency, but that approach has limited performance improvement. Although Micron officially announced in its latest earnings release that the issue had been “resolved,” skepticism remains within the industry.\n\n“The base die issue Micron faces isn’t something that can be fixed in a month or two,” another industry source said. “It’s likely they were referring to samples that happened to perform better than average.”\n\nSamsung’s HBM4 yield still requires improvement but has shown progress compared to before. Another industry insider noted, “Samsung’s current 1c DRAM sample yield exceeds 70% for DDR5 and has reached about 50% for HBM4. That’s definitely an improvement over the past,” adding, “However, it’s still insufficient for full-scale mass production. To begin mass production of HBM, yields need to surpass at least 70%.”\n\nEven if Samsung is waging an “aggressive pricing offensive” for HBM in the second half, ramping up wafer input at this stage would sharply erode profitability. The key task is to improve yields before shifting to volume production. If the company can also resolve heat dissipation issues, its competitiveness could rise significantly. “Samsung’s HBM4 design is structurally advantageous for bandwidth expansion,” the insider said. “If it manages thermal control as well, it could become a very competitive chip.”\n\nWith these positive signals around HBM4, morale has reportedly improved within Samsung’s semiconductor (DS) division. Earlier this year, the team had been demoralized by falling behind competitors — to the extent that, during executive gatherings, conversation was often minimal.\n\nOne company insider shared, “Honestly, up until the first half, the department’s mood was terrible. Many engineers felt a sense of relative deprivation watching SK hynix perform so well,” adding, “But recently, there’s been a renewed sense of energy and determination — people are motivated again.”\n\nMeanwhile, Samsung’s faster-than-expected emergence as a “catfish” (disruptor) in NVIDIA’s HBM supply chain is expected to put pressure on SK hynix’s profitability. The previous-generation HBM3E 12-layer product is already projected to see prices fall by around 30% within a year. Currently, SK hynix’s HBM3E 12-layer chips are priced roughly 30% higher than Samsung’s.\n\nSamsung’s “HBM volume offensive” is thus likely to drive SK hynix’s product prices downward, bringing both companies’ prices to similar levels by the end of next year. “HBM4 will likely follow a similar trend,” one industry figure explained. “SK hynix will be able to maintain a price premium as the ‘sole vendor’ in the first half of next year, but from the second half onward, it’ll be difficult to sustain that margin.”\n\n$MU\n\n---\n\nhttps://t.co/y0o4tv4MIm", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04119462522047679, "ret_signed_m1d": 0.04119462522047679, "alpha_spy_m1d": 0.030901206271412796, "alpha_c_m1d": 0.028551282693331737, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013388283854285032, "ret_signed_p1d": 0.013388283854285032, "alpha_spy_p1d": 0.013373287934272504, "alpha_c_p1d": 0.007858595812623381, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.1019566925800015, "ret_signed_p1w": -0.1019566925800015, "alpha_spy_p1w": -0.08119100090416365, "alpha_c_p1w": -0.0653228619626689, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": 0.17404730285143066, "ret_signed_p1m": -0.17404730285143066, "alpha_spy_p1m": -0.1907610786521854, "alpha_c_p1m": -0.2141659301769061, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.5438442835073052, "ret_signed_p3m": -0.5438442835073052, "alpha_spy_p3m": -0.5096040261919479, "alpha_c_p3m": -0.39862376074724315, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 1.2777070106448454, "ret_signed_p6m": -1.2777070106448454, "alpha_spy_p6m": -1.2373163230380224, "alpha_c_p6m": -0.9718881072228924, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.4467, -0.4457, -0.4529, -0.4611, -0.4601, -0.458, -0.4375, -0.4694, -0.4694, -0.458, -0.4683, -0.4611, -0.4724, -0.4508, -0.4467, -0.4282, -0.4313, -0.4313, -0.4498, -0.4591, -0.4745, -0.4961, -0.4838, -0.4663, -0.4622, -0.4652, -0.447, -0.4459, -0.4727, -0.4634, -0.4593, -0.4531, -0.4367, -0.4295, -0.4068, -0.3738, -0.3677, -0.3234, -0.3182, -0.2832, -0.3131, -0.2678, -0.2678, -0.277, -0.2564, -0.2636, -0.2657, -0.3069, -0.2812, -0.2842, -0.2585, -0.1854, -0.1854, -0.1854, -0.1854, -0.1854, -0.1854, -0.1184, -0.1452, -0.1524, -0.1298, -0.068, -0.0412, 0.0, -0.0134, -0.0082, -0.0144, 0.0505, 0.102, 0.0731, 0.1493, 0.1699, 0.1514, 0.277, 0.207, 0.1926, 0.2214, 0.1946, 0.2482, 0.275, 0.2709, 0.2606, 0.1535, 0.2482, 0.174, 0.1576, 0.1761, 0.0731, 0.0711, 0.069, 0.0793, 0.1213, 0.0924, 0.1089, 0.1502, 0.1378, 0.1172, 0.1213, 0.1893, 0.1666, 0.2099, 0.1646, 0.1769, 0.1419, 0.0924, 0.1357, 0.1378, 0.1275, 0.1955, 0.2037, 0.212, 0.212, 0.2347, 0.3192, 0.3418, 0.3418, 0.3418, 0.3954, 0.4346, 0.4964, 0.5294, 0.5583, 0.5335, 0.5438, 0.5212, 0.5294, 0.5438, 0.5583, 0.5748, 0.5315, 0.5253, 0.5562, 0.5809, 0.517, 0.649, 0.7335, 0.7747, 0.8736, 0.7108, 0.8695, 0.8551, 0.7355, 0.7294, 0.8283, 0.8056, 0.7726, 0.8304, 0.8139, 0.8139, 0.8139, 0.8139, 0.8427, 0.9561, 0.9602, 1.0715, 1.0983, 1.2694, 1.191, 1.191, 0.939, 0.7532, 0.9432, 0.9081, 0.7263, 0.937, 0.9721, 0.9205, 0.8792, 1.0113, 1.0031, 1.1807, 1.0919, 1.0795, 0.9267, 1.0361, 1.0547, 0.9267, 0.9267, 0.8028, 0.6665, 0.8523, 0.714, 0.8089, 0.8296, 0.8915, 1.1332, 1.0609, 1.1208, 1.1476, 1.2777]}
{"unit_id": "thread:1980411935284949049", "ticker_idx": 2, "ts": "2025-10-20T23:12:54+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron, meanwhile, has found itself at a disadvantage after failing to meet NVIDIA's requirements for higher HBM4 speeds… skepticism remains within the industry", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung set to capture ~30% of NVIDIA HBM4 in H2 2026; bearish on SK Hynix margins and Micron's HBM4 competitiveness as Samsung disrupts the supply chain.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Expected to Rank Second in NVIDIA-Bound HBM4 Market Share… “Positive Shift in Samsung DS Division’s Morale”\n\nSamsung Electronics, which had long struggled in the high-bandwidth memory (HBM) market, is seeking a turnaround. Although it did not achieve notable success with the previous-generation HBM3E, industry observers say the atmosphere is shifting with the next-generation HBM4. In particular, visible collaboration with NVIDIA — the largest HBM customer — is positioning Samsung as a potential “game changer” capable of reshaping the market landscape.\n\nAccording to industry sources, Samsung is expected to secure the No. 2 position in NVIDIA’s HBM4 supply chain next year. While SK hynix is likely to maintain exclusivity in the first half, the competitive landscape is projected to be reorganized in the second half. Semiconductor industry insiders anticipate that SK hynix’s share of NVIDIA’s HBM4 orders will fall from 100% in the first half of next year to around 60% in the second half, while Samsung could capture up to 30% and Micron between 10–15%.\n\nGiven that Samsung’s share of NVIDIA’s HBM supply was previously at 0%, this represents remarkable growth. If these projections materialize, Samsung will not only break SK hynix’s near-monopoly but also widen the gap with competitor Micron.\n\nAn industry official stated, “NVIDIA has largely completed negotiations with SK hynix for first-half HBM4 supply,” adding, “Since SK hynix was the first to submit customer (CS) samples, it’s likely to complete qualification testing by year-end, making a monopoly structure inevitable for the first half.” The official continued, “However, from the second half onward, both Samsung and Micron will enter the supply chain, which will completely reshape the landscape.”\n\nThe fact that NVIDIA only finalized supply for the first half with SK hynix — the No. 1 HBM supplier — already signals a shift in its procurement strategy. Normally, NVIDIA signs one-year HBM supply contracts with vendors, but in the case of HBM4, the company is cautiously adjusting volumes and prices, anticipating new entrants.\n\nMicron, meanwhile, has found itself at a disadvantage after failing to meet NVIDIA’s requirements for higher HBM4 speeds. The company’s strategy of producing base dies in-house prioritized power efficiency, but that approach has limited performance improvement. Although Micron officially announced in its latest earnings release that the issue had been “resolved,” skepticism remains within the industry.\n\n“The base die issue Micron faces isn’t something that can be fixed in a month or two,” another industry source said. “It’s likely they were referring to samples that happened to perform better than average.”\n\nSamsung’s HBM4 yield still requires improvement but has shown progress compared to before. Another industry insider noted, “Samsung’s current 1c DRAM sample yield exceeds 70% for DDR5 and has reached about 50% for HBM4. That’s definitely an improvement over the past,” adding, “However, it’s still insufficient for full-scale mass production. To begin mass production of HBM, yields need to surpass at least 70%.”\n\nEven if Samsung is waging an “aggressive pricing offensive” for HBM in the second half, ramping up wafer input at this stage would sharply erode profitability. The key task is to improve yields before shifting to volume production. If the company can also resolve heat dissipation issues, its competitiveness could rise significantly. “Samsung’s HBM4 design is structurally advantageous for bandwidth expansion,” the insider said. “If it manages thermal control as well, it could become a very competitive chip.”\n\nWith these positive signals around HBM4, morale has reportedly improved within Samsung’s semiconductor (DS) division. Earlier this year, the team had been demoralized by falling behind competitors — to the extent that, during executive gatherings, conversation was often minimal.\n\nOne company insider shared, “Honestly, up until the first half, the department’s mood was terrible. Many engineers felt a sense of relative deprivation watching SK hynix perform so well,” adding, “But recently, there’s been a renewed sense of energy and determination — people are motivated again.”\n\nMeanwhile, Samsung’s faster-than-expected emergence as a “catfish” (disruptor) in NVIDIA’s HBM supply chain is expected to put pressure on SK hynix’s profitability. The previous-generation HBM3E 12-layer product is already projected to see prices fall by around 30% within a year. Currently, SK hynix’s HBM3E 12-layer chips are priced roughly 30% higher than Samsung’s.\n\nSamsung’s “HBM volume offensive” is thus likely to drive SK hynix’s product prices downward, bringing both companies’ prices to similar levels by the end of next year. “HBM4 will likely follow a similar trend,” one industry figure explained. “SK hynix will be able to maintain a price premium as the ‘sole vendor’ in the first half of next year, but from the second half onward, it’ll be difficult to sustain that margin.”\n\n$MU\n\n---\n\nhttps://t.co/y0o4tv4MIm", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021231312112164225, "ret_signed_m1d": 0.021231312112164225, "alpha_spy_m1d": 0.010937893163100232, "alpha_c_m1d": 0.008587969585019173, "bench_spy_m1d": -0.010293418949063993, "bench_c_m1d": -0.012643342527145052, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021666697809607394, "ret_signed_p1d": 0.021666697809607394, "alpha_spy_p1d": 0.021651701889594865, "alpha_c_p1d": 0.016137009767945742, "bench_spy_p1d": -1.4995920012528252e-05, "bench_c_p1d": -0.005529688041661651, "ret_p1w": 0.06446773378383996, "ret_signed_p1w": -0.06446773378383996, "alpha_spy_p1w": -0.043702042108002104, "alpha_c_p1w": -0.027833903166507357, "bench_spy_p1w": 0.02076569167583786, "bench_c_p1w": 0.03663383061733261, "ret_p1m": 0.10509257983862152, "ret_signed_p1m": -0.10509257983862152, "alpha_spy_p1m": -0.12180635563937625, "alpha_c_p1m": -0.14521120716409697, "bench_spy_p1m": -0.016713775800754727, "bench_c_p1m": -0.04011862732547544, "ret_p3m": 0.628698498100293, "ret_signed_p3m": -0.628698498100293, "alpha_spy_p3m": -0.5944582407849357, "alpha_c_p3m": -0.48347797534023096, "bench_spy_p3m": 0.03424025731535729, "bench_c_p3m": 0.14522052276006203, "ret_p6m": 1.2539237335723143, "ret_signed_p6m": -1.2539237335723143, "alpha_spy_p6m": -1.2135330459654914, "alpha_c_p6m": -0.9481048301503614, "bench_spy_p6m": 0.04039068760682296, "bench_c_p6m": 0.30581890342195295, "price_path": [-0.4692, -0.46, -0.4623, -0.4623, -0.4589, -0.4454, -0.4725, -0.4931, -0.4791, -0.4729, -0.4742, -0.4593, -0.4254, -0.402, -0.3826, -0.3994, -0.3944, -0.4158, -0.4029, -0.4101, -0.4335, -0.4404, -0.4312, -0.4373, -0.4369, -0.4309, -0.4103, -0.4248, -0.4248, -0.4274, -0.4262, -0.3997, -0.3651, -0.3646, -0.3463, -0.3233, -0.2723, -0.2401, -0.2375, -0.2324, -0.2267, -0.1837, -0.2135, -0.2043, -0.1957, -0.2184, -0.242, -0.2399, -0.2078, -0.1913, -0.1196, -0.1119, -0.0916, -0.0765, -0.1019, -0.0495, -0.0698, -0.1217, -0.0677, -0.0953, -0.0717, -0.0205, -0.0212, 0.0, -0.0217, -0.0401, -0.0003, 0.0592, 0.0645, 0.0732, 0.096, 0.0834, 0.0822, 0.1351, 0.0545, 0.1486, 0.1526, 0.1507, 0.225, 0.1661, 0.1844, 0.146, 0.1937, 0.1701, 0.1051, 0.0926, -0.0261, 0.0029, 0.083, 0.0859, 0.1136, 0.1136, 0.1437, 0.1629, 0.1582, 0.1325, 0.0961, 0.1473, 0.1942, 0.2208, 0.2754, 0.25, 0.1662, 0.1486, 0.1245, 0.0907, 0.2021, 0.2861, 0.3377, 0.3361, 0.3865, 0.3865, 0.3773, 0.4242, 0.4158, 0.3809, 0.3809, 0.5261, 0.5103, 0.6616, 0.6428, 0.5822, 0.6696, 0.6734, 0.636, 0.6128, 0.6287, 0.7551, 0.7551, 0.766, 0.8826, 0.9236, 0.9336, 0.8825, 0.9848, 1.106, 1.1085, 1.0073, 1.1182, 1.0294, 0.8356, 0.8525, 0.9096, 0.8555, 0.8059, 0.9853, 1.0029, 0.9917, 0.9917, 0.9342, 1.0367, 1.0192, 1.0716, 1.0368, 1.0224, 1.0756, 1.0106, 0.9951, 0.9966, 0.837, 0.939, 0.921, 0.7916, 0.8836, 0.9503, 1.0257, 0.9612, 1.0617, 1.1375, 1.2338, 1.234, 1.1495, 1.0461, 0.9563, 0.9137, 0.8486, 0.7198, 0.7283, 0.5576, 0.6352, 0.7805, 0.7727, 0.7727, 0.8285, 0.8276, 0.9687, 1.0402, 1.0358, 1.0647, 1.2539]}
{"unit_id": "reply:1997643767063441876", "ticker_idx": 0, "ts": "2025-12-07T12:26:03+00:00", "symbol": "Xiaomi", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If I have more capital, I'm going to buy Xiaomi. They're a truly excellent company.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Would buy Xiaomi; views it as a truly excellent company.", "resolved_tickers": ["1810.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "Xiaomi Corp listed on HKEX as 1810.HK", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@TacticzH If I have more capital, I’m going to buy Xiaomi.\nThey’re a truly excellent company.", "reply_text": null, "parent_text": "@jukan05 Thoughts on Xiaomi?", "quoted_text": null, "parent_author": "TacticzH", "ret_m1d": 0.004696968991117378, "ret_signed_m1d": 0.004696968991117378, "alpha_spy_m1d": 0.001683638595691317, "alpha_c_m1d": 0.011673855386105991, "bench_spy_m1d": 0.0030133303954260615, "bench_c_m1d": -0.006976886394988613, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03100055079280839, "ret_signed_p1d": -0.03100055079280839, "alpha_spy_p1d": -0.030137405825846786, "alpha_c_p1d": -0.033642431388926086, "bench_spy_p1d": -0.0008631449669616043, "bench_c_p1d": 0.002641880596117696, "ret_p1w": -0.017379089869533404, "ret_signed_p1w": -0.017379089869533404, "alpha_spy_p1w": -0.013136992297069106, "alpha_c_p1w": 0.01872460783050689, "bench_spy_p1w": -0.004242097572464298, "bench_c_p1w": -0.036103697700040294, "ret_p1m": -0.08971355907118483, "ret_signed_p1m": -0.08971355907118483, "alpha_spy_p1m": -0.10466932120807781, "alpha_c_p1m": -0.08458205062887869, "bench_spy_p1m": 0.01495576213689298, "bench_c_p1m": -0.005131508442306143, "ret_p3m": -0.24377643545672267, "ret_signed_p3m": -0.24377643545672267, "alpha_spy_p3m": -0.24332766831380437, "alpha_c_p3m": -0.19475270828038416, "bench_spy_p3m": -0.0004487671429183049, "bench_c_p3m": -0.049023727176338516, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.2905, 0.2835, 0.2952, 0.3199, 0.3257, 0.3586, 0.3351, 0.3316, 0.3175, 0.3023, 0.3363, 0.3962, 0.2835, 0.2576, 0.2682, 0.2682, 0.3105, 0.2917, 0.2647, 0.2647, 0.2647, 0.2529, 0.2224, 0.1527, 0.1419, 0.162, 0.1202, 0.0794, 0.1071, 0.0911, 0.0883, 0.0977, 0.0784, 0.0756, 0.055, 0.055, 0.0371, 0.0146, 0.0503, 0.0197, 0.0174, 0.0202, -0.008, -0.0052, 0.0094, 0.0263, 0.0216, -0.0052, -0.0146, -0.0423, -0.0883, -0.1146, -0.1057, -0.0921, -0.0526, -0.0582, -0.0348, -0.0366, -0.0535, -0.0442, -0.0554, -0.0141, 0.0047, 0.0, -0.031, -0.0188, -0.0094, 0.0089, -0.0174, -0.0395, -0.0319, -0.0559, -0.0479, -0.0653, -0.0794, -0.0794, -0.0794, -0.0794, -0.0939, -0.0756, -0.0794, -0.0794, -0.054, -0.0761, -0.0897, -0.1038, -0.1071, -0.1118, -0.0902, -0.108, -0.1127, -0.1109, -0.1287, -0.1433, -0.1667, -0.1682, -0.1724, -0.1489, -0.1729, -0.1649, -0.147, -0.14, -0.1663, -0.1766, -0.1874, -0.2024, -0.1799, -0.1738, -0.1733, -0.1644, -0.1287, -0.1423, -0.1348, -0.1428, -0.1428, -0.1428, -0.1428, -0.1696, -0.1414, -0.1606, -0.1639, -0.1738, -0.1804, -0.2217, -0.2583, -0.2485, -0.2438, -0.2151, -0.209, -0.2104, -0.217, -0.2179, -0.2175, -0.1733, -0.1696, -0.1747, -0.147, -0.2203, -0.2471, -0.2325, -0.2363, -0.2381, -0.225, -0.2395, -0.2541, -0.248, -0.2748, -0.2748, -0.2748, -0.2748, -0.2306, -0.2635, -0.2743, -0.2799, -0.2748, -0.2743, -0.2471, -0.2485, -0.241, -0.2391, -0.2532, -0.2677, -0.2673, -0.2696, -0.2973, -0.2922, -0.3185, -0.3185, -0.2724, -0.2846, -0.2762, -0.2691, -0.256, -0.2555, -0.2612, -0.2532, -0.255, -0.279, -0.2799, -0.2804, -0.2922, -0.3034, -0.2954, -0.2954, -0.3011, -0.333, -0.3293, -0.3415, null, null]}
{"unit_id": "thread:2003702435840426238", "ticker_idx": 0, "ts": "2025-12-24T05:41:02+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "With performance differing by more than 6x while price differs by only 1.3x, from this perspective, the analysis is that the H200's cost-effectiveness and attractiveness will rise significantly.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "H200 China supply confirmed at only 1.3x H20 price but 6.7x performance; author endorses the bullish read on NVDA demand uplift.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "China-bound H200: China supply essentially confirmed, yet only ~1.3x pricier than the H20…\n\nAccording to Chinese media reports, Nvidia’s H200 sales to China are said to be virtually confirmed.\n\nAlso, Nvidia CEO Jensen Huang (黄仁勋) is reportedly scheduled to visit China in January 2026.\n\nAccording to Chinese media, the H200 8-card module is priced at 1.4 million yuan (about $200,000), which is slightly more expensive than the H20.\n\nNvidia’s H20 initially came in a 96GB HBM3 memory version, and earlier this year it released an upgraded 141GB HBM3e version. The price of this 8-card module once rose from 880,000 yuan to 1.05 million yuan.\n\nWu Zihao (吴梓豪), CEO of Ronghe (蓉和) Semiconductor Consulting, stated that on a performance-per-density (TPP) basis, the next-generation “special supply” product H200 has a compute value of 15,832, about 6.7x that of the H20.\n\nHe added, “TPP performance density is a more objective comparison at the physical layer, whereas TFLOPS floating-point operations often produce comparison errors because manufacturers use different calculation methodologies.”\n\nWith performance differing by more than 6x while price differs by only 1.3x, from this perspective, the analysis is that the H200’s cost-effectiveness and attractiveness will rise significantly.\n\n$NVDA\n\n---\n\nhttps://t.co/p48Dp35ms2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0031812102126231867, "ret_signed_m1d": 0.0031812102126231867, "alpha_spy_m1d": 0.006686500238782811, "alpha_c_m1d": 0.005954666917963425, "bench_spy_m1d": -0.0035052900261596243, "bench_c_m1d": -0.002773456705340238, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.011187067134538653, "ret_signed_p1w": -0.011187067134538653, "alpha_spy_p1w": 0.0010670424126671962, "alpha_c_p1w": -9.332411353735193e-05, "bench_spy_p1w": -0.01225410954720585, "bench_c_p1w": -0.011093743021001301, "ret_p1m": -0.01998838842091266, "ret_signed_p1m": -0.01998838842091266, "alpha_spy_p1m": -0.017960531023568094, "alpha_c_p1m": -0.12612011940198098, "bench_spy_p1m": -0.002027857397344568, "bench_c_p1m": 0.10613173098106832, "ret_p3m": -0.0687158396113422, "ret_signed_p3m": -0.0687158396113422, "alpha_spy_p3m": -0.020611693465188763, "alpha_c_p3m": -0.1433786919091019, "bench_spy_p3m": -0.04810414614615344, "bench_c_p3m": 0.07466285229775971, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0553, -0.0359, -0.0108, -0.0073, 0.0014, -0.0053, -0.0163, -0.019, 0.0026, 0.0209, -0.0289, -0.0016, -0.0455, -0.0466, -0.0361, -0.0286, -0.0317, -0.0396, -0.0442, -0.0343, -0.0125, 0.0152, 0.0658, 0.0977, 0.0757, 0.0735, 0.0968, 0.0534, 0.0349, -0.0029, -0.0025, 0.0553, 0.0241, 0.0275, -0.0093, 0.0082, -0.0107, -0.0385, -0.0111, -0.0423, -0.0516, -0.0322, -0.0573, -0.0443, -0.0443, -0.0616, -0.0461, -0.038, -0.0479, -0.0277, -0.0329, -0.0162, -0.0193, -0.0256, -0.0407, -0.0721, -0.0653, -0.0577, -0.0937, -0.0767, -0.0404, -0.0261, 0.0032, 0.0, 0.0, 0.0102, -0.0021, -0.0057, -0.0112, -0.0112, 0.0013, -0.0026, -0.0073, 0.0027, -0.0189, -0.0199, -0.0195, -0.0148, -0.029, -0.0083, -0.0126, -0.0126, -0.0559, -0.028, -0.02, -0.005, -0.0113, -0.0005, 0.0154, 0.0207, 0.0134, -0.0159, -0.0438, -0.0765, -0.0887, -0.017, 0.0076, -0.0004, 0.0076, -0.0089, -0.0308, -0.0308, -0.0193, -0.0033, -0.0038, 0.0064, 0.0156, 0.0225, 0.0368, -0.0197, -0.0605, -0.0325, -0.0454, -0.0295, -0.0279, -0.0572, -0.0316, -0.0204, -0.0136, -0.0289, -0.0443, -0.0285, -0.0354, -0.0435, -0.0532, -0.0843, -0.0687, -0.071, -0.0526, -0.092, -0.1118, -0.1242, -0.0753, -0.0681, -0.0594, -0.0594, -0.0581, -0.0557, -0.0346, -0.0249, 0.0002, 0.0038, 0.0419, 0.0545, 0.0517, 0.0694, 0.0714, 0.0598, 0.0737, 0.0585, 0.1043, 0.1485, 0.1303, 0.1095, 0.0582, 0.0522, 0.0524, 0.0419, 0.102, 0.1214, 0.141, 0.1635, 0.1706, 0.1974, 0.2499, 0.1947, 0.1788, 0.1697, 0.1849, 0.1639, 0.1417, 0.1417, 0.1392, 0.1273, 0.136, 0.1195, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:1983067165776589189", "ticker_idx": 0, "ts": "2025-10-28T07:03:50+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Google (through Broadcom) has been raising its planned TSMC CoWoS wafer starts every month for next year", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish TSMC on rising Google/Broadcom CoWoS orders; singled out Guangsheng as the most undervalued Google supply-chain play at 80% revenue exposure.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Rumor: Meta is reportedly considering purchasing Google’s TPUs.\n\nRegardless of whether the deal materializes, Google (through Broadcom) has been raising its planned TSMC CoWoS wafer starts every month for next year.", "reply_text": null, "parent_text": null, "quoted_text": "#市場小作文 #RUMOR\n轉\n\n繼Anthropic之後，Meta也在考慮買Google TPU來用，姑且不論會不會成，Google （透過Broadcom）在台積明年CoWoS 的投片每個月都在上修。\n\n台股裡面Google supply chain最被低估的鑽石就是光聖，80%的營收是Google", "parent_author": null, "ret_m1d": 0.003389900542257518, "ret_signed_m1d": 0.003389900542257518, "alpha_spy_m1d": 0.0060388638550537, "alpha_c_m1d": 0.012120008420869577, "bench_spy_m1d": -0.002648963312796182, "bench_c_m1d": -0.008730107878612059, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020338986721973207, "ret_signed_p1d": 0.020338986721973207, "alpha_spy_p1d": 0.019858689023167342, "alpha_c_p1d": 0.005192057333734956, "bench_spy_p1d": 0.0004802976988058649, "bench_c_p1d": 0.01514692938823825, "ret_p1w": 0.020338986721973207, "ret_signed_p1w": 0.020338986721973207, "alpha_spy_p1w": 0.037542685657509844, "alpha_c_p1w": 0.04884268103330702, "bench_spy_p1w": -0.017203698935536638, "bench_c_p1w": -0.028503694311333816, "ret_p1m": -0.02372872065160203, "ret_signed_p1m": -0.02372872065160203, "alpha_spy_p1m": -0.012987330765120597, "alpha_c_p1m": 0.018682681635465048, "bench_spy_p1m": -0.010741389886481434, "bench_c_p1m": -0.04241140228706708, "ret_p3m": 0.20400000266580198, "ret_signed_p3m": 0.20400000266580198, "alpha_spy_p3m": 0.19787745778128518, "alpha_c_p3m": 0.09872907526045283, "bench_spy_p3m": 0.006122544884516801, "bench_c_p3m": 0.10527092740534916, "ret_p6m": 0.39901292014290024, "ret_signed_p6m": 0.39901292014290024, "alpha_spy_p6m": 0.3579689980976468, "alpha_c_p6m": 0.0817434900211953, "bench_spy_p6m": 0.04104392204525342, "bench_c_p6m": 0.31726943012170494, "price_path": [-0.2167, -0.2167, -0.2336, -0.2234, -0.2403, -0.2032, -0.2066, -0.2032, -0.2032, -0.1897, -0.2066, -0.2032, -0.2032, -0.1998, -0.2336, -0.2234, -0.2336, -0.2099, -0.2066, -0.1964, -0.2167, -0.2167, -0.2133, -0.2167, -0.2167, -0.2167, -0.2032, -0.2032, -0.1897, -0.1728, -0.1627, -0.1492, -0.1525, -0.1322, -0.1424, -0.1288, -0.1424, -0.122, -0.0915, -0.0915, -0.1051, -0.1186, -0.1186, -0.1153, -0.1017, -0.0746, -0.0508, -0.0508, -0.0271, -0.0407, -0.0237, -0.0237, -0.0407, -0.0339, -0.0068, 0.0068, -0.0169, 0.0034, 0.0034, -0.0102, -0.0169, -0.0169, 0.0034, 0.0, 0.0203, 0.0203, 0.0169, 0.0237, 0.0203, -0.0102, -0.0068, -0.0102, 0.0, -0.0068, 0.0, -0.0102, -0.0305, -0.0203, -0.0475, -0.0542, -0.0136, -0.061, -0.0678, -0.0407, -0.0237, -0.0271, -0.0237, -0.0441, -0.0305, -0.0169, -0.0203, -0.0102, 0.0136, 0.0034, 0.0203, -0.0001, 0.0067, -0.0137, -0.0239, -0.0273, -0.0273, -0.0273, -0.0035, 0.0135, 0.0169, 0.0169, 0.0271, 0.0407, 0.0339, 0.0544, 0.0544, 0.0782, 0.136, 0.1598, 0.1394, 0.1462, 0.1428, 0.1496, 0.1632, 0.1632, 0.1496, 0.1836, 0.1972, 0.2074, 0.1836, 0.1972, 0.204, 0.1938, 0.2108, 0.238, 0.2278, 0.2074, 0.2006, 0.2244, 0.2142, 0.2006, 0.2108, 0.2346, 0.2788, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.3026, 0.2924, 0.3366, 0.3707, 0.3571, 0.3571, 0.3434, 0.3162, 0.2686, 0.2924, 0.2856, 0.2312, 0.2584, 0.3196, 0.2822, 0.2686, 0.255, 0.2762, 0.3001, 0.2625, 0.2557, 0.2352, 0.2352, 0.2591, 0.2557, 0.2421, 0.2148, 0.2011, 0.2659, 0.2352, 0.2352, 0.2352, 0.2693, 0.3308, 0.3342, 0.3649, 0.3581, 0.4024, 0.4195, 0.4229, 0.3854, 0.382, 0.399, 0.399]}
{"unit_id": "orig:2003005234017362224", "ticker_idx": 0, "ts": "2025-12-22T07:30:36+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe Street estimates are still far too low", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish AVGO: Street estimates far too low despite Google COT risk debate; diversification into Anthropic/OpenAI mitigates reliance.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: AVGO / Google COT Debate\n\nFollowing our meeting with AVGO CEO/CFO immediately after earnings (here), we have been fielding questions this past week about risk – real and perceived – to AVGO’s Google relationship. \n\nThe debate is around a potential longer-term desire to move to a customer owned tooling (COT) relationship where Google owns the mask sets and AVGO moves to more of a IP block provider. Google’s engagement with Mediatek for\nv7e (here) is part of this debate, but the debate is broader than just this engagement. We ultimately feel that it will be very hard for Google to move away from AVGO, but in our meeting, even AVGO management acknowledges that the Google relationship may evolve over the coming years. AVGO is willing to work with customers to monetize its IP and may be open to arrangements where it is paid royalties per chip vs recognizing\nrevenue on the entire ASIC under the current arrangement. AVGO also highlighted its strong efforts to diversify its customer base (Anthropic, OpenAI, etc..) to mitigate such a heavy reliance on Google for its XPU business. To this point we think that AVGO's engagement for Anthropic's TPU rack orders (at relatively lower GMs of 45-50% vs XPU 55%) is a part of that strategy and serves to build the relationship in light of Anthropic's own XPU program. Ultimately, AVGO expects the AI model competition to continue to\ndrive investments in hardware an increasingly complex system-level innovations where AVGO can add value with its combination of XPU and networking portfolio. \n\nFurthermore, we expect other LLM makers to have less internal know-how and continue using AVGO for ASIC programs. Our opinion is that after the 7th generation of TPU, Google has built more advanced internal XPU capabilities, but is still heavily reliant on key IP blocks from AVGO (SerDes, etc..). However, we do believe that this debate will remain front and center for AVGO going forward and may cap the multiple even as we believe Street estimates are still far too low.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005096006677167697, "ret_signed_m1d": -0.005096006677167697, "alpha_spy_m1d": 0.0010953143381664532, "alpha_c_m1d": 0.007622959271596397, "bench_spy_m1d": -0.00619132101533415, "bench_c_m1d": -0.012718965948764094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02304873180654954, "ret_signed_p1d": 0.02304873180654954, "alpha_spy_p1d": 0.01847827698811666, "alpha_c_p1d": 0.013429628838360852, "bench_spy_p1d": 0.004570454818432879, "bench_c_p1d": 0.009619102968188686, "ret_p1w": 0.023253743569424, "ret_signed_p1w": 0.023253743569424, "alpha_spy_p1w": 0.01884397229413648, "alpha_c_p1w": 0.010715575387625131, "bench_spy_p1w": 0.004409771275287522, "bench_c_p1w": 0.01253816818179887, "ret_p1m": -0.025918896486793352, "ret_signed_p1m": -0.025918896486793352, "alpha_spy_p1m": -0.015332281801282654, "alpha_c_p1m": -0.11124000129604794, "bench_spy_p1m": -0.010586614685510698, "bench_c_p1m": 0.08532110480925459, "ret_p3m": -0.06328898517206971, "ret_signed_p3m": -0.06328898517206971, "alpha_spy_p3m": -0.02673973486178427, "alpha_c_p3m": -0.1612039143504198, "bench_spy_p3m": -0.03654925031028544, "bench_c_p3m": 0.0979149291783501, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0082, -0.0175, -0.0221, -0.0415, -0.0356, -0.0255, -0.0115, -0.0109, -0.0193, -0.0166, 0.0099, 0.0085, -0.0511, 0.0427, 0.0059, 0.027, 0.0352, 0.0211, 0.0209, 0.0016, -0.0053, 0.0064, 0.0352, 0.0583, 0.0902, 0.1283, 0.1005, 0.0805, 0.0598, 0.0288, 0.0493, 0.0394, 0.0214, 0.0476, 0.0288, 0.0383, -0.0062, 0.001, 0.0016, -0.0047, 0.036, 0.0138, -0.0056, 0.1048, 0.1255, 0.1621, 0.1621, 0.1779, 0.1285, 0.1154, 0.1126, 0.1138, 0.1407, 0.1725, 0.1876, 0.2071, 0.1879, 0.0521, -0.0067, -0.0023, -0.047, -0.0357, -0.0051, 0.0, 0.023, 0.0257, 0.0257, 0.0313, 0.0233, 0.0246, 0.0136, 0.0136, 0.0181, 0.0058, 0.0068, 0.006, -0.0263, 0.0103, 0.0315, 0.0385, -0.0046, 0.0046, 0.03, 0.03, -0.0259, -0.037, -0.0467, -0.0627, -0.0486, -0.0254, -0.024, -0.0314, -0.0297, -0.0303, -0.0619, -0.0978, -0.0906, -0.025, 0.0073, -0.003, 0.0038, -0.0301, -0.0477, -0.0477, -0.0261, -0.0233, -0.0218, -0.0258, -0.0325, -0.0467, -0.0268, -0.0578, -0.0641, -0.0663, -0.0809, -0.0701, -0.0254, -0.0321, 0.0126, 0.0033, 0.0004, -0.016, -0.0565, -0.0484, -0.059, -0.0747, -0.0633, -0.0906, -0.0535, -0.0659, -0.0643, -0.0919, -0.1176, -0.1389, -0.0916, -0.08, -0.0768, -0.0768, -0.0772, -0.0199, 0.029, 0.0416, 0.0904, 0.1145, 0.1175, 0.1643, 0.1694, 0.1931, 0.1728, 0.1803, 0.2404, 0.2325, 0.2407, 0.2273, 0.1734, 0.1899, 0.2251, 0.2364, 0.2224, 0.2542, 0.2486, 0.2108, 0.262, 0.2574, 0.2306, 0.2232, 0.2907, 0.2479, 0.2347, 0.2064, 0.2261, 0.2167, 0.2154, 0.2154, 0.2385, 0.2381, 0.2519, 0.3112, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2028366919523594590", "ticker_idx": 0, "ts": "2026-03-02T07:08:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND companies will record higher profit margins than Nvidia this year!🤯", "tweet_type": "prediction", "timeline": "months", "summary": "Bullish Samsung and SK Hynix on AI-driven NAND demand pushing their OPM above NVDA's this year.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "NAND companies will record higher profit margins than Nvidia this year!🤯 https://t.co/PVEHxqbFXs", "reply_text": null, "parent_text": null, "quoted_text": "Samsung &amp; SK NAND Operating Margins Forecast at 70%+ … AI-Driven Memory Rally Extends Beyond HBM Across the Board\n\nA surge in AI-driven memory demand is expected to push Samsung Electronics and SK Hynix's NAND operating profit margins (OPM) up to the 70% range this year. While https://t.co/lVXfw1HC1t", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0005682556842063757, "alpha_c_m1d": 0.0055898388543234034, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -0.0055898388543234034, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09884526208146449, "ret_signed_p1d": -0.09884526208146449, "alpha_spy_p1d": -0.09003092359866882, "alpha_c_p1d": -0.084225843558279, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.01461941852318549, "ret_p1w": -0.1986143434258696, "ret_signed_p1w": -0.1986143434258696, "alpha_spy_p1w": -0.18679876250235472, "alpha_c_p1w": -0.2001909506044468, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": 0.0015766071785772162, "ret_p1m": -0.2261151425649689, "ret_signed_p1m": -0.2261151425649689, "alpha_spy_p1m": -0.17619540098978514, "alpha_c_p1m": -0.17971438894673752, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.04640075361823137, "ret_p3m": 0.3862351363743528, "ret_signed_p3m": 0.3862351363743528, "alpha_spy_p3m": 0.28384175720844484, "alpha_c_p3m": 0.04552758791859035, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.34070754845576245, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5197, -0.5169, -0.5017, -0.4967, -0.5017, -0.5036, -0.5068, -0.4994, -0.5183, -0.5275, -0.504, -0.5054, -0.5114, -0.4921, -0.4875, -0.4893, -0.4893, -0.4622, -0.448, -0.4462, -0.4462, -0.4462, -0.4065, -0.3621, -0.3584, -0.3487, -0.3589, -0.358, -0.3589, -0.3644, -0.352, -0.3353, -0.3122, -0.3104, -0.3293, -0.3095, -0.2965, -0.2975, -0.2975, -0.2633, -0.2499, -0.2577, -0.2587, -0.3053, -0.2263, -0.2189, -0.2642, -0.2674, -0.2314, -0.2342, -0.2249, -0.1751, -0.163, -0.163, -0.163, -0.163, -0.1224, -0.1219, -0.1085, -0.0762, -0.06, 0.0069, 0.0, 0.0, -0.0988, -0.2046, -0.115, -0.1307, -0.1986, -0.1321, -0.1224, -0.1321, -0.1524, -0.1284, -0.1044, -0.037, -0.0739, -0.079, -0.1395, -0.1238, -0.127, -0.1681, -0.1681, -0.184, -0.2261, -0.1222, -0.1743, -0.1382, -0.1062, -0.0905, -0.0257, -0.0558, -0.0465, -0.0697, -0.0442, -0.0234, 0.0067, -0.0002, -0.0072, 0.0136, 0.0067, 0.0391, 0.016, 0.0391, 0.0275, 0.046, 0.0206, 0.0206, 0.0761, 0.0761, 0.2312, 0.2566, 0.2428, 0.3214, 0.2914, 0.3145, 0.37, 0.252, 0.3006, 0.2752, 0.2775, 0.3862, 0.3538, 0.3538, 0.3839, 0.4209, 0.3862, 0.4672, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2028366919523594590", "ticker_idx": 1, "ts": "2026-03-02T07:08:53+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND companies will record higher profit margins than Nvidia this year!🤯", "tweet_type": "prediction", "timeline": "months", "summary": "Bullish Samsung and SK Hynix on AI-driven NAND demand pushing their OPM above NVDA's this year.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "NAND companies will record higher profit margins than Nvidia this year!🤯 https://t.co/PVEHxqbFXs", "reply_text": null, "parent_text": null, "quoted_text": "Samsung &amp; SK NAND Operating Margins Forecast at 70%+ … AI-Driven Memory Rally Extends Beyond HBM Across the Board\n\nA surge in AI-driven memory demand is expected to push Samsung Electronics and SK Hynix's NAND operating profit margins (OPM) up to the 70% range this year. While https://t.co/lVXfw1HC1t", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0005682556842063757, "alpha_c_m1d": 4.926186798004384e-05, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -4.926186798004384e-05, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.11498588182920777, "ret_signed_p1d": -0.11498588182920777, "alpha_spy_p1d": -0.1061715433464121, "alpha_c_p1d": -0.07726345623490394, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.03772242559430383, "ret_p1w": -0.21206408092964646, "ret_signed_p1w": -0.21206408092964646, "alpha_spy_p1w": -0.2002485000061316, "alpha_c_p1w": -0.1824865344890717, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.029577546440574753, "ret_p1m": -0.23939681952013425, "ret_signed_p1m": -0.23939681952013425, "alpha_spy_p1m": -0.1894770779449505, "alpha_c_p1m": -0.18282549659567338, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.056571322924460876, "ret_p3m": 1.157759408647605, "ret_signed_p3m": 1.157759408647605, "alpha_spy_p3m": 1.055366029481697, "alpha_c_p3m": 0.6817634553298333, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.4759959533177718, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4807, -0.4901, -0.4882, -0.4572, -0.4675, -0.4478, -0.4685, -0.4628, -0.4788, -0.5014, -0.4816, -0.4807, -0.4854, -0.4544, -0.4506, -0.4468, -0.4468, -0.4365, -0.3979, -0.3876, -0.3876, -0.3876, -0.3631, -0.3452, -0.317, -0.3019, -0.2888, -0.3001, -0.2954, -0.3057, -0.3019, -0.2954, -0.2888, -0.2813, -0.301, -0.3038, -0.2897, -0.2784, -0.3076, -0.2474, -0.2088, -0.19, -0.1448, -0.2192, -0.1467, -0.1533, -0.2079, -0.2107, -0.1655, -0.1759, -0.1909, -0.1646, -0.1721, -0.1721, -0.1721, -0.1721, -0.159, -0.1072, -0.1053, -0.0545, -0.0423, 0.0358, 0.0, 0.0, -0.115, -0.1998, -0.1131, -0.1291, -0.2121, -0.1159, -0.0999, -0.1235, -0.1423, -0.082, -0.0858, -0.0047, -0.0452, -0.0509, -0.1206, -0.0707, -0.0622, -0.1206, -0.1206, -0.1772, -0.2394, -0.1546, -0.2177, -0.1744, -0.1649, -0.1367, -0.0264, -0.0594, -0.032, -0.0198, 0.0396, 0.0707, 0.0886, 0.0631, 0.099, 0.1536, 0.1527, 0.1546, 0.1517, 0.2177, 0.2253, 0.2187, 0.2121, 0.2121, 0.3638, 0.3638, 0.509, 0.5589, 0.5891, 0.7719, 0.7295, 0.8624, 0.8567, 0.7144, 0.7342, 0.6447, 0.6447, 0.8285, 0.8294, 0.8294, 0.934, 1.114, 1.1578, 1.1992, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2019567733025960409", "ticker_idx": 0, "ts": "2026-02-06T00:24:03+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's basically a done deal that both Samsung and Hynix are going to make an absolute ton of money this year", "tweet_type": "prediction", "timeline": "months", "summary": "Bullish Samsung and SK Hynix on massive earnings ahead; sell-side estimates not yet fully reflecting SSD price increases.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Honestly, lately when I look at the operating profit estimates for Samsung and SK Hynix coming out from sell-side analysts, it feels like they’re not fully reflecting the SSD price increases.\n\nAnyway, it’s basically a done deal that both Samsung and Hynix are going to make an absolute ton of money this year.\n\nAmong the current numbers floating around, the median seems to be this KB Securities forecast, and according to it, Samsung’s operating profit in 2026 would rank as the 6th highest in the world.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004413573354641542, "ret_signed_m1d": 0.004413573354641542, "alpha_spy_m1d": 0.02323716010791821, "alpha_c_m1d": 0.043384692507755784, "bench_spy_m1d": -0.018823586753276667, "bench_c_m1d": -0.03897111915311424, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04918033919762932, "ret_signed_p1d": 0.04918033919762932, "alpha_spy_p1d": 0.04435850709461886, "alpha_c_p1d": 0.033450079963909696, "bench_spy_p1d": 0.004821832103010459, "bench_c_p1d": 0.015730259233719623, "ret_p1w": 0.1424968068304704, "ret_signed_p1w": 0.1424968068304704, "alpha_spy_p1w": 0.15534031506757862, "alpha_c_p1w": 0.15362132929894434, "bench_spy_p1w": -0.012843508237108225, "bench_c_p1w": -0.011124522468473952, "ret_p1m": 0.09394700631841824, "ret_signed_p1m": 0.09394700631841824, "alpha_spy_p1m": 0.11182937385561598, "alpha_c_p1m": 0.10365438734013277, "bench_spy_p1m": -0.01788236753719774, "bench_c_p1m": -0.009707381021714534, "ret_p3m": 0.6806467172795636, "ret_signed_p3m": 0.6806467172795636, "alpha_spy_p3m": 0.6151777914268302, "alpha_c_p3m": 0.47437301661456477, "bench_spy_p3m": 0.06546892585273345, "bench_c_p3m": 0.20627370066499884, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3506, -0.3531, -0.355, -0.3901, -0.3688, -0.3863, -0.3945, -0.3688, -0.4052, -0.3932, -0.3769, -0.355, -0.3506, -0.3694, -0.3675, -0.3512, -0.3443, -0.3405, -0.3198, -0.3129, -0.3198, -0.3223, -0.3267, -0.3167, -0.3424, -0.355, -0.323, -0.3248, -0.333, -0.3066, -0.3004, -0.3029, -0.3029, -0.2659, -0.2465, -0.244, -0.244, -0.244, -0.1898, -0.1293, -0.1242, -0.111, -0.1248, -0.1236, -0.1248, -0.1324, -0.1154, -0.0927, -0.0612, -0.0586, -0.0845, -0.0574, -0.0397, -0.041, -0.041, 0.0057, 0.024, 0.0132, 0.012, -0.0517, 0.0561, 0.0662, 0.0044, 0.0, 0.0492, 0.0454, 0.058, 0.1261, 0.1425, 0.1425, 0.1425, 0.1425, 0.198, 0.1986, 0.2169, 0.261, 0.2831, 0.3745, 0.3651, 0.3651, 0.2301, 0.0858, 0.2081, 0.1866, 0.0939, 0.1847, 0.198, 0.1847, 0.157, 0.1898, 0.2226, 0.3146, 0.2642, 0.2573, 0.1747, 0.1961, 0.1917, 0.1356, 0.1356, 0.1139, 0.0564, 0.1983, 0.1272, 0.1765, 0.22, 0.2415, 0.33, 0.2889, 0.3016, 0.27, 0.3047, 0.3331, 0.3742, 0.3647, 0.3553, 0.3837, 0.3742, 0.4184, 0.3868, 0.4184, 0.4026, 0.4279, 0.3932, 0.3932, 0.469, 0.469, 0.6806, 0.7154, 0.6964, 0.8039, 0.7628, 0.7944, 0.8702, 0.7091, 0.7754, 0.7407, 0.7438, 0.8923, 0.8481, 0.8481, 0.8891, 0.9397, 0.8923, 1.0029, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2019567733025960409", "ticker_idx": 1, "ts": "2026-02-06T00:24:03+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's basically a done deal that both Samsung and Hynix are going to make an absolute ton of money this year", "tweet_type": "prediction", "timeline": "months", "summary": "Bullish Samsung and SK Hynix on massive earnings ahead; sell-side estimates not yet fully reflecting SSD price increases.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Honestly, lately when I look at the operating profit estimates for Samsung and SK Hynix coming out from sell-side analysts, it feels like they’re not fully reflecting the SSD price increases.\n\nAnyway, it’s basically a done deal that both Samsung and Hynix are going to make an absolute ton of money this year.\n\nAmong the current numbers floating around, the median seems to be this KB Securities forecast, and according to it, Samsung’s operating profit in 2026 would rank as the 6th highest in the world.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003575716567135956, "ret_signed_m1d": 0.003575716567135956, "alpha_spy_m1d": 0.022399303320412622, "alpha_c_m1d": 0.054814324521826, "bench_spy_m1d": -0.018823586753276667, "bench_c_m1d": -0.05123860795469004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05721101721388355, "ret_signed_p1d": 0.05721101721388355, "alpha_spy_p1d": 0.05238918511087309, "alpha_c_p1d": 0.044762367703163886, "bench_spy_p1d": 0.004821832103010459, "bench_c_p1d": 0.012448649510719667, "ret_p1w": 0.04886767855723306, "ret_signed_p1w": 0.04886767855723306, "alpha_spy_p1w": 0.06171118679434129, "alpha_c_p1w": 0.03375499981589303, "bench_spy_p1w": -0.012843508237108225, "bench_c_p1w": 0.015112678741340035, "ret_p1m": -0.0017370261404855425, "ret_signed_p1m": -0.0017370261404855425, "alpha_spy_p1m": 0.0161453413967122, "alpha_c_p1m": 0.016388204507901127, "bench_spy_p1m": -0.01788236753719774, "bench_c_p1m": -0.01812523064838667, "ret_p3m": 0.9117452481090031, "ret_signed_p3m": 0.9117452481090031, "alpha_spy_p3m": 0.8462763222562697, "alpha_c_p3m": 0.5429913127925932, "bench_spy_p3m": 0.06546892585273345, "bench_c_p3m": 0.36875393531641, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2627, -0.2651, -0.2711, -0.333, -0.2782, -0.3211, -0.3306, -0.3199, -0.3795, -0.3807, -0.3818, -0.3759, -0.3516, -0.3683, -0.3588, -0.3349, -0.3421, -0.354, -0.3516, -0.3123, -0.3254, -0.3004, -0.3266, -0.3194, -0.3397, -0.3683, -0.3433, -0.3421, -0.348, -0.3087, -0.3039, -0.2992, -0.2992, -0.2861, -0.2372, -0.2241, -0.2241, -0.2241, -0.1931, -0.1704, -0.1347, -0.1156, -0.0989, -0.1132, -0.1073, -0.1204, -0.1156, -0.1073, -0.0989, -0.0894, -0.1144, -0.118, -0.1001, -0.0858, -0.1228, -0.0465, 0.0024, 0.0262, 0.0834, -0.0107, 0.081, 0.0727, 0.0036, 0.0, 0.0572, 0.0441, 0.025, 0.0584, 0.0489, 0.0489, 0.0489, 0.0489, 0.0656, 0.1311, 0.1335, 0.1979, 0.2133, 0.3123, 0.2669, 0.2669, 0.1213, 0.0138, 0.1236, 0.1033, -0.0017, 0.1201, 0.1404, 0.1105, 0.0866, 0.163, 0.1583, 0.261, 0.2096, 0.2025, 0.1141, 0.1774, 0.1881, 0.1141, 0.1141, 0.0424, -0.0364, 0.0711, -0.0089, 0.046, 0.058, 0.0938, 0.2335, 0.1917, 0.2263, 0.2419, 0.3171, 0.3565, 0.3792, 0.3469, 0.3923, 0.4616, 0.4604, 0.4628, 0.4592, 0.5428, 0.5523, 0.544, 0.5356, 0.5356, 0.7279, 0.7279, 0.9117, 0.975, 1.0132, 1.2449, 1.1912, 1.3595, 1.3524, 1.1721, 1.1971, 1.0837, 1.0837, 1.3165, 1.3177, 1.3177, 1.4503, 1.6784, 1.7337, 1.7863, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2005183355990802874", "ticker_idx": 0, "ts": "2025-12-28T07:45:41+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom's revenue from Google in 2026 is expected to roughly double versus 2025, and in 2027 it could grow again to nearly double compared with 2026", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Citi's 2026-27 AVGO outlook showing accelerating ASIC revenue from Google, Meta, xAI, OpenAI, AWS, and Microsoft; implied bullish.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I’m only just getting around to reading Citi’s 2026 semiconductor outlook… and this table is genuinely fascinating.\n\n1. Broadcom’s revenue from Google in 2026 is expected to roughly double versus 2025, and in 2027 it could grow again to nearly double compared with 2026.\n\n2. Meta-related revenue likely starts to show up meaningfully in 2027.\n\n3. ByteDance’s ASIC shipments look small through 2026, but it seems to enter true volume production toward late 2026.\n\n4. Revenue from Anthropic could be nearly on par with Google in 2026, but then drop sharply in 2027.\n\n5. xAI’s ASIC really starts to take off in 2027.\n\n6. OpenAI’s ASIC appears to start mass production in late 2026 and see large-scale deployment in 2027.\n\n7. Broadcom seems to have secured meaningful orders from AWS and Microsoft, with major revenue upside expected in 2027.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007842206363219306, "ret_signed_m1d": 0.007842206363219306, "alpha_spy_m1d": 0.004265833292209642, "alpha_c_m1d": 0.0033118696475999965, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.00453033671561931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0013165166515154425, "ret_signed_p1d": 0.0013165166515154425, "alpha_spy_p1d": 0.0025376173860977014, "alpha_c_p1d": 0.0037876017881113277, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0024710851365958852, "ret_p1w": -0.017086882396489145, "ret_signed_p1w": -0.017086882396489145, "alpha_spy_p1w": -0.01689789828396704, "alpha_c_p1w": -0.05382394397285384, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": 0.036737061576364693, "ret_p1m": -0.047511362516396316, "ret_signed_p1m": -0.047511362516396316, "alpha_spy_p1m": -0.0586184499544693, "alpha_c_p1m": -0.16568496992614934, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.11817360740975302, "ret_p3m": -0.11254164826379276, "ret_signed_p3m": -0.11254164826379276, "alpha_spy_p3m": -0.05293824797337654, "alpha_c_p3m": -0.15820213334925926, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": 0.045660485085466496, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0476, -0.0339, -0.0334, -0.0416, -0.039, -0.013, -0.0144, -0.0726, 0.019, -0.0169, 0.0036, 0.0117, -0.0021, -0.0023, -0.0211, -0.0279, -0.0165, 0.0116, 0.0343, 0.0655, 0.1026, 0.0754, 0.0559, 0.0357, 0.0054, 0.0255, 0.0158, -0.0018, 0.0238, 0.0054, 0.0147, -0.0288, -0.0217, -0.0212, -0.0273, 0.0125, -0.0093, -0.0282, 0.0797, 0.0999, 0.1357, 0.1357, 0.1511, 0.1029, 0.09, 0.0873, 0.0885, 0.1148, 0.1458, 0.1606, 0.1797, 0.1609, 0.0282, -0.0293, -0.025, -0.0687, -0.0576, -0.0277, -0.0227, -0.0002, 0.0024, 0.0024, 0.0078, 0.0, 0.0013, -0.0094, -0.0094, -0.0051, -0.0171, -0.0161, -0.0169, -0.0484, -0.0127, 0.0081, 0.0149, -0.0272, -0.0182, 0.0066, 0.0066, -0.0481, -0.0589, -0.0684, -0.084, -0.0702, -0.0475, -0.0462, -0.0534, -0.0518, -0.0523, -0.0832, -0.1183, -0.1113, -0.0471, -0.0156, -0.0256, -0.019, -0.0521, -0.0693, -0.0693, -0.0482, -0.0455, -0.0441, -0.0479, -0.0545, -0.0684, -0.0489, -0.0793, -0.0854, -0.0875, -0.1017, -0.0912, -0.0476, -0.0541, -0.0104, -0.0195, -0.0224, -0.0384, -0.0779, -0.07, -0.0804, -0.0958, -0.0846, -0.1113, -0.075, -0.0871, -0.0856, -0.1125, -0.1376, -0.1585, -0.1123, -0.1009, -0.0978, -0.0978, -0.0982, -0.0421, 0.0057, 0.0179, 0.0657, 0.0892, 0.0921, 0.1378, 0.1429, 0.166, 0.1462, 0.1535, 0.2122, 0.2044, 0.2125, 0.1995, 0.1468, 0.1629, 0.1972, 0.2083, 0.1946, 0.2257, 0.2202, 0.1833, 0.2333, 0.2288, 0.2026, 0.1954, 0.2614, 0.2195, 0.2067, 0.179, 0.1982, 0.189, 0.1878, 0.1878, 0.2104, 0.21, 0.2235, 0.2814, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007080754896048134", "ticker_idx": 0, "ts": "2026-01-02T13:25:16+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom's CoWoS allocation has been increased by ~20k in recent month and hence its 2026 total TPU volume has also risen from 4mn to 4.3mn", "tweet_type": "analysis", "timeline": "quarters", "summary": "Broadcom's 2026 CoWoS allocation raised ~20k and TPU volume up 4mn→4.3mn; bullish supply read.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "\"There’s recent rumor lately that Broadcom’s 2026 CoWoS allocation was cut meaningfully in order to compensate for NVDA’s CoWoS demand. This is apparently wrong…In fact, Broadcom’s CoWoS allocation has been increased by ~20k in recent month and hence its 2026 total TPU volume has also risen from 4mn to 4.3mn; Instead, AWS 2026 CoWoS allocation is cut in order to compensate for NVDA’s strong CoWoS demand.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004372494439369823, "ret_signed_m1d": -0.004372494439369823, "alpha_spy_m1d": -0.002542808592774559, "alpha_c_m1d": 0.030907396642384732, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.035279891081754555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01208202680627124, "ret_signed_p1d": -0.01208202680627124, "alpha_spy_p1d": -0.01874214062233437, "alpha_c_p1d": -0.02357414758100085, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.011492120774729608, "ret_p1w": -0.007623229661879005, "ret_signed_p1w": -0.007623229661879005, "alpha_spy_p1w": -0.023578244330541454, "alpha_c_p1w": -0.050269931907590326, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.04264670224571132, "ret_p1m": -0.047494375588492566, "ret_signed_p1m": -0.047494375588492566, "alpha_spy_p1m": -0.06541085576886674, "alpha_c_p1m": -0.1403687783783535, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.09287440278986092, "ret_p3m": -0.09629015374109906, "ret_signed_p3m": -0.09629015374109906, "alpha_spy_p3m": -0.05802657076040252, "alpha_c_p3m": -0.14630358539995225, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": 0.05001343165885319, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0341, -0.008, -0.0094, -0.0679, 0.0242, -0.0119, 0.0087, 0.0168, 0.003, 0.0027, -0.0162, -0.0229, -0.0115, 0.0168, 0.0395, 0.0709, 0.1082, 0.0809, 0.0613, 0.041, 0.0105, 0.0307, 0.021, 0.0033, 0.029, 0.0106, 0.0199, -0.0238, -0.0167, -0.0162, -0.0224, 0.0176, -0.0042, -0.0232, 0.0852, 0.1055, 0.1415, 0.1415, 0.157, 0.1085, 0.0956, 0.0928, 0.094, 0.1205, 0.1516, 0.1665, 0.1857, 0.1668, 0.0334, -0.0243, -0.0201, -0.0639, -0.0528, -0.0228, -0.0177, 0.0049, 0.0075, 0.0075, 0.013, 0.0051, 0.0064, -0.0044, -0.0044, 0.0, -0.0121, -0.0111, -0.0119, -0.0436, -0.0076, 0.0132, 0.0201, -0.0222, -0.0132, 0.0118, 0.0118, -0.0432, -0.0541, -0.0637, -0.0793, -0.0655, -0.0427, -0.0414, -0.0486, -0.0469, -0.0475, -0.0785, -0.1138, -0.1068, -0.0423, -0.0106, -0.0207, -0.014, -0.0473, -0.0646, -0.0646, -0.0434, -0.0406, -0.0392, -0.0431, -0.0497, -0.0637, -0.044, -0.0746, -0.0807, -0.0828, -0.0972, -0.0866, -0.0427, -0.0493, -0.0054, -0.0145, -0.0174, -0.0335, -0.0732, -0.0653, -0.0757, -0.0912, -0.0799, -0.1068, -0.0703, -0.0825, -0.081, -0.108, -0.1332, -0.1542, -0.1078, -0.0963, -0.0932, -0.0932, -0.0936, -0.0373, 0.0108, 0.0231, 0.0711, 0.0947, 0.0977, 0.1436, 0.1487, 0.1719, 0.152, 0.1594, 0.2184, 0.2106, 0.2187, 0.2056, 0.1526, 0.1688, 0.2033, 0.2144, 0.2007, 0.232, 0.2264, 0.1893, 0.2396, 0.2351, 0.2087, 0.2015, 0.2678, 0.2257, 0.2128, 0.185, 0.2043, 0.1951, 0.1939, 0.1939, 0.2165, 0.2161, 0.2297, 0.2879, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2019286921839800664", "ticker_idx": 0, "ts": "2026-02-05T05:48:13+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung recently passed the qual test for NVIDIA—the world's leading AI semiconductor company—beating its competitors to become the first to do so, and is now ahead of mass production shipments. As technological recovery and supply shortages coincide this year, the company is expected to proceed with facility investments quite aggressively", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish Samsung: first to pass NVIDIA HBM4 qual, 18% DRAM capacity expansion underway, AI-driven memory supercycle and supply tightening support aggressive investment.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung to Expand DRAM Capacity for HBM4 by 120,000 Wafers per Month\n\nSamsung Electronics is embarking on an initiative to increase its new DRAM production capacity by nearly 20% this year, centered around its Pyeongtaek Plant 4 (P4). The core of this expansion is to significantly boost the production of 6th-generation 10nm-class (1c) DRAM, which serves as the primary material for 6th-generation High Bandwidth Memory (HBM4). This move is interpreted as a strategic plan for aggressive facility investment, driven by a sharp recovery in the company's HBM4 technical competitiveness and the ever-increasing demand for DRAM from Artificial Intelligence (AI) customers.\n\nAccording to industry sources on the 5th, Samsung Electronics has established a strategy to build a new DRAM production line at 'P4' in Pyeongtaek, Gyeonggi Province—the company's latest semiconductor factory—capable of producing 100,000 to 120,000 wafers per month by the first quarter of next year. This line will specifically house the 1c DRAM production infrastructure used for HBM4.\n\nHBM is a type of memory created by vertically stacking multiple DRAMs. It is faster and has a larger capacity than single DRAM units, making it a highly sought-after product in the AI industry, which requires complex computations.\n\nHBM4 is the latest HBM product that Samsung Electronics will supply to AI semiconductor companies such as NVIDIA and AMD starting this year. Samsung produces HBM4 by vertically stacking 12 units of 1c DRAM, the most advanced product currently available.\n\nIndustry experts evaluate Samsung’s establishment of a new 1c DRAM line for HBM4 as a substantial investment. To date, Samsung Electronics possesses lines capable of producing 660,000 DRAM wafers per month. Adding up to 120,000 wafers of new capacity would mean a maximum 18% increase in DRAM production capability within a single year.\n\nThe production ratio of 1c DRAM for HBM4 is also expected to rise. Samsung Electronics is known to have established a production line for 60,000 to 70,000 1c DRAM wafers monthly as of last year. If the lines are completed as planned, the capacity for 1c DRAM for HBM4 will reach nearly 200,000 wafers per month, meaning HBM4-specific DRAM alone will account for 25% of the total production process.\n\nFurthermore, the 4nm process line at S5, the Pyeongtaek foundry plant that manufactures HBM4 base dies, is reportedly operating in \"full-throttle\" mode. Construction on the exterior of 'P5,' currently being built in Pyeongtaek, is also moving busily to align with the timing for semiconductor manufacturing equipment delivery as early as the first quarter of next year.\n\nThe company is also reportedly considering expanding the production of general-purpose 1c DRAM used in mobile devices and home appliances. This involves a large-scale process transition to 1c DRAM, primarily at Hwaseong Line 17, where previous generation processes were located.\n\nSamsung Electronics’ decision to significantly strengthen its 1c DRAM production capacity this year stems from growing confidence in its latest technology and the arrival of a memory \"super cycle\" driven by the expansion of the AI market.\n\nThe atmosphere within Samsung Electronics has recently seen a surge in technical confidence. Until last year, the company struggled significantly with the manufacturing of HBM4 and 1c DRAM. Consequently, Vice Chairman Jun Young-hyun, who was appointed as the new head of the Device Solutions (DS) division in 2024, decided on a redesign of the 1c DRAM. This strategy proved successful, leading to a rapid recovery in technology since the first half of last year.\n\nIn the HBM4 sector specifically, Samsung recently passed the \"qual test\" (qualification test) for NVIDIA—the world's leading AI semiconductor company—beating its competitors to become the first to do so, and is now ahead of mass production shipments. Faced with a situation where HBM4 production must be scaled up rapidly, the company has entered the stage of building new lines centered on P4, its most advanced facility.\n\nAdditionally, the company must increase production capacity to respond to tightening memory supply chains caused by the AI boom. According to market research firm Counterpoint Research, the price of 64GB DDR5 DRAM for servers rose by 98% in the first quarter of this year compared to the fourth quarter of last year, marking a significant price hike.\n\nAn industry official stated, \"Samsung Electronics possesses abundant production capacity and capital, allowing it to respond relatively flexibly to surging demand. As technological recovery and supply shortages coincide this year, the company is expected to proceed with facility investments quite aggressively.\"\n\n---\n\nhttps://t.co/Eo1kzeltQg", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.061519198524966345, "ret_signed_m1d": 0.061519198524966345, "alpha_spy_m1d": 0.04887203416123409, "alpha_c_m1d": 0.04316054118363244, "bench_spy_m1d": 0.012647164363732255, "bench_c_m1d": 0.018358657341333906, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004394179321871028, "ret_signed_p1d": -0.004394179321871028, "alpha_spy_p1d": -0.023578891162822413, "alpha_c_p1d": -0.04494563404898788, "bench_spy_p1d": 0.019184711840951385, "bench_c_p1d": 0.040551454727116854, "ret_p1w": 0.12115509309336292, "ret_signed_p1w": 0.12115509309336292, "alpha_spy_p1w": 0.11576856871617669, "alpha_c_p1w": 0.09475970063976336, "bench_spy_p1w": 0.00538652437718623, "bench_c_p1w": 0.026395392453599564, "ret_p1m": 0.1814187556472826, "ret_signed_p1m": 0.1814187556472826, "alpha_spy_p1m": 0.18915176308621195, "alpha_c_p1m": 0.16917965075305985, "bench_spy_p1m": -0.007733007438929351, "bench_c_p1m": 0.012239104894222752, "ret_p3m": 0.4625313331120533, "ret_signed_p3m": 0.4625313331120533, "alpha_spy_p3m": 0.39150831255973384, "alpha_c_p3m": 0.23982296903658695, "bench_spy_p3m": 0.07102302055231946, "bench_c_p3m": 0.22270836407546635, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3715, -0.3534, -0.3559, -0.3578, -0.3928, -0.3715, -0.389, -0.3972, -0.3715, -0.4078, -0.3959, -0.3797, -0.3578, -0.3534, -0.3722, -0.3703, -0.3541, -0.3472, -0.3434, -0.3228, -0.3159, -0.3228, -0.3253, -0.3297, -0.3197, -0.3453, -0.3578, -0.3259, -0.3278, -0.3359, -0.3097, -0.3034, -0.3059, -0.3059, -0.2691, -0.2498, -0.2473, -0.2473, -0.2473, -0.1933, -0.1331, -0.1281, -0.1149, -0.1287, -0.1274, -0.1287, -0.1362, -0.1193, -0.0967, -0.0653, -0.0628, -0.0885, -0.0615, -0.0439, -0.0452, -0.0452, 0.0013, 0.0195, 0.0088, 0.0075, -0.0559, 0.0515, 0.0615, 0.0, -0.0044, 0.0446, 0.0408, 0.0534, 0.1212, 0.1375, 0.1375, 0.1375, 0.1375, 0.1927, 0.1933, 0.2116, 0.2555, 0.2775, 0.3685, 0.3591, 0.3591, 0.2247, 0.081, 0.2028, 0.1814, 0.0891, 0.1795, 0.1927, 0.1795, 0.1519, 0.1846, 0.2172, 0.3089, 0.2586, 0.2517, 0.1695, 0.1908, 0.1864, 0.1306, 0.1306, 0.109, 0.0518, 0.193, 0.1222, 0.1713, 0.2147, 0.2361, 0.3241, 0.2833, 0.2958, 0.2644, 0.299, 0.3273, 0.3682, 0.3587, 0.3493, 0.3776, 0.3682, 0.4122, 0.3808, 0.4122, 0.3965, 0.4216, 0.387, 0.387, 0.4625, 0.4625, 0.6733, 0.7079, 0.689, 0.7959, 0.755, 0.7865, 0.862, 0.7016, 0.7676, 0.733, 0.7362, 0.884, 0.84, 0.84, 0.8808, 0.9312, 0.884, 0.9941, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012753767519895765", "ticker_idx": 0, "ts": "2026-01-18T05:07:47+00:00", "symbol": "EDA", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AI Agent EDA will dramatically lower the barriers to chip development... This will trigger a revaluation of EDA companies' valuations", "tweet_type": "prediction", "timeline": "years", "summary": "Bullish EDA sector: AI Agent EDA will slash chip-dev costs 70%, trigger valuation rerating; incumbents like Synopsys partnering with NVIDIA are best positioned.", "resolved_tickers": ["SNPS", "CDNS"], "resolution_method": "basket", "resolution_reason": "EDA sector basket: Synopsys and Cadence (ANSS is simulation, not EDA)", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Future Prediction 1\n\nWon’t EDA companies thrive going forward?\n\nI believe design automation and AI technology will inevitably merge with EDA.\nIn particular, Agent AI like Claude Code is bound to emerge in semiconductor design as well.\n\nBut can xAI or OpenAI actually pull off Agent AI in the semiconductor space?\n\nWhile EDA can be broadly seen as SaaS, the technical barriers are on a completely different level from typical SaaS.\n\nThe massive data and IP moat built up over decades of history makes it nearly impossible for new entrants to challenge—it’s a true economic fortress.\n\nSo I think the EDA industry will evolve along these lines:\n1.  EDA companies will partner with NVIDIA (Synopsys is already doing so) or collaborate with frontier labs.\n\n2.  Instead, the development of AI Agent EDA will most likely be led by the incumbent EDA companies themselves.\n\n3.  AI Agent EDA will dramatically lower the barriers to chip development—potentially slashing 70% of chip development costs.\n\n4.  This will trigger a revaluation of EDA companies’ valuations. It will also create enormous opportunities for foundries, as it removes the critical human resource bottleneck in the chip design industry.\n\n5.  Perhaps the semiconductor trend will shift from today’s low-variety, high-volume production to high-variety, low-volume production.\n\n6.  ASICs, currently the exclusive domain of big tech with massive capital, could become a viable option for mid-sized companies as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.021630611807367972, "ret_signed_p1d": -0.021630611807367972, "alpha_spy_p1d": -0.0012738260838229265, "alpha_c_p1d": -0.0012738260838229265, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.020356785723545046, "ret_p1w": -0.0056819001119607915, "ret_signed_p1w": -0.0056819001119607915, "alpha_spy_p1w": -0.007228891239436264, "alpha_c_p1w": -0.007228891239436264, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.0015469911274754722, "ret_p1m": -0.14491521695724768, "ret_signed_p1m": -0.14491521695724768, "alpha_spy_p1m": -0.13217773638754654, "alpha_c_p1m": -0.13217773638754654, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": -0.012737480569701143, "ret_p3m": -0.08930805241856576, "ret_signed_p3m": -0.08930805241856576, "alpha_spy_p3m": -0.10653652353244636, "alpha_c_p3m": -0.10653652353244636, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.017228471113880595, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0395, -0.0271, -0.0069, 0.0042, -0.02, -0.0213, -0.0417, -0.0271, -0.0401, -0.072, -0.0878, -0.1053, -0.107, -0.0938, -0.1152, -0.1168, -0.1208, -0.1265, -0.1318, -0.1507, -0.1427, -0.1507, -0.1505, -0.1286, -0.1328, -0.1207, -0.1207, -0.104, -0.0879, -0.0641, -0.0189, -0.0196, -0.0164, -0.0177, -0.0211, -0.0067, -0.0095, -0.0523, -0.0581, -0.048, -0.0682, -0.06, -0.0549, -0.0338, -0.0404, -0.0388, -0.0388, -0.0357, -0.0358, -0.0434, -0.0528, -0.0528, -0.0459, -0.047, -0.0117, 0.0081, 0.0004, 0.0241, 0.0293, 0.0045, -0.0174, -0.0029, 0.0, 0.0, -0.0216, -0.0002, -0.0072, -0.0131, -0.0057, -0.0117, -0.0, -0.0584, -0.0828, -0.1011, -0.1712, -0.1668, -0.177, -0.14, -0.1166, -0.1054, -0.1037, -0.1359, -0.1051, -0.1051, -0.1449, -0.0913, -0.1044, -0.1073, -0.1518, -0.1154, -0.0896, -0.1187, -0.1244, -0.1109, -0.1156, -0.1016, -0.0994, -0.1087, -0.1071, -0.1191, -0.1174, -0.1372, -0.1483, -0.1265, -0.1216, -0.1287, -0.1326, -0.1458, -0.1204, -0.1497, -0.1596, -0.1688, -0.2035, -0.2023, -0.1784, -0.1745, -0.1776, -0.1776, -0.1754, -0.1745, -0.1468, -0.1653, -0.2017, -0.1415, -0.1339, -0.0964, -0.0893, -0.0747, -0.052, -0.034, -0.0155, -0.0625, 0.0093, 0.0129, -0.019, -0.0143, -0.0135, 0.0106, 0.0323, 0.0436, 0.0475, 0.0514, 0.0714, 0.0735, 0.0609, 0.0516, 0.0496, 0.0335, 0.0276, 0.0108, 0.0358, 0.0527, 0.0966, 0.0966, 0.1189, 0.0985, 0.0543, 0.0511, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012753767519895765", "ticker_idx": 1, "ts": "2026-01-18T05:07:47+00:00", "symbol": "Synopsys", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Synopsys is already doing so", "tweet_type": "prediction", "timeline": "years", "summary": "Bullish EDA sector: AI Agent EDA will slash chip-dev costs 70%, trigger valuation rerating; incumbents like Synopsys partnering with NVIDIA are best positioned.", "resolved_tickers": ["SNPS"], "resolution_method": "exchange_ticker", "resolution_reason": "Synopsys Inc. listed on Nasdaq as SNPS", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Future Prediction 1\n\nWon’t EDA companies thrive going forward?\n\nI believe design automation and AI technology will inevitably merge with EDA.\nIn particular, Agent AI like Claude Code is bound to emerge in semiconductor design as well.\n\nBut can xAI or OpenAI actually pull off Agent AI in the semiconductor space?\n\nWhile EDA can be broadly seen as SaaS, the technical barriers are on a completely different level from typical SaaS.\n\nThe massive data and IP moat built up over decades of history makes it nearly impossible for new entrants to challenge—it’s a true economic fortress.\n\nSo I think the EDA industry will evolve along these lines:\n1.  EDA companies will partner with NVIDIA (Synopsys is already doing so) or collaborate with frontier labs.\n\n2.  Instead, the development of AI Agent EDA will most likely be led by the incumbent EDA companies themselves.\n\n3.  AI Agent EDA will dramatically lower the barriers to chip development—potentially slashing 70% of chip development costs.\n\n4.  This will trigger a revaluation of EDA companies’ valuations. It will also create enormous opportunities for foundries, as it removes the critical human resource bottleneck in the chip design industry.\n\n5.  Perhaps the semiconductor trend will shift from today’s low-variety, high-volume production to high-variety, low-volume production.\n\n6.  ASICs, currently the exclusive domain of big tech with massive capital, could become a viable option for mid-sized companies as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010342616573649877, "ret_signed_p1d": -0.010342616573649877, "alpha_spy_p1d": 0.010014169149895169, "alpha_c_p1d": 0.010014169149895169, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.020356785723545046, "ret_p1w": -0.02575969447784976, "ret_signed_p1w": -0.02575969447784976, "alpha_spy_p1w": -0.02730668560532523, "alpha_c_p1w": -0.02730668560532523, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.0015469911274754722, "ret_p1m": -0.1827583928216575, "ret_signed_p1m": -0.1827583928216575, "alpha_spy_p1m": -0.17002091225195637, "alpha_c_p1m": -0.17002091225195637, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": -0.012737480569701143, "ret_p3m": -0.14557146678837996, "ret_signed_p3m": -0.14557146678837996, "alpha_spy_p3m": -0.16279993790226055, "alpha_c_p3m": -0.16279993790226055, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.017228471113880595, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1201, -0.1166, -0.101, -0.0986, -0.1152, -0.1181, -0.1421, -0.121, -0.1367, -0.1936, -0.2076, -0.2327, -0.238, -0.2238, -0.2338, -0.2276, -0.237, -0.245, -0.2442, -0.2566, -0.2518, -0.2532, -0.2478, -0.2163, -0.2222, -0.2065, -0.2065, -0.1904, -0.1511, -0.1297, -0.0966, -0.1018, -0.096, -0.0979, -0.0977, -0.0784, -0.0756, -0.1227, -0.1194, -0.1026, -0.1225, -0.1127, -0.1018, -0.0679, -0.0791, -0.0786, -0.0786, -0.0759, -0.0723, -0.081, -0.0902, -0.0902, -0.0695, -0.0428, -0.0146, 0.0066, -0.0035, 0.0172, 0.0331, -0.0087, -0.0212, -0.0157, 0.0, 0.0, -0.0103, 0.0109, -0.0132, -0.0289, -0.0258, -0.0264, -0.0096, -0.0703, -0.0991, -0.1131, -0.1882, -0.1885, -0.2051, -0.1732, -0.1499, -0.1527, -0.1513, -0.18, -0.1534, -0.1534, -0.1828, -0.1433, -0.1431, -0.1479, -0.1849, -0.1464, -0.13, -0.1749, -0.1982, -0.1775, -0.1782, -0.1653, -0.1433, -0.1528, -0.1532, -0.162, -0.1614, -0.189, -0.2008, -0.1751, -0.1685, -0.1697, -0.1706, -0.1859, -0.1624, -0.195, -0.2057, -0.2215, -0.2631, -0.2579, -0.2321, -0.2316, -0.2331, -0.2331, -0.2309, -0.2293, -0.2056, -0.2157, -0.2403, -0.1909, -0.1889, -0.1508, -0.1456, -0.1292, -0.1072, -0.0944, -0.0756, -0.1152, -0.03, -0.0344, -0.0628, -0.068, -0.0653, -0.0529, -0.0364, -0.0267, -0.023, -0.0215, 0.0003, -0.0002, -0.006, -0.0136, -0.0122, -0.0269, -0.0346, -0.0435, -0.0337, -0.0239, 0.0163, 0.0163, 0.0353, 0.0186, -0.0691, -0.0788, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012753767519895765", "ticker_idx": 2, "ts": "2026-01-18T05:07:47+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EDA companies will partner with NVIDIA", "tweet_type": "prediction", "timeline": "years", "summary": "Bullish EDA sector: AI Agent EDA will slash chip-dev costs 70%, trigger valuation rerating; incumbents like Synopsys partnering with NVIDIA are best positioned.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Future Prediction 1\n\nWon’t EDA companies thrive going forward?\n\nI believe design automation and AI technology will inevitably merge with EDA.\nIn particular, Agent AI like Claude Code is bound to emerge in semiconductor design as well.\n\nBut can xAI or OpenAI actually pull off Agent AI in the semiconductor space?\n\nWhile EDA can be broadly seen as SaaS, the technical barriers are on a completely different level from typical SaaS.\n\nThe massive data and IP moat built up over decades of history makes it nearly impossible for new entrants to challenge—it’s a true economic fortress.\n\nSo I think the EDA industry will evolve along these lines:\n1.  EDA companies will partner with NVIDIA (Synopsys is already doing so) or collaborate with frontier labs.\n\n2.  Instead, the development of AI Agent EDA will most likely be led by the incumbent EDA companies themselves.\n\n3.  AI Agent EDA will dramatically lower the barriers to chip development—potentially slashing 70% of chip development costs.\n\n4.  This will trigger a revaluation of EDA companies’ valuations. It will also create enormous opportunities for foundries, as it removes the critical human resource bottleneck in the chip design industry.\n\n5.  Perhaps the semiconductor trend will shift from today’s low-variety, high-volume production to high-variety, low-volume production.\n\n6.  ASICs, currently the exclusive domain of big tech with massive capital, could become a viable option for mid-sized companies as well.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04381672138594639, "ret_signed_p1d": -0.04381672138594639, "alpha_spy_p1d": -0.023459935662401343, "alpha_c_p1d": -0.018841073557214982, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": 0.0012888283191201388, "ret_signed_p1w": 0.0012888283191201388, "alpha_spy_p1w": -0.00025816280835533334, "alpha_c_p1w": 0.005210023320941781, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": -0.006765754613118791, "ret_signed_p1m": -0.006765754613118791, "alpha_spy_p1m": 0.005971725956582352, "alpha_c_p1m": -0.024548403672035146, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.06513855826251302, "ret_signed_p3m": 0.06513855826251302, "alpha_spy_p3m": 0.04791008714863243, "alpha_c_p3m": -0.07075387450034176, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.032, -0.0219, 0.0001, 0.0282, 0.0794, 0.1117, 0.0894, 0.0873, 0.1108, 0.0668, 0.0482, 0.0099, 0.0103, 0.0688, 0.0372, 0.0406, 0.0033, 0.0211, 0.0019, -0.0262, 0.0015, -0.0301, -0.0395, -0.0198, -0.0452, -0.0321, -0.0321, -0.0496, -0.0339, -0.0257, -0.0357, -0.0153, -0.0205, -0.0037, -0.0068, -0.0132, -0.0285, -0.0602, -0.0534, -0.0457, -0.0821, -0.0649, -0.0281, -0.0136, 0.016, 0.0128, 0.0128, 0.0231, 0.0107, 0.007, 0.0014, 0.0014, 0.0141, 0.0101, 0.0054, 0.0155, -0.0064, -0.0074, -0.0069, -0.0023, -0.0166, 0.0044, 0.0, 0.0, -0.0438, -0.0156, -0.0075, 0.0077, 0.0013, 0.0123, 0.0284, 0.0337, 0.0263, -0.0033, -0.0316, -0.0647, -0.0771, -0.0044, 0.0205, 0.0124, 0.0205, 0.0038, -0.0184, -0.0184, -0.0068, 0.0094, 0.009, 0.0193, 0.0286, 0.0355, 0.0501, -0.0072, -0.0485, -0.0201, -0.0332, -0.0171, -0.0155, -0.0452, -0.0192, -0.0078, -0.001, -0.0165, -0.0321, -0.0161, -0.023, -0.0313, -0.0411, -0.0726, -0.0568, -0.0592, -0.0405, -0.0804, -0.1004, -0.113, -0.0635, -0.0562, -0.0474, -0.0474, -0.0461, -0.0436, -0.0222, -0.0124, 0.0129, 0.0166, 0.0553, 0.0679, 0.0651, 0.083, 0.0851, 0.0734, 0.0874, 0.0721, 0.1184, 0.1632, 0.1447, 0.1237, 0.0717, 0.0657, 0.0658, 0.0552, 0.116, 0.1358, 0.1556, 0.1784, 0.1856, 0.2127, 0.2659, 0.21, 0.1939, 0.1847, 0.2, 0.1788, 0.1563, 0.1563, 0.1538, 0.1417, 0.1505, 0.1338, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2018524878337708183", "ticker_idx": 0, "ts": "2026-02-03T03:20:08+00:00", "symbol": "RMBS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This fucking company crashes every single time they announce earnings", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Avoid Rambus (crashes on earnings); buy actual memory manufacturers to ride the memory super cycle.", "resolved_tickers": ["RMBS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "The advice I’m giving you is this: if you want to fully ride the memory super cycle, buy companies that actually manufacture memory. Not IP companies like Rambus.\n\nThis fucking company crashes every single time they announce earnings. https://t.co/QhOVVYnk5U", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.1550025658634917, "ret_signed_m1d": -0.1550025658634917, "alpha_spy_m1d": -0.14647506604123106, "alpha_c_m1d": -0.12912746846247392, "bench_spy_m1d": 0.008527499822260642, "bench_c_m1d": 0.025875097401017788, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024377873034781894, "ret_signed_p1d": -0.024377873034781894, "alpha_spy_p1d": -0.029221794713679028, "alpha_c_p1d": -0.06375627765790959, "bench_spy_p1d": -0.004843921678897134, "bench_c_p1d": -0.039378404623127694, "ret_p1w": 0.08674455236415146, "ret_signed_p1w": -0.08674455236415146, "alpha_spy_p1w": -0.08298845829724111, "alpha_c_p1w": -0.06894125016077424, "bench_spy_p1w": 0.0037560940669103537, "bench_c_p1w": 0.017803302203377225, "ret_p1m": -0.06510915420428864, "ret_signed_p1m": 0.06510915420428864, "alpha_spy_p1m": 0.05872793044970526, "alpha_c_p1m": 0.06867989815391462, "bench_spy_p1m": -0.006381223754583387, "bench_c_p1m": 0.0035707439496259763, "ret_p3m": 0.13692233392369113, "ret_signed_p3m": -0.13692233392369113, "alpha_spy_p3m": -0.08893587567329186, "alpha_c_p3m": 0.14506322409723382, "bench_spy_p3m": 0.047986458250399266, "bench_c_p3m": 0.28198555802092495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0817, 0.0937, 0.1234, 0.0658, 0.0382, -0.0618, -0.0325, -0.0744, -0.1092, -0.0836, -0.1135, -0.1046, -0.0579, -0.0609, -0.0364, -0.0364, -0.0293, -0.0433, -0.0228, -0.0043, 0.0037, 0.032, 0.0571, 0.0636, 0.0852, 0.0768, -0.0429, -0.0305, -0.0382, -0.0796, -0.0707, -0.0208, -0.0324, -0.0433, -0.0403, -0.0403, -0.0441, -0.0496, -0.0382, -0.0666, -0.0666, 0.0084, -0.0096, -0.0147, -0.0691, -0.0722, -0.0515, -0.0564, -0.0302, 0.022, 0.0469, 0.0969, 0.0969, 0.1183, 0.2791, 0.2673, 0.1713, 0.1599, 0.1753, 0.264, 0.2351, 0.1562, 0.155, 0.0, 0.0244, -0.0036, 0.0879, 0.1267, 0.0867, 0.0094, -0.0269, 0.0356, 0.0356, 0.0356, 0.0598, 0.0577, 0.0426, -0.0323, 0.0012, 0.0268, 0.0378, 0.0123, 0.0044, -0.1103, -0.0651, -0.0664, -0.1049, -0.1009, -0.0898, -0.0601, -0.0881, -0.0451, -0.0576, -0.0503, -0.0521, -0.0389, -0.0686, -0.0518, -0.0585, -0.0256, -0.0712, -0.0886, -0.1901, -0.1262, -0.0863, -0.0551, -0.0551, -0.0633, -0.0668, 0.0303, 0.0724, 0.1218, 0.1494, 0.2365, 0.2191, 0.2192, 0.2893, 0.2887, 0.325, 0.3362, 0.4068, 0.6089, 0.4353, 0.1302, 0.1393, 0.1692, 0.1369, 0.1326, 0.1986, 0.3209, 0.2859, 0.3128, 0.3663, 0.3233, 0.3697, 0.3251, 0.2905, 0.2571, 0.2395, 0.3566, 0.4405, 0.4523, 0.4523, 0.5971, 0.51, 0.5035, 0.4775, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2018524878337708183", "ticker_idx": 1, "ts": "2026-02-03T03:20:08+00:00", "symbol": "memory manufacturers", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "if you want to fully ride the memory super cycle, buy companies that actually manufacture memory", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Avoid Rambus (crashes on earnings); buy actual memory manufacturers to ride the memory super cycle.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "2408.TW"], "resolution_method": "basket", "resolution_reason": "Memory manufacturers basket: Micron, SK Hynix, Samsung, Nanya", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "The advice I’m giving you is this: if you want to fully ride the memory super cycle, buy companies that actually manufacture memory. Not IP companies like Rambus.\n\nThis fucking company crashes every single time they announce earnings. https://t.co/QhOVVYnk5U", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020938177675290853, "ret_signed_m1d": -0.020938177675290853, "alpha_spy_m1d": -0.029465677497551496, "alpha_c_m1d": -0.04681327507630864, "bench_spy_m1d": 0.008527499822260642, "bench_c_m1d": 0.025875097401017788, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00764075658941496, "ret_signed_p1d": -0.00764075658941496, "alpha_spy_p1d": -0.002796834910517826, "alpha_c_p1d": 0.031737648033712734, "bench_spy_p1d": -0.004843921678897134, "bench_c_p1d": -0.039378404623127694, "ret_p1w": -0.04446857046425798, "ret_signed_p1w": -0.04446857046425798, "alpha_spy_p1w": -0.04822466453116833, "alpha_c_p1w": -0.0622718726676352, "bench_spy_p1w": 0.0037560940669103537, "bench_c_p1w": 0.017803302203377225, "ret_p1m": -0.058406712682431944, "ret_signed_p1m": -0.058406712682431944, "alpha_spy_p1m": -0.05202548892784856, "alpha_c_p1m": -0.06197745663205792, "bench_spy_p1m": -0.006381223754583387, "bench_c_p1m": 0.0035707439496259763, "ret_p3m": 0.2023598875997221, "ret_signed_p3m": 0.2023598875997221, "alpha_spy_p3m": 0.15437342934932283, "alpha_c_p3m": -0.07962567042120286, "bench_spy_p3m": 0.047986458250399266, "bench_c_p3m": 0.28198555802092495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.414, -0.4207, -0.3873, -0.3839, -0.3842, -0.3903, -0.4115, -0.3895, -0.417, -0.4231, -0.4332, -0.466, -0.4509, -0.4465, -0.4428, -0.4276, -0.4322, -0.424, -0.4152, -0.4171, -0.4234, -0.4099, -0.3839, -0.3866, -0.3801, -0.3904, -0.3912, -0.4037, -0.4204, -0.4053, -0.3868, -0.3757, -0.3491, -0.3498, -0.3318, -0.3318, -0.3212, -0.3, -0.292, -0.2976, -0.2976, -0.2471, -0.2286, -0.182, -0.1655, -0.172, -0.1858, -0.1649, -0.1825, -0.1743, -0.1642, -0.1279, -0.1026, -0.1159, -0.1148, -0.0875, -0.0788, -0.0806, -0.0474, 0.0012, 0.0045, 0.0315, -0.0209, 0.0, -0.0076, -0.0542, -0.0589, -0.0313, -0.0445, -0.0175, 0.0085, 0.0088, 0.0088, 0.0017, 0.0144, 0.0292, 0.051, 0.0615, 0.0931, 0.0995, 0.1337, 0.1191, 0.1188, 0.0079, -0.0584, 0.0145, -0.0333, -0.089, -0.0122, 0.0257, -0.0054, -0.0064, 0.0495, 0.077, 0.1267, 0.078, 0.0386, -0.0224, -0.0138, -0.0114, -0.0585, -0.0629, -0.1047, -0.1468, -0.0592, -0.105, -0.0806, -0.0606, -0.0424, 0.0435, 0.0167, 0.0362, 0.0453, 0.0893, 0.0919, 0.1092, 0.0951, 0.0967, 0.1317, 0.151, 0.1487, 0.1475, 0.2094, 0.2057, 0.2159, 0.1874, 0.2024, 0.3046, 0.3601, 0.4915, 0.5069, 0.5593, 0.6916, 0.6703, 0.7388, 0.7566, 0.6196, 0.6095, 0.5579, 0.5789, 0.702, 0.701, 0.6879, 0.8215, 0.9132, 0.9227, 1.0101, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2009213052697367034", "ticker_idx": 0, "ts": "2026-01-08T10:38:15+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sandisk could raise enterprise SSD NAND prices by more than 100% quarter-on-quarter (QoQ) in Q1", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Nomura thesis: Sandisk poised to raise enterprise SSD NAND prices 100%+ QoQ in Q1, bullish for SNDK.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura: Sandisk could raise enterprise SSD NAND prices by more than 100% quarter-on-quarter (QoQ) in Q1.\n\n$SNDK https://t.co/XHG3q8mIsn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.056854153547730046, "ret_signed_m1d": 0.056854153547730046, "alpha_spy_m1d": 0.056752610555812044, "alpha_c_m1d": 0.04097784409723171, "bench_spy_m1d": 0.0001015429919180022, "bench_c_m1d": 0.015876309450498338, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.12814609320915027, "ret_signed_p1d": 0.12814609320915027, "alpha_spy_p1d": 0.12153275081505766, "alpha_c_p1d": 0.11490430781254601, "bench_spy_p1d": 0.006613342394092614, "bench_c_p1d": 0.013241785396604255, "ret_p1w": 0.22329162366665756, "ret_signed_p1w": 0.22329162366665756, "alpha_spy_p1w": 0.21933233459541723, "alpha_c_p1w": 0.2148334594595367, "bench_spy_p1w": 0.003959289071240324, "bench_c_p1w": 0.008458164207120866, "ret_p1m": 0.7873796763735623, "ret_signed_p1m": 0.7873796763735623, "alpha_spy_p1m": 0.7857699004107768, "alpha_c_p1m": 0.8089409779659279, "bench_spy_p1m": 0.0016097759627855535, "bench_c_p1m": -0.021561301592365578, "ret_p3m": 1.1247084642718388, "ret_signed_p3m": 1.1247084642718388, "alpha_spy_p3m": 1.1660272308014226, "alpha_c_p3m": 1.1707368501484257, "bench_spy_p3m": -0.04131876652958366, "bench_c_p3m": -0.046028385876586975, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5976, -0.6195, -0.5687, -0.5688, -0.581, -0.5575, -0.5537, -0.5607, -0.5007, -0.4435, -0.4724, -0.4753, -0.3891, -0.4147, -0.4042, -0.3812, -0.4184, -0.3528, -0.3792, -0.2842, -0.199, -0.1882, -0.1538, -0.2719, -0.2403, -0.2052, -0.2679, -0.2648, -0.4142, -0.4014, -0.3216, -0.3409, -0.3572, -0.3572, -0.3326, -0.3718, -0.3862, -0.419, -0.3624, -0.3171, -0.326, -0.344, -0.3039, -0.2778, -0.3837, -0.3966, -0.3743, -0.3817, -0.344, -0.2897, -0.2795, -0.268, -0.2525, -0.2525, -0.2526, -0.2699, -0.2819, -0.2904, -0.2904, -0.1773, -0.1807, 0.0451, 0.0569, 0.0, 0.1281, 0.1636, 0.1652, 0.1592, 0.2233, 0.2364, 0.2364, 0.3545, 0.4984, 0.5049, 0.4164, 0.4073, 0.4391, 0.5772, 0.6121, 0.7225, 0.9885, 1.079, 0.7473, 0.7224, 0.7874, 0.7439, 0.6191, 0.7915, 0.884, 0.8729, 0.8729, 0.7654, 0.7947, 0.8565, 0.9429, 0.9923, 0.9087, 0.8903, 0.9486, 0.8992, 0.8505, 0.6901, 0.7907, 0.6906, 0.5763, 0.7598, 0.85, 0.9592, 0.8498, 0.9777, 1.1033, 1.1527, 1.2529, 1.3079, 1.1215, 1.0999, 1.0998, 1.0262, 0.803, 0.8408, 0.7113, 0.8991, 1.0707, 1.0972, 1.0972, 1.166, 1.1247, 1.3343, 1.5455, 1.5461, 1.8472, 1.8232, 1.6655, 1.7485, 1.753, 1.7292, 1.7007, 1.9266, 1.7872, 1.959, 2.199, 1.9962, 2.1811, 2.2777, 2.5482, 2.754, 3.2037, 3.2147, 3.0054, 3.6701, 3.6259, 3.3403, 3.326, 3.1332, 3.2076, 2.9846, 3.1349, 3.1626, 3.61, 3.4201, 3.4201, 3.7514, 3.7526, 3.9072, 4.0666, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2034798663164371250", "ticker_idx": 0, "ts": "2026-03-20T01:06:20+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I confidently told them I expected ASPs to rise by 50%... now it feels like the sell side is starting to catch up with me", "tweet_type": "prediction", "timeline": "quarters", "summary": "Author's 50% memory ASP increase forecast for Q2 was more aggressive than all sell/buy side; vindicated, bullish Micron.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "During my U.S. trip, quite a few people I met asked me what I thought memory prices would look like in the second quarter.\n\nSo I confidently told them I expected ASPs to rise by 50%.\n\nThen one of them said that was the most aggressive number he had heard across both the sell side and the buy side.\n\nBut… now it feels like the sell side is starting to catch up with me, doesn’t it?", "reply_text": null, "parent_text": null, "quoted_text": "GF Securities Overseas Electronics &amp; Communications | Micron FY2Q26 Review: No Need to Worry About the CAPEX Increase\n\n☀️ Results significantly beat expectations; target price raised to US$623: Micron delivered a very strong performance in fiscal 2Q26, with revenue reaching", "parent_author": null, "ret_m1d": 0.05053203316717414, "ret_signed_m1d": 0.05053203316717414, "alpha_spy_m1d": 0.03598775376493224, "alpha_c_m1d": 0.024072545066761375, "bench_spy_m1d": 0.0145442794022419, "bench_c_m1d": 0.026459488100412765, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04386378691875292, "ret_signed_p1d": -0.04386378691875292, "alpha_spy_p1d": -0.05436380616369074, "alpha_c_p1d": -0.061070199782554124, "bench_spy_p1d": 0.01050001924493782, "bench_c_p1d": 0.0172064128638012, "ret_p1w": -0.15530854978223718, "ret_signed_p1w": -0.15530854978223718, "alpha_spy_p1w": -0.1329825397829627, "alpha_c_p1w": -0.1280434083863955, "bench_spy_p1w": -0.022326009999274476, "bench_c_p1w": -0.027265141395841686, "ret_p1m": 0.060790723610155695, "ret_signed_p1m": 0.060790723610155695, "alpha_spy_p1m": -0.031951713905671975, "alpha_c_p1m": -0.1451145714761286, "bench_spy_p1m": 0.09274243751582767, "bench_c_p1m": 0.2059052950862843, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.347, -0.3224, -0.3224, -0.3269, -0.3039, -0.308, -0.3251, -0.3251, -0.2541, -0.2619, -0.1879, -0.1971, -0.2267, -0.184, -0.1821, -0.2004, -0.2118, -0.204, -0.1422, -0.1422, -0.1369, -0.0799, -0.0599, -0.055, -0.0799, -0.0299, 0.0293, 0.0305, -0.019, 0.0352, -0.0082, -0.1029, -0.0946, -0.0667, -0.0932, -0.1174, -0.0297, -0.0211, -0.0266, -0.0266, -0.0547, -0.0046, -0.0131, 0.0125, -0.0046, -0.0116, 0.0144, -0.0174, -0.0249, -0.0242, -0.1022, -0.0523, -0.0611, -0.1244, -0.0794, -0.0468, -0.01, -0.0415, 0.0076, 0.0447, 0.0917, 0.0918, 0.0505, 0.0, -0.0439, -0.0647, -0.0965, -0.1595, -0.1553, -0.2387, -0.2008, -0.1298, -0.1336, -0.1336, -0.1064, -0.1068, -0.0378, -0.0029, -0.005, 0.0091, 0.1016, 0.0793, 0.0816, 0.0765, 0.0608, 0.0631, 0.1532, 0.1396, 0.1751, 0.2409, 0.193, 0.2265, 0.2234, 0.2827, 0.3637, 0.5145, 0.5769, 0.5297, 0.7667, 0.8814, 0.8134, 0.9011, 0.8357, 0.7143, 0.6123, 0.653, 0.7316, 0.8028, 0.7766, 0.7766, 1.1193, 1.1963, 1.1847, 1.297, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011342432492011770", "ticker_idx": 0, "ts": "2026-01-14T07:39:39+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bernstein: Sandisk (SNDK - Outperform, $300 TP)", "tweet_type": "news_reaction", "timeline": "months", "summary": "Bernstein reiterates Outperform on SNDK with $300 TP; NAND undersupply tightening further on AI/datacenter KV cache demand, prices rising at unprecedented pace.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Sandisk (SNDK - Outperform, $300 TP) \n\nSandisk sees the CES keynote comments on Vera Rubin KV cache from Jensen Huang of Nvidia as incremental new NAND demand that is not previously in the demand forecast. Prior to the forecast, Sandisk estimated that NAND was 8% undersupplied and this incremental demand will just make the extremely tight situation even tighter. The company couldn't quantify the impact as it was quite new information but felt that the impact was significant. Sandisk forecasted Datacenter NAND demand growth at 45% vs. mid 20s% for overall NAND bit demand growth compared to supply growth in the teens%. We estimate the KV cache could indicate an additional 16TB per GPU, vs. 3-4TB per GPU today. This would equate to >5x jump in NAND content.\n\nThe company reaffirmed that they are not adding capacity and doesn't see any significant new capacity coming online besides some from YMTC. Sandisk is currently taking a cautious approach towards supply additions following a few years of low profits. They need to see demand consistently outstrip supply with margins significantly higher than this before even considering capacity additions. In addition Sandisk is careful to consider the long-term impact of supply additions, which are essentially permanent. There was some discussion around competitors and how YMTC is really the only NAND player that is adding capacity and isn't acting rationally (arguably).\n\nLong term agreements are becoming a hot topic but currently less than 10% of business is locked in on longer-term price deals. Therefore, over 90% of SNDK's business is on short term (i.e. one quarter or less) price agreements. In an up-market this is a good thing as they are therefore more levered to the market price increases.\n\nSandisk mentioned that they are rising prices faster than the company ever has. Prices are increasing substantially every week in a manner that is unprecedented in the memory industry.\n\nWhere NAND supply bit growth is coming from:\n\n• For SNDK its entirely from technology migration to BiCS 8, which is offset by loss in wafer throughput due to the added layers. This in turn is offset by expansion in clean room space.\n\n• NAND industry supply growth is in mid to high teens, which is SNDK's goal. For 2026, SNDK may be below that.\n\n• Very few cleanrooms available for NAND expansion besides SNDK/Kioxia's Kitakami Fab (but this is planned to offset wafer throughput loss from BiCS8 ramp), and YMTC. Any new cleanrooms would take 2-3 years before having significant impact on the market due to the long lead times.\n\nWhere all the NAND demand growth is coming from:\n\n• Disproportionally most of the incremental demand growth is coming from datacenters, which in turn is coming from AI. Datacenters will be the largest portion of overall NAND demand for the first time this year.\n\n• Traditional NAND markets like mobile, have more price sensitivity than datacenter customers, which don't appear to be sensitive to price increases.\n\n• Although the rapid price increase is not expected to impact datacenter demand, overall NAND demand will be impacted due to some demand impact in mobile and PC markets (in the form of lower content growth or lower unit growth or both. Even with this weakness the market is still expected to be very tight due to the strong and price insensitive datacenter customers.\n\nHigh Bandwidth Flash (HBF) showing promise and remaining on track:\n\n• HBF is a flash based replacement for HBM used for inference.\n\n• Currently working with SK Hynix to standardize the interface for HBF\n\n• SNDK is on target to deliver HBF component samples in 2H26 and full solution samples by 1H27.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005157164623374122, "ret_signed_m1d": 0.005157164623374122, "alpha_spy_m1d": 0.00021764838710147671, "alpha_c_m1d": -0.007144125497510201, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.012301290120884323, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05525900005335327, "ret_signed_p1d": 0.05525900005335327, "alpha_spy_p1d": 0.05253578565682382, "alpha_c_p1d": 0.050006702133419845, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.005252297919933424, "ret_p1w": 0.2926175490594014, "ret_signed_p1w": 0.2926175490594014, "alpha_spy_p1w": 0.29980217905933404, "alpha_c_p1w": 0.2987682469110019, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": -0.006150697851600495, "ret_p1m": 0.6252545885749476, "ret_signed_p1m": 0.6252545885749476, "alpha_spy_p1m": 0.6384215491527889, "alpha_c_p1m": 0.6631950661216538, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": -0.03794047754670615, "ret_p3m": 1.456099651881945, "ret_signed_p3m": 1.456099651881945, "alpha_spy_p3m": 1.4595561962312087, "alpha_c_p3m": 1.4485595963266416, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.007540055555303349, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6386, -0.6183, -0.615, -0.6211, -0.5692, -0.52, -0.5449, -0.5474, -0.473, -0.4951, -0.486, -0.4662, -0.4983, -0.4417, -0.4645, -0.3825, -0.3091, -0.2997, -0.27, -0.3719, -0.3446, -0.3144, -0.3684, -0.3658, -0.4947, -0.4836, -0.4148, -0.4314, -0.4455, -0.4455, -0.4243, -0.4581, -0.4705, -0.4988, -0.45, -0.4109, -0.4186, -0.4341, -0.3996, -0.377, -0.4683, -0.4795, -0.4603, -0.4667, -0.4341, -0.3873, -0.3784, -0.3685, -0.3551, -0.3551, -0.3552, -0.3702, -0.3806, -0.3879, -0.3879, -0.2903, -0.2933, -0.0985, -0.0883, -0.1374, -0.0268, 0.0038, 0.0052, 0.0, 0.0553, 0.0666, 0.0666, 0.1684, 0.2926, 0.2982, 0.2218, 0.214, 0.2414, 0.3605, 0.3906, 0.4859, 0.7154, 0.7934, 0.5073, 0.4858, 0.5419, 0.5043, 0.3967, 0.5454, 0.6253, 0.6156, 0.6156, 0.5229, 0.5482, 0.6015, 0.676, 0.7186, 0.6465, 0.6306, 0.681, 0.6383, 0.5963, 0.458, 0.5447, 0.4584, 0.3598, 0.5181, 0.5959, 0.6901, 0.5957, 0.706, 0.8144, 0.857, 0.9435, 0.9909, 0.83, 0.8114, 0.8114, 0.7479, 0.5553, 0.588, 0.4762, 0.6383, 0.7863, 0.8091, 0.8091, 0.8685, 0.8329, 1.0136, 1.1958, 1.1964, 1.4561, 1.4354, 1.2994, 1.3709, 1.3748, 1.3543, 1.3297, 1.5246, 1.4043, 1.5525, 1.7596, 1.5846, 1.7442, 1.8274, 2.0608, 2.2383, 2.6263, 2.6357, 2.4552, 3.0286, 2.9905, 2.7442, 2.7318, 2.5655, 2.6296, 2.4373, 2.5669, 2.5908, 2.9768, 2.8129, 2.8129, 3.0988, 3.0998, 3.2331, 3.3706, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2029120738440167894", "ticker_idx": 0, "ts": "2026-03-04T09:04:17+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's Feynman GPU could potentially adopt Intel's EMIB-T packaging, which could translate into an approximately $1 billion annual revenue opportunity for Intel Foundry", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Wells Fargo note that NVIDIA Feynman GPU may use Intel EMIB-T packaging, creating ~$1B annual revenue opportunity for Intel Foundry.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Wells Fargo noted in its NVIDIA GTC 2026 preview report that NVIDIA’s Feynman GPU could potentially adopt Intel’s EMIB-T packaging, which could translate into an approximately $1 billion annual revenue opportunity for Intel Foundry.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05440990033580717, "ret_signed_m1d": -0.05440990033580717, "alpha_spy_m1d": -0.047404020624371324, "alpha_c_m1d": -0.034264552247697866, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.020145348088109305, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008117571676635515, "ret_signed_p1d": 0.008117571676635515, "alpha_spy_p1d": 0.013693118523029346, "alpha_c_p1d": 0.017513712850903862, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.009396141174268346, "ret_p1w": 0.05265462077156968, "ret_signed_p1w": 0.05265462077156968, "alpha_spy_p1w": 0.06549885268094524, "alpha_c_p1w": 0.0478187584657177, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004835862305851979, "ret_p1m": 0.10530932523548509, "ret_signed_p1m": 0.10530932523548509, "alpha_spy_p1m": 0.14546065895477844, "alpha_c_p1m": 0.12229754541992255, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.016988220184437464, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0915, -0.1158, -0.1115, -0.1053, -0.1332, -0.1705, -0.1771, -0.1814, -0.2091, -0.204, -0.1922, -0.2021, -0.2025, -0.2067, -0.2067, -0.2058, -0.1953, -0.1817, -0.1904, -0.1904, -0.136, -0.1362, -0.1215, -0.0647, -0.0981, -0.0007, -0.0333, 0.0375, 0.0689, 0.0601, 0.0303, 0.0303, 0.0654, 0.1902, 0.1918, -0.0112, -0.0678, -0.0362, 0.0702, 0.0676, 0.0195, 0.0709, 0.0805, 0.0663, 0.0584, 0.1099, 0.1022, 0.034, 0.0595, 0.0197, 0.0265, 0.0265, 0.0132, -0.0026, -0.0211, -0.0323, -0.0428, 0.0118, 0.0285, -0.0026, 0.0007, -0.0018, -0.0544, 0.0, 0.0081, -0.0474, 0.0, 0.0263, 0.0527, -0.0072, 0.0042, 0.0039, -0.0333, -0.0121, 0.0132, -0.0375, -0.0344, -0.0333, 0.0351, -0.0325, -0.0538, -0.0963, -0.0318, 0.0538, 0.1053, 0.1053, 0.1141, 0.1608, 0.2933, 0.3541, 0.3686, 0.43, 0.4, 0.4247, 0.5029, 0.5029, 0.4414, 0.4537, 0.432, 0.4651, 0.8109, 0.8646, 0.8543, 1.0788, 1.0728, 1.1856, 1.1014, 1.3728, 1.4794, 1.405, 1.7407, 1.8398, 1.6461, 1.6391, 1.5434, 1.3864, 1.3732, 1.4309, 1.6099, 1.5998, 1.6292, 1.6292, 1.71, 1.6716, 1.6523, 1.516, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014299492334653880", "ticker_idx": 0, "ts": "2026-01-22T11:29:57+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML's DUV upside is currently overlooked by the market", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish ASML: DUV upside ignored by market, implying undervaluation.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "\"ASML’s DUV upside is currently overlooked by the market.\"\n\n$ASML https://t.co/ugdWA4I1B5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018361185972599947, "ret_signed_m1d": -0.018361185972599947, "alpha_spy_m1d": -0.013165096324308512, "alpha_c_m1d": -0.016151726084176032, "bench_spy_m1d": -0.005196089648291435, "bench_c_m1d": -0.0022094598884239147, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.001529951306477706, "ret_signed_p1d": 0.001529951306477706, "alpha_spy_p1d": 0.0011671710667371382, "alpha_c_p1d": 0.008307199051599823, "bench_spy_p1d": 0.0003627802397405677, "bench_c_p1d": -0.006777247745122117, "ret_p1w": 0.013260654051059362, "ret_signed_p1w": 0.013260654051059362, "alpha_spy_p1w": 0.005916486107614638, "alpha_c_p1w": -0.023232026109242465, "bench_spy_p1w": 0.007344167943444724, "bench_c_p1w": 0.03649268016030183, "ret_p1m": 0.06875637826827674, "ret_signed_p1m": 0.06875637826827674, "alpha_spy_p1m": 0.06810321394339613, "alpha_c_p1m": 0.03844509473451074, "bench_spy_p1m": 0.0006531643248806063, "bench_c_p1m": 0.030311283533765998, "ret_p3m": 0.05292421471802333, "ret_signed_p3m": 0.05292421471802333, "alpha_spy_p3m": 0.028216872841884566, "alpha_c_p3m": -0.10059347361297677, "bench_spy_p3m": 0.024707341876138766, "bench_c_p3m": 0.1535176883310001, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2294, -0.2287, -0.2185, -0.2031, -0.2196, -0.2124, -0.2197, -0.2293, -0.2389, -0.2598, -0.2463, -0.2461, -0.2419, -0.2507, -0.257, -0.255, -0.2642, -0.2463, -0.2436, -0.2911, -0.27, -0.2687, -0.227, -0.2373, -0.2321, -0.2123, -0.2017, -0.1811, -0.1862, -0.1911, -0.1812, -0.19, -0.1959, -0.2002, -0.2134, -0.2086, -0.2276, -0.2571, -0.2412, -0.2336, -0.2384, -0.2322, -0.2358, -0.2358, -0.2358, -0.2292, -0.2193, -0.2168, -0.2168, -0.1616, -0.1047, -0.0981, -0.1059, -0.1389, -0.0806, -0.0765, -0.0627, -0.0784, -0.023, -0.0078, -0.0476, -0.0309, -0.0184, 0.0, 0.0015, 0.0014, 0.035, 0.0153, 0.0133, 0.0333, 0.0411, 0.0119, -0.0304, -0.0228, 0.0148, 0.0255, 0.0155, 0.0281, 0.0042, 0.0133, 0.0172, 0.0207, 0.0596, 0.0539, 0.0688, 0.0633, 0.0754, 0.0967, 0.049, 0.0499, 0.0303, -0.0111, 0.0213, 0.0095, -0.0237, -0.0232, 0.0214, 0.0204, 0.0047, 0.0041, 0.0192, 0.0162, 0.0138, -0.0053, -0.0397, 0.0005, 0.0228, 0.0313, -0.0077, -0.0242, -0.0535, -0.0473, 0.0109, -0.0118, -0.0118, -0.0118, -0.0519, 0.0322, 0.0511, 0.081, 0.0723, 0.0931, 0.047, 0.0407, 0.0589, 0.0597, 0.0529, 0.0638, 0.0417, 0.066, 0.0344, -0.0004, 0.0174, 0.0428, 0.0428, 0.0129, 0.0484, 0.1133, 0.109, 0.1255, 0.1138, 0.0798, 0.132, 0.1662, 0.1146, 0.0786, 0.0655, 0.1371, 0.1476, 0.202, 0.2204, 0.1856, 0.1742, 0.1866, 0.1813, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034824046987419851", "ticker_idx": 0, "ts": "2026-03-20T02:47:12+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long-term supply agreements would provide more than three years of demand visibility, allowing companies to invest with confidence and maintaining margins at reasonable levels by preventing dramatic price declines", "tweet_type": "analysis", "timeline": "years", "summary": "Samsung, Micron, and SK Hynix all benefit from historic LTAs with hyperscalers providing multi-year demand visibility, enabling confident capex and margin stabilization.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "BIG: Samsung Pursues Long-Term Supply 'Lock-In' with Google and Microsoft… A New Memory Paradigm Emerges\n\nSamsung Electronics is reportedly in talks to establish long-term supply agreements (LTAs) with Google and Microsoft. While various contract structures are under discussion, the most likely arrangement is one where large volumes are committed over an extended period with pricing remaining variable. If finalized, this would mark the first binding long-term supply contract in the industry's history — a significant historical inflection point.\n\n◆ Big Tech Gets Desperate, Seeks Long-Term Supply Commitments\n\nAccording to industry sources on the 20th, Samsung Electronics has recently been in negotiations with Google and Microsoft over long-term supply contracts. As memory has emerged as a critical bottleneck for AI data center buildouts, major tech companies are racing to secure commitments from Korean memory makers.\n\nDespite concerns over an AI bubble, U.S. hyperscalers have continued — and even accelerated — their capital expenditure programs, exacerbating memory supply shortages. Commodity DRAM prices have surged to the point of rivaling HBM margins, prompting Big Tech to pursue LTAs with memory manufacturers in order to secure stable supply.\n\nThe exact structure of the contracts remains under discussion, but the most likely format involves fixed volume commitments with pricing linked to spot market rates. Under this arrangement, Big Tech customers would pay large upfront prepayments to Samsung, which would be drawn down if the customer fails to take the agreed volumes within the three-to-five year contract window. However, pricing would remain tied to spot prices, adjusting upward or downward if spot prices move beyond a defined range.\n\nMemory semiconductor pricing is typically divided into spot prices and contract prices. Spot prices are determined daily by immediate supply and demand in B2C markets such as distributors and small-to-medium PC assemblers. Contract prices, on the other hand, serve as the benchmark for large-scale supply deals between manufacturers like Samsung Electronics and SK Hynix and major global firms such as Apple and Amazon.\n\nUnder this type of structure, Samsung would gain long-term demand visibility, enabling it to accelerate capacity expansion while avoiding inventory buildup — thereby preventing the steep price declines seen in past cycles. \"This long-term supply agreement is essentially a green light to expand capacity,\" said one industry official. \"By committing to large volumes over an extended period, manufacturers will be able to invest with confidence.\"\n\nMicron is also inking long-term supply agreements. In its fiscal Q2 2026 earnings call, Micron disclosed that it had \"signed its first-ever five-year Strategic Customer Agreement (SCA).\" SK Hynix is also said to have already reached LTAs with select customers, and it is expected that memory manufacturers will have concluded agreements with all major hyperscalers by mid-year.\n\nIn response, a Samsung Electronics representative said, \"We are unable to confirm details related to specific customers,\" while adding, \"All major customers are currently requesting large volumes of memory.\"\n\n◆ \"This Time Is Different\" — Prepayment First, Supply Later\n\nThe significance of these contracts lies in the potential paradigm shift they represent for the memory industry. Historically, the sector has been plagued by periodic down-cycles driven by the mismatch between large-scale capacity additions and fluctuating demand.\n\nLong-term supply agreements, however, would provide more than three years of demand visibility, allowing companies to invest with confidence and maintaining margins at reasonable levels by preventing dramatic price declines.\n\nThe defining feature of this round of LTAs is their enforceability. While long-term contracts were attempted around 2019, they carried no binding force — customers could cancel orders at will. This year's agreements, by contrast, enforce commitment through large upfront prepayments. \"My understanding is that Samsung is in discussions to receive over $10 billion in prepayments from Microsoft, with the prepayment balance drawn down if Microsoft fails to take the agreed volumes,\" said another industry source.\n\nThe successful conclusion of LTAs is expected to drive further expansion of Samsung's investment plans. With demand visibility materially improved, there is little reason to hesitate on capex. Micron, for its part, has already announced plans to invest over $25 billion in fiscal 2026 — nearly double the $13.8 billion invested the prior year.\n\nAt Samsung's annual general meeting on the 18th, Vice Chairman and CEO Jeon Young-hyun stated: \"For the semiconductor division, it is critically important to minimize uncertainty in our medium-to-long-term business and maintain a healthy supply-demand balance in memory. By securing long-term supply through multi-year agreements, both our customers and ourselves can enhance predictable business stability and visibility.\"\n\n---\n\nhttps://t.co/WfcyBuFWca", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005516582333744857, "ret_signed_m1d": 0.005516582333744857, "alpha_spy_m1d": -0.009027697068497043, "alpha_c_m1d": -0.01769276387541341, "bench_spy_m1d": 0.0145442794022419, "bench_c_m1d": 0.023209346209158266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06569707363190647, "ret_signed_p1d": -0.06569707363190647, "alpha_spy_p1d": -0.07619709287684429, "alpha_c_p1d": -0.07922561088422442, "bench_spy_p1d": 0.01050001924493782, "bench_c_p1d": 0.013528537252317951, "ret_p1w": -0.0967903235431763, "ret_signed_p1w": -0.0967903235431763, "alpha_spy_p1w": -0.07446431354390182, "alpha_c_p1w": -0.058291810496550855, "bench_spy_p1w": -0.022326009999274476, "bench_c_p1w": -0.03849851304662544, "ret_p1m": 0.07795377236706957, "ret_signed_p1m": 0.07795377236706957, "alpha_spy_p1m": -0.014788665148758096, "alpha_c_p1m": -0.06590144292356537, "bench_spy_p1m": 0.09274243751582767, "bench_c_p1m": 0.14385521529063494, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4435, -0.4455, -0.4455, -0.4161, -0.4007, -0.3987, -0.3987, -0.3987, -0.3556, -0.3074, -0.3034, -0.2929, -0.3039, -0.3029, -0.3039, -0.3099, -0.2964, -0.2783, -0.2533, -0.2513, -0.2718, -0.2503, -0.2362, -0.2372, -0.2372, -0.2001, -0.1856, -0.1941, -0.1951, -0.2457, -0.16, -0.152, -0.2011, -0.2046, -0.1655, -0.1685, -0.1585, -0.1043, -0.0913, -0.0913, -0.0913, -0.0913, -0.0471, -0.0466, -0.0321, 0.003, 0.0206, 0.0933, 0.0858, 0.0858, -0.0216, -0.1364, -0.0391, -0.0562, -0.1299, -0.0577, -0.0471, -0.0577, -0.0797, -0.0537, -0.0276, 0.0456, 0.0055, 0.0, -0.0657, -0.0486, -0.0522, -0.0968, -0.0968, -0.114, -0.1597, -0.0469, -0.1035, -0.0643, -0.0296, -0.0125, 0.0579, 0.0252, 0.0352, 0.0101, 0.0378, 0.0604, 0.093, 0.0855, 0.078, 0.1006, 0.093, 0.1282, 0.1031, 0.1282, 0.1156, 0.1357, 0.1081, 0.1081, 0.1684, 0.1684, 0.3368, 0.3644, 0.3493, 0.4348, 0.4021, 0.4272, 0.4875, 0.3594, 0.4121, 0.3845, 0.387, 0.5051, 0.4699, 0.4699, 0.5026, 0.5428, 0.5051, 0.5931, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034824046987419851", "ticker_idx": 1, "ts": "2026-03-20T02:47:12+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is also inking long-term supply agreements. In its fiscal Q2 2026 earnings call, Micron disclosed that it had 'signed its first-ever five-year Strategic Customer Agreement'", "tweet_type": "analysis", "timeline": "years", "summary": "Samsung, Micron, and SK Hynix all benefit from historic LTAs with hyperscalers providing multi-year demand visibility, enabling confident capex and margin stabilization.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "BIG: Samsung Pursues Long-Term Supply 'Lock-In' with Google and Microsoft… A New Memory Paradigm Emerges\n\nSamsung Electronics is reportedly in talks to establish long-term supply agreements (LTAs) with Google and Microsoft. While various contract structures are under discussion, the most likely arrangement is one where large volumes are committed over an extended period with pricing remaining variable. If finalized, this would mark the first binding long-term supply contract in the industry's history — a significant historical inflection point.\n\n◆ Big Tech Gets Desperate, Seeks Long-Term Supply Commitments\n\nAccording to industry sources on the 20th, Samsung Electronics has recently been in negotiations with Google and Microsoft over long-term supply contracts. As memory has emerged as a critical bottleneck for AI data center buildouts, major tech companies are racing to secure commitments from Korean memory makers.\n\nDespite concerns over an AI bubble, U.S. hyperscalers have continued — and even accelerated — their capital expenditure programs, exacerbating memory supply shortages. Commodity DRAM prices have surged to the point of rivaling HBM margins, prompting Big Tech to pursue LTAs with memory manufacturers in order to secure stable supply.\n\nThe exact structure of the contracts remains under discussion, but the most likely format involves fixed volume commitments with pricing linked to spot market rates. Under this arrangement, Big Tech customers would pay large upfront prepayments to Samsung, which would be drawn down if the customer fails to take the agreed volumes within the three-to-five year contract window. However, pricing would remain tied to spot prices, adjusting upward or downward if spot prices move beyond a defined range.\n\nMemory semiconductor pricing is typically divided into spot prices and contract prices. Spot prices are determined daily by immediate supply and demand in B2C markets such as distributors and small-to-medium PC assemblers. Contract prices, on the other hand, serve as the benchmark for large-scale supply deals between manufacturers like Samsung Electronics and SK Hynix and major global firms such as Apple and Amazon.\n\nUnder this type of structure, Samsung would gain long-term demand visibility, enabling it to accelerate capacity expansion while avoiding inventory buildup — thereby preventing the steep price declines seen in past cycles. \"This long-term supply agreement is essentially a green light to expand capacity,\" said one industry official. \"By committing to large volumes over an extended period, manufacturers will be able to invest with confidence.\"\n\nMicron is also inking long-term supply agreements. In its fiscal Q2 2026 earnings call, Micron disclosed that it had \"signed its first-ever five-year Strategic Customer Agreement (SCA).\" SK Hynix is also said to have already reached LTAs with select customers, and it is expected that memory manufacturers will have concluded agreements with all major hyperscalers by mid-year.\n\nIn response, a Samsung Electronics representative said, \"We are unable to confirm details related to specific customers,\" while adding, \"All major customers are currently requesting large volumes of memory.\"\n\n◆ \"This Time Is Different\" — Prepayment First, Supply Later\n\nThe significance of these contracts lies in the potential paradigm shift they represent for the memory industry. Historically, the sector has been plagued by periodic down-cycles driven by the mismatch between large-scale capacity additions and fluctuating demand.\n\nLong-term supply agreements, however, would provide more than three years of demand visibility, allowing companies to invest with confidence and maintaining margins at reasonable levels by preventing dramatic price declines.\n\nThe defining feature of this round of LTAs is their enforceability. While long-term contracts were attempted around 2019, they carried no binding force — customers could cancel orders at will. This year's agreements, by contrast, enforce commitment through large upfront prepayments. \"My understanding is that Samsung is in discussions to receive over $10 billion in prepayments from Microsoft, with the prepayment balance drawn down if Microsoft fails to take the agreed volumes,\" said another industry source.\n\nThe successful conclusion of LTAs is expected to drive further expansion of Samsung's investment plans. With demand visibility materially improved, there is little reason to hesitate on capex. Micron, for its part, has already announced plans to invest over $25 billion in fiscal 2026 — nearly double the $13.8 billion invested the prior year.\n\nAt Samsung's annual general meeting on the 18th, Vice Chairman and CEO Jeon Young-hyun stated: \"For the semiconductor division, it is critically important to minimize uncertainty in our medium-to-long-term business and maintain a healthy supply-demand balance in memory. By securing long-term supply through multi-year agreements, both our customers and ourselves can enhance predictable business stability and visibility.\"\n\n---\n\nhttps://t.co/WfcyBuFWca", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05053203316717414, "ret_signed_m1d": 0.05053203316717414, "alpha_spy_m1d": 0.03598775376493224, "alpha_c_m1d": 0.024072545066761375, "bench_spy_m1d": 0.0145442794022419, "bench_c_m1d": 0.026459488100412765, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04386378691875292, "ret_signed_p1d": -0.04386378691875292, "alpha_spy_p1d": -0.05436380616369074, "alpha_c_p1d": -0.061070199782554124, "bench_spy_p1d": 0.01050001924493782, "bench_c_p1d": 0.0172064128638012, "ret_p1w": -0.15530854978223718, "ret_signed_p1w": -0.15530854978223718, "alpha_spy_p1w": -0.1329825397829627, "alpha_c_p1w": -0.1280434083863955, "bench_spy_p1w": -0.022326009999274476, "bench_c_p1w": -0.027265141395841686, "ret_p1m": 0.060790723610155695, "ret_signed_p1m": 0.060790723610155695, "alpha_spy_p1m": -0.031951713905671975, "alpha_c_p1m": -0.1451145714761286, "bench_spy_p1m": 0.09274243751582767, "bench_c_p1m": 0.2059052950862843, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.347, -0.3224, -0.3224, -0.3269, -0.3039, -0.308, -0.3251, -0.3251, -0.2541, -0.2619, -0.1879, -0.1971, -0.2267, -0.184, -0.1821, -0.2004, -0.2118, -0.204, -0.1422, -0.1422, -0.1369, -0.0799, -0.0599, -0.055, -0.0799, -0.0299, 0.0293, 0.0305, -0.019, 0.0352, -0.0082, -0.1029, -0.0946, -0.0667, -0.0932, -0.1174, -0.0297, -0.0211, -0.0266, -0.0266, -0.0547, -0.0046, -0.0131, 0.0125, -0.0046, -0.0116, 0.0144, -0.0174, -0.0249, -0.0242, -0.1022, -0.0523, -0.0611, -0.1244, -0.0794, -0.0468, -0.01, -0.0415, 0.0076, 0.0447, 0.0917, 0.0918, 0.0505, 0.0, -0.0439, -0.0647, -0.0965, -0.1595, -0.1553, -0.2387, -0.2008, -0.1298, -0.1336, -0.1336, -0.1064, -0.1068, -0.0378, -0.0029, -0.005, 0.0091, 0.1016, 0.0793, 0.0816, 0.0765, 0.0608, 0.0631, 0.1532, 0.1396, 0.1751, 0.2409, 0.193, 0.2265, 0.2234, 0.2827, 0.3637, 0.5145, 0.5769, 0.5297, 0.7667, 0.8814, 0.8134, 0.9011, 0.8357, 0.7143, 0.6123, 0.653, 0.7316, 0.8028, 0.7766, 0.7766, 1.1193, 1.1963, 1.1847, 1.297, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034824046987419851", "ticker_idx": 2, "ts": "2026-03-20T02:47:12+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is also said to have already reached LTAs with select customers", "tweet_type": "analysis", "timeline": "years", "summary": "Samsung, Micron, and SK Hynix all benefit from historic LTAs with hyperscalers providing multi-year demand visibility, enabling confident capex and margin stabilization.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "BIG: Samsung Pursues Long-Term Supply 'Lock-In' with Google and Microsoft… A New Memory Paradigm Emerges\n\nSamsung Electronics is reportedly in talks to establish long-term supply agreements (LTAs) with Google and Microsoft. While various contract structures are under discussion, the most likely arrangement is one where large volumes are committed over an extended period with pricing remaining variable. If finalized, this would mark the first binding long-term supply contract in the industry's history — a significant historical inflection point.\n\n◆ Big Tech Gets Desperate, Seeks Long-Term Supply Commitments\n\nAccording to industry sources on the 20th, Samsung Electronics has recently been in negotiations with Google and Microsoft over long-term supply contracts. As memory has emerged as a critical bottleneck for AI data center buildouts, major tech companies are racing to secure commitments from Korean memory makers.\n\nDespite concerns over an AI bubble, U.S. hyperscalers have continued — and even accelerated — their capital expenditure programs, exacerbating memory supply shortages. Commodity DRAM prices have surged to the point of rivaling HBM margins, prompting Big Tech to pursue LTAs with memory manufacturers in order to secure stable supply.\n\nThe exact structure of the contracts remains under discussion, but the most likely format involves fixed volume commitments with pricing linked to spot market rates. Under this arrangement, Big Tech customers would pay large upfront prepayments to Samsung, which would be drawn down if the customer fails to take the agreed volumes within the three-to-five year contract window. However, pricing would remain tied to spot prices, adjusting upward or downward if spot prices move beyond a defined range.\n\nMemory semiconductor pricing is typically divided into spot prices and contract prices. Spot prices are determined daily by immediate supply and demand in B2C markets such as distributors and small-to-medium PC assemblers. Contract prices, on the other hand, serve as the benchmark for large-scale supply deals between manufacturers like Samsung Electronics and SK Hynix and major global firms such as Apple and Amazon.\n\nUnder this type of structure, Samsung would gain long-term demand visibility, enabling it to accelerate capacity expansion while avoiding inventory buildup — thereby preventing the steep price declines seen in past cycles. \"This long-term supply agreement is essentially a green light to expand capacity,\" said one industry official. \"By committing to large volumes over an extended period, manufacturers will be able to invest with confidence.\"\n\nMicron is also inking long-term supply agreements. In its fiscal Q2 2026 earnings call, Micron disclosed that it had \"signed its first-ever five-year Strategic Customer Agreement (SCA).\" SK Hynix is also said to have already reached LTAs with select customers, and it is expected that memory manufacturers will have concluded agreements with all major hyperscalers by mid-year.\n\nIn response, a Samsung Electronics representative said, \"We are unable to confirm details related to specific customers,\" while adding, \"All major customers are currently requesting large volumes of memory.\"\n\n◆ \"This Time Is Different\" — Prepayment First, Supply Later\n\nThe significance of these contracts lies in the potential paradigm shift they represent for the memory industry. Historically, the sector has been plagued by periodic down-cycles driven by the mismatch between large-scale capacity additions and fluctuating demand.\n\nLong-term supply agreements, however, would provide more than three years of demand visibility, allowing companies to invest with confidence and maintaining margins at reasonable levels by preventing dramatic price declines.\n\nThe defining feature of this round of LTAs is their enforceability. While long-term contracts were attempted around 2019, they carried no binding force — customers could cancel orders at will. This year's agreements, by contrast, enforce commitment through large upfront prepayments. \"My understanding is that Samsung is in discussions to receive over $10 billion in prepayments from Microsoft, with the prepayment balance drawn down if Microsoft fails to take the agreed volumes,\" said another industry source.\n\nThe successful conclusion of LTAs is expected to drive further expansion of Samsung's investment plans. With demand visibility materially improved, there is little reason to hesitate on capex. Micron, for its part, has already announced plans to invest over $25 billion in fiscal 2026 — nearly double the $13.8 billion invested the prior year.\n\nAt Samsung's annual general meeting on the 18th, Vice Chairman and CEO Jeon Young-hyun stated: \"For the semiconductor division, it is critically important to minimize uncertainty in our medium-to-long-term business and maintain a healthy supply-demand balance in memory. By securing long-term supply through multi-year agreements, both our customers and ourselves can enhance predictable business stability and visibility.\"\n\n---\n\nhttps://t.co/WfcyBuFWca", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.005958295161815119, "ret_signed_m1d": 0.005958295161815119, "alpha_spy_m1d": -0.00858598424042678, "alpha_c_m1d": -0.020501192938597645, "bench_spy_m1d": 0.0145442794022419, "bench_c_m1d": 0.026459488100412765, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07348559894510664, "ret_signed_p1d": -0.07348559894510664, "alpha_spy_p1d": -0.08398561819004446, "alpha_c_p1d": -0.09069201180890785, "bench_spy_p1d": 0.01050001924493782, "bench_c_p1d": 0.0172064128638012, "ret_p1w": -0.07348559894510664, "ret_signed_p1w": -0.07348559894510664, "alpha_spy_p1w": -0.05115958894583217, "alpha_c_p1w": -0.04622045754926496, "bench_spy_p1w": -0.022326009999274476, "bench_c_p1w": -0.027265141395841686, "ret_p1m": 0.15789469763626074, "ret_signed_p1m": 0.15789469763626074, "alpha_spy_p1m": 0.06515226012043307, "alpha_c_p1m": -0.04801059745002356, "bench_spy_p1m": 0.09274243751582767, "bench_c_p1m": 0.2059052950862843, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4211, -0.4172, -0.4172, -0.4063, -0.3656, -0.3547, -0.3547, -0.3547, -0.3289, -0.3101, -0.2804, -0.2645, -0.2506, -0.2625, -0.2576, -0.2685, -0.2645, -0.2576, -0.2506, -0.2427, -0.2635, -0.2665, -0.2516, -0.2397, -0.2705, -0.207, -0.1664, -0.1466, -0.099, -0.1773, -0.101, -0.1079, -0.1654, -0.1684, -0.1208, -0.1317, -0.1476, -0.1198, -0.1277, -0.1277, -0.1277, -0.1277, -0.1138, -0.0593, -0.0574, -0.0038, 0.0091, 0.0914, 0.0536, 0.0536, -0.0675, -0.1569, -0.0655, -0.0824, -0.1698, -0.0685, -0.0516, -0.0765, -0.0963, -0.0328, -0.0367, 0.0487, 0.006, 0.0, -0.0735, -0.0209, -0.0119, -0.0735, -0.0735, -0.1331, -0.1986, -0.1092, -0.1758, -0.1301, -0.1202, -0.0904, 0.0258, -0.0089, 0.0199, 0.0328, 0.0953, 0.1281, 0.147, 0.1202, 0.1579, 0.2155, 0.2145, 0.2165, 0.2135, 0.283, 0.291, 0.284, 0.2771, 0.2771, 0.4369, 0.4369, 0.5899, 0.6425, 0.6743, 0.8669, 0.8222, 0.9623, 0.9563, 0.8064, 0.8272, 0.7329, 0.7329, 0.9265, 0.9275, 0.9275, 1.0377, 1.2274, 1.2735, 1.3172, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2008681046284661059", "ticker_idx": 0, "ts": "2026-01-06T23:24:15+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Kioxia will likely soar as well once the Japanese market opens today—similar to SNDK", "tweet_type": "prediction", "timeline": "intraday", "summary": "Kioxia expected to soar when the Japanese market opens, following SNDK's move.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@savignyxyz That said, those companies also make HBM, so they aren’t pure-play NAND names.\n\nKioxia will likely soar as well once the Japanese market opens today—similar to SNDK.", "reply_text": null, "parent_text": "@jukan05 doesnt sndk have &lt;5% market share and samsung 35+, hynix 25+, then MU and kioxa?", "quoted_text": null, "parent_author": "savignyxyz", "ret_m1d": -0.02155172413793105, "ret_signed_m1d": -0.02155172413793105, "alpha_spy_m1d": -0.015639700652614996, "alpha_c_m1d": 0.004324130837673468, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.025875854975604518, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09396551724137936, "ret_signed_p1d": 0.09396551724137936, "alpha_spy_p1d": 0.09718887403360676, "alpha_c_p1d": 0.10067313329903282, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.006707616057653465, "ret_p1w": 0.17974137931034484, "ret_signed_p1w": 0.17974137931034484, "alpha_spy_p1w": 0.17690815874200916, "alpha_c_p1w": 0.16965417093264668, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": 0.010087208377698165, "ret_p1m": 0.8017241379310345, "ret_signed_p1m": 0.8017241379310345, "alpha_spy_p1m": 0.8098477295481129, "alpha_c_p1m": 0.8161712928096743, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.01444715487863979, "ret_p3m": 0.8836206896551724, "ret_signed_p3m": 0.8836206896551724, "alpha_spy_p3m": 0.9330401563380697, "alpha_c_p3m": 0.8714953806766963, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": 0.01212530897847608, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4638, -0.469, -0.469, -0.4828, -0.4552, -0.4069, -0.4345, -0.3836, -0.3931, -0.3698, -0.3647, -0.2431, -0.1543, -0.1914, -0.131, -0.0603, -0.0668, -0.0668, -0.0724, -0.0909, -0.0086, 0.0379, 0.1483, 0.1375, 0.1414, 0.1228, -0.1358, -0.0375, -0.1155, -0.0659, -0.0228, -0.1353, -0.1353, -0.1506, -0.2771, -0.2199, -0.1891, -0.2343, -0.2054, -0.2232, -0.2181, -0.1871, -0.125, -0.1488, -0.1836, -0.1716, -0.1474, -0.2069, -0.251, -0.1983, -0.1802, -0.1948, -0.1272, -0.1421, -0.0858, -0.0694, -0.0159, -0.0797, -0.1004, -0.1004, -0.1004, -0.1004, -0.0216, 0.0, 0.094, 0.1207, 0.094, 0.094, 0.1797, 0.1448, 0.175, 0.2716, 0.3112, 0.3108, 0.4224, 0.544, 0.4944, 0.5043, 0.5905, 0.6332, 0.6591, 0.8414, 0.5828, 0.7918, 0.8017, 0.6836, 0.6461, 0.6293, 0.6246, 0.6246, 0.8254, 0.9694, 0.9224, 0.9284, 0.8466, 0.8319, 0.7724, 0.7724, 0.9198, 0.8504, 0.831, 0.8284, 0.8547, 0.7414, 0.6591, 0.75, 0.722, 0.5543, 0.6871, 0.8431, 0.8315, 0.8159, 0.9496, 0.8629, 1.0164, 0.9276, 0.9276, 0.8504, 0.819, 0.9349, 0.8246, 0.7474, 0.7259, 0.6448, 0.8793, 0.8185, 0.8836, 0.9655, 1.006, 1.3793, 1.3879, 1.5983, 1.6957, 2.0172, 1.794, 1.9198, 1.6319, 1.6302, 1.8224, 2.0, 2.05, 1.981, 2.0578, 2.131, 2.131, 2.2379, 2.1388, 2.1388, 2.1388, 2.1388, 2.7422, 2.8353, 2.9603, 2.9741, 3.3534, 3.1776, 2.8319, 3.4353, 3.2905, 3.4216, 3.7707, 3.9483, 4.6422, 4.3845, 4.2198, 4.2828, 4.6767, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2018848328373653783", "ticker_idx": 0, "ts": "2026-02-04T00:45:24+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND margin rates for Samsung and SK Hynix in the first half of this year are projected to approach the 40–50% mark", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung and SK Hynix NAND margins projected to hit 40–50% in H1 on surging enterprise SSD demand and 55–60% ASP rise; decade-high profitability outlook.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung and SK to Hit Record NAND Margins: \"Unprecedented in a Decade\"\n\nSamsung Electronics and SK Hynix are set to aggressively hike NAND flash prices through the first half of this year. Consequently, NAND margin rates for both companies are highly likely to reach the 40–50% range. The industry anticipates that NAND products will achieve their highest profitability in nearly ten years, a level not seen since the memory super-cycle of 2017.\n\nAccording to industry sources on the 4th, the NAND margin rates for Samsung and SK Hynix in the first half of this year are projected to approach the 40–50% mark.\n\nAs of the fourth quarter of last year, NAND margins for both companies were estimated to have climbed into the 20% range. Specifically, SK Hynix, which maintains a higher proportion of Quad-Level Cell (QLC) technology, demonstrated stronger profitability compared to Samsung Electronics. QLC, which stores 4 bits of data per cell (the minimum storage unit of memory), is being actively adopted in enterprise SSDs due to its advantages in high-capacity implementation.\n\nWith global Big Tech firms aggressively expanding investments in AI infrastructure, the demand for enterprise SSDs is surging. Notably, Nvidia is increasingly placing orders not only for QLC but also for Triple-Level Cell (TLC; 3 bits per cell) products.\n\nSamsung Electronics and SK Hynix plan to significantly raise NAND prices in the first half of this year, centering on high-value-added products. Market research firm TrendForce analyzed that the Average Selling Price (ASP) of NAND increased by 33–38% in the fourth quarter of last year and is expected to see an even sharper rise of 55–60% in the first quarter of this year.\n\nAccordingly, it is highly probable that NAND margins for both companies will reach 40–50% in the first half of the year. The industry regards this as an unprecedented level of profitability, even when compared to past memory super-cycles.\n\n\"NAND margin rates hitting the 40–50% range is a phenomenon unseen since 2017, when the memory super-cycle and 3D NAND growth were in full swing,\" said a semiconductor industry official. \"Achieving even a 30% margin is considered very difficult; seeing profitability soar this much in such a short period is truly remarkable.\"\n\nAnother official added, \"The outlook for NAND prices to rise in a stepwise fashion through the first and second quarters is already a certainty. The conservative approach to NAND facility investment previously taken by memory suppliers is now resulting in a severe supply shortage.\"\n\n---\n\nhttps://t.co/snql0TZzdH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009461904143741595, "ret_signed_m1d": -0.009461904143741595, "alpha_spy_m1d": -0.014329403608819735, "alpha_c_m1d": -0.038132706749456124, "bench_spy_m1d": 0.00486749946507814, "bench_c_m1d": 0.02867080260571453, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05795391982589704, "ret_signed_p1d": -0.05795391982589704, "alpha_spy_p1d": -0.04546470856969531, "alpha_c_p1d": -0.03992622672485968, "bench_spy_p1d": -0.012489211256201727, "bench_c_p1d": -0.018027693101037356, "ret_p1w": -0.007687745033678528, "ret_signed_p1w": -0.007687745033678528, "alpha_spy_p1w": -0.01609656575541596, "alpha_c_p1w": -0.042802263542479624, "bench_spy_p1w": 0.008408820721737431, "bench_c_p1w": 0.035114518508801096, "ret_p1m": 0.13305739623701607, "ret_signed_p1m": 0.13305739623701607, "alpha_spy_p1m": 0.14016911495559026, "alpha_c_p1m": 0.11814277395181372, "bench_spy_p1m": -0.007111718718574189, "bench_c_p1m": 0.01491462228520235, "ret_p3m": 0.37777190949001516, "ret_signed_p3m": 0.37777190949001516, "alpha_spy_p3m": 0.3285422371690374, "alpha_c_p3m": 0.2030578891344874, "bench_spy_p3m": 0.049229672320977746, "bench_c_p3m": 0.17471402035552774, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4239, -0.408, -0.3909, -0.3933, -0.395, -0.428, -0.408, -0.4244, -0.4321, -0.408, -0.4421, -0.4309, -0.4156, -0.395, -0.3909, -0.4086, -0.4068, -0.3915, -0.385, -0.3815, -0.3621, -0.3556, -0.3621, -0.3644, -0.3685, -0.3591, -0.3832, -0.395, -0.365, -0.3668, -0.3744, -0.3497, -0.3438, -0.3462, -0.3462, -0.3114, -0.2933, -0.291, -0.291, -0.291, -0.2401, -0.1833, -0.1786, -0.1662, -0.1792, -0.178, -0.1792, -0.1863, -0.1703, -0.149, -0.1195, -0.1171, -0.1413, -0.1159, -0.0993, -0.1005, -0.1005, -0.0568, -0.0396, -0.0497, -0.0509, -0.1106, -0.0095, 0.0, -0.058, -0.0621, -0.016, -0.0195, -0.0077, 0.0562, 0.0716, 0.0716, 0.0716, 0.0716, 0.1236, 0.1242, 0.1413, 0.1827, 0.2034, 0.2892, 0.2803, 0.2803, 0.1538, 0.0183, 0.1331, 0.113, 0.026, 0.1112, 0.1236, 0.1112, 0.0852, 0.1159, 0.1467, 0.233, 0.1857, 0.1792, 0.1017, 0.1218, 0.1177, 0.0651, 0.0651, 0.0447, -0.0092, 0.1238, 0.0572, 0.1034, 0.1443, 0.1644, 0.2474, 0.2089, 0.2207, 0.1911, 0.2237, 0.2504, 0.2889, 0.28, 0.2711, 0.2978, 0.2889, 0.3304, 0.3007, 0.3304, 0.3155, 0.3393, 0.3067, 0.3067, 0.3778, 0.3778, 0.5763, 0.6089, 0.5911, 0.6918, 0.6533, 0.683, 0.7541, 0.603, 0.6652, 0.6326, 0.6355, 0.7748, 0.7333, 0.7333, 0.7718, 0.8193, 0.7748, 0.8785, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2018848328373653783", "ticker_idx": 1, "ts": "2026-02-04T00:45:24+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND margin rates for both companies are highly likely to reach the 40–50% range", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung and SK Hynix NAND margins projected to hit 40–50% in H1 on surging enterprise SSD demand and 55–60% ASP rise; decade-high profitability outlook.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung and SK to Hit Record NAND Margins: \"Unprecedented in a Decade\"\n\nSamsung Electronics and SK Hynix are set to aggressively hike NAND flash prices through the first half of this year. Consequently, NAND margin rates for both companies are highly likely to reach the 40–50% range. The industry anticipates that NAND products will achieve their highest profitability in nearly ten years, a level not seen since the memory super-cycle of 2017.\n\nAccording to industry sources on the 4th, the NAND margin rates for Samsung and SK Hynix in the first half of this year are projected to approach the 40–50% mark.\n\nAs of the fourth quarter of last year, NAND margins for both companies were estimated to have climbed into the 20% range. Specifically, SK Hynix, which maintains a higher proportion of Quad-Level Cell (QLC) technology, demonstrated stronger profitability compared to Samsung Electronics. QLC, which stores 4 bits of data per cell (the minimum storage unit of memory), is being actively adopted in enterprise SSDs due to its advantages in high-capacity implementation.\n\nWith global Big Tech firms aggressively expanding investments in AI infrastructure, the demand for enterprise SSDs is surging. Notably, Nvidia is increasingly placing orders not only for QLC but also for Triple-Level Cell (TLC; 3 bits per cell) products.\n\nSamsung Electronics and SK Hynix plan to significantly raise NAND prices in the first half of this year, centering on high-value-added products. Market research firm TrendForce analyzed that the Average Selling Price (ASP) of NAND increased by 33–38% in the fourth quarter of last year and is expected to see an even sharper rise of 55–60% in the first quarter of this year.\n\nAccordingly, it is highly probable that NAND margins for both companies will reach 40–50% in the first half of the year. The industry regards this as an unprecedented level of profitability, even when compared to past memory super-cycles.\n\n\"NAND margin rates hitting the 40–50% range is a phenomenon unseen since 2017, when the memory super-cycle and 3D NAND growth were in full swing,\" said a semiconductor industry official. \"Achieving even a 30% margin is considered very difficult; seeing profitability soar this much in such a short period is truly remarkable.\"\n\nAnother official added, \"The outlook for NAND prices to rise in a stepwise fashion through the first and second quarters is already a certainty. The conservative approach to NAND facility investment previously taken by memory suppliers is now resulting in a severe supply shortage.\"\n\n---\n\nhttps://t.co/snql0TZzdH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007777776386093338, "ret_signed_m1d": 0.007777776386093338, "alpha_spy_m1d": 0.002910276921015198, "alpha_c_m1d": -0.033214852566495034, "bench_spy_m1d": 0.00486749946507814, "bench_c_m1d": 0.04099262895258837, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0644443832103293, "ret_signed_p1d": -0.0644443832103293, "alpha_spy_p1d": -0.05195517195412758, "alpha_c_p1d": -0.061957650272856934, "bench_spy_p1d": -0.012489211256201727, "bench_c_p1d": -0.002486732937472369, "ret_p1w": -0.04444441661075571, "ret_signed_p1w": -0.04444441661075571, "alpha_spy_p1w": -0.05285323733249314, "alpha_c_p1w": -0.13019911717029264, "bench_spy_p1w": 0.008408820721737431, "bench_c_p1w": 0.08575470055953693, "ret_p1m": 0.04748489918011689, "ret_signed_p1m": 0.04748489918011689, "alpha_spy_p1m": 0.05459661789869108, "alpha_c_p1m": 0.012591392365826293, "bench_spy_p1m": -0.007111718718574189, "bench_c_p1m": 0.0348935068142906, "ret_p3m": 0.6107446110151544, "ret_signed_p3m": 0.6107446110151544, "alpha_spy_p3m": 0.5615149386941767, "alpha_c_p3m": 0.2841386136342565, "bench_spy_p3m": 0.049229672320977746, "bench_c_p3m": 0.3266059973808979, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.356, -0.3271, -0.3127, -0.3149, -0.3205, -0.3782, -0.3271, -0.3671, -0.376, -0.366, -0.4215, -0.4226, -0.4237, -0.4182, -0.3956, -0.4111, -0.4022, -0.38, -0.3867, -0.3978, -0.3956, -0.3589, -0.3711, -0.3478, -0.3722, -0.3656, -0.3844, -0.4111, -0.3878, -0.3867, -0.3922, -0.3556, -0.3511, -0.3467, -0.3467, -0.3344, -0.2889, -0.2767, -0.2767, -0.2767, -0.2478, -0.2267, -0.1933, -0.1756, -0.16, -0.1733, -0.1678, -0.18, -0.1756, -0.1678, -0.16, -0.1511, -0.1744, -0.1778, -0.1611, -0.1478, -0.1822, -0.1111, -0.0656, -0.0433, 0.01, -0.0778, 0.0078, 0.0, -0.0644, -0.0678, -0.0144, -0.0267, -0.0444, -0.0133, -0.0222, -0.0222, -0.0222, -0.0222, -0.0067, 0.0544, 0.0567, 0.1167, 0.1311, 0.2234, 0.1811, 0.1811, 0.0453, -0.0549, 0.0475, 0.0286, -0.0694, 0.0441, 0.0631, 0.0352, 0.013, 0.0842, 0.0798, 0.1755, 0.1276, 0.121, 0.0386, 0.0976, 0.1076, 0.0386, 0.0386, -0.0282, -0.1017, -0.0015, -0.0761, -0.0249, -0.0137, 0.0197, 0.1499, 0.1109, 0.1432, 0.1577, 0.2278, 0.2646, 0.2857, 0.2556, 0.2979, 0.3625, 0.3614, 0.3636, 0.3603, 0.4382, 0.4471, 0.4393, 0.4315, 0.4315, 0.6107, 0.6107, 0.7822, 0.8412, 0.8768, 1.0927, 1.0427, 1.1996, 1.1929, 1.0248, 1.0482, 0.9425, 0.9425, 1.1595, 1.1606, 1.1606, 1.2842, 1.4968, 1.5485, 1.5974, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027314509740970335", "ticker_idx": 0, "ts": "2026-02-27T09:26:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK hynix have informed their customers of plans to significantly raise DRAM prices again in Q2… cumulative price increases could reach 130% for the full year", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM prices up 130% by year-end; Samsung and SK Hynix raising prices into Q2 with volumes sold out through next year — structural bull case for both memory makers.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung & SK Hynix to Sharply Raise DRAM Prices in Q2… Surge Expected to Continue Through Next Year\n\nThe world's two largest DRAM manufacturers, Samsung Electronics and SK hynix, have informed their customers of plans to significantly raise DRAM prices again in Q2. With DRAM demand surging on the back of expanding global AI investment far outpacing the two companies' supply capacity, some observers project that cumulative price increases could reach 130% for the full year.\n\nAccording to semiconductor industry sources on the 27th, Samsung Electronics and SK hynix have notified customers of Q2 DRAM supply price increases and have begun price negotiations. While the magnitude of increases varies by customer, non-major clients are reportedly being told they must accept substantial price hikes to secure DRAM volumes.\n\nAn industry source said, \"Samsung Electronics' current production capacity can only cover about 60% of continuously surging DRAM demand,\" adding, \"Customers are also more focused on securing volumes than resisting price increases, so there should be no difficulty pushing through significant hikes.\"\n\nSome in the industry suggest that certain customers may need to accept price increases of more than double their previous contract prices to receive DRAM allocations. The DRAM price surge that began in H2 last year amid the AI infrastructure buildout frenzy has completely upended traditional trading practices, widening the purchasing power gap between large and small-to-mid-sized customers.\n\nTypically, global big tech companies such as NVIDIA and Apple hold structural advantages in DRAM procurement. Big tech firms purchasing DRAM in large volumes negotiate contracts on a quarterly or semi-annual basis, setting volume and price ranges with final supply prices adjusted to market rates at the time of delivery. Samsung Electronics and SK hynix also extend preferential pricing to big tech customers who purchase at scale regardless of market cycles, in order to maintain their supply chain relationships.\n\nHowever, small-to-mid-sized customers have traditionally contracted on a monthly basis reflecting spot market prices. Since Q4 last year, when DRAM prices began surging, Samsung and SK hynix have shifted these customers to quarterly volume-based \"forward commitment\" contracts that favor the suppliers. As a result, Samsung and SK hynix have presented steep price increase proposals to customers whose contracts come up for renewal in Q2.\n\nAnother industry source explained, \"Contract volumes for both large and small customers will see price increases reflecting market prices,\" adding, \"However, when existing contracts expire and new ones are signed, the absence of a previously agreed price range means the magnitude of increases will inevitably be even greater.\"\n\nAnalysis suggests DRAM prices could more than double over the course of this year. On the same day, global market research firm Gartner released a forecast projecting that memory prices, including DRAM and solid-state drives (SSDs), will rise 130% by year-end.\n\nAccording to DRAMeXchange, DDR4 8Gb product prices, which stood at around $1.30 in March last year, surged to $9.30 by the end of last year, and have climbed a further ~40% to $13.00 as of February this year.\n\nThe industry views the price surge not as a temporary supply-demand imbalance but as a structural problem where supply chronically cannot keep pace with demand. This is driven by the AI industry's transition from simple search-based applications to \"agentic AI\" that replaces specialized human functions. As agentic AI is adopted across enterprises, governments, and even personal smartphones, demand for DRAM—which supports inference and computation—continues to explode.\n\nMeanwhile, the three major memory makers' DRAM production capacity this year is expected to remain flat or increase by only about 7% (SK hynix) compared to last year. An industry source said, \"As companies adopt not just one but two or three AI models simultaneously for different use cases, DRAM demand is growing even further,\" adding, \"Volumes through next year are already sold out, so the price increase trend is expected to continue.\"\n\n---\n\nhttps://t.co/RUVTOu10cX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006928427511213364, "ret_signed_m1d": 0.006928427511213364, "alpha_spy_m1d": 0.002103314332988271, "alpha_c_m1d": -0.009286633526067067, "bench_spy_m1d": 0.004825113178225093, "bench_c_m1d": 0.01621506103728043, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0005685787823312971, "alpha_c_p1d": -0.005621260796333072, "bench_spy_p1d": 0.0005685787823312971, "bench_c_p1d": 0.005621260796333072, "ret_p1w": -0.13071594184870883, "ret_signed_p1w": -0.13071594184870883, "alpha_spy_p1w": -0.11087599503620715, "alpha_c_p1w": -0.12012214160776147, "bench_spy_p1w": -0.019839946812501674, "bench_c_p1w": -0.01059380024094736, "ret_p1m": -0.18399581120935415, "ret_signed_p1m": -0.18399581120935415, "alpha_spy_p1m": -0.10776423325557938, "alpha_c_p1m": -0.10399102837576302, "bench_spy_p1m": -0.07623157795377478, "bench_c_p1m": -0.08000478283359114, "ret_p3m": 0.4209488710081013, "ret_signed_p3m": 0.4209488710081013, "alpha_spy_p3m": 0.32398018191809586, "alpha_c_p3m": 0.09016682266492815, "bench_spy_p3m": 0.09696868909000544, "bench_c_p3m": 0.33078204834317315, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5247, -0.5197, -0.5169, -0.5017, -0.4967, -0.5017, -0.5036, -0.5068, -0.4994, -0.5183, -0.5275, -0.504, -0.5054, -0.5114, -0.4921, -0.4875, -0.4893, -0.4893, -0.4622, -0.448, -0.4462, -0.4462, -0.4462, -0.4065, -0.3621, -0.3584, -0.3487, -0.3589, -0.358, -0.3589, -0.3644, -0.352, -0.3353, -0.3122, -0.3104, -0.3293, -0.3095, -0.2965, -0.2975, -0.2975, -0.2633, -0.2499, -0.2577, -0.2587, -0.3053, -0.2263, -0.2189, -0.2642, -0.2674, -0.2314, -0.2342, -0.2249, -0.1751, -0.163, -0.163, -0.163, -0.163, -0.1224, -0.1219, -0.1085, -0.0762, -0.06, 0.0069, 0.0, 0.0, -0.0988, -0.2046, -0.115, -0.1307, -0.1986, -0.1321, -0.1224, -0.1321, -0.1524, -0.1284, -0.1044, -0.037, -0.0739, -0.079, -0.1395, -0.1238, -0.127, -0.1681, -0.1681, -0.184, -0.2261, -0.1222, -0.1743, -0.1382, -0.1062, -0.0905, -0.0257, -0.0558, -0.0465, -0.0697, -0.0442, -0.0234, 0.0067, -0.0002, -0.0072, 0.0136, 0.0067, 0.0391, 0.016, 0.0391, 0.0275, 0.046, 0.0206, 0.0206, 0.0761, 0.0761, 0.2312, 0.2566, 0.2428, 0.3214, 0.2914, 0.3145, 0.37, 0.252, 0.3006, 0.2752, 0.2775, 0.3862, 0.3538, 0.3538, 0.3839, 0.4209, 0.3862, 0.4672, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027314509740970335", "ticker_idx": 1, "ts": "2026-02-27T09:26:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Volumes through next year are already sold out, so the price increase trend is expected to continue", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM prices up 130% by year-end; Samsung and SK Hynix raising prices into Q2 with volumes sold out through next year — structural bull case for both memory makers.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung & SK Hynix to Sharply Raise DRAM Prices in Q2… Surge Expected to Continue Through Next Year\n\nThe world's two largest DRAM manufacturers, Samsung Electronics and SK hynix, have informed their customers of plans to significantly raise DRAM prices again in Q2. With DRAM demand surging on the back of expanding global AI investment far outpacing the two companies' supply capacity, some observers project that cumulative price increases could reach 130% for the full year.\n\nAccording to semiconductor industry sources on the 27th, Samsung Electronics and SK hynix have notified customers of Q2 DRAM supply price increases and have begun price negotiations. While the magnitude of increases varies by customer, non-major clients are reportedly being told they must accept substantial price hikes to secure DRAM volumes.\n\nAn industry source said, \"Samsung Electronics' current production capacity can only cover about 60% of continuously surging DRAM demand,\" adding, \"Customers are also more focused on securing volumes than resisting price increases, so there should be no difficulty pushing through significant hikes.\"\n\nSome in the industry suggest that certain customers may need to accept price increases of more than double their previous contract prices to receive DRAM allocations. The DRAM price surge that began in H2 last year amid the AI infrastructure buildout frenzy has completely upended traditional trading practices, widening the purchasing power gap between large and small-to-mid-sized customers.\n\nTypically, global big tech companies such as NVIDIA and Apple hold structural advantages in DRAM procurement. Big tech firms purchasing DRAM in large volumes negotiate contracts on a quarterly or semi-annual basis, setting volume and price ranges with final supply prices adjusted to market rates at the time of delivery. Samsung Electronics and SK hynix also extend preferential pricing to big tech customers who purchase at scale regardless of market cycles, in order to maintain their supply chain relationships.\n\nHowever, small-to-mid-sized customers have traditionally contracted on a monthly basis reflecting spot market prices. Since Q4 last year, when DRAM prices began surging, Samsung and SK hynix have shifted these customers to quarterly volume-based \"forward commitment\" contracts that favor the suppliers. As a result, Samsung and SK hynix have presented steep price increase proposals to customers whose contracts come up for renewal in Q2.\n\nAnother industry source explained, \"Contract volumes for both large and small customers will see price increases reflecting market prices,\" adding, \"However, when existing contracts expire and new ones are signed, the absence of a previously agreed price range means the magnitude of increases will inevitably be even greater.\"\n\nAnalysis suggests DRAM prices could more than double over the course of this year. On the same day, global market research firm Gartner released a forecast projecting that memory prices, including DRAM and solid-state drives (SSDs), will rise 130% by year-end.\n\nAccording to DRAMeXchange, DDR4 8Gb product prices, which stood at around $1.30 in March last year, surged to $9.30 by the end of last year, and have climbed a further ~40% to $13.00 as of February this year.\n\nThe industry views the price surge not as a temporary supply-demand imbalance but as a structural problem where supply chronically cannot keep pace with demand. This is driven by the AI industry's transition from simple search-based applications to \"agentic AI\" that replaces specialized human functions. As agentic AI is adopted across enterprises, governments, and even personal smartphones, demand for DRAM—which supports inference and computation—continues to explode.\n\nMeanwhile, the three major memory makers' DRAM production capacity this year is expected to remain flat or increase by only about 7% (SK hynix) compared to last year. An industry source said, \"As companies adopt not just one but two or three AI models simultaneously for different use cases, DRAM demand is growing even further,\" adding, \"Volumes through next year are already sold out, so the price increase trend is expected to continue.\"\n\n---\n\nhttps://t.co/RUVTOu10cX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03581524762445554, "ret_signed_m1d": 0.03581524762445554, "alpha_spy_m1d": 0.030990134446230444, "alpha_c_m1d": 0.02193623412393597, "bench_spy_m1d": 0.004825113178225093, "bench_c_m1d": 0.013879013500519566, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0005685787823312971, "alpha_c_p1d": -4.926429483109729e-05, "bench_spy_p1d": 0.0005685787823312971, "bench_c_p1d": 4.926429483109729e-05, "ret_p1w": -0.12912349508005638, "ret_signed_p1w": -0.12912349508005638, "alpha_spy_p1w": -0.10928354826755471, "alpha_c_p1w": -0.06560995355701471, "bench_spy_p1w": -0.019839946812501674, "bench_c_p1w": -0.06351354152304167, "ret_p1m": -0.17719132121640035, "ret_signed_p1m": -0.17719132121640035, "alpha_spy_p1m": -0.10095974326262558, "alpha_c_p1m": -0.06930934945505463, "bench_spy_p1m": -0.07623157795377478, "bench_c_p1m": -0.10788197176134573, "ret_p3m": 1.114043335001457, "ret_signed_p3m": 1.114043335001457, "alpha_spy_p3m": 1.0170746459114515, "alpha_c_p3m": 0.6486300240893856, "bench_spy_p3m": 0.09696868909000544, "bench_c_p3m": 0.4654133109120713, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.475, -0.4807, -0.4901, -0.4882, -0.4572, -0.4675, -0.4478, -0.4685, -0.4628, -0.4788, -0.5014, -0.4816, -0.4807, -0.4854, -0.4544, -0.4506, -0.4468, -0.4468, -0.4365, -0.3979, -0.3876, -0.3876, -0.3876, -0.3631, -0.3452, -0.317, -0.3019, -0.2888, -0.3001, -0.2954, -0.3057, -0.3019, -0.2954, -0.2888, -0.2813, -0.301, -0.3038, -0.2897, -0.2784, -0.3076, -0.2474, -0.2088, -0.19, -0.1448, -0.2192, -0.1467, -0.1533, -0.2079, -0.2107, -0.1655, -0.1759, -0.1909, -0.1646, -0.1721, -0.1721, -0.1721, -0.1721, -0.159, -0.1072, -0.1053, -0.0545, -0.0423, 0.0358, 0.0, 0.0, -0.115, -0.1998, -0.1131, -0.1291, -0.2121, -0.1159, -0.0999, -0.1235, -0.1423, -0.082, -0.0858, -0.0047, -0.0452, -0.0509, -0.1206, -0.0707, -0.0622, -0.1206, -0.1206, -0.1772, -0.2394, -0.1546, -0.2177, -0.1744, -0.1649, -0.1367, -0.0264, -0.0594, -0.032, -0.0198, 0.0396, 0.0707, 0.0886, 0.0631, 0.099, 0.1536, 0.1527, 0.1546, 0.1517, 0.2177, 0.2253, 0.2187, 0.2121, 0.2121, 0.3638, 0.3638, 0.509, 0.5589, 0.5891, 0.7719, 0.7295, 0.8624, 0.8567, 0.7144, 0.7342, 0.6447, 0.6447, 0.8285, 0.8294, 0.8294, 0.934, 1.114, 1.1578, 1.1992, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2034824813160931672", "ticker_idx": 0, "ts": "2026-03-20T02:50:15+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory is finally breaking out of the cycle!", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish Samsung on structural shift: hyperscalers paying upfront prepayments signals memory escaping its commodity cycle.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Did you see this? Hyperscalers are now literally paying upfront prepayments to memory companies! This is it — memory is finally breaking out of the cycle!\n\n\"Under this arrangement, Big Tech customers would pay large upfront prepayments to Samsung, which would be drawn down if the customer fails to take the agreed volumes within the three-to-five year contract window.\"", "reply_text": null, "parent_text": null, "quoted_text": "BIG: Samsung Pursues Long-Term Supply 'Lock-In' with Google and Microsoft… A New Memory Paradigm Emerges\n\nSamsung Electronics is reportedly in talks to establish long-term supply agreements (LTAs) with Google and Microsoft. While various contract structures are under discussion, https://t.co/nxdAn9g2V4", "parent_author": null, "ret_m1d": 0.005516582333744857, "ret_signed_m1d": 0.005516582333744857, "alpha_spy_m1d": -0.009027697068497043, "alpha_c_m1d": -0.01769276387541341, "bench_spy_m1d": 0.0145442794022419, "bench_c_m1d": 0.023209346209158266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06569707363190647, "ret_signed_p1d": -0.06569707363190647, "alpha_spy_p1d": -0.07619709287684429, "alpha_c_p1d": -0.07922561088422442, "bench_spy_p1d": 0.01050001924493782, "bench_c_p1d": 0.013528537252317951, "ret_p1w": -0.0967903235431763, "ret_signed_p1w": -0.0967903235431763, "alpha_spy_p1w": -0.07446431354390182, "alpha_c_p1w": -0.058291810496550855, "bench_spy_p1w": -0.022326009999274476, "bench_c_p1w": -0.03849851304662544, "ret_p1m": 0.07795377236706957, "ret_signed_p1m": 0.07795377236706957, "alpha_spy_p1m": -0.014788665148758096, "alpha_c_p1m": -0.06590144292356537, "bench_spy_p1m": 0.09274243751582767, "bench_c_p1m": 0.14385521529063494, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4435, -0.4455, -0.4455, -0.4161, -0.4007, -0.3987, -0.3987, -0.3987, -0.3556, -0.3074, -0.3034, -0.2929, -0.3039, -0.3029, -0.3039, -0.3099, -0.2964, -0.2783, -0.2533, -0.2513, -0.2718, -0.2503, -0.2362, -0.2372, -0.2372, -0.2001, -0.1856, -0.1941, -0.1951, -0.2457, -0.16, -0.152, -0.2011, -0.2046, -0.1655, -0.1685, -0.1585, -0.1043, -0.0913, -0.0913, -0.0913, -0.0913, -0.0471, -0.0466, -0.0321, 0.003, 0.0206, 0.0933, 0.0858, 0.0858, -0.0216, -0.1364, -0.0391, -0.0562, -0.1299, -0.0577, -0.0471, -0.0577, -0.0797, -0.0537, -0.0276, 0.0456, 0.0055, 0.0, -0.0657, -0.0486, -0.0522, -0.0968, -0.0968, -0.114, -0.1597, -0.0469, -0.1035, -0.0643, -0.0296, -0.0125, 0.0579, 0.0252, 0.0352, 0.0101, 0.0378, 0.0604, 0.093, 0.0855, 0.078, 0.1006, 0.093, 0.1282, 0.1031, 0.1282, 0.1156, 0.1357, 0.1081, 0.1081, 0.1684, 0.1684, 0.3368, 0.3644, 0.3493, 0.4348, 0.4021, 0.4272, 0.4875, 0.3594, 0.4121, 0.3845, 0.387, 0.5051, 0.4699, 0.4699, 0.5026, 0.5428, 0.5051, 0.5931, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2032013723888714090", "ticker_idx": 0, "ts": "2026-03-12T08:39:59+00:00", "symbol": "ASX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Initiating with Buy, TP NT$420... We view ASE as a core AI beneficiary, with its margin-accretive LEAP (advanced packaging) business growing rapidly... ASE has pulled back ~12% over the past two weeks due to systemic risk, and we see current valuations as attractive.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities initiates ASE (ASX US / 3711TT) at Buy, PT NT$420, on LEAP advanced packaging growth to $4.7B by 2027 and OSAT upcycle driving margin expansion.", "resolved_tickers": ["ASX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics & Communications》 | ASE (3711TT, Buy) Initiation: Packaging the AI Boom\n\nInitiating with Buy, TP NT$420: We initiate coverage on ASE (3711TT / ASX US) with a Buy rating and a target price of NT$420, based on 25x 2026/2027E EPS (or 22x 2027E EPS). We view ASE as a core AI beneficiary, with its margin-accretive LEAP (advanced packaging) business growing rapidly toward revenue of US$3.2bn / US$4.7bn in 2026E / 2027E. Meanwhile, driven by advanced node demand, traditional business recovery, and pricing increases, the OSAT upcycle should be sustained, keeping utilization and margins elevated. Given tight CoWoS supply, we expect TSMC to raise CoWoS prices 10–15% in 2026, acting as a sector catalyst. We further expect AMD-driven Full Process business and CoWoP to become the next wave of growth from 2027 onward. We forecast ASE's 2026/2027E EPS at NT$14.6 / NT$19.2, up +57% / +31% YoY. ASE has pulled back ~12% over the past two weeks due to systemic risk, and we see current valuations as attractive.\n\nMultiple Tailwinds for LEAP: On February 5th, management raised its 2026 LEAP revenue guidance to US$3.2bn — above the prior \"at least US$1bn incremental\" commentary — representing a doubling from 2025's US$1.6bn. We forecast LEAP revenue of US$1.6bn / US$3.2bn / US$4.7bn in 2025/2026/2027E, representing 13% / 20% / 24% of ATM revenue, driven by: 1) oS (outsourced substrate/interposer) — the primary driver, benefiting from larger interposer sizes and rising interconnect complexity alongside TSMC outsourcing and Rubin platform ramp; 2) CP (wafer probe/testing) — benefiting from TSMC outsourcing probe testing for most AI chips including NVIDIA; 3) FT (final test) — orders from non-NVIDIA customers already secured from 2025, with potential NVIDIA/Google supply chain entry expected in 2H26; 4) Full Process — forecast to triple YoY in 2026, reaching ~10% of LEAP revenue, with further acceleration in 2027 driven by AMD Venice CPU adoption in general servers and agentic AI applications. Finally, we continue to expect NVIDIA CoWoP to enter mass production in 2H27, with ASE/SPIL as the primary packaging suppliers — commanding higher value-add given the novel processes and equipment required.\n\nStructural Gross Margin Improvement from OSAT Upcycle: Driven by AI demand and chiplet design trends, we expect structural growth in the OSAT industry to persist. Accordingly, we expect LEAP acceleration to drive a meaningful improvement in ATM profitability in 2026, with management guiding for ATM gross margin to remain within a structural range of 24–30% and improve sequentially each quarter, reaching the upper end of that range by 2H26. Traditional business is also expected to recover modestly on restocking and industrial demand recovery, although mobile/PC weakness remains a watch item. Overall, we forecast ASE's consolidated and ATM-level gross/operating margins to trend upward.\n\n$ASX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04335397128799645, "ret_signed_m1d": 0.04335397128799645, "alpha_spy_m1d": 0.02793488437220426, "alpha_c_m1d": 0.010117700620442038, "bench_spy_m1d": 0.01541908691579219, "bench_c_m1d": 0.03323627066755441, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02429729559042615, "ret_signed_p1d": 0.02429729559042615, "alpha_spy_p1d": 0.02995740509143252, "alpha_c_p1d": 0.026358507814377874, "bench_spy_p1d": -0.005660109501006372, "bench_c_p1d": -0.002061212223951725, "ret_p1w": 0.04335397128799645, "ret_signed_p1w": 0.04335397128799645, "alpha_spy_p1w": 0.05275247840988051, "alpha_c_p1w": 0.025859810930638938, "bench_spy_p1w": -0.009398507121884059, "bench_c_p1w": 0.017494160357357513, "ret_p1m": 0.18342069668658367, "ret_signed_p1m": 0.18342069668658367, "alpha_spy_p1m": 0.16051636098150968, "alpha_c_p1m": 0.05781845159677412, "bench_spy_p1m": 0.022904335705073997, "bench_c_p1m": 0.12560224508980955, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2577, -0.2711, -0.2973, -0.2916, -0.2778, -0.2687, -0.2601, -0.2592, -0.2592, -0.2511, -0.2368, -0.2363, -0.233, -0.233, -0.1968, -0.192, -0.1663, -0.1606, -0.1744, -0.1596, -0.1296, -0.1091, -0.1005, -0.0858, -0.0753, -0.0753, -0.0996, -0.0958, -0.0919, -0.0762, -0.0553, -0.0438, -0.0348, -0.0576, -0.0958, -0.0753, -0.0762, -0.0953, -0.0348, -0.0048, 0.0553, 0.0715, 0.131, 0.1177, 0.1148, 0.1148, 0.1182, 0.112, 0.1043, 0.1415, 0.1186, 0.1825, 0.1963, 0.1544, 0.1572, 0.1529, 0.0629, 0.0529, 0.051, 0.0062, 0.0319, 0.0343, 0.0434, 0.0, 0.0243, 0.0253, 0.04, 0.0319, 0.0434, 0.0152, 0.0329, 0.0257, 0.0672, 0.021, 0.0243, -0.0048, 0.0329, 0.071, 0.0615, 0.0615, 0.0743, 0.0572, 0.1501, 0.1677, 0.1834, 0.2649, 0.2801, 0.2973, 0.3149, 0.3621, 0.395, 0.4102, 0.4269, 0.4202, 0.5245, 0.4697, 0.4331, 0.4574, 0.4964, 0.505, 0.5379, 0.5931, 0.6274, 0.5884, 0.6308, 0.6818, 0.6332, 0.6899, 0.6794, 0.6108, 0.5079, 0.4717, 0.5098, 0.555, 0.6584, 0.6584, 0.8556, 0.8614, 0.9343, 0.8271, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030582686876844472", "ticker_idx": 0, "ts": "2026-03-08T09:53:33+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron Eliminated from Vera Rubin", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Micron excluded from Vera Rubin HBM4 supply chain; Samsung and SK Hynix confirmed as sole HBM4 suppliers, securing dominance for 1-2 years.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA's Vera Rubin to Use Only Samsung and SK Hynix HBM4\n\nSamsung Electronics and SK Hynix's sixth-generation High Bandwidth Memory (HBM4) will be incorporated into NVIDIA's next-generation AI accelerator \"Vera Rubin,\" slated for release in the second half of this year. Micron, the world's third-largest memory company, has been excluded from the HBM4 supply chain for Vera Rubin — a result widely interpreted as a recognition of Korean memory makers' technological superiority in key HBM4 performance metrics such as operating speed and bandwidth.\n\nAccording to industry sources on March 8, Samsung Electronics and SK Hynix have been confirmed as suppliers on NVIDIA's Vera Rubin vendor list. The two companies have been provisionally selected as HBM4 suppliers for the flagship next-gen AI accelerator, whose performance hinges critically on this component. HBM4 is an AI server memory chip manufactured by stacking 8 to 16 layers of advanced DRAM built on 11–13nm process nodes atop a base die, enabling rapid delivery of large volumes of data to the GPU in AI accelerators.\n\nMicron, which currently supplies HBM3E to NVIDIA, has been dropped from the Vera Rubin HBM4 vendor list. The company is expected to supply HBM4 for mid-tier AI accelerators optimized for inference — such as the \"Rubin CPX\" — rather than for Vera Rubin itself.\n\nThe selection of Samsung and SK Hynix as the sole HBM4 suppliers reflects their ability to meet NVIDIA's performance and yield requirements. Samsung has effectively passed NVIDIA's bifurcated HBM4 qualification tests targeting operating speeds of 10Gb/s and 11Gb/s. SK Hynix is in the process of optimizing its product for the 11Gb/s test. Samsung's foundry division has also won an order to produce NVIDIA's GPU \"RTX 3060,\" which the company has decided to resume manufacturing, with production set to begin shortly on an 8nm process.\n\nOnly Samsung and SK Hynix at the Heart of NVIDIA's \"Monster AI Chip\" — Micron Eliminated\n\nSince around 2022, when the AI era began, NVIDIA has been the company that determines the fate of memory semiconductor firms. Those that made it into NVIDIA's HBM supply chain became central players in the AI industry; those that didn't saw their standing diminish. Samsung was a prime example — the two-year crisis narrative that plagued Samsung Semiconductor was triggered by delays in HBM3 supply to NVIDIA and only subsided last September when Samsung passed NVIDIA's HBM3E 12-layer qualification test.\n\nAgainst this backdrop, the selection of Samsung and SK Hynix — and exclusion of Micron — as Vera Rubin HBM4 suppliers is being assessed as highly significant. It opens the door to large-scale NVIDIA shipments over the next one to two years, securing their HBM dominance.\n\nHigh-Spec HBM4 Orders\n\nAccording to industry sources on March 8, Vera Rubin hardware will be unveiled for the first time at NVIDIA's developer conference GTC 2026, scheduled for March 16 in Silicon Valley. While no official launch date has been set, a second-half 2026 release is expected. NVIDIA is putting everything into making Vera Rubin's performance more than five times greater than its predecessor, in order to decisively outpace rivals including AMD and Broadcom. More than 80 partner companies worldwide are aligning with NVIDIA to support the launch of this \"monster AI accelerator.\"\n\nSince last year, NVIDIA has identified HBM4 as the critical component underpinning Vera Rubin's success, pushing memory companies to develop high-performance products. It demanded operating speeds exceeding 10Gb/s — well above the 8Gb/s JEDEC standard — for Vera Rubin's HBM4. Capacity has also been scaled up: Vera Rubin will incorporate 16 HBM4 stacks totaling 576GB, surpassing the 432GB HBM4 capacity in AMD's competing next-gen AI accelerator, the MI450.\n\nMicron Eliminated from Vera Rubin\n\nGlobal memory companies competed fiercely for Vera Rubin HBM4 supply contracts. Winning over NVIDIA — which commands over 80% share in the AI accelerator market where HBM is heavily integrated — means both technological validation and guaranteed strong earnings.\n\nThe race for Vera Rubin HBM4 has narrowed to Samsung and SK Hynix. Micron does not appear on the vendor list. \"Micron isn't even being discussed as a Vera Rubin HBM4 supplier,\" one industry source said.\n\nBetween the two, Samsung has recently pulled ahead. Samsung has effectively passed NVIDIA's bifurcated qualification tests for both 10Gb/s and 11Gb/s variants of HBM4, and began shipping finished products to NVIDIA last month, albeit in limited volumes. SK Hynix is continuing product optimization with NVIDIA to pass the 11Gb/s test. Given that the process from HBM4 DRAM wafer input to final packaging takes over six months, both companies are expected to begin full-scale HBM4 production as early as this month.\n\nMicron is not entirely out of the HBM4 picture, however — it is likely to supply HBM4 for mid-tier products in the Rubin series rather than for Vera Rubin itself.\n\nCommodity DRAM Pricing Is a Wild Card\n\nVolume allocations and pricing for Vera Rubin HBM4 have yet to be finalized. Some industry observers suggest that while SK Hynix will retain more than half of NVIDIA's total HBM shipments — including HBM3E — this year, Samsung may emerge as the largest supplier when it comes to Vera Rubin HBM4 specifically. Samsung recently expressed strong confidence, stating that its HBM revenue this year will be triple that of last year.\n\nA key variable is commodity DRAM pricing, which has been roughly doubling quarter-over-quarter. The per-Gb price of server DRAM modules such as SOCAMM2 has reportedly risen to approximately $1.30 — approaching the level of HBM3E, the flagship HBM product. From Samsung's perspective, producing commodity DRAM — which does not require the additional expensive stacking processes that HBM4 demands — may be more profitable.\n\nJensen Huang's meeting with SK Hynix engineers in Silicon Valley to encourage HBM4 development is also being interpreted as a move to check Samsung's growing negotiating leverage. \"Samsung holds HBM4, commodity DRAM, and other products, giving it a diverse set of negotiating cards it can play with NVIDIA,\" one industry observer noted.\n\n$MU\n\n---\n\nhttps://t.co/FVWACiweAR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04885443959208846, "ret_signed_m1d": 0.04885443959208846, "alpha_spy_m1d": 0.040170486227733226, "alpha_c_m1d": 0.013836568735838162, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.0350178708562503, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03542065676428452, "ret_signed_p1d": -0.03542065676428452, "alpha_spy_p1d": -0.03702778366965165, "alpha_c_p1d": -0.027915036428745044, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.007505620335539476, "ret_p1w": 0.13479909356117736, "ret_signed_p1w": -0.13479909356117736, "alpha_spy_p1w": -0.1484220002853761, "alpha_c_p1w": -0.13594010757092578, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.00114101400974842, "ret_p1m": -0.02974776819200242, "ret_signed_p1m": 0.02974776819200242, "alpha_spy_p1m": 0.004315803772389937, "alpha_c_p1m": 0.04377013005463248, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": 0.014022361862630062, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3229, -0.3364, -0.3809, -0.3902, -0.403, -0.421, -0.3618, -0.3172, -0.2898, -0.2907, -0.2639, -0.2639, -0.2688, -0.2439, -0.2484, -0.2669, -0.2669, -0.1898, -0.1982, -0.1179, -0.1278, -0.16, -0.1136, -0.1116, -0.1315, -0.1438, -0.1353, -0.0682, -0.0682, -0.0625, -0.0005, 0.0212, 0.0265, -0.0006, 0.0537, 0.1181, 0.1194, 0.0657, 0.1245, 0.0774, -0.0255, -0.0165, 0.0138, -0.0149, -0.0413, 0.054, 0.0633, 0.0574, 0.0574, 0.0269, 0.0812, 0.072, 0.0998, 0.0813, 0.0737, 0.1019, 0.0674, 0.0592, 0.06, -0.0248, 0.0294, 0.0199, -0.0489, 0.0, 0.0354, 0.0754, 0.0412, 0.0945, 0.1348, 0.1859, 0.186, 0.1411, 0.0863, 0.0386, 0.016, -0.0186, -0.087, -0.0825, -0.1731, -0.1319, -0.0548, -0.0589, -0.0589, -0.0293, -0.0297, 0.0452, 0.0831, 0.0808, 0.0961, 0.1966, 0.1724, 0.1749, 0.1694, 0.1523, 0.1548, 0.2527, 0.2379, 0.2764, 0.3479, 0.2959, 0.3323, 0.3289, 0.3933, 0.4813, 0.6451, 0.7129, 0.6616, 0.919, 1.0437, 0.9699, 1.0651, 0.9941, 0.8621, 0.7513, 0.7955, 0.881, 0.9583, 0.9298, 0.9298, 1.3021, 1.3857, 1.3731, 1.4951, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030582686876844472", "ticker_idx": 1, "ts": "2026-03-08T09:53:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "It opens the door to large-scale NVIDIA shipments over the next one to two years, securing their HBM dominance", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Micron excluded from Vera Rubin HBM4 supply chain; Samsung and SK Hynix confirmed as sole HBM4 suppliers, securing dominance for 1-2 years.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA's Vera Rubin to Use Only Samsung and SK Hynix HBM4\n\nSamsung Electronics and SK Hynix's sixth-generation High Bandwidth Memory (HBM4) will be incorporated into NVIDIA's next-generation AI accelerator \"Vera Rubin,\" slated for release in the second half of this year. Micron, the world's third-largest memory company, has been excluded from the HBM4 supply chain for Vera Rubin — a result widely interpreted as a recognition of Korean memory makers' technological superiority in key HBM4 performance metrics such as operating speed and bandwidth.\n\nAccording to industry sources on March 8, Samsung Electronics and SK Hynix have been confirmed as suppliers on NVIDIA's Vera Rubin vendor list. The two companies have been provisionally selected as HBM4 suppliers for the flagship next-gen AI accelerator, whose performance hinges critically on this component. HBM4 is an AI server memory chip manufactured by stacking 8 to 16 layers of advanced DRAM built on 11–13nm process nodes atop a base die, enabling rapid delivery of large volumes of data to the GPU in AI accelerators.\n\nMicron, which currently supplies HBM3E to NVIDIA, has been dropped from the Vera Rubin HBM4 vendor list. The company is expected to supply HBM4 for mid-tier AI accelerators optimized for inference — such as the \"Rubin CPX\" — rather than for Vera Rubin itself.\n\nThe selection of Samsung and SK Hynix as the sole HBM4 suppliers reflects their ability to meet NVIDIA's performance and yield requirements. Samsung has effectively passed NVIDIA's bifurcated HBM4 qualification tests targeting operating speeds of 10Gb/s and 11Gb/s. SK Hynix is in the process of optimizing its product for the 11Gb/s test. Samsung's foundry division has also won an order to produce NVIDIA's GPU \"RTX 3060,\" which the company has decided to resume manufacturing, with production set to begin shortly on an 8nm process.\n\nOnly Samsung and SK Hynix at the Heart of NVIDIA's \"Monster AI Chip\" — Micron Eliminated\n\nSince around 2022, when the AI era began, NVIDIA has been the company that determines the fate of memory semiconductor firms. Those that made it into NVIDIA's HBM supply chain became central players in the AI industry; those that didn't saw their standing diminish. Samsung was a prime example — the two-year crisis narrative that plagued Samsung Semiconductor was triggered by delays in HBM3 supply to NVIDIA and only subsided last September when Samsung passed NVIDIA's HBM3E 12-layer qualification test.\n\nAgainst this backdrop, the selection of Samsung and SK Hynix — and exclusion of Micron — as Vera Rubin HBM4 suppliers is being assessed as highly significant. It opens the door to large-scale NVIDIA shipments over the next one to two years, securing their HBM dominance.\n\nHigh-Spec HBM4 Orders\n\nAccording to industry sources on March 8, Vera Rubin hardware will be unveiled for the first time at NVIDIA's developer conference GTC 2026, scheduled for March 16 in Silicon Valley. While no official launch date has been set, a second-half 2026 release is expected. NVIDIA is putting everything into making Vera Rubin's performance more than five times greater than its predecessor, in order to decisively outpace rivals including AMD and Broadcom. More than 80 partner companies worldwide are aligning with NVIDIA to support the launch of this \"monster AI accelerator.\"\n\nSince last year, NVIDIA has identified HBM4 as the critical component underpinning Vera Rubin's success, pushing memory companies to develop high-performance products. It demanded operating speeds exceeding 10Gb/s — well above the 8Gb/s JEDEC standard — for Vera Rubin's HBM4. Capacity has also been scaled up: Vera Rubin will incorporate 16 HBM4 stacks totaling 576GB, surpassing the 432GB HBM4 capacity in AMD's competing next-gen AI accelerator, the MI450.\n\nMicron Eliminated from Vera Rubin\n\nGlobal memory companies competed fiercely for Vera Rubin HBM4 supply contracts. Winning over NVIDIA — which commands over 80% share in the AI accelerator market where HBM is heavily integrated — means both technological validation and guaranteed strong earnings.\n\nThe race for Vera Rubin HBM4 has narrowed to Samsung and SK Hynix. Micron does not appear on the vendor list. \"Micron isn't even being discussed as a Vera Rubin HBM4 supplier,\" one industry source said.\n\nBetween the two, Samsung has recently pulled ahead. Samsung has effectively passed NVIDIA's bifurcated qualification tests for both 10Gb/s and 11Gb/s variants of HBM4, and began shipping finished products to NVIDIA last month, albeit in limited volumes. SK Hynix is continuing product optimization with NVIDIA to pass the 11Gb/s test. Given that the process from HBM4 DRAM wafer input to final packaging takes over six months, both companies are expected to begin full-scale HBM4 production as early as this month.\n\nMicron is not entirely out of the HBM4 picture, however — it is likely to supply HBM4 for mid-tier products in the Rubin series rather than for Vera Rubin itself.\n\nCommodity DRAM Pricing Is a Wild Card\n\nVolume allocations and pricing for Vera Rubin HBM4 have yet to be finalized. Some industry observers suggest that while SK Hynix will retain more than half of NVIDIA's total HBM shipments — including HBM3E — this year, Samsung may emerge as the largest supplier when it comes to Vera Rubin HBM4 specifically. Samsung recently expressed strong confidence, stating that its HBM revenue this year will be triple that of last year.\n\nA key variable is commodity DRAM pricing, which has been roughly doubling quarter-over-quarter. The per-Gb price of server DRAM modules such as SOCAMM2 has reportedly risen to approximately $1.30 — approaching the level of HBM3E, the flagship HBM product. From Samsung's perspective, producing commodity DRAM — which does not require the additional expensive stacking processes that HBM4 demands — may be more profitable.\n\nJensen Huang's meeting with SK Hynix engineers in Silicon Valley to encourage HBM4 development is also being interpreted as a move to check Samsung's growing negotiating leverage. \"Samsung holds HBM4, commodity DRAM, and other products, giving it a diverse set of negotiating cards it can play with NVIDIA,\" one industry observer noted.\n\n$MU\n\n---\n\nhttps://t.co/FVWACiweAR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08472625011460999, "ret_signed_m1d": 0.08472625011460999, "alpha_spy_m1d": 0.09341020347896523, "alpha_c_m1d": 0.10239941511345074, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.01767316499884075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08299717427857112, "ret_signed_p1d": 0.08299717427857112, "alpha_spy_p1d": 0.08460430118393825, "alpha_c_p1d": 0.08299717427857112, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.0, "ret_p1w": 0.08760810333736901, "ret_signed_p1w": 0.08760810333736901, "alpha_spy_p1w": 0.10123101006156776, "alpha_c_p1w": 0.09462006679973212, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.007011963462363102, "ret_p1m": 0.1349090664939756, "ret_signed_p1m": 0.1349090664939756, "alpha_spy_p1m": 0.1603410309135881, "alpha_c_p1m": 0.15035790484955236, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": -0.015448838355576755, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3805, -0.3845, -0.3754, -0.3989, -0.4104, -0.3811, -0.3828, -0.3903, -0.3662, -0.3605, -0.3628, -0.3628, -0.3289, -0.3112, -0.3089, -0.3089, -0.3089, -0.2594, -0.204, -0.1994, -0.1873, -0.2, -0.1988, -0.2, -0.2069, -0.1914, -0.1706, -0.1418, -0.1395, -0.1631, -0.1383, -0.1222, -0.1233, -0.1233, -0.0807, -0.064, -0.0738, -0.0749, -0.1331, -0.0346, -0.0254, -0.0818, -0.0859, -0.0409, -0.0444, -0.0329, 0.0294, 0.0444, 0.0444, 0.0444, 0.0444, 0.0951, 0.0957, 0.1124, 0.1527, 0.1729, 0.2565, 0.2478, 0.2478, 0.1245, -0.0075, 0.1043, 0.0847, 0.0, 0.083, 0.0951, 0.083, 0.0576, 0.0876, 0.1176, 0.2017, 0.1556, 0.1493, 0.0738, 0.0934, 0.0893, 0.038, 0.038, 0.0182, -0.0343, 0.0953, 0.0304, 0.0754, 0.1153, 0.1349, 0.2158, 0.1782, 0.1898, 0.1609, 0.1927, 0.2187, 0.2562, 0.2475, 0.2389, 0.2649, 0.2562, 0.2966, 0.2677, 0.2966, 0.2822, 0.3053, 0.2735, 0.2735, 0.3428, 0.3428, 0.5363, 0.5681, 0.5508, 0.6489, 0.6114, 0.6403, 0.7096, 0.5623, 0.6229, 0.5912, 0.5941, 0.7298, 0.6894, 0.6894, 0.7269, 0.7731, 0.7298, 0.8309, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030582686876844472", "ticker_idx": 2, "ts": "2026-03-08T09:53:33+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "It opens the door to large-scale NVIDIA shipments over the next one to two years, securing their HBM dominance", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Micron excluded from Vera Rubin HBM4 supply chain; Samsung and SK Hynix confirmed as sole HBM4 suppliers, securing dominance for 1-2 years.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA's Vera Rubin to Use Only Samsung and SK Hynix HBM4\n\nSamsung Electronics and SK Hynix's sixth-generation High Bandwidth Memory (HBM4) will be incorporated into NVIDIA's next-generation AI accelerator \"Vera Rubin,\" slated for release in the second half of this year. Micron, the world's third-largest memory company, has been excluded from the HBM4 supply chain for Vera Rubin — a result widely interpreted as a recognition of Korean memory makers' technological superiority in key HBM4 performance metrics such as operating speed and bandwidth.\n\nAccording to industry sources on March 8, Samsung Electronics and SK Hynix have been confirmed as suppliers on NVIDIA's Vera Rubin vendor list. The two companies have been provisionally selected as HBM4 suppliers for the flagship next-gen AI accelerator, whose performance hinges critically on this component. HBM4 is an AI server memory chip manufactured by stacking 8 to 16 layers of advanced DRAM built on 11–13nm process nodes atop a base die, enabling rapid delivery of large volumes of data to the GPU in AI accelerators.\n\nMicron, which currently supplies HBM3E to NVIDIA, has been dropped from the Vera Rubin HBM4 vendor list. The company is expected to supply HBM4 for mid-tier AI accelerators optimized for inference — such as the \"Rubin CPX\" — rather than for Vera Rubin itself.\n\nThe selection of Samsung and SK Hynix as the sole HBM4 suppliers reflects their ability to meet NVIDIA's performance and yield requirements. Samsung has effectively passed NVIDIA's bifurcated HBM4 qualification tests targeting operating speeds of 10Gb/s and 11Gb/s. SK Hynix is in the process of optimizing its product for the 11Gb/s test. Samsung's foundry division has also won an order to produce NVIDIA's GPU \"RTX 3060,\" which the company has decided to resume manufacturing, with production set to begin shortly on an 8nm process.\n\nOnly Samsung and SK Hynix at the Heart of NVIDIA's \"Monster AI Chip\" — Micron Eliminated\n\nSince around 2022, when the AI era began, NVIDIA has been the company that determines the fate of memory semiconductor firms. Those that made it into NVIDIA's HBM supply chain became central players in the AI industry; those that didn't saw their standing diminish. Samsung was a prime example — the two-year crisis narrative that plagued Samsung Semiconductor was triggered by delays in HBM3 supply to NVIDIA and only subsided last September when Samsung passed NVIDIA's HBM3E 12-layer qualification test.\n\nAgainst this backdrop, the selection of Samsung and SK Hynix — and exclusion of Micron — as Vera Rubin HBM4 suppliers is being assessed as highly significant. It opens the door to large-scale NVIDIA shipments over the next one to two years, securing their HBM dominance.\n\nHigh-Spec HBM4 Orders\n\nAccording to industry sources on March 8, Vera Rubin hardware will be unveiled for the first time at NVIDIA's developer conference GTC 2026, scheduled for March 16 in Silicon Valley. While no official launch date has been set, a second-half 2026 release is expected. NVIDIA is putting everything into making Vera Rubin's performance more than five times greater than its predecessor, in order to decisively outpace rivals including AMD and Broadcom. More than 80 partner companies worldwide are aligning with NVIDIA to support the launch of this \"monster AI accelerator.\"\n\nSince last year, NVIDIA has identified HBM4 as the critical component underpinning Vera Rubin's success, pushing memory companies to develop high-performance products. It demanded operating speeds exceeding 10Gb/s — well above the 8Gb/s JEDEC standard — for Vera Rubin's HBM4. Capacity has also been scaled up: Vera Rubin will incorporate 16 HBM4 stacks totaling 576GB, surpassing the 432GB HBM4 capacity in AMD's competing next-gen AI accelerator, the MI450.\n\nMicron Eliminated from Vera Rubin\n\nGlobal memory companies competed fiercely for Vera Rubin HBM4 supply contracts. Winning over NVIDIA — which commands over 80% share in the AI accelerator market where HBM is heavily integrated — means both technological validation and guaranteed strong earnings.\n\nThe race for Vera Rubin HBM4 has narrowed to Samsung and SK Hynix. Micron does not appear on the vendor list. \"Micron isn't even being discussed as a Vera Rubin HBM4 supplier,\" one industry source said.\n\nBetween the two, Samsung has recently pulled ahead. Samsung has effectively passed NVIDIA's bifurcated qualification tests for both 10Gb/s and 11Gb/s variants of HBM4, and began shipping finished products to NVIDIA last month, albeit in limited volumes. SK Hynix is continuing product optimization with NVIDIA to pass the 11Gb/s test. Given that the process from HBM4 DRAM wafer input to final packaging takes over six months, both companies are expected to begin full-scale HBM4 production as early as this month.\n\nMicron is not entirely out of the HBM4 picture, however — it is likely to supply HBM4 for mid-tier products in the Rubin series rather than for Vera Rubin itself.\n\nCommodity DRAM Pricing Is a Wild Card\n\nVolume allocations and pricing for Vera Rubin HBM4 have yet to be finalized. Some industry observers suggest that while SK Hynix will retain more than half of NVIDIA's total HBM shipments — including HBM3E — this year, Samsung may emerge as the largest supplier when it comes to Vera Rubin HBM4 specifically. Samsung recently expressed strong confidence, stating that its HBM revenue this year will be triple that of last year.\n\nA key variable is commodity DRAM pricing, which has been roughly doubling quarter-over-quarter. The per-Gb price of server DRAM modules such as SOCAMM2 has reportedly risen to approximately $1.30 — approaching the level of HBM3E, the flagship HBM product. From Samsung's perspective, producing commodity DRAM — which does not require the additional expensive stacking processes that HBM4 demands — may be more profitable.\n\nJensen Huang's meeting with SK Hynix engineers in Silicon Valley to encourage HBM4 development is also being interpreted as a move to check Samsung's growing negotiating leverage. \"Samsung holds HBM4, commodity DRAM, and other products, giving it a diverse set of negotiating cards it can play with NVIDIA,\" one industry observer noted.\n\n$MU\n\n---\n\nhttps://t.co/FVWACiweAR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.10526311066951677, "ret_signed_m1d": 0.10526311066951677, "alpha_spy_m1d": 0.113947064033872, "alpha_c_m1d": 0.14028098152576707, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.0350178708562503, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.12200955542508463, "ret_signed_p1d": 0.12200955542508463, "alpha_spy_p1d": 0.12361668233045175, "alpha_c_p1d": 0.11450393508954515, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.007505620335539476, "ret_p1w": 0.16507170307476637, "ret_signed_p1w": 0.16507170307476637, "alpha_spy_p1w": 0.17869460979896512, "alpha_c_p1w": 0.1662127170845148, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.00114101400974842, "ret_p1m": 0.09569377775770538, "ret_signed_p1m": 0.09569377775770538, "alpha_spy_p1m": 0.12112574217731786, "alpha_c_p1m": 0.08167141589507532, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": 0.014022361862630062, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2991, -0.3254, -0.3182, -0.3385, -0.3672, -0.3421, -0.3409, -0.3469, -0.3075, -0.3027, -0.2979, -0.2979, -0.2848, -0.2359, -0.2227, -0.2227, -0.2227, -0.1917, -0.169, -0.1332, -0.1141, -0.0974, -0.1117, -0.1057, -0.1189, -0.1141, -0.1057, -0.0974, -0.0878, -0.1129, -0.1165, -0.0986, -0.0842, -0.1212, -0.0448, 0.0041, 0.028, 0.0853, -0.009, 0.0829, 0.0746, 0.0053, 0.0017, 0.0591, 0.0459, 0.0268, 0.0602, 0.0507, 0.0507, 0.0507, 0.0507, 0.0674, 0.1331, 0.1355, 0.1999, 0.2155, 0.3146, 0.2691, 0.2691, 0.1232, 0.0156, 0.1256, 0.1053, 0.0, 0.122, 0.1423, 0.1124, 0.0885, 0.1651, 0.1603, 0.2632, 0.2117, 0.2045, 0.116, 0.1794, 0.1902, 0.116, 0.116, 0.0443, -0.0347, 0.073, -0.0072, 0.0478, 0.0598, 0.0957, 0.2356, 0.1938, 0.2285, 0.244, 0.3194, 0.3589, 0.3816, 0.3493, 0.3947, 0.4641, 0.4629, 0.4653, 0.4617, 0.5455, 0.555, 0.5467, 0.5383, 0.5383, 0.7309, 0.7309, 0.9151, 0.9785, 1.0167, 1.2488, 1.195, 1.3636, 1.3565, 1.1758, 1.201, 1.0873, 1.0873, 1.3206, 1.3218, 1.3218, 1.4545, 1.683, 1.7385, 1.7911, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2023712691484062081", "ticker_idx": 0, "ts": "2026-02-17T10:54:39+00:00", "symbol": "6981.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "村田製社長、データセンター用コンデンサーで値上げの議論-株上昇", "tweet_type": "news_reaction", "timeline": "months", "summary": "Murata Mfg discussing MLCC price hikes for data centers amid 2x order-to-supply ratio; stock up 6.9%; AI demand expected to continue 3–5 years.", "resolved_tickers": ["6981.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "ja", "tweet_text": "Murata Manufacturing President Discusses Price Hikes for Data Center Capacitors — Stock Rises\n\n▶ 村田製社長、データセンター用コンデンサーで値上げの議論-株上昇\n\n∙ In an interview on the 17th, Murata Manufacturing emphasized that MLCC demand for data centers continues to grow on the back of the AI boom, and that the company has begun full-scale discussions on price increases.\n\n∙ On whether MLCC prices will be raised, management noted: “It is important to accurately assess real demand, and we expect to have a clearer read on demand levels by next quarter.”\n\n∙ The company added that it will carefully determine price hikes, taking into account the broader impact on the market and industry.\n\n∙ Murata also projected that data center investment, led by hyperscalers, will continue over the next 3–5 years.\n\n∙ Current order inquiries for MLCCs are running at approximately 2x the company’s supply capacity, underscoring that AI is not a temporary boom.\n\n∙ Following the report, Murata’s stock reversed into positive territory during afternoon trading, rising as much as 9.2% intraday to ¥3,585 (the highest since July 29, 2024), before closing up 6.9% at ¥3,509.\n\n---\n\nhttps://t.co/4geXbON1g6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06440583545825873, "ret_signed_m1d": -0.06440583545825873, "alpha_spy_m1d": -0.06279496806805662, "alpha_c_m1d": -0.06497935393276089, "bench_spy_m1d": -0.0016108673902021087, "bench_c_m1d": 0.0005735184745021549, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02878307949994019, "ret_signed_p1d": 0.02878307949994019, "alpha_spy_p1d": 0.023745320958323157, "alpha_c_p1d": 0.018530724799137088, "bench_spy_p1d": 0.005037758541617032, "bench_c_p1d": 0.010252354700803101, "ret_p1w": 0.15275001178140535, "ret_signed_p1w": 0.15275001178140535, "alpha_spy_p1w": 0.1461600019555893, "alpha_c_p1w": 0.14672762966431607, "bench_spy_p1w": 0.0065900098258160344, "bench_c_p1w": 0.0060223821170892755, "ret_p1m": 0.09546878370035006, "ret_signed_p1m": 0.09546878370035006, "alpha_spy_p1m": 0.12683727349222973, "alpha_c_p1m": 0.1063664014518203, "bench_spy_p1m": -0.03136848979187967, "bench_c_p1m": -0.010897617751470245, "ret_p3m": 0.7694579478976038, "ret_signed_p3m": 0.7694579478976038, "alpha_spy_p3m": 0.6840238082462549, "alpha_c_p3m": 0.50419305706437, "bench_spy_p3m": 0.08543413965134894, "bench_c_p3m": 0.2652648908332338, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1228, -0.1357, -0.1357, -0.1291, -0.0969, -0.0798, -0.0844, -0.0812, -0.0675, -0.0553, -0.0177, -0.037, -0.0313, -0.0308, -0.0294, -0.0576, -0.0311, -0.0678, -0.092, -0.0909, -0.0997, -0.0878, -0.0869, -0.0883, -0.0983, -0.0972, -0.0861, -0.0889, -0.075, -0.075, -0.075, -0.075, -0.051, -0.0504, -0.0479, -0.0918, -0.0898, -0.0898, -0.0573, -0.0182, -0.0311, -0.0271, -0.0217, -0.0579, -0.0596, -0.0499, -0.0504, -0.094, -0.0801, -0.0918, -0.1066, -0.1063, -0.1208, -0.0544, -0.0533, -0.0638, -0.0832, -0.0684, -0.0541, -0.0541, -0.061, -0.0712, -0.0644, 0.0, 0.0288, 0.0447, 0.0473, 0.0473, 0.1528, 0.2035, 0.165, 0.1707, 0.1584, 0.063, 0.0422, 0.0852, 0.0781, -0.0103, 0.0465, 0.071, 0.0601, 0.0402, 0.0279, 0.0835, 0.0955, 0.0681, 0.0681, -0.0003, 0.0097, 0.043, 0.061, 0.0673, 0.0166, -0.0193, 0.056, 0.0005, 0.09, 0.0952, 0.0891, 0.1507, 0.1478, 0.1927, 0.1996, 0.2453, 0.3003, 0.3785, 0.3233, 0.363, 0.3837, 0.4127, 0.3846, 0.4199, 0.4142, 0.3923, 0.3923, 0.4832, 0.478, 0.478, 0.478, 0.478, 0.6248, 0.6829, 0.7145, 0.7485, 0.7516, 0.8235, 0.7695, 0.7752, 0.7778, 0.7732, 0.9352, 1.0511, 1.2806, 1.3235, 1.2496, 1.4561, 1.7688, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013762133516001716", "ticker_idx": 0, "ts": "2026-01-20T23:54:41+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "current earnings power is insufficient to support further re-rating of the stock. Unless Intel convincingly demonstrates recovery of server market share, an upgrade to investment grade appears difficult", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's bearish INTC view: capacity constraints, server share loss, margin pressure block re-rating; CPU supply shortage may benefit AMD more.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley's Intel Earnings Outlook: CPU Supply Shortages Could Boost Results, But It Might Actually Benefit AMD More\n\nMorgan Stanley pointed out in a January 20 report that, amid the currently complex semiconductor cycle, Intel is in a very unusual position. While the supply shortage in the server CPU market could temporarily lift the stock price and lead to better-than-expected earnings in the short term, serious long-term risks—such as production capacity constraints and loss of market share—lie behind this shortage. Morgan Stanley expressed concern that this situation makes it difficult to support a long-term revaluation of Intel's value.\n\nShort-term Tailwind vs. Long-term Concerns\n\nFor the earnings report to be released this week, Morgan Stanley believes Intel has a chance of delivering a slight positive surprise in EPS (earnings per share). Analysts expect the data to exceed market expectations thanks to the supply shortage. Specifically, non-GAAP revenue for the December quarter (Q4) is projected at $13.31 billion, down 6.7% year-over-year, which is slightly below Wall Street consensus ($13.407 billion). However, the Data Center & AI (DCAI) segment is forecast to grow 11.5% quarter-over-quarter.\n\nYet this short-term earnings support cannot mask deeper structural problems. Analyst Joseph Moore highlighted in the report that “even with modest growth, the fact that internal production capacity is already fully sold out is itself concerning.” This means Intel is unable to fully capture the fruits of this demand rebound, and most of the incremental market is being handed over to competitors. AMD has publicly stated it will further expand its leadership in the server market this year with the Venice series, and analysts note that Intel’s current products still lag significantly in performance.\n\nProduct Roadmap Gaps\n\nIn terms of product competitiveness, Intel is currently in an awkward transition period. While Panther Lake (based on the 18A process) has shown some potential in the notebook segment and demonstrated the viability of the new process, the server and high-end desktop markets still require considerable waiting, according to the analyst.\n\nThe report states that regaining competitiveness in the high-end desktop market will require waiting for Nova Lake, expected in the second half of 2026. The server side is even more serious. Management has hinted that Diamond Rapids may lack SMT (hyper-threading), meaning Intel may need to wait until Coral Rapids in the second half of 2027 to truly overtake competitors in performance. During this multi-year gap, current product lines cannot properly capitalize on surging market demand, which will continue to limit upside potential for the stock price.\n\nFoundry Business Credibility Crisis\n\nIntel’s foundry business (wafer foundry), a core part of its transformation strategy, is also being overshadowed by the current supply shortage situation. Morgan Stanley has long been skeptical of Intel’s foundry story, and the present production capacity constraints are amplifying those concerns.\n\nFrom an external customer’s perspective, seeing that Intel is struggling to meet even the modest growth needs of its own products raises significant doubts about entrusting orders to them. Customers were already worried about having to compete with Intel’s internal products for production resources—and reality has now confirmed those fears. While the foundry narrative has been supporting the stock price to some extent, unless Intel’s product division can capitalize on surging demand, it remains a major hurdle for the foundry business to attract real customers and drive a meaningful valuation increase.\n\nFinancial Outlook and Valuation Perspective\n\nFor next quarter (March quarter / Q1), Morgan Stanley expects revenue of $12.552 billion, slightly above Wall Street consensus ($12.525 billion). However, they forecast gross margin at 34.9%, below market expectations (36.1%), due to continued pressure from Lunar Lake’s lower margins and early-stage 18A process costs.\n\nIntel currently trades at approximately 35× forward 2027 estimated EPS, a multiple higher than the average for large logic semiconductor companies. This reflects the market assigning a premium to Intel’s future recovery leverage and foundry business optionality. However, considering the lack of server product competitiveness, the reality that most incremental demand is flowing to AMD, and ongoing margin pressure, Morgan Stanley’s analysts judge that current earnings power is insufficient to support further re-rating of the stock. Unless Intel convincingly demonstrates recovery of server market share, an upgrade to investment grade appears difficult...\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.032948975361816846, "ret_signed_m1d": 0.032948975361816846, "alpha_spy_m1d": 0.05372877093114581, "alpha_c_m1d": 0.058564384631455635, "bench_spy_m1d": 0.020779795569328963, "bench_c_m1d": 0.02561540926963879, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.1171745977305223, "ret_signed_p1d": -0.1171745977305223, "alpha_spy_p1d": -0.10563351852842784, "alpha_c_p1d": -0.08761447171721737, "bench_spy_p1d": 0.011541079202094462, "bench_c_p1d": 0.029560126013304933, "ret_p1w": -0.09534598305554887, "ret_signed_p1w": 0.09534598305554887, "alpha_spy_p1w": 0.12177826454898655, "alpha_c_p1w": 0.1385335255615877, "bench_spy_p1w": 0.02643228149343768, "bench_c_p1w": 0.043187542506038845, "ret_p1m": -0.06383859557556404, "ret_signed_p1m": 0.06383859557556404, "alpha_spy_p1m": 0.07669316871035892, "alpha_c_p1m": 0.12060227372570131, "bench_spy_p1m": 0.012854573134794878, "bench_c_p1m": 0.056763678150137276, "ret_p3m": 0.4106259897611202, "ret_signed_p3m": -0.4106259897611202, "alpha_spy_p3m": -0.359710351171568, "alpha_c_p3m": -0.22166114372092416, "bench_spy_p3m": 0.05091563858955217, "bench_c_p3m": 0.18896484604019603, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2142, -0.2117, -0.1857, -0.1448, -0.1487, -0.173, -0.1765, -0.1866, -0.2374, -0.2096, -0.2331, -0.2148, -0.2082, -0.2199, -0.2197, -0.2605, -0.2685, -0.2852, -0.293, -0.277, -0.3077, -0.2895, -0.263, -0.2621, -0.242, -0.242, -0.1647, -0.1761, -0.1048, -0.0988, -0.166, -0.1472, -0.1701, -0.166, -0.1602, -0.1864, -0.2214, -0.2276, -0.2317, -0.2576, -0.2529, -0.2418, -0.251, -0.2514, -0.2554, -0.2554, -0.2545, -0.2446, -0.2319, -0.2401, -0.2401, -0.189, -0.1893, -0.1755, -0.1221, -0.1534, -0.062, -0.0927, -0.0262, 0.0033, -0.0049, -0.0329, -0.0329, 0.0, 0.1172, 0.1186, -0.0719, -0.125, -0.0953, 0.0045, 0.0021, -0.043, 0.0051, 0.0142, 0.0008, -0.0066, 0.0418, 0.0346, -0.0294, -0.0056, -0.0428, -0.0364, -0.0364, -0.049, -0.0638, -0.0811, -0.0916, -0.1015, -0.0502, -0.0346, -0.0638, -0.0607, -0.063, -0.1124, -0.0614, -0.0537, -0.1058, -0.0614, -0.0367, -0.0119, -0.0682, -0.0575, -0.0577, -0.0927, -0.0727, -0.049, -0.0966, -0.0937, -0.0927, -0.0284, -0.0918, -0.1118, -0.1518, -0.0912, -0.0109, 0.0375, 0.0375, 0.0457, 0.0896, 0.214, 0.271, 0.2846, 0.3423, 0.314, 0.3373, 0.4106, 0.4106, 0.353, 0.3645, 0.3441, 0.3752, 0.6998, 0.7502, 0.7405, 0.9512, 0.9456, 1.0515, 0.9724, 1.2271, 1.3272, 1.2574, 1.5725, 1.6656, 1.4837, 1.4771, 1.3874, 1.2399, 1.2276, 1.2817, 1.4498, 1.4403, 1.4679, 1.4679, 1.5437, 1.5076, 1.4895, 1.3616, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013762133516001716", "ticker_idx": 1, "ts": "2026-01-20T23:54:41+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It Might Actually Benefit AMD More", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's bearish INTC view: capacity constraints, server share loss, margin pressure block re-rating; CPU supply shortage may benefit AMD more.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley's Intel Earnings Outlook: CPU Supply Shortages Could Boost Results, But It Might Actually Benefit AMD More\n\nMorgan Stanley pointed out in a January 20 report that, amid the currently complex semiconductor cycle, Intel is in a very unusual position. While the supply shortage in the server CPU market could temporarily lift the stock price and lead to better-than-expected earnings in the short term, serious long-term risks—such as production capacity constraints and loss of market share—lie behind this shortage. Morgan Stanley expressed concern that this situation makes it difficult to support a long-term revaluation of Intel's value.\n\nShort-term Tailwind vs. Long-term Concerns\n\nFor the earnings report to be released this week, Morgan Stanley believes Intel has a chance of delivering a slight positive surprise in EPS (earnings per share). Analysts expect the data to exceed market expectations thanks to the supply shortage. Specifically, non-GAAP revenue for the December quarter (Q4) is projected at $13.31 billion, down 6.7% year-over-year, which is slightly below Wall Street consensus ($13.407 billion). However, the Data Center & AI (DCAI) segment is forecast to grow 11.5% quarter-over-quarter.\n\nYet this short-term earnings support cannot mask deeper structural problems. Analyst Joseph Moore highlighted in the report that “even with modest growth, the fact that internal production capacity is already fully sold out is itself concerning.” This means Intel is unable to fully capture the fruits of this demand rebound, and most of the incremental market is being handed over to competitors. AMD has publicly stated it will further expand its leadership in the server market this year with the Venice series, and analysts note that Intel’s current products still lag significantly in performance.\n\nProduct Roadmap Gaps\n\nIn terms of product competitiveness, Intel is currently in an awkward transition period. While Panther Lake (based on the 18A process) has shown some potential in the notebook segment and demonstrated the viability of the new process, the server and high-end desktop markets still require considerable waiting, according to the analyst.\n\nThe report states that regaining competitiveness in the high-end desktop market will require waiting for Nova Lake, expected in the second half of 2026. The server side is even more serious. Management has hinted that Diamond Rapids may lack SMT (hyper-threading), meaning Intel may need to wait until Coral Rapids in the second half of 2027 to truly overtake competitors in performance. During this multi-year gap, current product lines cannot properly capitalize on surging market demand, which will continue to limit upside potential for the stock price.\n\nFoundry Business Credibility Crisis\n\nIntel’s foundry business (wafer foundry), a core part of its transformation strategy, is also being overshadowed by the current supply shortage situation. Morgan Stanley has long been skeptical of Intel’s foundry story, and the present production capacity constraints are amplifying those concerns.\n\nFrom an external customer’s perspective, seeing that Intel is struggling to meet even the modest growth needs of its own products raises significant doubts about entrusting orders to them. Customers were already worried about having to compete with Intel’s internal products for production resources—and reality has now confirmed those fears. While the foundry narrative has been supporting the stock price to some extent, unless Intel’s product division can capitalize on surging demand, it remains a major hurdle for the foundry business to attract real customers and drive a meaningful valuation increase.\n\nFinancial Outlook and Valuation Perspective\n\nFor next quarter (March quarter / Q1), Morgan Stanley expects revenue of $12.552 billion, slightly above Wall Street consensus ($12.525 billion). However, they forecast gross margin at 34.9%, below market expectations (36.1%), due to continued pressure from Lunar Lake’s lower margins and early-stage 18A process costs.\n\nIntel currently trades at approximately 35× forward 2027 estimated EPS, a multiple higher than the average for large logic semiconductor companies. This reflects the market assigning a premium to Intel’s future recovery leverage and foundry business optionality. However, considering the lack of server product competitiveness, the reality that most incremental demand is flowing to AMD, and ongoing margin pressure, Morgan Stanley’s analysts judge that current earnings power is insufficient to support further re-rating of the stock. Unless Intel convincingly demonstrates recovery of server market share, an upgrade to investment grade appears difficult...\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00038804906261280525, "ret_signed_m1d": -0.00038804906261280525, "alpha_spy_m1d": -0.021167844631941768, "alpha_c_m1d": -0.026003458332251594, "bench_spy_m1d": 0.020779795569328963, "bench_c_m1d": 0.02561540926963879, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.077095571852271, "ret_signed_p1d": 0.077095571852271, "alpha_spy_p1d": 0.06555449265017654, "alpha_c_p1d": 0.04753544583896607, "bench_spy_p1d": 0.011541079202094462, "bench_c_p1d": 0.029560126013304933, "ret_p1w": 0.0867109381214386, "ret_signed_p1w": 0.0867109381214386, "alpha_spy_p1w": 0.060278656628000915, "alpha_c_p1w": 0.04352339561539975, "bench_spy_p1w": 0.02643228149343768, "bench_c_p1w": 0.043187542506038845, "ret_p1m": -0.13711626122294407, "ret_signed_p1m": -0.13711626122294407, "alpha_spy_p1m": -0.14997083435773895, "alpha_c_p1m": -0.19387993937308134, "bench_spy_p1m": 0.012854573134794878, "bench_c_p1m": 0.056763678150137276, "ret_p3m": 0.20037089016204823, "ret_signed_p3m": 0.20037089016204823, "alpha_spy_p3m": 0.14945525157249606, "alpha_c_p3m": 0.011406044121852199, "bench_spy_p3m": 0.05091563858955217, "bench_c_p3m": 0.18896484604019603, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0132, 0.0905, 0.1197, 0.1125, 0.1397, 0.0988, 0.1043, 0.1196, 0.0782, 0.1053, 0.0249, 0.007, 0.052, 0.0241, 0.1163, 0.0692, 0.0642, 0.0371, -0.007, -0.0361, -0.1117, -0.1213, -0.0727, -0.1112, -0.0762, -0.0762, -0.062, -0.0524, -0.0719, -0.0617, -0.0687, -0.0602, -0.0466, -0.0444, -0.0453, -0.0452, -0.0912, -0.1049, -0.0981, -0.1458, -0.1331, -0.0797, -0.0732, -0.0734, -0.0728, -0.0728, -0.073, -0.0703, -0.0715, -0.0766, -0.0766, -0.0364, -0.0467, -0.0758, -0.0944, -0.1175, -0.124, -0.1045, -0.0472, -0.0359, -0.0172, -0.0004, -0.0004, 0.0, 0.0771, 0.094, 0.1197, 0.0836, 0.0867, 0.0898, 0.0874, 0.0207, 0.0619, 0.0439, -0.1368, -0.17, -0.1012, -0.0686, -0.0791, -0.0791, -0.112, -0.1061, -0.1061, -0.1244, -0.1371, -0.1231, -0.137, -0.1523, -0.078, -0.0908, -0.1218, -0.1367, -0.1436, -0.1767, -0.1287, -0.14, -0.1703, -0.1261, -0.1237, -0.1168, -0.1474, -0.1661, -0.1524, -0.1535, -0.14, -0.1149, -0.1319, -0.1261, -0.1145, -0.0502, -0.1214, -0.1291, -0.1547, -0.1228, -0.0936, -0.0622, -0.0622, -0.0506, -0.0448, -0.0004, 0.0204, 0.0566, 0.0643, 0.0998, 0.113, 0.1998, 0.2004, 0.1855, 0.2267, 0.3085, 0.3165, 0.4997, 0.4429, 0.3936, 0.4536, 0.5285, 0.5546, 0.4727, 0.5318, 0.817, 0.7612, 0.9627, 0.9782, 0.933, 0.9209, 0.939, 0.8286, 0.8152, 0.7853, 0.9299, 0.9386, 1.0158, 1.0158, 1.1727, 1.1367, 1.2339, 1.2253, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011576039747850491", "ticker_idx": 0, "ts": "2026-01-14T23:07:55+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung may hold a relative advantage due to the rising importance of the 'logic die' (base die) starting from HBM4. Samsung possesses this technology in-house", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung favored over SK Hynix in HBM4 race: in-house logic die capability vs. SK Hynix's TSMC dependency and Wuxi fab legacy-fication risk.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia Ups HBM4 Specs; \"SK Hynix’s Structural Vulnerabilities Exposed\"\n\nNvidia has significantly raised the specifications for the 6th generation High Bandwidth Memory (HBM4), a major inflection point in this year's AI memory market. This forces Samsung Electronics and SK Hynix back to the starting line to modify their next-gen HBM to meet these demanding requirements. Experts observe that Samsung may hold a relative advantage due to the rising importance of the \"logic die\" (base die) starting from HBM4. Samsung possesses this technology in-house, whereas SK Hynix relies on Taiwan's TSMC, creating a structural differentiation. The logic die serves as the \"brain\" of the HBM, managing the \"traffic control\" of power and signals.\n\nParticularly in the case of TSMC, the consensus among domestic and international experts is that they are struggling to keep up with the overwhelming foundry (contract manufacturing) demand fueled by the AI semiconductor boom, making it difficult for them to prioritize investments in HBM logic die production capacity or process migration. This is cited as a vulnerability for SK Hynix, which lacks the independent capability to design and produce its own logic dies.\n\nEscalating Requirements for the \"Brain\" of HBM\nAccording to industry sources on the 15th, Nvidia updated and delivered its HBM4 supply standards to Samsung Electronics and SK Hynix in the fourth quarter of last year. In response, Samsung is accelerating development, focusing on heat control and performance enhancement by revising logic die designs in collaboration with its Foundry Division.\n\nGenerally, HBM is a structure where multiple DRAM dies are stacked vertically. In HBM4, the logic die is the chip attached to the very bottom of this stack, performing functions that DRAM dies alone cannot handle. Consequently, the timing and routing of data transmission, as well as power management, depend on the performance of the logic die.\n\nWhile the initial standard for HBM4 targeted data transfer speeds of 8–10 Gb/s, Nvidia raised this to 13 Gb/s, significantly increasing the importance of the logic die. At such high-speed intervals, heat and power fluctuations generated by the HBM can lead directly to data processing errors. Therefore, the specifications of the logic die that controls these factors must be improved. This implies that the competition for HBM performance has expanded from DRAM density to logic die design capabilities.\n\nSamsung Confident in In-House Synergy; SK Hynix Faces \"TSMC Bottleneck\"\nA source familiar with Samsung Electronics explained, \"Increasing data transfer speeds is manageable through minor revisions to logic die design and processes. Reaching the 13 Gbps per pin level requested by Nvidia is achievable; the real challenge is heat management. However, this is not an issue that will take long to resolve.\" Previously, Young-hyun Jun, Head of Samsung's DS Division (Vice Chairman), also expressed confidence in HBM4 leadership during his New Year's address.\n\nOn the other hand, experts anticipate that SK Hynix will face some difficulties in this process. A professor at a private university in Korea noted, \"As HBM generations evolve, technological gaps are beginning to emerge at the logic die level. In SK Hynix's case, they are entirely dependent on TSMC for this technology. From the perspective of TSMC, which is already struggling to meet demand, facility investment and line expansion for logic dies remain a lower priority.\"\n\nGeopolitical Risks and the \"Legacy-fication\" of the Wuxi Fab\nSK Hynix is also facing a disadvantage compared to Samsung regarding DRAM process migration—the raw material for HBM. In particular, concerns are being raised about the widening technology gap between the Wuxi fab in China, which handles more than 30% of SK Hynix's DRAM production, and its core Icheon fab in Korea. The Wuxi fab faces the risk of needing to obtain \"Validated End-User\" (VEU) status from the U.S. annually due to sanctions, and equipment imports must comply with standards set by the U.S. government.\n\nObservations suggest that if export licensing procedures become permanent, it will take longer to procure equipment and parts, acting as a barrier to transitioning to advanced processes within China. Korea Ratings recently analyzed in a report that \"if restrictions on bringing semiconductor equipment into China persist in the long term, both DRAM and NAND flash fabs located in China will undergo 'legacy-fication' (becoming outdated), which could act as a downward factor for operating performance due to decreased production capacity.\"\n\n---\n\nhttps://t.co/f6hHBBdolj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.019244468009195326, "ret_signed_m1d": -0.019244468009195326, "alpha_spy_m1d": -0.02418398424546797, "alpha_c_m1d": -0.03154575813007965, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.012301290120884323, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025659290678927027, "ret_signed_p1d": 0.025659290678927027, "alpha_spy_p1d": 0.022936076282397577, "alpha_c_p1d": 0.020406992758993603, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.005252297919933424, "ret_p1w": 0.06557379244570671, "ret_signed_p1w": 0.06557379244570671, "alpha_spy_p1w": 0.07275842244563935, "alpha_c_p1w": 0.07172449029730721, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": -0.006150697851600495, "ret_p1m": 0.27298646649970704, "ret_signed_p1m": 0.27298646649970704, "alpha_spy_p1m": 0.2861534270775483, "alpha_c_p1m": 0.3109269440464132, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": -0.03794047754670615, "ret_p3m": 0.43560965925139605, "ret_signed_p3m": 0.43560965925139605, "alpha_spy_p3m": 0.43906620360065973, "alpha_c_p3m": 0.4280696036960927, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.007540055555303349, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3056, -0.3042, -0.3084, -0.3006, -0.3155, -0.2992, -0.2765, -0.2942, -0.2871, -0.2616, -0.2375, -0.212, -0.2559, -0.2864, -0.2964, -0.3056, -0.2864, -0.2659, -0.2687, -0.2708, -0.3106, -0.2864, -0.3063, -0.3155, -0.2864, -0.3276, -0.3141, -0.2957, -0.2708, -0.2659, -0.2871, -0.285, -0.2666, -0.2588, -0.2545, -0.2311, -0.2233, -0.2311, -0.2339, -0.2389, -0.2276, -0.2566, -0.2708, -0.2347, -0.2368, -0.246, -0.2162, -0.2091, -0.212, -0.212, -0.1701, -0.1483, -0.1454, -0.1454, -0.1454, -0.0841, -0.0157, -0.01, 0.005, -0.0107, -0.0093, -0.0107, -0.0192, 0.0, 0.0257, 0.0613, 0.0641, 0.0349, 0.0656, 0.0855, 0.0841, 0.0841, 0.1368, 0.1575, 0.1454, 0.144, 0.072, 0.1939, 0.2053, 0.1354, 0.1304, 0.186, 0.1818, 0.196, 0.273, 0.2915, 0.2915, 0.2915, 0.2915, 0.3542, 0.355, 0.3756, 0.4255, 0.4505, 0.5538, 0.5431, 0.5431, 0.3906, 0.2274, 0.3656, 0.3414, 0.2366, 0.3393, 0.3542, 0.3393, 0.3079, 0.345, 0.382, 0.4861, 0.4291, 0.4212, 0.3279, 0.3521, 0.3471, 0.2837, 0.2837, 0.2592, 0.1942, 0.3545, 0.2742, 0.3299, 0.3792, 0.4035, 0.5035, 0.457, 0.4713, 0.4356, 0.4749, 0.507, 0.5535, 0.5427, 0.532, 0.5642, 0.5535, 0.6035, 0.5677, 0.6035, 0.5856, 0.6142, 0.5749, 0.5749, 0.6606, 0.6606, 0.8999, 0.9391, 0.9177, 1.0391, 0.9927, 1.0284, 1.1141, 0.932, 1.007, 0.9677, 0.9713, 1.1391, 1.0891, 1.0891, 1.1356, 1.1927, 1.1391, 1.2641, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011576039747850491", "ticker_idx": 1, "ts": "2026-01-14T23:07:55+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix will face some difficulties in this process… entirely dependent on TSMC for this technology… facility investment and line expansion for logic dies remain a lower priority", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung favored over SK Hynix in HBM4 race: in-house logic die capability vs. SK Hynix's TSMC dependency and Wuxi fab legacy-fication risk.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Nvidia Ups HBM4 Specs; \"SK Hynix’s Structural Vulnerabilities Exposed\"\n\nNvidia has significantly raised the specifications for the 6th generation High Bandwidth Memory (HBM4), a major inflection point in this year's AI memory market. This forces Samsung Electronics and SK Hynix back to the starting line to modify their next-gen HBM to meet these demanding requirements. Experts observe that Samsung may hold a relative advantage due to the rising importance of the \"logic die\" (base die) starting from HBM4. Samsung possesses this technology in-house, whereas SK Hynix relies on Taiwan's TSMC, creating a structural differentiation. The logic die serves as the \"brain\" of the HBM, managing the \"traffic control\" of power and signals.\n\nParticularly in the case of TSMC, the consensus among domestic and international experts is that they are struggling to keep up with the overwhelming foundry (contract manufacturing) demand fueled by the AI semiconductor boom, making it difficult for them to prioritize investments in HBM logic die production capacity or process migration. This is cited as a vulnerability for SK Hynix, which lacks the independent capability to design and produce its own logic dies.\n\nEscalating Requirements for the \"Brain\" of HBM\nAccording to industry sources on the 15th, Nvidia updated and delivered its HBM4 supply standards to Samsung Electronics and SK Hynix in the fourth quarter of last year. In response, Samsung is accelerating development, focusing on heat control and performance enhancement by revising logic die designs in collaboration with its Foundry Division.\n\nGenerally, HBM is a structure where multiple DRAM dies are stacked vertically. In HBM4, the logic die is the chip attached to the very bottom of this stack, performing functions that DRAM dies alone cannot handle. Consequently, the timing and routing of data transmission, as well as power management, depend on the performance of the logic die.\n\nWhile the initial standard for HBM4 targeted data transfer speeds of 8–10 Gb/s, Nvidia raised this to 13 Gb/s, significantly increasing the importance of the logic die. At such high-speed intervals, heat and power fluctuations generated by the HBM can lead directly to data processing errors. Therefore, the specifications of the logic die that controls these factors must be improved. This implies that the competition for HBM performance has expanded from DRAM density to logic die design capabilities.\n\nSamsung Confident in In-House Synergy; SK Hynix Faces \"TSMC Bottleneck\"\nA source familiar with Samsung Electronics explained, \"Increasing data transfer speeds is manageable through minor revisions to logic die design and processes. Reaching the 13 Gbps per pin level requested by Nvidia is achievable; the real challenge is heat management. However, this is not an issue that will take long to resolve.\" Previously, Young-hyun Jun, Head of Samsung's DS Division (Vice Chairman), also expressed confidence in HBM4 leadership during his New Year's address.\n\nOn the other hand, experts anticipate that SK Hynix will face some difficulties in this process. A professor at a private university in Korea noted, \"As HBM generations evolve, technological gaps are beginning to emerge at the logic die level. In SK Hynix's case, they are entirely dependent on TSMC for this technology. From the perspective of TSMC, which is already struggling to meet demand, facility investment and line expansion for logic dies remain a lower priority.\"\n\nGeopolitical Risks and the \"Legacy-fication\" of the Wuxi Fab\nSK Hynix is also facing a disadvantage compared to Samsung regarding DRAM process migration—the raw material for HBM. In particular, concerns are being raised about the widening technology gap between the Wuxi fab in China, which handles more than 30% of SK Hynix's DRAM production, and its core Icheon fab in Korea. The Wuxi fab faces the risk of needing to obtain \"Validated End-User\" (VEU) status from the U.S. annually due to sanctions, and equipment imports must comply with standards set by the U.S. government.\n\nObservations suggest that if export licensing procedures become permanent, it will take longer to procure equipment and parts, acting as a barrier to transitioning to advanced processes within China. Korea Ratings recently analyzed in a report that \"if restrictions on bringing semiconductor equipment into China persist in the long term, both DRAM and NAND flash fabs located in China will undergo 'legacy-fication' (becoming outdated), which could act as a downward factor for operating performance due to decreased production capacity.\"\n\n---\n\nhttps://t.co/f6hHBBdolj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005390882443871203, "ret_signed_m1d": 0.005390882443871203, "alpha_spy_m1d": 0.010330398680143849, "alpha_c_m1d": 0.013579353219724921, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009433875474092801, "ret_signed_p1d": -0.009433875474092801, "alpha_spy_p1d": -0.006710661077563351, "alpha_c_p1d": 0.011320591970013227, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": -0.002695441221935657, "ret_signed_p1w": 0.002695441221935657, "alpha_spy_p1w": -0.0044891887779969775, "alpha_c_p1w": 0.037663862385410085, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": 0.19676543677314084, "ret_signed_p1m": -0.19676543677314084, "alpha_spy_p1m": -0.20993239735098213, "alpha_c_p1m": -0.1510335480930287, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.404203507500283, "ret_signed_p3m": -0.404203507500283, "alpha_spy_p3m": -0.4076600518495467, "alpha_c_p3m": -0.26260446187102326, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3731, -0.3462, -0.3549, -0.3515, -0.3556, -0.3132, -0.2795, -0.2983, -0.2485, -0.235, -0.2472, -0.165, -0.2108, -0.2202, -0.2014, -0.2189, -0.1839, -0.1664, -0.1691, -0.1758, -0.2458, -0.1839, -0.2324, -0.2431, -0.231, -0.2983, -0.2997, -0.301, -0.2943, -0.2668, -0.2857, -0.2749, -0.248, -0.2561, -0.2695, -0.2668, -0.2224, -0.2372, -0.2089, -0.2385, -0.2305, -0.2534, -0.2857, -0.2574, -0.2561, -0.2628, -0.2183, -0.2129, -0.2075, -0.2075, -0.1927, -0.1375, -0.1226, -0.1226, -0.1226, -0.0876, -0.062, -0.0216, 0.0, 0.0189, 0.0027, 0.0094, -0.0054, 0.0, 0.0094, 0.0189, 0.0296, 0.0013, -0.0027, 0.0175, 0.0337, -0.0081, 0.0782, 0.1334, 0.1604, 0.2251, 0.1186, 0.2224, 0.2129, 0.1348, 0.1307, 0.1954, 0.1806, 0.159, 0.1968, 0.186, 0.186, 0.186, 0.186, 0.2049, 0.279, 0.2817, 0.3544, 0.372, 0.4839, 0.4326, 0.4326, 0.2678, 0.1463, 0.2705, 0.2476, 0.1288, 0.2665, 0.2894, 0.2557, 0.2287, 0.3151, 0.3097, 0.4258, 0.3677, 0.3596, 0.2597, 0.3313, 0.3434, 0.2597, 0.2597, 0.1787, 0.0896, 0.2111, 0.1207, 0.1828, 0.1963, 0.2368, 0.3948, 0.3475, 0.3867, 0.4042, 0.4893, 0.5338, 0.5595, 0.523, 0.5743, 0.6526, 0.6513, 0.654, 0.6499, 0.7445, 0.7553, 0.7458, 0.7364, 0.7364, 0.9537, 0.9537, 1.1617, 1.2332, 1.2764, 1.5384, 1.4776, 1.668, 1.6599, 1.456, 1.4844, 1.3561, 1.3561, 1.6194, 1.6207, 1.6207, 1.7706, 2.0285, 2.0911, 2.1505, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2019263889544999267", "ticker_idx": 0, "ts": "2026-02-05T04:16:42+00:00", "symbol": "MSTR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "short MSTR", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Short MSTR.", "resolved_tickers": ["MSTR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Application", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@zephyr_z9 short MSTR", "reply_text": null, "parent_text": "Are crypto bros OK??", "quoted_text": null, "parent_author": "zephyr_z9", "ret_m1d": 0.2065613507373001, "ret_signed_m1d": -0.2065613507373001, "alpha_spy_m1d": -0.19391418637356783, "alpha_c_m1d": -0.19391418637356783, "bench_spy_m1d": 0.012647164363732255, "bench_c_m1d": 0.012647164363732255, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.2611458582099211, "ret_signed_p1d": -0.2611458582099211, "alpha_spy_p1d": -0.2419611463689697, "alpha_c_p1d": -0.2419611463689697, "bench_spy_p1d": 0.019184711840951385, "bench_c_p1d": 0.019184711840951385, "ret_p1w": 0.14964017623980164, "ret_signed_p1w": -0.14964017623980164, "alpha_spy_p1w": -0.1442536518626154, "alpha_c_p1w": -0.1442536518626154, "bench_spy_p1w": 0.00538652437718623, "bench_c_p1w": 0.00538652437718623, "ret_p1m": 0.248060579918141, "ret_signed_p1m": -0.248060579918141, "alpha_spy_p1m": -0.25579358735707036, "alpha_c_p1m": -0.25579358735707036, "bench_spy_p1m": -0.007733007438929351, "bench_c_p1m": -0.007733007438929351, "ret_p3m": 0.746892210750993, "ret_signed_p3m": -0.746892210750993, "alpha_spy_p3m": -0.6758691901986735, "alpha_c_p3m": -0.6758691901986735, "bench_spy_p3m": 0.07102302055231946, "bench_c_p3m": 0.07102302055231946, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [1.2327, 1.1624, 1.0994, 0.9492, 0.867, 0.8265, 0.9329, 0.7432, 0.6556, 0.5936, 0.6734, 0.6094, 0.6416, 0.6416, 0.656, 0.6022, 0.6948, 0.7608, 0.7386, 0.673, 0.7169, 0.7664, 0.7258, 0.7132, 0.6492, 0.5149, 0.5656, 0.499, 0.479, 0.5405, 0.5358, 0.4757, 0.4834, 0.4834, 0.4843, 0.4524, 0.4544, 0.4202, 0.4202, 0.4689, 0.5396, 0.4765, 0.5126, 0.5606, 0.4705, 0.5163, 0.6169, 0.6761, 0.5974, 0.6236, 0.6236, 0.4976, 0.5311, 0.5046, 0.5245, 0.5009, 0.5102, 0.481, 0.3383, 0.3993, 0.3051, 0.2455, 0.2066, 0.0, 0.2611, 0.294, 0.2431, 0.1783, 0.1496, 0.2513, 0.2513, 0.2026, 0.1702, 0.2099, 0.2249, 0.1563, 0.1647, 0.2679, 0.2468, 0.2104, 0.2866, 0.2401, 0.3687, 0.3068, 0.2481, 0.2987, 0.2941, 0.2929, 0.2837, 0.3054, 0.3788, 0.4046, 0.3138, 0.2921, 0.268, 0.2917, 0.2735, 0.3004, 0.2425, 0.178, 0.1351, 0.1665, 0.1476, 0.12, 0.12, 0.1935, 0.1564, 0.1992, 0.2044, 0.2024, 0.2371, 0.2843, 0.3416, 0.3921, 0.5564, 0.5965, 0.5326, 0.6764, 0.612, 0.5985, 0.5815, 0.5488, 0.4785, 0.5464, 0.6559, 0.7179, 0.7469, 0.7461, 0.6809, 0.7533, 0.8314, 0.7237, 0.664, 0.7475, 0.6583, 0.5574, 0.5387, 0.5498, 0.5408, 0.4944, 0.4944, 0.4948, 0.4413, 0.4173, 0.487, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2026592400584618433", "ticker_idx": 0, "ts": "2026-02-25T09:37:35+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is expected to push the highest mobile DRAM price increases among the three major memory makers. Even if Samsung MX declines, Micron judges that demand from other customers is more than sufficient.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: pushing the largest mobile DRAM price hikes among major memory makers with sufficient demand regardless of Samsung MX's response.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung to Use 50% Micron DRAM in Galaxy S26 — Pricing Negotiations a Key Variable\n\nSamsung Electronics has been found to be producing the initial batch of its Galaxy S26 series — set to launch on the 26th — with Micron DRAM comprising 50% of the mix. Samsung's Mobile eXperience (MX) division has managed to secure volumes more easily amid the recent memory supply shortage, but has been unable to fend off component price pressure.\n\nThe DS (Device Solutions) division has already doubled the price of LPDDR5X supplied to Apple, and is unlikely to offer parts to MX at a discount either. With Micron also signaling steep price hikes, significant pressure on MX's profitability is expected, which is predicted to drive price increases for the Galaxy S26 as well.\n\nA source familiar with Samsung's internal affairs said, \"Samsung's MX division is producing the Galaxy S26 series initial batch with LPDDR5X from the DS division and Micron at a 50/50 split,\" adding, \"production volumes beyond the initial batch have not yet been determined, so there is effectively no concept of a 'first vendor' at this point.\"\n\nThe mobile DRAM being supplied this time has seen prices rise across the board for both Samsung DS and Micron products, though pricing is understood to have been set at levels below what each company charges competitors. In DS's case, the price increase is reportedly smaller than what it is charging Apple for iPhone DRAM.\n\nFor the preceding Galaxy S25 series, Micron DRAM was used exclusively for the first three months after launch. At the time, Micron's products were assessed as superior to Samsung DS's offerings in terms of performance and yield. DS subsequently improved quality, rapidly boosting its mobile DRAM competitiveness, and MX readjusted its sourcing mix. The Galaxy S25 series ultimately settled at a supply structure of roughly 60% DS and 40% Micron.\n\nAn industry source noted, \"Up until about two years ago, DS's consumer DRAM had intermittent recall (refund) issues, but quality has improved significantly through design margin enhancements.\"\n\nStarting with the Galaxy S26 series, however, a conventional memory supply shortage has compounded the situation, making volume procurement MX's top priority. The division engaged in intense negotiations with both DS and Micron during Q4 last year, ultimately settling on a 50/50 supply split for the initial batch.\n\nBoth suppliers are currently providing LPDDR5X based on the 1b process node. Micron has also submitted LPDDR5X samples based on its 1-gamma (γ; equivalent to 1c) process to MX, but adoption was deferred as it was deemed premature for actual deployment. Instead, for this initial batch, Micron is supplying a product with modestly refined circuit linewidths compared to the LPDDR5X used in the Galaxy S25 series, achieving a 15% improvement in productivity — an opportunity for Micron to enhance profitability.\n\nWith Galaxy Unpacked and the S26 reveal approaching, MX is reportedly relieved to have secured initial memory volumes smoothly. The supply mix beyond the initial batch is still under negotiation, though industry observers believe a similar structure is likely to be maintained.\n\nExperts, however, view pricing negotiations as the critical issue. Both DS and Micron plan to push for substantial LPDDR5X price increases in post-initial-batch negotiations. MX plans to partially offset cost pressures by raising Galaxy S26 series pricing and incorporating its in-house Exynos 2600 application processor (AP) in roughly 30% of units. Even so, the consensus is that the steep memory price trajectory will be insufficient to fully relieve profitability pressure.\n\nSamsung's DS division plans to push through with its price hike stance despite being part of the same corporate family. It has already shifted from long-term agreements (LTAs) to quarterly contracts with MX to maximize profitability. Apple — MX's smartphone rival — has also been confirmed to have secured DS's LPDDR5X for the iPhone 17 series at double the previous price. Apple recently held an emergency meeting with DS to negotiate first-half delivery volumes in order to stabilize memory procurement.\n\nA semiconductor industry source said, \"Samsung's DS division originally strategized to raise LPDDR5X prices for Apple's iPhone by about 60%. However, at the first negotiation table, they opened with a 100% increase (intending to converge toward 60%), and Apple accepted it on the spot — so that became the final price. That's how desperate smartphone makers are to secure memory inventory.\"\n\nMicron is also preparing to notify MX of elevated pricing. With limited production capacity, Micron is expected to push the highest mobile DRAM price increases among the three major memory makers. Even if Samsung MX declines, Micron judges that demand from other customers is more than sufficient. Samsung Electronics CEO and DX Division head Roh Tae-moon held an emergency meeting with Micron CEO Sanjay Mehrotra on the first day of CES 2026 to discuss memory supply issues.\n\nAn industry source explained, \"The pricing negotiations following the initial batch will determine the share that DS and Micron each hold within Samsung MX's supply chain. The shortage is so severe that even if MX reduces purchase volumes, neither DS nor Micron would have any reason to be concerned.\"\n\nMeanwhile, Samsung's MX division will hold its \"Galaxy Unpacked 2026\" event at 3:00 AM KST (Korean time) on the 26th in San Francisco, California, to unveil the Galaxy S26 series. Domestic pre-orders will run from the 27th — the day after Unpacked — through March 5, with the official launch scheduled for March 11.\n\n$MU $AAPL\n\n---\n\nhttps://t.co/eqkR2xXDZt", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025617708598947186, "ret_signed_m1d": -0.025617708598947186, "alpha_spy_m1d": -0.01724996955726832, "alpha_c_m1d": -0.00919195315534127, "bench_spy_m1d": -0.008367739041678868, "bench_c_m1d": -0.016425755443605916, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03132867226428593, "ret_signed_p1d": -0.03132867226428593, "alpha_spy_p1d": -0.02577419587266483, "alpha_c_p1d": 0.001874804774595229, "bench_spy_p1d": -0.005554476391621099, "bench_c_p1d": -0.03320347703888116, "ret_p1w": -0.06580425289161662, "ret_signed_p1w": -0.06580425289161662, "alpha_spy_p1w": -0.054233780935947884, "alpha_c_p1w": -0.002306981148468945, "bench_spy_p1w": -0.011570471955668737, "bench_c_p1w": -0.06349727174314768, "ret_p1m": -0.17142192448402027, "ret_signed_p1m": -0.17142192448402027, "alpha_spy_p1m": -0.10462790666251087, "alpha_c_p1m": -0.06507687348258528, "bench_spy_p1m": -0.0667940178215094, "bench_c_p1m": -0.10634505100143499, "ret_p3m": 0.7513181624059206, "ret_signed_p3m": 0.7513181624059206, "alpha_spy_p3m": 0.6726536454338112, "alpha_c_p3m": 0.3989622341823815, "bench_spy_p3m": 0.07866451697210941, "bench_c_p3m": 0.3523559282235391, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.449, -0.4397, -0.442, -0.4544, -0.4719, -0.4473, -0.4247, -0.4118, -0.3855, -0.3978, -0.4381, -0.4466, -0.4582, -0.4745, -0.4209, -0.3804, -0.3555, -0.3563, -0.332, -0.332, -0.3364, -0.3138, -0.3179, -0.3347, -0.3347, -0.2648, -0.2724, -0.1995, -0.2085, -0.2377, -0.1956, -0.1938, -0.2118, -0.223, -0.2153, -0.1544, -0.1544, -0.1492, -0.093, -0.0732, -0.0684, -0.093, -0.0437, 0.0146, 0.0158, -0.0329, 0.0205, -0.0223, -0.1156, -0.1075, -0.08, -0.1061, -0.13, -0.0435, -0.035, -0.0404, -0.0404, -0.0681, -0.0188, -0.0272, -0.0019, -0.0187, -0.0256, 0.0, -0.0313, -0.0388, -0.0381, -0.115, -0.0658, -0.0745, -0.1368, -0.0925, -0.0603, -0.024, -0.0551, -0.0067, 0.0298, 0.0762, 0.0763, 0.0356, -0.0142, -0.0575, -0.078, -0.1093, -0.1714, -0.1673, -0.2496, -0.2122, -0.1422, -0.1459, -0.1459, -0.1191, -0.1195, -0.0515, -0.017, -0.0192, -0.0053, 0.0859, 0.0639, 0.0663, 0.0612, 0.0457, 0.0479, 0.1368, 0.1234, 0.1583, 0.2233, 0.176, 0.209, 0.206, 0.2644, 0.3443, 0.4929, 0.5545, 0.5079, 0.7415, 0.8547, 0.7877, 0.8741, 0.8096, 0.6899, 0.5893, 0.6294, 0.707, 0.7772, 0.7513, 0.7513, 1.0892, 1.165, 1.1536, 1.2644, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012469559207592120", "ticker_idx": 0, "ts": "2026-01-17T10:18:27+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we raise our PT from $40 to $49 but maintain our Neutral rating", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS raises INTC PT from $40 to $49 on improved product OpM outlook and foundry 14A narrative, while maintaining Neutral; shared as an endorsement of the raised target.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: Intel FQ4 Earnings Preview\n\nSummary Thoughts\n\nDespite supply that remains tight, we still see an upside bias for FQ4 (Dec) results on PC/server demand strength and see Street numbers (which are about in-line w/normal seasonal) as okay for FQ1 (March) guidance - especially off the stronger Dec Q base. \n\nFundamentally, we see a mixed (albeit improving) picture this year as CQ1 should also be the worst of INTC's capacity issues (mix shift off Intel 10/7 to more volume for Granite Rapids (Intel 3) and ramp of Panther Lake (18A)). On the flipside though, history tells us that we should expect memory price increases to arrest some of the recent positive momentum in the PC market (see Figure 2). \n\nIf INTC guides 2026 revenue (unlikely), we would see low single digits as a reasonable (if not conservative) bar but expect some broader commentary on gross margin (drop-through 40-60%, UBSe full year ~38%), gross/net capex (UBSe ~$19B/$12B or both up slightly Y/Y), and opex $16B (in-line w/ prior guide). \n\nMore important though than fundamentals will be the \"narrative\" around foundry - where we believe things are improving and several potential customers may be engaged for 14A (NVDA (gaming) and potentially Google, AVGO, AAPL (maybe M-Series?)). Momentum on the 14A narrative should start to build once the PDK for the engineering version of the process is out (we think this Q) and PDK 1.0 is released more toward year-end (this would enable customer designs by 1H:28, now only ~2yrs out). \n\nWe maintain our estimates for INTC's C25/C26 Revs of $53B/$55B (vs Street at $54B/$56B) and our C25/C26 EPSε at $0.37/$0.51 (vs consensus $0.34/$0.59) as we remain cautious on margins in light of DC/Client share losses. Despite no change to estimates, we are feeling a little better about the OpM that the standalone product business may be able to sustain and now see something more in the mid 20%s (see Figure 12). When considered together with current foundry valuations per installed wafer capacity + higher MBLY valuation, we raise our PT from $40 to $49 but maintain our Neutral rating.\n\nData Center Demand Likely More Durable vs PCs\n\nAt our Conference last month, INTC reiterated a strong demand environment for both DC and Client products. Given the limited supply on INTC 7/10 nodes, the company is de-emphasizing lower-end PC products which we think will offer near-term tailwinds for the Client business. \n\nWhile we remain bullish on Data Center demand (UBS forecasts traditional servers growing 6% Y/Y - Figure 1), we believe PC demand may moderate from increasing memory pricing given memory accounts for 25-30% of PC BOM. UBS recently revised PC growth forecasts for C2025F/2026F to 7%/4% from prior 4%/3% and another 2% growth Y/Y in C2027F (Figure 4). Total PC units are expected to decline 8% and 8% in Dec Q and Mar Q, but we expect INTC revenues to outperform PC units in Dec Q as it benefits from a richer mix.\n\nWe Expect Upwards Bias for Capex\n\nAccording to INTC, earlier in C2025 the team was emphatic that C2026 Capex would have been down Y/Y from $18B, but we noted a shift in tone at the last earnings call and at our Conference as INTC was suggesting that it continues seeing durability of demand environment as well as CEO's increasingly bullish 14A narrative: we model C2026E Gross Capex at $19B but think it may be biased higher, particularly as INTC's potential customers make decisions on 14A platform in 2H26.\n\nPT $49 (from $40)\n\nWe raise our PT from $40 to $49 based on our SOTP valuation framework that values each business separately. For our SOTP valuation, we previously modelled eventual run rate INTC Products OpM at 20% (below peer group and INTC's historical Products segment OpM), assuming a long-term significant margin reduction for Product business in light of INTC's struggles to control OpEx coupled with competitive pressures in the CPU market. \n\nWe now increase our OpM estimate to mid-20%, closer to INTC's previously reported 24% (in C23) and as we see broader peer group OpM Avg at 25% in C26 and expanding 300bps to 28% in C27 as we think the mgmt. team is focusing on improving long-term competitiveness of INTC Products. \n\nWe value INTC's product business on a consistent 15x EV/EBIT multiple, and netting off the non-controlling interest (basically debt represented by the SCIP agreements for Arizona and Ireland). Our framework looks at an independent valuation for Intel's manufacturing assets based on how other global foundry comps are valued per installed wafer of capacity. \n\nWe value INTC's wafer capacity using a higher EV per capacity of $193 (vs. prior $158) based on the same group of peer foundries as well as TSMC (applying a 45% haircut to reflect node and yield gaps, customer stickiness, and utilization risk). See Valuation and Multiple Justification.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.034071599659536655, "ret_signed_p1d": 0.034071599659536655, "alpha_spy_p1d": 0.0544283853830817, "alpha_c_p1d": 0.05904724748826806, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": -0.09518734014379204, "ret_signed_p1w": -0.09518734014379204, "alpha_spy_p1w": -0.09673433127126752, "alpha_c_p1w": -0.0912661451419704, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": -0.016609855078868252, "ret_signed_p1m": -0.016609855078868252, "alpha_spy_p1m": -0.003872374509167109, "alpha_c_p1m": -0.03439250413778461, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.4586882737535989, "ret_signed_p3m": 0.4586882737535989, "alpha_spy_p3m": 0.4414598026397183, "alpha_c_p3m": 0.3227958409907441, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2138, -0.1874, -0.1848, -0.158, -0.1156, -0.1197, -0.1448, -0.1484, -0.1589, -0.2115, -0.1827, -0.207, -0.188, -0.1812, -0.1934, -0.1931, -0.2353, -0.2436, -0.2609, -0.269, -0.2523, -0.2841, -0.2653, -0.2379, -0.237, -0.2161, -0.2161, -0.1363, -0.148, -0.0743, -0.0681, -0.1376, -0.1182, -0.1418, -0.1376, -0.1316, -0.1586, -0.1948, -0.2012, -0.2055, -0.2323, -0.2274, -0.2159, -0.2255, -0.2259, -0.23, -0.23, -0.2291, -0.2189, -0.2057, -0.2142, -0.2142, -0.1614, -0.1616, -0.1474, -0.0922, -0.1246, -0.03, -0.0618, 0.007, 0.0375, 0.029, 0.0, 0.0, 0.0341, 0.1552, 0.1567, -0.0402, -0.0952, -0.0645, 0.0388, 0.0362, -0.0104, 0.0394, 0.0488, 0.0349, 0.0273, 0.0773, 0.0698, 0.0036, 0.0283, -0.0102, -0.0036, -0.0036, -0.0166, -0.0319, -0.0498, -0.0607, -0.0709, -0.0179, -0.0017, -0.0319, -0.0287, -0.0311, -0.0822, -0.0294, -0.0215, -0.0754, -0.0294, -0.0038, 0.0217, -0.0364, -0.0253, -0.0256, -0.0618, -0.0411, -0.0166, -0.0658, -0.0628, -0.0618, 0.0047, -0.0609, -0.0816, -0.1229, -0.0603, 0.0228, 0.0728, 0.0728, 0.0813, 0.1267, 0.2553, 0.3143, 0.3284, 0.388, 0.3588, 0.3829, 0.4587, 0.4587, 0.3991, 0.411, 0.3899, 0.4221, 0.7577, 0.8098, 0.7998, 1.0177, 1.0119, 1.1214, 1.0396, 1.303, 1.4065, 1.3343, 1.6601, 1.7564, 1.5684, 1.5615, 1.4687, 1.3162, 1.3034, 1.3595, 1.5332, 1.5234, 1.552, 1.552, 1.6303, 1.5931, 1.5743, 1.4421, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2021057099434950959", "ticker_idx": 0, "ts": "2026-02-10T03:02:16+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "maintaining a 'Buy' rating and a target price of 210,000 won", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Kiwoom Securities Buy on Samsung Electronics, PT 210,000 KRW, on non-memory profitability return, Exynos 2700 ramp, and HBM4 pricing tailwinds.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Samsung Expected to Turn Non-Memory Business Profitable Next Year\" – Kiwoom Securities\n\nKiwoom Securities forecast on the 10th that Samsung Electronics' non-memory division \"will return to operating profit next year,\" maintaining a 'Buy' rating and a target price of 210,000 won.\n\nAnalyst Park Yu-ak of the firm said, \"Yield improvements in the 2nm second-generation (SF2P) process to be used for the Exynos 2700, growing customer demand for cost reduction, and improved benchmark performance will lead to greater market share for the Exynos 2700.\" He added, \"The Exynos 2700 will enter full-scale mass production on Samsung Foundry's SF2P process in the second half of this year, and is expected to achieve around 50% adoption within the Galaxy S27.\"\n\nAccordingly, Kiwoom Securities estimates Samsung Electronics' non-memory division will post revenue of 36.4 trillion won next year, up 21% year-over-year, with operating profit turning to a 1.8 trillion won surplus.\n\nPark noted, \"As the market raises its expectations for non-memory earnings, this will serve as additional upside momentum for the stock.\"\n\nKiwoom Securities projects Samsung Electronics' total revenue and operating profit this year at 500 trillion won and 173 trillion won, respectively—up 50% and 295% year-over-year. This is based on expectations that commodity DRAM and NAND prices will rise 109% and 105%, respectively, over the same period. The firm also believes the surge in commodity DRAM prices and profitability will create favorable conditions for price negotiations on sixth-generation High Bandwidth Memory (HBM4).\n\n---\n\nhttps://t.co/rL8U97j4lR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.003618847812344672, "ret_signed_m1d": 0.003618847812344672, "alpha_spy_m1d": 0.000974714376188901, "alpha_c_m1d": -0.0019932507592730353, "bench_spy_m1d": 0.002644133436155771, "bench_c_m1d": 0.005612098571617707, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012062763084409234, "ret_signed_p1d": 0.012062763084409234, "alpha_spy_p1d": 0.012293821898825552, "alpha_c_p1d": 0.009116422051784756, "bench_spy_p1d": -0.00023105881441631837, "bench_c_p1d": 0.0029463410326244777, "ret_p1w": 0.09288297355760289, "ret_signed_p1w": 0.09288297355760289, "alpha_spy_p1w": 0.10627660324928834, "alpha_c_p1w": 0.11441931072764933, "bench_spy_p1w": -0.01339362969168545, "bench_c_p1w": -0.021536337170046438, "ret_p1m": 0.1459589989198513, "ret_signed_p1m": 0.1459589989198513, "alpha_spy_p1m": 0.1687728955910216, "alpha_c_p1m": 0.16083110300453818, "bench_spy_p1m": -0.022813896671170286, "bench_c_p1m": -0.014872104084686866, "ret_p3m": 0.6227729132154569, "ret_signed_p3m": 0.6227729132154569, "alpha_spy_p3m": 0.5541222710479219, "alpha_c_p3m": 0.3899547403220087, "bench_spy_p3m": 0.06865064216753503, "bench_c_p3m": 0.2328181728934482, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.383, -0.4166, -0.3962, -0.413, -0.4208, -0.3962, -0.431, -0.4196, -0.404, -0.383, -0.3788, -0.3968, -0.395, -0.3794, -0.3728, -0.3692, -0.3494, -0.3428, -0.3494, -0.3518, -0.356, -0.3464, -0.371, -0.383, -0.3524, -0.3542, -0.362, -0.3368, -0.3308, -0.3332, -0.3332, -0.2977, -0.2793, -0.2768, -0.2768, -0.2768, -0.225, -0.1671, -0.1622, -0.1496, -0.1628, -0.1616, -0.1628, -0.1701, -0.1538, -0.1321, -0.1019, -0.0995, -0.1242, -0.0983, -0.0814, -0.0826, -0.0826, -0.038, -0.0205, -0.0308, -0.032, -0.0929, 0.0103, 0.0199, -0.0392, -0.0434, 0.0036, 0.0, 0.0121, 0.0772, 0.0929, 0.0929, 0.0929, 0.0929, 0.146, 0.1466, 0.1641, 0.2063, 0.2274, 0.3148, 0.3058, 0.3058, 0.1767, 0.0386, 0.1556, 0.1351, 0.0464, 0.1333, 0.146, 0.1333, 0.1068, 0.1381, 0.1695, 0.2575, 0.2093, 0.2027, 0.1236, 0.1441, 0.1399, 0.0862, 0.0862, 0.0655, 0.0105, 0.1462, 0.0782, 0.1254, 0.1671, 0.1876, 0.2722, 0.2329, 0.245, 0.2148, 0.2481, 0.2753, 0.3145, 0.3055, 0.2964, 0.3236, 0.3145, 0.3568, 0.3266, 0.3568, 0.3417, 0.3659, 0.3327, 0.3327, 0.4052, 0.4052, 0.6077, 0.6409, 0.6228, 0.7255, 0.6862, 0.7165, 0.789, 0.6349, 0.6983, 0.6651, 0.6681, 0.8101, 0.7678, 0.7678, 0.8071, 0.8555, 0.8101, 0.9159, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2017153719067169249", "ticker_idx": 0, "ts": "2026-01-30T08:31:38+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley: Meta Price Target Raised to $825, Meta's Ad Revenue to Overtake Google Search Ads in Q2 2026", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's bullish META call: PT raised to $825, ad revenue to surpass Google Search in Q2 2026 on AI-driven growth.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: Meta Price Target Raised to $825, Meta's Ad Revenue to Overtake Google Search Ads in Q2 2026\n\nIn its latest report, Morgan Stanley predicts that Meta's advertising revenue will surpass Google Search advertising revenue for the first time in the second quarter of 2026.\n\n- While Google Search had a 15% growth rate when its quarterly revenue was at the $57 billion scale, Meta is recording high-speed growth of 34–35% at the same revenue scale.\n\n- In Zuckerberg's eyes, the current system is still at a 'primitive' level. Morgan Stanley believes that if the current system is primitive, long-term opportunities and improvements are likely to bring even more growth.\n\n- According to Morgan Stanley's model measurements, Meta's quarterly ad revenue will officially surpass Google Search ad revenue in Q2 2026, and the gap will widen further from then on.\n\n- This not only symbolizes Meta's strengthening dominance in the advertising market but also proves that AI-based recommendation algorithms have an absolute advantage in securing user time and advertiser budgets.\n\nMorgan Stanley analyzed in detail how Meta is converting AI technology into actual revenue.\n\n- Through more efficient model scaling, Instagram Reels watch time increased by over 30%. Facebook increased views by 7% through architecture simplification.\n\n- The application of GEM, an AI ad ranking model, was expanded to all Reels and trained using double the GPU resources. As a result, Facebook's ad click-through rate (CTR) rose by 3.5%, and Instagram's conversion rate improved by over 1%.\n\n- The introduction of a new runtime model improved Instagram conversion rates by 3%. The rollout of internal AI coding tools for employees increased engineer productivity by 30% in 2025, with some power users showing productivity gains of up to 80%.\n\nRegarding capital expenditure (Capex), a key market interest, Morgan Stanley provides a clear outlook. Meta's 2026 Capex guidance is $115 billion to $135 billion, which will be primarily used for the Super Intelligence Lab and core business infrastructure. Infrastructure spending, including cloud expenses and depreciation, is expected to increase by approximately $36 billion, accounting for about 75% of the increase in operating expenses for 2026.\n\n$META", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03043960599173312, "ret_signed_m1d": 0.03043960599173312, "alpha_spy_m1d": 0.027448103334555718, "alpha_c_m1d": 0.03285466463769404, "bench_spy_m1d": 0.002991502657177403, "bench_c_m1d": -0.0024150586459609213, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014082417618956788, "ret_signed_p1d": -0.014082417618956788, "alpha_spy_p1d": -0.019053779749892308, "alpha_c_p1d": -0.010668000184132831, "bench_spy_p1d": 0.004971362130935519, "bench_c_p1d": -0.0034144174348239575, "ret_p1w": -0.07681786845599503, "ret_signed_p1w": -0.07681786845599503, "alpha_spy_p1w": -0.07486693076235396, "alpha_c_p1w": -0.04084184439846705, "bench_spy_p1w": -0.001950937693641075, "bench_c_p1w": -0.03597602405752798, "ret_p1m": -0.0878436987545671, "ret_signed_p1m": -0.0878436987545671, "alpha_spy_p1m": -0.07976541218341837, "alpha_c_p1m": -0.06943934565756826, "bench_spy_p1m": -0.00807828657114873, "bench_c_p1m": -0.018404353096998838, "ret_p3m": -0.06532746689556757, "ret_signed_p3m": -0.06532746689556757, "alpha_spy_p3m": -0.09647533962346111, "alpha_c_p3m": -0.028501232564128953, "bench_spy_p3m": 0.031147872727893544, "bench_c_p3m": -0.03682623433143861, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1252, -0.1131, -0.1369, -0.133, -0.119, -0.1255, -0.1507, -0.1495, -0.1501, -0.1605, -0.1665, -0.1768, -0.1784, -0.1713, -0.1451, -0.1128, -0.1164, -0.1164, -0.0964, -0.1063, -0.0976, -0.1081, -0.0775, -0.0609, -0.0701, -0.0838, -0.0934, -0.0898, -0.1016, -0.0963, -0.0828, -0.0935, -0.0726, -0.0806, -0.0768, -0.072, -0.0683, -0.0683, -0.0743, -0.0807, -0.0706, -0.0787, -0.0787, -0.0922, -0.0805, -0.078, -0.0946, -0.0983, -0.0885, -0.104, -0.1192, -0.1409, -0.1336, -0.1343, -0.1343, -0.1568, -0.1445, -0.0961, -0.0806, -0.0616, -0.0608, -0.0667, 0.0304, 0.0, -0.0141, -0.0346, -0.0663, -0.0646, -0.0768, -0.0548, -0.0639, -0.0667, -0.0931, -0.1071, -0.1071, -0.1078, -0.1023, -0.1001, -0.0849, -0.1106, -0.1077, -0.0877, -0.083, -0.0954, -0.0878, -0.0857, -0.0681, -0.0781, -0.1, -0.0965, -0.0871, -0.086, -0.1093, -0.1435, -0.1235, -0.1302, -0.14, -0.1525, -0.1707, -0.1562, -0.1718, -0.169, -0.2352, -0.2656, -0.2507, -0.2008, -0.1909, -0.1976, -0.1976, -0.1996, -0.1967, -0.1445, -0.1222, -0.1202, -0.1136, -0.0746, -0.0619, -0.0545, -0.0382, -0.0628, -0.0657, -0.0575, -0.0793, -0.0571, -0.0521, -0.0622, -0.0653, -0.1452, -0.1497, -0.1473, -0.155, -0.1439, -0.1384, -0.1484, -0.1635, -0.1577, -0.1386, -0.1361, -0.142, -0.1462, -0.1582, -0.1548, -0.1516, -0.1475, -0.1475, -0.1446, -0.1126, -0.1126, -0.1165, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2020316143618310650", "ticker_idx": 0, "ts": "2026-02-08T01:57:58+00:00", "symbol": "Silicon Motion Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm now positive on Silicon Motion as well", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author is now bullish on Silicon Motion, excluding Fadu as a comparable.", "resolved_tickers": ["SIMO"], "resolution_method": "exchange_ticker", "resolution_reason": "Silicon Motion Technology listed on Nasdaq as SIMO", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@chipflare I’m now positive on Silicon Motion as well—excluding Korea’s Fadu.", "reply_text": null, "parent_text": "@jukan05 SIMO?", "quoted_text": null, "parent_author": "chipflare", "ret_m1d": -0.06899325751172336, "ret_signed_m1d": -0.06899325751172336, "alpha_spy_m1d": -0.06419456390360512, "alpha_c_m1d": -0.05669767144250826, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.012295586069215103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04529267197250564, "ret_signed_p1d": -0.04529267197250564, "alpha_spy_p1d": -0.042655511540424795, "alpha_c_p1d": -0.040644979462419184, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.004647692510086454, "ret_p1w": -0.005446435296382113, "ret_signed_p1w": -0.005446435296382113, "alpha_spy_p1w": 0.012134134487961079, "alpha_c_p1w": -0.008077708726306643, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": 0.00263127342992453, "ret_p1m": -0.10396760704432417, "ret_signed_p1m": -0.10396760704432417, "alpha_spy_p1m": -0.07980154442927034, "alpha_c_p1m": -0.08104861660019491, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.02291899044412926, "ret_p3m": 0.7748888200784996, "ret_signed_p3m": 0.7748888200784996, "alpha_spy_p3m": 0.7177839163553161, "alpha_c_p3m": 0.4467196729189309, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.3281691471595687, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3404, -0.374, -0.3694, -0.3786, -0.3854, -0.3835, -0.4172, -0.4108, -0.3905, -0.3842, -0.3633, -0.3633, -0.3533, -0.3551, -0.3557, -0.3388, -0.3383, -0.3296, -0.325, -0.3125, -0.3041, -0.322, -0.3623, -0.367, -0.3784, -0.3854, -0.3707, -0.3548, -0.349, -0.3484, -0.3524, -0.3524, -0.344, -0.352, -0.3423, -0.3261, -0.3261, -0.3184, -0.3197, -0.2351, -0.1194, -0.1928, -0.1776, -0.1616, -0.1744, -0.1876, -0.1757, -0.1794, -0.1794, -0.1784, -0.1435, -0.162, -0.179, -0.1774, -0.1549, -0.1487, -0.1554, -0.1354, -0.1294, -0.1245, -0.0667, -0.0827, -0.069, 0.0, -0.0453, 0.0213, 0.0097, -0.0054, -0.0054, -0.0414, -0.0311, -0.0344, -0.0157, -0.0063, -0.0083, 0.0437, -0.0434, -0.0569, -0.0572, -0.1284, -0.1075, -0.0968, -0.1363, -0.1124, -0.104, -0.0977, -0.1235, -0.1005, -0.0807, -0.0792, -0.0848, -0.0695, -0.0968, -0.097, -0.1342, -0.1904, -0.1883, -0.1968, -0.2304, -0.1805, -0.1463, -0.1688, -0.1688, -0.1491, -0.1461, -0.1115, -0.0732, -0.0753, -0.0805, -0.0502, 0.0146, -0.0002, 0.0196, 0.0122, 0.0396, 0.0359, 0.0363, 0.1199, 0.0767, 0.0887, 0.5873, 0.5966, 0.7115, 0.6578, 0.7399, 0.7876, 0.7749, 0.8545, 0.9163, 0.8919, 0.9735, 0.9295, 0.9013, 0.7533, 0.8496, 0.8868, 0.927, 1.0194, 1.0194, 1.1231, 1.084, 1.0654, 1.0247, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2019635381701079514", "ticker_idx": 0, "ts": "2026-02-06T04:52:52+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS raised its price target on Micron to $450", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS Buy-equivalent call on MU, PT $450; memory supply deficit through 2027 + DC demand + margins peaking ~80%.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS raised its price target on Micron to $450 with the following commentary...\n\n\"More importantly, synthesizing industry conversations suggests that memory supply shortages could extend deeper through C2027 — DRAM could remain in a supply deficit through C4Q27, and NAND could be in shortage through C1Q27.\n\nThis is occurring despite our view that accelerating memory price inflation will cause some demand destruction and spec-downs in PCs and smartphones (accordingly, we have lowered related estimates). The key takeaway is that strong data center demand is more than offsetting these headwinds.\n\n(omitted)...\n\nWe are modeling gross margins of 69.9%, with an increase sequentially in the May quarter (MayQ) to approximately 74.4%. Accordingly, we expect gross margins to peak in the low 80% range across C1Q27E/C2Q27E.\"\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029896801235108517, "ret_signed_m1d": -0.029896801235108517, "alpha_spy_m1d": -0.01107321448183185, "alpha_c_m1d": 0.021341806719581524, "bench_spy_m1d": -0.018823586753276667, "bench_c_m1d": -0.05123860795469004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028351368538319544, "ret_signed_p1d": -0.028351368538319544, "alpha_spy_p1d": -0.03317320064133, "alpha_c_p1d": -0.04080001804903921, "bench_spy_p1d": 0.004821832103010459, "bench_c_p1d": 0.012448649510719667, "ret_p1w": 0.042995811110824844, "ret_signed_p1w": 0.042995811110824844, "alpha_spy_p1w": 0.05583931934793307, "alpha_c_p1w": 0.02788313236948481, "bench_spy_p1w": -0.012843508237108225, "bench_c_p1w": 0.015112678741340035, "ret_p1m": -0.013605593725321996, "ret_signed_p1m": -0.013605593725321996, "alpha_spy_p1m": 0.004276773811875745, "alpha_c_p1m": 0.004519636923064674, "bench_spy_p1m": -0.01788236753719774, "bench_c_p1m": -0.01812523064838667, "ret_p3m": 0.6896046503298983, "ret_signed_p3m": 0.6896046503298983, "alpha_spy_p3m": 0.6241357244771648, "alpha_c_p3m": 0.3208507150134883, "bench_spy_p3m": 0.06546892585273345, "bench_c_p3m": 0.36875393531641, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3894, -0.3798, -0.3999, -0.3749, -0.3872, -0.4213, -0.4278, -0.49, -0.4748, -0.4329, -0.4314, -0.4168, -0.4168, -0.4011, -0.391, -0.3935, -0.407, -0.426, -0.3992, -0.3746, -0.3607, -0.3321, -0.3454, -0.3893, -0.3985, -0.4111, -0.4288, -0.3705, -0.3265, -0.2995, -0.3003, -0.274, -0.274, -0.2787, -0.2542, -0.2586, -0.2769, -0.2769, -0.2008, -0.2091, -0.1299, -0.1397, -0.1715, -0.1257, -0.1237, -0.1433, -0.1554, -0.1471, -0.0809, -0.0809, -0.0752, -0.0141, 0.0073, 0.0126, -0.0142, 0.0394, 0.1028, 0.1041, 0.0512, 0.1092, 0.0627, -0.0387, -0.0299, 0.0, -0.0284, -0.0543, 0.0397, 0.0488, 0.043, 0.043, 0.0129, 0.0665, 0.0574, 0.0848, 0.0666, 0.0591, 0.0869, 0.0529, 0.0448, 0.0456, -0.038, 0.0154, 0.006, -0.0618, -0.0136, 0.0213, 0.0608, 0.027, 0.0797, 0.1194, 0.1698, 0.1699, 0.1256, 0.0715, 0.0245, 0.0021, -0.0319, -0.0994, -0.0949, -0.1843, -0.1437, -0.0676, -0.0717, -0.0717, -0.0425, -0.0429, 0.0309, 0.0684, 0.0661, 0.0812, 0.1803, 0.1564, 0.1589, 0.1535, 0.1366, 0.139, 0.2356, 0.221, 0.259, 0.3296, 0.2782, 0.3141, 0.3108, 0.3743, 0.4611, 0.6227, 0.6896, 0.639, 0.8929, 1.0159, 0.943, 1.037, 0.967, 0.8368, 0.7275, 0.7711, 0.8554, 0.9317, 0.9036, 0.9036, 1.2708, 1.3532, 1.3408, 1.4612, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014877118702956833", "ticker_idx": 0, "ts": "2026-01-24T01:45:14+00:00", "symbol": "VNET", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$VNET looks poised to benefit significantly from this", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long VNET as a key beneficiary of ByteDance's RMB 300B capex increase.", "resolved_tickers": ["VNET"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Information Technology Services'", "bench_c_industry": "Information Technology Services", "unit_kind": "original", "lang": "en", "tweet_text": "ByteDance reportedly increased its capex to RMB 300 billion this year.\n\nFrom my perspective, $VNET looks poised to benefit significantly from this. https://t.co/bslzCf4OT5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.024816218283426617, "ret_signed_m1d": -0.024816218283426617, "alpha_spy_m1d": -0.019763728508913947, "alpha_c_m1d": -0.01803901306141953, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.042279414826146944, "ret_signed_p1d": 0.042279414826146944, "alpha_spy_p1d": 0.03829515009646234, "alpha_c_p1d": 0.028793716652481427, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": 0.0036764670434088487, "ret_signed_p1w": 0.0036764670434088487, "alpha_spy_p1w": -0.00019232583666939362, "alpha_c_p1w": 0.009289956234717733, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.10110288752068963, "ret_signed_p1m": 0.10110288752068963, "alpha_spy_p1m": 0.10886927498166599, "alpha_c_p1m": 0.14053337679702604, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": -0.18749999452163202, "ret_signed_p3m": -0.18749999452163202, "alpha_spy_p3m": -0.2129858774474891, "alpha_c_p3m": -0.25563903510799335, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0147, -0.0432, -0.0441, -0.068, -0.1471, -0.0965, -0.0836, -0.1131, -0.1085, -0.1002, -0.1756, -0.1939, -0.1912, -0.2114, -0.2031, -0.2353, -0.2445, -0.2289, -0.1829, -0.1783, -0.1811, -0.1811, -0.1783, -0.1756, -0.1939, -0.1829, -0.1866, -0.1783, -0.1425, -0.125, -0.1645, -0.1507, -0.1562, -0.1811, -0.1958, -0.2233, -0.2188, -0.1958, -0.2004, -0.2086, -0.2114, -0.2114, -0.2013, -0.2059, -0.2206, -0.2224, -0.2224, -0.1627, -0.1415, -0.1461, -0.1443, -0.0515, -0.0423, 0.046, 0.0221, 0.0974, 0.0561, -0.0294, -0.0294, -0.1287, -0.0699, -0.0432, -0.0248, 0.0, 0.0423, 0.0377, 0.011, -0.0322, 0.0037, -0.0018, -0.0561, -0.0699, 0.0294, 0.0892, 0.1094, 0.2445, 0.2895, 0.2684, 0.2684, 0.2638, 0.1452, 0.1443, 0.0892, 0.0322, 0.1011, 0.0836, 0.0469, -0.0092, -0.0395, -0.0938, -0.057, -0.0993, -0.1158, -0.1002, 0.0322, 0.0248, -0.0285, -0.034, -0.1241, -0.1425, -0.1296, -0.1397, -0.1893, -0.1756, -0.1903, -0.1627, -0.1994, -0.2233, -0.2702, -0.2289, -0.2105, -0.2472, -0.2472, -0.2555, -0.2693, -0.2335, -0.2243, -0.2261, -0.2243, -0.1719, -0.1719, -0.148, -0.1369, -0.1544, -0.2022, -0.1939, -0.1875, -0.1526, -0.1875, -0.2188, -0.2371, -0.2362, -0.2335, -0.2215, -0.2206, -0.1719, -0.1765, -0.182, -0.1599, -0.171, 0.0368, -0.0395, -0.0726, -0.0331, -0.0726, -0.0616, -0.0873, -0.1232, -0.1232, -0.0882, -0.0156, -0.0239, -0.0735, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2009245297525051675", "ticker_idx": 0, "ts": "2026-01-08T12:46:23+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect this ongoing supply-demand tightness to continue to support pricing strength through CY26 at a minimum (JPMe forecasts average DRAM pricing up almost 60% Y/Y in CY26)", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM/Micron mgmt: DRAM/HBM supply deficit persists beyond 2026; DRAM pricing up ~60% Y/Y in CY26; bullish MU on sustained supply-demand tightness.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "From JPMorgan’s meeting with Micron management:\n\nExpect DRAM/HBM demand to outstrip supply beyond 2026, even as new capacity comes online, underpinning further strength in pricing. Mgmt reiterated its previous assertion that it is only able to serve 50% to two-thirds of key customers’ medium-term bit demand, adding that even with greenfield clean room capacity starting to come online in CY27, demand will likely continue to outstrip supply. In terms of capacity additions, MU has already pulled forward the timing of first wafer-out for its Idaho 1 fab by about one quarter to mid-CY27, but emphasized that the ramp-up of this greenfield capacity will be gradual (not constrained by capital but rather by physical limitations), and this will be similar for competitors ramping greenfield capacity in late-2027 and 2028, all while demand continues to move up and to the right. On net, we expect this ongoing supply-demand tightness to continue to support pricing strength through CY26 at a minimum (JPMe forecasts average DRAM pricing up almost 60% Y/Y in CY26).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.038315736170364545, "ret_signed_m1d": 0.038315736170364545, "alpha_spy_m1d": 0.03821419317844654, "alpha_c_m1d": 0.022378278119149053, "bench_spy_m1d": 0.0001015429919180022, "bench_c_m1d": 0.015937458051215492, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05525664460054358, "ret_signed_p1d": 0.05525664460054358, "alpha_spy_p1d": 0.04864330220645097, "alpha_c_p1d": 0.028236775643010636, "bench_spy_p1d": 0.006613342394092614, "bench_c_p1d": 0.027019868957532944, "ret_p1w": 0.029386653765638426, "ret_signed_p1w": 0.029386653765638426, "alpha_spy_p1w": 0.025427364694398102, "alpha_c_p1w": -0.01660519849701303, "bench_spy_p1w": 0.003959289071240324, "bench_c_p1w": 0.045991852262651456, "ret_p1m": 0.2069293047868832, "ret_signed_p1m": 0.2069293047868832, "alpha_spy_p1m": 0.20531952882409765, "alpha_c_p1m": 0.14711089153823798, "bench_spy_p1m": 0.0016097759627855535, "bench_c_p1m": 0.05981841324864523, "ret_p3m": 0.15509334962517585, "ret_signed_p3m": 0.15509334962517585, "alpha_spy_p3m": 0.1964121161547595, "alpha_c_p3m": 0.09989259416656959, "bench_spy_p3m": -0.04131876652958366, "bench_c_p3m": 0.055200755458606254, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4108, -0.4282, -0.4133, -0.3809, -0.3814, -0.368, -0.3817, -0.3933, -0.3682, -0.3305, -0.3272, -0.3217, -0.3073, -0.3153, -0.316, -0.2826, -0.3336, -0.274, -0.2715, -0.2728, -0.2257, -0.263, -0.2514, -0.2757, -0.2455, -0.2604, -0.3015, -0.3094, -0.3845, -0.3661, -0.3155, -0.3137, -0.2962, -0.2962, -0.2772, -0.265, -0.268, -0.2842, -0.3072, -0.2749, -0.2452, -0.2284, -0.1939, -0.21, -0.2629, -0.274, -0.2893, -0.3107, -0.2403, -0.1872, -0.1546, -0.1555, -0.1237, -0.1237, -0.1295, -0.0998, -0.1052, -0.1272, -0.1272, -0.0355, -0.0455, 0.0502, 0.0383, 0.0, 0.0553, 0.0576, 0.034, 0.0194, 0.0294, 0.1093, 0.1093, 0.1161, 0.1899, 0.2158, 0.2221, 0.1898, 0.2545, 0.3311, 0.3326, 0.2687, 0.3388, 0.2826, 0.1602, 0.1708, 0.2069, 0.1727, 0.1414, 0.2548, 0.2659, 0.2588, 0.2588, 0.2225, 0.2872, 0.2762, 0.3093, 0.2873, 0.2782, 0.3118, 0.2707, 0.261, 0.2619, 0.161, 0.2255, 0.2141, 0.1323, 0.1905, 0.2327, 0.2803, 0.2395, 0.3031, 0.351, 0.4118, 0.4119, 0.3585, 0.2932, 0.2365, 0.2095, 0.1684, 0.087, 0.0923, -0.0155, 0.0335, 0.1253, 0.1204, 0.1204, 0.1556, 0.1551, 0.2443, 0.2895, 0.2867, 0.3049, 0.4245, 0.3957, 0.3988, 0.3922, 0.3718, 0.3747, 0.4913, 0.4737, 0.5196, 0.6047, 0.5427, 0.5861, 0.5821, 0.6587, 0.7635, 0.9585, 1.0392, 0.9782, 1.2846, 1.4331, 1.3451, 1.4585, 1.374, 1.2169, 1.085, 1.1376, 1.2393, 1.3314, 1.2975, 1.2975, 1.7407, 1.8402, 1.8252, 1.9705, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2009245297525051675", "ticker_idx": 1, "ts": "2026-01-08T12:46:23+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Expect DRAM/HBM demand to outstrip supply beyond 2026, even as new capacity comes online, underpinning further strength in pricing", "tweet_type": "news_reaction", "timeline": "years", "summary": "JPM/Micron mgmt: DRAM/HBM supply deficit persists beyond 2026; DRAM pricing up ~60% Y/Y in CY26; bullish MU on sustained supply-demand tightness.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "From JPMorgan’s meeting with Micron management:\n\nExpect DRAM/HBM demand to outstrip supply beyond 2026, even as new capacity comes online, underpinning further strength in pricing. Mgmt reiterated its previous assertion that it is only able to serve 50% to two-thirds of key customers’ medium-term bit demand, adding that even with greenfield clean room capacity starting to come online in CY27, demand will likely continue to outstrip supply. In terms of capacity additions, MU has already pulled forward the timing of first wafer-out for its Idaho 1 fab by about one quarter to mid-CY27, but emphasized that the ramp-up of this greenfield capacity will be gradual (not constrained by capital but rather by physical limitations), and this will be similar for competitors ramping greenfield capacity in late-2027 and 2028, all while demand continues to move up and to the right. On net, we expect this ongoing supply-demand tightness to continue to support pricing strength through CY26 at a minimum (JPMe forecasts average DRAM pricing up almost 60% Y/Y in CY26).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01188251392102905, "ret_signed_m1d": 0.01188251392102905, "alpha_spy_m1d": 0.011780970929111048, "alpha_c_m1d": -0.004054944130186441, "bench_spy_m1d": 0.0001015429919180022, "bench_c_m1d": 0.015937458051215492, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013608208930443308, "ret_signed_p1d": 0.013608208930443308, "alpha_spy_p1d": 0.006994866536350694, "alpha_c_p1d": -0.013411660027089637, "bench_spy_p1d": 0.006613342394092614, "bench_c_p1d": 0.027019868957532944, "ret_p1w": 0.018956978405177145, "ret_signed_p1w": 0.018956978405177145, "alpha_spy_p1w": 0.01499768933393682, "alpha_c_p1w": -0.02703487385747431, "bench_spy_p1w": 0.003959289071240324, "bench_c_p1w": 0.045991852262651456, "ret_p1m": 0.15312302386597829, "ret_signed_p1m": 0.15312302386597829, "alpha_spy_p1m": 0.15151324790319273, "alpha_c_p1m": 0.09330461061733306, "bench_spy_p1m": 0.0016097759627855535, "bench_c_p1m": 0.05981841324864523, "ret_p3m": 0.2625352565581081, "ret_signed_p3m": 0.2625352565581081, "alpha_spy_p3m": 0.30385402308769177, "alpha_c_p3m": 0.20733450109950186, "bench_spy_p3m": -0.04131876652958366, "bench_c_p3m": 0.055200755458606254, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3978, -0.4092, -0.3912, -0.3608, -0.3547, -0.341, -0.3498, -0.35, -0.3479, -0.316, -0.2962, -0.3065, -0.283, -0.2727, -0.2688, -0.2222, -0.269, -0.2625, -0.2588, -0.2681, -0.2345, -0.2342, -0.2322, -0.2432, -0.2695, -0.246, -0.2823, -0.2916, -0.3028, -0.3326, -0.3116, -0.3052, -0.2888, -0.2782, -0.2852, -0.2769, -0.2628, -0.2683, -0.2789, -0.2594, -0.2323, -0.2342, -0.2144, -0.2311, -0.2423, -0.2633, -0.2837, -0.2694, -0.2462, -0.2338, -0.1984, -0.1945, -0.1831, -0.1831, -0.1661, -0.1308, -0.1267, -0.1341, -0.1341, -0.0714, -0.0433, 0.0037, 0.0119, 0.0, 0.0136, 0.0161, 0.0005, 0.0039, 0.019, 0.0607, 0.0652, 0.0484, 0.0819, 0.1039, 0.1108, 0.0864, 0.1539, 0.2045, 0.2098, 0.2091, 0.1734, 0.2297, 0.1896, 0.1441, 0.1531, 0.1816, 0.1649, 0.2004, 0.2424, 0.2428, 0.2428, 0.2307, 0.2522, 0.2759, 0.3114, 0.3119, 0.3495, 0.3748, 0.4326, 0.4089, 0.4092, 0.2703, 0.1971, 0.2805, 0.2376, 0.1828, 0.2765, 0.3049, 0.2752, 0.277, 0.3337, 0.3647, 0.4378, 0.3818, 0.3548, 0.2717, 0.2943, 0.2829, 0.207, 0.2088, 0.138, 0.1034, 0.2277, 0.1694, 0.2085, 0.2413, 0.2625, 0.3776, 0.3616, 0.3783, 0.3781, 0.459, 0.4748, 0.4999, 0.4821, 0.4885, 0.526, 0.5608, 0.5726, 0.5745, 0.6459, 0.6227, 0.6437, 0.6261, 0.6516, 0.7865, 0.8515, 1.0271, 1.0434, 1.1525, 1.3285, 1.2637, 1.3758, 1.3739, 1.1934, 1.184, 1.1463, 1.1815, 1.3548, 1.3271, 1.3271, 1.5395, 1.6763, 1.6738, 1.7838, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2004701853062496651", "ticker_idx": 0, "ts": "2025-12-26T23:52:21+00:00", "symbol": "SMIC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Within China, companies such as SMIC, Hua Hong Semiconductor, United Nova Technology, and CR Micro that possess BCD platforms are at the center of this benefit cycle", "tweet_type": "analysis", "timeline": "quarters", "summary": "AI-driven PMIC/BCD demand is tightening legacy-node capacity; SMIC and Hua Hong Semiconductor named as prime beneficiaries of the ongoing price-hike cycle.", "resolved_tickers": ["0981.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "SMIC listed on HKEX as 0981.HK (also 688981.SS)", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI Infrastructure Investment Surge: Is the Next Bottleneck PMICs?\n\nWhile the semiconductor market’s attention is fixated on leading-edge technologies such as HBM and GPUs, subtle but meaningful tectonic shifts are being detected in the legacy-node market. Signals are emerging that the boom in AI infrastructure investment is tightening not only the leading-edge supply chain, but also the legacy-node ecosystem.\n\nLooking across recent industry developments, a new wave of foundry price hikes appears to be spreading. According to reports by China’s Security Times and Kechuangban Daily, China’s largest foundry SMIC and TSMC’s subsidiary VIS (Vanguard International Semiconductor) have moved in tandem to raise prices. In particular, SMIC has sent formal price-increase notices to downstream customers, primarily targeting 8-inch BCD (Bipolar-CMOS-DMOS) processes, with an average hike of around 10%. VIS has likewise notified customers of a similar increase and has joined this broader trend.\n\nEveryone who saw this news must have been pretty stunned. Weren’t legacy foundries the kind of business where prices collapsed day after day because of a China-driven chicken game? Apparently not. The surge in AI infrastructure investment is bringing warmth to legacy foundries as well. High-performance AI servers and data centers consume enormous amounts of power, and precisely controlling that power requires large volumes of power management ICs (PMICs). The BCD process—central to PMIC production—is a technology that maximizes power efficiency by integrating analog, digital, and power circuits into a single platform. As demand for AI servers explodes, BCD manufacturing capacity (Capa) is being rapidly depleted.\n\nWithin China, companies such as SMIC, Hua Hong Semiconductor, United Nova Technology, and CR Micro that possess BCD platforms are at the center of this benefit cycle. The key point is that this round of price hikes is unlikely to stop at BCD. Industry sources are identifying high-voltage CMOS (HV-CMOS) as the next target for price increases. As BCD demand surges and 8-inch line utilization reaches saturation—pushing BCD pricing higher—HV-CMOS, which had been relatively cheaper, is also likely to rise in a “level-setting” move.\n\nIn conclusion, amid structural supply constraints driven by TSMC scaling back 8-inch investment in order to focus on leading-edge nodes, AI-triggered PMIC demand is creating an unexpected “supplier’s market” in the legacy-node space. This suggests that the semiconductor cycle is not confined to cutting-edge memory alone, but that warmth is spreading into legacy processes that underpin power efficiency as well.\n\n---\n\nhttps://t.co/nnivZz9Nyq", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.00010142531829515633, "alpha_c_m1d": 0.00461917671101475, "bench_spy_m1d": 0.00010142531829515633, "bench_c_m1d": -0.00461917671101475, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019732120601860914, "ret_signed_p1d": -0.019732120601860914, "alpha_spy_p1d": -0.01616849239480611, "alpha_c_p1d": -0.0152222152759236, "bench_spy_p1d": -0.003563628207054803, "bench_c_p1d": -0.004509905325937313, "ret_p1w": 0.058491919723496544, "ret_signed_p1w": 0.058491919723496544, "alpha_spy_p1w": 0.0688350857272193, "alpha_c_p1w": 0.03815626472078559, "bench_spy_p1w": -0.010343166003722759, "bench_c_p1w": 0.02033565500271095, "ret_p1m": 0.06483447923008812, "ret_signed_p1m": 0.06483447923008812, "alpha_spy_p1m": 0.061328833678771844, "alpha_c_p1m": -0.025254689556665877, "bench_spy_p1m": 0.003505645551316272, "bench_c_p1m": 0.09008916878675399, "ret_p3m": -0.2036645182491469, "ret_signed_p3m": -0.2036645182491469, "alpha_spy_p3m": -0.15774863831580033, "alpha_c_p3m": -0.29430029420110093, "bench_spy_p3m": -0.04591587993334656, "bench_c_p3m": 0.09063577595195405, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1212, 0.1212, 0.2636, 0.2812, 0.2833, 0.2833, 0.2615, 0.1769, 0.093, 0.1297, 0.0338, 0.0712, 0.0416, -0.0261, 0.012, 0.0437, 0.055, 0.0437, 0.1276, 0.167, 0.129, 0.129, 0.1163, 0.0571, 0.0268, 0.0127, 0.0106, 0.0846, 0.0634, 0.0529, 0.024, 0.0324, 0.0655, 0.0359, 0.0282, 0.043, 0.0289, 0.0359, -0.0303, -0.0409, -0.0423, -0.0296, -0.0366, -0.0303, -0.0211, -0.0324, -0.0529, -0.0162, -0.0169, 0.012, -0.0296, -0.0317, -0.0536, -0.0458, -0.0881, -0.1057, -0.0874, -0.0881, -0.0832, -0.0289, -0.0289, 0.0, 0.0, 0.0, -0.0197, 0.0218, 0.0056, 0.0056, 0.0585, 0.0782, 0.0655, 0.0529, 0.0564, 0.0486, 0.0613, 0.0493, 0.0705, 0.0902, 0.1163, 0.0853, 0.05, 0.0888, 0.0973, 0.1008, 0.0648, 0.0796, 0.1177, 0.0888, 0.0627, 0.0176, -0.007, -0.031, -0.0493, -0.0472, -0.0085, 0.0085, -0.0134, -0.0162, -0.0085, -0.0078, -0.0078, -0.0078, -0.0078, -0.0458, 0.0021, -0.0218, -0.0155, -0.0472, -0.0416, -0.0895, -0.1184, -0.1367, -0.1332, -0.129, -0.1424, -0.0951, -0.1015, -0.1106, -0.1233, -0.1191, -0.1283, -0.1304, -0.1579, -0.198, -0.2368, -0.222, -0.2037, -0.2509, -0.26, -0.2755, -0.2854, -0.2551, -0.2812, -0.2812, -0.2812, -0.2812, -0.2086, -0.2149, -0.179, -0.1945, -0.1832, -0.1642, -0.155, -0.1635, -0.1572, -0.1529, -0.1642, -0.1762, -0.0937, -0.0381, -0.0691, -0.0726, -0.0007, -0.0007, 0.0169, -0.0021, 0.055, 0.0817, 0.0338, 0.0796, 0.0796, 0.0451, 0.0078, 0.0028, -0.0317, -0.0345, 0.0592, 0.0458, 0.1254, 0.1254, 0.1896, 0.2008, 0.2438, 0.1501, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2004701853062496651", "ticker_idx": 1, "ts": "2025-12-26T23:52:21+00:00", "symbol": "Hua Hong Semiconductor", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Within China, companies such as SMIC, Hua Hong Semiconductor, United Nova Technology, and CR Micro that possess BCD platforms are at the center of this benefit cycle", "tweet_type": "analysis", "timeline": "quarters", "summary": "AI-driven PMIC/BCD demand is tightening legacy-node capacity; SMIC and Hua Hong Semiconductor named as prime beneficiaries of the ongoing price-hike cycle.", "resolved_tickers": ["1347.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "Hua Hong Semiconductor 1347.HK Hong Kong", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI Infrastructure Investment Surge: Is the Next Bottleneck PMICs?\n\nWhile the semiconductor market’s attention is fixated on leading-edge technologies such as HBM and GPUs, subtle but meaningful tectonic shifts are being detected in the legacy-node market. Signals are emerging that the boom in AI infrastructure investment is tightening not only the leading-edge supply chain, but also the legacy-node ecosystem.\n\nLooking across recent industry developments, a new wave of foundry price hikes appears to be spreading. According to reports by China’s Security Times and Kechuangban Daily, China’s largest foundry SMIC and TSMC’s subsidiary VIS (Vanguard International Semiconductor) have moved in tandem to raise prices. In particular, SMIC has sent formal price-increase notices to downstream customers, primarily targeting 8-inch BCD (Bipolar-CMOS-DMOS) processes, with an average hike of around 10%. VIS has likewise notified customers of a similar increase and has joined this broader trend.\n\nEveryone who saw this news must have been pretty stunned. Weren’t legacy foundries the kind of business where prices collapsed day after day because of a China-driven chicken game? Apparently not. The surge in AI infrastructure investment is bringing warmth to legacy foundries as well. High-performance AI servers and data centers consume enormous amounts of power, and precisely controlling that power requires large volumes of power management ICs (PMICs). The BCD process—central to PMIC production—is a technology that maximizes power efficiency by integrating analog, digital, and power circuits into a single platform. As demand for AI servers explodes, BCD manufacturing capacity (Capa) is being rapidly depleted.\n\nWithin China, companies such as SMIC, Hua Hong Semiconductor, United Nova Technology, and CR Micro that possess BCD platforms are at the center of this benefit cycle. The key point is that this round of price hikes is unlikely to stop at BCD. Industry sources are identifying high-voltage CMOS (HV-CMOS) as the next target for price increases. As BCD demand surges and 8-inch line utilization reaches saturation—pushing BCD pricing higher—HV-CMOS, which had been relatively cheaper, is also likely to rise in a “level-setting” move.\n\nIn conclusion, amid structural supply constraints driven by TSMC scaling back 8-inch investment in order to focus on leading-edge nodes, AI-triggered PMIC demand is creating an unexpected “supplier’s market” in the legacy-node space. This suggests that the semiconductor cycle is not confined to cutting-edge memory alone, but that warmth is spreading into legacy processes that underpin power efficiency as well.\n\n---\n\nhttps://t.co/nnivZz9Nyq", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.00010142531829515633, "alpha_c_m1d": 0.00461917671101475, "bench_spy_m1d": 0.00010142531829515633, "bench_c_m1d": -0.00461917671101475, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013995801857408008, "ret_signed_p1d": 0.013995801857408008, "alpha_spy_p1d": 0.01755943006446281, "alpha_c_p1d": 0.01850570718334532, "bench_spy_p1d": -0.003563628207054803, "bench_c_p1d": -0.004509905325937313, "ret_p1w": 0.13785873371906354, "ret_signed_p1w": 0.13785873371906354, "alpha_spy_p1w": 0.1482018997227863, "alpha_c_p1w": 0.11752307871635259, "bench_spy_p1w": -0.010343166003722759, "bench_c_p1w": 0.02033565500271095, "ret_p1m": 0.5563331238319658, "ret_signed_p1m": 0.5563331238319658, "alpha_spy_p1m": 0.5528274782806495, "alpha_c_p1m": 0.46624395504521177, "bench_spy_p1m": 0.003505645551316272, "bench_c_p1m": 0.09008916878675399, "ret_p3m": 0.24072785601511315, "ret_signed_p3m": 0.24072785601511315, "alpha_spy_p3m": 0.2866437359484597, "alpha_c_p3m": 0.1500920800631591, "bench_spy_p3m": -0.04591587993334656, "bench_c_p3m": 0.09063577595195405, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1197, 0.1197, 0.1994, 0.2246, 0.2806, 0.2806, 0.2743, 0.1924, 0.1442, 0.2358, 0.0742, 0.1316, 0.14, 0.0609, 0.0903, 0.0875, 0.063, 0.014, 0.1533, 0.2106, 0.2057, 0.2057, 0.2057, 0.1162, 0.077, 0.0609, 0.028, 0.1211, 0.112, 0.0931, 0.0532, 0.042, 0.1008, 0.0812, 0.0854, 0.1232, 0.126, 0.0917, 0.0252, -0.0252, -0.0014, 0.0189, 0.0154, 0.0427, 0.0483, 0.0189, 0.035, 0.0672, 0.0644, 0.112, 0.0518, 0.0644, 0.0133, 0.0063, -0.0581, -0.0721, -0.0455, -0.0588, -0.0518, -0.0028, -0.0161, 0.0, 0.0, 0.0, 0.014, 0.0532, 0.0371, 0.0371, 0.1379, 0.1756, 0.1903, 0.2498, 0.2827, 0.2771, 0.2813, 0.2554, 0.3093, 0.3919, 0.4948, 0.436, 0.408, 0.4808, 0.5283, 0.5367, 0.5563, 0.5983, 0.7173, 0.6263, 0.6305, 0.4472, 0.5073, 0.4332, 0.394, 0.3891, 0.4346, 0.4178, 0.394, 0.3933, 0.3821, 0.4024, 0.4024, 0.4024, 0.4024, 0.3177, 0.3772, 0.4192, 0.4388, 0.3695, 0.3541, 0.3009, 0.2288, 0.2071, 0.2211, 0.2316, 0.2218, 0.2687, 0.2841, 0.3296, 0.2316, 0.3226, 0.2785, 0.3275, 0.2512, 0.2372, 0.1659, 0.2155, 0.2407, 0.1638, 0.1589, 0.1407, 0.0868, 0.1707, 0.1106, 0.1106, 0.1106, 0.1106, 0.2736, 0.2708, 0.2876, 0.247, 0.2631, 0.2652, 0.282, 0.2666, 0.3219, 0.3352, 0.3954, 0.3135, 0.5129, 0.6039, 0.5857, 0.5129, 0.5969, 0.5969, 0.6935, 0.6655, 0.8209, 0.979, 0.8307, 0.8754, 0.8125, 0.8167, 0.7775, 0.6221, 0.6207, 0.6319, 0.8586, 0.7915, 0.8209, 0.8209, 1.0112, 1.133, 1.3793, 1.2575, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2038381961778221151", "ticker_idx": 0, "ts": "2026-03-29T22:25:05+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Literally, everything is becoming a bottleneck, isn't it?", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish broad semiconductors: price increases spreading across automotive, analog, power, networking, and CPUs — everything bottlenecked.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "Global semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "Automotive chips, industrial analog semiconductors, power switches, high-performance networking chips and connectivity components, CPUs…\n\nLiterally, everything is becoming a bottleneck, isn’t it?", "reply_text": null, "parent_text": null, "quoted_text": "Beyond Memory to Non-Memory… Semiconductor Prices Rising Across the Board\n\nPrice increases are spreading not only in memory semiconductors but also across the non-memory space. Following the AI infrastructure investment boom that has driven up DRAM and NAND flash prices, supply https://t.co/2nOSEX1wA3", "parent_author": null, "ret_m1d": 0.03232836250838944, "ret_signed_m1d": 0.03232836250838944, "alpha_spy_m1d": 0.028973683883414703, "alpha_c_m1d": 0.028973683883414703, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.0033546786249747385, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05756763629895589, "ret_signed_p1d": 0.05756763629895589, "alpha_spy_p1d": 0.028499710600041084, "alpha_c_p1d": 0.028499710600041084, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.02906792569891481, "ret_p1w": 0.09226826006030775, "ret_signed_p1w": 0.09226826006030775, "alpha_spy_p1w": 0.04960797045404575, "alpha_c_p1w": 0.04960797045404575, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.042660289606262, "ret_p1m": 0.3549499161698886, "ret_signed_p1m": 0.3549499161698886, "alpha_spy_p1m": 0.22880463168811915, "alpha_c_p1m": 0.22880463168811915, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.12614528448176943, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0066, -0.0066, 0.0297, 0.0415, 0.0692, 0.062, 0.0454, 0.0736, 0.0776, 0.08, 0.0712, 0.0935, 0.1044, 0.1044, 0.0768, 0.1087, 0.1111, 0.1036, 0.1001, 0.1234, 0.1492, 0.1517, 0.1129, 0.1253, 0.097, 0.0538, 0.0511, 0.1079, 0.1217, 0.1165, 0.1441, 0.1202, 0.1247, 0.1247, 0.1241, 0.138, 0.1315, 0.1448, 0.1389, 0.1562, 0.1755, 0.1365, 0.1209, 0.121, 0.0787, 0.1009, 0.0905, 0.0497, 0.0878, 0.096, 0.1062, 0.0706, 0.0684, 0.0866, 0.0948, 0.0859, 0.0893, 0.0613, 0.0795, 0.0884, 0.1007, 0.0505, 0.0323, 0.0, 0.0576, 0.0812, 0.0822, 0.0822, 0.0923, 0.1031, 0.1666, 0.187, 0.2051, 0.2229, 0.2468, 0.2496, 0.2545, 0.2803, 0.2798, 0.2817, 0.3153, 0.3291, 0.397, 0.3965, 0.3549, 0.378, 0.3977, 0.4063, 0.3979, 0.4418, 0.5165, 0.4898, 0.5627, 0.5897, 0.5481, 0.5791, 0.5953, 0.5346, 0.5065, 0.5005, 0.5576, 0.5664, 0.5897, 0.5897, 0.6609, 0.6426, 0.6546, 0.6521, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2010356647445819547", "ticker_idx": 0, "ts": "2026-01-11T14:22:30+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Holy shit, lol — some customers are willing to pay up to 100% more just to have TSMC make their chips.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Endorses JPM's bullish TSMC estimate revisions; impressed by pricing power with 50–100% premium expedited orders.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Holy shit, lol — some customers are willing to pay up to 100% more just to have TSMC make their chips.\n\n\"The backdrop is that ~10% of total orders are being processed on an expedited basis with substantial pricing premiums (half at ~50% premium “hot runs” and the other half at ~100% premium “super hot runs”).\"", "reply_text": null, "parent_text": null, "quoted_text": "J.P. Morgan: Upward Revision of TSMC Estimates\n\nOur analyst Gokul has recently upgraded his TSMC estimates. The key differences versus consensus are that J.P. Morgan (JPM) forecasts 2026/27 revenue growth of +31%/+25%, which exceeds consensus (~+27%/+22%), and JPM’s 2026/27 gross https://t.co/1bdZumwOMi", "parent_author": null, "ret_m1d": -0.0059171168836570365, "ret_signed_m1d": -0.0059171168836570365, "alpha_spy_m1d": -0.004349125529883535, "alpha_c_m1d": -0.002231040940618545, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01183437870055326, "ret_signed_p1d": 0.01183437870055326, "alpha_spy_p1d": 0.013833853806423546, "alpha_c_p1d": 0.00960739075955086, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.04142010805207774, "ret_signed_p1w": 0.04142010805207774, "alpha_spy_p1w": 0.046454936212638964, "alpha_c_p1w": 0.016513511791601987, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.11242609038891893, "ret_signed_p1m": 0.11242609038891893, "alpha_spy_p1m": 0.11679921035099838, "alpha_c_p1m": 0.07633330901549984, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.16057899959396948, "ret_signed_p3m": 0.16057899959396948, "alpha_spy_p3m": 0.1798453081787178, "alpha_c_p3m": 0.059084110112581634, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.136, -0.1242, -0.1449, -0.1272, -0.1272, -0.139, -0.1449, -0.1449, -0.1272, -0.1301, -0.1124, -0.1124, -0.1154, -0.1095, -0.1124, -0.139, -0.136, -0.139, -0.1301, -0.136, -0.1301, -0.139, -0.1567, -0.1478, -0.1714, -0.1773, -0.1419, -0.1832, -0.1891, -0.1655, -0.1508, -0.1537, -0.1508, -0.1685, -0.1567, -0.1449, -0.1478, -0.139, -0.1183, -0.1272, -0.1124, -0.1302, -0.1243, -0.142, -0.1509, -0.1538, -0.1538, -0.1538, -0.1331, -0.1183, -0.1154, -0.1154, -0.1065, -0.0947, -0.1006, -0.0828, -0.0828, -0.0621, -0.0118, 0.0089, -0.0089, -0.003, -0.0059, 0.0, 0.0118, 0.0118, 0.0, 0.0296, 0.0414, 0.0503, 0.0296, 0.0414, 0.0473, 0.0385, 0.0533, 0.0769, 0.068, 0.0503, 0.0444, 0.0651, 0.0562, 0.0444, 0.0533, 0.074, 0.1124, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1243, 0.1627, 0.1923, 0.1805, 0.1805, 0.1686, 0.145, 0.1036, 0.1243, 0.1183, 0.071, 0.0947, 0.1479, 0.1154, 0.1036, 0.0917, 0.1101, 0.1309, 0.0982, 0.0923, 0.0745, 0.0745, 0.0953, 0.0923, 0.0804, 0.0567, 0.0448, 0.1012, 0.0745, 0.0745, 0.0745, 0.1042, 0.1576, 0.1606, 0.1873, 0.1814, 0.2199, 0.2348, 0.2378, 0.2051, 0.2021, 0.217, 0.217, 0.2348, 0.2971, 0.3446, 0.3149, 0.2941, 0.2674, 0.2674, 0.3505, 0.3357, 0.3357, 0.3713, 0.3595, 0.3268, 0.3387, 0.3179, 0.3476, 0.3446, 0.3298, 0.309, 0.2971, 0.3238, 0.3387, 0.3713, 0.3476, 0.3654, 0.3624, 0.398, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2037869712189866029", "ticker_idx": 0, "ts": "2026-03-28T12:29:35+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "200 pods per week would imply roughly $120 billion of revenue per month, versus our current CY27 estimate of $482 billion", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Wolfe Research note suggesting Jensen's pod-production comments imply revenue far above current CY27 estimates; bullish NVDA upside scenario.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Wolfe Research’s analysis on Jensen Huang’s remarks from the Lex Fridman podcast:\n\n“Lastly, we think it is worth paying attention to CEO Jensen Huang’s comments on last week’s Lex Fridman podcast. In that podcast, Jensen said, ‘Contextually, we probably need to make about 200 of them a week.’ We generally take Jensen’s comments seriously, though not always literally. Still, we could not resist running the math: 200 pods per week would imply roughly $120 billion of revenue per month, versus our current CY27 estimate of $482 billion.”\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014227802443346205, "ret_signed_m1d": 0.014227802443346205, "alpha_spy_m1d": 0.010873123818371466, "alpha_c_m1d": -0.018100560065043236, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.03232836250838944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05588179348466227, "ret_signed_p1d": 0.05588179348466227, "alpha_spy_p1d": 0.02681386778574746, "alpha_c_p1d": -0.0016858428142936255, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.05756763629895589, "ret_p1w": 0.07549798001419172, "ret_signed_p1w": 0.07549798001419172, "alpha_spy_p1w": 0.03283769040792972, "alpha_c_p1w": -0.016770280046116026, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.09226826006030775, "ret_p1m": 0.29060967810209015, "ret_signed_p1m": 0.29060967810209015, "alpha_spy_p1m": 0.16446439362032073, "alpha_c_p1m": -0.06434023806779843, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.3549499161698886, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1291, 0.1291, 0.1433, 0.1389, 0.1336, 0.1449, 0.1202, 0.1191, 0.1196, 0.1249, 0.1087, 0.1324, 0.1274, 0.1274, 0.078, 0.1098, 0.119, 0.1362, 0.1289, 0.1413, 0.1595, 0.1655, 0.1571, 0.1237, 0.0918, 0.0546, 0.0406, 0.1225, 0.1505, 0.1414, 0.1506, 0.1317, 0.1067, 0.1067, 0.1198, 0.138, 0.1376, 0.1492, 0.1597, 0.1675, 0.1839, 0.1193, 0.0727, 0.1047, 0.09, 0.1081, 0.1099, 0.0765, 0.1058, 0.1186, 0.1263, 0.1088, 0.0913, 0.1093, 0.1015, 0.0922, 0.0811, 0.0456, 0.0634, 0.0607, 0.0818, 0.0368, 0.0142, 0.0, 0.0559, 0.0641, 0.074, 0.074, 0.0755, 0.0783, 0.1024, 0.1135, 0.142, 0.1462, 0.1897, 0.204, 0.2009, 0.221, 0.2233, 0.2101, 0.226, 0.2087, 0.2609, 0.3114, 0.2906, 0.2669, 0.2083, 0.2015, 0.2017, 0.1897, 0.2583, 0.2805, 0.3029, 0.3286, 0.3367, 0.3673, 0.4273, 0.3642, 0.346, 0.3357, 0.353, 0.329, 0.3037, 0.3037, 0.3008, 0.2872, 0.2971, 0.2783, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011380454352728283", "ticker_idx": 0, "ts": "2026-01-14T10:10:44+00:00", "symbol": "DB HiTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DB HiTek Positioned as a Key Beneficiary... The retreat of giants like Samsung and TSMC from the 8-inch space could provide a significant tailwind for DB HiTek.", "tweet_type": "analysis", "timeline": "months", "summary": "DB HiTek is key beneficiary of Samsung/TSMC 8-inch retreat; already at full capacity with backlog as customers migrate toward it.", "resolved_tickers": ["000990.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "DB HiTek 000990.KS Korea KOSPI 8-inch foundry", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung to Shutter Parts of 8-Inch Foundry This Year to Focus on 12-Inch Operations\n\nSamsung Electronics is set to close a portion of its 8-inch foundry fabs within the year. The move is aimed at phasing out low-profit legacy processes and concentrating resources on advanced 12-inch (300mm) production. As TSMC is also scaling back its 8-inch footprint, the global 8-inch foundry market appears to be in a period of contraction.\n\nAccording to industry sources on the 14th, Samsung Electronics plans to shut down its 'S7' 8-inch line in Giheung during the second half of this year. Samsung currently operates three 8-inch fabs in Giheung (S6, S7, and S8), and has decided to retire one while maintaining the remaining two.\n\nThe closure of S7 will reduce Samsung's total 8-inch production capacity from approximately 250,000 wafers per month to below 200,000. S7's monthly capacity is estimated to be around 50,000 wafers. It is reported that the process of migrating existing customers to other lines is already underway. The future utility of the S7 facility has not yet been determined.\n\nCurrently, the utilization rate of Samsung's 8-inch fabs is known to hover around 70%. As flagship products such as CMOS Image Sensors (CIS) and Display Driver ICs (DDI) migrate to 12-inch wafers, 8-inch utilization has declined. Furthermore, the need to reassign development personnel to 12-inch advanced processes has increased the burden of maintaining underutilized 8-inch lines.\n\nAn industry insider commented, \"Unlike specialized pure-play foundries, Samsung's foundry model focuses on high-volume production for a smaller number of large-scale customers. From their perspective, there is little incentive to maintain three separate fabs while utilization remains low.\"\n\nA Global Shift in the 8-Inch Market\nThis reduction is not unique to Samsung. Market research firm TrendForce predicts that global 8-inch fab capacity will decrease by 2.4% year-on-year. TSMC began reducing its 8-inch capacity last year and plans to fully decommission some fabs by 2027. TrendForce estimates that Samsung also began its 8-inch scale-back initiatives last year.\n\nInterestingly, while supply is being consolidated, demand for certain segments remains resilient. Driven by the surge in power semiconductors (PMICs) for AI servers, the average global 8-inch utilization rate is projected to rise from 75–80% last year to 85–90% this year. Consequently, some foundry players are considering price hikes of 5% to 20%.\n\nDB HiTek Positioned as a Key Beneficiary\nThe retreat of giants like Samsung and TSMC from the 8-inch space could provide a significant tailwind for DB HiTek.\n\nDB HiTek is currently operating at full capacity, with a substantial order backlog making it difficult to accept new contracts. The company specializes in analog processes, such as BCD (Bipolar-CMOS-DMOS), and maintains a high-variety, low-volume production system tailored for PMICs and DDIs. As Samsung and TSMC reduce their 8-inch footprint, customers are expected to migrate toward DB HiTek.\n\nHowever, competition remains intense. Players like Taiwan's Vanguard (VIS) and UMC, China's SMIC, and Israel's Tower Semiconductor are all vying for market share. In particular, SMIC is aggressively expanding its volume based on price competitiveness. Meanwhile, Key Foundry is maintaining a utilization rate of around 90%, though its profitability is reported to be relatively low.\n\n---\n\nhttps://t.co/4UhEznYTQp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06715506844789665, "ret_signed_m1d": -0.06715506844789665, "alpha_spy_m1d": -0.0720945846841693, "alpha_c_m1d": -0.07534353922375037, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0027232143965294497, "alpha_c_p1d": -0.02075446744410603, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": 0.10256408528719874, "ret_signed_p1w": 0.10256408528719874, "alpha_spy_p1w": 0.10974871528713137, "alpha_c_p1w": 0.06759566412372431, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": 0.15750914129002336, "ret_signed_p1m": 0.15750914129002336, "alpha_spy_p1m": 0.17067610186786464, "alpha_c_p1m": 0.11177725260991123, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.18896690307715458, "ret_signed_p3m": 0.18896690307715458, "alpha_spy_p3m": 0.19242344742641826, "alpha_c_p3m": 0.047367857447894846, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2784, -0.2723, -0.3065, -0.3065, -0.2943, -0.2637, -0.2576, -0.2747, -0.2759, -0.2808, -0.2821, -0.2808, -0.2796, -0.3162, -0.3211, -0.348, -0.2454, -0.1563, -0.127, -0.1758, -0.1746, -0.1404, -0.1758, -0.1978, -0.1893, -0.2442, -0.2576, -0.2442, -0.2442, -0.2234, -0.2234, -0.2125, -0.2063, -0.199, -0.2137, -0.2088, -0.1929, -0.2076, -0.1917, -0.188, -0.1746, -0.2015, -0.2027, -0.2002, -0.2186, -0.2027, -0.1661, -0.1526, -0.1636, -0.1636, -0.1783, -0.1807, -0.1746, -0.1746, -0.1746, -0.1001, -0.0818, -0.0818, -0.094, -0.1245, -0.1221, -0.0965, -0.0672, 0.0, 0.0, 0.0403, 0.0464, 0.0342, 0.1026, 0.1368, 0.1111, 0.1074, 0.1673, 0.2784, 0.2344, 0.2918, 0.2576, 0.3785, 0.3529, 0.2747, 0.2576, 0.2088, 0.1819, 0.1453, 0.1575, 0.1355, 0.1355, 0.1355, 0.1355, 0.116, 0.1355, 0.1355, 0.116, 0.1099, 0.1477, 0.1685, 0.1685, 0.105, -0.0562, 0.16, 0.0989, 0.0251, 0.0805, 0.0473, 0.0584, 0.0658, 0.083, 0.0559, 0.083, 0.083, 0.0855, 0.0288, 0.0276, 0.0152, -0.0402, -0.0402, -0.0045, -0.0328, 0.0596, 0.0263, 0.1163, 0.1286, 0.0818, 0.1631, 0.1409, 0.168, 0.189, 0.1914, 0.2148, 0.2752, 0.3171, 0.5611, 0.5031, 0.5758, 0.6929, 0.8728, 0.959, 0.906, 0.8802, 0.9529, 0.9529, 0.9923, 0.9923, 0.996, 0.9405, 1.0711, 1.3003, 1.0428, 1.2868, 1.2523, 1.0145, 1.1056, 0.9381, 0.996, 1.2301, 1.2313, 1.2313, 1.7168, 1.4888, 1.394, 1.2855, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2025539507622490278", "ticker_idx": 0, "ts": "2026-02-22T11:53:45+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' non-memory division (Foundry + System LSI) is expected to achieve an earnings turnaround as early as Q4. In particular, starting next year—when production of Tesla's AI5 chip, Apple's image sensors, and other products ramps—overall operating profit is likely to swing to positive.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung Foundry utilization surpassed 80% in Q1; non-memory division expected to hit breakeven in Q4 with full profitability ramp in 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Foundry Revival… Q1 Utilization Surpasses 80%, Breakeven as Early as Q4 This Year\n\nSamsung Electronics' (005930) semiconductor foundry utilization rates have surpassed 80%, led by leading-edge nodes, with earnings improving rapidly. As chips developed in-house using Samsung's foundry—including 6th-generation HBM4 and Exynos 2600—have proven their performance, orders from global big tech companies are flooding in, making quarterly profitability within the year highly likely.\n\nAccording to industry sources on the 22nd, utilization at Samsung Electronics' Pyeongtaek Campus P2 and P3 foundry production lines climbed to the 80% range in Q1. These lines, which produce semiconductors on mature nodes including 4nm, 5nm, and 7nm, struggled with insufficient production volume in H2 2024, with utilization falling well below 50% last year.\n\nHowever, since H2 last year, Samsung's foundry division has applied its 4nm process to HBM4 base dies for the memory division, achieving the industry's best-in-class performance and driving a surge in order volume. On top of this, big tech companies from both the U.S. and China are placing successive chip production requests, with Samsung's semiconductor business spreading both wings wide. A source well-versed in Samsung's semiconductor production noted, \"With orders from global big tech continuing to grow, centered on mature nodes, foundry utilization will comfortably exceed 80% in the first half of this year.\"\n\nProduction volume on Samsung Foundry's cutting-edge 2nm process is also increasing, brightening the earnings outlook. Samsung's in-house mobile application processor (AP), Exynos 2600, is being produced on the 2nm node, and with its performance reportedly surpassing Qualcomm's chips, supply volume for the Galaxy S26 is expected to increase.\n\nAccordingly, Samsung Electronics' non-memory division (Foundry + System LSI) is expected to achieve an earnings turnaround as early as Q4. In particular, starting next year—when production of Tesla's AI5 chip, Apple's image sensors, and other products ramps—overall operating profit is likely to swing to positive.\n\nAn industry source noted, \"With the foundry business recovering, Samsung Electronics' competitiveness as an integrated semiconductor company will be significantly strengthened.\"\n\n---\n\nhttps://t.co/yu5F3wfzY4", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015025855706159885, "ret_signed_m1d": -0.015025855706159885, "alpha_spy_m1d": -0.025342517093223438, "alpha_c_m1d": -0.032063117250343054, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.01703726154418317, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03626946073682302, "ret_signed_p1d": 0.03626946073682302, "alpha_spy_p1d": 0.029001007806428003, "alpha_c_p1d": 0.023274982215576, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.012994478521247022, "ret_p1w": 0.12176170889254245, "ret_signed_p1w": 0.12176170889254245, "alpha_spy_p1w": 0.11591462317965084, "alpha_c_p1w": 0.11439822988666881, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": 0.007363479005873641, "ret_p1m": -0.01709838473024672, "ret_signed_p1m": -0.01709838473024672, "alpha_spy_p1m": 0.023092997594216014, "alpha_c_p1m": -0.0012109257509007598, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.01588745897934596, "ret_p3m": 0.5550254955061809, "ret_signed_p3m": 0.5550254955061809, "alpha_spy_p3m": 0.4636432786683964, "alpha_c_p3m": 0.264077910400387, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.2909475851057939, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4699, -0.4663, -0.4818, -0.4802, -0.4668, -0.4612, -0.4581, -0.4411, -0.4354, -0.4411, -0.4431, -0.4467, -0.4385, -0.4596, -0.4699, -0.4436, -0.4452, -0.4519, -0.4302, -0.4251, -0.4271, -0.4271, -0.3967, -0.3808, -0.3788, -0.3788, -0.3788, -0.3342, -0.2845, -0.2803, -0.2694, -0.2808, -0.2798, -0.2808, -0.287, -0.2731, -0.2544, -0.2285, -0.2264, -0.2477, -0.2254, -0.2109, -0.2119, -0.2119, -0.1736, -0.1585, -0.1674, -0.1684, -0.2207, -0.1321, -0.1238, -0.1746, -0.1782, -0.1378, -0.1409, -0.1306, -0.0746, -0.0611, -0.0611, -0.0611, -0.0611, -0.0155, -0.015, 0.0, 0.0363, 0.0544, 0.1295, 0.1218, 0.1218, 0.0109, -0.1078, -0.0073, -0.0249, -0.101, -0.0264, -0.0155, -0.0264, -0.0492, -0.0223, 0.0047, 0.0803, 0.0389, 0.0332, -0.0347, -0.0171, -0.0207, -0.0668, -0.0668, -0.0846, -0.1319, -0.0153, -0.0737, -0.0332, 0.0026, 0.0202, 0.0929, 0.0592, 0.0696, 0.0436, 0.0722, 0.0955, 0.1293, 0.1215, 0.1137, 0.1371, 0.1293, 0.1656, 0.1397, 0.1656, 0.1526, 0.1734, 0.1449, 0.1449, 0.2072, 0.2072, 0.3811, 0.4096, 0.3941, 0.4823, 0.4486, 0.4745, 0.5369, 0.4045, 0.459, 0.4304, 0.433, 0.555, 0.5187, 0.5187, 0.5524, 0.594, 0.555, 0.6459, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2017492440685715682", "ticker_idx": 0, "ts": "2026-01-31T06:57:35+00:00", "symbol": "EMC", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "EMC and TUC are projected to maintain their positions as dominant suppliers", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan CCL suppliers — especially EMC and TUC — set to lead AI ASIC server build-out as M8-grade spec upgrades accelerate and margin improves into 2026.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "EMC = Elite Material Co. Taiwan CCL maker 2383.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ASIC server demand boosts Taiwan's high-end CCL shipments\n\n- Cloud Service Providers (CSPs) continue to maintain robust capital expenditures (CAPEX), driving steady growth in the AI server and high-speed network switch markets.\n\n- Against this backdrop, the advancement of CCL (Copper Clad Laminate) material specifications is accelerating, with a gradual increase in orders for high-value-added products.\n\n- Taiwan’s top three CCL manufacturers achieved significant revenue growth in 2025, with all three recording double-digit year-over-year increases.\n\n- Looking ahead to 2026, the penetration of Taiwanese firms into the high-end market is expected to expand further as AI GPU and ASIC server customers begin upgrading to M8-grade or higher CCL materials in earnest.\n\n- Consequently, Taiwanese CCL suppliers are positioned to be key beneficiaries of the expanding investment in AI infrastructure.\n\n- At the same time, price hikes in glass fiber fabrics and copper foils are pushing up prices for low-to-mid-range product lines, which is expected to further improve profit margins.\n\n- However, some concerns have been raised that the benefits for Taiwanese suppliers might fall short of initial expectations due to design change rumors suggesting a potential downgrade in CCL specs for Nvidia’s next-generation products, alongside pressure from competitors such as South Korea's Doosan and China's Sytech.\n\n- Nevertheless, when focusing specifically on the AI ASIC market, EMC and TUC are projected to maintain their positions as dominant suppliers.\n\n- EMC successfully entered the AI server and 800G switch markets, recording a 46.39% YoY revenue increase in 2025—the highest growth rate among the top three.\n\n- While shipment momentum slowed slightly at year-end as some CSP customers entered a product transition phase in Q4, overall performance remained in line with company guidance.\n\n- TUC is also benefiting from increased ASIC server shipments; with the revenue share of high-spec CCL (M7-grade and above) expanding, its 2025 revenue rose 31.53% YoY.\n\n- Iteq is targeting both AI and non-AI server markets simultaneously, achieving a 12.66% revenue increase in 2025 compared to 2024.\n\n- The company identified M7-grade CCL materials for PCIe Gen6 server platforms as the key driver for expanding shipments in 2026.\n\n---\n\nhttps://t.co/A8umkTBnRm", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022408963585434205, "ret_signed_m1d": -0.022408963585434205, "alpha_spy_m1d": -0.017462193639279477, "alpha_c_m1d": -0.012908900638025078, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03921568627450989, "ret_signed_p1d": 0.03921568627450989, "alpha_spy_p1d": 0.04767108270521048, "alpha_c_p1d": 0.0611073963268588, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": 0.1652661064425771, "ret_signed_p1w": 0.1652661064425771, "alpha_spy_p1w": 0.16736554419281768, "alpha_c_p1w": 0.17841480526125886, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": 0.2997198879551821, "ret_signed_p1m": 0.2997198879551821, "alpha_spy_p1m": 0.3214048665066819, "alpha_c_p1m": 0.35314125580816647, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 1.5462184873949578, "ret_signed_p3m": 1.5462184873949578, "alpha_spy_p3m": 1.5099626803879842, "alpha_c_p3m": 1.4468220568128032, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1933, -0.1877, -0.2325, -0.2381, -0.2605, -0.2465, -0.2297, -0.2269, -0.1877, -0.2437, -0.2549, -0.1989, -0.2661, -0.2605, -0.1877, -0.1765, -0.1625, -0.1457, -0.1737, -0.1933, -0.1989, -0.2017, -0.1625, -0.1681, -0.1597, -0.1036, -0.1064, -0.1008, -0.1008, -0.1625, -0.1625, -0.1597, -0.1317, -0.1036, -0.0952, -0.0532, -0.0532, -0.0812, -0.0616, -0.0784, -0.0784, -0.0784, -0.1092, -0.0924, -0.0504, -0.1148, -0.1345, -0.1232, -0.098, -0.084, -0.0644, -0.1204, -0.0896, -0.1148, -0.0952, -0.1261, -0.07, 0.0084, -0.0252, 0.0196, 0.0644, 0.0028, -0.0224, 0.0, 0.0392, 0.1429, 0.0924, 0.0784, 0.1653, 0.1793, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2185, 0.2941, 0.3557, 0.3669, 0.3669, 0.3333, 0.2997, 0.2269, 0.3249, 0.3221, 0.2185, 0.3305, 0.3725, 0.4566, 0.5154, 0.4482, 0.5126, 0.5126, 0.619, 0.5742, 0.5658, 0.5042, 0.5938, 0.6499, 0.6303, 0.5602, 0.4566, 0.6022, 0.5602, 0.5602, 0.5602, 0.6162, 0.7759, 0.8039, 0.916, 0.8627, 1.0196, 1.1345, 1.1485, 1.1289, 1.1569, 1.2129, 1.2773, 1.2969, 1.507, 1.6499, 1.451, 1.4874, 1.5462, 1.5462, 1.6639, 1.6499, 1.7731, 1.7787, 1.6611, 1.7171, 1.7451, 1.7619, 1.7563, 1.5854, 1.5462, 1.5406, 1.4622, 1.6275, 1.8039, 2.084, 1.9636, 1.972, 1.8319, 1.8683, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2017492440685715682", "ticker_idx": 2, "ts": "2026-01-31T06:57:35+00:00", "symbol": "Taiwanese CCL suppliers", "kind": "sector", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Taiwanese CCL suppliers are positioned to be key beneficiaries of the expanding investment in AI infrastructure", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwan CCL suppliers — especially EMC and TUC — set to lead AI ASIC server build-out as M8-grade spec upgrades accelerate and margin improves into 2026.", "resolved_tickers": ["2383.TW"], "resolution_method": "basket", "resolution_reason": "Taiwan CCL sector; EMC 2383.TW confirmed; TUC/Lianmao tickers rejected", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ASIC server demand boosts Taiwan's high-end CCL shipments\n\n- Cloud Service Providers (CSPs) continue to maintain robust capital expenditures (CAPEX), driving steady growth in the AI server and high-speed network switch markets.\n\n- Against this backdrop, the advancement of CCL (Copper Clad Laminate) material specifications is accelerating, with a gradual increase in orders for high-value-added products.\n\n- Taiwan’s top three CCL manufacturers achieved significant revenue growth in 2025, with all three recording double-digit year-over-year increases.\n\n- Looking ahead to 2026, the penetration of Taiwanese firms into the high-end market is expected to expand further as AI GPU and ASIC server customers begin upgrading to M8-grade or higher CCL materials in earnest.\n\n- Consequently, Taiwanese CCL suppliers are positioned to be key beneficiaries of the expanding investment in AI infrastructure.\n\n- At the same time, price hikes in glass fiber fabrics and copper foils are pushing up prices for low-to-mid-range product lines, which is expected to further improve profit margins.\n\n- However, some concerns have been raised that the benefits for Taiwanese suppliers might fall short of initial expectations due to design change rumors suggesting a potential downgrade in CCL specs for Nvidia’s next-generation products, alongside pressure from competitors such as South Korea's Doosan and China's Sytech.\n\n- Nevertheless, when focusing specifically on the AI ASIC market, EMC and TUC are projected to maintain their positions as dominant suppliers.\n\n- EMC successfully entered the AI server and 800G switch markets, recording a 46.39% YoY revenue increase in 2025—the highest growth rate among the top three.\n\n- While shipment momentum slowed slightly at year-end as some CSP customers entered a product transition phase in Q4, overall performance remained in line with company guidance.\n\n- TUC is also benefiting from increased ASIC server shipments; with the revenue share of high-spec CCL (M7-grade and above) expanding, its 2025 revenue rose 31.53% YoY.\n\n- Iteq is targeting both AI and non-AI server markets simultaneously, achieving a 12.66% revenue increase in 2025 compared to 2024.\n\n- The company identified M7-grade CCL materials for PCIe Gen6 server platforms as the key driver for expanding shipments in 2026.\n\n---\n\nhttps://t.co/A8umkTBnRm", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022408963585434205, "ret_signed_m1d": -0.022408963585434205, "alpha_spy_m1d": -0.017462193639279477, "alpha_c_m1d": -0.012908900638025078, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03921568627450989, "ret_signed_p1d": 0.03921568627450989, "alpha_spy_p1d": 0.04767108270521048, "alpha_c_p1d": 0.0611073963268588, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": 0.1652661064425771, "ret_signed_p1w": 0.1652661064425771, "alpha_spy_p1w": 0.16736554419281768, "alpha_c_p1w": 0.17841480526125886, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": 0.2997198879551821, "ret_signed_p1m": 0.2997198879551821, "alpha_spy_p1m": 0.3214048665066819, "alpha_c_p1m": 0.35314125580816647, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 1.5462184873949578, "ret_signed_p3m": 1.5462184873949578, "alpha_spy_p3m": 1.5099626803879842, "alpha_c_p3m": 1.4468220568128032, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1933, -0.1877, -0.2325, -0.2381, -0.2605, -0.2465, -0.2297, -0.2269, -0.1877, -0.2437, -0.2549, -0.1989, -0.2661, -0.2605, -0.1877, -0.1765, -0.1625, -0.1457, -0.1737, -0.1933, -0.1989, -0.2017, -0.1625, -0.1681, -0.1597, -0.1036, -0.1064, -0.1008, -0.1008, -0.1625, -0.1625, -0.1597, -0.1317, -0.1036, -0.0952, -0.0532, -0.0532, -0.0812, -0.0616, -0.0784, -0.0784, -0.0784, -0.1092, -0.0924, -0.0504, -0.1148, -0.1345, -0.1232, -0.098, -0.084, -0.0644, -0.1204, -0.0896, -0.1148, -0.0952, -0.1261, -0.07, 0.0084, -0.0252, 0.0196, 0.0644, 0.0028, -0.0224, 0.0, 0.0392, 0.1429, 0.0924, 0.0784, 0.1653, 0.1793, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2297, 0.2185, 0.2941, 0.3557, 0.3669, 0.3669, 0.3333, 0.2997, 0.2269, 0.3249, 0.3221, 0.2185, 0.3305, 0.3725, 0.4566, 0.5154, 0.4482, 0.5126, 0.5126, 0.619, 0.5742, 0.5658, 0.5042, 0.5938, 0.6499, 0.6303, 0.5602, 0.4566, 0.6022, 0.5602, 0.5602, 0.5602, 0.6162, 0.7759, 0.8039, 0.916, 0.8627, 1.0196, 1.1345, 1.1485, 1.1289, 1.1569, 1.2129, 1.2773, 1.2969, 1.507, 1.6499, 1.451, 1.4874, 1.5462, 1.5462, 1.6639, 1.6499, 1.7731, 1.7787, 1.6611, 1.7171, 1.7451, 1.7619, 1.7563, 1.5854, 1.5462, 1.5406, 1.4622, 1.6275, 1.8039, 2.084, 1.9636, 1.972, 1.8319, 1.8683, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025539056432828670", "ticker_idx": 0, "ts": "2026-02-22T11:51:58+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix at KRW 1,250,000", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares CLSA Buy calls on SK Hynix (PT KRW 1.25M), Samsung (PT KRW 260K), and Micron (PT $495) on HBM4 ramp and memory mix shift driving re-rating.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CLSA estimates that Samsung Electronics began commercial shipments of HBM4 to NVIDIA earlier this month, making a preemptive move ahead of competitors. However, insufficient yields on the 1c nm process are expected to limit meaningful market share gains in the near term.\n\nCLSA projects the following HBM4 market share breakdown going forward: SK Hynix at 55%, Samsung Electronics at 25–30%, and Micron at 20%.\n\nThe firm notes that rising commodity DRAM prices are improving profitability, which in turn provides positive support for HBM pricing as players compete over limited production capacity. CLSA assesses that the ongoing shift in revenue mix toward high-value and customized products will continue to drive valuation re-rating for memory stocks.\n\nTarget prices: SK Hynix at KRW 1,250,000; Samsung Electronics at KRW 260,000; Micron at USD 495.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0021031335550720787, "ret_signed_m1d": -0.0021031335550720787, "alpha_spy_m1d": -0.012419794942135631, "alpha_c_m1d": -0.007310442872690004, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05678230114696747, "ret_signed_p1d": 0.05678230114696747, "alpha_spy_p1d": 0.049513848216572454, "alpha_c_p1d": 0.04157207370793392, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": 0.11772635306618784, "ret_signed_p1w": 0.11772635306618784, "alpha_spy_p1w": 0.11187926735329623, "alpha_c_p1w": 0.13344518216331758, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": 0.03871643856653173, "ret_signed_p1m": 0.03871643856653173, "alpha_spy_p1m": 0.07890782089099446, "alpha_c_p1m": 0.0830150442839801, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 1.0437220241815908, "ret_signed_p3m": 1.0437220241815908, "alpha_spy_p3m": 0.9523398073438063, "alpha_c_p3m": 0.6683102867108126, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4494, -0.428, -0.4427, -0.4343, -0.4132, -0.4196, -0.4301, -0.428, -0.3933, -0.4048, -0.3828, -0.4059, -0.3996, -0.4175, -0.4427, -0.4206, -0.4196, -0.4248, -0.3901, -0.3859, -0.3817, -0.3817, -0.3701, -0.327, -0.3155, -0.3155, -0.3155, -0.2881, -0.2681, -0.2366, -0.2198, -0.205, -0.2177, -0.2124, -0.224, -0.2198, -0.2124, -0.205, -0.1966, -0.2187, -0.2219, -0.2061, -0.1935, -0.2261, -0.1588, -0.1157, -0.0946, -0.0442, -0.1272, -0.0463, -0.0536, -0.1146, -0.1178, -0.0673, -0.0789, -0.0957, -0.0662, -0.0747, -0.0747, -0.0747, -0.0747, -0.0599, -0.0021, 0.0, 0.0568, 0.0705, 0.1578, 0.1177, 0.1177, -0.0108, -0.1056, -0.0087, -0.0266, -0.1193, -0.0118, 0.0061, -0.0203, -0.0413, 0.0261, 0.0219, 0.1125, 0.0672, 0.0608, -0.0171, 0.0387, 0.0482, -0.0171, -0.0171, -0.0803, -0.1499, -0.055, -0.1256, -0.0772, -0.0666, -0.035, 0.0882, 0.0514, 0.0819, 0.0956, 0.162, 0.1967, 0.2168, 0.1883, 0.2283, 0.2894, 0.2884, 0.2905, 0.2873, 0.3611, 0.3695, 0.3621, 0.3548, 0.3548, 0.5244, 0.5244, 0.6866, 0.7424, 0.7761, 0.9805, 0.9331, 1.0816, 1.0753, 0.9163, 0.9384, 0.8383, 0.8383, 1.0437, 1.0448, 1.0448, 1.1617, 1.3629, 1.4118, 1.4581, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025539056432828670", "ticker_idx": 1, "ts": "2026-02-22T11:51:58+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics at KRW 260,000", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares CLSA Buy calls on SK Hynix (PT KRW 1.25M), Samsung (PT KRW 260K), and Micron (PT $495) on HBM4 ramp and memory mix shift driving re-rating.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "CLSA estimates that Samsung Electronics began commercial shipments of HBM4 to NVIDIA earlier this month, making a preemptive move ahead of competitors. However, insufficient yields on the 1c nm process are expected to limit meaningful market share gains in the near term.\n\nCLSA projects the following HBM4 market share breakdown going forward: SK Hynix at 55%, Samsung Electronics at 25–30%, and Micron at 20%.\n\nThe firm notes that rising commodity DRAM prices are improving profitability, which in turn provides positive support for HBM pricing as players compete over limited production capacity. CLSA assesses that the ongoing shift in revenue mix toward high-value and customized products will continue to drive valuation re-rating for memory stocks.\n\nTarget prices: SK Hynix at KRW 1,250,000; Samsung Electronics at KRW 260,000; Micron at USD 495.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015025855706159885, "ret_signed_m1d": -0.015025855706159885, "alpha_spy_m1d": -0.025342517093223438, "alpha_c_m1d": -0.032063117250343054, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.01703726154418317, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03626946073682302, "ret_signed_p1d": 0.03626946073682302, "alpha_spy_p1d": 0.029001007806428003, "alpha_c_p1d": 0.023274982215576, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.012994478521247022, "ret_p1w": 0.12176170889254245, "ret_signed_p1w": 0.12176170889254245, "alpha_spy_p1w": 0.11591462317965084, "alpha_c_p1w": 0.11439822988666881, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": 0.007363479005873641, "ret_p1m": -0.01709838473024672, "ret_signed_p1m": -0.01709838473024672, "alpha_spy_p1m": 0.023092997594216014, "alpha_c_p1m": -0.0012109257509007598, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.01588745897934596, "ret_p3m": 0.5550254955061809, "ret_signed_p3m": 0.5550254955061809, "alpha_spy_p3m": 0.4636432786683964, "alpha_c_p3m": 0.264077910400387, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.2909475851057939, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4699, -0.4663, -0.4818, -0.4802, -0.4668, -0.4612, -0.4581, -0.4411, -0.4354, -0.4411, -0.4431, -0.4467, -0.4385, -0.4596, -0.4699, -0.4436, -0.4452, -0.4519, -0.4302, -0.4251, -0.4271, -0.4271, -0.3967, -0.3808, -0.3788, -0.3788, -0.3788, -0.3342, -0.2845, -0.2803, -0.2694, -0.2808, -0.2798, -0.2808, -0.287, -0.2731, -0.2544, -0.2285, -0.2264, -0.2477, -0.2254, -0.2109, -0.2119, -0.2119, -0.1736, -0.1585, -0.1674, -0.1684, -0.2207, -0.1321, -0.1238, -0.1746, -0.1782, -0.1378, -0.1409, -0.1306, -0.0746, -0.0611, -0.0611, -0.0611, -0.0611, -0.0155, -0.015, 0.0, 0.0363, 0.0544, 0.1295, 0.1218, 0.1218, 0.0109, -0.1078, -0.0073, -0.0249, -0.101, -0.0264, -0.0155, -0.0264, -0.0492, -0.0223, 0.0047, 0.0803, 0.0389, 0.0332, -0.0347, -0.0171, -0.0207, -0.0668, -0.0668, -0.0846, -0.1319, -0.0153, -0.0737, -0.0332, 0.0026, 0.0202, 0.0929, 0.0592, 0.0696, 0.0436, 0.0722, 0.0955, 0.1293, 0.1215, 0.1137, 0.1371, 0.1293, 0.1656, 0.1397, 0.1656, 0.1526, 0.1734, 0.1449, 0.1449, 0.2072, 0.2072, 0.3811, 0.4096, 0.3941, 0.4823, 0.4486, 0.4745, 0.5369, 0.4045, 0.459, 0.4304, 0.433, 0.555, 0.5187, 0.5187, 0.5524, 0.594, 0.555, 0.6459, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025539056432828670", "ticker_idx": 2, "ts": "2026-02-22T11:51:58+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron at USD 495", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares CLSA Buy calls on SK Hynix (PT KRW 1.25M), Samsung (PT KRW 260K), and Micron (PT $495) on HBM4 ramp and memory mix shift driving re-rating.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "CLSA estimates that Samsung Electronics began commercial shipments of HBM4 to NVIDIA earlier this month, making a preemptive move ahead of competitors. However, insufficient yields on the 1c nm process are expected to limit meaningful market share gains in the near term.\n\nCLSA projects the following HBM4 market share breakdown going forward: SK Hynix at 55%, Samsung Electronics at 25–30%, and Micron at 20%.\n\nThe firm notes that rising commodity DRAM prices are improving profitability, which in turn provides positive support for HBM pricing as players compete over limited production capacity. CLSA assesses that the ongoing shift in revenue mix toward high-value and customized products will continue to drive valuation re-rating for memory stocks.\n\nTarget prices: SK Hynix at KRW 1,250,000; Samsung Electronics at KRW 260,000; Micron at USD 495.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017103317512722427, "ret_signed_m1d": 0.017103317512722427, "alpha_spy_m1d": 0.006786656125658874, "alpha_c_m1d": 0.011896008195104502, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007031412215625532, "ret_signed_p1d": -0.007031412215625532, "alpha_spy_p1d": -0.014299865146020552, "alpha_c_p1d": -0.022241639654659084, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": -0.019716354573246386, "ret_signed_p1w": -0.019716354573246386, "alpha_spy_p1w": -0.025563440286137995, "alpha_c_p1w": -0.003997525476116648, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": -0.06043186692613789, "ret_signed_p1m": -0.06043186692613789, "alpha_spy_p1m": -0.020240484601675157, "alpha_c_p1m": -0.01613326120868952, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 0.811103166517908, "ret_signed_p3m": 0.811103166517908, "alpha_spy_p3m": 0.7197209496801236, "alpha_c_p3m": 0.4356914290471299, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4532, -0.4532, -0.4385, -0.429, -0.4313, -0.444, -0.4618, -0.4367, -0.4137, -0.4006, -0.3738, -0.3863, -0.4274, -0.4361, -0.4479, -0.4645, -0.4098, -0.3686, -0.3432, -0.344, -0.3193, -0.3193, -0.3238, -0.3007, -0.3049, -0.322, -0.322, -0.2507, -0.2585, -0.1842, -0.1934, -0.2232, -0.1803, -0.1784, -0.1968, -0.2081, -0.2003, -0.1383, -0.1383, -0.133, -0.0757, -0.0556, -0.0506, -0.0757, -0.0255, 0.034, 0.0352, -0.0145, 0.04, -0.0036, -0.0987, -0.0905, -0.0624, -0.089, -0.1134, -0.0253, -0.0166, -0.0221, -0.0221, -0.0503, -0.0, -0.0086, 0.0171, 0.0, -0.007, 0.0191, -0.0129, -0.0204, -0.0197, -0.0981, -0.048, -0.0568, -0.1204, -0.0752, -0.0424, -0.0054, -0.0371, 0.0123, 0.0495, 0.0967, 0.0968, 0.0553, 0.0046, -0.0395, -0.0604, -0.0924, -0.1556, -0.1514, -0.2353, -0.1971, -0.1258, -0.1296, -0.1296, -0.1023, -0.1027, -0.0334, 0.0017, -0.0005, 0.0137, 0.1066, 0.0842, 0.0866, 0.0815, 0.0657, 0.0679, 0.1585, 0.1448, 0.1804, 0.2466, 0.1984, 0.2321, 0.229, 0.2885, 0.3699, 0.5214, 0.5841, 0.5367, 0.7748, 0.8901, 0.8217, 0.9098, 0.8442, 0.7221, 0.6197, 0.6605, 0.7395, 0.8111, 0.7847, 0.7847, 1.129, 1.2063, 1.1947, 1.3075, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030657806920229226", "ticker_idx": 0, "ts": "2026-03-08T14:52:03+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CPUs and memory are emerging as the primary bottlenecks constraining compute performance in data centers", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares HSBC thesis that agentic AI lifts CPU demand 3-4x vs conventional AI; names INTC and AMD as plays on the trend through 2028.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> 'Agentic AI' Rise: The CPU's Comeback?\n\n• According to HSBC, approximately 44% of tasks in agentic AI workflows rely on the CPU — roughly 3–4x higher than in conventional AI compute architectures. This signals that CPUs and memory are emerging as the primary bottlenecks constraining compute performance in data centers.\n\n• HSBC has accordingly raised its 2026 global server shipment forecast to +20% YoY. The firm also notes that given supply chain capacity constraints, actual market demand could surge by as much as 60%, with growth tailwinds expected to extend through 2028.\n\n$INTC $AMD\n\n---\n\nhttps://t.co/Mfg4EbLv0r", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.047389284232931206, "ret_signed_m1d": -0.047389284232931206, "alpha_spy_m1d": -0.03870533086857597, "alpha_c_m1d": -0.012371413376680906, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.0350178708562503, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026327268539612092, "ret_signed_p1d": 0.026327268539612092, "alpha_spy_p1d": 0.02793439544497922, "alpha_c_p1d": 0.018821648204072616, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.007505620335539476, "ret_p1w": 0.003949023327065371, "ret_signed_p1w": 0.003949023327065371, "alpha_spy_p1w": 0.017571930051264117, "alpha_c_p1w": 0.005090037336813791, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.00114101400974842, "ret_p1m": 0.1608160974527062, "ret_signed_p1m": 0.1608160974527062, "alpha_spy_p1m": 0.18624806187231868, "alpha_c_p1m": 0.14679373559007614, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": 0.014022361862630062, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1053, -0.1332, -0.1705, -0.1771, -0.1814, -0.2091, -0.204, -0.1922, -0.2021, -0.2025, -0.2067, -0.2067, -0.2058, -0.1953, -0.1817, -0.1904, -0.1904, -0.136, -0.1362, -0.1215, -0.0647, -0.0981, -0.0007, -0.0333, 0.0375, 0.0689, 0.0601, 0.0303, 0.0303, 0.0654, 0.1902, 0.1918, -0.0112, -0.0678, -0.0362, 0.0702, 0.0676, 0.0195, 0.0709, 0.0805, 0.0663, 0.0584, 0.1099, 0.1022, 0.034, 0.0595, 0.0197, 0.0265, 0.0265, 0.0132, -0.0026, -0.0211, -0.0323, -0.0428, 0.0118, 0.0285, -0.0026, 0.0007, -0.0018, -0.0544, 0.0, 0.0081, -0.0474, 0.0, 0.0263, 0.0527, -0.0072, 0.0042, 0.0039, -0.0333, -0.0121, 0.0132, -0.0375, -0.0344, -0.0333, 0.0351, -0.0325, -0.0538, -0.0963, -0.0318, 0.0538, 0.1053, 0.1053, 0.1141, 0.1608, 0.2933, 0.3541, 0.3686, 0.43, 0.4, 0.4247, 0.5029, 0.5029, 0.4414, 0.4537, 0.432, 0.4651, 0.8109, 0.8646, 0.8543, 1.0788, 1.0728, 1.1856, 1.1014, 1.3728, 1.4794, 1.405, 1.7407, 1.8398, 1.6461, 1.6391, 1.5434, 1.3864, 1.3732, 1.4309, 1.6099, 1.5998, 1.6292, 1.6292, 1.71, 1.6716, 1.6523, 1.516, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030657806920229226", "ticker_idx": 1, "ts": "2026-03-08T14:52:03+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CPUs and memory are emerging as the primary bottlenecks constraining compute performance in data centers", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares HSBC thesis that agentic AI lifts CPU demand 3-4x vs conventional AI; names INTC and AMD as plays on the trend through 2028.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> 'Agentic AI' Rise: The CPU's Comeback?\n\n• According to HSBC, approximately 44% of tasks in agentic AI workflows rely on the CPU — roughly 3–4x higher than in conventional AI compute architectures. This signals that CPUs and memory are emerging as the primary bottlenecks constraining compute performance in data centers.\n\n• HSBC has accordingly raised its 2026 global server shipment forecast to +20% YoY. The firm also notes that given supply chain capacity constraints, actual market demand could surge by as much as 60%, with growth tailwinds expected to extend through 2028.\n\n$INTC $AMD\n\n---\n\nhttps://t.co/Mfg4EbLv0r", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05057233259523797, "ret_signed_m1d": -0.05057233259523797, "alpha_spy_m1d": -0.04188837923088273, "alpha_c_m1d": -0.015554461738987668, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.0350178708562503, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002713652415794243, "ret_signed_p1d": 0.002713652415794243, "alpha_spy_p1d": 0.004320779321161372, "alpha_c_p1d": -0.0047919679197452325, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.007505620335539476, "ret_p1w": -0.03009666008072376, "ret_signed_p1w": -0.03009666008072376, "alpha_spy_p1w": -0.01647375335652501, "alpha_c_p1w": -0.02895564607097534, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.00114101400974842, "ret_p1m": 0.09300378322822023, "ret_signed_p1m": 0.09300378322822023, "alpha_spy_p1m": 0.11843574764783271, "alpha_c_p1m": 0.07898142136559017, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": 0.014022361862630062, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0925, 0.0925, 0.04, 0.0242, 0.032, -0.0225, -0.008, 0.053, 0.0605, 0.0603, 0.061, 0.061, 0.0607, 0.0638, 0.0625, 0.0566, 0.0566, 0.1026, 0.0908, 0.0576, 0.0362, 0.0099, 0.0024, 0.0247, 0.0902, 0.1032, 0.1245, 0.1438, 0.1438, 0.1443, 0.2325, 0.2519, 0.2812, 0.2399, 0.2435, 0.247, 0.2442, 0.168, 0.2151, 0.1945, -0.0123, -0.0502, 0.0284, 0.0657, 0.0537, 0.0538, 0.0161, 0.0229, 0.0229, 0.002, -0.0126, 0.0034, -0.0125, -0.03, 0.0551, 0.0404, 0.0049, -0.0122, -0.02, -0.0579, -0.003, -0.0159, -0.0506, 0.0, 0.0027, 0.0106, -0.0244, -0.0458, -0.0301, -0.0314, -0.0159, 0.0128, -0.0067, 0.0, 0.0133, 0.0868, 0.0054, -0.0034, -0.0328, 0.0037, 0.0372, 0.0731, 0.0731, 0.0863, 0.093, 0.1438, 0.1676, 0.209, 0.2178, 0.2585, 0.2735, 0.3729, 0.3735, 0.3566, 0.4036, 0.4972, 0.5065, 0.7161, 0.651, 0.5947, 0.6633, 0.749, 0.7789, 0.6851, 0.7528, 1.0791, 1.0153, 1.2459, 1.2636, 1.2118, 1.198, 1.2188, 1.0925, 1.0771, 1.0429, 1.2083, 1.2182, 1.3066, 1.3066, 1.4861, 1.4449, 1.5562, 1.5464, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2036341266410299511", "ticker_idx": 0, "ts": "2026-03-24T07:16:05+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura has revised up SK hynix's 2026F operating profit to KRW 256 trillion... and its 2027F operating profit to KRW 365 trillion", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Nomura's sharply raised SK Hynix profit estimates on DRAM/NAND price surge; implied bullish endorsement.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura’s latest memory semiconductor comment:\n\n\"Second-quarter memory price increases are now expected to significantly exceed previous forecasts. Commodity DRAM is projected to rise 51% QoQ and NAND 50% QoQ, far above prior estimates of +6% for DRAM and +20% for NAND. Accordingly, Nomura has revised up SK hynix’s 2026F operating profit to KRW 256 trillion (approximately US$175.3 billion, up 36% from the previous KRW 189 trillion estimate) and its 2027F operating profit to KRW 365 trillion (approximately US$253.1 billion, up 37% from the previous KRW 267 trillion estimate). Operating margin is projected at 74% in both years. Net profit is forecast at KRW 205.5 trillion for 2026F and KRW 293.5 trillion for 2027F.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0537525227659138, "ret_signed_m1d": -0.0537525227659138, "alpha_spy_m1d": -0.05712067891230588, "alpha_c_m1d": -0.045566783096116836, "bench_spy_m1d": 0.003368156146392076, "bench_c_m1d": -0.008185739669796965, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009127794061912775, "ret_signed_p1d": 0.009127794061912775, "alpha_spy_p1d": 0.0035550351145268166, "alpha_c_p1d": -0.0020990305329708914, "bench_spy_p1d": 0.0055727589473859585, "bench_c_p1d": 0.011226824594883666, "ret_p1w": -0.18154157623466705, "ret_signed_p1w": -0.18154157623466705, "alpha_spy_p1w": -0.17719366929912472, "alpha_c_p1w": -0.15318302040391008, "bench_spy_p1w": -0.004347906935542323, "bench_c_p1w": -0.02835855583075697, "ret_p1m": 0.2403651168343186, "ret_signed_p1m": 0.2403651168343186, "alpha_spy_p1m": 0.15152279167498173, "alpha_c_p1m": 0.0319462744976462, "bench_spy_p1m": 0.08884232515933688, "bench_c_p1m": 0.2084188423366724, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4047, -0.3936, -0.3521, -0.341, -0.341, -0.341, -0.3147, -0.2954, -0.265, -0.2489, -0.2347, -0.2468, -0.2418, -0.2529, -0.2489, -0.2418, -0.2347, -0.2266, -0.2478, -0.2509, -0.2357, -0.2235, -0.2549, -0.1901, -0.1486, -0.1284, -0.0798, -0.1598, -0.0818, -0.0889, -0.1476, -0.1507, -0.1021, -0.1132, -0.1294, -0.1011, -0.1091, -0.1091, -0.1091, -0.1091, -0.095, -0.0393, -0.0373, 0.0174, 0.0306, 0.1146, 0.0761, 0.0761, -0.0477, -0.1389, -0.0456, -0.0629, -0.1521, -0.0487, -0.0314, -0.0568, -0.0771, -0.0122, -0.0162, 0.071, 0.0274, 0.0213, -0.0538, 0.0, 0.0091, -0.0538, -0.0538, -0.1146, -0.1815, -0.0903, -0.1582, -0.1116, -0.1014, -0.071, 0.0477, 0.0122, 0.0416, 0.0548, 0.1187, 0.1521, 0.1714, 0.144, 0.1826, 0.2414, 0.2404, 0.2424, 0.2394, 0.3103, 0.3185, 0.3114, 0.3043, 0.3043, 0.4675, 0.4675, 0.6237, 0.6775, 0.7099, 0.9067, 0.8611, 1.0041, 0.998, 0.8448, 0.8661, 0.7698, 0.7698, 0.9675, 0.9686, 0.9686, 1.0811, 1.2748, 1.3219, 1.3665, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007041987917312438", "ticker_idx": 0, "ts": "2026-01-02T10:51:13+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Analysts expect the imbalance to persist as AI servers take up a growing share of industry output and suppliers allocate capacity toward higher-margin products", "tweet_type": "analysis", "timeline": "years", "summary": "Memory shortages to persist through 2026 as AI server demand outpaces supply for DRAM and NAND; pricing power remains with producers Samsung, Hynix, Micron, Nanya, Winbond, Phison.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory shortages to persist through 2026 as AI server demand surges\n\nGlobal memory markets are expected to remain tight through 2026 as aggressive spending by cloud service providers (CSPs) on AI infrastructure continues to outstrip supply growth for both DRAM and NAND flash, pushing prices higher, according to industry estimates.\n\nThe outlook follows a volatile 2025 for memory makers, who entered the year grappling with geopolitical uncertainty and lingering oversupply before conditions tightened sharply in the second half. Analysts expect the imbalance to persist as AI servers take up a growing share of industry output and suppliers allocate capacity toward higher-margin products.\n\nIndustry estimates show DRAM bit supply rising about 15% to 20% in 2026, while demand is projected to grow faster at roughly 20% to 25%. NAND flash shows a similar pattern, with bit supply growth of 13% to 18% compared with demand growth of 18% to 23%.\n\nThe gap is most pronounced in server applications. Analysts estimate DRAM and NAND consumption tied to servers will expand by 40% to 50% year on year in 2026, driven by sustained investment in AI training and inference at major cloud platforms.\n\nDDR4 exit tightens legacy supply\nA rapid pullback from legacy DRAM production has emerged as a central driver of the shortage. Leading suppliers are accelerating the phaseout of DDR4 processes, reallocating wafer capacity to newer and higher-margin products. Industry sources said the share of DDR4 wafer starts at Samsung Electronics and SK Hynix is expected to fall to low single digits by the second half of 2026, sharply reducing available supply.\n\nThe contraction has fueled a sharp rebound in DDR4 prices. Despite a gradual shift toward DDR5 in some segments, the spot price gap between DDR4 and DDR5 widened into the fourth quarter of 2025, pointing to sustained demand for the older standard. Market participants estimate DDR4 supply will continue to trail demand by around 10% in 2026, keeping prices elevated through at least the second half of the year.\n\nTaiwan-based suppliers have benefited from the shift. Nanya Technology has reinforced its position as the world's largest DDR4 supplier, while Winbond Electronics has raised capital spending and expanded capacity. Winbond plans to lift monthly output at its Kaohsiung facility from about 14,000 wafers to roughly 24,000 to 25,000 wafers, industry sources said.\n\nHBM crowds out DDR5\nSupply constraints are intensifying as high-bandwidth memory (HBM) takes up a growing share of advanced production. Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out, according to industry estimates. With SK Hynix's HBM4 already passing customer validation, rivals are expected to move quickly into HBM4, further tightening supplies of standard DDR5.\n\nAt the same time, AI servers and new CPU platforms are adopting higher capacity DDR5 RDIMM configurations, reinforcing suppliers' focus on enterprise and AI products. Market estimates show the contract price for Samsung's 64GB DDR5 RDIMM rose from about US$265 in the third quarter of 2025 to around US$450 in the fourth quarter and could approach US$480 in the first quarter of 2026, with additional increases possible.\n\nExcluding HBM, prices for conventional DRAM climbed by close to 50% or more in the fourth quarter of 2025, analysts said. That momentum is expected to carry into the first half of 2026, and the scale of the supply gap suggests limited scope for price declines even later in the year.\n\nDemand for 128GB and higher capacity DDR5 modules, along with SOCAMM2 configurations using LPDDR5X, is expected to absorb a larger share of DRAM output in 2026. HBM4 also involves higher wafer consumption and more complex yield management, adding to structural supply pressure.\n\nAs memory makers increasingly suspend formal quotations and push through repeated price increases, buyers in non-AI and consumer segments face rising procurement costs and difficulty securing supply.\n\nNAND capacity offers limited relief\nOn the NAND side, capacity additions remain constrained. While Kioxia and YMTC are building new facilities, industry sources said meaningful output contributions are unlikely before the second quarter of 2026, limiting any near-term impact on supply conditions.\n\nThe surge in AI inference is also reshaping NAND demand. Capital spending tied to inference infrastructure overtook training workloads in 2025 and is expected to rise further in 2026, driving rapid growth in enterprise SSD demand. North American cloud operators are increasingly seeking 128TB to 256TB SSDs, prompting a shift from TLC to QLC NAND to balance cost and storage density.\n\nAggressive capacity reservations by cloud providers have already pushed NAND prices sharply higher. Wafer prices jumped by roughly 95% to 100% quarter on quarter in the fourth quarter of 2025, according to market estimates, and some buyers were unable to secure supply even at higher prices. Analysts expect NAND shortages and price increases to extend into 2026, although the pace of increases is likely to moderate.\n\nModule makers brace for margin pressure\nTight supply conditions are forcing memory module makers to ration shipments and prioritize strategic customers. While higher average selling prices support revenue growth, rising input costs are squeezing margins and intensifying competition for supply.\n\nAdata said the market is likely to become increasingly polarized in 2026, with some module vendors able to secure a stable supply while others face persistent shortages. Industrial memory supplier Innodisk said cloud AI will be the key driver next year, adding that faster adoption of AI inference and edge applications could benefit vendors with established AI-focused portfolios.\n\nController maker Phison Electronics said it secured part of its 2026 supply during the seasonal slowdown, but volumes remain insufficient to meet demand. The company expects the supply imbalance to persist for several years and plans to scale back retail shipments while concentrating resources on higher-value enterprise customers.\n\nAnalysts cautioned that risks remain, including execution challenges in advanced memory production and potential shifts in cloud spending. Heading into 2026, AI-driven demand and structural supply constraints continue to leave pricing power firmly with memory producers.\n\n---\n\nhttps://t.co/e4erPPclae", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06692605458283052, "ret_signed_m1d": -0.06692605458283052, "alpha_spy_m1d": -0.06509636873623525, "alpha_c_m1d": -0.06463915473919735, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.0022868998436331722, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07470817945578978, "ret_signed_p1d": 0.07470817945578978, "alpha_spy_p1d": 0.06804806563972665, "alpha_c_p1d": 0.07249062773241488, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.0022175517233749, "ret_p1w": 0.08171206747205106, "ret_signed_p1w": 0.08171206747205106, "alpha_spy_p1w": 0.06575705280338862, "alpha_c_p1w": 0.06889155872365338, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.012820508748397685, "ret_p1m": 0.1704280229603632, "ret_signed_p1m": 0.1704280229603632, "alpha_spy_p1m": 0.15251154277998902, "alpha_c_p1m": 0.16377526191524572, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.006652761045117472, "ret_p3m": 0.4789302758506657, "ret_signed_p3m": 0.4789302758506657, "alpha_spy_p3m": 0.5171938588313623, "alpha_c_p3m": 0.5428405506058007, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": -0.063910274755135, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3049, -0.3049, -0.3049, -0.2689, -0.2774, -0.2906, -0.2643, -0.2434, -0.2418, -0.2403, -0.2449, -0.2364, -0.2527, -0.2348, -0.2101, -0.2294, -0.2217, -0.1938, -0.1675, -0.1396, -0.1876, -0.2209, -0.2318, -0.2418, -0.2209, -0.1984, -0.2015, -0.2039, -0.2472, -0.2209, -0.2426, -0.2527, -0.2209, -0.2658, -0.2511, -0.231, -0.2039, -0.1984, -0.2217, -0.2194, -0.1992, -0.1907, -0.1861, -0.1605, -0.152, -0.1605, -0.1636, -0.169, -0.1566, -0.1884, -0.2039, -0.1644, -0.1667, -0.1768, -0.1442, -0.1365, -0.1396, -0.1396, -0.0939, -0.07, -0.0669, -0.0669, -0.0669, 0.0, 0.0747, 0.0809, 0.0973, 0.0802, 0.0817, 0.0802, 0.0708, 0.0918, 0.1198, 0.1588, 0.1619, 0.13, 0.1634, 0.1852, 0.1837, 0.1837, 0.2412, 0.2638, 0.2506, 0.249, 0.1704, 0.3035, 0.316, 0.2397, 0.2342, 0.2949, 0.2903, 0.3058, 0.3899, 0.4101, 0.4101, 0.4101, 0.4101, 0.4786, 0.4794, 0.5019, 0.5564, 0.5837, 0.6965, 0.6848, 0.6848, 0.5183, 0.3401, 0.4911, 0.4646, 0.3502, 0.4623, 0.4786, 0.4623, 0.428, 0.4685, 0.5089, 0.6226, 0.5603, 0.5518, 0.4498, 0.4763, 0.4708, 0.4016, 0.4016, 0.3748, 0.3039, 0.4789, 0.3912, 0.452, 0.5058, 0.5323, 0.6415, 0.5908, 0.6064, 0.5674, 0.6103, 0.6454, 0.6961, 0.6844, 0.6727, 0.7078, 0.6961, 0.7507, 0.7117, 0.7507, 0.7312, 0.7624, 0.7195, 0.7195, 0.8131, 0.8131, 1.0743, 1.1172, 1.0938, 1.2264, 1.1757, 1.2147, 1.3083, 1.1094, 1.1913, 1.1484, 1.1523, 1.3356, 1.281, 1.281, 1.3317, 1.3941, 1.3356, 1.472, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007041987917312438", "ticker_idx": 1, "ts": "2026-01-02T10:51:13+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out", "tweet_type": "analysis", "timeline": "years", "summary": "Memory shortages to persist through 2026 as AI server demand outpaces supply for DRAM and NAND; pricing power remains with producers Samsung, Hynix, Micron, Nanya, Winbond, Phison.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory shortages to persist through 2026 as AI server demand surges\n\nGlobal memory markets are expected to remain tight through 2026 as aggressive spending by cloud service providers (CSPs) on AI infrastructure continues to outstrip supply growth for both DRAM and NAND flash, pushing prices higher, according to industry estimates.\n\nThe outlook follows a volatile 2025 for memory makers, who entered the year grappling with geopolitical uncertainty and lingering oversupply before conditions tightened sharply in the second half. Analysts expect the imbalance to persist as AI servers take up a growing share of industry output and suppliers allocate capacity toward higher-margin products.\n\nIndustry estimates show DRAM bit supply rising about 15% to 20% in 2026, while demand is projected to grow faster at roughly 20% to 25%. NAND flash shows a similar pattern, with bit supply growth of 13% to 18% compared with demand growth of 18% to 23%.\n\nThe gap is most pronounced in server applications. Analysts estimate DRAM and NAND consumption tied to servers will expand by 40% to 50% year on year in 2026, driven by sustained investment in AI training and inference at major cloud platforms.\n\nDDR4 exit tightens legacy supply\nA rapid pullback from legacy DRAM production has emerged as a central driver of the shortage. Leading suppliers are accelerating the phaseout of DDR4 processes, reallocating wafer capacity to newer and higher-margin products. Industry sources said the share of DDR4 wafer starts at Samsung Electronics and SK Hynix is expected to fall to low single digits by the second half of 2026, sharply reducing available supply.\n\nThe contraction has fueled a sharp rebound in DDR4 prices. Despite a gradual shift toward DDR5 in some segments, the spot price gap between DDR4 and DDR5 widened into the fourth quarter of 2025, pointing to sustained demand for the older standard. Market participants estimate DDR4 supply will continue to trail demand by around 10% in 2026, keeping prices elevated through at least the second half of the year.\n\nTaiwan-based suppliers have benefited from the shift. Nanya Technology has reinforced its position as the world's largest DDR4 supplier, while Winbond Electronics has raised capital spending and expanded capacity. Winbond plans to lift monthly output at its Kaohsiung facility from about 14,000 wafers to roughly 24,000 to 25,000 wafers, industry sources said.\n\nHBM crowds out DDR5\nSupply constraints are intensifying as high-bandwidth memory (HBM) takes up a growing share of advanced production. Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out, according to industry estimates. With SK Hynix's HBM4 already passing customer validation, rivals are expected to move quickly into HBM4, further tightening supplies of standard DDR5.\n\nAt the same time, AI servers and new CPU platforms are adopting higher capacity DDR5 RDIMM configurations, reinforcing suppliers' focus on enterprise and AI products. Market estimates show the contract price for Samsung's 64GB DDR5 RDIMM rose from about US$265 in the third quarter of 2025 to around US$450 in the fourth quarter and could approach US$480 in the first quarter of 2026, with additional increases possible.\n\nExcluding HBM, prices for conventional DRAM climbed by close to 50% or more in the fourth quarter of 2025, analysts said. That momentum is expected to carry into the first half of 2026, and the scale of the supply gap suggests limited scope for price declines even later in the year.\n\nDemand for 128GB and higher capacity DDR5 modules, along with SOCAMM2 configurations using LPDDR5X, is expected to absorb a larger share of DRAM output in 2026. HBM4 also involves higher wafer consumption and more complex yield management, adding to structural supply pressure.\n\nAs memory makers increasingly suspend formal quotations and push through repeated price increases, buyers in non-AI and consumer segments face rising procurement costs and difficulty securing supply.\n\nNAND capacity offers limited relief\nOn the NAND side, capacity additions remain constrained. While Kioxia and YMTC are building new facilities, industry sources said meaningful output contributions are unlikely before the second quarter of 2026, limiting any near-term impact on supply conditions.\n\nThe surge in AI inference is also reshaping NAND demand. Capital spending tied to inference infrastructure overtook training workloads in 2025 and is expected to rise further in 2026, driving rapid growth in enterprise SSD demand. North American cloud operators are increasingly seeking 128TB to 256TB SSDs, prompting a shift from TLC to QLC NAND to balance cost and storage density.\n\nAggressive capacity reservations by cloud providers have already pushed NAND prices sharply higher. Wafer prices jumped by roughly 95% to 100% quarter on quarter in the fourth quarter of 2025, according to market estimates, and some buyers were unable to secure supply even at higher prices. Analysts expect NAND shortages and price increases to extend into 2026, although the pace of increases is likely to moderate.\n\nModule makers brace for margin pressure\nTight supply conditions are forcing memory module makers to ration shipments and prioritize strategic customers. While higher average selling prices support revenue growth, rising input costs are squeezing margins and intensifying competition for supply.\n\nAdata said the market is likely to become increasingly polarized in 2026, with some module vendors able to secure a stable supply while others face persistent shortages. Industrial memory supplier Innodisk said cloud AI will be the key driver next year, adding that faster adoption of AI inference and edge applications could benefit vendors with established AI-focused portfolios.\n\nController maker Phison Electronics said it secured part of its 2026 supply during the seasonal slowdown, but volumes remain insufficient to meet demand. The company expects the supply imbalance to persist for several years and plans to scale back retail shipments while concentrating resources on higher-value enterprise customers.\n\nAnalysts cautioned that risks remain, including execution challenges in advanced memory production and potential shifts in cloud spending. Heading into 2026, AI-driven demand and structural supply constraints continue to leave pricing power firmly with memory producers.\n\n---\n\nhttps://t.co/e4erPPclae", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.038404694215420365, "ret_signed_m1d": -0.038404694215420365, "alpha_spy_m1d": -0.0365750083688251, "alpha_c_m1d": -0.00312480313366581, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.035279891081754555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02806505423884631, "ret_signed_p1d": 0.02806505423884631, "alpha_spy_p1d": 0.02140494042278318, "alpha_c_p1d": 0.0165729334641167, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.011492120774729608, "ret_p1w": 0.09896606375380523, "ret_signed_p1w": 0.09896606375380523, "alpha_spy_p1w": 0.08301104908514279, "alpha_c_p1w": 0.05631936150809391, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.04264670224571132, "ret_p1m": 0.22599708924158124, "ret_signed_p1m": 0.22599708924158124, "alpha_spy_p1m": 0.20808060906120707, "alpha_c_p1m": 0.13312268645172032, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.09287440278986092, "ret_p3m": 0.3274079643552812, "ret_signed_p3m": 0.3274079643552812, "alpha_spy_p3m": 0.36567154733597773, "alpha_c_p3m": 0.277394532696428, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": 0.05001343165885319, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4162, -0.4162, -0.4162, -0.3683, -0.3874, -0.3926, -0.3764, -0.3321, -0.3129, -0.2834, -0.293, -0.2893, -0.2937, -0.2472, -0.2103, -0.231, -0.1764, -0.1616, -0.1749, -0.0849, -0.135, -0.1454, -0.1247, -0.1439, -0.1055, -0.0863, -0.0893, -0.0967, -0.1734, -0.1055, -0.1587, -0.1705, -0.1572, -0.231, -0.2325, -0.2339, -0.2266, -0.1965, -0.2171, -0.2053, -0.1758, -0.1846, -0.1994, -0.1965, -0.1477, -0.164, -0.1329, -0.1654, -0.1566, -0.1817, -0.2171, -0.1861, -0.1846, -0.192, -0.1433, -0.1374, -0.1315, -0.1315, -0.1152, -0.0547, -0.0384, -0.0384, -0.0384, 0.0, 0.0281, 0.0724, 0.096, 0.1167, 0.099, 0.1064, 0.0901, 0.096, 0.1064, 0.1167, 0.1285, 0.0975, 0.0931, 0.1152, 0.1329, 0.0871, 0.1817, 0.2422, 0.2718, 0.3427, 0.226, 0.3397, 0.3294, 0.2437, 0.2393, 0.3102, 0.2939, 0.2703, 0.3117, 0.2999, 0.2999, 0.2999, 0.2999, 0.3205, 0.4018, 0.4047, 0.4845, 0.5037, 0.6263, 0.5701, 0.5701, 0.3896, 0.2564, 0.3925, 0.3674, 0.2371, 0.3881, 0.4132, 0.3762, 0.3466, 0.4414, 0.4354, 0.5627, 0.4991, 0.4902, 0.3807, 0.4591, 0.4724, 0.3807, 0.3807, 0.2919, 0.1942, 0.3274, 0.2283, 0.2963, 0.3111, 0.3555, 0.5287, 0.4769, 0.5198, 0.539, 0.6323, 0.6811, 0.7092, 0.6692, 0.7255, 0.8113, 0.8098, 0.8128, 0.8084, 0.9119, 0.9238, 0.9134, 0.9031, 0.9031, 1.1413, 1.1413, 1.3692, 1.4476, 1.495, 1.7821, 1.7155, 1.9241, 1.9153, 1.6918, 1.7229, 1.5823, 1.5823, 1.8709, 1.8724, 1.8724, 2.0366, 2.3193, 2.3879, 2.453, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007041987917312438", "ticker_idx": 2, "ts": "2026-01-02T10:51:13+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out", "tweet_type": "analysis", "timeline": "years", "summary": "Memory shortages to persist through 2026 as AI server demand outpaces supply for DRAM and NAND; pricing power remains with producers Samsung, Hynix, Micron, Nanya, Winbond, Phison.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory shortages to persist through 2026 as AI server demand surges\n\nGlobal memory markets are expected to remain tight through 2026 as aggressive spending by cloud service providers (CSPs) on AI infrastructure continues to outstrip supply growth for both DRAM and NAND flash, pushing prices higher, according to industry estimates.\n\nThe outlook follows a volatile 2025 for memory makers, who entered the year grappling with geopolitical uncertainty and lingering oversupply before conditions tightened sharply in the second half. Analysts expect the imbalance to persist as AI servers take up a growing share of industry output and suppliers allocate capacity toward higher-margin products.\n\nIndustry estimates show DRAM bit supply rising about 15% to 20% in 2026, while demand is projected to grow faster at roughly 20% to 25%. NAND flash shows a similar pattern, with bit supply growth of 13% to 18% compared with demand growth of 18% to 23%.\n\nThe gap is most pronounced in server applications. Analysts estimate DRAM and NAND consumption tied to servers will expand by 40% to 50% year on year in 2026, driven by sustained investment in AI training and inference at major cloud platforms.\n\nDDR4 exit tightens legacy supply\nA rapid pullback from legacy DRAM production has emerged as a central driver of the shortage. Leading suppliers are accelerating the phaseout of DDR4 processes, reallocating wafer capacity to newer and higher-margin products. Industry sources said the share of DDR4 wafer starts at Samsung Electronics and SK Hynix is expected to fall to low single digits by the second half of 2026, sharply reducing available supply.\n\nThe contraction has fueled a sharp rebound in DDR4 prices. Despite a gradual shift toward DDR5 in some segments, the spot price gap between DDR4 and DDR5 widened into the fourth quarter of 2025, pointing to sustained demand for the older standard. Market participants estimate DDR4 supply will continue to trail demand by around 10% in 2026, keeping prices elevated through at least the second half of the year.\n\nTaiwan-based suppliers have benefited from the shift. Nanya Technology has reinforced its position as the world's largest DDR4 supplier, while Winbond Electronics has raised capital spending and expanded capacity. Winbond plans to lift monthly output at its Kaohsiung facility from about 14,000 wafers to roughly 24,000 to 25,000 wafers, industry sources said.\n\nHBM crowds out DDR5\nSupply constraints are intensifying as high-bandwidth memory (HBM) takes up a growing share of advanced production. Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out, according to industry estimates. With SK Hynix's HBM4 already passing customer validation, rivals are expected to move quickly into HBM4, further tightening supplies of standard DDR5.\n\nAt the same time, AI servers and new CPU platforms are adopting higher capacity DDR5 RDIMM configurations, reinforcing suppliers' focus on enterprise and AI products. Market estimates show the contract price for Samsung's 64GB DDR5 RDIMM rose from about US$265 in the third quarter of 2025 to around US$450 in the fourth quarter and could approach US$480 in the first quarter of 2026, with additional increases possible.\n\nExcluding HBM, prices for conventional DRAM climbed by close to 50% or more in the fourth quarter of 2025, analysts said. That momentum is expected to carry into the first half of 2026, and the scale of the supply gap suggests limited scope for price declines even later in the year.\n\nDemand for 128GB and higher capacity DDR5 modules, along with SOCAMM2 configurations using LPDDR5X, is expected to absorb a larger share of DRAM output in 2026. HBM4 also involves higher wafer consumption and more complex yield management, adding to structural supply pressure.\n\nAs memory makers increasingly suspend formal quotations and push through repeated price increases, buyers in non-AI and consumer segments face rising procurement costs and difficulty securing supply.\n\nNAND capacity offers limited relief\nOn the NAND side, capacity additions remain constrained. While Kioxia and YMTC are building new facilities, industry sources said meaningful output contributions are unlikely before the second quarter of 2026, limiting any near-term impact on supply conditions.\n\nThe surge in AI inference is also reshaping NAND demand. Capital spending tied to inference infrastructure overtook training workloads in 2025 and is expected to rise further in 2026, driving rapid growth in enterprise SSD demand. North American cloud operators are increasingly seeking 128TB to 256TB SSDs, prompting a shift from TLC to QLC NAND to balance cost and storage density.\n\nAggressive capacity reservations by cloud providers have already pushed NAND prices sharply higher. Wafer prices jumped by roughly 95% to 100% quarter on quarter in the fourth quarter of 2025, according to market estimates, and some buyers were unable to secure supply even at higher prices. Analysts expect NAND shortages and price increases to extend into 2026, although the pace of increases is likely to moderate.\n\nModule makers brace for margin pressure\nTight supply conditions are forcing memory module makers to ration shipments and prioritize strategic customers. While higher average selling prices support revenue growth, rising input costs are squeezing margins and intensifying competition for supply.\n\nAdata said the market is likely to become increasingly polarized in 2026, with some module vendors able to secure a stable supply while others face persistent shortages. Industrial memory supplier Innodisk said cloud AI will be the key driver next year, adding that faster adoption of AI inference and edge applications could benefit vendors with established AI-focused portfolios.\n\nController maker Phison Electronics said it secured part of its 2026 supply during the seasonal slowdown, but volumes remain insufficient to meet demand. The company expects the supply imbalance to persist for several years and plans to scale back retail shipments while concentrating resources on higher-value enterprise customers.\n\nAnalysts cautioned that risks remain, including execution challenges in advanced memory production and potential shifts in cloud spending. Heading into 2026, AI-driven demand and structural supply constraints continue to leave pricing power firmly with memory producers.\n\n---\n\nhttps://t.co/e4erPPclae", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09514300318942082, "ret_signed_m1d": -0.09514300318942082, "alpha_spy_m1d": -0.09331331734282555, "alpha_c_m1d": -0.05986311210766626, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.035279891081754555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010367190403028914, "ret_signed_p1d": -0.010367190403028914, "alpha_spy_p1d": -0.017027304219092043, "alpha_c_p1d": -0.021859311177758523, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.011492120774729608, "ret_p1w": 0.09406502136266304, "ret_signed_p1w": 0.09406502136266304, "alpha_spy_p1w": 0.07811000669400059, "alpha_c_p1w": 0.051418319116951716, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.04264670224571132, "ret_p1m": 0.3879905061713662, "ret_signed_p1m": 0.3879905061713662, "alpha_spy_p1m": 0.37007402599099204, "alpha_c_p1m": 0.2951161033815053, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.09287440278986092, "ret_p3m": 0.16671269166070668, "ret_signed_p3m": 0.16671269166070668, "alpha_spy_p3m": 0.20497627464140322, "alpha_c_p3m": 0.11669926000185349, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": 0.05001343165885319, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4115, -0.3771, -0.3905, -0.4245, -0.3891, -0.4072, -0.3917, -0.3582, -0.3586, -0.3447, -0.3589, -0.371, -0.3449, -0.3059, -0.3025, -0.2967, -0.2818, -0.2901, -0.2909, -0.2562, -0.309, -0.2473, -0.2447, -0.246, -0.1973, -0.2359, -0.2239, -0.2491, -0.2178, -0.2332, -0.2759, -0.284, -0.3618, -0.3428, -0.2903, -0.2884, -0.2703, -0.2703, -0.2506, -0.238, -0.241, -0.2579, -0.2817, -0.2482, -0.2175, -0.2001, -0.1643, -0.1809, -0.2358, -0.2473, -0.2632, -0.2853, -0.2123, -0.1573, -0.1235, -0.1245, -0.0915, -0.0915, -0.0975, -0.0667, -0.0723, -0.0951, -0.0951, 0.0, -0.0104, 0.0888, 0.0765, 0.0368, 0.0941, 0.0965, 0.072, 0.0568, 0.0672, 0.1501, 0.1501, 0.1572, 0.2336, 0.2605, 0.267, 0.2336, 0.3006, 0.38, 0.3816, 0.3153, 0.388, 0.3298, 0.2028, 0.2139, 0.2513, 0.2158, 0.1833, 0.3009, 0.3124, 0.3051, 0.3051, 0.2675, 0.3346, 0.3232, 0.3575, 0.3346, 0.3252, 0.3601, 0.3175, 0.3074, 0.3083, 0.2037, 0.2706, 0.2588, 0.174, 0.2343, 0.278, 0.3274, 0.2851, 0.351, 0.4007, 0.4637, 0.4639, 0.4085, 0.3408, 0.2819, 0.254, 0.2114, 0.1269, 0.1325, 0.0207, 0.0715, 0.1667, 0.1616, 0.1616, 0.1981, 0.1976, 0.29, 0.3369, 0.334, 0.3529, 0.4769, 0.447, 0.4502, 0.4433, 0.4223, 0.4253, 0.5461, 0.5279, 0.5755, 0.6638, 0.5995, 0.6444, 0.6403, 0.7197, 0.8283, 1.0305, 1.1142, 1.0509, 1.3687, 1.5226, 1.4314, 1.5489, 1.4613, 1.2984, 1.1616, 1.2162, 1.3217, 1.4172, 1.382, 1.382, 1.8415, 1.9446, 1.9291, 2.0797, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007041987917312438", "ticker_idx": 3, "ts": "2026-01-02T10:51:13+00:00", "symbol": "Nanya Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nanya Technology has reinforced its position as the world's largest DDR4 supplier", "tweet_type": "analysis", "timeline": "years", "summary": "Memory shortages to persist through 2026 as AI server demand outpaces supply for DRAM and NAND; pricing power remains with producers Samsung, Hynix, Micron, Nanya, Winbond, Phison.", "resolved_tickers": ["2408.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya Technology Corp listed on TWSE as 2408.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory shortages to persist through 2026 as AI server demand surges\n\nGlobal memory markets are expected to remain tight through 2026 as aggressive spending by cloud service providers (CSPs) on AI infrastructure continues to outstrip supply growth for both DRAM and NAND flash, pushing prices higher, according to industry estimates.\n\nThe outlook follows a volatile 2025 for memory makers, who entered the year grappling with geopolitical uncertainty and lingering oversupply before conditions tightened sharply in the second half. Analysts expect the imbalance to persist as AI servers take up a growing share of industry output and suppliers allocate capacity toward higher-margin products.\n\nIndustry estimates show DRAM bit supply rising about 15% to 20% in 2026, while demand is projected to grow faster at roughly 20% to 25%. NAND flash shows a similar pattern, with bit supply growth of 13% to 18% compared with demand growth of 18% to 23%.\n\nThe gap is most pronounced in server applications. Analysts estimate DRAM and NAND consumption tied to servers will expand by 40% to 50% year on year in 2026, driven by sustained investment in AI training and inference at major cloud platforms.\n\nDDR4 exit tightens legacy supply\nA rapid pullback from legacy DRAM production has emerged as a central driver of the shortage. Leading suppliers are accelerating the phaseout of DDR4 processes, reallocating wafer capacity to newer and higher-margin products. Industry sources said the share of DDR4 wafer starts at Samsung Electronics and SK Hynix is expected to fall to low single digits by the second half of 2026, sharply reducing available supply.\n\nThe contraction has fueled a sharp rebound in DDR4 prices. Despite a gradual shift toward DDR5 in some segments, the spot price gap between DDR4 and DDR5 widened into the fourth quarter of 2025, pointing to sustained demand for the older standard. Market participants estimate DDR4 supply will continue to trail demand by around 10% in 2026, keeping prices elevated through at least the second half of the year.\n\nTaiwan-based suppliers have benefited from the shift. Nanya Technology has reinforced its position as the world's largest DDR4 supplier, while Winbond Electronics has raised capital spending and expanded capacity. Winbond plans to lift monthly output at its Kaohsiung facility from about 14,000 wafers to roughly 24,000 to 25,000 wafers, industry sources said.\n\nHBM crowds out DDR5\nSupply constraints are intensifying as high-bandwidth memory (HBM) takes up a growing share of advanced production. Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out, according to industry estimates. With SK Hynix's HBM4 already passing customer validation, rivals are expected to move quickly into HBM4, further tightening supplies of standard DDR5.\n\nAt the same time, AI servers and new CPU platforms are adopting higher capacity DDR5 RDIMM configurations, reinforcing suppliers' focus on enterprise and AI products. Market estimates show the contract price for Samsung's 64GB DDR5 RDIMM rose from about US$265 in the third quarter of 2025 to around US$450 in the fourth quarter and could approach US$480 in the first quarter of 2026, with additional increases possible.\n\nExcluding HBM, prices for conventional DRAM climbed by close to 50% or more in the fourth quarter of 2025, analysts said. That momentum is expected to carry into the first half of 2026, and the scale of the supply gap suggests limited scope for price declines even later in the year.\n\nDemand for 128GB and higher capacity DDR5 modules, along with SOCAMM2 configurations using LPDDR5X, is expected to absorb a larger share of DRAM output in 2026. HBM4 also involves higher wafer consumption and more complex yield management, adding to structural supply pressure.\n\nAs memory makers increasingly suspend formal quotations and push through repeated price increases, buyers in non-AI and consumer segments face rising procurement costs and difficulty securing supply.\n\nNAND capacity offers limited relief\nOn the NAND side, capacity additions remain constrained. While Kioxia and YMTC are building new facilities, industry sources said meaningful output contributions are unlikely before the second quarter of 2026, limiting any near-term impact on supply conditions.\n\nThe surge in AI inference is also reshaping NAND demand. Capital spending tied to inference infrastructure overtook training workloads in 2025 and is expected to rise further in 2026, driving rapid growth in enterprise SSD demand. North American cloud operators are increasingly seeking 128TB to 256TB SSDs, prompting a shift from TLC to QLC NAND to balance cost and storage density.\n\nAggressive capacity reservations by cloud providers have already pushed NAND prices sharply higher. Wafer prices jumped by roughly 95% to 100% quarter on quarter in the fourth quarter of 2025, according to market estimates, and some buyers were unable to secure supply even at higher prices. Analysts expect NAND shortages and price increases to extend into 2026, although the pace of increases is likely to moderate.\n\nModule makers brace for margin pressure\nTight supply conditions are forcing memory module makers to ration shipments and prioritize strategic customers. While higher average selling prices support revenue growth, rising input costs are squeezing margins and intensifying competition for supply.\n\nAdata said the market is likely to become increasingly polarized in 2026, with some module vendors able to secure a stable supply while others face persistent shortages. Industrial memory supplier Innodisk said cloud AI will be the key driver next year, adding that faster adoption of AI inference and edge applications could benefit vendors with established AI-focused portfolios.\n\nController maker Phison Electronics said it secured part of its 2026 supply during the seasonal slowdown, but volumes remain insufficient to meet demand. The company expects the supply imbalance to persist for several years and plans to scale back retail shipments while concentrating resources on higher-value enterprise customers.\n\nAnalysts cautioned that risks remain, including execution challenges in advanced memory production and potential shifts in cloud spending. Heading into 2026, AI-driven demand and structural supply constraints continue to leave pricing power firmly with memory producers.\n\n---\n\nhttps://t.co/e4erPPclae", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06763285024154586, "ret_signed_m1d": -0.06763285024154586, "alpha_spy_m1d": -0.0658031643949506, "alpha_c_m1d": -0.032352959159791306, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.035279891081754555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004830917874396157, "ret_signed_p1d": 0.004830917874396157, "alpha_spy_p1d": -0.0018291959416669723, "alpha_c_p1d": -0.006661202900333452, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.011492120774729608, "ret_p1w": 0.050724637681159424, "ret_signed_p1w": 0.050724637681159424, "alpha_spy_p1w": 0.034769623012496975, "alpha_c_p1w": 0.008077935435448103, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.04264670224571132, "ret_p1m": 0.42028985507246386, "ret_signed_p1m": 0.42028985507246386, "alpha_spy_p1m": 0.4023733748920897, "alpha_c_p1m": 0.32741545228260294, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.09287440278986092, "ret_p3m": 0.019323671497584627, "ret_signed_p3m": 0.019323671497584627, "alpha_spy_p3m": 0.05758725447828117, "alpha_c_p3m": -0.030689760161268564, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": 0.05001343165885319, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5594, -0.5614, -0.5242, -0.5242, -0.5382, -0.5696, -0.5831, -0.5415, -0.4976, -0.4831, -0.4903, -0.4686, -0.471, -0.471, -0.4203, -0.372, -0.3575, -0.3502, -0.3599, -0.3382, -0.3647, -0.3333, -0.285, -0.2899, -0.2198, -0.2053, -0.2126, -0.2101, -0.2343, -0.1957, -0.2246, -0.2271, -0.2488, -0.3237, -0.3092, -0.3068, -0.3406, -0.2947, -0.2947, -0.2778, -0.2778, -0.2705, -0.2705, -0.2609, -0.2101, -0.2174, -0.2464, -0.2464, -0.2174, -0.215, -0.2367, -0.2053, -0.1787, -0.157, -0.1304, -0.1473, -0.087, -0.087, -0.087, -0.0894, -0.0604, -0.0676, -0.0676, 0.0, 0.0048, 0.1039, 0.1643, 0.1667, 0.0507, 0.1546, 0.1111, 0.1425, 0.1473, 0.2077, 0.3285, 0.314, 0.2126, 0.2874, 0.3116, 0.3816, 0.3382, 0.4348, 0.4348, 0.5773, 0.4203, 0.3406, 0.4251, 0.3285, 0.2754, 0.3261, 0.3092, 0.343, 0.343, 0.343, 0.343, 0.343, 0.343, 0.343, 0.343, 0.3961, 0.4396, 0.3913, 0.3792, 0.3792, 0.3768, 0.2391, 0.1329, 0.244, 0.1256, 0.0145, 0.1159, 0.2271, 0.157, 0.1498, 0.2633, 0.3116, 0.3333, 0.256, 0.1304, 0.0773, 0.0459, 0.0942, 0.0894, 0.0604, 0.0652, -0.0411, 0.0193, -0.0314, -0.0314, -0.0314, -0.0048, 0.0773, -0.0097, 0.0386, 0.0894, 0.0628, 0.0217, 0.0314, 0.0145, 0.0, 0.0628, 0.058, 0.0048, -0.0048, 0.0942, 0.1473, 0.1353, 0.0411, 0.0411, 0.1449, 0.2391, 0.3623, 0.3865, 0.3237, 0.4541, 0.5507, 0.5507, 0.6473, 0.5048, 0.4734, 0.3261, 0.3285, 0.4155, 0.5, 0.43, 0.4662, 0.5072, 0.5652, 0.6763, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007041987917312438", "ticker_idx": 4, "ts": "2026-01-02T10:51:13+00:00", "symbol": "Winbond Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Winbond Electronics has raised capital spending and expanded capacity", "tweet_type": "analysis", "timeline": "years", "summary": "Memory shortages to persist through 2026 as AI server demand outpaces supply for DRAM and NAND; pricing power remains with producers Samsung, Hynix, Micron, Nanya, Winbond, Phison.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory shortages to persist through 2026 as AI server demand surges\n\nGlobal memory markets are expected to remain tight through 2026 as aggressive spending by cloud service providers (CSPs) on AI infrastructure continues to outstrip supply growth for both DRAM and NAND flash, pushing prices higher, according to industry estimates.\n\nThe outlook follows a volatile 2025 for memory makers, who entered the year grappling with geopolitical uncertainty and lingering oversupply before conditions tightened sharply in the second half. Analysts expect the imbalance to persist as AI servers take up a growing share of industry output and suppliers allocate capacity toward higher-margin products.\n\nIndustry estimates show DRAM bit supply rising about 15% to 20% in 2026, while demand is projected to grow faster at roughly 20% to 25%. NAND flash shows a similar pattern, with bit supply growth of 13% to 18% compared with demand growth of 18% to 23%.\n\nThe gap is most pronounced in server applications. Analysts estimate DRAM and NAND consumption tied to servers will expand by 40% to 50% year on year in 2026, driven by sustained investment in AI training and inference at major cloud platforms.\n\nDDR4 exit tightens legacy supply\nA rapid pullback from legacy DRAM production has emerged as a central driver of the shortage. Leading suppliers are accelerating the phaseout of DDR4 processes, reallocating wafer capacity to newer and higher-margin products. Industry sources said the share of DDR4 wafer starts at Samsung Electronics and SK Hynix is expected to fall to low single digits by the second half of 2026, sharply reducing available supply.\n\nThe contraction has fueled a sharp rebound in DDR4 prices. Despite a gradual shift toward DDR5 in some segments, the spot price gap between DDR4 and DDR5 widened into the fourth quarter of 2025, pointing to sustained demand for the older standard. Market participants estimate DDR4 supply will continue to trail demand by around 10% in 2026, keeping prices elevated through at least the second half of the year.\n\nTaiwan-based suppliers have benefited from the shift. Nanya Technology has reinforced its position as the world's largest DDR4 supplier, while Winbond Electronics has raised capital spending and expanded capacity. Winbond plans to lift monthly output at its Kaohsiung facility from about 14,000 wafers to roughly 24,000 to 25,000 wafers, industry sources said.\n\nHBM crowds out DDR5\nSupply constraints are intensifying as high-bandwidth memory (HBM) takes up a growing share of advanced production. Capacity for HBM3E at SK Hynix, Micron Technology, and Samsung Electronics is largely booked out, according to industry estimates. With SK Hynix's HBM4 already passing customer validation, rivals are expected to move quickly into HBM4, further tightening supplies of standard DDR5.\n\nAt the same time, AI servers and new CPU platforms are adopting higher capacity DDR5 RDIMM configurations, reinforcing suppliers' focus on enterprise and AI products. Market estimates show the contract price for Samsung's 64GB DDR5 RDIMM rose from about US$265 in the third quarter of 2025 to around US$450 in the fourth quarter and could approach US$480 in the first quarter of 2026, with additional increases possible.\n\nExcluding HBM, prices for conventional DRAM climbed by close to 50% or more in the fourth quarter of 2025, analysts said. That momentum is expected to carry into the first half of 2026, and the scale of the supply gap suggests limited scope for price declines even later in the year.\n\nDemand for 128GB and higher capacity DDR5 modules, along with SOCAMM2 configurations using LPDDR5X, is expected to absorb a larger share of DRAM output in 2026. HBM4 also involves higher wafer consumption and more complex yield management, adding to structural supply pressure.\n\nAs memory makers increasingly suspend formal quotations and push through repeated price increases, buyers in non-AI and consumer segments face rising procurement costs and difficulty securing supply.\n\nNAND capacity offers limited relief\nOn the NAND side, capacity additions remain constrained. While Kioxia and YMTC are building new facilities, industry sources said meaningful output contributions are unlikely before the second quarter of 2026, limiting any near-term impact on supply conditions.\n\nThe surge in AI inference is also reshaping NAND demand. Capital spending tied to inference infrastructure overtook training workloads in 2025 and is expected to rise further in 2026, driving rapid growth in enterprise SSD demand. North American cloud operators are increasingly seeking 128TB to 256TB SSDs, prompting a shift from TLC to QLC NAND to balance cost and storage density.\n\nAggressive capacity reservations by cloud providers have already pushed NAND prices sharply higher. Wafer prices jumped by roughly 95% to 100% quarter on quarter in the fourth quarter of 2025, according to market estimates, and some buyers were unable to secure supply even at higher prices. Analysts expect NAND shortages and price increases to extend into 2026, although the pace of increases is likely to moderate.\n\nModule makers brace for margin pressure\nTight supply conditions are forcing memory module makers to ration shipments and prioritize strategic customers. While higher average selling prices support revenue growth, rising input costs are squeezing margins and intensifying competition for supply.\n\nAdata said the market is likely to become increasingly polarized in 2026, with some module vendors able to secure a stable supply while others face persistent shortages. Industrial memory supplier Innodisk said cloud AI will be the key driver next year, adding that faster adoption of AI inference and edge applications could benefit vendors with established AI-focused portfolios.\n\nController maker Phison Electronics said it secured part of its 2026 supply during the seasonal slowdown, but volumes remain insufficient to meet demand. The company expects the supply imbalance to persist for several years and plans to scale back retail shipments while concentrating resources on higher-value enterprise customers.\n\nAnalysts cautioned that risks remain, including execution challenges in advanced memory production and potential shifts in cloud spending. Heading into 2026, AI-driven demand and structural supply constraints continue to leave pricing power firmly with memory producers.\n\n---\n\nhttps://t.co/e4erPPclae", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09030840539245732, "ret_signed_m1d": -0.09030840539245732, "alpha_spy_m1d": -0.08847871954586206, "alpha_c_m1d": -0.05502851431070277, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.035279891081754555, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.051762123467568255, "ret_signed_p1d": 0.051762123467568255, "alpha_spy_p1d": 0.045102009651505126, "alpha_c_p1d": 0.04027000269283865, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.011492120774729608, "ret_p1w": 0.07709247938569619, "ret_signed_p1w": 0.07709247938569619, "alpha_spy_p1w": 0.06113746471703374, "alpha_c_p1w": 0.03444577713998487, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.04264670224571132, "ret_p1m": 0.27753301025258814, "ret_signed_p1m": 0.27753301025258814, "alpha_spy_p1m": 0.25961653007221397, "alpha_c_p1m": 0.18465860746272722, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.09287440278986092, "ret_p3m": 0.03069855512362496, "ret_signed_p3m": 0.03069855512362496, "alpha_spy_p3m": 0.0689621381043215, "alpha_c_p3m": -0.01931487653522823, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": 0.05001343165885319, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5303, -0.5457, -0.5215, -0.5215, -0.5264, -0.5518, -0.5595, -0.5154, -0.516, -0.4895, -0.5088, -0.4994, -0.4901, -0.4901, -0.4394, -0.4053, -0.4053, -0.3921, -0.4031, -0.3767, -0.413, -0.3833, -0.3502, -0.3601, -0.2963, -0.2896, -0.2863, -0.293, -0.3359, -0.2709, -0.3238, -0.3458, -0.3612, -0.4251, -0.3987, -0.3667, -0.4097, -0.3733, -0.3612, -0.3656, -0.3767, -0.3855, -0.3634, -0.3194, -0.2522, -0.2148, -0.2566, -0.2467, -0.1784, -0.217, -0.2225, -0.1861, -0.1795, -0.2148, -0.1949, -0.196, -0.1531, -0.1531, -0.1575, -0.1531, -0.0782, -0.0903, -0.0903, 0.0, 0.0518, 0.1454, 0.1729, 0.1949, 0.0771, 0.1233, 0.0991, 0.0991, 0.0925, 0.1729, 0.283, 0.2445, 0.1619, 0.1509, 0.1344, 0.2445, 0.272, 0.3987, 0.4317, 0.4152, 0.2775, 0.1619, 0.1949, 0.1289, 0.0771, 0.1839, 0.1399, 0.1619, 0.1619, 0.1619, 0.1619, 0.1619, 0.1619, 0.1619, 0.1619, 0.2665, 0.3436, 0.2996, 0.3491, 0.3491, 0.3711, 0.239, 0.1454, 0.2335, 0.1729, 0.1123, 0.1619, 0.2555, 0.2004, 0.2004, 0.2885, 0.3601, 0.4097, 0.3436, 0.2115, 0.1839, 0.0958, 0.0815, 0.0573, 0.0229, 0.0296, -0.0014, 0.0307, 0.0163, 0.0163, 0.0163, -0.0103, 0.0584, 0.0074, 0.0351, 0.0451, 0.0351, -0.0014, -0.008, -0.028, -0.0512, 0.0086, 0.003, -0.028, -0.0235, 0.0396, 0.0573, 0.0263, -0.0058, -0.0058, 0.0562, 0.0938, 0.2012, 0.2621, 0.1846, 0.3008, 0.3451, 0.3506, 0.4835, 0.4337, 0.4447, 0.3008, 0.2787, 0.2621, 0.3839, 0.4226, 0.561, 0.716, 0.5942, 0.7492, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027195656604094953", "ticker_idx": 0, "ts": "2026-02-27T01:34:42+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Macquarie raised its target price from KRW 240,000 to KRW 340,000 in just one month… Samsung Electronics is the only company capable of smoothly ramping new fabs over the next three years… stands to be the biggest beneficiary of the upcycle", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Macquarie's bullish thesis on Samsung: PT raised to KRW 340K, biggest upcycle beneficiary via P4/P5 ramp, DRAM/NAND prices expected to double in Q1 2026.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] \"Responding Faster to the Semiconductor Supercycle\"... Samsung Electronics' Pyeongtaek Campus Also in 'Speed Mode'\n\nBuoyed by semiconductor supercycle expectations, target price upgrades, and National Growth Fund support, Samsung Electronics is accelerating its investments. In particular, the Pyeongtaek Campus P4, P5, and P6 fabs are progressing faster than originally planned, taking on the character of an all-out sprint. The convergence of an industry upturn and policy funding is lending significant momentum to these mega-scale semiconductor investment projects.\n\nSamsung Pyeongtaek P5: Preempting the Supercycle\n\nAccording to investment industry sources on the 27th, Fab 6 (P6) is understood to have a first temporary occupancy permit scheduled for May 2029. A temporary occupancy permit marks the stage at which equipment move-in and trial operations become possible — effectively the start of the countdown to fab activation.\n\nTypically, cleanroom interior finishing and equipment setup have taken roughly one year per line, but the recent trend is toward drastically compressed timelines. Sources say Samsung Electronics, as the commissioning party, is pressing hard for schedule acceleration, placing growing strain on contractors. The possibility of workforce expansion and overnight shifts is even being discussed. A source familiar with local conditions said: \"Because the client's demand for schedule compression is so intense, contractors are reviewing options such as adding headcount and expanding weekend shift rotations.\"\n\nFab 5 (P5) has also seen notably accelerated progress. Column installation, originally scheduled for early April, reportedly began in early this month — roughly two months ahead of schedule. At the P6 site, demolition of the existing logistics warehouse (CDC) has also been moved up from the original plan, and workforce deployment timing has been brought forward as well. The source said: \"The P6 CDC was originally supposed to be demolished in May before handing over the site, but the schedule was pulled forward so that personnel will be deployed to the site starting next month.\"\n\nFab 4 (P4), which has already entered the operational phase, has Phase 1 in operation, while Phase 3 is reported to have largely completed equipment move-in and setup. An investment industry source commented: \"Given that yield stabilization typically takes about a year after equipment setup due to the nature of semiconductor processes, some lines appear to be effectively approaching mass production.\"\n\nOn the eastern side of the P5 site, an office building (Office Building 5) was originally planned, but the plan has been changed to the construction of a data center. This is interpreted as an infrastructure investment aimed at advancing semiconductor design automation and process optimization. Data centers are considered core facilities for realizing unmanned, autonomous fab operations. An industry source said: \"Overall, the trend is toward compressing schedules after they've been set. While this is still at a pre-disclosure stage, judging from the design work and on-site progress, it's reasonable to view P6 as also moving into full-scale development.\"\n\nNational Growth Fund Support to Enable Early P5 Activation\n\nMeanwhile, Samsung Electronics' Pyeongtaek P5 has been selected as the first mega-project under the National Growth Fund and will receive KRW 2.5 trillion in low-interest loans. Pyeongtaek P5 is a large-scale semiconductor investment with total spending exceeding KRW 60 trillion. Of the KRW 8.8 trillion in first-phase capex, KRW 6.3 trillion will be self-funded by the company, while the remaining KRW 2.5 trillion will be supplied through KRW 2 trillion from the Advanced Strategic Industries Fund (at government bond-level interest rates) and KRW 500 billion from five major commercial banks (at rates in the 3% range), each with five-year maturities.\n\nWith this financial support, Samsung Electronics plans to bring forward the equipment activation timeline from 2030 to 2028 and strengthen its competitiveness in next-generation High Bandwidth Memory (HBM) and AI semiconductor foundry services. The Financial Services Commission noted: \"Unlike typical semiconductor fabs recently built with two floors, the P5 fab is being constructed as a three-story structure, which will enable more efficient expansion of semiconductor production capacity.\"\n\nIn particular, Samsung Electronics has also agreed to establish a KRW 200 billion special guarantee program in partnership with the Korea Credit Guarantee Fund, and to pilot domestic equipment at the Pyeongtaek P5 process line — expanding its co-prosperity model with materials, parts, and equipment (MPE) companies.\n\nStreet expectations are also rising. Global investment bank Macquarie raised its target price from KRW 240,000 to KRW 340,000 in just one month. Macquarie stated: \"Commodity DRAM and NAND prices are expected to double in Q1 2026,\" forecasting that memory semiconductor price momentum will persist throughout the year.\n\nMacquarie observed that Samsung Electronics stands to be the biggest beneficiary of the upcycle through its Pyeongtaek P4 and P5 lines. It added: \"The company appears to be carefully allocating shipment volumes considering customers' inventory levels and its role within the technology ecosystem,\" and assessed that \"Samsung Electronics is the only company capable of smoothly ramping new fabs over the next three years.\"\n\n---\n\nhttps://t.co/LJVuKIzkyE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006928427511213364, "ret_signed_m1d": 0.006928427511213364, "alpha_spy_m1d": 0.002103314332988271, "alpha_c_m1d": -0.009286633526067067, "bench_spy_m1d": 0.004825113178225093, "bench_c_m1d": 0.01621506103728043, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0005685787823312971, "alpha_c_p1d": -0.005621260796333072, "bench_spy_p1d": 0.0005685787823312971, "bench_c_p1d": 0.005621260796333072, "ret_p1w": -0.13071594184870883, "ret_signed_p1w": -0.13071594184870883, "alpha_spy_p1w": -0.11087599503620715, "alpha_c_p1w": -0.12012214160776147, "bench_spy_p1w": -0.019839946812501674, "bench_c_p1w": -0.01059380024094736, "ret_p1m": -0.18399581120935415, "ret_signed_p1m": -0.18399581120935415, "alpha_spy_p1m": -0.10776423325557938, "alpha_c_p1m": -0.10399102837576302, "bench_spy_p1m": -0.07623157795377478, "bench_c_p1m": -0.08000478283359114, "ret_p3m": 0.4209488710081013, "ret_signed_p3m": 0.4209488710081013, "alpha_spy_p3m": 0.32398018191809586, "alpha_c_p3m": 0.09016682266492815, "bench_spy_p3m": 0.09696868909000544, "bench_c_p3m": 0.33078204834317315, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5247, -0.5197, -0.5169, -0.5017, -0.4967, -0.5017, -0.5036, -0.5068, -0.4994, -0.5183, -0.5275, -0.504, -0.5054, -0.5114, -0.4921, -0.4875, -0.4893, -0.4893, -0.4622, -0.448, -0.4462, -0.4462, -0.4462, -0.4065, -0.3621, -0.3584, -0.3487, -0.3589, -0.358, -0.3589, -0.3644, -0.352, -0.3353, -0.3122, -0.3104, -0.3293, -0.3095, -0.2965, -0.2975, -0.2975, -0.2633, -0.2499, -0.2577, -0.2587, -0.3053, -0.2263, -0.2189, -0.2642, -0.2674, -0.2314, -0.2342, -0.2249, -0.1751, -0.163, -0.163, -0.163, -0.163, -0.1224, -0.1219, -0.1085, -0.0762, -0.06, 0.0069, 0.0, 0.0, -0.0988, -0.2046, -0.115, -0.1307, -0.1986, -0.1321, -0.1224, -0.1321, -0.1524, -0.1284, -0.1044, -0.037, -0.0739, -0.079, -0.1395, -0.1238, -0.127, -0.1681, -0.1681, -0.184, -0.2261, -0.1222, -0.1743, -0.1382, -0.1062, -0.0905, -0.0257, -0.0558, -0.0465, -0.0697, -0.0442, -0.0234, 0.0067, -0.0002, -0.0072, 0.0136, 0.0067, 0.0391, 0.016, 0.0391, 0.0275, 0.046, 0.0206, 0.0206, 0.0761, 0.0761, 0.2312, 0.2566, 0.2428, 0.3214, 0.2914, 0.3145, 0.37, 0.252, 0.3006, 0.2752, 0.2775, 0.3862, 0.3538, 0.3538, 0.3839, 0.4209, 0.3862, 0.4672, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015227653960376540", "ticker_idx": 0, "ts": "2026-01-25T00:58:08+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung Electronics increased the supply price of NAND flash by more than 100% in the first quarter", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "NAND prices up 100%+ in Q1, supply can't keep up with AI-driven demand; price hikes expected to continue into Q2, benefiting Samsung, Hynix, SanDisk, and Micron.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Raises NAND Prices by 100% in Q1... \"Supply Can't Keep Up with Demand\"\n\nSamsung Electronics has raised the price of NAND flash memory by more than double in the first quarter. This is due to a surge in demand for NAND-based storage devices driven by the proliferation of artificial intelligence (AI), while supply has not increased significantly. The memory shortage and price hike trend are expected to continue for some time.\n\nAccording to industry sources on the 25th, Samsung Electronics increased the supply price of NAND flash by more than 100% in the first quarter of this year. It is understood that the company completed supply contracts with major customers at the end of last year and applied the increased prices starting from January. Following DRAM prices that were raised by up to nearly 70%, the upward trend for NAND has also intensified significantly.\n\nNAND price increases are not a recent phenomenon. Since the end of last year, as memory supply shortages became full-blown, NAND has also been affected.\n\nAccording to market research firm TrendForce, the NAND price increase in the fourth quarter of last year was 33-38% compared to the previous quarter. A similar level of increase was forecasted for the first quarter of this year, but the actual supply prices have exceeded that.\n\nThe doubling of NAND prices is not limited to Samsung Electronics. SK Hynix is also observed to be at a similar level. The two companies hold the No. 1 and No. 2 market shares in NAND. Overseas securities firms like Nomura have reported that SanDisk, the No. 5 player in the market, also plans to raise NAND prices by 100% this new year.\n\nAn industry official said, \"Similar to DRAM, NAND manufacturers are successively joining the price hikes,\" adding, \"In fact, price increases across the entire NAND market are inevitable.\"\n\nThe sharp rise in NAND prices is due to an imbalance between demand and supply. Demand continues to grow, but supply is insufficient. First, demand has surged, particularly for enterprise solid-state drives (eSSDs), driven by expanded investments in AI infrastructure. Additionally, the adoption of high-performance, high-capacity storage devices in mobiles and PCs has spread. This is fueled by the strong trend of 'on-device AI,' where AI computations are performed directly on the end device.\n\nOn the other hand, supply has been limited. There were no large-scale expansions of NAND flash production last year, and supply volumes did not increase significantly due to process transitions and other factors. Samsung Electronics has also been conservative in NAND investments. The consensus is that while it did not reach the level of production cuts, shipment expansions were restricted.\n\nNAND prices are expected to rise in the second quarter as well. Samsung Electronics is reportedly in discussions with customers for NAND supply price increases in Q2.\n\nThe market impact from price hikes in memory such as DRAM and NAND is significant. First, the cost of building AI infrastructure is rising sharply. There's even talk that memory is becoming a bottleneck in AI proliferation. Price increases for various devices equipped with memory, such as smartphones and PCs, are also inevitable.\n\nAn industry official stated, \"Not only AI data centers but also smartphone and PC manufacturers are planning to raise final product prices due to memory costs,\" adding, \"Since memory supply cannot be drastically increased immediately, this trend is likely to continue for a while.\"\n\n$SNDK $MU\n\n---\n\nhttps://t.co/PsCPq9p06p", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0050524897745126696, "alpha_c_m1d": 0.006777205222007088, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048652228216711224, "ret_signed_p1d": 0.048652228216711224, "alpha_spy_p1d": 0.04466796348702662, "alpha_c_p1d": 0.03516653004304571, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": -0.011176829852671766, "ret_signed_p1w": -0.011176829852671766, "alpha_spy_p1w": -0.015045622732750008, "alpha_c_p1w": -0.005563340661362881, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.31492441086551537, "ret_signed_p1m": 0.31492441086551537, "alpha_spy_p1m": 0.3226907983264917, "alpha_c_p1m": 0.3543549001418518, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": 0.4790577142361261, "ret_signed_p3m": 0.4790577142361261, "alpha_spy_p3m": 0.453571831310269, "alpha_c_p3m": 0.41091867364976475, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3424, -0.3189, -0.2966, -0.2731, -0.3137, -0.3418, -0.351, -0.3595, -0.3418, -0.3228, -0.3254, -0.3274, -0.364, -0.3418, -0.3601, -0.3686, -0.3418, -0.3797, -0.3673, -0.3503, -0.3274, -0.3228, -0.3424, -0.3405, -0.3235, -0.3163, -0.3124, -0.2908, -0.2836, -0.2908, -0.2934, -0.298, -0.2875, -0.3143, -0.3274, -0.294, -0.296, -0.3045, -0.277, -0.2705, -0.2731, -0.2731, -0.2345, -0.2143, -0.2117, -0.2117, -0.2117, -0.1552, -0.092, -0.0868, -0.073, -0.0874, -0.0861, -0.0874, -0.0953, -0.0776, -0.0539, -0.021, -0.0184, -0.0454, -0.0171, 0.0013, 0.0, 0.0, 0.0487, 0.0677, 0.0565, 0.0552, -0.0112, 0.1012, 0.1118, 0.0473, 0.0427, 0.094, 0.0901, 0.1032, 0.1742, 0.1913, 0.1913, 0.1913, 0.1913, 0.2492, 0.2498, 0.2689, 0.3149, 0.3379, 0.4333, 0.4234, 0.4234, 0.2827, 0.1321, 0.2597, 0.2373, 0.1407, 0.2354, 0.2492, 0.2354, 0.2064, 0.2406, 0.2748, 0.3708, 0.3182, 0.311, 0.2249, 0.2472, 0.2426, 0.1841, 0.1841, 0.1615, 0.1016, 0.2495, 0.1753, 0.2267, 0.2722, 0.2946, 0.3868, 0.344, 0.3572, 0.3242, 0.3605, 0.3901, 0.4329, 0.4231, 0.4132, 0.4428, 0.4329, 0.4791, 0.4461, 0.4791, 0.4626, 0.4889, 0.4527, 0.4527, 0.5318, 0.5318, 0.7525, 0.7887, 0.7689, 0.8809, 0.8381, 0.8711, 0.9501, 0.7821, 0.8513, 0.8151, 0.8184, 0.9732, 0.9271, 0.9271, 0.9699, 1.0226, 0.9732, 1.0885, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015227653960376540", "ticker_idx": 1, "ts": "2026-01-25T00:58:08+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "SK Hynix is also observed to be at a similar level", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "NAND prices up 100%+ in Q1, supply can't keep up with AI-driven demand; price hikes expected to continue into Q2, benefiting Samsung, Hynix, SanDisk, and Micron.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Raises NAND Prices by 100% in Q1... \"Supply Can't Keep Up with Demand\"\n\nSamsung Electronics has raised the price of NAND flash memory by more than double in the first quarter. This is due to a surge in demand for NAND-based storage devices driven by the proliferation of artificial intelligence (AI), while supply has not increased significantly. The memory shortage and price hike trend are expected to continue for some time.\n\nAccording to industry sources on the 25th, Samsung Electronics increased the supply price of NAND flash by more than 100% in the first quarter of this year. It is understood that the company completed supply contracts with major customers at the end of last year and applied the increased prices starting from January. Following DRAM prices that were raised by up to nearly 70%, the upward trend for NAND has also intensified significantly.\n\nNAND price increases are not a recent phenomenon. Since the end of last year, as memory supply shortages became full-blown, NAND has also been affected.\n\nAccording to market research firm TrendForce, the NAND price increase in the fourth quarter of last year was 33-38% compared to the previous quarter. A similar level of increase was forecasted for the first quarter of this year, but the actual supply prices have exceeded that.\n\nThe doubling of NAND prices is not limited to Samsung Electronics. SK Hynix is also observed to be at a similar level. The two companies hold the No. 1 and No. 2 market shares in NAND. Overseas securities firms like Nomura have reported that SanDisk, the No. 5 player in the market, also plans to raise NAND prices by 100% this new year.\n\nAn industry official said, \"Similar to DRAM, NAND manufacturers are successively joining the price hikes,\" adding, \"In fact, price increases across the entire NAND market are inevitable.\"\n\nThe sharp rise in NAND prices is due to an imbalance between demand and supply. Demand continues to grow, but supply is insufficient. First, demand has surged, particularly for enterprise solid-state drives (eSSDs), driven by expanded investments in AI infrastructure. Additionally, the adoption of high-performance, high-capacity storage devices in mobiles and PCs has spread. This is fueled by the strong trend of 'on-device AI,' where AI computations are performed directly on the end device.\n\nOn the other hand, supply has been limited. There were no large-scale expansions of NAND flash production last year, and supply volumes did not increase significantly due to process transitions and other factors. Samsung Electronics has also been conservative in NAND investments. The consensus is that while it did not reach the level of production cuts, shipment expansions were restricted.\n\nNAND prices are expected to rise in the second quarter as well. Samsung Electronics is reportedly in discussions with customers for NAND supply price increases in Q2.\n\nThe market impact from price hikes in memory such as DRAM and NAND is significant. First, the cost of building AI infrastructure is rising sharply. There's even talk that memory is becoming a bottleneck in AI proliferation. Price increases for various devices equipped with memory, such as smartphones and PCs, are also inevitable.\n\nAn industry official stated, \"Not only AI data centers but also smartphone and PC manufacturers are planning to raise final product prices due to memory costs,\" adding, \"Since memory supply cannot be drastically increased immediately, this trend is likely to continue for a while.\"\n\n$SNDK $MU\n\n---\n\nhttps://t.co/PsCPq9p06p", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04211959400443055, "ret_signed_m1d": 0.04211959400443055, "alpha_spy_m1d": 0.04717208377894322, "alpha_c_m1d": 0.038935227647806325, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08695651803959126, "ret_signed_p1d": 0.08695651803959126, "alpha_spy_p1d": 0.08297225330990665, "alpha_c_p1d": 0.06581918159466182, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.1277174044839562, "ret_signed_p1w": 0.1277174044839562, "alpha_spy_p1w": 0.12384861160387794, "alpha_c_p1w": 0.10477473930054693, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.36548918534981745, "ret_signed_p1m": 0.36548918534981745, "alpha_spy_p1m": 0.3732555728107938, "alpha_c_p1m": 0.31448874413736005, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.6674735127329907, "ret_signed_p3m": 0.6674735127329907, "alpha_spy_p3m": 0.6419876298071336, "alpha_c_p3m": 0.4592843616009663, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2424, -0.2288, -0.241, -0.1582, -0.2044, -0.2139, -0.1949, -0.2125, -0.1772, -0.1596, -0.1623, -0.1691, -0.2397, -0.1772, -0.2261, -0.237, -0.2247, -0.2926, -0.294, -0.2953, -0.2886, -0.2609, -0.2799, -0.269, -0.2418, -0.25, -0.2636, -0.2609, -0.216, -0.231, -0.2024, -0.2323, -0.2242, -0.2473, -0.2799, -0.2514, -0.25, -0.2568, -0.212, -0.2065, -0.2011, -0.2011, -0.1861, -0.1304, -0.1155, -0.1155, -0.1155, -0.0802, -0.0543, -0.0136, 0.0082, 0.0272, 0.0109, 0.0177, 0.0027, 0.0082, 0.0177, 0.0272, 0.038, 0.0095, 0.0054, 0.0258, 0.0421, 0.0, 0.087, 0.1427, 0.1698, 0.2351, 0.1277, 0.2323, 0.2228, 0.144, 0.1399, 0.2052, 0.1902, 0.1685, 0.2065, 0.1957, 0.1957, 0.1957, 0.1957, 0.2147, 0.2894, 0.2921, 0.3655, 0.3832, 0.496, 0.4442, 0.4442, 0.2782, 0.1557, 0.2809, 0.2578, 0.138, 0.2768, 0.2999, 0.2659, 0.2387, 0.3258, 0.3204, 0.4374, 0.3789, 0.3707, 0.27, 0.3421, 0.3544, 0.27, 0.27, 0.1883, 0.0985, 0.221, 0.1298, 0.1924, 0.206, 0.2469, 0.4061, 0.3585, 0.398, 0.4157, 0.5014, 0.5463, 0.5722, 0.5354, 0.5872, 0.6661, 0.6648, 0.6675, 0.6634, 0.7587, 0.7696, 0.76, 0.7505, 0.7505, 0.9697, 0.9697, 1.1793, 1.2514, 1.295, 1.5591, 1.4978, 1.6897, 1.6816, 1.476, 1.5046, 1.3753, 1.3753, 1.6407, 1.6421, 1.6421, 1.7932, 2.0532, 2.1163, 2.1762, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015227653960376540", "ticker_idx": 2, "ts": "2026-01-25T00:58:08+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SanDisk, the No. 5 player in the market, also plans to raise NAND prices by 100% this new year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "NAND prices up 100%+ in Q1, supply can't keep up with AI-driven demand; price hikes expected to continue into Q2, benefiting Samsung, Hynix, SanDisk, and Micron.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Raises NAND Prices by 100% in Q1... \"Supply Can't Keep Up with Demand\"\n\nSamsung Electronics has raised the price of NAND flash memory by more than double in the first quarter. This is due to a surge in demand for NAND-based storage devices driven by the proliferation of artificial intelligence (AI), while supply has not increased significantly. The memory shortage and price hike trend are expected to continue for some time.\n\nAccording to industry sources on the 25th, Samsung Electronics increased the supply price of NAND flash by more than 100% in the first quarter of this year. It is understood that the company completed supply contracts with major customers at the end of last year and applied the increased prices starting from January. Following DRAM prices that were raised by up to nearly 70%, the upward trend for NAND has also intensified significantly.\n\nNAND price increases are not a recent phenomenon. Since the end of last year, as memory supply shortages became full-blown, NAND has also been affected.\n\nAccording to market research firm TrendForce, the NAND price increase in the fourth quarter of last year was 33-38% compared to the previous quarter. A similar level of increase was forecasted for the first quarter of this year, but the actual supply prices have exceeded that.\n\nThe doubling of NAND prices is not limited to Samsung Electronics. SK Hynix is also observed to be at a similar level. The two companies hold the No. 1 and No. 2 market shares in NAND. Overseas securities firms like Nomura have reported that SanDisk, the No. 5 player in the market, also plans to raise NAND prices by 100% this new year.\n\nAn industry official said, \"Similar to DRAM, NAND manufacturers are successively joining the price hikes,\" adding, \"In fact, price increases across the entire NAND market are inevitable.\"\n\nThe sharp rise in NAND prices is due to an imbalance between demand and supply. Demand continues to grow, but supply is insufficient. First, demand has surged, particularly for enterprise solid-state drives (eSSDs), driven by expanded investments in AI infrastructure. Additionally, the adoption of high-performance, high-capacity storage devices in mobiles and PCs has spread. This is fueled by the strong trend of 'on-device AI,' where AI computations are performed directly on the end device.\n\nOn the other hand, supply has been limited. There were no large-scale expansions of NAND flash production last year, and supply volumes did not increase significantly due to process transitions and other factors. Samsung Electronics has also been conservative in NAND investments. The consensus is that while it did not reach the level of production cuts, shipment expansions were restricted.\n\nNAND prices are expected to rise in the second quarter as well. Samsung Electronics is reportedly in discussions with customers for NAND supply price increases in Q2.\n\nThe market impact from price hikes in memory such as DRAM and NAND is significant. First, the cost of building AI infrastructure is rising sharply. There's even talk that memory is becoming a bottleneck in AI proliferation. Price increases for various devices equipped with memory, such as smartphones and PCs, are also inevitable.\n\nAn industry official stated, \"Not only AI data centers but also smartphone and PC manufacturers are planning to raise final product prices due to memory costs,\" adding, \"Since memory supply cannot be drastically increased immediately, this trend is likely to continue for a while.\"\n\n$SNDK $MU\n\n---\n\nhttps://t.co/PsCPq9p06p", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006435851439803519, "ret_signed_m1d": 0.006435851439803519, "alpha_spy_m1d": 0.011488341214316189, "alpha_c_m1d": 0.013213056661810607, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022578600591397935, "ret_signed_p1d": 0.022578600591397935, "alpha_spy_p1d": 0.01859433586171333, "alpha_c_p1d": 0.009092902417732418, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": 0.41299916627633815, "ret_signed_p1w": 0.41299916627633815, "alpha_spy_p1w": 0.4091303733962599, "alpha_c_p1w": 0.41861265546764703, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.35624476653987314, "ret_signed_p1m": 0.35624476653987314, "alpha_spy_p1m": 0.3640111540008495, "alpha_c_p1m": 0.39567525581620955, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": 0.9805225506628759, "ret_signed_p3m": 0.9805225506628759, "alpha_spy_p3m": 0.9550366677370188, "alpha_c_p3m": 0.9123835100765145, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5659, -0.5841, -0.5766, -0.5603, -0.5867, -0.5401, -0.5589, -0.4913, -0.4309, -0.4232, -0.3987, -0.4826, -0.4602, -0.4353, -0.4798, -0.4776, -0.5838, -0.5746, -0.5179, -0.5316, -0.5432, -0.5432, -0.5257, -0.5536, -0.5638, -0.5871, -0.5469, -0.5147, -0.5211, -0.5339, -0.5054, -0.4868, -0.5621, -0.5712, -0.5554, -0.5607, -0.5339, -0.4953, -0.488, -0.4798, -0.4688, -0.4688, -0.4689, -0.4812, -0.4898, -0.4958, -0.4958, -0.4154, -0.4178, -0.2574, -0.249, -0.2894, -0.1984, -0.1732, -0.172, -0.1763, -0.1308, -0.1215, -0.1215, -0.0376, 0.0648, 0.0693, 0.0064, 0.0, 0.0226, 0.1207, 0.1455, 0.224, 0.413, 0.4773, 0.2416, 0.2239, 0.2701, 0.2392, 0.1505, 0.273, 0.3388, 0.3308, 0.3308, 0.2544, 0.2753, 0.3192, 0.3806, 0.4157, 0.3562, 0.3432, 0.3847, 0.3495, 0.315, 0.201, 0.2724, 0.2013, 0.1201, 0.2505, 0.3145, 0.3922, 0.3144, 0.4053, 0.4945, 0.5297, 0.6009, 0.64, 0.5075, 0.4921, 0.4921, 0.4398, 0.2812, 0.3081, 0.216, 0.3495, 0.4714, 0.4902, 0.4902, 0.5391, 0.5098, 0.6587, 0.8088, 0.8092, 1.0232, 1.0061, 0.8941, 0.953, 0.9562, 0.9393, 0.9191, 1.0796, 0.9805, 1.1026, 1.2732, 1.129, 1.2604, 1.329, 1.5212, 1.6675, 1.9871, 1.9949, 1.8461, 2.3185, 2.2871, 2.0842, 2.074, 1.937, 1.9898, 1.8314, 1.9382, 1.9579, 2.2758, 2.1408, 2.1408, 2.3763, 2.3771, 2.4869, 2.6002, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015227653960376540", "ticker_idx": 3, "ts": "2026-01-25T00:58:08+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND manufacturers are successively joining the price hikes", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "NAND prices up 100%+ in Q1, supply can't keep up with AI-driven demand; price hikes expected to continue into Q2, benefiting Samsung, Hynix, SanDisk, and Micron.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Raises NAND Prices by 100% in Q1... \"Supply Can't Keep Up with Demand\"\n\nSamsung Electronics has raised the price of NAND flash memory by more than double in the first quarter. This is due to a surge in demand for NAND-based storage devices driven by the proliferation of artificial intelligence (AI), while supply has not increased significantly. The memory shortage and price hike trend are expected to continue for some time.\n\nAccording to industry sources on the 25th, Samsung Electronics increased the supply price of NAND flash by more than 100% in the first quarter of this year. It is understood that the company completed supply contracts with major customers at the end of last year and applied the increased prices starting from January. Following DRAM prices that were raised by up to nearly 70%, the upward trend for NAND has also intensified significantly.\n\nNAND price increases are not a recent phenomenon. Since the end of last year, as memory supply shortages became full-blown, NAND has also been affected.\n\nAccording to market research firm TrendForce, the NAND price increase in the fourth quarter of last year was 33-38% compared to the previous quarter. A similar level of increase was forecasted for the first quarter of this year, but the actual supply prices have exceeded that.\n\nThe doubling of NAND prices is not limited to Samsung Electronics. SK Hynix is also observed to be at a similar level. The two companies hold the No. 1 and No. 2 market shares in NAND. Overseas securities firms like Nomura have reported that SanDisk, the No. 5 player in the market, also plans to raise NAND prices by 100% this new year.\n\nAn industry official said, \"Similar to DRAM, NAND manufacturers are successively joining the price hikes,\" adding, \"In fact, price increases across the entire NAND market are inevitable.\"\n\nThe sharp rise in NAND prices is due to an imbalance between demand and supply. Demand continues to grow, but supply is insufficient. First, demand has surged, particularly for enterprise solid-state drives (eSSDs), driven by expanded investments in AI infrastructure. Additionally, the adoption of high-performance, high-capacity storage devices in mobiles and PCs has spread. This is fueled by the strong trend of 'on-device AI,' where AI computations are performed directly on the end device.\n\nOn the other hand, supply has been limited. There were no large-scale expansions of NAND flash production last year, and supply volumes did not increase significantly due to process transitions and other factors. Samsung Electronics has also been conservative in NAND investments. The consensus is that while it did not reach the level of production cuts, shipment expansions were restricted.\n\nNAND prices are expected to rise in the second quarter as well. Samsung Electronics is reportedly in discussions with customers for NAND supply price increases in Q2.\n\nThe market impact from price hikes in memory such as DRAM and NAND is significant. First, the cost of building AI infrastructure is rising sharply. There's even talk that memory is becoming a bottleneck in AI proliferation. Price increases for various devices equipped with memory, such as smartphones and PCs, are also inevitable.\n\nAn industry official stated, \"Not only AI data centers but also smartphone and PC manufacturers are planning to raise final product prices due to memory costs,\" adding, \"Since memory supply cannot be drastically increased immediately, this trend is likely to continue for a while.\"\n\n$SNDK $MU\n\n---\n\nhttps://t.co/PsCPq9p06p", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02714026854416307, "ret_signed_m1d": 0.02714026854416307, "alpha_spy_m1d": 0.03219275831867574, "alpha_c_m1d": 0.023955902187538847, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05435759085969649, "ret_signed_p1d": 0.05435759085969649, "alpha_spy_p1d": 0.050373326130011886, "alpha_c_p1d": 0.03322025441476706, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.12518954482328515, "ret_signed_p1w": 0.12518954482328515, "alpha_spy_p1w": 0.1213207519432069, "alpha_c_p1w": 0.10224687963987589, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.07432730762902162, "ret_signed_p1m": 0.07432730762902162, "alpha_spy_p1m": 0.08209369508999798, "alpha_c_p1m": 0.02332686641656423, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.2385884502360145, "ret_signed_p3m": 0.2385884502360145, "alpha_spy_p3m": 0.21310256731015742, "alpha_c_p3m": 0.030399299103990085, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4178, -0.4245, -0.4251, -0.397, -0.4399, -0.3898, -0.3877, -0.3888, -0.3493, -0.3806, -0.3708, -0.3913, -0.3659, -0.3784, -0.413, -0.4196, -0.4827, -0.4673, -0.4247, -0.4232, -0.4084, -0.4084, -0.3925, -0.3822, -0.3847, -0.3984, -0.4177, -0.3906, -0.3656, -0.3515, -0.3225, -0.336, -0.3805, -0.3898, -0.4027, -0.4206, -0.3615, -0.3168, -0.2894, -0.2902, -0.2635, -0.2635, -0.2684, -0.2434, -0.2479, -0.2665, -0.2665, -0.1893, -0.1977, -0.1174, -0.1273, -0.1595, -0.1131, -0.1111, -0.131, -0.1433, -0.1348, -0.0677, -0.0677, -0.0619, 0.0001, 0.0218, 0.0271, 0.0, 0.0544, 0.1187, 0.12, 0.0663, 0.1252, 0.078, -0.0249, -0.0159, 0.0144, -0.0144, -0.0407, 0.0546, 0.0639, 0.058, 0.058, 0.0275, 0.0819, 0.0726, 0.1004, 0.0819, 0.0743, 0.1026, 0.068, 0.0598, 0.0606, -0.0242, 0.03, 0.0205, -0.0483, 0.0006, 0.036, 0.0761, 0.0418, 0.0952, 0.1355, 0.1866, 0.1867, 0.1418, 0.0869, 0.0392, 0.0166, -0.018, -0.0864, -0.0819, -0.1726, -0.1314, -0.0542, -0.0583, -0.0583, -0.0287, -0.0292, 0.0458, 0.0838, 0.0814, 0.0968, 0.1973, 0.173, 0.1756, 0.1701, 0.153, 0.1554, 0.2534, 0.2386, 0.2772, 0.3487, 0.2966, 0.3331, 0.3297, 0.3941, 0.4822, 0.6461, 0.7139, 0.6626, 0.9202, 1.0449, 0.971, 1.0663, 0.9953, 0.8632, 0.7524, 0.7966, 0.8821, 0.9595, 0.931, 0.931, 1.3035, 1.3871, 1.3745, 1.4966, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015427379725058467", "ticker_idx": 0, "ts": "2026-01-25T14:11:46+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics has recently passed the final quality tests for HBM4 from NVIDIA and AMD, and will enter mass production next month", "tweet_type": "news_reaction", "timeline": "months", "summary": "Samsung passed NVIDIA/AMD HBM4 quality tests and enters mass production, signaling technology gap vs rivals closed and product leadership recovery.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung's HBM4 to Make Official Debut at March GTC 'Rubin' Announcement\n\nSamsung Electronics' next-generation high-bandwidth memory (HBM), HBM4 (6th generation), will make its official debut alongside the next-generation AI accelerator 'Rubin' at NVIDIA's technology conference 'GTC 2026' in March.\n\nAccording to the semiconductor industry today (25th), Samsung Electronics has recently passed the final quality tests for HBM4 from NVIDIA and AMD, and will enter mass production next month. The HBM4 mass-produced products shipped from Samsung Electronics in February are scheduled to be received by NVIDIA and immediately used in the performance demonstration of the next-generation AI accelerator Rubin, which will be unveiled at the March GTC event.\n\nSamsung's HBM4 is reported to have the industry's highest specifications, operating at 11.7 Gb per second, faster than the 10 Gb required by NVIDIA and AMD. It is said that last year, it passed verification without redesign despite customer requests for performance upgrades, proving its technological completeness.\n\nIn the industry, evaluations are emerging that Samsung's memory technology has normalized starting with this shipment. It is assessed that the technological gap with competitors exposed in HBM3 and HBM3E has been resolved in HBM4, entering a phase of recovering its past product leadership.\n\nHowever, full-scale large-volume supply is expected around June. Since HBM4 is a component integrated into the main body of AI accelerators like NVIDIA's 'Rubin', it is linked to the customer's final finished product mass production schedule. Considering that major customers are currently in the process of producing next-generation chips through foundries, Samsung Electronics' HBM4 shipment volume is also expected to be adjusted according to the customer's actual mass production schedule and required quantities.\n\n---\n\nhttps://t.co/eAMTGoLXVE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0050524897745126696, "alpha_c_m1d": 0.006777205222007088, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048652228216711224, "ret_signed_p1d": 0.048652228216711224, "alpha_spy_p1d": 0.04466796348702662, "alpha_c_p1d": 0.03516653004304571, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": -0.011176829852671766, "ret_signed_p1w": -0.011176829852671766, "alpha_spy_p1w": -0.015045622732750008, "alpha_c_p1w": -0.005563340661362881, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.31492441086551537, "ret_signed_p1m": 0.31492441086551537, "alpha_spy_p1m": 0.3226907983264917, "alpha_c_p1m": 0.3543549001418518, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": 0.4790577142361261, "ret_signed_p3m": 0.4790577142361261, "alpha_spy_p3m": 0.453571831310269, "alpha_c_p3m": 0.41091867364976475, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3424, -0.3189, -0.2966, -0.2731, -0.3137, -0.3418, -0.351, -0.3595, -0.3418, -0.3228, -0.3254, -0.3274, -0.364, -0.3418, -0.3601, -0.3686, -0.3418, -0.3797, -0.3673, -0.3503, -0.3274, -0.3228, -0.3424, -0.3405, -0.3235, -0.3163, -0.3124, -0.2908, -0.2836, -0.2908, -0.2934, -0.298, -0.2875, -0.3143, -0.3274, -0.294, -0.296, -0.3045, -0.277, -0.2705, -0.2731, -0.2731, -0.2345, -0.2143, -0.2117, -0.2117, -0.2117, -0.1552, -0.092, -0.0868, -0.073, -0.0874, -0.0861, -0.0874, -0.0953, -0.0776, -0.0539, -0.021, -0.0184, -0.0454, -0.0171, 0.0013, 0.0, 0.0, 0.0487, 0.0677, 0.0565, 0.0552, -0.0112, 0.1012, 0.1118, 0.0473, 0.0427, 0.094, 0.0901, 0.1032, 0.1742, 0.1913, 0.1913, 0.1913, 0.1913, 0.2492, 0.2498, 0.2689, 0.3149, 0.3379, 0.4333, 0.4234, 0.4234, 0.2827, 0.1321, 0.2597, 0.2373, 0.1407, 0.2354, 0.2492, 0.2354, 0.2064, 0.2406, 0.2748, 0.3708, 0.3182, 0.311, 0.2249, 0.2472, 0.2426, 0.1841, 0.1841, 0.1615, 0.1016, 0.2495, 0.1753, 0.2267, 0.2722, 0.2946, 0.3868, 0.344, 0.3572, 0.3242, 0.3605, 0.3901, 0.4329, 0.4231, 0.4132, 0.4428, 0.4329, 0.4791, 0.4461, 0.4791, 0.4626, 0.4889, 0.4527, 0.4527, 0.5318, 0.5318, 0.7525, 0.7887, 0.7689, 0.8809, 0.8381, 0.8711, 0.9501, 0.7821, 0.8513, 0.8151, 0.8184, 0.9732, 0.9271, 0.9271, 0.9699, 1.0226, 0.9732, 1.0885, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011683318597894307", "ticker_idx": 0, "ts": "2026-01-15T06:14:13+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Revenue: $34.6–35.8bn vs. consensus: $33.2bn (+6% vs. consensus)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "TSMC 1Q26 guidance significantly above consensus on revenue (+6%), gross margin (+4.4pp), and operating margin (+5pp); implicitly bullish.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "es", "tweet_text": "▶ TSMC 1Q26 Guidance\n\n- Revenue: $34.6–35.8bnvs. consensus: $33.2bn (+6% vs. consensus)\n- Gross margin (GPM): 63.0%–65.0%vs. consensus: 59.6% (+4.4pp above consensus)\n- Operating margin (OPM): 54.0%–56.0%vs. consensus: 50.0% (+5.0pp above consensus)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04253021748563435, "ret_signed_m1d": -0.04253021748563435, "alpha_spy_m1d": -0.03981439884552407, "alpha_c_m1d": -0.0221977397879628, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.020332477697671547, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0022245006988907967, "ret_signed_p1d": 0.0022245006988907967, "alpha_spy_p1d": 0.003062374689786207, "alpha_c_p1d": -0.007815636910049495, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.010040137608940292, "ret_p1w": -0.04176915929943947, "ret_signed_p1w": -0.04176915929943947, "alpha_spy_p1w": -0.037059842427113354, "alpha_c_p1w": -0.0579392952770702, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": 0.01617013597763073, "ret_p1m": 0.0723567829310845, "ret_signed_p1m": 0.0723567829310845, "alpha_spy_p1m": 0.08751048157405927, "alpha_c_p1m": 0.043825723033226405, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": 0.02853105989785809, "ret_p3m": 0.11509320992055838, "ret_signed_p3m": 0.11509320992055838, "alpha_spy_p3m": 0.10914639993326314, "alpha_c_p3m": -0.025140376074597848, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.14023358599515623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.131, -0.1403, -0.1567, -0.1513, -0.139, -0.1294, -0.1198, -0.1094, -0.1148, -0.123, -0.1101, -0.1416, -0.1428, -0.1557, -0.1637, -0.1381, -0.15, -0.1516, -0.1762, -0.1686, -0.1768, -0.1887, -0.1757, -0.1899, -0.1971, -0.1691, -0.169, -0.1536, -0.1536, -0.149, -0.1602, -0.1473, -0.1375, -0.1449, -0.1397, -0.1188, -0.1143, -0.0946, -0.1077, -0.1452, -0.1578, -0.1603, -0.1893, -0.1667, -0.1542, -0.1416, -0.1308, -0.1254, -0.1254, -0.1136, -0.1192, -0.1231, -0.1105, -0.1105, -0.0645, -0.0568, -0.0416, -0.0672, -0.0692, -0.0527, -0.0289, -0.0305, -0.0425, 0.0, 0.0022, 0.0022, -0.0424, -0.0454, -0.0418, -0.0198, -0.0261, -0.0097, 0.0019, -0.0061, -0.0324, -0.0008, -0.0172, -0.0465, -0.0319, 0.0211, 0.0403, 0.0593, 0.095, 0.0774, 0.0724, 0.0724, 0.066, 0.0604, 0.0549, 0.0846, 0.0831, 0.1291, 0.1349, 0.1029, 0.0964, 0.0804, 0.0336, 0.0462, 0.0358, -0.008, 0.0207, 0.016, 0.0378, -0.0144, -0.0097, -0.0041, 0.0156, -0.0033, -0.0055, -0.0336, -0.0065, 0.0075, 0.0208, -0.0428, -0.0409, -0.071, -0.008, 0.0024, -0.0048, -0.0048, 0.0032, 0.0136, 0.074, 0.0728, 0.0878, 0.0848, 0.1151, 0.101, 0.0665, 0.0875, 0.075, 0.0804, 0.1373, 0.1232, 0.1813, 0.1887, 0.1516, 0.156, 0.1626, 0.1673, 0.1788, 0.1577, 0.2314, 0.2157, 0.2084, 0.1874, 0.1661, 0.1735, 0.2261, 0.1869, 0.1622, 0.1524, 0.1789, 0.1951, 0.1874, 0.1874, 0.2103, 0.2408, 0.2471, 0.2283, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2023908722062618769", "ticker_idx": 0, "ts": "2026-02-17T23:53:36+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung is emerging as the 'super supplier' that will shape the global memory market this year, as its recovery in HBM technological competitiveness coincides with rising commodity DRAM prices… HBM revenue tripling year-over-year, Samsung is expected to narrow the gap with SK hynix", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish Samsung on HBM4 pricing power, commodity DRAM recovery, ~30% HBM share, and tripling HBM revenue driving a profitability supercycle.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "HBM4, 20% More Expensive Than Its Predecessor… Samsung, the 'Super Supplier,' Devises Maximum Profitability Strategy\n\nSamsung Electronics, which became the first in the world to begin mass production shipments of HBM4, is crunching the numbers to maximize profits.\n\nSamsung is emerging as the \"super supplier\" that will shape the global memory market this year, as its recovery in HBM technological competitiveness coincides with rising commodity DRAM prices.\n\nAccording to industry sources on the 18th, Samsung's recently mass-produced HBM4 products are priced at approximately $700, 20–30% more expensive than HBM3E.\n\nUnlike with previous generations, Samsung is reportedly in a stronger position in price negotiations in the HBM4 market. As commodity DRAM profitability has converged with that of HBM products, Samsung's influence has grown as the company holds competitive advantages in both segments.\n\nAccordingly, Samsung is recalibrating its production capacity (capa) allocation between commodity DRAM and HBM to achieve maximum profitability this year.\n\nAn industry source said, \"With commodity DRAM profitability now exceeding that of HBM, Samsung no longer has a reason to maximize HBM4 volumes at the expense of its commodity DRAM lines. Having proven its HBM competitiveness recovery through best-in-class performance and industry-first mass production shipments, Samsung is now reportedly adjusting capacity carefully from a profitability standpoint.\"\n\nSamsung holds the largest capacity among the three major memory makers and still derives a significant share of revenue from commodity DRAM. With memory prices—both DRAM and NAND—expected to continue rising this year, some forecasts project annual operating profit of around KRW 250 trillion.\n\nThis year's HBM market is expected to remain centered on HBM3E, while all three memory makers are expected to compete fiercely for HBM4 market share.\n\nAccording to industry sources, NVIDIA is the only big tech company that has requested HBM4 supply this year. Google, Broadcom, and others are expected to continue adopting HBM3E as their primary product this year, as they did last year. If AMD—which positions itself as a competitor to NVIDIA—equips its next-generation AI accelerators with HBM4, Samsung is considered the most likely supplier given their long-standing partnership.\n\nSamsung is projected to capture approximately 30% of the global HBM market this year. With HBM revenue tripling year-over-year, Samsung is expected to narrow the gap with SK hynix.\n\nMeanwhile, AI accelerators equipped with Samsung's HBM4 are set to be unveiled at NVIDIA's technology conference \"GTC 2026,\" which opens on March 16. As Samsung was the first among the three memory makers to successfully deliver, attention is focused on whether CEO Jensen Huang will mention Samsung when introducing the \"Rubin\" GPU during his keynote address.\n\n---\n\nhttps://t.co/AesILLnC0P", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0016108673902021087, "alpha_c_m1d": -0.0005735184745021549, "bench_spy_m1d": -0.0016108673902021087, "bench_c_m1d": 0.0005735184745021549, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.005037758541617032, "alpha_c_p1d": -0.010252354700803101, "bench_spy_p1d": 0.005037758541617032, "bench_c_p1d": 0.010252354700803101, "ret_p1w": 0.10375278475147764, "ret_signed_p1w": 0.10375278475147764, "alpha_spy_p1w": 0.09716277492566161, "alpha_c_p1w": 0.09773040263438837, "bench_spy_p1w": 0.0065900098258160344, "bench_c_p1w": 0.0060223821170892755, "ret_p1m": 0.15066224850827536, "ret_signed_p1m": 0.15066224850827536, "alpha_spy_p1m": 0.18203073830015504, "alpha_c_p1m": 0.1615598662597456, "bench_spy_p1m": -0.03136848979187967, "bench_c_p1m": -0.010897617751470245, "ret_p3m": 0.495915525014331, "ret_signed_p3m": 0.495915525014331, "alpha_spy_p3m": 0.41048138536298207, "alpha_c_p3m": 0.2306506341810972, "bench_spy_p3m": 0.08543413965134894, "bench_c_p3m": 0.2652648908332338, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4475, -0.4794, -0.4689, -0.4546, -0.4354, -0.4316, -0.448, -0.4464, -0.4321, -0.4261, -0.4228, -0.4047, -0.3986, -0.4047, -0.4069, -0.4107, -0.4019, -0.4244, -0.4354, -0.4074, -0.4091, -0.4162, -0.3931, -0.3876, -0.3898, -0.3898, -0.3574, -0.3405, -0.3383, -0.3383, -0.3383, -0.2908, -0.2379, -0.2334, -0.2219, -0.234, -0.2329, -0.234, -0.2406, -0.2257, -0.2058, -0.1783, -0.176, -0.1987, -0.1749, -0.1595, -0.1606, -0.1606, -0.1198, -0.1038, -0.1131, -0.1142, -0.17, -0.0756, -0.0668, -0.1209, -0.1247, -0.0817, -0.085, -0.074, -0.0143, 0.0, 0.0, 0.0, 0.0, 0.0486, 0.0491, 0.0651, 0.1038, 0.1231, 0.2031, 0.1948, 0.1948, 0.0767, -0.0497, 0.0574, 0.0386, -0.0425, 0.037, 0.0486, 0.037, 0.0127, 0.0414, 0.0701, 0.1507, 0.1065, 0.1004, 0.0281, 0.0469, 0.043, -0.0061, -0.0061, -0.025, -0.0754, 0.0488, -0.0134, 0.0297, 0.0679, 0.0867, 0.1641, 0.1282, 0.1392, 0.1116, 0.142, 0.1669, 0.2028, 0.1945, 0.1862, 0.2111, 0.2028, 0.2415, 0.2139, 0.2415, 0.2277, 0.2498, 0.2194, 0.2194, 0.2858, 0.2858, 0.471, 0.5014, 0.4849, 0.5789, 0.5429, 0.5706, 0.6369, 0.4959, 0.554, 0.5236, 0.5263, 0.6563, 0.6176, 0.6176, 0.6535, 0.6978, 0.6563, 0.7531, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008038177702117545", "ticker_idx": 0, "ts": "2026-01-05T04:49:43+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "What an insanely bullish report.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Reacts to AMD's report as insanely bullish; implied long.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "What an insanely bullish report.\n\n$AMD https://t.co/8Xedh15823", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010810563460528888, "ret_signed_m1d": 0.010810563460528888, "alpha_spy_m1d": 0.01742661362945419, "alpha_c_m1d": 0.022172115902407974, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03044144957390582, "ret_signed_p1d": -0.03044144957390582, "alpha_spy_p1d": -0.036388632946995125, "alpha_c_p1d": -0.057004650073321006, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": -0.06056630757530401, "ret_signed_p1w": -0.06056630757530401, "alpha_spy_p1w": -0.0713846673459031, "alpha_c_p1w": -0.09518059145723945, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.09512393072697556, "ret_signed_p1m": 0.09512393072697556, "alpha_spy_p1m": 0.09249198082185917, "alpha_c_p1m": 0.04191808170360023, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": -0.016193241366944, "ret_signed_p3m": -0.016193241366944, "alpha_spy_p3m": 0.027572948573427514, "alpha_c_p3m": -0.055203837858347815, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0655, 0.0534, -0.028, -0.0211, -0.0135, 0.0792, 0.061, 0.0543, 0.0881, 0.0767, 0.0414, 0.0629, 0.144, 0.1746, 0.167, 0.1956, 0.1527, 0.1585, 0.1745, 0.131, 0.1594, 0.0752, 0.0564, 0.1036, 0.0744, 0.171, 0.1216, 0.1164, 0.0879, 0.0417, 0.0112, -0.0681, -0.0783, -0.0273, -0.0676, -0.0309, -0.0309, -0.0161, -0.006, -0.0264, -0.0157, -0.0231, -0.0141, 0.0001, 0.0024, 0.0015, 0.0016, -0.0466, -0.0611, -0.0539, -0.1039, -0.0906, -0.0346, -0.0277, -0.028, -0.0273, -0.0273, -0.0275, -0.0247, -0.026, -0.0313, -0.0313, 0.0108, 0.0, -0.0304, -0.05, -0.0742, -0.081, -0.0606, -0.0005, 0.0114, 0.0309, 0.0486, 0.0486, 0.049, 0.1299, 0.1477, 0.1746, 0.1367, 0.14, 0.1432, 0.1407, 0.0708, 0.1139, 0.0951, -0.0945, -0.1293, -0.0572, -0.023, -0.034, -0.0339, -0.0685, -0.0622, -0.0622, -0.0814, -0.0948, -0.0801, -0.0947, -0.1107, -0.0327, -0.0462, -0.0787, -0.0944, -0.1016, -0.1363, -0.086, -0.0978, -0.1296, -0.0832, -0.0807, -0.0735, -0.1056, -0.1252, -0.1108, -0.112, -0.0978, -0.0715, -0.0893, -0.0832, -0.0711, -0.0037, -0.0783, -0.0863, -0.1133, -0.0798, -0.0492, -0.0162, -0.0162, -0.0041, 0.002, 0.0486, 0.0704, 0.1084, 0.1165, 0.1537, 0.1675, 0.2586, 0.2592, 0.2437, 0.2868, 0.3726, 0.3811, 0.5732, 0.5136, 0.462, 0.5248, 0.6034, 0.6308, 0.5449, 0.6069, 0.9061, 0.8476, 1.0589, 1.0752, 1.0277, 1.0151, 1.0341, 0.9183, 0.9042, 0.8729, 1.0245, 1.0336, 1.1147, 1.1147, 1.2792, 1.2415, 1.3435, 1.3344, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2014571187377209534", "ticker_idx": 0, "ts": "2026-01-23T05:29:34+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Raising Target Price to $308", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares GF Securities Buy on AMD with raised PT to $308, driven by CPU/GPU tailwinds.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"To alleviate wafer supply concerns, we maintain our view that server CPUs may adopt Intel's 14A process by 2028.\"\n\nJeff Pu still believes that AMD will use Intel 14A.", "reply_text": null, "parent_text": null, "quoted_text": "《GF Overseas Electronics &amp; Communications》 \nAMD (Buy, AMD US): Performance Outlook, Tailwinds in CPU and GPU\n\nRaising Target Price to $308: Since last November, AMD's stock has been consolidating, but it rebounded 15% in the past week, mainly driven by reports of CPU price", "parent_author": null, "ret_m1d": -0.022912804667516107, "ret_signed_m1d": -0.022912804667516107, "alpha_spy_m1d": -0.022550155989549903, "alpha_c_m1d": -0.029736296910052862, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.032231959924759024, "ret_signed_p1d": -0.032231959924759024, "alpha_spy_p1d": -0.03731010698536885, "alpha_c_p1d": -0.029057701569648042, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": -0.08837799443754601, "ret_signed_p1w": -0.08837799443754601, "alpha_spy_p1w": -0.09235345509194048, "alpha_c_p1w": -0.09680108711177882, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": -0.24291431127319463, "ret_signed_p1m": -0.24291431127319463, "alpha_spy_p1m": -0.232990311212102, "alpha_c_p1m": -0.27488214016667256, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.16859211342460867, "ret_signed_p3m": 0.16859211342460867, "alpha_spy_p3m": 0.1338832428816621, "alpha_c_p3m": -0.023214707382691158, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0064, 0.0179, -0.0186, -0.0137, -0.0001, -0.0371, -0.0129, -0.0846, -0.1007, -0.0605, -0.0853, -0.003, -0.0451, -0.0496, -0.0738, -0.1132, -0.1391, -0.2066, -0.2153, -0.1719, -0.2062, -0.175, -0.175, -0.1623, -0.1537, -0.1711, -0.162, -0.1683, -0.1606, -0.1485, -0.1466, -0.1473, -0.1473, -0.1883, -0.2006, -0.1945, -0.2371, -0.2257, -0.1781, -0.1723, -0.1724, -0.1719, -0.1719, -0.1721, -0.1697, -0.1707, -0.1753, -0.1753, -0.1394, -0.1486, -0.1746, -0.1912, -0.2118, -0.2176, -0.2002, -0.1491, -0.1389, -0.1223, -0.1072, -0.1072, -0.1069, -0.038, -0.0229, 0.0, -0.0322, -0.0295, -0.0267, -0.0289, -0.0884, -0.0516, -0.0677, -0.2291, -0.2587, -0.1973, -0.1682, -0.1776, -0.1775, -0.2069, -0.2016, -0.2016, -0.218, -0.2294, -0.2168, -0.2292, -0.2429, -0.1765, -0.188, -0.2157, -0.229, -0.2351, -0.2647, -0.2218, -0.2319, -0.259, -0.2195, -0.2174, -0.2112, -0.2385, -0.2553, -0.243, -0.244, -0.2319, -0.2095, -0.2247, -0.2195, -0.2091, -0.1518, -0.2153, -0.2222, -0.2451, -0.2166, -0.1905, -0.1624, -0.1624, -0.1521, -0.1469, -0.1073, -0.0887, -0.0564, -0.0495, -0.0178, -0.006, 0.0715, 0.0721, 0.0588, 0.0955, 0.1686, 0.1758, 0.3394, 0.2886, 0.2446, 0.2982, 0.3651, 0.3884, 0.3152, 0.3681, 0.6227, 0.5729, 0.7529, 0.7668, 0.7263, 0.7156, 0.7317, 0.6332, 0.6212, 0.5945, 0.7236, 0.7313, 0.8003, 0.8003, 0.9404, 0.9083, 0.9951, 0.9874, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2036070719927550112", "ticker_idx": 0, "ts": "2026-03-23T13:21:02+00:00", "symbol": "DISCO Corporation", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I'm a big fan of DISCO.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish DISCO Corp as beneficiary of Samsung's next-gen wafer dicing process adoption for HBM4.", "resolved_tickers": ["6146.T"], "resolution_method": "exchange_ticker", "resolution_reason": "DISCO Corp 6146.T Japan dicing/grinding equipment", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "I’m a big fan of DISCO. https://t.co/x5CZdBhrm3", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics to Enhance HBM4 Quality with Next-Generation Wafer Dicing Process\n\nSamsung Electronics is transitioning its wafer dicing process — a critical step directly tied to semiconductor quality and yield — to a next-generation approach. The company plans to expand its https://t.co/WBl2cdKd02", "parent_author": null, "ret_m1d": 0.06554372695079991, "ret_signed_m1d": 0.06554372695079991, "alpha_spy_m1d": 0.07593464139411044, "alpha_c_m1d": 0.08245908714338757, "bench_spy_m1d": -0.010390914443310528, "bench_c_m1d": -0.01691536019258766, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0013782670448649892, "ret_signed_p1d": 0.0013782670448649892, "alpha_spy_p1d": 0.004735116797131456, "alpha_c_p1d": -0.006875031982184421, "bench_spy_p1d": -0.0033568497522664664, "bench_c_p1d": 0.00825329902704941, "ret_p1w": -0.010487282674028209, "ret_signed_p1w": -0.010487282674028209, "alpha_spy_p1w": 0.025232504327824867, "alpha_c_p1w": 0.06317887547028833, "bench_spy_p1w": -0.035719787001853076, "bench_c_p1w": -0.07366615814431654, "ret_p1m": 0.1433684690160335, "ret_signed_p1m": 0.1433684690160335, "alpha_spy_p1m": 0.06906041591066114, "alpha_c_p1m": -0.043927169409291755, "bench_spy_p1m": 0.07430805310537236, "bench_c_p1m": 0.18729563842532526, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2758, -0.2692, -0.2516, -0.2688, -0.2623, -0.2623, -0.2623, -0.2623, -0.2175, -0.17, -0.1475, -0.1487, -0.1577, -0.1577, -0.1205, -0.0812, -0.0949, -0.072, -0.0441, -0.0812, -0.1031, 0.0501, 0.0406, 0.0184, 0.0556, 0.0793, 0.0312, 0.0136, -0.0459, 0.0248, 0.0435, -0.002, 0.0077, 0.1061, 0.1542, 0.1542, 0.1153, 0.1188, 0.1066, 0.0983, 0.1, 0.1343, 0.1342, 0.1342, 0.158, 0.2251, 0.2109, 0.1562, 0.1472, 0.1184, 0.056, 0.0974, 0.1257, 0.0184, 0.0688, 0.1083, 0.091, 0.066, 0.0708, 0.0314, 0.0824, 0.0655, 0.0655, 0.0, 0.0014, 0.041, 0.0335, 0.0188, -0.0105, -0.0568, -0.0193, -0.021, 0.0111, 0.0316, -0.0319, 0.0388, -0.0051, 0.0302, 0.0376, 0.1032, 0.087, 0.1247, 0.1227, 0.1235, 0.1434, 0.1525, 0.1089, 0.1042, 0.1709, 0.1773, 0.1773, 0.1398, 0.1127, 0.1127, 0.1127, 0.1127, 0.1796, 0.1551, 0.1301, 0.1212, 0.1218, 0.0781, -0.0005, -0.022, -0.0513, -0.0553, 0.0111, 0.0143, 0.0747, 0.0305, 0.0416, 0.0219, 0.0024, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014871958274585025", "ticker_idx": 0, "ts": "2026-01-24T01:24:44+00:00", "symbol": "ASML", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy Rating, Target Price Raised to 2059 USD", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities reiterates Buy on ASML, raises TP to $2059 on seller's market dynamics, TSMC-driven demand, memory supercycle, and 2027 EPS of €43.62 vs buy-side €38-40.", "resolved_tickers": ["ASML"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics Communication》 ASML Analysis  \n\n☄️ASML: Customers Scrambling for Orders  \n\n☀️SPE Now Shifts to a Seller's Market: We believe the WFE market has recently turned into a seller's market. Core customers like TSMC are chasing ASML's sales team every day to demand machines, with an enthusiasm unmatched even in the previous industry upcycle before 2024. ASML's EUV capacity is fully booked for 2026-2027, and customers are still applying for additional capacity to meet their own needs. All production factories are operating at full capacity, and the entire industry is gripped by supply anxiety.  \n\n☀️More Optimistic Outlook for 4Q25 Performance: On the other hand, we are increasingly bullish on ASML's order volume for 4Q25. Driven by TSMC's procurement needs and strong demand from the memory industry, we currently predict that the quarter's new order total will reach 10 billion euros. At the same time, ASML is continuously expanding EUV capacity, and we believe this capacity expansion is expected to support the company in raising its full-year performance guidance, with revenue growth of at least 15% year-over-year.  \n\n☀️Maintain Buy Rating, Target Price Raised to 2059 USD: Historical data shows that SPE companies outperform SOX during industry upcycles; and under the dual drivers of TSMC's accelerated expansion and the memory industry's super cycle, we believe the prosperity of this industry upcycle is significantly higher than the 2023-2024 cycle. Clear performance growth expectations should support higher valuation levels. Specifically for ASML, previously when intense competition among foundries like TSMC, Intel, and Samsung triggered EUV supply shortages, the company's Forward P/E reached a peak of 40x. Now, AI-driven demand is tightening the EUV market again, putting ASML into a seller's market, with capacity fully sold out for the next two years. Based on this, combined with our prediction that the company's 2027 EPS will reach 43.62 euros (vs buy-side expectations of 38-40), we are optimistic about ASML's valuation repair to the high level of 40x Forward P/E from the previous cycle, thereby raising the company's new TP to 2059 USD. Text Views\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017200205482699293, "ret_signed_m1d": -0.017200205482699293, "alpha_spy_m1d": -0.012147715708186624, "alpha_c_m1d": -0.020384571839323518, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029178837986148176, "ret_signed_p1d": 0.029178837986148176, "alpha_spy_p1d": 0.02519457325646357, "alpha_c_p1d": 0.008041501541218743, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.01983937715201911, "ret_signed_p1w": 0.01983937715201911, "alpha_spy_p1w": 0.01597058427194087, "alpha_c_p1w": -0.00310328803139015, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.06114947756968436, "ret_signed_p1m": 0.06114947756968436, "alpha_spy_p1m": 0.06891586503066072, "alpha_c_p1m": 0.01014903635722697, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.00447177087378936, "ret_signed_p3m": 0.00447177087378936, "alpha_spy_p3m": -0.021014112052067713, "alpha_c_p3m": -0.20371738025823505, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2423, -0.2391, -0.2506, -0.2452, -0.2711, -0.2615, -0.2718, -0.2805, -0.265, -0.2766, -0.266, -0.2784, -0.2875, -0.2783, -0.2896, -0.2646, -0.3059, -0.3161, -0.3011, -0.2902, -0.2635, -0.2635, -0.25, -0.2302, -0.2155, -0.1928, -0.2146, -0.2221, -0.2078, -0.2136, -0.208, -0.2055, -0.2353, -0.2303, -0.2387, -0.2815, -0.2668, -0.2528, -0.2521, -0.2487, -0.2461, -0.2461, -0.241, -0.2458, -0.2414, -0.243, -0.243, -0.1766, -0.131, -0.1211, -0.1308, -0.155, -0.0987, -0.0935, -0.1013, -0.1059, -0.0578, -0.0388, -0.0388, -0.0618, -0.0377, -0.013, -0.0172, 0.0, 0.0292, 0.0068, 0.0296, 0.0068, 0.0198, -0.0124, -0.0525, -0.0447, -0.0002, 0.0114, 0.0015, 0.0171, -0.0033, -0.0035, -0.0035, 0.0059, 0.0405, 0.0336, 0.0412, 0.0528, 0.0611, 0.0815, 0.0371, 0.0277, 0.0085, -0.0358, -0.0086, -0.0306, -0.0841, -0.0383, -0.0199, -0.0176, -0.0424, -0.0466, -0.0255, -0.0158, -0.0399, -0.032, -0.0668, -0.0297, -0.0085, -0.0125, -0.0581, -0.0772, -0.1116, -0.0642, -0.0366, -0.0668, -0.0668, -0.0761, -0.0744, 0.0068, 0.0263, 0.0473, 0.0628, 0.0757, 0.0498, -0.0005, 0.0342, 0.0461, 0.0336, 0.0228, 0.0045, 0.0327, 0.0171, -0.0169, -0.0102, 0.0217, 0.0132, -0.0158, 0.0245, 0.0968, 0.0768, 0.1304, 0.1118, 0.0799, 0.123, 0.125, 0.0663, 0.0454, 0.0362, 0.1006, 0.1304, 0.1594, 0.1594, 0.1588, 0.1345, 0.1401, 0.1451, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014871958274585025", "ticker_idx": 1, "ts": "2026-01-24T01:24:44+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe this capacity expansion is expected to support the company in raising its full-year performance guidance, with revenue growth of at least 15% year-over-year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities reiterates Buy on ASML, raises TP to $2059 on seller's market dynamics, TSMC-driven demand, memory supercycle, and 2027 EPS of €43.62 vs buy-side €38-40.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics Communication》 ASML Analysis  \n\n☄️ASML: Customers Scrambling for Orders  \n\n☀️SPE Now Shifts to a Seller's Market: We believe the WFE market has recently turned into a seller's market. Core customers like TSMC are chasing ASML's sales team every day to demand machines, with an enthusiasm unmatched even in the previous industry upcycle before 2024. ASML's EUV capacity is fully booked for 2026-2027, and customers are still applying for additional capacity to meet their own needs. All production factories are operating at full capacity, and the entire industry is gripped by supply anxiety.  \n\n☀️More Optimistic Outlook for 4Q25 Performance: On the other hand, we are increasingly bullish on ASML's order volume for 4Q25. Driven by TSMC's procurement needs and strong demand from the memory industry, we currently predict that the quarter's new order total will reach 10 billion euros. At the same time, ASML is continuously expanding EUV capacity, and we believe this capacity expansion is expected to support the company in raising its full-year performance guidance, with revenue growth of at least 15% year-over-year.  \n\n☀️Maintain Buy Rating, Target Price Raised to 2059 USD: Historical data shows that SPE companies outperform SOX during industry upcycles; and under the dual drivers of TSMC's accelerated expansion and the memory industry's super cycle, we believe the prosperity of this industry upcycle is significantly higher than the 2023-2024 cycle. Clear performance growth expectations should support higher valuation levels. Specifically for ASML, previously when intense competition among foundries like TSMC, Intel, and Samsung triggered EUV supply shortages, the company's Forward P/E reached a peak of 40x. Now, AI-driven demand is tightening the EUV market again, putting ASML into a seller's market, with capacity fully sold out for the next two years. Based on this, combined with our prediction that the company's 2027 EPS will reach 43.62 euros (vs buy-side expectations of 38-40), we are optimistic about ASML's valuation repair to the high level of 40x Forward P/E from the previous cycle, thereby raising the company's new TP to 2059 USD. Text Views\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00016971749109284673, "ret_signed_m1d": 0.00016971749109284673, "alpha_spy_m1d": 0.005222207265605516, "alpha_c_m1d": -0.0030146488655313775, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03361623048300855, "ret_signed_p1d": 0.03361623048300855, "alpha_spy_p1d": 0.02963196575332394, "alpha_c_p1d": 0.012478894038079114, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.039728348020690474, "ret_signed_p1w": 0.039728348020690474, "alpha_spy_p1w": 0.03585955514061223, "alpha_c_p1w": 0.016785682837281213, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.0739351579350207, "ret_signed_p1m": 0.0739351579350207, "alpha_spy_p1m": 0.08170154539599706, "alpha_c_p1m": 0.02293471672256331, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.04027369025309957, "ret_signed_p3m": 0.04027369025309957, "alpha_spy_p3m": 0.014787807327242497, "alpha_c_p3m": -0.16791546087892484, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2195, -0.2042, -0.2206, -0.2135, -0.2207, -0.2304, -0.24, -0.2608, -0.2473, -0.2471, -0.243, -0.2517, -0.258, -0.256, -0.2652, -0.2474, -0.2447, -0.292, -0.271, -0.2697, -0.2281, -0.2384, -0.2331, -0.2133, -0.2028, -0.1822, -0.1874, -0.1922, -0.1823, -0.1911, -0.1969, -0.2013, -0.2145, -0.2097, -0.2287, -0.2581, -0.2422, -0.2346, -0.2394, -0.2333, -0.2368, -0.2368, -0.2368, -0.2302, -0.2204, -0.2178, -0.2178, -0.1627, -0.1059, -0.0993, -0.1071, -0.1401, -0.0818, -0.0778, -0.064, -0.0796, -0.0243, -0.0092, -0.0489, -0.0323, -0.0197, -0.0014, 0.0002, 0.0, 0.0336, 0.0139, 0.0119, 0.0319, 0.0397, 0.0105, -0.0317, -0.0241, 0.0134, 0.0241, 0.0141, 0.0267, 0.0029, 0.0119, 0.0158, 0.0194, 0.0581, 0.0525, 0.0673, 0.0619, 0.0739, 0.0952, 0.0476, 0.0484, 0.0289, -0.0124, 0.0199, 0.0081, -0.025, -0.0245, 0.02, 0.019, 0.0034, 0.0027, 0.0178, 0.0148, 0.0124, -0.0066, -0.041, -0.0009, 0.0214, 0.0299, -0.009, -0.0255, -0.0548, -0.0486, 0.0095, -0.0131, -0.0131, -0.0131, -0.0532, 0.0308, 0.0496, 0.0795, 0.0709, 0.0916, 0.0455, 0.0393, 0.0574, 0.0583, 0.0515, 0.0624, 0.0403, 0.0646, 0.033, -0.0017, 0.016, 0.0414, 0.0414, 0.0116, 0.047, 0.1117, 0.1075, 0.124, 0.1123, 0.0784, 0.1305, 0.1646, 0.1131, 0.0772, 0.064, 0.1356, 0.146, 0.2003, 0.2187, 0.184, 0.1726, 0.185, 0.1797, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008482733392359532", "ticker_idx": 0, "ts": "2026-01-06T10:16:14+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA is just universally the right answer, because we're flexible", "tweet_type": "other", "timeline": "unspecified", "summary": "Jensen Huang argues NVIDIA's flexible, unified GPU architecture delivers superior TCO vs specialized chips, implying durable competitive advantage.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Jensen Huang: Why NVIDIA’s Solution Is the Right Answer? — From the CES 2026 Analyst Q&A\n\n\"And so, my point is, because the workloads are changing so fast and because the world is innovating so fast, that's one of the reasons why NVIDIA is just universally the right answer, because we're exible. Does that make sense?\n\nIf your workload is changing from morning to night and depending on what customer you have, we're equally -- we're versatile, we're good at almost everything, and you might be able to take one particular workload and push it to the extreme, but that 10% of the workload or even 5% of the workload or 12% of the workload, if it's not being used, then all of a sudden, that part of the data center could have been used for something else. It's for 90% of the workload, and you deprived it because you only have 1 gigawatt.\n\nThe trick is to think through that one data center, not as innite money and innite space, but you have nite power. And so you have to utilize that nite power for the overall consumption of the data center, and the more flexible it is, the better it is. The more unied architecture it is, like, for example, if we updated a new DeepSeek model, and every single GPU in the data center, all of a sudden, every one of their performance go up. \n\nI uploaded the -- updated the library for a Qwen model, and the whole data center go up. I do it for -- does that make sense? But if you have 17 dierent architectures, one is good for this thing, one's good for that thing, then as it turns out, the overall TCO is not as good. So that's the challenge.\n\nAnd even when I'm building these things, I know what the challenge is. It's very hard. It's very -- I'm exploring ways to beat myself all the time and it's hard. We're exploring all kinds of dierent chips to try to build a better solution than even the one we have. It is extremely hard. But the exercise is worth it. Isn't that right?\n\nIf I'm constantly trying to come up with a new way to even do better than what I currently have, if I'm doing that myself, then I'm exploring all of the nooks and crannies. I'm trying to disrupt myself all the time, okay? So, I think that at the moment, CPX is exactly as you say, but it also reduces the exibility of the data center.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004699804739852764, "ret_signed_m1d": 0.004699804739852764, "alpha_spy_m1d": 0.010611828225168818, "alpha_c_m1d": 0.030575659715457282, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.025875854975604518, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009987125820996923, "ret_signed_p1d": 0.009987125820996923, "alpha_spy_p1d": 0.013210482613224328, "alpha_c_p1d": 0.016694741878650388, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.006707616057653465, "ret_p1w": -0.007637304948690304, "ret_signed_p1w": -0.007637304948690304, "alpha_spy_p1w": -0.010470525517025986, "alpha_c_p1w": -0.01772451332638847, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": 0.010087208377698165, "ret_p1m": -0.06969668297352538, "ret_signed_p1m": -0.06969668297352538, "alpha_spy_p1m": -0.06157309135644695, "alpha_c_p1m": -0.05524952809488559, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.01444715487863979, "ret_p3m": -0.05255504410365941, "ret_signed_p3m": -0.05255504410365941, "alpha_spy_p3m": -0.0031355774207620923, "alpha_c_p3m": -0.06468035308213549, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": 0.01212530897847608, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0284, -0.0218, 0.0057, -0.0386, -0.0396, -0.0291, -0.0215, -0.0246, -0.0325, -0.0372, -0.0272, -0.0053, 0.0226, 0.0736, 0.1057, 0.0835, 0.0814, 0.1048, 0.0611, 0.0425, 0.0044, 0.0048, 0.063, 0.0316, 0.035, -0.0021, 0.0156, -0.0035, -0.0315, -0.0039, -0.0353, -0.0447, -0.0251, -0.0504, -0.0373, -0.0373, -0.0547, -0.0391, -0.0309, -0.0409, -0.0206, -0.0258, -0.009, -0.0121, -0.0185, -0.0337, -0.0653, -0.0585, -0.0508, -0.0871, -0.07, -0.0334, -0.019, 0.0105, 0.0073, 0.0073, 0.0176, 0.0052, 0.0016, -0.004, -0.004, 0.0086, 0.0047, 0.0, 0.01, -0.0117, -0.0127, -0.0123, -0.0076, -0.0219, -0.001, -0.0054, -0.0054, -0.049, -0.0209, -0.0128, 0.0023, -0.0041, 0.0068, 0.0229, 0.0281, 0.0208, -0.0087, -0.0369, -0.0697, -0.082, -0.0098, 0.015, 0.0069, 0.015, -0.0016, -0.0237, -0.0237, -0.0121, 0.004, 0.0035, 0.0138, 0.023, 0.03, 0.0444, -0.0126, -0.0537, -0.0254, -0.0384, -0.0224, -0.0208, -0.0503, -0.0245, -0.0132, -0.0064, -0.0218, -0.0373, -0.0214, -0.0283, -0.0365, -0.0463, -0.0776, -0.0619, -0.0643, -0.0457, -0.0854, -0.1053, -0.1178, -0.0685, -0.0613, -0.0526, -0.0526, -0.0512, -0.0488, -0.0275, -0.0177, 0.0075, 0.0111, 0.0496, 0.0622, 0.0594, 0.0772, 0.0792, 0.0676, 0.0816, 0.0663, 0.1124, 0.1569, 0.1385, 0.1176, 0.0659, 0.0599, 0.0601, 0.0495, 0.11, 0.1296, 0.1494, 0.172, 0.1792, 0.2062, 0.2591, 0.2034, 0.1874, 0.1783, 0.1936, 0.1724, 0.1501, 0.1501, 0.1476, 0.1355, 0.1443, 0.1277, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015938329599254888", "ticker_idx": 0, "ts": "2026-01-27T00:02:06+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If profit growth through 2027 remains highly visible, the potential for long-term appreciation in DRAM stock prices is significant. We expect the risk of multiple de-rating to be extremely limited until the first half of 2027", "tweet_type": "analysis", "timeline": "years", "summary": "Shares Samsung Securities' bullish DRAM call: profit visibility through 2027, bit-growth cap limits competition/demand risk, top-of-band P/B justified for DRAM stocks including Hynix and Micron.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Securities’ January 27 memory report:\n\nLTAs as a Catalyst for Increased DRAM Price Volatility\nMany investors perceive that Long-Term Agreements (LTAs) have enhanced the stability of the DRAM business. However, for LTAs to actually provide industry stability, two prerequisites must be met:\n\n1. Fixed pricing must be maintained even during market booms.\n\n2. Order volumes must be guaranteed even during market downturns.\n\nIn reality, most current LTAs do not fix prices at the time of contract. Similar to how stocks with low trading volumes experience higher price swings, a higher proportion of LTAs in total DRAM volume increases the price volatility of the remaining \"free\" supply. When LTA prices are indexed to rising market rates, overall market volatility intensifies. Notably, TSMC maintains a dual-pricing structure by applying a premium to \"Hot runs\" (urgent orders), separating them from standard LTA pricing.\n\nThe sustainability of LTAs during a slump remains questionable. It is still unclear whether memory suppliers can effectively enforce contract fulfillment through litigation or penalties when customers attempt to cancel orders. The crux of the matter is whether suppliers can increase the sunk cost of switching to a competitor. If enforcement is impossible, LTAs will ironically amplify DRAM price volatility during downturns.\n\nSustainability of DRAM Profits through 2027 via Production Constraints\nDespite skepticism surrounding LTAs, the reason DRAM manufacturers' profits are expected to persist through 2027 is the clear limit on bit growth, capped at a maximum of 20% for the three major players in 2026. When production capacity is inherently limited, companies do not need to fear strategic shifts from competitors and become less vulnerable to demand fluctuations.\n\nIf profit growth through 2027 remains highly visible, the potential for long-term appreciation in DRAM stock prices is significant. Regardless of the cyclical nature of the industry, high visibility for 2027 earnings will prevent a valuation de-rating in 2026. Based on this outlook, it is justifiable to apply a P/B multiple at the top of the historical band to the BPS (Book Value Per Share). We expect the risk of multiple de-rating to be extremely limited until the first half of 2027, when SK Hynix’s Yongin Phase 1 (Y1) and Micron’s Idaho Phase 1 (ID1) fabs are scheduled to open.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.059316726764566195, "ret_signed_m1d": -0.059316726764566195, "alpha_spy_m1d": -0.05534827340363898, "alpha_c_m1d": -0.03861692891120524, "bench_spy_m1d": -0.0039684533609272155, "bench_c_m1d": -0.020699797853360957, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04348977957283114, "ret_signed_p1d": 0.04348977957283114, "alpha_spy_p1d": 0.04359044947648102, "alpha_c_p1d": 0.02045723228538495, "bench_spy_p1d": -0.00010066990364987483, "bench_c_p1d": 0.023032547287446192, "ret_p1w": 0.06877755536907504, "ret_signed_p1w": 0.06877755536907504, "alpha_spy_p1w": 0.07734699291910752, "alpha_c_p1w": 0.09227665254333839, "bench_spy_p1w": -0.008569437550032477, "bench_c_p1w": -0.02349909717426335, "ret_p1m": 0.1980304767721802, "ret_signed_p1m": 0.1980304767721802, "alpha_spy_p1m": 0.20139491071120783, "alpha_c_p1m": 0.15159707305920453, "bench_spy_p1m": -0.0033644339390276334, "bench_c_p1m": 0.04643340371297566, "ret_p3m": 0.37355160594108555, "ret_signed_p3m": 0.37355160594108555, "alpha_spy_p3m": 0.34422005674126616, "alpha_c_p3m": 0.12999113340233476, "bench_spy_p3m": 0.029331549199819396, "bench_c_p3m": 0.2435604725387508, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3651, -0.3619, -0.3202, -0.3608, -0.3568, -0.3532, -0.3617, -0.3327, -0.3312, -0.3298, -0.3389, -0.3642, -0.3419, -0.3737, -0.3818, -0.3895, -0.4175, -0.4005, -0.395, -0.381, -0.3711, -0.3781, -0.3709, -0.3579, -0.3625, -0.3715, -0.3552, -0.3313, -0.3337, -0.3166, -0.3315, -0.3397, -0.3583, -0.3765, -0.3628, -0.3443, -0.335, -0.3039, -0.3004, -0.2911, -0.2911, -0.2758, -0.2444, -0.2404, -0.2463, -0.2463, -0.1931, -0.1678, -0.1282, -0.1203, -0.1292, -0.1191, -0.1168, -0.1302, -0.1268, -0.1137, -0.0791, -0.0749, -0.0904, -0.0631, -0.0441, -0.0378, -0.0593, 0.0, 0.0435, 0.0487, 0.0513, 0.0159, 0.0688, 0.0367, -0.0051, 0.0017, 0.0289, 0.0148, 0.0424, 0.0796, 0.0798, 0.0798, 0.0702, 0.0874, 0.1087, 0.1406, 0.1416, 0.1764, 0.198, 0.252, 0.2304, 0.2307, 0.1082, 0.0399, 0.1158, 0.0799, 0.0279, 0.1118, 0.1359, 0.1103, 0.1096, 0.1599, 0.1853, 0.2517, 0.2029, 0.1807, 0.1074, 0.1294, 0.1208, 0.0547, 0.0561, -0.0048, -0.0384, 0.0706, 0.0178, 0.0533, 0.0813, 0.1008, 0.2027, 0.1864, 0.202, 0.2018, 0.2714, 0.2869, 0.3093, 0.2931, 0.3004, 0.3349, 0.3623, 0.3731, 0.3736, 0.4359, 0.4175, 0.4345, 0.419, 0.4393, 0.5595, 0.6113, 0.7672, 0.7846, 0.8731, 1.0292, 0.9734, 1.0729, 1.073, 0.9149, 0.9106, 0.8734, 0.9014, 1.0565, 1.0333, 1.0333, 1.211, 1.3339, 1.3336, 1.4272, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015938329599254888", "ticker_idx": 1, "ts": "2026-01-27T00:02:06+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the risk of multiple de-rating to be extremely limited until the first half of 2027, when SK Hynix's Yongin Phase 1 (Y1) and Micron's Idaho Phase 1 (ID1) fabs are scheduled to open", "tweet_type": "analysis", "timeline": "years", "summary": "Shares Samsung Securities' bullish DRAM call: profit visibility through 2027, bit-growth cap limits competition/demand risk, top-of-band P/B justified for DRAM stocks including Hynix and Micron.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Securities’ January 27 memory report:\n\nLTAs as a Catalyst for Increased DRAM Price Volatility\nMany investors perceive that Long-Term Agreements (LTAs) have enhanced the stability of the DRAM business. However, for LTAs to actually provide industry stability, two prerequisites must be met:\n\n1. Fixed pricing must be maintained even during market booms.\n\n2. Order volumes must be guaranteed even during market downturns.\n\nIn reality, most current LTAs do not fix prices at the time of contract. Similar to how stocks with low trading volumes experience higher price swings, a higher proportion of LTAs in total DRAM volume increases the price volatility of the remaining \"free\" supply. When LTA prices are indexed to rising market rates, overall market volatility intensifies. Notably, TSMC maintains a dual-pricing structure by applying a premium to \"Hot runs\" (urgent orders), separating them from standard LTA pricing.\n\nThe sustainability of LTAs during a slump remains questionable. It is still unclear whether memory suppliers can effectively enforce contract fulfillment through litigation or penalties when customers attempt to cancel orders. The crux of the matter is whether suppliers can increase the sunk cost of switching to a competitor. If enforcement is impossible, LTAs will ironically amplify DRAM price volatility during downturns.\n\nSustainability of DRAM Profits through 2027 via Production Constraints\nDespite skepticism surrounding LTAs, the reason DRAM manufacturers' profits are expected to persist through 2027 is the clear limit on bit growth, capped at a maximum of 20% for the three major players in 2026. When production capacity is inherently limited, companies do not need to fear strategic shifts from competitors and become less vulnerable to demand fluctuations.\n\nIf profit growth through 2027 remains highly visible, the potential for long-term appreciation in DRAM stock prices is significant. Regardless of the cyclical nature of the industry, high visibility for 2027 earnings will prevent a valuation de-rating in 2026. Based on this outlook, it is justifiable to apply a P/B multiple at the top of the historical band to the BPS (Book Value Per Share). We expect the risk of multiple de-rating to be extremely limited until the first half of 2027, when SK Hynix’s Yongin Phase 1 (Y1) and Micron’s Idaho Phase 1 (ID1) fabs are scheduled to open.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07999999686871007, "ret_signed_m1d": -0.07999999686871007, "alpha_spy_m1d": -0.07603154350778285, "alpha_c_m1d": -0.05930019901534911, "bench_spy_m1d": -0.0039684533609272155, "bench_c_m1d": -0.020699797853360957, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05124997964661504, "ret_signed_p1d": 0.05124997964661504, "alpha_spy_p1d": 0.051350649550264915, "alpha_c_p1d": 0.028217432359168848, "bench_spy_p1d": -0.00010066990364987483, "bench_c_p1d": 0.023032547287446192, "ret_p1w": 0.1337499984343551, "ret_signed_p1w": 0.1337499984343551, "alpha_spy_p1w": 0.14231943598438757, "alpha_c_p1w": 0.15724909560861844, "bench_spy_p1w": -0.008569437550032477, "bench_c_p1w": -0.02349909717426335, "ret_p1m": 0.2725000407067697, "ret_signed_p1m": 0.2725000407067697, "alpha_spy_p1m": 0.27586447464579733, "alpha_c_p1m": 0.22606663699379403, "bench_spy_p1m": -0.0033644339390276334, "bench_c_p1m": 0.04643340371297566, "ret_p3m": 0.5303187152191966, "ret_signed_p3m": 0.5303187152191966, "alpha_spy_p3m": 0.5009871660193772, "alpha_c_p3m": 0.2867582426804458, "bench_spy_p3m": 0.029331549199819396, "bench_c_p3m": 0.2435604725387508, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2905, -0.3018, -0.2256, -0.268, -0.2768, -0.2593, -0.2755, -0.243, -0.2268, -0.2293, -0.2355, -0.3005, -0.243, -0.288, -0.298, -0.2868, -0.3492, -0.3505, -0.3517, -0.3455, -0.32, -0.3375, -0.3275, -0.3025, -0.31, -0.3225, -0.32, -0.2788, -0.2925, -0.2662, -0.2937, -0.2863, -0.3075, -0.3375, -0.3112, -0.31, -0.3163, -0.275, -0.27, -0.265, -0.265, -0.2512, -0.2, -0.1862, -0.1862, -0.1862, -0.1538, -0.13, -0.0925, -0.0725, -0.055, -0.07, -0.0638, -0.0775, -0.0725, -0.0638, -0.055, -0.045, -0.0712, -0.075, -0.0562, -0.0412, -0.08, 0.0, 0.0512, 0.0763, 0.1363, 0.0375, 0.1337, 0.125, 0.0525, 0.0488, 0.1088, 0.095, 0.075, 0.11, 0.1, 0.1, 0.1, 0.1, 0.1175, 0.1862, 0.1888, 0.2563, 0.2725, 0.3763, 0.3287, 0.3287, 0.1759, 0.0632, 0.1784, 0.1571, 0.0469, 0.1747, 0.196, 0.1646, 0.1396, 0.2197, 0.2147, 0.3224, 0.2686, 0.2611, 0.1684, 0.2348, 0.246, 0.1684, 0.1684, 0.0933, 0.0106, 0.1233, 0.0394, 0.097, 0.1095, 0.1471, 0.2936, 0.2498, 0.2861, 0.3024, 0.3813, 0.4226, 0.4464, 0.4126, 0.4602, 0.5328, 0.5316, 0.5341, 0.5303, 0.618, 0.628, 0.6192, 0.6105, 0.6105, 0.8121, 0.8121, 1.0049, 1.0713, 1.1114, 1.3543, 1.298, 1.4746, 1.467, 1.2779, 1.3042, 1.1853, 1.1853, 1.4295, 1.4307, 1.4307, 1.5697, 1.8089, 1.867, 1.9221, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015938329599254888", "ticker_idx": 2, "ts": "2026-01-27T00:02:06+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the risk of multiple de-rating to be extremely limited until the first half of 2027, when SK Hynix's Yongin Phase 1 (Y1) and Micron's Idaho Phase 1 (ID1) fabs are scheduled to open", "tweet_type": "analysis", "timeline": "years", "summary": "Shares Samsung Securities' bullish DRAM call: profit visibility through 2027, bit-growth cap limits competition/demand risk, top-of-band P/B justified for DRAM stocks including Hynix and Micron.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Securities’ January 27 memory report:\n\nLTAs as a Catalyst for Increased DRAM Price Volatility\nMany investors perceive that Long-Term Agreements (LTAs) have enhanced the stability of the DRAM business. However, for LTAs to actually provide industry stability, two prerequisites must be met:\n\n1. Fixed pricing must be maintained even during market booms.\n\n2. Order volumes must be guaranteed even during market downturns.\n\nIn reality, most current LTAs do not fix prices at the time of contract. Similar to how stocks with low trading volumes experience higher price swings, a higher proportion of LTAs in total DRAM volume increases the price volatility of the remaining \"free\" supply. When LTA prices are indexed to rising market rates, overall market volatility intensifies. Notably, TSMC maintains a dual-pricing structure by applying a premium to \"Hot runs\" (urgent orders), separating them from standard LTA pricing.\n\nThe sustainability of LTAs during a slump remains questionable. It is still unclear whether memory suppliers can effectively enforce contract fulfillment through litigation or penalties when customers attempt to cancel orders. The crux of the matter is whether suppliers can increase the sunk cost of switching to a competitor. If enforcement is impossible, LTAs will ironically amplify DRAM price volatility during downturns.\n\nSustainability of DRAM Profits through 2027 via Production Constraints\nDespite skepticism surrounding LTAs, the reason DRAM manufacturers' profits are expected to persist through 2027 is the clear limit on bit growth, capped at a maximum of 20% for the three major players in 2026. When production capacity is inherently limited, companies do not need to fear strategic shifts from competitors and become less vulnerable to demand fluctuations.\n\nIf profit growth through 2027 remains highly visible, the potential for long-term appreciation in DRAM stock prices is significant. Regardless of the cyclical nature of the industry, high visibility for 2027 earnings will prevent a valuation de-rating in 2026. Based on this outlook, it is justifiable to apply a P/B multiple at the top of the historical band to the BPS (Book Value Per Share). We expect the risk of multiple de-rating to be extremely limited until the first half of 2027, when SK Hynix’s Yongin Phase 1 (Y1) and Micron’s Idaho Phase 1 (ID1) fabs are scheduled to open.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05155517571166213, "ret_signed_m1d": -0.05155517571166213, "alpha_spy_m1d": -0.04758672235073491, "alpha_c_m1d": -0.03085537785830117, "bench_spy_m1d": -0.0039684533609272155, "bench_c_m1d": -0.020699797853360957, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06103750519364337, "ret_signed_p1d": 0.06103750519364337, "alpha_spy_p1d": 0.061138175097293246, "alpha_c_p1d": 0.03800495790619718, "bench_spy_p1d": -0.00010066990364987483, "bench_c_p1d": 0.023032547287446192, "ret_p1w": 0.02242597155829551, "ret_signed_p1w": 0.02242597155829551, "alpha_spy_p1w": 0.030995409108327987, "alpha_c_p1w": 0.04592506873255886, "bench_spy_p1w": -0.008569437550032477, "bench_c_p1w": -0.02349909717426335, "ret_p1m": 0.04572936464931909, "ret_signed_p1m": 0.04572936464931909, "alpha_spy_p1m": 0.049093798588346726, "alpha_c_p1m": -0.0007040390636565697, "bench_spy_p1m": -0.0033644339390276334, "bench_c_p1m": 0.04643340371297566, "ret_p3m": 0.2113121292028881, "ret_signed_p3m": 0.2113121292028881, "alpha_spy_p3m": 0.18198058000306871, "alpha_c_p3m": -0.032248343335862684, "bench_spy_p3m": 0.029331549199819396, "bench_c_p3m": 0.2435604725387508, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4542, -0.4548, -0.4281, -0.4687, -0.4213, -0.4193, -0.4203, -0.3828, -0.4125, -0.4033, -0.4226, -0.3986, -0.4105, -0.4432, -0.4495, -0.5093, -0.4947, -0.4544, -0.4529, -0.4389, -0.4389, -0.4238, -0.4141, -0.4165, -0.4294, -0.4477, -0.422, -0.3984, -0.385, -0.3574, -0.3702, -0.4124, -0.4213, -0.4335, -0.4505, -0.3944, -0.3521, -0.3261, -0.3268, -0.3015, -0.3015, -0.3061, -0.2824, -0.2867, -0.3043, -0.3043, -0.2311, -0.2391, -0.1629, -0.1723, -0.2029, -0.1588, -0.1569, -0.1758, -0.1874, -0.1794, -0.1158, -0.1158, -0.1103, -0.0515, -0.0309, -0.0258, -0.0516, 0.0, 0.061, 0.0623, 0.0113, 0.0672, 0.0224, -0.0752, -0.0667, -0.0379, -0.0652, -0.0902, 0.0002, 0.0091, 0.0035, 0.0035, -0.0255, 0.0261, 0.0173, 0.0437, 0.0262, 0.0189, 0.0457, 0.013, 0.0052, 0.0059, -0.0745, -0.0231, -0.0322, -0.0974, -0.051, -0.0174, 0.0206, -0.0119, 0.0387, 0.0769, 0.1254, 0.1255, 0.083, 0.0309, -0.0144, -0.0359, -0.0686, -0.1335, -0.1292, -0.2153, -0.1761, -0.103, -0.1069, -0.1069, -0.0788, -0.0792, -0.0081, 0.0279, 0.0257, 0.0402, 0.1356, 0.1126, 0.115, 0.1097, 0.0935, 0.0959, 0.1888, 0.1747, 0.2113, 0.2792, 0.2298, 0.2643, 0.2612, 0.3222, 0.4057, 0.5612, 0.6256, 0.5769, 0.8212, 0.9395, 0.8694, 0.9597, 0.8924, 0.7672, 0.662, 0.704, 0.785, 0.8585, 0.8314, 0.8314, 1.1847, 1.264, 1.2521, 1.3679, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2013793519597416927", "ticker_idx": 0, "ts": "2026-01-21T01:59:24+00:00", "symbol": "SBUX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bearish Starbucks", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish SBUX on cheap Chinese bubble tea and ice cream flooding US market, eroding competitive position.", "resolved_tickers": ["SBUX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Restaurants'", "bench_c_industry": "Restaurants", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@firstadopter Bearish Starbucks", "reply_text": null, "parent_text": "Wow. China is dumping insanely cheap bubble tea and flooding the U.S. with delicious, affordable ice cream now. The key stronghold Times Square has fallen. Our trade defenses are crumbling! https://t.co/ViaIvc3Laj", "quoted_text": null, "parent_author": "firstadopter", "ret_m1d": -0.028725424619909123, "ret_signed_m1d": 0.028725424619909123, "alpha_spy_m1d": 0.017316022234394435, "alpha_c_m1d": 0.009944537177401958, "bench_spy_m1d": -0.011409402385514689, "bench_c_m1d": -0.018780887442507166, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00622213329849719, "ret_signed_p1d": 0.00622213329849719, "alpha_spy_p1d": 0.011445363318224877, "alpha_c_p1d": 0.01627160358358115, "bench_spy_p1d": 0.005223230019727687, "bench_c_p1d": 0.010049470285083961, "ret_p1w": -0.013170091373666204, "ret_signed_p1w": 0.013170091373666204, "alpha_spy_p1w": 0.027789242050301977, "alpha_c_p1w": 0.016217841281728096, "bench_spy_p1w": 0.014619150676635773, "bench_c_p1w": 0.003047749908061892, "ret_p1m": -0.00012639058680075088, "ret_signed_p1m": 0.00012639058680075088, "alpha_spy_p1m": -0.0012159603301420407, "alpha_c_p1m": -0.04237775292740309, "bench_spy_p1m": -0.0013423509169427916, "bench_c_p1m": -0.04250414351420384, "ret_p3m": 0.03275038011288167, "ret_signed_p3m": -0.03275038011288167, "alpha_spy_p3m": 0.004097433832084274, "alpha_c_p3m": -0.04336980400068191, "bench_spy_p3m": 0.036847813944965946, "bench_c_p3m": -0.010619423887800239, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1136, -0.102, -0.1204, -0.1334, -0.1439, -0.1674, -0.1664, -0.1805, -0.1467, -0.1535, -0.119, -0.129, -0.1102, -0.1016, -0.11, -0.1208, -0.1321, -0.1358, -0.1322, -0.1432, -0.1149, -0.1357, -0.1091, -0.1009, -0.1009, -0.0967, -0.1195, -0.1165, -0.0967, -0.1185, -0.1173, -0.135, -0.1467, -0.1305, -0.1212, -0.1149, -0.1144, -0.1177, -0.1164, -0.0727, -0.084, -0.1064, -0.1304, -0.123, -0.123, -0.1177, -0.1126, -0.1159, -0.1267, -0.1267, -0.1292, -0.1024, -0.0723, -0.101, -0.0856, -0.0783, -0.0671, -0.0609, -0.0548, -0.0327, -0.0357, -0.0357, -0.0287, 0.0, -0.0062, 0.0123, -0.001, -0.0074, -0.0132, -0.0264, -0.0465, -0.0492, -0.0352, 0.0056, -0.0037, 0.0313, 0.0264, 0.0114, 0.0279, 0.0034, -0.0211, -0.0211, -0.0044, -0.0005, -0.0001, 0.0175, -0.0108, 0.0073, 0.0221, 0.0237, 0.023, 0.0099, 0.0091, 0.014, 0.03, 0.0332, 0.0413, 0.0517, 0.0587, 0.0456, 0.0348, 0.021, 0.0183, -0.0329, 0.0002, -0.034, -0.0207, -0.04, -0.0325, -0.0479, -0.094, -0.0949, -0.0649, -0.0562, -0.0568, -0.0568, -0.0108, -0.0063, 0.0146, 0.0116, 0.0082, 0.0174, 0.0277, 0.0264, 0.0266, 0.0437, 0.0328, 0.0207, 0.0387, 0.0389, 0.0298, 0.0217, 0.0153, 0.1011, 0.0993, 0.1053, 0.0956, 0.0953, 0.1109, 0.0882, 0.0952, 0.1036, 0.1124, 0.1058, 0.1105, 0.1214, 0.1191, 0.1168, 0.1181, 0.0932, 0.0825, 0.0825, 0.0647, 0.0719, 0.0577, 0.041, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013069397565530170", "ticker_idx": 1, "ts": "2026-01-19T02:02:00+00:00", "symbol": "Visual Photonics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Taiwan's three leading InP epitaxy companies—IET-KY (4971), Visual Photonics (2455), and LandMark (3081)—are riding this price increase wave", "tweet_type": "analysis", "timeline": "months", "summary": "Long Taiwan InP epitaxy trio (IET, Visual Photonics, LandMark) on AI-driven InP substrate shortage forcing inevitable epitaxy price hikes and capacity lock-in demand.", "resolved_tickers": ["2455.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Visual Photonics Epitaxy listed on TWSE as 2455.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI Drives InP Epitaxy Price Hike; IET, Visual Photonics, LandMark Ride the Price Increase Wave\n\nDriven by AI, demand for optical communications remains strong. A shortage has emerged in the upstream key material indium phosphide (InP) substrates for optical communications, prompting InP epitaxy manufacturers to prepare price increases. Following memory and copper-clad laminates, InP has become the latest AI-critical material to see price hikes. Taiwan’s three leading InP epitaxy companies—IET-KY (4971), Visual Photonics (2455), and LandMark (3081)—are riding this price increase wave.\n\nIET, LandMark, and Visual Photonics are all positive about the outlook and are actively expanding InP epitaxy capacity to capture the shortage-driven price increase opportunity. Among them, IET’s second-phase U.S. plant began construction in July last year and is expected to be completed this year; LandMark has recently been successively expanding epitaxy capacity, with output projected to continue rising in the coming months; Visual Photonics is continuing to purchase new equipment for capacity expansion.\n\nIndustry analysis indicates that demand for optical communications from AI data centers continues to grow, with NVIDIA’s newly launched Rubin platform particularly igniting an optical communications “super cycle.” Combined with constrained substrate capacity and other factors, InP substrate prices have already risen 3% to 5% recently, and InP epitaxy prices are showing signs of movement. To ensure on-time delivery to customers, optical communication chip manufacturers are actively negotiating capacity lock-in agreements with InP epitaxy suppliers, driving intense market demand.\n\nIndustry sources note that InP substrates experienced supply bottlenecks last year due to geopolitical issues. Although related export restrictions were gradually lifted by the end of Q3 last year, the rapid surge in AI market demand has prevented upstream substrate suppliers from expanding capacity in the short term, making the supply-demand imbalance even more severe. To secure sufficient substrates and epitaxy, buyers are willing to pay higher prices, with “highest bidder wins” becoming the market norm.\n\nSubstrates are the raw material for epitaxy, and epitaxy is the raw material for optical communication chips—the supply chain is tightly interlinked, with every link affecting the whole. Now that substrate shortages and price increases are a reality, epitaxy prices are naturally poised to follow. As the substrate price rise continues, an epitaxy price increase is virtually inevitable.\n\n---\n\nhttps://t.co/XnrlzrSzvx", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014880952380952328, "ret_signed_m1d": 0.014880952380952328, "alpha_spy_m1d": 0.014880952380952328, "alpha_c_m1d": 0.014880952380952328, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09821428571428581, "ret_signed_p1d": 0.09821428571428581, "alpha_spy_p1d": 0.11857107143783086, "alpha_c_p1d": 0.12318993354301722, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": 0.0, "ret_signed_p1w": 0.0, "alpha_spy_p1w": -0.0015469911274754722, "alpha_c_p1w": 0.003921195001821642, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": 0.0714285714285714, "ret_signed_p1m": 0.0714285714285714, "alpha_spy_p1m": 0.08416605199827254, "alpha_c_p1m": 0.05364592236965504, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.7261904761904763, "ret_signed_p3m": 0.7261904761904763, "alpha_spy_p3m": 0.7089620050765957, "alpha_c_p3m": 0.5902980434276215, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1161, -0.1429, -0.1429, -0.1369, -0.1131, -0.1429, -0.1548, -0.1756, -0.1399, -0.122, -0.125, -0.0655, -0.0714, -0.0655, -0.0952, -0.1131, -0.0685, -0.1101, -0.1012, -0.1488, -0.1667, -0.1369, -0.1994, -0.1786, -0.1756, -0.1518, -0.1786, -0.1518, -0.1726, -0.1696, -0.1696, -0.1786, -0.1875, -0.1667, -0.1607, -0.1577, -0.1607, -0.1726, -0.1935, -0.2173, -0.2083, -0.2202, -0.2024, -0.125, -0.0893, -0.0804, -0.0804, -0.0685, -0.0744, -0.0833, -0.0982, -0.0982, -0.1042, -0.1161, -0.119, -0.0833, -0.1071, -0.122, -0.0357, -0.0714, -0.0476, 0.003, 0.0149, 0.0, 0.0982, 0.0506, 0.0565, 0.0119, 0.0, 0.0208, 0.0327, 0.0179, 0.0, -0.0685, 0.0238, 0.0804, 0.0149, -0.0208, 0.0298, 0.131, 0.0714, 0.0714, 0.0714, 0.0714, 0.0714, 0.0714, 0.0714, 0.0714, 0.1131, 0.1905, 0.1786, 0.2946, 0.2946, 0.4226, 0.497, 0.3482, 0.2976, 0.2917, 0.1637, 0.1994, 0.3185, 0.3274, 0.4137, 0.3631, 0.2857, 0.3542, 0.3512, 0.369, 0.3185, 0.4435, 0.5863, 0.619, 0.6101, 0.628, 0.4673, 0.5685, 0.5417, 0.5417, 0.5417, 0.6935, 0.7292, 0.7143, 0.7798, 0.744, 0.6696, 0.6369, 0.7262, 0.8482, 0.9821, 1.0238, 1.1696, 1.0595, 1.0595, 0.8542, 0.8929, 0.881, 1.0685, 1.0685, 1.1131, 1.3006, 1.0714, 1.0357, 1.006, 1.1429, 1.2321, 1.2262, 1.1339, 1.1339, 1.3036, 1.2798, 1.3839, 1.3006, 1.5298, 1.6429, 1.5208, 1.4911, 1.3304, 1.5119, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2025763551718338923", "ticker_idx": 0, "ts": "2026-02-23T02:44:02+00:00", "symbol": "industrial humanoids", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Industrial humanoids, on the other hand — I've become far more bullish on them", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish industrial humanoids; skeptical/bearish on non-industrial/consumer humanoids as wage destruction eliminates purchasing power for consumer robots.", "resolved_tickers": ["BOTZ"], "resolution_method": "etf", "resolution_reason": "Industrial humanoids/robotics proxied by BOTZ ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "self_thread", "lang": "en", "tweet_text": "Thoughts after reading Citrini’s piece.\n\nUltimately, this is one enormous game of chicken.\n\nAI frontier labs can never reduce investment. The moment they do, they fall from a galloping horse and suffer irreversible damage. The dilemma Citrini described for incumbents — “sit still and die slower?” — applies equally to the AI labs themselves.\n\nAnd the structure of this game is self-reinforcing. The more I cut to the bone, the more capacity I free up to invest, and more investment translates into stronger intelligence that can overwhelm competitors. Because the other side is running the same calculation, no one can turn the wheel first.\n\nAs Citrini precisely identified, AI investment occurs not as CapEx but as OpEx substitution, and this is how the game of chicken spreads from the labs to the entire economy. AI may not replace every job that relies on intelligence, but it structurally reduces the necessity of human intelligence. For companies, adopting AI is not a choice — it is a condition of survival.\n\nRight now, NVIDIA is hiring aggressively. But someday even NVIDIA will freeze hiring, cut headcount, and pursue maximum efficiency through AI. The reason is simple. Competitors like AMD will aggressively leverage AI to build the capability to challenge NVIDIA. Standing still while competitors ride AI upward means obsolescence.\n\nThis is the most uncomfortable truth of this scenario. Each company’s individual response is rational, but the collective result is catastrophic — in Citrini’s words, “a feedback loop with no natural brake.” We have no choice but to invest in AI as intelligence. Even if it means letting our people go. And that decision to let people go forces someone else to make the same one.\n\nIn a game where the one who turns the wheel loses, no one turns the wheel.\n\n---\n\nIf I were to make one prediction, it would be that no employee is safe from this wave. Not even researchers at AI frontier labs.\n\nEven the frontier labs themselves will have to let people go. To buy more compute.\n\nAnd on non-industrial humanoids… I’ve grown skeptical. If wages are disappearing, where is the purchasing power for such toys? Industrial humanoids, on the other hand — I’ve become far more bullish on them.\n\n---\n\nThe core point is this: excessive efficiency improvement destroys the market.\n\nThe wheel, the steam engine, the internet — these were efficiency improvements within a range that humans could absorb. The pace of progress stayed within the speed at which society could adapt, so we embraced them willingly and built new economies on top of them.\n\nAI is different. The improvement is exponential. And more importantly, it is not a one-time event. The steam engine was invented once, and society adapted over decades. AI gets better every quarter and cheaper every quarter. It gives no room to adapt.\n\nAs a result, we are not wielding AI as a tool — we are being displaced by it.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011404836853968536, "ret_signed_m1d": 0.011404836853968536, "alpha_spy_m1d": 0.0010881754669049837, "alpha_c_m1d": 0.0010881754669049837, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.010316661387063553, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011145579835934338, "ret_signed_p1d": 0.011145579835934338, "alpha_spy_p1d": 0.0038771269055393187, "alpha_c_p1d": 0.0038771269055393187, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.00726845293039502, "ret_p1w": -0.010108947274123037, "ret_signed_p1w": -0.010108947274123037, "alpha_spy_p1w": -0.015956032987014646, "alpha_c_p1w": -0.015956032987014646, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": 0.005847085712891609, "ret_p1m": -0.12338003914761198, "ret_signed_p1m": -0.12338003914761198, "alpha_spy_p1m": -0.08318865682314924, "alpha_c_p1m": -0.08318865682314924, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04019138232446273, "ret_p3m": 0.027734568274772053, "ret_signed_p3m": 0.027734568274772053, "alpha_spy_p3m": -0.06364764856301242, "alpha_c_p3m": -0.06364764856301242, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.09138221683778447, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1, -0.1, -0.0923, -0.0902, -0.0928, -0.0835, -0.0482, -0.051, -0.0479, -0.0482, -0.0443, -0.0469, -0.0611, -0.0673, -0.0763, -0.0967, -0.0874, -0.0719, -0.0564, -0.0531, -0.0533, -0.0533, -0.0557, -0.0515, -0.0518, -0.0609, -0.0609, -0.0485, -0.0251, -0.0168, -0.0109, -0.022, -0.0091, -0.0036, -0.0189, -0.015, -0.0062, -0.006, -0.006, -0.0365, -0.0226, -0.0181, -0.021, -0.0161, -0.0083, -0.0153, -0.0161, -0.0327, -0.0443, -0.0441, -0.0511, -0.0728, -0.0285, -0.0114, -0.0057, -0.0119, -0.0161, -0.0062, -0.0062, 0.001, -0.0057, 0.0016, 0.0114, 0.0, 0.0111, 0.028, 0.0238, 0.0114, -0.0101, -0.0435, -0.0376, -0.0518, -0.0645, -0.0645, -0.0596, -0.0591, -0.0775, -0.0972, -0.0827, -0.0829, -0.0967, -0.0959, -0.1291, -0.1125, -0.1234, -0.1089, -0.1371, -0.1586, -0.1708, -0.1389, -0.1213, -0.1343, -0.1343, -0.1293, -0.1345, -0.0897, -0.0949, -0.0816, -0.0677, -0.0516, -0.048, -0.0485, -0.0303, -0.028, -0.0477, -0.0314, -0.0498, -0.0316, -0.0023, -0.0238, -0.0324, -0.0034, 0.0005, -0.0052, 0.0044, 0.0337, 0.0324, 0.0744, 0.0708, 0.0555, 0.0791, 0.0658, 0.0428, 0.0324, 0.008, 0.0174, 0.0277, 0.0443, 0.0443, 0.0601, 0.0386, 0.0441, 0.0407, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007613325514027183", "ticker_idx": 0, "ts": "2026-01-04T00:41:31+00:00", "symbol": "oil", "kind": "commodity", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Investing in U.S. oil companies", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Venezuela invasion play: long US oil cos, long Korean shipbuilders, short US shale gas, long ENI on Venezuelan gas fields.", "resolved_tickers": ["USO"], "resolution_method": "etf", "resolution_reason": "US oil commodity proxied by USO ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Investment ideas that came to mind while watching the U.S. invasion of Venezuela\n\n1. Investing in U.S. oil companies (this is probably an obvious move that everyone will consider).\n\n2. Investing in South Korean shipbuilders (if oil field development ramps up in Venezuela, won’t they need ships?).\n\n3. Weakness in U.S. shale gas companies (short).\n\n4. Europe’s ENI (ticker: E) holds natural gas fields in Venezuela.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01794363592034487, "ret_signed_m1d": -0.01794363592034487, "alpha_spy_m1d": -0.011327585751419567, "alpha_c_m1d": -0.011327585751419567, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.006616050168925303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0243520229955293, "ret_signed_p1d": -0.0243520229955293, "alpha_spy_p1d": -0.030299206368618603, "alpha_c_p1d": -0.030299206368618603, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.005947183373089304, "ret_p1w": 0.020364572490980892, "ret_signed_p1w": 0.020364572490980892, "alpha_spy_p1w": 0.009546212720381808, "alpha_c_p1w": 0.009546212720381808, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.010818359770599084, "ret_p1m": 0.10324693639940397, "ret_signed_p1m": 0.10324693639940397, "alpha_spy_p1m": 0.10061498649428757, "alpha_c_p1m": 0.10061498649428757, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.002631949905116393, "ret_p3m": 0.9641127281593103, "ret_signed_p3m": 0.9641127281593103, "alpha_spy_p3m": 1.0078789180996819, "alpha_c_p3m": 1.0078789180996819, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.043766189940371514, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0471, 0.0326, -0.0118, 0.0009, -0.0174, -0.0175, -0.0343, -0.0319, -0.034, -0.0278, 0.0058, 0.0437, 0.0422, 0.0413, 0.0167, 0.0225, 0.0212, 0.0333, 0.0359, 0.0244, 0.0115, 0.0094, 0.0148, 0.0208, 0.0356, -0.0061, -0.0053, 0.0165, 0.0154, 0.0325, 0.0094, -0.001, -0.0131, 0.0028, -0.0138, -0.0026, -0.0026, 0.0121, 0.012, -0.0003, 0.0063, 0.0167, 0.0242, 0.0038, -0.0051, 0.0046, -0.0138, -0.0201, -0.0332, -0.0577, -0.0319, -0.0432, -0.0312, -0.007, 0.0011, -0.0003, -0.0003, -0.0248, -0.0087, -0.0068, -0.0151, -0.0151, -0.0179, 0.0, -0.0244, -0.0346, 0.0046, 0.008, 0.0204, 0.0464, 0.034, 0.013, 0.0204, 0.0204, 0.0234, 0.0444, 0.0228, 0.0531, 0.0464, 0.0775, 0.0911, 0.127, 0.1324, 0.0728, 0.1032, 0.1091, 0.0921, 0.0964, 0.1111, 0.1112, 0.1235, 0.0877, 0.0854, 0.0854, 0.0785, 0.1307, 0.1562, 0.1514, 0.1521, 0.1501, 0.1354, 0.136, 0.167, 0.2417, 0.2845, 0.3039, 0.3715, 0.549, 0.4858, 0.5075, 0.5387, 0.686, 0.7073, 0.6381, 0.6924, 0.7327, 0.6713, 0.7293, 0.5745, 0.6312, 0.6148, 0.6699, 0.7687, 0.8489, 0.8122, 0.7672, 0.9641, 0.9641, 0.9786, 0.9664, 0.7741, 0.808, 0.7776, 0.8295, 0.7637, 0.7458, 0.7921, 0.6525, 0.7277, 0.8264, 0.8428, 0.9185, 0.8855, 0.9185, 0.988, 1.1451, 1.0947, 1.0336, 1.1021, 1.0531, 0.9076, 0.9221, 0.9024, 0.9747, 1.055, 1.0228, 1.0365, 1.1109, 1.126, 1.1783, 1.0545, 1.0299, 1.0068, 1.0068, 0.951, 0.866, 0.8624, 0.8384, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007613325514027183", "ticker_idx": 1, "ts": "2026-01-04T00:41:31+00:00", "symbol": "Korean shipbuilders", "kind": "sector", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Investing in South Korean shipbuilders", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Venezuela invasion play: long US oil cos, long Korean shipbuilders, short US shale gas, long ENI on Venezuelan gas fields.", "resolved_tickers": ["009540.KS", "010140.KS", "000720.KS"], "resolution_method": "basket", "resolution_reason": "Korea shipbuilders: HD Hyundai Heavy, Samsung Heavy, Hyundai E&C", "bench_c": "XLI", "bench_c_reason": "industry='Aerospace & Defense'", "bench_c_industry": "Aerospace & Defense", "unit_kind": "original", "lang": "en", "tweet_text": "Investment ideas that came to mind while watching the U.S. invasion of Venezuela\n\n1. Investing in U.S. oil companies (this is probably an obvious move that everyone will consider).\n\n2. Investing in South Korean shipbuilders (if oil field development ramps up in Venezuela, won’t they need ships?).\n\n3. Weakness in U.S. shale gas companies (short).\n\n4. Europe’s ENI (ticker: E) holds natural gas fields in Venezuela.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0328469191370091, "ret_signed_m1d": -0.0328469191370091, "alpha_spy_m1d": -0.026230868968083798, "alpha_c_m1d": -0.02114844491923673, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.01169847421777237, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025487865942071863, "ret_signed_p1d": 0.025487865942071863, "alpha_spy_p1d": 0.01954068256898256, "alpha_c_p1d": 0.011912604535846416, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.013575261406225447, "ret_p1w": 0.16861726764087095, "ret_signed_p1w": 0.16861726764087095, "alpha_spy_p1w": 0.15779890787027187, "alpha_c_p1w": 0.14784778297910242, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.020769484661768534, "ret_p1m": 0.22683128492482627, "ret_signed_p1m": 0.22683128492482627, "alpha_spy_p1m": 0.22419933501970987, "alpha_c_p1m": 0.16996547549597615, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05686580942885011, "ret_p3m": 0.3376809903877698, "ret_signed_p3m": 0.3376809903877698, "alpha_spy_p3m": 0.3814471803281413, "alpha_c_p3m": 0.3102791890702854, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.027401801317484376, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1058, -0.1058, -0.0965, -0.0863, -0.1124, -0.0738, -0.0691, -0.0908, -0.0849, -0.0489, -0.0221, -0.0462, 0.0097, 0.1012, 0.1011, 0.1241, 0.1071, 0.1135, 0.0964, 0.0761, 0.0201, -0.0028, -0.0172, 0.0059, -0.0115, -0.0087, 0.0199, 0.0005, 0.0079, -0.0271, -0.034, -0.0016, -0.0382, -0.0288, -0.0389, -0.0036, -0.0064, -0.022, -0.0463, -0.0258, 0.0059, -0.008, 0.0444, 0.0518, 0.0691, 0.0487, 0.0536, 0.0836, 0.0429, 0.0037, 0.0026, -0.0229, 0.0089, 0.0069, 0.0173, -0.0049, -0.0049, -0.029, -0.0081, -0.0174, -0.0174, -0.0174, -0.0328, 0.0, 0.0255, 0.0296, 0.0499, 0.1036, 0.1686, 0.1936, 0.1693, 0.2268, 0.2399, 0.2897, 0.2735, 0.2614, 0.2317, 0.2688, 0.2385, 0.2446, 0.2391, 0.2459, 0.2074, 0.1523, 0.2268, 0.2606, 0.2186, 0.1839, 0.2197, 0.2017, 0.2221, 0.2336, 0.2229, 0.2229, 0.2229, 0.2229, 0.3051, 0.3558, 0.401, 0.3995, 0.4824, 0.5042, 0.5181, 0.5181, 0.4339, 0.2133, 0.3645, 0.4234, 0.3878, 0.4069, 0.419, 0.4822, 0.5117, 0.458, 0.4787, 0.5232, 0.5026, 0.4709, 0.3589, 0.3569, 0.4037, 0.3636, 0.3568, 0.3002, 0.2684, 0.4073, 0.3377, 0.3825, 0.3709, 0.3901, 0.5705, 0.5457, 0.5361, 0.5026, 0.5166, 0.5345, 0.55, 0.5532, 0.5286, 0.5661, 0.5904, 0.6486, 0.6559, 0.6383, 0.6085, 0.6135, 0.565, 0.565, 0.5595, 0.5595, 0.5591, 0.6138, 0.5614, 0.6155, 0.5283, 0.5414, 0.5486, 0.454, 0.4066, 0.3591, 0.332, 0.4085, 0.4397, 0.4397, 0.4807, 0.4252, 0.3793, 0.4053, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007613325514027183", "ticker_idx": 2, "ts": "2026-01-04T00:41:31+00:00", "symbol": "US shale gas", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Weakness in U.S. shale gas companies (short)", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Venezuela invasion play: long US oil cos, long Korean shipbuilders, short US shale gas, long ENI on Venezuelan gas fields.", "resolved_tickers": ["XOP"], "resolution_method": "etf", "resolution_reason": "US shale gas companies proxied by XOP ETF (oil & gas E&P)", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "Investment ideas that came to mind while watching the U.S. invasion of Venezuela\n\n1. Investing in U.S. oil companies (this is probably an obvious move that everyone will consider).\n\n2. Investing in South Korean shipbuilders (if oil field development ramps up in Venezuela, won’t they need ships?).\n\n3. Weakness in U.S. shale gas companies (short).\n\n4. Europe’s ENI (ticker: E) holds natural gas fields in Venezuela.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011451034750586064, "ret_signed_m1d": -0.011451034750586064, "alpha_spy_m1d": -0.018067084919511367, "alpha_c_m1d": -0.018067084919511367, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.006616050168925303, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015372503482172006, "ret_signed_p1d": 0.015372503482172006, "alpha_spy_p1d": 0.02131968685526131, "alpha_c_p1d": 0.02131968685526131, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.005947183373089304, "ret_p1w": 0.0003137319174584263, "ret_signed_p1w": -0.0003137319174584263, "alpha_spy_p1w": 0.010504627853140658, "alpha_c_p1w": 0.010504627853140658, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.010818359770599084, "ret_p1m": 0.09631377613551484, "ret_signed_p1m": -0.09631377613551484, "alpha_spy_p1m": -0.09368182623039845, "alpha_c_p1m": -0.09368182623039845, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.002631949905116393, "ret_p3m": 0.3994844270733702, "ret_signed_p3m": -0.3994844270733702, "alpha_spy_p3m": -0.4432506170137417, "alpha_c_p3m": -0.4432506170137417, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.043766189940371514, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0427, 0.02, -0.0348, -0.0095, -0.0178, -0.0195, -0.04, -0.0378, -0.0246, -0.0319, -0.0275, 0.0096, 0.0002, -0.0006, -0.0203, -0.0161, -0.02, -0.0125, -0.006, -0.0195, -0.0287, -0.0124, 0.0082, 0.0232, 0.0444, 0.0282, 0.0272, 0.0434, 0.0234, 0.0394, 0.0269, 0.001, 0.0102, 0.0187, 0.0139, 0.0253, 0.0253, 0.0435, 0.0518, 0.036, 0.0627, 0.0627, 0.0643, 0.0444, 0.0476, 0.0529, 0.0357, 0.0257, 0.0113, -0.027, -0.0041, -0.026, -0.0224, -0.0176, -0.008, -0.0124, -0.0124, -0.0199, -0.0082, -0.0005, -0.0097, -0.0097, 0.0115, 0.0, -0.0154, -0.0278, 0.0075, -0.0007, 0.0003, 0.0176, 0.0348, 0.0218, 0.0219, 0.0219, 0.0177, 0.0559, 0.0551, 0.0605, 0.0572, 0.062, 0.0753, 0.09, 0.0999, 0.0648, 0.0963, 0.1279, 0.1095, 0.1483, 0.1489, 0.1421, 0.1691, 0.1344, 0.1599, 0.1599, 0.1404, 0.1612, 0.1871, 0.1911, 0.1725, 0.173, 0.1638, 0.1748, 0.2056, 0.2515, 0.247, 0.2609, 0.2849, 0.2873, 0.2692, 0.2471, 0.2917, 0.3075, 0.3168, 0.3127, 0.334, 0.3597, 0.3729, 0.3892, 0.377, 0.4193, 0.4302, 0.4593, 0.4819, 0.4606, 0.4318, 0.3769, 0.3995, 0.3995, 0.41, 0.4201, 0.352, 0.3211, 0.3266, 0.3284, 0.2869, 0.2935, 0.3215, 0.2575, 0.2687, 0.2993, 0.325, 0.3352, 0.3233, 0.3346, 0.3517, 0.3995, 0.4033, 0.3912, 0.4254, 0.4221, 0.3334, 0.3077, 0.3006, 0.3281, 0.3366, 0.3256, 0.3325, 0.3712, 0.3876, 0.4061, 0.3759, 0.3438, 0.354, 0.354, 0.308, 0.2864, 0.299, 0.2914, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007613325514027183", "ticker_idx": 3, "ts": "2026-01-04T00:41:31+00:00", "symbol": "E", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Europe's ENI (ticker: E) holds natural gas fields in Venezuela", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Venezuela invasion play: long US oil cos, long Korean shipbuilders, short US shale gas, long ENI on Venezuelan gas fields.", "resolved_tickers": ["E"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLE", "bench_c_reason": "industry='Oil & Gas Integrated'", "bench_c_industry": "Oil & Gas Integrated", "unit_kind": "original", "lang": "en", "tweet_text": "Investment ideas that came to mind while watching the U.S. invasion of Venezuela\n\n1. Investing in U.S. oil companies (this is probably an obvious move that everyone will consider).\n\n2. Investing in South Korean shipbuilders (if oil field development ramps up in Venezuela, won’t they need ships?).\n\n3. Weakness in U.S. shale gas companies (short).\n\n4. Europe’s ENI (ticker: E) holds natural gas fields in Venezuela.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005350272947642676, "ret_signed_m1d": -0.005350272947642676, "alpha_spy_m1d": 0.0012657772212826268, "alpha_c_m1d": 0.02109455765167645, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.026444830599319125, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03566878461561862, "ret_signed_p1d": -0.03566878461561862, "alpha_spy_p1d": -0.041615967988707925, "alpha_c_p1d": -0.009010641706852618, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": -0.026658142908766003, "ret_p1w": -0.040254818084455524, "ret_signed_p1w": -0.040254818084455524, "alpha_spy_p1w": -0.05107317785505461, "alpha_c_p1w": -0.028525261408030023, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": -0.011729556676425501, "ret_p1m": 0.05859865366180195, "ret_signed_p1m": 0.05859865366180195, "alpha_spy_p1m": 0.055966703756685554, "alpha_c_p1m": -0.043342029794112946, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.1019406834559149, "ret_p3m": 0.47169554671908487, "ret_signed_p3m": 0.47169554671908487, "alpha_spy_p3m": 0.5154617366594564, "alpha_c_p3m": 0.19984387058792974, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.27185167613115513, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0945, -0.0978, -0.1236, -0.1187, -0.133, -0.1261, -0.1287, -0.1134, -0.1134, -0.1187, -0.0973, -0.0777, -0.0622, -0.0558, -0.0543, -0.0538, -0.0657, -0.0596, -0.0668, -0.0746, -0.0701, -0.0619, -0.0558, -0.0413, -0.027, -0.0318, -0.0298, -0.0262, -0.0268, -0.0321, -0.0443, -0.0489, -0.0497, -0.0586, -0.0693, -0.0545, -0.0545, -0.0466, -0.0428, -0.0423, -0.025, -0.0301, -0.052, -0.051, -0.0456, -0.0418, -0.0474, -0.042, -0.0438, -0.0701, -0.0594, -0.0662, -0.051, -0.0375, -0.0364, -0.0362, -0.0362, -0.0377, -0.0382, -0.0288, -0.0334, -0.0334, -0.0054, 0.0, -0.0357, -0.0637, -0.0538, -0.0492, -0.0403, -0.0206, -0.0061, -0.027, -0.014, -0.014, -0.026, -0.0064, -0.012, 0.0186, 0.0196, 0.0443, 0.0413, 0.0619, 0.0425, 0.0395, 0.0586, 0.0662, 0.0571, 0.0782, 0.0866, 0.0917, 0.1139, 0.0925, 0.107, 0.107, 0.1011, 0.1167, 0.1172, 0.1261, 0.1241, 0.133, 0.1325, 0.1603, 0.1959, 0.2066, 0.1781, 0.1921, 0.1885, 0.2168, 0.2313, 0.2321, 0.2576, 0.2805, 0.3085, 0.3144, 0.3682, 0.3501, 0.3944, 0.3896, 0.3608, 0.3498, 0.3882, 0.3944, 0.4232, 0.4374, 0.4591, 0.4145, 0.4717, 0.4717, 0.4848, 0.4833, 0.4256, 0.4351, 0.4544, 0.4436, 0.4276, 0.3975, 0.4276, 0.3467, 0.3544, 0.3683, 0.3977, 0.4085, 0.3946, 0.3859, 0.4122, 0.4289, 0.4596, 0.4475, 0.4482, 0.4562, 0.3895, 0.3555, 0.3771, 0.4294, 0.439, 0.431, 0.4261, 0.4302, 0.4534, 0.4532, 0.418, 0.4373, 0.4177, 0.4177, 0.3963, 0.3674, 0.3635, 0.3596, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2005212554654568502", "ticker_idx": 0, "ts": "2025-12-28T09:41:42+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "expect AMD, given its higher OpenAI exposure, to be more volatile", "tweet_type": "analysis", "timeline": "quarters", "summary": "Author endorses Citi view: favor lower-OpenAI-exposure names (Broadcom benefits from Google doubling); AMD faces volatility risk from high OpenAI exposure amid unsustainable capex.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Citi argues that as OpenAI’s bills start to come due in 2H26 and concerns rise over the debt burden required to fund the AI infrastructure build-out, investors should favor stocks with lower OpenAI exposure (they also expect AMD, given its higher OpenAI exposure, to be more volatile).\n\nWhat’s especially interesting is Citi’s view that OpenAI’s 2029 capex could exceed that of the four major CSPs combined.\n\nIn my view as well, OpenAI’s level of capex is not sustainable.", "reply_text": null, "parent_text": null, "quoted_text": "I’m only just getting around to reading Citi’s 2026 semiconductor outlook… and this table is genuinely fascinating.\n\n1. Broadcom’s revenue from Google in 2026 is expected to roughly double versus 2025, and in 2027 it could grow again to nearly double compared with 2026.\n\n2. https://t.co/91MKDiO5Y7", "parent_author": null, "ret_m1d": -0.0028755397033181085, "ret_signed_m1d": 0.0028755397033181085, "alpha_spy_m1d": 0.006451912774327773, "alpha_c_m1d": 0.007405876418937418, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.00453033671561931, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001252280838999198, "ret_signed_p1d": 0.001252280838999198, "alpha_spy_p1d": 3.118010441693908e-05, "alpha_c_p1d": -0.0012188042975966873, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0024710851365958852, "ret_p1w": 0.025369886393110663, "ret_signed_p1w": -0.025369886393110663, "alpha_spy_p1w": -0.02555887050563277, "alpha_c_p1w": 0.01136717518325403, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": 0.036737061576364693, "ret_p1m": 0.16891608954059234, "ret_signed_p1m": -0.16891608954059234, "alpha_spy_p1m": -0.15780900210251936, "alpha_c_p1m": -0.05074248213083932, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.11817360740975302, "ret_p3m": -0.05491394788912296, "ret_signed_p3m": 0.05491394788912296, "alpha_spy_p3m": -0.004689452401293259, "alpha_c_p3m": 0.10057443297458946, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": 0.045660485085466496, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2393, -0.2128, -0.2363, -0.0552, -0.019, 0.0925, 0.0801, -0.0033, 0.0038, 0.0115, 0.1066, 0.0879, 0.081, 0.1157, 0.104, 0.0678, 0.0899, 0.173, 0.2044, 0.1967, 0.226, 0.1819, 0.1879, 0.2043, 0.1597, 0.1889, 0.1025, 0.0832, 0.1316, 0.1016, 0.2007, 0.15, 0.1447, 0.1155, 0.0681, 0.0368, -0.0445, -0.0549, -0.0026, -0.044, -0.0064, -0.0064, 0.0089, 0.0192, -0.0017, 0.0092, 0.0017, 0.0109, 0.0255, 0.0279, 0.0269, 0.027, -0.0224, -0.0372, -0.0299, -0.0812, -0.0675, -0.0101, -0.0031, -0.0033, -0.0026, -0.0026, -0.0029, 0.0, -0.0013, -0.0067, -0.0067, 0.0365, 0.0254, -0.0058, -0.0259, -0.0507, -0.0577, -0.0367, 0.0249, 0.0371, 0.0571, 0.0752, 0.0752, 0.0756, 0.1586, 0.1768, 0.2044, 0.1656, 0.1689, 0.1722, 0.1696, 0.098, 0.1422, 0.1229, -0.0715, -0.1072, -0.0333, 0.0018, -0.0095, -0.0094, -0.0448, -0.0384, -0.0384, -0.0581, -0.0718, -0.0568, -0.0717, -0.0882, -0.0082, -0.022, -0.0553, -0.0714, -0.0788, -0.1144, -0.0628, -0.075, -0.1075, -0.06, -0.0574, -0.05, -0.0829, -0.1031, -0.0883, -0.0895, -0.0749, -0.048, -0.0662, -0.06, -0.0475, 0.0216, -0.0549, -0.0632, -0.0908, -0.0565, -0.025, 0.0088, 0.0088, 0.0212, 0.0275, 0.0752, 0.0975, 0.1365, 0.1448, 0.183, 0.1972, 0.2906, 0.2912, 0.2752, 0.3195, 0.4074, 0.4161, 0.6131, 0.552, 0.499, 0.5635, 0.6441, 0.6722, 0.5841, 0.6477, 0.9544, 0.8944, 1.1112, 1.1279, 1.0792, 1.0662, 1.0857, 0.967, 0.9526, 0.9204, 1.0759, 1.0852, 1.1683, 1.1683, 1.337, 1.2983, 1.4029, 1.3937, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2005212554654568502", "ticker_idx": 1, "ts": "2025-12-28T09:41:42+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom's revenue from Google in 2026 is expected to roughly double versus 2025, and in 2027 it could grow again to nearly double compared with 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Author endorses Citi view: favor lower-OpenAI-exposure names (Broadcom benefits from Google doubling); AMD faces volatility risk from high OpenAI exposure amid unsustainable capex.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Citi argues that as OpenAI’s bills start to come due in 2H26 and concerns rise over the debt burden required to fund the AI infrastructure build-out, investors should favor stocks with lower OpenAI exposure (they also expect AMD, given its higher OpenAI exposure, to be more volatile).\n\nWhat’s especially interesting is Citi’s view that OpenAI’s 2029 capex could exceed that of the four major CSPs combined.\n\nIn my view as well, OpenAI’s level of capex is not sustainable.", "reply_text": null, "parent_text": null, "quoted_text": "I’m only just getting around to reading Citi’s 2026 semiconductor outlook… and this table is genuinely fascinating.\n\n1. Broadcom’s revenue from Google in 2026 is expected to roughly double versus 2025, and in 2027 it could grow again to nearly double compared with 2026.\n\n2. https://t.co/91MKDiO5Y7", "parent_author": null, "ret_m1d": 0.007842206363219306, "ret_signed_m1d": 0.007842206363219306, "alpha_spy_m1d": 0.004265833292209642, "alpha_c_m1d": 0.0033118696475999965, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.00453033671561931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0013165166515154425, "ret_signed_p1d": 0.0013165166515154425, "alpha_spy_p1d": 0.0025376173860977014, "alpha_c_p1d": 0.0037876017881113277, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0024710851365958852, "ret_p1w": -0.017086882396489145, "ret_signed_p1w": -0.017086882396489145, "alpha_spy_p1w": -0.01689789828396704, "alpha_c_p1w": -0.05382394397285384, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": 0.036737061576364693, "ret_p1m": -0.047511362516396316, "ret_signed_p1m": -0.047511362516396316, "alpha_spy_p1m": -0.0586184499544693, "alpha_c_p1m": -0.16568496992614934, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.11817360740975302, "ret_p3m": -0.11254164826379276, "ret_signed_p3m": -0.11254164826379276, "alpha_spy_p3m": -0.05293824797337654, "alpha_c_p3m": -0.15820213334925926, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": 0.045660485085466496, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0476, -0.0339, -0.0334, -0.0416, -0.039, -0.013, -0.0144, -0.0726, 0.019, -0.0169, 0.0036, 0.0117, -0.0021, -0.0023, -0.0211, -0.0279, -0.0165, 0.0116, 0.0343, 0.0655, 0.1026, 0.0754, 0.0559, 0.0357, 0.0054, 0.0255, 0.0158, -0.0018, 0.0238, 0.0054, 0.0147, -0.0288, -0.0217, -0.0212, -0.0273, 0.0125, -0.0093, -0.0282, 0.0797, 0.0999, 0.1357, 0.1357, 0.1511, 0.1029, 0.09, 0.0873, 0.0885, 0.1148, 0.1458, 0.1606, 0.1797, 0.1609, 0.0282, -0.0293, -0.025, -0.0687, -0.0576, -0.0277, -0.0227, -0.0002, 0.0024, 0.0024, 0.0078, 0.0, 0.0013, -0.0094, -0.0094, -0.0051, -0.0171, -0.0161, -0.0169, -0.0484, -0.0127, 0.0081, 0.0149, -0.0272, -0.0182, 0.0066, 0.0066, -0.0481, -0.0589, -0.0684, -0.084, -0.0702, -0.0475, -0.0462, -0.0534, -0.0518, -0.0523, -0.0832, -0.1183, -0.1113, -0.0471, -0.0156, -0.0256, -0.019, -0.0521, -0.0693, -0.0693, -0.0482, -0.0455, -0.0441, -0.0479, -0.0545, -0.0684, -0.0489, -0.0793, -0.0854, -0.0875, -0.1017, -0.0912, -0.0476, -0.0541, -0.0104, -0.0195, -0.0224, -0.0384, -0.0779, -0.07, -0.0804, -0.0958, -0.0846, -0.1113, -0.075, -0.0871, -0.0856, -0.1125, -0.1376, -0.1585, -0.1123, -0.1009, -0.0978, -0.0978, -0.0982, -0.0421, 0.0057, 0.0179, 0.0657, 0.0892, 0.0921, 0.1378, 0.1429, 0.166, 0.1462, 0.1535, 0.2122, 0.2044, 0.2125, 0.1995, 0.1468, 0.1629, 0.1972, 0.2083, 0.1946, 0.2257, 0.2202, 0.1833, 0.2333, 0.2288, 0.2026, 0.1954, 0.2614, 0.2195, 0.2067, 0.179, 0.1982, 0.189, 0.1878, 0.1878, 0.2104, 0.21, 0.2235, 0.2814, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012041286224945166", "ticker_idx": 0, "ts": "2026-01-16T05:56:39+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML's revenue from TSMC would grow 27% YoY based on TSMC's new capex guidance. This is substantially higher than consensus expectations", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein relays bullish ASML thesis: TSMC capex surge drives 27% ASML revenue growth vs 10% consensus; EUV shipments have upside to even Bernstein's above-consensus 62-unit forecast.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Why would TSMC’s strong capex guidance benefit ASML the most among semicap names?\n\nYesterday (Jan 15th), TSMC smashed market expectations with a strong capex guidance of $52-56bn. This is great news for all semi equipment names, and especially for ASML given the high litho intensity in advanced logic. We quantify the implications for ASML.\n\nStrong TSMC capex guidance. TSMC 2026 capex guidance of $52-56bn, up 32% YoY from 2025 capex of $40.8bn at the midpoint (Exhibit 1). Revenue guidance is 30% YoY, so the capex intensity remains high at 33%. The allocation to advanced logic increased to 70-80% (up from 70%) last year. This is offset by the decreased allocation to equipment, as TSMC needs to accelerate the build out of new clean rooms in anticipation to stronger capacity needs in 2028/2029. We conclude that 1. advanced logic equipment growth in 2026 is strong at around 30%; 2. equipment demand will further accelerate in 2028/29 when the new clean rooms are ready to be filled with equipments.\n\nLitho intensity for advanced logic is above average.\nTSMC’s capacity addition is mostly for N3 and N2 this year, which is very high on litho intensity and hence benefits ASML more than other companies. We estimate that the litho intensity is as high as ~36% for N3, before coming down slightly to 32% in N2 (Exhibit 2). As currently ASML litho intensity is around 25% (Exhibit 3), the increased spending in N3 is more positive for ASML relative to other processes. We estimate that ASML’s revenue from TSMC would grow 27% YoY based on TSMC’s new capex guidance. This is substantially higher than consensus expectations, which imply TW system sales will grow only 10% YoY (Exhibit 4).\n\nTSMC alone could buy 30 EUV... upside to EUV shipment. When we upgraded ASML (link), we forecasted 62 EUV unit shipment in 2026. Client feedback was the number was too high. Now with the TSMC capex guidance, we think there is upside even to our number. Based on the capex, we see the possibility for TSMC to add 120-130k wpm N3+N2 capacity (current model: 100k). As every 10k wpm needs 2.5 EUV machines at N3/N2, the capacity addition means TSMC alone may need to purchase 30 EUV machines, presenting upside to our forecast in EUV model (Exhibit 5, Exhibit 6).\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015250183147876406, "ret_signed_m1d": -0.015250183147876406, "alpha_spy_m1d": -0.01608875976030477, "alpha_c_m1d": -0.00530984787297506, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04009595059520432, "ret_signed_p1d": -0.04009595059520432, "alpha_spy_p1d": -0.04009595059520432, "alpha_c_p1d": -0.04009595059520432, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.009424191472970689, "ret_signed_p1w": 0.009424191472970689, "alpha_spy_p1w": 0.012937506276860855, "alpha_c_p1w": 0.010173506639609697, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.02519170177438279, "ret_signed_p1m": 0.02519170177438279, "alpha_spy_p1m": 0.039519531342202985, "alpha_c_p1m": 0.006884585452559211, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.055218192619088, "ret_signed_p3m": 0.055218192619088, "alpha_spy_p3m": 0.04048324149446891, "alpha_c_p3m": -0.0761786540224465, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.239, -0.2559, -0.2408, -0.2351, -0.2234, -0.2226, -0.2123, -0.1968, -0.2134, -0.2062, -0.2135, -0.2233, -0.233, -0.2539, -0.2403, -0.2401, -0.2359, -0.2448, -0.2511, -0.2491, -0.2584, -0.2404, -0.2377, -0.2855, -0.2642, -0.2629, -0.221, -0.2313, -0.226, -0.206, -0.1954, -0.1746, -0.1798, -0.1847, -0.1748, -0.1836, -0.1895, -0.1939, -0.2072, -0.2024, -0.2216, -0.2512, -0.2352, -0.2276, -0.2324, -0.2262, -0.2298, -0.2298, -0.2298, -0.2231, -0.2132, -0.2106, -0.2106, -0.155, -0.0977, -0.091, -0.0989, -0.1321, -0.0733, -0.0692, -0.0553, -0.0711, -0.0153, 0.0, -0.0401, -0.0233, -0.0106, 0.0079, 0.0094, 0.0093, 0.0432, 0.0233, 0.0212, 0.0415, 0.0493, 0.0199, -0.0228, -0.0151, 0.0228, 0.0336, 0.0235, 0.0362, 0.0122, 0.0212, 0.0252, 0.0288, 0.0679, 0.0623, 0.0772, 0.0717, 0.0839, 0.1053, 0.0573, 0.0581, 0.0384, -0.0033, 0.0293, 0.0175, -0.016, -0.0155, 0.0295, 0.0285, 0.0127, 0.012, 0.0273, 0.0242, 0.0218, 0.0025, -0.0321, 0.0084, 0.0309, 0.0394, 0.0001, -0.0165, -0.046, -0.0398, 0.0188, -0.004, -0.004, -0.004, -0.0445, 0.0403, 0.0593, 0.0895, 0.0808, 0.1017, 0.0552, 0.0489, 0.0672, 0.0681, 0.0612, 0.0722, 0.0499, 0.0744, 0.0426, 0.0075, 0.0254, 0.051, 0.051, 0.0209, 0.0567, 0.122, 0.1177, 0.1344, 0.1226, 0.0883, 0.141, 0.1753, 0.1234, 0.0871, 0.0739, 0.1461, 0.1566, 0.2115, 0.23, 0.1949, 0.1834, 0.196, 0.1906, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2032626174560710762", "ticker_idx": 0, "ts": "2026-03-14T01:13:39+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SNDK is looking very strong", "tweet_type": "prediction", "timeline": "weeks", "summary": "Bullish SNDK ahead of GTC; expects potential ICMS rack unveil as catalyst.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Do you think there will be any mention of SNDK at GTC? SNDK is looking very strong.\n\nMy guess is… they might unveil an ICMS rack.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05970468779628246, "ret_signed_m1d": -0.05970468779628246, "alpha_spy_m1d": -0.04963034690322943, "alpha_c_m1d": -0.045437518314638714, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.014267169481643749, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023506641150271124, "ret_signed_p1d": 0.023506641150271124, "alpha_spy_p1d": 0.020876086280295914, "alpha_c_p1d": 0.018030403913663395, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.005476237236607728, "ret_p1w": -0.0016201904985949644, "ret_signed_p1w": -0.0016201904985949644, "alpha_spy_p1w": 0.01610723915583412, "alpha_c_p1w": 0.010339242302156104, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.011959432800751069, "ret_p1m": 0.34226797524069963, "ret_signed_p1m": 0.34226797524069963, "alpha_spy_p1m": 0.30142293007137577, "alpha_c_p1m": 0.2749388245952422, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.06732915064545741, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.7061, -0.6881, -0.6623, -0.6574, -0.6519, -0.6446, -0.6446, -0.6446, -0.6529, -0.6586, -0.6626, -0.6626, -0.6088, -0.6105, -0.5031, -0.4975, -0.5246, -0.4636, -0.4468, -0.446, -0.4488, -0.4184, -0.4122, -0.4122, -0.356, -0.2876, -0.2845, -0.3266, -0.3309, -0.3158, -0.2501, -0.2335, -0.181, -0.0546, -0.0115, -0.1692, -0.1811, -0.1502, -0.1709, -0.2302, -0.1482, -0.1042, -0.1095, -0.1095, -0.1607, -0.1467, -0.1173, -0.0763, -0.0528, -0.0925, -0.1013, -0.0735, -0.097, -0.1202, -0.1964, -0.1486, -0.1962, -0.2506, -0.1633, -0.1204, -0.0685, -0.1205, -0.0597, 0.0, 0.0235, 0.0711, 0.0973, 0.0086, -0.0016, -0.0016, -0.0366, -0.1428, -0.1248, -0.1864, -0.0971, -0.0155, -0.0029, -0.0029, 0.0298, 0.0102, 0.1098, 0.2103, 0.2105, 0.3537, 0.3423, 0.2673, 0.3068, 0.3089, 0.2976, 0.284, 0.3915, 0.3252, 0.4068, 0.521, 0.4245, 0.5125, 0.5584, 0.687, 0.7848, 0.9987, 1.0039, 0.9044, 1.2204, 1.1994, 1.0636, 1.0568, 0.9651, 1.0005, 0.8945, 0.9659, 0.9791, 1.1918, 1.1015, 1.1015, 1.2591, 1.2596, 1.3331, 1.4089, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015622185784865142", "ticker_idx": 0, "ts": "2026-01-26T03:05:52+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy Rating, Raise Target Price to $60 ... we remain bullish on the stock and recommend investors accumulate positions during the pullback", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF Securities reiterates Buy on INTC, raises PT to $60; pullback is a buying opportunity on foundry progress and server CPU tailwinds.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics Communication》  \n☄️ Intel (INTC US Buy): Pullback is a Buying Opportunity \n \n☀️ Maintain Buy Rating, Raise Target Price to $60: Intel's stock price has risen 38% year-to-date, benefiting from the release of Panther Lake at CES and strong momentum in server CPUs. Therefore, given the high expectations driven by multiple sell-side upgrades, Intel's weak 1Q26 guidance triggered a 12% after-hours stock drop. However, we remain bullish on the stock and recommend investors accumulate positions during the pullback, considering positive progress in its foundry business: 1) Steady improvement in 18A yields; 2) Discussions on 18A-P with internal/external customers; 3) 14A has released PDK 0.5, with customer formal orders expected as early as 2H26 to 1H27. Indeed, we believe memory/substrate shortages and PC weakness will persist, dragging gross margins in the short term. Nonetheless, we expect profitability to gradually recover in future quarters, benefiting from higher 18A yields/sales and continued tailwinds in server CPUs, as well as easing capacity constraints. Longer-term, we reiterate expectations for improved foundry execution and a solid external pipeline for 14A (Apple, Nvidia, AMD, etc.). Considering performance, guidance, and product mix changes, we lower our 2026E/2027E earnings estimates by 11% and 1%, respectively. Target price of $60 based on 2.3x 2027E BVPS.  \n\n☀️ Weak 1Q26 Guidance: 4Q25 revenue of $13.6 billion, flat QoQ, mainly driven by strong data center performance. Gross margin of 37.9%, above the guided 36%, primarily due to higher revenue and lower inventory provisions, as well as gradual improvement in data center profitability. EPS of $0.15, with adjusted free cash flow expanding to $2.2 billion. For 1Q26, the company guides: 1) Revenue of $11.7-12.7 billion, midpoint below consensus of $12.6 billion; 2) Gross margin of 34.5% below expectations; 3) EPS at breakeven. By segment, CCG operating margin declined to 27%, while DCAI operating margin continued to rise to 26.4%. On capital expenditures, Intel raised its 2026 capex guidance to flat to slightly down YoY, compared to prior guidance of a decline YoY.  \n\n☀️ Steady Progress in Foundry: Based on our previous reports, we believe 18A yields are steadily progressing toward 70% in 1Q26, benefiting from the successful release of Panther Lake and the addition of more product categories. For 14A, the earnings call provided more positive updates, including the release of PDK 0.5 and potential customer commitments as early as 2H26 to 1H27. We reiterate expectations for potential order wins, such as Apple's mobile SoC and Nvidia/AMD x86 server chips. Additionally, we do not rule out the possibility of upward revisions to capex in subsequent earnings calls.  \n\n☀️ Tailwinds in Server CPUs: CPU demand is now very strong, mainly due to their increasingly important role in data centers, managing GPU loads and handling I/O, networking, storage, and other tasks. Therefore, benefiting from strong demand from major US CSPs and data center modernization, we expect DCAI segment revenue to grow 15% YoY in 2026, accompanied by potential CPU price increases and continued gross margin improvements.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06072012036810337, "ret_signed_m1d": 0.06072012036810337, "alpha_spy_m1d": 0.06577261014261604, "alpha_c_m1d": 0.057535754011479145, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.033890293476706246, "ret_signed_p1d": 0.033890293476706246, "alpha_spy_p1d": 0.02990602874702164, "alpha_c_p1d": 0.012752957031776813, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.14874086714915546, "ret_signed_p1w": 0.14874086714915546, "alpha_spy_p1w": 0.14487207426907722, "alpha_c_p1w": 0.1257982019657462, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.08543179830603753, "ret_signed_p1m": 0.08543179830603753, "alpha_spy_p1m": 0.09319818576701389, "alpha_c_p1m": 0.03443135709358014, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.5716638301085273, "ret_signed_p3m": 0.5716638301085273, "alpha_spy_p3m": 0.5461779471826702, "alpha_c_p3m": 0.3634746789765029, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0271, -0.0548, -0.0588, -0.0704, -0.1285, -0.0967, -0.1236, -0.1026, -0.0951, -0.1085, -0.1083, -0.1549, -0.164, -0.1831, -0.192, -0.1737, -0.2088, -0.188, -0.1577, -0.1567, -0.1337, -0.1337, -0.0454, -0.0584, 0.0231, 0.0299, -0.0468, -0.0254, -0.0515, -0.0468, -0.0402, -0.0701, -0.1101, -0.1172, -0.1219, -0.1516, -0.1462, -0.1334, -0.144, -0.1445, -0.149, -0.149, -0.148, -0.1367, -0.1221, -0.1316, -0.1316, -0.0732, -0.0734, -0.0577, 0.0033, -0.0325, 0.072, 0.0369, 0.113, 0.1466, 0.1372, 0.1052, 0.1052, 0.1429, 0.2768, 0.2784, 0.0607, 0.0, 0.0339, 0.148, 0.1452, 0.0937, 0.1487, 0.1591, 0.1438, 0.1353, 0.1906, 0.1824, 0.1092, 0.1365, 0.0939, 0.1012, 0.1012, 0.0868, 0.0699, 0.0501, 0.0381, 0.0268, 0.0854, 0.1033, 0.0699, 0.0734, 0.0708, 0.0144, 0.0727, 0.0814, 0.0219, 0.0727, 0.101, 0.1292, 0.065, 0.0772, 0.077, 0.0369, 0.0598, 0.0868, 0.0325, 0.0358, 0.0369, 0.1104, 0.0379, 0.0151, -0.0306, 0.0386, 0.1304, 0.1857, 0.1857, 0.1951, 0.2452, 0.3874, 0.4526, 0.4681, 0.534, 0.5018, 0.5284, 0.6121, 0.6121, 0.5462, 0.5594, 0.5361, 0.5717, 0.9426, 1.0002, 0.9892, 1.2299, 1.2236, 1.3446, 1.2542, 1.5453, 1.6597, 1.5799, 1.94, 2.0464, 1.8386, 1.831, 1.7284, 1.5599, 1.5458, 1.6077, 1.7997, 1.7889, 1.8204, 1.8204, 1.907, 1.8659, 1.8451, 1.699, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2029536912198029383", "ticker_idx": 0, "ts": "2026-03-05T12:38:01+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung management presented a constructive memory market outlook, expecting tight supply-demand conditions to continue through 2027... HBM3E pricing in 2026 will come in above 2025 levels. This diverges from sell-side expectations — including JPM's own — that HBM3E prices would decline in 2026. (Additional upside?)", "tweet_type": "news_reaction", "timeline": "years", "summary": "Relays Samsung mgmt's bullish memory/HBM3E pricing outlook through 2027, flagging potential upside vs consensus on HBM3E price trajectory.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "JPM Samsung Korea Conference – Key Takeaways\n\neSSD supply shortage to persist through 2027.\n\nSamsung management presented a constructive memory market outlook, expecting tight supply-demand conditions to continue through 2027 (with DRAM/NAND bit demand growth significantly outpacing supply growth). Management also noted that given capex spending by suppliers is concentrated more heavily on DRAM than NAND, the eSSD NAND supply shortage is expected to be severe and prolonged.\n\nLimited cleanroom space at P4 is cited as the primary driver of slowing supply additions, and management noted that technology migration will be the key driver of bit supply growth — particularly in NAND.\n\nP5 construction has commenced, with management expecting approximately three years until wafer output begins.\n\nAnother notable point: Samsung management projected that HBM3E pricing in 2026 will come in above 2025 levels.\n\nThis diverges from sell-side expectations — including JPM's own — that HBM3E prices would decline in 2026. (Additional upside?)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10125265259532668, "ret_signed_m1d": -0.10125265259532668, "alpha_spy_m1d": -0.10685946046163786, "alpha_c_m1d": -0.09882715266224451, "bench_spy_m1d": 0.005606807866311181, "bench_c_m1d": -0.0024254999330821647, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.017745339734864696, "ret_signed_p1d": -0.017745339734864696, "alpha_spy_p1d": -0.0046382046640282, "alpha_c_p1d": 0.002871028242738105, "bench_spy_p1d": -0.013107135070836495, "bench_c_p1d": -0.0206163679776028, "ret_p1w": -0.019311073760128195, "ret_signed_p1w": -0.019311073760128195, "alpha_spy_p1w": 0.003072320832539721, "alpha_c_p1w": -0.0026182284781424148, "bench_spy_p1w": -0.022383394592667916, "bench_c_p1w": -0.01669284528198578, "ret_p1m": -0.02617227389035548, "ret_signed_p1m": -0.02617227389035548, "alpha_spy_p1m": 0.008597372776366852, "alpha_c_p1m": 0.002511619656626718, "bench_spy_p1m": -0.03476964666672233, "bench_c_p1m": -0.028683893546982198, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4313, -0.437, -0.4391, -0.4427, -0.4344, -0.4557, -0.4661, -0.4396, -0.4411, -0.4479, -0.4261, -0.4209, -0.423, -0.423, -0.3923, -0.3763, -0.3742, -0.3742, -0.3742, -0.3293, -0.2792, -0.2751, -0.2641, -0.2756, -0.2745, -0.2756, -0.2818, -0.2677, -0.249, -0.2229, -0.2208, -0.2422, -0.2197, -0.2051, -0.2062, -0.2062, -0.1675, -0.1524, -0.1613, -0.1623, -0.215, -0.1258, -0.1174, -0.1686, -0.1722, -0.1315, -0.1347, -0.1242, -0.0678, -0.0543, -0.0543, -0.0543, -0.0543, -0.0084, -0.0078, 0.0073, 0.0438, 0.0621, 0.1378, 0.13, 0.13, 0.0183, -0.1013, 0.0, -0.0177, -0.0945, -0.0193, -0.0084, -0.0193, -0.0423, -0.0151, 0.012, 0.0882, 0.0465, 0.0407, -0.0277, -0.0099, -0.0136, -0.06, -0.06, -0.0779, -0.1255, -0.0081, -0.067, -0.0262, 0.0099, 0.0277, 0.1009, 0.0669, 0.0774, 0.0512, 0.08, 0.1035, 0.1375, 0.1297, 0.1218, 0.1454, 0.1375, 0.1741, 0.148, 0.1741, 0.1611, 0.182, 0.1532, 0.1532, 0.216, 0.216, 0.3912, 0.4199, 0.4043, 0.4932, 0.4592, 0.4853, 0.5481, 0.4147, 0.4696, 0.4409, 0.4435, 0.5664, 0.5298, 0.5298, 0.5638, 0.6056, 0.5664, 0.6579, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025728266473213974", "ticker_idx": 0, "ts": "2026-02-23T00:23:49+00:00", "symbol": "HBM", "kind": "sector", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Should this latent demand materialize, the HBM supply shortage is expected to intensify further", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin CPX pivoting to HBM from GDDR7 signals deeper-than-modeled HBM demand; supply shortage to intensify, bullish for HBM producers including MU.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS: Rubin CPX pivots from GDDR7 to HBM adoption\n\nAccording to our channel checks, Rubin CPX is estimated to adopt HBM instead of GDDR7. NVIDIA had previously announced plans to offload prefill operations—a relatively less memory-intensive segment of AI inference workloads—onto a separate GPU, Rubin CPX, equipped with GDDR7. This design strategy was presumably based on the assessment that prefill operations require relatively lower memory bandwidth, making HBM unnecessary. The fact that Rubin CPX's memory specification has ultimately been changed to HBM is therefore highly significant. It demonstrates that even prefill operations demand high memory bandwidth and capacity in real-world production environments, and that GDDR7 is ultimately insufficient to deliver adequate performance efficiency. As NVIDIA has not yet officially announced this spec change, we have not incorporated Rubin CPX-related HBM demand into our HBM model. Should this latent demand materialize, the HBM supply shortage is expected to intensify further.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -8.557249503179007e-06, "ret_signed_m1d": -8.557249503179007e-06, "alpha_spy_m1d": -0.010325218636566732, "alpha_c_m1d": -0.005215866567121104, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02867344988938832, "ret_signed_p1d": 0.02867344988938832, "alpha_spy_p1d": 0.0214049969589933, "alpha_c_p1d": 0.013463222450354768, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": 0.0732572357951613, "ret_signed_p1w": 0.0732572357951613, "alpha_spy_p1w": 0.0674101500822697, "alpha_c_p1w": 0.08897606489229104, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": -0.012937937696617626, "ret_signed_p1m": -0.012937937696617626, "alpha_spy_p1m": 0.027253444627845104, "alpha_c_p1m": 0.03136066802083074, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 0.8032835620685598, "ret_signed_p3m": 0.8032835620685598, "alpha_spy_p3m": 0.7119013452307753, "alpha_c_p3m": 0.4278718245977816, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4575, -0.4492, -0.4543, -0.4479, -0.4371, -0.4416, -0.45, -0.4352, -0.4141, -0.4155, -0.3999, -0.413, -0.4218, -0.4377, -0.4535, -0.4429, -0.4249, -0.4151, -0.3879, -0.385, -0.376, -0.376, -0.3635, -0.3362, -0.333, -0.3387, -0.3387, -0.291, -0.2704, -0.2337, -0.2275, -0.2364, -0.2259, -0.2239, -0.2359, -0.2337, -0.2224, -0.1906, -0.1871, -0.1998, -0.1743, -0.1575, -0.152, -0.1712, -0.1193, -0.0801, -0.0756, -0.0757, -0.1027, -0.0607, -0.0921, -0.1266, -0.1195, -0.098, -0.1111, -0.0838, -0.0525, -0.0526, -0.0526, -0.062, -0.0453, -0.028, -0.0, 0.0, 0.0287, 0.048, 0.0915, 0.073, 0.0733, -0.0327, -0.0871, -0.0243, -0.0573, -0.0985, -0.0269, -0.005, -0.0279, -0.0261, 0.0178, 0.0411, 0.0965, 0.0538, 0.0329, -0.0304, -0.0129, -0.0216, -0.0799, -0.0785, -0.1334, -0.1596, -0.0654, -0.1097, -0.08, -0.0554, -0.0392, 0.0492, 0.0374, 0.0503, 0.051, 0.1136, 0.1255, 0.1442, 0.1304, 0.1359, 0.1648, 0.192, 0.2003, 0.2025, 0.2578, 0.2402, 0.2559, 0.2429, 0.2627, 0.3671, 0.4176, 0.5506, 0.5629, 0.6483, 0.7843, 0.7345, 0.822, 0.8188, 0.6809, 0.6723, 0.6431, 0.6703, 0.8033, 0.7827, 0.7827, 0.9477, 1.0544, 1.0538, 1.1372, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025728266473213974", "ticker_idx": 1, "ts": "2026-02-23T00:23:49+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Should this latent demand materialize, the HBM supply shortage is expected to intensify further", "tweet_type": "analysis", "timeline": "quarters", "summary": "Rubin CPX pivoting to HBM from GDDR7 signals deeper-than-modeled HBM demand; supply shortage to intensify, bullish for HBM producers including MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS: Rubin CPX pivots from GDDR7 to HBM adoption\n\nAccording to our channel checks, Rubin CPX is estimated to adopt HBM instead of GDDR7. NVIDIA had previously announced plans to offload prefill operations—a relatively less memory-intensive segment of AI inference workloads—onto a separate GPU, Rubin CPX, equipped with GDDR7. This design strategy was presumably based on the assessment that prefill operations require relatively lower memory bandwidth, making HBM unnecessary. The fact that Rubin CPX's memory specification has ultimately been changed to HBM is therefore highly significant. It demonstrates that even prefill operations demand high memory bandwidth and capacity in real-world production environments, and that GDDR7 is ultimately insufficient to deliver adequate performance efficiency. As NVIDIA has not yet officially announced this spec change, we have not incorporated Rubin CPX-related HBM demand into our HBM model. Should this latent demand materialize, the HBM supply shortage is expected to intensify further.\n\n$NVDA $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.017103317512722427, "ret_signed_m1d": 0.017103317512722427, "alpha_spy_m1d": 0.006786656125658874, "alpha_c_m1d": 0.011896008195104502, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007031412215625532, "ret_signed_p1d": -0.007031412215625532, "alpha_spy_p1d": -0.014299865146020552, "alpha_c_p1d": -0.022241639654659084, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": -0.019716354573246386, "ret_signed_p1w": -0.019716354573246386, "alpha_spy_p1w": -0.025563440286137995, "alpha_c_p1w": -0.003997525476116648, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": -0.06043186692613789, "ret_signed_p1m": -0.06043186692613789, "alpha_spy_p1m": -0.020240484601675157, "alpha_c_p1m": -0.01613326120868952, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 0.811103166517908, "ret_signed_p3m": 0.811103166517908, "alpha_spy_p3m": 0.7197209496801236, "alpha_c_p3m": 0.4356914290471299, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4532, -0.4532, -0.4385, -0.429, -0.4313, -0.444, -0.4618, -0.4367, -0.4137, -0.4006, -0.3738, -0.3863, -0.4274, -0.4361, -0.4479, -0.4645, -0.4098, -0.3686, -0.3432, -0.344, -0.3193, -0.3193, -0.3238, -0.3007, -0.3049, -0.322, -0.322, -0.2507, -0.2585, -0.1842, -0.1934, -0.2232, -0.1803, -0.1784, -0.1968, -0.2081, -0.2003, -0.1383, -0.1383, -0.133, -0.0757, -0.0556, -0.0506, -0.0757, -0.0255, 0.034, 0.0352, -0.0145, 0.04, -0.0036, -0.0987, -0.0905, -0.0624, -0.089, -0.1134, -0.0253, -0.0166, -0.0221, -0.0221, -0.0503, -0.0, -0.0086, 0.0171, 0.0, -0.007, 0.0191, -0.0129, -0.0204, -0.0197, -0.0981, -0.048, -0.0568, -0.1204, -0.0752, -0.0424, -0.0054, -0.0371, 0.0123, 0.0495, 0.0967, 0.0968, 0.0553, 0.0046, -0.0395, -0.0604, -0.0924, -0.1556, -0.1514, -0.2353, -0.1971, -0.1258, -0.1296, -0.1296, -0.1023, -0.1027, -0.0334, 0.0017, -0.0005, 0.0137, 0.1066, 0.0842, 0.0866, 0.0815, 0.0657, 0.0679, 0.1585, 0.1448, 0.1804, 0.2466, 0.1984, 0.2321, 0.229, 0.2885, 0.3699, 0.5214, 0.5841, 0.5367, 0.7748, 0.8901, 0.8217, 0.9098, 0.8442, 0.7221, 0.6197, 0.6605, 0.7395, 0.8111, 0.7847, 0.7847, 1.129, 1.2063, 1.1947, 1.3075, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008049076089270517", "ticker_idx": 0, "ts": "2026-01-05T05:33:02+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Simon Woo at BofA raised his price target on SK hynix to KRW 1,000,000", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "BofA raises SK Hynix PT to KRW 1,000,000; memory oversupply not a concern until 2028+, capacity growth constrained to single digits.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Simon Woo at BofA raised his price target on SK hynix to KRW 1,000,000 and made the following comments:\n\n- The possibility of a memory oversupply driven by new fabs should not really be debated until 2028. Even then, if the increase in memory content per NVIDIA GPU—such as with Rubin Ultra—continues, it would be hard to argue that 2028 would immediately signal a memory downturn.\n\n- Despite the memory chipmakers’ record-high capex, DRAM and NAND wafer capacity is likely to grow only at a single-digit (%) rate per year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027298908880455053, "ret_signed_m1d": -0.027298908880455053, "alpha_spy_m1d": -0.02068285871152975, "alpha_c_m1d": -0.015937356438575967, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0431033745856777, "ret_signed_p1d": 0.0431033745856777, "alpha_spy_p1d": 0.037156191212588396, "alpha_c_p1d": 0.016540174086262516, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.07614937008435851, "ret_signed_p1w": 0.07614937008435851, "alpha_spy_p1w": 0.06533101031375943, "alpha_c_p1w": 0.04153508620242308, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.30316086614716653, "ret_signed_p1m": 0.30316086614716653, "alpha_spy_p1m": 0.30052891624205014, "alpha_c_p1m": 0.2499550171237912, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.19472924735399255, "ret_signed_p3m": 0.19472924735399255, "alpha_spy_p3m": 0.23849543729436407, "alpha_c_p3m": 0.15571865086258874, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4322, -0.4322, -0.3855, -0.4042, -0.4092, -0.3934, -0.3503, -0.3317, -0.3029, -0.3123, -0.3087, -0.313, -0.2678, -0.2319, -0.252, -0.1988, -0.1845, -0.1974, -0.1098, -0.1586, -0.1687, -0.1486, -0.1673, -0.1299, -0.1113, -0.1141, -0.1213, -0.196, -0.1299, -0.1816, -0.1931, -0.1802, -0.252, -0.2534, -0.2548, -0.2477, -0.2184, -0.2385, -0.227, -0.1983, -0.2069, -0.2213, -0.2184, -0.171, -0.1868, -0.1566, -0.1882, -0.1796, -0.204, -0.2385, -0.2083, -0.2069, -0.2141, -0.1667, -0.1609, -0.1552, -0.1552, -0.1394, -0.0805, -0.0647, -0.0647, -0.0647, -0.0273, 0.0, 0.0431, 0.0661, 0.0862, 0.069, 0.0761, 0.0603, 0.0661, 0.0761, 0.0862, 0.0977, 0.0675, 0.0632, 0.0848, 0.102, 0.0575, 0.1494, 0.2083, 0.2371, 0.306, 0.1925, 0.3032, 0.2931, 0.2098, 0.2055, 0.2744, 0.2586, 0.2356, 0.2759, 0.2644, 0.2644, 0.2644, 0.2644, 0.2845, 0.3635, 0.3664, 0.444, 0.4626, 0.5819, 0.5272, 0.5272, 0.3516, 0.2221, 0.3545, 0.33, 0.2034, 0.3502, 0.3747, 0.3387, 0.3099, 0.402, 0.3962, 0.52, 0.4581, 0.4495, 0.343, 0.4193, 0.4322, 0.343, 0.343, 0.2566, 0.1616, 0.2912, 0.1947, 0.2609, 0.2753, 0.3185, 0.4869, 0.4366, 0.4783, 0.497, 0.5877, 0.6352, 0.6625, 0.6237, 0.6784, 0.7619, 0.7604, 0.7633, 0.759, 0.8597, 0.8713, 0.8612, 0.8511, 0.8511, 1.0829, 1.0829, 1.3045, 1.3808, 1.4269, 1.7061, 1.6414, 1.8443, 1.8357, 1.6183, 1.6486, 1.5118, 1.5118, 1.7925, 1.7939, 1.7939, 1.9537, 2.2286, 2.2954, 2.3588, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2021130348466499735", "ticker_idx": 0, "ts": "2026-02-10T07:53:20+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I've been bullish on Micron since it was at $100, and that hasn't changed", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Reaffirms ongoing bullish stance on Micron; conviction unchanged despite recent 15% cap decline.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@BullishMostly I’ve been bullish on Micron since it was at $100, and that hasn’t changed.", "reply_text": null, "parent_text": "@jukan05 there are basically three companies in the world that make memory at the scale required by customers. \nIt seems ridiculous the market has removed 15% of $MU cap in a month for what? Because Samsung puts out a headline?", "quoted_text": null, "parent_author": "BullishMostly", "ret_m1d": 0.027461490390592758, "ret_signed_m1d": 0.027461490390592758, "alpha_spy_m1d": 0.024817356954436987, "alpha_c_m1d": 0.02279209597103593, "bench_spy_m1d": 0.002644133436155771, "bench_c_m1d": 0.004669394419556827, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09937041532793689, "ret_signed_p1d": 0.09937041532793689, "alpha_spy_p1d": 0.09960147414235321, "alpha_c_p1d": 0.07461503245433265, "bench_spy_p1d": -0.00023105881441631837, "bench_c_p1d": 0.024755382873604237, "ret_p1w": 0.07107836721873584, "ret_signed_p1w": 0.07107836721873584, "alpha_spy_p1w": 0.08447199691042129, "alpha_c_p1w": 0.06428421776313042, "bench_spy_p1w": -0.01339362969168545, "bench_c_p1w": 0.006794149455605414, "ret_p1m": 0.12174148829661235, "ret_signed_p1m": 0.12174148829661235, "alpha_spy_p1m": 0.14455538496778264, "alpha_c_p1m": 0.13095685270103552, "bench_spy_p1m": -0.022813896671170286, "bench_c_p1m": -0.009215364404423165, "ret_p3m": 1.0016710945628486, "ret_signed_p3m": 1.0016710945628486, "alpha_spy_p3m": 0.9330204523953136, "alpha_c_p3m": 0.6019775370016733, "bench_spy_p3m": 0.06865064216753503, "bench_c_p3m": 0.3996935575611753, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3654, -0.339, -0.352, -0.3881, -0.395, -0.4607, -0.4446, -0.4003, -0.3987, -0.3833, -0.3833, -0.3667, -0.356, -0.3586, -0.3729, -0.393, -0.3647, -0.3387, -0.324, -0.2938, -0.3078, -0.3542, -0.364, -0.3773, -0.396, -0.3344, -0.2878, -0.2593, -0.2601, -0.2322, -0.2322, -0.2373, -0.2113, -0.216, -0.2353, -0.2353, -0.1549, -0.1637, -0.0799, -0.0903, -0.1239, -0.0754, -0.0734, -0.0941, -0.1069, -0.0981, -0.0281, -0.0281, -0.0221, 0.0425, 0.0652, 0.0707, 0.0424, 0.0991, 0.1662, 0.1676, 0.1115, 0.1729, 0.1238, 0.0165, 0.0258, 0.0574, 0.0275, 0.0, 0.0994, 0.1091, 0.1029, 0.1029, 0.0711, 0.1278, 0.1182, 0.1471, 0.1279, 0.1199, 0.1494, 0.1134, 0.1048, 0.1056, 0.0172, 0.0737, 0.0638, -0.0079, 0.0431, 0.08, 0.1217, 0.086, 0.1417, 0.1837, 0.2369, 0.2371, 0.1903, 0.133, 0.0833, 0.0597, 0.0237, -0.0477, -0.0429, -0.1375, -0.0945, -0.0141, -0.0184, -0.0184, 0.0125, 0.012, 0.0902, 0.1298, 0.1273, 0.1433, 0.2481, 0.2228, 0.2255, 0.2197, 0.2019, 0.2045, 0.3066, 0.2912, 0.3314, 0.406, 0.3516, 0.3896, 0.3861, 0.4533, 0.5451, 0.7159, 0.7867, 0.7332, 1.0017, 1.1317, 1.0547, 1.154, 1.0799, 0.9423, 0.8267, 0.8728, 0.9619, 1.0427, 1.0129, 1.0129, 1.4012, 1.4884, 1.4753, 1.6026, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2016672324687450563", "ticker_idx": 0, "ts": "2026-01-29T00:38:44+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Most customers are experiencing difficulties in securing memory volumes and continue to demand supply expansion. Especially server customers are in a situation where secured volumes immediately lead to set builds. Inventory levels are continuously decreasing. As we move into the second half of this year, inventory levels will decline to even lower levels.", "tweet_type": "analysis", "timeline": "quarters", "summary": "SK Hynix Q&A reveals persistent memory supply shortage, tightening inventory into 2H25, strong server demand, HBM4 on track with overwhelming share target — all pointing bullish for the stock.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SK Hynix 4Q25 Q/A\n\n1) Regarding HBM4\nHBM4 also aims for overwhelming market share, just like HBM3 and 3E.\nPreparations for HBM4 are proceeding with mass production of customer-requested volumes according to the schedule agreed upon with customers.\nHBM4 has achieved customer-required performance even based on 1b nano.\nPlans to secure yields at the level of HBM3E using MR-MUF.\nEven with current production maximization, demand cannot be fully met at 100%, so competitors' entry is expected, but competitiveness will be maintained.\n\n2) Regarding LTA\nIn the past, LTAs existed, but they were somewhat loose contracts regarding volumes and flexible depending on market conditions.\nIn contrast, recent LTAs under discussion reflect strong commitments.\nThis is because suppliers also need high visibility on demand.\nCustomers want LTA periods to be multi-year contracts unlike before, but it is difficult to respond to all due to CAPA constraints.\n\n3) Inventory status by major applications, owned inventory status\nMost customers are experiencing difficulties in securing memory volumes and continue to demand supply expansion.\nEspecially server customers are in a situation where secured volumes immediately lead to set builds.\nInventory levels are continuously decreasing.\nThere is not enough volume to accumulate inventory, and the speed of using memory through set builds is very fast.\nAs it is recognized as a bottleneck in data center expansion, server companies' purchases to secure volumes are expected to continue.\nPC/mobile customers are also directly or indirectly affected, experiencing difficulties in supply and demand, with inventory levels continuing to decrease.\n\nOur company's inventory also decreased in the 4th quarter compared to the previous quarter.\nAs we move into the second half of this year, inventory levels will decline to even lower levels.\n\nFor NAND as well, rapid inventory reduction by server companies is observed, especially centered on eSSD.\nOur company's inventory level for NAND at the end of last year was also at the same level as DRAM in terms of inventory weeks.\n\n---\n\nSK Hynix 4Q25 Q/A (2)\n\n4) CAPA Optimization Strategy\nRather than focusing only on short-term performance, it is important to respond to customer demand while building trust.\nMaximizing production volume as much as possible.\nCurrently expanding 1b nano CAPA at M15X and making efforts to improve yields.\nFor general DRAM, accelerating the transition speed to 1c nano and 321 layers.\n\n5) Regarding Additional Dividends, Treasury Stock Purchase and Cancellation, ADR Plans\nThe financial soundness goal is to continuously hold an appropriate level of cash that allows stable business operations even during industry fluctuations and execution of CAPEX despite such fluctuations.\nThe required CAPEX scale will continue to increase.\nThere is no change in the belief that reinvesting available resources into the business to increase corporate value is the best way to utilize cash.\n\nDue to achieving record-high performance last year, financial soundness improved faster than expected when announcing the current shareholder return policy.\nPerformance improvement is expected to continue this year as well.\nBased on the financial capacity secured last year, as we are implementing additional shareholder returns, we will continue to review the timing and methods for additional shareholder returns in the future according to performance and cash flow.\nFor the return method, we will maintain the current return policy's direction while flexibly seeking the optimal approach.\n\nThere are no specifically confirmed matters regarding ADR.\n\n6) Storage Market Outlook, Mid-to-Long-Term NAND Business Direction\nWe are paying attention to the fundamental changes in the AI server architecture itself.\nSSD is changing into a device that performs a core role in the computation pipeline.\nAlong with developing conventional storage lineup, preparing next-generation storage.\nAdditionally, concretizing the concept of HBF technology.\n\n---\n\nSK Hynix 4Q25 Q/A (3)\n\n7) Impact on Sets Due to Price Increases\nPC and mobile customers are showing purchase adjustments due to increased cost burdens.\nSome customers are conservatively adjusting shipment plans, reviewing spec adjustments centered on low-spec products with low price sensitivity.\nHowever, this will not expand into a contraction of overall market demand.\nThis is because on-device AI is driving demand centered on high-spec products.\n\n8) Regarding AI Corporation Establishment and Synergies\nAiming to leap beyond a simple component supplier to become a partner in the AI data center ecosystem.\nWill perform the role of discovering and investing in companies that have secured core AI capabilities.\nThe investment commitment scale of the AI corporation will not be large compared to SK Hynix's performance or cash.\nActual investment expenditures will be made sequentially after the investment scale is confirmed.\n\n9) Regarding CAPEX\nThe CAPEX scale for 2026 is expected to increase to a considerable level.\nWill maintain CAPEX by comprehensively considering demand visibility and investment efficiency.\nEven with the absolute CAPEX scale increasing, sales are also expected to increase, so there will be no issue in maintaining it at mid-30% of sales.\nInvestments in the AI corporation are not included in CAPEX and will have no impact on FCF calculations.\n\n10) Regarding Tariffs and Additional Fab Construction in the US\nOverseas semiconductor factory construction has many variables to consider, both domestically and internationally.\nCurrently, we are in a situation of watching the consultations being conducted by the governments of both countries.\nWe will share related matters in the future.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02322886099313315, "ret_signed_m1d": -0.02322886099313315, "alpha_spy_m1d": -0.025217189045392607, "alpha_c_m1d": -0.021097264532418003, "bench_spy_m1d": 0.001988328052259458, "bench_c_m1d": -0.002131596460715146, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05574910636406116, "ret_signed_p1d": 0.05574910636406116, "alpha_spy_p1d": 0.058731686624464374, "alpha_c_p1d": 0.08942413493409052, "bench_spy_p1d": -0.0029825802604032114, "bench_c_p1d": -0.03367502857002935, "ret_p1w": -0.022067337933747266, "ret_signed_p1w": -0.022067337933747266, "alpha_spy_p1w": 0.001591190973331047, "alpha_c_p1w": 0.06523382763574015, "bench_spy_p1w": -0.023658528907078313, "bench_c_p1w": -0.08730116556948742, "ret_p1m": 0.23456190454577452, "ret_signed_p1m": 0.23456190454577452, "alpha_spy_p1m": 0.24616066121464952, "alpha_c_p1m": 0.2612671981601582, "bench_spy_p1m": -0.011598756668874999, "bench_c_p1m": -0.026705293614383674, "ret_p3m": 0.5126582665179358, "ret_signed_p3m": 0.5126582665179358, "alpha_spy_p3m": 0.48442697109152233, "alpha_c_p3m": 0.336163716842631, "bench_spy_p3m": 0.02823129542641345, "bench_c_p3m": 0.17649454967530476, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2804, -0.3199, -0.328, -0.3118, -0.3268, -0.2967, -0.2816, -0.2839, -0.2897, -0.3501, -0.2967, -0.3385, -0.3477, -0.3373, -0.3953, -0.3965, -0.3976, -0.3918, -0.3682, -0.3844, -0.3751, -0.3519, -0.3589, -0.3705, -0.3682, -0.3298, -0.3426, -0.3182, -0.3438, -0.3368, -0.3566, -0.3844, -0.36, -0.3589, -0.3647, -0.3264, -0.3217, -0.3171, -0.3171, -0.3043, -0.2567, -0.2439, -0.2439, -0.2439, -0.2137, -0.1916, -0.1568, -0.1382, -0.122, -0.1359, -0.1301, -0.1429, -0.1382, -0.1301, -0.122, -0.1127, -0.137, -0.1405, -0.1231, -0.1092, -0.1452, -0.0708, -0.0232, 0.0, 0.0557, -0.036, 0.0534, 0.0453, -0.0221, -0.0256, 0.0302, 0.0174, -0.0012, 0.0314, 0.0221, 0.0221, 0.0221, 0.0221, 0.0383, 0.1022, 0.1045, 0.1672, 0.1823, 0.2788, 0.2346, 0.2346, 0.0926, -0.0121, 0.0949, 0.0752, -0.0272, 0.0914, 0.1112, 0.0821, 0.0589, 0.1333, 0.1287, 0.2287, 0.1787, 0.1717, 0.0856, 0.1473, 0.1578, 0.0856, 0.0856, 0.0158, -0.061, 0.0437, -0.0342, 0.0193, 0.0309, 0.0658, 0.202, 0.1613, 0.195, 0.2101, 0.2834, 0.3218, 0.3439, 0.3125, 0.3567, 0.4242, 0.4231, 0.4254, 0.4219, 0.5033, 0.5127, 0.5045, 0.4964, 0.4964, 0.6837, 0.6837, 0.8629, 0.9246, 0.9618, 1.1875, 1.1352, 1.2992, 1.2923, 1.1166, 1.141, 1.0305, 1.0305, 1.2574, 1.2585, 1.2585, 1.3877, 1.6099, 1.6639, 1.7151, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007631406265229468", "ticker_idx": 0, "ts": "2026-01-04T01:53:21+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron unveiled large-scale mass-production substrates (93×93mm) with increased layer counts from 22 to 24 layers", "tweet_type": "analysis", "timeline": "quarters", "summary": "CCL and IC substrate supply chain entering ASP upgrade cycle driven by AI platform spec inflation; structural long on CCL/substrate names including Unimicron, Nanya PCB, Kinsus.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Entry into a price increase cycle for CCL and substrate companies alongside the advancement of AI platform specifications\n\n- As AI server platforms rapidly evolve toward high-speed transmission and multi-layer stacking structures, the PCB and IC substrate supply chain is entering a phase of comprehensive upgrades across materials, structures, and sizes.\n\n- As process specifications are raised to the next level, so-called \"Specification Inflation\" is becoming a full-fledged reality, and this is expected to act as a structural background for future product price adjustments.\n\n- With the continuous rise in power consumption and operating frequencies of AI GPU and ASIC platforms, CCL is shifting from the existing M8 to M9 structure, and glass fiber is transitioning from Low Dk (Low Dielectric Constant) 2nd Generation to the application of Quartz Glass (Q-Glass).\n\n- According to the supply chain, next-generation platforms such as the Vera Rubin CPX server and 1.6T switches are scheduled to adopt the combination of M9-grade Quartz Glass + HVLP4 copper foil, notably starting from the second half of 2026.\n\n- As \"Specification Inflation\" led by AI servers emerges as the key keyword for next-generation technology transitions, the physical specifications of IC substrates are also continuously expanding, alongside material upgrades.\n\n- Unimicron unveiled large-scale mass-production substrates (93×93mm) with increased layer counts from 22 to 24 layers at a 2025 exhibition, and development specifications have already expanded to 120×150mm.\n\n- The industry is also raising the possibility of further expansion to over 150×150mm after 2027, which demands a much higher level of technical difficulty regarding process equipment and yield management.\n\n- In addition to mainboards, demand for substrates for CPO (Co-Packaged Optics) and optical modules is also rapidly increasing alongside the spread of Heterogeneous Integration.\n\n- For CPO substrates, trends include upsizing to over 110×110mm and miniaturization to line/space widths of 5/5μm levels; consequently, demand for low-loss dielectric materials and Ultra Low Roughness (ULR) copper foil is expanding rapidly.\n\n- Considering the rise in package module manufacturing costs, the market assesses that the entire supply chain has already secured the capacity for price adjustments, and mentions the possibility that price normalization for some product groups may precede the recovery of final demand.\n\n- Furthermore, major substrate companies such as Nanya PCB and Kinsus are sequentially introducing high-spec materials and design standards to respond to structural changes in BT and ABF.\n\n- Although subject to macroeconomic variables in the short term, structural drivers such as rising technical entry barriers and increased material costs remain valid. The CCL and substrate supply chain is judged to have currently entered a \"new cycle where specification upgrades drive ASP adjustments.\"\n\n---\n\nhttps://t.co/q5bHYL36WP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.047826086956521685, "ret_signed_m1d": -0.047826086956521685, "alpha_spy_m1d": -0.04121003678759638, "alpha_c_m1d": -0.045613441888031625, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0021739130434783593, "ret_signed_p1d": 0.0021739130434783593, "alpha_spy_p1d": -0.003773270329610945, "alpha_c_p1d": -0.011862887538306222, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": 0.05869565217391304, "ret_signed_p1w": 0.05869565217391304, "alpha_spy_p1w": 0.047877292403313954, "alpha_c_p1w": 0.04369075995175398, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.673913043478261, "ret_signed_p1m": 0.673913043478261, "alpha_spy_p1m": 0.6712810935731446, "alpha_c_p1m": 0.6914762372341333, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": 1.2565217391304349, "ret_signed_p3m": 1.2565217391304349, "alpha_spy_p3m": 1.3002879290708065, "alpha_c_p3m": 1.3150261030433081, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3043, -0.3, -0.3, -0.3326, -0.3652, -0.3109, -0.2804, -0.3022, -0.3043, -0.3043, -0.3087, -0.3217, -0.3217, -0.2891, -0.2935, -0.2783, -0.3043, -0.2891, -0.2239, -0.2674, -0.2925, -0.286, -0.3032, -0.2903, -0.2817, -0.2774, -0.2299, -0.2543, -0.2717, -0.2609, -0.3022, -0.2543, -0.3043, -0.287, -0.2348, -0.2239, -0.187, -0.1891, -0.1435, -0.1478, -0.1457, -0.137, -0.0652, -0.0413, -0.0174, -0.0109, -0.0217, -0.0217, -0.037, -0.0696, -0.0783, -0.1, -0.0652, -0.0413, -0.063, -0.0565, -0.0565, -0.0413, -0.0413, -0.0457, -0.0435, -0.0435, -0.0478, 0.0, 0.0022, -0.0174, -0.0109, -0.0196, 0.0587, 0.1043, 0.087, 0.0848, 0.1913, 0.3087, 0.3391, 0.3957, 0.4217, 0.4739, 0.4478, 0.5391, 0.5978, 0.4978, 0.6457, 0.587, 0.6739, 0.687, 0.5891, 0.4783, 0.5109, 0.6152, 0.6043, 0.6043, 0.6043, 0.6043, 0.6043, 0.6043, 0.6043, 0.6043, 0.763, 0.9391, 0.9935, 1.0935, 1.0935, 1.0913, 1.013, 0.813, 0.9565, 0.8783, 0.6913, 0.8435, 1.0261, 1.087, 1.2087, 1.387, 1.3609, 1.5217, 1.5261, 1.4087, 1.1696, 1.0, 1.2, 1.1826, 1.1957, 1.1457, 0.9326, 1.1239, 1.2565, 1.2565, 1.2565, 1.4522, 1.6957, 1.7174, 1.7739, 1.7217, 1.5957, 1.687, 1.8043, 1.7957, 1.9478, 2.087, 2.113, 2.1652, 2.4348, 2.6478, 2.587, 2.4913, 2.8391, 2.8391, 2.9609, 2.9261, 2.7304, 2.8957, 2.5565, 2.7435, 2.8043, 2.8783, 2.8304, 2.5696, 2.5478, 2.5565, 2.5783, 2.9348, 3.2174, 3.3043, 3.7174, 3.6957, 3.4565, 3.587, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007631406265229468", "ticker_idx": 1, "ts": "2026-01-04T01:53:21+00:00", "symbol": "Nanya PCB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "major substrate companies such as Nanya PCB and Kinsus are sequentially introducing high-spec materials and design standards", "tweet_type": "analysis", "timeline": "quarters", "summary": "CCL and IC substrate supply chain entering ASP upgrade cycle driven by AI platform spec inflation; structural long on CCL/substrate names including Unimicron, Nanya PCB, Kinsus.", "resolved_tickers": ["8046.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nanya PCB = Nan Ya PCB, TWSE 8046.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Entry into a price increase cycle for CCL and substrate companies alongside the advancement of AI platform specifications\n\n- As AI server platforms rapidly evolve toward high-speed transmission and multi-layer stacking structures, the PCB and IC substrate supply chain is entering a phase of comprehensive upgrades across materials, structures, and sizes.\n\n- As process specifications are raised to the next level, so-called \"Specification Inflation\" is becoming a full-fledged reality, and this is expected to act as a structural background for future product price adjustments.\n\n- With the continuous rise in power consumption and operating frequencies of AI GPU and ASIC platforms, CCL is shifting from the existing M8 to M9 structure, and glass fiber is transitioning from Low Dk (Low Dielectric Constant) 2nd Generation to the application of Quartz Glass (Q-Glass).\n\n- According to the supply chain, next-generation platforms such as the Vera Rubin CPX server and 1.6T switches are scheduled to adopt the combination of M9-grade Quartz Glass + HVLP4 copper foil, notably starting from the second half of 2026.\n\n- As \"Specification Inflation\" led by AI servers emerges as the key keyword for next-generation technology transitions, the physical specifications of IC substrates are also continuously expanding, alongside material upgrades.\n\n- Unimicron unveiled large-scale mass-production substrates (93×93mm) with increased layer counts from 22 to 24 layers at a 2025 exhibition, and development specifications have already expanded to 120×150mm.\n\n- The industry is also raising the possibility of further expansion to over 150×150mm after 2027, which demands a much higher level of technical difficulty regarding process equipment and yield management.\n\n- In addition to mainboards, demand for substrates for CPO (Co-Packaged Optics) and optical modules is also rapidly increasing alongside the spread of Heterogeneous Integration.\n\n- For CPO substrates, trends include upsizing to over 110×110mm and miniaturization to line/space widths of 5/5μm levels; consequently, demand for low-loss dielectric materials and Ultra Low Roughness (ULR) copper foil is expanding rapidly.\n\n- Considering the rise in package module manufacturing costs, the market assesses that the entire supply chain has already secured the capacity for price adjustments, and mentions the possibility that price normalization for some product groups may precede the recovery of final demand.\n\n- Furthermore, major substrate companies such as Nanya PCB and Kinsus are sequentially introducing high-spec materials and design standards to respond to structural changes in BT and ABF.\n\n- Although subject to macroeconomic variables in the short term, structural drivers such as rising technical entry barriers and increased material costs remain valid. The CCL and substrate supply chain is judged to have currently entered a \"new cycle where specification upgrades drive ASP adjustments.\"\n\n---\n\nhttps://t.co/q5bHYL36WP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018633540372670843, "ret_signed_m1d": -0.018633540372670843, "alpha_spy_m1d": -0.01201749020374554, "alpha_c_m1d": -0.016420895304180783, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012422360248447228, "ret_signed_p1d": 0.012422360248447228, "alpha_spy_p1d": 0.006475176875357924, "alpha_c_p1d": -0.0016144403333373525, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": 0.053830227743271175, "ret_signed_p1w": 0.053830227743271175, "alpha_spy_p1w": 0.04301186797267209, "alpha_c_p1w": 0.038825335521112114, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.6335403726708075, "ret_signed_p1m": 0.6335403726708075, "alpha_spy_p1m": 0.6309084227656911, "alpha_c_p1m": 0.6511035664266799, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": 1.351966873706004, "ret_signed_p3m": 1.351966873706004, "alpha_spy_p3m": 1.3957330636463756, "alpha_c_p3m": 1.4104712376188773, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0787, 0.0973, 0.0973, 0.1077, 0.06, 0.1656, 0.1656, 0.1615, 0.1491, 0.1325, 0.1408, 0.1284, 0.1284, 0.1532, 0.1429, 0.2029, 0.1677, 0.2029, 0.1801, 0.1594, 0.1263, 0.1304, 0.1304, 0.2029, 0.1967, 0.1781, 0.2091, 0.1698, 0.1284, 0.0994, 0.0393, 0.1139, 0.0352, 0.0228, 0.0352, 0.0642, 0.0787, 0.087, 0.1035, 0.0849, 0.1139, 0.0704, 0.118, 0.118, 0.176, 0.1594, 0.1511, 0.1242, 0.1014, 0.0311, 0.0083, -0.029, -0.0083, 0.0104, -0.0104, -0.0104, -0.0104, 0.0207, -0.0041, -0.0041, -0.0021, -0.0021, -0.0186, 0.0, 0.0124, 0.0, -0.0186, -0.0041, 0.0538, 0.1097, 0.0828, 0.0911, 0.1988, 0.3168, 0.4141, 0.4948, 0.5342, 0.5818, 0.5321, 0.6356, 0.6398, 0.5487, 0.6087, 0.5238, 0.6335, 0.6977, 0.646, 0.5694, 0.7246, 0.6957, 0.7226, 0.7226, 0.7226, 0.7226, 0.7226, 0.7226, 0.7226, 0.7226, 0.8944, 1.0828, 1.2899, 1.2981, 1.2981, 1.2402, 1.0704, 0.8634, 0.9089, 0.9441, 0.7516, 0.7723, 0.9482, 0.94, 1.1325, 1.3437, 1.2526, 1.3478, 1.323, 1.2402, 1.2112, 1.0331, 1.236, 1.3561, 1.323, 1.4017, 1.1615, 1.3768, 1.352, 1.352, 1.352, 1.5839, 1.6377, 1.7702, 1.7619, 1.7122, 1.6749, 1.7164, 1.9855, 1.9979, 2.2961, 2.2671, 2.3251, 2.2919, 2.619, 2.7805, 2.8592, 2.793, 3.1615, 3.1615, 3.0828, 3.1284, 2.8923, 2.9213, 2.5362, 2.7598, 2.7267, 2.6439, 2.5072, 2.2547, 2.2754, 2.3043, 2.4327, 2.6108, 2.8675, 2.6605, 2.8178, 2.7474, 2.5197, 2.5114, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007631406265229468", "ticker_idx": 2, "ts": "2026-01-04T01:53:21+00:00", "symbol": "Kinsus", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "major substrate companies such as Nanya PCB and Kinsus are sequentially introducing high-spec materials and design standards", "tweet_type": "analysis", "timeline": "quarters", "summary": "CCL and IC substrate supply chain entering ASP upgrade cycle driven by AI platform spec inflation; structural long on CCL/substrate names including Unimicron, Nanya PCB, Kinsus.", "resolved_tickers": ["3189.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Kinsus Interconnect Technology 3189.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Entry into a price increase cycle for CCL and substrate companies alongside the advancement of AI platform specifications\n\n- As AI server platforms rapidly evolve toward high-speed transmission and multi-layer stacking structures, the PCB and IC substrate supply chain is entering a phase of comprehensive upgrades across materials, structures, and sizes.\n\n- As process specifications are raised to the next level, so-called \"Specification Inflation\" is becoming a full-fledged reality, and this is expected to act as a structural background for future product price adjustments.\n\n- With the continuous rise in power consumption and operating frequencies of AI GPU and ASIC platforms, CCL is shifting from the existing M8 to M9 structure, and glass fiber is transitioning from Low Dk (Low Dielectric Constant) 2nd Generation to the application of Quartz Glass (Q-Glass).\n\n- According to the supply chain, next-generation platforms such as the Vera Rubin CPX server and 1.6T switches are scheduled to adopt the combination of M9-grade Quartz Glass + HVLP4 copper foil, notably starting from the second half of 2026.\n\n- As \"Specification Inflation\" led by AI servers emerges as the key keyword for next-generation technology transitions, the physical specifications of IC substrates are also continuously expanding, alongside material upgrades.\n\n- Unimicron unveiled large-scale mass-production substrates (93×93mm) with increased layer counts from 22 to 24 layers at a 2025 exhibition, and development specifications have already expanded to 120×150mm.\n\n- The industry is also raising the possibility of further expansion to over 150×150mm after 2027, which demands a much higher level of technical difficulty regarding process equipment and yield management.\n\n- In addition to mainboards, demand for substrates for CPO (Co-Packaged Optics) and optical modules is also rapidly increasing alongside the spread of Heterogeneous Integration.\n\n- For CPO substrates, trends include upsizing to over 110×110mm and miniaturization to line/space widths of 5/5μm levels; consequently, demand for low-loss dielectric materials and Ultra Low Roughness (ULR) copper foil is expanding rapidly.\n\n- Considering the rise in package module manufacturing costs, the market assesses that the entire supply chain has already secured the capacity for price adjustments, and mentions the possibility that price normalization for some product groups may precede the recovery of final demand.\n\n- Furthermore, major substrate companies such as Nanya PCB and Kinsus are sequentially introducing high-spec materials and design standards to respond to structural changes in BT and ABF.\n\n- Although subject to macroeconomic variables in the short term, structural drivers such as rising technical entry barriers and increased material costs remain valid. The CCL and substrate supply chain is judged to have currently entered a \"new cycle where specification upgrades drive ASP adjustments.\"\n\n---\n\nhttps://t.co/q5bHYL36WP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0714285714285714, "ret_signed_m1d": -0.0714285714285714, "alpha_spy_m1d": -0.0648125212596461, "alpha_c_m1d": -0.06006701898669231, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008928571428571397, "ret_signed_p1d": 0.008928571428571397, "alpha_spy_p1d": 0.0029813880554820926, "alpha_c_p1d": -0.017634629070843788, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.07738095238095233, "ret_signed_p1w": 0.07738095238095233, "alpha_spy_p1w": 0.06656259261035324, "alpha_c_p1w": 0.04276666849901689, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.4642857142857142, "ret_signed_p1m": 0.4642857142857142, "alpha_spy_p1m": 0.4616537643805978, "alpha_c_p1m": 0.41107986526233886, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 1.0, "ret_signed_p3m": 1.0, "alpha_spy_p3m": 1.0437661899403716, "alpha_c_p3m": 0.9609894035085962, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.256, -0.1815, -0.1815, -0.2054, -0.2708, -0.1994, -0.122, -0.1012, -0.1012, -0.1071, -0.128, -0.1369, -0.1369, -0.1161, -0.1815, -0.1726, -0.1815, -0.1875, -0.1726, -0.1994, -0.2113, -0.2262, -0.2381, -0.2381, -0.2381, -0.2173, -0.1399, -0.1845, -0.1577, -0.1339, -0.1994, -0.1696, -0.2321, -0.2143, -0.1815, -0.1786, -0.1607, -0.1518, -0.1577, -0.1607, -0.119, -0.119, -0.0655, -0.0714, -0.0506, -0.0923, -0.0982, -0.0982, -0.131, -0.1756, -0.1905, -0.2143, -0.1964, -0.1429, -0.1518, -0.1429, -0.1429, -0.0893, -0.1042, -0.119, -0.0536, -0.0536, -0.0714, 0.0, 0.0089, 0.0417, -0.0119, 0.003, 0.0774, 0.0506, 0.0238, 0.0417, 0.1458, 0.2171, 0.1883, 0.1509, 0.1826, 0.2315, 0.2381, 0.3601, 0.4494, 0.3661, 0.5, 0.4256, 0.4643, 0.6101, 0.5685, 0.4435, 0.5179, 0.5089, 0.5804, 0.5804, 0.5804, 0.5804, 0.5804, 0.5804, 0.5804, 0.5804, 0.7381, 0.9107, 0.9375, 0.872, 0.872, 0.8125, 0.8393, 0.6935, 0.7708, 0.7351, 0.5744, 0.631, 0.7292, 0.7202, 0.8899, 1.006, 1.0327, 1.1756, 1.2827, 1.2619, 1.0595, 0.9077, 1.0923, 1.0923, 1.0952, 1.0625, 0.8631, 0.9762, 1.0, 1.0, 1.0, 1.1339, 1.3452, 1.3036, 1.2679, 1.1815, 1.2083, 1.2827, 1.3423, 1.3304, 1.5625, 1.8185, 1.9256, 1.8512, 2.131, 2.0655, 2.0238, 2.0238, 2.1429, 2.1429, 2.1607, 2.1369, 1.872, 1.9315, 1.7262, 1.994, 1.9405, 1.9613, 2.1071, 1.8929, 2.006, 2.0536, 2.0, 2.2976, 2.625, 2.9821, 3.1429, 3.1786, 3.0536, 3.3393, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2007631406265229468", "ticker_idx": 3, "ts": "2026-01-04T01:53:21+00:00", "symbol": "CCL substrate sector", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The CCL and substrate supply chain is judged to have currently entered a 'new cycle where specification upgrades drive ASP adjustments'", "tweet_type": "analysis", "timeline": "quarters", "summary": "CCL and IC substrate supply chain entering ASP upgrade cycle driven by AI platform spec inflation; structural long on CCL/substrate names including Unimicron, Nanya PCB, Kinsus.", "resolved_tickers": ["2383.TW", "3034.TW", "4062.T", "2369.TW"], "resolution_method": "basket", "resolution_reason": "CCL/substrate basket: EMC, Unimicron, Ibiden, Tripod", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Entry into a price increase cycle for CCL and substrate companies alongside the advancement of AI platform specifications\n\n- As AI server platforms rapidly evolve toward high-speed transmission and multi-layer stacking structures, the PCB and IC substrate supply chain is entering a phase of comprehensive upgrades across materials, structures, and sizes.\n\n- As process specifications are raised to the next level, so-called \"Specification Inflation\" is becoming a full-fledged reality, and this is expected to act as a structural background for future product price adjustments.\n\n- With the continuous rise in power consumption and operating frequencies of AI GPU and ASIC platforms, CCL is shifting from the existing M8 to M9 structure, and glass fiber is transitioning from Low Dk (Low Dielectric Constant) 2nd Generation to the application of Quartz Glass (Q-Glass).\n\n- According to the supply chain, next-generation platforms such as the Vera Rubin CPX server and 1.6T switches are scheduled to adopt the combination of M9-grade Quartz Glass + HVLP4 copper foil, notably starting from the second half of 2026.\n\n- As \"Specification Inflation\" led by AI servers emerges as the key keyword for next-generation technology transitions, the physical specifications of IC substrates are also continuously expanding, alongside material upgrades.\n\n- Unimicron unveiled large-scale mass-production substrates (93×93mm) with increased layer counts from 22 to 24 layers at a 2025 exhibition, and development specifications have already expanded to 120×150mm.\n\n- The industry is also raising the possibility of further expansion to over 150×150mm after 2027, which demands a much higher level of technical difficulty regarding process equipment and yield management.\n\n- In addition to mainboards, demand for substrates for CPO (Co-Packaged Optics) and optical modules is also rapidly increasing alongside the spread of Heterogeneous Integration.\n\n- For CPO substrates, trends include upsizing to over 110×110mm and miniaturization to line/space widths of 5/5μm levels; consequently, demand for low-loss dielectric materials and Ultra Low Roughness (ULR) copper foil is expanding rapidly.\n\n- Considering the rise in package module manufacturing costs, the market assesses that the entire supply chain has already secured the capacity for price adjustments, and mentions the possibility that price normalization for some product groups may precede the recovery of final demand.\n\n- Furthermore, major substrate companies such as Nanya PCB and Kinsus are sequentially introducing high-spec materials and design standards to respond to structural changes in BT and ABF.\n\n- Although subject to macroeconomic variables in the short term, structural drivers such as rising technical entry barriers and increased material costs remain valid. The CCL and substrate supply chain is judged to have currently entered a \"new cycle where specification upgrades drive ASP adjustments.\"\n\n---\n\nhttps://t.co/q5bHYL36WP", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.010770684472842823, "ret_signed_m1d": -0.010770684472842823, "alpha_spy_m1d": -0.00415463430391752, "alpha_c_m1d": -0.008558039404352763, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03212275515413707, "ret_signed_p1d": 0.03212275515413707, "alpha_spy_p1d": 0.026175571781047763, "alpha_c_p1d": 0.018085954572352486, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": 0.0317347058592877, "ret_signed_p1w": 0.0317347058592877, "alpha_spy_p1w": 0.020916346088688614, "alpha_c_p1w": 0.016729813637128638, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.1122717015966806, "ret_signed_p1m": 0.1122717015966806, "alpha_spy_p1m": 0.1096397516915642, "alpha_c_p1m": 0.12983489535255294, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": 0.28219880750797716, "ret_signed_p3m": 0.28219880750797716, "alpha_spy_p3m": 0.32596499744834867, "alpha_c_p3m": 0.34070317142085055, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.135, -0.1115, -0.1183, -0.1306, -0.1547, -0.1519, -0.1318, -0.1475, -0.1393, -0.1422, -0.1308, -0.1509, -0.1361, -0.1169, -0.0708, -0.066, -0.0661, -0.0389, -0.0277, -0.0194, -0.0418, -0.0305, -0.0723, -0.061, -0.056, -0.0303, -0.0262, -0.0595, -0.0318, -0.0723, -0.1005, -0.066, -0.1187, -0.1201, -0.09, -0.0955, -0.0784, -0.0685, -0.0856, -0.0892, -0.0825, -0.068, -0.0364, -0.0316, -0.0297, -0.0211, -0.0224, -0.0161, -0.037, -0.0523, -0.05, -0.0659, -0.0544, -0.0295, -0.0282, -0.0033, -0.0091, -0.0025, -0.003, -0.002, -0.0051, -0.0051, -0.0108, 0.0, 0.0321, 0.0645, 0.0131, 0.0112, 0.0317, 0.0622, 0.0733, 0.0664, 0.0927, 0.1221, 0.1502, 0.1255, 0.1525, 0.1403, 0.1341, 0.1376, 0.1677, 0.1374, 0.1136, 0.0804, 0.1123, 0.1055, 0.0765, 0.065, 0.1257, 0.1324, 0.1498, 0.1703, 0.175, 0.1799, 0.1857, 0.1993, 0.2099, 0.1931, 0.2134, 0.2705, 0.2774, 0.2843, 0.2927, 0.2758, 0.2218, 0.1458, 0.1984, 0.1947, 0.1268, 0.1894, 0.2172, 0.2288, 0.2637, 0.2538, 0.3025, 0.3215, 0.3476, 0.3787, 0.3234, 0.3, 0.3545, 0.3596, 0.3486, 0.301, 0.2285, 0.3243, 0.2822, 0.2974, 0.3095, 0.3236, 0.4407, 0.4484, 0.4832, 0.4617, 0.5284, 0.5386, 0.5507, 0.5302, 0.5459, 0.623, 0.6644, 0.6209, 0.7351, 0.7965, 0.7368, 0.7364, 0.778, 0.7744, 0.8287, 0.8504, 0.885, 1.0114, 0.9547, 0.9773, 1.012, 1.0259, 1.0232, 0.9064, 0.8981, 0.8686, 0.8665, 1.0111, 1.1309, 1.2939, 1.3009, 1.3066, 1.2352, 1.3643, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2020357025851757045", "ticker_idx": 0, "ts": "2026-02-08T04:40:26+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Vibe coding will boost Apple's revenue even further", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish AAPL: vibe coding lowers dev barriers → more App Store submissions → higher App Store profit.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Lately I’ve been building a thesis I call the “Apple Bull” theory.\n\nVibe coding will boost Apple’s revenue even further.\n\nBy lowering the barrier to entry for programming, vibe coding will lead to more apps being submitted to the App Store, which in turn will generate more App Store profit for Apple.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011798054939316183, "ret_signed_m1d": 0.011798054939316183, "alpha_spy_m1d": 0.016596748547434426, "alpha_c_m1d": 0.027284705150551036, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.015486650211234854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003422959769044165, "ret_signed_p1d": -0.003422959769044165, "alpha_spy_p1d": -0.000785799336963322, "alpha_c_p1d": 0.002157818922450594, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.005580778691494759, "ret_p1w": -0.06860390424744434, "ret_signed_p1w": -0.06860390424744434, "alpha_spy_p1w": -0.051023334463101144, "alpha_c_p1w": -0.042165013155971764, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": -0.02643889109147257, "ret_p1m": -0.050214866528038304, "ret_signed_p1m": -0.050214866528038304, "alpha_spy_p1m": -0.026048803912984475, "alpha_c_p1m": -0.025171170109439367, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.025043696418598937, "ret_p3m": 0.046682679928123916, "ret_signed_p3m": 0.046682679928123916, "alpha_spy_p3m": -0.010422223795059526, "alpha_c_p3m": -0.13853514138975997, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.1852178213178839, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0051, -0.007, -0.009, -0.027, -0.0271, -0.023, -0.0314, -0.0123, 0.0038, 0.0076, 0.0097, 0.0097, 0.0145, 0.0299, 0.0412, 0.0337, 0.0212, 0.0142, 0.011, 0.0084, 0.0142, 0.0115, 0.0124, -0.0028, -0.001, -0.011, -0.0098, -0.0044, -0.0142, -0.0092, -0.0039, -0.0039, -0.0054, -0.0041, -0.0065, -0.011, -0.011, -0.0141, -0.0277, -0.0455, -0.0529, -0.0576, -0.0564, -0.0532, -0.0503, -0.0543, -0.0606, -0.0704, -0.0704, -0.1025, -0.0991, -0.0965, -0.0976, -0.0708, -0.0604, -0.0671, -0.0604, -0.056, -0.0177, -0.0196, 0.0059, 0.0038, 0.0118, 0.0, -0.0034, 0.0032, -0.0469, -0.0686, -0.0686, -0.0391, -0.0374, -0.0511, -0.0366, -0.0307, -0.009, -0.0014, -0.0061, -0.038, -0.036, -0.0396, -0.0441, -0.0522, -0.0625, -0.0537, -0.0502, -0.0503, -0.0687, -0.0892, -0.0794, -0.0742, -0.0899, -0.0934, -0.097, -0.0842, -0.0837, -0.0801, -0.0791, -0.094, -0.1019, -0.0759, -0.0692, -0.0681, -0.0681, -0.0574, -0.0769, -0.0572, -0.0515, -0.0515, -0.0562, -0.0575, -0.0298, -0.0409, -0.016, -0.0057, -0.0308, -0.0053, -0.0043, -0.013, -0.0255, -0.0142, -0.0162, -0.0119, 0.0201, 0.008, 0.0348, 0.0469, 0.0467, 0.0681, 0.0667, 0.0745, 0.0893, 0.0869, 0.0943, 0.0856, 0.0897, 0.1016, 0.1116, 0.1256, 0.1256, 0.1238, 0.133, 0.139, 0.1374, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2016699270423072783", "ticker_idx": 0, "ts": "2026-01-29T02:25:49+00:00", "symbol": "2303.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Low Risk-Reward Attractiveness… Maintain Hold Rating… mild guidance for 1Q26 (gross margin and utilization) and 2026 TAM, which implies potential stock price pullback", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF note: Hold UMC (pullback risk from mild guidance), Buy TSMC on mature-capacity shift, Buy TXN on industrial restocking recovery.", "resolved_tickers": ["2303.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "【GF Overseas Electronics Communication】\nUMC (2303 TT Hold) Analysis\n\n☄️UMC (2303 TT Hold): Low Risk-Reward Attractiveness\n\n☀️Maintain Hold Rating, Raise Target Price to NT$67: UMC's stock price has risen strongly over the past 6 weeks, reflecting wins in silicon photonics orders and improvements in the pricing environment for mature processes. Nevertheless, the earnings call focused investor attention on the 2026 outlook and pricing. Although the company did not provide 2026 gross margin guidance, management believes the pricing environment has become more favorable, but it gave mild guidance for 1Q26 (gross margin and utilization) and 2026 TAM, which implies potential stock price pullback. At the same time, the company stated that cooperation with Intel is progressing smoothly and expects silicon photonics and advanced packaging to become new drivers. As mentioned in our previous report, TSMC's (2330 TT Buy) reduction in mature process capacity is a key observation point and may gradually accelerate. Considering the earnings, guidance, and higher gross margin assumptions (partly benefiting from NT$ depreciation), we raise our 2026E/2027E earnings by 22% and 51%, respectively. Our target price of NT$67 is based on 15x 2026E/2027E P/E, and we roll our valuation method from 15x 2026E to 2026E/2027E to capture the long-term improvement in the pricing environment.\n\n☀️Improving Pricing Environment: Despite UMC's moderate overall utilization, its gross margins remain resilient, benefiting from positioning in specialty processes such as high-voltage, NVM, and BCD. From a supply perspective, given TSMC's tight advanced packaging capacity, we expect TSMC to gradually adjust its 8-inch and 12-inch lagging process capacities over the next three years, shifting toward advanced packaging. According to TrendForce, global 8-inch capacity will decline by 2.4% in 2026 and may accelerate in 2027 due to TSMC closing more 8-inch capacities. At the same time, restocking demand has emerged, as shown in TI's (TXN Buy) earnings call, where industrial sector pull-in demand has recovered. On the other hand, 28nm capacity in the industry is relatively aggressive, which has both positive and negative impacts on UMC.\n\n☀️Mild 1Q26 Guidance: 4Q25 earnings exceeded expectations, with EPS of NT$0.8, down 33% QoQ and up 19% YoY, mainly benefiting from gross margin improvement. Gross margin reached 30.7%, up 90 bps QoQ, primarily driven by NT$ depreciation and product mix. For 1Q26, the company guides wafer shipments to be flat QoQ, mainly due to seasonality, increased consumer demand, and declines in communications and automotive sectors, with USD-denominated ASP stable. Based on the current NT$ exchange rate, management expects gross margin in the high-20% range, which is below expectations, possibly due to depreciation cost impacts. For 2026, the company guides its business to outperform the addressable market, with the addressable market expected to grow in the low single digits YoY. On the other hand, depreciation expenses are expected to grow in the low-teens.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.1096491203609462, "ret_signed_m1d": -0.1096491203609462, "alpha_spy_m1d": -0.10766079230868675, "alpha_c_m1d": -0.11178071682166135, "bench_spy_m1d": 0.001988328052259458, "bench_c_m1d": -0.002131596460715146, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08771929628875696, "ret_signed_p1d": 0.08771929628875696, "alpha_spy_p1d": 0.08473671602835375, "alpha_c_p1d": 0.05404426771872761, "bench_spy_p1d": -0.0029825802604032114, "bench_c_p1d": -0.03367502857002935, "ret_p1w": -0.10087721304024111, "ret_signed_p1w": 0.10087721304024111, "alpha_spy_p1w": 0.0772186841331628, "alpha_c_p1w": 0.013576047470753694, "bench_spy_p1w": -0.023658528907078313, "bench_c_p1w": -0.08730116556948742, "ret_p1m": -0.04385964814437848, "ret_signed_p1m": 0.04385964814437848, "alpha_spy_p1m": 0.032260891475503484, "alpha_c_p1m": 0.01715435452999481, "bench_spy_p1m": -0.011598756668874999, "bench_c_p1m": -0.026705293614383674, "ret_p3m": 0.09795316957277067, "ret_signed_p3m": -0.09795316957277067, "alpha_spy_p3m": -0.06972187414635722, "alpha_c_p3m": 0.07854138010253409, "bench_spy_p3m": 0.02823129542641345, "bench_c_p3m": 0.17649454967530476, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3041, -0.307, -0.3114, -0.3304, -0.3421, -0.3392, -0.3385, -0.3421, -0.3501, -0.3545, -0.3589, -0.3538, -0.3516, -0.3472, -0.3421, -0.3194, -0.3406, -0.3355, -0.3282, -0.3304, -0.3202, -0.3099, -0.3056, -0.2982, -0.2851, -0.2792, -0.2749, -0.2785, -0.2887, -0.2865, -0.2858, -0.2807, -0.2632, -0.2646, -0.2588, -0.2558, -0.269, -0.2741, -0.2741, -0.2697, -0.2741, -0.2763, -0.28, -0.28, -0.288, -0.2909, -0.2807, -0.2091, -0.2281, -0.1944, -0.1915, -0.1974, -0.2105, -0.2003, -0.1447, -0.0863, -0.0088, 0.0, -0.0117, -0.0526, 0.0132, 0.114, 0.1096, 0.0, -0.0877, -0.1126, -0.1111, -0.0965, -0.1009, -0.095, -0.0877, -0.0877, -0.0819, -0.0819, -0.0819, -0.0819, -0.0819, -0.0819, -0.0819, -0.0819, -0.0658, -0.0132, -0.0424, -0.0439, -0.0439, -0.057, -0.0848, -0.1228, -0.0921, -0.0936, -0.1272, -0.117, -0.0906, -0.114, -0.1301, -0.1155, -0.1126, -0.1067, -0.1418, -0.155, -0.1711, -0.174, -0.136, -0.1491, -0.155, -0.1667, -0.174, -0.1608, -0.2164, -0.2164, -0.2164, -0.1477, -0.1184, -0.1228, -0.0936, -0.1213, -0.1272, -0.0921, -0.0015, 0.0673, 0.1243, 0.114, 0.1491, 0.076, 0.0892, 0.0629, 0.098, 0.0892, 0.1301, 0.1301, 0.1871, 0.2149, 0.3363, 0.4108, 0.3348, 0.3889, 0.5278, 0.4386, 0.5789, 0.6082, 0.6228, 0.652, 0.5789, 0.6959, 0.6667, 0.8275, 0.9079, 1.098, 1.076, 1.1126, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2016699270423072783", "ticker_idx": 1, "ts": "2026-01-29T02:25:49+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's (2330 TT Buy) reduction in mature process capacity is a key observation point and may gradually accelerate", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF note: Hold UMC (pullback risk from mild guidance), Buy TSMC on mature-capacity shift, Buy TXN on industrial restocking recovery.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "【GF Overseas Electronics Communication】\nUMC (2303 TT Hold) Analysis\n\n☄️UMC (2303 TT Hold): Low Risk-Reward Attractiveness\n\n☀️Maintain Hold Rating, Raise Target Price to NT$67: UMC's stock price has risen strongly over the past 6 weeks, reflecting wins in silicon photonics orders and improvements in the pricing environment for mature processes. Nevertheless, the earnings call focused investor attention on the 2026 outlook and pricing. Although the company did not provide 2026 gross margin guidance, management believes the pricing environment has become more favorable, but it gave mild guidance for 1Q26 (gross margin and utilization) and 2026 TAM, which implies potential stock price pullback. At the same time, the company stated that cooperation with Intel is progressing smoothly and expects silicon photonics and advanced packaging to become new drivers. As mentioned in our previous report, TSMC's (2330 TT Buy) reduction in mature process capacity is a key observation point and may gradually accelerate. Considering the earnings, guidance, and higher gross margin assumptions (partly benefiting from NT$ depreciation), we raise our 2026E/2027E earnings by 22% and 51%, respectively. Our target price of NT$67 is based on 15x 2026E/2027E P/E, and we roll our valuation method from 15x 2026E to 2026E/2027E to capture the long-term improvement in the pricing environment.\n\n☀️Improving Pricing Environment: Despite UMC's moderate overall utilization, its gross margins remain resilient, benefiting from positioning in specialty processes such as high-voltage, NVM, and BCD. From a supply perspective, given TSMC's tight advanced packaging capacity, we expect TSMC to gradually adjust its 8-inch and 12-inch lagging process capacities over the next three years, shifting toward advanced packaging. According to TrendForce, global 8-inch capacity will decline by 2.4% in 2026 and may accelerate in 2027 due to TSMC closing more 8-inch capacities. At the same time, restocking demand has emerged, as shown in TI's (TXN Buy) earnings call, where industrial sector pull-in demand has recovered. On the other hand, 28nm capacity in the industry is relatively aggressive, which has both positive and negative impacts on UMC.\n\n☀️Mild 1Q26 Guidance: 4Q25 earnings exceeded expectations, with EPS of NT$0.8, down 33% QoQ and up 19% YoY, mainly benefiting from gross margin improvement. Gross margin reached 30.7%, up 90 bps QoQ, primarily driven by NT$ depreciation and product mix. For 1Q26, the company guides wafer shipments to be flat QoQ, mainly due to seasonality, increased consumer demand, and declines in communications and automotive sectors, with USD-denominated ASP stable. Based on the current NT$ exchange rate, management expects gross margin in the high-20% range, which is below expectations, possibly due to depreciation cost impacts. For 2026, the company guides its business to outperform the addressable market, with the addressable market expected to grow in the low single digits YoY. On the other hand, depreciation expenses are expected to grow in the low-teens.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008310222618705287, "ret_signed_m1d": 0.008310222618705287, "alpha_spy_m1d": 0.006321894566445829, "alpha_c_m1d": 0.010441819079420434, "bench_spy_m1d": 0.001988328052259458, "bench_c_m1d": -0.002131596460715146, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.016620513087040045, "ret_signed_p1d": -0.016620513087040045, "alpha_spy_p1d": -0.013637932826636834, "alpha_c_p1d": 0.017054515482989308, "bench_spy_p1d": -0.0029825802604032114, "bench_c_p1d": -0.03367502857002935, "ret_p1w": -0.022160706732596624, "ret_signed_p1w": -0.022160706732596624, "alpha_spy_p1w": 0.0014978221744816889, "alpha_c_p1w": 0.06514045883689079, "bench_spy_p1w": -0.023658528907078313, "bench_c_p1w": -0.08730116556948742, "ret_p1m": 0.10526313646853813, "ret_signed_p1m": 0.10526313646853813, "alpha_spy_p1m": 0.11686189313741313, "alpha_c_p1m": 0.1319684300829218, "bench_spy_p1m": -0.011598756668874999, "bench_c_p1m": -0.026705293614383674, "ret_p3m": 0.23115052481009912, "ret_signed_p3m": 0.23115052481009912, "alpha_spy_p3m": 0.20291922938368567, "alpha_c_p3m": 0.054655975134794366, "bench_spy_p3m": 0.02823129542641345, "bench_c_p3m": 0.17649454967530476, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1662, -0.169, -0.1938, -0.1911, -0.1938, -0.1855, -0.1911, -0.1855, -0.1938, -0.2104, -0.2021, -0.2242, -0.2297, -0.1966, -0.2352, -0.2408, -0.2187, -0.2049, -0.2076, -0.2049, -0.2214, -0.2104, -0.1993, -0.2021, -0.1938, -0.1745, -0.1828, -0.169, -0.1856, -0.1801, -0.1967, -0.205, -0.2078, -0.2078, -0.2078, -0.1884, -0.1745, -0.1717, -0.1717, -0.1634, -0.1524, -0.1579, -0.1413, -0.1413, -0.1219, -0.0748, -0.0554, -0.072, -0.0665, -0.0693, -0.0637, -0.0526, -0.0526, -0.0637, -0.036, -0.0249, -0.0166, -0.036, -0.0249, -0.0194, -0.0277, -0.0139, 0.0083, 0.0, -0.0166, -0.0222, -0.0028, -0.0111, -0.0222, -0.0139, 0.0055, 0.0416, 0.0609, 0.0609, 0.0609, 0.0609, 0.0609, 0.0609, 0.0609, 0.0609, 0.0526, 0.0886, 0.1163, 0.1053, 0.1053, 0.0942, 0.072, 0.0332, 0.0526, 0.0471, 0.0028, 0.0249, 0.0748, 0.0443, 0.0332, 0.0222, 0.0394, 0.0588, 0.0283, 0.0227, 0.006, 0.006, 0.0255, 0.0227, 0.0116, -0.0106, -0.0217, 0.0311, 0.006, 0.006, 0.006, 0.0338, 0.0839, 0.0866, 0.1116, 0.1061, 0.1422, 0.1561, 0.1589, 0.1283, 0.1255, 0.1394, 0.1394, 0.1561, 0.2145, 0.2589, 0.2312, 0.2117, 0.1867, 0.1867, 0.2645, 0.2506, 0.2506, 0.284, 0.2728, 0.2423, 0.2534, 0.2339, 0.2617, 0.2589, 0.245, 0.2256, 0.2145, 0.2395, 0.2534, 0.284, 0.2617, 0.2784, 0.2756, 0.309, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2016699270423072783", "ticker_idx": 2, "ts": "2026-01-29T02:25:49+00:00", "symbol": "TXN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TI's (TXN Buy) earnings call, where industrial sector pull-in demand has recovered", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares GF note: Hold UMC (pullback risk from mild guidance), Buy TSMC on mature-capacity shift, Buy TXN on industrial restocking recovery.", "resolved_tickers": ["TXN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "【GF Overseas Electronics Communication】\nUMC (2303 TT Hold) Analysis\n\n☄️UMC (2303 TT Hold): Low Risk-Reward Attractiveness\n\n☀️Maintain Hold Rating, Raise Target Price to NT$67: UMC's stock price has risen strongly over the past 6 weeks, reflecting wins in silicon photonics orders and improvements in the pricing environment for mature processes. Nevertheless, the earnings call focused investor attention on the 2026 outlook and pricing. Although the company did not provide 2026 gross margin guidance, management believes the pricing environment has become more favorable, but it gave mild guidance for 1Q26 (gross margin and utilization) and 2026 TAM, which implies potential stock price pullback. At the same time, the company stated that cooperation with Intel is progressing smoothly and expects silicon photonics and advanced packaging to become new drivers. As mentioned in our previous report, TSMC's (2330 TT Buy) reduction in mature process capacity is a key observation point and may gradually accelerate. Considering the earnings, guidance, and higher gross margin assumptions (partly benefiting from NT$ depreciation), we raise our 2026E/2027E earnings by 22% and 51%, respectively. Our target price of NT$67 is based on 15x 2026E/2027E P/E, and we roll our valuation method from 15x 2026E to 2026E/2027E to capture the long-term improvement in the pricing environment.\n\n☀️Improving Pricing Environment: Despite UMC's moderate overall utilization, its gross margins remain resilient, benefiting from positioning in specialty processes such as high-voltage, NVM, and BCD. From a supply perspective, given TSMC's tight advanced packaging capacity, we expect TSMC to gradually adjust its 8-inch and 12-inch lagging process capacities over the next three years, shifting toward advanced packaging. According to TrendForce, global 8-inch capacity will decline by 2.4% in 2026 and may accelerate in 2027 due to TSMC closing more 8-inch capacities. At the same time, restocking demand has emerged, as shown in TI's (TXN Buy) earnings call, where industrial sector pull-in demand has recovered. On the other hand, 28nm capacity in the industry is relatively aggressive, which has both positive and negative impacts on UMC.\n\n☀️Mild 1Q26 Guidance: 4Q25 earnings exceeded expectations, with EPS of NT$0.8, down 33% QoQ and up 19% YoY, mainly benefiting from gross margin improvement. Gross margin reached 30.7%, up 90 bps QoQ, primarily driven by NT$ depreciation and product mix. For 1Q26, the company guides wafer shipments to be flat QoQ, mainly due to seasonality, increased consumer demand, and declines in communications and automotive sectors, with USD-denominated ASP stable. Based on the current NT$ exchange rate, management expects gross margin in the high-20% range, which is below expectations, possibly due to depreciation cost impacts. For 2026, the company guides its business to outperform the addressable market, with the addressable market expected to grow in the low single digits YoY. On the other hand, depreciation expenses are expected to grow in the low-teens.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01278712862911413, "ret_signed_m1d": -0.01278712862911413, "alpha_spy_m1d": -0.014775456681373589, "alpha_c_m1d": -0.010655532168398985, "bench_spy_m1d": 0.001988328052259458, "bench_c_m1d": -0.002131596460715146, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009193262311353756, "ret_signed_p1d": -0.009193262311353756, "alpha_spy_p1d": -0.0062106820509505445, "alpha_c_p1d": 0.024481766258675597, "bench_spy_p1d": -0.0029825802604032114, "bench_c_p1d": -0.03367502857002935, "ret_p1w": 0.029556483904441766, "ret_signed_p1w": 0.029556483904441766, "alpha_spy_p1w": 0.05321501281152008, "alpha_c_p1w": 0.11685764947392918, "bench_spy_p1w": -0.023658528907078313, "bench_c_p1w": -0.08730116556948742, "ret_p1m": -0.025005710144600624, "ret_signed_p1m": -0.025005710144600624, "alpha_spy_p1m": -0.013406953475725625, "alpha_c_p1m": 0.0016995834697830503, "bench_spy_p1m": -0.011598756668874999, "bench_c_p1m": -0.026705293614383674, "ret_p3m": 0.21811088677840695, "ret_signed_p3m": 0.21811088677840695, "alpha_spy_p3m": 0.1898795913519935, "alpha_c_p3m": 0.0416163371031022, "bench_spy_p3m": 0.02823129542641345, "bench_c_p3m": 0.17649454967530476, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2626, -0.2722, -0.253, -0.263, -0.2668, -0.2667, -0.2705, -0.2552, -0.2591, -0.2724, -0.2922, -0.2815, -0.2826, -0.2998, -0.272, -0.2636, -0.2612, -0.2449, -0.2449, -0.2315, -0.232, -0.1996, -0.1661, -0.1774, -0.1664, -0.1737, -0.1802, -0.1703, -0.1703, -0.1806, -0.1872, -0.1891, -0.2031, -0.1954, -0.1949, -0.1834, -0.1913, -0.1911, -0.1911, -0.1922, -0.1977, -0.1989, -0.2077, -0.2077, -0.1893, -0.1909, -0.1227, -0.1519, -0.1394, -0.1309, -0.1365, -0.139, -0.1165, -0.1363, -0.1251, -0.1251, -0.1342, -0.1122, -0.1095, -0.1172, -0.1022, -0.102, -0.0128, 0.0, -0.0092, 0.0343, 0.0352, 0.0247, 0.0296, 0.0179, 0.0056, 0.0155, 0.0414, 0.0251, 0.0396, 0.0396, 0.0374, 0.0265, 0.0023, 0.01, 0.0106, -0.0193, -0.0168, -0.0226, -0.025, -0.0355, -0.0684, -0.0697, -0.09, -0.1118, -0.0981, -0.0923, -0.0868, -0.1264, -0.1231, -0.1077, -0.1062, -0.1231, -0.1345, -0.1396, -0.1329, -0.1054, -0.0955, -0.111, -0.1251, -0.1431, -0.1076, -0.0977, -0.1043, -0.1043, -0.0833, -0.0819, -0.0398, -0.0118, -0.013, -0.0039, 0.0061, -0.0058, 0.0255, 0.0564, 0.0742, 0.0717, 0.0862, 0.2973, 0.2739, 0.2388, 0.2181, 0.2375, 0.292, 0.2917, 0.2912, 0.2982, 0.3372, 0.3178, 0.3296, 0.3757, 0.3637, 0.4153, 0.4237, 0.3986, 0.3888, 0.3967, 0.4085, 0.3786, 0.4286, 0.4286, 0.501, 0.4666, 0.4597, 0.4122, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008403337423843575", "ticker_idx": 0, "ts": "2026-01-06T05:00:44+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Rubin is utilizing top-tier, high-spec HBM4... Nvidia's strategy of using fewer units of high-spec HBM4", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA Rubin superior on HBM4 bandwidth and cost structure; AMD MI455X faces higher costs, yield risks, and supply chain disadvantage with 12-stack HBM4.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts on NVIDIA Rubin and AMD MI455X\n\nLet’s dive into a comparison between Nvidia’s Rubin and AMD’s MI455X, both unveiled today.\n\nStarting with Rubin, it utilizes an 8-stack HBM4 configuration. It boasts a memory bandwidth of 22TB/s, leveraging memory with a per-pin Fmax of around 10.7Gbps.\n\nOn the flip side, the MI455X opts for a 12-stack HBM4 setup. However, it delivers a bandwidth of 19.6TB/s, using memory with a per-pin Fmax of roughly 6.4Gbps.\n\nConsidering the current JEDEC standard for HBM4 is 8Gbps, the difference is stark: Rubin is utilizing top-tier, high-spec HBM4, while the MI455X appears to be relying on HBM4 that falls below the standard spec.\n\nThis highlights a distinct divergence in corporate strategy: Using top-tier components vs. Brute-forcing capacity.\n\nAMD likely adopted this approach because securing top-speed HBM4 volume is challenging for them. However, this strategy carries two significant risks.\n\nFirst, the cost and yield implications. Mounting more HBM stacks requires a larger interposer area, which directly drives up unit costs. Furthermore, a larger footprint inevitably lowers the yield for 2.5D packaging assembly. In other words, the strategy of using more units of lower-spec HBM4 could paradoxically end up being more costly than Nvidia’s strategy of using fewer units of high-spec HBM4.\n\nSecond, the impact during memory shortages. This approach exacerbates supply chain bottlenecks. A 12-stack configuration consumes 50% more HBM chiplets/stacks per GPU compared to an 8-stack design. The tighter the global HBM4 supply, the more AMD’s shipment volume becomes capped by memory availability.\n\nOf course, in the early stages where yields for high-spec HBM4 are low, this isn't a major issue—low yields for top-bin parts naturally result in an abundance of lower-binned supply.\n\nBut what happens as the yield learning curve improves? As yields for high-spec HBM4 rise, suppliers will have more incentive to allocate wafers to the higher-margin chips destined for Nvidia. This makes it increasingly difficult for AMD to source large volumes of low-performance HBM4 at low prices. Furthermore, with Samsung performing well in the HBM4 space, AMD won't be able to pick up inventory at \"clearance\" prices like they did during the HBM3E cycle.\n\nUltimately, AMD is facing an inherently more disadvantageous cost structure at the chip level compared to Nvidia's Rubin.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004699804739852764, "ret_signed_m1d": 0.004699804739852764, "alpha_spy_m1d": 0.010611828225168818, "alpha_c_m1d": 0.030575659715457282, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.025875854975604518, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009987125820996923, "ret_signed_p1d": 0.009987125820996923, "alpha_spy_p1d": 0.013210482613224328, "alpha_c_p1d": 0.016694741878650388, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.006707616057653465, "ret_p1w": -0.007637304948690304, "ret_signed_p1w": -0.007637304948690304, "alpha_spy_p1w": -0.010470525517025986, "alpha_c_p1w": -0.01772451332638847, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": 0.010087208377698165, "ret_p1m": -0.06969668297352538, "ret_signed_p1m": -0.06969668297352538, "alpha_spy_p1m": -0.06157309135644695, "alpha_c_p1m": -0.05524952809488559, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.01444715487863979, "ret_p3m": -0.05255504410365941, "ret_signed_p3m": -0.05255504410365941, "alpha_spy_p3m": -0.0031355774207620923, "alpha_c_p3m": -0.06468035308213549, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": 0.01212530897847608, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0284, -0.0218, 0.0057, -0.0386, -0.0396, -0.0291, -0.0215, -0.0246, -0.0325, -0.0372, -0.0272, -0.0053, 0.0226, 0.0736, 0.1057, 0.0835, 0.0814, 0.1048, 0.0611, 0.0425, 0.0044, 0.0048, 0.063, 0.0316, 0.035, -0.0021, 0.0156, -0.0035, -0.0315, -0.0039, -0.0353, -0.0447, -0.0251, -0.0504, -0.0373, -0.0373, -0.0547, -0.0391, -0.0309, -0.0409, -0.0206, -0.0258, -0.009, -0.0121, -0.0185, -0.0337, -0.0653, -0.0585, -0.0508, -0.0871, -0.07, -0.0334, -0.019, 0.0105, 0.0073, 0.0073, 0.0176, 0.0052, 0.0016, -0.004, -0.004, 0.0086, 0.0047, 0.0, 0.01, -0.0117, -0.0127, -0.0123, -0.0076, -0.0219, -0.001, -0.0054, -0.0054, -0.049, -0.0209, -0.0128, 0.0023, -0.0041, 0.0068, 0.0229, 0.0281, 0.0208, -0.0087, -0.0369, -0.0697, -0.082, -0.0098, 0.015, 0.0069, 0.015, -0.0016, -0.0237, -0.0237, -0.0121, 0.004, 0.0035, 0.0138, 0.023, 0.03, 0.0444, -0.0126, -0.0537, -0.0254, -0.0384, -0.0224, -0.0208, -0.0503, -0.0245, -0.0132, -0.0064, -0.0218, -0.0373, -0.0214, -0.0283, -0.0365, -0.0463, -0.0776, -0.0619, -0.0643, -0.0457, -0.0854, -0.1053, -0.1178, -0.0685, -0.0613, -0.0526, -0.0526, -0.0512, -0.0488, -0.0275, -0.0177, 0.0075, 0.0111, 0.0496, 0.0622, 0.0594, 0.0772, 0.0792, 0.0676, 0.0816, 0.0663, 0.1124, 0.1569, 0.1385, 0.1176, 0.0659, 0.0599, 0.0601, 0.0495, 0.11, 0.1296, 0.1494, 0.172, 0.1792, 0.2062, 0.2591, 0.2034, 0.1874, 0.1783, 0.1936, 0.1724, 0.1501, 0.1501, 0.1476, 0.1355, 0.1443, 0.1277, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008403337423843575", "ticker_idx": 1, "ts": "2026-01-06T05:00:44+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD is facing an inherently more disadvantageous cost structure at the chip level compared to Nvidia's Rubin", "tweet_type": "analysis", "timeline": "unspecified", "summary": "NVDA Rubin superior on HBM4 bandwidth and cost structure; AMD MI455X faces higher costs, yield risks, and supply chain disadvantage with 12-stack HBM4.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts on NVIDIA Rubin and AMD MI455X\n\nLet’s dive into a comparison between Nvidia’s Rubin and AMD’s MI455X, both unveiled today.\n\nStarting with Rubin, it utilizes an 8-stack HBM4 configuration. It boasts a memory bandwidth of 22TB/s, leveraging memory with a per-pin Fmax of around 10.7Gbps.\n\nOn the flip side, the MI455X opts for a 12-stack HBM4 setup. However, it delivers a bandwidth of 19.6TB/s, using memory with a per-pin Fmax of roughly 6.4Gbps.\n\nConsidering the current JEDEC standard for HBM4 is 8Gbps, the difference is stark: Rubin is utilizing top-tier, high-spec HBM4, while the MI455X appears to be relying on HBM4 that falls below the standard spec.\n\nThis highlights a distinct divergence in corporate strategy: Using top-tier components vs. Brute-forcing capacity.\n\nAMD likely adopted this approach because securing top-speed HBM4 volume is challenging for them. However, this strategy carries two significant risks.\n\nFirst, the cost and yield implications. Mounting more HBM stacks requires a larger interposer area, which directly drives up unit costs. Furthermore, a larger footprint inevitably lowers the yield for 2.5D packaging assembly. In other words, the strategy of using more units of lower-spec HBM4 could paradoxically end up being more costly than Nvidia’s strategy of using fewer units of high-spec HBM4.\n\nSecond, the impact during memory shortages. This approach exacerbates supply chain bottlenecks. A 12-stack configuration consumes 50% more HBM chiplets/stacks per GPU compared to an 8-stack design. The tighter the global HBM4 supply, the more AMD’s shipment volume becomes capped by memory availability.\n\nOf course, in the early stages where yields for high-spec HBM4 are low, this isn't a major issue—low yields for top-bin parts naturally result in an abundance of lower-binned supply.\n\nBut what happens as the yield learning curve improves? As yields for high-spec HBM4 rise, suppliers will have more incentive to allocate wafers to the higher-margin chips destined for Nvidia. This makes it increasingly difficult for AMD to source large volumes of low-performance HBM4 at low prices. Furthermore, with Samsung performing well in the HBM4 space, AMD won't be able to pick up inventory at \"clearance\" prices like they did during the HBM3E cycle.\n\nUltimately, AMD is facing an inherently more disadvantageous cost structure at the chip level compared to Nvidia's Rubin.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03139722666622635, "ret_signed_m1d": -0.03139722666622635, "alpha_spy_m1d": -0.037309250151542406, "alpha_c_m1d": -0.05727308164183087, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.025875854975604518, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020200614451877374, "ret_signed_p1d": 0.020200614451877374, "alpha_spy_p1d": 0.01697725765964997, "alpha_c_p1d": 0.01349299839422391, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.006707616057653465, "ret_p1w": 0.030884044453866277, "ret_signed_p1w": -0.030884044453866277, "alpha_spy_p1w": -0.028050823885530596, "alpha_c_p1w": -0.020796836076168113, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": 0.010087208377698165, "ret_p1m": -0.06606019714910161, "ret_signed_p1m": 0.06606019714910161, "alpha_spy_p1m": 0.05793660553202318, "alpha_c_p1m": 0.05161304227046182, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.01444715487863979, "ret_p3m": 0.014695562429623443, "ret_signed_p3m": -0.014695562429623443, "alpha_spy_p3m": -0.06411502911252076, "alpha_c_p3m": -0.002570253451147364, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": 0.01212530897847608, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0865, 0.0026, 0.0097, 0.0174, 0.1131, 0.0943, 0.0874, 0.1223, 0.1105, 0.0741, 0.0963, 0.1799, 0.2114, 0.2037, 0.2332, 0.1889, 0.1949, 0.2113, 0.1666, 0.1958, 0.1089, 0.0895, 0.1382, 0.1081, 0.2078, 0.1568, 0.1514, 0.1221, 0.0744, 0.0429, -0.0389, -0.0493, 0.0033, -0.0383, -0.0005, -0.0005, 0.0148, 0.0252, 0.0042, 0.0152, 0.0076, 0.0169, 0.0315, 0.0339, 0.033, 0.033, -0.0167, -0.0316, -0.0242, -0.0758, -0.062, -0.0043, 0.0028, 0.0026, 0.0032, 0.0032, 0.003, 0.0059, 0.0046, -0.0009, -0.0009, 0.0425, 0.0314, 0.0, -0.0202, -0.0451, -0.0522, -0.0311, 0.0309, 0.0432, 0.0633, 0.0815, 0.0815, 0.082, 0.1654, 0.1837, 0.2115, 0.1724, 0.1758, 0.1791, 0.1765, 0.1044, 0.1489, 0.1295, -0.0661, -0.1019, -0.0276, 0.0077, -0.0036, -0.0036, -0.0392, -0.0328, -0.0328, -0.0526, -0.0664, -0.0512, -0.0662, -0.0828, -0.0024, -0.0163, -0.0498, -0.066, -0.0734, -0.1092, -0.0573, -0.0695, -0.1023, -0.0544, -0.0519, -0.0444, -0.0775, -0.0978, -0.0829, -0.0842, -0.0695, -0.0424, -0.0607, -0.0544, -0.0419, 0.0276, -0.0494, -0.0577, -0.0854, -0.0509, -0.0193, 0.0147, 0.0147, 0.0272, 0.0335, 0.0815, 0.104, 0.1432, 0.1515, 0.19, 0.2042, 0.2982, 0.2988, 0.2827, 0.3272, 0.4157, 0.4244, 0.6226, 0.5611, 0.5079, 0.5727, 0.6538, 0.682, 0.5934, 0.6574, 0.9659, 0.9056, 1.1236, 1.1404, 1.0914, 1.0784, 1.098, 0.9785, 0.964, 0.9317, 1.0881, 1.0975, 1.1811, 1.1811, 1.3508, 1.3118, 1.417, 1.4077, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2017918366544216513", "ticker_idx": 0, "ts": "2026-02-01T11:10:04+00:00", "symbol": "HPSP", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Something... something is happening.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish on HPSP: connects Thiel/Vance/Intel political ties, Intel fab proximity, ex-Intel exec appointment, and Thiel fund commitment to hold as signals of an undisclosed catalyst.", "resolved_tickers": ["403870.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "HPSP Co. 403870.KQ Korea KOSDAQ HP annealing equipment", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "quote", "lang": "en", "tweet_text": "I'm also researching HPSP.\n\nPeter Thiel sponsors JD Vance.\n\nJD Vance represents Ohio as his district, where Intel's largest fab is located.\n\nHPSP has unusually high sales to Intel among Korean semiconductor equipment companies.\n\nPeter Thiel's fund appointed Lee Chun-heung, the former senior vice president in charge of Intel's packaging, to HPSP.\n\nPeter Thiel's fund, Cresando, has stated that it will not sell HPSP but will continue to hold it.\n\nSomething... something is happening.", "reply_text": null, "parent_text": null, "quoted_text": "I learned of this niche Korean semiconductor equipment firm called HPSP that has a virtual monopoly in high-pressure
hydrogen annealing equipment which is used by the most advanced nodes. The most interesting part is Peter Thiel is the biggest investor.\n\nhttps://t.co/PQVhOUVoPJ", "parent_author": null, "ret_m1d": 0.09026963657678788, "ret_signed_m1d": 0.09026963657678788, "alpha_spy_m1d": 0.09521640652294261, "alpha_c_m1d": 0.10274497334036847, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.012475336763580591, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09730363423212185, "ret_signed_p1d": 0.09730363423212185, "alpha_spy_p1d": 0.10575903066282244, "alpha_c_p1d": 0.0888872163214014, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": 0.008416417910720453, "ret_p1w": 0.08909730363423218, "ret_signed_p1w": 0.08909730363423218, "alpha_spy_p1w": 0.09119674138447276, "alpha_c_p1w": 0.052268088406038915, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": 0.03682921522819327, "ret_p1m": 0.01289566236811246, "ret_signed_p1m": 0.01289566236811246, "alpha_spy_p1m": 0.034580640919612216, "alpha_c_p1m": -0.03431974155729023, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": 0.04721540392540269, "ret_p3m": 0.23329425556858152, "ret_signed_p3m": 0.23329425556858152, "alpha_spy_p3m": 0.1970384485616079, "alpha_c_p3m": 0.18828408886236603, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.04501016670621549, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.248, -0.2318, -0.2711, -0.2491, -0.2572, -0.2468, -0.2388, -0.2826, -0.2711, -0.3103, -0.3299, -0.3045, -0.346, -0.3495, -0.3483, -0.3356, -0.3356, -0.3022, -0.278, -0.2791, -0.2849, -0.2953, -0.301, -0.2987, -0.3057, -0.316, -0.3207, -0.3207, -0.3345, -0.3506, -0.323, -0.3403, -0.338, -0.3091, -0.2987, -0.3068, -0.3068, -0.286, -0.2216, -0.2145, -0.2145, -0.2145, -0.1923, -0.1794, -0.0821, -0.2216, -0.1934, -0.2239, -0.2122, -0.2157, -0.2005, -0.1911, -0.1618, -0.177, -0.2075, -0.2298, -0.2204, -0.2028, -0.1594, -0.1196, 0.0317, 0.0399, 0.0903, 0.0, 0.0973, 0.075, 0.0574, 0.075, 0.0891, 0.0387, 0.0129, -0.0035, -0.0012, -0.0012, -0.0012, -0.0012, 0.0844, 0.0762, 0.0551, -0.0399, -0.0387, 0.0152, 0.0516, 0.0516, 0.0129, -0.1102, 0.0539, 0.1219, 0.0199, 0.0715, 0.0223, 0.0141, 0.0, 0.0281, 0.0328, 0.1723, 0.2122, 0.2005, 0.1536, 0.143, 0.1665, 0.0703, 0.0586, 0.0059, -0.041, 0.0434, -0.0422, -0.0152, -0.0246, -0.0234, 0.0762, 0.0481, 0.068, -0.0293, 0.0, -0.0023, 0.0352, 0.0434, 0.0715, 0.0926, 0.1254, 0.1958, 0.2521, 0.2849, 0.2591, 0.2521, 0.2333, 0.2333, 0.2755, 0.2755, 0.3271, 0.2825, 0.2638, 0.2989, 0.2028, 0.2966, 0.2825, 0.1606, 0.2239, 0.1489, 0.1348, 0.2521, 0.2825, 0.2825, 0.4091, 0.2943, 0.1934, 0.1407, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2017060282963808368", "ticker_idx": 0, "ts": "2026-01-30T02:20:21+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "all bad except BESI", "tweet_type": "analysis", "timeline": "unspecified", "summary": "BESI singled out as the only good bonding company; implied bullish stance.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@RealPoliBear I’m not investing. Bonding companies are all bad except BESI.", "reply_text": null, "parent_text": "@jukan05 Hanwha Vision", "quoted_text": null, "parent_author": "RealPoliBear", "ret_m1d": -0.011854142014292135, "ret_signed_m1d": -0.011854142014292135, "alpha_spy_m1d": -0.014845644671469538, "alpha_c_m1d": -0.04670269665755766, "bench_spy_m1d": 0.002991502657177403, "bench_c_m1d": 0.03484855464326553, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0006078520634440476, "ret_signed_p1d": 0.0006078520634440476, "alpha_spy_p1d": -0.004363510067491472, "alpha_c_p1d": -0.010570480016918182, "bench_spy_p1d": 0.004971362130935519, "bench_c_p1d": 0.01117833208036223, "ret_p1w": 0.006078987351319176, "ret_signed_p1w": 0.006078987351319176, "alpha_spy_p1w": 0.008029925044960251, "alpha_c_p1w": 0.010565175813441519, "bench_spy_p1w": -0.001950937693641075, "bench_c_p1w": -0.0044861884621223425, "ret_p1m": 0.14589662986541851, "ret_signed_p1m": 0.14589662986541851, "alpha_spy_p1m": 0.15397491643656724, "alpha_c_p1m": 0.13863439010084555, "bench_spy_p1m": -0.00807828657114873, "bench_c_p1m": 0.007262239764572964, "ret_p3m": 0.46449390668444046, "ret_signed_p3m": 0.46449390668444046, "alpha_spy_p3m": 0.4333460339565469, "alpha_c_p3m": 0.22625468308625707, "bench_spy_p3m": 0.031147872727893544, "bench_c_p3m": 0.2382392235981834, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1334, -0.1398, -0.1626, -0.1845, -0.1705, -0.1623, -0.1772, -0.1866, -0.2012, -0.2085, -0.2271, -0.2152, -0.2094, -0.2535, -0.2438, -0.2426, -0.2122, -0.2103, -0.2103, -0.2112, -0.2015, -0.1736, -0.1641, -0.1517, -0.1073, -0.1289, -0.1462, -0.1678, -0.197, -0.1948, -0.1878, -0.214, -0.2009, -0.2052, -0.2024, -0.1982, -0.2003, -0.2003, -0.2003, -0.197, -0.1909, -0.1869, -0.1869, -0.0936, -0.0653, -0.0219, -0.0347, -0.0815, -0.0815, -0.0143, 0.0164, -0.0182, 0.0556, 0.0511, 0.0261, 0.0511, 0.0581, 0.0638, 0.0641, 0.0629, 0.0684, 0.0301, -0.0119, 0.0, 0.0006, -0.0264, -0.0422, -0.0228, 0.0061, 0.0176, 0.0277, 0.0389, 0.0255, 0.0711, 0.0851, 0.0948, 0.1368, 0.0517, 0.1216, 0.1407, 0.169, 0.2012, 0.1483, 0.1514, 0.1459, 0.1024, 0.1611, 0.1468, -0.0498, 0.0109, 0.0614, 0.0635, 0.0653, 0.1252, 0.1125, 0.1456, 0.1696, 0.1176, 0.0851, 0.1097, 0.1191, 0.1277, 0.1271, 0.0629, 0.059, 0.0878, 0.1401, 0.1559, 0.1559, 0.1559, 0.1632, 0.2584, 0.2529, 0.2857, 0.2578, 0.3252, 0.3228, 0.3404, 0.3787, 0.3799, 0.3854, 0.4122, 0.4699, 0.5331, 0.5171, 0.4394, 0.4645, 0.5122, 0.5122, 0.4945, 0.5538, 0.5599, 0.5581, 0.5954, 0.5985, 0.5281, 0.5795, 0.6315, 0.6015, 0.5648, 0.5593, 0.6211, 0.6541, 0.6737, 0.7251, 0.7373, 0.6908, 0.7594, 0.7398, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008010928638636058", "ticker_idx": 0, "ts": "2026-01-05T03:01:27+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe Google has meaningfully raised the volume expectations on MediaTek's TPU v7e for 2026 and 2027. We forecast MediaTek's TPU sales to reach US$0.8bn in 2026, US$4bn in 2027 and US$5bn in 2028... we see notable upside potential for our revenue estimates", "tweet_type": "news_reaction", "timeline": "years", "summary": "UBS bullish on MediaTek on Google TPU v7e volume ramp; TSMC benefits as foundry for trial production.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: MediaTek’s TPU-related update:\n\nWe have written extensively in recent months on Google and MediaTek's ongoing progress for TPU v7e (link, link). While many investors remain skeptical about execution,\nthe project made a meaningful breakthrough for tape out in early Oct 2025. With fast growing TPU demand, which may exceed the capacity that Broadcom could secure, we believe Google has meaningfully raised the volume expectations on MediaTek's TPU v7e for 2026 and 2027. We forecast MediaTek's TPU sales to reach US$0.8bn in 2026 (vs guidance of US$1bn), US$4bn in 2027 and US$5bn in 2028. Assuming TPU demand to remain robust and no major performance issues for the chips, we see notable upside potential for our revenue estimates—especially for 2027 and beyond. Our current\nestimates imply MediaTek's TPU shipments to reach ~1mn units in 2027E. In an upside scenario, volumes could be in the multimillions, in our view.\n\nTPU v7e is now through trial production at TSMC. We believe samples could be delivered to Google and MediaTek by Q126 for performance and functionality tests. Chip performance and competitiveness would be key to determining the overall project volume, in our view.\n\nBy its nature, ASIC / TPU business should have lower GM vs MediaTek's corp GM of mid 40%, but we are optimistic about the strong operating leverage if MediaTek could ship >1mn units of TPU in 2027. We estimate OPM of the project should be well above MediaTek's corp OPM of ~17% in 2025.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03606557591076198, "ret_signed_m1d": -0.03606557591076198, "alpha_spy_m1d": -0.02944952574183668, "alpha_c_m1d": -0.024704023468882896, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007352941176470562, "ret_signed_p1d": -0.007352941176470562, "alpha_spy_p1d": -0.013300124549559866, "alpha_c_p1d": -0.03391614167588575, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": -0.03409090909090906, "ret_signed_p1w": -0.03409090909090906, "alpha_spy_p1w": -0.044909268861508145, "alpha_c_p1w": -0.0687051929728445, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.19986631016042788, "ret_signed_p1m": 0.19986631016042788, "alpha_spy_p1m": 0.1972343602553115, "alpha_c_p1m": 0.14666046113705256, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": -0.020721925133689867, "ret_signed_p3m": -0.020721925133689867, "alpha_spy_p3m": 0.023044264806681647, "alpha_c_p3m": -0.05973252162509368, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1246, -0.118, -0.118, -0.1377, -0.1443, -0.1377, -0.1279, -0.1279, -0.1213, -0.1213, -0.1279, -0.1508, -0.1508, -0.1246, -0.1377, -0.1475, -0.141, -0.141, -0.1475, -0.1377, -0.1279, -0.1541, -0.1738, -0.1803, -0.1869, -0.1836, -0.1836, -0.1934, -0.1934, -0.2328, -0.2393, -0.223, -0.2492, -0.2459, -0.223, -0.1475, -0.1213, -0.0852, -0.0525, -0.0721, -0.082, -0.0787, -0.0656, -0.0557, -0.0689, -0.0426, -0.0852, -0.0787, -0.0689, -0.0689, -0.0656, -0.0689, -0.0754, -0.082, -0.0885, -0.0951, -0.0951, -0.0918, -0.0689, -0.0689, -0.0623, -0.0623, -0.0361, 0.0, -0.0074, -0.004, -0.0341, -0.0508, -0.0341, -0.0074, 0.0027, -0.004, 0.006, -0.0074, -0.0074, -0.0207, -0.0074, 0.0896, 0.1832, 0.1865, 0.1898, 0.1898, 0.1765, 0.1397, 0.1999, 0.2032, 0.1832, 0.143, 0.2233, 0.2333, 0.24, 0.24, 0.24, 0.24, 0.24, 0.24, 0.24, 0.24, 0.1999, 0.2199, 0.25, 0.3001, 0.3001, 0.2701, 0.2132, 0.1497, 0.1865, 0.1798, 0.113, 0.1397, 0.1798, 0.1932, 0.1497, 0.143, 0.1564, 0.1564, 0.123, 0.1364, 0.0862, 0.0829, 0.0829, 0.0628, 0.0595, 0.0094, -0.004, -0.0207, -0.0207, -0.0207, -0.0207, -0.0174, 0.0561, 0.0528, 0.0528, 0.0829, 0.1497, 0.1965, 0.2667, 0.2868, 0.2701, 0.3971, 0.5341, 0.4806, 0.6277, 0.6277, 0.748, 0.7213, 0.7447, 0.7447, 0.9184, 1.109, 1.2928, 1.2861, 1.4265, 1.5936, 1.4733, 1.3362, 1.2761, 1.1791, 1.2727, 1.109, 1.1591, 1.373, 1.5802, 1.8376, 1.8509, 2.1016, 1.9479, 1.881, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008010928638636058", "ticker_idx": 1, "ts": "2026-01-05T03:01:27+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TPU v7e is now through trial production at TSMC. We believe samples could be delivered to Google and MediaTek by Q126", "tweet_type": "news_reaction", "timeline": "years", "summary": "UBS bullish on MediaTek on Google TPU v7e volume ramp; TSMC benefits as foundry for trial production.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: MediaTek’s TPU-related update:\n\nWe have written extensively in recent months on Google and MediaTek's ongoing progress for TPU v7e (link, link). While many investors remain skeptical about execution,\nthe project made a meaningful breakthrough for tape out in early Oct 2025. With fast growing TPU demand, which may exceed the capacity that Broadcom could secure, we believe Google has meaningfully raised the volume expectations on MediaTek's TPU v7e for 2026 and 2027. We forecast MediaTek's TPU sales to reach US$0.8bn in 2026 (vs guidance of US$1bn), US$4bn in 2027 and US$5bn in 2028. Assuming TPU demand to remain robust and no major performance issues for the chips, we see notable upside potential for our revenue estimates—especially for 2027 and beyond. Our current\nestimates imply MediaTek's TPU shipments to reach ~1mn units in 2027E. In an upside scenario, volumes could be in the multimillions, in our view.\n\nTPU v7e is now through trial production at TSMC. We believe samples could be delivered to Google and MediaTek by Q126 for performance and functionality tests. Chip performance and competitiveness would be key to determining the overall project volume, in our view.\n\nBy its nature, ASIC / TPU business should have lower GM vs MediaTek's corp GM of mid 40%, but we are optimistic about the strong operating leverage if MediaTek could ship >1mn units of TPU in 2027. We estimate OPM of the project should be well above MediaTek's corp OPM of ~17% in 2025.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05089816407123715, "ret_signed_m1d": -0.05089816407123715, "alpha_spy_m1d": -0.04428211390231185, "alpha_c_m1d": -0.039536611629358065, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020958115010468514, "ret_signed_p1d": 0.020958115010468514, "alpha_spy_p1d": 0.01501093163737921, "alpha_c_p1d": -0.005605085488946671, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.011976034291203863, "ret_signed_p1w": 0.011976034291203863, "alpha_spy_p1w": 0.001157674520604779, "alpha_c_p1w": -0.022638249590731574, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.07784433289454862, "ret_signed_p1m": 0.07784433289454862, "alpha_spy_p1m": 0.07521238298943222, "alpha_c_p1m": 0.02463848387117329, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.08736850210653291, "ret_signed_p3m": 0.08736850210653291, "alpha_spy_p3m": 0.13113469204690442, "alpha_c_p3m": 0.04835790561512909, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1555, -0.1406, -0.1406, -0.1555, -0.1495, -0.1257, -0.1137, -0.1346, -0.1167, -0.1167, -0.1287, -0.1346, -0.1346, -0.1167, -0.1197, -0.1018, -0.1018, -0.1048, -0.0988, -0.1018, -0.1287, -0.1257, -0.1287, -0.1197, -0.1257, -0.1197, -0.1287, -0.1466, -0.1376, -0.1615, -0.1674, -0.1316, -0.1734, -0.1794, -0.1555, -0.1406, -0.1436, -0.1406, -0.1585, -0.1466, -0.1346, -0.1376, -0.1287, -0.1078, -0.1167, -0.1018, -0.1198, -0.1138, -0.1317, -0.1407, -0.1437, -0.1437, -0.1437, -0.1228, -0.1078, -0.1048, -0.1048, -0.0958, -0.0838, -0.0898, -0.0719, -0.0719, -0.0509, 0.0, 0.021, 0.003, 0.009, 0.006, 0.012, 0.024, 0.024, 0.012, 0.0419, 0.0539, 0.0629, 0.0419, 0.0539, 0.0599, 0.0509, 0.0659, 0.0898, 0.0808, 0.0629, 0.0569, 0.0778, 0.0689, 0.0569, 0.0659, 0.0868, 0.1257, 0.1467, 0.1467, 0.1467, 0.1467, 0.1467, 0.1467, 0.1467, 0.1467, 0.1377, 0.1766, 0.2066, 0.1946, 0.1946, 0.1826, 0.1587, 0.1168, 0.1377, 0.1317, 0.0838, 0.1078, 0.1617, 0.1287, 0.1168, 0.1048, 0.1234, 0.1444, 0.1114, 0.1054, 0.0874, 0.0874, 0.1084, 0.1054, 0.0934, 0.0693, 0.0573, 0.1144, 0.0874, 0.0874, 0.0874, 0.1174, 0.1715, 0.1745, 0.2015, 0.1955, 0.2346, 0.2496, 0.2526, 0.2195, 0.2165, 0.2315, 0.2315, 0.2496, 0.3127, 0.3607, 0.3307, 0.3096, 0.2826, 0.2826, 0.3667, 0.3517, 0.3517, 0.3877, 0.3757, 0.3427, 0.3547, 0.3337, 0.3637, 0.3607, 0.3457, 0.3247, 0.3127, 0.3397, 0.3547, 0.3877, 0.3637, 0.3817, 0.3787, 0.4148, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2009972495332569159", "ticker_idx": 0, "ts": "2026-01-10T12:56:01+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I still feel like $QCOM has no real future ahead", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish QCOM: losing Apple modem share to zero, misallocated capital into auto semis instead of ASIC, and Samsung Foundry yield improvement means Exynos chips outcompete Qualcomm in Galaxy lineup.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "After reading the KeyBanc meeting report with Qualcomm's management, I still feel like $QCOM has no real future ahead.\nFirst, let me share the key points from the report:\n\n1. Qualcomm acknowledged in discussions with smartphone customers that the rise in memory prices and its potential impact is currently the biggest concern. \nHowever, they say they have already reflected the increased foundry costs for the next-generation AP in their negotiations, so at this point they believe there will be no impact on gross margin (GM).\n\n2. Qualcomm is maintaining its existing forecast that its modem share in iPhone 18 will be 20%, and in iPhone 19 it will drop to 0%.\n\n3. They hope that the 2nm contract with Samsung Foundry will help them recover their share in Galaxy S27 after having only 75% share in Galaxy S26.\n\nWhat I find difficult to understand is this: Qualcomm made an enormous amount of cash by selling modems to Apple, yet instead of investing that money into something like the ASIC business, they poured it into the automotive semiconductor business. And now they're trying to increase sales to Samsung by ordering their own flagship chips from Samsung Foundry...\nBut if Samsung Foundry's yield improves, wouldn't it make far more sense for Samsung to just produce more of their own Exynos chips rather than manufacturing Qualcomm chips? For that reason, I personally find it extremely, extremely difficult to agree with the outlook that Qualcomm's chip share will increase in the S27.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05027466520783386, "ret_signed_m1d": -0.05027466520783386, "alpha_spy_m1d": -0.05184265656160736, "alpha_c_m1d": -0.05396074115087235, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023512793365576434, "ret_signed_p1d": 0.023512793365576434, "alpha_spy_p1d": 0.02151331825970615, "alpha_c_p1d": 0.025739781306578835, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": -0.05819110166377717, "ret_signed_p1w": 0.05819110166377717, "alpha_spy_p1w": 0.053156273503215945, "alpha_c_p1w": 0.08309769792425292, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": -0.1723874144818195, "ret_signed_p1m": 0.1723874144818195, "alpha_spy_p1m": 0.16801429451974004, "alpha_c_p1m": 0.20848019585523858, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": -0.2404430569806696, "ret_signed_p3m": 0.2404430569806696, "alpha_spy_p3m": 0.22117674839592127, "alpha_c_p3m": 0.34193794646205744, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0421, -0.0356, -0.0393, -0.0182, -0.0077, -0.0051, -0.0006, -0.007, 0.1031, 0.064, 0.0502, 0.0419, 0.0633, 0.0622, 0.0159, 0.0563, 0.018, 0.0044, 0.0084, 0.0226, 0.0384, 0.0257, 0.0226, -0.0199, -0.0298, -0.0237, -0.062, -0.0402, -0.0298, -0.0402, -0.0294, -0.0294, -0.012, -0.0123, 0.0033, 0.029, 0.03, 0.0327, 0.0357, 0.0398, 0.0764, 0.0709, 0.0533, 0.059, 0.0405, 0.0181, 0.0291, 0.0353, 0.0292, 0.0324, 0.0325, 0.0325, 0.0327, 0.0246, 0.0259, 0.0105, 0.0105, 0.0219, 0.0416, 0.0779, 0.0645, 0.0744, 0.0503, 0.0, -0.0235, -0.0279, -0.0466, -0.0582, -0.0582, -0.0898, -0.0762, -0.0678, -0.0795, -0.0871, -0.0959, -0.0979, -0.1007, -0.1044, -0.0984, -0.1305, -0.1204, -0.1948, -0.1886, -0.1792, -0.1724, -0.1668, -0.182, -0.1688, -0.1688, -0.1574, -0.1538, -0.1654, -0.1559, -0.1705, -0.1447, -0.1385, -0.1399, -0.159, -0.1668, -0.184, -0.1758, -0.1854, -0.1932, -0.1788, -0.1961, -0.2026, -0.2202, -0.2281, -0.2307, -0.2176, -0.2243, -0.2195, -0.2277, -0.2369, -0.235, -0.225, -0.2239, -0.2442, -0.2445, -0.2343, -0.2432, -0.2461, -0.2461, -0.2525, -0.2623, -0.2419, -0.2404, -0.2386, -0.2197, -0.2102, -0.2089, -0.2005, -0.1902, -0.1824, -0.194, -0.191, -0.2036, -0.115, -0.1066, -0.1082, -0.0725, 0.0677, 0.0524, 0.0011, 0.1092, 0.145, 0.2043, 0.3026, 0.4123, 0.2504, 0.2674, 0.1896, 0.198, 0.2108, 0.163, 0.2041, 0.2689, 0.416, 0.416, 0.4794, 0.3877, 0.4465, 0.4925, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2031576734961922475", "ticker_idx": 0, "ts": "2026-03-11T03:43:33+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raised its target price on TSMC to NT$2,288, citing four bullish catalysts: capacity expansion, faster-than-expected production ramp-up, rising capex, and strong revenue growth. KGI Securities also lifted its target to NT$2,420", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley and KGI buy-side upgrades on TSMC with raised PTs, projecting +25% revenue in 2027 and +20% in 2028 on four bullish catalysts.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Morgan Stanley Raises TSMC Target Price, Projects Strong Revenue Growth in 2027–2028 on Four Key Catalysts (Commercial Times)\n\n- Morgan Stanley raised its target price on TSMC to NT$2,288, citing four bullish catalysts: capacity expansion, faster-than-expected production ramp-up, rising capex, and strong revenue growth. KGI Securities also lifted its target to NT$2,420, reflecting the potential for expanded CoWoS advanced packaging capacity.\n\n- Morgan Stanley noted that TSMC is expanding capacity more aggressively than anticipated, with infrastructure buildout — including cleanroom construction — proceeding ahead of schedule. The company plans to build a total of 12 new front-end wafer fabs by end-2027, including 3nm facilities in Arizona and Kumamoto, Japan, along with four advanced packaging plants.\n\n- Wafer fab equipment (WFE) capex in 2027 is projected to grow +20–30% YoY, prompting upward revisions to TSMC's capex outlook — now estimated at $65B in 2027 and $70B in 2028.\n\n- Morgan Stanley noted that significant efficiency gains across process node transitions, cleanroom construction, and equipment installation have shortened the investment-to-revenue conversion cycle by approximately nine months. As a result, TSMC revenue is now projected to grow +25% YoY in 2027 and +20% YoY in 2028.\n\n$TSM\n\n---\n\nhttps://t.co/3HUQozvCZB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021068352579103555, "ret_signed_m1d": -0.021068352579103555, "alpha_spy_m1d": -0.022325092921509304, "alpha_c_m1d": -0.011842079686068652, "bench_spy_m1d": 0.0012567403424057488, "bench_c_m1d": -0.009226272893034904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05034406575900929, "ret_signed_p1d": -0.05034406575900929, "alpha_spy_p1d": -0.03515911688964579, "alpha_c_p1d": -0.018176911342255186, "bench_spy_p1d": -0.015184948869363502, "bench_c_p1d": -0.03216715441675411, "ret_p1w": -0.03957904207783891, "ret_signed_p1w": -0.03957904207783891, "alpha_spy_p1w": -0.017548351298673492, "alpha_c_p1w": -0.021226336858863548, "bench_spy_p1w": -0.02203069077916542, "bench_c_p1w": -0.018352705218975363, "ret_p1m": 0.03373159867062281, "ret_signed_p1m": 0.03373159867062281, "alpha_spy_p1m": 0.02569291264115403, "alpha_c_p1m": -0.039280392622853544, "bench_spy_p1m": 0.008038686029468778, "bench_c_p1m": 0.07301199129347635, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1763, -0.1885, -0.1909, -0.2189, -0.1971, -0.185, -0.1728, -0.1625, -0.1573, -0.1573, -0.1459, -0.1513, -0.1551, -0.1429, -0.1429, -0.0986, -0.0911, -0.0765, -0.1012, -0.1031, -0.0872, -0.0643, -0.0659, -0.0774, -0.0364, -0.0343, -0.0343, -0.0773, -0.0802, -0.0767, -0.0555, -0.0616, -0.0457, -0.0346, -0.0423, -0.0677, -0.0372, -0.0531, -0.0813, -0.0672, -0.0161, 0.0024, 0.0207, 0.0551, 0.0382, 0.0333, 0.0333, 0.0272, 0.0217, 0.0164, 0.0451, 0.0437, 0.088, 0.0936, 0.0628, 0.0565, 0.041, -0.004, 0.0081, -0.002, -0.0442, -0.0165, -0.0211, 0.0, -0.0503, -0.0458, -0.0404, -0.0214, -0.0396, -0.0418, -0.0688, -0.0427, -0.0292, -0.0164, -0.0776, -0.0759, -0.1048, -0.0442, -0.0341, -0.0411, -0.0411, -0.0334, -0.0233, 0.0349, 0.0337, 0.0482, 0.0453, 0.0745, 0.0609, 0.0277, 0.0479, 0.0359, 0.0411, 0.0958, 0.0823, 0.1383, 0.1454, 0.1097, 0.1139, 0.1202, 0.1247, 0.1359, 0.1155, 0.1865, 0.1714, 0.1644, 0.1442, 0.1236, 0.1308, 0.1815, 0.1436, 0.1199, 0.1104, 0.1359, 0.1516, 0.1441, 0.1441, 0.1662, 0.1956, 0.2017, 0.1835, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2026453490630078476", "ticker_idx": 0, "ts": "2026-02-25T00:25:36+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics • Target price: KRW 340,000 (+24% vs. prior)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Macquarie sharply raises PTs for Samsung (+24% to KRW 340k) and SK Hynix (+21% to KRW 1.7M) on surging earnings forecasts.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Macquarie sharply raises earnings forecasts for Samsung Electronics / SK hynix\n\n1. Samsung Electronics\n\n• Target price: KRW 340,000 (+24% vs. prior)\n\n• Operating profit (OP): FY26E KRW 301tn (US$208.8B) / FY27E KRW 477tn (US$330.9B)\n\n2. SK hynix\n\n• Target price: KRW 1,700,000 (+21% vs. prior)\n\n• Operating profit (OP): FY26E KRW 272tn (US$188.7B) / FY27E KRW 447tn (US$310.2B)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017198993190633782, "ret_signed_m1d": -0.017198993190633782, "alpha_spy_m1d": -0.008831254148954915, "alpha_c_m1d": 0.001610783551699857, "bench_spy_m1d": -0.008367739041678868, "bench_c_m1d": -0.01880977674233364, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07125313666173794, "ret_signed_p1d": 0.07125313666173794, "alpha_spy_p1d": 0.07680761305335904, "alpha_c_p1d": 0.08523814883608793, "bench_spy_p1d": -0.005554476391621099, "bench_c_p1d": -0.013985012174349998, "ret_p1w": -0.15380833475288103, "ret_signed_p1w": -0.15380833475288103, "alpha_spy_p1w": -0.1422378627972123, "alpha_c_p1w": -0.1316420653514796, "bench_spy_p1w": -0.011570471955668737, "bench_c_p1w": -0.02216626940140143, "ret_p1m": -0.11498766878575506, "ret_signed_p1m": -0.11498766878575506, "alpha_spy_p1m": -0.04819365096424566, "alpha_c_p1m": -0.04264835610451789, "bench_spy_p1m": -0.0667940178215094, "bench_c_p1m": -0.07233931268123717, "ret_p3m": 0.4403215240817604, "ret_signed_p3m": 0.4403215240817604, "alpha_spy_p3m": 0.361657007109651, "alpha_c_p3m": 0.17737276318189266, "bench_spy_p3m": 0.07866451697210941, "bench_c_p3m": 0.26294876089986774, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5085, -0.5071, -0.4943, -0.489, -0.486, -0.4699, -0.4645, -0.4699, -0.4719, -0.4753, -0.4675, -0.4875, -0.4973, -0.4723, -0.4738, -0.4802, -0.4596, -0.4547, -0.4567, -0.4567, -0.4278, -0.4128, -0.4108, -0.4108, -0.4108, -0.3686, -0.3214, -0.3174, -0.3071, -0.3179, -0.317, -0.3179, -0.3238, -0.3106, -0.2929, -0.2683, -0.2663, -0.2865, -0.2654, -0.2516, -0.2526, -0.2526, -0.2162, -0.202, -0.2103, -0.2113, -0.2609, -0.1769, -0.169, -0.2172, -0.2206, -0.1823, -0.1853, -0.1754, -0.1224, -0.1096, -0.1096, -0.1096, -0.1096, -0.0663, -0.0658, -0.0516, -0.0172, 0.0, 0.0713, 0.0639, 0.0639, -0.0413, -0.1538, -0.0585, -0.0752, -0.1474, -0.0767, -0.0663, -0.0767, -0.0983, -0.0727, -0.0472, 0.0246, -0.0147, -0.0201, -0.0845, -0.0678, -0.0713, -0.115, -0.115, -0.1319, -0.1767, -0.0661, -0.1215, -0.0831, -0.0491, -0.0324, 0.0365, 0.0045, 0.0144, -0.0102, 0.0168, 0.039, 0.071, 0.0636, 0.0562, 0.0784, 0.071, 0.1055, 0.0809, 0.1055, 0.0932, 0.1129, 0.0858, 0.0858, 0.1449, 0.1449, 0.3098, 0.3369, 0.3221, 0.4059, 0.3738, 0.3985, 0.4576, 0.332, 0.3837, 0.3566, 0.3591, 0.4748, 0.4403, 0.4403, 0.4723, 0.5117, 0.4748, 0.561, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2026453490630078476", "ticker_idx": 1, "ts": "2026-02-25T00:25:36+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix • Target price: KRW 1,700,000 (+21% vs. prior)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Macquarie sharply raises PTs for Samsung (+24% to KRW 340k) and SK Hynix (+21% to KRW 1.7M) on surging earnings forecasts.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Macquarie sharply raises earnings forecasts for Samsung Electronics / SK hynix\n\n1. Samsung Electronics\n\n• Target price: KRW 340,000 (+24% vs. prior)\n\n• Operating profit (OP): FY26E KRW 301tn (US$208.8B) / FY27E KRW 477tn (US$330.9B)\n\n2. SK hynix\n\n• Target price: KRW 1,700,000 (+21% vs. prior)\n\n• Operating profit (OP): FY26E KRW 272tn (US$188.7B) / FY27E KRW 447tn (US$310.2B)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012770126042722652, "ret_signed_m1d": -0.012770126042722652, "alpha_spy_m1d": -0.004402387001043784, "alpha_c_m1d": 0.0036556294008832646, "bench_spy_m1d": -0.008367739041678868, "bench_c_m1d": -0.016425755443605916, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08155982775321902, "ret_signed_p1d": 0.08155982775321902, "alpha_spy_p1d": 0.08711430414484012, "alpha_c_p1d": 0.11476330479210017, "bench_spy_p1d": -0.005554476391621099, "bench_c_p1d": -0.03320347703888116, "ret_p1w": -0.16447287718539738, "ret_signed_p1w": -0.16447287718539738, "alpha_spy_p1w": -0.15290240522972864, "alpha_c_p1w": -0.1009756054422497, "bench_spy_p1w": -0.011570471955668737, "bench_c_p1w": -0.06349727174314768, "ret_p1m": -0.0818059016140108, "ret_signed_p1m": -0.0818059016140108, "alpha_spy_p1m": -0.015011883792501401, "alpha_c_p1m": 0.024539149387424186, "bench_spy_p1m": -0.0667940178215094, "bench_c_p1m": -0.10634505100143499, "ret_p3m": 0.9101980513164889, "ret_signed_p3m": 0.9101980513164889, "alpha_spy_p3m": 0.8315335343443795, "alpha_c_p3m": 0.5578421230929498, "bench_spy_p3m": 0.07866451697210941, "bench_c_p3m": 0.3523559282235391, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4794, -0.4715, -0.4519, -0.4578, -0.4676, -0.4656, -0.4332, -0.444, -0.4234, -0.445, -0.4391, -0.4558, -0.4794, -0.4587, -0.4578, -0.4627, -0.4303, -0.4263, -0.4224, -0.4224, -0.4116, -0.3713, -0.3605, -0.3605, -0.3605, -0.335, -0.3163, -0.2868, -0.2711, -0.2574, -0.2692, -0.2642, -0.275, -0.2711, -0.2642, -0.2574, -0.2495, -0.2701, -0.2731, -0.2583, -0.2466, -0.277, -0.2141, -0.1739, -0.1542, -0.1071, -0.1847, -0.109, -0.1159, -0.1729, -0.1758, -0.1287, -0.1395, -0.1552, -0.1277, -0.1356, -0.1356, -0.1356, -0.1356, -0.1218, -0.0678, -0.0658, -0.0128, 0.0, 0.0816, 0.0442, 0.0442, -0.0759, -0.1645, -0.0739, -0.0907, -0.1773, -0.0769, -0.0602, -0.0848, -0.1044, -0.0415, -0.0454, 0.0392, -0.0031, -0.009, -0.0818, -0.0296, -0.0208, -0.0818, -0.0818, -0.1409, -0.2058, -0.1172, -0.1832, -0.1379, -0.1281, -0.0985, 0.0166, -0.0178, 0.0107, 0.0235, 0.0855, 0.118, 0.1367, 0.1101, 0.1475, 0.2046, 0.2036, 0.2056, 0.2026, 0.2715, 0.2794, 0.2725, 0.2656, 0.2656, 0.424, 0.424, 0.5756, 0.6278, 0.6592, 0.8502, 0.8059, 0.9446, 0.9387, 0.7901, 0.8108, 0.7173, 0.7173, 0.9092, 0.9102, 0.9102, 1.0194, 1.2074, 1.2531, 1.2964, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2028364899274084389", "ticker_idx": 0, "ts": "2026-03-02T07:00:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Macquarie projected in a recent report that SK Hynix's NAND OPM will reach 79% this year… SK Hynix's operating profit this year will surge 476.8% year-over-year", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Macquarie bullish thesis: NAND OPMs surging to 75-79% at Samsung/Hynix on AI inference demand; SNDK tagged as beneficiary.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung & SK NAND Operating Margins Forecast at 70%+ … AI-Driven Memory Rally Extends Beyond HBM Across the Board\n\nA surge in AI-driven memory demand is expected to push Samsung Electronics and SK Hynix's NAND operating profit margins (OPM) up to the 70% range this year. While high-bandwidth memory (HBM) has been the primary earnings driver so far, analysts say we have now entered a phase of a \"full-spectrum memory rally,\" with conventional DRAM and NAND both surging in tandem.\n\nAccording to industry sources on the 2nd, global investment bank Macquarie projected in a recent report that SK Hynix's NAND OPM will reach 79% this year. Considering that NAND OPM was around 12% last year, this represents a more than sixfold improvement in profitability within a single year.\n\nSamsung Electronics' NAND OPM, which had been hovering near cash-cost levels as recently as 2023, is now expected to reach approximately 75.9% this year.\n\n\"Cash-cost level\" refers to a stage where a company barely covers its minimum manufacturing cash costs, excluding depreciation and similar charges — effectively a breakeven zone generating almost no profit. The key driver behind NAND's emergence as a strategic product for memory makers, alongside HBM, is the shift in AI workloads from training to inference.\n\nWhile HBM played a critical role in the ultra-high-speed computation required for training, large-scale storage capable of storing and retrieving massive datasets is essential for running inference at scale for global users. NVIDIA's upcoming \"Rubin\" GPU, for instance, incorporates 16 terabytes (TB) of eSSD per unit. On a server rack basis — with 72 GPUs per rack — that translates to 1,152 TB of storage capacity per rack, structurally driving NAND demand higher.\n\nRather than expanding cleanroom capacity, manufacturers are focusing on process transitions such as higher layer-count stacking, leaving supply significantly short of demand. SK Hynix, which began mass production of its 321-layer (321L) QLC NAND in the second half of last year, is aggressively ramping up the product's share, with the new product expected to account for half of total NAND shipments (by capacity) by year-end.\n\nDRAM is also on a steep upward trajectory alongside NAND. SK Hynix's DRAM contract prices are estimated to have risen approximately 100% quarter-over-quarter in Q1 2025 alone. The company's full-year DRAM average selling price (ASP) increase is projected at 160%.\n\nAccordingly, Macquarie forecasts SK Hynix's operating profit this year will surge 476.8% year-over-year to KRW 272.269 trillion.\n\nSamsung Electronics is likewise expected to see a significant boost in profitability, centered on its memory business, mirroring SK Hynix's trajectory. Macquarie stated that \"Samsung Electronics' net profit will increase tenfold between 2025 and 2028.\"\n\n$SNDK\n\n---\n\nhttps://t.co/8XyLSomG8K", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0005682556842063757, "alpha_c_m1d": 4.926186798004384e-05, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -4.926186798004384e-05, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.11498588182920777, "ret_signed_p1d": -0.11498588182920777, "alpha_spy_p1d": -0.1061715433464121, "alpha_c_p1d": -0.07726345623490394, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.03772242559430383, "ret_p1w": -0.21206408092964646, "ret_signed_p1w": -0.21206408092964646, "alpha_spy_p1w": -0.2002485000061316, "alpha_c_p1w": -0.1824865344890717, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.029577546440574753, "ret_p1m": -0.23939681952013425, "ret_signed_p1m": -0.23939681952013425, "alpha_spy_p1m": -0.1894770779449505, "alpha_c_p1m": -0.18282549659567338, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.056571322924460876, "ret_p3m": 1.157759408647605, "ret_signed_p3m": 1.157759408647605, "alpha_spy_p3m": 1.055366029481697, "alpha_c_p3m": 0.6817634553298333, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.4759959533177718, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4807, -0.4901, -0.4882, -0.4572, -0.4675, -0.4478, -0.4685, -0.4628, -0.4788, -0.5014, -0.4816, -0.4807, -0.4854, -0.4544, -0.4506, -0.4468, -0.4468, -0.4365, -0.3979, -0.3876, -0.3876, -0.3876, -0.3631, -0.3452, -0.317, -0.3019, -0.2888, -0.3001, -0.2954, -0.3057, -0.3019, -0.2954, -0.2888, -0.2813, -0.301, -0.3038, -0.2897, -0.2784, -0.3076, -0.2474, -0.2088, -0.19, -0.1448, -0.2192, -0.1467, -0.1533, -0.2079, -0.2107, -0.1655, -0.1759, -0.1909, -0.1646, -0.1721, -0.1721, -0.1721, -0.1721, -0.159, -0.1072, -0.1053, -0.0545, -0.0423, 0.0358, 0.0, 0.0, -0.115, -0.1998, -0.1131, -0.1291, -0.2121, -0.1159, -0.0999, -0.1235, -0.1423, -0.082, -0.0858, -0.0047, -0.0452, -0.0509, -0.1206, -0.0707, -0.0622, -0.1206, -0.1206, -0.1772, -0.2394, -0.1546, -0.2177, -0.1744, -0.1649, -0.1367, -0.0264, -0.0594, -0.032, -0.0198, 0.0396, 0.0707, 0.0886, 0.0631, 0.099, 0.1536, 0.1527, 0.1546, 0.1517, 0.2177, 0.2253, 0.2187, 0.2121, 0.2121, 0.3638, 0.3638, 0.509, 0.5589, 0.5891, 0.7719, 0.7295, 0.8624, 0.8567, 0.7144, 0.7342, 0.6447, 0.6447, 0.8285, 0.8294, 0.8294, 0.934, 1.114, 1.1578, 1.1992, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2028364899274084389", "ticker_idx": 1, "ts": "2026-03-02T07:00:51+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung Electronics' net profit will increase tenfold between 2025 and 2028", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Macquarie bullish thesis: NAND OPMs surging to 75-79% at Samsung/Hynix on AI inference demand; SNDK tagged as beneficiary.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung & SK NAND Operating Margins Forecast at 70%+ … AI-Driven Memory Rally Extends Beyond HBM Across the Board\n\nA surge in AI-driven memory demand is expected to push Samsung Electronics and SK Hynix's NAND operating profit margins (OPM) up to the 70% range this year. While high-bandwidth memory (HBM) has been the primary earnings driver so far, analysts say we have now entered a phase of a \"full-spectrum memory rally,\" with conventional DRAM and NAND both surging in tandem.\n\nAccording to industry sources on the 2nd, global investment bank Macquarie projected in a recent report that SK Hynix's NAND OPM will reach 79% this year. Considering that NAND OPM was around 12% last year, this represents a more than sixfold improvement in profitability within a single year.\n\nSamsung Electronics' NAND OPM, which had been hovering near cash-cost levels as recently as 2023, is now expected to reach approximately 75.9% this year.\n\n\"Cash-cost level\" refers to a stage where a company barely covers its minimum manufacturing cash costs, excluding depreciation and similar charges — effectively a breakeven zone generating almost no profit. The key driver behind NAND's emergence as a strategic product for memory makers, alongside HBM, is the shift in AI workloads from training to inference.\n\nWhile HBM played a critical role in the ultra-high-speed computation required for training, large-scale storage capable of storing and retrieving massive datasets is essential for running inference at scale for global users. NVIDIA's upcoming \"Rubin\" GPU, for instance, incorporates 16 terabytes (TB) of eSSD per unit. On a server rack basis — with 72 GPUs per rack — that translates to 1,152 TB of storage capacity per rack, structurally driving NAND demand higher.\n\nRather than expanding cleanroom capacity, manufacturers are focusing on process transitions such as higher layer-count stacking, leaving supply significantly short of demand. SK Hynix, which began mass production of its 321-layer (321L) QLC NAND in the second half of last year, is aggressively ramping up the product's share, with the new product expected to account for half of total NAND shipments (by capacity) by year-end.\n\nDRAM is also on a steep upward trajectory alongside NAND. SK Hynix's DRAM contract prices are estimated to have risen approximately 100% quarter-over-quarter in Q1 2025 alone. The company's full-year DRAM average selling price (ASP) increase is projected at 160%.\n\nAccordingly, Macquarie forecasts SK Hynix's operating profit this year will surge 476.8% year-over-year to KRW 272.269 trillion.\n\nSamsung Electronics is likewise expected to see a significant boost in profitability, centered on its memory business, mirroring SK Hynix's trajectory. Macquarie stated that \"Samsung Electronics' net profit will increase tenfold between 2025 and 2028.\"\n\n$SNDK\n\n---\n\nhttps://t.co/8XyLSomG8K", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0005682556842063757, "alpha_c_m1d": 0.0055898388543234034, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -0.0055898388543234034, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09884526208146449, "ret_signed_p1d": -0.09884526208146449, "alpha_spy_p1d": -0.09003092359866882, "alpha_c_p1d": -0.084225843558279, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.01461941852318549, "ret_p1w": -0.1986143434258696, "ret_signed_p1w": -0.1986143434258696, "alpha_spy_p1w": -0.18679876250235472, "alpha_c_p1w": -0.2001909506044468, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": 0.0015766071785772162, "ret_p1m": -0.2261151425649689, "ret_signed_p1m": -0.2261151425649689, "alpha_spy_p1m": -0.17619540098978514, "alpha_c_p1m": -0.17971438894673752, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.04640075361823137, "ret_p3m": 0.3862351363743528, "ret_signed_p3m": 0.3862351363743528, "alpha_spy_p3m": 0.28384175720844484, "alpha_c_p3m": 0.04552758791859035, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.34070754845576245, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5197, -0.5169, -0.5017, -0.4967, -0.5017, -0.5036, -0.5068, -0.4994, -0.5183, -0.5275, -0.504, -0.5054, -0.5114, -0.4921, -0.4875, -0.4893, -0.4893, -0.4622, -0.448, -0.4462, -0.4462, -0.4462, -0.4065, -0.3621, -0.3584, -0.3487, -0.3589, -0.358, -0.3589, -0.3644, -0.352, -0.3353, -0.3122, -0.3104, -0.3293, -0.3095, -0.2965, -0.2975, -0.2975, -0.2633, -0.2499, -0.2577, -0.2587, -0.3053, -0.2263, -0.2189, -0.2642, -0.2674, -0.2314, -0.2342, -0.2249, -0.1751, -0.163, -0.163, -0.163, -0.163, -0.1224, -0.1219, -0.1085, -0.0762, -0.06, 0.0069, 0.0, 0.0, -0.0988, -0.2046, -0.115, -0.1307, -0.1986, -0.1321, -0.1224, -0.1321, -0.1524, -0.1284, -0.1044, -0.037, -0.0739, -0.079, -0.1395, -0.1238, -0.127, -0.1681, -0.1681, -0.184, -0.2261, -0.1222, -0.1743, -0.1382, -0.1062, -0.0905, -0.0257, -0.0558, -0.0465, -0.0697, -0.0442, -0.0234, 0.0067, -0.0002, -0.0072, 0.0136, 0.0067, 0.0391, 0.016, 0.0391, 0.0275, 0.046, 0.0206, 0.0206, 0.0761, 0.0761, 0.2312, 0.2566, 0.2428, 0.3214, 0.2914, 0.3145, 0.37, 0.252, 0.3006, 0.2752, 0.2775, 0.3862, 0.3538, 0.3538, 0.3839, 0.4209, 0.3862, 0.4672, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2028364899274084389", "ticker_idx": 2, "ts": "2026-03-02T07:00:51+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$SNDK", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Macquarie bullish thesis: NAND OPMs surging to 75-79% at Samsung/Hynix on AI inference demand; SNDK tagged as beneficiary.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung & SK NAND Operating Margins Forecast at 70%+ … AI-Driven Memory Rally Extends Beyond HBM Across the Board\n\nA surge in AI-driven memory demand is expected to push Samsung Electronics and SK Hynix's NAND operating profit margins (OPM) up to the 70% range this year. While high-bandwidth memory (HBM) has been the primary earnings driver so far, analysts say we have now entered a phase of a \"full-spectrum memory rally,\" with conventional DRAM and NAND both surging in tandem.\n\nAccording to industry sources on the 2nd, global investment bank Macquarie projected in a recent report that SK Hynix's NAND OPM will reach 79% this year. Considering that NAND OPM was around 12% last year, this represents a more than sixfold improvement in profitability within a single year.\n\nSamsung Electronics' NAND OPM, which had been hovering near cash-cost levels as recently as 2023, is now expected to reach approximately 75.9% this year.\n\n\"Cash-cost level\" refers to a stage where a company barely covers its minimum manufacturing cash costs, excluding depreciation and similar charges — effectively a breakeven zone generating almost no profit. The key driver behind NAND's emergence as a strategic product for memory makers, alongside HBM, is the shift in AI workloads from training to inference.\n\nWhile HBM played a critical role in the ultra-high-speed computation required for training, large-scale storage capable of storing and retrieving massive datasets is essential for running inference at scale for global users. NVIDIA's upcoming \"Rubin\" GPU, for instance, incorporates 16 terabytes (TB) of eSSD per unit. On a server rack basis — with 72 GPUs per rack — that translates to 1,152 TB of storage capacity per rack, structurally driving NAND demand higher.\n\nRather than expanding cleanroom capacity, manufacturers are focusing on process transitions such as higher layer-count stacking, leaving supply significantly short of demand. SK Hynix, which began mass production of its 321-layer (321L) QLC NAND in the second half of last year, is aggressively ramping up the product's share, with the new product expected to account for half of total NAND shipments (by capacity) by year-end.\n\nDRAM is also on a steep upward trajectory alongside NAND. SK Hynix's DRAM contract prices are estimated to have risen approximately 100% quarter-over-quarter in Q1 2025 alone. The company's full-year DRAM average selling price (ASP) increase is projected at 160%.\n\nAccordingly, Macquarie forecasts SK Hynix's operating profit this year will surge 476.8% year-over-year to KRW 272.269 trillion.\n\nSamsung Electronics is likewise expected to see a significant boost in profitability, centered on its memory business, mirroring SK Hynix's trajectory. Macquarie stated that \"Samsung Electronics' net profit will increase tenfold between 2025 and 2028.\"\n\n$SNDK\n\n---\n\nhttps://t.co/8XyLSomG8K", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.026297034005793307, "ret_signed_m1d": 0.026297034005793307, "alpha_spy_m1d": 0.026865289689999683, "alpha_c_m1d": 0.03188687286011671, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -0.0055898388543234034, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08669322618036246, "ret_signed_p1d": -0.08669322618036246, "alpha_spy_p1d": -0.07787888769756679, "alpha_c_p1d": -0.07207380765717697, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.01461941852318549, "ret_p1w": -0.04902441652657841, "ret_signed_p1w": -0.04902441652657841, "alpha_spy_p1w": -0.03720883560306354, "alpha_c_p1w": -0.05060102370515562, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": 0.0015766071785772162, "ret_p1m": 0.02626479504549284, "ret_signed_p1m": 0.02626479504549284, "alpha_spy_p1m": 0.07618453662067659, "alpha_c_p1m": 0.07266554866372421, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.04640075361823137, "ret_p3m": 1.651741243992043, "ret_signed_p3m": 1.651741243992043, "alpha_spy_p3m": 1.549347864826135, "alpha_c_p3m": 1.3110336955362805, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.34070754845576245, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.686, -0.6554, -0.631, -0.6358, -0.6455, -0.6239, -0.6097, -0.667, -0.6739, -0.6619, -0.6659, -0.6455, -0.6162, -0.6106, -0.6044, -0.596, -0.596, -0.5961, -0.6055, -0.612, -0.6166, -0.6166, -0.5554, -0.5573, -0.4352, -0.4289, -0.4596, -0.3904, -0.3712, -0.3703, -0.3736, -0.339, -0.3319, -0.3319, -0.2681, -0.1903, -0.1868, -0.2346, -0.2395, -0.2223, -0.1477, -0.1289, -0.0692, 0.0746, 0.1235, -0.0558, -0.0693, -0.0341, -0.0576, -0.1251, -0.0319, 0.0181, 0.0121, 0.0121, -0.046, -0.0302, 0.0032, 0.0499, 0.0766, 0.0314, 0.0215, 0.053, 0.0263, 0.0, -0.0867, -0.0323, -0.0864, -0.1482, -0.049, -0.0003, 0.0587, -0.0004, 0.0687, 0.1366, 0.1633, 0.2174, 0.2472, 0.1464, 0.1347, 0.1347, 0.0949, -0.0257, -0.0052, -0.0752, 0.0263, 0.119, 0.1333, 0.1333, 0.1705, 0.1482, 0.2614, 0.3755, 0.3759, 0.5386, 0.5256, 0.4404, 0.4852, 0.4877, 0.4748, 0.4594, 0.5815, 0.5062, 0.599, 0.7287, 0.6191, 0.719, 0.7712, 0.9174, 1.0286, 1.2716, 1.2775, 1.1644, 1.5236, 1.4998, 1.3454, 1.3377, 1.2335, 1.2737, 1.1532, 1.2344, 1.2494, 1.4912, 1.3885, 1.3885, 1.5676, 1.5682, 1.6517, 1.7379, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2032425174545637463", "ticker_idx": 0, "ts": "2026-03-13T11:54:56+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "When 3D NAND was introduced, Lam Research's revenue doubled. We expect a similar revenue opportunity with DRAM. SAM to expand by 1.7x to 1.8x, new dry photoresist equipment to generate cumulative new revenue of $1.5 billion over the next five years", "tweet_type": "analysis", "timeline": "years", "summary": "Long LRCX: 3D DRAM transition to expand SAM 1.7-1.8x, new etch/TSV/dry-PR equipment poised to replicate NAND revenue-doubling opportunity.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Lam Research: \"DRAM Architecture Evolving from 4F² to 3D\"\n\nDRAM memory is expected to evolve from its current planar structure to 3D DRAM, driven by the need to overcome the limits of miniaturization.\n\nDoug Bettinger, CFO of Lam Research, said at a recent Morgan Stanley Media Conference: \"As we've emphasized for some time, semiconductor architectures are largely evolving in the 3D direction. DRAM will transition from 6F² to 4F², and ultimately evolve into 3D DRAM.\"\n\nF refers to the minimum feature size achievable in a given semiconductor manufacturing process, and F² (F-square) denotes the cell area defined by that feature size. In the 6F² structure, a single cell is rectangular at 3F × 2F, whereas in the 4F² structure, the cell is square at 2F × 2F.\n\nThe transition from 6F² to 4F² reduces cell area, which is known to improve wafer-level chip productivity by approximately 20%. For this reason, major DRAM manufacturers such as Samsung Electronics and SK Hynix are developing 4F² DRAM with a vertical transistor structure as their next-generation product. The planned approach involves bonding a cell array wafer — which stores data — with a peripheral circuit (peri) wafer via hybrid bonding.\n\nThe 3D DRAM that Bettinger referenced represents a further step beyond this. It involves vertically stacking multiple cell layers of the 4F² DRAM structure, analogous to 3D NAND, which stacks flash memory cells. This enables higher density per unit area. The physical limits of scaling in conventional planar (2D) semiconductor structures are well established, and following NAND, DRAM has now also committed to a 3D architecture for vertical integration.\n\nLam Research expects the 4F² structure to be commercially adopted one to two process nodes from now. Over this transition, the company projects its serviceable addressable market (SAM) to expand by 1.7x to 1.8x, as 3D stacking requires high-aspect-ratio etch equipment to drill through-silicon vias (TSVs) and electroplating equipment to fill TSVs with copper.\n\nBettinger noted: \"When 3D NAND was introduced, Lam Research's revenue doubled. We expect a similar revenue opportunity with DRAM.\" The company has launched a new etch system, Akara, which is specifically designed for 4F² and 3D DRAM processes. It also offers Syndion, an etch system for TSV processing, and Sabre 3D, an electroplating system.\n\nLam Research forecasts the front-end equipment market to grow 23% this year. To meet customer demand, the company is expanding production capacity at facilities including its Malaysia plant. That said, the current industry constraint is not equipment production capacity per se, but rather a shortage of cleanroom space. This bottleneck is carrying over equipment demand into next year, with high performance expected to continue.\n\nLam Research has recently been pursuing its \"Beyond NAND\" strategy, which centers on diversifying its product portfolio beyond memory equipment into foundry, logic, and advanced packaging equipment. Last year, foundry and logic revenue accounted for approximately 59% of total revenue — a reversal from 2022, when memory represented 60% of total equipment sales. The Akara, Syndion, and Sabre 3D systems for 3D DRAM manufacturing are all part of this diversification strategy.\n\nThe company has also entered the track equipment market for lithography processes for the first time. Bettinger stated that \"new dry photoresist (PR) technology equipment has been deployed and is now running in the actual production lines of our largest DRAM customer.\" Lam Research uses its deposition technology to form the photoresist on wafers, offering higher productivity compared to conventional wet chemical methods. The company expects the new equipment to generate cumulative new revenue of $1.5 billion over the next five years.\n\n$LRCX\n\n---\n\nhttps://t.co/66oqLiDE48", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012770930697699856, "ret_signed_m1d": -0.012770930697699856, "alpha_spy_m1d": -0.01846325940249105, "alpha_c_m1d": -0.014836400292829288, "bench_spy_m1d": 0.005692328704791194, "bench_c_m1d": 0.002065469595129432, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03393024058336036, "ret_signed_p1d": 0.03393024058336036, "alpha_spy_p1d": 0.023753374471877198, "alpha_c_p1d": 0.016916254876268244, "bench_spy_p1d": 0.010176866111483163, "bench_c_p1d": 0.017013985707092116, "ret_p1w": 0.07615458951232212, "ret_signed_p1w": 0.07615458951232212, "alpha_spy_p1w": 0.09419614529536913, "alpha_c_p1w": 0.0828413848229852, "bench_spy_p1w": -0.018041555783047003, "bench_c_p1w": -0.006686795310663074, "ret_p1m": 0.2597550007949383, "ret_signed_p1m": 0.2597550007949383, "alpha_spy_p1m": 0.22097482749767905, "alpha_c_p1m": 0.11514959530776214, "bench_spy_p1m": 0.038780173297259246, "bench_c_p1m": 0.14460540548717615, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2316, -0.2705, -0.2248, -0.1891, -0.1751, -0.1755, -0.1653, -0.1653, -0.1618, -0.1722, -0.182, -0.1943, -0.1943, -0.1289, -0.0833, -0.0259, -0.0441, -0.0541, 0.0278, 0.0374, 0.0091, -0.0172, 0.0236, 0.0494, 0.0494, 0.0469, 0.075, 0.0388, 0.0258, 0.049, 0.1224, 0.1277, 0.1681, 0.0989, 0.1179, 0.0831, -0.0126, 0.004, 0.0873, 0.0792, 0.0666, 0.1067, 0.0887, 0.1086, 0.1086, 0.1089, 0.1301, 0.1174, 0.1528, 0.1403, 0.1497, 0.1743, 0.1253, 0.1009, 0.0873, 0.0227, 0.0508, 0.0117, -0.0607, -0.0049, 0.0143, 0.0314, -0.0128, 0.0, 0.0339, 0.0672, 0.059, 0.1027, 0.0762, 0.0995, 0.1255, 0.1001, -0.0027, -0.0037, -0.0578, 0.0069, 0.0462, 0.0294, 0.0294, 0.0398, 0.0573, 0.1616, 0.2194, 0.2425, 0.2598, 0.2837, 0.2496, 0.2298, 0.2611, 0.2402, 0.2176, 0.2514, 0.2185, 0.2619, 0.2228, 0.1839, 0.1722, 0.2152, 0.2098, 0.2185, 0.2997, 0.4004, 0.3502, 0.3857, 0.3951, 0.3631, 0.3923, 0.4098, 0.3418, 0.3099, 0.2883, 0.3765, 0.4243, 0.439, 0.439, 0.5206, 0.503, 0.4986, 0.4994, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021129556036645202", "ticker_idx": 0, "ts": "2026-02-10T07:50:11+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA (NVDA, Buy)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF International: Buy NVDA on CPO roadmap acceleration; bullish Lumentum on CW Laser demand; neutral Coherent; expect Browave (3163 TT) to benefit from CPO Shuttle Box ramp.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF International Electronics & Communications》\n\nAI Networking: NVIDIA's Accelerated Roadmap, CPO Technology Debate Heats Up\n\nKey CPO Takeaways from Recent Earnings Calls: Coherent mentioned it has secured a significant order from a leading AI data center customer to supply CW Lasers for its CPO systems. Initial revenue is expected to begin in 4Q26, with more meaningful contributions starting in 2027. Lumentum stated it has received hundreds of millions of dollars in CW Laser commitments for Scale-out CPO, with volume production contributions beginning in 1H27. Assuming combined order value of $400 million for laser suppliers such as LITE and COHR in 2027, this implies approximately 80,000 Scale-out CPO switches will be shipped that year. We believe this CPO switch shipment volume is consistent with downstream suppliers' capacity buildout plans.\n\nNVIDIA May Accelerate Scale-out CPO Switches: At GTC 2025, NVIDIA (NVDA, Buy) unveiled its CPO switches, with Quantum-X planned for launch in 2H25 and Spectrum-X in 2H26. NVIDIA's first-generation Scale-out CPO (Quantum-X) adopts a pluggable architecture rather than true CPO. The system supports 115.2T switching bandwidth and integrates 36 pluggable optical engines, making it more accurately classified as NPO. However, due to its lack of cost-performance advantage over traditional pluggable solutions, customer adoption has been limited, with minimal shipments in 2025. For GTC 2026, we expect NVIDIA may unveil a next-generation CPO switch. This generation is expected to feature 115.2T bandwidth, with the CPO portion manufactured by TSMC. We believe this CPO design offers significant improvements in thermal performance and bandwidth, with cost-performance far exceeding the previous generation. The supply chain is expected to begin ramping in 2Q26 and accelerate through 2H26/2027. Driven by NVIDIA's aggressive push and bundling sales strategy, we are raising our NVIDIA Scale-out CPO switch shipment estimates for 2025/2026/2027 to 2,000/20,000/80,000 units. As for 2027, despite the surge in shipments, our estimate represents only a single-digit percentage of total Scale-out switch shipments for the year.\n\nNVIDIA Scale-up CPO Progress Update: According to recent earnings calls, Lumentum emphasized that as copper interconnects approach their limits, optical Scale-up represents a long-term structural opportunity beginning in late 2027. As we noted in our report, we expect NVIDIA to introduce CPO/NPO within the NVL576 architecture starting with Rubin Ultra in 2H27, specifically for Scale-up interconnects. Within the NVL576 architecture, Compute trays and Switch trays are expected to continue relying on backplane connections, while inter-rack interconnects may transition to CPO- or NPO-based optical interconnects. Technically, CPO holds clear advantages in power consumption and bandwidth density, while NPO is easier to manufacture and maintain. We believe TSMC's Scale-up CPO solution has already provided samples, but mass production readiness remains uncertain, indicating that the NPO vs. CPO debate will persist in the near term.\n\nKey Beneficiary Stocks: Overall, Scale-up CPO represents a purely incremental opportunity for the optical interconnect supply chain, as it aims to replace copper interconnects without affecting Scale-out. Furthermore, Scale-up and Scale-out CPO solutions share the same suppliers. Accordingly, we expect the primary beneficiaries to be component suppliers of FAUs, CW Lasers, ELS FP modules, and Shuffles. Our views: 1) Bullish on Lumentum, given upside potential in CW Laser demand; 2) Neutral impact on Coherent; 3) We expect Browave (3163 TT) to benefit, given its approximately 50% market share in CPO Shuttle Boxes, unit price of $5,000–6,000, and volume production ramp beginning in 2Q26.\n\n$COHR $LITE $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00795589781641981, "ret_signed_m1d": 0.00795589781641981, "alpha_spy_m1d": 0.005311764380264039, "alpha_c_m1d": 0.003286503396862983, "bench_spy_m1d": 0.002644133436155771, "bench_c_m1d": 0.004669394419556827, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008008991631896611, "ret_signed_p1d": 0.008008991631896611, "alpha_spy_p1d": 0.00824005044631293, "alpha_c_p1d": -0.016746391241707625, "bench_spy_p1d": -0.00023105881441631837, "bench_c_p1d": 0.024755382873604237, "ret_p1w": -0.01893490568725409, "ret_signed_p1w": -0.01893490568725409, "alpha_spy_p1w": -0.00554127599556864, "alpha_c_p1w": -0.025729055142859503, "bench_spy_p1w": -0.01339362969168545, "bench_c_p1w": 0.006794149455605414, "ret_p1m": -0.013259371058431668, "ret_signed_p1m": -0.013259371058431668, "alpha_spy_p1m": 0.009554525612738618, "alpha_c_p1m": -0.004044006654008503, "bench_spy_p1m": -0.022813896671170286, "bench_c_p1m": -0.009215364404423165, "ret_p3m": 0.14146418174660513, "ret_signed_p3m": 0.14146418174660513, "alpha_spy_p3m": 0.07281353957907011, "alpha_c_p3m": -0.2582293758145702, "bench_spy_p3m": 0.06865064216753503, "bench_c_p3m": 0.3996935575611753, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.009, 0.0086, -0.0103, -0.0381, -0.0108, -0.042, -0.0513, -0.0318, -0.0569, -0.044, -0.044, -0.0613, -0.0458, -0.0376, -0.0475, -0.0274, -0.0325, -0.0159, -0.0189, -0.0252, -0.0404, -0.0717, -0.065, -0.0574, -0.0933, -0.0764, -0.04, -0.0257, 0.0036, 0.0004, 0.0004, 0.0106, -0.0017, -0.0053, -0.0108, -0.0108, 0.0016, -0.0022, -0.0069, 0.003, -0.0186, -0.0195, -0.0191, -0.0145, -0.0286, -0.0079, -0.0123, -0.0123, -0.0555, -0.0277, -0.0196, -0.0046, -0.011, -0.0001, 0.0158, 0.0211, 0.0137, -0.0155, -0.0435, -0.0761, -0.0884, -0.0166, 0.008, 0.0, 0.008, -0.0085, -0.0304, -0.0304, -0.0189, -0.003, -0.0034, 0.0068, 0.016, 0.0229, 0.0372, -0.0194, -0.0602, -0.0321, -0.045, -0.0292, -0.0276, -0.0569, -0.0312, -0.02, -0.0133, -0.0286, -0.0439, -0.0282, -0.035, -0.0431, -0.0529, -0.084, -0.0684, -0.0707, -0.0522, -0.0917, -0.1114, -0.1239, -0.0749, -0.0678, -0.0591, -0.0591, -0.0578, -0.0553, -0.0342, -0.0245, 0.0005, 0.0041, 0.0423, 0.0548, 0.0521, 0.0698, 0.0718, 0.0602, 0.0741, 0.0589, 0.1047, 0.1489, 0.1307, 0.1099, 0.0586, 0.0526, 0.0528, 0.0423, 0.1024, 0.1218, 0.1415, 0.164, 0.1711, 0.1978, 0.2504, 0.1951, 0.1792, 0.1702, 0.1853, 0.1643, 0.1422, 0.1422, 0.1397, 0.1277, 0.1364, 0.1199, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021129556036645202", "ticker_idx": 1, "ts": "2026-02-10T07:50:11+00:00", "symbol": "LITE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bullish on Lumentum, given upside potential in CW Laser demand", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF International: Buy NVDA on CPO roadmap acceleration; bullish Lumentum on CW Laser demand; neutral Coherent; expect Browave (3163 TT) to benefit from CPO Shuttle Box ramp.", "resolved_tickers": ["LITE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "《GF International Electronics & Communications》\n\nAI Networking: NVIDIA's Accelerated Roadmap, CPO Technology Debate Heats Up\n\nKey CPO Takeaways from Recent Earnings Calls: Coherent mentioned it has secured a significant order from a leading AI data center customer to supply CW Lasers for its CPO systems. Initial revenue is expected to begin in 4Q26, with more meaningful contributions starting in 2027. Lumentum stated it has received hundreds of millions of dollars in CW Laser commitments for Scale-out CPO, with volume production contributions beginning in 1H27. Assuming combined order value of $400 million for laser suppliers such as LITE and COHR in 2027, this implies approximately 80,000 Scale-out CPO switches will be shipped that year. We believe this CPO switch shipment volume is consistent with downstream suppliers' capacity buildout plans.\n\nNVIDIA May Accelerate Scale-out CPO Switches: At GTC 2025, NVIDIA (NVDA, Buy) unveiled its CPO switches, with Quantum-X planned for launch in 2H25 and Spectrum-X in 2H26. NVIDIA's first-generation Scale-out CPO (Quantum-X) adopts a pluggable architecture rather than true CPO. The system supports 115.2T switching bandwidth and integrates 36 pluggable optical engines, making it more accurately classified as NPO. However, due to its lack of cost-performance advantage over traditional pluggable solutions, customer adoption has been limited, with minimal shipments in 2025. For GTC 2026, we expect NVIDIA may unveil a next-generation CPO switch. This generation is expected to feature 115.2T bandwidth, with the CPO portion manufactured by TSMC. We believe this CPO design offers significant improvements in thermal performance and bandwidth, with cost-performance far exceeding the previous generation. The supply chain is expected to begin ramping in 2Q26 and accelerate through 2H26/2027. Driven by NVIDIA's aggressive push and bundling sales strategy, we are raising our NVIDIA Scale-out CPO switch shipment estimates for 2025/2026/2027 to 2,000/20,000/80,000 units. As for 2027, despite the surge in shipments, our estimate represents only a single-digit percentage of total Scale-out switch shipments for the year.\n\nNVIDIA Scale-up CPO Progress Update: According to recent earnings calls, Lumentum emphasized that as copper interconnects approach their limits, optical Scale-up represents a long-term structural opportunity beginning in late 2027. As we noted in our report, we expect NVIDIA to introduce CPO/NPO within the NVL576 architecture starting with Rubin Ultra in 2H27, specifically for Scale-up interconnects. Within the NVL576 architecture, Compute trays and Switch trays are expected to continue relying on backplane connections, while inter-rack interconnects may transition to CPO- or NPO-based optical interconnects. Technically, CPO holds clear advantages in power consumption and bandwidth density, while NPO is easier to manufacture and maintain. We believe TSMC's Scale-up CPO solution has already provided samples, but mass production readiness remains uncertain, indicating that the NPO vs. CPO debate will persist in the near term.\n\nKey Beneficiary Stocks: Overall, Scale-up CPO represents a purely incremental opportunity for the optical interconnect supply chain, as it aims to replace copper interconnects without affecting Scale-out. Furthermore, Scale-up and Scale-out CPO solutions share the same suppliers. Accordingly, we expect the primary beneficiaries to be component suppliers of FAUs, CW Lasers, ELS FP modules, and Shuffles. Our views: 1) Bullish on Lumentum, given upside potential in CW Laser demand; 2) Neutral impact on Coherent; 3) We expect Browave (3163 TT) to benefit, given its approximately 50% market share in CPO Shuttle Boxes, unit price of $5,000–6,000, and volume production ramp beginning in 2Q26.\n\n$COHR $LITE $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.028549568534649294, "ret_signed_m1d": 0.028549568534649294, "alpha_spy_m1d": 0.025905435098493523, "alpha_c_m1d": 0.025905435098493523, "bench_spy_m1d": 0.002644133436155771, "bench_c_m1d": 0.002644133436155771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023131859561601464, "ret_signed_p1d": 0.023131859561601464, "alpha_spy_p1d": 0.023362918376017783, "alpha_c_p1d": 0.023362918376017783, "bench_spy_p1d": -0.00023105881441631837, "bench_c_p1d": -0.00023105881441631837, "ret_p1w": 0.0700193853214981, "ret_signed_p1w": 0.0700193853214981, "alpha_spy_p1w": 0.08341301501318354, "alpha_c_p1w": 0.08341301501318354, "bench_spy_p1w": -0.01339362969168545, "bench_c_p1w": -0.01339362969168545, "ret_p1m": 0.19758343726484884, "ret_signed_p1m": 0.19758343726484884, "alpha_spy_p1m": 0.22039733393601912, "alpha_c_p1m": 0.22039733393601912, "bench_spy_p1m": -0.022813896671170286, "bench_c_p1m": -0.022813896671170286, "ret_p3m": 0.6106784166384405, "ret_signed_p3m": 0.6106784166384405, "alpha_spy_p3m": 0.5420277744709054, "alpha_c_p3m": 0.5420277744709054, "bench_spy_p3m": 0.06865064216753503, "bench_c_p3m": 0.06865064216753503, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5957, -0.5863, -0.5686, -0.5591, -0.5208, -0.5843, -0.5445, -0.4665, -0.4809, -0.4506, -0.4506, -0.4205, -0.4334, -0.4604, -0.4601, -0.4157, -0.4094, -0.3895, -0.3578, -0.3477, -0.3369, -0.422, -0.4035, -0.4366, -0.4293, -0.3992, -0.3381, -0.3052, -0.3096, -0.2944, -0.2944, -0.3036, -0.336, -0.3385, -0.3431, -0.3431, -0.3119, -0.3637, -0.2918, -0.2998, -0.3794, -0.3737, -0.3943, -0.3561, -0.409, -0.3883, -0.4221, -0.4221, -0.3641, -0.3541, -0.3683, -0.3955, -0.4075, -0.3394, -0.3139, -0.3202, -0.3017, -0.2454, -0.2246, -0.1704, -0.1011, -0.0163, 0.0285, 0.0, 0.0231, 0.0398, 0.0029, 0.0029, 0.07, 0.059, 0.1328, 0.19, 0.2024, 0.2266, 0.2892, 0.2065, 0.2491, 0.3958, 0.2376, 0.2133, 0.1598, -0.0048, 0.1418, 0.1976, 0.1976, 0.0979, 0.1094, 0.1135, 0.1576, 0.2489, 0.376, 0.2588, 0.2991, 0.4292, 0.385, 0.2275, 0.2523, 0.1669, 0.2524, 0.3627, 0.4736, 0.4736, 0.3763, 0.4538, 0.5968, 0.5934, 0.5991, 0.5525, 0.5198, 0.4685, 0.5883, 0.5933, 0.5952, 0.4915, 0.5569, 0.5093, 0.5712, 0.5321, 0.4103, 0.5296, 0.608, 0.6929, 0.7397, 0.7724, 0.6828, 0.5907, 0.6107, 0.8767, 0.7685, 0.8362, 0.7853, 0.7299, 0.5771, 0.5862, 0.547, 0.7189, 0.6875, 0.6875, 0.6232, 0.608, 0.5337, 0.5236, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021129556036645202", "ticker_idx": 2, "ts": "2026-02-10T07:50:11+00:00", "symbol": "COHR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Coherent mentioned it has secured a significant order from a leading AI data center customer to supply CW Lasers for its CPO systems. Initial revenue is expected to begin in 4Q26, with more meaningful contributions starting in 2027", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF International: Buy NVDA on CPO roadmap acceleration; bullish Lumentum on CW Laser demand; neutral Coherent; expect Browave (3163 TT) to benefit from CPO Shuttle Box ramp.", "resolved_tickers": ["COHR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Scientific & Technical Instruments'", "bench_c_industry": "Scientific & Technical Instruments", "unit_kind": "original", "lang": "en", "tweet_text": "《GF International Electronics & Communications》\n\nAI Networking: NVIDIA's Accelerated Roadmap, CPO Technology Debate Heats Up\n\nKey CPO Takeaways from Recent Earnings Calls: Coherent mentioned it has secured a significant order from a leading AI data center customer to supply CW Lasers for its CPO systems. Initial revenue is expected to begin in 4Q26, with more meaningful contributions starting in 2027. Lumentum stated it has received hundreds of millions of dollars in CW Laser commitments for Scale-out CPO, with volume production contributions beginning in 1H27. Assuming combined order value of $400 million for laser suppliers such as LITE and COHR in 2027, this implies approximately 80,000 Scale-out CPO switches will be shipped that year. We believe this CPO switch shipment volume is consistent with downstream suppliers' capacity buildout plans.\n\nNVIDIA May Accelerate Scale-out CPO Switches: At GTC 2025, NVIDIA (NVDA, Buy) unveiled its CPO switches, with Quantum-X planned for launch in 2H25 and Spectrum-X in 2H26. NVIDIA's first-generation Scale-out CPO (Quantum-X) adopts a pluggable architecture rather than true CPO. The system supports 115.2T switching bandwidth and integrates 36 pluggable optical engines, making it more accurately classified as NPO. However, due to its lack of cost-performance advantage over traditional pluggable solutions, customer adoption has been limited, with minimal shipments in 2025. For GTC 2026, we expect NVIDIA may unveil a next-generation CPO switch. This generation is expected to feature 115.2T bandwidth, with the CPO portion manufactured by TSMC. We believe this CPO design offers significant improvements in thermal performance and bandwidth, with cost-performance far exceeding the previous generation. The supply chain is expected to begin ramping in 2Q26 and accelerate through 2H26/2027. Driven by NVIDIA's aggressive push and bundling sales strategy, we are raising our NVIDIA Scale-out CPO switch shipment estimates for 2025/2026/2027 to 2,000/20,000/80,000 units. As for 2027, despite the surge in shipments, our estimate represents only a single-digit percentage of total Scale-out switch shipments for the year.\n\nNVIDIA Scale-up CPO Progress Update: According to recent earnings calls, Lumentum emphasized that as copper interconnects approach their limits, optical Scale-up represents a long-term structural opportunity beginning in late 2027. As we noted in our report, we expect NVIDIA to introduce CPO/NPO within the NVL576 architecture starting with Rubin Ultra in 2H27, specifically for Scale-up interconnects. Within the NVL576 architecture, Compute trays and Switch trays are expected to continue relying on backplane connections, while inter-rack interconnects may transition to CPO- or NPO-based optical interconnects. Technically, CPO holds clear advantages in power consumption and bandwidth density, while NPO is easier to manufacture and maintain. We believe TSMC's Scale-up CPO solution has already provided samples, but mass production readiness remains uncertain, indicating that the NPO vs. CPO debate will persist in the near term.\n\nKey Beneficiary Stocks: Overall, Scale-up CPO represents a purely incremental opportunity for the optical interconnect supply chain, as it aims to replace copper interconnects without affecting Scale-out. Furthermore, Scale-up and Scale-out CPO solutions share the same suppliers. Accordingly, we expect the primary beneficiaries to be component suppliers of FAUs, CW Lasers, ELS FP modules, and Shuffles. Our views: 1) Bullish on Lumentum, given upside potential in CW Laser demand; 2) Neutral impact on Coherent; 3) We expect Browave (3163 TT) to benefit, given its approximately 50% market share in CPO Shuttle Boxes, unit price of $5,000–6,000, and volume production ramp beginning in 2Q26.\n\n$COHR $LITE $NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06169817356894636, "ret_signed_m1d": 0.06169817356894636, "alpha_spy_m1d": 0.059054040132790586, "alpha_c_m1d": 0.05608607499732865, "bench_spy_m1d": 0.002644133436155771, "bench_c_m1d": 0.005612098571617707, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020493027875136183, "ret_signed_p1d": -0.020493027875136183, "alpha_spy_p1d": -0.020261969060719864, "alpha_c_p1d": -0.02343936890776066, "bench_spy_p1d": -0.00023105881441631837, "bench_c_p1d": 0.0029463410326244777, "ret_p1w": -0.036826153098636016, "ret_signed_p1w": -0.036826153098636016, "alpha_spy_p1w": -0.023432523406950567, "alpha_c_p1w": -0.015289815928589578, "bench_spy_p1w": -0.01339362969168545, "bench_c_p1w": -0.021536337170046438, "ret_p1m": 0.10088894792461223, "ret_signed_p1m": 0.10088894792461223, "alpha_spy_p1m": 0.12370284459578251, "alpha_c_p1m": 0.11576105200929909, "bench_spy_p1m": -0.022813896671170286, "bench_c_p1m": -0.014872104084686866, "ret_p3m": 0.46805629869890364, "ret_signed_p3m": 0.46805629869890364, "alpha_spy_p3m": 0.3994056565313686, "alpha_c_p3m": 0.23523812580545544, "bench_spy_p3m": 0.06865064216753503, "bench_c_p3m": 0.2328181728934482, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3871, -0.3899, -0.391, -0.3951, -0.3741, -0.4062, -0.3891, -0.3352, -0.3482, -0.3257, -0.3257, -0.2807, -0.2841, -0.278, -0.2514, -0.2234, -0.204, -0.1861, -0.1561, -0.1354, -0.1308, -0.2191, -0.2186, -0.2328, -0.2537, -0.2306, -0.1863, -0.1637, -0.1598, -0.162, -0.162, -0.1605, -0.1723, -0.182, -0.1918, -0.1918, -0.1491, -0.184, -0.15, -0.1609, -0.2418, -0.2203, -0.1891, -0.1679, -0.1938, -0.1419, -0.1635, -0.1635, -0.1529, -0.1178, -0.1123, -0.1376, -0.134, -0.0629, -0.0317, -0.0548, -0.0709, -0.026, 0.0035, -0.0761, -0.0838, -0.003, 0.0617, 0.0, -0.0205, -0.0537, -0.0488, -0.0488, -0.0368, -0.0196, 0.018, 0.0867, 0.0899, 0.116, 0.1731, 0.0953, 0.1338, 0.3089, 0.2296, 0.2036, 0.1117, 0.0322, 0.1049, 0.1413, 0.1009, 0.0565, 0.063, 0.0832, 0.0763, 0.1263, 0.2067, 0.1106, 0.1168, 0.1925, 0.1912, 0.0653, 0.0662, -0.0382, 0.0431, 0.0851, 0.1304, 0.1304, 0.1088, 0.117, 0.2339, 0.2443, 0.3465, 0.3484, 0.3724, 0.3496, 0.4363, 0.5108, 0.5217, 0.5054, 0.5347, 0.4787, 0.4717, 0.4079, 0.331, 0.3352, 0.4, 0.4428, 0.4445, 0.4701, 0.5093, 0.3977, 0.4681, 0.6626, 0.6377, 0.7678, 0.7732, 0.6747, 0.5888, 0.5485, 0.5698, 0.6552, 0.6533, 0.6533, 0.6699, 0.6648, 0.6506, 0.5828, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2028803497630835049", "ticker_idx": 0, "ts": "2026-03-03T12:03:41+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "memory pricing will cool off", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Memory pricing expected to cool off; sustainability of price rises in doubt as hyperscalers signal budget constraints.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "IMO, memory pricing will cool off. The hyperscalers keep complaining that they’re out of money.\n\n---\n\nNow, what matters isn’t how much memory prices rise, but whether they can be sustained.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.10883394033501388, "ret_signed_m1d": -0.10883394033501388, "alpha_spy_m1d": -0.0999412183909701, "alpha_c_m1d": -0.06963275078468974, "bench_spy_m1d": 0.00889272194404378, "bench_c_m1d": 0.039201189550324145, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05255855072879941, "ret_signed_p1d": 0.05255855072879941, "alpha_spy_p1d": 0.05961385908188407, "alpha_c_p1d": 0.0731180776431247, "bench_spy_p1d": 0.007055308353084655, "bench_c_p1d": 0.020559526914325277, "ret_p1w": 0.007913571996201943, "ret_signed_p1w": -0.007913571996201943, "alpha_spy_p1w": -0.012543764286369533, "alpha_c_p1w": 0.00811974527689093, "bench_spy_p1w": -0.00463019229016759, "bench_c_p1w": 0.016033317273092873, "ret_p1m": -0.03380055713461242, "ret_signed_p1m": 0.03380055713461242, "alpha_spy_p1m": -0.0004483824056194727, "alpha_c_p1m": 0.03612757515688673, "bench_spy_p1m": -0.03424893954023189, "bench_c_p1m": 0.0023270180222743075, "ret_p3m": 1.2238784847431432, "ret_signed_p3m": -1.2238784847431432, "alpha_spy_p3m": -1.1089109091332365, "alpha_c_p3m": -0.6923232338483194, "bench_spy_p3m": 0.11496757560990667, "bench_c_p3m": 0.5315552508948238, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4304, -0.4148, -0.3927, -0.3936, -0.377, -0.3905, -0.4009, -0.4171, -0.4334, -0.4234, -0.4033, -0.3921, -0.3639, -0.361, -0.3512, -0.3512, -0.3389, -0.3106, -0.3076, -0.3139, -0.3139, -0.2637, -0.2434, -0.2039, -0.1981, -0.2079, -0.1959, -0.1938, -0.2066, -0.2047, -0.1932, -0.1593, -0.1557, -0.1682, -0.1408, -0.1232, -0.1175, -0.1377, -0.0839, -0.0424, -0.0378, -0.0395, -0.0646, -0.0242, -0.0591, -0.0933, -0.0852, -0.0647, -0.0786, -0.0483, -0.0168, -0.0172, -0.0172, -0.0276, -0.009, 0.0078, 0.037, 0.0363, 0.0648, 0.085, 0.1274, 0.1086, 0.1088, 0.0, -0.0526, 0.01, -0.0253, -0.065, 0.0079, 0.0312, 0.007, 0.0107, 0.056, 0.081, 0.1365, 0.0922, 0.0694, 0.0045, 0.0214, 0.0116, -0.049, -0.0475, -0.1056, -0.1305, -0.0338, -0.0783, -0.0486, -0.0231, -0.0068, 0.0843, 0.0738, 0.0867, 0.088, 0.1541, 0.1652, 0.184, 0.1699, 0.175, 0.2041, 0.2347, 0.2423, 0.2459, 0.3037, 0.2845, 0.3013, 0.2882, 0.3102, 0.418, 0.474, 0.6092, 0.6199, 0.7141, 0.8547, 0.8024, 0.8935, 0.8877, 0.7453, 0.7329, 0.7048, 0.7349, 0.8708, 0.8494, 0.8494, 1.0272, 1.1373, 1.1366, 1.2239, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030881520282702021", "ticker_idx": 0, "ts": "2026-03-09T05:41:01+00:00", "symbol": "CIEN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Cisco and Ciena are identified as the leading beneficiaries of this optical network upgrade cycle. Ciena holds approximately 30% market share in optical transport and maintains a competitive advantage through its WaveLogic DSP technology", "tweet_type": "analysis", "timeline": "years", "summary": "Long Ciena and Cisco as leading beneficiaries of AI-driven optical networking supercycle; Scale-Across DCI and 800G coherent module demand to drive prolonged growth.", "resolved_tickers": ["CIEN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "self_thread", "lang": "en", "tweet_text": "OFC 2026 Preview: The New Optical Module Supercycle Driven by Scale-Across\n\n■ As AI infrastructure investment expands, datacenter networking is emerging as a new bottleneck. While the previous AI investment cycle was centered on GPUs and servers, the optical network infrastructure responsible for data movement is now rising as the critical limiting factor on performance. Accordingly, analysts are arguing that the optical communications industry is entering a new supercycle.\n\n■ Datacenter architecture is also transitioning from the traditional Scale-Up model to a Scale-Across paradigm. Previously, the dominant approach was to increase GPU counts within a single datacenter, but as model sizes have grown dramatically, limitations in power, cooling, and network architecture have become apparent. As a result, configurations that connect multiple datacenters via high-speed optical networks are proliferating, and demand for datacenter interconnect (DCI) is surging.\n\n■ This structural shift is driving rapid growth in the optical module market. In particular, demand for coherent optical modules — used for long-distance datacenter connectivity — is growing quickly, and the technology generation is rapidly transitioning from 400G to 800G. Market data suggests that the 800G optical module market will grow approximately 10x year-over-year in 2026.\n\n■ Changes are also visible at the network architecture level. Rather than the integrated optical systems traditionally supplied by telecom equipment vendors, hyperscalers are adopting ZR/ZR+ pluggable optical module architectures that insert directly into routers and switches. This approach offers advantages in cost reduction, power efficiency, and network design flexibility, and is becoming the standard architecture for hyperscale datacenters.\n\n■ Cisco and Ciena are identified as the leading beneficiaries of this optical network upgrade cycle. Ciena holds approximately 30% market share in optical transport and maintains a competitive advantage through its WaveLogic DSP technology, while Cisco has secured a strong position in the pluggable optical module market. As AI infrastructure investment enters a phase where both compute and networking grow in tandem — rather than compute alone — the optical communications industry is expected to enter a prolonged growth phase.\n\n---\n\nhttps://t.co/PKn7hZQTL7", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07650528807514223, "ret_signed_m1d": -0.07650528807514223, "alpha_spy_m1d": -0.067821334710787, "alpha_c_m1d": -0.067821334710787, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.008683953364355235, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.059081987511112155, "ret_signed_p1d": 0.059081987511112155, "alpha_spy_p1d": 0.060689114416479284, "alpha_c_p1d": 0.060689114416479284, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": -0.0016071269053671289, "ret_p1w": 0.14233689684696738, "ret_signed_p1w": 0.14233689684696738, "alpha_spy_p1w": 0.15595980357116612, "alpha_c_p1w": 0.15595980357116612, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.013622906724198747, "ret_p1m": 0.40604003537486744, "ret_signed_p1m": 0.40604003537486744, "alpha_spy_p1m": 0.4314719997944799, "alpha_c_p1m": 0.4314719997944799, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": -0.025431964419612485, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3035, -0.2391, -0.3142, -0.3213, -0.3412, -0.3599, -0.3385, -0.2769, -0.2568, -0.2457, -0.2457, -0.2457, -0.2413, -0.2497, -0.2517, -0.2658, -0.2658, -0.2275, -0.2751, -0.202, -0.1861, -0.293, -0.2762, -0.2646, -0.2264, -0.2516, -0.2347, -0.2358, -0.2358, -0.2428, -0.272, -0.2779, -0.2807, -0.271, -0.223, -0.1923, -0.2041, -0.2095, -0.1571, -0.1319, -0.2042, -0.2058, -0.1482, -0.0906, -0.0623, -0.0668, -0.0803, -0.0459, -0.0459, -0.0509, -0.0238, -0.0004, 0.0515, 0.0826, 0.0758, 0.1092, 0.071, 0.0947, 0.1105, 0.0456, 0.0785, -0.0604, -0.0765, 0.0, 0.0591, 0.0672, 0.0574, 0.0591, 0.1423, 0.1617, 0.2095, 0.2952, 0.2052, 0.2805, 0.348, 0.3741, 0.218, 0.2608, 0.1459, 0.2188, 0.304, 0.4057, 0.4057, 0.3633, 0.406, 0.5509, 0.53, 0.5572, 0.5066, 0.4667, 0.4936, 0.5545, 0.593, 0.5799, 0.5883, 0.5664, 0.6194, 0.635, 0.5896, 0.4871, 0.4924, 0.6562, 0.6804, 0.6906, 0.7101, 0.8107, 0.6913, 0.7207, 0.8254, 0.8119, 0.8142, 0.8571, 0.7406, 0.6467, 0.7013, 0.7416, 0.8435, 0.8325, 0.8325, 0.8911, 0.8273, 0.79, 0.8215, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030881520282702021", "ticker_idx": 1, "ts": "2026-03-09T05:41:01+00:00", "symbol": "CSCO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Cisco has secured a strong position in the pluggable optical module market", "tweet_type": "analysis", "timeline": "years", "summary": "Long Ciena and Cisco as leading beneficiaries of AI-driven optical networking supercycle; Scale-Across DCI and 800G coherent module demand to drive prolonged growth.", "resolved_tickers": ["CSCO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "self_thread", "lang": "en", "tweet_text": "OFC 2026 Preview: The New Optical Module Supercycle Driven by Scale-Across\n\n■ As AI infrastructure investment expands, datacenter networking is emerging as a new bottleneck. While the previous AI investment cycle was centered on GPUs and servers, the optical network infrastructure responsible for data movement is now rising as the critical limiting factor on performance. Accordingly, analysts are arguing that the optical communications industry is entering a new supercycle.\n\n■ Datacenter architecture is also transitioning from the traditional Scale-Up model to a Scale-Across paradigm. Previously, the dominant approach was to increase GPU counts within a single datacenter, but as model sizes have grown dramatically, limitations in power, cooling, and network architecture have become apparent. As a result, configurations that connect multiple datacenters via high-speed optical networks are proliferating, and demand for datacenter interconnect (DCI) is surging.\n\n■ This structural shift is driving rapid growth in the optical module market. In particular, demand for coherent optical modules — used for long-distance datacenter connectivity — is growing quickly, and the technology generation is rapidly transitioning from 400G to 800G. Market data suggests that the 800G optical module market will grow approximately 10x year-over-year in 2026.\n\n■ Changes are also visible at the network architecture level. Rather than the integrated optical systems traditionally supplied by telecom equipment vendors, hyperscalers are adopting ZR/ZR+ pluggable optical module architectures that insert directly into routers and switches. This approach offers advantages in cost reduction, power efficiency, and network design flexibility, and is becoming the standard architecture for hyperscale datacenters.\n\n■ Cisco and Ciena are identified as the leading beneficiaries of this optical network upgrade cycle. Ciena holds approximately 30% market share in optical transport and maintains a competitive advantage through its WaveLogic DSP technology, while Cisco has secured a strong position in the pluggable optical module market. As AI infrastructure investment enters a phase where both compute and networking grow in tandem — rather than compute alone — the optical communications industry is expected to enter a prolonged growth phase.\n\n---\n\nhttps://t.co/PKn7hZQTL7", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03188553946075556, "ret_signed_m1d": 0.03188553946075556, "alpha_spy_m1d": 0.0405694928251108, "alpha_c_m1d": 0.0405694928251108, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.008683953364355235, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019551165849132524, "ret_signed_p1d": 0.019551165849132524, "alpha_spy_p1d": 0.021158292754499652, "alpha_c_p1d": 0.021158292754499652, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": -0.0016071269053671289, "ret_p1w": 0.03529726037683134, "ret_signed_p1w": 0.03529726037683134, "alpha_spy_p1w": 0.04892016710103009, "alpha_c_p1w": 0.04892016710103009, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.013622906724198747, "ret_p1m": 0.06439019313424388, "ret_signed_p1m": 0.06439019313424388, "alpha_spy_p1m": 0.08982215755385636, "alpha_c_p1m": 0.08982215755385636, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": -0.025431964419612485, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0474, 0.0346, 0.0154, 0.0213, 0.0122, -0.0081, 0.0043, 0.0235, 0.0195, 0.0183, 0.0183, 0.0183, 0.0201, 0.0153, 0.0103, 0.0054, 0.0054, -0.0022, -0.0083, -0.0129, -0.0196, -0.0295, -0.0306, -0.0289, -0.0097, -0.0236, -0.0126, -0.0134, -0.0134, -0.0375, -0.0331, -0.0247, -0.0213, 0.0105, 0.0324, 0.0361, 0.0291, 0.0277, 0.0581, 0.0905, 0.065, 0.0807, 0.113, 0.1387, 0.1323, 0.1224, -0.0159, 0.0084, 0.0084, 0.0084, 0.0258, 0.0308, 0.0392, 0.0201, 0.0253, 0.0382, 0.0248, 0.0426, 0.0421, 0.0361, 0.0611, 0.0499, 0.0319, 0.0, 0.0196, 0.0248, 0.0201, 0.0278, 0.0353, 0.0402, 0.0182, 0.0302, 0.0189, 0.0342, 0.061, 0.0737, 0.0781, 0.0487, 0.0109, 0.0181, 0.0226, 0.0425, 0.0425, 0.0612, 0.0644, 0.1042, 0.0972, 0.0847, 0.0864, 0.0899, 0.0866, 0.1148, 0.1379, 0.1571, 0.1834, 0.1847, 0.1687, 0.1743, 0.1644, 0.1459, 0.1817, 0.2071, 0.2118, 0.222, 0.2441, 0.209, 0.2158, 0.274, 0.3024, 0.3099, 0.3439, 0.5242, 0.5595, 0.5684, 0.5222, 0.5086, 0.5594, 0.5885, 0.5885, 0.5611, 0.5788, 0.5652, 0.5887, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014572273756471368", "ticker_idx": 0, "ts": "2026-01-23T05:33:53+00:00", "symbol": "CLS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we forecast 2026E/2027E EPS at 8.7/12.3 USD. The target price remains at 270 USD, based on a 22x 2027E P/E ratio", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF maintains Hold/PT $270 on CLS but flags TPU + networking upside; also flags Hon Hai capturing ~20% TPU share vs 5% consensus.", "resolved_tickers": ["CLS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics & Communications》\n\nCelestica (CLS US Hold): TPU and Network Upside Potential Offsets ODM Share Rebalancing  \n\nMaintain Hold Rating and Target Price of 270 USD: We expect Google's TPU server supply chain to further expand in 2026, with Hon Hai starting L10/L11 production in 1Q26 and potentially capturing a higher-than-expected share. Celestica remains a key supplier for Google's TPU L6 and rack-level solutions. Meanwhile, CLS's communications segment benefits from accelerating network demand, including continued growth in 800G and the ramp-up of 1.6T in 2026, which will support upside potential in gross margins and EPS. Therefore, we forecast 2026E/2027E EPS at 8.7/12.3 USD. The target price remains at 270 USD, based on a 22x 2027E P/E ratio.\n\nGoogle's TPU Server Supply Chain Further Expands: Hon Hai is expected to start TPU server L10/L11 production in 1Q26. We anticipate Hon Hai could capture about 20% market share, higher than the current consensus of about 5%, considering Hon Hai's strong manufacturing capabilities and scale. We believe this reflects a broader supply chain rebalancing, influenced by overall volume allocation. CLS remains the key leading provider for Google's TPU L6 and rack-level solutions. At the same time, we believe Meta has halted further development of the Iris project (retaining Iris Plus) due to increased design complexity. In 2026, Iris's CoWoS capacity is expected to shift to Google, implying a downward revision in Meta AI server revenue and a higher contribution from TPU.\n\nCommunications Segment to Continue Benefiting Significantly from Rapid Industry Growth in 2026/27: We expect continued upward revisions in Google's TPU demand to drive corresponding upgrades in the network supply chain, including optical modules, lasers, and switches. We anticipate sustained growth in 800G switch demand next year, while 1.6T switches will begin ramping up in the second half of this year and accelerate rapidly in 2026. CLS is well-positioned to capture this upside potential; moreover, given that switch gross margins are higher than servers, we believe the network business will provide a more significant positive contribution to overall gross margins and EPS. Additionally, we are optimistic about emerging O trends; as a key contract manufacturer in the Google ecosystem, we expect CLS to also have growth potential in O switch assembly.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.040120108908355934, "ret_signed_m1d": -0.040120108908355934, "alpha_spy_m1d": -0.03975746023038973, "alpha_c_m1d": -0.03867278080555092, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": -0.0014473281028050122, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017024658432992057, "ret_signed_p1d": 0.017024658432992057, "alpha_spy_p1d": 0.01194651137238223, "alpha_c_p1d": 0.01020120929586077, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": 0.006823449137131288, "ret_p1w": -0.0729156566384267, "ret_signed_p1w": -0.0729156566384267, "alpha_spy_p1w": -0.07689111729282117, "alpha_c_p1w": -0.06457611946752406, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": -0.008339537170902633, "ret_p1m": -0.021148846017878364, "ret_signed_p1m": -0.021148846017878364, "alpha_spy_p1m": -0.011224845956785745, "alpha_c_p1m": 0.024133306620744532, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": -0.0452821526386229, "ret_p3m": 0.3254808640616018, "ret_signed_p3m": 0.3254808640616018, "alpha_spy_p3m": 0.29077199351865524, "alpha_c_p3m": 0.23452653308098825, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.09095433098061356, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0771, 0.1144, 0.1222, 0.1366, 0.1515, 0.1079, 0.1634, 0.1243, 0.0627, 0.137, 0.0949, 0.1039, -0.0341, 0.0257, 0.0207, -0.0105, 0.031, -0.0687, -0.076, 0.0642, 0.0814, 0.0962, 0.0962, 0.1363, 0.0504, 0.004, 0.0176, 0.065, 0.0726, 0.1243, 0.1285, 0.1504, 0.1594, 0.0113, -0.0004, -0.0469, -0.0899, -0.1061, -0.0356, 0.0124, 0.0012, 0.0181, 0.0181, 0.0016, 0.0004, -0.0122, -0.0247, -0.0247, -0.0023, -0.0325, 0.0063, 0.0092, -0.0478, -0.0026, 0.0383, 0.084, 0.0229, 0.0344, 0.0347, 0.0347, 0.0253, 0.022, -0.0401, 0.0, 0.017, 0.0992, 0.139, -0.0102, -0.0729, -0.0593, -0.0186, -0.0898, -0.0271, 0.0146, 0.0532, -0.017, -0.0244, -0.0934, -0.074, -0.074, -0.0542, -0.0386, -0.0388, -0.0343, -0.0211, -0.0222, -0.0301, -0.079, -0.084, -0.1191, -0.1522, -0.1071, -0.1229, -0.1767, -0.1159, -0.1118, -0.123, -0.126, -0.1308, -0.1074, -0.0697, -0.0974, -0.062, -0.1121, -0.0532, -0.0052, -0.0029, -0.098, -0.0755, -0.1512, -0.0706, -0.0474, -0.0272, -0.0272, -0.0356, -0.019, 0.0581, 0.0833, 0.1591, 0.2058, 0.2681, 0.2602, 0.2614, 0.3066, 0.3234, 0.3297, 0.3255, 0.292, 0.3534, 0.393, 0.1928, 0.2423, 0.3514, 0.3822, 0.3881, 0.378, 0.3655, 0.2709, 0.2391, 0.257, 0.2342, 0.2292, 0.259, 0.183, 0.1306, 0.1189, 0.1432, 0.1705, 0.2121, 0.2121, 0.2235, 0.1802, 0.1581, 0.2715, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014572273756471368", "ticker_idx": 1, "ts": "2026-01-23T05:33:53+00:00", "symbol": "Hon Hai Precision", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hon Hai is expected to start TPU server L10/L11 production in 1Q26. We anticipate Hon Hai could capture about 20% market share, higher than the current consensus of about 5%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF maintains Hold/PT $270 on CLS but flags TPU + networking upside; also flags Hon Hai capturing ~20% TPU share vs 5% consensus.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision 2317.TW Taiwan (Foxconn)", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics & Communications》\n\nCelestica (CLS US Hold): TPU and Network Upside Potential Offsets ODM Share Rebalancing  \n\nMaintain Hold Rating and Target Price of 270 USD: We expect Google's TPU server supply chain to further expand in 2026, with Hon Hai starting L10/L11 production in 1Q26 and potentially capturing a higher-than-expected share. Celestica remains a key supplier for Google's TPU L6 and rack-level solutions. Meanwhile, CLS's communications segment benefits from accelerating network demand, including continued growth in 800G and the ramp-up of 1.6T in 2026, which will support upside potential in gross margins and EPS. Therefore, we forecast 2026E/2027E EPS at 8.7/12.3 USD. The target price remains at 270 USD, based on a 22x 2027E P/E ratio.\n\nGoogle's TPU Server Supply Chain Further Expands: Hon Hai is expected to start TPU server L10/L11 production in 1Q26. We anticipate Hon Hai could capture about 20% market share, higher than the current consensus of about 5%, considering Hon Hai's strong manufacturing capabilities and scale. We believe this reflects a broader supply chain rebalancing, influenced by overall volume allocation. CLS remains the key leading provider for Google's TPU L6 and rack-level solutions. At the same time, we believe Meta has halted further development of the Iris project (retaining Iris Plus) due to increased design complexity. In 2026, Iris's CoWoS capacity is expected to shift to Google, implying a downward revision in Meta AI server revenue and a higher contribution from TPU.\n\nCommunications Segment to Continue Benefiting Significantly from Rapid Industry Growth in 2026/27: We expect continued upward revisions in Google's TPU demand to drive corresponding upgrades in the network supply chain, including optical modules, lasers, and switches. We anticipate sustained growth in 800G switch demand next year, while 1.6T switches will begin ramping up in the second half of this year and accelerate rapidly in 2026. CLS is well-positioned to capture this upside potential; moreover, given that switch gross margins are higher than servers, we believe the network business will provide a more significant positive contribution to overall gross margins and EPS. Additionally, we are optimistic about emerging O trends; as a key contract manufacturer in the Google ecosystem, we expect CLS to also have growth potential in O switch assembly.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009029345372460584, "ret_signed_m1d": 0.009029345372460584, "alpha_spy_m1d": 0.009391994050426788, "alpha_c_m1d": 0.010476673475265597, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": -0.0014473281028050122, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011286681715575675, "ret_signed_p1d": 0.011286681715575675, "alpha_spy_p1d": 0.006208534654965847, "alpha_c_p1d": 0.004463232578444387, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": 0.006823449137131288, "ret_p1w": -0.004514672686230292, "ret_signed_p1w": -0.004514672686230292, "alpha_spy_p1w": -0.008490133340624761, "alpha_c_p1w": 0.003824864484672341, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": -0.008339537170902633, "ret_p1m": 0.03160270880361171, "ret_signed_p1m": 0.03160270880361171, "alpha_spy_p1m": 0.04152670886470433, "alpha_c_p1m": 0.07688486144223461, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": -0.0452821526386229, "ret_p3m": -0.0022573363431150906, "ret_signed_p3m": -0.0022573363431150906, "alpha_spy_p3m": -0.03696620688606167, "alpha_c_p3m": -0.09321166732372865, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.09095433098061356, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1287, 0.1738, 0.1828, 0.1625, 0.1354, 0.1016, 0.1151, 0.1196, 0.1016, 0.1264, 0.1129, 0.1332, 0.1377, 0.088, 0.0655, 0.0384, 0.0339, 0.0677, 0.0158, -0.0068, -0.0113, 0.0248, 0.0429, 0.0181, 0.0023, 0.0023, 0.0248, 0.0316, 0.0429, 0.0474, 0.0609, 0.0542, 0.0203, 0.0248, 0.0, -0.0158, -0.0226, -0.0248, 0.0, 0.0113, 0.0113, 0.0113, 0.0113, 0.0181, 0.0429, 0.0293, 0.0406, 0.0406, 0.0474, 0.0587, 0.0655, 0.0767, 0.0361, 0.0406, 0.0316, 0.0226, 0.0587, 0.0542, 0.0587, 0.0361, 0.009, -0.0113, 0.009, 0.0, 0.0113, 0.0181, 0.0181, 0.0113, -0.0045, -0.0316, -0.0226, -0.0113, -0.0271, -0.0293, -0.0135, -0.0023, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0316, 0.0474, 0.1129, 0.0971, 0.0971, 0.079, 0.0339, -0.0203, 0.0113, 0.0068, -0.0497, -0.0542, -0.0113, -0.0316, -0.0316, -0.0226, -0.0429, -0.0519, -0.0745, -0.0835, -0.1151, -0.1196, -0.0971, -0.0948, -0.0993, -0.1242, -0.1535, -0.1106, -0.1287, -0.1287, -0.1287, -0.1332, -0.0903, -0.0971, -0.0948, -0.0971, -0.0632, -0.0632, -0.0655, -0.07, -0.0632, -0.0474, -0.0023, 0.0158, 0.0, 0.0293, 0.0181, 0.0158, -0.009, -0.009, 0.0271, 0.0813, 0.1377, 0.1445, 0.1287, 0.1377, 0.1287, 0.1332, 0.1038, 0.1219, 0.1219, 0.1061, 0.0835, 0.1174, 0.1287, 0.1783, 0.1693, 0.1919, 0.1874, 0.3047, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027032116614287439", "ticker_idx": 0, "ts": "2026-02-26T14:44:51+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It's still cheap, and the competitive edge is really strong. Still, I can't bring myself to sell after holding it this long, so I'm just going to keep HOLDing…", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Continuing to hold NVDA; views it as still cheap with strong competitive edge despite underperformance vs Korean stocks.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "My biggest mistake this year was holding Nvidia. I should’ve bought Korean stocks with that money instead!\n\n---\n\nStill, I can’t bring myself to sell after holding it this long, so I’m just going to keep HOLDing…\n\nIt’s still cheap, and the competitive edge is really strong.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0577099682700033, "ret_signed_m1d": 0.0577099682700033, "alpha_spy_m1d": 0.052124467345360914, "alpha_c_m1d": 0.023366157291882406, "bench_spy_m1d": 0.005585500924642384, "bench_c_m1d": 0.03434381097812089, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.041646311846979644, "ret_signed_p1d": -0.041646311846979644, "alpha_spy_p1d": -0.0368443685884523, "alpha_c_p1d": -0.027957288486489262, "bench_spy_p1d": -0.004801943258527341, "bench_c_p1d": -0.013689023360490382, "ret_p1w": -0.008383338305975596, "ret_signed_p1w": -0.008383338305975596, "alpha_spy_p1w": 0.003208076636686652, "alpha_c_p1w": 0.032052581373710254, "bench_spy_p1w": -0.011591414942662248, "bench_c_p1w": -0.04043591967968585, "ret_p1m": -0.09389867116969342, "ret_signed_p1m": -0.09389867116969342, "alpha_spy_p1m": -0.016315274885381537, "alpha_c_p1m": -0.002250388675039794, "bench_spy_p1m": -0.07758339628431188, "bench_c_p1m": -0.09164828249465362, "ret_p3m": 0.16215930694985548, "ret_signed_p3m": 0.16215930694985548, "alpha_spy_p3m": 0.07026907996052456, "alpha_c_p3m": -0.2993100757745337, "bench_spy_p3m": 0.09189022698933091, "bench_c_p3m": 0.4614693827243892, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0269, -0.0186, -0.0287, -0.0082, -0.0134, 0.0036, 0.0004, -0.006, -0.0214, -0.0534, -0.0465, -0.0388, -0.0755, -0.0581, -0.0211, -0.0065, 0.0234, 0.0201, 0.0201, 0.0305, 0.018, 0.0143, 0.0087, 0.0087, 0.0214, 0.0175, 0.0127, 0.0228, 0.0008, -0.0002, 0.0003, 0.005, -0.0095, 0.0117, 0.0072, 0.0072, -0.0369, -0.0085, -0.0003, 0.015, 0.0085, 0.0196, 0.0359, 0.0412, 0.0337, 0.0039, -0.0246, -0.0579, -0.0704, 0.0028, 0.0279, 0.0197, 0.0279, 0.0111, -0.0112, -0.0112, 0.0004, 0.0167, 0.0163, 0.0267, 0.036, 0.0431, 0.0577, 0.0, -0.0416, -0.013, -0.0262, -0.01, -0.0084, -0.0382, -0.0121, -0.0006, 0.0062, -0.0094, -0.025, -0.009, -0.016, -0.0242, -0.0342, -0.0659, -0.05, -0.0524, -0.0335, -0.0738, -0.0939, -0.1066, -0.0567, -0.0494, -0.0405, -0.0405, -0.0392, -0.0367, -0.0151, -0.0052, 0.0203, 0.024, 0.0629, 0.0757, 0.0729, 0.0909, 0.0929, 0.0811, 0.0953, 0.0798, 0.1265, 0.1716, 0.153, 0.1318, 0.0795, 0.0734, 0.0736, 0.0629, 0.1241, 0.144, 0.164, 0.1869, 0.1942, 0.2215, 0.2751, 0.2187, 0.2025, 0.1933, 0.2087, 0.1873, 0.1647, 0.1647, 0.1622, 0.1499, 0.1589, 0.142, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2024305826597593343", "ticker_idx": 0, "ts": "2026-02-19T02:11:33+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics—which holds a technical edge in specifications within the 6th-generation High Bandwidth Memory (HBM4) market—stands to benefit", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Samsung to benefit as NVIDIA's HBM4 dual-bin strategy favors top-spec suppliers; Samsung's 11.7 Gbps HBM4 positions it as the preferred top-bin source.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA's AI Accelerator 'Tiering' Coming... Focus on HBM4 Processing Speed\n\nBig tech companies including NVIDIA are expected to devise a 'high-performance-centric' memory supply strategy to maximize next-generation artificial intelligence (AI) accelerator performance. In this scenario, Samsung Electronics—which holds a technical edge in specifications within the 6th-generation High Bandwidth Memory (HBM4) market—stands to benefit.\n\nAccording to industry sources on the 19th, NVIDIA is reviewing a 'Dual Bin' structure to simultaneously secure HBM4 supply stability and performance. Dual binning is a strategy where chips based on the same design are supplied in top-tier and second-tier grades based on speed, power efficiency, and other criteria.\n\nBig tech companies are expected to apply 'top bin' chips rated at 11.7 Gbps or above to their flagship products, while using 'second bin' chips in the 10 Gbps range for complementary products in parallel.\n\nIn NVIDIA's case, rather than expanding lower-spec volume, the company is likely to focus on high-performance bins for its flagship products—driven by the symbolic importance of achieving peak AI accelerator performance.\n\nBig tech firms are grappling with memory bottlenecks as they develop next-generation AI accelerators. While compute units (GPUs) operate at high speeds, memory transfer speeds cannot keep pace, degrading overall system efficiency.\n\nSamsung Electronics' HBM4 operates at 11.7 Gbps on paper. It utilizes 1c (6th-generation 10nm-class) DRAM and a 4nm-based base die. This exceeds the JEDEC standard of 8 Gbps by approximately 46%. It represents a 1.22x improvement over HBM3E (9.6 Gbps), with prospects of achieving up to 13 Gbps going forward.\n\nAn industry official explained, \"Rather than simply focusing on volume procurement, NVIDIA is prioritizing accelerator performance maximization,\" adding that \"Samsung Electronics' influence will strengthen as it can stably produce top-bin parts.\"\n\n---\n\nhttps://t.co/xcTrDp3JsC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04631581532347695, "ret_signed_m1d": -0.04631581532347695, "alpha_spy_m1d": -0.0489602237163419, "alpha_c_m1d": -0.051308301206176465, "bench_spy_m1d": 0.0026444083928649498, "bench_c_m1d": 0.0049924858826995155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0005263338756542613, "ret_signed_p1d": 0.0005263338756542613, "alpha_spy_p1d": -0.006705530713468688, "alpha_c_p1d": -0.0042522568176495135, "bench_spy_p1d": 0.007231864589122949, "bench_c_p1d": 0.004778590693303775, "ret_p1w": 0.14736845835266799, "ret_signed_p1w": 0.14736845835266799, "alpha_spy_p1w": 0.14032659774781697, "alpha_c_p1w": 0.14166280689961908, "bench_spy_p1w": 0.0070418606048510135, "bench_c_p1w": 0.00570565145304891, "ret_p1m": 0.04947365376328716, "ret_signed_p1m": 0.04947365376328716, "alpha_spy_p1m": 0.09934902903233278, "alpha_c_p1m": 0.08456390598308994, "bench_spy_p1m": -0.04987537526904562, "bench_c_p1m": -0.035090252219802776, "ret_p3m": 0.45300123618826804, "ret_signed_p3m": 0.45300123618826804, "alpha_spy_p3m": 0.37812124717541606, "alpha_c_p3m": 0.21588975615123784, "bench_spy_p3m": 0.07487998901285198, "bench_c_p3m": 0.2371114800370302, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4935, -0.4799, -0.4616, -0.4579, -0.4736, -0.472, -0.4584, -0.4527, -0.4495, -0.4322, -0.4265, -0.4322, -0.4343, -0.438, -0.4296, -0.4511, -0.4616, -0.4349, -0.4364, -0.4432, -0.4212, -0.416, -0.4181, -0.4181, -0.3872, -0.3711, -0.3689, -0.3689, -0.3689, -0.3237, -0.2732, -0.2689, -0.2579, -0.2695, -0.2684, -0.2695, -0.2758, -0.2616, -0.2426, -0.2163, -0.2142, -0.2358, -0.2132, -0.1984, -0.1995, -0.1995, -0.1605, -0.1453, -0.1542, -0.1553, -0.2084, -0.1184, -0.11, -0.1616, -0.1653, -0.1242, -0.1274, -0.1168, -0.06, -0.0463, -0.0463, -0.0463, -0.0463, 0.0, 0.0005, 0.0158, 0.0526, 0.0711, 0.1474, 0.1395, 0.1395, 0.0268, -0.0937, 0.0084, -0.0095, -0.0868, -0.0111, 0.0, -0.0111, -0.0342, -0.0068, 0.0205, 0.0974, 0.0553, 0.0495, -0.0195, -0.0016, -0.0053, -0.0521, -0.0521, -0.0702, -0.1182, 0.0002, -0.0591, -0.018, 0.0184, 0.0364, 0.1102, 0.0759, 0.0865, 0.0601, 0.0891, 0.1128, 0.1471, 0.1392, 0.1313, 0.155, 0.1471, 0.184, 0.1577, 0.184, 0.1708, 0.1919, 0.1629, 0.1629, 0.2262, 0.2262, 0.4029, 0.4319, 0.4161, 0.5057, 0.4715, 0.4978, 0.5611, 0.4266, 0.482, 0.453, 0.4556, 0.5796, 0.5427, 0.5427, 0.5769, 0.6191, 0.5796, 0.6719, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034643710835818954", "ticker_idx": 0, "ts": "2026-03-19T14:50:37+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics will become the first and sole supplier of next-generation High Bandwidth Memory 4 (HBM4) to OpenAI", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung wins sole HBM4 supplier deal with OpenAI for Titan chip, following earlier wins with NVIDIA and AMD; cements leadership in advanced AI memory.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[EXCLUSIVE] Samsung Electronics Breaks Into OpenAI — Sole Supplier of 800 Million Gb HBM4\n\nSamsung Electronics will become the first and sole supplier of next-generation High Bandwidth Memory 4 (HBM4) to OpenAI, the world’s largest artificial intelligence company. OpenAI plans to integrate Samsung’s HBM4 into its first-generation in-house AI chip, codenamed “Titan.” With this win following its earlier HBM4 supply agreement with NVIDIA, Samsung is being credited with cementing its leadership in the advanced AI chip market.\n\nAccording to industry sources on the 19th, Samsung Electronics has agreed to supply OpenAI with up to 800 million gigabits (Gb) of HBM4 (12-layer product) in the second half of this year. That volume represents approximately 7% of Samsung’s total planned HBM output for the year (over 11 billion Gb), and roughly 15% of its HBM4-specific production (approximately 5.5 billion Gb). The allocation is understood to be the third largest after NVIDIA and AMD.\nThe HBM4 units Samsung will deliver are destined to sit alongside OpenAI’s first-ever AI chip, Titan Gen 1, developed in partnership with Broadcom. TSMC is expected to begin production in Q3, with a launch targeted for year-end.\n\nThe deal is seen as particularly significant given that OpenAI — the company that ignited the generative AI boom with the launch of ChatGPT in 2022 — selected Samsung as its inaugural HBM supplier. OpenAI also sits at the center of the United States’ Stargate Project, a planned $500 billion AI infrastructure initiative.\n\nFirst Fruits After the JY Lee–Altman Meeting… AI Chip Orders Keep Coming\nOpenAI operates hundreds of thousands of AI chips across its data centers to deliver generative AI services, having relied heavily on NVIDIA’s general-purpose AI semiconductors. More recently, the company concluded that it needed its own custom chips optimized for inference workloads — a trend that has become central to next-generation AI development.\nAn industry insider noted that “OpenAI has been investing heavily in R&D to successfully mass-produce its Titan chip,” adding that “Samsung satisfied the stringent HBM4 requirements that OpenAI set out, which is what made this deal possible.”\n\nGiven that OpenAI has chosen Samsung as its first-ever HBM supplier, observers believe Samsung HBM is likely to be incorporated into future Titan generations as well.\n\nHBM stacks multiple DRAM dies vertically — similar to floors in an apartment building — delivering greater capacity and faster data transfer speeds than conventional DRAM, making it the memory of choice for AI applications. Micron Technology has projected that the global HBM market will grow from approximately $35 billion in revenue last year to around $100 billion by 2028.\n\nSamsung’s HBM business had a difficult stretch through last year, suffering back-to-back failures in NVIDIA’s qualification tests for HBM3 and HBM3E — a significant blow to the pride of the world’s top memory chipmaker. The turnaround came in May 2024, when Vice Chairman Jeon Young-hyun made the bold decision to redesign the DRAM at the core of Samsung’s HBM.\n\nThis year, Samsung passed NVIDIA’s HBM4 qualification without a single design revision, enabling direct shipment of mass-production units. On March 18th, AMD announced it had designated Samsung as its preferred HBM4 supplier. Demand for the prior-generation HBM3E is also said to be surging. Samsung is reportedly targeting over 5 billion Gb of HBM3E for Google’s Tensor Processing Units (TPUs) — also developed in partnership with Broadcom — in addition to its NVIDIA supply commitments.\n\nBehind the OpenAI HBM4 deal, JY Lee, Chairman of Samsung Electronics, reportedly played a pivotal role. In October of last year, Lee met with OpenAI CEO Sam Altman and exchanged a Letter of Intent (LOI) covering the supply of cutting-edge AI memory including HBM, laying the groundwork for the agreement that has now come to fruition.\n\n---\n\nhttps://t.co/l9Qt2mh5nd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03990021471251204, "ret_signed_m1d": 0.03990021471251204, "alpha_spy_m1d": 0.03742978625044535, "alpha_c_m1d": 0.04329535880554847, "bench_spy_m1d": 0.002470428462066687, "bench_c_m1d": -0.0033951440930364285, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005486316616421294, "ret_signed_p1d": -0.005486316616421294, "alpha_spy_p1d": 0.008849459256082937, "alpha_c_p1d": 0.01719657451934231, "bench_spy_p1d": -0.01433577587250423, "bench_c_p1d": -0.022682891135763605, "ret_p1w": -0.10174561779923386, "ret_signed_p1w": -0.10174561779923386, "alpha_spy_p1w": -0.08212114184813557, "alpha_c_p1w": -0.06009823733450648, "bench_spy_p1w": -0.019624475951098286, "bench_c_p1w": -0.04164738046472738, "ret_p1m": 0.07953655832906104, "ret_signed_p1m": 0.07953655832906104, "alpha_spy_p1m": 0.0003013197953860569, "alpha_c_p1m": -0.03685387695085973, "bench_spy_p1m": 0.07923523853367498, "bench_c_p1m": 0.11639043527992077, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4515, -0.4466, -0.4486, -0.4486, -0.4193, -0.404, -0.402, -0.402, -0.402, -0.3591, -0.3112, -0.3072, -0.2968, -0.3077, -0.3067, -0.3077, -0.3137, -0.3002, -0.2823, -0.2574, -0.2554, -0.2758, -0.2544, -0.2404, -0.2414, -0.2414, -0.2045, -0.19, -0.1985, -0.1995, -0.2499, -0.1646, -0.1566, -0.2055, -0.209, -0.1701, -0.1731, -0.1631, -0.1092, -0.0963, -0.0963, -0.0963, -0.0963, -0.0524, -0.0519, -0.0374, -0.0025, 0.015, 0.0873, 0.0798, 0.0798, -0.0269, -0.1411, -0.0444, -0.0613, -0.1347, -0.0628, -0.0524, -0.0628, -0.0848, -0.0589, -0.0329, 0.0399, 0.0, -0.0055, -0.0708, -0.0539, -0.0574, -0.1017, -0.1017, -0.1189, -0.1644, -0.0522, -0.1084, -0.0694, -0.0349, -0.0179, 0.052, 0.0196, 0.0296, 0.0046, 0.0321, 0.0545, 0.087, 0.0795, 0.072, 0.0945, 0.087, 0.122, 0.097, 0.122, 0.1095, 0.1295, 0.102, 0.102, 0.162, 0.162, 0.3294, 0.3569, 0.3419, 0.4269, 0.3944, 0.4194, 0.4794, 0.3519, 0.4044, 0.3769, 0.3794, 0.4969, 0.4619, 0.4619, 0.4944, 0.5343, 0.4969, 0.5843, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2036963841582784947", "ticker_idx": 0, "ts": "2026-03-26T00:29:59+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I find it hard to recommend AMD outright. AMD's x86 CPU division is full of excellent people, but the business environment is not especially favorable. On top of that, AMD does not have its own fabs, so it has to compete with other companies for TSMC capacity.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Prefers TSMC and Intel as CPU bottleneck plays on tightening wafer capacity; cautious on AMD (no fabs, unfavorable environment) and ARM (smartphone royalty headwinds).", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Yesterday’s CPU bottleneck trade — among AMD, ARM, and Intel, which was actually the best trade?\n\nLet me share my view.\n\nObviously, CPU price increases are positive. They should help earnings materially.\n\nThat said, what we need to understand is that hyperscalers have been moving toward in-house CPU ASICs for quite a long time. They still procure x86 CPUs for external customers, but for internal workloads, many are already relying on their own CPU ASIC solutions.\n\nThat is why I find it hard to recommend AMD outright. AMD’s x86 CPU division is full of excellent people, but the business environment is not especially favorable. On top of that, AMD does not have its own fabs, so it has to compete with other companies for TSMC capacity.\n\nAs for ARM, I do not think it is really a pure CPU trade at this point. Because of the memory apocalypse, it seems quite obvious that smartphone royalties — still its largest revenue source — are likely to take a meaningful hit.\n\nTherefore, in my view, the better CPU bottleneck trade is foundry companies.\n\nFirst, TSMC. If everyone starts rushing to produce CPUs as well, TSMC wafer capacity will become even tighter, which will further strengthen TSMC’s bargaining power. I think a lot of people are missing this point.\n\nSecond, Intel.\n\nBecause it has its own fabs, even if not immediately, Intel should be in a better position than competitors to expand CPU supply over time. Also, as TSMC capacity becomes increasingly constrained, it is possible that more companies will look to manufacture ASICs and CPUs at Intel Foundry, which is another positive.\n\nI also want to comment on GUC. GUC’s core business is building server CPUs using licensed ARM IP, but I am skeptical about whether GUC can really compete against ARM’s own CPU efforts.\n\n$AMD $INTC $ARM $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08097364506081983, "ret_signed_m1d": -0.08097364506081983, "alpha_spy_m1d": -0.06279016681681648, "alpha_c_m1d": -0.033237081541762636, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.047736563519057196, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00873533268869564, "ret_signed_p1d": 0.00873533268869564, "alpha_spy_p1d": -0.008316550834632341, "alpha_c_p1d": -0.008568512498940084, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.017303845187635725, "ret_p1w": 0.06737986671080742, "ret_signed_p1w": -0.06737986671080742, "alpha_spy_p1w": -0.05073104284551255, "alpha_c_p1w": -0.03723594512505368, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.03014392158575374, "ret_p1m": 0.7068753509644967, "ret_signed_p1m": -0.7068753509644967, "alpha_spy_p1m": -0.6001460996667345, "alpha_c_p1m": -0.3770780415661419, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.32979730939835483, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0581, 0.0568, 0.051, 0.051, 0.0967, 0.0849, 0.0519, 0.0307, 0.0045, -0.0029, 0.0192, 0.0844, 0.0973, 0.1185, 0.1377, 0.1377, 0.1381, 0.2259, 0.2452, 0.2744, 0.2333, 0.2368, 0.2403, 0.2376, 0.1618, 0.2086, 0.1882, -0.0176, -0.0553, 0.0229, 0.06, 0.0481, 0.0481, 0.0106, 0.0174, 0.0174, -0.0034, -0.0179, -0.002, -0.0178, -0.0352, 0.0494, 0.0348, -0.0004, -0.0175, -0.0253, -0.0629, -0.0083, -0.0212, -0.0557, -0.0053, -0.0027, 0.0052, -0.0296, -0.0509, -0.0353, -0.0366, -0.0212, 0.0074, -0.012, -0.0053, 0.0079, 0.081, 0.0, -0.0087, -0.0379, -0.0017, 0.0316, 0.0674, 0.0674, 0.0805, 0.0872, 0.1377, 0.1613, 0.2025, 0.2113, 0.2518, 0.2667, 0.3656, 0.3662, 0.3493, 0.3961, 0.4892, 0.4984, 0.7069, 0.6422, 0.5862, 0.6544, 0.7397, 0.7693, 0.6761, 0.7434, 1.068, 1.0045, 1.2338, 1.2515, 1.2, 1.1863, 1.2069, 1.0813, 1.066, 1.0319, 1.1965, 1.2064, 1.2943, 1.2943, 1.4728, 1.4319, 1.5425, 1.5328, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2036963841582784947", "ticker_idx": 1, "ts": "2026-03-26T00:29:59+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I do not think it is really a pure CPU trade at this point. Because of the memory apocalypse, it seems quite obvious that smartphone royalties — still its largest revenue source — are likely to take a meaningful hit.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Prefers TSMC and Intel as CPU bottleneck plays on tightening wafer capacity; cautious on AMD (no fabs, unfavorable environment) and ARM (smartphone royalty headwinds).", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Yesterday’s CPU bottleneck trade — among AMD, ARM, and Intel, which was actually the best trade?\n\nLet me share my view.\n\nObviously, CPU price increases are positive. They should help earnings materially.\n\nThat said, what we need to understand is that hyperscalers have been moving toward in-house CPU ASICs for quite a long time. They still procure x86 CPUs for external customers, but for internal workloads, many are already relying on their own CPU ASIC solutions.\n\nThat is why I find it hard to recommend AMD outright. AMD’s x86 CPU division is full of excellent people, but the business environment is not especially favorable. On top of that, AMD does not have its own fabs, so it has to compete with other companies for TSMC capacity.\n\nAs for ARM, I do not think it is really a pure CPU trade at this point. Because of the memory apocalypse, it seems quite obvious that smartphone royalties — still its largest revenue source — are likely to take a meaningful hit.\n\nTherefore, in my view, the better CPU bottleneck trade is foundry companies.\n\nFirst, TSMC. If everyone starts rushing to produce CPUs as well, TSMC wafer capacity will become even tighter, which will further strengthen TSMC’s bargaining power. I think a lot of people are missing this point.\n\nSecond, Intel.\n\nBecause it has its own fabs, even if not immediately, Intel should be in a better position than competitors to expand CPU supply over time. Also, as TSMC capacity becomes increasingly constrained, it is possible that more companies will look to manufacture ASICs and CPUs at Intel Foundry, which is another positive.\n\nI also want to comment on GUC. GUC’s core business is building server CPUs using licensed ARM IP, but I am skeptical about whether GUC can really compete against ARM’s own CPU efforts.\n\n$AMD $INTC $ARM $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014664109998124175, "ret_signed_m1d": -0.014664109998124175, "alpha_spy_m1d": 0.003519368245879173, "alpha_c_m1d": 0.03307245352093302, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.047736563519057196, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06892763539143965, "ret_signed_p1d": 0.06892763539143965, "alpha_spy_p1d": 0.051875751868111664, "alpha_c_p1d": 0.05162379020380392, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.017303845187635725, "ret_p1w": -0.03675712098987349, "ret_signed_p1w": 0.03675712098987349, "alpha_spy_p1w": 0.053405944855168364, "alpha_c_p1w": 0.06690104257562723, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.03014392158575374, "ret_p1m": 0.5168604194412345, "ret_signed_p1m": -0.5168604194412345, "alpha_spy_p1m": -0.4101311681434723, "alpha_c_p1m": -0.18706311004287968, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.32979730939835483, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2861, -0.2839, -0.2939, -0.2939, -0.2589, -0.2499, -0.2537, -0.2527, -0.2695, -0.2778, -0.282, -0.3034, -0.3218, -0.321, -0.3167, -0.3167, -0.3077, -0.2641, -0.23, -0.2502, -0.2589, -0.2579, -0.2897, -0.2995, -0.3194, -0.3092, -0.3246, -0.3224, -0.2837, -0.2009, -0.195, -0.1864, -0.1907, -0.2107, -0.1907, -0.1907, -0.1803, -0.178, -0.18, -0.1888, -0.2004, -0.1722, -0.149, -0.165, -0.1767, -0.1966, -0.2137, -0.1983, -0.2208, -0.2611, -0.2401, -0.2213, -0.2242, -0.2563, -0.2523, -0.2138, -0.1776, -0.1708, -0.1614, -0.145, -0.1157, -0.1282, 0.0147, 0.0, -0.0689, -0.1152, -0.0227, 0.0017, -0.0368, -0.0368, -0.039, -0.0707, -0.038, -0.0323, -0.0379, 0.018, 0.0415, 0.0293, 0.0486, 0.0771, 0.1311, 0.1337, 0.2698, 0.3218, 0.5169, 0.3946, 0.2833, 0.3029, 0.3587, 0.3642, 0.313, 0.3491, 0.5329, 0.378, 0.3777, 0.3737, 0.3432, 0.429, 0.4761, 0.3512, 0.3897, 0.4415, 0.6585, 0.9266, 0.98, 0.98, 1.0751, 0.9555, 1.1658, 1.2822, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2036963841582784947", "ticker_idx": 2, "ts": "2026-03-26T00:29:59+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If everyone starts rushing to produce CPUs as well, TSMC wafer capacity will become even tighter, which will further strengthen TSMC's bargaining power.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Prefers TSMC and Intel as CPU bottleneck plays on tightening wafer capacity; cautious on AMD (no fabs, unfavorable environment) and ARM (smartphone royalty headwinds).", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Yesterday’s CPU bottleneck trade — among AMD, ARM, and Intel, which was actually the best trade?\n\nLet me share my view.\n\nObviously, CPU price increases are positive. They should help earnings materially.\n\nThat said, what we need to understand is that hyperscalers have been moving toward in-house CPU ASICs for quite a long time. They still procure x86 CPUs for external customers, but for internal workloads, many are already relying on their own CPU ASIC solutions.\n\nThat is why I find it hard to recommend AMD outright. AMD’s x86 CPU division is full of excellent people, but the business environment is not especially favorable. On top of that, AMD does not have its own fabs, so it has to compete with other companies for TSMC capacity.\n\nAs for ARM, I do not think it is really a pure CPU trade at this point. Because of the memory apocalypse, it seems quite obvious that smartphone royalties — still its largest revenue source — are likely to take a meaningful hit.\n\nTherefore, in my view, the better CPU bottleneck trade is foundry companies.\n\nFirst, TSMC. If everyone starts rushing to produce CPUs as well, TSMC wafer capacity will become even tighter, which will further strengthen TSMC’s bargaining power. I think a lot of people are missing this point.\n\nSecond, Intel.\n\nBecause it has its own fabs, even if not immediately, Intel should be in a better position than competitors to expand CPU supply over time. Also, as TSMC capacity becomes increasingly constrained, it is possible that more companies will look to manufacture ASICs and CPUs at Intel Foundry, which is another positive.\n\nI also want to comment on GUC. GUC’s core business is building server CPUs using licensed ARM IP, but I am skeptical about whether GUC can really compete against ARM’s own CPU efforts.\n\n$AMD $INTC $ARM $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06635802526901613, "ret_signed_m1d": 0.06635802526901613, "alpha_spy_m1d": 0.04817454702501278, "alpha_c_m1d": 0.018621461749958934, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.047736563519057196, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0019318785413249273, "ret_signed_p1d": 0.0019318785413249273, "alpha_spy_p1d": 0.01898376206465291, "alpha_c_p1d": 0.019235723728960652, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.017303845187635725, "ret_p1w": 0.03964927102560245, "ret_signed_p1w": 0.03964927102560245, "alpha_spy_p1w": 0.02300044716030758, "alpha_c_p1w": 0.00950534943984871, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.03014392158575374, "ret_p1m": 0.2341234845084752, "ret_signed_p1m": 0.2341234845084752, "alpha_spy_p1m": 0.12739423321071297, "alpha_c_p1m": -0.09567382488987963, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.32979730939835483, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0798, -0.0839, -0.0708, -0.0708, -0.0227, -0.0146, 0.0012, -0.0255, -0.0276, -0.0104, 0.0145, 0.0128, 0.0002, 0.0447, 0.047, 0.047, 0.0004, -0.0028, 0.001, 0.024, 0.0174, 0.0346, 0.0467, 0.0383, 0.0108, 0.0438, 0.0267, -0.0039, 0.0113, 0.0667, 0.0868, 0.1067, 0.1439, 0.1256, 0.1203, 0.1203, 0.1137, 0.1077, 0.102, 0.133, 0.1315, 0.1796, 0.1856, 0.1522, 0.1454, 0.1287, 0.0798, 0.093, 0.082, 0.0363, 0.0663, 0.0613, 0.0842, 0.0296, 0.0345, 0.0404, 0.0609, 0.0413, 0.0389, 0.0096, 0.0378, 0.0526, 0.0664, 0.0, 0.0019, -0.0295, 0.0363, 0.0472, 0.0396, 0.0396, 0.048, 0.0589, 0.122, 0.1208, 0.1364, 0.1333, 0.1649, 0.1502, 0.1142, 0.1361, 0.1231, 0.1287, 0.1881, 0.1734, 0.2341, 0.2419, 0.2031, 0.2077, 0.2145, 0.2194, 0.2315, 0.2094, 0.2864, 0.27, 0.2624, 0.2405, 0.2182, 0.226, 0.2809, 0.2399, 0.2142, 0.2039, 0.2315, 0.2485, 0.2404, 0.2404, 0.2644, 0.2963, 0.3028, 0.2832, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2036963841582784947", "ticker_idx": 3, "ts": "2026-03-26T00:29:59+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel should be in a better position than competitors to expand CPU supply over time. Also, as TSMC capacity becomes increasingly constrained, it is possible that more companies will look to manufacture ASICs and CPUs at Intel Foundry, which is another positive.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Prefers TSMC and Intel as CPU bottleneck plays on tightening wafer capacity; cautious on AMD (no fabs, unfavorable environment) and ARM (smartphone royalty headwinds).", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Yesterday’s CPU bottleneck trade — among AMD, ARM, and Intel, which was actually the best trade?\n\nLet me share my view.\n\nObviously, CPU price increases are positive. They should help earnings materially.\n\nThat said, what we need to understand is that hyperscalers have been moving toward in-house CPU ASICs for quite a long time. They still procure x86 CPUs for external customers, but for internal workloads, many are already relying on their own CPU ASIC solutions.\n\nThat is why I find it hard to recommend AMD outright. AMD’s x86 CPU division is full of excellent people, but the business environment is not especially favorable. On top of that, AMD does not have its own fabs, so it has to compete with other companies for TSMC capacity.\n\nAs for ARM, I do not think it is really a pure CPU trade at this point. Because of the memory apocalypse, it seems quite obvious that smartphone royalties — still its largest revenue source — are likely to take a meaningful hit.\n\nTherefore, in my view, the better CPU bottleneck trade is foundry companies.\n\nFirst, TSMC. If everyone starts rushing to produce CPUs as well, TSMC wafer capacity will become even tighter, which will further strengthen TSMC’s bargaining power. I think a lot of people are missing this point.\n\nSecond, Intel.\n\nBecause it has its own fabs, even if not immediately, Intel should be in a better position than competitors to expand CPU supply over time. Also, as TSMC capacity becomes increasingly constrained, it is possible that more companies will look to manufacture ASICs and CPUs at Intel Foundry, which is another positive.\n\nI also want to comment on GUC. GUC’s core business is building server CPUs using licensed ARM IP, but I am skeptical about whether GUC can really compete against ARM’s own CPU efforts.\n\n$AMD $INTC $ARM $TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06984131377832359, "ret_signed_m1d": 0.06984131377832359, "alpha_spy_m1d": 0.05165783553432024, "alpha_c_m1d": 0.022104750259266392, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.047736563519057196, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02199540679292944, "ret_signed_p1d": -0.02199540679292944, "alpha_spy_p1d": -0.0049435232696014575, "alpha_c_p1d": -0.004691561605293715, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.017303845187635725, "ret_p1w": 0.14240369186587154, "ret_signed_p1w": 0.14240369186587154, "alpha_spy_p1w": 0.12575486800057667, "alpha_c_p1w": 0.1122597702801178, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.03014392158575374, "ret_p1m": 0.8716554143185229, "ret_signed_p1m": 0.8716554143185229, "alpha_spy_p1m": 0.7649261630207607, "alpha_c_p1m": 0.5418581049201681, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.32979730939835483, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1683, -0.1542, -0.1633, -0.1633, -0.107, -0.1073, -0.0921, -0.0333, -0.0678, 0.0329, -0.0009, 0.0723, 0.1048, 0.0957, 0.0649, 0.0649, 0.1011, 0.2302, 0.2317, 0.022, -0.0365, -0.0039, 0.1061, 0.1034, 0.0537, 0.1068, 0.1168, 0.102, 0.0939, 0.1472, 0.1392, 0.0687, 0.095, 0.054, 0.061, 0.061, 0.0472, 0.0308, 0.0118, 0.0002, -0.0107, 0.0458, 0.063, 0.0308, 0.0342, 0.0317, -0.0227, 0.0336, 0.042, -0.0154, 0.0336, 0.0608, 0.088, 0.0261, 0.0379, 0.0376, -0.0009, 0.0211, 0.0472, -0.0052, -0.002, -0.0009, 0.0698, 0.0, -0.022, -0.066, 0.0007, 0.0891, 0.1424, 0.1424, 0.1515, 0.1998, 0.3367, 0.3995, 0.4145, 0.478, 0.4469, 0.4726, 0.5533, 0.5533, 0.4898, 0.5025, 0.48, 0.5143, 0.8717, 0.9272, 0.9166, 1.1485, 1.1424, 1.259, 1.1719, 1.4524, 1.5626, 1.4857, 1.8327, 1.9351, 1.7349, 1.7277, 1.6288, 1.4664, 1.4528, 1.5125, 1.6975, 1.6871, 1.7175, 1.7175, 1.8009, 1.7612, 1.7413, 1.6005, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013386435928433102", "ticker_idx": 0, "ts": "2026-01-19T23:01:47+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Analysts say this will be a major positive factor for improving the operating margins of Samsung Electronics and SK Hynix, on par with DRAM.", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung and SK Hynix cutting NAND production into strong AI-driven demand will drive price increases and margin expansion; $SNDK also flagged as a beneficiary.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung and SK Hynix to Cut, Rather Than Increase, NAND Flash Production This Year… “Maximizing Profits in Super Boom Period”\n\nSamsung Electronics and SK Hynix, which together hold more than 60% of the global NAND flash market, are expected to reduce their NAND flash production this year. As competition in inference-type artificial intelligence (AI), led by Nvidia, intensifies, tight supply of NAND flash—a key component—is likely to drive price increases across servers, PCs, mobiles, and other sectors. Analysts say this will be a major positive factor for improving the operating margins of Samsung Electronics and SK Hynix, on par with DRAM.\n\nAccording to data from market research firm Omdia obtained by Chosun Biz on the 20th, Samsung Electronics has slightly lowered its NAND wafer production from 4.9 million wafers last year to 4.68 million this year. This is a further reduction from the production cuts implemented last year due to a sharp drop in NAND profitability in 2024. SK Hynix is also projected to follow a similar path, reducing NAND production from around 1.9 million wafers last year to 1.7 million this year.\n\nGiven that the NAND market this year is seeing strong AI-driven demand, supply adjustments by the leading suppliers—Samsung Electronics and SK Hynix—are highly likely to exacerbate shortages not only in AI servers but also in mobiles and PCs. According to Citigroup, Nvidia’s next-generation AI accelerator “Vera Rubin,” set for mass production in the second half of this year, will feature solid-state drives (SSDs) with a capacity of 1,152 terabytes (TB)—more than 10 times that of the current “Blackwell” product. With Vera Rubin shipments projected at 30,000 units this year and 100,000 units next year, this translates to new demand of 34.6 million TB in 2026 and 115.2 million TB in 2027.\n\nThe reduction in NAND production by Samsung Electronics and SK Hynix is largely attributed to lower investment priority for NAND facilities compared to DRAM, which currently shows the highest profitability. Additionally, as demand grows for high-capacity SSDs in AI data centers, the transition to quad-level cell (QLC) technology inevitably leads to natural production losses. Switching from existing triple-level cell (TLC) technology to QLC, which is better suited for AI data centers, involves various factors such as facility setup, stabilization periods, and initial yield rates.\n\nManagement at Samsung Electronics and SK Hynix reportedly see no reason to aggressively increase NAND production. For years, NAND has suffered from poor profitability, requiring both companies to focus on price defense. Now, in this memory super-boom cycle, they aim to maximize profits as much as possible. An industry insider commented, “Whether the NAND production cuts by Samsung Electronics and SK Hynix are intentional or natural, the benefits from reduced supply will be greatest this year.”\n\nSome analysts suggest the moves are also a response to growing supply of commodity NAND from China. Unlike Samsung and SK Hynix, China’s YMTC has been increasing its presence and output in the NAND market since last year. This is interpreted as a strategy to counter low-price competition from China by reducing supply in mobile and PC segments to protect profitability while increasing higher-margin server and enterprise volumes to manage the product mix.\n\nMajor market research firms are forecasting a strong NAND price uptrend starting in the first quarter and closely monitoring production discipline by key suppliers. TrendForce predicts NAND flash contract prices will rise 33–38% quarter-on-quarter in Q1, noting that Samsung Electronics and SK Hynix are maintaining a conservative production stance. IDC also expects NAND supply growth this year to be around 17%—lower than the average of recent years.\n\n$SNDK\n\n---\n\nhttps://t.co/iSiEmOtSAU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0026791567436211805, "ret_signed_m1d": -0.0026791567436211805, "alpha_spy_m1d": -0.0026791567436211805, "alpha_c_m1d": -0.0026791567436211805, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02746148771171364, "ret_signed_p1d": -0.02746148771171364, "alpha_spy_p1d": -0.0071047019881685936, "alpha_c_p1d": -0.001503561462741887, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.025957926248971752, "ret_p1w": 0.018754202077025495, "ret_signed_p1w": 0.018754202077025495, "alpha_spy_p1w": 0.017207210949550023, "alpha_c_p1w": 0.015595203300232674, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.0031589987767928207, "ret_p1m": 0.21366377280264204, "ret_signed_p1m": 0.21366377280264204, "alpha_spy_p1m": 0.22640125337234318, "alpha_c_p1m": 0.25582827553940934, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": -0.042164502736767306, "ret_p3m": 0.45981374999829594, "ret_signed_p3m": 0.45981374999829594, "alpha_spy_p3m": 0.44258527888441535, "alpha_c_p3m": 0.41456568256379644, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.045248067434499495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3428, -0.3568, -0.3414, -0.3201, -0.3368, -0.3301, -0.3061, -0.2835, -0.2595, -0.3008, -0.3294, -0.3388, -0.3474, -0.3294, -0.3101, -0.3128, -0.3148, -0.3521, -0.3294, -0.3481, -0.3568, -0.3294, -0.3681, -0.3554, -0.3381, -0.3148, -0.3101, -0.3301, -0.3281, -0.3108, -0.3035, -0.2995, -0.2775, -0.2701, -0.2775, -0.2801, -0.2848, -0.2741, -0.3015, -0.3148, -0.2808, -0.2828, -0.2915, -0.2635, -0.2568, -0.2595, -0.2595, -0.2201, -0.1996, -0.1969, -0.1969, -0.1969, -0.1393, -0.075, -0.0697, -0.0556, -0.0703, -0.069, -0.0703, -0.0784, -0.0603, -0.0362, -0.0027, 0.0, -0.0275, 0.0013, 0.0201, 0.0188, 0.0188, 0.0683, 0.0877, 0.0764, 0.075, 0.0074, 0.1219, 0.1326, 0.067, 0.0623, 0.1145, 0.1105, 0.1239, 0.1962, 0.2137, 0.2137, 0.2137, 0.2137, 0.2726, 0.2733, 0.2927, 0.3396, 0.363, 0.4601, 0.4501, 0.4501, 0.3068, 0.1534, 0.2833, 0.2605, 0.1621, 0.2585, 0.2726, 0.2585, 0.2291, 0.2639, 0.2987, 0.3965, 0.3429, 0.3356, 0.2478, 0.2706, 0.2659, 0.2063, 0.2063, 0.1833, 0.1222, 0.2729, 0.1974, 0.2497, 0.296, 0.3189, 0.4128, 0.3692, 0.3826, 0.3491, 0.386, 0.4162, 0.4598, 0.4497, 0.4397, 0.4699, 0.4598, 0.5068, 0.4732, 0.5068, 0.49, 0.5169, 0.4799, 0.4799, 0.5605, 0.5605, 0.7853, 0.8223, 0.8021, 0.9162, 0.8726, 0.9061, 0.9867, 0.8155, 0.886, 0.8491, 0.8525, 1.0102, 0.9632, 0.9632, 1.0068, 1.0605, 1.0102, 1.1276, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013386435928433102", "ticker_idx": 1, "ts": "2026-01-19T23:01:47+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Analysts say this will be a major positive factor for improving the operating margins of Samsung Electronics and SK Hynix, on par with DRAM.", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung and SK Hynix cutting NAND production into strong AI-driven demand will drive price increases and margin expansion; $SNDK also flagged as a beneficiary.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung and SK Hynix to Cut, Rather Than Increase, NAND Flash Production This Year… “Maximizing Profits in Super Boom Period”\n\nSamsung Electronics and SK Hynix, which together hold more than 60% of the global NAND flash market, are expected to reduce their NAND flash production this year. As competition in inference-type artificial intelligence (AI), led by Nvidia, intensifies, tight supply of NAND flash—a key component—is likely to drive price increases across servers, PCs, mobiles, and other sectors. Analysts say this will be a major positive factor for improving the operating margins of Samsung Electronics and SK Hynix, on par with DRAM.\n\nAccording to data from market research firm Omdia obtained by Chosun Biz on the 20th, Samsung Electronics has slightly lowered its NAND wafer production from 4.9 million wafers last year to 4.68 million this year. This is a further reduction from the production cuts implemented last year due to a sharp drop in NAND profitability in 2024. SK Hynix is also projected to follow a similar path, reducing NAND production from around 1.9 million wafers last year to 1.7 million this year.\n\nGiven that the NAND market this year is seeing strong AI-driven demand, supply adjustments by the leading suppliers—Samsung Electronics and SK Hynix—are highly likely to exacerbate shortages not only in AI servers but also in mobiles and PCs. According to Citigroup, Nvidia’s next-generation AI accelerator “Vera Rubin,” set for mass production in the second half of this year, will feature solid-state drives (SSDs) with a capacity of 1,152 terabytes (TB)—more than 10 times that of the current “Blackwell” product. With Vera Rubin shipments projected at 30,000 units this year and 100,000 units next year, this translates to new demand of 34.6 million TB in 2026 and 115.2 million TB in 2027.\n\nThe reduction in NAND production by Samsung Electronics and SK Hynix is largely attributed to lower investment priority for NAND facilities compared to DRAM, which currently shows the highest profitability. Additionally, as demand grows for high-capacity SSDs in AI data centers, the transition to quad-level cell (QLC) technology inevitably leads to natural production losses. Switching from existing triple-level cell (TLC) technology to QLC, which is better suited for AI data centers, involves various factors such as facility setup, stabilization periods, and initial yield rates.\n\nManagement at Samsung Electronics and SK Hynix reportedly see no reason to aggressively increase NAND production. For years, NAND has suffered from poor profitability, requiring both companies to focus on price defense. Now, in this memory super-boom cycle, they aim to maximize profits as much as possible. An industry insider commented, “Whether the NAND production cuts by Samsung Electronics and SK Hynix are intentional or natural, the benefits from reduced supply will be greatest this year.”\n\nSome analysts suggest the moves are also a response to growing supply of commodity NAND from China. Unlike Samsung and SK Hynix, China’s YMTC has been increasing its presence and output in the NAND market since last year. This is interpreted as a strategy to counter low-price competition from China by reducing supply in mobile and PC segments to protect profitability while increasing higher-margin server and enterprise volumes to manage the product mix.\n\nMajor market research firms are forecasting a strong NAND price uptrend starting in the first quarter and closely monitoring production discipline by key suppliers. TrendForce predicts NAND flash contract prices will rise 33–38% quarter-on-quarter in Q1, noting that Samsung Electronics and SK Hynix are maintaining a conservative production stance. IDC also expects NAND supply growth this year to be around 17%—lower than the average of recent years.\n\n$SNDK\n\n---\n\nhttps://t.co/iSiEmOtSAU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01047121405932816, "ret_signed_m1d": -0.01047121405932816, "alpha_spy_m1d": -0.01047121405932816, "alpha_c_m1d": -0.01047121405932816, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.027486906166632963, "ret_signed_p1d": -0.027486906166632963, "alpha_spy_p1d": -0.007130120443087917, "alpha_c_p1d": -0.0025112583379015563, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": -0.03664920822217732, "ret_signed_p1w": -0.03664920822217732, "alpha_spy_p1w": -0.03819619934965279, "alpha_c_p1w": -0.03272801322035568, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": 0.1518324809038445, "ret_signed_p1m": 0.1518324809038445, "alpha_spy_p1m": 0.16456996147354563, "alpha_c_p1m": 0.13404983184492814, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.5145697201798933, "ret_signed_p3m": 0.5145697201798933, "alpha_spy_p3m": 0.49734124906601274, "alpha_c_p3m": 0.37867728741703854, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3702, -0.3741, -0.3329, -0.3002, -0.3186, -0.2702, -0.2571, -0.2688, -0.1891, -0.2335, -0.2427, -0.2244, -0.2414, -0.2074, -0.1904, -0.193, -0.1995, -0.2675, -0.2074, -0.2545, -0.2649, -0.2532, -0.3186, -0.3199, -0.3212, -0.3146, -0.288, -0.3063, -0.2958, -0.2696, -0.2775, -0.2906, -0.288, -0.2448, -0.2592, -0.2317, -0.2605, -0.2526, -0.2749, -0.3063, -0.2788, -0.2775, -0.284, -0.2408, -0.2356, -0.2304, -0.2304, -0.216, -0.1623, -0.1479, -0.1479, -0.1479, -0.1139, -0.089, -0.0497, -0.0288, -0.0105, -0.0262, -0.0196, -0.034, -0.0288, -0.0196, -0.0105, 0.0, -0.0275, -0.0314, -0.0118, 0.0039, -0.0366, 0.0471, 0.1008, 0.127, 0.1898, 0.0864, 0.1872, 0.178, 0.1021, 0.0982, 0.161, 0.1466, 0.1257, 0.1623, 0.1518, 0.1518, 0.1518, 0.1518, 0.1702, 0.2421, 0.2448, 0.3154, 0.3325, 0.4411, 0.3913, 0.3913, 0.2313, 0.1133, 0.2339, 0.2117, 0.0963, 0.23, 0.2523, 0.2195, 0.1933, 0.2772, 0.272, 0.3847, 0.3284, 0.3205, 0.2235, 0.293, 0.3048, 0.2235, 0.2235, 0.1448, 0.0582, 0.1763, 0.0884, 0.1487, 0.1618, 0.2012, 0.3546, 0.3087, 0.3467, 0.3638, 0.4464, 0.4897, 0.5146, 0.4792, 0.529, 0.6051, 0.6037, 0.6064, 0.6024, 0.6942, 0.7047, 0.6955, 0.6864, 0.6864, 0.8975, 0.8975, 1.0994, 1.1689, 1.2109, 1.4653, 1.4063, 1.5912, 1.5833, 1.3853, 1.4128, 1.2882, 1.2882, 1.544, 1.5453, 1.5453, 1.6908, 1.9413, 2.0021, 2.0598, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013386435928433102", "ticker_idx": 2, "ts": "2026-01-19T23:01:47+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$SNDK", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung and SK Hynix cutting NAND production into strong AI-driven demand will drive price increases and margin expansion; $SNDK also flagged as a beneficiary.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung and SK Hynix to Cut, Rather Than Increase, NAND Flash Production This Year… “Maximizing Profits in Super Boom Period”\n\nSamsung Electronics and SK Hynix, which together hold more than 60% of the global NAND flash market, are expected to reduce their NAND flash production this year. As competition in inference-type artificial intelligence (AI), led by Nvidia, intensifies, tight supply of NAND flash—a key component—is likely to drive price increases across servers, PCs, mobiles, and other sectors. Analysts say this will be a major positive factor for improving the operating margins of Samsung Electronics and SK Hynix, on par with DRAM.\n\nAccording to data from market research firm Omdia obtained by Chosun Biz on the 20th, Samsung Electronics has slightly lowered its NAND wafer production from 4.9 million wafers last year to 4.68 million this year. This is a further reduction from the production cuts implemented last year due to a sharp drop in NAND profitability in 2024. SK Hynix is also projected to follow a similar path, reducing NAND production from around 1.9 million wafers last year to 1.7 million this year.\n\nGiven that the NAND market this year is seeing strong AI-driven demand, supply adjustments by the leading suppliers—Samsung Electronics and SK Hynix—are highly likely to exacerbate shortages not only in AI servers but also in mobiles and PCs. According to Citigroup, Nvidia’s next-generation AI accelerator “Vera Rubin,” set for mass production in the second half of this year, will feature solid-state drives (SSDs) with a capacity of 1,152 terabytes (TB)—more than 10 times that of the current “Blackwell” product. With Vera Rubin shipments projected at 30,000 units this year and 100,000 units next year, this translates to new demand of 34.6 million TB in 2026 and 115.2 million TB in 2027.\n\nThe reduction in NAND production by Samsung Electronics and SK Hynix is largely attributed to lower investment priority for NAND facilities compared to DRAM, which currently shows the highest profitability. Additionally, as demand grows for high-capacity SSDs in AI data centers, the transition to quad-level cell (QLC) technology inevitably leads to natural production losses. Switching from existing triple-level cell (TLC) technology to QLC, which is better suited for AI data centers, involves various factors such as facility setup, stabilization periods, and initial yield rates.\n\nManagement at Samsung Electronics and SK Hynix reportedly see no reason to aggressively increase NAND production. For years, NAND has suffered from poor profitability, requiring both companies to focus on price defense. Now, in this memory super-boom cycle, they aim to maximize profits as much as possible. An industry insider commented, “Whether the NAND production cuts by Samsung Electronics and SK Hynix are intentional or natural, the benefits from reduced supply will be greatest this year.”\n\nSome analysts suggest the moves are also a response to growing supply of commodity NAND from China. Unlike Samsung and SK Hynix, China’s YMTC has been increasing its presence and output in the NAND market since last year. This is interpreted as a strategy to counter low-price competition from China by reducing supply in mobile and PC segments to protect profitability while increasing higher-margin server and enterprise volumes to manage the product mix.\n\nMajor market research firms are forecasting a strong NAND price uptrend starting in the first quarter and closely monitoring production discipline by key suppliers. TrendForce predicts NAND flash contract prices will rise 33–38% quarter-on-quarter in Q1, noting that Samsung Electronics and SK Hynix are maintaining a conservative production stance. IDC also expects NAND supply growth this year to be around 17%—lower than the average of recent years.\n\n$SNDK\n\n---\n\nhttps://t.co/iSiEmOtSAU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09549828457593978, "ret_signed_p1d": 0.09549828457593978, "alpha_spy_p1d": 0.11585507029948483, "alpha_c_p1d": 0.12145621082491154, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.025957926248971752, "ret_p1w": 0.13824281550890904, "ret_signed_p1w": 0.13824281550890904, "alpha_spy_p1w": 0.13669582438143357, "alpha_c_p1w": 0.13508381673211622, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.0031589987767928207, "ret_p1m": 0.4278565684140627, "ret_signed_p1m": 0.4278565684140627, "alpha_spy_p1m": 0.44059404898376386, "alpha_c_p1m": 0.47002107115083, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": -0.042164502736767306, "ret_p3m": 1.222982403069318, "ret_signed_p3m": 1.222982403069318, "alpha_spy_p3m": 1.2057539319554373, "alpha_c_p3m": 1.1777343356348184, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.045248067434499495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6447, -0.5961, -0.5499, -0.5733, -0.5756, -0.5059, -0.5266, -0.5181, -0.4995, -0.5296, -0.4766, -0.4979, -0.421, -0.3522, -0.3434, -0.3156, -0.4111, -0.3855, -0.3572, -0.4078, -0.4053, -0.5262, -0.5158, -0.4513, -0.4669, -0.4801, -0.4801, -0.4602, -0.4919, -0.5035, -0.5301, -0.4843, -0.4476, -0.4549, -0.4694, -0.437, -0.4159, -0.5015, -0.5119, -0.494, -0.5, -0.4694, -0.4255, -0.4172, -0.4079, -0.3954, -0.3954, -0.3955, -0.4095, -0.4192, -0.4261, -0.4261, -0.3346, -0.3374, -0.1547, -0.1452, -0.1912, -0.0875, -0.0589, -0.0576, -0.0624, -0.0106, 0.0, 0.0, 0.0955, 0.212, 0.2172, 0.1456, 0.1382, 0.1639, 0.2756, 0.3039, 0.3932, 0.6083, 0.6815, 0.4133, 0.3931, 0.4457, 0.4105, 0.3095, 0.449, 0.5238, 0.5148, 0.5148, 0.4279, 0.4516, 0.5016, 0.5714, 0.6114, 0.5437, 0.5289, 0.5761, 0.5361, 0.4967, 0.367, 0.4483, 0.3674, 0.2749, 0.4234, 0.4963, 0.5846, 0.4961, 0.5996, 0.7012, 0.7411, 0.8222, 0.8667, 0.7159, 0.6984, 0.6984, 0.6388, 0.4583, 0.4889, 0.3841, 0.536, 0.6748, 0.6962, 0.6962, 0.7519, 0.7185, 0.888, 1.0588, 1.0593, 1.3028, 1.2834, 1.1559, 1.223, 1.2267, 1.2074, 1.1843, 1.3671, 1.2543, 1.3933, 1.5874, 1.4234, 1.5729, 1.651, 1.8698, 2.0363, 2.4, 2.4089, 2.2396, 2.7772, 2.7415, 2.5105, 2.4989, 2.343, 2.4031, 2.2228, 2.3444, 2.3668, 2.7286, 2.575, 2.575, 2.843, 2.844, 2.969, 3.0979, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022085214626074657", "ticker_idx": 0, "ts": "2026-02-12T23:07:38+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our Buy recommendations on Samsung Electronics and SK Hynix", "tweet_type": "analysis", "timeline": "quarters", "summary": "Maintains Buy on Samsung & SK Hynix; bullish NAND theme benefits SNDK and MU on surging ASPs (mid-to-high 40% QoQ) and tightening supply-demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: There Won't Be a Second Shock\n[Daishin Securities Semiconductor / Ryu Hyung-geun]\n\nThis is a Kioxia FY3Q25 (CY4Q25) Review. We maintain our existing view that a surprise will emerge from NAND. This value needs to be further reflected in Korean semiconductor stock prices. We maintain our Buy recommendations on Samsung Electronics and SK Hynix.\n\nKioxia FY3Q25 (CY4Q25) Review: Please Forget Last November's Disaster\n\n- Results were in line with consensus. The guidance was the surprise. The company issued guidance significantly above consensus, creating a markedly different atmosphere from last November's earnings release.\n\n- Throughout the recent upcycle, Kioxia's high revenue concentration toward a single customer has been a double-edged sword. Pre-negotiated long-term contracts meant products were supplied at prices below market rates, which was the key reason behind Kioxia's lower ASP growth relative to peers.\n\n- Recently, the company appears to have aggressively leveraged the tight supply-demand environment to push through price revisions. Starting from CY1Q26, the revised pricing policy is expected to take effect (estimated ~50% QoQ ASP increase for the North American customer in CY1Q26). This marks the moment Kioxia breaks free from its legacy constraints and the moment mobile NAND ASPs normalize.\n\n- Opportunities are also expanding on the server side. On top of a favorable market backdrop, competitiveness in high-capacity products of 128TB and above is expected to drive outperformance.\n\nNAND Is Just as Strong as DRAM\n\n- AI inference is strengthening the role of storage and driving higher content per unit. In addition, new products are set to be unveiled. At the upcoming GTC 2026 event, new products based on a next-generation solution (Gen 6 Controller) are expected to be announced.\n\n- The supply-demand balance is tightening to historically unprecedented levels. While concerns over weak set-level demand persist, opportunities in the server segment are more than offsetting them.\n\n- Meanwhile, supply discipline is strengthening across the industry. The prevailing approach is to respond through conversion investments and product mix shifts—which carry lower depreciation burdens—rather than new capacity additions. This signals a management philosophy increasingly centered on profitability and long-term supply-demand stability.\n\n- As a result, NAND ASPs are surging. 1Q26 NAND ASP is expected to rise in the mid-to-high 40% range QoQ, and we believe we will soon see operating margins reach the 50%+ levels that marked the previous peak.\n\n$SNDK $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06047029883078414, "ret_signed_m1d": -0.06047029883078414, "alpha_spy_m1d": -0.07616158577719445, "alpha_c_m1d": -0.08747981096749968, "bench_spy_m1d": 0.01569128694641031, "bench_c_m1d": 0.02700951213671554, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014557645051763046, "ret_signed_p1d": 0.014557645051763046, "alpha_spy_p1d": 0.013853159557947636, "alpha_c_p1d": 0.012043580535520793, "bench_spy_p1d": 0.0007044854938154099, "bench_c_p1d": 0.0025140645162422537, "ret_p1w": 0.0638297890992996, "ret_signed_p1w": 0.0638297890992996, "alpha_spy_p1w": 0.05911812687287399, "alpha_c_p1w": 0.056646465419968095, "bench_spy_p1w": 0.004711662226425606, "bench_c_p1w": 0.007183323679331499, "ret_p1m": 0.027435588802809496, "ret_signed_p1m": 0.027435588802809496, "alpha_spy_p1m": 0.055295375886846854, "alpha_c_p1m": 0.044747602997813885, "bench_spy_p1m": -0.027859787084037357, "bench_c_p1m": -0.017312014195004388, "ret_p3m": 0.56538342513538, "ret_signed_p3m": 0.56538342513538, "alpha_spy_p3m": 0.4788891900222849, "alpha_c_p3m": 0.30528848155696675, "bench_spy_p3m": 0.08649423511309506, "bench_c_p3m": 0.26009494357841323, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4395, -0.4551, -0.4623, -0.4395, -0.4718, -0.4612, -0.4467, -0.4272, -0.4233, -0.44, -0.4383, -0.4239, -0.4177, -0.4144, -0.396, -0.3899, -0.396, -0.3982, -0.4021, -0.3932, -0.4161, -0.4272, -0.3988, -0.4005, -0.4077, -0.3843, -0.3787, -0.3809, -0.3809, -0.3481, -0.3309, -0.3287, -0.3287, -0.3287, -0.2805, -0.2268, -0.2223, -0.2105, -0.2228, -0.2217, -0.2228, -0.2296, -0.2144, -0.1943, -0.1663, -0.1641, -0.187, -0.1629, -0.1473, -0.1484, -0.1484, -0.1069, -0.0907, -0.1002, -0.1013, -0.1579, -0.0622, -0.0532, -0.1081, -0.112, -0.0683, -0.0717, -0.0605, 0.0, 0.0146, 0.0146, 0.0146, 0.0146, 0.0638, 0.0644, 0.0806, 0.1198, 0.1394, 0.2206, 0.2122, 0.2122, 0.0924, -0.0358, 0.0728, 0.0538, -0.0286, 0.0521, 0.0638, 0.0521, 0.0274, 0.0566, 0.0857, 0.1674, 0.1226, 0.1165, 0.0431, 0.0622, 0.0582, 0.0084, 0.0084, -0.0108, -0.0619, 0.0641, 0.0009, 0.0447, 0.0834, 0.1025, 0.1811, 0.1446, 0.1558, 0.1277, 0.1586, 0.1839, 0.2203, 0.2119, 0.2035, 0.2287, 0.2203, 0.2596, 0.2315, 0.2596, 0.2456, 0.268, 0.2372, 0.2372, 0.3045, 0.3045, 0.4924, 0.5233, 0.5065, 0.6019, 0.5654, 0.5934, 0.6608, 0.5177, 0.5766, 0.5457, 0.5486, 0.6804, 0.6411, 0.6411, 0.6776, 0.7225, 0.6804, 0.7786, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022085214626074657", "ticker_idx": 1, "ts": "2026-02-12T23:07:38+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We maintain our Buy recommendations on Samsung Electronics and SK Hynix", "tweet_type": "analysis", "timeline": "quarters", "summary": "Maintains Buy on Samsung & SK Hynix; bullish NAND theme benefits SNDK and MU on surging ASPs (mid-to-high 40% QoQ) and tightening supply-demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: There Won't Be a Second Shock\n[Daishin Securities Semiconductor / Ryu Hyung-geun]\n\nThis is a Kioxia FY3Q25 (CY4Q25) Review. We maintain our existing view that a surprise will emerge from NAND. This value needs to be further reflected in Korean semiconductor stock prices. We maintain our Buy recommendations on Samsung Electronics and SK Hynix.\n\nKioxia FY3Q25 (CY4Q25) Review: Please Forget Last November's Disaster\n\n- Results were in line with consensus. The guidance was the surprise. The company issued guidance significantly above consensus, creating a markedly different atmosphere from last November's earnings release.\n\n- Throughout the recent upcycle, Kioxia's high revenue concentration toward a single customer has been a double-edged sword. Pre-negotiated long-term contracts meant products were supplied at prices below market rates, which was the key reason behind Kioxia's lower ASP growth relative to peers.\n\n- Recently, the company appears to have aggressively leveraged the tight supply-demand environment to push through price revisions. Starting from CY1Q26, the revised pricing policy is expected to take effect (estimated ~50% QoQ ASP increase for the North American customer in CY1Q26). This marks the moment Kioxia breaks free from its legacy constraints and the moment mobile NAND ASPs normalize.\n\n- Opportunities are also expanding on the server side. On top of a favorable market backdrop, competitiveness in high-capacity products of 128TB and above is expected to drive outperformance.\n\nNAND Is Just as Strong as DRAM\n\n- AI inference is strengthening the role of storage and driving higher content per unit. In addition, new products are set to be unveiled. At the upcoming GTC 2026 event, new products based on a next-generation solution (Gen 6 Controller) are expected to be announced.\n\n- The supply-demand balance is tightening to historically unprecedented levels. While concerns over weak set-level demand persist, opportunities in the server segment are more than offsetting them.\n\n- Meanwhile, supply discipline is strengthening across the industry. The prevailing approach is to respond through conversion investments and product mix shifts—which carry lower depreciation burdens—rather than new capacity additions. This signals a management philosophy increasingly centered on profitability and long-term supply-demand stability.\n\n- As a result, NAND ASPs are surging. 1Q26 NAND ASP is expected to rise in the mid-to-high 40% range QoQ, and we believe we will soon see operating margins reach the 50%+ levels that marked the previous peak.\n\n$SNDK $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.031531525177968645, "ret_signed_m1d": -0.031531525177968645, "alpha_spy_m1d": -0.047222812124378954, "alpha_c_m1d": -0.05288045463171964, "bench_spy_m1d": 0.01569128694641031, "bench_c_m1d": 0.021348929453750998, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009009017268640673, "ret_signed_p1d": -0.009009017268640673, "alpha_spy_p1d": -0.009713502762456083, "alpha_c_p1d": -0.012973499618777473, "bench_spy_p1d": 0.0007044854938154099, "bench_c_p1d": 0.0039644823501368, "ret_p1w": 0.006756745320343649, "ret_signed_p1w": 0.006756745320343649, "alpha_spy_p1w": 0.002045083093918043, "alpha_c_p1w": -0.0033144841568830685, "bench_spy_p1w": 0.004711662226425606, "bench_c_p1w": 0.010071229477226717, "ret_p1m": 0.02666575454918596, "ret_signed_p1m": 0.02666575454918596, "alpha_spy_p1m": 0.05452554163322332, "alpha_c_p1m": 0.07290938178489492, "bench_spy_p1m": -0.027859787084037357, "bench_c_p1m": -0.04624362723570896, "ret_p3m": 1.0702545795114906, "ret_signed_p3m": 1.0702545795114906, "alpha_spy_p3m": 0.9837603443983955, "alpha_c_p3m": 0.6882398045875611, "bench_spy_p3m": 0.08649423511309506, "bench_c_p3m": 0.38201477492392955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3181, -0.3586, -0.3676, -0.3574, -0.4137, -0.4148, -0.416, -0.4103, -0.3874, -0.4032, -0.3941, -0.3716, -0.3784, -0.3896, -0.3874, -0.3502, -0.3626, -0.339, -0.3637, -0.357, -0.3761, -0.4032, -0.3795, -0.3784, -0.384, -0.3468, -0.3423, -0.3378, -0.3378, -0.3255, -0.2793, -0.2669, -0.2669, -0.2669, -0.2376, -0.2162, -0.1824, -0.1644, -0.1486, -0.1622, -0.1565, -0.1689, -0.1644, -0.1565, -0.1486, -0.1396, -0.1633, -0.1667, -0.1498, -0.1363, -0.1712, -0.0991, -0.0529, -0.0304, 0.0236, -0.0653, 0.0214, 0.0135, -0.0518, -0.0552, -0.0011, -0.0135, -0.0315, 0.0, -0.009, -0.009, -0.009, -0.009, 0.0068, 0.0687, 0.0709, 0.1318, 0.1464, 0.2399, 0.197, 0.197, 0.0594, -0.0422, 0.0616, 0.0425, -0.0568, 0.0583, 0.0774, 0.0492, 0.0267, 0.0989, 0.0944, 0.1914, 0.1429, 0.1361, 0.0526, 0.1124, 0.1226, 0.0526, 0.0526, -0.0151, -0.0895, 0.012, -0.0636, -0.0117, -0.0004, 0.0334, 0.1654, 0.1259, 0.1587, 0.1733, 0.2444, 0.2816, 0.3031, 0.2726, 0.3155, 0.3809, 0.3798, 0.3821, 0.3787, 0.4576, 0.4667, 0.4588, 0.4509, 0.4509, 0.6325, 0.6325, 0.8063, 0.866, 0.9022, 1.121, 1.0703, 1.2293, 1.2226, 1.0522, 1.0759, 0.9687, 0.9687, 1.1887, 1.1898, 1.1898, 1.3151, 1.5306, 1.5829, 1.6325, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022085214626074657", "ticker_idx": 2, "ts": "2026-02-12T23:07:38+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND ASPs are surging. 1Q26 NAND ASP is expected to rise in the mid-to-high 40% range QoQ", "tweet_type": "analysis", "timeline": "quarters", "summary": "Maintains Buy on Samsung & SK Hynix; bullish NAND theme benefits SNDK and MU on surging ASPs (mid-to-high 40% QoQ) and tightening supply-demand.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: There Won't Be a Second Shock\n[Daishin Securities Semiconductor / Ryu Hyung-geun]\n\nThis is a Kioxia FY3Q25 (CY4Q25) Review. We maintain our existing view that a surprise will emerge from NAND. This value needs to be further reflected in Korean semiconductor stock prices. We maintain our Buy recommendations on Samsung Electronics and SK Hynix.\n\nKioxia FY3Q25 (CY4Q25) Review: Please Forget Last November's Disaster\n\n- Results were in line with consensus. The guidance was the surprise. The company issued guidance significantly above consensus, creating a markedly different atmosphere from last November's earnings release.\n\n- Throughout the recent upcycle, Kioxia's high revenue concentration toward a single customer has been a double-edged sword. Pre-negotiated long-term contracts meant products were supplied at prices below market rates, which was the key reason behind Kioxia's lower ASP growth relative to peers.\n\n- Recently, the company appears to have aggressively leveraged the tight supply-demand environment to push through price revisions. Starting from CY1Q26, the revised pricing policy is expected to take effect (estimated ~50% QoQ ASP increase for the North American customer in CY1Q26). This marks the moment Kioxia breaks free from its legacy constraints and the moment mobile NAND ASPs normalize.\n\n- Opportunities are also expanding on the server side. On top of a favorable market backdrop, competitiveness in high-capacity products of 128TB and above is expected to drive outperformance.\n\nNAND Is Just as Strong as DRAM\n\n- AI inference is strengthening the role of storage and driving higher content per unit. In addition, new products are set to be unveiled. At the upcoming GTC 2026 event, new products based on a next-generation solution (Gen 6 Controller) are expected to be announced.\n\n- The supply-demand balance is tightening to historically unprecedented levels. While concerns over weak set-level demand persist, opportunities in the server segment are more than offsetting them.\n\n- Meanwhile, supply discipline is strengthening across the industry. The prevailing approach is to respond through conversion investments and product mix shifts—which carry lower depreciation burdens—rather than new capacity additions. This signals a management philosophy increasingly centered on profitability and long-term supply-demand stability.\n\n- As a result, NAND ASPs are surging. 1Q26 NAND ASP is expected to rise in the mid-to-high 40% range QoQ, and we believe we will soon see operating margins reach the 50%+ levels that marked the previous peak.\n\n$SNDK $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.049104304765785645, "ret_signed_m1d": -0.049104304765785645, "alpha_spy_m1d": -0.06479559171219595, "alpha_c_m1d": -0.07611381690250119, "bench_spy_m1d": 0.01569128694641031, "bench_c_m1d": 0.02700951213671554, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005917880021548139, "ret_signed_p1d": -0.005917880021548139, "alpha_spy_p1d": -0.0066223655153635486, "alpha_c_p1d": -0.008431944537790392, "bench_spy_p1d": 0.0007044854938154099, "bench_c_p1d": 0.0025140645162422537, "ret_p1w": -0.014596378639350482, "ret_signed_p1w": -0.014596378639350482, "alpha_spy_p1w": -0.019308040865776088, "alpha_c_p1w": -0.02177970231868198, "bench_spy_p1w": 0.004711662226425606, "bench_c_p1w": 0.007183323679331499, "ret_p1m": 0.04970730644948396, "ret_signed_p1m": 0.04970730644948396, "alpha_spy_p1m": 0.07756709353352131, "alpha_c_p1m": 0.06701932064448834, "bench_spy_p1m": -0.027859787084037357, "bench_c_p1m": -0.017312014195004388, "ret_p3m": 1.3037333134753686, "ret_signed_p3m": 1.3037333134753686, "alpha_spy_p3m": 1.2172390783622735, "alpha_c_p3m": 1.0436383698969554, "bench_spy_p3m": 0.08649423511309506, "bench_c_p3m": 0.26009494357841323, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5782, -0.6114, -0.6098, -0.6891, -0.6823, -0.6399, -0.6502, -0.6588, -0.6588, -0.6458, -0.6666, -0.6742, -0.6916, -0.6616, -0.6375, -0.6423, -0.6518, -0.6306, -0.6167, -0.6729, -0.6797, -0.6679, -0.6718, -0.6518, -0.623, -0.6176, -0.6114, -0.6032, -0.6032, -0.6033, -0.6125, -0.6189, -0.6234, -0.6234, -0.5633, -0.5652, -0.4453, -0.4391, -0.4692, -0.4012, -0.3824, -0.3815, -0.3847, -0.3507, -0.3438, -0.3438, -0.2811, -0.2047, -0.2013, -0.2482, -0.253, -0.2362, -0.1629, -0.1444, -0.0857, 0.0555, 0.1035, -0.0726, -0.0858, -0.0513, -0.0744, -0.1406, -0.0491, 0.0, -0.0059, -0.0059, -0.063, -0.0474, -0.0146, 0.0312, 0.0574, 0.0131, 0.0033, 0.0343, 0.008, -0.0178, -0.1029, -0.0495, -0.1027, -0.1634, -0.0659, -0.0181, 0.0399, -0.0182, 0.0497, 0.1164, 0.1426, 0.1958, 0.225, 0.126, 0.1146, 0.1145, 0.0755, -0.043, -0.0229, -0.0917, 0.008, 0.0991, 0.1131, 0.1131, 0.1497, 0.1277, 0.239, 0.3511, 0.3514, 0.5112, 0.4985, 0.4148, 0.4588, 0.4612, 0.4486, 0.4335, 0.5534, 0.4794, 0.5705, 0.6979, 0.5903, 0.6884, 0.7397, 0.8833, 0.9925, 1.2312, 1.237, 1.1259, 1.4788, 1.4553, 1.3037, 1.2961, 1.1938, 1.2333, 1.1149, 1.1947, 1.2094, 1.4469, 1.346, 1.346, 1.5219, 1.5226, 1.6046, 1.6892, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022085214626074657", "ticker_idx": 3, "ts": "2026-02-12T23:07:38+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND Is Just as Strong as DRAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Maintains Buy on Samsung & SK Hynix; bullish NAND theme benefits SNDK and MU on surging ASPs (mid-to-high 40% QoQ) and tightening supply-demand.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia: There Won't Be a Second Shock\n[Daishin Securities Semiconductor / Ryu Hyung-geun]\n\nThis is a Kioxia FY3Q25 (CY4Q25) Review. We maintain our existing view that a surprise will emerge from NAND. This value needs to be further reflected in Korean semiconductor stock prices. We maintain our Buy recommendations on Samsung Electronics and SK Hynix.\n\nKioxia FY3Q25 (CY4Q25) Review: Please Forget Last November's Disaster\n\n- Results were in line with consensus. The guidance was the surprise. The company issued guidance significantly above consensus, creating a markedly different atmosphere from last November's earnings release.\n\n- Throughout the recent upcycle, Kioxia's high revenue concentration toward a single customer has been a double-edged sword. Pre-negotiated long-term contracts meant products were supplied at prices below market rates, which was the key reason behind Kioxia's lower ASP growth relative to peers.\n\n- Recently, the company appears to have aggressively leveraged the tight supply-demand environment to push through price revisions. Starting from CY1Q26, the revised pricing policy is expected to take effect (estimated ~50% QoQ ASP increase for the North American customer in CY1Q26). This marks the moment Kioxia breaks free from its legacy constraints and the moment mobile NAND ASPs normalize.\n\n- Opportunities are also expanding on the server side. On top of a favorable market backdrop, competitiveness in high-capacity products of 128TB and above is expected to drive outperformance.\n\nNAND Is Just as Strong as DRAM\n\n- AI inference is strengthening the role of storage and driving higher content per unit. In addition, new products are set to be unveiled. At the upcoming GTC 2026 event, new products based on a next-generation solution (Gen 6 Controller) are expected to be announced.\n\n- The supply-demand balance is tightening to historically unprecedented levels. While concerns over weak set-level demand persist, opportunities in the server segment are more than offsetting them.\n\n- Meanwhile, supply discipline is strengthening across the industry. The prevailing approach is to respond through conversion investments and product mix shifts—which carry lower depreciation burdens—rather than new capacity additions. This signals a management philosophy increasingly centered on profitability and long-term supply-demand stability.\n\n- As a result, NAND ASPs are surging. 1Q26 NAND ASP is expected to rise in the mid-to-high 40% range QoQ, and we believe we will soon see operating margins reach the 50%+ levels that marked the previous peak.\n\n$SNDK $MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008768759210025512, "ret_signed_m1d": -0.008768759210025512, "alpha_spy_m1d": -0.024460046156435822, "alpha_c_m1d": -0.03011768866377651, "bench_spy_m1d": 0.01569128694641031, "bench_c_m1d": 0.021348929453750998, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005580092679010162, "ret_signed_p1d": -0.005580092679010162, "alpha_spy_p1d": -0.006284578172825572, "alpha_c_p1d": -0.009544575029146962, "bench_spy_p1d": 0.0007044854938154099, "bench_c_p1d": 0.0039644823501368, "ret_p1w": 0.00816481827602411, "ret_signed_p1w": 0.00816481827602411, "alpha_spy_p1w": 0.0034531560495985048, "alpha_c_p1w": -0.0019064112012026069, "bench_spy_p1w": 0.004711662226425606, "bench_c_p1w": 0.010071229477226717, "ret_p1m": 0.029374141224238093, "ret_signed_p1m": 0.029374141224238093, "alpha_spy_p1m": 0.05723392830827545, "alpha_c_p1m": 0.07561776845994705, "bench_spy_p1m": -0.027859787084037357, "bench_c_p1m": -0.04624362723570896, "ret_p3m": 0.8525546318006874, "ret_signed_p3m": 0.8525546318006874, "alpha_spy_p3m": 0.7660603966875923, "alpha_c_p3m": 0.4705398568767578, "bench_spy_p3m": 0.08649423511309506, "bench_c_p3m": 0.38201477492392955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4158, -0.4483, -0.4545, -0.5138, -0.4993, -0.4593, -0.4578, -0.444, -0.444, -0.429, -0.4194, -0.4217, -0.4346, -0.4527, -0.4272, -0.4038, -0.3905, -0.3632, -0.3759, -0.4177, -0.4265, -0.4386, -0.4554, -0.3998, -0.3579, -0.3321, -0.3329, -0.3078, -0.3078, -0.3123, -0.2889, -0.2931, -0.3106, -0.3106, -0.2381, -0.246, -0.1704, -0.1798, -0.21, -0.1664, -0.1645, -0.1832, -0.1947, -0.1868, -0.1237, -0.1237, -0.1183, -0.0601, -0.0396, -0.0346, -0.0601, -0.009, 0.0515, 0.0527, 0.0022, 0.0576, 0.0132, -0.0835, -0.0751, -0.0466, -0.0736, -0.0984, -0.0088, 0.0, -0.0056, -0.0056, -0.0343, 0.0169, 0.0082, 0.0343, 0.0169, 0.0098, 0.0363, 0.0038, -0.0039, -0.0031, -0.0828, -0.0319, -0.0409, -0.1055, -0.0595, -0.0262, 0.0114, -0.0208, 0.0294, 0.0672, 0.1153, 0.1154, 0.0732, 0.0216, -0.0232, -0.0445, -0.077, -0.1413, -0.1371, -0.2223, -0.1836, -0.111, -0.1149, -0.1149, -0.0871, -0.0875, -0.0171, 0.0186, 0.0164, 0.0308, 0.1253, 0.1025, 0.105, 0.0997, 0.0837, 0.086, 0.1781, 0.1641, 0.2004, 0.2677, 0.2187, 0.2529, 0.2498, 0.3103, 0.3931, 0.5471, 0.6109, 0.5627, 0.8048, 0.922, 0.8526, 0.9421, 0.8753, 0.7512, 0.647, 0.6886, 0.769, 0.8417, 0.8149, 0.8149, 1.165, 1.2436, 1.2318, 1.3466, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029877007287533675", "ticker_idx": 0, "ts": "2026-03-06T11:09:26+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung to Double Supply Prices in Q2", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND prices doubling Q1→Q2 with uptrend expected to persist; bullish Samsung, SK Hynix, and NAND pricing broadly.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NAND Now Commands Whatever Price It Wants… Samsung to Double Supply Prices in Q2\n\nSamsung Electronics is reportedly set to raise its NAND flash supply prices by approximately 100% in the second quarter. The company has been cutting NAND production to focus on high-value products such as High Bandwidth Memory (HBM), while surging global AI investment has driven a sharp spike in demand for NAND, the storage memory segment.\n\nAccording to industry sources on the 6th, Samsung Electronics — the world's largest NAND supplier — has decided to set its Q2 price increase at a level similar to Q1. Samsung is said to have already raised NAND prices by approximately 100% quarter-on-quarter in Q1, meaning that by Q2, Samsung's NAND supply prices will have roughly doubled compared to late last year.\n\nFinal supply prices may vary slightly by customer, but the industry does not expect Samsung's decided increase to deviate significantly in practice. A senior semiconductor industry executive noted, \"Memory manufacturers now hold unprecedented negotiating leverage, so buyers will have little choice but to accept Samsung's asking prices. SK Hynix, Kioxia, and other producers are also preparing additional price hikes.\"\n\nMajor NAND makers have indeed been raising prices in unison, led by Samsung. According to market research firm DRAMeXchange, the average fixed contract price for 128Gb MLC NAND — a mainstream commodity product — surged to $12.67 last month, up 33.9% from the prior month and a staggering 452.3% higher than a year ago.\n\nThe steep price trajectory reflects NAND's growing role as AI evolves from the training phase into inference, where systems must process and respond to user requests in real time, requiring massive data storage at scale. Surging investment in AI data centers by Big Tech is further fueling storage demand.\n\nWith Samsung and SK Hynix — the #1 and #2 NAND suppliers — having sharply curtailed NAND output to prioritize high-margin products like HBM, the price uptrend is expected to persist for the foreseeable future.\n\n---\n\nhttps://t.co/qcgLzu1iSp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01806592572447019, "ret_signed_m1d": 0.01806592572447019, "alpha_spy_m1d": 0.004784711991328683, "alpha_c_m1d": -0.002984424009301323, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.021050349733771512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07810841685232373, "ret_signed_p1d": -0.07810841685232373, "alpha_spy_p1d": -0.0868684418653628, "alpha_c_p1d": -0.09609954197395376, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.017991125121630036, "ret_p1w": -0.024973458644195756, "ret_signed_p1w": -0.024973458644195756, "alpha_spy_p1w": -0.009967091606065459, "alpha_c_p1w": -0.021404455997489635, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": -0.003569002646706121, "ret_p1m": 0.028159835659485655, "ret_signed_p1m": 0.028159835659485655, "alpha_spy_p1m": 0.04548700857622623, "alpha_c_p1m": 0.030635836809685513, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": -0.002476001150199858, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4268, -0.4289, -0.4326, -0.4242, -0.4458, -0.4564, -0.4294, -0.431, -0.4379, -0.4157, -0.4104, -0.4125, -0.4125, -0.3813, -0.365, -0.3629, -0.3629, -0.3629, -0.3172, -0.2662, -0.262, -0.2508, -0.2625, -0.2614, -0.2625, -0.2689, -0.2545, -0.2354, -0.2088, -0.2067, -0.2285, -0.2056, -0.1908, -0.1918, -0.1918, -0.1525, -0.1371, -0.1461, -0.1472, -0.2009, -0.11, -0.1015, -0.1536, -0.1573, -0.1158, -0.119, -0.1084, -0.051, -0.0372, -0.0372, -0.0372, -0.0372, 0.0096, 0.0101, 0.0255, 0.0627, 0.0813, 0.1583, 0.1504, 0.1504, 0.0367, -0.085, 0.0181, 0.0, -0.0781, -0.0016, 0.0096, -0.0016, -0.025, 0.0027, 0.0303, 0.1079, 0.0654, 0.0595, -0.0101, 0.008, 0.0043, -0.043, -0.043, -0.0613, -0.1097, 0.0098, -0.0501, -0.0086, 0.0282, 0.0463, 0.1208, 0.0862, 0.0968, 0.0702, 0.0995, 0.1235, 0.1581, 0.1501, 0.1421, 0.1661, 0.1581, 0.1953, 0.1687, 0.1953, 0.182, 0.2033, 0.1741, 0.1741, 0.2379, 0.2379, 0.4163, 0.4456, 0.4296, 0.5201, 0.4855, 0.5122, 0.5761, 0.4403, 0.4962, 0.4669, 0.4696, 0.5947, 0.5574, 0.5574, 0.592, 0.6346, 0.5947, 0.6879, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029877007287533675", "ticker_idx": 1, "ts": "2026-03-06T11:09:26+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix, Kioxia, and other producers are also preparing additional price hikes", "tweet_type": "analysis", "timeline": "quarters", "summary": "NAND prices doubling Q1→Q2 with uptrend expected to persist; bullish Samsung, SK Hynix, and NAND pricing broadly.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NAND Now Commands Whatever Price It Wants… Samsung to Double Supply Prices in Q2\n\nSamsung Electronics is reportedly set to raise its NAND flash supply prices by approximately 100% in the second quarter. The company has been cutting NAND production to focus on high-value products such as High Bandwidth Memory (HBM), while surging global AI investment has driven a sharp spike in demand for NAND, the storage memory segment.\n\nAccording to industry sources on the 6th, Samsung Electronics — the world's largest NAND supplier — has decided to set its Q2 price increase at a level similar to Q1. Samsung is said to have already raised NAND prices by approximately 100% quarter-on-quarter in Q1, meaning that by Q2, Samsung's NAND supply prices will have roughly doubled compared to late last year.\n\nFinal supply prices may vary slightly by customer, but the industry does not expect Samsung's decided increase to deviate significantly in practice. A senior semiconductor industry executive noted, \"Memory manufacturers now hold unprecedented negotiating leverage, so buyers will have little choice but to accept Samsung's asking prices. SK Hynix, Kioxia, and other producers are also preparing additional price hikes.\"\n\nMajor NAND makers have indeed been raising prices in unison, led by Samsung. According to market research firm DRAMeXchange, the average fixed contract price for 128Gb MLC NAND — a mainstream commodity product — surged to $12.67 last month, up 33.9% from the prior month and a staggering 452.3% higher than a year ago.\n\nThe steep price trajectory reflects NAND's growing role as AI evolves from the training phase into inference, where systems must process and respond to user requests in real time, requiring massive data storage at scale. Surging investment in AI data centers by Big Tech is further fueling storage demand.\n\nWith Samsung and SK Hynix — the #1 and #2 NAND suppliers — having sharply curtailed NAND output to prioritize high-margin products like HBM, the price uptrend is expected to persist for the foreseeable future.\n\n---\n\nhttps://t.co/qcgLzu1iSp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0183983009063744, "ret_signed_m1d": 0.0183983009063744, "alpha_spy_m1d": 0.005117087173232893, "alpha_c_m1d": -0.020465501489579374, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.03886380239595377, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09523805657980688, "ret_signed_p1d": -0.09523805657980688, "alpha_spy_p1d": -0.10399808159284596, "alpha_c_p1d": -0.13152667768341486, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.036288621103607976, "ret_p1w": -0.015151478250421735, "ret_signed_p1w": -0.015151478250421735, "alpha_spy_p1w": -0.00014511121229143775, "alpha_c_p1w": -0.032941022756158955, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": 0.01778954450573722, "ret_p1m": -0.041125518282007056, "ret_signed_p1m": -0.041125518282007056, "alpha_spy_p1m": -0.023798345365266482, "alpha_c_p1m": -0.08164478863795199, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": 0.040519270355944936, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3886, -0.3659, -0.3897, -0.3832, -0.4015, -0.4275, -0.4048, -0.4037, -0.4091, -0.3735, -0.3691, -0.3648, -0.3648, -0.3529, -0.3086, -0.2968, -0.2968, -0.2968, -0.2687, -0.2481, -0.2157, -0.1984, -0.1833, -0.1963, -0.1909, -0.2028, -0.1984, -0.1909, -0.1833, -0.1747, -0.1974, -0.2006, -0.1844, -0.1714, -0.2049, -0.1358, -0.0915, -0.0699, -0.018, -0.1034, -0.0202, -0.0278, -0.0904, -0.0937, -0.0418, -0.0537, -0.071, -0.0407, -0.0494, -0.0494, -0.0494, -0.0494, -0.0342, 0.0252, 0.0273, 0.0857, 0.0997, 0.1894, 0.1483, 0.1483, 0.0162, -0.0812, 0.0184, 0.0, -0.0952, 0.0152, 0.0335, 0.0065, -0.0152, 0.0541, 0.0498, 0.1429, 0.0963, 0.0898, 0.0097, 0.0671, 0.0768, 0.0097, 0.0097, -0.0552, -0.1266, -0.0292, -0.1017, -0.0519, -0.0411, -0.0087, 0.118, 0.0801, 0.1115, 0.1255, 0.1937, 0.2294, 0.25, 0.2208, 0.2619, 0.3247, 0.3236, 0.3258, 0.3225, 0.3983, 0.4069, 0.3994, 0.3918, 0.3918, 0.566, 0.566, 0.7327, 0.79, 0.8247, 1.0346, 0.9859, 1.1385, 1.132, 0.9686, 0.9913, 0.8885, 0.8885, 1.0996, 1.1006, 1.1006, 1.2208, 1.4275, 1.4777, 1.5253, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2036809171065348382", "ticker_idx": 0, "ts": "2026-03-25T14:15:23+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy; raise target price to US$450", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses GF Buy on Broadcom with raised PT $450 on higher TPU shipment forecasts.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "So....\n\n1. MediaTek will ship only 400,000 TPUs this year.\n\n2. MediaTek’s Zebrafish has not even entered testing yet.\n\n3. AMD will use AVGO’s CPO? https://t.co/yd61AZSoiv", "reply_text": null, "parent_text": null, "quoted_text": "GF Overseas Electronics &amp; Communications\n☄️ Broadcom (AVGO US Buy): Raising 2027 TPU Forecast\n\n☀️ Maintain Buy; raise target price to US$450:\nWe forecast total TPU shipments to reach 4.5 million units in 2026 and 7.9 million units in 2027, above our previous estimates of 4.5", "parent_author": null, "ret_m1d": -0.0016310310769859226, "ret_signed_m1d": -0.0016310310769859226, "alpha_spy_m1d": 0.00391084433461486, "alpha_c_m1d": 0.009471151264133448, "bench_spy_m1d": -0.005541875411600783, "bench_c_m1d": -0.011102182341119371, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02945322983208387, "ret_signed_p1d": -0.02945322983208387, "alpha_spy_p1d": -0.011594485673945165, "alpha_c_p1d": 0.01610837905051854, "bench_spy_p1d": -0.017858744158138706, "bench_c_p1d": -0.04556160888260241, "ret_p1w": -0.016687078087132434, "ret_signed_p1w": -0.016687078087132434, "alpha_spy_p1w": -0.0142815224216154, "alpha_c_p1w": 0.0009811791106972834, "bench_spy_p1w": -0.002405555665517034, "bench_c_p1w": -0.017668257197829718, "ret_p1m": 0.31721089569729033, "ret_signed_p1m": 0.31721089569729033, "alpha_spy_p1m": 0.23860487836060007, "alpha_c_p1m": 0.10962726738185347, "bench_spy_p1m": 0.07860601733669026, "bench_c_p1m": 0.20758362831543686, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1022, 0.0936, 0.0951, 0.0833, 0.0833, 0.0881, 0.0749, 0.076, 0.0752, 0.0407, 0.0798, 0.1025, 0.11, 0.0639, 0.0737, 0.1009, 0.1009, 0.0411, 0.0292, 0.0188, 0.0018, 0.0168, 0.0417, 0.0431, 0.0352, 0.037, 0.0364, 0.0027, -0.0358, -0.0281, 0.0421, 0.0766, 0.0656, 0.0729, 0.0366, 0.0178, 0.0178, 0.0409, 0.0439, 0.0454, 0.0412, 0.034, 0.0188, 0.0402, 0.007, 0.0002, -0.0021, -0.0176, -0.0061, 0.0416, 0.0344, 0.0822, 0.0723, 0.0691, 0.0516, 0.0084, 0.017, 0.0057, -0.0111, 0.0011, -0.0281, 0.0116, -0.0016, 0.0, -0.0295, -0.0569, -0.0797, -0.0292, -0.0167, -0.0134, -0.0134, -0.0137, 0.0476, 0.0998, 0.1132, 0.1654, 0.1911, 0.1944, 0.2444, 0.2499, 0.2752, 0.2535, 0.2615, 0.3257, 0.3172, 0.3261, 0.3118, 0.2541, 0.2718, 0.3093, 0.3214, 0.3064, 0.3405, 0.3345, 0.2941, 0.3488, 0.3438, 0.3152, 0.3073, 0.3795, 0.3337, 0.3196, 0.2894, 0.3104, 0.3004, 0.299, 0.299, 0.3237, 0.3232, 0.338, 0.4014, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2029738792601674002", "ticker_idx": 0, "ts": "2026-03-06T02:00:13+00:00", "symbol": "1810.HK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Xiaomi (1810 HK, Buy): Stress Test... we lower our target price from HK$52 to HK$45, based on a SOTP valuation implying 32x/27x 2026E/2027E P/E", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF Securities reiterates Buy on Xiaomi with PT cut to HK$45 from HK$52; memory cost and EV demand headwinds weigh but Buy maintained.", "resolved_tickers": ["1810.HK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics & Telecom》 Xiaomi (1810 HK, Buy): Stress Test\n\nTarget price lowered to HK$45.0: Xiaomi's share price is down 11% YTD and 12% over the past three months, primarily pressured by rising memory prices and softening EV demand in China. We believe Xiaomi is entering a \"stress test\" phase, and investors are unlikely to re-engage with the stock until signs of a slowdown in the memory upcycle emerge or the company demonstrates margin resilience. Per TrendForce, memory price increases have exceeded expectations significantly, meaning the most acute impact on Xiaomi's smartphone gross margins will materialize in 1Q26/2Q26 — we await signals of price moderation driven by highly conservative consumer demand and capacity releases from CXMT. Accordingly, we cut our 2025/2026 smartphone shipment estimates to 165M/147.5M units; combined with lower gross margin assumptions and broader EV market demand softness, we reduce earnings estimates by 10% and 33% respectively, and lower our target price from HK$52 to HK$45, based on a SOTP valuation implying 32x/27x 2026E/2027E P/E.\n\n4Q25 results preview — adjusted net profit estimated at RMB 5.2bn: IDC data shows Xiaomi's 4Q25 shipments at 37.8M units, down 13% QoQ and 11% YoY, pressured by rising costs and weakening demand. Factoring in soft IoT performance, in-line internet services, and EV deliveries of 146K units, we estimate 4Q25 revenue of RMB 114bn, up 1% QoQ and 4% YoY. Overall gross margin is expected to decline to 20.7%, down 2.2ppt QoQ, dragged by smartphone gross margin of 8.1% and EV gross margin of 22.5%, resulting in adjusted net profit of RMB 5.2bn, down 54% QoQ and 38% YoY — broadly in line with already-lowered consensus. On the EV side, deliveries recovered to 146K units in 4Q25; however, due to a significant decline in the Ultra model mix and promotional discounts, we estimate gross margin fell to 22.5% (vs. 25.5% in the prior quarter) and expect margins to remain near this level in the near-to-medium term.\n\nMemory cost stress test: Per TrendForce and our estimates, LPDDR5X prices are expected to rise 45%/91% QoQ in 4Q25/1Q26, primarily driven by server applications crowding out supply, while eMMC prices are expected to rise 35%/60% QoQ — placing enormous pressure on smartphone OEM gross margins. Although the pace of price increases may moderate in 2Q26, memory prices will remain elevated. To mitigate margin pressure, we expect Xiaomi — which holds some memory inventory — to optimize its product mix, maintain lean inventories, and selectively downgrade eMMC specs on lower-end handsets. Nevertheless, accounting for inventory drawdowns, we project Xiaomi's smartphone gross margin will fall to 8.1% in 4Q25 and bottom out near this level through the current cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03650501115644922, "ret_signed_m1d": -0.03650501115644922, "alpha_spy_m1d": -0.049786224889590724, "alpha_c_m1d": -0.05755536089022073, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.021050349733771512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007779836938233808, "ret_signed_p1d": 0.007779836938233808, "alpha_spy_p1d": -0.000980188074805266, "alpha_c_p1d": -0.010211288183396228, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.017991125121630036, "ret_p1w": -0.0029921747336902893, "ret_signed_p1w": -0.0029921747336902893, "alpha_spy_p1w": 0.012014192304440008, "alpha_c_p1w": 0.0005768279130158316, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": -0.003569002646706121, "ret_p1m": -0.07600236826281281, "ret_signed_p1m": -0.07600236826281281, "alpha_spy_p1m": -0.05867519534607224, "alpha_c_p1m": -0.07352636711261296, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": -0.002476001150199858, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.2346, 0.2501, 0.2621, 0.2855, 0.2519, 0.2238, 0.2334, 0.2029, 0.213, 0.1909, 0.173, 0.173, 0.173, 0.173, 0.1544, 0.1777, 0.173, 0.173, 0.2053, 0.1771, 0.1598, 0.1418, 0.1376, 0.1317, 0.1592, 0.1364, 0.1305, 0.1329, 0.1101, 0.0916, 0.0616, 0.0598, 0.0545, 0.0844, 0.0539, 0.064, 0.0868, 0.0958, 0.0622, 0.0491, 0.0353, 0.0162, 0.0449, 0.0527, 0.0533, 0.0646, 0.1101, 0.0928, 0.1023, 0.0922, 0.0922, 0.0922, 0.0922, 0.058, 0.094, 0.0694, 0.0652, 0.0527, 0.0443, -0.0084, -0.0551, -0.0425, -0.0365, 0.0, 0.0078, 0.006, -0.0024, -0.0036, -0.003, 0.0533, 0.058, 0.0515, 0.0868, -0.0066, -0.0407, -0.0221, -0.0269, -0.0293, -0.0126, -0.0311, -0.0497, -0.0419, -0.076, -0.076, -0.076, -0.076, -0.0197, -0.0616, -0.0754, -0.0826, -0.076, -0.0754, -0.0407, -0.0425, -0.0329, -0.0305, -0.0485, -0.067, -0.0664, -0.0694, -0.1047, -0.0981, -0.1317, -0.1317, -0.073, -0.0886, -0.0778, -0.0688, -0.0521, -0.0515, -0.0586, -0.0485, -0.0509, -0.0814, -0.0826, -0.0832, -0.0981, -0.1125, -0.1023, -0.1023, -0.1095, -0.1502, -0.1454, -0.161, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2020285656401080408", "ticker_idx": 0, "ts": "2026-02-07T23:56:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "HBM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics will become the world's first to mass-produce and ship the next-generation High Bandwidth Memory (HBM), HBM4 — a 'game changer' for artificial intelligence (AI) semiconductors with the world's highest performance — after the Lunar New Year holiday", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares report that Samsung is first to mass-produce HBM4, shipping to NVIDIA post-Lunar New Year with 11.7 Gbps performance and 3x HBM sales growth expected YoY; implied bullish on stock.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics to Begin World's First HBM4 Mass Production After Lunar New Year Holiday…Seizing Initiative in Next-Generation Market\n\nSamsung Electronics will become the world's first to mass-produce and ship the next-generation High Bandwidth Memory (HBM), HBM4—a 'game changer' for artificial intelligence (AI) semiconductors with the world's highest performance—after the Lunar New Year holiday.\n\nThis strategy aims to overcome the setbacks with previous-generation products that sparked crisis concerns in its semiconductor business, seize the initiative in the next-generation market, and solidify its position as the world's number one memory manufacturer.\n\nAccording to industry sources on the 8th, Samsung Electronics has decided to begin mass production shipments of HBM4 to NVIDIA after this Lunar New Year holiday, possibly as early as the third week of this month.\n\nSamsung Electronics passed quality tests with NVIDIA early on and received purchase orders (PO), and reportedly finalized this schedule after comprehensively considering NVIDIA's launch plans for the 'Vera Rubin' AI accelerator that will incorporate HBM4.\n\nThe volume of HBM4 samples Samsung Electronics is supplying for customer product module testing is also observed to have expanded significantly with this PO.\n\nNVIDIA is expected to unveil 'Vera Rubin' featuring Samsung Electronics' HBM4 for the first time at its technology conference 'GTC 2026' next month.\n\nThis marks the world's first mass production shipment of next-generation HBM4.\n\nMoreover, Samsung Electronics' HBM4 is evaluated as having world-class performance.\n\nFrom the start of HBM4 development, Samsung Electronics set its goal to achieve top performance exceeding JEDEC (international semiconductor standards organization) standards.\n\nAccordingly, it made the bold move of simultaneously applying 1c (10nm-class 6th generation) DRAM and 4nm foundry process to this HBM4.\n\nWith this industry-unique process combination, Samsung Electronics' HBM4 achieves data processing speeds of up to 11.7 Gbps (gigabits per second), surpassing the JEDEC standard of 8Gbps.\n\nThis represents a 37% improvement over the JEDEC standard and 22% over the previous generation HBM3E (9.6Gbps).\n\nAdditionally, Samsung Electronics' HBM4 achieves memory bandwidth of up to 3TB/s per single stack, a 2.4-fold improvement over its predecessor, and offers capacity of up to 36GB with 12-layer stacking technology.\n\nFuture application of 16-layer stacking technology could enable capacity expansion up to 48GB.\n\nIt can significantly reduce power consumption and cooling costs for servers and data centers through low-power design while maximizing computational performance.\n\nSamsung Electronics, as the world's only semiconductor company capable of providing one-stop solutions encompassing logic, memory, foundry, and packaging, plans to leverage the synergy between its advanced memory and foundry process capabilities to deliver top-tier HBM performance.\n\nThe company also expects HBM sales volume to increase more than threefold year-over-year this year and has decided to install new production lines at its Pyeongtaek Campus Plant 4 to expand HBM production capacity.\n\nFollowing the achievement of stable yields despite applying cutting-edge processes and commencing mass production shipments, yields are expected to improve further as production scales up.\n\nFurthermore, Samsung Electronics is likely to establish optimal HBM4 production plans based on thorough review of overall memory market conditions.\n\nIn a situation where prices are soaring not only for HBM but all memory products, the company intends to efficiently distribute and utilize its world's largest production capacity.\n\nThis confidence is also based on the belief that Samsung Electronics can control market share while maintaining leadership, given that its HBM4 enters the market first with top performance.\n\nAn industry official stated, \"Samsung Electronics, with the world's largest production capacity and most diverse product lineup, has proven its restored technological competitiveness through the world's first mass production of top-performance HBM4,\" adding, \"Building on this foundation, it will be able to lead the market from the most advantageous position.\"\n\n---\n\nhttps://t.co/22hXlZ79VT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04687501029159635, "ret_signed_m1d": -0.04687501029159635, "alpha_spy_m1d": -0.04207631668347811, "alpha_c_m1d": -0.0313883600803615, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.015486650211234854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00360579897461355, "ret_signed_p1d": -0.00360579897461355, "alpha_spy_p1d": -0.0009686385425327071, "alpha_c_p1d": 0.001974979716881209, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.005580778691494759, "ret_p1w": 0.0889422572521763, "ret_signed_p1w": 0.0889422572521763, "alpha_spy_p1w": 0.1065228270365195, "alpha_c_p1w": 0.11538114834364888, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": -0.02643889109147257, "ret_p1m": 0.12920674586734182, "ret_signed_p1m": 0.12920674586734182, "alpha_spy_p1m": 0.15337280848239565, "alpha_c_p1m": 0.15425044228594076, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.025043696418598937, "ret_p3m": 0.6349876825470429, "ret_signed_p3m": 0.6349876825470429, "alpha_spy_p3m": 0.5778827788238594, "alpha_c_p3m": 0.44976986122915896, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.1852178213178839, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3834, -0.3852, -0.4187, -0.3984, -0.4151, -0.4229, -0.3984, -0.433, -0.4217, -0.4061, -0.3852, -0.381, -0.399, -0.3972, -0.3816, -0.375, -0.3714, -0.3517, -0.3451, -0.3517, -0.3541, -0.3583, -0.3487, -0.3732, -0.3852, -0.3547, -0.3565, -0.3643, -0.3391, -0.3332, -0.3356, -0.3356, -0.3003, -0.2819, -0.2794, -0.2794, -0.2794, -0.2278, -0.1701, -0.1653, -0.1526, -0.1659, -0.1647, -0.1659, -0.1731, -0.1569, -0.1352, -0.1052, -0.1028, -0.1274, -0.1016, -0.0847, -0.0859, -0.0859, -0.0415, -0.024, -0.0343, -0.0355, -0.0962, 0.0066, 0.0162, -0.0427, -0.0469, 0.0, -0.0036, 0.0084, 0.0733, 0.0889, 0.0889, 0.0889, 0.0889, 0.1418, 0.1424, 0.1599, 0.2019, 0.223, 0.3101, 0.3011, 0.3011, 0.1725, 0.0349, 0.1514, 0.131, 0.0427, 0.1292, 0.1418, 0.1292, 0.1028, 0.134, 0.1653, 0.253, 0.2049, 0.1983, 0.1196, 0.14, 0.1358, 0.0823, 0.0823, 0.0617, 0.0069, 0.1421, 0.0743, 0.1213, 0.1629, 0.1833, 0.2676, 0.2285, 0.2405, 0.2104, 0.2436, 0.2707, 0.3098, 0.3008, 0.2917, 0.3188, 0.3098, 0.352, 0.3218, 0.352, 0.3369, 0.361, 0.3279, 0.3279, 0.4001, 0.4001, 0.6019, 0.635, 0.6169, 0.7193, 0.6802, 0.7103, 0.7825, 0.629, 0.6922, 0.6591, 0.6621, 0.8036, 0.7615, 0.7615, 0.8006, 0.8488, 0.8036, 0.909, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010542624835551711", "ticker_idx": 0, "ts": "2026-01-12T02:41:30+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reiterate Buy Rating, Raise Target Price to $286: We recommend continuing to stay long NVIDIA, with key drivers being strong H200 demand and an unchanged Vera Rubin 200 (VR200) timeline", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF Securities reiterates Buy on NVDA, raises PT to $286 on H200 demand + VR200 ramp; also reiterates Buy on AVGO.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities Overseas Electronics Communication\n\n☄️ NVIDIA (NVDA US Buy): Upside Potential from H200, Followed by a Leap with Vera Rubin [Report Uploaded]\n\n☀️ Reiterate Buy Rating, Raise Target Price to $286: NVIDIA’s stock has declined 11% from its November peak last year, driven primarily by concerns over Google competition, OpenAI models and financing, and fears of an AI bubble. However, recent investor feedback indicates signs of reversal in the prior “long Google TPU, short OpenAI/NVIDIA” pair trade. We recommend continuing to stay long NVIDIA, with key drivers being strong H200 demand and an unchanged Vera Rubin 200 (VR200) timeline—both critical to widening its product leadership. In the near term, we expect revenue upside to be driven by H200, where we have observed significant demand and emerging supply-chain build-up plans. Incorporating H200 shipment volumes, we raise FY26/FY27 EPS by 0% and 6%, respectively, and increase the target price from $270 to $286, still based on 36x FY27 P/E.\n\n☀️ Both Gemini and OpenAI Provide Positive Signals:\n\nGoogle: DeepMind VP Oriol Vinyals stated that Gemini 3’s strength stems from pre-training and post-training improvements, implying that major performance gains are still possible through pre-training mechanisms without increasing parameter counts—this suggests continued scaling benefits in pre-training can drive significant model progress.\nOpenAI: User feedback on the latest GPT-5.2 is mixed, but investors are more focused on the upcoming new model trained on Blackwell. On the financing front, The Wall Street Journal reported that OpenAI is seeking a funding round of up to $100 billion at an $830 billion valuation, and our investor conversations reflect high optimism. With a potential post-funding IPO on the horizon, we expect concerns around OpenAI to ease.\n\n☀️ Strong Demand Leads to Upward Revision of H200 CoWoS Capacity Plans: Bloomberg reported on January 8 that China is expected to approve H200 purchases this quarter. Accordingly, we raise TSMC’s CoWoS-S capacity build-up plan by 30,000 wafers, primarily in 2Q26, implying roughly 800,000 H200 units. Combined with approximately 700,000 units in inventory, we estimate the current H200 shipment plan at around 1.5 million units. Factoring in a discount for policy uncertainty, we now forecast H200 volumes of 250k / 450k / 1.1 million units for the April quarter, July quarter, and FY2027, respectively. This translates to total sales of $74 billion / $89.5 billion for the April and July quarters, exceeding consensus by 4% and 14%. We concur with Reuters’ December 30 report that orders exceed 2 million units, with major customers accounting for roughly 40%.\n\n☀️ VR200 Moves to Full Production, Enhancing Product Competitiveness: At CES 2026, Jensen Huang stated that Vera Rubin delivers 5× higher inference performance and 3.5× higher training performance compared to Blackwell. Relative to Google’s TPU, Vera Rubin is expected to significantly outperform TPU v7 (targeted at Blackwell) and likely lead v8AX as well. While we remain positive on Broadcom (AVGO Buy) and highlighted potential TPU external sales since August last year, we believe external customers will be limited to non-Tier-1 CSPs such as Anthropic—contrary to the market view of broad adoption including Meta—due to AI competition dynamics and the TPU software learning curve. On timing, we reiterate that VR200 will start production in June, followed by a full ramp in 3Q26.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0004326060359227668, "ret_signed_m1d": -0.0004326060359227668, "alpha_spy_m1d": 0.0011353853178507345, "alpha_c_m1d": 0.0032534699071157247, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004704209026523465, "ret_signed_p1d": 0.004704209026523465, "alpha_spy_p1d": 0.00670368413239375, "alpha_c_p1d": 0.002477221085521064, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.006975163809415097, "ret_signed_p1w": 0.006975163809415097, "alpha_spy_p1w": 0.012009991969976319, "alpha_c_p1w": -0.017931432451060658, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.019465703904395282, "ret_signed_p1m": 0.019465703904395282, "alpha_spy_p1m": 0.023838823866474734, "alpha_c_p1m": -0.01662707746902381, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": -0.005515541307894334, "ret_signed_p3m": -0.005515541307894334, "alpha_spy_p3m": 0.013750767276853981, "alpha_c_p3m": -0.10701043078928218, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0277, -0.017, -0.0094, -0.0125, -0.0205, -0.0253, -0.0151, 0.0071, 0.0354, 0.0869, 0.1194, 0.097, 0.0948, 0.1186, 0.0743, 0.0555, 0.0169, 0.0173, 0.0762, 0.0444, 0.0478, 0.0103, 0.0282, 0.0089, -0.0194, 0.0085, -0.0233, -0.0328, -0.013, -0.0386, -0.0254, -0.0254, -0.043, -0.0272, -0.0189, -0.029, -0.0084, -0.0137, 0.0033, 0.0002, -0.0063, -0.0217, -0.0536, -0.0468, -0.039, -0.0757, -0.0584, -0.0214, -0.0068, 0.0231, 0.0198, 0.0198, 0.0302, 0.0177, 0.0141, 0.0084, 0.0084, 0.0211, 0.0172, 0.0124, 0.0225, 0.0005, -0.0004, 0.0, 0.0047, -0.0097, 0.0114, 0.007, 0.007, -0.0371, -0.0088, -0.0005, 0.0148, 0.0083, 0.0194, 0.0356, 0.0409, 0.0335, 0.0036, -0.0249, -0.0581, -0.0706, 0.0025, 0.0276, 0.0195, 0.0276, 0.0108, -0.0115, -0.0115, 0.0002, 0.0164, 0.016, 0.0264, 0.0357, 0.0428, 0.0574, -0.0003, -0.0419, -0.0133, -0.0264, -0.0103, -0.0087, -0.0385, -0.0124, -0.0009, 0.0059, -0.0097, -0.0253, -0.0092, -0.0162, -0.0245, -0.0344, -0.0661, -0.0502, -0.0526, -0.0338, -0.074, -0.0941, -0.1069, -0.0569, -0.0496, -0.0408, -0.0408, -0.0394, -0.0369, -0.0154, -0.0055, 0.02, 0.0237, 0.0626, 0.0754, 0.0726, 0.0906, 0.0926, 0.0808, 0.095, 0.0795, 0.1262, 0.1713, 0.1527, 0.1315, 0.0792, 0.0731, 0.0733, 0.0626, 0.1238, 0.1437, 0.1637, 0.1866, 0.1939, 0.2212, 0.2748, 0.2184, 0.2022, 0.1929, 0.2084, 0.187, 0.1644, 0.1644, 0.1618, 0.1496, 0.1585, 0.1417, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010542624835551711", "ticker_idx": 1, "ts": "2026-01-12T02:41:30+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we remain positive on Broadcom (AVGO Buy)", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF Securities reiterates Buy on NVDA, raises PT to $286 on H200 demand + VR200 ramp; also reiterates Buy on AVGO.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities Overseas Electronics Communication\n\n☄️ NVIDIA (NVDA US Buy): Upside Potential from H200, Followed by a Leap with Vera Rubin [Report Uploaded]\n\n☀️ Reiterate Buy Rating, Raise Target Price to $286: NVIDIA’s stock has declined 11% from its November peak last year, driven primarily by concerns over Google competition, OpenAI models and financing, and fears of an AI bubble. However, recent investor feedback indicates signs of reversal in the prior “long Google TPU, short OpenAI/NVIDIA” pair trade. We recommend continuing to stay long NVIDIA, with key drivers being strong H200 demand and an unchanged Vera Rubin 200 (VR200) timeline—both critical to widening its product leadership. In the near term, we expect revenue upside to be driven by H200, where we have observed significant demand and emerging supply-chain build-up plans. Incorporating H200 shipment volumes, we raise FY26/FY27 EPS by 0% and 6%, respectively, and increase the target price from $270 to $286, still based on 36x FY27 P/E.\n\n☀️ Both Gemini and OpenAI Provide Positive Signals:\n\nGoogle: DeepMind VP Oriol Vinyals stated that Gemini 3’s strength stems from pre-training and post-training improvements, implying that major performance gains are still possible through pre-training mechanisms without increasing parameter counts—this suggests continued scaling benefits in pre-training can drive significant model progress.\nOpenAI: User feedback on the latest GPT-5.2 is mixed, but investors are more focused on the upcoming new model trained on Blackwell. On the financing front, The Wall Street Journal reported that OpenAI is seeking a funding round of up to $100 billion at an $830 billion valuation, and our investor conversations reflect high optimism. With a potential post-funding IPO on the horizon, we expect concerns around OpenAI to ease.\n\n☀️ Strong Demand Leads to Upward Revision of H200 CoWoS Capacity Plans: Bloomberg reported on January 8 that China is expected to approve H200 purchases this quarter. Accordingly, we raise TSMC’s CoWoS-S capacity build-up plan by 30,000 wafers, primarily in 2Q26, implying roughly 800,000 H200 units. Combined with approximately 700,000 units in inventory, we estimate the current H200 shipment plan at around 1.5 million units. Factoring in a discount for policy uncertainty, we now forecast H200 volumes of 250k / 450k / 1.1 million units for the April quarter, July quarter, and FY2027, respectively. This translates to total sales of $74 billion / $89.5 billion for the April and July quarters, exceeding consensus by 4% and 14%. We concur with Reuters’ December 30 report that orders exceed 2 million units, with major customers accounting for roughly 40%.\n\n☀️ VR200 Moves to Full Production, Enhancing Product Competitiveness: At CES 2026, Jensen Huang stated that Vera Rubin delivers 5× higher inference performance and 3.5× higher training performance compared to Blackwell. Relative to Google’s TPU, Vera Rubin is expected to significantly outperform TPU v7 (targeted at Blackwell) and likely lead v8AX as well. While we remain positive on Broadcom (AVGO Buy) and highlighted potential TPU external sales since August last year, we believe external customers will be limited to non-Tier-1 CSPs such as Anthropic—contrary to the market view of broad adoption including Meta—due to AI competition dynamics and the TPU software learning curve. On timing, we reiterate that VR200 will start production in June, followed by a full ramp in 3Q26.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020555944061408704, "ret_signed_m1d": -0.020555944061408704, "alpha_spy_m1d": -0.018987952707635203, "alpha_c_m1d": -0.016869868118370213, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006814066578461286, "ret_signed_p1d": 0.006814066578461286, "alpha_spy_p1d": 0.008813541684331572, "alpha_c_p1d": 0.004587078637458886, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": -0.0014196333820668539, "ret_signed_p1w": -0.0014196333820668539, "alpha_spy_p1w": 0.0036151947784943683, "alpha_c_p1w": -0.02632622964254261, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": -0.033417562019744995, "ret_signed_p1m": -0.033417562019744995, "alpha_spy_p1m": -0.029044442057665543, "alpha_c_p1m": -0.06951034339316409, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.009779667689439497, "ret_signed_p3m": 0.009779667689439497, "alpha_spy_p3m": 0.029045976274187812, "alpha_c_p3m": -0.09171522179194835, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0044, 0.0036, -0.0101, -0.0103, -0.029, -0.0357, -0.0244, 0.0035, 0.026, 0.0569, 0.0938, 0.0668, 0.0475, 0.0274, -0.0027, 0.0173, 0.0077, -0.0098, 0.0156, -0.0026, 0.0066, -0.0366, -0.0295, -0.029, -0.0351, 0.0044, -0.0172, -0.0359, 0.0711, 0.0911, 0.1266, 0.1266, 0.1419, 0.0941, 0.0813, 0.0786, 0.0798, 0.1059, 0.1366, 0.1513, 0.1703, 0.1516, 0.02, -0.037, -0.0328, -0.0761, -0.0652, -0.0355, -0.0305, -0.0082, -0.0057, -0.0057, -0.0002, -0.008, -0.0067, -0.0173, -0.0173, -0.013, -0.025, -0.024, -0.0247, -0.056, -0.0206, 0.0, 0.0068, -0.035, -0.0261, -0.0014, -0.0014, -0.0557, -0.0665, -0.0759, -0.0913, -0.0777, -0.0551, -0.0539, -0.061, -0.0594, -0.0599, -0.0905, -0.1254, -0.1184, -0.0548, -0.0235, -0.0334, -0.0268, -0.0597, -0.0768, -0.0768, -0.0558, -0.0531, -0.0517, -0.0555, -0.0621, -0.0759, -0.0565, -0.0866, -0.0927, -0.0948, -0.1089, -0.0985, -0.0552, -0.0617, -0.0183, -0.0273, -0.0302, -0.0461, -0.0853, -0.0775, -0.0877, -0.103, -0.0919, -0.1184, -0.0824, -0.0944, -0.0929, -0.1196, -0.1445, -0.1652, -0.1194, -0.1081, -0.1051, -0.1051, -0.1054, -0.0498, -0.0024, 0.0098, 0.0571, 0.0805, 0.0834, 0.1287, 0.1337, 0.1567, 0.137, 0.1442, 0.2025, 0.1948, 0.2028, 0.1899, 0.1376, 0.1536, 0.1877, 0.1986, 0.185, 0.2159, 0.2104, 0.1738, 0.2234, 0.219, 0.193, 0.1858, 0.2513, 0.2097, 0.197, 0.1696, 0.1886, 0.1795, 0.1783, 0.1783, 0.2007, 0.2003, 0.2137, 0.2711, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2010543898800820293", "ticker_idx": 0, "ts": "2026-01-12T02:46:34+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$NVDA TP $286!", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares/endorses GF Securities Buy on NVDA with raised PT $286 on H200 upside and Vera Rubin catalyst.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "und", "tweet_text": "$NVDA TP $286! https://t.co/uPaRaz3Q9f", "reply_text": null, "parent_text": null, "quoted_text": "GF Securities Overseas Electronics Communication\n\n☄️ NVIDIA (NVDA US Buy): Upside Potential from H200, Followed by a Leap with Vera Rubin [Report Uploaded]\n\n☀️ Reiterate Buy Rating, Raise Target Price to $286: NVIDIA’s stock has declined 11% from its November peak last year,", "parent_author": null, "ret_m1d": -0.0004326060359227668, "ret_signed_m1d": -0.0004326060359227668, "alpha_spy_m1d": 0.0011353853178507345, "alpha_c_m1d": 0.0032534699071157247, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004704209026523465, "ret_signed_p1d": 0.004704209026523465, "alpha_spy_p1d": 0.00670368413239375, "alpha_c_p1d": 0.002477221085521064, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.006975163809415097, "ret_signed_p1w": 0.006975163809415097, "alpha_spy_p1w": 0.012009991969976319, "alpha_c_p1w": -0.017931432451060658, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.019465703904395282, "ret_signed_p1m": 0.019465703904395282, "alpha_spy_p1m": 0.023838823866474734, "alpha_c_p1m": -0.01662707746902381, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": -0.005515541307894334, "ret_signed_p3m": -0.005515541307894334, "alpha_spy_p3m": 0.013750767276853981, "alpha_c_p3m": -0.10701043078928218, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0277, -0.017, -0.0094, -0.0125, -0.0205, -0.0253, -0.0151, 0.0071, 0.0354, 0.0869, 0.1194, 0.097, 0.0948, 0.1186, 0.0743, 0.0555, 0.0169, 0.0173, 0.0762, 0.0444, 0.0478, 0.0103, 0.0282, 0.0089, -0.0194, 0.0085, -0.0233, -0.0328, -0.013, -0.0386, -0.0254, -0.0254, -0.043, -0.0272, -0.0189, -0.029, -0.0084, -0.0137, 0.0033, 0.0002, -0.0063, -0.0217, -0.0536, -0.0468, -0.039, -0.0757, -0.0584, -0.0214, -0.0068, 0.0231, 0.0198, 0.0198, 0.0302, 0.0177, 0.0141, 0.0084, 0.0084, 0.0211, 0.0172, 0.0124, 0.0225, 0.0005, -0.0004, 0.0, 0.0047, -0.0097, 0.0114, 0.007, 0.007, -0.0371, -0.0088, -0.0005, 0.0148, 0.0083, 0.0194, 0.0356, 0.0409, 0.0335, 0.0036, -0.0249, -0.0581, -0.0706, 0.0025, 0.0276, 0.0195, 0.0276, 0.0108, -0.0115, -0.0115, 0.0002, 0.0164, 0.016, 0.0264, 0.0357, 0.0428, 0.0574, -0.0003, -0.0419, -0.0133, -0.0264, -0.0103, -0.0087, -0.0385, -0.0124, -0.0009, 0.0059, -0.0097, -0.0253, -0.0092, -0.0162, -0.0245, -0.0344, -0.0661, -0.0502, -0.0526, -0.0338, -0.074, -0.0941, -0.1069, -0.0569, -0.0496, -0.0408, -0.0408, -0.0394, -0.0369, -0.0154, -0.0055, 0.02, 0.0237, 0.0626, 0.0754, 0.0726, 0.0906, 0.0926, 0.0808, 0.095, 0.0795, 0.1262, 0.1713, 0.1527, 0.1315, 0.0792, 0.0731, 0.0733, 0.0626, 0.1238, 0.1437, 0.1637, 0.1866, 0.1939, 0.2212, 0.2748, 0.2184, 0.2022, 0.1929, 0.2084, 0.187, 0.1644, 0.1644, 0.1618, 0.1496, 0.1585, 0.1417, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2012020780016840951", "ticker_idx": 0, "ts": "2026-01-16T04:35:10+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Don't bet against TSMC!", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Strong long on TSMC; rigorous supply-chain checks confirm demand is not a bubble, author endorses the bull case.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "CC Wei’s comment about being “scared by the scale” should basically be taken as a light joke to shift the tone of the call in the moment. What we should focus on is that, after conducting extremely rigorous supply-chain checks—including validating demand all the way down to their customers’ customers—TSMC has concluded that the demand we’re seeing right now is not a bubble.\n\nTSMC is also well known for having an outstanding analyst team. Many former TSMC analysts have continued to build excellent track records across the semiconductor industry.\n\nDon’t bet against TSMC!", "reply_text": null, "parent_text": null, "quoted_text": "If they did it on purpose, it’s moral hazard; if they didn’t know, it’s incompetence!!\n\n@theinformation Shame on you!!", "parent_author": null, "ret_m1d": -0.028735635015458172, "ret_signed_m1d": -0.028735635015458172, "alpha_spy_m1d": -0.029574211627886537, "alpha_c_m1d": -0.018795299740556826, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011494239929334293, "ret_signed_p1d": 0.011494239929334293, "alpha_spy_p1d": 0.011494239929334293, "alpha_c_p1d": 0.011494239929334293, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.01724139508612388, "ret_signed_p1w": 0.01724139508612388, "alpha_spy_p1w": 0.020754709890014045, "alpha_c_p1w": 0.017990710252762887, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.10057472255410338, "ret_signed_p1m": 0.10057472255410338, "alpha_spy_p1m": 0.11490255212192357, "alpha_c_p1m": 0.0822676062322798, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.1993024921301616, "ret_signed_p3m": 0.1993024921301616, "alpha_spy_p3m": 0.1845675410055425, "alpha_c_p3m": 0.0679056454886271, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1523, -0.1637, -0.1694, -0.1694, -0.1523, -0.1551, -0.1379, -0.1379, -0.1408, -0.1351, -0.1379, -0.1637, -0.1608, -0.1637, -0.1551, -0.1608, -0.1551, -0.1637, -0.1809, -0.1723, -0.1952, -0.2009, -0.1666, -0.2067, -0.2124, -0.1895, -0.1752, -0.178, -0.1752, -0.1923, -0.1809, -0.1694, -0.1723, -0.1637, -0.1437, -0.1523, -0.1379, -0.1552, -0.1494, -0.1667, -0.1753, -0.1782, -0.1782, -0.1782, -0.158, -0.1437, -0.1408, -0.1408, -0.1322, -0.1207, -0.1264, -0.1092, -0.1092, -0.0891, -0.0402, -0.0201, -0.0374, -0.0316, -0.0345, -0.0287, -0.0172, -0.0172, -0.0287, 0.0, 0.0115, 0.0201, 0.0, 0.0115, 0.0172, 0.0086, 0.023, 0.046, 0.0374, 0.0201, 0.0144, 0.0345, 0.0259, 0.0144, 0.023, 0.0431, 0.0805, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.092, 0.1293, 0.158, 0.1466, 0.1466, 0.1351, 0.1121, 0.0718, 0.092, 0.0862, 0.0402, 0.0632, 0.1149, 0.0833, 0.0718, 0.0603, 0.0782, 0.0984, 0.0667, 0.0609, 0.0436, 0.0436, 0.0638, 0.0609, 0.0494, 0.0263, 0.0148, 0.0696, 0.0436, 0.0436, 0.0436, 0.0725, 0.1243, 0.1272, 0.1532, 0.1474, 0.1849, 0.1993, 0.2022, 0.1705, 0.1676, 0.182, 0.182, 0.1993, 0.2598, 0.306, 0.2771, 0.257, 0.231, 0.231, 0.3117, 0.2973, 0.2973, 0.3319, 0.3204, 0.2887, 0.3002, 0.28, 0.3089, 0.306, 0.2916, 0.2714, 0.2598, 0.2858, 0.3002, 0.3319, 0.3089, 0.3262, 0.3233, 0.3579, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021812054613799248", "ticker_idx": 0, "ts": "2026-02-12T05:02:11+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We are currently passing through the most favorable risk-reward phase of the memory cycle. We define the inflection point as the moment when the rate of investment growth exceeds the rate of earnings growth, which we expect to occur after 2027. In the near term, caution is warranted around 3Q as the pace of DRAM price increases decelerates. However, given the durability of earnings, any correction is likely to be temporary.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: most favorable risk-reward phase of memory cycle now; supply inflection not until 2027, any near-term correction on DRAM price deceleration seen as temporary.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron Actively Refutes HBM4 Disqualification Rumors\n\n[Samsung Securities Semiconductor, IT / Jong-wook Lee]\n\nMicron held a conference call to address the HBM4 disqualification controversy, and its stock rose 10%.\nHow should we interpret this?\n\n1. What Micron Said\n1.Guidance: Noted that DRAM price increases will drive results above what was indicated at the December 2025 earnings call. No specific figures provided.\n2. DRAM market: Some customers are receiving only 50–66% of requested volumes. HBM is sold out. LTA (long-term agreement) commitments expanding. Inventory levels are very lean.\n3. AI technology: AI demand is driving an increasing share of HBM and high-performance LPDDR. Hyperscaler capex is expected to reach $800 billion. CY26 server demand growth in the mid-teens percent (maintaining prior guidance).\n4. HBM4: HBM4 shipments have begun — one quarter ahead of prior guidance.\n5. NAND: Gaining NAND market share in both TLC and QLC.\n6. New capacity additions: Idaho (ID1) to come online mid-2027, Taiwan Taoyuan (P5) by end of 2027, Singapore NAND fab in 2H 2028.\n7.HBM’s capacity cannibalization effect: As HBM generations advance, wafer cannibalization intensifies, exacerbating memory supply shortages.\n\n2. Implications of Micron’s Comments\n1.HBM4: Micron aggressively countered market rumors of HBM4 disqualification by using the phrase “shipments have begun.” However, we still believe Micron’s shipments are limited to lower tranches with modest volumes, and its market share will decline year-over-year. Starting with HBM4E, Micron’s competitiveness could recover as it adopts foundry-based base dies. The equity market appears largely indifferent to the HBM competitive dynamics.\n2. Bullish market outlook: Profits ultimately come from the commodity business. The company officially flagged DRAM price increases and strong earnings, and the market is cheering robust demand and consecutive earnings surprises.\n3. Timing of the supply inflection point: Micron’s investments will not impact supply until mid-2027. Until then, the company has limited strategic options to address high-end, AI-driven demand.\n\n3. Our View\n1.We are currently passing through the most favorable risk-reward phase of the memory cycle. We define the inflection point as the moment when the rate of investment growth exceeds the rate of earnings growth, which we expect to occur after 2027.\n2. In the near term, caution is warranted around 3Q as the pace of DRAM price increases decelerates. However, given the durability of earnings, any correction is likely to be temporary.\n\nThank you.\n(Published 2/12/2026)\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008768759210025512, "ret_signed_m1d": -0.008768759210025512, "alpha_spy_m1d": -0.024460046156435822, "alpha_c_m1d": -0.03011768866377651, "bench_spy_m1d": 0.01569128694641031, "bench_c_m1d": 0.021348929453750998, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005580092679010162, "ret_signed_p1d": -0.005580092679010162, "alpha_spy_p1d": -0.006284578172825572, "alpha_c_p1d": -0.009544575029146962, "bench_spy_p1d": 0.0007044854938154099, "bench_c_p1d": 0.0039644823501368, "ret_p1w": 0.00816481827602411, "ret_signed_p1w": 0.00816481827602411, "alpha_spy_p1w": 0.0034531560495985048, "alpha_c_p1w": -0.0019064112012026069, "bench_spy_p1w": 0.004711662226425606, "bench_c_p1w": 0.010071229477226717, "ret_p1m": 0.029374141224238093, "ret_signed_p1m": 0.029374141224238093, "alpha_spy_p1m": 0.05723392830827545, "alpha_c_p1m": 0.07561776845994705, "bench_spy_p1m": -0.027859787084037357, "bench_c_p1m": -0.04624362723570896, "ret_p3m": 0.8525546318006874, "ret_signed_p3m": 0.8525546318006874, "alpha_spy_p3m": 0.7660603966875923, "alpha_c_p3m": 0.4705398568767578, "bench_spy_p3m": 0.08649423511309506, "bench_c_p3m": 0.38201477492392955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4158, -0.4483, -0.4545, -0.5138, -0.4993, -0.4593, -0.4578, -0.444, -0.444, -0.429, -0.4194, -0.4217, -0.4346, -0.4527, -0.4272, -0.4038, -0.3905, -0.3632, -0.3759, -0.4177, -0.4265, -0.4386, -0.4554, -0.3998, -0.3579, -0.3321, -0.3329, -0.3078, -0.3078, -0.3123, -0.2889, -0.2931, -0.3106, -0.3106, -0.2381, -0.246, -0.1704, -0.1798, -0.21, -0.1664, -0.1645, -0.1832, -0.1947, -0.1868, -0.1237, -0.1237, -0.1183, -0.0601, -0.0396, -0.0346, -0.0601, -0.009, 0.0515, 0.0527, 0.0022, 0.0576, 0.0132, -0.0835, -0.0751, -0.0466, -0.0736, -0.0984, -0.0088, 0.0, -0.0056, -0.0056, -0.0343, 0.0169, 0.0082, 0.0343, 0.0169, 0.0098, 0.0363, 0.0038, -0.0039, -0.0031, -0.0828, -0.0319, -0.0409, -0.1055, -0.0595, -0.0262, 0.0114, -0.0208, 0.0294, 0.0672, 0.1153, 0.1154, 0.0732, 0.0216, -0.0232, -0.0445, -0.077, -0.1413, -0.1371, -0.2223, -0.1836, -0.111, -0.1149, -0.1149, -0.0871, -0.0875, -0.0171, 0.0186, 0.0164, 0.0308, 0.1253, 0.1025, 0.105, 0.0997, 0.0837, 0.086, 0.1781, 0.1641, 0.2004, 0.2677, 0.2187, 0.2529, 0.2498, 0.3103, 0.3931, 0.5471, 0.6109, 0.5627, 0.8048, 0.922, 0.8526, 0.9421, 0.8753, 0.7512, 0.647, 0.6886, 0.769, 0.8417, 0.8149, 0.8149, 1.165, 1.2436, 1.2318, 1.3466, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2032362362947453402", "ticker_idx": 0, "ts": "2026-03-13T07:45:21+00:00", "symbol": "EMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we forecast EMC's 2026/2027 EPS at NT$78.65/139.96, representing 12.0%/24.9% above consensus. We set our target price at NT$3,499, based on 25x FY2027E P/E", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities initiates bullish thesis on EMC: above-consensus EPS forecasts for 2026/2027, PT NT$3,499 on AI CCL capacity expansion and margin recovery from 2Q26.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "CCL context: EMC = Elite Material Co. 2383.TW not US EMC", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics & Communications》\n\n☄️ EMC 4Q25 Review: AI CCL to Accelerate from 2Q Onward\n\n☀️ Aggressive Capacity Expansion; Profitability Structure to Improve in 2Q: EMC announced aggressive CCL capacity expansion plans on its 4Q25 earnings call, targeting monthly capacity of 9.45 million sheets — a 54% increase from its current 2026 monthly capacity of 6.15 million sheets. Meanwhile, EMC's 4Q25 gross margin declined 1.5ppt due to an AI platform transition. Although upstream raw material costs (particularly copper and fiberglass) surged significantly in 1Q26, we believe price increases will materially alleviate margin pressure from 2Q26 onward. Our supply chain checks indicate that M7+ material price hike negotiations stalled in 1Q26, as PCB customers were unable to renegotiate pricing for existing AI products. However, as new AI platforms ramp in 2Q26, new PCB pricing should support at least a 10% price increase on EMC's AI CCL. Based on the combination of price increases and capacity expansion, we forecast EMC's 2026/2027 EPS at NT$78.65/139.96, representing 12.0%/24.9% above consensus. We set our target price at NT$3,499, based on 25x FY2027E P/E.\n\n☀️ New Platforms Ramping Gradually from 2Q:\n\n1. AWS: The T3 platform begins ramping in 2Q26, with volumes expected to recover to 3Q25 run-rate levels; EMC holds ~90% market share.\n\n2. Google: Google's TPUv7 switch boards use M8 materials, with EMC capturing ~80% market share — significantly higher than the sub-30% share on prior platforms. Google became EMC's largest AI customer in 1Q26, with monthly shipments exceeding 800,000 sheets.\n\n3. NVIDIA: The Rubin platform's midplane transitions to M9 materials and begins ramping in 2Q26; EMC remains the second-tier supplier for switch trays. EMC's revenue growth is expected to accelerate in 2Q, driven by the new AI platform ramps.\n\n☀️ CCL Material Upgrades Continue to Progress: The Rubin platform midplane upgrades to M9, and driven by higher data throughput and signal integrity requirements, switch trays are also expected to adopt M9 materials alongside the introduction of CPX boards. Looking ahead to Rubin Ultra, CCL is expected to fully transition to M9, and the Kyber platform midplane should meaningfully increase CCL content value. Additionally, M10-grade CCL samples for the Feynman platform are expected to begin qualification in 2Q26. On the ASIC side, new M9 design wins are also emerging in products such as Google TPUv8e/p. The ongoing CCL material upgrade cycle should benefit EMC, given its industry-leading R&D pace.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.038817005545286554, "ret_signed_m1d": -0.038817005545286554, "alpha_spy_m1d": -0.04450933425007775, "alpha_c_m1d": -0.04641925486046761, "bench_spy_m1d": 0.005692328704791194, "bench_c_m1d": 0.007602249315181053, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.044362292051756014, "ret_signed_p1d": -0.044362292051756014, "alpha_spy_p1d": -0.054539158163239176, "alpha_c_p1d": -0.0588359598042455, "bench_spy_p1d": 0.010176866111483163, "bench_c_p1d": 0.014473667752489483, "ret_p1w": 0.038817005545286554, "ret_signed_p1w": 0.038817005545286554, "alpha_spy_p1w": 0.05685856132833356, "alpha_c_p1w": 0.049855075758657, "bench_spy_p1w": -0.018041555783047003, "bench_c_p1w": -0.011038070213370443, "ret_p1m": 0.22920517560073939, "ret_signed_p1m": 0.22920517560073939, "alpha_spy_p1m": 0.19042500230348014, "alpha_c_p1m": 0.16348121078068645, "bench_spy_p1m": 0.038780173297259246, "bench_c_p1m": 0.06572396482005294, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4473, -0.4473, -0.4455, -0.427, -0.4085, -0.403, -0.3752, -0.3752, -0.3937, -0.3808, -0.3919, -0.3919, -0.3919, -0.4122, -0.4011, -0.3734, -0.4159, -0.4288, -0.4214, -0.4048, -0.3956, -0.3826, -0.4196, -0.3993, -0.4159, -0.403, -0.4233, -0.3863, -0.3346, -0.3567, -0.3272, -0.2976, -0.3383, -0.3549, -0.3401, -0.3142, -0.2458, -0.2791, -0.2884, -0.2311, -0.2218, -0.1885, -0.1885, -0.1885, -0.1885, -0.1885, -0.1885, -0.1885, -0.1885, -0.1959, -0.146, -0.1054, -0.098, -0.098, -0.1201, -0.1423, -0.1904, -0.1257, -0.1275, -0.1959, -0.122, -0.0943, -0.0388, 0.0, -0.0444, -0.0018, -0.0018, 0.0684, 0.0388, 0.0333, -0.0074, 0.0518, 0.0887, 0.0758, 0.0296, -0.0388, 0.0573, 0.0296, 0.0296, 0.0296, 0.0665, 0.1719, 0.1904, 0.2643, 0.2292, 0.3327, 0.4085, 0.4177, 0.4048, 0.4233, 0.4603, 0.5028, 0.5157, 0.6543, 0.7486, 0.6174, 0.6414, 0.6802, 0.6802, 0.7579, 0.7486, 0.8299, 0.8336, 0.756, 0.793, 0.8115, 0.8226, 0.8189, 0.7061, 0.6802, 0.6765, 0.6248, 0.7338, 0.8503, 1.0351, 0.9556, 0.9612, 0.8688, 0.8928, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2009164727411462303", "ticker_idx": 0, "ts": "2026-01-08T07:26:14+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CHINA TO APPROVE SOME NVIDIA H200 PURCHASES AS SOON AS THIS QTR", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "China approving Nvidia H200 purchases as soon as this quarter — bullish catalyst for NVDA near-term.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "*CHINA TO APPROVE SOME NVIDIA H200 PURCHASES AS SOON AS THIS QTR\n\n---\n\nhttps://t.co/J48u4eniEm", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.021995237555278324, "ret_signed_m1d": 0.021995237555278324, "alpha_spy_m1d": 0.021893694563360322, "alpha_c_m1d": 0.006057779504062832, "bench_spy_m1d": 0.0001015429919180022, "bench_c_m1d": 0.015937458051215492, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.000972775721840291, "ret_signed_p1d": -0.000972775721840291, "alpha_spy_p1d": -0.007586118115932905, "alpha_c_p1d": -0.027992644679373235, "bench_spy_p1d": 0.006613342394092614, "bench_c_p1d": 0.027019868957532944, "ret_p1w": 0.010862579760640756, "ret_signed_p1w": 0.010862579760640756, "alpha_spy_p1w": 0.006903290689400432, "alpha_c_p1w": -0.0351292725020107, "bench_spy_p1w": 0.003959289071240324, "bench_c_p1w": 0.045991852262651456, "ret_p1m": 0.001999567050479767, "ret_signed_p1m": 0.001999567050479767, "alpha_spy_p1m": 0.00038979108769421345, "alpha_c_p1m": -0.05781884619816546, "bench_spy_p1m": 0.0016097759627855535, "bench_c_p1m": 0.05981841324864523, "ret_p3m": -0.03745326722262887, "ret_signed_p3m": -0.03745326722262887, "alpha_spy_p3m": 0.0038654993069547894, "alpha_c_p3m": -0.09265402268123513, "bench_spy_p3m": -0.04131876652958366, "bench_c_p3m": 0.055200755458606254, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0177, -0.0271, -0.0282, -0.0175, -0.0099, -0.013, -0.021, -0.0258, -0.0156, 0.0065, 0.0348, 0.0864, 0.1188, 0.0964, 0.0942, 0.118, 0.0737, 0.0549, 0.0164, 0.0168, 0.0757, 0.0438, 0.0473, 0.0098, 0.0277, 0.0084, -0.0199, 0.0079, -0.0238, -0.0333, -0.0135, -0.0391, -0.0259, -0.0259, -0.0435, -0.0277, -0.0194, -0.0295, -0.009, -0.0142, 0.0028, -0.0004, -0.0068, -0.0222, -0.0542, -0.0473, -0.0396, -0.0762, -0.0589, -0.0219, -0.0073, 0.0225, 0.0193, 0.0193, 0.0297, 0.0172, 0.0135, 0.0079, 0.0079, 0.0206, 0.0166, 0.0119, 0.022, 0.0, -0.001, -0.0005, 0.0042, -0.0103, 0.0109, 0.0064, 0.0064, -0.0377, -0.0093, -0.0011, 0.0142, 0.0077, 0.0188, 0.035, 0.0404, 0.0329, 0.0031, -0.0254, -0.0586, -0.0711, 0.002, 0.027, 0.0189, 0.0271, 0.0103, -0.0121, -0.0121, -0.0004, 0.0159, 0.0155, 0.0258, 0.0352, 0.0422, 0.0569, -0.0008, -0.0424, -0.0138, -0.027, -0.0108, -0.0092, -0.039, -0.0129, -0.0015, 0.0054, -0.0102, -0.0258, -0.0098, -0.0168, -0.025, -0.035, -0.0666, -0.0507, -0.0531, -0.0343, -0.0745, -0.0946, -0.1073, -0.0575, -0.0502, -0.0413, -0.0413, -0.0399, -0.0375, -0.0159, -0.0061, 0.0195, 0.0231, 0.062, 0.0748, 0.072, 0.09, 0.092, 0.0803, 0.0944, 0.079, 0.1256, 0.1707, 0.1521, 0.1309, 0.0786, 0.0725, 0.0727, 0.062, 0.1232, 0.1431, 0.1631, 0.186, 0.1932, 0.2205, 0.2741, 0.2177, 0.2015, 0.1923, 0.2078, 0.1863, 0.1638, 0.1638, 0.1612, 0.149, 0.1579, 0.1411, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013156310985158725", "ticker_idx": 0, "ts": "2026-01-19T07:47:21+00:00", "symbol": "Unimicron Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura: Unimicron Technology reiterate Buy with a higher TP of TWD418, implying ~53% upside", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura reiterates Buy on Unimicron, raises TP to TWD418 (+53%) on AI substrate bottleneck cycle; also carries Buy on TSMC and Resonac.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura: Unimicron Technology reiterate Buy with a higher TP of TWD418, implying ~53% upside\n\nRecently, TSMC (2330 TT, Buy) announced record capex (USD52-56bn) for 2026E, well-above consensus of USD48-50bn, on booming AI demand (report), and Resonac (4004 JP, Buy) announced to raise its prices for all kinds of CCLs and PPs by 30%, effective 1 March, given the tight supply of T-glass and E-glass, and rising costs (link). We think the announcements indicate faster-than-expected demand growth for AI chips; and, the challenging supply-demand (S-D) situation for key materials and cost increases have led to earlier and faster price hikes of materials. We believe this will likely drive more price hikes for both BT and ABF substrates in the coming months. We expect the substrate bottlenecks will likely shift from the current T-glass level to the ABF substrate level from late-2026F to 2028F. We raise our earnings by 7.8%/24.0% for 2026F/27F, on stronger AI growth and faster price hikes. We raise TP to TWD418, based on 25x 2027F EPS of 16.72 (vs prior TWD270, on 20x 2027F EPS of TWD13.5). We raise our target PE from 20x to 25x, at the high end of its historical band (7-25x), as we are in the early stage of the bull cycle. The stock currently trades at 16x 2027F EPS of TWD16.72.\n\nT-glass S-D worsening, despite some supply additions from 1Q26F. The bottleneck could shift to “ABF substrate” level from late-2026, and worsen in 2027-28F\n\nIn our 2 Dec report, we had highlighted that T-glass shortage, the root cause of substrate material shortage, will likely continue for another year, until end-2026F, and Nittobo (3110 JP, Not rated) is converting some low-Dk/Df glass for PCB to support T-glass for substrate from late-1Q26F. However, despite increased supply, of late, the T-glass S-D gap continues to deteriorate into 2Q-3Q26F, as demand growth from AI is even faster. TSMC’s surging capex also implies tighter S-D dynamics for “ABF substrates” in 2027-28F; while T-glass supply bottlenecks may ease after Nittobo completes its 2-3x capacity expansion in 1H27F. ABF substrate expansion lags T-glass expansion through 2027F. Meanwhile, E-glass lead time has doubled from 8 weeks to 16 weeks, and we expect E-glass supply will likely deteriorate sharply in the coming months, underpinned by Resonac’s price hikes for both T-glass and E-glass CCLs. The potential E-glass shortage stems from many glass fiber cloth makers shifting production from E-glass to T-glass — and the conversion takes 6 months with less output, while new supplier/product qualifications could take another 3-4 months (if not longer). This indicates unavailability of glass fiber cloth capacity for 10-12 months.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0897009966777409, "ret_signed_m1d": -0.0897009966777409, "alpha_spy_m1d": -0.0897009966777409, "alpha_c_m1d": -0.0897009966777409, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023255813953488413, "ret_signed_p1d": 0.023255813953488413, "alpha_spy_p1d": 0.04361259967703346, "alpha_c_p1d": 0.049213740202460166, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.025957926248971752, "ret_p1w": 0.10631229235880402, "ret_signed_p1w": 0.10631229235880402, "alpha_spy_p1w": 0.10476530123132854, "alpha_c_p1w": 0.1031532935820112, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.0031589987767928207, "ret_p1m": 0.2259136212624584, "ret_signed_p1m": 0.2259136212624584, "alpha_spy_p1m": 0.23865110183215954, "alpha_c_p1m": 0.2680781239992257, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": -0.042164502736767306, "ret_p3m": 1.1428571428571428, "ret_signed_p3m": 1.1428571428571428, "alpha_spy_p3m": 1.1256286717432622, "alpha_c_p3m": 1.0976090754226433, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.045248067434499495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4718, -0.4817, -0.4817, -0.4568, -0.4601, -0.4485, -0.4684, -0.4568, -0.407, -0.4402, -0.4594, -0.4544, -0.4676, -0.4577, -0.4511, -0.4478, -0.4116, -0.4302, -0.4435, -0.4352, -0.4668, -0.4302, -0.4684, -0.4551, -0.4153, -0.407, -0.3787, -0.3804, -0.3455, -0.3488, -0.3472, -0.3405, -0.2857, -0.2674, -0.2492, -0.2442, -0.2525, -0.2525, -0.2641, -0.289, -0.2957, -0.3123, -0.2857, -0.2674, -0.2841, -0.2791, -0.2791, -0.2674, -0.2674, -0.2708, -0.2691, -0.2691, -0.2724, -0.2359, -0.2342, -0.2492, -0.2442, -0.2508, -0.191, -0.1561, -0.1694, -0.1711, -0.0897, 0.0, 0.0233, 0.0664, 0.0864, 0.1262, 0.1063, 0.1761, 0.2209, 0.1445, 0.2575, 0.2126, 0.2791, 0.289, 0.2143, 0.1296, 0.1545, 0.2342, 0.2259, 0.2259, 0.2259, 0.2259, 0.2259, 0.2259, 0.2259, 0.2259, 0.3472, 0.4817, 0.5233, 0.5997, 0.5997, 0.598, 0.5382, 0.3854, 0.495, 0.4352, 0.2924, 0.4086, 0.5482, 0.5947, 0.6877, 0.8239, 0.804, 0.9269, 0.9302, 0.8405, 0.6578, 0.5282, 0.6811, 0.6678, 0.6777, 0.6395, 0.4767, 0.6229, 0.7243, 0.7243, 0.7243, 0.8738, 1.0598, 1.0764, 1.1196, 1.0797, 0.9834, 1.0532, 1.1429, 1.1362, 1.2525, 1.3588, 1.3787, 1.4186, 1.6246, 1.7874, 1.7409, 1.6678, 1.9336, 1.9336, 2.0266, 2.0, 1.8505, 1.9767, 1.7176, 1.8605, 1.907, 1.9635, 1.9269, 1.7276, 1.711, 1.7176, 1.7342, 2.0066, 2.2226, 2.289, 2.6047, 2.588, 2.4053, 2.505, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013156310985158725", "ticker_idx": 1, "ts": "2026-01-19T07:47:21+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC (2330 TT, Buy) announced record capex (USD52-56bn) for 2026E, well-above consensus", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura reiterates Buy on Unimicron, raises TP to TWD418 (+53%) on AI substrate bottleneck cycle; also carries Buy on TSMC and Resonac.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura: Unimicron Technology reiterate Buy with a higher TP of TWD418, implying ~53% upside\n\nRecently, TSMC (2330 TT, Buy) announced record capex (USD52-56bn) for 2026E, well-above consensus of USD48-50bn, on booming AI demand (report), and Resonac (4004 JP, Buy) announced to raise its prices for all kinds of CCLs and PPs by 30%, effective 1 March, given the tight supply of T-glass and E-glass, and rising costs (link). We think the announcements indicate faster-than-expected demand growth for AI chips; and, the challenging supply-demand (S-D) situation for key materials and cost increases have led to earlier and faster price hikes of materials. We believe this will likely drive more price hikes for both BT and ABF substrates in the coming months. We expect the substrate bottlenecks will likely shift from the current T-glass level to the ABF substrate level from late-2026F to 2028F. We raise our earnings by 7.8%/24.0% for 2026F/27F, on stronger AI growth and faster price hikes. We raise TP to TWD418, based on 25x 2027F EPS of 16.72 (vs prior TWD270, on 20x 2027F EPS of TWD13.5). We raise our target PE from 20x to 25x, at the high end of its historical band (7-25x), as we are in the early stage of the bull cycle. The stock currently trades at 16x 2027F EPS of TWD16.72.\n\nT-glass S-D worsening, despite some supply additions from 1Q26F. The bottleneck could shift to “ABF substrate” level from late-2026, and worsen in 2027-28F\n\nIn our 2 Dec report, we had highlighted that T-glass shortage, the root cause of substrate material shortage, will likely continue for another year, until end-2026F, and Nittobo (3110 JP, Not rated) is converting some low-Dk/Df glass for PCB to support T-glass for substrate from late-1Q26F. However, despite increased supply, of late, the T-glass S-D gap continues to deteriorate into 2Q-3Q26F, as demand growth from AI is even faster. TSMC’s surging capex also implies tighter S-D dynamics for “ABF substrates” in 2027-28F; while T-glass supply bottlenecks may ease after Nittobo completes its 2-3x capacity expansion in 1H27F. ABF substrate expansion lags T-glass expansion through 2027F. Meanwhile, E-glass lead time has doubled from 8 weeks to 16 weeks, and we expect E-glass supply will likely deteriorate sharply in the coming months, underpinned by Resonac’s price hikes for both T-glass and E-glass CCLs. The potential E-glass shortage stems from many glass fiber cloth makers shifting production from E-glass to T-glass — and the conversion takes 6 months with less output, while new supplier/product qualifications could take another 3-4 months (if not longer). This indicates unavailability of glass fiber cloth capacity for 10-12 months.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011363623711922677, "ret_signed_m1d": -0.011363623711922677, "alpha_spy_m1d": -0.011363623711922677, "alpha_c_m1d": -0.011363623711922677, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008522769972260757, "ret_signed_p1d": 0.008522769972260757, "alpha_spy_p1d": 0.028879555695805803, "alpha_c_p1d": 0.033498417800992164, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": -0.0028408537396618083, "ret_signed_p1w": -0.0028408537396618083, "alpha_spy_p1w": -0.0043878448671372805, "alpha_c_p1w": 0.001080341262159834, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": 0.08806820554014494, "ret_signed_p1m": 0.08806820554014494, "alpha_spy_p1m": 0.10080568610984608, "alpha_c_p1m": 0.07028555648122858, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.1885242479454503, "ret_signed_p3m": 0.1885242479454503, "alpha_spy_p3m": 0.1712957768315697, "alpha_c_p3m": 0.05263181518259552, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1732, -0.1789, -0.1789, -0.1619, -0.1647, -0.1477, -0.1477, -0.1506, -0.1449, -0.1477, -0.1732, -0.1704, -0.1732, -0.1647, -0.1704, -0.1647, -0.1732, -0.1902, -0.1817, -0.2044, -0.21, -0.176, -0.2157, -0.2213, -0.1987, -0.1845, -0.1874, -0.1845, -0.2015, -0.1902, -0.1789, -0.1817, -0.1732, -0.1534, -0.1619, -0.1477, -0.1648, -0.1591, -0.1761, -0.1847, -0.1875, -0.1875, -0.1875, -0.1676, -0.1534, -0.1506, -0.1506, -0.142, -0.1307, -0.1364, -0.1193, -0.1193, -0.0994, -0.0511, -0.0312, -0.0483, -0.0426, -0.0455, -0.0398, -0.0284, -0.0284, -0.0398, -0.0114, 0.0, 0.0085, -0.0114, 0.0, 0.0057, -0.0028, 0.0114, 0.0341, 0.0256, 0.0085, 0.0028, 0.0227, 0.0142, 0.0028, 0.0114, 0.0313, 0.0682, 0.0881, 0.0881, 0.0881, 0.0881, 0.0881, 0.0881, 0.0881, 0.0881, 0.0795, 0.1165, 0.1449, 0.1335, 0.1335, 0.1222, 0.0994, 0.0597, 0.0795, 0.0739, 0.0284, 0.0511, 0.1023, 0.071, 0.0597, 0.0483, 0.066, 0.0859, 0.0546, 0.0489, 0.0318, 0.0318, 0.0517, 0.0489, 0.0375, 0.0147, 0.0033, 0.0574, 0.0318, 0.0318, 0.0318, 0.0603, 0.1116, 0.1144, 0.1401, 0.1344, 0.1714, 0.1857, 0.1885, 0.1572, 0.1543, 0.1686, 0.1686, 0.1857, 0.2455, 0.2911, 0.2626, 0.2427, 0.217, 0.217, 0.2968, 0.2826, 0.2826, 0.3168, 0.3054, 0.274, 0.2854, 0.2655, 0.294, 0.2911, 0.2769, 0.2569, 0.2455, 0.2712, 0.2854, 0.3168, 0.294, 0.3111, 0.3082, 0.3424, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013156310985158725", "ticker_idx": 2, "ts": "2026-01-19T07:47:21+00:00", "symbol": "Resonac", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Resonac (4004 JP, Buy) announced to raise its prices for all kinds of CCLs and PPs by 30%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura reiterates Buy on Unimicron, raises TP to TWD418 (+53%) on AI substrate bottleneck cycle; also carries Buy on TSMC and Resonac.", "resolved_tickers": ["4004.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Resonac Holdings (formerly Showa Denko) listed on TSE as 4004.T", "bench_c": "XLB", "bench_c_reason": "industry='Chemicals'", "bench_c_industry": "Chemicals", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura: Unimicron Technology reiterate Buy with a higher TP of TWD418, implying ~53% upside\n\nRecently, TSMC (2330 TT, Buy) announced record capex (USD52-56bn) for 2026E, well-above consensus of USD48-50bn, on booming AI demand (report), and Resonac (4004 JP, Buy) announced to raise its prices for all kinds of CCLs and PPs by 30%, effective 1 March, given the tight supply of T-glass and E-glass, and rising costs (link). We think the announcements indicate faster-than-expected demand growth for AI chips; and, the challenging supply-demand (S-D) situation for key materials and cost increases have led to earlier and faster price hikes of materials. We believe this will likely drive more price hikes for both BT and ABF substrates in the coming months. We expect the substrate bottlenecks will likely shift from the current T-glass level to the ABF substrate level from late-2026F to 2028F. We raise our earnings by 7.8%/24.0% for 2026F/27F, on stronger AI growth and faster price hikes. We raise TP to TWD418, based on 25x 2027F EPS of 16.72 (vs prior TWD270, on 20x 2027F EPS of TWD13.5). We raise our target PE from 20x to 25x, at the high end of its historical band (7-25x), as we are in the early stage of the bull cycle. The stock currently trades at 16x 2027F EPS of TWD16.72.\n\nT-glass S-D worsening, despite some supply additions from 1Q26F. The bottleneck could shift to “ABF substrate” level from late-2026, and worsen in 2027-28F\n\nIn our 2 Dec report, we had highlighted that T-glass shortage, the root cause of substrate material shortage, will likely continue for another year, until end-2026F, and Nittobo (3110 JP, Not rated) is converting some low-Dk/Df glass for PCB to support T-glass for substrate from late-1Q26F. However, despite increased supply, of late, the T-glass S-D gap continues to deteriorate into 2Q-3Q26F, as demand growth from AI is even faster. TSMC’s surging capex also implies tighter S-D dynamics for “ABF substrates” in 2027-28F; while T-glass supply bottlenecks may ease after Nittobo completes its 2-3x capacity expansion in 1H27F. ABF substrate expansion lags T-glass expansion through 2027F. Meanwhile, E-glass lead time has doubled from 8 weeks to 16 weeks, and we expect E-glass supply will likely deteriorate sharply in the coming months, underpinned by Resonac’s price hikes for both T-glass and E-glass CCLs. The potential E-glass shortage stems from many glass fiber cloth makers shifting production from E-glass to T-glass — and the conversion takes 6 months with less output, while new supplier/product qualifications could take another 3-4 months (if not longer). This indicates unavailability of glass fiber cloth capacity for 10-12 months.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.045794966236955226, "ret_signed_m1d": -0.045794966236955226, "alpha_spy_m1d": -0.045794966236955226, "alpha_c_m1d": -0.045794966236955226, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026028238182934316, "ret_signed_p1d": -0.026028238182934316, "alpha_spy_p1d": -0.00567145245938927, "alpha_c_p1d": -0.0159624127869048, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.010065825396029515, "ret_p1w": -0.015469613259668558, "ret_signed_p1w": -0.015469613259668558, "alpha_spy_p1w": -0.01701660438714403, "alpha_c_p1w": -0.04320169699485543, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.027732083735186874, "ret_p1m": 0.29036218538980973, "ret_signed_p1m": 0.29036218538980973, "alpha_spy_p1m": 0.3030996659595109, "alpha_c_p1m": 0.20778208551945498, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.08258009987035475, "ret_p3m": 0.6206261510128914, "ret_signed_p3m": 0.6206261510128914, "alpha_spy_p3m": 0.6033976798990108, "alpha_c_p3m": 0.5528336831874752, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.06779246782541626, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2898, -0.3092, -0.2762, -0.2668, -0.284, -0.2861, -0.2781, -0.2658, -0.2658, -0.2855, -0.3127, -0.2968, -0.3044, -0.2989, -0.3065, -0.2843, -0.2802, -0.2657, -0.2547, -0.3027, -0.3015, -0.2549, -0.3144, -0.3144, -0.3209, -0.2939, -0.2314, -0.2066, -0.2255, -0.2339, -0.2429, -0.253, -0.2446, -0.2521, -0.2395, -0.2212, -0.229, -0.2237, -0.2296, -0.2675, -0.268, -0.2964, -0.2697, -0.2093, -0.2126, -0.1929, -0.2017, -0.1956, -0.1835, -0.1988, -0.1988, -0.1988, -0.1988, -0.1711, -0.1519, -0.1261, -0.1626, -0.1639, -0.1639, -0.1651, -0.1632, -0.1244, -0.0458, 0.0, -0.026, -0.0201, 0.0193, 0.0128, -0.0155, 0.0529, 0.0953, 0.1117, 0.0988, 0.0754, 0.1449, 0.1481, 0.0809, 0.0791, 0.1316, 0.1686, 0.1686, 0.2443, 0.2449, 0.2879, 0.2904, 0.2744, 0.3333, 0.3775, 0.3775, 0.461, 0.4586, 0.4027, 0.4647, 0.5335, 0.5396, 0.4438, 0.4721, 0.4966, 0.3082, 0.326, 0.4641, 0.4138, 0.3628, 0.3579, 0.3039, 0.4119, 0.3757, 0.3757, 0.2793, 0.2971, 0.3229, 0.3517, 0.3204, 0.2812, 0.2034, 0.3186, 0.2959, 0.3874, 0.4543, 0.4469, 0.5678, 0.5998, 0.6458, 0.6329, 0.6562, 0.5807, 0.6206, 0.5298, 0.5347, 0.6366, 0.7747, 0.6882, 0.6906, 0.7176, 0.7379, 0.7379, 0.7452, 0.7244, 0.7244, 0.7244, 0.7244, 0.9141, 0.9644, 1.0688, 1.1872, 1.1872, 1.2947, 1.1117, 1.0491, 1.0012, 0.9533, 1.1252, 1.2112, 1.388, 1.3413, 1.345, 1.2713, 1.2983, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2010865597069099153", "ticker_idx": 0, "ts": "2026-01-13T00:04:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This would propel FCF to Won105tn in 2026E, and could pave the way for Won45tn of exceptional dividend.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS thesis: Samsung back to peak DRAM profitability, FCF surging, $30.7B special dividend possible — implicitly bullish.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "\"This would propel FCF to Won105tn in 2026E, and could pave the way for Won45tn of exceptional dividend.\"\n\nUBS expects Samsung to pay $30.7 billion in special dividends this year alone (paid separately from its regular recurring dividends).\n\nThat’s insane.", "reply_text": null, "parent_text": null, "quoted_text": "UBS: Samsung Electronics\nBack to peak DRAM profitability in earnest\n\nDDR and NAND contract pricing off to a roaring start \n4Q25 DRAM blended ASPs ended up stronger than we initially expected post prelims (link), as Samsung successfully raised server DRAM contract pricing by about", "parent_author": null, "ret_m1d": 0.008720888554096806, "ret_signed_m1d": 0.008720888554096806, "alpha_spy_m1d": 0.006717407537809361, "alpha_c_m1d": 0.006604551445406948, "bench_spy_m1d": 0.002003481016287445, "bench_c_m1d": 0.0021163371086898586, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019622084588329036, "ret_signed_p1d": 0.019622084588329036, "alpha_spy_p1d": 0.02453732192994762, "alpha_c_p1d": 0.031773891803170584, "bench_spy_p1d": -0.0049152373416185835, "bench_c_p1d": -0.012151807214841548, "ret_p1w": 0.055232521753576735, "ret_signed_p1w": 0.055232521753576735, "alpha_spy_p1w": 0.07856882797715048, "alpha_c_p1w": 0.08690915804272448, "bench_spy_p1w": -0.023336306223573744, "bench_c_p1w": -0.03167663628914774, "ret_p1m": 0.21947677627184592, "ret_signed_p1m": 0.21947677627184592, "alpha_spy_p1m": 0.22208568594450162, "alpha_c_p1m": 0.24343901909063048, "bench_spy_p1m": -0.0026089096726557015, "bench_c_p1m": -0.023962242818784563, "ret_p3m": 0.5001917341529163, "ret_signed_p3m": 0.5001917341529163, "alpha_spy_p3m": 0.5181434921720011, "alpha_c_p3m": 0.5253328719665026, "bench_spy_p3m": -0.017951758019084774, "bench_c_p3m": -0.02514113781358629, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2934, -0.292, -0.2905, -0.2949, -0.2869, -0.3021, -0.2855, -0.2623, -0.2804, -0.2732, -0.2471, -0.2225, -0.1965, -0.2413, -0.2724, -0.2826, -0.292, -0.2724, -0.2515, -0.2544, -0.2565, -0.297, -0.2724, -0.2927, -0.3021, -0.2724, -0.3144, -0.3006, -0.2818, -0.2565, -0.2515, -0.2732, -0.271, -0.2522, -0.2442, -0.2399, -0.216, -0.2081, -0.216, -0.2189, -0.224, -0.2124, -0.2421, -0.2565, -0.2196, -0.2218, -0.2312, -0.2008, -0.1936, -0.1965, -0.1965, -0.1538, -0.1315, -0.1286, -0.1286, -0.1286, -0.0661, 0.0036, 0.0094, 0.0247, 0.0087, 0.0102, 0.0087, 0.0, 0.0196, 0.0458, 0.0821, 0.085, 0.0552, 0.0865, 0.1068, 0.1054, 0.1054, 0.1592, 0.1802, 0.1679, 0.1664, 0.093, 0.2173, 0.2289, 0.1577, 0.1526, 0.2093, 0.2049, 0.2195, 0.298, 0.3169, 0.3169, 0.3169, 0.3169, 0.3808, 0.3815, 0.4026, 0.4535, 0.4789, 0.5843, 0.5734, 0.5734, 0.4179, 0.2515, 0.3924, 0.3677, 0.2609, 0.3656, 0.3808, 0.3656, 0.3336, 0.3714, 0.4092, 0.5153, 0.4571, 0.4491, 0.3539, 0.3786, 0.3735, 0.3089, 0.3089, 0.2839, 0.2176, 0.3811, 0.2992, 0.356, 0.4062, 0.431, 0.533, 0.4856, 0.5002, 0.4638, 0.5038, 0.5366, 0.5839, 0.573, 0.5621, 0.5949, 0.5839, 0.6349, 0.5985, 0.6349, 0.6167, 0.6458, 0.6058, 0.6058, 0.6932, 0.6932, 0.9371, 0.9772, 0.9553, 1.0791, 1.0318, 1.0682, 1.1556, 0.9699, 1.0464, 1.0063, 1.01, 1.1811, 1.1301, 1.1301, 1.1775, 1.2357, 1.1811, 1.3085, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2038529317500850227", "ticker_idx": 0, "ts": "2026-03-30T08:10:37+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "forecasts are gaining traction that Samsung Electronics could become the first Korean company to post quarterly operating profit of KRW 40 trillion", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Samsung HBM revenue tripled YoY in Q1, DRAM sold out, Q1 operating profit forecast at KRW 40T — bullish into next week's earnings.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Samsung HBM Revenue Triples… Operating Profit Nearing $26.33B\n\nSamsung Electronics has confirmed that its high-bandwidth memory (HBM) revenue more than tripled year-over-year in Q1 this year, riding the global wave of artificial intelligence (AI) investment. With conventional DRAM already sold out through next year's volumes and HBM orders surging, forecasts are gaining traction that Samsung Electronics could become the first Korean company to post quarterly operating profit of KRW 40 trillion.\n\nA source with direct knowledge of Samsung Electronics said on the 30th, \"Samsung's Q1 HBM revenue grew by more than 300%,\" adding that \"it was driven by a sharp increase in supply volumes to NVIDIA.\" Samsung Electronics is set to report its Q1 earnings next week.\n\nSamsung Electronics has seen a vertical rise in HBM-related revenue since it began supplying its 5th-generation HBM3E to NVIDIA in the second half of last year. In February this year, it became the first in the world to begin mass production of 6th-generation HBM4, which will be mounted on NVIDIA's next-generation AI accelerator \"Vera Rubin.\"\n\nWith conventional DRAM prices (based on DDR4 8Gb) rising approximately tenfold compared to Q1 last year, combined with the surge in high-margin HBM revenue, the possibility is being raised that Samsung Electronics could exceed KRW 120 trillion in revenue and KRW 40 trillion in operating profit in Q1. An industry source expressed confidence, saying, \"We expect Samsung's Q1 operating profit to come in at around KRW 40 trillion.\"\n\n---\n\nhttps://t.co/TljOqZ5o1A", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01944412932501427, "ret_signed_m1d": 0.01944412932501427, "alpha_spy_m1d": 0.01608945070003953, "alpha_c_m1d": 0.000463751529364842, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.018980377795649428, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.051616562677254674, "ret_signed_p1d": -0.051616562677254674, "alpha_spy_p1d": -0.08068448837616948, "alpha_c_p1d": -0.09396945598301443, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.04235289330575975, "ret_p1w": 0.0952921157118547, "ret_signed_p1w": 0.0952921157118547, "alpha_spy_p1w": 0.052631826105592694, "alpha_c_p1w": 0.02250781156050663, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.07278430415134807, "ret_p1m": 0.259217243335224, "ret_signed_p1m": 0.259217243335224, "alpha_spy_p1m": 0.13307195885345457, "alpha_c_p1m": 0.021177979778238765, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.23803926355698524, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3213, -0.3213, -0.2726, -0.2183, -0.2138, -0.2019, -0.2143, -0.2132, -0.2143, -0.2211, -0.2058, -0.1855, -0.1572, -0.1549, -0.1781, -0.1538, -0.1379, -0.139, -0.139, -0.0972, -0.0807, -0.0904, -0.0915, -0.1487, -0.0519, -0.0428, -0.0983, -0.1023, -0.0581, -0.0615, -0.0502, 0.011, 0.0257, 0.0257, 0.0257, 0.0257, 0.0755, 0.076, 0.0925, 0.1321, 0.1519, 0.234, 0.2255, 0.2255, 0.1044, -0.0253, 0.0845, 0.0653, -0.0179, 0.0636, 0.0755, 0.0636, 0.0387, 0.0681, 0.0976, 0.1802, 0.1349, 0.1287, 0.0545, 0.0738, 0.0698, 0.0194, 0.0194, 0.0, -0.0516, 0.0757, 0.0119, 0.0562, 0.0953, 0.1146, 0.194, 0.1571, 0.1685, 0.1401, 0.1713, 0.1968, 0.2337, 0.2252, 0.2167, 0.2422, 0.2337, 0.2734, 0.245, 0.2734, 0.2592, 0.2819, 0.2507, 0.2507, 0.3188, 0.3188, 0.5088, 0.54, 0.523, 0.6194, 0.5825, 0.6109, 0.679, 0.5343, 0.5939, 0.5627, 0.5655, 0.6988, 0.6591, 0.6591, 0.696, 0.7413, 0.6988, 0.7981, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013859334334357553", "ticker_idx": 0, "ts": "2026-01-21T06:20:55+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Fundamentals and valuation keep us sidelined even amid Trump's blessing. Bumping PT to $36 on higher numbers; MP.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein reiterates Market-Perform on INTC with $36 PT; estimates below consensus, sidelined on fundamentals/valuation despite positive sentiment.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Intel Earnings Preview\n\nFor Intel, we up our server market assumptions but lower our PC assumptions going forward; overall our estimates move up slightly. We now model Q425 at $13.5B/$0.10, up from $13.3B/$0.08 prior, and slightly above consensus at $13.4B/$0.09. For Q126 we now model $12.2B/$0.04, mostly unchanged, and below consensus at $12.5B/$0.07.\n\nFor this year (2026) we now model $52.9B/$0.47, up from $51.9B/$0.41 prior, and below consensus at $54.1B/$0.62. For next year (2027), we now model $56.4B/$0.68, up from $55.9B/$0.64 prior, and below consensus at $56.7B/$1.09.\n\nSentiment on Intel has certainly turned upward in recent weeks amid the 18A launch, foundry hopes, and (yes) Presidential tweets. On the positive market dynamics at least in servers seem supportive, getting 18A out the door helps the narrative, and of course the favor of the administration helps. At the same time, however, share losses seem ripe to continue, supply constraints seem more due to customers opting for older products vs new, and while it feels like beating a dead horse we still believe CPUs have been over-shipping PCs for some time (perhaps this will all come to a head amid potential PC headwinds this year). In the meantime 18A yields (while presumably improving) likely have quite a ways to go, and a bigger decision to go in on foundry is going to have to be accompanied by higher capex. And Street estimates appear to mis-model SCIP-related NCI with overall numbers going forward still looking high to us. Fundamentals and valuation keep us sidelined even amid Trump's blessing. Bumping PT to $36 on higher numbers; MP.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10488476731260798, "ret_signed_m1d": 0.10488476731260798, "alpha_spy_m1d": 0.0934753649270933, "alpha_c_m1d": 0.0761733542863785, "bench_spy_m1d": -0.011409402385514689, "bench_c_m1d": -0.028711413026229482, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0012903169552851423, "ret_signed_p1d": -0.0012903169552851423, "alpha_spy_p1d": 0.003932913064442545, "alpha_c_p1d": 0.0009240354559703068, "bench_spy_p1d": 0.005223230019727687, "bench_c_p1d": 0.002214352411255449, "ret_p1w": -0.10082951558899766, "ret_signed_p1w": 0.10082951558899766, "alpha_spy_p1w": 0.11544866626563344, "alpha_c_p1w": 0.13740307933120166, "bench_spy_p1w": 0.014619150676635773, "bench_c_p1w": 0.036573563742203996, "ret_p1m": -0.1775115404260873, "ret_signed_p1m": 0.1775115404260873, "alpha_spy_p1m": 0.1761691895091445, "alpha_c_p1m": 0.19808731153536374, "bench_spy_p1m": -0.0013423509169427916, "bench_c_p1m": 0.02057577110927644, "ret_p3m": 0.2110598515805011, "ret_signed_p3m": -0.2110598515805011, "alpha_spy_p3m": -0.17421203763553517, "alpha_c_p3m": -0.05672949577315611, "bench_spy_p3m": 0.036847813944965946, "bench_c_p3m": 0.154330355807345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2944, -0.2712, -0.2345, -0.238, -0.2597, -0.2629, -0.2719, -0.3174, -0.2925, -0.3135, -0.2971, -0.2912, -0.3018, -0.3016, -0.3381, -0.3453, -0.3602, -0.3672, -0.3528, -0.3803, -0.3641, -0.3403, -0.3395, -0.3215, -0.3215, -0.2524, -0.2625, -0.1987, -0.1934, -0.2535, -0.2367, -0.2571, -0.2535, -0.2483, -0.2717, -0.303, -0.3086, -0.3123, -0.3355, -0.3312, -0.3213, -0.3296, -0.33, -0.3335, -0.3335, -0.3327, -0.3239, -0.3124, -0.3198, -0.3198, -0.2741, -0.2743, -0.2619, -0.2142, -0.2422, -0.1604, -0.1878, -0.1283, -0.1019, -0.1093, -0.1344, -0.1344, -0.1049, 0.0, 0.0013, -0.1692, -0.2168, -0.1902, -0.1008, -0.103, -0.1434, -0.1003, -0.0922, -0.1041, -0.1108, -0.0675, -0.0739, -0.1312, -0.1099, -0.1432, -0.1375, -0.1375, -0.1488, -0.162, -0.1775, -0.1869, -0.1958, -0.1499, -0.1359, -0.162, -0.1593, -0.1613, -0.2055, -0.1598, -0.153, -0.1996, -0.1598, -0.1377, -0.1156, -0.1659, -0.1563, -0.1565, -0.1878, -0.17, -0.1488, -0.1913, -0.1888, -0.1878, -0.1303, -0.1871, -0.205, -0.2407, -0.1865, -0.1147, -0.0713, -0.0713, -0.064, -0.0247, 0.0866, 0.1377, 0.1499, 0.2015, 0.1762, 0.1971, 0.2627, 0.2627, 0.2111, 0.2214, 0.2031, 0.231, 0.5215, 0.5666, 0.558, 0.7465, 0.7416, 0.8363, 0.7655, 0.9935, 1.0831, 1.0206, 1.3027, 1.386, 1.2232, 1.2173, 1.137, 1.005, 0.9939, 1.0424, 1.1928, 1.1843, 1.209, 1.209, 1.2769, 1.2446, 1.2284, 1.1139, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025807935151231326", "ticker_idx": 0, "ts": "2026-02-23T05:40:24+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "How about we stop with the bullshit about H100 prices dropping already? Even A100 prices are going up?", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish NVDA: dismisses H100 price-drop narrative, noting even A100 prices are rising.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "How about we stop with the bullshit about H100 prices dropping already? Even A100 prices are going up?\n\n$NVDA https://t.co/K4ir0Zvksu", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009031572544070499, "ret_signed_m1d": -0.009031572544070499, "alpha_spy_m1d": -0.01934823393113405, "alpha_c_m1d": -0.014238881861688424, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006786724366676156, "ret_signed_p1d": 0.006786724366676156, "alpha_spy_p1d": -0.00048172856371886397, "alpha_c_p1d": -0.008423503072357397, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": -0.04735061072712243, "ret_signed_p1w": -0.04735061072712243, "alpha_spy_p1w": -0.05319769644001404, "alpha_c_p1w": -0.031631781629992695, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": -0.0853068596486728, "ret_signed_p1m": -0.0853068596486728, "alpha_spy_p1m": -0.04511547732421006, "alpha_c_p1m": -0.041008253931224425, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 0.1460290508639468, "ret_signed_p3m": 0.1460290508639468, "alpha_spy_p3m": 0.05464683402616233, "alpha_c_p3m": -0.22938268660683137, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.059, -0.059, -0.076, -0.0608, -0.0527, -0.0625, -0.0427, -0.0477, -0.0313, -0.0344, -0.0406, -0.0554, -0.0863, -0.0797, -0.0722, -0.1076, -0.0909, -0.0551, -0.041, -0.0122, -0.0153, -0.0153, -0.0053, -0.0174, -0.0209, -0.0264, -0.0264, -0.0141, -0.0179, -0.0225, -0.0127, -0.034, -0.0349, -0.0345, -0.03, -0.0439, -0.0235, -0.0278, -0.0278, -0.0704, -0.043, -0.035, -0.0203, -0.0265, -0.0158, -0.0002, 0.005, -0.0022, -0.031, -0.0585, -0.0906, -0.1027, -0.0321, -0.0079, -0.0157, -0.0078, -0.0241, -0.0456, -0.0456, -0.0344, -0.0186, -0.0191, -0.009, 0.0, 0.0068, 0.0209, -0.0348, -0.075, -0.0474, -0.06, -0.0444, -0.0429, -0.0717, -0.0465, -0.0354, -0.0288, -0.0439, -0.0589, -0.0434, -0.0502, -0.0582, -0.0678, -0.0984, -0.083, -0.0853, -0.0671, -0.106, -0.1254, -0.1377, -0.0895, -0.0824, -0.0739, -0.0739, -0.0726, -0.0702, -0.0494, -0.0398, -0.0152, -0.0116, 0.0259, 0.0383, 0.0356, 0.0529, 0.0549, 0.0435, 0.0572, 0.0423, 0.0873, 0.1309, 0.1129, 0.0925, 0.0419, 0.0361, 0.0362, 0.0259, 0.085, 0.1042, 0.1235, 0.1457, 0.1527, 0.179, 0.2308, 0.1764, 0.1607, 0.1518, 0.1667, 0.146, 0.1242, 0.1242, 0.1218, 0.11, 0.1186, 0.1023, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2004514480408195408", "ticker_idx": 0, "ts": "2025-12-26T11:27:48+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix recently raised HBM prices by more than 50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "HBM3E prices surging 50%+ on AI demand; supply constraints persist into 2026, keeping prices elevated — bullish for Samsung and SK Hynix.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "China Exports & AI Competition Heat Up… HBM Prices Hike by 50%\n\nAs the competition for artificial intelligence (AI) supremacy among global big tech companies intensifies, the prices of High Bandwidth Memory (HBM) semiconductors, a key component, are soaring. Although Samsung Electronics and SK Hynix recently raised HBM prices by more than 50%, selling prices continue to skyrocket as big tech firms keep demanding larger volumes.\n\nAccording to the semiconductor industry on the 26th, Samsung Electronics and SK Hynix are proposing prices that are more than 50% higher when renewing contracts for HBM3E 12-layer products with existing clients. It is also reported that new customers must offer even higher prices to secure volume.\n\nWhile the HBM3E 12-layer was known to be around the $300 range per chip, companies that recently renewed contracts are understood to be paying in the $500 range. This implies that the market price of HBM3E has effectively jumped to the level of the next-generation product, HBM4.\n\nHBM, made by stacking DRAM, has seen its value fly high as global prices for general-purpose DRAM have risen since the second half of this year. Particularly, with the AI performance competition among big tech firms catching fire recently, industry insiders are even going as far as to say that prices are \"skyrocketing.\"\n\nOpenAI's GPT and Google's Gemini, which are at the center of the AI performance competition, released their latest versions this month, each claiming superior performance. Additionally, xAI, created by Tesla CEO Elon Musk, recently unveiled the 'Grok Voice Agent' surpassing the performance of Google and OpenAI, while Anthropic's Claude and China's DeepSeek AI also released new versions to advance their capabilities. To improve AI chip performance, high-performance, low-power memory—HBM3E—is essential for all of them.\n\nThe United States is also fueling the rise in HBM prices. As the Donald Trump administration allowed the export of Nvidia's AI chip 'H200' to China this month, the prices of even the HBM3E 8-layer products equipped in the H200 are rising.\n\nThe trend of high HBM prices is highly likely to continue into next year. Samsung Electronics is converting NAND flash production lines to DRAM lines, but with HBM4 mass production scheduled for next February, expanding supply volume is a challenging situation. SK Hynix will also be able to supply new volume only after its M15X plant in Cheongju begins mass production in the second half of next year. An industry official predicted, \"Due to the shortage of supply compared to increasing demand, the rise in HBM prices will continue for the time being next year as well.\"\n\n---\n\nhttps://t.co/cIkhqYNewi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0504273985851188, "ret_signed_m1d": -0.0504273985851188, "alpha_spy_m1d": -0.050528823903413955, "alpha_c_m1d": -0.04885781272577994, "bench_spy_m1d": 0.00010142531829515633, "bench_c_m1d": -0.0015695858593388623, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026332556253146677, "ret_signed_p1d": 0.026332556253146677, "alpha_spy_p1d": 0.02989618446020148, "alpha_c_p1d": 0.03083674843618367, "bench_spy_p1d": -0.003563628207054803, "bench_c_p1d": -0.004504192183036992, "ret_p1w": 0.10362953284822307, "ret_signed_p1w": 0.10362953284822307, "alpha_spy_p1w": 0.11397269885194583, "alpha_c_p1w": 0.11884828071407527, "bench_spy_p1w": -0.010343166003722759, "bench_c_p1w": -0.0152187478658522, "ret_p1m": 0.3063194423526272, "ret_signed_p1m": 0.3063194423526272, "alpha_spy_p1m": 0.30281379680131093, "alpha_c_p1m": 0.3093904256063118, "bench_spy_p1m": 0.003505645551316272, "bench_c_p1m": -0.0030709832536846005, "ret_p3m": 0.6232371808663997, "ret_signed_p3m": 0.6232371808663997, "alpha_spy_p3m": 0.6691530607997462, "alpha_c_p3m": 0.6887525329154162, "bench_spy_p3m": -0.04591587993334656, "bench_c_p3m": -0.06551535204901648, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2829, -0.265, -0.2329, -0.2329, -0.2329, -0.2329, -0.2329, -0.2329, -0.1932, -0.2026, -0.2171, -0.188, -0.165, -0.1632, -0.1615, -0.1667, -0.1573, -0.1752, -0.1556, -0.1282, -0.1496, -0.141, -0.1103, -0.0812, -0.0504, -0.1034, -0.1402, -0.1521, -0.1632, -0.1402, -0.1154, -0.1188, -0.1214, -0.1692, -0.1402, -0.1641, -0.1752, -0.1402, -0.1897, -0.1735, -0.1513, -0.1214, -0.1154, -0.141, -0.1385, -0.1162, -0.1068, -0.1017, -0.0735, -0.0641, -0.0735, -0.0769, -0.0829, -0.0692, -0.1043, -0.1214, -0.0778, -0.0803, -0.0915, -0.0556, -0.047, -0.0504, -0.0504, 0.0, 0.0263, 0.0298, 0.0298, 0.0298, 0.1036, 0.1861, 0.193, 0.211, 0.1921, 0.1938, 0.1921, 0.1818, 0.205, 0.2359, 0.2788, 0.2823, 0.2471, 0.284, 0.308, 0.3063, 0.3063, 0.3699, 0.3948, 0.3802, 0.3785, 0.2917, 0.4386, 0.4523, 0.3682, 0.3621, 0.4291, 0.424, 0.4412, 0.5339, 0.5562, 0.5562, 0.5562, 0.5562, 0.6318, 0.6327, 0.6576, 0.7177, 0.7478, 0.8723, 0.8594, 0.8594, 0.6756, 0.4789, 0.6456, 0.6164, 0.4901, 0.6138, 0.6318, 0.6138, 0.576, 0.6207, 0.6653, 0.7907, 0.722, 0.7126, 0.6, 0.6292, 0.6232, 0.5468, 0.5468, 0.5173, 0.439, 0.6322, 0.5354, 0.6025, 0.6619, 0.6911, 0.8116, 0.7557, 0.7729, 0.7299, 0.7772, 0.8159, 0.8719, 0.859, 0.8461, 0.8848, 0.8719, 0.9321, 0.8891, 0.9321, 0.9106, 0.945, 0.8977, 0.8977, 1.001, 1.001, 1.2893, 1.3366, 1.3108, 1.4571, 1.4012, 1.4442, 1.5475, 1.328, 1.4184, 1.371, 1.3753, 1.5776, 1.5174, 1.5174, 1.5733, 1.6421, 1.5776, 1.7282, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2004514480408195408", "ticker_idx": 1, "ts": "2025-12-26T11:27:48+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix recently raised HBM prices by more than 50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "HBM3E prices surging 50%+ on AI demand; supply constraints persist into 2026, keeping prices elevated — bullish for Samsung and SK Hynix.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "China Exports & AI Competition Heat Up… HBM Prices Hike by 50%\n\nAs the competition for artificial intelligence (AI) supremacy among global big tech companies intensifies, the prices of High Bandwidth Memory (HBM) semiconductors, a key component, are soaring. Although Samsung Electronics and SK Hynix recently raised HBM prices by more than 50%, selling prices continue to skyrocket as big tech firms keep demanding larger volumes.\n\nAccording to the semiconductor industry on the 26th, Samsung Electronics and SK Hynix are proposing prices that are more than 50% higher when renewing contracts for HBM3E 12-layer products with existing clients. It is also reported that new customers must offer even higher prices to secure volume.\n\nWhile the HBM3E 12-layer was known to be around the $300 range per chip, companies that recently renewed contracts are understood to be paying in the $500 range. This implies that the market price of HBM3E has effectively jumped to the level of the next-generation product, HBM4.\n\nHBM, made by stacking DRAM, has seen its value fly high as global prices for general-purpose DRAM have risen since the second half of this year. Particularly, with the AI performance competition among big tech firms catching fire recently, industry insiders are even going as far as to say that prices are \"skyrocketing.\"\n\nOpenAI's GPT and Google's Gemini, which are at the center of the AI performance competition, released their latest versions this month, each claiming superior performance. Additionally, xAI, created by Tesla CEO Elon Musk, recently unveiled the 'Grok Voice Agent' surpassing the performance of Google and OpenAI, while Anthropic's Claude and China's DeepSeek AI also released new versions to advance their capabilities. To improve AI chip performance, high-performance, low-power memory—HBM3E—is essential for all of them.\n\nThe United States is also fueling the rise in HBM prices. As the Donald Trump administration allowed the export of Nvidia's AI chip 'H200' to China this month, the prices of even the HBM3E 8-layer products equipped in the H200 are rising.\n\nThe trend of high HBM prices is highly likely to continue into next year. Samsung Electronics is converting NAND flash production lines to DRAM lines, but with HBM4 mass production scheduled for next February, expanding supply volume is a challenging situation. SK Hynix will also be able to supply new volume only after its M15X plant in Cheongju begins mass production in the second half of next year. An industry official predicted, \"Due to the shortage of supply compared to increasing demand, the rise in HBM prices will continue for the time being next year as well.\"\n\n---\n\nhttps://t.co/cIkhqYNewi", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01836399470597716, "ret_signed_m1d": -0.01836399470597716, "alpha_spy_m1d": -0.018465420024272317, "alpha_c_m1d": -0.01374481799496241, "bench_spy_m1d": 0.00010142531829515633, "bench_c_m1d": -0.00461917671101475, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06844737901642794, "ret_signed_p1d": 0.06844737901642794, "alpha_spy_p1d": 0.07201100722348275, "alpha_c_p1d": 0.07295728434236526, "bench_spy_p1d": -0.003563628207054803, "bench_c_p1d": -0.004509905325937313, "ret_p1w": 0.1302169037577463, "ret_signed_p1w": 0.1302169037577463, "alpha_spy_p1w": 0.14056006976146906, "alpha_c_p1w": 0.10988124875503535, "bench_spy_p1w": -0.010343166003722759, "bench_c_p1w": 0.02033565500271095, "ret_p1m": 0.22871441791101899, "ret_signed_p1m": 0.22871441791101899, "alpha_spy_p1m": 0.2252087723597027, "alpha_c_p1m": 0.138625249124265, "bench_spy_p1m": 0.003505645551316272, "bench_c_p1m": 0.09008916878675399, "ret_p3m": 0.6641668275874255, "ret_signed_p3m": 0.6641668275874255, "alpha_spy_p3m": 0.7100827075207721, "alpha_c_p3m": 0.5735310516354715, "bench_spy_p3m": -0.04591587993334656, "bench_c_p3m": 0.09063577595195405, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4203, -0.3994, -0.3402, -0.3402, -0.3402, -0.3402, -0.3402, -0.3402, -0.286, -0.3077, -0.3135, -0.2952, -0.2451, -0.2234, -0.1901, -0.2009, -0.1967, -0.2017, -0.1492, -0.1075, -0.1308, -0.0691, -0.0524, -0.0674, 0.0343, -0.0224, -0.0341, -0.0107, -0.0324, 0.011, 0.0326, 0.0293, 0.021, -0.0658, 0.011, -0.0491, -0.0624, -0.0474, -0.1308, -0.1325, -0.1342, -0.1258, -0.0918, -0.1152, -0.1018, -0.0684, -0.0785, -0.0952, -0.0918, -0.0367, -0.0551, -0.02, -0.0568, -0.0467, -0.0751, -0.1152, -0.0801, -0.0785, -0.0868, -0.0317, -0.025, -0.0184, -0.0184, 0.0, 0.0684, 0.0868, 0.0868, 0.0868, 0.1302, 0.1619, 0.212, 0.2387, 0.2621, 0.2421, 0.2504, 0.2321, 0.2387, 0.2504, 0.2621, 0.2755, 0.2404, 0.2354, 0.2604, 0.2805, 0.2287, 0.3356, 0.404, 0.4374, 0.5175, 0.3856, 0.5142, 0.5025, 0.4057, 0.4007, 0.4808, 0.4624, 0.4357, 0.4825, 0.4691, 0.4691, 0.4691, 0.4691, 0.4925, 0.5843, 0.5876, 0.6778, 0.6995, 0.8381, 0.7746, 0.7746, 0.5705, 0.42, 0.5739, 0.5454, 0.3982, 0.5688, 0.5973, 0.5555, 0.522, 0.629, 0.6224, 0.7662, 0.6943, 0.6842, 0.5605, 0.6491, 0.6642, 0.5605, 0.5605, 0.4601, 0.3497, 0.5003, 0.3882, 0.4651, 0.4819, 0.532, 0.7277, 0.6692, 0.7177, 0.7394, 0.8448, 0.9, 0.9318, 0.8866, 0.9502, 1.0472, 1.0455, 1.0488, 1.0438, 1.1609, 1.1743, 1.1626, 1.1509, 1.1509, 1.4202, 1.4202, 1.6777, 1.7664, 1.8199, 2.1444, 2.0691, 2.3049, 2.2949, 2.0423, 2.0775, 1.9186, 1.9186, 2.2447, 2.2464, 2.2464, 2.432, 2.7515, 2.8291, 2.9027, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021412576635322716", "ticker_idx": 0, "ts": "2026-02-11T02:34:48+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM being in short supply may become the new normal", "tweet_type": "analysis", "timeline": "years", "summary": "DRAM supply shortage expected through at least 2028-2030 as CAPA grows only 4.8%/yr while AI demand surges; shortage may become structural.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Someone asked me today how long the DRAM supply shortage would last... I answered 2028, but I may need to revise that answer.\n\nAccording to SEMI's data, while AI infrastructure investment is expanding rapidly each year, DRAM production capacity (CAPA) is expected to grow at a relatively low annual rate of just 4.8% from last year through 2030.\n\nWhat's more, even this incremental capacity will be concentrated on HBM, meaning supply for legacy and special-purpose DRAM will be even more constrained.\n\nPerhaps... DRAM being in short supply may become the new normal.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027900931364043924, "ret_signed_m1d": -0.027900931364043924, "alpha_spy_m1d": -0.028132043578974475, "alpha_c_m1d": -0.0037435731424304564, "bench_spy_m1d": 0.00023111221493055112, "bench_c_m1d": -0.024157358221613467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03525559002233739, "ret_signed_p1d": 0.03525559002233739, "alpha_spy_p1d": 0.05070446425023859, "alpha_c_p1d": 0.056158269643462634, "bench_spy_p1d": -0.0154488742279012, "bench_c_p1d": -0.020902679621125242, "ret_p1w": 0.04298978625287564, "ret_signed_p1w": 0.04298978625287564, "alpha_spy_p1w": 0.051183965798321296, "alpha_c_p1w": 0.04836615684184109, "bench_spy_p1w": -0.008194179545445657, "bench_c_p1w": -0.005376370588965451, "ret_p1m": 0.06367185281004666, "ret_signed_p1m": 0.06367185281004666, "alpha_spy_p1m": 0.10110186021483607, "alpha_c_p1m": 0.12792276648697337, "bench_spy_p1m": -0.03743000740478941, "bench_c_p1m": -0.0642509136769267, "ret_p3m": 0.9446894417632201, "ret_signed_p3m": 0.9446894417632201, "alpha_spy_p3m": 0.873357315555608, "alpha_c_p3m": 0.5552540802349839, "bench_spy_p3m": 0.07133212620761209, "bench_c_p3m": 0.3894353615282362, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3905, -0.3699, -0.4003, -0.4081, -0.4165, -0.4424, -0.4256, -0.4204, -0.4068, -0.3976, -0.4039, -0.3969, -0.3848, -0.3893, -0.3981, -0.3822, -0.3594, -0.3614, -0.3448, -0.359, -0.3676, -0.3852, -0.4026, -0.39, -0.3715, -0.3619, -0.3322, -0.3289, -0.3197, -0.3197, -0.3053, -0.2754, -0.2718, -0.2776, -0.2776, -0.2261, -0.2023, -0.1637, -0.1565, -0.1656, -0.1552, -0.153, -0.1659, -0.1629, -0.1504, -0.1165, -0.1126, -0.1271, -0.1001, -0.0819, -0.0759, -0.0965, -0.0398, 0.0022, 0.007, 0.0082, -0.0239, 0.025, -0.007, -0.0462, -0.0391, -0.0141, -0.0279, 0.0, 0.0353, 0.0354, 0.0354, 0.0258, 0.043, 0.063, 0.0933, 0.094, 0.1264, 0.1473, 0.1974, 0.1771, 0.1773, 0.0606, -0.0027, 0.0686, 0.0335, -0.0145, 0.065, 0.0884, 0.0637, 0.064, 0.112, 0.1369, 0.1993, 0.1525, 0.1307, 0.0608, 0.081, 0.0722, 0.0088, 0.0102, -0.0485, -0.0792, 0.0248, -0.0249, 0.0084, 0.0354, 0.0537, 0.1507, 0.1362, 0.1507, 0.1506, 0.2178, 0.2319, 0.253, 0.2378, 0.2442, 0.2764, 0.304, 0.3141, 0.3151, 0.3749, 0.3565, 0.3733, 0.3586, 0.3789, 0.4932, 0.545, 0.6929, 0.7082, 0.7961, 0.9447, 0.8909, 0.9857, 0.9848, 0.8337, 0.8277, 0.7939, 0.8219, 0.9689, 0.9463, 0.9463, 1.1201, 1.2366, 1.2357, 1.3262, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021359981476577478", "ticker_idx": 0, "ts": "2026-02-10T23:05:49+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Short-term share price correction is a buying opportunity… We maintain a positive outlook on Micron… Micron's earnings are expected to continue trending upward", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Korea Investment & Securities Buy thesis on MU: HBM4 delay immaterial, commodity DRAM/NAND ASP upside drives 2026 earnings higher; dip is a buying opportunity.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Micron Technology: It's Okay If HBM4 Is a Bit Delayed – Chae Min-suk, Korea Investment & Securities\n\nMicron's HBM4 schedule delay is inevitable\n- Micron's stock weakened after SemiAnalysis reported it would adjust Micron's share in NVIDIA's HBM4 to 0%\n- However, we had already updated in our August 2025 report that Micron's HBM4 schedule would be delayed due to HBM4 speed upgrade issues\n- Our HBM model already adjusted Micron's 2026 HBM4 market share to around 1%, factoring in redesign and sample schedule delays\n\nImpact on earnings expected to be minimal\n- The delay in Micron's HBM4 is not expected to significantly impact earnings\n- This is because most of the upside in the 2026 memory market will be driven by commodity DRAM and NAND ASP increases\n- As of Q4 2025, commodity DRAM operating margins likely already exceeded HBM operating margins\n- Unlike HBM, where prices are determined on an annual basis, commodity DRAM ASPs are rising significantly on a quarterly basis, meaning the share of commodity DRAM in revenue and operating profit will expand considerably in 2026 compared to 2025\n- According to our HBM model, HBM is estimated to account for 9% of Micron's DRAM by volume, 11% by revenue, and approximately 8% of DRAM operating profit in 2026\n- Additionally, the current tight supply environment will also work in Micron's favor\n- Even if Micron cannot supply HBM4, Samsung Electronics and SK hynix have limited additional supply capacity due to constrained capacity\n- While market entry for HBM4 may be limited in 2026, full-scale supply should be possible from 2027 onward following the submission of improved samples\n\nShort-term share price correction is a buying opportunity\n- We maintain a positive outlook on Micron\n- HBM remains a strategically important product, but its contribution to earnings in 2026 is relatively less significant compared to 2025\n- In particular, the current tight supply environment limits competitors' ability to expand supply further, serving as a floor for Micron's earnings\n- Data center memory demand will translate into rising commodity DRAM ASPs, and based on this, Micron's earnings are expected to continue trending upward\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.027461490390592758, "ret_signed_m1d": 0.027461490390592758, "alpha_spy_m1d": 0.024817356954436987, "alpha_c_m1d": 0.02279209597103593, "bench_spy_m1d": 0.002644133436155771, "bench_c_m1d": 0.004669394419556827, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09937041532793689, "ret_signed_p1d": 0.09937041532793689, "alpha_spy_p1d": 0.09960147414235321, "alpha_c_p1d": 0.07461503245433265, "bench_spy_p1d": -0.00023105881441631837, "bench_c_p1d": 0.024755382873604237, "ret_p1w": 0.07107836721873584, "ret_signed_p1w": 0.07107836721873584, "alpha_spy_p1w": 0.08447199691042129, "alpha_c_p1w": 0.06428421776313042, "bench_spy_p1w": -0.01339362969168545, "bench_c_p1w": 0.006794149455605414, "ret_p1m": 0.12174148829661235, "ret_signed_p1m": 0.12174148829661235, "alpha_spy_p1m": 0.14455538496778264, "alpha_c_p1m": 0.13095685270103552, "bench_spy_p1m": -0.022813896671170286, "bench_c_p1m": -0.009215364404423165, "ret_p3m": 1.0016710945628486, "ret_signed_p3m": 1.0016710945628486, "alpha_spy_p3m": 0.9330204523953136, "alpha_c_p3m": 0.6019775370016733, "bench_spy_p3m": 0.06865064216753503, "bench_c_p3m": 0.3996935575611753, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3654, -0.339, -0.352, -0.3881, -0.395, -0.4607, -0.4446, -0.4003, -0.3987, -0.3833, -0.3833, -0.3667, -0.356, -0.3586, -0.3729, -0.393, -0.3647, -0.3387, -0.324, -0.2938, -0.3078, -0.3542, -0.364, -0.3773, -0.396, -0.3344, -0.2878, -0.2593, -0.2601, -0.2322, -0.2322, -0.2373, -0.2113, -0.216, -0.2353, -0.2353, -0.1549, -0.1637, -0.0799, -0.0903, -0.1239, -0.0754, -0.0734, -0.0941, -0.1069, -0.0981, -0.0281, -0.0281, -0.0221, 0.0425, 0.0652, 0.0707, 0.0424, 0.0991, 0.1662, 0.1676, 0.1115, 0.1729, 0.1238, 0.0165, 0.0258, 0.0574, 0.0275, 0.0, 0.0994, 0.1091, 0.1029, 0.1029, 0.0711, 0.1278, 0.1182, 0.1471, 0.1279, 0.1199, 0.1494, 0.1134, 0.1048, 0.1056, 0.0172, 0.0737, 0.0638, -0.0079, 0.0431, 0.08, 0.1217, 0.086, 0.1417, 0.1837, 0.2369, 0.2371, 0.1903, 0.133, 0.0833, 0.0597, 0.0237, -0.0477, -0.0429, -0.1375, -0.0945, -0.0141, -0.0184, -0.0184, 0.0125, 0.012, 0.0902, 0.1298, 0.1273, 0.1433, 0.2481, 0.2228, 0.2255, 0.2197, 0.2019, 0.2045, 0.3066, 0.2912, 0.3314, 0.406, 0.3516, 0.3896, 0.3861, 0.4533, 0.5451, 0.7159, 0.7867, 0.7332, 1.0017, 1.1317, 1.0547, 1.154, 1.0799, 0.9423, 0.8267, 0.8728, 0.9619, 1.0427, 1.0129, 1.0129, 1.4012, 1.4884, 1.4753, 1.6026, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015705362075398598", "ticker_idx": 0, "ts": "2026-01-26T08:36:22+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "foundry", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "There are growing forecasts that Samsung Foundry will post better results this year than in previous years", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung foundry expected to outperform YoY as Chinese fabless clients return with non-regulated chip orders and diversify away from TSMC.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Signs of Thawing U.S.-China Tensions: Chinese Fabless Firms Increase Orders to Samsung Foundry\n\n\"Last year, many Chinese clients utilizing Samsung Electronics’ foundry services frequently abandoned their projects just prior to mass production. This was primarily due to the high level of uncertainty caused by peak U.S. regulatory pressure on China. However, we expect a different landscape this year.\"\n\nAn industry official shared this insight on the 26th, identifying the U.S.-China conflict as the primary factor behind the slump in Samsung's foundry performance last year.\n\nAccording to industry sources, Samsung Foundry faced significant hurdles in securing Chinese clients during the first half of last year as U.S. semiconductor export controls tightened. During this period, extreme regulatory uncertainty led a succession of Chinese fabless companies to halt projects or defer decisions immediately before the mass production phase.\n\n\"The first half of last year was exceptionally challenging,\" another industry representative noted. \"Many cases resulted in the forced abandonment of mass production due to U.S. sanctions.\"\n\nAt that time, the Bureau of Industry and Security (BIS) under the U.S. Department of Commerce had intensified export controls on China, focusing on chips for AI and High-Performance Computing (HPC). Consequently, concerns arose that certain chips slated for production at Samsung Foundry might fall under these regulations. These products were reportedly advanced-process chips equipped with High Bandwidth Memory (HBM).\n\nThis situation weighed heavily on Samsung Foundry's financial performance. Industry insiders suggest that had these projects transitioned to mass production, the foundry division's revenue might have significantly improved, potentially outperforming the System LSI division.\n\nA Turnaround in Late 2023: Re-approaching via 'Non-Regulated' Chips\nThe industry consensus is that the atmosphere began to shift in the second half of last year. As U.S. regulatory criteria became more specific, inquiries and orders from Chinese customers began to rebound, focusing on chip specifications that do not directly fall under the scope of the restrictions.\n\nSpecifically, there is a visible move toward re-evaluating Samsung Foundry for general-purpose and low-power chipsets, which carry less regulatory burden compared to high-performance AI accelerators or HBM-integrated chips.\n\n\"Starting from the latter half of last year, Chinese clients have been knocking on our doors again with designs and applications that can bypass regulations,\" said an industry source. \"Compared to the first half, the sales environment is showing signs of stabilization.\"\n\nHeightened Expectations for Samsung Foundry Amid TSMC’s Geopolitical Risks\nThere are growing forecasts that Samsung Foundry will post better results this year than in previous years. This is largely driven by Chinese fabless firms seeking Samsung as a viable alternative due to the geopolitical risks associated with relying on Taiwan's TSMC.\n\nAs the U.S.-China rivalry persists alongside ongoing geopolitical instability in the Taiwan Strait, Chinese companies are feeling the pressure of over-reliance on a single production region. In this context, Samsung Foundry is being re-evaluated as a practical and strategic alternative.\n\nA representative from a domestic design house stated, \"Recently, we’ve seen a surge of interest in Samsung Foundry from Chinese clients. We anticipate a much stronger momentum this year compared to last.\"\n\nHowever, the U.S.-China technological hegemony race and export restrictions remain critical variables. Experts believe Samsung Foundry's success this year will depend on its ability to manage risks in the Chinese market while simultaneously expanding its non-Chinese customer base.\n\n---\n\nhttps://t.co/n9IlxYI2It", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0050524897745126696, "alpha_c_m1d": 0.006777205222007088, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048652228216711224, "ret_signed_p1d": 0.048652228216711224, "alpha_spy_p1d": 0.04466796348702662, "alpha_c_p1d": 0.03516653004304571, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": -0.011176829852671766, "ret_signed_p1w": -0.011176829852671766, "alpha_spy_p1w": -0.015045622732750008, "alpha_c_p1w": -0.005563340661362881, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.31492441086551537, "ret_signed_p1m": 0.31492441086551537, "alpha_spy_p1m": 0.3226907983264917, "alpha_c_p1m": 0.3543549001418518, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": 0.4790577142361261, "ret_signed_p3m": 0.4790577142361261, "alpha_spy_p3m": 0.453571831310269, "alpha_c_p3m": 0.41091867364976475, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3424, -0.3189, -0.2966, -0.2731, -0.3137, -0.3418, -0.351, -0.3595, -0.3418, -0.3228, -0.3254, -0.3274, -0.364, -0.3418, -0.3601, -0.3686, -0.3418, -0.3797, -0.3673, -0.3503, -0.3274, -0.3228, -0.3424, -0.3405, -0.3235, -0.3163, -0.3124, -0.2908, -0.2836, -0.2908, -0.2934, -0.298, -0.2875, -0.3143, -0.3274, -0.294, -0.296, -0.3045, -0.277, -0.2705, -0.2731, -0.2731, -0.2345, -0.2143, -0.2117, -0.2117, -0.2117, -0.1552, -0.092, -0.0868, -0.073, -0.0874, -0.0861, -0.0874, -0.0953, -0.0776, -0.0539, -0.021, -0.0184, -0.0454, -0.0171, 0.0013, 0.0, 0.0, 0.0487, 0.0677, 0.0565, 0.0552, -0.0112, 0.1012, 0.1118, 0.0473, 0.0427, 0.094, 0.0901, 0.1032, 0.1742, 0.1913, 0.1913, 0.1913, 0.1913, 0.2492, 0.2498, 0.2689, 0.3149, 0.3379, 0.4333, 0.4234, 0.4234, 0.2827, 0.1321, 0.2597, 0.2373, 0.1407, 0.2354, 0.2492, 0.2354, 0.2064, 0.2406, 0.2748, 0.3708, 0.3182, 0.311, 0.2249, 0.2472, 0.2426, 0.1841, 0.1841, 0.1615, 0.1016, 0.2495, 0.1753, 0.2267, 0.2722, 0.2946, 0.3868, 0.344, 0.3572, 0.3242, 0.3605, 0.3901, 0.4329, 0.4231, 0.4132, 0.4428, 0.4329, 0.4791, 0.4461, 0.4791, 0.4626, 0.4889, 0.4527, 0.4527, 0.5318, 0.5318, 0.7525, 0.7887, 0.7689, 0.8809, 0.8381, 0.8711, 0.9501, 0.7821, 0.8513, 0.8151, 0.8184, 0.9732, 0.9271, 0.9271, 0.9699, 1.0226, 0.9732, 1.0885, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2016479977659838801", "ticker_idx": 0, "ts": "2026-01-28T11:54:25+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We continue to be optimistic about the super cycle of global semiconductor equipment, and ASML is also expected to return to the 40-50x forward PE of the 2020-2022 shortage cycle", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Reiterates bullish ASML thesis: record orders, fully booked 26/27 capacity, targets 40-50x forward PE as equipment market turns into seller's market.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics Communication》- ASML Performance Commentary  \n\n🌟Event: ASML released its 4Q performance, with 4Q25 revenue of €9.7bn and a gross margin of 52.2%. On the other hand, the company's orders greatly exceeded expectations, with new orders amounting to €13.2bn (vs. market expectation of 8bn+, Guangfa forecast of 10bn🌹). Among the new orders, €7.4bn was for EUV. The company expects full-year revenue for 2026 to be €34-39bn. Additionally, ASML announced a €12 euro stock repurchase plan, which will be executed by December 31, 2028.  \n\n🌟Commentary: The company's orders greatly exceeded expectations, with TSMC, storage, and China all showing significant upward revisions. We see that ASML's 26/27 capacity has been fully booked, and the semiconductor equipment market has turned into a seller's market. We continue to be optimistic about the super cycle of global semiconductor equipment, and ASML is also expected to return to the 40-50x forward PE of the 2020-2022 shortage cycle.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.019423882595973607, "ret_signed_m1d": 0.019423882595973607, "alpha_spy_m1d": 0.01932320255687392, "alpha_c_m1d": 0.04193787528224413, "bench_spy_m1d": 0.00010068003909968759, "bench_c_m1d": -0.022513992686270523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0020094661940031067, "ret_signed_p1d": -0.0020094661940031067, "alpha_spy_p1d": -2.5083745033294136e-05, "alpha_c_p1d": -0.004145616064221258, "bench_spy_p1d": -0.0019843824489698125, "bench_c_p1d": 0.0021361498702181514, "ret_p1w": -0.04504355091302403, "ret_signed_p1w": -0.04504355091302403, "alpha_spy_p1w": -0.03177103512918378, "alpha_c_p1w": 0.03802779647799148, "bench_spy_p1w": -0.013272515783840255, "bench_c_p1w": -0.08307134739101552, "ret_p1m": 0.0331998851674995, "ret_signed_p1m": 0.0331998851674995, "alpha_spy_p1m": 0.04200032386492858, "alpha_c_p1m": 0.04428884914763098, "bench_spy_p1m": -0.00880043869742908, "bench_c_p1m": -0.011088963980131483, "ret_p3m": 0.0188480790428327, "ret_signed_p3m": 0.0188480790428327, "alpha_spy_p3m": -0.012360620909441122, "alpha_c_p3m": -0.1962828616776351, "bench_spy_p3m": 0.031208699952273822, "bench_c_p3m": 0.2151309407204678, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2313, -0.2243, -0.2314, -0.241, -0.2504, -0.2709, -0.2576, -0.2575, -0.2533, -0.262, -0.2682, -0.2662, -0.2753, -0.2577, -0.255, -0.3017, -0.281, -0.2797, -0.2387, -0.2488, -0.2436, -0.2241, -0.2137, -0.1934, -0.1985, -0.2033, -0.1936, -0.2022, -0.208, -0.2122, -0.2253, -0.2205, -0.2393, -0.2683, -0.2526, -0.2451, -0.2498, -0.2438, -0.2473, -0.2473, -0.2473, -0.2408, -0.2311, -0.2286, -0.2286, -0.1742, -0.1182, -0.1117, -0.1194, -0.1519, -0.0944, -0.0904, -0.0769, -0.0923, -0.0377, -0.0228, -0.062, -0.0455, -0.0332, -0.0151, -0.0136, -0.0137, 0.0194, 0.0, -0.002, 0.0177, 0.0255, -0.0033, -0.045, -0.0375, -0.0005, 0.0101, 0.0002, 0.0126, -0.0109, -0.002, 0.0018, 0.0054, 0.0436, 0.0381, 0.0526, 0.0473, 0.0592, 0.0801, 0.0332, 0.034, 0.0148, -0.026, 0.0059, -0.0057, -0.0384, -0.0379, 0.006, 0.005, -0.0104, -0.0111, 0.0039, 0.0008, -0.0015, -0.0203, -0.0542, -0.0146, 0.0074, 0.0158, -0.0226, -0.0389, -0.0677, -0.0617, -0.0044, -0.0267, -0.0267, -0.0267, -0.0662, 0.0166, 0.0352, 0.0647, 0.0562, 0.0766, 0.0312, 0.025, 0.0429, 0.0438, 0.0371, 0.0478, 0.026, 0.0499, 0.0188, -0.0154, 0.002, 0.0271, 0.0271, -0.0023, 0.0326, 0.0965, 0.0923, 0.1086, 0.097, 0.0635, 0.115, 0.1486, 0.0978, 0.0624, 0.0494, 0.12, 0.1303, 0.1839, 0.202, 0.1677, 0.1565, 0.1687, 0.1635, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2018615701465186799", "ticker_idx": 0, "ts": "2026-02-03T09:21:01+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley projected this year's operating profits for Samsung and SK Hynix at 245 trillion KRW and 179 trillion KRW, respectively. This represents a 4 to 5-fold increase compared to last year's operating profits", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish thesis on Samsung and SK Hynix: NVIDIA skip-qual HBM4 demand, Morgan Stanley projects 4-5x profit growth, Korean memory firms now hold AI supply-chain bottleneck.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] NVIDIA to Samsung: Skip HBM4 Qual Tests, ‘Supply First’ Requested\n\nNVIDIA, the undisputed leader in Artificial Intelligence (AI) semiconductors, has reportedly requested that Samsung Electronics accelerate the supply of 6th-generation High Bandwidth Memory (HBM4). Samsung is currently undergoing final quality testing for mass production shipments scheduled for February, yet NVIDIA has requested supply regardless of the test results. Analysts suggest that NVIDIA is racing to secure HBM4 volumes amidst fierce pursuit from competitors like AMD and Google. This move is being evaluated as a shift in the status of Korean memory companies, elevated to the realm of a \"Super Supplier\" (Super-Eul) riding the wave of a \"Memory Super Cycle.\"\n\nThe Peerless Stature of K-Memory\nNVIDIA recently requested that Samsung Electronics supply HBM4 even before reliability or quality evaluations are fully finalized. Currently, both Samsung Electronics and SK Hynix are conducting final quality tests for HBM4. It is interpreted that NVIDIA is pushing to bypass detailed testing and expedite supply now that a certain level of quality has been verified.\n\nThe positions have flipped in just one generation; previously, Samsung was anxious over whether it would pass NVIDIA’s HBM3E quality tests. NVIDIA plans to integrate HBM4 into its next-generation AI chip, \"Rubin.\"\n\nThis shift demonstrates that Korean memory companies have risen to a peerless position in the global AI and semiconductor supply chain. Previously, the initiative in the AI ecosystem was held by NVIDIA, which produces Graphics Processing Units (GPUs), or Big Tech firms operating AI models. However, the situation has changed as AI competition intensifies. Without HBM to back up computational performance, cutting-edge AI accelerators cannot function. As securing HBM directly impacts AI chip launches and data center expansions, Korean memory giants like Samsung and SK Hynix now hold the \"bottleneck\" of the global AI industry.\n\nAccording to market research firm Counterpoint Research, SK Hynix and Samsung Electronics are projected to hold 54% and 28% of the global HBM4 market share this year, respectively, combined accounting for over 80%. Kim Dong-won, head of research at KB Securities, noted, \"The recent performance of Samsung and SK Hynix mirrors TSMC, which controls the majority of advanced semiconductor production and can raise prices at will. While NVIDIA's GPUs were the essential assets of the AI era, this necessity is now extending to memory.\"\n\nThe general memory shortage caused by firms concentrating on HBM production is also increasing their bargaining power. According to DRAMeXchange, consumer DRAM prices skyrocketed 750% from $1.35 in January last year to $11.5 this past January. NAND flash prices also rose from $2.18 to $9.46 during the same period. Jensen Huang, CEO of NVIDIA, remarked after a dinner with Taiwanese supply chain executives last month, \"We will need a tremendous amount of memory this year. Demand is so high that the entire supply chain is likely to face challenges.\"\n\nK-Semiconductor Earnings Soar\nThe changing status of memory semiconductors is reflected in the earnings forecasts for Samsung Electronics and SK Hynix. Global investment bank Morgan Stanley projected this year's operating profits for Samsung and SK Hynix at 245 trillion KRW and 179 trillion KRW, respectively. This represents a 4 to 5-fold increase compared to last year's operating profits of 43.6 trillion KRW and 47.2 trillion KRW.\n\nMorgan Stanley stated, \"In 2026, both companies entered an unprecedented stage of supply constraints, and this year's entire memory volume is already sold out. Compared to the past, capital efficiency is higher, and a high-margin structure has been established.\" Professor Kim Jung-ho of KAIST added, \"As memory semiconductors gradually transition into customized products, the influence of domestic companies with superior design and production capabilities will grow even further. In the AI era, memory will come to dominate the industry.\"\n\n---\n\nhttps://t.co/RS9Lv9UU2c", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10208951194617766, "ret_signed_m1d": -0.10208951194617766, "alpha_spy_m1d": -0.1106170117684383, "alpha_c_m1d": -0.12447119532146178, "bench_spy_m1d": 0.008527499822260642, "bench_c_m1d": 0.022381683375284123, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00955228696738053, "ret_signed_p1d": 0.00955228696738053, "alpha_spy_p1d": 0.014396208646277664, "alpha_c_p1d": 0.03742398560322102, "bench_spy_p1d": -0.004843921678897134, "bench_c_p1d": -0.02787169863584049, "ret_p1w": -0.010149228953051037, "ret_signed_p1w": -0.010149228953051037, "alpha_spy_p1w": -0.01390532301996139, "alpha_c_p1w": -0.013457259928367438, "bench_spy_p1w": 0.0037560940669103537, "bench_c_p1w": 0.0033080309753164006, "ret_p1m": 0.028059731963139933, "ret_signed_p1m": 0.028059731963139933, "alpha_spy_p1m": 0.03444095571772332, "alpha_c_p1m": 0.04382556844636454, "bench_spy_p1m": -0.006381223754583387, "bench_c_p1m": -0.015765836483224605, "ret_p3m": 0.31914270306660475, "ret_signed_p3m": 0.31914270306660475, "alpha_spy_p3m": 0.2711562448162055, "alpha_c_p3m": 0.17843848086464575, "bench_spy_p3m": 0.047986458250399266, "bench_c_p3m": 0.140704222201959, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4106, -0.4183, -0.4023, -0.3851, -0.3875, -0.3892, -0.4225, -0.4023, -0.4189, -0.4267, -0.4023, -0.4368, -0.4255, -0.41, -0.3892, -0.3851, -0.4029, -0.4011, -0.3857, -0.3791, -0.3756, -0.356, -0.3494, -0.356, -0.3583, -0.3625, -0.353, -0.3774, -0.3892, -0.3589, -0.3607, -0.3684, -0.3435, -0.3375, -0.3399, -0.3399, -0.3049, -0.2866, -0.2842, -0.2842, -0.2842, -0.2328, -0.1755, -0.1707, -0.1582, -0.1713, -0.1701, -0.1713, -0.1785, -0.1624, -0.1409, -0.111, -0.1087, -0.1331, -0.1075, -0.0907, -0.0919, -0.0919, -0.0478, -0.0304, -0.0406, -0.0418, -0.1021, 0.0, 0.0096, -0.049, -0.0531, -0.0066, -0.0101, 0.0018, 0.0663, 0.0818, 0.0818, 0.0818, 0.0818, 0.1343, 0.1349, 0.1522, 0.194, 0.2149, 0.3015, 0.2925, 0.2925, 0.1648, 0.0281, 0.1439, 0.1236, 0.0358, 0.1218, 0.1343, 0.1218, 0.0955, 0.1266, 0.1576, 0.2448, 0.197, 0.1904, 0.1122, 0.1325, 0.1284, 0.0752, 0.0752, 0.0547, 0.0003, 0.1346, 0.0673, 0.1139, 0.1552, 0.1756, 0.2593, 0.2204, 0.2324, 0.2025, 0.2354, 0.2623, 0.3012, 0.2922, 0.2832, 0.3102, 0.3012, 0.3431, 0.3132, 0.3431, 0.3281, 0.352, 0.3191, 0.3191, 0.3909, 0.3909, 0.5913, 0.6243, 0.6063, 0.708, 0.6691, 0.699, 0.7708, 0.6183, 0.6811, 0.6482, 0.6512, 0.7918, 0.7499, 0.7499, 0.7888, 0.8366, 0.7918, 0.8965, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2018615701465186799", "ticker_idx": 1, "ts": "2026-02-03T09:21:01+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley projected this year's operating profits for Samsung and SK Hynix at 245 trillion KRW and 179 trillion KRW, respectively. This represents a 4 to 5-fold increase compared to last year's operating profits", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares bullish thesis on Samsung and SK Hynix: NVIDIA skip-qual HBM4 demand, Morgan Stanley projects 4-5x profit growth, Korean memory firms now hold AI supply-chain bottleneck.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] NVIDIA to Samsung: Skip HBM4 Qual Tests, ‘Supply First’ Requested\n\nNVIDIA, the undisputed leader in Artificial Intelligence (AI) semiconductors, has reportedly requested that Samsung Electronics accelerate the supply of 6th-generation High Bandwidth Memory (HBM4). Samsung is currently undergoing final quality testing for mass production shipments scheduled for February, yet NVIDIA has requested supply regardless of the test results. Analysts suggest that NVIDIA is racing to secure HBM4 volumes amidst fierce pursuit from competitors like AMD and Google. This move is being evaluated as a shift in the status of Korean memory companies, elevated to the realm of a \"Super Supplier\" (Super-Eul) riding the wave of a \"Memory Super Cycle.\"\n\nThe Peerless Stature of K-Memory\nNVIDIA recently requested that Samsung Electronics supply HBM4 even before reliability or quality evaluations are fully finalized. Currently, both Samsung Electronics and SK Hynix are conducting final quality tests for HBM4. It is interpreted that NVIDIA is pushing to bypass detailed testing and expedite supply now that a certain level of quality has been verified.\n\nThe positions have flipped in just one generation; previously, Samsung was anxious over whether it would pass NVIDIA’s HBM3E quality tests. NVIDIA plans to integrate HBM4 into its next-generation AI chip, \"Rubin.\"\n\nThis shift demonstrates that Korean memory companies have risen to a peerless position in the global AI and semiconductor supply chain. Previously, the initiative in the AI ecosystem was held by NVIDIA, which produces Graphics Processing Units (GPUs), or Big Tech firms operating AI models. However, the situation has changed as AI competition intensifies. Without HBM to back up computational performance, cutting-edge AI accelerators cannot function. As securing HBM directly impacts AI chip launches and data center expansions, Korean memory giants like Samsung and SK Hynix now hold the \"bottleneck\" of the global AI industry.\n\nAccording to market research firm Counterpoint Research, SK Hynix and Samsung Electronics are projected to hold 54% and 28% of the global HBM4 market share this year, respectively, combined accounting for over 80%. Kim Dong-won, head of research at KB Securities, noted, \"The recent performance of Samsung and SK Hynix mirrors TSMC, which controls the majority of advanced semiconductor production and can raise prices at will. While NVIDIA's GPUs were the essential assets of the AI era, this necessity is now extending to memory.\"\n\nThe general memory shortage caused by firms concentrating on HBM production is also increasing their bargaining power. According to DRAMeXchange, consumer DRAM prices skyrocketed 750% from $1.35 in January last year to $11.5 this past January. NAND flash prices also rose from $2.18 to $9.46 during the same period. Jensen Huang, CEO of NVIDIA, remarked after a dinner with Taiwanese supply chain executives last month, \"We will need a tremendous amount of memory this year. Demand is so high that the entire supply chain is likely to face challenges.\"\n\nK-Semiconductor Earnings Soar\nThe changing status of memory semiconductors is reflected in the earnings forecasts for Samsung Electronics and SK Hynix. Global investment bank Morgan Stanley projected this year's operating profits for Samsung and SK Hynix at 245 trillion KRW and 179 trillion KRW, respectively. This represents a 4 to 5-fold increase compared to last year's operating profits of 43.6 trillion KRW and 47.2 trillion KRW.\n\nMorgan Stanley stated, \"In 2026, both companies entered an unprecedented stage of supply constraints, and this year's entire memory volume is already sold out. Compared to the past, capital efficiency is higher, and a high-margin structure has been established.\" Professor Kim Jung-ho of KAIST added, \"As memory semiconductors gradually transition into customized products, the influence of domestic companies with superior design and production capabilities will grow even further. In the AI era, memory will come to dominate the industry.\"\n\n---\n\nhttps://t.co/RS9Lv9UU2c", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08489524402277482, "ret_signed_m1d": -0.08489524402277482, "alpha_spy_m1d": -0.09342274384503546, "alpha_c_m1d": -0.1107703414237926, "bench_spy_m1d": 0.008527499822260642, "bench_c_m1d": 0.025875097401017788, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007717749456615852, "ret_signed_p1d": -0.007717749456615852, "alpha_spy_p1d": -0.0028738277777187182, "alpha_c_p1d": 0.03166065516651184, "bench_spy_p1d": -0.004843921678897134, "bench_c_p1d": -0.039378404623127694, "ret_p1w": -0.03417856528090568, "ret_signed_p1w": -0.03417856528090568, "alpha_spy_p1w": -0.037934659347816035, "alpha_c_p1w": -0.051981867484282906, "bench_spy_p1w": 0.0037560940669103537, "bench_c_p1w": 0.017803302203377225, "ret_p1m": -0.06221980219499634, "ret_signed_p1m": -0.06221980219499634, "alpha_spy_p1m": -0.055838578440412956, "alpha_c_p1m": -0.06579054614462232, "bench_spy_p1m": -0.006381223754583387, "bench_c_p1m": 0.0035707439496259763, "ret_p3m": 0.4204774212587592, "ret_signed_p3m": 0.4204774212587592, "alpha_spy_p3m": 0.37249096300835993, "alpha_c_p3m": 0.13849186323783425, "bench_spy_p3m": 0.047986458250399266, "bench_c_p3m": 0.28198555802092495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3467, -0.361, -0.3323, -0.318, -0.3202, -0.3257, -0.383, -0.3323, -0.372, -0.3808, -0.3709, -0.426, -0.4271, -0.4282, -0.4227, -0.4002, -0.4157, -0.4068, -0.3848, -0.3914, -0.4024, -0.4002, -0.3638, -0.376, -0.3528, -0.3771, -0.3705, -0.3892, -0.4157, -0.3925, -0.3914, -0.3969, -0.3605, -0.3561, -0.3517, -0.3517, -0.3396, -0.2944, -0.2822, -0.2822, -0.2822, -0.2536, -0.2326, -0.1996, -0.1819, -0.1665, -0.1797, -0.1742, -0.1863, -0.1819, -0.1742, -0.1665, -0.1577, -0.1808, -0.1841, -0.1676, -0.1544, -0.1885, -0.118, -0.0728, -0.0507, 0.0022, -0.0849, 0.0, -0.0077, -0.0717, -0.075, -0.0221, -0.0342, -0.0518, -0.0209, -0.0298, -0.0298, -0.0298, -0.0298, -0.0143, 0.0463, 0.0485, 0.108, 0.1224, 0.2139, 0.1719, 0.1719, 0.0372, -0.0622, 0.0394, 0.0206, -0.0766, 0.0361, 0.0549, 0.0273, 0.0052, 0.0759, 0.0714, 0.1664, 0.1189, 0.1123, 0.0306, 0.0891, 0.099, 0.0306, 0.0306, -0.0357, -0.1086, -0.0092, -0.0832, -0.0324, -0.0214, 0.0118, 0.141, 0.1024, 0.1344, 0.1488, 0.2183, 0.2548, 0.2758, 0.246, 0.2879, 0.352, 0.3509, 0.3531, 0.3498, 0.4271, 0.4359, 0.4282, 0.4205, 0.4205, 0.5983, 0.5983, 0.7684, 0.827, 0.8623, 1.0766, 1.0269, 1.1826, 1.176, 1.0092, 1.0324, 0.9275, 0.9275, 1.1429, 1.144, 1.144, 1.2666, 1.4776, 1.5288, 1.5774, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2027652635873198297", "ticker_idx": 0, "ts": "2026-02-28T07:50:35+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadcom's margins should improve, then.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish AVGO: growing customer base distributes IP costs, improving margins as XDSiP scales.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"As the customer base grows, IP development costs are distributed and Broadcom's margins improve — a scalable structure.\"\n\nThis is really interesting. I didn’t realize it was structured this way. Broadcom’s margins should improve, then.\n\n$AVGO https://t.co/9qxAw4TBYU", "reply_text": null, "parent_text": null, "quoted_text": "The Next Big HBM Customer's Move… Broadcom Ships First AI 3.5D XDSiP\n\nBroadcom has completed development of its system-in-package (SiP) technology for custom artificial intelligence (AI) data centers utilizing hybrid bonding — an industry first — and is beginning initial mass https://t.co/BOJmp3lwTz", "parent_author": null, "ret_m1d": 0.0022896388327349193, "ret_signed_m1d": 0.0022896388327349193, "alpha_spy_m1d": 0.002857894516941295, "alpha_c_m1d": 0.002338900700714963, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -4.926186798004384e-05, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015620094530333861, "ret_signed_p1d": -0.015620094530333861, "alpha_spy_p1d": -0.006805756047538192, "alpha_c_p1d": 0.02210233106396997, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.03772242559430383, "ret_p1w": 0.08446772132016345, "ret_signed_p1w": 0.08446772132016345, "alpha_spy_p1w": 0.09628330224367831, "alpha_c_p1w": 0.1140452677607382, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.029577546440574753, "ret_p1m": -0.027164977254687583, "ret_signed_p1m": -0.027164977254687583, "alpha_spy_p1m": 0.022754764320496168, "alpha_c_p1m": 0.029406345669773293, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.056571322924460876, "ret_p3m": 0.34080300425170673, "ret_signed_p3m": 0.34080300425170673, "alpha_spy_p3m": 0.23840962508579877, "alpha_c_p3m": -0.13519294906606505, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.4759959533177718, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1915, 0.1928, 0.2217, 0.2557, 0.2719, 0.2928, 0.2722, 0.1268, 0.0638, 0.0685, 0.0206, 0.0327, 0.0655, 0.071, 0.0957, 0.0985, 0.0985, 0.1045, 0.0959, 0.0973, 0.0856, 0.0856, 0.0903, 0.0772, 0.0783, 0.0774, 0.0428, 0.082, 0.1047, 0.1123, 0.0661, 0.0759, 0.1032, 0.1032, 0.0432, 0.0313, 0.0209, 0.0039, 0.0189, 0.0438, 0.0452, 0.0374, 0.0391, 0.0385, 0.0047, -0.0338, -0.0261, 0.0442, 0.0788, 0.0678, 0.0751, 0.0387, 0.0199, 0.0199, 0.043, 0.0461, 0.0476, 0.0434, 0.0361, 0.0209, 0.0423, 0.009, 0.0023, 0.0, -0.0156, -0.004, 0.0438, 0.0366, 0.0845, 0.0745, 0.0714, 0.0538, 0.0105, 0.0191, 0.0078, -0.0091, 0.0032, -0.0261, 0.0137, 0.0004, 0.0021, -0.0274, -0.0549, -0.0778, -0.0272, -0.0147, -0.0113, -0.0113, -0.0117, 0.0497, 0.1021, 0.1155, 0.1678, 0.1936, 0.1968, 0.2469, 0.2524, 0.2778, 0.2561, 0.2641, 0.3285, 0.3199, 0.3288, 0.3145, 0.2567, 0.2744, 0.312, 0.3241, 0.3091, 0.3433, 0.3372, 0.2967, 0.3516, 0.3466, 0.3179, 0.31, 0.3823, 0.3364, 0.3224, 0.2921, 0.3131, 0.3031, 0.3017, 0.3017, 0.3264, 0.326, 0.3408, 0.4043, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2011648540800664035", "ticker_idx": 0, "ts": "2026-01-15T03:56:01+00:00", "symbol": "USDKRW=X", "kind": "fx", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "1500원대 진짜 돌파할거 같네요...", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Expects USD/KRW to break above 1,500 won level.", "resolved_tickers": ["USDKRW=X"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": null, "bench_c_reason": "no_country_fallback (all FX/crypto/index)", "bench_c_industry": null, "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@blazingbees 1500원대 진짜 돌파할거 같네요...", "reply_text": null, "parent_text": "[단독]정부, 은행에 환전우대 서비스 자제령… 환율방어 총력\n\n환율을 방어하랬지, \n누가 원화 팔고 싶게 만들랬냐\n이러면 달러를 더 갖고 싶어지잖아.\n\n부동산도 똑같다\n투기과열지구 지정은 사실상\n쪽집게 과외선생이 여기 오른다고 찍어준 격. https://t.co/qE62vM3NlC", "quoted_text": null, "parent_author": "blazingbees", "ret_m1d": 0.006910920215215555, "ret_signed_m1d": 0.006910920215215555, "alpha_spy_m1d": 0.009626738855325834, "alpha_c_m1d": null, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": null, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": null, "bench_spy_t0": 0.0, "bench_c_t0": null, "ret_p1d": 0.0040603877962399615, "ret_signed_p1d": 0.0040603877962399615, "alpha_spy_p1d": 0.004898261787135372, "alpha_c_p1d": null, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": null, "ret_p1w": 0.0013192505355230644, "ret_signed_p1w": 0.0013192505355230644, "alpha_spy_p1w": 0.0060285674078491835, "alpha_c_p1w": null, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": null, "ret_p1m": -0.015038621664396845, "ret_signed_p1m": -0.015038621664396845, "alpha_spy_p1m": 0.00011507697857793264, "alpha_c_p1m": null, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": null, "ret_p3m": 0.00952897749238435, "ret_signed_p3m": 0.00952897749238435, "alpha_spy_p3m": 0.0035821675050891066, "alpha_c_p3m": null, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0295, -0.0295, -0.022, -0.0222, -0.0185, -0.0177, -0.0218, -0.0252, -0.0266, -0.0266, -0.0244, -0.0232, -0.0158, -0.0158, -0.0103, -0.0061, -0.0047, -0.0021, 0.0035, 0.004, -0.0112, -0.0008, -0.0043, 0.0022, 0.0071, 0.0033, 0.0082, 0.0014, 0.0053, -0.001, 0.0019, 0.005, 0.003, 0.0015, 0.0065, 0.0061, 0.0038, 0.0034, 0.0042, 0.0054, 0.0068, 0.0031, 0.006, 0.0076, 0.0078, 0.0074, 0.0095, 0.0108, 0.0108, -0.0112, -0.0147, -0.0193, -0.0171, -0.0171, -0.0132, -0.0153, -0.0119, -0.0123, -0.0115, -0.0088, -0.0052, 0.0014, 0.0069, 0.0, 0.0041, 0.0059, 0.0059, 0.0102, 0.0013, 0.0002, -0.0124, -0.015, -0.0208, -0.023, -0.0233, -0.0096, -0.0071, -0.0111, -0.0019, 0.0053, -0.0003, -0.003, -0.0052, -0.0128, -0.015, -0.016, -0.0155, -0.0157, -0.0155, -0.0104, -0.0133, -0.0142, -0.0157, -0.0246, -0.0209, -0.0167, -0.0073, 0.0138, -0.0015, 0.0114, 0.0143, 0.0011, 0.0068, 0.0085, 0.0052, 0.0266, 0.0178, 0.015, 0.0289, 0.0184, 0.0282, 0.0154, 0.0235, 0.026, 0.0311, 0.0302, 0.0364, 0.0276, 0.0336, 0.0317, 0.0317, 0.0307, 0.0248, 0.01, 0.0071, 0.0132, 0.0095, 0.0045, 0.0076, 0.0103, 0.0015, 0.0043, 0.0157, 0.0103, 0.0117, 0.0081, 0.0069, 0.0062, 0.0167, 0.0076, 0.006, 0.009, 0.0076, -0.0126, -0.0055, -0.0016, 0.0079, 0.0206, 0.0184, 0.0208, 0.0234, 0.0201, 0.0307, 0.0252, 0.0284, 0.0341, 0.0355, 0.0291, 0.0275, 0.0302, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2036959231065415899", "ticker_idx": 0, "ts": "2026-03-26T00:11:40+00:00", "symbol": "GLW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$GLW", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Tags GLW as beneficiary of fiber optic prices hitting all-time highs on AI/drone demand surge and supply constraints.", "resolved_tickers": ["GLW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Fiber Optic Prices Break to an All-Time High… Tight Supply Amid AI and Drone Demand (Shanghai Securities News)\n\n• In March, with demand from both drones and AIDC construction rising at the same time, the price of G.657.A2 optical fiber surpassed RMB 210 per core-kilometer, reaching a record high. China Galaxy Securities noted that as AI and computing network deployment accelerates, demand for optical fiber and cable is surging, while near-term capacity expansion remains difficult, causing supply constraints to further drive price increases.\n\n• In particular, as AI infrastructure investment expands, demand is rising rapidly for data center interconnects (DCI) and all-optical network buildouts. In other words, this is not merely a simple price increase, but a sign that network infrastructure bottlenecks in the AI era are becoming a reality.\n\n$GLW\n\n---\n\nhttps://t.co/l1eClcmhSl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08151050039634056, "ret_signed_m1d": 0.08151050039634056, "alpha_spy_m1d": 0.06332702215233721, "alpha_c_m1d": 0.04935959845795623, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.032150901938384324, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011010799726836273, "ret_signed_p1d": 0.011010799726836273, "alpha_spy_p1d": 0.028062683250164255, "alpha_c_p1d": 0.030482511659324474, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.0194717119324882, "ret_p1w": 0.0931125074973791, "ret_signed_p1w": 0.0931125074973791, "alpha_spy_p1w": 0.07646368363208422, "alpha_c_p1w": 0.06677284339802769, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.0263396640993514, "ret_p1m": 0.299807753139808, "ret_signed_p1m": 0.299807753139808, "alpha_spy_p1m": 0.19307850184204578, "alpha_c_p1m": 0.09060019675714281, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.2092075563826652, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3435, -0.3446, -0.3541, -0.3541, -0.3312, -0.3458, -0.3417, -0.351, -0.3706, -0.3713, -0.3518, -0.337, -0.3346, -0.3104, -0.3052, -0.3052, -0.3172, -0.3102, -0.3046, -0.3118, -0.2996, -0.1905, -0.2308, -0.2403, -0.2384, -0.186, -0.168, -0.1909, -0.168, -0.0989, -0.0309, -0.0551, -0.0197, -0.03, -0.0156, -0.0156, -0.0373, -0.0263, -0.0412, 0.029, 0.0714, 0.1181, 0.1834, 0.1086, 0.1113, 0.1666, 0.0905, 0.0704, -0.0043, -0.0889, -0.0463, 0.0067, -0.0263, -0.041, -0.0458, -0.0242, -0.0401, -0.0403, -0.0166, -0.0794, -0.0321, 0.0494, 0.0815, 0.0, 0.011, -0.05, 0.0048, 0.0522, 0.0931, 0.0931, 0.0826, 0.0975, 0.2201, 0.2548, 0.2654, 0.2945, 0.2771, 0.2435, 0.2273, 0.2148, 0.2221, 0.2227, 0.2471, 0.2526, 0.2998, 0.2416, 0.131, 0.1225, 0.2137, 0.1695, 0.1821, 0.1979, 0.3418, 0.3479, 0.3815, 0.5326, 0.465, 0.5261, 0.5392, 0.4175, 0.3195, 0.2994, 0.3353, 0.418, 0.434, 0.434, 0.4497, 0.4107, 0.3521, 0.3408, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2026267362564202916", "ticker_idx": 0, "ts": "2026-02-24T12:06:00+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Jeff Pu was right again. $AMD", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Author endorses GF Securities Buy on AMD with raised PT $308, citing CPU/GPU tailwinds.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Jeff Pu was right again.\n\n$AMD https://t.co/EbOPR9xGco", "reply_text": null, "parent_text": null, "quoted_text": "《GF Overseas Electronics &amp; Communications》 \nAMD (Buy, AMD US): Performance Outlook, Tailwinds in CPU and GPU\n\nRaising Target Price to $308: Since last November, AMD's stock has been consolidating, but it rebounded 15% in the past week, mainly driven by reports of CPU price", "parent_author": null, "ret_m1d": -0.08062098077823532, "ret_signed_m1d": -0.08062098077823532, "alpha_spy_m1d": -0.07340497703141957, "alpha_c_m1d": -0.06563863817772497, "bench_spy_m1d": -0.007216003746815747, "bench_c_m1d": -0.014982342600510345, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013935633083487109, "ret_signed_p1d": -0.013935633083487109, "alpha_spy_p1d": -0.022373982027068418, "alpha_c_p1d": -0.03063569973784952, "bench_spy_p1d": 0.008438348943581309, "bench_c_p1d": 0.01670006665436241, "ret_p1w": -0.10704264768822636, "ret_signed_p1w": -0.10704264768822636, "alpha_spy_p1w": -0.09682963658579169, "alpha_c_p1w": -0.0400037941286977, "bench_spy_p1w": -0.010213011102434666, "bench_c_p1w": -0.06703885355952865, "ret_p1m": 0.030069248244889657, "ret_signed_p1m": 0.030069248244889657, "alpha_spy_p1m": 0.07187642792019533, "alpha_c_p1m": 0.07811776068042409, "bench_spy_p1m": -0.04180717967530567, "bench_c_p1m": -0.04804851243553443, "ret_p3m": 1.1862608388138387, "ret_signed_p3m": 1.1862608388138387, "alpha_spy_p3m": 1.0984941742544592, "alpha_c_p3m": 0.8113204764485444, "bench_spy_p3m": 0.08776666455937954, "bench_c_p3m": 0.37494036236529427, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0019, 0.0173, 0.0277, 0.0065, 0.0176, 0.01, 0.0193, 0.034, 0.0364, 0.0354, 0.0355, -0.0143, -0.0293, -0.0218, -0.0736, -0.0598, -0.0019, 0.0052, 0.005, 0.0056, 0.0056, 0.0054, 0.0083, 0.007, 0.0015, 0.0015, 0.045, 0.0339, 0.0024, -0.0179, -0.0428, -0.0499, -0.0288, 0.0333, 0.0456, 0.0658, 0.0841, 0.0841, 0.0845, 0.1682, 0.1865, 0.2144, 0.1752, 0.1786, 0.1819, 0.1793, 0.107, 0.1517, 0.1322, -0.0638, -0.0998, -0.0253, 0.0101, -0.0013, -0.0012, -0.0369, -0.0305, -0.0305, -0.0503, -0.0642, -0.049, -0.064, -0.0806, 0.0, -0.0139, -0.0475, -0.0637, -0.0712, -0.107, -0.055, -0.0673, -0.1001, -0.0522, -0.0496, -0.0421, -0.0753, -0.0956, -0.0807, -0.082, -0.0672, -0.0401, -0.0585, -0.0522, -0.0396, 0.0301, -0.0471, -0.0554, -0.0832, -0.0487, -0.017, 0.0171, 0.0171, 0.0296, 0.036, 0.0841, 0.1066, 0.1459, 0.1543, 0.1928, 0.2071, 0.3013, 0.3019, 0.2858, 0.3304, 0.4191, 0.4278, 0.6265, 0.5649, 0.5115, 0.5765, 0.6577, 0.686, 0.5972, 0.6613, 0.9706, 0.9101, 1.1286, 1.1455, 1.0964, 1.0833, 1.103, 0.9833, 0.9687, 0.9363, 1.0931, 1.1025, 1.1863, 1.1863, 1.3564, 1.3173, 1.4228, 1.4135, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013100653124702526", "ticker_idx": 0, "ts": "2026-01-19T04:06:11+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "spending $1.8 billion for that simply doesn't make sense to me. In my view, the PSMC fab is worth $1 billion at the absolute most. I just can't rationally make sense of this deal.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish on Micron's PSMC deal: $1.8B price tag for a tiny, legacy-process fab worth at most $1B is value-destructive.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts on the Micron-PSMC Deal\n\nHonestly... I just don’t understand why SJ made this deal.\n\nThe production capacity secured through this acquisition is only around 20,000 wpm (wafers per month).\n\nThat’s less than 1% of global DRAM production capacity—an extremely insignificant volume.\n\nThe PSMC fab that Micron is buying is tiny compared to competitors: Samsung’s P4 fab (200,000 wpm) or SK Hynix’s M15X fab (900,000 wpm).\n\nEven if the deal closes, actual wafer starts and chip output won’t happen until mid-2027 or later.\n\nOn top of that, despite being a new fab, the PSMC facility was built for legacy processes, making it very difficult to introduce EUV equipment.\n\nIn other words, the most advanced DRAM this fab can realistically produce is around the 1a node, and spending $1.8 billion for that simply doesn’t make sense to me.\n\nIn my view, the PSMC fab is worth $1 billion at the absolute most.\n\nI just can’t rationally make sense of this deal.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006202615287283164, "ret_signed_p1d": -0.006202615287283164, "alpha_spy_p1d": -0.02655940101082821, "alpha_c_p1d": -0.03117826311601457, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": 0.07261193167486435, "ret_signed_p1w": -0.07261193167486435, "alpha_spy_p1w": -0.07106494054738888, "alpha_c_p1w": -0.07653312667668599, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": 0.10208127675723166, "ret_signed_p1m": -0.10208127675723166, "alpha_spy_p1m": -0.11481875732693281, "alpha_c_p1m": -0.08429862769831531, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.2609843572427655, "ret_signed_p3m": -0.2609843572427655, "alpha_spy_p3m": -0.24375588612888488, "alpha_c_p3m": -0.1250919244799107, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4531, -0.4304, -0.3965, -0.3935, -0.3885, -0.3755, -0.3827, -0.3834, -0.3533, -0.3992, -0.3455, -0.3433, -0.3444, -0.302, -0.3356, -0.3252, -0.3471, -0.3198, -0.3333, -0.3703, -0.3775, -0.4451, -0.4286, -0.3829, -0.3813, -0.3655, -0.3655, -0.3484, -0.3374, -0.3401, -0.3547, -0.3754, -0.3463, -0.3196, -0.3044, -0.2733, -0.2878, -0.3355, -0.3455, -0.3593, -0.3786, -0.3151, -0.2672, -0.2378, -0.2387, -0.21, -0.21, -0.2152, -0.1885, -0.1933, -0.2132, -0.2132, -0.1305, -0.1395, -0.0533, -0.064, -0.0985, -0.0487, -0.0465, -0.0679, -0.081, -0.072, 0.0, 0.0, 0.0062, 0.0727, 0.096, 0.1017, 0.0726, 0.1309, 0.1999, 0.2014, 0.1437, 0.2069, 0.1563, 0.0459, 0.0555, 0.088, 0.0572, 0.0289, 0.1312, 0.1412, 0.1348, 0.1348, 0.1021, 0.1604, 0.1505, 0.1803, 0.1605, 0.1523, 0.1826, 0.1456, 0.1368, 0.1376, 0.0467, 0.1048, 0.0946, 0.0208, 0.0732, 0.1113, 0.1542, 0.1174, 0.1747, 0.2179, 0.2727, 0.2729, 0.2247, 0.1658, 0.1147, 0.0904, 0.0533, -0.0201, -0.0152, -0.1125, -0.0683, 0.0145, 0.01, 0.01, 0.0418, 0.0413, 0.1217, 0.1625, 0.1599, 0.1764, 0.2842, 0.2582, 0.261, 0.255, 0.2367, 0.2393, 0.3444, 0.3285, 0.3699, 0.4467, 0.3908, 0.4298, 0.4263, 0.4953, 0.5898, 0.7656, 0.8384, 0.7833, 1.0596, 1.1934, 1.1141, 1.2163, 1.1401, 0.9985, 0.8796, 0.927, 1.0187, 1.1018, 1.0712, 1.0712, 1.4707, 1.5604, 1.547, 1.6779, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008684781383635242", "ticker_idx": 0, "ts": "2026-01-06T23:39:06+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "맥쿼리, 삼성전자 목표가 24만원 제시", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Macquarie's ₩240,000 price target for Samsung Electronics; implied bullish.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "ko", "tweet_text": "Macquarie sets a ₩240,000 price target for Samsung Electronics.\n\n맥쿼리, 삼성전자 목표가 24만원 제시 https://t.co/aU9y0OUg3J", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005759511806582074, "ret_signed_m1d": -0.005759511806582074, "alpha_spy_m1d": 0.00015251167873397975, "alpha_c_m1d": 0.008082984416169725, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.013842496222751799, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015118859397076356, "ret_signed_p1d": 0.015118859397076356, "alpha_spy_p1d": 0.01834221618930376, "alpha_c_p1d": 0.01593707653173493, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.000818217134658572, "ret_p1w": -0.009359234866655575, "ret_signed_p1w": -0.009359234866655575, "alpha_spy_p1w": -0.012192455434991256, "alpha_c_p1w": -0.00820004183669043, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": -0.0011591930299651443, "ret_p1m": 0.2174226415749414, "ret_signed_p1m": 0.2174226415749414, "alpha_spy_p1m": 0.22554623319201983, "alpha_c_p1m": 0.27558835794362835, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.05816571636868695, "ret_p3m": 0.34330743916463513, "ret_signed_p3m": 0.34330743916463513, "alpha_spy_p3m": 0.39272690584753245, "alpha_c_p3m": 0.4148444528637819, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": -0.07153701369914678, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.357, -0.3237, -0.3315, -0.3437, -0.3194, -0.3, -0.2986, -0.2972, -0.3015, -0.2936, -0.3086, -0.2921, -0.2692, -0.2871, -0.28, -0.2542, -0.2298, -0.204, -0.2484, -0.2792, -0.2893, -0.2986, -0.2792, -0.2585, -0.2613, -0.2635, -0.3036, -0.2792, -0.2993, -0.3086, -0.2792, -0.3208, -0.3072, -0.2886, -0.2635, -0.2585, -0.28, -0.2778, -0.2592, -0.2513, -0.247, -0.2234, -0.2155, -0.2234, -0.2262, -0.2312, -0.2198, -0.2492, -0.2635, -0.2269, -0.2291, -0.2384, -0.2083, -0.2011, -0.204, -0.204, -0.1617, -0.1397, -0.1368, -0.1368, -0.1368, -0.0749, -0.0058, 0.0, 0.0151, -0.0007, 0.0007, -0.0007, -0.0094, 0.0101, 0.036, 0.072, 0.0749, 0.0454, 0.0763, 0.0965, 0.095, 0.095, 0.1483, 0.1692, 0.1569, 0.1555, 0.0828, 0.2059, 0.2174, 0.1469, 0.1418, 0.198, 0.1937, 0.2081, 0.2858, 0.3045, 0.3045, 0.3045, 0.3045, 0.3679, 0.3686, 0.3895, 0.4399, 0.4651, 0.5695, 0.5587, 0.5587, 0.4046, 0.2397, 0.3794, 0.3549, 0.2491, 0.3528, 0.3679, 0.3528, 0.3211, 0.3585, 0.396, 0.5011, 0.4435, 0.4356, 0.3413, 0.3657, 0.3607, 0.2966, 0.2966, 0.2719, 0.2062, 0.3682, 0.287, 0.3433, 0.3931, 0.4176, 0.5186, 0.4717, 0.4862, 0.4501, 0.4898, 0.5222, 0.5691, 0.5583, 0.5475, 0.5799, 0.5691, 0.6196, 0.5835, 0.6196, 0.6016, 0.6304, 0.5908, 0.5908, 0.6773, 0.6773, 0.919, 0.9587, 0.937, 1.0597, 1.0128, 1.0489, 1.1354, 0.9515, 1.0272, 0.9875, 0.9912, 1.1607, 1.1102, 1.1102, 1.1571, 1.2148, 1.1607, 1.2869, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2029154783224062329", "ticker_idx": 0, "ts": "2026-03-04T11:19:34+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Xiaomi plans to release its own smartphone application processor (AP) every year going forward, and ultimately aims to incorporate it into overseas products as well. $QCOM", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Xiaomi's annual in-house AP roadmap and overseas expansion implies reduced Qualcomm content; bearish QCOM.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Xiaomi President Lu Weibing said in an interview with CNBC that Xiaomi plans to release its own smartphone application processor (AP) every year going forward, and ultimately aims to incorporate it into overseas products as well.\n\n$QCOM\nhttps://t.co/v9PxaOLzSO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009891684829271497, "ret_signed_m1d": 0.009891684829271497, "alpha_spy_m1d": 0.002885805117835649, "alpha_c_m1d": -0.010253663258837808, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.020145348088109305, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011686590493802718, "ret_signed_p1d": 0.011686590493802718, "alpha_spy_p1d": 0.006111043647408887, "alpha_c_p1d": 0.002290449319534371, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.009396141174268346, "ret_p1w": -0.03246284921735687, "ret_signed_p1w": 0.03246284921735687, "alpha_spy_p1w": 0.019618617307981312, "alpha_c_p1w": 0.03729871152320885, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004835862305851979, "ret_p1m": -0.08526902670453296, "ret_signed_p1m": 0.08526902670453296, "alpha_spy_p1m": 0.0451176929852396, "alpha_c_p1m": 0.0682808065200955, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.016988220184437464, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.253, 0.2566, 0.2616, 0.3061, 0.2993, 0.278, 0.2849, 0.2624, 0.2353, 0.2486, 0.2562, 0.2488, 0.2526, 0.2527, 0.2527, 0.253, 0.2431, 0.2447, 0.2261, 0.2261, 0.2399, 0.2638, 0.3078, 0.2916, 0.3036, 0.2743, 0.2133, 0.1848, 0.1794, 0.1568, 0.1427, 0.1427, 0.1044, 0.1209, 0.1311, 0.1169, 0.1076, 0.097, 0.0945, 0.0911, 0.0866, 0.094, 0.055, 0.0672, -0.023, -0.0156, -0.0042, 0.0042, 0.011, -0.0075, 0.0085, 0.0085, 0.0224, 0.0267, 0.0126, 0.0242, 0.0065, 0.0378, 0.0452, 0.0436, 0.0204, 0.0109, -0.0099, 0.0, -0.0117, -0.0211, -0.0037, -0.0247, -0.0325, -0.0539, -0.0635, -0.0666, -0.0507, -0.0588, -0.053, -0.0629, -0.0741, -0.0718, -0.0597, -0.0583, -0.083, -0.0833, -0.071, -0.0818, -0.0853, -0.0853, -0.093, -0.105, -0.0801, -0.0784, -0.0762, -0.0532, -0.0417, -0.0402, -0.0299, -0.0175, -0.0079, -0.0221, -0.0184, -0.0337, 0.0738, 0.084, 0.0821, 0.1254, 0.2955, 0.2769, 0.2147, 0.3458, 0.3892, 0.4612, 0.5805, 0.7135, 0.5172, 0.5378, 0.4434, 0.4535, 0.4691, 0.4111, 0.4609, 0.5395, 0.7181, 0.7181, 0.795, 0.6837, 0.7551, 0.8108, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2019244472404697179", "ticker_idx": 0, "ts": "2026-02-05T02:59:32+00:00", "symbol": "semiconductors", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the chip industry has already benefited from the AI buildout cycle—and that tailwind should continue", "tweet_type": "analysis", "timeline": "months", "summary": "Author endorses BofA bullish chip/NVDA thesis and personally views the selloff as excessive, regretting inability to add more exposure.", "resolved_tickers": ["SMH"], "resolution_method": "etf", "resolution_reason": "US semiconductor sector proxied by SMH ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "original", "lang": "en", "tweet_text": "BofA (Bank of America) analyst Vivek Arya concluded that it doesn’t make logical sense for the chip sector and the software sector to sell off together. He argued that a decline in chip stocks implies deteriorating AI investment returns (ROI), while a decline in software stocks implies AI adoption will become extremely widespread—so those two conclusions are hard to reconcile at the same time. In his view, the market is overreacting.\n\nArya noted that the market’s implicit judgment behind the selloff in AI-related chip stocks is that “AI investments have deteriorated to the point where achieving ideal returns is difficult, limiting future growth runway.” By contrast, the selloff in software stocks reflects a different judgment: “AI adoption will be broadly disseminated and meaningfully boost productivity, replacing software businesses and application business models.”\n\nArya wrote that “these two outcomes cannot occur simultaneously.” He said this positioning resembles the fear seen during the “DeepSeek shock” in January this year—an episode that ultimately proved to lack factual grounding. At the time, the market worried that DeepSeek could build competitive models at lower cost, reducing demand for high-performance chips.\n\nArya believes what the market actually witnessed after the “DeepSeek shock” was more AI spending and accelerated growth in AI tokens. In his report, he estimated that 2025 cloud capex growth could reach 69%, far above the initial expectation of 20–30%.\n\nHe also argued that even if current AI models are highly capable, it may still take years to truly “prove value” from a productivity standpoint—so he does not expect AI investment to slow in the near term. Even if a major productivity breakthrough emerges in the future, models will need continuous improvement to sustain user engagement, meaning investment is unlikely to decline.\n\nHis conclusion is that while the ultimate end-state of the software industry is uncertain, the chip industry has already benefited from the AI buildout cycle—and that tailwind should continue. He added that enterprise AI adoption is still early, and more sovereign nations are beginning to increase AI investment. Arya also pointed out that chip-sector valuations already reflect concerns about slowing spend or downward revisions to earnings expectations, but those concerns “may not actually materialize.”\n\nFrom a trading perspective, this drawdown looks closer to a “narrative-driven correlation shock.” When the market sells software on the thesis that AI will displace it, while simultaneously selling AI chips on the thesis that AI ROI is worsening, core AI assets like Nvidia can easily get dragged down in the process—through position unwinds and broad risk re-rating.\n\nArya concluded that what Nvidia investors most need in the near term is clarity on whether this price action reflects a real change in AI capex and demand fundamentals, or whether it’s coming from indiscriminate, cross-sector de-risking triggered by the narrative that “AI beats software.”\n\nThat was Arya’s interpretation, and I also don’t really understand this selloff.\n\nIf I had to find a reason, you could argue that there was simply too much leverage in the system recently, and mechanical selling—like CTA-driven flows—may have contributed.\n\nStill, I think the magnitude of the drop is excessive.\n\nI’m really bummed that I don’t have any cash right now, so I can’t buy more even after this drop.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0024929321940703186, "ret_signed_m1d": 0.0024929321940703186, "alpha_spy_m1d": -0.010154232169661936, "alpha_c_m1d": -0.010154232169661936, "bench_spy_m1d": 0.012647164363732255, "bench_c_m1d": 0.012647164363732255, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05400578942639256, "ret_signed_p1d": 0.05400578942639256, "alpha_spy_p1d": 0.034821077585441174, "alpha_c_p1d": 0.034821077585441174, "bench_spy_p1d": 0.019184711840951385, "bench_c_p1d": 0.019184711840951385, "ret_p1w": 0.06570965320301214, "ret_signed_p1w": 0.06570965320301214, "alpha_spy_p1w": 0.06032312882582591, "alpha_c_p1w": 0.06032312882582591, "bench_spy_p1w": 0.00538652437718623, "bench_c_p1w": 0.00538652437718623, "ret_p1m": -0.0013383623896463526, "ret_signed_p1m": -0.0013383623896463526, "alpha_spy_p1m": 0.006394645049282999, "alpha_c_p1m": 0.006394645049282999, "bench_spy_p1m": -0.007733007438929351, "bench_c_p1m": -0.007733007438929351, "ret_p3m": 0.37163773688621893, "ret_signed_p3m": 0.37163773688621893, "alpha_spy_p3m": 0.30061471633389947, "alpha_c_p3m": 0.30061471633389947, "bench_spy_p3m": 0.07102302055231946, "bench_c_p3m": 0.07102302055231946, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0614, -0.0816, -0.0699, -0.0979, -0.0975, -0.1098, -0.1281, -0.112, -0.1495, -0.1468, -0.1128, -0.1105, -0.0904, -0.0904, -0.0784, -0.0766, -0.0597, -0.0468, -0.054, -0.0467, -0.0359, -0.0347, -0.0213, -0.0297, -0.0736, -0.0768, -0.0793, -0.1125, -0.0916, -0.0681, -0.0561, -0.047, -0.0443, -0.0443, -0.0399, -0.0442, -0.0466, -0.055, -0.055, -0.0204, -0.0091, 0.0172, 0.0104, -0.0055, 0.0214, 0.0252, 0.0274, 0.0191, 0.0403, 0.0507, 0.0507, 0.0245, 0.0547, 0.0571, 0.0499, 0.0466, 0.0687, 0.0933, 0.0957, 0.0588, 0.0706, 0.0436, 0.0025, 0.0, 0.054, 0.0671, 0.0622, 0.0885, 0.0657, 0.0699, 0.0699, 0.0694, 0.0826, 0.0764, 0.0891, 0.0835, 0.1, 0.1183, 0.0812, 0.0664, 0.0664, 0.0262, 0.0473, 0.0375, -0.0013, 0.0349, 0.0427, 0.0524, 0.0185, 0.0164, 0.0337, 0.0415, 0.0331, 0.0363, 0.0096, 0.027, 0.0355, 0.0471, -0.0006, -0.0179, -0.0487, 0.0061, 0.0286, 0.0295, 0.0295, 0.0391, 0.0494, 0.1098, 0.1292, 0.1465, 0.1634, 0.1861, 0.1888, 0.1935, 0.218, 0.2175, 0.2194, 0.2513, 0.2645, 0.329, 0.3285, 0.289, 0.311, 0.3297, 0.3379, 0.3299, 0.3716, 0.4427, 0.4173, 0.4867, 0.5123, 0.4728, 0.5022, 0.5177, 0.4599, 0.4332, 0.4275, 0.4818, 0.4902, 0.5124, 0.5124, 0.5801, 0.5627, 0.5741, 0.5717, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2019244472404697179", "ticker_idx": 1, "ts": "2026-02-05T02:59:32+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I also don't really understand this selloff… I think the magnitude of the drop is excessive. I'm really bummed that I don't have any cash right now, so I can't buy more even after this drop", "tweet_type": "analysis", "timeline": "months", "summary": "Author endorses BofA bullish chip/NVDA thesis and personally views the selloff as excessive, regretting inability to add more exposure.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "BofA (Bank of America) analyst Vivek Arya concluded that it doesn’t make logical sense for the chip sector and the software sector to sell off together. He argued that a decline in chip stocks implies deteriorating AI investment returns (ROI), while a decline in software stocks implies AI adoption will become extremely widespread—so those two conclusions are hard to reconcile at the same time. In his view, the market is overreacting.\n\nArya noted that the market’s implicit judgment behind the selloff in AI-related chip stocks is that “AI investments have deteriorated to the point where achieving ideal returns is difficult, limiting future growth runway.” By contrast, the selloff in software stocks reflects a different judgment: “AI adoption will be broadly disseminated and meaningfully boost productivity, replacing software businesses and application business models.”\n\nArya wrote that “these two outcomes cannot occur simultaneously.” He said this positioning resembles the fear seen during the “DeepSeek shock” in January this year—an episode that ultimately proved to lack factual grounding. At the time, the market worried that DeepSeek could build competitive models at lower cost, reducing demand for high-performance chips.\n\nArya believes what the market actually witnessed after the “DeepSeek shock” was more AI spending and accelerated growth in AI tokens. In his report, he estimated that 2025 cloud capex growth could reach 69%, far above the initial expectation of 20–30%.\n\nHe also argued that even if current AI models are highly capable, it may still take years to truly “prove value” from a productivity standpoint—so he does not expect AI investment to slow in the near term. Even if a major productivity breakthrough emerges in the future, models will need continuous improvement to sustain user engagement, meaning investment is unlikely to decline.\n\nHis conclusion is that while the ultimate end-state of the software industry is uncertain, the chip industry has already benefited from the AI buildout cycle—and that tailwind should continue. He added that enterprise AI adoption is still early, and more sovereign nations are beginning to increase AI investment. Arya also pointed out that chip-sector valuations already reflect concerns about slowing spend or downward revisions to earnings expectations, but those concerns “may not actually materialize.”\n\nFrom a trading perspective, this drawdown looks closer to a “narrative-driven correlation shock.” When the market sells software on the thesis that AI will displace it, while simultaneously selling AI chips on the thesis that AI ROI is worsening, core AI assets like Nvidia can easily get dragged down in the process—through position unwinds and broad risk re-rating.\n\nArya concluded that what Nvidia investors most need in the near term is clarity on whether this price action reflects a real change in AI capex and demand fundamentals, or whether it’s coming from indiscriminate, cross-sector de-risking triggered by the narrative that “AI beats software.”\n\nThat was Arya’s interpretation, and I also don’t really understand this selloff.\n\nIf I had to find a reason, you could argue that there was simply too much leverage in the system recently, and mechanical selling—like CTA-driven flows—may have contributed.\n\nStill, I think the magnitude of the drop is excessive.\n\nI’m really bummed that I don’t have any cash right now, so I can’t buy more even after this drop.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013439612035741622, "ret_signed_m1d": 0.013439612035741622, "alpha_spy_m1d": 0.0007924476720093665, "alpha_c_m1d": 0.010946679841671303, "bench_spy_m1d": 0.012647164363732255, "bench_c_m1d": 0.0024929321940703186, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07871770227202068, "ret_signed_p1d": 0.07871770227202068, "alpha_spy_p1d": 0.0595329904310693, "alpha_c_p1d": 0.024711912845628126, "bench_spy_p1d": 0.019184711840951385, "bench_c_p1d": 0.05400578942639256, "ret_p1w": 0.08761929231537935, "ret_signed_p1w": 0.08761929231537935, "alpha_spy_p1w": 0.08223276793819312, "alpha_c_p1w": 0.021909639112367207, "bench_spy_p1w": 0.00538652437718623, "bench_c_p1w": 0.06570965320301214, "ret_p1m": 0.0345590150605688, "ret_signed_p1m": 0.0345590150605688, "alpha_spy_p1m": 0.04229202249949815, "alpha_c_p1m": 0.03589737745021515, "bench_spy_p1m": -0.007733007438929351, "bench_c_p1m": -0.0013383623896463526, "ret_p3m": 0.14330135040667735, "ret_signed_p3m": 0.14330135040667735, "alpha_spy_p3m": 0.07227832985435789, "alpha_c_p3m": -0.22833638647954158, "bench_spy_p3m": 0.07102302055231946, "bench_c_p3m": 0.37163773688621893, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.158, 0.1237, 0.1275, 0.0871, 0.1063, 0.0856, 0.0551, 0.0851, 0.0509, 0.0407, 0.062, 0.0345, 0.0487, 0.0487, 0.0297, 0.0467, 0.0557, 0.0448, 0.0669, 0.0613, 0.0795, 0.0762, 0.0692, 0.0527, 0.0183, 0.0257, 0.034, -0.0055, 0.0131, 0.053, 0.0687, 0.1008, 0.0973, 0.0973, 0.1085, 0.0951, 0.0911, 0.0851, 0.0851, 0.0987, 0.0945, 0.0894, 0.1002, 0.0766, 0.0755, 0.076, 0.081, 0.0655, 0.0883, 0.0835, 0.0835, 0.036, 0.0666, 0.0754, 0.0919, 0.0849, 0.0968, 0.1143, 0.12, 0.112, 0.0799, 0.0492, 0.0134, 0.0, 0.0787, 0.1057, 0.0969, 0.1057, 0.0876, 0.0636, 0.0636, 0.0762, 0.0937, 0.0932, 0.1044, 0.1144, 0.122, 0.1378, 0.0757, 0.0309, 0.0617, 0.0475, 0.0649, 0.0667, 0.0346, 0.0627, 0.075, 0.0824, 0.0656, 0.0488, 0.066, 0.0585, 0.0496, 0.0389, 0.0048, 0.0219, 0.0194, 0.0396, -0.0037, -0.0253, -0.039, 0.0147, 0.0226, 0.0321, 0.0321, 0.0336, 0.0362, 0.0594, 0.07, 0.0975, 0.1015, 0.1434, 0.1571, 0.1541, 0.1734, 0.1757, 0.163, 0.1782, 0.1616, 0.2118, 0.2603, 0.2403, 0.2175, 0.1612, 0.1546, 0.1548, 0.1433, 0.2092, 0.2306, 0.2521, 0.2768, 0.2846, 0.314, 0.3716, 0.311, 0.2935, 0.2836, 0.3002, 0.2772, 0.2529, 0.2529, 0.2501, 0.237, 0.2466, 0.2285, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2030660410794451303", "ticker_idx": 0, "ts": "2026-03-08T15:02:24+00:00", "symbol": "3037.TW", "kind": "equity", "geography": null, "sub_sector": "ABF substrate", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Institutional investors, however, continue to maintain a Buy rating with a target price of TWD 520", "tweet_type": "news_reaction", "timeline": "months", "summary": "Relays institutional Buy on Unimicron (PT TWD 520) on EPS beat, ABF capacity expansion ~40%, LTA margin protection, and AI substrate demand into 2026–2027.", "resolved_tickers": ["3037.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Is the Stock Pullback a 'Diamond-Buying' Opportunity? IC Substrate Maker Unimicron Locks in Volume—Not Price—via Long-Term Contracts; Institutions Maintain Buy (Commercial Times Taiwan)\n\n- IC substrate and PCB major Unimicron Technology previously broke above TWD 500, hitting an all-time high, but fell to TWD 432 on March 6 amid a broader Taiwan market correction. Institutional investors, however, continue to maintain a Buy rating with a target price of TWD 520.\n\n- According to institutions, the company's preliminary January EPS came in at TWD 1.23, significantly above the market consensus of TWD 0.80, driven largely by gains on financial asset revaluations. Meanwhile, Mitsubishi Gas Chemical (MGC) raised CCL prices effective March 1, and Unimicron passed through the raw material cost increase to customers nearly simultaneously. As a result, margins are expected to show a notable improvement vs. Q4 last year, with a continued quarter-over-quarter improvement trend anticipated.\n\n- Another key factor cited by institutions is the company's long-term supply agreements (LTAs). Unimicron has signed multi-year contracts with several customers—primarily 5-year terms, with some at 3 or 7 years. Prepayments will be recognized this year and used to fund ABF substrate capacity expansion. Critically, these contracts lock in capacity but not price, allowing the company to adjust selling prices in line with market supply-demand dynamics—a structure that is highlighted as enabling effective margin defense.\n\n- Looking at 2026, the company plans a significant expansion of ABF substrate capacity. Through the Phase 3 ramp at its Guangfu plant, debottlenecking at the Yangmei facility, and expansion of mid-tier substrate lines, ABF capacity is expected to grow approximately 40%. Conversely, reflecting softening smartphone demand, BT substrate capacity will be trimmed by 15–17%, concentrating resources toward higher-value products.\n\n- The AI HDI segment is also viewed favorably. Once key equipment at the new Luzhu plant comes online in H2 and capacity adjustments are completed, both revenue and margins are expected to improve simultaneously. AI substrates currently account for roughly 60% of ABF product mix, serving GPUs, ASICs, AI peripherals, and general-purpose server CPUs. The market has also begun to raise the possibility of AI substrates entering a new supply shortage cycle by 2027.\n\n- Institutions characterize Unimicron as a key supplier to multiple ASIC companies, with de facto sole-source status on certain products. As AI accelerator demand broadens and next-generation GPU substrate mass production commences, earnings growth drivers are expected to strengthen. Given the strategy of expanding high-value ABF mix, the 2026 earnings outlook remains optimistic.\n\n---\n\nhttps://t.co/HHImNFw5gu", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.11053984575835485, "ret_signed_m1d": 0.11053984575835485, "alpha_spy_m1d": 0.11922379912271008, "alpha_c_m1d": 0.1282130107571956, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.01767316499884075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08997429305912585, "ret_signed_p1d": 0.08997429305912585, "alpha_spy_p1d": 0.09158141996449298, "alpha_c_p1d": 0.08997429305912585, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.0, "ret_p1w": 0.4113110539845759, "ret_signed_p1w": 0.4113110539845759, "alpha_spy_p1w": 0.42493396070877465, "alpha_c_p1w": 0.418323017446939, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.007011963462363102, "ret_p1m": 0.4498714652956297, "ret_signed_p1m": 0.4498714652956297, "alpha_spy_p1m": 0.4753034297152422, "alpha_c_p1m": 0.46532030365120647, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": -0.015448838355576755, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4152, -0.4216, -0.4216, -0.4306, -0.4499, -0.455, -0.4679, -0.4473, -0.4332, -0.446, -0.4422, -0.4422, -0.4332, -0.4332, -0.4357, -0.4344, -0.4344, -0.437, -0.4087, -0.4075, -0.419, -0.4152, -0.4203, -0.374, -0.347, -0.3573, -0.3586, -0.2956, -0.2262, -0.2082, -0.1748, -0.1594, -0.1285, -0.144, -0.09, -0.0553, -0.1144, -0.027, -0.0617, -0.0103, -0.0026, -0.0604, -0.126, -0.1067, -0.045, -0.0514, -0.0514, -0.0514, -0.0514, -0.0514, -0.0514, -0.0514, -0.0514, 0.0424, 0.1465, 0.1787, 0.2378, 0.2378, 0.2365, 0.1902, 0.072, 0.1568, 0.1105, 0.0, 0.09, 0.1979, 0.2339, 0.3059, 0.4113, 0.3959, 0.491, 0.4936, 0.4242, 0.2828, 0.1825, 0.3008, 0.2905, 0.2982, 0.2686, 0.1427, 0.2558, 0.3342, 0.3342, 0.3342, 0.4499, 0.5938, 0.6067, 0.6401, 0.6093, 0.5347, 0.5887, 0.6581, 0.653, 0.7429, 0.8252, 0.8406, 0.8715, 1.0308, 1.1568, 1.1208, 1.0643, 1.2699, 1.2699, 1.3419, 1.3213, 1.2057, 1.3033, 1.1028, 1.2134, 1.2494, 1.2931, 1.2648, 1.1105, 1.0977, 1.1028, 1.1157, 1.3265, 1.4936, 1.545, 1.7892, 1.7763, 1.635, 1.7121, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2017494633904017674", "ticker_idx": 0, "ts": "2026-01-31T07:06:18+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If it hasn't been fully priced in yet, then the stock price could still have a lot more room to run going forward", "tweet_type": "prediction", "timeline": "years", "summary": "Bullish SNDK if 2027 margin outlook not yet priced in; stock could have more room to run.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "Does Sandisk's recent sharp stock price surge already reflect the company's tremendous operating profit margin in 2027?\n\nIf it hasn't been fully priced in yet, then the stock price could still have a lot more room to run going forward. https://t.co/sOqJkyo7cA", "reply_text": null, "parent_text": null, "quoted_text": "Over the next three years, NAND companies are basically going to be printing money. Just look at these insane gross margins.\n\n$SNDK $MU https://t.co/aGz3aW28io", "parent_author": null, "ret_m1d": -0.13377125780280041, "ret_signed_m1d": -0.13377125780280041, "alpha_spy_m1d": -0.1288244878566457, "alpha_c_m1d": -0.12427119485539129, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04550240511035031, "ret_signed_p1d": 0.04550240511035031, "alpha_spy_p1d": 0.0539578015410509, "alpha_c_p1d": 0.06739411516269922, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": -0.12302322022384571, "ret_signed_p1w": -0.12302322022384571, "alpha_spy_p1w": -0.12092378247360513, "alpha_c_p1w": -0.10987452140516396, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": -0.1500661694355927, "ret_signed_p1m": -0.1500661694355927, "alpha_spy_p1m": -0.12838119088409294, "alpha_c_p1m": -0.09664480158260835, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 0.6482923844961672, "ret_signed_p3m": 0.6482923844961672, "alpha_spy_p3m": 0.6120365774891936, "alpha_c_p3m": 0.5488959539140126, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6746, -0.6878, -0.64, -0.5972, -0.5918, -0.5744, -0.6339, -0.6179, -0.6003, -0.6318, -0.6303, -0.7054, -0.699, -0.6588, -0.6685, -0.6767, -0.6767, -0.6644, -0.6841, -0.6913, -0.7078, -0.6793, -0.6566, -0.6611, -0.6701, -0.65, -0.6368, -0.6901, -0.6965, -0.6854, -0.6891, -0.6701, -0.6428, -0.6376, -0.6319, -0.6241, -0.6241, -0.6241, -0.6328, -0.6389, -0.6432, -0.6432, -0.5863, -0.588, -0.4744, -0.4685, -0.4971, -0.4327, -0.4148, -0.414, -0.417, -0.3848, -0.3782, -0.3782, -0.3189, -0.2465, -0.2432, -0.2877, -0.2923, -0.2763, -0.2069, -0.1893, -0.1338, 0.0, 0.0455, -0.1213, -0.1338, -0.1012, -0.123, -0.1858, -0.0991, -0.0525, -0.0581, -0.0581, -0.1122, -0.0975, -0.0664, -0.023, 0.0019, -0.0402, -0.0494, -0.0201, -0.0449, -0.0694, -0.1501, -0.0995, -0.1498, -0.2073, -0.115, -0.0697, -0.0147, -0.0698, -0.0054, 0.0577, 0.0826, 0.133, 0.1606, 0.0668, 0.056, 0.056, 0.019, -0.0933, -0.0743, -0.1394, -0.0449, 0.0413, 0.0546, 0.0546, 0.0893, 0.0685, 0.1739, 0.2801, 0.2804, 0.4318, 0.4197, 0.3404, 0.3822, 0.3844, 0.3725, 0.3581, 0.4718, 0.4016, 0.488, 0.6087, 0.5067, 0.5997, 0.6483, 0.7843, 0.8878, 1.114, 1.1195, 1.0143, 1.3485, 1.3263, 1.1827, 1.1755, 1.0785, 1.1159, 1.0038, 1.0794, 1.0933, 1.3183, 1.2228, 1.2228, 1.3894, 1.39, 1.4677, 1.5479, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008105238130303200", "ticker_idx": 0, "ts": "2026-01-05T09:16:12+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi presented Samsung Electronics' operating profit forecast for this year at 155 trillion KRW, 253% higher than last year", "tweet_type": "analysis", "timeline": "months", "summary": "DRAM boom drives 60-70% QoQ price hikes; Samsung and SK Hynix operating profits forecast up 253%/224% YoY; server DRAM ASP +144% this year.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Boom Shakes Up Hotels… ‘Business Trip War’ to Secure LTAs\n\nBusiness hotels in Pangyo, Gyeonggi Province, such as DoubleTree and Nine Tree, have been enjoying a \"semiconductor special boom\" since last month. This is due to a flood of demand for long-term stays from procurement staff visiting from the headquarters of global cloud companies like Amazon and Google, as well as smartphone and PC companies like Apple and Dell. The reason they visited Korea is the \"war to secure DRAM\" taking place among big tech companies. It is reported that they visit Samsung Electronics' DS Division in Hwaseong, Gyeonggi Province, and SK Hynix's headquarters in Icheon—located 30 to 40 minutes away—almost every day to request Long-Term Agreements (LTAs) that guarantee a certain volume for 2 to 3 years.\n\n◇ Samsung and SK Reject Long-Term Contracts\n\nAccording to the semiconductor industry on the 5th, Samsung Electronics and SK Hynix are rejecting LTAs with clients and sticking to \"quarterly\" contracts. This is because it is obvious that DRAM prices will rise in a stepwise manner every quarter until 2027. Far from long-term contracts, Samsung and SK have proposed supply prices for the first quarter of this year to major server DRAM clients that are raised by 60-70% compared to the fourth quarter of last year. They are also proposing similar price increase levels to PC and smartphone DRAM clients.\n\nBehind the unprecedentedly high price increase proposal lies a greater-than-expected demand for 5th generation High Bandwidth Memory (HBM3E). As memory companies struggle to respond to HBM3E orders, the shortage of server DRAM is intensifying. The US government's permission to export Nvidia's 'H200' AI accelerator to China served as the 'trigger.' The H200 contains eight HBM3E products. The volume of new H200 orders placed by Chinese clients since last month is estimated at $3 billion (approx. 4.3 trillion KRW). In addition, Broadcom, which develops custom AI accelerators for clients like Google's Tensor Processing Unit (TPU), is increasing its HBM3E orders, contributing to the DRAM shortage.\n\nSome analyze that weight should be placed on structural changes in the memory industry rather than immediate supply and demand. As AI inference-based services like Gemini and Copilot gain popularity, big tech companies are accelerating investment in inference servers. Whenever this happens, demand for general-purpose memory also increases.\n\n◇ “Server DRAM to Rise 144% This Year”\n\nClients seem to be accepting the price increase margins proposed by Samsung and SK. Reflecting this, market research firm DRAMeXchange presented a forecast for the fixed transaction price growth rate of server DRAM in the first quarter of this year at 60-65%. An industry insider explained, \"Clients are aware that it is difficult for Samsung and SK to increase production capacity in the short term.\"\n\nThis trend is expected to continue until the end of the year. Global investment banks (IBs) forecast that the average selling price (ASP) of server DRAM will rise by up to 144% year-on-year on an annual basis this year. Earnings forecasts are also being raised. Citi presented Samsung Electronics' operating profit forecast for this year at 155 trillion KRW, 253% higher than last year (approx. 44 trillion KRW). Morgan Stanley also viewed SK Hynix's operating profit for this year at 148 trillion KRW. This is a 224% surge compared to last year's estimate (46 trillion KRW).\n\n◇ Growing Burden on Finished Product Makers\n\nThe burden on smartphone, PC, and server companies is growing due to the sharp rise in memory prices. This is because costs are surging. According to market research firm IDC, the proportion of memory semiconductors in smartphone costs has recently risen from the 15% level to over 20%.\n\nSmartphone companies like China's Xiaomi, Oppo, and Vivo, as well as PC companies like Dell and HP, have set out to defend profitability by raising prices. Samsung Electronics, which is about to launch its premium AI phone, the Galaxy S26, next month, has also entered deep contemplation. Roh Tae-moon, Head of the Device Experience (DX) Division at Samsung Electronics, said in an interview with Reuters on this day, \"The global memory chip shortage is putting pressure on the profitability of Samsung's smartphone business.\"\n\n---\n\nhttps://t.co/fnZces6Swa", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06951485145820746, "ret_signed_m1d": -0.06951485145820746, "alpha_spy_m1d": -0.06289880128928216, "alpha_c_m1d": -0.0673022063897174, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005792875943975595, "ret_signed_p1d": 0.005792875943975595, "alpha_spy_p1d": -0.0001543074291137092, "alpha_c_p1d": -0.008243924637808986, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": 0.005068738106770221, "ret_signed_p1w": 0.005068738106770221, "alpha_spy_p1w": -0.005749621663828863, "alpha_c_p1w": -0.00993615411538884, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.21288915464418712, "ret_signed_p1m": 0.21288915464418712, "alpha_spy_p1m": 0.21025720473907072, "alpha_c_p1m": 0.23045234840005946, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": 0.29449133710288633, "ret_signed_p3m": 0.29449133710288633, "alpha_spy_p3m": 0.33825752704325784, "alpha_c_p3m": 0.3529957010157597, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3533, -0.3533, -0.3197, -0.3277, -0.3399, -0.3154, -0.296, -0.2945, -0.2931, -0.2974, -0.2895, -0.3046, -0.288, -0.265, -0.283, -0.2758, -0.2498, -0.2253, -0.1994, -0.2441, -0.2751, -0.2852, -0.2945, -0.2751, -0.2542, -0.2571, -0.2592, -0.2996, -0.2751, -0.2952, -0.3046, -0.2751, -0.3169, -0.3032, -0.2844, -0.2592, -0.2542, -0.2758, -0.2736, -0.2549, -0.247, -0.2426, -0.2189, -0.2109, -0.2189, -0.2217, -0.2268, -0.2153, -0.2448, -0.2592, -0.2225, -0.2246, -0.234, -0.2037, -0.1965, -0.1994, -0.1994, -0.1569, -0.1347, -0.1318, -0.1318, -0.1318, -0.0695, 0.0, 0.0058, 0.021, 0.0051, 0.0065, 0.0051, -0.0036, 0.0159, 0.042, 0.0782, 0.0811, 0.0514, 0.0825, 0.1028, 0.1014, 0.1014, 0.155, 0.176, 0.1636, 0.1622, 0.0891, 0.2129, 0.2245, 0.1535, 0.1484, 0.2049, 0.2006, 0.2151, 0.2933, 0.3121, 0.3121, 0.3121, 0.3121, 0.3758, 0.3765, 0.3975, 0.4482, 0.4736, 0.5786, 0.5677, 0.5677, 0.4127, 0.2469, 0.3874, 0.3628, 0.2563, 0.3606, 0.3758, 0.3606, 0.3287, 0.3664, 0.4041, 0.5098, 0.4518, 0.4439, 0.349, 0.3736, 0.3686, 0.3041, 0.3041, 0.2793, 0.2132, 0.3761, 0.2945, 0.3511, 0.4012, 0.4258, 0.5274, 0.4802, 0.4948, 0.4585, 0.4984, 0.531, 0.5782, 0.5673, 0.5564, 0.5891, 0.5782, 0.629, 0.5927, 0.629, 0.6109, 0.6399, 0.6, 0.6, 0.687, 0.687, 0.9301, 0.97, 0.9483, 1.0716, 1.0245, 1.0607, 1.1478, 0.9628, 1.039, 0.9991, 1.0027, 1.1732, 1.1224, 1.1224, 1.1696, 1.2276, 1.1732, 1.3002, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008105238130303200", "ticker_idx": 1, "ts": "2026-01-05T09:16:12+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley also viewed SK Hynix's operating profit for this year at 148 trillion KRW. This is a 224% surge compared to last year's estimate", "tweet_type": "analysis", "timeline": "months", "summary": "DRAM boom drives 60-70% QoQ price hikes; Samsung and SK Hynix operating profits forecast up 253%/224% YoY; server DRAM ASP +144% this year.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Boom Shakes Up Hotels… ‘Business Trip War’ to Secure LTAs\n\nBusiness hotels in Pangyo, Gyeonggi Province, such as DoubleTree and Nine Tree, have been enjoying a \"semiconductor special boom\" since last month. This is due to a flood of demand for long-term stays from procurement staff visiting from the headquarters of global cloud companies like Amazon and Google, as well as smartphone and PC companies like Apple and Dell. The reason they visited Korea is the \"war to secure DRAM\" taking place among big tech companies. It is reported that they visit Samsung Electronics' DS Division in Hwaseong, Gyeonggi Province, and SK Hynix's headquarters in Icheon—located 30 to 40 minutes away—almost every day to request Long-Term Agreements (LTAs) that guarantee a certain volume for 2 to 3 years.\n\n◇ Samsung and SK Reject Long-Term Contracts\n\nAccording to the semiconductor industry on the 5th, Samsung Electronics and SK Hynix are rejecting LTAs with clients and sticking to \"quarterly\" contracts. This is because it is obvious that DRAM prices will rise in a stepwise manner every quarter until 2027. Far from long-term contracts, Samsung and SK have proposed supply prices for the first quarter of this year to major server DRAM clients that are raised by 60-70% compared to the fourth quarter of last year. They are also proposing similar price increase levels to PC and smartphone DRAM clients.\n\nBehind the unprecedentedly high price increase proposal lies a greater-than-expected demand for 5th generation High Bandwidth Memory (HBM3E). As memory companies struggle to respond to HBM3E orders, the shortage of server DRAM is intensifying. The US government's permission to export Nvidia's 'H200' AI accelerator to China served as the 'trigger.' The H200 contains eight HBM3E products. The volume of new H200 orders placed by Chinese clients since last month is estimated at $3 billion (approx. 4.3 trillion KRW). In addition, Broadcom, which develops custom AI accelerators for clients like Google's Tensor Processing Unit (TPU), is increasing its HBM3E orders, contributing to the DRAM shortage.\n\nSome analyze that weight should be placed on structural changes in the memory industry rather than immediate supply and demand. As AI inference-based services like Gemini and Copilot gain popularity, big tech companies are accelerating investment in inference servers. Whenever this happens, demand for general-purpose memory also increases.\n\n◇ “Server DRAM to Rise 144% This Year”\n\nClients seem to be accepting the price increase margins proposed by Samsung and SK. Reflecting this, market research firm DRAMeXchange presented a forecast for the fixed transaction price growth rate of server DRAM in the first quarter of this year at 60-65%. An industry insider explained, \"Clients are aware that it is difficult for Samsung and SK to increase production capacity in the short term.\"\n\nThis trend is expected to continue until the end of the year. Global investment banks (IBs) forecast that the average selling price (ASP) of server DRAM will rise by up to 144% year-on-year on an annual basis this year. Earnings forecasts are also being raised. Citi presented Samsung Electronics' operating profit forecast for this year at 155 trillion KRW, 253% higher than last year (approx. 44 trillion KRW). Morgan Stanley also viewed SK Hynix's operating profit for this year at 148 trillion KRW. This is a 224% surge compared to last year's estimate (46 trillion KRW).\n\n◇ Growing Burden on Finished Product Makers\n\nThe burden on smartphone, PC, and server companies is growing due to the sharp rise in memory prices. This is because costs are surging. According to market research firm IDC, the proportion of memory semiconductors in smartphone costs has recently risen from the 15% level to over 20%.\n\nSmartphone companies like China's Xiaomi, Oppo, and Vivo, as well as PC companies like Dell and HP, have set out to defend profitability by raising prices. Samsung Electronics, which is about to launch its premium AI phone, the Galaxy S26, next month, has also entered deep contemplation. Roh Tae-moon, Head of the Device Experience (DX) Division at Samsung Electronics, said in an interview with Reuters on this day, \"The global memory chip shortage is putting pressure on the profitability of Samsung's smartphone business.\"\n\n---\n\nhttps://t.co/fnZces6Swa", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027298908880455053, "ret_signed_m1d": -0.027298908880455053, "alpha_spy_m1d": -0.02068285871152975, "alpha_c_m1d": -0.015937356438575967, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0431033745856777, "ret_signed_p1d": 0.0431033745856777, "alpha_spy_p1d": 0.037156191212588396, "alpha_c_p1d": 0.016540174086262516, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.07614937008435851, "ret_signed_p1w": 0.07614937008435851, "alpha_spy_p1w": 0.06533101031375943, "alpha_c_p1w": 0.04153508620242308, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.30316086614716653, "ret_signed_p1m": 0.30316086614716653, "alpha_spy_p1m": 0.30052891624205014, "alpha_c_p1m": 0.2499550171237912, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.19472924735399255, "ret_signed_p3m": 0.19472924735399255, "alpha_spy_p3m": 0.23849543729436407, "alpha_c_p3m": 0.15571865086258874, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4322, -0.4322, -0.3855, -0.4042, -0.4092, -0.3934, -0.3503, -0.3317, -0.3029, -0.3123, -0.3087, -0.313, -0.2678, -0.2319, -0.252, -0.1988, -0.1845, -0.1974, -0.1098, -0.1586, -0.1687, -0.1486, -0.1673, -0.1299, -0.1113, -0.1141, -0.1213, -0.196, -0.1299, -0.1816, -0.1931, -0.1802, -0.252, -0.2534, -0.2548, -0.2477, -0.2184, -0.2385, -0.227, -0.1983, -0.2069, -0.2213, -0.2184, -0.171, -0.1868, -0.1566, -0.1882, -0.1796, -0.204, -0.2385, -0.2083, -0.2069, -0.2141, -0.1667, -0.1609, -0.1552, -0.1552, -0.1394, -0.0805, -0.0647, -0.0647, -0.0647, -0.0273, 0.0, 0.0431, 0.0661, 0.0862, 0.069, 0.0761, 0.0603, 0.0661, 0.0761, 0.0862, 0.0977, 0.0675, 0.0632, 0.0848, 0.102, 0.0575, 0.1494, 0.2083, 0.2371, 0.306, 0.1925, 0.3032, 0.2931, 0.2098, 0.2055, 0.2744, 0.2586, 0.2356, 0.2759, 0.2644, 0.2644, 0.2644, 0.2644, 0.2845, 0.3635, 0.3664, 0.444, 0.4626, 0.5819, 0.5272, 0.5272, 0.3516, 0.2221, 0.3545, 0.33, 0.2034, 0.3502, 0.3747, 0.3387, 0.3099, 0.402, 0.3962, 0.52, 0.4581, 0.4495, 0.343, 0.4193, 0.4322, 0.343, 0.343, 0.2566, 0.1616, 0.2912, 0.1947, 0.2609, 0.2753, 0.3185, 0.4869, 0.4366, 0.4783, 0.497, 0.5877, 0.6352, 0.6625, 0.6237, 0.6784, 0.7619, 0.7604, 0.7633, 0.759, 0.8597, 0.8713, 0.8612, 0.8511, 0.8511, 1.0829, 1.0829, 1.3045, 1.3808, 1.4269, 1.7061, 1.6414, 1.8443, 1.8357, 1.6183, 1.6486, 1.5118, 1.5118, 1.7925, 1.7939, 1.7939, 1.9537, 2.2286, 2.2954, 2.3588, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008105238130303200", "ticker_idx": 2, "ts": "2026-01-05T09:16:12+00:00", "symbol": "DRAM", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "server DRAM will rise by up to 144% year-on-year on an annual basis this year", "tweet_type": "analysis", "timeline": "months", "summary": "DRAM boom drives 60-70% QoQ price hikes; Samsung and SK Hynix operating profits forecast up 253%/224% YoY; server DRAM ASP +144% this year.", "resolved_tickers": ["005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: Samsung Electronics + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Boom Shakes Up Hotels… ‘Business Trip War’ to Secure LTAs\n\nBusiness hotels in Pangyo, Gyeonggi Province, such as DoubleTree and Nine Tree, have been enjoying a \"semiconductor special boom\" since last month. This is due to a flood of demand for long-term stays from procurement staff visiting from the headquarters of global cloud companies like Amazon and Google, as well as smartphone and PC companies like Apple and Dell. The reason they visited Korea is the \"war to secure DRAM\" taking place among big tech companies. It is reported that they visit Samsung Electronics' DS Division in Hwaseong, Gyeonggi Province, and SK Hynix's headquarters in Icheon—located 30 to 40 minutes away—almost every day to request Long-Term Agreements (LTAs) that guarantee a certain volume for 2 to 3 years.\n\n◇ Samsung and SK Reject Long-Term Contracts\n\nAccording to the semiconductor industry on the 5th, Samsung Electronics and SK Hynix are rejecting LTAs with clients and sticking to \"quarterly\" contracts. This is because it is obvious that DRAM prices will rise in a stepwise manner every quarter until 2027. Far from long-term contracts, Samsung and SK have proposed supply prices for the first quarter of this year to major server DRAM clients that are raised by 60-70% compared to the fourth quarter of last year. They are also proposing similar price increase levels to PC and smartphone DRAM clients.\n\nBehind the unprecedentedly high price increase proposal lies a greater-than-expected demand for 5th generation High Bandwidth Memory (HBM3E). As memory companies struggle to respond to HBM3E orders, the shortage of server DRAM is intensifying. The US government's permission to export Nvidia's 'H200' AI accelerator to China served as the 'trigger.' The H200 contains eight HBM3E products. The volume of new H200 orders placed by Chinese clients since last month is estimated at $3 billion (approx. 4.3 trillion KRW). In addition, Broadcom, which develops custom AI accelerators for clients like Google's Tensor Processing Unit (TPU), is increasing its HBM3E orders, contributing to the DRAM shortage.\n\nSome analyze that weight should be placed on structural changes in the memory industry rather than immediate supply and demand. As AI inference-based services like Gemini and Copilot gain popularity, big tech companies are accelerating investment in inference servers. Whenever this happens, demand for general-purpose memory also increases.\n\n◇ “Server DRAM to Rise 144% This Year”\n\nClients seem to be accepting the price increase margins proposed by Samsung and SK. Reflecting this, market research firm DRAMeXchange presented a forecast for the fixed transaction price growth rate of server DRAM in the first quarter of this year at 60-65%. An industry insider explained, \"Clients are aware that it is difficult for Samsung and SK to increase production capacity in the short term.\"\n\nThis trend is expected to continue until the end of the year. Global investment banks (IBs) forecast that the average selling price (ASP) of server DRAM will rise by up to 144% year-on-year on an annual basis this year. Earnings forecasts are also being raised. Citi presented Samsung Electronics' operating profit forecast for this year at 155 trillion KRW, 253% higher than last year (approx. 44 trillion KRW). Morgan Stanley also viewed SK Hynix's operating profit for this year at 148 trillion KRW. This is a 224% surge compared to last year's estimate (46 trillion KRW).\n\n◇ Growing Burden on Finished Product Makers\n\nThe burden on smartphone, PC, and server companies is growing due to the sharp rise in memory prices. This is because costs are surging. According to market research firm IDC, the proportion of memory semiconductors in smartphone costs has recently risen from the 15% level to over 20%.\n\nSmartphone companies like China's Xiaomi, Oppo, and Vivo, as well as PC companies like Dell and HP, have set out to defend profitability by raising prices. Samsung Electronics, which is about to launch its premium AI phone, the Galaxy S26, next month, has also entered deep contemplation. Roh Tae-moon, Head of the Device Experience (DX) Division at Samsung Electronics, said in an interview with Reuters on this day, \"The global memory chip shortage is putting pressure on the profitability of Samsung's smartphone business.\"\n\n---\n\nhttps://t.co/fnZces6Swa", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04840688016933126, "ret_signed_m1d": -0.04840688016933126, "alpha_spy_m1d": -0.041790830000405954, "alpha_c_m1d": -0.0461942351008412, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024448125264826648, "ret_signed_p1d": 0.024448125264826648, "alpha_spy_p1d": 0.018500941891737344, "alpha_c_p1d": 0.010411324683042067, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": 0.04060905409556437, "ret_signed_p1w": 0.04060905409556437, "alpha_spy_p1w": 0.029790694324965283, "alpha_c_p1w": 0.025604161873405307, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.2580250103956768, "ret_signed_p1m": 0.2580250103956768, "alpha_spy_p1m": 0.25539306049056043, "alpha_c_p1m": 0.27558820415154917, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": 0.24461029222843944, "ret_signed_p3m": 0.24461029222843944, "alpha_spy_p3m": 0.28837648216881095, "alpha_c_p3m": 0.30311465614131283, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3927, -0.3927, -0.3526, -0.3659, -0.3746, -0.3544, -0.3231, -0.3131, -0.298, -0.3048, -0.2991, -0.3088, -0.2779, -0.2484, -0.2675, -0.2373, -0.2172, -0.2114, -0.1546, -0.2014, -0.2219, -0.2169, -0.2309, -0.2025, -0.1827, -0.1856, -0.1903, -0.2478, -0.2025, -0.2384, -0.2489, -0.2276, -0.2844, -0.2783, -0.2696, -0.2534, -0.2363, -0.2571, -0.2503, -0.2266, -0.2269, -0.232, -0.2186, -0.191, -0.2028, -0.1892, -0.2075, -0.1974, -0.2244, -0.2489, -0.2154, -0.2158, -0.224, -0.1852, -0.1787, -0.1773, -0.1773, -0.1481, -0.1076, -0.0982, -0.0982, -0.0982, -0.0484, 0.0, 0.0244, 0.0435, 0.0456, 0.0377, 0.0406, 0.0284, 0.041, 0.0591, 0.0822, 0.0894, 0.0595, 0.0729, 0.0938, 0.1017, 0.0794, 0.1522, 0.1921, 0.2004, 0.2341, 0.1408, 0.258, 0.2588, 0.1816, 0.177, 0.2397, 0.2296, 0.2253, 0.2846, 0.2882, 0.2882, 0.2882, 0.2882, 0.3301, 0.37, 0.382, 0.4461, 0.4681, 0.5803, 0.5475, 0.5475, 0.3822, 0.2345, 0.371, 0.3464, 0.2299, 0.3554, 0.3752, 0.3496, 0.3193, 0.3842, 0.4002, 0.5149, 0.455, 0.4467, 0.346, 0.3965, 0.4004, 0.3236, 0.3236, 0.2679, 0.1874, 0.3336, 0.2446, 0.306, 0.3382, 0.3722, 0.5072, 0.4584, 0.4865, 0.4777, 0.543, 0.5831, 0.6204, 0.5955, 0.6174, 0.6755, 0.6693, 0.6962, 0.6759, 0.7444, 0.7411, 0.7505, 0.7255, 0.7255, 0.885, 0.885, 1.1173, 1.1754, 1.1876, 1.3889, 1.3329, 1.4525, 1.4917, 1.2906, 1.3438, 1.2554, 1.2572, 1.4829, 1.4582, 1.4582, 1.5616, 1.7281, 1.7343, 1.8295, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029098861789364740", "ticker_idx": 0, "ts": "2026-03-04T07:37:22+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM and NAND prices will continue to rise in Q2 as well", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM prices doubled Q/Q and set to keep rising into Q2; Samsung, Hynix, Micron all benefiting from AI-driven supply crunch.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM memory basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Confirms Q1 DRAM Price Hike at '70→100%'… Another Jump in Just One Month\n\nSamsung Electronics has confirmed a Q1 DRAM price increase of over 100%. The price hike, initially negotiated at around 70% in January, widened further in just one month. The aftershock of surging memory demand driven by the spread of artificial intelligence (AI) has now triggered month-over-month price swings in DRAM.\n\nAccording to industry sources on the 4th, Samsung Electronics finalized Q1 DRAM supply pricing with major customers last month. The average price increase for commodity DRAM used in servers, PCs, and mobile devices is estimated at approximately 100% quarter-over-quarter. This means prices have effectively doubled compared to Q4 last year, with some customers and products reportedly seeing hikes exceeding 100%.\n\nAn industry source familiar with the matter said, \"Samsung Electronics has wrapped up DRAM supply price negotiations, and some overseas customers have already completed payment,\" adding that \"additional increases were made to reflect price movements between January and February.\"\n\nAs of January, Samsung's quarterly DRAM price increase stood at 70%, while NAND was at around 100% (refer to the January 26 front page of ETNews). Even then, the steep hikes caused significant ripples across the industry. Despite the sharp rise in memory prices, customers continued to line up. Some overseas Big Tech firms reportedly visited Korea to proactively secure supply volumes by meeting with Samsung Electronics and other memory manufacturers.\n\nSamsung Electronics has been conducting ongoing price negotiations with customers in consideration of its supply capacity. Demand continued to surge even during supply discussions, leading to further price increases following the January round.\n\nMemory supply contracts for DRAM and other products were traditionally done on an annual basis. Due to recent supply shortages, memory manufacturers have been pushing for quarterly contracts — but with prices now fluctuating so rapidly, even monthly pricing is becoming a consideration.\n\nSamsung Electronics is not the only one continuously raising memory prices. The industry estimates that SK Hynix and Micron have also closed Q1 supply contracts at similar levels.\n\nThe global expansion of AI infrastructure investment is what has driven this extraordinary situation. As more AI service providers (data centers) purchase AI semiconductor chips, demand for high-bandwidth memory (HBM) to support computation has surged. With Samsung Electronics, SK Hynix, and Micron concentrating their production capacity on HBM, supply of commodity DRAM for servers, PCs, and mobile devices has become constrained. Demand is overflowing — not only for AI servers but also AI PCs and AI smartphones — yet supply remains tight, pushing prices through the roof.\n\nWith AI growth still ongoing, memory price increases are expected to persist for some time. NVIDIA, the world's largest AI chip company, posted record-breaking earnings on the 25th of last month, dispelling concerns over an AI bubble. Market research firm Gartner forecasts that DRAM and solid-state drive (SSD) prices will rise 130% year-over-year (on a combined basis) this year.\n\nAn industry source said, \"DRAM and NAND prices will continue to rise in Q2 as well,\" adding that \"while the pace of increase may moderate, price hikes themselves are unavoidable at this point.\"\n\n---\n\nhttps://t.co/GoF4iSVQFR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06212278052658804, "ret_signed_m1d": 0.06212278052658804, "alpha_spy_m1d": 0.06912866023802389, "alpha_c_m1d": 0.08226812861469734, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.020145348088109305, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07058015472998487, "ret_signed_p1d": 0.07058015472998487, "alpha_spy_p1d": 0.0761557015763787, "alpha_c_p1d": 0.07997629590425322, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.009396141174268346, "ret_p1w": 0.09097830775741689, "ret_signed_p1w": 0.09097830775741689, "alpha_spy_p1w": 0.10382253966679245, "alpha_c_p1w": 0.08614244545156491, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004835862305851979, "ret_p1m": -0.02333515845309227, "ret_signed_p1m": -0.02333515845309227, "alpha_spy_p1m": 0.016816175266201088, "alpha_c_p1m": -0.006346938268654807, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.016988220184437464, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3808, -0.3577, -0.3595, -0.3427, -0.357, -0.366, -0.3835, -0.401, -0.3887, -0.3698, -0.3598, -0.3299, -0.3267, -0.3172, -0.3172, -0.3031, -0.273, -0.2694, -0.2754, -0.2754, -0.2236, -0.2003, -0.161, -0.1539, -0.1631, -0.1523, -0.1501, -0.1632, -0.1604, -0.1479, -0.1138, -0.1099, -0.1242, -0.097, -0.0786, -0.0726, -0.0935, -0.0365, 0.006, 0.011, 0.012, -0.0195, 0.0285, -0.0044, -0.0432, -0.0359, -0.0113, -0.0253, 0.0031, 0.038, 0.038, 0.038, 0.0281, 0.0457, 0.0653, 0.096, 0.0964, 0.1287, 0.1497, 0.1991, 0.1786, 0.1789, 0.0621, 0.0, 0.0706, 0.0351, -0.0121, 0.0673, 0.091, 0.066, 0.0669, 0.1151, 0.1402, 0.2022, 0.1554, 0.1331, 0.0633, 0.0833, 0.0743, 0.0106, 0.0121, -0.0475, -0.0777, 0.0261, -0.0233, 0.0099, 0.0368, 0.055, 0.1523, 0.1383, 0.1528, 0.1531, 0.2211, 0.2349, 0.2558, 0.2405, 0.247, 0.2793, 0.3077, 0.3173, 0.3189, 0.3792, 0.3606, 0.3774, 0.3629, 0.3838, 0.4988, 0.5518, 0.6992, 0.7141, 0.8042, 0.9537, 0.8995, 0.9954, 0.9933, 0.8418, 0.8346, 0.8009, 0.8296, 0.9768, 0.9543, 0.9543, 1.1311, 1.2487, 1.2483, 1.339, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029098861789364740", "ticker_idx": 1, "ts": "2026-03-04T07:37:22+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics has confirmed a Q1 DRAM price increase of over 100%", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM prices doubled Q/Q and set to keep rising into Q2; Samsung, Hynix, Micron all benefiting from AI-driven supply crunch.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Confirms Q1 DRAM Price Hike at '70→100%'… Another Jump in Just One Month\n\nSamsung Electronics has confirmed a Q1 DRAM price increase of over 100%. The price hike, initially negotiated at around 70% in January, widened further in just one month. The aftershock of surging memory demand driven by the spread of artificial intelligence (AI) has now triggered month-over-month price swings in DRAM.\n\nAccording to industry sources on the 4th, Samsung Electronics finalized Q1 DRAM supply pricing with major customers last month. The average price increase for commodity DRAM used in servers, PCs, and mobile devices is estimated at approximately 100% quarter-over-quarter. This means prices have effectively doubled compared to Q4 last year, with some customers and products reportedly seeing hikes exceeding 100%.\n\nAn industry source familiar with the matter said, \"Samsung Electronics has wrapped up DRAM supply price negotiations, and some overseas customers have already completed payment,\" adding that \"additional increases were made to reflect price movements between January and February.\"\n\nAs of January, Samsung's quarterly DRAM price increase stood at 70%, while NAND was at around 100% (refer to the January 26 front page of ETNews). Even then, the steep hikes caused significant ripples across the industry. Despite the sharp rise in memory prices, customers continued to line up. Some overseas Big Tech firms reportedly visited Korea to proactively secure supply volumes by meeting with Samsung Electronics and other memory manufacturers.\n\nSamsung Electronics has been conducting ongoing price negotiations with customers in consideration of its supply capacity. Demand continued to surge even during supply discussions, leading to further price increases following the January round.\n\nMemory supply contracts for DRAM and other products were traditionally done on an annual basis. Due to recent supply shortages, memory manufacturers have been pushing for quarterly contracts — but with prices now fluctuating so rapidly, even monthly pricing is becoming a consideration.\n\nSamsung Electronics is not the only one continuously raising memory prices. The industry estimates that SK Hynix and Micron have also closed Q1 supply contracts at similar levels.\n\nThe global expansion of AI infrastructure investment is what has driven this extraordinary situation. As more AI service providers (data centers) purchase AI semiconductor chips, demand for high-bandwidth memory (HBM) to support computation has surged. With Samsung Electronics, SK Hynix, and Micron concentrating their production capacity on HBM, supply of commodity DRAM for servers, PCs, and mobile devices has become constrained. Demand is overflowing — not only for AI servers but also AI PCs and AI smartphones — yet supply remains tight, pushing prices through the roof.\n\nWith AI growth still ongoing, memory price increases are expected to persist for some time. NVIDIA, the world's largest AI chip company, posted record-breaking earnings on the 25th of last month, dispelling concerns over an AI bubble. Market research firm Gartner forecasts that DRAM and solid-state drive (SSD) prices will rise 130% year-over-year (on a combined basis) this year.\n\nAn industry source said, \"DRAM and NAND prices will continue to rise in Q2 as well,\" adding that \"while the pace of increase may moderate, price hikes themselves are unavoidable at this point.\"\n\n---\n\nhttps://t.co/GoF4iSVQFR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.13298492567218356, "ret_signed_m1d": 0.13298492567218356, "alpha_spy_m1d": 0.1399908053836194, "alpha_c_m1d": 0.14971824864116934, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.016733322968985775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11265975125013239, "ret_signed_p1d": 0.11265975125013239, "alpha_spy_p1d": 0.11823529809652622, "alpha_c_p1d": 0.11022835396309327, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": 0.0024313972870391165, "ret_p1w": 0.10336818319198082, "ret_signed_p1w": 0.10336818319198082, "alpha_spy_p1w": 0.11621241510135638, "alpha_c_p1w": 0.09914909982056419, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004219083371416632, "ret_p1m": 0.03814899314750386, "ret_signed_p1m": 0.03814899314750386, "alpha_spy_p1m": 0.07830032686679722, "alpha_c_p1m": 0.06447123134839883, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.026322238200894965, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3735, -0.3672, -0.3735, -0.3759, -0.3799, -0.3707, -0.3943, -0.4059, -0.3764, -0.3782, -0.3857, -0.3614, -0.3556, -0.3579, -0.3579, -0.3238, -0.306, -0.3037, -0.3037, -0.3037, -0.2538, -0.198, -0.1934, -0.1812, -0.194, -0.1928, -0.194, -0.2009, -0.1852, -0.1643, -0.1353, -0.133, -0.1568, -0.1318, -0.1156, -0.1167, -0.1167, -0.0738, -0.0569, -0.0668, -0.0679, -0.1266, -0.0273, -0.018, -0.0749, -0.079, -0.0337, -0.0372, -0.0256, 0.0372, 0.0523, 0.0523, 0.0523, 0.0523, 0.1034, 0.1039, 0.1208, 0.1614, 0.1818, 0.266, 0.2573, 0.2573, 0.133, 0.0, 0.1127, 0.0929, 0.0075, 0.0912, 0.1034, 0.0912, 0.0656, 0.0958, 0.126, 0.2108, 0.1643, 0.158, 0.0819, 0.1016, 0.0976, 0.0459, 0.0459, 0.0259, -0.027, 0.1036, 0.0381, 0.0835, 0.1237, 0.1435, 0.2249, 0.1871, 0.1988, 0.1697, 0.2017, 0.2279, 0.2657, 0.257, 0.2482, 0.2744, 0.2657, 0.3064, 0.2773, 0.3064, 0.2919, 0.3151, 0.2831, 0.2831, 0.353, 0.353, 0.5479, 0.5799, 0.5625, 0.6614, 0.6236, 0.6527, 0.7225, 0.5741, 0.6352, 0.6032, 0.6061, 0.7429, 0.7021, 0.7021, 0.7399, 0.7865, 0.7429, 0.8447, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029098861789364740", "ticker_idx": 2, "ts": "2026-03-04T07:37:22+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron have also closed Q1 supply contracts at similar levels", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM prices doubled Q/Q and set to keep rising into Q2; Samsung, Hynix, Micron all benefiting from AI-driven supply crunch.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Confirms Q1 DRAM Price Hike at '70→100%'… Another Jump in Just One Month\n\nSamsung Electronics has confirmed a Q1 DRAM price increase of over 100%. The price hike, initially negotiated at around 70% in January, widened further in just one month. The aftershock of surging memory demand driven by the spread of artificial intelligence (AI) has now triggered month-over-month price swings in DRAM.\n\nAccording to industry sources on the 4th, Samsung Electronics finalized Q1 DRAM supply pricing with major customers last month. The average price increase for commodity DRAM used in servers, PCs, and mobile devices is estimated at approximately 100% quarter-over-quarter. This means prices have effectively doubled compared to Q4 last year, with some customers and products reportedly seeing hikes exceeding 100%.\n\nAn industry source familiar with the matter said, \"Samsung Electronics has wrapped up DRAM supply price negotiations, and some overseas customers have already completed payment,\" adding that \"additional increases were made to reflect price movements between January and February.\"\n\nAs of January, Samsung's quarterly DRAM price increase stood at 70%, while NAND was at around 100% (refer to the January 26 front page of ETNews). Even then, the steep hikes caused significant ripples across the industry. Despite the sharp rise in memory prices, customers continued to line up. Some overseas Big Tech firms reportedly visited Korea to proactively secure supply volumes by meeting with Samsung Electronics and other memory manufacturers.\n\nSamsung Electronics has been conducting ongoing price negotiations with customers in consideration of its supply capacity. Demand continued to surge even during supply discussions, leading to further price increases following the January round.\n\nMemory supply contracts for DRAM and other products were traditionally done on an annual basis. Due to recent supply shortages, memory manufacturers have been pushing for quarterly contracts — but with prices now fluctuating so rapidly, even monthly pricing is becoming a consideration.\n\nSamsung Electronics is not the only one continuously raising memory prices. The industry estimates that SK Hynix and Micron have also closed Q1 supply contracts at similar levels.\n\nThe global expansion of AI infrastructure investment is what has driven this extraordinary situation. As more AI service providers (data centers) purchase AI semiconductor chips, demand for high-bandwidth memory (HBM) to support computation has surged. With Samsung Electronics, SK Hynix, and Micron concentrating their production capacity on HBM, supply of commodity DRAM for servers, PCs, and mobile devices has become constrained. Demand is overflowing — not only for AI servers but also AI PCs and AI smartphones — yet supply remains tight, pushing prices through the roof.\n\nWith AI growth still ongoing, memory price increases are expected to persist for some time. NVIDIA, the world's largest AI chip company, posted record-breaking earnings on the 25th of last month, dispelling concerns over an AI bubble. Market research firm Gartner forecasts that DRAM and solid-state drive (SSD) prices will rise 130% year-over-year (on a combined basis) this year.\n\nAn industry source said, \"DRAM and NAND prices will continue to rise in Q2 as well,\" adding that \"while the pace of increase may moderate, price hikes themselves are unavoidable at this point.\"\n\n---\n\nhttps://t.co/GoF4iSVQFR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.10600704814534301, "ret_signed_m1d": 0.10600704814534301, "alpha_spy_m1d": 0.11301292785677886, "alpha_c_m1d": 0.1261523962334523, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.020145348088109305, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10836278650531161, "ret_signed_p1d": 0.10836278650531161, "alpha_spy_p1d": 0.11393833335170545, "alpha_c_p1d": 0.11775892767957996, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.009396141174268346, "ret_p1w": 0.12485273414010845, "ret_signed_p1w": 0.12485273414010845, "alpha_spy_p1w": 0.137696966049484, "alpha_c_p1w": 0.12001687183425647, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004835862305851979, "ret_p1m": -0.02237925672055452, "ret_signed_p1m": -0.02237925672055452, "alpha_spy_p1m": 0.01777207699873884, "alpha_c_p1m": -0.005391036536117055, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.016988220184437464, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3604, -0.3216, -0.3346, -0.3099, -0.3357, -0.3287, -0.3487, -0.3769, -0.3522, -0.351, -0.3569, -0.3181, -0.3134, -0.3087, -0.3087, -0.2958, -0.2476, -0.2346, -0.2346, -0.2346, -0.2041, -0.1817, -0.1465, -0.1276, -0.1112, -0.1253, -0.1194, -0.1323, -0.1276, -0.1194, -0.1112, -0.1018, -0.1265, -0.13, -0.1124, -0.0982, -0.1347, -0.0595, -0.0112, 0.0123, 0.0687, -0.0242, 0.0663, 0.0581, -0.0101, -0.0136, 0.0428, 0.0299, 0.0111, 0.044, 0.0346, 0.0346, 0.0346, 0.0346, 0.0511, 0.1157, 0.1181, 0.1816, 0.1968, 0.2945, 0.2497, 0.2497, 0.106, 0.0, 0.1084, 0.0883, -0.0153, 0.1048, 0.1249, 0.0954, 0.0718, 0.1472, 0.1425, 0.2438, 0.1932, 0.1861, 0.0989, 0.1614, 0.172, 0.0989, 0.0989, 0.0283, -0.0495, 0.0565, -0.0224, 0.0318, 0.0436, 0.0789, 0.2167, 0.1755, 0.2097, 0.225, 0.2992, 0.338, 0.3604, 0.3286, 0.3734, 0.4417, 0.4405, 0.4429, 0.4393, 0.5218, 0.5312, 0.523, 0.5147, 0.5147, 0.7044, 0.7044, 0.8857, 0.9482, 0.9859, 1.2144, 1.1614, 1.3274, 1.3204, 1.1425, 1.1673, 1.0554, 1.0554, 1.285, 1.2862, 1.2862, 1.417, 1.6419, 1.6966, 1.7484, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029098861789364740", "ticker_idx": 3, "ts": "2026-03-04T07:37:22+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron have also closed Q1 supply contracts at similar levels", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM prices doubled Q/Q and set to keep rising into Q2; Samsung, Hynix, Micron all benefiting from AI-driven supply crunch.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Confirms Q1 DRAM Price Hike at '70→100%'… Another Jump in Just One Month\n\nSamsung Electronics has confirmed a Q1 DRAM price increase of over 100%. The price hike, initially negotiated at around 70% in January, widened further in just one month. The aftershock of surging memory demand driven by the spread of artificial intelligence (AI) has now triggered month-over-month price swings in DRAM.\n\nAccording to industry sources on the 4th, Samsung Electronics finalized Q1 DRAM supply pricing with major customers last month. The average price increase for commodity DRAM used in servers, PCs, and mobile devices is estimated at approximately 100% quarter-over-quarter. This means prices have effectively doubled compared to Q4 last year, with some customers and products reportedly seeing hikes exceeding 100%.\n\nAn industry source familiar with the matter said, \"Samsung Electronics has wrapped up DRAM supply price negotiations, and some overseas customers have already completed payment,\" adding that \"additional increases were made to reflect price movements between January and February.\"\n\nAs of January, Samsung's quarterly DRAM price increase stood at 70%, while NAND was at around 100% (refer to the January 26 front page of ETNews). Even then, the steep hikes caused significant ripples across the industry. Despite the sharp rise in memory prices, customers continued to line up. Some overseas Big Tech firms reportedly visited Korea to proactively secure supply volumes by meeting with Samsung Electronics and other memory manufacturers.\n\nSamsung Electronics has been conducting ongoing price negotiations with customers in consideration of its supply capacity. Demand continued to surge even during supply discussions, leading to further price increases following the January round.\n\nMemory supply contracts for DRAM and other products were traditionally done on an annual basis. Due to recent supply shortages, memory manufacturers have been pushing for quarterly contracts — but with prices now fluctuating so rapidly, even monthly pricing is becoming a consideration.\n\nSamsung Electronics is not the only one continuously raising memory prices. The industry estimates that SK Hynix and Micron have also closed Q1 supply contracts at similar levels.\n\nThe global expansion of AI infrastructure investment is what has driven this extraordinary situation. As more AI service providers (data centers) purchase AI semiconductor chips, demand for high-bandwidth memory (HBM) to support computation has surged. With Samsung Electronics, SK Hynix, and Micron concentrating their production capacity on HBM, supply of commodity DRAM for servers, PCs, and mobile devices has become constrained. Demand is overflowing — not only for AI servers but also AI PCs and AI smartphones — yet supply remains tight, pushing prices through the roof.\n\nWith AI growth still ongoing, memory price increases are expected to persist for some time. NVIDIA, the world's largest AI chip company, posted record-breaking earnings on the 25th of last month, dispelling concerns over an AI bubble. Market research firm Gartner forecasts that DRAM and solid-state drive (SSD) prices will rise 130% year-over-year (on a combined basis) this year.\n\nAn industry source said, \"DRAM and NAND prices will continue to rise in Q2 as well,\" adding that \"while the pace of increase may moderate, price hikes themselves are unavoidable at this point.\"\n\n---\n\nhttps://t.co/GoF4iSVQFR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05262363223776245, "ret_signed_m1d": -0.05262363223776245, "alpha_spy_m1d": -0.045617752526326605, "alpha_c_m1d": -0.03247828414965315, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.020145348088109305, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00928207356548938, "ret_signed_p1d": -0.00928207356548938, "alpha_spy_p1d": -0.0037065267190955486, "alpha_c_p1d": 0.00011406760877896716, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.009396141174268346, "ret_p1w": 0.044714005940161394, "ret_signed_p1w": 0.044714005940161394, "alpha_spy_p1w": 0.05755823784953695, "alpha_c_p1w": 0.039878143634309415, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004835862305851979, "ret_p1m": -0.08577521178622616, "ret_signed_p1m": -0.08577521178622616, "alpha_spy_p1m": -0.0456238780669328, "alpha_c_p1m": -0.0687869916017887, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.016988220184437464, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4083, -0.3841, -0.3704, -0.3423, -0.3554, -0.3986, -0.4076, -0.4201, -0.4375, -0.3801, -0.3367, -0.3101, -0.3109, -0.285, -0.285, -0.2897, -0.2655, -0.2698, -0.2878, -0.2878, -0.213, -0.2211, -0.1431, -0.1528, -0.184, -0.1389, -0.137, -0.1563, -0.1682, -0.16, -0.0949, -0.0949, -0.0893, -0.0291, -0.008, -0.0028, -0.0291, 0.0236, 0.0861, 0.0874, 0.0352, 0.0924, 0.0466, -0.0533, -0.0446, -0.0152, -0.0431, -0.0687, 0.0239, 0.0329, 0.0272, 0.0272, -0.0025, 0.0504, 0.0414, 0.0684, 0.0504, 0.043, 0.0704, 0.0369, 0.0289, 0.0297, -0.0526, 0.0, -0.0093, -0.076, -0.0286, 0.0058, 0.0447, 0.0114, 0.0633, 0.1024, 0.152, 0.1521, 0.1085, 0.0552, 0.0089, -0.0131, -0.0466, -0.1131, -0.1087, -0.1967, -0.1567, -0.0818, -0.0858, -0.0858, -0.057, -0.0575, 0.0153, 0.0522, 0.0499, 0.0648, 0.1624, 0.1389, 0.1414, 0.136, 0.1194, 0.1218, 0.2169, 0.2025, 0.2399, 0.3094, 0.2588, 0.2942, 0.291, 0.3535, 0.439, 0.5981, 0.664, 0.6141, 0.8642, 0.9853, 0.9136, 1.0061, 0.9371, 0.8089, 0.7013, 0.7442, 0.8272, 0.9024, 0.8747, 0.8747, 1.2363, 1.3175, 1.3053, 1.4239, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2026285577264615460", "ticker_idx": 0, "ts": "2026-02-24T13:18:22+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi forecasts Samsung's 2026 operating profit at $173.9B (KRW 251 trillion) and set a target price of KRW 280,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's Buy-equivalent call on Samsung with KRW 280K PT and sharply raised DRAM/NAND ASP forecasts.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Citi forecasts Samsung’s 2026 operating profit at $173.9B (KRW 251 trillion) and set a target price of KRW 280,000.\n\nCiti also expects 2026 global DRAM/NAND ASP growth (forecast) to reach +171% / +127% YoY (vs. its previous estimates of +120% / +90%).\n\nThe outlook is even stronger for the server segment. Citi forecasts 2026 server DRAM and SSD ASP growth at +290% / +153% YoY.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.035000028574647235, "ret_signed_m1d": -0.035000028574647235, "alpha_spy_m1d": -0.02778402482783149, "alpha_c_m1d": -0.02217224047039612, "bench_spy_m1d": -0.007216003746815747, "bench_c_m1d": -0.012827788104251114, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01749997514397128, "ret_signed_p1d": 0.01749997514397128, "alpha_spy_p1d": 0.009061626200389972, "alpha_c_p1d": -0.001670391922959924, "bench_spy_p1d": 0.008438348943581309, "bench_c_p1d": 0.019170367066931204, "ret_p1w": -0.024499980858900705, "ret_signed_p1w": -0.024499980858900705, "alpha_spy_p1w": -0.01428696975646604, "alpha_c_p1w": -0.0044030620191743974, "bench_spy_p1w": -0.010213011102434666, "bench_c_p1w": -0.020096918839726308, "ret_p1m": -0.055000011351572176, "ret_signed_p1m": -0.055000011351572176, "alpha_spy_p1m": -0.013192831676266503, "alpha_c_p1m": -0.030841180997431272, "bench_spy_p1m": -0.04180717967530567, "bench_c_p1m": -0.024158830354140903, "ret_p3m": 0.465527114952518, "ret_signed_p3m": 0.465527114952518, "alpha_spy_p3m": 0.37776045039313844, "alpha_c_p3m": 0.17836716271947384, "bench_spy_p3m": 0.08776666455937954, "bench_c_p3m": 0.28715995223304414, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.485, -0.4999, -0.4984, -0.4855, -0.48, -0.477, -0.4606, -0.4551, -0.4606, -0.4626, -0.4661, -0.4581, -0.4785, -0.4885, -0.4631, -0.4646, -0.4711, -0.4502, -0.4452, -0.4472, -0.4472, -0.4178, -0.4025, -0.4005, -0.4005, -0.4005, -0.3575, -0.3095, -0.3055, -0.295, -0.306, -0.305, -0.306, -0.312, -0.2985, -0.2805, -0.2555, -0.2535, -0.274, -0.2525, -0.2385, -0.2395, -0.2395, -0.2025, -0.188, -0.1965, -0.1975, -0.248, -0.1625, -0.1545, -0.2035, -0.207, -0.168, -0.171, -0.161, -0.107, -0.094, -0.094, -0.094, -0.094, -0.05, -0.0495, -0.035, 0.0, 0.0175, 0.09, 0.0825, 0.0825, -0.0245, -0.139, -0.042, -0.059, -0.1325, -0.0605, -0.05, -0.0605, -0.0825, -0.0565, -0.0305, 0.0425, 0.0025, -0.003, -0.0685, -0.0515, -0.055, -0.0995, -0.0995, -0.1167, -0.1623, -0.0498, -0.1062, -0.0671, -0.0325, -0.0155, 0.0547, 0.0221, 0.0321, 0.0071, 0.0346, 0.0572, 0.0898, 0.0822, 0.0747, 0.0973, 0.0898, 0.1248, 0.0998, 0.1248, 0.1123, 0.1323, 0.1048, 0.1048, 0.1649, 0.1649, 0.3328, 0.3603, 0.3453, 0.4305, 0.3979, 0.4229, 0.4831, 0.3553, 0.4079, 0.3804, 0.3829, 0.5006, 0.4655, 0.4655, 0.4981, 0.5382, 0.5006, 0.5883, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2015986278647689720", "ticker_idx": 0, "ts": "2026-01-27T03:12:38+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If timing the cycle feels daunting, buy semicap instead. Among semicap names, $LRCX has the highest NAND exposure.", "tweet_type": "analysis", "timeline": "years", "summary": "Buy LRCX as the highest-NAND-exposure semicap play; supercycle expected to benefit the industry through 2026-2027.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Ultimately, memory makers have no choice but to expand capacity, and that will keep the memory industry cyclical.\n\nThat said, it doesn’t mean memory stocks will fall right away.\n\nMicron’s Singapore NAND fab won’t come online until 2028, so it doesn’t undermine the view that the supercycle should continue to benefit the industry throughout 2027—and at least all of 2026.\n\nIf timing the cycle feels daunting, buy semicap instead. Among semicap names, $LRCX has the highest NAND exposure.", "reply_text": null, "parent_text": null, "quoted_text": "Micron is the first to announce a NAND capacity increase, excluding Kioxia and YMTC.\n\nhttps://t.co/QGhxhNdqU2", "parent_author": null, "ret_m1d": -0.06537792842007828, "ret_signed_m1d": -0.06537792842007828, "alpha_spy_m1d": -0.061409475059151064, "alpha_c_m1d": -0.04467813056671732, "bench_spy_m1d": -0.0039684533609272155, "bench_c_m1d": -0.020699797853360957, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004696773287537592, "ret_signed_p1d": 0.004696773287537592, "alpha_spy_p1d": 0.004797443191187467, "alpha_c_p1d": -0.0183357739999086, "bench_spy_p1d": -0.00010066990364987483, "bench_c_p1d": 0.023032547287446192, "ret_p1w": -0.03505833232002997, "ret_signed_p1w": -0.03505833232002997, "alpha_spy_p1w": -0.026488894769997495, "alpha_c_p1w": -0.011559235145766622, "bench_spy_p1w": -0.008569437550032477, "bench_c_p1w": -0.02349909717426335, "ret_p1m": 0.04621310481238594, "ret_signed_p1m": 0.04621310481238594, "alpha_spy_p1m": 0.049577538751413575, "alpha_c_p1m": -0.00022029890058972157, "bench_spy_p1m": -0.0033644339390276334, "bench_c_p1m": 0.04643340371297566, "ret_p3m": 0.1243009968470783, "ret_signed_p3m": 0.1243009968470783, "alpha_spy_p3m": 0.0949694476472589, "alpha_c_p3m": -0.1192594756916725, "bench_spy_p3m": 0.029331549199819396, "bench_c_p3m": 0.2435604725387508, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3259, -0.3408, -0.3249, -0.3478, -0.309, -0.321, -0.3329, -0.3035, -0.3336, -0.3242, -0.3581, -0.3793, -0.3826, -0.4003, -0.377, -0.4156, -0.4028, -0.3704, -0.3639, -0.3505, -0.3505, -0.3469, -0.3519, -0.3377, -0.3301, -0.3412, -0.3345, -0.3175, -0.3047, -0.2944, -0.2925, -0.3268, -0.311, -0.3154, -0.3501, -0.3093, -0.2776, -0.265, -0.2655, -0.2564, -0.2564, -0.2533, -0.2625, -0.2712, -0.2821, -0.2821, -0.2239, -0.1833, -0.1321, -0.1484, -0.1573, -0.0843, -0.0757, -0.101, -0.1244, -0.088, -0.065, -0.065, -0.0673, -0.0422, -0.0745, -0.0861, -0.0654, 0.0, 0.0047, 0.0407, -0.021, -0.004, -0.0351, -0.1203, -0.1055, -0.0312, -0.0385, -0.0497, -0.014, -0.0301, -0.0123, -0.0123, -0.0121, 0.0068, -0.0045, 0.0271, 0.016, 0.0243, 0.0462, 0.0026, -0.0192, -0.0313, -0.0889, -0.0638, -0.0986, -0.1631, -0.1135, -0.0963, -0.0811, -0.1204, -0.1091, -0.0788, -0.0491, -0.0565, -0.0176, -0.0412, -0.0204, 0.0028, -0.0198, -0.1115, -0.1124, -0.1606, -0.1029, -0.0679, -0.0829, -0.0829, -0.0736, -0.058, 0.0349, 0.0864, 0.107, 0.1224, 0.1437, 0.1133, 0.0957, 0.1235, 0.1049, 0.0848, 0.1149, 0.0856, 0.1243, 0.0894, 0.0548, 0.0444, 0.0827, 0.0779, 0.0856, 0.158, 0.2477, 0.203, 0.2346, 0.243, 0.2144, 0.2404, 0.256, 0.1954, 0.167, 0.1478, 0.2264, 0.269, 0.282, 0.282, 0.3548, 0.3391, 0.3352, 0.3359, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2006665030281081308", "ticker_idx": 0, "ts": "2026-01-01T09:53:19+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics also sees advanced packaging as the next key battleground following HBM and is strengthening its related business... observations suggest that an opportunity will soon come to Samsung", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish Samsung as a beneficiary of TSMC's CoWoS bottleneck; turnkey advanced packaging could win Broadcom/AMD customers.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "A Major Opportunity Approaches Samsung... Going All-In on Advanced Packaging\n\nEvery December, senior executives from AI semiconductor giants like NVIDIA and Broadcom gather at TSMC’s headquarters in the Hsinchu Science Park, Taiwan. Their goal is to secure even just one more slot on TSMC’s \"Advanced Packaging\" production lines, which integrate GPUs and High Bandwidth Memory (HBM) to create AI accelerators. The allocation of these lines, known as \"CoWoS\" (Chip-on-Wafer-on-Substrate), determines the volume of AI accelerators they can produce the following year. In fact, Google reportedly lowered the production target for its proprietary AI accelerator, the Tensor Processing Unit (TPU), by 1 million units from its original goal of 4 million because it lost out to NVIDIA in the battle for CoWoS capacity.\n\n\"Advanced Packaging,\" which mounts high-performance semiconductors like GPUs and HBM onto a special material called a \"silicon interposer\" to make them function seamlessly as a single chip, has emerged as the decisive battleground for supremacy in the AI semiconductor market this year. As ultra-fine processes—narrowing circuit linewidths to the 1-nanometer (nm) range to pack multiple functions into a small chip—reach their technical limits, semiconductor companies are turning to advanced packaging to connect multiple chips and operate them as a single unit.\n\nConsequently, the related market is projected to grow from $43 billion (approx. 62 trillion KRW) last year to $64.3 billion (approx. 93 trillion KRW) by 2028. Currently, the advanced packaging market is dominated by TSMC. The situation has reached a point where companies cannot produce AI accelerators on time even if they have secured GPUs and HBM, simply because they haven't been allocated enough of TSMC's CoWoS lines. In response to repeated demands from NVIDIA, AMD, Broadcom, and others, TSMC has decided to invest $7.5 billion (approx. 10.85 trillion KRW) this year—its largest amount ever—to significantly expand its advanced packaging capacity.\n\nSamsung Electronics also sees advanced packaging as the next key battleground following HBM and is strengthening its related business. It is reported to be marketing a \"turnkey\" solution—offering advanced packaging bundled with DRAM and foundry services—to companies like Google, AMD, and Amazon, who were unable to secure sufficient CoWoS lines due to NVIDIA’s dominance.\n\nTo produce the AI accelerators—often called the \"pickaxes\" of the gold rush in the AI era—chips must undergo two stages of advanced packaging (a process that makes multiple chips work like one). The first is the fabrication of HBM, which involves stacking up to 16 DRAMs. Once this process, which is far more difficult than standard DRAM production, is cleared, an even greater challenge awaits: \"2.5D Packaging,\" which connects the HBM and GPU on a special substrate called a silicon interposer to function as a single chip.\n\nNo matter how good the GPU and HBM are, if the advanced packaging fails, the AI accelerator cannot perform properly. It is a high-difficulty process, with industry leader TSMC's yield remaining at only 50–60%. As TSMC fails to keep up with the flood of orders from NVIDIA and AMD, 2.5D packaging has become the biggest bottleneck hindering the expansion of the AI market.\n\n◇ Alternatives to the Limits of Ultra-Fine Processes\n\nPackaging is broadly divided into traditional packaging and advanced packaging. Traditional packaging refers to the process of placing a single chip on a mainboard and connecting it electrically. It is part of the \"back-end\" process, which was previously evaluated as a \"less critical\" technology in the semiconductor ecosystem.\n\nHowever, the situation changed as demand for high-performance chips exploded in the AI era. Until the early 2020s, semiconductor companies went \"all-in\" on \"ultra-fine processes,\" narrowing circuit widths to below 2nm to pack more functions into smaller chips. But this competition backfired. Profitability declined due to the need to purchase Extreme Ultraviolet (EUV) lithography equipment costing up to 500 billion KRW per unit. Technical challenges were also formidable; as linewidths narrowed, interference increased, and leakage current grew, making it difficult to control heat generation.\n\nThe solution found was packaging. Instead of cramming complex functions into a single chip via ultra-fine processes, connecting several moderately advanced chips could achieve the same performance. The industry attached the modifier \"Advanced\" to this technology because its technical difficulty far exceeds that of traditional packaging.\n\n◇ Severe Shortage of CoWoS Capacity\n\nThe frontrunner is TSMC. Its main weapon is a 2.5D advanced packaging technology called \"CoWoS-S.\" It places a silicon interposer—a special material layer acting as a bridge—on top of a substrate and arranges multiple chips horizontally. The silicon interposer consists of Through-Silicon Vias (TSV), which are vertical passages connecting the bottom substrate to the top chips, and a Redistribution Layer (RDL), which connects signals between chips. It is called 2.5D packaging because the interposer and chips are stacked vertically (3D) on the substrate, while the chips are arranged horizontally (2D).\n\nAs demand for high-performance chips grew with the AI era, NVIDIA and AMD recognized the power of CoWoS. Thus, AI accelerators like the B200 and H100, connecting HBM and GPUs, were born.\n\nAccording to Samsung Securities, TSMC’s CoWoS production capacity (converted to wafers) increased from 35,000 sheets per month in 2024 to about 70,000 sheets last year, and is expected to rise to about 110,000 sheets this year. However, evaluations suggest this is still insufficient. Considering that the CoWoS allocation for NVIDIA by TSMC is around 55%, the calculation suggests that only 8.91 million \"Blackwell\" AI accelerators can be produced this year. This volume can support data centers with a maximum capacity of 18 gigawatts (GW), which accounts for only 50% of global data center investment capacity this year. Samsung Securities analyzed, \"There is a possibility that TSMC will not be able to meet even NVIDIA's demand this year.\"\n\n◇ Samsung Electronics Aiming for a Niche\n\nSamsung Electronics is looking for an opportunity. Although the domestic company Rebellions verified Samsung's technical capabilities to some extent by developing its latest AI accelerator, \"Rebel Quad,\" using Samsung's 2.5D packaging service \"I-Cube,\" Samsung has not yet secured significant global customers.\n\nHowever, as the \"TSMC bottleneck\" shows no signs of resolving, observations suggest that an opportunity will soon come to Samsung. Samsung is reportedly proposing a \"turnkey service\"—providing AI accelerator-related memory, foundry, and advanced packaging all at once—to companies like Broadcom (maker of Google's TPU) and AMD, who have not been allocated sufficient CoWoS volume due to NVIDIA.\n\nSK Hynix is also paying attention to 2.5D packaging. It has begun full-scale R&D centered on its Future Technology Research Institute and Packaging & Test (P&T) organization. Although SK Hynix draws a line by stating it is \"to make better customer-customized HBM,\" the industry observes a high possibility that it will enter the 2.5D packaging business in earnest in the future.\n\n---\n\nhttps://t.co/gSoGDDimh3", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07172642094610038, "ret_signed_p1d": 0.07172642094610038, "alpha_spy_p1d": 0.06989338121264854, "alpha_c_p1d": 0.06943427920387535, "bench_spy_p1d": 0.0018330397334518356, "bench_c_p1d": 0.002292141742225029, "ret_p1w": 0.1576312885638731, "ret_signed_p1w": 0.1576312885638731, "alpha_spy_p1w": 0.1465009391913823, "alpha_c_p1w": 0.15575587092450527, "bench_spy_p1w": 0.0111303493724908, "bench_c_p1w": 0.0018754176393678268, "ret_p1m": 0.33861552022460906, "ret_signed_p1m": 0.33861552022460906, "alpha_spy_p1m": 0.32387778742838336, "alpha_c_p1m": 0.3392405533200584, "bench_spy_p1m": 0.0147377327962257, "bench_c_p1m": -0.0006250330954493677, "ret_p3m": 0.39738173745551886, "ret_signed_p3m": 0.39738173745551886, "alpha_spy_p3m": 0.4410875830429226, "alpha_c_p3m": 0.47312503073898116, "bench_spy_p3m": -0.04370584558740376, "bench_c_p3m": -0.0757432932834623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2551, -0.2551, -0.2551, -0.2551, -0.2165, -0.2256, -0.2397, -0.2115, -0.1891, -0.1874, -0.1858, -0.1908, -0.1816, -0.1991, -0.18, -0.1534, -0.1742, -0.1659, -0.136, -0.1078, -0.0779, -0.1293, -0.165, -0.1766, -0.1874, -0.165, -0.141, -0.1443, -0.1468, -0.1932, -0.165, -0.1883, -0.1991, -0.165, -0.2132, -0.1974, -0.1758, -0.1468, -0.141, -0.1659, -0.1634, -0.1418, -0.1327, -0.1277, -0.1003, -0.0912, -0.1003, -0.1036, -0.1094, -0.0961, -0.1302, -0.1468, -0.1044, -0.1069, -0.1177, -0.0829, -0.0746, -0.0779, -0.0779, -0.0289, -0.0033, 0.0, 0.0, 0.0, 0.0717, 0.1518, 0.1585, 0.176, 0.1576, 0.1593, 0.1576, 0.1476, 0.1701, 0.2002, 0.2419, 0.2452, 0.211, 0.2469, 0.2702, 0.2686, 0.2686, 0.3303, 0.3545, 0.3403, 0.3386, 0.2544, 0.397, 0.4103, 0.3286, 0.3228, 0.3878, 0.3828, 0.3995, 0.4896, 0.5113, 0.5113, 0.5113, 0.5113, 0.5847, 0.5855, 0.6097, 0.6681, 0.6972, 0.8182, 0.8057, 0.8057, 0.6272, 0.4362, 0.598, 0.5696, 0.447, 0.5671, 0.5847, 0.5671, 0.5304, 0.5738, 0.6172, 0.7389, 0.6722, 0.6631, 0.5538, 0.5822, 0.5763, 0.5021, 0.5021, 0.4734, 0.3974, 0.585, 0.491, 0.5562, 0.6138, 0.6423, 0.7593, 0.7049, 0.7217, 0.6799, 0.7258, 0.7634, 0.8178, 0.8052, 0.7927, 0.8303, 0.8178, 0.8763, 0.8345, 0.8763, 0.8554, 0.8888, 0.8428, 0.8428, 0.9431, 0.9431, 1.2231, 1.2691, 1.244, 1.3861, 1.3318, 1.3735, 1.4738, 1.2607, 1.3485, 1.3025, 1.3067, 1.5031, 1.4446, 1.4446, 1.4989, 1.5658, 1.5031, 1.6493, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2011343940595875991", "ticker_idx": 0, "ts": "2026-01-14T07:45:39+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bernstein: Sandisk (SNDK - Outperform, $300 TP)", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Endorses Bernstein Outperform/$300 TP on SNDK; NAND makers prioritizing profitability over supply adds, structural bullish setup for the sector.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "\"They need to see demand consistently outstrip supply with margins significantly higher than this before even considering capacity additions.\"\n\nMy take is that the NAND companies that “came back from the dead” have no choice but to prioritize extreme profitability right now.\n\nLet me share one anecdote.\n\nDuring a severe downturn in the NAND industry, one NAND manufacturer was on the verge of bankruptcy — and at that point, none of its customers wanted to buy its products.\n\nWhy? Because once the company went bankrupt, customers could procure NAND far more cheaply by buying up its inventory at distressed auction prices.\n\nAnd after the company actually went bankrupt, DRAM buyers swooped in like hyenas and feasted on that inventory.\n\nNAND makers remember that period vividly.\n\nThey also remember exactly how poorly they were treated by demand-side customers during the brutal 2022–2024 NAND downturn.\n\nNow, NAND manufacturers have no intention of compromising. They have no intention of adding supply.", "reply_text": null, "parent_text": null, "quoted_text": "Bernstein: Sandisk (SNDK - Outperform, $300 TP) \n\nSandisk sees the CES keynote comments on Vera Rubin KV cache from Jensen Huang of Nvidia as incremental new NAND demand that is not previously in the demand forecast. Prior to the forecast, Sandisk estimated that NAND was 8%", "parent_author": null, "ret_m1d": 0.005157164623374122, "ret_signed_m1d": 0.005157164623374122, "alpha_spy_m1d": 0.00021764838710147671, "alpha_c_m1d": -0.007144125497510201, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.012301290120884323, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05525900005335327, "ret_signed_p1d": 0.05525900005335327, "alpha_spy_p1d": 0.05253578565682382, "alpha_c_p1d": 0.050006702133419845, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.005252297919933424, "ret_p1w": 0.2926175490594014, "ret_signed_p1w": 0.2926175490594014, "alpha_spy_p1w": 0.29980217905933404, "alpha_c_p1w": 0.2987682469110019, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": -0.006150697851600495, "ret_p1m": 0.6252545885749476, "ret_signed_p1m": 0.6252545885749476, "alpha_spy_p1m": 0.6384215491527889, "alpha_c_p1m": 0.6631950661216538, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": -0.03794047754670615, "ret_p3m": 1.456099651881945, "ret_signed_p3m": 1.456099651881945, "alpha_spy_p3m": 1.4595561962312087, "alpha_c_p3m": 1.4485595963266416, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.007540055555303349, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6386, -0.6183, -0.615, -0.6211, -0.5692, -0.52, -0.5449, -0.5474, -0.473, -0.4951, -0.486, -0.4662, -0.4983, -0.4417, -0.4645, -0.3825, -0.3091, -0.2997, -0.27, -0.3719, -0.3446, -0.3144, -0.3684, -0.3658, -0.4947, -0.4836, -0.4148, -0.4314, -0.4455, -0.4455, -0.4243, -0.4581, -0.4705, -0.4988, -0.45, -0.4109, -0.4186, -0.4341, -0.3996, -0.377, -0.4683, -0.4795, -0.4603, -0.4667, -0.4341, -0.3873, -0.3784, -0.3685, -0.3551, -0.3551, -0.3552, -0.3702, -0.3806, -0.3879, -0.3879, -0.2903, -0.2933, -0.0985, -0.0883, -0.1374, -0.0268, 0.0038, 0.0052, 0.0, 0.0553, 0.0666, 0.0666, 0.1684, 0.2926, 0.2982, 0.2218, 0.214, 0.2414, 0.3605, 0.3906, 0.4859, 0.7154, 0.7934, 0.5073, 0.4858, 0.5419, 0.5043, 0.3967, 0.5454, 0.6253, 0.6156, 0.6156, 0.5229, 0.5482, 0.6015, 0.676, 0.7186, 0.6465, 0.6306, 0.681, 0.6383, 0.5963, 0.458, 0.5447, 0.4584, 0.3598, 0.5181, 0.5959, 0.6901, 0.5957, 0.706, 0.8144, 0.857, 0.9435, 0.9909, 0.83, 0.8114, 0.8114, 0.7479, 0.5553, 0.588, 0.4762, 0.6383, 0.7863, 0.8091, 0.8091, 0.8685, 0.8329, 1.0136, 1.1958, 1.1964, 1.4561, 1.4354, 1.2994, 1.3709, 1.3748, 1.3543, 1.3297, 1.5246, 1.4043, 1.5525, 1.7596, 1.5846, 1.7442, 1.8274, 2.0608, 2.2383, 2.6263, 2.6357, 2.4552, 3.0286, 2.9905, 2.7442, 2.7318, 2.5655, 2.6296, 2.4373, 2.5669, 2.5908, 2.9768, 2.8129, 2.8129, 3.0988, 3.0998, 3.2331, 3.3706, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2011343940595875991", "ticker_idx": 1, "ts": "2026-01-14T07:45:39+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "NAND manufacturers have no intention of compromising. They have no intention of adding supply.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Endorses Bernstein Outperform/$300 TP on SNDK; NAND makers prioritizing profitability over supply adds, structural bullish setup for the sector.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"They need to see demand consistently outstrip supply with margins significantly higher than this before even considering capacity additions.\"\n\nMy take is that the NAND companies that “came back from the dead” have no choice but to prioritize extreme profitability right now.\n\nLet me share one anecdote.\n\nDuring a severe downturn in the NAND industry, one NAND manufacturer was on the verge of bankruptcy — and at that point, none of its customers wanted to buy its products.\n\nWhy? Because once the company went bankrupt, customers could procure NAND far more cheaply by buying up its inventory at distressed auction prices.\n\nAnd after the company actually went bankrupt, DRAM buyers swooped in like hyenas and feasted on that inventory.\n\nNAND makers remember that period vividly.\n\nThey also remember exactly how poorly they were treated by demand-side customers during the brutal 2022–2024 NAND downturn.\n\nNow, NAND manufacturers have no intention of compromising. They have no intention of adding supply.", "reply_text": null, "parent_text": null, "quoted_text": "Bernstein: Sandisk (SNDK - Outperform, $300 TP) \n\nSandisk sees the CES keynote comments on Vera Rubin KV cache from Jensen Huang of Nvidia as incremental new NAND demand that is not previously in the demand forecast. Prior to the forecast, Sandisk estimated that NAND was 8%", "parent_author": null, "ret_m1d": 0.005071612111988788, "ret_signed_m1d": 0.005071612111988788, "alpha_spy_m1d": 0.00013209587571614244, "alpha_c_m1d": -0.00311685866386493, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02530942086608894, "ret_signed_p1d": 0.02530942086608894, "alpha_spy_p1d": 0.02258620646955949, "alpha_c_p1d": 0.004554953421982911, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": 0.1530474752125065, "ret_signed_p1w": 0.1530474752125065, "alpha_spy_p1w": 0.16023210521243914, "alpha_c_p1w": 0.11807905404903207, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": 0.3862714850107667, "ret_signed_p1m": 0.3862714850107667, "alpha_spy_p1m": 0.39943844558860797, "alpha_c_p1m": 0.34053959633065456, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.7861470116884337, "ret_signed_p3m": 0.7861470116884337, "alpha_spy_p3m": 0.7896035560376974, "alpha_c_p3m": 0.644547966059174, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4433, -0.422, -0.4283, -0.4255, -0.4131, -0.3629, -0.3404, -0.3536, -0.314, -0.2998, -0.2969, -0.2649, -0.3002, -0.2884, -0.2763, -0.2574, -0.2034, -0.2031, -0.1953, -0.2254, -0.2812, -0.2437, -0.2899, -0.2862, -0.311, -0.3465, -0.3204, -0.3226, -0.3377, -0.3213, -0.3159, -0.3256, -0.3146, -0.3266, -0.3223, -0.2975, -0.2719, -0.2804, -0.2677, -0.2712, -0.2917, -0.3169, -0.3331, -0.3164, -0.2931, -0.2791, -0.2442, -0.2425, -0.2233, -0.2204, -0.2009, -0.1938, -0.197, -0.2028, -0.2028, -0.146, -0.116, -0.0453, -0.0218, -0.0339, -0.0085, -0.0009, 0.0051, 0.0, 0.0253, 0.0691, 0.0788, 0.0889, 0.153, 0.1885, 0.1688, 0.1542, 0.2153, 0.2768, 0.2906, 0.3416, 0.3204, 0.4066, 0.3275, 0.275, 0.285, 0.2919, 0.2596, 0.3101, 0.3863, 0.4097, 0.4015, 0.3768, 0.3803, 0.4026, 0.4285, 0.4374, 0.4715, 0.4713, 0.5129, 0.4896, 0.486, 0.3553, 0.314, 0.3629, 0.3127, 0.2818, 0.3769, 0.4399, 0.4013, 0.4214, 0.5005, 0.5122, 0.6004, 0.5608, 0.5127, 0.4457, 0.454, 0.455, 0.3518, 0.3459, 0.2775, 0.2745, 0.4195, 0.3783, 0.4132, 0.4589, 0.4717, 0.6422, 0.6702, 0.7172, 0.7861, 0.8865, 0.83, 0.8813, 0.8211, 0.8208, 0.8721, 0.9626, 0.9543, 0.973, 1.0705, 1.0348, 1.079, 1.1038, 1.1482, 1.2649, 1.3807, 1.4879, 1.5674, 1.7628, 1.8828, 1.7973, 1.9285, 1.8635, 1.7079, 1.7697, 1.7471, 1.7954, 2.0379, 2.0198, 2.141, 2.2794, 2.3333, 2.3699, 2.5316, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021471213617545510", "ticker_idx": 0, "ts": "2026-02-11T06:27:49+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "selling stocks solely on concerns about supply increases in the second half of 2027 appears premature", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (shared/adopted): memory supply structurally tight through 2026+; selling memory stocks on 2H27 supply fears is premature; bullish MU, Samsung, Hynix, DRAM broadly.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Interesting parts from Morgan Stanley's Micron commentary:\n\n\"Memory producers currently have virtually no inventory. Customers lower on the supply priority list—such as Chinese Android OEMs and consumer PC makers—are struggling to secure any inventory at all, and even the highest-priority customers face a similar situation. These top-tier customers are absorbing every incremental unit of supply and paying premiums over contracted prices just to receive shipments even 30 days sooner.\n\nThis tight supply-demand environment is reflected in Samsung and SK Hynix posting only low-single-digit (LSD) quarter-over-quarter bit growth heading into Q1—a clear sign that near-term fab output growth alone cannot keep pace with demand.\n\nEven over the medium to long term, the pace of supply expansion remains constrained. Factoring in new fabs such as CXMT Shanghai, Hynix M15, and Samsung P4L, wafer starts are expected to grow only about 7% year-over-year by the end of CY26. Micron's Boise project and its partnership with PSMC are also unlikely to contribute meaningful capacity before 2027.\n\nDemand, on the other hand, is overwhelming. Summing up the expected quarterly revenue increases from key players through the end of 2026:\n\n- NVIDIA: +$30 billion in quarterly revenue\n- AMD Data Center: doubling to ~$10 billion per quarter\n- Broadcom Semiconductors: doubling to ~$25 billion per quarter\n- Marvell + Intel: ~$1 billion in incremental revenue\n\nAnnualized, the memory industry faces roughly $200 billion in incremental revenue it must support over the next 12 months—a figure that exceeds the entire logic semiconductor market in 2020. The surge in HBM demand, in particular, means DRAM suppliers will need to operate at significantly higher capital intensity.\n\nIn short, as long as AI growth maintains its current CAGR, closing the supply-demand gap will be no easy task. In this structurally robust growth environment, selling stocks solely on concerns about supply increases in the second half of 2027 appears premature.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09038847502394831, "ret_signed_m1d": -0.09038847502394831, "alpha_spy_m1d": -0.09061958723887886, "alpha_c_m1d": -0.06623111680233484, "bench_spy_m1d": 0.00023111221493055112, "bench_c_m1d": -0.024157358221613467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008846330552533033, "ret_signed_p1d": 0.008846330552533033, "alpha_spy_p1d": 0.024295204780434232, "alpha_c_p1d": 0.029749010173658275, "bench_spy_p1d": -0.0154488742279012, "bench_c_p1d": -0.020902679621125242, "ret_p1w": 0.025856644624870917, "ret_signed_p1w": 0.025856644624870917, "alpha_spy_p1w": 0.034050824170316574, "alpha_c_p1w": 0.031233015213836368, "bench_spy_p1w": -0.008194179545445657, "bench_c_p1w": -0.005376370588965451, "ret_p1m": -0.012160672600280864, "ret_signed_p1m": -0.012160672600280864, "alpha_spy_p1m": 0.025269334804508548, "alpha_c_p1m": 0.05209024107664584, "bench_spy_p1m": -0.03743000740478941, "bench_c_p1m": -0.0642509136769267, "ret_p3m": 0.9390362248077155, "ret_signed_p3m": 0.9390362248077155, "alpha_spy_p3m": 0.8677040986001034, "alpha_c_p3m": 0.5496008632794793, "bench_spy_p3m": 0.07133212620761209, "bench_c_p3m": 0.3894353615282362, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3987, -0.4106, -0.4434, -0.4497, -0.5095, -0.4948, -0.4545, -0.453, -0.4391, -0.4391, -0.4239, -0.4142, -0.4166, -0.4296, -0.4479, -0.4221, -0.3985, -0.3851, -0.3576, -0.3704, -0.4126, -0.4214, -0.4336, -0.4506, -0.3945, -0.3522, -0.3262, -0.327, -0.3016, -0.3016, -0.3062, -0.2826, -0.2869, -0.3045, -0.3045, -0.2313, -0.2393, -0.1631, -0.1725, -0.2031, -0.159, -0.1571, -0.176, -0.1876, -0.1796, -0.116, -0.116, -0.1105, -0.0517, -0.0311, -0.0261, -0.0518, -0.0002, 0.0608, 0.062, 0.0111, 0.0669, 0.0222, -0.0754, -0.0669, -0.0381, -0.0654, -0.0904, 0.0, 0.0088, 0.0032, 0.0032, -0.0257, 0.0259, 0.0171, 0.0435, 0.0259, 0.0187, 0.0455, 0.0127, 0.0049, 0.0057, -0.0747, -0.0233, -0.0324, -0.0976, -0.0512, -0.0176, 0.0203, -0.0122, 0.0385, 0.0767, 0.1251, 0.1252, 0.0827, 0.0306, -0.0146, -0.0361, -0.0688, -0.1337, -0.1295, -0.2154, -0.1763, -0.1032, -0.1071, -0.1071, -0.079, -0.0794, -0.0084, 0.0277, 0.0254, 0.04, 0.1353, 0.1123, 0.1147, 0.1095, 0.0933, 0.0956, 0.1885, 0.1744, 0.211, 0.2789, 0.2295, 0.264, 0.2609, 0.3219, 0.4054, 0.5608, 0.6252, 0.5765, 0.8207, 0.939, 0.8689, 0.9593, 0.8919, 0.7667, 0.6616, 0.7035, 0.7846, 0.858, 0.831, 0.831, 1.1842, 1.2635, 1.2516, 1.3673, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021471213617545510", "ticker_idx": 1, "ts": "2026-02-11T06:27:49+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and SK Hynix posting only low-single-digit (LSD) quarter-over-quarter bit growth heading into Q1", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (shared/adopted): memory supply structurally tight through 2026+; selling memory stocks on 2H27 supply fears is premature; bullish MU, Samsung, Hynix, DRAM broadly.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Interesting parts from Morgan Stanley's Micron commentary:\n\n\"Memory producers currently have virtually no inventory. Customers lower on the supply priority list—such as Chinese Android OEMs and consumer PC makers—are struggling to secure any inventory at all, and even the highest-priority customers face a similar situation. These top-tier customers are absorbing every incremental unit of supply and paying premiums over contracted prices just to receive shipments even 30 days sooner.\n\nThis tight supply-demand environment is reflected in Samsung and SK Hynix posting only low-single-digit (LSD) quarter-over-quarter bit growth heading into Q1—a clear sign that near-term fab output growth alone cannot keep pace with demand.\n\nEven over the medium to long term, the pace of supply expansion remains constrained. Factoring in new fabs such as CXMT Shanghai, Hynix M15, and Samsung P4L, wafer starts are expected to grow only about 7% year-over-year by the end of CY26. Micron's Boise project and its partnership with PSMC are also unlikely to contribute meaningful capacity before 2027.\n\nDemand, on the other hand, is overwhelming. Summing up the expected quarterly revenue increases from key players through the end of 2026:\n\n- NVIDIA: +$30 billion in quarterly revenue\n- AMD Data Center: doubling to ~$10 billion per quarter\n- Broadcom Semiconductors: doubling to ~$25 billion per quarter\n- Marvell + Intel: ~$1 billion in incremental revenue\n\nAnnualized, the memory industry faces roughly $200 billion in incremental revenue it must support over the next 12 months—a figure that exceeds the entire logic semiconductor market in 2020. The surge in HBM demand, in particular, means DRAM suppliers will need to operate at significantly higher capital intensity.\n\nIn short, as long as AI growth maintains its current CAGR, closing the supply-demand gap will be no easy task. In this structurally robust growth environment, selling stocks solely on concerns about supply increases in the second half of 2027 appears premature.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011918987166019468, "ret_signed_m1d": -0.011918987166019468, "alpha_spy_m1d": -0.012150099380950019, "alpha_c_m1d": -0.008981301557045684, "bench_spy_m1d": 0.00023111221493055112, "bench_c_m1d": -0.0029376856089737835, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06436230675361365, "ret_signed_p1d": 0.06436230675361365, "alpha_spy_p1d": 0.07981118098151485, "alpha_c_p1d": 0.09066149076070407, "bench_spy_p1d": -0.0154488742279012, "bench_c_p1d": -0.026299184007090415, "ret_p1w": 0.0798569154218085, "ret_signed_p1w": 0.0798569154218085, "alpha_spy_p1w": 0.08805109496725416, "alpha_c_p1w": 0.09426558423913756, "bench_spy_p1w": -0.008194179545445657, "bench_c_p1w": -0.014408668817329051, "ret_p1m": 0.11978546644225108, "ret_signed_p1m": 0.11978546644225108, "alpha_spy_p1m": 0.1572154738470405, "alpha_c_p1m": 0.1556672054418522, "bench_spy_p1m": -0.03743000740478941, "bench_c_p1m": -0.03588173899960112, "ret_p3m": 0.7049518812042679, "ret_signed_p3m": 0.7049518812042679, "alpha_spy_p3m": 0.6336197549966558, "alpha_c_p3m": 0.45922785353054496, "bench_spy_p3m": 0.07133212620761209, "bench_c_p3m": 0.24572402767372292, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4235, -0.4034, -0.42, -0.4277, -0.4034, -0.4378, -0.4265, -0.4111, -0.3903, -0.3862, -0.404, -0.4022, -0.3868, -0.3802, -0.3767, -0.3571, -0.3506, -0.3571, -0.3595, -0.3636, -0.3542, -0.3785, -0.3903, -0.3601, -0.3619, -0.3696, -0.3447, -0.3387, -0.3411, -0.3411, -0.3061, -0.2878, -0.2855, -0.2855, -0.2855, -0.2342, -0.177, -0.1722, -0.1597, -0.1728, -0.1716, -0.1728, -0.18, -0.1639, -0.1424, -0.1126, -0.1103, -0.1347, -0.1091, -0.0924, -0.0936, -0.0936, -0.0495, -0.0322, -0.0423, -0.0435, -0.1037, -0.0018, 0.0077, -0.0507, -0.0548, -0.0083, -0.0119, 0.0, 0.0644, 0.0799, 0.0799, 0.0799, 0.0799, 0.1323, 0.1329, 0.1502, 0.1919, 0.2128, 0.2992, 0.2902, 0.2902, 0.1627, 0.0262, 0.1418, 0.1216, 0.034, 0.1198, 0.1323, 0.1198, 0.0936, 0.1246, 0.1555, 0.2426, 0.1949, 0.1883, 0.1103, 0.1305, 0.1263, 0.0733, 0.0733, 0.0528, -0.0015, 0.1326, 0.0654, 0.112, 0.1532, 0.1735, 0.2571, 0.2182, 0.2302, 0.2003, 0.2332, 0.2601, 0.2989, 0.2899, 0.281, 0.3078, 0.2989, 0.3407, 0.3108, 0.3407, 0.3257, 0.3496, 0.3168, 0.3168, 0.3884, 0.3884, 0.5885, 0.6213, 0.6034, 0.705, 0.6661, 0.696, 0.7677, 0.6154, 0.6781, 0.6452, 0.6482, 0.7886, 0.7468, 0.7468, 0.7856, 0.8333, 0.7886, 0.8931, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021471213617545510", "ticker_idx": 2, "ts": "2026-02-11T06:27:49+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and SK Hynix posting only low-single-digit (LSD) quarter-over-quarter bit growth heading into Q1", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (shared/adopted): memory supply structurally tight through 2026+; selling memory stocks on 2H27 supply fears is premature; bullish MU, Samsung, Hynix, DRAM broadly.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Interesting parts from Morgan Stanley's Micron commentary:\n\n\"Memory producers currently have virtually no inventory. Customers lower on the supply priority list—such as Chinese Android OEMs and consumer PC makers—are struggling to secure any inventory at all, and even the highest-priority customers face a similar situation. These top-tier customers are absorbing every incremental unit of supply and paying premiums over contracted prices just to receive shipments even 30 days sooner.\n\nThis tight supply-demand environment is reflected in Samsung and SK Hynix posting only low-single-digit (LSD) quarter-over-quarter bit growth heading into Q1—a clear sign that near-term fab output growth alone cannot keep pace with demand.\n\nEven over the medium to long term, the pace of supply expansion remains constrained. Factoring in new fabs such as CXMT Shanghai, Hynix M15, and Samsung P4L, wafer starts are expected to grow only about 7% year-over-year by the end of CY26. Micron's Boise project and its partnership with PSMC are also unlikely to contribute meaningful capacity before 2027.\n\nDemand, on the other hand, is overwhelming. Summing up the expected quarterly revenue increases from key players through the end of 2026:\n\n- NVIDIA: +$30 billion in quarterly revenue\n- AMD Data Center: doubling to ~$10 billion per quarter\n- Broadcom Semiconductors: doubling to ~$25 billion per quarter\n- Marvell + Intel: ~$1 billion in incremental revenue\n\nAnnualized, the memory industry faces roughly $200 billion in incremental revenue it must support over the next 12 months—a figure that exceeds the entire logic semiconductor market in 2020. The surge in HBM demand, in particular, means DRAM suppliers will need to operate at significantly higher capital intensity.\n\nIn short, as long as AI growth maintains its current CAGR, closing the supply-demand gap will be no easy task. In this structurally robust growth environment, selling stocks solely on concerns about supply increases in the second half of 2027 appears premature.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.018604668097836008, "ret_signed_m1d": 0.018604668097836008, "alpha_spy_m1d": 0.018373555882905457, "alpha_c_m1d": 0.042762026319449475, "bench_spy_m1d": 0.00023111221493055112, "bench_c_m1d": -0.024157358221613467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0325581327608655, "ret_signed_p1d": 0.0325581327608655, "alpha_spy_p1d": 0.0480070069887667, "alpha_c_p1d": 0.05346081238199074, "bench_spy_p1d": -0.0154488742279012, "bench_c_p1d": -0.020902679621125242, "ret_p1w": 0.023255798711947495, "ret_signed_p1w": 0.023255798711947495, "alpha_spy_p1w": 0.03144997825739315, "alpha_c_p1w": 0.028632169300912946, "bench_spy_p1w": -0.008194179545445657, "bench_c_p1w": -0.005376370588965451, "ret_p1m": 0.08339076458816974, "ret_signed_p1m": 0.08339076458816974, "alpha_spy_p1m": 0.12082077199295915, "alpha_c_p1m": 0.14764167826509644, "bench_spy_p1m": -0.03743000740478941, "bench_c_p1m": -0.0642509136769267, "ret_p3m": 1.190080219277677, "ret_signed_p3m": 1.190080219277677, "alpha_spy_p3m": 1.1187480930700648, "alpha_c_p3m": 0.8006448577494407, "bench_spy_p3m": 0.07133212620761209, "bench_c_p3m": 0.3894353615282362, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3493, -0.2959, -0.3377, -0.347, -0.3365, -0.3946, -0.3958, -0.3969, -0.3911, -0.3674, -0.3837, -0.3744, -0.3512, -0.3581, -0.3698, -0.3674, -0.3291, -0.3419, -0.3174, -0.343, -0.336, -0.3558, -0.3837, -0.3593, -0.3581, -0.364, -0.3256, -0.3209, -0.3163, -0.3163, -0.3035, -0.2558, -0.243, -0.243, -0.243, -0.2128, -0.1907, -0.1558, -0.1372, -0.1209, -0.1349, -0.1291, -0.1419, -0.1372, -0.1291, -0.1209, -0.1116, -0.136, -0.1395, -0.1221, -0.1081, -0.1442, -0.0698, -0.0221, 0.0012, 0.057, -0.0349, 0.0547, 0.0465, -0.0209, -0.0244, 0.0314, 0.0186, 0.0, 0.0326, 0.0233, 0.0233, 0.0233, 0.0233, 0.0395, 0.1035, 0.1058, 0.1686, 0.1837, 0.2803, 0.236, 0.236, 0.0939, -0.011, 0.0962, 0.0764, -0.0261, 0.0927, 0.1125, 0.0834, 0.0601, 0.1346, 0.13, 0.2302, 0.1801, 0.1731, 0.0869, 0.1486, 0.1591, 0.0869, 0.0869, 0.017, -0.0599, 0.0449, -0.0331, 0.0205, 0.0321, 0.0671, 0.2034, 0.1626, 0.1964, 0.2115, 0.2849, 0.3234, 0.3455, 0.314, 0.3583, 0.4259, 0.4247, 0.427, 0.4236, 0.5051, 0.5144, 0.5063, 0.4981, 0.4981, 0.6857, 0.6857, 0.8651, 0.9268, 0.9641, 1.1901, 1.1377, 1.3019, 1.2949, 1.119, 1.1435, 1.0328, 1.0328, 1.26, 1.2611, 1.2611, 1.3904, 1.613, 1.667, 1.7183, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021471213617545510", "ticker_idx": 3, "ts": "2026-02-11T06:27:49+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "as long as AI growth maintains its current CAGR, closing the supply-demand gap will be no easy task", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley (shared/adopted): memory supply structurally tight through 2026+; selling memory stocks on 2H27 supply fears is premature; bullish MU, Samsung, Hynix, DRAM broadly.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Interesting parts from Morgan Stanley's Micron commentary:\n\n\"Memory producers currently have virtually no inventory. Customers lower on the supply priority list—such as Chinese Android OEMs and consumer PC makers—are struggling to secure any inventory at all, and even the highest-priority customers face a similar situation. These top-tier customers are absorbing every incremental unit of supply and paying premiums over contracted prices just to receive shipments even 30 days sooner.\n\nThis tight supply-demand environment is reflected in Samsung and SK Hynix posting only low-single-digit (LSD) quarter-over-quarter bit growth heading into Q1—a clear sign that near-term fab output growth alone cannot keep pace with demand.\n\nEven over the medium to long term, the pace of supply expansion remains constrained. Factoring in new fabs such as CXMT Shanghai, Hynix M15, and Samsung P4L, wafer starts are expected to grow only about 7% year-over-year by the end of CY26. Micron's Boise project and its partnership with PSMC are also unlikely to contribute meaningful capacity before 2027.\n\nDemand, on the other hand, is overwhelming. Summing up the expected quarterly revenue increases from key players through the end of 2026:\n\n- NVIDIA: +$30 billion in quarterly revenue\n- AMD Data Center: doubling to ~$10 billion per quarter\n- Broadcom Semiconductors: doubling to ~$25 billion per quarter\n- Marvell + Intel: ~$1 billion in incremental revenue\n\nAnnualized, the memory industry faces roughly $200 billion in incremental revenue it must support over the next 12 months—a figure that exceeds the entire logic semiconductor market in 2020. The surge in HBM demand, in particular, means DRAM suppliers will need to operate at significantly higher capital intensity.\n\nIn short, as long as AI growth maintains its current CAGR, closing the supply-demand gap will be no easy task. In this structurally robust growth environment, selling stocks solely on concerns about supply increases in the second half of 2027 appears premature.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027900931364043924, "ret_signed_m1d": -0.027900931364043924, "alpha_spy_m1d": -0.028132043578974475, "alpha_c_m1d": -0.0037435731424304564, "bench_spy_m1d": 0.00023111221493055112, "bench_c_m1d": -0.024157358221613467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03525559002233739, "ret_signed_p1d": 0.03525559002233739, "alpha_spy_p1d": 0.05070446425023859, "alpha_c_p1d": 0.056158269643462634, "bench_spy_p1d": -0.0154488742279012, "bench_c_p1d": -0.020902679621125242, "ret_p1w": 0.04298978625287564, "ret_signed_p1w": 0.04298978625287564, "alpha_spy_p1w": 0.051183965798321296, "alpha_c_p1w": 0.04836615684184109, "bench_spy_p1w": -0.008194179545445657, "bench_c_p1w": -0.005376370588965451, "ret_p1m": 0.06367185281004666, "ret_signed_p1m": 0.06367185281004666, "alpha_spy_p1m": 0.10110186021483607, "alpha_c_p1m": 0.12792276648697337, "bench_spy_p1m": -0.03743000740478941, "bench_c_p1m": -0.0642509136769267, "ret_p3m": 0.9446894417632201, "ret_signed_p3m": 0.9446894417632201, "alpha_spy_p3m": 0.873357315555608, "alpha_c_p3m": 0.5552540802349839, "bench_spy_p3m": 0.07133212620761209, "bench_c_p3m": 0.3894353615282362, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3905, -0.3699, -0.4003, -0.4081, -0.4165, -0.4424, -0.4256, -0.4204, -0.4068, -0.3976, -0.4039, -0.3969, -0.3848, -0.3893, -0.3981, -0.3822, -0.3594, -0.3614, -0.3448, -0.359, -0.3676, -0.3852, -0.4026, -0.39, -0.3715, -0.3619, -0.3322, -0.3289, -0.3197, -0.3197, -0.3053, -0.2754, -0.2718, -0.2776, -0.2776, -0.2261, -0.2023, -0.1637, -0.1565, -0.1656, -0.1552, -0.153, -0.1659, -0.1629, -0.1504, -0.1165, -0.1126, -0.1271, -0.1001, -0.0819, -0.0759, -0.0965, -0.0398, 0.0022, 0.007, 0.0082, -0.0239, 0.025, -0.007, -0.0462, -0.0391, -0.0141, -0.0279, 0.0, 0.0353, 0.0354, 0.0354, 0.0258, 0.043, 0.063, 0.0933, 0.094, 0.1264, 0.1473, 0.1974, 0.1771, 0.1773, 0.0606, -0.0027, 0.0686, 0.0335, -0.0145, 0.065, 0.0884, 0.0637, 0.064, 0.112, 0.1369, 0.1993, 0.1525, 0.1307, 0.0608, 0.081, 0.0722, 0.0088, 0.0102, -0.0485, -0.0792, 0.0248, -0.0249, 0.0084, 0.0354, 0.0537, 0.1507, 0.1362, 0.1507, 0.1506, 0.2178, 0.2319, 0.253, 0.2378, 0.2442, 0.2764, 0.304, 0.3141, 0.3151, 0.3749, 0.3565, 0.3733, 0.3586, 0.3789, 0.4932, 0.545, 0.6929, 0.7082, 0.7961, 0.9447, 0.8909, 0.9857, 0.9848, 0.8337, 0.8277, 0.7939, 0.8219, 0.9689, 0.9463, 0.9463, 1.1201, 1.2366, 1.2357, 1.3262, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2016024794312147358", "ticker_idx": 0, "ts": "2026-01-27T05:45:41+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix are now positioned to maximize the profitability of their DRAM businesses starting in the first quarter of this year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung and SK Hynix hiking LPDDR prices ~80-100% QoQ for Apple; profitability set to surge in Q1 with margins expected in the 60%+ range.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Even 'Whale' Apple Surrenders... Samsung and SK Hynix Hike LPDDR Prices for iPhone\n\nIt has been reported that Samsung Electronics and SK Hynix have recently increased the price of Low Power Double Data Rate (LPDDR) DRAM supplied to Apple by nearly double compared to the previous quarter. Apple, a major smartphone manufacturer, has traditionally utilized its dominant market position to procure LPDDR at relatively low prices.\n\nHowever, as the memory supply shortage has intensified since mid-last year, Apple appears to have been unable to resist the overall trend of rising prices. Conversely, Samsung Electronics and SK Hynix are now positioned to maximize the profitability of their DRAM businesses starting in the first quarter of this year.\n\nAccording to industry sources on the 27th, Samsung and SK Hynix have agreed to significantly raise LPDDR prices for the iPhone through negotiations with Apple.\n\nLPDDR is a type of DRAM that focuses on power efficiency compared to standard DRAM (DDR). It is a critical component for mobile devices, and the 7th-generation product, LPDDR5X, is currently the most widely used. Apple is a major customer for this memory, with annual iPhone shipments estimated at approximately 250 million units as of last year.\n\nRecently, memory prices, including LPDDR5, have seen a sharp upward trajectory. This is because demand for DRAM has surged due to aggressive AI infrastructure investments by global big tech companies, while suppliers have shifted a larger portion of their production capacity to High Bandwidth Memory (HBM), exacerbating the supply shortage of conventional chips.\n\nThrough these negotiations, Samsung and SK Hynix have significantly raised the prices of LPDDR supplied for iPhones in the first quarter. Specifically, Samsung Electronics reportedly proposed an increase of over 80% compared to the previous quarter, while SK Hynix suggested an increase of around 100%.\n\nThere is also a possibility that LPDDR prices for Apple could rise further in the second half of the year, coinciding with the launch of Apple's latest flagship, the 'iPhone 18.'\n\nAn industry official noted, \"Apple typically signs annual Long-Term Agreements (LTA), but due to the recent memory crisis, they have reportedly finalized unit prices only through the first half of the year. Prices may see additional hikes in the second half to align with new product releases.\"\n\nDespite the hike, the price of LPDDR for Apple is not considered objectively high. As a 'whale' in the memory market, Apple has historically used its leverage to secure LPDDR at lower prices than its competitors. While exact unit prices remain confidential, the industry views these negotiations as having significantly addressed the long-standing price imbalance between Apple and memory suppliers.\n\nThe reflection of these price hikes is expected to further boost the profitability of memory suppliers in Q1. Market research firm TrendForce anticipates that general-purpose DRAM prices will rise by 55–60% this quarter compared to the previous one.\n\nAnother official remarked, \"LPDDR prices already rose by about 40% in the fourth quarter of last year. With an even larger increase in the first quarter, profit margins are expected to reach at least the 60% range.\"\n\n---\n\nhttps://t.co/p1KjlFsQwY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.046395007713326386, "ret_signed_m1d": -0.046395007713326386, "alpha_spy_m1d": -0.04242655435239917, "alpha_c_m1d": -0.03308875366557107, "bench_spy_m1d": -0.0039684533609272155, "bench_c_m1d": -0.013306254047755317, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01818185387823501, "ret_signed_p1d": 0.01818185387823501, "alpha_spy_p1d": 0.018282523781884885, "alpha_c_p1d": 0.010211619765765123, "bench_spy_p1d": -0.00010066990364987483, "bench_c_p1d": 0.007970234112469887, "ret_p1w": 0.05015669611457452, "ret_signed_p1w": 0.05015669611457452, "alpha_spy_p1w": 0.058726133664606994, "alpha_c_p1w": 0.0904809045500864, "bench_spy_p1w": -0.008569437550032477, "bench_c_p1w": -0.04032420843551188, "ret_p1m": 0.27586202496045176, "ret_signed_p1m": 0.27586202496045176, "alpha_spy_p1m": 0.2792264588994794, "alpha_c_p1m": 0.30990465368062015, "bench_spy_p1m": -0.0033644339390276334, "bench_c_p1m": -0.03404262872016839, "ret_p3m": 0.379023973401172, "ret_signed_p3m": 0.379023973401172, "alpha_spy_p3m": 0.3496924242013526, "alpha_c_p3m": 0.2954764452717231, "bench_spy_p3m": 0.029331549199819396, "bench_c_p3m": 0.08354752812944888, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3505, -0.3293, -0.3068, -0.3455, -0.3723, -0.3811, -0.3892, -0.3723, -0.3542, -0.3567, -0.3586, -0.3935, -0.3723, -0.3898, -0.3979, -0.3723, -0.4085, -0.3967, -0.3804, -0.3586, -0.3542, -0.373, -0.3711, -0.3549, -0.348, -0.3443, -0.3237, -0.3168, -0.3237, -0.3262, -0.3305, -0.3205, -0.3461, -0.3586, -0.3268, -0.3287, -0.3368, -0.3106, -0.3043, -0.3068, -0.3068, -0.27, -0.2508, -0.2483, -0.2483, -0.2483, -0.1944, -0.1342, -0.1292, -0.116, -0.1298, -0.1285, -0.1298, -0.1373, -0.1204, -0.0978, -0.0665, -0.0639, -0.0897, -0.0627, -0.0451, -0.0464, -0.0464, 0.0, 0.0182, 0.0075, 0.0063, -0.0571, 0.0502, 0.0602, -0.0013, -0.0056, 0.0433, 0.0395, 0.052, 0.1197, 0.1361, 0.1361, 0.1361, 0.1361, 0.1912, 0.1918, 0.21, 0.2539, 0.2759, 0.3668, 0.3574, 0.3574, 0.2232, 0.0796, 0.2013, 0.1799, 0.0878, 0.1781, 0.1912, 0.1781, 0.1505, 0.1831, 0.2157, 0.3072, 0.2571, 0.2502, 0.168, 0.1893, 0.185, 0.1292, 0.1292, 0.1076, 0.0504, 0.1915, 0.1208, 0.1698, 0.2132, 0.2345, 0.3225, 0.2816, 0.2942, 0.2628, 0.2974, 0.3256, 0.3665, 0.357, 0.3476, 0.3759, 0.3665, 0.4104, 0.379, 0.4104, 0.3947, 0.4199, 0.3853, 0.3853, 0.4607, 0.4607, 0.6712, 0.7057, 0.6869, 0.7937, 0.7528, 0.7842, 0.8596, 0.6994, 0.7654, 0.7308, 0.734, 0.8816, 0.8377, 0.8377, 0.8785, 0.9287, 0.8816, 0.9916, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2016024794312147358", "ticker_idx": 1, "ts": "2026-01-27T05:45:41+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and SK Hynix have significantly raised the prices of LPDDR supplied for iPhones in the first quarter", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung and SK Hynix hiking LPDDR prices ~80-100% QoQ for Apple; profitability set to surge in Q1 with margins expected in the 60%+ range.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Even 'Whale' Apple Surrenders... Samsung and SK Hynix Hike LPDDR Prices for iPhone\n\nIt has been reported that Samsung Electronics and SK Hynix have recently increased the price of Low Power Double Data Rate (LPDDR) DRAM supplied to Apple by nearly double compared to the previous quarter. Apple, a major smartphone manufacturer, has traditionally utilized its dominant market position to procure LPDDR at relatively low prices.\n\nHowever, as the memory supply shortage has intensified since mid-last year, Apple appears to have been unable to resist the overall trend of rising prices. Conversely, Samsung Electronics and SK Hynix are now positioned to maximize the profitability of their DRAM businesses starting in the first quarter of this year.\n\nAccording to industry sources on the 27th, Samsung and SK Hynix have agreed to significantly raise LPDDR prices for the iPhone through negotiations with Apple.\n\nLPDDR is a type of DRAM that focuses on power efficiency compared to standard DRAM (DDR). It is a critical component for mobile devices, and the 7th-generation product, LPDDR5X, is currently the most widely used. Apple is a major customer for this memory, with annual iPhone shipments estimated at approximately 250 million units as of last year.\n\nRecently, memory prices, including LPDDR5, have seen a sharp upward trajectory. This is because demand for DRAM has surged due to aggressive AI infrastructure investments by global big tech companies, while suppliers have shifted a larger portion of their production capacity to High Bandwidth Memory (HBM), exacerbating the supply shortage of conventional chips.\n\nThrough these negotiations, Samsung and SK Hynix have significantly raised the prices of LPDDR supplied for iPhones in the first quarter. Specifically, Samsung Electronics reportedly proposed an increase of over 80% compared to the previous quarter, while SK Hynix suggested an increase of around 100%.\n\nThere is also a possibility that LPDDR prices for Apple could rise further in the second half of the year, coinciding with the launch of Apple's latest flagship, the 'iPhone 18.'\n\nAn industry official noted, \"Apple typically signs annual Long-Term Agreements (LTA), but due to the recent memory crisis, they have reportedly finalized unit prices only through the first half of the year. Prices may see additional hikes in the second half to align with new product releases.\"\n\nDespite the hike, the price of LPDDR for Apple is not considered objectively high. As a 'whale' in the memory market, Apple has historically used its leverage to secure LPDDR at lower prices than its competitors. While exact unit prices remain confidential, the industry views these negotiations as having significantly addressed the long-standing price imbalance between Apple and memory suppliers.\n\nThe reflection of these price hikes is expected to further boost the profitability of memory suppliers in Q1. Market research firm TrendForce anticipates that general-purpose DRAM prices will rise by 55–60% this quarter compared to the previous one.\n\nAnother official remarked, \"LPDDR prices already rose by about 40% in the fourth quarter of last year. With an even larger increase in the first quarter, profit margins are expected to reach at least the 60% range.\"\n\n---\n\nhttps://t.co/p1KjlFsQwY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07999999686871007, "ret_signed_m1d": -0.07999999686871007, "alpha_spy_m1d": -0.07603154350778285, "alpha_c_m1d": -0.05930019901534911, "bench_spy_m1d": -0.0039684533609272155, "bench_c_m1d": -0.020699797853360957, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05124997964661504, "ret_signed_p1d": 0.05124997964661504, "alpha_spy_p1d": 0.051350649550264915, "alpha_c_p1d": 0.028217432359168848, "bench_spy_p1d": -0.00010066990364987483, "bench_c_p1d": 0.023032547287446192, "ret_p1w": 0.1337499984343551, "ret_signed_p1w": 0.1337499984343551, "alpha_spy_p1w": 0.14231943598438757, "alpha_c_p1w": 0.15724909560861844, "bench_spy_p1w": -0.008569437550032477, "bench_c_p1w": -0.02349909717426335, "ret_p1m": 0.2725000407067697, "ret_signed_p1m": 0.2725000407067697, "alpha_spy_p1m": 0.27586447464579733, "alpha_c_p1m": 0.22606663699379403, "bench_spy_p1m": -0.0033644339390276334, "bench_c_p1m": 0.04643340371297566, "ret_p3m": 0.5303187152191966, "ret_signed_p3m": 0.5303187152191966, "alpha_spy_p3m": 0.5009871660193772, "alpha_c_p3m": 0.2867582426804458, "bench_spy_p3m": 0.029331549199819396, "bench_c_p3m": 0.2435604725387508, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2905, -0.3018, -0.2256, -0.268, -0.2768, -0.2593, -0.2755, -0.243, -0.2268, -0.2293, -0.2355, -0.3005, -0.243, -0.288, -0.298, -0.2868, -0.3492, -0.3505, -0.3517, -0.3455, -0.32, -0.3375, -0.3275, -0.3025, -0.31, -0.3225, -0.32, -0.2788, -0.2925, -0.2662, -0.2937, -0.2863, -0.3075, -0.3375, -0.3112, -0.31, -0.3163, -0.275, -0.27, -0.265, -0.265, -0.2512, -0.2, -0.1862, -0.1862, -0.1862, -0.1538, -0.13, -0.0925, -0.0725, -0.055, -0.07, -0.0638, -0.0775, -0.0725, -0.0638, -0.055, -0.045, -0.0712, -0.075, -0.0562, -0.0412, -0.08, 0.0, 0.0512, 0.0763, 0.1363, 0.0375, 0.1337, 0.125, 0.0525, 0.0488, 0.1088, 0.095, 0.075, 0.11, 0.1, 0.1, 0.1, 0.1, 0.1175, 0.1862, 0.1888, 0.2563, 0.2725, 0.3763, 0.3287, 0.3287, 0.1759, 0.0632, 0.1784, 0.1571, 0.0469, 0.1747, 0.196, 0.1646, 0.1396, 0.2197, 0.2147, 0.3224, 0.2686, 0.2611, 0.1684, 0.2348, 0.246, 0.1684, 0.1684, 0.0933, 0.0106, 0.1233, 0.0394, 0.097, 0.1095, 0.1471, 0.2936, 0.2498, 0.2861, 0.3024, 0.3813, 0.4226, 0.4464, 0.4126, 0.4602, 0.5328, 0.5316, 0.5341, 0.5303, 0.618, 0.628, 0.6192, 0.6105, 0.6105, 0.8121, 0.8121, 1.0049, 1.0713, 1.1114, 1.3543, 1.298, 1.4746, 1.467, 1.2779, 1.3042, 1.1853, 1.1853, 1.4295, 1.4307, 1.4307, 1.5697, 1.8089, 1.867, 1.9221, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034042994199019582", "ticker_idx": 0, "ts": "2026-03-17T23:03:35+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": "advanced packaging, SoIC", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's SoIC (System on Integrated Chips) as a critical enabling technology", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC SoIC adoption for NVDA Rubin Ultra/Feynman identified as a structural tailwind for TSMC, Besi, AMAT, and TEL on hybrid bonding/advanced packaging demand.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA’s Next-Generation Chips: TSMC Prepares with SoIC (Taiwan Commercial Times)\n\n• NVIDIA unveiled its AI chip roadmap at GTC 2026, presenting an enhanced Rubin Ultra platform for 2027 and the next-generation Feynman architecture for 2028. The Rubin Ultra platform is expected to feature seven chips and five rack configurations, while Feynman is set to adopt custom HBM, high-density chiplets, and advanced 3D stacking technologies.\n\n• Industry sources view TSMC’s SoIC (System on Integrated Chips) as a critical enabling technology. SoIC vertically stacks dies to significantly increase transistor density, helping extend scaling beyond the limits of traditional Moore’s Law. As the Rubin Ultra generation drives larger chip sizes and greater HBM integration, packaging complexity is expected to rise sharply. With manufacturing efficiency for single-package solutions becoming more constrained, the importance of 3D stacking is becoming even more pronounced.\n\n• In particular, the Feynman generation is expected to mark a major inflection point for SoIC demand. Based on copper-to-copper hybrid bonding, SoIC enables shorter interconnects and lower power consumption, making it a key technology for addressing AI compute bottlenecks. Compared with conventional 2.5D packaging such as CoWoS, it offers higher bandwidth and better energy efficiency, making it well suited for next-generation AI training and inference chips.\n•On the manufacturing side, TSMC is expected to build SoIC capacity of around 11,000–15,000 wafers per month by 2026, with further expansion planned to meet demand from major customers such as Broadcom and AMD. Relative to CoWoS, SoIC requires significantly higher capital investment, with roughly $6.8 billion to $7.0 billion in capex needed for every 10,000 wafers of capacity. This underscores both the high technical barriers and the strategic importance of the technology.\n\n• From an equipment standpoint, hybrid bonding and related process requirements are expected to benefit companies such as Besi, Applied Materials, and Tokyo Electron. Among Taiwanese firms, Hongsu (All Ring Tech), GMM, and CSUN are also expected to capture opportunities across relevant process segments.\n\n---\n\nhttps://t.co/eClRW1COeU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016577540106951894, "ret_signed_m1d": -0.016577540106951894, "alpha_spy_m1d": -0.013953886900695545, "alpha_c_m1d": -0.009119387973658744, "bench_spy_m1d": -0.002623653206256349, "bench_c_m1d": -0.00745815213329315, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018716577540107027, "ret_signed_p1d": 0.018716577540107027, "alpha_spy_p1d": 0.03267022292722721, "alpha_c_p1d": 0.026804642989576966, "bench_spy_p1d": -0.013953645387120184, "bench_c_p1d": -0.008088065449469939, "ret_p1w": -0.03208556149732622, "ret_signed_p1w": -0.03208556149732622, "alpha_spy_p1w": -0.008492298910296125, "alpha_c_p1w": -0.02631553399090636, "bench_spy_p1w": -0.023593262587030095, "bench_c_p1w": -0.005770027506419861, "ret_p1m": 0.11229946524064172, "ret_signed_p1m": 0.11229946524064172, "alpha_spy_p1m": 0.0659934824609889, "alpha_c_p1m": -0.029103463671733554, "bench_spy_p1m": 0.04630598277965281, "bench_c_p1m": 0.14140292891237527, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2378, -0.2378, -0.2191, -0.2058, -0.2031, -0.2031, -0.1951, -0.1845, -0.1898, -0.1738, -0.1738, -0.1552, -0.1099, -0.0912, -0.1072, -0.1019, -0.1045, -0.0992, -0.0885, -0.0885, -0.0992, -0.0725, -0.0619, -0.0539, -0.0725, -0.0619, -0.0566, -0.0645, -0.0512, -0.0299, -0.0379, -0.0539, -0.0592, -0.0406, -0.0486, -0.0592, -0.0512, -0.0326, 0.0021, 0.0207, 0.0207, 0.0207, 0.0207, 0.0207, 0.0207, 0.0207, 0.0207, 0.0127, 0.0474, 0.074, 0.0634, 0.0634, 0.0527, 0.0314, -0.0059, 0.0127, 0.0074, -0.0352, -0.0139, 0.0341, 0.0047, -0.0059, -0.0166, 0.0, 0.0187, -0.0107, -0.016, -0.0321, -0.0321, -0.0134, -0.016, -0.0267, -0.0481, -0.0588, -0.008, -0.0321, -0.0321, -0.0321, -0.0053, 0.0428, 0.0455, 0.0695, 0.0642, 0.0989, 0.1123, 0.115, 0.0856, 0.0829, 0.0963, 0.0963, 0.1123, 0.1684, 0.2112, 0.1845, 0.1658, 0.1417, 0.1417, 0.2166, 0.2032, 0.2032, 0.2353, 0.2246, 0.1952, 0.2059, 0.1872, 0.2139, 0.2112, 0.1979, 0.1791, 0.1684, 0.1925, 0.2059, 0.2353, 0.2139, 0.2299, 0.2273, 0.2594, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034042994199019582", "ticker_idx": 1, "ts": "2026-03-17T23:03:35+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "hybrid bonding and related process requirements are expected to benefit companies such as Besi", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC SoIC adoption for NVDA Rubin Ultra/Feynman identified as a structural tailwind for TSMC, Besi, AMAT, and TEL on hybrid bonding/advanced packaging demand.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA’s Next-Generation Chips: TSMC Prepares with SoIC (Taiwan Commercial Times)\n\n• NVIDIA unveiled its AI chip roadmap at GTC 2026, presenting an enhanced Rubin Ultra platform for 2027 and the next-generation Feynman architecture for 2028. The Rubin Ultra platform is expected to feature seven chips and five rack configurations, while Feynman is set to adopt custom HBM, high-density chiplets, and advanced 3D stacking technologies.\n\n• Industry sources view TSMC’s SoIC (System on Integrated Chips) as a critical enabling technology. SoIC vertically stacks dies to significantly increase transistor density, helping extend scaling beyond the limits of traditional Moore’s Law. As the Rubin Ultra generation drives larger chip sizes and greater HBM integration, packaging complexity is expected to rise sharply. With manufacturing efficiency for single-package solutions becoming more constrained, the importance of 3D stacking is becoming even more pronounced.\n\n• In particular, the Feynman generation is expected to mark a major inflection point for SoIC demand. Based on copper-to-copper hybrid bonding, SoIC enables shorter interconnects and lower power consumption, making it a key technology for addressing AI compute bottlenecks. Compared with conventional 2.5D packaging such as CoWoS, it offers higher bandwidth and better energy efficiency, making it well suited for next-generation AI training and inference chips.\n•On the manufacturing side, TSMC is expected to build SoIC capacity of around 11,000–15,000 wafers per month by 2026, with further expansion planned to meet demand from major customers such as Broadcom and AMD. Relative to CoWoS, SoIC requires significantly higher capital investment, with roughly $6.8 billion to $7.0 billion in capex needed for every 10,000 wafers of capacity. This underscores both the high technical barriers and the strategic importance of the technology.\n\n• From an equipment standpoint, hybrid bonding and related process requirements are expected to benefit companies such as Besi, Applied Materials, and Tokyo Electron. Among Taiwanese firms, Hongsu (All Ring Tech), GMM, and CSUN are also expected to capture opportunities across relevant process segments.\n\n---\n\nhttps://t.co/eClRW1COeU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02892009263346862, "ret_signed_m1d": -0.02892009263346862, "alpha_spy_m1d": -0.02629643942721227, "alpha_c_m1d": -0.02146194050017547, "bench_spy_m1d": -0.002623653206256349, "bench_c_m1d": -0.00745815213329315, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02096051409586952, "ret_signed_p1d": 0.02096051409586952, "alpha_spy_p1d": 0.034914159482989704, "alpha_c_p1d": 0.02904857954533946, "bench_spy_p1d": -0.013953645387120184, "bench_c_p1d": -0.008088065449469939, "ret_p1w": -0.023082997756191936, "ret_signed_p1w": -0.023082997756191936, "alpha_spy_p1w": 0.0005102648308381585, "alpha_c_p1w": -0.017312970249772075, "bench_spy_p1w": -0.023593262587030095, "bench_c_p1w": -0.005770027506419861, "ret_p1m": 0.1546830142478306, "ret_signed_p1m": 0.1546830142478306, "alpha_spy_p1m": 0.10837703146817779, "alpha_c_p1m": 0.013280085335455327, "bench_spy_p1m": 0.04630598277965281, "bench_c_p1m": 0.14140292891237527, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3025, -0.3062, -0.3038, -0.3001, -0.3019, -0.3019, -0.3019, -0.299, -0.2937, -0.2903, -0.2903, -0.2088, -0.1841, -0.1462, -0.1573, -0.1982, -0.1982, -0.1396, -0.1128, -0.143, -0.0785, -0.0825, -0.1043, -0.0825, -0.0764, -0.0714, -0.0711, -0.0722, -0.0674, -0.1008, -0.1374, -0.1271, -0.1266, -0.1502, -0.164, -0.147, -0.1218, -0.1117, -0.1029, -0.0931, -0.1048, -0.065, -0.0528, -0.0443, -0.0077, -0.082, -0.021, -0.0042, 0.0204, 0.0486, 0.0024, 0.005, 0.0003, -0.0377, 0.0135, 0.0011, -0.1706, -0.1175, -0.0735, -0.0716, -0.07, -0.0178, -0.0289, 0.0, 0.021, -0.0244, -0.0528, -0.0313, -0.0231, -0.0157, -0.0162, -0.0722, -0.0756, -0.0504, -0.0048, 0.009, 0.009, 0.009, 0.0154, 0.0984, 0.0937, 0.1223, 0.0979, 0.1568, 0.1547, 0.1701, 0.2035, 0.2046, 0.2093, 0.2327, 0.2831, 0.3383, 0.3243, 0.2565, 0.2784, 0.32, 0.32, 0.3045, 0.3563, 0.3617, 0.3601, 0.3926, 0.3953, 0.3339, 0.3788, 0.4242, 0.398, 0.3659, 0.3611, 0.4151, 0.4439, 0.461, 0.5059, 0.5165, 0.4759, 0.5358, 0.5187, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034042994199019582", "ticker_idx": 2, "ts": "2026-03-17T23:03:35+00:00", "symbol": "Applied Materials", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "hybrid bonding and related process requirements are expected to benefit companies such as Besi, Applied Materials, and Tokyo Electron", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC SoIC adoption for NVDA Rubin Ultra/Feynman identified as a structural tailwind for TSMC, Besi, AMAT, and TEL on hybrid bonding/advanced packaging demand.", "resolved_tickers": ["AMAT"], "resolution_method": "exchange_ticker", "resolution_reason": "Applied Materials AMAT US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA’s Next-Generation Chips: TSMC Prepares with SoIC (Taiwan Commercial Times)\n\n• NVIDIA unveiled its AI chip roadmap at GTC 2026, presenting an enhanced Rubin Ultra platform for 2027 and the next-generation Feynman architecture for 2028. The Rubin Ultra platform is expected to feature seven chips and five rack configurations, while Feynman is set to adopt custom HBM, high-density chiplets, and advanced 3D stacking technologies.\n\n• Industry sources view TSMC’s SoIC (System on Integrated Chips) as a critical enabling technology. SoIC vertically stacks dies to significantly increase transistor density, helping extend scaling beyond the limits of traditional Moore’s Law. As the Rubin Ultra generation drives larger chip sizes and greater HBM integration, packaging complexity is expected to rise sharply. With manufacturing efficiency for single-package solutions becoming more constrained, the importance of 3D stacking is becoming even more pronounced.\n\n• In particular, the Feynman generation is expected to mark a major inflection point for SoIC demand. Based on copper-to-copper hybrid bonding, SoIC enables shorter interconnects and lower power consumption, making it a key technology for addressing AI compute bottlenecks. Compared with conventional 2.5D packaging such as CoWoS, it offers higher bandwidth and better energy efficiency, making it well suited for next-generation AI training and inference chips.\n•On the manufacturing side, TSMC is expected to build SoIC capacity of around 11,000–15,000 wafers per month by 2026, with further expansion planned to meet demand from major customers such as Broadcom and AMD. Relative to CoWoS, SoIC requires significantly higher capital investment, with roughly $6.8 billion to $7.0 billion in capex needed for every 10,000 wafers of capacity. This underscores both the high technical barriers and the strategic importance of the technology.\n\n• From an equipment standpoint, hybrid bonding and related process requirements are expected to benefit companies such as Besi, Applied Materials, and Tokyo Electron. Among Taiwanese firms, Hongsu (All Ring Tech), GMM, and CSUN are also expected to capture opportunities across relevant process segments.\n\n---\n\nhttps://t.co/eClRW1COeU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017817665416652417, "ret_signed_m1d": -0.017817665416652417, "alpha_spy_m1d": -0.015194012210396068, "alpha_c_m1d": -0.010359513283359267, "bench_spy_m1d": -0.002623653206256349, "bench_c_m1d": -0.00745815213329315, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008483174222885026, "ret_signed_p1d": -0.008483174222885026, "alpha_spy_p1d": 0.005470471164235158, "alpha_c_p1d": -0.0003951087734150871, "bench_spy_p1d": -0.013953645387120184, "bench_c_p1d": -0.008088065449469939, "ret_p1w": 0.06108494019411581, "ret_signed_p1w": 0.06108494019411581, "alpha_spy_p1w": 0.0846782027811459, "alpha_c_p1w": 0.06685496770053567, "bench_spy_p1w": -0.023593262587030095, "bench_c_p1w": -0.005770027506419861, "ret_p1m": 0.11859504264777265, "ret_signed_p1m": 0.11859504264777265, "alpha_spy_p1m": 0.07228905986811984, "alpha_c_p1m": -0.022807886264602617, "bench_spy_p1m": 0.04630598277965281, "bench_c_p1m": 0.14140292891237527, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2817, -0.2734, -0.2661, -0.2626, -0.261, -0.261, -0.2579, -0.2546, -0.2633, -0.2718, -0.2718, -0.2381, -0.1943, -0.1612, -0.172, -0.2019, -0.1466, -0.1294, -0.1361, -0.1445, -0.0958, -0.0734, -0.0734, -0.0982, -0.0784, -0.0967, -0.0865, -0.0948, -0.0572, -0.0458, -0.0328, -0.0867, -0.0694, -0.097, -0.1567, -0.1386, -0.0861, -0.0633, -0.0675, -0.0369, -0.0695, 0.0057, 0.0057, 0.0177, 0.0465, 0.0493, 0.065, 0.0598, 0.0723, 0.1206, 0.066, 0.0563, 0.0559, -0.0032, 0.015, -0.0168, -0.0786, -0.0384, -0.0187, -0.0039, -0.0431, -0.031, -0.0178, 0.0, -0.0085, 0.0135, 0.0131, 0.0265, 0.0611, 0.0479, -0.0395, -0.0434, -0.0832, -0.0303, 0.0038, -0.0113, -0.0113, 0.0005, 0.0052, 0.0944, 0.1287, 0.1334, 0.1228, 0.1225, 0.1186, 0.1062, 0.1262, 0.1111, 0.1188, 0.1448, 0.146, 0.1832, 0.1487, 0.0813, 0.0855, 0.1192, 0.1039, 0.1104, 0.1656, 0.2161, 0.1651, 0.2354, 0.2586, 0.2234, 0.2388, 0.25, 0.2388, 0.1734, 0.1545, 0.2111, 0.214, 0.2276, 0.2276, 0.2922, 0.2734, 0.2774, 0.2785, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2034042994199019582", "ticker_idx": 3, "ts": "2026-03-17T23:03:35+00:00", "symbol": "Tokyo Electron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "hybrid bonding and related process requirements are expected to benefit companies such as Besi, Applied Materials, and Tokyo Electron", "tweet_type": "analysis", "timeline": "quarters", "summary": "TSMC SoIC adoption for NVDA Rubin Ultra/Feynman identified as a structural tailwind for TSMC, Besi, AMAT, and TEL on hybrid bonding/advanced packaging demand.", "resolved_tickers": ["8035.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Tokyo Electron listed on TSE as 8035.T", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA’s Next-Generation Chips: TSMC Prepares with SoIC (Taiwan Commercial Times)\n\n• NVIDIA unveiled its AI chip roadmap at GTC 2026, presenting an enhanced Rubin Ultra platform for 2027 and the next-generation Feynman architecture for 2028. The Rubin Ultra platform is expected to feature seven chips and five rack configurations, while Feynman is set to adopt custom HBM, high-density chiplets, and advanced 3D stacking technologies.\n\n• Industry sources view TSMC’s SoIC (System on Integrated Chips) as a critical enabling technology. SoIC vertically stacks dies to significantly increase transistor density, helping extend scaling beyond the limits of traditional Moore’s Law. As the Rubin Ultra generation drives larger chip sizes and greater HBM integration, packaging complexity is expected to rise sharply. With manufacturing efficiency for single-package solutions becoming more constrained, the importance of 3D stacking is becoming even more pronounced.\n\n• In particular, the Feynman generation is expected to mark a major inflection point for SoIC demand. Based on copper-to-copper hybrid bonding, SoIC enables shorter interconnects and lower power consumption, making it a key technology for addressing AI compute bottlenecks. Compared with conventional 2.5D packaging such as CoWoS, it offers higher bandwidth and better energy efficiency, making it well suited for next-generation AI training and inference chips.\n•On the manufacturing side, TSMC is expected to build SoIC capacity of around 11,000–15,000 wafers per month by 2026, with further expansion planned to meet demand from major customers such as Broadcom and AMD. Relative to CoWoS, SoIC requires significantly higher capital investment, with roughly $6.8 billion to $7.0 billion in capex needed for every 10,000 wafers of capacity. This underscores both the high technical barriers and the strategic importance of the technology.\n\n• From an equipment standpoint, hybrid bonding and related process requirements are expected to benefit companies such as Besi, Applied Materials, and Tokyo Electron. Among Taiwanese firms, Hongsu (All Ring Tech), GMM, and CSUN are also expected to capture opportunities across relevant process segments.\n\n---\n\nhttps://t.co/eClRW1COeU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009541054759165668, "ret_signed_m1d": 0.009541054759165668, "alpha_spy_m1d": 0.012164707965422017, "alpha_c_m1d": 0.016999206892458818, "bench_spy_m1d": -0.002623653206256349, "bench_c_m1d": -0.00745815213329315, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.038937625721796154, "ret_signed_p1d": 0.038937625721796154, "alpha_spy_p1d": 0.05289127110891634, "alpha_c_p1d": 0.04702569117126609, "bench_spy_p1d": -0.013953645387120184, "bench_c_p1d": -0.008088065449469939, "ret_p1w": 0.009025280851753914, "ret_signed_p1w": 0.009025280851753914, "alpha_spy_p1w": 0.03261854343878401, "alpha_c_p1w": 0.014795308358173775, "bench_spy_p1w": -0.023593262587030095, "bench_c_p1w": -0.005770027506419861, "ret_p1m": 0.13220150462015678, "ret_signed_p1m": 0.13220150462015678, "alpha_spy_p1m": 0.08589552184050397, "alpha_c_p1m": -0.009201424292218485, "bench_spy_p1m": 0.04630598277965281, "bench_c_p1m": 0.14140292891237527, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2184, -0.1955, -0.1447, -0.149, -0.1434, -0.1287, -0.1147, -0.1173, -0.115, -0.115, -0.115, -0.115, -0.0477, -0.0369, -0.0157, -0.0552, -0.0224, -0.0224, 0.058, 0.091, 0.0982, 0.0869, 0.0877, 0.0593, 0.0627, 0.0959, 0.0758, 0.0596, 0.0864, 0.1294, 0.0658, 0.0652, 0.0227, 0.0717, 0.0495, 0.0312, 0.058, 0.0469, 0.0701, 0.0701, 0.065, 0.0828, 0.067, 0.0869, 0.1189, 0.1509, 0.1336, 0.1336, 0.1444, 0.1921, 0.1684, 0.1349, 0.1225, 0.0985, 0.0493, 0.0758, 0.0776, 0.0036, 0.0322, 0.0451, 0.025, -0.0113, 0.0095, 0.0, 0.0389, 0.0142, 0.0142, -0.0124, 0.009, 0.0407, 0.0451, 0.0132, 0.0109, -0.031, 0.0224, -0.0104, -0.0, -0.0013, 0.0005, 0.1038, 0.0981, 0.1463, 0.1051, 0.1374, 0.1322, 0.1926, 0.1455, 0.1517, 0.1915, 0.1843, 0.1837, 0.1934, 0.2259, 0.1751, 0.1751, 0.1554, 0.235, 0.235, 0.235, 0.235, 0.3461, 0.3651, 0.3555, 0.3576, 0.3363, 0.3326, 0.3089, 0.2821, 0.2275, 0.1999, 0.2701, 0.297, 0.3581, 0.3383, 0.3665, 0.3618, 0.3644, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2033240029934579726", "ticker_idx": 0, "ts": "2026-03-15T17:52:53+00:00", "symbol": "semiconductor equipment", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I still like equipment names", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish on semiconductor equipment names long-term; near-term headwind from memory capex discipline, but durable demand ahead.", "resolved_tickers": ["AMAT", "LRCX", "KLAC", "ASML"], "resolution_method": "basket", "resolution_reason": "Semiconductor equipment basket: AMAT, Lam, KLA, ASML", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "Well, when you look at the fact that the profits memory companies are making are not flowing straight into equipment CAPEX, that is negative for equipment makers in the short term. But over the longer run, it will still lead to more disciplined investment. It may not be reckless, massive-scale spending, but even that disciplined CAPEX will still be quite meaningful. I still like equipment names.", "reply_text": null, "parent_text": "@jukan05 But this is not good for machine manufacturers like ASML, KLA, NVMI, etc. ?", "quoted_text": null, "parent_author": "nav09284172", "ret_m1d": -0.020397903960987424, "ret_signed_m1d": -0.020397903960987424, "alpha_spy_m1d": -0.010323563067934388, "alpha_c_m1d": -0.0036685512222805194, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.016729352738706904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022556541001829356, "ret_signed_p1d": 0.022556541001829356, "alpha_spy_p1d": 0.019925986131854145, "alpha_c_p1d": 0.015042346865508538, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.0075141941363208176, "ret_p1w": 0.038765692705574895, "ret_signed_p1w": 0.038765692705574895, "alpha_spy_p1w": 0.05649312236000398, "alpha_c_p1w": 0.04526454523630957, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.006498852530734678, "ret_p1m": 0.18423545149713982, "ret_signed_p1m": 0.18423545149713982, "alpha_spy_p1m": 0.14339040632781597, "alpha_c_p1m": 0.03679438581764605, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.14744106567949378, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2568, -0.2294, -0.2111, -0.2022, -0.2, -0.195, -0.195, -0.1916, -0.1978, -0.2042, -0.2147, -0.2147, -0.163, -0.1156, -0.0833, -0.0991, -0.1214, -0.06, -0.0471, -0.0553, -0.0662, -0.0125, 0.0084, 0.0084, -0.0187, 0.0052, -0.0053, -0.0047, 0.0085, 0.056, 0.0563, 0.0852, 0.0044, 0.0136, -0.0196, -0.0766, -0.0618, 0.0025, 0.0088, 0.0006, 0.0305, 0.0076, 0.0339, 0.0339, 0.0407, 0.0639, 0.0579, 0.077, 0.0742, 0.085, 0.1154, 0.0744, 0.0637, 0.0572, -0.0011, 0.0233, -0.008, -0.0697, -0.0195, -0.001, 0.0096, -0.0271, -0.0204, 0.0, 0.0226, 0.0124, 0.0357, 0.018, 0.0388, 0.0688, 0.0544, -0.0205, -0.028, -0.0706, -0.0137, 0.0198, 0.0036, 0.0036, 0.0108, 0.0182, 0.1084, 0.1457, 0.1596, 0.1705, 0.1842, 0.16, 0.1369, 0.1683, 0.1648, 0.1547, 0.1713, 0.1596, 0.2076, 0.1792, 0.1281, 0.1294, 0.1451, 0.1335, 0.1276, 0.175, 0.2452, 0.2058, 0.2644, 0.2636, 0.2329, 0.2615, 0.2767, 0.2269, 0.1893, 0.1741, 0.2418, 0.264, 0.2865, 0.2865, 0.3439, 0.3192, 0.3155, 0.3162, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2018916685957452011", "ticker_idx": 0, "ts": "2026-02-04T05:17:02+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we remain bullish on Apple (AAPL Buy), benefiting from its formidable procurement capabilities and ongoing negotiations for favorable memory supply terms with Korean vendors", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "GF Securities reiterates Buy on AAPL (strong memory procurement) and NVDA (performance/cost-per-token lead); cautious on Hyperscaler Capex boost and ODM/OEM demand.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "《 GF Securities Overseas Electronic Communications 》\n\nAnalysis of Memory Price Impact on Consumer and AI Segments\n\nHigh Memory Prices Leading to Smartphone \"Downgrading\"\nDue to the continuous price hikes of LPDDR and NAND, smartphone manufacturers' profit margins are being squeezed amidst moderate market demand. Memory now accounts for 20-30% of the smartphone Bill of Materials (BOM). While upgrade plans for high-end and flagship models remain unchanged, a \"downgrading\" trend for mid-to-low-end models has become inevitable. We observe that some Chinese manufacturers are evaluating external storage solutions or re-introducing MicroSD card slots in flagship models to reduce internal storage capacity. Consequently, we have revised down the 2026 global smartphone shipment forecast to a 6% year-on-year (YoY) decline. We believe price wars will gradually ease, as further price cuts are unlikely to stimulate new demand. On the other hand, we remain bullish on Apple (AAPL Buy), benefiting from its formidable procurement capabilities and ongoing negotiations for favorable memory supply terms with Korean vendors.\n\n1GW GB300 NVL72 Rack Deployment Cost to Reach $36.8B by Year-End (Previously ~$35B)\nDuring Microsoft’s earnings call, management noted that \"rising memory prices will affect capital expenditure (Capex),\" triggering investor concerns regarding Capex expansion. However, our analysis shows that recent memory price increases will add approximately $300,000 in costs per rack, or an additional $1.8 billion per 1GW data center deployment. This represents a 7% increase in rack BOM and a 5% increase in total investment. Therefore, we believe recent Capex upward revisions by Hyperscalers (e.g., Meta) primarily reflect rising memory costs and higher prices for general-purpose servers. General-purpose servers typically account for 20-30% of Hyperscaler Capex; if their BOM rises by 30%, it would result in a 7-8% overall increase in Capex. Notably, the entire AI supply chain is prioritizing this issue, particularly Neo-clouds, which are more sensitive to these costs.\n\nHigher Cost Pressure on HGX 8-GPU and General-Purpose Servers\nAccording to our estimates, rising memory costs will lead to an 11% BOM increase for HGX 8-GPU B300 AI servers, aligning with recent reports of 15-20% hikes in system prices. For general-purpose servers, memory and storage costs now represent 72% of the total BOM (up from 42% previously), putting significant pricing pressure on server brands. Consequently, we expect OEMs to further raise prices to protect their profitability levels.\n\nImpact on Equities\nWe believe the Capex upward revisions by Meta and potentially other Hyperscalers have no substantial positive impact, as the majority of the increase merely reflects rising memory costs. Furthermore, while subsequent price hikes for 8-GPU and general-purpose servers are a short-term positive for ODMs/OEMs, demand uncertainty from small-to-mid-sized cloud customers may dampen this effect. Overall, we maintain a positive outlook on Apple (AAPL Buy) due to its strong procurement power, and NVIDIA (NVDA Buy) for its leading advantage in performance and cost per token.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025353353041648208, "ret_signed_m1d": -0.025353353041648208, "alpha_spy_m1d": -0.030220852506726348, "alpha_c_m1d": -0.05402415564736274, "bench_spy_m1d": 0.00486749946507814, "bench_c_m1d": 0.02867080260571453, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002097593092373451, "ret_signed_p1d": -0.002097593092373451, "alpha_spy_p1d": 0.010391618163828276, "alpha_c_p1d": 0.015930100008663906, "bench_spy_p1d": -0.012489211256201727, "bench_c_p1d": -0.018027693101037356, "ret_p1w": -0.002648171193851079, "ret_signed_p1w": -0.002648171193851079, "alpha_spy_p1w": -0.01105699191558851, "alpha_c_p1w": -0.037762689702652175, "bench_spy_p1w": 0.008408820721737431, "bench_c_p1w": 0.035114518508801096, "ret_p1m": -0.057710625703980334, "ret_signed_p1m": -0.057710625703980334, "alpha_spy_p1m": -0.050598906985406145, "alpha_c_p1m": -0.07262524798918268, "bench_spy_p1m": -0.007111718718574189, "bench_c_p1m": 0.01491462228520235, "ret_p3m": 0.002166595047650155, "ret_signed_p3m": 0.002166595047650155, "alpha_spy_p3m": -0.04706307727332759, "alpha_c_p3m": -0.1725474253078776, "bench_spy_p3m": 0.049229672320977746, "bench_c_p3m": 0.17471402035552774, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0299, -0.0255, -0.0045, -0.0109, -0.0128, -0.0148, -0.0327, -0.0327, -0.0287, -0.037, -0.0181, -0.0021, 0.0017, 0.0038, 0.0038, 0.0085, 0.0239, 0.0351, 0.0277, 0.0152, 0.0083, 0.0051, 0.0025, 0.0083, 0.0056, 0.0065, -0.0086, -0.0068, -0.0168, -0.0156, -0.0102, -0.02, -0.0149, -0.0097, -0.0097, -0.0112, -0.0099, -0.0123, -0.0167, -0.0167, -0.0198, -0.0334, -0.0511, -0.0584, -0.0631, -0.0619, -0.0587, -0.0558, -0.0598, -0.0661, -0.0758, -0.0758, -0.1077, -0.1043, -0.1018, -0.1029, -0.0762, -0.0659, -0.0725, -0.0659, -0.0615, -0.0234, -0.0254, 0.0, -0.0021, 0.0059, -0.0058, -0.0092, -0.0026, -0.0525, -0.074, -0.074, -0.0447, -0.043, -0.0567, -0.0422, -0.0364, -0.0148, -0.0072, -0.0119, -0.0436, -0.0417, -0.0452, -0.0496, -0.0577, -0.068, -0.0592, -0.0558, -0.0558, -0.0741, -0.0945, -0.0848, -0.0796, -0.0952, -0.0987, -0.1022, -0.0896, -0.089, -0.0855, -0.0845, -0.0993, -0.1072, -0.0812, -0.0746, -0.0735, -0.0735, -0.0629, -0.0823, -0.0627, -0.057, -0.057, -0.0617, -0.063, -0.0355, -0.0465, -0.0217, -0.0115, -0.0364, -0.0111, -0.0101, -0.0187, -0.0312, -0.02, -0.0219, -0.0177, 0.0141, 0.0022, 0.0288, 0.0408, 0.0406, 0.0619, 0.0605, 0.0682, 0.083, 0.0806, 0.0879, 0.0792, 0.0833, 0.0952, 0.1051, 0.119, 0.119, 0.1172, 0.1264, 0.1324, 0.1307, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2018916685957452011", "ticker_idx": 1, "ts": "2026-02-04T05:17:02+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA (NVDA Buy) for its leading advantage in performance and cost per token", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "GF Securities reiterates Buy on AAPL (strong memory procurement) and NVDA (performance/cost-per-token lead); cautious on Hyperscaler Capex boost and ODM/OEM demand.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《 GF Securities Overseas Electronic Communications 》\n\nAnalysis of Memory Price Impact on Consumer and AI Segments\n\nHigh Memory Prices Leading to Smartphone \"Downgrading\"\nDue to the continuous price hikes of LPDDR and NAND, smartphone manufacturers' profit margins are being squeezed amidst moderate market demand. Memory now accounts for 20-30% of the smartphone Bill of Materials (BOM). While upgrade plans for high-end and flagship models remain unchanged, a \"downgrading\" trend for mid-to-low-end models has become inevitable. We observe that some Chinese manufacturers are evaluating external storage solutions or re-introducing MicroSD card slots in flagship models to reduce internal storage capacity. Consequently, we have revised down the 2026 global smartphone shipment forecast to a 6% year-on-year (YoY) decline. We believe price wars will gradually ease, as further price cuts are unlikely to stimulate new demand. On the other hand, we remain bullish on Apple (AAPL Buy), benefiting from its formidable procurement capabilities and ongoing negotiations for favorable memory supply terms with Korean vendors.\n\n1GW GB300 NVL72 Rack Deployment Cost to Reach $36.8B by Year-End (Previously ~$35B)\nDuring Microsoft’s earnings call, management noted that \"rising memory prices will affect capital expenditure (Capex),\" triggering investor concerns regarding Capex expansion. However, our analysis shows that recent memory price increases will add approximately $300,000 in costs per rack, or an additional $1.8 billion per 1GW data center deployment. This represents a 7% increase in rack BOM and a 5% increase in total investment. Therefore, we believe recent Capex upward revisions by Hyperscalers (e.g., Meta) primarily reflect rising memory costs and higher prices for general-purpose servers. General-purpose servers typically account for 20-30% of Hyperscaler Capex; if their BOM rises by 30%, it would result in a 7-8% overall increase in Capex. Notably, the entire AI supply chain is prioritizing this issue, particularly Neo-clouds, which are more sensitive to these costs.\n\nHigher Cost Pressure on HGX 8-GPU and General-Purpose Servers\nAccording to our estimates, rising memory costs will lead to an 11% BOM increase for HGX 8-GPU B300 AI servers, aligning with recent reports of 15-20% hikes in system prices. For general-purpose servers, memory and storage costs now represent 72% of the total BOM (up from 42% previously), putting significant pricing pressure on server brands. Consequently, we expect OEMs to further raise prices to protect their profitability levels.\n\nImpact on Equities\nWe believe the Capex upward revisions by Meta and potentially other Hyperscalers have no substantial positive impact, as the majority of the increase merely reflects rising memory costs. Furthermore, while subsequent price hikes for 8-GPU and general-purpose servers are a short-term positive for ODMs/OEMs, demand uncertainty from small-to-mid-sized cloud customers may dampen this effect. Overall, we maintain a positive outlook on Apple (AAPL Buy) due to its strong procurement power, and NVIDIA (NVDA Buy) for its leading advantage in performance and cost per token.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03530622338940925, "ret_signed_m1d": 0.03530622338940925, "alpha_spy_m1d": 0.030438723924331113, "alpha_c_m1d": -0.00568640556317912, "bench_spy_m1d": 0.00486749946507814, "bench_c_m1d": 0.04099262895258837, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01326138417734124, "ret_signed_p1d": -0.01326138417734124, "alpha_spy_p1d": -0.0007721729211395134, "alpha_c_p1d": -0.010774651239868871, "bench_spy_p1d": -0.012489211256201727, "bench_c_p1d": -0.002486732937472369, "ret_p1w": 0.0910500286550322, "ret_signed_p1w": 0.0910500286550322, "alpha_spy_p1w": 0.08264120793329477, "alpha_c_p1w": 0.005295328095495266, "bench_spy_p1w": 0.008408820721737431, "bench_c_p1w": 0.08575470055953693, "ret_p1m": 0.05252885257752493, "ret_signed_p1m": 0.05252885257752493, "alpha_spy_p1m": 0.05964057129609912, "alpha_c_p1m": 0.017635345763234334, "bench_spy_p1m": -0.007111718718574189, "bench_c_p1m": 0.0348935068142906, "ret_p3m": 0.1395070808853327, "ret_signed_p3m": 0.1395070808853327, "alpha_spy_p3m": 0.09027740856435496, "alpha_c_p3m": -0.1870989164955652, "bench_spy_p3m": 0.049229672320977746, "bench_c_p3m": 0.3266059973808979, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0801, 0.1427, 0.1088, 0.1125, 0.0727, 0.0917, 0.0712, 0.0411, 0.0707, 0.037, 0.0269, 0.0479, 0.0208, 0.0348, 0.0348, 0.0161, 0.0328, 0.0417, 0.0309, 0.0528, 0.0472, 0.0652, 0.0619, 0.0551, 0.0387, 0.0048, 0.0121, 0.0203, -0.0187, -0.0003, 0.039, 0.0545, 0.0862, 0.0828, 0.0828, 0.0938, 0.0805, 0.0766, 0.0707, 0.0707, 0.0842, 0.08, 0.0749, 0.0857, 0.0623, 0.0613, 0.0617, 0.0667, 0.0514, 0.0738, 0.0691, 0.0691, 0.0223, 0.0524, 0.0611, 0.0774, 0.0705, 0.0823, 0.0995, 0.1052, 0.0973, 0.0656, 0.0353, 0.0, -0.0133, 0.0644, 0.091, 0.0824, 0.0911, 0.0732, 0.0495, 0.0495, 0.0619, 0.0792, 0.0787, 0.0897, 0.0997, 0.1071, 0.1227, 0.0614, 0.0172, 0.0476, 0.0336, 0.0508, 0.0525, 0.0208, 0.0486, 0.0607, 0.068, 0.0514, 0.0348, 0.0519, 0.0445, 0.0357, 0.0251, -0.0085, 0.0084, 0.0059, 0.0258, -0.0169, -0.0382, -0.0517, 0.0013, 0.009, 0.0184, 0.0184, 0.0199, 0.0225, 0.0454, 0.0559, 0.083, 0.0869, 0.1282, 0.1417, 0.1388, 0.1579, 0.1601, 0.1475, 0.1626, 0.1462, 0.1957, 0.2436, 0.2238, 0.2013, 0.1458, 0.1393, 0.1395, 0.1281, 0.1932, 0.2143, 0.2355, 0.2598, 0.2675, 0.2965, 0.3534, 0.2936, 0.2764, 0.2666, 0.283, 0.2602, 0.2362, 0.2362, 0.2335, 0.2206, 0.23, 0.2122, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2032638476336902243", "ticker_idx": 0, "ts": "2026-03-14T02:02:32+00:00", "symbol": "HBM memory makers", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "That's hugely bullish for the HBM makers", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Feynman GPU expected to use 1TB/1.3TB HBM5 per JPM, hugely bullish for HBM makers; MU called out explicitly.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM memory makers: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Looks like Feynman’s specs may have changed lol.\n\nAccording to JPM’s estimate, it’s expected to use even more HBM capacity than Rubin Ultra.\n\nJPM appears to expect the Feynman GPU to use 1TB/1.3TB of HBM5.\n\nThat’s hugely bullish for the HBM makers.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04291128195265215, "ret_signed_m1d": -0.04291128195265215, "alpha_spy_m1d": -0.03283694105959911, "alpha_c_m1d": -0.026181929213945243, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.016729352738706904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.022823515339485862, "ret_signed_p1d": 0.022823515339485862, "alpha_spy_p1d": 0.020192960469510652, "alpha_c_p1d": 0.015309321203165045, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.0075141941363208176, "ret_p1w": -0.04652663458477713, "ret_signed_p1w": -0.04652663458477713, "alpha_spy_p1w": -0.028799204930348044, "alpha_c_p1w": -0.04002778205404245, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.006498852530734678, "ret_p1m": 0.09449580941850193, "ret_signed_p1m": 0.09449580941850193, "alpha_spy_p1m": 0.05365076424917807, "alpha_c_p1m": -0.05294525626099185, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.14744106567949378, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.452, -0.4348, -0.4257, -0.399, -0.3961, -0.3876, -0.3876, -0.3749, -0.3482, -0.345, -0.3505, -0.3505, -0.3038, -0.2828, -0.2475, -0.2413, -0.2498, -0.2399, -0.238, -0.2497, -0.2472, -0.236, -0.205, -0.2016, -0.2143, -0.1895, -0.1731, -0.1678, -0.1863, -0.1354, -0.0974, -0.0932, -0.0929, -0.1205, -0.0778, -0.1076, -0.1421, -0.1354, -0.1137, -0.1263, -0.1002, -0.0688, -0.0687, -0.0687, -0.0777, -0.0617, -0.0441, -0.017, -0.0166, 0.012, 0.0309, 0.0747, 0.0567, 0.0569, -0.0475, -0.1029, -0.0399, -0.0719, -0.1137, -0.0429, -0.0216, -0.044, -0.0429, 0.0, 0.0228, 0.0781, 0.0361, 0.0159, -0.0465, -0.029, -0.0373, -0.0944, -0.093, -0.1463, -0.1729, -0.0797, -0.1237, -0.0942, -0.0698, -0.0537, 0.0331, 0.0208, 0.0336, 0.0337, 0.0945, 0.1066, 0.1254, 0.1119, 0.1172, 0.1457, 0.1715, 0.1802, 0.1817, 0.2355, 0.2185, 0.2339, 0.2208, 0.2397, 0.3419, 0.39, 0.5219, 0.5347, 0.616, 0.7491, 0.7005, 0.7856, 0.7839, 0.6483, 0.6415, 0.6123, 0.6383, 0.7693, 0.7489, 0.7489, 0.9077, 1.0118, 1.0107, 1.0926, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2032638476336902243", "ticker_idx": 1, "ts": "2026-03-14T02:02:32+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Feynman GPU expected to use 1TB/1.3TB HBM5 per JPM, hugely bullish for HBM makers; MU called out explicitly.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Looks like Feynman’s specs may have changed lol.\n\nAccording to JPM’s estimate, it’s expected to use even more HBM capacity than Rubin Ultra.\n\nJPM appears to expect the Feynman GPU to use 1TB/1.3TB of HBM5.\n\nThat’s hugely bullish for the HBM makers.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03546847335497094, "ret_signed_m1d": -0.03546847335497094, "alpha_spy_m1d": -0.025394132461917907, "alpha_c_m1d": -0.018739120616264038, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.016729352738706904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.045020383429470456, "ret_signed_p1d": 0.045020383429470456, "alpha_spy_p1d": 0.042389828559495246, "alpha_c_p1d": 0.03750618929314964, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.0075141941363208176, "ret_p1w": -0.08476684429074977, "ret_signed_p1w": -0.08476684429074977, "alpha_spy_p1w": -0.06703941463632068, "alpha_c_p1w": -0.07826799176001509, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.006498852530734678, "ret_p1m": 0.05444914916034849, "ret_signed_p1m": 0.05444914916034849, "alpha_spy_p1m": 0.01360410399102463, "alpha_c_p1m": -0.09299191651914529, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.14744106567949378, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4897, -0.4376, -0.3983, -0.3742, -0.3749, -0.3514, -0.3514, -0.3556, -0.3337, -0.3376, -0.354, -0.354, -0.2861, -0.2935, -0.2227, -0.2314, -0.2598, -0.2189, -0.2171, -0.2347, -0.2455, -0.238, -0.1789, -0.1789, -0.1738, -0.1193, -0.1001, -0.0954, -0.1193, -0.0714, -0.0148, -0.0136, -0.0609, -0.0091, -0.0506, -0.1412, -0.1333, -0.1066, -0.132, -0.1552, -0.0712, -0.063, -0.0682, -0.0682, -0.0951, -0.0472, -0.0553, -0.0309, -0.0471, -0.0538, -0.029, -0.0594, -0.0666, -0.0659, -0.1406, -0.0929, -0.1013, -0.1618, -0.1188, -0.0876, -0.0523, -0.0825, -0.0355, 0.0, 0.045, 0.0451, 0.0056, -0.0428, -0.0848, -0.1047, -0.1352, -0.1954, -0.1914, -0.2713, -0.235, -0.167, -0.1707, -0.1707, -0.1446, -0.145, -0.079, -0.0455, -0.0476, -0.0341, 0.0544, 0.0331, 0.0354, 0.0305, 0.0154, 0.0176, 0.1039, 0.0908, 0.1248, 0.1878, 0.1419, 0.174, 0.1711, 0.2278, 0.3053, 0.4497, 0.5094, 0.4642, 0.6911, 0.801, 0.7359, 0.8198, 0.7572, 0.6409, 0.5433, 0.5822, 0.6575, 0.7257, 0.7006, 0.7006, 1.0286, 1.1023, 1.0912, 1.1988, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2033411211371676021", "ticker_idx": 0, "ts": "2026-03-16T05:13:06+00:00", "symbol": "Zhen Ding Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we have raised our 2027 earnings forecast. We now forecast Zhen Ding's 2026/2027 EPS at NT$14.00/22.42, which is 15%/33% above market consensus. Based on 15x the average of 2026E/2027E EPS, we set our target price at NT$273", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish Zhen Ding on AI/substrate supercycle; 2026/2027 EPS 15%/33% above consensus, PT NT$273 on optical module 10x growth and ABF ramp.", "resolved_tickers": ["4935.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Zhen Ding Technology listed on TWSE as 4935.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "《 GF Overseas Electronics & Communications 》\n\n☄️Zhen Ding 4Q25 Review: Entering the AI Era\n\n☀️Rising confidence in AI and substrate businesses:\nZhen Ding reported 4Q25 revenue of NT$56.9 billion and a gross margin of 22.6%, both above market expectations. On the earnings call, management expressed strong confidence in AI-driven growth, targeting 6–7x revenue growth in optical modules and AI server PCB businesses. In addition, management reiterated its optimistic outlook for the IC substrate business in both revenue and profitability. Given improving profitability in the substrate segment, alongside cost pressure from upstream PCB materials, we have slightly adjusted our 2026 earnings forecast. Meanwhile, with better visibility in the AI business, we have raised our 2027 earnings forecast. We now forecast Zhen Ding’s 2026/2027 EPS at NT$14.00/22.42, which is 15%/33% above market consensus. Based on 15x the average of 2026E/2027E EPS, we set our target price at NT$273.\n\n☀️A substrate supercycle is coming:\nWe estimate AI demand accounted for 30% / around 50% of total ABF demand in 2025/2026, and AI-related demand is expected to double again in 2027. To support surging customer orders, Zhen Ding is pushing ahead with an aggressive capacity expansion plan. Fab 1 expansion is set to be completed in 2026, while equipment move-in for Fab 2 will begin in 2H26. Together, the two facilities will support annual output value of RMB 7 billion. In the first two months of 2026, Zhen Ding’s ABF revenue grew more than 50% YoY. Higher prices for low-CTE glass, together with better utilization, should further drive earnings upside. Management also stated that the ABF business is expected to turn profit-contributive in 2026, after dragging down 2025 EPS by NT$1.5.\n\n☀️AI business continues to ramp:\nWe reiterate our forecast of RMB 5.7 billion AI revenue in 2026, above consensus expectations. Optical modules will be the key growth driver in 2026, with management indicating this segment could achieve nearly 10x growth this year, followed by another doubling in 2027. In AI PCB, driven by demand ramp from ASIC customers (AWS/Google), scaling of NVIDIA compute trays, and capacity expansion at the Huai’an plant in 3Q26, Zhen Ding’s AI business revenue is expected to accelerate starting in 2Q. Looking ahead to 2027, CoWoP will become another important revenue contributor. Product R&D is progressing smoothly, and FII is advancing the SLT project, with Zhen Ding being one of the key suppliers.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014005601941584622, "ret_signed_m1d": 0.014005601941584622, "alpha_spy_m1d": 0.024079942834637658, "alpha_c_m1d": 0.02827277142322837, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.014267169481643749, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021008402912377155, "ret_signed_p1d": 0.021008402912377155, "alpha_spy_p1d": 0.018377848042401945, "alpha_c_p1d": 0.015532165675769427, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.005476237236607728, "ret_p1w": -0.009803985471667054, "ret_signed_p1w": -0.009803985471667054, "alpha_spy_p1w": 0.00792344418276203, "alpha_c_p1w": 0.002155447329084015, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.011959432800751069, "ret_p1m": 0.021008402912377155, "ret_signed_p1m": 0.021008402912377155, "alpha_spy_p1m": -0.0198366422569467, "alpha_c_p1m": -0.04632074773308026, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.06732915064545741, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1401, 0.1457, 0.1555, 0.1555, 0.1485, 0.1485, 0.1485, 0.1443, 0.1429, 0.1429, 0.1457, 0.1457, 0.1443, 0.1261, 0.1204, 0.1303, 0.1345, 0.1289, 0.1401, 0.1331, 0.1387, 0.1373, 0.1485, 0.1625, 0.1415, 0.1415, 0.1485, 0.1429, 0.1289, 0.1246, 0.1261, 0.119, 0.1204, 0.105, 0.0924, 0.0938, 0.091, 0.0826, 0.0784, 0.0756, 0.084, 0.084, 0.084, 0.084, 0.084, 0.084, 0.084, 0.084, 0.1078, 0.112, 0.1022, 0.0924, 0.0924, 0.098, 0.091, 0.056, 0.0616, 0.0728, 0.0252, 0.0014, 0.0028, 0.0098, 0.014, 0.0, 0.021, 0.028, 0.007, 0.0126, -0.0098, 0.0042, 0.021, 0.084, 0.084, 0.0434, 0.007, 0.014, -0.0084, -0.0084, -0.0084, -0.0112, 0.0112, 0.0182, 0.0098, -0.0028, 0.021, 0.0084, 0.0588, 0.0742, 0.0924, 0.0952, 0.119, 0.0882, 0.0868, 0.0756, 0.0658, 0.0616, 0.0504, 0.0504, 0.0574, 0.0658, 0.0518, 0.0616, 0.0434, 0.0322, 0.0098, 0.0098, 0.0098, 0.0084, 0.0126, 0.014, 0.0112, 0.0252, 0.0364, 0.0434, 0.0406, 0.0238, 0.042, 0.1457, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2033441858769371425", "ticker_idx": 0, "ts": "2026-03-16T07:14:53+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Should Samsung Foundry secure AMD chip orders, the earnings recovery trajectory would likely accelerate further. It could serve as a pivotal validation of Samsung Foundry's technical capabilities", "tweet_type": "analysis", "timeline": "quarters", "summary": "AMD-Samsung partnership on HBM4, DRAM/NAND supply, and foundry could accelerate AMD earnings and Samsung's foundry profitability recovery.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung·AMD 'Memory·Foundry' Partnership Takes Shape\n\nAMD CEO Lisa Su is making her first-ever visit to Korea, drawing significant attention to the scope of potential collaboration with Samsung Semiconductor.\n\nAccording to industry sources on the 16th, Su is set to arrive in Korea on the 18th. She is scheduled to meet successively with Naver CEO Choi Soo-yeon and Samsung Electronics DS Division Head Vice Chairman Jun Young-hyun, followed by a meeting with Samsung Electronics Chairman Lee Jae-yong. It will be the second meeting between Chairman Lee and CEO Su in roughly three months, following their last encounter in the U.S. in December. AMD is widely regarded as NVIDIA's primary competitor in the AI accelerator market.\n\nSu's visit to Korea is highly unusual — it marks her first trip to the country since taking office in 2014. Historically, Su's Asia itinerary has centered on Japan, China, and Taiwan. Japan has been active ground for automotive semiconductor collaboration through AMD's acquisition of Xilinx; Taiwan is home to TSMC, the world's largest foundry; and China remains one of the world's largest markets. Korea's emergence on this circuit underscores how dramatically Samsung Memory's strategic importance has grown.\n\nIndustry sources indicate that the visit will cover discussions on long-term supply agreements spanning commodity DRAM and NAND flash in addition to broader memory categories — a signal that even hyperscale customers like AMD are facing serious memory procurement constraints. The agenda also includes HBM4 supply terms: Samsung's sixth-generation High Bandwidth Memory is expected to be incorporated into AMD's latest MI450 series AI accelerators.\n\nFoundry collaboration is attracting attention on par with the memory discussions. The two companies are reportedly negotiating a contract for AMD to allocate volume to Samsung's leading-edge foundry process. If realized, a formal foundry partnership between the two would be essentially unprecedented. Samsung did produce some AMD graphics chips on its 14nm process back in 2016, but that engagement was not considered meaningful in scale — unlike more recent wins such as the Tesla order.\n\nShould Samsung Foundry secure AMD chip orders, the earnings recovery trajectory would likely accelerate further. It could serve as a pivotal validation of Samsung Foundry's technical capabilities as a credible \"alternative to TSMC\" in the eyes of other major fabless customers. With Samsung Memory already entering an unprecedented upcycle, the question now is whether Foundry can pull forward its own path to profitability.\n\nA senior industry official noted, \"Samsung and AMD are understood to be coordinating plans to jointly disclose the details of this collaboration directly,\" adding that \"Samsung Semiconductor's strategic importance is set to be reaffirmed.\"\n\nSu is also reportedly pursuing meetings with Blue House officials or other government figures during her stay. This follows SoftBank Group Chairman Masayoshi Son's visit to Korea last December, during which he visited the Blue House directly to discuss AI cooperation.\n\n$AMD\n\n---\n\nhttps://t.co/89wLB3otbf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016227502348625578, "ret_signed_m1d": -0.016227502348625578, "alpha_spy_m1d": -0.006153161455572542, "alpha_c_m1d": 0.0005018503900813265, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.016729352738706904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0013735083423845884, "ret_signed_p1d": -0.0013735083423845884, "alpha_spy_p1d": -0.0040040632123597986, "alpha_c_p1d": -0.008887702478705406, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.0075141941363208176, "ret_p1w": 0.03103057680286847, "ret_signed_p1w": 0.03103057680286847, "alpha_spy_p1w": 0.048758006457297554, "alpha_c_p1w": 0.03752942933360315, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.006498852530734678, "ret_p1m": 0.2975379232289952, "ret_signed_p1m": 0.2975379232289952, "alpha_spy_p1m": 0.25669287805967134, "alpha_c_p1m": 0.15009685754950142, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.14744106567949378, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0078, 0.0228, 0.0857, 0.0934, 0.0932, 0.0939, 0.0939, 0.0937, 0.0968, 0.0954, 0.0894, 0.0894, 0.1368, 0.1246, 0.0904, 0.0684, 0.0412, 0.0335, 0.0565, 0.1241, 0.1375, 0.1594, 0.1793, 0.1793, 0.1798, 0.2707, 0.2907, 0.321, 0.2784, 0.2821, 0.2857, 0.2828, 0.2042, 0.2528, 0.2316, 0.0184, -0.0208, 0.0603, 0.0988, 0.0864, 0.0865, 0.0476, 0.0546, 0.0546, 0.0331, 0.018, 0.0345, 0.0182, 0.0001, 0.0878, 0.0726, 0.0361, 0.0185, 0.0104, -0.0286, 0.0279, 0.0146, -0.0211, 0.031, 0.0338, 0.042, 0.0059, -0.0162, 0.0, -0.0014, 0.0147, 0.0442, 0.0242, 0.031, 0.0447, 0.1205, 0.0366, 0.0275, -0.0027, 0.0348, 0.0693, 0.1064, 0.1064, 0.1201, 0.1269, 0.1793, 0.2038, 0.2465, 0.2556, 0.2975, 0.3131, 0.4155, 0.4162, 0.3987, 0.4472, 0.5437, 0.5532, 0.7693, 0.7023, 0.6442, 0.7149, 0.8033, 0.8341, 0.7374, 0.8072, 1.1436, 1.0778, 1.3155, 1.3339, 1.2804, 1.2663, 1.2876, 1.1574, 1.1416, 1.1063, 1.2768, 1.2871, 1.3782, 1.3782, 1.5633, 1.5208, 1.6355, 1.6254, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2033441858769371425", "ticker_idx": 1, "ts": "2026-03-16T07:14:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Memory already entering an unprecedented upcycle, the question now is whether Foundry can pull forward its own path to profitability", "tweet_type": "analysis", "timeline": "quarters", "summary": "AMD-Samsung partnership on HBM4, DRAM/NAND supply, and foundry could accelerate AMD earnings and Samsung's foundry profitability recovery.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung·AMD 'Memory·Foundry' Partnership Takes Shape\n\nAMD CEO Lisa Su is making her first-ever visit to Korea, drawing significant attention to the scope of potential collaboration with Samsung Semiconductor.\n\nAccording to industry sources on the 16th, Su is set to arrive in Korea on the 18th. She is scheduled to meet successively with Naver CEO Choi Soo-yeon and Samsung Electronics DS Division Head Vice Chairman Jun Young-hyun, followed by a meeting with Samsung Electronics Chairman Lee Jae-yong. It will be the second meeting between Chairman Lee and CEO Su in roughly three months, following their last encounter in the U.S. in December. AMD is widely regarded as NVIDIA's primary competitor in the AI accelerator market.\n\nSu's visit to Korea is highly unusual — it marks her first trip to the country since taking office in 2014. Historically, Su's Asia itinerary has centered on Japan, China, and Taiwan. Japan has been active ground for automotive semiconductor collaboration through AMD's acquisition of Xilinx; Taiwan is home to TSMC, the world's largest foundry; and China remains one of the world's largest markets. Korea's emergence on this circuit underscores how dramatically Samsung Memory's strategic importance has grown.\n\nIndustry sources indicate that the visit will cover discussions on long-term supply agreements spanning commodity DRAM and NAND flash in addition to broader memory categories — a signal that even hyperscale customers like AMD are facing serious memory procurement constraints. The agenda also includes HBM4 supply terms: Samsung's sixth-generation High Bandwidth Memory is expected to be incorporated into AMD's latest MI450 series AI accelerators.\n\nFoundry collaboration is attracting attention on par with the memory discussions. The two companies are reportedly negotiating a contract for AMD to allocate volume to Samsung's leading-edge foundry process. If realized, a formal foundry partnership between the two would be essentially unprecedented. Samsung did produce some AMD graphics chips on its 14nm process back in 2016, but that engagement was not considered meaningful in scale — unlike more recent wins such as the Tesla order.\n\nShould Samsung Foundry secure AMD chip orders, the earnings recovery trajectory would likely accelerate further. It could serve as a pivotal validation of Samsung Foundry's technical capabilities as a credible \"alternative to TSMC\" in the eyes of other major fabless customers. With Samsung Memory already entering an unprecedented upcycle, the question now is whether Foundry can pull forward its own path to profitability.\n\nA senior industry official noted, \"Samsung and AMD are understood to be coordinating plans to jointly disclose the details of this collaboration directly,\" adding that \"Samsung Semiconductor's strategic importance is set to be reaffirmed.\"\n\nSu is also reportedly pursuing meetings with Blue House officials or other government figures during her stay. This follows SoftBank Group Chairman Masayoshi Son's visit to Korea last December, during which he visited the Blue House directly to discuss AI cooperation.\n\n$AMD\n\n---\n\nhttps://t.co/89wLB3otbf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027557002635693828, "ret_signed_m1d": -0.027557002635693828, "alpha_spy_m1d": -0.017482661742640793, "alpha_c_m1d": -0.01328983315405008, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.014267169481643749, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027556919660915336, "ret_signed_p1d": 0.027556919660915336, "alpha_spy_p1d": 0.024926364790940125, "alpha_c_p1d": 0.022080682424307607, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.005476237236607728, "ret_p1w": -0.012718622983764694, "ret_signed_p1w": -0.012718622983764694, "alpha_spy_p1w": 0.00500880667066439, "alpha_c_p1w": -0.000759190183013625, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.011959432800751069, "ret_p1m": 0.09659467360003915, "ret_signed_p1m": 0.09659467360003915, "alpha_spy_p1m": 0.05574962843071529, "alpha_c_p1m": 0.029265522954581735, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.06732915064545741, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.431, -0.4325, -0.4394, -0.4172, -0.412, -0.4141, -0.4141, -0.383, -0.3667, -0.3646, -0.3646, -0.3646, -0.319, -0.2682, -0.2639, -0.2528, -0.2644, -0.2634, -0.2644, -0.2708, -0.2565, -0.2374, -0.2109, -0.2088, -0.2305, -0.2077, -0.1929, -0.194, -0.194, -0.1547, -0.1394, -0.1484, -0.1494, -0.203, -0.1123, -0.1039, -0.1558, -0.1595, -0.1182, -0.1214, -0.1108, -0.0535, -0.0397, -0.0397, -0.0397, -0.0397, 0.0069, 0.0074, 0.0228, 0.0599, 0.0784, 0.1553, 0.1473, 0.1473, 0.0339, -0.0874, 0.0154, -0.0026, -0.0806, -0.0042, 0.0069, -0.0042, -0.0276, 0.0, 0.0276, 0.1049, 0.0625, 0.0567, -0.0127, 0.0053, 0.0016, -0.0456, -0.0456, -0.0638, -0.1121, 0.0071, -0.0526, -0.0112, 0.0254, 0.0435, 0.1178, 0.0833, 0.0939, 0.0674, 0.0966, 0.1205, 0.155, 0.147, 0.1391, 0.163, 0.155, 0.1922, 0.1656, 0.1922, 0.1789, 0.2001, 0.1709, 0.1709, 0.2347, 0.2347, 0.4126, 0.4418, 0.4258, 0.5161, 0.4816, 0.5081, 0.5719, 0.4365, 0.4922, 0.463, 0.4657, 0.5905, 0.5533, 0.5533, 0.5878, 0.6303, 0.5905, 0.6834, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2031697141681238411", "ticker_idx": 0, "ts": "2026-03-11T11:42:00+00:00", "symbol": "Wistron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Preferred ODM ranking: Wistron, Foxconn, Quanta Computer, in that order", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's Positive ratings on Wistron, Foxconn, and Quanta on GB200/300 NVL72 rack shipment acceleration, doubling YoY in 2026.", "resolved_tickers": ["3231.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Wistron Corp listed on TWSE as 3231.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Morgan Stanley: GB200/300 NVL72 Rack Shipments Accelerate in February; Positive on Wistron, Foxconn, Quanta (Taiwan Economic Daily)\n\n• As AI server demand continues to expand, GB200/300 NVL72 rack shipments are rising rapidly. Morgan Stanley estimates total production of these racks reached approximately 6,500 units in February 2026, up roughly +6% MoM. Full-year 2026 shipments are projected at 70,000–80,000 units, more than doubling from approximately 29,000 units in 2025. Preferred ODM ranking: Wistron, Foxconn, Quanta Computer, in that order.\n\n• By vendor, Foxconn led February rack shipments at approximately 2,400 units, followed by Wistron at 1,400–1,500 units and Quanta Computer at 1,300–1,400 units. However, some figures include Wistron's L10 compute trays converted into rack-equivalent counts, meaning actual end-customer delivery volumes may be somewhat lower.\n\n• On the company outlook, Wistron has shipped a cumulative 2,400–2,500 units through December, putting it on track for 3,000+ units in Q1. Foxconn's March shipments are expected to exceed February levels, bringing Q1 total to approximately 8,500 units. Quanta saw February revenue decline MoM due to a drop in notebook shipments, but AI server shipments are normalizing, with Q1 rack shipments expected at approximately 4,500 units.\n\n$NVDA\n\n---\n\nhttps://t.co/Cea1WY5T4i", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03007518796992481, "ret_signed_m1d": -0.03007518796992481, "alpha_spy_m1d": -0.03133192831233056, "alpha_c_m1d": -0.025304191762548056, "bench_spy_m1d": 0.0012567403424057488, "bench_c_m1d": -0.004770996207376754, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003759398496240629, "ret_signed_p1d": -0.003759398496240629, "alpha_spy_p1d": 0.011425550373122872, "alpha_c_p1d": 0.014683905969668176, "bench_spy_p1d": -0.015184948869363502, "bench_c_p1d": -0.018443304465908805, "ret_p1w": -0.018796992481203034, "ret_signed_p1w": -0.018796992481203034, "alpha_spy_p1w": 0.003233698297962384, "alpha_c_p1w": -0.001208254878270143, "bench_spy_p1w": -0.02203069077916542, "bench_c_p1w": -0.01758873760293289, "ret_p1m": -0.003759398496240629, "ret_signed_p1m": -0.003759398496240629, "alpha_spy_p1m": -0.011798084525709407, "alpha_c_p1m": -0.016695832755896922, "bench_spy_p1m": 0.008038686029468778, "bench_c_p1m": 0.012936434259656293, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0789, 0.0639, 0.0451, 0.0376, 0.0489, 0.0827, 0.1015, 0.0902, 0.094, 0.094, 0.1015, 0.0977, 0.1053, 0.1316, 0.1316, 0.1654, 0.1654, 0.1617, 0.1917, 0.1353, 0.1203, 0.0977, 0.0789, 0.0977, 0.0789, 0.0902, 0.0414, 0.0226, 0.0038, 0.0188, 0.0263, 0.0263, 0.0113, 0.0113, -0.0038, -0.0188, -0.0376, -0.0338, -0.0301, -0.0414, -0.0301, -0.015, -0.0263, -0.0113, -0.0113, -0.0113, -0.0113, -0.0113, -0.0113, -0.0113, -0.0113, -0.0263, -0.015, 0.0376, 0.0226, 0.0226, 0.015, -0.0113, -0.0602, -0.0338, -0.0188, -0.0414, -0.0301, 0.0, -0.0038, 0.0075, 0.015, 0.0075, -0.0188, -0.0414, -0.0226, -0.0564, -0.0301, -0.0376, -0.0451, -0.0451, -0.0564, -0.0789, -0.0451, -0.0602, -0.0602, -0.0602, -0.0714, -0.015, -0.0038, 0.015, 0.0113, 0.0263, 0.0, 0.0075, 0.0263, 0.0564, 0.0865, 0.0827, 0.0489, 0.0639, 0.0714, 0.0564, 0.0564, 0.0301, 0.0301, 0.0564, 0.0677, 0.1015, 0.0977, 0.1015, 0.0451, 0.0639, 0.0564, 0.0376, 0.0188, 0.0188, -0.0038, -0.0038, 0.0526, 0.0865, 0.1203, 0.0977, 0.0902, 0.0865, 0.1917, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2031697141681238411", "ticker_idx": 1, "ts": "2026-03-11T11:42:00+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Positive on Wistron, Foxconn, Quanta", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's Positive ratings on Wistron, Foxconn, and Quanta on GB200/300 NVL72 rack shipment acceleration, doubling YoY in 2026.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Morgan Stanley: GB200/300 NVL72 Rack Shipments Accelerate in February; Positive on Wistron, Foxconn, Quanta (Taiwan Economic Daily)\n\n• As AI server demand continues to expand, GB200/300 NVL72 rack shipments are rising rapidly. Morgan Stanley estimates total production of these racks reached approximately 6,500 units in February 2026, up roughly +6% MoM. Full-year 2026 shipments are projected at 70,000–80,000 units, more than doubling from approximately 29,000 units in 2025. Preferred ODM ranking: Wistron, Foxconn, Quanta Computer, in that order.\n\n• By vendor, Foxconn led February rack shipments at approximately 2,400 units, followed by Wistron at 1,400–1,500 units and Quanta Computer at 1,300–1,400 units. However, some figures include Wistron's L10 compute trays converted into rack-equivalent counts, meaning actual end-customer delivery volumes may be somewhat lower.\n\n• On the company outlook, Wistron has shipped a cumulative 2,400–2,500 units through December, putting it on track for 3,000+ units in Q1. Foxconn's March shipments are expected to exceed February levels, bringing Q1 total to approximately 8,500 units. Quanta saw February revenue decline MoM due to a drop in notebook shipments, but AI server shipments are normalizing, with Q1 rack shipments expected at approximately 4,500 units.\n\n$NVDA\n\n---\n\nhttps://t.co/Cea1WY5T4i", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04337899543378998, "ret_signed_m1d": -0.04337899543378998, "alpha_spy_m1d": -0.04463573577619573, "alpha_c_m1d": -0.038607999226413225, "bench_spy_m1d": 0.0012567403424057488, "bench_c_m1d": -0.004770996207376754, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.020547945205479423, "ret_signed_p1d": -0.020547945205479423, "alpha_spy_p1d": -0.005362996336115922, "alpha_c_p1d": -0.002104640739570618, "bench_spy_p1d": -0.015184948869363502, "bench_c_p1d": -0.018443304465908805, "ret_p1w": -0.04109589041095896, "ret_signed_p1w": -0.04109589041095896, "alpha_spy_p1w": -0.01906519963179354, "alpha_c_p1w": -0.023507152808026066, "bench_spy_p1w": -0.02203069077916542, "bench_c_p1w": -0.01758873760293289, "ret_p1m": -0.08675799086757996, "ret_signed_p1m": -0.08675799086757996, "alpha_spy_p1m": -0.09479667689704874, "alpha_c_p1m": -0.09969442512723625, "bench_spy_p1m": 0.008038686029468778, "bench_c_p1m": 0.012936434259656293, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0365, 0.0114, -0.0046, -0.0114, -0.0137, 0.0114, 0.0228, 0.0228, 0.0228, 0.0228, 0.0297, 0.0548, 0.0411, 0.0525, 0.0525, 0.0594, 0.0708, 0.0776, 0.089, 0.0479, 0.0525, 0.0434, 0.0342, 0.0708, 0.0662, 0.0708, 0.0479, 0.0205, 0.0, 0.0205, 0.0114, 0.0228, 0.0297, 0.0297, 0.0228, 0.0068, -0.0205, -0.0114, 0.0, -0.016, -0.0183, -0.0023, 0.0091, 0.0365, 0.0365, 0.0365, 0.0365, 0.0365, 0.0365, 0.0365, 0.0365, 0.0434, 0.0594, 0.1256, 0.1096, 0.1096, 0.0913, 0.0457, -0.0091, 0.0228, 0.0183, -0.0388, -0.0434, 0.0, -0.0205, -0.0205, -0.0114, -0.032, -0.0411, -0.0639, -0.0731, -0.105, -0.1096, -0.0868, -0.0845, -0.089, -0.1142, -0.1438, -0.1005, -0.1187, -0.1187, -0.1187, -0.1233, -0.0799, -0.0868, -0.0845, -0.0868, -0.0525, -0.0525, -0.0548, -0.0594, -0.0525, -0.0365, 0.0091, 0.0274, 0.0114, 0.0411, 0.0297, 0.0274, 0.0023, 0.0023, 0.0388, 0.0936, 0.1507, 0.1575, 0.1416, 0.1507, 0.1416, 0.1461, 0.1164, 0.1347, 0.1347, 0.1187, 0.0959, 0.1301, 0.1416, 0.1918, 0.1826, 0.2055, 0.2009, 0.3196, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2031697141681238411", "ticker_idx": 2, "ts": "2026-03-11T11:42:00+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Positive on Wistron, Foxconn, Quanta", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's Positive ratings on Wistron, Foxconn, and Quanta on GB200/300 NVL72 rack shipment acceleration, doubling YoY in 2026.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Morgan Stanley: GB200/300 NVL72 Rack Shipments Accelerate in February; Positive on Wistron, Foxconn, Quanta (Taiwan Economic Daily)\n\n• As AI server demand continues to expand, GB200/300 NVL72 rack shipments are rising rapidly. Morgan Stanley estimates total production of these racks reached approximately 6,500 units in February 2026, up roughly +6% MoM. Full-year 2026 shipments are projected at 70,000–80,000 units, more than doubling from approximately 29,000 units in 2025. Preferred ODM ranking: Wistron, Foxconn, Quanta Computer, in that order.\n\n• By vendor, Foxconn led February rack shipments at approximately 2,400 units, followed by Wistron at 1,400–1,500 units and Quanta Computer at 1,300–1,400 units. However, some figures include Wistron's L10 compute trays converted into rack-equivalent counts, meaning actual end-customer delivery volumes may be somewhat lower.\n\n• On the company outlook, Wistron has shipped a cumulative 2,400–2,500 units through December, putting it on track for 3,000+ units in Q1. Foxconn's March shipments are expected to exceed February levels, bringing Q1 total to approximately 8,500 units. Quanta saw February revenue decline MoM due to a drop in notebook shipments, but AI server shipments are normalizing, with Q1 rack shipments expected at approximately 4,500 units.\n\n$NVDA\n\n---\n\nhttps://t.co/Cea1WY5T4i", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02249134948096887, "ret_signed_m1d": -0.02249134948096887, "alpha_spy_m1d": -0.02374808982337462, "alpha_c_m1d": -0.017720353273592115, "bench_spy_m1d": 0.0012567403424057488, "bench_c_m1d": -0.004770996207376754, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0017301038062284002, "ret_signed_p1d": -0.0017301038062284002, "alpha_spy_p1d": 0.013454845063135101, "alpha_c_p1d": 0.016713200659680405, "bench_spy_p1d": -0.015184948869363502, "bench_c_p1d": -0.018443304465908805, "ret_p1w": 0.01384083044982698, "ret_signed_p1w": 0.01384083044982698, "alpha_spy_p1w": 0.0358715212289924, "alpha_c_p1w": 0.03142956805275987, "bench_spy_p1w": -0.02203069077916542, "bench_c_p1w": -0.01758873760293289, "ret_p1m": 0.08304498269896188, "ret_signed_p1m": 0.08304498269896188, "alpha_spy_p1m": 0.0750062966694931, "alpha_c_p1m": 0.07010854843930558, "bench_spy_p1m": 0.008038686029468778, "bench_c_p1m": 0.012936434259656293, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0087, -0.0138, -0.0554, -0.083, -0.1003, -0.083, -0.0761, -0.0848, -0.0865, -0.0865, -0.09, -0.09, -0.09, -0.0588, -0.0588, -0.0415, -0.0381, -0.0242, -0.0242, -0.0554, -0.0208, -0.0294, -0.0346, -0.0104, -0.0208, 0.0017, -0.0242, -0.0398, -0.0484, -0.0208, -0.0311, -0.0069, -0.0052, -0.0069, -0.0208, -0.0311, -0.0554, -0.0433, -0.0381, -0.0554, -0.0329, -0.0173, -0.0138, -0.0104, -0.0104, -0.0104, -0.0104, -0.0104, -0.0104, -0.0104, -0.0104, -0.019, -0.0173, 0.0294, 0.0087, 0.0087, 0.0415, 0.019, -0.0277, 0.0017, 0.0, -0.0242, -0.0225, 0.0, -0.0017, 0.0, 0.0, 0.0, 0.0138, -0.0087, -0.019, -0.0311, -0.026, -0.0138, -0.0069, 0.0, -0.0277, -0.0363, 0.0052, -0.0156, -0.0156, -0.0156, 0.0, 0.064, 0.083, 0.1125, 0.1038, 0.1176, 0.0675, 0.1021, 0.1176, 0.1384, 0.1765, 0.1592, 0.1142, 0.1176, 0.1263, 0.1107, 0.1142, 0.0813, 0.0813, 0.1003, 0.1107, 0.199, 0.1903, 0.1782, 0.1886, 0.1765, 0.1782, 0.1574, 0.0588, 0.0381, 0.0035, 0.0035, 0.0657, 0.0934, 0.0952, 0.0952, 0.0796, 0.0675, 0.173, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2009455481392336999", "ticker_idx": 0, "ts": "2026-01-09T02:41:35+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SanDisk expects revenue of between US$2.55 billion and US$2.65 billion for the October to December 2025 period, with gross margin forecast to rise to between 41% and 43%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SNDK: 100% prepay long-term NAND contracts, surging AI demand, margins expanding from 29.9% to 41-43%, enterprise SSD prices up 25-40%.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: SanDisk seeks prepaid cash deals for long-term HAND supply\n\nSanDisk is approaching customers with proposals for long-term NAND flash supply contracts that require 100% cash prepayment, according to supply chain sources. Industry participants said the terms are without precedent, particularly as they require buyers to commit cash upfront at a time of rapidly rising prices and severe supply constraints.\n\nLocking in supply amid surging AI demand\nThe proposal aims to lock in supply for as long as three years as demand for artificial intelligence servers continues to surge and growth in NAND bit output remains slow. While some large cloud service providers are reported to be considering the terms, supply chain sources said the requirement for full advance cash payment has drawn strong pushback from across the memory industry.\n\nAccording to company statements, SanDisk expects the current supply-demand imbalance for its NAND products to persist through the end of 2026. The company has told investors that its manufacturing lines are operating at full utilization and that it is negotiating multi-quarter transactions with selected customers, including discussions that incorporate demand forecasts extending into 2027.\n\nMore recently, SanDisk has expanded its long-term contract proposals beyond cloud service providers to include customers in the PC, smartphone, and memory module sectors, sources said. The contracts would require buyers to commit to specific purchase volumes and provide full cash prepayments, with contract terms ranging from one to three years.\n\nIndustry participants said the key challenge for 2026 is no longer limited to contract price increases but stems from upstream suppliers conditioning shipment availability on cash-only payment terms. Such requirements are expected to place significant pressure on buyers' financial positions and cash flow, particularly as prices continue to climb.\n\nSources said severe NAND shortages have already triggered aggressive stockpiling across the market. Requiring full cash prepayment shifts future price risk and inventory responsibility entirely onto customers, while allowing suppliers to collect cash upfront and pursue investment or capacity expansion without bearing pricing or inventory risk.\n\nAI demand drives margin expansion\nRising AI server demand has continued to push NAND prices sharply higher. For its fiscal first quarter of 2026, which ended on October 3, 2025, SanDisk reported revenue of approximately US$2.308 billion, up 21% from the prior quarter, with a gross margin of 29.9%.\n\nCompany statements indicate SanDisk expects revenue of between US$2.55 billion and US$2.65 billion for the October to December 2025 period, with gross margin forecast to rise to between 41% and 43%.\n\nNAND flash prices began climbing rapidly in the fourth quarter of 2025. Sources said SanDisk issued multiple price increase notices during the year, with November 2025 contract prices rising by as much as 50%. Industry participants said the sharp increases prompted several memory module makers to temporarily suspend price quotations.\n\nPressure on downstream markets\nSome cloud service providers are said to have agreed to long-term contracts with prepayment terms, reflecting their stronger financial resources. Sources cautioned, however, that server operators face rapidly expanding storage capacity requirements, raising questions over how much idle cash they can realistically allocate upfront.\n\nMemory module makers face an even more difficult position, industry sources said. With the market firmly in a seller-driven phase, module suppliers have limited leverage to reject upstream terms. Providing hundreds of millions of dollars in upfront cash would be challenging, while borrowing to fund purchases could result in substantial interest costs.\n\nIndustry participants also warned that concentrating large sums with a single supplier could trigger reactions from other major NAND vendors, potentially prompting similar demands across the market. Locking into long-term contracts carries significant risk given uncertainty over future industry conditions, complicating capital allocation decisions and board-level risk management, they said.\n\nAs AI workloads increasingly shift from training to inference, demand for enterprise solid-state drives continues to rise sharply. Enterprise SSD contract prices increased between 5% and 8% in the third quarter of 2025, before jumping 25% to 30% in the fourth quarter, according to market participants. Industry sources said NAND wafer prices have risen even more sharply and expect enterprise SSD contract prices to climb a further 30% to 40% in the first quarter of 2026.\n\nWith suppliers prioritizing higher-margin enterprise products, capacity for consumer applications has been squeezed. Industry participants said higher NAND costs are forcing specification downgrades for PCs, smartphones, and other consumer devices. Retail demand is expected to weaken further, fueling concern over consumer market conditions in the second half of 2026.\n\n$SNDK\n\n---\n\nhttps://t.co/WE58YnB5Om", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.1135899809258063, "ret_signed_m1d": -0.1135899809258063, "alpha_spy_m1d": -0.10702008748652037, "alpha_c_m1d": -0.10052124887443203, "bench_spy_m1d": -0.006569893439285934, "bench_c_m1d": -0.013068732051374266, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.031424671409029825, "ret_signed_p1d": 0.031424671409029825, "alpha_spy_p1d": 0.029854217597258037, "alpha_c_p1d": 0.027045603959829467, "bench_spy_p1d": 0.0015704538117717881, "bench_c_p1d": 0.004379067449200358, "ret_p1w": 0.09594338015347503, "ret_signed_p1w": 0.09594338015347503, "alpha_spy_p1w": 0.09941566146734082, "alpha_c_p1w": 0.09956979701820035, "bench_spy_p1w": -0.003472281313865788, "bench_c_p1w": -0.0036264168647253125, "ret_p1m": 0.5457990481258508, "ret_signed_p1m": 0.5457990481258508, "alpha_spy_p1m": 0.5459718774055292, "alpha_c_p1m": 0.5649573506144481, "bench_spy_p1m": -0.00017282927967843253, "bench_c_p1m": -0.01915830248859729, "ret_p3m": 1.0691026121109202, "ret_signed_p3m": 1.0691026121109202, "alpha_spy_p3m": 1.092463046427373, "alpha_c_p3m": 1.0984137528448716, "bench_spy_p3m": -0.02336043431645274, "bench_c_p3m": -0.029311140733951357, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6627, -0.6177, -0.6177, -0.6286, -0.6077, -0.6044, -0.6106, -0.5574, -0.5067, -0.5324, -0.5349, -0.4585, -0.4811, -0.4718, -0.4515, -0.4845, -0.4264, -0.4497, -0.3655, -0.29, -0.2804, -0.2499, -0.3546, -0.3266, -0.2955, -0.351, -0.3483, -0.4808, -0.4694, -0.3986, -0.4158, -0.4302, -0.4302, -0.4084, -0.4431, -0.4559, -0.485, -0.4348, -0.3946, -0.4026, -0.4185, -0.383, -0.3598, -0.4537, -0.4651, -0.4454, -0.452, -0.4185, -0.3704, -0.3613, -0.3511, -0.3374, -0.3374, -0.3375, -0.3528, -0.3635, -0.371, -0.371, -0.2707, -0.2738, -0.0736, -0.0632, -0.1136, 0.0, 0.0314, 0.0329, 0.0276, 0.0843, 0.0959, 0.0959, 0.2006, 0.3282, 0.3339, 0.2555, 0.2474, 0.2756, 0.398, 0.4289, 0.5269, 0.7626, 0.8428, 0.5488, 0.5267, 0.5844, 0.5458, 0.4352, 0.588, 0.67, 0.6602, 0.6602, 0.5648, 0.5908, 0.6457, 0.7222, 0.766, 0.6918, 0.6756, 0.7273, 0.6835, 0.6403, 0.4981, 0.5873, 0.4986, 0.3972, 0.5599, 0.6398, 0.7367, 0.6396, 0.7531, 0.8644, 0.9082, 0.997, 1.0458, 0.8805, 0.8613, 0.8613, 0.7961, 0.5982, 0.6317, 0.5169, 0.6834, 0.8355, 0.859, 0.859, 0.92, 0.8834, 1.0691, 1.2564, 1.2569, 1.5238, 1.5025, 1.3627, 1.4363, 1.4403, 1.4192, 1.3939, 1.5942, 1.4706, 1.6229, 1.8356, 1.6559, 1.8198, 1.9054, 2.1451, 2.3276, 2.7262, 2.7359, 2.5504, 3.1396, 3.1005, 2.8473, 2.8346, 2.6637, 2.7297, 2.532, 2.6652, 2.6898, 3.0864, 2.918, 2.918, 3.2117, 3.2128, 3.3498, 3.4911, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2010696025468772587", "ticker_idx": 0, "ts": "2026-01-12T12:51:04+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "UBS: SK Hynix Significant profit uplift potential - PT to Won1m", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS Buy on SK Hynix with PT Won1m; 82% DRAM OPM by 4Q26 on unprecedented memory upcycle.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Delta Force might actually storm the Blue House at this rate. Word is SK Hynix's operating profit margin just overtook Nvidia's. https://t.co/ZPi33NW5jC", "reply_text": null, "parent_text": null, "quoted_text": "UBS: SK Hynix\n\nSignificant profit uplift potential - PT to Won1m\n\nWe forecast 82% DRAM OPM by 4Q26\nThe memory semis industry is in the midst of an unprecedented upcycle. By 4Q26, we now forecast SK Hynix DDR ASPs to be a whisker away from the 3Q18 peak at US$0.97 per Gb. With an", "parent_author": null, "ret_m1d": -0.006675542417608371, "ret_signed_m1d": -0.006675542417608371, "alpha_spy_m1d": -0.00510755106383487, "alpha_c_m1d": -0.0029894664745698796, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014686210041248593, "ret_signed_p1d": -0.014686210041248593, "alpha_spy_p1d": -0.012686734935378308, "alpha_c_p1d": -0.016913197982250994, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.020026710865375774, "ret_signed_p1w": 0.020026710865375774, "alpha_spy_p1w": 0.025061539025936996, "alpha_c_p1w": -0.004879885395099981, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.16955947975267738, "ret_signed_p1m": 0.16955947975267738, "alpha_spy_p1m": 0.17393259971475683, "alpha_c_p1m": 0.1334666983792583, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.33490196052199983, "ret_signed_p3m": 0.33490196052199983, "alpha_spy_p3m": 0.35416826910674815, "alpha_c_p3m": 0.23340707104061198, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4363, -0.3963, -0.3789, -0.3523, -0.3609, -0.3576, -0.3616, -0.3196, -0.2862, -0.3049, -0.2555, -0.2422, -0.2542, -0.1728, -0.2182, -0.2275, -0.2088, -0.2262, -0.1915, -0.1742, -0.1768, -0.1835, -0.2529, -0.1915, -0.2395, -0.2502, -0.2382, -0.3049, -0.3062, -0.3076, -0.3009, -0.2737, -0.2924, -0.2817, -0.255, -0.263, -0.2764, -0.2737, -0.2296, -0.2443, -0.2163, -0.2457, -0.2377, -0.2603, -0.2924, -0.2644, -0.263, -0.2697, -0.2256, -0.2203, -0.215, -0.215, -0.2003, -0.1455, -0.1308, -0.1308, -0.1308, -0.0961, -0.0708, -0.0307, -0.0093, 0.0093, -0.0067, 0.0, -0.0147, -0.0093, 0.0, 0.0093, 0.02, -0.008, -0.012, 0.008, 0.024, -0.0174, 0.0681, 0.1228, 0.1495, 0.2136, 0.1081, 0.2109, 0.2016, 0.1242, 0.1202, 0.1842, 0.1696, 0.1482, 0.1856, 0.1749, 0.1749, 0.1749, 0.1749, 0.1936, 0.267, 0.2697, 0.3418, 0.3591, 0.47, 0.4192, 0.4192, 0.256, 0.1356, 0.2587, 0.2359, 0.1182, 0.2546, 0.2774, 0.2439, 0.2172, 0.3028, 0.2974, 0.4125, 0.355, 0.3469, 0.248, 0.3189, 0.3309, 0.248, 0.248, 0.1677, 0.0794, 0.1998, 0.1102, 0.1717, 0.1851, 0.2252, 0.3817, 0.3349, 0.3737, 0.3911, 0.4753, 0.5195, 0.5449, 0.5088, 0.5596, 0.6372, 0.6359, 0.6385, 0.6345, 0.7281, 0.7389, 0.7295, 0.7201, 0.7201, 0.9355, 0.9355, 1.1415, 1.2124, 1.2552, 1.5146, 1.4545, 1.6431, 1.635, 1.4331, 1.4611, 1.3341, 1.3341, 1.5949, 1.5962, 1.5962, 1.7447, 2.0002, 2.0622, 2.1211, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2028633028185178216", "ticker_idx": 0, "ts": "2026-03-03T00:46:18+00:00", "symbol": "6965.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I bought Hamamatsu Photonics today.", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Disclosed fresh long position in Hamamatsu Photonics.", "resolved_tickers": ["6965.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@pre09576802 I bought Hamamatsu Photonics today.", "reply_text": null, "parent_text": "@jukan05 @jukan05 After nvidia investment in photon/laser companies.  what do you think about below japan photon companies\nHamamatsu photonics\nSantech holdings\nQD Laser", "quoted_text": null, "parent_author": "pre09576802", "ret_m1d": -0.03709756674781228, "ret_signed_m1d": -0.03709756674781228, "alpha_spy_m1d": -0.04599028869185606, "alpha_c_m1d": -0.051933883596349095, "bench_spy_m1d": 0.00889272194404378, "bench_c_m1d": 0.014836316848536812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08912745520683152, "ret_signed_p1d": -0.08912745520683152, "alpha_spy_p1d": -0.09618276355991617, "alpha_c_p1d": -0.10614554740895121, "bench_spy_p1d": 0.007055308353084655, "bench_c_p1d": 0.017018092202119695, "ret_p1w": -0.033597824520009256, "ret_signed_p1w": -0.033597824520009256, "alpha_spy_p1w": -0.028967632229841667, "alpha_c_p1w": -0.05003413959077019, "bench_spy_p1w": -0.00463019229016759, "bench_c_p1w": 0.016436315070760932, "ret_p1m": -0.12347558607940401, "ret_signed_p1m": -0.12347558607940401, "alpha_spy_p1m": -0.08922664653917212, "alpha_c_p1m": -0.10585916120876493, "bench_spy_p1m": -0.03424893954023189, "bench_c_p1m": -0.01761642487063908, "ret_p3m": 0.34918784787885304, "ret_signed_p3m": 0.34918784787885304, "alpha_spy_p3m": 0.23422027226894637, "alpha_c_p3m": -0.04177582882817288, "bench_spy_p3m": 0.11496757560990667, "bench_c_p3m": 0.3909636767070259, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2336, -0.2207, -0.1759, -0.1661, -0.1575, -0.1776, -0.179, -0.2004, -0.2165, -0.2172, -0.2324, -0.2298, -0.2149, -0.2133, -0.2095, -0.2123, -0.2149, -0.2142, -0.2249, -0.2249, -0.2249, -0.2249, -0.2032, -0.2044, -0.206, -0.2088, -0.2191, -0.2191, -0.1976, -0.1717, -0.1631, -0.1654, -0.1778, -0.1955, -0.1981, -0.1778, -0.1804, -0.2018, -0.1934, -0.2053, -0.2163, -0.2011, -0.2144, -0.1806, -0.1764, -0.1647, -0.2172, -0.2002, -0.1759, -0.1759, -0.2014, -0.2088, -0.1974, -0.1745, -0.1778, -0.1722, -0.1302, -0.1302, -0.0952, -0.0476, -0.0665, -0.0558, -0.0371, 0.0, -0.0891, -0.0579, -0.0098, -0.084, -0.0336, -0.0205, -0.0665, -0.0345, -0.0369, -0.0709, -0.0632, -0.1101, -0.1101, -0.1703, -0.136, -0.0826, -0.0667, -0.0945, -0.1353, -0.1675, -0.1235, -0.143, -0.1289, -0.1225, -0.1334, -0.0733, -0.0811, -0.0846, -0.0813, -0.0877, -0.1037, -0.0764, -0.0872, -0.0801, -0.0832, -0.0837, -0.0983, -0.0945, -0.0323, -0.0563, -0.0563, -0.0467, -0.0547, -0.0547, -0.0547, -0.0547, 0.0085, 0.0337, 0.0188, 0.0259, 0.0191, 0.0165, 0.2521, 0.3511, 0.2655, 0.241, 0.2599, 0.3716, 0.4013, 0.3819, 0.2724, 0.3303, 0.3492, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010930781565448429", "ticker_idx": 0, "ts": "2026-01-13T04:23:54+00:00", "symbol": "Zhen Ding Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Initiating coverage on Zhen Ding with a \"Buy\" rating and target price of NT$227", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF initiates Zhen Ding with Buy, PT NT$227 on Apple share gains, AI platform wins, and early CoWoP opportunity on NVIDIA Rubin; author shares/adopts the call.", "resolved_tickers": ["4935.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Zhen Ding Technology listed on TWSE as 4935.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics Communication Zhen Ding First Coverage: Undervalued Giant Embraces AI and CoWoP Opportunities\n\nInitiating coverage on Zhen Ding with a “Buy” rating and target price of NT$227: We believe Zhen Ding’s core Apple business will benefit from higher market share in foldable devices and stable iPhone volume growth. More importantly, we are bullish on Zhen Ding’s AI potential, with expected share gains across multiple AI platforms. Additionally, NVIDIA’s CoWoP may launch earlier than expected, and with its mSAP+HDI capabilities, Zhen Ding is positioned to be one of the leading participants. We forecast Zhen Ding’s 2026 and 2027 EPS at NT$15.13 and NT$22.54, respectively. Our NT$227 target price is based on 15x FY2026E P/E. Zhen Ding has underperformed the Taiwan Weighted Index by 16.7% over the past three months; we view its current risk/reward as attractive.\n\nCoWoP may launch ahead of schedule: We believe CoWoP could enter trial production as early as the Rubin platform, while the market expects late-2027/2028. The architecture places an mSAP PCB on top of two independent 7-layer HDI boards (using M9Q glass-based CCL), with the three boards connected via copper paste lamination technology. We see Zhen Ding as one of the leaders in CoWoP; its CoWoP solution has already been sent to OSATs for chip-level testing. We estimate CoWoP costs ~US$600 per GPU versus ~US$200 for a GB200 compute tray. Our scenario analysis suggests that, assuming 30% CoWoP adoption on the Rubin Ultra platform and 35% market share for Zhen Ding, CoWoP could add ~NT$1.8 to EPS by 2027.\n\nSteady AI server market share gains: We expect Zhen Ding to make solid progress in AI market share, driven by: 1) NVIDIA compute tray scheduled for 1Q26 launch, with switch tray verification underway; 2) AWS UBB starting in 4Q25; 3) potential Google cooperation supported by the Hon Hai Group; and 4) 1.6T optical module PCBs. Zhen Ding’s aggressive NT$30 billion annual capex plan for 2025–2026 also strongly supports share gains. Including potential CoWoP growth, we forecast server and optical module PCB revenue of NT$21.7 billion in 2026 and NT$51.9 billion in 2027.\n\nIC substrate upgrade cycle to continue: First, we believe CoWoP solutions will not cannibalize the ABF substrate business, as demand remains robust due to increasing chip size and layer count. Separately, we estimate ~75% of Zhen Ding’s ABF revenue comes from AI, with its Kaohsiung capacity largely locked in by AMD.\n\nDriven by the memory supercycle, BT substrate utilization has recovered to above 95%. Due to Low-CTE shortages, we have observed price increases: BT substrates up 5% monthly since 3Q25, and ABF substrates up 10–15% in 4Q25. With favorable pricing and strong utilization, we expect the IC substrate business to return to profitability in 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00618046959912677, "ret_signed_m1d": 0.00618046959912677, "alpha_spy_m1d": 0.004176988582839325, "alpha_c_m1d": 0.004064132490436911, "bench_spy_m1d": 0.002003481016287445, "bench_c_m1d": 0.0021163371086898586, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004944394540597807, "ret_signed_p1d": 0.004944394540597807, "alpha_spy_p1d": 0.00985963188221639, "alpha_c_p1d": 0.017096201755439355, "bench_spy_p1d": -0.0049152373416185835, "bench_c_p1d": -0.012151807214841548, "ret_p1w": 0.0074165446576557326, "ret_signed_p1w": 0.0074165446576557326, "alpha_spy_p1w": 0.030752850881229477, "alpha_c_p1w": 0.039093180946803474, "bench_spy_p1w": -0.023336306223573744, "bench_c_p1w": -0.03167663628914774, "ret_p1m": -0.04326328719388706, "ret_signed_p1m": -0.04326328719388706, "alpha_spy_p1m": -0.040654377521231355, "alpha_c_p1m": -0.019301044375102494, "bench_spy_p1m": -0.0026089096726557015, "bench_c_p1m": -0.023962242818784563, "ret_p3m": -0.10877630266722305, "ret_signed_p3m": -0.10877630266722305, "alpha_spy_p3m": -0.09082454464813827, "alpha_c_p3m": -0.08363516485363676, "bench_spy_p3m": -0.017951758019084774, "bench_c_p3m": -0.02514113781358629, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0667, 0.063, 0.0618, 0.0606, 0.0692, 0.0729, 0.0729, 0.1174, 0.1187, 0.1088, 0.047, 0.0346, 0.0272, 0.0037, -0.0049, 0.0087, 0.0074, 0.0247, 0.0185, 0.0457, 0.0556, 0.0346, 0.0371, 0.0124, 0.0049, 0.0074, 0.0025, -0.0025, 0.0049, 0.0655, 0.0581, 0.0532, 0.0445, 0.047, 0.0494, 0.0519, 0.0457, 0.0185, 0.0037, 0.0012, 0.0025, 0.0062, 0.0148, 0.0062, 0.0062, 0.0111, 0.0198, 0.0198, 0.0136, 0.0136, 0.0136, 0.0099, 0.0087, 0.0087, 0.0111, 0.0111, 0.0099, -0.0062, -0.0111, -0.0025, 0.0012, -0.0037, 0.0062, 0.0, 0.0049, 0.0037, 0.0136, 0.026, 0.0074, 0.0074, 0.0136, 0.0087, -0.0037, -0.0074, -0.0062, -0.0124, -0.0111, -0.0247, -0.0358, -0.0346, -0.0371, -0.0445, -0.0482, -0.0507, -0.0433, -0.0433, -0.0433, -0.0433, -0.0433, -0.0433, -0.0433, -0.0433, -0.0222, -0.0185, -0.0272, -0.0358, -0.0358, -0.0309, -0.0371, -0.068, -0.063, -0.0532, -0.0952, -0.1162, -0.115, -0.1088, -0.1051, -0.1174, -0.0989, -0.0927, -0.1112, -0.1063, -0.1261, -0.1137, -0.0989, -0.0433, -0.0433, -0.0791, -0.1112, -0.1051, -0.1248, -0.1248, -0.1248, -0.1273, -0.1075, -0.1014, -0.1088, -0.1199, -0.0989, -0.11, -0.0655, -0.0519, -0.0358, -0.0334, -0.0124, -0.0396, -0.0408, -0.0507, -0.0593, -0.063, -0.0729, -0.0729, -0.0667, -0.0593, -0.0717, -0.063, -0.0791, -0.089, -0.1088, -0.1088, -0.1088, -0.11, -0.1063, -0.1051, -0.1075, -0.0952, -0.0853, -0.0791, -0.0816, -0.0964, -0.0803, 0.0111, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2010847273073459373", "ticker_idx": 0, "ts": "2026-01-12T22:52:04+00:00", "symbol": "LITE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Lumentum (LITE Buy) … Both COHR and LITE continue to offer significant upside potential through 2027", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish LITE/COHR into 2027 on 800G/1.6T demand ramp; InP shortage near-term positive for AXTI and Sumitomo; Win Semi benefits from optical growth.", "resolved_tickers": ["LITE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "self_thread", "lang": "en", "tweet_text": "GF Overseas Electronics & Communication\n\nAI Network: Short-term Disturbances Do Not Alter the Long-term Trend\n\nRecent Supply Chain Concerns\nRecently, due to China’s slower approval of export licenses to Taiwan and tighter export controls to Japan, AXT Inc. (AXTI) lowered its 4Q25 revenue guidance. This has heightened market concerns over the supply chain for related suppliers including AXTI (AXTI), Sumitomo Electric (5802 JP), Lumentum (LITE Buy), and Coherent (COHR Hold).\n\nHowever, we believe inventories have already been built up in advance along the supply chain, and recent comments from hyperscalers continue to point to a constructive outlook for overseas optical module manufacturers through 2027. Therefore, we view the short-term supply chain impact as not altering the medium- to long-term trend. Both COHR and LITE continue to offer significant upside potential through 2027.\n\nStock Impact\nLITE and COHR have corrected recently, primarily due to concerns over optical utilization rates, upstream material export controls, and some profit-taking after LITE’s 35–40x P/E had priced in near-term expectations.\n\nLooking toward 2027, we remain bullish on robust growth in 800G and 1.6T demand. We are also positive on Google’s stance toward optical architectures, given its advantage in interconnect chip count relative to Nvidia, as well as the expected optical ramp-up from other hyperscalers. We believe upstream material supply concerns are already partially priced in.\n\nMeanwhile, given the ongoing shortage of InP substrates, we expect short-term price increases that will benefit AXTI and Sumitomo. As highlighted in LITE’s latest earnings call, the laser supply gap continues to widen. We believe laser manufacturers such as Sumitomo are expanding capacity through external foundries, including Win Semiconductors (3105 TT). Win Semiconductors is also expected to benefit from incremental optical communication growth.\n\nStrong Demand in 2027\nWe forecast total 800G/1.6T demand to reach 58 million / 64 million+ units in 2027, representing YoY growth of 40% / 213%+, primarily driven by:\n\n1. Ramp-up in ASIC demand from Nvidia, Google, AWS, and others;\n\n2. Continued increase in GPU/ASIC scale-out bandwidth, driving a higher proportion of GPUs/ASICs connected via optical modules;\n\n3. Google’s full transition to large-scale clusters in 2026 and adoption of optical interconnects for scale-up.\n\nAdditionally, as noted in our previous report, scale-up CPO represents a pure incremental opportunity for the optical interconnect supply chain, benefiting FAU, laser, shuffle, and module vendors—including Lumentum, Sumitomo, and Browave.\n\n$AXTI $LITE\n\n---\n\n@aleabitoreddit", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03398363632118917, "ret_signed_m1d": 0.03398363632118917, "alpha_spy_m1d": 0.03555162767496267, "alpha_c_m1d": 0.03555162767496267, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0015679913537735013, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06314176527315607, "ret_signed_p1d": 0.06314176527315607, "alpha_spy_p1d": 0.06514124037902636, "alpha_c_p1d": 0.06514124037902636, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": -0.0019994751058702853, "ret_p1w": -0.045958735226985015, "ret_signed_p1w": -0.045958735226985015, "alpha_spy_p1w": -0.04092390706642379, "alpha_c_p1w": -0.04092390706642379, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": -0.005034828160561222, "ret_p1m": 0.6510136609421326, "ret_signed_p1m": 0.6510136609421326, "alpha_spy_p1m": 0.6553867809042121, "alpha_c_p1m": 0.6553867809042121, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": -0.0043731199620794525, "ret_p3m": 1.6308000639318445, "ret_signed_p3m": 1.6308000639318445, "alpha_spy_p3m": 1.650066372516593, "alpha_c_p3m": 1.650066372516593, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": -0.019266308584748315, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5393, -0.5197, -0.5152, -0.5263, -0.5206, -0.5349, -0.5042, -0.4724, -0.4298, -0.4163, -0.3695, -0.4112, -0.4069, -0.4128, -0.4458, -0.3152, -0.2948, -0.2935, -0.2353, -0.2572, -0.2532, -0.3325, -0.3169, -0.2878, -0.272, -0.2088, -0.3137, -0.248, -0.1192, -0.143, -0.0929, -0.0929, -0.0433, -0.0646, -0.109, -0.1085, -0.0354, -0.0249, 0.0079, 0.0602, 0.0769, 0.0948, -0.0457, -0.0152, -0.0698, -0.0577, -0.0081, 0.0929, 0.1471, 0.1399, 0.1649, 0.1649, 0.1498, 0.0963, 0.0921, 0.0845, 0.0845, 0.1361, 0.0505, 0.1693, 0.156, 0.0247, 0.034, 0.0, 0.0631, -0.0243, 0.01, -0.046, -0.046, 0.0499, 0.0664, 0.043, -0.002, -0.0218, 0.0906, 0.1328, 0.1223, 0.1529, 0.2458, 0.2802, 0.3698, 0.4842, 0.6241, 0.6981, 0.651, 0.6892, 0.7167, 0.6558, 0.6558, 0.7666, 0.7485, 0.8702, 0.9648, 0.9853, 1.0251, 1.1284, 0.9919, 1.0623, 1.3046, 1.0432, 1.0031, 0.9149, 0.6431, 0.8851, 0.9772, 0.9772, 0.8127, 0.8316, 0.8385, 0.9112, 1.062, 1.2718, 1.0783, 1.1448, 1.3597, 1.2867, 1.0267, 1.0676, 0.9266, 1.0677, 1.2498, 1.4329, 1.4329, 1.2723, 1.4002, 1.6364, 1.6308, 1.6401, 1.5633, 1.5092, 1.4245, 1.6222, 1.6306, 1.6337, 1.4625, 1.5704, 1.4918, 1.5941, 1.5294, 1.3284, 1.5254, 1.6549, 1.795, 1.8722, 1.9263, 1.7784, 1.6262, 1.6593, 2.0985, 1.9199, 2.0317, 1.9476, 1.8561, 1.6039, 1.6189, 1.5541, 1.8378, 1.7861, 1.7861, 1.6799, 1.6549, 1.5322, 1.5156, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2010847273073459373", "ticker_idx": 1, "ts": "2026-01-12T22:52:04+00:00", "symbol": "COHR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Both COHR and LITE continue to offer significant upside potential through 2027", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish LITE/COHR into 2027 on 800G/1.6T demand ramp; InP shortage near-term positive for AXTI and Sumitomo; Win Semi benefits from optical growth.", "resolved_tickers": ["COHR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Scientific & Technical Instruments'", "bench_c_industry": "Scientific & Technical Instruments", "unit_kind": "self_thread", "lang": "en", "tweet_text": "GF Overseas Electronics & Communication\n\nAI Network: Short-term Disturbances Do Not Alter the Long-term Trend\n\nRecent Supply Chain Concerns\nRecently, due to China’s slower approval of export licenses to Taiwan and tighter export controls to Japan, AXT Inc. (AXTI) lowered its 4Q25 revenue guidance. This has heightened market concerns over the supply chain for related suppliers including AXTI (AXTI), Sumitomo Electric (5802 JP), Lumentum (LITE Buy), and Coherent (COHR Hold).\n\nHowever, we believe inventories have already been built up in advance along the supply chain, and recent comments from hyperscalers continue to point to a constructive outlook for overseas optical module manufacturers through 2027. Therefore, we view the short-term supply chain impact as not altering the medium- to long-term trend. Both COHR and LITE continue to offer significant upside potential through 2027.\n\nStock Impact\nLITE and COHR have corrected recently, primarily due to concerns over optical utilization rates, upstream material export controls, and some profit-taking after LITE’s 35–40x P/E had priced in near-term expectations.\n\nLooking toward 2027, we remain bullish on robust growth in 800G and 1.6T demand. We are also positive on Google’s stance toward optical architectures, given its advantage in interconnect chip count relative to Nvidia, as well as the expected optical ramp-up from other hyperscalers. We believe upstream material supply concerns are already partially priced in.\n\nMeanwhile, given the ongoing shortage of InP substrates, we expect short-term price increases that will benefit AXTI and Sumitomo. As highlighted in LITE’s latest earnings call, the laser supply gap continues to widen. We believe laser manufacturers such as Sumitomo are expanding capacity through external foundries, including Win Semiconductors (3105 TT). Win Semiconductors is also expected to benefit from incremental optical communication growth.\n\nStrong Demand in 2027\nWe forecast total 800G/1.6T demand to reach 58 million / 64 million+ units in 2027, representing YoY growth of 40% / 213%+, primarily driven by:\n\n1. Ramp-up in ASIC demand from Nvidia, Google, AWS, and others;\n\n2. Continued increase in GPU/ASIC scale-out bandwidth, driving a higher proportion of GPUs/ASICs connected via optical modules;\n\n3. Google’s full transition to large-scale clusters in 2026 and adoption of optical interconnects for scale-up.\n\nAdditionally, as noted in our previous report, scale-up CPO represents a pure incremental opportunity for the optical interconnect supply chain, benefiting FAU, laser, shuffle, and module vendors—including Lumentum, Sumitomo, and Browave.\n\n$AXTI $LITE\n\n---\n\n@aleabitoreddit", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0384490517269509, "ret_signed_m1d": -0.0384490517269509, "alpha_spy_m1d": -0.0368810603731774, "alpha_c_m1d": -0.03408907690158769, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.004359974825363211, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026190767336335075, "ret_signed_p1d": 0.026190767336335075, "alpha_spy_p1d": 0.02819024244220536, "alpha_c_p1d": 0.028302635021075107, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": -0.002111867684740032, "ret_p1w": 0.031644890604415554, "ret_signed_p1w": 0.031644890604415554, "alpha_spy_p1w": 0.036679718764976776, "alpha_c_p1w": 0.03961547120826758, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": -0.007970580603852029, "ret_p1m": 0.2332325529196051, "ret_signed_p1m": 0.2332325529196051, "alpha_spy_p1m": 0.23760567288168455, "alpha_c_p1m": 0.26211729506848125, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": -0.028884742148876152, "ret_p3m": 0.5345611009136817, "ret_signed_p3m": 0.5345611009136817, "alpha_spy_p3m": 0.55382740949843, "alpha_c_p3m": 0.5655124363673825, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": -0.030951335453700768, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4038, -0.3738, -0.3717, -0.3509, -0.3477, -0.377, -0.3438, -0.3015, -0.271, -0.2751, -0.2545, -0.2833, -0.2874, -0.2872, -0.305, -0.273, -0.1398, -0.1656, -0.0997, -0.1467, -0.154, -0.2441, -0.2476, -0.249, -0.254, -0.2281, -0.2677, -0.2466, -0.1802, -0.1962, -0.1684, -0.1684, -0.113, -0.1171, -0.1096, -0.0768, -0.0423, -0.0183, 0.0037, 0.0408, 0.0663, 0.0719, -0.0369, -0.0363, -0.0539, -0.0796, -0.0511, 0.0035, 0.0313, 0.0361, 0.0334, 0.0334, 0.0353, 0.0207, 0.0088, -0.0033, -0.0033, 0.0494, 0.0064, 0.0482, 0.0348, -0.065, -0.0384, 0.0, 0.0262, -0.0058, 0.0582, 0.0316, 0.0316, 0.0447, 0.0879, 0.0947, 0.0635, 0.0679, 0.1556, 0.1942, 0.1657, 0.1458, 0.2012, 0.2376, 0.1394, 0.1299, 0.2295, 0.3093, 0.2332, 0.208, 0.167, 0.1731, 0.1731, 0.1878, 0.209, 0.2554, 0.3402, 0.344, 0.3763, 0.4467, 0.3508, 0.3983, 0.6142, 0.5164, 0.4843, 0.3709, 0.2729, 0.3626, 0.4075, 0.3577, 0.3029, 0.3109, 0.3358, 0.3274, 0.389, 0.4881, 0.3696, 0.3773, 0.4706, 0.4691, 0.3138, 0.3148, 0.1861, 0.2864, 0.3382, 0.3941, 0.3941, 0.3674, 0.3776, 0.5217, 0.5346, 0.6605, 0.6629, 0.6925, 0.6643, 0.7712, 0.8632, 0.8766, 0.8565, 0.8926, 0.8235, 0.8149, 0.7363, 0.6415, 0.6467, 0.7265, 0.7793, 0.7815, 0.813, 0.8613, 0.7237, 0.8105, 1.0504, 1.0197, 1.1801, 1.1867, 1.0653, 0.9593, 0.9097, 0.936, 1.0413, 1.0389, 1.0389, 1.0593, 1.053, 1.0356, 0.952, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2010847273073459373", "ticker_idx": 2, "ts": "2026-01-12T22:52:04+00:00", "symbol": "AXTI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect short-term price increases that will benefit AXTI and Sumitomo", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish LITE/COHR into 2027 on 800G/1.6T demand ramp; InP shortage near-term positive for AXTI and Sumitomo; Win Semi benefits from optical growth.", "resolved_tickers": ["AXTI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "GF Overseas Electronics & Communication\n\nAI Network: Short-term Disturbances Do Not Alter the Long-term Trend\n\nRecent Supply Chain Concerns\nRecently, due to China’s slower approval of export licenses to Taiwan and tighter export controls to Japan, AXT Inc. (AXTI) lowered its 4Q25 revenue guidance. This has heightened market concerns over the supply chain for related suppliers including AXTI (AXTI), Sumitomo Electric (5802 JP), Lumentum (LITE Buy), and Coherent (COHR Hold).\n\nHowever, we believe inventories have already been built up in advance along the supply chain, and recent comments from hyperscalers continue to point to a constructive outlook for overseas optical module manufacturers through 2027. Therefore, we view the short-term supply chain impact as not altering the medium- to long-term trend. Both COHR and LITE continue to offer significant upside potential through 2027.\n\nStock Impact\nLITE and COHR have corrected recently, primarily due to concerns over optical utilization rates, upstream material export controls, and some profit-taking after LITE’s 35–40x P/E had priced in near-term expectations.\n\nLooking toward 2027, we remain bullish on robust growth in 800G and 1.6T demand. We are also positive on Google’s stance toward optical architectures, given its advantage in interconnect chip count relative to Nvidia, as well as the expected optical ramp-up from other hyperscalers. We believe upstream material supply concerns are already partially priced in.\n\nMeanwhile, given the ongoing shortage of InP substrates, we expect short-term price increases that will benefit AXTI and Sumitomo. As highlighted in LITE’s latest earnings call, the laser supply gap continues to widen. We believe laser manufacturers such as Sumitomo are expanding capacity through external foundries, including Win Semiconductors (3105 TT). Win Semiconductors is also expected to benefit from incremental optical communication growth.\n\nStrong Demand in 2027\nWe forecast total 800G/1.6T demand to reach 58 million / 64 million+ units in 2027, representing YoY growth of 40% / 213%+, primarily driven by:\n\n1. Ramp-up in ASIC demand from Nvidia, Google, AWS, and others;\n\n2. Continued increase in GPU/ASIC scale-out bandwidth, driving a higher proportion of GPUs/ASICs connected via optical modules;\n\n3. Google’s full transition to large-scale clusters in 2026 and adoption of optical interconnects for scale-up.\n\nAdditionally, as noted in our previous report, scale-up CPO represents a pure incremental opportunity for the optical interconnect supply chain, benefiting FAU, laser, shuffle, and module vendors—including Lumentum, Sumitomo, and Browave.\n\n$AXTI $LITE\n\n---\n\n@aleabitoreddit", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.033723021929792596, "ret_signed_m1d": 0.033723021929792596, "alpha_spy_m1d": 0.0352910132835661, "alpha_c_m1d": 0.03740909787283109, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03282372076209894, "ret_signed_p1d": -0.03282372076209894, "alpha_spy_p1d": -0.030824245656228655, "alpha_c_p1d": -0.03505070870310134, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": -0.006744587233542543, "ret_signed_p1w": -0.006744587233542543, "alpha_spy_p1w": -0.0017097590729813206, "alpha_c_p1w": -0.0316511834940183, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.09487412913968196, "ret_signed_p1m": 0.09487412913968196, "alpha_spy_p1m": 0.09924724910176141, "alpha_c_p1m": 0.05878134776626287, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 1.8381294775845798, "ret_signed_p3m": 1.8381294775845798, "alpha_spy_p3m": 1.8573957861693282, "alpha_c_p3m": 1.736634588103192, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.7936, -0.7783, -0.7972, -0.7792, -0.7675, -0.7779, -0.7671, -0.7275, -0.7217, -0.7032, -0.6821, -0.6709, -0.6425, -0.6196, -0.616, -0.5908, -0.6012, -0.5746, -0.5306, -0.5252, -0.5063, -0.5306, -0.5306, -0.5391, -0.5549, -0.5454, -0.5751, -0.5985, -0.58, -0.585, -0.5301, -0.5301, -0.5189, -0.5148, -0.4712, -0.4838, -0.4559, -0.4793, -0.4285, -0.3273, -0.3026, -0.2635, -0.3341, -0.3332, -0.4155, -0.4442, -0.3701, -0.3413, -0.3156, -0.3417, -0.3251, -0.3251, -0.3089, -0.344, -0.2896, -0.2648, -0.2648, -0.2464, -0.2504, -0.0931, 0.0841, 0.1614, 0.0337, 0.0, -0.0328, -0.0063, 0.1565, -0.0067, -0.0067, -0.0373, -0.1106, -0.1942, -0.2176, -0.1996, -0.2266, -0.2433, -0.2635, -0.1664, -0.0585, -0.1124, -0.1569, -0.0814, 0.0818, 0.2487, 0.0949, 0.205, 0.1147, 0.0899, 0.0899, 0.0436, 0.0585, 0.0706, 0.3345, 0.2783, 0.5773, 0.8422, 0.6691, 0.7041, 1.0827, 0.8777, 0.7594, 0.7446, 0.4555, 0.7338, 0.9919, 1.1295, 1.1012, 1.1969, 1.1758, 0.9946, 1.1924, 1.612, 1.4388, 1.8975, 2.0773, 2.0283, 1.6308, 1.7262, 1.371, 1.5621, 1.1196, 1.3759, 1.3759, 0.888, 1.0441, 1.3912, 1.8381, 1.8858, 1.9879, 2.0261, 1.8296, 2.6772, 2.7122, 2.5414, 2.371, 2.9092, 2.3844, 2.4245, 2.1542, 2.0895, 2.1956, 2.5621, 3.3165, 3.7662, 3.8359, 3.7136, 3.875, 4.232, 4.6569, 4.5261, 4.4829, 4.17, 4.5656, 3.7608, 4.0755, 3.7037, 4.4415, 5.3323, 5.3323, 4.9622, 4.5202, 4.2023, 3.6385, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2010847273073459373", "ticker_idx": 3, "ts": "2026-01-12T22:52:04+00:00", "symbol": "Sumitomo Electric", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect short-term price increases that will benefit AXTI and Sumitomo", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish LITE/COHR into 2027 on 800G/1.6T demand ramp; InP shortage near-term positive for AXTI and Sumitomo; Win Semi benefits from optical growth.", "resolved_tickers": ["5802.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Sumitomo Electric Industries listed on TSE as 5802.T", "bench_c": "XLY", "bench_c_reason": "industry='Auto Parts'", "bench_c_industry": "Auto Parts", "unit_kind": "self_thread", "lang": "en", "tweet_text": "GF Overseas Electronics & Communication\n\nAI Network: Short-term Disturbances Do Not Alter the Long-term Trend\n\nRecent Supply Chain Concerns\nRecently, due to China’s slower approval of export licenses to Taiwan and tighter export controls to Japan, AXT Inc. (AXTI) lowered its 4Q25 revenue guidance. This has heightened market concerns over the supply chain for related suppliers including AXTI (AXTI), Sumitomo Electric (5802 JP), Lumentum (LITE Buy), and Coherent (COHR Hold).\n\nHowever, we believe inventories have already been built up in advance along the supply chain, and recent comments from hyperscalers continue to point to a constructive outlook for overseas optical module manufacturers through 2027. Therefore, we view the short-term supply chain impact as not altering the medium- to long-term trend. Both COHR and LITE continue to offer significant upside potential through 2027.\n\nStock Impact\nLITE and COHR have corrected recently, primarily due to concerns over optical utilization rates, upstream material export controls, and some profit-taking after LITE’s 35–40x P/E had priced in near-term expectations.\n\nLooking toward 2027, we remain bullish on robust growth in 800G and 1.6T demand. We are also positive on Google’s stance toward optical architectures, given its advantage in interconnect chip count relative to Nvidia, as well as the expected optical ramp-up from other hyperscalers. We believe upstream material supply concerns are already partially priced in.\n\nMeanwhile, given the ongoing shortage of InP substrates, we expect short-term price increases that will benefit AXTI and Sumitomo. As highlighted in LITE’s latest earnings call, the laser supply gap continues to widen. We believe laser manufacturers such as Sumitomo are expanding capacity through external foundries, including Win Semiconductors (3105 TT). Win Semiconductors is also expected to benefit from incremental optical communication growth.\n\nStrong Demand in 2027\nWe forecast total 800G/1.6T demand to reach 58 million / 64 million+ units in 2027, representing YoY growth of 40% / 213%+, primarily driven by:\n\n1. Ramp-up in ASIC demand from Nvidia, Google, AWS, and others;\n\n2. Continued increase in GPU/ASIC scale-out bandwidth, driving a higher proportion of GPUs/ASICs connected via optical modules;\n\n3. Google’s full transition to large-scale clusters in 2026 and adoption of optical interconnects for scale-up.\n\nAdditionally, as noted in our previous report, scale-up CPO represents a pure incremental opportunity for the optical interconnect supply chain, benefiting FAU, laser, shuffle, and module vendors—including Lumentum, Sumitomo, and Browave.\n\n$AXTI $LITE\n\n---\n\n@aleabitoreddit", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0015679913537735013, "alpha_c_m1d": 0.0008833299648755011, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0008833299648755011, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027760137339330138, "ret_signed_p1d": 0.027760137339330138, "alpha_spy_p1d": 0.029759612445200423, "alpha_c_p1d": 0.03024964457283741, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": -0.002489507233507271, "ret_p1w": 0.07074869153369368, "ret_signed_p1w": 0.07074869153369368, "alpha_spy_p1w": 0.0757835196942549, "alpha_c_p1w": 0.08857706275624244, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": -0.017828371222548767, "ret_p1m": 0.384517780113129, "ret_signed_p1m": 0.384517780113129, "alpha_spy_p1m": 0.38889090007520843, "alpha_c_p1m": 0.4342286043555146, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": -0.0497108242423856, "ret_p3m": 0.6696492908543743, "ret_signed_p3m": 0.6696492908543743, "alpha_spy_p3m": 0.6889155994391226, "alpha_c_p3m": 0.7624337472921111, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": -0.0927844564377368, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2962, -0.2792, -0.2928, -0.274, -0.2808, -0.2871, -0.2814, -0.2687, -0.2433, -0.2452, -0.2078, -0.1594, -0.1037, -0.1037, -0.0384, -0.0638, -0.0154, -0.0712, -0.0435, -0.0704, -0.0554, 0.0113, -0.0011, 0.0363, -0.0576, -0.0574, 0.0133, -0.0961, -0.0961, -0.0384, -0.0401, -0.0341, -0.0259, 0.0041, 0.0414, 0.0849, 0.066, 0.0589, 0.101, 0.1277, 0.1003, 0.0582, 0.0563, 0.052, -0.0098, 0.0027, -0.0355, -0.0002, 0.0725, 0.0482, 0.0314, 0.0374, -0.0078, 0.0051, 0.0033, 0.0033, 0.0033, 0.0033, 0.0644, 0.0354, 0.0544, 0.0082, 0.0, 0.0, 0.0278, 0.0278, 0.0654, 0.0831, 0.0707, 0.0465, 0.0612, 0.0642, 0.0668, 0.0385, 0.0595, 0.112, 0.0882, 0.0666, 0.0793, 0.2143, 0.2425, 0.175, 0.1961, 0.3273, 0.3845, 0.3845, 0.3912, 0.3631, 0.3726, 0.3723, 0.4001, 0.4784, 0.5463, 0.5463, 0.6482, 0.6997, 0.6244, 0.6458, 0.7235, 0.6378, 0.5776, 0.5823, 0.57, 0.4475, 0.5753, 0.6997, 0.6442, 0.6482, 0.6101, 0.5109, 0.5871, 0.578, 0.578, 0.4616, 0.5073, 0.6196, 0.5681, 0.5038, 0.4435, 0.3441, 0.4866, 0.4602, 0.5077, 0.5304, 0.4741, 0.6544, 0.6696, 0.6969, 0.592, 0.6159, 0.5319, 0.6263, 0.6, 0.6, 0.652, 0.7162, 0.6368, 0.6063, 0.6135, 0.6344, 0.6344, 0.6344, 0.5879, 0.5879, 0.5879, 0.5879, 0.8196, 0.8132, 0.798, 0.8685, 1.049, 0.9415, 0.7611, 0.8252, 0.7386, 0.6881, 0.721, 0.8958, 0.9672, 0.915, 0.9271, 0.9175, 1.0185, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2037317421389037723", "ticker_idx": 0, "ts": "2026-03-26T23:54:59+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TurboQuant will narrow the gap between GPU compute capability and memory bandwidth, increasing the number of tokens that can be processed per unit time on the same hardware. This lowers cost per token, encouraging broader AI usage, drawing in more services and users, and ultimately leading to an increase — not a decrease — in KV Cache consumption", "tweet_type": "analysis", "timeline": "unspecified", "summary": "KIS argues TurboQuant increases HBM demand (not decreases it); memory semis oversold on misread — bullish MU and SNDK.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[KIS — Chae Min-sook / Kim Yeon-jun] Semiconductor Industry Note: After TurboQuant and DeepSeek, the Conclusion is Clear\n\n● TurboQuant's Opening Shot\n\nOn March 25 (local time), Google officially published the TurboQuant algorithm via its official blog, claiming it can compress KV Cache by up to 6x with no performance degradation. The market interpreted this as a reduction in memory requirements, causing memory semiconductor stocks to fall sharply. However, this reflects a misreading stemming from a confusion between the roles of memory capacity and memory bandwidth. The bottleneck in AI inference is not a shortage of memory capacity, but rather a problem determined by memory access speed and data movement efficiency. TurboQuant should instead be understood as a technology that partially alleviates this bottleneck and improves GPU efficiency, enabling more tokens to be processed per unit time with the same GPU resources.\n\n● The Structure of AI Inference: Prefill and Decode\n\nLLM inference consists of two stages: Prefill and Decode. The Prefill stage is a compute-intensive task where GPU computational power limits performance, while the Decode stage is a memory-intensive task where the speed of data movement determines performance. In Decode, the existing KV cache must be repeatedly referenced with each new token generated, making it structurally sensitive to memory bandwidth and access latency. Since the Decode stage largely determines the response speed that users experience, it sits at the heart of AI inference optimization.\n\n● How TurboQuant Actually Works\n\nWhen TurboQuant claims to compress KV Cache by up to 6x, this is less about reducing the required memory capacity itself and more about significantly lowering the data size occupied by the KV Cache and the resulting memory access burden. This means the volume of data that must be handled within the same HBM bandwidth is reduced. As a result, memory access latency is alleviated and the time GPUs spend waiting for data can decrease. Because KV Cache accesses occur repeatedly during the Decode stage of LLM inference, reducing data size has a direct impact on relieving the memory bottleneck. The proportion of time GPUs spend waiting for memory responses diminishes, and compute resources are utilized more efficiently. This improves actual GPU utilization rates and increases the number of tokens that can be processed per unit time (throughput) within the same hardware environment — which can also be interpreted as a reduction in cost per token.\n\n● The Real Bottleneck in AI Inference: Bandwidth, Not Capacity\n\nIt appears the market interpreted TurboQuant as reducing memory capacity usage and therefore decreasing HBM demand. However, the core bottleneck in AI inference is not a shortage of memory capacity but rather the speed at which data is read from memory — i.e., memory bandwidth and access latency. Because GPU compute cores process data far faster than HBM can supply it, GPUs sit in a wait state while expecting data from memory. Industry research indicates that in the Decode stage, more than 50% of attention computation cycles are spent in a wait state due to memory access latency — meaning GPUs waste more than half their theoretical performance waiting for memory responses. According to research published by Google DeepMind in January 2026, Nvidia GPU 64-bit FLOPS grew approximately 80x between 2012 and 2022, while memory bandwidth grew only about 17x over the same period; this gap is expected to continue widening going forward (arXiv:2601.05047, forthcoming in IEEE Computer). TurboQuant will narrow the gap between GPU compute capability and memory bandwidth, increasing the number of tokens that can be processed per unit time on the same hardware. This lowers cost per token, encouraging broader AI usage, drawing in more services and users, and ultimately leading to an increase — not a decrease — in KV Cache consumption.\n\n● A Bottleneck TurboQuant Cannot Solve: Inter-Chip Communication Latency\n\nThere is another bottleneck in AI inference that TurboQuant does nothing to address. Large models exceed the memory capacity of a single GPU and are therefore distributed across multiple GPUs. During the Decode stage, because the model is spread across multiple GPUs, intermediate computation results must be exchanged between GPUs with each token generated. These data transfers are small in size but occur at very high frequency, which can introduce inter-chip communication latency. To reduce this latency, a single GPU must be able to handle more parameters and KV Cache — which ultimately demands more HBM capacity per GPU. In the current AI environment, where model sizes and context lengths continue to expand, this need for greater HBM capacity will only intensify. The fact that Nvidia has consistently increased HBM capacity with each successive GPU generation is likely not unrelated to this dynamic.\n\n● TurboQuant's Limitations: Still in Early Validation\n\nWhile TurboQuant carries significant technical implications, there are also limitations in its scope of application and validation that deserve consideration. The published performance benchmarks are confined to relatively simple test environments centered on single-query information retrieval tasks such as LongBench and Needle-In-A-Haystack. Experiments were also conducted primarily on comparatively small models with around 8 billion parameters; whether the same effects can be reproduced in the large-scale models of hundreds of billions of parameters used in real-world industry settings has yet to be verified. More importantly, no validation has been conducted in the rapidly expanding Agentic AI environments. In these environments, models execute repeated multi-step judgments across longer contexts and more complex KV Cache structures, giving rise to memory usage patterns that differ entirely from those seen in single-response benchmarks.\n\n$MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07491706048585134, "ret_signed_m1d": 0.07491706048585134, "alpha_spy_m1d": 0.05673358224184799, "alpha_c_m1d": 0.027180496966794143, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.047736563519057196, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0049513769084506, "ret_signed_p1d": 0.0049513769084506, "alpha_spy_p1d": 0.022003260431778582, "alpha_c_p1d": 0.022255222096086325, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.017303845187635725, "ret_p1w": 0.030759724747144723, "ret_signed_p1w": 0.030759724747144723, "alpha_spy_p1w": 0.014110900881849853, "alpha_c_p1w": 0.000615803161390982, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.03014392158575374, "ret_p1m": 0.39798761846569497, "ret_signed_p1m": 0.39798761846569497, "alpha_spy_p1m": 0.29125836716793274, "alpha_c_p1m": 0.06819030906734014, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.32979730939835483, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1719, -0.1768, -0.1971, -0.1971, -0.1126, -0.1218, -0.0338, -0.0448, -0.08, -0.0292, -0.027, -0.0488, -0.0622, -0.053, 0.0205, 0.0205, 0.0268, 0.0947, 0.1185, 0.1243, 0.0946, 0.1541, 0.2246, 0.226, 0.1672, 0.2316, 0.18, 0.0673, 0.0772, 0.1104, 0.0789, 0.05, 0.1544, 0.1646, 0.1581, 0.1581, 0.1247, 0.1842, 0.1741, 0.2046, 0.1843, 0.176, 0.2069, 0.1691, 0.1601, 0.1609, 0.0681, 0.1275, 0.117, 0.0417, 0.0953, 0.1341, 0.1779, 0.1404, 0.1988, 0.2429, 0.2989, 0.299, 0.2498, 0.1897, 0.1375, 0.1127, 0.0749, 0.0, 0.005, -0.0943, -0.0492, 0.0353, 0.0308, 0.0308, 0.0632, 0.0627, 0.1447, 0.1863, 0.1837, 0.2005, 0.3106, 0.284, 0.2868, 0.2808, 0.2621, 0.2648, 0.372, 0.3558, 0.398, 0.4763, 0.4193, 0.4592, 0.4555, 0.526, 0.6224, 0.8018, 0.8761, 0.8199, 1.1019, 1.2384, 1.1575, 1.2618, 1.184, 1.0395, 0.9182, 0.9666, 1.0601, 1.1449, 1.1136, 1.1136, 1.5214, 1.613, 1.5992, 1.7328, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2037317421389037723", "ticker_idx": 1, "ts": "2026-03-26T23:54:59+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU $SNDK", "tweet_type": "analysis", "timeline": "unspecified", "summary": "KIS argues TurboQuant increases HBM demand (not decreases it); memory semis oversold on misread — bullish MU and SNDK.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "[KIS — Chae Min-sook / Kim Yeon-jun] Semiconductor Industry Note: After TurboQuant and DeepSeek, the Conclusion is Clear\n\n● TurboQuant's Opening Shot\n\nOn March 25 (local time), Google officially published the TurboQuant algorithm via its official blog, claiming it can compress KV Cache by up to 6x with no performance degradation. The market interpreted this as a reduction in memory requirements, causing memory semiconductor stocks to fall sharply. However, this reflects a misreading stemming from a confusion between the roles of memory capacity and memory bandwidth. The bottleneck in AI inference is not a shortage of memory capacity, but rather a problem determined by memory access speed and data movement efficiency. TurboQuant should instead be understood as a technology that partially alleviates this bottleneck and improves GPU efficiency, enabling more tokens to be processed per unit time with the same GPU resources.\n\n● The Structure of AI Inference: Prefill and Decode\n\nLLM inference consists of two stages: Prefill and Decode. The Prefill stage is a compute-intensive task where GPU computational power limits performance, while the Decode stage is a memory-intensive task where the speed of data movement determines performance. In Decode, the existing KV cache must be repeatedly referenced with each new token generated, making it structurally sensitive to memory bandwidth and access latency. Since the Decode stage largely determines the response speed that users experience, it sits at the heart of AI inference optimization.\n\n● How TurboQuant Actually Works\n\nWhen TurboQuant claims to compress KV Cache by up to 6x, this is less about reducing the required memory capacity itself and more about significantly lowering the data size occupied by the KV Cache and the resulting memory access burden. This means the volume of data that must be handled within the same HBM bandwidth is reduced. As a result, memory access latency is alleviated and the time GPUs spend waiting for data can decrease. Because KV Cache accesses occur repeatedly during the Decode stage of LLM inference, reducing data size has a direct impact on relieving the memory bottleneck. The proportion of time GPUs spend waiting for memory responses diminishes, and compute resources are utilized more efficiently. This improves actual GPU utilization rates and increases the number of tokens that can be processed per unit time (throughput) within the same hardware environment — which can also be interpreted as a reduction in cost per token.\n\n● The Real Bottleneck in AI Inference: Bandwidth, Not Capacity\n\nIt appears the market interpreted TurboQuant as reducing memory capacity usage and therefore decreasing HBM demand. However, the core bottleneck in AI inference is not a shortage of memory capacity but rather the speed at which data is read from memory — i.e., memory bandwidth and access latency. Because GPU compute cores process data far faster than HBM can supply it, GPUs sit in a wait state while expecting data from memory. Industry research indicates that in the Decode stage, more than 50% of attention computation cycles are spent in a wait state due to memory access latency — meaning GPUs waste more than half their theoretical performance waiting for memory responses. According to research published by Google DeepMind in January 2026, Nvidia GPU 64-bit FLOPS grew approximately 80x between 2012 and 2022, while memory bandwidth grew only about 17x over the same period; this gap is expected to continue widening going forward (arXiv:2601.05047, forthcoming in IEEE Computer). TurboQuant will narrow the gap between GPU compute capability and memory bandwidth, increasing the number of tokens that can be processed per unit time on the same hardware. This lowers cost per token, encouraging broader AI usage, drawing in more services and users, and ultimately leading to an increase — not a decrease — in KV Cache consumption.\n\n● A Bottleneck TurboQuant Cannot Solve: Inter-Chip Communication Latency\n\nThere is another bottleneck in AI inference that TurboQuant does nothing to address. Large models exceed the memory capacity of a single GPU and are therefore distributed across multiple GPUs. During the Decode stage, because the model is spread across multiple GPUs, intermediate computation results must be exchanged between GPUs with each token generated. These data transfers are small in size but occur at very high frequency, which can introduce inter-chip communication latency. To reduce this latency, a single GPU must be able to handle more parameters and KV Cache — which ultimately demands more HBM capacity per GPU. In the current AI environment, where model sizes and context lengths continue to expand, this need for greater HBM capacity will only intensify. The fact that Nvidia has consistently increased HBM capacity with each successive GPU generation is likely not unrelated to this dynamic.\n\n● TurboQuant's Limitations: Still in Early Validation\n\nWhile TurboQuant carries significant technical implications, there are also limitations in its scope of application and validation that deserve consideration. The published performance benchmarks are confined to relatively simple test environments centered on single-query information retrieval tasks such as LongBench and Needle-In-A-Haystack. Experiments were also conducted primarily on comparatively small models with around 8 billion parameters; whether the same effects can be reproduced in the large-scale models of hundreds of billions of parameters used in real-world industry settings has yet to be verified. More importantly, no validation has been conducted in the rapidly expanding Agentic AI environments. In these environments, models execute repeated multi-step judgments across longer contexts and more complex KV Cache structures, giving rise to memory usage patterns that differ entirely from those seen in single-response benchmarks.\n\n$MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.12382911046243406, "ret_signed_m1d": 0.12382911046243406, "alpha_spy_m1d": 0.10564563221843071, "alpha_c_m1d": 0.09167820852404973, "bench_spy_m1d": 0.018183478244003348, "bench_c_m1d": 0.032150901938384324, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02098916480990276, "ret_signed_p1d": 0.02098916480990276, "alpha_spy_p1d": 0.03804104833323074, "alpha_c_p1d": 0.04046087674239096, "bench_spy_p1d": -0.017051883523327982, "bench_c_p1d": -0.0194717119324882, "ret_p1w": 0.16317132273469315, "ret_signed_p1w": 0.16317132273469315, "alpha_spy_p1w": 0.14652249886939828, "alpha_c_p1w": 0.13683165863534175, "bench_spy_p1w": 0.01664882386529487, "bench_c_p1w": 0.0263396640993514, "ret_p1m": 0.6411626116373081, "ret_signed_p1m": 0.6411626116373081, "alpha_spy_p1m": 0.5344333603395459, "alpha_c_p1m": 0.43195505525464295, "bench_spy_p1m": 0.10672925129776223, "bench_c_p1m": 0.2092075563826652, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5951, -0.6017, -0.6064, -0.6064, -0.5437, -0.5456, -0.4203, -0.4138, -0.4454, -0.3743, -0.3546, -0.3537, -0.357, -0.3215, -0.3143, -0.3143, -0.2488, -0.1689, -0.1653, -0.2144, -0.2195, -0.2018, -0.1252, -0.1059, -0.0446, 0.1029, 0.1531, -0.0309, -0.0447, -0.0087, -0.0328, -0.102, -0.0063, 0.045, 0.0388, 0.0388, -0.0209, -0.0046, 0.0297, 0.0776, 0.105, 0.0586, 0.0484, 0.0808, 0.0534, 0.0264, -0.0626, -0.0068, -0.0623, -0.1257, -0.0239, 0.0261, 0.0866, 0.0259, 0.0969, 0.1666, 0.194, 0.2495, 0.2801, 0.1766, 0.1647, 0.1646, 0.1238, 0.0, 0.021, -0.0508, 0.0533, 0.1485, 0.1632, 0.1632, 0.2014, 0.1784, 0.2947, 0.4118, 0.4122, 0.5792, 0.5658, 0.4784, 0.5244, 0.5269, 0.5137, 0.4979, 0.6232, 0.5459, 0.6412, 0.7743, 0.6618, 0.7644, 0.8179, 0.9679, 1.0821, 1.3315, 1.3376, 1.2215, 1.5902, 1.5657, 1.4073, 1.3994, 1.2924, 1.3337, 1.21, 1.2934, 1.3087, 1.5569, 1.4515, 1.4515, 1.6353, 1.636, 1.7217, 1.8101, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012049814574526655", "ticker_idx": 0, "ts": "2026-01-16T06:30:32+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Raising Target Price to NT$2,325 ... we raise our 2026E/2027E EPS estimates by 11% and 17%, respectively. With a higher 25x valuation (up from 23x), we raise our target price from NT$1,930 to NT$2,325", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF raises TSMC Buy PT to NT$2,325 on AI cycle strength, capex beat, raised AI CAGR guidance, and higher EPS estimates for 2026-27.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics Communication \n\nTSMC (2330TT Buy): Strong Capex in an Unprecedented AI Cycle\n\nRaising Target Price to NT$2,325  \nTSMC reported strong 4Q25 results, with full-year revenue guidance in line with expectations, while the capital expenditure guidance of US$52-56 billion far exceeded expectations. At the same time, TSMC raised its AI revenue CAGR guidance from mid-40% to mid/high-50%, which is the key focus for long-term investors. From a broader perspective, TSMC’s comments on the significant shortage of AI-related silicon supply indicate that the risk of an AI bubble is limited (at least until new capital expenditure is converted into actual capacity). For TSMC, we expect the unprecedented AI cycle to continue, with the company capturing an increasing share of value through N2/N3 capacity expansion and rapidly rising ASPs (driven by N2/A16 and advanced packaging). Taking into account the results, higher gross margins, and upgraded long-term CAGR guidance, we raise our 2026E/2027E EPS estimates by 11% and 17%, respectively. With a higher 25x valuation (up from 23x), we raise our target price from NT$1,930 to NT$2,325.\n\n30% YoY growth guidance in line with expectations, but gross margin far exceeded expectations  \n4Q25 revenue was NT$1.046 trillion, up 6% QoQ and 20% YoY. Gross margin reached 62.3%, up 180 bps QoQ, mainly benefiting from NTD depreciation, higher utilization rates, and cost optimization, though partially offset by overseas capacity ramp-up, and far above the market’s expected 60-61%. EPS was NT$19.5, up 12% QoQ and 35% YoY.  \n\nFor 1Q26 and full-year 2026, the company guided:  \n\n1) Revenue midpoint implying ~4% QoQ growth, with full-year growth close to 30% YoY, largely in line with expectations (vs. our estimate of 30% and Bloomberg consensus of high-20%);  \n2) 1Q26 gross margin of 63-65%, stronger than expectations and recent market rumors of 61-62%;  \n3) 2026 capital expenditure of US$52-56 billion, significantly above expectations (vs. our US$49-50 billion, sell-side US$46-49 billion, buy-side US$47-52 billion);  \n4) AI revenue CAGR guidance raised from mid-40% to mid/high-50%;  \n5) Long-term revenue CAGR raised to 25%, higher than the previous 15-20%.\n\nStrong US$52-56 billion capex guidance, with an aggressive tone for 2026-2028  \nDriven by the long-term AI trend, the company’s US$52-56 billion guidance for 2026 was the key highlight of the earnings call. The company also expects capital expenditure over the next three years to be significantly higher than the US$101 billion spent in the past three years. Based on our estimates of TSMC’s EUV demand of 25/36 tools in 2026/2027, we expect capex to reach US$54 billion in 2026 and US$64 billion in 2027, maintaining a normal level of capital intensity.  \n\nThe scale of advanced-node capacity and capital expenditure will further strengthen TSMC’s leading position, while migration to the new N2P node (5-10% higher ASP than N2) and A16 (20% higher ASP than N2) will continue.\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0022195632788257758, "ret_signed_m1d": -0.0022195632788257758, "alpha_spy_m1d": -0.0030581398912541413, "alpha_c_m1d": 0.0077207719960755705, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.021991847330370295, "ret_signed_p1w": -0.021991847330370295, "alpha_spy_p1w": -0.01847853252648013, "alpha_c_p1w": -0.021242532163731287, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.06997661919389087, "ret_signed_p1m": 0.06997661919389087, "alpha_spy_p1m": 0.08430444876171106, "alpha_c_p1m": 0.051669502872067286, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.09858925941593832, "ret_signed_p3m": 0.09858925941593832, "alpha_spy_p3m": 0.08385430829131924, "alpha_c_p3m": -0.03280758722559618, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1422, -0.1586, -0.1532, -0.1409, -0.1313, -0.1217, -0.1114, -0.1168, -0.1249, -0.112, -0.1435, -0.1447, -0.1575, -0.1655, -0.14, -0.1519, -0.1535, -0.178, -0.1704, -0.1786, -0.1905, -0.1775, -0.1917, -0.1988, -0.1709, -0.1708, -0.1554, -0.1554, -0.1509, -0.1621, -0.1492, -0.1394, -0.1468, -0.1416, -0.1207, -0.1163, -0.0967, -0.1097, -0.1471, -0.1596, -0.1622, -0.1911, -0.1686, -0.1561, -0.1435, -0.1327, -0.1273, -0.1273, -0.1155, -0.1211, -0.1251, -0.1125, -0.1125, -0.0666, -0.0588, -0.0437, -0.0693, -0.0712, -0.0548, -0.031, -0.0327, -0.0447, -0.0022, 0.0, 0.0, -0.0445, -0.0475, -0.0439, -0.022, -0.0283, -0.0119, -0.0003, -0.0083, -0.0346, -0.003, -0.0194, -0.0487, -0.0341, 0.0188, 0.038, 0.057, 0.0926, 0.0751, 0.07, 0.07, 0.0637, 0.058, 0.0525, 0.0822, 0.0807, 0.1266, 0.1324, 0.1005, 0.094, 0.078, 0.0313, 0.0439, 0.0335, -0.0103, 0.0184, 0.0137, 0.0355, -0.0166, -0.0119, -0.0063, 0.0133, -0.0055, -0.0078, -0.0357, -0.0088, 0.0053, 0.0185, -0.0449, -0.043, -0.073, -0.0102, 0.0002, -0.007, -0.007, 0.0009, 0.0114, 0.0716, 0.0704, 0.0854, 0.0824, 0.1126, 0.0986, 0.0642, 0.0851, 0.0726, 0.078, 0.1347, 0.1207, 0.1787, 0.1861, 0.1491, 0.1534, 0.16, 0.1647, 0.1762, 0.1551, 0.2286, 0.213, 0.2057, 0.1848, 0.1635, 0.1709, 0.2234, 0.1843, 0.1597, 0.1499, 0.1763, 0.1925, 0.1848, 0.1848, 0.2076, 0.2381, 0.2443, 0.2256, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2014513214948835696", "ticker_idx": 0, "ts": "2026-01-23T01:39:12+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "strong price increases in the first quarter are certain. It is estimated that RDIMM module prices could rise by more than 70%... first-quarter price increases may well exceed 80%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Structural AI-driven memory shortage to persist through 2027-28; RDIMM prices up 80%+ Q1, DRAM +20-30% and SSDs +30-50% beyond Q1; bullish Samsung and SK Hynix.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "RDIMM spot prices blow past US$2,000, raising odds of 80% Samsung memory hike\n\nStructural imbalances in the memory industry have emerged due to AI data center demand, sending server memory prices sky high and upstream manufacturers to go all out to increase profits. Severe shortages are expected to persist through 2027-2028. As pressure increases due to insufficient inventory, memory prices also continue to rise.\n\nConfirmation elusive of price hike\nSpot prices for server-grade 64GB RDIMM modules have surged to US$2,550, rising more than 20% in just the past half month. This suggests that official contract prices could approach a sky-high US$1,000 level, implying a potential quarterly increase of 90-100%. Recently, online reports circulated documents claiming that Samsung Electronics had issued notices of price increases, citing major changes in the global semiconductor market, including supply constraints and sharply rising upstream manufacturing costs.\n\nThe reports claimed that, effective immediately, prices for all Samsung memory products would increase by as much as 80%. Samsung Taiwan said it would not comment on market rumors.\n\nWhile the alleged price hike notice apparently originated from a Samsung distributor, related companies stated that they have not yet received any confirmed price increase notifications and are unable to verify the documents' sources. With Samsung and SK Hynix still having not fully finalized official pricing, the supply chain is anxiously awaiting the outcome.\n\nEven steeper first-quarter increases\nNonetheless, strong price increases in the first quarter are certain. It is estimated that RDIMM module prices could rise by more than 70%. Some module makers also believe that the manufacturers' continued delay in announcing prices indicates that headquarters are still consolidating views on end-market demand. With nearly all manufacturers' 2026 capacity allocations already sold out in advance, first-quarter price increases may well exceed 80%.\n\nMemory supply chain players note that surging demand for AI computing and inference has driven cloud service providers' (CSPs) procurement of DRAM and NAND Flash ever higher. With capacity expansion unable to keep pace, the demand gap is estimated at at least 25–30%. Server memory module pricing has been particularly strong, with the spot market being the most price-sensitive and likely to reflect upstream manufacturers' contract price increases. Using 64GB RDIMM modules as an example, spot prices reached US$1,850 in mid-January, but have since surged to US$2,250, representing a 20% increase within half a month.\n\nTo compare, contract prices for 64GB RDIMM modules were around US$450-460 in the fourth quarter of 2025 and were originally expected to rise to US$800-900 in the first quarter. However, as spot prices continue to spike, server memory modules could reach the US$1,000 threshold. Based on this, quarterly price increases could potentially double.\n\nServer storage cost pressures\nNAND Flash has also seen severe supply tightness as enterprise server storage capacities expand. There have recently been plans to mobilize hundreds of millions of US dollars to approach US-based NAND supplier SanDisk with prepayment offers for long-term supply contracts. However, contract negotiations have been repeatedly delayed, with pricing and supply volumes still under negotiation. While contract price increases are expected to be more moderate than anticipated, with wafer-level Flash prices rising by around 30–50%, order fulfillment rates remain below 50%.\n\nMemory contract prices will see their most dramatic, step-like increases in the first quarter of 2026. From the second quarter onward, the pace of increases is expected to moderate, but DRAM is still projected to have room for 20-30% price hikes, while solid-state drives (SSDs) are expected to continue seeing increases of 30-50%.\n\n---\n\nhttps://t.co/3ZoR6wZszl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0013149696020080537, "ret_signed_m1d": 0.0013149696020080537, "alpha_spy_m1d": 0.0016776182799742578, "alpha_c_m1d": 0.002762297704813066, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": -0.0014473281028050122, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.005078147060609828, "alpha_c_p1d": -0.006823449137131288, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": 0.006823449137131288, "ret_p1w": 0.05522687034455309, "ret_signed_p1w": 0.05522687034455309, "alpha_spy_p1w": 0.05125140969015862, "alpha_c_p1w": 0.06356640751545573, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": -0.008339537170902633, "ret_p1m": 0.26890201891172105, "ret_signed_p1m": 0.26890201891172105, "alpha_spy_p1m": 0.27882601897281367, "alpha_c_p1m": 0.31418417155034395, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": -0.0452821526386229, "ret_p3m": 0.4329401017655119, "ret_signed_p3m": 0.4329401017655119, "alpha_spy_p3m": 0.3982312312225653, "alpha_c_p3m": 0.34198577078489834, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.09095433098061356, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.349, -0.3424, -0.3189, -0.2966, -0.2731, -0.3137, -0.3418, -0.351, -0.3595, -0.3418, -0.3228, -0.3254, -0.3274, -0.364, -0.3418, -0.3601, -0.3686, -0.3418, -0.3797, -0.3673, -0.3503, -0.3274, -0.3228, -0.3424, -0.3405, -0.3235, -0.3163, -0.3124, -0.2908, -0.2836, -0.2908, -0.2934, -0.298, -0.2875, -0.3143, -0.3274, -0.294, -0.296, -0.3045, -0.277, -0.2705, -0.2731, -0.2731, -0.2345, -0.2143, -0.2117, -0.2117, -0.2117, -0.1552, -0.092, -0.0868, -0.073, -0.0874, -0.0861, -0.0874, -0.0953, -0.0776, -0.0539, -0.021, -0.0184, -0.0454, -0.0171, 0.0013, 0.0, 0.0, 0.0487, 0.0677, 0.0565, 0.0552, -0.0112, 0.1012, 0.1118, 0.0473, 0.0427, 0.094, 0.0901, 0.1032, 0.1742, 0.1913, 0.1913, 0.1913, 0.1913, 0.2492, 0.2498, 0.2689, 0.3149, 0.3379, 0.4333, 0.4234, 0.4234, 0.2827, 0.1321, 0.2597, 0.2373, 0.1407, 0.2354, 0.2492, 0.2354, 0.2064, 0.2406, 0.2748, 0.3708, 0.3182, 0.311, 0.2249, 0.2472, 0.2426, 0.1841, 0.1841, 0.1615, 0.1016, 0.2495, 0.1753, 0.2267, 0.2722, 0.2946, 0.3868, 0.344, 0.3572, 0.3242, 0.3605, 0.3901, 0.4329, 0.4231, 0.4132, 0.4428, 0.4329, 0.4791, 0.4461, 0.4791, 0.4626, 0.4889, 0.4527, 0.4527, 0.5318, 0.5318, 0.7525, 0.7887, 0.7689, 0.8809, 0.8381, 0.8711, 0.9501, 0.7821, 0.8513, 0.8151, 0.8184, 0.9732, 0.9271, 0.9271, 0.9699, 1.0226, 0.9732, 1.0885, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2014513214948835696", "ticker_idx": 1, "ts": "2026-01-23T01:39:12+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "With Samsung and SK Hynix still having not fully finalized official pricing, the supply chain is anxiously awaiting the outcome", "tweet_type": "analysis", "timeline": "quarters", "summary": "Structural AI-driven memory shortage to persist through 2027-28; RDIMM prices up 80%+ Q1, DRAM +20-30% and SSDs +30-50% beyond Q1; bullish Samsung and SK Hynix.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "RDIMM spot prices blow past US$2,000, raising odds of 80% Samsung memory hike\n\nStructural imbalances in the memory industry have emerged due to AI data center demand, sending server memory prices sky high and upstream manufacturers to go all out to increase profits. Severe shortages are expected to persist through 2027-2028. As pressure increases due to insufficient inventory, memory prices also continue to rise.\n\nConfirmation elusive of price hike\nSpot prices for server-grade 64GB RDIMM modules have surged to US$2,550, rising more than 20% in just the past half month. This suggests that official contract prices could approach a sky-high US$1,000 level, implying a potential quarterly increase of 90-100%. Recently, online reports circulated documents claiming that Samsung Electronics had issued notices of price increases, citing major changes in the global semiconductor market, including supply constraints and sharply rising upstream manufacturing costs.\n\nThe reports claimed that, effective immediately, prices for all Samsung memory products would increase by as much as 80%. Samsung Taiwan said it would not comment on market rumors.\n\nWhile the alleged price hike notice apparently originated from a Samsung distributor, related companies stated that they have not yet received any confirmed price increase notifications and are unable to verify the documents' sources. With Samsung and SK Hynix still having not fully finalized official pricing, the supply chain is anxiously awaiting the outcome.\n\nEven steeper first-quarter increases\nNonetheless, strong price increases in the first quarter are certain. It is estimated that RDIMM module prices could rise by more than 70%. Some module makers also believe that the manufacturers' continued delay in announcing prices indicates that headquarters are still consolidating views on end-market demand. With nearly all manufacturers' 2026 capacity allocations already sold out in advance, first-quarter price increases may well exceed 80%.\n\nMemory supply chain players note that surging demand for AI computing and inference has driven cloud service providers' (CSPs) procurement of DRAM and NAND Flash ever higher. With capacity expansion unable to keep pace, the demand gap is estimated at at least 25–30%. Server memory module pricing has been particularly strong, with the spot market being the most price-sensitive and likely to reflect upstream manufacturers' contract price increases. Using 64GB RDIMM modules as an example, spot prices reached US$1,850 in mid-January, but have since surged to US$2,250, representing a 20% increase within half a month.\n\nTo compare, contract prices for 64GB RDIMM modules were around US$450-460 in the fourth quarter of 2025 and were originally expected to rise to US$800-900 in the first quarter. However, as spot prices continue to spike, server memory modules could reach the US$1,000 threshold. Based on this, quarterly price increases could potentially double.\n\nServer storage cost pressures\nNAND Flash has also seen severe supply tightness as enterprise server storage capacities expand. There have recently been plans to mobilize hundreds of millions of US dollars to approach US-based NAND supplier SanDisk with prepayment offers for long-term supply contracts. However, contract negotiations have been repeatedly delayed, with pricing and supply volumes still under negotiation. While contract price increases are expected to be more moderate than anticipated, with wafer-level Flash prices rising by around 30–50%, order fulfillment rates remain below 50%.\n\nMemory contract prices will see their most dramatic, step-like increases in the first quarter of 2026. From the second quarter onward, the pace of increases is expected to moderate, but DRAM is still projected to have room for 20-30% price hikes, while solid-state drives (SSDs) are expected to continue seeing increases of 30-50%.\n\n---\n\nhttps://t.co/3ZoR6wZszl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015645344964040797, "ret_signed_m1d": -0.015645344964040797, "alpha_spy_m1d": -0.015282696286074593, "alpha_c_m1d": -0.022468837206577552, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.040417236415815405, "ret_signed_p1d": -0.040417236415815405, "alpha_spy_p1d": -0.04549538347642523, "alpha_c_p1d": -0.037242978060704424, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.18513688145899976, "ret_signed_p1w": 0.18513688145899976, "alpha_spy_p1w": 0.1811614208046053, "alpha_c_p1w": 0.17671378878476696, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": 0.23989575213378833, "ret_signed_p1m": 0.23989575213378833, "alpha_spy_p1m": 0.24981975219488095, "alpha_c_p1m": 0.2079279232403104, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.5974663578215988, "ret_signed_p3m": 0.5974663578215988, "alpha_spy_p3m": 0.5627574872786523, "alpha_c_p3m": 0.405659537014299, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3212, -0.273, -0.26, -0.2717, -0.1922, -0.2365, -0.2457, -0.2274, -0.2443, -0.2105, -0.1935, -0.1961, -0.2027, -0.2704, -0.2105, -0.2574, -0.2678, -0.2561, -0.3212, -0.3225, -0.3238, -0.3173, -0.2907, -0.309, -0.2986, -0.2725, -0.2803, -0.2934, -0.2907, -0.2477, -0.2621, -0.2347, -0.2634, -0.2555, -0.2777, -0.309, -0.2816, -0.2803, -0.2868, -0.2438, -0.2386, -0.2334, -0.2334, -0.219, -0.1656, -0.1512, -0.1512, -0.1512, -0.1173, -0.0926, -0.0535, -0.0326, -0.0143, -0.03, -0.0235, -0.0378, -0.0326, -0.0235, -0.0143, -0.0039, -0.0313, -0.0352, -0.0156, 0.0, -0.0404, 0.043, 0.0965, 0.1226, 0.1851, 0.0821, 0.1825, 0.1734, 0.0978, 0.0939, 0.1565, 0.1421, 0.1213, 0.1578, 0.1473, 0.1473, 0.1473, 0.1473, 0.1656, 0.2373, 0.2399, 0.3103, 0.3272, 0.4355, 0.3859, 0.3859, 0.2265, 0.109, 0.2291, 0.2069, 0.092, 0.2252, 0.2474, 0.2148, 0.1886, 0.2722, 0.267, 0.3793, 0.3232, 0.3153, 0.2187, 0.2879, 0.2997, 0.2187, 0.2187, 0.1403, 0.0541, 0.1716, 0.0841, 0.1442, 0.1573, 0.1965, 0.3493, 0.3036, 0.3415, 0.3584, 0.4407, 0.4838, 0.5086, 0.4734, 0.523, 0.5988, 0.5975, 0.6001, 0.5962, 0.6876, 0.698, 0.6889, 0.6798, 0.6798, 0.8901, 0.8901, 1.0912, 1.1604, 1.2022, 1.4556, 1.3969, 1.581, 1.5732, 1.376, 1.4034, 1.2793, 1.2793, 1.534, 1.5353, 1.5353, 1.6803, 1.9298, 1.9904, 2.0478, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2014513214948835696", "ticker_idx": 2, "ts": "2026-01-23T01:39:12+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM is still projected to have room for 20-30% price hikes", "tweet_type": "analysis", "timeline": "quarters", "summary": "Structural AI-driven memory shortage to persist through 2027-28; RDIMM prices up 80%+ Q1, DRAM +20-30% and SSDs +30-50% beyond Q1; bullish Samsung and SK Hynix.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "RDIMM spot prices blow past US$2,000, raising odds of 80% Samsung memory hike\n\nStructural imbalances in the memory industry have emerged due to AI data center demand, sending server memory prices sky high and upstream manufacturers to go all out to increase profits. Severe shortages are expected to persist through 2027-2028. As pressure increases due to insufficient inventory, memory prices also continue to rise.\n\nConfirmation elusive of price hike\nSpot prices for server-grade 64GB RDIMM modules have surged to US$2,550, rising more than 20% in just the past half month. This suggests that official contract prices could approach a sky-high US$1,000 level, implying a potential quarterly increase of 90-100%. Recently, online reports circulated documents claiming that Samsung Electronics had issued notices of price increases, citing major changes in the global semiconductor market, including supply constraints and sharply rising upstream manufacturing costs.\n\nThe reports claimed that, effective immediately, prices for all Samsung memory products would increase by as much as 80%. Samsung Taiwan said it would not comment on market rumors.\n\nWhile the alleged price hike notice apparently originated from a Samsung distributor, related companies stated that they have not yet received any confirmed price increase notifications and are unable to verify the documents' sources. With Samsung and SK Hynix still having not fully finalized official pricing, the supply chain is anxiously awaiting the outcome.\n\nEven steeper first-quarter increases\nNonetheless, strong price increases in the first quarter are certain. It is estimated that RDIMM module prices could rise by more than 70%. Some module makers also believe that the manufacturers' continued delay in announcing prices indicates that headquarters are still consolidating views on end-market demand. With nearly all manufacturers' 2026 capacity allocations already sold out in advance, first-quarter price increases may well exceed 80%.\n\nMemory supply chain players note that surging demand for AI computing and inference has driven cloud service providers' (CSPs) procurement of DRAM and NAND Flash ever higher. With capacity expansion unable to keep pace, the demand gap is estimated at at least 25–30%. Server memory module pricing has been particularly strong, with the spot market being the most price-sensitive and likely to reflect upstream manufacturers' contract price increases. Using 64GB RDIMM modules as an example, spot prices reached US$1,850 in mid-January, but have since surged to US$2,250, representing a 20% increase within half a month.\n\nTo compare, contract prices for 64GB RDIMM modules were around US$450-460 in the fourth quarter of 2025 and were originally expected to rise to US$800-900 in the first quarter. However, as spot prices continue to spike, server memory modules could reach the US$1,000 threshold. Based on this, quarterly price increases could potentially double.\n\nServer storage cost pressures\nNAND Flash has also seen severe supply tightness as enterprise server storage capacities expand. There have recently been plans to mobilize hundreds of millions of US dollars to approach US-based NAND supplier SanDisk with prepayment offers for long-term supply contracts. However, contract negotiations have been repeatedly delayed, with pricing and supply volumes still under negotiation. While contract price increases are expected to be more moderate than anticipated, with wafer-level Flash prices rising by around 30–50%, order fulfillment rates remain below 50%.\n\nMemory contract prices will see their most dramatic, step-like increases in the first quarter of 2026. From the second quarter onward, the pace of increases is expected to moderate, but DRAM is still projected to have room for 20-30% price hikes, while solid-state drives (SSDs) are expected to continue seeing increases of 30-50%.\n\n---\n\nhttps://t.co/3ZoR6wZszl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006503320902081611, "ret_signed_m1d": -0.006503320902081611, "alpha_spy_m1d": -0.006140672224115407, "alpha_c_m1d": -0.013326813144618366, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02228012463751272, "ret_signed_p1d": -0.02228012463751272, "alpha_spy_p1d": -0.02735827169812255, "alpha_c_p1d": -0.01910586628240174, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.09282403030236404, "ret_signed_p1w": 0.09282403030236404, "alpha_spy_p1w": 0.08884856964796957, "alpha_c_p1w": 0.08440093762813124, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": 0.18738149851147531, "ret_signed_p1m": 0.18738149851147531, "alpha_spy_p1m": 0.19730549857256793, "alpha_c_p1m": 0.15541366961799738, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.41689540237060996, "ret_signed_p3m": 0.41689540237060996, "alpha_spy_p3m": 0.3821865318276634, "alpha_c_p3m": 0.22508858156331013, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3717, -0.3495, -0.3395, -0.3362, -0.2928, -0.335, -0.3311, -0.3274, -0.3362, -0.3062, -0.3044, -0.303, -0.3125, -0.339, -0.3157, -0.3487, -0.3571, -0.3647, -0.3941, -0.3766, -0.3708, -0.3563, -0.3459, -0.3533, -0.3459, -0.3323, -0.337, -0.3463, -0.3294, -0.3046, -0.3072, -0.2895, -0.305, -0.3133, -0.3327, -0.3516, -0.3372, -0.3182, -0.3087, -0.2763, -0.2727, -0.2631, -0.2631, -0.2471, -0.2144, -0.2102, -0.2163, -0.2163, -0.1611, -0.1345, -0.0936, -0.0853, -0.0945, -0.0842, -0.0818, -0.0957, -0.092, -0.0784, -0.0426, -0.0382, -0.0545, -0.0262, -0.0065, 0.0, -0.0223, 0.0394, 0.0845, 0.0898, 0.0928, 0.0555, 0.1111, 0.0782, 0.0344, 0.0414, 0.07, 0.0554, 0.0837, 0.1226, 0.1229, 0.1229, 0.113, 0.1306, 0.153, 0.1862, 0.1874, 0.2237, 0.2462, 0.3029, 0.2804, 0.2806, 0.1531, 0.0813, 0.1608, 0.1236, 0.0689, 0.1564, 0.1814, 0.1548, 0.1538, 0.2061, 0.2324, 0.3018, 0.251, 0.2282, 0.1518, 0.1749, 0.1661, 0.0974, 0.0989, 0.0358, 0.0004, 0.114, 0.0588, 0.0959, 0.125, 0.1454, 0.2514, 0.2342, 0.2505, 0.2502, 0.3223, 0.3387, 0.362, 0.3452, 0.3529, 0.3888, 0.4169, 0.4283, 0.4286, 0.4933, 0.4743, 0.4919, 0.4757, 0.4966, 0.6216, 0.6748, 0.8374, 0.8559, 0.9469, 1.1092, 1.0513, 1.1546, 1.1553, 0.9907, 0.9869, 0.9478, 0.9767, 1.1383, 1.1141, 1.1141, 1.2976, 1.4255, 1.4251, 1.5223, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2014513214948835696", "ticker_idx": 3, "ts": "2026-01-23T01:39:12+00:00", "symbol": "NAND Flash", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "solid-state drives (SSDs) are expected to continue seeing increases of 30-50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Structural AI-driven memory shortage to persist through 2027-28; RDIMM prices up 80%+ Q1, DRAM +20-30% and SSDs +30-50% beyond Q1; bullish Samsung and SK Hynix.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket: major NAND producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "RDIMM spot prices blow past US$2,000, raising odds of 80% Samsung memory hike\n\nStructural imbalances in the memory industry have emerged due to AI data center demand, sending server memory prices sky high and upstream manufacturers to go all out to increase profits. Severe shortages are expected to persist through 2027-2028. As pressure increases due to insufficient inventory, memory prices also continue to rise.\n\nConfirmation elusive of price hike\nSpot prices for server-grade 64GB RDIMM modules have surged to US$2,550, rising more than 20% in just the past half month. This suggests that official contract prices could approach a sky-high US$1,000 level, implying a potential quarterly increase of 90-100%. Recently, online reports circulated documents claiming that Samsung Electronics had issued notices of price increases, citing major changes in the global semiconductor market, including supply constraints and sharply rising upstream manufacturing costs.\n\nThe reports claimed that, effective immediately, prices for all Samsung memory products would increase by as much as 80%. Samsung Taiwan said it would not comment on market rumors.\n\nWhile the alleged price hike notice apparently originated from a Samsung distributor, related companies stated that they have not yet received any confirmed price increase notifications and are unable to verify the documents' sources. With Samsung and SK Hynix still having not fully finalized official pricing, the supply chain is anxiously awaiting the outcome.\n\nEven steeper first-quarter increases\nNonetheless, strong price increases in the first quarter are certain. It is estimated that RDIMM module prices could rise by more than 70%. Some module makers also believe that the manufacturers' continued delay in announcing prices indicates that headquarters are still consolidating views on end-market demand. With nearly all manufacturers' 2026 capacity allocations already sold out in advance, first-quarter price increases may well exceed 80%.\n\nMemory supply chain players note that surging demand for AI computing and inference has driven cloud service providers' (CSPs) procurement of DRAM and NAND Flash ever higher. With capacity expansion unable to keep pace, the demand gap is estimated at at least 25–30%. Server memory module pricing has been particularly strong, with the spot market being the most price-sensitive and likely to reflect upstream manufacturers' contract price increases. Using 64GB RDIMM modules as an example, spot prices reached US$1,850 in mid-January, but have since surged to US$2,250, representing a 20% increase within half a month.\n\nTo compare, contract prices for 64GB RDIMM modules were around US$450-460 in the fourth quarter of 2025 and were originally expected to rise to US$800-900 in the first quarter. However, as spot prices continue to spike, server memory modules could reach the US$1,000 threshold. Based on this, quarterly price increases could potentially double.\n\nServer storage cost pressures\nNAND Flash has also seen severe supply tightness as enterprise server storage capacities expand. There have recently been plans to mobilize hundreds of millions of US dollars to approach US-based NAND supplier SanDisk with prepayment offers for long-term supply contracts. However, contract negotiations have been repeatedly delayed, with pricing and supply volumes still under negotiation. While contract price increases are expected to be more moderate than anticipated, with wafer-level Flash prices rising by around 30–50%, order fulfillment rates remain below 50%.\n\nMemory contract prices will see their most dramatic, step-like increases in the first quarter of 2026. From the second quarter onward, the pace of increases is expected to moderate, but DRAM is still projected to have room for 20-30% price hikes, while solid-state drives (SSDs) are expected to continue seeing increases of 30-50%.\n\n---\n\nhttps://t.co/3ZoR6wZszl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015230145359475134, "ret_signed_m1d": 0.015230145359475134, "alpha_spy_m1d": 0.015592794037441338, "alpha_c_m1d": 0.00840665311693838, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013320218507182346, "ret_signed_p1d": -0.013320218507182346, "alpha_spy_p1d": -0.018398365567792174, "alpha_c_p1d": -0.010145960152071365, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.14536296214230598, "ret_signed_p1w": 0.14536296214230598, "alpha_spy_p1w": 0.14138750148791152, "alpha_c_p1w": 0.13693986946807318, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": 0.2309571684807982, "ret_signed_p1m": 0.2309571684807982, "alpha_spy_p1m": 0.24088116854189082, "alpha_c_p1m": 0.19898933958732026, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.664895031515156, "ret_signed_p3m": 0.664895031515156, "alpha_spy_p3m": 0.6301861609722095, "alpha_c_p3m": 0.4730882107078562, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4407, -0.4072, -0.3953, -0.3927, -0.3634, -0.3947, -0.3856, -0.3761, -0.3618, -0.317, -0.3158, -0.3096, -0.3344, -0.3805, -0.3484, -0.3874, -0.3854, -0.4054, -0.4362, -0.4144, -0.4158, -0.4262, -0.4124, -0.4092, -0.4163, -0.4064, -0.4162, -0.4131, -0.3924, -0.3705, -0.3777, -0.3661, -0.3701, -0.3868, -0.4082, -0.4224, -0.4077, -0.3886, -0.3772, -0.3472, -0.3454, -0.33, -0.3278, -0.311, -0.3024, -0.3044, -0.3092, -0.3092, -0.2601, -0.2341, -0.1748, -0.1555, -0.1655, -0.1448, -0.1384, -0.135, -0.1383, -0.117, -0.0808, -0.0729, -0.066, -0.0138, 0.0152, 0.0, -0.0133, 0.0397, 0.092, 0.1036, 0.1454, 0.1259, 0.2, 0.1348, 0.0892, 0.0975, 0.1063, 0.0793, 0.1206, 0.1839, 0.2018, 0.1955, 0.1752, 0.1789, 0.1991, 0.2233, 0.231, 0.2607, 0.2623, 0.3019, 0.2811, 0.2779, 0.1636, 0.1237, 0.1694, 0.1272, 0.0979, 0.1809, 0.2322, 0.1992, 0.2146, 0.2816, 0.2927, 0.3691, 0.3345, 0.2944, 0.2352, 0.2449, 0.2447, 0.1572, 0.1531, 0.0941, 0.0886, 0.2123, 0.1748, 0.2058, 0.2439, 0.2557, 0.3989, 0.4196, 0.4576, 0.5129, 0.5958, 0.5535, 0.5961, 0.5481, 0.5491, 0.5924, 0.6649, 0.6588, 0.6739, 0.7569, 0.7267, 0.7634, 0.7816, 0.819, 0.9231, 1.0185, 1.1177, 1.18, 1.3409, 1.4487, 1.3755, 1.4863, 1.4359, 1.3014, 1.3484, 1.3268, 1.3655, 1.5724, 1.5548, 1.6477, 1.7701, 1.825, 1.855, 1.9886, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2024103033106481615", "ticker_idx": 0, "ts": "2026-02-18T12:45:43+00:00", "symbol": "META", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I've been building a bull case for Meta lately. Even without reaching $60, a move to the $30–40 range would be transformative against Meta's ~$160B annual ad revenue base — potentially unlocking tens of billions in incremental revenue.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long META: platform-level AI deployment across 4B MAU, Ray-Ban hardware momentum, and CPM expansion from $10–20 toward $30–40 could unlock tens of billions in incremental ad revenue.", "resolved_tickers": ["META"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "original", "lang": "en", "tweet_text": "I've been building a bull case for Meta lately. This is a casual memo — would love to hear your thoughts.\n\n---------------------------------\nMeta: An Underappreciated AI Platform Strategy (Memo)\n\n-----## 1. Does Meta Really Need the Best LLM?ByteDance’s Doubao offers a telling precedent. Despite not being China’s most powerful model, Doubao leads the Chinese AI chatbot market with 170M MAU — far ahead of DeepSeek (72M) and Tencent’s Yuanbao (73M). The edge isn’t model performance but platform integration — ByteDance embedded AI across its entire ecosystem: Douyin, Jianying, Ola Friend, and even a ZTE Nubia AI-native phone.\nMeta could become the Western ByteDance. Facebook, Instagram, WhatsApp, and Messenger combined reach ~4B MAU. Layer even a “good enough” AI on top of that, and Meta holds an overwhelming advantage in mass-market agent AI deployment.\nTo be clear, Meta hasn’t abandoned the frontier model race. It hired Scale AI’s Alexandr Wang for ~$14.3B and is developing “Avocado,” a proprietary next-gen model targeting H1 2026 release (reportedly achieving 10x+ compute efficiency vs. Llama 4). But even if Avocado falls short of the latest models from OpenAI or Anthropic, Meta’s AI strategy doesn’t fail. Platform-level utilization may matter more than benchmarks.\n\n-----## 2. AI Hardware: Narrative vs. Results\nMarket attention is fixated on OpenAI’s hardware play (Jony Ive’s io acquisition for $6.4B, Foxconn partnership), but the AI hardware that’s actually selling is Meta’s Ray-Ban smart glasses.\n- Ray-Ban Meta: 2M+ cumulative units sold; Q2 2025 sales tripled QoQ\n- Ray-Ban Display ($799): US demand exceeded supply → UK, France, Italy, Canada launches postponed; waitlists extend through end of 2026\n- EssilorLuxottica reviewing capacity expansion from 10M to 20–30M units/year (Bloomberg)\n- All this demand before Avocado (next-gen LLM) has even been integrated\n\nMeanwhile, OpenAI hardware remains at prototype stage; court filings indicate customer shipments delayed to post-February 2027.The key differentiator: Meta’s partnership with EssilorLuxottica — the world’s largest eyewear company — secured a form factor people actually wear in public. Compare this to Google Glass, which failed largely on social acceptability. The Ray-Ban brand itself is a moat.\n\n-----## 3. Ad CPM Upside Potential\nPer The Information: OpenAI is pricing ChatGPT ads at ~$60 CPM — roughly 3x Meta’s average of $10–20. OpenAI’s rationale: a “high-intent” environment commands a premium.\nBut OpenAI still lacks basic ad infrastructure (only high-level metrics like total impressions/clicks; no conversion tracking; $200K minimum commitment). Meta’s decades of precision targeting and conversion attribution remain assets OpenAI simply doesn’t have.\nAdd AI-driven hyper-personalization → contextual awareness, real-time interest signals, purchase-journey stage optimization → meaningful room to push current $10–20 CPMs higher.\nEven without reaching $60, a move to the $30–40 range would be transformative against Meta’s ~$160B annual ad revenue base — potentially unlocking tens of billions in incremental revenue. Ray-Ban glasses could also emerge as an entirely new ad channel via location-based AI ads and gaze-tracking engagement metrics.\n\n$META", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006109881180504151, "ret_signed_m1d": -0.006109881180504151, "alpha_spy_m1d": -0.0010973744375464278, "alpha_c_m1d": -0.004451826614719212, "bench_spy_m1d": -0.005012506742957723, "bench_c_m1d": -0.0016580545657849388, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002425376093321052, "ret_signed_p1d": 0.002425376093321052, "alpha_spy_p1d": 0.005062810033738319, "alpha_c_p1d": -0.002374392644669321, "bench_spy_p1d": -0.0026374339404172664, "bench_c_p1d": 0.004799768737990373, "ret_p1w": 0.016277504508523277, "ret_signed_p1w": 0.016277504508523277, "alpha_spy_p1w": 0.006281652168796059, "alpha_c_p1w": -0.00039067277961568614, "bench_spy_p1w": 0.009995852339727218, "bench_c_p1w": 0.016668177288138963, "ret_p1m": -0.055969223992665906, "ret_signed_p1m": -0.055969223992665906, "alpha_spy_p1m": -0.01737038942188629, "alpha_c_p1m": -0.0433154703899673, "bench_spy_p1m": -0.03859883457077962, "bench_c_p1m": -0.012653753602698603, "ret_p3m": -0.04895160748833205, "ret_signed_p3m": -0.04895160748833205, "alpha_spy_p3m": -0.12818529537185808, "alpha_c_p3m": -0.0730663359942536, "bench_spy_p3m": 0.07923368788352603, "bench_c_p3m": 0.024114728505921557, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0769, -0.0477, -0.0117, -0.0157, -0.0157, 0.0065, -0.0045, 0.0052, -0.0064, 0.0276, 0.0461, 0.0358, 0.0205, 0.0099, 0.0139, 0.0008, 0.0067, 0.0217, 0.0098, 0.033, 0.0242, 0.0284, 0.0338, 0.0378, 0.0378, 0.0312, 0.0241, 0.0353, 0.0262, 0.0262, 0.0112, 0.0242, 0.0271, 0.0085, 0.0044, 0.0153, -0.0019, -0.0189, -0.0431, -0.0349, -0.0357, -0.0357, -0.0608, -0.047, 0.0069, 0.0242, 0.0453, 0.0463, 0.0397, 0.1478, 0.1139, 0.0982, 0.0754, 0.0401, 0.042, 0.0284, 0.0529, 0.0428, 0.0396, 0.0102, -0.0054, -0.0054, -0.0061, 0.0, 0.0024, 0.0193, -0.0093, -0.0061, 0.0163, 0.0214, 0.0077, 0.0161, 0.0184, 0.0381, 0.027, 0.0025, 0.0065, 0.0169, 0.0181, -0.0078, -0.0459, -0.0237, -0.0311, -0.042, -0.056, -0.0763, -0.0601, -0.0774, -0.0743, -0.148, -0.182, -0.1654, -0.1098, -0.0987, -0.1061, -0.1061, -0.1084, -0.1052, -0.0471, -0.0222, -0.0199, -0.0127, 0.0308, 0.045, 0.0532, 0.0714, 0.0439, 0.0407, 0.0499, 0.0256, 0.0504, 0.0559, 0.0446, 0.0412, -0.0479, -0.0528, -0.0502, -0.0587, -0.0464, -0.0402, -0.0514, -0.0682, -0.0617, -0.0405, -0.0377, -0.0443, -0.049, -0.0623, -0.0585, -0.0549, -0.0504, -0.0504, -0.0472, -0.0115, -0.0115, -0.0158, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029037422248702244", "ticker_idx": 0, "ts": "2026-03-04T03:33:13+00:00", "symbol": "KOSPI", "kind": "index", "geography": "Korea", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Sell KOSPI and watch from the sidelines for a while", "tweet_type": "news_reaction", "timeline": "months", "summary": "Relays foreign brokerage consensus to sell KOSPI and sit on sidelines; strong-H1/weak-H2 pattern with technical correction targets as low as 5,000.", "resolved_tickers": ["EWY"], "resolution_method": "etf", "resolution_reason": "KOSPI index proxy ETF EWY", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Sell KOSPI and watch from the sidelines for a while\"… The global sentiment conveyed by foreign brokerages\n\nForeign investors are aggressively selling domestic equities. Their intention is reportedly to lock in profits first and then stay on the sidelines for the time being. Additionally, these investors are said to have been anticipating a \"strong first half, weak second half\" pattern and a technical correction in KOSPI.\n\nForeign investors' cumulative net selling in KOSPI this year: KRW 27 trillion\n\nAccording to Yonhap Infomax trading data (screen no. 3302) as of the 4th, foreign investors were cumulative net buyers from the start of the year through January 30, but have been cumulative net sellers from February 2 to the present. Year-to-date cumulative net selling now approaches KRW 27 trillion. Notably, foreigners net sold roughly KRW 7 trillion and KRW 5 trillion on February 27 and March 3, respectively — as war clouds gathered over the Middle East and conflict broke out.\n\nAn official at a foreign brokerage said, \"Foreign investors are locking in gains and de-risking.\" The explanation is that leading stocks including Samsung Electronics and SK Hynix had rallied significantly, and foreigners used the Middle East crisis as a catalyst to take profits.\n\nKOSPI has plunged for two consecutive days amid the foreign selling onslaught. Since breaching the 6,300 level late last month, the index has nosedived to 5,386 as of now.\n\nThe \"strong first half, weak second half\" narrative circulating among foreigners\n\nHe also noted that foreigners intend to \"sell first and then watch.\" The story that Korea's market would follow a \"strong first half, weak second half\" trajectory this year had been widely circulated among global investors.\n\nOn top of that, some foreign investors hold a neutral view on the Korean market — meaning they are reluctant to overweight Korea within their global equity portfolios relative to benchmark levels.\n\nThe brokerage official said there have \"always been many institutions that don't hold Korea,\" adding that \"while it's hard to generalize, there have been cases globally where investors played the tech theme through other avenues and came into Korea only belatedly.\"\n\nForeigners had been anticipating a KOSPI correction all along\n\nSome foreign investors had in fact foreseen a KOSPI correction.\n\nAn official at another foreign financial firm said, \"There was considerable consensus on the overseas side that KOSPI could undergo a significant correction from a technical standpoint.\" He added, \"Looking at technical analyses of Korean equities, there were even calls for KOSPI to fall to 5,000.\" His interpretation is that the correction foreigners had long anticipated is now materializing on a larger scale due to geopolitical risk.\n\n---\n\nhttps://t.co/WbujAruhVC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015107530349513398, "ret_signed_m1d": 0.015107530349513398, "alpha_spy_m1d": 0.008101650638077551, "alpha_c_m1d": 0.008101650638077551, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.007005879711435847, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06422562749884397, "ret_signed_p1d": 0.06422562749884397, "alpha_spy_p1d": 0.05865008065245014, "alpha_c_p1d": 0.05865008065245014, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.005575546846393831, "ret_p1w": -0.011311967469718631, "ret_signed_p1w": 0.011311967469718631, "alpha_spy_p1w": -0.0015322644396569274, "alpha_c_p1w": -0.0015322644396569274, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": -0.012844231909375559, "ret_p1m": -0.08558452622236112, "ret_signed_p1m": 0.08558452622236112, "alpha_spy_p1m": 0.045433192503067765, "alpha_c_p1m": 0.045433192503067765, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.04015133371929336, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3136, -0.308, -0.3044, -0.2958, -0.3088, -0.3255, -0.3251, -0.3338, -0.3408, -0.3297, -0.3202, -0.3172, -0.3101, -0.296, -0.296, -0.2867, -0.2653, -0.2661, -0.2765, -0.2765, -0.2393, -0.2198, -0.1992, -0.2035, -0.2067, -0.1896, -0.1893, -0.1997, -0.184, -0.1736, -0.1648, -0.1648, -0.1741, -0.1392, -0.1368, -0.1221, -0.1269, -0.0868, -0.0656, -0.0735, -0.089, -0.1, -0.0748, -0.105, -0.1054, -0.0719, -0.0619, -0.0738, -0.0281, -0.0266, -0.003, -0.003, -0.0275, -0.0107, 0.0063, 0.0559, 0.0367, 0.0758, 0.1081, 0.1194, 0.1265, 0.098, -0.0151, 0.0, -0.0642, -0.0569, -0.0036, -0.0303, -0.0113, -0.0809, -0.0761, -0.0095, 0.0004, -0.0185, 0.0033, -0.0639, -0.0042, -0.0418, -0.0493, -0.107, -0.0997, -0.1335, -0.0845, -0.0607, -0.0856, -0.0856, -0.0627, -0.0534, 0.0424, 0.0366, 0.0324, 0.0511, 0.0916, 0.0785, 0.0975, 0.1337, 0.117, 0.0924, 0.1595, 0.1207, 0.1503, 0.1664, 0.1488, 0.1458, 0.1964, 0.2056, 0.2175, 0.291, 0.3543, 0.3154, 0.4155, 0.4352, 0.3284, 0.4038, 0.4179, 0.3311, 0.3106, 0.2951, 0.3404, 0.3874, 0.3547, 0.3547, 0.4933, 0.4757, 0.5361, 0.5318, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2006870008471883841", "ticker_idx": 0, "ts": "2026-01-01T23:27:50+00:00", "symbol": "Server DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the 4Q25 fixed price is expanding to a level 53–58% higher QoQ... suppliers are expected to propose initial quotes reflecting a 60–70% increase... the growth rate for Server DRAM fixed prices in 2026 is projected as follows: 1Q (+60–65%) / 2Q (+10–15%) / 3Q (+3–8%) / 4Q (+0–5%)", "tweet_type": "analysis", "timeline": "quarters", "summary": "Server DRAM prices surging 53-58% QoQ in 4Q25 with 60-65% more in 1Q26; HBM3E prices also rising as shortages persist through 2026.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Server DRAM sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[Commentary on DRAMeXchange December Server DRAM Fixed Prices & Industry Trends]\n\n- December Pricing & Additional Orders: Negotiations for December Server DRAM fixed prices are proceeding at elevated levels, with additional orders for December delivery becoming a key topic. Major customers include US CSPs, Chinese CSPs, and OEMs. As supplier inventories are depleted, additional orders can only be fulfilled by rescheduling delivery timelines with existing customers. Consequently, actual transaction prices in December surged, rising 88–93% compared to the 3Q25 average fixed price. Furthermore, the 4Q25 fixed price is expanding to a level 53–58% higher QoQ.\n\n- 1Q26 Outlook & Supplier Stance: From December, the negotiation focus is gradually shifting to 1Q26 contracts. Given the current tight supply-demand conditions, suppliers are expected to propose initial quotes reflecting a 60–70% increase, exceeding the levels seen in December's add-on orders. Suppliers maintain a bullish outlook for continued quarterly price hikes throughout 2026. Anticipating further increases in future quarters, they are currently reluctant to sign multi-year lock-in contracts with CSPs.\n\n- 2026 Price Forecast: Considering the above, the growth rate for Server DRAM fixed prices in 2026 is projected as follows: 1Q (+60–65%) / 2Q (+10–15%) / 3Q (+3–8%) / 4Q (+0–5%). This represents an upward revision from last month's estimates, and further upward adjustments cannot be ruled out.\n\n- CSP Financials & Demand: CSPs assess that cash flows generated from existing businesses can sufficiently cover CapEx for large-scale infrastructure build-outs, with a focus on realizing monetization from inference AI. Therefore, they are not significantly pushing back against DRAM price hikes, accepting them to secure stable volumes. Analysis suggests that considering their margin structures, customers still have room to absorb these price increases.\n\n- Market Monitoring: Market participants are closely watching these purchasing strategies to determine when the peak of the price (P) cycle will occur. OEMs are expected to maintain margin headroom until at least 1H26, while CSPs are projected to be capable of absorbing potential supply gluts in the market.\n\n- HBM3E Impact & Supply Allocation: Entering December, GPU/ASIC manufacturers retracted previous expectations for a decline in HBM3E purchase prices in 2026, as suppliers successfully induced price hikes against the backdrop of shortages. Consequently, suppliers are reducing legacy commodity DRAM supply to focus on HBM3E. Amid this trend, adjustments to supply and procurement volumes for CSPs and OEMs continue within 2026 LTAs (Long-Term Agreements). Some OEMs and Chinese CSPs with relatively lower transaction stability have reportedly received notices of supply cuts. Overall, while Server DRAM supply volumes are being revised upward, PC DRAM remains an application area where suppliers are concentrating on production cuts.\n\n- Customer Prioritization: Supplier allocation decisions are based on the customer's actual demand levels, past transaction history, and partnership evaluation. With US CSPs showing rising transaction prices for Server DRAM and HBM while actively securing 1H26 deliveries, the possibility cannot be excluded that suppliers may cut Server DRAM supply for lower-priority customers. This increases the likelihood of intensifying tight supply conditions for OEMs and Chinese CSPs.\n\n- AI-Driven Demand Growth: As inference AI applications like Gemini and Copilot drive demand for commodity servers, CSPs (led by MSFT and Google) are aggressively increasing server assembly orders to ODMs. Accordingly, the 2026 server shipment growth forecast has been revised upward from 9% to 13% YoY. Additionally, the expanded adoption of multi-core CPUs is increasing procurement of high-capacity RDIMMs, pushing Server DRAM content per box up 20% YoY. Total Server DRAM bit demand growth is expected to reach 39%, significantly exceeding the previous forecast of 23%.\n\n- Contract Terms (NCNR): Suppliers are negotiating LTA contracts that include NCNR (Non-Cancellable, Non-Returnable) clauses with major customers, aiming to secure long-term CapEx visibility through 2028 based on these agreements. However, despite these contracts, securing sufficient supply in a timely manner for 2026 is expected to remain challenging.\n\n(Source: DRAMeXchange)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07227064895592512, "ret_signed_p1d": 0.07227064895592512, "alpha_spy_p1d": 0.07043760922247329, "alpha_c_p1d": 0.03570056945243645, "bench_spy_p1d": 0.0018330397334518356, "bench_c_p1d": 0.03657007950348867, "ret_p1w": 0.15490391400329662, "ret_signed_p1w": 0.15490391400329662, "alpha_spy_p1w": 0.14377356463080582, "alpha_c_p1w": 0.1025616880072997, "bench_spy_p1w": 0.0111303493724908, "bench_c_p1w": 0.052342225995996916, "ret_p1m": 0.3961856461483211, "ret_signed_p1m": 0.3961856461483211, "alpha_spy_p1m": 0.3814479133520954, "alpha_c_p1m": 0.27586802199932703, "bench_spy_p1m": 0.0147377327962257, "bench_c_p1m": 0.12031762414899405, "ret_p3m": 0.27449944577797564, "ret_signed_p3m": 0.27449944577797564, "alpha_spy_p3m": 0.3182052913653794, "alpha_c_p3m": 0.20988392166677577, "bench_spy_p3m": -0.04370584558740376, "bench_c_p3m": 0.06461552411119986, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3264, -0.3325, -0.3199, -0.3248, -0.3078, -0.3045, -0.3176, -0.2969, -0.2617, -0.2547, -0.2388, -0.249, -0.2491, -0.2469, -0.21, -0.1871, -0.1991, -0.1719, -0.1598, -0.1553, -0.1014, -0.1554, -0.1482, -0.1439, -0.1546, -0.1159, -0.1154, -0.1132, -0.1258, -0.1564, -0.1291, -0.171, -0.1817, -0.1944, -0.2291, -0.205, -0.1976, -0.1787, -0.1663, -0.1745, -0.1649, -0.1486, -0.1549, -0.1671, -0.1446, -0.1133, -0.1156, -0.0928, -0.1121, -0.1248, -0.1491, -0.1728, -0.1561, -0.1295, -0.1154, -0.0744, -0.07, -0.0569, -0.0569, -0.0371, 0.0037, 0.0084, 0.0, 0.0, 0.0723, 0.1049, 0.159, 0.1685, 0.1549, 0.1704, 0.1733, 0.1553, 0.1593, 0.1767, 0.2247, 0.2299, 0.2104, 0.249, 0.2743, 0.2823, 0.2541, 0.3322, 0.3905, 0.3966, 0.3962, 0.3544, 0.4199, 0.374, 0.3212, 0.3315, 0.3647, 0.3454, 0.3861, 0.4347, 0.4351, 0.4351, 0.4212, 0.446, 0.4734, 0.5145, 0.5152, 0.5588, 0.588, 0.6552, 0.6278, 0.6281, 0.4675, 0.3823, 0.4791, 0.4297, 0.3659, 0.4743, 0.5071, 0.4729, 0.4746, 0.5402, 0.5759, 0.6606, 0.5959, 0.5648, 0.4688, 0.4951, 0.4821, 0.3944, 0.3965, 0.315, 0.2745, 0.4183, 0.3507, 0.396, 0.4338, 0.4585, 0.5916, 0.5728, 0.5921, 0.5918, 0.6852, 0.7036, 0.7326, 0.7121, 0.7196, 0.763, 0.8029, 0.8167, 0.8187, 0.9011, 0.8745, 0.8987, 0.8782, 0.9075, 1.0635, 1.138, 1.3412, 1.3603, 1.4855, 1.689, 1.6142, 1.7438, 1.7419, 1.5334, 1.523, 1.4791, 1.5193, 1.72, 1.688, 1.688, 1.9323, 2.0906, 2.0878, 2.2146, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2006870008471883841", "ticker_idx": 1, "ts": "2026-01-01T23:27:50+00:00", "symbol": "HBM", "kind": "sector", "geography": "global", "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GPU/ASIC manufacturers retracted previous expectations for a decline in HBM3E purchase prices in 2026, as suppliers successfully induced price hikes against the backdrop of shortages", "tweet_type": "analysis", "timeline": "quarters", "summary": "Server DRAM prices surging 53-58% QoQ in 4Q25 with 60-65% more in 1Q26; HBM3E prices also rising as shortages persist through 2026.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "Global HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "[Commentary on DRAMeXchange December Server DRAM Fixed Prices & Industry Trends]\n\n- December Pricing & Additional Orders: Negotiations for December Server DRAM fixed prices are proceeding at elevated levels, with additional orders for December delivery becoming a key topic. Major customers include US CSPs, Chinese CSPs, and OEMs. As supplier inventories are depleted, additional orders can only be fulfilled by rescheduling delivery timelines with existing customers. Consequently, actual transaction prices in December surged, rising 88–93% compared to the 3Q25 average fixed price. Furthermore, the 4Q25 fixed price is expanding to a level 53–58% higher QoQ.\n\n- 1Q26 Outlook & Supplier Stance: From December, the negotiation focus is gradually shifting to 1Q26 contracts. Given the current tight supply-demand conditions, suppliers are expected to propose initial quotes reflecting a 60–70% increase, exceeding the levels seen in December's add-on orders. Suppliers maintain a bullish outlook for continued quarterly price hikes throughout 2026. Anticipating further increases in future quarters, they are currently reluctant to sign multi-year lock-in contracts with CSPs.\n\n- 2026 Price Forecast: Considering the above, the growth rate for Server DRAM fixed prices in 2026 is projected as follows: 1Q (+60–65%) / 2Q (+10–15%) / 3Q (+3–8%) / 4Q (+0–5%). This represents an upward revision from last month's estimates, and further upward adjustments cannot be ruled out.\n\n- CSP Financials & Demand: CSPs assess that cash flows generated from existing businesses can sufficiently cover CapEx for large-scale infrastructure build-outs, with a focus on realizing monetization from inference AI. Therefore, they are not significantly pushing back against DRAM price hikes, accepting them to secure stable volumes. Analysis suggests that considering their margin structures, customers still have room to absorb these price increases.\n\n- Market Monitoring: Market participants are closely watching these purchasing strategies to determine when the peak of the price (P) cycle will occur. OEMs are expected to maintain margin headroom until at least 1H26, while CSPs are projected to be capable of absorbing potential supply gluts in the market.\n\n- HBM3E Impact & Supply Allocation: Entering December, GPU/ASIC manufacturers retracted previous expectations for a decline in HBM3E purchase prices in 2026, as suppliers successfully induced price hikes against the backdrop of shortages. Consequently, suppliers are reducing legacy commodity DRAM supply to focus on HBM3E. Amid this trend, adjustments to supply and procurement volumes for CSPs and OEMs continue within 2026 LTAs (Long-Term Agreements). Some OEMs and Chinese CSPs with relatively lower transaction stability have reportedly received notices of supply cuts. Overall, while Server DRAM supply volumes are being revised upward, PC DRAM remains an application area where suppliers are concentrating on production cuts.\n\n- Customer Prioritization: Supplier allocation decisions are based on the customer's actual demand levels, past transaction history, and partnership evaluation. With US CSPs showing rising transaction prices for Server DRAM and HBM while actively securing 1H26 deliveries, the possibility cannot be excluded that suppliers may cut Server DRAM supply for lower-priority customers. This increases the likelihood of intensifying tight supply conditions for OEMs and Chinese CSPs.\n\n- AI-Driven Demand Growth: As inference AI applications like Gemini and Copilot drive demand for commodity servers, CSPs (led by MSFT and Google) are aggressively increasing server assembly orders to ODMs. Accordingly, the 2026 server shipment growth forecast has been revised upward from 9% to 13% YoY. Additionally, the expanded adoption of multi-core CPUs is increasing procurement of high-capacity RDIMMs, pushing Server DRAM content per box up 20% YoY. Total Server DRAM bit demand growth is expected to reach 39%, significantly exceeding the previous forecast of 23%.\n\n- Contract Terms (NCNR): Suppliers are negotiating LTA contracts that include NCNR (Non-Cancellable, Non-Returnable) clauses with major customers, aiming to secure long-term CapEx visibility through 2028 based on these agreements. However, despite these contracts, securing sufficient supply in a timely manner for 2026 is expected to remain challenging.\n\n(Source: DRAMeXchange)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07227064895592512, "ret_signed_p1d": 0.07227064895592512, "alpha_spy_p1d": 0.07043760922247329, "alpha_c_p1d": 0.03570056945243645, "bench_spy_p1d": 0.0018330397334518356, "bench_c_p1d": 0.03657007950348867, "ret_p1w": 0.15490391400329662, "ret_signed_p1w": 0.15490391400329662, "alpha_spy_p1w": 0.14377356463080582, "alpha_c_p1w": 0.1025616880072997, "bench_spy_p1w": 0.0111303493724908, "bench_c_p1w": 0.052342225995996916, "ret_p1m": 0.3961856461483211, "ret_signed_p1m": 0.3961856461483211, "alpha_spy_p1m": 0.3814479133520954, "alpha_c_p1m": 0.27586802199932703, "bench_spy_p1m": 0.0147377327962257, "bench_c_p1m": 0.12031762414899405, "ret_p3m": 0.27449944577797564, "ret_signed_p3m": 0.27449944577797564, "alpha_spy_p3m": 0.3182052913653794, "alpha_c_p3m": 0.20988392166677577, "bench_spy_p3m": -0.04370584558740376, "bench_c_p3m": 0.06461552411119986, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3264, -0.3325, -0.3199, -0.3248, -0.3078, -0.3045, -0.3176, -0.2969, -0.2617, -0.2547, -0.2388, -0.249, -0.2491, -0.2469, -0.21, -0.1871, -0.1991, -0.1719, -0.1598, -0.1553, -0.1014, -0.1554, -0.1482, -0.1439, -0.1546, -0.1159, -0.1154, -0.1132, -0.1258, -0.1564, -0.1291, -0.171, -0.1817, -0.1944, -0.2291, -0.205, -0.1976, -0.1787, -0.1663, -0.1745, -0.1649, -0.1486, -0.1549, -0.1671, -0.1446, -0.1133, -0.1156, -0.0928, -0.1121, -0.1248, -0.1491, -0.1728, -0.1561, -0.1295, -0.1154, -0.0744, -0.07, -0.0569, -0.0569, -0.0371, 0.0037, 0.0084, 0.0, 0.0, 0.0723, 0.1049, 0.159, 0.1685, 0.1549, 0.1704, 0.1733, 0.1553, 0.1593, 0.1767, 0.2247, 0.2299, 0.2104, 0.249, 0.2743, 0.2823, 0.2541, 0.3322, 0.3905, 0.3966, 0.3962, 0.3544, 0.4199, 0.374, 0.3212, 0.3315, 0.3647, 0.3454, 0.3861, 0.4347, 0.4351, 0.4351, 0.4212, 0.446, 0.4734, 0.5145, 0.5152, 0.5588, 0.588, 0.6552, 0.6278, 0.6281, 0.4675, 0.3823, 0.4791, 0.4297, 0.3659, 0.4743, 0.5071, 0.4729, 0.4746, 0.5402, 0.5759, 0.6606, 0.5959, 0.5648, 0.4688, 0.4951, 0.4821, 0.3944, 0.3965, 0.315, 0.2745, 0.4183, 0.3507, 0.396, 0.4338, 0.4585, 0.5916, 0.5728, 0.5921, 0.5918, 0.6852, 0.7036, 0.7326, 0.7121, 0.7196, 0.763, 0.8029, 0.8167, 0.8187, 0.9011, 0.8745, 0.8987, 0.8782, 0.9075, 1.0635, 1.138, 1.3412, 1.3603, 1.4855, 1.689, 1.6142, 1.7438, 1.7419, 1.5334, 1.523, 1.4791, 1.5193, 1.72, 1.688, 1.688, 1.9323, 2.0906, 2.0878, 2.2146, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2030454034344399320", "ticker_idx": 0, "ts": "2026-03-08T01:22:20+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HSBC raises Samsung Electronics target price to KRW 300,000", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares/endorses HSBC raising Samsung Electronics PT to KRW 300,000 with 2026 OP forecast of KRW 220T.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "ko", "tweet_text": "HSBC raises Samsung Electronics target price to KRW 300,000\n\nHSBC forecasts Samsung Electronics' 2026 operating profit at KRW 220 trillion (approximately $148.1 billion USD)\n\nHSBC, 삼성전자 목표가 30만원으로 상향\n\nHSBC, 올해 삼성전자 영업이익 148.1 billion USD로 전망 https://t.co/RSz18g1pMz", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08472625011460999, "ret_signed_m1d": 0.08472625011460999, "alpha_spy_m1d": 0.09341020347896523, "alpha_c_m1d": 0.10239941511345074, "bench_spy_m1d": -0.008683953364355235, "bench_c_m1d": -0.01767316499884075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08299717427857112, "ret_signed_p1d": 0.08299717427857112, "alpha_spy_p1d": 0.08460430118393825, "alpha_c_p1d": 0.08299717427857112, "bench_spy_p1d": -0.0016071269053671289, "bench_c_p1d": 0.0, "ret_p1w": 0.08760810333736901, "ret_signed_p1w": 0.08760810333736901, "alpha_spy_p1w": 0.10123101006156776, "alpha_c_p1w": 0.09462006679973212, "bench_spy_p1w": -0.013622906724198747, "bench_c_p1w": -0.007011963462363102, "ret_p1m": 0.1349090664939756, "ret_signed_p1m": 0.1349090664939756, "alpha_spy_p1m": 0.1603410309135881, "alpha_c_p1m": 0.15035790484955236, "bench_spy_p1m": -0.025431964419612485, "bench_c_p1m": -0.015448838355576755, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3805, -0.3845, -0.3754, -0.3989, -0.4104, -0.3811, -0.3828, -0.3903, -0.3662, -0.3605, -0.3628, -0.3628, -0.3289, -0.3112, -0.3089, -0.3089, -0.3089, -0.2594, -0.204, -0.1994, -0.1873, -0.2, -0.1988, -0.2, -0.2069, -0.1914, -0.1706, -0.1418, -0.1395, -0.1631, -0.1383, -0.1222, -0.1233, -0.1233, -0.0807, -0.064, -0.0738, -0.0749, -0.1331, -0.0346, -0.0254, -0.0818, -0.0859, -0.0409, -0.0444, -0.0329, 0.0294, 0.0444, 0.0444, 0.0444, 0.0444, 0.0951, 0.0957, 0.1124, 0.1527, 0.1729, 0.2565, 0.2478, 0.2478, 0.1245, -0.0075, 0.1043, 0.0847, 0.0, 0.083, 0.0951, 0.083, 0.0576, 0.0876, 0.1176, 0.2017, 0.1556, 0.1493, 0.0738, 0.0934, 0.0893, 0.038, 0.038, 0.0182, -0.0343, 0.0953, 0.0304, 0.0754, 0.1153, 0.1349, 0.2158, 0.1782, 0.1898, 0.1609, 0.1927, 0.2187, 0.2562, 0.2475, 0.2389, 0.2649, 0.2562, 0.2966, 0.2677, 0.2966, 0.2822, 0.3053, 0.2735, 0.2735, 0.3428, 0.3428, 0.5363, 0.5681, 0.5508, 0.6489, 0.6114, 0.6403, 0.7096, 0.5623, 0.6229, 0.5912, 0.5941, 0.7298, 0.6894, 0.6894, 0.7269, 0.7731, 0.7298, 0.8309, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029826157810700582", "ticker_idx": 0, "ts": "2026-03-06T07:47:23+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel's advanced packaging strategy is set to reach a critical turning point in second half 2026, boosting confidence across its long-struggling supply chain and triggering renewed investment momentum", "tweet_type": "analysis", "timeline": "years", "summary": "EMIB gaining traction as CoWoS alternative; INTC packaging revenues to scale to tens of billions from H2 2026, driving renewed investment by Ibiden and Unimicron.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Intel’s EMIB gains traction as CoWoS alternative, reviving IC substrate investment\n\nThe global AI surge has sharply increased demand for advanced chips, leaving TSMC's CoWoS capacity in tight supply. Intel's advanced packaging strategy is set to reach a critical turning point in second half 2026, boosting confidence across its long-struggling supply chain and triggering renewed investment momentum.\n\nIntel has reportedly maintained close collaboration on its Embedded Multi-die Interconnect Bridge (EMIB) advanced packaging technology with two major IC substrate suppliers: Japan's Ibiden and Taiwan's Unimicron. Both companies also serve the CoWoS supply chain, capturing business from both TSMC and Intel ecosystems and positioning themselves to benefit from an expected order surge over the coming years.\n\nUnder intense pressure from AI chipmakers locking down TSMC capacity and amid manufacturing reshoring trends to the US, many previously sidelined customers are now turning to Intel's EMIB solution. The industry views EMIB as the most powerful alternative advanced packaging technology outside of TSMC's domain.\n\nIC substrate players note that while TSMC's CoWoS will remain the dominant industry standard, the supply chain actively seeks secondary sources.\n\nAlthough the back-end processes differ significantly between the two advanced packaging technologies, their front-end production shares commonalities, allowing some capacity flexibility depending on end-customer allocation.\n\nInvestment moves by the two leading IC substrate suppliers illustrate this trend. Ibiden announced plans to invest about JPY500 billion (US$3.2 billion) over three years at its Gifu sites—JPY220 billion and JPY280 billion respectively—focusing on expanding advanced IC substrate capacity for AI servers. By 2028, Ibiden aims to increase output to 2.5x current levels to meet orders from Intel, TSMC, and others.\n\nMeanwhile, Unimicron has repeatedly raised its 2026 capital expenditure budget to approximately NT$34 billion (US$1.1 billion), concentrating ABF substrate expansion at its Hsinchu plants. These expansions are expected to contribute significant revenue starting second half of 2027. The company's board has also expanded long-lead equipment procurement for 2027, increasing orders to around NT$9.2 billion.\n\nNotably, Intel's cooperation with these two IC substrate suppliers goes beyond traditional purchasing models. Years ago, they jointly invested in dedicated production facilities for EMIB orders, including Unimicron's Taoyuan Yangmei plant.\n\nHowever, due to slower-than-expected adoption of EMIB advanced packaging by end products, these dedicated lines remained underutilized for a long time. With Intel's consent, both suppliers prioritized reallocating part of this capacity to CoWoS customers to alleviate low utilization pressures.\n\nIntel CFO David Zinsner recently confirmed during earnings calls that Intel's EMIB and upgraded EMIB-T solutions have attracted proactive inquiries from multiple IC design customers, including heavyweights Apple, Nvidia, and Qualcomm.\n\nZinsner stated that based on ongoing negotiations, Intel expects EMIB-related advanced packaging revenues to reach several hundred million dollars initially, then scale up to tens of billions of dollars starting second half of 2026—far exceeding earlier projections. This reflects surging market interest and accelerating mass production commitments among key customers.\n\n$INTC\n\n---\n\nhttps://t.co/xCBhAU5lK5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05826814142529457, "ret_signed_m1d": 0.05826814142529457, "alpha_spy_m1d": 0.04498692769215307, "alpha_c_m1d": 0.0194043390293408, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.03886380239595377, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.049746746964495436, "ret_signed_p1d": 0.049746746964495436, "alpha_spy_p1d": 0.04098672195145636, "alpha_c_p1d": 0.01345812586088746, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.036288621103607976, "ret_p1w": 0.05412257917825336, "ret_signed_p1w": 0.05412257917825336, "alpha_spy_p1w": 0.06912894621638366, "alpha_c_p1w": 0.03633303467251614, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": 0.01778954450573722, "ret_p1m": 0.16950716077218897, "ret_signed_p1m": 0.16950716077218897, "alpha_spy_p1m": 0.18683433368892954, "alpha_c_p1m": 0.12898789041624403, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": 0.040519270355944936, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0673, -0.0608, -0.0901, -0.1292, -0.1361, -0.1407, -0.1697, -0.1644, -0.152, -0.1624, -0.1628, -0.1672, -0.1672, -0.1663, -0.1552, -0.1409, -0.1502, -0.1502, -0.093, -0.0933, -0.0778, -0.0182, -0.0532, 0.0491, 0.0147, 0.0891, 0.1221, 0.1129, 0.0815, 0.0815, 0.1184, 0.2494, 0.251, 0.038, -0.0214, 0.0117, 0.1234, 0.1207, 0.0702, 0.1241, 0.1343, 0.1193, 0.111, 0.1651, 0.1571, 0.0854, 0.1122, 0.0705, 0.0776, 0.0776, 0.0636, 0.047, 0.0276, 0.0159, 0.0048, 0.0622, 0.0797, 0.047, 0.0504, 0.0479, -0.0074, 0.0497, 0.0583, 0.0, 0.0497, 0.0774, 0.105, 0.0421, 0.0541, 0.0539, 0.0147, 0.0371, 0.0636, 0.0104, 0.0136, 0.0147, 0.0866, 0.0157, -0.0067, -0.0514, 0.0164, 0.1062, 0.1603, 0.1603, 0.1695, 0.2186, 0.3577, 0.4215, 0.4367, 0.5012, 0.4696, 0.4956, 0.5776, 0.5776, 0.5131, 0.526, 0.5032, 0.538, 0.901, 0.9574, 0.9466, 1.1822, 1.176, 1.2943, 1.2059, 1.4908, 1.6027, 1.5246, 1.877, 1.9811, 1.7778, 1.7704, 1.67, 1.5051, 1.4912, 1.5518, 1.7398, 1.7292, 1.76, 1.76, 1.8448, 1.8045, 1.7842, 1.6412, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029826157810700582", "ticker_idx": 1, "ts": "2026-03-06T07:47:23+00:00", "symbol": "Ibiden", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Ibiden announced plans to invest about JPY500 billion (US$3.2 billion) over three years at its Gifu sites—focusing on expanding advanced IC substrate capacity for AI servers. By 2028, Ibiden aims to increase output to 2.5x current levels", "tweet_type": "analysis", "timeline": "years", "summary": "EMIB gaining traction as CoWoS alternative; INTC packaging revenues to scale to tens of billions from H2 2026, driving renewed investment by Ibiden and Unimicron.", "resolved_tickers": ["4062.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ibiden Co. Ltd 4062.T Japan ABF substrate", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Intel’s EMIB gains traction as CoWoS alternative, reviving IC substrate investment\n\nThe global AI surge has sharply increased demand for advanced chips, leaving TSMC's CoWoS capacity in tight supply. Intel's advanced packaging strategy is set to reach a critical turning point in second half 2026, boosting confidence across its long-struggling supply chain and triggering renewed investment momentum.\n\nIntel has reportedly maintained close collaboration on its Embedded Multi-die Interconnect Bridge (EMIB) advanced packaging technology with two major IC substrate suppliers: Japan's Ibiden and Taiwan's Unimicron. Both companies also serve the CoWoS supply chain, capturing business from both TSMC and Intel ecosystems and positioning themselves to benefit from an expected order surge over the coming years.\n\nUnder intense pressure from AI chipmakers locking down TSMC capacity and amid manufacturing reshoring trends to the US, many previously sidelined customers are now turning to Intel's EMIB solution. The industry views EMIB as the most powerful alternative advanced packaging technology outside of TSMC's domain.\n\nIC substrate players note that while TSMC's CoWoS will remain the dominant industry standard, the supply chain actively seeks secondary sources.\n\nAlthough the back-end processes differ significantly between the two advanced packaging technologies, their front-end production shares commonalities, allowing some capacity flexibility depending on end-customer allocation.\n\nInvestment moves by the two leading IC substrate suppliers illustrate this trend. Ibiden announced plans to invest about JPY500 billion (US$3.2 billion) over three years at its Gifu sites—JPY220 billion and JPY280 billion respectively—focusing on expanding advanced IC substrate capacity for AI servers. By 2028, Ibiden aims to increase output to 2.5x current levels to meet orders from Intel, TSMC, and others.\n\nMeanwhile, Unimicron has repeatedly raised its 2026 capital expenditure budget to approximately NT$34 billion (US$1.1 billion), concentrating ABF substrate expansion at its Hsinchu plants. These expansions are expected to contribute significant revenue starting second half of 2027. The company's board has also expanded long-lead equipment procurement for 2027, increasing orders to around NT$9.2 billion.\n\nNotably, Intel's cooperation with these two IC substrate suppliers goes beyond traditional purchasing models. Years ago, they jointly invested in dedicated production facilities for EMIB orders, including Unimicron's Taoyuan Yangmei plant.\n\nHowever, due to slower-than-expected adoption of EMIB advanced packaging by end products, these dedicated lines remained underutilized for a long time. With Intel's consent, both suppliers prioritized reallocating part of this capacity to CoWoS customers to alleviate low utilization pressures.\n\nIntel CFO David Zinsner recently confirmed during earnings calls that Intel's EMIB and upgraded EMIB-T solutions have attracted proactive inquiries from multiple IC design customers, including heavyweights Apple, Nvidia, and Qualcomm.\n\nZinsner stated that based on ongoing negotiations, Intel expects EMIB-related advanced packaging revenues to reach several hundred million dollars initially, then scale up to tens of billions of dollars starting second half of 2026—far exceeding earlier projections. This reflects surging market interest and accelerating mass production commitments among key customers.\n\n$INTC\n\n---\n\nhttps://t.co/xCBhAU5lK5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.027791915867965367, "ret_signed_m1d": 0.027791915867965367, "alpha_spy_m1d": 0.01451070213482386, "alpha_c_m1d": 0.006741566134193855, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.021050349733771512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08147208712664034, "ret_signed_p1d": -0.08147208712664034, "alpha_spy_p1d": -0.09023211213967941, "alpha_c_p1d": -0.09946321224827037, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.017991125121630036, "ret_p1w": -0.0032994670979793472, "ret_signed_p1w": -0.0032994670979793472, "alpha_spy_p1w": 0.01170689994015095, "alpha_c_p1w": 0.00026953554872677365, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": -0.003569002646706121, "ret_p1m": 0.08750574602403072, "ret_signed_p1m": 0.08750574602403072, "alpha_spy_p1m": 0.10483291894077129, "alpha_c_p1m": 0.08998174717423058, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": -0.002476001150199858, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1796, -0.1999, -0.1954, -0.1907, -0.2456, -0.2782, -0.2602, -0.2922, -0.2608, -0.2227, -0.2021, -0.158, -0.1793, -0.1599, -0.1591, -0.1458, -0.1458, -0.1458, -0.1458, -0.0916, -0.1102, -0.0835, -0.1306, -0.148, -0.148, -0.0774, -0.0563, -0.0505, -0.0057, -0.0062, 0.0066, 0.0555, 0.0845, 0.0513, 0.0052, 0.0152, 0.047, 0.0478, 0.0443, -0.019, 0.0648, -0.0863, -0.0971, -0.1009, -0.0109, 0.0048, 0.0048, 0.0792, 0.0964, 0.1142, 0.1352, 0.1845, 0.2231, 0.1621, 0.1621, 0.2543, 0.2249, 0.1801, 0.2103, 0.1888, 0.0937, 0.0089, 0.0278, 0.0, -0.0815, -0.006, -0.0105, -0.0272, -0.0033, 0.0341, 0.0786, 0.0763, 0.0492, 0.0492, -0.0431, -0.0057, 0.0759, 0.0736, 0.0494, -0.0114, -0.0628, 0.026, -0.0115, 0.0438, 0.0875, 0.1084, 0.2452, 0.2472, 0.2571, 0.1933, 0.2605, 0.1925, 0.2169, 0.1741, 0.2173, 0.3425, 0.4239, 0.4156, 0.5943, 0.6108, 0.5968, 0.5968, 0.7138, 0.7004, 0.7004, 0.7004, 0.7004, 1.0818, 0.9725, 0.9903, 1.1041, 1.1689, 1.1562, 0.9909, 0.9877, 0.9477, 0.9884, 1.2725, 1.4473, 1.6609, 1.6654, 1.5967, 1.5096, 1.9241, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029826157810700582", "ticker_idx": 2, "ts": "2026-03-06T07:47:23+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron has repeatedly raised its 2026 capital expenditure budget to approximately NT$34 billion (US$1.1 billion), concentrating ABF substrate expansion at its Hsinchu plants. These expansions are expected to contribute significant revenue starting second half of 2027", "tweet_type": "analysis", "timeline": "years", "summary": "EMIB gaining traction as CoWoS alternative; INTC packaging revenues to scale to tens of billions from H2 2026, driving renewed investment by Ibiden and Unimicron.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Intel’s EMIB gains traction as CoWoS alternative, reviving IC substrate investment\n\nThe global AI surge has sharply increased demand for advanced chips, leaving TSMC's CoWoS capacity in tight supply. Intel's advanced packaging strategy is set to reach a critical turning point in second half 2026, boosting confidence across its long-struggling supply chain and triggering renewed investment momentum.\n\nIntel has reportedly maintained close collaboration on its Embedded Multi-die Interconnect Bridge (EMIB) advanced packaging technology with two major IC substrate suppliers: Japan's Ibiden and Taiwan's Unimicron. Both companies also serve the CoWoS supply chain, capturing business from both TSMC and Intel ecosystems and positioning themselves to benefit from an expected order surge over the coming years.\n\nUnder intense pressure from AI chipmakers locking down TSMC capacity and amid manufacturing reshoring trends to the US, many previously sidelined customers are now turning to Intel's EMIB solution. The industry views EMIB as the most powerful alternative advanced packaging technology outside of TSMC's domain.\n\nIC substrate players note that while TSMC's CoWoS will remain the dominant industry standard, the supply chain actively seeks secondary sources.\n\nAlthough the back-end processes differ significantly between the two advanced packaging technologies, their front-end production shares commonalities, allowing some capacity flexibility depending on end-customer allocation.\n\nInvestment moves by the two leading IC substrate suppliers illustrate this trend. Ibiden announced plans to invest about JPY500 billion (US$3.2 billion) over three years at its Gifu sites—JPY220 billion and JPY280 billion respectively—focusing on expanding advanced IC substrate capacity for AI servers. By 2028, Ibiden aims to increase output to 2.5x current levels to meet orders from Intel, TSMC, and others.\n\nMeanwhile, Unimicron has repeatedly raised its 2026 capital expenditure budget to approximately NT$34 billion (US$1.1 billion), concentrating ABF substrate expansion at its Hsinchu plants. These expansions are expected to contribute significant revenue starting second half of 2027. The company's board has also expanded long-lead equipment procurement for 2027, increasing orders to around NT$9.2 billion.\n\nNotably, Intel's cooperation with these two IC substrate suppliers goes beyond traditional purchasing models. Years ago, they jointly invested in dedicated production facilities for EMIB orders, including Unimicron's Taoyuan Yangmei plant.\n\nHowever, due to slower-than-expected adoption of EMIB advanced packaging by end products, these dedicated lines remained underutilized for a long time. With Intel's consent, both suppliers prioritized reallocating part of this capacity to CoWoS customers to alleviate low utilization pressures.\n\nIntel CFO David Zinsner recently confirmed during earnings calls that Intel's EMIB and upgraded EMIB-T solutions have attracted proactive inquiries from multiple IC design customers, including heavyweights Apple, Nvidia, and Qualcomm.\n\nZinsner stated that based on ongoing negotiations, Intel expects EMIB-related advanced packaging revenues to reach several hundred million dollars initially, then scale up to tens of billions of dollars starting second half of 2026—far exceeding earlier projections. This reflects surging market interest and accelerating mass production commitments among key customers.\n\n$INTC\n\n---\n\nhttps://t.co/xCBhAU5lK5", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04166666666666674, "ret_signed_m1d": 0.04166666666666674, "alpha_spy_m1d": 0.028385452933525235, "alpha_c_m1d": 0.02061631693289523, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.021050349733771512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09953703703703709, "ret_signed_p1d": -0.09953703703703709, "alpha_spy_p1d": -0.10829706205007616, "alpha_c_p1d": -0.11752816215866713, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.017991125121630036, "ret_p1w": 0.17592592592592582, "ret_signed_p1w": 0.17592592592592582, "alpha_spy_p1w": 0.19093229296405612, "alpha_c_p1w": 0.17949492857263194, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": -0.003569002646706121, "ret_p1m": 0.20138888888888884, "ret_signed_p1m": 0.20138888888888884, "alpha_spy_p1m": 0.2187160618056294, "alpha_c_p1m": 0.2038648900390887, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": -0.002476001150199858, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4769, -0.4734, -0.4792, -0.4792, -0.4873, -0.5046, -0.5093, -0.5208, -0.5023, -0.4896, -0.5012, -0.4977, -0.4977, -0.4896, -0.4896, -0.4919, -0.4907, -0.4907, -0.4931, -0.4676, -0.4664, -0.4769, -0.4734, -0.478, -0.4363, -0.412, -0.4213, -0.4225, -0.3657, -0.3032, -0.287, -0.2569, -0.2431, -0.2153, -0.2292, -0.1806, -0.1493, -0.2025, -0.1238, -0.1551, -0.1088, -0.1019, -0.1539, -0.213, -0.1956, -0.14, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.0613, 0.0324, 0.0613, 0.1146, 0.1146, 0.1134, 0.0718, -0.0347, 0.0417, 0.0, -0.0995, -0.0185, 0.0787, 0.1111, 0.1759, 0.2708, 0.2569, 0.3426, 0.3449, 0.2824, 0.1551, 0.0648, 0.1713, 0.162, 0.169, 0.1424, 0.0289, 0.1308, 0.2014, 0.2014, 0.2014, 0.3056, 0.4352, 0.4468, 0.4769, 0.4491, 0.3819, 0.4306, 0.4931, 0.4884, 0.5694, 0.6435, 0.6574, 0.6852, 0.8287, 0.9421, 0.9097, 0.8588, 1.044, 1.044, 1.1088, 1.0903, 0.9861, 1.0741, 0.8935, 0.9931, 1.0255, 1.0648, 1.0394, 0.9005, 0.8889, 0.8935, 0.9051, 1.0949, 1.2454, 1.2917, 1.5116, 1.5, 1.3727, 1.4421, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008328141354471930", "ticker_idx": 0, "ts": "2026-01-06T00:01:56+00:00", "symbol": "Ibiden", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "high-end ABF substrate supply tight at Japan's Ibiden and Taiwan's Nanya PCB and Unimicron, double-digit ABF price hike requests have already become the dominant trend", "tweet_type": "analysis", "timeline": "quarters", "summary": "AI-driven structural tightness in high-end ABF substrates expected to drive 15–20%+ price hikes in 2026; Ibiden, Unimicron, and Kinsus are key beneficiaries.", "resolved_tickers": ["4062.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ibiden Co. Ltd 4062.T Japan ABF substrate", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ AI Platform Push Drives High-End Substrate Price Hikes \n\n- As demand for AI servers and high-performance computing (HPC) continues to expand, supply-demand conditions for high-end IC substrates remain persistently tight.\n\n- In particular, shortages of key materials such as low-coefficient-of-thermal-expansion (Low CTE) fiberglass are driving a shift in pricing negotiations—from one-off quotations in the past to a more structural model based on quarterly adjustments and cumulative demand.\n\n- The market expects high-end substrate prices in 2026 to rise by at least 15–20%, and some specialty products are even rumored to see increases of over 40%. The first round of price adjustments is anticipated to begin in the late part of 1Q.\n\n- With high-end ABF substrate supply tight at Japan’s Ibiden and Taiwan’s Nanya PCB and Unimicron, double-digit ABF price hike requests have already become the dominant trend.\n\n- As a spillover effect, some of Kinsus’ high-end production lines have reportedly been pre-booked in advance by major AI customers, leading to an operating model resembling a “quasi-packaging factory.” This suggests that demand intensity from AI and advanced computing platforms is far outpacing that of mainstream consumer applications.\n\n- However, Kinsus noted that the kind of long-term contracts or “blanket capacity lock-ups” seen during the peak shortage period in 2022 no longer exist, and at present there are no cases where customers are formally tying up capacity through contract structures.\nKinsus further explained that the current cooperation model has shifted toward preparing and allocating capacity based on annual demand forecasts.\n\n- Industry players view this round of price adjustments as somewhat different from the broad-based substrate super-cycle driven by the 2022 economic cycle.\n\n- This time, the key driver is structural change caused by rapid product specification upgrades as AI platform generations advance.\n\n- In other words, as AI chips see significant increases in pin count and packaging complexity, high-end substrates are simultaneously being pushed to higher layer counts, tighter line/space, and more intensive material usage, which in turn raises the barrier to achieving acceptable process yields.\n\n- According to industry sources, shortages of the key material—Low CTE fiberglass cloth—are ongoing, and lead times for Nittobo’s high-end fiberglass products have lengthened noticeably.\n\n- In addition, Nanya Plastics has officially issued price increase notices to customers: standard E-glass products were raised by roughly 15%, while ultra-thin fiberglass products below the 1037 spec posted an increase of around 25%.\n\n- Considering these raw-material price trends, industry participants expect high-end substrate prices in 2026 to be adjusted step-by-step starting from 1Q, with quarterly increases of at least 3–5%, and an annual increase of at least ~20%.\n\n- That said, the magnitude of increases may be negotiated in stages depending on demand strength, material supply conditions, and capacity allocation.\n\n- The industry also notes that as AI servers move toward more modular designs and demand for high-speed transmission rises, some applications have begun adopting high-spec BT substrates as core supporting structures—an additional growth driver expected to underpin the 2026 substrate market.\n\n---\n\nhttps://t.co/jES1ZpYbaJ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.020821407000133796, "ret_signed_m1d": 0.020821407000133796, "alpha_spy_m1d": 0.02673343048544985, "alpha_c_m1d": 0.034663903222885595, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.013842496222751799, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029948670782696363, "ret_signed_p1d": 0.029948670782696363, "alpha_spy_p1d": 0.03317202757492377, "alpha_c_p1d": 0.030766887917354935, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.000818217134658572, "ret_p1w": 0.03679403141757187, "ret_signed_p1w": 0.03679403141757187, "alpha_spy_p1w": 0.03396081084923619, "alpha_c_p1w": 0.037953224447537015, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": -0.0011591930299651443, "ret_p1m": 0.026811141156673157, "ret_signed_p1m": 0.026811141156673157, "alpha_spy_p1m": 0.03493473277375159, "alpha_c_p1m": 0.0849768575253601, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.05816571636868695, "ret_p3m": 0.17297774617730655, "ret_signed_p3m": 0.17297774617730655, "alpha_spy_p3m": 0.22239721286020386, "alpha_c_p3m": 0.24451475987645332, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": -0.07153701369914678, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.263, -0.2905, -0.2905, -0.3028, -0.272, -0.2548, -0.2827, -0.2723, -0.2869, -0.2752, -0.3108, -0.2506, -0.226, -0.099, -0.032, -0.0306, 0.0421, 0.0421, 0.0236, -0.0374, -0.0121, -0.0859, -0.0705, -0.0463, -0.0227, -0.0288, -0.1112, -0.0541, -0.1343, -0.17, -0.1062, -0.1921, -0.1921, -0.1921, -0.2224, -0.1835, -0.1529, -0.1928, -0.2035, -0.1582, -0.1522, -0.0941, -0.0916, -0.078, -0.1008, -0.0958, -0.0905, -0.1522, -0.1889, -0.1686, -0.2046, -0.1693, -0.1265, -0.1033, -0.0538, -0.0777, -0.0559, -0.055, -0.0401, -0.0401, -0.0401, -0.0401, 0.0208, 0.0, 0.0299, -0.023, -0.0425, -0.0425, 0.0368, 0.0605, 0.067, 0.1174, 0.1168, 0.1312, 0.1861, 0.2188, 0.1814, 0.1296, 0.1409, 0.1766, 0.1776, 0.1736, 0.1024, 0.1967, 0.0268, 0.0147, 0.0104, 0.1115, 0.1292, 0.1292, 0.2128, 0.2322, 0.2521, 0.2757, 0.3311, 0.3745, 0.3059, 0.3059, 0.4096, 0.3765, 0.3262, 0.3601, 0.336, 0.229, 0.1338, 0.155, 0.1238, 0.0322, 0.1171, 0.112, 0.0933, 0.1201, 0.1622, 0.2121, 0.2095, 0.1791, 0.1791, 0.0753, 0.1174, 0.2091, 0.2065, 0.1793, 0.111, 0.0533, 0.153, 0.1108, 0.173, 0.2221, 0.2456, 0.3993, 0.4016, 0.4127, 0.341, 0.4166, 0.3401, 0.3676, 0.3194, 0.368, 0.5087, 0.6002, 0.5909, 0.7916, 0.8102, 0.7945, 0.7945, 0.9259, 0.9109, 0.9109, 0.9109, 0.9109, 1.3395, 1.2167, 1.2367, 1.3645, 1.4374, 1.4231, 1.2374, 1.2338, 1.1888, 1.2345, 1.5538, 1.7503, 1.9903, 1.9953, 1.9182, 1.8203, 2.2861, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008328141354471930", "ticker_idx": 1, "ts": "2026-01-06T00:01:56+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "high-end ABF substrate supply tight at Japan's Ibiden and Taiwan's Nanya PCB and Unimicron, double-digit ABF price hike requests have already become the dominant trend", "tweet_type": "analysis", "timeline": "quarters", "summary": "AI-driven structural tightness in high-end ABF substrates expected to drive 15–20%+ price hikes in 2026; Ibiden, Unimicron, and Kinsus are key beneficiaries.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ AI Platform Push Drives High-End Substrate Price Hikes \n\n- As demand for AI servers and high-performance computing (HPC) continues to expand, supply-demand conditions for high-end IC substrates remain persistently tight.\n\n- In particular, shortages of key materials such as low-coefficient-of-thermal-expansion (Low CTE) fiberglass are driving a shift in pricing negotiations—from one-off quotations in the past to a more structural model based on quarterly adjustments and cumulative demand.\n\n- The market expects high-end substrate prices in 2026 to rise by at least 15–20%, and some specialty products are even rumored to see increases of over 40%. The first round of price adjustments is anticipated to begin in the late part of 1Q.\n\n- With high-end ABF substrate supply tight at Japan’s Ibiden and Taiwan’s Nanya PCB and Unimicron, double-digit ABF price hike requests have already become the dominant trend.\n\n- As a spillover effect, some of Kinsus’ high-end production lines have reportedly been pre-booked in advance by major AI customers, leading to an operating model resembling a “quasi-packaging factory.” This suggests that demand intensity from AI and advanced computing platforms is far outpacing that of mainstream consumer applications.\n\n- However, Kinsus noted that the kind of long-term contracts or “blanket capacity lock-ups” seen during the peak shortage period in 2022 no longer exist, and at present there are no cases where customers are formally tying up capacity through contract structures.\nKinsus further explained that the current cooperation model has shifted toward preparing and allocating capacity based on annual demand forecasts.\n\n- Industry players view this round of price adjustments as somewhat different from the broad-based substrate super-cycle driven by the 2022 economic cycle.\n\n- This time, the key driver is structural change caused by rapid product specification upgrades as AI platform generations advance.\n\n- In other words, as AI chips see significant increases in pin count and packaging complexity, high-end substrates are simultaneously being pushed to higher layer counts, tighter line/space, and more intensive material usage, which in turn raises the barrier to achieving acceptable process yields.\n\n- According to industry sources, shortages of the key material—Low CTE fiberglass cloth—are ongoing, and lead times for Nittobo’s high-end fiberglass products have lengthened noticeably.\n\n- In addition, Nanya Plastics has officially issued price increase notices to customers: standard E-glass products were raised by roughly 15%, while ultra-thin fiberglass products below the 1037 spec posted an increase of around 25%.\n\n- Considering these raw-material price trends, industry participants expect high-end substrate prices in 2026 to be adjusted step-by-step starting from 1Q, with quarterly increases of at least 3–5%, and an annual increase of at least ~20%.\n\n- That said, the magnitude of increases may be negotiated in stages depending on demand strength, material supply conditions, and capacity allocation.\n\n- The industry also notes that as AI servers move toward more modular designs and demand for high-speed transmission rises, some applications have begun adopting high-spec BT substrates as core supporting structures—an additional growth driver expected to underpin the 2026 substrate market.\n\n---\n\nhttps://t.co/jES1ZpYbaJ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002169197396963085, "ret_signed_m1d": -0.002169197396963085, "alpha_spy_m1d": 0.0037428260883529685, "alpha_c_m1d": 0.011673298825788714, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.013842496222751799, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0195227765726681, "ret_signed_p1d": -0.0195227765726681, "alpha_spy_p1d": -0.016299419780440694, "alpha_c_p1d": -0.018704559438009527, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.000818217134658572, "ret_p1w": 0.10195227765726678, "ret_signed_p1w": 0.10195227765726678, "alpha_spy_p1w": 0.0991190570889311, "alpha_c_p1w": 0.10311147068723192, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": -0.0011591930299651443, "ret_p1m": 0.683297180043384, "ret_signed_p1m": 0.683297180043384, "alpha_spy_p1m": 0.6914207716604625, "alpha_c_p1m": 0.741462896412071, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.05816571636868695, "ret_p3m": 1.2516268980477223, "ret_signed_p3m": 1.2516268980477223, "alpha_spy_p3m": 1.3010463647306196, "alpha_c_p3m": 1.3231639117468692, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": -0.07153701369914678, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3015, -0.3015, -0.3341, -0.3666, -0.3124, -0.282, -0.3037, -0.3059, -0.3059, -0.3102, -0.3232, -0.3232, -0.2907, -0.295, -0.2798, -0.3059, -0.2907, -0.2256, -0.269, -0.294, -0.2875, -0.3048, -0.2918, -0.2832, -0.2789, -0.2316, -0.256, -0.2733, -0.2625, -0.3037, -0.256, -0.3059, -0.2885, -0.2364, -0.2256, -0.1887, -0.1909, -0.1453, -0.1497, -0.1475, -0.1388, -0.0672, -0.0434, -0.0195, -0.013, -0.0239, -0.0239, -0.039, -0.0716, -0.0803, -0.102, -0.0672, -0.0434, -0.0651, -0.0586, -0.0586, -0.0434, -0.0434, -0.0477, -0.0456, -0.0456, -0.0499, -0.0022, 0.0, -0.0195, -0.013, -0.0217, 0.0564, 0.102, 0.0846, 0.0824, 0.1887, 0.3059, 0.3362, 0.3926, 0.4187, 0.4707, 0.4447, 0.5358, 0.5944, 0.4946, 0.6421, 0.5835, 0.6703, 0.6833, 0.5857, 0.4751, 0.5076, 0.6117, 0.6009, 0.6009, 0.6009, 0.6009, 0.6009, 0.6009, 0.6009, 0.6009, 0.7592, 0.9349, 0.9892, 1.0889, 1.0889, 1.0868, 1.0087, 0.8091, 0.9523, 0.8742, 0.6876, 0.8395, 1.0217, 1.0824, 1.2039, 1.3818, 1.3557, 1.5163, 1.5206, 1.4035, 1.1649, 0.9957, 1.1952, 1.1779, 1.1909, 1.141, 0.9284, 1.1193, 1.2516, 1.2516, 1.2516, 1.4469, 1.6898, 1.7115, 1.7679, 1.7158, 1.59, 1.6811, 1.7983, 1.7896, 1.9414, 2.0803, 2.1063, 2.1584, 2.4273, 2.6399, 2.5792, 2.4837, 2.8308, 2.8308, 2.9523, 2.9176, 2.7223, 2.8872, 2.5488, 2.7354, 2.7961, 2.8698, 2.8221, 2.5618, 2.5401, 2.5488, 2.5705, 2.9262, 3.2082, 3.295, 3.7072, 3.6855, 3.4469, 3.577, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008328141354471930", "ticker_idx": 2, "ts": "2026-01-06T00:01:56+00:00", "symbol": "Kinsus", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "some of Kinsus' high-end production lines have reportedly been pre-booked in advance by major AI customers, leading to an operating model resembling a 'quasi-packaging factory'", "tweet_type": "analysis", "timeline": "quarters", "summary": "AI-driven structural tightness in high-end ABF substrates expected to drive 15–20%+ price hikes in 2026; Ibiden, Unimicron, and Kinsus are key beneficiaries.", "resolved_tickers": ["3189.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Kinsus Interconnect Technology 3189.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ AI Platform Push Drives High-End Substrate Price Hikes \n\n- As demand for AI servers and high-performance computing (HPC) continues to expand, supply-demand conditions for high-end IC substrates remain persistently tight.\n\n- In particular, shortages of key materials such as low-coefficient-of-thermal-expansion (Low CTE) fiberglass are driving a shift in pricing negotiations—from one-off quotations in the past to a more structural model based on quarterly adjustments and cumulative demand.\n\n- The market expects high-end substrate prices in 2026 to rise by at least 15–20%, and some specialty products are even rumored to see increases of over 40%. The first round of price adjustments is anticipated to begin in the late part of 1Q.\n\n- With high-end ABF substrate supply tight at Japan’s Ibiden and Taiwan’s Nanya PCB and Unimicron, double-digit ABF price hike requests have already become the dominant trend.\n\n- As a spillover effect, some of Kinsus’ high-end production lines have reportedly been pre-booked in advance by major AI customers, leading to an operating model resembling a “quasi-packaging factory.” This suggests that demand intensity from AI and advanced computing platforms is far outpacing that of mainstream consumer applications.\n\n- However, Kinsus noted that the kind of long-term contracts or “blanket capacity lock-ups” seen during the peak shortage period in 2022 no longer exist, and at present there are no cases where customers are formally tying up capacity through contract structures.\nKinsus further explained that the current cooperation model has shifted toward preparing and allocating capacity based on annual demand forecasts.\n\n- Industry players view this round of price adjustments as somewhat different from the broad-based substrate super-cycle driven by the 2022 economic cycle.\n\n- This time, the key driver is structural change caused by rapid product specification upgrades as AI platform generations advance.\n\n- In other words, as AI chips see significant increases in pin count and packaging complexity, high-end substrates are simultaneously being pushed to higher layer counts, tighter line/space, and more intensive material usage, which in turn raises the barrier to achieving acceptable process yields.\n\n- According to industry sources, shortages of the key material—Low CTE fiberglass cloth—are ongoing, and lead times for Nittobo’s high-end fiberglass products have lengthened noticeably.\n\n- In addition, Nanya Plastics has officially issued price increase notices to customers: standard E-glass products were raised by roughly 15%, while ultra-thin fiberglass products below the 1037 spec posted an increase of around 25%.\n\n- Considering these raw-material price trends, industry participants expect high-end substrate prices in 2026 to be adjusted step-by-step starting from 1Q, with quarterly increases of at least 3–5%, and an annual increase of at least ~20%.\n\n- That said, the magnitude of increases may be negotiated in stages depending on demand strength, material supply conditions, and capacity allocation.\n\n- The industry also notes that as AI servers move toward more modular designs and demand for high-speed transmission rises, some applications have begun adopting high-spec BT substrates as core supporting structures—an additional growth driver expected to underpin the 2026 substrate market.\n\n---\n\nhttps://t.co/jES1ZpYbaJ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008849557522123908, "ret_signed_m1d": -0.008849557522123908, "alpha_spy_m1d": -0.002937534036807854, "alpha_c_m1d": 0.01702629745348061, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.025875854975604518, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.032448377581120846, "ret_signed_p1d": 0.032448377581120846, "alpha_spy_p1d": 0.03567173437334825, "alpha_c_p1d": 0.03915599363877431, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.006707616057653465, "ret_p1w": 0.041297935103244754, "ret_signed_p1w": 0.041297935103244754, "alpha_spy_p1w": 0.03846471453490907, "alpha_c_p1w": 0.03121072672554659, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": 0.010087208377698165, "ret_p1m": 0.5958702064896755, "ret_signed_p1m": 0.5958702064896755, "alpha_spy_p1m": 0.6039937981067539, "alpha_c_p1m": 0.6103173613683153, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.01444715487863979, "ret_p3m": 0.9823008849557522, "ret_signed_p3m": 0.9823008849557522, "alpha_spy_p3m": 1.0317203516386495, "alpha_c_p3m": 0.9701755759772761, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": 0.01212530897847608, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1888, -0.1888, -0.2124, -0.2773, -0.2065, -0.1298, -0.1091, -0.1091, -0.115, -0.1357, -0.1445, -0.1445, -0.1239, -0.1888, -0.1799, -0.1888, -0.1947, -0.1799, -0.2065, -0.2183, -0.233, -0.2448, -0.2448, -0.2448, -0.2242, -0.1475, -0.1917, -0.1652, -0.1416, -0.2065, -0.177, -0.2389, -0.2212, -0.1888, -0.1858, -0.1681, -0.1593, -0.1652, -0.1681, -0.1268, -0.1268, -0.0737, -0.0796, -0.059, -0.1003, -0.1062, -0.1062, -0.1386, -0.1829, -0.1976, -0.2212, -0.2035, -0.1504, -0.1593, -0.1504, -0.1504, -0.0973, -0.1121, -0.1268, -0.0619, -0.0619, -0.0796, -0.0088, 0.0, 0.0324, -0.0206, -0.0059, 0.0678, 0.0413, 0.0147, 0.0324, 0.1357, 0.2063, 0.1778, 0.1407, 0.1721, 0.2206, 0.2271, 0.3481, 0.4366, 0.354, 0.4867, 0.413, 0.4513, 0.5959, 0.5546, 0.4307, 0.5044, 0.4956, 0.5664, 0.5664, 0.5664, 0.5664, 0.5664, 0.5664, 0.5664, 0.5664, 0.7227, 0.8938, 0.9204, 0.8555, 0.8555, 0.7965, 0.823, 0.6785, 0.7552, 0.7198, 0.5605, 0.6165, 0.7139, 0.705, 0.8732, 0.9882, 1.0147, 1.1563, 1.2625, 1.2419, 1.0413, 0.8909, 1.0737, 1.0737, 1.0767, 1.0442, 0.8466, 0.9587, 0.9823, 0.9823, 0.9823, 1.115, 1.3245, 1.2832, 1.2478, 1.1622, 1.1888, 1.2625, 1.3215, 1.3097, 1.5398, 1.7935, 1.8997, 1.826, 2.1032, 2.0383, 1.9971, 1.9971, 2.115, 2.115, 2.1327, 2.1091, 1.8466, 1.9056, 1.7021, 1.9676, 1.9145, 1.9351, 2.0796, 1.8673, 1.9794, 2.0265, 1.9735, 2.2684, 2.5929, 2.9469, 3.1062, 3.1416, 3.0177, 3.3009, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008869191059984596", "ticker_idx": 0, "ts": "2026-01-07T11:51:52+00:00", "symbol": "STX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The HDD supply shortage is now widening to 200EB over the next 12 months… STX is raising consumer/client HDD prices by 10% every quarter to better align the margin profile with nearline drives", "tweet_type": "analysis", "timeline": "months", "summary": "Morgan Stanley takeaways bullish on HDD names: widening supply shortage, rising prices, strong nearline demand, and constrained supply support STX and WDC.", "resolved_tickers": ["STX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s Key Takeaways on HDD/Storage:\n\n• The HDD supply shortage is now widening to 200EB over the next 12 months (vs. 100-150EB 3 months ago). Hyperscalers are using/qualifying eSSDs to help fulfill / backstop storage demand despite being far less efficient in terms of TCO (total cost of ownership) vs. using HDDs. Our checks suggest this is still a temporary measure and not a re-architecting towards eSSD.\n\n• Memory makers have been proactively trying to find opportunities to lower the price per GB for eSSDs to try and replace HDDs as early as CY28. However, since NAND prices per GB are now well above $0.10 vs. overall\nHDD at $0.013 on average — implying up to 10x of acquisition cost difference by the end of C4Q25 — introducing high-layer (400+ layers) NAND for eSSDs to replace HDDs no longer is a key focus for memory makers.\n\n• CSPs' SSD demand fulfillment rate is now at just 40%, likely because of both increasing customer demand and strengthening forward-looking forecasts amidst limited supply upside from SSDs and HDDs. To be clear, we have not seen over-shipping so far as current SSD/HDD industry supply cannot fulfill the full demand.\n\n• HDD makers remain reluctant to expand unit capacity, but they are shifting capacity allocation from client/consumer applications to support expanding\ncloud/nearline demand.\n\n• STX is raising consumer/client HDD prices by 10% every quarter to better align the margin profile with nearline drives. For what it's worth, STX mentioned on its F1Q26 earnings call that it has been \"taking some supply out of Edge IoT products and putting it into Cloud as we can pivot demand.\"\n\n$STX $WDC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07188732059136349, "ret_signed_m1d": 0.07188732059136349, "alpha_spy_m1d": 0.06865354017105374, "alpha_c_m1d": 0.07106843342919755, "bench_spy_m1d": 0.003233780420309751, "bench_c_m1d": 0.0008188871621659377, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07717518018341896, "ret_signed_p1d": -0.07717518018341896, "alpha_spy_p1d": -0.0770736475014332, "alpha_c_p1d": -0.061546988736483677, "bench_spy_p1d": -0.00010153268198576093, "bench_c_p1d": -0.01562819144693528, "ret_p1w": 0.013040841574953, "ret_signed_p1w": 0.013040841574953, "alpha_spy_p1w": 0.011909781698006272, "alpha_c_p1w": 0.025529757039502554, "bench_spy_p1w": 0.0011310598769467273, "bench_c_p1w": -0.012488915464549555, "ret_p1m": 0.31528575218364696, "ret_signed_p1m": 0.31528575218364696, "alpha_spy_p1m": 0.3326296465726173, "alpha_c_p1m": 0.38967321593245896, "bench_spy_p1m": -0.017343894388970327, "bench_c_p1m": -0.074387463748812, "ret_p3m": 0.47307689065525516, "ret_signed_p3m": 0.47307689065525516, "alpha_spy_p3m": 0.5149146570498545, "alpha_c_p3m": 0.5384555529831614, "bench_spy_p3m": -0.04183776639459935, "bench_c_p3m": -0.06537866232790623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3064, -0.2898, -0.3153, -0.2902, -0.2687, -0.2707, -0.3063, -0.3058, -0.3042, -0.2674, -0.2425, -0.2548, -0.2785, -0.1406, -0.1318, -0.1721, -0.1408, -0.1899, -0.1077, -0.099, -0.0962, -0.0488, -0.0682, -0.0835, -0.1505, -0.1646, -0.1543, -0.1786, -0.1615, -0.2219, -0.2316, -0.1802, -0.1526, -0.119, -0.119, -0.1048, -0.1261, -0.1365, -0.1631, -0.1405, -0.098, -0.0765, -0.0848, -0.0328, -0.0039, -0.0693, -0.076, -0.0677, -0.1017, -0.0552, -0.0411, -0.0848, -0.085, -0.0746, -0.0746, -0.0715, -0.0875, -0.0914, -0.1066, -0.1066, -0.0672, -0.0598, 0.0719, 0.0, -0.0772, -0.0138, 0.0429, 0.033, 0.013, 0.0391, 0.0583, 0.0583, 0.0575, 0.1167, 0.1241, 0.1228, 0.1623, 0.206, 0.4369, 0.4487, 0.3226, 0.4045, 0.4418, 0.358, 0.3153, 0.3927, 0.3787, 0.2854, 0.3211, 0.3987, 0.3819, 0.3819, 0.3493, 0.3759, 0.3267, 0.3336, 0.3216, 0.2847, 0.3685, 0.329, 0.323, 0.2312, 0.1601, 0.2165, 0.1917, 0.1445, 0.2143, 0.2466, 0.2521, 0.2132, 0.2448, 0.2936, 0.366, 0.3196, 0.4098, 0.334, 0.3106, 0.3786, 0.3428, 0.2309, 0.2351, 0.1778, 0.2731, 0.375, 0.3953, 0.3953, 0.4731, 0.5232, 0.6128, 0.6273, 0.635, 0.668, 0.7335, 0.6885, 0.7282, 0.78, 0.754, 0.8195, 0.8844, 0.9096, 0.9051, 0.9364, 0.8817, 1.0905, 1.1891, 1.3623, 1.4, 1.5055, 1.5556, 1.4907, 1.5433, 1.7103, 1.6283, 1.6561, 1.6152, 1.585, 1.4075, 1.3831, 1.4407, 1.6337, 1.6411, 1.6411, 1.7484, 1.8294, 1.8621, 1.8591, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008869191059984596", "ticker_idx": 1, "ts": "2026-01-07T11:51:52+00:00", "symbol": "WDC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HDD makers remain reluctant to expand unit capacity, but they are shifting capacity allocation from client/consumer applications to support expanding cloud/nearline demand", "tweet_type": "analysis", "timeline": "months", "summary": "Morgan Stanley takeaways bullish on HDD names: widening supply shortage, rising prices, strong nearline demand, and constrained supply support STX and WDC.", "resolved_tickers": ["WDC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s Key Takeaways on HDD/Storage:\n\n• The HDD supply shortage is now widening to 200EB over the next 12 months (vs. 100-150EB 3 months ago). Hyperscalers are using/qualifying eSSDs to help fulfill / backstop storage demand despite being far less efficient in terms of TCO (total cost of ownership) vs. using HDDs. Our checks suggest this is still a temporary measure and not a re-architecting towards eSSD.\n\n• Memory makers have been proactively trying to find opportunities to lower the price per GB for eSSDs to try and replace HDDs as early as CY28. However, since NAND prices per GB are now well above $0.10 vs. overall\nHDD at $0.013 on average — implying up to 10x of acquisition cost difference by the end of C4Q25 — introducing high-layer (400+ layers) NAND for eSSDs to replace HDDs no longer is a key focus for memory makers.\n\n• CSPs' SSD demand fulfillment rate is now at just 40%, likely because of both increasing customer demand and strengthening forward-looking forecasts amidst limited supply upside from SSDs and HDDs. To be clear, we have not seen over-shipping so far as current SSD/HDD industry supply cannot fulfill the full demand.\n\n• HDD makers remain reluctant to expand unit capacity, but they are shifting capacity allocation from client/consumer applications to support expanding\ncloud/nearline demand.\n\n• STX is raising consumer/client HDD prices by 10% every quarter to better align the margin profile with nearline drives. For what it's worth, STX mentioned on its F1Q26 earnings call that it has been \"taking some supply out of Edge IoT products and putting it into Cloud as we can pivot demand.\"\n\n$STX $WDC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.09755856323677814, "ret_signed_m1d": 0.09755856323677814, "alpha_spy_m1d": 0.09432478281646839, "alpha_c_m1d": 0.0967396760746122, "bench_spy_m1d": 0.003233780420309751, "bench_c_m1d": 0.0008188871621659377, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06103664465527525, "ret_signed_p1d": -0.06103664465527525, "alpha_spy_p1d": -0.06093511197328949, "alpha_c_p1d": -0.04540845320833997, "bench_spy_p1d": -0.00010153268198576093, "bench_c_p1d": -0.01562819144693528, "ret_p1w": 0.07564536626210394, "ret_signed_p1w": 0.07564536626210394, "alpha_spy_p1w": 0.07451430638515721, "alpha_c_p1w": 0.0881342817266535, "bench_spy_p1w": 0.0011310598769467273, "bench_c_p1w": -0.012488915464549555, "ret_p1m": 0.3017310150889756, "ret_signed_p1m": 0.3017310150889756, "alpha_spy_p1m": 0.3190749094779459, "alpha_c_p1m": 0.3761184788377876, "bench_spy_p1m": -0.017343894388970327, "bench_c_p1m": -0.074387463748812, "ret_p3m": 0.5223912743452279, "ret_signed_p3m": 0.5223912743452279, "alpha_spy_p3m": 0.5642290407398273, "alpha_c_p3m": 0.5877699366731342, "bench_spy_p3m": -0.04183776639459935, "bench_c_p3m": -0.06537866232790623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.423, -0.4058, -0.4345, -0.3979, -0.3705, -0.3691, -0.3925, -0.3931, -0.3978, -0.3715, -0.353, -0.3668, -0.3755, -0.2932, -0.3095, -0.2491, -0.2101, -0.2393, -0.1997, -0.1822, -0.1854, -0.1291, -0.1502, -0.1696, -0.2144, -0.211, -0.1879, -0.2359, -0.2303, -0.299, -0.3042, -0.2455, -0.2231, -0.2115, -0.2115, -0.1835, -0.1825, -0.2002, -0.2222, -0.1945, -0.155, -0.1506, -0.1518, -0.0897, -0.0634, -0.1178, -0.1393, -0.1266, -0.1682, -0.1244, -0.0941, -0.1157, -0.1082, -0.1017, -0.1017, -0.0918, -0.1011, -0.1192, -0.1381, -0.1381, -0.0609, -0.06, 0.0976, 0.0, -0.061, 0.0029, 0.0613, 0.0706, 0.0756, 0.1112, 0.1082, 0.1082, 0.1155, 0.2102, 0.2172, 0.1827, 0.205, 0.2641, 0.3993, 0.3929, 0.2519, 0.352, 0.4521, 0.3479, 0.3017, 0.4137, 0.4308, 0.3136, 0.3695, 0.4214, 0.4087, 0.4087, 0.4214, 0.4837, 0.4242, 0.4285, 0.4029, 0.3537, 0.4556, 0.4121, 0.3993, 0.3512, 0.2538, 0.3073, 0.2965, 0.2276, 0.3117, 0.3325, 0.3455, 0.3073, 0.3629, 0.4326, 0.5707, 0.5261, 0.5864, 0.4671, 0.4755, 0.5069, 0.4823, 0.3682, 0.3782, 0.2597, 0.3539, 0.4903, 0.4764, 0.4764, 0.5224, 0.5615, 0.6957, 0.6912, 0.719, 0.7527, 0.8331, 0.827, 0.8104, 0.8646, 0.8726, 0.9211, 0.9476, 1.0178, 1.0222, 1.0058, 0.9571, 1.066, 1.1749, 1.1599, 1.2142, 1.3288, 1.4184, 1.3221, 1.4026, 1.5819, 1.4463, 1.4731, 1.4484, 1.4127, 1.2959, 1.2815, 1.3006, 1.4349, 1.424, 1.424, 1.6261, 1.6559, 1.6588, 1.6589, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2028400638686838927", "ticker_idx": 0, "ts": "2026-03-02T09:22:52+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I believe the market is currently not pricing in the normalization of Samsung Foundry into Samsung's stock price", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Samsung Electronics: foundry profitability normalization not yet priced into the stock.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "I believe the market is currently not pricing in the normalization of Samsung Foundry into Samsung's stock price. https://t.co/eYQjaTkAQw", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Foundry \"Targeting Profitability This Year\"\n\nSamsung Electronics has set a goal of turning its semiconductor foundry business profitable this year. It has been confirmed that the company moved up the timeline, which was originally scheduled to break even next year. After https://t.co/WgHc13YbSp", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0005682556842063757, "alpha_c_m1d": 0.0055898388543234034, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -0.0055898388543234034, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09884526208146449, "ret_signed_p1d": -0.09884526208146449, "alpha_spy_p1d": -0.09003092359866882, "alpha_c_p1d": -0.084225843558279, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.01461941852318549, "ret_p1w": -0.1986143434258696, "ret_signed_p1w": -0.1986143434258696, "alpha_spy_p1w": -0.18679876250235472, "alpha_c_p1w": -0.2001909506044468, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": 0.0015766071785772162, "ret_p1m": -0.2261151425649689, "ret_signed_p1m": -0.2261151425649689, "alpha_spy_p1m": -0.17619540098978514, "alpha_c_p1m": -0.17971438894673752, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.04640075361823137, "ret_p3m": 0.3862351363743528, "ret_signed_p3m": 0.3862351363743528, "alpha_spy_p3m": 0.28384175720844484, "alpha_c_p3m": 0.04552758791859035, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.34070754845576245, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5197, -0.5169, -0.5017, -0.4967, -0.5017, -0.5036, -0.5068, -0.4994, -0.5183, -0.5275, -0.504, -0.5054, -0.5114, -0.4921, -0.4875, -0.4893, -0.4893, -0.4622, -0.448, -0.4462, -0.4462, -0.4462, -0.4065, -0.3621, -0.3584, -0.3487, -0.3589, -0.358, -0.3589, -0.3644, -0.352, -0.3353, -0.3122, -0.3104, -0.3293, -0.3095, -0.2965, -0.2975, -0.2975, -0.2633, -0.2499, -0.2577, -0.2587, -0.3053, -0.2263, -0.2189, -0.2642, -0.2674, -0.2314, -0.2342, -0.2249, -0.1751, -0.163, -0.163, -0.163, -0.163, -0.1224, -0.1219, -0.1085, -0.0762, -0.06, 0.0069, 0.0, 0.0, -0.0988, -0.2046, -0.115, -0.1307, -0.1986, -0.1321, -0.1224, -0.1321, -0.1524, -0.1284, -0.1044, -0.037, -0.0739, -0.079, -0.1395, -0.1238, -0.127, -0.1681, -0.1681, -0.184, -0.2261, -0.1222, -0.1743, -0.1382, -0.1062, -0.0905, -0.0257, -0.0558, -0.0465, -0.0697, -0.0442, -0.0234, 0.0067, -0.0002, -0.0072, 0.0136, 0.0067, 0.0391, 0.016, 0.0391, 0.0275, 0.046, 0.0206, 0.0206, 0.0761, 0.0761, 0.2312, 0.2566, 0.2428, 0.3214, 0.2914, 0.3145, 0.37, 0.252, 0.3006, 0.2752, 0.2775, 0.3862, 0.3538, 0.3538, 0.3839, 0.4209, 0.3862, 0.4672, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2026593942180143489", "ticker_idx": 0, "ts": "2026-02-25T09:43:42+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung could become the most profitable company in the world in 2027, and yet its market cap is still under $1 trillion!", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish Samsung: potential to be world's most profitable company by 2027 yet trades under $1T market cap — implying deep undervaluation.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung could become the most profitable company in the world in 2027, and yet its market cap is still under $1 trillion! https://t.co/uAMgYrJlEa", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017198993190633782, "ret_signed_m1d": -0.017198993190633782, "alpha_spy_m1d": -0.008831254148954915, "alpha_c_m1d": 0.001610783551699857, "bench_spy_m1d": -0.008367739041678868, "bench_c_m1d": -0.01880977674233364, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07125313666173794, "ret_signed_p1d": 0.07125313666173794, "alpha_spy_p1d": 0.07680761305335904, "alpha_c_p1d": 0.08523814883608793, "bench_spy_p1d": -0.005554476391621099, "bench_c_p1d": -0.013985012174349998, "ret_p1w": -0.15380833475288103, "ret_signed_p1w": -0.15380833475288103, "alpha_spy_p1w": -0.1422378627972123, "alpha_c_p1w": -0.1316420653514796, "bench_spy_p1w": -0.011570471955668737, "bench_c_p1w": -0.02216626940140143, "ret_p1m": -0.11498766878575506, "ret_signed_p1m": -0.11498766878575506, "alpha_spy_p1m": -0.04819365096424566, "alpha_c_p1m": -0.04264835610451789, "bench_spy_p1m": -0.0667940178215094, "bench_c_p1m": -0.07233931268123717, "ret_p3m": 0.4403215240817604, "ret_signed_p3m": 0.4403215240817604, "alpha_spy_p3m": 0.361657007109651, "alpha_c_p3m": 0.17737276318189266, "bench_spy_p3m": 0.07866451697210941, "bench_c_p3m": 0.26294876089986774, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5085, -0.5071, -0.4943, -0.489, -0.486, -0.4699, -0.4645, -0.4699, -0.4719, -0.4753, -0.4675, -0.4875, -0.4973, -0.4723, -0.4738, -0.4802, -0.4596, -0.4547, -0.4567, -0.4567, -0.4278, -0.4128, -0.4108, -0.4108, -0.4108, -0.3686, -0.3214, -0.3174, -0.3071, -0.3179, -0.317, -0.3179, -0.3238, -0.3106, -0.2929, -0.2683, -0.2663, -0.2865, -0.2654, -0.2516, -0.2526, -0.2526, -0.2162, -0.202, -0.2103, -0.2113, -0.2609, -0.1769, -0.169, -0.2172, -0.2206, -0.1823, -0.1853, -0.1754, -0.1224, -0.1096, -0.1096, -0.1096, -0.1096, -0.0663, -0.0658, -0.0516, -0.0172, 0.0, 0.0713, 0.0639, 0.0639, -0.0413, -0.1538, -0.0585, -0.0752, -0.1474, -0.0767, -0.0663, -0.0767, -0.0983, -0.0727, -0.0472, 0.0246, -0.0147, -0.0201, -0.0845, -0.0678, -0.0713, -0.115, -0.115, -0.1319, -0.1767, -0.0661, -0.1215, -0.0831, -0.0491, -0.0324, 0.0365, 0.0045, 0.0144, -0.0102, 0.0168, 0.039, 0.071, 0.0636, 0.0562, 0.0784, 0.071, 0.1055, 0.0809, 0.1055, 0.0932, 0.1129, 0.0858, 0.0858, 0.1449, 0.1449, 0.3098, 0.3369, 0.3221, 0.4059, 0.3738, 0.3985, 0.4576, 0.332, 0.3837, 0.3566, 0.3591, 0.4748, 0.4403, 0.4403, 0.4723, 0.5117, 0.4748, 0.561, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2034726127030476973", "ticker_idx": 0, "ts": "2026-03-19T20:18:06+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a development that is positive for ASML", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Micron's EUV pivot benefits ASML but is modestly negative for AMAT and LRCX as multi-patterning steps decline.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Micron's strategic pivot toward more aggressive EUV adoption is interpreted as an effort to save fab space and thereby expand overall output capacity.\n\nThis implies a departure from its conventional multi-patterning approach — a development that is positive for ASML, but modestly negative for AMAT and LRCX.", "reply_text": null, "parent_text": null, "quoted_text": "First 5-year SCA signed. New EUV layer roadmap. ✍️ https://t.co/52JZpZMK6l", "parent_author": null, "ret_m1d": 0.01916824137227091, "ret_signed_m1d": 0.01916824137227091, "alpha_spy_m1d": 0.016697812910204224, "alpha_c_m1d": 0.022333439279438316, "bench_spy_m1d": 0.002470428462066687, "bench_c_m1d": -0.0031651979071674052, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03457132238629401, "ret_signed_p1d": -0.03457132238629401, "alpha_spy_p1d": -0.02023554651378978, "alpha_c_p1d": -0.008793891901064388, "bench_spy_p1d": -0.01433577587250423, "bench_c_p1d": -0.025777430485229624, "ret_p1w": -0.002395925481964656, "ret_signed_p1w": -0.002395925481964656, "alpha_spy_p1w": 0.01722855046913363, "alpha_c_p1w": 0.033256907018559434, "bench_spy_p1w": -0.019624475951098286, "bench_c_p1w": -0.03565283250052409, "ret_p1m": 0.06452164718568265, "ret_signed_p1m": 0.06452164718568265, "alpha_spy_p1m": -0.014713591347992327, "alpha_c_p1m": -0.11080497056002714, "bench_spy_p1m": 0.07923523853367498, "bench_c_p1m": 0.1753266177457098, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2343, -0.2281, -0.2317, -0.2317, -0.2317, -0.2251, -0.2152, -0.2126, -0.2126, -0.1571, -0.1, -0.0933, -0.1012, -0.1343, -0.0757, -0.0716, -0.0577, -0.0735, -0.0177, -0.0025, -0.0425, -0.0258, -0.0131, 0.0053, 0.0069, 0.0067, 0.0405, 0.0207, 0.0187, 0.0388, 0.0467, 0.0173, -0.0253, -0.0176, 0.0202, 0.031, 0.0209, 0.0335, 0.0096, 0.0187, 0.0226, 0.0262, 0.0652, 0.0596, 0.0744, 0.069, 0.0811, 0.1025, 0.0546, 0.0555, 0.0358, -0.0058, 0.0267, 0.0149, -0.0185, -0.018, 0.0269, 0.0258, 0.0101, 0.0094, 0.0246, 0.0216, 0.0192, 0.0, -0.0346, 0.0058, 0.0282, 0.0368, -0.0024, -0.019, -0.0484, -0.0423, 0.0163, -0.0065, -0.0065, -0.0065, -0.0469, 0.0377, 0.0566, 0.0868, 0.078, 0.0989, 0.0525, 0.0462, 0.0645, 0.0654, 0.0585, 0.0695, 0.0472, 0.0717, 0.0399, 0.005, 0.0228, 0.0484, 0.0484, 0.0183, 0.054, 0.1192, 0.1149, 0.1315, 0.1197, 0.0856, 0.1381, 0.1724, 0.1206, 0.0844, 0.0712, 0.1432, 0.1537, 0.2084, 0.2269, 0.1919, 0.1804, 0.1929, 0.1876, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2034726127030476973", "ticker_idx": 1, "ts": "2026-03-19T20:18:06+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "modestly negative for AMAT", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Micron's EUV pivot benefits ASML but is modestly negative for AMAT and LRCX as multi-patterning steps decline.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Micron's strategic pivot toward more aggressive EUV adoption is interpreted as an effort to save fab space and thereby expand overall output capacity.\n\nThis implies a departure from its conventional multi-patterning approach — a development that is positive for ASML, but modestly negative for AMAT and LRCX.", "reply_text": null, "parent_text": null, "quoted_text": "First 5-year SCA signed. New EUV layer roadmap. ✍️ https://t.co/52JZpZMK6l", "parent_author": null, "ret_m1d": -0.021667891020111685, "ret_signed_m1d": 0.021667891020111685, "alpha_spy_m1d": 0.024138319482178372, "alpha_c_m1d": 0.01850269311294428, "bench_spy_m1d": 0.002470428462066687, "bench_c_m1d": -0.0031651979071674052, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.000419912743004458, "ret_signed_p1d": 0.000419912743004458, "alpha_spy_p1d": -0.013915863129499773, "alpha_c_p1d": -0.025357517742225166, "bench_spy_p1d": -0.01433577587250423, "bench_c_p1d": -0.025777430485229624, "ret_p1w": -0.05223820591788697, "ret_signed_p1w": 0.05223820591788697, "alpha_spy_p1w": 0.032613729966788685, "alpha_c_p1w": 0.01658537341736288, "bench_spy_p1w": -0.019624475951098286, "bench_c_p1w": -0.03565283250052409, "ret_p1m": 0.11122307263558073, "ret_signed_p1m": -0.11122307263558073, "alpha_spy_p1m": -0.031987834101905754, "alpha_c_p1m": 0.06410354511012906, "bench_spy_p1m": 0.07923523853367498, "bench_c_p1m": 0.1753266177457098, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2758, -0.2724, -0.2709, -0.2709, -0.2677, -0.2645, -0.2731, -0.2815, -0.2815, -0.2482, -0.205, -0.1724, -0.183, -0.2125, -0.1579, -0.141, -0.1476, -0.1559, -0.1079, -0.0857, -0.0857, -0.1102, -0.0906, -0.1087, -0.0986, -0.1068, -0.0697, -0.0585, -0.0456, -0.0988, -0.0818, -0.109, -0.1679, -0.15, -0.0983, -0.0757, -0.0799, -0.0497, -0.0818, -0.0077, -0.0077, 0.0041, 0.0326, 0.0353, 0.0509, 0.0457, 0.058, 0.1057, 0.0518, 0.0422, 0.0419, -0.0165, 0.0015, -0.0299, -0.0909, -0.0511, -0.0317, -0.0172, -0.0558, -0.0439, -0.0309, -0.0133, -0.0217, 0.0, -0.0004, 0.0128, 0.047, 0.034, -0.0522, -0.0561, -0.0954, -0.0432, -0.0095, -0.0245, -0.0245, -0.0128, -0.0081, 0.0798, 0.1137, 0.1184, 0.1078, 0.1076, 0.1037, 0.0915, 0.1112, 0.0963, 0.1039, 0.1295, 0.1307, 0.1675, 0.1334, 0.0669, 0.0711, 0.1044, 0.0892, 0.0957, 0.1501, 0.1999, 0.1496, 0.219, 0.2419, 0.2071, 0.2223, 0.2333, 0.2223, 0.1578, 0.1391, 0.195, 0.1979, 0.2113, 0.2113, 0.275, 0.2564, 0.2604, 0.2615, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2034726127030476973", "ticker_idx": 2, "ts": "2026-03-19T20:18:06+00:00", "symbol": "LRCX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "modestly negative for AMAT and LRCX", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Micron's EUV pivot benefits ASML but is modestly negative for AMAT and LRCX as multi-patterning steps decline.", "resolved_tickers": ["LRCX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Micron's strategic pivot toward more aggressive EUV adoption is interpreted as an effort to save fab space and thereby expand overall output capacity.\n\nThis implies a departure from its conventional multi-patterning approach — a development that is positive for ASML, but modestly negative for AMAT and LRCX.", "reply_text": null, "parent_text": null, "quoted_text": "First 5-year SCA signed. New EUV layer roadmap. ✍️ https://t.co/52JZpZMK6l", "parent_author": null, "ret_m1d": -0.03965980837402894, "ret_signed_m1d": 0.03965980837402894, "alpha_spy_m1d": 0.042130236836095625, "alpha_c_m1d": 0.03649461046686153, "bench_spy_m1d": 0.002470428462066687, "bench_c_m1d": -0.0031651979071674052, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024060877604351272, "ret_signed_p1d": 0.024060877604351272, "alpha_spy_p1d": 0.009725101731847041, "alpha_c_p1d": -0.0017165528808783526, "bench_spy_p1d": -0.01433577587250423, "bench_c_p1d": -0.025777430485229624, "ret_p1w": -0.09560241826922855, "ret_signed_p1w": 0.09560241826922855, "alpha_spy_p1w": 0.07597794231813026, "alpha_c_p1w": 0.05994958576870446, "bench_spy_p1w": -0.019624475951098286, "bench_c_p1w": -0.03565283250052409, "ret_p1m": 0.14363861627129815, "ret_signed_p1m": -0.14363861627129815, "alpha_spy_p1m": -0.06440337773762317, "alpha_c_p1m": 0.03168800147441164, "bench_spy_p1m": 0.07923523853367498, "bench_c_p1m": 0.1753266177457098, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2519, -0.2523, -0.2431, -0.2431, -0.2399, -0.2493, -0.2582, -0.2693, -0.2693, -0.2101, -0.1687, -0.1166, -0.1331, -0.1422, -0.0679, -0.0592, -0.0849, -0.1088, -0.0717, -0.0483, -0.0483, -0.0506, -0.0251, -0.0579, -0.0697, -0.0487, 0.0179, 0.0227, 0.0593, -0.0035, 0.0138, -0.0178, -0.1045, -0.0895, -0.0139, -0.0213, -0.0327, 0.0036, -0.0127, 0.0054, 0.0054, 0.0056, 0.0248, 0.0133, 0.0455, 0.0341, 0.0426, 0.0649, 0.0205, -0.0016, -0.014, -0.0726, -0.047, -0.0825, -0.1481, -0.0976, -0.0802, -0.0646, -0.1047, -0.0931, -0.0624, -0.0321, -0.0397, 0.0, -0.0241, -0.0029, 0.0207, -0.0023, -0.0956, -0.0965, -0.1456, -0.0869, -0.0512, -0.0665, -0.0665, -0.057, -0.0412, 0.0534, 0.1059, 0.1268, 0.1424, 0.1642, 0.1332, 0.1153, 0.1436, 0.1247, 0.1042, 0.1349, 0.105, 0.1444, 0.1089, 0.0737, 0.0631, 0.102, 0.0971, 0.105, 0.1787, 0.27, 0.2245, 0.2567, 0.2652, 0.2361, 0.2626, 0.2785, 0.2168, 0.1879, 0.1683, 0.2483, 0.2917, 0.305, 0.305, 0.379, 0.363, 0.359, 0.3598, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2017424430323175849", "ticker_idx": 0, "ts": "2026-01-31T02:27:20+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "In 2026, the global IC substrate industry is projected to focus on expanding ABF substrate production for AI chips, which will be a primary capacity expansion goal for Taiwanese suppliers", "tweet_type": "analysis", "timeline": "quarters", "summary": "ABF substrate structural shortage drives growth for Unimicron, Kinsus, Nan Ya PCB, and Ibiden into 2026 on AI server demand; Q1 2026 expected to hold steady despite seasonal weakness.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF substrate crunch reshapes market: Unimicron leads and rivals close in\n\n- As AI server architectures rapidly shift toward ultra-high-density designs, advanced CoWoS packaging capacity faces persistent shortages.\n\n- Simultaneously, warping and thermal stress issues caused by expanding die sizes have led to a surge in ABF substrate demand, resulting in structural supply pressure.\n\n- As AI GPU and ASIC customers accelerate market expansion, Japan’s Ibiden recently announced a large-scale capacity expansion, highlighting the ongoing shortage of ABF substrates for AI servers.\n\n- Alongside Unimicron, an early supply chain entrant, Nan Ya PCB and Kinsus have also begun securing related orders, with shipments expected to increase steadily.\n\n- Kinsus recorded the strongest performance, with 2025 annual revenue surging 28.87%, driven by increased orders for ABF and BT substrates.\n\n- Nan Ya PCB also achieved robust growth, with 2025 revenue increasing 24.44%, as the effects of product mix shifts continued to manifest.\n\n- Although Unimicron’s annual revenue for 2025 grew by a relatively modest 13.75%, its December net profit recorded explosive growth, increasing 2,770% year-on-year.\n\n- In the case of BT substrates, while they benefited from active purchasing by downstream companies due to past memory shortages, growth is expected to slow due to deteriorating conditions in the consumer electronics sector.\n\n- In 2026, the global IC substrate industry is projected to focus on expanding ABF substrate production for AI chips, which will be a primary capacity expansion goal for Taiwanese suppliers.\n\n- Looking ahead to Q1 2026, IC substrate manufacturers are expected to maintain steady momentum even during the traditional off-season, supported by rising AI-driven demand and extended order lead times from memory customers.\n\n- Utilization rates for both ABF and BT substrates are expected to steadily recover, offsetting the revenue decline that typically occurs as the consumer device launch cycle ends.\n\n- Simultaneously, rising product prices are contributing to improved profitability, laying a solid foundation for annual growth.\n\n---\n\nhttps://t.co/6mYzd24jIj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03698630136986303, "ret_signed_m1d": 0.03698630136986303, "alpha_spy_m1d": 0.041933071316017756, "alpha_c_m1d": 0.046486364317272155, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0547945205479452, "ret_signed_p1d": 0.0547945205479452, "alpha_spy_p1d": 0.06324991697864579, "alpha_c_p1d": 0.07668623060029411, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": -0.047945205479452024, "ret_signed_p1w": -0.047945205479452024, "alpha_spy_p1w": -0.04584576772921145, "alpha_c_p1w": -0.03479650666077028, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": 0.2684931506849315, "ret_signed_p1m": 0.2684931506849315, "alpha_spy_p1m": 0.29017812923643127, "alpha_c_p1m": 0.32191451853791586, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 1.419178082191781, "ret_signed_p3m": 1.419178082191781, "alpha_spy_p3m": 1.3829222751848074, "alpha_c_p3m": 1.3197816516096263, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5542, -0.5501, -0.561, -0.5528, -0.5474, -0.5446, -0.5147, -0.5301, -0.5411, -0.5342, -0.5603, -0.5301, -0.5616, -0.5507, -0.5178, -0.511, -0.4877, -0.489, -0.4603, -0.463, -0.4616, -0.4562, -0.411, -0.3959, -0.3808, -0.3767, -0.3836, -0.3836, -0.3932, -0.4137, -0.4192, -0.4329, -0.411, -0.3959, -0.4096, -0.4055, -0.4055, -0.3959, -0.3959, -0.3986, -0.3973, -0.3973, -0.4, -0.3699, -0.3685, -0.3808, -0.3767, -0.3822, -0.3329, -0.3041, -0.3151, -0.3164, -0.2493, -0.1753, -0.1562, -0.1205, -0.1041, -0.0712, -0.0877, -0.0301, 0.0068, -0.0562, 0.037, 0.0, 0.0548, 0.063, 0.0014, -0.0685, -0.0479, 0.0178, 0.011, 0.011, 0.011, 0.011, 0.011, 0.011, 0.011, 0.011, 0.111, 0.2219, 0.2562, 0.3192, 0.3192, 0.3178, 0.2685, 0.1425, 0.2329, 0.1836, 0.0658, 0.1616, 0.2767, 0.3151, 0.3918, 0.5041, 0.4877, 0.589, 0.5918, 0.5178, 0.3671, 0.2603, 0.3863, 0.3753, 0.3836, 0.3521, 0.2178, 0.3384, 0.4219, 0.4219, 0.4219, 0.5452, 0.6986, 0.7123, 0.7479, 0.7151, 0.6356, 0.6932, 0.7671, 0.7616, 0.8575, 0.9452, 0.9616, 0.9945, 1.1644, 1.2986, 1.2603, 1.2, 1.4192, 1.4192, 1.4959, 1.474, 1.3507, 1.4548, 1.2411, 1.3589, 1.3973, 1.4438, 1.4137, 1.2493, 1.2356, 1.2411, 1.2548, 1.4795, 1.6575, 1.7123, 1.9726, 1.9589, 1.8082, 1.8904, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2017424430323175849", "ticker_idx": 1, "ts": "2026-01-31T02:27:20+00:00", "symbol": "Nan Ya PCB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nan Ya PCB also achieved robust growth, with 2025 revenue increasing 24.44%, as the effects of product mix shifts continued to manifest", "tweet_type": "analysis", "timeline": "quarters", "summary": "ABF substrate structural shortage drives growth for Unimicron, Kinsus, Nan Ya PCB, and Ibiden into 2026 on AI server demand; Q1 2026 expected to hold steady despite seasonal weakness.", "resolved_tickers": ["8046.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nan Ya PCB (NYPCB) listed on TWSE as 8046.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF substrate crunch reshapes market: Unimicron leads and rivals close in\n\n- As AI server architectures rapidly shift toward ultra-high-density designs, advanced CoWoS packaging capacity faces persistent shortages.\n\n- Simultaneously, warping and thermal stress issues caused by expanding die sizes have led to a surge in ABF substrate demand, resulting in structural supply pressure.\n\n- As AI GPU and ASIC customers accelerate market expansion, Japan’s Ibiden recently announced a large-scale capacity expansion, highlighting the ongoing shortage of ABF substrates for AI servers.\n\n- Alongside Unimicron, an early supply chain entrant, Nan Ya PCB and Kinsus have also begun securing related orders, with shipments expected to increase steadily.\n\n- Kinsus recorded the strongest performance, with 2025 annual revenue surging 28.87%, driven by increased orders for ABF and BT substrates.\n\n- Nan Ya PCB also achieved robust growth, with 2025 revenue increasing 24.44%, as the effects of product mix shifts continued to manifest.\n\n- Although Unimicron’s annual revenue for 2025 grew by a relatively modest 13.75%, its December net profit recorded explosive growth, increasing 2,770% year-on-year.\n\n- In the case of BT substrates, while they benefited from active purchasing by downstream companies due to past memory shortages, growth is expected to slow due to deteriorating conditions in the consumer electronics sector.\n\n- In 2026, the global IC substrate industry is projected to focus on expanding ABF substrate production for AI chips, which will be a primary capacity expansion goal for Taiwanese suppliers.\n\n- Looking ahead to Q1 2026, IC substrate manufacturers are expected to maintain steady momentum even during the traditional off-season, supported by rising AI-driven demand and extended order lead times from memory customers.\n\n- Utilization rates for both ABF and BT substrates are expected to steadily recover, offsetting the revenue decline that typically occurs as the consumer device launch cycle ends.\n\n- Simultaneously, rising product prices are contributing to improved profitability, laying a solid foundation for annual growth.\n\n---\n\nhttps://t.co/6mYzd24jIj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05570652173913038, "ret_signed_m1d": 0.05570652173913038, "alpha_spy_m1d": 0.060653291685285105, "alpha_c_m1d": 0.0652065846865395, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0720108695652173, "ret_signed_p1d": 0.0720108695652173, "alpha_spy_p1d": 0.08046626599591788, "alpha_c_p1d": 0.0939025796175662, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": 0.13179347826086962, "ret_signed_p1w": 0.13179347826086962, "alpha_spy_p1w": 0.1338929160111102, "alpha_c_p1w": 0.14494217707955137, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": 0.3586956521739131, "ret_signed_p1m": 0.3586956521739131, "alpha_spy_p1m": 0.38038063072541284, "alpha_c_p1m": 0.41211702002689743, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 1.7309782608695654, "ret_signed_p3m": 1.7309782608695654, "alpha_spy_p3m": 1.6947224538625918, "alpha_c_p3m": 1.6315818302874108, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2609, -0.2582, -0.2582, -0.2106, -0.2147, -0.2269, -0.2065, -0.2323, -0.2595, -0.2785, -0.3179, -0.269, -0.3207, -0.3288, -0.3207, -0.3016, -0.2921, -0.2867, -0.2758, -0.288, -0.269, -0.2976, -0.2663, -0.2663, -0.2283, -0.2391, -0.2446, -0.2622, -0.2772, -0.3234, -0.3383, -0.3628, -0.3492, -0.337, -0.3505, -0.3505, -0.3505, -0.3302, -0.3465, -0.3465, -0.3451, -0.3451, -0.356, -0.3438, -0.3356, -0.3438, -0.356, -0.3465, -0.3084, -0.2717, -0.2894, -0.284, -0.2133, -0.1359, -0.072, -0.019, 0.0068, 0.038, 0.0054, 0.0734, 0.0761, 0.0163, 0.0557, 0.0, 0.072, 0.1141, 0.0802, 0.0299, 0.1318, 0.1128, 0.1304, 0.1304, 0.1304, 0.1304, 0.1304, 0.1304, 0.1304, 0.1304, 0.2432, 0.3668, 0.5027, 0.5082, 0.5082, 0.4701, 0.3587, 0.2228, 0.2527, 0.2758, 0.1495, 0.163, 0.2785, 0.2731, 0.3995, 0.538, 0.4783, 0.5408, 0.5245, 0.4701, 0.4511, 0.3342, 0.4674, 0.5462, 0.5245, 0.5761, 0.4185, 0.5598, 0.5435, 0.5435, 0.5435, 0.6957, 0.731, 0.8179, 0.8125, 0.7799, 0.7554, 0.7826, 0.9592, 0.9674, 1.163, 1.144, 1.1821, 1.1603, 1.375, 1.481, 1.5326, 1.4891, 1.731, 1.731, 1.6793, 1.7092, 1.5543, 1.5734, 1.3207, 1.4674, 1.4457, 1.3913, 1.3016, 1.1359, 1.1495, 1.1685, 1.2527, 1.3696, 1.538, 1.4022, 1.5054, 1.4592, 1.3098, 1.3043, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2017424430323175849", "ticker_idx": 2, "ts": "2026-01-31T02:27:20+00:00", "symbol": "Kinsus", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Kinsus recorded the strongest performance, with 2025 annual revenue surging 28.87%, driven by increased orders for ABF and BT substrates", "tweet_type": "analysis", "timeline": "quarters", "summary": "ABF substrate structural shortage drives growth for Unimicron, Kinsus, Nan Ya PCB, and Ibiden into 2026 on AI server demand; Q1 2026 expected to hold steady despite seasonal weakness.", "resolved_tickers": ["3189.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Kinsus Interconnect Technology 3189.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF substrate crunch reshapes market: Unimicron leads and rivals close in\n\n- As AI server architectures rapidly shift toward ultra-high-density designs, advanced CoWoS packaging capacity faces persistent shortages.\n\n- Simultaneously, warping and thermal stress issues caused by expanding die sizes have led to a surge in ABF substrate demand, resulting in structural supply pressure.\n\n- As AI GPU and ASIC customers accelerate market expansion, Japan’s Ibiden recently announced a large-scale capacity expansion, highlighting the ongoing shortage of ABF substrates for AI servers.\n\n- Alongside Unimicron, an early supply chain entrant, Nan Ya PCB and Kinsus have also begun securing related orders, with shipments expected to increase steadily.\n\n- Kinsus recorded the strongest performance, with 2025 annual revenue surging 28.87%, driven by increased orders for ABF and BT substrates.\n\n- Nan Ya PCB also achieved robust growth, with 2025 revenue increasing 24.44%, as the effects of product mix shifts continued to manifest.\n\n- Although Unimicron’s annual revenue for 2025 grew by a relatively modest 13.75%, its December net profit recorded explosive growth, increasing 2,770% year-on-year.\n\n- In the case of BT substrates, while they benefited from active purchasing by downstream companies due to past memory shortages, growth is expected to slow due to deteriorating conditions in the consumer electronics sector.\n\n- In 2026, the global IC substrate industry is projected to focus on expanding ABF substrate production for AI chips, which will be a primary capacity expansion goal for Taiwanese suppliers.\n\n- Looking ahead to Q1 2026, IC substrate manufacturers are expected to maintain steady momentum even during the traditional off-season, supported by rising AI-driven demand and extended order lead times from memory customers.\n\n- Utilization rates for both ABF and BT substrates are expected to steadily recover, offsetting the revenue decline that typically occurs as the consumer device launch cycle ends.\n\n- Simultaneously, rising product prices are contributing to improved profitability, laying a solid foundation for annual growth.\n\n---\n\nhttps://t.co/6mYzd24jIj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05219206680584554, "ret_signed_m1d": 0.05219206680584554, "alpha_spy_m1d": 0.057138836752000266, "alpha_c_m1d": 0.06324682512662994, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.011054758320784397, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02713987473903967, "ret_signed_p1d": 0.02713987473903967, "alpha_spy_p1d": 0.03559527116974026, "alpha_c_p1d": 0.052362338435224975, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.025222463696185304, "ret_p1w": 0.06471816283924836, "ret_signed_p1w": 0.06471816283924836, "alpha_spy_p1w": 0.06681760058948893, "alpha_c_p1w": 0.06795371280681728, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.003235549967568918, "ret_p1m": 0.29018789144050094, "ret_signed_p1m": 0.29018789144050094, "alpha_spy_p1m": 0.3118728699920007, "alpha_c_p1m": 0.33163702645413495, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.04144913501363401, "ret_p3m": 1.2045929018789145, "ret_signed_p3m": 1.2045929018789145, "alpha_spy_p3m": 1.168337094871941, "alpha_c_p3m": 0.9625407909282901, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.2420521109506244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4468, -0.4572, -0.4656, -0.4656, -0.4656, -0.4509, -0.3967, -0.428, -0.4092, -0.3925, -0.4384, -0.4175, -0.4614, -0.4489, -0.4259, -0.4238, -0.4113, -0.405, -0.4092, -0.4113, -0.382, -0.382, -0.3445, -0.3486, -0.334, -0.3633, -0.3674, -0.3674, -0.3904, -0.4217, -0.4322, -0.4489, -0.4363, -0.3987, -0.405, -0.3987, -0.3987, -0.3612, -0.3716, -0.382, -0.3361, -0.3361, -0.3486, -0.2985, -0.2923, -0.2693, -0.3069, -0.2965, -0.2443, -0.263, -0.2818, -0.2693, -0.1962, -0.1463, -0.1664, -0.1927, -0.1705, -0.1362, -0.1315, -0.0459, 0.0167, -0.0418, 0.0522, 0.0, 0.0271, 0.1294, 0.1002, 0.0125, 0.0647, 0.0585, 0.1086, 0.1086, 0.1086, 0.1086, 0.1086, 0.1086, 0.1086, 0.1086, 0.2192, 0.3403, 0.3591, 0.3132, 0.3132, 0.2714, 0.2902, 0.1879, 0.2422, 0.2171, 0.1044, 0.1441, 0.2129, 0.2067, 0.3257, 0.4071, 0.4259, 0.5261, 0.6013, 0.5866, 0.4447, 0.3382, 0.4676, 0.4676, 0.4697, 0.4468, 0.3069, 0.3862, 0.4029, 0.4029, 0.4029, 0.4969, 0.6451, 0.6159, 0.5908, 0.5303, 0.5491, 0.6013, 0.643, 0.6347, 0.7975, 0.977, 1.0522, 1.0, 1.1962, 1.1503, 1.1211, 1.1211, 1.2046, 1.2046, 1.2171, 1.2004, 1.0146, 1.0564, 0.9123, 1.1002, 1.0626, 1.0772, 1.1795, 1.0292, 1.1086, 1.142, 1.1044, 1.3132, 1.5428, 1.7933, 1.9061, 1.9311, 1.8434, 2.0438, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2017424430323175849", "ticker_idx": 3, "ts": "2026-01-31T02:27:20+00:00", "symbol": "Ibiden", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Japan's Ibiden recently announced a large-scale capacity expansion, highlighting the ongoing shortage of ABF substrates for AI servers", "tweet_type": "analysis", "timeline": "quarters", "summary": "ABF substrate structural shortage drives growth for Unimicron, Kinsus, Nan Ya PCB, and Ibiden into 2026 on AI server demand; Q1 2026 expected to hold steady despite seasonal weakness.", "resolved_tickers": ["4062.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ibiden Co. Ltd 4062.T Japan ABF substrate", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF substrate crunch reshapes market: Unimicron leads and rivals close in\n\n- As AI server architectures rapidly shift toward ultra-high-density designs, advanced CoWoS packaging capacity faces persistent shortages.\n\n- Simultaneously, warping and thermal stress issues caused by expanding die sizes have led to a surge in ABF substrate demand, resulting in structural supply pressure.\n\n- As AI GPU and ASIC customers accelerate market expansion, Japan’s Ibiden recently announced a large-scale capacity expansion, highlighting the ongoing shortage of ABF substrates for AI servers.\n\n- Alongside Unimicron, an early supply chain entrant, Nan Ya PCB and Kinsus have also begun securing related orders, with shipments expected to increase steadily.\n\n- Kinsus recorded the strongest performance, with 2025 annual revenue surging 28.87%, driven by increased orders for ABF and BT substrates.\n\n- Nan Ya PCB also achieved robust growth, with 2025 revenue increasing 24.44%, as the effects of product mix shifts continued to manifest.\n\n- Although Unimicron’s annual revenue for 2025 grew by a relatively modest 13.75%, its December net profit recorded explosive growth, increasing 2,770% year-on-year.\n\n- In the case of BT substrates, while they benefited from active purchasing by downstream companies due to past memory shortages, growth is expected to slow due to deteriorating conditions in the consumer electronics sector.\n\n- In 2026, the global IC substrate industry is projected to focus on expanding ABF substrate production for AI chips, which will be a primary capacity expansion goal for Taiwanese suppliers.\n\n- Looking ahead to Q1 2026, IC substrate manufacturers are expected to maintain steady momentum even during the traditional off-season, supported by rising AI-driven demand and extended order lead times from memory customers.\n\n- Utilization rates for both ABF and BT substrates are expected to steadily recover, offsetting the revenue decline that typically occurs as the consumer device launch cycle ends.\n\n- Simultaneously, rising product prices are contributing to improved profitability, laying a solid foundation for annual growth.\n\n---\n\nhttps://t.co/6mYzd24jIj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06455367761775777, "ret_signed_m1d": 0.06455367761775777, "alpha_spy_m1d": 0.06950044756391249, "alpha_c_m1d": 0.07405374056516689, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08551097743736658, "ret_signed_p1d": 0.08551097743736658, "alpha_spy_p1d": 0.09396637386806717, "alpha_c_p1d": 0.10740268748971549, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": 0.00827947943346552, "ret_signed_p1w": 0.00827947943346552, "alpha_spy_p1w": 0.010378917183706093, "alpha_c_p1w": 0.021428178252147267, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": 0.11487716805163184, "ret_signed_p1m": 0.11487716805163184, "alpha_spy_p1m": 0.1365621466031316, "alpha_c_p1m": 0.1682985359046162, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 0.7470242664337559, "ret_signed_p3m": 0.7470242664337559, "alpha_spy_p3m": 0.7107684594267822, "alpha_c_p3m": 0.6476278358516012, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1268, -0.1038, -0.1708, -0.1569, -0.1349, -0.1135, -0.119, -0.1937, -0.142, -0.2147, -0.2471, -0.1892, -0.2671, -0.2671, -0.2671, -0.2946, -0.2594, -0.2316, -0.2678, -0.2775, -0.2364, -0.2309, -0.1782, -0.1759, -0.1636, -0.1843, -0.1798, -0.175, -0.2309, -0.2642, -0.2458, -0.2785, -0.2464, -0.2076, -0.1866, -0.1417, -0.1633, -0.1436, -0.1428, -0.1292, -0.1292, -0.1292, -0.1292, -0.074, -0.0929, -0.0657, -0.1137, -0.1314, -0.1314, -0.0595, -0.038, -0.0321, 0.0136, 0.0131, 0.0261, 0.0759, 0.1056, 0.0717, 0.0247, 0.0349, 0.0673, 0.0682, 0.0646, 0.0, 0.0855, -0.0686, -0.0796, -0.0834, 0.0083, 0.0243, 0.0243, 0.1001, 0.1177, 0.1358, 0.1572, 0.2075, 0.2468, 0.1846, 0.1846, 0.2787, 0.2486, 0.203, 0.2338, 0.2119, 0.1149, 0.0285, 0.0477, 0.0194, -0.0636, 0.0133, 0.0087, -0.0083, 0.016, 0.0542, 0.0995, 0.0972, 0.0696, 0.0696, -0.0246, 0.0136, 0.0968, 0.0944, 0.0697, 0.0078, -0.0446, 0.0459, 0.0076, 0.064, 0.1086, 0.1299, 0.2693, 0.2714, 0.2815, 0.2164, 0.285, 0.2157, 0.2405, 0.1969, 0.2409, 0.3686, 0.4515, 0.4431, 0.6252, 0.642, 0.6278, 0.6278, 0.747, 0.7334, 0.7334, 0.7334, 0.7334, 1.1222, 1.0108, 1.0289, 1.1449, 1.211, 1.198, 1.0296, 1.0263, 0.9855, 1.027, 1.3166, 1.4948, 1.7126, 1.7171, 1.6471, 1.5583, 1.9808, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2029391707062833510", "ticker_idx": 0, "ts": "2026-03-05T03:01:01+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the decline in overall semiconductor output likely means commodity DRAM production has been reduced to ramp up HBM, which is actually a bullish signal for memory", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish memory: no signs of falling prices; commodity DRAM supply squeeze from HBM ramp is a positive signal.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@MikeFritzell I still haven’t seen any signs that memory prices are falling. If anything, as you pointed out, the decline in overall semiconductor output likely means commodity DRAM production has been reduced to ramp up HBM, which is actually a bullish signal for memory.", "reply_text": null, "parent_text": "@jukan05 curious to hear your view, as I don't hear anything about the memory bear case, from anywhere", "quoted_text": null, "parent_author": "MikeFritzell", "ret_m1d": -0.06321731733386289, "ret_signed_m1d": -0.06321731733386289, "alpha_spy_m1d": -0.06882412520017407, "alpha_c_m1d": -0.0727025834072322, "bench_spy_m1d": 0.005606807866311181, "bench_c_m1d": 0.009485266073369303, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03439437735528935, "ret_signed_p1d": -0.03439437735528935, "alpha_spy_p1d": -0.021287242284452855, "alpha_c_p1d": 0.0030155336516410128, "bench_spy_p1d": -0.013107135070836495, "bench_c_p1d": -0.03740991100693036, "ret_p1w": -0.0033655467365576084, "ret_signed_p1w": -0.0033655467365576084, "alpha_spy_p1w": 0.01901784785611031, "alpha_c_p1w": 0.014896755297687672, "bench_spy_p1w": -0.022383394592667916, "bench_c_p1w": -0.01826230203424528, "ret_p1m": -0.05748584058147702, "ret_signed_p1m": -0.05748584058147702, "alpha_spy_p1m": -0.02271619391475469, "alpha_c_p1m": -0.049821748681846335, "bench_spy_p1m": -0.03476964666672233, "bench_c_p1m": -0.007664091899630687, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3992, -0.4004, -0.3842, -0.3976, -0.4072, -0.4234, -0.4395, -0.4291, -0.41, -0.3994, -0.3715, -0.3686, -0.3592, -0.3592, -0.3466, -0.3187, -0.3156, -0.3216, -0.3216, -0.2723, -0.2516, -0.2133, -0.2073, -0.2167, -0.2054, -0.2033, -0.2158, -0.2137, -0.2022, -0.1691, -0.1656, -0.1783, -0.1516, -0.1343, -0.1287, -0.1485, -0.0952, -0.0547, -0.0501, -0.0511, -0.0773, -0.0358, -0.0691, -0.1037, -0.0961, -0.0749, -0.0885, -0.0595, -0.0278, -0.028, -0.028, -0.038, -0.0202, -0.003, 0.0257, 0.0254, 0.0542, 0.0741, 0.1174, 0.0987, 0.0989, -0.0092, -0.0632, 0.0, -0.0344, -0.0752, -0.0024, 0.0203, -0.0034, -0.0007, 0.0442, 0.0685, 0.1244, 0.0806, 0.0587, -0.0059, 0.0114, 0.002, -0.0578, -0.0563, -0.1131, -0.1389, -0.0427, -0.0874, -0.0575, -0.0322, -0.0158, 0.0745, 0.0632, 0.0762, 0.0771, 0.1418, 0.1534, 0.1723, 0.1583, 0.1636, 0.1928, 0.2218, 0.2299, 0.2327, 0.2896, 0.2711, 0.2875, 0.2743, 0.2953, 0.402, 0.4556, 0.5907, 0.6023, 0.6926, 0.8317, 0.7802, 0.87, 0.8656, 0.7245, 0.7141, 0.6853, 0.7141, 0.8494, 0.8282, 0.8282, 1.0006, 1.1095, 1.1087, 1.1947, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014529638970937859", "ticker_idx": 0, "ts": "2026-01-23T02:44:28+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel Foundry (IFS) investments are set to be front-loaded in 1H, with a greater allocation toward equipment purchases. This suggests Intel is becoming more confident in its ability to secure foundry customers.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long semicap on Intel's revised flat 2026 capex with front-loaded IFS equipment purchases signaling foundry confidence.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Commenting on $INTC, Intel’s 2026 capex was originally expected to decline vs. 2025. However, note that Intel has revised this plan to keep capex roughly flat or only slightly down.\n\nIntel Foundry (IFS) investments are set to be front-loaded in 1H, with a greater allocation toward equipment purchases. This suggests Intel is becoming more confident in its ability to secure foundry customers.\n\nLong Semicap.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.20523630047998997, "ret_signed_m1d": 0.20523630047998997, "alpha_spy_m1d": 0.20559894915795618, "alpha_c_m1d": 0.19841280823745322, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05724424304031461, "ret_signed_p1d": -0.05724424304031461, "alpha_spy_p1d": -0.06232239010092444, "alpha_c_p1d": -0.05406998468520363, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.031062825279754236, "ret_signed_p1w": 0.031062825279754236, "alpha_spy_p1w": 0.027087364625359767, "alpha_c_p1w": 0.022639732605521434, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": -0.03195026928021816, "ret_signed_p1m": -0.03195026928021816, "alpha_spy_p1m": -0.022026269219125538, "alpha_c_p1m": -0.06391809817369609, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.4481916371204664, "ret_signed_p3m": 0.4481916371204664, "alpha_spy_p3m": 0.4134827665775198, "alpha_c_p3m": 0.25638481631316656, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0785, -0.0828, -0.1089, -0.1127, -0.1236, -0.1784, -0.1484, -0.1737, -0.154, -0.1469, -0.1595, -0.1593, -0.2032, -0.2119, -0.2299, -0.2383, -0.221, -0.254, -0.2345, -0.2059, -0.205, -0.1833, -0.1833, -0.1001, -0.1123, -0.0355, -0.0291, -0.1014, -0.0812, -0.1058, -0.1014, -0.0952, -0.1234, -0.1611, -0.1677, -0.1722, -0.2001, -0.195, -0.183, -0.193, -0.1935, -0.1977, -0.1977, -0.1968, -0.1862, -0.1724, -0.1813, -0.1813, -0.1262, -0.1265, -0.1116, -0.0541, -0.0879, 0.0107, -0.0224, 0.0493, 0.081, 0.0721, 0.0419, 0.0419, 0.0774, 0.2037, 0.2052, 0.0, -0.0572, -0.0253, 0.0823, 0.0797, 0.0311, 0.083, 0.0927, 0.0783, 0.0703, 0.1225, 0.1147, 0.0457, 0.0714, 0.0313, 0.0382, 0.0382, 0.0246, 0.0087, -0.01, -0.0213, -0.032, 0.0233, 0.0402, 0.0087, 0.012, 0.0095, -0.0437, 0.0113, 0.0195, -0.0366, 0.0113, 0.0379, 0.0646, 0.004, 0.0155, 0.0153, -0.0224, -0.0009, 0.0246, -0.0266, -0.0235, -0.0224, 0.0468, -0.0215, -0.043, -0.0861, -0.0209, 0.0657, 0.1178, 0.1178, 0.1267, 0.174, 0.308, 0.3694, 0.3841, 0.4462, 0.4158, 0.4409, 0.5199, 0.5199, 0.4577, 0.4702, 0.4482, 0.4817, 0.8314, 0.8857, 0.8753, 1.1023, 1.0963, 1.2103, 1.1251, 1.3996, 1.5074, 1.4322, 1.7717, 1.872, 1.6761, 1.669, 1.5722, 1.4134, 1.4, 1.4584, 1.6394, 1.6292, 1.659, 1.659, 1.7406, 1.7018, 1.6823, 1.5445, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014529638970937859", "ticker_idx": 1, "ts": "2026-01-23T02:44:28+00:00", "symbol": "semiconductor equipment", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Long Semicap.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long semicap on Intel's revised flat 2026 capex with front-loaded IFS equipment purchases signaling foundry confidence.", "resolved_tickers": ["AMAT", "LRCX", "KLAC", "KLIC"], "resolution_method": "basket", "resolution_reason": "US semiconductor equipment basket: AMAT, Lam, KLA, Kulicke", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Commenting on $INTC, Intel’s 2026 capex was originally expected to decline vs. 2025. However, note that Intel has revised this plan to keep capex roughly flat or only slightly down.\n\nIntel Foundry (IFS) investments are set to be front-loaded in 1H, with a greater allocation toward equipment purchases. This suggests Intel is becoming more confident in its ability to secure foundry customers.\n\nLong Semicap.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006784330431116287, "ret_signed_m1d": 0.006784330431116287, "alpha_spy_m1d": 0.007146979109082491, "alpha_c_m1d": -3.916181142046837e-05, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008433324832431255, "ret_signed_p1d": 0.008433324832431255, "alpha_spy_p1d": 0.003355177771821427, "alpha_c_p1d": 0.011607583187542236, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.002780317355215889, "ret_signed_p1w": 0.002780317355215889, "alpha_spy_p1w": -0.0011951432991785804, "alpha_c_p1w": -0.005642775319016913, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": 0.11420682081052858, "ret_signed_p1m": 0.11420682081052858, "alpha_spy_p1m": 0.1241308208716212, "alpha_c_p1m": 0.08223899191705064, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.2893790528151034, "ret_signed_p3m": 0.2893790528151034, "alpha_spy_p3m": 0.2546701822721568, "alpha_c_p3m": 0.09757223200780357, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2694, -0.258, -0.2641, -0.2672, -0.2563, -0.2805, -0.2513, -0.2683, -0.2764, -0.2571, -0.2771, -0.2701, -0.2973, -0.3074, -0.3126, -0.3223, -0.3021, -0.3184, -0.3048, -0.2812, -0.2602, -0.2425, -0.2425, -0.2364, -0.2403, -0.2213, -0.2049, -0.2055, -0.1997, -0.1931, -0.1882, -0.1747, -0.1787, -0.2096, -0.1997, -0.2076, -0.2381, -0.2143, -0.1993, -0.188, -0.1863, -0.1815, -0.1815, -0.1781, -0.1849, -0.1921, -0.2056, -0.2056, -0.1586, -0.1096, -0.0696, -0.0812, -0.0978, -0.0442, -0.0282, -0.0387, -0.0438, 0.0093, 0.0184, 0.0184, -0.006, 0.0237, 0.0068, 0.0, 0.0084, 0.0512, 0.0583, 0.0834, 0.0028, 0.0118, -0.0164, -0.0707, -0.0112, 0.0616, 0.0741, 0.0767, 0.103, 0.0729, 0.0985, 0.0985, 0.1038, 0.1064, 0.1044, 0.1277, 0.1142, 0.1322, 0.1648, 0.1255, 0.1125, 0.1132, 0.0482, 0.0724, 0.0279, -0.0294, 0.0265, 0.0405, 0.0513, 0.0111, 0.0233, 0.0461, 0.0672, 0.0602, 0.0871, 0.0659, 0.0835, 0.1203, 0.1102, 0.0282, 0.0253, -0.0215, 0.0408, 0.07, 0.0517, 0.0517, 0.0723, 0.0818, 0.1746, 0.2265, 0.238, 0.2468, 0.2654, 0.2478, 0.2366, 0.2688, 0.2686, 0.264, 0.2894, 0.2799, 0.3337, 0.3035, 0.2454, 0.248, 0.2647, 0.2588, 0.2672, 0.3207, 0.3833, 0.3674, 0.4325, 0.4442, 0.3938, 0.4388, 0.4525, 0.4092, 0.3643, 0.3437, 0.4111, 0.4251, 0.4545, 0.4545, 0.5306, 0.4992, 0.4906, 0.4777, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2008473953073779184", "ticker_idx": 0, "ts": "2026-01-06T09:41:20+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Vera Rubin: Entered mass production regime. Increased visibility for 2H26 shipments… Reasoning created 5x; Agentic will create 50x… overall platform strategy remains centered on general-purpose GPU/systems", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish NVDA: Vera Rubin mass production on track for 2H26, agentic AI driving 50x token demand, and general-purpose GPU platform retains structural TCO advantage over SRAM/CPX alternatives.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "260106_NVIDIA: CES Financial Analyst Q&A\n\nQ1) Bank of America: What does it mean that Vera Rubin has entered Full Production? Does this accelerate shipments and revenue recognition?\n\nA1) In the past, this term referred to Tape-Out, but now it literally means progress into the mass production phase. Market launch is in the second half of this year, and the cycle time should be considered around 9 months or more.\n\nQ2) Bank of America: What is the short-, medium-, and long-term significance of the Groq license for us? Does this mean we have reached the stage where we believe we need more \"ASIC-like\" chips specialized for certain types of inference?\n\nA2) Our chips are inherently ASICs. The concept is that we design them in-house and manufacture them through TSMC foundries. Most AI workloads are highly likely to run on our architecture. What Groq pursues is extreme low latency, which is in a trade-off relationship with our strength in extreme high throughput. In the future, when AI-powered devices (such as glasses) emerge, diverse demand for low latency will arise.\n\nQ3) Could you elaborate a bit more to help us better understand the core market sizes and timing for data centers, autonomous driving, robotics, etc.? In particular, how do you view the revenue model for Alpamayo?\n\nA3) Autonomous driving took 8 years to reach this point, and we provide data center training infrastructure and the full software stack (simulation + synthetic data + world models, etc.) across the entire ecosystem. Most companies use us—some extend into vehicles, while others deploy the full stack. In the future, autonomous driving and robotaxis will become a very large business.\n\nQ4) Melius Research: Regarding the \"better token economics\" comment, I want to ask about Anthropic. They could have obtained as much Trainium as they wanted, or TPUs if they wished—why choose us? Are there cases where workloads previously tuned for Trainium or TPUs are migrating to Rubin? Is it ultimately because of token economics?\n\nA4) I regret not being able to invest in Anthropic. Anthropic excels at producing high-quality tokens, especially in enterprise coding, where iterative tasks make token generation speed (low latency/high-speed generation) critical—and that's where we deliver significant value. AI companies need platforms that can run anything with a rich ecosystem. With last year's Anthropic announcement, the last major player not running on NVIDIA has now joined.\n\nQ5) Wells Fargo: Supply chain variables seem numerous—1) DRAM pricing, 2) available supply volumes, etc. Please explain. Also, could power become a bottleneck in the future?\n\nA5) The reason we are advantaged is that we were already at massive scale and growing very rapidly from that large base, so we prepared for this kind of major ramp-up with partners well in advance. We invested heavily in partners, much of it in the form of prepayments, enabling them to expand capacity. Thanks to long-standing relationships, we believe we are in a strong position. We are essentially the only semiconductor company in the world that directly purchases DRAM at global scale. People ask, \"Why buy DRAM?\" The reason is simple: transforming that DRAM into CoWoS supercomputers is extraordinarily difficult. Properly connecting that supply chain like \"plumbing\" is a major advantage for us. In tight situations like now, having this capability is also fortunate.\n\nQ6) Bernstein: If China-related restrictions are eased, can ramp-up proceed smoothly? Regarding DGX, you previously said it wasn't the original target market, but the stance seems different now. What changed?\n\nA6) H20 is dedicated volume for China, so it does not overlap with existing orders. H200 is currently in the licensing process. For DGX, it was never about competing with CSPs; initially, when CSPs were unfamiliar with AI Cloud, it was a strategic vehicle to land our models and partner ecosystem with them. Now the market flywheel is spinning, and the need to force token leasing via DGX Cloud has diminished. The resale model for profit was somewhat inefficient to begin with.\n\nQ7) Goldman Sachs: How important is context memory storage control?\n\nA7) With inference volumes surging, context/token memory and KV Cache throughput have exploded, reaching limits with existing storage architectures—this makes it extremely important. The demand is large enough to create a \"new storage tier,\" potentially becoming a massive storage market.\n\nQ8) Have there been performance bottlenecks in specific customer/workload segments? Like networking in the past, can we expect continued innovation in storage/memory going forward?\n\nA8) AI DB and KV Cache have become extremely heavy workloads; handling them over traditional north-south network-based storage wastes traffic. We are already the world's largest in networking and expect to become the largest in storage processors as well. Based on BlueField-4 + DOCA adoption, we will continue expanding and innovating the infrastructure stack from networking into storage. In other words, we will strengthen BlueField-4 and DOCA as the SmartNIC for AI factories, establishing a dominant position in storage processing following networking.\n\nQ9) Truist: When ramping Rubin, what happens to Hopper? The ramp speed seems quite fast—what impact on margins?\n\nA9) Rubin targets \"zero-downtime AI factories\" with features like RAS and NVLink hot-swap. The challenge with Rubin is that it is historically unique as a computer where \"every chip inside is brand new\"—even high-temperature capacitors were newly developed. HBM4 did not exist, nor did LPDDR5 SOCAMM. Co-Packaged Optics are integrated into switches. This was possible because we de-risked technologies, components, and supply chains piece by piece over about 5 years, and now everything is verified and in mass production readiness.\n\nQ10) Truist: How do you see the frontier AI model market evolving going forward?\n\nA10) I agree that in general-purpose models, most are now at a similar \"good enough\" level. General-purpose reaches \"good enough\" quickly, but in specialized models, it is still unrealistic for one model to excel across all domains. If consolidation occurs, it will likely be around \"channels\"—platforms like Alphabet, Meta, social media, etc.\n\nQ11) Daiwa: Where is the industry at present with Agentic/Physical AI?\n\nA11) Agentic AI is already a core trend. Most future AI applications will likely be agentic. Citing Cursor and OpenEvidence as examples, in the future, instead of libraries, combining models + agentic systems and simply giving goals will enable automated collaboration and easier application deployment. Autonomous driving/Physical AI is evolving toward fully agentic forms like Alpamayo's VLA (vision-language-action) robotic systems that perform situation decomposition, commonsense reasoning, and user Q&A.\n\nQ12) Morgan Stanley: In the LLM era, CSPs shifted CAPEX— who will fund Physical AI?\n\nA12) By creating and freely distributing Cosmos, we significantly lower the initial cost/barrier that the industry must bear, so companies will invest money and effort in fine-tuning and domain adaptation on top of it.\n\nQ13) Loop Capital: What role does Neo-Cloud play in the market? Why are we strengthening partnerships with them? Does enterprise/sovereign adoption help NVIDIA Enterprise OS sales?\n\nA13) In an environment where technology/market changes are rapid and core data center resources (land/power/space) are constrained, agile operators will succeed. AI infrastructure will be built regionally, evolving into communities that include researchers, startups, and large enterprises. Providing infrastructure close to customers is becoming important, leading to growth in regionally distributed AI infrastructure.\n\nQ14) Wolfe Research: Why can initial Rubin ramp margins be maintained? How will high margins be sustained amid competition?\n\nA14) Precise predictions like 50BP increments are difficult. Focus on mid- to long-term customer value rather than short-term numbers. Key factors include training efficiency, inference token cost and speed, and maximizing data center revenue within constrained power. We will also achieve world-leading RAS (reliability, availability, serviceability).\n\nQ15) If there is a game-changer in the oligopoly landscape?\n\nA15) There are about four. 1) Memory/context processing innovation—as context length grows (100K–1M → effectively infinite), attention/memory handling becomes the bottleneck, and research solving this could reshape competition. 2) Hybrid models (Transformer + SSM, etc.)—SSM is advantageous for context compression, making hybrids important. 3) System architecture innovation (bringing memory closer)—approaches like BlueField-4 to place long-term memory nearer to the compute fabric. Finally, 4) Continuous learning: automating \"recognize unknown → research/learn → re-infer\" is a major opportunity.\n\nQ16) Tokens are growing 5x—when will re-acceleration/slowdown occur?\n\nA16) The 5x growth driver is Reasoning. The next re-acceleration factor is Agentic Systems—with tool use, planning, and simulation, token usage will surge like \"real-time AlphaGo.\" It could grow 50x.\n\nQ17) Stifel: What does the pace/numbers like Spectrum X and 400G SerDes imply for future R&D intensity? Are there immediate benefits felt at EDA companies or Cursor, etc., and what is the relationship with Grok, Fabrica (Infabrica), etc.?\n\nA17) We will continue investing across the entire stack (land/power/shell → chips → infrastructure → models → applications), and building such systems is critical. For Infabrica, scale-up has high integration that makes an independent company difficult, so internal integration was rational. For Groq, collaboration was natural because innovation is needed at the team level.\n\nQ18) UBS: Can the software technology gained from the Grok deal be applied across the NVIDIA portfolio?\n\nA18) Yes. Grok's programming model differs from NVIDIA's, and that difference underlies extreme low latency vs. extreme high throughput. We are considering reconfiguring some elements to absorb and extend them into NVIDIA capabilities. The product portfolio will remain mostly (~90%) centered on Grace Blackwell/Vera Rubin, though some (~10%) could take new forms specialized for extreme low latency.\n\nQ19) Robocap: NVIDIA already has CPX technology and, through the Groq acquisition, secured SRAM accessibility for inference. Your team published a paper a month ago on using CPX to reduce HBM usage, suggesting GDDR7 could replace HBM. Going forward, could the combination of Grok (=SRAM) + internal CPX allow HBM usage to be controlled at a more \"manageable level\"? And could this positively impact our margins?\n\nA19) First explain the strengths of each, then why it is not that simple. CPX offers better prefill performance per dollar. If everything could fit in SRAM, HBM would indeed be unnecessary. However, the issue is that model size would shrink by roughly 100x. For certain workloads, SRAM can be dramatically faster than HBM access. But workloads (models/tasks) keep changing—sometimes MoE, sometimes multimodal, diffusion, autoregressive, or SSM. These vary in shape and size, stressing NVLink, HBM, or all three differently. We are the universal answer because of flexibility. When workloads differ morning vs. night or customer to customer, our broad strength and flexibility across nearly everything is key. Extreme optimization for one workload is possible, but if it only covers 10% of data center workloads and is underutilized, the power/resources could have been used elsewhere—power is constrained. The key is viewing data centers not as \"infinite money and space\" but as finite power. Finite power must be optimally utilized across the entire data center, and greater flexibility is better. More integrated architecture is also better—for example, updating DeepSeek instantly improves all GPUs in the data center. Updating model libraries improves the whole data center. Understand? But with 17 fragmented architectures, \"this is good for that, that is good for this\" results in overall TCO not improving as much. In summary, CPX clearly has advantages but also reduces data center flexibility.\n\n[Implication]\n\n(1) Vera Rubin: Entered mass production regime. Increased visibility for 2H26 shipments\n\n■ All components—HBM4, LPDDR5 SOCAMM, CPO, etc.—incorporate new parts and technologies\n\n■ This reflects years of building technology + component supply chains, now fully verified and ready for mass production\n\n■ Market launch/revenue recognition in the second half of 2026. Cycle time approximately 9 months+\n\n(2) Agentic AI, Reasoning: Next level of token economics\n\n■ Recent 5x token demand growth triggered by Reasoning\n\n■ Agentic is not a specific feature but likely to become the default framework for future application development; combined with multimodal/large model proliferation, it will chain-drive compute demand in both training and inference\n\n■ Reasoning created 5x; Agentic will create 50x\n\n(3) AI Data Center: Era continues where generality is critical from TCO perspective\n\n■ CPX and SRAM-based approaches have clear strengths in specific segments (especially prefill, latency-sensitive workloads)\n\n■ However, as workloads rapidly shift to MoE/multimodal/diffusion/AR/SSM etc., bottlenecks move across NVLink/HBM/system-wide, worsening data center TCO\n\n■ CPX–SRAM is closer to a localized option for some workloads rather than a \"mainstream HBM replacement\"; overall platform strategy remains centered on general-purpose GPU/systems", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004699804739852764, "ret_signed_m1d": 0.004699804739852764, "alpha_spy_m1d": 0.010611828225168818, "alpha_c_m1d": 0.030575659715457282, "bench_spy_m1d": -0.005912023485316054, "bench_c_m1d": -0.025875854975604518, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009987125820996923, "ret_signed_p1d": 0.009987125820996923, "alpha_spy_p1d": 0.013210482613224328, "alpha_c_p1d": 0.016694741878650388, "bench_spy_p1d": -0.0032233567922274053, "bench_c_p1d": -0.006707616057653465, "ret_p1w": -0.007637304948690304, "ret_signed_p1w": -0.007637304948690304, "alpha_spy_p1w": -0.010470525517025986, "alpha_c_p1w": -0.01772451332638847, "bench_spy_p1w": 0.0028332205683356815, "bench_c_p1w": 0.010087208377698165, "ret_p1m": -0.06969668297352538, "ret_signed_p1m": -0.06969668297352538, "alpha_spy_p1m": -0.06157309135644695, "alpha_c_p1m": -0.05524952809488559, "bench_spy_p1m": -0.00812359161707843, "bench_c_p1m": -0.01444715487863979, "ret_p3m": -0.05255504410365941, "ret_signed_p3m": -0.05255504410365941, "alpha_spy_p3m": -0.0031355774207620923, "alpha_c_p3m": -0.06468035308213549, "bench_spy_p3m": -0.04941946668289732, "bench_c_p3m": 0.01212530897847608, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0284, -0.0218, 0.0057, -0.0386, -0.0396, -0.0291, -0.0215, -0.0246, -0.0325, -0.0372, -0.0272, -0.0053, 0.0226, 0.0736, 0.1057, 0.0835, 0.0814, 0.1048, 0.0611, 0.0425, 0.0044, 0.0048, 0.063, 0.0316, 0.035, -0.0021, 0.0156, -0.0035, -0.0315, -0.0039, -0.0353, -0.0447, -0.0251, -0.0504, -0.0373, -0.0373, -0.0547, -0.0391, -0.0309, -0.0409, -0.0206, -0.0258, -0.009, -0.0121, -0.0185, -0.0337, -0.0653, -0.0585, -0.0508, -0.0871, -0.07, -0.0334, -0.019, 0.0105, 0.0073, 0.0073, 0.0176, 0.0052, 0.0016, -0.004, -0.004, 0.0086, 0.0047, 0.0, 0.01, -0.0117, -0.0127, -0.0123, -0.0076, -0.0219, -0.001, -0.0054, -0.0054, -0.049, -0.0209, -0.0128, 0.0023, -0.0041, 0.0068, 0.0229, 0.0281, 0.0208, -0.0087, -0.0369, -0.0697, -0.082, -0.0098, 0.015, 0.0069, 0.015, -0.0016, -0.0237, -0.0237, -0.0121, 0.004, 0.0035, 0.0138, 0.023, 0.03, 0.0444, -0.0126, -0.0537, -0.0254, -0.0384, -0.0224, -0.0208, -0.0503, -0.0245, -0.0132, -0.0064, -0.0218, -0.0373, -0.0214, -0.0283, -0.0365, -0.0463, -0.0776, -0.0619, -0.0643, -0.0457, -0.0854, -0.1053, -0.1178, -0.0685, -0.0613, -0.0526, -0.0526, -0.0512, -0.0488, -0.0275, -0.0177, 0.0075, 0.0111, 0.0496, 0.0622, 0.0594, 0.0772, 0.0792, 0.0676, 0.0816, 0.0663, 0.1124, 0.1569, 0.1385, 0.1176, 0.0659, 0.0599, 0.0601, 0.0495, 0.11, 0.1296, 0.1494, 0.172, 0.1792, 0.2062, 0.2591, 0.2034, 0.1874, 0.1783, 0.1936, 0.1724, 0.1501, 0.1501, 0.1476, 0.1355, 0.1443, 0.1277, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008099914849702032", "ticker_idx": 0, "ts": "2026-01-05T08:55:03+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Global investment banks (IBs) such as Citi and Morgan Stanley have significantly raised their target prices and operating profit forecasts for Samsung Electronics", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares IBs' raised targets on Samsung and SK Hynix as server DRAM prices surge 60–70% into a memory super-cycle driven by AI/HBM demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Prices Hike 70%... Korean Stock Market Hits All-Time High\n\nSamsung Electronics and SK Hynix have proposed server DRAM prices up to 70% higher than the fourth quarter of last year to global cloud companies such as Microsoft (MS), Amazon Web Services (AWS), and Google. This surge is driven by the server DRAM market shifting into a complete \"seller's market,\" as AI accelerator manufacturers like Nvidia and Broadcom increase orders for HBM3E (5th generation High Bandwidth Memory)—which is made by stacking DRAM—and Big Tech companies expand their purchases of DRAM for inference AI. Market forecasts suggest that, driven by this \"memory super-cycle,\" the combined operating profit of Samsung Electronics and SK Hynix this year could reach 150 trillion won, a 2.5 to 3-fold increase compared to last year.\n\nAccording to industry sources on the 5th, Samsung Electronics and SK Hynix are reportedly negotiating with major clients after presenting server DRAM price quotes that are 60–70% higher than in the fourth quarter of last year. DRAM suppliers and clients typically sign quarterly supply contracts at the end of the previous quarter or the beginning of the respective quarter.\n\nSamsung and SK have significantly raised prices because they anticipate the server DRAM supply shortage will worsen. Memory manufacturers are focusing heavily on HBM3E production for Nvidia and AMD's AI accelerators (which the U.S. government has authorized for export to China), leaving them unable to produce enough server DRAM to meet demand. The explosive growth in demand for general-purpose DRAM used in servers, as Google (Gemini) and MS (Copilot) expand their inference AI services, has also contributed to the price hike.\n\nThe industry expects clients to accept the proposals from Samsung and SK. Market research firm DRAMeXchange analyzed, \"Large clients view the spending required for AI infrastructure as 'sufficiently bearable,'\" adding, \"There is an atmosphere of not strongly resisting the DRAM price increase, based on the judgment that the monetization of inference AI is more important.\"\n\nGlobal investment banks (IBs) such as Citi and Morgan Stanley have significantly raised their target prices and operating profit forecasts for Samsung Electronics and SK Hynix, reflecting this DRAM \"mega-boom.\" On this day, the KOSPI index closed at 4,457.52, up 3.43%, marking an all-time high driven by the rally of the \"two tops,\" Samsung Electronics (7.47%) and SK Hynix (2.81%).\n\n---\n\nhttps://t.co/SnjYvRGRVx", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06951485145820746, "ret_signed_m1d": -0.06951485145820746, "alpha_spy_m1d": -0.06289880128928216, "alpha_c_m1d": -0.0673022063897174, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005792875943975595, "ret_signed_p1d": 0.005792875943975595, "alpha_spy_p1d": -0.0001543074291137092, "alpha_c_p1d": -0.008243924637808986, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": 0.005068738106770221, "ret_signed_p1w": 0.005068738106770221, "alpha_spy_p1w": -0.005749621663828863, "alpha_c_p1w": -0.00993615411538884, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.21288915464418712, "ret_signed_p1m": 0.21288915464418712, "alpha_spy_p1m": 0.21025720473907072, "alpha_c_p1m": 0.23045234840005946, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": 0.29449133710288633, "ret_signed_p3m": 0.29449133710288633, "alpha_spy_p3m": 0.33825752704325784, "alpha_c_p3m": 0.3529957010157597, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3533, -0.3533, -0.3197, -0.3277, -0.3399, -0.3154, -0.296, -0.2945, -0.2931, -0.2974, -0.2895, -0.3046, -0.288, -0.265, -0.283, -0.2758, -0.2498, -0.2253, -0.1994, -0.2441, -0.2751, -0.2852, -0.2945, -0.2751, -0.2542, -0.2571, -0.2592, -0.2996, -0.2751, -0.2952, -0.3046, -0.2751, -0.3169, -0.3032, -0.2844, -0.2592, -0.2542, -0.2758, -0.2736, -0.2549, -0.247, -0.2426, -0.2189, -0.2109, -0.2189, -0.2217, -0.2268, -0.2153, -0.2448, -0.2592, -0.2225, -0.2246, -0.234, -0.2037, -0.1965, -0.1994, -0.1994, -0.1569, -0.1347, -0.1318, -0.1318, -0.1318, -0.0695, 0.0, 0.0058, 0.021, 0.0051, 0.0065, 0.0051, -0.0036, 0.0159, 0.042, 0.0782, 0.0811, 0.0514, 0.0825, 0.1028, 0.1014, 0.1014, 0.155, 0.176, 0.1636, 0.1622, 0.0891, 0.2129, 0.2245, 0.1535, 0.1484, 0.2049, 0.2006, 0.2151, 0.2933, 0.3121, 0.3121, 0.3121, 0.3121, 0.3758, 0.3765, 0.3975, 0.4482, 0.4736, 0.5786, 0.5677, 0.5677, 0.4127, 0.2469, 0.3874, 0.3628, 0.2563, 0.3606, 0.3758, 0.3606, 0.3287, 0.3664, 0.4041, 0.5098, 0.4518, 0.4439, 0.349, 0.3736, 0.3686, 0.3041, 0.3041, 0.2793, 0.2132, 0.3761, 0.2945, 0.3511, 0.4012, 0.4258, 0.5274, 0.4802, 0.4948, 0.4585, 0.4984, 0.531, 0.5782, 0.5673, 0.5564, 0.5891, 0.5782, 0.629, 0.5927, 0.629, 0.6109, 0.6399, 0.6, 0.6, 0.687, 0.687, 0.9301, 0.97, 0.9483, 1.0716, 1.0245, 1.0607, 1.1478, 0.9628, 1.039, 0.9991, 1.0027, 1.1732, 1.1224, 1.1224, 1.1696, 1.2276, 1.1732, 1.3002, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008099914849702032", "ticker_idx": 1, "ts": "2026-01-05T08:55:03+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Global investment banks (IBs) such as Citi and Morgan Stanley have significantly raised their target prices and operating profit forecasts for Samsung Electronics and SK Hynix", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares IBs' raised targets on Samsung and SK Hynix as server DRAM prices surge 60–70% into a memory super-cycle driven by AI/HBM demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Prices Hike 70%... Korean Stock Market Hits All-Time High\n\nSamsung Electronics and SK Hynix have proposed server DRAM prices up to 70% higher than the fourth quarter of last year to global cloud companies such as Microsoft (MS), Amazon Web Services (AWS), and Google. This surge is driven by the server DRAM market shifting into a complete \"seller's market,\" as AI accelerator manufacturers like Nvidia and Broadcom increase orders for HBM3E (5th generation High Bandwidth Memory)—which is made by stacking DRAM—and Big Tech companies expand their purchases of DRAM for inference AI. Market forecasts suggest that, driven by this \"memory super-cycle,\" the combined operating profit of Samsung Electronics and SK Hynix this year could reach 150 trillion won, a 2.5 to 3-fold increase compared to last year.\n\nAccording to industry sources on the 5th, Samsung Electronics and SK Hynix are reportedly negotiating with major clients after presenting server DRAM price quotes that are 60–70% higher than in the fourth quarter of last year. DRAM suppliers and clients typically sign quarterly supply contracts at the end of the previous quarter or the beginning of the respective quarter.\n\nSamsung and SK have significantly raised prices because they anticipate the server DRAM supply shortage will worsen. Memory manufacturers are focusing heavily on HBM3E production for Nvidia and AMD's AI accelerators (which the U.S. government has authorized for export to China), leaving them unable to produce enough server DRAM to meet demand. The explosive growth in demand for general-purpose DRAM used in servers, as Google (Gemini) and MS (Copilot) expand their inference AI services, has also contributed to the price hike.\n\nThe industry expects clients to accept the proposals from Samsung and SK. Market research firm DRAMeXchange analyzed, \"Large clients view the spending required for AI infrastructure as 'sufficiently bearable,'\" adding, \"There is an atmosphere of not strongly resisting the DRAM price increase, based on the judgment that the monetization of inference AI is more important.\"\n\nGlobal investment banks (IBs) such as Citi and Morgan Stanley have significantly raised their target prices and operating profit forecasts for Samsung Electronics and SK Hynix, reflecting this DRAM \"mega-boom.\" On this day, the KOSPI index closed at 4,457.52, up 3.43%, marking an all-time high driven by the rally of the \"two tops,\" Samsung Electronics (7.47%) and SK Hynix (2.81%).\n\n---\n\nhttps://t.co/SnjYvRGRVx", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.027298908880455053, "ret_signed_m1d": -0.027298908880455053, "alpha_spy_m1d": -0.02068285871152975, "alpha_c_m1d": -0.015937356438575967, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0431033745856777, "ret_signed_p1d": 0.0431033745856777, "alpha_spy_p1d": 0.037156191212588396, "alpha_c_p1d": 0.016540174086262516, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.07614937008435851, "ret_signed_p1w": 0.07614937008435851, "alpha_spy_p1w": 0.06533101031375943, "alpha_c_p1w": 0.04153508620242308, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.30316086614716653, "ret_signed_p1m": 0.30316086614716653, "alpha_spy_p1m": 0.30052891624205014, "alpha_c_p1m": 0.2499550171237912, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.19472924735399255, "ret_signed_p3m": 0.19472924735399255, "alpha_spy_p3m": 0.23849543729436407, "alpha_c_p3m": 0.15571865086258874, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4322, -0.4322, -0.3855, -0.4042, -0.4092, -0.3934, -0.3503, -0.3317, -0.3029, -0.3123, -0.3087, -0.313, -0.2678, -0.2319, -0.252, -0.1988, -0.1845, -0.1974, -0.1098, -0.1586, -0.1687, -0.1486, -0.1673, -0.1299, -0.1113, -0.1141, -0.1213, -0.196, -0.1299, -0.1816, -0.1931, -0.1802, -0.252, -0.2534, -0.2548, -0.2477, -0.2184, -0.2385, -0.227, -0.1983, -0.2069, -0.2213, -0.2184, -0.171, -0.1868, -0.1566, -0.1882, -0.1796, -0.204, -0.2385, -0.2083, -0.2069, -0.2141, -0.1667, -0.1609, -0.1552, -0.1552, -0.1394, -0.0805, -0.0647, -0.0647, -0.0647, -0.0273, 0.0, 0.0431, 0.0661, 0.0862, 0.069, 0.0761, 0.0603, 0.0661, 0.0761, 0.0862, 0.0977, 0.0675, 0.0632, 0.0848, 0.102, 0.0575, 0.1494, 0.2083, 0.2371, 0.306, 0.1925, 0.3032, 0.2931, 0.2098, 0.2055, 0.2744, 0.2586, 0.2356, 0.2759, 0.2644, 0.2644, 0.2644, 0.2644, 0.2845, 0.3635, 0.3664, 0.444, 0.4626, 0.5819, 0.5272, 0.5272, 0.3516, 0.2221, 0.3545, 0.33, 0.2034, 0.3502, 0.3747, 0.3387, 0.3099, 0.402, 0.3962, 0.52, 0.4581, 0.4495, 0.343, 0.4193, 0.4322, 0.343, 0.343, 0.2566, 0.1616, 0.2912, 0.1947, 0.2609, 0.2753, 0.3185, 0.4869, 0.4366, 0.4783, 0.497, 0.5877, 0.6352, 0.6625, 0.6237, 0.6784, 0.7619, 0.7604, 0.7633, 0.759, 0.8597, 0.8713, 0.8612, 0.8511, 0.8511, 1.0829, 1.0829, 1.3045, 1.3808, 1.4269, 1.7061, 1.6414, 1.8443, 1.8357, 1.6183, 1.6486, 1.5118, 1.5118, 1.7925, 1.7939, 1.7939, 1.9537, 2.2286, 2.2954, 2.3588, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011582419783336029", "ticker_idx": 0, "ts": "2026-01-14T23:33:16+00:00", "symbol": "Nitto Boseki", "kind": "equity", "geography": "Japan", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Currently, high-end glass fabric is supplied almost entirely by Japan's Nitto Boseki (Nittobo), giving it absolute influence in the global supply chain", "tweet_type": "analysis", "timeline": "years", "summary": "Long Nitto Boseki as the dominant supplier of high-end glass fabric with absolute pricing power; structural shortage persists until 2H27 when new capacity comes online.", "resolved_tickers": ["3110.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Nitto Boseki (Nittobo) listed on TSE as 3110.T", "bench_c": "EWJ", "bench_c_reason": "country_fallback (suffix='T')", "bench_c_industry": "Textile Manufacturing", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Fiberglass: Will the supply shortage only ease in 2H27?\n\n- Glass fabric is an essential core material used in iPhones. As the supply shortage intensifies, Apple has officially entered the competition to secure components alongside global Big Tech companies such as Nvidia, Google, and Amazon.\n\n- According to industry outlooks, the glass fabric supply shortage is expected to persist until at least the second half of 2027, which is evaluated as a structural bottleneck reminiscent of the memory semiconductor supply shortage phase.\n\n- Currently, high-end glass fabric is supplied almost entirely by Japan's Nitto Boseki (Nittobo), giving it absolute influence in the global supply chain.\n\n- As major manufacturers such as Nvidia, Google, and Amazon widely adopt high-specification substrate materials, the supply-demand imbalance of high-quality glass fabric is rapidly worsening.\n\n- This material shortage shows a pattern very similar to recent rapid price fluctuations in the memory market and is emerging as a key risk across the industry.\n\n- Recently, Apple, AMD, and Nvidia dispatched personnel to Japan to secure the supply of glass fabric, but they failed to obtain actual additional volume.\n\n- A senior executive of a global substrate manufacturer mentioned that, regardless of customer pressure, the current production capacity of leading manufacturer Nittobo has reached its limit, making it impossible to secure additional volume.\n\n- From the current perspective, the supply shortage is expected to improve only around the second half of 2027, when Nittobo’s new production facilities begin full-scale operation.\n\n- Apple is actively building alternative supply chains to mitigate supply risks.\n\n- According to sources, Apple has dispatched a dedicated team to GFT, a small glass fabric factory in China, and requested technical support from Mitsubishi Gas Chemical to monitor and improve the factory's product quality.\n\n- This material shortage is acting as a factor prompting many new companies to enter the market to capture business opportunities.\n\n- In addition to the Taiwanese manufacturer Taiwan Glass, Chinese companies such as Taishan Fiberglass, Hung Ho Electronics, and Kingboard Laminates are also expanding investments in the market to benefit from the global supply chain imbalance.\n\n---\n\nhttps://t.co/Lf0K1vBTvH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.035222046660384865, "ret_signed_m1d": -0.035222046660384865, "alpha_spy_m1d": -0.04016156289665751, "alpha_c_m1d": -0.02828408851206332, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": -0.0069379581483215436, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019908126394606596, "ret_signed_p1d": 0.019908126394606596, "alpha_spy_p1d": 0.017184911998077146, "alpha_c_p1d": 0.015439673321617775, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.0044684530729888206, "ret_p1w": 0.2320061324192999, "ret_signed_p1w": 0.2320061324192999, "alpha_spy_p1w": 0.23919076241923254, "alpha_c_p1w": 0.23506354226679904, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": -0.0030574098474991374, "ret_p1m": 0.5666157259384315, "ret_signed_p1m": 0.5666157259384315, "alpha_spy_p1m": 0.5797826865162727, "alpha_c_p1m": 0.46866180141614566, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.0979539245222858, "ret_p3m": 1.227767782697566, "ret_signed_p3m": 1.227767782697566, "alpha_spy_p3m": 1.2312243270468297, "alpha_c_p3m": 1.1888449519948514, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.03892283070271452, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4242, -0.3951, -0.4158, -0.4349, -0.4395, -0.3905, -0.3783, -0.389, -0.3683, -0.3476, -0.33, -0.33, -0.3392, -0.3729, -0.366, -0.2511, -0.1363, -0.0613, 0.0337, 0.0429, 0.0245, 0.0911, 0.0475, 0.0352, 0.1608, 0.0972, 0.0972, 0.0842, 0.0168, 0.0368, 0.0459, 0.0421, -0.0421, -0.0926, -0.1103, -0.059, -0.0919, -0.1256, -0.0995, -0.1217, -0.111, -0.1516, -0.1968, -0.1876, -0.2052, -0.2175, -0.1723, -0.1784, -0.1838, -0.1914, -0.219, -0.2067, -0.219, -0.219, -0.219, -0.219, -0.1983, -0.1922, -0.1815, -0.0904, -0.0398, -0.0398, -0.0352, 0.0, 0.0199, 0.1003, 0.1891, 0.1715, 0.232, 0.3461, 0.307, 0.2029, 0.1822, 0.1715, 0.1348, 0.1355, 0.0972, 0.1639, 0.255, 0.3155, 0.294, 0.3637, 0.4387, 0.4387, 0.5666, 0.5459, 0.5858, 0.6348, 0.6194, 0.6271, 0.6233, 0.6233, 1.0061, 1.1194, 0.9502, 0.9296, 0.9242, 0.7864, 0.6646, 0.7764, 0.8231, 0.5835, 0.621, 0.7136, 0.6616, 0.6953, 0.6753, 0.5926, 0.7075, 0.6524, 0.6524, 0.5406, 0.5391, 0.7121, 0.67, 0.6462, 0.5201, 0.4416, 0.6054, 0.5085, 0.5754, 0.597, 0.6662, 0.9116, 0.9455, 1.1101, 1.2278, 1.2647, 1.0385, 1.0801, 1.1162, 1.0555, 1.0355, 1.1124, 0.9655, 1.1062, 1.1039, 1.1701, 1.1701, 1.1955, 1.1662, 1.1662, 1.1662, 1.1662, 1.4578, 1.3385, 1.2378, 1.2308, 0.9839, 1.0147, 0.8385, 0.7485, 0.6601, 0.6224, 0.6939, 0.947, 1.1201, 0.9916, 0.7924, 0.747, 0.7439, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025475383660085313", "ticker_idx": 0, "ts": "2026-02-22T07:38:57+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "My estimates are FQ4 revenue of $67.0–67.5B and FQ1 guidance of $75.0B", "tweet_type": "prediction", "timeline": "weeks", "summary": "Own NVDA FQ4 revenue estimate $67.0–67.5B above $65.7B consensus; FQ1 guidance estimate $75.0B above $70.8B consensus; implied bullish into the print.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "What do you think NVIDIA's earnings will look like? \nMy estimates are FQ4 revenue of $67.0–67.5B and FQ1 guidance of $75.0B. \nNVIDIA's own FQ4 guidance was $65.0B, and the current consensus sits above that at ~$65.7B. Consensus for FQ1 revenue guidance is $70.8B.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009031572544070499, "ret_signed_m1d": -0.009031572544070499, "alpha_spy_m1d": -0.01934823393113405, "alpha_c_m1d": -0.014238881861688424, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006786724366676156, "ret_signed_p1d": 0.006786724366676156, "alpha_spy_p1d": -0.00048172856371886397, "alpha_c_p1d": -0.008423503072357397, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": -0.04735061072712243, "ret_signed_p1w": -0.04735061072712243, "alpha_spy_p1w": -0.05319769644001404, "alpha_c_p1w": -0.031631781629992695, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": -0.0853068596486728, "ret_signed_p1m": -0.0853068596486728, "alpha_spy_p1m": -0.04511547732421006, "alpha_c_p1m": -0.041008253931224425, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 0.1460290508639468, "ret_signed_p3m": 0.1460290508639468, "alpha_spy_p3m": 0.05464683402616233, "alpha_c_p3m": -0.22938268660683137, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.059, -0.059, -0.076, -0.0608, -0.0527, -0.0625, -0.0427, -0.0477, -0.0313, -0.0344, -0.0406, -0.0554, -0.0863, -0.0797, -0.0722, -0.1076, -0.0909, -0.0551, -0.041, -0.0122, -0.0153, -0.0153, -0.0053, -0.0174, -0.0209, -0.0264, -0.0264, -0.0141, -0.0179, -0.0225, -0.0127, -0.034, -0.0349, -0.0345, -0.03, -0.0439, -0.0235, -0.0278, -0.0278, -0.0704, -0.043, -0.035, -0.0203, -0.0265, -0.0158, -0.0002, 0.005, -0.0022, -0.031, -0.0585, -0.0906, -0.1027, -0.0321, -0.0079, -0.0157, -0.0078, -0.0241, -0.0456, -0.0456, -0.0344, -0.0186, -0.0191, -0.009, 0.0, 0.0068, 0.0209, -0.0348, -0.075, -0.0474, -0.06, -0.0444, -0.0429, -0.0717, -0.0465, -0.0354, -0.0288, -0.0439, -0.0589, -0.0434, -0.0502, -0.0582, -0.0678, -0.0984, -0.083, -0.0853, -0.0671, -0.106, -0.1254, -0.1377, -0.0895, -0.0824, -0.0739, -0.0739, -0.0726, -0.0702, -0.0494, -0.0398, -0.0152, -0.0116, 0.0259, 0.0383, 0.0356, 0.0529, 0.0549, 0.0435, 0.0572, 0.0423, 0.0873, 0.1309, 0.1129, 0.0925, 0.0419, 0.0361, 0.0362, 0.0259, 0.085, 0.1042, 0.1235, 0.1457, 0.1527, 0.179, 0.2308, 0.1764, 0.1607, 0.1518, 0.1667, 0.146, 0.1242, 0.1242, 0.1218, 0.11, 0.1186, 0.1023, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015675123450425798", "ticker_idx": 0, "ts": "2026-01-26T06:36:13+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we maintain our Overweight rating, driven by Broadcom's overwhelming technological superiority and market leadership. We recommend active accumulation at current price levels.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares JPM Overweight on AVGO with active accumulation rec; AI ASIC backlog strengthening for FY2026-27 on Google TPU dominance and Meta programs.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM TPU Commentary: \n\nStrengthening AI ASIC Outlook for FY2026-2027 Broadcom’s outlook for its AI ASIC business in fiscal years 2026 and 2027 is strengthening, driven by deepened collaboration with its largest AI customer, Google. Broadcom maintains a technical lead of over 18 months compared to Google's internal COT (Customer Owned Tooling) program. This gap is becoming even more entrenched as the complexity of chip design and packaging intensifies. To support AI workloads for external clients such as Anthropic, OpenAI, and Apple, Google is expanding its TPU deployment, with plans to ship 6 to 7 million units in 2027—surpassing previous market expectations. Crucially, over 95% of this volume will be powered by Broadcom’s next-generation 3nm chip, 'Sunfish' (codename Hellcat), underscoring Broadcom’s formidable market dominance.\n\nDelays in Internal Programs and Supply Chain AdvantagesIn contrast, Google’s internal chip program, 'Zebrafish', has faced multiple delays and is expected to secure its first prototypes only by the end of this quarter or early next. This timeline lags at least 18 months behind Broadcom’s Sunfish, which completed tape-out in the first half of last year and has already secured a multi-billion dollar backlog. Furthermore, given the extremely tight supply chain for HBM and advanced packaging (CoWoS), it is highly unlikely that Google’s late-entry internal chip will secure meaningful volume within 2027. Broadcom has already commenced design work for Google’s next-gen 2nm TPU aimed at 2028 production, maintaining a 1.5-generation lead over both competitors and internal initiatives.\n\nInvestment Thesis: Market Underestimation and Overweight RatingThe market continues to underestimate Broadcom’s 30-plus years of ASIC leadership, its vast IP portfolio, and its proven execution in supporting Google’s cutting-edge silicon development. Additionally, Meta’s 3nm and 2nm programs are progressing as planned, contributing to a steadily growing AI-related backlog. Consequently, we maintain our Overweight rating, driven by Broadcom’s overwhelming technological superiority and market leadership. We recommend active accumulation at current price levels.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014776036531097159, "ret_signed_m1d": -0.014776036531097159, "alpha_spy_m1d": -0.009723546756584489, "alpha_c_m1d": -0.017960402887721383, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.024442117705581135, "ret_signed_p1d": 0.024442117705581135, "alpha_spy_p1d": 0.02045785297589653, "alpha_c_p1d": 0.0033047812606517013, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.019270406071997703, "ret_signed_p1w": 0.019270406071997703, "alpha_spy_p1w": 0.015401613191919461, "alpha_c_p1w": -0.0036722591114115577, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.0019700817197465526, "ret_signed_p1m": 0.0019700817197465526, "alpha_spy_p1m": 0.009736469180722906, "alpha_c_p1m": -0.04903035949271084, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.2954314852704689, "ret_signed_p3m": 0.2954314852704689, "alpha_spy_p3m": 0.26994560234461185, "alpha_c_p3m": 0.08724233413844451, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1859, 0.1567, 0.1357, 0.1139, 0.0813, 0.103, 0.0925, 0.0736, 0.1011, 0.0814, 0.0914, 0.0446, 0.0522, 0.0528, 0.0462, 0.0889, 0.0656, 0.0453, 0.1613, 0.183, 0.2215, 0.2215, 0.2381, 0.1862, 0.1724, 0.1694, 0.1707, 0.199, 0.2324, 0.2483, 0.2688, 0.2486, 0.1059, 0.0441, 0.0486, 0.0017, 0.0135, 0.0457, 0.0511, 0.0753, 0.0781, 0.0781, 0.084, 0.0755, 0.077, 0.0654, 0.0654, 0.0701, 0.0572, 0.0582, 0.0574, 0.0235, 0.0619, 0.0842, 0.0916, 0.0463, 0.0559, 0.0827, 0.0827, 0.0239, 0.0122, 0.002, -0.0148, 0.0, 0.0244, 0.0258, 0.0181, 0.0199, 0.0193, -0.0139, -0.0517, -0.0441, 0.0248, 0.0588, 0.048, 0.0551, 0.0195, 0.001, 0.001, 0.0237, 0.0267, 0.0281, 0.024, 0.0169, 0.002, 0.023, -0.0097, -0.0163, -0.0186, -0.0339, -0.0225, 0.0244, 0.0173, 0.0643, 0.0546, 0.0515, 0.0342, -0.0083, 0.0002, -0.0109, -0.0275, -0.0154, -0.0441, -0.0051, -0.0181, -0.0165, -0.0455, -0.0725, -0.0949, -0.0452, -0.0329, -0.0297, -0.0297, -0.03, 0.0302, 0.0816, 0.0948, 0.1462, 0.1715, 0.1746, 0.2238, 0.2292, 0.2541, 0.2328, 0.2406, 0.3038, 0.2954, 0.3041, 0.2901, 0.2334, 0.2507, 0.2877, 0.2996, 0.2848, 0.3183, 0.3124, 0.2727, 0.3265, 0.3216, 0.2935, 0.2857, 0.3567, 0.3116, 0.2978, 0.2681, 0.2887, 0.2789, 0.2775, 0.2775, 0.3018, 0.3014, 0.3159, 0.3782, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012045084146221340", "ticker_idx": 1, "ts": "2026-01-16T06:11:44+00:00", "symbol": "ASMPT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "vs potential share gainers ASMPT (Buy)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Sell Hanmi on HBM4 TCB share loss; Buy ASMPT/BESI as share gainers; sentiment positive for Teradyne, Advantest, Disco on HBM4 back-end spend.", "resolved_tickers": ["0522.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "ASMPT listed on Hong Kong Exchange 0522.HK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: regarding HBM4 TCB bonding equipment\n\nHanmi, Hanwha may have both secured HBM4 TCB tool orders from SK Hynix\n\nPre-market yesterday, Hanmi disclosed having received HBM thermo-compression bonding (TCB) tool orders from SK Hynix (link) for a contract amount of Won9.7bn (ex. VAT) and duration till the end of 1Q26E. Given the size of order, we believe this may have been for c.3 units of HBM4 TCB tools. Based on our channel checks and industry conversations, we believe competitor Hanwha Semitech may have also secured TCB tool orders ‘of equivalent scale’ or c.3 units for HBM4 12Hi, suggesting a 50/50 split. With ASMPT also seeing renewed engagement to supply HBM TCB tools to SK Hynix (link, link), we believe all three vendors will be engaged in supplying HBM4 TCB tools to SK Hynix in 2026. Given our expectation of 2026 HBM4 TCB unit TAM for SK Hynix at c. 86 tools, we believe follow-up TCB orders may come through shortly (within 1Q26) as SK Hynix continues to ramp up capacity at M15X. Following orders for HBM TCB tools, we expect visibility may also soon extend to other areas of HBM back-end including dicing/grinding, descum, test etc.\n\nHBM4 TCB tool specs largely unchanged vs HBM3E; any ASP uplift limited\n\nIn the case for HBM4 12Hi, SK Hynix will continue to adopt a flux-type TCB tool (based on advanced MR-MUF) – with TCB tool specs, we believe, largely unchanged in versus HBM3E 12Hi. This was enabled, we believe, due to SK Hynix in particular having been able to meaningfully improve/control TSV back-end yields for HBM3E allowing for extension of the same process/tools to HBM4. This suggests to us that 1/ we may see a more competitive set up for HBM4 TCB tools (lower entry barrier + time for competitors to catch up on offering) but also 2/ any potential ASP uplift on tool spec improvement for HBM4 may be challenging. This may also extend beyond MR-MUF type TCB tools to potentially NCF type TCB tools, as well where we believe ASMPT may be set to gain share at Micron at the expense of Hanmi.\n\nUpdate on fluxless TCB and hybrid bonding adoption timing\n\nIndustry preparation for fluxless and hybrid bonding continues to progress though adoption timing remains uncertain. For SK Hynix, we believe advanced MR-MUF will remain the mainstream packaging solution till HBM4E (base scenario flux TCB while keeping fluxless ‘optionality’). Hybrid bonding adoption for SK Hynix, we believe, may only occur afterwards from HBM5 with Hanmi, Hanwha and BESI all engaged. For Hanwha, we believe the company is set to ship their 2nd gen hybrid bonder tools for eval to SK Hynix within 1H26. In the case for Micron, we believe the company is actively weighing the option of adopting fluxless TCB starting HBM4E, with Hanmi, ASMPT and BESI engaged. For Samsung, we believe both fluxless TCB and hybrid bonding tracks are in consideration for HBM4E with SEMES and BESI engaged for hybrid bonding.\n\nStock implications\n\nIntensifying competition for HBM4 TCB tools into 2026 bodes negatively for TCB tool ASP/margins. Within UBS coverage, we see Hanmi (Sell) as most exposed given market share loss off high base vs potential share gainers ASMPT (Buy) / BESI (Buy). Broadening visibility on HBM4 back-end spend should be sentiment positive for test tool vendors Teradyne (Buy), Advantest (Neutral) as well as dicing/grinding vendors Disco (Neutral).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026060257030865897, "ret_signed_m1d": -0.026060257030865897, "alpha_spy_m1d": -0.026898833643294262, "alpha_c_m1d": -0.01611992175596455, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004598873504373269, "ret_signed_p1d": -0.004598873504373269, "alpha_spy_p1d": -0.004598873504373269, "alpha_c_p1d": -0.004598873504373269, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.06489522372635115, "ret_signed_p1w": 0.06489522372635115, "alpha_spy_p1w": 0.06840853853024131, "alpha_c_p1w": 0.06564453889299016, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.05263158720860517, "ret_signed_p1m": 0.05263158720860517, "alpha_spy_p1m": 0.06695941677642536, "alpha_c_p1m": 0.03432447088678159, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.29586104523876355, "ret_signed_p3m": 0.29586104523876355, "alpha_spy_p3m": 0.28112609411414446, "alpha_c_p3m": 0.16446419859722905, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1226, -0.1461, -0.1492, -0.0802, -0.0419, -0.1083, -0.1083, -0.1431, -0.162, -0.1559, -0.1656, -0.1666, -0.1589, -0.1707, -0.1584, -0.1732, -0.1896, -0.1814, -0.2039, -0.2034, -0.2233, -0.233, -0.2248, -0.2724, -0.2759, -0.2509, -0.2591, -0.2432, -0.2305, -0.2182, -0.2187, -0.2095, -0.2024, -0.1967, -0.2044, -0.1916, -0.2121, -0.1942, -0.1952, -0.2315, -0.2361, -0.2238, -0.2289, -0.2402, -0.2095, -0.2085, -0.209, -0.209, -0.209, -0.2218, -0.2105, -0.2059, -0.2059, -0.1702, -0.1599, -0.1288, -0.0751, -0.0904, -0.0807, -0.0797, -0.0465, -0.0358, -0.0261, 0.0, -0.0046, -0.0026, 0.0404, 0.1058, 0.0649, 0.0506, 0.0864, 0.1191, 0.0649, 0.0618, 0.0204, 0.0649, 0.0537, -0.0158, 0.0061, 0.0322, 0.0312, 0.0731, 0.068, 0.0649, 0.0526, 0.0526, 0.0526, 0.0526, 0.072, 0.1272, 0.1405, 0.1293, 0.165, 0.1456, 0.1078, 0.0567, 0.1037, 0.1487, 0.1354, 0.0976, 0.1446, 0.1303, 0.117, 0.1058, 0.0986, 0.0608, 0.1139, 0.0966, 0.0751, 0.0485, 0.0659, 0.1191, 0.0956, 0.0864, 0.0526, 0.0148, 0.0506, 0.0281, 0.0281, 0.0281, 0.0281, 0.1344, 0.1272, 0.1671, 0.2029, 0.2826, 0.2959, 0.3572, 0.4113, 0.4165, 0.5135, 0.5554, 0.5789, 0.6883, 0.6944, 0.7353, 0.6811, 0.6668, 0.6668, 0.6873, 0.6689, 0.7741, 0.8109, 0.7261, 0.7721, 0.8125, 0.8753, 0.7724, 0.7693, 0.8146, 0.7405, 0.8403, 0.7992, 0.8989, 0.8989, 1.1047, 1.0964, 1.0543, 0.9956, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012045084146221340", "ticker_idx": 2, "ts": "2026-01-16T06:11:44+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BESI (Buy)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Sell Hanmi on HBM4 TCB share loss; Buy ASMPT/BESI as share gainers; sentiment positive for Teradyne, Advantest, Disco on HBM4 back-end spend.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: regarding HBM4 TCB bonding equipment\n\nHanmi, Hanwha may have both secured HBM4 TCB tool orders from SK Hynix\n\nPre-market yesterday, Hanmi disclosed having received HBM thermo-compression bonding (TCB) tool orders from SK Hynix (link) for a contract amount of Won9.7bn (ex. VAT) and duration till the end of 1Q26E. Given the size of order, we believe this may have been for c.3 units of HBM4 TCB tools. Based on our channel checks and industry conversations, we believe competitor Hanwha Semitech may have also secured TCB tool orders ‘of equivalent scale’ or c.3 units for HBM4 12Hi, suggesting a 50/50 split. With ASMPT also seeing renewed engagement to supply HBM TCB tools to SK Hynix (link, link), we believe all three vendors will be engaged in supplying HBM4 TCB tools to SK Hynix in 2026. Given our expectation of 2026 HBM4 TCB unit TAM for SK Hynix at c. 86 tools, we believe follow-up TCB orders may come through shortly (within 1Q26) as SK Hynix continues to ramp up capacity at M15X. Following orders for HBM TCB tools, we expect visibility may also soon extend to other areas of HBM back-end including dicing/grinding, descum, test etc.\n\nHBM4 TCB tool specs largely unchanged vs HBM3E; any ASP uplift limited\n\nIn the case for HBM4 12Hi, SK Hynix will continue to adopt a flux-type TCB tool (based on advanced MR-MUF) – with TCB tool specs, we believe, largely unchanged in versus HBM3E 12Hi. This was enabled, we believe, due to SK Hynix in particular having been able to meaningfully improve/control TSV back-end yields for HBM3E allowing for extension of the same process/tools to HBM4. This suggests to us that 1/ we may see a more competitive set up for HBM4 TCB tools (lower entry barrier + time for competitors to catch up on offering) but also 2/ any potential ASP uplift on tool spec improvement for HBM4 may be challenging. This may also extend beyond MR-MUF type TCB tools to potentially NCF type TCB tools, as well where we believe ASMPT may be set to gain share at Micron at the expense of Hanmi.\n\nUpdate on fluxless TCB and hybrid bonding adoption timing\n\nIndustry preparation for fluxless and hybrid bonding continues to progress though adoption timing remains uncertain. For SK Hynix, we believe advanced MR-MUF will remain the mainstream packaging solution till HBM4E (base scenario flux TCB while keeping fluxless ‘optionality’). Hybrid bonding adoption for SK Hynix, we believe, may only occur afterwards from HBM5 with Hanmi, Hanwha and BESI all engaged. For Hanwha, we believe the company is set to ship their 2nd gen hybrid bonder tools for eval to SK Hynix within 1H26. In the case for Micron, we believe the company is actively weighing the option of adopting fluxless TCB starting HBM4E, with Hanmi, ASMPT and BESI engaged. For Samsung, we believe both fluxless TCB and hybrid bonding tracks are in consideration for HBM4E with SEMES and BESI engaged for hybrid bonding.\n\nStock implications\n\nIntensifying competition for HBM4 TCB tools into 2026 bodes negatively for TCB tool ASP/margins. Within UBS coverage, we see Hanmi (Sell) as most exposed given market share loss off high base vs potential share gainers ASMPT (Buy) / BESI (Buy). Broadening visibility on HBM4 back-end spend should be sentiment positive for test tool vendors Teradyne (Buy), Advantest (Neutral) as well as dicing/grinding vendors Disco (Neutral).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004337761383337746, "ret_signed_m1d": 0.004337761383337746, "alpha_spy_m1d": 0.0034991847709093804, "alpha_c_m1d": 0.014278096658239092, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023713107403376843, "ret_signed_p1d": -0.023713107403376843, "alpha_spy_p1d": -0.023713107403376843, "alpha_c_p1d": -0.023713107403376843, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.012434963330090643, "ret_signed_p1w": 0.012434963330090643, "alpha_spy_p1w": 0.01594827813398081, "alpha_c_p1w": 0.013184278496729651, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.032388718978531905, "ret_signed_p1m": 0.032388718978531905, "alpha_spy_p1m": 0.0467165485463521, "alpha_c_p1m": 0.014081602656708325, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.2585309869659347, "ret_signed_p3m": 0.2585309869659347, "alpha_spy_p3m": 0.2437960358413156, "alpha_c_p3m": 0.12713414032440018, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1654, -0.199, -0.1657, -0.1495, -0.144, -0.1521, -0.146, -0.1495, -0.1463, -0.1567, -0.1755, -0.1816, -0.2033, -0.2241, -0.2108, -0.203, -0.2172, -0.2261, -0.24, -0.247, -0.2646, -0.2533, -0.2478, -0.2898, -0.2805, -0.2794, -0.2504, -0.2487, -0.2487, -0.2496, -0.2403, -0.2137, -0.2047, -0.1929, -0.1507, -0.1712, -0.1877, -0.2082, -0.236, -0.234, -0.2273, -0.2522, -0.2397, -0.2438, -0.2412, -0.2371, -0.2392, -0.2392, -0.2392, -0.236, -0.2302, -0.2264, -0.2264, -0.1377, -0.1108, -0.0694, -0.0815, -0.1261, -0.1261, -0.0622, -0.033, -0.0659, 0.0043, 0.0, -0.0237, 0.0, 0.0067, 0.0121, 0.0124, 0.0113, 0.0165, -0.02, -0.0599, -0.0486, -0.048, -0.0737, -0.0888, -0.0703, -0.0428, -0.0318, -0.0223, -0.0116, -0.0243, 0.0191, 0.0324, 0.0416, 0.0816, 0.0006, 0.0671, 0.0853, 0.1122, 0.1429, 0.0925, 0.0954, 0.0902, 0.0489, 0.1047, 0.0911, -0.096, -0.0382, 0.0098, 0.0119, 0.0136, 0.0706, 0.0584, 0.0899, 0.1128, 0.0633, 0.0324, 0.0558, 0.0648, 0.0729, 0.0723, 0.0113, 0.0075, 0.035, 0.0847, 0.0998, 0.0998, 0.0998, 0.1067, 0.1972, 0.192, 0.2233, 0.1966, 0.2608, 0.2585, 0.2753, 0.3117, 0.3129, 0.3181, 0.3436, 0.3985, 0.4586, 0.4434, 0.3695, 0.3933, 0.4387, 0.4387, 0.4219, 0.4783, 0.4841, 0.4824, 0.5179, 0.5208, 0.4539, 0.5028, 0.5522, 0.5237, 0.4888, 0.4836, 0.5423, 0.5738, 0.5924, 0.6413, 0.6529, 0.6087, 0.6739, 0.6553, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012045084146221340", "ticker_idx": 3, "ts": "2026-01-16T06:11:44+00:00", "symbol": "TER", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Broadening visibility on HBM4 back-end spend should be sentiment positive for test tool vendors Teradyne (Buy)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Sell Hanmi on HBM4 TCB share loss; Buy ASMPT/BESI as share gainers; sentiment positive for Teradyne, Advantest, Disco on HBM4 back-end spend.", "resolved_tickers": ["TER"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: regarding HBM4 TCB bonding equipment\n\nHanmi, Hanwha may have both secured HBM4 TCB tool orders from SK Hynix\n\nPre-market yesterday, Hanmi disclosed having received HBM thermo-compression bonding (TCB) tool orders from SK Hynix (link) for a contract amount of Won9.7bn (ex. VAT) and duration till the end of 1Q26E. Given the size of order, we believe this may have been for c.3 units of HBM4 TCB tools. Based on our channel checks and industry conversations, we believe competitor Hanwha Semitech may have also secured TCB tool orders ‘of equivalent scale’ or c.3 units for HBM4 12Hi, suggesting a 50/50 split. With ASMPT also seeing renewed engagement to supply HBM TCB tools to SK Hynix (link, link), we believe all three vendors will be engaged in supplying HBM4 TCB tools to SK Hynix in 2026. Given our expectation of 2026 HBM4 TCB unit TAM for SK Hynix at c. 86 tools, we believe follow-up TCB orders may come through shortly (within 1Q26) as SK Hynix continues to ramp up capacity at M15X. Following orders for HBM TCB tools, we expect visibility may also soon extend to other areas of HBM back-end including dicing/grinding, descum, test etc.\n\nHBM4 TCB tool specs largely unchanged vs HBM3E; any ASP uplift limited\n\nIn the case for HBM4 12Hi, SK Hynix will continue to adopt a flux-type TCB tool (based on advanced MR-MUF) – with TCB tool specs, we believe, largely unchanged in versus HBM3E 12Hi. This was enabled, we believe, due to SK Hynix in particular having been able to meaningfully improve/control TSV back-end yields for HBM3E allowing for extension of the same process/tools to HBM4. This suggests to us that 1/ we may see a more competitive set up for HBM4 TCB tools (lower entry barrier + time for competitors to catch up on offering) but also 2/ any potential ASP uplift on tool spec improvement for HBM4 may be challenging. This may also extend beyond MR-MUF type TCB tools to potentially NCF type TCB tools, as well where we believe ASMPT may be set to gain share at Micron at the expense of Hanmi.\n\nUpdate on fluxless TCB and hybrid bonding adoption timing\n\nIndustry preparation for fluxless and hybrid bonding continues to progress though adoption timing remains uncertain. For SK Hynix, we believe advanced MR-MUF will remain the mainstream packaging solution till HBM4E (base scenario flux TCB while keeping fluxless ‘optionality’). Hybrid bonding adoption for SK Hynix, we believe, may only occur afterwards from HBM5 with Hanmi, Hanwha and BESI all engaged. For Hanwha, we believe the company is set to ship their 2nd gen hybrid bonder tools for eval to SK Hynix within 1H26. In the case for Micron, we believe the company is actively weighing the option of adopting fluxless TCB starting HBM4E, with Hanmi, ASMPT and BESI engaged. For Samsung, we believe both fluxless TCB and hybrid bonding tracks are in consideration for HBM4E with SEMES and BESI engaged for hybrid bonding.\n\nStock implications\n\nIntensifying competition for HBM4 TCB tools into 2026 bodes negatively for TCB tool ASP/margins. Within UBS coverage, we see Hanmi (Sell) as most exposed given market share loss off high base vs potential share gainers ASMPT (Buy) / BESI (Buy). Broadening visibility on HBM4 back-end spend should be sentiment positive for test tool vendors Teradyne (Buy), Advantest (Neutral) as well as dicing/grinding vendors Disco (Neutral).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.001972336816521092, "ret_signed_m1d": -0.001972336816521092, "alpha_spy_m1d": -0.0028109134289494575, "alpha_c_m1d": 0.007967998458380254, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.004514614515815252, "ret_signed_p1w": 0.004514614515815252, "alpha_spy_p1w": 0.008027929319705418, "alpha_c_p1w": 0.00526392968245426, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.3797570652924007, "ret_signed_p1m": 0.3797570652924007, "alpha_spy_p1m": 0.3940848948602209, "alpha_c_p1m": 0.36144994897057714, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.6003182298218506, "ret_signed_p3m": 0.6003182298218506, "alpha_spy_p3m": 0.5855832786972315, "alpha_c_p3m": 0.46892138318031606, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3722, -0.3919, -0.3648, -0.3681, -0.354, -0.3676, -0.2382, -0.2253, -0.2039, -0.1982, -0.2307, -0.1784, -0.1897, -0.2017, -0.1938, -0.2238, -0.2148, -0.2572, -0.2556, -0.2667, -0.2829, -0.2632, -0.3172, -0.304, -0.2725, -0.2651, -0.2138, -0.2138, -0.2028, -0.2127, -0.1675, -0.1449, -0.1294, -0.12, -0.1103, -0.1235, -0.1058, -0.106, -0.1524, -0.1466, -0.1572, -0.1882, -0.1652, -0.1445, -0.1356, -0.13, -0.1298, -0.1298, -0.1282, -0.135, -0.138, -0.1516, -0.1516, -0.0902, -0.0379, 0.003, -0.0249, -0.0519, -0.0477, -0.0166, 0.0049, 0.0089, -0.002, 0.0, 0.0, -0.0183, 0.0164, 0.0043, 0.0045, 0.0158, 0.0473, 0.0979, 0.104, 0.0565, 0.0937, 0.2403, 0.1794, 0.1884, 0.3154, 0.3588, 0.3364, 0.4089, 0.3633, 0.3798, 0.3798, 0.3397, 0.3805, 0.3852, 0.4244, 0.3966, 0.443, 0.5032, 0.4589, 0.4033, 0.4287, 0.334, 0.3383, 0.3399, 0.1973, 0.2999, 0.3189, 0.3242, 0.2568, 0.2559, 0.3079, 0.3128, 0.3158, 0.326, 0.2753, 0.3327, 0.4038, 0.4179, 0.3038, 0.2962, 0.2118, 0.3, 0.369, 0.3576, 0.3576, 0.3831, 0.4053, 0.5711, 0.597, 0.6136, 0.623, 0.6027, 0.6003, 0.6045, 0.6679, 0.6453, 0.666, 0.689, 0.7583, 0.8332, 0.7627, 0.6668, 0.3432, 0.5061, 0.5146, 0.4796, 0.5659, 0.6771, 0.5527, 0.5776, 0.6077, 0.5718, 0.5934, 0.5634, 0.4816, 0.4078, 0.4098, 0.5099, 0.5504, 0.5723, 0.5723, 0.707, 0.6486, 0.6785, 0.6419, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012045084146221340", "ticker_idx": 4, "ts": "2026-01-16T06:11:44+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Advantest (Neutral)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Sell Hanmi on HBM4 TCB share loss; Buy ASMPT/BESI as share gainers; sentiment positive for Teradyne, Advantest, Disco on HBM4 back-end spend.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: regarding HBM4 TCB bonding equipment\n\nHanmi, Hanwha may have both secured HBM4 TCB tool orders from SK Hynix\n\nPre-market yesterday, Hanmi disclosed having received HBM thermo-compression bonding (TCB) tool orders from SK Hynix (link) for a contract amount of Won9.7bn (ex. VAT) and duration till the end of 1Q26E. Given the size of order, we believe this may have been for c.3 units of HBM4 TCB tools. Based on our channel checks and industry conversations, we believe competitor Hanwha Semitech may have also secured TCB tool orders ‘of equivalent scale’ or c.3 units for HBM4 12Hi, suggesting a 50/50 split. With ASMPT also seeing renewed engagement to supply HBM TCB tools to SK Hynix (link, link), we believe all three vendors will be engaged in supplying HBM4 TCB tools to SK Hynix in 2026. Given our expectation of 2026 HBM4 TCB unit TAM for SK Hynix at c. 86 tools, we believe follow-up TCB orders may come through shortly (within 1Q26) as SK Hynix continues to ramp up capacity at M15X. Following orders for HBM TCB tools, we expect visibility may also soon extend to other areas of HBM back-end including dicing/grinding, descum, test etc.\n\nHBM4 TCB tool specs largely unchanged vs HBM3E; any ASP uplift limited\n\nIn the case for HBM4 12Hi, SK Hynix will continue to adopt a flux-type TCB tool (based on advanced MR-MUF) – with TCB tool specs, we believe, largely unchanged in versus HBM3E 12Hi. This was enabled, we believe, due to SK Hynix in particular having been able to meaningfully improve/control TSV back-end yields for HBM3E allowing for extension of the same process/tools to HBM4. This suggests to us that 1/ we may see a more competitive set up for HBM4 TCB tools (lower entry barrier + time for competitors to catch up on offering) but also 2/ any potential ASP uplift on tool spec improvement for HBM4 may be challenging. This may also extend beyond MR-MUF type TCB tools to potentially NCF type TCB tools, as well where we believe ASMPT may be set to gain share at Micron at the expense of Hanmi.\n\nUpdate on fluxless TCB and hybrid bonding adoption timing\n\nIndustry preparation for fluxless and hybrid bonding continues to progress though adoption timing remains uncertain. For SK Hynix, we believe advanced MR-MUF will remain the mainstream packaging solution till HBM4E (base scenario flux TCB while keeping fluxless ‘optionality’). Hybrid bonding adoption for SK Hynix, we believe, may only occur afterwards from HBM5 with Hanmi, Hanwha and BESI all engaged. For Hanwha, we believe the company is set to ship their 2nd gen hybrid bonder tools for eval to SK Hynix within 1H26. In the case for Micron, we believe the company is actively weighing the option of adopting fluxless TCB starting HBM4E, with Hanmi, ASMPT and BESI engaged. For Samsung, we believe both fluxless TCB and hybrid bonding tracks are in consideration for HBM4E with SEMES and BESI engaged for hybrid bonding.\n\nStock implications\n\nIntensifying competition for HBM4 TCB tools into 2026 bodes negatively for TCB tool ASP/margins. Within UBS coverage, we see Hanmi (Sell) as most exposed given market share loss off high base vs potential share gainers ASMPT (Buy) / BESI (Buy). Broadening visibility on HBM4 back-end spend should be sentiment positive for test tool vendors Teradyne (Buy), Advantest (Neutral) as well as dicing/grinding vendors Disco (Neutral).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013596469707830017, "ret_signed_m1d": -0.013596469707830017, "alpha_spy_m1d": -0.014435046320258382, "alpha_c_m1d": -0.0036561344329286705, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028070125475521346, "ret_signed_p1d": -0.028070125475521346, "alpha_spy_p1d": -0.028070125475521346, "alpha_c_p1d": -0.028070125475521346, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.03201754852487593, "ret_signed_p1w": 0.03201754852487593, "alpha_spy_p1w": 0.035530863328766094, "alpha_c_p1w": 0.032766863691514936, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.18859654690981054, "ret_signed_p1m": 0.18859654690981054, "alpha_spy_p1m": 0.20292437647763073, "alpha_c_p1m": 0.17028943058798696, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.21151524155813495, "ret_signed_p3m": 0.21151524155813495, "alpha_spy_p3m": 0.19678029043351586, "alpha_c_p3m": 0.08011839491660044, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2368, -0.25, -0.2779, -0.2509, -0.202, -0.2053, -0.0298, -0.0232, 0.0147, 0.0147, -0.0447, -0.1015, -0.0732, -0.1246, -0.0912, -0.1283, -0.1303, -0.0934, -0.1434, -0.1228, -0.1553, -0.1601, -0.0862, -0.1967, -0.1967, -0.1632, -0.1465, -0.1048, -0.0976, -0.136, -0.1311, -0.0851, -0.0921, -0.114, -0.1118, -0.1116, -0.1162, -0.0772, -0.0884, -0.1469, -0.159, -0.1469, -0.1752, -0.1581, -0.1206, -0.1371, -0.1158, -0.1318, -0.1121, -0.1325, -0.1388, -0.1388, -0.1388, -0.1388, -0.0713, -0.0546, -0.0963, -0.1182, -0.1116, -0.1116, -0.0357, 0.0114, -0.0136, 0.0, -0.0281, -0.0577, -0.0452, 0.0022, 0.032, 0.034, 0.0945, 0.1202, 0.1781, 0.1186, 0.066, 0.1417, 0.1173, 0.0636, 0.0759, 0.1998, 0.2138, 0.2138, 0.1759, 0.1899, 0.1886, 0.1752, 0.1825, 0.1401, 0.1173, 0.1173, 0.1675, 0.255, 0.2336, 0.1776, 0.1316, 0.1283, 0.0746, 0.12, 0.1276, 0.0033, 0.0559, 0.0914, 0.0746, 0.0373, 0.0588, 0.0327, 0.1022, 0.0518, 0.0518, -0.0031, -0.02, 0.0274, 0.0072, -0.032, -0.081, -0.1071, -0.0118, -0.0722, -0.0533, -0.0368, -0.025, 0.1076, 0.0892, 0.0975, 0.0927, 0.1858, 0.2115, 0.2572, 0.224, 0.1902, 0.1946, 0.2253, 0.2253, 0.293, 0.3835, 0.3066, 0.3066, 0.2412, 0.2216, 0.2216, 0.2216, 0.2216, 0.3051, 0.3125, 0.264, 0.2607, 0.2425, 0.2568, 0.1577, 0.1485, 0.1107, 0.1254, 0.1746, 0.179, 0.219, 0.1452, 0.1915, 0.1568, 0.1494, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012045084146221340", "ticker_idx": 5, "ts": "2026-01-16T06:11:44+00:00", "symbol": "6146.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "dicing/grinding vendors Disco (Neutral)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS: Sell Hanmi on HBM4 TCB share loss; Buy ASMPT/BESI as share gainers; sentiment positive for Teradyne, Advantest, Disco on HBM4 back-end spend.", "resolved_tickers": ["6146.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: regarding HBM4 TCB bonding equipment\n\nHanmi, Hanwha may have both secured HBM4 TCB tool orders from SK Hynix\n\nPre-market yesterday, Hanmi disclosed having received HBM thermo-compression bonding (TCB) tool orders from SK Hynix (link) for a contract amount of Won9.7bn (ex. VAT) and duration till the end of 1Q26E. Given the size of order, we believe this may have been for c.3 units of HBM4 TCB tools. Based on our channel checks and industry conversations, we believe competitor Hanwha Semitech may have also secured TCB tool orders ‘of equivalent scale’ or c.3 units for HBM4 12Hi, suggesting a 50/50 split. With ASMPT also seeing renewed engagement to supply HBM TCB tools to SK Hynix (link, link), we believe all three vendors will be engaged in supplying HBM4 TCB tools to SK Hynix in 2026. Given our expectation of 2026 HBM4 TCB unit TAM for SK Hynix at c. 86 tools, we believe follow-up TCB orders may come through shortly (within 1Q26) as SK Hynix continues to ramp up capacity at M15X. Following orders for HBM TCB tools, we expect visibility may also soon extend to other areas of HBM back-end including dicing/grinding, descum, test etc.\n\nHBM4 TCB tool specs largely unchanged vs HBM3E; any ASP uplift limited\n\nIn the case for HBM4 12Hi, SK Hynix will continue to adopt a flux-type TCB tool (based on advanced MR-MUF) – with TCB tool specs, we believe, largely unchanged in versus HBM3E 12Hi. This was enabled, we believe, due to SK Hynix in particular having been able to meaningfully improve/control TSV back-end yields for HBM3E allowing for extension of the same process/tools to HBM4. This suggests to us that 1/ we may see a more competitive set up for HBM4 TCB tools (lower entry barrier + time for competitors to catch up on offering) but also 2/ any potential ASP uplift on tool spec improvement for HBM4 may be challenging. This may also extend beyond MR-MUF type TCB tools to potentially NCF type TCB tools, as well where we believe ASMPT may be set to gain share at Micron at the expense of Hanmi.\n\nUpdate on fluxless TCB and hybrid bonding adoption timing\n\nIndustry preparation for fluxless and hybrid bonding continues to progress though adoption timing remains uncertain. For SK Hynix, we believe advanced MR-MUF will remain the mainstream packaging solution till HBM4E (base scenario flux TCB while keeping fluxless ‘optionality’). Hybrid bonding adoption for SK Hynix, we believe, may only occur afterwards from HBM5 with Hanmi, Hanwha and BESI all engaged. For Hanwha, we believe the company is set to ship their 2nd gen hybrid bonder tools for eval to SK Hynix within 1H26. In the case for Micron, we believe the company is actively weighing the option of adopting fluxless TCB starting HBM4E, with Hanmi, ASMPT and BESI engaged. For Samsung, we believe both fluxless TCB and hybrid bonding tracks are in consideration for HBM4E with SEMES and BESI engaged for hybrid bonding.\n\nStock implications\n\nIntensifying competition for HBM4 TCB tools into 2026 bodes negatively for TCB tool ASP/margins. Within UBS coverage, we see Hanmi (Sell) as most exposed given market share loss off high base vs potential share gainers ASMPT (Buy) / BESI (Buy). Broadening visibility on HBM4 back-end spend should be sentiment positive for test tool vendors Teradyne (Buy), Advantest (Neutral) as well as dicing/grinding vendors Disco (Neutral).", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.024752424863004996, "ret_signed_m1d": -0.024752424863004996, "alpha_spy_m1d": -0.02559100147543336, "alpha_c_m1d": -0.01481208958810365, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03003301650086021, "ret_signed_p1d": 0.03003301650086021, "alpha_spy_p1d": 0.03003301650086021, "alpha_c_p1d": 0.03003301650086021, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.12128719947584332, "ret_signed_p1w": 0.12128719947584332, "alpha_spy_p1w": 0.12480051427973349, "alpha_c_p1w": 0.12203651464248233, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.19240929699960252, "ret_signed_p1m": 0.19240929699960252, "alpha_spy_p1m": 0.2067371265674227, "alpha_c_p1m": 0.17410218067777894, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.17130624857909638, "ret_signed_p3m": 0.17130624857909638, "alpha_spy_p3m": 0.1565712974544773, "alpha_c_p3m": 0.03990940193756187, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.133, -0.1271, -0.1602, -0.118, -0.1048, -0.1109, -0.0695, -0.1459, -0.1483, -0.1483, -0.1441, -0.2155, -0.1663, -0.1993, -0.1746, -0.212, -0.2102, -0.2008, -0.2196, -0.2078, -0.2442, -0.2701, -0.22, -0.2751, -0.2751, -0.2944, -0.2741, -0.2779, -0.2784, -0.2995, -0.2985, -0.2667, -0.2498, -0.2393, -0.2168, -0.1804, -0.1987, -0.186, -0.1946, -0.2129, -0.2356, -0.2437, -0.271, -0.262, -0.2206, -0.2269, -0.2196, -0.2125, -0.1936, -0.212, -0.2051, -0.2051, -0.2051, -0.2051, -0.1568, -0.1056, -0.0814, -0.0827, -0.0924, -0.0924, -0.0523, -0.0099, -0.0248, 0.0, 0.03, -0.0099, -0.0335, 0.1315, 0.1213, 0.0974, 0.1375, 0.163, 0.1112, 0.0922, 0.0281, 0.1043, 0.1244, 0.0754, 0.0858, 0.1919, 0.2437, 0.2437, 0.2018, 0.2056, 0.1924, 0.1835, 0.1853, 0.2223, 0.2221, 0.2221, 0.2479, 0.3201, 0.3048, 0.2459, 0.2361, 0.2051, 0.138, 0.1825, 0.213, 0.0974, 0.1517, 0.1942, 0.1756, 0.1487, 0.1538, 0.1114, 0.1663, 0.1482, 0.1482, 0.0776, 0.079, 0.1218, 0.1137, 0.0979, 0.0663, 0.0163, 0.0568, 0.055, 0.0895, 0.1116, 0.0432, 0.1194, 0.0721, 0.1101, 0.118, 0.1887, 0.1713, 0.212, 0.2098, 0.2106, 0.232, 0.2418, 0.1949, 0.1899, 0.2618, 0.2686, 0.2686, 0.2282, 0.199, 0.199, 0.199, 0.199, 0.271, 0.2447, 0.2178, 0.2081, 0.2088, 0.1617, 0.077, 0.0538, 0.0223, 0.018, 0.0895, 0.093, 0.158, 0.1104, 0.1223, 0.1011, 0.0802, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2038207631152787578", "ticker_idx": 0, "ts": "2026-03-29T10:52:21+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the Big 3 trade at valuations that remain absurdly cheap by comparison. Repeating the old playbook in a changed environment isn't sophistication — it's a failure to price in structural change.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish memory sector (Big 3): TSMC-like contract structures + disciplined capacity make current valuations absurdly cheap; market misprices structural change.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "2408.TW"], "resolution_method": "basket", "resolution_reason": "Global memory sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Dan wrote a good piece, but I see things differently on a few important points.\n\n1. Buyer resistance to smartphone memory price hikes is largely confined to legacy memory like DDR4 — not memory broadly.\n\nThis distinction matters. It's true that some buyers have recently pushed back on price increases, but the target of that resistance has primarily been legacy memory like DDR4.\n\nThe market has been behaving rather abnormally of late. At one point, legacy memory like DDR4 was actually trading above DDR5 in price — a genuinely strange phenomenon. That distortion has since moderated as DDR5 prices have risen, but the spike at the time was difficult to explain through normal demand alone.\n\nAccording to industry sources, a significant portion of the DDR4 price surge was attributable to Chinese stockpiling. That's what gave smartphone OEMs the room to respond — by downgrading specs on entry-level devices or trimming production volumes. In other words, there was flexibility on the legacy side.\n\nDDR5 is a different story. As I noted in an earlier post, smartphone and PC makers accepted substantial DDR5 price increases in Q1 of this year, and even into Q2.\n\nThe implication is clear: DDR5 is not a negotiating target right now — it's closer to an essential input that buyers need to secure even at a premium.\n\nWhat's more, flagship products built around DDR5 aren't easy to spec down. At most, a spec freeze is on the table.\n\nI've also heard some buyers say they simply can't keep absorbing additional costs on legacy memory. What I failed to account for was that most market participants weren't drawing a clear enough distinction — that this dynamic was specific to legacy — and I didn't anticipate that gap in understanding.\n\n2. Repeating a familiar playbook and adapting to structural change are two different things.\n\nDan portrayed investors who reflexively sell on spot price declines as sophisticated. I'd disagree.\n\nMemory companies are no longer operating the way they used to. If anything, they're moving toward a model closer to what TSMC does — building out capacity after securing advance paymets and long-term demand visibility from key customers. Just a few days ago, Korean media reported that Samsung is in discussions to pursue prepayment-based agreements with the likes of Microsoft.\n\nAnd yet the market still treats memory companies as textbook cyclicals. Even as contract structures and demand visibility that resemble TSMC's begin to emerge, the Big 3 trade at valuations that remain absurdly cheap by comparison.\n\nRepeating the old playbook in a changed environment isn't sophistication — it's a failure to price in structural change.\n\nI'm well aware of the counterargument. LTAs existed in prior cycles, and many were torn up when downturns hit.\n\nBut the memory companies know that better than anyone.\n\nThere's no reason to think they don't want TSMC-like valuations. And precisely because of that, the incentive to avoid repeating the old pattern — overcapacity that destroys their own cycle — is far stronger than it's ever been. They're working considerably harder than most people realize to pursue disciplined expansion rather than reflexive overbuilding.\n\nIt's disappointing to see so many participants still running the same old framework, watching spot prices tick on a screen rather than paying attention to how the industry itself is changing on the ground.", "reply_text": null, "parent_text": null, "quoted_text": "https://t.co/xs7CYRIwcR", "parent_author": null, "ret_m1d": 0.04830994234499161, "ret_signed_m1d": 0.04830994234499161, "alpha_spy_m1d": 0.044955263720016875, "alpha_c_m1d": 0.015981579836602172, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.03232836250838944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.044286641334337046, "ret_signed_p1d": -0.044286641334337046, "alpha_spy_p1d": -0.07335456703325186, "alpha_c_p1d": -0.10185427763329294, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.05756763629895589, "ret_p1w": 0.04834430912698079, "ret_signed_p1w": 0.04834430912698079, "alpha_spy_p1w": 0.005684019520718786, "alpha_c_p1w": -0.04392395093332696, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.09226826006030775, "ret_p1m": 0.34813102873393653, "ret_signed_p1m": 0.34813102873393653, "alpha_spy_p1m": 0.2219857442521671, "alpha_c_p1m": -0.006818887435952048, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.3549499161698886, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2038, -0.2038, -0.145, -0.1274, -0.0702, -0.0515, -0.0597, -0.0761, -0.0499, -0.071, -0.0624, -0.0516, -0.0081, 0.0231, 0.0097, 0.0098, 0.0422, 0.0527, 0.052, 0.087, 0.145, 0.1487, 0.1793, 0.1234, 0.1366, 0.1256, 0.0752, 0.0701, 0.098, 0.0821, 0.1171, 0.1432, 0.1428, 0.1428, 0.1336, 0.15, 0.1637, 0.188, 0.1995, 0.2328, 0.2386, 0.2696, 0.2541, 0.2538, 0.1306, 0.0639, 0.1409, 0.0827, 0.0253, 0.1095, 0.1555, 0.1185, 0.121, 0.1855, 0.2185, 0.2689, 0.2136, 0.1643, 0.0977, 0.1034, 0.1059, 0.0537, 0.0483, 0.0, -0.0443, 0.0508, 0.0025, 0.0267, 0.0483, 0.0679, 0.1631, 0.135, 0.1567, 0.1699, 0.2199, 0.2188, 0.2365, 0.221, 0.2211, 0.2596, 0.2857, 0.2792, 0.2807, 0.3527, 0.3481, 0.36, 0.327, 0.3465, 0.4606, 0.5322, 0.6733, 0.6864, 0.7544, 0.9024, 0.8806, 0.9569, 0.9734, 0.8206, 0.8007, 0.7444, 0.7716, 0.9045, 0.9061, 0.8897, 1.0517, 1.1527, 1.1651, 1.2655, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2038207631152787578", "ticker_idx": 1, "ts": "2026-03-29T10:52:21+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Korean media reported that Samsung is in discussions to pursue prepayment-based agreements with the likes of Microsoft", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish memory sector (Big 3): TSMC-like contract structures + disciplined capacity make current valuations absurdly cheap; market misprices structural change.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Dan wrote a good piece, but I see things differently on a few important points.\n\n1. Buyer resistance to smartphone memory price hikes is largely confined to legacy memory like DDR4 — not memory broadly.\n\nThis distinction matters. It's true that some buyers have recently pushed back on price increases, but the target of that resistance has primarily been legacy memory like DDR4.\n\nThe market has been behaving rather abnormally of late. At one point, legacy memory like DDR4 was actually trading above DDR5 in price — a genuinely strange phenomenon. That distortion has since moderated as DDR5 prices have risen, but the spike at the time was difficult to explain through normal demand alone.\n\nAccording to industry sources, a significant portion of the DDR4 price surge was attributable to Chinese stockpiling. That's what gave smartphone OEMs the room to respond — by downgrading specs on entry-level devices or trimming production volumes. In other words, there was flexibility on the legacy side.\n\nDDR5 is a different story. As I noted in an earlier post, smartphone and PC makers accepted substantial DDR5 price increases in Q1 of this year, and even into Q2.\n\nThe implication is clear: DDR5 is not a negotiating target right now — it's closer to an essential input that buyers need to secure even at a premium.\n\nWhat's more, flagship products built around DDR5 aren't easy to spec down. At most, a spec freeze is on the table.\n\nI've also heard some buyers say they simply can't keep absorbing additional costs on legacy memory. What I failed to account for was that most market participants weren't drawing a clear enough distinction — that this dynamic was specific to legacy — and I didn't anticipate that gap in understanding.\n\n2. Repeating a familiar playbook and adapting to structural change are two different things.\n\nDan portrayed investors who reflexively sell on spot price declines as sophisticated. I'd disagree.\n\nMemory companies are no longer operating the way they used to. If anything, they're moving toward a model closer to what TSMC does — building out capacity after securing advance paymets and long-term demand visibility from key customers. Just a few days ago, Korean media reported that Samsung is in discussions to pursue prepayment-based agreements with the likes of Microsoft.\n\nAnd yet the market still treats memory companies as textbook cyclicals. Even as contract structures and demand visibility that resemble TSMC's begin to emerge, the Big 3 trade at valuations that remain absurdly cheap by comparison.\n\nRepeating the old playbook in a changed environment isn't sophistication — it's a failure to price in structural change.\n\nI'm well aware of the counterargument. LTAs existed in prior cycles, and many were torn up when downturns hit.\n\nBut the memory companies know that better than anyone.\n\nThere's no reason to think they don't want TSMC-like valuations. And precisely because of that, the incentive to avoid repeating the old pattern — overcapacity that destroys their own cycle — is far stronger than it's ever been. They're working considerably harder than most people realize to pursue disciplined expansion rather than reflexive overbuilding.\n\nIt's disappointing to see so many participants still running the same old framework, watching spot prices tick on a screen rather than paying attention to how the industry itself is changing on the ground.", "reply_text": null, "parent_text": null, "quoted_text": "https://t.co/xs7CYRIwcR", "parent_author": null, "ret_m1d": 0.01944412932501427, "ret_signed_m1d": 0.01944412932501427, "alpha_spy_m1d": 0.01608945070003953, "alpha_c_m1d": 0.000463751529364842, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.018980377795649428, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.051616562677254674, "ret_signed_p1d": -0.051616562677254674, "alpha_spy_p1d": -0.08068448837616948, "alpha_c_p1d": -0.09396945598301443, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.04235289330575975, "ret_p1w": 0.0952921157118547, "ret_signed_p1w": 0.0952921157118547, "alpha_spy_p1w": 0.052631826105592694, "alpha_c_p1w": 0.02250781156050663, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.07278430415134807, "ret_p1m": 0.259217243335224, "ret_signed_p1m": 0.259217243335224, "alpha_spy_p1m": 0.13307195885345457, "alpha_c_p1m": 0.021177979778238765, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.23803926355698524, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3213, -0.3213, -0.2726, -0.2183, -0.2138, -0.2019, -0.2143, -0.2132, -0.2143, -0.2211, -0.2058, -0.1855, -0.1572, -0.1549, -0.1781, -0.1538, -0.1379, -0.139, -0.139, -0.0972, -0.0807, -0.0904, -0.0915, -0.1487, -0.0519, -0.0428, -0.0983, -0.1023, -0.0581, -0.0615, -0.0502, 0.011, 0.0257, 0.0257, 0.0257, 0.0257, 0.0755, 0.076, 0.0925, 0.1321, 0.1519, 0.234, 0.2255, 0.2255, 0.1044, -0.0253, 0.0845, 0.0653, -0.0179, 0.0636, 0.0755, 0.0636, 0.0387, 0.0681, 0.0976, 0.1802, 0.1349, 0.1287, 0.0545, 0.0738, 0.0698, 0.0194, 0.0194, 0.0, -0.0516, 0.0757, 0.0119, 0.0562, 0.0953, 0.1146, 0.194, 0.1571, 0.1685, 0.1401, 0.1713, 0.1968, 0.2337, 0.2252, 0.2167, 0.2422, 0.2337, 0.2734, 0.245, 0.2734, 0.2592, 0.2819, 0.2507, 0.2507, 0.3188, 0.3188, 0.5088, 0.54, 0.523, 0.6194, 0.5825, 0.6109, 0.679, 0.5343, 0.5939, 0.5627, 0.5655, 0.6988, 0.6591, 0.6591, 0.696, 0.7413, 0.6988, 0.7981, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011744483197796601", "ticker_idx": 0, "ts": "2026-01-15T10:17:15+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the strategy isn't just about buying TSMC; it's about recognizing that 2026 is an acceleration of the AI cycle, not a deceleration", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses hedge fund review: TSMC capex surge signals 2026 AI cycle acceleration; buy TSMC, ASML, Japan equipment, and Asian AI hardware.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Translation: Review of the TSMC Earnings Call (from a Hedge Fund Manager's Perspective)\n\nThe most critical takeaway from this TSMC earnings call is not just the generic \"AI is good\" sentiment, but the fact that hard data proves \"AI has redefined the system.\" With 4Q net income at $16 billion and a gross profit margin (GPM) of 62.3%, the company comfortably beat estimates. More importantly, the Capex guidance for 2026 has jumped to $52B–$56B (midpoint of $54B), representing a 31% YoY increase and sitting 17% above consensus. CFO Wendell Huang emphasized that the company will maintain \"meaningfully high spending for the next three years.\"\n\nThe interpretation of these figures is straightforward. The market knows TSMC is typically not a company to over-invest without clear demand visibility. The fact that they are betting over $50 billion now for supply 2–3 years down the line suggests the AI cycle is not about a \"peak debate\" but rather a stage of \"capacity and time constraints.\"\n\nCEO C.C. Wei’s comment—\"Our headache, if I may call it that, is the gap between supply and demand. We must work hard to bridge that gap\"—reveals that fundamental physical constraints (cleanrooms, tools, power, and packaging) are determining the sector’s ceiling, rather than traditional cyclical signals like inventory or unit price. Thus, the focus shifts from \"demand slowdown\" to the \"speed of supply catch-up.\"\n\nTSMC is accelerating fab expansion in Taiwan and Arizona. Tool-in and installation for Arizona Fab 2 will proceed this year, with mass production (HVM) pulled forward to 2H 2027 due to customer demand. Wei sought to reduce the \"US expansion = yield risk\" premium by stating that everything in Arizona, including yields, is \"very close to Taiwan.\"\n\nStrategy-wise, capital is naturally gravitating toward \"bottleneck locations\" rather than just the \"direction of demand.\" The core of 2026 lies not in total foundry volume, but in where the $52B–$56B Capex flows: 70–80% is allocated to advanced nodes (7nm and below), with the rest distributed among advanced packaging, testing, masks, and others. The moment the CFO mentioned that \"tools are becoming more expensive and complexity is increasing,\" the supply chain shifted from a simple leverage play to a long-term game of \"securing capacity and improving productivity.\"\n\nAn interesting tension arose when Wei dismissed two major market fears for the new year:\n\n1. Memory Chip Shortage: Wei stated that a \"memory chip crunch does not affect TSMC at all,\" noting that their revenue increasingly stems from high-end AI hardware and premium smartphones, which are less sensitive than low-end electronics. The platform mix supports this: 4Q revenue saw HPC (essentially AI) at 55%, Smartphones at 32%, and Others (Auto, etc.) at 13%. Despite an 11% growth in smartphones, HPC grew by 48% in 2025, dominating the landscape.\n\n2. Power Constraints in AI Data Centers: Wei noted they have audited the supply chain down to racks and cooling, stating, \"So far, so good.\" He emphasized that Cloud Service Providers (CSPs) are \"very smart\" and that their data center investments are operating on plans designed 5–6 years ago. In essence, power and memory constraints act as a \"brake\" for low-end and commodity products, while high-end products absorb the shock through pricing and mix.\n\nThe implications extend beyond TSMC. High-end smartphones have low demand elasticity despite rising memory costs. This context suggests robust demand for the iPhone 17 Pro/Pro Max, while low-end devices may suffer. Investment strategies must move past the simple \"Memory Shortage = Industry Downside\" frame and instead bifurcate the winners (Premium phones, AI accelerators, networking, advanced packaging) from the losers (price-sensitive consumer electronics).\n\nFinally, the core of this call was found in Wei’s attitude. While acknowledging the risk—\"Is AI demand real? I am also very nervous. Investing $52B–$56B could be a disaster if we aren't careful\"—he confirmed demand signals by meeting with \"customers' customers.\" He drew laughs by adding, \"I checked their financials. They are very rich.\"\n\nThis explains why data is projecting supply needs for 2028–2029. Since fabs take 2–3 years to build, today’s investment targets 2028. From the moment Wei mentioned they are \"looking at 2028 and 2029 supply,\" the debate shifted from short-term demand to \"how such long-term visibility is even possible.\" Notably, while 2nm ramp-up may dilute margins by 2–3% in 2H 2026, the 1Q GPM guidance of 63–65% (vs. 59.6% estimate) suggests margins are shifting toward a structural upper bound rather than being impaired.\n\nIn conclusion, the strategy isn't just about buying TSMC; it's about recognizing that 2026 is an acceleration of the AI cycle, not a deceleration. Capital should follow the path where profits migrate: utilization, efficiency, node transition, and packaging. The strengthening TWD, foreign inflows, and the positive reaction of ASML and Japanese equipment makers indicate that this guidance is a signal resetting the risk appetite for the entire Asian AI hardware asset class. 2026 tech is not about \"whether demand exists,\" but about \"how slowly supply can catch up.\"\n\n---\n\nI happened to come across a review of today’s TSMC call by a Korean hedge fund manager, and there were a few points that really resonated with me, so I’m sharing them here.\n\nSource: https://t.co/BotS7dza5W", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01183437870055326, "ret_signed_m1d": 0.01183437870055326, "alpha_spy_m1d": 0.01455019734066354, "alpha_c_m1d": 0.03216685639822481, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.020332477697671547, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029585801818144075, "ret_signed_p1d": 0.029585801818144075, "alpha_spy_p1d": 0.030423675809039485, "alpha_c_p1d": 0.019545664209203784, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.010040137608940292, "ret_p1w": 0.04142010805207774, "ret_signed_p1w": 0.04142010805207774, "alpha_spy_p1w": 0.04612942492440386, "alpha_c_p1w": 0.025249972074447014, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": 0.01617013597763073, "ret_p1m": 0.133136108181648, "ret_signed_p1m": 0.133136108181648, "alpha_spy_p1m": 0.14828980682462278, "alpha_c_p1m": 0.10460504828378991, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": 0.02853105989785809, "ret_p3m": 0.21994365430465845, "ret_signed_p3m": 0.21994365430465845, "alpha_spy_p3m": 0.2139968443173632, "alpha_c_p3m": 0.07971006830950222, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.14023358599515623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1272, -0.1272, -0.139, -0.1449, -0.1449, -0.1272, -0.1301, -0.1124, -0.1124, -0.1154, -0.1095, -0.1124, -0.139, -0.136, -0.139, -0.1301, -0.136, -0.1301, -0.139, -0.1567, -0.1478, -0.1714, -0.1773, -0.1419, -0.1832, -0.1891, -0.1655, -0.1508, -0.1537, -0.1508, -0.1685, -0.1567, -0.1449, -0.1478, -0.139, -0.1183, -0.1272, -0.1124, -0.1302, -0.1243, -0.142, -0.1509, -0.1538, -0.1538, -0.1538, -0.1331, -0.1183, -0.1154, -0.1154, -0.1065, -0.0947, -0.1006, -0.0828, -0.0828, -0.0621, -0.0118, 0.0089, -0.0089, -0.003, -0.0059, 0.0, 0.0118, 0.0118, 0.0, 0.0296, 0.0414, 0.0503, 0.0296, 0.0414, 0.0473, 0.0385, 0.0533, 0.0769, 0.068, 0.0503, 0.0444, 0.0651, 0.0562, 0.0444, 0.0533, 0.074, 0.1124, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1331, 0.1243, 0.1627, 0.1923, 0.1805, 0.1805, 0.1686, 0.145, 0.1036, 0.1243, 0.1183, 0.071, 0.0947, 0.1479, 0.1154, 0.1036, 0.0917, 0.1101, 0.1309, 0.0982, 0.0923, 0.0745, 0.0745, 0.0953, 0.0923, 0.0804, 0.0567, 0.0448, 0.1012, 0.0745, 0.0745, 0.0745, 0.1042, 0.1576, 0.1606, 0.1873, 0.1814, 0.2199, 0.2348, 0.2378, 0.2051, 0.2021, 0.217, 0.217, 0.2348, 0.2971, 0.3446, 0.3149, 0.2941, 0.2674, 0.2674, 0.3505, 0.3357, 0.3357, 0.3713, 0.3595, 0.3268, 0.3387, 0.3179, 0.3476, 0.3446, 0.3298, 0.309, 0.2971, 0.3238, 0.3387, 0.3713, 0.3476, 0.3654, 0.3624, 0.398, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011744483197796601", "ticker_idx": 1, "ts": "2026-01-15T10:17:15+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the positive reaction of ASML and Japanese equipment makers indicate that this guidance is a signal resetting the risk appetite for the entire Asian AI hardware asset class", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses hedge fund review: TSMC capex surge signals 2026 AI cycle acceleration; buy TSMC, ASML, Japan equipment, and Asian AI hardware.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Translation: Review of the TSMC Earnings Call (from a Hedge Fund Manager's Perspective)\n\nThe most critical takeaway from this TSMC earnings call is not just the generic \"AI is good\" sentiment, but the fact that hard data proves \"AI has redefined the system.\" With 4Q net income at $16 billion and a gross profit margin (GPM) of 62.3%, the company comfortably beat estimates. More importantly, the Capex guidance for 2026 has jumped to $52B–$56B (midpoint of $54B), representing a 31% YoY increase and sitting 17% above consensus. CFO Wendell Huang emphasized that the company will maintain \"meaningfully high spending for the next three years.\"\n\nThe interpretation of these figures is straightforward. The market knows TSMC is typically not a company to over-invest without clear demand visibility. The fact that they are betting over $50 billion now for supply 2–3 years down the line suggests the AI cycle is not about a \"peak debate\" but rather a stage of \"capacity and time constraints.\"\n\nCEO C.C. Wei’s comment—\"Our headache, if I may call it that, is the gap between supply and demand. We must work hard to bridge that gap\"—reveals that fundamental physical constraints (cleanrooms, tools, power, and packaging) are determining the sector’s ceiling, rather than traditional cyclical signals like inventory or unit price. Thus, the focus shifts from \"demand slowdown\" to the \"speed of supply catch-up.\"\n\nTSMC is accelerating fab expansion in Taiwan and Arizona. Tool-in and installation for Arizona Fab 2 will proceed this year, with mass production (HVM) pulled forward to 2H 2027 due to customer demand. Wei sought to reduce the \"US expansion = yield risk\" premium by stating that everything in Arizona, including yields, is \"very close to Taiwan.\"\n\nStrategy-wise, capital is naturally gravitating toward \"bottleneck locations\" rather than just the \"direction of demand.\" The core of 2026 lies not in total foundry volume, but in where the $52B–$56B Capex flows: 70–80% is allocated to advanced nodes (7nm and below), with the rest distributed among advanced packaging, testing, masks, and others. The moment the CFO mentioned that \"tools are becoming more expensive and complexity is increasing,\" the supply chain shifted from a simple leverage play to a long-term game of \"securing capacity and improving productivity.\"\n\nAn interesting tension arose when Wei dismissed two major market fears for the new year:\n\n1. Memory Chip Shortage: Wei stated that a \"memory chip crunch does not affect TSMC at all,\" noting that their revenue increasingly stems from high-end AI hardware and premium smartphones, which are less sensitive than low-end electronics. The platform mix supports this: 4Q revenue saw HPC (essentially AI) at 55%, Smartphones at 32%, and Others (Auto, etc.) at 13%. Despite an 11% growth in smartphones, HPC grew by 48% in 2025, dominating the landscape.\n\n2. Power Constraints in AI Data Centers: Wei noted they have audited the supply chain down to racks and cooling, stating, \"So far, so good.\" He emphasized that Cloud Service Providers (CSPs) are \"very smart\" and that their data center investments are operating on plans designed 5–6 years ago. In essence, power and memory constraints act as a \"brake\" for low-end and commodity products, while high-end products absorb the shock through pricing and mix.\n\nThe implications extend beyond TSMC. High-end smartphones have low demand elasticity despite rising memory costs. This context suggests robust demand for the iPhone 17 Pro/Pro Max, while low-end devices may suffer. Investment strategies must move past the simple \"Memory Shortage = Industry Downside\" frame and instead bifurcate the winners (Premium phones, AI accelerators, networking, advanced packaging) from the losers (price-sensitive consumer electronics).\n\nFinally, the core of this call was found in Wei’s attitude. While acknowledging the risk—\"Is AI demand real? I am also very nervous. Investing $52B–$56B could be a disaster if we aren't careful\"—he confirmed demand signals by meeting with \"customers' customers.\" He drew laughs by adding, \"I checked their financials. They are very rich.\"\n\nThis explains why data is projecting supply needs for 2028–2029. Since fabs take 2–3 years to build, today’s investment targets 2028. From the moment Wei mentioned they are \"looking at 2028 and 2029 supply,\" the debate shifted from short-term demand to \"how such long-term visibility is even possible.\" Notably, while 2nm ramp-up may dilute margins by 2–3% in 2H 2026, the 1Q GPM guidance of 63–65% (vs. 59.6% estimate) suggests margins are shifting toward a structural upper bound rather than being impaired.\n\nIn conclusion, the strategy isn't just about buying TSMC; it's about recognizing that 2026 is an acceleration of the AI cycle, not a deceleration. Capital should follow the path where profits migrate: utilization, efficiency, node transition, and packaging. The strengthening TWD, foreign inflows, and the positive reaction of ASML and Japanese equipment makers indicate that this guidance is a signal resetting the risk appetite for the entire Asian AI hardware asset class. 2026 tech is not about \"whether demand exists,\" but about \"how slowly supply can catch up.\"\n\n---\n\nI happened to come across a review of today’s TSMC call by a Korean hedge fund manager, and there were a few points that really resonated with me, so I’m sharing them here.\n\nSource: https://t.co/BotS7dza5W", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05672527814758488, "ret_signed_m1d": -0.05672527814758488, "alpha_spy_m1d": -0.0540094595074746, "alpha_c_m1d": -0.03639280044991333, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.020332477697671547, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015486352865365927, "ret_signed_p1d": 0.015486352865365927, "alpha_spy_p1d": 0.016324226856261337, "alpha_c_p1d": 0.005446215256425635, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.010040137608940292, "ret_p1w": 0.02349059991245417, "ret_signed_p1w": 0.02349059991245417, "alpha_spy_p1w": 0.028199916784780288, "alpha_c_p1w": 0.00732046393482344, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": 0.01617013597763073, "ret_p1m": 0.037060729639875944, "ret_signed_p1m": 0.037060729639875944, "alpha_spy_p1m": 0.05221442828285072, "alpha_c_p1m": 0.008529669742017854, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": 0.02853105989785809, "ret_p3m": 0.11877801686306211, "ret_signed_p3m": 0.11877801686306211, "alpha_spy_p3m": 0.11283120687576687, "alpha_c_p3m": -0.02145556913209412, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.14023358599515623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2192, -0.2272, -0.2443, -0.229, -0.2232, -0.2113, -0.2105, -0.2001, -0.1844, -0.2012, -0.1939, -0.2013, -0.2112, -0.2211, -0.2424, -0.2286, -0.2284, -0.2241, -0.2331, -0.2395, -0.2375, -0.2469, -0.2286, -0.2259, -0.2744, -0.2528, -0.2515, -0.2089, -0.2194, -0.214, -0.1938, -0.183, -0.1618, -0.1671, -0.1721, -0.162, -0.171, -0.177, -0.1814, -0.195, -0.19, -0.2095, -0.2396, -0.2233, -0.2156, -0.2205, -0.2142, -0.2179, -0.2179, -0.2179, -0.2111, -0.201, -0.1984, -0.1984, -0.1419, -0.0837, -0.0769, -0.0849, -0.1187, -0.059, -0.0548, -0.0407, -0.0567, 0.0, 0.0155, -0.0252, -0.0082, 0.0047, 0.0235, 0.0251, 0.0249, 0.0593, 0.0392, 0.0371, 0.0576, 0.0656, 0.0357, -0.0077, 0.0002, 0.0386, 0.0496, 0.0393, 0.0522, 0.0278, 0.0371, 0.0411, 0.0447, 0.0845, 0.0787, 0.0939, 0.0883, 0.1007, 0.1224, 0.0737, 0.0745, 0.0545, 0.0121, 0.0452, 0.0332, -0.0007, -0.0002, 0.0454, 0.0444, 0.0283, 0.0277, 0.0432, 0.04, 0.0376, 0.0181, -0.0171, 0.024, 0.0468, 0.0555, 0.0156, -0.0013, -0.0312, -0.025, 0.0346, 0.0114, 0.0114, 0.0114, -0.0297, 0.0564, 0.0757, 0.1064, 0.0975, 0.1188, 0.0716, 0.0651, 0.0838, 0.0846, 0.0777, 0.0888, 0.0662, 0.091, 0.0587, 0.0231, 0.0413, 0.0673, 0.0673, 0.0367, 0.0731, 0.1394, 0.135, 0.152, 0.1399, 0.1052, 0.1586, 0.1935, 0.1408, 0.104, 0.0905, 0.1639, 0.1745, 0.2302, 0.2491, 0.2135, 0.2018, 0.2145, 0.2091, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011744483197796601", "ticker_idx": 2, "ts": "2026-01-15T10:17:15+00:00", "symbol": "Japanese equipment makers", "kind": "sector", "geography": "Japan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the positive reaction of ASML and Japanese equipment makers indicate that this guidance is a signal resetting the risk appetite for the entire Asian AI hardware asset class", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses hedge fund review: TSMC capex surge signals 2026 AI cycle acceleration; buy TSMC, ASML, Japan equipment, and Asian AI hardware.", "resolved_tickers": ["6857.T", "6146.T", "4063.T", "7735.T"], "resolution_method": "basket", "resolution_reason": "Japan semi equip basket: Advantest, DISCO, Shin-Etsu, SCR", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Translation: Review of the TSMC Earnings Call (from a Hedge Fund Manager's Perspective)\n\nThe most critical takeaway from this TSMC earnings call is not just the generic \"AI is good\" sentiment, but the fact that hard data proves \"AI has redefined the system.\" With 4Q net income at $16 billion and a gross profit margin (GPM) of 62.3%, the company comfortably beat estimates. More importantly, the Capex guidance for 2026 has jumped to $52B–$56B (midpoint of $54B), representing a 31% YoY increase and sitting 17% above consensus. CFO Wendell Huang emphasized that the company will maintain \"meaningfully high spending for the next three years.\"\n\nThe interpretation of these figures is straightforward. The market knows TSMC is typically not a company to over-invest without clear demand visibility. The fact that they are betting over $50 billion now for supply 2–3 years down the line suggests the AI cycle is not about a \"peak debate\" but rather a stage of \"capacity and time constraints.\"\n\nCEO C.C. Wei’s comment—\"Our headache, if I may call it that, is the gap between supply and demand. We must work hard to bridge that gap\"—reveals that fundamental physical constraints (cleanrooms, tools, power, and packaging) are determining the sector’s ceiling, rather than traditional cyclical signals like inventory or unit price. Thus, the focus shifts from \"demand slowdown\" to the \"speed of supply catch-up.\"\n\nTSMC is accelerating fab expansion in Taiwan and Arizona. Tool-in and installation for Arizona Fab 2 will proceed this year, with mass production (HVM) pulled forward to 2H 2027 due to customer demand. Wei sought to reduce the \"US expansion = yield risk\" premium by stating that everything in Arizona, including yields, is \"very close to Taiwan.\"\n\nStrategy-wise, capital is naturally gravitating toward \"bottleneck locations\" rather than just the \"direction of demand.\" The core of 2026 lies not in total foundry volume, but in where the $52B–$56B Capex flows: 70–80% is allocated to advanced nodes (7nm and below), with the rest distributed among advanced packaging, testing, masks, and others. The moment the CFO mentioned that \"tools are becoming more expensive and complexity is increasing,\" the supply chain shifted from a simple leverage play to a long-term game of \"securing capacity and improving productivity.\"\n\nAn interesting tension arose when Wei dismissed two major market fears for the new year:\n\n1. Memory Chip Shortage: Wei stated that a \"memory chip crunch does not affect TSMC at all,\" noting that their revenue increasingly stems from high-end AI hardware and premium smartphones, which are less sensitive than low-end electronics. The platform mix supports this: 4Q revenue saw HPC (essentially AI) at 55%, Smartphones at 32%, and Others (Auto, etc.) at 13%. Despite an 11% growth in smartphones, HPC grew by 48% in 2025, dominating the landscape.\n\n2. Power Constraints in AI Data Centers: Wei noted they have audited the supply chain down to racks and cooling, stating, \"So far, so good.\" He emphasized that Cloud Service Providers (CSPs) are \"very smart\" and that their data center investments are operating on plans designed 5–6 years ago. In essence, power and memory constraints act as a \"brake\" for low-end and commodity products, while high-end products absorb the shock through pricing and mix.\n\nThe implications extend beyond TSMC. High-end smartphones have low demand elasticity despite rising memory costs. This context suggests robust demand for the iPhone 17 Pro/Pro Max, while low-end devices may suffer. Investment strategies must move past the simple \"Memory Shortage = Industry Downside\" frame and instead bifurcate the winners (Premium phones, AI accelerators, networking, advanced packaging) from the losers (price-sensitive consumer electronics).\n\nFinally, the core of this call was found in Wei’s attitude. While acknowledging the risk—\"Is AI demand real? I am also very nervous. Investing $52B–$56B could be a disaster if we aren't careful\"—he confirmed demand signals by meeting with \"customers' customers.\" He drew laughs by adding, \"I checked their financials. They are very rich.\"\n\nThis explains why data is projecting supply needs for 2028–2029. Since fabs take 2–3 years to build, today’s investment targets 2028. From the moment Wei mentioned they are \"looking at 2028 and 2029 supply,\" the debate shifted from short-term demand to \"how such long-term visibility is even possible.\" Notably, while 2nm ramp-up may dilute margins by 2–3% in 2H 2026, the 1Q GPM guidance of 63–65% (vs. 59.6% estimate) suggests margins are shifting toward a structural upper bound rather than being impaired.\n\nIn conclusion, the strategy isn't just about buying TSMC; it's about recognizing that 2026 is an acceleration of the AI cycle, not a deceleration. Capital should follow the path where profits migrate: utilization, efficiency, node transition, and packaging. The strengthening TWD, foreign inflows, and the positive reaction of ASML and Japanese equipment makers indicate that this guidance is a signal resetting the risk appetite for the entire Asian AI hardware asset class. 2026 tech is not about \"whether demand exists,\" but about \"how slowly supply can catch up.\"\n\n---\n\nI happened to come across a review of today’s TSMC call by a Korean hedge fund manager, and there were a few points that really resonated with me, so I’m sharing them here.\n\nSource: https://t.co/BotS7dza5W", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004256807213389863, "ret_signed_m1d": 0.004256807213389863, "alpha_spy_m1d": 0.006972625853500142, "alpha_c_m1d": 0.02458928491106141, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.020332477697671547, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026000761379967485, "ret_signed_p1d": 0.026000761379967485, "alpha_spy_p1d": 0.026838635370862896, "alpha_c_p1d": 0.015960623771027194, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.010040137608940292, "ret_p1w": 0.08572358852365128, "ret_signed_p1w": 0.08572358852365128, "alpha_spy_p1w": 0.0904329053959774, "alpha_c_p1w": 0.06955345254602055, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": 0.01617013597763073, "ret_p1m": 0.16223552243554173, "ret_signed_p1m": 0.16223552243554173, "alpha_spy_p1m": 0.1773892210785165, "alpha_c_p1m": 0.13370446253768364, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": 0.02853105989785809, "ret_p3m": 0.23855758717849745, "ret_signed_p3m": 0.23855758717849745, "alpha_spy_p3m": 0.2326107771912022, "alpha_c_p3m": 0.09832400118334123, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.14023358599515623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1599, -0.1623, -0.1719, -0.1944, -0.1691, -0.1557, -0.1631, -0.1071, -0.1193, -0.1012, -0.1012, -0.1427, -0.1813, -0.1663, -0.1897, -0.1752, -0.1974, -0.1982, -0.1741, -0.1879, -0.1797, -0.2085, -0.2235, -0.1805, -0.2279, -0.2279, -0.2253, -0.2077, -0.1905, -0.188, -0.2008, -0.1984, -0.1652, -0.1547, -0.1662, -0.159, -0.1492, -0.154, -0.1576, -0.1533, -0.1673, -0.19, -0.1851, -0.203, -0.1964, -0.1642, -0.166, -0.138, -0.1384, -0.1282, -0.139, -0.1368, -0.1368, -0.1368, -0.1368, -0.0929, -0.0621, -0.0668, -0.0911, -0.0814, -0.0814, -0.0306, 0.0043, 0.0, 0.026, 0.0302, 0.0032, -0.0014, 0.0857, 0.0871, 0.0661, 0.0962, 0.1004, 0.0963, 0.0814, 0.0496, 0.1087, 0.1126, 0.0727, 0.0752, 0.1588, 0.1811, 0.1811, 0.1765, 0.1622, 0.1641, 0.1553, 0.1609, 0.1766, 0.1762, 0.1762, 0.1997, 0.2649, 0.2623, 0.23, 0.2126, 0.1788, 0.1187, 0.1641, 0.181, 0.0814, 0.1206, 0.1457, 0.1395, 0.1212, 0.1369, 0.1189, 0.1724, 0.1373, 0.1373, 0.0764, 0.0773, 0.1186, 0.1085, 0.0808, 0.056, 0.0331, 0.093, 0.0613, 0.0816, 0.0904, 0.0834, 0.1558, 0.1444, 0.1762, 0.1803, 0.2386, 0.2376, 0.2746, 0.2425, 0.2301, 0.2386, 0.237, 0.2229, 0.2516, 0.3069, 0.2843, 0.2843, 0.2583, 0.2443, 0.2443, 0.2443, 0.2443, 0.3382, 0.3254, 0.3113, 0.3062, 0.2907, 0.3145, 0.2251, 0.1993, 0.1716, 0.1666, 0.2035, 0.2144, 0.2496, 0.2083, 0.2553, 0.2326, 0.2465, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011744483197796601", "ticker_idx": 3, "ts": "2026-01-15T10:17:15+00:00", "symbol": "AI hardware", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Capital should follow the path where profits migrate: utilization, efficiency, node transition, and packaging", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses hedge fund review: TSMC capex surge signals 2026 AI cycle acceleration; buy TSMC, ASML, Japan equipment, and Asian AI hardware.", "resolved_tickers": ["SMH", "SOXX"], "resolution_method": "etf", "resolution_reason": "AI hardware proxied by semiconductor ETFs SMH/SOXX", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "self_thread", "lang": "en", "tweet_text": "Translation: Review of the TSMC Earnings Call (from a Hedge Fund Manager's Perspective)\n\nThe most critical takeaway from this TSMC earnings call is not just the generic \"AI is good\" sentiment, but the fact that hard data proves \"AI has redefined the system.\" With 4Q net income at $16 billion and a gross profit margin (GPM) of 62.3%, the company comfortably beat estimates. More importantly, the Capex guidance for 2026 has jumped to $52B–$56B (midpoint of $54B), representing a 31% YoY increase and sitting 17% above consensus. CFO Wendell Huang emphasized that the company will maintain \"meaningfully high spending for the next three years.\"\n\nThe interpretation of these figures is straightforward. The market knows TSMC is typically not a company to over-invest without clear demand visibility. The fact that they are betting over $50 billion now for supply 2–3 years down the line suggests the AI cycle is not about a \"peak debate\" but rather a stage of \"capacity and time constraints.\"\n\nCEO C.C. Wei’s comment—\"Our headache, if I may call it that, is the gap between supply and demand. We must work hard to bridge that gap\"—reveals that fundamental physical constraints (cleanrooms, tools, power, and packaging) are determining the sector’s ceiling, rather than traditional cyclical signals like inventory or unit price. Thus, the focus shifts from \"demand slowdown\" to the \"speed of supply catch-up.\"\n\nTSMC is accelerating fab expansion in Taiwan and Arizona. Tool-in and installation for Arizona Fab 2 will proceed this year, with mass production (HVM) pulled forward to 2H 2027 due to customer demand. Wei sought to reduce the \"US expansion = yield risk\" premium by stating that everything in Arizona, including yields, is \"very close to Taiwan.\"\n\nStrategy-wise, capital is naturally gravitating toward \"bottleneck locations\" rather than just the \"direction of demand.\" The core of 2026 lies not in total foundry volume, but in where the $52B–$56B Capex flows: 70–80% is allocated to advanced nodes (7nm and below), with the rest distributed among advanced packaging, testing, masks, and others. The moment the CFO mentioned that \"tools are becoming more expensive and complexity is increasing,\" the supply chain shifted from a simple leverage play to a long-term game of \"securing capacity and improving productivity.\"\n\nAn interesting tension arose when Wei dismissed two major market fears for the new year:\n\n1. Memory Chip Shortage: Wei stated that a \"memory chip crunch does not affect TSMC at all,\" noting that their revenue increasingly stems from high-end AI hardware and premium smartphones, which are less sensitive than low-end electronics. The platform mix supports this: 4Q revenue saw HPC (essentially AI) at 55%, Smartphones at 32%, and Others (Auto, etc.) at 13%. Despite an 11% growth in smartphones, HPC grew by 48% in 2025, dominating the landscape.\n\n2. Power Constraints in AI Data Centers: Wei noted they have audited the supply chain down to racks and cooling, stating, \"So far, so good.\" He emphasized that Cloud Service Providers (CSPs) are \"very smart\" and that their data center investments are operating on plans designed 5–6 years ago. In essence, power and memory constraints act as a \"brake\" for low-end and commodity products, while high-end products absorb the shock through pricing and mix.\n\nThe implications extend beyond TSMC. High-end smartphones have low demand elasticity despite rising memory costs. This context suggests robust demand for the iPhone 17 Pro/Pro Max, while low-end devices may suffer. Investment strategies must move past the simple \"Memory Shortage = Industry Downside\" frame and instead bifurcate the winners (Premium phones, AI accelerators, networking, advanced packaging) from the losers (price-sensitive consumer electronics).\n\nFinally, the core of this call was found in Wei’s attitude. While acknowledging the risk—\"Is AI demand real? I am also very nervous. Investing $52B–$56B could be a disaster if we aren't careful\"—he confirmed demand signals by meeting with \"customers' customers.\" He drew laughs by adding, \"I checked their financials. They are very rich.\"\n\nThis explains why data is projecting supply needs for 2028–2029. Since fabs take 2–3 years to build, today’s investment targets 2028. From the moment Wei mentioned they are \"looking at 2028 and 2029 supply,\" the debate shifted from short-term demand to \"how such long-term visibility is even possible.\" Notably, while 2nm ramp-up may dilute margins by 2–3% in 2H 2026, the 1Q GPM guidance of 63–65% (vs. 59.6% estimate) suggests margins are shifting toward a structural upper bound rather than being impaired.\n\nIn conclusion, the strategy isn't just about buying TSMC; it's about recognizing that 2026 is an acceleration of the AI cycle, not a deceleration. Capital should follow the path where profits migrate: utilization, efficiency, node transition, and packaging. The strengthening TWD, foreign inflows, and the positive reaction of ASML and Japanese equipment makers indicate that this guidance is a signal resetting the risk appetite for the entire Asian AI hardware asset class. 2026 tech is not about \"whether demand exists,\" but about \"how slowly supply can catch up.\"\n\n---\n\nI happened to come across a review of today’s TSMC call by a Korean hedge fund manager, and there were a few points that really resonated with me, so I’m sharing them here.\n\nSource: https://t.co/BotS7dza5W", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018054245436768757, "ret_signed_m1d": -0.018054245436768757, "alpha_spy_m1d": -0.015338426796658478, "alpha_c_m1d": -0.015338426796658478, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.002715818640110279, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.012804301321414302, "ret_signed_p1d": 0.012804301321414302, "alpha_spy_p1d": 0.013642175312309712, "alpha_c_p1d": 0.013642175312309712, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": -0.0008378739908954103, "ret_p1w": 0.025062134734106456, "ret_signed_p1w": 0.025062134734106456, "alpha_spy_p1w": 0.029771451606432575, "alpha_c_p1w": 0.029771451606432575, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.004709316872326119, "ret_p1m": 0.04012401815630984, "ret_signed_p1m": 0.04012401815630984, "alpha_spy_p1m": 0.055277716799284615, "alpha_c_p1m": 0.055277716799284615, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.015153698642974778, "ret_p3m": 0.1654065187035112, "ret_signed_p3m": 0.1654065187035112, "alpha_spy_p3m": 0.15945970871621595, "alpha_c_p3m": 0.15945970871621595, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.005946809987295243, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1296, -0.1345, -0.1536, -0.1355, -0.1204, -0.0972, -0.0927, -0.0781, -0.0905, -0.0897, -0.0831, -0.1178, -0.0957, -0.1181, -0.1264, -0.1017, -0.1216, -0.1093, -0.138, -0.1391, -0.1523, -0.1708, -0.1562, -0.1943, -0.1889, -0.155, -0.1526, -0.1314, -0.1314, -0.1177, -0.1166, -0.0997, -0.0838, -0.0919, -0.0837, -0.0735, -0.0738, -0.0613, -0.068, -0.1113, -0.115, -0.1183, -0.1505, -0.1304, -0.1075, -0.0958, -0.0898, -0.0867, -0.0867, -0.0848, -0.0879, -0.0896, -0.0992, -0.0992, -0.064, -0.0521, -0.0242, -0.0326, -0.0481, -0.0216, -0.0174, -0.012, -0.0181, 0.0, 0.0128, 0.0128, -0.0074, 0.0232, 0.0251, 0.0157, 0.0117, 0.0342, 0.06, 0.0621, 0.0224, 0.0375, 0.0141, -0.0283, -0.0288, 0.0233, 0.036, 0.0316, 0.0572, 0.0334, 0.0401, 0.0401, 0.039, 0.051, 0.045, 0.0564, 0.0498, 0.0655, 0.0832, 0.0487, 0.0349, 0.0346, -0.0104, 0.0098, -0.0005, -0.0403, -0.0038, 0.0035, 0.013, -0.0209, -0.0202, -0.0022, 0.0054, -0.0016, 0.0028, -0.0214, -0.007, 0.0037, 0.0155, -0.0317, -0.048, -0.0831, -0.0288, -0.0033, -0.0013, -0.0013, 0.01, 0.0203, 0.083, 0.1045, 0.1246, 0.1428, 0.1654, 0.1676, 0.1759, 0.2022, 0.2047, 0.2102, 0.242, 0.262, 0.3236, 0.3142, 0.2705, 0.2977, 0.3237, 0.334, 0.3247, 0.3755, 0.4454, 0.4117, 0.4865, 0.5173, 0.4735, 0.5058, 0.5158, 0.4562, 0.4246, 0.4231, 0.4842, 0.4948, 0.5241, 0.5241, 0.6053, 0.5879, 0.6015, 0.5997, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2032336572344123830", "ticker_idx": 0, "ts": "2026-03-13T06:02:52+00:00", "symbol": "Nittobo", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bullish Nittobo", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Nittobo as a beneficiary of quartz cloth / Low Dk glass substrate trend.", "resolved_tickers": ["3110.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Nittobo = Nitto Boseki, TSE 3110.T", "bench_c": "EWJ", "bench_c_reason": "country_fallback (suffix='T')", "bench_c_industry": "Textile Manufacturing", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@Arronwei3n Bullish Nittobo", "reply_text": null, "parent_text": "It seems Q Glass is really a nightmare for drills.\n\n6. From a manufacturability and commercialization perspective, the quartz cloth used in M10 may be replaced with Low Dk-2 Glass.", "quoted_text": null, "parent_author": "Aaronwei3n", "ret_m1d": -0.019873582186645256, "ret_signed_m1d": -0.019873582186645256, "alpha_spy_m1d": -0.02556591089143645, "alpha_c_m1d": -0.02983041789894758, "bench_spy_m1d": 0.005692328704791194, "bench_c_m1d": 0.009956835712302325, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011743456211967818, "ret_signed_p1d": -0.011743456211967818, "alpha_spy_p1d": -0.02192032232345098, "alpha_c_p1d": -0.03225688077953337, "bench_spy_p1d": 0.010176866111483163, "bench_c_p1d": 0.020513424567565552, "ret_p1w": -0.02529357754258088, "ret_signed_p1w": -0.02529357754258088, "alpha_spy_p1w": -0.007252021759533878, "alpha_c_p1w": 0.0006181744523099386, "bench_spy_p1w": -0.018041555783047003, "bench_c_p1w": -0.02591175199489082, "ret_p1m": 0.314121323522895, "ret_signed_p1m": 0.314121323522895, "alpha_spy_p1m": 0.27534115022563577, "alpha_c_p1m": 0.25426050505800313, "bench_spy_p1m": 0.038780173297259246, "bench_c_p1m": 0.059860818464891885, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5262, -0.5208, -0.5312, -0.5384, -0.5117, -0.5154, -0.5185, -0.523, -0.5393, -0.5321, -0.5393, -0.5393, -0.5393, -0.5393, -0.5271, -0.5235, -0.5172, -0.4634, -0.4336, -0.4336, -0.4309, -0.4101, -0.3984, -0.3509, -0.2986, -0.3089, -0.2733, -0.206, -0.229, -0.2904, -0.3026, -0.3089, -0.3306, -0.3302, -0.3528, -0.3135, -0.2597, -0.224, -0.2367, -0.1956, -0.1513, -0.1513, -0.0759, -0.0881, -0.0646, -0.0357, -0.0447, -0.0402, -0.0425, -0.0425, 0.1834, 0.2502, 0.1504, 0.1382, 0.135, 0.0537, -0.0181, 0.0479, 0.0754, -0.0659, -0.0438, 0.0108, -0.0199, 0.0, -0.0117, -0.0605, 0.0072, -0.0253, -0.0253, -0.0912, -0.0921, 0.0099, -0.0149, -0.0289, -0.1033, -0.1496, -0.053, -0.1102, -0.0707, -0.058, -0.0171, 0.1276, 0.1476, 0.2447, 0.3141, 0.3359, 0.2025, 0.227, 0.2483, 0.2125, 0.2007, 0.2461, 0.1594, 0.2424, 0.2411, 0.2801, 0.2801, 0.2951, 0.2778, 0.2778, 0.2778, 0.2778, 0.4498, 0.3795, 0.32, 0.3159, 0.1703, 0.1884, 0.0845, 0.0314, -0.0208, -0.043, -0.0008, 0.1485, 0.2506, 0.1748, 0.0573, 0.0305, 0.0287, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2024824598093562220", "ticker_idx": 0, "ts": "2026-02-20T12:32:58+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Memory prices are expected to continue rising throughout the year, driven by robust demand from AI customers... No customer is able to fully meet their memory demand this year, and Hynix's own DRAM and NAND inventories remain very lean at approximately 4 weeks—reinforcing supplier leverage. I really like how it dispels concerns about double ordering.", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish SK Hynix: memory prices rising all year on AI demand + lean inventories + supplier leverage; double-ordering concern dispelled.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SK Hynix disclosed the following at today's virtual group meeting with Goldman Sachs:\n\n1. Memory prices are expected to continue rising throughout the year, driven by robust demand from AI customers—underpinned by advancements in AI services—coupled with limited supply growth.\n\n2. A particularly noteworthy point: customers recognize that near-term production ramp-ups are difficult, and understand that double-ordering only drives price increases rather than higher allocations.\n\n3. While there is potential for despeccing among PC and mobile customers, constrained supply growth is expected to offset this.\n\n4. Server customer inventories have reached healthy levels, while PC and mobile customer inventories are on a declining trend.\n\n5. No customer is able to fully meet their memory demand this year, and Hynix's own DRAM and NAND inventories remain very lean at approximately 4 weeks—reinforcing supplier leverage.\n\n---\n\nI really like how it dispels concerns about double ordering.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.057955723625282896, "ret_signed_m1d": -0.057955723625282896, "alpha_spy_m1d": -0.050775783391687423, "alpha_c_m1d": -0.04631804300828701, "bench_spy_m1d": -0.007179940233595472, "bench_c_m1d": -0.011637680616995882, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0021075660479470404, "ret_signed_p1d": 0.0021075660479470404, "alpha_spy_p1d": 0.012318880758821704, "alpha_c_p1d": 0.007287899765529016, "bench_spy_p1d": -0.010211314710874664, "bench_c_p1d": -0.005180333717581975, "ret_p1w": 0.12008203517880589, "ret_signed_p1w": 0.12008203517880589, "alpha_spy_p1w": 0.1250717123668963, "alpha_c_p1w": 0.14094800556065912, "bench_spy_p1w": -0.004989677188090402, "bench_c_p1w": -0.020865970381853227, "ret_p1m": -0.015045700006315332, "ret_signed_p1m": -0.015045700006315332, "alpha_spy_p1m": 0.03174681586849781, "alpha_c_p1m": 0.04198635429663522, "bench_spy_p1m": -0.046792515874813145, "bench_c_p1m": -0.05703205430295055, "ret_p3m": 0.842170709220359, "ret_signed_p3m": 0.842170709220359, "alpha_spy_p3m": 0.7640709162092272, "alpha_c_p3m": 0.48164261533477926, "bench_spy_p3m": 0.07809979301113179, "bench_c_p3m": 0.3605280938855797, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4535, -0.4482, -0.4268, -0.4415, -0.4331, -0.412, -0.4183, -0.4289, -0.4268, -0.392, -0.4036, -0.3815, -0.4046, -0.3983, -0.4162, -0.4415, -0.4194, -0.4183, -0.4236, -0.3888, -0.3846, -0.3804, -0.3804, -0.3688, -0.3256, -0.314, -0.314, -0.314, -0.2866, -0.2666, -0.235, -0.2181, -0.2034, -0.216, -0.2107, -0.2223, -0.2181, -0.2107, -0.2034, -0.1949, -0.2171, -0.2202, -0.2044, -0.1918, -0.2244, -0.157, -0.1138, -0.0927, -0.0421, -0.1254, -0.0443, -0.0516, -0.1128, -0.1159, -0.0653, -0.0769, -0.0938, -0.0643, -0.0727, -0.0727, -0.0727, -0.0727, -0.058, 0.0, 0.0021, 0.059, 0.0727, 0.1602, 0.1201, 0.1201, -0.0087, -0.1037, -0.0066, -0.0245, -0.1174, -0.0098, 0.0082, -0.0182, -0.0393, 0.0282, 0.024, 0.1148, 0.0694, 0.0631, -0.015, 0.0409, 0.0504, -0.015, -0.015, -0.0784, -0.1481, -0.0531, -0.1238, -0.0752, -0.0647, -0.033, 0.0905, 0.0536, 0.0842, 0.0979, 0.1644, 0.1993, 0.2193, 0.1908, 0.2309, 0.2922, 0.2911, 0.2932, 0.29, 0.3639, 0.3724, 0.365, 0.3576, 0.3576, 0.5276, 0.5276, 0.6902, 0.7461, 0.7799, 0.9847, 0.9372, 1.086, 1.0797, 0.9203, 0.9425, 0.8422, 0.8422, 1.048, 1.0491, 1.0491, 1.1663, 1.3679, 1.4169, 1.4633, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2006129584690135328", "ticker_idx": 0, "ts": "2025-12-30T22:25:39+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "got back in yesterday at an average price of 187 dollars", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Re-entered NVDA at $187 avg after selling at $204; initiated new long INTC position at $36 avg.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I sold Nvidia at 204 dollars on November 4, and got back in yesterday at an average price of 187 dollars.\n\nOn December 30, I also took a position in Intel at an average price of 36 dollars.\n\n$NVDA $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0036258876760828507, "ret_signed_m1d": 0.0036258876760828507, "alpha_spy_m1d": 0.002403294031503078, "alpha_c_m1d": 0.0011486811512630979, "bench_spy_m1d": 0.0012225936445797725, "bench_c_m1d": 0.002477206524819753, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005545422619901985, "ret_signed_p1d": -0.005545422619901985, "alpha_spy_p1d": 0.0018635447534887106, "alpha_c_p1d": 0.003207412967201284, "bench_spy_p1d": -0.007408967373390696, "bench_c_p1d": -0.00875283558710327, "ret_p1w": -0.0015995988919732529, "ret_signed_p1w": -0.0015995988919732529, "alpha_spy_p1w": -0.008586306437667224, "alpha_c_p1w": -0.06851214663895089, "bench_spy_p1w": 0.006986707545693971, "bench_c_p1w": 0.06691254774697764, "ret_p1m": 0.02122220779309547, "ret_signed_p1m": 0.02122220779309547, "alpha_spy_p1m": 0.008980859754430082, "alpha_c_p1m": -0.125539531995305, "bench_spy_p1m": 0.01224134803866539, "bench_c_p1m": 0.14676173978840046, "ret_p3m": -0.106702172082634, "ret_signed_p3m": -0.106702172082634, "alpha_spy_p3m": -0.03219335647334509, "alpha_c_p3m": -0.1368142045478069, "bench_spy_p3m": -0.0745088156092889, "bench_c_p3m": 0.03011203246517291, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0071, 0.0004, -0.0107, -0.0134, 0.0083, 0.0268, -0.0234, 0.0041, -0.0401, -0.0412, -0.0306, -0.0231, -0.0262, -0.0341, -0.0388, -0.0287, -0.0069, 0.021, 0.0719, 0.1039, 0.0818, 0.0797, 0.1031, 0.0594, 0.0408, 0.0028, 0.0032, 0.0613, 0.0299, 0.0333, -0.0037, 0.014, -0.0051, -0.033, -0.0055, -0.0368, -0.0462, -0.0267, -0.0519, -0.0389, -0.0389, -0.0563, -0.0407, -0.0325, -0.0424, -0.0222, -0.0274, -0.0106, -0.0137, -0.02, -0.0352, -0.0668, -0.06, -0.0524, -0.0885, -0.0715, -0.0349, -0.0205, 0.0089, 0.0057, 0.0057, 0.0159, 0.0036, 0.0, -0.0055, -0.0055, 0.007, 0.0031, -0.0016, 0.0084, -0.0133, -0.0143, -0.0139, -0.0092, -0.0235, -0.0026, -0.007, -0.007, -0.0505, -0.0225, -0.0144, 0.0007, -0.0057, 0.0052, 0.0212, 0.0265, 0.0191, -0.0103, -0.0384, -0.0712, -0.0835, -0.0114, 0.0133, 0.0053, 0.0134, -0.0032, -0.0252, -0.0252, -0.0137, 0.0023, 0.0019, 0.0122, 0.0214, 0.0283, 0.0428, -0.0141, -0.0552, -0.027, -0.0399, -0.024, -0.0224, -0.0518, -0.0261, -0.0148, -0.008, -0.0234, -0.0388, -0.023, -0.0299, -0.038, -0.0478, -0.0791, -0.0634, -0.0657, -0.0472, -0.0869, -0.1067, -0.1192, -0.07, -0.0628, -0.0541, -0.0541, -0.0527, -0.0503, -0.0291, -0.0193, 0.0059, 0.0095, 0.0479, 0.0605, 0.0577, 0.0755, 0.0775, 0.0659, 0.0798, 0.0646, 0.1106, 0.1551, 0.1367, 0.1158, 0.0642, 0.0582, 0.0584, 0.0478, 0.1083, 0.1278, 0.1476, 0.1702, 0.1773, 0.2042, 0.2571, 0.2015, 0.1855, 0.1764, 0.1917, 0.1705, 0.1482, 0.1482, 0.1457, 0.1337, 0.1425, 0.1259, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2006129584690135328", "ticker_idx": 1, "ts": "2025-12-30T22:25:39+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I also took a position in Intel at an average price of 36 dollars", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Re-entered NVDA at $187 avg after selling at $204; initiated new long INTC position at $36 avg.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I sold Nvidia at 204 dollars on November 4, and got back in yesterday at an average price of 187 dollars.\n\nOn December 30, I also took a position in Intel at an average price of 36 dollars.\n\n$NVDA $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.016621955618399742, "ret_signed_m1d": -0.016621955618399742, "alpha_spy_m1d": -0.017844549262979514, "alpha_c_m1d": -0.019099162143219495, "bench_spy_m1d": 0.0012225936445797725, "bench_c_m1d": 0.002477206524819753, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01072379944673596, "ret_signed_p1d": -0.01072379944673596, "alpha_spy_p1d": -0.003314832073345264, "alpha_c_p1d": -0.0019709638596326906, "bench_spy_p1d": -0.007408967373390696, "bench_c_p1d": -0.00875283558710327, "ret_p1w": 0.07345849154185458, "ret_signed_p1w": 0.07345849154185458, "alpha_spy_p1w": 0.0664717839961606, "alpha_c_p1w": 0.0065459437948769406, "bench_spy_p1w": 0.006986707545693971, "bench_c_p1w": 0.06691254774697764, "ret_p1m": 0.30777479295039845, "ret_signed_p1m": 0.30777479295039845, "alpha_spy_p1m": 0.29553344491173306, "alpha_c_p1m": 0.161013053161998, "bench_spy_p1m": 0.01224134803866539, "bench_c_p1m": 0.14676173978840046, "ret_p3m": 0.15630032038344144, "ret_signed_p3m": 0.15630032038344144, "alpha_spy_p3m": 0.23080913599273034, "alpha_c_p3m": 0.12618828791826853, "bench_spy_p3m": -0.0745088156092889, "bench_c_p3m": 0.03011203246517291, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0, -0.0126, -0.019, -0.0035, 0.0035, 0.0134, -0.0249, -0.0021, -0.0448, -0.004, -0.0123, -0.0078, 0.0214, 0.022, -0.0102, 0.0231, 0.0263, 0.0601, 0.1134, 0.1083, 0.0767, 0.0721, 0.059, -0.0072, 0.029, -0.0016, 0.0223, 0.0308, 0.0155, 0.0158, -0.0373, -0.0477, -0.0694, -0.0796, -0.0587, -0.0987, -0.0751, -0.0405, -0.0394, -0.0131, -0.0131, 0.0874, 0.0727, 0.1654, 0.1732, 0.0858, 0.1102, 0.0804, 0.0858, 0.0933, 0.0592, 0.0137, 0.0056, 0.0003, -0.0335, -0.0273, -0.0129, -0.0249, -0.0255, -0.0306, -0.0306, -0.0295, -0.0166, 0.0, -0.0107, -0.0107, 0.0558, 0.0555, 0.0735, 0.1429, 0.1021, 0.2212, 0.1812, 0.2678, 0.3062, 0.2954, 0.259, 0.259, 0.3019, 0.4544, 0.4563, 0.2083, 0.1391, 0.1777, 0.3078, 0.3046, 0.2458, 0.3086, 0.3204, 0.3029, 0.2933, 0.3563, 0.3469, 0.2635, 0.2946, 0.2461, 0.2544, 0.2544, 0.2381, 0.2188, 0.1962, 0.1826, 0.1697, 0.2365, 0.2568, 0.2188, 0.2228, 0.2198, 0.1555, 0.222, 0.2319, 0.1641, 0.222, 0.2542, 0.2863, 0.2131, 0.2271, 0.2268, 0.1812, 0.2072, 0.2381, 0.1761, 0.1799, 0.1812, 0.2649, 0.1823, 0.1563, 0.1043, 0.1831, 0.2877, 0.3507, 0.3507, 0.3614, 0.4185, 0.5804, 0.6547, 0.6724, 0.7475, 0.7107, 0.741, 0.8365, 0.8365, 0.7614, 0.7764, 0.7499, 0.7903, 1.2129, 1.2786, 1.266, 1.5402, 1.533, 1.6708, 1.5678, 1.8995, 2.0298, 1.9389, 2.3491, 2.4702, 2.2335, 2.2249, 2.108, 1.9161, 1.9, 1.9705, 2.1893, 2.1769, 2.2129, 2.2129, 2.3115, 2.2646, 2.241, 2.0745, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014571006585655533", "ticker_idx": 0, "ts": "2026-01-23T05:28:51+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Raising Target Price to $308 ... Buy, AMD US ... we are raising our 2025/2026 EPS estimates by 1% and 2%, respectively", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF raises AMD PT to $308 (Buy) on CPU price tailwinds, strong server demand, and $17.8B AI revenue estimate; also carries Buy on NVDA.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics & Communications》 \nAMD (Buy, AMD US): Performance Outlook, Tailwinds in CPU and GPU\n\nRaising Target Price to $308: Since last November, AMD's stock has been consolidating, but it rebounded 15% in the past week, mainly driven by reports of CPU price increases. We agree with this, due to substrate shortages and tight wafer capacity. At the same time, we expect server CPU demand to remain strong, with AMD's share gain momentum continuing. For AI, we still maintain a higher-than-consensus revenue estimate of $17.8 billion, primarily driven by the MI455X and the resilient MI325X, while also anticipating potential upside as we have removed Meta's Iris from our ASIC expectations. Therefore, we believe the company is on track to achieve its guidance of data center revenue CAGR exceeding 60%. In the short term, we expect the company to provide optimistic guidance at its earnings call on February 3. Considering the short-term outlook and higher CPU prices, we are raising our 2025/2026 EPS estimates by 1% and 2%, respectively. We are shifting our valuation benchmark from 38x 2026E to 33x 2026E/2027E EPS to capture the long-term CAGR momentum, resulting in a higher target price of $308.\n\nAI Demand/Supply Tailwinds and Meta ASIC Progress: In the short term, we believe AI revenue will be supported by MI325X, based on an estimated 160,000–180,000 units for 2H26 supply chain buildout. If the Chinese government approves imports of NVIDIA's (Buy, NVDA US) H200 and AMD's MI325X, there could be upside potential in 2Q26. For 2026, we maintain our $17.8 billion estimate (compared to $6.3 billion in 2025), mainly driven by: 1) Solid customer pipeline for MI455X, primarily from OpenAI's first 1GW project, followed by Oracle, Meta, Microsoft, etc.; 2) 90,000 AI GPU CoWoS capacity in 2026, with MI430X/MI455X accounting for approximately 470,000 units; 3) Helios progressing as planned, with an expected ramp-up of about 4,000 racks in 2026; 4) We have removed Meta's Iris ASIC project from expectations due to design/ramp complexity, while retaining the high-end/testing project for Iris Plus—this implies upside potential in AMD's allocation at Meta.\n\nCPU Price Increase Tailwinds, Server Demand, and 2nm Process: Objectively speaking, PC demand may remain soft due to rising memory prices, yet the supply chain continues to manage inventory and profitability. On January 15, reports emerged that AMD will raise prices for its new PC CPUs to reflect substrate costs and tight wafer capacity—we expect this round of PC CPU price increases to be 10–15%. For servers, strong AI demand has already driven demand for general-purpose servers; meanwhile, TSMC's N3 capacity remains tight, so we do not rule out the possibility of server CPU price increases. For AMD, the Venice CPU ramp is progressing as planned, which will benefit from N2 and CoWoS to boost ASP. To alleviate wafer supply concerns, we maintain our view that server CPUs may adopt Intel's 14A process by 2028.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022912804667516107, "ret_signed_m1d": -0.022912804667516107, "alpha_spy_m1d": -0.022550155989549903, "alpha_c_m1d": -0.029736296910052862, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.032231959924759024, "ret_signed_p1d": -0.032231959924759024, "alpha_spy_p1d": -0.03731010698536885, "alpha_c_p1d": -0.029057701569648042, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": -0.08837799443754601, "ret_signed_p1w": -0.08837799443754601, "alpha_spy_p1w": -0.09235345509194048, "alpha_c_p1w": -0.09680108711177882, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": -0.24291431127319463, "ret_signed_p1m": -0.24291431127319463, "alpha_spy_p1m": -0.232990311212102, "alpha_c_p1m": -0.27488214016667256, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.16859211342460867, "ret_signed_p3m": 0.16859211342460867, "alpha_spy_p3m": 0.1338832428816621, "alpha_c_p3m": -0.023214707382691158, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0064, 0.0179, -0.0186, -0.0137, -0.0001, -0.0371, -0.0129, -0.0846, -0.1007, -0.0605, -0.0853, -0.003, -0.0451, -0.0496, -0.0738, -0.1132, -0.1391, -0.2066, -0.2153, -0.1719, -0.2062, -0.175, -0.175, -0.1623, -0.1537, -0.1711, -0.162, -0.1683, -0.1606, -0.1485, -0.1466, -0.1473, -0.1473, -0.1883, -0.2006, -0.1945, -0.2371, -0.2257, -0.1781, -0.1723, -0.1724, -0.1719, -0.1719, -0.1721, -0.1697, -0.1707, -0.1753, -0.1753, -0.1394, -0.1486, -0.1746, -0.1912, -0.2118, -0.2176, -0.2002, -0.1491, -0.1389, -0.1223, -0.1072, -0.1072, -0.1069, -0.038, -0.0229, 0.0, -0.0322, -0.0295, -0.0267, -0.0289, -0.0884, -0.0516, -0.0677, -0.2291, -0.2587, -0.1973, -0.1682, -0.1776, -0.1775, -0.2069, -0.2016, -0.2016, -0.218, -0.2294, -0.2168, -0.2292, -0.2429, -0.1765, -0.188, -0.2157, -0.229, -0.2351, -0.2647, -0.2218, -0.2319, -0.259, -0.2195, -0.2174, -0.2112, -0.2385, -0.2553, -0.243, -0.244, -0.2319, -0.2095, -0.2247, -0.2195, -0.2091, -0.1518, -0.2153, -0.2222, -0.2451, -0.2166, -0.1905, -0.1624, -0.1624, -0.1521, -0.1469, -0.1073, -0.0887, -0.0564, -0.0495, -0.0178, -0.006, 0.0715, 0.0721, 0.0588, 0.0955, 0.1686, 0.1758, 0.3394, 0.2886, 0.2446, 0.2982, 0.3651, 0.3884, 0.3152, 0.3681, 0.6227, 0.5729, 0.7529, 0.7668, 0.7263, 0.7156, 0.7317, 0.6332, 0.6212, 0.5945, 0.7236, 0.7313, 0.8003, 0.8003, 0.9404, 0.9083, 0.9951, 0.9874, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014571006585655533", "ticker_idx": 1, "ts": "2026-01-23T05:28:51+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy, NVDA US", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF raises AMD PT to $308 (Buy) on CPU price tailwinds, strong server demand, and $17.8B AI revenue estimate; also carries Buy on NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics & Communications》 \nAMD (Buy, AMD US): Performance Outlook, Tailwinds in CPU and GPU\n\nRaising Target Price to $308: Since last November, AMD's stock has been consolidating, but it rebounded 15% in the past week, mainly driven by reports of CPU price increases. We agree with this, due to substrate shortages and tight wafer capacity. At the same time, we expect server CPU demand to remain strong, with AMD's share gain momentum continuing. For AI, we still maintain a higher-than-consensus revenue estimate of $17.8 billion, primarily driven by the MI455X and the resilient MI325X, while also anticipating potential upside as we have removed Meta's Iris from our ASIC expectations. Therefore, we believe the company is on track to achieve its guidance of data center revenue CAGR exceeding 60%. In the short term, we expect the company to provide optimistic guidance at its earnings call on February 3. Considering the short-term outlook and higher CPU prices, we are raising our 2025/2026 EPS estimates by 1% and 2%, respectively. We are shifting our valuation benchmark from 38x 2026E to 33x 2026E/2027E EPS to capture the long-term CAGR momentum, resulting in a higher target price of $308.\n\nAI Demand/Supply Tailwinds and Meta ASIC Progress: In the short term, we believe AI revenue will be supported by MI325X, based on an estimated 160,000–180,000 units for 2H26 supply chain buildout. If the Chinese government approves imports of NVIDIA's (Buy, NVDA US) H200 and AMD's MI325X, there could be upside potential in 2Q26. For 2026, we maintain our $17.8 billion estimate (compared to $6.3 billion in 2025), mainly driven by: 1) Solid customer pipeline for MI455X, primarily from OpenAI's first 1GW project, followed by Oracle, Meta, Microsoft, etc.; 2) 90,000 AI GPU CoWoS capacity in 2026, with MI430X/MI455X accounting for approximately 470,000 units; 3) Helios progressing as planned, with an expected ramp-up of about 4,000 racks in 2026; 4) We have removed Meta's Iris ASIC project from expectations due to design/ramp complexity, while retaining the high-end/testing project for Iris Plus—this implies upside potential in AMD's allocation at Meta.\n\nCPU Price Increase Tailwinds, Server Demand, and 2nm Process: Objectively speaking, PC demand may remain soft due to rising memory prices, yet the supply chain continues to manage inventory and profitability. On January 15, reports emerged that AMD will raise prices for its new PC CPUs to reflect substrate costs and tight wafer capacity—we expect this round of PC CPU price increases to be 10–15%. For servers, strong AI demand has already driven demand for general-purpose servers; meanwhile, TSMC's N3 capacity remains tight, so we do not rule out the possibility of server CPU price increases. For AMD, the Venice CPU ramp is progressing as planned, which will benefit from N2 and CoWoS to boost ASP. To alleviate wafer supply concerns, we maintain our view that server CPUs may adopt Intel's 14A process by 2028.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01507967451249037, "ret_signed_m1d": -0.01507967451249037, "alpha_spy_m1d": -0.014717025834524167, "alpha_c_m1d": -0.021903166755027126, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006394126101000319, "ret_signed_p1d": -0.006394126101000319, "alpha_spy_p1d": -0.011472273161610147, "alpha_c_p1d": -0.0032198677458893377, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.01843667609195454, "ret_signed_p1w": 0.01843667609195454, "alpha_spy_p1w": 0.014461215437560071, "alpha_c_p1w": 0.010013583417721739, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": 0.020674677144930653, "ret_signed_p1m": 0.020674677144930653, "alpha_spy_p1m": 0.030598677206023273, "alpha_c_p1m": -0.011293151748547281, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.079080129853498, "ret_signed_p3m": 0.079080129853498, "alpha_spy_p3m": 0.044371259310551414, "alpha_c_p3m": -0.11272669095380183, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0711, 0.1032, 0.081, 0.0789, 0.1023, 0.0587, 0.0401, 0.0021, 0.0025, 0.0606, 0.0292, 0.0326, -0.0044, 0.0133, -0.0058, -0.0337, -0.0062, -0.0375, -0.0469, -0.0273, -0.0525, -0.0395, -0.0395, -0.0569, -0.0413, -0.0331, -0.0431, -0.0229, -0.028, -0.0113, -0.0144, -0.0207, -0.0359, -0.0674, -0.0606, -0.053, -0.0891, -0.0721, -0.0356, -0.0212, 0.0082, 0.005, 0.005, 0.0152, 0.0029, -0.0007, -0.0062, -0.0062, 0.0063, 0.0024, -0.0023, 0.0077, -0.014, -0.015, -0.0145, -0.0099, -0.0241, -0.0033, -0.0077, -0.0077, -0.0512, -0.0232, -0.0151, 0.0, -0.0064, 0.0045, 0.0205, 0.0258, 0.0184, -0.011, -0.0391, -0.0718, -0.0841, -0.012, 0.0126, 0.0046, 0.0127, -0.0039, -0.0259, -0.0259, -0.0144, 0.0017, 0.0012, 0.0115, 0.0207, 0.0276, 0.042, -0.0148, -0.0558, -0.0277, -0.0406, -0.0247, -0.0231, -0.0525, -0.0267, -0.0155, -0.0087, -0.0241, -0.0395, -0.0237, -0.0305, -0.0387, -0.0485, -0.0797, -0.0641, -0.0664, -0.0479, -0.0875, -0.1073, -0.1198, -0.0707, -0.0635, -0.0547, -0.0547, -0.0534, -0.0509, -0.0297, -0.02, 0.0052, 0.0088, 0.0472, 0.0597, 0.057, 0.0747, 0.0767, 0.0651, 0.0791, 0.0638, 0.1098, 0.1543, 0.1359, 0.115, 0.0635, 0.0575, 0.0577, 0.0471, 0.1075, 0.127, 0.1468, 0.1693, 0.1765, 0.2034, 0.2562, 0.2007, 0.1847, 0.1756, 0.1908, 0.1697, 0.1474, 0.1474, 0.1449, 0.1329, 0.1417, 0.1251, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025447897320620231", "ticker_idx": 0, "ts": "2026-02-22T05:49:44+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix's combined Q1 operating profit could exceed USD 20 billion, surpassing TSMC", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA's Simon Woo bullish on Samsung and SK Hynix: combined Q1 OP could top $20B on rising DRAM and NAND prices.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Bank of America's Simon Woo, in his latest memory report, argued that if improved optimism on not only DRAM but also NAND is fully reflected, Samsung Electronics and SK Hynix's combined Q1 operating profit (OP) could exceed USD 20 billion (over KRW 30 trillion), surpassing TSMC. He added that with memory chip prices expected to rise further in April, each Korean memory company's quarterly operating profit could easily exceed USD 20 billion or KRW 30 trillion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015025855706159885, "ret_signed_m1d": -0.015025855706159885, "alpha_spy_m1d": -0.025342517093223438, "alpha_c_m1d": -0.032063117250343054, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.01703726154418317, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03626946073682302, "ret_signed_p1d": 0.03626946073682302, "alpha_spy_p1d": 0.029001007806428003, "alpha_c_p1d": 0.023274982215576, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.012994478521247022, "ret_p1w": 0.12176170889254245, "ret_signed_p1w": 0.12176170889254245, "alpha_spy_p1w": 0.11591462317965084, "alpha_c_p1w": 0.11439822988666881, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": 0.007363479005873641, "ret_p1m": -0.01709838473024672, "ret_signed_p1m": -0.01709838473024672, "alpha_spy_p1m": 0.023092997594216014, "alpha_c_p1m": -0.0012109257509007598, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.01588745897934596, "ret_p3m": 0.5550254955061809, "ret_signed_p3m": 0.5550254955061809, "alpha_spy_p3m": 0.4636432786683964, "alpha_c_p3m": 0.264077910400387, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.2909475851057939, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4699, -0.4663, -0.4818, -0.4802, -0.4668, -0.4612, -0.4581, -0.4411, -0.4354, -0.4411, -0.4431, -0.4467, -0.4385, -0.4596, -0.4699, -0.4436, -0.4452, -0.4519, -0.4302, -0.4251, -0.4271, -0.4271, -0.3967, -0.3808, -0.3788, -0.3788, -0.3788, -0.3342, -0.2845, -0.2803, -0.2694, -0.2808, -0.2798, -0.2808, -0.287, -0.2731, -0.2544, -0.2285, -0.2264, -0.2477, -0.2254, -0.2109, -0.2119, -0.2119, -0.1736, -0.1585, -0.1674, -0.1684, -0.2207, -0.1321, -0.1238, -0.1746, -0.1782, -0.1378, -0.1409, -0.1306, -0.0746, -0.0611, -0.0611, -0.0611, -0.0611, -0.0155, -0.015, 0.0, 0.0363, 0.0544, 0.1295, 0.1218, 0.1218, 0.0109, -0.1078, -0.0073, -0.0249, -0.101, -0.0264, -0.0155, -0.0264, -0.0492, -0.0223, 0.0047, 0.0803, 0.0389, 0.0332, -0.0347, -0.0171, -0.0207, -0.0668, -0.0668, -0.0846, -0.1319, -0.0153, -0.0737, -0.0332, 0.0026, 0.0202, 0.0929, 0.0592, 0.0696, 0.0436, 0.0722, 0.0955, 0.1293, 0.1215, 0.1137, 0.1371, 0.1293, 0.1656, 0.1397, 0.1656, 0.1526, 0.1734, 0.1449, 0.1449, 0.2072, 0.2072, 0.3811, 0.4096, 0.3941, 0.4823, 0.4486, 0.4745, 0.5369, 0.4045, 0.459, 0.4304, 0.433, 0.555, 0.5187, 0.5187, 0.5524, 0.594, 0.555, 0.6459, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025447897320620231", "ticker_idx": 1, "ts": "2026-02-22T05:49:44+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "each Korean memory company's quarterly operating profit could easily exceed USD 20 billion or KRW 30 trillion", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA's Simon Woo bullish on Samsung and SK Hynix: combined Q1 OP could top $20B on rising DRAM and NAND prices.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bank of America's Simon Woo, in his latest memory report, argued that if improved optimism on not only DRAM but also NAND is fully reflected, Samsung Electronics and SK Hynix's combined Q1 operating profit (OP) could exceed USD 20 billion (over KRW 30 trillion), surpassing TSMC. He added that with memory chip prices expected to rise further in April, each Korean memory company's quarterly operating profit could easily exceed USD 20 billion or KRW 30 trillion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0021031335550720787, "ret_signed_m1d": -0.0021031335550720787, "alpha_spy_m1d": -0.012419794942135631, "alpha_c_m1d": -0.007310442872690004, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05678230114696747, "ret_signed_p1d": 0.05678230114696747, "alpha_spy_p1d": 0.049513848216572454, "alpha_c_p1d": 0.04157207370793392, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": 0.11772635306618784, "ret_signed_p1w": 0.11772635306618784, "alpha_spy_p1w": 0.11187926735329623, "alpha_c_p1w": 0.13344518216331758, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": 0.03871643856653173, "ret_signed_p1m": 0.03871643856653173, "alpha_spy_p1m": 0.07890782089099446, "alpha_c_p1m": 0.0830150442839801, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 1.0437220241815908, "ret_signed_p3m": 1.0437220241815908, "alpha_spy_p3m": 0.9523398073438063, "alpha_c_p3m": 0.6683102867108126, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4494, -0.428, -0.4427, -0.4343, -0.4132, -0.4196, -0.4301, -0.428, -0.3933, -0.4048, -0.3828, -0.4059, -0.3996, -0.4175, -0.4427, -0.4206, -0.4196, -0.4248, -0.3901, -0.3859, -0.3817, -0.3817, -0.3701, -0.327, -0.3155, -0.3155, -0.3155, -0.2881, -0.2681, -0.2366, -0.2198, -0.205, -0.2177, -0.2124, -0.224, -0.2198, -0.2124, -0.205, -0.1966, -0.2187, -0.2219, -0.2061, -0.1935, -0.2261, -0.1588, -0.1157, -0.0946, -0.0442, -0.1272, -0.0463, -0.0536, -0.1146, -0.1178, -0.0673, -0.0789, -0.0957, -0.0662, -0.0747, -0.0747, -0.0747, -0.0747, -0.0599, -0.0021, 0.0, 0.0568, 0.0705, 0.1578, 0.1177, 0.1177, -0.0108, -0.1056, -0.0087, -0.0266, -0.1193, -0.0118, 0.0061, -0.0203, -0.0413, 0.0261, 0.0219, 0.1125, 0.0672, 0.0608, -0.0171, 0.0387, 0.0482, -0.0171, -0.0171, -0.0803, -0.1499, -0.055, -0.1256, -0.0772, -0.0666, -0.035, 0.0882, 0.0514, 0.0819, 0.0956, 0.162, 0.1967, 0.2168, 0.1883, 0.2283, 0.2894, 0.2884, 0.2905, 0.2873, 0.3611, 0.3695, 0.3621, 0.3548, 0.3548, 0.5244, 0.5244, 0.6866, 0.7424, 0.7761, 0.9805, 0.9331, 1.0816, 1.0753, 0.9163, 0.9384, 0.8383, 0.8383, 1.0437, 1.0448, 1.0448, 1.1617, 1.3629, 1.4118, 1.4581, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010490942944592349", "ticker_idx": 0, "ts": "2026-01-11T23:16:08+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "significant upside potential for demand … the global NAND supply shortage is expected to intensify further", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Citi view that ICMS-driven NAND demand (2.8%→9.3% of global demand) will tighten supply and create significant upside; tags SNDK as beneficiary.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Citi: We estimate that approximately 1,152TB of additional SSD NAND will be required per Vera Rubin server system to support NVIDIA's ICMS operations.\n\nAccordingly, assuming Vera Rubin server shipments of 30,000 units in 2026 and 100,000 units in 2027, NAND demand driven by ICMS is projected to reach 34.6 million TB in 2026 and 115.2 million TB in 2027.\n\nThis represents 2.8% of expected global NAND demand in 2026 and 9.3% in 2027, a meaningful scale that is likely to create significant upside potential for demand.\n\nAs a result, with this structural demand increase factored in, the global NAND supply shortage is expected to intensify further.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030467248147264536, "ret_signed_m1d": -0.030467248147264536, "alpha_spy_m1d": -0.028899256793491035, "alpha_c_m1d": -0.026107273321901325, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.004359974825363211, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0013872339511455944, "ret_signed_p1d": 0.0013872339511455944, "alpha_spy_p1d": 0.0033867090570158798, "alpha_c_p1d": 0.0034991016358856264, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": -0.002111867684740032, "ret_p1w": 0.06255300123498708, "ret_signed_p1w": 0.06255300123498708, "alpha_spy_p1w": 0.0675878293955483, "alpha_c_p1w": 0.07052358183883911, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": -0.007970580603852029, "ret_p1m": 0.3914250518536997, "ret_signed_p1m": 0.3914250518536997, "alpha_spy_p1m": 0.39579817181577914, "alpha_c_p1m": 0.42030979400257584, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": -0.028884742148876152, "ret_p3m": 1.1876076537056393, "ret_signed_p3m": 1.1876076537056393, "alpha_spy_p3m": 1.2068739622903877, "alpha_c_p3m": 1.21855898915934, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": -0.030951335453700768, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6293, -0.6294, -0.6399, -0.6197, -0.6165, -0.6225, -0.5709, -0.5218, -0.5466, -0.5491, -0.475, -0.497, -0.4879, -0.4682, -0.5002, -0.4438, -0.4665, -0.3848, -0.3117, -0.3023, -0.2727, -0.3743, -0.3471, -0.317, -0.3708, -0.3682, -0.4966, -0.4855, -0.417, -0.4336, -0.4476, -0.4476, -0.4264, -0.4601, -0.4725, -0.5007, -0.452, -0.4131, -0.4208, -0.4362, -0.4018, -0.3793, -0.4703, -0.4814, -0.4623, -0.4687, -0.4362, -0.3896, -0.3808, -0.3709, -0.3576, -0.3576, -0.3576, -0.3725, -0.3829, -0.3902, -0.3902, -0.2929, -0.2959, -0.1018, -0.0917, -0.1406, -0.0305, 0.0, 0.0014, -0.0038, 0.0513, 0.0626, 0.0626, 0.164, 0.2878, 0.2933, 0.2172, 0.2094, 0.2368, 0.3554, 0.3854, 0.4803, 0.7089, 0.7867, 0.5017, 0.4802, 0.5361, 0.4987, 0.3914, 0.5397, 0.6192, 0.6096, 0.6096, 0.5172, 0.5424, 0.5955, 0.6697, 0.7122, 0.6403, 0.6245, 0.6747, 0.6322, 0.5904, 0.4525, 0.5389, 0.453, 0.3547, 0.5124, 0.5899, 0.6837, 0.5897, 0.6996, 0.8076, 0.8501, 0.9362, 0.9834, 0.8232, 0.8046, 0.8046, 0.7414, 0.5495, 0.582, 0.4707, 0.6321, 0.7796, 0.8023, 0.8023, 0.8615, 0.826, 1.0061, 1.1876, 1.1881, 1.4469, 1.4262, 1.2907, 1.362, 1.3659, 1.3455, 1.321, 1.5151, 1.3953, 1.543, 1.7492, 1.5749, 1.7339, 1.8168, 2.0493, 2.2262, 2.6127, 2.6221, 2.4422, 3.0135, 2.9755, 2.7301, 2.7178, 2.5521, 2.616, 2.4244, 2.5535, 2.5774, 2.9619, 2.7986, 2.7986, 3.0834, 3.0844, 3.2172, 3.3543, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010490942944592349", "ticker_idx": 1, "ts": "2026-01-11T23:16:08+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "significant upside potential for demand … the global NAND supply shortage is expected to intensify further", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Citi view that ICMS-driven NAND demand (2.8%→9.3% of global demand) will tighten supply and create significant upside; tags SNDK as beneficiary.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citi: We estimate that approximately 1,152TB of additional SSD NAND will be required per Vera Rubin server system to support NVIDIA's ICMS operations.\n\nAccordingly, assuming Vera Rubin server shipments of 30,000 units in 2026 and 100,000 units in 2027, NAND demand driven by ICMS is projected to reach 34.6 million TB in 2026 and 115.2 million TB in 2027.\n\nThis represents 2.8% of expected global NAND demand in 2026 and 9.3% in 2027, a meaningful scale that is likely to create significant upside potential for demand.\n\nAs a result, with this structural demand increase factored in, the global NAND supply shortage is expected to intensify further.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007591394852567568, "ret_signed_m1d": -0.007591394852567568, "alpha_spy_m1d": -0.006023403498794067, "alpha_c_m1d": -0.0039053189095290767, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006817072166661187, "ret_signed_p1d": 0.006817072166661187, "alpha_spy_p1d": 0.008816547272531472, "alpha_c_p1d": 0.004590084225658786, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.08112283949711552, "ret_signed_p1w": 0.08112283949711552, "alpha_spy_p1w": 0.08615766765767674, "alpha_c_p1w": 0.05621624323663976, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.2639398656668463, "ret_signed_p1m": 0.2639398656668463, "alpha_spy_p1m": 0.26831298562892575, "alpha_c_p1m": 0.2278470842934272, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.6794640817048946, "ret_signed_p3m": 0.6794640817048946, "alpha_spy_p3m": 0.6987303902896429, "alpha_c_p3m": 0.5779691922235067, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4663, -0.4395, -0.443, -0.4215, -0.4278, -0.4247, -0.4125, -0.3616, -0.3385, -0.352, -0.3121, -0.2973, -0.2943, -0.2626, -0.2976, -0.2865, -0.2738, -0.2545, -0.2, -0.1996, -0.1918, -0.2219, -0.2799, -0.2416, -0.288, -0.2839, -0.3077, -0.3442, -0.3185, -0.3208, -0.337, -0.3201, -0.3147, -0.3248, -0.3135, -0.3255, -0.321, -0.2962, -0.2704, -0.2791, -0.2671, -0.2703, -0.2902, -0.3158, -0.3322, -0.315, -0.292, -0.2785, -0.2434, -0.2418, -0.2223, -0.2194, -0.1993, -0.193, -0.1963, -0.2019, -0.2019, -0.1458, -0.115, -0.045, -0.0207, -0.0323, -0.0076, 0.0, 0.0068, 0.0016, 0.0271, 0.0712, 0.0811, 0.0911, 0.1556, 0.1919, 0.1717, 0.1576, 0.2188, 0.2799, 0.2939, 0.3466, 0.3227, 0.411, 0.3331, 0.2796, 0.289, 0.296, 0.2639, 0.3136, 0.3914, 0.4161, 0.4075, 0.3832, 0.3856, 0.4078, 0.4329, 0.4419, 0.4773, 0.4763, 0.5181, 0.495, 0.4916, 0.3607, 0.3181, 0.3679, 0.3182, 0.2854, 0.3812, 0.4451, 0.4067, 0.4262, 0.5059, 0.5165, 0.6058, 0.5659, 0.5183, 0.4511, 0.4593, 0.4615, 0.3581, 0.3515, 0.2839, 0.2799, 0.4261, 0.3844, 0.4199, 0.466, 0.4791, 0.6518, 0.6795, 0.7281, 0.7974, 0.8995, 0.8414, 0.8937, 0.8313, 0.831, 0.8838, 0.9747, 0.9672, 0.9848, 1.0821, 1.0475, 1.0913, 1.1169, 1.1598, 1.2754, 1.3895, 1.4963, 1.5812, 1.7746, 1.8944, 1.8098, 1.943, 1.8775, 1.7202, 1.788, 1.7639, 1.8126, 2.0569, 2.0404, 2.1673, 2.3, 2.3516, 2.3884, 2.5523, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010355785487045100", "ticker_idx": 0, "ts": "2026-01-11T14:19:04+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "J.P. Morgan forecasts 2026/27 revenue growth of +31%/+25%, which exceeds consensus (~+27%/+22%), and JPM's 2026/27 gross margin estimates of ~62% are also above consensus (~60%)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "JPM upgrades TSMC estimates above consensus on 2026/27 revenue growth, gross margins, and data-center AI CAGR; expects positive 1Q26 GM guidance surprise.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "J.P. Morgan: Upward Revision of TSMC Estimates\n\nOur analyst Gokul has recently upgraded his TSMC estimates. The key differences versus consensus are that J.P. Morgan (JPM) forecasts 2026/27 revenue growth of +31%/+25%, which exceeds consensus (~+27%/+22%), and JPM’s 2026/27 gross margin (GM) estimates of ~62% are also above consensus (~60%).\n\nJPM’s 2026/27 Capex estimates are US$48bn/US$55bn, which are higher than consensus (~US$46bn/US$52bn) despite no upward revision to the prior 2026 figure in light of cleanroom space constraints.\n\nWe estimate 4Q25 gross margin at 62%, above consensus (~61%). We also expect TSMC to guide 1Q26 gross margin to 61-63%, which is below our estimate of 63.1% but above consensus (~60%) and could be a positive surprise for the market. The backdrop is that ~10% of total orders are being processed on an expedited basis with substantial pricing premiums (half at ~50% premium “hot runs” and the other half at ~100% premium “super hot runs”).\n\nWhile the market is modeling gross margins below 60% for 2Q-4Q, Gokul forecasts 62.7%/61.2%/60.8% respectively, reflecting the high base from 1Q, which may lead market participants to question the sustainability of this pricing model.\n\nWe expect TSMC to guide 1Q26 revenue roughly flat QoQ (our forecast +3% / consensus +1%), full-year 2026 USD revenue growth in the mid-20% range, and Capex of US$46-50bn. We also expect no specific guidance on CoWoS capacity growth (our estimate +69%), and for the 2024-29 data center AI revenue CAGR to be guided to mid-50% (versus our model of 57%).\n\n$TSM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.024535065933688815, "ret_signed_m1d": -0.024535065933688815, "alpha_spy_m1d": -0.022967074579915314, "alpha_c_m1d": -0.020848989990650324, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0016879607323825319, "ret_signed_p1d": -0.0016879607323825319, "alpha_spy_p1d": 0.00031151437348775346, "alpha_c_p1d": -0.003914948673384933, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.03204026015571415, "ret_signed_p1w": 0.03204026015571415, "alpha_spy_p1w": 0.03707508831627537, "alpha_c_p1w": 0.007133663895238396, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.09084603001302849, "ret_signed_p1m": 0.09084603001302849, "alpha_spy_p1m": 0.09521914997510794, "alpha_c_p1m": 0.0547532486396094, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.1047408275997952, "ret_signed_p3m": 0.1047408275997952, "alpha_spy_p3m": 0.12400713618454351, "alpha_c_p3m": 0.0032459381184073433, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.084, -0.0987, -0.113, -0.1051, -0.1147, -0.1316, -0.1261, -0.1133, -0.1035, -0.0936, -0.0829, -0.0885, -0.0969, -0.0836, -0.1161, -0.1173, -0.1305, -0.1388, -0.1124, -0.1247, -0.1264, -0.1517, -0.1438, -0.1523, -0.1646, -0.1512, -0.1658, -0.1732, -0.1444, -0.1442, -0.1284, -0.1284, -0.1237, -0.1352, -0.122, -0.1119, -0.1194, -0.1141, -0.0926, -0.0879, -0.0677, -0.0811, -0.1198, -0.1327, -0.1353, -0.1652, -0.1419, -0.1291, -0.116, -0.105, -0.0994, -0.0994, -0.0872, -0.093, -0.097, -0.084, -0.084, -0.0367, -0.0287, -0.0131, -0.0395, -0.0415, -0.0245, 0.0, -0.0017, -0.014, 0.0297, 0.032, 0.032, -0.0139, -0.017, -0.0133, 0.0093, 0.0028, 0.0198, 0.0317, 0.0234, -0.0036, 0.0289, 0.012, -0.0182, -0.0031, 0.0515, 0.0713, 0.0908, 0.1276, 0.1095, 0.1043, 0.1043, 0.0977, 0.0919, 0.0863, 0.1169, 0.1154, 0.1627, 0.1687, 0.1358, 0.129, 0.1125, 0.0644, 0.0774, 0.0666, 0.0215, 0.051, 0.0462, 0.0687, 0.0149, 0.0197, 0.0255, 0.0458, 0.0264, 0.024, -0.0048, 0.023, 0.0375, 0.0511, -0.0143, -0.0124, -0.0433, 0.0215, 0.0322, 0.0248, 0.0248, 0.033, 0.0438, 0.106, 0.1047, 0.1202, 0.1171, 0.1483, 0.1338, 0.0983, 0.1199, 0.107, 0.1126, 0.1711, 0.1566, 0.2165, 0.2241, 0.1859, 0.1904, 0.1971, 0.202, 0.2139, 0.1922, 0.268, 0.2518, 0.2444, 0.2228, 0.2008, 0.2084, 0.2626, 0.2222, 0.1968, 0.1867, 0.2139, 0.2307, 0.2227, 0.2227, 0.2463, 0.2778, 0.2842, 0.2648, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2035581499597275615", "ticker_idx": 0, "ts": "2026-03-22T04:57:03+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Companies capable of system-level integration will have a durable structural advantage — the same logic that governed the mobile SoC era. Nvidia's attempted ARM acquisition (blocked), Mellanox acquisition, and Groq acquisition are just the opening moves of this new historical cycle.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long NVDA: Rubin+LPX closes the startup speed gap, system-level integration gives durable structural advantage; AI accelerator startups' addressable space is shrinking.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "GTC 2026 Recap: Nvidia Closes Its Biggest Gap — Dramatically Narrowing the Key Advantage of AI Accelerator Startups\n\nLooking back at my pre-GTC preview, the directional calls and technical roadmap were largely correct. What Nvidia ultimately delivered was more elegant than I anticipated: not only is prefill handled on the GPU, but the Attention phase of decode is also kept on the GPU (that part I didn't see coming) — only the MLP phase of decode is offloaded to the LPU.\n\nThis bears a striking resemblance to MatX's approach: Weights sit in SRAM, KV cache sits in HBM.\n\nThe logic is sound. The Attention phase demands enormous KV cache — easily tens to hundreds of GB — which is simply beyond what LPU SRAM can hold. Keeping that portion on HBM is the more rational choice.\n\nThis maps neatly onto the emerging trend in agentic workflows: longer multi-turn conversations, exploding long-context KV cache. Even when high batch concurrency generates massive KV cache, HBM can absorb it. As context length grows, all incremental cost falls on GPU HBM — LPX is entirely static, unaffected, scaling only with model weight size.\n\nThis lets the LPU's precious 128 GB of SRAM focus exclusively on the fixed weights of the FFN/MLP phase. FFN accounts for over 50% of the entire decode stage on a GPU — in short-context scenarios, it can exceed 60%. Accelerating FFN on the LPU by several multiples is a highly favorable trade.\n\nThe potential drawback of this design is that a typical transformer decode pass spans many layers — say 80. That means Attention and FFN alternate 80 times, i.e., tensors must shuttle between GPU and LPU 80 times per token. Even over low-latency Nvidia Spectrum-X Ethernet, generating a single token accumulates 80 round-trip GPU–LPU latencies. That is non-trivial overhead.\n\nUnder this new architecture, assuming a 40/60 Attention–FFN split, with FFN accelerated by several multiples on the LPU, peak theoretical throughput improvement reaches well over 2× compared to Rubin NVL72.\n\nBreaching 1,000 tokens/sec at peak speed, while still maintaining commercially meaningful throughput, is significant. On Blackwell, hitting 400–500 tokens/sec forced you down to handling very few concurrent requests — a massive waste of GPU resources. Now, even at 1,000 tokens/sec, a meaningful batch size can be sustained, finally unlocking real commercial value.\n\nThe slide shows that at 400 tokens/sec, Rubin + LPX delivers a 35× throughput improvement — classic token economics. At that token velocity, it represents 35× more commercial value versus Blackwell.\n\nSo what does this mean for startups — Cerebras, d-Matrix, MatX, SambaNova — now that Nvidia has closed this gap?\n\nThe core value proposition of these startups has always been a speed advantage or cost advantage in specific scenarios.\n\nIn large-batch (high-concurrency) settings, GPUs exceed the ridge point and approach high arithmetic intensity utilization — both cost and speed metrics become unfavorable for startups. Their survival niche has always been: workloads concentrated in small-batch, ultra-low latency, extreme token speed, where cost is no object. GPUs are deeply inefficient here and simply cannot match the required token rates.\n\nCerebras: Pure speed. Wafer-scale die with massive on-chip SRAM (40 GB) eliminates inter-chip communication — the dominant bottleneck — delivering extremely high token rates at small batch sizes. Cost competitiveness is nonexistent: a single CS-3 system runs $2.3M, far exceeding an equivalent GPU cluster — roughly 10× the cost for ~10× the speed versus H100.\n\nd-Matrix: High speed + small-batch scenarios. In-memory compute reduces data movement; utilization exceeds GPU efficiency in small-batch decode, giving it competitive perf/watt in that regime. The recently introduced 3D stacked DRAM addresses capacity and bandwidth scaling as larger reasoning models and higher token consumption put pressure on the stack.\n\nSambaNova: Targets enterprise on-premises deployments running multiple mid-size models concurrently. GPUs suffer significant context-switching overhead in this scenario, while SambaNova's RDU delivers better perf/dollar. The advantage is fundamentally a cost advantage in a specific use case — the raw speed lead is not as pronounced.\n\nMatX: Partitionable systolic array + hybrid SRAM/HBM — conceptually similar to Nvidia's new heterogeneous architecture. The standout feature is achieving Weights-in-SRAM and KV-cache-in-HBM within a single chip. Critically, by keeping everything on one chip, it eliminates the 80-layer GPU–LPU inter-chip round trips described above — so it retains a speed edge, but scalability may be more constrained than a GPU + LPU array.\n\nBottom line: With Rubin + LPX, the small-batch, low-latency, extreme-speed gap that previously favored startups has been largely closed. The viable addressable space for these companies is shrinking.\n\nOne more observation from my preview: speculative decoding using the LPU as the draft model and the GPU for verification — I predicted this would yield dramatic acceleration. That call landed squarely. Nvidia's official blog devoted an entire dedicated section to it: \"LPX generates draft tokens rapidly using its low-latency architecture. Rubin GPUs verify and finalize tokens efficiently.\"\n\nThe other item from my preview — CPX (Content Phase aXcelerator, a compute module purpose-built for the compute-bound characteristics of prefill) — appears to have vanished entirely from GTC. Not a single word. Does that mean CPX is cancelled?\n\nI don't think so, necessarily.\n\nCurrent prefill and decode already run in a disaggregated structure — a subset of GPUs handles prefill, another handles decode. Replacing the prefill-side GPUs with CPX is architecturally cleaner and would accelerate the prefill phase, though it adds cost as an additional chip. CPX has no fundamental conflict with the current Rubin + LPX architecture — it is simply a drop-in replacement for the prefill GPU partition. When speed optimization demand resurfaces, CPX may well return.\n\nSame reflection as last time: every shift in computing paradigm triggers a wave of semiconductor startups. But as the software and application landscape gradually converges, the outcome is invariably that the platform incumbents absorb the innovations through acquisition — scaling parameters, deepening system integration, lowering cost, improving power efficiency and benchmarks — until startups lose their independent reason to exist.\n\nIn the early days of mobile internet, it was the same: a fragmented ecosystem of independent AP, baseband, ISP, and GPU vendors. But the ultimate winners were the companies that integrated GPU, ISP, modem — everything — into a single SoC, completing the heterogeneous compute platform at the system level.\n\nApple acquired PA Semi's CPU team, Infineon's modem, and hollowed out Imagination's GPU. Qualcomm acquired ATI's mobile GPU, Atheros's Wi-Fi, Nuvia's CPU, CSR's Bluetooth/DSP. These are all canonical examples.\n\nThe complexity of heterogeneous inference is only increasing. Companies capable of system-level integration will have a durable structural advantage — the same logic that governed the mobile SoC era. Nvidia's attempted ARM acquisition (blocked), Mellanox acquisition, and Groq acquisition are just the opening moves of this new historical cycle.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": "复盘GTC 2026：Nvidia补上了短板，大幅削弱了各个AI 加速器 startup最大的优势--token速度\n\n回顾这篇GTC前瞻，方向预测和技术路线写的没啥大问题，最后Nvidia给出的解法比我想象的更为精巧：不仅是prefill放在GPU上，decode阶段Attention阶段也放在GPU上(这点没想到)，只把decode的MLP阶段放在LPU上做 https://t.co/0nX5A10pXS", "parent_author": null, "ret_m1d": -0.016738797834336183, "ret_signed_m1d": -0.016738797834336183, "alpha_spy_m1d": -0.006347883391025655, "alpha_c_m1d": 0.00017656235825147526, "bench_spy_m1d": -0.010390914443310528, "bench_c_m1d": -0.01691536019258766, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002505138026276832, "ret_signed_p1d": -0.002505138026276832, "alpha_spy_p1d": 0.0008517117259896345, "alpha_c_p1d": -0.010758437053326242, "bench_spy_p1d": -0.0033568497522664664, "bench_c_p1d": 0.00825329902704941, "ret_p1w": -0.059610574226181545, "ret_signed_p1w": -0.059610574226181545, "alpha_spy_p1w": -0.02389078722432847, "alpha_c_p1w": 0.014055583918134995, "bench_spy_p1w": -0.035719787001853076, "bench_c_p1w": -0.07366615814431654, "ret_p1m": 0.13800959677407443, "ret_signed_p1m": 0.13800959677407443, "alpha_spy_p1m": 0.06370154366870207, "alpha_c_p1m": -0.049286041651250834, "bench_spy_p1m": 0.07430805310537236, "bench_c_p1m": 0.18729563842532526, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0738, 0.0738, 0.0847, 0.0716, 0.0677, 0.0618, 0.0618, 0.0752, 0.071, 0.066, 0.0766, 0.0535, 0.0524, 0.0529, 0.0578, 0.0426, 0.0649, 0.0602, 0.0602, 0.0138, 0.0437, 0.0523, 0.0684, 0.0616, 0.0733, 0.0904, 0.096, 0.0881, 0.0567, 0.0267, -0.0083, -0.0215, 0.0556, 0.0819, 0.0734, 0.082, 0.0643, 0.0408, 0.0408, 0.0531, 0.0702, 0.0697, 0.0807, 0.0905, 0.0979, 0.1134, 0.0526, 0.0088, 0.0389, 0.0251, 0.0421, 0.0438, 0.0124, 0.0399, 0.0519, 0.0592, 0.0427, 0.0262, 0.0432, 0.0358, 0.0271, 0.0166, -0.0167, 0.0, -0.0025, 0.0173, -0.0251, -0.0462, -0.0596, -0.0071, 0.0006, 0.01, 0.01, 0.0114, 0.014, 0.0367, 0.0471, 0.074, 0.0778, 0.1188, 0.1323, 0.1293, 0.1483, 0.1504, 0.138, 0.1529, 0.1366, 0.1858, 0.2333, 0.2137, 0.1914, 0.1362, 0.1299, 0.13, 0.1188, 0.1833, 0.2042, 0.2252, 0.2494, 0.257, 0.2858, 0.3422, 0.2829, 0.2658, 0.256, 0.2723, 0.2498, 0.226, 0.226, 0.2233, 0.2104, 0.2198, 0.2021, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2020690749231370337", "ticker_idx": 0, "ts": "2026-02-09T02:46:31+00:00", "symbol": "Nikkei 225", "kind": "index", "geography": "Japan", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the Nikkei has no choice but to go up", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Nikkei on Takaichi landslide win securing mandate for fiscal and monetary stimulus.", "resolved_tickers": ["EWJ"], "resolution_method": "etf", "resolution_reason": "Nikkei 225 proxied by EWJ ETF", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "reply_other", "lang": "en", "tweet_text": "@SteveMatthews12 @takaichi_sanae Since Takaichi won a landslide victory in the general election and has secured the political mandate and governing capacity to push through more fiscal and monetary stimulus, the Nikkei has no choice but to go up.", "reply_text": null, "parent_text": "@jukan05 @takaichi_sanae For someone not following Japan, why is this bullish?", "quoted_text": null, "parent_author": "SteveMatthews12", "ret_m1d": -0.020703246474904735, "ret_signed_m1d": -0.020703246474904735, "alpha_spy_m1d": -0.015904552866786492, "alpha_c_m1d": -0.015904552866786492, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.004798693608118243, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01982689824122774, "ret_signed_p1d": 0.01982689824122774, "alpha_spy_p1d": 0.022464058673308585, "alpha_c_p1d": 0.022464058673308585, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.002637160432080843, "ret_p1w": 0.028042474892322256, "ret_signed_p1w": 0.028042474892322256, "alpha_spy_p1w": 0.04562304467666545, "alpha_c_p1w": 0.04562304467666545, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": -0.017580569784343192, "ret_p1m": -0.052908333690608655, "ret_signed_p1m": -0.052908333690608655, "alpha_spy_p1m": -0.028742271075554826, "alpha_c_p1m": -0.028742271075554826, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.02416606261505383, "ret_p3m": -0.0032862640897047823, "ret_signed_p3m": -0.0032862640897047823, "alpha_spy_p3m": -0.060391167812888225, "alpha_c_p3m": -0.060391167812888225, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.05710490372318344, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1095, -0.1235, -0.1133, -0.1292, -0.147, -0.1481, -0.1607, -0.1427, -0.1369, -0.135, -0.1209, -0.1209, -0.1235, -0.1286, -0.1292, -0.1277, -0.1171, -0.1162, -0.1216, -0.1203, -0.114, -0.1107, -0.113, -0.1039, -0.1144, -0.1265, -0.124, -0.1173, -0.1186, -0.1112, -0.1127, -0.1127, -0.113, -0.1117, -0.1129, -0.1156, -0.1156, -0.1091, -0.0894, -0.09, -0.0917, -0.0923, -0.0732, -0.0657, -0.0749, -0.0685, -0.0643, -0.0636, -0.0636, -0.0829, -0.0713, -0.0704, -0.0737, -0.073, -0.0597, -0.0666, -0.057, -0.061, -0.0592, -0.0505, -0.0416, -0.0569, -0.0207, 0.0, 0.0198, 0.0294, 0.0228, 0.028, 0.028, 0.0064, 0.001, -0.001, 0.002, 0.0011, -0.0007, 0.0126, 0.013, 0.0118, -0.0119, -0.0489, -0.0296, -0.0592, -0.0714, -0.0619, -0.0529, -0.0609, -0.0778, -0.0869, -0.0681, -0.068, -0.0782, -0.079, -0.1105, -0.0848, -0.0877, -0.0716, -0.094, -0.108, -0.107, -0.075, -0.0527, -0.0657, -0.0657, -0.0627, -0.0633, -0.0206, -0.0336, -0.0346, -0.0322, -0.0206, -0.0243, -0.0206, -0.012, -0.0215, -0.0455, -0.0387, -0.0462, -0.0435, -0.0395, -0.0409, -0.0491, -0.024, -0.0328, -0.0347, -0.0222, 0.0043, -0.0033, 0.0102, 0.0106, 0.0084, 0.0197, 0.0084, -0.0024, -0.0041, -0.011, -0.0009, 0.0009, 0.0035, 0.0035, 0.0176, 0.011, 0.0154, 0.0183, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012128444042383860", "ticker_idx": 0, "ts": "2026-01-16T11:42:59+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'd pick TSMC and the memory names. But the foundries that can actually manufacture those 'brain' logic chips, and the memory suppliers, feel much harder for China to truly penetrate in the long run.", "tweet_type": "analysis", "timeline": "years", "summary": "Long TSMC and memory names as robotic supply chain plays — logic foundry and memory are hardest for China to penetrate long-term.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Someone asked me, “Jukan, what is the best robotic supply chain play, in your opinion?” So I tried to think it through and come to my own conclusion.\n\nI’d pick TSMC and the memory names.\n\nI think most robot components can ultimately be produced in-house. People love to talk about reducers having a huge technological moat, but when I ask people in the industry, the answer to “is it really that hard to make?” is usually no — it’s not that impossible.\n\nIn the end, components other than the robot’s “brain” — meaning the logic chip and the memory chip paired with it — will have to compete head-on with Chinese companies. And honestly, I’m not sure you can maintain attractive profitability while competing with China in those areas.\n\nEven for the logic chip itself, China might eventually be able to design it domestically.\n\nBut the foundries that can actually manufacture those “brain” logic chips, and the memory suppliers, feel much harder for China to truly penetrate in the long run. That’s my conclusion.\n\nAnd yeah — it seems I ended up recommending the “safer” plays rather than the best return opportunities. But I genuinely worry about China dominating humanoids the way it did EVs. The components that go into EVs and humanoid robots are highly similar.\n\nChina has already taken over large parts of the supply chain — starting with smartphones and extending into EVs — and I can’t shake the feeling that it won’t take long before it also takes over the humanoid supply chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028735635015458172, "ret_signed_m1d": -0.028735635015458172, "alpha_spy_m1d": -0.029574211627886537, "alpha_c_m1d": -0.018795299740556826, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011494239929334293, "ret_signed_p1d": 0.011494239929334293, "alpha_spy_p1d": 0.011494239929334293, "alpha_c_p1d": 0.011494239929334293, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.01724139508612388, "ret_signed_p1w": 0.01724139508612388, "alpha_spy_p1w": 0.020754709890014045, "alpha_c_p1w": 0.017990710252762887, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.10057472255410338, "ret_signed_p1m": 0.10057472255410338, "alpha_spy_p1m": 0.11490255212192357, "alpha_c_p1m": 0.0822676062322798, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.1993024921301616, "ret_signed_p3m": 0.1993024921301616, "alpha_spy_p3m": 0.1845675410055425, "alpha_c_p3m": 0.0679056454886271, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1523, -0.1637, -0.1694, -0.1694, -0.1523, -0.1551, -0.1379, -0.1379, -0.1408, -0.1351, -0.1379, -0.1637, -0.1608, -0.1637, -0.1551, -0.1608, -0.1551, -0.1637, -0.1809, -0.1723, -0.1952, -0.2009, -0.1666, -0.2067, -0.2124, -0.1895, -0.1752, -0.178, -0.1752, -0.1923, -0.1809, -0.1694, -0.1723, -0.1637, -0.1437, -0.1523, -0.1379, -0.1552, -0.1494, -0.1667, -0.1753, -0.1782, -0.1782, -0.1782, -0.158, -0.1437, -0.1408, -0.1408, -0.1322, -0.1207, -0.1264, -0.1092, -0.1092, -0.0891, -0.0402, -0.0201, -0.0374, -0.0316, -0.0345, -0.0287, -0.0172, -0.0172, -0.0287, 0.0, 0.0115, 0.0201, 0.0, 0.0115, 0.0172, 0.0086, 0.023, 0.046, 0.0374, 0.0201, 0.0144, 0.0345, 0.0259, 0.0144, 0.023, 0.0431, 0.0805, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.1006, 0.092, 0.1293, 0.158, 0.1466, 0.1466, 0.1351, 0.1121, 0.0718, 0.092, 0.0862, 0.0402, 0.0632, 0.1149, 0.0833, 0.0718, 0.0603, 0.0782, 0.0984, 0.0667, 0.0609, 0.0436, 0.0436, 0.0638, 0.0609, 0.0494, 0.0263, 0.0148, 0.0696, 0.0436, 0.0436, 0.0436, 0.0725, 0.1243, 0.1272, 0.1532, 0.1474, 0.1849, 0.1993, 0.2022, 0.1705, 0.1676, 0.182, 0.182, 0.1993, 0.2598, 0.306, 0.2771, 0.257, 0.231, 0.231, 0.3117, 0.2973, 0.2973, 0.3319, 0.3204, 0.2887, 0.3002, 0.28, 0.3089, 0.306, 0.2916, 0.2714, 0.2598, 0.2858, 0.3002, 0.3319, 0.3089, 0.3262, 0.3233, 0.3579, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012128444042383860", "ticker_idx": 1, "ts": "2026-01-16T11:42:59+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I'd pick TSMC and the memory names. the memory suppliers, feel much harder for China to truly penetrate in the long run", "tweet_type": "analysis", "timeline": "years", "summary": "Long TSMC and memory names as robotic supply chain plays — logic foundry and memory are hardest for China to penetrate long-term.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Someone asked me, “Jukan, what is the best robotic supply chain play, in your opinion?” So I tried to think it through and come to my own conclusion.\n\nI’d pick TSMC and the memory names.\n\nI think most robot components can ultimately be produced in-house. People love to talk about reducers having a huge technological moat, but when I ask people in the industry, the answer to “is it really that hard to make?” is usually no — it’s not that impossible.\n\nIn the end, components other than the robot’s “brain” — meaning the logic chip and the memory chip paired with it — will have to compete head-on with Chinese companies. And honestly, I’m not sure you can maintain attractive profitability while competing with China in those areas.\n\nEven for the logic chip itself, China might eventually be able to design it domestically.\n\nBut the foundries that can actually manufacture those “brain” logic chips, and the memory suppliers, feel much harder for China to truly penetrate in the long run. That’s my conclusion.\n\nAnd yeah — it seems I ended up recommending the “safer” plays rather than the best return opportunities. But I genuinely worry about China dominating humanoids the way it did EVs. The components that go into EVs and humanoid robots are highly similar.\n\nChina has already taken over large parts of the supply chain — starting with smartphones and extending into EVs — and I can’t shake the feeling that it won’t take long before it also takes over the humanoid supply chain.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03828145706827416, "ret_signed_m1d": -0.03828145706827416, "alpha_spy_m1d": -0.03912003368070253, "alpha_c_m1d": -0.028341121793372816, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004422791523225185, "ret_signed_p1d": 0.004422791523225185, "alpha_spy_p1d": 0.004422791523225185, "alpha_c_p1d": 0.004422791523225185, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.045921384475958206, "ret_signed_p1w": 0.045921384475958206, "alpha_spy_p1w": 0.04943469927984837, "alpha_c_p1w": 0.046670699642597215, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.17192548513627437, "ret_signed_p1m": 0.17192548513627437, "alpha_spy_p1m": 0.18625331470409456, "alpha_c_p1m": 0.1536183688144508, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.39454545535865587, "ret_signed_p3m": 0.39454545535865587, "alpha_spy_p3m": 0.3798105042340368, "alpha_c_p3m": 0.26314860871712137, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3859, -0.3859, -0.3843, -0.354, -0.3349, -0.3449, -0.3221, -0.3121, -0.3087, -0.2637, -0.3078, -0.3026, -0.2988, -0.3078, -0.2762, -0.2752, -0.2735, -0.2837, -0.31, -0.2866, -0.3211, -0.3299, -0.3393, -0.3688, -0.3498, -0.3439, -0.3286, -0.3181, -0.3252, -0.3174, -0.3036, -0.3087, -0.3187, -0.3008, -0.2748, -0.2771, -0.2584, -0.2744, -0.2841, -0.3041, -0.3237, -0.3095, -0.2886, -0.2777, -0.244, -0.2403, -0.2299, -0.2299, -0.2136, -0.1798, -0.1757, -0.1823, -0.1823, -0.124, -0.0971, -0.0534, -0.0452, -0.0554, -0.0437, -0.0412, -0.0559, -0.0524, -0.0383, 0.0, 0.0044, -0.0119, 0.0185, 0.0392, 0.0459, 0.0225, 0.0868, 0.1343, 0.1398, 0.1413, 0.1049, 0.1603, 0.124, 0.0797, 0.0877, 0.116, 0.1004, 0.1319, 0.1718, 0.1719, 0.1719, 0.161, 0.1805, 0.203, 0.2374, 0.2382, 0.275, 0.2986, 0.3553, 0.3323, 0.3325, 0.2004, 0.1288, 0.2094, 0.1697, 0.1154, 0.2054, 0.2319, 0.2039, 0.2043, 0.2586, 0.2868, 0.3575, 0.3046, 0.2798, 0.2008, 0.2237, 0.2137, 0.1419, 0.1436, 0.0769, 0.0421, 0.1598, 0.1035, 0.1413, 0.1718, 0.1925, 0.3024, 0.286, 0.3024, 0.3024, 0.3785, 0.3945, 0.4184, 0.4012, 0.4085, 0.4451, 0.4763, 0.4876, 0.4888, 0.5566, 0.5358, 0.5548, 0.5381, 0.5612, 0.6907, 0.7493, 0.9167, 0.9341, 1.0336, 1.202, 1.1412, 1.2487, 1.2476, 1.0765, 1.0697, 1.0312, 1.0629, 1.2294, 1.2039, 1.2039, 1.4007, 1.533, 1.5321, 1.6345, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2005778239357346131", "ticker_idx": 0, "ts": "2025-12-29T23:09:32+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This is a very bullish piece on ASML", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish ASML and Intel: Broadcom-designed AWS chips use Intel EMIB packaging; Nvidia's next chip reportedly headed to Intel 14A.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "This is a very bullish piece on ASML, but if you look closely, it’s also bullish on Intel.\n\nAccording to the article, Broadcom-designed AWS chips will be packaged using Intel’s EMIB. It even claims that Nvidia’s “Nintendo Switch 3” chip will go into mass production on Intel 14A.\n\nYou should definitely read it.\n\n---\n\nhttps://t.co/guzhfDNyK6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008601652863979847, "ret_signed_m1d": -0.008601652863979847, "alpha_spy_m1d": -0.012178025934989511, "alpha_c_m1d": -0.013131989579599157, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.00453033671561931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0127923090785933, "ret_signed_p1d": 0.0127923090785933, "alpha_spy_p1d": 0.014013409813175559, "alpha_c_p1d": 0.015263394215189185, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0024710851365958852, "ret_p1w": 0.1614467193592375, "ret_signed_p1w": 0.1614467193592375, "alpha_spy_p1w": 0.1616357034717596, "alpha_c_p1w": 0.1247096577828728, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": 0.036737061576364693, "ret_p1m": 0.3427436666647754, "ret_signed_p1m": 0.3427436666647754, "alpha_spy_p1m": 0.3316365792267024, "alpha_c_p1m": 0.22457005925502238, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.11817360740975302, "ret_p3m": 0.2873452719752094, "ret_signed_p3m": 0.2873452719752094, "alpha_spy_p3m": 0.34694867226562565, "alpha_c_p3m": 0.24168478688974293, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": 0.045660485085466496, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0738, -0.0339, -0.0312, -0.0122, -0.037, -0.0623, -0.0645, -0.104, -0.0709, -0.068, -0.039, -0.0341, -0.036, -0.0103, -0.0205, -0.0422, -0.0228, -0.0154, -0.0003, 0.0007, 0.0139, 0.0339, 0.0125, 0.0217, 0.0124, -0.0002, -0.0127, -0.0397, -0.0222, -0.0219, -0.0165, -0.0279, -0.0361, -0.0335, -0.0454, -0.0223, -0.0187, -0.0803, -0.0529, -0.0513, 0.0028, -0.0106, -0.0037, 0.0219, 0.0356, 0.0624, 0.0557, 0.0494, 0.0622, 0.0508, 0.0432, 0.0376, 0.0204, 0.0267, 0.002, -0.0362, -0.0155, -0.0057, -0.0119, -0.004, -0.0086, -0.0086, -0.0086, 0.0, 0.0128, 0.0161, 0.0161, 0.0877, 0.1614, 0.17, 0.1599, 0.1171, 0.1928, 0.1981, 0.2159, 0.1956, 0.2675, 0.2872, 0.2356, 0.2572, 0.2735, 0.2973, 0.2993, 0.2991, 0.3427, 0.3172, 0.3145, 0.3405, 0.3507, 0.3127, 0.2578, 0.2678, 0.3165, 0.3304, 0.3174, 0.3337, 0.3028, 0.3145, 0.3196, 0.3242, 0.3746, 0.3673, 0.3865, 0.3794, 0.3951, 0.4227, 0.3609, 0.362, 0.3366, 0.2829, 0.3249, 0.3097, 0.2666, 0.2672, 0.3251, 0.3238, 0.3035, 0.3026, 0.3222, 0.3183, 0.3152, 0.2904, 0.2458, 0.2979, 0.3269, 0.3379, 0.2873, 0.2659, 0.2279, 0.2359, 0.3114, 0.282, 0.282, 0.282, 0.2299, 0.339, 0.3635, 0.4024, 0.3911, 0.4181, 0.3582, 0.3501, 0.3737, 0.3748, 0.366, 0.3801, 0.3514, 0.3829, 0.342, 0.2968, 0.3199, 0.3528, 0.3528, 0.3141, 0.3601, 0.4442, 0.4387, 0.4602, 0.4449, 0.4009, 0.4686, 0.5129, 0.446, 0.3993, 0.3823, 0.4752, 0.4887, 0.5593, 0.5832, 0.5381, 0.5233, 0.5394, 0.5326, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2005778239357346131", "ticker_idx": 1, "ts": "2025-12-29T23:09:32+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's also bullish on Intel", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish ASML and Intel: Broadcom-designed AWS chips use Intel EMIB packaging; Nvidia's next chip reportedly headed to Intel 14A.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "This is a very bullish piece on ASML, but if you look closely, it’s also bullish on Intel.\n\nAccording to the article, Broadcom-designed AWS chips will be packaged using Intel’s EMIB. It even claims that Nvidia’s “Nintendo Switch 3” chip will go into mass production on Intel 14A.\n\nYou should definitely read it.\n\n---\n\nhttps://t.co/guzhfDNyK6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.01308613790192903, "ret_signed_m1d": -0.01308613790192903, "alpha_spy_m1d": -0.016662510972938693, "alpha_c_m1d": -0.01761647461754834, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.00453033671561931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016902915123402584, "ret_signed_p1d": 0.016902915123402584, "alpha_spy_p1d": 0.018124015857984843, "alpha_c_p1d": 0.01937400025999847, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0024710851365958852, "ret_p1w": 0.07333693032519983, "ret_signed_p1w": 0.07333693032519983, "alpha_spy_p1w": 0.07352591443772194, "alpha_c_p1w": 0.03659986874883514, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": 0.036737061576364693, "ret_p1m": 0.19765539639259422, "ret_signed_p1m": 0.19765539639259422, "alpha_spy_p1m": 0.18654830895452124, "alpha_c_p1m": 0.0794817889828412, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.11817360740975302, "ret_p3m": 0.2022900247331323, "ret_signed_p3m": 0.2022900247331323, "alpha_spy_p3m": 0.2618934250235485, "alpha_c_p3m": 0.1566295396476658, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": 0.045660485085466496, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0202, 0.0169, 0.0041, -0.0025, 0.0134, 0.0204, 0.0305, -0.0085, 0.0147, -0.0286, 0.0128, 0.0044, 0.009, 0.0387, 0.0393, 0.0065, 0.0403, 0.0436, 0.078, 0.1322, 0.127, 0.0949, 0.0902, 0.0769, 0.0095, 0.0463, 0.0153, 0.0395, 0.0483, 0.0327, 0.033, -0.021, -0.0316, -0.0537, -0.0641, -0.0428, -0.0834, -0.0594, -0.0243, -0.0232, 0.0035, 0.0035, 0.1058, 0.0908, 0.1851, 0.193, 0.1041, 0.129, 0.0987, 0.1041, 0.1118, 0.0772, 0.0308, 0.0226, 0.0172, -0.0172, -0.0109, 0.0038, -0.0085, -0.009, -0.0142, -0.0142, -0.0131, 0.0, 0.0169, 0.006, 0.006, 0.0736, 0.0733, 0.0916, 0.1622, 0.1208, 0.2418, 0.2012, 0.2893, 0.3282, 0.3173, 0.2803, 0.2803, 0.3239, 0.479, 0.4809, 0.2287, 0.1584, 0.1977, 0.3299, 0.3266, 0.2669, 0.3307, 0.3427, 0.325, 0.3152, 0.3792, 0.3697, 0.2849, 0.3165, 0.2672, 0.2756, 0.2756, 0.259, 0.2394, 0.2165, 0.2026, 0.1895, 0.2574, 0.2781, 0.2394, 0.2435, 0.2405, 0.175, 0.2426, 0.2527, 0.1838, 0.2426, 0.2754, 0.3081, 0.2336, 0.2478, 0.2475, 0.2012, 0.2276, 0.259, 0.196, 0.1998, 0.2012, 0.2863, 0.2023, 0.1758, 0.123, 0.2031, 0.3094, 0.3735, 0.3735, 0.3844, 0.4425, 0.6071, 0.6827, 0.7007, 0.777, 0.7396, 0.7704, 0.8675, 0.8675, 0.7912, 0.8064, 0.7794, 0.8206, 1.2503, 1.3171, 1.3043, 1.5832, 1.5758, 1.7159, 1.6112, 1.9485, 2.081, 1.9885, 2.4057, 2.5289, 2.2882, 2.2794, 2.1606, 1.9654, 1.949, 2.0207, 2.2432, 2.2306, 2.2672, 2.2672, 2.3675, 2.3198, 2.2958, 2.1265, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2016691994312266097", "ticker_idx": 0, "ts": "2026-01-29T01:56:54+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$BESI", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Author highlights Samsung's hybrid bonding commercialization plan for HBM4E, tagging BESI as the implied beneficiary.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "* Samsung: Planning to commercialize hybrid bonding in part for HBM4E.\n\n$BESI", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04244843586962488, "ret_signed_m1d": 0.04244843586962488, "alpha_spy_m1d": 0.04046010781736542, "alpha_c_m1d": 0.044580032330340025, "bench_spy_m1d": 0.001988328052259458, "bench_c_m1d": -0.002131596460715146, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011996348432261161, "ret_signed_p1d": 0.011996348432261161, "alpha_spy_p1d": 0.014978928692664373, "alpha_c_p1d": 0.045671377002290514, "bench_spy_p1d": -0.0029825802604032114, "bench_c_p1d": -0.03367502857002935, "ret_p1w": -0.01107353786621712, "ret_signed_p1w": -0.01107353786621712, "alpha_spy_p1w": 0.012584991040861193, "alpha_c_p1w": 0.0762276277032703, "bench_spy_p1w": -0.023658528907078313, "bench_c_p1w": -0.08730116556948742, "ret_p1m": 0.1651798795745909, "ret_signed_p1m": 0.1651798795745909, "alpha_spy_p1m": 0.1767786362434659, "alpha_c_p1m": 0.19188517318897458, "bench_spy_p1m": -0.011598756668874999, "bench_c_p1m": -0.026705293614383674, "ret_p3m": 0.45668051937354304, "ret_signed_p3m": 0.45668051937354304, "alpha_spy_p3m": 0.4284492239471296, "alpha_c_p3m": 0.2801859696982383, "bench_spy_p3m": 0.02823129542641345, "bench_c_p3m": 0.17649454967530476, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.103, -0.123, -0.1295, -0.1526, -0.1747, -0.1606, -0.1523, -0.1673, -0.1769, -0.1916, -0.199, -0.2178, -0.2058, -0.1999, -0.2445, -0.2347, -0.2335, -0.2027, -0.2009, -0.2009, -0.2018, -0.1919, -0.1636, -0.1541, -0.1415, -0.0966, -0.1184, -0.136, -0.1578, -0.1873, -0.1852, -0.1781, -0.2046, -0.1913, -0.1956, -0.1929, -0.1886, -0.1907, -0.1907, -0.1907, -0.1873, -0.1812, -0.1772, -0.1772, -0.0827, -0.0541, -0.0102, -0.0231, -0.0704, -0.0704, -0.0025, 0.0286, -0.0065, 0.0683, 0.0637, 0.0384, 0.0637, 0.0707, 0.0766, 0.0769, 0.0757, 0.0812, 0.0424, 0.0, 0.012, 0.0126, -0.0148, -0.0308, -0.0111, 0.0181, 0.0298, 0.04, 0.0514, 0.0378, 0.084, 0.0981, 0.108, 0.1504, 0.0643, 0.135, 0.1544, 0.183, 0.2156, 0.1621, 0.1652, 0.1596, 0.1157, 0.175, 0.1606, -0.0384, 0.0231, 0.0741, 0.0763, 0.0781, 0.1387, 0.1258, 0.1593, 0.1836, 0.131, 0.0981, 0.123, 0.1326, 0.1412, 0.1406, 0.0757, 0.0717, 0.1009, 0.1538, 0.1698, 0.1698, 0.1698, 0.1772, 0.2735, 0.2679, 0.3011, 0.2728, 0.3411, 0.3387, 0.3565, 0.3953, 0.3965, 0.402, 0.4291, 0.4875, 0.5515, 0.5353, 0.4567, 0.4821, 0.5304, 0.5304, 0.5124, 0.5724, 0.5786, 0.5768, 0.6145, 0.6176, 0.5464, 0.5984, 0.6511, 0.6207, 0.5836, 0.578, 0.6405, 0.674, 0.6938, 0.7458, 0.7582, 0.7111, 0.7805, 0.7606, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015255130317467966", "ticker_idx": 0, "ts": "2026-01-25T02:47:19+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we remain skeptical... Intel has a severe shortage with no revenue growth... foundry losses remain at a $10 bn run rate... AMD captured all of the unit growth... AMD's server products on 5 nm and even 7 nm likely outperform Granite Rapids, and even potentially Diamond Rapids", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's skeptical INTC review: AMD winning server share on superior products while Intel faces foundry losses, no revenue growth, and structural challenges.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s comment from its Intel earnings review:\n\nAs the microprocessor market has been in shortage since September, some of the focus has shifted to supply, which could favor Intel - but we remain skeptical. The server market is strong, but at least some of this is a function of Intel's lack of supply growth. With no revenue growth in 2H25, AMD captured all of the unit growth, which surprised them - our sense has been that AMD's tight supply demand situation on server CPU has more to do with demand upside than any real constraints.\n\nLooking forward, Intel should have better than seasonal supply growth beyond 1q, but AMD should as well. While 3 nm wafers remain tight, we believe that AMD's server products on 5 nm and even 7 nm likely outperform Granite Rapids, and even potentially Diamond Rapids given Intel's concerns around the lack of symmetric multithreading - the best processors will still likely win.\n\nFull year commentary on the call and the callback points to double digit growth for data center in CY26 (we were at +12% coming into the year), with declines in PCs given the memory shortage impact (we were up slightly). There will be price increases given ongoing shortages but we would think of them as mild, given competitive issues and supply shortfalls, and if the shortage eases in PCs prices will be weaker there as well.\n\nThis backdrop does create incremental challenges for the foundry prospects. In addition to just proving their basic foundry capabilities, customer objections to using Intel have been an on-again, off-again commitment to the foundry business, and the need to compete with the Intel product group for wafers. The fact that Intel has a severe shortage with no revenue growth will not ease those concerns - and while the current management team isn't to blame, we have had several shortages over the last 8 years as revenues have declined from $70 bn to $52 bn. Samsung remains the most viable alternative to TSMC, and TSMC's buildout in Arizona reduces the geopolitical imperative. The foundry losses remain at a $10 bn run rate, and plans to improve that are going to come mostly from charging more for the wafers, which puts the burden on the product group to build higher value products - which remains the key driver here.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06072012036810337, "ret_signed_m1d": -0.06072012036810337, "alpha_spy_m1d": -0.06577261014261604, "alpha_c_m1d": -0.057535754011479145, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.033890293476706246, "ret_signed_p1d": -0.033890293476706246, "alpha_spy_p1d": -0.02990602874702164, "alpha_c_p1d": -0.012752957031776813, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.14874086714915546, "ret_signed_p1w": -0.14874086714915546, "alpha_spy_p1w": -0.14487207426907722, "alpha_c_p1w": -0.1257982019657462, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.08543179830603753, "ret_signed_p1m": -0.08543179830603753, "alpha_spy_p1m": -0.09319818576701389, "alpha_c_p1m": -0.03443135709358014, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.5716638301085273, "ret_signed_p3m": -0.5716638301085273, "alpha_spy_p3m": -0.5461779471826702, "alpha_c_p3m": -0.3634746789765029, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0271, -0.0548, -0.0588, -0.0704, -0.1285, -0.0967, -0.1236, -0.1026, -0.0951, -0.1085, -0.1083, -0.1549, -0.164, -0.1831, -0.192, -0.1737, -0.2088, -0.188, -0.1577, -0.1567, -0.1337, -0.1337, -0.0454, -0.0584, 0.0231, 0.0299, -0.0468, -0.0254, -0.0515, -0.0468, -0.0402, -0.0701, -0.1101, -0.1172, -0.1219, -0.1516, -0.1462, -0.1334, -0.144, -0.1445, -0.149, -0.149, -0.148, -0.1367, -0.1221, -0.1316, -0.1316, -0.0732, -0.0734, -0.0577, 0.0033, -0.0325, 0.072, 0.0369, 0.113, 0.1466, 0.1372, 0.1052, 0.1052, 0.1429, 0.2768, 0.2784, 0.0607, 0.0, 0.0339, 0.148, 0.1452, 0.0937, 0.1487, 0.1591, 0.1438, 0.1353, 0.1906, 0.1824, 0.1092, 0.1365, 0.0939, 0.1012, 0.1012, 0.0868, 0.0699, 0.0501, 0.0381, 0.0268, 0.0854, 0.1033, 0.0699, 0.0734, 0.0708, 0.0144, 0.0727, 0.0814, 0.0219, 0.0727, 0.101, 0.1292, 0.065, 0.0772, 0.077, 0.0369, 0.0598, 0.0868, 0.0325, 0.0358, 0.0369, 0.1104, 0.0379, 0.0151, -0.0306, 0.0386, 0.1304, 0.1857, 0.1857, 0.1951, 0.2452, 0.3874, 0.4526, 0.4681, 0.534, 0.5018, 0.5284, 0.6121, 0.6121, 0.5462, 0.5594, 0.5361, 0.5717, 0.9426, 1.0002, 0.9892, 1.2299, 1.2236, 1.3446, 1.2542, 1.5453, 1.6597, 1.5799, 1.94, 2.0464, 1.8386, 1.831, 1.7284, 1.5599, 1.5458, 1.6077, 1.7997, 1.7889, 1.8204, 1.8204, 1.907, 1.8659, 1.8451, 1.699, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2015255130317467966", "ticker_idx": 1, "ts": "2026-01-25T02:47:19+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's tight supply demand situation on server CPU has more to do with demand upside than any real constraints... AMD should as well [have better than seasonal supply growth]... the best processors will still likely win", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley's skeptical INTC review: AMD winning server share on superior products while Intel faces foundry losses, no revenue growth, and structural challenges.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s comment from its Intel earnings review:\n\nAs the microprocessor market has been in shortage since September, some of the focus has shifted to supply, which could favor Intel - but we remain skeptical. The server market is strong, but at least some of this is a function of Intel's lack of supply growth. With no revenue growth in 2H25, AMD captured all of the unit growth, which surprised them - our sense has been that AMD's tight supply demand situation on server CPU has more to do with demand upside than any real constraints.\n\nLooking forward, Intel should have better than seasonal supply growth beyond 1q, but AMD should as well. While 3 nm wafers remain tight, we believe that AMD's server products on 5 nm and even 7 nm likely outperform Granite Rapids, and even potentially Diamond Rapids given Intel's concerns around the lack of symmetric multithreading - the best processors will still likely win.\n\nFull year commentary on the call and the callback points to double digit growth for data center in CY26 (we were at +12% coming into the year), with declines in PCs given the memory shortage impact (we were up slightly). There will be price increases given ongoing shortages but we would think of them as mild, given competitive issues and supply shortfalls, and if the shortage eases in PCs prices will be weaker there as well.\n\nThis backdrop does create incremental challenges for the foundry prospects. In addition to just proving their basic foundry capabilities, customer objections to using Intel have been an on-again, off-again commitment to the foundry business, and the need to compete with the Intel product group for wafers. The fact that Intel has a severe shortage with no revenue growth will not ease those concerns - and while the current management team isn't to blame, we have had several shortages over the last 8 years as revenues have declined from $70 bn to $52 bn. Samsung remains the most viable alternative to TSMC, and TSMC's buildout in Arizona reduces the geopolitical imperative. The foundry losses remain at a $10 bn run rate, and plans to improve that are going to come mostly from charging more for the wafers, which puts the burden on the product group to build higher value products - which remains the key driver here.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03330546018264169, "ret_signed_m1d": 0.03330546018264169, "alpha_spy_m1d": 0.03835794995715436, "alpha_c_m1d": 0.030121093826017464, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.002864992350872342, "ret_signed_p1d": 0.002864992350872342, "alpha_spy_p1d": -0.0011192723788122638, "alpha_c_p1d": -0.01827234409405709, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": -0.020054885739106854, "ret_signed_p1w": -0.020054885739106854, "alpha_spy_p1w": -0.023923678619185096, "alpha_c_p1w": -0.042997550922516115, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": -0.14909872899890053, "ret_signed_p1m": -0.14909872899890053, "alpha_spy_p1m": -0.14133234153792418, "alpha_c_p1m": -0.20009917021135792, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.21495360128312013, "ret_signed_p3m": 0.21495360128312013, "alpha_spy_p3m": 0.18946771835726306, "alpha_c_p3m": 0.0067644501510957245, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0518, 0.014, 0.0191, 0.0332, -0.005, 0.02, -0.0542, -0.0707, -0.0292, -0.0549, 0.0302, -0.0133, -0.0179, -0.0429, -0.0836, -0.1105, -0.1802, -0.1891, -0.1443, -0.1798, -0.1475, -0.1475, -0.1344, -0.1255, -0.1435, -0.1341, -0.1406, -0.1327, -0.1202, -0.1181, -0.1189, -0.1189, -0.1613, -0.174, -0.1677, -0.2117, -0.2, -0.1507, -0.1447, -0.1449, -0.1443, -0.1443, -0.1445, -0.1421, -0.1431, -0.1478, -0.1478, -0.1108, -0.1203, -0.1471, -0.1643, -0.1855, -0.1916, -0.1736, -0.1207, -0.1103, -0.0931, -0.0775, -0.0775, -0.0772, -0.006, 0.0096, 0.0333, 0.0, 0.0029, 0.0057, 0.0035, -0.058, -0.0201, -0.0366, -0.2034, -0.234, -0.1706, -0.1405, -0.1502, -0.1501, -0.1805, -0.175, -0.175, -0.1919, -0.2037, -0.1908, -0.2036, -0.2177, -0.1491, -0.161, -0.1895, -0.2033, -0.2097, -0.2402, -0.1959, -0.2064, -0.2343, -0.1935, -0.1913, -0.185, -0.2132, -0.2305, -0.2178, -0.2189, -0.2063, -0.1832, -0.1989, -0.1935, -0.1828, -0.1235, -0.1892, -0.1963, -0.2199, -0.1905, -0.1635, -0.1345, -0.1345, -0.1239, -0.1185, -0.0776, -0.0584, -0.0249, -0.0178, 0.015, 0.0271, 0.1072, 0.1078, 0.0941, 0.132, 0.2075, 0.215, 0.384, 0.3315, 0.2861, 0.3414, 0.4106, 0.4346, 0.359, 0.4136, 0.6768, 0.6253, 0.8113, 0.8256, 0.7838, 0.7727, 0.7894, 0.6876, 0.6752, 0.6476, 0.781, 0.789, 0.8603, 0.8603, 1.0051, 0.9718, 1.0616, 1.0536, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010924680144986481", "ticker_idx": 0, "ts": "2026-01-13T03:59:39+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Fundamentals clearly remain unattractive here... We don't find ourselves tempted by the President's whims. We rate Intel Market-Perform with a $35 TP.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Bernstein Market-Perform on INTC with $35 TP; fundamentals unattractive on PC demand, share losses to AMD/Arm, margin headwinds, and weak foundry progress.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Intel (INTC, Market-Perform, $35 TP) - Trump wants the stock to go up...\n\nFundamentals clearly remain unattractive here. Sustainability of core markets, especially PC, remains questionable as the current windows EOL won't last forever and the trajectory of the AI PC refresh cycle remains up for debate, plus rapidly increasing memory prices seem likely to put a further damper on demand into 2026. Furthermore, the company has continued to lose substantial amounts of share both to traditional competitors like AMD as well as ARM (with AAPL taking 10%+ away over the last several years, and now with new entrants like QCOM, and potentially NVDA and Mediatek, entering the mix). And in datacenter while we do believe server CPU growth can be strong as upgrades finally come through the company continues to bleed share in traditional x86 CPUs to AMD, while x86 in general loses share to Arm as hyperscalers increasingly deploy their own internal solutions, and CPUs continue to lose share to GPUs and accelerators with the company having effectively zero response to the advent of AI (Exhibit 9).\n\nWhile the company is ramping 18A and suggesting they feel more positive around 14A, supply constraints appear primarily on 10/7nm (where demand is higher because customers are less enthused by Intel's newer products) and seems likely to cost further share. It also sounds like 18A yields, while “adequate” to ship, still leave something to be desired (they suggested they wouldn't really be at mature levels until the end of next year), and 14A remains a long way off. And it increasingly sounds like 2026 margins could be faced with a variety of headwinds.\n\nThat being said, the stock is clearly NOT trading on fundamentals. And we admit that “Donald Trump wants the stock to go up” may indeed be viable bull case (for now it clearly has been). (Exhibit 10). But while we wouldn't necessarily talk you out of it if you are feeling a bit degenerate it's not our kind of bull case, and fundamentals there appear unattractive to us regardless of what the President might desire.\n\nWe don't find ourselves tempted by the President's whims. We rate Intel Market-Perform with a $35 TP.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06830195558688978, "ret_signed_m1d": 0.06830195558688978, "alpha_spy_m1d": 0.07030543660317723, "alpha_c_m1d": 0.0660799161010267, "bench_spy_m1d": 0.002003481016287445, "bench_c_m1d": -0.002222039485863081, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03023895702032542, "ret_signed_p1d": -0.03023895702032542, "alpha_spy_p1d": -0.035154194361944, "alpha_c_p1d": -0.03836092132879665, "bench_spy_p1d": -0.0049152373416185835, "bench_c_p1d": -0.00812196430847123, "ret_p1w": 0.02685558116254283, "ret_signed_p1w": -0.02685558116254283, "alpha_spy_p1w": -0.050191887386116574, "alpha_c_p1w": -0.02976719491988855, "bench_spy_p1w": -0.023336306223573744, "bench_c_p1w": -0.0029116137573457213, "ret_p1m": 0.021146119277651643, "ret_signed_p1m": -0.021146119277651643, "alpha_spy_p1m": -0.023755028950307344, "alpha_c_p1m": 0.038236303710339925, "bench_spy_p1m": -0.0026089096726557015, "bench_c_p1m": 0.05938242298799157, "ret_p3m": 0.31909494312640607, "ret_signed_p3m": -0.31909494312640607, "alpha_spy_p3m": -0.33704670114549085, "alpha_c_p3m": -0.20326727667707445, "bench_spy_p3m": -0.017951758019084774, "bench_c_p3m": 0.11582766644933162, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.221, -0.2174, -0.1943, -0.1939, -0.2193, -0.1931, -0.1905, -0.1639, -0.1218, -0.1258, -0.1508, -0.1544, -0.1647, -0.217, -0.1884, -0.2125, -0.1937, -0.1869, -0.199, -0.1988, -0.2406, -0.2489, -0.266, -0.2741, -0.2576, -0.2891, -0.2705, -0.2432, -0.2423, -0.2216, -0.2216, -0.1423, -0.1539, -0.0808, -0.0746, -0.1436, -0.1243, -0.1478, -0.1436, -0.1377, -0.1645, -0.2005, -0.2068, -0.211, -0.2377, -0.2328, -0.2214, -0.2309, -0.2313, -0.2354, -0.2354, -0.2345, -0.2244, -0.2112, -0.2197, -0.2197, -0.1673, -0.1675, -0.1533, -0.0985, -0.1307, -0.0368, -0.0683, 0.0, 0.0302, 0.0218, -0.007, -0.007, 0.0269, 0.1472, 0.1487, -0.0469, -0.1015, -0.0711, 0.0315, 0.029, -0.0173, 0.0321, 0.0414, 0.0277, 0.0201, 0.0698, 0.0624, -0.0034, 0.0211, -0.0171, -0.0106, -0.0106, -0.0235, -0.0387, -0.0565, -0.0672, -0.0774, -0.0247, -0.0087, -0.0387, -0.0355, -0.0379, -0.0886, -0.0362, -0.0283, -0.0818, -0.0362, -0.0108, 0.0146, -0.0431, -0.0321, -0.0324, -0.0683, -0.0478, -0.0235, -0.0723, -0.0694, -0.0683, -0.0023, -0.0675, -0.088, -0.129, -0.0668, 0.0156, 0.0653, 0.0653, 0.0738, 0.1188, 0.2466, 0.3051, 0.3191, 0.3783, 0.3493, 0.3732, 0.4485, 0.4485, 0.3893, 0.4011, 0.3802, 0.4121, 0.7454, 0.7972, 0.7873, 1.0036, 0.9979, 1.1066, 1.0254, 1.287, 1.3897, 1.318, 1.6416, 1.7372, 1.5504, 1.5437, 1.4515, 1.3001, 1.2874, 1.343, 1.5155, 1.5058, 1.5342, 1.5342, 1.612, 1.575, 1.5564, 1.425, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2033238007709577277", "ticker_idx": 0, "ts": "2026-03-15T17:44:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron look like the obvious beneficiaries", "tweet_type": "analysis", "timeline": "weeks", "summary": "Samsung strike risk could halt memory lines for months; SK Hynix and Micron are near-term beneficiaries.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I feel like I’ve been talking too much about myself, so to shift the conversation back to memory…\n\nIt looks like Samsung’s semiconductor division may genuinely be heading toward a strike. The union is demanding a performance bonus equal to 20% of operating profit, while management is holding firm, saying it cannot remove the cap on bonuses.\n\nThe union seems willing to do whatever it takes to eliminate the bonus gap with SK Hynix. There is a real possibility that Samsung’s memory production lines could actually be halted.\n\nOnce a line is shut down, it can take as long as two months to restart. The losses could amount to tens of billions of dollars.\n\nThat would be good news for Samsung’s competitors. In the near term, SK Hynix and Micron look like the obvious beneficiaries.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06570836986729167, "ret_signed_m1d": -0.06570836986729167, "alpha_spy_m1d": -0.055634028974238636, "alpha_c_m1d": -0.04897901712858477, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.016729352738706904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004106757071928202, "ret_signed_p1d": -0.004106757071928202, "alpha_spy_p1d": -0.006737311941903412, "alpha_c_p1d": -0.01162095120824902, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.0075141941363208176, "ret_p1w": -0.04209443647981692, "ret_signed_p1w": -0.04209443647981692, "alpha_spy_p1w": -0.024367006825387838, "alpha_c_p1w": -0.03559558394908224, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.006498852530734678, "ret_p1m": 0.13244360549511813, "ret_signed_p1m": 0.13244360549511813, "alpha_spy_p1m": 0.09159856032579428, "alpha_c_p1m": -0.014997460184375644, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.14744106567949378, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4353, -0.4343, -0.4394, -0.4056, -0.4015, -0.3974, -0.3974, -0.3861, -0.3441, -0.3329, -0.3329, -0.3329, -0.3062, -0.2867, -0.256, -0.2396, -0.2252, -0.2375, -0.2324, -0.2437, -0.2396, -0.2324, -0.2252, -0.2171, -0.2386, -0.2416, -0.2263, -0.214, -0.2457, -0.1802, -0.1381, -0.1176, -0.0685, -0.1494, -0.0705, -0.0777, -0.1371, -0.1402, -0.091, -0.1023, -0.1187, -0.09, -0.0982, -0.0982, -0.0982, -0.0982, -0.0838, -0.0275, -0.0254, 0.0299, 0.0432, 0.1283, 0.0893, 0.0893, -0.0359, -0.1283, -0.0339, -0.0513, -0.1417, -0.037, -0.0195, -0.0452, -0.0657, 0.0, -0.0041, 0.0842, 0.04, 0.0339, -0.0421, 0.0123, 0.0216, -0.0421, -0.0421, -0.1037, -0.1715, -0.0791, -0.1478, -0.1006, -0.0903, -0.0595, 0.0606, 0.0246, 0.0544, 0.0678, 0.1324, 0.1663, 0.1858, 0.1581, 0.1971, 0.2567, 0.2556, 0.2577, 0.2546, 0.3265, 0.3347, 0.3275, 0.3203, 0.3203, 0.4856, 0.4856, 0.6437, 0.6982, 0.731, 0.9302, 0.884, 1.0287, 1.0226, 0.8676, 0.8891, 0.7916, 0.7916, 0.9918, 0.9928, 0.9928, 1.1068, 1.3029, 1.3505, 1.3957, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2033238007709577277", "ticker_idx": 1, "ts": "2026-03-15T17:44:51+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix and Micron look like the obvious beneficiaries", "tweet_type": "analysis", "timeline": "weeks", "summary": "Samsung strike risk could halt memory lines for months; SK Hynix and Micron are near-term beneficiaries.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I feel like I’ve been talking too much about myself, so to shift the conversation back to memory…\n\nIt looks like Samsung’s semiconductor division may genuinely be heading toward a strike. The union is demanding a performance bonus equal to 20% of operating profit, while management is holding firm, saying it cannot remove the cap on bonuses.\n\nThe union seems willing to do whatever it takes to eliminate the bonus gap with SK Hynix. There is a real possibility that Samsung’s memory production lines could actually be halted.\n\nOnce a line is shut down, it can take as long as two months to restart. The losses could amount to tens of billions of dollars.\n\nThat would be good news for Samsung’s competitors. In the near term, SK Hynix and Micron look like the obvious beneficiaries.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03546847335497094, "ret_signed_m1d": -0.03546847335497094, "alpha_spy_m1d": -0.025394132461917907, "alpha_c_m1d": -0.018739120616264038, "bench_spy_m1d": -0.010074340893053035, "bench_c_m1d": -0.016729352738706904, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.045020383429470456, "ret_signed_p1d": 0.045020383429470456, "alpha_spy_p1d": 0.042389828559495246, "alpha_c_p1d": 0.03750618929314964, "bench_spy_p1d": 0.00263055486997521, "bench_c_p1d": 0.0075141941363208176, "ret_p1w": -0.08476684429074977, "ret_signed_p1w": -0.08476684429074977, "alpha_spy_p1w": -0.06703941463632068, "alpha_c_p1w": -0.07826799176001509, "bench_spy_p1w": -0.017727429654429083, "bench_c_p1w": -0.006498852530734678, "ret_p1m": 0.05444914916034849, "ret_signed_p1m": 0.05444914916034849, "alpha_spy_p1m": 0.01360410399102463, "alpha_c_p1m": -0.09299191651914529, "bench_spy_p1m": 0.04084504516932386, "bench_c_p1m": 0.14744106567949378, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4897, -0.4376, -0.3983, -0.3742, -0.3749, -0.3514, -0.3514, -0.3556, -0.3337, -0.3376, -0.354, -0.354, -0.2861, -0.2935, -0.2227, -0.2314, -0.2598, -0.2189, -0.2171, -0.2347, -0.2455, -0.238, -0.1789, -0.1789, -0.1738, -0.1193, -0.1001, -0.0954, -0.1193, -0.0714, -0.0148, -0.0136, -0.0609, -0.0091, -0.0506, -0.1412, -0.1333, -0.1066, -0.132, -0.1552, -0.0712, -0.063, -0.0682, -0.0682, -0.0951, -0.0472, -0.0553, -0.0309, -0.0471, -0.0538, -0.029, -0.0594, -0.0666, -0.0659, -0.1406, -0.0929, -0.1013, -0.1618, -0.1188, -0.0876, -0.0523, -0.0825, -0.0355, 0.0, 0.045, 0.0451, 0.0056, -0.0428, -0.0848, -0.1047, -0.1352, -0.1954, -0.1914, -0.2713, -0.235, -0.167, -0.1707, -0.1707, -0.1446, -0.145, -0.079, -0.0455, -0.0476, -0.0341, 0.0544, 0.0331, 0.0354, 0.0305, 0.0154, 0.0176, 0.1039, 0.0908, 0.1248, 0.1878, 0.1419, 0.174, 0.1711, 0.2278, 0.3053, 0.4497, 0.5094, 0.4642, 0.6911, 0.801, 0.7359, 0.8198, 0.7572, 0.6409, 0.5433, 0.5822, 0.6575, 0.7257, 0.7006, 0.7006, 1.0286, 1.1023, 1.0912, 1.1988, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013754475396345954", "ticker_idx": 0, "ts": "2026-01-20T23:24:15+00:00", "symbol": "ASMPT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics started preparations to expand its TC bonder suppliers… The leading candidate within Samsung is Singapore-based ASMPT", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung diversifying TC/hybrid bonder supply chain to ASMPT and Besi beyond existing suppliers, implying new revenue opportunities for both.", "resolved_tickers": ["0522.HK"], "resolution_method": "exchange_ticker", "resolution_reason": "ASMPT listed on Hong Kong Exchange 0522.HK", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Eyes Diversification of TC Bonder Supply Chain… In Discussions with ASMPT\n\nSamsung Electronics has begun efforts to diversify its supply chain for thermo-compression (TC) bonders used in high-bandwidth memory (HBM) and 3D packaging. Attention is now focused on whether this will lead to changes in the supply of next-generation hybrid bonders.\n\nAccording to industry sources on the 21st, Samsung Electronics started preparations to expand its TC bonder suppliers as early as the fourth quarter of last year.\n\nThe leading candidate within Samsung is Singapore-based ASMPT. The company is better known for supplying equipment to SK Hynix’s HBM lines than to Samsung Electronics.\n\nAn industry source familiar with the matter said, “I understand that Samsung Electronics is actively negotiating with ASMPT to supply a range of bonding equipment it needs, including TC-NCF (non-conductive film) bonders.” The industry sees a strong possibility that cooperation could extend to “fluxless bonders,” which improve process efficiency in TC bonding.\n\nSamsung Electronics is also reportedly showing strong interest in diversifying the supply chain for next-generation hybrid bonders. In addition to exploring cooperation with Netherlands-based BE Semiconductor Industries (Besi)—which has already supplied hybrid bonder prototypes—centered around its subsidiary Semes, Samsung is understood to be holding related discussions with ASMPT as well.\n\nBonders are taking on growing importance in Samsung Electronics’ semiconductor production lines. The currently commercialized TC-NCF bonder bonds multiple semiconductor chips together using heat and pressure, much like an iron.\n\nSamsung used the TC-NCF process to stack and bond multiple DRAM chips in its latest HBM4, which is built with 10nm-class 6th-generation (1c) DRAM. As future HBM stacks grow taller, equipment capable of bonding chips with uniform pressure without causing warpage will become increasingly critical.\n\nTo date, Samsung Electronics has primarily relied on equipment from its subsidiary Semes and Japan’s Shinkawa in the TC-NCF bonder segment. It is known that equipment from both companies is extensively deployed at Samsung’s Cheonan campus, where advanced packaging (AVP) processes are concentrated.\n\nSamsung’s move to broaden ties with foreign bonder makers such as ASMPT and Besi, beyond its existing two suppliers, is interpreted as an effort to secure alternatives to overcome the sharply rising bonding difficulty expected after HBM4. It also reflects a judgment that advanced heterogeneous integration packaging—connecting substrates to dies and dies to dies—is needed not only for HBM but also for system semiconductors.\n\nOne industry source explained, “For both the already-commercialized TC-NCF bonder and the early-stage hybrid bonder, the key challenge for suppliers is how much high-level synergy they can create with Samsung Electronics.”\n\n---\n\nhttps://t.co/9UGzRs7AdH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002049135665716517, "ret_signed_m1d": -0.002049135665716517, "alpha_spy_m1d": -0.02282893123504548, "alpha_c_m1d": -0.027664544935355306, "bench_spy_m1d": 0.020779795569328963, "bench_c_m1d": 0.02561540926963879, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04303287185620008, "ret_signed_p1d": 0.04303287185620008, "alpha_spy_p1d": 0.03149179265410562, "alpha_c_p1d": 0.01347274584289515, "bench_spy_p1d": 0.011541079202094462, "bench_c_p1d": 0.029560126013304933, "ret_p1w": 0.08913940786958419, "ret_signed_p1w": 0.08913940786958419, "alpha_spy_p1w": 0.0627071263761465, "alpha_c_p1w": 0.04595186536354534, "bench_spy_p1w": 0.02643228149343768, "bench_c_p1w": 0.043187542506038845, "ret_p1m": 0.05532792189884206, "ret_signed_p1m": 0.05532792189884206, "alpha_spy_p1m": 0.04247334876404718, "alpha_c_p1m": -0.0014357562512952171, "bench_spy_p1m": 0.012854573134794878, "bench_c_p1m": 0.056763678150137276, "ret_p3m": 0.41495921753436327, "ret_signed_p3m": 0.41495921753436327, "alpha_spy_p3m": 0.3640435789448111, "alpha_c_p3m": 0.22599437149416723, "bench_spy_p3m": 0.05091563858955217, "bench_c_p3m": 0.18896484604019603, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.147, -0.0779, -0.0394, -0.106, -0.106, -0.1409, -0.1598, -0.1537, -0.1634, -0.1644, -0.1568, -0.1685, -0.1562, -0.1711, -0.1875, -0.1793, -0.2018, -0.2013, -0.2213, -0.231, -0.2228, -0.2705, -0.2741, -0.249, -0.2572, -0.2413, -0.2285, -0.2162, -0.2167, -0.2075, -0.2003, -0.1947, -0.2024, -0.1895, -0.21, -0.1921, -0.1931, -0.2295, -0.2341, -0.2218, -0.2269, -0.2382, -0.2075, -0.2065, -0.207, -0.207, -0.207, -0.2198, -0.2085, -0.2039, -0.2039, -0.168, -0.1578, -0.1265, -0.0727, -0.0881, -0.0784, -0.0774, -0.0441, -0.0333, -0.0236, 0.0026, -0.002, 0.0, 0.043, 0.1086, 0.0676, 0.0533, 0.0891, 0.1219, 0.0676, 0.0645, 0.0231, 0.0676, 0.0564, -0.0133, 0.0087, 0.0348, 0.0338, 0.0758, 0.0707, 0.0676, 0.0553, 0.0553, 0.0553, 0.0553, 0.0748, 0.1301, 0.1434, 0.1322, 0.168, 0.1486, 0.1107, 0.0594, 0.1066, 0.1516, 0.1383, 0.1004, 0.1475, 0.1332, 0.1199, 0.1086, 0.1014, 0.0635, 0.1168, 0.0994, 0.0779, 0.0512, 0.0686, 0.1219, 0.0984, 0.0891, 0.0553, 0.0174, 0.0533, 0.0307, 0.0307, 0.0307, 0.0307, 0.1373, 0.1301, 0.1701, 0.2059, 0.2859, 0.2992, 0.3607, 0.415, 0.4201, 0.5174, 0.5594, 0.583, 0.6926, 0.6988, 0.7398, 0.6855, 0.6711, 0.6711, 0.6916, 0.6732, 0.7787, 0.8156, 0.7305, 0.7766, 0.8172, 0.8801, 0.7769, 0.7739, 0.8192, 0.745, 0.845, 0.8038, 0.9038, 0.9038, 1.1101, 1.1018, 1.0595, 1.0007, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013754475396345954", "ticker_idx": 1, "ts": "2026-01-20T23:24:15+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "exploring cooperation with Netherlands-based BE Semiconductor Industries (Besi)—which has already supplied hybrid bonder prototypes", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung diversifying TC/hybrid bonder supply chain to ASMPT and Besi beyond existing suppliers, implying new revenue opportunities for both.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Eyes Diversification of TC Bonder Supply Chain… In Discussions with ASMPT\n\nSamsung Electronics has begun efforts to diversify its supply chain for thermo-compression (TC) bonders used in high-bandwidth memory (HBM) and 3D packaging. Attention is now focused on whether this will lead to changes in the supply of next-generation hybrid bonders.\n\nAccording to industry sources on the 21st, Samsung Electronics started preparations to expand its TC bonder suppliers as early as the fourth quarter of last year.\n\nThe leading candidate within Samsung is Singapore-based ASMPT. The company is better known for supplying equipment to SK Hynix’s HBM lines than to Samsung Electronics.\n\nAn industry source familiar with the matter said, “I understand that Samsung Electronics is actively negotiating with ASMPT to supply a range of bonding equipment it needs, including TC-NCF (non-conductive film) bonders.” The industry sees a strong possibility that cooperation could extend to “fluxless bonders,” which improve process efficiency in TC bonding.\n\nSamsung Electronics is also reportedly showing strong interest in diversifying the supply chain for next-generation hybrid bonders. In addition to exploring cooperation with Netherlands-based BE Semiconductor Industries (Besi)—which has already supplied hybrid bonder prototypes—centered around its subsidiary Semes, Samsung is understood to be holding related discussions with ASMPT as well.\n\nBonders are taking on growing importance in Samsung Electronics’ semiconductor production lines. The currently commercialized TC-NCF bonder bonds multiple semiconductor chips together using heat and pressure, much like an iron.\n\nSamsung used the TC-NCF process to stack and bond multiple DRAM chips in its latest HBM4, which is built with 10nm-class 6th-generation (1c) DRAM. As future HBM stacks grow taller, equipment capable of bonding chips with uniform pressure without causing warpage will become increasingly critical.\n\nTo date, Samsung Electronics has primarily relied on equipment from its subsidiary Semes and Japan’s Shinkawa in the TC-NCF bonder segment. It is known that equipment from both companies is extensively deployed at Samsung’s Cheonan campus, where advanced packaging (AVP) processes are concentrated.\n\nSamsung’s move to broaden ties with foreign bonder makers such as ASMPT and Besi, beyond its existing two suppliers, is interpreted as an effort to secure alternatives to overcome the sharply rising bonding difficulty expected after HBM4. It also reflects a judgment that advanced heterogeneous integration packaging—connecting substrates to dies and dies to dies—is needed not only for HBM but also for system semiconductors.\n\nOne industry source explained, “For both the already-commercialized TC-NCF bonder and the early-stage hybrid bonder, the key challenge for suppliers is how much high-level synergy they can create with Samsung Electronics.”\n\n---\n\nhttps://t.co/9UGzRs7AdH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.023713107403376843, "ret_signed_m1d": -0.023713107403376843, "alpha_spy_m1d": -0.044492902972705806, "alpha_c_m1d": -0.04932851667301563, "bench_spy_m1d": 0.020779795569328963, "bench_c_m1d": 0.02561540926963879, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006651311088466949, "ret_signed_p1d": 0.006651311088466949, "alpha_spy_p1d": -0.004889768113627513, "alpha_c_p1d": -0.022908814924837984, "bench_spy_p1d": 0.011541079202094462, "bench_c_p1d": 0.029560126013304933, "ret_p1w": 0.01648356430346709, "ret_signed_p1w": 0.01648356430346709, "alpha_spy_p1w": -0.00994871718997059, "alpha_c_p1w": -0.026703978202571754, "bench_spy_p1w": 0.02643228149343768, "bench_c_p1w": 0.043187542506038845, "ret_p1m": 0.08155007386201252, "ret_signed_p1m": 0.08155007386201252, "alpha_spy_p1m": 0.06869550072721764, "alpha_c_p1m": 0.024786395711875242, "bench_spy_p1m": 0.012854573134794878, "bench_c_p1m": 0.056763678150137276, "ret_p3m": 0.31174094282279174, "ret_signed_p3m": 0.31174094282279174, "alpha_spy_p3m": 0.2608253042332396, "alpha_c_p3m": 0.12277609678259571, "bench_spy_p3m": 0.05091563858955217, "bench_c_p3m": 0.18896484604019603, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1657, -0.1495, -0.144, -0.1521, -0.146, -0.1495, -0.1463, -0.1567, -0.1755, -0.1816, -0.2033, -0.2241, -0.2108, -0.203, -0.2172, -0.2261, -0.24, -0.247, -0.2646, -0.2533, -0.2478, -0.2898, -0.2805, -0.2794, -0.2504, -0.2487, -0.2487, -0.2496, -0.2403, -0.2137, -0.2047, -0.1929, -0.1507, -0.1712, -0.1877, -0.2082, -0.236, -0.234, -0.2273, -0.2522, -0.2397, -0.2438, -0.2412, -0.2371, -0.2392, -0.2392, -0.2392, -0.236, -0.2302, -0.2264, -0.2264, -0.1377, -0.1108, -0.0694, -0.0815, -0.1261, -0.1261, -0.0622, -0.033, -0.0659, 0.0043, 0.0, -0.0237, 0.0, 0.0067, 0.0121, 0.0124, 0.0113, 0.0165, -0.02, -0.0599, -0.0486, -0.048, -0.0737, -0.0888, -0.0703, -0.0428, -0.0318, -0.0223, -0.0116, -0.0243, 0.0191, 0.0324, 0.0416, 0.0816, 0.0006, 0.0671, 0.0853, 0.1122, 0.1429, 0.0925, 0.0954, 0.0902, 0.0489, 0.1047, 0.0911, -0.096, -0.0382, 0.0098, 0.0119, 0.0136, 0.0706, 0.0584, 0.0899, 0.1128, 0.0633, 0.0324, 0.0558, 0.0648, 0.0729, 0.0723, 0.0113, 0.0075, 0.035, 0.0847, 0.0998, 0.0998, 0.0998, 0.1067, 0.1972, 0.192, 0.2233, 0.1966, 0.2608, 0.2585, 0.2753, 0.3117, 0.3129, 0.3181, 0.3436, 0.3985, 0.4586, 0.4434, 0.3695, 0.3933, 0.4387, 0.4387, 0.4219, 0.4783, 0.4841, 0.4824, 0.5179, 0.5208, 0.4539, 0.5028, 0.5522, 0.5237, 0.4888, 0.4836, 0.5423, 0.5738, 0.5924, 0.6413, 0.6529, 0.6087, 0.6739, 0.6553, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2009165542717051311", "ticker_idx": 0, "ts": "2026-01-08T07:29:28+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "CHINA TO APPROVE SOME NVIDIA H200 PURCHASES AS SOON AS THIS QTR", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "China set to approve Nvidia H200 purchases this quarter; Alibaba and ByteDance each eyeing 200k chips — bullish demand signal for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "It was reported that Alibaba and ByteDance each said they wanted to order 200,000 H200 chips.\n\n$NVDA https://t.co/vX3uGFq1j5", "reply_text": null, "parent_text": null, "quoted_text": "*CHINA TO APPROVE SOME NVIDIA H200 PURCHASES AS SOON AS THIS QTR", "parent_author": null, "ret_m1d": 0.021995237555278324, "ret_signed_m1d": 0.021995237555278324, "alpha_spy_m1d": 0.021893694563360322, "alpha_c_m1d": 0.006057779504062832, "bench_spy_m1d": 0.0001015429919180022, "bench_c_m1d": 0.015937458051215492, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.000972775721840291, "ret_signed_p1d": -0.000972775721840291, "alpha_spy_p1d": -0.007586118115932905, "alpha_c_p1d": -0.027992644679373235, "bench_spy_p1d": 0.006613342394092614, "bench_c_p1d": 0.027019868957532944, "ret_p1w": 0.010862579760640756, "ret_signed_p1w": 0.010862579760640756, "alpha_spy_p1w": 0.006903290689400432, "alpha_c_p1w": -0.0351292725020107, "bench_spy_p1w": 0.003959289071240324, "bench_c_p1w": 0.045991852262651456, "ret_p1m": 0.001999567050479767, "ret_signed_p1m": 0.001999567050479767, "alpha_spy_p1m": 0.00038979108769421345, "alpha_c_p1m": -0.05781884619816546, "bench_spy_p1m": 0.0016097759627855535, "bench_c_p1m": 0.05981841324864523, "ret_p3m": -0.03745326722262887, "ret_signed_p3m": -0.03745326722262887, "alpha_spy_p3m": 0.0038654993069547894, "alpha_c_p3m": -0.09265402268123513, "bench_spy_p3m": -0.04131876652958366, "bench_c_p3m": 0.055200755458606254, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0177, -0.0271, -0.0282, -0.0175, -0.0099, -0.013, -0.021, -0.0258, -0.0156, 0.0065, 0.0348, 0.0864, 0.1188, 0.0964, 0.0942, 0.118, 0.0737, 0.0549, 0.0164, 0.0168, 0.0757, 0.0438, 0.0473, 0.0098, 0.0277, 0.0084, -0.0199, 0.0079, -0.0238, -0.0333, -0.0135, -0.0391, -0.0259, -0.0259, -0.0435, -0.0277, -0.0194, -0.0295, -0.009, -0.0142, 0.0028, -0.0004, -0.0068, -0.0222, -0.0542, -0.0473, -0.0396, -0.0762, -0.0589, -0.0219, -0.0073, 0.0225, 0.0193, 0.0193, 0.0297, 0.0172, 0.0135, 0.0079, 0.0079, 0.0206, 0.0166, 0.0119, 0.022, 0.0, -0.001, -0.0005, 0.0042, -0.0103, 0.0109, 0.0064, 0.0064, -0.0377, -0.0093, -0.0011, 0.0142, 0.0077, 0.0188, 0.035, 0.0404, 0.0329, 0.0031, -0.0254, -0.0586, -0.0711, 0.002, 0.027, 0.0189, 0.0271, 0.0103, -0.0121, -0.0121, -0.0004, 0.0159, 0.0155, 0.0258, 0.0352, 0.0422, 0.0569, -0.0008, -0.0424, -0.0138, -0.027, -0.0108, -0.0092, -0.039, -0.0129, -0.0015, 0.0054, -0.0102, -0.0258, -0.0098, -0.0168, -0.025, -0.035, -0.0666, -0.0507, -0.0531, -0.0343, -0.0745, -0.0946, -0.1073, -0.0575, -0.0502, -0.0413, -0.0413, -0.0399, -0.0375, -0.0159, -0.0061, 0.0195, 0.0231, 0.062, 0.0748, 0.072, 0.09, 0.092, 0.0803, 0.0944, 0.079, 0.1256, 0.1707, 0.1521, 0.1309, 0.0786, 0.0725, 0.0727, 0.062, 0.1232, 0.1431, 0.1631, 0.186, 0.1932, 0.2205, 0.2741, 0.2177, 0.2015, 0.1923, 0.2078, 0.1863, 0.1638, 0.1638, 0.1612, 0.149, 0.1579, 0.1411, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2021483195884241177", "ticker_idx": 0, "ts": "2026-02-11T07:15:25+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM supply shortage is expected to persist, with memory prices projected to surge 90–95% this quarter. Furthermore, the magnitude of price increases next quarter is likely to remain at a similar level.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: DRAM prices up ~90-95% this quarter and again next, implying memory prices ~4x end-of-year levels by June.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwan's Winbond: The DRAM supply shortage is expected to persist, with memory prices projected to surge 90–95% this quarter. Furthermore, the magnitude of price increases next quarter is likely to remain at a similar level.\n\nThis means prices are not just nearly doubling this quarter—memory prices could effectively double again in Q2. This implies that by the end of June, memory prices could reach nearly 4x the levels seen at the end of last year.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09038847502394831, "ret_signed_m1d": -0.09038847502394831, "alpha_spy_m1d": -0.09061958723887886, "alpha_c_m1d": -0.06623111680233484, "bench_spy_m1d": 0.00023111221493055112, "bench_c_m1d": -0.024157358221613467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008846330552533033, "ret_signed_p1d": 0.008846330552533033, "alpha_spy_p1d": 0.024295204780434232, "alpha_c_p1d": 0.029749010173658275, "bench_spy_p1d": -0.0154488742279012, "bench_c_p1d": -0.020902679621125242, "ret_p1w": 0.025856644624870917, "ret_signed_p1w": 0.025856644624870917, "alpha_spy_p1w": 0.034050824170316574, "alpha_c_p1w": 0.031233015213836368, "bench_spy_p1w": -0.008194179545445657, "bench_c_p1w": -0.005376370588965451, "ret_p1m": -0.012160672600280864, "ret_signed_p1m": -0.012160672600280864, "alpha_spy_p1m": 0.025269334804508548, "alpha_c_p1m": 0.05209024107664584, "bench_spy_p1m": -0.03743000740478941, "bench_c_p1m": -0.0642509136769267, "ret_p3m": 0.9390362248077155, "ret_signed_p3m": 0.9390362248077155, "alpha_spy_p3m": 0.8677040986001034, "alpha_c_p3m": 0.5496008632794793, "bench_spy_p3m": 0.07133212620761209, "bench_c_p3m": 0.3894353615282362, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3987, -0.4106, -0.4434, -0.4497, -0.5095, -0.4948, -0.4545, -0.453, -0.4391, -0.4391, -0.4239, -0.4142, -0.4166, -0.4296, -0.4479, -0.4221, -0.3985, -0.3851, -0.3576, -0.3704, -0.4126, -0.4214, -0.4336, -0.4506, -0.3945, -0.3522, -0.3262, -0.327, -0.3016, -0.3016, -0.3062, -0.2826, -0.2869, -0.3045, -0.3045, -0.2313, -0.2393, -0.1631, -0.1725, -0.2031, -0.159, -0.1571, -0.176, -0.1876, -0.1796, -0.116, -0.116, -0.1105, -0.0517, -0.0311, -0.0261, -0.0518, -0.0002, 0.0608, 0.062, 0.0111, 0.0669, 0.0222, -0.0754, -0.0669, -0.0381, -0.0654, -0.0904, 0.0, 0.0088, 0.0032, 0.0032, -0.0257, 0.0259, 0.0171, 0.0435, 0.0259, 0.0187, 0.0455, 0.0127, 0.0049, 0.0057, -0.0747, -0.0233, -0.0324, -0.0976, -0.0512, -0.0176, 0.0203, -0.0122, 0.0385, 0.0767, 0.1251, 0.1252, 0.0827, 0.0306, -0.0146, -0.0361, -0.0688, -0.1337, -0.1295, -0.2154, -0.1763, -0.1032, -0.1071, -0.1071, -0.079, -0.0794, -0.0084, 0.0277, 0.0254, 0.04, 0.1353, 0.1123, 0.1147, 0.1095, 0.0933, 0.0956, 0.1885, 0.1744, 0.211, 0.2789, 0.2295, 0.264, 0.2609, 0.3219, 0.4054, 0.5608, 0.6252, 0.5765, 0.8207, 0.939, 0.8689, 0.9593, 0.8919, 0.7667, 0.6616, 0.7035, 0.7846, 0.858, 0.831, 0.831, 1.1842, 1.2635, 1.2516, 1.3673, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014494793121038840", "ticker_idx": 0, "ts": "2026-01-23T00:26:00+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It doesn't make sense for Intel to be more expensive than TSMC.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "INTC is overvalued relative to TSMC; implied bearish on Intel's relative valuation.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "\"It doesn’t make sense for Intel to be more expensive than TSMC.\"\n\n$INTC https://t.co/shpRuHGHqy", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.20523630047998997, "ret_signed_m1d": -0.20523630047998997, "alpha_spy_m1d": -0.20559894915795618, "alpha_c_m1d": -0.19841280823745322, "bench_spy_m1d": -0.00036264867796620415, "bench_c_m1d": 0.006823492242536755, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05724424304031461, "ret_signed_p1d": 0.05724424304031461, "alpha_spy_p1d": 0.06232239010092444, "alpha_c_p1d": 0.05406998468520363, "bench_spy_p1d": 0.005078147060609828, "bench_c_p1d": -0.0031742583551109815, "ret_p1w": 0.031062825279754236, "ret_signed_p1w": -0.031062825279754236, "alpha_spy_p1w": -0.027087364625359767, "alpha_c_p1w": -0.022639732605521434, "bench_spy_p1w": 0.003975460654394469, "bench_c_p1w": 0.008423092674232802, "ret_p1m": -0.03195026928021816, "ret_signed_p1m": 0.03195026928021816, "alpha_spy_p1m": 0.022026269219125538, "alpha_c_p1m": 0.06391809817369609, "bench_spy_p1m": -0.00992400006109262, "bench_c_p1m": 0.031967828893477934, "ret_p3m": 0.4481916371204664, "ret_signed_p3m": -0.4481916371204664, "alpha_spy_p3m": -0.4134827665775198, "alpha_c_p3m": -0.25638481631316656, "bench_spy_p3m": 0.03470887054294658, "bench_c_p3m": 0.19180682080729983, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0785, -0.0828, -0.1089, -0.1127, -0.1236, -0.1784, -0.1484, -0.1737, -0.154, -0.1469, -0.1595, -0.1593, -0.2032, -0.2119, -0.2299, -0.2383, -0.221, -0.254, -0.2345, -0.2059, -0.205, -0.1833, -0.1833, -0.1001, -0.1123, -0.0355, -0.0291, -0.1014, -0.0812, -0.1058, -0.1014, -0.0952, -0.1234, -0.1611, -0.1677, -0.1722, -0.2001, -0.195, -0.183, -0.193, -0.1935, -0.1977, -0.1977, -0.1968, -0.1862, -0.1724, -0.1813, -0.1813, -0.1262, -0.1265, -0.1116, -0.0541, -0.0879, 0.0107, -0.0224, 0.0493, 0.081, 0.0721, 0.0419, 0.0419, 0.0774, 0.2037, 0.2052, 0.0, -0.0572, -0.0253, 0.0823, 0.0797, 0.0311, 0.083, 0.0927, 0.0783, 0.0703, 0.1225, 0.1147, 0.0457, 0.0714, 0.0313, 0.0382, 0.0382, 0.0246, 0.0087, -0.01, -0.0213, -0.032, 0.0233, 0.0402, 0.0087, 0.012, 0.0095, -0.0437, 0.0113, 0.0195, -0.0366, 0.0113, 0.0379, 0.0646, 0.004, 0.0155, 0.0153, -0.0224, -0.0009, 0.0246, -0.0266, -0.0235, -0.0224, 0.0468, -0.0215, -0.043, -0.0861, -0.0209, 0.0657, 0.1178, 0.1178, 0.1267, 0.174, 0.308, 0.3694, 0.3841, 0.4462, 0.4158, 0.4409, 0.5199, 0.5199, 0.4577, 0.4702, 0.4482, 0.4817, 0.8314, 0.8857, 0.8753, 1.1023, 1.0963, 1.2103, 1.1251, 1.3996, 1.5074, 1.4322, 1.7717, 1.872, 1.6761, 1.669, 1.5722, 1.4134, 1.4, 1.4584, 1.6394, 1.6292, 1.659, 1.659, 1.7406, 1.7018, 1.6823, 1.5445, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015768447242965273", "ticker_idx": 0, "ts": "2026-01-26T12:47:03+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Foundry's performance is expected to shift from a 7 trillion won loss last year to a profit next year, driven by increased capacity utilization from Tesla AI chip supplies", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares analyst thesis that Samsung Foundry swings to profit next year on Tesla AI chips, Qualcomm/AMD orders, and Taylor fab ramp — implied long Samsung Electronics.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Foundry Eyeing 'Turnaround to Profit' Next Year Amid Potential Orders from Qualcomm and AMD for Tesla Chips\n\nSamsung Electronics' Foundry Division, long considered the \"troubled child\" of the Device Solutions (DS) department, is projected to swing back into the black next year. This optimistic outlook is fueled by the mass production of Tesla's next-generation AI chips and the increasing likelihood of securing orders from global tech giants such as Qualcomm and AMD. As yields for advanced processes below 3nm stabilize and the Taylor, Texas fab nears operation, Samsung is aggressively pursuing contracts with global big tech firms.\n\nAccording to industry sources on the 26th, expectations are rising for a turnaround in the foundry business, which has recorded trillions of won in losses since 2022. Kim Dong-won, head of research at KB Securities, stated, \"Samsung Foundry's performance is expected to shift from a 7 trillion won loss last year to a profit next year, driven by increased capacity utilization from Tesla AI chip supplies.\" A source familiar with Samsung Foundry's operations added, \"We are accelerating our order-taking efforts with the goal of turning a profit next year. Beyond the stabilization of sub-3nm process yields, the utilization rates for highly profitable 4-8nm nodes are currently at maximum capacity.\"\n\nThe turnaround projection aligns with the operational timeline of the Taylor fab. Samsung Electronics is investing over $37 billion (approx. 54 trillion KRW) to establish a massive foundry production base in Taylor, Texas, by 2030, with mass production slated to begin in the second half of this year. Final preparations, including the deployment of engineers for pilot production, are currently underway. The primary goal is to manufacture chips for major U.S. tech companies.\n\nProfitability is expected to improve significantly as the Taylor fab begins producing Tesla's AI chips. Samsung Foundry plans to mass-produce the AI5 and AI6 chips next year—the \"brains\" behind Tesla’s Optimus humanoid robots, Robotaxis, and autonomous driving systems. Notably, reports suggest that Samsung Foundry will be the exclusive producer of the AI6 chip.\n\nThe possibility of securing orders from Qualcomm and AMD is also increasing. Qualcomm, a long-standing partner and a key supplier for Samsung’s Mobile eXperience (MX) division, had previously shifted orders away due to low yields in Samsung's advanced processes. However, as Samsung's 2nm process stabilizes, the chances of a return are growing. Additionally, AMD, a long-term customer of TSMC, is reportedly seeking to diversify its foundry partners.\n\nTSMC’s current capacity saturation, despite its 70% market share, is also working in Samsung's favor. As big tech demand concentrated on TSMC for stable sub-3nm nodes, its production lines reached their limit, causing prices to skyrocket and lead times to lengthen. This has prompted customers to seek Samsung Foundry as a viable alternative. Furthermore, utilizing the Taylor fab allows companies to align with the \"Made in USA\" policy of the Trump administration.\n\nIncreased utilization of 4-8nm nodes has also bolstered profitability. These mature nodes offer stable yields compared to cutting-edge processes and are considered competitive against TSMC in terms of performance and pricing. Samsung currently utilizes these nodes to manufacture High Bandwidth Memory (HBM) logic dies, as well as chips for IBM and Nintendo.\n\n---\n\nhttps://t.co/215pZNhRMS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0050524897745126696, "alpha_c_m1d": 0.006777205222007088, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048652228216711224, "ret_signed_p1d": 0.048652228216711224, "alpha_spy_p1d": 0.04466796348702662, "alpha_c_p1d": 0.03516653004304571, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": -0.011176829852671766, "ret_signed_p1w": -0.011176829852671766, "alpha_spy_p1w": -0.015045622732750008, "alpha_c_p1w": -0.005563340661362881, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.31492441086551537, "ret_signed_p1m": 0.31492441086551537, "alpha_spy_p1m": 0.3226907983264917, "alpha_c_p1m": 0.3543549001418518, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": 0.4790577142361261, "ret_signed_p3m": 0.4790577142361261, "alpha_spy_p3m": 0.453571831310269, "alpha_c_p3m": 0.41091867364976475, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3424, -0.3189, -0.2966, -0.2731, -0.3137, -0.3418, -0.351, -0.3595, -0.3418, -0.3228, -0.3254, -0.3274, -0.364, -0.3418, -0.3601, -0.3686, -0.3418, -0.3797, -0.3673, -0.3503, -0.3274, -0.3228, -0.3424, -0.3405, -0.3235, -0.3163, -0.3124, -0.2908, -0.2836, -0.2908, -0.2934, -0.298, -0.2875, -0.3143, -0.3274, -0.294, -0.296, -0.3045, -0.277, -0.2705, -0.2731, -0.2731, -0.2345, -0.2143, -0.2117, -0.2117, -0.2117, -0.1552, -0.092, -0.0868, -0.073, -0.0874, -0.0861, -0.0874, -0.0953, -0.0776, -0.0539, -0.021, -0.0184, -0.0454, -0.0171, 0.0013, 0.0, 0.0, 0.0487, 0.0677, 0.0565, 0.0552, -0.0112, 0.1012, 0.1118, 0.0473, 0.0427, 0.094, 0.0901, 0.1032, 0.1742, 0.1913, 0.1913, 0.1913, 0.1913, 0.2492, 0.2498, 0.2689, 0.3149, 0.3379, 0.4333, 0.4234, 0.4234, 0.2827, 0.1321, 0.2597, 0.2373, 0.1407, 0.2354, 0.2492, 0.2354, 0.2064, 0.2406, 0.2748, 0.3708, 0.3182, 0.311, 0.2249, 0.2472, 0.2426, 0.1841, 0.1841, 0.1615, 0.1016, 0.2495, 0.1753, 0.2267, 0.2722, 0.2946, 0.3868, 0.344, 0.3572, 0.3242, 0.3605, 0.3901, 0.4329, 0.4231, 0.4132, 0.4428, 0.4329, 0.4791, 0.4461, 0.4791, 0.4626, 0.4889, 0.4527, 0.4527, 0.5318, 0.5318, 0.7525, 0.7887, 0.7689, 0.8809, 0.8381, 0.8711, 0.9501, 0.7821, 0.8513, 0.8151, 0.8184, 0.9732, 0.9271, 0.9271, 0.9699, 1.0226, 0.9732, 1.0885, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014907751949074727", "ticker_idx": 0, "ts": "2026-01-24T03:46:57+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs maintains a $320 price target and continues to see the current weakness as a buying opportunity ahead of the earnings report", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Relays Goldman Sachs Buy call on AAPL with $320 PT; recommends buying dip before Jan 29 earnings on +11% revenue growth forecast and AI super-cycle thesis.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": ">> Apple Stock Falls for 8 Consecutive Weeks… Goldman Sachs Recommends ‘Buy the Dip Before January 29’ Despite Memory Cost Pressures\n\n• Apple shares have declined for 8 straight weeks, marking the longest consecutive losing streak since May 2022. Notably, among the ‘Magnificent 7’ stocks since July last year, Apple has experienced the most pronounced fund outflows.\n\n• Market concerns are mounting over margin pressure on Apple products, including the iPhone, as memory prices have surged 40–50% due to explosive demand for AI hardware. This is creating short-term caution around earnings and profitability.\n\n• Nevertheless, Goldman Sachs is taking a contrarian stance. The firm forecasts +11% YoY revenue growth for Apple’s fiscal Q1 results scheduled for release on January 29, and views the full rollout of AI features and replacement demand (super cycle) as key mid- to long-term growth drivers. Goldman Sachs maintains a $320 price target and continues to see the current weakness as a buying opportunity ahead of the earnings report.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028855588226826656, "ret_signed_m1d": -0.028855588226826656, "alpha_spy_m1d": -0.023803098452313987, "alpha_c_m1d": -0.02207838300481957, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": -0.006777205222007088, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011197539980376536, "ret_signed_p1d": 0.011197539980376536, "alpha_spy_p1d": 0.007213275250691931, "alpha_c_p1d": -0.0022881581932889805, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.013485698173665517, "ret_p1w": 0.05716309953805809, "ret_signed_p1w": 0.05716309953805809, "alpha_spy_p1w": 0.05329430665797985, "alpha_c_p1w": 0.06277658872936698, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": -0.005613489191308885, "ret_p1m": 0.0664996200211534, "ret_signed_p1m": 0.0664996200211534, "alpha_spy_p1m": 0.07426600748212975, "alpha_c_p1m": 0.1059301092974898, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": -0.03943048927633641, "ret_p3m": 0.07155495172859094, "ret_signed_p3m": 0.07155495172859094, "alpha_spy_p3m": 0.04606906880273387, "alpha_c_p3m": 0.0034159111422296107, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.06813904058636133, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0549, 0.0616, 0.0575, 0.0524, 0.0563, 0.0566, 0.0552, 0.0501, 0.0549, 0.0777, 0.0707, 0.0687, 0.0666, 0.0472, 0.0471, 0.0515, 0.0424, 0.063, 0.0803, 0.0844, 0.0867, 0.0867, 0.0918, 0.1084, 0.1205, 0.1125, 0.099, 0.0915, 0.088, 0.0852, 0.0915, 0.0886, 0.0895, 0.0732, 0.0752, 0.0643, 0.0657, 0.0715, 0.0609, 0.0664, 0.072, 0.072, 0.0704, 0.0718, 0.0692, 0.0644, 0.0644, 0.0611, 0.0464, 0.0272, 0.0193, 0.0142, 0.0155, 0.0189, 0.0221, 0.0178, 0.011, 0.0005, 0.0005, -0.0341, -0.0304, -0.0276, -0.0289, 0.0, 0.0112, 0.004, 0.0112, 0.0159, 0.0572, 0.0551, 0.0825, 0.0803, 0.0889, 0.0762, 0.0725, 0.0797, 0.0257, 0.0024, 0.0024, 0.0341, 0.036, 0.0212, 0.0369, 0.0431, 0.0665, 0.0747, 0.0697, 0.0353, 0.0374, 0.0336, 0.0288, 0.0201, 0.009, 0.0185, 0.0222, 0.0221, 0.0023, -0.0198, -0.0092, -0.0037, -0.0205, -0.0243, -0.0281, -0.0144, -0.0138, -0.01, -0.0089, -0.025, -0.0335, -0.0054, 0.0018, 0.0029, 0.0029, 0.0145, -0.0065, 0.0146, 0.0208, 0.0208, 0.0158, 0.0143, 0.0441, 0.0322, 0.059, 0.0701, 0.0431, 0.0705, 0.0716, 0.0623, 0.0487, 0.0609, 0.0588, 0.0634, 0.0979, 0.0849, 0.1137, 0.1267, 0.1265, 0.1495, 0.1481, 0.1564, 0.1723, 0.1697, 0.1777, 0.1683, 0.1727, 0.1856, 0.1963, 0.2114, 0.2114, 0.2094, 0.2193, 0.2258, 0.2241, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2036336914412753147", "ticker_idx": 0, "ts": "2026-03-24T06:58:48+00:00", "symbol": "Kinsus", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Kinsus is expected to benefit from new ABF substrate capacity coming online, along with improving revenue and product mix, with double-digit revenue growth projected for 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Kinsus on AI-driven ABF substrate demand, double-digit 2026 revenue growth, and K6 expansion; also bullish Unimicron on accelerating price hikes from 2Q26.", "resolved_tickers": ["3189.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Kinsus Interconnect Technology 3189.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ AI boom lifts substrate maker Kinsus as prices and demand rise\n\n∙ Driven by growing demand for AI GPUs, CPUs, and ASICs, Kinsus is expected to benefit from new ABF substrate capacity coming online, along with improving revenue and product mix, with double-digit revenue growth projected for 2026.\n\n∙ The share of AI-related revenue is expected to expand from around 5–10% currently to 10–15% by the end of 2026, with price hike benefits expected to be reflected more meaningfully in the second half of 2026.\n\n∙ Upstream, shortages of high-end fiberglass cloth continue, leading to longer CCL lead times and rising raw material costs, including precious metals and prepreg, which are increasing pricing pressure for both ABF and BT substrates.\n\n∙ Unimicron is expected to accelerate price hikes starting in 2Q 2026, while Kinsus has indicated ABF price increases of around 5% in 1H26 and up to 7% in 2H26, with profitability likely to improve alongside a better product mix.\n\n∙ The biggest bottleneck remains the shortage of T-glass fiberglass cloth. With supply concentrated around Ibiden and capacity expansion limited, CCL lead times have extended to as long as 24 weeks.\n\n∙ As a result, customers are placing orders earlier and becoming more directly involved in procurement.\n\n∙ Taiwan Glass, Nanya Technology, and Fulltech Fiber Glass are attempting to enter the high-end fiberglass cloth market, but their impact on easing supply constraints is expected to remain limited due to scale and yield challenges.\n\n∙ There is growing discussion across the supply chain that fiberglass material shortages could persist through 2027, potentially capping upside for ABF and BT substrate market growth.\n\n∙ Kinsus’s current revenue mix is approximately 30% ABF substrates, 40% BT substrates, and 20% contact lenses, and all business segments are expected to grow in 2026.\n\n∙ Over the next three years, the company plans to pursue an ABF substrate-focused expansion strategy, with the K6 plant in Yangmei, Taoyuan set to serve as a key production hub.\n\n∙ Phase 1 is already running at full utilization and is expected to drive growth in 2026.\n\n∙ Phase 2 is scheduled to begin operations in 2027 and is expected to raise monthly capacity by around 25% to roughly 5,000 units.\n\n∙ In addition, Kinsus has approved NT$23.5 billion in capex, with investment set to rise from NT$7–8 billion in 2026 to around NT$10 billion in 2027.\n\n∙ The Yangmei site also has space secured for a Phase 3 expansion, leaving room for additional investment after 2028 depending on AI chip demand.\n\n---\n\nhttps://t.co/W1vf9iMJHp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0795631825273011, "ret_signed_m1d": 0.0795631825273011, "alpha_spy_m1d": 0.07619502638090903, "alpha_c_m1d": 0.08774892219709807, "bench_spy_m1d": 0.003368156146392076, "bench_c_m1d": -0.008185739669796965, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09672386895475826, "ret_signed_p1d": 0.09672386895475826, "alpha_spy_p1d": 0.0911511100073723, "alpha_c_p1d": 0.0854970443598746, "bench_spy_p1d": 0.0055727589473859585, "bench_c_p1d": 0.011226824594883666, "ret_p1w": -0.023400936037441533, "ret_signed_p1w": -0.023400936037441533, "alpha_spy_p1w": -0.01905302910189921, "alpha_c_p1w": 0.0049576197933154376, "bench_spy_p1w": -0.004347906935542323, "bench_c_p1w": -0.02835855583075697, "ret_p1m": 0.5335413416536661, "ret_signed_p1m": 0.5335413416536661, "alpha_spy_p1m": 0.4446990164943292, "alpha_c_p1m": 0.32512249931699366, "bench_spy_p1m": 0.08884232515933688, "bench_c_p1m": 0.2084188423366724, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5507, -0.5226, -0.5304, -0.5382, -0.5039, -0.5039, -0.5133, -0.4758, -0.4711, -0.454, -0.4821, -0.4743, -0.4353, -0.4493, -0.4633, -0.454, -0.3994, -0.362, -0.3771, -0.3967, -0.3801, -0.3545, -0.351, -0.2871, -0.2402, -0.2839, -0.2137, -0.2527, -0.2324, -0.156, -0.1778, -0.2434, -0.2044, -0.209, -0.1716, -0.1716, -0.1716, -0.1716, -0.1716, -0.1716, -0.1716, -0.1716, -0.0889, 0.0016, 0.0156, -0.0187, -0.0187, -0.0499, -0.0359, -0.1123, -0.0718, -0.0905, -0.1747, -0.1451, -0.0936, -0.0983, -0.0094, 0.0515, 0.0655, 0.1404, 0.1966, 0.1856, 0.0796, 0.0, 0.0967, 0.0967, 0.0983, 0.0811, -0.0234, 0.0359, 0.0484, 0.0484, 0.0484, 0.1186, 0.2293, 0.2075, 0.1888, 0.1435, 0.1576, 0.1966, 0.2278, 0.2215, 0.3432, 0.4774, 0.5335, 0.4945, 0.6412, 0.6069, 0.585, 0.585, 0.6474, 0.6474, 0.6568, 0.6443, 0.5055, 0.5367, 0.429, 0.5694, 0.5413, 0.5523, 0.6287, 0.5164, 0.5757, 0.6006, 0.5725, 0.7285, 0.9002, 1.0874, 1.1716, 1.1903, 1.1248, 1.2746, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2036336914412753147", "ticker_idx": 1, "ts": "2026-03-24T06:58:48+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron is expected to accelerate price hikes starting in 2Q 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Kinsus on AI-driven ABF substrate demand, double-digit 2026 revenue growth, and K6 expansion; also bullish Unimicron on accelerating price hikes from 2Q26.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ AI boom lifts substrate maker Kinsus as prices and demand rise\n\n∙ Driven by growing demand for AI GPUs, CPUs, and ASICs, Kinsus is expected to benefit from new ABF substrate capacity coming online, along with improving revenue and product mix, with double-digit revenue growth projected for 2026.\n\n∙ The share of AI-related revenue is expected to expand from around 5–10% currently to 10–15% by the end of 2026, with price hike benefits expected to be reflected more meaningfully in the second half of 2026.\n\n∙ Upstream, shortages of high-end fiberglass cloth continue, leading to longer CCL lead times and rising raw material costs, including precious metals and prepreg, which are increasing pricing pressure for both ABF and BT substrates.\n\n∙ Unimicron is expected to accelerate price hikes starting in 2Q 2026, while Kinsus has indicated ABF price increases of around 5% in 1H26 and up to 7% in 2H26, with profitability likely to improve alongside a better product mix.\n\n∙ The biggest bottleneck remains the shortage of T-glass fiberglass cloth. With supply concentrated around Ibiden and capacity expansion limited, CCL lead times have extended to as long as 24 weeks.\n\n∙ As a result, customers are placing orders earlier and becoming more directly involved in procurement.\n\n∙ Taiwan Glass, Nanya Technology, and Fulltech Fiber Glass are attempting to enter the high-end fiberglass cloth market, but their impact on easing supply constraints is expected to remain limited due to scale and yield challenges.\n\n∙ There is growing discussion across the supply chain that fiberglass material shortages could persist through 2027, potentially capping upside for ABF and BT substrate market growth.\n\n∙ Kinsus’s current revenue mix is approximately 30% ABF substrates, 40% BT substrates, and 20% contact lenses, and all business segments are expected to grow in 2026.\n\n∙ Over the next three years, the company plans to pursue an ABF substrate-focused expansion strategy, with the K6 plant in Yangmei, Taoyuan set to serve as a key production hub.\n\n∙ Phase 1 is already running at full utilization and is expected to drive growth in 2026.\n\n∙ Phase 2 is scheduled to begin operations in 2027 and is expected to raise monthly capacity by around 25% to roughly 5,000 units.\n\n∙ In addition, Kinsus has approved NT$23.5 billion in capex, with investment set to rise from NT$7–8 billion in 2026 to around NT$10 billion in 2027.\n\n∙ The Yangmei site also has space secured for a Phase 3 expansion, leaving room for additional investment after 2028 depending on AI chip demand.\n\n---\n\nhttps://t.co/W1vf9iMJHp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08478260869565224, "ret_signed_m1d": 0.08478260869565224, "alpha_spy_m1d": 0.08141445254926016, "alpha_c_m1d": 0.07890671381778636, "bench_spy_m1d": 0.003368156146392076, "bench_c_m1d": 0.005875894877865884, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10000000000000009, "ret_signed_p1d": 0.10000000000000009, "alpha_spy_p1d": 0.09442724105261413, "alpha_c_p1d": 0.0955196427638818, "bench_spy_p1d": 0.0055727589473859585, "bench_c_p1d": 0.004480357236118282, "ret_p1w": -0.033695652173913015, "ret_signed_p1w": -0.033695652173913015, "alpha_spy_p1w": -0.029347745238370693, "alpha_c_p1w": -0.009824920292198835, "bench_spy_p1w": -0.004347906935542323, "bench_c_p1w": -0.02387073188171418, "ret_p1m": 0.5565217391304347, "ret_signed_p1m": 0.5565217391304347, "alpha_spy_p1m": 0.4676794139710978, "alpha_c_p1m": 0.3953759188956769, "bench_spy_p1m": 0.08884232515933688, "bench_c_p1m": 0.16114582023475776, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5283, -0.5207, -0.5207, -0.5228, -0.5217, -0.5217, -0.5239, -0.5, -0.4989, -0.5087, -0.5054, -0.5098, -0.4707, -0.4478, -0.4565, -0.4576, -0.4043, -0.3457, -0.3304, -0.3022, -0.2891, -0.263, -0.2761, -0.2304, -0.2011, -0.2511, -0.1772, -0.2065, -0.163, -0.1565, -0.2054, -0.2609, -0.2446, -0.1924, -0.1978, -0.1978, -0.1978, -0.1978, -0.1978, -0.1978, -0.1978, -0.1978, -0.1185, -0.0304, -0.0033, 0.0467, 0.0467, 0.0457, 0.0065, -0.0935, -0.0217, -0.0609, -0.1543, -0.0783, 0.013, 0.0435, 0.1043, 0.1935, 0.1804, 0.2609, 0.263, 0.2043, 0.0848, 0.0, 0.1, 0.0913, 0.0978, 0.0728, -0.0337, 0.062, 0.1283, 0.1283, 0.1283, 0.2261, 0.3478, 0.3587, 0.387, 0.3609, 0.2978, 0.3435, 0.4022, 0.3978, 0.4739, 0.5435, 0.5565, 0.5826, 0.7174, 0.8239, 0.7935, 0.7457, 0.9196, 0.9196, 0.9804, 0.963, 0.8652, 0.9478, 0.7783, 0.8717, 0.9022, 0.9391, 0.9152, 0.7848, 0.7739, 0.7783, 0.7891, 0.9674, 1.1087, 1.1522, 1.3587, 1.3478, 1.2283, 1.2935, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2012050002806505647", "ticker_idx": 0, "ts": "2026-01-16T06:31:17+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bullish $ASML", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Endorses bullish stance on ASML citing TSMC's rising EUV tool demand (25/36 tools in 2026/2027) driving capex cycle.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Based on our estimates of TSMC’s EUV demand of 25/36 tools in 2026/2027, we expect capex to reach US$54 billion in 2026 and US$64 billion in 2027, maintaining a normal level of capital intensity.\"\n\nBullish $ASML", "reply_text": null, "parent_text": null, "quoted_text": "GF Overseas Electronics Communication \n\nTSMC (2330TT Buy): Strong Capex in an Unprecedented AI Cycle\n\nRaising Target Price to NT$2,325  \nTSMC reported strong 4Q25 results, with full-year revenue guidance in line with expectations, while the capital expenditure guidance of", "parent_author": null, "ret_m1d": -0.015250183147876406, "ret_signed_m1d": -0.015250183147876406, "alpha_spy_m1d": -0.01608875976030477, "alpha_c_m1d": -0.00530984787297506, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04009595059520432, "ret_signed_p1d": -0.04009595059520432, "alpha_spy_p1d": -0.04009595059520432, "alpha_c_p1d": -0.04009595059520432, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.009424191472970689, "ret_signed_p1w": 0.009424191472970689, "alpha_spy_p1w": 0.012937506276860855, "alpha_c_p1w": 0.010173506639609697, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.02519170177438279, "ret_signed_p1m": 0.02519170177438279, "alpha_spy_p1m": 0.039519531342202985, "alpha_c_p1m": 0.006884585452559211, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.055218192619088, "ret_signed_p3m": 0.055218192619088, "alpha_spy_p3m": 0.04048324149446891, "alpha_c_p3m": -0.0761786540224465, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.239, -0.2559, -0.2408, -0.2351, -0.2234, -0.2226, -0.2123, -0.1968, -0.2134, -0.2062, -0.2135, -0.2233, -0.233, -0.2539, -0.2403, -0.2401, -0.2359, -0.2448, -0.2511, -0.2491, -0.2584, -0.2404, -0.2377, -0.2855, -0.2642, -0.2629, -0.221, -0.2313, -0.226, -0.206, -0.1954, -0.1746, -0.1798, -0.1847, -0.1748, -0.1836, -0.1895, -0.1939, -0.2072, -0.2024, -0.2216, -0.2512, -0.2352, -0.2276, -0.2324, -0.2262, -0.2298, -0.2298, -0.2298, -0.2231, -0.2132, -0.2106, -0.2106, -0.155, -0.0977, -0.091, -0.0989, -0.1321, -0.0733, -0.0692, -0.0553, -0.0711, -0.0153, 0.0, -0.0401, -0.0233, -0.0106, 0.0079, 0.0094, 0.0093, 0.0432, 0.0233, 0.0212, 0.0415, 0.0493, 0.0199, -0.0228, -0.0151, 0.0228, 0.0336, 0.0235, 0.0362, 0.0122, 0.0212, 0.0252, 0.0288, 0.0679, 0.0623, 0.0772, 0.0717, 0.0839, 0.1053, 0.0573, 0.0581, 0.0384, -0.0033, 0.0293, 0.0175, -0.016, -0.0155, 0.0295, 0.0285, 0.0127, 0.012, 0.0273, 0.0242, 0.0218, 0.0025, -0.0321, 0.0084, 0.0309, 0.0394, 0.0001, -0.0165, -0.046, -0.0398, 0.0188, -0.004, -0.004, -0.004, -0.0445, 0.0403, 0.0593, 0.0895, 0.0808, 0.1017, 0.0552, 0.0489, 0.0672, 0.0681, 0.0612, 0.0722, 0.0499, 0.0744, 0.0426, 0.0075, 0.0254, 0.051, 0.051, 0.0209, 0.0567, 0.122, 0.1177, 0.1344, 0.1226, 0.0883, 0.141, 0.1753, 0.1234, 0.0871, 0.0739, 0.1461, 0.1566, 0.2115, 0.23, 0.1949, 0.1834, 0.196, 0.1906, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2036069841866752430", "ticker_idx": 0, "ts": "2026-03-23T13:17:33+00:00", "symbol": "EO Technics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's femtosecond laser dicing supply chain is expected to be handled by domestic firm EO Technics and Japan's Disco", "tweet_type": "analysis", "timeline": "months", "summary": "EO Technics and Disco named as sole beneficiaries of Samsung's femtosecond laser dicing ramp for HBM4, with demand set to intensify.", "resolved_tickers": ["039030.KQ"], "resolution_method": "exchange_ticker", "resolution_reason": "EO Technics 039030.KQ Korea KOSDAQ laser equipment", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics to Enhance HBM4 Quality with Next-Generation Wafer Dicing Process\n\nSamsung Electronics is transitioning its wafer dicing process — a critical step directly tied to semiconductor quality and yield — to a next-generation approach. The company plans to expand its application of \"femtosecond (one-quadrillionth of a second) laser\" technology, which was first introduced last year targeting High Bandwidth Memory (HBM4), in a move to assert dominance in the HBM4 market through dramatically improved cutting precision.\n\nAccording to industry sources on the 23rd, Samsung Electronics is placing orders for wafer dicing equipment equipped with femtosecond laser technology. The equipment covers both \"grooving\" — which scores the wafer surface — and \"full-cut\" tools, with the initial procurement volume standing at a minimum of 10 units. The equipment is slated for installation at Samsung's Cheonan campus, where advanced semiconductor packaging operations are conducted, and purchase orders (POs) are currently being prepared.\n\nSamsung is also reviewing the option of placing additional orders to secure as large a volume as possible. Given that lead times for wafer dicing equipment exceed eight months, the intent is to secure supply proactively while substantially expanding the scope of femtosecond laser-based dicing processes. A well-placed industry source noted, \"Even after the initial introduction, Samsung Electronics is in ongoing discussions with partners to continue scaling up femtosecond grooving and full-cut equipment,\" adding that it represents \"an attempt to convert the majority of wafer dicing processes to the next-generation approach.\"\n\nIn semiconductor manufacturing, a dicing process — which separates wafers into individual chip units (dies) — is essential following the front-end process where circuits are etched onto wafers. Conventional dicing has largely relied on mechanical cutting via diamond wheel blades. While lasers have been used in some cases, the laser pulse duration — the key measure of precision — has remained at the nanosecond (one-billionth of a second) level.\n\nFemtosecond lasers generate pulses at the one-quadrillionth-of-a-second scale, enabling far finer cuts compared to both conventional mechanical methods and nanosecond lasers. The technology can cut without affecting the ultra-fine circuitry and interconnects of advanced semiconductors, while also minimizing particle contamination generated during the dicing process — ultimately maximizing the quality of the finished semiconductor.\n\nSamsung Electronics first introduced femtosecond laser dicing in Q2 of last year. At the time, only a handful of units were deployed, but having demonstrated improvements in performance and productivity, the company has decided to scale up adoption. Samsung plans to apply femtosecond laser dicing to HBM4 processes as a priority, with the goal of improving HBM4 yield and productivity.\n\nSamsung recently commenced mass production shipments of HBM4. Demand for femtosecond laser dicing is expected to grow in tandem with the ramp-up schedule for full-scale mass production. The company is also reportedly evaluating the introduction of femtosecond laser dicing for products beyond DRAM, including NAND flash and system semiconductors.\n\nSamsung's femtosecond laser dicing supply chain is expected to be handled by domestic firm EO Technics and Japan's Disco. An industry source commented, \"Competition between EO Technics and Disco for femtosecond laser dicing orders is set to intensify,\" adding that \"demand for femtosecond laser dicing is rising not only at Samsung Electronics but also at other domestic and international semiconductor manufacturers.\"\n\n---\n\nhttps://t.co/pZmxwDOXv0", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02505966587112174, "ret_signed_m1d": 0.02505966587112174, "alpha_spy_m1d": 0.03545058031443227, "alpha_c_m1d": 0.0419750260637094, "bench_spy_m1d": -0.010390914443310528, "bench_c_m1d": -0.01691536019258766, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010739856801909253, "ret_signed_p1d": 0.010739856801909253, "alpha_spy_p1d": 0.01409670655417572, "alpha_c_p1d": 0.0024865577748598433, "bench_spy_p1d": -0.0033568497522664664, "bench_c_p1d": 0.00825329902704941, "ret_p1w": -0.07875894988066823, "ret_signed_p1w": -0.07875894988066823, "alpha_spy_p1w": -0.04303916287881515, "alpha_c_p1w": -0.005092791736351687, "bench_spy_p1w": -0.035719787001853076, "bench_c_p1w": -0.07366615814431654, "ret_p1m": 0.1527446300715991, "ret_signed_p1m": 0.1527446300715991, "alpha_spy_p1m": 0.07843657696622675, "alpha_c_p1m": -0.034551008353726154, "bench_spy_p1m": 0.07430805310537236, "bench_c_p1m": 0.18729563842532526, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3942, -0.3942, -0.3823, -0.3628, -0.3508, -0.3508, -0.3508, -0.3103, -0.2434, -0.2351, -0.2661, -0.2589, -0.2613, -0.2589, -0.2649, -0.2506, -0.2804, -0.284, -0.2864, -0.3091, -0.3019, -0.2852, -0.2852, -0.2291, -0.2017, -0.1647, -0.1396, -0.1026, -0.1539, -0.0549, -0.0668, -0.0847, -0.0561, -0.0024, -0.0668, -0.1038, -0.0967, -0.1289, -0.1289, -0.1289, -0.1289, -0.0979, -0.111, -0.1217, -0.0776, -0.1074, 0.0048, -0.0012, -0.0012, 0.006, -0.1372, 0.0644, 0.0811, -0.0143, 0.0, -0.0489, -0.0203, -0.0203, -0.0215, -0.0585, -0.0203, -0.0155, 0.0251, 0.0, 0.0107, 0.0656, -0.0263, -0.0048, -0.0788, -0.0823, 0.0346, -0.0513, -0.0549, -0.0573, -0.0167, 0.0549, 0.0179, 0.0692, 0.1695, 0.1492, 0.1265, 0.1277, 0.1229, 0.1217, 0.1527, 0.1527, 0.1563, 0.2076, 0.2053, 0.2243, 0.2315, 0.1169, 0.1169, 0.1337, 0.1337, 0.1432, 0.1062, 0.142, 0.1957, 0.0215, 0.1444, 0.1432, 0.0549, 0.1575, 0.1086, 0.1026, 0.3103, 0.3675, 0.3675, 0.3055, 0.2291, 0.1563, 0.0931, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2036069841866752430", "ticker_idx": 1, "ts": "2026-03-23T13:17:33+00:00", "symbol": "Disco", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's femtosecond laser dicing supply chain is expected to be handled by domestic firm EO Technics and Japan's Disco", "tweet_type": "analysis", "timeline": "months", "summary": "EO Technics and Disco named as sole beneficiaries of Samsung's femtosecond laser dicing ramp for HBM4, with demand set to intensify.", "resolved_tickers": ["6146.T"], "resolution_method": "exchange_ticker", "resolution_reason": "DISCO Corp 6146.T Japan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics to Enhance HBM4 Quality with Next-Generation Wafer Dicing Process\n\nSamsung Electronics is transitioning its wafer dicing process — a critical step directly tied to semiconductor quality and yield — to a next-generation approach. The company plans to expand its application of \"femtosecond (one-quadrillionth of a second) laser\" technology, which was first introduced last year targeting High Bandwidth Memory (HBM4), in a move to assert dominance in the HBM4 market through dramatically improved cutting precision.\n\nAccording to industry sources on the 23rd, Samsung Electronics is placing orders for wafer dicing equipment equipped with femtosecond laser technology. The equipment covers both \"grooving\" — which scores the wafer surface — and \"full-cut\" tools, with the initial procurement volume standing at a minimum of 10 units. The equipment is slated for installation at Samsung's Cheonan campus, where advanced semiconductor packaging operations are conducted, and purchase orders (POs) are currently being prepared.\n\nSamsung is also reviewing the option of placing additional orders to secure as large a volume as possible. Given that lead times for wafer dicing equipment exceed eight months, the intent is to secure supply proactively while substantially expanding the scope of femtosecond laser-based dicing processes. A well-placed industry source noted, \"Even after the initial introduction, Samsung Electronics is in ongoing discussions with partners to continue scaling up femtosecond grooving and full-cut equipment,\" adding that it represents \"an attempt to convert the majority of wafer dicing processes to the next-generation approach.\"\n\nIn semiconductor manufacturing, a dicing process — which separates wafers into individual chip units (dies) — is essential following the front-end process where circuits are etched onto wafers. Conventional dicing has largely relied on mechanical cutting via diamond wheel blades. While lasers have been used in some cases, the laser pulse duration — the key measure of precision — has remained at the nanosecond (one-billionth of a second) level.\n\nFemtosecond lasers generate pulses at the one-quadrillionth-of-a-second scale, enabling far finer cuts compared to both conventional mechanical methods and nanosecond lasers. The technology can cut without affecting the ultra-fine circuitry and interconnects of advanced semiconductors, while also minimizing particle contamination generated during the dicing process — ultimately maximizing the quality of the finished semiconductor.\n\nSamsung Electronics first introduced femtosecond laser dicing in Q2 of last year. At the time, only a handful of units were deployed, but having demonstrated improvements in performance and productivity, the company has decided to scale up adoption. Samsung plans to apply femtosecond laser dicing to HBM4 processes as a priority, with the goal of improving HBM4 yield and productivity.\n\nSamsung recently commenced mass production shipments of HBM4. Demand for femtosecond laser dicing is expected to grow in tandem with the ramp-up schedule for full-scale mass production. The company is also reportedly evaluating the introduction of femtosecond laser dicing for products beyond DRAM, including NAND flash and system semiconductors.\n\nSamsung's femtosecond laser dicing supply chain is expected to be handled by domestic firm EO Technics and Japan's Disco. An industry source commented, \"Competition between EO Technics and Disco for femtosecond laser dicing orders is set to intensify,\" adding that \"demand for femtosecond laser dicing is rising not only at Samsung Electronics but also at other domestic and international semiconductor manufacturers.\"\n\n---\n\nhttps://t.co/pZmxwDOXv0", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06554372695079991, "ret_signed_m1d": 0.06554372695079991, "alpha_spy_m1d": 0.07593464139411044, "alpha_c_m1d": 0.08245908714338757, "bench_spy_m1d": -0.010390914443310528, "bench_c_m1d": -0.01691536019258766, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0013782670448649892, "ret_signed_p1d": 0.0013782670448649892, "alpha_spy_p1d": 0.004735116797131456, "alpha_c_p1d": -0.006875031982184421, "bench_spy_p1d": -0.0033568497522664664, "bench_c_p1d": 0.00825329902704941, "ret_p1w": -0.010487282674028209, "ret_signed_p1w": -0.010487282674028209, "alpha_spy_p1w": 0.025232504327824867, "alpha_c_p1w": 0.06317887547028833, "bench_spy_p1w": -0.035719787001853076, "bench_c_p1w": -0.07366615814431654, "ret_p1m": 0.1433684690160335, "ret_signed_p1m": 0.1433684690160335, "alpha_spy_p1m": 0.06906041591066114, "alpha_c_p1m": -0.043927169409291755, "bench_spy_p1m": 0.07430805310537236, "bench_c_p1m": 0.18729563842532526, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2758, -0.2692, -0.2516, -0.2688, -0.2623, -0.2623, -0.2623, -0.2623, -0.2175, -0.17, -0.1475, -0.1487, -0.1577, -0.1577, -0.1205, -0.0812, -0.0949, -0.072, -0.0441, -0.0812, -0.1031, 0.0501, 0.0406, 0.0184, 0.0556, 0.0793, 0.0312, 0.0136, -0.0459, 0.0248, 0.0435, -0.002, 0.0077, 0.1061, 0.1542, 0.1542, 0.1153, 0.1188, 0.1066, 0.0983, 0.1, 0.1343, 0.1342, 0.1342, 0.158, 0.2251, 0.2109, 0.1562, 0.1472, 0.1184, 0.056, 0.0974, 0.1257, 0.0184, 0.0688, 0.1083, 0.091, 0.066, 0.0708, 0.0314, 0.0824, 0.0655, 0.0655, 0.0, 0.0014, 0.041, 0.0335, 0.0188, -0.0105, -0.0568, -0.0193, -0.021, 0.0111, 0.0316, -0.0319, 0.0388, -0.0051, 0.0302, 0.0376, 0.1032, 0.087, 0.1247, 0.1227, 0.1235, 0.1434, 0.1525, 0.1089, 0.1042, 0.1709, 0.1773, 0.1773, 0.1398, 0.1127, 0.1127, 0.1127, 0.1127, 0.1796, 0.1551, 0.1301, 0.1212, 0.1218, 0.0781, -0.0005, -0.022, -0.0513, -0.0553, 0.0111, 0.0143, 0.0747, 0.0305, 0.0416, 0.0219, 0.0024, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007303808259645854", "ticker_idx": 0, "ts": "2026-01-03T04:11:36+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm going to buy more Intel with this money", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author plans to buy more Intel with received funds; bullish on INTC.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thank you, @elonmusk! I’m going to buy more Intel with this money. https://t.co/SV97sJc9O6", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0002540547752560851, "ret_signed_m1d": 0.0002540547752560851, "alpha_spy_m1d": 0.006870104944181388, "alpha_c_m1d": 0.011615607217135171, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01701808488239398, "ret_signed_p1d": 0.01701808488239398, "alpha_spy_p1d": 0.011070901509304676, "alpha_c_p1d": -0.009545115617021205, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.11912630349623754, "ret_signed_p1w": 0.11912630349623754, "alpha_spy_p1w": 0.10830794372563846, "alpha_c_p1w": 0.0845120196143021, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.2509525358435727, "ret_signed_p1m": 0.2509525358435727, "alpha_spy_p1m": 0.2483205859384563, "alpha_c_p1m": 0.19774668682019736, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.2796546211565616, "ret_signed_p3m": 0.2796546211565616, "alpha_spy_p3m": 0.3234208110969331, "alpha_c_p3m": 0.24064402466515777, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0493, -0.0399, -0.0762, -0.0546, -0.095, -0.0564, -0.0643, -0.0599, -0.0323, -0.0318, -0.0622, -0.0307, -0.0277, 0.0043, 0.0549, 0.05, 0.0201, 0.0157, 0.0033, -0.0594, -0.0251, -0.0541, -0.0315, -0.0234, -0.0378, -0.0376, -0.0879, -0.0978, -0.1184, -0.128, -0.1082, -0.1461, -0.1237, -0.0909, -0.0899, -0.065, -0.065, 0.0302, 0.0163, 0.1041, 0.1115, 0.0287, 0.0518, 0.0236, 0.0287, 0.0358, 0.0036, -0.0396, -0.0472, -0.0523, -0.0843, -0.0785, -0.0648, -0.0762, -0.0767, -0.0815, -0.0815, -0.0805, -0.0683, -0.0526, -0.0627, -0.0627, 0.0003, 0.0, 0.017, 0.0828, 0.0442, 0.157, 0.1191, 0.2012, 0.2375, 0.2273, 0.1928, 0.1928, 0.2334, 0.378, 0.3797, 0.1448, 0.0792, 0.1158, 0.239, 0.236, 0.1803, 0.2398, 0.251, 0.2344, 0.2253, 0.285, 0.2761, 0.1971, 0.2266, 0.1806, 0.1885, 0.1885, 0.173, 0.1547, 0.1334, 0.1204, 0.1082, 0.1715, 0.1908, 0.1547, 0.1585, 0.1557, 0.0947, 0.1577, 0.1671, 0.1029, 0.1577, 0.1882, 0.2187, 0.1494, 0.1626, 0.1623, 0.1191, 0.1438, 0.173, 0.1143, 0.1179, 0.1191, 0.1984, 0.1201, 0.0955, 0.0462, 0.1209, 0.22, 0.2797, 0.2797, 0.2898, 0.3439, 0.4973, 0.5677, 0.5845, 0.6556, 0.6208, 0.6495, 0.7399, 0.7399, 0.6688, 0.683, 0.6579, 0.6962, 1.0965, 1.1588, 1.1468, 1.4067, 1.3998, 1.5304, 1.4328, 1.747, 1.8705, 1.7844, 2.173, 2.2878, 2.0635, 2.0554, 1.9446, 1.7628, 1.7475, 1.8143, 2.0216, 2.0099, 2.0439, 2.0439, 2.1374, 2.093, 2.0706, 1.9129, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2020822218973094226", "ticker_idx": 0, "ts": "2026-02-09T11:28:56+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory suppliers are leading the LTAs. NCNR contracts with no price negotiation are expanding to PCs, mobile, and commodity DRAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply shortage flips LTA power to memory chipmakers; shorter contracts + NCNR expansion locks in volume and pricing power for Samsung, Hynix, and MU.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Supply Shortage Reshapes LTA Dynamics… Contract Periods Shrink and Power Balance Flips\n\nAs the supply shortage drags on, the trading order in the memory semiconductor market is rapidly being reorganized. With the semiconductor supply crunch persisting over an extended period, the initiative in long-term supply agreements (LTAs) is shifting from customers to semiconductor suppliers. In the past, smartphone and PC manufacturers seeking stable procurement were the ones to propose long-term contracts first, but recently memory chipmakers such as Samsung Electronics and SK hynix have been presenting contract terms first, signaling a shift in the market's balance of power.\n\nAccording to industry sources on the 9th, the contract periods of LTAs recently signed by Korean memory chipmakers are getting shorter. LTA durations vary from six months to one year depending on the product and end customer, but industry insiders universally say these periods have been shrinking lately.\n\nThe analysis is that as the semiconductor supply shortage continues, suppliers are opting for shorter contract periods. In a rising-price environment, shorter contract durations allow suppliers to reflect updated pricing more quickly, giving them an advantage in negotiations.\n\nThe initiative in LTAs is also shifting from customers to memory suppliers. Previously, it was common for smartphone and PC makers to propose long-term supply agreements first to secure volume, but recently Samsung Electronics and SK hynix are increasingly taking the lead in setting contract terms and structures. As the semiconductor supply shortage extends, the center of gravity in business relationships is tilting toward suppliers.\n\nLee Jong-wook, a researcher at Samsung Securities, analyzed: \"It was customers who first proposed long-term supply agreements, but now memory suppliers are leading the LTAs. NCNR (non-cancellable, non-returnable) contracts with no price negotiation are expanding to PCs, mobile, and commodity DRAM, causing the volume of available spot supply in the market to shrink dramatically.\"\n\nStockpiling behavior is also emerging. According to market research firm TrendForce, smartphone manufacturers are being extremely cautious about immediately reducing their LTA-based memory purchase volumes this year. TrendForce noted: \"Concerned about further price increases and potential supply shortages, they are preemptively prioritizing resource procurement to avoid future allocation restrictions.\" In other words, despite slowing global smartphone demand and a negative market outlook, companies are stocking up on semiconductors out of fear of supply constraints and additional procurement cost increases ahead.\n\nFor smartphone makers, rather than cutting production or trimming contracted volumes right away, the more pressing concern has become whether they can receive memory shipments on time going forward. The worry is that tampering with contracts now could mean failing to secure adequate volume next quarter or next year, or having to re-contract at even higher prices. So even in an unfavorable market, they appear to be maintaining their long-term supply agreements as-is.\n\nMemory chipmakers share this view. During the conference call following its Q4 earnings announcement, SK hynix explained: \"In the past, LTAs had loose volume commitments and were highly flexible depending on market conditions. The LTAs being discussed recently are not mere purchase intentions but arrangements where customers and suppliers firmly commit volumes to each other.\" This implies that from the supplier's perspective, demand visibility has become far more important than before.\n\nAn industry insider said: \"In this unprecedented memory supply shortage, the power dynamic between buyer and seller is reversing. Even as consumer electronics demand slows, semiconductor supply contracts are being maintained — and these contracts are, in turn, entrenching the supply shortage and price increases in a self-reinforcing cycle.\"\n\n$MU $SNDK\n\n---\n\nhttps://t.co/Yk7ywEN1Er", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04687501029159635, "ret_signed_m1d": -0.04687501029159635, "alpha_spy_m1d": -0.04207631668347811, "alpha_c_m1d": -0.0313883600803615, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.015486650211234854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00360579897461355, "ret_signed_p1d": -0.00360579897461355, "alpha_spy_p1d": -0.0009686385425327071, "alpha_c_p1d": 0.001974979716881209, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.005580778691494759, "ret_p1w": 0.0889422572521763, "ret_signed_p1w": 0.0889422572521763, "alpha_spy_p1w": 0.1065228270365195, "alpha_c_p1w": 0.11538114834364888, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": -0.02643889109147257, "ret_p1m": 0.12920674586734182, "ret_signed_p1m": 0.12920674586734182, "alpha_spy_p1m": 0.15337280848239565, "alpha_c_p1m": 0.15425044228594076, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.025043696418598937, "ret_p3m": 0.6349876825470429, "ret_signed_p3m": 0.6349876825470429, "alpha_spy_p3m": 0.5778827788238594, "alpha_c_p3m": 0.44976986122915896, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.1852178213178839, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3834, -0.3852, -0.4187, -0.3984, -0.4151, -0.4229, -0.3984, -0.433, -0.4217, -0.4061, -0.3852, -0.381, -0.399, -0.3972, -0.3816, -0.375, -0.3714, -0.3517, -0.3451, -0.3517, -0.3541, -0.3583, -0.3487, -0.3732, -0.3852, -0.3547, -0.3565, -0.3643, -0.3391, -0.3332, -0.3356, -0.3356, -0.3003, -0.2819, -0.2794, -0.2794, -0.2794, -0.2278, -0.1701, -0.1653, -0.1526, -0.1659, -0.1647, -0.1659, -0.1731, -0.1569, -0.1352, -0.1052, -0.1028, -0.1274, -0.1016, -0.0847, -0.0859, -0.0859, -0.0415, -0.024, -0.0343, -0.0355, -0.0962, 0.0066, 0.0162, -0.0427, -0.0469, 0.0, -0.0036, 0.0084, 0.0733, 0.0889, 0.0889, 0.0889, 0.0889, 0.1418, 0.1424, 0.1599, 0.2019, 0.223, 0.3101, 0.3011, 0.3011, 0.1725, 0.0349, 0.1514, 0.131, 0.0427, 0.1292, 0.1418, 0.1292, 0.1028, 0.134, 0.1653, 0.253, 0.2049, 0.1983, 0.1196, 0.14, 0.1358, 0.0823, 0.0823, 0.0617, 0.0069, 0.1421, 0.0743, 0.1213, 0.1629, 0.1833, 0.2676, 0.2285, 0.2405, 0.2104, 0.2436, 0.2707, 0.3098, 0.3008, 0.2917, 0.3188, 0.3098, 0.352, 0.3218, 0.352, 0.3369, 0.361, 0.3279, 0.3279, 0.4001, 0.4001, 0.6019, 0.635, 0.6169, 0.7193, 0.6802, 0.7103, 0.7825, 0.629, 0.6922, 0.6591, 0.6621, 0.8036, 0.7615, 0.7615, 0.8006, 0.8488, 0.8036, 0.909, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2020822218973094226", "ticker_idx": 1, "ts": "2026-02-09T11:28:56+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory suppliers are leading the LTAs. NCNR contracts with no price negotiation are expanding to PCs, mobile, and commodity DRAM", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply shortage flips LTA power to memory chipmakers; shorter contracts + NCNR expansion locks in volume and pricing power for Samsung, Hynix, and MU.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Supply Shortage Reshapes LTA Dynamics… Contract Periods Shrink and Power Balance Flips\n\nAs the supply shortage drags on, the trading order in the memory semiconductor market is rapidly being reorganized. With the semiconductor supply crunch persisting over an extended period, the initiative in long-term supply agreements (LTAs) is shifting from customers to semiconductor suppliers. In the past, smartphone and PC manufacturers seeking stable procurement were the ones to propose long-term contracts first, but recently memory chipmakers such as Samsung Electronics and SK hynix have been presenting contract terms first, signaling a shift in the market's balance of power.\n\nAccording to industry sources on the 9th, the contract periods of LTAs recently signed by Korean memory chipmakers are getting shorter. LTA durations vary from six months to one year depending on the product and end customer, but industry insiders universally say these periods have been shrinking lately.\n\nThe analysis is that as the semiconductor supply shortage continues, suppliers are opting for shorter contract periods. In a rising-price environment, shorter contract durations allow suppliers to reflect updated pricing more quickly, giving them an advantage in negotiations.\n\nThe initiative in LTAs is also shifting from customers to memory suppliers. Previously, it was common for smartphone and PC makers to propose long-term supply agreements first to secure volume, but recently Samsung Electronics and SK hynix are increasingly taking the lead in setting contract terms and structures. As the semiconductor supply shortage extends, the center of gravity in business relationships is tilting toward suppliers.\n\nLee Jong-wook, a researcher at Samsung Securities, analyzed: \"It was customers who first proposed long-term supply agreements, but now memory suppliers are leading the LTAs. NCNR (non-cancellable, non-returnable) contracts with no price negotiation are expanding to PCs, mobile, and commodity DRAM, causing the volume of available spot supply in the market to shrink dramatically.\"\n\nStockpiling behavior is also emerging. According to market research firm TrendForce, smartphone manufacturers are being extremely cautious about immediately reducing their LTA-based memory purchase volumes this year. TrendForce noted: \"Concerned about further price increases and potential supply shortages, they are preemptively prioritizing resource procurement to avoid future allocation restrictions.\" In other words, despite slowing global smartphone demand and a negative market outlook, companies are stocking up on semiconductors out of fear of supply constraints and additional procurement cost increases ahead.\n\nFor smartphone makers, rather than cutting production or trimming contracted volumes right away, the more pressing concern has become whether they can receive memory shipments on time going forward. The worry is that tampering with contracts now could mean failing to secure adequate volume next quarter or next year, or having to re-contract at even higher prices. So even in an unfavorable market, they appear to be maintaining their long-term supply agreements as-is.\n\nMemory chipmakers share this view. During the conference call following its Q4 earnings announcement, SK hynix explained: \"In the past, LTAs had loose volume commitments and were highly flexible depending on market conditions. The LTAs being discussed recently are not mere purchase intentions but arrangements where customers and suppliers firmly commit volumes to each other.\" This implies that from the supplier's perspective, demand visibility has become far more important than before.\n\nAn industry insider said: \"In this unprecedented memory supply shortage, the power dynamic between buyer and seller is reversing. Even as consumer electronics demand slows, semiconductor supply contracts are being maintained — and these contracts are, in turn, entrenching the supply shortage and price increases in a self-reinforcing cycle.\"\n\n$MU $SNDK\n\n---\n\nhttps://t.co/Yk7ywEN1Er", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.054115040689468374, "ret_signed_m1d": -0.054115040689468374, "alpha_spy_m1d": -0.04931634708135013, "alpha_c_m1d": -0.04181945462025327, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.012295586069215103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012401389967856113, "ret_signed_p1d": -0.012401389967856113, "alpha_spy_p1d": -0.00976422953577527, "alpha_c_p1d": -0.007753697457769659, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.004647692510086454, "ret_p1w": -0.00789183854575981, "ret_signed_p1w": -0.00789183854575981, "alpha_spy_p1w": 0.009688731238583381, "alpha_c_p1w": -0.01052311197568434, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": 0.00263127342992453, "ret_p1m": 0.05944847079742255, "ret_signed_p1m": 0.05944847079742255, "alpha_spy_p1m": 0.08361453341247638, "alpha_c_p1m": 0.08236746124155181, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.02291899044412926, "ret_p3m": 0.8681532867633712, "ret_signed_p3m": 0.8681532867633712, "alpha_spy_p3m": 0.8110483830401878, "alpha_c_p3m": 0.5399841396038025, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.3281691471595687, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3049, -0.3105, -0.3691, -0.3173, -0.3578, -0.3669, -0.3567, -0.413, -0.4142, -0.4153, -0.4097, -0.3867, -0.4025, -0.3935, -0.3709, -0.3777, -0.389, -0.3867, -0.3495, -0.3619, -0.3382, -0.363, -0.3563, -0.3754, -0.4025, -0.3788, -0.3777, -0.3833, -0.3461, -0.3416, -0.3371, -0.3371, -0.3247, -0.2785, -0.2661, -0.2661, -0.2661, -0.2368, -0.2153, -0.1815, -0.1635, -0.1477, -0.1612, -0.1556, -0.168, -0.1635, -0.1556, -0.1477, -0.1387, -0.1623, -0.1657, -0.1488, -0.1353, -0.1702, -0.0981, -0.0519, -0.0293, 0.0248, -0.0643, 0.0225, 0.0147, -0.0507, -0.0541, 0.0, -0.0124, -0.0304, 0.0011, -0.0079, -0.0079, -0.0079, -0.0079, 0.0079, 0.0699, 0.0722, 0.133, 0.1477, 0.2413, 0.1984, 0.1984, 0.0606, -0.0411, 0.0628, 0.0436, -0.0558, 0.0594, 0.0786, 0.0504, 0.0278, 0.1001, 0.0956, 0.1927, 0.1442, 0.1374, 0.0538, 0.1137, 0.1238, 0.0538, 0.0538, -0.014, -0.0885, 0.0131, -0.0625, -0.0106, 0.0007, 0.0346, 0.1667, 0.1272, 0.16, 0.1747, 0.2458, 0.2831, 0.3045, 0.274, 0.317, 0.3825, 0.3813, 0.3836, 0.3802, 0.4593, 0.4683, 0.4604, 0.4525, 0.4525, 0.6344, 0.6344, 0.8083, 0.8682, 0.9043, 1.1234, 1.0726, 1.2318, 1.2251, 1.0545, 1.0782, 0.9709, 0.9709, 1.1912, 1.1923, 1.1923, 1.3177, 1.5334, 1.5858, 1.6355, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2020822218973094226", "ticker_idx": 2, "ts": "2026-02-09T11:28:56+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$MU $SNDK", "tweet_type": "analysis", "timeline": "quarters", "summary": "Supply shortage flips LTA power to memory chipmakers; shorter contracts + NCNR expansion locks in volume and pricing power for Samsung, Hynix, and MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Supply Shortage Reshapes LTA Dynamics… Contract Periods Shrink and Power Balance Flips\n\nAs the supply shortage drags on, the trading order in the memory semiconductor market is rapidly being reorganized. With the semiconductor supply crunch persisting over an extended period, the initiative in long-term supply agreements (LTAs) is shifting from customers to semiconductor suppliers. In the past, smartphone and PC manufacturers seeking stable procurement were the ones to propose long-term contracts first, but recently memory chipmakers such as Samsung Electronics and SK hynix have been presenting contract terms first, signaling a shift in the market's balance of power.\n\nAccording to industry sources on the 9th, the contract periods of LTAs recently signed by Korean memory chipmakers are getting shorter. LTA durations vary from six months to one year depending on the product and end customer, but industry insiders universally say these periods have been shrinking lately.\n\nThe analysis is that as the semiconductor supply shortage continues, suppliers are opting for shorter contract periods. In a rising-price environment, shorter contract durations allow suppliers to reflect updated pricing more quickly, giving them an advantage in negotiations.\n\nThe initiative in LTAs is also shifting from customers to memory suppliers. Previously, it was common for smartphone and PC makers to propose long-term supply agreements first to secure volume, but recently Samsung Electronics and SK hynix are increasingly taking the lead in setting contract terms and structures. As the semiconductor supply shortage extends, the center of gravity in business relationships is tilting toward suppliers.\n\nLee Jong-wook, a researcher at Samsung Securities, analyzed: \"It was customers who first proposed long-term supply agreements, but now memory suppliers are leading the LTAs. NCNR (non-cancellable, non-returnable) contracts with no price negotiation are expanding to PCs, mobile, and commodity DRAM, causing the volume of available spot supply in the market to shrink dramatically.\"\n\nStockpiling behavior is also emerging. According to market research firm TrendForce, smartphone manufacturers are being extremely cautious about immediately reducing their LTA-based memory purchase volumes this year. TrendForce noted: \"Concerned about further price increases and potential supply shortages, they are preemptively prioritizing resource procurement to avoid future allocation restrictions.\" In other words, despite slowing global smartphone demand and a negative market outlook, companies are stocking up on semiconductors out of fear of supply constraints and additional procurement cost increases ahead.\n\nFor smartphone makers, rather than cutting production or trimming contracted volumes right away, the more pressing concern has become whether they can receive memory shipments on time going forward. The worry is that tampering with contracts now could mean failing to secure adequate volume next quarter or next year, or having to re-contract at even higher prices. So even in an unfavorable market, they appear to be maintaining their long-term supply agreements as-is.\n\nMemory chipmakers share this view. During the conference call following its Q4 earnings announcement, SK hynix explained: \"In the past, LTAs had loose volume commitments and were highly flexible depending on market conditions. The LTAs being discussed recently are not mere purchase intentions but arrangements where customers and suppliers firmly commit volumes to each other.\" This implies that from the supplier's perspective, demand visibility has become far more important than before.\n\nAn industry insider said: \"In this unprecedented memory supply shortage, the power dynamic between buyer and seller is reversing. Even as consumer electronics demand slows, semiconductor supply contracts are being maintained — and these contracts are, in turn, entrenching the supply shortage and price increases in a self-reinforcing cycle.\"\n\n$MU $SNDK\n\n---\n\nhttps://t.co/Yk7ywEN1Er", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029178622415872324, "ret_signed_m1d": 0.029178622415872324, "alpha_spy_m1d": 0.03397731602399057, "alpha_c_m1d": 0.04147420848508743, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.012295586069215103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02672751304786447, "ret_signed_p1d": -0.02672751304786447, "alpha_spy_p1d": -0.024090352615783628, "alpha_c_p1d": -0.022079820537778017, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.004647692510086454, "ret_p1w": 0.07342899206456432, "ret_signed_p1w": 0.07342899206456432, "alpha_spy_p1w": 0.09100956184890752, "alpha_c_p1w": 0.0707977186346398, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": 0.00263127342992453, "ret_p1m": 0.0511342381423634, "ret_signed_p1m": 0.0511342381423634, "alpha_spy_p1m": 0.07530030075741723, "alpha_c_p1m": 0.07405322858649266, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.02291899044412926, "ret_p3m": 0.6868361280257833, "ret_signed_p3m": 0.6868361280257833, "alpha_spy_p3m": 0.6297312243025999, "alpha_c_p3m": 0.3586669808662146, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.3281691471595687, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3617, -0.3824, -0.3566, -0.3694, -0.4044, -0.4111, -0.4751, -0.4595, -0.4163, -0.4148, -0.3998, -0.3998, -0.3836, -0.3732, -0.3758, -0.3897, -0.4092, -0.3817, -0.3564, -0.3421, -0.3126, -0.3263, -0.3715, -0.381, -0.394, -0.4122, -0.3522, -0.3069, -0.2791, -0.2799, -0.2528, -0.2528, -0.2577, -0.2324, -0.2369, -0.2558, -0.2558, -0.1775, -0.186, -0.1045, -0.1146, -0.1473, -0.1002, -0.0981, -0.1183, -0.1308, -0.1222, -0.0541, -0.0541, -0.0482, 0.0146, 0.0367, 0.0421, 0.0146, 0.0697, 0.135, 0.1363, 0.0818, 0.1416, 0.0937, -0.0107, -0.0016, 0.0292, 0.0, -0.0267, 0.07, 0.0795, 0.0734, 0.0734, 0.0425, 0.0977, 0.0883, 0.1165, 0.0977, 0.09, 0.1186, 0.0836, 0.0753, 0.0761, -0.01, 0.045, 0.0353, -0.0344, 0.0152, 0.0511, 0.0918, 0.057, 0.1112, 0.152, 0.2039, 0.204, 0.1585, 0.1027, 0.0544, 0.0314, -0.0037, -0.0731, -0.0685, -0.1605, -0.1187, -0.0404, -0.0446, -0.0446, -0.0146, -0.015, 0.061, 0.0996, 0.0972, 0.1127, 0.2147, 0.1901, 0.1928, 0.1871, 0.1698, 0.1723, 0.2717, 0.2566, 0.2958, 0.3684, 0.3155, 0.3525, 0.3491, 0.4144, 0.5038, 0.6701, 0.7389, 0.6868, 0.9482, 1.0747, 0.9997, 1.0964, 1.0243, 0.8904, 0.7779, 0.8228, 0.9095, 0.9881, 0.9591, 0.9591, 1.337, 1.4219, 1.4091, 1.533, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2036806187925024952", "ticker_idx": 0, "ts": "2026-03-25T14:03:31+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy; raise target price to US$450", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF reiterates Buy on Broadcom, raises PT to $450 on higher TPU forecasts, AI revenue upgrades, and strong external/Anthropic demand.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications\n☄️ Broadcom (AVGO US Buy): Raising 2027 TPU Forecast\n\n☀️ Maintain Buy; raise target price to US$450:\nWe forecast total TPU shipments to reach 4.5 million units in 2026 and 7.9 million units in 2027, above our previous estimates of 4.5 million and 6.0 million, respectively. This upward revision is mainly driven by strong demand from external customers, with TPU shipments designed by Broadcom expected to reach 4.1 million units in 2026 and 5.8 million units in 2027. Based on this, we raise our AI-related revenue forecasts for FY2026/FY2027/FY2028 to US$50bn/US$110bn/US$130bn, and revise FY2026/FY2027 EPS forecasts to US$10.8/US$18.0. Given lower market risk appetite and multiple compression across AI fabless names, we reduce our target P/E from 27x to 25x, implying a target price of US$450.\n\n☀️ External demand and ASP upside drive TPU forecast revision:\nWe now expect Google TPU shipments to reach 4.5 million units in 2026 and 7.9 million units in 2027, versus our previous forecast of 4.5 million and 6.0 million, mainly due to stronger demand from external customers. Of these, TPU shipments designed by Broadcom are expected to total 4.1 million units in 2026 and 5.8 million units in 2027. On October 23, 2025, Anthropic announced a major expansion in TPU usage, with contract value reaching tens of billions of dollars, and this is expected to contribute more than 1GW of compute capacity in 2026. We believe Broadcom will be the key beneficiary of this wave of external demand, as its Ironwood and Sunfish products are already available for customer testing, while MediaTek’s Zebrafish has not yet begun. On ASP, we expect the uptrend to continue. Pumafish, scheduled for launch in 2027, features greater design complexity and could carry an ASP of more than US$20,000. In comparison, our shipment forecasts for Meta ASICs have been lowered to 300,000 units in 2026 and 500,000 units in 2027.\n\n☀️ AI networking business continues to support growth:\nAccording to the latest earnings results, AI networking is becoming an increasingly important growth driver alongside compute. In 1Q, AI networking revenue grew by approximately 60% YoY, accounting for about one-third of total AI revenue. Management expects that share to rise to around 40% in 2Q, and to remain within the 33%–40% range over the longer term. Looking ahead, Tomahawk 7 is expected to launch in 2027, with performance improving by roughly 2x. On architecture, Broadcom reiterated that copper interconnect technology in scale-up systems can support the transition from 200G to 400G SerDes through 2028. The TH series, used in scale-up products for Meta and AMD, is expected to begin contributing revenue in 2026. As such, the company views CPO as a long-term evolution path rather than a near-term necessity. We forecast shipments of its scale-up CPO switches to reach 5,000 units in 2026 and 20,000 units in 2027.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0016310310769859226, "ret_signed_m1d": -0.0016310310769859226, "alpha_spy_m1d": 0.00391084433461486, "alpha_c_m1d": 0.009471151264133448, "bench_spy_m1d": -0.005541875411600783, "bench_c_m1d": -0.011102182341119371, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02945322983208387, "ret_signed_p1d": -0.02945322983208387, "alpha_spy_p1d": -0.011594485673945165, "alpha_c_p1d": 0.01610837905051854, "bench_spy_p1d": -0.017858744158138706, "bench_c_p1d": -0.04556160888260241, "ret_p1w": -0.016687078087132434, "ret_signed_p1w": -0.016687078087132434, "alpha_spy_p1w": -0.0142815224216154, "alpha_c_p1w": 0.0009811791106972834, "bench_spy_p1w": -0.002405555665517034, "bench_c_p1w": -0.017668257197829718, "ret_p1m": 0.31721089569729033, "ret_signed_p1m": 0.31721089569729033, "alpha_spy_p1m": 0.23860487836060007, "alpha_c_p1m": 0.10962726738185347, "bench_spy_p1m": 0.07860601733669026, "bench_c_p1m": 0.20758362831543686, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1022, 0.0936, 0.0951, 0.0833, 0.0833, 0.0881, 0.0749, 0.076, 0.0752, 0.0407, 0.0798, 0.1025, 0.11, 0.0639, 0.0737, 0.1009, 0.1009, 0.0411, 0.0292, 0.0188, 0.0018, 0.0168, 0.0417, 0.0431, 0.0352, 0.037, 0.0364, 0.0027, -0.0358, -0.0281, 0.0421, 0.0766, 0.0656, 0.0729, 0.0366, 0.0178, 0.0178, 0.0409, 0.0439, 0.0454, 0.0412, 0.034, 0.0188, 0.0402, 0.007, 0.0002, -0.0021, -0.0176, -0.0061, 0.0416, 0.0344, 0.0822, 0.0723, 0.0691, 0.0516, 0.0084, 0.017, 0.0057, -0.0111, 0.0011, -0.0281, 0.0116, -0.0016, 0.0, -0.0295, -0.0569, -0.0797, -0.0292, -0.0167, -0.0134, -0.0134, -0.0137, 0.0476, 0.0998, 0.1132, 0.1654, 0.1911, 0.1944, 0.2444, 0.2499, 0.2752, 0.2535, 0.2615, 0.3257, 0.3172, 0.3261, 0.3118, 0.2541, 0.2718, 0.3093, 0.3214, 0.3064, 0.3405, 0.3345, 0.2941, 0.3488, 0.3438, 0.3152, 0.3073, 0.3795, 0.3337, 0.3196, 0.2894, 0.3104, 0.3004, 0.299, 0.299, 0.3237, 0.3232, 0.338, 0.4014, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2006546180961087643", "ticker_idx": 0, "ts": "2026-01-01T02:01:04+00:00", "symbol": "LG Electronics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm worried about LG Electronics. I'm genuinely very concerned.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish LG Electronics into 2026: legacy memory price surge will hammer HA costs, demand may weaken more than laptops.", "resolved_tickers": ["066570.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "LG Electronics 066570.KS Korea", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "In 2026, one group of companies that could face major headwinds would be HA (home appliance) companies.\n\nHome appliances tend to use older-generation memory, and since the prices of legacy memory have surged sharply, it looks like this will hit costs hard.\n\nIt might even be that HA ends up suffering more from demand weakness caused by steep price hikes than laptops do.\n\nI’m worried about LG Electronics. I’m genuinely very concerned.\n\n---\n\nhttps://t.co/Y55d048eyk", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005440757462399581, "ret_signed_p1d": 0.005440757462399581, "alpha_spy_p1d": 0.007273797195851417, "alpha_c_p1d": 0.00773289920462461, "bench_spy_p1d": 0.0018330397334518356, "bench_c_p1d": 0.002292141742225029, "ret_p1w": 0.0032643688164546436, "ret_signed_p1w": -0.0032643688164546436, "alpha_spy_p1w": 0.007865980556036156, "alpha_c_p1w": -0.0013889511770868168, "bench_spy_p1w": 0.0111303493724908, "bench_c_p1w": 0.0018754176393678268, "ret_p1m": 0.0783459651877183, "ret_signed_p1m": -0.0783459651877183, "alpha_spy_p1m": -0.0636082323914926, "alpha_c_p1m": -0.07897099828316767, "bench_spy_p1m": 0.0147377327962257, "bench_c_p1m": -0.0006250330954493677, "ret_p3m": 0.15786240207450342, "ret_signed_p3m": -0.15786240207450342, "alpha_spy_p3m": -0.20156824766190717, "alpha_c_p3m": -0.23360569535796571, "bench_spy_p3m": -0.04370584558740376, "bench_c_p3m": -0.0757432932834623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1687, -0.1687, -0.1687, -0.1687, -0.1404, -0.1186, -0.0968, -0.0827, -0.0871, -0.0838, -0.0098, -0.0294, -0.025, -0.0185, -0.0424, -0.0316, -0.0305, -0.0337, -0.0468, -0.0468, -0.0359, -0.037, -0.0522, -0.0022, -0.0392, -0.025, -0.0141, 0.0109, 0.0065, -0.0185, -0.0054, -0.0468, -0.0522, -0.0501, -0.0598, -0.086, -0.0729, -0.0588, -0.0707, -0.0686, -0.0707, -0.0511, -0.0261, 0.0316, 0.0849, 0.0566, 0.0392, 0.0283, 0.0305, 0.049, 0.0413, 0.0131, 0.0479, 0.012, 0.0261, 0.0218, 0.0087, 0.0065, 0.0065, -0.0033, -0.0011, 0.0, 0.0, 0.0, -0.0054, 0.012, 0.0229, 0.0044, 0.0033, -0.0305, -0.0424, -0.012, 0.0065, 0.0229, 0.0707, 0.1632, 0.1425, 0.1893, 0.1556, 0.1491, 0.1153, 0.1023, 0.0871, 0.0979, 0.0783, 0.0511, 0.0914, 0.1186, 0.0914, 0.0794, 0.1164, 0.1317, 0.3917, 0.321, 0.2753, 0.2753, 0.2753, 0.2753, 0.3177, 0.3449, 0.4429, 0.4635, 0.4505, 0.5963, 0.5854, 0.5854, 0.3885, 0.1861, 0.2557, 0.284, 0.2339, 0.2612, 0.2829, 0.2405, 0.2307, 0.2427, 0.2459, 0.3058, 0.2644, 0.2829, 0.1915, 0.2437, 0.2481, 0.2296, 0.2187, 0.1875, 0.1579, 0.2324, 0.1929, 0.1875, 0.1995, 0.1743, 0.2796, 0.2851, 0.2982, 0.2642, 0.3158, 0.3399, 0.3815, 0.3618, 0.3651, 0.3881, 0.4561, 0.4243, 0.398, 0.4254, 0.535, 0.489, 0.5449, 0.5449, 0.5701, 0.5701, 0.6984, 0.6304, 0.6896, 0.7182, 1.0274, 1.0986, 1.3793, 1.637, 1.3793, 1.1019, 0.9846, 1.5767, 1.5986, 1.5986, 1.626, 1.5767, 1.4725, 2.2126, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2016703239128043903", "ticker_idx": 0, "ts": "2026-01-29T02:41:35+00:00", "symbol": "UMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "What we and GFHK agree on is that UMC has a very solid positioning in specialty processes. Its long-term growth potential also looks compelling. We believe UMC has a meaningful chance of winning key silicon photonics orders.", "tweet_type": "news_reaction", "timeline": "years", "summary": "Endorses GFHK PT raise on UMC; bullish on specialty process positioning, long-term growth, and silicon photonics order wins.", "resolved_tickers": ["UMC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "GFHK has raised its price target on UMC.\n\nThey noted that UMC should benefit as TSMC scales back legacy capacity.\n\nWhat we and GFHK agree on is that UMC has a very solid positioning in specialty processes. Its long-term growth potential also looks compelling. We believe UMC has a meaningful chance of winning key silicon photonics orders.\n\nFor more details, please refer to the link below.\n\n---\n\nhttps://t.co/TKa2Emhunf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07561438530281506, "ret_signed_m1d": 0.07561438530281506, "alpha_spy_m1d": 0.0736260572505556, "alpha_c_m1d": 0.0777459817635302, "bench_spy_m1d": 0.001988328052259458, "bench_c_m1d": -0.002131596460715146, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.037807147581732536, "ret_signed_p1d": -0.037807147581732536, "alpha_spy_p1d": -0.034824567321329325, "alpha_c_p1d": -0.004132119011703184, "bench_spy_p1d": -0.0029825802604032114, "bench_c_p1d": -0.03367502857002935, "ret_p1w": -0.0576559225969796, "ret_signed_p1w": -0.0576559225969796, "alpha_spy_p1w": -0.03399739368990129, "alpha_c_p1w": 0.029645242972507813, "bench_spy_p1w": -0.023658528907078313, "bench_c_p1w": -0.08730116556948742, "ret_p1m": -0.013232546723281335, "ret_signed_p1m": -0.013232546723281335, "alpha_spy_p1m": -0.0016337900544063366, "alpha_c_p1m": 0.013472746891102338, "bench_spy_p1m": -0.011598756668874999, "bench_c_p1m": -0.026705293614383674, "ret_p3m": 0.10207938861002774, "ret_signed_p3m": 0.10207938861002774, "alpha_spy_p3m": 0.0738480931836143, "alpha_c_p3m": -0.07441516106527701, "bench_spy_p3m": 0.02823129542641345, "bench_c_p3m": 0.17649454967530476, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.259, -0.2817, -0.2665, -0.3091, -0.2996, -0.2977, -0.3025, -0.3157, -0.328, -0.3251, -0.3365, -0.3327, -0.3166, -0.327, -0.3157, -0.2968, -0.3129, -0.3072, -0.3072, -0.2949, -0.2883, -0.2647, -0.258, -0.2561, -0.242, -0.2391, -0.2353, -0.2439, -0.2543, -0.2561, -0.259, -0.2543, -0.2439, -0.2372, -0.2344, -0.2316, -0.2467, -0.2439, -0.2439, -0.242, -0.2467, -0.2476, -0.2571, -0.2571, -0.259, -0.2656, -0.2325, -0.156, -0.1919, -0.1692, -0.1635, -0.1815, -0.1692, -0.1711, -0.121, -0.121, 0.0189, 0.052, 0.0302, -0.0038, 0.0709, 0.1796, 0.0756, 0.0, -0.0378, -0.051, -0.069, -0.0595, -0.0577, -0.0491, -0.0501, -0.0633, -0.0265, -0.0274, -0.0151, -0.0151, -0.0293, -0.0331, -0.0425, -0.017, -0.0151, 0.0321, -0.0028, -0.0265, -0.0132, -0.0066, -0.0643, -0.0662, -0.0784, -0.0926, -0.0803, -0.104, -0.0832, -0.1163, -0.1238, -0.1049, -0.1002, -0.121, -0.1248, -0.138, -0.1408, -0.1446, -0.12, -0.1474, -0.1616, -0.1862, -0.1512, -0.1522, -0.1805, -0.1805, -0.1786, -0.12, -0.1049, -0.0992, -0.0794, -0.0832, -0.0936, -0.0595, 0.0038, 0.1021, 0.1947, 0.1654, 0.2013, 0.1342, 0.1607, 0.1049, 0.1021, 0.2051, 0.2344, 0.2335, 0.2268, 0.3242, 0.4395, 0.4329, 0.4575, 0.4735, 0.518, 0.5047, 0.6191, 0.6257, 0.6181, 0.6616, 0.6597, 0.7344, 0.7221, 0.7221, 0.9924, 1.1002, 1.1437, 1.0964, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2031252413538250795", "ticker_idx": 0, "ts": "2026-03-10T06:14:48+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "When NVIDIA's ambitions extend beyond the GPU toward system-level optimization of the entire agentic workflow, the difficulty of competing is no longer on any single dimension... NVIDIA is operating from a systems perspective — orchestrating heterogeneous hardware across the entire agentic flow from the ground up. In terms of system complexity and software stack depth (Dynamo / ICMS / CMX), this move raises the competitive bar from 'build a great chip' to 'define an entire heterogeneous system as a universal accelerated computing solution.'", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long NVDA: post-Groq heterogeneous inference stack (LPU+CPX+HBM GPU+ICMS) raises competitive bar to system-level, mirroring mobile SoC consolidation winners.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "GTC 2026 Preview: The New Era of Heterogeneous AI Inference Through the Lens of Groq's Niche\n\nWhere does Groq's SRAM approach fit in the ecosystem? Will SRAM displace HBM? How will NVIDIA integrate Groq into its existing product lineup — technical integration or product line integration? And what upgrades will the acquisition bring to the Groq LPU?\n\nHere I attempt to piece together a logical chain from first principles.\n---------------------------------\nGroq's Design Philosophy: Compiler-First, Not SRAM-First\n\nGroq is fundamentally a compiler-first architecture taken to its extreme — SRAM is a byproduct, not the primary thesis.\n\nUnlike CPUs designed for general-purpose workloads, AI inference is characterized by high determinism: virtually no data-dependent branching, fixed tensor shapes, and predictable memory access patterns.\n\nWhen Groq examined the hardware-software interface through this lens and asked \"what should be done at compile time versus runtime?\", the answer for AI inference was: almost everything can be done at compile time.\n\nThis is Groq's most radical and distinctive insight: a fully deterministic compiler, scheduled down to the exact clock cycle. Complete determinism enables extreme efficiency. To achieve this, the compiler must account for every state of the hardware at every moment of execution — acting as an omniscient god — which eliminates wasted hardware resources. This requires every computation, memory access, and communication latency within the LPU to be precise to the clock cycle, which is enormously complex for a compiler.\n\nThe high determinism of AI workloads, combined with Groq's fully deterministic compiler approach, naturally avoids VLIW's weaknesses (unpredictable memory and branch behavior) while amplifying its strengths. The next step — increasing efficiency and parallelism — naturally leads to VLIW-style instruction encoding: since the compiler controls what each functional unit does every cycle, packing multiple operations into a single wide instruction yields higher efficiency. This is VLIW.\n\nIn Groq's chips, there is no out-of-order execution or speculation. Hardware is dramatically simplified (instruction dispatch occupies less than 3% of die area), and complexity is shifted entirely to the static compiler. This is the essence of the VLIW philosophy.\n\nTo enable cycle-accurate deterministic scheduling by the compiler, all sources of non-determinism in hardware must be eliminated — arbiters, crossbars, replays, and any component that makes runtime decisions via autonomous algorithms are removed.\n\nMemory latency must also be deterministic, which is why all caches and DRAM are eliminated. Caches are replaced with scratchpad SRAM, because cache replacement policies are runtime decisions — non-deterministic. Addresses are fully controlled by the compiler, guaranteeing determinism.\n\nCommunication must also be cycle-precise. Send and receive instructions are executed at compiler-scheduled moments — there are no traditional \"I want to send a packet, please allocate memory\" operations. Instead, SRAM allocation and data transfer are executed synchronously according to a pre-determined timetable decided by the compiler; the hardware simply follows it.\n\nThe fully deterministic compiler also delivers extremely low overhead latency for inter-chip communication — arguably Groq's most underappreciated advantage. In traditional interconnect architectures, packet routing, arbiter contention, and buffer queuing are the primary sources of latency variance.\n\nSo Groq is not fundamentally an SRAM-first architecture, nor strictly a VLIW-first architecture — it is a compiler-first architecture. More precisely, the fully deterministic compiler is the core of the entire Groq design.\n\nSRAM-only became a necessary consequence of the deterministic compiler: at the core decode stage, the non-determinism inherent to HBM/DRAM is incompatible. This is why Groq is better described as compiler-first.\n------------------------------------\nCan Groq's Deterministic Compiler Be Applied to NVIDIA's GPU + HBM Architecture?\n\nNo. For two reasons:\n\n1. HBM/DRAM's physical characteristics make deterministic latency impossible.\n\n- DRAM cells must be refreshed at regular intervals (tREFI) to replenish charge, blocking bank access. This is dictated by DRAM cell physics, and refresh frequency varies with temperature.\n\n- To maximize DRAM bandwidth, controllers perform dynamic optimizations such as batch scheduling — grouping traffic to the same page to minimize page misses, spreading access across banks, and minimizing read-write switching. These optimizations occur in real time and are fundamentally unpredictable.\n\n- System-level DRAM optimizations like bank address hashing make it extremely difficult for a compiler to statically pre-locate specific data, and the complexity of enforcing cycle-level determinism becomes prohibitive.\n\nThese non-deterministic behaviors can technically be resolved — but only by abandoning most DRAM optimization strategies, severely degrading efficiency and utilization. Groq itself explored this direction and filed a patent for deterministic DRAM, but concluded that practical implementation was infeasible. This is one of the core reasons Groq chose SRAM-only.\n\nApplying a deterministic compiler to DRAM is not a yes-or-no question — it's simply a bad choice, because it structurally and unavoidably sacrifices HBM efficiency and bandwidth. It would essentially require rewriting a full memory controller as a compiler-defined software construct — extraordinarily complex, requiring major updates with each memory generation, and a validation nightmare across every DRAM vendor and generation.\n\n2. NVIDIA's SIMT architecture is philosophically opposed to Groq's compiler-first approach.\n\nThese two architectures give opposite answers to the same problem — runtime non-determinism:\n\n- Groq eliminates all non-determinism at compile time.\n\n- NVIDIA hides unpredictable latency through warp switching.\n\nNVIDIA GPUs are built on SIMT (Single Instruction, Multiple Threads) and hardware warp schedulers. When a warp stalls on a memory access, the scheduler instantly switches to another ready warp, hiding the stall latency. This entire mechanism presupposes that latency is unpredictable, so sufficient concurrent threads are needed to statistically fill the pipeline.\nReplacing this with a deterministic compiler would mean abandoning NVIDIA's most fundamental hardware scheduling unit — and with it, the need for large register files to support multi-warp rotation.\n\nHistorically, AMD's transition from TeraScale (VLIW) to GCN (scalar SIMT) was precisely a large-scale VLIW-to-SIMT migration: when workloads became less predictable, the compiler burden of VLIW became too heavy, and scheduling authority had to be returned to hardware.\n\nGroq's only viable path within NVIDIA is as a standalone, purpose-built product for low-latency decode.\n--------------------------------------------------\nWhat Will NVIDIA Bring to Groq?\n\nGroq's current bottlenecks, simply stated:\n\n1. SRAM capacity is too small to hold large model weights + KV cache\n\n2. The primary decode bottleneck is not SRAM bandwidth (80 TB/s) but interconnect latency (accounting for ~80% of total latency)\n\n3. Prefill (a compute-bound task) is relatively slow\n\nGroq's core architecture was essentially finalized in 2017–2018, the era of CNNs. The architecture targeted CNN/LSTM workloads, with benchmarks like ResNet50, for which 230MB of SRAM per TSP compute card was more than sufficient.\n\nBut in the LLM era, 230MB of on-chip SRAM per card is woefully inadequate. A LLaMA 70B model's parameter footprint is equivalent to ~3,000 ResNet50 models, and the KV cache grows further with long context. Scale-out became the only option.\n\nRunning a 70B model thus requires a 576-card cluster: 16-stage pipeline parallelism (PP) × 36-way tensor parallelism (TP), with 80 transformer layers split into 16 pipeline stages and each stage spread across 36 cards. The PP and TP communication overhead constitutes over 80% of total latency — PP alone accounts for more than 50%.\n\nAfter joining NVIDIA's product line, Groq can play to its strengths by focusing exclusively on decode (where it excels) and avoiding prefill (where it is weak).\n\nThe most important improvement NVIDIA can bring is likely expanding Groq LPU SRAM capacity through process node upgrades and hybrid bonding technology (analogous to AMD's 3D V-Cache). Moving from 14nm to 3nm alone could grow SRAM from 230MB to ~500MB per chip; adding 3D SRAM stacking could double capacity again.\n\nWith larger SRAM, a 70B inference job that previously required 576 LPUs might need only 256. With 32-way tensor parallel × 8-stage pipeline parallel, pipeline interconnect latency could be cut roughly in half.\n\nNVIDIA's primary contribution: expand SRAM capacity → reduce scale-out card count → reduce communication latency → increase token throughput.\n---------------------------------------------\nWill SRAM Displace HBM?\n\nNo.\n\nSRAM is fundamentally an exchange of 10× the cost for a few times the speed. It is viable only for a subset of customers willing to pay a premium for low latency. The dominant metric in the AI hardware market remains TCO (total cost of ownership).\n\nA simple cost comparison makes this clear: running LLaMA 70B including KV cache requires 576 Groq cards. At a retail price of ~$20,000 per card (the CEO has stated actual prices are far lower, so let's use $2,000), that's over $1.1M in hardware costs. Two H100s can run the same model for under $100,000 — a 10× cost gap.\n\nGroq pivoted to selling token API services, and its pricing is indeed cheap — but for two reasons: (1) NVIDIA GPU cloud providers typically sell capacity at 2× hardware cost; and (2) Groq itself is operating at a loss. In 2025, Groq's LPU-based API inference business generated ~$40M in revenue against ~$60M in costs — a –50% gross margin. Groq's low token prices are not evidence of SRAM's economic superiority; they are VC subsidies.\n\nIs there a market willing to pay a speed premium? Yes.\n\nClaude Opus 4.6's Fast mode is a telling market signal: 2.5× faster output, priced at $30/$150 per million tokens versus $5/$25 — a 6× price increase, presumably achieved by sacrificing batch efficiency. This market segment is real, and SRAM has a genuine niche here.\n\nBut how large is this niche?\n\nDifferent workloads have vastly different hardware requirements:\n\nPrefill: low bandwidth requirement, high compute requirement\n\nDecode: the reverse — high bandwidth, lower compute\n\nRanking & Recommendation: modest compute and bandwidth, but enormous capacity requirements\n\nLatency-sensitive decode (the red zone in the accompanying chart) is SRAM's domain: real-time, interactive LLM applications — chat, copilot, agents. Reasoning models in particular show dramatic improvements: H100 might take 2–3 minutes to complete a reasoning chain; Cerebras handles it in 10 seconds.\n\nThe share of inference workloads demanding extreme speed appears to be around 10% of hyperscaler compute, based on one data point.\n\nAgentic workloads are more complicated: profiling common agentic frameworks (SWE-Agent, LangChain, Toolformer) shows CPU can account for up to 90% of end-to-end latency, with throughput bottlenecks more often in CPU than in decode. The SRAM speed advantage is substantially diluted here.\n\nThe larger volume workloads — batch inference, offline processing, ranking, recommendation — are insensitive to latency; throughput and cost-per-token are the only metrics. SRAM has no cost advantage here whatsoever.\n\nAn analogy: H100/B200 is a bus — high passenger capacity (large batches), low per-seat cost, but slower. Groq/Cerebras is a Ferrari — extreme speed, few passengers, per-seat cost 10× or more.\n\nStructurally, SRAM's cost disadvantage will not converge over time. A 6T SRAM cell is inherently more expensive than a 1T1C DRAM cell — this is determined by physics, not process node. And SRAM scaling has stalled: from N5 to N3E, SRAM cell area has barely shrunk.\n\nEven the speed advantage is eroding: SRAM access speeds are approaching process limits with little room for generational improvement, while HBM bandwidth is scaling exponentially. A decade ago, SRAM's 2-order-of-magnitude speed lead over HBM was a crushing advantage. Today the gap is less than one order of magnitude (Rubin HBM4: 22 TB/s). In another decade, the gap may be negligible.\n\nConclusion: SRAM will not displace HBM, but it has an irreplaceable role in the low-latency, low-batch, real-time inference segment. Long-term, as HBM bandwidth scales exponentially, even that advantage will gradually diminish.\n-------------------------------------------------------\nThe New Era of Heterogeneous Inference\n\nPiecing these fragments together, the outline of NVIDIA's post-Groq acquisition strategy comes into focus: the dawn of heterogeneous inference.\n\nInference workloads have already diverged — no single architecture can cover the optimal operating point for all of them. The most important principle in computer architecture is the tradeoff and the range of scales it covers. A single architecture may be extraordinary within its optimal tradeoff space; using multiple architectures tailored to different workload types is the essence of heterogeneous computing.\n\nThe defining theme of GTC 2026 is the systematization of heterogeneous inference. Inference will no longer be handled by a single piece of hardware — it will be decomposed:\n\n- Control, scheduling, and agent runtime → Vera CPU\n\n- Long-context prefill → CPX (Content Phase Accelerator, a compute-bound prefill-optimized module)\n\n- Small models / low-latency / low-batch decode → Groq LPU (256-card clusters)\n\n- High-throughput / high-concurrency batch decode → HBM GPU (still the workhorse)\n\n- ICMS / CMX (Inference Context Memory Storage): KV cache is now core infrastructure. Heterogeneity has extended beyond compute into the memory hierarchy itself\n\nLPU and GPU will likely become two different tiers within the inference stack: small models, low-latency, low-batch workloads to LPU; long context, high-batch workloads to HBM GPU.\n\nHow CPX connects to LPU/GPU remains unclear. The general workflow would be: CPU handles control and scheduling → CPX completes prefill and generates tens of GB of KV cache → distributed to Groq LPU SRAM arrays or HBM GPU → decode begins.\n\nThere is also a more speculative possibility: speculative decoding, where LPU runs the typically smaller draft model at blazing speed and HBM GPU serves as the verifier. This heterogeneous decoding structure could substantially increase token rates — potentially doubling them in some scenarios (e.g., coding tasks, where syntax is predictable and small draft models achieve high acceptance rates).\n\nWhen NVIDIA's ambitions extend beyond the GPU toward system-level optimization of the entire agentic workflow, the difficulty of competing is no longer on any single dimension. NVIDIA used to take big leaps through brute-force advances in GPU architecture and parameters. Now, with CPX, LPU, and ICMS joining the heterogeneous inference stack, NVIDIA is operating from a systems perspective — orchestrating heterogeneous hardware across the entire agentic flow from the ground up.\n\nIn terms of system complexity and software stack depth (Dynamo / ICMS / CMX), this move raises the competitive bar from \"build a great chip\" to \"define an entire heterogeneous system as a universal accelerated computing solution.\"\n---------------------------------------------------------\nA Historical Parallel\n\nEvery paradigm shift in computing has spawned a wave of semiconductor startups. But as software and application forms converge, large incumbents ultimately absorb the functionality through acquisitions — scaling parameters higher, integrating systems more deeply and comprehensively, reducing costs, improving power efficiency and benchmark scores — gradually crowding out independent startups.\n\nIn the early mobile era, there was a profusion of independent companies building application processors, discrete baseband chips, ISPs, and GPUs. The ultimate winners were those who integrated GPU, ISP, and modem into a single SoC with complete system-level heterogeneous compute platforms.\n\nApple acquired PA Semi's CPU team, Infineon's modem, and hollowed out Imagination's GPU. Qualcomm acquired ATI's mobile GPU, Atheros for Wi-Fi, Nuvia for CPU, CSR for Bluetooth/DSP — all canonical examples.\n\nAs heterogeneous inference grows in complexity, companies capable of system-level integration will have an increasing advantage. The logic is identical to the mobile SoC era. In the AI era, NVIDIA's acquisitions of Arm (failed), Mellanox, and Groq are merely the beginning of this new historical cycle.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": "GTC 2026 preview： 从Groq生态位看AI异构推理(Heterogeneous Inference)新时代\n\nGroq的SRAM路线的生态位在哪里？SRAM会不会替代HBM路线？\n\nNvidia如何整合groq到现有的产品线？是技术整合还是产品线整合？收购之后会给groq LPU产品带来怎样的升级？\n\n这里尝试从基本原理出发去拼凑一个逻辑链 https://t.co/oAS2suk2uo", "parent_author": null, "ret_m1d": -0.011473822539532774, "ret_signed_m1d": -0.011473822539532774, "alpha_spy_m1d": -0.013083536459450595, "alpha_c_m1d": -0.004024116866387595, "bench_spy_m1d": 0.0016097139199178212, "bench_c_m1d": -0.007449705673145179, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006873722377926494, "ret_signed_p1d": 0.006873722377926494, "alpha_spy_p1d": 0.008128885306443623, "alpha_c_p1d": -0.002438467318479187, "bench_spy_p1d": -0.0012551629285171284, "bench_c_p1d": 0.009312189696405682, "ret_p1w": -0.015317259906154623, "ret_signed_p1w": -0.015317259906154623, "alpha_spy_p1w": -0.005881033941038005, "alpha_c_p1w": -0.0141847461149478, "bench_spy_p1w": -0.009436225965116618, "bench_c_p1w": -0.001132513791206824, "ret_p1m": -0.014505346356984972, "ret_signed_p1m": -0.014505346356984972, "alpha_spy_p1m": -0.015503927520376659, "alpha_c_p1m": -0.07891031884907374, "bench_spy_p1m": 0.0009985811633916875, "bench_c_p1m": 0.06440497249208876, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0208, -0.0528, -0.0459, -0.0382, -0.0748, -0.0575, -0.0205, -0.0058, 0.024, 0.0208, 0.0208, 0.0312, 0.0187, 0.015, 0.0094, 0.0094, 0.0221, 0.0181, 0.0134, 0.0235, 0.0015, 0.0005, 0.0009, 0.0056, -0.0088, 0.0123, 0.0079, 0.0079, -0.0363, -0.0078, 0.0004, 0.0157, 0.0092, 0.0203, 0.0365, 0.0419, 0.0344, 0.0045, -0.024, -0.0573, -0.0698, 0.0035, 0.0285, 0.0204, 0.0286, 0.0117, -0.0106, -0.0106, 0.0011, 0.0174, 0.0169, 0.0273, 0.0367, 0.0437, 0.0584, 0.0006, -0.041, -0.0124, -0.0255, -0.0094, -0.0077, -0.0376, -0.0115, 0.0, 0.0069, -0.0088, -0.0244, -0.0083, -0.0153, -0.0236, -0.0336, -0.0653, -0.0494, -0.0517, -0.0329, -0.0732, -0.0933, -0.106, -0.0561, -0.0488, -0.0399, -0.0399, -0.0385, -0.036, -0.0145, -0.0046, 0.0209, 0.0246, 0.0636, 0.0764, 0.0736, 0.0916, 0.0936, 0.0818, 0.096, 0.0805, 0.1272, 0.1724, 0.1538, 0.1326, 0.0802, 0.0741, 0.0743, 0.0635, 0.1249, 0.1447, 0.1648, 0.1877, 0.195, 0.2223, 0.2759, 0.2195, 0.2033, 0.194, 0.2095, 0.1881, 0.1655, 0.1655, 0.1629, 0.1507, 0.1596, 0.1428, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038446519477801177", "ticker_idx": 0, "ts": "2026-03-30T02:41:37+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We believe this is a correction, not a peak-out... riding through this correction offers a more favorable risk-return profile", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities analyst calls current selloff in Samsung Electronics and SK Hynix a correction not a peak; recommends riding through it as DRAM cycle plateaus rather than peaks out.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Views on Memory Stock Price Decline\n[Samsung Securities Semiconductor, IT / Jong-wook Lee]\n\nAs of 10:30 AM, Samsung Electronics and SK Hynix shares are down 3.9% and 5.6%, respectively.\n\n1. Spot Price Decline\n\nD5 16G spot prices peaked on March 19 and have since corrected over the past week, with the premium over contract prices narrowing from a 1Q average of 30% to roughly 20% recently. The equity market is concerned this may mark an inflection point signaling DRAM price declines.\n\nWe believe the pace of DRAM price increases will begin to moderate from 3Q onward, and that early signs of this deceleration could trigger a share price correction around May–June. From this same perspective, the narrowing of spot price premiums appears to be interpreted as evidence of slowing DRAM price momentum.\n\n2. Views on the Correction\n\nWe believe this is a correction, not a peak-out. The market narrative will shift from the magnitude of DRAM price increases to the sustainability of DRAM prices. Because it is difficult to pinpoint the exact inflection between these two themes, we believe riding through this correction offers a more favorable risk-return profile.\n\nWe see a high probability that DRAM profits will trace a broad plateau rather than the typical cyclical peak, as the supply-side response has lagged roughly one year behind compared to previous cycles. We believe this plateau-shaped earnings profile will instill greater confidence among investors in the sustainability of the cycle.\n\n3. The Cycle\n\nWe maintain that DRAM remains a cyclical industry. That said, we caution against selling on the vague fear that \"it's cyclical, so it has to come down eventually.\"\n\nCyclical peak-outs originate from changes in downstream industries. Therefore, early signs of a peak-out can be identified through business trajectory shifts at OpenAI, Anthropic, and the ISPs — and we believe there is still upside to be demonstrated from the demand side.\n\nLooking back at 2025, the equity market has wobbled every time spot prices stalled. While spot price declines can induce fear among investors, we do not believe they should be extrapolated as evidence of a cyclical peak-out.\n\n4. Natural Gas Supply Issues\n\nNatural gas is a cost inflation issue, not a production disruption issue. In memory, cost is dominated by the variability of technology, yields, and capex — which has historically overwhelmed raw material cost fluctuations. Moreover, in a commodity industry, raw material supply disruptions cannot be viewed as purely negative. What matters is whether demand is strong enough to absorb cost pass-throughs.\n\n5. Sentiment\n\nExamining the drivers behind the recent share price correction, this is a sentiment issue rather than a fundamental one — though sentiment itself is a meaningful factor in equity markets. Investors are feeling significant anxiety about sustainability following the sharp run-up in both DRAM prices and stock prices, and they want concrete evidence to alleviate that anxiety.\n\nIf tangible evidence emerges regarding the direction from 3Q onward, we believe share prices can stabilize and begin to reflect earnings with greater confidence.\n\nThank you.\n\n(Published 2026/3/30)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01944412932501427, "ret_signed_m1d": 0.01944412932501427, "alpha_spy_m1d": 0.01608945070003953, "alpha_c_m1d": 0.000463751529364842, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.018980377795649428, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.051616562677254674, "ret_signed_p1d": -0.051616562677254674, "alpha_spy_p1d": -0.08068448837616948, "alpha_c_p1d": -0.09396945598301443, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.04235289330575975, "ret_p1w": 0.0952921157118547, "ret_signed_p1w": 0.0952921157118547, "alpha_spy_p1w": 0.052631826105592694, "alpha_c_p1w": 0.02250781156050663, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.07278430415134807, "ret_p1m": 0.259217243335224, "ret_signed_p1m": 0.259217243335224, "alpha_spy_p1m": 0.13307195885345457, "alpha_c_p1m": 0.021177979778238765, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.23803926355698524, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3213, -0.3213, -0.2726, -0.2183, -0.2138, -0.2019, -0.2143, -0.2132, -0.2143, -0.2211, -0.2058, -0.1855, -0.1572, -0.1549, -0.1781, -0.1538, -0.1379, -0.139, -0.139, -0.0972, -0.0807, -0.0904, -0.0915, -0.1487, -0.0519, -0.0428, -0.0983, -0.1023, -0.0581, -0.0615, -0.0502, 0.011, 0.0257, 0.0257, 0.0257, 0.0257, 0.0755, 0.076, 0.0925, 0.1321, 0.1519, 0.234, 0.2255, 0.2255, 0.1044, -0.0253, 0.0845, 0.0653, -0.0179, 0.0636, 0.0755, 0.0636, 0.0387, 0.0681, 0.0976, 0.1802, 0.1349, 0.1287, 0.0545, 0.0738, 0.0698, 0.0194, 0.0194, 0.0, -0.0516, 0.0757, 0.0119, 0.0562, 0.0953, 0.1146, 0.194, 0.1571, 0.1685, 0.1401, 0.1713, 0.1968, 0.2337, 0.2252, 0.2167, 0.2422, 0.2337, 0.2734, 0.245, 0.2734, 0.2592, 0.2819, 0.2507, 0.2507, 0.3188, 0.3188, 0.5088, 0.54, 0.523, 0.6194, 0.5825, 0.6109, 0.679, 0.5343, 0.5939, 0.5627, 0.5655, 0.6988, 0.6591, 0.6591, 0.696, 0.7413, 0.6988, 0.7981, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038446519477801177", "ticker_idx": 1, "ts": "2026-03-30T02:41:37+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "riding through this correction offers a more favorable risk-return profile", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung Securities analyst calls current selloff in Samsung Electronics and SK Hynix a correction not a peak; recommends riding through it as DRAM cycle plateaus rather than peaks out.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Views on Memory Stock Price Decline\n[Samsung Securities Semiconductor, IT / Jong-wook Lee]\n\nAs of 10:30 AM, Samsung Electronics and SK Hynix shares are down 3.9% and 5.6%, respectively.\n\n1. Spot Price Decline\n\nD5 16G spot prices peaked on March 19 and have since corrected over the past week, with the premium over contract prices narrowing from a 1Q average of 30% to roughly 20% recently. The equity market is concerned this may mark an inflection point signaling DRAM price declines.\n\nWe believe the pace of DRAM price increases will begin to moderate from 3Q onward, and that early signs of this deceleration could trigger a share price correction around May–June. From this same perspective, the narrowing of spot price premiums appears to be interpreted as evidence of slowing DRAM price momentum.\n\n2. Views on the Correction\n\nWe believe this is a correction, not a peak-out. The market narrative will shift from the magnitude of DRAM price increases to the sustainability of DRAM prices. Because it is difficult to pinpoint the exact inflection between these two themes, we believe riding through this correction offers a more favorable risk-return profile.\n\nWe see a high probability that DRAM profits will trace a broad plateau rather than the typical cyclical peak, as the supply-side response has lagged roughly one year behind compared to previous cycles. We believe this plateau-shaped earnings profile will instill greater confidence among investors in the sustainability of the cycle.\n\n3. The Cycle\n\nWe maintain that DRAM remains a cyclical industry. That said, we caution against selling on the vague fear that \"it's cyclical, so it has to come down eventually.\"\n\nCyclical peak-outs originate from changes in downstream industries. Therefore, early signs of a peak-out can be identified through business trajectory shifts at OpenAI, Anthropic, and the ISPs — and we believe there is still upside to be demonstrated from the demand side.\n\nLooking back at 2025, the equity market has wobbled every time spot prices stalled. While spot price declines can induce fear among investors, we do not believe they should be extrapolated as evidence of a cyclical peak-out.\n\n4. Natural Gas Supply Issues\n\nNatural gas is a cost inflation issue, not a production disruption issue. In memory, cost is dominated by the variability of technology, yields, and capex — which has historically overwhelmed raw material cost fluctuations. Moreover, in a commodity industry, raw material supply disruptions cannot be viewed as purely negative. What matters is whether demand is strong enough to absorb cost pass-throughs.\n\n5. Sentiment\n\nExamining the drivers behind the recent share price correction, this is a sentiment issue rather than a fundamental one — though sentiment itself is a meaningful factor in equity markets. Investors are feeling significant anxiety about sustainability following the sharp run-up in both DRAM prices and stock prices, and they want concrete evidence to alleviate that anxiety.\n\nIf tangible evidence emerges regarding the direction from 3Q onward, we believe share prices can stabilize and begin to reflect earnings with greater confidence.\n\nThank you.\n\n(Published 2026/3/30)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0687284851813359, "ret_signed_m1d": 0.0687284851813359, "alpha_spy_m1d": 0.06537380655636116, "alpha_c_m1d": 0.036400122672946456, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.03232836250838944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07560141246406815, "ret_signed_p1d": -0.07560141246406815, "alpha_spy_p1d": -0.10466933816298296, "alpha_c_p1d": -0.13316904876302404, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.05756763629895589, "ret_p1w": 0.014891163435468258, "ret_signed_p1w": 0.014891163435468258, "alpha_spy_p1w": -0.027769126170793745, "alpha_c_p1w": -0.07737709662483949, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.09226826006030775, "ret_p1m": 0.4891179188388093, "ret_signed_p1m": 0.4891179188388093, "alpha_spy_p1m": 0.36297263435703986, "alpha_c_p1m": 0.1341680026689207, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.3549499161698886, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2557, -0.2557, -0.2259, -0.2042, -0.1699, -0.1516, -0.1356, -0.1493, -0.1436, -0.1562, -0.1516, -0.1436, -0.1356, -0.1265, -0.1505, -0.1539, -0.1368, -0.123, -0.1585, -0.0853, -0.0384, -0.0156, 0.0393, -0.051, 0.037, 0.029, -0.0373, -0.0407, 0.0142, 0.0016, -0.0167, 0.0153, 0.0062, 0.0062, 0.0062, 0.0062, 0.0222, 0.0851, 0.0873, 0.1491, 0.1639, 0.2589, 0.2153, 0.2153, 0.0756, -0.0275, 0.0779, 0.0584, -0.0424, 0.0745, 0.0939, 0.0653, 0.0424, 0.1157, 0.1111, 0.2096, 0.1604, 0.1535, 0.0687, 0.1294, 0.1397, 0.0687, 0.0687, 0.0, -0.0756, 0.0275, -0.0493, 0.0034, 0.0149, 0.0493, 0.1833, 0.1432, 0.1764, 0.1913, 0.2635, 0.3013, 0.323, 0.2921, 0.3356, 0.4021, 0.4009, 0.4032, 0.3998, 0.48, 0.4891, 0.4811, 0.4731, 0.4731, 0.6575, 0.6575, 0.8339, 0.8946, 0.9313, 1.1535, 1.1019, 1.2635, 1.2566, 1.0836, 1.1077, 0.9989, 0.9989, 1.2222, 1.2234, 1.2234, 1.3505, 1.5693, 1.6224, 1.6728, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2018968434550444495", "ticker_idx": 0, "ts": "2026-02-04T08:42:40+00:00", "symbol": "4062.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I picked up a small amount of Ibiden shares today. I agreed with Bernstein's view that even if Intel wins EMIB-T orders from external customers, it's actually Ibiden—the one supplying the substrates to Intel—that stands to benefit far more. On top of that, I believe ABF substrates are going to face severe shortages going forward.", "tweet_type": "position_disclosure", "timeline": "months", "summary": "Long Ibiden: ABF substrate supplier benefits most from Intel EMIB-T wins; 42% supply shortage projected over next two years.", "resolved_tickers": ["4062.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "I picked up a small amount of Ibiden shares today.\n\nI agreed with Bernstein’s view that even if Intel wins EMIB-T orders from external customers, it’s actually Ibiden—the one supplying the substrates to Intel—that stands to benefit far more. On top of that, I believe ABF substrates are going to face severe shortages going forward.\n\nAnd coincidentally, Goldman Sachs just put out a report saying that ABF substrate supply is projected to fall short by a whopping 42% over the next two years.\n\nI’m really disappointed that I suddenly ran short on cash and couldn’t buy more Ibiden.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.16541664835115188, "ret_signed_m1d": 0.16541664835115188, "alpha_spy_m1d": 0.16054914888607374, "alpha_c_m1d": 0.13674584574543736, "bench_spy_m1d": 0.00486749946507814, "bench_c_m1d": 0.02867080260571453, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011805541071106651, "ret_signed_p1d": -0.011805541071106651, "alpha_spy_p1d": 0.0006836701850950755, "alpha_c_p1d": 0.006222152029930705, "bench_spy_p1d": -0.012489211256201727, "bench_c_p1d": -0.018027693101037356, "ret_p1w": 0.09972225413519031, "ret_signed_p1w": 0.09972225413519031, "alpha_spy_p1w": 0.09131343341345288, "alpha_c_p1w": 0.06460773562638922, "bench_spy_p1w": 0.008408820721737431, "bench_c_p1w": 0.035114518508801096, "ret_p1m": 0.12486117292715315, "ret_signed_p1m": 0.12486117292715315, "alpha_spy_p1m": 0.13197289164572734, "alpha_c_p1m": 0.1099465506419508, "bench_spy_p1m": -0.007111718718574189, "bench_c_p1m": 0.01491462228520235, "ret_p3m": 0.861014830573392, "ret_signed_p3m": 0.861014830573392, "alpha_spy_p3m": 0.8117851582524143, "alpha_c_p3m": 0.6863008102178643, "bench_spy_p3m": 0.049229672320977746, "bench_c_p3m": 0.17471402035552774, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1097, -0.0948, -0.0712, -0.0483, -0.0542, -0.1344, -0.0788, -0.1569, -0.1917, -0.1295, -0.2132, -0.2132, -0.2132, -0.2427, -0.2049, -0.175, -0.2139, -0.2243, -0.1802, -0.1743, -0.1177, -0.1153, -0.1021, -0.1243, -0.1194, -0.1142, -0.1743, -0.2101, -0.1903, -0.2253, -0.191, -0.1493, -0.1267, -0.0785, -0.1017, -0.0806, -0.0797, -0.0651, -0.0651, -0.0651, -0.0651, -0.0058, -0.0261, 0.0031, -0.0485, -0.0675, -0.0675, 0.0097, 0.0328, 0.0392, 0.0882, 0.0876, 0.1017, 0.1551, 0.1869, 0.1506, 0.1001, 0.1111, 0.1458, 0.1468, 0.1429, 0.0736, 0.1654, 0.0, -0.0118, -0.016, 0.0825, 0.0997, 0.0997, 0.1811, 0.2, 0.2194, 0.2424, 0.2964, 0.3386, 0.2718, 0.2718, 0.3728, 0.3406, 0.2915, 0.3246, 0.3011, 0.1969, 0.1042, 0.1249, 0.0944, 0.0053, 0.0879, 0.0829, 0.0647, 0.0908, 0.1318, 0.1804, 0.1779, 0.1483, 0.1483, 0.0472, 0.0882, 0.1775, 0.175, 0.1485, 0.082, 0.0257, 0.1229, 0.0818, 0.1424, 0.1902, 0.213, 0.3627, 0.365, 0.3758, 0.306, 0.3796, 0.3051, 0.3319, 0.285, 0.3323, 0.4693, 0.5584, 0.5493, 0.7448, 0.7629, 0.7476, 0.7476, 0.8756, 0.861, 0.861, 0.861, 0.861, 1.2784, 1.1588, 1.1783, 1.3028, 1.3738, 1.3598, 1.179, 1.1755, 1.1316, 1.1762, 1.4872, 1.6785, 1.9122, 1.9171, 1.842, 1.7466, 2.2002, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2009782526207041908", "ticker_idx": 0, "ts": "2026-01-10T00:21:08+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "인텔 사세요~", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Terse buy instruction on Intel.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@StarAndCompass_ 인텔 사세요~", "reply_text": null, "parent_text": "@jukan05 대 주 칸 https://t.co/ig0nWUa5Sl", "quoted_text": null, "parent_author": "StarAndCompass_", "ret_m1d": 0.033817472022884765, "ret_signed_m1d": 0.033817472022884765, "alpha_spy_m1d": 0.035385463376658266, "alpha_c_m1d": 0.037503547965923256, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07330911124742578, "ret_signed_p1d": 0.07330911124742578, "alpha_spy_p1d": 0.07530858635329607, "alpha_c_p1d": 0.07108212330642338, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.06581928326810282, "ret_signed_p1w": 0.06581928326810282, "alpha_spy_p1w": 0.07085411142866405, "alpha_c_p1w": 0.04091268700762707, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.06967770311249333, "ret_signed_p1m": 0.06967770311249333, "alpha_spy_p1m": 0.07405082307457278, "alpha_c_p1m": 0.033584921739074236, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.40081705167893, "ret_signed_p3m": 0.40081705167893, "alpha_spy_p3m": 0.4200833602636783, "alpha_c_p3m": 0.29932216219754215, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1568, -0.1639, -0.16, -0.1353, -0.1348, -0.1621, -0.1339, -0.1312, -0.1026, -0.0574, -0.0617, -0.0885, -0.0924, -0.1035, -0.1596, -0.1289, -0.1548, -0.1346, -0.1273, -0.1403, -0.14, -0.185, -0.1938, -0.2122, -0.2208, -0.2031, -0.2369, -0.217, -0.1877, -0.1868, -0.1645, -0.1645, -0.0794, -0.0919, -0.0134, -0.0068, -0.0808, -0.0601, -0.0853, -0.0808, -0.0744, -0.1033, -0.1419, -0.1487, -0.1532, -0.1818, -0.1766, -0.1643, -0.1745, -0.175, -0.1793, -0.1793, -0.1784, -0.1675, -0.1534, -0.1625, -0.1625, -0.1062, -0.1064, -0.0912, -0.0325, -0.067, 0.0338, 0.0, 0.0733, 0.1058, 0.0967, 0.0658, 0.0658, 0.1021, 0.2313, 0.2329, 0.0229, -0.0356, -0.003, 0.1071, 0.1044, 0.0547, 0.1078, 0.1178, 0.103, 0.0949, 0.1482, 0.1403, 0.0697, 0.096, 0.0549, 0.062, 0.062, 0.0481, 0.0318, 0.0127, 0.0011, -0.0098, 0.0468, 0.064, 0.0318, 0.0352, 0.0327, -0.0218, 0.0345, 0.0429, -0.0145, 0.0345, 0.0617, 0.089, 0.027, 0.0388, 0.0386, 0.0, 0.022, 0.0481, -0.0043, -0.0011, 0.0, 0.0708, 0.0009, -0.0211, -0.0651, 0.0016, 0.0901, 0.1434, 0.1434, 0.1525, 0.2009, 0.3379, 0.4008, 0.4158, 0.4793, 0.4483, 0.4739, 0.5547, 0.5547, 0.4911, 0.5039, 0.4814, 0.5157, 0.8734, 0.929, 0.9183, 1.1505, 1.1443, 1.261, 1.1739, 1.4546, 1.5649, 1.488, 1.8352, 1.9378, 1.7374, 1.7301, 1.6312, 1.4687, 1.4551, 1.5148, 1.7, 1.6895, 1.7199, 1.7199, 1.8034, 1.7637, 1.7438, 1.6028, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2019777050727653653", "ticker_idx": 0, "ts": "2026-02-06T14:15:49+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "selling Micron because of this is foolish, too. Micron can earn even higher margins from server DDR5 instead of HBM4", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Don't sell MU on HBM4 supply news; server DDR5 margins compensate and the selloff is unwarranted.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Wasn’t it already widely known news that Micron wouldn’t be able to supply HBM4? I don’t understand why everyone is making such a big fuss about it.\n\nAnd selling Micron because of this is foolish, too. Micron can earn even higher margins from server DDR5 instead of HBM4.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029896801235108517, "ret_signed_m1d": -0.029896801235108517, "alpha_spy_m1d": -0.01107321448183185, "alpha_c_m1d": 0.021341806719581524, "bench_spy_m1d": -0.018823586753276667, "bench_c_m1d": -0.05123860795469004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.028351368538319544, "ret_signed_p1d": -0.028351368538319544, "alpha_spy_p1d": -0.03317320064133, "alpha_c_p1d": -0.04080001804903921, "bench_spy_p1d": 0.004821832103010459, "bench_c_p1d": 0.012448649510719667, "ret_p1w": 0.042995811110824844, "ret_signed_p1w": 0.042995811110824844, "alpha_spy_p1w": 0.05583931934793307, "alpha_c_p1w": 0.02788313236948481, "bench_spy_p1w": -0.012843508237108225, "bench_c_p1w": 0.015112678741340035, "ret_p1m": -0.013605593725321996, "ret_signed_p1m": -0.013605593725321996, "alpha_spy_p1m": 0.004276773811875745, "alpha_c_p1m": 0.004519636923064674, "bench_spy_p1m": -0.01788236753719774, "bench_c_p1m": -0.01812523064838667, "ret_p3m": 0.6896046503298983, "ret_signed_p3m": 0.6896046503298983, "alpha_spy_p3m": 0.6241357244771648, "alpha_c_p3m": 0.3208507150134883, "bench_spy_p3m": 0.06546892585273345, "bench_c_p3m": 0.36875393531641, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3894, -0.3798, -0.3999, -0.3749, -0.3872, -0.4213, -0.4278, -0.49, -0.4748, -0.4329, -0.4314, -0.4168, -0.4168, -0.4011, -0.391, -0.3935, -0.407, -0.426, -0.3992, -0.3746, -0.3607, -0.3321, -0.3454, -0.3893, -0.3985, -0.4111, -0.4288, -0.3705, -0.3265, -0.2995, -0.3003, -0.274, -0.274, -0.2787, -0.2542, -0.2586, -0.2769, -0.2769, -0.2008, -0.2091, -0.1299, -0.1397, -0.1715, -0.1257, -0.1237, -0.1433, -0.1554, -0.1471, -0.0809, -0.0809, -0.0752, -0.0141, 0.0073, 0.0126, -0.0142, 0.0394, 0.1028, 0.1041, 0.0512, 0.1092, 0.0627, -0.0387, -0.0299, 0.0, -0.0284, -0.0543, 0.0397, 0.0488, 0.043, 0.043, 0.0129, 0.0665, 0.0574, 0.0848, 0.0666, 0.0591, 0.0869, 0.0529, 0.0448, 0.0456, -0.038, 0.0154, 0.006, -0.0618, -0.0136, 0.0213, 0.0608, 0.027, 0.0797, 0.1194, 0.1698, 0.1699, 0.1256, 0.0715, 0.0245, 0.0021, -0.0319, -0.0994, -0.0949, -0.1843, -0.1437, -0.0676, -0.0717, -0.0717, -0.0425, -0.0429, 0.0309, 0.0684, 0.0661, 0.0812, 0.1803, 0.1564, 0.1589, 0.1535, 0.1366, 0.139, 0.2356, 0.221, 0.259, 0.3296, 0.2782, 0.3141, 0.3108, 0.3743, 0.4611, 0.6227, 0.6896, 0.639, 0.8929, 1.0159, 0.943, 1.037, 0.967, 0.8368, 0.7275, 0.7711, 0.8554, 0.9317, 0.9036, 0.9036, 1.2708, 1.3532, 1.3408, 1.4612, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2020392242121437265", "ticker_idx": 0, "ts": "2026-02-08T07:00:22+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it feels like Apple only has upside left in smartphone market share going forward", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish AAPL on growing smartphone market share gains in India and resilience vs Samsung in China.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Is it believable that Apple has already broken out of the “Others” category and climbed to 5th place in the Indian market?\n\nUnlike Samsung, Apple didn’t collapse in China, and it has successfully established itself in India.\n\nAt this point, it feels like Apple only has upside left in smartphone market share going forward.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011798054939316183, "ret_signed_m1d": 0.011798054939316183, "alpha_spy_m1d": 0.016596748547434426, "alpha_c_m1d": 0.027284705150551036, "bench_spy_m1d": -0.004798693608118243, "bench_c_m1d": -0.015486650211234854, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.003422959769044165, "ret_signed_p1d": -0.003422959769044165, "alpha_spy_p1d": -0.000785799336963322, "alpha_c_p1d": 0.002157818922450594, "bench_spy_p1d": -0.002637160432080843, "bench_c_p1d": -0.005580778691494759, "ret_p1w": -0.06860390424744434, "ret_signed_p1w": -0.06860390424744434, "alpha_spy_p1w": -0.051023334463101144, "alpha_c_p1w": -0.042165013155971764, "bench_spy_p1w": -0.017580569784343192, "bench_c_p1w": -0.02643889109147257, "ret_p1m": -0.050214866528038304, "ret_signed_p1m": -0.050214866528038304, "alpha_spy_p1m": -0.026048803912984475, "alpha_c_p1m": -0.025171170109439367, "bench_spy_p1m": -0.02416606261505383, "bench_c_p1m": -0.025043696418598937, "ret_p3m": 0.046682679928123916, "ret_signed_p3m": 0.046682679928123916, "alpha_spy_p3m": -0.010422223795059526, "alpha_c_p3m": -0.13853514138975997, "bench_spy_p3m": 0.05710490372318344, "bench_c_p3m": 0.1852178213178839, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0051, -0.007, -0.009, -0.027, -0.0271, -0.023, -0.0314, -0.0123, 0.0038, 0.0076, 0.0097, 0.0097, 0.0145, 0.0299, 0.0412, 0.0337, 0.0212, 0.0142, 0.011, 0.0084, 0.0142, 0.0115, 0.0124, -0.0028, -0.001, -0.011, -0.0098, -0.0044, -0.0142, -0.0092, -0.0039, -0.0039, -0.0054, -0.0041, -0.0065, -0.011, -0.011, -0.0141, -0.0277, -0.0455, -0.0529, -0.0576, -0.0564, -0.0532, -0.0503, -0.0543, -0.0606, -0.0704, -0.0704, -0.1025, -0.0991, -0.0965, -0.0976, -0.0708, -0.0604, -0.0671, -0.0604, -0.056, -0.0177, -0.0196, 0.0059, 0.0038, 0.0118, 0.0, -0.0034, 0.0032, -0.0469, -0.0686, -0.0686, -0.0391, -0.0374, -0.0511, -0.0366, -0.0307, -0.009, -0.0014, -0.0061, -0.038, -0.036, -0.0396, -0.0441, -0.0522, -0.0625, -0.0537, -0.0502, -0.0503, -0.0687, -0.0892, -0.0794, -0.0742, -0.0899, -0.0934, -0.097, -0.0842, -0.0837, -0.0801, -0.0791, -0.094, -0.1019, -0.0759, -0.0692, -0.0681, -0.0681, -0.0574, -0.0769, -0.0572, -0.0515, -0.0515, -0.0562, -0.0575, -0.0298, -0.0409, -0.016, -0.0057, -0.0308, -0.0053, -0.0043, -0.013, -0.0255, -0.0142, -0.0162, -0.0119, 0.0201, 0.008, 0.0348, 0.0469, 0.0467, 0.0681, 0.0667, 0.0745, 0.0893, 0.0869, 0.0943, 0.0856, 0.0897, 0.1016, 0.1116, 0.1256, 0.1256, 0.1238, 0.133, 0.139, 0.1374, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2007305740864917849", "ticker_idx": 0, "ts": "2026-01-03T04:19:17+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I invest in Intel based more on the U.S. government backing Intel than on Intel's own competitiveness", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Long INTC premised on U.S. government backing rather than Intel's own competitive position.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@sidsomanii I invest in Intel based more on the U.S. government backing Intel than on Intel’s own competitiveness.", "reply_text": null, "parent_text": "@jukan05 Does this play into your INTC thesis?", "quoted_text": null, "parent_author": "sidsomanii", "ret_m1d": 0.0002540547752560851, "ret_signed_m1d": 0.0002540547752560851, "alpha_spy_m1d": 0.006870104944181388, "alpha_c_m1d": 0.011615607217135171, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01701808488239398, "ret_signed_p1d": 0.01701808488239398, "alpha_spy_p1d": 0.011070901509304676, "alpha_c_p1d": -0.009545115617021205, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.11912630349623754, "ret_signed_p1w": 0.11912630349623754, "alpha_spy_p1w": 0.10830794372563846, "alpha_c_p1w": 0.0845120196143021, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.2509525358435727, "ret_signed_p1m": 0.2509525358435727, "alpha_spy_p1m": 0.2483205859384563, "alpha_c_p1m": 0.19774668682019736, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.2796546211565616, "ret_signed_p3m": 0.2796546211565616, "alpha_spy_p3m": 0.3234208110969331, "alpha_c_p3m": 0.24064402466515777, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0493, -0.0399, -0.0762, -0.0546, -0.095, -0.0564, -0.0643, -0.0599, -0.0323, -0.0318, -0.0622, -0.0307, -0.0277, 0.0043, 0.0549, 0.05, 0.0201, 0.0157, 0.0033, -0.0594, -0.0251, -0.0541, -0.0315, -0.0234, -0.0378, -0.0376, -0.0879, -0.0978, -0.1184, -0.128, -0.1082, -0.1461, -0.1237, -0.0909, -0.0899, -0.065, -0.065, 0.0302, 0.0163, 0.1041, 0.1115, 0.0287, 0.0518, 0.0236, 0.0287, 0.0358, 0.0036, -0.0396, -0.0472, -0.0523, -0.0843, -0.0785, -0.0648, -0.0762, -0.0767, -0.0815, -0.0815, -0.0805, -0.0683, -0.0526, -0.0627, -0.0627, 0.0003, 0.0, 0.017, 0.0828, 0.0442, 0.157, 0.1191, 0.2012, 0.2375, 0.2273, 0.1928, 0.1928, 0.2334, 0.378, 0.3797, 0.1448, 0.0792, 0.1158, 0.239, 0.236, 0.1803, 0.2398, 0.251, 0.2344, 0.2253, 0.285, 0.2761, 0.1971, 0.2266, 0.1806, 0.1885, 0.1885, 0.173, 0.1547, 0.1334, 0.1204, 0.1082, 0.1715, 0.1908, 0.1547, 0.1585, 0.1557, 0.0947, 0.1577, 0.1671, 0.1029, 0.1577, 0.1882, 0.2187, 0.1494, 0.1626, 0.1623, 0.1191, 0.1438, 0.173, 0.1143, 0.1179, 0.1191, 0.1984, 0.1201, 0.0955, 0.0462, 0.1209, 0.22, 0.2797, 0.2797, 0.2898, 0.3439, 0.4973, 0.5677, 0.5845, 0.6556, 0.6208, 0.6495, 0.7399, 0.7399, 0.6688, 0.683, 0.6579, 0.6962, 1.0965, 1.1588, 1.1468, 1.4067, 1.3998, 1.5304, 1.4328, 1.747, 1.8705, 1.7844, 2.173, 2.2878, 2.0635, 2.0554, 1.9446, 1.7628, 1.7475, 1.8143, 2.0216, 2.0099, 2.0439, 2.0439, 2.1374, 2.093, 2.0706, 1.9129, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2028363932755452153", "ticker_idx": 1, "ts": "2026-03-02T06:57:01+00:00", "symbol": "Keyware", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Keyware has also stated that demand for high-end drill bits exceeds supply, and is pursuing capacity expansion... order visibility extends through Q3 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Topoint and Keyware as AI-driven PCB drill bit demand surges, supply tightens, and both companies expand capacity with visible order books into 2026.", "resolved_tickers": ["6152.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Keyware Taiwan PCB drill bit maker 6152.TW", "bench_c": "EWT", "bench_c_reason": "country_fallback (suffix='TW')", "bench_c_industry": "Communication Equipment", "unit_kind": "self_thread", "lang": "en", "tweet_text": "PCB Boom Triggers Drill Bit Shortage.. Topoint & Keyware Begin Capacity Expansion — Once a Consumable, Now a Major AI Business Opportunity\n\n- Rising demand for AI servers is triggering bottlenecks in PCB materials. MLB designs for NVIDIA's GPU and ASIC platforms have expanded to 40+ layers, with growing adoption of ultra-low-loss CCL specifications such as M7 and M8.\n\n- As a result, the processing difficulty of advanced substrates has surged dramatically, elevating drill bits — once considered mere \"consumables\" — into critical strategic materials.\n\n- To address capacity shortfalls, Taiwan's Topoint and Keyware have begun expansion efforts, actively responding to growing AI demand.\n\n- According to industry sources, high-end CCL (copper clad laminate) has high hardness, which drastically shortens drill bit lifespan.\n\n- Drill bits that previously lasted ~3,000 hits per unit have now dropped to under 800, and with the potential introduction of M9-grade materials, usage could fall to as few as 100 hits.\n\n- Processing methods are also shifting from \"1 hole, 1 bit\" to \"1 hole, 3–5 bits.\" To manage yields, some PCB manufacturers have banned the re-grinding and reuse of worn bits, causing overall consumption to surge.\n\n- In light of tight supply conditions, Topoint has begun expanding capacity with a focus on high-end products. The company recently invested approximately NT$560 million to secure a second factory in Zhongli District, Taoyuan, primarily for mass production of premium coated bits.\n\n- The facility is located near major PCB manufacturers including Kinsus, GoldCircuit, and Compeq. Equipment installation will begin this year, with full-scale production expansion expected from next year.\n\n- Keyware has also stated that demand for high-end drill bits exceeds supply, and is pursuing capacity expansion.\n\n- Keyware plans to expand production capacity to 15 million units/month (vs. Topoint's 35 million units/month) by the end of Q1 2026, and emphasized that its order visibility extends through Q3 2026.\n\n- The growth rate of drill bit demand driven by AI advanced substrates significantly outpaces the overall PCB industry growth rate. As capacity is being rapidly depleted, early signs of price increases have also been detected.\n\n- Price hikes are likely to begin with lower-end \"white bits,\" but if specifications are further upgraded alongside new platform mass production in 2026, price increases could extend across premium products more broadly.\n\n---\n\nhttps://t.co/h4suO4ryY9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08977552646153808, "ret_signed_m1d": -0.08977552646153808, "alpha_spy_m1d": -0.0892072707773317, "alpha_c_m1d": -0.09776804646367232, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": 0.00799252000213424, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09975062723709183, "ret_signed_p1d": 0.09975062723709183, "alpha_spy_p1d": 0.1085649657198875, "alpha_c_p1d": 0.15010361504398384, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.05035298780689201, "ret_p1w": -0.08977552646153808, "ret_signed_p1w": -0.08977552646153808, "alpha_spy_p1w": -0.07795994553802321, "alpha_c_p1w": -0.032362477625587305, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.057413048835950775, "ret_p1m": -0.21695758570179113, "ret_signed_p1m": -0.21695758570179113, "alpha_spy_p1m": -0.16703784412660738, "alpha_c_p1m": -0.16167582810555348, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.05528175759623766, "ret_p3m": -0.2518702957218121, "ret_signed_p3m": -0.2518702957218121, "alpha_spy_p3m": -0.35426367488772004, "alpha_c_p3m": -0.6151312089218232, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.3632609132000111, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5761, -0.5761, -0.5751, -0.5686, -0.5496, -0.5591, -0.5546, -0.5496, -0.5207, -0.4738, -0.4863, -0.5122, -0.4838, -0.4339, -0.4065, -0.4339, -0.4339, -0.4339, -0.4464, -0.4613, -0.4813, -0.4813, -0.4763, -0.4838, -0.4788, -0.4314, -0.4713, -0.4888, -0.4489, -0.4414, -0.4165, -0.3591, -0.2968, -0.2618, -0.2893, -0.2843, -0.3217, -0.2618, -0.2918, -0.3242, -0.3142, -0.3491, -0.3591, -0.3741, -0.3416, -0.2768, -0.2045, -0.1446, -0.2269, -0.2594, -0.2419, -0.2419, -0.2419, -0.2419, -0.2419, -0.2419, -0.2419, -0.2419, -0.2344, -0.2444, -0.1696, -0.0898, -0.0898, 0.0, 0.0998, -0.01, -0.0499, 0.01, -0.0898, -0.0998, -0.0474, -0.015, -0.0474, -0.0623, -0.0399, -0.0524, 0.015, -0.0499, -0.0698, -0.0923, -0.0648, -0.0773, -0.1022, -0.1796, -0.217, -0.1995, -0.1771, -0.1771, -0.1771, -0.192, -0.1721, -0.0898, -0.1072, -0.1022, -0.1047, -0.1222, -0.1172, -0.1322, -0.1945, -0.192, -0.2145, -0.2594, -0.2943, -0.3217, -0.3017, -0.3067, -0.3267, -0.3267, -0.2993, -0.3017, -0.3167, -0.3042, -0.3267, -0.2594, -0.1995, -0.2269, -0.212, -0.2618, -0.2219, -0.2444, -0.2544, -0.2319, -0.197, -0.2195, -0.2444, -0.2544, -0.2519, -0.2369, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010694374464844018", "ticker_idx": 0, "ts": "2026-01-12T12:44:30+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Raise PT from Won853,000 to Won1,000,000 - Key Call Buy", "tweet_type": "news_reaction", "timeline": "months", "summary": "UBS reiterates Buy on SK Hynix, raises PT to Won1m on 82% DRAM OPM forecast by 4Q26, DDR upcycle, and HBM leadership; 2026E OP 52% above consensus.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: SK Hynix\n\nSignificant profit uplift potential - PT to Won1m\n\nWe forecast 82% DRAM OPM by 4Q26\nThe memory semis industry is in the midst of an unprecedented upcycle. By 4Q26, we now forecast SK Hynix DDR ASPs to be a whisker away from the 3Q18 peak at US$0.97 per Gb. With an upcycle of that magnitude, we believe conventional memory, especially DDR (52% of 4Q25E revs.), will likely be the key profit driver in 2026 and 2027. We forecast DDR operating profit to increase 2.9x YoY in 2026 and account for 67% of total, with operating margin reaching 82% by 4Q26. We also expect SK Hynix to continue to strongly benefit from AI compute semis growth, due to its HBM leadership position (UBSe 56% bit market share in 2026). We note that in its recently published \"2026 Market Outlook\" (5 Jan 2026; link), SK Hynix referenced UBS research assessing that it will ship first to Google for TPU7e/7p, and will likely secure c70% share for HBM4 at Nvidia (Rubin) in 2026. All in all, we forecast SK Hynix to generate Won75trn in FCF in 2026, and Won98trn in 2027. We forecast its book value per share to grow 104% YoY from end 2025 to 2027. At 2.23x NTM (to March 2027E) book, SK Hynix shares should continue to re-rate to reflect longer-term economics which have fundamentally shifted.\n\nStrong uplift to DDR and NAND contract pricing in 1Q26E and beyond\nOur industry checks indicate very strong contract pricing for 1Q26E, especially DDR. We forecast SK Hynix DDR contract pricing to rise 60% QoQ in 1Q26 (was +30%), with server DDR up 65%, and Mobile DDR up 60%. We forecast NAND contract ASPs up 27% QoQ. Our checks also show strong confidence in positive momentum carrying into 2Q26E and beyond, in particular for DDR. We continue to forecast DDR contract pricing to increase til 1Q27, and NAND til 3Q26. This leads us to forecast SK Hynix DRAM operating margin to reach 77% by 4Q26, well above the prior peak of 65% in 3Q18.\n\nSignificantly upping estimates, increasing 2026E capex\nWe increase our 4Q25E OP to Won18.2trn, 12% above VA consensus of Won16.2trn, on the back of higher DRAM ASPs (+21% QoQ vs. 20% prior), NAND (+30% vs. +20%), and NAND bit shipments (+11% vs. +4%). We increase our 2026E OP by 211% to Won150.2trn, 52% above consensus of Won99trn, and 2027E by 19% to Won157.8trn, 29% above consensus of Won122.3trn. These changes lead us to raise 2026E/27E EPS by 21%, on the back of raised DDR and NAND flash contract pricing forecasts. We also increase our 2026 capex forecast to Won38trn from Won35trn.\n\nValuation: Raise PT from Won853,000 to Won1,000,000 - Key Call Buy\nWe value SK Hynix shares at 3.00x NTM (to March 2027E) book (was 2.99x) based on our 2026-30 LT ROE forecast of 34.3% (was 33.7%) and CoE of 11.5% (was 11.3%). Our revised price target also reflects our increased book value per share forecast.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006675542417608371, "ret_signed_m1d": -0.006675542417608371, "alpha_spy_m1d": -0.00510755106383487, "alpha_c_m1d": -0.0029894664745698796, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.0036860759430384915, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014686210041248593, "ret_signed_p1d": -0.014686210041248593, "alpha_spy_p1d": -0.012686734935378308, "alpha_c_p1d": -0.016913197982250994, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": 0.0022269879410024007, "ret_p1w": 0.020026710865375774, "ret_signed_p1w": 0.020026710865375774, "alpha_spy_p1w": 0.025061539025936996, "alpha_c_p1w": -0.004879885395099981, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": 0.024906596260475755, "ret_p1m": 0.16955947975267738, "ret_signed_p1m": 0.16955947975267738, "alpha_spy_p1m": 0.17393259971475683, "alpha_c_p1m": 0.1334666983792583, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": 0.03609278137341909, "ret_p3m": 0.33490196052199983, "ret_signed_p3m": 0.33490196052199983, "alpha_spy_p3m": 0.35416826910674815, "alpha_c_p3m": 0.23340707104061198, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": 0.10149488948138785, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4363, -0.3963, -0.3789, -0.3523, -0.3609, -0.3576, -0.3616, -0.3196, -0.2862, -0.3049, -0.2555, -0.2422, -0.2542, -0.1728, -0.2182, -0.2275, -0.2088, -0.2262, -0.1915, -0.1742, -0.1768, -0.1835, -0.2529, -0.1915, -0.2395, -0.2502, -0.2382, -0.3049, -0.3062, -0.3076, -0.3009, -0.2737, -0.2924, -0.2817, -0.255, -0.263, -0.2764, -0.2737, -0.2296, -0.2443, -0.2163, -0.2457, -0.2377, -0.2603, -0.2924, -0.2644, -0.263, -0.2697, -0.2256, -0.2203, -0.215, -0.215, -0.2003, -0.1455, -0.1308, -0.1308, -0.1308, -0.0961, -0.0708, -0.0307, -0.0093, 0.0093, -0.0067, 0.0, -0.0147, -0.0093, 0.0, 0.0093, 0.02, -0.008, -0.012, 0.008, 0.024, -0.0174, 0.0681, 0.1228, 0.1495, 0.2136, 0.1081, 0.2109, 0.2016, 0.1242, 0.1202, 0.1842, 0.1696, 0.1482, 0.1856, 0.1749, 0.1749, 0.1749, 0.1749, 0.1936, 0.267, 0.2697, 0.3418, 0.3591, 0.47, 0.4192, 0.4192, 0.256, 0.1356, 0.2587, 0.2359, 0.1182, 0.2546, 0.2774, 0.2439, 0.2172, 0.3028, 0.2974, 0.4125, 0.355, 0.3469, 0.248, 0.3189, 0.3309, 0.248, 0.248, 0.1677, 0.0794, 0.1998, 0.1102, 0.1717, 0.1851, 0.2252, 0.3817, 0.3349, 0.3737, 0.3911, 0.4753, 0.5195, 0.5449, 0.5088, 0.5596, 0.6372, 0.6359, 0.6385, 0.6345, 0.7281, 0.7389, 0.7295, 0.7201, 0.7201, 0.9355, 0.9355, 1.1415, 1.2124, 1.2552, 1.5146, 1.4545, 1.6431, 1.635, 1.4331, 1.4611, 1.3341, 1.3341, 1.5949, 1.5962, 1.5962, 1.7447, 2.0002, 2.0622, 2.1211, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2026812001910157793", "ticker_idx": 0, "ts": "2026-02-26T00:10:12+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron is well behind Samsung and Hynix and we believe they are effectively out of the picture for Rubin HBM4", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish MU: seen as effectively disqualified from Rubin HBM4, trailing Samsung and Hynix.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SA: Micron is well behind Samsung and Hynix and we believe they are effectively out of the picture for Rubin HBM4. We have more details on qualifications and pin speeds in the Accelerator and HBM model. https://t.co/2ysVLO4DUV\n\n---\n\nhttps://t.co/HxVTfbV1ao", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.032341901083742464, "ret_signed_m1d": -0.032341901083742464, "alpha_spy_m1d": -0.02675640015910008, "alpha_c_m1d": 0.0020019098943784286, "bench_spy_m1d": 0.005585500924642384, "bench_c_m1d": 0.03434381097812089, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007676386592602946, "ret_signed_p1d": 0.007676386592602946, "alpha_spy_p1d": 0.002874443334075605, "alpha_c_p1d": -0.006012636767887436, "bench_spy_p1d": -0.004801943258527341, "bench_c_p1d": -0.013689023360490382, "ret_p1w": -0.04454230556964511, "ret_signed_p1w": 0.04454230556964511, "alpha_spy_p1w": 0.032950890626982865, "alpha_c_p1w": 0.004106385889959263, "bench_spy_p1w": -0.011591414942662248, "bench_c_p1w": -0.04043591967968585, "ret_p1m": -0.1403888460161783, "ret_signed_p1m": 0.1403888460161783, "alpha_spy_p1m": 0.06280544973186641, "alpha_c_p1m": 0.04874056352152467, "bench_spy_p1m": -0.07758339628431188, "bench_c_p1m": -0.09164828249465362, "ret_p3m": 1.1567435769789807, "ret_signed_p3m": -1.1567435769789807, "alpha_spy_p3m": -1.0648533499896498, "alpha_c_p3m": -0.6952741942545915, "bench_spy_p3m": 0.09189022698933091, "bench_c_p3m": 0.4614693827243892, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4216, -0.4239, -0.4367, -0.4548, -0.4294, -0.4061, -0.3928, -0.3657, -0.3783, -0.42, -0.4287, -0.4407, -0.4575, -0.4021, -0.3604, -0.3347, -0.3355, -0.3104, -0.3104, -0.315, -0.2916, -0.2958, -0.3132, -0.3132, -0.241, -0.2488, -0.1736, -0.1829, -0.2131, -0.1696, -0.1677, -0.1863, -0.1978, -0.1899, -0.1271, -0.1271, -0.1217, -0.0636, -0.0433, -0.0383, -0.0637, -0.0128, 0.0475, 0.0487, -0.0016, 0.0535, 0.0093, -0.087, -0.0786, -0.0502, -0.0771, -0.1018, -0.0126, -0.0038, -0.0094, -0.0094, -0.038, 0.013, 0.0043, 0.0303, 0.013, 0.0059, 0.0323, 0.0, -0.0077, -0.007, -0.0863, -0.0356, -0.0445, -0.1089, -0.0631, -0.03, 0.0075, -0.0246, 0.0254, 0.0631, 0.111, 0.1111, 0.0691, 0.0177, -0.027, -0.0482, -0.0805, -0.1446, -0.1404, -0.2253, -0.1867, -0.1144, -0.1183, -0.1183, -0.0906, -0.091, -0.0208, 0.0147, 0.0125, 0.0269, 0.121, 0.0983, 0.1007, 0.0955, 0.0795, 0.0818, 0.1736, 0.1597, 0.1958, 0.2628, 0.214, 0.2481, 0.245, 0.3053, 0.3877, 0.5412, 0.6048, 0.5567, 0.7979, 0.9147, 0.8455, 0.9347, 0.8682, 0.7445, 0.6407, 0.6821, 0.7622, 0.8347, 0.808, 0.808, 1.1567, 1.2351, 1.2233, 1.3376, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027950257200632170", "ticker_idx": 0, "ts": "2026-03-01T03:33:13+00:00", "symbol": "9988.HK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "it's difficult to generate additional returns from the short side, given these names have already seen significant declines over recent months", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish HK/China tech (Tencent, Alibaba, HST Index) as capital rotation from Korea nears peak; bearish near-term on Samsung/Hynix on overbought/crowded signals.", "resolved_tickers": ["9988.HK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Is Korea to Blame for the Hang Seng Tech Index's Underperformance? (Haitong International Securities Strategy)\n\n- Many investors are now concerned that the Hong Kong market may be re-entering a bear market. Looking at recent trends, however, it's less that Hong Kong has underperformed on its own merits and more that other markets have simply been stronger. Since the start of 2025, foreign capital within emerging markets has been flowing back into Chinese assets, benefiting both A-shares and Hong Kong equities — with particularly strong foreign inflows into the Hong Kong market.\nHowever, since February, capital has shifted en masse out of the Hong Kong market and the Hang Seng Tech Index into Japan, Korea, and Taiwan — with a notably outsized tilt toward Korea. That said, Korea's market strength has essentially been concentrated in just two names: Samsung Electronics and SK hynix. On the surface, the Korean market as a whole appears strong, but in reality it's closer to a \"two-man show,\" with the rest of the market largely being dragged up passively.\n\n- So when could this capital dispersion trend start to ease? First, there's a possibility that the current phase of Japanese, Korean, and Taiwanese markets absorbing Hong Kong's capital flows is entering its later stages. Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded. While the market's high retail investor participation drove strong upside momentum, sentiment now appears to be near a short-term peak — suggesting the \"K-wave\" shock could begin to moderate.\n\n- On the other hand, looking at the Hong Kong market, discussions with multiple hedge funds regarding core Hang Seng Tech heavyweights like Tencent and Alibaba suggest that the current environment is one where it's difficult to generate additional returns from the short side, given these names have already seen significant declines over recent months. Tencent is currently trading at only low-teens P/E. Alibaba still possesses one of China's top-tier AI large language models, giving it genuine technology momentum — yet the market's focus remains fixated solely on traditional economy and domestic consumption recovery narratives.\n$BABA\n\n---\n\nhttps://t.co/QgP7mTP3K9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04765396107666198, "ret_signed_m1d": 0.04765396107666198, "alpha_spy_m1d": 0.04822221676086835, "alpha_c_m1d": 0.03517780612078414, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": 0.012476154955877838, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01173013868271, "ret_signed_p1d": -0.01173013868271, "alpha_spy_p1d": -0.0029158001999143313, "alpha_c_p1d": -0.0025462738783561534, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.009183864804353847, "ret_p1w": -0.05645159305568448, "ret_signed_p1w": -0.05645159305568448, "alpha_spy_p1w": -0.04463601213216961, "alpha_c_p1w": -0.04926044238613558, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.007191150669548896, "ret_p1m": -0.12756594336572658, "ret_signed_p1m": -0.12756594336572658, "alpha_spy_p1m": -0.07764620179054282, "alpha_c_p1m": -0.0736664900539935, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.05389945331173307, "ret_p3m": -0.10703806083603407, "ret_signed_p3m": -0.10703806083603407, "alpha_spy_p3m": -0.20943144000194203, "alpha_c_p3m": -0.166691454685988, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.05965339384995394, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1261, 0.132, 0.1364, 0.1246, 0.1063, 0.1232, 0.1041, 0.1298, 0.0894, 0.0572, 0.0704, 0.0565, 0.0652, 0.0733, 0.0792, 0.0711, 0.0711, 0.0711, 0.0506, 0.0594, 0.0469, 0.0469, 0.0924, 0.1202, 0.1056, 0.0696, 0.0455, 0.074, 0.1312, 0.1723, 0.239, 0.2067, 0.2185, 0.176, 0.1708, 0.1965, 0.2082, 0.2353, 0.2111, 0.2456, 0.272, 0.2705, 0.2405, 0.1972, 0.1804, 0.1694, 0.1701, 0.1364, 0.1576, 0.1767, 0.1738, 0.1628, 0.1393, 0.1342, 0.1342, 0.1342, 0.1342, 0.0784, 0.1158, 0.085, 0.0872, 0.0484, 0.0477, 0.0, -0.0117, -0.0477, -0.074, -0.0418, -0.0565, -0.0213, -0.0235, -0.0352, -0.0286, -0.0176, -0.0132, 0.0095, -0.0323, -0.0931, -0.1224, -0.0968, -0.055, -0.0982, -0.1012, -0.1166, -0.1276, -0.1004, -0.1312, -0.1312, -0.1312, -0.1312, -0.0726, -0.099, -0.0799, -0.0968, -0.0872, -0.0572, -0.0044, 0.0, 0.0044, -0.0007, -0.0359, -0.044, -0.0337, -0.0455, -0.0726, -0.0425, -0.0762, -0.0762, -0.0345, -0.0381, -0.0161, 0.033, 0.0191, -0.0183, -0.0227, -0.0264, 0.011, -0.0301, -0.0345, -0.0227, -0.033, -0.0762, -0.0689, -0.0689, -0.0645, -0.0887, -0.107, -0.1136, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027950257200632170", "ticker_idx": 1, "ts": "2026-03-01T03:33:13+00:00", "symbol": "0700.HK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Tencent is currently trading at only low-teens P/E", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish HK/China tech (Tencent, Alibaba, HST Index) as capital rotation from Korea nears peak; bearish near-term on Samsung/Hynix on overbought/crowded signals.", "resolved_tickers": ["0700.HK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Is Korea to Blame for the Hang Seng Tech Index's Underperformance? (Haitong International Securities Strategy)\n\n- Many investors are now concerned that the Hong Kong market may be re-entering a bear market. Looking at recent trends, however, it's less that Hong Kong has underperformed on its own merits and more that other markets have simply been stronger. Since the start of 2025, foreign capital within emerging markets has been flowing back into Chinese assets, benefiting both A-shares and Hong Kong equities — with particularly strong foreign inflows into the Hong Kong market.\nHowever, since February, capital has shifted en masse out of the Hong Kong market and the Hang Seng Tech Index into Japan, Korea, and Taiwan — with a notably outsized tilt toward Korea. That said, Korea's market strength has essentially been concentrated in just two names: Samsung Electronics and SK hynix. On the surface, the Korean market as a whole appears strong, but in reality it's closer to a \"two-man show,\" with the rest of the market largely being dragged up passively.\n\n- So when could this capital dispersion trend start to ease? First, there's a possibility that the current phase of Japanese, Korean, and Taiwanese markets absorbing Hong Kong's capital flows is entering its later stages. Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded. While the market's high retail investor participation drove strong upside momentum, sentiment now appears to be near a short-term peak — suggesting the \"K-wave\" shock could begin to moderate.\n\n- On the other hand, looking at the Hong Kong market, discussions with multiple hedge funds regarding core Hang Seng Tech heavyweights like Tencent and Alibaba suggest that the current environment is one where it's difficult to generate additional returns from the short side, given these names have already seen significant declines over recent months. Tencent is currently trading at only low-teens P/E. Alibaba still possesses one of China's top-tier AI large language models, giving it genuine technology momentum — yet the market's focus remains fixated solely on traditional economy and domestic consumption recovery narratives.\n$BABA\n\n---\n\nhttps://t.co/QgP7mTP3K9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0077820726542212615, "ret_signed_m1d": 0.0077820726542212615, "alpha_spy_m1d": 0.008350328338427637, "alpha_c_m1d": 0.00625496083513033, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": 0.0015271118190909316, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006809328588560648, "ret_signed_p1d": -0.006809328588560648, "alpha_spy_p1d": 0.0020050098942350214, "alpha_c_p1d": -0.007572819564360955, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": 0.0007634909758003072, "ret_p1w": 0.0038910363271105197, "ret_signed_p1w": 0.0038910363271105197, "alpha_spy_p1w": 0.015706617250625388, "alpha_c_p1w": 0.006521112741791235, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.0026300764146807154, "ret_p1m": -0.058365725100064436, "ret_signed_p1m": -0.058365725100064436, "alpha_spy_p1m": -0.008445983524880685, "alpha_c_p1m": -0.00189472999756235, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.056470995102502086, "ret_p3m": -0.16351818734083667, "ret_signed_p3m": -0.16351818734083667, "alpha_spy_p3m": -0.26591156650674463, "alpha_c_p3m": -0.1564960573360764, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": -0.007022130004760263, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1887, 0.1907, 0.1868, 0.177, 0.1722, 0.1732, 0.1702, 0.1984, 0.1732, 0.1605, 0.177, 0.177, 0.1946, 0.1955, 0.1712, 0.1722, 0.1722, 0.1722, 0.1605, 0.1673, 0.1654, 0.1654, 0.2121, 0.215, 0.2305, 0.215, 0.1984, 0.1887, 0.2121, 0.2208, 0.2315, 0.2101, 0.2014, 0.1868, 0.1693, 0.1722, 0.1625, 0.1576, 0.1663, 0.1809, 0.2082, 0.2101, 0.179, 0.1644, 0.1304, 0.0856, 0.0866, 0.0652, 0.0895, 0.072, 0.0661, 0.0418, 0.035, 0.0399, 0.0399, 0.0399, 0.0399, 0.0156, 0.0467, 0.0117, 0.0165, -0.0039, 0.0078, 0.0, -0.0068, -0.0156, -0.0233, 0.0097, 0.0039, 0.0768, 0.0739, 0.0632, 0.0652, 0.0866, 0.07, 0.071, -0.0019, -0.0117, -0.0304, 0.0, -0.0165, -0.0358, -0.0401, -0.063, -0.0584, -0.0339, -0.0482, -0.0482, -0.0482, -0.0482, -0.0117, -0.0107, -0.0185, -0.0467, -0.0405, -0.0292, 0.0058, -0.0068, 0.0165, 0.0097, -0.0195, -0.0366, -0.0401, -0.0689, -0.0782, -0.0677, -0.0899, -0.0899, -0.0798, -0.0813, -0.0992, -0.0712, -0.0829, -0.0965, -0.1105, -0.1, -0.1047, -0.1017, -0.1159, -0.0946, -0.1041, -0.136, -0.1312, -0.1312, -0.136, -0.145, -0.1635, -0.1592, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027950257200632170", "ticker_idx": 2, "ts": "2026-03-01T03:33:13+00:00", "symbol": "BABA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Alibaba still possesses one of China's top-tier AI large language models, giving it genuine technology momentum", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish HK/China tech (Tencent, Alibaba, HST Index) as capital rotation from Korea nears peak; bearish near-term on Samsung/Hynix on overbought/crowded signals.", "resolved_tickers": ["BABA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Is Korea to Blame for the Hang Seng Tech Index's Underperformance? (Haitong International Securities Strategy)\n\n- Many investors are now concerned that the Hong Kong market may be re-entering a bear market. Looking at recent trends, however, it's less that Hong Kong has underperformed on its own merits and more that other markets have simply been stronger. Since the start of 2025, foreign capital within emerging markets has been flowing back into Chinese assets, benefiting both A-shares and Hong Kong equities — with particularly strong foreign inflows into the Hong Kong market.\nHowever, since February, capital has shifted en masse out of the Hong Kong market and the Hang Seng Tech Index into Japan, Korea, and Taiwan — with a notably outsized tilt toward Korea. That said, Korea's market strength has essentially been concentrated in just two names: Samsung Electronics and SK hynix. On the surface, the Korean market as a whole appears strong, but in reality it's closer to a \"two-man show,\" with the rest of the market largely being dragged up passively.\n\n- So when could this capital dispersion trend start to ease? First, there's a possibility that the current phase of Japanese, Korean, and Taiwanese markets absorbing Hong Kong's capital flows is entering its later stages. Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded. While the market's high retail investor participation drove strong upside momentum, sentiment now appears to be near a short-term peak — suggesting the \"K-wave\" shock could begin to moderate.\n\n- On the other hand, looking at the Hong Kong market, discussions with multiple hedge funds regarding core Hang Seng Tech heavyweights like Tencent and Alibaba suggest that the current environment is one where it's difficult to generate additional returns from the short side, given these names have already seen significant declines over recent months. Tencent is currently trading at only low-teens P/E. Alibaba still possesses one of China's top-tier AI large language models, giving it genuine technology momentum — yet the market's focus remains fixated solely on traditional economy and domestic consumption recovery narratives.\n$BABA\n\n---\n\nhttps://t.co/QgP7mTP3K9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.010872636632311528, "ret_signed_m1d": 0.010872636632311528, "alpha_spy_m1d": 0.011440892316517903, "alpha_c_m1d": -0.0016035183235663109, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": 0.012476154955877838, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.048891704125053614, "ret_signed_p1d": -0.048891704125053614, "alpha_spy_p1d": -0.040077365642257945, "alpha_c_p1d": -0.03970783932069977, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.009183864804353847, "ret_p1w": -0.06958472459897513, "ret_signed_p1w": -0.06958472459897513, "alpha_spy_p1w": -0.057769143675460266, "alpha_c_p1w": -0.06239357392942624, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.007191150669548896, "ret_p1m": -0.11994948630027025, "ret_signed_p1m": -0.11994948630027025, "alpha_spy_p1m": -0.0700297447250865, "alpha_c_p1m": -0.06605003298853718, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.05389945331173307, "ret_p3m": -0.11503924086238704, "ret_signed_p3m": -0.11503924086238704, "alpha_spy_p3m": -0.217432620028295, "alpha_c_p3m": -0.17469263471234098, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.05965339384995394, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1089, 0.1044, 0.1106, 0.1092, 0.094, 0.1141, 0.1006, 0.092, 0.0528, 0.0472, 0.0318, 0.0334, 0.0507, 0.0589, 0.0608, 0.0526, 0.0526, 0.0679, 0.0416, 0.0337, 0.0282, 0.0282, 0.0925, 0.0961, 0.0585, 0.0294, 0.0835, 0.0589, 0.1666, 0.1715, 0.1918, 0.199, 0.1602, 0.1602, 0.1391, 0.1832, 0.2428, 0.2151, 0.2021, 0.2116, 0.2322, 0.2223, 0.1894, 0.1812, 0.1479, 0.1163, 0.1066, 0.1399, 0.1434, 0.168, 0.1526, 0.1134, 0.0924, 0.0924, 0.0903, 0.0927, 0.0821, 0.0834, 0.0717, 0.074, 0.0682, 0.0385, 0.0109, 0.0, -0.0489, -0.0652, -0.0856, -0.0826, -0.0696, -0.0401, -0.044, -0.0586, -0.0516, -0.041, -0.042, -0.057, -0.1239, -0.1413, -0.1157, -0.1198, -0.089, -0.1203, -0.1394, -0.1444, -0.1199, -0.1321, -0.1439, -0.1439, -0.142, -0.1602, -0.1209, -0.1044, -0.1068, -0.1021, -0.0786, -0.0651, -0.0278, -0.0109, -0.0168, -0.0504, -0.0431, -0.0762, -0.0473, -0.0704, -0.0821, -0.0851, -0.0749, -0.0776, -0.0652, -0.0723, -0.0079, -0.0109, -0.0175, -0.0369, -0.0546, 0.0228, -0.0101, -0.0699, -0.0652, -0.0485, -0.0567, -0.0778, -0.0881, -0.0881, -0.0918, -0.1038, -0.115, -0.1286, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027950257200632170", "ticker_idx": 4, "ts": "2026-03-01T03:33:13+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish HK/China tech (Tencent, Alibaba, HST Index) as capital rotation from Korea nears peak; bearish near-term on Samsung/Hynix on overbought/crowded signals.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Is Korea to Blame for the Hang Seng Tech Index's Underperformance? (Haitong International Securities Strategy)\n\n- Many investors are now concerned that the Hong Kong market may be re-entering a bear market. Looking at recent trends, however, it's less that Hong Kong has underperformed on its own merits and more that other markets have simply been stronger. Since the start of 2025, foreign capital within emerging markets has been flowing back into Chinese assets, benefiting both A-shares and Hong Kong equities — with particularly strong foreign inflows into the Hong Kong market.\nHowever, since February, capital has shifted en masse out of the Hong Kong market and the Hang Seng Tech Index into Japan, Korea, and Taiwan — with a notably outsized tilt toward Korea. That said, Korea's market strength has essentially been concentrated in just two names: Samsung Electronics and SK hynix. On the surface, the Korean market as a whole appears strong, but in reality it's closer to a \"two-man show,\" with the rest of the market largely being dragged up passively.\n\n- So when could this capital dispersion trend start to ease? First, there's a possibility that the current phase of Japanese, Korean, and Taiwanese markets absorbing Hong Kong's capital flows is entering its later stages. Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded. While the market's high retail investor participation drove strong upside momentum, sentiment now appears to be near a short-term peak — suggesting the \"K-wave\" shock could begin to moderate.\n\n- On the other hand, looking at the Hong Kong market, discussions with multiple hedge funds regarding core Hang Seng Tech heavyweights like Tencent and Alibaba suggest that the current environment is one where it's difficult to generate additional returns from the short side, given these names have already seen significant declines over recent months. Tencent is currently trading at only low-teens P/E. Alibaba still possesses one of China's top-tier AI large language models, giving it genuine technology momentum — yet the market's focus remains fixated solely on traditional economy and domestic consumption recovery narratives.\n$BABA\n\n---\n\nhttps://t.co/QgP7mTP3K9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.0005682556842063757, "alpha_c_m1d": -0.0055898388543234034, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -0.0055898388543234034, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09884526208146449, "ret_signed_p1d": 0.09884526208146449, "alpha_spy_p1d": 0.09003092359866882, "alpha_c_p1d": 0.084225843558279, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.01461941852318549, "ret_p1w": -0.1986143434258696, "ret_signed_p1w": 0.1986143434258696, "alpha_spy_p1w": 0.18679876250235472, "alpha_c_p1w": 0.2001909506044468, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": 0.0015766071785772162, "ret_p1m": -0.2261151425649689, "ret_signed_p1m": 0.2261151425649689, "alpha_spy_p1m": 0.17619540098978514, "alpha_c_p1m": 0.17971438894673752, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.04640075361823137, "ret_p3m": 0.3862351363743528, "ret_signed_p3m": -0.3862351363743528, "alpha_spy_p3m": -0.28384175720844484, "alpha_c_p3m": -0.04552758791859035, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.34070754845576245, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5197, -0.5169, -0.5017, -0.4967, -0.5017, -0.5036, -0.5068, -0.4994, -0.5183, -0.5275, -0.504, -0.5054, -0.5114, -0.4921, -0.4875, -0.4893, -0.4893, -0.4622, -0.448, -0.4462, -0.4462, -0.4462, -0.4065, -0.3621, -0.3584, -0.3487, -0.3589, -0.358, -0.3589, -0.3644, -0.352, -0.3353, -0.3122, -0.3104, -0.3293, -0.3095, -0.2965, -0.2975, -0.2975, -0.2633, -0.2499, -0.2577, -0.2587, -0.3053, -0.2263, -0.2189, -0.2642, -0.2674, -0.2314, -0.2342, -0.2249, -0.1751, -0.163, -0.163, -0.163, -0.163, -0.1224, -0.1219, -0.1085, -0.0762, -0.06, 0.0069, 0.0, 0.0, -0.0988, -0.2046, -0.115, -0.1307, -0.1986, -0.1321, -0.1224, -0.1321, -0.1524, -0.1284, -0.1044, -0.037, -0.0739, -0.079, -0.1395, -0.1238, -0.127, -0.1681, -0.1681, -0.184, -0.2261, -0.1222, -0.1743, -0.1382, -0.1062, -0.0905, -0.0257, -0.0558, -0.0465, -0.0697, -0.0442, -0.0234, 0.0067, -0.0002, -0.0072, 0.0136, 0.0067, 0.0391, 0.016, 0.0391, 0.0275, 0.046, 0.0206, 0.0206, 0.0761, 0.0761, 0.2312, 0.2566, 0.2428, 0.3214, 0.2914, 0.3145, 0.37, 0.252, 0.3006, 0.2752, 0.2775, 0.3862, 0.3538, 0.3538, 0.3839, 0.4209, 0.3862, 0.4672, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2027950257200632170", "ticker_idx": 5, "ts": "2026-03-01T03:33:13+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish HK/China tech (Tencent, Alibaba, HST Index) as capital rotation from Korea nears peak; bearish near-term on Samsung/Hynix on overbought/crowded signals.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> Is Korea to Blame for the Hang Seng Tech Index's Underperformance? (Haitong International Securities Strategy)\n\n- Many investors are now concerned that the Hong Kong market may be re-entering a bear market. Looking at recent trends, however, it's less that Hong Kong has underperformed on its own merits and more that other markets have simply been stronger. Since the start of 2025, foreign capital within emerging markets has been flowing back into Chinese assets, benefiting both A-shares and Hong Kong equities — with particularly strong foreign inflows into the Hong Kong market.\nHowever, since February, capital has shifted en masse out of the Hong Kong market and the Hang Seng Tech Index into Japan, Korea, and Taiwan — with a notably outsized tilt toward Korea. That said, Korea's market strength has essentially been concentrated in just two names: Samsung Electronics and SK hynix. On the surface, the Korean market as a whole appears strong, but in reality it's closer to a \"two-man show,\" with the rest of the market largely being dragged up passively.\n\n- So when could this capital dispersion trend start to ease? First, there's a possibility that the current phase of Japanese, Korean, and Taiwanese markets absorbing Hong Kong's capital flows is entering its later stages. Korea's market is flashing strong short-term overbought signals, and positioning has become quite crowded. While the market's high retail investor participation drove strong upside momentum, sentiment now appears to be near a short-term peak — suggesting the \"K-wave\" shock could begin to moderate.\n\n- On the other hand, looking at the Hong Kong market, discussions with multiple hedge funds regarding core Hang Seng Tech heavyweights like Tencent and Alibaba suggest that the current environment is one where it's difficult to generate additional returns from the short side, given these names have already seen significant declines over recent months. Tencent is currently trading at only low-teens P/E. Alibaba still possesses one of China's top-tier AI large language models, giving it genuine technology momentum — yet the market's focus remains fixated solely on traditional economy and domestic consumption recovery narratives.\n$BABA\n\n---\n\nhttps://t.co/QgP7mTP3K9", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.0005682556842063757, "alpha_c_m1d": -4.926186798004384e-05, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -4.926186798004384e-05, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.11498588182920777, "ret_signed_p1d": 0.11498588182920777, "alpha_spy_p1d": 0.1061715433464121, "alpha_c_p1d": 0.07726345623490394, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.03772242559430383, "ret_p1w": -0.21206408092964646, "ret_signed_p1w": 0.21206408092964646, "alpha_spy_p1w": 0.2002485000061316, "alpha_c_p1w": 0.1824865344890717, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.029577546440574753, "ret_p1m": -0.23939681952013425, "ret_signed_p1m": 0.23939681952013425, "alpha_spy_p1m": 0.1894770779449505, "alpha_c_p1m": 0.18282549659567338, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.056571322924460876, "ret_p3m": 1.157759408647605, "ret_signed_p3m": -1.157759408647605, "alpha_spy_p3m": -1.055366029481697, "alpha_c_p3m": -0.6817634553298333, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.4759959533177718, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4807, -0.4901, -0.4882, -0.4572, -0.4675, -0.4478, -0.4685, -0.4628, -0.4788, -0.5014, -0.4816, -0.4807, -0.4854, -0.4544, -0.4506, -0.4468, -0.4468, -0.4365, -0.3979, -0.3876, -0.3876, -0.3876, -0.3631, -0.3452, -0.317, -0.3019, -0.2888, -0.3001, -0.2954, -0.3057, -0.3019, -0.2954, -0.2888, -0.2813, -0.301, -0.3038, -0.2897, -0.2784, -0.3076, -0.2474, -0.2088, -0.19, -0.1448, -0.2192, -0.1467, -0.1533, -0.2079, -0.2107, -0.1655, -0.1759, -0.1909, -0.1646, -0.1721, -0.1721, -0.1721, -0.1721, -0.159, -0.1072, -0.1053, -0.0545, -0.0423, 0.0358, 0.0, 0.0, -0.115, -0.1998, -0.1131, -0.1291, -0.2121, -0.1159, -0.0999, -0.1235, -0.1423, -0.082, -0.0858, -0.0047, -0.0452, -0.0509, -0.1206, -0.0707, -0.0622, -0.1206, -0.1206, -0.1772, -0.2394, -0.1546, -0.2177, -0.1744, -0.1649, -0.1367, -0.0264, -0.0594, -0.032, -0.0198, 0.0396, 0.0707, 0.0886, 0.0631, 0.099, 0.1536, 0.1527, 0.1546, 0.1517, 0.2177, 0.2253, 0.2187, 0.2121, 0.2121, 0.3638, 0.3638, 0.509, 0.5589, 0.5891, 0.7719, 0.7295, 0.8624, 0.8567, 0.7144, 0.7342, 0.6447, 0.6447, 0.8285, 0.8294, 0.8294, 0.934, 1.114, 1.1578, 1.1992, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2037769643629396194", "ticker_idx": 0, "ts": "2026-03-28T05:51:57+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TrendForce Raises Q1/Q2 2026 DRAM & NAND Price Forecasts", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares TrendForce's sharply upward-revised DRAM and NAND price forecasts for Q1/Q2 2026, implying bullish memory pricing outlook.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TrendForce Raises Q1/Q2 2026 DRAM & NAND Price Forecasts\n\nQ1 2026\n\n- PC DRAM Blended ASP: revised up from +50~55% to +110~115%\n- Server DRAM Blended ASP: revised up from +60~65% to +93~98%\n- Mobile DRAM Blended ASP: revised up from +45~50% to +58~63%\n- Overall DRAM Blended ASP: revised up from +55~60% to +93~98%\n\n- eSSD prices: revised up from +33~38% to +75~80%\nTLC/QLC NAND prices: revised up from +30~35% to +75~80%\n- Overall NAND Blended ASP: revised up from +33~38% to +85~90%\n\nQ2 2026\n\n- PC DRAM prices: revised up from +10~15% to +40~45%\n- Server DRAM prices: revised up from +10~15% to +43~48%\n- Mobile DRAM (LP5X) prices: revised up from +13~18% to +58~63%\n\n- eSSD prices: revised up from +15~20% to +68~73%\n- TLC/QLC NAND prices: revised up from +15~20% to +60~65%\n- Overall NAND Blended ASP: revised up from +18~23% to +70~75%", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06592497225741889, "ret_signed_m1d": 0.06592497225741889, "alpha_spy_m1d": 0.06257029363244415, "alpha_c_m1d": 0.03359660974902945, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.03232836250838944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.025791107568987576, "ret_signed_p1d": -0.025791107568987576, "alpha_spy_p1d": -0.05485903326790238, "alpha_c_p1d": -0.08335874386794347, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.05756763629895589, "ret_p1w": 0.09469339478457606, "ret_signed_p1w": 0.09469339478457606, "alpha_spy_p1w": 0.052033105178314054, "alpha_c_p1w": 0.0024251347242683075, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.09226826006030775, "ret_p1m": 0.4384755364222706, "ret_signed_p1m": 0.4384755364222706, "alpha_spy_p1m": 0.3123302519405012, "alpha_c_p1m": 0.08352562025238203, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.3549499161698886, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2301, -0.2301, -0.1729, -0.151, -0.1056, -0.0996, -0.1114, -0.0969, -0.0945, -0.109, -0.1073, -0.0945, -0.0553, -0.0515, -0.0649, -0.033, -0.0132, -0.0069, -0.0296, 0.0306, 0.0776, 0.0826, 0.0788, 0.0534, 0.096, 0.0549, 0.0179, 0.0277, 0.0491, 0.0332, 0.0692, 0.104, 0.1035, 0.1035, 0.0912, 0.1131, 0.1313, 0.1637, 0.1625, 0.1932, 0.2161, 0.2612, 0.2406, 0.2409, 0.1198, 0.064, 0.1319, 0.0913, 0.0497, 0.1301, 0.1567, 0.1293, 0.1349, 0.1854, 0.2143, 0.2747, 0.2251, 0.1986, 0.1264, 0.1439, 0.1321, 0.0641, 0.0659, 0.0, -0.0258, 0.0821, 0.0336, 0.0659, 0.0947, 0.1124, 0.2137, 0.2034, 0.2173, 0.219, 0.2939, 0.3053, 0.3259, 0.3105, 0.3153, 0.3469, 0.3832, 0.3912, 0.3961, 0.4611, 0.4385, 0.458, 0.4436, 0.4696, 0.5892, 0.6552, 0.8047, 0.8147, 0.925, 1.0815, 1.0222, 1.1239, 1.1157, 0.9566, 0.9398, 0.911, 0.9463, 1.0964, 1.0721, 1.0721, 1.2768, 1.3986, 1.397, 1.4961, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2037769643629396194", "ticker_idx": 1, "ts": "2026-03-28T05:51:57+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TrendForce Raises Q1/Q2 2026 DRAM & NAND Price Forecasts", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares TrendForce's sharply upward-revised DRAM and NAND price forecasts for Q1/Q2 2026, implying bullish memory pricing outlook.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "TrendForce Raises Q1/Q2 2026 DRAM & NAND Price Forecasts\n\nQ1 2026\n\n- PC DRAM Blended ASP: revised up from +50~55% to +110~115%\n- Server DRAM Blended ASP: revised up from +60~65% to +93~98%\n- Mobile DRAM Blended ASP: revised up from +45~50% to +58~63%\n- Overall DRAM Blended ASP: revised up from +55~60% to +93~98%\n\n- eSSD prices: revised up from +33~38% to +75~80%\nTLC/QLC NAND prices: revised up from +30~35% to +75~80%\n- Overall NAND Blended ASP: revised up from +33~38% to +85~90%\n\nQ2 2026\n\n- PC DRAM prices: revised up from +10~15% to +40~45%\n- Server DRAM prices: revised up from +10~15% to +43~48%\n- Mobile DRAM (LP5X) prices: revised up from +13~18% to +58~63%\n\n- eSSD prices: revised up from +15~20% to +68~73%\n- TLC/QLC NAND prices: revised up from +15~20% to +60~65%\n- Overall NAND Blended ASP: revised up from +18~23% to +70~75%", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05718960972613507, "ret_signed_m1d": 0.05718960972613507, "alpha_spy_m1d": 0.05383493110116033, "alpha_c_m1d": 0.02486124721774563, "bench_spy_m1d": 0.0033546786249747385, "bench_c_m1d": 0.03232836250838944, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002912426122231171, "ret_signed_p1d": -0.002912426122231171, "alpha_spy_p1d": -0.03198035182114598, "alpha_c_p1d": -0.060480062421187065, "bench_spy_p1d": 0.02906792569891481, "bench_c_p1d": 0.05756763629895589, "ret_p1w": 0.1377341178771437, "ret_signed_p1w": 0.1377341178771437, "alpha_spy_p1w": 0.0950738282708817, "alpha_c_p1w": 0.04546585781683596, "bench_spy_p1w": 0.042660289606262, "bench_c_p1w": 0.09226826006030775, "ret_p1m": 0.5760884150262239, "ret_signed_p1m": 0.5760884150262239, "alpha_spy_p1m": 0.44994313054445445, "alpha_c_p1m": 0.2211384988563353, "bench_spy_p1m": 0.12614528448176943, "bench_c_p1m": 0.3549499161698886, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3509, -0.3509, -0.3034, -0.2814, -0.2254, -0.2095, -0.2201, -0.1995, -0.194, -0.1925, -0.1963, -0.1776, -0.1414, -0.1345, -0.1288, -0.0798, -0.0531, -0.0654, -0.079, -0.0291, 0.0202, 0.0302, 0.062, 0.0479, 0.1082, 0.0459, 0.0071, 0.0163, 0.0221, -0.0026, 0.0392, 0.0942, 0.1092, 0.1038, 0.0845, 0.0916, 0.1081, 0.1307, 0.1357, 0.1615, 0.165, 0.1967, 0.1782, 0.1757, 0.0712, 0.04, 0.0795, 0.0386, 0.0156, 0.0898, 0.1365, 0.106, 0.1225, 0.183, 0.196, 0.2618, 0.2282, 0.1905, 0.1357, 0.1426, 0.1403, 0.0606, 0.0572, 0.0, -0.0029, 0.109, 0.076, 0.1029, 0.1377, 0.1482, 0.2768, 0.2962, 0.329, 0.3765, 0.456, 0.4185, 0.4551, 0.413, 0.4129, 0.4508, 0.5196, 0.5139, 0.5289, 0.6049, 0.5761, 0.6094, 0.6245, 0.6602, 0.756, 0.8482, 0.9391, 0.9906, 1.1452, 1.2485, 1.1811, 1.2844, 1.2366, 1.1098, 1.1436, 1.127, 1.1667, 1.3495, 1.3332, 1.4137, 1.5454, 1.5995, 1.6239, 1.7476, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011616052858515671", "ticker_idx": 0, "ts": "2026-01-15T01:46:55+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "positive for vendors like SNDK, STX, and WDC", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein sees AI infrastructure/memory cycle intact for several quarters; names SNDK, STX, WDC, Dell, HPE, SMCI as structural beneficiaries.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Even if the scaling laws break down, hyperscalers’ capex is unlikely to collapse abruptly and would likely remain resilient for several quarters\n\nThe intelligence revolution has substantial room to run as ongoing model improvements unlock new use cases and boost productivity, implying an estimated $1.3T in enterprise willingness to pay for AI in our base case. These improvements drive both training and inference needs, structurally benefiting AI infrastructure such as servers and memory. We believe that the most watched signpost - hyperscaler capex - may be more of a lagging indicator, while consistent model improvement should be a better sign post. We see near-to-medium term data points as healthy since models keep improving, and further note that if a digestion cycle happens, there is likely time to take profits before it fully unwinds.\n\nWe believe the intelligence revolution still has substantial room to run, driven by continued model improvement that expands the set of AI-addressable tasks. Using the Anthropic Economic Index, we estimate a $16T global productivity opportunity, assume that enterprises demand a 10x ROI requirement for IT projects (in line with historical precedents), and assume 80% user penetration, which leads to a $1.3T enterprise AI willingness to pay in the base case (Exhibit 1-Exhibit 3). Importantly, this willingness to pay rises as models improve: better models address more complex tasks, unlock incremental productivity gains, justify continued investment in both training and inference, and expands willingness to pay. Based on the OSWorld Agentic Task Benchmark, we can see that model performance has been getting closer to the human benchmark (Exhibit 4), which we see as a positive leading indicator. While inference is where monetization ultimately occurs, training is strategically essential, since ongoing model improvements are required to expand addressable use cases and sustain future demand for inference and AI infrastructure.\n\nEven if a digestion happens, there is likely still time to reduce exposure before things fully unwind. If the scaling laws truly started to break down, the early indicator would be that AI labs start training new models which end up not being better than the smaller models of the previous generation. However, AI labs are unlikely to immediately publish those results. More likely, they would attempt to fix the issues by re-attempting a different run, which could take months. Even when it becomes clear that something is wrong, AI labs will be left guessing whether this is solely their own problem or an industry-wide breakdown in the scaling laws¹. Word would likely get out gradually, through a trickle of researchers exiting or complaining to the press². It will likely take months before it becomes clear that it is an industry-wide problem. Even then, VCs and hyperscalers will appear less concerned about this risk, in part because hyperscalers can redeploy excess GPUs to other workloads while VCs appear willing to accept these risks. Therefore, they are likely to be slow to react, and GPU-buying can continue for some time. On net, we believe it would likely be at least a year before it is clear that it is a problem, and hyperscaler capex would likely continue to be healthy for a few quarters after.\n\nWe view the Intelligence Revolution as structurally positive for AI server and memory vendors. As model size and complexity continue to expand, larger workloads and the growing need for nearline storage will further drive the demand for memory, which is positive for vendors like SNDK, STX, and WDC. Moreover, the expansion of AI use cases is driving growth in both training and inference over time, supporting the sustained demand for AI infrastructure including AI servers, supporting names like Dell, HPE, and SMCI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.052365343532307596, "ret_signed_m1d": -0.052365343532307596, "alpha_spy_m1d": -0.049649524892197316, "alpha_c_m1d": -0.047140488109642664, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.005224855422664931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010702778289834347, "ret_signed_p1d": 0.010702778289834347, "alpha_spy_p1d": 0.011540652280729757, "alpha_c_p1d": 0.009602897174043168, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.0010998811157911792, "ret_p1w": 0.23018281315343136, "ret_signed_p1w": 0.23018281315343136, "alpha_spy_p1w": 0.23489213002575748, "alpha_c_p1w": 0.23417019729016697, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.0039873841367356055, "ret_p1m": 0.5310331651600271, "ret_signed_p1m": 0.5310331651600271, "alpha_spy_p1m": 0.5461868638030019, "alpha_c_p1m": 0.5715942221634245, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.040561057003397405, "ret_p3m": 1.3078390296895388, "ret_signed_p3m": 1.3078390296895388, "alpha_spy_p3m": 1.3018922197022436, "alpha_c_p3m": 1.2895251478601693, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.01831388182936955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6383, -0.6352, -0.6409, -0.5918, -0.5451, -0.5687, -0.5711, -0.5006, -0.5215, -0.5129, -0.4942, -0.5246, -0.471, -0.4925, -0.4148, -0.3452, -0.3364, -0.3082, -0.4048, -0.3789, -0.3503, -0.4015, -0.399, -0.5212, -0.5106, -0.4454, -0.4612, -0.4745, -0.4745, -0.4544, -0.4864, -0.4982, -0.525, -0.4788, -0.4417, -0.4491, -0.4637, -0.431, -0.4096, -0.4962, -0.5067, -0.4885, -0.4946, -0.4637, -0.4194, -0.411, -0.4016, -0.3889, -0.3889, -0.389, -0.4032, -0.413, -0.4199, -0.4199, -0.3274, -0.3303, -0.1457, -0.1361, -0.1825, -0.0778, -0.0488, -0.0475, -0.0524, 0.0, 0.0107, 0.0107, 0.1072, 0.2249, 0.2302, 0.1578, 0.1504, 0.1764, 0.2893, 0.3178, 0.4081, 0.6255, 0.6995, 0.4284, 0.408, 0.4611, 0.4256, 0.3235, 0.4645, 0.5401, 0.531, 0.531, 0.4431, 0.4671, 0.5177, 0.5882, 0.6286, 0.5603, 0.5453, 0.593, 0.5525, 0.5128, 0.3816, 0.4638, 0.382, 0.2886, 0.4386, 0.5123, 0.6016, 0.5121, 0.6167, 0.7194, 0.7598, 0.8417, 0.8866, 0.7342, 0.7166, 0.7165, 0.6564, 0.4739, 0.5048, 0.3989, 0.5525, 0.6927, 0.7144, 0.7144, 0.7707, 0.7369, 0.9082, 1.0809, 1.0813, 1.3275, 1.3078, 1.179, 1.2468, 1.2505, 1.231, 1.2077, 1.3924, 1.2784, 1.4189, 1.6151, 1.4493, 1.6005, 1.6794, 1.9005, 2.0688, 2.4364, 2.4454, 2.2743, 2.8177, 2.7815, 2.5481, 2.5364, 2.3788, 2.4396, 2.2573, 2.3801, 2.4028, 2.7685, 2.6133, 2.6133, 2.8842, 2.8851, 3.0114, 3.1418, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011616052858515671", "ticker_idx": 1, "ts": "2026-01-15T01:46:55+00:00", "symbol": "STX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "positive for vendors like SNDK, STX, and WDC", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein sees AI infrastructure/memory cycle intact for several quarters; names SNDK, STX, WDC, Dell, HPE, SMCI as structural beneficiaries.", "resolved_tickers": ["STX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Even if the scaling laws break down, hyperscalers’ capex is unlikely to collapse abruptly and would likely remain resilient for several quarters\n\nThe intelligence revolution has substantial room to run as ongoing model improvements unlock new use cases and boost productivity, implying an estimated $1.3T in enterprise willingness to pay for AI in our base case. These improvements drive both training and inference needs, structurally benefiting AI infrastructure such as servers and memory. We believe that the most watched signpost - hyperscaler capex - may be more of a lagging indicator, while consistent model improvement should be a better sign post. We see near-to-medium term data points as healthy since models keep improving, and further note that if a digestion cycle happens, there is likely time to take profits before it fully unwinds.\n\nWe believe the intelligence revolution still has substantial room to run, driven by continued model improvement that expands the set of AI-addressable tasks. Using the Anthropic Economic Index, we estimate a $16T global productivity opportunity, assume that enterprises demand a 10x ROI requirement for IT projects (in line with historical precedents), and assume 80% user penetration, which leads to a $1.3T enterprise AI willingness to pay in the base case (Exhibit 1-Exhibit 3). Importantly, this willingness to pay rises as models improve: better models address more complex tasks, unlock incremental productivity gains, justify continued investment in both training and inference, and expands willingness to pay. Based on the OSWorld Agentic Task Benchmark, we can see that model performance has been getting closer to the human benchmark (Exhibit 4), which we see as a positive leading indicator. While inference is where monetization ultimately occurs, training is strategically essential, since ongoing model improvements are required to expand addressable use cases and sustain future demand for inference and AI infrastructure.\n\nEven if a digestion happens, there is likely still time to reduce exposure before things fully unwind. If the scaling laws truly started to break down, the early indicator would be that AI labs start training new models which end up not being better than the smaller models of the previous generation. However, AI labs are unlikely to immediately publish those results. More likely, they would attempt to fix the issues by re-attempting a different run, which could take months. Even when it becomes clear that something is wrong, AI labs will be left guessing whether this is solely their own problem or an industry-wide breakdown in the scaling laws¹. Word would likely get out gradually, through a trickle of researchers exiting or complaining to the press². It will likely take months before it becomes clear that it is an industry-wide problem. Even then, VCs and hyperscalers will appear less concerned about this risk, in part because hyperscalers can redeploy excess GPUs to other workloads while VCs appear willing to accept these risks. Therefore, they are likely to be slow to react, and GPU-buying can continue for some time. On net, we believe it would likely be at least a year before it is clear that it is a problem, and hyperscaler capex would likely continue to be healthy for a few quarters after.\n\nWe view the Intelligence Revolution as structurally positive for AI server and memory vendors. As model size and complexity continue to expand, larger workloads and the growing need for nearline storage will further drive the demand for memory, which is positive for vendors like SNDK, STX, and WDC. Moreover, the expansion of AI use cases is driving growth in both training and inference over time, supporting the sustained demand for AI infrastructure including AI servers, supporting names like Dell, HPE, and SMCI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025099904867050604, "ret_signed_m1d": -0.025099904867050604, "alpha_spy_m1d": -0.022384086226940325, "alpha_c_m1d": -0.019875049444385673, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.005224855422664931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.018450333118024353, "ret_signed_p1d": 0.018450333118024353, "alpha_spy_p1d": 0.019288207108919764, "alpha_c_p1d": 0.017350452002233174, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.0010998811157911792, "ret_p1w": 0.08182445298554075, "ret_signed_p1w": 0.08182445298554075, "alpha_spy_p1w": 0.08653376985786687, "alpha_c_p1w": 0.08581183712227636, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.0039873841367356055, "ret_p1m": 0.3298888677652003, "ret_signed_p1m": 0.3298888677652003, "alpha_spy_p1m": 0.3450425664081751, "alpha_c_p1m": 0.3704499247685977, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.040561057003397405, "ret_p3m": 0.668239681587117, "ret_signed_p3m": 0.668239681587117, "alpha_spy_p3m": 0.6622928715998218, "alpha_c_p3m": 0.6499257997577474, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.01831388182936955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3324, -0.3319, -0.3304, -0.295, -0.271, -0.2829, -0.3056, -0.1729, -0.1645, -0.2033, -0.1732, -0.2204, -0.1413, -0.1329, -0.1302, -0.0846, -0.1033, -0.118, -0.1825, -0.196, -0.1861, -0.2096, -0.1931, -0.2512, -0.2605, -0.211, -0.1846, -0.1522, -0.1522, -0.1385, -0.159, -0.169, -0.1946, -0.1729, -0.1319, -0.1113, -0.1193, -0.0693, -0.0414, -0.1044, -0.1108, -0.1028, -0.1355, -0.0908, -0.0772, -0.1193, -0.1194, -0.1094, -0.1094, -0.1065, -0.1218, -0.1256, -0.1403, -0.1403, -0.1023, -0.0952, 0.0315, -0.0376, -0.1119, -0.0509, 0.0036, -0.0059, -0.0251, 0.0, 0.0185, 0.0185, 0.0177, 0.0746, 0.0818, 0.0805, 0.1185, 0.1606, 0.3828, 0.3941, 0.2728, 0.3516, 0.3875, 0.3069, 0.2658, 0.3403, 0.3268, 0.237, 0.2714, 0.3461, 0.3299, 0.3299, 0.2985, 0.3241, 0.2768, 0.2834, 0.2719, 0.2363, 0.317, 0.2789, 0.2732, 0.1848, 0.1164, 0.1707, 0.1468, 0.1014, 0.1686, 0.1997, 0.205, 0.1675, 0.1979, 0.2449, 0.3146, 0.2699, 0.3568, 0.2838, 0.2613, 0.3267, 0.2923, 0.1846, 0.1886, 0.1334, 0.2252, 0.3232, 0.3427, 0.3427, 0.4176, 0.4658, 0.5521, 0.5661, 0.5735, 0.6052, 0.6682, 0.625, 0.6631, 0.713, 0.688, 0.751, 0.8135, 0.8377, 0.8334, 0.8634, 0.8108, 1.0118, 1.1067, 1.2733, 1.3097, 1.4112, 1.4594, 1.3969, 1.4476, 1.6082, 1.5294, 1.5561, 1.5167, 1.4877, 1.3168, 1.2934, 1.3488, 1.5346, 1.5417, 1.5417, 1.645, 1.7228, 1.7543, 1.7514, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011616052858515671", "ticker_idx": 2, "ts": "2026-01-15T01:46:55+00:00", "symbol": "WDC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "positive for vendors like SNDK, STX, and WDC", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein sees AI infrastructure/memory cycle intact for several quarters; names SNDK, STX, WDC, Dell, HPE, SMCI as structural beneficiaries.", "resolved_tickers": ["WDC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Even if the scaling laws break down, hyperscalers’ capex is unlikely to collapse abruptly and would likely remain resilient for several quarters\n\nThe intelligence revolution has substantial room to run as ongoing model improvements unlock new use cases and boost productivity, implying an estimated $1.3T in enterprise willingness to pay for AI in our base case. These improvements drive both training and inference needs, structurally benefiting AI infrastructure such as servers and memory. We believe that the most watched signpost - hyperscaler capex - may be more of a lagging indicator, while consistent model improvement should be a better sign post. We see near-to-medium term data points as healthy since models keep improving, and further note that if a digestion cycle happens, there is likely time to take profits before it fully unwinds.\n\nWe believe the intelligence revolution still has substantial room to run, driven by continued model improvement that expands the set of AI-addressable tasks. Using the Anthropic Economic Index, we estimate a $16T global productivity opportunity, assume that enterprises demand a 10x ROI requirement for IT projects (in line with historical precedents), and assume 80% user penetration, which leads to a $1.3T enterprise AI willingness to pay in the base case (Exhibit 1-Exhibit 3). Importantly, this willingness to pay rises as models improve: better models address more complex tasks, unlock incremental productivity gains, justify continued investment in both training and inference, and expands willingness to pay. Based on the OSWorld Agentic Task Benchmark, we can see that model performance has been getting closer to the human benchmark (Exhibit 4), which we see as a positive leading indicator. While inference is where monetization ultimately occurs, training is strategically essential, since ongoing model improvements are required to expand addressable use cases and sustain future demand for inference and AI infrastructure.\n\nEven if a digestion happens, there is likely still time to reduce exposure before things fully unwind. If the scaling laws truly started to break down, the early indicator would be that AI labs start training new models which end up not being better than the smaller models of the previous generation. However, AI labs are unlikely to immediately publish those results. More likely, they would attempt to fix the issues by re-attempting a different run, which could take months. Even when it becomes clear that something is wrong, AI labs will be left guessing whether this is solely their own problem or an industry-wide breakdown in the scaling laws¹. Word would likely get out gradually, through a trickle of researchers exiting or complaining to the press². It will likely take months before it becomes clear that it is an industry-wide problem. Even then, VCs and hyperscalers will appear less concerned about this risk, in part because hyperscalers can redeploy excess GPUs to other workloads while VCs appear willing to accept these risks. Therefore, they are likely to be slow to react, and GPU-buying can continue for some time. On net, we believe it would likely be at least a year before it is clear that it is a problem, and hyperscaler capex would likely continue to be healthy for a few quarters after.\n\nWe view the Intelligence Revolution as structurally positive for AI server and memory vendors. As model size and complexity continue to expand, larger workloads and the growing need for nearline storage will further drive the demand for memory, which is positive for vendors like SNDK, STX, and WDC. Moreover, the expansion of AI use cases is driving growth in both training and inference over time, supporting the sustained demand for AI infrastructure including AI servers, supporting names like Dell, HPE, and SMCI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03196764935801655, "ret_signed_m1d": -0.03196764935801655, "alpha_spy_m1d": -0.02925183071790627, "alpha_c_m1d": -0.026742793935351616, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.005224855422664931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002656547376469587, "ret_signed_p1d": -0.002656547376469587, "alpha_spy_p1d": -0.0018186733855741766, "alpha_c_p1d": -0.003756428492260766, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.0010998811157911792, "ret_p1w": 0.09540736716421105, "ret_signed_p1w": 0.09540736716421105, "alpha_spy_p1w": 0.10011668403653717, "alpha_c_p1w": 0.09939475130094666, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.0039873841367356055, "ret_p1m": 0.26780718094676303, "ret_signed_p1m": 0.26780718094676303, "alpha_spy_p1m": 0.2829608795897378, "alpha_c_p1m": 0.30836823795016044, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.040561057003397405, "ret_p3m": 0.6496860209685025, "ret_signed_p3m": 0.6496860209685025, "alpha_spy_p3m": 0.6437392109812072, "alpha_c_p3m": 0.6313721391391329, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.01831388182936955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4533, -0.4538, -0.458, -0.4344, -0.4177, -0.4301, -0.438, -0.364, -0.3786, -0.3242, -0.2891, -0.3154, -0.2797, -0.264, -0.2669, -0.2162, -0.2352, -0.2527, -0.293, -0.2899, -0.2692, -0.3123, -0.3073, -0.3691, -0.3738, -0.321, -0.3008, -0.2903, -0.2903, -0.2652, -0.2643, -0.2802, -0.3, -0.2751, -0.2396, -0.2356, -0.2367, -0.1808, -0.1571, -0.206, -0.2254, -0.214, -0.2514, -0.212, -0.1847, -0.2041, -0.1974, -0.1915, -0.1915, -0.1826, -0.191, -0.2073, -0.2244, -0.2244, -0.1549, -0.1541, -0.0122, -0.1, -0.155, -0.0974, -0.0448, -0.0365, -0.032, 0.0, -0.0027, -0.0027, 0.0039, 0.0891, 0.0954, 0.0643, 0.0844, 0.1376, 0.2593, 0.2535, 0.1267, 0.2167, 0.3068, 0.213, 0.1715, 0.2723, 0.2877, 0.1822, 0.2325, 0.2792, 0.2678, 0.2678, 0.2792, 0.3353, 0.2817, 0.2855, 0.2626, 0.2182, 0.31, 0.2708, 0.2593, 0.216, 0.1284, 0.1765, 0.1668, 0.1048, 0.1805, 0.1992, 0.2109, 0.1765, 0.2266, 0.2893, 0.4136, 0.3735, 0.4277, 0.3203, 0.3279, 0.3561, 0.334, 0.2313, 0.2403, 0.1337, 0.2185, 0.3412, 0.3287, 0.3287, 0.3701, 0.4053, 0.5261, 0.522, 0.547, 0.5773, 0.6497, 0.6442, 0.6293, 0.6781, 0.6852, 0.7289, 0.7528, 0.8159, 0.8199, 0.8051, 0.7613, 0.8593, 0.9574, 0.9438, 0.9927, 1.0958, 1.1764, 1.0897, 1.1622, 1.3236, 1.2016, 1.2257, 1.2034, 1.1713, 1.0662, 1.0532, 1.0704, 1.1913, 1.1815, 1.1815, 1.3634, 1.3902, 1.3928, 1.3929, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011616052858515671", "ticker_idx": 3, "ts": "2026-01-15T01:46:55+00:00", "symbol": "Dell", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "supporting names like Dell, HPE, and SMCI", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein sees AI infrastructure/memory cycle intact for several quarters; names SNDK, STX, WDC, Dell, HPE, SMCI as structural beneficiaries.", "resolved_tickers": ["DELL"], "resolution_method": "exchange_ticker", "resolution_reason": "Dell Technologies DELL US", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Even if the scaling laws break down, hyperscalers’ capex is unlikely to collapse abruptly and would likely remain resilient for several quarters\n\nThe intelligence revolution has substantial room to run as ongoing model improvements unlock new use cases and boost productivity, implying an estimated $1.3T in enterprise willingness to pay for AI in our base case. These improvements drive both training and inference needs, structurally benefiting AI infrastructure such as servers and memory. We believe that the most watched signpost - hyperscaler capex - may be more of a lagging indicator, while consistent model improvement should be a better sign post. We see near-to-medium term data points as healthy since models keep improving, and further note that if a digestion cycle happens, there is likely time to take profits before it fully unwinds.\n\nWe believe the intelligence revolution still has substantial room to run, driven by continued model improvement that expands the set of AI-addressable tasks. Using the Anthropic Economic Index, we estimate a $16T global productivity opportunity, assume that enterprises demand a 10x ROI requirement for IT projects (in line with historical precedents), and assume 80% user penetration, which leads to a $1.3T enterprise AI willingness to pay in the base case (Exhibit 1-Exhibit 3). Importantly, this willingness to pay rises as models improve: better models address more complex tasks, unlock incremental productivity gains, justify continued investment in both training and inference, and expands willingness to pay. Based on the OSWorld Agentic Task Benchmark, we can see that model performance has been getting closer to the human benchmark (Exhibit 4), which we see as a positive leading indicator. While inference is where monetization ultimately occurs, training is strategically essential, since ongoing model improvements are required to expand addressable use cases and sustain future demand for inference and AI infrastructure.\n\nEven if a digestion happens, there is likely still time to reduce exposure before things fully unwind. If the scaling laws truly started to break down, the early indicator would be that AI labs start training new models which end up not being better than the smaller models of the previous generation. However, AI labs are unlikely to immediately publish those results. More likely, they would attempt to fix the issues by re-attempting a different run, which could take months. Even when it becomes clear that something is wrong, AI labs will be left guessing whether this is solely their own problem or an industry-wide breakdown in the scaling laws¹. Word would likely get out gradually, through a trickle of researchers exiting or complaining to the press². It will likely take months before it becomes clear that it is an industry-wide problem. Even then, VCs and hyperscalers will appear less concerned about this risk, in part because hyperscalers can redeploy excess GPUs to other workloads while VCs appear willing to accept these risks. Therefore, they are likely to be slow to react, and GPU-buying can continue for some time. On net, we believe it would likely be at least a year before it is clear that it is a problem, and hyperscaler capex would likely continue to be healthy for a few quarters after.\n\nWe view the Intelligence Revolution as structurally positive for AI server and memory vendors. As model size and complexity continue to expand, larger workloads and the growing need for nearline storage will further drive the demand for memory, which is positive for vendors like SNDK, STX, and WDC. Moreover, the expansion of AI use cases is driving growth in both training and inference over time, supporting the sustained demand for AI infrastructure including AI servers, supporting names like Dell, HPE, and SMCI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00810632807606515, "ret_signed_m1d": -0.00810632807606515, "alpha_spy_m1d": -0.005390509435954871, "alpha_c_m1d": -0.002881472653400219, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.005224855422664931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007270553292413773, "ret_signed_p1d": 0.007270553292413773, "alpha_spy_p1d": 0.008108427283309183, "alpha_c_p1d": 0.006170672176622594, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.0010998811157911792, "ret_p1w": -0.016525164362296008, "ret_signed_p1w": -0.016525164362296008, "alpha_spy_p1w": -0.011815847489969888, "alpha_c_p1w": -0.012537780225560402, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.0039873841367356055, "ret_p1m": -0.0138392638912469, "ret_signed_p1m": -0.0138392638912469, "alpha_spy_p1m": 0.0013144347517278776, "alpha_c_p1m": 0.026721793112150505, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.040561057003397405, "ret_p3m": 0.5486979327758548, "ret_signed_p3m": 0.5486979327758548, "alpha_spy_p3m": 0.5427511227885595, "alpha_c_p3m": 0.5303840509464852, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.01831388182936955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.2314, 0.2488, 0.2546, 0.2889, 0.3258, 0.3554, 0.3779, 0.3672, 0.3456, 0.3539, 0.338, 0.2923, 0.2737, 0.2467, 0.226, 0.1925, 0.1596, 0.1759, 0.1193, 0.1178, 0.0236, 0.0253, -0.0023, -0.0189, 0.0238, 0.0632, 0.0523, 0.1137, 0.1137, 0.1144, 0.1039, 0.1361, 0.1167, 0.1615, 0.1609, 0.1734, 0.1551, 0.1752, 0.1583, 0.0862, 0.0907, 0.1178, 0.0688, 0.0274, 0.0565, 0.0581, 0.0665, 0.0729, 0.0729, 0.0801, 0.0652, 0.069, 0.052, 0.052, 0.068, 0.0364, 0.0357, 0.0034, -0.0097, 0.008, 0.0068, 0.0, -0.0081, 0.0, 0.0073, 0.0073, -0.0677, -0.0493, -0.0165, -0.0311, -0.0269, -0.0376, -0.0153, -0.0054, -0.0394, 0.0002, -0.0167, 0.0244, -0.0315, 0.016, 0.0149, 0.0577, 0.0421, -0.053, -0.0138, -0.0138, -0.0256, -0.0198, -0.0007, 0.0263, 0.0, 0.0054, 0.0364, 0.0194, 0.2429, 0.2888, 0.2186, 0.2347, 0.2298, 0.2295, 0.2297, 0.207, 0.237, 0.2583, 0.2726, 0.3139, 0.2843, 0.2524, 0.3158, 0.3234, 0.3815, 0.4849, 0.5445, 0.4758, 0.4421, 0.3821, 0.3776, 0.4217, 0.4636, 0.4636, 0.4536, 0.4915, 0.5568, 0.5231, 0.4924, 0.593, 0.5487, 0.488, 0.6207, 0.6498, 0.7143, 0.788, 0.8073, 0.7861, 0.8194, 0.8184, 0.7338, 0.7316, 0.7593, 0.7695, 0.7819, 0.8213, 1.0106, 0.9388, 1.193, 1.08, 1.0118, 1.0533, 1.0871, 1.0374, 1.0041, 0.9808, 1.0454, 1.1285, 1.4854, 1.4854, 1.5686, 1.5707, 1.6694, 2.5439, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011616052858515671", "ticker_idx": 4, "ts": "2026-01-15T01:46:55+00:00", "symbol": "HPE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "supporting names like Dell, HPE, and SMCI", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein sees AI infrastructure/memory cycle intact for several quarters; names SNDK, STX, WDC, Dell, HPE, SMCI as structural beneficiaries.", "resolved_tickers": ["HPE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Even if the scaling laws break down, hyperscalers’ capex is unlikely to collapse abruptly and would likely remain resilient for several quarters\n\nThe intelligence revolution has substantial room to run as ongoing model improvements unlock new use cases and boost productivity, implying an estimated $1.3T in enterprise willingness to pay for AI in our base case. These improvements drive both training and inference needs, structurally benefiting AI infrastructure such as servers and memory. We believe that the most watched signpost - hyperscaler capex - may be more of a lagging indicator, while consistent model improvement should be a better sign post. We see near-to-medium term data points as healthy since models keep improving, and further note that if a digestion cycle happens, there is likely time to take profits before it fully unwinds.\n\nWe believe the intelligence revolution still has substantial room to run, driven by continued model improvement that expands the set of AI-addressable tasks. Using the Anthropic Economic Index, we estimate a $16T global productivity opportunity, assume that enterprises demand a 10x ROI requirement for IT projects (in line with historical precedents), and assume 80% user penetration, which leads to a $1.3T enterprise AI willingness to pay in the base case (Exhibit 1-Exhibit 3). Importantly, this willingness to pay rises as models improve: better models address more complex tasks, unlock incremental productivity gains, justify continued investment in both training and inference, and expands willingness to pay. Based on the OSWorld Agentic Task Benchmark, we can see that model performance has been getting closer to the human benchmark (Exhibit 4), which we see as a positive leading indicator. While inference is where monetization ultimately occurs, training is strategically essential, since ongoing model improvements are required to expand addressable use cases and sustain future demand for inference and AI infrastructure.\n\nEven if a digestion happens, there is likely still time to reduce exposure before things fully unwind. If the scaling laws truly started to break down, the early indicator would be that AI labs start training new models which end up not being better than the smaller models of the previous generation. However, AI labs are unlikely to immediately publish those results. More likely, they would attempt to fix the issues by re-attempting a different run, which could take months. Even when it becomes clear that something is wrong, AI labs will be left guessing whether this is solely their own problem or an industry-wide breakdown in the scaling laws¹. Word would likely get out gradually, through a trickle of researchers exiting or complaining to the press². It will likely take months before it becomes clear that it is an industry-wide problem. Even then, VCs and hyperscalers will appear less concerned about this risk, in part because hyperscalers can redeploy excess GPUs to other workloads while VCs appear willing to accept these risks. Therefore, they are likely to be slow to react, and GPU-buying can continue for some time. On net, we believe it would likely be at least a year before it is clear that it is a problem, and hyperscaler capex would likely continue to be healthy for a few quarters after.\n\nWe view the Intelligence Revolution as structurally positive for AI server and memory vendors. As model size and complexity continue to expand, larger workloads and the growing need for nearline storage will further drive the demand for memory, which is positive for vendors like SNDK, STX, and WDC. Moreover, the expansion of AI use cases is driving growth in both training and inference over time, supporting the sustained demand for AI infrastructure including AI servers, supporting names like Dell, HPE, and SMCI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00637812662011239, "ret_signed_m1d": 0.00637812662011239, "alpha_spy_m1d": 0.009093945260222669, "alpha_c_m1d": 0.009093945260222669, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.002715818640110279, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02323464159703459, "ret_signed_p1d": -0.02323464159703459, "alpha_spy_p1d": -0.02239676760613918, "alpha_c_p1d": -0.02239676760613918, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": -0.0008378739908954103, "ret_p1w": -0.02687926039268196, "ret_signed_p1w": -0.02687926039268196, "alpha_spy_p1w": -0.02216994352035584, "alpha_c_p1w": -0.02216994352035584, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.004709316872326119, "ret_p1m": 0.03371294249838552, "ret_signed_p1m": 0.03371294249838552, "alpha_spy_p1m": 0.0488666411413603, "alpha_c_p1m": 0.0488666411413603, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.015153698642974778, "ret_p3m": 0.12199468060916585, "ret_signed_p3m": 0.12199468060916585, "alpha_spy_p3m": 0.11604787062187061, "alpha_c_p3m": 0.11604787062187061, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.005946809987295243, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0357, 0.0538, 0.0456, 0.0502, 0.0642, 0.095, 0.0982, 0.0991, 0.1054, 0.1059, 0.1095, 0.0669, 0.0737, 0.0597, 0.062, 0.0814, 0.0366, 0.0633, 0.0384, 0.0339, -0.0386, -0.0535, -0.0671, -0.0934, -0.0707, -0.0449, -0.0327, -0.025, -0.025, -0.0096, -0.0064, -0.0073, 0.0081, 0.037, 0.0565, 0.0805, 0.1217, 0.1439, 0.1113, 0.081, 0.0891, 0.1004, 0.0878, 0.0832, 0.113, 0.1253, 0.1198, 0.1134, 0.1134, 0.1157, 0.1084, 0.0966, 0.0943, 0.0943, 0.1011, 0.0993, 0.0838, 0.0219, 0.0032, 0.01, 0.0068, 0.0155, 0.0064, 0.0, -0.0232, -0.0232, -0.0711, -0.0424, -0.0269, -0.041, -0.0264, -0.0159, -0.0109, 0.0027, -0.0196, 0.0027, -0.0077, 0.0588, 0.0346, 0.0756, 0.0893, 0.092, 0.0847, 0.0114, 0.0337, 0.0337, 0.0009, -0.0182, -0.0255, -0.0264, -0.0888, -0.0774, -0.0601, -0.0506, -0.0219, 0.0082, -0.0141, -0.0182, -0.0219, -0.0374, -0.0064, -0.0387, -0.0237, -0.0169, -0.0169, -0.0046, -0.0128, -0.0191, 0.0077, -0.0118, 0.0169, 0.0959, 0.1821, 0.1495, 0.0977, 0.0367, 0.0917, 0.0995, 0.1284, 0.1284, 0.127, 0.1293, 0.1454, 0.1385, 0.1413, 0.1376, 0.122, 0.1289, 0.1871, 0.2123, 0.2751, 0.3187, 0.3072, 0.2806, 0.2912, 0.3123, 0.2816, 0.2976, 0.3192, 0.31, 0.3164, 0.3774, 0.3925, 0.3618, 0.4375, 0.4154, 0.3852, 0.4705, 0.5649, 0.5177, 0.5131, 0.4957, 0.5498, 0.5576, 0.7231, 0.7231, 0.7451, 0.7057, 0.752, 0.9735, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011616052858515671", "ticker_idx": 5, "ts": "2026-01-15T01:46:55+00:00", "symbol": "SMCI", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "supporting names like Dell, HPE, and SMCI", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Bernstein sees AI infrastructure/memory cycle intact for several quarters; names SNDK, STX, WDC, Dell, HPE, SMCI as structural beneficiaries.", "resolved_tickers": ["SMCI"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein: Even if the scaling laws break down, hyperscalers’ capex is unlikely to collapse abruptly and would likely remain resilient for several quarters\n\nThe intelligence revolution has substantial room to run as ongoing model improvements unlock new use cases and boost productivity, implying an estimated $1.3T in enterprise willingness to pay for AI in our base case. These improvements drive both training and inference needs, structurally benefiting AI infrastructure such as servers and memory. We believe that the most watched signpost - hyperscaler capex - may be more of a lagging indicator, while consistent model improvement should be a better sign post. We see near-to-medium term data points as healthy since models keep improving, and further note that if a digestion cycle happens, there is likely time to take profits before it fully unwinds.\n\nWe believe the intelligence revolution still has substantial room to run, driven by continued model improvement that expands the set of AI-addressable tasks. Using the Anthropic Economic Index, we estimate a $16T global productivity opportunity, assume that enterprises demand a 10x ROI requirement for IT projects (in line with historical precedents), and assume 80% user penetration, which leads to a $1.3T enterprise AI willingness to pay in the base case (Exhibit 1-Exhibit 3). Importantly, this willingness to pay rises as models improve: better models address more complex tasks, unlock incremental productivity gains, justify continued investment in both training and inference, and expands willingness to pay. Based on the OSWorld Agentic Task Benchmark, we can see that model performance has been getting closer to the human benchmark (Exhibit 4), which we see as a positive leading indicator. While inference is where monetization ultimately occurs, training is strategically essential, since ongoing model improvements are required to expand addressable use cases and sustain future demand for inference and AI infrastructure.\n\nEven if a digestion happens, there is likely still time to reduce exposure before things fully unwind. If the scaling laws truly started to break down, the early indicator would be that AI labs start training new models which end up not being better than the smaller models of the previous generation. However, AI labs are unlikely to immediately publish those results. More likely, they would attempt to fix the issues by re-attempting a different run, which could take months. Even when it becomes clear that something is wrong, AI labs will be left guessing whether this is solely their own problem or an industry-wide breakdown in the scaling laws¹. Word would likely get out gradually, through a trickle of researchers exiting or complaining to the press². It will likely take months before it becomes clear that it is an industry-wide problem. Even then, VCs and hyperscalers will appear less concerned about this risk, in part because hyperscalers can redeploy excess GPUs to other workloads while VCs appear willing to accept these risks. Therefore, they are likely to be slow to react, and GPU-buying can continue for some time. On net, we believe it would likely be at least a year before it is clear that it is a problem, and hyperscaler capex would likely continue to be healthy for a few quarters after.\n\nWe view the Intelligence Revolution as structurally positive for AI server and memory vendors. As model size and complexity continue to expand, larger workloads and the growing need for nearline storage will further drive the demand for memory, which is positive for vendors like SNDK, STX, and WDC. Moreover, the expansion of AI use cases is driving growth in both training and inference over time, supporting the sustained demand for AI infrastructure including AI servers, supporting names like Dell, HPE, and SMCI.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.039089041996873464, "ret_signed_m1d": -0.039089041996873464, "alpha_spy_m1d": -0.036373223356763185, "alpha_c_m1d": -0.03386418657420853, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.005224855422664931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.10944933055758566, "ret_signed_p1d": 0.10944933055758566, "alpha_spy_p1d": 0.11028720454848107, "alpha_c_p1d": 0.10834944944179448, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.0010998811157911792, "ret_p1w": 0.10299118554683573, "ret_signed_p1w": 0.10299118554683573, "alpha_spy_p1w": 0.10770050241916185, "alpha_c_p1w": 0.10697856968357133, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.0039873841367356055, "ret_p1m": 0.03806936901186053, "ret_signed_p1m": 0.03806936901186053, "alpha_spy_p1m": 0.05322306765483531, "alpha_c_p1m": 0.07863042601525794, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.040561057003397405, "ret_p3m": -0.07545884798087832, "ret_signed_p3m": -0.07545884798087832, "alpha_spy_p3m": -0.08140565796817356, "alpha_c_p3m": -0.09377272981024787, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.01831388182936955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.8708, 0.8634, 0.7845, 0.6288, 0.6414, 0.7529, 0.7797, 0.791, 0.7121, 0.7661, 0.725, 0.6111, 0.4286, 0.3712, 0.3515, 0.3661, 0.3195, 0.2886, 0.1927, 0.2379, 0.1591, 0.1863, 0.1465, 0.0727, 0.0942, 0.1326, 0.104, 0.1159, 0.1159, 0.1506, 0.1356, 0.119, 0.1448, 0.1635, 0.1791, 0.2022, 0.1903, 0.1863, 0.1564, 0.0989, 0.0663, 0.0761, 0.0143, -0.0017, 0.0574, 0.0561, 0.0455, 0.0384, 0.0384, 0.0415, 0.0224, 0.0078, -0.0051, -0.0051, 0.0523, 0.0221, 0.0381, 0.0201, 0.0163, 0.0252, 0.0238, -0.0279, -0.0391, 0.0, 0.1094, 0.1094, 0.0676, 0.0959, 0.103, 0.0775, 0.0469, 0.0605, 0.0608, 0.0238, -0.0105, 0.0099, 0.0085, 0.1475, 0.0486, 0.1686, 0.1397, 0.1329, 0.0891, 0.0343, 0.0381, 0.0381, 0.0235, 0.0099, 0.0931, 0.102, 0.0438, 0.0581, 0.1421, 0.0972, 0.101, 0.0819, 0.0428, 0.1098, 0.0959, 0.0642, 0.087, 0.0806, 0.0806, 0.0503, 0.0452, 0.0829, 0.071, 0.0316, 0.0466, -0.3022, -0.2665, -0.2444, -0.1825, -0.2451, -0.2532, -0.2842, -0.226, -0.2349, -0.2107, -0.2107, -0.2505, -0.2294, -0.2056, -0.2107, -0.1414, -0.1173, -0.0755, -0.0724, -0.0347, -0.0292, -0.0207, -0.0337, -0.0082, -0.0908, -0.0116, -0.0534, -0.0738, -0.1054, -0.0687, -0.0792, -0.051, -0.054, 0.1781, 0.1428, 0.2022, 0.1394, 0.1145, 0.0877, 0.1227, 0.0551, 0.0486, 0.0387, 0.1373, 0.1373, 0.2094, 0.2094, 0.261, 0.2981, 0.4038, 0.5666, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2016445200168800631", "ticker_idx": 0, "ts": "2026-01-28T09:36:14+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The market is way too optimistic about MediaTek's TPU volume. MediaTek's smartphone business is collapsing amid brutal price competition with Qualcomm and others.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Short MediaTek on collapsing smartphone biz + TPU volume over-optimism; long Broadcom as essential to Google's NVIDIA-alternative ambitions.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "1.  The market is way too optimistic about MediaTek’s TPU volume.\n\n2.  Google is pushing ahead with a fully in-house, completely CoT-designed 2nm TPU project, completely excluding both Broadcom and MediaTek. (WTF?)\n\n3.  Google wants to become a real alternative to NVIDIA, and for that, Broadcom is absolutely essential.\n\n4.  MediaTek’s smartphone business is collapsing amid brutal price competition with Qualcomm and others.", "reply_text": null, "parent_text": null, "quoted_text": "Gokul just kicked MediaTek’s ass again. https://t.co/eU0VnUvhXq", "parent_author": null, "ret_m1d": -0.002808988764044895, "ret_signed_m1d": 0.002808988764044895, "alpha_spy_m1d": 0.0029096688031445828, "alpha_c_m1d": -0.019705003922225628, "bench_spy_m1d": 0.00010068003909968759, "bench_c_m1d": -0.022513992686270523, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": -0.0019843824489698125, "alpha_c_p1d": 0.0021361498702181514, "bench_spy_p1d": -0.0019843824489698125, "bench_c_p1d": 0.0021361498702181514, "ret_p1w": 0.011235955056179803, "ret_signed_p1w": -0.011235955056179803, "alpha_spy_p1w": -0.024508470840020058, "alpha_c_p1w": -0.09430730244719532, "bench_spy_p1w": -0.013272515783840255, "bench_c_p1w": -0.08307134739101552, "ret_p1m": 0.0926966292134832, "ret_signed_p1m": -0.0926966292134832, "alpha_spy_p1m": -0.10149706791091229, "alpha_c_p1m": -0.10378559319361469, "bench_spy_p1m": -0.00880043869742908, "bench_c_p1m": -0.011088963980131483, "ret_p3m": 0.3679775280898876, "ret_signed_p3m": -0.3679775280898876, "alpha_spy_p3m": -0.3367688281376138, "alpha_c_p3m": -0.1528465873694198, "bench_spy_p3m": 0.031208699952273822, "bench_c_p3m": 0.2151309407204678, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.278, -0.2836, -0.2753, -0.267, -0.2891, -0.3056, -0.3111, -0.3166, -0.3139, -0.3139, -0.3221, -0.3221, -0.3552, -0.3607, -0.3469, -0.369, -0.3662, -0.3469, -0.2836, -0.2615, -0.2312, -0.2036, -0.2202, -0.2284, -0.2257, -0.2147, -0.2064, -0.2174, -0.1954, -0.2312, -0.2257, -0.2174, -0.2174, -0.2147, -0.2174, -0.2229, -0.2284, -0.234, -0.2395, -0.2395, -0.2367, -0.2174, -0.2174, -0.2119, -0.2119, -0.1899, -0.1596, -0.1657, -0.1629, -0.1882, -0.2022, -0.1882, -0.1657, -0.1573, -0.1629, -0.1545, -0.1657, -0.1657, -0.177, -0.1657, -0.0843, -0.0056, -0.0028, 0.0, 0.0, -0.0112, -0.0421, 0.0084, 0.0112, -0.0056, -0.0393, 0.0281, 0.0365, 0.0421, 0.0421, 0.0421, 0.0421, 0.0421, 0.0421, 0.0421, 0.0421, 0.0084, 0.0253, 0.0506, 0.0927, 0.0927, 0.0674, 0.0197, -0.0337, -0.0028, -0.0084, -0.0646, -0.0421, -0.0084, 0.0028, -0.0337, -0.0393, -0.0281, -0.0281, -0.0562, -0.0449, -0.0871, -0.0899, -0.0899, -0.1067, -0.1096, -0.1517, -0.1629, -0.177, -0.177, -0.177, -0.177, -0.1742, -0.1124, -0.1152, -0.1152, -0.0899, -0.0337, 0.0056, 0.0646, 0.0815, 0.0674, 0.1742, 0.2893, 0.2444, 0.368, 0.368, 0.4691, 0.4466, 0.4663, 0.4663, 0.6124, 0.7725, 0.927, 0.9213, 1.0393, 1.1798, 1.0787, 0.9635, 0.9129, 0.8315, 0.9101, 0.7725, 0.8146, 0.9944, 1.1685, 1.3848, 1.3961, 1.6067, 1.4775, 1.4213, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2016445200168800631", "ticker_idx": 1, "ts": "2026-01-28T09:36:14+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Google wants to become a real alternative to NVIDIA, and for that, Broadcom is absolutely essential.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Short MediaTek on collapsing smartphone biz + TPU volume over-optimism; long Broadcom as essential to Google's NVIDIA-alternative ambitions.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "1.  The market is way too optimistic about MediaTek’s TPU volume.\n\n2.  Google is pushing ahead with a fully in-house, completely CoT-designed 2nm TPU project, completely excluding both Broadcom and MediaTek. (WTF?)\n\n3.  Google wants to become a real alternative to NVIDIA, and for that, Broadcom is absolutely essential.\n\n4.  MediaTek’s smartphone business is collapsing amid brutal price competition with Qualcomm and others.", "reply_text": null, "parent_text": null, "quoted_text": "Gokul just kicked MediaTek’s ass again. https://t.co/eU0VnUvhXq", "parent_author": null, "ret_m1d": -0.0013503108725078095, "ret_signed_m1d": -0.0013503108725078095, "alpha_spy_m1d": -0.001450990911607497, "alpha_c_m1d": 0.021163681813762714, "bench_spy_m1d": 0.00010068003909968759, "bench_c_m1d": -0.022513992686270523, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.007532062312157306, "ret_signed_p1d": -0.007532062312157306, "alpha_spy_p1d": -0.0055476798631874935, "alpha_c_p1d": -0.009668212182375457, "bench_spy_p1d": -0.0019843824489698125, "bench_c_p1d": 0.0021361498702181514, "ret_p1w": -0.07559116250258102, "ret_signed_p1w": -0.07559116250258102, "alpha_spy_p1w": -0.06231864671874077, "alpha_c_p1w": 0.0074801848884344935, "bench_spy_p1w": -0.013272515783840255, "bench_c_p1w": -0.08307134739101552, "ret_p1m": -0.034629683162410774, "ret_signed_p1m": -0.034629683162410774, "alpha_spy_p1m": -0.025829244464981693, "alpha_c_p1m": -0.02354071918227929, "bench_spy_p1m": -0.00880043869742908, "bench_c_p1m": -0.011088963980131483, "ret_p3m": 0.25758395851974214, "ret_signed_p3m": 0.25758395851974214, "alpha_spy_p3m": 0.22637525856746832, "alpha_c_p3m": 0.04245301779927435, "bench_spy_p3m": 0.031208699952273822, "bench_c_p3m": 0.2151309407204678, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1071, 0.0859, 0.0541, 0.0752, 0.065, 0.0466, 0.0734, 0.0542, 0.0639, 0.0183, 0.0257, 0.0263, 0.0198, 0.0615, 0.0388, 0.0189, 0.132, 0.1532, 0.1908, 0.1908, 0.2069, 0.1564, 0.1428, 0.14, 0.1412, 0.1688, 0.2013, 0.2169, 0.2369, 0.2171, 0.078, 0.0178, 0.0222, -0.0235, -0.012, 0.0194, 0.0246, 0.0483, 0.051, 0.051, 0.0567, 0.0485, 0.0498, 0.0386, 0.0386, 0.0432, 0.0305, 0.0316, 0.0308, -0.0023, 0.0352, 0.0569, 0.0641, 0.02, 0.0293, 0.0554, 0.0554, -0.0019, -0.0133, -0.0233, -0.0396, -0.0252, -0.0014, 0.0, -0.0075, -0.0058, -0.0064, -0.0387, -0.0756, -0.0682, -0.001, 0.0321, 0.0216, 0.0286, -0.0062, -0.0242, -0.0242, -0.0021, 0.0008, 0.0023, -0.0018, -0.0087, -0.0233, -0.0028, -0.0346, -0.0411, -0.0433, -0.0582, -0.0471, -0.0014, -0.0083, 0.0375, 0.028, 0.025, 0.0082, -0.0332, -0.025, -0.0358, -0.0519, -0.0402, -0.0682, -0.0302, -0.0429, -0.0413, -0.0695, -0.0958, -0.1177, -0.0693, -0.0573, -0.0541, -0.0541, -0.0545, 0.0043, 0.0544, 0.0673, 0.1173, 0.142, 0.1451, 0.193, 0.1983, 0.2225, 0.2017, 0.2094, 0.271, 0.2628, 0.2713, 0.2576, 0.2023, 0.2192, 0.2553, 0.2668, 0.2525, 0.2851, 0.2794, 0.2406, 0.2931, 0.2883, 0.2609, 0.2533, 0.3225, 0.2786, 0.2651, 0.2361, 0.2563, 0.2467, 0.2454, 0.2454, 0.269, 0.2686, 0.2828, 0.3435, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2005407134931104215", "ticker_idx": 0, "ts": "2025-12-28T22:34:54+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Institutions believe this round of price increases will positively contribute to TSMC's 1Q26 results... TSMC's performance during what is traditionally a seasonal slow period could remain in line with last year or even post slight growth", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares report of TSMC consecutive annual price hikes 2026-2029 on tight advanced-node supply; institutions see positive 1Q26 contribution and resilient performance.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "TSMC to Raise Prices Again for Advanced Nodes at 3nm and Below Starting January 2026\n(Taiwan Economic Daily)\n\n• As global AI demand continues to surge, TSMC’s production capacity for advanced processes at 3nm and below remains tight, with supply still unable to catch up with demand. Accordingly, TSMC is reportedly coordinating with customers to implement consecutive annual price hikes for advanced nodes for four years, from January 1, 2026 through 2029.\n\n• Institutions believe this round of price increases will positively contribute to TSMC’s 1Q26 results. In particular, with major AI-chip customers such as Nvidia and AMD continuing to roll out new products, TSMC’s performance during what is traditionally a seasonal slow period could remain in line with last year or even post slight growth.\n\n• Industry estimates suggest the 2026 price increase will “vary by process,” but will generally fall within a range of roughly 3% to 10%. The actual increase applied will differ depending on each customer’s purchase volume and bargaining power.\n\n• While TSMC has not issued an official statement on its pricing policy, the report emphasizes that, amid supply shortages and expanding AI demand, customers are still actively seeking to secure advanced-node capacity.\n\n$TSM\n\n---\n\nhttps://t.co/aLY6PYwxDX", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006380353997420984, "ret_signed_m1d": 0.006380353997420984, "alpha_spy_m1d": 0.0028039809264113202, "alpha_c_m1d": 0.0018500172818016747, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.00453033671561931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004453138665939882, "ret_signed_p1d": -0.004453138665939882, "alpha_spy_p1d": -0.003232037931357623, "alpha_c_p1d": -0.0019820535293439967, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0024710851365958852, "ret_p1w": 0.07088257309180257, "ret_signed_p1w": 0.07088257309180257, "alpha_spy_p1w": 0.07107155720432468, "alpha_c_p1w": 0.034145511515437876, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": 0.036737061576364693, "ret_p1m": 0.1243519674885385, "ret_signed_p1m": 0.1243519674885385, "alpha_spy_p1m": 0.11324488005046551, "alpha_c_p1m": 0.006178360078785472, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.11817360740975302, "ret_p3m": 0.08676353932332881, "ret_signed_p3m": 0.08676353932332881, "alpha_spy_p3m": 0.14636693961374503, "alpha_c_p3m": 0.04110305423786231, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": 0.045660485085466496, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.044, -0.0451, -0.0316, 0.0022, -0.0255, 0.0092, -0.0061, -0.0698, 0.0038, -0.0192, 0.0099, -0.0063, -0.022, -0.0134, -0.0239, -0.0426, -0.0365, -0.0224, -0.0115, -0.0007, 0.0111, 0.0049, -0.0043, 0.0104, -0.0255, -0.0268, -0.0414, -0.0505, -0.0214, -0.035, -0.0368, -0.0647, -0.0561, -0.0654, -0.079, -0.0642, -0.0803, -0.0884, -0.0566, -0.0565, -0.039, -0.039, -0.0339, -0.0466, -0.032, -0.0208, -0.0292, -0.0232, 0.0005, 0.0056, 0.0279, 0.0131, -0.0295, -0.0438, -0.0467, -0.0796, -0.054, -0.0398, -0.0254, -0.0132, -0.007, -0.007, 0.0064, 0.0, -0.0045, 0.0099, 0.0099, 0.0621, 0.0709, 0.0881, 0.059, 0.0568, 0.0755, 0.1025, 0.1007, 0.087, 0.1353, 0.1378, 0.1378, 0.0872, 0.0837, 0.0879, 0.1128, 0.1056, 0.1244, 0.1375, 0.1284, 0.0985, 0.1344, 0.1157, 0.0825, 0.0991, 0.1593, 0.1811, 0.2027, 0.2432, 0.2232, 0.2175, 0.2175, 0.2103, 0.2038, 0.1976, 0.2314, 0.2297, 0.2819, 0.2885, 0.2522, 0.2448, 0.2266, 0.1735, 0.1878, 0.1759, 0.1262, 0.1588, 0.1534, 0.1783, 0.1189, 0.1243, 0.1306, 0.153, 0.1316, 0.129, 0.0972, 0.1279, 0.1439, 0.1589, 0.0868, 0.0889, 0.0547, 0.1262, 0.138, 0.1299, 0.1299, 0.1389, 0.1508, 0.2194, 0.218, 0.235, 0.2316, 0.266, 0.25, 0.2109, 0.2347, 0.2205, 0.2266, 0.2911, 0.2752, 0.3412, 0.3496, 0.3075, 0.3124, 0.3199, 0.3252, 0.3384, 0.3144, 0.398, 0.3802, 0.3719, 0.3481, 0.3239, 0.3323, 0.3921, 0.3475, 0.3195, 0.3084, 0.3384, 0.3568, 0.3481, 0.3481, 0.3741, 0.4088, 0.4158, 0.3945, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2023741304816189663", "ticker_idx": 0, "ts": "2026-02-17T12:48:21+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Evercore: Vera Rubin timeline pulled forward!! Some sources believe that China export bans have enabled NVDA to leverage suppliers that previously served China-bound volumes for global product development, pulling Rubin's schedule forward by 3-6 months.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Evercore bullish channel check: Vera Rubin pulled forward 3-6 months, shipments possibly starting end of 2Q26.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Evercore: Vera Rubin timeline pulled forward!!\n\nSome sources believe that China export bans have enabled NVDA to leverage suppliers that previously served China-bound volumes for global product development, pulling Rubin's schedule forward by 3-6 months.\n\nSome would even not be surprised if Rubin shipments begin by end of 2Q26.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": "Evercore out with a 4Q25 AI Channel Checks, saying $NVDA Vera Rubin appears to be ahead of schedule https://t.co/CqWOUnl5Q8", "parent_author": null, "ret_m1d": -0.011677561048771157, "ret_signed_m1d": -0.011677561048771157, "alpha_spy_m1d": -0.010066693658569048, "alpha_c_m1d": -0.012192864845109508, "bench_spy_m1d": -0.0016108673902021087, "bench_c_m1d": 0.0005153037963383511, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016272852504360502, "ret_signed_p1d": 0.016272852504360502, "alpha_spy_p1d": 0.01123509396274347, "alpha_c_p1d": 0.003905112064702454, "bench_spy_p1d": 0.005037758541617032, "bench_c_p1d": 0.012367740439658048, "ret_p1w": 0.04260152465800959, "ret_signed_p1w": 0.04260152465800959, "alpha_spy_p1w": 0.03601151483219356, "alpha_c_p1w": 0.014013284817754545, "bench_spy_p1w": 0.0065900098258160344, "bench_c_p1w": 0.028588239840255047, "ret_p1m": -0.02465396717809243, "ret_signed_p1m": -0.02465396717809243, "alpha_spy_p1m": 0.006714522613787244, "alpha_c_p1m": 0.009308379773459885, "bench_spy_p1m": -0.03136848979187967, "bench_c_p1m": -0.033962346951552314, "ret_p3m": 0.21820943843924767, "ret_signed_p3m": 0.21820943843924767, "alpha_spy_p3m": 0.13277529878789873, "alpha_c_p3m": -0.1470086262052055, "bench_spy_p3m": 0.08543413965134894, "bench_c_p3m": 0.36521806464445317, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0235, -0.033, -0.0131, -0.0387, -0.0255, -0.0255, -0.0431, -0.0274, -0.019, -0.0291, -0.0086, -0.0138, 0.0031, 0.0, -0.0064, -0.0218, -0.0538, -0.0469, -0.0392, -0.0759, -0.0586, -0.0215, -0.0069, 0.0229, 0.0197, 0.0197, 0.0301, 0.0176, 0.0139, 0.0083, 0.0083, 0.021, 0.017, 0.0123, 0.0224, 0.0004, -0.0006, -0.0002, 0.0045, -0.0099, 0.0112, 0.0068, 0.0068, -0.0373, -0.0089, -0.0007, 0.0146, 0.0081, 0.0192, 0.0354, 0.0408, 0.0333, 0.0035, -0.025, -0.0583, -0.0708, 0.0024, 0.0274, 0.0193, 0.0275, 0.0107, -0.0117, -0.0117, 0.0, 0.0163, 0.0158, 0.0262, 0.0356, 0.0426, 0.0573, -0.0004, -0.0421, -0.0135, -0.0266, -0.0104, -0.0088, -0.0387, -0.0125, -0.0011, 0.0058, -0.0098, -0.0255, -0.0094, -0.0164, -0.0247, -0.0346, -0.0663, -0.0504, -0.0528, -0.034, -0.0742, -0.0943, -0.107, -0.0571, -0.0498, -0.0409, -0.0409, -0.0396, -0.0371, -0.0156, -0.0057, 0.0198, 0.0235, 0.0624, 0.0752, 0.0724, 0.0904, 0.0925, 0.0807, 0.0948, 0.0794, 0.126, 0.1711, 0.1525, 0.1313, 0.079, 0.0729, 0.0731, 0.0624, 0.1236, 0.1435, 0.1635, 0.1864, 0.1937, 0.221, 0.2745, 0.2182, 0.202, 0.1927, 0.2082, 0.1868, 0.1642, 0.1642, 0.1617, 0.1494, 0.1584, 0.1415, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2022131245552107624", "ticker_idx": 0, "ts": "2026-02-13T02:10:32+00:00", "symbol": "6754.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Anritsu (6754T) in particular caught my eye as especially intriguing", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Anritsu as a standout Japanese player in optics, flagged as especially intriguing for sector investment.", "resolved_tickers": ["6754.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Scientific & Technical Instruments'", "bench_c_industry": "Scientific & Technical Instruments", "unit_kind": "quote", "lang": "en", "tweet_text": "People often say that optics will be the next big money-making field after memory, right?\n\nThis is an interesting article introducing noteworthy Japanese players in optics that you should pay attention to when considering investments in the sector.\n\nAmong the companies featured here, Anritsu (6754T) in particular caught my eye as especially intriguing.", "reply_text": null, "parent_text": null, "quoted_text": "https://t.co/LLKaGrwD1X", "parent_author": null, "ret_m1d": 0.02897021990014892, "ret_signed_m1d": 0.02897021990014892, "alpha_spy_m1d": 0.029674209443543154, "alpha_c_m1d": 0.031477979746346696, "bench_spy_m1d": -0.0007039895433942345, "bench_c_m1d": -0.002507759846197777, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05984355331906244, "ret_signed_p1d": 0.05984355331906244, "alpha_spy_p1d": 0.05984355331906244, "alpha_c_p1d": 0.05984355331906244, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.10425033556269492, "ret_signed_p1w": 0.10425033556269492, "alpha_spy_p1w": 0.0929851562931312, "alpha_c_p1w": 0.0947919385628988, "bench_spy_p1w": 0.011265179269563719, "bench_c_p1w": 0.009458396999796115, "ret_p1m": 0.22161133339323946, "ret_signed_p1m": 0.22161133339323946, "alpha_spy_p1m": 0.24026912104191256, "alpha_c_p1m": 0.22720026176707298, "bench_spy_p1m": -0.018657787648673096, "bench_c_p1m": -0.005588928373833513, "ret_p3m": 0.7907263944781215, "ret_signed_p3m": 0.7907263944781215, "alpha_spy_p3m": 0.6989225491978228, "alpha_c_p3m": 0.5219538284162526, "bench_spy_p3m": 0.09180384528029872, "bench_c_p3m": 0.26877256606186894, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.044, -0.0649, -0.0423, -0.077, -0.077, -0.0391, -0.0125, 0.0055, 0.0025, 0.0042, -0.0307, -0.0186, -0.0123, -0.0406, -0.0165, -0.0146, -0.04, -0.0584, -0.04, -0.0385, -0.0613, -0.066, -0.0765, -0.0685, -0.055, -0.0398, -0.0527, -0.0455, -0.0417, -0.037, -0.0508, -0.0508, -0.0508, -0.0508, -0.0324, -0.0436, -0.0429, -0.0429, -0.0603, -0.0603, -0.0385, -0.014, 0.0133, 0.0192, 0.0207, -0.0034, -0.0118, -0.0044, -0.0015, -0.0338, -0.0165, -0.0262, 0.0017, -0.0939, -0.0869, -0.0209, -0.0093, 0.0025, 0.0173, 0.0438, 0.0605, 0.0605, 0.029, 0.0, 0.0598, 0.0658, 0.0634, 0.0636, 0.1043, 0.1043, 0.2197, 0.3123, 0.2683, 0.2802, 0.3047, 0.2584, 0.207, 0.2793, 0.2461, 0.1417, 0.1909, 0.239, 0.214, 0.2227, 0.2216, 0.1895, 0.2241, 0.1874, 0.1874, 0.1239, 0.1984, 0.2616, 0.2296, 0.1893, 0.2081, 0.1704, 0.2767, 0.2446, 0.2833, 0.2842, 0.2884, 0.4103, 0.4009, 0.4526, 0.45, 0.4517, 0.4547, 0.4423, 0.4611, 0.4817, 0.5654, 0.5539, 0.527, 0.5364, 0.6052, 0.6988, 0.6988, 0.7266, 0.6492, 0.6492, 0.6492, 0.6492, 0.7484, 0.8095, 0.7706, 0.7168, 0.7907, 0.7715, 0.8121, 0.7976, 0.7715, 0.8728, 0.91, 0.9668, 1.0228, 0.9963, 0.9724, 0.9579, 0.9438, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2034798222783529090", "ticker_idx": 0, "ts": "2026-03-20T01:04:35+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We remain confident in the DRAM upcycle, and our US$623 target price continues to be based on 4x the average FY2026/FY2027 book value per share", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Securities reiterates Buy on MU, raises PT to $623; confident in DRAM upcycle extending through 1H27, strong HBM ramp, and rising ASPs.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Communications | Micron FY2Q26 Review: No Need to Worry About the CAPEX Increase\n\n☀️ Results significantly beat expectations; target price raised to US$623: Micron delivered a very strong performance in fiscal 2Q26, with revenue reaching US$23.9 billion and non-GAAP EPS of US$12.20.\n\n☀️ Fiscal 3Q26 guidance came in well above expectations: midpoint guidance is US$33.5 billion in revenue and US$19.15 in EPS (vs. around US$15 in buy-side checks and GF’s forecast of US$19.78).\n\n☀️ Micron also announced that it began volume shipments of fourth-generation HBM to NVIDIA in calendar 1Q26, with yield maturation progressing faster than for HBM3E.\n\n☀️ The stock fell 4% after hours, mainly because the market is concerned that Micron’s capacity expansion plans may be more aggressive than expected.\n\n☀️ We remain confident in the DRAM upcycle, and our US$623 target price continues to be based on 4x the average FY2026/FY2027 book value per share.\n\n☀️ The DRAM upcycle is continuing, so there is no need for excessive concern going forward: on the one hand, management’s comments on the earnings call clearly supported our view that the DRAM price upcycle will extend through 1H27.\n\n☀️ Demand from both AI and traditional servers remains strong, while incremental supply in 1H27 will remain limited. Wafer output from new capacity will be concentrated in 2H27, and will only be sufficient to meet 50% to two-thirds of customer demand.\n\n☀️ On the other hand, although the market is concerned about capacity expansion beyond 2027, we remain optimistic on the long-term memory cycle.\n\n☀️ In 2026, we expect pure AI server memory — including HBM and LPDDR5X — to account for roughly 50% of total DRAM wafer demand, supported by the fact that the HBM-to-conventional-DRAM content ratio reaches 3.5x–3.8x.\n\n☀️ Entering 2027, as next-generation products transition to 1TB HBM4E and CPU-side LPDDR5X memory usage increases, AI memory demand is likely to double.\n\n☀️ In 2028, agentic AI could further boost memory demand by increasing requirements for CPU memory and workflow-related memory.\n\n☀️ In addition, newly signed long-term agreements should provide earnings support even during a downcycle.\n\n☀️ Pricing remained strong in fiscal 2Q26: according to market research data, average contract prices for conventional DRAM rose 90–95% in fiscal 1Q26, and we expect contract prices to increase by at least 50% in fiscal 2Q26.\n\n☀️ For example, Micron’s 64GB DDR5 server DRAM was priced at around US$800 in fiscal 1Q26, while current negotiations for fiscal 2Q26 indicate pricing of at least US$1,250, implying a sequential increase of more than 50%.\n\n☀️ Even after these price increases, Micron’s pricing still remains below that of SK Hynix and Samsung, leaving room for further hikes.\n\n☀️ Meanwhile, as the share of server products in the mix continues to rise, we also expect blended ASPs to trend higher.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05053203316717414, "ret_signed_m1d": 0.05053203316717414, "alpha_spy_m1d": 0.03598775376493224, "alpha_c_m1d": 0.024072545066761375, "bench_spy_m1d": 0.0145442794022419, "bench_c_m1d": 0.026459488100412765, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04386378691875292, "ret_signed_p1d": -0.04386378691875292, "alpha_spy_p1d": -0.05436380616369074, "alpha_c_p1d": -0.061070199782554124, "bench_spy_p1d": 0.01050001924493782, "bench_c_p1d": 0.0172064128638012, "ret_p1w": -0.15530854978223718, "ret_signed_p1w": -0.15530854978223718, "alpha_spy_p1w": -0.1329825397829627, "alpha_c_p1w": -0.1280434083863955, "bench_spy_p1w": -0.022326009999274476, "bench_c_p1w": -0.027265141395841686, "ret_p1m": 0.060790723610155695, "ret_signed_p1m": 0.060790723610155695, "alpha_spy_p1m": -0.031951713905671975, "alpha_c_p1m": -0.1451145714761286, "bench_spy_p1m": 0.09274243751582767, "bench_c_p1m": 0.2059052950862843, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.347, -0.3224, -0.3224, -0.3269, -0.3039, -0.308, -0.3251, -0.3251, -0.2541, -0.2619, -0.1879, -0.1971, -0.2267, -0.184, -0.1821, -0.2004, -0.2118, -0.204, -0.1422, -0.1422, -0.1369, -0.0799, -0.0599, -0.055, -0.0799, -0.0299, 0.0293, 0.0305, -0.019, 0.0352, -0.0082, -0.1029, -0.0946, -0.0667, -0.0932, -0.1174, -0.0297, -0.0211, -0.0266, -0.0266, -0.0547, -0.0046, -0.0131, 0.0125, -0.0046, -0.0116, 0.0144, -0.0174, -0.0249, -0.0242, -0.1022, -0.0523, -0.0611, -0.1244, -0.0794, -0.0468, -0.01, -0.0415, 0.0076, 0.0447, 0.0917, 0.0918, 0.0505, 0.0, -0.0439, -0.0647, -0.0965, -0.1595, -0.1553, -0.2387, -0.2008, -0.1298, -0.1336, -0.1336, -0.1064, -0.1068, -0.0378, -0.0029, -0.005, 0.0091, 0.1016, 0.0793, 0.0816, 0.0765, 0.0608, 0.0631, 0.1532, 0.1396, 0.1751, 0.2409, 0.193, 0.2265, 0.2234, 0.2827, 0.3637, 0.5145, 0.5769, 0.5297, 0.7667, 0.8814, 0.8134, 0.9011, 0.8357, 0.7143, 0.6123, 0.653, 0.7316, 0.8028, 0.7766, 0.7766, 1.1193, 1.1963, 1.1847, 1.297, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2005866451304079476", "ticker_idx": 0, "ts": "2025-12-30T05:00:03+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the export of NVIDIA's H200 chipset will be a major variable shaking up the landscape of next year's Chinese AI accelerator market... The resumption of H200 exports can serve as a containment measure for NVIDIA to build an ecosystem defense line... NVIDIA to recover a meaningful market share in a short period", "tweet_type": "analysis", "timeline": "months", "summary": "H200 re-entry into China gives NVIDIA opportunity to recover 10–25% lost market share while maintaining tech dominance via Blackwell/Rubin; bullish for NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Research] Resumption of NVIDIA H200 Exports to China: Shifting Market Landscape and Strategic Implications\n\nHong Kong-based research firm Counterpoint Research, through its latest report, projected that the export of NVIDIA's H200 chipset will be a major variable shaking up the landscape of next year's Chinese AI accelerator market. The report analyzed that the H200's entry into the Chinese market offers NVIDIA an opportunity to recover 10–25% of its lost market share, while simultaneously establishing a new strategic balance point in the US-China technological hegemony competition.\n\n1. US Confidence and NVIDIA's Pragmatic RecoveryThe permission for NVIDIA H200 exports to China is not merely a relaxation of regulations but is underpinned by the United States' sophisticated strategic judgment.\n\nFirst is commercial recovery and containment. As CEO Jensen Huang mentioned, due to strong export controls, NVIDIA's market share in China plummeted from 95% to nearly 0%. The resumption of H200 exports can serve as a containment measure for NVIDIA to build an ecosystem defense line against rapidly emerging local Chinese chipmakers and to slow down China's 'AI chip self-sufficiency' through a sophisticated tiering strategy.\n\nSecond is the US's confidence regarding the technological gap. While the H200 remains a powerful chip, from the US perspective, it is already considered 'previous generation' technology. The next-generation flagship, Blackwell (B200), boasts three times the throughput of the H200, and Rubin, scheduled for 2026, will widen that gap even further. In other words, the US intends to maintain market control by releasing older models (H200) to China while maintaining technological dominance by monopolizing cutting-edge chips (B200, Rubin).\n\n2. Persistent Performance Gap: H200 vs. Chinese Local ChipsThe issue is that even the H200, treated as 'outdated,' overwhelms competitors within China. Despite being two generations old, the H200 holds a clear performance advantage over Chinese-made chips.\n\nEven Huawei's '910C', China's strongest contender, offers only about 76% of the computing performance and two-thirds of the memory bandwidth compared to the H200. The gap with other companies like Cambricon or Hygon is even larger. Even bearing the price premium, the proven H200 is an irresistible option for Chinese companies that prioritize 'Time to Market.' Depending on the scope permitted by Chinese authorities, this will be the driving force for NVIDIA to recover a meaningful market share in a short period.\n\n3. 'Win-Win' Scenario: A Duet of Coexistence and CompetitionCounterpoint Research views a scenario where NVIDIA and the Chinese ecosystem coexist for the time being as the most likely outcome.\n\n1) Dual AI Ecosystem: NVIDIA for Training, Local for InferenceChina does not need to win on every front. A pragmatic division of labor will likely form where the 'Inference' market—characterized by high cost sensitivity and massive volume—is dominated by competitive domestic chips, while the 'Training' sector—with high technological barriers—utilizes NVIDIA H200s. This is a realistic alternative to secure the competitiveness of domestic AI models while maintaining US dependency in cutting-edge areas at a manageable level.\n\n2) H200: Not Dependence, but a 'Target to Overcome'The influx of H200s will not dampen China's will for self-reliance. Rather, the H200 acts as a clear and concrete 'performance benchmark' for Chinese chip manufacturers. The appearance of a tangible competitor, rather than an insurmountable wall, will serve as a catalyst providing stronger feedback loops and development motivation for Chinese companies.\n\n3) ASIC-Centric Differentiation StrategyIf it is difficult to break through NVIDIA's GPU architecture and CUDA ecosystem head-on, China is likely to detour via ASICs (Application-Specific Integrated Circuits). As model architectures stabilize, ASICs, which are advantageous in terms of power efficiency and cost, demonstrate their strength. China's rapid manufacturing cycle provides a structural advantage for companies like Huawei and Cambricon to develop dedicated accelerators optimized for LLMs. However, restricted access to TSMC and yield issues at SMIC remain challenges to be solved.\n\n4) Expansion into Full-Stack CompetitionUltimately, victory or defeat is determined not by a single chip but by the ecosystem. NVIDIA's strength comes from vertical integration combining hardware, CUDA, and developer tools. China also aims to build an independent full-stack ranging from computation to frameworks, models, and applications. This will take a long time, but once completed, it will be a decisive turning point to lower US dependency to the extent that individual chip performance gaps become meaningless.\n\n$NVDA\n\n---\n\nhttps://t.co/xxnmjcLKpw", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0036258876760828507, "ret_signed_m1d": 0.0036258876760828507, "alpha_spy_m1d": 0.002403294031503078, "alpha_c_m1d": 0.0011486811512630979, "bench_spy_m1d": 0.0012225936445797725, "bench_c_m1d": 0.002477206524819753, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005545422619901985, "ret_signed_p1d": -0.005545422619901985, "alpha_spy_p1d": 0.0018635447534887106, "alpha_c_p1d": 0.003207412967201284, "bench_spy_p1d": -0.007408967373390696, "bench_c_p1d": -0.00875283558710327, "ret_p1w": -0.0015995988919732529, "ret_signed_p1w": -0.0015995988919732529, "alpha_spy_p1w": -0.008586306437667224, "alpha_c_p1w": -0.06851214663895089, "bench_spy_p1w": 0.006986707545693971, "bench_c_p1w": 0.06691254774697764, "ret_p1m": 0.02122220779309547, "ret_signed_p1m": 0.02122220779309547, "alpha_spy_p1m": 0.008980859754430082, "alpha_c_p1m": -0.125539531995305, "bench_spy_p1m": 0.01224134803866539, "bench_c_p1m": 0.14676173978840046, "ret_p3m": -0.106702172082634, "ret_signed_p3m": -0.106702172082634, "alpha_spy_p3m": -0.03219335647334509, "alpha_c_p3m": -0.1368142045478069, "bench_spy_p3m": -0.0745088156092889, "bench_c_p3m": 0.03011203246517291, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0071, 0.0004, -0.0107, -0.0134, 0.0083, 0.0268, -0.0234, 0.0041, -0.0401, -0.0412, -0.0306, -0.0231, -0.0262, -0.0341, -0.0388, -0.0287, -0.0069, 0.021, 0.0719, 0.1039, 0.0818, 0.0797, 0.1031, 0.0594, 0.0408, 0.0028, 0.0032, 0.0613, 0.0299, 0.0333, -0.0037, 0.014, -0.0051, -0.033, -0.0055, -0.0368, -0.0462, -0.0267, -0.0519, -0.0389, -0.0389, -0.0563, -0.0407, -0.0325, -0.0424, -0.0222, -0.0274, -0.0106, -0.0137, -0.02, -0.0352, -0.0668, -0.06, -0.0524, -0.0885, -0.0715, -0.0349, -0.0205, 0.0089, 0.0057, 0.0057, 0.0159, 0.0036, 0.0, -0.0055, -0.0055, 0.007, 0.0031, -0.0016, 0.0084, -0.0133, -0.0143, -0.0139, -0.0092, -0.0235, -0.0026, -0.007, -0.007, -0.0505, -0.0225, -0.0144, 0.0007, -0.0057, 0.0052, 0.0212, 0.0265, 0.0191, -0.0103, -0.0384, -0.0712, -0.0835, -0.0114, 0.0133, 0.0053, 0.0134, -0.0032, -0.0252, -0.0252, -0.0137, 0.0023, 0.0019, 0.0122, 0.0214, 0.0283, 0.0428, -0.0141, -0.0552, -0.027, -0.0399, -0.024, -0.0224, -0.0518, -0.0261, -0.0148, -0.008, -0.0234, -0.0388, -0.023, -0.0299, -0.038, -0.0478, -0.0791, -0.0634, -0.0657, -0.0472, -0.0869, -0.1067, -0.1192, -0.07, -0.0628, -0.0541, -0.0541, -0.0527, -0.0503, -0.0291, -0.0193, 0.0059, 0.0095, 0.0479, 0.0605, 0.0577, 0.0755, 0.0775, 0.0659, 0.0798, 0.0646, 0.1106, 0.1551, 0.1367, 0.1158, 0.0642, 0.0582, 0.0584, 0.0478, 0.1083, 0.1278, 0.1476, 0.1702, 0.1773, 0.2042, 0.2571, 0.2015, 0.1855, 0.1764, 0.1917, 0.1705, 0.1482, 0.1482, 0.1457, 0.1337, 0.1425, 0.1259, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2005158863021260887", "ticker_idx": 0, "ts": "2025-12-28T06:08:21+00:00", "symbol": "Samsung Electro-Mechanics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think Samsung Electro-Mechanics is still undervalued. I believe Samsung Electro-Mechanics will deliver explosive growth next year.", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish on Samsung Electro-Mechanics: undervalued MLCC/FC-BGA supplier to ASIC players, expects explosive growth next year.", "resolved_tickers": ["009150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electro-Mechanics listed on KOSPI as 009150.KS", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "I think Samsung Electro-Mechanics is still undervalued.\n\nA lot of people still see it as “just an MLCC company,” but in reality it supplies FC-BGA substrates to most ASIC players outside of NVIDIA.\n\nI believe Samsung Electro-Mechanics will deliver explosive growth next year. https://t.co/g8FjxUWOSN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0033009708737864463, "ret_signed_m1d": -0.0033009708737864463, "alpha_spy_m1d": -0.0068773439447961104, "alpha_c_m1d": -0.007825542597412638, "bench_spy_m1d": 0.003576373071009664, "bench_c_m1d": 0.004524571723626192, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009708737864077666, "ret_signed_p1d": -0.009708737864077666, "alpha_spy_p1d": -0.008487637129495407, "alpha_c_p1d": -0.006555258002349462, "bench_spy_p1d": -0.0012211007345822589, "bench_c_p1d": -0.0031534798617282034, "ret_p1w": 0.04854368932038833, "ret_signed_p1w": 0.04854368932038833, "alpha_spy_p1w": 0.048732673432910434, "alpha_c_p1w": 0.05711303964111447, "bench_spy_p1w": -0.0001889841125221059, "bench_c_p1w": -0.008569350320726143, "ret_p1m": 0.07572815533980592, "ret_signed_p1m": 0.07572815533980592, "alpha_spy_p1m": 0.06462106790173294, "alpha_c_p1m": 0.06078334830702947, "bench_spy_p1m": 0.011107087438072982, "bench_c_p1m": 0.014944807032776453, "ret_p3m": 0.7747572815533981, "ret_signed_p3m": 0.7747572815533981, "alpha_spy_p3m": 0.8343606818438143, "alpha_c_p3m": 0.8652848497584366, "bench_spy_p3m": -0.05960340029041622, "bench_c_p3m": -0.09052756820503849, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2561, -0.2438, -0.2438, -0.2438, -0.2438, -0.2438, -0.2438, -0.1899, -0.2169, -0.2373, -0.2303, -0.2246, -0.2034, -0.1822, -0.1688, -0.1842, -0.1457, -0.1457, -0.1207, -0.1072, -0.1072, -0.1341, -0.0572, -0.0649, -0.0803, -0.1168, -0.1072, -0.1438, -0.138, -0.1245, -0.1284, -0.1476, -0.1803, -0.1707, -0.215, -0.1726, -0.1592, -0.163, -0.1303, -0.0706, -0.011, -0.0187, -0.0341, -0.0379, -0.0302, 0.0044, 0.0198, 0.0121, 0.0082, 0.0371, 0.0333, 0.0371, 0.0467, 0.014, -0.0187, -0.0052, -0.0225, -0.0264, -0.0206, -0.0322, -0.0245, -0.0245, -0.0033, 0.0, -0.0097, -0.0097, -0.0097, 0.0485, 0.0485, 0.0408, 0.0388, 0.0563, 0.033, 0.0835, 0.1223, 0.1049, 0.1243, 0.1243, 0.134, 0.0951, 0.0874, 0.099, 0.0563, 0.068, 0.0757, 0.0874, 0.0854, 0.0835, 0.099, 0.1942, 0.1845, 0.1262, 0.1029, 0.1806, 0.2175, 0.2019, 0.2408, 0.2019, 0.2019, 0.2019, 0.2019, 0.3903, 0.4641, 0.6563, 0.732, 0.8058, 0.8117, 0.7417, 0.7417, 0.5883, 0.4078, 0.5592, 0.5709, 0.4835, 0.5612, 0.5767, 0.5515, 0.5553, 0.6039, 0.734, 0.8019, 0.8621, 0.8039, 0.6602, 0.6971, 0.765, 0.7748, 0.7748, 0.666, 0.5825, 0.7243, 0.6194, 0.7709, 0.7942, 0.7748, 0.9961, 1.0039, 1.1942, 1.2058, 1.268, 1.3845, 1.4816, 1.6369, 1.6408, 1.9981, 2.1534, 2.0058, 2.0602, 2.0835, 2.2583, 2.2117, 2.2311, 2.2311, 2.565, 2.565, 2.5417, 2.5612, 2.5495, 2.4951, 2.7204, 2.9961, 2.9767, 2.9223, 3.0039, 2.833, 3.1204, 3.6757, 4.2039, 4.2039, 5.1049, 5.3301, 6.1806, 7.2602, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022241673531769248", "ticker_idx": 0, "ts": "2026-02-13T09:29:21+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we anticipate SEC to benefit from high-speed HBM4 market and gain market share in HBM4. We raise our HBM ASP forecast for SEC, and raise 2026F HBM shipment growth from +112% y-y to +144% y-y", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura raises Samsung OP forecasts sharply on HBM4 share gains, pricing premium 30-40%, and commodity DRAM/NAND price beats; 2026/27F OP KRW243tn/322tn.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura on Samsung Electronics: Raised 2026/27 operating profit (OP) forecasts to $168.2bn / $222.9bn\n\n\"We estimate that the rise in commodity memory prices significantly outpaced our expectation in 1Q26. We estimate that commodity DRAM/NAND prices rose by 90%/60% in 1Q26, which strongly beat our previous forecasts (DRAM +56%/NAND +40% q-q). We believe Samsung Electronics’ (SEC) HBM4 has better performance but higher production costs vs. competitors’, as the company fabricates core dies through 1C node and adopts base dies from 4nm-node foundry. While some of the customers require speed faster than 11.7Gbps, supplies from HBM suppliers are likely to be limited; thus, we see high potential for pricing premium of 30-40%. (Since low-speed HBM4 is also in tight supply, we do not think it can be viewed as a discount factor for memory companies.) Consequently, we anticipate SEC to benefit from high-speed HBM4 market and gain market share in HBM4. We raise our HBM ASP forecast for SEC, and raise 2026F HBM shipment growth from +112% y-y to +144% y-y. Reflecting these factors, we expect SEC’s 1Q26F memory OP to be at KRW44tn (+153% q-q), which is substantially above our previous forecast of KRW33tn. Although we anticipate commodity memory price growth to decelerate from 2Q26F, we think HBM ASP should rise q-q, based on rising mix of HBM4. We expect SEC to record 2026/27F OP of KRW243tn/322tn (+457%/+33% y-y; Fig. 6), and 2026/27F ROE of 42%/42%, which is higher than our previous estimates of 34%/35%.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.014348760883882816, "ret_signed_m1d": -0.014348760883882816, "alpha_spy_m1d": -0.013644771340488582, "alpha_c_m1d": -0.01184100103768504, "bench_spy_m1d": -0.0007039895433942345, "bench_c_m1d": -0.002507759846197777, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.049117045193550535, "ret_signed_p1w": 0.049117045193550535, "alpha_spy_p1w": 0.037851865923986816, "alpha_c_p1w": 0.03965864819375442, "bench_spy_p1w": 0.011265179269563719, "bench_c_p1w": 0.009458396999796115, "ret_p1m": 0.04139076939242092, "ret_signed_p1m": 0.04139076939242092, "alpha_spy_p1m": 0.06004855704109402, "alpha_c_p1m": 0.046979697766254436, "bench_spy_p1m": -0.018657787648673096, "bench_c_p1m": -0.005588928373833513, "ret_p3m": 0.5705730465954528, "ret_signed_p3m": 0.5705730465954528, "alpha_spy_p3m": 0.4787692013151541, "alpha_c_p3m": 0.3018004805335839, "bench_spy_p3m": 0.09180384528029872, "bench_c_p3m": 0.26877256606186894, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4629, -0.47, -0.4475, -0.4794, -0.4689, -0.4546, -0.4354, -0.4316, -0.448, -0.4464, -0.4321, -0.4261, -0.4228, -0.4047, -0.3986, -0.4047, -0.4069, -0.4107, -0.4019, -0.4244, -0.4354, -0.4074, -0.4091, -0.4162, -0.3931, -0.3876, -0.3898, -0.3898, -0.3574, -0.3405, -0.3383, -0.3383, -0.3383, -0.2908, -0.2379, -0.2334, -0.2219, -0.234, -0.2329, -0.234, -0.2406, -0.2257, -0.2058, -0.1783, -0.176, -0.1987, -0.1749, -0.1595, -0.1606, -0.1606, -0.1198, -0.1038, -0.1131, -0.1142, -0.17, -0.0756, -0.0668, -0.1209, -0.1247, -0.0817, -0.085, -0.074, -0.0143, 0.0, 0.0, 0.0, 0.0, 0.0486, 0.0491, 0.0651, 0.1038, 0.1231, 0.2031, 0.1948, 0.1948, 0.0767, -0.0497, 0.0574, 0.0386, -0.0425, 0.037, 0.0486, 0.037, 0.0127, 0.0414, 0.0701, 0.1507, 0.1065, 0.1004, 0.0281, 0.0469, 0.043, -0.0061, -0.0061, -0.025, -0.0754, 0.0488, -0.0134, 0.0297, 0.0679, 0.0867, 0.1641, 0.1282, 0.1392, 0.1116, 0.142, 0.1669, 0.2028, 0.1945, 0.1862, 0.2111, 0.2028, 0.2415, 0.2139, 0.2415, 0.2277, 0.2498, 0.2194, 0.2194, 0.2858, 0.2858, 0.471, 0.5014, 0.4849, 0.5789, 0.5429, 0.5706, 0.6369, 0.4959, 0.554, 0.5236, 0.5263, 0.6563, 0.6176, 0.6176, 0.6535, 0.6978, 0.6563, 0.7531, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2008040189982388379", "ticker_idx": 0, "ts": "2026-01-05T04:57:43+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "What an insanely bullish report.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Endorses bullish AMD report citing MI455X GPU demand potentially reaching 600K–800K units in 2026.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "“We believe actual demand for MI455X GPUs could reach 600,000–800,000 units in 2026.”\n\nwow https://t.co/krOaPloQbB", "reply_text": null, "parent_text": null, "quoted_text": "What an insanely bullish report.\n\n$AMD https://t.co/8Xedh15823", "parent_author": null, "ret_m1d": 0.010810563460528888, "ret_signed_m1d": 0.010810563460528888, "alpha_spy_m1d": 0.01742661362945419, "alpha_c_m1d": 0.022172115902407974, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03044144957390582, "ret_signed_p1d": -0.03044144957390582, "alpha_spy_p1d": -0.036388632946995125, "alpha_c_p1d": -0.057004650073321006, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": -0.06056630757530401, "ret_signed_p1w": -0.06056630757530401, "alpha_spy_p1w": -0.0713846673459031, "alpha_c_p1w": -0.09518059145723945, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.09512393072697556, "ret_signed_p1m": 0.09512393072697556, "alpha_spy_p1m": 0.09249198082185917, "alpha_c_p1m": 0.04191808170360023, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": -0.016193241366944, "ret_signed_p3m": -0.016193241366944, "alpha_spy_p3m": 0.027572948573427514, "alpha_c_p3m": -0.055203837858347815, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0655, 0.0534, -0.028, -0.0211, -0.0135, 0.0792, 0.061, 0.0543, 0.0881, 0.0767, 0.0414, 0.0629, 0.144, 0.1746, 0.167, 0.1956, 0.1527, 0.1585, 0.1745, 0.131, 0.1594, 0.0752, 0.0564, 0.1036, 0.0744, 0.171, 0.1216, 0.1164, 0.0879, 0.0417, 0.0112, -0.0681, -0.0783, -0.0273, -0.0676, -0.0309, -0.0309, -0.0161, -0.006, -0.0264, -0.0157, -0.0231, -0.0141, 0.0001, 0.0024, 0.0015, 0.0016, -0.0466, -0.0611, -0.0539, -0.1039, -0.0906, -0.0346, -0.0277, -0.028, -0.0273, -0.0273, -0.0275, -0.0247, -0.026, -0.0313, -0.0313, 0.0108, 0.0, -0.0304, -0.05, -0.0742, -0.081, -0.0606, -0.0005, 0.0114, 0.0309, 0.0486, 0.0486, 0.049, 0.1299, 0.1477, 0.1746, 0.1367, 0.14, 0.1432, 0.1407, 0.0708, 0.1139, 0.0951, -0.0945, -0.1293, -0.0572, -0.023, -0.034, -0.0339, -0.0685, -0.0622, -0.0622, -0.0814, -0.0948, -0.0801, -0.0947, -0.1107, -0.0327, -0.0462, -0.0787, -0.0944, -0.1016, -0.1363, -0.086, -0.0978, -0.1296, -0.0832, -0.0807, -0.0735, -0.1056, -0.1252, -0.1108, -0.112, -0.0978, -0.0715, -0.0893, -0.0832, -0.0711, -0.0037, -0.0783, -0.0863, -0.1133, -0.0798, -0.0492, -0.0162, -0.0162, -0.0041, 0.002, 0.0486, 0.0704, 0.1084, 0.1165, 0.1537, 0.1675, 0.2586, 0.2592, 0.2437, 0.2868, 0.3726, 0.3811, 0.5732, 0.5136, 0.462, 0.5248, 0.6034, 0.6308, 0.5449, 0.6069, 0.9061, 0.8476, 1.0589, 1.0752, 1.0277, 1.0151, 1.0341, 0.9183, 0.9042, 0.8729, 1.0245, 1.0336, 1.1147, 1.1147, 1.2792, 1.2415, 1.3435, 1.3344, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011311326455591283", "ticker_idx": 0, "ts": "2026-01-14T05:36:03+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC N2 capacity to reach 100kwpm by 2026 and 140kwpm by 2027, with N3 capacity reaching 180kwpm in 2027", "tweet_type": "news_reaction", "timeline": "years", "summary": "Morgan Stanley relayed: robust 2nm demand pulling in TSMC EUV deliveries, implying bullish outlook for both TSMC capacity expansion and ASML scanner demand through 2027.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: Robust 2nm demand is driving a sharp increase in TSMC’s EUV tool deliveries \n\nIn a prior report, we mentioned that 2026 EUV scanner deliveries to TSMC have been pulled in given strong 3nm demand, as noted in AI Supply Chain: TSMC to expand 3nm capacity for major AI customer's growth? report. TSMC's 2025 EUV delivery has been pulled in 20%, and 2026 is increasing slightly vs. our prior checks. According to our recent checks, we also believe that it is possible that TSMC to have another 50% increase for its 2027 EUV scanners delivery given the huge demand for 2nm and pull-in schedule. We now expect TSMC N2 capacity to reach 100kwpm by 2026 and 140kwpm by 2027, with N3 capacity reaching 180kwpm in 2027.\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.004939516236272645, "alpha_c_m1d": -0.008188470775853718, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011695964230580391, "ret_signed_p1d": -0.011695964230580391, "alpha_spy_p1d": -0.014419178627109841, "alpha_c_p1d": -0.03245043167468642, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": 0.017543803107765488, "ret_signed_p1w": 0.017543803107765488, "alpha_spy_p1w": 0.024728433107698122, "alpha_c_p1w": -0.01742461805570894, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": 0.11988298879197634, "ret_signed_p1m": 0.11988298879197634, "alpha_spy_p1m": 0.13304994936981762, "alpha_c_p1m": 0.07415110011186421, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.16753952386937687, "ret_signed_p3m": 0.16753952386937687, "alpha_spy_p3m": 0.17099606821864055, "alpha_c_p3m": 0.025940478240117137, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1549, -0.1374, -0.1374, -0.149, -0.1549, -0.1549, -0.1374, -0.1403, -0.1228, -0.1228, -0.1257, -0.1199, -0.1228, -0.149, -0.1461, -0.149, -0.1403, -0.1461, -0.1403, -0.149, -0.1665, -0.1578, -0.1811, -0.1869, -0.1519, -0.1927, -0.1986, -0.1753, -0.1607, -0.1636, -0.1607, -0.1782, -0.1665, -0.1549, -0.1578, -0.149, -0.1286, -0.1374, -0.1228, -0.1404, -0.1345, -0.152, -0.1608, -0.1637, -0.1637, -0.1637, -0.1433, -0.1287, -0.1257, -0.1257, -0.117, -0.1053, -0.1111, -0.0936, -0.0936, -0.0731, -0.0234, -0.0029, -0.0205, -0.0146, -0.0175, -0.0117, 0.0, 0.0, -0.0117, 0.0175, 0.0292, 0.038, 0.0175, 0.0292, 0.0351, 0.0263, 0.0409, 0.0643, 0.0556, 0.038, 0.0322, 0.0526, 0.0439, 0.0322, 0.0409, 0.0614, 0.0994, 0.1199, 0.1199, 0.1199, 0.1199, 0.1199, 0.1199, 0.1199, 0.1199, 0.1111, 0.1491, 0.1784, 0.1667, 0.1667, 0.155, 0.1316, 0.0906, 0.1111, 0.1053, 0.0585, 0.0819, 0.1345, 0.1023, 0.0906, 0.0789, 0.0971, 0.1177, 0.0854, 0.0795, 0.0619, 0.0619, 0.0825, 0.0795, 0.0678, 0.0443, 0.0326, 0.0883, 0.0619, 0.0619, 0.0619, 0.0913, 0.1441, 0.147, 0.1734, 0.1675, 0.2057, 0.2203, 0.2233, 0.191, 0.1881, 0.2027, 0.2027, 0.2203, 0.2819, 0.3289, 0.2995, 0.279, 0.2526, 0.2526, 0.3347, 0.3201, 0.3201, 0.3553, 0.3436, 0.3113, 0.323, 0.3025, 0.3318, 0.3289, 0.3142, 0.2937, 0.2819, 0.3083, 0.323, 0.3553, 0.3318, 0.3494, 0.3465, 0.3817, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2011311326455591283", "ticker_idx": 1, "ts": "2026-01-14T05:36:03+00:00", "symbol": "ASML", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "another 50% increase for its 2027 EUV scanners delivery given the huge demand for 2nm", "tweet_type": "news_reaction", "timeline": "years", "summary": "Morgan Stanley relayed: robust 2nm demand pulling in TSMC EUV deliveries, implying bullish outlook for both TSMC capacity expansion and ASML scanner demand through 2027.", "resolved_tickers": ["ASML"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: Robust 2nm demand is driving a sharp increase in TSMC’s EUV tool deliveries \n\nIn a prior report, we mentioned that 2026 EUV scanner deliveries to TSMC have been pulled in given strong 3nm demand, as noted in AI Supply Chain: TSMC to expand 3nm capacity for major AI customer's growth? report. TSMC's 2025 EUV delivery has been pulled in 20%, and 2026 is increasing slightly vs. our prior checks. According to our recent checks, we also believe that it is possible that TSMC to have another 50% increase for its 2027 EUV scanners delivery given the huge demand for 2nm and pull-in schedule. We now expect TSMC N2 capacity to reach 100kwpm by 2026 and 140kwpm by 2027, with N3 capacity reaching 180kwpm in 2027.\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0050961475826813984, "ret_signed_m1d": 0.0050961475826813984, "alpha_spy_m1d": 0.0001566313464087532, "alpha_c_m1d": -0.0030923231931723194, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05371451802519989, "ret_signed_p1d": 0.05371451802519989, "alpha_spy_p1d": 0.05099130362867044, "alpha_c_p1d": 0.03296005058109386, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": 0.07625900717161627, "ret_signed_p1w": 0.07625900717161627, "alpha_spy_p1w": 0.0834436371715489, "alpha_c_p1w": 0.04129058600814184, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": 0.11474518195804984, "ret_signed_p1m": 0.11474518195804984, "alpha_spy_p1m": 0.12791214253589112, "alpha_c_p1m": 0.0690132932779377, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.18869588329993614, "ret_signed_p3m": 0.18869588329993614, "alpha_spy_p3m": 0.19215242764919982, "alpha_c_p3m": 0.047096837670676406, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.187, -0.1768, -0.1903, -0.2009, -0.1813, -0.1839, -0.1627, -0.1686, -0.1526, -0.149, -0.1618, -0.1558, -0.1848, -0.1741, -0.1856, -0.1953, -0.178, -0.1909, -0.1791, -0.193, -0.2032, -0.1929, -0.2055, -0.1776, -0.2237, -0.2351, -0.2183, -0.2061, -0.1763, -0.1763, -0.1612, -0.1391, -0.1226, -0.0972, -0.1216, -0.13, -0.114, -0.1205, -0.1143, -0.1115, -0.1447, -0.1392, -0.1485, -0.1965, -0.18, -0.1644, -0.1636, -0.1598, -0.1568, -0.1568, -0.1511, -0.1565, -0.1516, -0.1534, -0.1534, -0.0791, -0.0281, -0.017, -0.0279, -0.0549, 0.008, 0.0139, 0.0051, 0.0, 0.0537, 0.0751, 0.0751, 0.0493, 0.0763, 0.1039, 0.0992, 0.1184, 0.151, 0.126, 0.1515, 0.126, 0.1406, 0.1046, 0.0597, 0.0684, 0.1181, 0.1312, 0.1201, 0.1375, 0.1147, 0.1145, 0.1145, 0.125, 0.1638, 0.156, 0.1644, 0.1774, 0.1868, 0.2095, 0.1599, 0.1494, 0.128, 0.0784, 0.1088, 0.0842, 0.0244, 0.0756, 0.0961, 0.0987, 0.0709, 0.0663, 0.0899, 0.1007, 0.0738, 0.0827, 0.0437, 0.0852, 0.1088, 0.1045, 0.0534, 0.032, -0.0064, 0.0466, 0.0774, 0.0437, 0.0437, 0.0332, 0.0352, 0.126, 0.1478, 0.1713, 0.1887, 0.203, 0.1741, 0.1179, 0.1567, 0.1699, 0.156, 0.1439, 0.1234, 0.155, 0.1375, 0.0995, 0.107, 0.1427, 0.1332, 0.1008, 0.1458, 0.2267, 0.2043, 0.2642, 0.2434, 0.2078, 0.2559, 0.2582, 0.1926, 0.1692, 0.1589, 0.2309, 0.2642, 0.2967, 0.2967, 0.296, 0.2689, 0.2751, 0.2807, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2007305149304451397", "ticker_idx": 0, "ts": "2026-01-03T04:16:56+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel is a card that has to succeed. I believe the U.S. will do whatever it takes to make Intel successful.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long INTC: thesis rests on US government commitment to making Intel succeed regardless of competitive position.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@0xgoodie @elonmusk My investment thesis for Intel is less about Intel’s own competitiveness and more about the U.S. government’s commitment to semiconductors.\n\nFrom the U.S. perspective, Intel is a card that has to succeed. I believe the U.S. will do whatever it takes to make Intel successful.", "reply_text": null, "parent_text": "@jukan05 @elonmusk Could you share your thought process on intel, appreciate it", "quoted_text": null, "parent_author": "0xgoodie", "ret_m1d": 0.0002540547752560851, "ret_signed_m1d": 0.0002540547752560851, "alpha_spy_m1d": 0.006870104944181388, "alpha_c_m1d": 0.011615607217135171, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.011361552441879086, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01701808488239398, "ret_signed_p1d": 0.01701808488239398, "alpha_spy_p1d": 0.011070901509304676, "alpha_c_p1d": -0.009545115617021205, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.026563200499415185, "ret_p1w": 0.11912630349623754, "ret_signed_p1w": 0.11912630349623754, "alpha_spy_p1w": 0.10830794372563846, "alpha_c_p1w": 0.0845120196143021, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.03461428388193544, "ret_p1m": 0.2509525358435727, "ret_signed_p1m": 0.2509525358435727, "alpha_spy_p1m": 0.2483205859384563, "alpha_c_p1m": 0.19774668682019736, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": 0.05320584902337533, "ret_p3m": 0.2796546211565616, "ret_signed_p3m": 0.2796546211565616, "alpha_spy_p3m": 0.3234208110969331, "alpha_c_p3m": 0.24064402466515777, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": 0.039010596491403815, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0493, -0.0399, -0.0762, -0.0546, -0.095, -0.0564, -0.0643, -0.0599, -0.0323, -0.0318, -0.0622, -0.0307, -0.0277, 0.0043, 0.0549, 0.05, 0.0201, 0.0157, 0.0033, -0.0594, -0.0251, -0.0541, -0.0315, -0.0234, -0.0378, -0.0376, -0.0879, -0.0978, -0.1184, -0.128, -0.1082, -0.1461, -0.1237, -0.0909, -0.0899, -0.065, -0.065, 0.0302, 0.0163, 0.1041, 0.1115, 0.0287, 0.0518, 0.0236, 0.0287, 0.0358, 0.0036, -0.0396, -0.0472, -0.0523, -0.0843, -0.0785, -0.0648, -0.0762, -0.0767, -0.0815, -0.0815, -0.0805, -0.0683, -0.0526, -0.0627, -0.0627, 0.0003, 0.0, 0.017, 0.0828, 0.0442, 0.157, 0.1191, 0.2012, 0.2375, 0.2273, 0.1928, 0.1928, 0.2334, 0.378, 0.3797, 0.1448, 0.0792, 0.1158, 0.239, 0.236, 0.1803, 0.2398, 0.251, 0.2344, 0.2253, 0.285, 0.2761, 0.1971, 0.2266, 0.1806, 0.1885, 0.1885, 0.173, 0.1547, 0.1334, 0.1204, 0.1082, 0.1715, 0.1908, 0.1547, 0.1585, 0.1557, 0.0947, 0.1577, 0.1671, 0.1029, 0.1577, 0.1882, 0.2187, 0.1494, 0.1626, 0.1623, 0.1191, 0.1438, 0.173, 0.1143, 0.1179, 0.1191, 0.1984, 0.1201, 0.0955, 0.0462, 0.1209, 0.22, 0.2797, 0.2797, 0.2898, 0.3439, 0.4973, 0.5677, 0.5845, 0.6556, 0.6208, 0.6495, 0.7399, 0.7399, 0.6688, 0.683, 0.6579, 0.6962, 1.0965, 1.1588, 1.1468, 1.4067, 1.3998, 1.5304, 1.4328, 1.747, 1.8705, 1.7844, 2.173, 2.2878, 2.0635, 2.0554, 1.9446, 1.7628, 1.7475, 1.8143, 2.0216, 2.0099, 2.0439, 2.0439, 2.1374, 2.093, 2.0706, 1.9129, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2029344714533814418", "ticker_idx": 0, "ts": "2026-03-04T23:54:18+00:00", "symbol": "HBM memory", "kind": "sector", "geography": "global", "sub_sector": "HBM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM and DRAM will continue to serve as main memory for large-scale model training and general-purpose inference servers... the memory hierarchy spanning SRAM, HBM, and DRAM will become increasingly multi-layered, ultimately driving an expansion of the total addressable market (TAM) for the memory industry as a whole", "tweet_type": "analysis", "timeline": "quarters", "summary": "SRAM-based inference chips are not a substitute for HBM/DRAM; memory TAM expands as hierarchy becomes more granular, keeping HBM demand intact.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "Global HBM memory basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS: Commentary on KOSPI Decline\n\nKOSPI Falls on Reports of NVIDIA's Upcoming Inference Chip Using Groq Architecture\n\nMemory stocks weakened on news that NVIDIA is developing a new inference chip leveraging Groq's on-chip SRAM-based architecture. The concern was that SRAM usage could reduce demand for main memory, including HBM. NVIDIA is reportedly set to unveil the architecture at GTC in March.\n\nSRAM Is Far More Expensive and Lower in Capacity Than DRAM\n\nThe claim that the emergence of a \"low-cost\" SRAM-based inference chip will reduce usage of existing main memory such as HBM reflects a poor understanding of memory. SRAM has a larger cell area and lower density than DRAM, making its cost per bit significantly higher. SRAM generally requires 5–10x the die area of DRAM for the same capacity. Its structure also makes it difficult to scale density, limiting capacity expansion. For these reasons, SRAM has historically been used for cache or on-chip buffer applications requiring extreme low latency, rather than serving as main memory for storing large volumes of data.\n\nSRAM Architecture Is Not a Substitute for DRAM—It's a Separate Option\n\nSRAM-centric architectures offer the advantage of much lower access latency and minimized data movement compared to DRAM. NVIDIA therefore plans to leverage Groq's architecture in a form optimized for specific inference workloads that are difficult to address with GPUs. Ultimately, the adoption of an SRAM-centric architecture should be understood not as a strategy to replace HBM or DRAM, but as a distinct option for specific data center workloads requiring ultra-low latency and for Physical AI edge applications such as robotics and autonomous driving that demand real-time responsiveness. In fact, OpenAI is already deploying Cerebras's SRAM-centric chips in its data centers, and the inference services built on them charge higher API costs than standard GPU-based inference services.\n\nThe Memory Hierarchy Is Becoming More Granular, Expanding the Overall TAM\n\nThe proliferation of Groq-based SRAM-centric architectures will further segment the memory hierarchy within AI infrastructure. HBM and DRAM will continue to serve as main memory for large-scale model training and general-purpose inference servers, while SRAM-based architectures will fulfill a separate role for ultra-low-latency inference and edge AI. As the AI industry advances, the memory hierarchy spanning SRAM, HBM, and DRAM will become increasingly multi-layered, ultimately driving an expansion of the total addressable market (TAM) for the memory industry as a whole.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06212278052658804, "ret_signed_m1d": 0.06212278052658804, "alpha_spy_m1d": 0.06912866023802389, "alpha_c_m1d": 0.08226812861469734, "bench_spy_m1d": -0.007005879711435847, "bench_c_m1d": -0.020145348088109305, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07058015472998487, "ret_signed_p1d": 0.07058015472998487, "alpha_spy_p1d": 0.0761557015763787, "alpha_c_p1d": 0.07997629590425322, "bench_spy_p1d": -0.005575546846393831, "bench_c_p1d": -0.009396141174268346, "ret_p1w": 0.09097830775741689, "ret_signed_p1w": 0.09097830775741689, "alpha_spy_p1w": 0.10382253966679245, "alpha_c_p1w": 0.08614244545156491, "bench_spy_p1w": -0.012844231909375559, "bench_c_p1w": 0.004835862305851979, "ret_p1m": -0.02333515845309227, "ret_signed_p1m": -0.02333515845309227, "alpha_spy_p1m": 0.016816175266201088, "alpha_c_p1m": -0.006346938268654807, "bench_spy_p1m": -0.04015133371929336, "bench_c_p1m": -0.016988220184437464, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3808, -0.3577, -0.3595, -0.3427, -0.357, -0.366, -0.3835, -0.401, -0.3887, -0.3698, -0.3598, -0.3299, -0.3267, -0.3172, -0.3172, -0.3031, -0.273, -0.2694, -0.2754, -0.2754, -0.2236, -0.2003, -0.161, -0.1539, -0.1631, -0.1523, -0.1501, -0.1632, -0.1604, -0.1479, -0.1138, -0.1099, -0.1242, -0.097, -0.0786, -0.0726, -0.0935, -0.0365, 0.006, 0.011, 0.012, -0.0195, 0.0285, -0.0044, -0.0432, -0.0359, -0.0113, -0.0253, 0.0031, 0.038, 0.038, 0.038, 0.0281, 0.0457, 0.0653, 0.096, 0.0964, 0.1287, 0.1497, 0.1991, 0.1786, 0.1789, 0.0621, 0.0, 0.0706, 0.0351, -0.0121, 0.0673, 0.091, 0.066, 0.0669, 0.1151, 0.1402, 0.2022, 0.1554, 0.1331, 0.0633, 0.0833, 0.0743, 0.0106, 0.0121, -0.0475, -0.0777, 0.0261, -0.0233, 0.0099, 0.0368, 0.055, 0.1523, 0.1383, 0.1528, 0.1531, 0.2211, 0.2349, 0.2558, 0.2405, 0.247, 0.2793, 0.3077, 0.3173, 0.3189, 0.3792, 0.3606, 0.3774, 0.3629, 0.3838, 0.4988, 0.5518, 0.6992, 0.7141, 0.8042, 0.9537, 0.8995, 0.9954, 0.9933, 0.8418, 0.8346, 0.8009, 0.8296, 0.9768, 0.9543, 0.9543, 1.1311, 1.2487, 1.2483, 1.339, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2008046471489429890", "ticker_idx": 0, "ts": "2026-01-05T05:22:41+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bearish $AAPL", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bearish AAPL in reaction to Samsung doubling Google Gemini AI devices, implying Apple loses competitive ground.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@illyquid Bearish $AAPL", "reply_text": null, "parent_text": "$goog \n\nSEOUL, Jan 5 (Reuters) - Samsung Electronics plans to double this year the number of its mobile devices with AI features powered by Google's Gemini, its co-CEO said, which would give the U.S. firm an edge over rivals as the global race in artificial intelligence hots up.", "quoted_text": null, "parent_author": "illyquid", "ret_m1d": 0.014031266391232089, "ret_signed_m1d": -0.014031266391232089, "alpha_spy_m1d": -0.02064731656015739, "alpha_c_m1d": -0.01624391145972215, "bench_spy_m1d": -0.006616050168925303, "bench_c_m1d": -0.00221264506849006, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.018334270451824564, "ret_signed_p1d": 0.018334270451824564, "alpha_spy_p1d": 0.024281453824913868, "alpha_c_p1d": 0.032371071033609145, "bench_spy_p1d": 0.005947183373089304, "bench_c_p1d": 0.014036800581784581, "ret_p1w": -0.02622905997768077, "ret_signed_p1w": 0.02622905997768077, "alpha_spy_p1w": 0.03704741974827985, "alpha_c_p1w": 0.04123395219983983, "bench_spy_p1w": 0.010818359770599084, "bench_c_p1w": 0.01500489222215906, "ret_p1m": 0.00830662593299536, "ret_signed_p1m": -0.00830662593299536, "alpha_spy_p1m": -0.0056746760278789665, "alpha_c_p1m": -0.0258698196888677, "bench_spy_p1m": 0.002631949905116393, "bench_c_p1m": -0.01756319375587234, "ret_p3m": -0.041534559652501635, "ret_signed_p3m": 0.041534559652501635, "alpha_spy_p3m": -0.002231630287869879, "alpha_c_p3m": -0.016969804260371757, "bench_spy_p3m": -0.043766189940371514, "bench_c_p3m": -0.05850436391287339, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0354, -0.0504, -0.0832, -0.0742, -0.0738, -0.068, -0.075, -0.0569, -0.0197, -0.0178, -0.0339, -0.0297, -0.0176, 0.0048, 0.0055, 0.0082, 0.0145, 0.0107, 0.0057, 0.0094, 0.0098, 0.0084, 0.0036, 0.0081, 0.0299, 0.0232, 0.0213, 0.0193, 0.0007, 0.0007, 0.0049, -0.0038, 0.0158, 0.0324, 0.0363, 0.0385, 0.0385, 0.0434, 0.0593, 0.0708, 0.0632, 0.0503, 0.0431, 0.0398, 0.0371, 0.0431, 0.0403, 0.0412, 0.0256, 0.0275, 0.0171, 0.0184, 0.024, 0.0139, 0.0191, 0.0245, 0.0245, 0.023, 0.0243, 0.0218, 0.0172, 0.0172, 0.014, 0.0, -0.0183, -0.0259, -0.0308, -0.0295, -0.0262, -0.0232, -0.0273, -0.0339, -0.0439, -0.0439, -0.0769, -0.0734, -0.0708, -0.0719, -0.0443, -0.0336, -0.0405, -0.0336, -0.0291, 0.0103, 0.0083, 0.0345, 0.0324, 0.0406, 0.0285, 0.025, 0.0318, -0.0198, -0.0421, -0.0421, -0.0117, -0.01, -0.0241, -0.0091, -0.0031, 0.0192, 0.027, 0.0222, -0.0106, -0.0086, -0.0122, -0.0168, -0.0252, -0.0358, -0.0267, -0.0231, -0.0232, -0.0421, -0.0633, -0.0531, -0.0479, -0.0639, -0.0676, -0.0712, -0.0581, -0.0576, -0.0539, -0.0529, -0.0682, -0.0763, -0.0495, -0.0426, -0.0415, -0.0415, -0.0305, -0.0506, -0.0304, -0.0244, -0.0245, -0.0293, -0.0306, -0.0022, -0.0135, 0.0121, 0.0226, -0.0031, 0.0231, 0.024, 0.0152, 0.0022, 0.0139, 0.0118, 0.0163, 0.0492, 0.0368, 0.0643, 0.0768, 0.0765, 0.0985, 0.0971, 0.1051, 0.1204, 0.1179, 0.1255, 0.1165, 0.1207, 0.133, 0.1433, 0.1577, 0.1577, 0.1558, 0.1653, 0.1715, 0.1698, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010864327327445043", "ticker_idx": 0, "ts": "2026-01-12T23:59:50+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "PT Won172,000 from Won154,000 - Buy", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "UBS reiterates Buy on Samsung Electronics, raises PT to Won172,000 on peak DRAM profitability, strong DDR/NAND pricing cycle, and HBM4 progress; 32% above consensus OP forecast.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: Samsung Electronics\nBack to peak DRAM profitability in earnest\n\nDDR and NAND contract pricing off to a roaring start \n4Q25 DRAM blended ASPs ended up stronger than we initially expected post prelims (link), as Samsung successfully raised server DRAM contract pricing by about 65% QoQ. We hence now estimate total DRAM blended ASPs up 43% QoQ in 4Q25 (DDR: +47%). From this already high level, we believe that Samsung should be able to increase blended DDR contract pricing by 56% QoQ in 1Q26, and total blended DRAM ASPs by 50%. We forecast NAND ASPs up 27% QoQ in 1Q26. This means that Samsung DDR contract pricing could be close to the 3Q18 peak. Even from that elevated level, we expect DDR contract pricing to continue to increase till 1Q27, and for NAND, 3Q26. This would propel FCF to Won105tn in 2026E, and could pave the way for Won45tn of exceptional dividend. We see further upside to the stock from 1/ strong earnings upgrades ahead; 2/ investors re-thinking the duration of the cycle, but also, longer term economics - both offsetting HBM execution still in flux.\n\nGetting closer to deliver HBM4 working samples\nVice Chairman Jun Young-hyun reportedly messaged earlier this year that Samsung is \"back\" with HBM4 competitiveness (link). We believe that Samsung is getting closer to send HBM4 working samples, with a qualification window at Nvidia later in 1H26. We believe SK Hynix has been shipping HBM4 to Nvidia since last October. We maintain our forecast for HBM volume shipments to reach 7.5bn Gb (+77% YoY) in 2026, after 4.2bn Gb in '25E.\n\nSignificantly increasing forecasts, again \nWe increase our '26 OP forecast to Won171tn (was Won135tn), 32% above VA consensus of Won130tn. This is on the back of revised up DDR/NAND contract pricing forecasts, more than offsetting downside in Smartphone margins. For the same reasons, we raise '27E OP to Won181tn (was Won144tn), 10% above consensus of Won165tn. We now have DRAM operating margins peaking out at 75% in 4Q26E, compared to prior peak of 72% in 3Q18.\n\nValuation: PT Won172,000 from Won154,000 - Buy \nWe value Samsung ordinary shares at 2.03x NTM book (was 1.97x) considering our 2026-30 average ROE estimate of 18.0% (was 16.3%) and CoE of 8.9% (was 8.3%). We also raise our Samsung GDR PT to US$2,970, from US$2,610, with the latest FX.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0014409717659100707, "ret_signed_m1d": 0.0014409717659100707, "alpha_spy_m1d": 0.003008963119683572, "alpha_c_m1d": 0.005800946591273282, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.004359974825363211, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008645492180297198, "ret_signed_p1d": -0.008645492180297198, "alpha_spy_p1d": -0.006646017074426913, "alpha_c_p1d": -0.0065336244955571665, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": -0.002111867684740032, "ret_p1w": 0.07564842319402776, "ret_signed_p1w": 0.07564842319402776, "alpha_spy_p1w": 0.08068325135458898, "alpha_c_p1w": 0.08361900379787979, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": -0.007970580603852029, "ret_p1m": 0.19452453289965055, "ret_signed_p1m": 0.19452453289965055, "alpha_spy_p1m": 0.19889765286173, "alpha_c_p1m": 0.2234092750485267, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": -0.028884742148876152, "ret_p3m": 0.4727827912730851, "ret_signed_p3m": 0.4727827912730851, "alpha_spy_p3m": 0.4920490998578334, "alpha_c_p3m": 0.5037341267267859, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": -0.030951335453700768, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3189, -0.2995, -0.2981, -0.2966, -0.3009, -0.2931, -0.3081, -0.2916, -0.2687, -0.2866, -0.2794, -0.2536, -0.2293, -0.2034, -0.2479, -0.2787, -0.2888, -0.2981, -0.2787, -0.2579, -0.2608, -0.2629, -0.3031, -0.2787, -0.2988, -0.3081, -0.2787, -0.3203, -0.3067, -0.288, -0.2629, -0.2579, -0.2794, -0.2773, -0.2586, -0.2508, -0.2465, -0.2228, -0.2149, -0.2228, -0.2257, -0.2307, -0.2192, -0.2486, -0.2629, -0.2264, -0.2285, -0.2379, -0.2077, -0.2006, -0.2034, -0.2034, -0.1611, -0.139, -0.1362, -0.1362, -0.1362, -0.0742, -0.005, 0.0007, 0.0159, 0.0, 0.0014, 0.0, -0.0086, 0.0108, 0.0367, 0.0728, 0.0756, 0.0461, 0.0771, 0.0973, 0.0958, 0.0958, 0.1491, 0.17, 0.1578, 0.1563, 0.0836, 0.2068, 0.2183, 0.1477, 0.1427, 0.1988, 0.1945, 0.2089, 0.2867, 0.3055, 0.3055, 0.3055, 0.3055, 0.3689, 0.3696, 0.3905, 0.4409, 0.4661, 0.5706, 0.5598, 0.5598, 0.4056, 0.2406, 0.3804, 0.3559, 0.25, 0.3537, 0.3689, 0.3537, 0.322, 0.3595, 0.397, 0.5022, 0.4445, 0.4366, 0.3422, 0.3667, 0.3617, 0.2976, 0.2976, 0.2728, 0.2071, 0.3692, 0.288, 0.3443, 0.3941, 0.4186, 0.5197, 0.4728, 0.4872, 0.4511, 0.4908, 0.5233, 0.5702, 0.5594, 0.5486, 0.5811, 0.5702, 0.6208, 0.5847, 0.6208, 0.6027, 0.6316, 0.5919, 0.5919, 0.6785, 0.6785, 0.9204, 0.9601, 0.9384, 1.0612, 1.0142, 1.0503, 1.137, 0.9529, 1.0287, 0.989, 0.9926, 1.1622, 1.1117, 1.1117, 1.1586, 1.2164, 1.1622, 1.2886, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2008796694306320476", "ticker_idx": 0, "ts": "2026-01-07T07:03:48+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The memory market has entered a \"Hyper-Bull\" phase that surpasses the historic peak of 2018. Driven by relentless demand for AI and server capacity, suppliers' pricing power has reached an all-time high.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory sector in 'Hyper-Bull' phase with pricing power at all-time highs; RDIMM prices projected to hit $700 by March 2026, potentially $1,000 later in 2026.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Memory Prices Surge ~50% in Q4 2025, Additional Increases Expected in H1 2026\n\n- Memory Price Trend: The price of 64GB RDIMM, which was $255 in Q3 2025, surged to $450 in Q4 2025 and is projected to reach $700 by March 2026\n- Changes in BoM Cost Structure: Rising memory prices are fundamentally reshaping hardware manufacturers’ BoM structures\n- Memory Supply Outlook: Capex is increasing, but supply easing is expected to be delayed\n\nThe memory market has entered a “Hyper-Bull” phase that surpasses the historic peak of 2018. Driven by relentless demand for AI and server capacity, suppliers’ pricing power has reached an all-time high. According to Counterpoint Research’s recently released [Memory Monthly Pricing Tracker (Jan 2026)](https://t.co/1xMh4URa0Y), memory prices are forecast to surge 40–50% in Q4 2025, with an additional 40–50% increase in Q1 2026 and roughly 20% further rise in Q2 2026.\n\nMemory Price Trend: Breaking Historical Records\n\nThe 64GB RDIMM price, which stood at $255 in Q3 2025, jumped to $450 in Q4 2025 and is expected to hit $700 by March 2026.\n\n- The $1,000 Era: It would not be surprising to see prices reach $1,000 this year. This equates to approximately $1.95 per Gb—nearly double the 2018 peak of $1.00 per Gb.\n\nChanges in BoM Cost Structure\n\nRising memory prices are fundamentally altering hardware manufacturers’ BoM structures.\n\n- Impact on Smartphones: As of 2025, memory accounted for more than 10% of the iPhone 17 Pro Max BoM, a significant increase from 8% for the iPhone 12 Pro Max in 2020. For flagship models equipped with 16–24GB LPDDR5X RAM and 512GB–1TB UFS 4.0 storage, recent price surges could push memory costs to over 20% of the total BoM.\n- Pressure on Legacy Products: Major players including Samsung Electronics, SK hynix, and China’s CXMT are shifting production to high-margin server DDR5, rapidly reducing supply of older technologies such as LPDDR4 and eMMC.\n\nMemory Supply Outlook: Capex Rising, Supply Relief Delayed\n\n- Supply-Demand Gap: DRAM production in 2026 is projected to grow 24% year-over-year. Although capital expenditure (capex) is expanding, it is expected to take time before actual demand can be fully met.\n\n---\n\nhttps://t.co/BPCxlsjAOC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00834342703393065, "ret_signed_m1d": -0.00834342703393065, "alpha_spy_m1d": -0.011577207454240401, "alpha_c_m1d": -0.015096339032542039, "bench_spy_m1d": 0.003233780420309751, "bench_c_m1d": 0.006752911998611388, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011212302885210579, "ret_signed_p1d": -0.011212302885210579, "alpha_spy_p1d": -0.011110770203224818, "alpha_c_p1d": 0.004475137246967242, "bench_spy_p1d": -0.00010153268198576093, "bench_c_p1d": -0.01568744013217782, "ret_p1w": -0.007741344593708825, "ret_signed_p1w": -0.007741344593708825, "alpha_spy_p1w": -0.008872404470655553, "alpha_c_p1w": -0.01639029058317402, "bench_spy_p1w": 0.0011310598769467273, "bench_c_p1w": 0.008648945989465195, "ret_p1m": 0.13073255016798382, "ret_signed_p1m": 0.13073255016798382, "alpha_spy_p1m": 0.14807644455695415, "alpha_c_p1m": 0.14099171021723747, "bench_spy_p1m": -0.017343894388970327, "bench_c_p1m": -0.010259160049253646, "ret_p3m": 0.22720339611265938, "ret_signed_p3m": 0.22720339611265938, "alpha_spy_p3m": 0.26904116250725874, "alpha_c_p3m": 0.1987372845200297, "bench_spy_p3m": -0.04183776639459935, "bench_c_p3m": 0.028466111592629684, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4076, -0.405, -0.4162, -0.3985, -0.3683, -0.3621, -0.3484, -0.3571, -0.3571, -0.3553, -0.3237, -0.3039, -0.3143, -0.2907, -0.2803, -0.2766, -0.23, -0.2762, -0.2703, -0.2665, -0.2758, -0.2427, -0.242, -0.2401, -0.2509, -0.2777, -0.2539, -0.2898, -0.299, -0.3094, -0.3396, -0.3193, -0.3131, -0.297, -0.2862, -0.2934, -0.2852, -0.2711, -0.2764, -0.2868, -0.2678, -0.2409, -0.243, -0.2234, -0.2401, -0.2507, -0.2715, -0.2919, -0.2773, -0.255, -0.2432, -0.2081, -0.2042, -0.1932, -0.1932, -0.1762, -0.141, -0.1368, -0.1439, -0.1439, -0.0824, -0.0544, -0.0083, 0.0, -0.0112, 0.0016, 0.0041, -0.0112, -0.0077, 0.0071, 0.0477, 0.0523, 0.0354, 0.0678, 0.0895, 0.0965, 0.0722, 0.1392, 0.189, 0.1945, 0.1951, 0.1582, 0.2152, 0.1765, 0.1307, 0.1393, 0.1683, 0.1519, 0.1859, 0.2275, 0.2278, 0.2278, 0.2162, 0.2369, 0.2605, 0.2961, 0.2968, 0.3347, 0.3596, 0.4179, 0.3942, 0.3945, 0.2566, 0.1826, 0.2662, 0.2243, 0.1686, 0.2621, 0.29, 0.2607, 0.2617, 0.3182, 0.3482, 0.4215, 0.366, 0.3398, 0.2573, 0.2805, 0.2697, 0.1946, 0.1964, 0.1266, 0.0911, 0.2142, 0.1559, 0.195, 0.2272, 0.2486, 0.363, 0.3464, 0.3633, 0.3632, 0.4429, 0.4592, 0.4841, 0.4663, 0.4733, 0.511, 0.5444, 0.5563, 0.5578, 0.6285, 0.6063, 0.6265, 0.609, 0.6336, 0.7682, 0.8308, 1.0054, 1.0226, 1.1283, 1.3036, 1.2397, 1.3514, 1.35, 1.1712, 1.1631, 1.1242, 1.1581, 1.3311, 1.3041, 1.3041, 1.5117, 1.6486, 1.6469, 1.7548, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022158796379107403", "ticker_idx": 0, "ts": "2026-02-13T04:00:01+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley raises Kioxia target price to ¥33,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley PT raise to ¥33,000 on Kioxia; CY2026 NAND ASP forecast lifted to +100% YoY with sharply upgraded earnings outlook.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley raises Kioxia target price to ¥33,000\n\nThe firm significantly raised its CY2026 ASP growth forecast for Kioxia from +75% YoY to over +100% YoY, with Q1 ASP expected to surge approximately 90% QoQ, followed by an additional 10% QoQ increase in Q2.\n\nThis suggests Kioxia has achieved a breakthrough in price negotiations with its key OEM customer, implying that Apple will have to procure NAND at significantly higher prices than previously anticipated.\n\nAdditionally, Morgan Stanley sharply upgraded Kioxia's earnings outlook. For FY2027 (ending March 2027), the revenue forecast was raised from ¥2,689.8 billion to ¥4,280.1 billion, non-GAAP operating profit more than doubled from ¥1,171.8 billion to ¥2,723.2 billion, with an adjusted operating margin of 64%. EPS is projected to jump from ¥1,550.2 to ¥3,562.9, with net income expected to reach ¥1,930.3 billion.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07310133508426353, "ret_signed_m1d": -0.07310133508426353, "alpha_spy_m1d": -0.0723973455408693, "alpha_c_m1d": -0.06915250779023752, "bench_spy_m1d": -0.0007039895433942345, "bench_c_m1d": -0.003948827294026014, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023856423725103992, "ret_signed_p1d": -0.023856423725103992, "alpha_spy_p1d": -0.023856423725103992, "alpha_c_p1d": -0.023856423725103992, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.10002188662727074, "ret_signed_p1w": -0.10002188662727074, "alpha_spy_p1w": -0.11128706589683446, "alpha_c_p1w": -0.11795085144800022, "bench_spy_p1w": 0.011265179269563719, "bench_c_p1w": 0.017928964820729476, "ret_p1m": -0.010067848544539237, "ret_signed_p1m": -0.010067848544539237, "alpha_spy_p1m": 0.008589939104133859, "alpha_c_p1m": 0.023778878993113928, "bench_spy_p1m": -0.018657787648673096, "bench_c_p1m": -0.033846727537653165, "ret_p3m": 1.2105493543444954, "ret_signed_p3m": 1.2105493543444954, "alpha_spy_p3m": 1.1187455090641967, "alpha_c_p3m": 0.8064975044015792, "bench_spy_p3m": 0.09180384528029872, "bench_c_p3m": 0.4040518499429162, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5509, -0.5257, -0.5038, -0.561, -0.561, -0.5687, -0.6329, -0.6039, -0.5883, -0.6112, -0.5965, -0.6056, -0.603, -0.5872, -0.5557, -0.5678, -0.5855, -0.5793, -0.5671, -0.5973, -0.6197, -0.5929, -0.5837, -0.5912, -0.5568, -0.5644, -0.5358, -0.5275, -0.5003, -0.5327, -0.5432, -0.5432, -0.5432, -0.5432, -0.5032, -0.4922, -0.4445, -0.4309, -0.4445, -0.4445, -0.401, -0.4187, -0.4034, -0.3543, -0.3342, -0.3344, -0.2777, -0.216, -0.2412, -0.2362, -0.1924, -0.1707, -0.1576, -0.065, -0.1963, -0.0902, -0.0851, -0.1451, -0.1641, -0.1727, -0.1751, -0.1751, -0.0731, 0.0, -0.0239, -0.0208, -0.0624, -0.0698, -0.1, -0.1, -0.0252, -0.0604, -0.0703, -0.0716, -0.0582, -0.1158, -0.1576, -0.1114, -0.1256, -0.2108, -0.1434, -0.0641, -0.07, -0.0779, -0.0101, -0.0541, 0.0239, -0.0212, -0.0212, -0.0604, -0.0764, -0.0175, -0.0735, -0.1127, -0.1237, -0.1648, -0.0457, -0.0766, -0.0436, -0.002, 0.0186, 0.2081, 0.2125, 0.3193, 0.3688, 0.5321, 0.4187, 0.4826, 0.3364, 0.3355, 0.4331, 0.5233, 0.5487, 0.5137, 0.5526, 0.5898, 0.5898, 0.6441, 0.5938, 0.5938, 0.5938, 0.5938, 0.9002, 0.9475, 1.0109, 1.0179, 1.2105, 1.1213, 0.9457, 1.2521, 1.1786, 1.2451, 1.4224, 1.5126, 1.865, 1.7341, 1.6505, 1.6824, 1.8825, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2029745136499114216", "ticker_idx": 0, "ts": "2026-03-06T02:25:26+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Securities analysts have raised the possibility that a new substrate supply shortage cycle could emerge in 2027. AI-related demand remains robust, and Unimicron's Guangfu Fab 1 is actively ramping next-generation substrate production for GPU customers. Unimicron is a key substrate supplier to multiple ASIC companies and holds a de facto sole-source position with certain customers; company operating momentum is expected to grow in tandem with expanding AI accelerator shipments.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Unimicron: ABF substrate shortage cycle feared 2027, sole-source ASIC position, 40% capacity expansion, AI revenue ~60% of total, improving product mix.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Unimicron: Customers Rush to Lock In ABF Substrate LTAs — Supply Shortage Feared for Next Year\n\n- Amid continued upstream material cost inflation, IC substrate maker Unimicron will implement an additional ABF substrate price hike starting in March.\n\n- Customers are accelerating the signing of long-term agreements (LTAs) with Unimicron to secure production capacity; most contracts span 5 years, with some at 3 or 7 years.\n\n- LTA-related prepayments are expected to begin flowing in this year and will be used to fund ABF substrate capacity expansion investments.\n\n- Unimicron's ABF substrate production capacity is projected to grow approximately 40% by year-end versus end of last year.\n\n- Unimicron shares are currently subject to 5-minute trading restrictions; the stock hit the daily limit-down yesterday, but was trading up approximately 7.67% at NT$449 as of 9:42 AM today.\n\n- AI-related substrate revenue is expected to account for approximately 60% of total revenue this year, encompassing demand from GPUs, ASICs, AI peripherals, and general-purpose server CPUs.\n\n- Securities analysts have raised the possibility that a new substrate supply shortage cycle could emerge in 2027.\n\n- AI-related demand remains robust, and Unimicron's Guangfu Fab 1 is actively ramping next-generation substrate production for GPU customers.\n\n- Unimicron is a key substrate supplier to multiple ASIC companies and holds a de facto sole-source position with certain customers; company operating momentum is expected to grow in tandem with expanding AI accelerator shipments.\n\n- Unimicron is restructuring its production mix this year — planning to expand ABF substrate capacity by approximately 40% while reducing BT substrate capacity by approximately 15%.\n\n- The shift toward high-value-added products is expected to improve the overall product mix; the market currently forecasts full-year 2026 EPS of approximately NT$14.25.\n\n---\n\nhttps://t.co/irv8om7ZyL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04166666666666674, "ret_signed_m1d": 0.04166666666666674, "alpha_spy_m1d": 0.028385452933525235, "alpha_c_m1d": 0.02061631693289523, "bench_spy_m1d": 0.013281213733141506, "bench_c_m1d": 0.021050349733771512, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.09953703703703709, "ret_signed_p1d": -0.09953703703703709, "alpha_spy_p1d": -0.10829706205007616, "alpha_c_p1d": -0.11752816215866713, "bench_spy_p1d": 0.008760025013039074, "bench_c_p1d": 0.017991125121630036, "ret_p1w": 0.17592592592592582, "ret_signed_p1w": 0.17592592592592582, "alpha_spy_p1w": 0.19093229296405612, "alpha_c_p1w": 0.17949492857263194, "bench_spy_p1w": -0.015006367038130297, "bench_c_p1w": -0.003569002646706121, "ret_p1m": 0.20138888888888884, "ret_signed_p1m": 0.20138888888888884, "alpha_spy_p1m": 0.2187160618056294, "alpha_c_p1m": 0.2038648900390887, "bench_spy_p1m": -0.017327172916740574, "bench_c_p1m": -0.002476001150199858, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4769, -0.4734, -0.4792, -0.4792, -0.4873, -0.5046, -0.5093, -0.5208, -0.5023, -0.4896, -0.5012, -0.4977, -0.4977, -0.4896, -0.4896, -0.4919, -0.4907, -0.4907, -0.4931, -0.4676, -0.4664, -0.4769, -0.4734, -0.478, -0.4363, -0.412, -0.4213, -0.4225, -0.3657, -0.3032, -0.287, -0.2569, -0.2431, -0.2153, -0.2292, -0.1806, -0.1493, -0.2025, -0.1238, -0.1551, -0.1088, -0.1019, -0.1539, -0.213, -0.1956, -0.14, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.1458, -0.0613, 0.0324, 0.0613, 0.1146, 0.1146, 0.1134, 0.0718, -0.0347, 0.0417, 0.0, -0.0995, -0.0185, 0.0787, 0.1111, 0.1759, 0.2708, 0.2569, 0.3426, 0.3449, 0.2824, 0.1551, 0.0648, 0.1713, 0.162, 0.169, 0.1424, 0.0289, 0.1308, 0.2014, 0.2014, 0.2014, 0.3056, 0.4352, 0.4468, 0.4769, 0.4491, 0.3819, 0.4306, 0.4931, 0.4884, 0.5694, 0.6435, 0.6574, 0.6852, 0.8287, 0.9421, 0.9097, 0.8588, 1.044, 1.044, 1.1088, 1.0903, 0.9861, 1.0741, 0.8935, 0.9931, 1.0255, 1.0648, 1.0394, 0.9005, 0.8889, 0.8935, 0.9051, 1.0949, 1.2454, 1.2917, 1.5116, 1.5, 1.3727, 1.4421, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015657635266351267", "ticker_idx": 0, "ts": "2026-01-26T05:26:44+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the high-margin trend for DRAM is likely to continue for some time", "tweet_type": "analysis", "timeline": "quarters", "summary": "SK Hynix benefits from sustained high DRAM margins as conversion investments tighten supply; Micron flagged as a risk via aggressive capacity expansion.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Highlighting Profitability of 'General-Purpose' DRAM, SK Hynix Accelerates Conversion Investments\n\nSK Hynix is undertaking a major overhaul of its Icheon DRAM lines. Amid the strong upward trend in conventional DRAM prices, the company plans to convert to the most advanced products to enhance productivity.\n\nAt the same time, analysts suggest that strategic production cuts due to conversion investments could also be expected. In a situation where supply shortages persist, shipment volumes will be adjusted during the conversion investment period. As a result, the high-margin trend for DRAM is likely to continue for some time.\n\n◇ Expanding Conversion of Legacy DRAM Lines to Cutting-Edge Lines\n\nAccording to industry sources on the 22nd, SK Hynix is considering converting all of its 10nm-class 4th generation (1a) DRAM lines in Icheon to 10nm-class 6th generation (1c) DRAM lines. Initially, the plan was to maintain a certain volume of 1a DRAM lines, but the direction is shifting toward converting them all to 1c.\n\nThis is influenced by the recent strong upward trend in conventional DRAM prices. Currently, DRAM is facing an unprecedented shortage due to the expansion of artificial intelligence (AI) investments by big tech companies. Some consumers are even equipping PCs with DDR4 due to the burden of DDR5 prices.\n\nSpot prices for DRAM have shown sharp monthly increases in recent months. Market research firm TrendForce forecasts that DRAM prices will rise by 55-60% in the first quarter alone. Including high-bandwidth memory (HBM), the increase is expected to be 50-55%. Among these, LPDDR prices are reportedly rising the most sharply.\n\nThe background for SK Hynix's aggressive conversion investments includes the ability to naturally increase DRAM production through generational shifts. SK Hynix has stated that the productivity of 1c DDR5 has improved by more than 30% compared to the previous generation, 10nm-class 5th generation (1b) DDR5.\n\nConsidering that the existing lines were for 1a DRAM, the productivity improvement is expected to be even greater. Of course, 1c DRAM has drawbacks. The capital expenditure (CAPEX) required for conversion investments and the increased use of extreme ultraviolet (EUV) photoresist (PR) result in significantly higher cost burdens than previous generations.\n\nHowever, in the current phase where DRAM margins are high, these cost burdens are not seen as major constraints. The industry estimates that conventional DRAM margins have reached 60-70%. With DRAM prices expected to rise further in the first quarter, margins are likely to improve even more.\n\nThe 1b DRAM lines in Icheon will be maintained. Along with the M15X 1b DRAM lines in Cheongju, they are estimated to be used for HBM core die production.\n\nOnce SK Hynix's DRAM conversion investments are completed, only the Wuxi fab will remain producing 1a DRAM. Since advanced front-end equipment like EUV cannot be imported to Wuxi, this effectively reorganizes all fabs to the most advanced processes.\n\nA semiconductor industry official said, \"I understand that the decision was recently made to convert the 1a DRAM lines that were planned to be retained (in the Icheon DRAM fab) to 1c lines,\" adding, \"As a result, additional investments by SK Hynix are expected.\"\n\n◇ Warnings of Demand Contraction Amid DRAM Price Surge\n\nIn some parts of the memory industry, there are concerns that continued sharp rises in DRAM prices could lead to a contraction in customer demand.\n\nThis phenomenon is already appearing first in consumer product segments. Consumer resistance to rising laptop prices is reportedly significant. TrendForce forecasts that global laptop shipments in 2026 will be only 173 million units, a 5.4% decrease from the previous year.\n\nThe aggressive capacity expansions by companies like Micron and Kioxia are also pointed out as risk factors that could lower memory margins in the future. Micron recently acquired Taiwan's PSMC P5 fab. The plan is to produce DRAM there starting from the second half of next year.\n\nA memory industry official noted, \"It's true that orders from major customers are continuing,\" but added, \"If memory prices exceed a certain level, there will come a time when key customers reduce their CAPEX, and that could be a turning point for the market.\"\n\n---\n\nhttps://t.co/2PlZ84rvnn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04211959400443055, "ret_signed_m1d": 0.04211959400443055, "alpha_spy_m1d": 0.04717208377894322, "alpha_c_m1d": 0.038935227647806325, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08695651803959126, "ret_signed_p1d": 0.08695651803959126, "alpha_spy_p1d": 0.08297225330990665, "alpha_c_p1d": 0.06581918159466182, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.1277174044839562, "ret_signed_p1w": 0.1277174044839562, "alpha_spy_p1w": 0.12384861160387794, "alpha_c_p1w": 0.10477473930054693, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.36548918534981745, "ret_signed_p1m": 0.36548918534981745, "alpha_spy_p1m": 0.3732555728107938, "alpha_c_p1m": 0.31448874413736005, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.6674735127329907, "ret_signed_p3m": 0.6674735127329907, "alpha_spy_p3m": 0.6419876298071336, "alpha_c_p3m": 0.4592843616009663, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2424, -0.2288, -0.241, -0.1582, -0.2044, -0.2139, -0.1949, -0.2125, -0.1772, -0.1596, -0.1623, -0.1691, -0.2397, -0.1772, -0.2261, -0.237, -0.2247, -0.2926, -0.294, -0.2953, -0.2886, -0.2609, -0.2799, -0.269, -0.2418, -0.25, -0.2636, -0.2609, -0.216, -0.231, -0.2024, -0.2323, -0.2242, -0.2473, -0.2799, -0.2514, -0.25, -0.2568, -0.212, -0.2065, -0.2011, -0.2011, -0.1861, -0.1304, -0.1155, -0.1155, -0.1155, -0.0802, -0.0543, -0.0136, 0.0082, 0.0272, 0.0109, 0.0177, 0.0027, 0.0082, 0.0177, 0.0272, 0.038, 0.0095, 0.0054, 0.0258, 0.0421, 0.0, 0.087, 0.1427, 0.1698, 0.2351, 0.1277, 0.2323, 0.2228, 0.144, 0.1399, 0.2052, 0.1902, 0.1685, 0.2065, 0.1957, 0.1957, 0.1957, 0.1957, 0.2147, 0.2894, 0.2921, 0.3655, 0.3832, 0.496, 0.4442, 0.4442, 0.2782, 0.1557, 0.2809, 0.2578, 0.138, 0.2768, 0.2999, 0.2659, 0.2387, 0.3258, 0.3204, 0.4374, 0.3789, 0.3707, 0.27, 0.3421, 0.3544, 0.27, 0.27, 0.1883, 0.0985, 0.221, 0.1298, 0.1924, 0.206, 0.2469, 0.4061, 0.3585, 0.398, 0.4157, 0.5014, 0.5463, 0.5722, 0.5354, 0.5872, 0.6661, 0.6648, 0.6675, 0.6634, 0.7587, 0.7696, 0.76, 0.7505, 0.7505, 0.9697, 0.9697, 1.1793, 1.2514, 1.295, 1.5591, 1.4978, 1.6897, 1.6816, 1.476, 1.5046, 1.3753, 1.3753, 1.6407, 1.6421, 1.6421, 1.7932, 2.0532, 2.1163, 2.1762, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2015657635266351267", "ticker_idx": 1, "ts": "2026-01-26T05:26:44+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The aggressive capacity expansions by companies like Micron and Kioxia are also pointed out as risk factors that could lower memory margins in the future", "tweet_type": "analysis", "timeline": "quarters", "summary": "SK Hynix benefits from sustained high DRAM margins as conversion investments tighten supply; Micron flagged as a risk via aggressive capacity expansion.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Highlighting Profitability of 'General-Purpose' DRAM, SK Hynix Accelerates Conversion Investments\n\nSK Hynix is undertaking a major overhaul of its Icheon DRAM lines. Amid the strong upward trend in conventional DRAM prices, the company plans to convert to the most advanced products to enhance productivity.\n\nAt the same time, analysts suggest that strategic production cuts due to conversion investments could also be expected. In a situation where supply shortages persist, shipment volumes will be adjusted during the conversion investment period. As a result, the high-margin trend for DRAM is likely to continue for some time.\n\n◇ Expanding Conversion of Legacy DRAM Lines to Cutting-Edge Lines\n\nAccording to industry sources on the 22nd, SK Hynix is considering converting all of its 10nm-class 4th generation (1a) DRAM lines in Icheon to 10nm-class 6th generation (1c) DRAM lines. Initially, the plan was to maintain a certain volume of 1a DRAM lines, but the direction is shifting toward converting them all to 1c.\n\nThis is influenced by the recent strong upward trend in conventional DRAM prices. Currently, DRAM is facing an unprecedented shortage due to the expansion of artificial intelligence (AI) investments by big tech companies. Some consumers are even equipping PCs with DDR4 due to the burden of DDR5 prices.\n\nSpot prices for DRAM have shown sharp monthly increases in recent months. Market research firm TrendForce forecasts that DRAM prices will rise by 55-60% in the first quarter alone. Including high-bandwidth memory (HBM), the increase is expected to be 50-55%. Among these, LPDDR prices are reportedly rising the most sharply.\n\nThe background for SK Hynix's aggressive conversion investments includes the ability to naturally increase DRAM production through generational shifts. SK Hynix has stated that the productivity of 1c DDR5 has improved by more than 30% compared to the previous generation, 10nm-class 5th generation (1b) DDR5.\n\nConsidering that the existing lines were for 1a DRAM, the productivity improvement is expected to be even greater. Of course, 1c DRAM has drawbacks. The capital expenditure (CAPEX) required for conversion investments and the increased use of extreme ultraviolet (EUV) photoresist (PR) result in significantly higher cost burdens than previous generations.\n\nHowever, in the current phase where DRAM margins are high, these cost burdens are not seen as major constraints. The industry estimates that conventional DRAM margins have reached 60-70%. With DRAM prices expected to rise further in the first quarter, margins are likely to improve even more.\n\nThe 1b DRAM lines in Icheon will be maintained. Along with the M15X 1b DRAM lines in Cheongju, they are estimated to be used for HBM core die production.\n\nOnce SK Hynix's DRAM conversion investments are completed, only the Wuxi fab will remain producing 1a DRAM. Since advanced front-end equipment like EUV cannot be imported to Wuxi, this effectively reorganizes all fabs to the most advanced processes.\n\nA semiconductor industry official said, \"I understand that the decision was recently made to convert the 1a DRAM lines that were planned to be retained (in the Icheon DRAM fab) to 1c lines,\" adding, \"As a result, additional investments by SK Hynix are expected.\"\n\n◇ Warnings of Demand Contraction Amid DRAM Price Surge\n\nIn some parts of the memory industry, there are concerns that continued sharp rises in DRAM prices could lead to a contraction in customer demand.\n\nThis phenomenon is already appearing first in consumer product segments. Consumer resistance to rising laptop prices is reportedly significant. TrendForce forecasts that global laptop shipments in 2026 will be only 173 million units, a 5.4% decrease from the previous year.\n\nThe aggressive capacity expansions by companies like Micron and Kioxia are also pointed out as risk factors that could lower memory margins in the future. Micron recently acquired Taiwan's PSMC P5 fab. The plan is to produce DRAM there starting from the second half of next year.\n\nA memory industry official noted, \"It's true that orders from major customers are continuing,\" but added, \"If memory prices exceed a certain level, there will come a time when key customers reduce their CAPEX, and that could be a turning point for the market.\"\n\n---\n\nhttps://t.co/2PlZ84rvnn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02714026854416307, "ret_signed_m1d": -0.02714026854416307, "alpha_spy_m1d": -0.03219275831867574, "alpha_c_m1d": -0.023955902187538847, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05435759085969649, "ret_signed_p1d": -0.05435759085969649, "alpha_spy_p1d": -0.050373326130011886, "alpha_c_p1d": -0.03322025441476706, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.12518954482328515, "ret_signed_p1w": -0.12518954482328515, "alpha_spy_p1w": -0.1213207519432069, "alpha_c_p1w": -0.10224687963987589, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.07432730762902162, "ret_signed_p1m": -0.07432730762902162, "alpha_spy_p1m": -0.08209369508999798, "alpha_c_p1m": -0.02332686641656423, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.2385884502360145, "ret_signed_p3m": -0.2385884502360145, "alpha_spy_p3m": -0.21310256731015742, "alpha_c_p3m": -0.030399299103990085, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4178, -0.4245, -0.4251, -0.397, -0.4399, -0.3898, -0.3877, -0.3888, -0.3493, -0.3806, -0.3708, -0.3913, -0.3659, -0.3784, -0.413, -0.4196, -0.4827, -0.4673, -0.4247, -0.4232, -0.4084, -0.4084, -0.3925, -0.3822, -0.3847, -0.3984, -0.4177, -0.3906, -0.3656, -0.3515, -0.3225, -0.336, -0.3805, -0.3898, -0.4027, -0.4206, -0.3615, -0.3168, -0.2894, -0.2902, -0.2635, -0.2635, -0.2684, -0.2434, -0.2479, -0.2665, -0.2665, -0.1893, -0.1977, -0.1174, -0.1273, -0.1595, -0.1131, -0.1111, -0.131, -0.1433, -0.1348, -0.0677, -0.0677, -0.0619, 0.0001, 0.0218, 0.0271, 0.0, 0.0544, 0.1187, 0.12, 0.0663, 0.1252, 0.078, -0.0249, -0.0159, 0.0144, -0.0144, -0.0407, 0.0546, 0.0639, 0.058, 0.058, 0.0275, 0.0819, 0.0726, 0.1004, 0.0819, 0.0743, 0.1026, 0.068, 0.0598, 0.0606, -0.0242, 0.03, 0.0205, -0.0483, 0.0006, 0.036, 0.0761, 0.0418, 0.0952, 0.1355, 0.1866, 0.1867, 0.1418, 0.0869, 0.0392, 0.0166, -0.018, -0.0864, -0.0819, -0.1726, -0.1314, -0.0542, -0.0583, -0.0583, -0.0287, -0.0292, 0.0458, 0.0838, 0.0814, 0.0968, 0.1973, 0.173, 0.1756, 0.1701, 0.153, 0.1554, 0.2534, 0.2386, 0.2772, 0.3487, 0.2966, 0.3331, 0.3297, 0.3941, 0.4822, 0.6461, 0.7139, 0.6626, 0.9202, 1.0449, 0.971, 1.0663, 0.9953, 0.8632, 0.7524, 0.7966, 0.8821, 0.9595, 0.931, 0.931, 1.3035, 1.3871, 1.3745, 1.4966, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2014855907528867892", "ticker_idx": 0, "ts": "2026-01-24T00:20:57+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "there could be additional upside to the Street's expectation on near-term memory pricing and hence earnings for the memory names", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Goldman Sachs sees 1Q26 DRAM/NAND pricing rising above 4Q25 levels, with upside to consensus earnings for memory names.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Goldman Sachs: Memory prices are expected to post strong growth again in 1Q26\n\n\"Based on our discussion with investors, expectations for 1Q26 conventional memory pricing growth has been consistently rising throughout the past few weeks, and now at a level similar to or even higher than the increase seen in 4Q25. Based on our checks, certain customers in the mobile market have started to accept 1Q26 pricing quotes where for both DRAM and NAND the growth rates are meaningfully higher than what was agreed in 4Q25. As a result, we expect other customers to likely agree on a meaningful memory pricing growth in 1Q26, and along with the rapidly increasing spot pricing, there could be additional upside to the Street's expectation on near-term memory pricing and hence earnings for the memory names.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.023086620849531208, "ret_signed_m1d": 0.023086620849531208, "alpha_spy_m1d": 0.028139110624043878, "alpha_c_m1d": 0.019902254492906984, "bench_spy_m1d": -0.0050524897745126696, "bench_c_m1d": 0.003184366356624224, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06332211237199965, "ret_signed_p1d": 0.06332211237199965, "alpha_spy_p1d": 0.05933784764231505, "alpha_c_p1d": 0.04218477592707022, "bench_spy_p1d": 0.003984264729684606, "bench_c_p1d": 0.021137336444929433, "ret_p1w": 0.08057670648485653, "ret_signed_p1w": 0.08057670648485653, "alpha_spy_p1w": 0.07670791360477829, "alpha_c_p1w": 0.05763404130144727, "bench_spy_p1w": 0.0038687928800782423, "bench_c_p1w": 0.02294266518340926, "ret_p1m": 0.2515803012814515, "ret_signed_p1m": 0.2515803012814515, "alpha_spy_p1m": 0.25934668874242783, "alpha_c_p1m": 0.2005798600689941, "bench_spy_p1m": -0.007766387460976354, "bench_c_p1m": 0.05100044121245739, "ret_p3m": 0.4617065590683771, "ret_signed_p3m": 0.4617065590683771, "alpha_spy_p3m": 0.43622067614252, "alpha_c_p3m": 0.2535174079363527, "bench_spy_p3m": 0.025485882925857073, "bench_c_p3m": 0.2081891511320244, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3342, -0.3241, -0.3209, -0.2761, -0.3193, -0.3152, -0.3112, -0.3202, -0.2894, -0.2877, -0.2862, -0.2959, -0.3232, -0.2991, -0.3331, -0.3417, -0.3497, -0.3799, -0.362, -0.3563, -0.3415, -0.3307, -0.3383, -0.3306, -0.3167, -0.3216, -0.3312, -0.3141, -0.2884, -0.2911, -0.2728, -0.2888, -0.2974, -0.3171, -0.3367, -0.322, -0.3025, -0.2927, -0.2595, -0.2557, -0.2459, -0.2459, -0.2297, -0.1961, -0.1917, -0.1979, -0.1979, -0.1416, -0.1147, -0.0726, -0.064, -0.0733, -0.0628, -0.0603, -0.0745, -0.0709, -0.057, -0.0205, -0.016, -0.0326, -0.0039, 0.0163, 0.0231, 0.0, 0.0633, 0.1097, 0.1155, 0.1189, 0.0806, 0.1372, 0.1032, 0.0585, 0.0657, 0.0949, 0.0799, 0.1088, 0.1482, 0.1483, 0.1483, 0.1381, 0.1563, 0.1788, 0.2132, 0.2143, 0.2516, 0.2746, 0.3324, 0.3092, 0.3094, 0.1789, 0.1059, 0.187, 0.1489, 0.0931, 0.1827, 0.2084, 0.181, 0.1801, 0.234, 0.2606, 0.3316, 0.2796, 0.2562, 0.178, 0.202, 0.193, 0.1226, 0.1241, 0.0591, 0.0229, 0.1388, 0.0823, 0.1203, 0.1498, 0.1708, 0.2796, 0.2621, 0.2788, 0.2789, 0.3531, 0.3698, 0.3936, 0.3762, 0.3844, 0.4215, 0.4504, 0.4617, 0.4622, 0.5288, 0.5096, 0.5273, 0.511, 0.5324, 0.6612, 0.7158, 0.8819, 0.9009, 0.9947, 1.1616, 1.1023, 1.209, 1.209, 1.0405, 1.0361, 0.9956, 1.0252, 1.1911, 1.1667, 1.1667, 1.3555, 1.4876, 1.488, 1.5871, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2022239250746581020", "ticker_idx": 0, "ts": "2026-02-13T09:19:43+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we raise our 1Q26F operating profit (OP) estimate for Hynix from KRW 29T to KRW 36T. We also raise our full-year 2026F commodity DRAM/NAND price growth forecasts from +126%/+115% YoY to +176%/+146% YoY. Accordingly, we revise up our 2026/27F operating profit (OP) estimates to KRW 189T / KRW 267T", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Nomura raises SK Hynix 2026/27F OP estimates sharply on commodity DRAM/NAND price beats; author shares the bullish revision.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura SK Hynix Comment: SK Hynix 2026/27F Operating Profit Forecast at $130.8B / $184.8B\n\n\"We estimate that commodity memory price increases in 1Q26 significantly exceeded our initial expectations. We estimate commodity DRAM/NAND prices rose +90%/+60% QoQ in 1Q, substantially surpassing our previous forecasts of DRAM +56% and NAND +40% QoQ. Reflecting this, we raise our 1Q26F operating profit (OP) estimate for Hynix from KRW 29T to KRW 36T. We also raise our full-year 2026F commodity DRAM/NAND price growth forecasts from +126%/+115% YoY to +176%/+146% YoY. Accordingly, we revise up our 2026/27F operating profit (OP) estimates to KRW 189T / KRW 267T. We expect Hynix to achieve DRAM/NAND operating profit margins (OPM) of 76%/57% in 2026F. Factoring in higher quarterly performance bonus costs, we estimate 2026F DRAM/NAND cost per bit will increase +26%/+18% YoY, respectively.\"", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.009090917501398676, "ret_signed_m1d": 0.009090917501398676, "alpha_spy_m1d": 0.009794907044792911, "alpha_c_m1d": 0.01303974479542469, "bench_spy_m1d": -0.0007039895433942345, "bench_c_m1d": -0.003948827294026014, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.07840906559675243, "ret_signed_p1w": 0.07840906559675243, "alpha_spy_p1w": 0.0671438863271887, "alpha_c_p1w": 0.06048010077602295, "bench_spy_p1w": 0.011265179269563719, "bench_c_p1w": 0.017928964820729476, "ret_p1m": 0.10886050437821937, "ret_signed_p1m": 0.10886050437821937, "alpha_spy_p1m": 0.12751829202689247, "alpha_c_p1m": 0.14270723191587253, "bench_spy_p1m": -0.018657787648673096, "bench_c_p1m": -0.033846727537653165, "ret_p3m": 1.2495979672804398, "ret_signed_p3m": 1.2495979672804398, "alpha_spy_p3m": 1.157794122000141, "alpha_c_p3m": 0.8455461173375236, "bench_spy_p3m": 0.09180384528029872, "bench_c_p3m": 0.4040518499429162, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3527, -0.3618, -0.3516, -0.4084, -0.4095, -0.4106, -0.405, -0.3818, -0.3977, -0.3886, -0.3659, -0.3727, -0.3841, -0.3818, -0.3443, -0.3568, -0.333, -0.358, -0.3511, -0.3705, -0.3977, -0.3739, -0.3727, -0.3784, -0.3409, -0.3364, -0.3318, -0.3318, -0.3193, -0.2727, -0.2602, -0.2602, -0.2602, -0.2307, -0.2091, -0.175, -0.1568, -0.1409, -0.1545, -0.1489, -0.1614, -0.1568, -0.1489, -0.1409, -0.1318, -0.1557, -0.1591, -0.142, -0.1284, -0.1636, -0.0909, -0.0443, -0.0216, 0.033, -0.0568, 0.0307, 0.0227, -0.0432, -0.0466, 0.008, -0.0045, -0.0227, 0.0091, 0.0, 0.0, 0.0, 0.0, 0.0159, 0.0784, 0.0807, 0.142, 0.1568, 0.2512, 0.2079, 0.2079, 0.069, -0.0334, 0.0713, 0.0519, -0.0482, 0.0679, 0.0872, 0.0588, 0.036, 0.1089, 0.1043, 0.2022, 0.1533, 0.1464, 0.0622, 0.1225, 0.1328, 0.0622, 0.0622, -0.0061, -0.0813, 0.0212, -0.0551, -0.0027, 0.0087, 0.0428, 0.176, 0.1362, 0.1692, 0.184, 0.2557, 0.2933, 0.3149, 0.2842, 0.3274, 0.3935, 0.3923, 0.3946, 0.3912, 0.4709, 0.48, 0.472, 0.4641, 0.4641, 0.6474, 0.6474, 0.8227, 0.883, 0.9194, 1.1403, 1.0891, 1.2496, 1.2428, 1.0709, 1.0948, 0.9866, 0.9866, 1.2086, 1.2098, 1.2098, 1.3361, 1.5536, 1.6064, 1.6565, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2035295851409801660", "ticker_idx": 0, "ts": "2026-03-21T10:01:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix will see significant — or rather, substantial — margin compression. TSMC's 3nm is far more expensive than 12nm, after all", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish SK Hynix margins: HBM4E on TSMC 3nm drives substantial cost/margin compression vs Samsung's 4nm approach.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Samsung is reportedly using an improved 4nm process for HBM4E compared to HBM4... but can it really beat SK Hynix's HBM4E, which uses TSMC's 3nm? I'm worried about this.\n\nOf course, SK Hynix will see significant — or rather, substantial — margin compression. TSMC's 3nm is far more expensive than 12nm, after all...", "reply_text": null, "parent_text": null, "quoted_text": "\"No Choice\" — Samsung's 'Bold Gambit' Deepens SK Hynix's Dilemma\n\n\"Our stance is to dramatically increase supply of premium HBM4 (6th-generation High Bandwidth Memory) to NVIDIA.\"\n\nThese were the words of Samsung Electronics' Hwang Sang-joon, EVP and Head of Memory Development https://t.co/lw7fcvGmkB", "parent_author": null, "ret_m1d": 0.0793140385745097, "ret_signed_m1d": -0.0793140385745097, "alpha_spy_m1d": -0.08970495301782022, "alpha_c_m1d": -0.09622939876709735, "bench_spy_m1d": -0.010390914443310528, "bench_c_m1d": -0.01691536019258766, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.056805987925097945, "ret_signed_p1d": -0.056805987925097945, "alpha_spy_p1d": -0.06016283767736441, "alpha_c_p1d": -0.048552688898048535, "bench_spy_p1d": -0.0033568497522664664, "bench_c_p1d": 0.00825329902704941, "ret_p1w": -0.06430864914176437, "ret_signed_p1w": 0.06430864914176437, "alpha_spy_p1w": 0.028588862139911297, "alpha_c_p1w": -0.009357509002552167, "bench_spy_p1w": -0.035719787001853076, "bench_c_p1w": -0.07366615814431654, "ret_p1m": 0.3118971392858827, "ret_signed_p1m": -0.3118971392858827, "alpha_spy_p1m": -0.23758908618051033, "alpha_c_p1m": -0.12460150086055743, "bench_spy_p1m": 0.07430805310537236, "bench_c_p1m": 0.18729563842532526, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3709, -0.3709, -0.3592, -0.3153, -0.3035, -0.3035, -0.3035, -0.2757, -0.2554, -0.2233, -0.2062, -0.1912, -0.204, -0.1987, -0.2105, -0.2062, -0.1987, -0.1912, -0.1826, -0.2051, -0.2083, -0.1923, -0.1794, -0.2126, -0.1441, -0.1003, -0.0789, -0.0275, -0.112, -0.0297, -0.0371, -0.0992, -0.1024, -0.0511, -0.0628, -0.0799, -0.05, -0.0585, -0.0585, -0.0585, -0.0585, -0.0436, 0.0153, 0.0174, 0.0752, 0.0891, 0.1779, 0.1372, 0.1372, 0.0064, -0.09, 0.0086, -0.0096, -0.104, 0.0054, 0.0236, -0.0032, -0.0247, 0.0439, 0.0397, 0.1318, 0.0857, 0.0793, 0.0, 0.0568, 0.0665, 0.0, 0.0, -0.0643, -0.135, -0.0386, -0.1104, -0.0611, -0.0504, -0.0182, 0.1072, 0.0697, 0.1008, 0.1147, 0.1822, 0.2176, 0.2379, 0.209, 0.2497, 0.3119, 0.3108, 0.313, 0.3098, 0.3848, 0.3934, 0.3859, 0.3783, 0.3783, 0.5509, 0.5509, 0.716, 0.7728, 0.8071, 1.015, 0.9668, 1.1179, 1.1115, 0.9496, 0.9721, 0.8703, 0.8703, 1.0793, 1.0804, 1.0804, 1.1994, 1.4041, 1.4538, 1.501, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2025543896890052795", "ticker_idx": 0, "ts": "2026-02-22T12:11:12+00:00", "symbol": "ONTO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintaining Buy rating, target price raised to $282", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF Securities reiterates Buy on ONTO, raises PT to $282 on HBM VPA, GAA/2.5D-3D share gains, and new 2027 catalysts (SoIC, panel packaging, CPO).", "resolved_tickers": ["ONTO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Securities Overseas Electronics & Communications》\n\nOnto Innovation 4Q25 Review: Upside Potential Remains\n\nMaintaining Buy rating, target price raised to $282: Onto Innovation's 4Q25 revenue came in at $267M, largely in line with expectations. The company guided 1Q26 revenue to a range of $275–285M, with 2Q26 revenue expected to surpass $300M. Memory and GAA capacity buildouts, combined with the company's continued market share gains in 2.5D/3D, underpin robust growth. Management indicated that in 2026, advanced packaging revenue will grow 30% and advanced node revenue will grow 15%. G5 qualification is expected to be completed in 1H26, and potential G5 applications present additional upside to earnings. We forecast Onto Innovation's 2026/2027 revenue at $1.3B/$1.5B respectively, and derive our $282 target price based on 30x 2027E P/E.\n\nMarket share gains accelerating: The company's market share expansion is progressing well. Management announced a large-scale HBM Volume Purchase Agreement (VPA) with a leading HBM customer worth over $240M, of which more than $60M is earmarked specifically for 3D bump metrology. This single HBM order alone is nearly equivalent to the company's entire AI packaging revenue for full-year 2025. In CoWoS, G5 has completed qualification at a foundry customer, and the company is recapturing sub-micron inspection market share. Additionally, Firefly and Iris thin-film metrology continue to drive share gains.\n\nNew catalysts emerging for 2027: We highlight several new growth vectors for Onto Innovation. 1) SoIC: NVIDIA's Feynman platform will adopt a 3D architecture combining SoIC and CoWoS. SoIC is poised to become a core growth driver in 2027, and the company's G5 product holds distinct advantages in both sub-micron inspection and 3D applications. 2) Panel-level packaging: Onto Innovation secured orders for JetStep and 8 Firefly systems in 4Q25 to support capacity at a panel-level packaging facility; G5 equipment is expected to be delivered to TSMC by year-end for a panel packaging pilot line. 3) CPO: The company has initiated early-stage engagements with CPO customers, including V-groove metrology solutions. CPO is expected to become a meaningful revenue contributor starting in 2027.\n\n$ONTO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.014865869852207059, "ret_signed_m1d": 0.014865869852207059, "alpha_spy_m1d": 0.004549208465143506, "alpha_c_m1d": 0.009658560534589133, "bench_spy_m1d": 0.010316661387063553, "bench_c_m1d": 0.005207309317617925, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.051303633937337834, "ret_signed_p1d": 0.051303633937337834, "alpha_spy_p1d": 0.044035181006942814, "alpha_c_p1d": 0.03609340649830428, "bench_spy_p1d": 0.00726845293039502, "bench_c_p1d": 0.015210227439033552, "ret_p1w": 0.022556731545141373, "ret_signed_p1w": 0.022556731545141373, "alpha_spy_p1w": 0.016709645832249764, "alpha_c_p1w": 0.03827556064227111, "bench_spy_p1w": 0.005847085712891609, "bench_c_p1w": -0.01571882909712974, "ret_p1m": 0.02916901641564862, "ret_signed_p1m": 0.02916901641564862, "alpha_spy_p1m": 0.06936039874011135, "alpha_c_p1m": 0.07346762213309699, "bench_spy_p1m": -0.04019138232446273, "bench_c_p1m": -0.04429860571744837, "ret_p3m": 0.2260833120988297, "ret_signed_p3m": 0.2260833120988297, "alpha_spy_p3m": 0.13470109526104523, "alpha_c_p3m": -0.14932842537194846, "bench_spy_p3m": 0.09138221683778447, "bench_c_p3m": 0.37541173747077816, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3413, -0.3413, -0.3286, -0.3106, -0.2778, -0.2585, -0.2523, -0.2619, -0.2473, -0.2374, -0.2104, -0.2479, -0.2667, -0.2683, -0.2766, -0.3081, -0.2885, -0.2667, -0.2531, -0.2444, -0.2409, -0.2409, -0.2471, -0.254, -0.2496, -0.2597, -0.2597, -0.2221, -0.1644, -0.1218, -0.1371, -0.1541, -0.1106, -0.1042, -0.0741, -0.0517, 0.0216, 0.0329, 0.0329, 0.0137, 0.0157, -0.0068, -0.0064, -0.0305, -0.0224, 0.0022, -0.0006, -0.0525, -0.0507, -0.0624, -0.1156, -0.1143, -0.0222, 0.0039, 0.021, 0.0431, 0.0069, 0.0187, 0.0187, 0.0331, 0.0317, 0.0567, 0.0149, 0.0, 0.0513, 0.0562, 0.0218, 0.0124, 0.0226, -0.0269, -0.0197, -0.0767, -0.1572, -0.1012, -0.0911, -0.0961, -0.1344, -0.1112, -0.0872, -0.078, -0.0533, -0.0177, -0.0613, -0.0162, 0.0292, 0.0366, -0.0398, -0.0423, -0.1124, -0.0383, -0.0071, 0.0107, 0.0107, 0.0372, 0.0275, 0.1278, 0.1581, 0.2059, 0.1846, 0.2267, 0.2141, 0.2528, 0.3635, 0.3838, 0.3605, 0.3944, 0.3708, 0.4437, 0.3988, 0.3243, 0.3334, 0.3837, 0.3737, 0.4048, 0.4303, 0.379, 0.2857, 0.335, 0.3483, 0.3034, 0.2989, 0.3003, 0.2745, 0.1936, 0.1634, 0.2339, 0.2261, 0.2298, 0.2298, 0.2857, 0.2516, 0.2137, 0.211, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2011310945860272254", "ticker_idx": 0, "ts": "2026-01-14T05:34:32+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I expect Broadcom will be next.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author expects Broadcom to invest in Intel next, following Apple and Nvidia; implicitly bullish on both INTC and AVGO.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "It’s being reported that Apple has invested in Intel, following Nvidia.\n\nI expect Broadcom will be next.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": "BREAKING: President Trump says $AAPL Apple has invested in Intel.\n\nAbout the $INTC Intel investment, he says:\n\n“They did a great deal for themselves because their stock has gone up. As soon as we (US Government) went in, $AAPL Apple went in, $NVDA Nvidia went in. A lot of smart https://t.co/0Y3BwaNwme", "parent_author": null, "ret_m1d": 0.043308012577061294, "ret_signed_m1d": 0.043308012577061294, "alpha_spy_m1d": 0.03836849634078865, "alpha_c_m1d": 0.03511954180120758, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009208758307613918, "ret_signed_p1d": 0.009208758307613918, "alpha_spy_p1d": 0.006485543911084468, "alpha_c_p1d": -0.01154570913649211, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": -0.03262834344741983, "ret_signed_p1w": -0.03262834344741983, "alpha_spy_p1w": -0.025443713447487193, "alpha_c_p1w": -0.06759676461089426, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": -0.02565537318828648, "ret_signed_p1m": -0.02565537318828648, "alpha_spy_p1m": -0.012488412610445199, "alpha_c_p1m": -0.07138726186839861, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.11961686498049962, "ret_signed_p3m": 0.11961686498049962, "alpha_spy_p3m": 0.1230734093297633, "alpha_c_p3m": -0.02198218064876012, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0258, 0.0255, 0.0062, -0.0007, 0.011, 0.0399, 0.0632, 0.0952, 0.1334, 0.1055, 0.0854, 0.0646, 0.0335, 0.0541, 0.0442, 0.0261, 0.0524, 0.0335, 0.0431, -0.0016, 0.0056, 0.0062, -0.0001, 0.0408, 0.0184, -0.001, 0.1099, 0.1306, 0.1675, 0.1675, 0.1833, 0.1337, 0.1205, 0.1177, 0.1189, 0.1459, 0.1778, 0.1931, 0.2127, 0.1933, 0.0569, -0.0021, 0.0022, -0.0426, -0.0313, -0.0005, 0.0046, 0.0277, 0.0304, 0.0304, 0.036, 0.028, 0.0293, 0.0183, 0.0183, 0.0227, 0.0104, 0.0114, 0.0106, -0.0218, 0.0149, 0.0362, 0.0433, 0.0, 0.0092, 0.0348, 0.0348, -0.0214, -0.0326, -0.0424, -0.0584, -0.0442, -0.0209, -0.0196, -0.0269, -0.0253, -0.0258, -0.0575, -0.0937, -0.0864, -0.0205, 0.0119, 0.0016, 0.0084, -0.0257, -0.0433, -0.0433, -0.0216, -0.0188, -0.0174, -0.0213, -0.0281, -0.0424, -0.0223, -0.0535, -0.0598, -0.062, -0.0766, -0.0658, -0.0209, -0.0277, 0.0172, 0.0079, 0.0049, -0.0115, -0.0522, -0.044, -0.0547, -0.0705, -0.059, -0.0864, -0.0491, -0.0616, -0.0601, -0.0877, -0.1135, -0.1349, -0.0875, -0.0757, -0.0726, -0.0726, -0.073, -0.0154, 0.0338, 0.0464, 0.0954, 0.1196, 0.1227, 0.1696, 0.1748, 0.1986, 0.1782, 0.1857, 0.2461, 0.2381, 0.2464, 0.233, 0.1788, 0.1954, 0.2307, 0.2421, 0.228, 0.26, 0.2543, 0.2164, 0.2678, 0.2631, 0.2362, 0.2288, 0.2966, 0.2536, 0.2404, 0.212, 0.2317, 0.2223, 0.221, 0.221, 0.2442, 0.2438, 0.2577, 0.3172, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2011310945860272254", "ticker_idx": 1, "ts": "2026-01-14T05:34:32+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It's being reported that Apple has invested in Intel, following Nvidia.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author expects Broadcom to invest in Intel next, following Apple and Nvidia; implicitly bullish on both INTC and AVGO.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "It’s being reported that Apple has invested in Intel, following Nvidia.\n\nI expect Broadcom will be next.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": "BREAKING: President Trump says $AAPL Apple has invested in Intel.\n\nAbout the $INTC Intel investment, he says:\n\n“They did a great deal for themselves because their stock has gone up. As soon as we (US Government) went in, $AAPL Apple went in, $NVDA Nvidia went in. A lot of smart https://t.co/0Y3BwaNwme", "parent_author": null, "ret_m1d": -0.029351401259163246, "ret_signed_m1d": -0.029351401259163246, "alpha_spy_m1d": -0.03429091749543589, "alpha_c_m1d": -0.03753987203501696, "bench_spy_m1d": 0.004939516236272645, "bench_c_m1d": 0.008188470775853718, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00821021173761649, "ret_signed_p1d": -0.00821021173761649, "alpha_spy_p1d": -0.01093342613414594, "alpha_c_p1d": -0.02896467918172252, "bench_spy_p1d": 0.0027232143965294497, "bench_c_p1d": 0.02075446744410603, "ret_p1w": 0.11350571922701325, "ret_signed_p1w": 0.11350571922701325, "alpha_spy_p1w": 0.12069034922694588, "alpha_c_p1w": 0.07853729806353882, "bench_spy_p1w": -0.007184629999932635, "bench_c_p1w": 0.03496842116347443, "ret_p1m": -0.04597704479356479, "ret_signed_p1m": -0.04597704479356479, "alpha_spy_p1m": -0.03281008421572351, "alpha_c_p1m": -0.09170893347367692, "bench_spy_p1m": -0.01316696057784128, "bench_c_p1m": 0.04573188868011213, "ret_p3m": 0.33784890541993917, "ret_signed_p3m": 0.33784890541993917, "alpha_spy_p3m": 0.34130544976920285, "alpha_c_p3m": 0.19624985979067944, "bench_spy_p3m": -0.0034565443492636794, "bench_c_p3m": 0.14159904562925973, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2404, -0.218, -0.2176, -0.2422, -0.2167, -0.2143, -0.1884, -0.1476, -0.1515, -0.1757, -0.1792, -0.1892, -0.2399, -0.2122, -0.2356, -0.2174, -0.2108, -0.2225, -0.2223, -0.2629, -0.2709, -0.2876, -0.2954, -0.2794, -0.3099, -0.2919, -0.2654, -0.2646, -0.2445, -0.2445, -0.1675, -0.1788, -0.1078, -0.1018, -0.1687, -0.15, -0.1728, -0.1687, -0.163, -0.189, -0.2239, -0.2301, -0.2342, -0.2601, -0.2553, -0.2443, -0.2535, -0.2539, -0.2578, -0.2578, -0.257, -0.2471, -0.2344, -0.2426, -0.2426, -0.1917, -0.1919, -0.1782, -0.125, -0.1562, -0.0651, -0.0956, -0.0294, 0.0, -0.0082, -0.0361, -0.0361, -0.0033, 0.1135, 0.1149, -0.0749, -0.1279, -0.0983, 0.0012, -0.0012, -0.0462, 0.0018, 0.0109, -0.0025, -0.0099, 0.0384, 0.0312, -0.0326, -0.0088, -0.046, -0.0396, -0.0396, -0.0521, -0.0669, -0.0842, -0.0946, -0.1045, -0.0534, -0.0378, -0.0669, -0.0638, -0.0661, -0.1154, -0.0644, -0.0569, -0.1088, -0.0644, -0.0398, -0.0152, -0.0712, -0.0606, -0.0608, -0.0956, -0.0757, -0.0521, -0.0995, -0.0967, -0.0956, -0.0316, -0.0948, -0.1147, -0.1546, -0.0942, -0.0142, 0.0341, 0.0341, 0.0423, 0.086, 0.21, 0.2668, 0.2804, 0.3378, 0.3097, 0.3329, 0.406, 0.406, 0.3485, 0.36, 0.3397, 0.3707, 0.6942, 0.7445, 0.7348, 0.9448, 0.9392, 1.0447, 0.9659, 1.2198, 1.3196, 1.25, 1.564, 1.6568, 1.4756, 1.469, 1.3795, 1.2326, 1.2202, 1.2742, 1.4417, 1.4323, 1.4598, 1.4598, 1.5353, 1.4994, 1.4813, 1.3539, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2010337650151248146", "ticker_idx": 0, "ts": "2026-01-11T13:07:00+00:00", "symbol": "AAPL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think the street is way too optimistic about Apple. Samsung and SK hynix are already planning to raise DRAM prices to Apple starting this month.", "tweet_type": "analysis", "timeline": "months", "summary": "Bearish AAPL: street too optimistic; DRAM price hikes from Samsung and SK Hynix will pressure Apple margins.", "resolved_tickers": ["AAPL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung's DS division has recently raised the price of DRAM supplied to Samsung's MX division by about 60–70%.\n\nThey’re pushing such an absurd condition even on MX, which is part of the same family— so just imagine how much they’re raising prices for other companies.\nI think the street is way too optimistic about Apple. Samsung and SK hynix are already planning to raise DRAM prices to Apple starting this month.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0033814384253491037, "ret_signed_m1d": 0.0033814384253491037, "alpha_spy_m1d": 0.0018134470715756024, "alpha_c_m1d": -0.0009785364000141072, "bench_spy_m1d": -0.0015679913537735013, "bench_c_m1d": -0.004359974825363211, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003073874731586601, "ret_signed_p1d": -0.003073874731586601, "alpha_spy_p1d": -0.005073349837456886, "alpha_c_p1d": -0.005185742416326633, "bench_spy_p1d": -0.0019994751058702853, "bench_c_p1d": -0.002111867684740032, "ret_p1w": -0.0181364483183738, "ret_signed_p1w": 0.0181364483183738, "alpha_spy_p1w": 0.013101620157812577, "alpha_c_p1w": 0.01016586771452177, "bench_spy_p1w": -0.005034828160561222, "bench_c_p1w": -0.007970580603852029, "ret_p1m": 0.0525881050152619, "ret_signed_p1m": -0.0525881050152619, "alpha_spy_p1m": -0.05696122497734135, "alpha_c_p1m": -0.08147284716413805, "bench_spy_p1m": -0.0043731199620794525, "bench_c_p1m": -0.028884742148876152, "ret_p3m": 0.0018586574480579099, "ret_signed_p3m": -0.0018586574480579099, "alpha_spy_p3m": -0.021124966032806225, "alpha_c_p3m": -0.03280999290175868, "bench_spy_p3m": -0.019266308584748315, "bench_c_p3m": -0.030951335453700768, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0428, -0.0501, -0.0315, 0.0067, 0.0087, -0.0079, -0.0035, 0.0089, 0.0319, 0.0326, 0.0353, 0.0418, 0.0379, 0.0328, 0.0366, 0.037, 0.0356, 0.0306, 0.0353, 0.0576, 0.0508, 0.0488, 0.0467, 0.0277, 0.0276, 0.0319, 0.0231, 0.0432, 0.0602, 0.0642, 0.0665, 0.0665, 0.0715, 0.0878, 0.0997, 0.0918, 0.0786, 0.0712, 0.0678, 0.0651, 0.0712, 0.0683, 0.0693, 0.0533, 0.0552, 0.0445, 0.0459, 0.0516, 0.0412, 0.0465, 0.0521, 0.0521, 0.0505, 0.0519, 0.0493, 0.0446, 0.0446, 0.0413, 0.0269, 0.0081, 0.0003, -0.0046, -0.0034, 0.0, 0.0031, -0.0011, -0.0078, -0.0181, -0.0181, -0.0521, -0.0484, -0.0457, -0.0469, -0.0186, -0.0076, -0.0146, -0.0076, -0.003, 0.0375, 0.0355, 0.0624, 0.0602, 0.0687, 0.0562, 0.0526, 0.0596, 0.0066, -0.0163, -0.0163, 0.0149, 0.0167, 0.0022, 0.0176, 0.0237, 0.0467, 0.0547, 0.0498, 0.0161, 0.0181, 0.0144, 0.0097, 0.0011, -0.0098, -0.0005, 0.0032, 0.0031, -0.0163, -0.038, -0.0276, -0.0222, -0.0387, -0.0425, -0.0462, -0.0328, -0.0322, -0.0284, -0.0274, -0.0431, -0.0514, -0.0239, -0.0168, -0.0157, -0.0157, -0.0044, -0.025, -0.0043, 0.0019, 0.0018, -0.0031, -0.0045, 0.0247, 0.0131, 0.0393, 0.0502, 0.0237, 0.0506, 0.0516, 0.0425, 0.0292, 0.0412, 0.0391, 0.0436, 0.0774, 0.0647, 0.093, 0.1058, 0.1055, 0.1281, 0.1267, 0.1349, 0.1505, 0.148, 0.1558, 0.1466, 0.1509, 0.1635, 0.1741, 0.1888, 0.1888, 0.1869, 0.1966, 0.203, 0.2013, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2028446029671235665", "ticker_idx": 0, "ts": "2026-03-02T12:23:14+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The reason we named NVIDIA—not Micron—as our top pick is that investor sentiment toward NVIDIA is more negative than toward Micron.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Morgan Stanley report naming NVDA as top pick; strong customer preference, 80%+ NVIDIA growth among ASIC/AMD users, Rubin demand in CY26.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley has published an excellent report.\n\nTo quote a few particularly memorable lines:\n\n“There appears to be a cohort of investors who are positioning 2026 as the exit point for memory.”\n\n“The reason we named NVIDIA—not Micron—as our top pick is that investor sentiment toward NVIDIA is more negative than toward Micron.”\n\n“This is by no means a negative call on MU.”\n\n“Based on our checks, we continue to see a strong preference for NVIDIA in many cases—even in areas where competitors emphasize lower total cost of ownership (TCO). In addition, our checks suggest that the two largest ASIC users and the two potentially largest AMD users are each likely to grow their business with NVIDIA by 80%+ in CY26.”\n\n“However, the product that every customer is most focused on—including ASIC and AMD users—is NVIDIA’s Rubin, which will ship in the second half. Customers will only be able to secure a portion of the volumes they want, so much of the competitive dynamic is still likely to be ‘Blackwell vs. competitors,’ and we expect Blackwell to hold up very well.”\n\n“While GPUs (processors) themselves may become less scarce soon, the real bottleneck could shift to other components and infrastructure (memory, storage, optics, power, CPUs, etc.). This transition could cause some near-term volatility in the stock, but it may ultimately work in NVIDIA’s favor from a share perspective.”", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0289894217953931, "ret_signed_m1d": -0.0289894217953931, "alpha_spy_m1d": -0.028421166111186724, "alpha_c_m1d": -0.028940159927413056, "bench_spy_m1d": -0.0005682556842063757, "bench_c_m1d": -4.926186798004384e-05, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.013316539505995872, "ret_signed_p1d": -0.013316539505995872, "alpha_spy_p1d": -0.004502201023200203, "alpha_c_p1d": 0.02440588608830796, "bench_spy_p1d": -0.00881433848279567, "bench_c_p1d": -0.03772242559430383, "ret_p1w": 0.0009315657110897657, "ret_signed_p1w": 0.0009315657110897657, "alpha_spy_p1w": 0.012747146634604634, "alpha_c_p1w": 0.03050911215166452, "bench_spy_p1w": -0.011815580923514868, "bench_c_p1w": -0.029577546440574753, "ret_p1m": -0.04422712742765267, "ret_signed_p1m": -0.04422712742765267, "alpha_spy_p1m": 0.005692614147531083, "alpha_c_p1m": 0.012344195496808208, "bench_spy_p1m": -0.04991974157518375, "bench_c_p1m": -0.056571322924460876, "ret_p3m": 0.17416482290802038, "ret_signed_p3m": 0.17416482290802038, "alpha_spy_p3m": 0.07177144374211242, "alpha_c_p3m": -0.3018311304097514, "bench_spy_p3m": 0.10239337916590796, "bench_c_p3m": 0.4759959533177718, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0159, 0.0049, -0.0004, 0.0168, 0.0136, 0.0071, -0.0085, -0.0409, -0.0339, -0.0261, -0.0632, -0.0457, -0.0082, 0.0066, 0.0369, 0.0336, 0.0336, 0.0441, 0.0315, 0.0277, 0.022, 0.022, 0.0349, 0.0309, 0.0261, 0.0363, 0.014, 0.013, 0.0135, 0.0182, 0.0036, 0.025, 0.0206, 0.0206, -0.0242, 0.0046, 0.0129, 0.0284, 0.0219, 0.0331, 0.0495, 0.055, 0.0474, 0.0172, -0.0117, -0.0454, -0.0581, 0.0161, 0.0414, 0.0332, 0.0415, 0.0244, 0.0018, 0.0018, 0.0136, 0.0301, 0.0297, 0.0402, 0.0497, 0.0568, 0.0717, 0.0132, -0.029, 0.0, -0.0133, 0.0031, 0.0047, -0.0255, 0.0009, 0.0125, 0.0195, 0.0037, -0.0122, 0.0041, -0.003, -0.0113, -0.0214, -0.0535, -0.0374, -0.0398, -0.0208, -0.0615, -0.0819, -0.0948, -0.0442, -0.0368, -0.0278, -0.0278, -0.0265, -0.0239, -0.0021, 0.0079, 0.0338, 0.0375, 0.0769, 0.0899, 0.087, 0.1053, 0.1074, 0.0954, 0.1098, 0.0941, 0.1414, 0.1871, 0.1682, 0.1468, 0.0937, 0.0876, 0.0877, 0.0769, 0.139, 0.1591, 0.1794, 0.2026, 0.21, 0.2376, 0.2919, 0.2348, 0.2184, 0.209, 0.2247, 0.203, 0.1801, 0.1801, 0.1775, 0.1651, 0.1742, 0.1571, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011620175536398356", "ticker_idx": 0, "ts": "2026-01-15T02:03:18+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND companies are basically going to be printing money", "tweet_type": "prediction", "timeline": "years", "summary": "Long NAND makers (SNDK, MU) on 3-year supply discipline + fat margins; short NAND equipment makers on weak demand outlook.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Over the next three years, NAND companies are basically going to be printing money. Just look at these insane gross margins.\n\n$SNDK $MU https://t.co/aGz3aW28io\n\n---\n\nThis implies two major things:\n\n1. NAND makers have little incentive to ramp supply, so the NAND market is likely to remain meaningfully undersupplied over the next three years.\n\n2. As a result, NAND-focused equipment makers are unlikely to see strong performance, and we would expect their results to remain relatively weak.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.052365343532307596, "ret_signed_m1d": -0.052365343532307596, "alpha_spy_m1d": -0.049649524892197316, "alpha_c_m1d": -0.047140488109642664, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.005224855422664931, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.010702778289834347, "ret_signed_p1d": 0.010702778289834347, "alpha_spy_p1d": 0.011540652280729757, "alpha_c_p1d": 0.009602897174043168, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.0010998811157911792, "ret_p1w": 0.23018281315343136, "ret_signed_p1w": 0.23018281315343136, "alpha_spy_p1w": 0.23489213002575748, "alpha_c_p1w": 0.23417019729016697, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": -0.0039873841367356055, "ret_p1m": 0.5310331651600271, "ret_signed_p1m": 0.5310331651600271, "alpha_spy_p1m": 0.5461868638030019, "alpha_c_p1m": 0.5715942221634245, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": -0.040561057003397405, "ret_p3m": 1.3078390296895388, "ret_signed_p3m": 1.3078390296895388, "alpha_spy_p3m": 1.3018922197022436, "alpha_c_p3m": 1.2895251478601693, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.01831388182936955, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6383, -0.6352, -0.6409, -0.5918, -0.5451, -0.5687, -0.5711, -0.5006, -0.5215, -0.5129, -0.4942, -0.5246, -0.471, -0.4925, -0.4148, -0.3452, -0.3364, -0.3082, -0.4048, -0.3789, -0.3503, -0.4015, -0.399, -0.5212, -0.5106, -0.4454, -0.4612, -0.4745, -0.4745, -0.4544, -0.4864, -0.4982, -0.525, -0.4788, -0.4417, -0.4491, -0.4637, -0.431, -0.4096, -0.4962, -0.5067, -0.4885, -0.4946, -0.4637, -0.4194, -0.411, -0.4016, -0.3889, -0.3889, -0.389, -0.4032, -0.413, -0.4199, -0.4199, -0.3274, -0.3303, -0.1457, -0.1361, -0.1825, -0.0778, -0.0488, -0.0475, -0.0524, 0.0, 0.0107, 0.0107, 0.1072, 0.2249, 0.2302, 0.1578, 0.1504, 0.1764, 0.2893, 0.3178, 0.4081, 0.6255, 0.6995, 0.4284, 0.408, 0.4611, 0.4256, 0.3235, 0.4645, 0.5401, 0.531, 0.531, 0.4431, 0.4671, 0.5177, 0.5882, 0.6286, 0.5603, 0.5453, 0.593, 0.5525, 0.5128, 0.3816, 0.4638, 0.382, 0.2886, 0.4386, 0.5123, 0.6016, 0.5121, 0.6167, 0.7194, 0.7598, 0.8417, 0.8866, 0.7342, 0.7166, 0.7165, 0.6564, 0.4739, 0.5048, 0.3989, 0.5525, 0.6927, 0.7144, 0.7144, 0.7707, 0.7369, 0.9082, 1.0809, 1.0813, 1.3275, 1.3078, 1.179, 1.2468, 1.2505, 1.231, 1.2077, 1.3924, 1.2784, 1.4189, 1.6151, 1.4493, 1.6005, 1.6794, 1.9005, 2.0688, 2.4364, 2.4454, 2.2743, 2.8177, 2.7815, 2.5481, 2.5364, 2.3788, 2.4396, 2.2573, 2.3801, 2.4028, 2.7685, 2.6133, 2.6133, 2.8842, 2.8851, 3.0114, 3.1418, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011620175536398356", "ticker_idx": 1, "ts": "2026-01-15T02:03:18+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": "NAND", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND companies are basically going to be printing money", "tweet_type": "prediction", "timeline": "years", "summary": "Long NAND makers (SNDK, MU) on 3-year supply discipline + fat margins; short NAND equipment makers on weak demand outlook.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Over the next three years, NAND companies are basically going to be printing money. Just look at these insane gross margins.\n\n$SNDK $MU https://t.co/aGz3aW28io\n\n---\n\nThis implies two major things:\n\n1. NAND makers have little incentive to ramp supply, so the NAND market is likely to remain meaningfully undersupplied over the next three years.\n\n2. As a result, NAND-focused equipment makers are unlikely to see strong performance, and we would expect their results to remain relatively weak.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009743643695028004, "ret_signed_m1d": -0.009743643695028004, "alpha_spy_m1d": -0.007027825054917725, "alpha_c_m1d": 0.010588834002643543, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.020332477697671547, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07759263054980647, "ret_signed_p1d": 0.07759263054980647, "alpha_spy_p1d": 0.07843050454070188, "alpha_c_p1d": 0.06755249294086618, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.010040137608940292, "ret_p1w": 0.1810592360281007, "ret_signed_p1w": 0.1810592360281007, "alpha_spy_p1w": 0.18576855290042682, "alpha_c_p1w": 0.16488910005046997, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": 0.01617013597763073, "ret_p1m": 0.22288569080886544, "ret_signed_p1m": 0.22288569080886544, "alpha_spy_p1m": 0.23803938945184022, "alpha_c_p1m": 0.19435463091100735, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": 0.02853105989785809, "ret_p3m": 0.3838802799912209, "ret_signed_p3m": 0.3838802799912209, "alpha_spy_p3m": 0.37793347000392563, "alpha_c_p3m": 0.24364669399606464, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.14023358599515623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.386, -0.3993, -0.4107, -0.3862, -0.3496, -0.3464, -0.3411, -0.327, -0.3348, -0.3355, -0.3031, -0.3526, -0.2948, -0.2923, -0.2935, -0.2478, -0.284, -0.2728, -0.2964, -0.2671, -0.2815, -0.3215, -0.3291, -0.402, -0.3842, -0.3351, -0.3333, -0.3163, -0.3163, -0.2978, -0.286, -0.2889, -0.3047, -0.327, -0.2956, -0.2668, -0.2505, -0.2169, -0.2325, -0.284, -0.2948, -0.3096, -0.3303, -0.262, -0.2104, -0.1787, -0.1796, -0.1487, -0.1487, -0.1543, -0.1255, -0.1307, -0.1522, -0.1522, -0.063, -0.0727, 0.0202, 0.0087, -0.0285, 0.0251, 0.0274, 0.0045, -0.0097, 0.0, 0.0776, 0.0776, 0.0843, 0.1559, 0.1811, 0.1872, 0.1558, 0.2187, 0.2931, 0.2946, 0.2325, 0.3005, 0.246, 0.1271, 0.1374, 0.1725, 0.1392, 0.1088, 0.219, 0.2297, 0.2229, 0.2229, 0.1876, 0.2505, 0.2398, 0.2719, 0.2505, 0.2417, 0.2744, 0.2345, 0.225, 0.2259, 0.1279, 0.1905, 0.1795, 0.1, 0.1565, 0.1975, 0.2438, 0.2041, 0.2659, 0.3124, 0.3715, 0.3716, 0.3198, 0.2563, 0.2012, 0.175, 0.135, 0.0559, 0.0612, -0.0437, 0.004, 0.0932, 0.0884, 0.0884, 0.1227, 0.1221, 0.2087, 0.2527, 0.2499, 0.2677, 0.3839, 0.3559, 0.3588, 0.3524, 0.3326, 0.3355, 0.4487, 0.4316, 0.4762, 0.5589, 0.4987, 0.5408, 0.5369, 0.6114, 0.7131, 0.9026, 0.981, 0.9217, 1.2194, 1.3636, 1.2782, 1.3883, 1.3062, 1.1536, 1.0254, 1.0766, 1.1754, 1.2649, 1.2319, 1.2319, 1.6624, 1.7591, 1.7446, 1.8857, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2011620175536398356", "ticker_idx": 2, "ts": "2026-01-15T02:03:18+00:00", "symbol": "NAND-focused equipment makers", "kind": "sector", "geography": null, "sub_sector": "NAND equipment", "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "NAND-focused equipment makers are unlikely to see strong performance, and we would expect their results to remain relatively weak", "tweet_type": "prediction", "timeline": "years", "summary": "Long NAND makers (SNDK, MU) on 3-year supply discipline + fat margins; short NAND equipment makers on weak demand outlook.", "resolved_tickers": ["AMAT", "LRCX", "6857.T", "KLAC"], "resolution_method": "basket", "resolution_reason": "NAND equipment makers: AMAT, Lam, TEL, KLA", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Over the next three years, NAND companies are basically going to be printing money. Just look at these insane gross margins.\n\n$SNDK $MU https://t.co/aGz3aW28io\n\n---\n\nThis implies two major things:\n\n1. NAND makers have little incentive to ramp supply, so the NAND market is likely to remain meaningfully undersupplied over the next three years.\n\n2. As a result, NAND-focused equipment makers are unlikely to see strong performance, and we would expect their results to remain relatively weak.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03498490241285013, "ret_signed_m1d": 0.03498490241285013, "alpha_spy_m1d": 0.03226908377273985, "alpha_c_m1d": 0.01465242471517858, "bench_spy_m1d": -0.002715818640110279, "bench_c_m1d": -0.020332477697671547, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01966951126159988, "ret_signed_p1d": -0.01966951126159988, "alpha_spy_p1d": -0.02050738525249529, "alpha_c_p1d": -0.00962937365265959, "bench_spy_p1d": -0.0008378739908954103, "bench_c_p1d": 0.010040137608940292, "ret_p1w": 0.0002124384286056602, "ret_signed_p1w": -0.0002124384286056602, "alpha_spy_p1w": -0.004921755300931779, "alpha_c_p1w": 0.01595769754902507, "bench_spy_p1w": -0.004709316872326119, "bench_c_p1w": 0.01617013597763073, "ret_p1m": 0.08733325236121639, "ret_signed_p1m": -0.08733325236121639, "alpha_spy_p1m": -0.10248695100419117, "alpha_c_p1m": -0.0588021924633583, "bench_spy_p1m": -0.015153698642974778, "bench_c_p1m": 0.02853105989785809, "ret_p3m": 0.21543230411418535, "ret_signed_p3m": -0.21543230411418535, "alpha_spy_p3m": -0.2094854941268901, "alpha_c_p3m": -0.07519871811902912, "bench_spy_p3m": 0.005946809987295243, "bench_c_p3m": 0.14023358599515623, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2754, -0.278, -0.2953, -0.2818, -0.2661, -0.2405, -0.2471, -0.1858, -0.1896, -0.1846, -0.1749, -0.2063, -0.1963, -0.2015, -0.2225, -0.1982, -0.2252, -0.2202, -0.232, -0.2528, -0.2461, -0.2637, -0.2436, -0.2572, -0.2796, -0.259, -0.2381, -0.2222, -0.2117, -0.2044, -0.2165, -0.1978, -0.1779, -0.1827, -0.1865, -0.1796, -0.1766, -0.1665, -0.1589, -0.1881, -0.192, -0.1984, -0.2215, -0.2052, -0.1861, -0.1679, -0.1707, -0.1611, -0.1651, -0.158, -0.1679, -0.177, -0.187, -0.187, -0.1521, -0.0991, -0.0648, -0.0885, -0.1105, -0.0613, -0.0497, -0.037, -0.035, 0.0, 0.0197, 0.0126, -0.0157, 0.0053, 0.0002, 0.0095, 0.0183, 0.0738, 0.0865, 0.1239, 0.0355, 0.0287, 0.0229, -0.031, -0.0317, 0.0244, 0.0597, 0.0575, 0.0836, 0.056, 0.0873, 0.087, 0.0883, 0.1049, 0.0901, 0.1016, 0.0958, 0.1173, 0.1653, 0.1292, 0.1064, 0.093, 0.0449, 0.0488, 0.0344, -0.0122, -0.0053, 0.0221, 0.0413, 0.0064, 0.0049, 0.0255, 0.0389, 0.0523, 0.061, 0.0523, 0.0499, 0.0704, 0.069, -0.0004, -0.013, -0.0595, -0.0211, 0.0298, 0.0057, 0.0105, 0.0242, 0.0342, 0.138, 0.1658, 0.1765, 0.1817, 0.2154, 0.2048, 0.2059, 0.2199, 0.2043, 0.1988, 0.2263, 0.2192, 0.2766, 0.2747, 0.2124, 0.2119, 0.2044, 0.19, 0.1919, 0.2301, 0.2822, 0.2684, 0.3155, 0.3081, 0.2842, 0.2972, 0.3151, 0.256, 0.2204, 0.1978, 0.2531, 0.2801, 0.2961, 0.3062, 0.3452, 0.3387, 0.3251, 0.3228, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2006699047214788677", "ticker_idx": 0, "ts": "2026-01-01T12:08:30+00:00", "symbol": "HBM makers", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Very bullish on HBM makers", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Very bullish on HBM makers; NVIDIA's bandwidth-driven architecture pursuit means HBM demand will keep growing with each GPU generation.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM makers: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Seeing this remark from Ian Buck, I definitely realized something. (Very bullish on HBM makers)\n\nQuote (Ian Buck): reducing token cost is the goal that every generation of NVIDIA architecture pursues.\n\nWe keep thinking about what technologies to adopt and scale, and where we should place our bets.\n\nIncluding which communities or partners we should bring technology in from.\n\nBecause the goal is not to improve performance by a few percent, but by “multiples.”\n\nThen wouldn’t today’s MoE models also be able to achieve multiple-fold performance gains?\n\nWe don’t hesitate to spend more money per GPU.\n\nHBM is far more expensive than conventional memory.\n\nBut that only increases cost by a few percent.\n\nOn the other hand, if you attach HBM and connect that enormous bandwidth to the floating-point compute units,\n\nend to end, you can achieve multiple-fold performance improvements overall.", "reply_text": null, "parent_text": null, "quoted_text": "Did you know NVIDIA also runs a podcast?\n\nIn it, Ian Buck explains MoE himself, and it’s genuinely fascinating.\n\nIt’s a perfect choice for the first podcast of the new year.\n\nThe link is below. https://t.co/RTs81lzC1v", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07227064895592512, "ret_signed_p1d": 0.07227064895592512, "alpha_spy_p1d": 0.07043760922247329, "alpha_c_p1d": 0.03570056945243645, "bench_spy_p1d": 0.0018330397334518356, "bench_c_p1d": 0.03657007950348867, "ret_p1w": 0.15490391400329662, "ret_signed_p1w": 0.15490391400329662, "alpha_spy_p1w": 0.14377356463080582, "alpha_c_p1w": 0.1025616880072997, "bench_spy_p1w": 0.0111303493724908, "bench_c_p1w": 0.052342225995996916, "ret_p1m": 0.3961856461483211, "ret_signed_p1m": 0.3961856461483211, "alpha_spy_p1m": 0.3814479133520954, "alpha_c_p1m": 0.27586802199932703, "bench_spy_p1m": 0.0147377327962257, "bench_c_p1m": 0.12031762414899405, "ret_p3m": 0.27449944577797564, "ret_signed_p3m": 0.27449944577797564, "alpha_spy_p3m": 0.3182052913653794, "alpha_c_p3m": 0.20988392166677577, "bench_spy_p3m": -0.04370584558740376, "bench_c_p3m": 0.06461552411119986, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3264, -0.3325, -0.3199, -0.3248, -0.3078, -0.3045, -0.3176, -0.2969, -0.2617, -0.2547, -0.2388, -0.249, -0.2491, -0.2469, -0.21, -0.1871, -0.1991, -0.1719, -0.1598, -0.1553, -0.1014, -0.1554, -0.1482, -0.1439, -0.1546, -0.1159, -0.1154, -0.1132, -0.1258, -0.1564, -0.1291, -0.171, -0.1817, -0.1944, -0.2291, -0.205, -0.1976, -0.1787, -0.1663, -0.1745, -0.1649, -0.1486, -0.1549, -0.1671, -0.1446, -0.1133, -0.1156, -0.0928, -0.1121, -0.1248, -0.1491, -0.1728, -0.1561, -0.1295, -0.1154, -0.0744, -0.07, -0.0569, -0.0569, -0.0371, 0.0037, 0.0084, 0.0, 0.0, 0.0723, 0.1049, 0.159, 0.1685, 0.1549, 0.1704, 0.1733, 0.1553, 0.1593, 0.1767, 0.2247, 0.2299, 0.2104, 0.249, 0.2743, 0.2823, 0.2541, 0.3322, 0.3905, 0.3966, 0.3962, 0.3544, 0.4199, 0.374, 0.3212, 0.3315, 0.3647, 0.3454, 0.3861, 0.4347, 0.4351, 0.4351, 0.4212, 0.446, 0.4734, 0.5145, 0.5152, 0.5588, 0.588, 0.6552, 0.6278, 0.6281, 0.4675, 0.3823, 0.4791, 0.4297, 0.3659, 0.4743, 0.5071, 0.4729, 0.4746, 0.5402, 0.5759, 0.6606, 0.5959, 0.5648, 0.4688, 0.4951, 0.4821, 0.3944, 0.3965, 0.315, 0.2745, 0.4183, 0.3507, 0.396, 0.4338, 0.4585, 0.5916, 0.5728, 0.5921, 0.5918, 0.6852, 0.7036, 0.7326, 0.7121, 0.7196, 0.763, 0.8029, 0.8167, 0.8187, 0.9011, 0.8745, 0.8987, 0.8782, 0.9075, 1.0635, 1.138, 1.3412, 1.3603, 1.4855, 1.689, 1.6142, 1.7438, 1.7419, 1.5334, 1.523, 1.4791, 1.5193, 1.72, 1.688, 1.688, 1.9323, 2.0906, 2.0878, 2.2146, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012071572610433038", "ticker_idx": 0, "ts": "2026-01-16T07:57:00+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Raising ASML's price target to €1,400", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises ASML PT to €1,400 on surging EUV demand; raises TSMC unit estimates and sees Intel/Samsung momentum building in 2027.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: Raising ASML’s price target to €1,400:\n\nDemand in advanced logic seeing upside, at least at TSMC.\n\nAt its 4Q25 print earlier this week, TSMC gave a 2026 capex guide to $52-56bn, up 32% y/y at the mid-point (2025: $40.9bn; 2024: $29.8bn), with 70-80% allocated to advanced processes. The company intimated that this could go higher again over the next few years. In alignment with this we raise our TSMC estimate to 29 units in 2026 (prior c.20) and raise our 2027 units from 28 to 40 as we pull forward more capacity build at A14 next year in readiness for 2028 use. We continue to see c.18 low NA EUV tools shipping to TSMC in 2025. Aside TSMC, we expect improving momentum with process development at Intel and Samsung to materialise and shift both to taking 5-6 EUV tools each for foundry/logic in 2027 given our expectations for emerging customer momentum. Altogether we expect c.52 tools to ship to logic/foundry in 2027 vs our prior estimate of 25-30.\n\n$ASML $TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015250183147876406, "ret_signed_m1d": -0.015250183147876406, "alpha_spy_m1d": -0.01608875976030477, "alpha_c_m1d": -0.00530984787297506, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04009595059520432, "ret_signed_p1d": -0.04009595059520432, "alpha_spy_p1d": -0.04009595059520432, "alpha_c_p1d": -0.04009595059520432, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.009424191472970689, "ret_signed_p1w": 0.009424191472970689, "alpha_spy_p1w": 0.012937506276860855, "alpha_c_p1w": 0.010173506639609697, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.02519170177438279, "ret_signed_p1m": 0.02519170177438279, "alpha_spy_p1m": 0.039519531342202985, "alpha_c_p1m": 0.006884585452559211, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.055218192619088, "ret_signed_p3m": 0.055218192619088, "alpha_spy_p3m": 0.04048324149446891, "alpha_c_p3m": -0.0761786540224465, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.239, -0.2559, -0.2408, -0.2351, -0.2234, -0.2226, -0.2123, -0.1968, -0.2134, -0.2062, -0.2135, -0.2233, -0.233, -0.2539, -0.2403, -0.2401, -0.2359, -0.2448, -0.2511, -0.2491, -0.2584, -0.2404, -0.2377, -0.2855, -0.2642, -0.2629, -0.221, -0.2313, -0.226, -0.206, -0.1954, -0.1746, -0.1798, -0.1847, -0.1748, -0.1836, -0.1895, -0.1939, -0.2072, -0.2024, -0.2216, -0.2512, -0.2352, -0.2276, -0.2324, -0.2262, -0.2298, -0.2298, -0.2298, -0.2231, -0.2132, -0.2106, -0.2106, -0.155, -0.0977, -0.091, -0.0989, -0.1321, -0.0733, -0.0692, -0.0553, -0.0711, -0.0153, 0.0, -0.0401, -0.0233, -0.0106, 0.0079, 0.0094, 0.0093, 0.0432, 0.0233, 0.0212, 0.0415, 0.0493, 0.0199, -0.0228, -0.0151, 0.0228, 0.0336, 0.0235, 0.0362, 0.0122, 0.0212, 0.0252, 0.0288, 0.0679, 0.0623, 0.0772, 0.0717, 0.0839, 0.1053, 0.0573, 0.0581, 0.0384, -0.0033, 0.0293, 0.0175, -0.016, -0.0155, 0.0295, 0.0285, 0.0127, 0.012, 0.0273, 0.0242, 0.0218, 0.0025, -0.0321, 0.0084, 0.0309, 0.0394, 0.0001, -0.0165, -0.046, -0.0398, 0.0188, -0.004, -0.004, -0.004, -0.0445, 0.0403, 0.0593, 0.0895, 0.0808, 0.1017, 0.0552, 0.0489, 0.0672, 0.0681, 0.0612, 0.0722, 0.0499, 0.0744, 0.0426, 0.0075, 0.0254, 0.051, 0.051, 0.0209, 0.0567, 0.122, 0.1177, 0.1344, 0.1226, 0.0883, 0.141, 0.1753, 0.1234, 0.0871, 0.0739, 0.1461, 0.1566, 0.2115, 0.23, 0.1949, 0.1834, 0.196, 0.1906, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012071572610433038", "ticker_idx": 1, "ts": "2026-01-16T07:57:00+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we raise our TSMC estimate to 29 units in 2026 (prior c.20) and raise our 2027 units from 28 to 40", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises ASML PT to €1,400 on surging EUV demand; raises TSMC unit estimates and sees Intel/Samsung momentum building in 2027.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: Raising ASML’s price target to €1,400:\n\nDemand in advanced logic seeing upside, at least at TSMC.\n\nAt its 4Q25 print earlier this week, TSMC gave a 2026 capex guide to $52-56bn, up 32% y/y at the mid-point (2025: $40.9bn; 2024: $29.8bn), with 70-80% allocated to advanced processes. The company intimated that this could go higher again over the next few years. In alignment with this we raise our TSMC estimate to 29 units in 2026 (prior c.20) and raise our 2027 units from 28 to 40 as we pull forward more capacity build at A14 next year in readiness for 2028 use. We continue to see c.18 low NA EUV tools shipping to TSMC in 2025. Aside TSMC, we expect improving momentum with process development at Intel and Samsung to materialise and shift both to taking 5-6 EUV tools each for foundry/logic in 2027 given our expectations for emerging customer momentum. Altogether we expect c.52 tools to ship to logic/foundry in 2027 vs our prior estimate of 25-30.\n\n$ASML $TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0022195632788257758, "ret_signed_m1d": -0.0022195632788257758, "alpha_spy_m1d": -0.0030581398912541413, "alpha_c_m1d": 0.0077207719960755705, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.021991847330370295, "ret_signed_p1w": -0.021991847330370295, "alpha_spy_p1w": -0.01847853252648013, "alpha_c_p1w": -0.021242532163731287, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": 0.06997661919389087, "ret_signed_p1m": 0.06997661919389087, "alpha_spy_p1m": 0.08430444876171106, "alpha_c_p1m": 0.051669502872067286, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.09858925941593832, "ret_signed_p3m": 0.09858925941593832, "alpha_spy_p3m": 0.08385430829131924, "alpha_c_p3m": -0.03280758722559618, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1422, -0.1586, -0.1532, -0.1409, -0.1313, -0.1217, -0.1114, -0.1168, -0.1249, -0.112, -0.1435, -0.1447, -0.1575, -0.1655, -0.14, -0.1519, -0.1535, -0.178, -0.1704, -0.1786, -0.1905, -0.1775, -0.1917, -0.1988, -0.1709, -0.1708, -0.1554, -0.1554, -0.1509, -0.1621, -0.1492, -0.1394, -0.1468, -0.1416, -0.1207, -0.1163, -0.0967, -0.1097, -0.1471, -0.1596, -0.1622, -0.1911, -0.1686, -0.1561, -0.1435, -0.1327, -0.1273, -0.1273, -0.1155, -0.1211, -0.1251, -0.1125, -0.1125, -0.0666, -0.0588, -0.0437, -0.0693, -0.0712, -0.0548, -0.031, -0.0327, -0.0447, -0.0022, 0.0, 0.0, -0.0445, -0.0475, -0.0439, -0.022, -0.0283, -0.0119, -0.0003, -0.0083, -0.0346, -0.003, -0.0194, -0.0487, -0.0341, 0.0188, 0.038, 0.057, 0.0926, 0.0751, 0.07, 0.07, 0.0637, 0.058, 0.0525, 0.0822, 0.0807, 0.1266, 0.1324, 0.1005, 0.094, 0.078, 0.0313, 0.0439, 0.0335, -0.0103, 0.0184, 0.0137, 0.0355, -0.0166, -0.0119, -0.0063, 0.0133, -0.0055, -0.0078, -0.0357, -0.0088, 0.0053, 0.0185, -0.0449, -0.043, -0.073, -0.0102, 0.0002, -0.007, -0.007, 0.0009, 0.0114, 0.0716, 0.0704, 0.0854, 0.0824, 0.1126, 0.0986, 0.0642, 0.0851, 0.0726, 0.078, 0.1347, 0.1207, 0.1787, 0.1861, 0.1491, 0.1534, 0.16, 0.1647, 0.1762, 0.1551, 0.2286, 0.213, 0.2057, 0.1848, 0.1635, 0.1709, 0.2234, 0.1843, 0.1597, 0.1499, 0.1763, 0.1925, 0.1848, 0.1848, 0.2076, 0.2381, 0.2443, 0.2256, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2012071572610433038", "ticker_idx": 2, "ts": "2026-01-16T07:57:00+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we expect improving momentum with process development at Intel and Samsung to materialise and shift both to taking 5-6 EUV tools each for foundry/logic in 2027", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Morgan Stanley raises ASML PT to €1,400 on surging EUV demand; raises TSMC unit estimates and sees Intel/Samsung momentum building in 2027.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley: Raising ASML’s price target to €1,400:\n\nDemand in advanced logic seeing upside, at least at TSMC.\n\nAt its 4Q25 print earlier this week, TSMC gave a 2026 capex guide to $52-56bn, up 32% y/y at the mid-point (2025: $40.9bn; 2024: $29.8bn), with 70-80% allocated to advanced processes. The company intimated that this could go higher again over the next few years. In alignment with this we raise our TSMC estimate to 29 units in 2026 (prior c.20) and raise our 2027 units from 28 to 40 as we pull forward more capacity build at A14 next year in readiness for 2028 use. We continue to see c.18 low NA EUV tools shipping to TSMC in 2025. Aside TSMC, we expect improving momentum with process development at Intel and Samsung to materialise and shift both to taking 5-6 EUV tools each for foundry/logic in 2027 given our expectations for emerging customer momentum. Altogether we expect c.52 tools to ship to logic/foundry in 2027 vs our prior estimate of 25-30.\n\n$ASML $TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02896083127908855, "ret_signed_m1d": 0.02896083127908855, "alpha_spy_m1d": 0.028122254666660185, "alpha_c_m1d": 0.0389011665539899, "bench_spy_m1d": 0.0008385766124283656, "bench_c_m1d": -0.009940335274901346, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": -0.04024700652673918, "ret_signed_p1w": -0.04024700652673918, "alpha_spy_p1w": -0.036733691722849016, "alpha_c_p1w": -0.03949769136010017, "bench_spy_p1w": -0.0035133148038901663, "bench_c_p1w": -0.0007493151666390085, "ret_p1m": -0.0036200632934322297, "ret_signed_p1m": -0.0036200632934322297, "alpha_spy_p1m": 0.010707766274387964, "alpha_c_p1m": -0.02192717961525581, "bench_spy_p1m": -0.014327829567820194, "bench_c_p1m": 0.01830711632182358, "ret_p3m": 0.3828791249461223, "ret_signed_p3m": 0.3828791249461223, "alpha_spy_p3m": 0.3681441738215032, "alpha_c_p3m": 0.2514822783045878, "bench_spy_p3m": 0.014734951124619089, "bench_c_p3m": 0.1313968466415345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1882, -0.2138, -0.1874, -0.1848, -0.158, -0.1156, -0.1197, -0.1448, -0.1484, -0.1589, -0.2115, -0.1827, -0.207, -0.188, -0.1812, -0.1934, -0.1931, -0.2353, -0.2436, -0.2609, -0.269, -0.2523, -0.2841, -0.2653, -0.2379, -0.237, -0.2161, -0.2161, -0.1363, -0.148, -0.0743, -0.0681, -0.1376, -0.1182, -0.1418, -0.1376, -0.1316, -0.1586, -0.1948, -0.2012, -0.2055, -0.2323, -0.2274, -0.2159, -0.2255, -0.2259, -0.23, -0.23, -0.2291, -0.2189, -0.2057, -0.2142, -0.2142, -0.1614, -0.1616, -0.1474, -0.0922, -0.1246, -0.03, -0.0618, 0.007, 0.0375, 0.029, 0.0, 0.0, 0.0341, 0.1552, 0.1567, -0.0402, -0.0952, -0.0645, 0.0388, 0.0362, -0.0104, 0.0394, 0.0488, 0.0349, 0.0273, 0.0773, 0.0698, 0.0036, 0.0283, -0.0102, -0.0036, -0.0036, -0.0166, -0.0319, -0.0498, -0.0607, -0.0709, -0.0179, -0.0017, -0.0319, -0.0287, -0.0311, -0.0822, -0.0294, -0.0215, -0.0754, -0.0294, -0.0038, 0.0217, -0.0364, -0.0253, -0.0256, -0.0618, -0.0411, -0.0166, -0.0658, -0.0628, -0.0618, 0.0047, -0.0609, -0.0816, -0.1229, -0.0603, 0.0228, 0.0728, 0.0728, 0.0813, 0.1267, 0.2553, 0.3143, 0.3284, 0.388, 0.3588, 0.3829, 0.4587, 0.4587, 0.3991, 0.411, 0.3899, 0.4221, 0.7577, 0.8098, 0.7998, 1.0177, 1.0119, 1.1214, 1.0396, 1.303, 1.4065, 1.3343, 1.6601, 1.7564, 1.5684, 1.5615, 1.4687, 1.3162, 1.3034, 1.3595, 1.5332, 1.5234, 1.552, 1.552, 1.6303, 1.5931, 1.5743, 1.4421, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2022944089495449653", "ticker_idx": 0, "ts": "2026-02-15T08:00:30+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bernstein remains very bearish on Kioxia. They've effectively issued a sell call and are recommending investors take profits. (Target price: ¥7,000 vs. current share price: ¥22,845.)", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Bernstein's sell call on Kioxia with PT ¥7,000 vs ¥22,845 current price; author adopts the bearish direction by relaying it.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Bernstein remains very bearish on Kioxia.\n\nThey’ve effectively issued a sell call and are recommending investors take profits. (Target price: ¥7,000 vs. current share price: ¥22,845.) https://t.co/b9MEVySUmM\n\n---\n\nI still don’t really understand why Bernstein sees it this way. For context, Bernstein has been warning about YMTC for a long time, and they’ve generally had a negative view on memory stocks.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.024439461883408065, "ret_signed_m1d": -0.024439461883408065, "alpha_spy_m1d": -0.024439461883408065, "alpha_c_m1d": -0.024439461883408065, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003139013452914785, "ret_signed_p1d": -0.003139013452914785, "alpha_spy_p1d": -0.0015255469821897627, "alpha_c_p1d": -0.003654051848012929, "bench_spy_p1d": 0.0016134664707250224, "bench_c_p1d": -0.0005150383950981441, "ret_p1w": -0.07802690582959637, "ret_signed_p1w": 0.07802690582959637, "alpha_spy_p1w": 0.07896573809748952, "alpha_c_p1w": 0.09068265891176175, "bench_spy_p1w": 0.0009388322678931527, "bench_c_p1w": 0.012655753082165377, "ret_p1m": -0.030941704035874484, "ret_signed_p1m": 0.030941704035874484, "alpha_spy_p1m": 0.014865390923014421, "alpha_c_p1m": 0.00435483975294515, "bench_spy_p1m": -0.016076313112860063, "bench_c_p1m": -0.026586864282929334, "ret_p3m": 1.1730941704035875, "ret_signed_p3m": -1.1730941704035875, "alpha_spy_p3m": -1.0726713457734693, "alpha_c_p3m": -0.7546206466031027, "bench_spy_p3m": 0.1004228246301182, "bench_c_p3m": 0.41847352380048486, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5141, -0.4917, -0.5502, -0.5502, -0.5582, -0.6239, -0.5942, -0.5782, -0.6017, -0.5867, -0.5959, -0.5933, -0.5771, -0.5448, -0.5572, -0.5753, -0.5691, -0.5565, -0.5874, -0.6104, -0.583, -0.5735, -0.5812, -0.546, -0.5537, -0.5244, -0.5159, -0.4881, -0.5213, -0.5321, -0.5321, -0.5321, -0.5321, -0.491, -0.4798, -0.4309, -0.417, -0.4309, -0.4309, -0.3863, -0.4045, -0.3888, -0.3386, -0.3179, -0.3182, -0.2601, -0.1969, -0.2226, -0.2175, -0.1726, -0.1504, -0.137, -0.0422, -0.1767, -0.0679, -0.0628, -0.1242, -0.1437, -0.1525, -0.1549, -0.1549, -0.0504, 0.0244, 0.0, 0.0031, -0.0395, -0.0471, -0.078, -0.078, -0.0013, -0.0374, -0.0475, -0.0489, -0.0352, -0.0942, -0.137, -0.0897, -0.1043, -0.1915, -0.1224, -0.0413, -0.0473, -0.0554, 0.0141, -0.0309, 0.0489, 0.0027, 0.0027, -0.0374, -0.0538, 0.0065, -0.0509, -0.091, -0.1022, -0.1444, -0.0224, -0.054, -0.0202, 0.0224, 0.0435, 0.2377, 0.2422, 0.3516, 0.4022, 0.5695, 0.4534, 0.5188, 0.3691, 0.3682, 0.4682, 0.5605, 0.5865, 0.5507, 0.5906, 0.6287, 0.6287, 0.6843, 0.6327, 0.6327, 0.6327, 0.6327, 0.9466, 0.9951, 1.0601, 1.0673, 1.2646, 1.1731, 0.9933, 1.3072, 1.2318, 1.3, 1.4816, 1.574, 1.935, 1.8009, 1.7152, 1.748, 1.9529, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013091516840116729", "ticker_idx": 0, "ts": "2026-01-19T03:29:53+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's share of the advanced-node foundry market is expected to decline from 95% to 90%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Deutsche Bank view: TSMC capacity crunch forces top customers to Samsung/Intel alternatives, reducing TSMC's advanced-node share from 95% to 90%.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Deutsche Bank: As TSMC's capacity fills up, Samsung and Intel are expected to benefit from the \"6 major customers\" including Apple and Nvidia\n\nDeutsche Bank stated in a recently released research report that global semiconductor foundry giant TSMC is facing a \"happy dilemma.\" Its 3nm process production capacity is extremely tight, with orders fully booked not only through 2026 but extending into 2027, leaving TSMC no choice but to significantly increase its capital expenditure plans.\n\nThis capacity bottleneck is reshaping the market landscape, prompting TSMC's top-tier customers—who have traditionally relied heavily on the company—to seriously consider Samsung and Intel as alternatives, Deutsche Bank noted.\n\nAnalysts including Robert Sanders pointed out in the report that TSMC has guided 2026 capital expenditures at $52–56 billion. This substantially exceeds Deutsche Bank's estimate of $50 billion and the market consensus of $46 billion.\n\nThe analysts assessed that this guidance effectively acknowledges that TSMC's 2025 capital expenditure plan was \"too conservative\" relative to the explosive AI demand. The current situation goes beyond mere excess demand for CoWoS packaging capacity; it reflects a severe supply shortage in core wafer fabrication capacity—particularly the 3nm process.\n\nThis supply-demand imbalance is creating a direct market overflow effect. Despite Samsung and Intel having a \"chequered record\" in foundry performance, Deutsche Bank emphasized that the \"6 major customers\"—Apple, Nvidia, AMD, Broadcom, Qualcomm, and MediaTek—have no choice but to seek alternative production capacity. As a result, TSMC's share of the advanced-node foundry market is expected to decline from 95% to 90%.\n\nDeutsche Bank noted that, given TSMC's full capacity booking for 2026 and order backlog extending into 2027, the sharp increase in spending is not surprising.\n\nPreviously, the initial bottleneck from the AI demand surge was concentrated in CoWoS packaging, causing some orders to overflow to ASE and Amkor. However, the situation has now become more severe, with demand significantly outstripping supply—especially for the 3nm process.\n\nAccording to Deutsche Bank estimates, TSMC plans to expand 3nm monthly production capacity to 190,000 wafers (190k wpm) by the end of 2026, but this will still fall short of customer demand.\n\nDeutsche Bank remarked, \"It is common for customers to want more capacity under TSMC's high utilization rates, but this level is unprecedented.\"\n\nDue to the extreme capacity tightness, TSMC is adopting a more aggressive strategy. Deutsche Bank reported that TSMC is delaying new 3nm development projects and encouraging customers to shift more 2027/2028 mass-production products to the 2nm GAA process. At the same time, TSMC is using pricing strategies to \"facilitate\" this transition, which is expected to be reflected in strong guidance for Q1 2026.\n\nIn this capacity competition, Deutsche Bank analyzed that Samsung's Taylor fab (SF2P) is more likely to be preferred over Intel.\n\nThe report stated, \"For customers seeking alternative supply, Samsung's Taylor fab is more likely to become the first destination.\"\n\nSpecifically, Qualcomm and AMD are the most likely to consider Samsung, while previous discussions indicate that Apple and Broadcom are evaluating Intel. However, the analyst added that although Intel has potential with its 14A process, \"there is still much work to be done.\"\n\n$TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04450933601066909, "ret_signed_p1d": 0.04450933601066909, "alpha_spy_p1d": 0.02415255028712404, "alpha_c_p1d": 0.01953368818193768, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": -0.02830025856620666, "ret_signed_p1w": 0.02830025856620666, "alpha_spy_p1w": 0.029847249693682132, "alpha_c_p1w": 0.024379063564385017, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": 0.06366829733764123, "ret_signed_p1m": -0.06366829733764123, "alpha_spy_p1m": -0.07640577790734238, "alpha_c_p1m": -0.04588564827872488, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.06417597339062442, "ret_signed_p3m": -0.06417597339062442, "alpha_spy_p3m": -0.046947502276743824, "alpha_c_p3m": 0.07171645937223037, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1586, -0.1532, -0.1409, -0.1313, -0.1217, -0.1114, -0.1168, -0.1249, -0.112, -0.1435, -0.1447, -0.1575, -0.1655, -0.14, -0.1519, -0.1535, -0.178, -0.1704, -0.1786, -0.1905, -0.1775, -0.1917, -0.1988, -0.1709, -0.1708, -0.1554, -0.1554, -0.1509, -0.1621, -0.1492, -0.1394, -0.1468, -0.1416, -0.1207, -0.1163, -0.0967, -0.1097, -0.1471, -0.1596, -0.1622, -0.1911, -0.1686, -0.1561, -0.1435, -0.1327, -0.1273, -0.1273, -0.1155, -0.1211, -0.1251, -0.1125, -0.1125, -0.0666, -0.0588, -0.0437, -0.0693, -0.0712, -0.0548, -0.031, -0.0327, -0.0447, -0.0022, 0.0, 0.0, -0.0445, -0.0475, -0.0439, -0.022, -0.0283, -0.0119, -0.0003, -0.0083, -0.0346, -0.003, -0.0194, -0.0487, -0.0341, 0.0188, 0.038, 0.057, 0.0926, 0.0751, 0.07, 0.07, 0.0637, 0.058, 0.0525, 0.0822, 0.0807, 0.1266, 0.1324, 0.1005, 0.094, 0.078, 0.0313, 0.0439, 0.0335, -0.0103, 0.0184, 0.0137, 0.0355, -0.0166, -0.0119, -0.0063, 0.0133, -0.0055, -0.0078, -0.0357, -0.0088, 0.0053, 0.0185, -0.0449, -0.043, -0.073, -0.0102, 0.0002, -0.007, -0.007, 0.0009, 0.0114, 0.0716, 0.0704, 0.0854, 0.0824, 0.1126, 0.0986, 0.0642, 0.0851, 0.0726, 0.078, 0.1347, 0.1207, 0.1787, 0.1861, 0.1491, 0.1534, 0.16, 0.1647, 0.1762, 0.1551, 0.2286, 0.213, 0.2057, 0.1848, 0.1635, 0.1709, 0.2234, 0.1843, 0.1597, 0.1499, 0.1763, 0.1925, 0.1848, 0.1848, 0.2076, 0.2381, 0.2443, 0.2256, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013091516840116729", "ticker_idx": 1, "ts": "2026-01-19T03:29:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Samsung and Intel are expected to benefit from the '6 major customers' including Apple and Nvidia", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Deutsche Bank view: TSMC capacity crunch forces top customers to Samsung/Intel alternatives, reducing TSMC's advanced-node share from 95% to 90%.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Deutsche Bank: As TSMC's capacity fills up, Samsung and Intel are expected to benefit from the \"6 major customers\" including Apple and Nvidia\n\nDeutsche Bank stated in a recently released research report that global semiconductor foundry giant TSMC is facing a \"happy dilemma.\" Its 3nm process production capacity is extremely tight, with orders fully booked not only through 2026 but extending into 2027, leaving TSMC no choice but to significantly increase its capital expenditure plans.\n\nThis capacity bottleneck is reshaping the market landscape, prompting TSMC's top-tier customers—who have traditionally relied heavily on the company—to seriously consider Samsung and Intel as alternatives, Deutsche Bank noted.\n\nAnalysts including Robert Sanders pointed out in the report that TSMC has guided 2026 capital expenditures at $52–56 billion. This substantially exceeds Deutsche Bank's estimate of $50 billion and the market consensus of $46 billion.\n\nThe analysts assessed that this guidance effectively acknowledges that TSMC's 2025 capital expenditure plan was \"too conservative\" relative to the explosive AI demand. The current situation goes beyond mere excess demand for CoWoS packaging capacity; it reflects a severe supply shortage in core wafer fabrication capacity—particularly the 3nm process.\n\nThis supply-demand imbalance is creating a direct market overflow effect. Despite Samsung and Intel having a \"chequered record\" in foundry performance, Deutsche Bank emphasized that the \"6 major customers\"—Apple, Nvidia, AMD, Broadcom, Qualcomm, and MediaTek—have no choice but to seek alternative production capacity. As a result, TSMC's share of the advanced-node foundry market is expected to decline from 95% to 90%.\n\nDeutsche Bank noted that, given TSMC's full capacity booking for 2026 and order backlog extending into 2027, the sharp increase in spending is not surprising.\n\nPreviously, the initial bottleneck from the AI demand surge was concentrated in CoWoS packaging, causing some orders to overflow to ASE and Amkor. However, the situation has now become more severe, with demand significantly outstripping supply—especially for the 3nm process.\n\nAccording to Deutsche Bank estimates, TSMC plans to expand 3nm monthly production capacity to 190,000 wafers (190k wpm) by the end of 2026, but this will still fall short of customer demand.\n\nDeutsche Bank remarked, \"It is common for customers to want more capacity under TSMC's high utilization rates, but this level is unprecedented.\"\n\nDue to the extreme capacity tightness, TSMC is adopting a more aggressive strategy. Deutsche Bank reported that TSMC is delaying new 3nm development projects and encouraging customers to shift more 2027/2028 mass-production products to the 2nm GAA process. At the same time, TSMC is using pricing strategies to \"facilitate\" this transition, which is expected to be reflected in strong guidance for Q1 2026.\n\nIn this capacity competition, Deutsche Bank analyzed that Samsung's Taylor fab (SF2P) is more likely to be preferred over Intel.\n\nThe report stated, \"For customers seeking alternative supply, Samsung's Taylor fab is more likely to become the first destination.\"\n\nSpecifically, Qualcomm and AMD are the most likely to consider Samsung, while previous discussions indicate that Apple and Broadcom are evaluating Intel. However, the analyst added that although Intel has potential with its 14A process, \"there is still much work to be done.\"\n\n$TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0026791567436211805, "ret_signed_m1d": -0.0026791567436211805, "alpha_spy_m1d": -0.0026791567436211805, "alpha_c_m1d": -0.0026791567436211805, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02746148771171364, "ret_signed_p1d": -0.02746148771171364, "alpha_spy_p1d": -0.0071047019881685936, "alpha_c_p1d": -0.001503561462741887, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.025957926248971752, "ret_p1w": 0.018754202077025495, "ret_signed_p1w": 0.018754202077025495, "alpha_spy_p1w": 0.017207210949550023, "alpha_c_p1w": 0.015595203300232674, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": 0.0031589987767928207, "ret_p1m": 0.21366377280264204, "ret_signed_p1m": 0.21366377280264204, "alpha_spy_p1m": 0.22640125337234318, "alpha_c_p1m": 0.25582827553940934, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": -0.042164502736767306, "ret_p3m": 0.45981374999829594, "ret_signed_p3m": 0.45981374999829594, "alpha_spy_p3m": 0.44258527888441535, "alpha_c_p3m": 0.41456568256379644, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.045248067434499495, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3428, -0.3568, -0.3414, -0.3201, -0.3368, -0.3301, -0.3061, -0.2835, -0.2595, -0.3008, -0.3294, -0.3388, -0.3474, -0.3294, -0.3101, -0.3128, -0.3148, -0.3521, -0.3294, -0.3481, -0.3568, -0.3294, -0.3681, -0.3554, -0.3381, -0.3148, -0.3101, -0.3301, -0.3281, -0.3108, -0.3035, -0.2995, -0.2775, -0.2701, -0.2775, -0.2801, -0.2848, -0.2741, -0.3015, -0.3148, -0.2808, -0.2828, -0.2915, -0.2635, -0.2568, -0.2595, -0.2595, -0.2201, -0.1996, -0.1969, -0.1969, -0.1969, -0.1393, -0.075, -0.0697, -0.0556, -0.0703, -0.069, -0.0703, -0.0784, -0.0603, -0.0362, -0.0027, 0.0, -0.0275, 0.0013, 0.0201, 0.0188, 0.0188, 0.0683, 0.0877, 0.0764, 0.075, 0.0074, 0.1219, 0.1326, 0.067, 0.0623, 0.1145, 0.1105, 0.1239, 0.1962, 0.2137, 0.2137, 0.2137, 0.2137, 0.2726, 0.2733, 0.2927, 0.3396, 0.363, 0.4601, 0.4501, 0.4501, 0.3068, 0.1534, 0.2833, 0.2605, 0.1621, 0.2585, 0.2726, 0.2585, 0.2291, 0.2639, 0.2987, 0.3965, 0.3429, 0.3356, 0.2478, 0.2706, 0.2659, 0.2063, 0.2063, 0.1833, 0.1222, 0.2729, 0.1974, 0.2497, 0.296, 0.3189, 0.4128, 0.3692, 0.3826, 0.3491, 0.386, 0.4162, 0.4598, 0.4497, 0.4397, 0.4699, 0.4598, 0.5068, 0.4732, 0.5068, 0.49, 0.5169, 0.4799, 0.4799, 0.5605, 0.5605, 0.7853, 0.8223, 0.8021, 0.9162, 0.8726, 0.9061, 0.9867, 0.8155, 0.886, 0.8491, 0.8525, 1.0102, 0.9632, 0.9632, 1.0068, 1.0605, 1.0102, 1.1276, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2013091516840116729", "ticker_idx": 2, "ts": "2026-01-19T03:29:53+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and Intel are expected to benefit from the '6 major customers' including Apple and Nvidia", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Deutsche Bank view: TSMC capacity crunch forces top customers to Samsung/Intel alternatives, reducing TSMC's advanced-node share from 95% to 90%.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Deutsche Bank: As TSMC's capacity fills up, Samsung and Intel are expected to benefit from the \"6 major customers\" including Apple and Nvidia\n\nDeutsche Bank stated in a recently released research report that global semiconductor foundry giant TSMC is facing a \"happy dilemma.\" Its 3nm process production capacity is extremely tight, with orders fully booked not only through 2026 but extending into 2027, leaving TSMC no choice but to significantly increase its capital expenditure plans.\n\nThis capacity bottleneck is reshaping the market landscape, prompting TSMC's top-tier customers—who have traditionally relied heavily on the company—to seriously consider Samsung and Intel as alternatives, Deutsche Bank noted.\n\nAnalysts including Robert Sanders pointed out in the report that TSMC has guided 2026 capital expenditures at $52–56 billion. This substantially exceeds Deutsche Bank's estimate of $50 billion and the market consensus of $46 billion.\n\nThe analysts assessed that this guidance effectively acknowledges that TSMC's 2025 capital expenditure plan was \"too conservative\" relative to the explosive AI demand. The current situation goes beyond mere excess demand for CoWoS packaging capacity; it reflects a severe supply shortage in core wafer fabrication capacity—particularly the 3nm process.\n\nThis supply-demand imbalance is creating a direct market overflow effect. Despite Samsung and Intel having a \"chequered record\" in foundry performance, Deutsche Bank emphasized that the \"6 major customers\"—Apple, Nvidia, AMD, Broadcom, Qualcomm, and MediaTek—have no choice but to seek alternative production capacity. As a result, TSMC's share of the advanced-node foundry market is expected to decline from 95% to 90%.\n\nDeutsche Bank noted that, given TSMC's full capacity booking for 2026 and order backlog extending into 2027, the sharp increase in spending is not surprising.\n\nPreviously, the initial bottleneck from the AI demand surge was concentrated in CoWoS packaging, causing some orders to overflow to ASE and Amkor. However, the situation has now become more severe, with demand significantly outstripping supply—especially for the 3nm process.\n\nAccording to Deutsche Bank estimates, TSMC plans to expand 3nm monthly production capacity to 190,000 wafers (190k wpm) by the end of 2026, but this will still fall short of customer demand.\n\nDeutsche Bank remarked, \"It is common for customers to want more capacity under TSMC's high utilization rates, but this level is unprecedented.\"\n\nDue to the extreme capacity tightness, TSMC is adopting a more aggressive strategy. Deutsche Bank reported that TSMC is delaying new 3nm development projects and encouraging customers to shift more 2027/2028 mass-production products to the 2nm GAA process. At the same time, TSMC is using pricing strategies to \"facilitate\" this transition, which is expected to be reflected in strong guidance for Q1 2026.\n\nIn this capacity competition, Deutsche Bank analyzed that Samsung's Taylor fab (SF2P) is more likely to be preferred over Intel.\n\nThe report stated, \"For customers seeking alternative supply, Samsung's Taylor fab is more likely to become the first destination.\"\n\nSpecifically, Qualcomm and AMD are the most likely to consider Samsung, while previous discussions indicate that Apple and Broadcom are evaluating Intel. However, the analyst added that although Intel has potential with its 14A process, \"there is still much work to be done.\"\n\n$TSM $INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.034071599659536655, "ret_signed_p1d": 0.034071599659536655, "alpha_spy_p1d": 0.0544283853830817, "alpha_c_p1d": 0.05904724748826806, "bench_spy_p1d": -0.020356785723545046, "bench_c_p1d": -0.024975647828731407, "ret_p1w": -0.09518734014379204, "ret_signed_p1w": -0.09518734014379204, "alpha_spy_p1w": -0.09673433127126752, "alpha_c_p1w": -0.0912661451419704, "bench_spy_p1w": 0.0015469911274754722, "bench_c_p1w": -0.003921195001821642, "ret_p1m": -0.016609855078868252, "ret_signed_p1m": -0.016609855078868252, "alpha_spy_p1m": -0.003872374509167109, "alpha_c_p1m": -0.03439250413778461, "bench_spy_p1m": -0.012737480569701143, "bench_c_p1m": 0.017782649058916356, "ret_p3m": 0.4586882737535989, "ret_signed_p3m": 0.4586882737535989, "alpha_spy_p3m": 0.4414598026397183, "alpha_c_p3m": 0.3227958409907441, "bench_spy_p3m": 0.017228471113880595, "bench_c_p3m": 0.1358924327628548, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2138, -0.1874, -0.1848, -0.158, -0.1156, -0.1197, -0.1448, -0.1484, -0.1589, -0.2115, -0.1827, -0.207, -0.188, -0.1812, -0.1934, -0.1931, -0.2353, -0.2436, -0.2609, -0.269, -0.2523, -0.2841, -0.2653, -0.2379, -0.237, -0.2161, -0.2161, -0.1363, -0.148, -0.0743, -0.0681, -0.1376, -0.1182, -0.1418, -0.1376, -0.1316, -0.1586, -0.1948, -0.2012, -0.2055, -0.2323, -0.2274, -0.2159, -0.2255, -0.2259, -0.23, -0.23, -0.2291, -0.2189, -0.2057, -0.2142, -0.2142, -0.1614, -0.1616, -0.1474, -0.0922, -0.1246, -0.03, -0.0618, 0.007, 0.0375, 0.029, 0.0, 0.0, 0.0341, 0.1552, 0.1567, -0.0402, -0.0952, -0.0645, 0.0388, 0.0362, -0.0104, 0.0394, 0.0488, 0.0349, 0.0273, 0.0773, 0.0698, 0.0036, 0.0283, -0.0102, -0.0036, -0.0036, -0.0166, -0.0319, -0.0498, -0.0607, -0.0709, -0.0179, -0.0017, -0.0319, -0.0287, -0.0311, -0.0822, -0.0294, -0.0215, -0.0754, -0.0294, -0.0038, 0.0217, -0.0364, -0.0253, -0.0256, -0.0618, -0.0411, -0.0166, -0.0658, -0.0628, -0.0618, 0.0047, -0.0609, -0.0816, -0.1229, -0.0603, 0.0228, 0.0728, 0.0728, 0.0813, 0.1267, 0.2553, 0.3143, 0.3284, 0.388, 0.3588, 0.3829, 0.4587, 0.4587, 0.3991, 0.411, 0.3899, 0.4221, 0.7577, 0.8098, 0.7998, 1.0177, 1.0119, 1.1214, 1.0396, 1.303, 1.4065, 1.3343, 1.6601, 1.7564, 1.5684, 1.5615, 1.4687, 1.3162, 1.3034, 1.3595, 1.5332, 1.5234, 1.552, 1.552, 1.6303, 1.5931, 1.5743, 1.4421, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2031167396875219382", "ticker_idx": 0, "ts": "2026-03-10T00:36:59+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy rating; raise target price to $571. We believe Micron's valuation still lags the pace of earnings upgrades. We expect the company's near-term operating margin to rise into the 80% range.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares GF Securities Buy on MU with PT raised to $571; estimates FY2Q26 revenue $23B / GM 77% well above consensus, memory upcycle intact into 2027.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics & Communications》\n\n☄️ Micron FY2Q26 Earnings Preview: Memory Cycle Moving Higher\n\n☀️ Strong earnings expected, easing investor concerns:\n\nWe expect DRAM pricing to remain firm over the next several quarters. We forecast DRAM contract prices to rise 100% in 1Q26 and at least 30% QoQ in 2Q26. Based on current pricing trends, there is still further upside ahead. For Micron, we estimate FY2Q26 revenue of $23 billion and gross margin of 77% (vs. consensus of $19 billion revenue and 69% GM). Looking ahead, the company’s FY3Q26 revenue guidance could reach $29 billion, with gross margin further improving to 83% (vs. consensus of $26–27 billion revenue and 76–77% GM). On HBM, Micron has already secured HBM4 orders, with small-scale volume shipments set to begin in March.\n\n☀️ The memory upcycle remains intact:\n\n1. Inventory restocking continues: CSPs are targeting inventory days equivalent to 4–6 months, versus the current level of only 2 months. Smartphone OEM inventory stands at around 7 weeks, up from 4–5 weeks in 4Q25, but still below the historical average of 7–9 weeks.\n\n2. Supply shortages are likely to persist into 2027: Although Micron’s Idaho fab, SK Hynix’s M15X, and Samsung’s P4L are all accelerating capacity expansion, meaningful wafer output is expected to be concentrated in 2H27. At the same time, AI-related demand is expected to account for more than 75% of total DRAM demand in 2027, meaning supply shortages could persist through 2H27. As a result, earnings visibility for memory makers has improved significantly, and negotiations for 3–5 year LTAs are already underway.\n\n3. Memory BOM impact remains manageable: As noted in our previous report on memory cost impact, we expect memory’s share of AI server BOM to rise from 8–9% in 2025 to 15% by 4Q26, with most of the increase concentrated in 2026.\n\n☀️ Maintain Buy rating; raise target price to $571:\n\nWe believe Micron’s valuation still lags the pace of earnings upgrades. We expect the company’s near-term operating margin to rise into the 80% range. Even under a historical downcycle scenario where DRAM prices fall by about 60%, Micron could still sustain an operating profit margin of around 50%. Historically, Micron has traded at an average 1.4x P/B, corresponding to an average 15% OPM. Given our forecast for 50%+ OPM, we believe the current 4x P/B remains reasonable even within a downcycle valuation framework. We have revised our BVPS base to the average of FY26/27E, and raise our target price to $571.\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.034208953175586476, "ret_signed_m1d": -0.034208953175586476, "alpha_spy_m1d": -0.0358186670955043, "alpha_c_m1d": -0.026759247502441297, "bench_spy_m1d": 0.0016097139199178212, "bench_c_m1d": -0.007449705673145179, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.038649573399176074, "ret_signed_p1d": 0.038649573399176074, "alpha_spy_p1d": 0.0399047363276932, "alpha_c_p1d": 0.029337383702770392, "bench_spy_p1d": -0.0012551629285171284, "bench_c_p1d": 0.009312189696405682, "ret_p1w": 0.1453201905152708, "ret_signed_p1w": 0.1453201905152708, "alpha_spy_p1w": 0.1547564164803874, "alpha_c_p1w": 0.1464527043064776, "bench_spy_p1w": -0.009436225965116618, "bench_c_p1w": -0.001132513791206824, "ret_p1m": 0.009404129688952878, "ret_signed_p1m": 0.009404129688952878, "alpha_spy_p1m": 0.00840554852556119, "alpha_c_p1m": -0.055000842803135885, "bench_spy_p1m": 0.0009985811633916875, "bench_c_p1m": 0.06440497249208876, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3591, -0.402, -0.4111, -0.4234, -0.4408, -0.3837, -0.3406, -0.3141, -0.3149, -0.2891, -0.2891, -0.2938, -0.2698, -0.2741, -0.292, -0.292, -0.2175, -0.2256, -0.148, -0.1577, -0.1888, -0.1439, -0.142, -0.1612, -0.1731, -0.1649, -0.1001, -0.1001, -0.0945, -0.0347, -0.0137, -0.0086, -0.0348, 0.0177, 0.0798, 0.0811, 0.0292, 0.0861, 0.0405, -0.0588, -0.0502, -0.0209, -0.0486, -0.0741, 0.0179, 0.0269, 0.0212, 0.0212, -0.0083, 0.0443, 0.0353, 0.0622, 0.0443, 0.037, 0.0642, 0.0309, 0.023, 0.0237, -0.0581, -0.0058, -0.015, -0.0814, -0.0342, 0.0, 0.0386, 0.0056, 0.0571, 0.096, 0.1453, 0.1454, 0.1021, 0.0491, 0.0031, -0.0188, -0.0521, -0.1182, -0.1138, -0.2014, -0.1616, -0.0871, -0.0911, -0.0911, -0.0625, -0.0629, 0.0094, 0.0461, 0.0438, 0.0586, 0.1557, 0.1323, 0.1347, 0.1294, 0.1129, 0.1153, 0.2098, 0.1955, 0.2327, 0.3018, 0.2515, 0.2867, 0.2835, 0.3456, 0.4306, 0.5888, 0.6543, 0.6048, 0.8534, 0.9738, 0.9025, 0.9944, 0.9259, 0.7984, 0.6914, 0.7341, 0.8166, 0.8913, 0.8638, 0.8638, 1.2234, 1.3041, 1.292, 1.4098, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2007019452504871309", "ticker_idx": 0, "ts": "2026-01-02T09:21:40+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "씨티, 삼성전자 목표가 20만원 제시", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Citi's price target of 200,000 KRW on Samsung Electronics, implying significant upside.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "ko", "tweet_text": "Citi sets Samsung Electronics’ price target at 200,000 KRW.\n\n씨티, 삼성전자 목표가 20만원 제시 https://t.co/2KNpDnUjwD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06692605458283052, "ret_signed_m1d": -0.06692605458283052, "alpha_spy_m1d": -0.06509636873623525, "alpha_c_m1d": -0.06463915473919735, "bench_spy_m1d": -0.0018296858465952637, "bench_c_m1d": -0.0022868998436331722, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07470817945578978, "ret_signed_p1d": 0.07470817945578978, "alpha_spy_p1d": 0.06804806563972665, "alpha_c_p1d": 0.07249062773241488, "bench_spy_p1d": 0.006660113816063129, "bench_c_p1d": 0.0022175517233749, "ret_p1w": 0.08171206747205106, "ret_signed_p1w": 0.08171206747205106, "alpha_spy_p1w": 0.06575705280338862, "alpha_c_p1w": 0.06889155872365338, "bench_spy_p1w": 0.01595501466866245, "bench_c_p1w": 0.012820508748397685, "ret_p1m": 0.1704280229603632, "ret_signed_p1m": 0.1704280229603632, "alpha_spy_p1m": 0.15251154277998902, "alpha_c_p1m": 0.16377526191524572, "bench_spy_p1m": 0.017916480180374172, "bench_c_p1m": 0.006652761045117472, "ret_p3m": 0.4789302758506657, "ret_signed_p3m": 0.4789302758506657, "alpha_spy_p3m": 0.5171938588313623, "alpha_c_p3m": 0.5428405506058007, "bench_spy_p3m": -0.03826358298069654, "bench_c_p3m": -0.063910274755135, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3049, -0.3049, -0.3049, -0.2689, -0.2774, -0.2906, -0.2643, -0.2434, -0.2418, -0.2403, -0.2449, -0.2364, -0.2527, -0.2348, -0.2101, -0.2294, -0.2217, -0.1938, -0.1675, -0.1396, -0.1876, -0.2209, -0.2318, -0.2418, -0.2209, -0.1984, -0.2015, -0.2039, -0.2472, -0.2209, -0.2426, -0.2527, -0.2209, -0.2658, -0.2511, -0.231, -0.2039, -0.1984, -0.2217, -0.2194, -0.1992, -0.1907, -0.1861, -0.1605, -0.152, -0.1605, -0.1636, -0.169, -0.1566, -0.1884, -0.2039, -0.1644, -0.1667, -0.1768, -0.1442, -0.1365, -0.1396, -0.1396, -0.0939, -0.07, -0.0669, -0.0669, -0.0669, 0.0, 0.0747, 0.0809, 0.0973, 0.0802, 0.0817, 0.0802, 0.0708, 0.0918, 0.1198, 0.1588, 0.1619, 0.13, 0.1634, 0.1852, 0.1837, 0.1837, 0.2412, 0.2638, 0.2506, 0.249, 0.1704, 0.3035, 0.316, 0.2397, 0.2342, 0.2949, 0.2903, 0.3058, 0.3899, 0.4101, 0.4101, 0.4101, 0.4101, 0.4786, 0.4794, 0.5019, 0.5564, 0.5837, 0.6965, 0.6848, 0.6848, 0.5183, 0.3401, 0.4911, 0.4646, 0.3502, 0.4623, 0.4786, 0.4623, 0.428, 0.4685, 0.5089, 0.6226, 0.5603, 0.5518, 0.4498, 0.4763, 0.4708, 0.4016, 0.4016, 0.3748, 0.3039, 0.4789, 0.3912, 0.452, 0.5058, 0.5323, 0.6415, 0.5908, 0.6064, 0.5674, 0.6103, 0.6454, 0.6961, 0.6844, 0.6727, 0.7078, 0.6961, 0.7507, 0.7117, 0.7507, 0.7312, 0.7624, 0.7195, 0.7195, 0.8131, 0.8131, 1.0743, 1.1172, 1.0938, 1.2264, 1.1757, 1.2147, 1.3083, 1.1094, 1.1913, 1.1484, 1.1523, 1.3356, 1.281, 1.281, 1.3317, 1.3941, 1.3356, 1.472, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2013803830102872099", "ticker_idx": 0, "ts": "2026-01-21T02:40:22+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Memory companies are literally minting gold.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SK Hynix on extraordinary profitability: server DDR5 OPM potentially 90% in H1, memory companies minting gold.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Comment from Korea's BNK Investment & Securities:\n\n\"SK Hynix's server DDR5 OPM was 70%, but considering the recent price increases, the server DDR5 OPM could reach 90% within the first half of this year.\"\n\nNo way lol OPM 90%? Does that make sense?\n\nMemory companies are literally minting gold.\n\n---\n\nWhy make HBM? LOL—DDR5 is way more profitable.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0040540047631150156, "ret_signed_m1d": 0.0040540047631150156, "alpha_spy_m1d": 0.015463407148629704, "alpha_c_m1d": 0.0327654177893445, "bench_spy_m1d": -0.011409402385514689, "bench_c_m1d": -0.028711413026229482, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020270277703938522, "ret_signed_p1d": 0.020270277703938522, "alpha_spy_p1d": 0.015047047684210835, "alpha_c_p1d": 0.018055925292683073, "bench_spy_p1d": 0.005223230019727687, "bench_c_p1d": 0.002214352411255449, "ret_p1w": 0.13648640677468937, "ret_signed_p1w": 0.13648640677468937, "alpha_spy_p1w": 0.1218672560980536, "alpha_c_p1w": 0.09991284303248538, "bench_spy_p1w": 0.014619150676635773, "bench_c_p1w": 0.036573563742203996, "ret_p1m": 0.20810806030390272, "ret_signed_p1m": 0.20810806030390272, "alpha_spy_p1m": 0.2094504112208455, "alpha_c_p1m": 0.18753228919462628, "bench_spy_p1m": -0.0013423509169427916, "bench_c_p1m": 0.02057577110927644, "ret_p3m": 0.5785830608899616, "ret_signed_p3m": 0.5785830608899616, "alpha_spy_p3m": 0.5417352469449956, "alpha_c_p3m": 0.4242527050826166, "bench_spy_p3m": 0.036847813944965946, "bench_c_p3m": 0.154330355807345, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3113, -0.2775, -0.2964, -0.2465, -0.233, -0.2451, -0.1628, -0.2087, -0.2181, -0.1992, -0.2168, -0.1817, -0.1641, -0.1668, -0.1736, -0.2438, -0.1817, -0.2303, -0.2411, -0.2289, -0.2964, -0.2978, -0.2992, -0.2924, -0.2649, -0.2838, -0.273, -0.2459, -0.2541, -0.2676, -0.2649, -0.2203, -0.2351, -0.2068, -0.2365, -0.2284, -0.2514, -0.2838, -0.2554, -0.2541, -0.2608, -0.2162, -0.2108, -0.2054, -0.2054, -0.1905, -0.1351, -0.1203, -0.1203, -0.1203, -0.0851, -0.0595, -0.0189, 0.0027, 0.0216, 0.0054, 0.0122, -0.0027, 0.0027, 0.0122, 0.0216, 0.0324, 0.0041, 0.0, 0.0203, 0.0365, -0.0054, 0.0811, 0.1365, 0.1635, 0.2284, 0.1216, 0.2257, 0.2162, 0.1378, 0.1338, 0.1986, 0.1838, 0.1622, 0.2, 0.1892, 0.1892, 0.1892, 0.1892, 0.2081, 0.2824, 0.2851, 0.3581, 0.3757, 0.4879, 0.4364, 0.4364, 0.2713, 0.1494, 0.274, 0.251, 0.1318, 0.2699, 0.2929, 0.2591, 0.232, 0.3186, 0.3132, 0.4297, 0.3714, 0.3633, 0.2631, 0.3349, 0.3471, 0.2631, 0.2631, 0.1819, 0.0926, 0.2144, 0.1237, 0.186, 0.1995, 0.2401, 0.3985, 0.3511, 0.3904, 0.408, 0.4933, 0.538, 0.5637, 0.5271, 0.5786, 0.6571, 0.6558, 0.6585, 0.6544, 0.7492, 0.76, 0.7505, 0.741, 0.741, 0.959, 0.959, 1.1675, 1.2393, 1.2826, 1.5452, 1.4843, 1.6752, 1.6671, 1.4626, 1.4911, 1.3625, 1.3625, 1.6265, 1.6278, 1.6278, 1.7781, 2.0367, 2.0995, 2.159, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2017490821441065276", "ticker_idx": 0, "ts": "2026-01-31T06:51:09+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bernstein raises SanDisk price target to $1,000", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Bernstein raises SanDisk PT to $1,000; author shares/adopts the bullish call.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein raises SanDisk price target to $1,000… https://t.co/Z2M7yQZbuz", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.13377125780280041, "ret_signed_m1d": -0.13377125780280041, "alpha_spy_m1d": -0.1288244878566457, "alpha_c_m1d": -0.12427119485539129, "bench_spy_m1d": -0.004946769946154728, "bench_c_m1d": -0.009500062947409127, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04550240511035031, "ret_signed_p1d": 0.04550240511035031, "alpha_spy_p1d": 0.0539578015410509, "alpha_c_p1d": 0.06739411516269922, "bench_spy_p1d": -0.00845539643070059, "bench_c_p1d": -0.02189171005234891, "ret_p1w": -0.12302322022384571, "ret_signed_p1w": -0.12302322022384571, "alpha_spy_p1w": -0.12092378247360513, "alpha_c_p1w": -0.10987452140516396, "bench_spy_p1w": -0.0020994377502405737, "bench_c_p1w": -0.013148698818681748, "ret_p1m": -0.1500661694355927, "ret_signed_p1m": -0.1500661694355927, "alpha_spy_p1m": -0.12838119088409294, "alpha_c_p1m": -0.09664480158260835, "bench_spy_p1m": -0.021684978551499756, "bench_c_p1m": -0.05342136785298435, "ret_p3m": 0.6482923844961672, "ret_signed_p3m": 0.6482923844961672, "alpha_spy_p3m": 0.6120365774891936, "alpha_c_p3m": 0.5488959539140126, "bench_spy_p3m": 0.03625580700697362, "bench_c_p3m": 0.09939643058215464, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6746, -0.6878, -0.64, -0.5972, -0.5918, -0.5744, -0.6339, -0.6179, -0.6003, -0.6318, -0.6303, -0.7054, -0.699, -0.6588, -0.6685, -0.6767, -0.6767, -0.6644, -0.6841, -0.6913, -0.7078, -0.6793, -0.6566, -0.6611, -0.6701, -0.65, -0.6368, -0.6901, -0.6965, -0.6854, -0.6891, -0.6701, -0.6428, -0.6376, -0.6319, -0.6241, -0.6241, -0.6241, -0.6328, -0.6389, -0.6432, -0.6432, -0.5863, -0.588, -0.4744, -0.4685, -0.4971, -0.4327, -0.4148, -0.414, -0.417, -0.3848, -0.3782, -0.3782, -0.3189, -0.2465, -0.2432, -0.2877, -0.2923, -0.2763, -0.2069, -0.1893, -0.1338, 0.0, 0.0455, -0.1213, -0.1338, -0.1012, -0.123, -0.1858, -0.0991, -0.0525, -0.0581, -0.0581, -0.1122, -0.0975, -0.0664, -0.023, 0.0019, -0.0402, -0.0494, -0.0201, -0.0449, -0.0694, -0.1501, -0.0995, -0.1498, -0.2073, -0.115, -0.0697, -0.0147, -0.0698, -0.0054, 0.0577, 0.0826, 0.133, 0.1606, 0.0668, 0.056, 0.056, 0.019, -0.0933, -0.0743, -0.1394, -0.0449, 0.0413, 0.0546, 0.0546, 0.0893, 0.0685, 0.1739, 0.2801, 0.2804, 0.4318, 0.4197, 0.3404, 0.3822, 0.3844, 0.3725, 0.3581, 0.4718, 0.4016, 0.488, 0.6087, 0.5067, 0.5997, 0.6483, 0.7843, 0.8878, 1.114, 1.1195, 1.0143, 1.3485, 1.3263, 1.1827, 1.1755, 1.0785, 1.1159, 1.0038, 1.0794, 1.0933, 1.3183, 1.2228, 1.2228, 1.3894, 1.39, 1.4677, 1.5479, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2019626726058520801", "ticker_idx": 0, "ts": "2026-02-06T04:18:28+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If Samsung Electronics' chip quality meets expectations, there is a high possibility this could expand into a $100 billion business", "tweet_type": "news_reaction", "timeline": "years", "summary": "Sharing bullish article on Samsung foundry turnaround via Tesla AI5/AI6 contracts; potential $100B business upside cited, with Tesla's deep involvement signaling durable partnership.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Begins Tesla AI5 Wafer Processing… \"Musk Also Attending Meetings\"\n\nSamsung Electronics will soon begin pilot production of Tesla's AI5 chips. While the AI5 and AI6 chip foundry contracts are loss-making in the short term, the level of interest is high enough that Tesla CEO Elon Musk is personally involved in working-level meetings with Samsung Electronics staff, raising expectations for revenue growth through expanded cooperation.\n\nSamsung Electronics was previously reported to have secured a $16.5 billion (approximately 24 trillion won) contract from Tesla in July last year for foundry production of AI6 chips. The contract runs through 2033. Additionally, Samsung Electronics has also secured a portion of AI5 chip production, which had been expected to be produced entirely by TSMC. The opportunity arose for Samsung due to TSMC's limited production capacity and lower price competitiveness. Samsung Electronics will manufacture the AI5 and AI6 chips using its cutting-edge 2-nanometer process.\n\nAccording to industry sources, Samsung Electronics will begin wafer processing for pilot production of Tesla AI5 chips as early as this month. Samsung's Taylor foundry fab in the United States had been expected to begin operations in the second half of this year, but the timeline has been moved up as Tesla recently completed the AI5 chip design. Considering that it takes three to four months from the first wafer input to produce finished products, initial AI5 chip samples are expected to be available as early as the first half of this year. Full-scale mass production is expected in 2027.\n\nTesla is investing significant effort in the AI chip development and production preparation process with Samsung Electronics. In particular, Tesla CEO Elon Musk is reportedly leading the project by personally overseeing key progress updates.\n\nOne semiconductor industry insider said, \"Tesla CEO Elon Musk is paying close attention to the collaboration with Samsung Electronics. He personally participates in working-level technical meetings of Samsung Electronics' foundry division related to AI5 and AI6 chip work,\" adding, \"I've heard he actively checks on progress point by point and offers his input during meetings.\"\n\nSamsung Electronics had previously allowed Tesla to be involved in the manufacturing process, considering the risks that could arise during the initial mass production phase. CEO Elon Musk also expressed his willingness to be deeply involved in the project, emphasizing a close cooperative relationship with Samsung Electronics. This has gone beyond public statements to actual participation at the working level.\n\nFor the next-generation AI6 chip, which Samsung Electronics is exclusively responsible for, pilot production is expected to begin next year with mass production starting in 2028. During its recent fourth-quarter earnings conference call, the company expressed confidence, stating, \"Our second-generation 2-nanometer process is on track for mass production in the second half of the year, meeting yield and performance targets,\" and \"We are conducting power, performance, and area (PPA) evaluations and test chip collaborations with major clients for product design, with technology verification proceeding as planned.\"\n\nThe industry is paying attention to the mid-to-long-term value of Samsung Electronics' Tesla contract. While this deal is essentially a loss-making contract in the short term, it holds significant strategic importance considering future scalability. The contract volume for AI5 chips has not been disclosed but is known to be small. The AI6 chip contract is valued at $16.5 billion. Considering that Samsung Electronics is spending approximately $37 billion (about 54.2 trillion won) to build the Taylor foundry fab through 2030, annual revenue of around $3 billion is not a large proportion relative to the investment.\n\nHowever, the potential expansion of applications for Tesla's AI5 and AI6 chips is a point where future revenue growth is expected. The chips could be deployed across Tesla's entire AI value chain, including not only autonomous vehicles but also robotaxis, humanoid robots, and proprietary AI servers. Another semiconductor industry insider said, \"The specific applications for the AI chips are expected to become clearer in the first half of this year,\" adding, \"If Samsung Electronics' chip quality meets expectations, there is a high possibility this could expand into a $100 billion business.\"\n\nThe industry is also raising the possibility that Samsung Electronics' foundry division could achieve a turnaround to profitability next year. The division has continued to post annual losses of several trillion won until recently. Potential turning points for improved performance include the adoption of the next-generation mobile application processor \"Exynos 2600\" in the Galaxy S26 series, potential orders from big tech companies like Qualcomm and AMD, and increased utilization rates for 4-nanometer and 8-nanometer processes.\nKiwoom Securities noted in a recent report, \"Samsung Electronics' foundry and System LSI divisions are expected to post annual operating losses of around 4 trillion won (this year), but there is also a possibility that the deficit could narrow further, driven by increased utilization of advanced process nodes.\" Heungkuk Securities projected, \"As TSMC capacity shortages intensify, the visibility of Samsung Foundry's turnaround is increasing, with the Tesla contract as a catalyst.\"\n\n---\n\nhttps://t.co/ZFzsb8YLPR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004413573354641542, "ret_signed_m1d": 0.004413573354641542, "alpha_spy_m1d": 0.02323716010791821, "alpha_c_m1d": 0.043384692507755784, "bench_spy_m1d": -0.018823586753276667, "bench_c_m1d": -0.03897111915311424, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04918033919762932, "ret_signed_p1d": 0.04918033919762932, "alpha_spy_p1d": 0.04435850709461886, "alpha_c_p1d": 0.033450079963909696, "bench_spy_p1d": 0.004821832103010459, "bench_c_p1d": 0.015730259233719623, "ret_p1w": 0.1424968068304704, "ret_signed_p1w": 0.1424968068304704, "alpha_spy_p1w": 0.15534031506757862, "alpha_c_p1w": 0.15362132929894434, "bench_spy_p1w": -0.012843508237108225, "bench_c_p1w": -0.011124522468473952, "ret_p1m": 0.09394700631841824, "ret_signed_p1m": 0.09394700631841824, "alpha_spy_p1m": 0.11182937385561598, "alpha_c_p1m": 0.10365438734013277, "bench_spy_p1m": -0.01788236753719774, "bench_c_p1m": -0.009707381021714534, "ret_p3m": 0.6806467172795636, "ret_signed_p3m": 0.6806467172795636, "alpha_spy_p3m": 0.6151777914268302, "alpha_c_p3m": 0.47437301661456477, "bench_spy_p3m": 0.06546892585273345, "bench_c_p3m": 0.20627370066499884, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3506, -0.3531, -0.355, -0.3901, -0.3688, -0.3863, -0.3945, -0.3688, -0.4052, -0.3932, -0.3769, -0.355, -0.3506, -0.3694, -0.3675, -0.3512, -0.3443, -0.3405, -0.3198, -0.3129, -0.3198, -0.3223, -0.3267, -0.3167, -0.3424, -0.355, -0.323, -0.3248, -0.333, -0.3066, -0.3004, -0.3029, -0.3029, -0.2659, -0.2465, -0.244, -0.244, -0.244, -0.1898, -0.1293, -0.1242, -0.111, -0.1248, -0.1236, -0.1248, -0.1324, -0.1154, -0.0927, -0.0612, -0.0586, -0.0845, -0.0574, -0.0397, -0.041, -0.041, 0.0057, 0.024, 0.0132, 0.012, -0.0517, 0.0561, 0.0662, 0.0044, 0.0, 0.0492, 0.0454, 0.058, 0.1261, 0.1425, 0.1425, 0.1425, 0.1425, 0.198, 0.1986, 0.2169, 0.261, 0.2831, 0.3745, 0.3651, 0.3651, 0.2301, 0.0858, 0.2081, 0.1866, 0.0939, 0.1847, 0.198, 0.1847, 0.157, 0.1898, 0.2226, 0.3146, 0.2642, 0.2573, 0.1747, 0.1961, 0.1917, 0.1356, 0.1356, 0.1139, 0.0564, 0.1983, 0.1272, 0.1765, 0.22, 0.2415, 0.33, 0.2889, 0.3016, 0.27, 0.3047, 0.3331, 0.3742, 0.3647, 0.3553, 0.3837, 0.3742, 0.4184, 0.3868, 0.4184, 0.4026, 0.4279, 0.3932, 0.3932, 0.469, 0.469, 0.6806, 0.7154, 0.6964, 0.8039, 0.7628, 0.7944, 0.8702, 0.7091, 0.7754, 0.7407, 0.7438, 0.8923, 0.8481, 0.8481, 0.8891, 0.9397, 0.8923, 1.0029, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2019626726058520801", "ticker_idx": 1, "ts": "2026-02-06T04:18:28+00:00", "symbol": "TSLA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Tesla CEO Elon Musk is personally involved in working-level meetings with Samsung Electronics staff, raising expectations for revenue growth through expanded cooperation", "tweet_type": "news_reaction", "timeline": "years", "summary": "Sharing bullish article on Samsung foundry turnaround via Tesla AI5/AI6 contracts; potential $100B business upside cited, with Tesla's deep involvement signaling durable partnership.", "resolved_tickers": ["TSLA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Auto Manufacturers'", "bench_c_industry": "Auto Manufacturers", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Begins Tesla AI5 Wafer Processing… \"Musk Also Attending Meetings\"\n\nSamsung Electronics will soon begin pilot production of Tesla's AI5 chips. While the AI5 and AI6 chip foundry contracts are loss-making in the short term, the level of interest is high enough that Tesla CEO Elon Musk is personally involved in working-level meetings with Samsung Electronics staff, raising expectations for revenue growth through expanded cooperation.\n\nSamsung Electronics was previously reported to have secured a $16.5 billion (approximately 24 trillion won) contract from Tesla in July last year for foundry production of AI6 chips. The contract runs through 2033. Additionally, Samsung Electronics has also secured a portion of AI5 chip production, which had been expected to be produced entirely by TSMC. The opportunity arose for Samsung due to TSMC's limited production capacity and lower price competitiveness. Samsung Electronics will manufacture the AI5 and AI6 chips using its cutting-edge 2-nanometer process.\n\nAccording to industry sources, Samsung Electronics will begin wafer processing for pilot production of Tesla AI5 chips as early as this month. Samsung's Taylor foundry fab in the United States had been expected to begin operations in the second half of this year, but the timeline has been moved up as Tesla recently completed the AI5 chip design. Considering that it takes three to four months from the first wafer input to produce finished products, initial AI5 chip samples are expected to be available as early as the first half of this year. Full-scale mass production is expected in 2027.\n\nTesla is investing significant effort in the AI chip development and production preparation process with Samsung Electronics. In particular, Tesla CEO Elon Musk is reportedly leading the project by personally overseeing key progress updates.\n\nOne semiconductor industry insider said, \"Tesla CEO Elon Musk is paying close attention to the collaboration with Samsung Electronics. He personally participates in working-level technical meetings of Samsung Electronics' foundry division related to AI5 and AI6 chip work,\" adding, \"I've heard he actively checks on progress point by point and offers his input during meetings.\"\n\nSamsung Electronics had previously allowed Tesla to be involved in the manufacturing process, considering the risks that could arise during the initial mass production phase. CEO Elon Musk also expressed his willingness to be deeply involved in the project, emphasizing a close cooperative relationship with Samsung Electronics. This has gone beyond public statements to actual participation at the working level.\n\nFor the next-generation AI6 chip, which Samsung Electronics is exclusively responsible for, pilot production is expected to begin next year with mass production starting in 2028. During its recent fourth-quarter earnings conference call, the company expressed confidence, stating, \"Our second-generation 2-nanometer process is on track for mass production in the second half of the year, meeting yield and performance targets,\" and \"We are conducting power, performance, and area (PPA) evaluations and test chip collaborations with major clients for product design, with technology verification proceeding as planned.\"\n\nThe industry is paying attention to the mid-to-long-term value of Samsung Electronics' Tesla contract. While this deal is essentially a loss-making contract in the short term, it holds significant strategic importance considering future scalability. The contract volume for AI5 chips has not been disclosed but is known to be small. The AI6 chip contract is valued at $16.5 billion. Considering that Samsung Electronics is spending approximately $37 billion (about 54.2 trillion won) to build the Taylor foundry fab through 2030, annual revenue of around $3 billion is not a large proportion relative to the investment.\n\nHowever, the potential expansion of applications for Tesla's AI5 and AI6 chips is a point where future revenue growth is expected. The chips could be deployed across Tesla's entire AI value chain, including not only autonomous vehicles but also robotaxis, humanoid robots, and proprietary AI servers. Another semiconductor industry insider said, \"The specific applications for the AI chips are expected to become clearer in the first half of this year,\" adding, \"If Samsung Electronics' chip quality meets expectations, there is a high possibility this could expand into a $100 billion business.\"\n\nThe industry is also raising the possibility that Samsung Electronics' foundry division could achieve a turnaround to profitability next year. The division has continued to post annual losses of several trillion won until recently. Potential turning points for improved performance include the adoption of the next-generation mobile application processor \"Exynos 2600\" in the Galaxy S26 series, potential orders from big tech companies like Qualcomm and AMD, and increased utilization rates for 4-nanometer and 8-nanometer processes.\nKiwoom Securities noted in a recent report, \"Samsung Electronics' foundry and System LSI divisions are expected to post annual operating losses of around 4 trillion won (this year), but there is also a possibility that the deficit could narrow further, driven by increased utilization of advanced process nodes.\" Heungkuk Securities projected, \"As TSMC capacity shortages intensify, the visibility of Samsung Foundry's turnaround is increasing, with the Tesla contract as a catalyst.\"\n\n---\n\nhttps://t.co/ZFzsb8YLPR", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03381088854992842, "ret_signed_m1d": -0.03381088854992842, "alpha_spy_m1d": -0.014987301796651753, "alpha_c_m1d": -0.029742774792022253, "bench_spy_m1d": -0.018823586753276667, "bench_c_m1d": -0.004068113757906167, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015105500216312517, "ret_signed_p1d": 0.015105500216312517, "alpha_spy_p1d": 0.010283668113302058, "alpha_c_p1d": 0.018834561300218544, "bench_spy_p1d": 0.004821832103010459, "bench_c_p1d": -0.0037290610839060268, "ret_p1w": 0.015397381030360036, "ret_signed_p1w": 0.015397381030360036, "alpha_spy_p1w": 0.02824088926746826, "alpha_c_p1w": 0.03073761938445463, "bench_spy_p1w": -0.012843508237108225, "bench_c_p1w": -0.015340238354094593, "ret_p1m": -0.030235200113545524, "ret_signed_p1m": -0.030235200113545524, "alpha_spy_p1m": -0.012352832576347783, "alpha_c_p1m": -0.0014192215516782847, "bench_spy_p1m": -0.01788236753719774, "bench_c_p1m": -0.02881597856186724, "ret_p3m": -0.030113533619592348, "ret_signed_p3m": -0.030113533619592348, "alpha_spy_p3m": -0.0955824594723258, "alpha_c_p3m": -0.04808799031246491, "bench_spy_p3m": 0.06546892585273345, "bench_c_p3m": 0.017974456692872565, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0693, 0.0474, -0.0222, -0.0164, -0.0053, -0.024, -0.0173, -0.0386, -0.0487, 0.0162, 0.0202, 0.0376, 0.0376, 0.0464, 0.0463, 0.0441, 0.0867, 0.1056, 0.1068, 0.0693, 0.0828, 0.0981, 0.087, 0.1164, 0.1562, 0.1916, 0.1366, 0.1758, 0.1705, 0.1888, 0.1811, 0.1807, 0.1807, 0.1559, 0.118, 0.1054, 0.0939, 0.0939, 0.0656, 0.0987, 0.0531, 0.0494, 0.0601, 0.0825, 0.0921, 0.0878, 0.0683, 0.0668, 0.0642, 0.0642, 0.0198, 0.0495, 0.093, 0.0923, 0.0586, 0.0481, 0.0495, 0.0133, 0.0469, 0.026, 0.0264, -0.0124, -0.0338, 0.0, 0.0151, 0.0343, 0.0417, 0.0145, 0.0154, 0.0154, -0.0012, 0.0005, 0.0015, 0.0017, -0.0274, -0.0042, 0.0153, -0.0062, -0.0209, -0.0189, -0.0454, -0.0126, -0.0135, -0.035, -0.0302, -0.0289, -0.008, -0.0392, -0.0484, -0.0378, -0.0288, -0.0446, -0.0749, -0.105, -0.0736, -0.0683, -0.0612, -0.0949, -0.1199, -0.1358, -0.0957, -0.0726, -0.1229, -0.1229, -0.1418, -0.1568, -0.1651, -0.1593, -0.1512, -0.1428, -0.1141, -0.0466, -0.054, -0.0255, -0.0453, -0.0601, -0.0574, -0.0909, -0.0847, -0.0789, -0.0854, -0.0932, -0.0717, -0.0494, -0.0452, -0.0529, -0.0301, 0.0017, 0.0419, 0.0824, 0.0543, 0.0831, 0.0783, 0.0271, -0.0027, -0.017, 0.015, 0.0164, 0.0362, 0.0362, 0.0547, 0.0711, 0.0754, 0.06, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2024451374889783501", "ticker_idx": 1, "ts": "2026-02-19T11:49:54+00:00", "symbol": "3037.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron is seeing a continued increase in AI PCB revenue mix, driven by expanding share in ASIC projects for Amazon Web Services, Google, and Meta. 2025–2027 EPS is projected at NT$19.24, NT$37.90, and NT$67.13, respectively, with the price target raised from NT$805 to NT$925.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman/BofA Buy calls on Taiwan PCB/CCL names — Taiwan Union, Unimicron, Elite Material — with raised PTs on AI server ABF supercycle.", "resolved_tickers": ["3037.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Taiwan Stock Market 2026 Barometer: ABF 'Supercycle' Ignites… AI Infrastructure Driving PCB & CCL Demand Explosion\n\n- As AI server and high-performance computing (HPC) demand ramps up in earnest, the long-struggling PCB sector has entered a strong rebound phase. After enduring weak consumer electronics and inventory destocking pressure, the advancement and higher specifications of AI server architectures are driving across-the-board increases in substrate layer counts, material grades, and per-unit value, serving as key catalysts for both industry fundamentals and share prices. Related names have surged sharply year-to-date, drawing significant market attention.\n\n- Capital is flowing back into high-end materials and process supply chains, with PCB, ABF, and CCL simultaneously entering a new growth cycle. Most major players posted January revenues well above year-ago levels, hitting all-time highs. High-end PCB makers exposed to AI server demand led the earnings recovery, while ABF substrates are also seeing a rapidly tightening supply-demand structure as AI/HPC proliferates, prompting assessments of a 'supercycle' entry.\n\n- In a January report, Goldman Sachs identified PCB and CCL as key beneficiary sectors of next-generation AI infrastructure. The global AI server PCB market is projected to surge from $3.1B in 2024 to $27.1B in 2027, with 2026–2027 YoY growth of +113% and +117%, respectively. The upstream CCL market is expected to expand from $1.5B in 2024 to $18.7B in 2027, with even steeper 2026–2027 YoY growth of +142% and +222%. As AI servers rapidly evolve toward M9-grade materials and higher-layer designs, R&D and CAPEX barriers are rising, creating a competitive environment that favors leading players with the requisite technology and production capabilities.\n\n- On a company basis, Taiwan Union Technology is expected to benefit from ASP uplift as M9-grade materials are adopted in VR200/300, with a stable share within the NVIDIA supply chain cited as a key strength. The price target was raised from NT$2,060 to NT$2,250.\n\n- Unimicron is seeing a continued increase in AI PCB revenue mix, driven by expanding share in ASIC projects for Amazon Web Services, Google, and Meta. 2025–2027 EPS is projected at NT$19.24, NT$37.90, and NT$67.13, respectively, with the price target raised from NT$805 to NT$925.\n\n- Elite Material is considered a primary beneficiary of the supply shortage, expected to absorb transition demand and expand its share within the AI server supply chain. 2025–2027 EPS is estimated at NT$13.03, NT$28.39, and NT$48.51, with the price target adjusted from NT$615 to NT$650.\n\n- BofA also views Unimicron positively. Citing full utilization of high-end capacity and strong ASIC demand, the company has raised this year's CAPEX to a record level. January revenue beat expectations, leading to a 2026 revenue forecast of YoY +44%. 2026–2028 EPS estimates were revised upward by 5–10%, and the valuation multiple was lifted from 18.5x to 20x.\n\n- At the broader industry level, the Taiwan Printed Circuit Association (TPCA) and the Industrial Technology Research Institute (ITRI) forecast global PCB output value at $92.36B in 2025 (YoY +15.4%) and $105.2B in 2026 (YoY +13.9%). AI is now firmly established as the core driver of industry premiumization and value-added growth.\n\n- On the supply chain front, U.S.–China trade tensions and tariff barriers are reshaping the global division of labor. According to TPCA, amid the 'de-China' trend, U.S.-based customers are expanding non-China production in sensitive areas such as AI servers and low-earth orbit satellites, with Taiwanese PCB makers absorbing the resulting transition demand. Under the latest Taiwan–U.S. tariff regime, the export duty on Taiwan-made PCBs to the U.S. stands at ~15%, highlighting a competitive edge versus China's 45%. Taiwan has already emerged as the largest PCB import source for the U.S. in 2025. However, establishing local U.S. factories requires careful consideration of ROI and customer negotiation terms, and in the near term, Asia-based manufacturing is judged to retain a competitive advantage.\n\n---\n\nhttps://t.co/mLTPWaylcy", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.0026444083928649498, "alpha_c_m1d": -0.0049924858826995155, "bench_spy_m1d": 0.0026444083928649498, "bench_c_m1d": 0.0049924858826995155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.007231864589122949, "alpha_c_p1d": -0.004778590693303775, "bench_spy_p1d": 0.007231864589122949, "bench_c_p1d": 0.004778590693303775, "ret_p1w": 0.30487804878048785, "ret_signed_p1w": 0.30487804878048785, "alpha_spy_p1w": 0.29783618817563684, "alpha_c_p1w": 0.29917239732743894, "bench_spy_p1w": 0.0070418606048510135, "bench_c_p1w": 0.00570565145304891, "ret_p1m": 0.5013550135501355, "ret_signed_p1m": 0.5013550135501355, "alpha_spy_p1m": 0.5512303888191811, "alpha_c_p1m": 0.5364452657699382, "bench_spy_p1m": -0.04987537526904562, "bench_c_p1m": -0.035090252219802776, "ret_p3m": 1.21680216802168, "ret_signed_p3m": 1.21680216802168, "alpha_spy_p3m": 1.141922179008828, "alpha_c_p3m": 0.9796906879846499, "bench_spy_p3m": 0.07487998901285198, "bench_c_p3m": 0.2371114800370302, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5556, -0.523, -0.5163, -0.4932, -0.4946, -0.4661, -0.4688, -0.4675, -0.4621, -0.4173, -0.4024, -0.3875, -0.3835, -0.3902, -0.3902, -0.3997, -0.4201, -0.4255, -0.439, -0.4173, -0.4024, -0.416, -0.4119, -0.4119, -0.4024, -0.4024, -0.4051, -0.4038, -0.4038, -0.4065, -0.3767, -0.3753, -0.3875, -0.3835, -0.3889, -0.3401, -0.3117, -0.3225, -0.3238, -0.2575, -0.1843, -0.1653, -0.1301, -0.1138, -0.0813, -0.0976, -0.0407, -0.0041, -0.0664, 0.0257, -0.0108, 0.0434, 0.0515, -0.0095, -0.0786, -0.0583, 0.0068, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0989, 0.2087, 0.2425, 0.3049, 0.3049, 0.3035, 0.2547, 0.1301, 0.2195, 0.1707, 0.0542, 0.1491, 0.2629, 0.3008, 0.3767, 0.4878, 0.4715, 0.5718, 0.5745, 0.5014, 0.3523, 0.2466, 0.3713, 0.3604, 0.3686, 0.3374, 0.2046, 0.3238, 0.4065, 0.4065, 0.4065, 0.5285, 0.6802, 0.6938, 0.729, 0.6965, 0.6179, 0.6748, 0.748, 0.7425, 0.8374, 0.9241, 0.9404, 0.9729, 1.1409, 1.2737, 1.2358, 1.1762, 1.393, 1.393, 1.4688, 1.4472, 1.3252, 1.4282, 1.2168, 1.3333, 1.3713, 1.4173, 1.3875, 1.2249, 1.2114, 1.2168, 1.2304, 1.4526, 1.6287, 1.6829, 1.9404, 1.9268, 1.7778, 1.8591, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2024451374889783501", "ticker_idx": 2, "ts": "2026-02-19T11:49:54+00:00", "symbol": "Elite Material", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Elite Material is considered a primary beneficiary of the supply shortage, expected to absorb transition demand and expand its share within the AI server supply chain. The price target adjusted from NT$615 to NT$650.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Goldman/BofA Buy calls on Taiwan PCB/CCL names — Taiwan Union, Unimicron, Elite Material — with raised PTs on AI server ABF supercycle.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Elite Material Co. 2383.TW Taiwan CCL", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Taiwan Stock Market 2026 Barometer: ABF 'Supercycle' Ignites… AI Infrastructure Driving PCB & CCL Demand Explosion\n\n- As AI server and high-performance computing (HPC) demand ramps up in earnest, the long-struggling PCB sector has entered a strong rebound phase. After enduring weak consumer electronics and inventory destocking pressure, the advancement and higher specifications of AI server architectures are driving across-the-board increases in substrate layer counts, material grades, and per-unit value, serving as key catalysts for both industry fundamentals and share prices. Related names have surged sharply year-to-date, drawing significant market attention.\n\n- Capital is flowing back into high-end materials and process supply chains, with PCB, ABF, and CCL simultaneously entering a new growth cycle. Most major players posted January revenues well above year-ago levels, hitting all-time highs. High-end PCB makers exposed to AI server demand led the earnings recovery, while ABF substrates are also seeing a rapidly tightening supply-demand structure as AI/HPC proliferates, prompting assessments of a 'supercycle' entry.\n\n- In a January report, Goldman Sachs identified PCB and CCL as key beneficiary sectors of next-generation AI infrastructure. The global AI server PCB market is projected to surge from $3.1B in 2024 to $27.1B in 2027, with 2026–2027 YoY growth of +113% and +117%, respectively. The upstream CCL market is expected to expand from $1.5B in 2024 to $18.7B in 2027, with even steeper 2026–2027 YoY growth of +142% and +222%. As AI servers rapidly evolve toward M9-grade materials and higher-layer designs, R&D and CAPEX barriers are rising, creating a competitive environment that favors leading players with the requisite technology and production capabilities.\n\n- On a company basis, Taiwan Union Technology is expected to benefit from ASP uplift as M9-grade materials are adopted in VR200/300, with a stable share within the NVIDIA supply chain cited as a key strength. The price target was raised from NT$2,060 to NT$2,250.\n\n- Unimicron is seeing a continued increase in AI PCB revenue mix, driven by expanding share in ASIC projects for Amazon Web Services, Google, and Meta. 2025–2027 EPS is projected at NT$19.24, NT$37.90, and NT$67.13, respectively, with the price target raised from NT$805 to NT$925.\n\n- Elite Material is considered a primary beneficiary of the supply shortage, expected to absorb transition demand and expand its share within the AI server supply chain. 2025–2027 EPS is estimated at NT$13.03, NT$28.39, and NT$48.51, with the price target adjusted from NT$615 to NT$650.\n\n- BofA also views Unimicron positively. Citing full utilization of high-end capacity and strong ASIC demand, the company has raised this year's CAPEX to a record level. January revenue beat expectations, leading to a 2026 revenue forecast of YoY +44%. 2026–2028 EPS estimates were revised upward by 5–10%, and the valuation multiple was lifted from 18.5x to 20x.\n\n- At the broader industry level, the Taiwan Printed Circuit Association (TPCA) and the Industrial Technology Research Institute (ITRI) forecast global PCB output value at $92.36B in 2025 (YoY +15.4%) and $105.2B in 2026 (YoY +13.9%). AI is now firmly established as the core driver of industry premiumization and value-added growth.\n\n- On the supply chain front, U.S.–China trade tensions and tariff barriers are reshaping the global division of labor. According to TPCA, amid the 'de-China' trend, U.S.-based customers are expanding non-China production in sensitive areas such as AI servers and low-earth orbit satellites, with Taiwanese PCB makers absorbing the resulting transition demand. Under the latest Taiwan–U.S. tariff regime, the export duty on Taiwan-made PCBs to the U.S. stands at ~15%, highlighting a competitive edge versus China's 45%. Taiwan has already emerged as the largest PCB import source for the U.S. in 2025. However, establishing local U.S. factories requires careful consideration of ROI and customer negotiation terms, and in the near term, Asia-based manufacturing is judged to retain a competitive advantage.\n\n---\n\nhttps://t.co/mLTPWaylcy", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": -0.0026444083928649498, "alpha_c_m1d": -0.0049924858826995155, "bench_spy_m1d": 0.0026444083928649498, "bench_c_m1d": 0.0049924858826995155, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.007231864589122949, "alpha_c_p1d": -0.004778590693303775, "bench_spy_p1d": 0.007231864589122949, "bench_c_p1d": 0.004778590693303775, "ret_p1w": 0.11161731207289294, "ret_signed_p1w": 0.11161731207289294, "alpha_spy_p1w": 0.10457545146804192, "alpha_c_p1w": 0.10591166061984403, "bench_spy_p1w": 0.0070418606048510135, "bench_c_p1w": 0.00570565145304891, "ret_p1m": 0.2801822323462415, "ret_signed_p1m": 0.2801822323462415, "alpha_spy_p1m": 0.33005760761528713, "alpha_c_p1m": 0.3152724845660443, "bench_spy_p1m": -0.04987537526904562, "bench_c_p1m": -0.035090252219802776, "ret_p3m": 1.0660592255125283, "ret_signed_p3m": 1.0660592255125283, "alpha_spy_p3m": 0.9911792364996763, "alpha_c_p3m": 0.8289477454754981, "bench_spy_p3m": 0.07487998901285198, "bench_c_p3m": 0.2371114800370302, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3986, -0.3394, -0.3303, -0.3189, -0.3052, -0.328, -0.344, -0.3485, -0.3508, -0.3189, -0.3235, -0.3166, -0.2711, -0.2733, -0.2688, -0.2688, -0.3189, -0.3189, -0.3166, -0.2938, -0.2711, -0.2642, -0.2301, -0.2301, -0.2528, -0.2369, -0.2506, -0.2506, -0.2506, -0.2756, -0.262, -0.2278, -0.2802, -0.2961, -0.287, -0.2665, -0.2551, -0.2392, -0.2847, -0.2597, -0.2802, -0.2642, -0.2893, -0.2437, -0.18, -0.2073, -0.1708, -0.1344, -0.1845, -0.205, -0.1868, -0.1549, -0.0706, -0.1116, -0.123, -0.0524, -0.041, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0091, 0.0524, 0.1025, 0.1116, 0.1116, 0.0843, 0.0569, -0.0023, 0.0774, 0.0752, -0.0091, 0.082, 0.1162, 0.1845, 0.2323, 0.1777, 0.2301, 0.2301, 0.3166, 0.2802, 0.2733, 0.2232, 0.2961, 0.3417, 0.3257, 0.2688, 0.1845, 0.303, 0.2688, 0.2688, 0.2688, 0.3144, 0.4442, 0.467, 0.5581, 0.5148, 0.6424, 0.7358, 0.7472, 0.7312, 0.754, 0.7995, 0.8519, 0.8679, 1.0387, 1.1549, 0.9932, 1.0228, 1.0706, 1.0706, 1.1663, 1.1549, 1.2551, 1.2597, 1.164, 1.2096, 1.2323, 1.246, 1.2415, 1.1025, 1.0706, 1.0661, 1.0023, 1.1367, 1.2802, 1.508, 1.41, 1.4169, 1.303, 1.3326, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2028731098755146036", "ticker_idx": 0, "ts": "2026-03-03T07:16:00+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics' foundry division has set a target of exceeding 2 trillion won in operating profit next year… management believes a return to profit is possible as early as Q4 of this year, with full-year profitability achievable next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung Foundry targets 2T won OP next year on Tesla AI chip ramp, Exynos volume, and mature-node expansion; first full-year profit in 5 years possible.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Foundry Targets Over $1.3 Billion in Operating Profit Next Year… Attempting Return to Black After 5 Years\n\nSamsung Electronics' foundry (semiconductor contract manufacturing) division has reportedly established an internal target of surpassing 2 trillion won in operating profit next year. This comes as big tech orders are ramping up in earnest—starting with Tesla's AI chips—and as mature-node customer production volumes expand. Attention is now focused on whether the foundry division, which had been posting trillion-won-scale operating losses amid an order drought, can achieve its first full-year profitability in five years.\n\nAccording to industry sources on the 3rd, Samsung Electronics' foundry division has set a target of exceeding 2 trillion won in operating profit next year. The division had been recording losses in the trillions since 2023, plagued by poor order intake due to low yields at advanced nodes. However, with advanced processes now stabilizing and mature-node demand expanding, management believes a return to profit is possible as early as Q4 of this year, with full-year profitability achievable next year.\n\nNext Year: Tesla AI Chip Mass Production Begins, Exynos Volumes Expand\n\nSamsung's foundry division had been unable to supply Exynos application processors (APs) for the Galaxy series produced by Samsung's MX (Mobile eXperience) division for several years due to poor yields. However, as 2-nanometer process yields have begun to stabilize, the division has been supplying the Exynos 2600 starting this year. Galaxy models launching next year are expected to include more Exynos-equipped variants, driving higher production volumes for the foundry division.\n\nSamsung's foundry division plans to mass-produce Tesla's AI chips on the 2nm process at its Taylor, Texas fab next year. It will manufacture AI5 and AI6—the AI chips serving as the brains behind Tesla's Optimus humanoid robot, robotaxi, and autonomous driving systems. Notably, the AI6 chip is reported to be exclusively manufactured by Samsung's foundry division.\n\nBuilding on stabilized advanced-node yields, the division is also actively pursuing orders from Qualcomm, AMD, and others. Qualcomm was previously a Samsung foundry customer, and the likelihood of renewed orders is growing as the 2nm process stabilizes. AMD is also seen as likely to come knocking at Samsung's door, given TSMC's constrained production capacity.\n\nMature Node Orders Also Expanding… Rising Demand for HBM Logic Dies, Grok AI Chips, and More\n\nIncreased utilization at mature nodes (4–8nm) is also seen as contributing to improved profitability. Through its mature processes, Samsung manufactures the logic die that serves as the brain of 6th-generation High Bandwidth Memory (HBM4), as well as AI chips for xAI's Grok (the AI semiconductor startup backed by Nvidia), IBM chips, Nintendo chips, and more. As the HBM market expands, demand for foundry-produced logic dies is growing, while order volumes from tech companies are also reportedly increasing.\n\nAn industry source noted: \"The 4–8nm mature processes have achieved stability through repeated high-volume production. Existing customers leveraging these nodes are increasing their production volumes, which is driving improved profitability.\"\n\n---\n\nhttps://t.co/uXyCJNnCAn", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.10968733550663545, "ret_signed_m1d": 0.10968733550663545, "alpha_spy_m1d": 0.10079461356259167, "alpha_c_m1d": 0.09485101865809864, "bench_spy_m1d": 0.00889272194404378, "bench_c_m1d": 0.014836316848536812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.11737572377080441, "ret_signed_p1d": -0.11737572377080441, "alpha_spy_p1d": -0.12443103212388906, "alpha_c_p1d": -0.1343938159729241, "bench_spy_p1d": 0.007055308353084655, "bench_c_p1d": 0.017018092202119695, "ret_p1w": -0.03690413526337277, "ret_signed_p1w": -0.03690413526337277, "alpha_spy_p1w": -0.03227394297320518, "alpha_c_p1w": -0.053340450334133704, "bench_spy_p1w": -0.00463019229016759, "bench_c_p1w": 0.016436315070760932, "ret_p1m": -0.02592240876830343, "ret_signed_p1m": -0.02592240876830343, "alpha_spy_p1m": 0.008326530771928464, "alpha_c_p1m": -0.008305983897664349, "bench_spy_p1m": -0.03424893954023189, "bench_c_p1m": -0.01761642487063908, "ret_p3m": 0.6281708221484199, "ret_signed_p3m": 0.6281708221484199, "alpha_spy_p3m": 0.5132032465385132, "alpha_c_p3m": 0.23720714544139399, "bench_spy_p3m": 0.11496757560990667, "bench_c_p3m": 0.3909636767070259, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4639, -0.4471, -0.4415, -0.4471, -0.4491, -0.4527, -0.4445, -0.4654, -0.4756, -0.4496, -0.4512, -0.4578, -0.4364, -0.4313, -0.4333, -0.4333, -0.4032, -0.3875, -0.3854, -0.3854, -0.3854, -0.3414, -0.2922, -0.2881, -0.2773, -0.2886, -0.2875, -0.2886, -0.2947, -0.2809, -0.2624, -0.2368, -0.2348, -0.2558, -0.2337, -0.2194, -0.2204, -0.2204, -0.1825, -0.1676, -0.1763, -0.1773, -0.2291, -0.1415, -0.1333, -0.1835, -0.1871, -0.1471, -0.1502, -0.1399, -0.0846, -0.0712, -0.0712, -0.0712, -0.0712, -0.0261, -0.0256, -0.0108, 0.0251, 0.0431, 0.1174, 0.1097, 0.1097, 0.0, -0.1174, -0.0179, -0.0354, -0.1107, -0.0369, -0.0261, -0.0369, -0.0595, -0.0328, -0.0062, 0.0687, 0.0277, 0.022, -0.0451, -0.0277, -0.0313, -0.0769, -0.0769, -0.0945, -0.1412, -0.0259, -0.0837, -0.0436, -0.0082, 0.0093, 0.0812, 0.0478, 0.0581, 0.0324, 0.0606, 0.0837, 0.1171, 0.1094, 0.1017, 0.1248, 0.1171, 0.1531, 0.1274, 0.1531, 0.1402, 0.1608, 0.1325, 0.1325, 0.1942, 0.1942, 0.3662, 0.3945, 0.3791, 0.4664, 0.433, 0.4587, 0.5203, 0.3893, 0.4433, 0.415, 0.4176, 0.5383, 0.5023, 0.5023, 0.5357, 0.5768, 0.5383, 0.6282, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2059190597328757064", "ticker_idx": 0, "ts": "2026-05-26T08:31:10+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HOLY SMOKES", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Reacts bullishly to UBS raising MU PT, endorsing the memory LTA pricing thesis.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Our supply chain work on Long Term Agreements (LTAs) across the memory industry suggests that up to 30% of DDR volumes industry-wide will be soon locked in at pricing that is just slightly below current levels\"\n\nHOLY SMOKES", "reply_text": null, "parent_text": null, "quoted_text": "UBS raising $MU pt to... $1,625 https://t.co/jVO4G5ldq1", "parent_author": null, "ret_m1d": -0.16171809181271302, "ret_signed_m1d": -0.16171809181271302, "alpha_spy_m1d": -0.15512326370425955, "alpha_c_m1d": -0.11883768742820466, "bench_spy_m1d": -0.0065948281084534655, "bench_c_m1d": -0.04288040438450835, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03631063098229759, "ret_signed_p1d": 0.03631063098229759, "alpha_spy_p1d": 0.03648383456635951, "alpha_c_p1d": 0.04733799087684687, "bench_spy_p1d": -0.0001732035840619206, "bench_c_p1d": -0.011027359894549282, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5363, -0.5399, -0.5396, -0.5764, -0.5528, -0.557, -0.5868, -0.5656, -0.5502, -0.5328, -0.5477, -0.5245, -0.5071, -0.4849, -0.4848, -0.5043, -0.5281, -0.5488, -0.5587, -0.5737, -0.6034, -0.6014, -0.6408, -0.6229, -0.5894, -0.5912, -0.5912, -0.5783, -0.5785, -0.546, -0.5295, -0.5305, -0.5239, -0.4802, -0.4907, -0.4896, -0.492, -0.4995, -0.4984, -0.4559, -0.4623, -0.4456, -0.4145, -0.4371, -0.4213, -0.4227, -0.3948, -0.3566, -0.2854, -0.2559, -0.2782, -0.1664, -0.1122, -0.1443, -0.103, -0.1338, -0.1911, -0.2393, -0.2201, -0.1829, -0.1493, -0.1617, -0.1617, 0.0, 0.0363, 0.0309, 0.0839, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2038927755988160898", "ticker_idx": 0, "ts": "2026-03-31T10:33:53+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung may end up buying more BESI equipment than expected.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish BESI: Samsung likely to buy more Besi hybrid bonding equipment than expected as SEMES alternative is less mature.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Samsung Electronics has likewise already introduced Besi equipment for development purposes, and has recently begun quality-testing a hybrid bonder from SEMES as well. However, the SEMES tool is understood to be less mature in terms of completeness.\"\n\nSamsung may end up buying more BESI equipment than expected.", "reply_text": null, "parent_text": null, "quoted_text": "SK Hynix Places First Order for Applied Materials–Besi Hybrid Bonding Equipment\n\nSK Hynix has decided to purchase production-grade hybrid bonding equipment for the development of next-generation High Bandwidth Memory (HBM).\n\nAccording to industry sources on March 31, SK Hynix https://t.co/BvIErBRI6C", "parent_author": null, "ret_m1d": -0.026543746693787118, "ret_signed_m1d": -0.026543746693787118, "alpha_spy_m1d": 0.001703101714370714, "alpha_c_m1d": 0.02789025291348013, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04805814558620303, "ret_signed_p1d": 0.04805814558620303, "alpha_spy_p1d": 0.040523681803313005, "alpha_c_p1d": 0.025705491802560543, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.06929316010243514, "ret_signed_p1w": 0.06929316010243514, "alpha_spy_p1w": 0.055638850847332666, "alpha_c_p1w": 0.026257165677106897, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.34623781393738473, "ret_signed_p1m": 0.34623781393738473, "alpha_spy_p1m": 0.25207170435547366, "alpha_c_p1m": 0.04321226744468154, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2526, -0.1668, -0.1408, -0.1009, -0.1126, -0.1556, -0.1556, -0.0939, -0.0657, -0.0975, -0.0296, -0.0338, -0.0567, -0.0338, -0.0274, -0.0221, -0.0218, -0.0229, -0.0179, -0.0531, -0.0916, -0.0807, -0.0802, -0.1051, -0.1196, -0.1017, -0.0752, -0.0645, -0.0553, -0.045, -0.0573, -0.0154, -0.0025, 0.0064, 0.045, -0.0332, 0.031, 0.0486, 0.0746, 0.1042, 0.0556, 0.0584, 0.0534, 0.0134, 0.0673, 0.0542, -0.1266, -0.0707, -0.0243, -0.0224, -0.0207, 0.0344, 0.0226, 0.0531, 0.0752, 0.0274, -0.0025, 0.0201, 0.0288, 0.0366, 0.036, -0.0229, -0.0265, 0.0, 0.0481, 0.0626, 0.0626, 0.0626, 0.0693, 0.1567, 0.1517, 0.1819, 0.1562, 0.2182, 0.216, 0.2322, 0.2674, 0.2685, 0.2735, 0.2981, 0.3512, 0.4093, 0.3946, 0.3232, 0.3462, 0.3901, 0.3901, 0.3738, 0.4283, 0.434, 0.4323, 0.4666, 0.4694, 0.4047, 0.452, 0.4998, 0.4722, 0.4385, 0.4334, 0.4902, 0.5206, 0.5386, 0.5858, 0.597, 0.5543, 0.6173, 0.5993, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2053635859086459244", "ticker_idx": 0, "ts": "2026-05-11T00:38:38+00:00", "symbol": "Ajinomoto", "kind": "equity", "geography": "Japan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Ajinomoto is finally raising ABF film prices by 30%, apparently", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Ajinomoto on 30% ABF film price hike; rising prices are a positive earnings catalyst.", "resolved_tickers": ["2802.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ajinomoto Japan 2802.T; ABF film maker", "bench_c": "XLP", "bench_c_reason": "industry='Packaged Foods'", "bench_c_industry": "Packaged Foods", "unit_kind": "quote", "lang": "en", "tweet_text": "Congratulations to the hedge funds that bought stakes in Ajinomoto and pushed for ABF film price hikes.\n\nAjinomoto is finally raising ABF film prices by 30%, apparently. https://t.co/NY2UeAxtef", "reply_text": null, "parent_text": null, "quoted_text": "Prices of ABF substrates, key to heat control in AI chip packages, are expected to rise further in the 2nd half amid rising material costs, higher spot prices, and tighter capacity, media report, citing unnamed supply chain sources. Taiwan’s 3 top makers all reported record high", "parent_author": null, "ret_m1d": -0.08528823422440446, "ret_signed_m1d": -0.08528823422440446, "alpha_spy_m1d": -0.08301582423722653, "alpha_c_m1d": -0.09500392970130023, "bench_spy_m1d": -0.0022724099871779257, "bench_c_m1d": 0.00971569547689577, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.010911074740861948, "ret_signed_p1d": -0.010911074740861948, "alpha_spy_p1d": -0.009396134749410034, "alpha_c_p1d": -0.023745423542272825, "bench_spy_p1d": -0.0015149399914519135, "bench_c_p1d": 0.012834348801410878, "ret_p1w": -0.0556464811783961, "ret_signed_p1w": -0.0556464811783961, "alpha_spy_p1w": -0.0547673206351279, "alpha_c_p1w": -0.08599311301175772, "bench_spy_p1w": -0.0008791605432681981, "bench_c_p1w": 0.030346631833361615, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2213, -0.2121, -0.1976, -0.1773, -0.1612, -0.1607, -0.1654, -0.1676, -0.1676, -0.1504, -0.1068, -0.1191, -0.1013, -0.1073, -0.1473, -0.18, -0.1871, -0.1827, -0.214, -0.209, -0.1918, -0.2045, -0.2032, -0.19, -0.1878, -0.1721, -0.2032, -0.2032, -0.2446, -0.2234, -0.1894, -0.1835, -0.1806, -0.1913, -0.2004, -0.1595, -0.1528, -0.1613, -0.1531, -0.1638, -0.1315, -0.1429, -0.154, -0.1622, -0.1604, -0.1513, -0.1471, -0.1682, -0.1622, -0.1415, -0.1529, -0.1653, -0.1348, -0.1297, -0.1331, -0.1331, -0.0744, -0.1122, -0.1122, -0.1122, -0.1122, -0.0984, -0.0853, 0.0, -0.0109, -0.0029, 0.0078, -0.0418, -0.0556, -0.0584, -0.0784, -0.0587, -0.0362, -0.0044, -0.0282, -0.0382, -0.0649, -0.0631, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039620475169694168", "ticker_idx": 0, "ts": "2026-04-02T08:26:30+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics is estimated to have recorded Q1 operating profit of KRW 40–45 trillion… both companies would be setting new records roughly double their prior highs", "tweet_type": "analysis", "timeline": "quarters", "summary": "Record-breaking memory earnings for Samsung, SK Hynix, and Micron; continued DRAM/NAND price surge and upward revisions expected through at least 2027.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Prices Double… Samsung and SK Hynix Profits Likely Doubled Too\n\nSamsung Electronics and SK Hynix are estimated to have posted record-breaking earnings in Q1.\n\nThis is attributed to strong HBM sales momentum, compounded by steep price increases in commodity DRAM and NAND flash.\n\nAccording to sell-side estimates as of April 2, Samsung Electronics is estimated to have recorded Q1 operating profit of KRW 40–45 trillion, while SK Hynix is estimated at KRW 36–39 trillion. At the start of the year, Samsung's OP consensus stood at KRW 36–37 trillion and SK Hynix at around KRW 31 trillion, but continued memory price increases led to upward revisions throughout the quarter.\n\nSamsung's previous all-time quarterly OP record was KRW 20.1 trillion, achieved in Q4 last year. SK Hynix's was KRW 19.17 trillion, also set in Q4. If sell-side estimates prove accurate, both companies would be setting new records roughly double their prior highs. Companies generating over KRW 10 trillion in monthly operating profit are exceedingly rare globally — limited to the likes of NVIDIA, Apple, Alphabet, Microsoft, and Saudi Aramco.\n\nMicron's earnings, reported last month, were also a blowout. FY2026 Q2 (Dec 2025–Feb 2026) revenue came in at $23.86 billion, beating its own guidance of $18.7 billion by 27%. Non-GAAP operating income was $16.5 billion, exceeding the guidance-implied estimate of $11.3 billion by 46%.\n\nAccording to TrendForce, Q1 DRAM and NAND flash contract prices rose 90–95% and 55–60% QoQ, respectively. Further increases of 58–63% for DRAM and 70–75% for NAND are forecast for Q2. As such, upward earnings revisions across the memory industry are likely to continue for the foreseeable future.\n\nThe industry expects new cleanroom expansions to translate into actual supply additions no earlier than late 2027 to 2028. AI cloud service providers are currently negotiating 5-year long-term purchase agreements to secure volumes. Supply shortages in the commodity market are set to persist.\n\nThe consumer electronics industry is feeling the pain. IDC forecasts that global smartphone shipments will decline 12.9% and PC shipments 11.3% this year due to component cost pressures from surging memory prices. IDC described the impact of soaring memory prices as \"tsunami-level.\" Memory prices have risen to the point where OEMs are actively downgrading product specs or eliminating budget lineups entirely. Display panel manufacturers are also reportedly lowering their shipment targets for the year as a result.\n\nJitesh Ubrani, Research Manager at IDC, said, \"The memory shortage will persist through 2027. Even if prices begin to moderate from 2028, they will not return to 2025 levels.\"\n\nAn industry insider noted, \"The key mid-to-long-term variables for the memory industry are whether memory efficiency technologies like Google's TurboQuant actually reduce server purchase volumes, and whether cloud datacenter players maintain their current pace of capex.\"\n\n---\n\nhttps://t.co/XWSJO8CEN8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06306053811659185, "ret_signed_m1d": 0.06306053811659185, "alpha_spy_m1d": 0.0639602024199839, "alpha_c_m1d": 0.07100231168398075, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.007941773567388899, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043721973094170474, "ret_signed_p1d": 0.043721973094170474, "alpha_spy_p1d": 0.043721973094170474, "alpha_c_p1d": 0.043721973094170474, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.1434977578475336, "ret_signed_p1w": 0.1434977578475336, "alpha_spy_p1w": 0.10678099370560079, "alpha_c_p1w": 0.09878857639698713, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.044709181450546476, "ret_p1m": 0.23598654708520184, "ret_signed_p1m": 0.23598654708520184, "alpha_spy_p1m": 0.13714994974272465, "alpha_c_p1m": 0.045678520181530846, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.190308026903671, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2275, -0.223, -0.2113, -0.2236, -0.2225, -0.2236, -0.2303, -0.2152, -0.1951, -0.1671, -0.1648, -0.1878, -0.1637, -0.1481, -0.1492, -0.1492, -0.1078, -0.0916, -0.1011, -0.1022, -0.1587, -0.063, -0.0541, -0.1089, -0.1128, -0.0692, -0.0725, -0.0614, -0.0009, 0.0136, 0.0136, 0.0136, 0.0136, 0.0628, 0.0634, 0.0796, 0.1188, 0.1383, 0.2194, 0.2111, 0.2111, 0.0914, -0.0367, 0.0718, 0.0528, -0.0295, 0.0511, 0.0628, 0.0511, 0.0265, 0.0556, 0.0846, 0.1663, 0.1216, 0.1154, 0.0421, 0.0611, 0.0572, 0.0074, 0.0074, -0.0118, -0.0628, 0.0631, 0.0, 0.0437, 0.0824, 0.1015, 0.1799, 0.1435, 0.1547, 0.1267, 0.1575, 0.1827, 0.2192, 0.2108, 0.2024, 0.2276, 0.2192, 0.2584, 0.2304, 0.2584, 0.2444, 0.2668, 0.236, 0.236, 0.3033, 0.3033, 0.491, 0.5219, 0.505, 0.6003, 0.5639, 0.5919, 0.6592, 0.5163, 0.5751, 0.5443, 0.5471, 0.6788, 0.6396, 0.6396, 0.676, 0.7209, 0.6788, 0.7769, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039620475169694168", "ticker_idx": 1, "ts": "2026-04-02T08:26:30+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is estimated at KRW 36–39 trillion… upward earnings revisions across the memory industry are likely to continue for the foreseeable future", "tweet_type": "analysis", "timeline": "quarters", "summary": "Record-breaking memory earnings for Samsung, SK Hynix, and Micron; continued DRAM/NAND price surge and upward revisions expected through at least 2027.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Prices Double… Samsung and SK Hynix Profits Likely Doubled Too\n\nSamsung Electronics and SK Hynix are estimated to have posted record-breaking earnings in Q1.\n\nThis is attributed to strong HBM sales momentum, compounded by steep price increases in commodity DRAM and NAND flash.\n\nAccording to sell-side estimates as of April 2, Samsung Electronics is estimated to have recorded Q1 operating profit of KRW 40–45 trillion, while SK Hynix is estimated at KRW 36–39 trillion. At the start of the year, Samsung's OP consensus stood at KRW 36–37 trillion and SK Hynix at around KRW 31 trillion, but continued memory price increases led to upward revisions throughout the quarter.\n\nSamsung's previous all-time quarterly OP record was KRW 20.1 trillion, achieved in Q4 last year. SK Hynix's was KRW 19.17 trillion, also set in Q4. If sell-side estimates prove accurate, both companies would be setting new records roughly double their prior highs. Companies generating over KRW 10 trillion in monthly operating profit are exceedingly rare globally — limited to the likes of NVIDIA, Apple, Alphabet, Microsoft, and Saudi Aramco.\n\nMicron's earnings, reported last month, were also a blowout. FY2026 Q2 (Dec 2025–Feb 2026) revenue came in at $23.86 billion, beating its own guidance of $18.7 billion by 27%. Non-GAAP operating income was $16.5 billion, exceeding the guidance-implied estimate of $11.3 billion by 46%.\n\nAccording to TrendForce, Q1 DRAM and NAND flash contract prices rose 90–95% and 55–60% QoQ, respectively. Further increases of 58–63% for DRAM and 70–75% for NAND are forecast for Q2. As such, upward earnings revisions across the memory industry are likely to continue for the foreseeable future.\n\nThe industry expects new cleanroom expansions to translate into actual supply additions no earlier than late 2027 to 2028. AI cloud service providers are currently negotiating 5-year long-term purchase agreements to secure volumes. Supply shortages in the commodity market are set to persist.\n\nThe consumer electronics industry is feeling the pain. IDC forecasts that global smartphone shipments will decline 12.9% and PC shipments 11.3% this year due to component cost pressures from surging memory prices. IDC described the impact of soaring memory prices as \"tsunami-level.\" Memory prices have risen to the point where OEMs are actively downgrading product specs or eliminating budget lineups entirely. Display panel manufacturers are also reportedly lowering their shipment targets for the year as a result.\n\nJitesh Ubrani, Research Manager at IDC, said, \"The memory shortage will persist through 2027. Even if prices begin to moderate from 2028, they will not return to 2025 levels.\"\n\nAn industry insider noted, \"The key mid-to-long-term variables for the memory industry are whether memory efficiency technologies like Google's TurboQuant actually reduce server purchase volumes, and whether cloud datacenter players maintain their current pace of capex.\"\n\n---\n\nhttps://t.co/XWSJO8CEN8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08072283167818717, "ret_signed_m1d": 0.08072283167818717, "alpha_spy_m1d": 0.08162249598157922, "alpha_c_m1d": 0.08161497609341295, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.0008921444152257818, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05542168946917325, "ret_signed_p1d": 0.05542168946917325, "alpha_spy_p1d": 0.05542168946917325, "alpha_c_p1d": 0.05542168946917325, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.2024095895548803, "ret_signed_p1w": 0.2024095895548803, "alpha_spy_p1w": 0.1656928254129475, "alpha_c_p1w": 0.10557539928892345, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.09683419026595685, "ret_p1m": 0.5493975649543825, "ret_signed_p1m": 0.5493975649543825, "alpha_spy_p1m": 0.45056096761190534, "alpha_c_p1m": 0.24989716271541051, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.299500402238972, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.163, -0.1269, -0.1077, -0.0908, -0.1053, -0.0993, -0.1125, -0.1077, -0.0993, -0.0908, -0.0812, -0.1065, -0.1101, -0.092, -0.0776, -0.1149, -0.0379, 0.0114, 0.0354, 0.0932, -0.0018, 0.0908, 0.0823, 0.0126, 0.009, 0.0667, 0.0535, 0.0342, 0.0679, 0.0583, 0.0583, 0.0583, 0.0583, 0.0751, 0.1413, 0.1437, 0.2086, 0.2242, 0.3241, 0.2783, 0.2783, 0.1313, 0.0229, 0.1337, 0.1133, 0.0072, 0.1301, 0.1506, 0.1205, 0.0964, 0.1735, 0.1687, 0.2723, 0.2205, 0.2133, 0.1241, 0.188, 0.1988, 0.1241, 0.1241, 0.0518, -0.0277, 0.0807, 0.0, 0.0554, 0.0675, 0.1036, 0.2446, 0.2024, 0.2373, 0.253, 0.3289, 0.3687, 0.3916, 0.359, 0.4048, 0.4747, 0.4735, 0.4759, 0.4723, 0.5566, 0.5663, 0.5578, 0.5494, 0.5494, 0.7434, 0.7434, 0.9289, 0.9928, 1.0313, 1.2651, 1.2108, 1.3807, 1.3735, 1.1916, 1.2169, 1.1024, 1.1024, 1.3373, 1.3386, 1.3386, 1.4723, 1.7024, 1.7583, 1.8113, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039620475169694168", "ticker_idx": 2, "ts": "2026-04-02T08:26:30+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron's earnings, reported last month, were also a blowout. FY2026 Q2 revenue came in at $23.86 billion, beating its own guidance of $18.7 billion by 27%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Record-breaking memory earnings for Samsung, SK Hynix, and Micron; continued DRAM/NAND price surge and upward revisions expected through at least 2027.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DRAM Prices Double… Samsung and SK Hynix Profits Likely Doubled Too\n\nSamsung Electronics and SK Hynix are estimated to have posted record-breaking earnings in Q1.\n\nThis is attributed to strong HBM sales momentum, compounded by steep price increases in commodity DRAM and NAND flash.\n\nAccording to sell-side estimates as of April 2, Samsung Electronics is estimated to have recorded Q1 operating profit of KRW 40–45 trillion, while SK Hynix is estimated at KRW 36–39 trillion. At the start of the year, Samsung's OP consensus stood at KRW 36–37 trillion and SK Hynix at around KRW 31 trillion, but continued memory price increases led to upward revisions throughout the quarter.\n\nSamsung's previous all-time quarterly OP record was KRW 20.1 trillion, achieved in Q4 last year. SK Hynix's was KRW 19.17 trillion, also set in Q4. If sell-side estimates prove accurate, both companies would be setting new records roughly double their prior highs. Companies generating over KRW 10 trillion in monthly operating profit are exceedingly rare globally — limited to the likes of NVIDIA, Apple, Alphabet, Microsoft, and Saudi Aramco.\n\nMicron's earnings, reported last month, were also a blowout. FY2026 Q2 (Dec 2025–Feb 2026) revenue came in at $23.86 billion, beating its own guidance of $18.7 billion by 27%. Non-GAAP operating income was $16.5 billion, exceeding the guidance-implied estimate of $11.3 billion by 46%.\n\nAccording to TrendForce, Q1 DRAM and NAND flash contract prices rose 90–95% and 55–60% QoQ, respectively. Further increases of 58–63% for DRAM and 70–75% for NAND are forecast for Q2. As such, upward earnings revisions across the memory industry are likely to continue for the foreseeable future.\n\nThe industry expects new cleanroom expansions to translate into actual supply additions no earlier than late 2027 to 2028. AI cloud service providers are currently negotiating 5-year long-term purchase agreements to secure volumes. Supply shortages in the commodity market are set to persist.\n\nThe consumer electronics industry is feeling the pain. IDC forecasts that global smartphone shipments will decline 12.9% and PC shipments 11.3% this year due to component cost pressures from surging memory prices. IDC described the impact of soaring memory prices as \"tsunami-level.\" Memory prices have risen to the point where OEMs are actively downgrading product specs or eliminating budget lineups entirely. Display panel manufacturers are also reportedly lowering their shipment targets for the year as a result.\n\nJitesh Ubrani, Research Manager at IDC, said, \"The memory shortage will persist through 2027. Even if prices begin to moderate from 2028, they will not return to 2025 levels.\"\n\nAn industry insider noted, \"The key mid-to-long-term variables for the memory industry are whether memory efficiency technologies like Google's TurboQuant actually reduce server purchase volumes, and whether cloud datacenter players maintain their current pace of capex.\"\n\n---\n\nhttps://t.co/XWSJO8CEN8", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.004396067911945689, "ret_signed_m1d": 0.004396067911945689, "alpha_spy_m1d": 0.005295732215337745, "alpha_c_m1d": 0.005288212327171471, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.0008921444152257818, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.1509120276458067, "ret_signed_p1w": 0.1509120276458067, "alpha_spy_p1w": 0.11419526350387388, "alpha_c_p1w": 0.054077837379849836, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.09683419026595685, "ret_p1m": 0.4804773821276844, "ret_signed_p1m": 0.4804773821276844, "alpha_spy_p1m": 0.38164078478520724, "alpha_c_p1m": 0.18097697988871242, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.299500402238972, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.148, -0.0627, -0.0733, -0.1075, -0.0581, -0.056, -0.0771, -0.0902, -0.0812, -0.0099, -0.0099, -0.0038, 0.062, 0.0851, 0.0908, 0.0619, 0.1197, 0.188, 0.1894, 0.1323, 0.1949, 0.1448, 0.0355, 0.045, 0.0772, 0.0467, 0.0187, 0.1199, 0.1298, 0.1235, 0.1235, 0.0911, 0.1489, 0.1391, 0.1686, 0.149, 0.1409, 0.1709, 0.1342, 0.1255, 0.1263, 0.0363, 0.0938, 0.0837, 0.0107, 0.0626, 0.1002, 0.1427, 0.1063, 0.163, 0.2058, 0.2601, 0.2602, 0.2125, 0.1542, 0.1036, 0.0795, 0.0428, -0.0298, -0.025, -0.1213, -0.0775, 0.0044, 0.0, 0.0, 0.0315, 0.031, 0.1106, 0.1509, 0.1484, 0.1647, 0.2715, 0.2457, 0.2484, 0.2425, 0.2244, 0.227, 0.331, 0.3153, 0.3563, 0.4323, 0.3769, 0.4156, 0.4121, 0.4805, 0.574, 0.748, 0.8201, 0.7656, 1.0391, 1.1716, 1.0931, 1.1943, 1.1189, 0.9786, 0.8609, 0.9079, 0.9987, 1.0809, 1.0506, 1.0506, 1.4462, 1.535, 1.5216, 1.6513, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2041076972177285495", "ticker_idx": 0, "ts": "2026-04-06T08:54:06+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML is still cheap", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long ASML: EUV slots sold out through 2027, Samsung ordered 20 tools for one fab, still cheap with DRAM capacity expansion thesis playing out.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "I've been calling ASML since October last year, predicting that DRAM would face a severe supply shortage, and that as the three major DRAM makers ramp up capacity expansions, ASML's EUV tools would become the subject of an intense scramble. \n\nMy thesis is gradually playing out. ASML's EUV slots are already fully sold out through 2027, and negotiations for 2028 allocations are now underway. \n\nMost recently, Samsung alone ordered 20 EUV systems for a single fab. \n\nASML is still cheap.", "reply_text": null, "parent_text": null, "quoted_text": "Samsung Electronics Orders ~20 EUV Tools for P5… \"First Cleanroom Completion Next Year\"\n\nSamsung Electronics has reportedly placed orders for approximately 20 extreme ultraviolet (EUV) lithography tools—critical equipment for sub-10nm advanced processes—with Dutch semiconductor https://t.co/ThQQiLGq1y", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.004704559847623746, "alpha_c_m1d": 0.009242900046880775, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04065457609372802, "ret_signed_p1d": -0.04065457609372802, "alpha_spy_p1d": -0.04109465019957681, "alpha_c_p1d": -0.050554022533946474, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.08509917040466486, "ret_signed_p1w": 0.08509917040466486, "alpha_spy_p1w": 0.043865680941202356, "alpha_c_p1w": -0.03450278938516327, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.06091089416063, "ret_signed_p1m": 0.06091089416063, "alpha_spy_p1m": -0.037491102373427454, "alpha_c_p1m": -0.2590799840139075, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0953, -0.1286, -0.0696, -0.0655, -0.0516, -0.0674, -0.0113, 0.004, -0.0363, -0.0194, -0.0067, 0.0119, 0.0135, 0.0133, 0.0473, 0.0274, 0.0253, 0.0456, 0.0535, 0.0239, -0.0189, -0.0111, 0.0269, 0.0377, 0.0276, 0.0403, 0.0162, 0.0253, 0.0293, 0.0329, 0.0722, 0.0665, 0.0815, 0.076, 0.0882, 0.1097, 0.0615, 0.0624, 0.0425, 0.0007, 0.0334, 0.0215, -0.0121, -0.0115, 0.0336, 0.0326, 0.0167, 0.016, 0.0314, 0.0283, 0.0258, 0.0065, -0.0283, 0.0124, 0.035, 0.0436, 0.0041, -0.0126, -0.0422, -0.036, 0.0229, 0.0, 0.0, 0.0, -0.0407, 0.0444, 0.0636, 0.0939, 0.0851, 0.1061, 0.0594, 0.0531, 0.0715, 0.0724, 0.0655, 0.0765, 0.0541, 0.0787, 0.0468, 0.0115, 0.0295, 0.0552, 0.0552, 0.025, 0.0609, 0.1265, 0.1222, 0.1389, 0.127, 0.0927, 0.1455, 0.18, 0.1279, 0.0915, 0.0782, 0.1507, 0.1612, 0.2163, 0.2349, 0.1997, 0.1882, 0.2008, 0.1954, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2050439372126986554", "ticker_idx": 0, "ts": "2026-05-02T04:56:56+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Are you still bearish on NAND, anon?", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish NAND: Samsung LTA floor price above 0.5, implying eSSD pricing will be stronger than Q3; challenges bears.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "LTAs : We believe that Samsung’s floor price of LTA is above 0.5!\n\nThis means the LTA eSSD price is even higher than this year’s Q3 eSSD price…\n\nAre you still bearish on NAND, anon?", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.055421315490927214, "ret_signed_m1d": -0.055421315490927214, "alpha_spy_m1d": -0.059098164855881374, "alpha_c_m1d": -0.06140012077163204, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.046079400030769824, "ret_signed_p1d": 0.046079400030769824, "alpha_spy_p1d": 0.03805721414577756, "alpha_c_p1d": 0.014705470733828996, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.28018250459069777, "ret_signed_p1w": 0.28018250459069777, "alpha_spy_p1w": 0.25053113798329724, "alpha_c_p1w": 0.14300539406857976, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3912, -0.4153, -0.4111, -0.4051, -0.4176, -0.396, -0.3639, -0.355, -0.358, -0.3675, -0.3638, -0.3532, -0.3409, -0.338, -0.3207, -0.3175, -0.2964, -0.3068, -0.3076, -0.3701, -0.3947, -0.3662, -0.3886, -0.4077, -0.3632, -0.3379, -0.3543, -0.3478, -0.3139, -0.306, -0.2664, -0.288, -0.3071, -0.3411, -0.3359, -0.3365, -0.3808, -0.3831, -0.4149, -0.4214, -0.3552, -0.3771, -0.3599, -0.3397, -0.3323, -0.2591, -0.2526, -0.2337, -0.2119, -0.1657, -0.1832, -0.1621, -0.186, -0.1858, -0.1628, -0.1277, -0.1276, -0.1224, -0.081, -0.0953, -0.078, -0.0722, -0.0554, 0.0, 0.0461, 0.1059, 0.1384, 0.2163, 0.2802, 0.2441, 0.3041, 0.2825, 0.2039, 0.2274, 0.2143, 0.2361, 0.3398, 0.3312, 0.3755, 0.4479, 0.482, 0.4925, 0.5637, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044659280620687757", "ticker_idx": 0, "ts": "2026-04-16T06:08:55+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Revenue: US$39.0–40.2bn vs. consensus: US$38.1bn (midpoint +4% above)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "TSMC 2Q26 guidance midpoints beat consensus on revenue (+4%), GPM (+2.4ppt), and OPM (+2.3ppt); all metrics above Street.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "▶ TSMC 2Q26 Guidance\n- Revenue: US$39.0–40.2bnvs. consensus: US$38.1bn (midpoint +4% above)\n- GPM: 65.5%–67.5%vs. consensus: 64.1% (midpoint +2.4ppt above)\n- OPM: 56.5%–58.5%vs. consensus: 55.2% (midpoint +2.3ppt above)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0023980815347721673, "ret_signed_m1d": -0.0023980815347721673, "alpha_spy_m1d": 5.320650923135162e-05, "alpha_c_m1d": 0.0015596753721786039, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.026378896882493952, "ret_signed_p1d": -0.026378896882493952, "alpha_spy_p1d": -0.03846458201271796, "alpha_c_p1d": -0.04695940726149039, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": -0.0023980815347721673, "ret_signed_p1w": -0.0023980815347721673, "alpha_spy_p1w": -0.012075189140141163, "alpha_c_p1w": -0.06187481644370196, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.08633093525179847, "ret_signed_p1m": 0.08633093525179847, "alpha_spy_p1m": 0.032871979060352, "alpha_c_p1m": -0.13693212936535026, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1586, -0.1514, -0.1682, -0.1586, -0.1538, -0.161, -0.1491, -0.1299, -0.1371, -0.1514, -0.1562, -0.1395, -0.1467, -0.1562, -0.1491, -0.1323, -0.1013, -0.0845, -0.0845, -0.0845, -0.0845, -0.0845, -0.0845, -0.0845, -0.0845, -0.0917, -0.0606, -0.0367, -0.0463, -0.0463, -0.0558, -0.075, -0.1084, -0.0917, -0.0965, -0.1347, -0.1156, -0.0726, -0.0989, -0.1084, -0.118, -0.1031, -0.0863, -0.1127, -0.1175, -0.1319, -0.1319, -0.1151, -0.1175, -0.1271, -0.1463, -0.1559, -0.1103, -0.1319, -0.1319, -0.1319, -0.1079, -0.0647, -0.0624, -0.0408, -0.0456, -0.0144, -0.0024, 0.0, -0.0264, -0.0288, -0.0168, -0.0168, -0.0024, 0.048, 0.0863, 0.0624, 0.0456, 0.024, 0.024, 0.0911, 0.0791, 0.0791, 0.1079, 0.0983, 0.0719, 0.0815, 0.0647, 0.0887, 0.0863, 0.0743, 0.0576, 0.048, 0.0695, 0.0815, 0.1079, 0.0887, 0.1031, 0.1007, 0.1295, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042043368927322442", "ticker_idx": 0, "ts": "2026-04-09T00:54:13+00:00", "symbol": "memory", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "the adoption of CXL will inevitably reduce memory usage", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish on memory sector: CXL adoption will reduce overall memory consumption, contrary to consensus bullish view.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "2408.TW"], "resolution_method": "basket", "resolution_reason": "Global memory sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Many people seem to think CXL is a positive for memory companies, but I see it a bit differently.\n\nCXL, by its very origin, came from the idea of how to use less memory more efficiently. Because of that, the adoption of CXL will inevitably reduce memory usage. That is precisely why Google is seriously considering CXL, especially in a memory apocalypse like the one we are seeing now.\n\nBelow is a passage quoted from The Diligence Stack:\n\n\"If CXL-attached memory pools become cost-effective at scale, they would absorb some of the overflow we currently expect to land in NVMe flash. This would not eliminate the storage opportunity but would resize it, potentially materially.\"\n\n---\n\nhttps://t.co/dtWzNPZI0m", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029918348942173267, "ret_signed_m1d": -0.029918348942173267, "alpha_spy_m1d": -0.0356543486082308, "alpha_c_m1d": -0.0470919776581854, "bench_spy_m1d": -0.005735999666057534, "bench_c_m1d": -0.017173628716012135, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021364975333430458, "ret_signed_p1d": -0.021364975333430458, "alpha_spy_p1d": -0.02202675584798003, "alpha_c_p1d": -0.006096895666459379, "bench_spy_p1d": -0.0006617805145495703, "bench_c_p1d": 0.01526807966697108, "ret_p1w": 0.08742436631729594, "ret_signed_p1w": -0.08742436631729594, "alpha_spy_p1w": -0.055434837027987294, "alpha_c_p1w": -0.030511930228612483, "bench_spy_p1w": 0.03198952928930865, "bench_c_p1w": 0.05691243608868346, "ret_p1m": 0.5284724436822463, "ret_signed_p1m": -0.5284724436822463, "alpha_spy_p1m": -0.44359352698427257, "alpha_c_p1m": -0.21188677000126233, "bench_spy_p1m": 0.08487891669797376, "bench_c_p1m": 0.316585673680984, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1465, -0.1662, -0.1568, -0.1475, -0.1089, -0.076, -0.0885, -0.0954, -0.0643, -0.0542, -0.0505, -0.0239, 0.0291, 0.0324, 0.0677, 0.0096, 0.0187, 0.0165, -0.0323, -0.0403, -0.0126, -0.0264, 0.0025, 0.0249, 0.0247, 0.0247, 0.0177, 0.0302, 0.0424, 0.0626, 0.0758, 0.1071, 0.109, 0.1364, 0.1232, 0.1228, 0.0117, -0.0532, 0.0195, -0.0347, -0.0915, -0.0145, 0.0296, -0.0049, -0.0048, 0.0556, 0.0849, 0.1298, 0.0794, 0.0322, -0.0268, -0.0225, -0.0169, -0.0603, -0.0666, -0.1055, -0.1505, -0.0674, -0.1117, -0.0906, -0.0729, -0.0546, 0.0299, 0.0, 0.0214, 0.0348, 0.0738, 0.0717, 0.0874, 0.0733, 0.0734, 0.1098, 0.1291, 0.1214, 0.1209, 0.1861, 0.1864, 0.1949, 0.1619, 0.1768, 0.2783, 0.3399, 0.4663, 0.4806, 0.5285, 0.6596, 0.6477, 0.7111, 0.7323, 0.5968, 0.5815, 0.5239, 0.5449, 0.6623, 0.6688, 0.6511, 0.7819, 0.8692, 0.8834, 0.9721, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050136199621456079", "ticker_idx": 0, "ts": "2026-05-01T08:52:14+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek insiders reportedly expect that his arrival will make it much easier to manage the technical and mass-production risks associated with the first adoption of EMIB-T than previously feared", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish MediaTek (EMIB-T risk reduced by VP hire) and INTC packaging (TPU v9 order held); bearish TSMC on lost packaging share.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Former TSMC packaging vice president Douglas Yu has joined MediaTek, and MediaTek insiders reportedly expect that his arrival will make it much easier to manage the technical and mass-production risks associated with the first adoption of EMIB-T than previously feared.\n\nThe media outlet that reported the news cited an industry source as saying, “TSMC assumed customers would not leave, but they did.” The outlet also commented that while TSMC is now rushing to fill the packaging gap between CoWoS and SoW, based on the current timeline, it appears difficult for TSMC to reclaim the TPU v9 packaging order.\n\n$INTC\n\n---\n\nhttps://t.co/F0qID1wsNQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0027614670118818463, "alpha_c_m1d": 0.006080589343258502, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09961685823754785, "ret_signed_p1d": 0.09961685823754785, "alpha_spy_p1d": 0.1032802379072898, "alpha_c_p1d": 0.10556012985452456, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.3908045977011494, "ret_signed_p1w": 0.3908045977011494, "alpha_spy_p1w": 0.3672564534865943, "alpha_c_p1w": 0.2795497040361896, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3123, -0.3103, -0.3218, -0.3448, -0.2989, -0.2931, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.3123, -0.3008, -0.2835, -0.2548, -0.2548, -0.272, -0.3046, -0.341, -0.3199, -0.3238, -0.3621, -0.3467, -0.3238, -0.3161, -0.341, -0.3448, -0.3372, -0.3372, -0.3563, -0.3487, -0.3774, -0.3793, -0.3793, -0.3908, -0.3927, -0.4215, -0.4291, -0.4387, -0.4387, -0.4387, -0.4387, -0.4368, -0.3946, -0.3966, -0.3966, -0.3793, -0.341, -0.3142, -0.2739, -0.2625, -0.272, -0.1992, -0.1207, -0.1513, -0.067, -0.067, 0.0019, -0.0134, 0.0, 0.0, 0.0996, 0.2088, 0.3142, 0.3103, 0.3908, 0.4866, 0.4176, 0.3391, 0.3046, 0.249, 0.3027, 0.2088, 0.2375, 0.3602, 0.4789, 0.6264, 0.6341, 0.7778, 0.6897, 0.6513, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050136199621456079", "ticker_idx": 1, "ts": "2026-05-01T08:52:14+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC assumed customers would not leave, but they did. based on the current timeline, it appears difficult for TSMC to reclaim the TPU v9 packaging order", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish MediaTek (EMIB-T risk reduced by VP hire) and INTC packaging (TPU v9 order held); bearish TSMC on lost packaging share.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Former TSMC packaging vice president Douglas Yu has joined MediaTek, and MediaTek insiders reportedly expect that his arrival will make it much easier to manage the technical and mass-production risks associated with the first adoption of EMIB-T than previously feared.\n\nThe media outlet that reported the news cited an industry source as saying, “TSMC assumed customers would not leave, but they did.” The outlet also commented that while TSMC is now rushing to fill the packaging gap between CoWoS and SoW, based on the current timeline, it appears difficult for TSMC to reclaim the TPU v9 packaging order.\n\n$INTC\n\n---\n\nhttps://t.co/F0qID1wsNQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.051596057497848125, "ret_signed_m1d": -0.051596057497848125, "alpha_spy_m1d": -0.04883459048596628, "alpha_c_m1d": -0.04551546815458962, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0385465153725556, "ret_signed_p1d": -0.0385465153725556, "alpha_spy_p1d": -0.03488313570281365, "alpha_c_p1d": -0.03260324375557888, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.2539650143026253, "ret_signed_p1w": 0.2539650143026253, "alpha_spy_p1w": 0.2304168700880702, "alpha_c_p1w": 0.14271012063766553, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5056, -0.5121, -0.5158, -0.4922, -0.4957, -0.5269, -0.5153, -0.5334, -0.5303, -0.5303, -0.5364, -0.5437, -0.5521, -0.5572, -0.562, -0.537, -0.5294, -0.5437, -0.5422, -0.5433, -0.5674, -0.5425, -0.5387, -0.5641, -0.5425, -0.5304, -0.5184, -0.5458, -0.5406, -0.5407, -0.5577, -0.548, -0.5364, -0.5596, -0.5582, -0.5577, -0.5264, -0.5573, -0.5671, -0.5865, -0.557, -0.5179, -0.4943, -0.4943, -0.4903, -0.4689, -0.4083, -0.3804, -0.3738, -0.3457, -0.3595, -0.3481, -0.3124, -0.3124, -0.3405, -0.3349, -0.3448, -0.3297, -0.1715, -0.1469, -0.1516, -0.0489, -0.0516, 0.0, -0.0385, 0.0856, 0.1344, 0.1004, 0.254, 0.2993, 0.2107, 0.2075, 0.1637, 0.0918, 0.0858, 0.1122, 0.1941, 0.1895, 0.203, 0.203, 0.2399, 0.2223, 0.2135, 0.1512, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050136199621456079", "ticker_idx": 2, "ts": "2026-05-01T08:52:14+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "based on the current timeline, it appears difficult for TSMC to reclaim the TPU v9 packaging order", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish MediaTek (EMIB-T risk reduced by VP hire) and INTC packaging (TPU v9 order held); bearish TSMC on lost packaging share.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Former TSMC packaging vice president Douglas Yu has joined MediaTek, and MediaTek insiders reportedly expect that his arrival will make it much easier to manage the technical and mass-production risks associated with the first adoption of EMIB-T than previously feared.\n\nThe media outlet that reported the news cited an industry source as saying, “TSMC assumed customers would not leave, but they did.” The outlet also commented that while TSMC is now rushing to fill the packaging gap between CoWoS and SoW, based on the current timeline, it appears difficult for TSMC to reclaim the TPU v9 packaging order.\n\n$INTC\n\n---\n\nhttps://t.co/F0qID1wsNQ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.0027614670118818463, "alpha_c_m1d": -0.006080589343258502, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06557377049180335, "ret_signed_p1d": -0.06557377049180335, "alpha_spy_p1d": -0.0692371501615453, "alpha_c_p1d": -0.07151704210878007, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.07259953161592514, "ret_signed_p1w": -0.07259953161592514, "alpha_spy_p1w": -0.04905138740137005, "alpha_c_p1w": 0.038655362049034636, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1597, -0.1667, -0.176, -0.169, -0.1526, -0.1223, -0.106, -0.106, -0.106, -0.106, -0.106, -0.106, -0.106, -0.106, -0.113, -0.0826, -0.0593, -0.0686, -0.0686, -0.0779, -0.0966, -0.1293, -0.113, -0.1176, -0.155, -0.1363, -0.0943, -0.12, -0.1293, -0.1386, -0.1241, -0.1077, -0.1335, -0.1382, -0.1522, -0.1522, -0.1358, -0.1382, -0.1475, -0.1663, -0.1756, -0.1311, -0.1522, -0.1522, -0.1522, -0.1288, -0.0867, -0.0843, -0.0632, -0.0679, -0.0375, -0.0258, -0.0234, -0.0492, -0.0515, -0.0398, -0.0398, -0.0258, 0.0234, 0.0609, 0.0375, 0.0211, 0.0, 0.0, 0.0656, 0.0539, 0.0539, 0.082, 0.0726, 0.0468, 0.0562, 0.0398, 0.0632, 0.0609, 0.0492, 0.0328, 0.0234, 0.0445, 0.0562, 0.082, 0.0632, 0.0773, 0.0749, 0.103, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2051616159527817628", "ticker_idx": 0, "ts": "2026-05-05T10:53:04+00:00", "symbol": "LITE", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Maintain Buy rating; raise target price to US$1,174", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/adopts GF Buy on Lumentum, PT raised to $1,174 on laser supply tightness, booked-out capacity through 2028, and EML pricing power.", "resolved_tickers": ["LITE"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Communication Equipment", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications\n\n☄ Lumentum (LITE US, Buy): Earnings Preview — Outlook Becoming More Certain\n\n☀ Maintain Buy rating; raise target price to US$1,174:\nWe forecast FY3Q/FY4Q26 revenue of US$840 million / US$1.02 billion, versus consensus of US$830 million / US$970 million. We estimate EPS of US$2.4 / US$3.1, versus consensus of around US$2.3 / US$2.7. We expect earnings visibility to improve further as laser capacity approaches being largely booked out through 2028, alongside tightening EML supply and stronger pricing power. Accordingly, we raise our CY26/CY27 EPS estimates to US$13 / US$31. To reflect improved medium- to long-term visibility, we roll our valuation forward to CY28 and raise our target price to US$1,174, based on 20x CY28 P/E.\n\n☀ Structural supply tightness, combined with improving demand visibility, strengthens EML pricing:\nUpstream laser supply remains structurally tight. The increase in prepayments from optical module makers in 1Q indicates that lasers continue to be in short supply. We expect laser capacity to be largely booked out through 2028. Given the limited number of suppliers and the long lead times required for capacity expansion, we believe EML prices still have room to rise.\n\n☀ CPO / NPO / OCS:\nAlthough the market has recently been discussing CPO yield issues, we believe the industry trend remains intact. At the same time, visibility is improving for NPO solutions in CSP ASICs. As NPO/CPO solutions begin to scale, laser suppliers should fully benefit from their critical position in the value chain. In addition, Google recently disclosed at its conference the use of OCS in TPU v8i, enabling a cluster-level, high-radix, reconfigurable optical interconnect architecture to improve network utilization and system scalability. We expect OCS to be adopted by more cloud service providers over time.\n\n$LITE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018480539060419687, "ret_signed_m1d": -0.018480539060419687, "alpha_spy_m1d": -0.010522196478952361, "alpha_c_m1d": -0.010522196478952361, "bench_spy_m1d": -0.007958342581467326, "bench_c_m1d": -0.007958342581467326, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05055498751723775, "ret_signed_p1d": -0.05055498751723775, "alpha_spy_p1d": -0.06445442696225534, "alpha_c_p1d": -0.06445442696225534, "bench_spy_p1d": 0.013899439445017592, "bench_c_p1d": 0.013899439445017592, "ret_p1w": -0.0022019812246442116, "ret_signed_p1w": -0.0022019812246442116, "alpha_spy_p1w": -0.022111583385098332, "alpha_c_p1w": -0.022111583385098332, "bench_spy_p1w": 0.01990960216045412, "bench_c_p1w": 0.01990960216045412, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4928, -0.445, -0.4197, -0.4358, -0.4227, -0.4133, -0.4342, -0.4342, -0.3963, -0.4025, -0.3609, -0.3286, -0.3216, -0.308, -0.2727, -0.3193, -0.2953, -0.2125, -0.3018, -0.3155, -0.3456, -0.4385, -0.3558, -0.3243, -0.3243, -0.3805, -0.3741, -0.3717, -0.3469, -0.2954, -0.2236, -0.2898, -0.2671, -0.1936, -0.2186, -0.3074, -0.2934, -0.3416, -0.2934, -0.2312, -0.1686, -0.1686, -0.2235, -0.1798, -0.0991, -0.101, -0.0978, -0.1241, -0.1425, -0.1715, -0.1039, -0.101, -0.1, -0.1585, -0.1216, -0.1485, -0.1135, -0.1356, -0.2043, -0.137, -0.0927, -0.0449, -0.0185, 0.0, -0.0506, -0.1025, -0.0913, 0.0589, -0.0022, 0.036, 0.0073, -0.024, -0.1102, -0.105, -0.1272, -0.0302, -0.0479, -0.0479, -0.0842, -0.0928, -0.1347, -0.1404, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2050024550176522452", "ticker_idx": 0, "ts": "2026-05-01T01:28:35+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "If even Apple — arguably the strongest buyer in the world — is acknowledging memory cost pressure, then this cycle is not over yet", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory cycle not over — AAPL acknowledging rising costs signals further price increases; TSMC leading-edge tight; Amkor raised as a bullish idea on Apple supply chain.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts from Apple’s earnings call:\n\nMemory does not seem to be the bottleneck preventing Apple from shipping products. Rather, it is a cost variable pressuring margins. In contrast, SoC / leading-edge node capacity appears to be the supply constraint that is actually limiting unit shipments.\n\nApple seems to be able to secure enough memory. The issue is that it cannot ship enough products because it is not getting sufficient SoC supply due to tight TSMC leading-edge capacity.\n\nMac mini and Mac Studio demand has been stronger than expected, as they are increasingly being recognized as strong platforms for AI and agentic tools.\n\n→ There is currently no AI platform that can really match Apple’s unified memory architecture.\n\nAmkor bullish?\n\nMemory costs were already higher in the March quarter, and Apple expects significantly higher memory costs in the June quarter.\n\n→ Memory prices are likely to rise even more sharply in calendar 2Q. If even Apple — arguably the strongest buyer in the world — is acknowledging memory cost pressure, then this cycle is not over yet.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.015399966687562938, "ret_signed_m1d": -0.015399966687562938, "alpha_spy_m1d": -0.012638499675681092, "alpha_c_m1d": -0.009319377344304436, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08092171543720195, "ret_signed_p1d": 0.08092171543720195, "alpha_spy_p1d": 0.0845850951069439, "alpha_c_p1d": 0.08686498705417867, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.3020245252385733, "ret_signed_p1w": 0.3020245252385733, "alpha_spy_p1w": 0.2784763810240182, "alpha_c_p1w": 0.19076963157361354, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2549, -0.2789, -0.3065, -0.3011, -0.2838, -0.2939, -0.2722, -0.2464, -0.246, -0.246, -0.2533, -0.2403, -0.2256, -0.2046, -0.2041, -0.1814, -0.166, -0.1309, -0.145, -0.1448, -0.229, -0.2739, -0.2231, -0.249, -0.2823, -0.2257, -0.2085, -0.2264, -0.2254, -0.1914, -0.1723, -0.128, -0.1619, -0.1783, -0.2286, -0.2152, -0.2222, -0.268, -0.267, -0.3094, -0.3304, -0.2547, -0.29, -0.2663, -0.2462, -0.2334, -0.164, -0.1738, -0.1638, -0.1643, -0.1157, -0.1061, -0.0907, -0.1013, -0.0978, -0.0754, -0.0545, -0.047, -0.0461, -0.0032, -0.0174, -0.0045, -0.0154, 0.0, 0.0809, 0.1201, 0.2269, 0.2367, 0.302, 0.4078, 0.3687, 0.4356, 0.4352, 0.3259, 0.3207, 0.2983, 0.3195, 0.4241, 0.407, 0.407, 0.5346, 0.6162, 0.6139, 0.681, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2050024550176522452", "ticker_idx": 1, "ts": "2026-05-01T01:28:35+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "SoC / leading-edge node capacity appears to be the supply constraint that is actually limiting unit shipments... it is not getting sufficient SoC supply due to tight TSMC leading-edge capacity", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory cycle not over — AAPL acknowledging rising costs signals further price increases; TSMC leading-edge tight; Amkor raised as a bullish idea on Apple supply chain.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts from Apple’s earnings call:\n\nMemory does not seem to be the bottleneck preventing Apple from shipping products. Rather, it is a cost variable pressuring margins. In contrast, SoC / leading-edge node capacity appears to be the supply constraint that is actually limiting unit shipments.\n\nApple seems to be able to secure enough memory. The issue is that it cannot ship enough products because it is not getting sufficient SoC supply due to tight TSMC leading-edge capacity.\n\nMac mini and Mac Studio demand has been stronger than expected, as they are increasingly being recognized as strong platforms for AI and agentic tools.\n\n→ There is currently no AI platform that can really match Apple’s unified memory architecture.\n\nAmkor bullish?\n\nMemory costs were already higher in the March quarter, and Apple expects significantly higher memory costs in the June quarter.\n\n→ Memory prices are likely to rise even more sharply in calendar 2Q. If even Apple — arguably the strongest buyer in the world — is acknowledging memory cost pressure, then this cycle is not over yet.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0027614670118818463, "alpha_c_m1d": 0.006080589343258502, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06557377049180335, "ret_signed_p1d": 0.06557377049180335, "alpha_spy_p1d": 0.0692371501615453, "alpha_c_p1d": 0.07151704210878007, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.07259953161592514, "ret_signed_p1w": 0.07259953161592514, "alpha_spy_p1w": 0.04905138740137005, "alpha_c_p1w": -0.038655362049034636, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1597, -0.1667, -0.176, -0.169, -0.1526, -0.1223, -0.106, -0.106, -0.106, -0.106, -0.106, -0.106, -0.106, -0.106, -0.113, -0.0826, -0.0593, -0.0686, -0.0686, -0.0779, -0.0966, -0.1293, -0.113, -0.1176, -0.155, -0.1363, -0.0943, -0.12, -0.1293, -0.1386, -0.1241, -0.1077, -0.1335, -0.1382, -0.1522, -0.1522, -0.1358, -0.1382, -0.1475, -0.1663, -0.1756, -0.1311, -0.1522, -0.1522, -0.1522, -0.1288, -0.0867, -0.0843, -0.0632, -0.0679, -0.0375, -0.0258, -0.0234, -0.0492, -0.0515, -0.0398, -0.0398, -0.0258, 0.0234, 0.0609, 0.0375, 0.0211, 0.0, 0.0, 0.0656, 0.0539, 0.0539, 0.082, 0.0726, 0.0468, 0.0562, 0.0398, 0.0632, 0.0609, 0.0492, 0.0328, 0.0234, 0.0445, 0.0562, 0.082, 0.0632, 0.0773, 0.0749, 0.103, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2050024550176522452", "ticker_idx": 2, "ts": "2026-05-01T01:28:35+00:00", "symbol": "Amkor", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Amkor bullish?", "tweet_type": "analysis", "timeline": "quarters", "summary": "Memory cycle not over — AAPL acknowledging rising costs signals further price increases; TSMC leading-edge tight; Amkor raised as a bullish idea on Apple supply chain.", "resolved_tickers": ["AMKR"], "resolution_method": "exchange_ticker", "resolution_reason": "Amkor Technology AMKR US", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts from Apple’s earnings call:\n\nMemory does not seem to be the bottleneck preventing Apple from shipping products. Rather, it is a cost variable pressuring margins. In contrast, SoC / leading-edge node capacity appears to be the supply constraint that is actually limiting unit shipments.\n\nApple seems to be able to secure enough memory. The issue is that it cannot ship enough products because it is not getting sufficient SoC supply due to tight TSMC leading-edge capacity.\n\nMac mini and Mac Studio demand has been stronger than expected, as they are increasingly being recognized as strong platforms for AI and agentic tools.\n\n→ There is currently no AI platform that can really match Apple’s unified memory architecture.\n\nAmkor bullish?\n\nMemory costs were already higher in the March quarter, and Apple expects significantly higher memory costs in the June quarter.\n\n→ Memory prices are likely to rise even more sharply in calendar 2Q. If even Apple — arguably the strongest buyer in the world — is acknowledging memory cost pressure, then this cycle is not over yet.\n\n$AAPL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.018849295356854756, "ret_signed_m1d": -0.018849295356854756, "alpha_spy_m1d": -0.01608782834497291, "alpha_c_m1d": -0.012768706013596254, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.001828629344698407, "ret_signed_p1d": -0.001828629344698407, "alpha_spy_p1d": 0.0018347503250435437, "alpha_c_p1d": 0.004114642272278313, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.07764811586463405, "ret_signed_p1w": 0.07764811586463405, "alpha_spy_p1w": 0.05409997165007896, "alpha_c_p1w": -0.03360677780032573, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3508, -0.3841, -0.3785, -0.307, -0.2626, -0.2494, -0.2114, -0.2757, -0.3334, -0.3334, -0.3424, -0.3438, -0.3213, -0.3269, -0.3391, -0.3186, -0.2848, -0.3188, -0.3286, -0.3288, -0.374, -0.3574, -0.3744, -0.4158, -0.3932, -0.3831, -0.3834, -0.4195, -0.3953, -0.3722, -0.3482, -0.3407, -0.323, -0.3556, -0.3512, -0.293, -0.2977, -0.365, -0.3747, -0.4197, -0.3666, -0.346, -0.3431, -0.3431, -0.3384, -0.3301, -0.2626, -0.2244, -0.1847, -0.1498, -0.1374, -0.1599, -0.1152, -0.0523, -0.0232, -0.0124, 0.0163, 0.0256, 0.0986, 0.0637, 0.0038, -0.0068, -0.0188, 0.0, -0.0018, 0.0791, 0.0862, 0.0166, 0.0776, 0.0788, 0.0332, 0.0495, 0.0141, -0.0104, -0.071, -0.0781, -0.0366, -0.073, -0.0751, -0.0751, 0.0333, 0.0155, -0.0072, -0.0215, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2039702363418878123", "ticker_idx": 0, "ts": "2026-04-02T13:51:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi raises its price target on Samsung Electronics to KRW 300,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Citi's raised PT on Samsung to KRW 300,000 with Q1 2026 operating profit forecast of KRW 51T; implied long.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "* Citi raises its price target on Samsung Electronics to KRW 300,000.\n\n* Citi expects Samsung Electronics to post a market-high KRW 51 trillion in operating profit for Q1 2026. ($33.62 billion)\n\n* Citi raises its 2026 global DRAM/NAND ASP forecast to +190% / +172% YoY.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06306053811659185, "ret_signed_m1d": 0.06306053811659185, "alpha_spy_m1d": 0.0639602024199839, "alpha_c_m1d": 0.07100231168398075, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.007941773567388899, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043721973094170474, "ret_signed_p1d": 0.043721973094170474, "alpha_spy_p1d": 0.043721973094170474, "alpha_c_p1d": 0.043721973094170474, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.1434977578475336, "ret_signed_p1w": 0.1434977578475336, "alpha_spy_p1w": 0.10678099370560079, "alpha_c_p1w": 0.09878857639698713, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.044709181450546476, "ret_p1m": 0.23598654708520184, "ret_signed_p1m": 0.23598654708520184, "alpha_spy_p1m": 0.13714994974272465, "alpha_c_p1m": 0.045678520181530846, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.190308026903671, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2275, -0.223, -0.2113, -0.2236, -0.2225, -0.2236, -0.2303, -0.2152, -0.1951, -0.1671, -0.1648, -0.1878, -0.1637, -0.1481, -0.1492, -0.1492, -0.1078, -0.0916, -0.1011, -0.1022, -0.1587, -0.063, -0.0541, -0.1089, -0.1128, -0.0692, -0.0725, -0.0614, -0.0009, 0.0136, 0.0136, 0.0136, 0.0136, 0.0628, 0.0634, 0.0796, 0.1188, 0.1383, 0.2194, 0.2111, 0.2111, 0.0914, -0.0367, 0.0718, 0.0528, -0.0295, 0.0511, 0.0628, 0.0511, 0.0265, 0.0556, 0.0846, 0.1663, 0.1216, 0.1154, 0.0421, 0.0611, 0.0572, 0.0074, 0.0074, -0.0118, -0.0628, 0.0631, 0.0, 0.0437, 0.0824, 0.1015, 0.1799, 0.1435, 0.1547, 0.1267, 0.1575, 0.1827, 0.2192, 0.2108, 0.2024, 0.2276, 0.2192, 0.2584, 0.2304, 0.2584, 0.2444, 0.2668, 0.236, 0.236, 0.3033, 0.3033, 0.491, 0.5219, 0.505, 0.6003, 0.5639, 0.5919, 0.6592, 0.5163, 0.5751, 0.5443, 0.5471, 0.6788, 0.6396, 0.6396, 0.676, 0.7209, 0.6788, 0.7769, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050815047375712603", "ticker_idx": 0, "ts": "2026-05-03T05:49:44+00:00", "symbol": "Hung Ching Precision", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Hung Ching Precision has secured a leading position in optical-electrical co-test equipment and expects large-scale mass production demand from late this year into next year. ASPs for this equipment are significantly higher than conventional electrical test systems", "tweet_type": "analysis", "timeline": "quarters", "summary": "Hung Ching Precision is the standout long on CPO test equipment + SLT dominance (60–70% share); Auras Technology leads MCCP ramp; Jentech pursues MCL.", "resolved_tickers": ["3443.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hung Ching Precision 3443.TW Taiwan test equipment", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> AI Thermal War Heats Up: MCCP and MCL to Coexist, While CPO Test Demand Surges (Commercial Times, Taiwan)\n\n• As AI chip power density rises sharply, cooling technology competition is becoming a key variable. Ahead of mass production for NVIDIA’s Vera Rubin platform, test equipment maker Hung Ching Precision said MCCP and MCL are more likely to coexist rather than compete directly. MCCP is currently moving quickly into mass production, led by Auras Technology, while MCL is being developed by companies such as Jentech Precision Industrial. MCL carries higher cost and technical difficulty, but also has stronger performance potential. Cooler Master is pursuing both technologies in parallel. Hung Ching Precision expects MCCP to enter mass production first in Q3, improving thermal density efficiency by 2x, while MCL should follow after validation. From the customer perspective, however, adoption will likely be mixed depending on cost and product-specific requirements: some products may use MCL, others MCCP, and others conventional liquid cooling.\n\n• At the same time, the real growth opportunity in the industry is shifting beyond cooling and toward CPO. Hung Ching Precision has secured a leading position in optical-electrical co-test equipment and expects large-scale mass production demand from late this year into next year. ASPs for this equipment are significantly higher than conventional electrical test systems. In addition, the integration of vision inspection systems to protect high-value AI chips is pushing equipment prices up by another 10–15%.\n\n• On top of this, the expansion of ASICs is also driving higher demand for SLT. Hung Ching Precision currently holds roughly 60–70% share in this market. Ultimately, in the AI era, the key battleground extends beyond the chip itself to the full-stack infrastructure of cooling, packaging, and testing. Cooling will likely be a coexistence story, while profitability should expand meaningfully in CPO and test equipment.\n\n---\n\nhttps://t.co/FrX4oQ2bDL", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0907150480256137, "ret_signed_m1d": -0.0907150480256137, "alpha_spy_m1d": -0.09439189739056786, "alpha_c_m1d": -0.09669385330631852, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03201707577374591, "ret_signed_p1d": 0.03201707577374591, "alpha_spy_p1d": 0.023994889888753645, "alpha_c_p1d": 0.0006431464768050787, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.22305229455709719, "ret_signed_p1w": 0.22305229455709719, "alpha_spy_p1w": 0.19340092794969665, "alpha_c_p1w": 0.08587518403497918, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.46, -0.5005, -0.4931, -0.4781, -0.4653, -0.4717, -0.4717, -0.4717, -0.4717, -0.4717, -0.4717, -0.4717, -0.4717, -0.4941, -0.4664, -0.4504, -0.4088, -0.4088, -0.4354, -0.4685, -0.5133, -0.4973, -0.4909, -0.5326, -0.5208, -0.4931, -0.5112, -0.5123, -0.5176, -0.5112, -0.4632, -0.4429, -0.4354, -0.4472, -0.4653, -0.4717, -0.4578, -0.4749, -0.5069, -0.5379, -0.4973, -0.5048, -0.5048, -0.5048, -0.4952, -0.4642, -0.4333, -0.4312, -0.4077, -0.349, -0.3447, -0.286, -0.3052, -0.2359, -0.2028, -0.2156, -0.1654, -0.1387, -0.1558, -0.1057, -0.1206, -0.0907, -0.0907, 0.0, 0.032, 0.0576, 0.1633, 0.1121, 0.2231, 0.1889, 0.1377, 0.0832, 0.0342, 0.0117, -0.047, -0.0171, 0.0811, 0.0971, 0.1334, 0.1281, 0.0886, -0.0203, -0.0032, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038841723972137247", "ticker_idx": 0, "ts": "2026-03-31T04:52:01+00:00", "symbol": "memory semiconductors", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we believe these market concerns are excessive", "tweet_type": "analysis", "timeline": "months", "summary": "Daishin Securities bullish on memory semis: spot price softness is overstated, server/mobile demand remains robust, and market concerns are excessive.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory semiconductor basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "On the Decline in Memory Semiconductor Spot Prices [Daishin Securities Semiconductor / Ryu Hyung-geun]\n\nWhat is the problem?\n\nMemory semiconductor share prices continue to decline. The fundamental backdrop, if there is one, would be macro uncertainty, but from a sector perspective, the market appears to take issue with four key factors:\n\n1. Pricing: Decline in spot prices and a deceleration in memory semiconductor price growth rates.\n\n2. Demand: Concerns that the emergence of AI efficiency solutions such as Turbo Quant could dampen medium-to-long-term memory semiconductor demand.\n\n3. Supply: Concerns over capex increases across the supply industry and the potential recurrence of oversupply.\n\n4. Profitability: The possibility that the pace of profitability improvement may slow amid elevated base effects.\n\nWhere problems exist, a proper diagnosis is warranted. In this report, we examine the decline in spot prices.\n\nA moderation in spot prices does not signal a trend reversal in pricing\n\nRecently, a moderation in spot prices has been observed in certain product segments. The market has historically used spot prices as a tool for diagnosing industry conditions and informing share price decisions, and has been selling shares on the back of elevated price base effects and the recent softening in spot prices. In conclusion, we believe these market concerns are excessive.\n\n- As the demand landscape shifts toward a server-centric structure, the correlation between spot prices and contract prices continues to diminish. The spot market is primarily composed of PC and consumer products, and its share of total market transactions is in the low-single-digit percentage range at best.\n\n- Under the current demand structure, what matters for memory semiconductor earnings are mobile and server products. It is worth reiterating that these segments account for a negligible share of spot market transactions.\n\n- Given the nature of the spot market, the sharp price run-up may have been felt as a burden by consumers. However, we do not believe this phenomenon will drive contract price declines in other application segments.\n\n- According to our channel checks, one major CSP has decided to purchase server DDR4 at a price exceeding that of HBM3e. If genuine demand had begun to weaken, it would be impossible to explain why buyers are paying premiums for legacy products.\n\n- Even in the Chinese market, where spot price declines have been observed, an acute memory semiconductor supply shortage persists. For some mobile customers, downward revisions to annual smartphone shipment targets are unavoidable due to memory supply constraints.\n\n- Some large accounts are seizing this shift as an opportunity to expand market share. Considering the changing power dynamics at the set-maker level, there is a reasonable possibility of an unexpected surprise emerging from the mobile segment.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029577382111775224, "ret_signed_m1d": 0.029577382111775224, "alpha_spy_m1d": 0.05782423051993306, "alpha_c_m1d": 0.08401138171904247, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11154118095690742, "ret_signed_p1d": 0.11154118095690742, "alpha_spy_p1d": 0.1040067171740174, "alpha_c_p1d": 0.08918852717326493, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.14264568502045907, "ret_signed_p1w": 0.14264568502045907, "alpha_spy_p1w": 0.1289913757653566, "alpha_c_p1w": 0.09960969059513083, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.49617902620393045, "ret_signed_p1m": 0.49617902620393045, "alpha_spy_p1m": 0.4020129166220194, "alpha_c_p1m": 0.19315347971122726, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2116, -0.1541, -0.1304, -0.0856, -0.0787, -0.0896, -0.0764, -0.0739, -0.0885, -0.0862, -0.0729, -0.0343, -0.0302, -0.0448, -0.0137, 0.0064, 0.013, -0.0102, 0.0518, 0.0991, 0.1045, 0.1033, 0.0732, 0.1209, 0.0817, 0.0417, 0.0507, 0.075, 0.0591, 0.0931, 0.1297, 0.1293, 0.1293, 0.1176, 0.1385, 0.1582, 0.1918, 0.1912, 0.2245, 0.2477, 0.2975, 0.2757, 0.276, 0.1505, 0.0885, 0.1615, 0.1213, 0.0745, 0.1588, 0.1854, 0.1577, 0.1612, 0.2135, 0.2418, 0.3064, 0.2555, 0.2297, 0.1548, 0.1748, 0.1638, 0.0943, 0.096, 0.0296, 0.0, 0.1115, 0.0598, 0.0944, 0.1237, 0.1426, 0.2476, 0.2348, 0.2499, 0.2512, 0.3267, 0.34, 0.3618, 0.3455, 0.3517, 0.3856, 0.4198, 0.4289, 0.4324, 0.4988, 0.4771, 0.4962, 0.481, 0.5058, 0.63, 0.6929, 0.8493, 0.8625, 0.9685, 1.1304, 1.0705, 1.1753, 1.1695, 1.0056, 0.9927, 0.9594, 0.9932, 1.1503, 1.1259, 1.1259, 1.3276, 1.4545, 1.4539, 1.5538, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039673040712384778", "ticker_idx": 0, "ts": "2026-04-02T11:55:22+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek, Chinese smartphone brands, and KYEC will all take a big hit", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Entry-level smartphones are finished; MediaTek, Chinese smartphone brands, and KYEC all face significant headwinds.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Entry-level smartphones are literally finished.\n\nMediaTek, Chinese smartphone brands, and KYEC will all take a big hit. https://t.co/ANLH0Vsz8O\n\n---\n\nhttps://t.co/EwfuqGLTkh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": -0.0, "alpha_spy_m1d": -0.0008996643033920559, "alpha_c_m1d": -0.0008921444152257818, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.0008921444152257818, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.07508532423208192, "ret_signed_p1w": -0.07508532423208192, "alpha_spy_p1w": -0.038368560090149106, "alpha_c_p1w": 0.021748866033874936, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.09683419026595685, "ret_p1m": 0.7815699658703072, "ret_signed_p1m": -0.7815699658703072, "alpha_spy_p1m": -0.6827333685278301, "alpha_c_p1m": -0.48206956363133524, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.299500402238972, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0212, 0.0137, 0.0171, -0.0137, -0.0307, -0.0137, 0.0137, 0.0239, 0.0171, 0.0273, 0.0137, 0.0137, 0.0, 0.0137, 0.1126, 0.2082, 0.2116, 0.215, 0.215, 0.2014, 0.1638, 0.2253, 0.2287, 0.2082, 0.1672, 0.2491, 0.2594, 0.2662, 0.2662, 0.2662, 0.2662, 0.2662, 0.2662, 0.2662, 0.2662, 0.2253, 0.2457, 0.2765, 0.3276, 0.3276, 0.2969, 0.2389, 0.1741, 0.2116, 0.2048, 0.1365, 0.1638, 0.2048, 0.2184, 0.1741, 0.1672, 0.1809, 0.1809, 0.1468, 0.1604, 0.1092, 0.1058, 0.1058, 0.0853, 0.0819, 0.0307, 0.0171, 0.0, 0.0, 0.0, 0.0, 0.0034, 0.0785, 0.0751, 0.0751, 0.1058, 0.1741, 0.2218, 0.2935, 0.314, 0.2969, 0.4266, 0.5666, 0.5119, 0.6621, 0.6621, 0.785, 0.7577, 0.7816, 0.7816, 0.959, 1.1536, 1.3413, 1.3345, 1.4778, 1.6485, 1.5256, 1.3857, 1.3242, 1.2253, 1.3208, 1.1536, 1.2048, 1.4232, 1.6348, 1.8976, 1.9113, 2.1672, 2.0102, 1.942, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039673040712384778", "ticker_idx": 1, "ts": "2026-04-02T11:55:22+00:00", "symbol": "Chinese smartphone brands", "kind": "sector", "geography": "China", "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek, Chinese smartphone brands, and KYEC will all take a big hit", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Entry-level smartphones are finished; MediaTek, Chinese smartphone brands, and KYEC all face significant headwinds.", "resolved_tickers": ["1810.HK", "2382.HK"], "resolution_method": "basket", "resolution_reason": "Chinese smartphone basket: Xiaomi+Sunny Optical HK", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Entry-level smartphones are literally finished.\n\nMediaTek, Chinese smartphone brands, and KYEC will all take a big hit. https://t.co/ANLH0Vsz8O\n\n---\n\nhttps://t.co/EwfuqGLTkh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.006165808972710074, "ret_signed_m1d": 0.006165808972710074, "alpha_spy_m1d": 0.005266144669318018, "alpha_c_m1d": -0.001775964594678825, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.007941773567388899, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.022379723501534055, "ret_signed_p1w": -0.022379723501534055, "alpha_spy_p1w": 0.014337040640398757, "alpha_c_p1w": 0.02232945794901242, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.044709181450546476, "ret_p1m": 0.0016029652490325286, "ret_signed_p1m": -0.0016029652490325286, "alpha_spy_p1m": 0.09723363209344466, "alpha_c_p1m": 0.18870506165463846, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.190308026903671, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.2025, 0.1877, 0.1738, 0.1486, 0.1524, 0.1665, 0.1579, 0.1522, 0.1652, 0.1537, 0.1307, 0.097, 0.1073, 0.1036, 0.1206, 0.0878, 0.1029, 0.1206, 0.118, 0.0973, 0.0685, 0.0648, 0.0478, 0.0646, 0.0584, 0.0603, 0.0686, 0.0965, 0.0713, 0.0735, 0.071, 0.071, 0.071, 0.071, 0.0479, 0.0832, 0.0632, 0.0647, 0.0504, 0.0513, -0.0018, -0.0442, -0.0365, -0.03, -0.001, 0.0044, 0.0126, 0.0152, 0.0171, 0.0103, 0.0487, 0.0521, 0.0444, 0.0589, -0.0079, -0.041, -0.0293, -0.0169, -0.0336, -0.017, -0.0333, -0.0396, -0.0062, 0.0, 0.0, 0.0, 0.0, 0.0572, 0.0224, 0.0245, 0.0148, 0.0188, 0.0179, 0.0487, 0.0532, 0.0763, 0.0643, 0.0562, 0.0332, 0.0348, 0.0595, 0.0057, 0.021, 0.0016, 0.0016, 0.0534, 0.0491, 0.0454, 0.0648, 0.0764, 0.0575, 0.0394, 0.0641, 0.0679, 0.0392, 0.0265, 0.0191, 0.0118, 0.0529, 0.0972, 0.0972, 0.0991, 0.0871, 0.0772, 0.1535, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039673040712384778", "ticker_idx": 2, "ts": "2026-04-02T11:55:22+00:00", "symbol": "KYEC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek, Chinese smartphone brands, and KYEC will all take a big hit", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Entry-level smartphones are finished; MediaTek, Chinese smartphone brands, and KYEC all face significant headwinds.", "resolved_tickers": ["2449.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "King Yuan Electronics 2449.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Entry-level smartphones are literally finished.\n\nMediaTek, Chinese smartphone brands, and KYEC will all take a big hit. https://t.co/ANLH0Vsz8O\n\n---\n\nhttps://t.co/EwfuqGLTkh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08301158301158296, "ret_signed_m1d": -0.08301158301158296, "alpha_spy_m1d": -0.08391124731497501, "alpha_c_m1d": -0.08390372742680874, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.0008921444152257818, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.061776061776061875, "ret_signed_p1w": -0.061776061776061875, "alpha_spy_p1w": -0.025059297634129063, "alpha_c_p1w": 0.03505812848989498, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.09683419026595685, "ret_p1m": 0.16795366795366795, "ret_signed_p1m": -0.16795366795366795, "alpha_spy_p1m": -0.06911707061119077, "alpha_c_p1m": 0.13154673428530406, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.299500402238972, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0154, 0.0309, 0.029, 0.0154, 0.0174, 0.0425, 0.0444, 0.0444, 0.0328, 0.056, 0.0483, 0.0598, 0.0309, 0.0598, 0.1062, 0.1486, 0.1583, 0.2008, 0.1776, 0.1429, 0.1004, 0.1332, 0.1467, 0.083, 0.0946, 0.1255, 0.166, 0.2104, 0.2104, 0.2104, 0.2104, 0.2104, 0.2104, 0.2104, 0.2104, 0.222, 0.2587, 0.2355, 0.2625, 0.2625, 0.2201, 0.1776, 0.0734, 0.139, 0.0985, 0.0502, 0.1313, 0.1815, 0.1583, 0.1409, 0.112, 0.1351, 0.2471, 0.2201, 0.1892, 0.1042, 0.0753, 0.1139, 0.112, 0.0753, 0.0502, 0.0077, 0.083, 0.0, 0.0, 0.0, 0.0347, 0.0811, 0.0618, 0.0714, 0.139, 0.1622, 0.1236, 0.0946, 0.0618, 0.0463, 0.0849, 0.11, 0.0714, 0.1042, 0.0946, 0.0946, 0.0946, 0.168, 0.168, 0.2838, 0.3668, 0.2683, 0.3089, 0.2008, 0.1564, 0.1622, 0.1776, 0.1429, 0.1564, 0.139, 0.1158, 0.0714, 0.1042, 0.1429, 0.2239, 0.3069, 0.2181, 0.2355, 0.2683, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2048426095297626498", "ticker_idx": 0, "ts": "2026-04-26T15:36:53+00:00", "symbol": "Hyundai Motor", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "They have a Buy rating on Hyundai Motor", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Confirms/endorses Buy rating on Hyundai Motor.", "resolved_tickers": ["005380.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Hyundai Motor 005380.KS Korea", "bench_c": "XLY", "bench_c_reason": "industry='Auto Manufacturers'", "bench_c_industry": "Auto Manufacturers", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@zephyr_z9 Why not? They have a Buy rating on Hyundai Motor. https://t.co/OLWRNhrmdV", "reply_text": null, "parent_text": "@jukan05 do they have buy rating on Hyundai motors or not??", "quoted_text": null, "parent_author": "zephyr_z9", "ret_m1d": -0.020992378749182028, "ret_signed_m1d": -0.020992378749182028, "alpha_spy_m1d": -0.019272535218074927, "alpha_c_m1d": -0.02820560116882931, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.0072132224196472805, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05916028569574716, "ret_signed_p1d": 0.05916028569574716, "alpha_spy_p1d": 0.06402623442389332, "alpha_c_p1d": 0.06620368545350219, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.007043399757755031, "ret_p1w": 0.028625960138495055, "ret_signed_p1w": 0.028625960138495055, "alpha_spy_p1w": 0.02465483869417473, "alpha_c_p1w": 0.029644248673089235, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": -0.0010182885345941806, "ret_p1m": 0.31488544180916067, "ret_signed_p1m": 0.31488544180916067, "alpha_spy_p1m": 0.26535869321551586, "alpha_c_p1m": 0.3012228428581636, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.013662598950997085, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0642, 0.0032, -0.05, -0.0918, -0.0661, -0.0424, -0.0718, -0.1117, -0.0918, -0.087, -0.0329, -0.0386, -0.0519, -0.0519, -0.0519, -0.0519, -0.0253, -0.0329, -0.0063, -0.0044, 0.0868, 0.1622, 0.2863, 0.2863, 0.1355, -0.0439, 0.0458, 0.0553, -0.0324, 0.0019, 0.0115, -0.0057, -0.0134, -0.0344, -0.0038, 0.0401, -0.0038, -0.0134, -0.0744, -0.0611, -0.0439, -0.0649, -0.0553, -0.104, -0.1498, -0.0697, -0.1116, -0.1011, -0.105, -0.0973, -0.0305, -0.0658, -0.0658, -0.0868, -0.062, -0.0305, 0.0191, 0.0267, 0.0057, 0.042, 0.0324, 0.0153, -0.021, 0.0, 0.0592, 0.0611, 0.0134, 0.0134, 0.0286, 0.0286, 0.0496, 0.0916, 0.1698, 0.2328, 0.2328, 0.355, 0.3588, 0.3359, 0.2653, 0.1527, 0.1298, 0.271, 0.25, 0.25, 0.3149, 0.2996, 0.2967, 0.3849, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046786630296674532", "ticker_idx": 0, "ts": "2026-04-22T03:02:14+00:00", "symbol": "power semiconductors", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The era of power semiconductors is arriving. This is only the beginning.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish on power semiconductors sector as NVIDIA's 800V DC push signals explosive, durable growth in AI server power demand.", "resolved_tickers": ["WOLF", "ON", "STM", "IFNNY"], "resolution_method": "basket", "resolution_reason": "Power semicon basket: Wolfspeed, ON Semi, STMicro, Infineon", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "NVIDIA has reportedly visited Korean power equipment companies and asked them to design data center infrastructure around an 800V-class DC architecture, according to Korean media.\n\nThis is being interpreted as an effort to minimize power conversion losses in order to handle the explosive growth in AI server power consumption.\n\nThe era of power semiconductors is arriving. This is only the beginning.\n\n---\n\nhttps://t.co/r43vnpYosT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04424341757890432, "ret_signed_m1d": -0.04424341757890432, "alpha_spy_m1d": -0.03421824264810641, "alpha_c_m1d": -0.018720729717678852, "bench_spy_m1d": -0.010025174930797909, "bench_c_m1d": -0.025522687861225468, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06496681018381692, "ret_signed_p1d": 0.06496681018381692, "alpha_spy_p1d": 0.06884753413364211, "alpha_c_p1d": 0.054438907844165746, "bench_spy_p1d": -0.0038807239498251933, "bench_c_p1d": 0.010527902339651174, "ret_p1w": 0.0803594751122444, "ret_signed_p1w": 0.0803594751122444, "alpha_spy_p1w": 0.07983924183971369, "alpha_c_p1w": 0.03264854953992369, "bench_spy_p1w": 0.0005202332725307013, "bench_c_p1w": 0.047710925572320706, "ret_p1m": 0.6450061922864299, "ret_signed_p1m": 0.6450061922864299, "alpha_spy_p1m": 0.6007014893759297, "alpha_c_p1m": 0.45405759241948784, "bench_spy_p1m": 0.04430470291050015, "bench_c_p1m": 0.19094859986694201, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2902, -0.3012, -0.2825, -0.271, -0.294, -0.3272, -0.3071, -0.3186, -0.306, -0.3156, -0.2967, -0.29, -0.2664, -0.2229, -0.2548, -0.2314, -0.2314, -0.2166, -0.2069, -0.2236, -0.2228, -0.218, -0.1967, -0.1891, -0.1975, -0.2172, -0.2123, -0.2475, -0.2344, -0.2625, -0.3213, -0.2867, -0.2941, -0.2732, -0.3026, -0.3075, -0.2849, -0.2828, -0.3004, -0.3134, -0.3311, -0.3139, -0.3068, -0.2736, -0.305, -0.333, -0.3546, -0.2927, -0.2854, -0.2832, -0.2832, -0.2645, -0.2575, -0.197, -0.1548, -0.1627, -0.1388, -0.1371, -0.1282, -0.0722, -0.0441, -0.0304, -0.0442, 0.0, 0.065, 0.1131, 0.0951, 0.0385, 0.0804, 0.1432, 0.2196, 0.2038, 0.2467, 0.3087, 0.2958, 0.3541, 0.4008, 0.3952, 0.5688, 0.6574, 0.5435, 0.5123, 0.4913, 0.5421, 0.645, 0.6922, 0.6922, 0.7981, 0.6815, 0.7206, 0.6694, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2050043448376660230", "ticker_idx": 0, "ts": "2026-05-01T02:43:40+00:00", "symbol": "DRAM", "kind": "sector", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Who the hell said DRAM prices were going to drop? 😂 On the contrary, they've risen 26% compared to March prices?", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Author pushes back on bearish DRAM price forecasts; data shows prices up 26% MoM, implying bullish DRAM outlook.", "resolved_tickers": ["005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Korea DRAM sector: Samsung Electronics + SK Hynix", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Who the hell said DRAM prices were going to drop? 😂 On the contrary, they've risen 26% compared to March prices? https://t.co/zDrMbJkkNX", "reply_text": null, "parent_text": null, "quoted_text": "Preliminary April Korea semiconductor export figures:\n\n* DRAM: $9.25bn (+15% MoM, +83% QoQ)\n* DRAM modules: $6.14bn (-17% MoM, +70% QoQ)\n* NAND: $1.67bn (-34% MoM, +17% QoQ)\n* MCP: $8.16bn (-5% MoM, +67% QoQ)\n* SSD: $3.84bn (+20% MoM, +181% QoQ) https://t.co/XHNwN0bkIR", "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0027614670118818463, "alpha_c_m1d": 0.01464134916092208, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.01464134916092208, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0898081012290356, "ret_signed_p1d": 0.0898081012290356, "alpha_spy_p1d": 0.09347148089877755, "alpha_c_p1d": 0.0886960487177596, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": 0.001112052511275996, "ret_p1w": 0.2643645323713545, "ret_signed_p1w": 0.2643645323713545, "alpha_spy_p1w": 0.2408163881567994, "alpha_c_p1w": 0.18003754425109564, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.08432698812025885, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.269, -0.2681, -0.3128, -0.3155, -0.2792, -0.2848, -0.2865, -0.2512, -0.2484, -0.2484, -0.2484, -0.2484, -0.2231, -0.2015, -0.1942, -0.1574, -0.1444, -0.0794, -0.0976, -0.0976, -0.1934, -0.2802, -0.2006, -0.2149, -0.2823, -0.2101, -0.1987, -0.2132, -0.2309, -0.1943, -0.1841, -0.1176, -0.1524, -0.1573, -0.2157, -0.1874, -0.1855, -0.2297, -0.2297, -0.2608, -0.3071, -0.2212, -0.2728, -0.2372, -0.2177, -0.1983, -0.121, -0.1494, -0.1336, -0.1399, -0.1029, -0.0799, -0.0577, -0.0716, -0.0603, -0.0275, -0.0313, -0.0146, -0.0272, 0.0114, 0.0088, 0.0152, 0.0, 0.0, 0.0898, 0.0898, 0.2256, 0.2587, 0.2644, 0.3783, 0.3461, 0.4123, 0.4371, 0.3206, 0.3526, 0.3032, 0.3043, 0.4334, 0.4179, 0.4179, 0.4758, 0.5682, 0.5693, 0.626, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2047440842856743388", "ticker_idx": 0, "ts": "2026-04-23T22:21:51+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This might be the first time I've ever agreed with Burry.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Author endorses Burry's long position in Microsoft, agreeing with the bullish thesis.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "quote", "lang": "en", "tweet_text": "* Michael Burry initiated a new long position in Microsoft.\n\nThis might be the first time I’ve ever agreed with Burry.", "reply_text": null, "parent_text": null, "quoted_text": "Why does it feel like people talk so little about GitHub Copilot? I think $MSFT is in a stronger position than Google when it comes to coding.", "parent_author": null, "ret_m1d": 0.04129888707785789, "ret_signed_m1d": 0.04129888707785789, "alpha_spy_m1d": 0.03740304443819631, "alpha_c_m1d": 0.03740304443819631, "bench_spy_m1d": 0.003895842639661584, "bench_c_m1d": 0.003895842639661584, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021334937195510273, "ret_signed_p1d": 0.021334937195510273, "alpha_spy_p1d": 0.013585639235486058, "alpha_c_p1d": 0.013585639235486058, "bench_spy_p1d": 0.007749297960024215, "bench_c_p1d": 0.007749297960024215, "ret_p1w": -0.019170135207435268, "ret_signed_p1w": -0.019170135207435268, "alpha_spy_p1w": -0.03358182376919128, "alpha_c_p1w": -0.03358182376919128, "bench_spy_p1w": 0.014411688561756009, "bench_c_p1w": 0.014411688561756009, "ret_p1m": 0.00896353668088068, "ret_signed_p1m": 0.00896353668088068, "alpha_spy_p1m": -0.04353134905643263, "alpha_c_p1m": -0.04353134905643263, "bench_spy_p1m": 0.05249488573731331, "bench_c_p1m": 0.05249488573731331, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1286, 0.1533, 0.1558, 0.0403, 0.0326, 0.016, -0.0132, -0.006, -0.0553, -0.0373, -0.0074, -0.0082, -0.0296, -0.0357, -0.0369, -0.0369, -0.0476, -0.041, -0.0416, -0.0445, -0.0752, -0.0643, -0.0364, -0.0337, -0.0553, -0.0414, -0.0284, -0.0254, -0.0122, -0.0163, -0.0152, -0.024, -0.0261, -0.0334, -0.0486, -0.038, -0.0393, -0.0576, -0.0643, -0.0815, -0.0788, -0.1035, -0.1075, -0.1197, -0.1419, -0.1366, -0.1096, -0.1116, -0.1017, -0.1017, -0.1031, -0.1045, -0.0996, -0.1027, -0.1079, -0.0755, -0.0545, -0.0109, 0.0108, 0.0169, 0.0056, 0.0202, 0.0413, 0.0, 0.0213, 0.0218, 0.0325, 0.021, -0.0192, -0.0032, -0.0051, -0.0105, -0.0043, 0.0121, -0.0015, -0.0074, -0.0192, -0.0254, -0.0152, 0.0148, 0.0187, 0.004, 0.0128, 0.0102, 0.009, 0.009, 0.0028, -0.0053, 0.0293, 0.0853, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2051149046711210066", "ticker_idx": 0, "ts": "2026-05-04T03:56:55+00:00", "symbol": "USDKRW=X", "kind": "fx", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "환율하방압력 꽤 커보이는데", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Expects downward pressure on USD/KRW (KRW appreciation) as foreign retail investors buy Korean equities.", "resolved_tickers": ["USDKRW=X"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": null, "bench_c_reason": "no_country_fallback (all FX/crypto/index)", "bench_c_industry": null, "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@Alisvolatprop12 환율은 어떻게 될까요? 해외 리테일 이렇게 한국 주식 막 사는거보면....환율하방압력 꽤 커보이는데...", "reply_text": null, "parent_text": "“2%도 쉽지 않다더니, 앞다퉈 상승 버튼”…韓경제 눈높이 높이는 해외IB - 매일경제\n\n이 추세가 안 꺾이면… 아마 하반기엔 하이 싱글을 말하는 곳들이 나타날 겁니다. https://t.co/ztyp9jrp7a", "quoted_text": null, "parent_author": "Alisvolatprop12", "ret_m1d": 0.0015492171302706748, "ret_signed_m1d": -0.0015492171302706748, "alpha_spy_m1d": 0.0021276322346834853, "alpha_c_m1d": null, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": null, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": null, "bench_spy_t0": 0.0, "bench_c_t0": null, "ret_p1d": 0.0029353675040302285, "ret_signed_p1d": -0.0029353675040302285, "alpha_spy_p1d": 0.005086818380962033, "alpha_c_p1d": null, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": null, "ret_p1w": -0.007596455728287932, "ret_signed_p1w": 0.007596455728287932, "alpha_spy_p1w": 0.03724782233568846, "alpha_c_p1w": null, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.017, -0.0079, -0.0007, -0.0063, -0.009, -0.0111, -0.0187, -0.0209, -0.0219, -0.0214, -0.0216, -0.0215, -0.0163, -0.0192, -0.0201, -0.0216, -0.0304, -0.0268, -0.0226, -0.0132, 0.0077, -0.0075, 0.0053, 0.0083, -0.0049, 0.0007, 0.0025, -0.0008, 0.0204, 0.0117, 0.0089, 0.0228, 0.0123, 0.0221, 0.0093, 0.0174, 0.0199, 0.0249, 0.024, 0.0302, 0.0215, 0.0274, 0.0255, 0.0255, 0.0245, 0.0186, 0.004, 0.0011, 0.0071, 0.0035, -0.0016, 0.0016, 0.0042, -0.0045, -0.0017, 0.0096, 0.0042, 0.0056, 0.002, 0.0009, 0.0002, 0.0106, 0.0015, 0.0, 0.0029, 0.0015, -0.0185, -0.0115, -0.0076, 0.0018, 0.0144, 0.0123, 0.0147, 0.0173, 0.014, 0.0245, 0.0191, 0.0222, 0.0279, 0.0293, 0.0229, 0.0213, 0.0241, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046748095254528214", "ticker_idx": 0, "ts": "2026-04-22T00:29:07+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "DRAM", "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung has scrapped its mass production plan for its 10nm-class 7th-generation DRAM (1d DRAM)… management concluded that mass production was not viable… raises the possibility of partial delays in Samsung Electronics' next-generation artificial intelligence (AI) memory roadmap", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung scraps 1d DRAM mass production on yield/ROI failure, threatening HBM5E roadmap; SK Hynix leads with 50%+ D1d yields.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Scraps 7th-Generation DRAM Mass Production Plan… Red Flag for Future HBM Leadership\n\nSamsung Electronics has scrapped its mass production plan for its 10nm-class 7th-generation DRAM (1d DRAM, hereafter D1d). Although the company had initially planned to begin actual D1d production after going through the Production Release Approval (PRA) process in Q1 of this year, it is understood that Samsung did not proceed with mass production after considering the deterioration in Return on Investment (ROI) stemming from the failure to meet yield targets.\n\nAccording to a source familiar with Samsung Electronics' internal affairs on the 22nd, Samsung's management recently reviewed the yield and ROI of D1d and concluded that mass production was not viable. With process difficulty rising exponentially and stable product quality difficult to guarantee, management determined not to force line operations.\n\nD1d is a core product that Samsung plans to apply to HBM5E, the 9th-generation High Bandwidth Memory. While 1c DRAM can handle the 6th through 8th generations (HBM4, HBM4E, HBM5), stable supply of D1d is essential starting from HBM5E.\n\nThis mass production cancellation raises the possibility of partial delays in Samsung Electronics' next-generation artificial intelligence (AI) memory roadmap. Previously, at GTC 2026 which opened on March 16 (local time) in San Jose, California, Samsung Electronics Memory Development VP Hwang Sang-joon stated, \"D1d will be used as the core DRAM for HBM5E.\" The decision contradicts Hwang's presentation at the time, in which he expressed confidence in leveraging leading-edge processes.\n\nGiven the nature of the semiconductor business, a 1% difference in process yield is an indicator that determines trillions of won in operating profit. Samsung Electronics' current D1d appears to lack sufficient customer demand for the raw product to justify recouping the massive investment.\n\nA source familiar with Samsung's internal affairs explained, \"Along with the yield shortfall, management made the cold assessment that there are no customers willing to take or wait for D1d products at the current level.\" The judgment is that forcing mass production could lead to company-wide ROI deterioration.\n\nWith the mass production plan falling through, the approximately 400-person D1d mass production Task Force (TF) assembled for the project has effectively been left idle. The D1d mass production TF has been operating with mass production in mind for the past two years, but with the development product acceptance approval failing to come through, key engineers are reportedly confined to merely exchanging information at the development stage.\n\nInside Samsung Electronics, concerns over the DRAM technology gap with rival SK Hynix have reignited. SK Hynix is understood to have already completed D1d development and secured stable mass production yields of over 50%. Industry observers note that Samsung's strategy last year of simultaneously developing 6th-generation (1c) and 7th-generation (1d) DRAM dispersed its capabilities, leaving it at risk of falling behind rivals that relatively concentrated their firepower on a single generation.\n\nA source familiar with Samsung's internal affairs said, \"Samsung Electronics plans to indefinitely postpone the mass production schedule until D1d yields reach target levels, and as of now has not confirmed a timeline for resumption,\" adding, \"The company is comprehensively re-examining its process roadmap and focusing on further boosting yields.\"\n\n---\n\nhttps://t.co/ejUhyu0bAu", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.006896551724137945, "ret_signed_m1d": -0.006896551724137945, "alpha_spy_m1d": -0.016921726654935854, "alpha_c_m1d": -0.028403250283466575, "bench_spy_m1d": -0.010025174930797909, "bench_c_m1d": -0.02150669855932863, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03218390804597693, "ret_signed_p1d": -0.03218390804597693, "alpha_spy_p1d": -0.03606463199580212, "alpha_c_p1d": -0.04641630764190707, "bench_spy_p1d": -0.0038807239498251933, "bench_c_p1d": -0.014232399595930145, "ret_p1w": 0.03908045977011487, "ret_signed_p1w": -0.03908045977011487, "alpha_spy_p1w": -0.03856022649758417, "alpha_c_p1w": -0.032628411594461504, "bench_spy_p1w": 0.0005202332725307013, "bench_c_p1w": 0.006452048175653369, "ret_p1m": 0.3770114942528735, "ret_signed_p1m": -0.3770114942528735, "alpha_spy_p1m": -0.33270679134237335, "alpha_c_p1m": -0.24727520327473762, "bench_spy_p1m": 0.04430470291050015, "bench_c_p1m": 0.12973629097813588, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3021, -0.3021, -0.2682, -0.2549, -0.2627, -0.2636, -0.3099, -0.2315, -0.2241, -0.2691, -0.2723, -0.2365, -0.2393, -0.2301, -0.1805, -0.1686, -0.1686, -0.1686, -0.1686, -0.1282, -0.1278, -0.1145, -0.0824, -0.0663, 0.0002, -0.0067, -0.0067, -0.1048, -0.2099, -0.1209, -0.1365, -0.2039, -0.1379, -0.1282, -0.1379, -0.1581, -0.1342, -0.1103, -0.0434, -0.0801, -0.0851, -0.1452, -0.1296, -0.1328, -0.1737, -0.1737, -0.1894, -0.2313, -0.128, -0.1798, -0.1439, -0.1122, -0.0966, -0.0322, -0.0621, -0.0529, -0.0759, -0.0506, -0.0299, 0.0, -0.0069, -0.0138, 0.0069, 0.0, 0.0322, 0.0092, 0.0322, 0.0207, 0.0391, 0.0138, 0.0138, 0.069, 0.069, 0.223, 0.2483, 0.2345, 0.3126, 0.2828, 0.3057, 0.3609, 0.2437, 0.292, 0.2667, 0.269, 0.377, 0.3448, 0.3448, 0.3747, 0.4115, 0.377, 0.4575, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046748095254528214", "ticker_idx": 1, "ts": "2026-04-22T00:29:07+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is understood to have already completed D1d development and secured stable mass production yields of over 50%", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung scraps 1d DRAM mass production on yield/ROI failure, threatening HBM5E roadmap; SK Hynix leads with 50%+ D1d yields.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Scraps 7th-Generation DRAM Mass Production Plan… Red Flag for Future HBM Leadership\n\nSamsung Electronics has scrapped its mass production plan for its 10nm-class 7th-generation DRAM (1d DRAM, hereafter D1d). Although the company had initially planned to begin actual D1d production after going through the Production Release Approval (PRA) process in Q1 of this year, it is understood that Samsung did not proceed with mass production after considering the deterioration in Return on Investment (ROI) stemming from the failure to meet yield targets.\n\nAccording to a source familiar with Samsung Electronics' internal affairs on the 22nd, Samsung's management recently reviewed the yield and ROI of D1d and concluded that mass production was not viable. With process difficulty rising exponentially and stable product quality difficult to guarantee, management determined not to force line operations.\n\nD1d is a core product that Samsung plans to apply to HBM5E, the 9th-generation High Bandwidth Memory. While 1c DRAM can handle the 6th through 8th generations (HBM4, HBM4E, HBM5), stable supply of D1d is essential starting from HBM5E.\n\nThis mass production cancellation raises the possibility of partial delays in Samsung Electronics' next-generation artificial intelligence (AI) memory roadmap. Previously, at GTC 2026 which opened on March 16 (local time) in San Jose, California, Samsung Electronics Memory Development VP Hwang Sang-joon stated, \"D1d will be used as the core DRAM for HBM5E.\" The decision contradicts Hwang's presentation at the time, in which he expressed confidence in leveraging leading-edge processes.\n\nGiven the nature of the semiconductor business, a 1% difference in process yield is an indicator that determines trillions of won in operating profit. Samsung Electronics' current D1d appears to lack sufficient customer demand for the raw product to justify recouping the massive investment.\n\nA source familiar with Samsung's internal affairs explained, \"Along with the yield shortfall, management made the cold assessment that there are no customers willing to take or wait for D1d products at the current level.\" The judgment is that forcing mass production could lead to company-wide ROI deterioration.\n\nWith the mass production plan falling through, the approximately 400-person D1d mass production Task Force (TF) assembled for the project has effectively been left idle. The D1d mass production TF has been operating with mass production in mind for the past two years, but with the development product acceptance approval failing to come through, key engineers are reportedly confined to merely exchanging information at the development stage.\n\nInside Samsung Electronics, concerns over the DRAM technology gap with rival SK Hynix have reignited. SK Hynix is understood to have already completed D1d development and secured stable mass production yields of over 50%. Industry observers note that Samsung's strategy last year of simultaneously developing 6th-generation (1c) and 7th-generation (1d) DRAM dispersed its capabilities, leaving it at risk of falling behind rivals that relatively concentrated their firepower on a single generation.\n\nA source familiar with Samsung's internal affairs said, \"Samsung Electronics plans to indefinitely postpone the mass production schedule until D1d yields reach target levels, and as of now has not confirmed a timeline for resumption,\" adding, \"The company is comprehensively re-examining its process roadmap and focusing on further boosting yields.\"\n\n---\n\nhttps://t.co/ejUhyu0bAu", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0008176960088583929, "ret_signed_m1d": 0.0008176960088583929, "alpha_spy_m1d": 0.010842870939656302, "alpha_c_m1d": 0.02634038387008386, "bench_spy_m1d": -0.010025174930797909, "bench_c_m1d": -0.025522687861225468, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0016353920177167858, "ret_signed_p1d": 0.0016353920177167858, "alpha_spy_p1d": 0.005516115967541979, "alpha_c_p1d": -0.008892510321934388, "bench_spy_p1d": -0.0038807239498251933, "bench_c_p1d": 0.010527902339651174, "ret_p1w": 0.057236369450165636, "ret_signed_p1w": 0.057236369450165636, "alpha_spy_p1w": 0.056716136177634935, "alpha_c_p1w": 0.00952544387784493, "bench_spy_p1w": 0.0005202332725307013, "bench_c_p1w": 0.047710925572320706, "ret_p1m": 0.5862632999548985, "ret_signed_p1m": 0.5862632999548985, "alpha_spy_p1m": 0.5419585970443983, "alpha_c_p1m": 0.39531470008795644, "bench_spy_p1m": 0.04430470291050015, "bench_c_p1m": 0.19094859986694201, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.374, -0.3993, -0.3471, -0.3136, -0.2973, -0.2581, -0.3226, -0.2597, -0.2655, -0.3128, -0.3152, -0.2761, -0.285, -0.2981, -0.2753, -0.2818, -0.2818, -0.2818, -0.2818, -0.2704, -0.2255, -0.2238, -0.1798, -0.1692, -0.1014, -0.1325, -0.1325, -0.2322, -0.3058, -0.2306, -0.2445, -0.3164, -0.233, -0.2191, -0.2396, -0.2559, -0.2036, -0.2069, -0.1365, -0.1717, -0.1766, -0.2371, -0.1938, -0.1864, -0.2371, -0.2371, -0.2862, -0.3401, -0.2666, -0.3213, -0.2837, -0.2756, -0.251, -0.1554, -0.184, -0.1603, -0.1496, -0.0981, -0.0711, -0.0556, -0.0777, -0.0466, 0.0008, 0.0, 0.0016, -0.0008, 0.0564, 0.063, 0.0572, 0.0515, 0.0515, 0.1832, 0.1832, 0.3091, 0.3524, 0.3786, 0.5372, 0.5004, 0.6157, 0.6108, 0.4873, 0.5045, 0.4268, 0.4268, 0.5863, 0.5871, 0.5871, 0.6778, 0.834, 0.8719, 0.9079, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2056238134481076408", "ticker_idx": 0, "ts": "2026-05-18T04:59:08+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Reiterate Buy rating; target price modestly raised to $308", "tweet_type": "news_reaction", "timeline": "months", "summary": "GF reiterates Buy on NVDA, raises PT to $308; above-consensus revenue forecasts of $80.6B F1Q and $91–93B F2Q guidance on Blackwell ramp + LPX contribution.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications\n NVIDIA (NVDA Buy): Earnings Preview — Product Mix Offsets the Impact of Rubin Delay\n\nReiterate Buy rating; target price modestly raised to $308:\nDriven by rising market expectations ahead of the upcoming earnings call and the U.S. approval on May 14 to export H200 chips to 10 Chinese companies, NVIDIA’s share price has reached a new all-time high. For the May 20 earnings call, we expect results to be in line with expectations, with guidance slightly above expectations. Given the company’s sizeable cash position and free cash flow, the announcement of a new share repurchase program during the earnings call would be a reasonable expectation.\n\nIn addition, Rubin’s timeline remains a key market focus. We reiterate our view that mass production will be delayed by about one month to September, while the 2300W specification remains unchanged.\nBeyond GPUs, with Vera CPU and LPX, the company is expected to capture a larger share of value within the data center silicon TAM. For product re-rating, we still believe NVIDIA needs to launch an inference-focused GPU, potentially Feynman.\n\nTaking into account higher Blackwell shipments and ASPs, lower Rubin shipments, and LPX contribution, we adjust our FY27E/28E EPS forecasts by +0%/+13%, respectively, and modestly raise our target price from $292 to $308, based on 28x FY27E/28E P/E.\n\nEarnings preview — expected to be moderately positive:\nDriven by normal GB300 NVL72 production, approximately 10% QoQ growth in Blackwell shipments, and a small contribution from RTX6000, we now expect F1Q revenue of $80.6 billion, compared with Bloomberg consensus of $78.8 billion and buy-side expectations of $80.0 billion.\n\nAs Blackwell shipments continue to increase, we expect F2Q guidance of $91.0 billion, or actual revenue of $93.0 billion, compared with Bloomberg consensus of $86.2 billion and buy-side expectations of around $90.0 billion. Our forecast does not include any Rubin contribution, as we believe it has already been delayed to mass production in September. Given the company’s sizeable cash position and free cash flow, its capital return plan is also worth watching.\n\nBlackwell, Vera, and LPX will offset the widely known impact of Rubin’s delay:\nAccording to our monthly report published on the 12th, due to earlier heatsink design changes, we expect Rubin’s timeline to be: QS in July, MP in September, and rack mass production in October. In terms of performance, we believe the redesigned version will still maintain the 2300W specification this year.\nOn the other hand, LPX rack ramp-up is faster than expected. We still expect 16,000 LPX racks to ship from 4Q26 through 2027, contributing approximately $70 billion in revenue during this period.\n\nIn addition, we believe NVIDIA may have shifted N3 wafer capacity from Rubin to Vera CPU in order to capture the Agentic AI trend. Overall, with deeper software–chip integration, the company is likely to capture a larger share of the data center silicon TAM.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013494062167895615, "ret_signed_m1d": 0.013494062167895615, "alpha_spy_m1d": 0.01279013137671181, "alpha_c_m1d": -0.005145263691752389, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.018639325859648004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00769164563481417, "ret_signed_p1d": -0.00769164563481417, "alpha_spy_p1d": -0.0010307862525710254, "alpha_c_p1d": -0.003663611212575746, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.0040280344222384246, "ret_p1w": -0.03144118955956243, "ret_signed_p1w": -0.03144118955956243, "alpha_spy_p1w": -0.040904385935843446, "alpha_c_p1w": -0.08666316784925698, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.05522197828969455, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1545, -0.1549, -0.1462, -0.1385, -0.1326, -0.1204, -0.1684, -0.203, -0.1792, -0.1902, -0.1767, -0.1754, -0.2002, -0.1785, -0.1689, -0.1632, -0.1762, -0.1892, -0.1759, -0.1817, -0.1886, -0.1968, -0.2232, -0.21, -0.2119, -0.1963, -0.2298, -0.2465, -0.2571, -0.2155, -0.2095, -0.2021, -0.2021, -0.201, -0.1989, -0.181, -0.1728, -0.1515, -0.1485, -0.1161, -0.1055, -0.1078, -0.0928, -0.0911, -0.1009, -0.0892, -0.102, -0.0632, -0.0257, -0.0412, -0.0588, -0.1023, -0.1074, -0.1072, -0.1161, -0.0652, -0.0487, -0.032, -0.013, -0.0069, 0.0158, 0.0604, 0.0135, 0.0, -0.0077, 0.0052, -0.0126, -0.0314, -0.0314, -0.0336, -0.0437, -0.0363, -0.0503, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047974859125465105", "ticker_idx": 0, "ts": "2026-04-25T09:43:50+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Amazon is still trading near a 10-year low on CY26 EV/EBITDA. But as the rivalry between the two AI labs escalates and Amazon begins to collect its 'toll-gate revenue,' we believe the market is likely to gradually move toward a re-rating.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Long AMZN: shares Mirae Asset thesis that market will re-rate Amazon as AI toll-gate beneficiary collecting cloud fees + silicon adoption from OpenAI/Anthropic rivalry, currently near 10-yr low EV/EBITDA.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "The most interesting comment I read in today’s report:\n\nAll of Anthropic's unusual moves around the GPT-5.5 launch ultimately converge on a single conclusion: in order to secure compute, Anthropic must bind itself far more deeply — and far more dependently — to those who possess these physical resources. This conclusion was crystallized in the $100B deal formalized a few days ago. The most representative example is the deal with Amazon. The key terms are as follows:\n\n- Amazon will inject an additional $5B into Anthropic immediately, with up to $20B more in follow-on investment to come. Combined with the existing $8B, cumulative investment reaches up to $33B.\n- Up to 5GW of compute capacity secured, of which roughly 1GW is scheduled to come online by year-end 2026.\n- Anthropic has committed to spending $100B+ on AWS technology over the next 10 years (including usage requirements for AWS silicon such as Trainium and Graviton).\n- Purchase rights secured for Trainium2/3/4 and future Trainium generations. Trainium2 is currently sold out, and Trainium3 is largely booked as well.\n- The Claude Platform will be offered natively on AWS.\n\nJust how large is the 5GW Anthropic has committed to? It is equivalent to roughly five nuclear power plants. Given that Microsoft's total global data center footprint in 2024 is estimated at around 5–6GW, this means Anthropic alone is locking in incremental capacity — for AI training and serving — that rivals the entirety of MSFT's historical physical infrastructure. Coupled with Anthropic's announcement of having reached $30B ARR, the market is reading this deal as \"Anthropic pre-signing Amazon's invoice in order to keep its growth going.\"\n\nIt is also worth noting Dario Amodei's (Anthropic CEO) remarks accompanying the announcement: \"Users tell us that Claude is becoming increasingly essential to how they work. We need to build infrastructure to keep up with rapidly growing demand.\" Last month's blog post from Anthropic — directly admitting that \"surging enterprise/developer demand is placing unavoidable strain on our infrastructure, affecting reliability and performance\" — was effectively a teaser for this deal.\n\nWhat matters is that the structure of this deal is more favorable to Amazon than to Anthropic. Amazon has already invested in OpenAI as well (up to $50B). In other words, the more fiercely OpenAI and Anthropic compete to eat each other's lunch, the more Amazon benefits simultaneously along three axes: cloud usage fees from both, adoption rates for its in-house silicon (Trainium as XPU, Graviton as CPU), and visibility on the recovery of data center CAPEX. This is structurally almost identical to the valuation premium Google historically enjoyed under the \"full-stack player\" framing.\n\nThe market, however, has not yet fully come around to recognizing Amazon in this light. (Sentiment has admittedly improved compared to two weeks ago.) Amazon is still trading near a 10-year low on CY26 EV/EBITDA. But as the rivalry between the two AI labs escalates and Amazon begins to collect its \"toll-gate revenue,\" we believe the market is likely to gradually move toward a re-rating.\n\n(Excerpt from Mirae Asset Securities’ AI Hot Issue report, dated April 24, 2026)\n\n$AMZN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01099109670211007, "ret_signed_m1d": 0.01099109670211007, "alpha_spy_m1d": 0.012710940233217172, "alpha_c_m1d": 0.0037778742824627898, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.0072132224196472805, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005438047398549362, "ret_signed_p1d": -0.005438047398549362, "alpha_spy_p1d": -0.0005720986704031983, "alpha_c_p1d": 0.0016053523592056695, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.007043399757755031, "ret_p1w": 0.04185812224328522, "ret_signed_p1w": 0.04185812224328522, "alpha_spy_p1w": 0.0378870007989649, "alpha_c_p1w": 0.0428764107778794, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": -0.0010182885345941806, "ret_p1m": 0.015969720839887902, "ret_signed_p1m": 0.015969720839887902, "alpha_spy_p1m": -0.03355702775375691, "alpha_c_p1m": 0.0023071218888908174, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.013662598950997085, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0694, -0.0743, -0.0836, -0.0695, -0.0862, -0.1077, -0.1472, -0.1945, -0.2007, -0.2074, -0.2184, -0.2356, -0.2387, -0.2387, -0.2297, -0.2157, -0.2155, -0.1954, -0.2139, -0.2013, -0.1933, -0.2037, -0.1958, -0.2019, -0.2006, -0.1697, -0.1615, -0.1835, -0.1824, -0.1792, -0.1856, -0.1976, -0.2047, -0.1891, -0.1759, -0.1963, -0.2005, -0.2135, -0.1952, -0.2063, -0.1892, -0.2052, -0.2366, -0.2304, -0.2024, -0.1936, -0.1967, -0.1967, -0.1851, -0.1813, -0.1527, -0.1052, -0.0871, -0.0813, -0.0463, -0.0483, -0.0437, -0.0404, -0.0492, -0.0429, -0.0221, -0.0231, 0.011, 0.0, -0.0054, 0.0074, 0.0151, 0.0273, 0.0419, 0.0476, 0.0531, 0.0385, 0.0443, 0.0301, 0.018, 0.0345, 0.0234, 0.0116, 0.0143, -0.0068, 0.0149, 0.0281, 0.0199, 0.0199, 0.016, 0.0411, 0.0493, 0.0365, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2046920825694277828", "ticker_idx": 0, "ts": "2026-04-22T11:55:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Morgan Stanley's latest SK hynix operating profit forecasts... KRW 222 trillion → KRW 277 trillion... KRW 298 trillion → KRW 408 trillion... KRW 226 trillion → KRW 403 trillion", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares Morgan Stanley's raised SK Hynix operating profit forecasts for 2026–2028, implying a bullish stance.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Morgan Stanley’s latest SK hynix operating profit forecasts (as of April 19, 2026):\n\n- Previous forecast for 2026: KRW 222 trillion (US$150.177 billion) → KRW 277 trillion (US$187.382 billion)\n- Previous forecast for 2027: KRW 298 trillion (US$201.588 billion) → KRW 408 trillion (US$276.0 billion)\n- Previous forecast for 2028: KRW 226 trillion (US$152.882 billion) → KRW 403 trillion (US$272.618 billion)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0008176960088583929, "ret_signed_m1d": 0.0008176960088583929, "alpha_spy_m1d": 0.010842870939656302, "alpha_c_m1d": 0.02634038387008386, "bench_spy_m1d": -0.010025174930797909, "bench_c_m1d": -0.025522687861225468, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0016353920177167858, "ret_signed_p1d": 0.0016353920177167858, "alpha_spy_p1d": 0.005516115967541979, "alpha_c_p1d": -0.008892510321934388, "bench_spy_p1d": -0.0038807239498251933, "bench_c_p1d": 0.010527902339651174, "ret_p1w": 0.057236369450165636, "ret_signed_p1w": 0.057236369450165636, "alpha_spy_p1w": 0.056716136177634935, "alpha_c_p1w": 0.00952544387784493, "bench_spy_p1w": 0.0005202332725307013, "bench_c_p1w": 0.047710925572320706, "ret_p1m": 0.5862632999548985, "ret_signed_p1m": 0.5862632999548985, "alpha_spy_p1m": 0.5419585970443983, "alpha_c_p1m": 0.39531470008795644, "bench_spy_p1m": 0.04430470291050015, "bench_c_p1m": 0.19094859986694201, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.374, -0.3993, -0.3471, -0.3136, -0.2973, -0.2581, -0.3226, -0.2597, -0.2655, -0.3128, -0.3152, -0.2761, -0.285, -0.2981, -0.2753, -0.2818, -0.2818, -0.2818, -0.2818, -0.2704, -0.2255, -0.2238, -0.1798, -0.1692, -0.1014, -0.1325, -0.1325, -0.2322, -0.3058, -0.2306, -0.2445, -0.3164, -0.233, -0.2191, -0.2396, -0.2559, -0.2036, -0.2069, -0.1365, -0.1717, -0.1766, -0.2371, -0.1938, -0.1864, -0.2371, -0.2371, -0.2862, -0.3401, -0.2666, -0.3213, -0.2837, -0.2756, -0.251, -0.1554, -0.184, -0.1603, -0.1496, -0.0981, -0.0711, -0.0556, -0.0777, -0.0466, 0.0008, 0.0, 0.0016, -0.0008, 0.0564, 0.063, 0.0572, 0.0515, 0.0515, 0.1832, 0.1832, 0.3091, 0.3524, 0.3786, 0.5372, 0.5004, 0.6157, 0.6108, 0.4873, 0.5045, 0.4268, 0.4268, 0.5863, 0.5871, 0.5871, 0.6778, 0.834, 0.8719, 0.9079, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042022163734208700", "ticker_idx": 0, "ts": "2026-04-08T23:29:57+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory companies' margins will stabilize at levels structurally higher than in the past", "tweet_type": "analysis", "timeline": "years", "summary": "Samsung and SK Hynix shift to 3–5yr LTAs with big tech, structurally raising margins and reducing cyclical downside risk.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "One-year contracts are a thing of the past… Samsung and SK hynix won't sell without 3–5 year long-term supply agreements\n\nSamsung Electronics and SK hynix have effectively abandoned the practice of signing one-year short-term memory supply contracts with global big tech companies, opting instead to supply products exclusively through 3–5 year long-term agreements (LTAs). The strategy aims to build a stable, high-margin business model through strategic collaboration with customers from the earliest stages of AI memory development.\n\nAccording to industry sources on the 8th, Samsung Electronics has established a policy of applying a minimum three-year LTA to all new contracts starting this year. Until last year, the company had been accommodating even quarterly ultra-short-term contracts, but has now decided to fully transition its supply policy to an LTA framework. At the annual general meeting on March 18, Samsung Electronics DS Division CEO (Vice Chairman) Jeon Young-hyun stated: \"We are pursuing a shift in supply contracts with our customers from the current annual and quarterly basis to multi-year agreements spanning 3–5 years.\" Accordingly, Samsung is expected to secure stable three-year memory supply commitments with major customers including AMD — with which negotiations are in the final stages — as well as Microsoft and Google.\n\nFurthermore, SK hynix has reportedly been pursuing a five-year commodity DRAM long-term supply agreement with Google. While a three-year term was initially considered the leading option, an internal assessment that \"even three years is too short\" led to a decision to extend the contract period to five years to maximize supply stability. Given that SK hynix is currently the primary supplier of 5th-generation High Bandwidth Memory (HBM3E) to Google, the company is said to be negotiating an additional two-year extension by leveraging next-generation HBM supply as a condition. Negotiations that were originally slated for completion by year-end are now expected to be finalized within the first half.\n\nHistorically, the memory market was built on commoditization. Commodity DRAM for PCs and smartphones was mass-produced and sold based on quarterly fixed transaction prices or daily-fluctuating spot prices. As a result, even slight demand fluctuations triggered price collapses, leading to trillion-won-scale losses — the inescapable \"curse of the cycle.\" Suppliers were unilaterally exposed to market conditions in a classic subordinate position.\n\nBy contrast, a shift to an LTA framework minimizes the risk of order cliffs. By locking in volumes for three to five-plus years, companies can defend earnings with pre-committed prices and volumes even amid demand shocks from economic downturns — effectively securing downside rigidity where profits no longer plunge during downturns. Samsung Electronics is reportedly reviewing supply pricing policies to hedge against DRAM price volatility alongside minimum annual supply volume provisions as part of the LTA transition.\n\nThe shift is also effective from a capital expenditure efficiency standpoint. With volumes already committed, disciplined and calculated CAPEX becomes possible, replacing reckless capacity expansion races. The semiconductor industry expects that as the share of long-term contracts rises, memory companies' margins will stabilize at levels structurally higher than in the past. In particular, for products like HBM — where process complexity is high and customer requirements vary — a made-to-order supply system based on \"secure orders first, produce after,\" akin to the foundry model, is essential to eliminate inventory risk at the source and maximize production efficiency.\n\nAn industry official commented: \"The era of memory companies anxiously stockpiling inventory while worrying about price declines is over. Samsung Electronics and SK hynix have now been elevated beyond mere memory suppliers to strategic infrastructure partners for global big tech.\"\n\n---\n\nhttps://t.co/u9i3jEUT28", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0665083135391924, "ret_signed_m1d": -0.0665083135391924, "alpha_spy_m1d": -0.041671345417490846, "alpha_c_m1d": -0.036442608886624095, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.030065704652568304, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.030878859857482177, "ret_signed_p1d": -0.030878859857482177, "alpha_spy_p1d": -0.03664795102857055, "alpha_c_p1d": -0.03356080548712981, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.0026819456296476307, "ret_p1w": 0.002375296912114022, "ret_signed_p1w": 0.002375296912114022, "alpha_spy_p1w": -0.03302357637391595, "alpha_c_p1w": -0.05839116834299163, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.06076646525510565, "ret_p1m": 0.2897862232779098, "ret_signed_p1m": 0.2897862232779098, "alpha_spy_p1m": 0.2075832876387096, "alpha_c_p1m": 0.09217171613172659, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.19761450714618323, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.341, -0.342, -0.3477, -0.3349, -0.3178, -0.2941, -0.2922, -0.3116, -0.2913, -0.278, -0.2789, -0.2789, -0.2438, -0.2301, -0.2382, -0.2391, -0.287, -0.2059, -0.1983, -0.2448, -0.2481, -0.2111, -0.214, -0.2045, -0.1533, -0.141, -0.141, -0.141, -0.141, -0.0993, -0.0988, -0.085, -0.0518, -0.0353, 0.0335, 0.0264, 0.0264, -0.0751, -0.1836, -0.0917, -0.1078, -0.1775, -0.1092, -0.0993, -0.1092, -0.1301, -0.1054, -0.0808, -0.0115, -0.0495, -0.0547, -0.1168, -0.1007, -0.104, -0.1462, -0.1462, -0.1625, -0.2057, -0.099, -0.1525, -0.1154, -0.0827, -0.0665, 0.0, -0.0309, -0.0214, -0.0451, -0.019, 0.0024, 0.0333, 0.0261, 0.019, 0.0404, 0.0333, 0.0665, 0.0428, 0.0665, 0.0546, 0.0736, 0.0475, 0.0475, 0.1045, 0.1045, 0.2637, 0.2898, 0.2755, 0.3563, 0.3254, 0.3492, 0.4062, 0.285, 0.3349, 0.3088, 0.3112, 0.4228, 0.3895, 0.3895, 0.4204, 0.4584, 0.4228, 0.5059, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042022163734208700", "ticker_idx": 1, "ts": "2026-04-08T23:29:57+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory companies' margins will stabilize at levels structurally higher than in the past", "tweet_type": "analysis", "timeline": "years", "summary": "Samsung and SK Hynix shift to 3–5yr LTAs with big tech, structurally raising margins and reducing cyclical downside risk.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "One-year contracts are a thing of the past… Samsung and SK hynix won't sell without 3–5 year long-term supply agreements\n\nSamsung Electronics and SK hynix have effectively abandoned the practice of signing one-year short-term memory supply contracts with global big tech companies, opting instead to supply products exclusively through 3–5 year long-term agreements (LTAs). The strategy aims to build a stable, high-margin business model through strategic collaboration with customers from the earliest stages of AI memory development.\n\nAccording to industry sources on the 8th, Samsung Electronics has established a policy of applying a minimum three-year LTA to all new contracts starting this year. Until last year, the company had been accommodating even quarterly ultra-short-term contracts, but has now decided to fully transition its supply policy to an LTA framework. At the annual general meeting on March 18, Samsung Electronics DS Division CEO (Vice Chairman) Jeon Young-hyun stated: \"We are pursuing a shift in supply contracts with our customers from the current annual and quarterly basis to multi-year agreements spanning 3–5 years.\" Accordingly, Samsung is expected to secure stable three-year memory supply commitments with major customers including AMD — with which negotiations are in the final stages — as well as Microsoft and Google.\n\nFurthermore, SK hynix has reportedly been pursuing a five-year commodity DRAM long-term supply agreement with Google. While a three-year term was initially considered the leading option, an internal assessment that \"even three years is too short\" led to a decision to extend the contract period to five years to maximize supply stability. Given that SK hynix is currently the primary supplier of 5th-generation High Bandwidth Memory (HBM3E) to Google, the company is said to be negotiating an additional two-year extension by leveraging next-generation HBM supply as a condition. Negotiations that were originally slated for completion by year-end are now expected to be finalized within the first half.\n\nHistorically, the memory market was built on commoditization. Commodity DRAM for PCs and smartphones was mass-produced and sold based on quarterly fixed transaction prices or daily-fluctuating spot prices. As a result, even slight demand fluctuations triggered price collapses, leading to trillion-won-scale losses — the inescapable \"curse of the cycle.\" Suppliers were unilaterally exposed to market conditions in a classic subordinate position.\n\nBy contrast, a shift to an LTA framework minimizes the risk of order cliffs. By locking in volumes for three to five-plus years, companies can defend earnings with pre-committed prices and volumes even amid demand shocks from economic downturns — effectively securing downside rigidity where profits no longer plunge during downturns. Samsung Electronics is reportedly reviewing supply pricing policies to hedge against DRAM price volatility alongside minimum annual supply volume provisions as part of the LTA transition.\n\nThe shift is also effective from a capital expenditure efficiency standpoint. With volumes already committed, disciplined and calculated CAPEX becomes possible, replacing reckless capacity expansion races. The semiconductor industry expects that as the share of long-term contracts rises, memory companies' margins will stabilize at levels structurally higher than in the past. In particular, for products like HBM — where process complexity is high and customer requirements vary — a made-to-order supply system based on \"secure orders first, produce after,\" akin to the foundry model, is essential to eliminate inventory risk at the source and maximize production efficiency.\n\nAn industry official commented: \"The era of memory companies anxiously stockpiling inventory while worrying about price declines is over. Samsung Electronics and SK hynix have now been elevated beyond mere memory suppliers to strategic infrastructure partners for global big tech.\"\n\n---\n\nhttps://t.co/u9i3jEUT28", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.11326233929353025, "ret_signed_m1d": -0.11326233929353025, "alpha_spy_m1d": -0.08842537117182869, "alpha_c_m1d": -0.05883119675980497, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.054431142533725274, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03388193464335787, "ret_signed_p1d": -0.03388193464335787, "alpha_spy_p1d": -0.03965102581444624, "alpha_c_p1d": -0.05135565046721813, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.01747371582386026, "ret_p1w": 0.09970961384699062, "ret_signed_p1w": 0.09970961384699062, "alpha_spy_p1w": 0.06431074056096064, "alpha_c_p1w": 0.02858508529023207, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.07112452855675855, "ret_p1m": 0.6011616656390466, "ret_signed_p1m": 0.6011616656390466, "alpha_spy_p1m": 0.5189587299998464, "alpha_c_p1m": 0.3240880856765651, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.2770735799624815, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2811, -0.2763, -0.2869, -0.283, -0.2763, -0.2695, -0.2618, -0.2821, -0.285, -0.2705, -0.2589, -0.2888, -0.227, -0.1874, -0.168, -0.1217, -0.198, -0.1236, -0.1304, -0.1864, -0.1893, -0.1429, -0.1535, -0.169, -0.142, -0.1497, -0.1497, -0.1497, -0.1497, -0.1362, -0.083, -0.0811, -0.0289, -0.0163, 0.0639, 0.0271, 0.0271, -0.091, -0.1781, -0.0891, -0.1055, -0.1907, -0.092, -0.0755, -0.0997, -0.1191, -0.0571, -0.061, 0.0223, -0.0194, -0.0252, -0.0968, -0.0455, -0.0368, -0.0968, -0.0968, -0.1549, -0.2188, -0.1317, -0.1965, -0.152, -0.1423, -0.1133, 0.0, -0.0339, -0.0058, 0.0068, 0.0678, 0.0997, 0.1181, 0.092, 0.1288, 0.1849, 0.1839, 0.1859, 0.183, 0.2507, 0.2585, 0.2517, 0.2449, 0.2449, 0.4008, 0.4008, 0.5499, 0.6012, 0.6321, 0.8199, 0.7764, 0.9129, 0.9071, 0.7609, 0.7812, 0.6893, 0.6893, 0.878, 0.879, 0.879, 0.9864, 1.1713, 1.2162, 1.2588, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2048051254786687169", "ticker_idx": 0, "ts": "2026-04-25T14:47:24+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "People who think the DRAM supercycle will end in 1H next year don't know what they're talking about", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supercycle extends well beyond 1H next year; NVIDIA alone to absorb 72% of total LPDDR supply driven by GB200/VR200 LPDDR5X demand.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Hana Securities’ bull case estimates that next year, NVIDIA alone will account for 72% of total LPDDR supply.\n\nDid you hear me, anon? https://t.co/a9MdGHZeXT", "reply_text": null, "parent_text": null, "quoted_text": "People who think the DRAM supercycle will end in 1H next year don't know what they're talking about.\n\nThe GB200 uses 17TB of SOCAMM2, but the VR200 uses over 50TB of SOCAMM2. That alone has NVIDIA hoovering up so much LPDDR5X that even smartphone OEMs are panic-buying — and do", "parent_author": null, "ret_m1d": -0.04317476330084754, "ret_signed_m1d": -0.04317476330084754, "alpha_spy_m1d": -0.04145491976974044, "alpha_c_m1d": -0.0435302973590573, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01452860823398446, "ret_signed_p1d": -0.01452860823398446, "alpha_spy_p1d": -0.009662659505838297, "alpha_c_p1d": 0.015199235206611076, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.08484161420827219, "ret_signed_p1w": 0.08484161420827219, "alpha_spy_p1w": 0.08087049276395186, "alpha_c_p1w": 0.08379472373880326, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.5426510883260374, "ret_signed_p1m": 0.5426510883260374, "alpha_spy_p1m": 0.49312433973239256, "alpha_c_p1m": 0.3532622307561879, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2663, -0.2634, -0.2646, -0.2853, -0.2518, -0.2767, -0.3039, -0.2982, -0.2814, -0.2917, -0.2693, -0.2437, -0.2434, -0.2434, -0.251, -0.2375, -0.2232, -0.202, -0.2017, -0.1793, -0.1638, -0.1295, -0.1435, -0.1434, -0.2275, -0.2712, -0.2211, -0.2475, -0.2799, -0.2235, -0.2061, -0.2242, -0.2227, -0.1885, -0.1692, -0.1253, -0.1594, -0.1761, -0.2264, -0.2133, -0.2205, -0.2666, -0.2655, -0.3085, -0.3289, -0.2532, -0.2882, -0.2648, -0.2447, -0.232, -0.1625, -0.1718, -0.1619, -0.1622, -0.1129, -0.1037, -0.0885, -0.0991, -0.0957, -0.0735, -0.0518, -0.0445, -0.0432, 0.0, -0.0145, -0.0014, -0.0122, 0.0037, 0.0848, 0.1254, 0.2316, 0.2407, 0.3082, 0.4143, 0.3748, 0.4422, 0.4409, 0.3314, 0.325, 0.3033, 0.3252, 0.4295, 0.4123, 0.4123, 0.5427, 0.6245, 0.6222, 0.6897, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2050417440975372542", "ticker_idx": 0, "ts": "2026-05-02T03:29:47+00:00", "symbol": "FORM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we remain positive on FormFactor's four key growth drivers... We raise our 2026/2027 EPS forecasts to $2.75/$4.30. Applying an unchanged 40x 2027E P/E multiple, we set a target price of $193.", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF reiterates Buy on FormFactor; raises EPS estimates and sets $193 PT on AI probe card share gains, HBM ramp, and CPO testing commercialization.", "resolved_tickers": ["FORM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications — FormFactor 1Q26 Review: Sufficient Growth Momentum Ahead\n\nGross margin beat expectations; industry momentum continues\n\nFormFactor reported 1Q26 revenue of $230 million, with gross margin reaching 49.0%, up 5.1 percentage points QoQ and 2.5 percentage points above the high end of prior guidance.\n\nManagement guided for 2Q26 revenue of $240 million, with gross margin rising further to 49.5%.\n\nAs previously argued, we remain positive on FormFactor’s four key growth drivers: AI-driven share gains, accelerated HBM business ramp, and the commercialization of CPO testing. The May 11 Investor Day will be the next important catalyst, where we expect management to raise its long-term revenue target to around $1.6 billion.\n\nWe raise our 2026/2027 EPS forecasts to $2.75/$4.30. Applying an unchanged 40x 2027E P/E multiple, we set a target price of $193.\n\nAI probe card share is steadily rising\n\nThrough its deep partnership with Teradyne, FormFactor has become a core probe card supplier to Mellanox, helping Nvidia become a customer accounting for more than 10% of FormFactor’s revenue for the first time.\n\nLooking ahead, the company is expected to enter the AI probe card market starting with Nvidia’s Rubin platform. MediaTek’s Humufish platform scale-up in 2H27 should also bring another wave of incremental revenue.\n\nOn HBM, management stated that its second-largest customer is the biggest source of incremental revenue, which we believe refers to Samsung.\n\nNotably, HBM4 probe cards carry a 20% premium over HBM3E probe cards. Combined with 70% YoY growth in HBM wafer shipments, this should drive both revenue growth and gross margin improvement.\n\nInsertion 1 leadership remains solid; CPO probe cards open up new growth opportunities\n\nThe company’s 2026 CPO-related revenue is moving toward the high end of its $10–20 million guidance range.\n\nWith its Triton series, FormFactor has established a clear leadership position in the Insertion 1 probe station market. We expect shipments to exceed 100 units in 2027, corresponding to CPO revenue of more than $100 million.\n\nIn addition, the Keystone acquisition should further strengthen the company’s probe card technology capabilities in CPO testing. This incremental contribution has not yet been included in current revenue forecasts.\n\n$FORM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.022326412956877872, "ret_signed_m1d": 0.022326412956877872, "alpha_spy_m1d": 0.018649563591923712, "alpha_c_m1d": 0.016347607676173048, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07910995958243361, "ret_signed_p1d": 0.07910995958243361, "alpha_spy_p1d": 0.07108777369744135, "alpha_c_p1d": 0.04773603028549278, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.12673959513390298, "ret_signed_p1w": 0.12673959513390298, "alpha_spy_p1w": 0.09708822852650245, "alpha_c_p1w": -0.010437515388215024, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4674, -0.3758, -0.328, -0.2939, -0.3056, -0.285, -0.3028, -0.2801, -0.2801, -0.2902, -0.2954, -0.3132, -0.2963, -0.2847, -0.2515, -0.2124, -0.2725, -0.2641, -0.258, -0.3095, -0.3109, -0.355, -0.3673, -0.3374, -0.3225, -0.3167, -0.3421, -0.3308, -0.3276, -0.2952, -0.2842, -0.2923, -0.3137, -0.2766, -0.2146, -0.1957, -0.2532, -0.2712, -0.3197, -0.2782, -0.249, -0.2323, -0.2323, -0.2225, -0.2177, -0.1233, -0.099, -0.0787, -0.0458, -0.0453, -0.0354, -0.0498, 0.0211, 0.0719, 0.0571, 0.0879, 0.1042, 0.1541, 0.0711, -0.0014, 0.0086, 0.0116, 0.0223, 0.0, 0.0791, 0.1098, 0.0767, 0.1005, 0.1267, -0.0175, -0.0377, -0.0475, -0.0603, -0.1261, -0.1278, -0.0674, -0.0474, -0.04, -0.04, 0.0157, -0.01, -0.0309, -0.0728, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2054564780732498342", "ticker_idx": 0, "ts": "2026-05-13T14:09:50+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Apple and Intel already signed an agreement in December 2025. The M7 chip will use Intel 18A-P and is expected to enter production by the end of 2027, while the smartphone chip will use Intel 14A and is expected to enter production by the end of 2028.", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares bullish news: Apple signed deal to use Intel 18A-P for M7 and 14A for smartphone chips, implying meaningful foundry wins for INTC.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "From GFHK Monthly Call: Apple and Intel already signed an agreement in December 2025. The M7 chip will use Intel 18A-P and is expected to enter production by the end of 2027, while the smartphone chip will use Intel 14A and is expected to enter production by the end of 2028.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002660235201502159, "ret_signed_m1d": 0.002660235201502159, "alpha_spy_m1d": 0.008223955071092748, "alpha_c_m1d": 0.02224242704439683, "bench_spy_m1d": -0.005563719869590589, "bench_c_m1d": -0.01958219184289467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.036245744261098944, "ret_signed_p1d": -0.036245744261098944, "alpha_spy_p1d": -0.0441400005298217, "alpha_c_p1d": -0.04651721241105011, "bench_spy_p1d": 0.007894256268722755, "bench_c_p1d": 0.010271468149951168, "ret_p1w": -0.011056628322653972, "ret_signed_p1w": -0.011056628322653972, "alpha_spy_p1w": -0.009628656771576871, "alpha_c_p1w": 0.002568862591535215, "bench_spy_p1w": -0.0014279715510771007, "bench_c_p1w": -0.013625490914189187, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.611, -0.611, -0.6161, -0.6221, -0.6291, -0.6333, -0.6373, -0.6166, -0.6103, -0.6221, -0.6208, -0.6217, -0.6417, -0.6211, -0.618, -0.639, -0.6211, -0.6111, -0.6011, -0.6238, -0.6195, -0.6196, -0.6337, -0.6257, -0.6161, -0.6353, -0.6341, -0.6337, -0.6078, -0.6334, -0.6414, -0.6576, -0.6331, -0.6007, -0.5812, -0.5812, -0.5779, -0.5601, -0.5099, -0.4869, -0.4814, -0.4581, -0.4695, -0.4601, -0.4305, -0.4305, -0.4538, -0.4492, -0.4574, -0.4448, -0.3138, -0.2935, -0.2974, -0.2123, -0.2146, -0.1718, -0.2038, -0.1009, -0.0605, -0.0887, 0.0385, 0.0761, 0.0027, 0.0, -0.0362, -0.0958, -0.1008, -0.0789, -0.0111, -0.0149, -0.0037, -0.0037, 0.0269, 0.0123, 0.005, -0.0466, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2059498535918456961", "ticker_idx": 0, "ts": "2026-05-27T04:54:49+00:00", "symbol": "Murata Manufacturing", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Major Asian MLCC makers have likewise entered full-capacity production: Murata's and Samsung Electro-Mechanics' utilization rates have surpassed 90%", "tweet_type": "analysis", "timeline": "months", "summary": "High-spec MLCC/resistor shortage widening; full-capacity utilization + order suspensions signal price-hike cycle; high-voltage MLCC makers named biggest Vera Rubin beneficiaries.", "resolved_tickers": ["6981.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Murata Manufacturing 6981.T Japan MLCC leader", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "- China's largest passive component maker, Fenghua Advanced Technology, has suspended order intake across all 0402/0603-size chip resistors and MLCC product lines in response to surging orders.\n\n- The company said order volumes for its RC/RS series 0402/0603 resistors have spiked, with unfilled backlog now far exceeding current production capacity; utilization has already reached roughly 90%.\n\n- The supply chain attributes the order concentration to expanded preemptive (pull-in) ordering aimed at securing inventory, amplified by fears of supply shortages.\n\n- The industry reads the surge in forward orders as the product of shortage fears in high-spec passive components colliding with the need to rebuild inventory. An order suspension is typically interpreted as a signal of an oncoming price-hike cycle.\n\n- With inventory levels drawing down on recovering Chinese-market and automotive demand, and chip-resistor costs pressured by rising silver prices, the high-spec passive component shortage is expected to widen further.\n\n- Rising silver prices in particular are weighing on chip-resistor costs, and the high-spec MLCC shortage could intensify further in 2027.\n\n- Major Asian MLCC makers have likewise entered full-capacity production: Murata's and Samsung Electro-Mechanics' utilization rates have surpassed 90%, with Samsung Electro-Mechanics' Tianjin plant already running flat out.\n\n- Holy Stone and PDC are similarly seeing extended lead times and utilization approaching full capacity, led by high-voltage MLCCs for AI power.\n\n- Once the Vera Rubin platform launches in the second half, the supply squeeze on non-AI products is expected to intensify, positioning high-voltage MLCC makers as the biggest beneficiaries of the GPU platform generational transition.\n\n---\n\nhttps://t.co/gxf6c0rQgS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0328644501278772, "ret_signed_m1d": 0.0328644501278772, "alpha_spy_m1d": 0.03269121653913687, "alpha_c_m1d": 0.029014714498839078, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.00384973562903812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0918158567774936, "ret_signed_p1d": 0.0918158567774936, "alpha_spy_p1d": 0.08629930227483129, "alpha_c_p1d": 0.07869427393392314, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.013121582843570456, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4796, -0.485, -0.5275, -0.5367, -0.5176, -0.5208, -0.56, -0.5348, -0.5239, -0.5287, -0.5376, -0.5431, -0.5184, -0.513, -0.5252, -0.5252, -0.5556, -0.5512, -0.5363, -0.5284, -0.5256, -0.5481, -0.5641, -0.5306, -0.5552, -0.5155, -0.5132, -0.5159, -0.4885, -0.4898, -0.4698, -0.4668, -0.4464, -0.422, -0.3872, -0.4118, -0.3941, -0.3849, -0.372, -0.3845, -0.3688, -0.3714, -0.3811, -0.3811, -0.3407, -0.343, -0.343, -0.343, -0.343, -0.2777, -0.2519, -0.2379, -0.2228, -0.2214, -0.1894, -0.2134, -0.2109, -0.2097, -0.2118, -0.1398, -0.0882, 0.0138, 0.0329, 0.0, 0.0918, 0.2308, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2059498535918456961", "ticker_idx": 1, "ts": "2026-05-27T04:54:49+00:00", "symbol": "009150.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electro-Mechanics' utilization rates have surpassed 90%, with Samsung Electro-Mechanics' Tianjin plant already running flat out", "tweet_type": "analysis", "timeline": "months", "summary": "High-spec MLCC/resistor shortage widening; full-capacity utilization + order suspensions signal price-hike cycle; high-voltage MLCC makers named biggest Vera Rubin beneficiaries.", "resolved_tickers": ["009150.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "- China's largest passive component maker, Fenghua Advanced Technology, has suspended order intake across all 0402/0603-size chip resistors and MLCC product lines in response to surging orders.\n\n- The company said order volumes for its RC/RS series 0402/0603 resistors have spiked, with unfilled backlog now far exceeding current production capacity; utilization has already reached roughly 90%.\n\n- The supply chain attributes the order concentration to expanded preemptive (pull-in) ordering aimed at securing inventory, amplified by fears of supply shortages.\n\n- The industry reads the surge in forward orders as the product of shortage fears in high-spec passive components colliding with the need to rebuild inventory. An order suspension is typically interpreted as a signal of an oncoming price-hike cycle.\n\n- With inventory levels drawing down on recovering Chinese-market and automotive demand, and chip-resistor costs pressured by rising silver prices, the high-spec passive component shortage is expected to widen further.\n\n- Rising silver prices in particular are weighing on chip-resistor costs, and the high-spec MLCC shortage could intensify further in 2027.\n\n- Major Asian MLCC makers have likewise entered full-capacity production: Murata's and Samsung Electro-Mechanics' utilization rates have surpassed 90%, with Samsung Electro-Mechanics' Tianjin plant already running flat out.\n\n- Holy Stone and PDC are similarly seeing extended lead times and utilization approaching full capacity, led by high-voltage MLCCs for AI power.\n\n- Once the Vera Rubin platform launches in the second half, the supply squeeze on non-AI products is expected to intensify, positioning high-voltage MLCC makers as the biggest beneficiaries of the GPU platform generational transition.\n\n---\n\nhttps://t.co/gxf6c0rQgS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03558282208588959, "ret_signed_m1d": -0.03558282208588959, "alpha_spy_m1d": -0.035756055674629916, "alpha_c_m1d": -0.03943255771492771, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.00384973562903812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.1343558282208588, "ret_signed_p1d": 0.1343558282208588, "alpha_spy_p1d": 0.12883927371819648, "alpha_c_p1d": 0.12123424537728833, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.013121582843570456, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.7248, -0.7248, -0.7491, -0.7776, -0.7537, -0.7518, -0.7656, -0.7534, -0.7509, -0.7549, -0.7543, -0.7466, -0.7261, -0.7153, -0.7058, -0.715, -0.7377, -0.7319, -0.7212, -0.7196, -0.7196, -0.7368, -0.75, -0.7276, -0.7442, -0.7202, -0.7166, -0.7196, -0.6847, -0.6834, -0.6534, -0.6515, -0.6417, -0.6233, -0.608, -0.5834, -0.5828, -0.5264, -0.5018, -0.5252, -0.5166, -0.5129, -0.4853, -0.4926, -0.4896, -0.4896, -0.4368, -0.4368, -0.4405, -0.4374, -0.4393, -0.4479, -0.4123, -0.3687, -0.3718, -0.3804, -0.3675, -0.3945, -0.3491, -0.2613, -0.1779, -0.1779, -0.0356, 0.0, 0.1344, 0.3049, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2059498535918456961", "ticker_idx": 2, "ts": "2026-05-27T04:54:49+00:00", "symbol": "Fenghua Advanced Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "An order suspension is typically interpreted as a signal of an oncoming price-hike cycle", "tweet_type": "analysis", "timeline": "months", "summary": "High-spec MLCC/resistor shortage widening; full-capacity utilization + order suspensions signal price-hike cycle; high-voltage MLCC makers named biggest Vera Rubin beneficiaries.", "resolved_tickers": ["000636.SZ"], "resolution_method": "exchange_ticker", "resolution_reason": "Fenghua Advanced Technology 000636.SZ China MLCC", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "- China's largest passive component maker, Fenghua Advanced Technology, has suspended order intake across all 0402/0603-size chip resistors and MLCC product lines in response to surging orders.\n\n- The company said order volumes for its RC/RS series 0402/0603 resistors have spiked, with unfilled backlog now far exceeding current production capacity; utilization has already reached roughly 90%.\n\n- The supply chain attributes the order concentration to expanded preemptive (pull-in) ordering aimed at securing inventory, amplified by fears of supply shortages.\n\n- The industry reads the surge in forward orders as the product of shortage fears in high-spec passive components colliding with the need to rebuild inventory. An order suspension is typically interpreted as a signal of an oncoming price-hike cycle.\n\n- With inventory levels drawing down on recovering Chinese-market and automotive demand, and chip-resistor costs pressured by rising silver prices, the high-spec passive component shortage is expected to widen further.\n\n- Rising silver prices in particular are weighing on chip-resistor costs, and the high-spec MLCC shortage could intensify further in 2027.\n\n- Major Asian MLCC makers have likewise entered full-capacity production: Murata's and Samsung Electro-Mechanics' utilization rates have surpassed 90%, with Samsung Electro-Mechanics' Tianjin plant already running flat out.\n\n- Holy Stone and PDC are similarly seeing extended lead times and utilization approaching full capacity, led by high-voltage MLCCs for AI power.\n\n- Once the Vera Rubin platform launches in the second half, the supply squeeze on non-AI products is expected to intensify, positioning high-voltage MLCC makers as the biggest beneficiaries of the GPU platform generational transition.\n\n---\n\nhttps://t.co/gxf6c0rQgS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0569877971098709, "ret_signed_m1d": -0.0569877971098709, "alpha_spy_m1d": -0.057161030698611226, "alpha_c_m1d": -0.06083753273890902, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.00384973562903812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09995472742944966, "ret_signed_p1d": 0.09995472742944966, "alpha_spy_p1d": 0.09443817292678736, "alpha_c_p1d": 0.08683314458587921, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.013121582843570456, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.412, -0.4007, -0.4607, -0.4595, -0.4629, -0.4629, -0.4796, -0.4597, -0.4708, -0.4912, -0.5093, -0.5036, -0.4934, -0.4912, -0.4905, -0.5185, -0.5572, -0.5455, -0.5443, -0.5606, -0.5565, -0.5622, -0.5719, -0.5586, -0.5789, -0.5791, -0.5791, -0.5805, -0.5516, -0.5538, -0.5486, -0.5366, -0.5287, -0.5242, -0.4975, -0.5027, -0.4986, -0.4982, -0.4916, -0.4989, -0.486, -0.4733, -0.4473, -0.4283, -0.4326, -0.4326, -0.4326, -0.4326, -0.4389, -0.4313, -0.415, -0.3564, -0.334, -0.3313, -0.3621, -0.3894, -0.3345, -0.268, -0.2311, -0.2266, -0.1493, -0.0642, -0.057, 0.0, 0.1, 0.1992, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2059498535918456961", "ticker_idx": 3, "ts": "2026-05-27T04:54:49+00:00", "symbol": "high-spec passive components", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "high-voltage MLCC makers as the biggest beneficiaries of the GPU platform generational transition", "tweet_type": "analysis", "timeline": "months", "summary": "High-spec MLCC/resistor shortage widening; full-capacity utilization + order suspensions signal price-hike cycle; high-voltage MLCC makers named biggest Vera Rubin beneficiaries.", "resolved_tickers": ["009150.KS", "6981.T", "2327.TW", "6976.T"], "resolution_method": "basket", "resolution_reason": "High-spec passives basket: Samsung EM, Murata, Yageo, TDK", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "- China's largest passive component maker, Fenghua Advanced Technology, has suspended order intake across all 0402/0603-size chip resistors and MLCC product lines in response to surging orders.\n\n- The company said order volumes for its RC/RS series 0402/0603 resistors have spiked, with unfilled backlog now far exceeding current production capacity; utilization has already reached roughly 90%.\n\n- The supply chain attributes the order concentration to expanded preemptive (pull-in) ordering aimed at securing inventory, amplified by fears of supply shortages.\n\n- The industry reads the surge in forward orders as the product of shortage fears in high-spec passive components colliding with the need to rebuild inventory. An order suspension is typically interpreted as a signal of an oncoming price-hike cycle.\n\n- With inventory levels drawing down on recovering Chinese-market and automotive demand, and chip-resistor costs pressured by rising silver prices, the high-spec passive component shortage is expected to widen further.\n\n- Rising silver prices in particular are weighing on chip-resistor costs, and the high-spec MLCC shortage could intensify further in 2027.\n\n- Major Asian MLCC makers have likewise entered full-capacity production: Murata's and Samsung Electro-Mechanics' utilization rates have surpassed 90%, with Samsung Electro-Mechanics' Tianjin plant already running flat out.\n\n- Holy Stone and PDC are similarly seeing extended lead times and utilization approaching full capacity, led by high-voltage MLCCs for AI power.\n\n- Once the Vera Rubin platform launches in the second half, the supply squeeze on non-AI products is expected to intensify, positioning high-voltage MLCC makers as the biggest beneficiaries of the GPU platform generational transition.\n\n---\n\nhttps://t.co/gxf6c0rQgS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.007638273548622121, "ret_signed_m1d": -0.007638273548622121, "alpha_spy_m1d": -0.00781150713736245, "alpha_c_m1d": -0.011488009177660241, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.00384973562903812, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11365589706009516, "ret_signed_p1d": 0.11365589706009516, "alpha_spy_p1d": 0.10813934255743285, "alpha_c_p1d": 0.1005343142165247, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.013121582843570456, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5882, -0.5901, -0.62, -0.6419, -0.6253, -0.6275, -0.6552, -0.6389, -0.6332, -0.6423, -0.6434, -0.6412, -0.6161, -0.6105, -0.6173, -0.621, -0.6458, -0.6431, -0.6303, -0.6261, -0.6269, -0.6432, -0.6586, -0.6339, -0.6503, -0.6232, -0.6174, -0.6146, -0.5836, -0.5836, -0.5677, -0.5573, -0.5421, -0.515, -0.4926, -0.4965, -0.4904, -0.4761, -0.4651, -0.4826, -0.4691, -0.4652, -0.4521, -0.4555, -0.4508, -0.457, -0.4347, -0.4326, -0.4372, -0.3972, -0.3843, -0.3801, -0.363, -0.3506, -0.3202, -0.334, -0.3042, -0.2992, -0.278, -0.2132, -0.1376, -0.0562, -0.0076, 0.0, 0.1137, 0.2302, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2043583889118318896", "ticker_idx": 0, "ts": "2026-04-13T06:55:41+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML 들고 있어요", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Holds ASML as preferred play on selling equipment to all chipmakers regardless of who wins.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@Xgngu12 @Xterbase @PhotonCap 저는 오히려 메이커가 누가 이길지 찍는거보다 모든 메이커에게 장비를 파는 회사가 좋더라고요. 그래서 ASML 들고 있어요", "reply_text": null, "parent_text": "@Xterbase @jukan05 @PhotonCap ㅋㅋ 바쁘신데 알림 너무 가면 귀찮으실까봐😂\n\n저는 장비쪽은 싸이클 못맞추면 투자 관점에선 너무 재미가 없어지다 보니까 좀 무서워서 잘 안건드리게 되더라구요. 매출이 계절성 상관없이 잘 나오는 곳들을 투자처로 선호합니다ㅋㅋ\n\n그래도 올려주시면 또 제 편향일 수 있으니 열심히 살펴볼게요!", "quoted_text": null, "parent_author": "Xgngu12", "ret_m1d": 0.008096420951350103, "ret_signed_m1d": 0.008096420951350103, "alpha_spy_m1d": 0.017774243206934104, "alpha_c_m1d": 0.022667611253239017, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.014571190301888914, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01936807460209078, "ret_signed_p1d": 0.01936807460209078, "alpha_spy_p1d": 0.007183193849720659, "alpha_c_p1d": -0.00016547466854421877, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.019533549270634998, "ret_p1w": -0.011747982255521316, "ret_signed_p1w": -0.011747982255521316, "alpha_spy_p1w": -0.04471693127475107, "alpha_c_p1w": -0.058258550414302945, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.04651056815878163, "ret_p1m": 0.006984307764408726, "ret_signed_p1m": 0.006984307764408726, "alpha_spy_p1m": -0.06892302198788691, "alpha_c_p1m": -0.25897410933325493, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.26595841709766366, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1405, -0.0889, -0.0747, -0.1118, -0.0963, -0.0846, -0.0675, -0.066, -0.0662, -0.0348, -0.0532, -0.0551, -0.0364, -0.0291, -0.0564, -0.0958, -0.0887, -0.0537, -0.0437, -0.053, -0.0413, -0.0635, -0.0551, -0.0514, -0.0481, -0.0119, -0.0171, -0.0033, -0.0084, 0.0029, 0.0227, -0.0217, -0.021, -0.0392, -0.0778, -0.0476, -0.0586, -0.0895, -0.0891, -0.0475, -0.0484, -0.063, -0.0637, -0.0495, -0.0524, -0.0546, -0.0724, -0.1045, -0.067, -0.0462, -0.0383, -0.0746, -0.09, -0.1173, -0.1116, -0.0573, -0.0784, -0.0784, -0.0784, -0.1159, -0.0375, -0.0198, 0.0081, 0.0, 0.0194, -0.0237, -0.0295, -0.0125, -0.0117, -0.0181, -0.0079, -0.0286, -0.0059, -0.0353, -0.0678, -0.0512, -0.0275, -0.0275, -0.0554, -0.0223, 0.0382, 0.0342, 0.0496, 0.0386, 0.007, 0.0557, 0.0875, 0.0394, 0.0059, -0.0064, 0.0604, 0.0701, 0.1209, 0.1381, 0.1056, 0.095, 0.1066, 0.1017, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038892716428603841", "ticker_idx": 0, "ts": "2026-03-31T08:14:39+00:00", "symbol": "NAND", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "You know what to buy, right?", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Kioxia dropping SLC/MLC implies NAND supply tightening; author signals a buy in the space.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "NAND sector basket", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia has announced the discontinuation of SLC and MLC.\n\nYou know what to buy, right? https://t.co/fmei5RhONs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007818136828624955, "ret_signed_m1d": 0.007818136828624955, "alpha_spy_m1d": 0.03606498523678279, "alpha_c_m1d": 0.062252136435892204, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11350214106817669, "ret_signed_p1d": 0.11350214106817669, "alpha_spy_p1d": 0.10596767728528667, "alpha_c_p1d": 0.0911494872845342, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.15326194348547023, "ret_signed_p1w": 0.15326194348547023, "alpha_spy_p1w": 0.13960763423036776, "alpha_c_p1w": 0.11022594906014199, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.6134250568025192, "ret_signed_p1m": 0.6134250568025192, "alpha_spy_p1m": 0.5192589472206082, "alpha_c_p1m": 0.31039951030981605, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3428, -0.2965, -0.273, -0.2197, -0.2029, -0.2122, -0.194, -0.1888, -0.1869, -0.1904, -0.172, -0.1358, -0.1285, -0.1249, -0.0775, -0.0499, -0.0613, -0.075, -0.024, 0.0242, 0.0342, 0.0673, 0.0458, 0.1093, 0.0521, 0.0111, 0.0188, 0.0268, 0.0035, 0.0421, 0.0982, 0.1143, 0.1086, 0.091, 0.0966, 0.1132, 0.1352, 0.1401, 0.1691, 0.1727, 0.2064, 0.1877, 0.186, 0.08, 0.0434, 0.0877, 0.0482, 0.019, 0.0953, 0.1417, 0.1121, 0.1258, 0.1866, 0.1983, 0.2663, 0.2307, 0.1956, 0.139, 0.1472, 0.147, 0.0683, 0.0639, 0.0078, 0.0, 0.1135, 0.0779, 0.1065, 0.1413, 0.1533, 0.2837, 0.2993, 0.334, 0.3783, 0.4602, 0.4244, 0.4616, 0.4173, 0.4182, 0.4589, 0.5248, 0.5217, 0.5336, 0.608, 0.5825, 0.6134, 0.6275, 0.6588, 0.755, 0.8401, 0.9351, 0.9943, 1.1393, 1.247, 1.1826, 1.2901, 1.2449, 1.1124, 1.1545, 1.1328, 1.1719, 1.3558, 1.3427, 1.427, 1.5517, 1.6079, 1.6315, 1.7561, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042472966139887934", "ticker_idx": 0, "ts": "2026-04-10T05:21:17+00:00", "symbol": "009150.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The AI server MLCC market is forming an oligopolistic structure centered on Samsung Electro-Mechanics and Murata, positioning them as primary beneficiaries if prices rise", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Samsung Electro-Mechanics and Murata as oligopolistic MLCC winners in AI server upcycle; DRAM prices also rising with Q2 gains expected.", "resolved_tickers": ["009150.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Demand for AI infrastructure spreads shortages beyond memory chips to MLCCs\n\n• The supply shortage, previously centered on memory chips, is now spreading to MLCCs due to expanding AI infrastructure investment, with lead time increases confirmed across the industry\n\n• As of Q1 2026, most major suppliers—including Kyocera, Murata, Samsung Electro-Mechanics, TDK, Yageo, and Walsin—are experiencing delivery delays, with Vishay being the sole exception\n\n• The key drivers of the tight supply-demand balance are surging demand for high-spec MLCCs in AI servers and automotive electronics, with major manufacturers’ production lines already operating near capacity limits\n\n• Samsung Electro-Mechanics’ MLCC lines are effectively running at full utilization, leaving limited room to accommodate additional demand\n\n• Supply shortages are causing order delays and priority-based allocation; industry leader Murata has begun pricing discussions, signaling broader price increases across the sector\n\n• Rising lead times are generally interpreted as a leading indicator of price increases and an industry recovery, highlighting the potential onset of a new MLCC upcycle\n\n• The AI server MLCC market is forming an oligopolistic structure centered on Samsung Electro-Mechanics and Murata, positioning them as primary beneficiaries if prices rise\n\n• Meanwhile, DRAM prices rose sharply QoQ in Q1 2026, with further increases expected in Q2\n\n• The DRAM price uptrend is spilling over into other electronic components such as MLCCs, as AI-driven demand ripples across the broader component supply chain\n\n---\n\nhttps://t.co/QGNwxIS2vH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08672566371681412, "ret_signed_m1d": -0.08672566371681412, "alpha_spy_m1d": -0.0873878824748342, "alpha_c_m1d": -0.08286934754938724, "bench_spy_m1d": 0.0006622187580200833, "bench_c_m1d": -0.0038563161674268764, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005309734513274433, "ret_signed_p1d": 0.005309734513274433, "alpha_spy_p1d": -0.004462663271077361, "alpha_c_p1d": -0.015655106291181164, "bench_spy_p1d": 0.009772397784351794, "bench_c_p1d": 0.020964840804455598, "ret_p1w": 0.2017699115044247, "ret_signed_p1w": 0.2017699115044247, "alpha_spy_p1w": 0.15661641949070915, "alpha_c_p1w": 0.11952330241788078, "bench_spy_p1w": 0.045153492013715546, "bench_c_p1w": 0.08224660908654391, "ret_p1m": 0.5929203539823009, "ret_signed_p1m": 0.5929203539823009, "alpha_spy_p1m": 0.5048504692426794, "alpha_c_p1m": 0.3456898745768524, "bench_spy_p1m": 0.08806988473962152, "bench_c_p1m": 0.2472304794054485, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4885, -0.4965, -0.4876, -0.4876, -0.4832, -0.5009, -0.5044, -0.4991, -0.5186, -0.5133, -0.5097, -0.5044, -0.5053, -0.5062, -0.4991, -0.4558, -0.4602, -0.4867, -0.4973, -0.4619, -0.4451, -0.4522, -0.4345, -0.4522, -0.4522, -0.4522, -0.4522, -0.3664, -0.3327, -0.2451, -0.2106, -0.177, -0.1743, -0.2062, -0.2062, -0.2761, -0.3584, -0.2894, -0.2841, -0.3239, -0.2885, -0.2814, -0.2929, -0.2912, -0.269, -0.2097, -0.1788, -0.1513, -0.1779, -0.2434, -0.2265, -0.1956, -0.1912, -0.1912, -0.2407, -0.2788, -0.2142, -0.2619, -0.1929, -0.1823, -0.1912, -0.0903, -0.0867, 0.0, 0.0053, 0.0336, 0.0867, 0.131, 0.2018, 0.2035, 0.3664, 0.4372, 0.3699, 0.3947, 0.4053, 0.485, 0.4637, 0.4726, 0.4726, 0.6248, 0.6248, 0.6142, 0.623, 0.6177, 0.5929, 0.6956, 0.8212, 0.8124, 0.7876, 0.8248, 0.7469, 0.8779, 1.131, 1.3717, 1.3717, 1.7823, 1.885, 2.2726, 2.7646, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042472966139887934", "ticker_idx": 1, "ts": "2026-04-10T05:21:17+00:00", "symbol": "Murata", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The AI server MLCC market is forming an oligopolistic structure centered on Samsung Electro-Mechanics and Murata, positioning them as primary beneficiaries if prices rise", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Samsung Electro-Mechanics and Murata as oligopolistic MLCC winners in AI server upcycle; DRAM prices also rising with Q2 gains expected.", "resolved_tickers": ["6981.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Murata Manufacturing Co., listed on Tokyo Stock Exchange", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Demand for AI infrastructure spreads shortages beyond memory chips to MLCCs\n\n• The supply shortage, previously centered on memory chips, is now spreading to MLCCs due to expanding AI infrastructure investment, with lead time increases confirmed across the industry\n\n• As of Q1 2026, most major suppliers—including Kyocera, Murata, Samsung Electro-Mechanics, TDK, Yageo, and Walsin—are experiencing delivery delays, with Vishay being the sole exception\n\n• The key drivers of the tight supply-demand balance are surging demand for high-spec MLCCs in AI servers and automotive electronics, with major manufacturers’ production lines already operating near capacity limits\n\n• Samsung Electro-Mechanics’ MLCC lines are effectively running at full utilization, leaving limited room to accommodate additional demand\n\n• Supply shortages are causing order delays and priority-based allocation; industry leader Murata has begun pricing discussions, signaling broader price increases across the sector\n\n• Rising lead times are generally interpreted as a leading indicator of price increases and an industry recovery, highlighting the potential onset of a new MLCC upcycle\n\n• The AI server MLCC market is forming an oligopolistic structure centered on Samsung Electro-Mechanics and Murata, positioning them as primary beneficiaries if prices rise\n\n• Meanwhile, DRAM prices rose sharply QoQ in Q1 2026, with further increases expected in Q2\n\n• The DRAM price uptrend is spilling over into other electronic components such as MLCCs, as AI-driven demand ripples across the broader component supply chain\n\n---\n\nhttps://t.co/QGNwxIS2vH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0376266280752533, "ret_signed_m1d": -0.0376266280752533, "alpha_spy_m1d": -0.03828884683327338, "alpha_c_m1d": -0.03377031190782642, "bench_spy_m1d": 0.0006622187580200833, "bench_c_m1d": -0.0038563161674268764, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005788712011577379, "ret_signed_p1d": 0.005788712011577379, "alpha_spy_p1d": -0.003983685772774415, "alpha_c_p1d": -0.015176128792878218, "bench_spy_p1d": 0.009772397784351794, "bench_c_p1d": 0.020964840804455598, "ret_p1w": 0.1095031355523397, "ret_signed_p1w": 0.1095031355523397, "alpha_spy_p1w": 0.06434964353862416, "alpha_c_p1w": 0.027256526465795794, "bench_spy_p1w": 0.045153492013715546, "bench_c_p1w": 0.08224660908654391, "ret_p1m": 0.437530149541727, "ret_signed_p1m": 0.437530149541727, "alpha_spy_p1m": 0.34946026480210546, "alpha_c_p1m": 0.19029967013627846, "bench_spy_p1m": 0.08806988473962152, "bench_c_p1m": 0.2472304794054485, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2096, -0.1768, -0.1876, -0.1843, -0.1797, -0.2101, -0.2115, -0.2034, -0.2038, -0.2404, -0.2287, -0.2385, -0.2509, -0.2507, -0.2629, -0.2072, -0.2062, -0.2151, -0.2313, -0.2189, -0.207, -0.207, -0.2127, -0.2213, -0.2156, -0.1616, -0.1374, -0.124, -0.1219, -0.1219, -0.0335, 0.0091, -0.0232, -0.0184, -0.0287, -0.1087, -0.1262, -0.0901, -0.0961, -0.1702, -0.1226, -0.1021, -0.1111, -0.1279, -0.1381, -0.0915, -0.0815, -0.1044, -0.1044, -0.1618, -0.1534, -0.1255, -0.1104, -0.1052, -0.1476, -0.1778, -0.1146, -0.1611, -0.0861, -0.0818, -0.0868, -0.0352, -0.0376, 0.0, 0.0058, 0.0441, 0.0902, 0.1558, 0.1095, 0.1428, 0.1602, 0.1845, 0.1609, 0.1905, 0.1857, 0.1674, 0.1674, 0.2436, 0.2393, 0.2393, 0.2393, 0.2393, 0.3623, 0.411, 0.4375, 0.466, 0.4686, 0.5289, 0.4836, 0.4884, 0.4906, 0.4867, 0.6225, 0.7197, 0.9122, 0.9481, 0.8862, 1.0593, 1.3215, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042472966139887934", "ticker_idx": 2, "ts": "2026-04-10T05:21:17+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DRAM prices rose sharply QoQ in Q1 2026, with further increases expected in Q2", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Samsung Electro-Mechanics and Murata as oligopolistic MLCC winners in AI server upcycle; DRAM prices also rising with Q2 gains expected.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Demand for AI infrastructure spreads shortages beyond memory chips to MLCCs\n\n• The supply shortage, previously centered on memory chips, is now spreading to MLCCs due to expanding AI infrastructure investment, with lead time increases confirmed across the industry\n\n• As of Q1 2026, most major suppliers—including Kyocera, Murata, Samsung Electro-Mechanics, TDK, Yageo, and Walsin—are experiencing delivery delays, with Vishay being the sole exception\n\n• The key drivers of the tight supply-demand balance are surging demand for high-spec MLCCs in AI servers and automotive electronics, with major manufacturers’ production lines already operating near capacity limits\n\n• Samsung Electro-Mechanics’ MLCC lines are effectively running at full utilization, leaving limited room to accommodate additional demand\n\n• Supply shortages are causing order delays and priority-based allocation; industry leader Murata has begun pricing discussions, signaling broader price increases across the sector\n\n• Rising lead times are generally interpreted as a leading indicator of price increases and an industry recovery, highlighting the potential onset of a new MLCC upcycle\n\n• The AI server MLCC market is forming an oligopolistic structure centered on Samsung Electro-Mechanics and Murata, positioning them as primary beneficiaries if prices rise\n\n• Meanwhile, DRAM prices rose sharply QoQ in Q1 2026, with further increases expected in Q2\n\n• The DRAM price uptrend is spilling over into other electronic components such as MLCCs, as AI-driven demand ripples across the broader component supply chain\n\n---\n\nhttps://t.co/QGNwxIS2vH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011919640773676651, "ret_signed_m1d": -0.011919640773676651, "alpha_spy_m1d": -0.012581859531696734, "alpha_c_m1d": 0.00311883031858832, "bench_spy_m1d": 0.0006622187580200833, "bench_c_m1d": -0.015038471092264971, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0008602394610422633, "ret_signed_p1d": 0.0008602394610422633, "alpha_spy_p1d": -0.00891215832330953, "alpha_c_p1d": -0.01392640992294806, "bench_spy_p1d": 0.009772397784351794, "bench_c_p1d": 0.014786649383990325, "ret_p1w": 0.07628947782758504, "ret_signed_p1w": 0.07628947782758504, "alpha_spy_p1w": 0.03113598581386949, "alpha_c_p1w": 0.013846705271275558, "bench_spy_p1w": 0.045153492013715546, "bench_c_p1w": 0.06244277255630948, "ret_p1m": 0.7024944204819789, "ret_signed_p1m": 0.7024944204819789, "alpha_spy_p1m": 0.6144245357423573, "alpha_c_p1m": 0.3833450176312313, "bench_spy_p1m": 0.08806988473962152, "bench_c_p1m": 0.31914940285074755, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2708, -0.269, -0.2583, -0.2273, -0.224, -0.2357, -0.2106, -0.1945, -0.1893, -0.2077, -0.1583, -0.1205, -0.1163, -0.1177, -0.1414, -0.1034, -0.1348, -0.1667, -0.1594, -0.1401, -0.1528, -0.1254, -0.096, -0.0962, -0.0962, -0.1056, -0.0888, -0.0729, -0.0464, -0.0468, -0.0203, -0.0017, 0.0379, 0.0207, 0.0209, -0.0794, -0.1289, -0.0706, -0.1028, -0.1401, -0.0728, -0.0515, -0.0736, -0.0707, -0.0292, -0.0063, 0.0452, 0.0045, -0.0161, -0.076, -0.0603, -0.0692, -0.1248, -0.1234, -0.1763, -0.1998, -0.1104, -0.1517, -0.1241, -0.1006, -0.0855, -0.0018, -0.0119, 0.0, 0.0009, 0.0612, 0.0717, 0.0892, 0.0763, 0.0809, 0.1078, 0.1352, 0.1426, 0.1455, 0.1983, 0.1808, 0.1963, 0.1841, 0.2039, 0.3027, 0.3532, 0.4784, 0.4886, 0.5736, 0.7025, 0.6546, 0.7378, 0.7334, 0.6024, 0.592, 0.5659, 0.5931, 0.7183, 0.6985, 0.6985, 0.8599, 0.9606, 0.9596, 1.0398, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2046070726978076835", "ticker_idx": 0, "ts": "2026-04-20T03:37:30+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix's Q2 is going to be way stronger than Q1", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish SK Hynix Q2 on Trendforce-revised DDR5/DDR4 contract price forecasts up +18%/+23%, hitting $37.5 by end-April.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Hey guys, Trendforce just revised April DRAM contract price forecasts even higher. They bumped DDR5/DDR4 up by +18% and +23% respectively vs. the March estimates lol.\n\nThe contract prices to be announced at the end of April are expected to hit $37.5 for both DDR5 and DDR4 — that's +27%/+21% MoM gains.\n\nSK Hynix's Q2 is going to be way stronger than Q1.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03259003435394425, "ret_signed_m1d": -0.03259003435394425, "alpha_spy_m1d": -0.03459370858920496, "alpha_c_m1d": -0.03302113231651638, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": 0.0004310979625721245, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04974277430616603, "ret_signed_p1d": 0.04974277430616603, "alpha_spy_p1d": 0.05628972346110228, "alpha_c_p1d": 0.048233997213477764, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0015087770926882627, "ret_p1w": 0.10806181136639492, "ret_signed_p1w": 0.10806181136639492, "alpha_spy_p1w": 0.09896087944171361, "alpha_c_p1w": 0.01689011749278113, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.09117169387361379, "ret_p1m": 0.49656953778719415, "ret_signed_p1m": 0.49656953778719415, "alpha_spy_p1m": 0.46128055088195374, "alpha_c_p1m": 0.3241408111922979, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.17242872659489628, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3665, -0.3537, -0.3434, -0.3699, -0.3152, -0.2801, -0.2629, -0.2218, -0.2895, -0.2236, -0.2296, -0.2792, -0.2818, -0.2407, -0.2501, -0.2638, -0.2398, -0.2467, -0.2467, -0.2467, -0.2467, -0.2347, -0.1876, -0.1859, -0.1397, -0.1285, -0.0575, -0.0901, -0.0901, -0.1947, -0.2719, -0.193, -0.2075, -0.283, -0.1955, -0.181, -0.2024, -0.2196, -0.1647, -0.1681, -0.0943, -0.1312, -0.1364, -0.1998, -0.1544, -0.1467, -0.1998, -0.1998, -0.2513, -0.3079, -0.2307, -0.2882, -0.2487, -0.2401, -0.2144, -0.1141, -0.1441, -0.1192, -0.1081, -0.054, -0.0257, -0.0094, -0.0326, 0.0, 0.0497, 0.0489, 0.0506, 0.048, 0.1081, 0.1149, 0.1089, 0.1029, 0.1029, 0.241, 0.241, 0.3731, 0.4185, 0.446, 0.6123, 0.5738, 0.6947, 0.6895, 0.56, 0.578, 0.4966, 0.4966, 0.6638, 0.6647, 0.6647, 0.7599, 0.9237, 0.9635, 1.0012, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2056239458278928796", "ticker_idx": 0, "ts": "2026-05-18T05:04:24+00:00", "symbol": "AT&S", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "Bullish", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish on AT&S.", "resolved_tickers": ["ATS.VI"], "resolution_method": "exchange_ticker", "resolution_reason": "AT&S Austria listed on Vienna Exchange ATS.VI", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@mcrnhrdt Bullish", "reply_text": null, "parent_text": "@jukan05 What do you say to AT&amp;S?", "quoted_text": null, "parent_author": "mcrnhrdt", "ret_m1d": -0.041984746161752184, "ret_signed_m1d": -0.041984746161752184, "alpha_spy_m1d": -0.04268867695293599, "alpha_c_m1d": -0.052881705841916204, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.01089695968016402, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0114504250256946, "ret_signed_p1d": -0.0114504250256946, "alpha_spy_p1d": -0.004789565643451454, "alpha_c_p1d": -0.005026961425861343, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.006423463599833257, "ret_p1w": 0.37404576151474345, "ret_signed_p1w": 0.37404576151474345, "alpha_spy_p1w": 0.36458256513846243, "alpha_c_p1w": 0.33946214854050205, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.034583612974241396, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5019, -0.5143, -0.5086, -0.4828, -0.4733, -0.5, -0.5086, -0.5105, -0.5153, -0.5592, -0.5267, -0.5491, -0.5682, -0.5806, -0.5286, -0.5172, -0.5219, -0.5191, -0.4971, -0.4905, -0.4809, -0.5115, -0.5382, -0.5086, -0.5262, -0.4714, -0.4962, -0.5086, -0.5095, -0.5067, -0.4361, -0.4351, -0.4351, -0.4351, -0.4418, -0.3922, -0.3893, -0.3645, -0.3674, -0.3406, -0.3254, -0.2863, -0.1918, -0.1508, -0.1479, -0.1584, -0.1498, -0.0697, -0.1231, -0.1613, -0.1307, -0.105, -0.105, -0.0687, -0.0363, -0.0305, -0.042, -0.0401, -0.0191, -0.0716, -0.0267, 0.0019, -0.042, 0.0, -0.0115, 0.1088, 0.1126, 0.2481, 0.374, 0.3359, 0.3263, 0.3378, 0.3454, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2049717500691279979", "ticker_idx": 0, "ts": "2026-04-30T05:08:28+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Google aims to use MediaTek to reduce the 15–20% margin Broadcom captures under its turnkey ASIC model", "tweet_type": "analysis", "timeline": "years", "summary": "Bearish Broadcom as Google/MediaTek partnership erodes its 15-20% ASIC turnkey margin; bullish MediaTek as the beneficiary.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Google is adopting a disaggregated ASIC model through its partnership with MediaTek, aiming to eliminate the 15–20% markup embedded in Broadcom’s turnkey model.\"\n\nIn other words, Google aims to use MediaTek to reduce the 15–20% margin Broadcom captures under its turnkey ASIC model.", "reply_text": null, "parent_text": null, "quoted_text": "𝗠𝗲𝗱𝗶𝗮𝗧𝗲𝗸 𝘁𝗼 𝗖𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗲 𝟭 𝗶𝗻 𝟰 𝗔𝗜 𝗔𝗦𝗜𝗖 𝗦𝗲𝗿𝘃𝗲𝗿 𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗦𝗵𝗶𝗽𝗺𝗲𝗻𝘁𝘀 𝗶𝗻 𝟮𝟬𝟮𝟴\n\n@Google’𝗌 𝗅𝖺𝗍𝖾𝗌𝗍 𝖳𝖯𝖴 𝗏𝟪 𝗀𝖾𝗇𝖾𝗋𝖺𝗍𝗂𝗈𝗇 𝗆𝖺𝗋𝗄𝗌 𝖺 𝗌𝗍𝗋𝖺𝗍𝖾𝗀𝗂𝖼 𝗂𝗇𝖿𝗅𝖾𝖼𝗍𝗂𝗈𝗇 𝗉𝗈𝗂𝗇𝗍, 𝗌𝗁𝗂𝖿𝗍𝗂𝗇𝗀 𝖿𝗋𝗈𝗆 https://t.co/pKNyPhy2fx", "parent_author": null, "ret_m1d": -0.028699376371967733, "ret_signed_m1d": 0.028699376371967733, "alpha_spy_m1d": 0.018847768769962525, "alpha_c_m1d": 0.01460872545778047, "bench_spy_m1d": -0.009851607602005208, "bench_c_m1d": -0.014090650914187264, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0092231180582798, "ret_signed_p1d": -0.0092231180582798, "alpha_spy_p1d": -0.006454004229908827, "alpha_c_p1d": -0.003105328951775954, "bench_spy_p1d": 0.0027691138283709726, "bench_c_p1d": 0.006117789106503846, "ret_p1w": -0.011666615247194367, "ret_signed_p1w": 0.011666615247194367, "alpha_spy_p1w": 0.029644580388038633, "alpha_c_p1w": 0.07754120927312791, "bench_spy_p1w": 0.017977965140844265, "bench_c_p1w": 0.06587459402593354, "ret_p1m": 0.07028722624796879, "ret_signed_p1m": -0.07028722624796879, "alpha_spy_p1m": -0.01766149399974637, "alpha_c_p1m": 0.11168702231719863, "bench_spy_p1m": 0.05262573224822242, "bench_c_p1m": 0.18197424856516742, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2084, -0.2342, -0.2636, -0.2577, -0.2041, -0.1778, -0.1861, -0.1806, -0.2083, -0.2226, -0.2226, -0.205, -0.2027, -0.2016, -0.2048, -0.2103, -0.2219, -0.2056, -0.2309, -0.2361, -0.2378, -0.2497, -0.2409, -0.2045, -0.21, -0.1735, -0.181, -0.1834, -0.1968, -0.2298, -0.2232, -0.2319, -0.2447, -0.2354, -0.2577, -0.2274, -0.2375, -0.2363, -0.2587, -0.2797, -0.2971, -0.2585, -0.249, -0.2465, -0.2465, -0.2467, -0.1999, -0.16, -0.1498, -0.1099, -0.0903, -0.0878, -0.0496, -0.0454, -0.0261, -0.0426, -0.0366, 0.0125, 0.006, 0.0128, 0.0018, -0.0422, -0.0287, 0.0, 0.0092, -0.0022, 0.0238, 0.0192, -0.0117, 0.0301, 0.0264, 0.0045, -0.0015, 0.0536, 0.0186, 0.0079, -0.0152, 0.0008, -0.0069, -0.0079, -0.0079, 0.011, 0.0106, 0.0219, 0.0703, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2049717500691279979", "ticker_idx": 1, "ts": "2026-04-30T05:08:28+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Google is adopting a disaggregated ASIC model through its partnership with MediaTek", "tweet_type": "analysis", "timeline": "years", "summary": "Bearish Broadcom as Google/MediaTek partnership erodes its 15-20% ASIC turnkey margin; bullish MediaTek as the beneficiary.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"Google is adopting a disaggregated ASIC model through its partnership with MediaTek, aiming to eliminate the 15–20% markup embedded in Broadcom’s turnkey model.\"\n\nIn other words, Google aims to use MediaTek to reduce the 15–20% margin Broadcom captures under its turnkey ASIC model.", "reply_text": null, "parent_text": null, "quoted_text": "𝗠𝗲𝗱𝗶𝗮𝗧𝗲𝗸 𝘁𝗼 𝗖𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗲 𝟭 𝗶𝗻 𝟰 𝗔𝗜 𝗔𝗦𝗜𝗖 𝗦𝗲𝗿𝘃𝗲𝗿 𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗦𝗵𝗶𝗽𝗺𝗲𝗻𝘁𝘀 𝗶𝗻 𝟮𝟬𝟮𝟴\n\n@Google’𝗌 𝗅𝖺𝗍𝖾𝗌𝗍 𝖳𝖯𝖴 𝗏𝟪 𝗀𝖾𝗇𝖾𝗋𝖺𝗍𝗂𝗈𝗇 𝗆𝖺𝗋𝗄𝗌 𝖺 𝗌𝗍𝗋𝖺𝗍𝖾𝗀𝗂𝖼 𝗂𝗇𝖿𝗅𝖾𝖼𝗍𝗂𝗈𝗇 𝗉𝗈𝗂𝗇𝗍, 𝗌𝗁𝗂𝖿𝗍𝗂𝗇𝗀 𝖿𝗋𝗈𝗆 https://t.co/pKNyPhy2fx", "parent_author": null, "ret_m1d": -0.013409961685823757, "ret_signed_m1d": -0.013409961685823757, "alpha_spy_m1d": -0.003558354083818549, "alpha_c_m1d": 0.0006806892283635069, "bench_spy_m1d": -0.009851607602005208, "bench_c_m1d": -0.014090650914187264, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0027691138283709726, "alpha_c_p1d": -0.006117789106503846, "bench_spy_p1d": 0.0027691138283709726, "bench_c_p1d": 0.006117789106503846, "ret_p1w": 0.31034482758620685, "ret_signed_p1w": 0.31034482758620685, "alpha_spy_p1w": 0.2923668624453626, "alpha_c_p1w": 0.2444702335602733, "bench_spy_p1w": 0.017977965140844265, "bench_c_p1w": 0.06587459402593354, "ret_p1m": 0.6513409961685823, "ret_signed_p1m": 0.6513409961685823, "alpha_spy_p1m": 0.5987152639203599, "alpha_c_p1m": 0.4693667476034149, "bench_spy_p1m": 0.05262573224822242, "bench_c_p1m": 0.18197424856516742, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3467, -0.3123, -0.3103, -0.3218, -0.3448, -0.2989, -0.2931, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.2893, -0.3123, -0.3008, -0.2835, -0.2548, -0.2548, -0.272, -0.3046, -0.341, -0.3199, -0.3238, -0.3621, -0.3467, -0.3238, -0.3161, -0.341, -0.3448, -0.3372, -0.3372, -0.3563, -0.3487, -0.3774, -0.3793, -0.3793, -0.3908, -0.3927, -0.4215, -0.4291, -0.4387, -0.4387, -0.4387, -0.4387, -0.4368, -0.3946, -0.3966, -0.3966, -0.3793, -0.341, -0.3142, -0.2739, -0.2625, -0.272, -0.1992, -0.1207, -0.1513, -0.067, -0.067, 0.0019, -0.0134, 0.0, 0.0, 0.0996, 0.2088, 0.3142, 0.3103, 0.3908, 0.4866, 0.4176, 0.3391, 0.3046, 0.249, 0.3027, 0.2088, 0.2375, 0.3602, 0.4789, 0.6264, 0.6341, 0.7778, 0.6897, 0.6513, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2059895609642127856", "ticker_idx": 0, "ts": "2026-05-28T07:12:39+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This packaging investment is evidence that Intel has secured a certain degree of advanced foundry process customers… The 18A process combined with expanded advanced packaging investment could be the starting signal for Intel's foundry revival.", "tweet_type": "analysis", "timeline": "months", "summary": "Bullish INTC: large-scale EMIB packaging investment + 18A ramp signals foundry revival with secured customers.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Intel Accelerates Foundry Revival, Spearheaded by Advanced Packaging\n\nIntel is accelerating the revival of its foundry (contract manufacturing) business through large-scale investment in advanced semiconductor packaging. By expanding infrastructure such as materials, components, and equipment, Intel is significantly scaling up its advanced packaging capacity to strengthen its response to the foundry customers it has recently secured.\n\nAccording to industry sources on the 28th, Intel is currently placing large-scale purchase orders for materials, components, and equipment from its global supply chain partners. Multiple supply contracts for materials, components, and equipment have reportedly already been signed. Some Korean materials/components/equipment companies are also participating.\n\nAn industry official familiar with the matter said, \"Advanced packaging facility investment on the scale of several trillion won is underway,\" adding, \"Considering equipment lead times and the timing of materials and component supply, full-scale operation is expected next year.\" Intel is also discussing cooperation with some partners regarding facility investment in 2028. The investment locations cited as major packaging production hubs include the United States as well as Vietnam and Malaysia.\n\nThis investment is said to be focused on expanding capacity for \"EMIB (Embedded Multi-die Interconnect Bridge).\"\n\nEMIB is Intel's proprietary 2.5D packaging technology that connects different semiconductors (dies). Signals are exchanged between chips via a silicon bridge embedded in the semiconductor substrate, and it is known to offer superior cost and productivity compared to the silicon interposer-based 2.5D packaging led by TSMC. In particular, it enables precise 2.5D packaging by connecting only the necessary areas with bridges.\n\nIntel is advancing EMIB further with technologies such as EMIB-T, which applies through-silicon via (TSV) technology to the bridge, and packaging that integrates glass substrates. The aim is to broaden its customer base by diversifying its packaging technologies. This materials/components/equipment investment also includes numerous new technologies to be applied to next-generation packaging.\n\nThis packaging investment is a strategic move to strengthen foundry capabilities. Intel declared its re-entry into the foundry business in 2021 and pursued the development of advanced (front-end) processes, but it has struggled to secure major customers. This was largely due to TSMC's dominance in advanced chip foundry, including AI semiconductors.\n\nThe packaging investment is also interpreted as a strategy to strengthen Intel's differentiated capabilities to attract foundry customers. Because the limits of circuit miniaturization mean the entire industry faces constraints in boosting chip performance through front-end processes, Intel is seen as attempting to break through this bottleneck with advanced packaging (back-end) and place its entire foundry business on a stable trajectory.\n\nIntel is also pursuing front-end advancement in parallel to maximize synergy. Last year, Intel brought its \"Intel 18A\" process—equivalent to the 2-nanometer (㎚) class—into full-scale operation. It has continued facility investment not only for its own chips but also to serve foundry customers.\n\nAnother industry official predicted, \"This packaging investment is evidence that Intel has secured a certain degree of advanced foundry process customers,\" and \"The 18A process combined with expanded advanced packaging investment could be the starting signal for Intel's foundry revival.\"\n\n$INTC\n\n---\n\nhttps://t.co/bbp6JYEGBB", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007279322176035441, "ret_signed_m1d": 0.007279322176035441, "alpha_spy_m1d": 0.012765611265916021, "alpha_c_m1d": 0.014498062429901704, "bench_spy_m1d": -0.0054862890898805805, "bench_c_m1d": -0.007218740253866263, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05136900583857895, "ret_signed_p1d": -0.05136900583857895, "alpha_spy_p1d": -0.05386039855476965, "alpha_c_p1d": -0.049868540059241795, "bench_spy_p1d": 0.002491392716190699, "bench_c_p1d": -0.0015004657793371567, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6236, -0.6435, -0.623, -0.6199, -0.6408, -0.623, -0.613, -0.6031, -0.6257, -0.6214, -0.6215, -0.6355, -0.6275, -0.618, -0.6371, -0.636, -0.6355, -0.6097, -0.6352, -0.6432, -0.6593, -0.635, -0.6027, -0.5833, -0.5833, -0.5799, -0.5623, -0.5124, -0.4895, -0.484, -0.4608, -0.4722, -0.4628, -0.4334, -0.4334, -0.4565, -0.4519, -0.4601, -0.4476, -0.3172, -0.297, -0.3009, -0.2162, -0.2185, -0.1759, -0.2077, -0.1054, -0.0652, -0.0932, 0.0333, 0.0707, -0.0023, -0.005, -0.041, -0.1003, -0.1052, -0.0835, -0.016, -0.0198, -0.0087, -0.0087, 0.0218, 0.0073, 0.0, -0.0514, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2044624197885051287", "ticker_idx": 0, "ts": "2026-04-16T03:49:30+00:00", "symbol": "Android smartphones", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The smartphone situation is only going to get worse — it certainly won't get any better. I personally think this target is overly optimistic as well.", "tweet_type": "prediction", "timeline": "quarters", "summary": "Bearish Android smartphone sector; Xiaomi CFO's -10% YoY revenue floor seen as too optimistic, expects worse declines.", "resolved_tickers": ["QCOM", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Android phone sector proxy: Qualcomm + Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "The smartphone situation is only going to get worse — it certainly won't get any better.\n\nSomeone from the @TMTBreakout Chat mentioned that they met Xiaomi's CFO last week.\n\nAccording to them, the CFO said:\n\"Our goal is to make sure that the smartphone division's revenue doesn't fall more than -10% (YoY basis).\"\n\nI personally think this target is overly optimistic as well.", "reply_text": null, "parent_text": null, "quoted_text": "Long Memory, Short [Android] Smartphone https://t.co/5QvZWX1SXB https://t.co/ZwrGvmbs0b", "parent_author": null, "ret_m1d": -0.020222509980723002, "ret_signed_m1d": 0.020222509980723002, "alpha_spy_m1d": 0.017771221936719483, "alpha_c_m1d": 0.01626475307377223, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0029843697534377434, "ret_signed_p1d": -0.0029843697534377434, "alpha_spy_p1d": 0.009101315376786268, "alpha_c_p1d": 0.0175961406255587, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": 0.014158421384703357, "ret_signed_p1w": -0.014158421384703357, "alpha_spy_p1w": -0.004481313779334362, "alpha_c_p1w": 0.04531831352422644, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.37103965926567883, "ret_signed_p1m": -0.37103965926567883, "alpha_spy_p1m": -0.31758070307423236, "alpha_c_p1m": -0.1477765946485301, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0685, -0.0977, -0.0793, -0.0676, -0.0754, -0.0802, -0.0687, -0.0633, -0.0689, -0.0717, -0.0911, -0.072, -0.062, -0.131, -0.1287, -0.105, -0.1021, -0.094, -0.0787, -0.0645, -0.0645, -0.0574, -0.0551, -0.0422, -0.036, -0.0385, -0.0063, 0.0056, 0.038, 0.0226, 0.0177, -0.0421, -0.0895, -0.051, -0.0637, -0.0884, -0.0662, -0.0654, -0.0813, -0.0963, -0.086, -0.0659, -0.0366, -0.0519, -0.0595, -0.0954, -0.0864, -0.0817, -0.1014, -0.1142, -0.1222, -0.1368, -0.0908, -0.1184, -0.1005, -0.0886, -0.0869, -0.042, -0.056, -0.0503, -0.0499, -0.0313, -0.0202, 0.0, 0.003, 0.0044, 0.0075, 0.0059, 0.0142, 0.0581, 0.0748, 0.0681, 0.0996, 0.1746, 0.1651, 0.1606, 0.2281, 0.3275, 0.3773, 0.4319, 0.5395, 0.4234, 0.4455, 0.4244, 0.371, 0.4032, 0.3607, 0.3875, 0.482, 0.558, 0.558, 0.6125, 0.5736, 0.5931, 0.6621, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2039733794710896979", "ticker_idx": 0, "ts": "2026-04-02T15:56:47+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Long $ASML", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long ASML on surging DUV demand, with CXMT bidding up reserved slots.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "ASML’s DUV demand is apparently going through the roof, and CXMT is said to be buying up other companies’ reserved DUV slots at higher prices.\n\nInsane. Long $ASML", "reply_text": null, "parent_text": null, "quoted_text": "ASML的DUV需求完全爆表了，长鑫在到处加价“收条子”——之前定机器的\n\n要知道一年前市场还在担心DUV受中国+ intc影响呢\n\n有种感觉半导体研究没意义了，token指数级需求增长，供不应求才是常态，大家都没涨到位", "parent_author": null, "ret_m1d": 0.022911343479107504, "ret_signed_m1d": 0.022911343479107504, "alpha_spy_m1d": 0.02381100778249956, "alpha_c_m1d": 0.023803487894333286, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.0008921444152257818, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.06356602494771302, "ret_signed_p1w": 0.06356602494771302, "alpha_spy_p1w": 0.026849260805780206, "alpha_c_p1w": -0.033268165318243836, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.09683419026595685, "ret_p1m": 0.055213590661594036, "ret_signed_p1m": 0.055213590661594036, "alpha_spy_p1m": -0.04362300668088315, "alpha_c_p1m": -0.24428681157737797, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.299500402238972, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0941, -0.0874, -0.0953, -0.1286, -0.0696, -0.0655, -0.0516, -0.0674, -0.0113, 0.004, -0.0363, -0.0194, -0.0067, 0.0119, 0.0135, 0.0133, 0.0473, 0.0274, 0.0253, 0.0456, 0.0535, 0.0239, -0.0189, -0.0111, 0.0269, 0.0377, 0.0276, 0.0403, 0.0162, 0.0253, 0.0293, 0.0329, 0.0722, 0.0665, 0.0815, 0.076, 0.0882, 0.1097, 0.0615, 0.0624, 0.0425, 0.0007, 0.0334, 0.0215, -0.0121, -0.0115, 0.0336, 0.0326, 0.0167, 0.016, 0.0314, 0.0283, 0.0258, 0.0065, -0.0283, 0.0124, 0.035, 0.0436, 0.0041, -0.0126, -0.0422, -0.036, 0.0229, 0.0, 0.0, 0.0, -0.0407, 0.0444, 0.0636, 0.0939, 0.0851, 0.1061, 0.0594, 0.0531, 0.0715, 0.0724, 0.0655, 0.0765, 0.0541, 0.0787, 0.0468, 0.0115, 0.0295, 0.0552, 0.0552, 0.025, 0.0609, 0.1265, 0.1222, 0.1389, 0.127, 0.0927, 0.1455, 0.18, 0.1279, 0.0915, 0.0782, 0.1507, 0.1612, 0.2163, 0.2349, 0.1997, 0.1882, 0.2008, 0.1954, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2038802510346158513", "ticker_idx": 0, "ts": "2026-03-31T02:16:12+00:00", "symbol": "Samsung Electro-Mechanics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "FC-BGA is seeing a sustained sold-out trend from the second half onward in a demand environment where additional capacity expansion is inevitable… With a 10% upward revision to our FC-BGA price estimates, we are raising our substrate materials division operating profit forecasts for 2026 and 2027 to KRW 386.1 billion and KRW 504.6 billion, respectively—up 185% and 41% year-over-year", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Mirae Asset Buy thesis on Samsung Electro-Mechanics: FC-BGA price hikes + sold-out demand drive 185%/41% OP forecast upgrades for 2026/2027.", "resolved_tickers": ["009150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electro-Mechanics listed on KOSPI as 009150.KS", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electro-Mechanics Raises FC-BGA Prices… 185% Operating Profit Growth Forecast for This Year\n\nSamsung Electro-Mechanics has recently raised prices on its flip-chip ball grid array (FC-BGA), a high-performance semiconductor substrate. Market expectations for earnings improvement and mid-to-long-term growth are rising accordingly.\n\nAccording to industry sources on March 31, Samsung Electro-Mechanics is reported to have raised prices on select FC-BGA product lines. The move is interpreted as a measure to defend profitability amid rising raw material costs for FC-BGA.\n\nAdditionally, as demand surges rapidly for these components—essential to AI servers and high-performance computing—suppliers' pricing power has strengthened, which is another driver behind the price hike.\n\nThe decision to raise substrate prices, underpinned by rising raw material costs and enhanced bargaining power, is expected to have a positive impact on both revenue and operating profit.\n\nFC-BGA has recently been experiencing near-\"sold out\" levels of scarcity. Samsung Electro-Mechanics recently achieved the milestone of supplying FC-BGA to NVIDIA.\n\nThe structural shift in the semiconductor industry toward high-end compute devices such as CPUs and GPUs is also driving mid-to-long-term growth in the FC-BGA market.\n\nFC-BGA is an essential substrate for packaging top-tier semiconductors like CPUs and GPUs, and demand is surging alongside the proliferation of AI servers.\n\nHowever, a severe \"supply bottleneck\" persists. Production capacity for advanced package substrates has failed to keep pace with rapidly growing global demand, and the substrate shortage is now constraining the pace of AI server supply chain expansion.\n\nAt a recent shareholders' meeting, Samsung Electro-Mechanics CEO Jang Deok-hyun stated: \"Demand for server and data center FC-BGA exceeds production capacity by more than 50%,\" adding that \"we are making supplementary investments and expanding some facilities.\"\n\nAccording to market research firm Prismark, FC-BGA is projected to record a CAGR of over 15% from 2025 to 2030, the highest growth rate among all PCB product categories.\n\nEven as suppliers like Nittobo expand production, demand growth is outpacing capacity additions, and the surge in demand for low-loss materials used in AI high-performance substrates is creating supply chain constraints.\n\nThe sell-side also views Samsung Electro-Mechanics as entering a phase of full-scale profitability improvement driven by FC-BGA price increases.\n\nIn a report dated March 31, Mirae Asset analyst Park Jun-seo wrote: \"FC-BGA is seeing a sustained sold-out trend from the second half onward in a demand environment where additional capacity expansion is inevitable,\" and noted that \"negotiations for further price increases are understood to be progressing favorably.\" He added: \"With a 10% upward revision to our FC-BGA price estimates, we are raising our substrate materials division operating profit forecasts for 2026 and 2027 to KRW 386.1 billion and KRW 504.6 billion, respectively—up 185% and 41% year-over-year.\"\n\nPrices for high-end MLCCs (multilayer ceramic capacitors) used in AI servers are also being raised as supply tightens amid growing demand. Mirae Asset Securities estimates that a 15–25% increase in AI server MLCC prices could generate hundreds of billions of won in additional operating profit.\n\nSamsung Electro-Mechanics is restructuring its portfolio around high-end products such as FC-BGA and MLCC. It is also expanding supply capacity through facility expansion at its Philippine production base. In the aerospace sector, the company is building a new growth pillar by expanding supply of high-reliability MLCCs for satellites and aviation. As a result, Samsung Electro-Mechanics posted record-high revenue of KRW 11.3145 trillion last year.\n\n---\n\nhttps://t.co/EQ0MErnRQH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.052760736196318936, "ret_signed_m1d": 0.052760736196318936, "alpha_spy_m1d": 0.08100758460447677, "alpha_c_m1d": 0.09339274631281613, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.0406320101164972, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08957055214723919, "ret_signed_p1d": 0.08957055214723919, "alpha_spy_p1d": 0.08203608836434917, "alpha_c_p1d": 0.07444632447277755, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.015124227674461643, "ret_p1w": 0.12147239263803677, "ret_signed_p1w": 0.12147239263803677, "alpha_spy_p1w": 0.1078180833829343, "alpha_c_p1w": 0.08738662147671317, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.03408577116132361, "ret_p1m": 1.029447852760736, "ret_signed_p1m": 1.029447852760736, "alpha_spy_p1m": 0.935281743178825, "alpha_c_p1m": 0.832231841341788, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.19721601141894807, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3742, -0.3374, -0.3374, -0.3423, -0.3436, -0.3325, -0.3472, -0.3153, -0.2908, -0.3018, -0.2896, -0.2896, -0.2834, -0.308, -0.3129, -0.3055, -0.3325, -0.3252, -0.3202, -0.3129, -0.3141, -0.3153, -0.3055, -0.2454, -0.2515, -0.2883, -0.3031, -0.254, -0.2307, -0.2405, -0.216, -0.2405, -0.2405, -0.2405, -0.2405, -0.1215, -0.0748, 0.0466, 0.0945, 0.1411, 0.1448, 0.1006, 0.1006, 0.0037, -0.1104, -0.0147, -0.0074, -0.0626, -0.0135, -0.0037, -0.0196, -0.0172, 0.0135, 0.0957, 0.1387, 0.1767, 0.1399, 0.0491, 0.0724, 0.1153, 0.1215, 0.1215, 0.0528, 0.0, 0.0896, 0.0233, 0.119, 0.1337, 0.1215, 0.2613, 0.2663, 0.3865, 0.3939, 0.4331, 0.5067, 0.5681, 0.6663, 0.6687, 0.8945, 0.9926, 0.8994, 0.9337, 0.9485, 1.0589, 1.0294, 1.0417, 1.0417, 1.2528, 1.2528, 1.238, 1.2503, 1.2429, 1.2086, 1.3509, 1.5252, 1.5129, 1.4785, 1.5301, 1.4221, 1.6037, 1.9546, 2.2883, 2.2883, 2.8577, 3.0, 3.5374, 4.2196, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2056527538944713101", "ticker_idx": 0, "ts": "2026-05-19T00:09:08+00:00", "symbol": "Elite Material", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Elite Material is currently facing a shortage of its highest-end M7 and M8 materials, with supply unable to keep up with demand", "tweet_type": "analysis", "timeline": "months", "summary": "Severe glass fiber shortage drives spillover demand for Taiwanese CCL/glass fiber makers—Elite Material, FULLTECH FIBER, Baotek, and Grandsys—ahead of NVIDIA Rubin ramp in 2H26.", "resolved_tickers": ["2383.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Elite Material Co. 2383.TW Taiwan CCL", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe glass fiber shortage drives spillover benefits for Taiwanese glass fiber makers\n\n• Ahead of the mass production ramp of NVIDIA’s Rubin platform in 2H26, AI servers and high-speed switches are being upgraded from the existing 112G generation to the 224G PAM4 high-speed transmission generation. As a result, demand is surging for high-performance Low-Dk and Low-CTE materials that reduce signal loss and thermal deformation.\n\n• Taiwan’s major CCL supplier Elite Material is currently facing a shortage of its highest-end M7 and M8 materials, with supply unable to keep up with demand. The company plans to begin early shipments of next-generation M9-grade materials from 3Q this year, making 2027 likely the first year of full-scale mass production for high-end materials.\n\n• The high-end Low-CTE glass fiber market is tightly supplied, as Japanese companies such as Nittobo effectively dominate the segment. This is creating significant spillover benefits for Taiwanese glass fiber manufacturers. FULLTECH FIBER is entering the AI server and high-speed switch supply chain with Low-Dk and Low-CTE products, while Baotek is expanding its high-strength glass fiber, or T-Glass, portfolio. Grandsys is benefiting from price increases in general-purpose products and is developing next-generation new materials, including Q-cloth and Q-glass.\n\n---\n\nhttps://t.co/dSyBdtAJbc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0022050716648291946, "ret_signed_m1d": 0.0022050716648291946, "alpha_spy_m1d": -0.004500452269424304, "alpha_c_m1d": -0.0042599195708799, "bench_spy_m1d": 0.006705523934253499, "bench_c_m1d": 0.006464991235709094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.030871003307607503, "ret_signed_p1d": -0.030871003307607503, "alpha_spy_p1d": -0.04112003187591784, "alpha_c_p1d": -0.05338308927402291, "bench_spy_p1d": 0.010249028568310337, "bench_c_p1d": 0.022512085966415407, "ret_p1w": 0.16648291069459753, "ret_signed_p1w": 0.16648291069459753, "alpha_spy_p1w": 0.14350436704023095, "alpha_c_p1w": 0.09779211457848946, "bench_spy_p1w": 0.02297854365436658, "bench_c_p1w": 0.06869079611610807, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.516, -0.516, -0.5204, -0.4906, -0.4664, -0.462, -0.462, -0.4752, -0.4884, -0.5171, -0.4785, -0.4796, -0.5204, -0.4763, -0.4598, -0.4267, -0.4035, -0.43, -0.4046, -0.4046, -0.3627, -0.3804, -0.3837, -0.4079, -0.3727, -0.3506, -0.3583, -0.3859, -0.4267, -0.3693, -0.3859, -0.3859, -0.3859, -0.3638, -0.301, -0.29, -0.2459, -0.2668, -0.2051, -0.1599, -0.1544, -0.1621, -0.151, -0.129, -0.1036, -0.0959, -0.0132, 0.043, -0.0353, -0.0209, 0.0022, 0.0022, 0.0485, 0.043, 0.0915, 0.0937, 0.0474, 0.0695, 0.0805, 0.0871, 0.0849, 0.0176, 0.0022, 0.0, -0.0309, 0.0342, 0.1036, 0.2139, 0.1665, 0.1698, 0.1147, 0.129, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2056527538944713101", "ticker_idx": 1, "ts": "2026-05-19T00:09:08+00:00", "symbol": "FULLTECH FIBER", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "FULLTECH FIBER is entering the AI server and high-speed switch supply chain with Low-Dk and Low-CTE products", "tweet_type": "analysis", "timeline": "months", "summary": "Severe glass fiber shortage drives spillover demand for Taiwanese CCL/glass fiber makers—Elite Material, FULLTECH FIBER, Baotek, and Grandsys—ahead of NVIDIA Rubin ramp in 2H26.", "resolved_tickers": ["3035.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Fulltech Fiber Glass 3035.TW Taiwan glass fiber materials", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe glass fiber shortage drives spillover benefits for Taiwanese glass fiber makers\n\n• Ahead of the mass production ramp of NVIDIA’s Rubin platform in 2H26, AI servers and high-speed switches are being upgraded from the existing 112G generation to the 224G PAM4 high-speed transmission generation. As a result, demand is surging for high-performance Low-Dk and Low-CTE materials that reduce signal loss and thermal deformation.\n\n• Taiwan’s major CCL supplier Elite Material is currently facing a shortage of its highest-end M7 and M8 materials, with supply unable to keep up with demand. The company plans to begin early shipments of next-generation M9-grade materials from 3Q this year, making 2027 likely the first year of full-scale mass production for high-end materials.\n\n• The high-end Low-CTE glass fiber market is tightly supplied, as Japanese companies such as Nittobo effectively dominate the segment. This is creating significant spillover benefits for Taiwanese glass fiber manufacturers. FULLTECH FIBER is entering the AI server and high-speed switch supply chain with Low-Dk and Low-CTE products, while Baotek is expanding its high-strength glass fiber, or T-Glass, portfolio. Grandsys is benefiting from price increases in general-purpose products and is developing next-generation new materials, including Q-cloth and Q-glass.\n\n---\n\nhttps://t.co/dSyBdtAJbc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.023872679045092937, "ret_signed_m1d": 0.023872679045092937, "alpha_spy_m1d": 0.017167155110839438, "alpha_c_m1d": 0.0198283539421249, "bench_spy_m1d": 0.006705523934253499, "bench_c_m1d": 0.004044325102968038, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0053050397877985045, "ret_signed_p1d": 0.0053050397877985045, "alpha_spy_p1d": -0.004943988780511832, "alpha_c_p1d": -0.032749137607090084, "bench_spy_p1d": 0.010249028568310337, "bench_c_p1d": 0.03805417739488859, "ret_p1w": 0.10344827586206895, "ret_signed_p1w": 0.10344827586206895, "alpha_spy_p1w": 0.08046973220770237, "alpha_c_p1w": -0.00350810020540937, "bench_spy_p1w": 0.02297854365436658, "bench_c_p1w": 0.10695637606747832, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1406, -0.1406, -0.1167, -0.1167, -0.1114, -0.1088, -0.1088, -0.1167, -0.1273, -0.1618, -0.1565, -0.1592, -0.2361, -0.2255, -0.183, -0.1857, -0.1963, -0.1751, -0.1565, -0.13, -0.1485, -0.1645, -0.1989, -0.2149, -0.1963, -0.1989, -0.2016, -0.2228, -0.252, -0.2334, -0.2626, -0.2626, -0.2626, -0.2573, -0.2016, -0.1936, -0.2069, -0.2095, -0.1883, -0.2069, -0.1698, -0.1565, -0.0743, -0.0451, -0.0504, -0.1061, -0.0186, -0.0106, -0.0637, -0.1061, -0.1088, -0.1088, -0.1034, -0.061, -0.0371, -0.0398, -0.0557, 0.0371, 0.1061, 0.0292, 0.0531, 0.008, 0.0239, 0.0, 0.0053, 0.1034, 0.1194, 0.1592, 0.1034, 0.0743, 0.1061, 0.0981, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2056527538944713101", "ticker_idx": 2, "ts": "2026-05-19T00:09:08+00:00", "symbol": "Baotek", "kind": "equity", "geography": "Taiwan", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Baotek is expanding its high-strength glass fiber, or T-Glass, portfolio", "tweet_type": "analysis", "timeline": "months", "summary": "Severe glass fiber shortage drives spillover demand for Taiwanese CCL/glass fiber makers—Elite Material, FULLTECH FIBER, Baotek, and Grandsys—ahead of NVIDIA Rubin ramp in 2H26.", "resolved_tickers": ["3051.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Baotek Industrial Materials 3051.TW Taiwan glass fiber", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Severe glass fiber shortage drives spillover benefits for Taiwanese glass fiber makers\n\n• Ahead of the mass production ramp of NVIDIA’s Rubin platform in 2H26, AI servers and high-speed switches are being upgraded from the existing 112G generation to the 224G PAM4 high-speed transmission generation. As a result, demand is surging for high-performance Low-Dk and Low-CTE materials that reduce signal loss and thermal deformation.\n\n• Taiwan’s major CCL supplier Elite Material is currently facing a shortage of its highest-end M7 and M8 materials, with supply unable to keep up with demand. The company plans to begin early shipments of next-generation M9-grade materials from 3Q this year, making 2027 likely the first year of full-scale mass production for high-end materials.\n\n• The high-end Low-CTE glass fiber market is tightly supplied, as Japanese companies such as Nittobo effectively dominate the segment. This is creating significant spillover benefits for Taiwanese glass fiber manufacturers. FULLTECH FIBER is entering the AI server and high-speed switch supply chain with Low-Dk and Low-CTE products, while Baotek is expanding its high-strength glass fiber, or T-Glass, portfolio. Grandsys is benefiting from price increases in general-purpose products and is developing next-generation new materials, including Q-cloth and Q-glass.\n\n---\n\nhttps://t.co/dSyBdtAJbc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009960159211175323, "ret_signed_m1d": -0.009960159211175323, "alpha_spy_m1d": -0.016665683145428822, "alpha_c_m1d": -0.016425150446884418, "bench_spy_m1d": 0.006705523934253499, "bench_c_m1d": 0.006464991235709094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009960159211175323, "ret_signed_p1d": -0.009960159211175323, "alpha_spy_p1d": -0.02020918777948566, "alpha_c_p1d": -0.03247224517759073, "bench_spy_p1d": 0.010249028568310337, "bench_c_p1d": 0.022512085966415407, "ret_p1w": 0.12350598941657087, "ret_signed_p1w": 0.12350598941657087, "alpha_spy_p1w": 0.10052744576220429, "alpha_c_p1w": 0.054815193300462806, "bench_spy_p1w": 0.02297854365436658, "bench_c_p1w": 0.06869079611610807, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1175, -0.1175, -0.0956, -0.0777, -0.0717, -0.0777, -0.0777, -0.0817, -0.1155, -0.1693, -0.1633, -0.1375, -0.1932, -0.1912, -0.1614, -0.1394, -0.1514, -0.1414, -0.1315, -0.1434, -0.1713, -0.1494, -0.1952, -0.1952, -0.1713, -0.1912, -0.1793, -0.1992, -0.2151, -0.1992, -0.2072, -0.2072, -0.2072, -0.2131, -0.1673, -0.1793, -0.1992, -0.1992, -0.1793, -0.1793, -0.1733, -0.1813, -0.0996, -0.01, 0.004, -0.0777, -0.0777, -0.0398, 0.0558, 0.0936, 0.1295, 0.1295, 0.1713, 0.1892, 0.0876, 0.0817, 0.0359, 0.0498, 0.0359, 0.0159, 0.0518, 0.0179, -0.01, 0.0, -0.01, 0.0219, 0.1235, 0.2271, 0.1235, 0.0857, 0.0797, 0.1116, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039116249700143544", "ticker_idx": 0, "ts": "2026-03-31T23:02:53+00:00", "symbol": "Samsung Electro-Mechanics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "High-capacity MLCCs emerging as the alternative, enabling both performance compliance and multi-sourcing from suppliers such as Murata and Samsung Electro-Mechanics", "tweet_type": "analysis", "timeline": "quarters", "summary": "MLCC suppliers Murata and Samsung Electro-Mechanics named as beneficiaries of Nvidia's Vera Rubin shift away from Panasonic SP-CAPs toward multi-sourced high-cap MLCCs, with demand exceeding supply.", "resolved_tickers": ["009150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electro-Mechanics listed on KOSPI as 009150.KS", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Vera Rubin Compute Tray Design Unfinalized as Nvidia Pushes Supply Diversification\n\n- Nvidia's next-gen platform Vera Rubin is slated for mass production in Q3 2026, though the GPU compute tray design remains unfinalized, raising the possibility of partial shipment delays.\n\n- Supply chain diversification strategy intensified to mitigate component shortage risks — ongoing design revisions driven by the shift toward multi-vendor structures and elimination of single-source dependencies.\n\n- As a result, Panasonic SP-CAPs (polymer aluminum capacitors) will not be adopted in the Vera Rubin compute tray; the high supplier concentration inherent in that component conflicted with the broader supply diversification strategy.\n\n- High-capacity MLCCs emerging as the alternative, enabling both performance compliance and multi-sourcing from suppliers such as Murata and Samsung Electro-Mechanics.\n\n- AI server racks require approximately 15,000–25,000 MLCCs per unit; rising ASPs and increasing content per system are positioning MLCCs as a major BOM cost item alongside GPUs and memory.\n\n- On the Nvidia GB300 platform, up to 30,000 MLCCs are needed per system — roughly 3x an automobile, 10x a mining rig, and 30x a smartphone.\n\n- Murata reports high-spec MLCC demand exceeding production capacity, with lead times extending from 10–12 weeks to 14–16 weeks, signaling tightening supply conditions.\n\n- The Vera Rubin platform's elevated spec requirements and design complexity mean that not only supplier technical capability but also speed of response to design changes is becoming a key competitive differentiator.\n\n- Despite being in the pre-mass production stage, portions of the BOM remain unfinalized in terms of component specs and supply structure, with final adjustments still underway.\n\n- Near-term shipment timelines may be impacted, but longer-term this is viewed as a necessary process to strengthen supply chain resilience.\n\n---\n\nhttps://t.co/wzHPK5QUDc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.052760736196318936, "ret_signed_m1d": 0.052760736196318936, "alpha_spy_m1d": 0.08100758460447677, "alpha_c_m1d": 0.09339274631281613, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.0406320101164972, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08957055214723919, "ret_signed_p1d": 0.08957055214723919, "alpha_spy_p1d": 0.08203608836434917, "alpha_c_p1d": 0.07444632447277755, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.015124227674461643, "ret_p1w": 0.12147239263803677, "ret_signed_p1w": 0.12147239263803677, "alpha_spy_p1w": 0.1078180833829343, "alpha_c_p1w": 0.08738662147671317, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.03408577116132361, "ret_p1m": 1.029447852760736, "ret_signed_p1m": 1.029447852760736, "alpha_spy_p1m": 0.935281743178825, "alpha_c_p1m": 0.832231841341788, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.19721601141894807, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3742, -0.3374, -0.3374, -0.3423, -0.3436, -0.3325, -0.3472, -0.3153, -0.2908, -0.3018, -0.2896, -0.2896, -0.2834, -0.308, -0.3129, -0.3055, -0.3325, -0.3252, -0.3202, -0.3129, -0.3141, -0.3153, -0.3055, -0.2454, -0.2515, -0.2883, -0.3031, -0.254, -0.2307, -0.2405, -0.216, -0.2405, -0.2405, -0.2405, -0.2405, -0.1215, -0.0748, 0.0466, 0.0945, 0.1411, 0.1448, 0.1006, 0.1006, 0.0037, -0.1104, -0.0147, -0.0074, -0.0626, -0.0135, -0.0037, -0.0196, -0.0172, 0.0135, 0.0957, 0.1387, 0.1767, 0.1399, 0.0491, 0.0724, 0.1153, 0.1215, 0.1215, 0.0528, 0.0, 0.0896, 0.0233, 0.119, 0.1337, 0.1215, 0.2613, 0.2663, 0.3865, 0.3939, 0.4331, 0.5067, 0.5681, 0.6663, 0.6687, 0.8945, 0.9926, 0.8994, 0.9337, 0.9485, 1.0589, 1.0294, 1.0417, 1.0417, 1.2528, 1.2528, 1.238, 1.2503, 1.2429, 1.2086, 1.3509, 1.5252, 1.5129, 1.4785, 1.5301, 1.4221, 1.6037, 1.9546, 2.2883, 2.2883, 2.8577, 3.0, 3.5374, 4.2196, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039116249700143544", "ticker_idx": 1, "ts": "2026-03-31T23:02:53+00:00", "symbol": "Murata", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Murata reports high-spec MLCC demand exceeding production capacity, with lead times extending from 10–12 weeks to 14–16 weeks, signaling tightening supply conditions", "tweet_type": "analysis", "timeline": "quarters", "summary": "MLCC suppliers Murata and Samsung Electro-Mechanics named as beneficiaries of Nvidia's Vera Rubin shift away from Panasonic SP-CAPs toward multi-sourced high-cap MLCCs, with demand exceeding supply.", "resolved_tickers": ["6981.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Murata Manufacturing Co., listed on Tokyo Stock Exchange", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Vera Rubin Compute Tray Design Unfinalized as Nvidia Pushes Supply Diversification\n\n- Nvidia's next-gen platform Vera Rubin is slated for mass production in Q3 2026, though the GPU compute tray design remains unfinalized, raising the possibility of partial shipment delays.\n\n- Supply chain diversification strategy intensified to mitigate component shortage risks — ongoing design revisions driven by the shift toward multi-vendor structures and elimination of single-source dependencies.\n\n- As a result, Panasonic SP-CAPs (polymer aluminum capacitors) will not be adopted in the Vera Rubin compute tray; the high supplier concentration inherent in that component conflicted with the broader supply diversification strategy.\n\n- High-capacity MLCCs emerging as the alternative, enabling both performance compliance and multi-sourcing from suppliers such as Murata and Samsung Electro-Mechanics.\n\n- AI server racks require approximately 15,000–25,000 MLCCs per unit; rising ASPs and increasing content per system are positioning MLCCs as a major BOM cost item alongside GPUs and memory.\n\n- On the Nvidia GB300 platform, up to 30,000 MLCCs are needed per system — roughly 3x an automobile, 10x a mining rig, and 30x a smartphone.\n\n- Murata reports high-spec MLCC demand exceeding production capacity, with lead times extending from 10–12 weeks to 14–16 weeks, signaling tightening supply conditions.\n\n- The Vera Rubin platform's elevated spec requirements and design complexity mean that not only supplier technical capability but also speed of response to design changes is becoming a key competitive differentiator.\n\n- Despite being in the pre-mass production stage, portions of the BOM remain unfinalized in terms of component specs and supply structure, with final adjustments still underway.\n\n- Near-term shipment timelines may be impacted, but longer-term this is viewed as a necessary process to strengthen supply chain resilience.\n\n---\n\nhttps://t.co/wzHPK5QUDc", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03666764447051918, "ret_signed_m1d": 0.03666764447051918, "alpha_spy_m1d": 0.06491449287867701, "alpha_c_m1d": 0.07729965458701638, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.0406320101164972, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07685538281020832, "ret_signed_p1d": 0.07685538281020832, "alpha_spy_p1d": 0.0693209190273183, "alpha_c_p1d": 0.061731155135746674, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.015124227674461643, "ret_p1w": 0.1105896157230859, "ret_signed_p1w": 0.1105896157230859, "alpha_spy_p1w": 0.09693530646798343, "alpha_c_p1w": 0.07650384456176229, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.03408577116132361, "ret_p1m": 0.41977119389850404, "ret_signed_p1m": 0.41977119389850404, "alpha_spy_p1m": 0.32560508431659296, "alpha_c_p1m": 0.22255518247955597, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.19721601141894807, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0567, -0.0567, -0.0323, -0.0317, -0.0291, -0.0739, -0.0718, -0.0718, -0.0387, 0.0011, -0.012, -0.0079, -0.0024, -0.0393, -0.041, -0.0311, -0.0317, -0.0762, -0.0619, -0.0739, -0.089, -0.0887, -0.1035, -0.0358, -0.0346, -0.0454, -0.0651, -0.05, -0.0355, -0.0355, -0.0425, -0.0529, -0.046, 0.0197, 0.0491, 0.0653, 0.068, 0.068, 0.1755, 0.2272, 0.188, 0.1938, 0.1813, 0.0839, 0.0627, 0.1066, 0.0993, 0.0093, 0.0671, 0.0921, 0.081, 0.0607, 0.0482, 0.1049, 0.1171, 0.0892, 0.0892, 0.0194, 0.0296, 0.0636, 0.0819, 0.0883, 0.0367, 0.0, 0.0769, 0.0202, 0.1115, 0.1167, 0.1106, 0.1734, 0.1704, 0.2162, 0.2232, 0.2699, 0.3259, 0.4057, 0.3494, 0.3899, 0.411, 0.4406, 0.4119, 0.4479, 0.4421, 0.4198, 0.4198, 0.5125, 0.5072, 0.5072, 0.5072, 0.5072, 0.6568, 0.716, 0.7483, 0.7829, 0.7862, 0.8595, 0.8043, 0.8102, 0.8128, 0.8082, 0.9733, 1.0915, 1.3256, 1.3693, 1.2939, 1.5045, 1.8234, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2049672730119999519", "ticker_idx": 0, "ts": "2026-04-30T02:10:34+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "HBM revenue is expected to triple YoY this year. Capacity is already sold out. From Q3 of this year, HBM4 revenue will account for more than half of total HBM revenue. The supply-to-demand fulfillment rate is at a historic low. 2027 demand is already being booked in advance by customers.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Samsung: HBM revenue tripling YoY, capacity sold out, historic-low supply/demand ratio, 2027 demand already booking ahead, and 2nm foundry contracts materializing.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics 1Q26 Conference Call Q&A Key Takeaways (1)\n\n1. Is Samsung also pursuing LTAs?\nYes. Contracts have already been signed with some customers. Unlike prior trust-based agreements, these carry substantial binding force. By covering investment amounts, duration, and technical difficulty, they lower CAPA operational risk and enhance business stability and visibility for both customers and Samsung.\n\n2. Union bonus negotiations — were any provisions booked in Q1, and at what scale?\nLabor-management negotiations are still in progress, so no provisions were booked this quarter. The size of any Q2 provision will be determined based on the outcome of the negotiations.\n\n3. Detailed memory results\n• Q1 DRAM bit growth met guidance\n• NAND grew by high single digits, exceeding guidance\n• ASP rose by low 90% for DRAM and high 80% for NAND\n\n4. Impact of the May general strike on results and management response\nA general strike is scheduled for May 21 – June 7. The company will do its utmost to prevent any production disruption.\n\n5. Geopolitical risks\nProduction lines are operating normally, with no supply chain risk following the Middle East conflict. Some process gases are sourced from parts of the Middle East, but inventory has been secured and the supplier base is diversified, so there is no medium- to long-term supply risk.\nOn rising electricity prices, the company will continue to secure a stable power supply through close cooperation with the government. There is, however, some impact from higher freight costs driven by oil prices.\n\n6. Q2 memory business outlook\nIndustry-wide supply expansion constraints will persist for the time being. The supply-to-demand fulfillment rate is at a historic low. 2027 demand is already being booked in advance by customers. Based on demand received so far alone, the 2027 supply-demand gap is expected to widen further versus 2026.\nBit growth is expected to increase by mid-single digits for DRAM and low single digits for NAND QoQ.\n\n7. HBM sales expansion outlook and HBM4E business status\nHBM revenue is expected to triple YoY this year. Capacity is already sold out. From Q3 of this year, HBM4 revenue will account for more than half of total HBM revenue.\n\n8. Foundry business outlook\nCustomer demand to secure memory and foundry on a turnkey basis is increasing. The company is in discussions with multiple HPC customers on 2nm/4nm processes, and 2nm contracts with some customers are expected to materialize in the near future.\nBeyond silicon photonics devices, Samsung is also developing advanced-node processes, 3D packaging, and CPO technologies. Mass production for optical communication module customers is scheduled to begin in 2H26.\n\n---\n\nSamsung Electronics 1Q26 Conference Call Q&A Key Takeaways (2)\n\n9. How will Samsung manage the mix between HBM and conventional DRAM?\nConventional DRAM has actually been delivering higher profitability than HBM recently. Conventional pricing is negotiated on a quarterly basis while HBM is on an annual basis, and as conventional’s quarterly price increases have widened, a profitability reversal has occurred.\nHowever, given that the HBM supply-demand gap continues to widen, the profitability gap between the two products is expected to narrow significantly in 2027.\nRunning a P-mix weighted toward conventional based solely on short-term profitability could constrain AI infrastructure build-out itself. A balanced supply is needed to sustain continued AI demand creation.\nSamsung will execute a balanced P-mix taking into account medium- to long-term growth and long-term customer relationships in a comprehensive manner.\n\n10. MX division outlook\nThe 2026 market is expected to grow slightly in value terms, but unit volumes will drop sharply. Profitability deterioration is unavoidable.\n\n11. NAND market outlook\nDemand for high-performance storage such as TLC-based PCIe 6.0 SSDs is expected to grow. The need to address data movement latency stemming from performance gaps between HBM, DRAM, and storage will also increase.\nIn the QLC space, the V9 transition is also accelerating.\n\n12. Plans for additional investment in the Taylor fab, and how to address relatively low utilization at legacy fabs?\nTaylor Fab 1 held its equipment move-in ceremony on April 23. After fab ramp-up in 2026 and the start of mass production in 2027, 2nm capacity will be expanded in stages. For Fab 2, initial reviews for construction are underway in parallel with discussions on global customer orders.\nLess competitive processes are being decisively phased out. For CIS and DDI — where process migration is expected to continue — capacity will be converted to 17nm, while 8-inch lines currently mass-producing PMIC, DDI, and similar products will be sequentially closed.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.024943310657596474, "ret_signed_m1d": 0.024943310657596474, "alpha_spy_m1d": 0.03479491825960168, "alpha_c_m1d": 0.02738844789677164, "bench_spy_m1d": -0.009851607602005208, "bench_c_m1d": -0.002445137239175166, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0027691138283709726, "alpha_c_p1d": -0.014858903556034475, "bench_spy_p1d": 0.0027691138283709726, "bench_c_p1d": 0.014858903556034475, "ret_p1w": 0.23129251700680276, "ret_signed_p1w": 0.23129251700680276, "alpha_spy_p1w": 0.2133145518659585, "alpha_c_p1w": 0.16740535436475734, "bench_spy_p1w": 0.017977965140844265, "bench_c_p1w": 0.06388716264204541, "ret_p1m": 0.43764172335600904, "ret_signed_p1m": 0.43764172335600904, "alpha_spy_p1m": 0.3850159911077866, "alpha_c_p1m": 0.24002414171048603, "bench_spy_p1m": 0.05262573224822242, "bench_c_p1m": 0.197617581645523, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3193, -0.2419, -0.2347, -0.279, -0.2822, -0.2469, -0.2496, -0.2406, -0.1917, -0.1799, -0.1799, -0.1799, -0.1799, -0.1401, -0.1396, -0.1265, -0.0948, -0.079, -0.0134, -0.0202, -0.0202, -0.117, -0.2207, -0.1329, -0.1482, -0.2148, -0.1496, -0.1401, -0.1496, -0.1695, -0.146, -0.1225, -0.0564, -0.0926, -0.0976, -0.1568, -0.1415, -0.1446, -0.1849, -0.1849, -0.2005, -0.2417, -0.1399, -0.1909, -0.1556, -0.1243, -0.1088, -0.0454, -0.0748, -0.0658, -0.0884, -0.0635, -0.0431, -0.0136, -0.0204, -0.0272, -0.0068, -0.0136, 0.0181, -0.0045, 0.0181, 0.0068, 0.0249, 0.0, 0.0, 0.0544, 0.0544, 0.2063, 0.2313, 0.2177, 0.2948, 0.2653, 0.288, 0.3424, 0.2268, 0.2744, 0.2494, 0.2517, 0.3583, 0.3265, 0.3265, 0.356, 0.3923, 0.3583, 0.4376, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2042152949955359157", "ticker_idx": 0, "ts": "2026-04-09T08:09:39+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "buy names tied to Nvidia's CPO ecosystem", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish on stocks tied to Nvidia's CPO ecosystem as the leading play on co-packaged optics commercialization.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "If you want to make money from CPO stocks right now, it would be smart to buy names tied to Nvidia’s CPO ecosystem. Other companies are dragging their feet because of existing vested interests.\n\nDo your own research. https://t.co/nY1Q3NBJ1k", "reply_text": null, "parent_text": null, "quoted_text": "When will CPO actually see mass production?-Digitimes\n\nThe market continues to closely watch the deployment of co-packaged optics (CPO) technology in cloud AI, driven by the desire of the silicon photonics (SiPh) ecosystem to see a tangible revenue impact. Cloud AI vendors are https://t.co/qM5hoOp1vo", "parent_author": null, "ret_m1d": -0.009950529033846744, "ret_signed_m1d": -0.009950529033846744, "alpha_spy_m1d": -0.00421452936778921, "alpha_c_m1d": 0.00722309968216539, "bench_spy_m1d": -0.005735999666057534, "bench_c_m1d": -0.017173628716012135, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.025664733438725884, "ret_signed_p1d": 0.025664733438725884, "alpha_spy_p1d": 0.026326513953275454, "alpha_c_p1d": 0.010396653771754805, "bench_spy_p1d": -0.0006617805145495703, "bench_c_p1d": 0.01526807966697108, "ret_p1w": 0.07851667747196767, "ret_signed_p1w": 0.07851667747196767, "alpha_spy_p1w": 0.04652714818265902, "alpha_c_p1w": 0.021604241383284206, "bench_spy_p1w": 0.03198952928930865, "bench_c_p1w": 0.05691243608868346, "ret_p1m": 0.1701375273942176, "ret_signed_p1m": 0.1701375273942176, "alpha_spy_p1m": 0.08525861069624385, "alpha_c_p1m": -0.1464481462867664, "bench_spy_p1m": 0.08487891669797376, "bench_c_p1m": 0.316585673680984, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0055, 0.0103, -0.0042, 0.017, 0.0126, 0.0126, -0.0318, -0.0033, 0.005, 0.0204, 0.0139, 0.025, 0.0413, 0.0467, 0.0392, 0.0092, -0.0195, -0.0529, -0.0655, 0.0081, 0.0333, 0.0251, 0.0333, 0.0164, -0.006, -0.006, 0.0057, 0.0221, 0.0216, 0.0321, 0.0415, 0.0486, 0.0633, 0.0053, -0.0366, -0.0078, -0.021, -0.0048, -0.0032, -0.0332, -0.0069, 0.0046, 0.0115, -0.0042, -0.0199, -0.0038, -0.0108, -0.0191, -0.0291, -0.061, -0.045, -0.0474, -0.0284, -0.0689, -0.0891, -0.1019, -0.0517, -0.0444, -0.0355, -0.0355, -0.0341, -0.0316, -0.01, 0.0, 0.0257, 0.0294, 0.0685, 0.0813, 0.0785, 0.0966, 0.0987, 0.0868, 0.1011, 0.0855, 0.1325, 0.1778, 0.1591, 0.1378, 0.0852, 0.0791, 0.0792, 0.0685, 0.1301, 0.15, 0.1701, 0.1932, 0.2005, 0.2279, 0.2818, 0.2252, 0.2089, 0.1996, 0.2151, 0.1936, 0.1708, 0.1708, 0.1683, 0.156, 0.165, 0.1481, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2047850011288011025", "ticker_idx": 0, "ts": "2026-04-25T01:27:44+00:00", "symbol": "Lotes", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm also positive on Lotes", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish on Lotes; endorses the parent's thesis on CPU socket exposure and optionality.", "resolved_tickers": ["3533.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Lotes Co. 3533.TW Taiwan connector maker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@HighBorrowCost I’m also positive on Lotes.", "reply_text": null, "parent_text": "@jukan05 Why not Lotes? Geared to CPU via LGA sockets. And plenty of optionality to be second source when SOCAMM (finally) adopted, and second source for LGA NPO sockets + ELS connectors (most obvious second source to FIT?)", "quoted_text": null, "parent_author": "HighBorrowCost", "ret_m1d": -0.07011070110701112, "ret_signed_m1d": -0.07011070110701112, "alpha_spy_m1d": -0.06839085757590402, "alpha_c_m1d": -0.06793092849973392, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": -0.0021797726072771972, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014760147601476037, "ret_signed_p1d": -0.014760147601476037, "alpha_spy_p1d": -0.009894198873329874, "alpha_c_p1d": 0.0021795117161633737, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.01693965931763941, "ret_p1w": -0.03690036900369009, "ret_signed_p1w": -0.03690036900369009, "alpha_spy_p1w": -0.04087149044801042, "alpha_c_p1w": -0.04611750582894214, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.009217136825252048, "ret_p1m": -0.02583025830258301, "ret_signed_p1m": -0.02583025830258301, "alpha_spy_p1m": -0.07535700689622782, "alpha_c_p1m": -0.17884757750727476, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.15301731920469175, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4446, -0.4613, -0.4723, -0.4613, -0.4077, -0.3838, -0.4207, -0.4114, -0.4059, -0.4004, -0.393, -0.393, -0.393, -0.393, -0.393, -0.393, -0.393, -0.393, -0.4004, -0.3967, -0.3376, -0.345, -0.345, -0.3727, -0.3875, -0.4483, -0.4151, -0.4225, -0.4428, -0.4188, -0.3764, -0.3339, -0.3542, -0.3782, -0.3284, -0.2934, -0.2509, -0.238, -0.2675, -0.2786, -0.2066, -0.1624, -0.1661, -0.1808, -0.2435, -0.2159, -0.2454, -0.2454, -0.2454, -0.2509, -0.1771, -0.1476, -0.1624, -0.1697, -0.179, -0.1236, -0.0738, -0.0812, -0.0314, -0.0277, -0.0221, -0.0923, -0.0701, 0.0, -0.0148, -0.0221, -0.0351, -0.0351, -0.0369, -0.0812, 0.0092, 0.0517, -0.0037, 0.0498, 0.0332, -0.0424, -0.0959, -0.1402, -0.1421, -0.1421, -0.1292, -0.0683, -0.0572, -0.0461, -0.0258, -0.0314, -0.083, -0.0221, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042932686973796492", "ticker_idx": 0, "ts": "2026-04-11T11:48:03+00:00", "symbol": "ABF substrate", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think ABF and HBM stocks are still worth buying even now", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long ABF substrate and HBM stocks as chip bottlenecks; skeptical on gas turbine stocks as power theme looks overblown.", "resolved_tickers": ["4062.T", "3034.TW", "2383.TW", "3037.TW"], "resolution_method": "basket", "resolution_reason": "ABF substrate basket: Ibiden, Unimicron, EMC, Tripod", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "A striking sentence I read today:\n\nWe don’t think power is the constraint for TSMC’s chip demand, at least for 2027, but ABF substrate and HBM supply are major gating factors. -Morgan Stanley\n\n---\n\nMy view: hyperscalers will find a solution on the power side, whether that means buying transformers from China or buying natural gas generators instead of transformers.\n\nBut the chip bottleneck cannot be worked around.\n\n---\n\nIsn’t it a bit late to be buying gas turbine stocks at this point? I think ABF and HBM stocks are still worth buying even now.\n\n---\n\nFor now, the power shortage theme looks a bit overblown to me.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.024989310682911103, "ret_signed_m1d": 0.024989310682911103, "alpha_spy_m1d": 0.034667132938495104, "alpha_c_m1d": 0.0455236522846616, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.020534341601750494, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.026096150418754765, "ret_signed_p1d": 0.026096150418754765, "alpha_spy_p1d": 0.013911269666384646, "alpha_c_p1d": 0.010094489672322154, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.01600166074643261, "ret_p1w": 0.07826567871409723, "ret_signed_p1w": 0.07826567871409723, "alpha_spy_p1w": 0.04529672969486748, "alpha_c_p1w": 0.01680014124602086, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.06146553746807637, "ret_p1m": 0.46729656073492154, "ret_signed_p1m": 0.46729656073492154, "alpha_spy_p1m": 0.3913892309826259, "alpha_c_p1m": 0.26408252176878855, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.203214038966133, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3342, -0.3429, -0.3205, -0.3079, -0.2963, -0.2942, -0.2744, -0.26, -0.2746, -0.2588, -0.242, -0.2646, -0.2612, -0.281, -0.247, -0.2673, -0.2895, -0.2998, -0.2619, -0.2439, -0.2394, -0.2239, -0.2202, -0.2165, -0.2121, -0.2018, -0.1937, -0.2065, -0.1915, -0.1414, -0.1331, -0.128, -0.1216, -0.135, -0.171, -0.2223, -0.1898, -0.2029, -0.2485, -0.2002, -0.1762, -0.1622, -0.139, -0.1299, -0.1127, -0.0975, -0.0942, -0.1037, -0.1544, -0.1716, -0.1194, -0.1047, -0.1068, -0.1386, -0.1901, -0.1268, -0.1342, -0.1226, -0.1134, -0.082, -0.0051, 0.0062, 0.025, 0.0, 0.0261, 0.0398, 0.062, 0.0515, 0.0783, 0.1251, 0.164, 0.1529, 0.2509, 0.2921, 0.2544, 0.2479, 0.3126, 0.3098, 0.3409, 0.3387, 0.3509, 0.454, 0.4127, 0.4401, 0.4673, 0.4896, 0.4859, 0.3876, 0.3854, 0.371, 0.3751, 0.4972, 0.5841, 0.6766, 0.7009, 0.6816, 0.6184, 0.7212, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042932686973796492", "ticker_idx": 1, "ts": "2026-04-11T11:48:03+00:00", "symbol": "HBM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think ABF and HBM stocks are still worth buying even now", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long ABF substrate and HBM stocks as chip bottlenecks; skeptical on gas turbine stocks as power theme looks overblown.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "HBM sector basket: SK Hynix, Samsung, Micron", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "A striking sentence I read today:\n\nWe don’t think power is the constraint for TSMC’s chip demand, at least for 2027, but ABF substrate and HBM supply are major gating factors. -Morgan Stanley\n\n---\n\nMy view: hyperscalers will find a solution on the power side, whether that means buying transformers from China or buying natural gas generators instead of transformers.\n\nBut the chip bottleneck cannot be worked around.\n\n---\n\nIsn’t it a bit late to be buying gas turbine stocks at this point? I think ABF and HBM stocks are still worth buying even now.\n\n---\n\nFor now, the power shortage theme looks a bit overblown to me.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0005400179830016283, "ret_signed_m1d": -0.0005400179830016283, "alpha_spy_m1d": 0.009137804272582373, "alpha_c_m1d": 0.014031172318887286, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.014571190301888914, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.059867898143609675, "ret_signed_p1d": 0.059867898143609675, "alpha_spy_p1d": 0.047683017391239556, "alpha_c_p1d": 0.04033434887297468, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.019533549270634998, "ret_p1w": 0.07985506415001391, "ret_signed_p1w": 0.07985506415001391, "alpha_spy_p1w": 0.04688611513078415, "alpha_c_p1w": 0.03334449599123228, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.04651056815878163, "ret_p1m": 0.6498680072069773, "ret_signed_p1m": 0.6498680072069773, "alpha_spy_p1m": 0.5739606774546817, "alpha_c_p1m": 0.3839095901093137, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.26595841709766366, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.27, -0.2593, -0.2284, -0.2251, -0.2369, -0.2119, -0.1959, -0.1907, -0.2089, -0.1597, -0.1222, -0.1182, -0.1195, -0.1436, -0.105, -0.1359, -0.1679, -0.1608, -0.1413, -0.1538, -0.1266, -0.097, -0.097, -0.097, -0.1063, -0.0898, -0.0735, -0.0473, -0.0475, -0.021, -0.0024, 0.0376, 0.0205, 0.0207, -0.0796, -0.1299, -0.0712, -0.1031, -0.1408, -0.0735, -0.0524, -0.0743, -0.0718, -0.0304, -0.0076, 0.0442, 0.0035, -0.0169, -0.0768, -0.0611, -0.0698, -0.1252, -0.1239, -0.1764, -0.2001, -0.1105, -0.1519, -0.1242, -0.1006, -0.0855, -0.002, -0.0124, -0.0005, 0.0, 0.0599, 0.0705, 0.0882, 0.0754, 0.0799, 0.1067, 0.1336, 0.1414, 0.1438, 0.1963, 0.1789, 0.1944, 0.182, 0.2016, 0.2998, 0.3496, 0.4752, 0.4857, 0.5692, 0.6975, 0.6499, 0.7323, 0.7287, 0.5979, 0.5883, 0.5622, 0.589, 0.714, 0.6941, 0.6941, 0.8536, 0.9535, 0.9521, 1.0324, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042932686973796492", "ticker_idx": 2, "ts": "2026-04-11T11:48:03+00:00", "symbol": "gas turbine", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Isn't it a bit late to be buying gas turbine stocks at this point", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long ABF substrate and HBM stocks as chip bottlenecks; skeptical on gas turbine stocks as power theme looks overblown.", "resolved_tickers": ["GEV", "ENR.DE", "7011.T"], "resolution_method": "basket", "resolution_reason": "Gas turbine basket: GE Vernova, Siemens Energy, MHI", "bench_c": "XLI", "bench_c_reason": "industry='Specialty Industrial Machinery'", "bench_c_industry": "Specialty Industrial Machinery", "unit_kind": "self_thread", "lang": "en", "tweet_text": "A striking sentence I read today:\n\nWe don’t think power is the constraint for TSMC’s chip demand, at least for 2027, but ABF substrate and HBM supply are major gating factors. -Morgan Stanley\n\n---\n\nMy view: hyperscalers will find a solution on the power side, whether that means buying transformers from China or buying natural gas generators instead of transformers.\n\nBut the chip bottleneck cannot be worked around.\n\n---\n\nIsn’t it a bit late to be buying gas turbine stocks at this point? I think ABF and HBM stocks are still worth buying even now.\n\n---\n\nFor now, the power shortage theme looks a bit overblown to me.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0015992701703475183, "ret_signed_m1d": -0.0015992701703475183, "alpha_spy_m1d": -0.01127709242593152, "alpha_c_m1d": -0.008604373424021724, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.007005103253674205, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.008109157391768021, "ret_signed_p1d": -0.008109157391768021, "alpha_spy_p1d": 0.004075723360602098, "alpha_c_p1d": -0.004519680223282788, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.0035894771684852333, "ret_p1w": -0.012262355511089043, "ret_signed_p1w": 0.012262355511089043, "alpha_spy_p1w": 0.045231304530318796, "alpha_c_p1w": 0.0190359223372634, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.00677356682617436, "ret_p1m": 0.0021117782564766077, "ret_signed_p1m": -0.0021117782564766077, "alpha_spy_p1m": 0.07379555149581903, "alpha_c_p1m": 0.007267081328120114, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.009378859584596722, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2023, -0.1925, -0.1729, -0.1618, -0.1683, -0.1747, -0.1746, -0.1741, -0.1761, -0.165, -0.1662, -0.1565, -0.1512, -0.1357, -0.1033, -0.1202, -0.1331, -0.0976, -0.0677, -0.0764, -0.04, -0.0539, -0.0548, -0.0572, -0.0659, -0.0461, -0.0452, -0.0437, -0.0423, -0.0363, -0.0349, -0.0334, -0.0219, -0.0149, -0.0611, -0.0672, -0.0855, -0.1015, -0.1024, -0.0766, -0.0822, -0.0798, -0.1089, -0.0923, -0.0743, -0.0566, -0.0737, -0.0925, -0.0912, -0.0805, -0.0536, -0.0896, -0.1034, -0.1305, -0.1249, -0.0719, -0.0672, -0.0607, -0.0659, -0.0619, -0.0133, -0.0091, 0.0016, 0.0, 0.0081, -0.0089, -0.0155, -0.011, -0.0123, -0.0146, 0.0533, 0.0867, 0.0952, 0.0615, 0.0496, 0.0455, 0.0503, 0.0414, 0.0375, 0.0562, 0.0708, 0.0316, 0.0235, 0.0265, 0.0021, 0.022, 0.0138, -0.0208, -0.031, -0.0339, -0.0234, -0.023, -0.0203, -0.0128, 0.0004, -0.0297, -0.0636, -0.0799, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048792853149819225", "ticker_idx": 0, "ts": "2026-04-27T15:54:15+00:00", "symbol": "semiconductor equipment", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Semiconductor equipment broadly could become tight", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish semiconductor equipment sector broadly; author expects supply tightness beyond EUV alone, echoing Fubon/MS views on an equipment bottleneck.", "resolved_tickers": ["AMAT", "LRCX", "KLAC", "ASML"], "resolution_method": "basket", "resolution_reason": "Semiconductor equipment basket: AMAT, Lam, KLA, ASML", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Two sell side firms I usually pay close attention to came out with a similar view today.\n\nFubon Research: Even if TSMC wants to increase capex next year, EUV shortages may limit how much it can actually raise capex.\n\nMorgan Stanley: Around 2028, memory makers may try to execute unprecedented capex programs, and at that point EUV could emerge as the next bottleneck, potentially triggering a boom in the semiconductor equipment sector.\n\nPersonally, I think the shortage won’t be limited to EUV. Semiconductor equipment broadly could become tight.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.023987188870400633, "ret_signed_m1d": 0.023987188870400633, "alpha_spy_m1d": 0.025707032401507734, "alpha_c_m1d": 0.023631654812190872, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.042938815088782584, "ret_signed_p1d": -0.042938815088782584, "alpha_spy_p1d": -0.03807286636063642, "alpha_c_p1d": -0.013210971648187048, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": -0.041822561511517115, "ret_signed_p1w": -0.041822561511517115, "alpha_spy_p1w": -0.04579368295583744, "alpha_c_p1w": -0.042869451980986045, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.14197391431568662, "ret_signed_p1m": 0.14197391431568662, "alpha_spy_p1m": 0.09244716572204181, "alpha_c_p1m": -0.047414943254162845, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1005, -0.0762, -0.1414, -0.1329, -0.1611, -0.2099, -0.1975, -0.1434, -0.1376, -0.1448, -0.1197, -0.1392, -0.1171, -0.1171, -0.1111, -0.0906, -0.0957, -0.0797, -0.0817, -0.0725, -0.0467, -0.0824, -0.0918, -0.098, -0.1473, -0.1263, -0.1528, -0.2054, -0.1628, -0.147, -0.1381, -0.169, -0.1636, -0.1461, -0.1273, -0.1366, -0.1171, -0.1329, -0.1144, -0.0892, -0.1011, -0.1646, -0.1714, -0.2075, -0.1593, -0.1308, -0.1454, -0.1454, -0.1399, -0.1337, -0.0568, -0.0255, -0.0133, -0.0041, 0.0075, -0.013, -0.0338, -0.0071, -0.01, -0.0186, -0.0052, -0.0157, 0.024, 0.0, -0.0429, -0.0418, -0.0265, -0.0362, -0.0418, -0.0007, 0.0596, 0.0263, 0.076, 0.0752, 0.0487, 0.0737, 0.0859, 0.0435, 0.0116, -0.0013, 0.0566, 0.076, 0.0952, 0.0952, 0.142, 0.1212, 0.1186, 0.1195, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2047467201075245356", "ticker_idx": 0, "ts": "2026-04-24T00:06:35+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel 14A seems to be ramping much better than I had expected.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish INTC: 14A process ramping better than expected, implying TSMC technology transfer thesis may have legs.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "Were Jensen Huang and C.C. Wei wrong when they said that TSMC’s technology cannot be copied just because a few people leave the company?\n\nIntel 14A seems to be ramping much better than I had expected.", "reply_text": null, "parent_text": null, "quoted_text": "终于知道为什么台湾动用全国力量围剿罗唯仁了。\n这是真的决定人类命运的一次跳槽，英特尔真有能力去把那一册子N2 A16 A14机密变成世纪大翻盘。真有能力扭转美国半导体乃至美国国家的命运。\n\n这比梁孟松还要梁孟松，已经是这个世纪最难以置信的逆转故事 $INTC https://t.co/bkOuczPy2b", "parent_author": null, "ret_m1d": -0.19093775092587517, "ret_signed_m1d": -0.19093775092587517, "alpha_spy_m1d": -0.18324804280530305, "alpha_c_m1d": -0.14238314185204792, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029682540841616367, "ret_signed_p1d": 0.029682540841616367, "alpha_spy_p1d": 0.027959734352914323, "alpha_c_p1d": 0.03003794854028463, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.20692999323485117, "ret_signed_p1w": 0.20692999323485117, "alpha_spy_p1w": 0.19753141359270976, "alpha_c_p1w": 0.20025594523998813, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.45190204759673613, "ret_signed_p1m": 0.45190204759673613, "alpha_spy_p1m": 0.40750054032913785, "alpha_c_p1m": 0.31391925684967736, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4678, -0.409, -0.4105, -0.437, -0.4087, -0.4033, -0.4112, -0.4156, -0.3871, -0.3913, -0.429, -0.415, -0.4369, -0.4331, -0.4331, -0.4405, -0.4492, -0.4594, -0.4656, -0.4714, -0.4412, -0.432, -0.4492, -0.4474, -0.4488, -0.4778, -0.4478, -0.4433, -0.474, -0.4478, -0.4332, -0.4187, -0.4518, -0.4455, -0.4456, -0.4662, -0.4544, -0.4405, -0.4685, -0.4668, -0.4662, -0.4284, -0.4657, -0.4775, -0.501, -0.4654, -0.4181, -0.3896, -0.3896, -0.3848, -0.359, -0.2858, -0.2522, -0.2442, -0.2103, -0.2269, -0.2132, -0.1701, -0.1701, -0.204, -0.1972, -0.2092, -0.1909, 0.0, 0.0297, 0.024, 0.1479, 0.1447, 0.2069, 0.1604, 0.3103, 0.3692, 0.3281, 0.5134, 0.5682, 0.4612, 0.4574, 0.4045, 0.3178, 0.3105, 0.3424, 0.4412, 0.4357, 0.4519, 0.4519, 0.4965, 0.4753, 0.4646, 0.3894, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2042800654523986126", "ticker_idx": 0, "ts": "2026-04-11T03:03:24+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's earnings are projected to improve further in Q2, with 2026 full-year results expected to reach record levels", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Counterpoint Research bullish thesis on Samsung: record 2026 earnings, strong DRAM margins until 2027, HBM4/4E leadership.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Counterpoint Research: Memory prices to rise further in Q2… Samsung’s Q2 results likely to improve further\n\nAccording to Counterpoint Research's memory price tracker, memory demand relative to supply remains strong, with Q2 prices expected to rise over 80% for mobile and over 50% for PC. Samsung's earnings are projected to improve further in Q2, with 2026 full-year results expected to reach record levels.\n\nCounterpoint senior researcher Choi Jeong-gu noted: \"Samsung's strong performance is expected to continue for the foreseeable future. Memory demand is robust across the board, generating high margins in commodity DRAM, and this trend is likely to persist until supply expansion becomes visible around 2027. In HBM4, Samsung has solidified its leadership by adopting 1c-nano process core dies and 4nm foundry-based base dies, and it can be said to hold an advantageous position over competitors in the upcoming HBM4E as well.\"\n\n---\n\nhttps://t.co/QlFX1QBQwE", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.02487562189054726, "ret_signed_m1d": 0.02487562189054726, "alpha_spy_m1d": 0.03455344414613126, "alpha_c_m1d": 0.045409963492297756, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.020534341601750494, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.027363184079602032, "ret_signed_p1d": 0.027363184079602032, "alpha_spy_p1d": 0.015178303327231912, "alpha_c_p1d": 0.011361523333169421, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.01600166074643261, "ret_p1w": 0.0671641791044777, "ret_signed_p1w": 0.0671641791044777, "alpha_spy_p1w": 0.03419523008524794, "alpha_c_p1w": 0.005698641636401325, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.06146553746807637, "ret_p1m": 0.3880597014925373, "ret_signed_p1m": 0.3880597014925373, "alpha_spy_p1m": 0.3121523717402417, "alpha_c_p1m": 0.18484566252640433, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.203214038966133, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3034, -0.2856, -0.2607, -0.2587, -0.2791, -0.2578, -0.2439, -0.2448, -0.2448, -0.2081, -0.1937, -0.2021, -0.2031, -0.2533, -0.1684, -0.1604, -0.2091, -0.2126, -0.1738, -0.1768, -0.1669, -0.1133, -0.1004, -0.1004, -0.1004, -0.1004, -0.0567, -0.0562, -0.0418, -0.007, 0.0103, 0.0823, 0.0749, 0.0749, -0.0314, -0.1451, -0.0487, -0.0656, -0.1386, -0.0671, -0.0567, -0.0671, -0.089, -0.0631, -0.0373, 0.0352, -0.0045, -0.01, -0.075, -0.0582, -0.0616, -0.1058, -0.1058, -0.1229, -0.1682, -0.0565, -0.1124, -0.0736, -0.0393, -0.0224, 0.0473, 0.0149, 0.0249, 0.0, 0.0274, 0.0498, 0.0821, 0.0746, 0.0672, 0.0896, 0.0821, 0.1169, 0.092, 0.1169, 0.1045, 0.1244, 0.097, 0.097, 0.1567, 0.1567, 0.3234, 0.3507, 0.3358, 0.4204, 0.3881, 0.4129, 0.4726, 0.3458, 0.398, 0.3706, 0.3731, 0.49, 0.4552, 0.4552, 0.4876, 0.5274, 0.49, 0.5771, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2053967787815837865", "ticker_idx": 0, "ts": "2026-05-11T22:37:36+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "전 좀 영업이익 전망치를 낮춰야한다고 생각합니다", "tweet_type": "prediction", "timeline": "quarters", "summary": "Author believes SK Hynix operating profit estimates need to come down, implying bearish on the stock.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@treeapple33 전 좀 영업이익 전망치를 낮춰야한다고 생각합니다.", "reply_text": null, "parent_text": "@jukan05 그런 의미에서 내년 메모리 영업이익 추정치가 상당히 갈리고 있는 것처럼 보이네요. 필히 가격을 낮추고 장기 안정성을 얻는 방향으로 갈 것 같은데 그렇게 보면 닉스 400조 영업이익이 과연 가능할지..", "quoted_text": null, "parent_author": "treeapple33", "ret_m1d": -0.10319145788009099, "ret_signed_m1d": 0.10319145788009099, "alpha_spy_m1d": 0.10091904789291306, "alpha_c_m1d": 0.08623874220165673, "bench_spy_m1d": -0.0022724099871779257, "bench_c_m1d": -0.01695271567843426, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023936116977007815, "ret_signed_p1d": 0.023936116977007815, "alpha_spy_p1d": 0.022421176985555902, "alpha_c_p1d": -0.0021956482576610536, "bench_spy_p1d": -0.0015149399914519135, "bench_c_p1d": -0.02613176523466887, "ret_p1w": -0.02127656320189175, "ret_signed_p1w": 0.02127656320189175, "alpha_spy_p1w": 0.020397402658623554, "alpha_c_p1w": -0.031039073420388363, "bench_spy_p1w": -0.0008791605432681981, "bench_c_p1w": -0.052315636622280115, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5434, -0.5285, -0.5328, -0.5328, -0.5328, -0.5328, -0.5253, -0.4961, -0.4951, -0.4664, -0.4595, -0.4154, -0.4356, -0.4356, -0.5005, -0.5484, -0.4995, -0.5085, -0.5553, -0.5011, -0.492, -0.5053, -0.516, -0.4819, -0.484, -0.4383, -0.4612, -0.4644, -0.5037, -0.4755, -0.4707, -0.5037, -0.5037, -0.5356, -0.5707, -0.5229, -0.5585, -0.534, -0.5287, -0.5128, -0.4505, -0.4691, -0.4537, -0.4468, -0.4133, -0.3957, -0.3856, -0.4, -0.3798, -0.3489, -0.3495, -0.3484, -0.35, -0.3128, -0.3085, -0.3122, -0.316, -0.316, -0.2303, -0.2303, -0.1484, -0.1202, -0.1032, 0.0, -0.0239, 0.0511, 0.0479, -0.0324, -0.0213, -0.0718, -0.0718, 0.0319, 0.0324, 0.0324, 0.0915, 0.1931, 0.2178, 0.2412, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2045014557345882229", "ticker_idx": 0, "ts": "2026-04-17T05:40:39+00:00", "symbol": "legacy foundry", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Legacy foundry pricing is going up.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Legacy foundry prices rising into shortage; bullish on legacy foundry sector.", "resolved_tickers": ["UMC", "0981.HK", "2303.TW"], "resolution_method": "basket", "resolution_reason": "Legacy foundry basket: UMC, SMIC, United Micro", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "What did I say? Everything is going to go into shortage.\n\nLegacy foundry pricing is going up. https://t.co/bYTwui6b53", "reply_text": null, "parent_text": null, "quoted_text": "UMC Wafer Price 📈📈📈 https://t.co/xf8uu3xS4t https://t.co/Ly6cEfmurW", "parent_author": null, "ret_m1d": -0.047822594718317735, "ret_signed_m1d": -0.047822594718317735, "alpha_spy_m1d": -0.03588122917211012, "alpha_c_m1d": -0.027657100502331056, "bench_spy_m1d": -0.011941365546207616, "bench_c_m1d": -0.02016549421598668, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048351634136661215, "ret_signed_p1d": 0.048351634136661215, "alpha_spy_p1d": 0.05035130168956692, "alpha_c_p1d": 0.04878254633386303, "bench_spy_p1d": -0.001999667552905704, "bench_c_p1d": -0.00043091219720181595, "ret_p1w": 0.052374931848240935, "ret_signed_p1w": 0.052374931848240935, "alpha_spy_p1w": 0.04702389161028164, "alpha_c_p1w": -0.03871434436194088, "bench_spy_p1w": 0.0053510402379592925, "bench_c_p1w": 0.09108927621018181, "ret_p1m": 0.38211853768670473, "ret_signed_p1m": 0.38211853768670473, "alpha_spy_p1m": 0.341971511498159, "alpha_c_p1m": 0.20545538524904933, "bench_spy_p1m": 0.04014702618854571, "bench_c_p1m": 0.1766631524376554, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0362, 0.0644, 0.0575, 0.0358, 0.0647, 0.1349, 0.1173, 0.0487, -0.0006, -0.0303, -0.0451, -0.0472, -0.0553, -0.0501, -0.0326, -0.0299, -0.0256, -0.027, -0.0202, -0.02, -0.0243, -0.0254, -0.0283, -0.0357, -0.011, 0.0102, -0.007, -0.0272, -0.021, -0.0422, -0.0798, -0.0996, -0.0923, -0.0954, -0.1075, -0.0926, -0.0806, -0.1016, -0.114, -0.102, -0.1033, -0.1086, -0.1317, -0.1558, -0.1771, -0.1733, -0.1467, -0.1779, -0.1876, -0.2049, -0.2005, -0.1846, -0.221, -0.221, -0.2204, -0.1812, -0.1386, -0.1408, -0.1113, -0.1273, -0.1278, -0.0989, -0.0478, 0.0, 0.0484, 0.038, 0.0553, 0.0074, 0.0524, 0.0495, 0.0472, 0.0742, 0.1245, 0.1242, 0.1471, 0.1776, 0.2731, 0.3051, 0.2697, 0.3097, 0.3665, 0.3209, 0.3844, 0.3936, 0.3821, 0.4033, 0.4172, 0.471, 0.4899, 0.5401, 0.6725, 0.769, 0.7924, 0.7522, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2055026707321786430", "ticker_idx": 0, "ts": "2026-05-14T20:45:22+00:00", "symbol": "AMAT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMAT has an extremely broad equipment portfolio, allowing it to propose bundled solutions across multiple tools. That gives them a structural advantage", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish AMAT on structural advantage from broad equipment portfolio enabling bundled fab solutions vs single-tool specialists.", "resolved_tickers": ["AMAT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "AMAT’s strength is that even if each individual tool may lag competitors in performance, AMAT has an extremely broad equipment portfolio, allowing it to propose bundled solutions across multiple tools. That gives them a structural advantage over companies that focus on refining individual tools one by one. After all, a fab does not run on a single piece of equipment.\n\n$AMAT", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008965938256718209, "ret_signed_m1d": -0.008965938256718209, "alpha_spy_m1d": -0.0011335131589230718, "alpha_c_m1d": 0.0012010994888500548, "bench_spy_m1d": -0.007832425097795137, "bench_c_m1d": -0.010167037745568264, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.008943189461537382, "ret_signed_p1d": -0.008943189461537382, "alpha_spy_p1d": 0.0030861624311872005, "alpha_c_p1d": 0.02909671609301623, "bench_spy_p1d": -0.012029351892724582, "bench_c_p1d": -0.03803990555455361, "ret_p1w": -0.02875600294236247, "ret_signed_p1w": -0.02875600294236247, "alpha_spy_p1w": -0.0214715457581377, "alpha_c_p1w": -0.010669718945162554, "bench_spy_p1w": -0.0072844571842247685, "bench_c_p1w": -0.018086283997199915, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1954, -0.1858, -0.1628, -0.1605, -0.1479, -0.1521, -0.1422, -0.1035, -0.1472, -0.1549, -0.1552, -0.2026, -0.1879, -0.2134, -0.2629, -0.2307, -0.2149, -0.2031, -0.2345, -0.2248, -0.2142, -0.2, -0.2068, -0.1892, -0.1895, -0.1788, -0.1511, -0.1617, -0.2315, -0.2347, -0.2666, -0.2242, -0.1969, -0.209, -0.209, -0.1996, -0.1958, -0.1245, -0.097, -0.0932, -0.1018, -0.102, -0.1051, -0.115, -0.099, -0.1111, -0.1049, -0.0842, -0.0832, -0.0534, -0.081, -0.1349, -0.1316, -0.1046, -0.1169, -0.1116, -0.0675, -0.0271, -0.0679, -0.0116, 0.0069, -0.0212, -0.009, 0.0, -0.0089, -0.0613, -0.0764, -0.0311, -0.0288, -0.0178, -0.0178, 0.0338, 0.0187, 0.022, 0.0228, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2056576060045807819", "ticker_idx": 0, "ts": "2026-05-19T03:21:56+00:00", "symbol": "Ibiden", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "LFG Ibiden!!!", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Ibiden as a beneficiary of tight substrate supply chain, with customers prepaying to secure capacity.", "resolved_tickers": ["4062.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ibiden Co. Ltd 4062.T Japan ABF substrate", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "quote", "lang": "en", "tweet_text": "“A couple of customers have prepaid for substrates, because the substrate supply chain is very tight – so I need to put up the money to secure this material…(and it shows) the commitment to me – so that’s very exciting.”\n\nLFG Ibiden!!!", "reply_text": null, "parent_text": null, "quoted_text": "Intel CEO Lip-Bu Tan sat down with CNBC’s Jim Cramer to talk about Apple; Intel 14A rivaling TSMC's top chip production technology; Shortages of CPUs and substrates; and the state of Intel's turnaround.\n\nHow to know when Intel signs Apple or other foundry customers:\nIntel CEO", "parent_author": null, "ret_m1d": 0.020561357702349792, "ret_signed_m1d": 0.020561357702349792, "alpha_spy_m1d": 0.013855833768096293, "alpha_c_m1d": 0.014096366466640697, "bench_spy_m1d": 0.006705523934253499, "bench_c_m1d": 0.006464991235709094, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02088772845953013, "ret_signed_p1d": 0.02088772845953013, "alpha_spy_p1d": 0.010638699891219794, "alpha_c_p1d": -0.0016243575068852767, "bench_spy_p1d": 0.010249028568310337, "bench_c_p1d": 0.022512085966415407, "ret_p1w": 0.3684725848563968, "ret_signed_p1w": 0.3684725848563968, "alpha_spy_p1w": 0.3454940412020302, "alpha_c_p1w": 0.2997817887402887, "bench_spy_p1w": 0.02297854365436658, "bench_c_p1w": 0.06869079611610807, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.372, -0.4034, -0.4034, -0.356, -0.3711, -0.3941, -0.3786, -0.3896, -0.4385, -0.482, -0.4723, -0.4866, -0.5284, -0.4896, -0.492, -0.5005, -0.4883, -0.469, -0.4462, -0.4474, -0.4613, -0.4613, -0.5087, -0.4895, -0.4476, -0.4488, -0.4612, -0.4924, -0.5188, -0.4732, -0.4925, -0.4641, -0.4416, -0.4309, -0.3607, -0.3597, -0.3546, -0.3873, -0.3528, -0.3877, -0.3752, -0.3972, -0.375, -0.3107, -0.2689, -0.2732, -0.1815, -0.173, -0.1802, -0.1802, -0.1201, -0.127, -0.127, -0.127, -0.127, 0.0689, 0.0127, 0.0219, 0.0803, 0.1136, 0.107, 0.0222, 0.0206, 0.0, 0.0209, 0.1668, 0.2565, 0.3662, 0.3685, 0.3332, 0.2885, 0.5013, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2042024105122394460", "ticker_idx": 0, "ts": "2026-04-08T23:37:40+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "that essentially suggests that memory companies do not expect memory prices to rise much further from here", "tweet_type": "analysis", "timeline": "months", "summary": "Author reads Samsung/Hynix push for multi-year LTAs as a bearish signal — memory companies locking in prices because they expect prices to fall, not rise.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "To be honest, the fact that they are only signing new LTAs is somewhat concerning to me.\n\nBecause that essentially suggests that memory companies do not expect memory prices to rise much further from here.\n\nThink about it. If prices were going to keep rising, would they really try to lock them in now? The only reason to lock in prices would be if they believe prices are going to come down.", "reply_text": null, "parent_text": null, "quoted_text": "One-year contracts are a thing of the past… Samsung and SK hynix won't sell without 3–5 year long-term supply agreements\n\nSamsung Electronics and SK hynix have effectively abandoned the practice of signing one-year short-term memory supply contracts with global big tech https://t.co/hzN7eJhgm6", "parent_author": null, "ret_m1d": -0.0838132923998182, "ret_signed_m1d": 0.0838132923998182, "alpha_spy_m1d": 0.05897632427811665, "alpha_c_m1d": 0.02938214986609293, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.054431142533725274, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009474065159735404, "ret_signed_p1d": 0.009474065159735404, "alpha_spy_p1d": 0.015243156330823774, "alpha_c_p1d": 0.02694778098359566, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.01747371582386026, "ret_p1w": 0.07459575510042586, "ret_signed_p1w": -0.07459575510042586, "alpha_spy_p1w": -0.03919688181439589, "alpha_c_p1w": -0.0034712265436673134, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.07112452855675855, "ret_p1m": 0.493591344196932, "ret_signed_p1m": -0.493591344196932, "alpha_spy_p1m": -0.4113884085577318, "alpha_c_p1m": -0.21651776423445052, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.2770735799624815, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.258, -0.2561, -0.2679, -0.2662, -0.2556, -0.224, -0.2208, -0.2322, -0.2066, -0.1905, -0.1852, -0.2038, -0.1542, -0.1159, -0.1117, -0.1137, -0.1363, -0.0996, -0.1321, -0.1634, -0.1558, -0.1372, -0.1501, -0.1217, -0.0926, -0.093, -0.093, -0.1027, -0.0854, -0.0699, -0.0432, -0.0438, -0.0178, 0.0009, 0.0396, 0.0223, 0.0226, -0.0777, -0.1256, -0.0683, -0.1011, -0.1371, -0.0702, -0.0486, -0.0709, -0.0673, -0.0256, -0.0024, 0.0485, 0.0077, -0.0135, -0.0733, -0.058, -0.0673, -0.1231, -0.1217, -0.1754, -0.198, -0.1088, -0.1495, -0.1223, -0.0987, -0.0838, 0.0, -0.0095, 0.0023, 0.0035, 0.0645, 0.0746, 0.0918, 0.079, 0.0834, 0.11, 0.1386, 0.1456, 0.149, 0.2023, 0.1843, 0.2, 0.188, 0.2085, 0.3075, 0.3598, 0.4841, 0.4936, 0.5813, 0.7106, 0.6622, 0.746, 0.7404, 0.6092, 0.5973, 0.572, 0.6, 0.7249, 0.705, 0.705, 0.8698, 0.9708, 0.9699, 1.0507, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2042024105122394460", "ticker_idx": 1, "ts": "2026-04-08T23:37:40+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and SK hynix won't sell without 3–5 year long-term supply agreements", "tweet_type": "analysis", "timeline": "months", "summary": "Author reads Samsung/Hynix push for multi-year LTAs as a bearish signal — memory companies locking in prices because they expect prices to fall, not rise.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "To be honest, the fact that they are only signing new LTAs is somewhat concerning to me.\n\nBecause that essentially suggests that memory companies do not expect memory prices to rise much further from here.\n\nThink about it. If prices were going to keep rising, would they really try to lock them in now? The only reason to lock in prices would be if they believe prices are going to come down.", "reply_text": null, "parent_text": null, "quoted_text": "One-year contracts are a thing of the past… Samsung and SK hynix won't sell without 3–5 year long-term supply agreements\n\nSamsung Electronics and SK hynix have effectively abandoned the practice of signing one-year short-term memory supply contracts with global big tech https://t.co/hzN7eJhgm6", "parent_author": null, "ret_m1d": -0.0665083135391924, "ret_signed_m1d": 0.0665083135391924, "alpha_spy_m1d": 0.041671345417490846, "alpha_c_m1d": 0.036442608886624095, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.030065704652568304, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.030878859857482177, "ret_signed_p1d": 0.030878859857482177, "alpha_spy_p1d": 0.03664795102857055, "alpha_c_p1d": 0.03356080548712981, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.0026819456296476307, "ret_p1w": 0.002375296912114022, "ret_signed_p1w": -0.002375296912114022, "alpha_spy_p1w": 0.03302357637391595, "alpha_c_p1w": 0.05839116834299163, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.06076646525510565, "ret_p1m": 0.2897862232779098, "ret_signed_p1m": -0.2897862232779098, "alpha_spy_p1m": -0.2075832876387096, "alpha_c_p1m": -0.09217171613172659, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.19761450714618323, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.341, -0.342, -0.3477, -0.3349, -0.3178, -0.2941, -0.2922, -0.3116, -0.2913, -0.278, -0.2789, -0.2789, -0.2438, -0.2301, -0.2382, -0.2391, -0.287, -0.2059, -0.1983, -0.2448, -0.2481, -0.2111, -0.214, -0.2045, -0.1533, -0.141, -0.141, -0.141, -0.141, -0.0993, -0.0988, -0.085, -0.0518, -0.0353, 0.0335, 0.0264, 0.0264, -0.0751, -0.1836, -0.0917, -0.1078, -0.1775, -0.1092, -0.0993, -0.1092, -0.1301, -0.1054, -0.0808, -0.0115, -0.0495, -0.0547, -0.1168, -0.1007, -0.104, -0.1462, -0.1462, -0.1625, -0.2057, -0.099, -0.1525, -0.1154, -0.0827, -0.0665, 0.0, -0.0309, -0.0214, -0.0451, -0.019, 0.0024, 0.0333, 0.0261, 0.019, 0.0404, 0.0333, 0.0665, 0.0428, 0.0665, 0.0546, 0.0736, 0.0475, 0.0475, 0.1045, 0.1045, 0.2637, 0.2898, 0.2755, 0.3563, 0.3254, 0.3492, 0.4062, 0.285, 0.3349, 0.3088, 0.3112, 0.4228, 0.3895, 0.3895, 0.4204, 0.4584, 0.4228, 0.5059, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2042024105122394460", "ticker_idx": 2, "ts": "2026-04-08T23:37:40+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung and SK hynix won't sell without 3–5 year long-term supply agreements", "tweet_type": "analysis", "timeline": "months", "summary": "Author reads Samsung/Hynix push for multi-year LTAs as a bearish signal — memory companies locking in prices because they expect prices to fall, not rise.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "To be honest, the fact that they are only signing new LTAs is somewhat concerning to me.\n\nBecause that essentially suggests that memory companies do not expect memory prices to rise much further from here.\n\nThink about it. If prices were going to keep rising, would they really try to lock them in now? The only reason to lock in prices would be if they believe prices are going to come down.", "reply_text": null, "parent_text": null, "quoted_text": "One-year contracts are a thing of the past… Samsung and SK hynix won't sell without 3–5 year long-term supply agreements\n\nSamsung Electronics and SK hynix have effectively abandoned the practice of signing one-year short-term memory supply contracts with global big tech https://t.co/hzN7eJhgm6", "parent_author": null, "ret_m1d": -0.11326233929353025, "ret_signed_m1d": 0.11326233929353025, "alpha_spy_m1d": 0.08842537117182869, "alpha_c_m1d": 0.05883119675980497, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.054431142533725274, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03388193464335787, "ret_signed_p1d": 0.03388193464335787, "alpha_spy_p1d": 0.03965102581444624, "alpha_c_p1d": 0.05135565046721813, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.01747371582386026, "ret_p1w": 0.09970961384699062, "ret_signed_p1w": -0.09970961384699062, "alpha_spy_p1w": -0.06431074056096064, "alpha_c_p1w": -0.02858508529023207, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.07112452855675855, "ret_p1m": 0.6011616656390466, "ret_signed_p1m": -0.6011616656390466, "alpha_spy_p1m": -0.5189587299998464, "alpha_c_p1m": -0.3240880856765651, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.2770735799624815, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2811, -0.2763, -0.2869, -0.283, -0.2763, -0.2695, -0.2618, -0.2821, -0.285, -0.2705, -0.2589, -0.2888, -0.227, -0.1874, -0.168, -0.1217, -0.198, -0.1236, -0.1304, -0.1864, -0.1893, -0.1429, -0.1535, -0.169, -0.142, -0.1497, -0.1497, -0.1497, -0.1497, -0.1362, -0.083, -0.0811, -0.0289, -0.0163, 0.0639, 0.0271, 0.0271, -0.091, -0.1781, -0.0891, -0.1055, -0.1907, -0.092, -0.0755, -0.0997, -0.1191, -0.0571, -0.061, 0.0223, -0.0194, -0.0252, -0.0968, -0.0455, -0.0368, -0.0968, -0.0968, -0.1549, -0.2188, -0.1317, -0.1965, -0.152, -0.1423, -0.1133, 0.0, -0.0339, -0.0058, 0.0068, 0.0678, 0.0997, 0.1181, 0.092, 0.1288, 0.1849, 0.1839, 0.1859, 0.183, 0.2507, 0.2585, 0.2517, 0.2449, 0.2449, 0.4008, 0.4008, 0.5499, 0.6012, 0.6321, 0.8199, 0.7764, 0.9129, 0.9071, 0.7609, 0.7812, 0.6893, 0.6893, 0.878, 0.879, 0.879, 0.9864, 1.1713, 1.2162, 1.2588, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2052683022391259386", "ticker_idx": 0, "ts": "2026-05-08T09:32:24+00:00", "symbol": "Unimicron", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unimicron expects revenue growth to accelerate further in 2Q as the impact of price hikes is gradually reflected. Unimicron's utilization rates for ABF, BT substrates, and HDI all reached around 90% in 1Q and are expected to remain at high levels in 2Q", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Unimicron: revenue growth accelerating in 2Q on price hikes, 90% utilization, 40% ABF capacity expansion, AI revenue rising from 40% to 60%.", "resolved_tickers": ["3037.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Unimicron Technology listed on TWSE as 3037.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "▶ Taiwanese ABF + HDI substrate maker Unimicron reported April revenue of NT$13,932.8mn, up 6.5% MoM and 27.6% YoY\n\n- Monthly revenue hit a new all-time high.\n- Supply-demand imbalance for ABF substrates is intensifying, driven by AI server platform upgrades and rising demand for high-speed computing.\n- Unimicron expects revenue growth to accelerate further in 2Q as the impact of price hikes is gradually reflected.\n- Unimicron’s utilization rates for ABF, BT substrates, and HDI all reached around 90% in 1Q and are expected to remain at high levels in 2Q.\n- The company plans to increase ABF substrate capacity by around 40% this year, while AI-related revenue contribution is expected to rise from about 40% last year to 60% this year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0953545232273838, "ret_signed_m1d": 0.0953545232273838, "alpha_spy_m1d": 0.1035429916356787, "alpha_c_m1d": 0.12857012081821306, "bench_spy_m1d": -0.008188468408294902, "bench_c_m1d": -0.03321559759082926, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05256723716381417, "ret_signed_p1d": 0.05256723716381417, "alpha_spy_p1d": 0.05028965156838261, "alpha_c_p1d": 0.039121472846898486, "bench_spy_p1d": 0.002277585595431564, "bench_c_p1d": 0.013445764316915687, "ret_p1w": 0.003667481662591676, "ret_signed_p1w": 0.003667481662591676, "alpha_spy_p1w": 0.001566145197705815, "alpha_c_p1w": -0.000548506352234801, "bench_spy_p1w": 0.0021013364648858612, "bench_c_p1w": 0.004215988014826477, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5458, -0.5489, -0.5489, -0.5489, -0.5489, -0.5489, -0.5489, -0.5489, -0.5489, -0.5043, -0.4548, -0.4395, -0.4114, -0.4114, -0.412, -0.434, -0.4902, -0.4499, -0.4719, -0.5244, -0.4817, -0.4303, -0.4132, -0.379, -0.3289, -0.3362, -0.291, -0.2897, -0.3227, -0.39, -0.4377, -0.3814, -0.3863, -0.3826, -0.3967, -0.4566, -0.4028, -0.3655, -0.3655, -0.3655, -0.3105, -0.2421, -0.2359, -0.22, -0.2347, -0.2702, -0.2445, -0.2115, -0.2139, -0.1711, -0.132, -0.1247, -0.11, -0.0342, 0.0257, 0.0086, -0.0183, 0.0795, 0.0795, 0.1137, 0.1039, 0.0489, 0.0954, 0.0, 0.0526, 0.0697, 0.0905, 0.077, 0.0037, -0.0024, 0.0, 0.0061, 0.1064, 0.1858, 0.2103, 0.3264, 0.3203, 0.2531, 0.2897, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2043117943194890240", "ticker_idx": 0, "ts": "2026-04-12T00:04:11+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "my ASML shares can go up", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long ASML; hopes TSMC capex guidance boosts ASML shares.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "I hope C.C. Wei blesses us with even more capex on this earnings call, so my ASML shares can go up.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.008096420951350103, "ret_signed_m1d": 0.008096420951350103, "alpha_spy_m1d": 0.017774243206934104, "alpha_c_m1d": 0.022667611253239017, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.014571190301888914, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01936807460209078, "ret_signed_p1d": 0.01936807460209078, "alpha_spy_p1d": 0.007183193849720659, "alpha_c_p1d": -0.00016547466854421877, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.019533549270634998, "ret_p1w": -0.011747982255521316, "ret_signed_p1w": -0.011747982255521316, "alpha_spy_p1w": -0.04471693127475107, "alpha_c_p1w": -0.058258550414302945, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.04651056815878163, "ret_p1m": 0.006984307764408726, "ret_signed_p1m": 0.006984307764408726, "alpha_spy_p1m": -0.06892302198788691, "alpha_c_p1m": -0.25897410933325493, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.26595841709766366, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1405, -0.0889, -0.0747, -0.1118, -0.0963, -0.0846, -0.0675, -0.066, -0.0662, -0.0348, -0.0532, -0.0551, -0.0364, -0.0291, -0.0564, -0.0958, -0.0887, -0.0537, -0.0437, -0.053, -0.0413, -0.0635, -0.0551, -0.0514, -0.0481, -0.0119, -0.0171, -0.0033, -0.0084, 0.0029, 0.0227, -0.0217, -0.021, -0.0392, -0.0778, -0.0476, -0.0586, -0.0895, -0.0891, -0.0475, -0.0484, -0.063, -0.0637, -0.0495, -0.0524, -0.0546, -0.0724, -0.1045, -0.067, -0.0462, -0.0383, -0.0746, -0.09, -0.1173, -0.1116, -0.0573, -0.0784, -0.0784, -0.0784, -0.1159, -0.0375, -0.0198, 0.0081, 0.0, 0.0194, -0.0237, -0.0295, -0.0125, -0.0117, -0.0181, -0.0079, -0.0286, -0.0059, -0.0353, -0.0678, -0.0512, -0.0275, -0.0275, -0.0554, -0.0223, 0.0382, 0.0342, 0.0496, 0.0386, 0.007, 0.0557, 0.0875, 0.0394, 0.0059, -0.0064, 0.0604, 0.0701, 0.1209, 0.1381, 0.1056, 0.095, 0.1066, 0.1017, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2056988526550687855", "ticker_idx": 0, "ts": "2026-05-20T06:40:56+00:00", "symbol": "Murata Manufacturing", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I recently added to my Murata position", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Added to Murata position; ongoing long.", "resolved_tickers": ["6981.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Murata Manufacturing 6981.T Japan MLCC leader", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "It feels like everyone is gradually starting to realize this.\n\nP.S. I recently added to my Murata position. https://t.co/trMvuFiEeh", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002595717066839809, "ret_signed_m1d": 0.002595717066839809, "alpha_spy_m1d": 0.012740768711021921, "alpha_c_m1d": 0.024612166823654524, "bench_spy_m1d": -0.010145051644182113, "bench_c_m1d": -0.022016449756814715, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09133679428942254, "ret_signed_p1d": 0.09133679428942254, "alpha_spy_p1d": 0.08935369721943931, "alpha_c_p1d": 0.0830946879390908, "bench_spy_p1d": 0.0019830970699832307, "bench_c_p1d": 0.008242106350331735, "ret_p1w": 0.26865671641791056, "ret_signed_p1w": 0.26865671641791056, "alpha_spy_p1w": 0.2562317289337199, "alpha_c_p1w": 0.2275028645987216, "bench_spy_p1w": 0.01242498748419063, "bench_c_p1w": 0.04115385181918896, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4094, -0.4094, -0.3499, -0.3213, -0.343, -0.3398, -0.3467, -0.4005, -0.4123, -0.388, -0.392, -0.4418, -0.4099, -0.396, -0.4021, -0.4134, -0.4203, -0.389, -0.3822, -0.3976, -0.3976, -0.4362, -0.4306, -0.4118, -0.4017, -0.3981, -0.4267, -0.447, -0.4044, -0.4358, -0.3853, -0.3824, -0.3858, -0.3511, -0.3527, -0.3274, -0.3235, -0.2977, -0.2667, -0.2226, -0.2537, -0.2313, -0.2197, -0.2033, -0.2192, -0.1992, -0.2025, -0.2148, -0.2148, -0.1635, -0.1665, -0.1665, -0.1665, -0.1665, -0.0837, -0.0509, -0.0331, -0.014, -0.0122, 0.0284, -0.0021, 0.0011, 0.0026, 0.0, 0.0913, 0.1567, 0.2862, 0.3104, 0.2687, 0.3851, 0.5615, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2042181382311473295", "ticker_idx": 0, "ts": "2026-04-09T10:02:38+00:00", "symbol": "AMZN", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Anthropic-related revenue alone could contribute over $1.3 billion in quarter-over-quarter growth to AWS revenue in Q1. This significantly exceeds the current Street consensus", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "BofA: Anthropic cloud spend on AWS could drive $1.3B+ Q1 QoQ growth, well above ~$1B consensus, plus $1B+ more in Q2; bullish AMZN upside surprise.", "resolved_tickers": ["AMZN"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLY", "bench_c_reason": "industry='Internet Retail'", "bench_c_industry": "Internet Retail", "unit_kind": "original", "lang": "en", "tweet_text": "BofA: Anthropic's Claude Revenue Growth Trajectory Could Surprise AWS to the Upside\n\nAnthropic's ARR increased by approximately $21 billion during Q1 (from $9 billion to $30 billion, see Exhibit 1), implying a quarter-over-quarter revenue increase of over $3 billion. Given that a significant portion of Anthropic's workloads run on AWS (Anthropic has reaffirmed that AWS remains its primary cloud provider and training partner), and factoring in the quarter-over-quarter increase in training-related spending, we estimate that Anthropic-related revenue alone could contribute over $1.3 billion in quarter-over-quarter growth to AWS revenue in Q1. This significantly exceeds the current Street consensus, which implies approximately $1 billion in quarter-over-quarter growth (we estimate Amazon's total quarter-over-quarter growth could reach $2 billion). Looking ahead to Q2, we project that Anthropic alone could contribute an additional $1 billion-plus in quarter-over-quarter growth.\n\n$AMZN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.053070807705554834, "ret_signed_m1d": -0.053070807705554834, "alpha_spy_m1d": -0.0473348080394973, "alpha_c_m1d": -0.03604048833952678, "bench_spy_m1d": -0.005735999666057534, "bench_c_m1d": -0.017030319366028057, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020244002182271403, "ret_signed_p1d": 0.020244002182271403, "alpha_spy_p1d": 0.020905782696820974, "alpha_c_p1d": 0.01891349367844697, "bench_spy_p1d": -0.0006617805145495703, "bench_c_p1d": 0.0013305085038244346, "ret_p1w": 0.06869250362090118, "ret_signed_p1w": 0.06869250362090118, "alpha_spy_p1w": 0.03670297433159253, "alpha_c_p1w": 0.02531837303454698, "bench_spy_p1w": 0.03198952928930865, "bench_c_p1w": 0.0433741305863542, "ret_p1m": 0.16704472415517646, "ret_signed_p1m": 0.16704472415517646, "alpha_spy_p1m": 0.0821658074572027, "alpha_c_p1m": 0.10087478246787018, "bench_spy_p1m": 0.08487891669797376, "bench_c_p1m": 0.06616994168730628, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0549, 0.0383, 0.0128, 0.0194, 0.0234, 0.0234, -0.0113, -0.01, 0.003, 0.0236, 0.0204, 0.0472, 0.0401, 0.0346, 0.0242, 0.0398, 0.0213, -0.0028, -0.0469, -0.0999, -0.1067, -0.1142, -0.1266, -0.1457, -0.1492, -0.1492, -0.1391, -0.1235, -0.1232, -0.1007, -0.1215, -0.1074, -0.0985, -0.1101, -0.1012, -0.1081, -0.1067, -0.072, -0.063, -0.0875, -0.0863, -0.0827, -0.0899, -0.1032, -0.1112, -0.0938, -0.079, -0.1018, -0.1065, -0.121, -0.1006, -0.113, -0.0939, -0.1117, -0.1468, -0.14, -0.1086, -0.0988, -0.1022, -0.1022, -0.0893, -0.0851, -0.0531, 0.0, 0.0202, 0.0267, 0.0658, 0.0636, 0.0687, 0.0724, 0.0626, 0.0696, 0.0929, 0.0917, 0.1299, 0.1176, 0.1115, 0.1258, 0.1344, 0.1481, 0.1643, 0.1708, 0.1769, 0.1606, 0.167, 0.1513, 0.1377, 0.1561, 0.1437, 0.1305, 0.1336, 0.11, 0.1342, 0.149, 0.1398, 0.1398, 0.1354, 0.1635, 0.1727, 0.1583, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048660802736140321", "ticker_idx": 0, "ts": "2026-04-27T07:09:32+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Advantest clearly proved why its stock deserves to trade at a premium", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish Advantest: SoC share 56%→66%, first silicon photonics ATE order; stock deserves premium valuation.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Advantest earnings highlights:\n\n- SoC tester market share increased from 56% in CY24 to 66% in CY25, clearly reinforcing its advantage over Teradyne.\n- The company received its first high-volume ATE order for silicon photonics.\n\nOne-line take: Advantest clearly proved why its stock deserves to trade at a premium.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06539682539682534, "ret_signed_m1d": -0.06539682539682534, "alpha_spy_m1d": -0.06367698186571824, "alpha_c_m1d": -0.0657523594550351, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05555555555555558, "ret_signed_p1d": -0.05555555555555558, "alpha_spy_p1d": -0.05068960682740942, "alpha_c_p1d": -0.025827712114960044, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": -0.11698412698412697, "ret_signed_p1w": -0.11698412698412697, "alpha_spy_p1w": -0.12095524842844729, "alpha_c_p1w": -0.1180310174535959, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": -0.17222222222222228, "ret_signed_p1m": -0.17222222222222228, "alpha_spy_p1m": -0.22174897081586709, "alpha_c_p1m": -0.36161107979207174, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1903, -0.1485, -0.1914, -0.2295, -0.1748, -0.1924, -0.2312, -0.2223, -0.1328, -0.1226, -0.1226, -0.15, -0.1399, -0.1409, -0.1505, -0.1453, -0.1759, -0.1924, -0.1924, -0.1561, -0.0928, -0.1084, -0.1488, -0.1821, -0.1844, -0.2233, -0.1905, -0.1849, -0.2748, -0.2368, -0.2111, -0.2233, -0.2502, -0.2347, -0.2536, -0.2033, -0.2398, -0.2398, -0.2794, -0.2916, -0.2574, -0.2719, -0.3003, -0.3357, -0.3546, -0.2857, -0.3294, -0.3157, -0.3038, -0.2952, -0.1994, -0.2127, -0.2067, -0.2102, -0.1429, -0.1243, -0.0913, -0.1152, -0.1397, -0.1365, -0.1143, -0.1143, -0.0654, 0.0, -0.0556, -0.0556, -0.1029, -0.117, -0.117, -0.117, -0.117, -0.0567, -0.0513, -0.0863, -0.0887, -0.1019, -0.0916, -0.1632, -0.1698, -0.1971, -0.1865, -0.151, -0.1478, -0.1189, -0.1722, -0.1387, -0.1638, -0.1692, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2038890721026191659", "ticker_idx": 0, "ts": "2026-03-31T08:06:43+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think ASML will continue to sell a substantial amount of DUV equipment to China going forward", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish ASML on continued substantial DUV equipment sales to China going forward.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "I think ASML will continue to sell a substantial amount of DUV equipment to China going forward. ASML probably knows that too, which is why they set up such a large booth at Semicon China, right?", "reply_text": null, "parent_text": null, "quoted_text": "At Semicon Shanghai one of the most interesting booth’s on display was the ASML booth. Last year it was a single space in a corner of a hall, with a few staff. This year, it was front and center taking up several booths’ worth of space and many staffers. 2025 vs 2026 https://t.co/odaepYLvg3", "parent_author": null, "ret_m1d": -0.006433137325160709, "ret_signed_m1d": -0.006433137325160709, "alpha_spy_m1d": 0.021813711082997123, "alpha_c_m1d": 0.04800086228210654, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06111513252031142, "ret_signed_p1d": 0.06111513252031142, "alpha_spy_p1d": 0.0535806687374214, "alpha_c_p1d": 0.03876247873666894, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": -0.0048248529938704765, "ret_signed_p1w": -0.0048248529938704765, "alpha_spy_p1w": -0.018479162248972947, "alpha_c_p1w": -0.047860847419198715, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.06793872669529, "ret_signed_p1m": 0.06793872669529, "alpha_spy_p1m": -0.026227382886621076, "alpha_c_p1m": -0.2350868197974132, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1778, -0.1199, -0.0602, -0.0533, -0.0615, -0.0961, -0.0349, -0.0306, -0.0162, -0.0326, 0.0256, 0.0415, -0.0003, 0.0172, 0.0304, 0.0497, 0.0513, 0.0511, 0.0865, 0.0658, 0.0636, 0.0847, 0.0929, 0.0622, 0.0178, 0.0258, 0.0652, 0.0765, 0.0659, 0.0792, 0.0541, 0.0636, 0.0677, 0.0715, 0.1122, 0.1063, 0.1219, 0.1162, 0.1288, 0.1512, 0.1011, 0.102, 0.0815, 0.0381, 0.072, 0.0597, 0.0248, 0.0254, 0.0722, 0.0711, 0.0547, 0.054, 0.0699, 0.0667, 0.0642, 0.0441, 0.008, 0.0502, 0.0736, 0.0826, 0.0416, 0.0243, -0.0064, 0.0, 0.0611, 0.0373, 0.0373, 0.0373, -0.0048, 0.0835, 0.1033, 0.1347, 0.1256, 0.1474, 0.099, 0.0924, 0.1115, 0.1124, 0.1053, 0.1167, 0.0935, 0.119, 0.0858, 0.0493, 0.0679, 0.0946, 0.0946, 0.0633, 0.1005, 0.1686, 0.1641, 0.1815, 0.1691, 0.1335, 0.1883, 0.2241, 0.17, 0.1322, 0.1184, 0.1937, 0.2046, 0.2617, 0.2811, 0.2445, 0.2325, 0.2456, 0.24, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2051070393734303767", "ticker_idx": 0, "ts": "2026-05-03T22:44:23+00:00", "symbol": "Doosan", "kind": "equity", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Doosan, known to be the sole supplier of CCL for Blackwell, saw its share price jump from 152,300 won at the end of April 2024 to 1,596,000 won at the end of last month", "tweet_type": "analysis", "timeline": "unspecified", "summary": "CCL shortage crisis with prices +74.5% YoY drives bullish case for Doosan (sole Blackwell CCL supplier), Samsung Electro-Mechanics, and Daeduck Electronics as demand surges across AI/data center/auto.", "resolved_tickers": ["006280.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Doosan Electronics (CCL for Blackwell) 006280.KS", "bench_c": "EWY", "bench_c_reason": "country_fallback (suffix='KS')", "bench_c_industry": "Drug Manufacturers - Specialty & Generic", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Even China Has No Volumes\"... CCL Shortage Reaches Crisis Level\n\nA PCB manufacturer in the Seoul metropolitan area recently placed advance orders worth 10 billion won with two Taiwanese CCL (copper clad laminate) producers, EMC and TUC. The volume is more than five times the company's average monthly usage of 1.5 to 2 billion won. The company's CEO said, \"We placed bulk orders out of concern about a CCL shortage, but we have no idea when they'll actually arrive,\" adding, \"I've been in the PCB business for over 20 years, and this is the first time we've been unable to manufacture products because we couldn't get CCL.\"\n\nCCL Import Prices Break Through $20,000 Per Ton\n\nPrices for CCL — a core PCB material — are surging as supply falls dramatically short of demand. CCL is a panel made by laminating a thin copper foil onto an insulating substrate, and it functions as the key building-block material for PCBs. The shortage has spread as demand has surged simultaneously across major advanced industries, including AI semiconductors, data centers, and autonomous vehicles.\n\nAccording to data released by the Korea Customs Service on May 3, the import unit price of CCL came to $20,728 per ton in March — up 74.5% from $11,880 in the same month last year. This is the first time the CCL import unit price has broken above $20,000/ton since the relevant statistics began being compiled in 2000. The biggest driver behind the spike in CCL import prices is rising demand for CCL used in AI chips. On top of this, the use of high-spec CCL is also increasing across 5G/6G telecom infrastructure, automotive autonomous driving systems, and data center servers.\n\nCCL Players' Share Prices Rally Sharply\n\nKorea's CCL supply chain is built around copper foil suppliers (Lotte Energy Materials, SK Nexilis), CCL manufacturers (Doosan, LG Chem), and PCB substrate manufacturers (Samsung Electro-Mechanics, Daeduck Electronics). PCBs are split between premium products — which go into the most advanced GPUs such as Nvidia's Blackwell — and general-purpose products. Premium semiconductor substrates in this top tier use CCL based on high-grade T-glass material, while small- and mid-sized domestic PCB makers use E-glass-based CCL to produce general-purpose PCBs.\n\nAs demand for high-spec PCBs has surged and unit prices have jumped, CCL-related companies have moved quickly to reallocate their limited production lines toward high value-added products. This is the backdrop to the spike in CCL export prices: last month, the average CCL export unit price stood at $30,998, up 65.2% year-on-year.\n\nCompanies in the global CCL manufacturing supply chain are seeing both earnings and share prices rally sharply. Industry participants attribute the CCL price hikes primarily to PCBs going into Nvidia GPUs. Doosan, known to be the sole supplier of CCL for Blackwell, saw its share price jump from 152,300 won at the end of April 2024 to 1,596,000 won at the end of last month — a more than 10x gain over two years. Over the same period, Samsung Electro-Mechanics rose 5.3x and Daeduck Electronics rose 4.8x.\n\nEven Pivoting to China — \"No Volumes Available\"\n\nBy contrast, companies left out of the advanced industries are unable to secure adequate CCL supply, with some now facing the prospect of having to halt factory operations. Major Korean semiconductor equipment makers, in particular, are reportedly suffering from CCL supply constraints. Some companies are reportedly switching from ocean freight to air freight in an effort to pull forward their CCL delivery schedules. One industry source said, \"In the past, you could secure the volumes you wanted just by waiting about a month after placing an order, but now even if you place an order today, you have to wait at least six months to receive any volumes.\"\n\nSoaring CCL prices are also a heavy burden. With FX rates and oil prices also rising, not only product prices but also transport costs are climbing sharply. With CCL supply schedules becoming increasingly uncertain, many firms are worried about missing the PCB delivery timelines they originally committed to. Another industry source said, \"We've secured Chinese suppliers in addition to our existing channels, but we've been hit by a string of delivery-delay notices recently. There's no clear-cut countermeasure, and it feels hopeless.\"\n\n---\n\nhttps://t.co/lOgUHYzluY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002909090909090861, "ret_signed_m1d": -0.002909090909090861, "alpha_spy_m1d": -0.006585940274045021, "alpha_c_m1d": 0.00681030742506239, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": -0.00971939833415325, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.060394920863043655, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.060394920863043655, "ret_p1w": -0.02400000000000002, "ret_signed_p1w": -0.02400000000000002, "alpha_spy_p1w": -0.05365136660740055, "alpha_c_p1w": -0.2028618523176029, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.17886185231760288, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1794, 0.2241, 0.183, 0.2198, 0.2176, 0.2586, 0.2982, 0.255, 0.255, 0.255, 0.255, 0.2781, 0.2428, 0.2464, 0.2738, 0.214, 0.1989, 0.2176, 0.2176, 0.1672, -0.0013, 0.0772, 0.0952, 0.0383, 0.0909, 0.0902, 0.0743, 0.0959, 0.0765, 0.0678, 0.0866, 0.0455, 0.0578, 0.0282, 0.0405, 0.0729, 0.0678, 0.0873, 0.0545, 0.0393, 0.08, 0.0262, 0.0225, 0.0182, 0.0065, 0.0371, 0.0495, 0.0495, 0.0349, 0.0473, 0.0764, 0.088, 0.0996, 0.0742, 0.0473, 0.0349, 0.0342, 0.0393, 0.0298, 0.0189, 0.0124, -0.0029, -0.0029, 0.0, 0.0, -0.0124, -0.0065, 0.0044, -0.024, -0.0015, 0.0051, 0.0531, 0.0087, -0.0233, -0.0495, -0.0575, -0.0407, 0.0109, 0.0109, 0.0415, 0.0924, 0.0378, 0.0284, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2051070393734303767", "ticker_idx": 1, "ts": "2026-05-03T22:44:23+00:00", "symbol": "Samsung Electro-Mechanics", "kind": "equity", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electro-Mechanics rose 5.3x", "tweet_type": "analysis", "timeline": "unspecified", "summary": "CCL shortage crisis with prices +74.5% YoY drives bullish case for Doosan (sole Blackwell CCL supplier), Samsung Electro-Mechanics, and Daeduck Electronics as demand surges across AI/data center/auto.", "resolved_tickers": ["009150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electro-Mechanics listed on KOSPI as 009150.KS", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Even China Has No Volumes\"... CCL Shortage Reaches Crisis Level\n\nA PCB manufacturer in the Seoul metropolitan area recently placed advance orders worth 10 billion won with two Taiwanese CCL (copper clad laminate) producers, EMC and TUC. The volume is more than five times the company's average monthly usage of 1.5 to 2 billion won. The company's CEO said, \"We placed bulk orders out of concern about a CCL shortage, but we have no idea when they'll actually arrive,\" adding, \"I've been in the PCB business for over 20 years, and this is the first time we've been unable to manufacture products because we couldn't get CCL.\"\n\nCCL Import Prices Break Through $20,000 Per Ton\n\nPrices for CCL — a core PCB material — are surging as supply falls dramatically short of demand. CCL is a panel made by laminating a thin copper foil onto an insulating substrate, and it functions as the key building-block material for PCBs. The shortage has spread as demand has surged simultaneously across major advanced industries, including AI semiconductors, data centers, and autonomous vehicles.\n\nAccording to data released by the Korea Customs Service on May 3, the import unit price of CCL came to $20,728 per ton in March — up 74.5% from $11,880 in the same month last year. This is the first time the CCL import unit price has broken above $20,000/ton since the relevant statistics began being compiled in 2000. The biggest driver behind the spike in CCL import prices is rising demand for CCL used in AI chips. On top of this, the use of high-spec CCL is also increasing across 5G/6G telecom infrastructure, automotive autonomous driving systems, and data center servers.\n\nCCL Players' Share Prices Rally Sharply\n\nKorea's CCL supply chain is built around copper foil suppliers (Lotte Energy Materials, SK Nexilis), CCL manufacturers (Doosan, LG Chem), and PCB substrate manufacturers (Samsung Electro-Mechanics, Daeduck Electronics). PCBs are split between premium products — which go into the most advanced GPUs such as Nvidia's Blackwell — and general-purpose products. Premium semiconductor substrates in this top tier use CCL based on high-grade T-glass material, while small- and mid-sized domestic PCB makers use E-glass-based CCL to produce general-purpose PCBs.\n\nAs demand for high-spec PCBs has surged and unit prices have jumped, CCL-related companies have moved quickly to reallocate their limited production lines toward high value-added products. This is the backdrop to the spike in CCL export prices: last month, the average CCL export unit price stood at $30,998, up 65.2% year-on-year.\n\nCompanies in the global CCL manufacturing supply chain are seeing both earnings and share prices rally sharply. Industry participants attribute the CCL price hikes primarily to PCBs going into Nvidia GPUs. Doosan, known to be the sole supplier of CCL for Blackwell, saw its share price jump from 152,300 won at the end of April 2024 to 1,596,000 won at the end of last month — a more than 10x gain over two years. Over the same period, Samsung Electro-Mechanics rose 5.3x and Daeduck Electronics rose 4.8x.\n\nEven Pivoting to China — \"No Volumes Available\"\n\nBy contrast, companies left out of the advanced industries are unable to secure adequate CCL supply, with some now facing the prospect of having to halt factory operations. Major Korean semiconductor equipment makers, in particular, are reportedly suffering from CCL supply constraints. Some companies are reportedly switching from ocean freight to air freight in an effort to pull forward their CCL delivery schedules. One industry source said, \"In the past, you could secure the volumes you wanted just by waiting about a month after placing an order, but now even if you place an order today, you have to wait at least six months to receive any volumes.\"\n\nSoaring CCL prices are also a heavy burden. With FX rates and oil prices also rising, not only product prices but also transport costs are climbing sharply. With CCL supply schedules becoming increasingly uncertain, many firms are worried about missing the PCB delivery timelines they originally committed to. Another industry source said, \"We've secured Chinese suppliers in addition to our existing channels, but we've been hit by a string of delivery-delay notices recently. There's no clear-cut countermeasure, and it feels hopeless.\"\n\n---\n\nhttps://t.co/lOgUHYzluY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09368191721132901, "ret_signed_m1d": -0.09368191721132901, "alpha_spy_m1d": -0.09735876657628317, "alpha_c_m1d": -0.09257109998713664, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": -0.001110817224192373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.022091957813239027, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.022091957813239027, "ret_p1w": -0.019607843137254943, "ret_signed_p1w": -0.019607843137254943, "alpha_spy_p1w": -0.049259209744655474, "alpha_c_p1w": -0.11729375201069503, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.09768590887344009, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6678, -0.6841, -0.6906, -0.6688, -0.6585, -0.6629, -0.652, -0.6629, -0.6629, -0.6629, -0.6629, -0.61, -0.5893, -0.5354, -0.5142, -0.4935, -0.4918, -0.5114, -0.5114, -0.5545, -0.6051, -0.5626, -0.5594, -0.5839, -0.5621, -0.5577, -0.5648, -0.5637, -0.5501, -0.5136, -0.4946, -0.4777, -0.494, -0.5343, -0.524, -0.5049, -0.5022, -0.5022, -0.5327, -0.5561, -0.5163, -0.5458, -0.5033, -0.4967, -0.5022, -0.4401, -0.4379, -0.3845, -0.3813, -0.3638, -0.3312, -0.3039, -0.2603, -0.2593, -0.159, -0.1155, -0.1569, -0.1416, -0.1351, -0.0861, -0.0991, -0.0937, -0.0937, 0.0, 0.0, -0.0065, -0.0011, -0.0044, -0.0196, 0.0436, 0.1209, 0.1155, 0.1002, 0.1231, 0.0752, 0.1558, 0.3115, 0.4597, 0.4597, 0.7124, 0.7756, 1.0142, 1.317, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2051070393734303767", "ticker_idx": 2, "ts": "2026-05-03T22:44:23+00:00", "symbol": "Daeduck Electronics", "kind": "equity", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Daeduck Electronics rose 4.8x", "tweet_type": "analysis", "timeline": "unspecified", "summary": "CCL shortage crisis with prices +74.5% YoY drives bullish case for Doosan (sole Blackwell CCL supplier), Samsung Electro-Mechanics, and Daeduck Electronics as demand surges across AI/data center/auto.", "resolved_tickers": ["353200.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Daeduck Electronics 353200.KS Korea PCB maker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Even China Has No Volumes\"... CCL Shortage Reaches Crisis Level\n\nA PCB manufacturer in the Seoul metropolitan area recently placed advance orders worth 10 billion won with two Taiwanese CCL (copper clad laminate) producers, EMC and TUC. The volume is more than five times the company's average monthly usage of 1.5 to 2 billion won. The company's CEO said, \"We placed bulk orders out of concern about a CCL shortage, but we have no idea when they'll actually arrive,\" adding, \"I've been in the PCB business for over 20 years, and this is the first time we've been unable to manufacture products because we couldn't get CCL.\"\n\nCCL Import Prices Break Through $20,000 Per Ton\n\nPrices for CCL — a core PCB material — are surging as supply falls dramatically short of demand. CCL is a panel made by laminating a thin copper foil onto an insulating substrate, and it functions as the key building-block material for PCBs. The shortage has spread as demand has surged simultaneously across major advanced industries, including AI semiconductors, data centers, and autonomous vehicles.\n\nAccording to data released by the Korea Customs Service on May 3, the import unit price of CCL came to $20,728 per ton in March — up 74.5% from $11,880 in the same month last year. This is the first time the CCL import unit price has broken above $20,000/ton since the relevant statistics began being compiled in 2000. The biggest driver behind the spike in CCL import prices is rising demand for CCL used in AI chips. On top of this, the use of high-spec CCL is also increasing across 5G/6G telecom infrastructure, automotive autonomous driving systems, and data center servers.\n\nCCL Players' Share Prices Rally Sharply\n\nKorea's CCL supply chain is built around copper foil suppliers (Lotte Energy Materials, SK Nexilis), CCL manufacturers (Doosan, LG Chem), and PCB substrate manufacturers (Samsung Electro-Mechanics, Daeduck Electronics). PCBs are split between premium products — which go into the most advanced GPUs such as Nvidia's Blackwell — and general-purpose products. Premium semiconductor substrates in this top tier use CCL based on high-grade T-glass material, while small- and mid-sized domestic PCB makers use E-glass-based CCL to produce general-purpose PCBs.\n\nAs demand for high-spec PCBs has surged and unit prices have jumped, CCL-related companies have moved quickly to reallocate their limited production lines toward high value-added products. This is the backdrop to the spike in CCL export prices: last month, the average CCL export unit price stood at $30,998, up 65.2% year-on-year.\n\nCompanies in the global CCL manufacturing supply chain are seeing both earnings and share prices rally sharply. Industry participants attribute the CCL price hikes primarily to PCBs going into Nvidia GPUs. Doosan, known to be the sole supplier of CCL for Blackwell, saw its share price jump from 152,300 won at the end of April 2024 to 1,596,000 won at the end of last month — a more than 10x gain over two years. Over the same period, Samsung Electro-Mechanics rose 5.3x and Daeduck Electronics rose 4.8x.\n\nEven Pivoting to China — \"No Volumes Available\"\n\nBy contrast, companies left out of the advanced industries are unable to secure adequate CCL supply, with some now facing the prospect of having to halt factory operations. Major Korean semiconductor equipment makers, in particular, are reportedly suffering from CCL supply constraints. Some companies are reportedly switching from ocean freight to air freight in an effort to pull forward their CCL delivery schedules. One industry source said, \"In the past, you could secure the volumes you wanted just by waiting about a month after placing an order, but now even if you place an order today, you have to wait at least six months to receive any volumes.\"\n\nSoaring CCL prices are also a heavy burden. With FX rates and oil prices also rising, not only product prices but also transport costs are climbing sharply. With CCL supply schedules becoming increasingly uncertain, many firms are worried about missing the PCB delivery timelines they originally committed to. Another industry source said, \"We've secured Chinese suppliers in addition to our existing channels, but we've been hit by a string of delivery-delay notices recently. There's no clear-cut countermeasure, and it feels hopeless.\"\n\n---\n\nhttps://t.co/lOgUHYzluY", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013986013986013957, "ret_signed_m1d": -0.013986013986013957, "alpha_spy_m1d": -0.017662863350968117, "alpha_c_m1d": -0.012875196761821583, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": -0.001110817224192373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.022091957813239027, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.022091957813239027, "ret_p1w": 0.09965034965034958, "ret_signed_p1w": 0.09965034965034958, "alpha_spy_p1w": 0.06999898304294905, "alpha_c_p1w": 0.001964440776909493, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.09768590887344009, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4655, -0.4907, -0.4881, -0.4707, -0.4985, -0.502, -0.4837, -0.5002, -0.5002, -0.5002, -0.5002, -0.482, -0.4924, -0.4776, -0.4351, -0.4498, -0.4325, -0.4559, -0.4559, -0.4898, -0.5663, -0.4846, -0.4203, -0.4768, -0.4385, -0.4568, -0.4464, -0.4498, -0.4177, -0.4159, -0.3499, -0.3282, -0.2804, -0.3116, -0.3255, -0.2299, -0.2378, -0.2378, -0.2771, -0.3295, -0.2316, -0.2981, -0.2902, -0.292, -0.2788, -0.2255, -0.236, -0.2142, -0.2483, -0.2264, -0.2395, -0.1844, -0.1757, -0.1871, -0.1233, -0.1154, -0.1355, 0.007, -0.049, -0.0446, 0.0087, -0.014, -0.014, 0.0, 0.0, 0.0962, 0.0551, 0.0769, 0.0997, 0.1189, 0.2491, 0.264, 0.1871, 0.2273, 0.174, 0.1626, 0.2797, 0.3226, 0.3226, 0.479, 0.4292, 0.4406, 0.6687, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047816848058794440", "ticker_idx": 0, "ts": "2026-04-24T23:15:57+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Could there still be more upside here?", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Watchlist leans: Advantest for test-step upside; Montage over Rambus on CXL exposure; QCOM bullish if Nuvia data-center CPU launches in June.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "My current watchlist:\n\nAdvantest — As GPU architectures become increasingly complex, the number of required testing steps could rise exponentially. Could there still be more upside here?\n\nMontage — If CPU adoption increases, usage of memory interfaces should also surge. I prefer Montage over Rambus in particular because Montage also has exposure to CXL.\n\nDelton Technology — A specialized CPU substrate manufacturer with a meaningful presence in the CPU substrate market.\n\nQualcomm — If Qualcomm really launches its data-center Nuvia CPU in June, it could potentially reverse the current overly short-biased sentiment.\n\n---\n\nSorry, The reason I prefer Montage is that it has a higher share of revenue from physical chip sales than Rambus.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.052309782608695676, "ret_signed_m1d": -0.052309782608695676, "alpha_spy_m1d": -0.04462007448812355, "alpha_c_m1d": -0.003755173534868428, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06997282608695654, "ret_signed_p1d": 0.06997282608695654, "alpha_spy_p1d": 0.0682500195982545, "alpha_c_p1d": 0.0703282337856248, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": -0.05519701086956519, "ret_signed_p1w": -0.05519701086956519, "alpha_spy_p1w": -0.0645955905117066, "alpha_c_p1w": -0.06187105886442823, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": -0.05723505434782605, "ret_signed_p1m": -0.05723505434782605, "alpha_spy_p1m": -0.10163656161542434, "alpha_c_p1m": -0.19521784509488482, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1535, -0.1337, -0.0889, -0.1348, -0.1755, -0.117, -0.1359, -0.1774, -0.1679, -0.0721, -0.0612, -0.0612, -0.0906, -0.0797, -0.0807, -0.0911, -0.0855, -0.1182, -0.1359, -0.1359, -0.097, -0.0293, -0.046, -0.0892, -0.1248, -0.1274, -0.1689, -0.1338, -0.1279, -0.2241, -0.1833, -0.1559, -0.1689, -0.1978, -0.1811, -0.2013, -0.1476, -0.1866, -0.1866, -0.229, -0.242, -0.2054, -0.221, -0.2514, -0.2892, -0.3094, -0.2357, -0.2824, -0.2678, -0.2551, -0.2459, -0.1433, -0.1576, -0.1512, -0.1549, -0.0829, -0.063, -0.0277, -0.0533, -0.0795, -0.0761, -0.0523, -0.0523, 0.0, 0.07, 0.0105, 0.0105, -0.0401, -0.0552, -0.0552, -0.0552, -0.0552, 0.0093, 0.0151, -0.0224, -0.025, -0.0391, -0.028, -0.1046, -0.1118, -0.141, -0.1296, -0.0915, -0.0881, -0.0572, -0.1143, -0.0785, -0.1053, -0.1111, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047816848058794440", "ticker_idx": 1, "ts": "2026-04-24T23:15:57+00:00", "symbol": "Montage Technology", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I prefer Montage over Rambus in particular because Montage also has exposure to CXL", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Watchlist leans: Advantest for test-step upside; Montage over Rambus on CXL exposure; QCOM bullish if Nuvia data-center CPU launches in June.", "resolved_tickers": ["688008.SS"], "resolution_method": "exchange_ticker", "resolution_reason": "Montage Technology 688008.SS China STAR Market memory IC", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "My current watchlist:\n\nAdvantest — As GPU architectures become increasingly complex, the number of required testing steps could rise exponentially. Could there still be more upside here?\n\nMontage — If CPU adoption increases, usage of memory interfaces should also surge. I prefer Montage over Rambus in particular because Montage also has exposure to CXL.\n\nDelton Technology — A specialized CPU substrate manufacturer with a meaningful presence in the CPU substrate market.\n\nQualcomm — If Qualcomm really launches its data-center Nuvia CPU in June, it could potentially reverse the current overly short-biased sentiment.\n\n---\n\nSorry, The reason I prefer Montage is that it has a higher share of revenue from physical chip sales than Rambus.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02324160887925264, "ret_signed_m1d": -0.02324160887925264, "alpha_spy_m1d": -0.015551900758680515, "alpha_c_m1d": 0.02531300019457461, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0892354516078937, "ret_signed_p1d": 0.0892354516078937, "alpha_spy_p1d": 0.08751264511919166, "alpha_c_p1d": 0.08959085930656197, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.05993885658567466, "ret_signed_p1w": 0.05993885658567466, "alpha_spy_p1w": 0.05054027694353325, "alpha_c_p1w": 0.05326480859081162, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.7374922726132453, "ret_signed_p1m": 0.7374922726132453, "alpha_spy_p1m": 0.6930907653456471, "alpha_c_p1m": 0.5995094818661866, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0338, 0.0597, -0.0081, 0.1122, 0.0883, 0.1133, 0.0453, 0.0284, 0.0002, 0.0609, 0.0856, 0.0391, 0.0282, 0.0067, 0.0067, 0.0067, 0.0067, 0.0067, 0.0067, 0.0067, 0.068, 0.0259, 0.0099, 0.0003, -0.0248, -0.0955, -0.0807, -0.0665, -0.0727, -0.0937, -0.0643, -0.0838, -0.0957, -0.104, -0.0673, -0.1034, -0.0604, -0.0845, -0.1138, -0.1633, -0.1586, -0.1315, -0.1621, -0.1737, -0.1927, -0.2338, -0.2051, -0.211, -0.2217, -0.2217, -0.2177, -0.1486, -0.1488, -0.1221, -0.0685, -0.062, -0.0998, -0.0765, -0.0778, -0.0541, -0.0312, -0.0009, -0.0232, 0.0, 0.0892, 0.0618, 0.0391, 0.0599, 0.0599, 0.0599, 0.0599, 0.217, 0.2839, 0.2861, 0.5243, 0.4722, 0.5829, 0.6213, 0.5167, 0.5168, 0.5114, 0.5737, 0.6035, 0.6626, 0.7375, 0.6654, 0.6239, 0.5962, 0.5474, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047816848058794440", "ticker_idx": 2, "ts": "2026-04-24T23:15:57+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "could potentially reverse the current overly short-biased sentiment", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Watchlist leans: Advantest for test-step upside; Montage over Rambus on CXL exposure; QCOM bullish if Nuvia data-center CPU launches in June.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "My current watchlist:\n\nAdvantest — As GPU architectures become increasingly complex, the number of required testing steps could rise exponentially. Could there still be more upside here?\n\nMontage — If CPU adoption increases, usage of memory interfaces should also surge. I prefer Montage over Rambus in particular because Montage also has exposure to CXL.\n\nDelton Technology — A specialized CPU substrate manufacturer with a meaningful presence in the CPU substrate market.\n\nQualcomm — If Qualcomm really launches its data-center Nuvia CPU in June, it could potentially reverse the current overly short-biased sentiment.\n\n---\n\nSorry, The reason I prefer Montage is that it has a higher share of revenue from physical chip sales than Rambus.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10010082999198155, "ret_signed_m1d": -0.10010082999198155, "alpha_spy_m1d": -0.09241112187140943, "alpha_c_m1d": -0.051546220918154306, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00947254514950946, "ret_signed_p1d": 0.00947254514950946, "alpha_spy_p1d": 0.007749738660807415, "alpha_c_p1d": 0.009827952848177723, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.18918365635629764, "ret_signed_p1w": 0.18918365635629764, "alpha_spy_p1w": 0.17978507671415622, "alpha_c_p1w": 0.1825096083614346, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.5999999589955298, "ret_signed_p1m": 0.5999999589955298, "alpha_spy_p1m": 0.5555984517279315, "alpha_c_p1m": 0.46201716824847106, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0216, 0.0193, 0.0161, 0.0119, 0.0188, -0.0175, -0.0061, -0.0902, -0.0832, -0.0726, -0.0649, -0.0585, -0.0757, -0.0608, -0.0608, -0.0479, -0.0438, -0.057, -0.0462, -0.0627, -0.0335, -0.0266, -0.0281, -0.0497, -0.0586, -0.0779, -0.0687, -0.0796, -0.0884, -0.0722, -0.0917, -0.099, -0.1189, -0.1278, -0.1307, -0.116, -0.1235, -0.118, -0.1273, -0.1377, -0.1356, -0.1243, -0.123, -0.1461, -0.1463, -0.1348, -0.1449, -0.1481, -0.1481, -0.1553, -0.1665, -0.1434, -0.1418, -0.1397, -0.1183, -0.1076, -0.1061, -0.0966, -0.085, -0.0761, -0.0893, -0.0859, -0.1001, 0.0, 0.0095, 0.0077, 0.048, 0.2064, 0.1892, 0.1312, 0.2533, 0.2937, 0.3608, 0.4719, 0.5958, 0.4129, 0.4321, 0.3442, 0.3536, 0.3681, 0.3141, 0.3605, 0.4337, 0.6, 0.6, 0.6716, 0.568, 0.6345, 0.6864, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2041038584933961881", "ticker_idx": 0, "ts": "2026-04-06T06:21:33+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$INTC (positive, Overweight, increasing estimates, increasing PT to $70 from $65)", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares KeyBanc supply chain survey: INTC Overweight PT $70, negative AVGO/ARM, slightly negative NVDA.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "From KeyBanc's supply chain survey report:\n\n$INTC (positive, Overweight, increasing estimates, increasing PT to $70 from $65) – Positive takeaways include: 1) Server CPU demand remains outsized; 2) INTC increases pricing twice for both server and client CPUs; 3) 18A yields continue to improve to 65% as Panther Lake ramp continues; 4) Wins 14A design win at AAPL for low-end Mseries processor for MacBook and iPad; 5) Google’s “Humu Fish” TPU will use EMIB-T, which we size at $4B-$5B in revenues.", "reply_text": null, "parent_text": null, "quoted_text": "KeyB neg $AVGO, neg $ARM, slightly neg $NVDA: https://t.co/VM4ZiiVx8x", "parent_author": null, "ret_m1d": -0.007877072091319604, "ret_signed_m1d": -0.007877072091319604, "alpha_spy_m1d": -0.0031725122436958575, "alpha_c_m1d": 0.0013658279555611719, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04194566993537707, "ret_signed_p1d": 0.04194566993537707, "alpha_spy_p1d": 0.041505595829528286, "alpha_c_p1d": 0.03204622349515862, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.28357624797245595, "ret_signed_p1w": 0.28357624797245595, "alpha_spy_p1w": 0.24234275850899345, "alpha_c_p1w": 0.16397428818262783, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 1.1297755834128123, "ret_signed_p1m": 1.1297755834128123, "alpha_spy_p1m": 1.0313735868787548, "alpha_c_p1m": 0.8097847052382747, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1605, -0.1904, -0.103, -0.1323, -0.0687, -0.0406, -0.0484, -0.0752, -0.0752, -0.0437, 0.0683, 0.0697, -0.1124, -0.1633, -0.1349, -0.0394, -0.0417, -0.0849, -0.0388, -0.0301, -0.0429, -0.05, -0.0037, -0.0106, -0.0719, -0.049, -0.0847, -0.0786, -0.0786, -0.0906, -0.1048, -0.1213, -0.1314, -0.1408, -0.0918, -0.0768, -0.1048, -0.1018, -0.104, -0.1512, -0.1024, -0.0951, -0.1449, -0.1024, -0.0788, -0.0551, -0.1089, -0.0987, -0.0989, -0.1323, -0.1132, -0.0906, -0.1361, -0.1333, -0.1323, -0.0709, -0.1315, -0.1506, -0.1889, -0.131, -0.0542, -0.0079, -0.0079, 0.0, 0.0419, 0.1609, 0.2154, 0.2284, 0.2836, 0.2566, 0.2789, 0.349, 0.349, 0.2938, 0.3048, 0.2853, 0.3151, 0.6254, 0.6737, 0.6644, 0.8659, 0.8606, 0.9618, 0.8862, 1.1298, 1.2255, 1.1587, 1.46, 1.549, 1.3751, 1.3688, 1.283, 1.142, 1.1302, 1.182, 1.3427, 1.3336, 1.36, 1.36, 1.4325, 1.398, 1.3807, 1.2584, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2041038584933961881", "ticker_idx": 1, "ts": "2026-04-06T06:21:33+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KeyB neg $AVGO", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares KeyBanc supply chain survey: INTC Overweight PT $70, negative AVGO/ARM, slightly negative NVDA.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "From KeyBanc's supply chain survey report:\n\n$INTC (positive, Overweight, increasing estimates, increasing PT to $70 from $65) – Positive takeaways include: 1) Server CPU demand remains outsized; 2) INTC increases pricing twice for both server and client CPUs; 3) 18A yields continue to improve to 65% as Panther Lake ramp continues; 4) Wins 14A design win at AAPL for low-end Mseries processor for MacBook and iPad; 5) Google’s “Humu Fish” TPU will use EMIB-T, which we size at $4B-$5B in revenues.", "reply_text": null, "parent_text": null, "quoted_text": "KeyB neg $AVGO, neg $ARM, slightly neg $NVDA: https://t.co/VM4ZiiVx8x", "parent_author": null, "ret_m1d": 0.0003816274527959074, "ret_signed_m1d": -0.0003816274527959074, "alpha_spy_m1d": -0.005086187300419653, "alpha_c_m1d": -0.009624527499676683, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.062144226060108165, "ret_signed_p1d": -0.062144226060108165, "alpha_spy_p1d": -0.06170415195425938, "alpha_c_p1d": -0.052244779619889714, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.20774101976834092, "ret_signed_p1w": -0.20774101976834092, "alpha_spy_p1w": -0.16650753030487841, "alpha_c_p1w": -0.08813905997851279, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.35915782624537007, "ret_signed_p1m": -0.35915782624537007, "alpha_spy_p1m": -0.2607558297113126, "alpha_c_p1m": -0.03916694807083254, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0902, 0.0552, 0.0948, 0.1178, 0.1254, 0.0787, 0.0886, 0.1162, 0.1162, 0.0556, 0.0435, 0.033, 0.0157, 0.031, 0.0562, 0.0576, 0.0496, 0.0514, 0.0508, 0.0166, -0.0223, -0.0145, 0.0566, 0.0916, 0.0805, 0.0878, 0.051, 0.032, 0.032, 0.0554, 0.0585, 0.06, 0.0557, 0.0484, 0.033, 0.0547, 0.021, 0.0142, 0.0118, -0.004, 0.0077, 0.0561, 0.0488, 0.0973, 0.0872, 0.084, 0.0663, 0.0224, 0.0312, 0.0197, 0.0027, 0.0151, -0.0145, 0.0257, 0.0123, 0.0139, -0.0159, -0.0437, -0.0669, -0.0156, -0.003, 0.0004, 0.0004, 0.0, 0.0621, 0.1151, 0.1287, 0.1817, 0.2077, 0.211, 0.2617, 0.2673, 0.2929, 0.271, 0.279, 0.3442, 0.3356, 0.3445, 0.33, 0.2716, 0.2895, 0.3276, 0.3398, 0.3246, 0.3592, 0.3531, 0.3121, 0.3676, 0.3626, 0.3335, 0.3255, 0.3987, 0.3523, 0.338, 0.3073, 0.3286, 0.3185, 0.3171, 0.3171, 0.3421, 0.3417, 0.3567, 0.4209, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2041038584933961881", "ticker_idx": 2, "ts": "2026-04-06T06:21:33+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "neg $ARM", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares KeyBanc supply chain survey: INTC Overweight PT $70, negative AVGO/ARM, slightly negative NVDA.", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "From KeyBanc's supply chain survey report:\n\n$INTC (positive, Overweight, increasing estimates, increasing PT to $70 from $65) – Positive takeaways include: 1) Server CPU demand remains outsized; 2) INTC increases pricing twice for both server and client CPUs; 3) 18A yields continue to improve to 65% as Panther Lake ramp continues; 4) Wins 14A design win at AAPL for low-end Mseries processor for MacBook and iPad; 5) Google’s “Humu Fish” TPU will use EMIB-T, which we size at $4B-$5B in revenues.", "reply_text": null, "parent_text": null, "quoted_text": "KeyB neg $AVGO, neg $ARM, slightly neg $NVDA: https://t.co/VM4ZiiVx8x", "parent_author": null, "ret_m1d": 0.0022853823225543746, "ret_signed_m1d": -0.0022853823225543746, "alpha_spy_m1d": -0.006989942170178121, "alpha_c_m1d": -0.01152828236943515, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.033003989521416366, "ret_signed_p1d": 0.033003989521416366, "alpha_spy_p1d": 0.03344406362726515, "alpha_c_p1d": 0.04290343596163482, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.05921891043613137, "ret_signed_p1w": -0.05921891043613137, "alpha_spy_p1w": -0.017985420972668864, "alpha_c_p1w": 0.06038304935369676, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.4037775784116808, "ret_signed_p1m": -0.4037775784116808, "alpha_spy_p1m": -0.30537558187762337, "alpha_c_p1m": -0.0837867002371433, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2224, -0.2399, -0.2486, -0.2529, -0.2751, -0.2943, -0.2935, -0.289, -0.289, -0.2796, -0.2343, -0.1988, -0.2198, -0.2288, -0.2278, -0.2609, -0.2712, -0.2918, -0.2812, -0.2972, -0.2949, -0.2547, -0.1685, -0.1624, -0.1534, -0.1579, -0.1787, -0.1579, -0.1579, -0.1471, -0.1447, -0.1468, -0.1559, -0.168, -0.1387, -0.1145, -0.1311, -0.1433, -0.164, -0.1818, -0.1658, -0.1892, -0.2312, -0.2093, -0.1897, -0.1927, -0.2262, -0.222, -0.182, -0.1442, -0.1372, -0.1274, -0.1104, -0.0799, -0.0928, 0.0558, 0.0405, -0.0312, -0.0794, 0.0169, 0.0423, 0.0023, 0.0023, 0.0, -0.033, 0.0009, 0.0069, 0.0011, 0.0592, 0.0837, 0.071, 0.0911, 0.1207, 0.177, 0.1796, 0.3213, 0.3753, 0.5783, 0.4511, 0.3353, 0.3557, 0.4137, 0.4195, 0.3663, 0.4038, 0.5951, 0.4338, 0.4336, 0.4294, 0.3976, 0.4869, 0.5359, 0.4059, 0.446, 0.5, 0.7257, 1.0046, 1.0603, 1.0603, 1.1592, 1.0348, 1.2536, 1.3747, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2041038584933961881", "ticker_idx": 3, "ts": "2026-04-06T06:21:33+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "slightly neg $NVDA", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares KeyBanc supply chain survey: INTC Overweight PT $70, negative AVGO/ARM, slightly negative NVDA.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "From KeyBanc's supply chain survey report:\n\n$INTC (positive, Overweight, increasing estimates, increasing PT to $70 from $65) – Positive takeaways include: 1) Server CPU demand remains outsized; 2) INTC increases pricing twice for both server and client CPUs; 3) 18A yields continue to improve to 65% as Panther Lake ramp continues; 4) Wins 14A design win at AAPL for low-end Mseries processor for MacBook and iPad; 5) Google’s “Humu Fish” TPU will use EMIB-T, which we size at $4B-$5B in revenues.", "reply_text": null, "parent_text": null, "quoted_text": "KeyB neg $AVGO, neg $ARM, slightly neg $NVDA: https://t.co/VM4ZiiVx8x", "parent_author": null, "ret_m1d": -0.0014073406938694966, "ret_signed_m1d": 0.0014073406938694966, "alpha_spy_m1d": -0.0032972191537542495, "alpha_c_m1d": -0.007835559353011279, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0025895446715138437, "ret_signed_p1d": -0.0025895446715138437, "alpha_spy_p1d": -0.002149470565665057, "alpha_c_p1d": 0.0073099017687046075, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.06569465328215585, "ret_signed_p1w": -0.06569465328215585, "alpha_spy_p1w": -0.024461163818693343, "alpha_c_p1w": 0.05390730650767228, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.1061697853814032, "ret_signed_p1m": -0.1061697853814032, "alpha_spy_p1m": -0.007767788847345747, "alpha_c_p1m": 0.21382109279313433, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0645, 0.0416, 0.0406, 0.041, 0.0459, 0.0309, 0.0529, 0.0483, 0.0483, 0.0024, 0.0319, 0.0405, 0.0564, 0.0497, 0.0612, 0.0781, 0.0836, 0.0759, 0.0448, 0.0151, -0.0195, -0.0325, 0.0437, 0.0697, 0.0613, 0.0698, 0.0523, 0.029, 0.029, 0.0412, 0.0582, 0.0577, 0.0685, 0.0782, 0.0856, 0.1008, 0.0408, -0.0026, 0.0272, 0.0135, 0.0303, 0.032, 0.001, 0.0281, 0.0401, 0.0472, 0.031, 0.0147, 0.0314, 0.0241, 0.0155, 0.0052, -0.0278, -0.0113, -0.0137, 0.0059, -0.036, -0.057, -0.0702, -0.0182, -0.0106, -0.0014, -0.0014, 0.0, 0.0026, 0.025, 0.0353, 0.0619, 0.0657, 0.1062, 0.1195, 0.1166, 0.1353, 0.1375, 0.1252, 0.1399, 0.1238, 0.1724, 0.2194, 0.2, 0.1779, 0.1235, 0.1171, 0.1173, 0.1062, 0.17, 0.1906, 0.2114, 0.2353, 0.2429, 0.2713, 0.3271, 0.2684, 0.2515, 0.2419, 0.258, 0.2357, 0.2122, 0.2122, 0.2095, 0.1968, 0.2061, 0.1886, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044750394719715585", "ticker_idx": 0, "ts": "2026-04-16T12:10:58+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "JPM forecasts that in Google's TPU v9, $AVGO (Broadcom) is highly likely to be the sole supplier using TSMC's CoWoS-L technology", "tweet_type": "analysis", "timeline": "unspecified", "summary": "JPM survey: AVGO sole TPU v9 supplier (long); Google SRAM engine hurts MediaTek TAM (short); possible bearish read on INTC EMIB-T losing to CoWoS-L.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM's supply chain survey indicates that Google is developing an SRAM-based compute engine in collaboration with Marvell (a concept similar to NVIDIA's Groq LPU), and this is expected to negatively impact MediaTek's TAM. (@FundaAI made a really impressive call.)\n\nAdditionally, JPM forecasts that in Google's TPU v9, $AVGO (Broadcom) is highly likely to be the sole supplier using TSMC's CoWoS-L technology.\n(Bearish on $INTC's EMIB-T?)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004391798616304676, "ret_signed_m1d": -0.004391798616304676, "alpha_spy_m1d": -0.0019405105723011573, "alpha_c_m1d": -0.00043404170935390507, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020252484000041315, "ret_signed_p1d": 0.020252484000041315, "alpha_spy_p1d": 0.008166798869817304, "alpha_c_p1d": -0.0003280263789551263, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": 0.053881098087510404, "ret_signed_p1w": 0.053881098087510404, "alpha_spy_p1w": 0.04420399048214141, "alpha_c_p1w": -0.005595636821419392, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.06705649393642443, "ret_signed_p1m": 0.06705649393642443, "alpha_spy_p1m": 0.013597537744977961, "alpha_c_p1m": -0.1562065706807243, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1192, -0.1671, -0.1766, -0.1849, -0.1985, -0.1865, -0.1666, -0.1655, -0.1717, -0.1703, -0.1708, -0.1978, -0.2285, -0.2224, -0.1663, -0.1387, -0.1474, -0.1416, -0.1706, -0.1857, -0.1857, -0.1672, -0.1648, -0.1636, -0.1669, -0.1727, -0.1849, -0.1678, -0.1944, -0.1997, -0.2016, -0.214, -0.2048, -0.1666, -0.1724, -0.1341, -0.1421, -0.1446, -0.1586, -0.1932, -0.1863, -0.1953, -0.2088, -0.199, -0.2224, -0.1906, -0.2012, -0.1999, -0.2235, -0.2454, -0.2637, -0.2233, -0.2133, -0.2106, -0.2106, -0.2109, -0.1619, -0.1201, -0.1093, -0.0676, -0.047, -0.0444, -0.0044, 0.0, 0.0203, 0.0029, 0.0093, 0.0607, 0.0539, 0.061, 0.0495, 0.0034, 0.0175, 0.0476, 0.0572, 0.0452, 0.0725, 0.0677, 0.0354, 0.0791, 0.0752, 0.0523, 0.046, 0.1037, 0.0671, 0.0558, 0.0316, 0.0484, 0.0404, 0.0393, 0.0393, 0.0591, 0.0587, 0.0705, 0.1212, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044750394719715585", "ticker_idx": 1, "ts": "2026-04-16T12:10:58+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "this is expected to negatively impact MediaTek's TAM", "tweet_type": "analysis", "timeline": "unspecified", "summary": "JPM survey: AVGO sole TPU v9 supplier (long); Google SRAM engine hurts MediaTek TAM (short); possible bearish read on INTC EMIB-T losing to CoWoS-L.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM's supply chain survey indicates that Google is developing an SRAM-based compute engine in collaboration with Marvell (a concept similar to NVIDIA's Groq LPU), and this is expected to negatively impact MediaTek's TAM. (@FundaAI made a really impressive call.)\n\nAdditionally, JPM forecasts that in Google's TPU v9, $AVGO (Broadcom) is highly likely to be the sole supplier using TSMC's CoWoS-L technology.\n(Bearish on $INTC's EMIB-T?)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0554089709762533, "ret_signed_m1d": 0.0554089709762533, "alpha_spy_m1d": 0.05295768293224978, "alpha_c_m1d": 0.05145121406930253, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0158311345646438, "ret_signed_p1d": -0.0158311345646438, "alpha_spy_p1d": -0.0037454494344197897, "alpha_c_p1d": 0.004749375814352641, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": 0.16886543535620047, "ret_signed_p1w": -0.16886543535620047, "alpha_spy_p1w": -0.15918832775083147, "alpha_c_p1w": -0.10938870044727067, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.7203166226912929, "ret_signed_p1m": -0.7203166226912929, "alpha_spy_p1m": -0.6668576664998465, "alpha_c_p1m": -0.4970535580741442, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2164, -0.2164, -0.2269, -0.2164, -0.1398, -0.066, -0.0633, -0.0607, -0.0607, -0.0712, -0.1003, -0.0528, -0.0501, -0.066, -0.0976, -0.0343, -0.0264, -0.0211, -0.0211, -0.0211, -0.0211, -0.0211, -0.0211, -0.0211, -0.0211, -0.0528, -0.0369, -0.0132, 0.0264, 0.0264, 0.0026, -0.0422, -0.0923, -0.0633, -0.0686, -0.1214, -0.1003, -0.0686, -0.058, -0.0923, -0.0976, -0.0871, -0.0871, -0.1135, -0.1029, -0.1425, -0.1451, -0.1451, -0.1609, -0.1636, -0.2032, -0.2137, -0.2269, -0.2269, -0.2269, -0.2269, -0.2243, -0.1662, -0.1689, -0.1689, -0.1451, -0.0923, -0.0554, 0.0, 0.0158, 0.0026, 0.1029, 0.2111, 0.1689, 0.285, 0.285, 0.3799, 0.3588, 0.3773, 0.3773, 0.5145, 0.6649, 0.81, 0.8047, 0.9156, 1.0475, 0.9525, 0.8443, 0.7968, 0.7203, 0.7942, 0.6649, 0.7045, 0.8734, 1.0369, 1.2401, 1.2507, 1.4485, 1.3272, 1.2744, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044750394719715585", "ticker_idx": 2, "ts": "2026-04-16T12:10:58+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bearish on $INTC's EMIB-T?", "tweet_type": "analysis", "timeline": "unspecified", "summary": "JPM survey: AVGO sole TPU v9 supplier (long); Google SRAM engine hurts MediaTek TAM (short); possible bearish read on INTC EMIB-T losing to CoWoS-L.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "JPM's supply chain survey indicates that Google is developing an SRAM-based compute engine in collaboration with Marvell (a concept similar to NVIDIA's Groq LPU), and this is expected to negatively impact MediaTek's TAM. (@FundaAI made a really impressive call.)\n\nAdditionally, JPM forecasts that in Google's TPU v9, $AVGO (Broadcom) is highly likely to be the sole supplier using TSMC's CoWoS-L technology.\n(Bearish on $INTC's EMIB-T?)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05197076727874084, "ret_signed_m1d": 0.05197076727874084, "alpha_spy_m1d": 0.04951947923473732, "alpha_c_m1d": 0.048013010371790066, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": -0.0, "alpha_spy_p1d": 0.012085685130224011, "alpha_c_p1d": 0.02058051037899644, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": -0.025109506871578424, "ret_signed_p1w": 0.025109506871578424, "alpha_spy_p1w": 0.03478661447694742, "alpha_c_p1w": 0.08458624178050822, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.5878831626725023, "ret_signed_p1m": -0.5878831626725023, "alpha_spy_p1m": -0.5344242064810558, "alpha_c_p1m": -0.36462009805535356, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3145, -0.2911, -0.208, -0.207, -0.342, -0.3797, -0.3587, -0.2879, -0.2896, -0.3216, -0.2874, -0.281, -0.2905, -0.2958, -0.2615, -0.2666, -0.312, -0.295, -0.3215, -0.3169, -0.3169, -0.3258, -0.3364, -0.3486, -0.3561, -0.3631, -0.3267, -0.3156, -0.3364, -0.3342, -0.3358, -0.3708, -0.3346, -0.3292, -0.3661, -0.3346, -0.3171, -0.2996, -0.3394, -0.3318, -0.332, -0.3568, -0.3426, -0.3258, -0.3596, -0.3575, -0.3568, -0.3112, -0.3562, -0.3704, -0.3987, -0.3558, -0.2988, -0.2645, -0.2645, -0.2587, -0.2276, -0.1394, -0.099, -0.0893, -0.0485, -0.0685, -0.052, 0.0, 0.0, -0.0409, -0.0327, -0.0472, -0.0251, 0.205, 0.2407, 0.2339, 0.3832, 0.3793, 0.4543, 0.3982, 0.5788, 0.6498, 0.6003, 0.8236, 0.8896, 0.7607, 0.7561, 0.6924, 0.5879, 0.5791, 0.6175, 0.7366, 0.7299, 0.7495, 0.7495, 0.8032, 0.7777, 0.7648, 0.6742, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047503819890991301", "ticker_idx": 0, "ts": "2026-04-24T02:32:06+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Upgrade to Buy, new PT of $1,277: SNDK reports FY3Q26 on April 30. Combined with the recent contract price trajectory, we expect both results and guidance to come in strong.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Upgrades SNDK to Buy with $1,277 PT on surging NAND ASPs, KV cache demand, and superior valuation vs Micron; implies relative underperformance for MU.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "【GF Overseas Electronics & Communications】\nSanDisk FY3Q26 Preview: NAND Is in a Better Position\n\n☄️ SanDisk FY3Q26 Preview: NAND Is in a Better Position\n\n☀️ Upgrade to Buy, new PT of $1,277: SNDK reports FY3Q26 on April 30. Combined with the recent contract price trajectory, we expect both results and guidance to come in strong. NAND ASP rose in the 70% range in FY1Q26, and is accelerating further to +90–100% in FY2Q26. The biggest upside surprise this cycle is coming from consumer electronics: despite volume cuts from domestic Chinese Android OEMs, the supply-demand balance remains tight, pushing smartphone makers to accept price hikes to $0.25/GB — versus just $0.12/GB in FY1Q26. In eSSD, the current negotiation range is $0.35–0.37/GB, driving blended ASP up ~90% QoQ. On this basis, we forecast FY3Q revenue of $4.9B at 71.5% OPM, and guidance for FY4Q26 revenue north of $10B with OPM above 84%. Another important catalyst is the step-up in contract liabilities, reflecting LTA prepayments flowing through. We forecast FY2027 EPS of $286.52, and apply 4x FY2027E P/B to derive our $1,277 price target.\n\n☀️ KV cache demand growth accelerates into 2027: SNDK has indicated that KV cache will generate 100EB of incremental NAND demand in 2027, with demand potentially doubling in 2028. However, given the recent surge in token demand, OpenRouter traffic up 17x, and accelerating AI agent adoption, we view this forecast as conservative. 100EB already equates to ~7% of total 2027 NAND demand, and even accounting for improvements in compression rates, demand still looks set to grow exponentially. On the supply side, NAND capacity additions remain measured: SK Hynix's core focus is on technology migration, while new capacity is concentrated in SNDK/Kioxia's K2 fab, Samsung's P5 line (coming online in 2028), and YMTC.\n\n☀️ Cost optimization may reshape the NAND vs. DRAM competitive landscape: We expect the memory market to approach $1 trillion in size by 2027, with CSPs accounting for 40–45% of total demand and memory spend representing over 40% of CSP CAPEX. On this basis, we believe the cost advantage will progressively tilt the revenue mix toward NAND (server DRAM is priced at ~$2.2/Gb, vs. eSSD at just $0.35–0.37/GB). Tiered storage architecture is the core structural driver of rising NAND share in AI workloads, and the arrival of HBF from 2028 onward should further amplify this. On valuation, Micron currently trades at 4.3x FY2027E P/E versus SNDK at just 3.4x — on both multiple and long-term positioning, SNDK offers more compelling allocation value.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05805639996048961, "ret_signed_m1d": -0.05805639996048961, "alpha_spy_m1d": -0.05036669183991749, "alpha_c_m1d": -0.03071895863330454, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08111922899016322, "ret_signed_p1d": 0.08111922899016322, "alpha_spy_p1d": 0.07939642250146117, "alpha_c_p1d": 0.07893469459465097, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": 0.19911099174141755, "ret_signed_p1w": 0.19911099174141755, "alpha_spy_p1w": 0.18971241209927614, "alpha_c_p1w": 0.18881269013167334, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": 0.49377705317414256, "ret_signed_p1m": 0.49377705317414256, "alpha_spy_p1m": 0.44937554590654427, "alpha_c_p1m": 0.3678876634895105, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5137, -0.467, -0.4552, -0.4179, -0.328, -0.2974, -0.4095, -0.4179, -0.3959, -0.4106, -0.4528, -0.3945, -0.3633, -0.367, -0.367, -0.4034, -0.3935, -0.3726, -0.3434, -0.3267, -0.355, -0.3612, -0.3414, -0.3582, -0.3746, -0.4288, -0.3948, -0.4286, -0.4673, -0.4053, -0.3748, -0.3379, -0.3749, -0.3316, -0.2892, -0.2725, -0.2386, -0.22, -0.283, -0.2903, -0.2904, -0.3152, -0.3907, -0.3779, -0.4217, -0.3582, -0.3002, -0.2913, -0.2913, -0.268, -0.2819, -0.2111, -0.1397, -0.1395, -0.0378, -0.0459, -0.0992, -0.0711, -0.0696, -0.0777, -0.0873, -0.0109, -0.0581, 0.0, 0.0811, 0.0126, 0.0751, 0.1077, 0.1991, 0.2687, 0.4207, 0.4244, 0.3536, 0.5783, 0.5633, 0.4668, 0.462, 0.3968, 0.422, 0.3466, 0.3974, 0.4068, 0.558, 0.4938, 0.4938, 0.6058, 0.6062, 0.6584, 0.7123, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047503819890991301", "ticker_idx": 1, "ts": "2026-04-24T02:32:06+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron currently trades at 4.3x FY2027E P/E versus SNDK at just 3.4x — on both multiple and long-term positioning, SNDK offers more compelling allocation value", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Upgrades SNDK to Buy with $1,277 PT on surging NAND ASPs, KV cache demand, and superior valuation vs Micron; implies relative underperformance for MU.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "【GF Overseas Electronics & Communications】\nSanDisk FY3Q26 Preview: NAND Is in a Better Position\n\n☄️ SanDisk FY3Q26 Preview: NAND Is in a Better Position\n\n☀️ Upgrade to Buy, new PT of $1,277: SNDK reports FY3Q26 on April 30. Combined with the recent contract price trajectory, we expect both results and guidance to come in strong. NAND ASP rose in the 70% range in FY1Q26, and is accelerating further to +90–100% in FY2Q26. The biggest upside surprise this cycle is coming from consumer electronics: despite volume cuts from domestic Chinese Android OEMs, the supply-demand balance remains tight, pushing smartphone makers to accept price hikes to $0.25/GB — versus just $0.12/GB in FY1Q26. In eSSD, the current negotiation range is $0.35–0.37/GB, driving blended ASP up ~90% QoQ. On this basis, we forecast FY3Q revenue of $4.9B at 71.5% OPM, and guidance for FY4Q26 revenue north of $10B with OPM above 84%. Another important catalyst is the step-up in contract liabilities, reflecting LTA prepayments flowing through. We forecast FY2027 EPS of $286.52, and apply 4x FY2027E P/B to derive our $1,277 price target.\n\n☀️ KV cache demand growth accelerates into 2027: SNDK has indicated that KV cache will generate 100EB of incremental NAND demand in 2027, with demand potentially doubling in 2028. However, given the recent surge in token demand, OpenRouter traffic up 17x, and accelerating AI agent adoption, we view this forecast as conservative. 100EB already equates to ~7% of total 2027 NAND demand, and even accounting for improvements in compression rates, demand still looks set to grow exponentially. On the supply side, NAND capacity additions remain measured: SK Hynix's core focus is on technology migration, while new capacity is concentrated in SNDK/Kioxia's K2 fab, Samsung's P5 line (coming online in 2028), and YMTC.\n\n☀️ Cost optimization may reshape the NAND vs. DRAM competitive landscape: We expect the memory market to approach $1 trillion in size by 2027, with CSPs accounting for 40–45% of total demand and memory spend representing over 40% of CSP CAPEX. On this basis, we believe the cost advantage will progressively tilt the revenue mix toward NAND (server DRAM is priced at ~$2.2/Gb, vs. eSSD at just $0.35–0.37/GB). Tiered storage architecture is the core structural driver of rising NAND share in AI workloads, and the arrival of HBF from 2028 onward should further amplify this. On valuation, Micron currently trades at 4.3x FY2027E P/E versus SNDK at just 3.4x — on both multiple and long-term positioning, SNDK offers more compelling allocation value.\n\n$SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030198099458723426, "ret_signed_m1d": 0.030198099458723426, "alpha_spy_m1d": 0.0225083913381513, "alpha_c_m1d": -0.018356509615103822, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.056047665222807685, "ret_signed_p1d": -0.056047665222807685, "alpha_spy_p1d": -0.05432485873410564, "alpha_c_p1d": -0.05640307292147595, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.09158081140312468, "ret_signed_p1w": -0.09158081140312468, "alpha_spy_p1w": -0.08218223176098327, "alpha_c_p1w": -0.08490676340826164, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.5119181795667513, "ret_signed_p1m": -0.5119181795667513, "alpha_spy_p1m": -0.46751667229915306, "alpha_c_p1m": -0.37393538881969257, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1744, -0.1241, -0.123, -0.1651, -0.119, -0.1559, -0.2365, -0.2295, -0.2057, -0.2283, -0.2489, -0.1742, -0.1669, -0.1716, -0.1716, -0.1955, -0.1529, -0.1601, -0.1384, -0.1529, -0.1588, -0.1367, -0.1637, -0.1702, -0.1696, -0.2359, -0.1935, -0.201, -0.2548, -0.2165, -0.1888, -0.1574, -0.1843, -0.1425, -0.1109, -0.0709, -0.0708, -0.106, -0.149, -0.1863, -0.2041, -0.2311, -0.2847, -0.2811, -0.3522, -0.3199, -0.2594, -0.2627, -0.2627, -0.2395, -0.2399, -0.1812, -0.1514, -0.1533, -0.1412, -0.0625, -0.0815, -0.0795, -0.0839, -0.0972, -0.0953, -0.0186, -0.0302, 0.0, 0.056, 0.0152, 0.0438, 0.0411, 0.0916, 0.1605, 0.2889, 0.342, 0.3018, 0.5035, 0.6012, 0.5433, 0.6179, 0.5623, 0.4589, 0.3721, 0.4067, 0.4736, 0.5343, 0.5119, 0.5119, 0.8036, 0.8691, 0.8592, 0.9548, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050756475090550953", "ticker_idx": 0, "ts": "2026-05-03T01:56:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Foundry is expected to swing to profit as early as the second half of this year, or at the latest in the first half of next year. After a prolonged slump, it has clearly entered a recovery phase.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung foundry 4nm fully booked through next year; HBM4 + Big Tech orders drive expected H2 profitability turnaround after prolonged losses.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Foundry's 4nm \"Fully Booked\" Through Next Year… Set for H2 Profitability Turnaround\n\nSamsung Electronics' foundry flagship 4nm line has entered \"fully booked\" status through next year. The result reflects the convergence of HBM4 volume ramp and orders from global Big Tech. Industry observers expect the chronically loss-making foundry division to fire its first signal of recovery as early as the second half of this year.\n\nAccording to the semiconductor industry on the 3rd, Samsung Electronics' 4nm foundry process has recently secured order volumes extending into next year's production. A semiconductor industry source who requested anonymity said, \"The 4nm process has recently demonstrated better-than-expected stability among global customers, and demand is exploding,\" adding, \"The line is running so tightly that it is effectively impossible to take additional orders through next year.\"\n\nThe core driver of these orders is HBM4. Samsung Electronics produces the base die mounted in its HBM4 on the 4nm foundry process. As the company begins full-scale HBM4 supply to AI accelerator vendors such as NVIDIA and AMD, the utilization rate of the supporting 4nm foundry line has likewise reached its ceiling.\n\nDemand for the 4nm process is not confined to memory. Global fabless companies that previously relied on TSMC are now knocking on Samsung's door, factoring in supply-chain diversification and cost-effectiveness. NVIDIA and Google currently appear on the customer roster for Samsung's 4nm node. With improved yields and verified power efficiency (performance per watt), the \"love calls\" from Big Tech continue to come in.\n\nWith high-value-added HBM4 and global Big Tech volumes filling the 4nm line, expectations for an earnings turnaround are also rising. The 4nm process has already completed its large-scale investment phase, easing the depreciation burden. The structure is one in which profitability rises sharply as utilization is maximized.\n\nA source at a Samsung Foundry partner company assessed, \"Thanks to the stabilization of the 4nm process and the strong demand anchor of HBM4, Samsung Foundry is expected to swing to profit as early as the second half of this year, or at the latest in the first half of next year,\" adding, \"After a prolonged slump, it has clearly entered a recovery phase.\"\n\nThat said, the industry points to the new fab under construction in Taylor, Texas, as the biggest variable for future earnings. The substantial initial operating costs and labor expenses incurred during fab completion and ramp-up preparation could swing reported earnings depending on how they are accounted for on the books.\n\nAnother semiconductor industry source said, \"Since the Taylor fab currently being built sits on U.S. soil, whether the related performance is booked under the U.S. subsidiary (DSA) or consolidated into the domestic foundry results remains to be seen.\"\n\n---\n\nhttps://t.co/acLrMbO2s2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05161290322580647, "ret_signed_m1d": -0.05161290322580647, "alpha_spy_m1d": -0.05528975259076063, "alpha_c_m1d": -0.0505020860016141, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": -0.001110817224192373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.022091957813239027, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.022091957813239027, "ret_p1w": 0.2279569892473119, "ret_signed_p1w": 0.2279569892473119, "alpha_spy_p1w": 0.19830562263991136, "alpha_c_p1w": 0.1302710803738718, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.09768590887344009, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2742, -0.3163, -0.3193, -0.2858, -0.2884, -0.2798, -0.2334, -0.2223, -0.2223, -0.2223, -0.2223, -0.1845, -0.1841, -0.1716, -0.1416, -0.1265, -0.0643, -0.0707, -0.0707, -0.1626, -0.2609, -0.1776, -0.1922, -0.2553, -0.1935, -0.1845, -0.1935, -0.2124, -0.1901, -0.1677, -0.1051, -0.1394, -0.1441, -0.2004, -0.1858, -0.1888, -0.227, -0.227, -0.2417, -0.2809, -0.1843, -0.2327, -0.1991, -0.1695, -0.1548, -0.0946, -0.1226, -0.114, -0.1355, -0.1118, -0.0925, -0.0645, -0.071, -0.0774, -0.0581, -0.0645, -0.0344, -0.0559, -0.0344, -0.0452, -0.028, -0.0516, -0.0516, 0.0, 0.0, 0.1441, 0.1677, 0.1548, 0.228, 0.2, 0.2215, 0.2731, 0.1634, 0.2086, 0.1849, 0.1871, 0.2882, 0.2581, 0.2581, 0.286, 0.3204, 0.2882, 0.3634, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046521336106004977", "ticker_idx": 0, "ts": "2026-04-21T09:28:03+00:00", "symbol": "Samsung Electro-Mechanics", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electro-Mechanics has secured Broadcom of the U.S. as a new customer for Flip Chip Ball Grid Array (FC-BGA) AI semiconductor substrates", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung Electro-Mechanics wins Broadcom as FC-BGA substrate customer; LG Innotek also named as key AI supply-chain partner — bullish on both Korean component makers.", "resolved_tickers": ["009150.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electro-Mechanics listed on KOSPI as 009150.KS", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Electro-Mechanics to Supply AI Substrates to Broadcom\n\nRiding the artificial intelligence (AI) boom, not only memory semiconductor companies but also Korean electronic component makers have moved to the center of the global materials, parts, and equipment (MPE) supply chain. Major players including Samsung Electro-Mechanics and LG Innotek have emerged as key partners for Big Tech firms seeking to resolve AI bottlenecks. Analysts say an era of \"component-supplier supremacy\" has arrived, with Korean companies now holding the upper hand in the supply chain — a reversal from the past subcontractor-style structure in which they relied on orders from finished-goods makers. \n\nAccording to industry sources on the 21st, Samsung Electro-Mechanics has secured Broadcom of the U.S. as a new customer for Flip Chip Ball Grid Array (FC-BGA) AI semiconductor substrates. Following major Big Tech names such as NVIDIA, Google, Amazon, and Apple, Broadcom has now joined Samsung Electro-Mechanics' customer roster. Samsung Electro-Mechanics will begin supplying substrates for Broadcom's most advanced AI accelerators starting in the second half of this year.\n\nFC-BGA is a core component that connects semiconductor chips to the mainboard, transmitting electrical signals and power. Only a handful of companies are capable of manufacturing these substrates — Samsung Electro-Mechanics, LG Innotek, and Japan's Ibiden and Murata among them.\n\nA tectonic shift is also underway in the Multi-Layer Ceramic Capacitor (MLCC) market. Samsung Electro-Mechanics is supplying large volumes of ultra-high-capacitance MLCCs for AI servers to NVIDIA and other major global Big Tech firms. AI servers use three to four times more MLCCs than conventional servers. In particular, these components must withstand high-temperature, high-pressure environments, requiring advanced technical capabilities, and unit prices run three to five times higher than standard products. This is why Samsung Electro-Mechanics and other Korean companies with specialized technologies, such as Amotech, are being courted by Big Tech.\n\nThere is also a growing view that the global AI supply chain is being restructured in such a way that it can no longer be sustained without Korean component makers. This is because Korean firms possess unrivaled technological capabilities in addressing the critical bottlenecks that determine the performance of AI accelerators.\n\nAn industry official said, \"Korean companies have demonstrated irreplaceable technological capabilities in high-performance substrates and materials, and are now facing an unprecedented golden window of opportunity.\"\n\n---\n\nhttps://t.co/PtLYll9scl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.11917098445595853, "ret_signed_m1d": -0.11917098445595853, "alpha_spy_m1d": -0.12576107862232822, "alpha_c_m1d": -0.11833056244574014, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.000840422010218389, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.051813471502590636, "ret_signed_p1d": 0.051813471502590636, "alpha_spy_p1d": 0.04168677466451998, "alpha_c_p1d": 0.02983406854943449, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.021979402953156146, "ret_p1w": 0.08678756476683946, "ret_signed_p1w": 0.08678756476683946, "alpha_spy_p1w": 0.07597915497413332, "alpha_c_p1w": 0.06635958873583303, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.020427976031006434, "ret_p1m": 0.3743523316062176, "ret_signed_p1m": 0.3743523316062176, "alpha_spy_p1m": 0.32156006083024824, "alpha_c_p1m": 0.22922338601229764, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.14512894559391998, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6334, -0.6477, -0.6438, -0.6412, -0.6373, -0.638, -0.6386, -0.6334, -0.6017, -0.6049, -0.6244, -0.6321, -0.6062, -0.5939, -0.5991, -0.5861, -0.5991, -0.5991, -0.5991, -0.5991, -0.5363, -0.5117, -0.4475, -0.4223, -0.3977, -0.3957, -0.419, -0.419, -0.4702, -0.5304, -0.4799, -0.476, -0.5052, -0.4793, -0.4741, -0.4825, -0.4812, -0.465, -0.4216, -0.399, -0.3789, -0.3983, -0.4462, -0.4339, -0.4113, -0.408, -0.408, -0.4443, -0.4722, -0.4249, -0.4598, -0.4093, -0.4016, -0.408, -0.3342, -0.3316, -0.2681, -0.2642, -0.2435, -0.2047, -0.1723, -0.1205, -0.1192, 0.0, 0.0518, 0.0026, 0.0207, 0.0285, 0.0868, 0.0712, 0.0777, 0.0777, 0.1891, 0.1891, 0.1813, 0.1878, 0.1839, 0.1658, 0.2409, 0.3329, 0.3264, 0.3083, 0.3355, 0.2785, 0.3744, 0.5596, 0.7358, 0.7358, 1.0363, 1.1114, 1.3951, 1.7552, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046521336106004977", "ticker_idx": 1, "ts": "2026-04-21T09:28:03+00:00", "symbol": "LG Innotek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Major players including Samsung Electro-Mechanics and LG Innotek have emerged as key partners for Big Tech firms seeking to resolve AI bottlenecks", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung Electro-Mechanics wins Broadcom as FC-BGA substrate customer; LG Innotek also named as key AI supply-chain partner — bullish on both Korean component makers.", "resolved_tickers": ["011070.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "LG Innotek 011070.KS Korea camera modules", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Electro-Mechanics to Supply AI Substrates to Broadcom\n\nRiding the artificial intelligence (AI) boom, not only memory semiconductor companies but also Korean electronic component makers have moved to the center of the global materials, parts, and equipment (MPE) supply chain. Major players including Samsung Electro-Mechanics and LG Innotek have emerged as key partners for Big Tech firms seeking to resolve AI bottlenecks. Analysts say an era of \"component-supplier supremacy\" has arrived, with Korean companies now holding the upper hand in the supply chain — a reversal from the past subcontractor-style structure in which they relied on orders from finished-goods makers. \n\nAccording to industry sources on the 21st, Samsung Electro-Mechanics has secured Broadcom of the U.S. as a new customer for Flip Chip Ball Grid Array (FC-BGA) AI semiconductor substrates. Following major Big Tech names such as NVIDIA, Google, Amazon, and Apple, Broadcom has now joined Samsung Electro-Mechanics' customer roster. Samsung Electro-Mechanics will begin supplying substrates for Broadcom's most advanced AI accelerators starting in the second half of this year.\n\nFC-BGA is a core component that connects semiconductor chips to the mainboard, transmitting electrical signals and power. Only a handful of companies are capable of manufacturing these substrates — Samsung Electro-Mechanics, LG Innotek, and Japan's Ibiden and Murata among them.\n\nA tectonic shift is also underway in the Multi-Layer Ceramic Capacitor (MLCC) market. Samsung Electro-Mechanics is supplying large volumes of ultra-high-capacitance MLCCs for AI servers to NVIDIA and other major global Big Tech firms. AI servers use three to four times more MLCCs than conventional servers. In particular, these components must withstand high-temperature, high-pressure environments, requiring advanced technical capabilities, and unit prices run three to five times higher than standard products. This is why Samsung Electro-Mechanics and other Korean companies with specialized technologies, such as Amotech, are being courted by Big Tech.\n\nThere is also a growing view that the global AI supply chain is being restructured in such a way that it can no longer be sustained without Korean component makers. This is because Korean firms possess unrivaled technological capabilities in addressing the critical bottlenecks that determine the performance of AI accelerators.\n\nAn industry official said, \"Korean companies have demonstrated irreplaceable technological capabilities in high-performance substrates and materials, and are now facing an unprecedented golden window of opportunity.\"\n\n---\n\nhttps://t.co/PtLYll9scl", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.036470588235294144, "ret_signed_m1d": -0.036470588235294144, "alpha_spy_m1d": -0.04306068240166383, "alpha_c_m1d": -0.035630166225075754, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.000840422010218389, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.17647058823529416, "ret_signed_p1d": 0.17647058823529416, "alpha_spy_p1d": 0.1663438913972235, "alpha_c_p1d": 0.154491185282138, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.021979402953156146, "ret_p1w": 0.35764705882352943, "ret_signed_p1w": 0.35764705882352943, "alpha_spy_p1w": 0.3468386490308233, "alpha_c_p1w": 0.337219082792523, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.020427976031006434, "ret_p1m": 0.7882352941176471, "ret_signed_p1m": 0.7882352941176471, "alpha_spy_p1m": 0.7354430233416778, "alpha_c_p1m": 0.6431063485237272, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.14512894559391998, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.338, -0.3521, -0.3544, -0.3977, -0.3977, -0.4199, -0.4269, -0.4433, -0.4281, -0.4222, -0.448, -0.4562, -0.4491, -0.4304, -0.4129, -0.407, -0.4328, -0.4328, -0.4328, -0.4328, -0.4082, -0.4012, -0.3661, -0.3532, -0.3275, -0.193, -0.2538, -0.2538, -0.3333, -0.4246, -0.3836, -0.3684, -0.414, -0.3965, -0.3895, -0.4059, -0.4152, -0.4059, -0.4023, -0.3719, -0.331, -0.3076, -0.3216, -0.317, -0.2678, -0.2912, -0.2503, -0.2753, -0.3094, -0.1953, -0.2271, -0.2365, -0.2082, -0.2341, -0.1835, -0.1976, -0.1059, -0.1494, -0.1282, -0.0988, -0.1082, -0.0729, -0.0365, 0.0, 0.1765, 0.1765, 0.2753, 0.2612, 0.3576, 0.3929, 0.3482, 0.3482, 0.4, 0.4, 0.4871, 0.4988, 0.4682, 0.6188, 0.6376, 0.7718, 0.7882, 0.7224, 0.7882, 0.8659, 0.7882, 0.9718, 1.0329, 1.0329, 1.5129, 1.4565, 1.6682, 2.4306, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2042197886541455506", "ticker_idx": 0, "ts": "2026-04-09T11:08:13+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bernstein raises $SNDK target price to $1,250, with a blue-sky upside scenario of $3,000", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Bernstein's raised PT of $1,250 (blue-sky $3,000) on SNDK based on 11x through-cycle EPS multiple.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Bernstein raises $SNDK target price to $1,250, with a blue-sky upside scenario of $3,000.\n\nBase-case valuation applies an 11x multiple to through-cycle earnings, defined as the average of FY26-29 Bernstein base-case EPS of $113.93.\n\nIn the blue-sky scenario, applying a 13x multiple to peak earnings yields a valuation of approximately $3,000 per share, or roughly $2,950 based on bull-case peak EPS estimates.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08298787216827141, "ret_signed_m1d": -0.08298787216827141, "alpha_spy_m1d": -0.07725187250221388, "alpha_c_m1d": -0.08031310013179727, "bench_spy_m1d": -0.005735999666057534, "bench_c_m1d": -0.002674772036474149, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00023487464954263793, "ret_signed_p1d": 0.00023487464954263793, "alpha_spy_p1d": 0.0008966551640922082, "alpha_c_p1d": -0.003636370262225519, "bench_spy_p1d": -0.0006617805145495703, "bench_c_p1d": 0.003871244911768157, "ret_p1w": 0.07973503387262282, "ret_signed_p1w": 0.07973503387262282, "alpha_spy_p1w": 0.04774550458331417, "alpha_c_p1w": 0.009699161167026071, "bench_spy_p1w": 0.03198952928930865, "bench_c_p1w": 0.07003587270559675, "ret_p1m": 0.83465828103723, "ret_signed_p1m": 0.83465828103723, "alpha_spy_p1m": 0.7497793643392563, "alpha_c_p1m": 0.599210999952027, "bench_spy_p1m": 0.08487891669797376, "bench_c_p1m": 0.23544728108520308, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5429, -0.5422, -0.5446, -0.5194, -0.5143, -0.5143, -0.4679, -0.4113, -0.4088, -0.4436, -0.4471, -0.4347, -0.3804, -0.3667, -0.3233, -0.2188, -0.1833, -0.3136, -0.3234, -0.2978, -0.3149, -0.364, -0.2962, -0.2598, -0.2642, -0.2642, -0.3065, -0.2949, -0.2707, -0.2367, -0.2173, -0.2502, -0.2574, -0.2345, -0.2539, -0.273, -0.336, -0.2965, -0.3358, -0.3808, -0.3087, -0.2732, -0.2303, -0.2733, -0.2231, -0.1737, -0.1543, -0.1149, -0.0933, -0.1666, -0.1751, -0.1751, -0.204, -0.2917, -0.2768, -0.3277, -0.2539, -0.1865, -0.1761, -0.1761, -0.1491, -0.1653, -0.083, 0.0, 0.0002, 0.1185, 0.1091, 0.0471, 0.0797, 0.0815, 0.0722, 0.061, 0.1497, 0.095, 0.1624, 0.2567, 0.1771, 0.2497, 0.2876, 0.3939, 0.4748, 0.6514, 0.6557, 0.5735, 0.8347, 0.8173, 0.7051, 0.6995, 0.6237, 0.653, 0.5654, 0.6244, 0.6353, 0.8111, 0.7364, 0.7364, 0.8666, 0.8671, 0.9278, 0.9904, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048725249722917370", "ticker_idx": 0, "ts": "2026-04-27T11:25:37+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD's earnings could be worth getting excited about", "tweet_type": "analysis", "timeline": "weeks", "summary": "Bullish AMD into earnings: picked up TSMC 4/5nm capacity vacated by Qualcomm/MediaTek; CPU supply tight with customers lining up.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I didn’t realize this before. I naturally assumed Intel would have a supply advantage over AMD in CPUs.\n\nBut Qualcomm and MediaTek apparently cut their TSMC 4/5nm orders because mid- to low-end smartphone shipments literally collapsed. AMD seems to have taken over that capacity.\n\nRight now, even CPUs produced on older nodes are being bought up by customers lining up for supply. AMD’s earnings could be worth getting excited about.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03938676294254262, "ret_signed_m1d": 0.03938676294254262, "alpha_spy_m1d": 0.04110660647364972, "alpha_c_m1d": 0.03903122888433286, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03412728464601866, "ret_signed_p1d": -0.03412728464601866, "alpha_spy_p1d": -0.0292613359178725, "alpha_c_p1d": -0.004399441205423127, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.02064968341535689, "ret_signed_p1w": 0.02064968341535689, "alpha_spy_p1w": 0.016678561971036565, "alpha_c_p1w": 0.01960279294588796, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.5058124116063647, "ret_signed_p1m": 0.5058124116063647, "alpha_spy_p1m": 0.4562856630127199, "alpha_c_p1m": 0.31642355403651523, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2447, -0.2464, -0.2926, -0.2641, -0.2765, -0.4018, -0.4247, -0.3771, -0.3545, -0.3618, -0.3617, -0.3846, -0.3805, -0.3805, -0.3931, -0.402, -0.3923, -0.4019, -0.4125, -0.361, -0.3699, -0.3913, -0.4017, -0.4064, -0.4294, -0.3961, -0.404, -0.4249, -0.3943, -0.3927, -0.3879, -0.4091, -0.4221, -0.4125, -0.4134, -0.4039, -0.3866, -0.3984, -0.3943, -0.3863, -0.3418, -0.3911, -0.3964, -0.4142, -0.3921, -0.3718, -0.35, -0.35, -0.342, -0.338, -0.3072, -0.2928, -0.2677, -0.2624, -0.2378, -0.2286, -0.1685, -0.1681, -0.1783, -0.1498, -0.0931, -0.0876, 0.0394, 0.0, -0.0341, 0.0074, 0.0593, 0.0774, 0.0206, 0.0617, 0.2593, 0.2206, 0.3603, 0.371, 0.3397, 0.3313, 0.3439, 0.2674, 0.2581, 0.2373, 0.3375, 0.3435, 0.3971, 0.3971, 0.5058, 0.4809, 0.5482, 0.5423, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2050059631490400525", "ticker_idx": 0, "ts": "2026-05-01T03:47:59+00:00", "symbol": "6857.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Now I'm debating whether I should add more Advantest", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Author considers adding to Advantest position after Teradyne acknowledged its CPO testing capabilities on earnings call.", "resolved_tickers": ["6857.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "Interesting…\n\nAccording to GFHK’s analysis, low TSMC SoIC yields have led the industry to lower its near-term — i.e., 2026 — CPO shipment forecasts.\n\nFor what it’s worth, I was a bit disappointed that Teradyne seemed to acknowledge Advantest’s capabilities during this earnings call.\n\nI had thought Teradyne would be overwhelmingly dominant in CPO testing. Now I’m debating whether I should add more Advantest.", "reply_text": null, "parent_text": null, "quoted_text": "[GF Overseas Electronics &amp; Communications] \nTeradyne 1Q26 Review: Entering a Transition Period in 2H26\n\n☄️ Teradyne 1Q26 Review: Entering a Transition Period in 2H26\n\n☀️ Strong 1Q results, but full-year 2026 guidance fell short of expectations:\nTeradyne reported 1Q26 revenue of", "parent_author": null, "ret_m1d": 0.015998561927017763, "ret_signed_m1d": 0.015998561927017763, "alpha_spy_m1d": 0.01876002893889961, "alpha_c_m1d": 0.022079151270276265, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0036633796697419507, "alpha_c_p1d": 0.00594327161697672, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.07442027682904917, "ret_signed_p1w": 0.07442027682904917, "alpha_spy_p1w": 0.05087213261449408, "alpha_c_p1w": -0.036834616835910605, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0654, -0.0854, -0.1294, -0.1193, -0.0179, -0.0064, -0.0064, -0.0374, -0.026, -0.027, -0.038, -0.0321, -0.0667, -0.0854, -0.0854, -0.0443, 0.0274, 0.0098, -0.036, -0.0737, -0.0764, -0.1204, -0.0832, -0.0769, -0.1787, -0.1356, -0.1066, -0.1204, -0.1509, -0.1333, -0.1547, -0.0978, -0.139, -0.139, -0.1839, -0.1977, -0.159, -0.1755, -0.2076, -0.2477, -0.2691, -0.1911, -0.2405, -0.2251, -0.2116, -0.2019, -0.0933, -0.1084, -0.1016, -0.1055, -0.0293, -0.0083, 0.0291, 0.002, -0.0257, -0.0221, 0.0031, 0.0031, 0.0584, 0.1325, 0.0696, 0.0696, 0.016, 0.0, 0.0, 0.0, 0.0, 0.0683, 0.0744, 0.0347, 0.032, 0.0171, 0.0288, -0.0523, -0.0599, -0.0908, -0.0787, -0.0385, -0.0349, -0.0022, -0.0626, -0.0246, -0.053, -0.0591, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039649540085886996", "ticker_idx": 0, "ts": "2026-04-02T10:21:59+00:00", "symbol": "DRAM", "kind": "etf", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "if your goal is to take a pure bet on the memory cycle itself, $DRAM is the more fitting choice", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Positions DRAM ETF as the fitting pure-play vehicle on the memory cycle for investors wanting direct exposure to Samsung, SK Hynix, and Micron.", "resolved_tickers": ["DRAM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "self_thread", "lang": "en", "tweet_text": "Today, ticker 'DRAM' finally begins trading.\n\nWith roughly three-quarters of its portfolio concentrated in the memory Big 3 — Samsung Electronics, SK Hynix, and Micron — this ETF is effectively a direct investment vehicle for the memory semiconductor industry.\n\nIn particular, for overseas retail investors who have faced regulatory or institutional barriers to buying Samsung Electronics and SK Hynix directly, $DRAM could serve as a highly attractive alternative.\n\nOf course, $EWY also carries significant exposure to Samsung Electronics and SK Hynix, but it includes sizable allocations to other major Korean corporates such as Hyundai Motor.\n\nSo if your goal is to take a pure bet on the memory cycle itself, $DRAM is the more fitting choice.\n\n---\n\nMore info: https://t.co/WNeJz9xTdj", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": null, "ret_signed_m1d": null, "alpha_spy_m1d": null, "alpha_c_m1d": null, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.0008996643033920559, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.16642654182785965, "ret_signed_p1w": 0.16642654182785965, "alpha_spy_p1w": 0.12970977768592684, "alpha_c_p1w": 0.12970977768592684, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.03671676414193281, "ret_p1m": 0.45569162515240413, "ret_signed_p1m": 0.45569162515240413, "alpha_spy_p1m": 0.35685502780992695, "alpha_c_p1m": 0.35685502780992695, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.09883659734247718, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, 0.0, 0.0, 0.0504, 0.0605, 0.1697, 0.1664, 0.1682, 0.2086, 0.2813, 0.2417, 0.2622, 0.2821, 0.2633, 0.2583, 0.3437, 0.3098, 0.3401, 0.3894, 0.3447, 0.3772, 0.4168, 0.4557, 0.5299, 0.6675, 0.7536, 0.6769, 0.902, 0.9841, 0.848, 0.9647, 0.9377, 0.8408, 0.7767, 0.7929, 0.8555, 0.9575, 0.9027, 0.9027, 1.1798, 1.1877, 1.254, 1.2767, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2054528522593739180", "ticker_idx": 0, "ts": "2026-05-13T11:45:45+00:00", "symbol": "Ajinomoto", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Ajinomoto holds more than 95% share of the global ABF market, giving it strong pricing power across the supply chain", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish Ajinomoto on 30% ABF price hike + near-monopoly pricing power; bullish IC substrate sector on AI-driven Super Expansion Cycle through 2027–2028.", "resolved_tickers": ["2802.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Ajinomoto Japan 2802.T; dominant ABF market share", "bench_c": "XLP", "bench_c_reason": "industry='Packaged Foods'", "bench_c_industry": "Packaged Foods", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Ajinomoto Raises ABF Substrate Film Prices by 30%\n\n- Japan’s Ajinomoto has decided to raise prices for its core ABF build-up film by 30%, with the new prices set to take effect from Q3 2026.\n\n- Taiwanese package substrate makers said they have officially received notice of the price increase, at a time when cost pressures remain elevated across the IC substrate supply chain.\n\n- Specific new pricing is still under negotiation with suppliers, but in the near term, the hike is expected to increase production costs before being gradually passed through to product ASPs.\n\n- Despite the ABF price increase, the largest cost burden is still understood to come from repeated price hikes in upstream raw materials, particularly CCL.\n\n- With urgent demand from AI chip customers continuing, quarterly price increases for ABF and BT substrates are expected to persist through year-end. The magnitude of price hikes could widen further in the second half.\n\n- Ajinomoto holds more than 95% share of the global ABF market, giving it strong pricing power across the supply chain.\nIC substrate makers noted that the previous price increase took place around early 2025, roughly a year ago, and said the latest 30% increase is also a “reasonable level” given strong customer demand.\n\n- Unlike Nittobo, whose conservative capacity expansion stance has deepened the shortage of T-Glass fiberglass for IC substrates, Ajinomoto had already anticipated changes in ABF supply-demand dynamics three years ago and moved ahead with proactive capacity expansion.\n\n- As a result, despite its near-monopolistic market position, the supply bottleneck is viewed as relatively limited.\nAjinomoto recently announced plans to build a third ABF plant in Gifu Prefecture, Japan, in order to address demand beyond 2030.\n\n- The company has secured land for the project with an investment of around JPY 12 billion, or approximately USD 76 million, and plans to begin construction in 2028, with mass production targeted for 2032.\nThe new Gifu plant is expected to be much larger in scale than Ajinomoto’s existing production sites in Kanagawa and Gunma prefectures.\n\n- In addition, as AI chip packaging layer counts are expected to expand from the current 3+3 level to 11+11 layers, and potentially to 13+13 layers after 2030, ABF demand is likely to sustain structural growth.\n\n- According to the industry, as the AI CPU, GPU, and ASIC upgrade cycle accelerates, demand is rising for larger substrate area and higher layer counts. As a result, ABF substrates are understood to have re-entered a supply shortage phase from 1H 2026.\n\n- Shortages are also emerging simultaneously across the upstream value chain, including fiberglass, copper foil, and drill bits. Therefore, the ABF supply-demand imbalance in 2027–2028 is more likely to intensify than ease.\n\n- The industry expects the IC substrate sector to enter a so-called “Super Expansion Cycle” over the next two to three years, which should improve order visibility across the sector.\n\n---\n\nhttps://t.co/koAzxvobUe", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.008024803939449154, "ret_signed_m1d": -0.008024803939449154, "alpha_spy_m1d": -0.0024610840698585656, "alpha_c_m1d": -0.004719813674311513, "bench_spy_m1d": -0.005563719869590589, "bench_c_m1d": -0.0033049902651376417, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01076053255517051, "ret_signed_p1d": 0.01076053255517051, "alpha_spy_p1d": 0.0028662762864477553, "alpha_c_p1d": 0.0076915744285865895, "bench_spy_p1d": 0.007894256268722755, "bench_c_p1d": 0.0030689581265839205, "ret_p1w": -0.07568849170162317, "ret_signed_p1w": -0.07568849170162317, "alpha_spy_p1w": -0.07426052015054607, "alpha_c_p1w": -0.08513130816569792, "bench_spy_p1w": -0.0014279715510771007, "bench_c_p1w": 0.009442816464074744, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1952, -0.1749, -0.1588, -0.1582, -0.1629, -0.1651, -0.1651, -0.1479, -0.1042, -0.1165, -0.0987, -0.1047, -0.1448, -0.1776, -0.1847, -0.1804, -0.2117, -0.2067, -0.1894, -0.2021, -0.2009, -0.1876, -0.1854, -0.1697, -0.2009, -0.2009, -0.2424, -0.2212, -0.1871, -0.1811, -0.1782, -0.1889, -0.1981, -0.157, -0.1503, -0.1589, -0.1506, -0.1614, -0.1289, -0.1404, -0.1516, -0.1598, -0.1579, -0.1488, -0.1446, -0.1658, -0.1598, -0.139, -0.1505, -0.1629, -0.1322, -0.1271, -0.1306, -0.1306, -0.0717, -0.1096, -0.1096, -0.1096, -0.1096, -0.0958, -0.0826, 0.0029, -0.008, 0.0, 0.0108, -0.039, -0.0529, -0.0556, -0.0757, -0.056, -0.0334, -0.0015, -0.0254, -0.0354, -0.0622, -0.0604, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2054528522593739180", "ticker_idx": 1, "ts": "2026-05-13T11:45:45+00:00", "symbol": "IC substrates", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The industry expects the IC substrate sector to enter a so-called 'Super Expansion Cycle' over the next two to three years, which should improve order visibility across the sector", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish Ajinomoto on 30% ABF price hike + near-monopoly pricing power; bullish IC substrate sector on AI-driven Super Expansion Cycle through 2027–2028.", "resolved_tickers": ["3034.TW", "2367.TW", "3037.TW", "4062.T"], "resolution_method": "basket", "resolution_reason": "IC substrate basket: Unimicron, Unitech, Tripod, Ibiden", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Ajinomoto Raises ABF Substrate Film Prices by 30%\n\n- Japan’s Ajinomoto has decided to raise prices for its core ABF build-up film by 30%, with the new prices set to take effect from Q3 2026.\n\n- Taiwanese package substrate makers said they have officially received notice of the price increase, at a time when cost pressures remain elevated across the IC substrate supply chain.\n\n- Specific new pricing is still under negotiation with suppliers, but in the near term, the hike is expected to increase production costs before being gradually passed through to product ASPs.\n\n- Despite the ABF price increase, the largest cost burden is still understood to come from repeated price hikes in upstream raw materials, particularly CCL.\n\n- With urgent demand from AI chip customers continuing, quarterly price increases for ABF and BT substrates are expected to persist through year-end. The magnitude of price hikes could widen further in the second half.\n\n- Ajinomoto holds more than 95% share of the global ABF market, giving it strong pricing power across the supply chain.\nIC substrate makers noted that the previous price increase took place around early 2025, roughly a year ago, and said the latest 30% increase is also a “reasonable level” given strong customer demand.\n\n- Unlike Nittobo, whose conservative capacity expansion stance has deepened the shortage of T-Glass fiberglass for IC substrates, Ajinomoto had already anticipated changes in ABF supply-demand dynamics three years ago and moved ahead with proactive capacity expansion.\n\n- As a result, despite its near-monopolistic market position, the supply bottleneck is viewed as relatively limited.\nAjinomoto recently announced plans to build a third ABF plant in Gifu Prefecture, Japan, in order to address demand beyond 2030.\n\n- The company has secured land for the project with an investment of around JPY 12 billion, or approximately USD 76 million, and plans to begin construction in 2028, with mass production targeted for 2032.\nThe new Gifu plant is expected to be much larger in scale than Ajinomoto’s existing production sites in Kanagawa and Gunma prefectures.\n\n- In addition, as AI chip packaging layer counts are expected to expand from the current 3+3 level to 11+11 layers, and potentially to 13+13 layers after 2030, ABF demand is likely to sustain structural growth.\n\n- According to the industry, as the AI CPU, GPU, and ASIC upgrade cycle accelerates, demand is rising for larger substrate area and higher layer counts. As a result, ABF substrates are understood to have re-entered a supply shortage phase from 1H 2026.\n\n- Shortages are also emerging simultaneously across the upstream value chain, including fiberglass, copper foil, and drill bits. Therefore, the ABF supply-demand imbalance in 2027–2028 is more likely to intensify than ease.\n\n- The industry expects the IC substrate sector to enter a so-called “Super Expansion Cycle” over the next two to three years, which should improve order visibility across the sector.\n\n---\n\nhttps://t.co/koAzxvobUe", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002724632326834009, "ret_signed_m1d": 0.002724632326834009, "alpha_spy_m1d": 0.008288352196424598, "alpha_c_m1d": 0.012054624401064584, "bench_spy_m1d": -0.005563719869590589, "bench_c_m1d": -0.009329992074230575, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014314681682308567, "ret_signed_p1d": 0.014314681682308567, "alpha_spy_p1d": 0.006420425413585812, "alpha_c_p1d": -0.0006697333871105993, "bench_spy_p1d": 0.007894256268722755, "bench_c_p1d": 0.014984415069419166, "ret_p1w": -0.03590381411954088, "ret_signed_p1w": -0.03590381411954088, "alpha_spy_p1w": -0.03447584256846378, "alpha_c_p1w": -0.037543583846000955, "bench_spy_p1w": -0.0014279715510771007, "bench_c_p1w": 0.001639769726460072, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.363, -0.361, -0.3586, -0.3529, -0.3484, -0.3555, -0.3306, -0.2826, -0.2917, -0.2636, -0.2601, -0.2399, -0.2534, -0.2862, -0.2774, -0.2891, -0.3367, -0.3112, -0.273, -0.255, -0.2468, -0.2301, -0.214, -0.2064, -0.2298, -0.2277, -0.2674, -0.2873, -0.2334, -0.2265, -0.213, -0.2343, -0.2721, -0.2393, -0.2522, -0.2459, -0.2408, -0.2237, -0.178, -0.1592, -0.1674, -0.182, -0.1885, -0.1867, -0.1797, -0.1914, -0.1884, -0.1456, -0.136, -0.1746, -0.1455, -0.1209, -0.1289, -0.1461, -0.1091, -0.1106, -0.0937, -0.0851, -0.0879, -0.0253, -0.0354, -0.0184, 0.0027, 0.0, 0.0143, -0.0482, -0.0352, -0.0482, -0.0359, 0.0264, 0.0763, 0.1218, 0.1364, 0.1253, 0.0912, 0.156, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2042618014785507680", "ticker_idx": 0, "ts": "2026-04-10T14:57:39+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "consensus is that Anthropic's 10-trillion-parameter model, Mythos, was trained on NVL72", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares JPM note implying Anthropic's massive model trained on NVDA NVL72 with no ASIC threat; bullish for NVDA demand.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "*JPM: The consensus is that Anthropic’s 10-trillion-parameter model, Mythos, was trained on NVL72.\n*JPM: Our analysts are not currently seeing anything that would suggest Anthropic is developing its own ASIC.\n\n$NVDA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.025022536704250475, "ret_signed_m1d": -0.025022536704250475, "alpha_spy_m1d": -0.02568475546227056, "alpha_c_m1d": -0.009984065611985504, "bench_spy_m1d": 0.0006622187580200833, "bench_c_m1d": -0.015038471092264971, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0036049019677633076, "ret_signed_p1d": 0.0036049019677633076, "alpha_spy_p1d": -0.006167495816588486, "alpha_c_p1d": -0.011181747416227017, "bench_spy_p1d": 0.009772397784351794, "bench_c_p1d": 0.014786649383990325, "ret_p1w": 0.06918299027281538, "ret_signed_p1w": 0.06918299027281538, "alpha_spy_p1w": 0.024029498259099835, "alpha_c_p1w": 0.006740217716505903, "bench_spy_p1w": 0.045153492013715546, "bench_c_p1w": 0.06244277255630948, "ret_p1m": 0.16333561342870473, "ret_signed_p1m": 0.16333561342870473, "alpha_spy_p1m": 0.07526572868908321, "alpha_c_p1m": -0.15581378942204283, "bench_spy_p1m": 0.08806988473962152, "bench_c_p1m": 0.31914940285074755, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.015, -0.0292, -0.0084, -0.0128, -0.0128, -0.056, -0.0282, -0.0201, -0.0051, -0.0115, -0.0006, 0.0153, 0.0205, 0.0132, -0.0161, -0.044, -0.0766, -0.0888, -0.0171, 0.0074, -0.0005, 0.0075, -0.009, -0.0309, -0.0309, -0.0195, -0.0035, -0.0039, 0.0063, 0.0154, 0.0223, 0.0367, -0.0199, -0.0607, -0.0327, -0.0455, -0.0297, -0.0281, -0.0574, -0.0318, -0.0205, -0.0138, -0.0291, -0.0444, -0.0287, -0.0355, -0.0436, -0.0534, -0.0845, -0.0689, -0.0712, -0.0527, -0.0922, -0.1119, -0.1244, -0.0754, -0.0683, -0.0596, -0.0596, -0.0583, -0.0558, -0.0347, -0.025, 0.0, 0.0036, 0.0418, 0.0543, 0.0515, 0.0692, 0.0712, 0.0596, 0.0735, 0.0584, 0.1041, 0.1483, 0.1301, 0.1093, 0.058, 0.0521, 0.0522, 0.0417, 0.1018, 0.1212, 0.1409, 0.1633, 0.1704, 0.1972, 0.2497, 0.1945, 0.1786, 0.1695, 0.1847, 0.1637, 0.1415, 0.1415, 0.1391, 0.1271, 0.1358, 0.1193, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044323309983297714", "ticker_idx": 0, "ts": "2026-04-15T07:53:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics raised from KRW 270,000 to KRW 330,000... valuations remain remarkably cheap. On Daishin's revised estimates, Samsung and SK Hynix trade at just 4.8x and 4.0x 2027E P/E", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/adopts Daishin Buy upgrade on Samsung (PT KRW 330k) and SK Hynix (PT KRW 1.7M) on memory supercycle; bearish Kioxia and SanDisk on legacy DRAM shortage disadvantage.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities analyst Hyung-Geun Ryu published a memory report today, packed with significant alpha. \n\nKey takeaways:\n\n1. Apple is exploiting the memory shortage amid growth stagnation in the North American market. As I've noted previously, Daishin Securities estimates that Apple is capitalizing on competitors' struggles by pursuing an aggressive shipment strategy to expand the iOS ecosystem and strengthen its dominance in the end-device market. In other words, Apple is hoarding memory to prevent competitors from producing smartphones. They've even reportedly raised their iPhone shipment target for this year to 240 million units. As Apple stockpiles memory, Chinese smartphone OEMs have panicked into buying as well.\n\n- Some Chinese customers accepted mobile memory prices in 2Q26 that are 90–100% higher QoQ, following similar increases in 1Q26.\n- As a result, 2Q26 mobile memory prices are expected to rise ~80% QoQ.\n- Despite this, customers are not pushing back on pricing. Per Daishin's channel checks, some mobile customers are locking in annual contracts at the sharply elevated 2Q26 prices and seeking to increase purchase volumes in H2.\n\n2. DRAM capacity shortages are forcing a reduction in HBM4E stack counts. Originally expected at 16-hi and 20-hi, the products are now shifting to 12-hi and 16-hi.\n\n3. HBM4E I/O speed is increasing to 15–16 Gbps, pushing the commodity DRAM-to-HBM trade ratio from the previous 3:1 to as high as 5:1.\n\n4. Legacy DRAM shortages in SSDs are becoming severe. The three major DRAM makers, which can self-supply legacy DRAM (DDR4) used in SSDs, will gain a structural advantage in NAND as well. Bearish Kioxia, SanDisk.\n\n5. Kioxia is expected to increase this year's capex by 40% vs. prior plans.\n\n6. Samsung Electronics earnings revisions: 2026E OP raised from KRW 307T ($208.27B) to KRW 342T ($232.02B); 2027E OP raised from KRW 335T ($227.27B) to KRW 410T ($278.1B).\n\n7. SK Hynix earnings revisions: 2026E OP raised from KRW 232T ($157.3B) to KRW 263T ($178.4B); 2027E OP raised from KRW 260T ($176.3B) to KRW 332T ($225.2B).\n\n8. Target price upgrades: Samsung Electronics raised from KRW 270,000 to KRW 330,000; SK Hynix raised from KRW 1,450,000 to KRW 1,700,000.\n\n9. Despite the upgrades, valuations remain remarkably cheap. On Daishin's revised estimates, Samsung and SK Hynix trade at just 4.8x and 3.9x 2026E P/E, and 4.0x and 3.1x 2027E P/E, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.021327014218009532, "ret_signed_m1d": -0.021327014218009532, "alpha_spy_m1d": -0.013497799642994912, "alpha_c_m1d": -0.005625081001555343, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.01570193321645419, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0308056872037914, "ret_signed_p1d": 0.0308056872037914, "alpha_spy_p1d": 0.02834837558118708, "alpha_c_p1d": 0.019361900205923988, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.011443786997867411, "ret_p1w": 0.0308056872037914, "ret_signed_p1w": 0.0308056872037914, "alpha_spy_p1w": 0.014704279237480078, "alpha_c_p1w": -0.0210239410594244, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.0518296282632158, "ret_p1m": 0.40284360189573465, "ret_signed_p1m": 0.40284360189573465, "alpha_spy_p1m": 0.3339377238197241, "alpha_c_p1m": 0.20856551569911108, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.19427808619662357, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2958, -0.2939, -0.3133, -0.2929, -0.2797, -0.2806, -0.2806, -0.2456, -0.2319, -0.24, -0.2409, -0.2887, -0.2078, -0.2002, -0.2466, -0.2499, -0.213, -0.2158, -0.2064, -0.1553, -0.143, -0.143, -0.143, -0.143, -0.1014, -0.1009, -0.0872, -0.0541, -0.0375, 0.031, 0.0239, 0.0239, -0.0773, -0.1856, -0.0938, -0.1099, -0.1794, -0.1113, -0.1014, -0.1113, -0.1321, -0.1075, -0.0829, -0.0139, -0.0517, -0.0569, -0.1189, -0.1028, -0.1061, -0.1482, -0.1482, -0.1645, -0.2076, -0.1012, -0.1545, -0.1175, -0.0848, -0.0687, -0.0024, -0.0332, -0.0237, -0.0474, -0.0213, 0.0, 0.0308, 0.0237, 0.0166, 0.0379, 0.0308, 0.064, 0.0403, 0.064, 0.0521, 0.0711, 0.045, 0.045, 0.1019, 0.1019, 0.2607, 0.2867, 0.2725, 0.3531, 0.3223, 0.346, 0.4028, 0.282, 0.3318, 0.3057, 0.3081, 0.4194, 0.3863, 0.3863, 0.4171, 0.455, 0.4194, 0.5024, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044323309983297714", "ticker_idx": 1, "ts": "2026-04-15T07:53:53+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix raised from KRW 1,450,000 to KRW 1,700,000... valuations remain remarkably cheap... 3.9x and 3.1x 2027E P/E", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/adopts Daishin Buy upgrade on Samsung (PT KRW 330k) and SK Hynix (PT KRW 1.7M) on memory supercycle; bearish Kioxia and SanDisk on legacy DRAM shortage disadvantage.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities analyst Hyung-Geun Ryu published a memory report today, packed with significant alpha. \n\nKey takeaways:\n\n1. Apple is exploiting the memory shortage amid growth stagnation in the North American market. As I've noted previously, Daishin Securities estimates that Apple is capitalizing on competitors' struggles by pursuing an aggressive shipment strategy to expand the iOS ecosystem and strengthen its dominance in the end-device market. In other words, Apple is hoarding memory to prevent competitors from producing smartphones. They've even reportedly raised their iPhone shipment target for this year to 240 million units. As Apple stockpiles memory, Chinese smartphone OEMs have panicked into buying as well.\n\n- Some Chinese customers accepted mobile memory prices in 2Q26 that are 90–100% higher QoQ, following similar increases in 1Q26.\n- As a result, 2Q26 mobile memory prices are expected to rise ~80% QoQ.\n- Despite this, customers are not pushing back on pricing. Per Daishin's channel checks, some mobile customers are locking in annual contracts at the sharply elevated 2Q26 prices and seeking to increase purchase volumes in H2.\n\n2. DRAM capacity shortages are forcing a reduction in HBM4E stack counts. Originally expected at 16-hi and 20-hi, the products are now shifting to 12-hi and 16-hi.\n\n3. HBM4E I/O speed is increasing to 15–16 Gbps, pushing the commodity DRAM-to-HBM trade ratio from the previous 3:1 to as high as 5:1.\n\n4. Legacy DRAM shortages in SSDs are becoming severe. The three major DRAM makers, which can self-supply legacy DRAM (DDR4) used in SSDs, will gain a structural advantage in NAND as well. Bearish Kioxia, SanDisk.\n\n5. Kioxia is expected to increase this year's capex by 40% vs. prior plans.\n\n6. Samsung Electronics earnings revisions: 2026E OP raised from KRW 307T ($208.27B) to KRW 342T ($232.02B); 2027E OP raised from KRW 335T ($227.27B) to KRW 410T ($278.1B).\n\n7. SK Hynix earnings revisions: 2026E OP raised from KRW 232T ($157.3B) to KRW 263T ($178.4B); 2027E OP raised from KRW 260T ($176.3B) to KRW 332T ($225.2B).\n\n8. Target price upgrades: Samsung Electronics raised from KRW 270,000 to KRW 330,000; SK Hynix raised from KRW 1,450,000 to KRW 1,700,000.\n\n9. Despite the upgrades, valuations remain remarkably cheap. On Daishin's revised estimates, Samsung and SK Hynix trade at just 4.8x and 3.9x 2026E P/E, and 4.0x and 3.1x 2027E P/E, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029049309628226827, "ret_signed_m1d": -0.029049309628226827, "alpha_spy_m1d": -0.021220095053212207, "alpha_c_m1d": -0.02684180410946302, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.002207505518763808, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01672528671990836, "ret_signed_p1d": 0.01672528671990836, "alpha_spy_p1d": 0.014267975097304042, "alpha_c_p1d": 0.012751803733222378, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.003973482986685983, "ret_p1w": 0.07658443351171917, "ret_signed_p1w": 0.07658443351171917, "alpha_spy_p1w": 0.060483025545407854, "alpha_c_p1w": 0.02397960664137555, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.052604826870343624, "ret_p1m": 0.7341548394807627, "ret_signed_p1m": 0.7341548394807627, "alpha_spy_p1m": 0.6652489614047521, "alpha_c_p1m": 0.45746603847531286, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.2766888010054498, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3357, -0.3287, -0.3472, -0.3498, -0.3366, -0.3261, -0.3533, -0.2971, -0.261, -0.2435, -0.2013, -0.2707, -0.2031, -0.2092, -0.2602, -0.2628, -0.2206, -0.2303, -0.2444, -0.2197, -0.2268, -0.2268, -0.2268, -0.2268, -0.2145, -0.1662, -0.1644, -0.1169, -0.1055, -0.0326, -0.066, -0.066, -0.1734, -0.2526, -0.1717, -0.1866, -0.2641, -0.1743, -0.1593, -0.1813, -0.1989, -0.1426, -0.1461, -0.0704, -0.1083, -0.1136, -0.1787, -0.132, -0.1241, -0.1787, -0.1787, -0.2315, -0.2896, -0.2104, -0.2694, -0.2289, -0.2201, -0.1937, -0.0907, -0.1215, -0.096, -0.0845, -0.029, 0.0, 0.0167, -0.007, 0.0264, 0.0775, 0.0766, 0.0783, 0.0757, 0.1373, 0.1444, 0.1382, 0.132, 0.132, 0.2738, 0.2738, 0.4093, 0.456, 0.4842, 0.6549, 0.6153, 0.7394, 0.7342, 0.6012, 0.6197, 0.5361, 0.5361, 0.7077, 0.7086, 0.7086, 0.8063, 0.9745, 1.0153, 1.054, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044323309983297714", "ticker_idx": 2, "ts": "2026-04-15T07:53:53+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bearish Kioxia, SanDisk.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/adopts Daishin Buy upgrade on Samsung (PT KRW 330k) and SK Hynix (PT KRW 1.7M) on memory supercycle; bearish Kioxia and SanDisk on legacy DRAM shortage disadvantage.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities analyst Hyung-Geun Ryu published a memory report today, packed with significant alpha. \n\nKey takeaways:\n\n1. Apple is exploiting the memory shortage amid growth stagnation in the North American market. As I've noted previously, Daishin Securities estimates that Apple is capitalizing on competitors' struggles by pursuing an aggressive shipment strategy to expand the iOS ecosystem and strengthen its dominance in the end-device market. In other words, Apple is hoarding memory to prevent competitors from producing smartphones. They've even reportedly raised their iPhone shipment target for this year to 240 million units. As Apple stockpiles memory, Chinese smartphone OEMs have panicked into buying as well.\n\n- Some Chinese customers accepted mobile memory prices in 2Q26 that are 90–100% higher QoQ, following similar increases in 1Q26.\n- As a result, 2Q26 mobile memory prices are expected to rise ~80% QoQ.\n- Despite this, customers are not pushing back on pricing. Per Daishin's channel checks, some mobile customers are locking in annual contracts at the sharply elevated 2Q26 prices and seeking to increase purchase volumes in H2.\n\n2. DRAM capacity shortages are forcing a reduction in HBM4E stack counts. Originally expected at 16-hi and 20-hi, the products are now shifting to 12-hi and 16-hi.\n\n3. HBM4E I/O speed is increasing to 15–16 Gbps, pushing the commodity DRAM-to-HBM trade ratio from the previous 3:1 to as high as 5:1.\n\n4. Legacy DRAM shortages in SSDs are becoming severe. The three major DRAM makers, which can self-supply legacy DRAM (DDR4) used in SSDs, will gain a structural advantage in NAND as well. Bearish Kioxia, SanDisk.\n\n5. Kioxia is expected to increase this year's capex by 40% vs. prior plans.\n\n6. Samsung Electronics earnings revisions: 2026E OP raised from KRW 307T ($208.27B) to KRW 342T ($232.02B); 2027E OP raised from KRW 335T ($227.27B) to KRW 410T ($278.1B).\n\n7. SK Hynix earnings revisions: 2026E OP raised from KRW 232T ($157.3B) to KRW 263T ($178.4B); 2027E OP raised from KRW 260T ($176.3B) to KRW 332T ($225.2B).\n\n8. Target price upgrades: Samsung Electronics raised from KRW 270,000 to KRW 330,000; SK Hynix raised from KRW 1,450,000 to KRW 1,700,000.\n\n9. Despite the upgrades, valuations remain remarkably cheap. On Daishin's revised estimates, Samsung and SK Hynix trade at just 4.8x and 3.9x 2026E P/E, and 4.0x and 3.1x 2027E P/E, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.079913606911447, "ret_signed_m1d": -0.079913606911447, "alpha_spy_m1d": -0.08774282148646162, "alpha_c_m1d": -0.08212111243021081, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.002207505518763808, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04504782474544888, "ret_signed_p1d": -0.04504782474544888, "alpha_spy_p1d": -0.042590513122844564, "alpha_c_p1d": -0.0410743417587629, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.003973482986685983, "ret_p1w": 0.07374267201481022, "ret_signed_p1w": -0.07374267201481022, "alpha_spy_p1w": -0.0576412640484989, "alpha_c_p1w": -0.021137845144466594, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.052604826870343624, "ret_p1m": 0.49521752545510656, "ret_signed_p1m": -0.49521752545510656, "alpha_spy_p1m": -0.426311647379096, "alpha_c_p1m": -0.21852872444965676, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.2766888010054498, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5449, -0.5307, -0.5309, -0.4909, -0.4474, -0.4651, -0.4616, -0.4307, -0.4155, -0.4062, -0.3409, -0.4335, -0.3587, -0.3551, -0.3974, -0.4108, -0.4168, -0.4185, -0.4185, -0.3467, -0.2951, -0.3119, -0.3098, -0.3391, -0.3443, -0.3656, -0.3656, -0.3129, -0.3377, -0.3446, -0.3456, -0.3362, -0.3767, -0.4062, -0.3737, -0.3837, -0.4437, -0.3962, -0.3403, -0.3445, -0.35, -0.3022, -0.3332, -0.2783, -0.3101, -0.3101, -0.3377, -0.349, -0.3075, -0.347, -0.3746, -0.3823, -0.4113, -0.3274, -0.3491, -0.3258, -0.2965, -0.282, -0.1484, -0.1453, -0.07, -0.0352, 0.0799, 0.0, 0.045, -0.058, -0.0586, 0.0102, 0.0737, 0.0916, 0.067, 0.0944, 0.1206, 0.1206, 0.1589, 0.1234, 0.1234, 0.1234, 0.1234, 0.3394, 0.3727, 0.4175, 0.4224, 0.5582, 0.4952, 0.3715, 0.5875, 0.5356, 0.5825, 0.7075, 0.7711, 1.0194, 0.9272, 0.8683, 0.8908, 1.0318, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044323309983297714", "ticker_idx": 3, "ts": "2026-04-15T07:53:53+00:00", "symbol": "SanDisk", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Bearish Kioxia, SanDisk.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares/adopts Daishin Buy upgrade on Samsung (PT KRW 330k) and SK Hynix (PT KRW 1.7M) on memory supercycle; bearish Kioxia and SanDisk on legacy DRAM shortage disadvantage.", "resolved_tickers": ["SNDK"], "resolution_method": "exchange_ticker", "resolution_reason": "SanDisk (spun off from WD) listed on Nasdaq as SNDK", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "Daishin Securities analyst Hyung-Geun Ryu published a memory report today, packed with significant alpha. \n\nKey takeaways:\n\n1. Apple is exploiting the memory shortage amid growth stagnation in the North American market. As I've noted previously, Daishin Securities estimates that Apple is capitalizing on competitors' struggles by pursuing an aggressive shipment strategy to expand the iOS ecosystem and strengthen its dominance in the end-device market. In other words, Apple is hoarding memory to prevent competitors from producing smartphones. They've even reportedly raised their iPhone shipment target for this year to 240 million units. As Apple stockpiles memory, Chinese smartphone OEMs have panicked into buying as well.\n\n- Some Chinese customers accepted mobile memory prices in 2Q26 that are 90–100% higher QoQ, following similar increases in 1Q26.\n- As a result, 2Q26 mobile memory prices are expected to rise ~80% QoQ.\n- Despite this, customers are not pushing back on pricing. Per Daishin's channel checks, some mobile customers are locking in annual contracts at the sharply elevated 2Q26 prices and seeking to increase purchase volumes in H2.\n\n2. DRAM capacity shortages are forcing a reduction in HBM4E stack counts. Originally expected at 16-hi and 20-hi, the products are now shifting to 12-hi and 16-hi.\n\n3. HBM4E I/O speed is increasing to 15–16 Gbps, pushing the commodity DRAM-to-HBM trade ratio from the previous 3:1 to as high as 5:1.\n\n4. Legacy DRAM shortages in SSDs are becoming severe. The three major DRAM makers, which can self-supply legacy DRAM (DDR4) used in SSDs, will gain a structural advantage in NAND as well. Bearish Kioxia, SanDisk.\n\n5. Kioxia is expected to increase this year's capex by 40% vs. prior plans.\n\n6. Samsung Electronics earnings revisions: 2026E OP raised from KRW 307T ($208.27B) to KRW 342T ($232.02B); 2027E OP raised from KRW 335T ($227.27B) to KRW 410T ($278.1B).\n\n7. SK Hynix earnings revisions: 2026E OP raised from KRW 232T ($157.3B) to KRW 263T ($178.4B); 2027E OP raised from KRW 260T ($176.3B) to KRW 332T ($225.2B).\n\n8. Target price upgrades: Samsung Electronics raised from KRW 270,000 to KRW 330,000; SK Hynix raised from KRW 1,450,000 to KRW 1,700,000.\n\n9. Despite the upgrades, valuations remain remarkably cheap. On Daishin's revised estimates, Samsung and SK Hynix trade at just 4.8x and 3.9x 2026E P/E, and 4.0x and 3.1x 2027E P/E, respectively.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.05914418539706534, "ret_signed_m1d": -0.05914418539706534, "alpha_spy_m1d": -0.06697339997207996, "alpha_c_m1d": -0.07484611861351953, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.01570193321645419, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.031119634988234113, "ret_signed_p1d": -0.031119634988234113, "alpha_spy_p1d": -0.028662323365629794, "alpha_c_p1d": -0.0196758479903667, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.011443786997867411, "ret_p1w": 0.09795680201287627, "ret_signed_p1w": -0.09795680201287627, "alpha_spy_p1w": -0.08185539404656494, "alpha_c_p1w": -0.046127173749660466, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.0518296282632158, "ret_p1m": 0.5506212893413662, "ret_signed_p1m": -0.5506212893413662, "alpha_spy_p1m": -0.48171541126535566, "alpha_c_p1m": -0.35634320314474266, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.19427808619662357, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5362, -0.5362, -0.4919, -0.4378, -0.4354, -0.4686, -0.472, -0.4601, -0.4083, -0.3952, -0.3538, -0.254, -0.22, -0.3445, -0.3538, -0.3294, -0.3458, -0.3926, -0.3279, -0.2932, -0.2974, -0.2974, -0.3377, -0.3267, -0.3035, -0.2711, -0.2526, -0.2839, -0.2908, -0.2689, -0.2875, -0.3057, -0.3659, -0.3282, -0.3657, -0.4086, -0.3398, -0.306, -0.265, -0.306, -0.258, -0.2109, -0.1924, -0.1548, -0.1342, -0.2041, -0.2122, -0.2122, -0.2398, -0.3236, -0.3094, -0.358, -0.2875, -0.2232, -0.2132, -0.2132, -0.1874, -0.2029, -0.1243, -0.045, -0.0448, 0.0682, 0.0591, 0.0, 0.0311, 0.0328, 0.0239, 0.0132, 0.098, 0.0457, 0.1101, 0.2002, 0.1241, 0.1934, 0.2297, 0.3311, 0.4084, 0.5771, 0.5812, 0.5027, 0.7521, 0.7355, 0.6283, 0.623, 0.5506, 0.5785, 0.4949, 0.5513, 0.5617, 0.7295, 0.6582, 0.6582, 0.7826, 0.783, 0.841, 0.9008, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2058398967097139268", "ticker_idx": 0, "ts": "2026-05-24T04:05:31+00:00", "symbol": "ALAB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Demand is bound to grow exponentially... Astera Labs, a leading retimer company, is also seeing revenue grow. Its first-quarter revenue this year came in at $308.4 million, nearly double the figure from the same period a year earlier.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish ALAB as the leading retimer beneficiary of exponentially growing PCIe-driven demand in AI servers.", "resolved_tickers": ["ALAB"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "In the AI Era, the 'Retimer' Rises — The Key Component That Resuscitates Signals\n\nThe retimer is emerging as a hidden core component of the AI chip era. A retimer is a part that sits between chips, reviving weakened signals and retransmitting them. Its role is growing as communication speeds between the CPU and GPU increase.\n\nInside a server, CPU-to-CPU and CPU-to-GPU communication runs over PCIe (Peripheral Component Interconnect Express). As PCIe advances through each generation, communication speeds rise, giving way to channel loss — the phenomenon in which the signal weakens and its waveform becomes distorted.\n\nOn the 24th, a senior official at a domestic system semiconductor company said, \"CPUs talk to each other, and CPUs and GPUs communicate heavily over PCIe,\" adding, \"As these distances grow longer, speeds get faster, and channel counts increase, chips called 'retimers' have started to see widespread use.\"\n\nToday's high-performance servers use PCIe 5.0, which runs at 32 GT/s (gigatransfers per second). At this speed, even with ultra-low-loss printed circuit boards (PCBs), the distance over which a signal can arrive intact is extremely short. The retimer revives the dying signal at this point, enabling high-speed processor-to-processor communication even over longer distances.\n\nIn the past, redrivers were used. A redriver only amplifies a weakened signal and cannot fully restore it. The retimer, by contrast, restores the signal completely. This difference is what is driving up demand for retimers.\n\nThe key point is that as PCIe advances through each generation, faster speeds shorten the distance a signal can travel intact. As that distance shrinks, more retimers are needed. Demand is bound to grow exponentially. The same official forecast, \"As communication speeds rise, the number of retimers required will increase several times over from current levels.\"\n\nAstera Labs, a leading retimer company, is also seeing revenue grow. Its first-quarter revenue this year came in at $308.4 million, nearly double the figure from the same period a year earlier.\n\n$ALAB\n\n---\n\nhttps://t.co/YizYQmOrlH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0, "alpha_c_m1d": 0.0, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.038581843552863315, "ret_signed_p1d": 0.038581843552863315, "alpha_spy_p1d": 0.031943234961873124, "alpha_c_p1d": -0.006219667753498914, "bench_spy_p1d": 0.0066386085909901915, "bench_c_p1d": 0.04480151130636223, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5824, -0.5938, -0.6128, -0.6072, -0.6422, -0.6293, -0.609, -0.6116, -0.6014, -0.6204, -0.5936, -0.6093, -0.608, -0.5846, -0.5843, -0.5883, -0.5889, -0.6219, -0.5964, -0.6032, -0.6079, -0.6298, -0.6335, -0.6733, -0.6429, -0.6535, -0.6183, -0.6183, -0.6155, -0.6123, -0.5912, -0.5781, -0.5143, -0.4565, -0.4441, -0.4392, -0.4434, -0.4328, -0.4271, -0.3744, -0.3676, -0.3563, -0.3064, -0.3592, -0.4027, -0.3585, -0.3654, -0.3395, -0.3442, -0.2972, -0.303, -0.3625, -0.349, -0.3243, -0.3339, -0.2698, -0.255, -0.2418, -0.2975, -0.2041, -0.0632, -0.0295, 0.0, 0.0, 0.0386, 0.0601, 0.1378, 0.1172, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039353092500791534", "ticker_idx": 0, "ts": "2026-04-01T14:44:01+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It can also be interpreted as a sign that the company is very optimistic about its CPU business going forward", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish INTC: buying back Ireland fab stake signals improved financials and optimism about CPU business.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Intel buying back the stake in its Ireland fab is genuinely surprising. It is remarkable that the company’s financial position improved so much in such a short period of time. It can also be interpreted as a sign that the company is very optimistic about its CPU business going forward.\n\n---\n\nhttps://t.co/cWp95B7ZbZ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08119920487823784, "ret_signed_m1d": -0.08119920487823784, "alpha_spy_m1d": -0.07372108472083783, "alpha_c_m1d": -0.05933526810366485, "bench_spy_m1d": -0.0074781201574000145, "bench_c_m1d": -0.02186393677457299, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048927802384856944, "ret_signed_p1d": 0.048927802384856944, "alpha_spy_p1d": 0.048027327956765786, "alpha_c_p1d": 0.048034861337262535, "bench_spy_p1d": 0.0009004744280911581, "bench_c_p1d": 0.0008929410475944088, "ret_p1w": 0.22735794839015488, "ret_signed_p1w": 0.22735794839015488, "alpha_spy_p1w": 0.1956596090934024, "alpha_c_p1w": 0.1483977929939655, "bench_spy_p1w": 0.031698339296752476, "bench_c_p1w": 0.07896015539618939, "ret_p1m": 0.9671040132871873, "ret_signed_p1m": 0.9671040132871873, "alpha_spy_p1m": 0.870315075103681, "alpha_c_p1m": 0.6743520178726485, "bench_spy_p1m": 0.09678893818350631, "bench_c_p1m": 0.2927519954145388, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1801, -0.1803, -0.1664, -0.1124, -0.1441, -0.0516, -0.0827, -0.0154, 0.0144, 0.006, -0.0223, -0.0223, 0.011, 0.1295, 0.131, -0.0616, -0.1153, -0.0854, 0.0156, 0.0131, -0.0325, 0.0162, 0.0254, 0.0119, 0.0044, 0.0533, 0.046, -0.0187, 0.0054, -0.0323, -0.0258, -0.0258, -0.0385, -0.0535, -0.071, -0.0816, -0.0916, -0.0398, -0.0239, -0.0535, -0.0504, -0.0527, -0.1026, -0.051, -0.0433, -0.096, -0.051, -0.026, -0.001, -0.0579, -0.0471, -0.0473, -0.0827, -0.0625, -0.0385, -0.0866, -0.0837, -0.0827, -0.0177, -0.0818, -0.102, -0.1424, -0.0812, 0.0, 0.0489, 0.0489, 0.0573, 0.1016, 0.2274, 0.285, 0.2988, 0.3571, 0.3285, 0.3521, 0.4262, 0.4262, 0.3679, 0.3796, 0.3589, 0.3904, 0.7185, 0.7695, 0.7597, 0.9727, 0.9671, 1.0741, 0.9942, 1.2517, 1.3529, 1.2823, 1.6009, 1.695, 1.5111, 1.5045, 1.4137, 1.2646, 1.2521, 1.3069, 1.4768, 1.4672, 1.4951, 1.4951, 1.5717, 1.5353, 1.517, 1.3877, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2040716870471585807", "ticker_idx": 0, "ts": "2026-04-05T09:03:11+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "What makes the LTAs being discussed this time different from the ones in the past is that hyperscalers are paying upfront deposits. That's very interesting.", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Bullish SK Hynix: hyperscaler LTAs with upfront deposits signal stronger demand lock-in than prior agreements.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "What makes the LTAs being discussed this time different from the ones in the past is that hyperscalers are paying upfront deposits. That’s very interesting. https://t.co/0rfB9gyw92", "reply_text": null, "parent_text": null, "quoted_text": "MS and Google Pursuing LTA with SK Hynix\n\nSK Hynix is entering into long-term supply agreements (LTAs) for DRAM with major global AI companies including Microsoft (MS) and Google. Big Tech is moving to lock in DRAM—a product known for high price volatility—over extended periods. https://t.co/8qYYpygMQu", "parent_author": null, "ret_m1d": -0.011286664037370775, "ret_signed_m1d": -0.011286664037370775, "alpha_spy_m1d": -0.006582104189747029, "alpha_c_m1d": -0.0020437639904899996, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.033860062665668655, "ret_signed_p1d": 0.033860062665668655, "alpha_spy_p1d": 0.03341998855981987, "alpha_c_p1d": 0.023960616225450204, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.17381489427902386, "ret_signed_p1w": 0.17381489427902386, "alpha_spy_p1w": 0.13258140481556135, "alpha_c_p1w": 0.05421293448919573, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.6331828896337777, "ret_signed_p1m": 0.6331828896337777, "alpha_spy_p1m": 0.5347808930997202, "alpha_c_p1m": 0.31319201145924014, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1641, -0.1483, -0.1618, -0.1562, -0.1686, -0.1641, -0.1562, -0.1483, -0.1393, -0.1629, -0.1663, -0.1494, -0.1359, -0.1708, -0.0987, -0.0525, -0.03, 0.0241, -0.0649, 0.0218, 0.0139, -0.0514, -0.0548, -0.0007, -0.0131, -0.0311, 0.0004, -0.0086, -0.0086, -0.0086, -0.0086, 0.0072, 0.0691, 0.0714, 0.1322, 0.1469, 0.2404, 0.1975, 0.1975, 0.0598, -0.0418, 0.0621, 0.0429, -0.0564, 0.0587, 0.0779, 0.0497, 0.0271, 0.0993, 0.0948, 0.1919, 0.1433, 0.1366, 0.053, 0.1129, 0.123, 0.053, 0.053, -0.0147, -0.0892, 0.0124, -0.0632, -0.0113, 0.0, 0.0339, 0.1659, 0.1264, 0.1591, 0.1738, 0.2449, 0.2822, 0.3036, 0.2731, 0.316, 0.3815, 0.3804, 0.3826, 0.3792, 0.4582, 0.4673, 0.4594, 0.4515, 0.4515, 0.6332, 0.6332, 0.807, 0.8668, 0.9029, 1.1219, 1.0711, 1.2302, 1.2235, 1.053, 1.0767, 0.9695, 0.9695, 1.1896, 1.1907, 1.1907, 1.316, 1.5316, 1.584, 1.6336, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2050393594499150332", "ticker_idx": 0, "ts": "2026-05-02T01:55:02+00:00", "symbol": "memory", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Agentic AI will benefit CPUs as well, but it will make memory even more scarce", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish memory sector: agentic AI drives CPU DRAM integration 4x typical scale, prolonging shortage.", "resolved_tickers": ["MU", "000660.KS", "005930.KS"], "resolution_method": "basket", "resolution_reason": "Memory sector basket: Micron, SK Hynix, Samsung", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "'According to industry sources on the 2nd, CPU manufacturers are pursuing the integration of 300–400GB of DRAM into AI CPUs. This is up to four times the overwhelming scale compared to typical CPU products (96–256GB).'\n\nIt’s similar to what I’ve been thinking. Agentic AI will benefit CPUs as well, but it will make memory even more scarce.", "reply_text": null, "parent_text": null, "quoted_text": "AI CPUs Are Devouring DRAM — Memory 'Shortage' to Last Another Year\n\nThe memory industry has posted record-breaking results driven by commodity DRAM prices that have surged more than 100%, and with the proliferation of AI-purposed central processing units (CPUs) now coming into https://t.co/5UvPYMk0pz", "parent_author": null, "ret_m1d": -0.07409187866148274, "ret_signed_m1d": -0.07409187866148274, "alpha_spy_m1d": -0.0777687280264369, "alpha_c_m1d": -0.08007068394218757, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03686356067309456, "ret_signed_p1d": 0.03686356067309456, "alpha_spy_p1d": 0.0288413747881023, "alpha_c_p1d": 0.005489631376153735, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.3023000230111992, "ret_signed_p1w": 0.3023000230111992, "alpha_spy_p1w": 0.27264865640379865, "alpha_c_p1w": 0.16512291248908118, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3318, -0.3572, -0.352, -0.3363, -0.3456, -0.325, -0.301, -0.3005, -0.3005, -0.3073, -0.2951, -0.2814, -0.2623, -0.2619, -0.2412, -0.2268, -0.1947, -0.2075, -0.2073, -0.2851, -0.3264, -0.2796, -0.3038, -0.3342, -0.2821, -0.2662, -0.2826, -0.2815, -0.2503, -0.2323, -0.1915, -0.223, -0.2383, -0.2848, -0.2728, -0.2795, -0.3219, -0.3209, -0.3601, -0.379, -0.3088, -0.3413, -0.3195, -0.3006, -0.2889, -0.2251, -0.2339, -0.2249, -0.2256, -0.1806, -0.172, -0.1577, -0.1673, -0.1646, -0.1442, -0.1246, -0.1174, -0.1166, -0.0772, -0.0906, -0.0783, -0.0886, -0.0741, 0.0, 0.0369, 0.1356, 0.1442, 0.2052, 0.3023, 0.266, 0.3271, 0.3269, 0.2259, 0.2208, 0.201, 0.221, 0.317, 0.3008, 0.3008, 0.4194, 0.4937, 0.4908, 0.5535, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047942867029770502", "ticker_idx": 0, "ts": "2026-04-25T07:36:43+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "competitors such as SK Hynix and Micron could enjoy the windfall of a steeper rise in memory prices", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung strike could disrupt memory output 1–2 months, benefiting SK Hynix and Micron on tighter supply; Samsung itself faces production and shipment losses.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Samsung Union Strike Could Disrupt Memory Mass Production for At Least 1 Month\"… Industry Concerns Mount\n\nLabor-management conflict at Samsung Electronics is intensifying over the size of performance bonuses. The Samsung Group Cho-gi-eop Union's Samsung Electronics chapter (hereinafter \"the union\") reaffirmed on the 23rd at a resolution rally its intent to push ahead with an 18-day general strike starting on the 21st of next month.\n\nOn the 25th, concerns emerged within the semiconductor industry that the Samsung Electronics union strike could have a significant ripple effect on the memory semiconductor market.\n\nWhile the union has announced an 18-day strike period, when factoring in the time required to bring equipment back online to normal operating conditions, production disruption could last at least one month or more. The impact on mass production stretches to at least double that period. There are also forecasts that, amid the memory supercycle, only Samsung Electronics' competitors will reap windfall benefits.\n\nOne semiconductor engineer explained, \"If setup and maintenance (CS) work on semiconductor equipment is suspended for an extended period, restoring equipment to normal operation can take more than a month,\" adding, \"The impact could be even greater in a supercycle situation where memory shipments need to be aggressively ramped up.\"\n\nThe union strike is not an issue confined to Samsung Electronics alone. If Samsung Electronics—which holds the world's largest production capacity—experiences disruptions in mass production, the memory supply crunch could worsen further. For global Big Tech and IT companies, this means greater pressure from rising manufacturing costs.\n\nIn such a scenario, competitors such as SK Hynix and Micron could enjoy the windfall of a steeper rise in memory prices. Volatility is expected to be particularly pronounced in high-performance server DRAM and eSSD (enterprise SSD) segments, where only a handful of companies are capable of mass production.\n\nAnother semiconductor engineer said, \"When personnel are pulled out of the fab, equipment starts going offline one by one, creating problems,\" and added, \"If memory equipment isn't properly managed, output can immediately drop by 10–20%.\"\n\nThe union is demanding that 15% of annual operating profit be distributed as performance bonuses and that the bonus system be made more transparent. Approximately 40,000 people (police estimate) gathered at the resolution rally held on the 23rd at Samsung Electronics' Pyeongtaek campus, where the same message was emphasized. More than half of the 76,100 members of Samsung Electronics' first-ever majority union in the company's history attended the rally. If a labor-management agreement is not reached, the union will proceed with a general strike.\n\nAt the rally, Union Chairman Choi Seung-ho warned, \"If the company treats devoted union members as mere numbers, we too will respond with numbers—by demonstrating the massive production disruptions and losses that a strike will cause.\" Estimated losses from the strike have been pegged at up to 30 trillion won.\n\n---\n\nhttps://t.co/rQlCFt8C0w", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.054179530014305755, "ret_signed_m1d": -0.054179530014305755, "alpha_spy_m1d": -0.05245968648319865, "alpha_c_m1d": -0.054535064072515516, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006191921404816947, "ret_signed_p1d": 0.006191921404816947, "alpha_spy_p1d": 0.01105787013296311, "alpha_c_p1d": 0.03591976484541248, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.11996906990642398, "ret_signed_p1w": 0.11996906990642398, "alpha_spy_p1w": 0.11599794846210365, "alpha_c_p1w": 0.11892217943695504, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.5882352428889039, "ret_signed_p1m": 0.5882352428889039, "alpha_spy_p1m": 0.5387084942952591, "alpha_c_p1m": 0.3988463853190545, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3503, -0.3348, -0.2977, -0.3588, -0.2993, -0.3047, -0.3495, -0.3518, -0.3147, -0.3232, -0.3356, -0.314, -0.3201, -0.3201, -0.3201, -0.3201, -0.3093, -0.2668, -0.2653, -0.2236, -0.2135, -0.1494, -0.1788, -0.1788, -0.2732, -0.3429, -0.2717, -0.2848, -0.3529, -0.274, -0.2608, -0.2802, -0.2957, -0.2461, -0.2492, -0.1827, -0.2159, -0.2206, -0.2779, -0.2368, -0.2299, -0.2779, -0.2779, -0.3243, -0.3754, -0.3057, -0.3576, -0.322, -0.3142, -0.291, -0.2005, -0.2276, -0.2051, -0.195, -0.1463, -0.1207, -0.106, -0.1269, -0.0975, -0.0526, -0.0534, -0.0519, -0.0542, 0.0, 0.0062, 0.0008, -0.0046, -0.0046, 0.12, 0.12, 0.2392, 0.2802, 0.305, 0.4551, 0.4203, 0.5294, 0.5248, 0.4079, 0.4241, 0.3506, 0.3506, 0.5015, 0.5023, 0.5023, 0.5882, 0.7361, 0.772, 0.806, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047942867029770502", "ticker_idx": 1, "ts": "2026-04-25T07:36:43+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "competitors such as SK Hynix and Micron could enjoy the windfall of a steeper rise in memory prices", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung strike could disrupt memory output 1–2 months, benefiting SK Hynix and Micron on tighter supply; Samsung itself faces production and shipment losses.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Samsung Union Strike Could Disrupt Memory Mass Production for At Least 1 Month\"… Industry Concerns Mount\n\nLabor-management conflict at Samsung Electronics is intensifying over the size of performance bonuses. The Samsung Group Cho-gi-eop Union's Samsung Electronics chapter (hereinafter \"the union\") reaffirmed on the 23rd at a resolution rally its intent to push ahead with an 18-day general strike starting on the 21st of next month.\n\nOn the 25th, concerns emerged within the semiconductor industry that the Samsung Electronics union strike could have a significant ripple effect on the memory semiconductor market.\n\nWhile the union has announced an 18-day strike period, when factoring in the time required to bring equipment back online to normal operating conditions, production disruption could last at least one month or more. The impact on mass production stretches to at least double that period. There are also forecasts that, amid the memory supercycle, only Samsung Electronics' competitors will reap windfall benefits.\n\nOne semiconductor engineer explained, \"If setup and maintenance (CS) work on semiconductor equipment is suspended for an extended period, restoring equipment to normal operation can take more than a month,\" adding, \"The impact could be even greater in a supercycle situation where memory shipments need to be aggressively ramped up.\"\n\nThe union strike is not an issue confined to Samsung Electronics alone. If Samsung Electronics—which holds the world's largest production capacity—experiences disruptions in mass production, the memory supply crunch could worsen further. For global Big Tech and IT companies, this means greater pressure from rising manufacturing costs.\n\nIn such a scenario, competitors such as SK Hynix and Micron could enjoy the windfall of a steeper rise in memory prices. Volatility is expected to be particularly pronounced in high-performance server DRAM and eSSD (enterprise SSD) segments, where only a handful of companies are capable of mass production.\n\nAnother semiconductor engineer said, \"When personnel are pulled out of the fab, equipment starts going offline one by one, creating problems,\" and added, \"If memory equipment isn't properly managed, output can immediately drop by 10–20%.\"\n\nThe union is demanding that 15% of annual operating profit be distributed as performance bonuses and that the bonus system be made more transparent. Approximately 40,000 people (police estimate) gathered at the resolution rally held on the 23rd at Samsung Electronics' Pyeongtaek campus, where the same message was emphasized. More than half of the 76,100 members of Samsung Electronics' first-ever majority union in the company's history attended the rally. If a labor-management agreement is not reached, the union will proceed with a general strike.\n\nAt the rally, Union Chairman Choi Seung-ho warned, \"If the company treats devoted union members as mere numbers, we too will respond with numbers—by demonstrating the massive production disruptions and losses that a strike will cause.\" Estimated losses from the strike have been pegged at up to 30 trillion won.\n\n---\n\nhttps://t.co/rQlCFt8C0w", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05307304496618781, "ret_signed_m1d": -0.05307304496618781, "alpha_spy_m1d": -0.05135320143508071, "alpha_c_m1d": -0.05342857902439757, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0386418886457458, "ret_signed_p1d": -0.0386418886457458, "alpha_spy_p1d": -0.03377593991759964, "alpha_c_p1d": -0.008914045205150267, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.09892102884311416, "ret_signed_p1w": 0.09892102884311416, "alpha_spy_p1w": 0.09494990739879383, "alpha_c_p1w": 0.09787413837364523, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.707869469750678, "ret_signed_p1m": 0.707869469750678, "alpha_spy_p1m": 0.6583427211570332, "alpha_c_p1m": 0.5184806121808285, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1705, -0.1696, -0.2094, -0.1657, -0.2007, -0.277, -0.2704, -0.2479, -0.2692, -0.2888, -0.2181, -0.2112, -0.2156, -0.2156, -0.2382, -0.1979, -0.2047, -0.1841, -0.1978, -0.2035, -0.1825, -0.2081, -0.2142, -0.2136, -0.2765, -0.2363, -0.2434, -0.2944, -0.2581, -0.2319, -0.2022, -0.2276, -0.188, -0.1581, -0.1202, -0.1201, -0.1534, -0.1941, -0.2295, -0.2463, -0.2719, -0.3226, -0.3193, -0.3865, -0.356, -0.2987, -0.3018, -0.3018, -0.2799, -0.2802, -0.2246, -0.1965, -0.1982, -0.1868, -0.1123, -0.1303, -0.1284, -0.1325, -0.1452, -0.1433, -0.0707, -0.0817, -0.0531, 0.0, -0.0386, -0.0116, -0.0141, 0.0336, 0.0989, 0.2205, 0.2708, 0.2327, 0.4237, 0.5162, 0.4614, 0.532, 0.4794, 0.3815, 0.2993, 0.332, 0.3954, 0.4528, 0.4317, 0.4317, 0.7079, 0.7699, 0.7606, 0.8511, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047942867029770502", "ticker_idx": 2, "ts": "2026-04-25T07:36:43+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "production disruption could last at least one month or more. The impact on mass production stretches to at least double that period", "tweet_type": "analysis", "timeline": "months", "summary": "Samsung strike could disrupt memory output 1–2 months, benefiting SK Hynix and Micron on tighter supply; Samsung itself faces production and shipment losses.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "\"Samsung Union Strike Could Disrupt Memory Mass Production for At Least 1 Month\"… Industry Concerns Mount\n\nLabor-management conflict at Samsung Electronics is intensifying over the size of performance bonuses. The Samsung Group Cho-gi-eop Union's Samsung Electronics chapter (hereinafter \"the union\") reaffirmed on the 23rd at a resolution rally its intent to push ahead with an 18-day general strike starting on the 21st of next month.\n\nOn the 25th, concerns emerged within the semiconductor industry that the Samsung Electronics union strike could have a significant ripple effect on the memory semiconductor market.\n\nWhile the union has announced an 18-day strike period, when factoring in the time required to bring equipment back online to normal operating conditions, production disruption could last at least one month or more. The impact on mass production stretches to at least double that period. There are also forecasts that, amid the memory supercycle, only Samsung Electronics' competitors will reap windfall benefits.\n\nOne semiconductor engineer explained, \"If setup and maintenance (CS) work on semiconductor equipment is suspended for an extended period, restoring equipment to normal operation can take more than a month,\" adding, \"The impact could be even greater in a supercycle situation where memory shipments need to be aggressively ramped up.\"\n\nThe union strike is not an issue confined to Samsung Electronics alone. If Samsung Electronics—which holds the world's largest production capacity—experiences disruptions in mass production, the memory supply crunch could worsen further. For global Big Tech and IT companies, this means greater pressure from rising manufacturing costs.\n\nIn such a scenario, competitors such as SK Hynix and Micron could enjoy the windfall of a steeper rise in memory prices. Volatility is expected to be particularly pronounced in high-performance server DRAM and eSSD (enterprise SSD) segments, where only a handful of companies are capable of mass production.\n\nAnother semiconductor engineer said, \"When personnel are pulled out of the fab, equipment starts going offline one by one, creating problems,\" and added, \"If memory equipment isn't properly managed, output can immediately drop by 10–20%.\"\n\nThe union is demanding that 15% of annual operating profit be distributed as performance bonuses and that the bonus system be made more transparent. Approximately 40,000 people (police estimate) gathered at the resolution rally held on the 23rd at Samsung Electronics' Pyeongtaek campus, where the same message was emphasized. More than half of the 76,100 members of Samsung Electronics' first-ever majority union in the company's history attended the rally. If a labor-management agreement is not reached, the union will proceed with a general strike.\n\nAt the rally, Union Chairman Choi Seung-ho warned, \"If the company treats devoted union members as mere numbers, we too will respond with numbers—by demonstrating the massive production disruptions and losses that a strike will cause.\" Estimated losses from the strike have been pegged at up to 30 trillion won.\n\n---\n\nhttps://t.co/rQlCFt8C0w", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022271714922049046, "ret_signed_m1d": 0.022271714922049046, "alpha_spy_m1d": 0.020551871390941945, "alpha_c_m1d": 0.02009194231477185, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": -0.0021797726072771972, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011135857461024523, "ret_signed_p1d": 0.011135857461024523, "alpha_spy_p1d": 0.00626990873287836, "alpha_c_p1d": -0.0058038018566148875, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.01693965931763941, "ret_p1w": 0.03563474387527843, "ret_signed_p1w": -0.03563474387527843, "alpha_spy_p1w": -0.031663622430958105, "alpha_c_p1w": -0.02641760705002638, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.009217136825252048, "ret_p1m": 0.33184855233853017, "ret_signed_p1m": -0.33184855233853017, "alpha_spy_p1m": -0.28232180374488536, "alpha_c_p1m": -0.17883123313383842, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.15301731920469175, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2781, -0.2857, -0.2866, -0.3315, -0.2554, -0.2483, -0.2919, -0.295, -0.2603, -0.263, -0.2541, -0.2061, -0.1945, -0.1945, -0.1945, -0.1945, -0.1554, -0.155, -0.1421, -0.111, -0.0954, -0.031, -0.0376, -0.0376, -0.1328, -0.2345, -0.1483, -0.1634, -0.2288, -0.1648, -0.1554, -0.1648, -0.1843, -0.1612, -0.1381, -0.0732, -0.1087, -0.1136, -0.1719, -0.1568, -0.1599, -0.1994, -0.1994, -0.2147, -0.2552, -0.1552, -0.2053, -0.1706, -0.1399, -0.1247, -0.0624, -0.0913, -0.0824, -0.1047, -0.0802, -0.0601, -0.0312, -0.0379, -0.0445, -0.0245, -0.0312, 0.0, -0.0223, 0.0, -0.0111, 0.0067, -0.0178, -0.0178, 0.0356, 0.0356, 0.1849, 0.2094, 0.196, 0.2717, 0.2428, 0.265, 0.3185, 0.2049, 0.2517, 0.2272, 0.2294, 0.3341, 0.3029, 0.3029, 0.3318, 0.3675, 0.3341, 0.412, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2049637296136491214", "ticker_idx": 0, "ts": "2026-04-29T23:49:46+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Earnings surprise: +49.7% — Highest operating profit in the past five quarters — Highest operating margin in the past five quarters: 42.8%", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Samsung 1Q26 blowout: op profit KRW 57.2T vs consensus KRW 38.2T (+49.7%), highest margin in 5 quarters at 42.8%; durable beat implies bullish.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "🔴 Samsung Electronics (Market Cap: KRW 1,321.3T)\n📁 Preliminary Earnings Results Based on Consolidated Financial Statements — Fair Disclosure\n Apr. 30, 2026, 08:46 (Current price: KRW 226,000, +1.8%)\nRevenue: KRW 133.8734T (Consensus: KRW 117.4850T, +13.9%)\n Operating Profit: KRW 57.2328T (Consensus: KRW 38.2309T, +49.7%)\n Net Income: KRW 47.2253T (Consensus: KRW 31.5201T, +49.8%)\n* Recent Earnings\n (Period / Revenue / Operating Profit / Net Income)\n\n1Q26: KRW 133.8734T / KRW 57.2328T / KRW 47.2253T\n 4Q25: KRW 93.8374T / KRW 20.0737T / KRW 19.6417T\n 3Q25: KRW 86.0617T / KRW 12.1661T / KRW 12.2257T\n 2Q25: KRW 74.5663T / KRW 4.6761T / KRW 5.1164T\n 1Q25: KRW 79.1405T / KRW 6.6853T / KRW 8.2229T\n\n- Earnings surprise: +49.7%\n- Highest operating profit in the past five quarters\n- Highest operating margin in the past five quarters: 42.8%", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017699115044247815, "ret_signed_m1d": -0.017699115044247815, "alpha_spy_m1d": -0.017853680024883856, "alpha_c_m1d": -0.009780099884908022, "bench_spy_m1d": 0.00015456498063604052, "bench_c_m1d": -0.007919015159339793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02433628318584069, "ret_signed_p1d": -0.02433628318584069, "alpha_spy_p1d": -0.03428591061304487, "alpha_c_p1d": -0.026787413775699154, "bench_spy_p1d": 0.009949627427204177, "bench_c_p1d": 0.0024511305898584634, "ret_p1w": 0.17699115044247793, "ret_signed_p1w": 0.17699115044247793, "alpha_spy_p1w": 0.14572270632428008, "alpha_c_p1w": 0.10835939802555994, "bench_spy_p1w": 0.03126844411819785, "bench_c_p1w": 0.06863175241691799, "ret_p1m": 0.3252212389380531, "ret_signed_p1m": 0.3252212389380531, "alpha_spy_p1m": 0.2647642875581746, "alpha_c_p1m": 0.15087641217196412, "bench_spy_p1m": 0.0604569513798785, "bench_c_p1m": 0.17434482676608898, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2913, -0.3359, -0.2604, -0.2533, -0.2966, -0.2997, -0.2652, -0.2679, -0.2591, -0.2114, -0.1999, -0.1999, -0.1999, -0.1999, -0.161, -0.1606, -0.1478, -0.1169, -0.1014, -0.0374, -0.044, -0.044, -0.1385, -0.2396, -0.154, -0.169, -0.2339, -0.1703, -0.161, -0.1703, -0.1897, -0.1668, -0.1438, -0.0793, -0.1147, -0.1195, -0.1774, -0.1624, -0.1654, -0.2047, -0.2047, -0.2199, -0.2602, -0.1608, -0.2106, -0.1761, -0.1456, -0.1305, -0.0686, -0.0973, -0.0885, -0.1106, -0.0863, -0.0664, -0.0376, -0.0442, -0.0509, -0.031, -0.0376, -0.0066, -0.0288, -0.0066, -0.0177, 0.0, -0.0243, -0.0243, 0.0288, 0.0288, 0.177, 0.2013, 0.1881, 0.2633, 0.2345, 0.2566, 0.3097, 0.1969, 0.2434, 0.219, 0.2212, 0.3252, 0.2942, 0.2942, 0.323, 0.3584, 0.3252, 0.4027, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2052651707105181882", "ticker_idx": 0, "ts": "2026-05-08T07:27:58+00:00", "symbol": "Macronix", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Macronix returned to profitability in the first quarter of 2026, with rising prices continuing to drive operational growth... surpassing the previous monthly record... Revenue for the first four months of the year totaled NT$16.38 billion, up 93.5% year over year", "tweet_type": "analysis", "timeline": "months", "summary": "Macronix and Winbond named as dominant 2D NAND beneficiaries with expanding share, pricing power, and margin expansion as major producers exit the market.", "resolved_tickers": ["2337.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Macronix International 2337.TW Taiwan NOR flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "2D NAND shortage spirals after Samsung, Micron, and rivals exit market\n\nMajor international NAND manufacturers are gradually exiting the mature-process 2D NAND market, triggering panic buying and a sharp spike in prices. Industry sources suggest the supply shortage may be difficult to resolve. While vendors are seeking alternative solutions, product specifications are expected to become increasingly polarized as scarcity persists.\n\nSupply chain sources said multi-level cell (MLC) NAND has become extremely difficult to secure, with prices rising at least 20–30% in April 2026 alone. Spot market prices have climbed even faster. For example, MLC 64Gb 8GBx8 products are currently trading at US$19–20, with some offers reaching US$28, compared with only around US$6–7 at the end of 2025 — representing an increase of roughly 200–300%.\n\nSingle-level cell (SLC) NAND prices have also surged sharply. Spot prices for SLC 2Gb products recently reached around US$4–4.5, up roughly 120–150% from late 2025 levels.\n\nIndustry sources said that although other players are interested in entering the 2D NAND manufacturing market, second-hand equipment phased out by major NAND makers has recently become highly sought after. However, companies including Samsung, SK Hynix, Kioxia, and Micron Technology currently have no plans to license MLC technology externally, making it more difficult for new entrants to break into 2D NAND manufacturing.\n\nThe challenge is particularly acute for niche applications that remain dependent on 2D NAND, where transitioning to alternative products is difficult, and qualification cycles are typically lengthy. Although some Chinese manufacturers are interested in entering the market, their production capacity remains limited. End customers are also reluctant to take on higher supply risks, especially for critical applications.\n\nThe global NAND market is estimated to have reached around US$73 billion in 2025, with the six major NAND suppliers focusing heavily on 3D NAND development. Yet only around 1% of the overall NAND market remains tied to 2D NAND products. Despite the small share, price increases for both MLC and SLC products over the past six months have significantly outpaced those of 3D NAND.\n\nMacronix and Winbond gain pricing power in shrinking 2D NAND market\nMacronix and Winbond are expected to remain the leading suppliers of MLC and SLC NAND, respectively. Based on an estimated 2025 market value of around US$600 million, the two Taiwanese companies already account for roughly 59% of the 2D NAND market, with their combined share expected to expand further in 2026.\n\nMacronix returned to profitability in the first quarter of 2026, with rising prices continuing to drive operational growth. April revenue reached NT$5.91 billion (US$188 million), up 33.7% sequentially and 153.7% from a year earlier, surpassing the previous monthly record of NT$5.68 billion set in September 2021. Revenue for the first four months of the year totaled NT$16.38 billion, up 93.5% year over year.\n\nMacronix president Miin Wu said the company plans to gradually raise the proportion of NOR and NAND products, including eMMC, toward a 1:1 ratio through capacity expansion. He noted that current 4GB–32GB eMMC applications align closely with Macronix's product offerings.\n\nWu added that there remains a clear supply gap in MLC eMMC products, with shortages expected to persist for the foreseeable future. He estimated that meaningful new competition is unlikely to emerge for at least another three years.\n\nWinbond recently said it will focus on maintaining a stable supply of SLC NAND and NOR Flash products and does not plan to enter the MLC NAND market.\n\nThe company expects SLC NAND bit growth to exceed 80% over the next year. In particular, its new 24nm process accounted for only 1% of Flash-related business in the first quarter of 2026, but the contribution of 24nm SLC NAND is expected to increase from the second quarter onward, gradually becoming a meaningful revenue driver. Winbond also expects SLC NAND gross margins to surpass those of NOR Flash as it pushes toward becoming a leading SLC NAND supplier.\n\nWinbond president Chen Pei-ming said the current Flash shortage appears difficult to resolve and that the market is showing two clear trends.\n\nThe first involves products originally requiring only 16GB or 32GB storage being upgraded toward higher-capacity UFS solutions such as 36GB and 48GB. The second involves customers previously using 4GB eMMC solutions compressing or optimizing code to reduce storage requirements from above 1GB down to around 1GB, enabling the adoption of 8Gb SLC NAND products and creating new business opportunities.\n\n---\n\nhttps://t.co/W6a2xE8OFZ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07516339869281041, "ret_signed_m1d": 0.07516339869281041, "alpha_spy_m1d": 0.08335186710110531, "alpha_c_m1d": 0.1218326602927926, "bench_spy_m1d": -0.008188468408294902, "bench_c_m1d": -0.04666926159998219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.042483660130719025, "ret_signed_p1d": 0.042483660130719025, "alpha_spy_p1d": 0.04020607453528746, "alpha_c_p1d": 0.02523859374499171, "bench_spy_p1d": 0.002277585595431564, "bench_c_p1d": 0.017245066385727315, "ret_p1w": 0.06209150326797386, "ret_signed_p1w": 0.06209150326797386, "alpha_spy_p1w": 0.059990166803087996, "alpha_c_p1w": 0.08009544220868303, "bench_spy_p1w": 0.0021013364648858612, "bench_c_p1w": -0.018003938940709174, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4144, -0.3562, -0.3562, -0.3562, -0.3562, -0.3562, -0.3562, -0.3562, -0.3562, -0.3235, -0.2582, -0.2941, -0.281, -0.281, -0.2418, -0.317, -0.3595, -0.2974, -0.3301, -0.3967, -0.4072, -0.3484, -0.3529, -0.2908, -0.2222, -0.1471, -0.0621, 0.0261, 0.0458, -0.0294, -0.1078, -0.0719, -0.1634, -0.183, -0.1634, -0.2451, -0.1699, -0.1569, -0.1569, -0.1569, -0.1078, -0.0196, -0.0719, -0.0458, -0.0359, 0.0163, -0.0719, -0.0915, -0.1438, -0.2124, -0.134, -0.1209, -0.1765, -0.1373, -0.0523, 0.0425, 0.0915, 0.0065, 0.0065, 0.0098, 0.049, 0.1242, 0.0752, 0.0, 0.0425, 0.0, 0.098, 0.1046, 0.0621, 0.0458, -0.0588, -0.0784, -0.0784, -0.0229, -0.0033, 0.049, 0.0163, 0.0458, 0.0882, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2052651707105181882", "ticker_idx": 1, "ts": "2026-05-08T07:27:58+00:00", "symbol": "Winbond", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Winbond expects SLC NAND gross margins to surpass those of NOR Flash as it pushes toward becoming a leading SLC NAND supplier... SLC NAND bit growth to exceed 80% over the next year", "tweet_type": "analysis", "timeline": "months", "summary": "Macronix and Winbond named as dominant 2D NAND beneficiaries with expanding share, pricing power, and margin expansion as major producers exit the market.", "resolved_tickers": ["2344.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Winbond Electronics listed on TWSE as 2344.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "2D NAND shortage spirals after Samsung, Micron, and rivals exit market\n\nMajor international NAND manufacturers are gradually exiting the mature-process 2D NAND market, triggering panic buying and a sharp spike in prices. Industry sources suggest the supply shortage may be difficult to resolve. While vendors are seeking alternative solutions, product specifications are expected to become increasingly polarized as scarcity persists.\n\nSupply chain sources said multi-level cell (MLC) NAND has become extremely difficult to secure, with prices rising at least 20–30% in April 2026 alone. Spot market prices have climbed even faster. For example, MLC 64Gb 8GBx8 products are currently trading at US$19–20, with some offers reaching US$28, compared with only around US$6–7 at the end of 2025 — representing an increase of roughly 200–300%.\n\nSingle-level cell (SLC) NAND prices have also surged sharply. Spot prices for SLC 2Gb products recently reached around US$4–4.5, up roughly 120–150% from late 2025 levels.\n\nIndustry sources said that although other players are interested in entering the 2D NAND manufacturing market, second-hand equipment phased out by major NAND makers has recently become highly sought after. However, companies including Samsung, SK Hynix, Kioxia, and Micron Technology currently have no plans to license MLC technology externally, making it more difficult for new entrants to break into 2D NAND manufacturing.\n\nThe challenge is particularly acute for niche applications that remain dependent on 2D NAND, where transitioning to alternative products is difficult, and qualification cycles are typically lengthy. Although some Chinese manufacturers are interested in entering the market, their production capacity remains limited. End customers are also reluctant to take on higher supply risks, especially for critical applications.\n\nThe global NAND market is estimated to have reached around US$73 billion in 2025, with the six major NAND suppliers focusing heavily on 3D NAND development. Yet only around 1% of the overall NAND market remains tied to 2D NAND products. Despite the small share, price increases for both MLC and SLC products over the past six months have significantly outpaced those of 3D NAND.\n\nMacronix and Winbond gain pricing power in shrinking 2D NAND market\nMacronix and Winbond are expected to remain the leading suppliers of MLC and SLC NAND, respectively. Based on an estimated 2025 market value of around US$600 million, the two Taiwanese companies already account for roughly 59% of the 2D NAND market, with their combined share expected to expand further in 2026.\n\nMacronix returned to profitability in the first quarter of 2026, with rising prices continuing to drive operational growth. April revenue reached NT$5.91 billion (US$188 million), up 33.7% sequentially and 153.7% from a year earlier, surpassing the previous monthly record of NT$5.68 billion set in September 2021. Revenue for the first four months of the year totaled NT$16.38 billion, up 93.5% year over year.\n\nMacronix president Miin Wu said the company plans to gradually raise the proportion of NOR and NAND products, including eMMC, toward a 1:1 ratio through capacity expansion. He noted that current 4GB–32GB eMMC applications align closely with Macronix's product offerings.\n\nWu added that there remains a clear supply gap in MLC eMMC products, with shortages expected to persist for the foreseeable future. He estimated that meaningful new competition is unlikely to emerge for at least another three years.\n\nWinbond recently said it will focus on maintaining a stable supply of SLC NAND and NOR Flash products and does not plan to enter the MLC NAND market.\n\nThe company expects SLC NAND bit growth to exceed 80% over the next year. In particular, its new 24nm process accounted for only 1% of Flash-related business in the first quarter of 2026, but the contribution of 24nm SLC NAND is expected to increase from the second quarter onward, gradually becoming a meaningful revenue driver. Winbond also expects SLC NAND gross margins to surpass those of NOR Flash as it pushes toward becoming a leading SLC NAND supplier.\n\nWinbond president Chen Pei-ming said the current Flash shortage appears difficult to resolve and that the market is showing two clear trends.\n\nThe first involves products originally requiring only 16GB or 32GB storage being upgraded toward higher-capacity UFS solutions such as 36GB and 48GB. The second involves customers previously using 4GB eMMC solutions compressing or optimizing code to reduce storage requirements from above 1GB down to around 1GB, enabling the adoption of 8Gb SLC NAND products and creating new business opportunities.\n\n---\n\nhttps://t.co/W6a2xE8OFZ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06542056074766345, "ret_signed_m1d": 0.06542056074766345, "alpha_spy_m1d": 0.07360902915595835, "alpha_c_m1d": 0.11208982234764564, "bench_spy_m1d": -0.008188468408294902, "bench_c_m1d": -0.04666926159998219, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0981308411214954, "ret_signed_p1d": 0.0981308411214954, "alpha_spy_p1d": 0.09585325552606383, "alpha_c_p1d": 0.08088577473576808, "bench_spy_p1d": 0.002277585595431564, "bench_c_p1d": 0.017245066385727315, "ret_p1w": 0.21028037383177578, "ret_signed_p1w": 0.21028037383177578, "alpha_spy_p1w": 0.20817903736688992, "alpha_c_p1w": 0.22828431277248495, "bench_spy_p1w": 0.0021013364648858612, "bench_c_p1w": -0.018003938940709174, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0377, -0.0192, -0.0192, -0.0192, -0.0192, -0.0192, -0.0192, -0.0192, -0.0192, 0.0692, 0.1342, 0.0971, 0.1389, 0.1389, 0.1575, 0.0459, -0.0331, 0.0413, -0.0099, -0.061, -0.0192, 0.0599, 0.0134, 0.0134, 0.0878, 0.1482, 0.19, 0.1342, 0.0227, -0.0006, -0.0749, -0.087, -0.1075, -0.1364, -0.1308, -0.157, -0.1299, -0.1421, -0.1421, -0.1421, -0.1645, -0.1065, -0.1495, -0.1262, -0.1178, -0.1262, -0.157, -0.1626, -0.1794, -0.1991, -0.1486, -0.1533, -0.1794, -0.1757, -0.1224, -0.1075, -0.1336, -0.1607, -0.1607, -0.1084, -0.0766, 0.014, 0.0654, 0.0, 0.0981, 0.1355, 0.1402, 0.2523, 0.2103, 0.2196, 0.0981, 0.0794, 0.0654, 0.1682, 0.2009, 0.3178, 0.4486, 0.3458, 0.4766, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2047504600052498682", "ticker_idx": 0, "ts": "2026-04-24T02:35:12+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Upgrade to Buy, new PT of $1,277", "tweet_type": "news_reaction", "timeline": "weeks", "summary": "Shares/endorses GF upgrade to Buy on SNDK with PT $1,277 ahead of FY3Q26 earnings on April 30.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "quote", "lang": "en", "tweet_text": "So… smartphone NAND prices have already doubled in Q2, and even eSSD prices have surged massively?\n\n…OPM of 71.5%? And FY4Q26 guidance at 84% OPM?", "reply_text": null, "parent_text": null, "quoted_text": "【GF Overseas Electronics &amp; Communications】\nSanDisk FY3Q26 Preview: NAND Is in a Better Position\n\n☄️ SanDisk FY3Q26 Preview: NAND Is in a Better Position\n\n☀️ Upgrade to Buy, new PT of $1,277: SNDK reports FY3Q26 on April 30. Combined with the recent contract price trajectory, we", "parent_author": null, "ret_m1d": -0.05805639996048961, "ret_signed_m1d": -0.05805639996048961, "alpha_spy_m1d": -0.05036669183991749, "alpha_c_m1d": -0.03071895863330454, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08111922899016322, "ret_signed_p1d": 0.08111922899016322, "alpha_spy_p1d": 0.07939642250146117, "alpha_c_p1d": 0.07893469459465097, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": 0.19911099174141755, "ret_signed_p1w": 0.19911099174141755, "alpha_spy_p1w": 0.18971241209927614, "alpha_c_p1w": 0.18881269013167334, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": 0.49377705317414256, "ret_signed_p1m": 0.49377705317414256, "alpha_spy_p1m": 0.44937554590654427, "alpha_c_p1m": 0.3678876634895105, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5137, -0.467, -0.4552, -0.4179, -0.328, -0.2974, -0.4095, -0.4179, -0.3959, -0.4106, -0.4528, -0.3945, -0.3633, -0.367, -0.367, -0.4034, -0.3935, -0.3726, -0.3434, -0.3267, -0.355, -0.3612, -0.3414, -0.3582, -0.3746, -0.4288, -0.3948, -0.4286, -0.4673, -0.4053, -0.3748, -0.3379, -0.3749, -0.3316, -0.2892, -0.2725, -0.2386, -0.22, -0.283, -0.2903, -0.2904, -0.3152, -0.3907, -0.3779, -0.4217, -0.3582, -0.3002, -0.2913, -0.2913, -0.268, -0.2819, -0.2111, -0.1397, -0.1395, -0.0378, -0.0459, -0.0992, -0.0711, -0.0696, -0.0777, -0.0873, -0.0109, -0.0581, 0.0, 0.0811, 0.0126, 0.0751, 0.1077, 0.1991, 0.2687, 0.4207, 0.4244, 0.3536, 0.5783, 0.5633, 0.4668, 0.462, 0.3968, 0.422, 0.3466, 0.3974, 0.4068, 0.558, 0.4938, 0.4938, 0.6058, 0.6062, 0.6584, 0.7123, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2051612202403299609", "ticker_idx": 0, "ts": "2026-05-05T10:37:20+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Buy rating, target price raised to US$465", "tweet_type": "news_reaction", "timeline": "years", "summary": "Shares/adopts GF Overseas Buy on AMD with PT raised to $465, driven by data-center CPU growth, MI455/MI550 ramp, and new Anthropic partnership.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications — AMD (AMD US, Buy): Key Focus Areas and Catalysts Ahead of Earnings Call\n\nBuy rating, target price raised to US$465:\nFollowing the recent share-price rally and Intel’s better-than-expected results, market expectations have already moved higher. However, we remain optimistic and outline the key issues to watch during and after the earnings call. Incorporating our updated pre-earnings preview and higher CPU/GPU shipment assumptions, we raise our 2026E/2027E EPS forecasts by 2%/6%, respectively. We lift our target price from US$311 to US$465, based on 38x 2027E P/E.\n\nStrong CPU business and capacity:\nWe expect AMD’s data-center CPU business to grow 55%/48% YoY in 2026E/2027E, driven by:\n\n1. CPU price increases in April / 4Q26;\n2. ramp-up of the Venice platform in 3Q26;\n3. market-share gains from product leadership and customer expansion.\n\nIn the near term, driven by cloud demand, we expect ODM server shipments to grow around 10% QoQ in 2Q. On the demand side, agentic AI is expected to drive the CPU-to-GPU ratio back toward a more balanced level, while AMD’s competitive advantage continues to expand with its N2-based Venice platform. Based on our ASE report published on April 29, we believe AMD has secured 160k–180k wafers of CoWoS capacity for 2027.\n\nNew partnership with Anthropic:\nAccording to Wccftech’s April 17 report, supply constraints prompted Anthropic to sign an agreement with AMD to use MI450 chips. This could become a key focus of the upcoming earnings call. Unlike AMD’s five-year, 6GW partnerships with Meta/OpenAI in exchange for AMD warrants, we expect this collaboration to be more flexible in both GW scale and transaction structure. However, given Anthropic’s rapidly growing ARR, the first-year scale of the partnership could be larger.\n\nMI455 progress:\nFollowing our April 14 report, Demand Far Exceeds Supply, in which we noted a downward revision to MI455 shipments due to the complexity of its stacked design, we believe the supply chain is still maintaining its target of starting shipments in mid-3Q. Meanwhile, downstream readiness has improved with support from Wistron and Foxconn Industrial Internet. With multiple GW-scale customers including OpenAI, Meta, and Anthropic, we expect strong MI455 demand to continue into 2027/2028.\n\nMI550 versus Rubin Ultra:\nAs stated in our March 31 report, we believe Rubin Ultra has been changed to a dual-chip design, with a second MCM version used to connect two sets of dual chips, while 16-high HBM4e may be revised down to 12-high due to capacity constraints. For AMD, given its experience in 2.5D/3D packaging, we believe MI550 remains on track, using a four-chiplet / 8-XCD package with 12-high HBM4e. Combined with the 2nm process node, MI550, as a single GPU, could deliver superior performance.\n\n$AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03861960491909744, "ret_signed_m1d": -0.03861960491909744, "alpha_spy_m1d": -0.030661262337630113, "alpha_c_m1d": -0.008200056387046128, "bench_spy_m1d": -0.007958342581467326, "bench_c_m1d": -0.03041954853205131, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.1861453669565576, "ret_signed_p1d": 0.1861453669565576, "alpha_spy_p1d": 0.17224592751154, "alpha_c_p1d": 0.13435557339291782, "bench_spy_p1d": 0.013899439445017592, "bench_c_p1d": 0.051789793563639774, "ret_p1w": 0.2618645392727186, "ret_signed_p1w": 0.2618645392727186, "alpha_spy_p1w": 0.24195493711226446, "alpha_c_p1w": 0.18809232748660087, "bench_spy_p1w": 0.01990960216045412, "bench_c_p1w": 0.0737722117861177, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4581, -0.4133, -0.392, -0.3988, -0.3988, -0.4203, -0.4164, -0.4164, -0.4284, -0.4367, -0.4275, -0.4366, -0.4466, -0.3981, -0.4065, -0.4267, -0.4364, -0.4409, -0.4625, -0.4312, -0.4386, -0.4583, -0.4295, -0.4279, -0.4234, -0.4434, -0.4556, -0.4467, -0.4474, -0.4386, -0.4222, -0.4333, -0.4295, -0.4219, -0.38, -0.4264, -0.4314, -0.4482, -0.4274, -0.4083, -0.3878, -0.3878, -0.3802, -0.3764, -0.3475, -0.3339, -0.3103, -0.3052, -0.282, -0.2734, -0.2167, -0.2164, -0.2261, -0.1992, -0.1458, -0.1405, -0.021, -0.0581, -0.0902, -0.0511, -0.0022, 0.0149, -0.0386, 0.0, 0.1861, 0.1497, 0.2813, 0.2914, 0.2619, 0.254, 0.2658, 0.1938, 0.185, 0.1655, 0.2599, 0.2655, 0.316, 0.316, 0.4184, 0.3949, 0.4583, 0.4527, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2059967125020934181", "ticker_idx": 0, "ts": "2026-05-28T11:56:49+00:00", "symbol": "009150.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SEMCO is the only company in the world that operates both a silicon capacitor business and a substrate business... if a company can deliver both the substrate and the silicon capacitor to Intel's EMIB on a turnkey basis, that maximizes SEMCO's bargaining power and value. This is precisely why SEMCO's stock has been surging recently", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long SEMCO as unique turnkey supplier of substrate + silicon capacitor for Intel EMIB with $1B+ contract; also long Murata on its Ibiden/EMIB relationship.", "resolved_tickers": ["009150.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Why Are Silicon Capacitors Becoming Important? And Why Is SEMCO's Stock Surging?\n\nThe reason silicon capacitors have failed to grow into a large market until now is clear. TSMC, which holds an overwhelming share in substrates for AI packaging, has historically sourced silicon capacitors in-house. As a result, it has been effectively impossible for silicon capacitor suppliers outside of TSMC to break into the AI packaging value chain.\n\nBut with the rise of Intel's EMIB, the situation is changing.\n\nSilicon capacitor suppliers like Murata and SEMCO are now supplying silicon capacitors for Intel's EMIB, and have begun to benefit directly from AI demand.\n\nThe existing silicon capacitor market was effectively an oligopoly dominated by Murata and TSMC. SEMCO was closer to a latecomer.\n\nYet even SEMCO—the latecomer—has managed to land a supply contract worth more than $1 billion. For context, SEMCO's silicon capacitor revenue last year was a mere few million dollars.\n\nSo how did SEMCO ramp its revenue this quickly? It comes down to the fact that SEMCO is the only company in the world that operates both a silicon capacitor business and a substrate business.\n\nThe reason silicon capacitors have become important for Intel's EMIB is that Intel cannot make its own silicon capacitors and must source them externally. On top of that, within Intel's EMIB BoM, the highest value-added component is the substrate.\n\nSo if a company can deliver both the substrate and the silicon capacitor to Intel's EMIB on a turnkey basis, that maximizes SEMCO's bargaining power and value. This is precisely why SEMCO's stock has been surging recently.\n\nOf course, this isn't to say Murata will benefit less. Murata still maintains a very close relationship with Ibiden—Intel's first-vendor for substrates—and is clearly positioned to benefit from EMIB. That said, my view is that the emergence of a new player like SEMCO should by no means be underestimated.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.11844240129799888, "ret_signed_m1d": -0.11844240129799888, "alpha_spy_m1d": -0.1129561122081183, "alpha_c_m1d": -0.1054907644305445, "bench_spy_m1d": -0.0054862890898805805, "bench_c_m1d": -0.012951636867454375, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.15035154137371554, "ret_signed_p1d": 0.15035154137371554, "alpha_spy_p1d": 0.14786014865752484, "alpha_c_p1d": 0.1280341850304938, "bench_spy_p1d": 0.002491392716190699, "bench_c_p1d": 0.02231735634322174, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.7574, -0.7788, -0.8039, -0.7829, -0.7812, -0.7934, -0.7826, -0.7804, -0.7839, -0.7834, -0.7766, -0.7585, -0.7491, -0.7407, -0.7488, -0.7688, -0.7637, -0.7542, -0.7528, -0.7528, -0.768, -0.7796, -0.7599, -0.7745, -0.7534, -0.7501, -0.7528, -0.722, -0.7209, -0.6944, -0.6928, -0.6842, -0.6679, -0.6544, -0.6328, -0.6322, -0.5825, -0.5608, -0.5814, -0.5738, -0.5706, -0.5462, -0.5527, -0.55, -0.55, -0.5035, -0.5035, -0.5068, -0.5041, -0.5057, -0.5133, -0.4819, -0.4435, -0.4462, -0.4538, -0.4424, -0.4662, -0.4262, -0.3488, -0.2753, -0.2753, -0.1498, -0.1184, 0.0, 0.1504, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2059967125020934181", "ticker_idx": 1, "ts": "2026-05-28T11:56:49+00:00", "symbol": "Murata", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Murata still maintains a very close relationship with Ibiden—Intel's first-vendor for substrates—and is clearly positioned to benefit from EMIB", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long SEMCO as unique turnkey supplier of substrate + silicon capacitor for Intel EMIB with $1B+ contract; also long Murata on its Ibiden/EMIB relationship.", "resolved_tickers": ["6981.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Murata Manufacturing Co., listed on Tokyo Stock Exchange", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Why Are Silicon Capacitors Becoming Important? And Why Is SEMCO's Stock Surging?\n\nThe reason silicon capacitors have failed to grow into a large market until now is clear. TSMC, which holds an overwhelming share in substrates for AI packaging, has historically sourced silicon capacitors in-house. As a result, it has been effectively impossible for silicon capacitor suppliers outside of TSMC to break into the AI packaging value chain.\n\nBut with the rise of Intel's EMIB, the situation is changing.\n\nSilicon capacitor suppliers like Murata and SEMCO are now supplying silicon capacitors for Intel's EMIB, and have begun to benefit directly from AI demand.\n\nThe existing silicon capacitor market was effectively an oligopoly dominated by Murata and TSMC. SEMCO was closer to a latecomer.\n\nYet even SEMCO—the latecomer—has managed to land a supply contract worth more than $1 billion. For context, SEMCO's silicon capacitor revenue last year was a mere few million dollars.\n\nSo how did SEMCO ramp its revenue this quickly? It comes down to the fact that SEMCO is the only company in the world that operates both a silicon capacitor business and a substrate business.\n\nThe reason silicon capacitors have become important for Intel's EMIB is that Intel cannot make its own silicon capacitors and must source them externally. On top of that, within Intel's EMIB BoM, the highest value-added component is the substrate.\n\nSo if a company can deliver both the substrate and the silicon capacitor to Intel's EMIB on a turnkey basis, that maximizes SEMCO's bargaining power and value. This is precisely why SEMCO's stock has been surging recently.\n\nOf course, this isn't to say Murata will benefit less. Murata still maintains a very close relationship with Ibiden—Intel's first-vendor for substrates—and is clearly positioned to benefit from EMIB. That said, my view is that the emergence of a new player like SEMCO should by no means be underestimated.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08409463574607634, "ret_signed_m1d": -0.08409463574607634, "alpha_spy_m1d": -0.07860834665619576, "alpha_c_m1d": -0.07114299887862197, "bench_spy_m1d": -0.0054862890898805805, "bench_c_m1d": -0.012951636867454375, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.1273131881002576, "ret_signed_p1d": 0.1273131881002576, "alpha_spy_p1d": 0.1248217953840669, "alpha_c_p1d": 0.10499583175703586, "bench_spy_p1d": 0.002491392716190699, "bench_c_p1d": 0.02231735634322174, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5283, -0.5672, -0.5757, -0.5582, -0.5611, -0.597, -0.5739, -0.564, -0.5684, -0.5765, -0.5815, -0.5589, -0.554, -0.5651, -0.5651, -0.593, -0.5889, -0.5753, -0.568, -0.5655, -0.5861, -0.6007, -0.57, -0.5926, -0.5562, -0.5541, -0.5566, -0.5315, -0.5327, -0.5144, -0.5116, -0.493, -0.4706, -0.4387, -0.4612, -0.4451, -0.4366, -0.4248, -0.4363, -0.4219, -0.4242, -0.4331, -0.4331, -0.3961, -0.3982, -0.3982, -0.3982, -0.3982, -0.3385, -0.3148, -0.3019, -0.2881, -0.2868, -0.2576, -0.2796, -0.2772, -0.2762, -0.2781, -0.2121, -0.1649, -0.0714, -0.054, -0.0841, 0.0, 0.1273, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047338240613769426", "ticker_idx": 0, "ts": "2026-04-23T15:34:08+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think $MSFT is in a stronger position than Google when it comes to coding", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish MSFT: stronger coding AI position than Google via GitHub Copilot.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "Why does it feel like people talk so little about GitHub Copilot? I think $MSFT is in a stronger position than Google when it comes to coding.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04129888707785789, "ret_signed_m1d": 0.04129888707785789, "alpha_spy_m1d": 0.03740304443819631, "alpha_c_m1d": 0.03740304443819631, "bench_spy_m1d": 0.003895842639661584, "bench_c_m1d": 0.003895842639661584, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.021334937195510273, "ret_signed_p1d": 0.021334937195510273, "alpha_spy_p1d": 0.013585639235486058, "alpha_c_p1d": 0.013585639235486058, "bench_spy_p1d": 0.007749297960024215, "bench_c_p1d": 0.007749297960024215, "ret_p1w": -0.019170135207435268, "ret_signed_p1w": -0.019170135207435268, "alpha_spy_p1w": -0.03358182376919128, "alpha_c_p1w": -0.03358182376919128, "bench_spy_p1w": 0.014411688561756009, "bench_c_p1w": 0.014411688561756009, "ret_p1m": 0.00896353668088068, "ret_signed_p1m": 0.00896353668088068, "alpha_spy_p1m": -0.04353134905643263, "alpha_c_p1m": -0.04353134905643263, "bench_spy_p1m": 0.05249488573731331, "bench_c_p1m": 0.05249488573731331, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1286, 0.1533, 0.1558, 0.0403, 0.0326, 0.016, -0.0132, -0.006, -0.0553, -0.0373, -0.0074, -0.0082, -0.0296, -0.0357, -0.0369, -0.0369, -0.0476, -0.041, -0.0416, -0.0445, -0.0752, -0.0643, -0.0364, -0.0337, -0.0553, -0.0414, -0.0284, -0.0254, -0.0122, -0.0163, -0.0152, -0.024, -0.0261, -0.0334, -0.0486, -0.038, -0.0393, -0.0576, -0.0643, -0.0815, -0.0788, -0.1035, -0.1075, -0.1197, -0.1419, -0.1366, -0.1096, -0.1116, -0.1017, -0.1017, -0.1031, -0.1045, -0.0996, -0.1027, -0.1079, -0.0755, -0.0545, -0.0109, 0.0108, 0.0169, 0.0056, 0.0202, 0.0413, 0.0, 0.0213, 0.0218, 0.0325, 0.021, -0.0192, -0.0032, -0.0051, -0.0105, -0.0043, 0.0121, -0.0015, -0.0074, -0.0192, -0.0254, -0.0152, 0.0148, 0.0187, 0.004, 0.0128, 0.0102, 0.009, 0.009, 0.0028, -0.0053, 0.0293, 0.0853, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2052914079619297612", "ticker_idx": 0, "ts": "2026-05-09T00:50:32+00:00", "symbol": "ACMR", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$ACMR\n\nAs you know, China is going to keep pushing for localization.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Long ACMR as a beneficiary of China's semiconductor localization push.", "resolved_tickers": ["ACMR"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@EquityBrian $ACMR \n\nAs you know, China is going to keep pushing for localization.", "reply_text": null, "parent_text": "What’s your best under-$5B market cap idea right now?\n\nBonus points if you explain the bull case.", "quoted_text": null, "parent_author": "EquityBrian", "ret_m1d": -0.07453223799747677, "ret_signed_m1d": -0.07453223799747677, "alpha_spy_m1d": -0.07225982801029884, "alpha_c_m1d": -0.05757952231904251, "bench_spy_m1d": -0.0022724099871779257, "bench_c_m1d": -0.01695271567843426, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02597798883633262, "ret_signed_p1d": -0.02597798883633262, "alpha_spy_p1d": -0.024463048844880708, "alpha_c_p1d": 0.00015377639833624723, "bench_spy_p1d": -0.0015149399914519135, "bench_c_p1d": -0.02613176523466887, "ret_p1w": -0.021957603357830258, "ret_signed_p1w": -0.021957603357830258, "alpha_spy_p1w": -0.02107844281456206, "alpha_c_p1w": 0.030358033264449857, "bench_spy_p1w": -0.0008791605432681981, "bench_c_p1w": -0.052315636622280115, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0929, -0.0088, 0.0026, 0.0026, 0.0079, -0.0009, 0.0105, 0.0297, 0.007, 0.0478, 0.0588, -0.118, -0.139, -0.1169, -0.203, -0.1987, -0.253, -0.3054, -0.278, -0.2568, -0.2596, -0.3, -0.2828, -0.2745, -0.268, -0.2794, -0.2794, -0.3266, -0.3068, -0.2899, -0.2943, -0.3632, -0.3804, -0.4135, -0.3915, -0.3744, -0.3731, -0.3731, -0.3557, -0.3519, -0.2865, -0.2737, -0.2567, -0.2327, -0.2152, -0.2301, -0.2595, -0.2174, -0.2316, -0.2081, -0.1568, -0.1651, -0.1325, -0.1769, -0.2341, -0.2477, -0.2007, -0.1936, -0.2078, -0.1835, -0.1437, -0.0846, -0.0745, 0.0, -0.026, 0.0012, -0.0229, -0.0155, -0.022, 0.0275, 0.1072, 0.107, 0.1342, 0.1342, 0.3369, 0.3707, 0.4359, 0.3385, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047861732702662741", "ticker_idx": 0, "ts": "2026-04-25T02:14:19+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's moat is its ability to anticipate the trajectory of mainstream LLM technology and the new demands on accelerators 3–5 years out, plan and position accordingly, and bake that foresight into product design. Other GPU companies don't anticipate demand — they react to it.", "tweet_type": "analysis", "timeline": "years", "summary": "Long NVDA: deep structural moat in anticipating LLM compute demands 3-5 years ahead; Blackwell/Rubin/Feynman roadmap precisely aligned with next-gen model architectures like DeepSeek V4.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Thoughts after reading the DeepSeek V4 paper:\n\n- NVIDIA really is something else. Remember how back in 2024 people were bashing Blackwell as overspec'd and dismissing FP4 as just marketing? Turns out it was all groundwork for the next generation of models. Maybe NVIDIA's moat is its ability to anticipate the trajectory of mainstream LLM technology and the new demands on accelerators 3–5 years out, plan and position accordingly, and bake that foresight into product design. Other GPU companies don't anticipate demand — they react to it.\n\n- Have NVIDIA and DeepSeek been talking to each other? Looking at the 6144 FLOPs/Byte passage — I'd been wondering why NVIDIA was pushing HBM4 pin speeds up so aggressively, and it turns out raising Rubin's HBM4 pin speed isn't \"overkill\" from the perspective of a model like V4. It's a precisely balanced design.\n\n- NVIDIA is once again working hard to crank up bandwidth on Rubin Ultra, which is telling: it implies that Rubin Ultra's FP4 compute is getting too fast relative to HBM bandwidth, and that HBM bandwidth could once again become the bottleneck when training MoE models like DeepSeek-V4.\n\n- Why is the next-gen NVIDIA chip scaling up the NVL domain? Why are they moving toward Kyber? You can read this as an attempt to push up the bandwidth-density of the interconnect fabric, pulling compute — which has surged above the threshold — back into a communication-friendly balance.\n\n- The upshot is that the DeepSeek paper is essentially telling us NVIDIA's current chip design lines up with DeepSeek's model patterns. If Blackwell alone produced this much evolution, imagine what Rubin and Feynman will deliver.\n\n- NVIDIA's G3.5, unveiled this year, is almost uncanny. It's a new tier sitting between local NVMe SSDs and object/file storage — one that exists only for AI inference — meaning NVIDIA has created an entirely new memory-tier category exclusively for AI workloads. And in §3.6.2 of the V4 paper, DeepSeek argues that the KV cache can break out of GPU HBM's limits and be permanently offloaded to NVMe storage. That maps exactly onto the ICMS rack NVIDIA showed at CES. NVIDIA called it precisely — they saw that labs like DeepSeek would need this.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.038502360529941604, "ret_signed_m1d": -0.038502360529941604, "alpha_spy_m1d": -0.0367825169988345, "alpha_c_m1d": -0.038857894588151365, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01588108781548958, "ret_signed_p1d": -0.01588108781548958, "alpha_spy_p1d": -0.011015139087343417, "alpha_c_p1d": 0.013846755625105955, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": -0.08369883584195925, "ret_signed_p1w": -0.08369883584195925, "alpha_spy_p1w": -0.08766995728627958, "alpha_c_p1w": -0.08474572631142818, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": -0.00807903603281912, "ret_signed_p1m": -0.00807903603281912, "alpha_spy_p1m": -0.05760578462646393, "alpha_c_p1m": -0.1974678936026686, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1159, -0.1113, -0.1177, -0.1432, -0.1675, -0.1959, -0.2065, -0.1441, -0.1227, -0.1296, -0.1227, -0.137, -0.1561, -0.1561, -0.1461, -0.1322, -0.1326, -0.1237, -0.1157, -0.1097, -0.0972, -0.1465, -0.182, -0.1576, -0.1688, -0.155, -0.1536, -0.1791, -0.1568, -0.147, -0.1412, -0.1545, -0.1679, -0.1541, -0.1601, -0.1672, -0.1757, -0.2027, -0.1891, -0.1912, -0.1751, -0.2095, -0.2266, -0.2375, -0.1949, -0.1886, -0.1811, -0.1811, -0.1799, -0.1778, -0.1594, -0.151, -0.1292, -0.126, -0.0928, -0.0819, -0.0843, -0.0689, -0.0672, -0.0772, -0.0651, -0.0783, -0.0385, 0.0, -0.0159, -0.034, -0.0787, -0.0838, -0.0837, -0.0928, -0.0405, -0.0236, -0.0065, 0.0131, 0.0193, 0.0426, 0.0883, 0.0402, 0.0264, 0.0185, 0.0317, 0.0134, -0.0059, -0.0059, -0.0081, -0.0185, -0.0109, -0.0253, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2051088714357715392", "ticker_idx": 0, "ts": "2026-05-03T23:57:11+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Goldman Sachs raises its Samsung Electronics operating profit forecasts", "tweet_type": "news_reaction", "timeline": "years", "summary": "Goldman Sachs raises Samsung Electronics operating profit forecasts materially for 2026–2028; author shares/adopts bullish stance.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "in", "tweet_text": "Goldman Sachs raises its Samsung Electronics operating profit forecasts:\n\n2026: KRW 315tn ($213.7B) → KRW 355tn ($240.9B)\n\n2027: KRW 307tn ($208.3B) → KRW 438tn ($297.2B)\n\n2028: KRW 318tn ($215.8B) → KRW 495tn ($335.9B)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05161290322580647, "ret_signed_m1d": -0.05161290322580647, "alpha_spy_m1d": -0.05528975259076063, "alpha_c_m1d": -0.0505020860016141, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": -0.001110817224192373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.022091957813239027, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.022091957813239027, "ret_p1w": 0.2279569892473119, "ret_signed_p1w": 0.2279569892473119, "alpha_spy_p1w": 0.19830562263991136, "alpha_c_p1w": 0.1302710803738718, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.09768590887344009, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2742, -0.3163, -0.3193, -0.2858, -0.2884, -0.2798, -0.2334, -0.2223, -0.2223, -0.2223, -0.2223, -0.1845, -0.1841, -0.1716, -0.1416, -0.1265, -0.0643, -0.0707, -0.0707, -0.1626, -0.2609, -0.1776, -0.1922, -0.2553, -0.1935, -0.1845, -0.1935, -0.2124, -0.1901, -0.1677, -0.1051, -0.1394, -0.1441, -0.2004, -0.1858, -0.1888, -0.227, -0.227, -0.2417, -0.2809, -0.1843, -0.2327, -0.1991, -0.1695, -0.1548, -0.0946, -0.1226, -0.114, -0.1355, -0.1118, -0.0925, -0.0645, -0.071, -0.0774, -0.0581, -0.0645, -0.0344, -0.0559, -0.0344, -0.0452, -0.028, -0.0516, -0.0516, 0.0, 0.0, 0.1441, 0.1677, 0.1548, 0.228, 0.2, 0.2215, 0.2731, 0.1634, 0.2086, 0.1849, 0.1871, 0.2882, 0.2581, 0.2581, 0.286, 0.3204, 0.2882, 0.3634, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2045424807614038306", "ticker_idx": 0, "ts": "2026-04-18T08:50:50+00:00", "symbol": "UMC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura forecasts UMC's 2Q revenue to grow 7–9% QoQ, significantly exceeding the street consensus of 2%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares Nomura's bullish thesis: UMC and Vanguard raising chip prices, utilization climbing, with 2Q revenue guidance well above street consensus.", "resolved_tickers": ["UMC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Taiwan UMC, Vanguard Raise Chip Prices — Impact of \"De-China\" Supply Chain Restructuring\n\nAs the decoupling between China and non-China semiconductor supply chains peaks, Taiwan's tier-2 foundries UMC and Vanguard (VIS) have gained the leverage to raise chip prices, according to a recent assessment. Years of supply chain restructuring have eroded the cost-advantage argument for shifting production elsewhere, and TSMC's concentration on leading-edge nodes has further benefited these tier-2 players.\n\nIn its April 15 report titled \"Taiwan Tier-2 Foundries: Price Hikes Materializing,\" Nomura Securities noted that UMC and Vanguard plan to raise chip prices and that fab utilization is also climbing.\n\nOne Taiwanese media outlet reported that Vanguard plans to raise foundry prices across all platforms by 10% starting in April. UMC has notified customers that it will adjust foundry pricing in the second half. In both cases, the drivers are pass-through of rising input costs and tight fab utilization.\n\nNomura cited \"the peak-out of supply chain decoupling effects\" as one of the key rationales behind the price hikes at both companies. Over the past 2–3 years, customers looking to migrate volume to Chinese foundries have effectively completed that shift, reducing the customer-attrition risk for UMC and Vanguard.\n\nAs geopolitical risk has elevated the importance of supply chain diversification, flexible production migration has diminished, and the incentive offered by price cuts is no longer what it once was. Taiwan foundries now have less reason to engage in the chicken game being waged by Chinese competitors.\n\nThe shift in TSMC's strategy also matters. As TSMC concentrates its resources on preparing next-generation fabs set to come online from 2028 and on optimizing leading-edge nodes, mature nodes have been deprioritized.\n\nNomura pointed to Vanguard's expanding foundry business with Renesas for AI server PMICs and its growing share within Qualcomm as supporting the price hike. Products on ≤180nm nodes are now shipping as soon as they come off the line. Vanguard's 1Q fab utilization is estimated to have risen to 80–85%.\n\nUMC's 1Q utilization is estimated at roughly 70% for 8-inch and 80% for 12-inch fabs. Both are expected to break above 80% from 2Q onward.\n\nNomura noted that demand for analog and mixed-signal chips from IDMs — which account for roughly 20% of UMC's revenue — is driving performance. Strong Apple device sales this year are also contributing to higher UMC fab utilization. Representative examples include OLED DDIs designed by Novatek and Samsung System LSI for Apple products, as well as ISPs from Samsung System LSI. These products are primarily manufactured on 22/28nm nodes.\n\nNomura forecasts UMC's 2Q revenue to grow 7–9% QoQ, significantly exceeding the street consensus of 2%. Vanguard's 2Q revenue is expected to grow more than 15% QoQ as the price hikes fully flow through — also well above the consensus of 6%.\n\n---\n\nhttps://t.co/dF5Praw6WN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07753168269717192, "ret_signed_m1d": -0.07753168269717192, "alpha_spy_m1d": -0.07953535693243263, "alpha_c_m1d": -0.07796278065974405, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": 0.0004310979625721245, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.024525348987071927, "ret_signed_p1d": -0.024525348987071927, "alpha_spy_p1d": -0.017978399832135672, "alpha_c_p1d": -0.02603412607976019, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0015087770926882627, "ret_p1w": -0.0751582861658151, "ret_signed_p1w": -0.0751582861658151, "alpha_spy_p1w": -0.08425921809049641, "alpha_c_p1w": -0.1663299800394289, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.09117169387361379, "ret_p1m": 0.390822740997228, "ret_signed_p1m": 0.390822740997228, "alpha_spy_p1m": 0.3555337540919876, "alpha_c_p1m": 0.21839401440233175, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.17242872659489628, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1195, -0.1377, -0.1661, -0.1036, -0.0127, -0.0997, -0.163, -0.1946, -0.2057, -0.2207, -0.2128, -0.2112, -0.2041, -0.2049, -0.216, -0.1851, -0.1859, -0.1756, -0.1756, -0.1875, -0.1907, -0.1986, -0.1772, -0.1756, -0.1361, -0.1653, -0.1851, -0.1741, -0.1685, -0.2168, -0.2184, -0.2286, -0.2405, -0.2302, -0.25, -0.2326, -0.2603, -0.2666, -0.2508, -0.2468, -0.2642, -0.2674, -0.2785, -0.2809, -0.284, -0.2634, -0.2864, -0.2983, -0.3188, -0.2896, -0.2903, -0.3141, -0.3141, -0.3125, -0.2634, -0.2508, -0.246, -0.2294, -0.2326, -0.2413, -0.2128, -0.1598, -0.0775, 0.0, -0.0245, 0.0055, -0.0506, -0.0285, -0.0752, -0.0775, 0.0087, 0.0332, 0.0324, 0.0269, 0.1084, 0.2049, 0.1994, 0.2199, 0.2334, 0.2706, 0.2595, 0.3552, 0.3608, 0.3544, 0.3908, 0.3892, 0.4517, 0.4415, 0.4415, 0.6677, 0.7579, 0.7943, 0.7547, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2060140629649486165", "ticker_idx": 0, "ts": "2026-05-28T23:26:16+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung's foundry business, which has run at a loss for several years, has recently landed a string of major global customers, fueling growing expectations for a return to profitability next year", "tweet_type": "analysis", "timeline": "quarters", "summary": "Samsung foundry revival thesis: Anthropic chip order, Tesla AI5/AI6, NVIDIA LPU production signal return to foundry profitability next year.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics Eyed for Anthropic Chip Order… Foundry Revival Begins\n\nSamsung Electronics has made an investment in Anthropic, the U.S. AI company behind the AI model \"Claude,\" raising the possibility of a foundry partnership in which Samsung would manufacture Anthropic's chips on a contract basis.\n\nSamsung's foundry business, which has run at a loss for several years, has recently landed a string of major global customers, fueling growing expectations for a return to profitability next year.\n\nOn the 28th (local time), Anthropic announced that it had raised $65 billion in its recent Series H funding round, with its post-money valuation assessed at $965 billion (about 1,440 trillion won).\n\nThe three major memory semiconductor manufacturers—Samsung Electronics, SK Hynix, and Micron—participated in this funding round as \"strategic infrastructure partners.\"\n\nAnthropic emphasized that \"the technologies of these companies play a critical role in the global supply of memory, storage, and logic chips,\" adding that \"partnerships with them will be a great help in reliably scaling computing capacity to meet customer needs.\"\n\nAnthropic's mention of \"logic chips\" in its announcement has prompted speculation about a foundry partnership with Samsung Electronics. The process of manufacturing logic chips is foundry work, and aside from Samsung Electronics, neither Micron nor SK Hynix has a foundry division. For this reason, the industry analyzes that Samsung Electronics and Anthropic will newly collaborate not just at the memory level but also in foundry.\n\nThis implies a high likelihood that Samsung Electronics will produce the AI chips used in Anthropic's flagship model \"Claude.\" Anthropic, alongside OpenAI, is regarded as one of the world's leading AI model companies and a core AI player. The prevailing view is that Samsung's foundry has secured yet another major customer, following Tesla, NVIDIA, and others.\n\nEarlier, Samsung Electronics won orders for Tesla's next-generation AI chips, \"AI5\" and \"AI6.\" Samsung's foundry is also handling production of an upgraded version of \"AI4.\" NVIDIA's inference-dedicated language processing unit (LPU) chip \"Grok3\" is likewise being produced through Samsung's foundry. Samsung Electronics also plans to supply the image sensor to be mounted in Apple's new iPhone next year.\n\nAccordingly, attention is focused on whether Samsung Electronics' foundry business—which has run at a loss in recent years—can regain its vitality. As of last year, Samsung Electronics held second place (7.2%) in the global foundry market, but the gap with first-place TSMC (69.9%) reaches 62.7 percentage points.\n\nRegarding Samsung's investment in Anthropic, the industry views it as strategic cooperation spanning \"AI semiconductors, memory, foundry, and AI infrastructure.\" Samsung Electronics is positioning a \"one-stop solution\"—encompassing high-bandwidth memory (HBM), advanced foundry processes, and advanced packaging—as its differentiation point. As the AI market evolves into an ecosystem centered not only on HBM supply but also AI chip design, production, packaging, and system optimization, the move is interpreted as Samsung strengthening partnerships by leveraging its competitiveness across the semiconductor spectrum.\n\nA semiconductor industry official said, \"Samsung's investment is not a simple equity stake in an AI company, but is interpreted as a signal that Samsung is earnestly expanding strategic relationships with the key players of the AI era,\" adding, \"Expectations are growing as to whether Samsung's foundry business can seize another opportunity amid the expansion of the AI market.\"\n\n---\n\nhttps://t.co/rVUzI49t6v", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.025041736227044975, "ret_signed_m1d": 0.025041736227044975, "alpha_spy_m1d": 0.030528025316925556, "alpha_c_m1d": 0.03799337309449935, "bench_spy_m1d": -0.0054862890898805805, "bench_c_m1d": -0.012951636867454375, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.058430717863105164, "ret_signed_p1d": 0.058430717863105164, "alpha_spy_p1d": 0.055939325146914465, "alpha_c_p1d": 0.036113361519883425, "bench_spy_p1d": 0.002491392716190699, "bench_c_p1d": 0.02231735634322174, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2786, -0.3499, -0.4262, -0.3616, -0.3729, -0.4219, -0.3739, -0.3669, -0.3739, -0.3886, -0.3713, -0.3539, -0.3053, -0.3319, -0.3356, -0.3792, -0.3679, -0.3703, -0.3999, -0.3999, -0.4114, -0.4417, -0.3668, -0.4043, -0.3783, -0.3553, -0.3439, -0.2972, -0.3189, -0.3122, -0.3289, -0.3105, -0.2955, -0.2738, -0.2788, -0.2838, -0.2688, -0.2738, -0.2504, -0.2671, -0.2504, -0.2588, -0.2454, -0.2638, -0.2638, -0.2237, -0.2237, -0.1119, -0.0935, -0.1035, -0.0467, -0.0684, -0.0518, -0.0117, -0.0968, -0.0618, -0.0801, -0.0785, 0.0, -0.0234, -0.0234, -0.0017, 0.025, 0.0, 0.0584, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2043170566996480102", "ticker_idx": 0, "ts": "2026-04-12T03:33:18+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ASML株持ってるから、ラピダスがEUVをガッツリ買ってくれるといいな〜", "tweet_type": "position_disclosure", "timeline": "unspecified", "summary": "Holds ASML; hopes Rapidus buys EUV heavily, implying continued bullish position.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "ja", "tweet_text": "@dottomato ASML株持ってるから、ラピダスがEUVをガッツリ買ってくれるといいな〜 って思ってるよ（笑）", "reply_text": null, "parent_text": "@jukan05 日本人として応援しているし、チャレンジはするべきだ\n\nただ投資するかと言われれば…🤔", "quoted_text": null, "parent_author": "dottomato", "ret_m1d": 0.008096420951350103, "ret_signed_m1d": 0.008096420951350103, "alpha_spy_m1d": 0.017774243206934104, "alpha_c_m1d": 0.022667611253239017, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.014571190301888914, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01936807460209078, "ret_signed_p1d": 0.01936807460209078, "alpha_spy_p1d": 0.007183193849720659, "alpha_c_p1d": -0.00016547466854421877, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.019533549270634998, "ret_p1w": -0.011747982255521316, "ret_signed_p1w": -0.011747982255521316, "alpha_spy_p1w": -0.04471693127475107, "alpha_c_p1w": -0.058258550414302945, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.04651056815878163, "ret_p1m": 0.006984307764408726, "ret_signed_p1m": 0.006984307764408726, "alpha_spy_p1m": -0.06892302198788691, "alpha_c_p1m": -0.25897410933325493, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.26595841709766366, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1405, -0.0889, -0.0747, -0.1118, -0.0963, -0.0846, -0.0675, -0.066, -0.0662, -0.0348, -0.0532, -0.0551, -0.0364, -0.0291, -0.0564, -0.0958, -0.0887, -0.0537, -0.0437, -0.053, -0.0413, -0.0635, -0.0551, -0.0514, -0.0481, -0.0119, -0.0171, -0.0033, -0.0084, 0.0029, 0.0227, -0.0217, -0.021, -0.0392, -0.0778, -0.0476, -0.0586, -0.0895, -0.0891, -0.0475, -0.0484, -0.063, -0.0637, -0.0495, -0.0524, -0.0546, -0.0724, -0.1045, -0.067, -0.0462, -0.0383, -0.0746, -0.09, -0.1173, -0.1116, -0.0573, -0.0784, -0.0784, -0.0784, -0.1159, -0.0375, -0.0198, 0.0081, 0.0, 0.0194, -0.0237, -0.0295, -0.0125, -0.0117, -0.0181, -0.0079, -0.0286, -0.0059, -0.0353, -0.0678, -0.0512, -0.0275, -0.0275, -0.0554, -0.0223, 0.0382, 0.0342, 0.0496, 0.0386, 0.007, 0.0557, 0.0875, 0.0394, 0.0059, -0.0064, 0.0604, 0.0701, 0.1209, 0.1381, 0.1056, 0.095, 0.1066, 0.1017, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2044914319243280555", "ticker_idx": 0, "ts": "2026-04-16T23:02:21+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Positive for BESI. More hybrid bonding tool orders should follow.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long BESI: TSMC's SoIC capacity expansion to 50K wafers/month in 2027 drives more hybrid bonding tool orders.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "DigiTimes published a report on TSMC's packaging this morning, and there's a lot of alpha in it.\n\nTo summarize:\n\n1. TSMC is significantly expanding SoIC capacity in 2027 (from 10K wafers/month to 50K wafers/month), with NVIDIA locking up most of this SoIC capacity. Only 10% will be allocated to CPO. \n\n-> Positive for BESI. More hybrid bonding tool orders should follow.\n\n2. CoPoS is delayed… first product shipment expected in Q4 2030. \nCoWoS's lifespan is now expected to be longer than previously anticipated. CoPoS is just too difficult to push through. According to DigiTimes, with large orders pouring in not only from NVIDIA and AMD but also from ASIC customers, TSMC's CoWoS capacity for the next two years is already fully booked. CoPoS was originally expected to take the baton with mass production in 2028, but per the latest schedule: \nQ3 2026: R&D equipment move-in begins. R&D line buildout takes about 1 year. \nQ3 2027: Pilot line equipment orders placed. Equipment lead time of about 3 quarters. \nQ2 2028: Pilot tools moved into the Chiayi P7 fab. Validation and fine-tuning needed for about 1 year afterward. \nMid-to-late 2029: Mass production equipment finalized, supply chain orders placed. Lead time of 3 quarters. \nQ1 2030: Mass production equipment move-in. First packaging product shipment possible 6 months to 1 year later, in Q4 2030. \nIn particular, \"uniformity\" and \"warpage\" are reportedly causing significant difficulties in implementing CoPoS. \n\n-> Negative for the CoPoS supply chain. TSMC is reportedly even imposing a policy that anyone joining the CoPoS development effort will be barred from selling externally.\n\n3. Lastly, the CoWoP project led by NVIDIA and SPIL (矽品) is reportedly facing a potential suspension. Because of the higher technical difficulty and elevated cost, SPIL and Taiwanese PCB vendors are showing low willingness to participate, and only Chinese PCB vendors are showing intent to continue R&D.\n\n-> Those who were pushing CoWoP should be ashamed of themselves. The ones who were pumping mainland Chinese names...\n\n---\n\nhttps://t.co/itAP0GluPN", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.013151876623067271, "ret_signed_m1d": -0.013151876623067271, "alpha_spy_m1d": -0.010700588579063752, "alpha_c_m1d": -0.0091941197161165, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.028571486270762003, "ret_signed_p1d": 0.028571486270762003, "alpha_spy_p1d": 0.016485801140537992, "alpha_c_p1d": 0.007990975891765562, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": 0.09659867204785244, "ret_signed_p1w": 0.09659867204785244, "alpha_spy_p1w": 0.08692156444248345, "alpha_c_p1w": 0.037121937138922645, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.19478674187104938, "ret_signed_p1m": 0.19478674187104938, "alpha_spy_p1m": 0.1413277856796029, "alpha_c_p1m": -0.028476322746099347, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2345, -0.2159, -0.2107, -0.2063, -0.2061, -0.207, -0.2029, -0.2315, -0.2628, -0.254, -0.2535, -0.2737, -0.2855, -0.271, -0.2494, -0.2408, -0.2333, -0.2249, -0.2349, -0.2009, -0.1905, -0.1832, -0.1519, -0.2154, -0.1633, -0.149, -0.1279, -0.1039, -0.1433, -0.141, -0.1451, -0.1776, -0.1338, -0.1444, -0.2912, -0.2458, -0.2082, -0.2066, -0.2052, -0.1605, -0.1701, -0.1454, -0.1274, -0.1662, -0.1905, -0.1721, -0.1651, -0.1587, -0.1592, -0.207, -0.21, -0.1884, -0.1494, -0.1376, -0.1376, -0.1376, -0.1322, -0.0612, -0.0653, -0.0408, -0.0617, -0.0113, -0.0132, 0.0, 0.0286, 0.0295, 0.0336, 0.0535, 0.0966, 0.1438, 0.1318, 0.0738, 0.0926, 0.1282, 0.1282, 0.1149, 0.1592, 0.1638, 0.1624, 0.1902, 0.1925, 0.14, 0.1784, 0.2171, 0.1948, 0.1674, 0.1633, 0.2094, 0.234, 0.2486, 0.287, 0.2961, 0.2614, 0.3125, 0.2979, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2055602380750885005", "ticker_idx": 0, "ts": "2026-05-16T10:52:53+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a target price of KRW 590,000 for Samsung Electronics", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Nomura's DRAM PER-based valuation with Buy-implied targets: Samsung KRW 590K, SK Hynix KRW 4M.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura argues that DRAM makers should be valued on a PER basis, with a target price of KRW 590,000 for Samsung Electronics and KRW 4,000,000 for SK Hynix.🤯 https://t.co/aR970ogxgW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03736654804270467, "ret_signed_m1d": -0.03736654804270467, "alpha_spy_m1d": -0.03807047883388848, "alpha_c_m1d": -0.04826350772286869, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.01089695968016402, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.019572953736654797, "ret_signed_p1d": -0.019572953736654797, "alpha_spy_p1d": -0.012912094354411652, "alpha_c_p1d": -0.01314949013682154, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.006423463599833257, "ret_p1w": 0.040925266903914626, "ret_signed_p1w": 0.040925266903914626, "alpha_spy_p1w": 0.03146207052763361, "alpha_c_p1w": 0.00634165392967323, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.034583612974241396, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3565, -0.3252, -0.3249, -0.3146, -0.2897, -0.2773, -0.2258, -0.2311, -0.2311, -0.3071, -0.3885, -0.3196, -0.3316, -0.3838, -0.3327, -0.3252, -0.3327, -0.3483, -0.3299, -0.3114, -0.2595, -0.288, -0.2919, -0.3384, -0.3263, -0.3288, -0.3604, -0.3604, -0.3726, -0.405, -0.3251, -0.3651, -0.3374, -0.3128, -0.3007, -0.2509, -0.274, -0.2669, -0.2847, -0.2651, -0.2491, -0.226, -0.2313, -0.2367, -0.2206, -0.226, -0.2011, -0.2189, -0.2011, -0.21, -0.1957, -0.2153, -0.2153, -0.1726, -0.1726, -0.0534, -0.0338, -0.0445, 0.016, -0.0071, 0.0107, 0.0534, -0.0374, 0.0, -0.0196, -0.0178, 0.0658, 0.0409, 0.0409, 0.0641, 0.0925, 0.0658, 0.1281, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2055602380750885005", "ticker_idx": 1, "ts": "2026-05-16T10:52:53+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "KRW 4,000,000 for SK Hynix", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Nomura's DRAM PER-based valuation with Buy-implied targets: Samsung KRW 590K, SK Hynix KRW 4M.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura argues that DRAM makers should be valued on a PER basis, with a target price of KRW 590,000 for Samsung Electronics and KRW 4,000,000 for SK Hynix.🤯 https://t.co/aR970ogxgW", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011413049423548371, "ret_signed_m1d": -0.011413049423548371, "alpha_spy_m1d": -0.012116980214732176, "alpha_c_m1d": -0.030052375283196375, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.018639325859648004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05163043902924258, "ret_signed_p1d": -0.05163043902924258, "alpha_spy_p1d": -0.044969579646999436, "alpha_c_p1d": -0.047602404607004156, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.0040280344222384246, "ret_p1w": 0.05489131029311345, "ret_signed_p1w": 0.05489131029311345, "alpha_spy_p1w": 0.045428113916832435, "alpha_c_p1w": -0.00033066799658110213, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.05522197828969455, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5226, -0.515, -0.4852, -0.4841, -0.4548, -0.4478, -0.4027, -0.4234, -0.4234, -0.4897, -0.5386, -0.4886, -0.4978, -0.5457, -0.4902, -0.481, -0.4946, -0.5054, -0.4707, -0.4728, -0.4261, -0.4495, -0.4527, -0.4929, -0.4641, -0.4592, -0.4929, -0.4929, -0.5255, -0.5614, -0.5125, -0.5489, -0.5239, -0.5185, -0.5022, -0.4386, -0.4576, -0.4418, -0.4348, -0.4005, -0.3826, -0.3723, -0.387, -0.3663, -0.3348, -0.3353, -0.3342, -0.3359, -0.2978, -0.2935, -0.2973, -0.3011, -0.3011, -0.2136, -0.2136, -0.1299, -0.1011, -0.0837, 0.0217, -0.0027, 0.0739, 0.0707, -0.0114, 0.0, -0.0516, -0.0516, 0.0543, 0.0549, 0.0549, 0.1152, 0.219, 0.2442, 0.2681, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2043559550016733352", "ticker_idx": 0, "ts": "2026-04-13T05:18:58+00:00", "symbol": "TER", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "It sounds like bad news for Teradyne....", "tweet_type": "analysis", "timeline": "unspecified", "summary": "SiPh test bottleneck shift to FormFactor/MPI seen as bad news for Teradyne.", "resolved_tickers": ["TER"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "It sounds like bad news for Teradyne.... https://t.co/weCkWo74UW", "reply_text": null, "parent_text": null, "quoted_text": "FormFactor and MPI vie to break SiPh test bottleneck as ficonTEC reportedly exits\n\nAs AI data centers officially enter the silicon photonics (SiPh) era, the industry faces severe challenges in testing accuracy and throughput on the path toward mass production of co-packaged", "parent_author": null, "ret_m1d": -0.0057817485399223045, "ret_signed_m1d": 0.0057817485399223045, "alpha_spy_m1d": -0.0038960737156616965, "alpha_c_m1d": -0.00878944176196661, "bench_spy_m1d": -0.009677822255584001, "bench_c_m1d": -0.014571190301888914, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012482100350870606, "ret_signed_p1d": 0.012482100350870606, "alpha_spy_p1d": 0.024666981103240726, "alpha_c_p1d": 0.032015649621505604, "bench_spy_p1d": 0.01218488075237012, "bench_c_p1d": 0.019533549270634998, "ret_p1w": 0.013724858318415745, "ret_signed_p1w": -0.013724858318415745, "alpha_spy_p1w": 0.01924409070081401, "alpha_c_p1w": 0.032785709840365884, "bench_spy_p1w": 0.032968949019229754, "bench_c_p1w": 0.04651056815878163, "ret_p1m": -0.03155641846507229, "ret_signed_p1m": 0.03155641846507229, "alpha_spy_p1m": 0.10746374821736793, "alpha_c_p1m": 0.29751483556273595, "bench_spy_p1m": 0.07590732975229564, "bench_c_p1m": 0.26595841709766366, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3783, -0.3851, -0.3839, -0.3839, -0.3951, -0.3738, -0.3812, -0.3811, -0.3741, -0.3547, -0.3235, -0.3198, -0.349, -0.3261, -0.2358, -0.2733, -0.2678, -0.1895, -0.1628, -0.1766, -0.1319, -0.16, -0.1499, -0.1499, -0.1745, -0.1494, -0.1465, -0.1223, -0.1395, -0.1109, -0.0738, -0.1011, -0.1354, -0.1197, -0.1781, -0.1754, -0.1744, -0.2623, -0.1991, -0.1874, -0.1841, -0.2257, -0.2262, -0.1941, -0.1911, -0.1893, -0.183, -0.2142, -0.1789, -0.1351, -0.1264, -0.1967, -0.2013, -0.2534, -0.199, -0.1565, -0.1635, -0.1635, -0.1478, -0.1341, -0.032, -0.016, -0.0058, 0.0, -0.0125, -0.014, -0.0114, 0.0277, 0.0137, 0.0265, 0.0407, 0.0834, 0.1295, 0.0861, 0.027, -0.1724, -0.072, -0.0668, -0.0883, -0.0352, 0.0334, -0.0433, -0.028, -0.0094, -0.0316, -0.0182, -0.0367, -0.0871, -0.1326, -0.1313, -0.0697, -0.0447, -0.0312, -0.0312, 0.0518, 0.0158, 0.0342, 0.0117, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2044555350821769595", "ticker_idx": 0, "ts": "2026-04-15T23:15:56+00:00", "symbol": "MLCC", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Everything is in shortage.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Endorses MLCC long thesis by affirming the shortage backdrop driving the call.", "resolved_tickers": ["6981.T", "009150.KS", "2327.TW"], "resolution_method": "basket", "resolution_reason": "MLCC basket: Murata, Samsung Electro-Mechanics, Yageo", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@superposition_V \"Everything is in shortage.\"", "reply_text": null, "parent_text": "한줄 요약\nMLCC 롱\n\n요즘 뭣만 하면 계속 롱인가도 싶지만 🤔🤔", "quoted_text": null, "parent_author": "superposition_V", "ret_m1d": -0.040724157547559904, "ret_signed_m1d": -0.040724157547559904, "alpha_spy_m1d": -0.032894942972545284, "alpha_c_m1d": -0.025022224331105715, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.01570193321645419, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03673679123820935, "ret_signed_p1d": 0.03673679123820935, "alpha_spy_p1d": 0.034279479615605034, "alpha_c_p1d": 0.02529300424034194, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.011443786997867411, "ret_p1w": 0.12959788646920248, "ret_signed_p1w": 0.12959788646920248, "alpha_spy_p1w": 0.11349647850289116, "alpha_c_p1w": 0.07776825820598668, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.0518296282632158, "ret_p1m": 0.5016562266703263, "ret_signed_p1m": 0.5016562266703263, "alpha_spy_p1m": 0.4327503485943157, "alpha_c_p1m": 0.3073781404737027, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.19427808619662357, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3005, -0.2848, -0.3057, -0.3196, -0.2964, -0.3071, -0.3214, -0.3084, -0.3196, -0.3325, -0.3322, -0.3539, -0.3148, -0.3153, -0.3355, -0.3535, -0.3373, -0.3244, -0.3245, -0.3208, -0.3289, -0.3271, -0.3106, -0.3032, -0.2728, -0.2618, -0.2086, -0.1508, -0.1285, -0.1396, -0.1479, -0.1573, -0.2151, -0.2679, -0.2269, -0.2297, -0.2816, -0.2459, -0.2312, -0.251, -0.2488, -0.2431, -0.1838, -0.1769, -0.1812, -0.1935, -0.2497, -0.2446, -0.2167, -0.2087, -0.2092, -0.2394, -0.2753, -0.2279, -0.2625, -0.2184, -0.2138, -0.2077, -0.1449, -0.1342, -0.0925, -0.0699, -0.0407, 0.0, 0.0367, 0.0355, 0.0493, 0.1082, 0.1296, 0.0805, 0.0993, 0.1093, 0.1504, 0.1392, 0.1601, 0.1587, 0.2318, 0.238, 0.2239, 0.2989, 0.3303, 0.3334, 0.4026, 0.4425, 0.5017, 0.4735, 0.5334, 0.5061, 0.5689, 0.7419, 0.9044, 1.0275, 1.1333, 1.1873, 1.4016, 1.6285, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2046433937153761642", "ticker_idx": 0, "ts": "2026-04-21T03:40:46+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I sold Samsung in the short term", "tweet_type": "position_disclosure", "timeline": "weeks", "summary": "Sold Samsung / bought SK Hynix on strike risk weighing on Samsung's share price near-term.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@yzliu88 It’s hard to dismiss the strike factor as a reason for Samsung’s weak share price performance. I sold Samsung in the short term and bought SK Hynix instead.", "reply_text": null, "parent_text": "@jukan05 Hi Jukan, may I ask you what's your perspective on SEC Strike. Do you think they will reach an agreement similar to Sk Hynix say 10 percent? Or there might be chance of halting production. Do you think this risk is the main reason why SEC performer worse than Hynix recently?", "quoted_text": null, "parent_author": "yzliu88", "ret_m1d": -0.020547945205479423, "ret_signed_m1d": 0.020547945205479423, "alpha_spy_m1d": 0.02713803937184911, "alpha_c_m1d": 0.019707523195261034, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.000840422010218389, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006849315068493178, "ret_signed_p1d": 0.006849315068493178, "alpha_spy_p1d": 0.016976011906563837, "alpha_c_p1d": 0.028828718021649324, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.021979402953156146, "ret_p1w": 0.013698630136986356, "ret_signed_p1w": -0.013698630136986356, "alpha_spy_p1w": -0.0028902203442802143, "alpha_c_p1w": 0.006729345894020078, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.020427976031006434, "ret_p1m": 0.26027397260273966, "ret_signed_p1m": -0.26027397260273966, "alpha_spy_p1m": -0.20748170182677028, "alpha_c_p1m": -0.11514502700881968, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.14512894559391998, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.306, -0.3069, -0.3069, -0.2732, -0.26, -0.2677, -0.2686, -0.3147, -0.2367, -0.2294, -0.2741, -0.2773, -0.2418, -0.2445, -0.2354, -0.1862, -0.1743, -0.1743, -0.1743, -0.1743, -0.1342, -0.1338, -0.1205, -0.0886, -0.0727, -0.0066, -0.0135, -0.0135, -0.111, -0.2153, -0.1269, -0.1424, -0.2094, -0.1438, -0.1342, -0.1438, -0.1638, -0.1401, -0.1164, -0.0499, -0.0864, -0.0914, -0.1511, -0.1356, -0.1388, -0.1793, -0.1793, -0.195, -0.2365, -0.134, -0.1854, -0.1498, -0.1183, -0.1027, -0.0388, -0.0685, -0.0594, -0.0822, -0.0571, -0.0365, -0.0068, -0.0137, -0.0205, 0.0, -0.0068, 0.0251, 0.0023, 0.0251, 0.0137, 0.032, 0.0068, 0.0068, 0.0616, 0.0616, 0.2146, 0.2397, 0.226, 0.3037, 0.274, 0.2968, 0.3516, 0.2352, 0.2831, 0.258, 0.2603, 0.3676, 0.3356, 0.3356, 0.3653, 0.4018, 0.3676, 0.4475, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2046433937153761642", "ticker_idx": 1, "ts": "2026-04-21T03:40:46+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "bought SK Hynix instead", "tweet_type": "position_disclosure", "timeline": "weeks", "summary": "Sold Samsung / bought SK Hynix on strike risk weighing on Samsung's share price near-term.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@yzliu88 It’s hard to dismiss the strike factor as a reason for Samsung’s weak share price performance. I sold Samsung in the short term and bought SK Hynix instead.", "reply_text": null, "parent_text": "@jukan05 Hi Jukan, may I ask you what's your perspective on SEC Strike. Do you think they will reach an agreement similar to Sk Hynix say 10 percent? Or there might be chance of halting production. Do you think this risk is the main reason why SEC performer worse than Hynix recently?", "quoted_text": null, "parent_author": "yzliu88", "ret_m1d": -0.04738567916225378, "ret_signed_m1d": -0.04738567916225378, "alpha_spy_m1d": -0.053975773328623466, "alpha_c_m1d": -0.045879175048462506, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.0015065041137912738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0008170279283822302, "ret_signed_p1d": -0.0008170279283822302, "alpha_spy_p1d": -0.01094372476645289, "alpha_c_p1d": -0.02700818450360154, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.02619115657521931, "ret_p1w": 0.06209146688432288, "ret_signed_p1w": 0.06209146688432288, "alpha_spy_p1w": 0.051283057091616735, "alpha_c_p1w": 0.004952940685120755, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.05713852619920212, "ret_p1m": 0.425653573825608, "ret_signed_p1m": 0.425653573825608, "alpha_spy_p1m": 0.3728613030496386, "alpha_c_p1m": 0.21044251908817535, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.21521105473743263, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3843, -0.3745, -0.3998, -0.3476, -0.3142, -0.2979, -0.2587, -0.3231, -0.2604, -0.2661, -0.3134, -0.3158, -0.2767, -0.2856, -0.2987, -0.2758, -0.2824, -0.2824, -0.2824, -0.2824, -0.271, -0.2261, -0.2245, -0.1804, -0.1698, -0.1021, -0.1332, -0.1332, -0.2328, -0.3064, -0.2312, -0.2451, -0.317, -0.2337, -0.2198, -0.2402, -0.2565, -0.2042, -0.2075, -0.1373, -0.1724, -0.1773, -0.2377, -0.1944, -0.1871, -0.2377, -0.2377, -0.2868, -0.3407, -0.2672, -0.3219, -0.2843, -0.2761, -0.2516, -0.156, -0.1846, -0.1609, -0.1503, -0.0989, -0.0719, -0.0564, -0.0784, -0.0474, 0.0, -0.0008, 0.0008, -0.0016, 0.0556, 0.0621, 0.0564, 0.0507, 0.0507, 0.1822, 0.1822, 0.308, 0.3513, 0.3775, 0.5359, 0.4992, 0.6144, 0.6095, 0.4861, 0.5033, 0.4257, 0.4257, 0.585, 0.5858, 0.5858, 0.6765, 0.8325, 0.8704, 0.9064, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046721056602485090", "ticker_idx": 0, "ts": "2026-04-21T22:41:40+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel raised PC CPU prices in March and server CPU prices in April, improving 2Q margins, with room for an additional ~8–10% hike in 2H", "tweet_type": "analysis", "timeline": "quarters", "summary": "CPU supply shortage + price upcycle through 2026–2027 benefits INTC, AMD, TSMC, and the broader foundry/OSAT/equipment supply chain.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Strong CPU Pull-In; Supply Chain: Prices Continue Rising in Q3\n\n- CPU price increases are driven by two factors: 1) surging AI server demand, and 2) deepening supply shortages caused by capacity concentration on leading-edge nodes\n\n- Intel raised PC CPU prices in March and server CPU prices in April, improving 2Q margins, with room for an additional ~8–10% hike in 2H\n\n- AMD is likewise expected to raise prices in both 2Q and 3Q this year, centered on server CPUs, with cumulative annual increases projected at 16–17% — marking entry into a broad CPU price upcycle\n\n- As 2nm/3nm processes enter mass production and capacity expansion phases, capacity competition among AI chips, CPUs, GPUs, and TPUs is intensifying, putting upward pressure on foundry pricing and lead times\n\n- CPUs are currently in a clear state of undersupply, with the supply-demand imbalance persisting and the price upcycle likely to be prolonged\n\n- TSMC is, unusually, continuing to expand 3nm capacity — driven by simultaneous CPU and AI ASIC demand growth and the fact that Intel, AMD, and Nvidia's Vera CPU all adopt 3nm\n\n- Intel recently strengthened its control over leading-edge capacity by repurchasing the 49% stake in its Ireland Fab 34 ($14.2B)\n\n- The facility is the core production base for Intel 4 and Intel 3 processes, ranking among the highest-volume leading-edge fabs\n\n- For the 2026–2027 CPU market, capacity — rather than demand — will be the key variable; if AI infrastructure investment continues to expand and process/packaging bottlenecks persist, the price uptrend is expected to hold, with benefits spreading broadly across the foundry, OSAT, equipment, and materials supply chain\n\n---\n\nhttps://t.co/fFVy1wXIbd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.00845163250729919, "ret_signed_m1d": -0.00845163250729919, "alpha_spy_m1d": -0.015041726673668876, "alpha_c_m1d": -0.0069451283935079156, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.0015065041137912738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.014941223381318536, "ret_signed_p1d": -0.014941223381318536, "alpha_spy_p1d": -0.025067920219389195, "alpha_c_p1d": -0.041132379956537846, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.02619115657521931, "ret_p1w": 0.2755809525827273, "ret_signed_p1w": 0.2755809525827273, "alpha_spy_p1w": 0.26477254279002116, "alpha_c_p1w": 0.21844242638352518, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.05713852619920212, "ret_p1m": 0.7953515733351633, "ret_signed_p1m": 0.7953515733351633, "alpha_spy_p1m": 0.7425593025591939, "alpha_c_p1m": 0.5801405185977306, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.21521105473743263, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1802, -0.3198, -0.3587, -0.337, -0.2638, -0.2656, -0.2987, -0.2634, -0.2567, -0.2665, -0.272, -0.2365, -0.2418, -0.2887, -0.2712, -0.2985, -0.2938, -0.2938, -0.303, -0.3139, -0.3266, -0.3343, -0.3415, -0.304, -0.2925, -0.3139, -0.3117, -0.3133, -0.3495, -0.3121, -0.3065, -0.3447, -0.3121, -0.294, -0.2759, -0.3171, -0.3092, -0.3094, -0.335, -0.3204, -0.303, -0.3379, -0.3358, -0.335, -0.288, -0.3344, -0.3491, -0.3784, -0.334, -0.2751, -0.2397, -0.2397, -0.2336, -0.2015, -0.1103, -0.0685, -0.0586, -0.0163, -0.037, -0.0199, 0.0338, 0.0338, -0.0085, 0.0, -0.0149, 0.0078, 0.2457, 0.2827, 0.2756, 0.43, 0.4259, 0.5035, 0.4455, 0.6322, 0.7056, 0.6544, 0.8853, 0.9535, 0.8203, 0.8154, 0.7496, 0.6416, 0.6325, 0.6722, 0.7954, 0.7884, 0.8086, 0.8086, 0.8642, 0.8378, 0.8245, 0.7308, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046721056602485090", "ticker_idx": 1, "ts": "2026-04-21T22:41:40+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD is likewise expected to raise prices in both 2Q and 3Q this year, centered on server CPUs, with cumulative annual increases projected at 16–17% — marking entry into a broad CPU price upcycle", "tweet_type": "analysis", "timeline": "quarters", "summary": "CPU supply shortage + price upcycle through 2026–2027 benefits INTC, AMD, TSMC, and the broader foundry/OSAT/equipment supply chain.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Strong CPU Pull-In; Supply Chain: Prices Continue Rising in Q3\n\n- CPU price increases are driven by two factors: 1) surging AI server demand, and 2) deepening supply shortages caused by capacity concentration on leading-edge nodes\n\n- Intel raised PC CPU prices in March and server CPU prices in April, improving 2Q margins, with room for an additional ~8–10% hike in 2H\n\n- AMD is likewise expected to raise prices in both 2Q and 3Q this year, centered on server CPUs, with cumulative annual increases projected at 16–17% — marking entry into a broad CPU price upcycle\n\n- As 2nm/3nm processes enter mass production and capacity expansion phases, capacity competition among AI chips, CPUs, GPUs, and TPUs is intensifying, putting upward pressure on foundry pricing and lead times\n\n- CPUs are currently in a clear state of undersupply, with the supply-demand imbalance persisting and the price upcycle likely to be prolonged\n\n- TSMC is, unusually, continuing to expand 3nm capacity — driven by simultaneous CPU and AI ASIC demand growth and the fact that Intel, AMD, and Nvidia's Vera CPU all adopt 3nm\n\n- Intel recently strengthened its control over leading-edge capacity by repurchasing the 49% stake in its Ireland Fab 34 ($14.2B)\n\n- The facility is the core production base for Intel 4 and Intel 3 processes, ranking among the highest-volume leading-edge fabs\n\n- For the 2026–2027 CPU market, capacity — rather than demand — will be the key variable; if AI infrastructure investment continues to expand and process/packaging bottlenecks persist, the price uptrend is expected to hold, with benefits spreading broadly across the foundry, OSAT, equipment, and materials supply chain\n\n---\n\nhttps://t.co/fFVy1wXIbd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.033533615785512505, "ret_signed_m1d": -0.033533615785512505, "alpha_spy_m1d": -0.04012370995188219, "alpha_c_m1d": -0.03202711167172123, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.0015065041137912738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.06668073349461201, "ret_signed_p1d": 0.06668073349461201, "alpha_spy_p1d": 0.05655403665654135, "alpha_c_p1d": 0.0404895769193927, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.02619115657521931, "ret_p1w": 0.1361032111843512, "ret_signed_p1w": 0.1361032111843512, "alpha_spy_p1w": 0.12529480139164506, "alpha_c_p1w": 0.07896468498514908, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.05713852619920212, "ret_p1m": 0.5732714750474353, "ret_signed_p1m": 0.5732714750474353, "alpha_spy_p1m": 0.5204792042714659, "alpha_c_p1m": 0.3580604203100026, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.21521105473743263, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1081, -0.0872, -0.1166, -0.1141, -0.1116, -0.1136, -0.1679, -0.1343, -0.149, -0.2963, -0.3234, -0.2673, -0.2407, -0.2493, -0.2493, -0.2761, -0.2713, -0.2713, -0.2862, -0.2966, -0.2851, -0.2965, -0.3089, -0.2483, -0.2588, -0.2841, -0.2962, -0.3018, -0.3288, -0.2897, -0.2989, -0.3236, -0.2876, -0.2856, -0.28, -0.3049, -0.3202, -0.309, -0.31, -0.2989, -0.2785, -0.2923, -0.2876, -0.2781, -0.2257, -0.2837, -0.29, -0.3109, -0.2849, -0.2611, -0.2355, -0.2355, -0.2261, -0.2213, -0.1851, -0.1682, -0.1387, -0.1324, -0.1034, -0.0927, -0.0219, -0.0214, -0.0335, 0.0, 0.0667, 0.0733, 0.2226, 0.1762, 0.1361, 0.185, 0.2461, 0.2673, 0.2005, 0.2488, 0.4812, 0.4358, 0.6, 0.6127, 0.5758, 0.566, 0.5807, 0.4907, 0.4798, 0.4554, 0.5733, 0.5803, 0.6433, 0.6433, 0.7712, 0.7419, 0.8211, 0.8141, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046721056602485090", "ticker_idx": 2, "ts": "2026-04-21T22:41:40+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC is, unusually, continuing to expand 3nm capacity — driven by simultaneous CPU and AI ASIC demand growth", "tweet_type": "analysis", "timeline": "quarters", "summary": "CPU supply shortage + price upcycle through 2026–2027 benefits INTC, AMD, TSMC, and the broader foundry/OSAT/equipment supply chain.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Strong CPU Pull-In; Supply Chain: Prices Continue Rising in Q3\n\n- CPU price increases are driven by two factors: 1) surging AI server demand, and 2) deepening supply shortages caused by capacity concentration on leading-edge nodes\n\n- Intel raised PC CPU prices in March and server CPU prices in April, improving 2Q margins, with room for an additional ~8–10% hike in 2H\n\n- AMD is likewise expected to raise prices in both 2Q and 3Q this year, centered on server CPUs, with cumulative annual increases projected at 16–17% — marking entry into a broad CPU price upcycle\n\n- As 2nm/3nm processes enter mass production and capacity expansion phases, capacity competition among AI chips, CPUs, GPUs, and TPUs is intensifying, putting upward pressure on foundry pricing and lead times\n\n- CPUs are currently in a clear state of undersupply, with the supply-demand imbalance persisting and the price upcycle likely to be prolonged\n\n- TSMC is, unusually, continuing to expand 3nm capacity — driven by simultaneous CPU and AI ASIC demand growth and the fact that Intel, AMD, and Nvidia's Vera CPU all adopt 3nm\n\n- Intel recently strengthened its control over leading-edge capacity by repurchasing the 49% stake in its Ireland Fab 34 ($14.2B)\n\n- The facility is the core production base for Intel 4 and Intel 3 processes, ranking among the highest-volume leading-edge fabs\n\n- For the 2026–2027 CPU market, capacity — rather than demand — will be the key variable; if AI infrastructure investment continues to expand and process/packaging bottlenecks persist, the price uptrend is expected to hold, with benefits spreading broadly across the foundry, OSAT, equipment, and materials supply chain\n\n---\n\nhttps://t.co/fFVy1wXIbd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012195121951219523, "ret_signed_m1d": -0.012195121951219523, "alpha_spy_m1d": -0.01878521611758921, "alpha_c_m1d": -0.01068861783742825, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.0015065041137912738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.010126696838070659, "alpha_c_p1d": -0.02619115657521931, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.02619115657521931, "ret_p1w": 0.08048780487804885, "ret_signed_p1w": 0.08048780487804885, "alpha_spy_p1w": 0.06967939508534271, "alpha_c_p1w": 0.02334927867884673, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.05713852619920212, "ret_p1m": 0.06585365853658542, "ret_signed_p1m": 0.06585365853658542, "alpha_spy_p1m": 0.01306138776061605, "alpha_c_p1m": -0.1493573962008472, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.21521105473743263, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1443, -0.1394, -0.1467, -0.1345, -0.1151, -0.1224, -0.137, -0.1418, -0.1248, -0.1321, -0.1418, -0.1345, -0.1175, -0.0859, -0.0689, -0.0689, -0.0689, -0.0689, -0.0689, -0.0689, -0.0689, -0.0689, -0.0762, -0.0446, -0.0203, -0.03, -0.03, -0.0397, -0.0592, -0.0932, -0.0762, -0.081, -0.1199, -0.1005, -0.0567, -0.0835, -0.0932, -0.1029, -0.0878, -0.0707, -0.0976, -0.1024, -0.1171, -0.1171, -0.1, -0.1024, -0.1122, -0.1317, -0.1415, -0.0951, -0.1171, -0.1171, -0.1171, -0.0927, -0.0488, -0.0463, -0.0244, -0.0293, 0.0024, 0.0146, 0.0171, -0.0098, -0.0122, 0.0, 0.0, 0.0146, 0.0659, 0.1049, 0.0805, 0.0634, 0.0415, 0.0415, 0.1098, 0.0976, 0.0976, 0.1268, 0.1171, 0.0902, 0.1, 0.0829, 0.1073, 0.1049, 0.0927, 0.0756, 0.0659, 0.0878, 0.1, 0.1268, 0.1073, 0.122, 0.1195, 0.1488, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046721056602485090", "ticker_idx": 3, "ts": "2026-04-21T22:41:40+00:00", "symbol": "foundry", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "benefits spreading broadly across the foundry, OSAT, equipment, and materials supply chain", "tweet_type": "analysis", "timeline": "quarters", "summary": "CPU supply shortage + price upcycle through 2026–2027 benefits INTC, AMD, TSMC, and the broader foundry/OSAT/equipment supply chain.", "resolved_tickers": ["TSM", "005930.KS", "UMC", "0981.HK"], "resolution_method": "basket", "resolution_reason": "Foundry sector basket: TSMC, Samsung, UMC, SMIC", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Strong CPU Pull-In; Supply Chain: Prices Continue Rising in Q3\n\n- CPU price increases are driven by two factors: 1) surging AI server demand, and 2) deepening supply shortages caused by capacity concentration on leading-edge nodes\n\n- Intel raised PC CPU prices in March and server CPU prices in April, improving 2Q margins, with room for an additional ~8–10% hike in 2H\n\n- AMD is likewise expected to raise prices in both 2Q and 3Q this year, centered on server CPUs, with cumulative annual increases projected at 16–17% — marking entry into a broad CPU price upcycle\n\n- As 2nm/3nm processes enter mass production and capacity expansion phases, capacity competition among AI chips, CPUs, GPUs, and TPUs is intensifying, putting upward pressure on foundry pricing and lead times\n\n- CPUs are currently in a clear state of undersupply, with the supply-demand imbalance persisting and the price upcycle likely to be prolonged\n\n- TSMC is, unusually, continuing to expand 3nm capacity — driven by simultaneous CPU and AI ASIC demand growth and the fact that Intel, AMD, and Nvidia's Vera CPU all adopt 3nm\n\n- Intel recently strengthened its control over leading-edge capacity by repurchasing the 49% stake in its Ireland Fab 34 ($14.2B)\n\n- The facility is the core production base for Intel 4 and Intel 3 processes, ranking among the highest-volume leading-edge fabs\n\n- For the 2026–2027 CPU market, capacity — rather than demand — will be the key variable; if AI infrastructure investment continues to expand and process/packaging bottlenecks persist, the price uptrend is expected to hold, with benefits spreading broadly across the foundry, OSAT, equipment, and materials supply chain\n\n---\n\nhttps://t.co/fFVy1wXIbd", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0013491380609647052, "ret_signed_m1d": -0.0013491380609647052, "alpha_spy_m1d": -0.007939232227334392, "alpha_c_m1d": 0.00015736605282656857, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.0015065041137912738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.015814001369721498, "ret_signed_p1d": 0.015814001369721498, "alpha_spy_p1d": 0.005687304531650839, "alpha_c_p1d": -0.010377155205497812, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.02619115657521931, "ret_p1w": 0.031067742363071887, "ret_signed_p1w": 0.031067742363071887, "alpha_spy_p1w": 0.020259332570365746, "alpha_c_p1w": -0.026070783836130235, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.05713852619920212, "ret_p1m": 0.25649504618066493, "ret_signed_p1m": 0.25649504618066493, "alpha_spy_p1m": 0.20370277540469556, "alpha_c_p1m": 0.041283991443232304, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.21521105473743263, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0599, -0.0613, -0.0574, -0.0175, -0.0226, -0.0511, -0.0732, -0.0936, -0.089, -0.099, -0.1118, -0.0979, -0.0733, -0.0674, -0.0555, -0.0482, -0.0415, -0.0413, -0.0458, -0.048, -0.0412, -0.04, -0.0225, -0.0008, -0.0011, -0.0064, -0.0051, -0.0216, -0.0776, -0.1066, -0.0886, -0.1044, -0.1158, -0.0916, -0.0816, -0.1058, -0.1151, -0.1026, -0.0938, -0.0866, -0.1052, -0.1276, -0.1483, -0.1376, -0.1247, -0.1694, -0.1747, -0.1954, -0.1866, -0.1498, -0.1781, -0.1692, -0.1591, -0.1402, -0.0856, -0.094, -0.0733, -0.0851, -0.0707, -0.0559, -0.0402, -0.0185, -0.0013, 0.0, 0.0158, 0.0026, 0.0404, 0.0523, 0.0311, 0.0577, 0.0804, 0.0813, 0.1015, 0.1119, 0.2087, 0.2179, 0.2039, 0.2354, 0.2326, 0.227, 0.2664, 0.2281, 0.2226, 0.2225, 0.2565, 0.2991, 0.3102, 0.3102, 0.3999, 0.4425, 0.4574, 0.4352, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2045072922876977643", "ticker_idx": 0, "ts": "2026-04-17T09:32:35+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "TSMC 케파가 정말 빡빡해질거 같거든요", "tweet_type": "prediction", "timeline": "months", "summary": "TSMC capacity will get very tight as CPU demand adds to HBM base-die load; bullish on TSMC capacity utilization.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@treeapple33 지금 제 견해로는...삼성을 제외한 디램 2사가 과연 HBM 베이스 다이를 위한 케파를 넉넉히 확보할 수 있을지 조금 걱정됩니다. 이제 CPU마저 쇼티지되면 TSMC 케파가 정말 빡빡해질거 같거든요.\n\n물론, 삼성도 TSMC랑 베이스다이 논의하고 있습니다?", "reply_text": null, "parent_text": "@jukan05 삼성파운드리가 있다는게 삼성 메모리에 정말로 유리한게 맞는지 아직은 잘 모르겠네요 ㅎ", "quoted_text": null, "parent_author": "treeapple33", "ret_m1d": -0.019298229140308654, "ret_signed_m1d": -0.019298229140308654, "alpha_spy_m1d": -0.007356863594101037, "alpha_c_m1d": 0.0008672650756780254, "bench_spy_m1d": -0.011941365546207616, "bench_c_m1d": -0.02016549421598668, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01149800206646423, "ret_signed_p1d": -0.01149800206646423, "alpha_spy_p1d": -0.009498334513558526, "alpha_c_p1d": -0.011067089869262414, "bench_spy_p1d": -0.001999667552905704, "bench_c_p1d": -0.00043091219720181595, "ret_p1w": 0.08626178530385453, "ret_signed_p1w": 0.08626178530385453, "alpha_spy_p1w": 0.08091074506589524, "alpha_c_p1w": -0.004827490906327281, "bench_spy_p1w": 0.0053510402379592925, "bench_c_p1w": 0.09108927621018181, "ret_p1m": 0.0686909911120952, "ret_signed_p1m": 0.0686909911120952, "alpha_spy_p1m": 0.028543964923549492, "alpha_c_p1m": -0.1079721613255602, "bench_spy_p1m": 0.04014702618854571, "bench_c_p1m": 0.1766631524376554, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1195, -0.1223, -0.1189, -0.0987, -0.1045, -0.0894, -0.0787, -0.0861, -0.1103, -0.0812, -0.0963, -0.1233, -0.1098, -0.0611, -0.0434, -0.0259, 0.0069, -0.0093, -0.014, -0.014, -0.0198, -0.025, -0.03, -0.0027, -0.004, 0.0382, 0.0436, 0.0142, 0.0082, -0.0066, -0.0496, -0.038, -0.0476, -0.0879, -0.0615, -0.0658, -0.0457, -0.0938, -0.0894, -0.0843, -0.0662, -0.0835, -0.0856, -0.1114, -0.0865, -0.0735, -0.0614, -0.1198, -0.1181, -0.1457, -0.0879, -0.0783, -0.0849, -0.0849, -0.0776, -0.068, -0.0124, -0.0135, 0.0003, -0.0025, 0.0253, 0.0124, -0.0193, 0.0, -0.0115, -0.0065, 0.0457, 0.0328, 0.0863, 0.0931, 0.0589, 0.063, 0.069, 0.0733, 0.084, 0.0645, 0.1323, 0.1178, 0.1111, 0.0919, 0.0723, 0.0791, 0.1274, 0.0914, 0.0687, 0.0597, 0.084, 0.0989, 0.0918, 0.0918, 0.1129, 0.141, 0.1467, 0.1294, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2045547185954304500", "ticker_idx": 0, "ts": "2026-04-18T16:57:08+00:00", "symbol": "IGV", "kind": "sector", "geography": "US", "sub_sector": null, "direction": "short", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "Software will never return to the era when it commanded 50x revenue multiples… Software companies now have to fight not just for growth, but for survival itself.", "tweet_type": "prediction", "timeline": "years", "summary": "Bearish enterprise software sector: multiples will never recover to 50x revenue; companies fighting for survival, not just growth.", "resolved_tickers": ["IGV"], "resolution_method": "etf", "resolution_reason": "iShares Expanded Tech-Software ETF IGV US", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "quote", "lang": "en", "tweet_text": "Software will never return to the era when it commanded 50x revenue multiples…\n\nSoftware companies now have to fight not just for growth, but for survival itself.\n\nThis is a truly great piece. You should definitely read it.", "reply_text": null, "parent_text": null, "quoted_text": "In August I wrote a thesis I never published. The funds I was warning were key Crossover Research clients, so I stayed quiet. Since then, 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗺𝘂𝗹𝘁𝗶𝗽𝗹𝗲𝘀 𝗮𝗿𝗲 𝗱𝗼𝘄𝗻 𝟱𝟬%+. Salesforce $CRM, ServiceNow $NOW, Adobe $ADBE, Workday $WDAY all off 40% from", "parent_author": null, "ret_m1d": -0.014136745974057896, "ret_signed_m1d": 0.014136745974057896, "alpha_spy_m1d": 0.016140420209318607, "alpha_c_m1d": 0.016140420209318607, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": 0.0020036742352607106, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.004403212514257682, "ret_signed_p1d": -0.004403212514257682, "alpha_spy_p1d": -0.010950161669193936, "alpha_c_p1d": -0.010950161669193936, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": -0.006546949154936255, "ret_p1w": -0.010081143868097553, "ret_signed_p1w": 0.010081143868097553, "alpha_spy_p1w": 0.019182075792778863, "alpha_c_p1w": 0.019182075792778863, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.00910093192468131, "ret_p1m": 0.06546922012384893, "ret_signed_p1m": -0.06546922012384893, "alpha_spy_p1m": -0.03018023321860852, "alpha_c_p1m": -0.03018023321860852, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.035288986905240405, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1054, 0.1233, 0.1322, 0.145, 0.1328, 0.1247, 0.0692, 0.0465, 0.0373, -0.0105, -0.0285, -0.0768, -0.0445, -0.0144, -0.0103, -0.0356, -0.0619, -0.0409, -0.0409, -0.0619, -0.0498, -0.0524, -0.064, -0.1085, -0.0914, -0.0632, -0.0429, -0.0548, -0.0409, -0.0253, -0.0075, 0.0153, 0.0194, 0.0163, -0.0075, -0.0065, -0.0152, -0.0244, -0.0156, -0.0089, -0.0227, -0.0213, -0.0384, -0.0216, -0.0635, -0.0684, -0.0759, -0.109, -0.1006, -0.0724, -0.0757, -0.0691, -0.0691, -0.0677, -0.0672, -0.0759, -0.1119, -0.1348, -0.0881, -0.079, -0.0385, -0.0225, -0.0141, 0.0, 0.0044, 0.0283, -0.0316, -0.0127, -0.0101, -0.0151, -0.0209, -0.0277, 0.0038, 0.0248, 0.0228, 0.0176, 0.0527, 0.0562, 0.0509, 0.0364, 0.0267, 0.0503, 0.0635, 0.0761, 0.0655, 0.0813, 0.0716, 0.0893, 0.0893, 0.0898, 0.0782, 0.1087, 0.178, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2046395114952790308", "ticker_idx": 0, "ts": "2026-04-21T01:06:30+00:00", "symbol": "009150.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electro-Mechanics is the only company that can benefit from both the ABF substrate supercycle and the MLCC supercycle", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Long Samsung Electro-Mechanics as the sole play on both ABF substrate and MLCC supercycles.", "resolved_tickers": ["009150.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Samsung Electro-Mechanics… it’s up 70% in just one month.\n\nDid you hear me? Anon?\n\nSamsung Electro-Mechanics is the only company that can benefit from both the ABF substrate supercycle and the MLCC supercycle. https://t.co/Zd7nvCBOgw", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.11917098445595853, "ret_signed_m1d": -0.11917098445595853, "alpha_spy_m1d": -0.12576107862232822, "alpha_c_m1d": -0.11833056244574014, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.000840422010218389, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.051813471502590636, "ret_signed_p1d": 0.051813471502590636, "alpha_spy_p1d": 0.04168677466451998, "alpha_c_p1d": 0.02983406854943449, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.021979402953156146, "ret_p1w": 0.08678756476683946, "ret_signed_p1w": 0.08678756476683946, "alpha_spy_p1w": 0.07597915497413332, "alpha_c_p1w": 0.06635958873583303, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.020427976031006434, "ret_p1m": 0.3743523316062176, "ret_signed_p1m": 0.3743523316062176, "alpha_spy_p1m": 0.32156006083024824, "alpha_c_p1m": 0.22922338601229764, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.14512894559391998, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6334, -0.6477, -0.6438, -0.6412, -0.6373, -0.638, -0.6386, -0.6334, -0.6017, -0.6049, -0.6244, -0.6321, -0.6062, -0.5939, -0.5991, -0.5861, -0.5991, -0.5991, -0.5991, -0.5991, -0.5363, -0.5117, -0.4475, -0.4223, -0.3977, -0.3957, -0.419, -0.419, -0.4702, -0.5304, -0.4799, -0.476, -0.5052, -0.4793, -0.4741, -0.4825, -0.4812, -0.465, -0.4216, -0.399, -0.3789, -0.3983, -0.4462, -0.4339, -0.4113, -0.408, -0.408, -0.4443, -0.4722, -0.4249, -0.4598, -0.4093, -0.4016, -0.408, -0.3342, -0.3316, -0.2681, -0.2642, -0.2435, -0.2047, -0.1723, -0.1205, -0.1192, 0.0, 0.0518, 0.0026, 0.0207, 0.0285, 0.0868, 0.0712, 0.0777, 0.0777, 0.1891, 0.1891, 0.1813, 0.1878, 0.1839, 0.1658, 0.2409, 0.3329, 0.3264, 0.3083, 0.3355, 0.2785, 0.3744, 0.5596, 0.7358, 0.7358, 1.0363, 1.1114, 1.3951, 1.7552, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044287982086348936", "ticker_idx": 0, "ts": "2026-04-15T05:33:30+00:00", "symbol": "ASML", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "To those shorting ASML, you might want to be careful. TSMC earnings are tomorrow, right?", "tweet_type": "prediction", "timeline": "days", "summary": "Warning to ASML shorts ahead of TSMC earnings; implied bullish catalyst from C.C. Wei's expected commentary.", "resolved_tickers": ["ASML"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "To those shorting ASML, you might want to be careful. TSMC earnings are tomorrow, right?\n\nWe already know what Uncle C.C. Wei is going to say there.\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.024652986359873008, "ret_signed_m1d": 0.024652986359873008, "alpha_spy_m1d": 0.03248220093488763, "alpha_c_m1d": 0.026860491878636816, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.002207505518763808, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.04787518440079108, "ret_signed_p1d": -0.04787518440079108, "alpha_spy_p1d": -0.0503324960233954, "alpha_c_p1d": -0.051848667387477065, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.003973482986685983, "ret_p1w": -0.025719183725769534, "ret_signed_p1w": -0.025719183725769534, "alpha_spy_p1w": -0.041820591692080855, "alpha_c_p1w": -0.07832401059611316, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.052604826870343624, "ret_p1m": 0.07167167924186213, "ret_signed_p1m": 0.07167167924186213, "alpha_spy_p1m": 0.002765801165851567, "alpha_c_p1m": -0.20501712176358766, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.2766888010054498, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0844, -0.0844, -0.1063, -0.0833, -0.0598, -0.0638, -0.0474, -0.0196, -0.041, -0.0193, -0.0409, -0.0285, -0.0592, -0.0975, -0.09, -0.0477, -0.0366, -0.046, -0.0311, -0.0505, -0.0507, -0.0507, -0.0418, -0.0088, -0.0154, -0.0082, 0.0028, 0.0108, 0.0302, -0.0121, -0.0211, -0.0393, -0.0815, -0.0556, -0.0765, -0.1275, -0.0839, -0.0664, -0.0642, -0.0879, -0.0918, -0.0717, -0.0625, -0.0854, -0.0779, -0.111, -0.0757, -0.0556, -0.0593, -0.1028, -0.121, -0.1537, -0.1086, -0.0823, -0.111, -0.111, -0.12, -0.1183, -0.041, -0.0224, -0.0024, 0.0124, 0.0247, 0.0, -0.0479, -0.0148, -0.0036, -0.0154, -0.0257, -0.0432, -0.0162, -0.0312, -0.0636, -0.0571, -0.0267, -0.0348, -0.0624, -0.0241, 0.0448, 0.0257, 0.0768, 0.059, 0.0287, 0.0697, 0.0717, 0.0157, -0.0042, -0.0129, 0.0484, 0.0767, 0.1044, 0.1044, 0.1038, 0.0807, 0.0861, 0.0908, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2039315025752703328", "ticker_idx": 0, "ts": "2026-04-01T12:12:45+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel has carried out two rounds of CPU price hikes since the start of 2026... The full-year target is a cumulative 30% price increase... Server CPU demand is strong globally, with orders already up 50% YoY vs. 2025... Intel's production line utilization has reached 95%", "tweet_type": "analysis", "timeline": "months", "summary": "Shares bullish CPU channel check: INTC executing 30% cumulative price hikes, 95% utilization, surging server demand; AMD gaining server share on stable supply and price-performance.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "A very spicy piece of information here. A must-read for everyone (especially if you're interested in the CPU shortage).\n\n- Intel has carried out two rounds of CPU price hikes since the start of 2026. The first round in February was 10–15%, and the second on March 16th was 15%, bringing the cumulative increase to ~20% vs. January.\n\n- The full-year target is a cumulative 30% price increase, aimed at recouping investment and meeting capital market expectations.\n\n- A third round of price adjustments is reportedly planned for May.\n\n- Server CPU demand is strong globally, with orders already up 50% YoY vs. 2025.\n\n- Chinese CSPs in particular are placing additional orders — Tencent is scaling its server CPU procurement from ~300K units in 2025 to a 2026 target of 900K–1M units, deploying them toward AI infrastructure buildout.\n\n- Alibaba has the largest procurement scale among Chinese CSPs, and is also expected to grow 30–35% YoY.\n\n- Consumer CPU shipments are also picking up, driven by local AI PC demand, with a projected ~20% YoY increase in shipments (though I find this one a bit dubious). Intel's consumer CPU share remains stable at 65%.\n\n- Intel's production line utilization has reached 95%, but is not yet physically at full capacity. Server market deliveries are smooth (China Q1 orders +60%, deliveries +27%), but on the consumer side, despite orders doubling (+100%), deliveries actually declined 3.8%. This is due to PTL yield issues and new product supply being prioritized for North America — Intel is managing this by extending lead times to 3 months.\n\n- Intel's consumer CPU market share remains stable at 65%.\n\n- On the server side, Intel's share has dropped from 65% in 2025 to 50%, primarily due to AMD's stable supply and superior price-performance.\n\n- A particularly noteworthy point: the CPU-to-GPU ratio in AI servers has risen from 1:12 to 1:8, with a target of 1:4, meaning the share of CPU procurement is increasing. Very important.\n\n- Cloud providers' AI servers generally have excess compute capacity, and there are no plans for large-scale traditional server expansion for the time being.\n\n- AI applications are expected to drive ~10% of incremental demand for consumer CPUs, with long-term growth dependent on product usability.\n\n$INTC $AMD $ARM", "reply_text": null, "parent_text": null, "quoted_text": "CPU紀要\n\n漲價情況\nIntel 2026年已執行兩輪CPU漲價：2月首輪漲10-15%，3月16日第二輪漲15%，��計較1月上漲約20%。目標是全年累計漲30%，以回收投資並滿足資本市場預期。5月計劃第三輪調價，但受AMD與ARM競爭影響，進一步大漲空間有限。\n\n需求表現\n-", "parent_author": null, "ret_m1d": -0.08119920487823784, "ret_signed_m1d": -0.08119920487823784, "alpha_spy_m1d": -0.07372108472083783, "alpha_c_m1d": -0.05933526810366485, "bench_spy_m1d": -0.0074781201574000145, "bench_c_m1d": -0.02186393677457299, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.048927802384856944, "ret_signed_p1d": 0.048927802384856944, "alpha_spy_p1d": 0.048027327956765786, "alpha_c_p1d": 0.048034861337262535, "bench_spy_p1d": 0.0009004744280911581, "bench_c_p1d": 0.0008929410475944088, "ret_p1w": 0.22735794839015488, "ret_signed_p1w": 0.22735794839015488, "alpha_spy_p1w": 0.1956596090934024, "alpha_c_p1w": 0.1483977929939655, "bench_spy_p1w": 0.031698339296752476, "bench_c_p1w": 0.07896015539618939, "ret_p1m": 0.9671040132871873, "ret_signed_p1m": 0.9671040132871873, "alpha_spy_p1m": 0.870315075103681, "alpha_c_p1m": 0.6743520178726485, "bench_spy_p1m": 0.09678893818350631, "bench_c_p1m": 0.2927519954145388, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1801, -0.1803, -0.1664, -0.1124, -0.1441, -0.0516, -0.0827, -0.0154, 0.0144, 0.006, -0.0223, -0.0223, 0.011, 0.1295, 0.131, -0.0616, -0.1153, -0.0854, 0.0156, 0.0131, -0.0325, 0.0162, 0.0254, 0.0119, 0.0044, 0.0533, 0.046, -0.0187, 0.0054, -0.0323, -0.0258, -0.0258, -0.0385, -0.0535, -0.071, -0.0816, -0.0916, -0.0398, -0.0239, -0.0535, -0.0504, -0.0527, -0.1026, -0.051, -0.0433, -0.096, -0.051, -0.026, -0.001, -0.0579, -0.0471, -0.0473, -0.0827, -0.0625, -0.0385, -0.0866, -0.0837, -0.0827, -0.0177, -0.0818, -0.102, -0.1424, -0.0812, 0.0, 0.0489, 0.0489, 0.0573, 0.1016, 0.2274, 0.285, 0.2988, 0.3571, 0.3285, 0.3521, 0.4262, 0.4262, 0.3679, 0.3796, 0.3589, 0.3904, 0.7185, 0.7695, 0.7597, 0.9727, 0.9671, 1.0741, 0.9942, 1.2517, 1.3529, 1.2823, 1.6009, 1.695, 1.5111, 1.5045, 1.4137, 1.2646, 1.2521, 1.3069, 1.4768, 1.4672, 1.4951, 1.4951, 1.5717, 1.5353, 1.517, 1.3877, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2039315025752703328", "ticker_idx": 1, "ts": "2026-04-01T12:12:45+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "On the server side, Intel's share has dropped from 65% in 2025 to 50%, primarily due to AMD's stable supply and superior price-performance", "tweet_type": "analysis", "timeline": "months", "summary": "Shares bullish CPU channel check: INTC executing 30% cumulative price hikes, 95% utilization, surging server demand; AMD gaining server share on stable supply and price-performance.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "A very spicy piece of information here. A must-read for everyone (especially if you're interested in the CPU shortage).\n\n- Intel has carried out two rounds of CPU price hikes since the start of 2026. The first round in February was 10–15%, and the second on March 16th was 15%, bringing the cumulative increase to ~20% vs. January.\n\n- The full-year target is a cumulative 30% price increase, aimed at recouping investment and meeting capital market expectations.\n\n- A third round of price adjustments is reportedly planned for May.\n\n- Server CPU demand is strong globally, with orders already up 50% YoY vs. 2025.\n\n- Chinese CSPs in particular are placing additional orders — Tencent is scaling its server CPU procurement from ~300K units in 2025 to a 2026 target of 900K–1M units, deploying them toward AI infrastructure buildout.\n\n- Alibaba has the largest procurement scale among Chinese CSPs, and is also expected to grow 30–35% YoY.\n\n- Consumer CPU shipments are also picking up, driven by local AI PC demand, with a projected ~20% YoY increase in shipments (though I find this one a bit dubious). Intel's consumer CPU share remains stable at 65%.\n\n- Intel's production line utilization has reached 95%, but is not yet physically at full capacity. Server market deliveries are smooth (China Q1 orders +60%, deliveries +27%), but on the consumer side, despite orders doubling (+100%), deliveries actually declined 3.8%. This is due to PTL yield issues and new product supply being prioritized for North America — Intel is managing this by extending lead times to 3 months.\n\n- Intel's consumer CPU market share remains stable at 65%.\n\n- On the server side, Intel's share has dropped from 65% in 2025 to 50%, primarily due to AMD's stable supply and superior price-performance.\n\n- A particularly noteworthy point: the CPU-to-GPU ratio in AI servers has risen from 1:12 to 1:8, with a target of 1:4, meaning the share of CPU procurement is increasing. Very important.\n\n- Cloud providers' AI servers generally have excess compute capacity, and there are no plans for large-scale traditional server expansion for the time being.\n\n- AI applications are expected to drive ~10% of incremental demand for consumer CPUs, with long-term growth dependent on product usability.\n\n$INTC $AMD $ARM", "reply_text": null, "parent_text": null, "quoted_text": "CPU紀要\n\n漲價情況\nIntel 2026年已執行兩輪CPU漲價：2月首輪漲10-15%，3月16日第二輪漲15%，��計較1月上漲約20%。目標是全年累計漲30%，以回收投資並滿足資本市場預期。5月計劃第三輪調價，但受AMD與ARM競爭影響，進一步大漲空間有限。\n\n需求表現\n-", "parent_author": null, "ret_m1d": -0.03225352657599567, "ret_signed_m1d": -0.03225352657599567, "alpha_spy_m1d": -0.024775406418595658, "alpha_c_m1d": -0.010389589801422683, "bench_spy_m1d": -0.0074781201574000145, "bench_c_m1d": -0.02186393677457299, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.034679573061694224, "ret_signed_p1d": 0.034679573061694224, "alpha_spy_p1d": 0.033779098633603066, "alpha_c_p1d": 0.033786632014099816, "bench_spy_p1d": 0.0009004744280911581, "bench_c_p1d": 0.0008929410475944088, "ret_p1w": 0.10280195956497251, "ret_signed_p1w": 0.10280195956497251, "alpha_spy_p1w": 0.07110362026822004, "alpha_c_p1w": 0.02384180416878312, "bench_spy_p1w": 0.031698339296752476, "bench_c_p1w": 0.07896015539618939, "ret_p1m": 0.6863611574728612, "ret_signed_p1m": 0.6863611574728612, "alpha_spy_p1m": 0.5895722192893549, "alpha_c_p1m": 0.3936091620583224, "bench_spy_p1m": 0.09678893818350631, "bench_c_p1m": 0.2927519954145388, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0631, 0.0517, 0.0197, -0.0009, -0.0263, -0.0335, -0.012, 0.0512, 0.0637, 0.0842, 0.1028, 0.1028, 0.1033, 0.1883, 0.207, 0.2353, 0.1955, 0.1989, 0.2023, 0.1997, 0.1262, 0.1715, 0.1518, -0.0477, -0.0842, -0.0084, 0.0275, 0.016, 0.016, -0.0203, -0.0137, -0.0137, -0.0339, -0.048, -0.0325, -0.0479, -0.0647, 0.0173, 0.0031, -0.0311, -0.0476, -0.0551, -0.0916, -0.0387, -0.0512, -0.0846, -0.0358, -0.0332, -0.0256, -0.0593, -0.08, -0.0648, -0.0661, -0.0511, -0.0235, -0.0422, -0.0358, -0.023, 0.0479, -0.0306, -0.0391, -0.0674, -0.0323, 0.0, 0.0347, 0.0347, 0.0474, 0.0539, 0.1028, 0.1257, 0.1657, 0.1742, 0.2134, 0.2279, 0.3237, 0.3243, 0.308, 0.3534, 0.4436, 0.4525, 0.6546, 0.5919, 0.5376, 0.6037, 0.6864, 0.7151, 0.6248, 0.69, 1.0046, 0.9431, 1.1654, 1.1825, 1.1326, 1.1193, 1.1393, 1.0175, 1.0027, 0.9697, 1.1292, 1.1388, 1.224, 1.224, 1.3971, 1.3574, 1.4646, 1.4552, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2059568540840276017", "ticker_idx": 0, "ts": "2026-05-27T09:32:59+00:00", "symbol": "semiconductor equipment", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "I'm increasingly convinced that semiconductor equipment could become seriously scarce going forward", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish semi equipment on scarcity; SK Hynix outperforming Samsung in HBM4; HBM price renegotiations failing to lift prices.", "resolved_tickers": ["AMAT", "LRCX", "KLAC", "ASML", "8035.T"], "resolution_method": "basket", "resolution_reason": "Global semicon equipment basket: AMAT, Lam, KLA, ASML, TEL", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "Things I’ve looked into recently:\n\nI’m increasingly convinced that semiconductor equipment could become seriously scarce going forward. Based on channel checks, the “tera-fab” project appears to be much more serious than expected. Intel needs to absorb foundry customers from TSMC while also expanding capacity for its own CPUs. Samsung also seems to have recently secured multiple customers. And TSMC may raise its capex guidance in Q3 this year.\n\nSK hynix appears to be performing better than Samsung in HBM4.\n\nHBM price renegotiations do not seem to be going smoothly. Hyperscalers appear very reluctant to renegotiate the prices under the long-term agreements they signed last year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011014223219221918, "ret_signed_m1d": 0.011014223219221918, "alpha_spy_m1d": 0.01084098963048159, "alpha_c_m1d": -0.00013609525002660286, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.011150318469248521, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002662707589419977, "ret_signed_p1d": -0.002662707589419977, "alpha_spy_p1d": -0.008179262092082284, "alpha_c_p1d": -0.009933936959434763, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.007271229370014787, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1847, -0.1907, -0.2295, -0.2218, -0.2367, -0.2739, -0.2543, -0.2388, -0.2306, -0.2556, -0.2572, -0.2418, -0.2299, -0.2308, -0.2207, -0.2318, -0.2226, -0.2013, -0.205, -0.2493, -0.2589, -0.285, -0.2566, -0.2285, -0.2436, -0.242, -0.2383, -0.2337, -0.1638, -0.1425, -0.1269, -0.1264, -0.1135, -0.1288, -0.1348, -0.1225, -0.1232, -0.1234, -0.1151, -0.1223, -0.092, -0.1042, -0.1427, -0.1417, -0.1346, -0.1301, -0.1343, -0.1058, -0.0632, -0.0705, -0.0317, -0.034, -0.0525, -0.0378, -0.0295, -0.0632, -0.09, -0.107, -0.07, -0.0464, -0.0285, -0.0196, 0.011, 0.0, -0.0027, -0.0017, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2059568540840276017", "ticker_idx": 1, "ts": "2026-05-27T09:32:59+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix appears to be performing better than Samsung in HBM4", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish semi equipment on scarcity; SK Hynix outperforming Samsung in HBM4; HBM price renegotiations failing to lift prices.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Things I’ve looked into recently:\n\nI’m increasingly convinced that semiconductor equipment could become seriously scarce going forward. Based on channel checks, the “tera-fab” project appears to be much more serious than expected. Intel needs to absorb foundry customers from TSMC while also expanding capacity for its own CPUs. Samsung also seems to have recently secured multiple customers. And TSMC may raise its capex guidance in Q3 this year.\n\nSK hynix appears to be performing better than Samsung in HBM4.\n\nHBM price renegotiations do not seem to be going smoothly. Hyperscalers appear very reluctant to renegotiate the prices under the long-term agreements they signed last year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0851538375787303, "ret_signed_m1d": -0.0851538375787303, "alpha_spy_m1d": -0.08532707116747063, "alpha_c_m1d": -0.09630415604797882, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.011150318469248521, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02067889192352701, "ret_signed_p1d": 0.02067889192352701, "alpha_spy_p1d": 0.015162337420864702, "alpha_c_p1d": 0.013407662553512223, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.007271229370014787, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.527, -0.527, -0.5814, -0.6215, -0.5805, -0.5881, -0.6273, -0.5818, -0.5742, -0.5854, -0.5943, -0.5658, -0.5675, -0.5292, -0.5484, -0.551, -0.584, -0.5604, -0.5564, -0.584, -0.584, -0.6108, -0.6402, -0.6001, -0.63, -0.6095, -0.605, -0.5916, -0.5395, -0.5551, -0.5421, -0.5363, -0.5082, -0.4935, -0.4851, -0.4971, -0.4802, -0.4543, -0.4547, -0.4539, -0.4552, -0.424, -0.4204, -0.4235, -0.4267, -0.4267, -0.3549, -0.3549, -0.2862, -0.2626, -0.2483, -0.1618, -0.1819, -0.119, -0.1217, -0.189, -0.1797, -0.222, -0.222, -0.1351, -0.1346, -0.1346, -0.0852, 0.0, 0.0207, 0.0403, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2059568540840276017", "ticker_idx": 2, "ts": "2026-05-27T09:32:59+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK hynix appears to be performing better than Samsung in HBM4", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish semi equipment on scarcity; SK Hynix outperforming Samsung in HBM4; HBM price renegotiations failing to lift prices.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "Things I’ve looked into recently:\n\nI’m increasingly convinced that semiconductor equipment could become seriously scarce going forward. Based on channel checks, the “tera-fab” project appears to be much more serious than expected. Intel needs to absorb foundry customers from TSMC while also expanding capacity for its own CPUs. Samsung also seems to have recently secured multiple customers. And TSMC may raise its capex guidance in Q3 this year.\n\nSK hynix appears to be performing better than Samsung in HBM4.\n\nHBM price renegotiations do not seem to be going smoothly. Hyperscalers appear very reluctant to renegotiate the prices under the long-term agreements they signed last year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.026058631921824116, "ret_signed_m1d": 0.026058631921824116, "alpha_spy_m1d": 0.026231865510564445, "alpha_c_m1d": 0.029908367550862236, "bench_spy_m1d": 0.00017323358874032913, "bench_c_m1d": 0.00384973562903812, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02442996742671011, "ret_signed_p1d": 0.02442996742671011, "alpha_spy_p1d": 0.029946521929372416, "alpha_c_p1d": 0.037551550270280565, "bench_spy_p1d": 0.0055165545026623075, "bench_c_p1d": 0.013121582843570456, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2962, -0.2962, -0.3658, -0.4402, -0.3772, -0.3882, -0.436, -0.3892, -0.3824, -0.3892, -0.4035, -0.3866, -0.3697, -0.3222, -0.3483, -0.3518, -0.3944, -0.3834, -0.3856, -0.4146, -0.4146, -0.4257, -0.4554, -0.3822, -0.4189, -0.3935, -0.371, -0.3599, -0.3143, -0.3355, -0.329, -0.3453, -0.3274, -0.3127, -0.2915, -0.2964, -0.3013, -0.2866, -0.2915, -0.2687, -0.285, -0.2687, -0.2769, -0.2638, -0.2818, -0.2818, -0.2427, -0.2427, -0.1336, -0.1156, -0.1254, -0.07, -0.0912, -0.0749, -0.0358, -0.1189, -0.0847, -0.1026, -0.101, -0.0244, -0.0472, -0.0472, -0.0261, 0.0, -0.0244, 0.0326, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2049625970198589689", "ticker_idx": 0, "ts": "2026-04-29T23:04:46+00:00", "symbol": "QCOM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'm skeptical that Qualcomm can stage a meaningful rebound within this year.", "tweet_type": "analysis", "timeline": "months", "summary": "Skeptical QCOM rebounds this year; on-device AI thesis undermined by Apple scaling back WMCM adoption and no DRAM upgrade.", "resolved_tickers": ["QCOM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "I’m skeptical that Qualcomm can stage a meaningful rebound within this year.\n\nWhy did Apple choose to adopt WMCM packaging for the iPhone 18? It was precisely to enable the use of more DRAM for on-device AI.\n\nHowever, according to recent industry checks, even Apple appears to have walked back full-scale WMCM adoption for the iPhone 18 and decided to use it only in the higher-end lineup. Even in the high-end models that do use WMCM, there appears to be no RAM upgrade versus the previous generation.\n\nIf even Apple — the company with the strongest buying power in the industry — is facing this reality, how exactly is Qualcomm supposed to deliver on-device AI?\n\nIf Qualcomm can deliver on-device AI without additional mobile DRAM, that would imply it has achieved some extraordinary algorithmic breakthrough that bypasses the memory wall. I’m skeptical of that.", "reply_text": null, "parent_text": null, "quoted_text": "My hottest and most wrong (year to date) take is also probably the only semi stock that Zephyr and I currently disagree on. Since December I’ve been bullish QCOM and he’s been bearish. Of course, he has been hugely right and I have not been - but I really think that’s going to", "parent_author": null, "ret_m1d": -0.038461538461538436, "ret_signed_m1d": 0.038461538461538436, "alpha_spy_m1d": 0.038616103442174476, "alpha_c_m1d": 0.021707474363382717, "bench_spy_m1d": 0.00015456498063604052, "bench_c_m1d": -0.01675406409815572, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.15115385789137625, "ret_signed_p1d": -0.15115385789137625, "alpha_spy_p1d": -0.14120423046417208, "alpha_c_p1d": -0.13686182290129034, "bench_spy_p1d": 0.009949627427204177, "bench_c_p1d": 0.014292034990085911, "ret_p1w": 0.23442312387319708, "ret_signed_p1w": -0.23442312387319708, "alpha_spy_p1w": -0.20315467975499923, "alpha_c_p1w": -0.13397870147418112, "bench_spy_p1w": 0.03126844411819785, "bench_c_p1w": 0.10044442239901596, "ret_p1m": 0.5595512390136719, "ret_signed_p1m": -0.5595512390136719, "alpha_spy_p1m": -0.4990942876337934, "alpha_c_p1m": -0.3588826109418852, "bench_spy_p1m": 0.0604569513798785, "bench_c_p1m": 0.20066862807178665, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0345, -0.0279, -0.0626, -0.0517, -0.1319, -0.1252, -0.1151, -0.1077, -0.1017, -0.118, -0.1038, -0.1038, -0.0915, -0.0877, -0.1002, -0.0899, -0.1057, -0.0778, -0.0712, -0.0727, -0.0933, -0.1017, -0.1202, -0.1114, -0.1218, -0.1302, -0.1147, -0.1333, -0.1403, -0.1593, -0.1678, -0.1706, -0.1565, -0.1637, -0.1585, -0.1673, -0.1772, -0.1752, -0.1644, -0.1632, -0.1852, -0.1854, -0.1745, -0.1841, -0.1872, -0.1872, -0.194, -0.2047, -0.1826, -0.1811, -0.1791, -0.1587, -0.1485, -0.1471, -0.138, -0.1269, -0.1185, -0.131, -0.1278, -0.1413, -0.0458, -0.0368, -0.0385, 0.0, 0.1512, 0.1347, 0.0794, 0.1958, 0.2344, 0.2984, 0.4044, 0.5226, 0.3481, 0.3665, 0.2826, 0.2916, 0.3054, 0.2539, 0.2981, 0.368, 0.5267, 0.5267, 0.595, 0.4962, 0.5596, 0.6091, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044550082557841907", "ticker_idx": 0, "ts": "2026-04-15T22:55:00+00:00", "symbol": "Yageo", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The MLCC supercycle is coming", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish on Yageo and the MLCC supercycle as price hikes signal rising demand cycle ahead.", "resolved_tickers": ["2327.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Yageo Corp listed on TWSE as 2327.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "Taiwan’s Yageo mentioned for the first time on its earnings call that it has raised prices for some MLCC product lines in response to rising raw material cost pressures.\n\nThe MLCC supercycle is coming.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03105590062111796, "ret_signed_m1d": -0.03105590062111796, "alpha_spy_m1d": -0.02322668604610334, "alpha_c_m1d": -0.01535396740466377, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.01570193321645419, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.009316770186335477, "ret_signed_p1d": 0.009316770186335477, "alpha_spy_p1d": 0.006859458563731158, "alpha_c_p1d": -0.0021270168115319343, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.011443786997867411, "ret_p1w": -0.02018633540372672, "ret_signed_p1w": -0.02018633540372672, "alpha_spy_p1w": -0.03628774337003804, "alpha_c_p1w": -0.07201596366694252, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.0518296282632158, "ret_p1m": 0.4347826086956521, "ret_signed_p1m": 0.4347826086956521, "alpha_spy_p1m": 0.36587673061964154, "alpha_c_p1m": 0.24050452249902854, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.19427808619662357, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1211, -0.0823, -0.1009, -0.1382, -0.0807, -0.0947, -0.1087, -0.0839, -0.1134, -0.1398, -0.1382, -0.1988, -0.1724, -0.1708, -0.1988, -0.2283, -0.2236, -0.2112, -0.205, -0.205, -0.205, -0.205, -0.205, -0.205, -0.205, -0.205, -0.1258, -0.0652, -0.0683, -0.0745, -0.0745, -0.0932, -0.1289, -0.1957, -0.1693, -0.177, -0.2283, -0.1972, -0.1786, -0.2189, -0.1988, -0.1925, -0.1118, -0.1289, -0.146, -0.1584, -0.2143, -0.222, -0.1925, -0.1863, -0.1925, -0.1988, -0.2438, -0.2189, -0.236, -0.236, -0.236, -0.205, -0.1568, -0.1258, -0.1149, -0.0575, -0.0311, 0.0, 0.0093, -0.0171, -0.0078, 0.0031, -0.0202, -0.0839, -0.0776, -0.0528, 0.014, 0.0, -0.0155, -0.0155, 0.0637, 0.0823, 0.0497, 0.1537, 0.2081, 0.2158, 0.3028, 0.3043, 0.4348, 0.4146, 0.5559, 0.5435, 0.6149, 0.7764, 0.9534, 1.146, 1.0528, 1.177, 1.3043, 1.2919, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2055039613438402790", "ticker_idx": 0, "ts": "2026-05-14T21:36:39+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Memory Surge Cycle Set to Run Longer", "tweet_type": "analysis", "timeline": "quarters", "summary": "Strike risk at Samsung + structural supply cliff extends DRAM/NAND price peak well beyond Q3, bullish for memory producers into 2027.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Samsung Strike Fears Pour Fuel on Already-Burning Prices”… Memory Surge Cycle Set to Run Longer\n\nWith a strike at Samsung Electronics’ DS (Semiconductor) Division becoming increasingly likely, observers say the labor action could serve as a second catalyst that further inflames DRAM and NAND flash prices, which are already on a sharp uptrend. Major market research firms have previously forecast that memory contract prices will post their largest-ever quarterly increase in Q2. Analysts warn that the supply-side risk from a strike, coinciding with the cycle’s approach toward peak, could amplify the shock to the market well beyond current expectations.\n\nAccording to industry sources on the 15th, a scenario is emerging in which strike risk at Samsung Electronics’ semiconductor union delivers a second shock to this year’s memory pricing cycle. Until now, domestic and overseas market research firms and investment banks had projected that DRAM and NAND prices would peak in Q3 before gradually stabilizing from Q4 onward. The breakdown in Samsung’s labor-management negotiations, however, has introduced a new variable.\n\nAfter Samsung Electronics’ semiconductor union rejected the company’s proposal for additional negotiations the previous day, the situation has effectively moved toward strike action. Samsung’s major semiconductor customers are reportedly already taking pre-emptive steps to mitigate the risk. “Regardless of whether the strike actually happens, the current situation itself is already acting as a psychological driver pushing DRAM and NAND prices higher,” a semiconductor industry source said. “It is making Big Tech firms and PC and smartphone manufacturers — already struggling with memory shortages — even more anxious.”\n\nAccording to earlier reports from major foreign media, Samsung semiconductor customers including Apple and HP have inquired about the impact of a potential strike, while HBM (high-bandwidth memory) customers are also closely watching for developments. Samsung Electronics has been attempting to reassure key customers by stating that “there will be no supply disruptions,” but some customers remain concerned about production setbacks and further upside in DRAM and NAND prices.\n\nKim Dong-won, Head of Research at KB Securities, characterized the situation as “pouring fuel on an already-burning market.” The global memory market has entered a “structural supply cliff” phase, in which supply cannot keep up with demand amid an explosion in AI infrastructure spending. A significant portion of DRAM output from Samsung Electronics and SK hynix has been allocated to HBM for AI servers, reducing supply of commodity DRAM and NAND. On top of that, U.S. hyperscalers are now offering large prepayments to secure capacity for AI datacenter buildouts.\n\nAccording to KB Securities, if 30–40% of union members participate in the strike, the global supply disruption is estimated to reach 3–4% for DRAM and 2–3% for NAND. While these percentages may appear modest in absolute terms, experts argue that under current undersupply conditions, they could act as a decisive catalyst to drive prices sharply higher. Global DRAM inventory currently stands at only 4–6 weeks of demand.\n\nThis could extend the peak phase of memory prices well beyond previous forecasts. Major overseas investment banks and market research firms had built their supply-demand outlooks around the assumption that DRAM and NAND prices would peak in Q3 before gradually stabilizing from Q4. But if the Samsung semiconductor strike materializes, the peak could be pushed beyond Q4, potentially having an additional pass-through effect on consumer prices. TrendForce projects that the memory supply shortage will persist through 2027.\n\n---\n\nhttps://t.co/XHb1Gjs9aO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04054054054054057, "ret_signed_m1d": -0.04054054054054057, "alpha_spy_m1d": -0.032708115442745433, "alpha_c_m1d": -0.025777343345641524, "bench_spy_m1d": -0.007832425097795137, "bench_c_m1d": -0.014763197194899047, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.08614864864864868, "ret_signed_p1d": -0.08614864864864868, "alpha_spy_p1d": -0.0741192967559241, "alpha_c_p1d": -0.0680984787702974, "bench_spy_p1d": -0.012029351892724582, "bench_c_p1d": -0.018050169878351285, "ret_p1w": 0.011824324324324342, "ret_signed_p1w": 0.011824324324324342, "alpha_spy_p1w": 0.01910878150854911, "alpha_c_p1w": 0.016838217898053398, "bench_spy_p1w": -0.0072844571842247685, "bench_c_p1w": -0.005013893573729056, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3891, -0.3891, -0.3891, -0.3594, -0.3591, -0.3493, -0.3257, -0.3139, -0.265, -0.2701, -0.2701, -0.3422, -0.4194, -0.354, -0.3655, -0.4151, -0.3665, -0.3594, -0.3665, -0.3813, -0.3638, -0.3463, -0.2971, -0.324, -0.3277, -0.3719, -0.3604, -0.3628, -0.3928, -0.3928, -0.4044, -0.4351, -0.3593, -0.3973, -0.3709, -0.3476, -0.3361, -0.2889, -0.3108, -0.3041, -0.3209, -0.3024, -0.2872, -0.2652, -0.2703, -0.2753, -0.2601, -0.2652, -0.2416, -0.2584, -0.2416, -0.25, -0.2365, -0.2551, -0.2551, -0.2145, -0.2145, -0.1014, -0.0828, -0.0929, -0.0355, -0.0574, -0.0405, 0.0, -0.0861, -0.0507, -0.0693, -0.0676, 0.0118, -0.0118, -0.0118, 0.0101, 0.0372, 0.0118, 0.0709, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2055039613438402790", "ticker_idx": 1, "ts": "2026-05-14T21:36:39+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "a significant portion of DRAM output from Samsung Electronics and SK hynix has been allocated to HBM for AI servers, reducing supply of commodity DRAM and NAND", "tweet_type": "analysis", "timeline": "quarters", "summary": "Strike risk at Samsung + structural supply cliff extends DRAM/NAND price peak well beyond Q3, bullish for memory producers into 2027.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Samsung Strike Fears Pour Fuel on Already-Burning Prices”… Memory Surge Cycle Set to Run Longer\n\nWith a strike at Samsung Electronics’ DS (Semiconductor) Division becoming increasingly likely, observers say the labor action could serve as a second catalyst that further inflames DRAM and NAND flash prices, which are already on a sharp uptrend. Major market research firms have previously forecast that memory contract prices will post their largest-ever quarterly increase in Q2. Analysts warn that the supply-side risk from a strike, coinciding with the cycle’s approach toward peak, could amplify the shock to the market well beyond current expectations.\n\nAccording to industry sources on the 15th, a scenario is emerging in which strike risk at Samsung Electronics’ semiconductor union delivers a second shock to this year’s memory pricing cycle. Until now, domestic and overseas market research firms and investment banks had projected that DRAM and NAND prices would peak in Q3 before gradually stabilizing from Q4 onward. The breakdown in Samsung’s labor-management negotiations, however, has introduced a new variable.\n\nAfter Samsung Electronics’ semiconductor union rejected the company’s proposal for additional negotiations the previous day, the situation has effectively moved toward strike action. Samsung’s major semiconductor customers are reportedly already taking pre-emptive steps to mitigate the risk. “Regardless of whether the strike actually happens, the current situation itself is already acting as a psychological driver pushing DRAM and NAND prices higher,” a semiconductor industry source said. “It is making Big Tech firms and PC and smartphone manufacturers — already struggling with memory shortages — even more anxious.”\n\nAccording to earlier reports from major foreign media, Samsung semiconductor customers including Apple and HP have inquired about the impact of a potential strike, while HBM (high-bandwidth memory) customers are also closely watching for developments. Samsung Electronics has been attempting to reassure key customers by stating that “there will be no supply disruptions,” but some customers remain concerned about production setbacks and further upside in DRAM and NAND prices.\n\nKim Dong-won, Head of Research at KB Securities, characterized the situation as “pouring fuel on an already-burning market.” The global memory market has entered a “structural supply cliff” phase, in which supply cannot keep up with demand amid an explosion in AI infrastructure spending. A significant portion of DRAM output from Samsung Electronics and SK hynix has been allocated to HBM for AI servers, reducing supply of commodity DRAM and NAND. On top of that, U.S. hyperscalers are now offering large prepayments to secure capacity for AI datacenter buildouts.\n\nAccording to KB Securities, if 30–40% of union members participate in the strike, the global supply disruption is estimated to reach 3–4% for DRAM and 2–3% for NAND. While these percentages may appear modest in absolute terms, experts argue that under current undersupply conditions, they could act as a decisive catalyst to drive prices sharply higher. Global DRAM inventory currently stands at only 4–6 weeks of demand.\n\nThis could extend the peak phase of memory prices well beyond previous forecasts. Major overseas investment banks and market research firms had built their supply-demand outlooks around the assumption that DRAM and NAND prices would peak in Q3 before gradually stabilizing from Q4. But if the Samsung semiconductor strike materializes, the peak could be pushed beyond Q4, potentially having an additional pass-through effect on consumer prices. TrendForce projects that the memory supply shortage will persist through 2027.\n\n---\n\nhttps://t.co/XHb1Gjs9aO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.00304568688990825, "ret_signed_m1d": 0.00304568688990825, "alpha_spy_m1d": 0.010878111987703387, "alpha_c_m1d": 0.013212724635476514, "bench_spy_m1d": -0.007832425097795137, "bench_c_m1d": -0.010167037745568264, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07664974442110084, "ret_signed_p1d": -0.07664974442110084, "alpha_spy_p1d": -0.06462039252837626, "alpha_c_p1d": -0.03860983886654723, "bench_spy_p1d": -0.012029351892724582, "bench_c_p1d": -0.03803990555455361, "ret_p1w": -0.015228434449541584, "ret_signed_p1w": -0.015228434449541584, "alpha_spy_p1w": -0.007943977265316815, "alpha_c_p1w": 0.002857849547658331, "bench_spy_p1w": -0.0072844571842247685, "bench_c_p1w": -0.018086283997199915, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5541, -0.5541, -0.5541, -0.547, -0.5192, -0.5181, -0.4908, -0.4842, -0.4421, -0.4614, -0.4614, -0.5234, -0.569, -0.5223, -0.531, -0.5756, -0.5239, -0.5152, -0.5279, -0.5381, -0.5056, -0.5076, -0.464, -0.4858, -0.4888, -0.5264, -0.4995, -0.4949, -0.5264, -0.5264, -0.5569, -0.5904, -0.5447, -0.5787, -0.5553, -0.5503, -0.535, -0.4756, -0.4934, -0.4787, -0.4721, -0.4401, -0.4234, -0.4137, -0.4274, -0.4081, -0.3787, -0.3792, -0.3782, -0.3797, -0.3442, -0.3401, -0.3437, -0.3472, -0.3472, -0.2655, -0.2655, -0.1873, -0.1604, -0.1442, -0.0457, -0.0685, 0.003, 0.0, -0.0766, -0.066, -0.1142, -0.1142, -0.0152, -0.0147, -0.0147, 0.0416, 0.1386, 0.1621, 0.1845, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2055039613438402790", "ticker_idx": 2, "ts": "2026-05-14T21:36:39+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "The global memory market has entered a 'structural supply cliff' phase, in which supply cannot keep up with demand amid an explosion in AI infrastructure spending", "tweet_type": "analysis", "timeline": "quarters", "summary": "Strike risk at Samsung + structural supply cliff extends DRAM/NAND price peak well beyond Q3, bullish for memory producers into 2027.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "“Samsung Strike Fears Pour Fuel on Already-Burning Prices”… Memory Surge Cycle Set to Run Longer\n\nWith a strike at Samsung Electronics’ DS (Semiconductor) Division becoming increasingly likely, observers say the labor action could serve as a second catalyst that further inflames DRAM and NAND flash prices, which are already on a sharp uptrend. Major market research firms have previously forecast that memory contract prices will post their largest-ever quarterly increase in Q2. Analysts warn that the supply-side risk from a strike, coinciding with the cycle’s approach toward peak, could amplify the shock to the market well beyond current expectations.\n\nAccording to industry sources on the 15th, a scenario is emerging in which strike risk at Samsung Electronics’ semiconductor union delivers a second shock to this year’s memory pricing cycle. Until now, domestic and overseas market research firms and investment banks had projected that DRAM and NAND prices would peak in Q3 before gradually stabilizing from Q4 onward. The breakdown in Samsung’s labor-management negotiations, however, has introduced a new variable.\n\nAfter Samsung Electronics’ semiconductor union rejected the company’s proposal for additional negotiations the previous day, the situation has effectively moved toward strike action. Samsung’s major semiconductor customers are reportedly already taking pre-emptive steps to mitigate the risk. “Regardless of whether the strike actually happens, the current situation itself is already acting as a psychological driver pushing DRAM and NAND prices higher,” a semiconductor industry source said. “It is making Big Tech firms and PC and smartphone manufacturers — already struggling with memory shortages — even more anxious.”\n\nAccording to earlier reports from major foreign media, Samsung semiconductor customers including Apple and HP have inquired about the impact of a potential strike, while HBM (high-bandwidth memory) customers are also closely watching for developments. Samsung Electronics has been attempting to reassure key customers by stating that “there will be no supply disruptions,” but some customers remain concerned about production setbacks and further upside in DRAM and NAND prices.\n\nKim Dong-won, Head of Research at KB Securities, characterized the situation as “pouring fuel on an already-burning market.” The global memory market has entered a “structural supply cliff” phase, in which supply cannot keep up with demand amid an explosion in AI infrastructure spending. A significant portion of DRAM output from Samsung Electronics and SK hynix has been allocated to HBM for AI servers, reducing supply of commodity DRAM and NAND. On top of that, U.S. hyperscalers are now offering large prepayments to secure capacity for AI datacenter buildouts.\n\nAccording to KB Securities, if 30–40% of union members participate in the strike, the global supply disruption is estimated to reach 3–4% for DRAM and 2–3% for NAND. While these percentages may appear modest in absolute terms, experts argue that under current undersupply conditions, they could act as a decisive catalyst to drive prices sharply higher. Global DRAM inventory currently stands at only 4–6 weeks of demand.\n\nThis could extend the peak phase of memory prices well beyond previous forecasts. Major overseas investment banks and market research firms had built their supply-demand outlooks around the assumption that DRAM and NAND prices would peak in Q3 before gradually stabilizing from Q4. But if the Samsung semiconductor strike materializes, the peak could be pushed beyond Q4, potentially having an additional pass-through effect on consumer prices. TrendForce projects that the memory supply shortage will persist through 2027.\n\n---\n\nhttps://t.co/XHb1Gjs9aO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.03559231810106356, "ret_signed_m1d": 0.03559231810106356, "alpha_spy_m1d": 0.043424743198858695, "alpha_c_m1d": 0.04575935584663182, "bench_spy_m1d": -0.007832425097795137, "bench_c_m1d": -0.010167037745568264, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06617187404142222, "ret_signed_p1d": -0.06617187404142222, "alpha_spy_p1d": -0.054142522148697636, "alpha_c_p1d": -0.028131968486868608, "bench_spy_p1d": -0.012029351892724582, "bench_c_p1d": -0.03803990555455361, "ret_p1w": -0.017925070559191236, "ret_signed_p1w": -0.017925070559191236, "alpha_spy_p1w": -0.010640613374966468, "alpha_c_p1w": 0.00016121343800867827, "bench_spy_p1w": -0.0072844571842247685, "bench_c_p1w": -0.018086283997199915, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4697, -0.485, -0.4578, -0.4624, -0.4485, -0.4577, -0.4616, -0.4474, -0.4647, -0.4688, -0.4684, -0.5109, -0.4838, -0.4886, -0.523, -0.4985, -0.4808, -0.4607, -0.4779, -0.4511, -0.4309, -0.4053, -0.4052, -0.4277, -0.4553, -0.4792, -0.4905, -0.5078, -0.5421, -0.5399, -0.5853, -0.5646, -0.526, -0.528, -0.528, -0.5132, -0.5134, -0.4759, -0.4568, -0.458, -0.4503, -0.3999, -0.4121, -0.4108, -0.4136, -0.4221, -0.4209, -0.3718, -0.3792, -0.3599, -0.324, -0.3502, -0.3319, -0.3336, -0.3013, -0.2572, -0.175, -0.141, -0.1667, -0.0376, 0.0249, -0.0122, 0.0356, 0.0, -0.0662, -0.1217, -0.0996, -0.0567, -0.0179, -0.0322, -0.0322, 0.1545, 0.1964, 0.1901, 0.2513, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2048234037433770210", "ticker_idx": 0, "ts": "2026-04-26T02:53:43+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel (INTC US, Buy): Earnings Beat — From Recovery to Strength ☀️ Price target raised to $94.2", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares GF Securities Buy on INTC with PT raised to $94.2; beats add AI6 order win catalyst.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "According to Jeff Pu's estimates, Intel is expected to win a portion of the AI6 orders that were previously anticipated to be awarded exclusively to Samsung. https://t.co/cRM1i5BO2q", "reply_text": null, "parent_text": null, "quoted_text": "《GF Overseas Electronics &amp; Telecom》 \n☄️ Intel (INTC US, Buy): Earnings Beat — From Recovery to Strength\n\n☀️ Price target raised to $94.2: Despite already elevated market expectations, Intel's 1Q results still meaningfully beat, even though 2Q guidance was somewhat conservative.", "parent_author": null, "ret_m1d": -0.028826885631522026, "ret_signed_m1d": -0.028826885631522026, "alpha_spy_m1d": -0.027107042100414924, "alpha_c_m1d": -0.029182419689731787, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005530076862179545, "ret_signed_p1d": -0.005530076862179545, "alpha_spy_p1d": -0.0006641281340333816, "alpha_c_p1d": 0.02419776657841599, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.12695612644705134, "ret_signed_p1w": 0.12695612644705134, "alpha_spy_p1w": 0.12298500500273102, "alpha_c_p1w": 0.1259092359775824, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.4533474496735088, "ret_signed_p1m": 0.4533474496735088, "alpha_spy_p1m": 0.403820701079864, "alpha_c_p1m": 0.26395859210365935, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4261, -0.4275, -0.4532, -0.4257, -0.4205, -0.4282, -0.4324, -0.4048, -0.4089, -0.4455, -0.4318, -0.4531, -0.4495, -0.4495, -0.4566, -0.4651, -0.475, -0.481, -0.4866, -0.4573, -0.4484, -0.4651, -0.4633, -0.4646, -0.4929, -0.4637, -0.4593, -0.4891, -0.4637, -0.4496, -0.4355, -0.4676, -0.4615, -0.4616, -0.4816, -0.4702, -0.4566, -0.4838, -0.4822, -0.4816, -0.4449, -0.4811, -0.4925, -0.5154, -0.4808, -0.4349, -0.4072, -0.4072, -0.4025, -0.3775, -0.3064, -0.2738, -0.266, -0.2331, -0.2492, -0.2359, -0.194, -0.194, -0.227, -0.2204, -0.232, -0.2143, -0.0288, 0.0, -0.0055, 0.1148, 0.1117, 0.1721, 0.127, 0.2725, 0.3297, 0.2898, 0.4698, 0.523, 0.4191, 0.4153, 0.364, 0.2798, 0.2727, 0.3037, 0.3997, 0.3943, 0.41, 0.41, 0.4533, 0.4328, 0.4224, 0.3493, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048233256961876087", "ticker_idx": 0, "ts": "2026-04-26T02:50:37+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Price target raised to $94.2 … we believe Intel's core CPU business will remain strong … we revise our 2026E/2027E/2028E EPS forecasts to $1.5/$2.4/$3.1, and raise our price target from $78.5 to $94.2", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF reiterates Buy on INTC, raises PT to $94.2; bullish on CPU strength, AI inference tailwinds, foundry 18A execution, and double-digit server CPU shipment growth through 2027.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "《GF Overseas Electronics & Telecom》 \n☄️ Intel (INTC US, Buy): Earnings Beat — From Recovery to Strength\n\n☀️ Price target raised to $94.2: Despite already elevated market expectations, Intel's 1Q results still meaningfully beat, even though 2Q guidance was somewhat conservative. Many in the market view this earnings call as carrying far greater constructive significance than Intel's routine updates. In fact, several brokers have already followed up with rating upgrades, as the narrative shifts from a \"turnaround story\" to \"AI beneficiary + external foundry option.\" Overall, we believe Intel's core CPU business will remain strong, with the key watch points being the tightness of substrate and silicon capacitor supply, as well as order wins and execution in the foundry business — and as our prior updates have noted, both are progressing well. Reflecting the results, guidance, and higher margins, we revise our 2026E/2027E/2028E EPS forecasts to $1.5/$2.4/$3.1, and raise our price target from $78.5 to $94.2, based on 3.5x 2027 P/B.\n\n☀️ All-around beat: 1Q revenue of $13.6B came in $1.4B above the guidance midpoint and above consensus of $12.4B. Non-GAAP gross margin was 41%, above both guidance and consensus of 35%, with EPS of $0.29 (guidance: breakeven). Operating cash flow was $1.1B, while adjusted free cash flow was -$2.0B. 2Q guidance: revenue $13.8-14.8B (+2% to +9% QoQ) vs. consensus of $13.0B; Non-GAAP gross margin of 39%, reflecting a higher Panther Lake mix; EPS guidance of $0.20, with DCAI expected to deliver double-digit QoQ growth.\n\n☀️ Foundry on track: As we have repeatedly emphasized since our initiation report in July 2025 and our pre-earnings preview on April 16, we remain bullish on strong customer engagement from Apple, NVIDIA, and AMD on 18A-P (primarily 14A), and we believe Intel will secure a portion of Tesla's AI6 program on 14A by end-2028. On execution, management indicated that 18A yields are running better than internal expectations. External foundry revenue remains small at $174M, with an operating loss of $2.4B, though improving by $72M QoQ. Management expects losses to continue narrowing through the year.\n\n☀️ CPU continues to strengthen: Intel argues that AI workloads are shifting from training toward inference, agentic AI, robotics, physical AI, and edge AI, making the CPU increasingly important as the orchestration layer of the AI stack. Management noted that server CPU demand has improved materially over the past 90 days, and expects both the industry and Intel to deliver double-digit shipment growth in 2026, with the momentum extending into 2027. As flagged in our report, the 2Q26 CPU price hike was already within expectations, and we expect another 5-10% price increase by the end of 3Q. We now forecast DCAI to grow 39%/15% YoY in 2026E/2027E. \n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.028826885631522026, "ret_signed_m1d": -0.028826885631522026, "alpha_spy_m1d": -0.027107042100414924, "alpha_c_m1d": -0.029182419689731787, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.005530076862179545, "ret_signed_p1d": -0.005530076862179545, "alpha_spy_p1d": -0.0006641281340333816, "alpha_c_p1d": 0.02419776657841599, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.12695612644705134, "ret_signed_p1w": 0.12695612644705134, "alpha_spy_p1w": 0.12298500500273102, "alpha_c_p1w": 0.1259092359775824, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.4533474496735088, "ret_signed_p1m": 0.4533474496735088, "alpha_spy_p1m": 0.403820701079864, "alpha_c_p1m": 0.26395859210365935, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4261, -0.4275, -0.4532, -0.4257, -0.4205, -0.4282, -0.4324, -0.4048, -0.4089, -0.4455, -0.4318, -0.4531, -0.4495, -0.4495, -0.4566, -0.4651, -0.475, -0.481, -0.4866, -0.4573, -0.4484, -0.4651, -0.4633, -0.4646, -0.4929, -0.4637, -0.4593, -0.4891, -0.4637, -0.4496, -0.4355, -0.4676, -0.4615, -0.4616, -0.4816, -0.4702, -0.4566, -0.4838, -0.4822, -0.4816, -0.4449, -0.4811, -0.4925, -0.5154, -0.4808, -0.4349, -0.4072, -0.4072, -0.4025, -0.3775, -0.3064, -0.2738, -0.266, -0.2331, -0.2492, -0.2359, -0.194, -0.194, -0.227, -0.2204, -0.232, -0.2143, -0.0288, 0.0, -0.0055, 0.1148, 0.1117, 0.1721, 0.127, 0.2725, 0.3297, 0.2898, 0.4698, 0.523, 0.4191, 0.4153, 0.364, 0.2798, 0.2727, 0.3037, 0.3997, 0.3943, 0.41, 0.41, 0.4533, 0.4328, 0.4224, 0.3493, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2041774750448623618", "ticker_idx": 0, "ts": "2026-04-08T07:06:49+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We raise our blended HBM ASP assumption for SK Hynix in 2027 to +36% YoY from +6% prior, with further upside possible", "tweet_type": "news_reaction", "timeline": "years", "summary": "UBS relayed: SK Hynix leads HBM4/Rubin at 60% share, raises 2027 HBM ASP to +36% YoY with further upside; bullish.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "UBS: On SK Hynix's HBM4 Redesign\n\nSK Hynix is understood to be close to completing a partial redesign of the logic and DRAM die for HBM4, working towards final qualification for Nvidia/Rubin. We adjust our HBM4/Rubin market share assumptions to 60% for SK Hynix, 30% for Samsung, and 10% for Micron. We slightly lower our SK Hynix HBM bit shipment forecast to 18.4bn Gb in 2026 (from 19.0bn) and 24.7bn Gb in 2027 (from 26.3bn), reflecting partial capacity allocation toward LPDDR5X. This mix shift leads us to raise overall DRAM bit shipment growth to 25.7% YoY in 2026, up from 22.5% previously. We believe HBM suppliers will remain focused on rebuilding the HBM pricing premium over DDR into 2027. We raise our blended HBM ASP assumption for SK Hynix in 2027 to +36% YoY from +6% prior, with further upside possible.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.11326233929353025, "ret_signed_m1d": -0.11326233929353025, "alpha_spy_m1d": -0.08842537117182869, "alpha_c_m1d": -0.05883119675980497, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.054431142533725274, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.03388193464335787, "ret_signed_p1d": -0.03388193464335787, "alpha_spy_p1d": -0.03965102581444624, "alpha_c_p1d": -0.05135565046721813, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.01747371582386026, "ret_p1w": 0.09970961384699062, "ret_signed_p1w": 0.09970961384699062, "alpha_spy_p1w": 0.06431074056096064, "alpha_c_p1w": 0.02858508529023207, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.07112452855675855, "ret_p1m": 0.6011616656390466, "ret_signed_p1m": 0.6011616656390466, "alpha_spy_p1m": 0.5189587299998464, "alpha_c_p1m": 0.3240880856765651, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.2770735799624815, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2811, -0.2763, -0.2869, -0.283, -0.2763, -0.2695, -0.2618, -0.2821, -0.285, -0.2705, -0.2589, -0.2888, -0.227, -0.1874, -0.168, -0.1217, -0.198, -0.1236, -0.1304, -0.1864, -0.1893, -0.1429, -0.1535, -0.169, -0.142, -0.1497, -0.1497, -0.1497, -0.1497, -0.1362, -0.083, -0.0811, -0.0289, -0.0163, 0.0639, 0.0271, 0.0271, -0.091, -0.1781, -0.0891, -0.1055, -0.1907, -0.092, -0.0755, -0.0997, -0.1191, -0.0571, -0.061, 0.0223, -0.0194, -0.0252, -0.0968, -0.0455, -0.0368, -0.0968, -0.0968, -0.1549, -0.2188, -0.1317, -0.1965, -0.152, -0.1423, -0.1133, 0.0, -0.0339, -0.0058, 0.0068, 0.0678, 0.0997, 0.1181, 0.092, 0.1288, 0.1849, 0.1839, 0.1859, 0.183, 0.2507, 0.2585, 0.2517, 0.2449, 0.2449, 0.4008, 0.4008, 0.5499, 0.6012, 0.6321, 0.8199, 0.7764, 0.9129, 0.9071, 0.7609, 0.7812, 0.6893, 0.6893, 0.878, 0.879, 0.879, 0.9864, 1.1713, 1.2162, 1.2588, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2049459751659200798", "ticker_idx": 0, "ts": "2026-04-29T12:04:16+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we see a very favorable setup for AMD and expect revenue to be guided up at least $1B Q/Q to the low $11B range (vs. Street consensus of ~$10.4B)", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS AMD preview: above-consensus revenue guide, 80% server CPU growth, competitive advantage through 2026; INTC undershipping implies share loss.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citing UBS's AMD earnings preview:\n\n- The market is clearly aware that INTC's guidance reads very favorably for AMD.\n- This is particularly true for server CPU, given commentary implying that INTC is undershipping the market by roughly 20%.\n- The key question here is AMD's supply, but our field work throughout the quarter indicated that supply of INTC parts has been far more constrained than that of AMD parts.\n- We therefore see a very favorable setup for AMD and expect revenue to be guided up at least $1B Q/Q to the low $11B range (vs. Street consensus of ~$10.4B).\n- Intel's CY2026 outlook implies its Data Center & AI segment growing ~40% Y/Y, and we now see AMD server CPU growing as much as 80% this year, with units up ~40–45% and pricing up around 20%.\n- From a competitive standpoint, our checks remain constructive. AMD's CPU portfolio continues to compare favorably with Intel's offerings, and the lack of meaningful timeline updates for Diamond Rapids and Coral Rapids reinforces our view that AMD should maintain a competitive advantage across the x86 ecosystem through C2026.\n\n$INTC $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04123281569956605, "ret_signed_m1d": -0.04123281569956605, "alpha_spy_m1d": -0.04138738068020209, "alpha_c_m1d": -0.02447875160141033, "bench_spy_m1d": 0.00015456498063604052, "bench_c_m1d": -0.01675406409815572, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.051555888695161034, "ret_signed_p1d": 0.051555888695161034, "alpha_spy_p1d": 0.04160626126795686, "alpha_c_p1d": 0.03726385370507512, "bench_spy_p1d": 0.009949627427204177, "bench_c_p1d": 0.014292034990085911, "ret_p1w": 0.2500075137465352, "ret_signed_p1w": 0.2500075137465352, "alpha_spy_p1w": 0.21873906962833733, "alpha_c_p1w": 0.14956309134751922, "bench_spy_p1w": 0.03126844411819785, "bench_c_p1w": 0.10044442239901596, "ret_p1m": 0.5368575520394843, "ret_signed_p1m": 0.5368575520394843, "alpha_spy_p1m": 0.47640060065960577, "alpha_c_p1m": 0.3361889239676976, "bench_spy_p1m": 0.0604569513798785, "bench_c_p1m": 0.20066862807178665, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2978, -0.2695, -0.2818, -0.4062, -0.429, -0.3817, -0.3593, -0.3665, -0.3664, -0.3891, -0.385, -0.385, -0.3976, -0.4064, -0.3967, -0.4063, -0.4168, -0.3657, -0.3745, -0.3958, -0.4061, -0.4108, -0.4336, -0.4006, -0.4084, -0.4292, -0.3988, -0.3971, -0.3924, -0.4134, -0.4263, -0.4169, -0.4177, -0.4083, -0.3911, -0.4028, -0.3988, -0.3908, -0.3466, -0.3955, -0.4008, -0.4185, -0.3965, -0.3764, -0.3548, -0.3548, -0.3469, -0.3429, -0.3123, -0.298, -0.2731, -0.2678, -0.2434, -0.2343, -0.1746, -0.1742, -0.1844, -0.1561, -0.0998, -0.0943, 0.0317, -0.0074, -0.0412, 0.0, 0.0516, 0.0695, 0.0131, 0.0538, 0.25, 0.2117, 0.3503, 0.361, 0.3298, 0.3215, 0.334, 0.258, 0.2488, 0.2282, 0.3277, 0.3337, 0.3868, 0.3868, 0.4947, 0.47, 0.5369, 0.531, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2049459751659200798", "ticker_idx": 1, "ts": "2026-04-29T12:04:16+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "INTC is undershipping the market by roughly 20%", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares UBS AMD preview: above-consensus revenue guide, 80% server CPU growth, competitive advantage through 2026; INTC undershipping implies share loss.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Citing UBS's AMD earnings preview:\n\n- The market is clearly aware that INTC's guidance reads very favorably for AMD.\n- This is particularly true for server CPU, given commentary implying that INTC is undershipping the market by roughly 20%.\n- The key question here is AMD's supply, but our field work throughout the quarter indicated that supply of INTC parts has been far more constrained than that of AMD parts.\n- We therefore see a very favorable setup for AMD and expect revenue to be guided up at least $1B Q/Q to the low $11B range (vs. Street consensus of ~$10.4B).\n- Intel's CY2026 outlook implies its Data Center & AI segment growing ~40% Y/Y, and we now see AMD server CPU growing as much as 80% this year, with units up ~40–45% and pricing up around 20%.\n- From a competitive standpoint, our checks remain constructive. AMD's CPU portfolio continues to compare favorably with Intel's offerings, and the lack of meaningful timeline updates for Diamond Rapids and Coral Rapids reinforces our view that AMD should maintain a competitive advantage across the x86 ecosystem through C2026.\n\n$INTC $AMD", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.10796837316024899, "ret_signed_m1d": 0.10796837316024899, "alpha_spy_m1d": 0.10812293814088503, "alpha_c_m1d": 0.09121430906209327, "bench_spy_m1d": 0.00015456498063604052, "bench_c_m1d": -0.01675406409815572, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0028495687922576174, "ret_signed_p1d": 0.0028495687922576174, "alpha_spy_p1d": 0.012799196219461795, "alpha_c_p1d": 0.01714160378234353, "bench_spy_p1d": 0.009949627427204177, "bench_c_p1d": 0.014292034990085911, "ret_p1w": 0.19271770064623195, "ret_signed_p1w": -0.19271770064623195, "alpha_spy_p1w": -0.1614492565280341, "alpha_c_p1w": -0.09227327824721598, "bench_spy_p1w": 0.03126844411819785, "bench_c_p1w": 0.10044442239901596, "ret_p1m": 0.27588389857148754, "ret_signed_p1m": -0.27588389857148754, "alpha_spy_p1m": -0.21542694719160904, "alpha_c_p1m": -0.07521527049970089, "bench_spy_p1m": 0.0604569513798785, "bench_c_p1m": 0.20066862807178665, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5096, -0.4849, -0.4802, -0.4871, -0.4909, -0.4661, -0.4698, -0.5026, -0.4903, -0.5094, -0.5062, -0.5062, -0.5126, -0.5202, -0.5291, -0.5345, -0.5395, -0.5132, -0.5052, -0.5202, -0.5186, -0.5198, -0.5451, -0.5189, -0.515, -0.5417, -0.5189, -0.5063, -0.4936, -0.5224, -0.5169, -0.517, -0.535, -0.5247, -0.5126, -0.537, -0.5355, -0.535, -0.5021, -0.5346, -0.5448, -0.5653, -0.5342, -0.4931, -0.4683, -0.4683, -0.4641, -0.4416, -0.3778, -0.3486, -0.3416, -0.3121, -0.3265, -0.3146, -0.277, -0.277, -0.3066, -0.3007, -0.3111, -0.2952, -0.1289, -0.103, -0.108, 0.0, -0.0028, 0.0514, 0.0109, 0.1414, 0.1927, 0.1569, 0.3184, 0.3661, 0.2729, 0.2696, 0.2235, 0.148, 0.1416, 0.1694, 0.2555, 0.2507, 0.2648, 0.2648, 0.3036, 0.2852, 0.2759, 0.2103, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047438876445446619", "ticker_idx": 0, "ts": "2026-04-23T22:14:02+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Intel is insane… This is a massive earnings surprise.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish INTC on massive earnings surprise.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Intel is insane… This is a massive earnings surprise.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.022611592749812948, "ret_signed_m1d": -0.022611592749812948, "alpha_spy_m1d": -0.026507435389474532, "alpha_c_m1d": -0.012193372416385073, "bench_spy_m1d": 0.003895842639661584, "bench_c_m1d": -0.010418220333427874, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.2359988383395475, "ret_signed_p1d": 0.2359988383395475, "alpha_spy_p1d": 0.22824954037952327, "alpha_c_p1d": 0.18496636759883334, "bench_spy_p1d": 0.007749297960024215, "bench_c_p1d": 0.05103247074071415, "ret_p1w": 0.41479492488736414, "ret_signed_p1w": 0.41479492488736414, "alpha_spy_p1w": 0.4003832363256081, "alpha_c_p1w": 0.36318136298601855, "bench_spy_p1w": 0.014411688561756009, "bench_c_p1w": 0.051613561901345584, "ret_p1m": 0.7945492442123765, "ret_signed_p1m": 0.7945492442123765, "alpha_spy_p1m": 0.7420543584750632, "alpha_c_p1m": 0.5984923799930821, "bench_spy_p1m": 0.05249488573731331, "bench_c_p1m": 0.19605686421929436, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3637, -0.3422, -0.2695, -0.2713, -0.3041, -0.2691, -0.2625, -0.2722, -0.2776, -0.2424, -0.2477, -0.2942, -0.2769, -0.304, -0.2993, -0.2993, -0.3085, -0.3193, -0.3318, -0.3395, -0.3467, -0.3094, -0.298, -0.3193, -0.317, -0.3187, -0.3546, -0.3175, -0.3119, -0.3498, -0.3175, -0.2995, -0.2815, -0.3224, -0.3146, -0.3148, -0.3402, -0.3257, -0.3085, -0.3431, -0.341, -0.3402, -0.2935, -0.3396, -0.3541, -0.3832, -0.3392, -0.2808, -0.2456, -0.2456, -0.2396, -0.2077, -0.1173, -0.0758, -0.0659, -0.024, -0.0445, -0.0276, 0.0258, 0.0258, -0.0162, -0.0078, -0.0226, 0.0, 0.236, 0.2727, 0.2656, 0.4188, 0.4148, 0.4918, 0.4343, 0.6195, 0.6923, 0.6415, 0.8706, 0.9383, 0.8061, 0.8013, 0.736, 0.6288, 0.6198, 0.6592, 0.7814, 0.7745, 0.7945, 0.7945, 0.8497, 0.8235, 0.8103, 0.7173, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2040676339775045957", "ticker_idx": 0, "ts": "2026-04-05T06:22:07+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "If Samsung were to exit the appliance business, however, the consequences would not end at simply selling fewer TVs or refrigerators. The very sales foundation of Samsung Galaxy could weaken. And a decline in Galaxy sales would, in turn, mean reduced sales of Samsung's core components that go into those devices.", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bearish Samsung: appliance exit risks weakening Galaxy distribution, core component sales, and OLED/memory demand as Chinese rivals fill the void.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "quote", "lang": "en", "tweet_text": "Korean home appliances are highly likely to cede market leadership to China before long.\n\nWhat concerns me more is that this is not simply a matter of home appliances alone. Starting with appliances, this is a broader trend in which dominance across consumer electronics as a whole—including smartphones and laptops—shifts entirely to China.\n\nFor Samsung, home appliances are not inherently a high-margin business. Yet the reason Samsung has continued to maintain this low-profitability segment is clear: it secures offline retail presence and distribution networks, through which Samsung can cross-sell its broader ecosystem products, including Galaxy devices. In other words, home appliances have served not merely as a standalone business unit, but as a distribution anchor underpinning Samsung's entire consumer-facing business.\n\nIf Samsung were to exit the appliance business, however, the consequences would not end at simply selling fewer TVs or refrigerators. The very sales foundation of Samsung Galaxy could weaken. And a decline in Galaxy sales would, in turn, mean reduced sales of Samsung's core components that go into those devices.\n\nThe bigger problem lies further downstream. If Chinese players come to dominate consumer electronics markets such as laptops and smartphones, will they continue to actively adopt Samsung's OLEDs and memory? Memory demand may be partially sustained for a while, but even that is likely to face growing pressure for substitution with Chinese-made alternatives over the long term. In particular, as Chinese memory players like CXMT scale up, it is difficult to rule out the possibility that the resulting demand shifts will benefit local Chinese suppliers rather than Samsung.\n\nThe concerns are very real.", "reply_text": null, "parent_text": null, "quoted_text": "Chinese media report that Samsung has almost certainly decided to withdraw its home appliance business from China, and that its monitor business could also exit the market.\n\nThey also say that Samsung may eventually retain only its smartphone and memory businesses in China, while https://t.co/Ujwkww5u5a", "parent_author": null, "ret_m1d": -0.03573278094251686, "ret_signed_m1d": 0.03573278094251686, "alpha_spy_m1d": 0.031028221094893116, "alpha_c_m1d": 0.029957131774632884, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.005775649167883978, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017607457276022753, "ret_signed_p1d": -0.017607457276022753, "alpha_spy_p1d": -0.017167383170173967, "alpha_c_p1d": -0.012855345108415683, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.00475211216760707, "ret_p1w": 0.04091144484722942, "ret_signed_p1w": -0.04091144484722942, "alpha_spy_p1w": 0.0003220446162330859, "alpha_c_p1w": 0.02364479078560211, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.06455623563283153, "ret_p1m": 0.20403935784567584, "ret_signed_p1m": -0.20403935784567584, "alpha_spy_p1m": -0.10563736131161838, "alpha_c_p1m": 0.0068833381697475104, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.21092269601542335, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2713, -0.2827, -0.2817, -0.2827, -0.2889, -0.2749, -0.2563, -0.2305, -0.2284, -0.2496, -0.2274, -0.2129, -0.214, -0.214, -0.1757, -0.1607, -0.1695, -0.1705, -0.2227, -0.1344, -0.1261, -0.1767, -0.1804, -0.1401, -0.1432, -0.1328, -0.077, -0.0636, -0.0636, -0.0636, -0.0636, -0.0181, -0.0176, -0.0026, 0.0336, 0.0517, 0.1266, 0.1189, 0.1189, 0.0083, -0.1101, -0.0098, -0.0274, -0.1034, -0.0289, -0.0181, -0.0289, -0.0517, -0.0248, 0.0021, 0.0775, 0.0362, 0.0305, -0.0372, -0.0196, -0.0233, -0.0692, -0.0692, -0.087, -0.1341, -0.0179, -0.0761, -0.0357, 0.0, 0.0176, 0.0901, 0.0564, 0.0668, 0.0409, 0.0694, 0.0927, 0.1264, 0.1186, 0.1108, 0.1341, 0.1264, 0.1626, 0.1367, 0.1626, 0.1497, 0.1704, 0.1419, 0.1419, 0.204, 0.204, 0.3775, 0.406, 0.3905, 0.4785, 0.4448, 0.4707, 0.5329, 0.4008, 0.4552, 0.4267, 0.4293, 0.551, 0.5148, 0.5148, 0.5484, 0.5898, 0.551, 0.6416, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2049847048690938030", "ticker_idx": 0, "ts": "2026-04-30T13:43:15+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "memory purchases by hyperscalers locked in the AI arms race will continue", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Shares KIS thesis: sustained AI-driven demand for DRAM and NAND; memory ASP premium justified by system-level cost-per-token economics, supporting MU and SNDK.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS, which I consider one of the top three Korean sell-side firms, published a memory report, and there’s a comment from it that I wanted to share with you all:\n\nMemory is the critical variable that determines GPU utilization. In particular, when HBM and DRAM capacity is insufficient, memory bottlenecks prevent GPUs from being fully utilized, leading to a decline in overall system efficiency. Conversely, expanding memory capacity raises GPU utilization, allowing the same GPU resources to process a greater number of tokens. This translates directly into a lower cost per token. As cost per token falls, more users are drawn in, ultimately driving a larger expansion in inference demand. This is precisely why hyperscalers, despite rising memory ASPs, are willing to go as far as proposing long-term supply agreements in order to secure greater memory allocation. The cost of purchasing memory rises, but at the same time, that purchase improves GPU utilization—lifting overall system efficiency and lowering cost per unit of performance.\n\nWith DRAM prices having spiked roughly 3x year-over-year in a short span, the market is now bracing for a subsequent decline in demand or prices. But this view stems from looking at memory purely as a standalone cost line item. Even if DRAM is purchased at 3x the price, if that spending allows more GPUs to run and more tokens to be processed, system-wide profitability can actually improve. The fate of the chips that once tried to challenge NVIDIA GPUs by leaning on MLPerf benchmarks—touting per-chip performance-per-watt and price-performance comparisons—illustrates this well. As Jensen Huang has repeatedly emphasized, comparing single-chip specs in isolation is meaningless. Huang has stated that NVIDIA GPUs deliver the lowest cost per token in the world, adding that the lowest cost per token and the highest performance per watt are the decisive metrics of AI economics. In other words, his consistent argument is that competition should not be waged on GPU hardware specs alone, but rather on cost per token at the full-system level. The same logic extends to memory. If \"buying memory at a premium still pays off at the system level,\" then memory purchases by hyperscalers locked in the AI arms race will continue.\n\nThe same logic applies to NAND. Active efforts are underway to integrate NAND into AI infrastructure systems. Some argue that wider NAND adoption will eat into DRAM demand, but this too misreads the memory market as a zero-sum game with a fixed pie. As AI workloads have grown, the upper tiers of the memory hierarchy—HBM and DRAM—alone can no longer accommodate the demand, and NAND has emerged as a key element in extending that hierarchy. NAND, in particular, offers an overwhelmingly lower cost per unit of capacity than DRAM. As of Q1 2026, NAND prices stand at roughly $0.10–$0.12 per GB, whereas DRAM—even mobile DRAM, the lowest-priced application segment—has already surpassed $6 per GB on a contract basis. Even if NAND prices were to double or triple, they would still remain far below DRAM. If even a portion of AI workloads can be offloaded to NAND, it can directly contribute to lowering system-wide cost per token.\n\n$DRAM $MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0025138233731047954, "ret_signed_m1d": 0.0025138233731047954, "alpha_spy_m1d": 0.012365430975110003, "alpha_c_m1d": 0.01660447428729206, "bench_spy_m1d": -0.009851607602005208, "bench_c_m1d": -0.014090650914187264, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04843771778355377, "ret_signed_p1d": 0.04843771778355377, "alpha_spy_p1d": 0.045668603955182796, "alpha_c_p1d": 0.04231992867704992, "bench_spy_p1d": 0.0027691138283709726, "bench_c_p1d": 0.006117789106503846, "ret_p1w": 0.25034812913121196, "ret_signed_p1w": 0.25034812913121196, "alpha_spy_p1w": 0.2323701639903677, "alpha_c_p1w": 0.18447353510527842, "bench_spy_p1w": 0.017977965140844265, "bench_c_p1w": 0.06587459402593354, "ret_p1m": 0.8775621672650864, "ret_signed_p1m": 0.8775621672650864, "alpha_spy_p1m": 0.824936435016864, "alpha_c_p1m": 0.695587918699919, "bench_spy_p1m": 0.05262573224822242, "bench_c_p1m": 0.18197424856516742, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1538, -0.1893, -0.2667, -0.2599, -0.2371, -0.2588, -0.2786, -0.2069, -0.1999, -0.2043, -0.2043, -0.2273, -0.1864, -0.1933, -0.1724, -0.1863, -0.1921, -0.1708, -0.1968, -0.203, -0.2024, -0.2661, -0.2254, -0.2326, -0.2843, -0.2475, -0.2209, -0.1907, -0.2165, -0.1764, -0.1461, -0.1076, -0.1076, -0.1413, -0.1826, -0.2185, -0.2355, -0.2615, -0.313, -0.3096, -0.3778, -0.3467, -0.2887, -0.2918, -0.2918, -0.2695, -0.2699, -0.2135, -0.185, -0.1867, -0.1752, -0.0996, -0.1178, -0.1159, -0.1201, -0.1329, -0.1311, -0.0574, -0.0685, -0.0395, 0.0143, -0.0249, 0.0025, 0.0, 0.0484, 0.1146, 0.2379, 0.2889, 0.2503, 0.4441, 0.5379, 0.4823, 0.5539, 0.5005, 0.4012, 0.3179, 0.3511, 0.4154, 0.4736, 0.4522, 0.4522, 0.7323, 0.7952, 0.7858, 0.8776, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2049847048690938030", "ticker_idx": 1, "ts": "2026-04-30T13:43:15+00:00", "symbol": "SNDK", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NAND has emerged as a key element in extending that hierarchy", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Shares KIS thesis: sustained AI-driven demand for DRAM and NAND; memory ASP premium justified by system-level cost-per-token economics, supporting MU and SNDK.", "resolved_tickers": ["SNDK"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "KIS, which I consider one of the top three Korean sell-side firms, published a memory report, and there’s a comment from it that I wanted to share with you all:\n\nMemory is the critical variable that determines GPU utilization. In particular, when HBM and DRAM capacity is insufficient, memory bottlenecks prevent GPUs from being fully utilized, leading to a decline in overall system efficiency. Conversely, expanding memory capacity raises GPU utilization, allowing the same GPU resources to process a greater number of tokens. This translates directly into a lower cost per token. As cost per token falls, more users are drawn in, ultimately driving a larger expansion in inference demand. This is precisely why hyperscalers, despite rising memory ASPs, are willing to go as far as proposing long-term supply agreements in order to secure greater memory allocation. The cost of purchasing memory rises, but at the same time, that purchase improves GPU utilization—lifting overall system efficiency and lowering cost per unit of performance.\n\nWith DRAM prices having spiked roughly 3x year-over-year in a short span, the market is now bracing for a subsequent decline in demand or prices. But this view stems from looking at memory purely as a standalone cost line item. Even if DRAM is purchased at 3x the price, if that spending allows more GPUs to run and more tokens to be processed, system-wide profitability can actually improve. The fate of the chips that once tried to challenge NVIDIA GPUs by leaning on MLPerf benchmarks—touting per-chip performance-per-watt and price-performance comparisons—illustrates this well. As Jensen Huang has repeatedly emphasized, comparing single-chip specs in isolation is meaningless. Huang has stated that NVIDIA GPUs deliver the lowest cost per token in the world, adding that the lowest cost per token and the highest performance per watt are the decisive metrics of AI economics. In other words, his consistent argument is that competition should not be waged on GPU hardware specs alone, but rather on cost per token at the full-system level. The same logic extends to memory. If \"buying memory at a premium still pays off at the system level,\" then memory purchases by hyperscalers locked in the AI arms race will continue.\n\nThe same logic applies to NAND. Active efforts are underway to integrate NAND into AI infrastructure systems. Some argue that wider NAND adoption will eat into DRAM demand, but this too misreads the memory market as a zero-sum game with a fixed pie. As AI workloads have grown, the upper tiers of the memory hierarchy—HBM and DRAM—alone can no longer accommodate the demand, and NAND has emerged as a key element in extending that hierarchy. NAND, in particular, offers an overwhelmingly lower cost per unit of capacity than DRAM. As of Q1 2026, NAND prices stand at roughly $0.10–$0.12 per GB, whereas DRAM—even mobile DRAM, the lowest-priced application segment—has already surpassed $6 per GB on a contract basis. Even if NAND prices were to double or triple, they would still remain far below DRAM. If even a portion of AI workloads can be offloaded to NAND, it can directly contribute to lowering system-wide cost per token.\n\n$DRAM $MU $SNDK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.029457139962661194, "ret_signed_m1d": -0.029457139962661194, "alpha_spy_m1d": -0.019605532360655986, "alpha_c_m1d": -0.02701200272348603, "bench_spy_m1d": -0.009851607602005208, "bench_c_m1d": -0.002445137239175166, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08252545752292484, "ret_signed_p1d": 0.08252545752292484, "alpha_spy_p1d": 0.07975634369455387, "alpha_c_p1d": 0.06766655396689036, "bench_spy_p1d": 0.0027691138283709726, "bench_c_p1d": 0.014858903556034475, "ret_p1w": 0.22202255246526348, "ret_signed_p1w": 0.22202255246526348, "alpha_spy_p1w": 0.2040445873244192, "alpha_c_p1w": 0.15813538982321806, "bench_spy_p1w": 0.017977965140844265, "bench_c_p1w": 0.06388716264204541, "ret_p1m": 0.5457952644053343, "ret_signed_p1m": 0.5457952644053343, "alpha_spy_p1m": 0.4931695321571119, "alpha_c_p1m": 0.3481776827598113, "bench_spy_p1m": 0.05262573224822242, "bench_c_p1m": 0.197617581645523, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3933, -0.3657, -0.4669, -0.4745, -0.4547, -0.4679, -0.506, -0.4534, -0.4252, -0.4286, -0.4286, -0.4614, -0.4524, -0.4336, -0.4072, -0.3922, -0.4177, -0.4233, -0.4055, -0.4206, -0.4354, -0.4844, -0.4537, -0.4842, -0.5191, -0.4631, -0.4356, -0.4023, -0.4356, -0.3966, -0.3583, -0.3432, -0.3126, -0.2959, -0.3528, -0.3593, -0.3593, -0.3818, -0.4499, -0.4384, -0.4779, -0.4206, -0.3682, -0.3602, -0.3602, -0.3391, -0.3518, -0.2878, -0.2234, -0.2232, -0.1313, -0.1387, -0.1868, -0.1615, -0.1601, -0.1673, -0.176, -0.1071, -0.1496, -0.0972, -0.024, -0.0859, -0.0295, 0.0, 0.0825, 0.1453, 0.2825, 0.2859, 0.222, 0.4248, 0.4114, 0.3242, 0.3199, 0.261, 0.2837, 0.2157, 0.2615, 0.27, 0.4065, 0.3485, 0.3485, 0.4496, 0.45, 0.4972, 0.5458, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039827565343555737", "ticker_idx": 0, "ts": "2026-04-02T22:09:24+00:00", "symbol": "CWTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "lead frame manufacturers are expected to see meaningful gross margin expansion, benefiting from lower-cost inventory. All major Taiwanese lead frame players are expected to sustain steady sequential quarterly growth through 2026", "tweet_type": "analysis", "timeline": "quarters", "summary": "Taiwanese lead frame suppliers CWTC, SDI, Jih Lin Technology to see margin expansion and steady sequential growth through 2026 on double-digit price hikes and volume tailwinds.", "resolved_tickers": ["2425.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "CWTC = Chin-Wei Technology Corp 2425.TW lead frames", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Copper price surge drives quarterly lead frame price hikes\n\n∙ Lead frame packaging suppliers are pushing price increases to pass on rising costs across the semiconductor supply chain.\n\n∙ Taiwanese manufacturers have already laid out new pricing adjustment plans to take effect in Q2 2026, with double-digit percentage increases expected across all product lines.\n\n∙ Taiwan’s three major lead frame suppliers — CWTC, SDI, and Jih Lin Technology — initiated their first round of price negotiations in H2 2025.\n\n∙ Initial increases ranged from 10% to 30%, applied in phases from Q4 2025 through Q1 2026, and have begun flowing through to revenue and profitability.\n\n∙ CWTC noted that precious and base metal prices — specifically gold, silver, and copper — surged by over 30%, 90%, and 35% respectively since H2 2025, driving up overall process costs.\n\n∙ Accordingly, CWTC will raise LED lead frame prices by 20%, with an even larger increase for IC lead frames — both measures taking effect simultaneously on April 1.\n\n∙ SDI implemented pricing adjustments on roughly a quarter of its products earlier this year.\n\n∙ However, with international refined copper premiums having doubled, SDI is set to execute a second round of price hikes, with an average increase expected in the double digits.\n\n∙ SDI’s first round of increases primarily targeted IDM customers, but the second round will extend coverage to OSAT customers as well.\n\n∙ Jih Lin Technology has ongoing price adjustments underway, with its second round of increases also expected to reach double-digit percentages.\n\n∙ Once the initial phase of price increases is absorbed, lead frame manufacturers are expected to see meaningful gross margin expansion, benefiting from lower-cost inventory.\n\n∙ All major Taiwanese lead frame players are expected to sustain steady sequential quarterly growth through 2026, leveraging both volume and pricing tailwinds.\n\n---\n\nhttps://t.co/8FDXy5J7Iv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.024999985328087337, "ret_signed_m1d": 0.024999985328087337, "alpha_spy_m1d": 0.025899649631479393, "alpha_c_m1d": 0.032941758895476236, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.007941773567388899, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.026923106266901975, "ret_signed_p1w": 0.026923106266901975, "alpha_spy_p1w": -0.009793657875030837, "alpha_c_p1w": -0.0177860751836445, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.044709181450546476, "ret_p1m": 0.16538458604079032, "ret_signed_p1m": 0.16538458604079032, "alpha_spy_p1m": 0.06654798869831313, "alpha_c_p1m": -0.024923440862880675, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.190308026903671, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.1904, 0.1654, 0.1673, 0.1404, 0.1558, 0.2115, 0.1942, 0.2058, 0.2115, 0.1904, 0.275, 0.2135, 0.1923, 0.1923, 0.1769, 0.1731, 0.1692, 0.1596, 0.1288, 0.0981, 0.0827, 0.1038, 0.1231, 0.1038, 0.0769, 0.0846, 0.0962, 0.0462, 0.0462, 0.0462, 0.0462, 0.0462, 0.0462, 0.0462, 0.0462, 0.0808, 0.0808, 0.0788, 0.0885, 0.0885, 0.0769, 0.0615, 0.0, 0.0269, 0.0385, -0.0308, -0.0115, 0.0865, 0.1942, 0.1788, 0.1942, 0.1962, 0.1904, 0.1923, 0.1788, 0.1269, 0.1327, 0.1212, 0.0962, 0.0808, 0.05, 0.0096, 0.025, 0.0, 0.0, 0.0, 0.0077, 0.0385, 0.0269, 0.0308, 0.1327, 0.1654, 0.2038, 0.1942, 0.1615, 0.1538, 0.2, 0.1827, 0.1423, 0.1385, 0.1346, 0.1538, 0.1481, 0.1654, 0.1654, 0.1865, 0.2404, 0.2135, 0.1923, 0.1865, 0.2654, 0.3077, 0.2923, 0.3538, 0.3308, 0.3212, 0.25, 0.2558, 0.3096, 0.3346, 0.3462, 0.3423, 0.4038, 0.3308, 0.3692, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2044045150410919940", "ticker_idx": 0, "ts": "2026-04-14T13:28:35+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "You would need at least twice as many CPUs. $AMD $INTC $ARM", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish AMD, INTC, ARM on agentic AI driving structural CPU demand surge — 'at least twice as many CPUs' needed.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "You would need at least twice as many CPUs.\n\n$AMD $INTC $ARM https://t.co/5phuKGbXGM", "reply_text": null, "parent_text": null, "quoted_text": "🔥 CPUs are having a moment.\n\n#Nvidia launched a standalone CPU. #Arm made its first chip in 35 years. #Intel &amp; #AMD are raising prices amid a supply crunch.\n\nWhat's behind it: Agentic AI needs far more CPU than anyone planned for — driving a structural shift in CPU:GPU ratios https://t.co/KroXoKJKdK", "parent_author": null, "ret_m1d": -0.032304878098389, "ret_signed_m1d": -0.032304878098389, "alpha_spy_m1d": -0.020266681338017456, "alpha_c_m1d": -0.013145577960978927, "bench_spy_m1d": -0.012038196760371545, "bench_c_m1d": -0.019159300137410074, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.011957453661307627, "ret_signed_p1d": 0.011957453661307627, "alpha_spy_p1d": 0.004066458794268701, "alpha_c_p1d": 0.009745064280776594, "bench_spy_p1d": 0.007890994867038925, "bench_c_p1d": 0.0022123893805310324, "ret_p1w": 0.11534081650282224, "ret_signed_p1w": 0.11534081650282224, "alpha_spy_p1w": 0.10148833426797554, "alpha_c_p1w": 0.08733195884328815, "bench_spy_p1w": 0.013852482234846697, "bench_c_p1w": 0.028008857659534092, "ret_p1m": 0.7465793202166895, "ret_signed_p1m": 0.7465793202166895, "alpha_spy_p1m": 0.6776769009672463, "alpha_c_p1m": 0.4800748468258571, "bench_spy_p1m": 0.06890241924944318, "bench_c_p1m": 0.2665044733908324, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1064, -0.0911, -0.0911, -0.0908, -0.0207, -0.0053, 0.0181, -0.0147, -0.0119, -0.0091, -0.0113, -0.0719, -0.0345, -0.0508, -0.2152, -0.2453, -0.1828, -0.1532, -0.1627, -0.1627, -0.1926, -0.1872, -0.1872, -0.2038, -0.2154, -0.2027, -0.2153, -0.2292, -0.1616, -0.1733, -0.2015, -0.2151, -0.2213, -0.2514, -0.2078, -0.2181, -0.2456, -0.2054, -0.2032, -0.197, -0.2248, -0.2418, -0.2293, -0.2304, -0.218, -0.1952, -0.2107, -0.2054, -0.1948, -0.1364, -0.2011, -0.2081, -0.2314, -0.2025, -0.1759, -0.1473, -0.1473, -0.1368, -0.1315, -0.0912, -0.0723, -0.0393, -0.0323, 0.0, 0.012, 0.0909, 0.0914, 0.0779, 0.1153, 0.1897, 0.197, 0.3636, 0.3119, 0.2671, 0.3216, 0.3898, 0.4135, 0.339, 0.3928, 0.6521, 0.6014, 0.7846, 0.7987, 0.7575, 0.7466, 0.763, 0.6627, 0.6505, 0.6233, 0.7547, 0.7626, 0.8329, 0.8329, 0.9755, 0.9428, 1.0312, 1.0234, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2044045150410919940", "ticker_idx": 1, "ts": "2026-04-14T13:28:35+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "You would need at least twice as many CPUs. $AMD $INTC $ARM", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish AMD, INTC, ARM on agentic AI driving structural CPU demand surge — 'at least twice as many CPUs' needed.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "You would need at least twice as many CPUs.\n\n$AMD $INTC $ARM https://t.co/5phuKGbXGM", "reply_text": null, "parent_text": null, "quoted_text": "🔥 CPUs are having a moment.\n\n#Nvidia launched a standalone CPU. #Arm made its first chip in 35 years. #Intel &amp; #AMD are raising prices amid a supply crunch.\n\nWhat's behind it: Agentic AI needs far more CPU than anyone planned for — driving a structural shift in CPU:GPU ratios https://t.co/KroXoKJKdK", "parent_author": null, "ret_m1d": 0.021469971828870893, "ret_signed_m1d": 0.021469971828870893, "alpha_spy_m1d": 0.03350816858924244, "alpha_c_m1d": 0.040629271966280966, "bench_spy_m1d": -0.012038196760371545, "bench_c_m1d": -0.019159300137410074, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017708839426357148, "ret_signed_p1d": 0.017708839426357148, "alpha_spy_p1d": 0.009817844559318223, "alpha_c_p1d": 0.015496450045826116, "bench_spy_p1d": 0.007890994867038925, "bench_c_p1d": 0.0022123893805310324, "ret_p1w": 0.038395246986547615, "ret_signed_p1w": 0.038395246986547615, "alpha_spy_p1w": 0.02454276475170092, "alpha_c_p1w": 0.010386389327013523, "bench_spy_p1w": 0.013852482234846697, "bench_c_p1w": 0.028008857659534092, "ret_p1m": 0.8851276967042534, "ret_signed_p1m": 0.8851276967042534, "alpha_spy_p1m": 0.8162252774548102, "alpha_c_p1m": 0.618623223313421, "bench_spy_p1m": 0.06890241924944318, "bench_c_p1m": 0.2665044733908324, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2428, -0.2641, -0.2641, -0.239, -0.1498, -0.1487, -0.2937, -0.3341, -0.3115, -0.2355, -0.2374, -0.2717, -0.2351, -0.2282, -0.2384, -0.244, -0.2072, -0.2127, -0.2614, -0.2432, -0.2716, -0.2667, -0.2667, -0.2763, -0.2876, -0.3007, -0.3087, -0.3163, -0.2772, -0.2653, -0.2876, -0.2852, -0.2869, -0.3246, -0.2857, -0.2799, -0.3195, -0.2857, -0.2669, -0.2481, -0.2909, -0.2827, -0.2829, -0.3095, -0.2943, -0.2763, -0.3125, -0.3103, -0.3095, -0.2606, -0.3089, -0.3241, -0.3545, -0.3084, -0.2473, -0.2105, -0.2105, -0.2042, -0.1708, -0.0762, -0.0328, -0.0224, 0.0215, 0.0, 0.0177, 0.0735, 0.0735, 0.0296, 0.0384, 0.0229, 0.0465, 0.2935, 0.3319, 0.3246, 0.4849, 0.4806, 0.5612, 0.501, 0.6949, 0.771, 0.7179, 0.9577, 1.0285, 0.8901, 0.8851, 0.8168, 0.7046, 0.6952, 0.7364, 0.8643, 0.8571, 0.8781, 0.8781, 0.9357, 0.9083, 0.8945, 0.7972, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2044045150410919940", "ticker_idx": 2, "ts": "2026-04-14T13:28:35+00:00", "symbol": "ARM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "You would need at least twice as many CPUs. $AMD $INTC $ARM", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish AMD, INTC, ARM on agentic AI driving structural CPU demand surge — 'at least twice as many CPUs' needed.", "resolved_tickers": ["ARM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "You would need at least twice as many CPUs.\n\n$AMD $INTC $ARM https://t.co/5phuKGbXGM", "reply_text": null, "parent_text": null, "quoted_text": "🔥 CPUs are having a moment.\n\n#Nvidia launched a standalone CPU. #Arm made its first chip in 35 years. #Intel &amp; #AMD are raising prices amid a supply crunch.\n\nWhat's behind it: Agentic AI needs far more CPU than anyone planned for — driving a structural shift in CPU:GPU ratios https://t.co/KroXoKJKdK", "parent_author": null, "ret_m1d": -0.02257783998317575, "ret_signed_m1d": -0.02257783998317575, "alpha_spy_m1d": -0.010539643222804207, "alpha_c_m1d": -0.003418539845765678, "bench_spy_m1d": -0.012038196760371545, "bench_c_m1d": -0.019159300137410074, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.011661114431073982, "ret_signed_p1d": -0.011661114431073982, "alpha_spy_p1d": -0.019552109298112907, "alpha_c_p1d": -0.013873503811605015, "bench_spy_p1d": 0.007890994867038925, "bench_c_p1d": 0.0022123893805310324, "ret_p1w": 0.08851261732051419, "ret_signed_p1w": 0.08851261732051419, "alpha_spy_p1w": 0.07466013508566749, "alpha_c_p1w": 0.060503759660980094, "bench_spy_p1w": 0.013852482234846697, "bench_c_p1w": 0.028008857659534092, "ret_p1m": 0.37210026695782217, "ret_signed_p1m": 0.37210026695782217, "alpha_spy_p1m": 0.303197847708379, "alpha_c_p1m": 0.10559579356698978, "bench_spy_p1m": 0.06890241924944318, "bench_c_p1m": 0.2665044733908324, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.348, -0.3439, -0.3439, -0.3353, -0.2934, -0.2606, -0.2801, -0.2884, -0.2874, -0.318, -0.3274, -0.3465, -0.3367, -0.3515, -0.3493, -0.3122, -0.2327, -0.2271, -0.2188, -0.2229, -0.2421, -0.2229, -0.2229, -0.2129, -0.2108, -0.2127, -0.2211, -0.2322, -0.2052, -0.1829, -0.1982, -0.2095, -0.2286, -0.245, -0.2302, -0.2518, -0.2905, -0.2704, -0.2523, -0.2551, -0.2859, -0.282, -0.2451, -0.2103, -0.2038, -0.1948, -0.1791, -0.1509, -0.1629, -0.0257, -0.0398, -0.106, -0.1505, -0.0617, -0.0381, -0.0751, -0.0751, -0.0772, -0.1077, -0.0764, -0.0709, -0.0762, -0.0226, 0.0, -0.0117, 0.0069, 0.0342, 0.0861, 0.0885, 0.2193, 0.2691, 0.4565, 0.339, 0.2322, 0.251, 0.3046, 0.3099, 0.2608, 0.2954, 0.4719, 0.3231, 0.3229, 0.319, 0.2897, 0.3721, 0.4173, 0.2974, 0.3343, 0.3841, 0.5924, 0.8498, 0.9012, 0.9012, 0.9924, 0.8776, 1.0796, 1.1914, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038872846911648097", "ticker_idx": 0, "ts": "2026-03-31T06:55:41+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron (MU US, Buy) as an example, we estimate FY2028 BVPS could exceed $300, implying a current valuation of ~1.3x P/B", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF Securities Buy on MU and SK Hynix; DRAM market structurally tight through 2028 on AI demand, LTAs with floor prices >25% above Q2 2026; DDR5 over DDR4.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Communications: Memory Industry — It’s Too Early to Be Pessimistic\n\nSpot prices need not be a major concern: We believe the recent spot price pullback is primarily due to elevated premiums persisting even after Q2 price increases. For example, Huaqiangbei 64GB DDR5 spot prices are ~$2,500, nearly double the Micron/Samsung Q2 2026 contract price of $1,250–1,500. Similarly, eSSD spot market pricing is near $0.5/GB vs. contract prices near $0.3/GB. We believe channel inventory and speculative position unwinding triggered the price decline, but expect spot prices to gradually stabilize underpinned by strong contract pricing.\n\nAI is the key DRAM demand driver: We now estimate pure AI server demand (HBM + near-GPU DRAM) will account for ~37% of total DRAM demand in 2026, growing 74% in 2027 and 90% in 2028, primarily driven by platform transitions such as NVIDIA Rubin Ultra’s shift to 1TB HBM, higher CPU LPDDR capacity, and Google’s aggressive shipment plans. Despite investor concerns about 2028 expansion, we believe the DRAM market will remain structurally tight. AI demand alone can drive >40% incremental wafer-equivalent demand growth, far exceeding the ~15% annual supply growth from technology migration. Consequently, even with new capacity additions, memory capex in 2028 could approach $200B, and this still does not account for wafer output lag and robust general-purpose server demand.\n\nLTAs underpin earnings visibility; valuations digesting rapidly: We have recently observed positive progress on LTAs between memory manufacturers and [hyperscaler] partners. Under new LTA terms, pricing will fluctuate with market conditions but with floor prices set to ensure minimum pricing. Notably, 2027 floor prices are >25% above Q2 2026 DRAM pricing, while 2028 LTAs (e.g., the Samsung–Google agreement) show further YoY increases. More importantly, these LTAs carry higher commitment levels, with upfront payments reaching billions of dollars. Thanks to these agreements, we believe current profitability levels can be sustained through 2028. Taking Micron (MU US, Buy) as an example, we estimate FY2028 BVPS could exceed $300, implying a current valuation of ~1.3x P/B. We believe the near-term memory trading focus remains DRAM pricing, but longer-term investors will pivot toward earnings sustainability.\n\nDDR4 cycle has peaked: In contrast to DDR5, we believe the DDR4 cycle has peaked. On the demand side, DDR4 prices remain above post-Q2 DDR5 prices, which continues to incentivize customers to accelerate migration to DDR5 platforms. On the supply side, the DDR4 EOL timeline has been extended: 1) domestic (Chinese) DRAM manufacturers will not EOL DDR4 and will deliver an additional ~20% DDR4 supply uplift through technology migration; 2) Samsung and SK Hynix (000660 KS, Buy) have also pushed back their DDR4 EOL timelines to Q4 2026. We now expect DDR4 contract prices to rise 15–20% in Q2 2026, well below the 50–60% increase for DDR5, and anticipate meaningful pricing pressure in Q3 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04747812194764367, "ret_signed_m1d": -0.04747812194764367, "alpha_spy_m1d": -0.019231273539485838, "alpha_c_m1d": 0.006955877659623577, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08882906136315039, "ret_signed_p1d": 0.08882906136315039, "alpha_spy_p1d": 0.08129459758026036, "alpha_c_p1d": 0.0664764075795079, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.11762961953927165, "ret_signed_p1w": 0.11762961953927165, "alpha_spy_p1w": 0.10397531028416918, "alpha_c_p1w": 0.07459362511394341, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.5346318600303279, "ret_signed_p1m": 0.5346318600303279, "alpha_spy_p1m": 0.44046575044841685, "alpha_c_p1m": 0.23160631353762473, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1555, -0.0668, -0.0764, 0.0161, 0.0046, -0.0324, 0.021, 0.0233, 0.0004, -0.0137, -0.004, 0.0733, 0.0733, 0.0799, 0.1513, 0.1763, 0.1825, 0.1512, 0.2138, 0.2879, 0.2894, 0.2275, 0.2953, 0.241, 0.1225, 0.1329, 0.1678, 0.1347, 0.1043, 0.2141, 0.2248, 0.218, 0.218, 0.1828, 0.2455, 0.2348, 0.2668, 0.2455, 0.2368, 0.2693, 0.2295, 0.2201, 0.221, 0.1234, 0.1858, 0.1748, 0.0956, 0.1519, 0.1927, 0.2388, 0.1993, 0.2608, 0.3072, 0.366, 0.3661, 0.3145, 0.2513, 0.1964, 0.1703, 0.1305, 0.0517, 0.0569, -0.0475, 0.0, 0.0888, 0.0841, 0.0841, 0.1182, 0.1176, 0.2039, 0.2477, 0.2449, 0.2626, 0.3783, 0.3504, 0.3534, 0.347, 0.3273, 0.3302, 0.4429, 0.4259, 0.4703, 0.5527, 0.4927, 0.5346, 0.5308, 0.6049, 0.7063, 0.895, 0.9731, 0.914, 1.2105, 1.3542, 1.2691, 1.3787, 1.297, 1.145, 1.0173, 1.0683, 1.1667, 1.2558, 1.2229, 1.2229, 1.6518, 1.7481, 1.7336, 1.8741, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038872846911648097", "ticker_idx": 1, "ts": "2026-03-31T06:55:41+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix (000660 KS, Buy)", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF Securities Buy on MU and SK Hynix; DRAM market structurally tight through 2028 on AI demand, LTAs with floor prices >25% above Q2 2026; DDR5 over DDR4.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Communications: Memory Industry — It’s Too Early to Be Pessimistic\n\nSpot prices need not be a major concern: We believe the recent spot price pullback is primarily due to elevated premiums persisting even after Q2 price increases. For example, Huaqiangbei 64GB DDR5 spot prices are ~$2,500, nearly double the Micron/Samsung Q2 2026 contract price of $1,250–1,500. Similarly, eSSD spot market pricing is near $0.5/GB vs. contract prices near $0.3/GB. We believe channel inventory and speculative position unwinding triggered the price decline, but expect spot prices to gradually stabilize underpinned by strong contract pricing.\n\nAI is the key DRAM demand driver: We now estimate pure AI server demand (HBM + near-GPU DRAM) will account for ~37% of total DRAM demand in 2026, growing 74% in 2027 and 90% in 2028, primarily driven by platform transitions such as NVIDIA Rubin Ultra’s shift to 1TB HBM, higher CPU LPDDR capacity, and Google’s aggressive shipment plans. Despite investor concerns about 2028 expansion, we believe the DRAM market will remain structurally tight. AI demand alone can drive >40% incremental wafer-equivalent demand growth, far exceeding the ~15% annual supply growth from technology migration. Consequently, even with new capacity additions, memory capex in 2028 could approach $200B, and this still does not account for wafer output lag and robust general-purpose server demand.\n\nLTAs underpin earnings visibility; valuations digesting rapidly: We have recently observed positive progress on LTAs between memory manufacturers and [hyperscaler] partners. Under new LTA terms, pricing will fluctuate with market conditions but with floor prices set to ensure minimum pricing. Notably, 2027 floor prices are >25% above Q2 2026 DRAM pricing, while 2028 LTAs (e.g., the Samsung–Google agreement) show further YoY increases. More importantly, these LTAs carry higher commitment levels, with upfront payments reaching billions of dollars. Thanks to these agreements, we believe current profitability levels can be sustained through 2028. Taking Micron (MU US, Buy) as an example, we estimate FY2028 BVPS could exceed $300, implying a current valuation of ~1.3x P/B. We believe the near-term memory trading focus remains DRAM pricing, but longer-term investors will pivot toward earnings sustainability.\n\nDDR4 cycle has peaked: In contrast to DDR5, we believe the DDR4 cycle has peaked. On the demand side, DDR4 prices remain above post-Q2 DDR5 prices, which continues to incentivize customers to accelerate migration to DDR5 platforms. On the supply side, the DDR4 EOL timeline has been extended: 1) domestic (Chinese) DRAM manufacturers will not EOL DDR4 and will deliver an additional ~20% DDR4 supply uplift through technology migration; 2) Samsung and SK Hynix (000660 KS, Buy) have also pushed back their DDR4 EOL timelines to Q4 2026. We now expect DDR4 contract prices to rise 15–20% in Q2 2026, well below the 50–60% increase for DDR5, and anticipate meaningful pricing pressure in Q3 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.08178443096239518, "ret_signed_m1d": 0.08178443096239518, "alpha_spy_m1d": 0.11003127937055301, "alpha_c_m1d": 0.13621843056966243, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11152414657934218, "ret_signed_p1d": 0.11152414657934218, "alpha_spy_p1d": 0.10398968279645215, "alpha_c_p1d": 0.08917149279569969, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.13506820107234474, "ret_signed_p1w": 0.13506820107234474, "alpha_spy_p1w": 0.12141389181724227, "alpha_c_p1w": 0.0920322066470165, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.602230577433138, "ret_signed_p1m": 0.602230577433138, "alpha_spy_p1m": 0.5080644678512269, "alpha_c_p1m": 0.2992050309404348, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1948, -0.1626, -0.1391, -0.102, -0.0822, -0.0649, -0.0798, -0.0736, -0.0872, -0.0822, -0.0736, -0.0649, -0.055, -0.081, -0.0847, -0.0662, -0.0513, -0.0897, -0.0105, 0.0402, 0.0649, 0.1243, 0.0266, 0.1218, 0.1132, 0.0414, 0.0377, 0.0971, 0.0835, 0.0637, 0.0983, 0.0885, 0.0885, 0.0885, 0.0885, 0.1058, 0.1738, 0.1763, 0.2431, 0.2591, 0.3618, 0.3147, 0.3147, 0.1636, 0.052, 0.166, 0.145, 0.0359, 0.1623, 0.1834, 0.1524, 0.1276, 0.2069, 0.202, 0.3086, 0.2553, 0.2478, 0.1561, 0.2218, 0.233, 0.1561, 0.1561, 0.0818, 0.0, 0.1115, 0.0285, 0.0855, 0.0979, 0.1351, 0.28, 0.2367, 0.2726, 0.2887, 0.3668, 0.4077, 0.4312, 0.3978, 0.4449, 0.5167, 0.5155, 0.518, 0.5143, 0.601, 0.6109, 0.6022, 0.5936, 0.5936, 0.7931, 0.7931, 0.9839, 1.0496, 1.0892, 1.3296, 1.2739, 1.4486, 1.4411, 1.254, 1.28, 1.1623, 1.1623, 1.404, 1.4052, 1.4052, 1.5428, 1.7794, 1.8369, 1.8914, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038872846911648097", "ticker_idx": 2, "ts": "2026-03-31T06:55:41+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "we believe the DRAM market will remain structurally tight. AI demand alone can drive >40% incremental wafer-equivalent demand growth", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF Securities Buy on MU and SK Hynix; DRAM market structurally tight through 2028 on AI demand, LTAs with floor prices >25% above Q2 2026; DDR5 over DDR4.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Securities Overseas Electronics & Communications: Memory Industry — It’s Too Early to Be Pessimistic\n\nSpot prices need not be a major concern: We believe the recent spot price pullback is primarily due to elevated premiums persisting even after Q2 price increases. For example, Huaqiangbei 64GB DDR5 spot prices are ~$2,500, nearly double the Micron/Samsung Q2 2026 contract price of $1,250–1,500. Similarly, eSSD spot market pricing is near $0.5/GB vs. contract prices near $0.3/GB. We believe channel inventory and speculative position unwinding triggered the price decline, but expect spot prices to gradually stabilize underpinned by strong contract pricing.\n\nAI is the key DRAM demand driver: We now estimate pure AI server demand (HBM + near-GPU DRAM) will account for ~37% of total DRAM demand in 2026, growing 74% in 2027 and 90% in 2028, primarily driven by platform transitions such as NVIDIA Rubin Ultra’s shift to 1TB HBM, higher CPU LPDDR capacity, and Google’s aggressive shipment plans. Despite investor concerns about 2028 expansion, we believe the DRAM market will remain structurally tight. AI demand alone can drive >40% incremental wafer-equivalent demand growth, far exceeding the ~15% annual supply growth from technology migration. Consequently, even with new capacity additions, memory capex in 2028 could approach $200B, and this still does not account for wafer output lag and robust general-purpose server demand.\n\nLTAs underpin earnings visibility; valuations digesting rapidly: We have recently observed positive progress on LTAs between memory manufacturers and [hyperscaler] partners. Under new LTA terms, pricing will fluctuate with market conditions but with floor prices set to ensure minimum pricing. Notably, 2027 floor prices are >25% above Q2 2026 DRAM pricing, while 2028 LTAs (e.g., the Samsung–Google agreement) show further YoY increases. More importantly, these LTAs carry higher commitment levels, with upfront payments reaching billions of dollars. Thanks to these agreements, we believe current profitability levels can be sustained through 2028. Taking Micron (MU US, Buy) as an example, we estimate FY2028 BVPS could exceed $300, implying a current valuation of ~1.3x P/B. We believe the near-term memory trading focus remains DRAM pricing, but longer-term investors will pivot toward earnings sustainability.\n\nDDR4 cycle has peaked: In contrast to DDR5, we believe the DDR4 cycle has peaked. On the demand side, DDR4 prices remain above post-Q2 DDR5 prices, which continues to incentivize customers to accelerate migration to DDR5 platforms. On the supply side, the DDR4 EOL timeline has been extended: 1) domestic (Chinese) DRAM manufacturers will not EOL DDR4 and will deliver an additional ~20% DDR4 supply uplift through technology migration; 2) Samsung and SK Hynix (000660 KS, Buy) have also pushed back their DDR4 EOL timelines to Q4 2026. We now expect DDR4 contract prices to rise 15–20% in Q2 2026, well below the 50–60% increase for DDR5, and anticipate meaningful pricing pressure in Q3 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029577382111775224, "ret_signed_m1d": 0.029577382111775224, "alpha_spy_m1d": 0.05782423051993306, "alpha_c_m1d": 0.08401138171904247, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11154118095690742, "ret_signed_p1d": 0.11154118095690742, "alpha_spy_p1d": 0.1040067171740174, "alpha_c_p1d": 0.08918852717326493, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.14264568502045907, "ret_signed_p1w": 0.14264568502045907, "alpha_spy_p1w": 0.1289913757653566, "alpha_c_p1w": 0.09960969059513083, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.49617902620393045, "ret_signed_p1m": 0.49617902620393045, "alpha_spy_p1m": 0.4020129166220194, "alpha_c_p1m": 0.19315347971122726, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2116, -0.1541, -0.1304, -0.0856, -0.0787, -0.0896, -0.0764, -0.0739, -0.0885, -0.0862, -0.0729, -0.0343, -0.0302, -0.0448, -0.0137, 0.0064, 0.013, -0.0102, 0.0518, 0.0991, 0.1045, 0.1033, 0.0732, 0.1209, 0.0817, 0.0417, 0.0507, 0.075, 0.0591, 0.0931, 0.1297, 0.1293, 0.1293, 0.1176, 0.1385, 0.1582, 0.1918, 0.1912, 0.2245, 0.2477, 0.2975, 0.2757, 0.276, 0.1505, 0.0885, 0.1615, 0.1213, 0.0745, 0.1588, 0.1854, 0.1577, 0.1612, 0.2135, 0.2418, 0.3064, 0.2555, 0.2297, 0.1548, 0.1748, 0.1638, 0.0943, 0.096, 0.0296, 0.0, 0.1115, 0.0598, 0.0944, 0.1237, 0.1426, 0.2476, 0.2348, 0.2499, 0.2512, 0.3267, 0.34, 0.3618, 0.3455, 0.3517, 0.3856, 0.4198, 0.4289, 0.4324, 0.4988, 0.4771, 0.4962, 0.481, 0.5058, 0.63, 0.6929, 0.8493, 0.8625, 0.9685, 1.1304, 1.0705, 1.1753, 1.1695, 1.0056, 0.9927, 0.9594, 0.9932, 1.1503, 1.1259, 1.1259, 1.3276, 1.4545, 1.4539, 1.5538, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2052159092085641420", "ticker_idx": 0, "ts": "2026-05-06T22:50:29+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Securities raised its target price for Samsung Electronics to KRW 500,000, implying 88% upside.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares SK Securities raised PTs: Samsung Electronics to KRW 500k (+88% upside) and SK Hynix to KRW 3M (+87.5% upside).", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "original", "lang": "en", "tweet_text": "SK Securities raised its target price for Samsung Electronics to KRW 500,000, implying 88% upside.\n\nSK Securities raised its target price for SK hynix to KRW 3,000,000, implying 87.5% upside.\n\nSK Securities’ OP forecasts for Samsung Electronics and SK hynix:\n\n2026E OP\nSamsung Electronics: KRW 338tn ($233.83bn)\nSK hynix: KRW 262tn ($181.25bn)\n\n2027E OP\nSamsung Electronics: KRW 494tn ($341.77bn)\nSK hynix: KRW 376tn ($260.12bn)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.12593984962406013, "ret_signed_m1d": -0.12593984962406013, "alpha_spy_m1d": -0.1122309561142405, "alpha_c_m1d": -0.10006209906196273, "bench_spy_m1d": -0.01370889350981963, "bench_c_m1d": -0.0258777505620974, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.020676691729323293, "ret_signed_p1d": 0.020676691729323293, "alpha_spy_p1d": 0.023742796886649598, "alpha_c_p1d": 0.022676317326756568, "bench_spy_p1d": -0.0030661051573263043, "bench_c_p1d": -0.0019996255974332744, "ret_p1w": 0.06766917293233088, "ret_signed_p1w": 0.06766917293233088, "alpha_spy_p1w": 0.056113389888215126, "alpha_c_p1w": 0.02755856084515851, "bench_spy_p1w": 0.011555783044115753, "bench_c_p1w": 0.04011061208717237, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.405, -0.3757, -0.378, -0.3705, -0.33, -0.3202, -0.3202, -0.3202, -0.3202, -0.2872, -0.2868, -0.2759, -0.2497, -0.2365, -0.1821, -0.1878, -0.1878, -0.2681, -0.354, -0.2812, -0.2939, -0.3491, -0.2951, -0.2872, -0.2951, -0.3116, -0.2921, -0.2726, -0.2178, -0.2478, -0.2519, -0.3011, -0.2883, -0.2909, -0.3243, -0.3243, -0.3372, -0.3714, -0.287, -0.3293, -0.3, -0.2741, -0.2613, -0.2086, -0.2331, -0.2256, -0.2444, -0.2237, -0.2068, -0.1823, -0.188, -0.1936, -0.1767, -0.1823, -0.156, -0.1748, -0.156, -0.1654, -0.1504, -0.1711, -0.1711, -0.1259, -0.1259, 0.0, 0.0207, 0.0094, 0.0733, 0.0489, 0.0677, 0.1128, 0.0169, 0.0564, 0.0357, 0.0376, 0.1259, 0.0996, 0.0996, 0.1241, 0.1541, 0.1259, 0.1917, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2052159092085641420", "ticker_idx": 1, "ts": "2026-05-06T22:50:29+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Securities raised its target price for SK hynix to KRW 3,000,000, implying 87.5% upside.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares SK Securities raised PTs: Samsung Electronics to KRW 500k (+88% upside) and SK Hynix to KRW 3M (+87.5% upside).", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "SK Securities raised its target price for Samsung Electronics to KRW 500,000, implying 88% upside.\n\nSK Securities raised its target price for SK hynix to KRW 3,000,000, implying 87.5% upside.\n\nSK Securities’ OP forecasts for Samsung Electronics and SK hynix:\n\n2026E OP\nSamsung Electronics: KRW 338tn ($233.83bn)\nSK hynix: KRW 262tn ($181.25bn)\n\n2027E OP\nSamsung Electronics: KRW 494tn ($341.77bn)\nSK hynix: KRW 376tn ($260.12bn)", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09618987683684477, "ret_signed_m1d": -0.09618987683684477, "alpha_spy_m1d": -0.08248098332702514, "alpha_c_m1d": -0.046950196170076564, "bench_spy_m1d": -0.01370889350981963, "bench_c_m1d": -0.049239680666768204, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.033104300148861565, "ret_signed_p1d": 0.033104300148861565, "alpha_spy_p1d": 0.03617040530618787, "alpha_c_p1d": 0.050675665850426155, "bench_spy_p1d": -0.0030661051573263043, "bench_c_p1d": -0.01757136570156459, "ret_p1w": 0.23422856678337634, "ret_signed_p1w": 0.23422856678337634, "alpha_spy_p1w": 0.22267278373926058, "alpha_c_p1w": 0.19293780026016427, "bench_spy_p1w": 0.011555783044115753, "bench_c_p1w": 0.041290766523212064, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4769, -0.447, -0.4538, -0.4638, -0.4464, -0.4514, -0.4514, -0.4514, -0.4514, -0.4426, -0.4083, -0.4071, -0.3734, -0.3653, -0.3136, -0.3373, -0.3373, -0.4135, -0.4697, -0.4122, -0.4229, -0.4778, -0.4141, -0.4035, -0.4191, -0.4316, -0.3916, -0.3941, -0.3404, -0.3673, -0.371, -0.4172, -0.3841, -0.3785, -0.4172, -0.4172, -0.4547, -0.4959, -0.4397, -0.4816, -0.4528, -0.4466, -0.4279, -0.3548, -0.3766, -0.3585, -0.3504, -0.3111, -0.2904, -0.2786, -0.2954, -0.2717, -0.2355, -0.2361, -0.2349, -0.2367, -0.193, -0.188, -0.1924, -0.1968, -0.1968, -0.0962, -0.0962, 0.0, 0.0331, 0.0531, 0.1743, 0.1462, 0.2342, 0.2305, 0.1362, 0.1493, 0.0899, 0.0899, 0.2117, 0.2124, 0.2124, 0.2817, 0.401, 0.43, 0.4575, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047577112660541818", "ticker_idx": 0, "ts": "2026-04-24T07:23:20+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We remain positive on the stock.", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "GF Overseas reiterates Buy on INTC; 1Q26 beat, strong server CPU momentum, foundry on track; near- and long-term tailwinds intact.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications\n Intel (INTC US, Buy) — 1Q26 Results Review\n1Q26 results beat expectations\n\n- Revenue: US$13.6bn, +7% YoY, above consensus of US$12.4bn\n- Gross margin: 41%, +3.1 ppts QoQ, above guidance of 34.5%\n- EPS: US$0.29, above consensus of US$0.01\n\nSegment performance\n\n- CCG: Revenue US$7.7bn, operating margin 32.6% (+5.6 ppts)\n- DCAI: Revenue US$5.1bn, operating margin 30.5% (+4.1 ppts)\n- Foundry: Revenue US$5.4bn, operating loss US$2.4bn, improving by US$0.1bn QoQ\n\n2Q26 guidance also beat expectations\n\n- Revenue: US$13.8–14.8bn, +5% QoQ, above consensus of US$13bn\n- Gross margin: 39%, -2 ppts QoQ, above consensus of 36.5%\n- EPS: US$0.20, above consensus of US$0.12\n\nOther key takeaways\n\n- The 1Q gross margin beat reflected higher sales, product mix optimization, pricing, and improved 18A yield.\n- Gross margin is expected to decline modestly by 2 ppts QoQ in 2Q, due to a meaningful increase in the contribution from 18A.\n- Capex is expected to be roughly flat with last year, while spending on equipment/tools is increasing significantly and is expected to grow by around 25% YoY.\n- 14A is currently being developed with multiple customers on PDKs. Intel already has a 0.5 PDK and is targeting a 0.9 PDK. The company reiterated that early customer design commitments are expected in 2H26, extending into 1H27.\n- The ASIC business performed strongly, with 1Q revenue up more than 30% QoQ and nearly doubling YoY.\n- Panther Lake shipments are increasing. In 2Q, shipments could grow 6–7x QoQ versus 1Q. Panther Lake gross margin is improving quarter by quarter, though it remains below the company average.\n- Gross margin in the second half needs to be monitored due to rising raw material costs, including substrates, T-glass, and memory prices.\n- Intel increased its advanced packaging order backlog and announced an expansion of its backend facility in Malaysia to support order demand that is expected to convert into revenue in 2027.\n- Intel expects the full-year PC market TAM to decline by low double digits YoY.\n- Intel expects server CPU shipments to grow by double digits this year.\n- Annualized ASIC revenue has already exceeded US$1bn.\n- The GPU-to-CPU ratio is typically 7–8:1 during the model training stage, but changes to around 3:1 or 4:1 during the inference stage.\n\nGF HK view\nIntel rose 19% after hours, as both 1Q results and 2Q guidance came in far above expectations. Server CPU growth momentum is strong, and the foundry business is progressing according to plan, including rapid improvement in 18A yield and smooth progress on 14A.\n\nWe expect near-term results to continue benefiting from CPU tailwinds. Longer term, considering Intel’s favorable supply-chain positioning and improving execution in foundry, investor confidence and risk appetite toward Intel are gradually strengthening. We remain positive on the stock.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.19093775092587517, "ret_signed_m1d": -0.19093775092587517, "alpha_spy_m1d": -0.18324804280530305, "alpha_c_m1d": -0.14238314185204792, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029682540841616367, "ret_signed_p1d": 0.029682540841616367, "alpha_spy_p1d": 0.027959734352914323, "alpha_c_p1d": 0.03003794854028463, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.20692999323485117, "ret_signed_p1w": 0.20692999323485117, "alpha_spy_p1w": 0.19753141359270976, "alpha_c_p1w": 0.20025594523998813, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.45190204759673613, "ret_signed_p1m": 0.45190204759673613, "alpha_spy_p1m": 0.40750054032913785, "alpha_c_p1m": 0.31391925684967736, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4678, -0.409, -0.4105, -0.437, -0.4087, -0.4033, -0.4112, -0.4156, -0.3871, -0.3913, -0.429, -0.415, -0.4369, -0.4331, -0.4331, -0.4405, -0.4492, -0.4594, -0.4656, -0.4714, -0.4412, -0.432, -0.4492, -0.4474, -0.4488, -0.4778, -0.4478, -0.4433, -0.474, -0.4478, -0.4332, -0.4187, -0.4518, -0.4455, -0.4456, -0.4662, -0.4544, -0.4405, -0.4685, -0.4668, -0.4662, -0.4284, -0.4657, -0.4775, -0.501, -0.4654, -0.4181, -0.3896, -0.3896, -0.3848, -0.359, -0.2858, -0.2522, -0.2442, -0.2103, -0.2269, -0.2132, -0.1701, -0.1701, -0.204, -0.1972, -0.2092, -0.1909, 0.0, 0.0297, 0.024, 0.1479, 0.1447, 0.2069, 0.1604, 0.3103, 0.3692, 0.3281, 0.5134, 0.5682, 0.4612, 0.4574, 0.4045, 0.3178, 0.3105, 0.3424, 0.4412, 0.4357, 0.4519, 0.4519, 0.4965, 0.4753, 0.4646, 0.3894, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2057310816651923575", "ticker_idx": 0, "ts": "2026-05-21T04:01:36+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA is taking money from the hyperscalers and returning it to shareholders. That's bullish.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish NVDA: capital extraction from hyperscalers returned to shareholders via buybacks/dividends.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@TShirtnJeans2 @benitoz @firstadopter @zephyr_z9 NVIDIA is taking money from the hyperscalers and returning it to shareholders. That’s bullish.", "reply_text": null, "parent_text": "@benitoz @firstadopter @jukan05 @zephyr_z9 these are big promises. The rest of the Mag 7 are going to cancel buybacks while $NVDA will be buying back a trillion worth of shares. Could be even more if the dividends it pays out get auto reinvested.", "quoted_text": null, "parent_author": "TShirtnJeans2", "ret_m1d": 0.018040211438955067, "ret_signed_m1d": 0.018040211438955067, "alpha_spy_m1d": 0.020019383618388176, "alpha_c_m1d": 0.023710479295886766, "bench_spy_m1d": -0.0019791721794331085, "bench_c_m1d": -0.005670267856931699, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01904237975665879, "ret_signed_p1d": -0.01904237975665879, "alpha_spy_p1d": -0.02297393412505433, "alpha_c_p1d": -0.03390467877905157, "bench_spy_p1d": 0.00393155436839554, "bench_c_p1d": 0.014862299022392778, "ret_p1w": -0.023962437421828353, "ret_signed_p1w": -0.023962437421828353, "alpha_spy_p1w": -0.039957705290505574, "alpha_c_p1w": -0.0802243447478741, "bench_spy_p1w": 0.01599526786867722, "bench_c_p1w": 0.056261907326045746, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1274, -0.1215, -0.1092, -0.1578, -0.1928, -0.1687, -0.1798, -0.1662, -0.1648, -0.19, -0.168, -0.1583, -0.1525, -0.1657, -0.1789, -0.1653, -0.1712, -0.1782, -0.1866, -0.2132, -0.1999, -0.2019, -0.186, -0.2199, -0.2368, -0.2476, -0.2055, -0.1994, -0.1919, -0.1919, -0.1907, -0.1886, -0.1705, -0.1622, -0.1407, -0.1376, -0.1048, -0.094, -0.0964, -0.0812, -0.0795, -0.0894, -0.0775, -0.0905, -0.0512, -0.0132, -0.0289, -0.0467, -0.0908, -0.0959, -0.0958, -0.1048, -0.0532, -0.0365, -0.0196, -0.0003, 0.0058, 0.0288, 0.0739, 0.0265, 0.0128, 0.005, 0.018, 0.0, -0.019, -0.019, -0.0212, -0.0315, -0.024, -0.0381, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2051894739843362932", "ticker_idx": 0, "ts": "2026-05-06T05:20:02+00:00", "symbol": "AI infrastructure / hyperscalers", "kind": "sector", "geography": "global", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit_alias_present", "polarity_reversed": false, "quote": "Agentic AI could turn AI from a capex-heavy cost burden into a business where usage growth drives margin expansion... Goldman Sachs argues that this margin inflection is beginning to appear from 2026 onward", "tweet_type": "analysis", "timeline": "years", "summary": "Endorses Goldman's AI bull case: agentic AI drives token usage, improves hyperscaler unit economics, margin inflection from 2026 onward.", "resolved_tickers": ["NVDA", "MSFT", "GOOGL", "AMZN", "META"], "resolution_method": "basket", "resolution_reason": "AI infra/hyperscaler basket: NVDA, MSFT, GOOGL, AMZN, META", "bench_c": "XLC", "bench_c_reason": "industry='Internet Content & Information'", "bench_c_industry": "Internet Content & Information", "unit_kind": "quote", "lang": "en", "tweet_text": "I read Goldman Sachs’ AI report, and I was genuinely impressed.\n\nThe core insight is as follows:\n\nAgentic AI could turn AI from a capex-heavy cost burden into a business where usage growth drives margin expansion. As token costs fall, more complex agents become economically viable. These agents then consume far more tokens through longer context windows, repeated reasoning loops, validation, tool use, and always-on background monitoring.\n\nThis increase in token usage improves infrastructure utilization, strengthens unit economics, and gives hyperscalers and model providers more room to reinvest in model quality, distribution, and capacity.\n\nIn other words, the bull case for AI capex is not simply that usage will grow. It is that this usage growth can increasingly flow through at attractive incremental margins. Goldman Sachs argues that this margin inflection is beginning to appear from 2026 onward.", "reply_text": null, "parent_text": null, "quoted_text": "We have only just entered the early innings. https://t.co/keHgvgf20w", "parent_author": null, "ret_m1d": -0.02061012449056787, "ret_signed_m1d": -0.02061012449056787, "alpha_spy_m1d": -0.006901230980748241, "alpha_c_m1d": -0.005953104950150667, "bench_spy_m1d": -0.01370889350981963, "bench_c_m1d": -0.014657019540417204, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.005300975635880545, "ret_signed_p1d": 0.005300975635880545, "alpha_spy_p1d": 0.00836708079320685, "alpha_c_p1d": 0.005045340529898845, "bench_spy_p1d": -0.0030661051573263043, "bench_c_p1d": 0.0002556351059816997, "ret_p1w": 0.013084729333603185, "ret_signed_p1w": 0.013084729333603185, "alpha_spy_p1w": 0.001528946289487432, "alpha_c_p1w": 0.0181123931219477, "bench_spy_p1w": 0.011555783044115753, "bench_c_p1w": -0.005027663788344516, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0975, -0.0823, -0.0902, -0.0996, -0.1142, -0.124, -0.124, -0.1243, -0.1155, -0.1154, -0.1007, -0.1165, -0.1103, -0.0948, -0.1082, -0.1191, -0.1133, -0.1138, -0.1005, -0.0995, -0.1161, -0.1062, -0.1027, -0.1021, -0.1166, -0.1324, -0.1182, -0.116, -0.129, -0.1361, -0.155, -0.1442, -0.1612, -0.1545, -0.1876, -0.212, -0.209, -0.1707, -0.1608, -0.1602, -0.1602, -0.1564, -0.1521, -0.1237, -0.1078, -0.101, -0.0891, -0.0564, -0.0407, -0.0348, -0.023, -0.0345, -0.0357, -0.0196, -0.0361, -0.0091, 0.0011, -0.0038, -0.0081, -0.0251, -0.0212, -0.0195, -0.0206, 0.0, 0.0053, 0.0063, -0.0031, -0.0058, 0.0131, 0.0224, 0.0126, 0.0101, -0.006, 0.0041, 0.0024, -0.0049, -0.0049, -0.0036, 0.0048, 0.0155, 0.0155, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2046771689036677236", "ticker_idx": 0, "ts": "2026-04-22T02:02:52+00:00", "symbol": "DB HiTek", "kind": "equity", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DB HiTek plans to raise its 8-inch foundry process pricing by a single-digit percentage per wafer starting in Q2 of this year... Utilization rates at the Bucheon and Sangwoo campuses are also high at over 90%... high utilization rates are expected to be maintained throughout this year", "tweet_type": "analysis", "timeline": "quarters", "summary": "DB HiTek raising 8-inch wafer prices 3–5% from Q2 with 91%+ utilization; full benefit expected through year-end.", "resolved_tickers": ["000990.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "DB HiTek 000990.KS Korea KOSPI", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "8-inch Foundry Prices on the Rise… Korean Foundries to See Full Benefit from Q2\n\nThe 8-inch foundry industry is enjoying a boom driven by rising power semiconductor demand and constrained production capacity.\n\nOn the 22nd, it was learned that DB HiTek plans to raise its 8-inch foundry process pricing by a single-digit percentage per wafer starting in Q2 of this year.\n\n8-inch foundry refers to processes using wafers with a 200mm (millimeter) diameter. It is primarily used for mature (legacy) processes at 130 nanometers (nm) and above. While not cutting-edge, it mass-produces semiconductors essential across industries, including power management ICs (PMIC), display driver ICs (DDI), and CMOS image sensors (CIS).\n\nThe 8-inch foundry market began its full-fledged rebound last year. Power semiconductor demand has been rising across industries, while major players such as Samsung Electronics and TSMC have been gradually scaling down their 8-inch production capacity since last year.\n\nSamsung Electronics is planning to completely shut down its Giheung 6-2 line, which produces 8-inch foundry output. An industry source explained, \"Samsung Electronics is first clearing out products with poor profitability in 8-inch foundry,\" adding that \"the related ecosystem is already within the sphere of influence.\"\n\nAs the 8-inch supply crunch intensifies, 8-inch foundries are also facing pricing pressure. China's SMIC already sent a notice to customers late last year informing them of a 10% price increase for its 8-inch BCD process. BCD is a power semiconductor process that implements three different power devices (Bipolar, CMOS, and DMOS) on a single chip, enhancing durability and power efficiency.\n\nDB HiTek also mass-produces 8-inch BCD processes as its mainstay. It has been in continuous discussions with customers regarding process price increases. The price hike could be reflected as early as Q2 of this year, with the increase understood to be in the 3–5% range.\n\nUtilization rates at the Bucheon and Sangwoo campuses are also high at over 90%. According to DB HiTek's Q1 business report, the company's average utilization rate stands at 91.96%.\n\nAn industry source stated, \"Because there are various 8-inch processes, the situation differs slightly from case to case, but price increase discussions have been progressing steadily since late last year,\" and added, \"As customers continue their inventory build-up stance, high utilization rates are expected to be maintained throughout this year.\"\n\n---\n\nhttps://t.co/XeUX98Pj1x", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04612978889757624, "ret_signed_m1d": -0.04612978889757624, "alpha_spy_m1d": -0.03610461396677833, "alpha_c_m1d": -0.020607101036350772, "bench_spy_m1d": -0.010025174930797909, "bench_c_m1d": -0.025522687861225468, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07427677873338556, "ret_signed_p1d": 0.07427677873338556, "alpha_spy_p1d": 0.07815750268321076, "alpha_c_p1d": 0.06374887639373439, "bench_spy_p1d": -0.0038807239498251933, "bench_c_p1d": 0.010527902339651174, "ret_p1w": 0.1931196247068021, "ret_signed_p1w": 0.1931196247068021, "alpha_spy_p1w": 0.1925993914342714, "alpha_c_p1w": 0.1454086991344814, "bench_spy_p1w": 0.0005202332725307013, "bench_c_p1w": 0.047710925572320706, "ret_p1m": 0.41516810007818616, "ret_signed_p1m": 0.41516810007818616, "alpha_spy_p1m": 0.370863397167686, "alpha_c_p1m": 0.22421950021124415, "bench_spy_p1m": 0.04430470291050015, "bench_c_p1m": 0.19094859986694201, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2949, -0.2972, -0.2593, -0.1888, -0.2167, -0.1802, -0.2019, -0.1252, -0.1415, -0.1911, -0.2019, -0.2329, -0.25, -0.2732, -0.2655, -0.2794, -0.2794, -0.2794, -0.2794, -0.2918, -0.2794, -0.2794, -0.2918, -0.2957, -0.2717, -0.2585, -0.2585, -0.2988, -0.4011, -0.2639, -0.3027, -0.3495, -0.3143, -0.3354, -0.3284, -0.3237, -0.3127, -0.3299, -0.3127, -0.3127, -0.3112, -0.3471, -0.3479, -0.3557, -0.3909, -0.3909, -0.3683, -0.3862, -0.3276, -0.3487, -0.2916, -0.2838, -0.3135, -0.2619, -0.276, -0.2588, -0.2455, -0.2439, -0.2291, -0.1908, -0.1642, -0.0094, -0.0461, 0.0, 0.0743, 0.1884, 0.2432, 0.2095, 0.1931, 0.2392, 0.2392, 0.2643, 0.2643, 0.2666, 0.2314, 0.3143, 0.4597, 0.2963, 0.4511, 0.4292, 0.2783, 0.3362, 0.2299, 0.2666, 0.4152, 0.4159, 0.4159, 0.724, 0.5794, 0.5192, 0.4504, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044570503739044226", "ticker_idx": 0, "ts": "2026-04-16T00:16:09+00:00", "symbol": "Largan Precision", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "My pick is Largan Precision", "tweet_type": "analysis", "timeline": "months", "summary": "Author picks Largan Precision as the top Apple value-chain play on potential iPhone 2027 production increase above 260-270M units.", "resolved_tickers": ["3008.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Largan Precision 3008.TW Taiwan camera lens", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "original", "lang": "en", "tweet_text": "It seems like a good time to invest in Apple’s value chain…\n\nMy pick is Largan Precision.\n\nKorea’s SK Securities: Apple is considering increasing iPhone production by more than 5% in 2027 as well. Shipments could exceed 260 million units and potentially reach 270 million.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.002451288044003519, "alpha_c_m1d": 0.011314308461802258, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.011314308461802258, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.07624113475177308, "ret_signed_p1d": -0.07624113475177308, "alpha_spy_p1d": -0.08832681988199709, "alpha_c_p1d": -0.09156807703286063, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.015326942281087552, "ret_p1w": -0.11879432624113473, "ret_signed_p1w": -0.11879432624113473, "alpha_spy_p1w": -0.12847143384650372, "alpha_c_p1w": -0.14392254593637455, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.025128219695239817, "ret_p1m": 0.22163120567375882, "ret_signed_p1m": 0.22163120567375882, "alpha_spy_p1m": 0.16817224948231235, "alpha_c_p1m": 0.06217857080255862, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.1594526348712002, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1497, -0.1567, -0.1532, -0.1549, -0.1463, -0.1497, -0.1549, -0.1532, -0.1584, -0.1636, -0.1827, -0.1844, -0.1792, -0.2018, -0.1914, -0.2052, -0.1931, -0.2052, -0.2052, -0.2052, -0.2052, -0.2052, -0.2052, -0.2052, -0.2052, -0.1775, -0.1567, -0.1619, -0.1254, -0.1254, -0.148, -0.1497, -0.2, -0.181, -0.174, -0.207, -0.2018, -0.1705, -0.1827, -0.1862, -0.2, -0.1671, -0.1532, -0.1773, -0.1649, -0.2074, -0.2057, -0.2128, -0.2163, -0.2128, -0.227, -0.2358, -0.2234, -0.2411, -0.2411, -0.2411, -0.2305, -0.1915, -0.1915, -0.1809, -0.1383, -0.0532, 0.0, 0.0, -0.0762, -0.039, -0.0621, -0.0833, -0.1188, -0.1064, -0.0727, -0.0851, -0.0798, -0.1082, -0.1082, -0.0869, -0.0975, -0.1064, -0.0869, -0.0887, -0.0975, -0.0089, 0.0106, 0.1117, 0.2216, 0.1667, 0.1933, 0.1348, 0.227, 0.2553, 0.3635, 0.2996, 0.2429, 0.2145, 0.2465, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2046550105877270728", "ticker_idx": 0, "ts": "2026-04-21T11:22:22+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I'd been thinking the same for quite a while… we continue to prefer memory names — for example, MU, which we see as an inevitable re-rating story", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses UBS view: long MU as inevitable re-rating; short STX/WDC as overvalued cyclicals trading at peak multiples that will fall off a cliff when HDD demand flattens.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I’d been thinking the same for quite a while, and UBS finally called it out clearly this time: why are HDD stocks being rewarded with premium valuations while memory stocks remain undervalued? Sharing this here:\n\n\"While we do not see any negative catalysts for STX or WDC in the near term, we believe it is problematic that the market is taking a structurally very different view of HDD stocks versus memory stocks, and we continue to prefer memory names — for example, MU, which we see as an inevitable re-rating story. Put a different way, the market is treating HDD stocks as if they are no longer cyclical - something we push back on aggressively as we believe customers will 2x order HDD capacity until demand starts to flatten - at which case revenue and EPS for HDD stocks will fall off a proverbial cliff. Even if we take an aggressive ~30% EB growth Y/Y, HSD% price increases, and LDD% costdowns for the next 2-3 years, STX still trades at 11x - almost 2-3 turns higher than the peak of previous cycles and the exact opposite of the expectations placed on MU despite the memory industry structurally changing. So, with a still-positive cyclical view for now and a negative structural view relative to the current peaky multiples, we remain Neutral; we continue to prefer MU and semicaps as better cyclicals to gain exposure to the AI buildout.\"\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.002136257609700376, "ret_signed_m1d": -0.002136257609700376, "alpha_spy_m1d": -0.008726351776070063, "alpha_c_m1d": -0.0006297534959091022, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.0015065041137912738, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08478349212144232, "ret_signed_p1d": 0.08478349212144232, "alpha_spy_p1d": 0.07465679528337166, "alpha_c_p1d": 0.05859233554622301, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.02619115657521931, "ret_p1w": 0.12219058050085829, "ret_signed_p1w": 0.12219058050085829, "alpha_spy_p1w": 0.11138217070815215, "alpha_c_p1w": 0.06505205430165617, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.05713852619920212, "ret_p1m": 0.6288886516552075, "ret_signed_p1m": 0.6288886516552075, "alpha_spy_p1m": 0.5760963808792381, "alpha_c_p1m": 0.41367759691777484, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.21521105473743263, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1156, -0.111, -0.1345, -0.0875, -0.0318, -0.0306, -0.0772, -0.0262, -0.067, -0.1561, -0.1483, -0.1221, -0.147, -0.1698, -0.0873, -0.0792, -0.0843, -0.0843, -0.1107, -0.0637, -0.0717, -0.0476, -0.0636, -0.0702, -0.0458, -0.0756, -0.0827, -0.0821, -0.1555, -0.1085, -0.1168, -0.1763, -0.134, -0.1033, -0.0687, -0.0984, -0.0521, -0.0173, 0.027, 0.0271, -0.0118, -0.0593, -0.1006, -0.1202, -0.1501, -0.2093, -0.2054, -0.2839, -0.2482, -0.1814, -0.185, -0.185, -0.1594, -0.1598, -0.0949, -0.062, -0.0641, -0.0508, 0.0362, 0.0152, 0.0175, 0.0127, -0.0021, 0.0, 0.0848, 0.072, 0.1053, 0.1673, 0.1222, 0.1537, 0.1508, 0.2066, 0.2828, 0.4246, 0.4834, 0.4389, 0.6619, 0.7698, 0.7059, 0.7883, 0.7268, 0.6126, 0.5166, 0.5549, 0.6289, 0.6959, 0.6712, 0.6712, 0.9936, 1.066, 1.0551, 1.1608, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2046550105877270728", "ticker_idx": 1, "ts": "2026-04-21T11:22:22+00:00", "symbol": "STX", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "customers will 2x order HDD capacity until demand starts to flatten - at which case revenue and EPS for HDD stocks will fall off a proverbial cliff… STX still trades at 11x - almost 2-3 turns higher than the peak of previous cycles", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses UBS view: long MU as inevitable re-rating; short STX/WDC as overvalued cyclicals trading at peak multiples that will fall off a cliff when HDD demand flattens.", "resolved_tickers": ["STX"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "I’d been thinking the same for quite a while, and UBS finally called it out clearly this time: why are HDD stocks being rewarded with premium valuations while memory stocks remain undervalued? Sharing this here:\n\n\"While we do not see any negative catalysts for STX or WDC in the near term, we believe it is problematic that the market is taking a structurally very different view of HDD stocks versus memory stocks, and we continue to prefer memory names — for example, MU, which we see as an inevitable re-rating story. Put a different way, the market is treating HDD stocks as if they are no longer cyclical - something we push back on aggressively as we believe customers will 2x order HDD capacity until demand starts to flatten - at which case revenue and EPS for HDD stocks will fall off a proverbial cliff. Even if we take an aggressive ~30% EB growth Y/Y, HSD% price increases, and LDD% costdowns for the next 2-3 years, STX still trades at 11x - almost 2-3 turns higher than the peak of previous cycles and the exact opposite of the expectations placed on MU despite the memory industry structurally changing. So, with a still-positive cyclical view for now and a negative structural view relative to the current peaky multiples, we remain Neutral; we continue to prefer MU and semicaps as better cyclicals to gain exposure to the AI buildout.\"\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03598861142245813, "ret_signed_m1d": 0.03598861142245813, "alpha_spy_m1d": 0.042578705588827814, "alpha_c_m1d": 0.03514818941223974, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.000840422010218389, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03568490730047591, "ret_signed_p1d": -0.03568490730047591, "alpha_spy_p1d": -0.02555821046240525, "alpha_c_p1d": -0.013705504347319764, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.021979402953156146, "ret_p1w": 0.03416682273381233, "ret_signed_p1w": -0.03416682273381233, "alpha_spy_p1w": -0.023358412941106188, "alpha_c_p1w": -0.013738846702805896, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.020427976031006434, "ret_p1m": 0.34143592529794287, "ret_signed_p1m": -0.34143592529794287, "alpha_spy_p1m": -0.2886436545219735, "alpha_c_p1m": -0.1963069797040229, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.14512894559391998, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3822, -0.3829, -0.3612, -0.3372, -0.2103, -0.2038, -0.2731, -0.2281, -0.2076, -0.2536, -0.2771, -0.2346, -0.2423, -0.2936, -0.2739, -0.2313, -0.2405, -0.2405, -0.2584, -0.2438, -0.2708, -0.267, -0.2736, -0.2939, -0.2479, -0.2696, -0.2729, -0.3233, -0.3624, -0.3314, -0.3451, -0.371, -0.3326, -0.3148, -0.3118, -0.3332, -0.3159, -0.289, -0.2492, -0.2748, -0.2251, -0.2668, -0.2797, -0.2423, -0.262, -0.3235, -0.3212, -0.3527, -0.3003, -0.2443, -0.2331, -0.2331, -0.1904, -0.1629, -0.1136, -0.1056, -0.1014, -0.0833, -0.0473, -0.072, -0.0502, -0.0217, -0.036, 0.0, 0.0357, 0.0495, 0.0471, 0.0642, 0.0342, 0.149, 0.2031, 0.2983, 0.3191, 0.377, 0.4046, 0.3689, 0.3978, 0.4896, 0.4445, 0.4598, 0.4373, 0.4207, 0.3232, 0.3098, 0.3414, 0.4475, 0.4516, 0.4516, 0.5106, 0.555, 0.573, 0.5714, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2046550105877270728", "ticker_idx": 2, "ts": "2026-04-21T11:22:22+00:00", "symbol": "WDC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we do not see any negative catalysts for STX or WDC in the near term… we remain Neutral", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Author endorses UBS view: long MU as inevitable re-rating; short STX/WDC as overvalued cyclicals trading at peak multiples that will fall off a cliff when HDD demand flattens.", "resolved_tickers": ["WDC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "original", "lang": "en", "tweet_text": "I’d been thinking the same for quite a while, and UBS finally called it out clearly this time: why are HDD stocks being rewarded with premium valuations while memory stocks remain undervalued? Sharing this here:\n\n\"While we do not see any negative catalysts for STX or WDC in the near term, we believe it is problematic that the market is taking a structurally very different view of HDD stocks versus memory stocks, and we continue to prefer memory names — for example, MU, which we see as an inevitable re-rating story. Put a different way, the market is treating HDD stocks as if they are no longer cyclical - something we push back on aggressively as we believe customers will 2x order HDD capacity until demand starts to flatten - at which case revenue and EPS for HDD stocks will fall off a proverbial cliff. Even if we take an aggressive ~30% EB growth Y/Y, HSD% price increases, and LDD% costdowns for the next 2-3 years, STX still trades at 11x - almost 2-3 turns higher than the peak of previous cycles and the exact opposite of the expectations placed on MU despite the memory industry structurally changing. So, with a still-positive cyclical view for now and a negative structural view relative to the current peaky multiples, we remain Neutral; we continue to prefer MU and semicaps as better cyclicals to gain exposure to the AI buildout.\"\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.02527295346325731, "ret_signed_m1d": 0.02527295346325731, "alpha_spy_m1d": 0.031863047629626995, "alpha_c_m1d": 0.02443253145303892, "bench_spy_m1d": 0.006590094166369687, "bench_c_m1d": -0.000840422010218389, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.013782883662675438, "ret_signed_p1d": -0.013782883662675438, "alpha_spy_p1d": -0.003656186824604779, "alpha_c_p1d": 0.008196519290480708, "bench_spy_p1d": 0.010126696838070659, "bench_c_p1d": 0.021979402953156146, "ret_p1w": 0.01870715385595223, "ret_signed_p1w": -0.01870715385595223, "alpha_spy_p1w": -0.007898744063246088, "alpha_c_p1w": 0.001720822175054204, "bench_spy_p1w": 0.010808409792706142, "bench_c_p1w": 0.020427976031006434, "ret_p1m": 0.19751960095052068, "ret_signed_p1m": -0.19751960095052068, "alpha_spy_p1m": -0.1447273301745513, "alpha_c_p1m": -0.0523906553566007, "bench_spy_p1m": 0.052792270775969374, "bench_c_p1m": 0.14512894559391998, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3664, -0.3844, -0.3728, -0.342, -0.2716, -0.275, -0.3483, -0.2963, -0.2442, -0.2984, -0.3224, -0.2641, -0.2552, -0.3162, -0.2871, -0.2601, -0.2667, -0.2667, -0.2601, -0.2277, -0.2587, -0.2564, -0.2697, -0.2954, -0.2423, -0.265, -0.2716, -0.2967, -0.3474, -0.3195, -0.3251, -0.361, -0.3172, -0.3064, -0.2996, -0.3195, -0.2906, -0.2543, -0.1824, -0.2056, -0.1743, -0.2363, -0.2319, -0.2156, -0.2284, -0.2878, -0.2826, -0.3443, -0.2953, -0.2243, -0.2315, -0.2315, -0.2076, -0.1872, -0.1173, -0.1197, -0.1052, -0.0877, -0.0458, -0.049, -0.0576, -0.0294, -0.0253, 0.0, 0.0138, 0.0503, 0.0526, 0.0441, 0.0187, 0.0754, 0.1321, 0.1243, 0.1525, 0.2122, 0.2588, 0.2087, 0.2506, 0.344, 0.2734, 0.2873, 0.2745, 0.2559, 0.1951, 0.1876, 0.1975, 0.2675, 0.2618, 0.2618, 0.367, 0.3825, 0.384, 0.384, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039853707333144877", "ticker_idx": 0, "ts": "2026-04-02T23:53:17+00:00", "symbol": "SK Specialty", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "domestic WF6 suppliers stand to benefit from the windfall", "tweet_type": "analysis", "timeline": "months", "summary": "Korean WF6 suppliers SK Specialty and Foosung are windfall beneficiaries of China's tungsten export controls; Samsung faces higher supply disruption risk than SK Hynix.", "resolved_tickers": ["285130.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Specialty (gas supplier, WF6 context) listed on KOSPI 285130.KS", "bench_c": "XLB", "bench_c_reason": "industry='Specialty Chemicals'", "bench_c_industry": "Specialty Chemicals", "unit_kind": "self_thread", "lang": "en", "tweet_text": "EXCLUSIVE: \"Worse Than the Iran Helium Problem\"... Semiconductor Industry Faces 'Tungsten Crisis'\n\n\"Helium can be sourced from the U.S. or Russia. For tungsten, there's nowhere to go but China.\"\n\nWhile concerns over semiconductor process material procurement — including helium — have emerged in the wake of the Iran war, the semiconductor industry points to tungsten hexafluoride (WF6) as an even more urgent supply risk.\n\nJapanese suppliers, which account for roughly 25% of global WF6 supply, have signaled production cuts in the second half of the year. The reason: China's tightening of tungsten export controls toward Japan has made it difficult for them to secure raw materials.\n\nAccording to multiple industry sources on April 3, Japanese WF6 suppliers such as Kanto Denka and Central Glass began notifying Korean semiconductor companies — including Samsung Electronics and DB HiTek — of raw material supply disruptions starting last week. A source familiar with the matter said, \"They indicated they can maintain supply through May–June using existing tungsten inventory, but cannot guarantee anything beyond that for the second half.\" The source added, \"They asked Korean clients to seek alternatives such as SK Specialty or Foosung.\"\n\nSamsung Electronics has a higher dependence on Japanese WF6 than SK Hynix, making the switch to domestic suppliers more urgent. SK Hynix is reportedly in a relatively better position, as it already sources some volume from SK Specialty, Foosung, and the Chinese supplier Peric. Meanwhile, some foundries are now under pressure to fast-track process qualification of Korean-made WF6 — a procedure that typically takes over 18 months, parts of which are now being skipped given the urgency.\n\nWF6 is a critical precursor used to deposit tungsten films on semiconductor wafers. These tungsten films serve as pathways for electrical signals within chips. While WF6 is used across DRAM, NAND, and logic, consumption is overwhelmingly concentrated in 3D NAND — its vertically stacked cell structure means that at 200 layers, tungsten must be deposited 200 times. A technology transition to replace tungsten interconnects with molybdenum is partially underway in leading-edge NAND processes, but full-scale adoption will take considerable time.\n\nWF6 manufacturers import Chinese tungsten powder, react it with fluorine gas at high temperatures, then distill and purify it into high-purity gas for semiconductor use. China accounts for 80% of global tungsten powder supply.\n\nSince February of last year, China has imposed an export licensing regime on strategic minerals including tungsten. On top of that, following Japanese Prime Minister Sanae Takaichi's remarks on Taiwan in January of this year, China added export controls on \"dual-use\" items destined for Japan. Tungsten powder qualifies as a dual-use item — it is both a semiconductor material and a component used in weapons systems including tanks, missiles, and aircraft.\n\nAs China's export controls have taken hold, tungsten prices have surged relentlessly. Tungsten powder prices have risen 6–7x over the past year. According to Bloomberg, the international benchmark price for ammonium paratungstate (APT) — the raw material for tungsten powder — has climbed 557% since February of last year.\n\nSouth Korea remains relatively free from China's export controls, and SK Specialty and Foosung have maintained their tungsten powder imports. However, SK Specialty has reportedly notified customers of a more-than-double price increase for WF6. Foosung is taking the same stance, as the cost structure is identical. An industry source said, \"The bottom line is that domestic WF6 suppliers stand to benefit from the windfall.\"\n\n---\n\nhttps://t.co/qzXEFp0WSC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06262230919765166, "ret_signed_m1d": 0.06262230919765166, "alpha_spy_m1d": 0.06352197350104372, "alpha_c_m1d": 0.06163045763621766, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": 0.0009918515614339984, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.00391389432485334, "ret_signed_p1d": 0.00391389432485334, "alpha_spy_p1d": 0.00391389432485334, "alpha_c_p1d": 0.00391389432485334, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.08610567514677103, "ret_signed_p1w": 0.08610567514677103, "alpha_spy_p1w": 0.04938891100483822, "alpha_c_p1w": 0.06111066770088458, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.024995007445886452, "ret_p1m": 0.16242661448140905, "ret_signed_p1m": 0.16242661448140905, "alpha_spy_p1m": 0.06359001713893186, "alpha_c_p1m": 0.14377954783680869, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.018647066644600363, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.2991, 0.3029, 0.2952, 0.2856, 0.2778, 0.2836, 0.3126, 0.3473, 0.357, 0.3106, 0.3068, 0.2991, 0.2682, 0.2913, 0.3145, 0.3126, 0.3164, 0.33, 0.3203, 0.2971, 0.2392, 0.272, 0.3126, 0.2624, 0.2508, 0.2643, 0.272, 0.2412, 0.2354, 0.2161, 0.2161, 0.2161, 0.2161, 0.2412, 0.2431, 0.2373, 0.2489, 0.3184, 0.3184, 0.328, 0.328, 0.2431, 0.0578, 0.1466, 0.1427, 0.0655, 0.0983, 0.1582, 0.1369, 0.108, 0.0867, 0.1157, 0.1485, 0.1022, 0.1311, 0.0636, 0.0713, 0.0848, 0.0481, 0.0616, 0.0333, 0.0294, 0.0626, 0.0, 0.0039, 0.0098, 0.0059, 0.0626, 0.0861, 0.1155, 0.1194, 0.1135, 0.1292, 0.1663, 0.1605, 0.1311, 0.1311, 0.1389, 0.1429, 0.1487, 0.1292, 0.1292, 0.2211, 0.1624, 0.1624, 0.1683, 0.1683, 0.1389, 0.1037, 0.0841, 0.0391, -0.0117, -0.0313, 0.0078, -0.047, -0.0812, -0.1106, -0.1526, -0.138, -0.0939, -0.0939, -0.0949, -0.1429, -0.1477, -0.1556, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039853707333144877", "ticker_idx": 2, "ts": "2026-04-02T23:53:17+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics has a higher dependence on Japanese WF6 than SK Hynix, making the switch to domestic suppliers more urgent", "tweet_type": "analysis", "timeline": "months", "summary": "Korean WF6 suppliers SK Specialty and Foosung are windfall beneficiaries of China's tungsten export controls; Samsung faces higher supply disruption risk than SK Hynix.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "EXCLUSIVE: \"Worse Than the Iran Helium Problem\"... Semiconductor Industry Faces 'Tungsten Crisis'\n\n\"Helium can be sourced from the U.S. or Russia. For tungsten, there's nowhere to go but China.\"\n\nWhile concerns over semiconductor process material procurement — including helium — have emerged in the wake of the Iran war, the semiconductor industry points to tungsten hexafluoride (WF6) as an even more urgent supply risk.\n\nJapanese suppliers, which account for roughly 25% of global WF6 supply, have signaled production cuts in the second half of the year. The reason: China's tightening of tungsten export controls toward Japan has made it difficult for them to secure raw materials.\n\nAccording to multiple industry sources on April 3, Japanese WF6 suppliers such as Kanto Denka and Central Glass began notifying Korean semiconductor companies — including Samsung Electronics and DB HiTek — of raw material supply disruptions starting last week. A source familiar with the matter said, \"They indicated they can maintain supply through May–June using existing tungsten inventory, but cannot guarantee anything beyond that for the second half.\" The source added, \"They asked Korean clients to seek alternatives such as SK Specialty or Foosung.\"\n\nSamsung Electronics has a higher dependence on Japanese WF6 than SK Hynix, making the switch to domestic suppliers more urgent. SK Hynix is reportedly in a relatively better position, as it already sources some volume from SK Specialty, Foosung, and the Chinese supplier Peric. Meanwhile, some foundries are now under pressure to fast-track process qualification of Korean-made WF6 — a procedure that typically takes over 18 months, parts of which are now being skipped given the urgency.\n\nWF6 is a critical precursor used to deposit tungsten films on semiconductor wafers. These tungsten films serve as pathways for electrical signals within chips. While WF6 is used across DRAM, NAND, and logic, consumption is overwhelmingly concentrated in 3D NAND — its vertically stacked cell structure means that at 200 layers, tungsten must be deposited 200 times. A technology transition to replace tungsten interconnects with molybdenum is partially underway in leading-edge NAND processes, but full-scale adoption will take considerable time.\n\nWF6 manufacturers import Chinese tungsten powder, react it with fluorine gas at high temperatures, then distill and purify it into high-purity gas for semiconductor use. China accounts for 80% of global tungsten powder supply.\n\nSince February of last year, China has imposed an export licensing regime on strategic minerals including tungsten. On top of that, following Japanese Prime Minister Sanae Takaichi's remarks on Taiwan in January of this year, China added export controls on \"dual-use\" items destined for Japan. Tungsten powder qualifies as a dual-use item — it is both a semiconductor material and a component used in weapons systems including tanks, missiles, and aircraft.\n\nAs China's export controls have taken hold, tungsten prices have surged relentlessly. Tungsten powder prices have risen 6–7x over the past year. According to Bloomberg, the international benchmark price for ammonium paratungstate (APT) — the raw material for tungsten powder — has climbed 557% since February of last year.\n\nSouth Korea remains relatively free from China's export controls, and SK Specialty and Foosung have maintained their tungsten powder imports. However, SK Specialty has reportedly notified customers of a more-than-double price increase for WF6. Foosung is taking the same stance, as the cost structure is identical. An industry source said, \"The bottom line is that domestic WF6 suppliers stand to benefit from the windfall.\"\n\n---\n\nhttps://t.co/qzXEFp0WSC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.06306053811659185, "ret_signed_m1d": -0.06306053811659185, "alpha_spy_m1d": -0.0639602024199839, "alpha_c_m1d": -0.07100231168398075, "bench_spy_m1d": -0.0008996643033920559, "bench_c_m1d": -0.007941773567388899, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043721973094170474, "ret_signed_p1d": -0.043721973094170474, "alpha_spy_p1d": -0.043721973094170474, "alpha_c_p1d": -0.043721973094170474, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.1434977578475336, "ret_signed_p1w": -0.1434977578475336, "alpha_spy_p1w": -0.10678099370560079, "alpha_c_p1w": -0.09878857639698713, "bench_spy_p1w": 0.03671676414193281, "bench_c_p1w": 0.044709181450546476, "ret_p1m": 0.23598654708520184, "ret_signed_p1m": -0.23598654708520184, "alpha_spy_p1m": -0.13714994974272465, "alpha_c_p1m": -0.045678520181530846, "bench_spy_p1m": 0.09883659734247718, "bench_c_p1m": 0.190308026903671, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2275, -0.223, -0.2113, -0.2236, -0.2225, -0.2236, -0.2303, -0.2152, -0.1951, -0.1671, -0.1648, -0.1878, -0.1637, -0.1481, -0.1492, -0.1492, -0.1078, -0.0916, -0.1011, -0.1022, -0.1587, -0.063, -0.0541, -0.1089, -0.1128, -0.0692, -0.0725, -0.0614, -0.0009, 0.0136, 0.0136, 0.0136, 0.0136, 0.0628, 0.0634, 0.0796, 0.1188, 0.1383, 0.2194, 0.2111, 0.2111, 0.0914, -0.0367, 0.0718, 0.0528, -0.0295, 0.0511, 0.0628, 0.0511, 0.0265, 0.0556, 0.0846, 0.1663, 0.1216, 0.1154, 0.0421, 0.0611, 0.0572, 0.0074, 0.0074, -0.0118, -0.0628, 0.0631, 0.0, 0.0437, 0.0824, 0.1015, 0.1799, 0.1435, 0.1547, 0.1267, 0.1575, 0.1827, 0.2192, 0.2108, 0.2024, 0.2276, 0.2192, 0.2584, 0.2304, 0.2584, 0.2444, 0.2668, 0.236, 0.236, 0.3033, 0.3033, 0.491, 0.5219, 0.505, 0.6003, 0.5639, 0.5919, 0.6592, 0.5163, 0.5751, 0.5443, 0.5471, 0.6788, 0.6396, 0.6396, 0.676, 0.7209, 0.6788, 0.7769, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2050103918798287095", "ticker_idx": 0, "ts": "2026-05-01T06:43:57+00:00", "symbol": "SUMCO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I do not think my directional view was wrong; rather, I was probably too conservative on the timing", "tweet_type": "analysis", "timeline": "quarters", "summary": "Maintains bullish directional view on SUMCO; admits timing was too conservative but thesis on epi wafer tightening and 300mm constraints intact.", "resolved_tickers": ["3436.T"], "resolution_method": "exchange_ticker", "resolution_reason": "SUMCO Corp listed on TSE as 3436.T", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "quote", "lang": "en", "tweet_text": "I fully agree with the broad logic laid out by SemiAnalysis. That said, I did not expect the stock to bottom and rebound this quickly.\n\nIt is true that the supply-demand balance for epitaxial wafers is tightening. However, there is one important nuance in the near term. We are now in a period where profitability in commodity memory has become higher than in HBM, which gives the Memory Big 3 a stronger incentive to prioritize commodity DRAM/NAND production.\n\nIn that scenario, the shipment mix of relatively lower-margin polished wafers could grow faster than that of higher-margin epitaxial wafers in 2026. In other words, even if wafer shipments recover, the improvement in company-wide ASPs and margins may be more gradual than the market currently expects.\n\nThis is especially relevant given that 85–90% of SUMCO’s 300mm revenue is under LTAs, while spot-price exposure is limited to only some memory-related demand. As a result, it is difficult for the memory supercycle to translate immediately into an explosive increase in wafer makers’ ASPs. And considering that the wafer industry typically lags the memory cycle, this rerating has happened much faster than I had expected.\n\nUltimately, the market seems to be pricing in post-2027 epi wafer tightness and 300mm supply constraints ahead of 2026 earnings. I do not think my directional view was wrong; rather, I was probably too conservative on the timing.", "reply_text": null, "parent_text": null, "quoted_text": "SemiAnalysis最近发了一组硅晶圆行业的分析，逻辑很清晰，简单转述一下。\n\n需求端两条线。一条是先进制程逻辑芯片（7nm以下），晶圆需求预计明年开始加速，到CY28接近100万片/月，占300mm等效总需求约10%。用的是外延片（epi wafer），供需平衡收紧速度比市场预期快。另一条是HBM，8到16层DRAM", "parent_author": null, "ret_m1d": -0.022502960915909975, "ret_signed_m1d": -0.022502960915909975, "alpha_spy_m1d": -0.01974149390402813, "alpha_c_m1d": -0.016422371572651473, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0036633796697419507, "alpha_c_p1d": 0.00594327161697672, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.41255428345834977, "ret_signed_p1w": 0.41255428345834977, "alpha_spy_p1w": 0.3890061392437947, "alpha_c_p1w": 0.30129938979339, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3537, -0.3604, -0.3662, -0.3587, -0.331, -0.3034, -0.3034, -0.3624, -0.3976, -0.4062, -0.3918, -0.3879, -0.3593, -0.3628, -0.3628, -0.3439, -0.3462, -0.3062, -0.2785, -0.2957, -0.319, -0.3677, -0.3324, -0.3289, -0.3895, -0.3527, -0.3265, -0.3593, -0.3814, -0.362, -0.361, -0.3018, -0.3358, -0.3358, -0.3699, -0.3527, -0.3044, -0.269, -0.2694, -0.2965, -0.3448, -0.3097, -0.3235, -0.3139, -0.3107, -0.304, -0.2521, -0.2304, -0.1972, -0.1496, -0.1417, -0.1163, -0.0811, -0.1729, -0.167, -0.1212, -0.1153, -0.1461, -0.1149, -0.0606, -0.0732, -0.0732, -0.0225, 0.0, 0.0, 0.0, 0.0, 0.1974, 0.4126, 0.4264, 0.4844, 0.3837, 0.3529, 0.2199, 0.1812, 0.1686, 0.1923, 0.2152, 0.2542, 0.3529, 0.3241, 0.2969, 0.3218, 0.5768, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2057598767394959604", "ticker_idx": 0, "ts": "2026-05-21T23:05:49+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AMD is accelerating the introduction of its next-generation Helios rack-scale AI platform... AMD's MI300 and MI350 series, in particular, are establishing themselves as one of the core pillars of hyperscalers' future high-performance AI server expansion, positioning AMD as a key player in AI supply chain diversification", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long AMD on accelerating Helios rack-scale platform rollout, hyperscaler adoption (MSFT, Meta, xAI), and MI300/MI350 becoming core AI server pillars for diversification away from NVIDIA.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": ">> AMD Accelerates Helios Rollout… Taiwan AI Server Supply Chain Gears Up in Earnest\n\n• AMD is accelerating the introduction of its next-generation Helios rack-scale AI platform, with full-scale mass production deployment set to begin in 2H26. As North American CSPs broaden their adoption of AMD AI GPU platforms, Taiwanese ODMs are also moving into production-ready mode.\n\n• Major ODMs including Wistron, Wiwynn, and Inventec have already begun mass production preparations, with sequential shipments expected to start as early as late Q2 to Q3, followed by a full-scale ramp in Q4. The market views rising AMD platform penetration—alongside expanding rack-scale AI server demand—as a new growth engine for the Taiwan AI server supply chain.\n\n• TrendForce notes that the top five North American hyperscalers are sharply increasing rack-scale AI server purchases this year to expand their AI training and inference infrastructure. Accordingly, NVIDIA GB300, AMD Helios, and CSPs' own ASIC platforms are all expected to enter large-scale buildout phases simultaneously.\n\n• The industry view is that cloud companies have aggressively expanded AMD purchases over the past year to avoid excessive concentration of AI compute resources on NVIDIA. AMD's MI300 and MI350 series, in particular, are establishing themselves as one of the core pillars of hyperscalers' future high-performance AI server expansion, positioning AMD as a key player in AI supply chain diversification.\n\n• Microsoft is currently the most aggressive CSP in adopting AMD AI infrastructure, having designated AMD as its second AI accelerator platform. Helios rack systems will begin mass production shipments in 2H26, with MI450-based AI rack systems targeted for deployment before year-end and MI500 series adoption planned for 2027.\n\n• Meta is likewise leveraging AMD platforms as an important supplementary compute resource within its \"Meta Compute\" strategy, aiming to build a hybrid ASIC compute architecture beyond its NVIDIA-centric structure. xAI is also cited as a key customer that has adopted the AMD Helios rack system as core compute infrastructure for AI development.\n\n• Within the Taiwan supply chain, Wistron has secured a substantial volume of AMD front-end board orders and will support shipments of the new platform through its newly built Zhubei Plant 2. Inventec is expected to see a meaningful shipment ramp from Q4 on the back of growing orders from large CSP customers, while Wiwynn is also projected to gradually increase Helios rack system shipments from Q3 onward.\n\n$AMD\n\n---\n\nhttps://t.co/vAisBgldQV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.004470761765158038, "ret_signed_m1d": -0.004470761765158038, "alpha_spy_m1d": -0.0024915895857249293, "alpha_c_m1d": 0.001199506091773661, "bench_spy_m1d": -0.0019791721794331085, "bench_c_m1d": -0.005670267856931699, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.039858567970152414, "ret_signed_p1d": 0.039858567970152414, "alpha_spy_p1d": 0.035927013601756874, "alpha_c_p1d": 0.024996268947759637, "bench_spy_p1d": 0.00393155436839554, "bench_c_p1d": 0.014862299022392778, "ret_p1w": 0.15236110917845402, "ret_signed_p1w": 0.15236110917845402, "alpha_spy_p1w": 0.1363658413097768, "alpha_c_p1w": 0.09609920185240828, "bench_spy_p1w": 0.01599526786867722, "bench_c_p1w": 0.056261907326045746, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5627, -0.5244, -0.531, -0.547, -0.5547, -0.5582, -0.5753, -0.5505, -0.5564, -0.572, -0.5492, -0.548, -0.5444, -0.5602, -0.5699, -0.5628, -0.5634, -0.5564, -0.5434, -0.5522, -0.5492, -0.5432, -0.5101, -0.5468, -0.5507, -0.564, -0.5475, -0.5324, -0.5162, -0.5162, -0.5103, -0.5073, -0.4844, -0.4737, -0.455, -0.451, -0.4327, -0.4259, -0.3811, -0.3808, -0.3884, -0.3672, -0.325, -0.3209, -0.2264, -0.2557, -0.2811, -0.2502, -0.2115, -0.1981, -0.2403, -0.2098, -0.0627, -0.0915, 0.0125, 0.0205, -0.0029, -0.0091, 0.0002, -0.0567, -0.0636, -0.079, -0.0045, 0.0, 0.0399, 0.0399, 0.1208, 0.1022, 0.1524, 0.1479, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047219439385677997", "ticker_idx": 0, "ts": "2026-04-23T07:42:04+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is highly likely to deliver record profitability again in Q2, as DRAM and NAND prices continue to climb and shipments expand with a shift toward higher-value products", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long SK Hynix into Q2: record margins expected again as DRAM/NAND ASPs rise ~20-30% QoQ and bit shipments accelerate.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SK Hynix Set for Even Stronger Q2 — Memory Tailwinds Point to Record Profitability\n\nSK Hynix posted its highest-ever profitability in Q1, driven by an unprecedented memory supercycle. The operating margin reached 72%, placing the company among the top tier of global manufacturers. SK Hynix is highly likely to deliver record profitability again in Q2, as DRAM and NAND prices continue to climb and shipments expand with a shift toward higher-value products.\n\nAccording to industry sources on the 23rd, SK Hynix is expected to set a new all-time high operating margin in Q2 on the back of the continuing memory supercycle.\n\nIn Q1, SK Hynix reported revenue of KRW 52.5763 trillion and operating profit of KRW 37.6103 trillion. Revenue rose 198% YoY and 60% QoQ, while operating profit jumped 405% YoY and 96% QoQ.\n\nThese results mark a record high for the company. The 72% operating margin ranks at the top of the global manufacturing sector, exceeding the recent profitability of major players including TSMC, Nvidia, and Micron.\n\nMemory bit growth in Q1 was limited — DRAM was flat QoQ and NAND declined roughly 10% QoQ. However, average selling prices (ASP) for DRAM and NAND rose by the mid-60% and mid-70% range respectively, driving a sharp expansion in profitability.\n\nThis was supported by rising demand for high-value server memory tied to AI infrastructure investment. Notably, excluding HBM — where price volatility is limited — commodity DRAM ASP is estimated to have increased by around 100%.\n\nQ2 earnings are expected to expand further, as volume growth and price increases are set to materialize simultaneously.\n\nSK Hynix guided Q2 DRAM bit growth to the high single digits and NAND to the mid-teens, with ASPs expected to continue rising. Current industry estimates place Q2 QoQ ASP increases at roughly 20% for DRAM and 30% for NAND.\n\nA semiconductor industry source said, \"Given current conditions, a sharp rise in memory prices through Q2 is all but locked in, and SK Hynix's operating margin is highly likely to set a new record high in Q2.\"\n\n---\n\nhttps://t.co/LpLECpQ9qK", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0016327218773913854, "ret_signed_m1d": -0.0016327218773913854, "alpha_spy_m1d": -0.00552856451705297, "alpha_c_m1d": 0.008785498456036489, "bench_spy_m1d": 0.003895842639661584, "bench_c_m1d": -0.010418220333427874, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0024489807582124268, "ret_signed_p1d": -0.0024489807582124268, "alpha_spy_p1d": -0.010198278718236642, "alpha_c_p1d": -0.05348145149892658, "bench_spy_p1d": 0.007749297960024215, "bench_c_p1d": 0.05103247074071415, "ret_p1w": 0.04979587404506991, "ret_signed_p1w": 0.04979587404506991, "alpha_spy_p1w": 0.0353841854833139, "alpha_c_p1w": -0.0018176878562756738, "bench_spy_p1w": 0.014411688561756009, "bench_c_p1w": 0.051613561901345584, "ret_p1m": 0.5844897341004547, "ret_signed_p1m": 0.5844897341004547, "alpha_spy_p1m": 0.5319948483631414, "alpha_c_p1m": 0.3884328698811603, "bench_spy_p1m": 0.05249488573731331, "bench_c_p1m": 0.19605686421929436, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4003, -0.3481, -0.3147, -0.2984, -0.2593, -0.3237, -0.261, -0.2667, -0.3139, -0.3164, -0.2773, -0.2862, -0.2993, -0.2764, -0.283, -0.283, -0.283, -0.283, -0.2715, -0.2267, -0.2251, -0.1811, -0.1705, -0.1029, -0.1339, -0.1339, -0.2335, -0.3069, -0.2318, -0.2457, -0.3176, -0.2343, -0.2204, -0.2408, -0.2571, -0.2049, -0.2082, -0.138, -0.1731, -0.178, -0.2384, -0.1951, -0.1878, -0.2384, -0.2384, -0.2873, -0.3412, -0.2678, -0.3224, -0.2849, -0.2767, -0.2522, -0.1567, -0.1853, -0.1616, -0.151, -0.0996, -0.0727, -0.0571, -0.0792, -0.0482, -0.0008, -0.0016, 0.0, -0.0024, 0.0547, 0.0612, 0.0555, 0.0498, 0.0498, 0.1812, 0.1812, 0.3069, 0.3502, 0.3763, 0.5347, 0.498, 0.6131, 0.6082, 0.4849, 0.502, 0.4245, 0.4245, 0.5837, 0.5845, 0.5845, 0.6751, 0.831, 0.8689, 0.9048, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2042599777712427019", "ticker_idx": 0, "ts": "2026-04-10T13:45:11+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Lynx Equity raised Micron's price target to $825, a Street-high", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Relays Lynx Equity Street-high $825 PT on MU; HBM/DDR5/LPDDR5 capacity fully booked through 2027.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Lynx Equity raised Micron’s price target to $825, a Street-high, and commented as follows:\n\n\"Beyond Micron management’s statement that its 2026 HBM capacity is already sold out, the company’s HBM, DDR5, and LPDDR5 production capacity has now been fully reserved through 2027.\n\nWith 2027 allocation now complete, Micron is currently in discussions with key customers regarding 2028 supply volumes and pricing terms.\"\n\nOK… so 2027 is already fully booked?\n\n$MU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.002187435354490175, "ret_signed_m1d": 0.002187435354490175, "alpha_spy_m1d": 0.0015252165964700914, "alpha_c_m1d": 0.017225906446755146, "bench_spy_m1d": 0.0006622187580200833, "bench_c_m1d": -0.015038471092264971, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014194349063659129, "ret_signed_p1d": 0.014194349063659129, "alpha_spy_p1d": 0.0044219512793073346, "alpha_c_p1d": -0.0005923003203311961, "bench_spy_p1d": 0.009772397784351794, "bench_c_p1d": 0.014786649383990325, "ret_p1w": 0.08198010244311149, "ret_signed_p1w": 0.08198010244311149, "alpha_spy_p1w": 0.036826610429395945, "alpha_c_p1w": 0.019537329886802013, "bench_spy_p1w": 0.045153492013715546, "bench_c_p1w": 0.06244277255630948, "ret_p1m": 0.890986528483424, "ret_signed_p1m": 0.890986528483424, "alpha_spy_p1m": 0.8029166437438024, "alpha_c_p1m": 0.5718371256326764, "bench_spy_p1m": 0.08806988473962152, "bench_c_p1m": 0.31914940285074755, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1964, -0.2078, -0.2, -0.1379, -0.1379, -0.1325, -0.0752, -0.0551, -0.0502, -0.0753, -0.025, 0.0345, 0.0357, -0.014, 0.0405, -0.0032, -0.0983, -0.09, -0.062, -0.0886, -0.1129, -0.0248, -0.0162, -0.0216, -0.0216, -0.0499, 0.0004, -0.0081, 0.0176, 0.0005, -0.0066, 0.0196, -0.0124, -0.02, -0.0192, -0.0976, -0.0475, -0.0564, -0.1199, -0.0747, -0.042, -0.0049, -0.0366, 0.0127, 0.05, 0.0973, 0.0974, 0.0559, 0.0051, -0.039, -0.06, -0.0919, -0.1552, -0.151, -0.2349, -0.1967, -0.1254, -0.1292, -0.1292, -0.1018, -0.1023, -0.033, 0.0022, 0.0, 0.0142, 0.1072, 0.0847, 0.0871, 0.082, 0.0662, 0.0685, 0.159, 0.1453, 0.181, 0.2472, 0.199, 0.2327, 0.2296, 0.2892, 0.3706, 0.5221, 0.5849, 0.5374, 0.7756, 0.891, 0.8226, 0.9107, 0.8451, 0.723, 0.6204, 0.6613, 0.7404, 0.812, 0.7856, 0.7856, 1.1301, 1.2074, 1.1958, 1.3087, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2038861527344918844", "ticker_idx": 0, "ts": "2026-03-31T06:10:42+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Citi cuts Micron's price target to $425 from $510, citing weakness in memory spot prices", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares Citi PT cut on MU to $425 from $510 on memory spot price weakness; implied bearish.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Just in: Citi cuts Micron’s price target to $425 from $510, citing weakness in memory spot prices.\n\n$MU https://t.co/KAvNtTuBNs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04747812194764367, "ret_signed_m1d": 0.04747812194764367, "alpha_spy_m1d": 0.019231273539485838, "alpha_c_m1d": -0.006955877659623577, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.08882906136315039, "ret_signed_p1d": -0.08882906136315039, "alpha_spy_p1d": -0.08129459758026036, "alpha_c_p1d": -0.0664764075795079, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.11762961953927165, "ret_signed_p1w": -0.11762961953927165, "alpha_spy_p1w": -0.10397531028416918, "alpha_c_p1w": -0.07459362511394341, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.5346318600303279, "ret_signed_p1m": -0.5346318600303279, "alpha_spy_p1m": -0.44046575044841685, "alpha_c_p1m": -0.23160631353762473, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1555, -0.0668, -0.0764, 0.0161, 0.0046, -0.0324, 0.021, 0.0233, 0.0004, -0.0137, -0.004, 0.0733, 0.0733, 0.0799, 0.1513, 0.1763, 0.1825, 0.1512, 0.2138, 0.2879, 0.2894, 0.2275, 0.2953, 0.241, 0.1225, 0.1329, 0.1678, 0.1347, 0.1043, 0.2141, 0.2248, 0.218, 0.218, 0.1828, 0.2455, 0.2348, 0.2668, 0.2455, 0.2368, 0.2693, 0.2295, 0.2201, 0.221, 0.1234, 0.1858, 0.1748, 0.0956, 0.1519, 0.1927, 0.2388, 0.1993, 0.2608, 0.3072, 0.366, 0.3661, 0.3145, 0.2513, 0.1964, 0.1703, 0.1305, 0.0517, 0.0569, -0.0475, 0.0, 0.0888, 0.0841, 0.0841, 0.1182, 0.1176, 0.2039, 0.2477, 0.2449, 0.2626, 0.3783, 0.3504, 0.3534, 0.347, 0.3273, 0.3302, 0.4429, 0.4259, 0.4703, 0.5527, 0.4927, 0.5346, 0.5308, 0.6049, 0.7063, 0.895, 0.9731, 0.914, 1.2105, 1.3542, 1.2691, 1.3787, 1.297, 1.145, 1.0173, 1.0683, 1.1667, 1.2558, 1.2229, 1.2229, 1.6518, 1.7481, 1.7336, 1.8741, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2038784940452843825", "ticker_idx": 0, "ts": "2026-03-31T01:06:23+00:00", "symbol": "MSFT", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Short MSFT", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Short MSFT in reaction to criticism of Office 365/SharePoint product quality.", "resolved_tickers": ["MSFT"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@dylan522p Short MSFT", "reply_text": null, "parent_text": "Microsoft Office 365 is such a dog shit product.\nLiterally been breaking for the last 2 weeks when we try to buy more licenses. \nWe use Gmail and Slack, so it's silly to even use this.\nSharepoint is buggy a f too.\nWhy no one else make a functional collaborative file systems :( https://t.co/kO9ur2JYeP", "quoted_text": null, "parent_author": "dylan522p", "ret_m1d": -0.030283431690427798, "ret_signed_m1d": 0.030283431690427798, "alpha_spy_m1d": 0.002036583282269966, "alpha_c_m1d": 0.002036583282269966, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.028246848408157832, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0021611901863888283, "ret_signed_p1d": 0.0021611901863888283, "alpha_spy_p1d": 0.00969565396927885, "alpha_c_p1d": 0.00969565396927885, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.007534463782890022, "ret_p1w": 0.005727096159496181, "ret_signed_p1w": -0.005727096159496181, "alpha_spy_p1w": 0.00792721309560629, "alpha_c_p1w": 0.00792721309560629, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.01365430925510247, "ret_p1m": 0.14666225913197461, "ret_signed_p1m": -0.14666225913197461, "alpha_spy_p1m": -0.05249614955006354, "alpha_c_p1m": -0.05249614955006354, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.09416610958191107, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.3035, 0.2747, 0.2745, 0.2897, 0.3031, 0.2887, 0.2918, 0.2861, 0.2686, 0.2382, 0.2308, 0.2395, 0.2395, 0.2251, 0.197, 0.216, 0.2559, 0.2676, 0.2953, 0.2981, 0.1684, 0.1598, 0.1411, 0.1083, 0.1164, 0.0611, 0.0812, 0.1148, 0.1139, 0.0899, 0.0831, 0.0817, 0.0817, 0.0697, 0.077, 0.0764, 0.0731, 0.0386, 0.0509, 0.0822, 0.0852, 0.061, 0.0767, 0.0912, 0.0946, 0.1094, 0.1048, 0.106, 0.0961, 0.0938, 0.0856, 0.0686, 0.0804, 0.079, 0.0584, 0.0509, 0.0316, 0.0347, 0.0069, 0.0024, -0.0113, -0.0362, -0.0303, 0.0, -0.0022, 0.0089, 0.0089, 0.0073, 0.0057, 0.0112, 0.0078, 0.0019, 0.0384, 0.062, 0.1109, 0.1353, 0.1422, 0.1294, 0.1459, 0.1695, 0.1231, 0.1471, 0.1476, 0.1596, 0.1467, 0.1016, 0.1196, 0.1174, 0.1113, 0.1183, 0.1367, 0.1214, 0.1148, 0.1016, 0.0947, 0.1061, 0.1398, 0.1442, 0.1276, 0.1375, 0.1346, 0.1332, 0.1332, 0.1263, 0.1172, 0.156, 0.2189, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2050058291066376226", "ticker_idx": 0, "ts": "2026-05-01T03:42:39+00:00", "symbol": "TER", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "we remain constructive over the long term on TER's AI-related share gains and CPO growth potential ... we lower our FY27 earnings multiple to 30x P/E and set a new target price of US$381", "tweet_type": "analysis", "timeline": "years", "summary": "Remain constructive long-term on TER; cut FY27 multiple to 30x, new PT $381; near-term limited but CPO and AI upside materialize in 2027.", "resolved_tickers": ["TER"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "[GF Overseas Electronics & Communications] \nTeradyne 1Q26 Review: Entering a Transition Period in 2H26\n\n☄️ Teradyne 1Q26 Review: Entering a Transition Period in 2H26\n\n☀️ Strong 1Q results, but full-year 2026 guidance fell short of expectations:\nTeradyne reported 1Q26 revenue of US$1.28 billion and gross margin of 60.9%, beating market expectations. Despite the strong first-quarter performance, the company guided 2Q26 revenue of US$1.15–1.25 billion, with the midpoint implying a 6% QoQ decline.\nManagement noted that 55–60% of 2026 revenue will still be concentrated in 1H26, which is below the market’s optimistic expectations for CPO, memory, and AI businesses. Meanwhile, in the CPO testing segment, management expects the medium-term TAM to reach US$300–700 million.\n\nAfter the earnings release, the stock fell 20%. Although we remain constructive over the long term on TER’s AI-related share gains and CPO growth potential, near-term catalysts are limited. The company currently trades at 47x FY2026 P/E, representing a valuation premium versus SPE peers. Given uncertainty around 2H26 revenue, we lower our FY27 earnings multiple to 30x P/E and set a new target price of US$381.\n\n☀️ 2H26 enters a transition period:\nWe believe the high first-half revenue weighting is partly due to the Mellanox business, where all 160–170 equipment shipments are concentrated in 1H26. Looking ahead, 2H26 will be a transition period for Teradyne’s business, with AI-driven growth expected to materialize more meaningfully in 2027.\n\nAlthough Teradyne completed its first GPU tester shipments in 1Q26, these were mainly for Nvidia gaming GPUs. Testers for Nvidia’s Rubin platform are expected to begin shipping in 4Q26, with 2027 shipments ramping to 100–150 units. On the MediaTek side, a small number of tools will ship in 2026, while a more significant revenue contribution is expected in 2H27 as MediaTek’s Humufish capacity scales up.\n\n☀️ CPO revenue has upside potential:\nManagement stated that the medium-term TAM for CPO test equipment is US$300–700 million, and it is also positive on providing Insertion 2 testing solutions through its partnership with Ficontec.\n\nWe are more optimistic on the outlook for the CPO business. According to supply-chain checks, Teradyne’s 2027 CPO revenue could reach US$800 million. The company has an absolute competitive advantage in full-flow CPO testing, covering Insertion 1/2/4 stages.\n\nThe current bottleneck in the CPO industry is concentrated in the OE stage, where SoIC process yields remain low, leading the industry to revise down near-term CPO shipment expectations. We recommend closely tracking CPO industry progress, as this business represents Teradyne’s biggest potential upside driver in 2026.\n\n$TER", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.005645347321345695, "ret_signed_m1d": -0.005645347321345695, "alpha_spy_m1d": -0.0028838803094638488, "alpha_c_m1d": 0.00043524202191280725, "bench_spy_m1d": -0.0027614670118818463, "bench_c_m1d": -0.006080589343258502, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.023102404447410185, "ret_signed_p1d": -0.023102404447410185, "alpha_spy_p1d": -0.019439024777668235, "alpha_c_p1d": -0.017159132830433466, "bench_spy_p1d": -0.0036633796697419507, "bench_c_p1d": -0.00594327161697672, "ret_p1w": 0.04154351294141567, "ret_signed_p1w": 0.04154351294141567, "alpha_spy_p1w": 0.017995368726860583, "alpha_c_p1w": -0.0697113807235441, "bench_spy_p1w": 0.02354814421455509, "bench_c_p1w": 0.11125489366495978, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1811, -0.2214, -0.2154, -0.1315, -0.1029, -0.1177, -0.0698, -0.0999, -0.0891, -0.0891, -0.1155, -0.0886, -0.0855, -0.0596, -0.0779, -0.0473, -0.0075, -0.0368, -0.0735, -0.0567, -0.1193, -0.1164, -0.1153, -0.2095, -0.1418, -0.1293, -0.1257, -0.1703, -0.1708, -0.1365, -0.1332, -0.1313, -0.1245, -0.158, -0.1201, -0.0732, -0.0639, -0.1392, -0.1442, -0.2, -0.1417, -0.0962, -0.1037, -0.1037, -0.0868, -0.0722, 0.0373, 0.0544, 0.0653, 0.0715, 0.0582, 0.0566, 0.0593, 0.1012, 0.0862, 0.0999, 0.1151, 0.1609, 0.2104, 0.1638, 0.1005, -0.1132, -0.0056, 0.0, -0.0231, 0.0338, 0.1073, 0.0252, 0.0415, 0.0614, 0.0377, 0.052, 0.0322, -0.0218, -0.0706, -0.0692, -0.0031, 0.0236, 0.0381, 0.0381, 0.127, 0.0884, 0.1082, 0.084, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2057828710536114453", "ticker_idx": 0, "ts": "2026-05-22T14:19:31+00:00", "symbol": "DELL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Dell has significant growth potential thanks to CPUs. CPU servers for Agentic AI share many components with general-purpose servers, but come with a higher ASP. This is a major tailwind for Dell, given its strength in general-purpose servers.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bullish DELL: Agentic AI CPU servers carry higher ASP and play to Dell's general-purpose server strength.", "resolved_tickers": ["DELL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Dell is up a lot today…\n\nDell has significant growth potential thanks to CPUs.\n\nCPU servers for Agentic AI share many components with general-purpose servers, but come with a higher ASP.\n\nThis is a major tailwind for Dell, given its strength in general-purpose servers.\n\n$DELL\n\n---\n\nAnother major positive is that Agentic AI CPU servers do not necessarily require relying on Taiwanese ODMs or outsourcing work to them.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.14360242230108267, "ret_signed_m1d": -0.14360242230108267, "alpha_spy_m1d": -0.13968626451991895, "alpha_c_m1d": -0.1336795147552693, "bench_spy_m1d": -0.0039161577811637205, "bench_c_m1d": -0.009922907545813375, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": 0.0, "alpha_c_p1d": 0.0, "bench_spy_p1d": 0.0, "bench_c_p1d": 0.0, "ret_p1w": 0.4258951867641856, "ret_signed_p1w": 0.4258951867641856, "alpha_spy_p1w": 0.4113573862071387, "alpha_c_p1w": 0.3669672810112101, "bench_spy_p1w": 0.014537800557046898, "bench_c_p1w": 0.05892790575297546, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5955, -0.583, -0.5898, -0.4999, -0.4814, -0.5097, -0.5032, -0.5052, -0.5053, -0.5052, -0.5144, -0.5023, -0.4937, -0.4879, -0.4713, -0.4833, -0.4961, -0.4706, -0.4675, -0.4441, -0.4025, -0.3786, -0.4062, -0.4198, -0.4439, -0.4457, -0.428, -0.4111, -0.4111, -0.4151, -0.3999, -0.3736, -0.3872, -0.3995, -0.359, -0.3769, -0.4013, -0.3479, -0.3362, -0.3102, -0.2806, -0.2728, -0.2813, -0.268, -0.2684, -0.3024, -0.3033, -0.2922, -0.288, -0.283, -0.2672, -0.191, -0.2199, -0.1177, -0.1631, -0.1906, -0.1739, -0.1602, -0.1802, -0.1936, -0.203, -0.177, -0.1436, 0.0, 0.0, 0.0335, 0.0343, 0.0741, 0.4259, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2045853008081420654", "ticker_idx": 0, "ts": "2026-04-19T13:12:21+00:00", "symbol": "MRVL", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "GOOGLE IS DESIGNING TWO ASICS WITH MARVELL", "tweet_type": "news_reaction", "timeline": "unspecified", "summary": "Shares news that Google is designing two ASICs (MPU + TPU) with Marvell; implicit bullish endorsement via cashtag.", "resolved_tickers": ["MRVL"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "THE INFORMATION: GOOGLE IS DESIGNING TWO ASICS WITH MARVELL\n\nTHE INFORMATION: ONE IS A MEMORY PROCESSING UNIT DESIGNED TO WORK ALONGSIDE GOOGLE’S TENSOR PROCESSING UNITS (TPUS), AND THE OTHER IS A NEW TPU OPTIMIZED FOR RUNNING AI MODELS.\n\n$MRVL\n\n---\n\nhttps://t.co/DvJgUpZbma", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05512712458310287, "ret_signed_m1d": -0.05512712458310287, "alpha_spy_m1d": -0.05713079881836358, "alpha_c_m1d": -0.055558222545674996, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": 0.0004310979625721245, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.023471329184643563, "ret_signed_p1d": 0.023471329184643563, "alpha_spy_p1d": 0.030018278339579818, "alpha_c_p1d": 0.0219625520919553, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0015087770926882627, "ret_p1w": 0.07014347018972966, "ret_signed_p1w": 0.07014347018972966, "alpha_spy_p1w": 0.06104253826504835, "alpha_c_p1w": -0.02102822368388413, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.09117169387361379, "ret_p1m": 0.19230254761095278, "ret_signed_p1m": 0.19230254761095278, "alpha_spy_p1m": 0.15701356070571237, "alpha_c_p1m": 0.019873821016056503, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.17242872659489628, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4419, -0.4382, -0.4576, -0.4472, -0.4393, -0.4347, -0.4501, -0.4664, -0.4682, -0.4893, -0.5015, -0.4983, -0.4573, -0.4433, -0.4456, -0.4501, -0.4711, -0.4685, -0.4685, -0.4658, -0.4653, -0.4618, -0.4627, -0.4741, -0.47, -0.4529, -0.4639, -0.4477, -0.4533, -0.476, -0.4721, -0.4884, -0.3944, -0.3736, -0.3692, -0.3886, -0.4073, -0.406, -0.3809, -0.3862, -0.4076, -0.3947, -0.4057, -0.3905, -0.3756, -0.3344, -0.3396, -0.3585, -0.4063, -0.3304, -0.2786, -0.2759, -0.2759, -0.2596, -0.2605, -0.2262, -0.1892, -0.1309, -0.1119, -0.0948, -0.0896, -0.0979, -0.0551, 0.0, 0.0235, 0.0641, 0.1199, 0.1114, 0.0701, 0.0365, 0.0591, 0.1171, 0.1157, 0.107, 0.1414, 0.1644, 0.0823, 0.1508, 0.1556, 0.1127, 0.2037, 0.235, 0.1965, 0.1427, 0.1923, 0.2635, 0.2898, 0.328, 0.328, 0.4087, 0.344, 0.3855, 0.3866, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2057255667900248509", "ticker_idx": 0, "ts": "2026-05-21T00:22:27+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The shortage of LPDDR5 — and of DRAM overall — is likely to intensify further over time.", "tweet_type": "analysis", "timeline": "quarters", "summary": "SoCAMM/LPDDR5 demand from Vera CPU + NVL72 likely exceeds 2026–2027 supply; DRAM shortage set to intensify, bullish MU and DRAM sector.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS comment on NVIDIA Vera CPU:\n\nOn today’s conference call, NVIDIA stated that it expects the standalone Vera CPU market to reach $20 billion in FY2027.\n\nThe unit price of Grace CPU is estimated at around $3,000–$5,000. Since Vera is the successor to Grace and is optimized for AI agentic workloads, we expect Vera to carry a higher ASP of roughly $5,000–$8,000 per unit.\n\nAssuming a Vera CPU ASP of $8,000, a $20 billion market would imply 2.5 million CPUs.\n\nIt remains unclear whether standalone Vera CPU sales will include the same SoCAMM capacity as NVL72. However, assuming the same capacity is applied, each Vera CPU would have 8 SoCAMM slots. Assuming 192GB per module, SoCAMM capacity per Vera CPU would be 1,536GB.\n\nTherefore, FY2027 SoCAMM demand for Vera CPU would be:\n\n2.5 million CPUs × 1,536GB = 3.84 billion GB, or 30.72 billion Gb.\n\nCY2026, which broadly overlaps with NVIDIA’s FY2027, SoCAMM supply from the three major DRAM makers is estimated at 30 billion Gb. Therefore, combined SoCAMM demand from NVIDIA’s standalone Vera CPU sales and VR NVL72 sales already appears likely to exceed the annual supply capacity of 30 billion Gb.\n\nIn our previous in-depth report, assuming CY2027 VR NVL72 shipments of 100,000 servers, we estimated the SoCAMM TAM at 44 billion Gb based on 192GB modules. If additional SoCAMM demand from standalone Vera CPU sales is added, the CY2027 SoCAMM TAM could exceed 80 billion Gb.\n\nAn annual 80 billion Gb of LPDDR5 would be nearly equivalent to the annual LPDDR5 TAM used for smartphones.\n\nThe shortage of LPDDR5 — and of DRAM overall — is likely to intensify further over time.\n\n$MU $DRAM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0395092327990334, "ret_signed_m1d": -0.0395092327990334, "alpha_spy_m1d": -0.03753006061960029, "alpha_c_m1d": -0.0338389649421017, "bench_spy_m1d": -0.0019791721794331085, "bench_c_m1d": -0.005670267856931699, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01456498614555568, "ret_signed_p1d": -0.01456498614555568, "alpha_spy_p1d": -0.01849654051395122, "alpha_c_p1d": -0.029427285167948458, "bench_spy_p1d": 0.00393155436839554, "bench_c_p1d": 0.014862299022392778, "ret_p1w": 0.21180953827115045, "ret_signed_p1w": 0.21180953827115045, "alpha_spy_p1w": 0.19581427040247323, "alpha_c_p1w": 0.1555476309451047, "bench_spy_p1w": 0.01599526786867722, "bench_c_p1w": 0.056261907326045746, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4479, -0.4517, -0.4373, -0.4549, -0.4591, -0.4587, -0.502, -0.4743, -0.4792, -0.5143, -0.4894, -0.4713, -0.4508, -0.4683, -0.4411, -0.4205, -0.3944, -0.3944, -0.4173, -0.4453, -0.4696, -0.4812, -0.4988, -0.5338, -0.5315, -0.5777, -0.5567, -0.5173, -0.5194, -0.5194, -0.5043, -0.5046, -0.4663, -0.4469, -0.4481, -0.4403, -0.389, -0.4014, -0.4, -0.4029, -0.4116, -0.4103, -0.3603, -0.3679, -0.3482, -0.3117, -0.3383, -0.3197, -0.3214, -0.2885, -0.2436, -0.16, -0.1253, -0.1515, -0.0201, 0.0436, 0.0059, 0.0545, 0.0183, -0.0491, -0.1057, -0.0831, -0.0395, 0.0, -0.0146, -0.0146, 0.1755, 0.2182, 0.2118, 0.2741, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2057255667900248509", "ticker_idx": 1, "ts": "2026-05-21T00:22:27+00:00", "symbol": "DRAM", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "The shortage of LPDDR5 — and of DRAM overall — is likely to intensify further over time.", "tweet_type": "analysis", "timeline": "quarters", "summary": "SoCAMM/LPDDR5 demand from Vera CPU + NVL72 likely exceeds 2026–2027 supply; DRAM shortage set to intensify, bullish MU and DRAM sector.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "DRAM sector basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "KIS comment on NVIDIA Vera CPU:\n\nOn today’s conference call, NVIDIA stated that it expects the standalone Vera CPU market to reach $20 billion in FY2027.\n\nThe unit price of Grace CPU is estimated at around $3,000–$5,000. Since Vera is the successor to Grace and is optimized for AI agentic workloads, we expect Vera to carry a higher ASP of roughly $5,000–$8,000 per unit.\n\nAssuming a Vera CPU ASP of $8,000, a $20 billion market would imply 2.5 million CPUs.\n\nIt remains unclear whether standalone Vera CPU sales will include the same SoCAMM capacity as NVL72. However, assuming the same capacity is applied, each Vera CPU would have 8 SoCAMM slots. Assuming 192GB per module, SoCAMM capacity per Vera CPU would be 1,536GB.\n\nTherefore, FY2027 SoCAMM demand for Vera CPU would be:\n\n2.5 million CPUs × 1,536GB = 3.84 billion GB, or 30.72 billion Gb.\n\nCY2026, which broadly overlaps with NVIDIA’s FY2027, SoCAMM supply from the three major DRAM makers is estimated at 30 billion Gb. Therefore, combined SoCAMM demand from NVIDIA’s standalone Vera CPU sales and VR NVL72 sales already appears likely to exceed the annual supply capacity of 30 billion Gb.\n\nIn our previous in-depth report, assuming CY2027 VR NVL72 shipments of 100,000 servers, we estimated the SoCAMM TAM at 44 billion Gb based on 192GB modules. If additional SoCAMM demand from standalone Vera CPU sales is added, the CY2027 SoCAMM TAM could exceed 80 billion Gb.\n\nAn annual 80 billion Gb of LPDDR5 would be nearly equivalent to the annual LPDDR5 TAM used for smartphones.\n\nThe shortage of LPDDR5 — and of DRAM overall — is likely to intensify further over time.\n\n$MU $DRAM", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07282959769916442, "ret_signed_m1d": -0.07282959769916442, "alpha_spy_m1d": -0.07085042551973131, "alpha_c_m1d": -0.06715932984223272, "bench_spy_m1d": -0.0019791721794331085, "bench_c_m1d": -0.005670267856931699, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.012473929205065551, "ret_signed_p1d": -0.012473929205065551, "alpha_spy_p1d": -0.01640548357346109, "alpha_c_p1d": -0.027336228227458327, "bench_spy_p1d": 0.00393155436839554, "bench_c_p1d": 0.014862299022392778, "ret_p1w": 0.1306345871666137, "ret_signed_p1w": 0.1306345871666137, "alpha_spy_p1w": 0.11463931929793647, "alpha_c_p1w": 0.07437267984056795, "bench_spy_p1w": 0.01599526786867722, "bench_c_p1w": 0.056261907326045746, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4385, -0.4227, -0.4118, -0.3874, -0.3969, -0.3968, -0.456, -0.4876, -0.4519, -0.4703, -0.4934, -0.4539, -0.4418, -0.4543, -0.4535, -0.4299, -0.4161, -0.3851, -0.409, -0.4206, -0.456, -0.447, -0.4521, -0.4843, -0.4835, -0.513, -0.5275, -0.4739, -0.4986, -0.4821, -0.4676, -0.4588, -0.4103, -0.4171, -0.4103, -0.411, -0.377, -0.3704, -0.3595, -0.3667, -0.3648, -0.3494, -0.3346, -0.329, -0.3285, -0.2987, -0.309, -0.2995, -0.3074, -0.2965, -0.2405, -0.2126, -0.1373, -0.1308, -0.0848, -0.0114, -0.0389, 0.0071, 0.0073, -0.0694, -0.073, -0.0879, -0.0728, 0.0, -0.0125, -0.0125, 0.0772, 0.1332, 0.1306, 0.1784, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2056689560331890964", "ticker_idx": 0, "ts": "2026-05-19T10:52:57+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVDA's claimed 35x TCO advantage does not resonate strongly with the average AI engineer, and there is a widespread perception that its 70%+ gross margins are excessive. There is a clear willingness to improve economics by using ASICs or 'good enough' alternatives.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Evercore ISI report highlighting NVDA margin vulnerability and customer willingness to switch to ASICs/alternatives; implied bearish on pricing power.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "From Evercore ISI’s report:\n\nNVDA’s claimed 35x TCO advantage does not resonate strongly with the average AI engineer, and there is a widespread perception that its 70%+ gross margins are excessive. There is a clear willingness to improve economics by using ASICs or “good enough” alternatives.\n\nSome hyperscalers push back against NVDA’s 35x TCO advantage claim, arguing that the calculation does not account for power consumption around the chip, including cooling. The power component, including cooling, can account for 30–50% of total overhead costs.\n\nNo major issues have been observed at the hyperscaler level in the bring-up preparation for Rubin mass production.\n\nVera Rubin mass-production shipments are expected to be received by hyperscalers in 2Q26, while enterprise OEMs are expected to have access around September–October 2026.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.007751265623209269, "ret_signed_m1d": -0.007751265623209269, "alpha_spy_m1d": -0.0010457416889557702, "alpha_c_m1d": -0.003706940520241231, "bench_spy_m1d": 0.006705523934253499, "bench_c_m1d": 0.004044325102968038, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.01296405694410474, "ret_signed_p1d": -0.01296405694410474, "alpha_spy_p1d": -0.002715028375794404, "alpha_c_p1d": 0.025090120450783848, "bench_spy_p1d": 0.010249028568310337, "bench_c_p1d": 0.03805417739488859, "ret_p1w": -0.026064094937182092, "ret_signed_p1w": 0.026064094937182092, "alpha_spy_p1w": 0.04904263859154867, "alpha_c_p1w": 0.1330204710046604, "bench_spy_p1w": 0.02297854365436658, "bench_c_p1w": 0.10695637606747832, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1483, -0.1396, -0.1318, -0.1259, -0.1136, -0.162, -0.1969, -0.1729, -0.1839, -0.1703, -0.169, -0.194, -0.1721, -0.1625, -0.1567, -0.1698, -0.1829, -0.1695, -0.1753, -0.1823, -0.1906, -0.2172, -0.2038, -0.2058, -0.1901, -0.2238, -0.2407, -0.2513, -0.2095, -0.2033, -0.1959, -0.1959, -0.1948, -0.1927, -0.1747, -0.1664, -0.145, -0.1419, -0.1092, -0.0985, -0.1009, -0.0858, -0.0841, -0.094, -0.0821, -0.0951, -0.0559, -0.0181, -0.0337, -0.0515, -0.0954, -0.1004, -0.1003, -0.1093, -0.0579, -0.0413, -0.0245, -0.0053, 0.0008, 0.0237, 0.0686, 0.0213, 0.0078, 0.0, 0.013, -0.005, -0.0239, -0.0239, -0.0261, -0.0363, -0.0288, -0.0429, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2045806552490336763", "ticker_idx": 0, "ts": "2026-04-19T10:07:46+00:00", "symbol": "ASML.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$ASML", "tweet_type": "analysis", "timeline": "years", "summary": "TSMC's expected High-NA EUV adoption confirmation (A14 node, volume ~2029) is bullish for ASML as the sole High-NA EUV supplier.", "resolved_tickers": ["ASML.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "original", "lang": "en", "tweet_text": "An interesting sell-side prediction I came across: TSMC is expected to unveil its A14 enhanced node on April 22, and this node will likely signal the initial adoption of High-NA EUV at limited volume.\n\nIf realized, this would mark TSMC's first official confirmation of High-NA EUV adoption.\n\nFor context, TSMC announced the use of EUV for its N7+ node at its 2017 Symposium — roughly two years ahead of its 2019 volume production.\n\nGiven that TSMC's A14+ volume production is estimated for around 2029, it is plausible that TSMC will share details on this process upgrade in 2026–27, mirroring the 2017 precedent.\n\n$ASML", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0008032201127731176, "ret_signed_m1d": -0.0008032201127731176, "alpha_spy_m1d": -0.002806894348033828, "alpha_c_m1d": -0.001234318075345242, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": 0.0004310979625721245, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.006425662636939999, "ret_signed_p1d": -0.006425662636939999, "alpha_spy_p1d": 0.00012128651799625523, "alpha_c_p1d": -0.007934439729628262, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0015087770926882627, "ret_p1w": -0.023870495054408525, "ret_signed_p1w": -0.023870495054408525, "alpha_spy_p1w": -0.032971426979089835, "alpha_c_p1w": -0.11504218892802232, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.09117169387361379, "ret_p1m": 0.005431120092117769, "ret_signed_p1m": 0.005431120092117769, "alpha_spy_p1m": -0.029857866813122635, "alpha_c_p1m": -0.1669976065027785, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.17242872659489628, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0737, -0.0564, -0.0549, -0.0551, -0.0233, -0.0419, -0.0439, -0.0249, -0.0175, -0.0451, -0.0851, -0.0779, -0.0424, -0.0323, -0.0418, -0.0299, -0.0524, -0.0439, -0.0402, -0.0368, -0.0002, -0.0055, 0.0085, 0.0034, 0.0148, 0.0349, -0.0101, -0.0093, -0.0278, -0.0668, -0.0363, -0.0474, -0.0787, -0.0782, -0.0361, -0.0371, -0.0519, -0.0525, -0.0382, -0.0411, -0.0434, -0.0614, -0.0938, -0.0559, -0.0349, -0.0268, -0.0636, -0.0792, -0.1068, -0.101, -0.0461, -0.0675, -0.0675, -0.0675, -0.1054, -0.026, -0.0082, 0.0201, 0.0119, 0.0315, -0.012, -0.018, -0.0008, 0.0, -0.0064, 0.0039, -0.017, 0.0059, -0.0239, -0.0567, -0.04, -0.016, -0.016, -0.0442, -0.0107, 0.0505, 0.0465, 0.0621, 0.051, 0.019, 0.0682, 0.1004, 0.0518, 0.0178, 0.0054, 0.0731, 0.0829, 0.1342, 0.1516, 0.1188, 0.108, 0.1197, 0.1147, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2058118644983595231", "ticker_idx": 0, "ts": "2026-05-23T09:31:37+00:00", "symbol": "Nittobo", "kind": "equity", "geography": null, "sub_sector": "glass fiber materials", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nittobo is a leading major supplier that commands the critical global supply of high-end low-thermal-expansion glass fiber", "tweet_type": "analysis", "timeline": "quarters", "summary": "Long Nittobo: dominant supplier of T-glass accelerating CapEx 50% to meet surging AI server and smartphone demand while defending market share.", "resolved_tickers": ["3110.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Nittobo = Nitto Boseki, TSE 3110.T, glass fiber supplier", "bench_c": "EWJ", "bench_c_reason": "country_fallback (suffix='T')", "bench_c_industry": "Textile Manufacturing", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Why No Price Hike Despite a Severe Shortage? Why doesn't Nittobo raise prices to reap windfall profits?\n\nThe AI server boom has triggered an explosion in global demand for high-end glass fiber materials. Yet even as the market tightens into deepening shortage, Nittobo (Nitto Boseki, 日東紡) has opted for a counterintuitive strategy. Management made clear that the company currently has no plans for further price increases on its low-CTE (low coefficient of thermal expansion) glass fabric (T-glass), emphasizing that at this stage it will focus on expanding capacity to \"defend market share.\"\n\nAccording to Japanese media reports, Nittobo is a leading major supplier that commands the critical global supply of high-end low-thermal-expansion glass fiber. The company said it had initially planned to prioritize expanding low-Dk (low dielectric) glass fiber capacity, but as demand from AI servers and edge devices such as smartphones outpaced expectations, the supply-demand gap for low-thermal-expansion T-glass widened sharply. In response, it has decided to accelerate the pace of expansion, adding glass fabric production lines at its Fukushima site in Japan while also installing new glass melting furnace equipment in Taiwan.\n\nNittobo noted that T-glass demand is currently expanding rapidly. Not only is demand for the \"thick fabric\" used in semiconductor packaging substrates for AI servers continuing to climb, but the growth rate of the \"thin fabric\" used in edge devices such as smartphones is exceeding initial expectations by an even wider margin. The company emphasized that it has reached an internal consensus to further scale up the relevant capital investment in order to meet the strong demand from both thick-fabric and thin-fabric customers simultaneously.\n\nDefending Market Share Over Short-Term Price Hikes\n\nOn the future price trajectory that the market is watching closely, Nittobo management responded that it currently has no plans to make further price adjustments to either T-glass or low-dielectric glass, and will continue selling at the prices set at the start of this fiscal year. The company reiterated that its top priority at present is expanding capacity to meet customer demand, thereby cementing its leadership position in the high-end materials market. It added that, so long as its products remain competitive and market demand holds firm, it will not hesitate to commit to capacity expansion.\n\nNittobo had previously announced that it would raise the capital expenditure (CapEx) budget in its FY2024–2027 medium-term management plan by a substantial 50%, from the originally planned ¥80 billion to ¥120 billion—a move intended to respond to product demand and shifts in the market environment.\n\nGlass fabric is a critical component of IC substrates and printed circuit boards (PCBs), serving as the most fundamental chip-reinforcement layer material in electronic devices. As advanced chip packaging imposes extremely stringent requirements on micro-scale deformation, T-glass—with its low-thermal-expansion properties—can effectively prevent package warpage, establishing itself as a core element in ensuring the stability of high-end computing.\n\n---\n\nhttps://t.co/3d8WWJ5Xsf", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.0816400580551524, "ret_signed_m1d": -0.0816400580551524, "alpha_spy_m1d": -0.0816400580551524, "alpha_c_m1d": -0.0816400580551524, "bench_spy_m1d": 0.0, "bench_c_m1d": 0.0, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.06059506531204639, "ret_signed_p1d": -0.06059506531204639, "alpha_spy_p1d": -0.06723367390303658, "alpha_c_p1d": -0.07467650737003906, "bench_spy_p1d": 0.0066386085909901915, "bench_c_p1d": 0.014081442057992666, "ret_p1w": null, "ret_signed_p1w": null, "alpha_spy_p1w": null, "alpha_c_p1w": null, "bench_spy_p1w": null, "bench_c_p1w": null, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0003, -0.0801, -0.0899, -0.0924, -0.1574, -0.2148, -0.1621, -0.1401, -0.2531, -0.2354, -0.1917, -0.2163, -0.2004, -0.2098, -0.2488, -0.1946, -0.2206, -0.2206, -0.2733, -0.2741, -0.1924, -0.2123, -0.2235, -0.283, -0.32, -0.2427, -0.2885, -0.2569, -0.2467, -0.2141, -0.0983, -0.0824, -0.0047, 0.0508, 0.0682, -0.0385, -0.0189, -0.0018, -0.0305, -0.0399, -0.0036, -0.0729, -0.0065, -0.0076, 0.0236, 0.0236, 0.0356, 0.0218, 0.0218, 0.0218, 0.0218, 0.1593, 0.103, 0.0555, 0.0522, -0.0642, -0.0497, -0.1328, -0.1753, -0.217, -0.2348, -0.201, -0.0816, 0.0, -0.0606, -0.1546, -0.176, -0.1774, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2053807032223023576", "ticker_idx": 0, "ts": "2026-05-11T11:58:49+00:00", "symbol": "Kioxia", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Kioxia will surpass Toyota's market cap within this year", "tweet_type": "prediction", "timeline": "months", "summary": "Kioxia will surpass Toyota's market cap by year-end; strongly bullish on Kioxia's relative rise.", "resolved_tickers": ["285A.T"], "resolution_method": "exchange_ticker", "resolution_reason": "Kioxia Holdings 285A.T Japan NAND flash", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Kioxia will surpass Toyota’s market cap within this year.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03156290814105356, "ret_signed_m1d": -0.03156290814105356, "alpha_spy_m1d": -0.029290498153875633, "alpha_c_m1d": -0.014610192462619298, "bench_spy_m1d": -0.0022724099871779257, "bench_c_m1d": -0.01695271567843426, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.003482803656943867, "ret_signed_p1d": 0.003482803656943867, "alpha_spy_p1d": 0.0049977436483957804, "alpha_c_p1d": 0.029614568891612736, "bench_spy_p1d": -0.0015149399914519135, "bench_c_p1d": -0.02613176523466887, "ret_p1w": 0.11993905093600343, "ret_signed_p1w": 0.11993905093600343, "alpha_spy_p1w": 0.12081821147927163, "alpha_c_p1w": 0.17225468755828355, "bench_spy_p1w": -0.0008791605432681981, "bench_c_p1w": -0.052315636622280115, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5898, -0.5391, -0.5027, -0.5146, -0.5131, -0.5337, -0.5374, -0.5525, -0.5525, -0.5152, -0.5328, -0.5377, -0.5383, -0.5317, -0.5603, -0.5811, -0.5581, -0.5652, -0.6075, -0.574, -0.5346, -0.5375, -0.5415, -0.5077, -0.5296, -0.4909, -0.5133, -0.5133, -0.5328, -0.5407, -0.5114, -0.5393, -0.5588, -0.5642, -0.5847, -0.5255, -0.5408, -0.5244, -0.5037, -0.4935, -0.3992, -0.397, -0.3439, -0.3193, -0.2381, -0.2945, -0.2627, -0.3354, -0.3359, -0.2873, -0.2425, -0.2299, -0.2473, -0.2279, -0.2094, -0.2094, -0.1824, -0.2074, -0.2074, -0.2074, -0.2074, -0.0551, -0.0316, 0.0, 0.0035, 0.0993, 0.0549, -0.0324, 0.1199, 0.0834, 0.1165, 0.2046, 0.2495, 0.4247, 0.3596, 0.318, 0.3339, 0.4334, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2038928949917700579", "ticker_idx": 0, "ts": "2026-03-31T10:38:37+00:00", "symbol": "DB HiTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "DB HiTek, Korea's leading 8-inch wafer foundry, which has guided for 98% annual utilization this year. Utilization climbed from 76% in 2024 to 92% last year, with Q4 alone hitting 98%", "tweet_type": "analysis", "timeline": "quarters", "summary": "DB HiTek named beneficiary of Chinese fabless order diversification away from SMIC; 98% utilization guidance bullish for Korean 8-inch foundry.", "resolved_tickers": ["000990.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "DB HiTek 000990.KS Korea KOSPI 8-inch foundry", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "SMIC's 8-Inch Price Hike Faces 'Headwinds'… China-Origin Volume Shifts to Korean Foundries\n\nFollowing its price hike on 8-inch foundry services late last year, China's largest foundry SMIC reportedly resorted to rebates (returning overpaid amounts) after facing pushback from customers.\n\nAmong Chinese fabless companies, a clear trend is emerging of diversifying chip production away from domestic foundries toward overseas alternatives. Analysts note that as China's semiconductor localization policies intensify, wafer volumes are migrating to 8-inch foundries in Taiwan and Korea, creating a tailwind for those players.\n\nAccording to semiconductor industry sources on March 31, SMIC notified its major domestic fabless clients in H2 last year of a 10–15% price increase on 8-inch BCD (Bipolar-CMOS-DMOS) process foundry services. The hike was driven by rising chip volumes under Beijing's semiconductor self-sufficiency push, combined with tightening supply as Samsung and TSMC scaled back their 8-inch lines — prompting SMIC to raise prices to improve profitability. SMIC's fab utilization indeed remained in the mid-90% range during H2 last year.\n\nHowever, as fabless clients pushed back harder than expected, SMIC reportedly offered rebates to key customers. As a result, the market estimates the effective price increase was only around 2–4%.\n\nAn industry source explained: \"SMIC had to push the hike because utilization matters to them too, but they saw some slippage at that point. Customer pushback was fiercer than anticipated — not something they could brush off — so they gave rebates. The market view is that the headline 10% increase was never really applied in practice.\"\n\nIndustry observers say it will be difficult for Chinese foundries to sustain full utilization on their 8-inch lines. While Samsung and TSMC are reducing 8-inch wafer capacity, the current high utilization rates are largely attributed to set makers building inventory ahead of anticipated semiconductor price increases.\n\nTrendForce noted: \"Uncertainty around consumer electronics demand in H2 this year could lead to downward revisions in supply chain shipment forecasts. Utilization on 8-inch lines is likely to vary across foundries, making broad-based price hikes difficult.\"\n\nIn contrast, domestic Korean 8-inch foundry utilization is expected to remain elevated this year. Chinese fabless firms, under Beijing's semiconductor self-sufficiency mandate, are reportedly diversifying orders to overseas foundries including Korea as they continue ramping chip volumes. The Chinese government currently requires that more than half of computing chips for public data centers be sourced from domestic suppliers, with a target of 70% domestic adoption by 2027.\n\nA prime example is DB HiTek, Korea's leading 8-inch wafer foundry, which has guided for 98% annual utilization this year. Utilization climbed from 76% in 2024 to 92% last year, with Q4 alone hitting 98%. China accounted for 62.7% of DB HiTek's annual revenue last year — well over half.\n\nAn industry source noted: \"While SMIC's business is concentrated around roughly 10 large domestic clients, DB HiTek operates a high-mix, low-volume model across about 400 customers. SMIC's client base is primarily local consumer electronics and mobile set makers, making it directly exposed to domestic demand cycles, whereas DB HiTek's structure is relatively more insulated from such swings.\"\n\nA DB HiTek representative said: \"Foundry pricing is part of our internal strategy, so it's hard to define precisely, but we are watching how other companies respond to market price movements. As for maintaining full utilization through year-end, we see lingering market uncertainty and plan to monitor and adapt accordingly.\"\n\n---\n\nhttps://t.co/DCTWeOitdv", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.029299363057324834, "ret_signed_m1d": 0.029299363057324834, "alpha_spy_m1d": 0.057546211465482666, "alpha_c_m1d": 0.08373336266459208, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.09554140127388533, "ret_signed_p1d": 0.09554140127388533, "alpha_spy_p1d": 0.08800693749099531, "alpha_c_p1d": 0.07318874749024284, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.11847133757961781, "ret_signed_p1w": 0.11847133757961781, "alpha_spy_p1w": 0.10481702832451534, "alpha_c_p1w": 0.07543534315428957, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.9439490445859873, "ret_signed_p1m": 0.9439490445859873, "alpha_spy_p1m": 0.8497829350040762, "alpha_c_p1m": 0.6409234980932841, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1466, -0.0696, -0.0507, -0.0507, -0.0633, -0.0948, -0.0923, -0.0658, -0.0355, 0.0339, 0.0339, 0.0756, 0.0819, 0.0693, 0.14, 0.1753, 0.1488, 0.145, 0.2069, 0.3218, 0.2763, 0.3356, 0.3003, 0.4253, 0.3988, 0.318, 0.3003, 0.2498, 0.222, 0.1842, 0.1968, 0.1741, 0.1741, 0.1741, 0.1741, 0.1539, 0.1741, 0.1741, 0.1539, 0.1475, 0.1867, 0.2081, 0.2081, 0.1425, -0.0241, 0.1993, 0.1362, 0.0599, 0.1172, 0.0828, 0.0943, 0.1019, 0.1197, 0.0917, 0.1197, 0.1197, 0.1223, 0.0637, 0.0624, 0.0497, -0.0076, -0.0076, 0.0293, 0.0, 0.0955, 0.0611, 0.1541, 0.1669, 0.1185, 0.2025, 0.1796, 0.2076, 0.2293, 0.2318, 0.2561, 0.3185, 0.3618, 0.614, 0.5541, 0.6293, 0.7503, 0.9363, 1.0255, 0.9707, 0.9439, 1.0191, 1.0191, 1.0599, 1.0599, 1.0637, 1.0064, 1.1414, 1.3783, 1.1121, 1.3643, 1.3287, 1.0828, 1.1771, 1.0038, 1.0637, 1.3057, 1.307, 1.307, 1.8089, 1.5732, 1.4752, 1.3631, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2039356580978991350", "ticker_idx": 0, "ts": "2026-04-01T14:57:52+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "pricing power is strengthening, and the 'normal range' of margins itself could structurally shift higher versus the past", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish MU: HBM/AI DRAM transitioning from bit-supply to bandwidth-performance business, enabling structural margin expansion and pricing power without customer pushback.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Micron's management presented an interesting view at their NDR with Cantor.\n\nThe key insight: the market is no longer focused on pricing on a per-bit basis as in the past, but rather on a per-bandwidth basis. In other words, while price-per-bit may rise in 2026, price-per-bandwidth will decline.\n\nThis is a remarkably interesting view.\n\nHistorically, the memory industry has largely been viewed through this lens:\n\n- How many bits of memory chips were sold\n- Whether price-per-bit went up or down\n- When supply increases, bit pricing falls, and margins eventually collapse\n\nBut in the AI era, as products like HBM have become critical, what customers are buying is no longer simple storage capacity — it's \"how fast can data be processed,\" i.e., bandwidth.\n\nFor example, whereas in the past the question was \"what's the capacity of this memory?\", today it has shifted to \"how much data per second can this system feed to the GPU?\" — or so Micron's management argues.\n\nThis is especially true in today's AI infrastructure, where the latter matters far more. Memory vendors are therefore being reframed — from simply selling bits to selling units of performance that enable high-performance computing.\n\nSo what does it mean when price-per-bit rises but price-per-bandwidth falls?\n\nThe essence is this:\n\n- Even though it may look like customers are paying more,\n- They are actually getting higher performance and more bandwidth,\n- So price-per-performance can actually improve.\n\nIn other words, memory vendors can raise ASPs, and from the customer's perspective, it's not a rip-off — it can be perceived as an improvement in price-performance.\n\nWhy does this structure matter? Because even as memory companies raise prices, the magnitude of bandwidth improvement is larger, creating a logical framework where memory price hikes don't face customer pushback. In effect, memory is now being evaluated on a TCO basis.\n\nTo sum it up in one line: HBM and AI DRAM are no longer a simple bit-supply business — they are becoming an integral part of GPU system performance. That's why pricing power is strengthening, and the \"normal range\" of margins itself could structurally shift higher versus the past.\n\n---\n\nP.S.: As expected, American companies are incredibly skilled at putting a positive spin on their own business. It's a realm Korean companies wouldn't dare attempt to match. Who would've thought they'd come up with a bandwidth-based bull thesis lol", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08158219183821347, "ret_signed_m1d": -0.08158219183821347, "alpha_spy_m1d": -0.07410407168081345, "alpha_c_m1d": -0.05971825506364048, "bench_spy_m1d": -0.0074781201574000145, "bench_c_m1d": -0.02186393677457299, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.004376827082850587, "ret_signed_p1d": -0.004376827082850587, "alpha_spy_p1d": -0.005277301510941745, "alpha_c_p1d": -0.005269768130444996, "bench_spy_p1d": 0.0009004744280911581, "bench_c_p1d": 0.0008929410475944088, "ret_p1w": 0.10569526773875171, "ret_signed_p1w": 0.10569526773875171, "alpha_spy_p1w": 0.07399692844199923, "alpha_c_p1w": 0.026735112342562317, "bench_spy_p1w": 0.031698339296752476, "bench_c_p1w": 0.07896015539618939, "ret_p1m": 0.40589904733887305, "ret_signed_p1m": 0.40589904733887305, "alpha_spy_p1m": 0.30911010915536674, "alpha_c_p1m": 0.11314705192433427, "bench_spy_p1m": 0.09678893818350631, "bench_c_p1m": 0.2927519954145388, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1429, -0.1518, -0.0668, -0.0773, -0.1114, -0.0623, -0.0601, -0.0812, -0.0942, -0.0853, -0.0143, -0.0143, -0.0082, 0.0574, 0.0804, 0.086, 0.0573, 0.1148, 0.1828, 0.1842, 0.1274, 0.1897, 0.1398, 0.031, 0.0404, 0.0725, 0.0421, 0.0143, 0.115, 0.1249, 0.1186, 0.1186, 0.0863, 0.1439, 0.1341, 0.1635, 0.1439, 0.1359, 0.1657, 0.1292, 0.1206, 0.1214, 0.0317, 0.089, 0.0789, 0.0062, 0.0579, 0.0954, 0.1377, 0.1015, 0.1579, 0.2005, 0.2546, 0.2547, 0.2072, 0.1492, 0.0988, 0.0748, 0.0383, -0.0341, -0.0293, -0.1252, -0.0816, 0.0, -0.0044, -0.0044, 0.0269, 0.0265, 0.1057, 0.1459, 0.1434, 0.1596, 0.2659, 0.2403, 0.243, 0.2371, 0.219, 0.2216, 0.3252, 0.3096, 0.3503, 0.426, 0.3709, 0.4094, 0.4059, 0.474, 0.5671, 0.7404, 0.8121, 0.7579, 1.0302, 1.1621, 1.0839, 1.1847, 1.1096, 0.97, 0.8528, 0.8995, 0.9899, 1.0718, 1.0416, 1.0416, 1.4354, 1.5239, 1.5106, 1.6397, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2052002910477938982", "ticker_idx": 0, "ts": "2026-05-06T12:29:52+00:00", "symbol": "Sandisk", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "For now, it looks high. Even DeepSeek's approach increases the importance of NAND, if anything — it doesn't reduce it.", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Sandisk; DeepSeek's approach increases NAND importance rather than reducing it, ceiling looks high.", "resolved_tickers": ["SNDK"], "resolution_method": "exchange_ticker", "resolution_reason": "SanDisk listed on Nasdaq as SNDK", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "reply_other", "lang": "en", "tweet_text": "@teortaxesTex For now, it looks high. Even DeepSeek’s approach increases the importance of NAND, if anything — it doesn’t reduce it.", "reply_text": null, "parent_text": "@jukan05 what's the ceiling for Sandisk?", "quoted_text": null, "parent_author": "teortaxesTex", "ret_m1d": -0.0025958057776612753, "ret_signed_m1d": -0.0025958057776612753, "alpha_spy_m1d": 0.011113087732158355, "alpha_c_m1d": 0.023281944784436126, "bench_spy_m1d": -0.01370889350981963, "bench_c_m1d": -0.0258777505620974, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.049660293409252376, "ret_signed_p1d": -0.049660293409252376, "alpha_spy_p1d": -0.04659418825192607, "alpha_c_p1d": -0.0476606678118191, "bench_spy_p1d": -0.0030661051573263043, "bench_c_p1d": -0.0019996255974332744, "ret_p1w": 0.02641881481722619, "ret_signed_p1w": 0.02641881481722619, "alpha_spy_p1w": 0.014863031773110436, "alpha_c_p1w": -0.013691797269946182, "bench_spy_p1w": 0.011555783044115753, "bench_c_p1w": 0.04011061208717237, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5759, -0.5862, -0.6159, -0.5749, -0.553, -0.5556, -0.5556, -0.5811, -0.5742, -0.5595, -0.539, -0.5273, -0.5471, -0.5515, -0.5377, -0.5494, -0.5609, -0.599, -0.5751, -0.5989, -0.626, -0.5825, -0.5611, -0.5351, -0.5611, -0.5308, -0.501, -0.4892, -0.4655, -0.4524, -0.4967, -0.5018, -0.5018, -0.5192, -0.5722, -0.5632, -0.594, -0.5494, -0.5087, -0.5024, -0.5024, -0.4861, -0.4959, -0.4462, -0.396, -0.3959, -0.3245, -0.3302, -0.3676, -0.3479, -0.3468, -0.3525, -0.3592, -0.3056, -0.3387, -0.2979, -0.241, -0.2891, -0.2452, -0.2223, -0.1581, -0.1093, -0.0026, 0.0, -0.0497, 0.1081, 0.0976, 0.0298, 0.0264, -0.0193, -0.0017, -0.0546, -0.0189, -0.0124, 0.0938, 0.0487, 0.0487, 0.1274, 0.1276, 0.1643, 0.2021, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2040960323952505309", "ticker_idx": 0, "ts": "2026-04-06T01:10:35+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Microsoft has recently been expanding its adoption of AMD platforms... it is expected to become one of the key customers for Helios platform-based rack-scale AI systems by year-end... Microsoft's cumulative purchase volume projected to reach at least 1 million units or more", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Shares bullish AMD thesis: Microsoft expanding AMD platform adoption with Helios rack-scale systems and 1M+ unit cumulative purchases projected.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Cloud Big Tech Accelerates 'Dual-Track' Strategy for AI Infrastructure Build-Out (Taiwan Commercial Times)\n\n• As AI compute demand surges, global cloud big tech and major CSPs have been pursuing a 'dual-track strategy' in hardware procurement since this year to build out AI infrastructure. DIGITIMES forecasts that dedicated ASIC-based AI server growth will outpace GPU-based AI servers, with expected growth of YoY +64% and YoY +44%, respectively. Within the high-end AI server market, shipment share is estimated at 38% ASIC and 62% GPU.\n\n• Looking at Tier 1 CSP strategies, Microsoft is expanding its adoption of AMD platform servers, moving away from its existing dependence on NVIDIA architecture. Meanwhile, Google and Amazon are aggressively pursuing proprietary ASIC-based AI rack systems centered on TPU and Trainium, respectively. In the related key supply chain, beyond Wiwynn, Quanta, and Hon Hai, companies such as Inventec and Sunda are also expected to benefit from ASIC AI project expansion.\n\n• DIGITIMES forecasts AI server shipments to grow from approximately 1.844 million units in 2025 to 2.723 million units in 2026, with penetration rising to 16.7%. High-end AI servers in particular are expected to grow YoY +67.5%, from 1.142 million to 1.914 million units. Strong growth is projected to continue in 2027 and 2028, at approximately YoY +60% and YoY +49% respectively, with shipments surpassing 3 million and 4.5 million units each year.\n\n• By ASIC and GPU breakdown, ASIC servers are forecast at 660,700 units (YoY +64.2%) and GPU AI servers at 1,064,100 units (YoY +43.8%) this year.\n\n• Microsoft has recently been expanding its adoption of AMD platforms. Beyond the existing MI200 series, it is expected to become one of the key customers for Helios platform-based rack-scale AI systems by year-end. Additionally, the next-generation MI300 series is expected to enter mass production and supply in H2 next year, with Microsoft's cumulative purchase volume projected to reach at least 1 million units or more.\n\n• Google continues to expand its proprietary AI ecosystem, including its large language model 'Gemini,' and its TPU project is expected to serve as a core growth driver for the ASIC AI server market this year. At the same time, Amazon's Trainium-based proprietary ASIC servers are also expected to exhibit rapid growth this year.\n\n$AMD\n\n---\n\nhttps://t.co/jkgv6JGcxH", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.012171826527978769, "ret_signed_m1d": -0.012171826527978769, "alpha_spy_m1d": -0.0074672666803550225, "alpha_c_m1d": -0.002928926481097993, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.006131375004192696, "ret_signed_p1d": 0.006131375004192696, "alpha_spy_p1d": 0.0056913008983439095, "alpha_c_p1d": -0.003768071436025755, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.12103737869823639, "ret_signed_p1w": 0.12103737869823639, "alpha_spy_p1w": 0.07980388923477388, "alpha_c_p1w": 0.0014354189084082591, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.6134981450778381, "ret_signed_p1m": 0.6134981450778381, "alpha_spy_p1m": 0.5150961485437806, "alpha_c_p1m": 0.29350726690330053, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0461, -0.0704, -0.0773, -0.0567, 0.0036, 0.0155, 0.0352, 0.0529, 0.0529, 0.0533, 0.1345, 0.1524, 0.1794, 0.1414, 0.1447, 0.1479, 0.1453, 0.0752, 0.1185, 0.0996, -0.0908, -0.1257, -0.0533, -0.019, -0.03, -0.03, -0.0647, -0.0584, -0.0584, -0.0777, -0.0911, -0.0763, -0.091, -0.1071, -0.0288, -0.0423, -0.0749, -0.0907, -0.0979, -0.1328, -0.0823, -0.0942, -0.126, -0.0795, -0.077, -0.0697, -0.1019, -0.1217, -0.1072, -0.1084, -0.0941, -0.0677, -0.0856, -0.0795, -0.0673, 0.0004, -0.0745, -0.0826, -0.1096, -0.0761, -0.0453, -0.0122, -0.0122, 0.0, 0.0061, 0.0529, 0.0748, 0.1129, 0.121, 0.1585, 0.1723, 0.2638, 0.2644, 0.2488, 0.2921, 0.3782, 0.3867, 0.5797, 0.5198, 0.4679, 0.5311, 0.61, 0.6375, 0.5512, 0.6135, 0.9138, 0.8551, 1.0674, 1.0837, 1.036, 1.0233, 1.0424, 0.9262, 0.912, 0.8805, 1.0328, 1.0419, 1.1233, 1.1233, 1.2885, 1.2506, 1.353, 1.344, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2040716472570536273", "ticker_idx": 0, "ts": "2026-04-05T09:01:36+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "SK Hynix is entering into long-term supply agreements (LTAs) for DRAM with major global AI companies including Microsoft (MS) and Google", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supply crunch driving LTAs with Big Tech; Samsung, SK Hynix, and Micron all beneficiaries with record earnings projected on surging demand.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "MS and Google Pursuing LTA with SK Hynix\n\nSK Hynix is entering into long-term supply agreements (LTAs) for DRAM with major global AI companies including Microsoft (MS) and Google. Big Tech is moving to lock in DRAM—a product known for high price volatility—over extended periods. As concerns over DRAM supply shortages intensify amid the global AI infrastructure investment boom, Big Tech is strategizing to secure stable access to memory, a critical component for data centers. Samsung Electronics and SK Hynix, the world's No. 1 and No. 2 DRAM makers, are ramping up capital expenditure in response to the growing supply-demand imbalance.\n\n◇ MS and Google Arrive with Long-Term DRAM Contracts\n\nAccording to industry sources on the 5th, SK Hynix is in final-stage negotiations with MS on a long-term supply agreement for DDR5. The deal, valued in the tens of trillions of won, would cover a three-year period starting this year. Discussions are reportedly centered on provisions such as a price floor to hedge against significant DRAM ASP declines during the contract period, and terms under which 10–30% of the total contract value would be paid upfront.\n\nSK Hynix is also in discussions with Google on a long-term supply agreement. The negotiations reportedly cover not only its flagship High Bandwidth Memory (HBM) but also commodity DRAM used in servers. MS and Google are also pursuing long-term memory supply agreements with Samsung Electronics. Micron Technology, the U.S.-based No. 3 DRAM maker behind Samsung and SK Hynix, reportedly signed a similar contract last month.\n\n◇ \"Securing Volume Matters More Than Price\"\n\nA long-term supply agreement is a contract in which volume and pricing are predetermined over an extended period. This type of agreement is typically employed when prices for a specific product surge or supply becomes scarce. Until now, large companies like MS and Google had not entered into annual or multi-year memory supply contracts with memory manufacturers, given that memory is a highly cyclical product subject to significant price volatility.\n\nNevertheless, Big Tech has recently been approaching Samsung Electronics and SK Hynix with LTA proposals one after another due to an increasingly acute memory shortage. Memory supply has tightened significantly as AI infrastructure investment expands worldwide, and prices have surged accordingly. According to DRAMeXchange, the DDR4 contract price, which stood at just $1.35 in March of last year, skyrocketed to $13.00 by the end of last month.\n\nMoreover, as the AI infrastructure hegemony race appears set to become protracted, analysts note that AI companies have shifted to a strategy of securing volume first. A senior semiconductor industry executive explained, \"Right now, the fact that prices have risen so much is secondary—the real issue is that DRAM volume itself is extremely difficult to procure.\"\n\n◇ Samsung and SK Hynix to Significantly Expand DRAM CAPEX\n\nSamsung Electronics and SK Hynix have decided to aggressively invest in DRAM capacity this year in light of these conditions. Samsung Electronics is focusing on ramping up volume of 10nm-class 6th-generation (1c) DRAM—used in HBM4—at its main DRAM production base, the Pyeongtaek campus in Gyeonggi Province. At its Hwaseong campus, the company is accelerating process migration to 10nm-class 5th-generation (1b) DRAM, which serves as the base die for SOCAMM and commodity DRAM modules.\n\nSK Hynix is addressing new HBM volume demand centered on M15X, its new production base in Cheongju, North Chungcheong Province. Its headquarters campus in Icheon, Gyeonggi Province, is also aggressively pursuing process migration to its most advanced 1c DRAM node.\n\nDriven by the DRAM supply crunch, both companies' Q1 earnings are estimated to reach all-time highs. In a report on the 3rd, Meritz Securities projected Samsung Electronics' Q1 estimated revenue at KRW 122 trillion and operating profit at KRW 54 trillion. The estimated operating profit is more than triple the previous Q1 record of KRW 14.12 trillion set in 2022. FnGuide estimates SK Hynix's Q1 revenue at KRW 46.6252 trillion and operating profit at KRW 31.5627 trillion—4.2x the KRW 7.4405 trillion reported in the year-ago quarter.\n\n---\n\nhttps://t.co/B2Mm6qvfRU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.011286664037370775, "ret_signed_m1d": -0.011286664037370775, "alpha_spy_m1d": -0.006582104189747029, "alpha_c_m1d": -0.0020437639904899996, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.033860062665668655, "ret_signed_p1d": 0.033860062665668655, "alpha_spy_p1d": 0.03341998855981987, "alpha_c_p1d": 0.023960616225450204, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.17381489427902386, "ret_signed_p1w": 0.17381489427902386, "alpha_spy_p1w": 0.13258140481556135, "alpha_c_p1w": 0.05421293448919573, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.6331828896337777, "ret_signed_p1m": 0.6331828896337777, "alpha_spy_p1m": 0.5347808930997202, "alpha_c_p1m": 0.31319201145924014, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1641, -0.1483, -0.1618, -0.1562, -0.1686, -0.1641, -0.1562, -0.1483, -0.1393, -0.1629, -0.1663, -0.1494, -0.1359, -0.1708, -0.0987, -0.0525, -0.03, 0.0241, -0.0649, 0.0218, 0.0139, -0.0514, -0.0548, -0.0007, -0.0131, -0.0311, 0.0004, -0.0086, -0.0086, -0.0086, -0.0086, 0.0072, 0.0691, 0.0714, 0.1322, 0.1469, 0.2404, 0.1975, 0.1975, 0.0598, -0.0418, 0.0621, 0.0429, -0.0564, 0.0587, 0.0779, 0.0497, 0.0271, 0.0993, 0.0948, 0.1919, 0.1433, 0.1366, 0.053, 0.1129, 0.123, 0.053, 0.053, -0.0147, -0.0892, 0.0124, -0.0632, -0.0113, 0.0, 0.0339, 0.1659, 0.1264, 0.1591, 0.1738, 0.2449, 0.2822, 0.3036, 0.2731, 0.316, 0.3815, 0.3804, 0.3826, 0.3792, 0.4582, 0.4673, 0.4594, 0.4515, 0.4515, 0.6332, 0.6332, 0.807, 0.8668, 0.9029, 1.1219, 1.0711, 1.2302, 1.2235, 1.053, 1.0767, 0.9695, 0.9695, 1.1896, 1.1907, 1.1907, 1.316, 1.5316, 1.584, 1.6336, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2040716472570536273", "ticker_idx": 1, "ts": "2026-04-05T09:01:36+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Samsung Electronics and SK Hynix have decided to aggressively invest in DRAM capacity this year", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supply crunch driving LTAs with Big Tech; Samsung, SK Hynix, and Micron all beneficiaries with record earnings projected on surging demand.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "MS and Google Pursuing LTA with SK Hynix\n\nSK Hynix is entering into long-term supply agreements (LTAs) for DRAM with major global AI companies including Microsoft (MS) and Google. Big Tech is moving to lock in DRAM—a product known for high price volatility—over extended periods. As concerns over DRAM supply shortages intensify amid the global AI infrastructure investment boom, Big Tech is strategizing to secure stable access to memory, a critical component for data centers. Samsung Electronics and SK Hynix, the world's No. 1 and No. 2 DRAM makers, are ramping up capital expenditure in response to the growing supply-demand imbalance.\n\n◇ MS and Google Arrive with Long-Term DRAM Contracts\n\nAccording to industry sources on the 5th, SK Hynix is in final-stage negotiations with MS on a long-term supply agreement for DDR5. The deal, valued in the tens of trillions of won, would cover a three-year period starting this year. Discussions are reportedly centered on provisions such as a price floor to hedge against significant DRAM ASP declines during the contract period, and terms under which 10–30% of the total contract value would be paid upfront.\n\nSK Hynix is also in discussions with Google on a long-term supply agreement. The negotiations reportedly cover not only its flagship High Bandwidth Memory (HBM) but also commodity DRAM used in servers. MS and Google are also pursuing long-term memory supply agreements with Samsung Electronics. Micron Technology, the U.S.-based No. 3 DRAM maker behind Samsung and SK Hynix, reportedly signed a similar contract last month.\n\n◇ \"Securing Volume Matters More Than Price\"\n\nA long-term supply agreement is a contract in which volume and pricing are predetermined over an extended period. This type of agreement is typically employed when prices for a specific product surge or supply becomes scarce. Until now, large companies like MS and Google had not entered into annual or multi-year memory supply contracts with memory manufacturers, given that memory is a highly cyclical product subject to significant price volatility.\n\nNevertheless, Big Tech has recently been approaching Samsung Electronics and SK Hynix with LTA proposals one after another due to an increasingly acute memory shortage. Memory supply has tightened significantly as AI infrastructure investment expands worldwide, and prices have surged accordingly. According to DRAMeXchange, the DDR4 contract price, which stood at just $1.35 in March of last year, skyrocketed to $13.00 by the end of last month.\n\nMoreover, as the AI infrastructure hegemony race appears set to become protracted, analysts note that AI companies have shifted to a strategy of securing volume first. A senior semiconductor industry executive explained, \"Right now, the fact that prices have risen so much is secondary—the real issue is that DRAM volume itself is extremely difficult to procure.\"\n\n◇ Samsung and SK Hynix to Significantly Expand DRAM CAPEX\n\nSamsung Electronics and SK Hynix have decided to aggressively invest in DRAM capacity this year in light of these conditions. Samsung Electronics is focusing on ramping up volume of 10nm-class 6th-generation (1c) DRAM—used in HBM4—at its main DRAM production base, the Pyeongtaek campus in Gyeonggi Province. At its Hwaseong campus, the company is accelerating process migration to 10nm-class 5th-generation (1b) DRAM, which serves as the base die for SOCAMM and commodity DRAM modules.\n\nSK Hynix is addressing new HBM volume demand centered on M15X, its new production base in Cheongju, North Chungcheong Province. Its headquarters campus in Icheon, Gyeonggi Province, is also aggressively pursuing process migration to its most advanced 1c DRAM node.\n\nDriven by the DRAM supply crunch, both companies' Q1 earnings are estimated to reach all-time highs. In a report on the 3rd, Meritz Securities projected Samsung Electronics' Q1 estimated revenue at KRW 122 trillion and operating profit at KRW 54 trillion. The estimated operating profit is more than triple the previous Q1 record of KRW 14.12 trillion set in 2022. FnGuide estimates SK Hynix's Q1 revenue at KRW 46.6252 trillion and operating profit at KRW 31.5627 trillion—4.2x the KRW 7.4405 trillion reported in the year-ago quarter.\n\n---\n\nhttps://t.co/B2Mm6qvfRU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03573278094251686, "ret_signed_m1d": -0.03573278094251686, "alpha_spy_m1d": -0.031028221094893116, "alpha_c_m1d": -0.029957131774632884, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.005775649167883978, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.017607457276022753, "ret_signed_p1d": 0.017607457276022753, "alpha_spy_p1d": 0.017167383170173967, "alpha_c_p1d": 0.012855345108415683, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.00475211216760707, "ret_p1w": 0.04091144484722942, "ret_signed_p1w": 0.04091144484722942, "alpha_spy_p1w": -0.0003220446162330859, "alpha_c_p1w": -0.02364479078560211, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.06455623563283153, "ret_p1m": 0.20403935784567584, "ret_signed_p1m": 0.20403935784567584, "alpha_spy_p1m": 0.10563736131161838, "alpha_c_p1m": -0.0068833381697475104, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.21092269601542335, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2713, -0.2827, -0.2817, -0.2827, -0.2889, -0.2749, -0.2563, -0.2305, -0.2284, -0.2496, -0.2274, -0.2129, -0.214, -0.214, -0.1757, -0.1607, -0.1695, -0.1705, -0.2227, -0.1344, -0.1261, -0.1767, -0.1804, -0.1401, -0.1432, -0.1328, -0.077, -0.0636, -0.0636, -0.0636, -0.0636, -0.0181, -0.0176, -0.0026, 0.0336, 0.0517, 0.1266, 0.1189, 0.1189, 0.0083, -0.1101, -0.0098, -0.0274, -0.1034, -0.0289, -0.0181, -0.0289, -0.0517, -0.0248, 0.0021, 0.0775, 0.0362, 0.0305, -0.0372, -0.0196, -0.0233, -0.0692, -0.0692, -0.087, -0.1341, -0.0179, -0.0761, -0.0357, 0.0, 0.0176, 0.0901, 0.0564, 0.0668, 0.0409, 0.0694, 0.0927, 0.1264, 0.1186, 0.1108, 0.1341, 0.1264, 0.1626, 0.1367, 0.1626, 0.1497, 0.1704, 0.1419, 0.1419, 0.204, 0.204, 0.3775, 0.406, 0.3905, 0.4785, 0.4448, 0.4707, 0.5329, 0.4008, 0.4552, 0.4267, 0.4293, 0.551, 0.5148, 0.5148, 0.5484, 0.5898, 0.551, 0.6416, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2040716472570536273", "ticker_idx": 2, "ts": "2026-04-05T09:01:36+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": "DRAM", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Micron Technology, the U.S.-based No. 3 DRAM maker behind Samsung and SK Hynix, reportedly signed a similar contract last month", "tweet_type": "analysis", "timeline": "quarters", "summary": "DRAM supply crunch driving LTAs with Big Tech; Samsung, SK Hynix, and Micron all beneficiaries with record earnings projected on surging demand.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "MS and Google Pursuing LTA with SK Hynix\n\nSK Hynix is entering into long-term supply agreements (LTAs) for DRAM with major global AI companies including Microsoft (MS) and Google. Big Tech is moving to lock in DRAM—a product known for high price volatility—over extended periods. As concerns over DRAM supply shortages intensify amid the global AI infrastructure investment boom, Big Tech is strategizing to secure stable access to memory, a critical component for data centers. Samsung Electronics and SK Hynix, the world's No. 1 and No. 2 DRAM makers, are ramping up capital expenditure in response to the growing supply-demand imbalance.\n\n◇ MS and Google Arrive with Long-Term DRAM Contracts\n\nAccording to industry sources on the 5th, SK Hynix is in final-stage negotiations with MS on a long-term supply agreement for DDR5. The deal, valued in the tens of trillions of won, would cover a three-year period starting this year. Discussions are reportedly centered on provisions such as a price floor to hedge against significant DRAM ASP declines during the contract period, and terms under which 10–30% of the total contract value would be paid upfront.\n\nSK Hynix is also in discussions with Google on a long-term supply agreement. The negotiations reportedly cover not only its flagship High Bandwidth Memory (HBM) but also commodity DRAM used in servers. MS and Google are also pursuing long-term memory supply agreements with Samsung Electronics. Micron Technology, the U.S.-based No. 3 DRAM maker behind Samsung and SK Hynix, reportedly signed a similar contract last month.\n\n◇ \"Securing Volume Matters More Than Price\"\n\nA long-term supply agreement is a contract in which volume and pricing are predetermined over an extended period. This type of agreement is typically employed when prices for a specific product surge or supply becomes scarce. Until now, large companies like MS and Google had not entered into annual or multi-year memory supply contracts with memory manufacturers, given that memory is a highly cyclical product subject to significant price volatility.\n\nNevertheless, Big Tech has recently been approaching Samsung Electronics and SK Hynix with LTA proposals one after another due to an increasingly acute memory shortage. Memory supply has tightened significantly as AI infrastructure investment expands worldwide, and prices have surged accordingly. According to DRAMeXchange, the DDR4 contract price, which stood at just $1.35 in March of last year, skyrocketed to $13.00 by the end of last month.\n\nMoreover, as the AI infrastructure hegemony race appears set to become protracted, analysts note that AI companies have shifted to a strategy of securing volume first. A senior semiconductor industry executive explained, \"Right now, the fact that prices have risen so much is secondary—the real issue is that DRAM volume itself is extremely difficult to procure.\"\n\n◇ Samsung and SK Hynix to Significantly Expand DRAM CAPEX\n\nSamsung Electronics and SK Hynix have decided to aggressively invest in DRAM capacity this year in light of these conditions. Samsung Electronics is focusing on ramping up volume of 10nm-class 6th-generation (1c) DRAM—used in HBM4—at its main DRAM production base, the Pyeongtaek campus in Gyeonggi Province. At its Hwaseong campus, the company is accelerating process migration to 10nm-class 5th-generation (1b) DRAM, which serves as the base die for SOCAMM and commodity DRAM modules.\n\nSK Hynix is addressing new HBM volume demand centered on M15X, its new production base in Cheongju, North Chungcheong Province. Its headquarters campus in Icheon, Gyeonggi Province, is also aggressively pursuing process migration to its most advanced 1c DRAM node.\n\nDriven by the DRAM supply crunch, both companies' Q1 earnings are estimated to reach all-time highs. In a report on the 3rd, Meritz Securities projected Samsung Electronics' Q1 estimated revenue at KRW 122 trillion and operating profit at KRW 54 trillion. The estimated operating profit is more than triple the previous Q1 record of KRW 14.12 trillion set in 2022. FnGuide estimates SK Hynix's Q1 revenue at KRW 46.6252 trillion and operating profit at KRW 31.5627 trillion—4.2x the KRW 7.4405 trillion reported in the year-ago quarter.\n\n---\n\nhttps://t.co/B2Mm6qvfRU", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.030495603646339964, "ret_signed_m1d": -0.030495603646339964, "alpha_spy_m1d": -0.025791043798716218, "alpha_c_m1d": -0.02125270359945919, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00047655439619209705, "ret_signed_p1d": -0.00047655439619209705, "alpha_spy_p1d": -0.0009166285020408838, "alpha_c_p1d": -0.010376000836410548, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.12918251411324855, "ret_signed_p1w": 0.12918251411324855, "alpha_spy_p1w": 0.08794902464978605, "alpha_c_p1w": 0.00958055432342042, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.6947267991766328, "ret_signed_p1m": 0.6947267991766328, "alpha_spy_p1m": 0.5963248026425754, "alpha_c_p1m": 0.3747359210020953, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1015, -0.1347, -0.0869, -0.0848, -0.1053, -0.1179, -0.1093, -0.0401, -0.0401, -0.0342, 0.0296, 0.052, 0.0575, 0.0296, 0.0855, 0.1518, 0.1531, 0.0978, 0.1585, 0.1099, 0.0039, 0.0132, 0.0444, 0.0148, -0.0124, 0.0858, 0.0954, 0.0893, 0.0893, 0.0578, 0.1139, 0.1043, 0.133, 0.1139, 0.1061, 0.1352, 0.0996, 0.0912, 0.092, 0.0047, 0.0605, 0.0506, -0.0202, 0.0302, 0.0667, 0.1079, 0.0726, 0.1276, 0.169, 0.2217, 0.2218, 0.1756, 0.119, 0.0699, 0.0466, 0.011, -0.0594, -0.0548, -0.1481, -0.1057, -0.0262, -0.0305, -0.0305, 0.0, -0.0005, 0.0767, 0.1158, 0.1134, 0.1292, 0.2327, 0.2077, 0.2104, 0.2047, 0.187, 0.1896, 0.2904, 0.2752, 0.3149, 0.3886, 0.3349, 0.3725, 0.369, 0.4353, 0.526, 0.6947, 0.7646, 0.7117, 0.9769, 1.1054, 1.0293, 1.1274, 1.0542, 0.9183, 0.8042, 0.8497, 0.9377, 1.0174, 0.988, 0.988, 1.3716, 1.4577, 1.4447, 1.5704, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2044208909976776936", "ticker_idx": 0, "ts": "2026-04-15T00:19:18+00:00", "symbol": "3103.T", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Unitika (3103 JP) is definitely gonna go up even more", "tweet_type": "prediction", "timeline": "months", "summary": "Long Unitika; E-glass price surge driven by AI server PCB demand will push the stock higher.", "resolved_tickers": ["3103.T"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLB", "bench_c_reason": "industry='Specialty Chemicals'", "bench_c_industry": "Specialty Chemicals", "unit_kind": "quote", "lang": "en", "tweet_text": "Holy shit, E-glass is gonna double in price this year alone? Then Unitika (3103 JP) is definitely gonna go up even more. https://t.co/Q0QxhXflUB", "reply_text": null, "parent_text": null, "quoted_text": "Glass Fiber Cloth and Copper Foil Costs Continue to Rise… 2026 Full-Year CCL Price Hikes 'Far From Over'\n\nAs demand for high-end PCBs from AI servers and switches surges, the supply-demand imbalance in upstream materials is intensifying. Industry sources note that copper clad https://t.co/5HUAjqPxES", "parent_author": null, "ret_m1d": 0.03940413262854392, "ret_signed_m1d": 0.03940413262854392, "alpha_spy_m1d": 0.04723334720355854, "alpha_c_m1d": 0.027142605958220045, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": 0.012261526670323875, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.24026910139356072, "ret_signed_p1d": 0.24026910139356072, "alpha_spy_p1d": 0.2378117897709564, "alpha_c_p1d": 0.23306787671094042, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.0072012246826203, "ret_p1w": 0.3623258049014897, "ret_signed_p1w": 0.3623258049014897, "alpha_spy_p1w": 0.34622439693517837, "alpha_c_p1w": 0.35356751853355295, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.00875828636793674, "ret_p1m": -0.12782316194137433, "ret_signed_p1m": -0.12782316194137433, "alpha_spy_p1m": -0.1967290400173849, "alpha_c_p1m": -0.13346732493867797, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.005644162997303637, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.8121, -0.8174, -0.8333, -0.8251, -0.8289, -0.8126, -0.8059, -0.7674, -0.729, -0.7458, -0.6977, -0.7203, -0.6795, -0.6559, -0.6771, -0.629, -0.5569, -0.4849, -0.4849, -0.3407, -0.283, -0.1389, 0.0533, -0.1869, -0.3772, -0.4061, -0.4061, -0.3383, -0.3138, -0.223, -0.0783, -0.0625, -0.2076, -0.1519, -0.1725, -0.2018, -0.2734, -0.2859, -0.3099, -0.3758, -0.358, -0.3777, -0.3998, -0.2556, -0.284, -0.284, -0.3623, -0.3671, -0.3287, -0.4498, -0.4286, -0.4161, -0.4488, -0.4099, -0.4262, -0.4652, -0.4805, -0.4974, -0.4464, -0.4411, -0.297, -0.1528, 0.0394, 0.0, 0.2403, 0.4825, 0.7996, 0.4632, 0.3623, 0.37, 0.2326, 0.1927, 0.0735, 0.0735, 0.1716, 0.1446, 0.1446, 0.1446, 0.1446, 0.2182, 0.2345, 0.1091, 0.0865, 0.1124, -0.1278, -0.32, -0.3681, -0.3215, -0.383, -0.3878, -0.2436, -0.2432, -0.2763, -0.308, -0.3345, -0.322, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2047501746642272653", "ticker_idx": 0, "ts": "2026-04-24T02:23:51+00:00", "symbol": "000660.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Nomura sets a target price of KRW 2.34 million for SK hynix, implying 92% upside from the current share price", "tweet_type": "news_reaction", "timeline": "months", "summary": "Shares Nomura Buy-implied PT of KRW 2.34M on SK Hynix, 92% upside on DRAM/NAND price surge.", "resolved_tickers": ["000660.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Nomura sets a target price of KRW 2.34 million for SK hynix, implying 92% upside from the current share price.\n\n* Nomura expects DRAM and NAND prices to rise by 51% QoQ and 65% QoQ, respectively, in 2Q.\n* As a result, the company’s operating profit is expected to increase 79% QoQ to KRW 68 trillion, or USD 45.9 billion.\n* For the full year, Nomura now expects DRAM and NAND prices to rise by 175% and 280%, respectively, revised up from its previous forecasts of 166% and 206%.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0024549929888035837, "ret_signed_m1d": 0.0024549929888035837, "alpha_spy_m1d": 0.010144701109375709, "alpha_c_m1d": 0.05100960206263083, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.05728310153313276, "ret_signed_p1d": 0.05728310153313276, "alpha_spy_p1d": 0.05556029504443072, "alpha_c_p1d": 0.05763850923180103, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.052373115555525596, "ret_signed_p1w": 0.052373115555525596, "alpha_spy_p1w": 0.04297453591338418, "alpha_c_p1w": 0.045699067560662554, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.5883796452885026, "ret_signed_p1m": 0.5883796452885026, "alpha_spy_p1m": 0.5439781380209043, "alpha_c_p1m": 0.45039685454144385, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3465, -0.3131, -0.2967, -0.2575, -0.322, -0.2591, -0.2649, -0.3122, -0.3147, -0.2755, -0.2845, -0.2975, -0.2747, -0.2812, -0.2812, -0.2812, -0.2812, -0.2698, -0.2248, -0.2232, -0.1791, -0.1685, -0.1007, -0.1318, -0.1318, -0.2316, -0.3052, -0.23, -0.2439, -0.3159, -0.2324, -0.2185, -0.239, -0.2553, -0.2029, -0.2062, -0.1358, -0.171, -0.1759, -0.2365, -0.1931, -0.1858, -0.2365, -0.2365, -0.2856, -0.3396, -0.266, -0.3208, -0.2831, -0.275, -0.2504, -0.1547, -0.1833, -0.1596, -0.1489, -0.0974, -0.0704, -0.0548, -0.0769, -0.0458, 0.0016, 0.0008, 0.0025, 0.0, 0.0573, 0.0638, 0.0581, 0.0524, 0.0524, 0.1841, 0.1841, 0.3101, 0.3535, 0.3797, 0.5385, 0.5016, 0.617, 0.6121, 0.4885, 0.5057, 0.428, 0.428, 0.5876, 0.5884, 0.5884, 0.6792, 0.8355, 0.8735, 0.9095, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2051205348854542495", "ticker_idx": 0, "ts": "2026-05-04T07:40:39+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "BESI is regarded as the global leader in the hybrid bonding equipment market. It holds a technological edge in both precision and throughput", "tweet_type": "analysis", "timeline": "months", "summary": "Long BESI as global leader in hybrid bonding equipment, with Samsung adoption decision imminent and mass production of HBM expected next year.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Netherlands' BESI: Samsung's Hybrid Bonding Adoption Decision Coming Soon\n\nSamsung Electronics is expected to introduce hybrid bonding technology — projected to be at the heart of next-generation semiconductor production — around the middle of this year.\n\nDutch back-end equipment company BE Semiconductor Industries (BESI) stated during its Q1 earnings conference call on April 23 that \"we should see more definitive news around the middle of this year regarding Samsung Electronics' (hybrid bonder adoption).\" This is largely consistent with the outlook shared during the Q4/full-year 2025 earnings call in February, which indicated Samsung would adopt hybrid bonding in Q2.\n\nUnlike conventional thermo-compression (TC) bonding, hybrid bonding does not use fine metal bumps to connect stacked semiconductor dies. Instead, it is an advanced packaging technology that directly connects the copper (Cu) surfaces of two chips. When applied to High Bandwidth Memory (HBM) and new high-layer-count NAND flash, it can deliver performance improvements, reduced power consumption, and cost savings — because the direct copper-to-copper connection enables higher integration density and shorter interconnect lengths.\n\nThat said, some observers suggest that hybrid bonding adoption could be delayed as the Joint Electron Device Engineering Council (JEDEC) moves to relax packaging height limits for next-generation HBM. On this point, BESI CEO Richard Blickman explained, \"The only reason (JEDEC) is looking to relax HBM thickness specifications is because one of the three memory companies needs that height for its process,\" adding that \"the other two are not affected by this change.\" The major HBM players are Samsung Electronics, SK Hynix, and Micron. The implication is that two of these three are pursuing hybrid bonding adoption for reasons beyond height considerations.\n\nHybrid bonding produces chip-to-chip gaps of around 7 micrometers (㎛), which makes it advantageous for reducing the total thickness (height) of HBM packages. JEDEC is reportedly considering raising the height limit from the current 775㎛ to either 825㎛ or 900㎛. Analysts argue that if the package height limit is relaxed, the lifespan of existing TC bonding technology would be extended, while the introduction of hybrid bonding could be delayed — since hybrid bonding would no longer be required simply to reduce thickness.\n\nBlickman emphasized, \"The advantages of hybrid bonding over conventional TC bonding — faster circuit speeds, lower power consumption, and reduced heat — are being increasingly clearly demonstrated,\" and \"as Samsung Electronics has publicly disclosed, because of these technical benefits, (the relaxation of HBM thickness limits) has not changed the pace or rate of hybrid bonding adoption.\"\n\nSamsung Electronics introduced its hybrid bonding technology for HBM at NVIDIA's annual developer conference, GTC 2026, last March. At the time, Samsung stated that hybrid bonding improves thermal resistance by more than 20% versus TC bonding and supports 16-high and higher stacks. Interconnect density is also higher. According to BESI, hybrid copper bonding (HCB) offers 15× the interconnect density of TC bonding and more than 100× the energy-efficiency performance (EEP).\n\nBESI expects mass production of hybrid-bonding-based HBM to begin next year. Blickman explained, \"All three manufacturers are conducting equipment evaluations against the same criteria set by a specific end customer,\" and \"that end customer has asked the memory makers to secure hybrid-bonded stacked products by the end of this year, with the goal of bringing finished products to market next year.\" The end customer in question is most likely NVIDIA.\n\nBESI is regarded as the global leader in the hybrid bonding equipment market. It holds a technological edge in both precision and throughput, particularly in die bonding equipment. Applied Materials, the world's largest semiconductor equipment company, holds a 9% stake, making it BESI's largest shareholder.\n\n---\n\nhttps://t.co/biBH4yqVwq", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.011870625707155469, "ret_signed_m1d": 0.011870625707155469, "alpha_spy_m1d": 0.008193776342201309, "alpha_c_m1d": 0.005891820426450645, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03970526740512215, "ret_signed_p1d": 0.03970526740512215, "alpha_spy_p1d": 0.031683081520129885, "alpha_c_p1d": 0.008331338108181319, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.06958651055607756, "ret_signed_p1w": 0.06958651055607756, "alpha_spy_p1w": 0.03993514394867703, "alpha_c_p1w": -0.06759059996604044, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3591, -0.3461, -0.3268, -0.3191, -0.3124, -0.3048, -0.3138, -0.2833, -0.2739, -0.2674, -0.2393, -0.2963, -0.2495, -0.2367, -0.2178, -0.1962, -0.2316, -0.2296, -0.2332, -0.2623, -0.2231, -0.2326, -0.3642, -0.3235, -0.2898, -0.2884, -0.2871, -0.2471, -0.2556, -0.2334, -0.2174, -0.2522, -0.2739, -0.2574, -0.2511, -0.2454, -0.2459, -0.2888, -0.2914, -0.2721, -0.2371, -0.2265, -0.2265, -0.2265, -0.2216, -0.158, -0.1617, -0.1397, -0.1584, -0.1132, -0.1149, -0.1031, -0.0774, -0.0766, -0.073, -0.0551, -0.0164, 0.0259, 0.0151, -0.0368, -0.0201, 0.0119, 0.0119, 0.0, 0.0397, 0.0438, 0.0426, 0.0675, 0.0696, 0.0225, 0.0569, 0.0917, 0.0716, 0.0471, 0.0434, 0.0847, 0.1068, 0.1199, 0.1543, 0.1625, 0.1314, 0.1772, 0.1641, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2041302484682207692", "ticker_idx": 0, "ts": "2026-04-06T23:50:12+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "$AVGO", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Google TPU v8x tape-out delayed again at MediaTek; implied beneficiary is Broadcom (AVGO).", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "According to a DigiTimes report, the Google TPU product being handled by MediaTek is undergoing yet another engineering change order (ECO).\n\nThe tape-out is now expected around mid-2026, and the product undergoing the design revision is the v8x product that MediaTek is responsible for.\n\nIn other words, the tape-out has been delayed once again.\n\n$AVGO\n\n---\n\nhttps://t.co/y4nqMxhQ7U", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0003816274527959074, "ret_signed_m1d": 0.0003816274527959074, "alpha_spy_m1d": 0.005086187300419653, "alpha_c_m1d": 0.009624527499676683, "bench_spy_m1d": -0.004704559847623746, "bench_c_m1d": -0.009242900046880775, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.062144226060108165, "ret_signed_p1d": 0.062144226060108165, "alpha_spy_p1d": 0.06170415195425938, "alpha_c_p1d": 0.052244779619889714, "bench_spy_p1d": 0.0004400741058487867, "bench_c_p1d": 0.009899446440218451, "ret_p1w": 0.20774101976834092, "ret_signed_p1w": 0.20774101976834092, "alpha_spy_p1w": 0.16650753030487841, "alpha_c_p1w": 0.08813905997851279, "bench_spy_p1w": 0.0412334894634625, "bench_c_p1w": 0.11960195978982813, "ret_p1m": 0.35915782624537007, "ret_signed_p1m": 0.35915782624537007, "alpha_spy_p1m": 0.2607558297113126, "alpha_c_p1m": 0.03916694807083254, "bench_spy_p1m": 0.09840199653405746, "bench_c_p1m": 0.31999087817453753, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0902, 0.0552, 0.0948, 0.1178, 0.1254, 0.0787, 0.0886, 0.1162, 0.1162, 0.0556, 0.0435, 0.033, 0.0157, 0.031, 0.0562, 0.0576, 0.0496, 0.0514, 0.0508, 0.0166, -0.0223, -0.0145, 0.0566, 0.0916, 0.0805, 0.0878, 0.051, 0.032, 0.032, 0.0554, 0.0585, 0.06, 0.0557, 0.0484, 0.033, 0.0547, 0.021, 0.0142, 0.0118, -0.004, 0.0077, 0.0561, 0.0488, 0.0973, 0.0872, 0.084, 0.0663, 0.0224, 0.0312, 0.0197, 0.0027, 0.0151, -0.0145, 0.0257, 0.0123, 0.0139, -0.0159, -0.0437, -0.0669, -0.0156, -0.003, 0.0004, 0.0004, 0.0, 0.0621, 0.1151, 0.1287, 0.1817, 0.2077, 0.211, 0.2617, 0.2673, 0.2929, 0.271, 0.279, 0.3442, 0.3356, 0.3445, 0.33, 0.2716, 0.2895, 0.3276, 0.3398, 0.3246, 0.3592, 0.3531, 0.3121, 0.3676, 0.3626, 0.3335, 0.3255, 0.3987, 0.3523, 0.338, 0.3073, 0.3286, 0.3185, 0.3171, 0.3171, 0.3421, 0.3417, 0.3567, 0.4209, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2054660911311253804", "ticker_idx": 0, "ts": "2026-05-13T20:31:49+00:00", "symbol": "TSEM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think the customer Tower Semi secured this time is NVIDIA", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish Tower Semi on implied design win with NVIDIA as the newly secured customer.", "resolved_tickers": ["TSEM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I think the customer Tower Semi secured this time is NVIDIA.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.184436936177944, "ret_signed_m1d": -0.184436936177944, "alpha_spy_m1d": -0.1788732163083534, "alpha_c_m1d": -0.16485474433504932, "bench_spy_m1d": -0.005563719869590589, "bench_c_m1d": -0.01958219184289467, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03663631063350037, "ret_signed_p1d": 0.03663631063350037, "alpha_spy_p1d": 0.028742054364777614, "alpha_c_p1d": 0.0263648424835492, "bench_spy_p1d": 0.007894256268722755, "bench_c_p1d": 0.010271468149951168, "ret_p1w": -0.023229967435667853, "ret_signed_p1w": -0.023229967435667853, "alpha_spy_p1w": -0.021801995884590752, "alpha_c_p1w": -0.009604476521478666, "bench_spy_p1w": -0.0014279715510771007, "bench_c_p1w": -0.013625490914189187, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5245, -0.5245, -0.5281, -0.526, -0.5352, -0.5255, -0.5294, -0.5346, -0.5348, -0.5426, -0.5388, -0.4882, -0.5364, -0.5439, -0.5548, -0.5916, -0.5736, -0.5683, -0.5589, -0.5622, -0.5394, -0.4886, -0.4909, -0.4757, -0.3866, -0.3957, -0.3643, -0.3322, -0.303, -0.3549, -0.3651, -0.4143, -0.3519, -0.3091, -0.2695, -0.2695, -0.2976, -0.2951, -0.2478, -0.2473, -0.2481, -0.2264, -0.2093, -0.2036, -0.2069, -0.1637, -0.1709, -0.1996, -0.2388, -0.2567, -0.2594, -0.2491, -0.2852, -0.2644, -0.1836, -0.1949, -0.2057, -0.1671, -0.1991, -0.2244, -0.2207, -0.1547, -0.1844, 0.0, 0.0366, 0.0119, -0.0838, -0.072, -0.0232, 0.0304, 0.0468, 0.0468, 0.0656, 0.0259, 0.0175, -0.0574, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2057400706617987449", "ticker_idx": 0, "ts": "2026-05-21T09:58:47+00:00", "symbol": "CRWV", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "We initiate coverage with a Buy rating and a US$162 target price. CoreWeave is well positioned to become a long-term core beneficiary of the AI infrastructure boom.", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF Overseas initiates CRWV Buy, PT $162; bullish on AI compute demand, contract backlog, declining debt costs, and profitability inflection by 2028.", "resolved_tickers": ["CRWV"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": "Software - Infrastructure", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Communications\n\nCoreWeave Initiation: Riding the AI Infrastructure Supercycle\n\nWe initiate coverage with a Buy rating and a US$162 target price. Leveraging its first-mover advantage in GPU deployment, pure-play focus on AI workloads, industry-leading operational efficiency, and long-term supply agreements with leading AI cloud customers, CoreWeave is well positioned to become a long-term core beneficiary of the AI infrastructure boom.\n\nWe are bullish on five core drivers:\n\n1. AI compute demand continues to accelerate.\n2. The company’s remaining performance obligations provide strong support for 2026–2027 revenue expectations.\n3. Debt financing costs continue to decline.\n4. Large long-term contracts are gradually entering maturity, driving an inflection in profitability.\n5. Rising GPU rental prices are improving the profitability of newly signed contracts.\n\nLooking back, CoreWeave’s share-price moves have largely been driven by the signing of large contracts. We believe the next key catalysts will be new contract wins, or further disclosure of the scale of the Anthropic order.\n\nWe forecast CoreWeave’s 2026/2027/2028 revenue at US$12.3bn / US$25.6bn / US$39.6bn, respectively, with adjusted EBITDA reaching US$7.6bn / US$14.2bn / US$18.8bn. Our US$162 target price is based on 10x 2028E EV/EBITDA.\n\nAI data centers remain in a high-growth cycle, creating a vast market opportunity for neoclouds\n\n1. Driven by exponential growth in token consumption, global AI data center power capacity is expected to reach 283GW by 2030. Due to cash flow constraints and capex limitations, the top five global cloud providers are expected to self-build only around 53% of incremental compute capacity. After deducting compute capacity built by enterprises and sovereign AI projects, the TAM for neocloud service providers from 2026 to 2030 could reach as much as US$1.7tn.\n\n2. OpenAI and Anthropic have disclosed AI compute center buildout plans of more than 20GW and 10GW, respectively. CoreWeave is currently the only neocloud provider that has signed long-term agreements with both companies. Orders from OpenAI, Anthropic, Meta, and Microsoft are expected to contribute 60% / 56% / 47% of CoreWeave’s revenue in 2026/2027/2028, respectively, providing extremely strong earnings visibility.\n\nProfitability is better than expected\n\nWe expect the company to turn profitable in 2028.\n\n1. Most large contracts are currently still in the early construction and initial billing stages. As projects enter stable monthly billing and later commercial phases after contract completion, the pace of revenue recognition should accelerate significantly.\n\n2. New orders signed in 2026 are expected to account for around 40% of current remaining performance obligations. Benefiting from continued increases in GPU rental prices, once these orders formally enter the monthly billing stage, they should lift the company’s blended ASP by around 20%.\n\n3. CoreWeave’s financing costs continue to fall. The latest DDTL 4.0 interest rate is SOFR + 2.25%, down sharply from the rates on DDTL 1.0 / 2.0 / 3.0 of SOFR + 9.62% / +4.25% / +4.00%, respectively.\n\n$CRWV", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05856109820161037, "ret_signed_m1d": -0.05856109820161037, "alpha_spy_m1d": -0.056581926022177265, "alpha_c_m1d": -0.056581926022177265, "bench_spy_m1d": -0.0019791721794331085, "bench_c_m1d": -0.0019791721794331085, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.01942743940985736, "ret_signed_p1d": -0.01942743940985736, "alpha_spy_p1d": -0.0233589937782529, "alpha_c_p1d": -0.0233589937782529, "bench_spy_p1d": 0.00393155436839554, "bench_c_p1d": 0.00393155436839554, "ret_p1w": -0.006692705042279368, "ret_signed_p1w": -0.006692705042279368, "alpha_spy_p1w": -0.02268797291095659, "alpha_c_p1w": -0.02268797291095659, "bench_spy_p1w": 0.01599526786867722, "bench_c_p1w": 0.01599526786867722, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1556, -0.077, -0.089, -0.0925, -0.2605, -0.2745, -0.3142, -0.261, -0.3045, -0.3215, -0.3083, -0.3036, -0.2381, -0.2577, -0.246, -0.2019, -0.2367, -0.2302, -0.2502, -0.2427, -0.2381, -0.2283, -0.1859, -0.2522, -0.3046, -0.3572, -0.2799, -0.2709, -0.2355, -0.2355, -0.2476, -0.2077, -0.1736, -0.1448, -0.0519, 0.025, 0.0894, 0.1033, 0.1114, 0.0862, 0.0916, 0.0705, 0.1391, 0.0915, 0.0238, 0.0416, -0.0191, 0.0614, 0.0374, 0.1062, 0.1659, 0.1888, 0.2826, 0.1976, 0.0611, 0.0662, 0.0016, 0.0347, 0.0617, -0.0026, -0.0354, -0.0722, -0.0586, 0.0, -0.0194, -0.0194, -0.0157, -0.0308, -0.0067, 0.0181, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "quote:2051305126011576374", "ticker_idx": 0, "ts": "2026-05-04T14:17:08+00:00", "symbol": "AMD", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MI550 is positioning to challenge Rubin Ultra in 2H27. AMD's 2.5D/3D packaging expertise enables a 4-die/12 HBM4e design", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish AMD: MI550 packaging advantage + Rubin Ultra HBM4E delay means AMD can realistically compete in 2H27.", "resolved_tickers": ["AMD"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "quote", "lang": "en", "tweet_text": "\"MI550 is positioning to challenge Rubin Ultra in 2H27. AMD's 2.5D/3D packaging expertise enables a 4-die/12 HBM4e design, vs Rubin Ultra 2-die package (and MCM for 2 sets of 2-die).\"\n\nGiven that high-capacity HBM4E Rubin Ultra has been delayed, I think MI550 can realistically compete with Rubin Ultra.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.055630378651527757, "ret_signed_m1d": 0.055630378651527757, "alpha_spy_m1d": 0.051953529286573596, "alpha_c_m1d": 0.04965157337082293, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.04017099278984926, "ret_signed_p1d": 0.04017099278984926, "alpha_spy_p1d": 0.032148806904857, "alpha_c_p1d": 0.00879706349290843, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.34329799457324306, "ret_signed_p1w": 0.34329799457324306, "alpha_spy_p1w": 0.31364662796584253, "alpha_c_p1w": 0.20612088405112505, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4139, -0.4364, -0.3897, -0.3676, -0.3747, -0.3747, -0.397, -0.393, -0.393, -0.4054, -0.4141, -0.4046, -0.414, -0.4244, -0.3739, -0.3826, -0.4036, -0.4138, -0.4185, -0.4409, -0.4084, -0.416, -0.4366, -0.4066, -0.405, -0.4003, -0.421, -0.4338, -0.4244, -0.4252, -0.416, -0.399, -0.4105, -0.4066, -0.3987, -0.3551, -0.4034, -0.4086, -0.426, -0.4044, -0.3845, -0.3632, -0.3632, -0.3553, -0.3514, -0.3213, -0.3071, -0.2825, -0.2773, -0.2532, -0.2442, -0.1853, -0.1849, -0.195, -0.167, -0.1115, -0.106, 0.0184, -0.0202, -0.0537, -0.013, 0.0379, 0.0556, 0.0, 0.0402, 0.2338, 0.1959, 0.3328, 0.3433, 0.3126, 0.3044, 0.3167, 0.2417, 0.2326, 0.2123, 0.3105, 0.3164, 0.3688, 0.3688, 0.4753, 0.4509, 0.5169, 0.5111, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047558649179214315", "ticker_idx": 0, "ts": "2026-04-24T06:09:58+00:00", "symbol": "CLS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Celestica remains the primary assembly partner for TPU servers", "tweet_type": "analysis", "timeline": "months", "summary": "Google TPU8 ramp benefits Celestica, Inventec (share rising 10→30%), Foxconn, MediaTek, Quanta, and AVC in the ASIC server supply chain through 2026.", "resolved_tickers": ["CLS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Google ramps new TPU servers, Taiwan suppliers gain share\n\nGoogle's unveiling of its eighth-generation tensor processing unit (TPU) at Cloud Next 2026 is expected to drive the next wave of growth in the application-specific integrated circuit (ASIC) server supply chain, with Taiwanese manufacturers expanding their role, according to supply chain sources.\n\nAssembly partners, including Celestica and Foxconn, are expected to begin volume production in the second half of 2026. Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter, with Inventec reportedly set to begin production earlier.\n\nTaiwan gains share across supply chain\nTaiwanese vendors are taking on a larger role in Google's TPU server ecosystem. While Broadcom has historically dominated upstream IC design, MediaTek has also entered the supply chain, according to sources.\n\nIn server motherboard production, Inventec's share is reportedly rising from about 10 to 30%, gradually taking orders from Celestica. The companies declined to comment.\n\nSupply chain sources said Taiwanese firms remain competitive in terms of flexibility, quality, and pricing, supported by long-standing experience in server manufacturing. Inventec is currently the world's second-largest server motherboard supplier.\n\nASIC servers to outpace GPU growth\nASIC servers are expected to be the fastest-growing segment of the AI server market in 2026, outpacing systems based on GPUs led by Nvidia.\n\nDIGITIMES Research estimates ASIC server shipments will grow 64.2% year-over-year in 2026, compared with 43.8% growth for GPU servers.\n\nIndustry sources said rising ASIC shipments are also likely to support demand for general-purpose servers. Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026.\n\nDual-chip TPU highlights shift to inference\nGoogle's eighth-generation TPU introduces a dual-chip architecture for the first time, separating training and inference workloads.\n\nThe lineup includes TPU 8t for large-scale training and TPU 8i for low-latency inference. Both chips are expected to launch later in 2026, with shipments projected before year-end, according to supply chain estimates.\n\nThe move reflects a broader shift in AI development. Over the past three years, the industry has focused on large-scale training, but leading chipmakers, including Google and Nvidia, are placing greater emphasis on inference, indicating a widening range of AI applications.\n\nGoogle remains key ASIC player\nGoogle continues to be a major driver of ASIC-based AI infrastructure. DIGITIMES Research estimates the company will account for 46% of the ASIC accelerator market in 2026, with shipments reaching about 3.33 million units.\n\nThe TPU v7, code-named Ironwood, is expected to remain the main product, accounting for 89% of Google's total TPU shipments.\n\nIndustry sources said Celestica remains the primary assembly partner for TPU servers, but Taiwanese firms are expanding their role. This is most evident in CPU motherboard production, where Inventec's share is increasing, and in L10 to L11 assembly, where Foxconn is involved.\n\n$CLS\n\n---\n\nhttps://t.co/hHrcN2SlAp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.04536697237809839, "ret_signed_m1d": -0.04536697237809839, "alpha_spy_m1d": -0.037677264257526266, "alpha_c_m1d": -0.018029531050913317, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.029253309889976542, "ret_signed_p1d": 0.029253309889976542, "alpha_spy_p1d": 0.027530503401274498, "alpha_c_p1d": 0.027068775494464292, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": 0.021257408162516755, "ret_signed_p1w": 0.021257408162516755, "alpha_spy_p1w": 0.011858828520375342, "alpha_c_p1w": 0.010959106552772546, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": -0.1044343073798143, "ret_signed_p1m": -0.1044343073798143, "alpha_spy_p1m": -0.14883581464741258, "alpha_c_p1m": -0.23032369706444633, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1878, -0.1584, -0.2687, -0.315, -0.305, -0.2749, -0.3275, -0.2811, -0.2503, -0.2218, -0.2737, -0.2792, -0.3301, -0.3158, -0.3158, -0.3012, -0.2896, -0.2898, -0.2865, -0.2768, -0.2776, -0.2833, -0.3195, -0.3232, -0.3492, -0.3736, -0.3402, -0.352, -0.3917, -0.3468, -0.3438, -0.352, -0.3543, -0.3577, -0.3405, -0.3127, -0.3331, -0.3069, -0.344, -0.3004, -0.265, -0.2633, -0.3336, -0.3169, -0.3728, -0.3133, -0.2961, -0.2812, -0.2812, -0.2874, -0.2752, -0.2182, -0.1996, -0.1436, -0.1091, -0.063, -0.0689, -0.068, -0.0346, -0.0222, -0.0175, -0.0206, -0.0454, 0.0, 0.0293, -0.1186, -0.0821, -0.0015, 0.0213, 0.0256, 0.0182, 0.0089, -0.0609, -0.0845, -0.0712, -0.0881, -0.0918, -0.0698, -0.1259, -0.1646, -0.1733, -0.1554, -0.1352, -0.1044, -0.1044, -0.096, -0.128, -0.1443, -0.0605, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047558649179214315", "ticker_idx": 1, "ts": "2026-04-24T06:09:58+00:00", "symbol": "Inventec", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Inventec's share is reportedly rising from about 10 to 30%", "tweet_type": "analysis", "timeline": "months", "summary": "Google TPU8 ramp benefits Celestica, Inventec (share rising 10→30%), Foxconn, MediaTek, Quanta, and AVC in the ASIC server supply chain through 2026.", "resolved_tickers": ["2356.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Inventec Corporation 2356.TW Taiwan server ODM", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Google ramps new TPU servers, Taiwan suppliers gain share\n\nGoogle's unveiling of its eighth-generation tensor processing unit (TPU) at Cloud Next 2026 is expected to drive the next wave of growth in the application-specific integrated circuit (ASIC) server supply chain, with Taiwanese manufacturers expanding their role, according to supply chain sources.\n\nAssembly partners, including Celestica and Foxconn, are expected to begin volume production in the second half of 2026. Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter, with Inventec reportedly set to begin production earlier.\n\nTaiwan gains share across supply chain\nTaiwanese vendors are taking on a larger role in Google's TPU server ecosystem. While Broadcom has historically dominated upstream IC design, MediaTek has also entered the supply chain, according to sources.\n\nIn server motherboard production, Inventec's share is reportedly rising from about 10 to 30%, gradually taking orders from Celestica. The companies declined to comment.\n\nSupply chain sources said Taiwanese firms remain competitive in terms of flexibility, quality, and pricing, supported by long-standing experience in server manufacturing. Inventec is currently the world's second-largest server motherboard supplier.\n\nASIC servers to outpace GPU growth\nASIC servers are expected to be the fastest-growing segment of the AI server market in 2026, outpacing systems based on GPUs led by Nvidia.\n\nDIGITIMES Research estimates ASIC server shipments will grow 64.2% year-over-year in 2026, compared with 43.8% growth for GPU servers.\n\nIndustry sources said rising ASIC shipments are also likely to support demand for general-purpose servers. Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026.\n\nDual-chip TPU highlights shift to inference\nGoogle's eighth-generation TPU introduces a dual-chip architecture for the first time, separating training and inference workloads.\n\nThe lineup includes TPU 8t for large-scale training and TPU 8i for low-latency inference. Both chips are expected to launch later in 2026, with shipments projected before year-end, according to supply chain estimates.\n\nThe move reflects a broader shift in AI development. Over the past three years, the industry has focused on large-scale training, but leading chipmakers, including Google and Nvidia, are placing greater emphasis on inference, indicating a widening range of AI applications.\n\nGoogle remains key ASIC player\nGoogle continues to be a major driver of ASIC-based AI infrastructure. DIGITIMES Research estimates the company will account for 46% of the ASIC accelerator market in 2026, with shipments reaching about 3.33 million units.\n\nThe TPU v7, code-named Ironwood, is expected to remain the main product, accounting for 89% of Google's total TPU shipments.\n\nIndustry sources said Celestica remains the primary assembly partner for TPU servers, but Taiwanese firms are expanding their role. This is most evident in CPU motherboard production, where Inventec's share is increasing, and in L10 to L11 assembly, where Foxconn is involved.\n\n$CLS\n\n---\n\nhttps://t.co/hHrcN2SlAp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.03742202274765116, "ret_signed_m1d": -0.03742202274765116, "alpha_spy_m1d": -0.029732314627079037, "alpha_c_m1d": -0.010084581420466088, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.029105982444778467, "ret_signed_p1d": -0.029105982444778467, "alpha_spy_p1d": -0.03082878893348051, "alpha_c_p1d": -0.031290516840290716, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": -0.04573798374288596, "ret_signed_p1w": -0.04573798374288596, "alpha_spy_p1w": -0.05513656338502737, "alpha_c_p1w": -0.056036285352630166, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": 0.3596674345232176, "ret_signed_p1m": 0.3596674345232176, "alpha_spy_p1m": 0.3152659272556193, "alpha_c_p1m": 0.23377804483858555, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0541, -0.0416, -0.0634, -0.0644, -0.0884, -0.0842, -0.0686, -0.0842, -0.0894, -0.0686, -0.0634, -0.0572, -0.0572, -0.0572, -0.0572, -0.0572, -0.0572, -0.0572, -0.0572, -0.0676, -0.0655, -0.0353, -0.0385, -0.0385, -0.0696, -0.1091, -0.1632, -0.1289, -0.131, -0.1674, -0.1497, -0.106, -0.1071, -0.1133, -0.104, -0.0852, -0.078, -0.0967, -0.1102, -0.1279, -0.1216, -0.1081, -0.1071, -0.104, -0.1393, -0.1674, -0.1455, -0.1507, -0.1507, -0.1507, -0.1507, -0.1133, -0.0894, -0.079, -0.0738, -0.0728, -0.0759, -0.0644, -0.0489, -0.0312, -0.0156, -0.0073, -0.0374, 0.0, -0.0291, -0.0364, -0.0374, -0.0457, -0.0457, -0.0333, -0.0239, 0.0229, 0.0385, 0.0301, 0.0437, 0.0125, 0.0977, 0.106, 0.0852, 0.079, 0.0748, 0.0686, 0.1247, 0.237, 0.3597, 0.3015, 0.289, 0.3285, 0.4595, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047558649179214315", "ticker_idx": 2, "ts": "2026-04-24T06:09:58+00:00", "symbol": "Foxconn", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Foxconn is involved", "tweet_type": "analysis", "timeline": "months", "summary": "Google TPU8 ramp benefits Celestica, Inventec (share rising 10→30%), Foxconn, MediaTek, Quanta, and AVC in the ASIC server supply chain through 2026.", "resolved_tickers": ["2317.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Hon Hai Precision (Foxconn) 2317.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Google ramps new TPU servers, Taiwan suppliers gain share\n\nGoogle's unveiling of its eighth-generation tensor processing unit (TPU) at Cloud Next 2026 is expected to drive the next wave of growth in the application-specific integrated circuit (ASIC) server supply chain, with Taiwanese manufacturers expanding their role, according to supply chain sources.\n\nAssembly partners, including Celestica and Foxconn, are expected to begin volume production in the second half of 2026. Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter, with Inventec reportedly set to begin production earlier.\n\nTaiwan gains share across supply chain\nTaiwanese vendors are taking on a larger role in Google's TPU server ecosystem. While Broadcom has historically dominated upstream IC design, MediaTek has also entered the supply chain, according to sources.\n\nIn server motherboard production, Inventec's share is reportedly rising from about 10 to 30%, gradually taking orders from Celestica. The companies declined to comment.\n\nSupply chain sources said Taiwanese firms remain competitive in terms of flexibility, quality, and pricing, supported by long-standing experience in server manufacturing. Inventec is currently the world's second-largest server motherboard supplier.\n\nASIC servers to outpace GPU growth\nASIC servers are expected to be the fastest-growing segment of the AI server market in 2026, outpacing systems based on GPUs led by Nvidia.\n\nDIGITIMES Research estimates ASIC server shipments will grow 64.2% year-over-year in 2026, compared with 43.8% growth for GPU servers.\n\nIndustry sources said rising ASIC shipments are also likely to support demand for general-purpose servers. Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026.\n\nDual-chip TPU highlights shift to inference\nGoogle's eighth-generation TPU introduces a dual-chip architecture for the first time, separating training and inference workloads.\n\nThe lineup includes TPU 8t for large-scale training and TPU 8i for low-latency inference. Both chips are expected to launch later in 2026, with shipments projected before year-end, according to supply chain estimates.\n\nThe move reflects a broader shift in AI development. Over the past three years, the industry has focused on large-scale training, but leading chipmakers, including Google and Nvidia, are placing greater emphasis on inference, indicating a widening range of AI applications.\n\nGoogle remains key ASIC player\nGoogle continues to be a major driver of ASIC-based AI infrastructure. DIGITIMES Research estimates the company will account for 46% of the ASIC accelerator market in 2026, with shipments reaching about 3.33 million units.\n\nThe TPU v7, code-named Ironwood, is expected to remain the main product, accounting for 89% of Google's total TPU shipments.\n\nIndustry sources said Celestica remains the primary assembly partner for TPU servers, but Taiwanese firms are expanding their role. This is most evident in CPU motherboard production, where Inventec's share is increasing, and in L10 to L11 assembly, where Foxconn is involved.\n\n$CLS\n\n---\n\nhttps://t.co/hHrcN2SlAp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.015801354401805856, "ret_signed_m1d": 0.015801354401805856, "alpha_spy_m1d": 0.02349106252237798, "alpha_c_m1d": 0.04313879572899093, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02934537246049662, "ret_signed_p1d": 0.02934537246049662, "alpha_spy_p1d": 0.027622565971794577, "alpha_c_p1d": 0.02716083806498437, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": -0.009029345372460473, "ret_signed_p1w": -0.009029345372460473, "alpha_spy_p1w": -0.018427925014601887, "alpha_c_p1w": -0.019327646982204683, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": 0.17832957110609482, "ret_signed_p1m": 0.17832957110609482, "alpha_spy_p1m": 0.13392806383849654, "alpha_c_p1m": 0.052440181421462784, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [0.0181, 0.0181, 0.0113, -0.0045, -0.0316, -0.0226, -0.0113, -0.0271, -0.0293, -0.0135, -0.0023, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0248, 0.0316, 0.0474, 0.1129, 0.0971, 0.0971, 0.079, 0.0339, -0.0203, 0.0113, 0.0068, -0.0497, -0.0542, -0.0113, -0.0316, -0.0316, -0.0226, -0.0429, -0.0519, -0.0745, -0.0835, -0.1151, -0.1196, -0.0971, -0.0948, -0.0993, -0.1242, -0.1535, -0.1106, -0.1287, -0.1287, -0.1287, -0.1332, -0.0903, -0.0971, -0.0948, -0.0971, -0.0632, -0.0632, -0.0655, -0.07, -0.0632, -0.0474, -0.0023, 0.0158, 0.0, 0.0293, 0.0181, 0.0158, -0.009, -0.009, 0.0271, 0.0813, 0.1377, 0.1445, 0.1287, 0.1377, 0.1287, 0.1332, 0.1038, 0.1219, 0.1219, 0.1061, 0.0835, 0.1174, 0.1287, 0.1783, 0.1693, 0.1919, 0.1874, 0.3047, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047558649179214315", "ticker_idx": 3, "ts": "2026-04-24T06:09:58+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "MediaTek has also entered the supply chain", "tweet_type": "analysis", "timeline": "months", "summary": "Google TPU8 ramp benefits Celestica, Inventec (share rising 10→30%), Foxconn, MediaTek, Quanta, and AVC in the ASIC server supply chain through 2026.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Google ramps new TPU servers, Taiwan suppliers gain share\n\nGoogle's unveiling of its eighth-generation tensor processing unit (TPU) at Cloud Next 2026 is expected to drive the next wave of growth in the application-specific integrated circuit (ASIC) server supply chain, with Taiwanese manufacturers expanding their role, according to supply chain sources.\n\nAssembly partners, including Celestica and Foxconn, are expected to begin volume production in the second half of 2026. Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter, with Inventec reportedly set to begin production earlier.\n\nTaiwan gains share across supply chain\nTaiwanese vendors are taking on a larger role in Google's TPU server ecosystem. While Broadcom has historically dominated upstream IC design, MediaTek has also entered the supply chain, according to sources.\n\nIn server motherboard production, Inventec's share is reportedly rising from about 10 to 30%, gradually taking orders from Celestica. The companies declined to comment.\n\nSupply chain sources said Taiwanese firms remain competitive in terms of flexibility, quality, and pricing, supported by long-standing experience in server manufacturing. Inventec is currently the world's second-largest server motherboard supplier.\n\nASIC servers to outpace GPU growth\nASIC servers are expected to be the fastest-growing segment of the AI server market in 2026, outpacing systems based on GPUs led by Nvidia.\n\nDIGITIMES Research estimates ASIC server shipments will grow 64.2% year-over-year in 2026, compared with 43.8% growth for GPU servers.\n\nIndustry sources said rising ASIC shipments are also likely to support demand for general-purpose servers. Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026.\n\nDual-chip TPU highlights shift to inference\nGoogle's eighth-generation TPU introduces a dual-chip architecture for the first time, separating training and inference workloads.\n\nThe lineup includes TPU 8t for large-scale training and TPU 8i for low-latency inference. Both chips are expected to launch later in 2026, with shipments projected before year-end, according to supply chain estimates.\n\nThe move reflects a broader shift in AI development. Over the past three years, the industry has focused on large-scale training, but leading chipmakers, including Google and Nvidia, are placing greater emphasis on inference, indicating a widening range of AI applications.\n\nGoogle remains key ASIC player\nGoogle continues to be a major driver of ASIC-based AI infrastructure. DIGITIMES Research estimates the company will account for 46% of the ASIC accelerator market in 2026, with shipments reaching about 3.33 million units.\n\nThe TPU v7, code-named Ironwood, is expected to remain the main product, accounting for 89% of Google's total TPU shipments.\n\nIndustry sources said Celestica remains the primary assembly partner for TPU servers, but Taiwanese firms are expanding their role. This is most evident in CPU motherboard production, where Inventec's share is increasing, and in L10 to L11 assembly, where Foxconn is involved.\n\n$CLS\n\n---\n\nhttps://t.co/hHrcN2SlAp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.09034907597535935, "ret_signed_m1d": -0.09034907597535935, "alpha_spy_m1d": -0.08265936785478722, "alpha_c_m1d": -0.0417944669015321, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.04855460907382725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.0017228064887020444, "alpha_c_p1d": 0.00035540769866826416, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": -0.00035540769866826416, "ret_p1w": 0.07186858316221767, "ret_signed_p1w": 0.07186858316221767, "alpha_spy_p1w": 0.062470003520076256, "alpha_c_p1w": 0.06519453516735463, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.0066740479948630416, "ret_p1m": 0.7433264887063655, "ret_signed_p1m": 0.7433264887063655, "alpha_spy_p1m": 0.6989249814387672, "alpha_c_p1m": 0.6053436979593068, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.13798279074705877, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.271, -0.269, -0.269, -0.2772, -0.2998, -0.2628, -0.2608, -0.2731, -0.2977, -0.2485, -0.2423, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2628, -0.2505, -0.232, -0.2012, -0.2012, -0.2197, -0.2546, -0.2936, -0.271, -0.2752, -0.3162, -0.2998, -0.2752, -0.2669, -0.2936, -0.2977, -0.2895, -0.2895, -0.3101, -0.3018, -0.3326, -0.3347, -0.3347, -0.347, -0.3491, -0.3799, -0.3881, -0.3984, -0.3984, -0.3984, -0.3984, -0.3963, -0.3511, -0.3532, -0.3532, -0.3347, -0.2936, -0.2649, -0.2218, -0.2094, -0.2197, -0.1417, -0.0575, -0.0903, 0.0, 0.0, 0.0739, 0.0575, 0.0719, 0.0719, 0.1786, 0.2957, 0.4086, 0.4045, 0.4908, 0.5934, 0.5195, 0.4353, 0.3984, 0.3388, 0.3963, 0.2957, 0.3265, 0.4579, 0.5852, 0.7433, 0.7515, 0.9055, 0.8111, 0.77, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047558649179214315", "ticker_idx": 4, "ts": "2026-04-24T06:09:58+00:00", "symbol": "Quanta Computer", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026", "tweet_type": "analysis", "timeline": "months", "summary": "Google TPU8 ramp benefits Celestica, Inventec (share rising 10→30%), Foxconn, MediaTek, Quanta, and AVC in the ASIC server supply chain through 2026.", "resolved_tickers": ["2382.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Quanta Computer listed on TWSE as 2382.TW", "bench_c": "XLK", "bench_c_reason": "industry='Computer Hardware'", "bench_c_industry": "Computer Hardware", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Google ramps new TPU servers, Taiwan suppliers gain share\n\nGoogle's unveiling of its eighth-generation tensor processing unit (TPU) at Cloud Next 2026 is expected to drive the next wave of growth in the application-specific integrated circuit (ASIC) server supply chain, with Taiwanese manufacturers expanding their role, according to supply chain sources.\n\nAssembly partners, including Celestica and Foxconn, are expected to begin volume production in the second half of 2026. Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter, with Inventec reportedly set to begin production earlier.\n\nTaiwan gains share across supply chain\nTaiwanese vendors are taking on a larger role in Google's TPU server ecosystem. While Broadcom has historically dominated upstream IC design, MediaTek has also entered the supply chain, according to sources.\n\nIn server motherboard production, Inventec's share is reportedly rising from about 10 to 30%, gradually taking orders from Celestica. The companies declined to comment.\n\nSupply chain sources said Taiwanese firms remain competitive in terms of flexibility, quality, and pricing, supported by long-standing experience in server manufacturing. Inventec is currently the world's second-largest server motherboard supplier.\n\nASIC servers to outpace GPU growth\nASIC servers are expected to be the fastest-growing segment of the AI server market in 2026, outpacing systems based on GPUs led by Nvidia.\n\nDIGITIMES Research estimates ASIC server shipments will grow 64.2% year-over-year in 2026, compared with 43.8% growth for GPU servers.\n\nIndustry sources said rising ASIC shipments are also likely to support demand for general-purpose servers. Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026.\n\nDual-chip TPU highlights shift to inference\nGoogle's eighth-generation TPU introduces a dual-chip architecture for the first time, separating training and inference workloads.\n\nThe lineup includes TPU 8t for large-scale training and TPU 8i for low-latency inference. Both chips are expected to launch later in 2026, with shipments projected before year-end, according to supply chain estimates.\n\nThe move reflects a broader shift in AI development. Over the past three years, the industry has focused on large-scale training, but leading chipmakers, including Google and Nvidia, are placing greater emphasis on inference, indicating a widening range of AI applications.\n\nGoogle remains key ASIC player\nGoogle continues to be a major driver of ASIC-based AI infrastructure. DIGITIMES Research estimates the company will account for 46% of the ASIC accelerator market in 2026, with shipments reaching about 3.33 million units.\n\nThe TPU v7, code-named Ironwood, is expected to remain the main product, accounting for 89% of Google's total TPU shipments.\n\nIndustry sources said Celestica remains the primary assembly partner for TPU servers, but Taiwanese firms are expanding their role. This is most evident in CPU motherboard production, where Inventec's share is increasing, and in L10 to L11 assembly, where Foxconn is involved.\n\n$CLS\n\n---\n\nhttps://t.co/hHrcN2SlAp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.003095975232198178, "ret_signed_m1d": -0.003095975232198178, "alpha_spy_m1d": 0.004593732888373947, "alpha_c_m1d": 0.024241466094986897, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.007739938080495445, "ret_signed_p1d": 0.007739938080495445, "alpha_spy_p1d": 0.006017131591793401, "alpha_c_p1d": 0.005555403684983196, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": -0.03250773993808054, "ret_signed_p1w": -0.03250773993808054, "alpha_spy_p1w": -0.04190631958022195, "alpha_c_p1w": -0.04280604154782475, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": -0.020123839009287936, "ret_signed_p1m": -0.020123839009287936, "alpha_spy_p1m": -0.06452534627688622, "alpha_c_p1m": -0.14601322869391997, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1099, -0.1115, -0.1238, -0.1331, -0.1548, -0.144, -0.1393, -0.1548, -0.1347, -0.1207, -0.1176, -0.1146, -0.1146, -0.1146, -0.1146, -0.1146, -0.1146, -0.1146, -0.1146, -0.1223, -0.1207, -0.0789, -0.0975, -0.0975, -0.0681, -0.0882, -0.13, -0.1037, -0.1053, -0.1269, -0.1254, -0.1053, -0.1068, -0.1053, -0.1053, -0.1053, -0.0929, -0.113, -0.1223, -0.1331, -0.1285, -0.1176, -0.1115, -0.1053, -0.13, -0.1378, -0.1006, -0.1192, -0.1192, -0.1192, -0.1053, -0.048, -0.031, -0.0046, -0.0124, 0.0, -0.0449, -0.0139, 0.0, 0.0186, 0.0526, 0.0372, -0.0031, 0.0, 0.0077, -0.0062, -0.0031, -0.0325, -0.0325, -0.0155, -0.0062, 0.0728, 0.065, 0.0542, 0.0635, 0.0526, 0.0542, 0.0356, -0.0526, -0.0712, -0.1022, -0.1022, -0.0464, -0.0217, -0.0201, -0.0201, -0.0341, -0.0449, 0.0495, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2047558649179214315", "ticker_idx": 5, "ts": "2026-04-24T06:09:58+00:00", "symbol": "Asia Vital Components", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter", "tweet_type": "analysis", "timeline": "months", "summary": "Google TPU8 ramp benefits Celestica, Inventec (share rising 10→30%), Foxconn, MediaTek, Quanta, and AVC in the ASIC server supply chain through 2026.", "resolved_tickers": ["3017.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Asia Vital Components (AVC) 3017.TW Taiwan", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Exclusive: Google ramps new TPU servers, Taiwan suppliers gain share\n\nGoogle's unveiling of its eighth-generation tensor processing unit (TPU) at Cloud Next 2026 is expected to drive the next wave of growth in the application-specific integrated circuit (ASIC) server supply chain, with Taiwanese manufacturers expanding their role, according to supply chain sources.\n\nAssembly partners, including Celestica and Foxconn, are expected to begin volume production in the second half of 2026. Key component suppliers such as Asia Vital Components and Cooler Master are preparing shipments by the end of the second quarter, with Inventec reportedly set to begin production earlier.\n\nTaiwan gains share across supply chain\nTaiwanese vendors are taking on a larger role in Google's TPU server ecosystem. While Broadcom has historically dominated upstream IC design, MediaTek has also entered the supply chain, according to sources.\n\nIn server motherboard production, Inventec's share is reportedly rising from about 10 to 30%, gradually taking orders from Celestica. The companies declined to comment.\n\nSupply chain sources said Taiwanese firms remain competitive in terms of flexibility, quality, and pricing, supported by long-standing experience in server manufacturing. Inventec is currently the world's second-largest server motherboard supplier.\n\nASIC servers to outpace GPU growth\nASIC servers are expected to be the fastest-growing segment of the AI server market in 2026, outpacing systems based on GPUs led by Nvidia.\n\nDIGITIMES Research estimates ASIC server shipments will grow 64.2% year-over-year in 2026, compared with 43.8% growth for GPU servers.\n\nIndustry sources said rising ASIC shipments are also likely to support demand for general-purpose servers. Google's continued investment in ASIC infrastructure is expected to benefit its primary general server assembler, Quanta Computer, in 2026.\n\nDual-chip TPU highlights shift to inference\nGoogle's eighth-generation TPU introduces a dual-chip architecture for the first time, separating training and inference workloads.\n\nThe lineup includes TPU 8t for large-scale training and TPU 8i for low-latency inference. Both chips are expected to launch later in 2026, with shipments projected before year-end, according to supply chain estimates.\n\nThe move reflects a broader shift in AI development. Over the past three years, the industry has focused on large-scale training, but leading chipmakers, including Google and Nvidia, are placing greater emphasis on inference, indicating a widening range of AI applications.\n\nGoogle remains key ASIC player\nGoogle continues to be a major driver of ASIC-based AI infrastructure. DIGITIMES Research estimates the company will account for 46% of the ASIC accelerator market in 2026, with shipments reaching about 3.33 million units.\n\nThe TPU v7, code-named Ironwood, is expected to remain the main product, accounting for 89% of Google's total TPU shipments.\n\nIndustry sources said Celestica remains the primary assembly partner for TPU servers, but Taiwanese firms are expanding their role. This is most evident in CPU motherboard production, where Inventec's share is increasing, and in L10 to L11 assembly, where Foxconn is involved.\n\n$CLS\n\n---\n\nhttps://t.co/hHrcN2SlAp", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.08998302207130726, "ret_signed_m1d": -0.08998302207130726, "alpha_spy_m1d": -0.08229331395073514, "alpha_c_m1d": -0.06264558074412219, "bench_spy_m1d": -0.007689708120572125, "bench_c_m1d": -0.027337441327185075, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.0373514431239389, "ret_signed_p1d": -0.0373514431239389, "alpha_spy_p1d": -0.03907424961264094, "alpha_c_p1d": -0.03953597751945115, "bench_spy_p1d": 0.0017228064887020444, "bench_c_p1d": 0.00218453439551225, "ret_p1w": -0.0373514431239389, "ret_signed_p1w": -0.0373514431239389, "alpha_spy_p1w": -0.04675002276608031, "alpha_c_p1w": -0.04764974473368311, "bench_spy_p1w": 0.009398579642141414, "bench_c_p1w": 0.01029830160974421, "ret_p1m": -0.1256366723259762, "ret_signed_p1m": -0.1256366723259762, "alpha_spy_p1m": -0.17003817959357448, "alpha_c_p1m": -0.25152606201060823, "bench_spy_p1m": 0.044401507267598284, "bench_c_p1m": 0.12588938968463204, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.5178, -0.5025, -0.4924, -0.5042, -0.4941, -0.4448, -0.4312, -0.4465, -0.4584, -0.4567, -0.4499, -0.455, -0.455, -0.455, -0.455, -0.455, -0.455, -0.455, -0.455, -0.4839, -0.4737, -0.4312, -0.4007, -0.4007, -0.4007, -0.3905, -0.4058, -0.382, -0.3735, -0.4346, -0.4244, -0.3684, -0.3633, -0.3667, -0.3786, -0.3531, -0.3718, -0.3158, -0.3039, -0.3226, -0.3396, -0.2869, -0.2275, -0.2547, -0.275, -0.3243, -0.2835, -0.2971, -0.2971, -0.2971, -0.3107, -0.2428, -0.2411, -0.2309, -0.253, -0.2156, -0.2105, -0.1715, -0.1851, -0.1545, -0.129, -0.0849, -0.09, 0.0, -0.0374, -0.056, -0.0374, -0.0374, -0.0374, -0.0204, -0.0815, -0.1732, -0.18, -0.1698, -0.1324, -0.1443, -0.1222, -0.1324, -0.1664, -0.1817, -0.1885, -0.2054, -0.1528, -0.1358, -0.1256, -0.0747, -0.0832, -0.1239, -0.0951, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2049402423584710776", "ticker_idx": 0, "ts": "2026-04-29T08:16:28+00:00", "symbol": "005930.KS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "expectations are growing that the Foundry division's swing to profitability will come sooner than anticipated", "tweet_type": "news_reaction", "timeline": "quarters", "summary": "Samsung Foundry 4nm yield >80% triggers early-than-expected profitability swing; bullish for Samsung Electronics.", "resolved_tickers": ["005930.KS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "[Exclusive] Samsung Foundry’s 4nm Yield Crosses the ‘80% Threshold’… Head-to-Head Showdown with TSMC\n\nSamsung Electronics’ Foundry division has pushed its 4-nanometer (nm, one-billionth of a meter) process yield past 80%, entering the technically refined “mature process” stage. With its technical completeness now on par with Taiwan’s TSMC — the world’s largest foundry — Samsung’s 4nm is drawing a flood of orders from global Big Tech. As a result, expectations are growing that the Foundry division’s swing to profitability will come sooner than anticipated.\n\nAccording to industry sources on the 29th, Samsung Foundry’s 4nm process yield has recently surpassed 80%. Since mass production began in 2021, accumulated manufacturing know-how has lifted the mature-process yield to a world-class level.\n\nWith the 4nm node reaching maturity, global AI chip companies are also flocking to Samsung Foundry. Groq — acquired by Nvidia late last year — has entrusted production of its Language Processing Units (LPUs) entirely to Samsung Foundry’s 4nm. Groq has placed deep trust in Samsung, outsourcing everything from its first-generation LPU (LP1) three years ago through this year’s LP3. Other clients on Samsung’s 4nm process include U.S. Big Tech IBM, automotive chip specialist Ambarella, Chinese conglomerate Baidu, and Chinese mining rig maker Faraday. Domestic Korean AI chip fabless players Rebellions and FuriosaAI are likewise leveraging Samsung 4nm to accelerate product development.\n\nSynergies between the Foundry and Memory divisions are also expanding. The Foundry division is exclusively manufacturing the “base die” for the 6th-generation High Bandwidth Memory (HBM4), which Samsung Electronics has mass-produced as a world first. The Foundry division’s performance is in turn reinforcing the competitiveness of Samsung’s Memory division.\n\nAccordingly, forecasts are emerging that Samsung Electronics’ non-memory businesses (Foundry and System LSI) could see an earnings turnaround as early as the second half. An industry source said, “The price of 12-inch wafers for HBM4 base dies has surged more than 50% from prior levels, exceeding $20,000 (roughly 30 million won) per wafer,” adding that “HBM4-related revenue could exceed 30 trillion won.“\n\n---\n\nhttps://t.co/3tHQMHZnx2", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.017699115044247815, "ret_signed_m1d": -0.017699115044247815, "alpha_spy_m1d": -0.017853680024883856, "alpha_c_m1d": -0.009780099884908022, "bench_spy_m1d": 0.00015456498063604052, "bench_c_m1d": -0.007919015159339793, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.02433628318584069, "ret_signed_p1d": -0.02433628318584069, "alpha_spy_p1d": -0.03428591061304487, "alpha_c_p1d": -0.026787413775699154, "bench_spy_p1d": 0.009949627427204177, "bench_c_p1d": 0.0024511305898584634, "ret_p1w": 0.17699115044247793, "ret_signed_p1w": 0.17699115044247793, "alpha_spy_p1w": 0.14572270632428008, "alpha_c_p1w": 0.10835939802555994, "bench_spy_p1w": 0.03126844411819785, "bench_c_p1w": 0.06863175241691799, "ret_p1m": 0.3252212389380531, "ret_signed_p1m": 0.3252212389380531, "alpha_spy_p1m": 0.2647642875581746, "alpha_c_p1m": 0.15087641217196412, "bench_spy_p1m": 0.0604569513798785, "bench_c_p1m": 0.17434482676608898, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2913, -0.3359, -0.2604, -0.2533, -0.2966, -0.2997, -0.2652, -0.2679, -0.2591, -0.2114, -0.1999, -0.1999, -0.1999, -0.1999, -0.161, -0.1606, -0.1478, -0.1169, -0.1014, -0.0374, -0.044, -0.044, -0.1385, -0.2396, -0.154, -0.169, -0.2339, -0.1703, -0.161, -0.1703, -0.1897, -0.1668, -0.1438, -0.0793, -0.1147, -0.1195, -0.1774, -0.1624, -0.1654, -0.2047, -0.2047, -0.2199, -0.2602, -0.1608, -0.2106, -0.1761, -0.1456, -0.1305, -0.0686, -0.0973, -0.0885, -0.1106, -0.0863, -0.0664, -0.0376, -0.0442, -0.0509, -0.031, -0.0376, -0.0066, -0.0288, -0.0066, -0.0177, 0.0, -0.0243, -0.0243, 0.0288, 0.0288, 0.177, 0.2013, 0.1881, 0.2633, 0.2345, 0.2566, 0.3097, 0.1969, 0.2434, 0.219, 0.2212, 0.3252, 0.2942, 0.2942, 0.323, 0.3584, 0.3252, 0.4027, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048992506851868780", "ticker_idx": 0, "ts": "2026-04-28T05:07:36+00:00", "symbol": "AVGO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This would be a major negative for Broadcom and MTK. Broadcom/MediaTek's per-chip margin would fall, if this is true. That makes me more worried about Broadcom than MediaTek.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Broadcom (more) and MediaTek (less) on rumored Google COT shift to Annapurna-like model, compressing per-chip margins to design service fees.", "resolved_tickers": ["AVGO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Rumor: Google to shift to a COT model starting with TPU v9, downgrading the role of MediaTek/Broadcom.\n\nThis would be a major negative for Broadcom and MTK.\n\nAccording to this rumor, starting with TPU v9, Google plans to move to a model similar to AWS’s Annapurna, where Google directly places wafer orders with TSMC.\n\nIn other words, Google would be doing COT directly.\n\nWhat does that mean? Google would become TSMC’s direct customer, while Broadcom and MediaTek would be downgraded to IP support / design service roles.\n\nOwnership of the chip and control over supply would shift toward Google.\n\nIn other words, Broadcom/MediaTek’s per-chip margin would fall, if this is true.\n\nInstead of recognizing the full chip ASP as revenue, the model would shift toward design service fees / margin. That makes me more worried about Broadcom than MediaTek.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.045944592080877866, "ret_signed_m1d": -0.045944592080877866, "alpha_spy_m1d": -0.04105485011905152, "alpha_c_m1d": -0.015305927207703629, "bench_spy_m1d": 0.004889741961826344, "bench_c_m1d": 0.030638664873174237, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.014056038375075408, "ret_signed_p1d": -0.014056038375075408, "alpha_spy_p1d": -0.014210579469070228, "alpha_c_p1d": 0.002983507364606286, "bench_spy_p1d": -0.00015454109399481997, "bench_c_p1d": 0.017039545739681694, "ret_p1w": 0.06885426232112057, "ret_signed_p1w": -0.06885426232112057, "alpha_spy_p1w": -0.05188055600460073, "alpha_c_p1w": -0.004767595653211387, "bench_spy_p1w": 0.016973706316519843, "bench_c_p1w": 0.06408666666790919, "ret_p1m": 0.0550984156244998, "ret_signed_p1m": -0.0550984156244998, "alpha_spy_p1m": -0.0006224227086910616, "alpha_c_p1m": 0.15721405838685043, "bench_spy_p1m": 0.054475992915808735, "bench_c_p1m": 0.21231247401135023, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1746, -0.1731, -0.1736, -0.2005, -0.2312, -0.225, -0.1691, -0.1416, -0.1503, -0.1445, -0.1735, -0.1884, -0.1884, -0.17, -0.1676, -0.1664, -0.1698, -0.1755, -0.1876, -0.1706, -0.1971, -0.2025, -0.2043, -0.2167, -0.2075, -0.1695, -0.1752, -0.1371, -0.145, -0.1475, -0.1615, -0.1959, -0.1891, -0.1981, -0.2115, -0.2017, -0.225, -0.1934, -0.2039, -0.2026, -0.2261, -0.248, -0.2662, -0.2259, -0.2159, -0.2133, -0.2133, -0.2136, -0.1647, -0.1231, -0.1123, -0.0707, -0.0502, -0.0476, -0.0078, -0.0034, 0.0168, -0.0005, 0.0059, 0.0571, 0.0503, 0.0573, 0.0459, 0.0, 0.0141, 0.044, 0.0536, 0.0417, 0.0689, 0.0641, 0.0318, 0.0755, 0.0715, 0.0487, 0.0424, 0.0999, 0.0634, 0.0522, 0.0281, 0.0448, 0.0369, 0.0358, 0.0358, 0.0555, 0.0551, 0.0669, 0.1174, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048992506851868780", "ticker_idx": 1, "ts": "2026-04-28T05:07:36+00:00", "symbol": "MediaTek", "kind": "equity", "geography": null, "sub_sector": null, "direction": "short", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "This would be a major negative for Broadcom and MTK. Broadcom/MediaTek's per-chip margin would fall, if this is true.", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Bearish Broadcom (more) and MediaTek (less) on rumored Google COT shift to Annapurna-like model, compressing per-chip margins to design service fees.", "resolved_tickers": ["2454.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "MediaTek 2454.TW Taiwan", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "Rumor: Google to shift to a COT model starting with TPU v9, downgrading the role of MediaTek/Broadcom.\n\nThis would be a major negative for Broadcom and MTK.\n\nAccording to this rumor, starting with TPU v9, Google plans to move to a model similar to AWS’s Annapurna, where Google directly places wafer orders with TSMC.\n\nIn other words, Google would be doing COT directly.\n\nWhat does that mean? Google would become TSMC’s direct customer, while Broadcom and MediaTek would be downgraded to IP support / design service roles.\n\nOwnership of the chip and control over supply would shift toward Google.\n\nIn other words, Broadcom/MediaTek’s per-chip margin would fall, if this is true.\n\nInstead of recognizing the full chip ASP as revenue, the model would shift toward design service fees / margin. That makes me more worried about Broadcom than MediaTek.\n\n$AVGO", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.06883365200764824, "ret_signed_m1d": 0.06883365200764824, "alpha_spy_m1d": 0.07372339396947458, "alpha_c_m1d": 0.09947231688082248, "bench_spy_m1d": 0.004889741961826344, "bench_c_m1d": 0.030638664873174237, "ret_t0": 0.0, "ret_signed_t0": -0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015296367112810683, "ret_signed_p1d": 0.015296367112810683, "alpha_spy_p1d": 0.015141826018815863, "alpha_c_p1d": 0.03233591285249238, "bench_spy_p1d": -0.00015454109399481997, "bench_c_p1d": 0.017039545739681694, "ret_p1w": 0.2065009560229445, "ret_signed_p1w": -0.2065009560229445, "alpha_spy_p1w": -0.18952724970642465, "alpha_c_p1w": -0.1424142893550353, "bench_spy_p1w": 0.016973706316519843, "bench_c_p1w": 0.06408666666790919, "ret_p1m": 0.7743785850860421, "ret_signed_p1m": -0.7743785850860421, "alpha_spy_p1m": -0.7199025921702333, "alpha_c_p1m": -0.5620661110746918, "bench_spy_p1m": 0.054475992915808735, "bench_c_p1m": 0.21231247401135023, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3193, -0.327, -0.348, -0.3136, -0.3117, -0.3231, -0.3461, -0.3002, -0.2945, -0.2906, -0.2906, -0.2906, -0.2906, -0.2906, -0.2906, -0.2906, -0.2906, -0.3136, -0.3021, -0.2849, -0.2562, -0.2562, -0.2734, -0.3059, -0.3423, -0.3212, -0.325, -0.3633, -0.348, -0.325, -0.3174, -0.3423, -0.3461, -0.3384, -0.3384, -0.3576, -0.3499, -0.3786, -0.3805, -0.3805, -0.392, -0.3939, -0.4226, -0.4302, -0.4398, -0.4398, -0.4398, -0.4398, -0.4379, -0.3958, -0.3977, -0.3977, -0.3805, -0.3423, -0.3155, -0.2753, -0.2639, -0.2734, -0.2008, -0.1224, -0.153, -0.0688, -0.0688, 0.0, -0.0153, -0.0019, -0.0019, 0.0975, 0.2065, 0.3117, 0.3078, 0.3881, 0.4837, 0.4149, 0.3365, 0.3021, 0.2467, 0.3002, 0.2065, 0.2352, 0.3576, 0.4761, 0.6233, 0.631, 0.7744, 0.6864, 0.6482, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2045664195921010835", "ticker_idx": 0, "ts": "2026-04-19T00:42:05+00:00", "symbol": "ABF载板", "kind": "sector", "geography": "global", "sub_sector": "ABF substrate", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "现在 ABF 和 CCL 就跟去年9月的记忆体行情差不多", "tweet_type": "prediction", "timeline": "months", "summary": "Sell-side analyst analogy: ABF substrates and CCL materials now mirror memory's inflection point in Sep 2024, implying a similar upward run ahead.", "resolved_tickers": ["4062.T", "3034.TW", "2383.TW", "3037.TW"], "resolution_method": "basket", "resolution_reason": "Chinese term for ABF substrate; same basket", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "zh", "tweet_text": "@yiqifacai 卖方分析师提到，现在 ABF 和 CCL 就跟去年9月的记忆体行情差不多。你明白我现在在说什么吧？", "reply_text": null, "parent_text": "在X泡着还赚不到钱是要反思的，满世界专家不是每天都在送钱吗？ https://t.co/hO9HTCgq46", "quoted_text": null, "parent_author": "yiqifacai", "ret_m1d": -0.025029642561817528, "ret_signed_m1d": -0.025029642561817528, "alpha_spy_m1d": -0.02703331679707824, "alpha_c_m1d": -0.023671002173650396, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": -0.0013586403881671316, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.044312137008078456, "ret_signed_p1d": 0.044312137008078456, "alpha_spy_p1d": 0.05085908616301471, "alpha_c_p1d": 0.0434710080946078, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0008411289134706568, "ret_p1w": 0.19822165278940973, "ret_signed_p1w": 0.19822165278940973, "alpha_spy_p1w": 0.18912072086472842, "alpha_c_p1w": 0.15933701342244422, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.03888463936696551, "ret_p1m": 0.2768258299258406, "ret_signed_p1m": 0.2768258299258406, "alpha_spy_p1m": 0.2415368430206002, "alpha_c_p1m": 0.1559665634297427, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.1208592664960979, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3362, -0.3179, -0.3055, -0.3192, -0.3048, -0.2894, -0.3099, -0.3068, -0.3261, -0.2939, -0.3154, -0.3357, -0.3451, -0.3096, -0.2929, -0.2891, -0.2739, -0.2703, -0.2667, -0.2624, -0.2522, -0.2443, -0.2568, -0.2429, -0.196, -0.1891, -0.1849, -0.1787, -0.1911, -0.2254, -0.2733, -0.2438, -0.2562, -0.2984, -0.2537, -0.2318, -0.2197, -0.1986, -0.1893, -0.1734, -0.1594, -0.1576, -0.1658, -0.2139, -0.2289, -0.1803, -0.167, -0.169, -0.1986, -0.2462, -0.1876, -0.1947, -0.1833, -0.1743, -0.1454, -0.074, -0.0637, -0.0471, -0.0705, -0.0468, -0.0358, -0.015, -0.025, 0.0, 0.0443, 0.0809, 0.07, 0.1611, 0.1982, 0.1647, 0.1584, 0.2191, 0.2163, 0.2442, 0.2423, 0.253, 0.3529, 0.3152, 0.3403, 0.3663, 0.3875, 0.3841, 0.2923, 0.2907, 0.2768, 0.2818, 0.3968, 0.4778, 0.5636, 0.5871, 0.5682, 0.5099, 0.6094, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2045664195921010835", "ticker_idx": 1, "ts": "2026-04-19T00:42:05+00:00", "symbol": "CCL材料", "kind": "sector", "geography": "global", "sub_sector": "copper clad laminate", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "现在 ABF 和 CCL 就跟去年9月的记忆体行情差不多", "tweet_type": "prediction", "timeline": "months", "summary": "Sell-side analyst analogy: ABF substrates and CCL materials now mirror memory's inflection point in Sep 2024, implying a similar upward run ahead.", "resolved_tickers": ["2383.TW", "3034.TW", "4062.T"], "resolution_method": "basket", "resolution_reason": "CCL material global basket: EMC, Unimicron, Ibiden", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "zh", "tweet_text": "@yiqifacai 卖方分析师提到，现在 ABF 和 CCL 就跟去年9月的记忆体行情差不多。你明白我现在在说什么吧？", "reply_text": null, "parent_text": "在X泡着还赚不到钱是要反思的，满世界专家不是每天都在送钱吗？ https://t.co/hO9HTCgq46", "quoted_text": null, "parent_author": "yiqifacai", "ret_m1d": -0.016165383789404686, "ret_signed_m1d": -0.016165383789404686, "alpha_spy_m1d": -0.018169058024665397, "alpha_c_m1d": -0.014806743401237554, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": -0.0013586403881671316, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.043350302638106575, "ret_signed_p1d": 0.043350302638106575, "alpha_spy_p1d": 0.04989725179304283, "alpha_c_p1d": 0.04250917372463592, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0008411289134706568, "ret_p1w": 0.18514116143799372, "ret_signed_p1w": 0.18514116143799372, "alpha_spy_p1w": 0.1760402295133124, "alpha_c_p1w": 0.1462565220710282, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.03888463936696551, "ret_p1m": 0.30027121472904605, "ret_signed_p1m": 0.30027121472904605, "alpha_spy_p1m": 0.26498222782380565, "alpha_c_p1m": 0.17941194823294815, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.1208592664960979, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2727, -0.2513, -0.2406, -0.256, -0.2472, -0.2331, -0.2493, -0.2618, -0.2809, -0.2478, -0.278, -0.294, -0.294, -0.2503, -0.2398, -0.2336, -0.2132, -0.2085, -0.2036, -0.1979, -0.1844, -0.1738, -0.1905, -0.1899, -0.1473, -0.1442, -0.1499, -0.1417, -0.158, -0.1948, -0.2361, -0.213, -0.2206, -0.2558, -0.2133, -0.2049, -0.1956, -0.1812, -0.189, -0.1649, -0.1643, -0.1625, -0.1602, -0.1972, -0.1981, -0.1559, -0.1362, -0.1402, -0.1741, -0.2134, -0.157, -0.1814, -0.1662, -0.1543, -0.1379, -0.0701, -0.0588, -0.0432, -0.0685, -0.0226, -0.0183, -0.0038, -0.0162, 0.0, 0.0434, 0.0892, 0.0687, 0.1597, 0.1851, 0.1474, 0.1498, 0.1913, 0.1876, 0.211, 0.2125, 0.2488, 0.3634, 0.3514, 0.3637, 0.3916, 0.4115, 0.4124, 0.3194, 0.3197, 0.3003, 0.3045, 0.4175, 0.4934, 0.598, 0.5826, 0.5599, 0.5092, 0.6271, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048683483929919806", "ticker_idx": 0, "ts": "2026-04-27T08:39:39+00:00", "symbol": "2454.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Target price raised to NT$2,850; reiterating the views from our March 25 TPU upgrade report… we raise our target price from NT$2,750 to NT$2,850 and reiterate our Buy rating", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF reiterates Buy on MediaTek, raises PT to NT$2,850 on Meta AR, OpenAI edge AI win potential, and TPU 8t ramp; AP Memory benefits as SRAM supplier for Meta AR glasses.", "resolved_tickers": ["2454.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Telecom\n☄️ MediaTek (2454 TT, Buy): Two More Things!\n\n☀️ What’s new? Meta!\nSince we published our MediaTek report on March 25, MediaTek’s share price has risen 50%, driven by multiple brokerages upgrading the stock and highlighting its Google TPU potential, while our earnings forecasts remain the highest on the Street. We now see increasing opportunities for MediaTek to win more customers.\n\nRegarding Meta, we are highly positive on its next-generation AR glasses. The product will use two SRAM chips supplied by AP Memory (6531 TT) and one SoC supplied by MediaTek, replacing Qualcomm, with mass production targeted for the end of 2027. Following the success of Meta’s Ray-Ban Display glasses, we expect this product to reach annual shipments of 6–7 million units.\n\nOn the ASIC side, we expect MediaTek management to continue emphasizing its second ASIC customer during the upcoming earnings call, and we believe the company is targeting the Astrid and Apollo projects.\n\n☀️ What’s new? OpenAI!\nIn addition to the new Meta catalyst, we now believe MediaTek is highly likely to win the bid for OpenAI’s edge AI project, as the product SKU is focused on the mid-range market. We expect the project to be an N3-based SoC — with the modem still undecided — and to fully integrate OpenAI’s AI model.\n\nThe bidding process is ongoing, with tape-out targeted for 1H27. Regardless of shipment volume, we believe the symbolic significance of this cooperation would be very positive, as it could open the door for MediaTek to pursue ASIC collaboration opportunities with OpenAI.\nThat said, MediaTek’s capabilities have, at least for now, been validated through the TPU 8t tape-out, and the company can offer multiple backend solutions — including CoWoS, EMIB, and InFO — to compete for ASIC projects in what is currently the industry bottleneck.\n\n☀️ Target price raised to NT$2,850; reiterating the views from our March 25 TPU upgrade report\nWe reiterate the core views from our March 25 report:\n\n1. The early redesign issues for v8x, i.e. 8t, have been quickly resolved, and risk production is now targeted for May.\n2. MediaTek’s v8x CoWoS capacity allocation for 2027 has been raised from 70,000 wafers to 160,000 wafers, corresponding to 3.2 million packaged chips.\n3. Humufish has an ASP of US$15,000, with shipments reaching 3.5 million units by the end of 2028.\n4. Intel’s EMIB can support 9x to 12x reticle-size packages, but constraints in silicon capacitors and substrates are worth monitoring.\n\nBased on a minor earnings adjustment from raising the average selling price of TPU 8t to US$4,400, we raise our 2026/2027 EPS estimates by 1%/4%, respectively. Applying 25x 2027 P/E, we raise our target price from NT$2,750 to NT$2,850 and reiterate our Buy rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0, "ret_signed_m1d": 0.0, "alpha_spy_m1d": 0.0017198435311071014, "alpha_c_m1d": -0.00035553405820976103, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.07392197125256672, "ret_signed_p1d": 0.07392197125256672, "alpha_spy_p1d": 0.07878791998071288, "alpha_c_p1d": 0.10364981469316226, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.17864476386036965, "ret_signed_p1w": 0.17864476386036965, "alpha_spy_p1w": 0.17467364241604932, "alpha_c_p1w": 0.17759787339090072, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.7515400410677617, "ret_signed_p1m": 0.7515400410677617, "alpha_spy_p1m": 0.7020132924741169, "alpha_c_p1m": 0.5621511834979123, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.269, -0.269, -0.2772, -0.2998, -0.2628, -0.2608, -0.2731, -0.2977, -0.2485, -0.2423, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2382, -0.2628, -0.2505, -0.232, -0.2012, -0.2012, -0.2197, -0.2546, -0.2936, -0.271, -0.2752, -0.3162, -0.2998, -0.2752, -0.2669, -0.2936, -0.2977, -0.2895, -0.2895, -0.3101, -0.3018, -0.3326, -0.3347, -0.3347, -0.347, -0.3491, -0.3799, -0.3881, -0.3984, -0.3984, -0.3984, -0.3984, -0.3963, -0.3511, -0.3532, -0.3532, -0.3347, -0.2936, -0.2649, -0.2218, -0.2094, -0.2197, -0.1417, -0.0575, -0.0903, 0.0, 0.0, 0.0739, 0.0575, 0.0719, 0.0719, 0.1786, 0.2957, 0.4086, 0.4045, 0.4908, 0.5934, 0.5195, 0.4353, 0.3984, 0.3388, 0.3963, 0.2957, 0.3265, 0.4579, 0.5852, 0.7433, 0.7515, 0.9055, 0.8111, 0.77, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2048683483929919806", "ticker_idx": 1, "ts": "2026-04-27T08:39:39+00:00", "symbol": "AP Memory", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "two SRAM chips supplied by AP Memory (6531 TT)… with mass production targeted for the end of 2027", "tweet_type": "news_reaction", "timeline": "years", "summary": "GF reiterates Buy on MediaTek, raises PT to NT$2,850 on Meta AR, OpenAI edge AI win potential, and TPU 8t ramp; AP Memory benefits as SRAM supplier for Meta AR glasses.", "resolved_tickers": ["6531.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "AP Memory Technology 6531 TT = 6531.TW", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "GF Overseas Electronics & Telecom\n☄️ MediaTek (2454 TT, Buy): Two More Things!\n\n☀️ What’s new? Meta!\nSince we published our MediaTek report on March 25, MediaTek’s share price has risen 50%, driven by multiple brokerages upgrading the stock and highlighting its Google TPU potential, while our earnings forecasts remain the highest on the Street. We now see increasing opportunities for MediaTek to win more customers.\n\nRegarding Meta, we are highly positive on its next-generation AR glasses. The product will use two SRAM chips supplied by AP Memory (6531 TT) and one SoC supplied by MediaTek, replacing Qualcomm, with mass production targeted for the end of 2027. Following the success of Meta’s Ray-Ban Display glasses, we expect this product to reach annual shipments of 6–7 million units.\n\nOn the ASIC side, we expect MediaTek management to continue emphasizing its second ASIC customer during the upcoming earnings call, and we believe the company is targeting the Astrid and Apollo projects.\n\n☀️ What’s new? OpenAI!\nIn addition to the new Meta catalyst, we now believe MediaTek is highly likely to win the bid for OpenAI’s edge AI project, as the product SKU is focused on the mid-range market. We expect the project to be an N3-based SoC — with the modem still undecided — and to fully integrate OpenAI’s AI model.\n\nThe bidding process is ongoing, with tape-out targeted for 1H27. Regardless of shipment volume, we believe the symbolic significance of this cooperation would be very positive, as it could open the door for MediaTek to pursue ASIC collaboration opportunities with OpenAI.\nThat said, MediaTek’s capabilities have, at least for now, been validated through the TPU 8t tape-out, and the company can offer multiple backend solutions — including CoWoS, EMIB, and InFO — to compete for ASIC projects in what is currently the industry bottleneck.\n\n☀️ Target price raised to NT$2,850; reiterating the views from our March 25 TPU upgrade report\nWe reiterate the core views from our March 25 report:\n\n1. The early redesign issues for v8x, i.e. 8t, have been quickly resolved, and risk production is now targeted for May.\n2. MediaTek’s v8x CoWoS capacity allocation for 2027 has been raised from 70,000 wafers to 160,000 wafers, corresponding to 3.2 million packaged chips.\n3. Humufish has an ASP of US$15,000, with shipments reaching 3.5 million units by the end of 2028.\n4. Intel’s EMIB can support 9x to 12x reticle-size packages, but constraints in silicon capacitors and substrates are worth monitoring.\n\nBased on a minor earnings adjustment from raising the average selling price of TPU 8t to US$4,400, we raise our 2026/2027 EPS estimates by 1%/4%, respectively. Applying 25x 2027 P/E, we raise our target price from NT$2,750 to NT$2,850 and reiterate our Buy rating.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.04648074369189903, "ret_signed_m1d": 0.04648074369189903, "alpha_spy_m1d": 0.04820058722300613, "alpha_c_m1d": 0.04612520963368927, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.062416998671978696, "ret_signed_p1d": 0.062416998671978696, "alpha_spy_p1d": 0.06728294740012486, "alpha_c_p1d": 0.09214484211257423, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": 0.22709163346613548, "ret_signed_p1w": 0.22709163346613548, "alpha_spy_p1w": 0.22312051202181515, "alpha_c_p1w": 0.22604474299666655, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.4342629482071714, "ret_signed_p1m": 0.4342629482071714, "alpha_spy_p1m": 0.38473619961352656, "alpha_c_p1m": 0.2448740906373219, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.421, -0.4117, -0.4296, -0.4422, -0.429, -0.4336, -0.4515, -0.4595, -0.4615, -0.4555, -0.4741, -0.4741, -0.4741, -0.4741, -0.4741, -0.4741, -0.4741, -0.4741, -0.4655, -0.4529, -0.4323, -0.4243, -0.4243, -0.4243, -0.4396, -0.4635, -0.415, -0.423, -0.4648, -0.4323, -0.4137, -0.3552, -0.2948, -0.2988, -0.2297, -0.2404, -0.2683, -0.2656, -0.3054, -0.3347, -0.3214, -0.324, -0.3267, -0.3559, -0.413, -0.3891, -0.4157, -0.4157, -0.4157, -0.3738, -0.3652, -0.3619, -0.3147, -0.247, -0.1952, -0.2151, -0.2218, -0.1448, -0.0943, -0.0438, 0.0518, 0.0425, 0.0465, 0.0, 0.0624, 0.089, 0.1155, 0.1155, 0.2271, 0.2417, 0.2815, 0.2284, 0.162, 0.2776, 0.4011, 0.3811, 0.3944, 0.4343, 0.5073, 0.3612, 0.2922, 0.1633, 0.2789, 0.4011, 0.4343, 0.4343, 0.4542, 0.5803, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2044233874415464529", "ticker_idx": 0, "ts": "2026-04-15T01:58:30+00:00", "symbol": "DRAM", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "People who think the DRAM supercycle will end in 1H next year don't know what they're talking about", "tweet_type": "prediction", "timeline": "quarters", "summary": "DRAM supercycle extends well beyond 1H next year driven by NVIDIA SOCAMM/LPDDR5X demand; NAND also in shortage through 1H 2027 though less confident beyond that.", "resolved_tickers": ["MU", "005930.KS", "000660.KS"], "resolution_method": "basket", "resolution_reason": "Global DRAM basket: Micron, Samsung, SK Hynix", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "People who think the DRAM supercycle will end in 1H next year don't know what they're talking about.\n\nThe GB200 uses 17TB of SOCAMM2, but the VR200 uses over 50TB of SOCAMM2. That alone has NVIDIA hoovering up so much LPDDR5X that even smartphone OEMs are panic-buying — and do you know how much SOCAMM goes into the VR300? 220TB.\n\nA lot of people have only been focused on the VR300's 1TB of HBM4E, but you need to look at the SOCAMM too. Once NVIDIA starts consuming 220TB of SOCAMM, the LPDDR shortage could become so severe that it reaches a point where OEMs can't build smartphones even if they want to.\n\n---\n\nPS. As for NAND, I’m a bit skeptical. Samsung recently decided to build a new NAND fab at P5. There are also widespread rumors that even SK hynix may open a new NAND fab in China.\n\n---\n\nPS. Even so, I think NAND itself could remain in shortage through the first half of 2027. After that, I’m not confident enough to make a firm call.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009902313730404807, "ret_signed_m1d": -0.009902313730404807, "alpha_spy_m1d": -0.002073099155390187, "alpha_c_m1d": -0.007694808211640999, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.002207505518763808, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016574283591703615, "ret_signed_p1d": 0.016574283591703615, "alpha_spy_p1d": 0.014116971969099296, "alpha_c_p1d": 0.012600800605017632, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.003973482986685983, "ret_p1w": 0.05862875744070332, "ret_signed_p1w": 0.05862875744070332, "alpha_spy_p1w": 0.042527349474392, "alpha_c_p1w": 0.006023930570359694, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.052604826870343624, "ret_p1m": 0.6126389394150443, "ret_signed_p1m": 0.6126389394150443, "alpha_spy_p1m": 0.5437330613390338, "alpha_c_p1m": 0.33595013840959453, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.2766888010054498, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2789, -0.2759, -0.2869, -0.2634, -0.2484, -0.2437, -0.2605, -0.2146, -0.1798, -0.1762, -0.1777, -0.2001, -0.164, -0.1927, -0.2226, -0.216, -0.1978, -0.2095, -0.1839, -0.156, -0.156, -0.156, -0.1646, -0.1492, -0.1338, -0.1097, -0.1098, -0.0851, -0.0677, -0.0304, -0.0462, -0.046, -0.1396, -0.1867, -0.1319, -0.1617, -0.1968, -0.1341, -0.1145, -0.1349, -0.1325, -0.0941, -0.0725, -0.0242, -0.0622, -0.0813, -0.1372, -0.1228, -0.131, -0.1827, -0.1814, -0.2302, -0.2522, -0.1684, -0.207, -0.1812, -0.159, -0.1449, -0.0672, -0.0769, -0.0659, -0.0656, -0.0099, 0.0, 0.0166, 0.0047, 0.0086, 0.0335, 0.0586, 0.0661, 0.0682, 0.117, 0.1006, 0.1152, 0.1035, 0.1218, 0.2131, 0.2596, 0.377, 0.3867, 0.4645, 0.5838, 0.5393, 0.6156, 0.6126, 0.4905, 0.4818, 0.4578, 0.4829, 0.5992, 0.5803, 0.5803, 0.729, 0.8215, 0.8197, 0.8949, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2044233874415464529", "ticker_idx": 1, "ts": "2026-04-15T01:58:30+00:00", "symbol": "LPDDR5X", "kind": "sector", "geography": "global", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Once NVIDIA starts consuming 220TB of SOCAMM, the LPDDR shortage could become so severe that it reaches a point where OEMs can't build smartphones even if they want to", "tweet_type": "prediction", "timeline": "quarters", "summary": "DRAM supercycle extends well beyond 1H next year driven by NVIDIA SOCAMM/LPDDR5X demand; NAND also in shortage through 1H 2027 though less confident beyond that.", "resolved_tickers": ["000660.KS", "005930.KS", "MU"], "resolution_method": "basket", "resolution_reason": "LPDDR5X made by SK Hynix, Samsung, Micron; no direct ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "People who think the DRAM supercycle will end in 1H next year don't know what they're talking about.\n\nThe GB200 uses 17TB of SOCAMM2, but the VR200 uses over 50TB of SOCAMM2. That alone has NVIDIA hoovering up so much LPDDR5X that even smartphone OEMs are panic-buying — and do you know how much SOCAMM goes into the VR300? 220TB.\n\nA lot of people have only been focused on the VR300's 1TB of HBM4E, but you need to look at the SOCAMM too. Once NVIDIA starts consuming 220TB of SOCAMM, the LPDDR shortage could become so severe that it reaches a point where OEMs can't build smartphones even if they want to.\n\n---\n\nPS. As for NAND, I’m a bit skeptical. Samsung recently decided to build a new NAND fab at P5. There are also widespread rumors that even SK hynix may open a new NAND fab in China.\n\n---\n\nPS. Even so, I think NAND itself could remain in shortage through the first half of 2027. After that, I’m not confident enough to make a firm call.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.009902313730404807, "ret_signed_m1d": -0.009902313730404807, "alpha_spy_m1d": -0.002073099155390187, "alpha_c_m1d": -0.007694808211640999, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.002207505518763808, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.016574283591703615, "ret_signed_p1d": 0.016574283591703615, "alpha_spy_p1d": 0.014116971969099296, "alpha_c_p1d": 0.012600800605017632, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.003973482986685983, "ret_p1w": 0.05862875744070332, "ret_signed_p1w": 0.05862875744070332, "alpha_spy_p1w": 0.042527349474392, "alpha_c_p1w": 0.006023930570359694, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.052604826870343624, "ret_p1m": 0.6126389394150443, "ret_signed_p1m": 0.6126389394150443, "alpha_spy_p1m": 0.5437330613390338, "alpha_c_p1m": 0.33595013840959453, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.2766888010054498, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2789, -0.2759, -0.2869, -0.2634, -0.2484, -0.2437, -0.2605, -0.2146, -0.1798, -0.1762, -0.1777, -0.2001, -0.164, -0.1927, -0.2226, -0.216, -0.1978, -0.2095, -0.1839, -0.156, -0.156, -0.156, -0.1646, -0.1492, -0.1338, -0.1097, -0.1098, -0.0851, -0.0677, -0.0304, -0.0462, -0.046, -0.1396, -0.1867, -0.1319, -0.1617, -0.1968, -0.1341, -0.1145, -0.1349, -0.1325, -0.0941, -0.0725, -0.0242, -0.0622, -0.0813, -0.1372, -0.1228, -0.131, -0.1827, -0.1814, -0.2302, -0.2522, -0.1684, -0.207, -0.1812, -0.159, -0.1449, -0.0672, -0.0769, -0.0659, -0.0656, -0.0099, 0.0, 0.0166, 0.0047, 0.0086, 0.0335, 0.0586, 0.0661, 0.0682, 0.117, 0.1006, 0.1152, 0.1035, 0.1218, 0.2131, 0.2596, 0.377, 0.3867, 0.4645, 0.5838, 0.5393, 0.6156, 0.6126, 0.4905, 0.4818, 0.4578, 0.4829, 0.5992, 0.5803, 0.5803, 0.729, 0.8215, 0.8197, 0.8949, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2044233874415464529", "ticker_idx": 2, "ts": "2026-04-15T01:58:30+00:00", "symbol": "NAND", "kind": "sector", "geography": "global", "sub_sector": "NAND", "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "I think NAND itself could remain in shortage through the first half of 2027", "tweet_type": "prediction", "timeline": "quarters", "summary": "DRAM supercycle extends well beyond 1H next year driven by NVIDIA SOCAMM/LPDDR5X demand; NAND also in shortage through 1H 2027 though less confident beyond that.", "resolved_tickers": ["MU", "000660.KS", "005930.KS", "285A.T", "SNDK"], "resolution_method": "basket", "resolution_reason": "Global NAND sector basket: dominant NAND producers", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "People who think the DRAM supercycle will end in 1H next year don't know what they're talking about.\n\nThe GB200 uses 17TB of SOCAMM2, but the VR200 uses over 50TB of SOCAMM2. That alone has NVIDIA hoovering up so much LPDDR5X that even smartphone OEMs are panic-buying — and do you know how much SOCAMM goes into the VR300? 220TB.\n\nA lot of people have only been focused on the VR300's 1TB of HBM4E, but you need to look at the SOCAMM too. Once NVIDIA starts consuming 220TB of SOCAMM, the LPDDR shortage could become so severe that it reaches a point where OEMs can't build smartphones even if they want to.\n\n---\n\nPS. As for NAND, I’m a bit skeptical. Samsung recently decided to build a new NAND fab at P5. There are also widespread rumors that even SK hynix may open a new NAND fab in China.\n\n---\n\nPS. Even so, I think NAND itself could remain in shortage through the first half of 2027. After that, I’m not confident enough to make a firm call.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.021870170223459585, "ret_signed_m1d": 0.021870170223459585, "alpha_spy_m1d": 0.029699384798474206, "alpha_c_m1d": 0.024077675742223394, "bench_spy_m1d": -0.00782921457501462, "bench_c_m1d": -0.002207505518763808, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.02517806210175877, "ret_signed_p1d": 0.02517806210175877, "alpha_spy_p1d": 0.02272075047915445, "alpha_c_p1d": 0.021204579115072786, "bench_spy_p1d": 0.002457311622604319, "bench_c_p1d": 0.003973482986685983, "ret_p1w": 0.06951714926995929, "ret_signed_p1w": 0.06951714926995929, "alpha_spy_p1w": 0.05341574130364797, "alpha_c_p1w": 0.016912322399615667, "bench_spy_p1w": 0.01610140796631132, "bench_c_p1w": 0.052604826870343624, "ret_p1m": 0.5767511266083212, "ret_signed_p1m": 0.5767511266083212, "alpha_spy_p1m": 0.5078452485323106, "alpha_c_p1m": 0.3000623256028714, "bench_spy_p1m": 0.06890587807601056, "bench_c_p1m": 0.2766888010054498, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3836, -0.3789, -0.3767, -0.3438, -0.3256, -0.333, -0.343, -0.3069, -0.2726, -0.266, -0.2456, -0.2575, -0.2141, -0.2556, -0.2838, -0.2776, -0.2712, -0.2879, -0.2596, -0.2216, -0.2121, -0.2154, -0.2283, -0.2227, -0.2099, -0.1931, -0.1895, -0.1704, -0.1663, -0.1409, -0.1543, -0.156, -0.2323, -0.2589, -0.227, -0.2555, -0.2748, -0.2209, -0.1897, -0.211, -0.2011, -0.1591, -0.1486, -0.1012, -0.1262, -0.1516, -0.1923, -0.1859, -0.1881, -0.2437, -0.2456, -0.2862, -0.2911, -0.2112, -0.2367, -0.2165, -0.1922, -0.1839, -0.0948, -0.0842, -0.0625, -0.0328, 0.0219, 0.0, 0.0252, -0.0022, -0.0018, 0.0247, 0.0695, 0.0671, 0.0764, 0.1291, 0.1093, 0.132, 0.1398, 0.164, 0.2342, 0.2959, 0.3671, 0.4004, 0.5037, 0.5808, 0.5337, 0.6056, 0.5768, 0.4843, 0.5055, 0.492, 0.5186, 0.6469, 0.6341, 0.6837, 0.7794, 0.8231, 0.8381, 0.9235, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2041676053027172702", "ticker_idx": 0, "ts": "2026-04-08T00:34:38+00:00", "symbol": "Nan Ya PCB", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Continued earnings improvement driven by rising utilization and increasing mix of advanced products", "tweet_type": "analysis", "timeline": "quarters", "summary": "Bullish Nan Ya PCB: near-full ABF utilization, improving pricing, AI/HPC-driven ASP uplift, and sustained margin improvement over next 3–5 years.", "resolved_tickers": ["8046.TW"], "resolution_method": "exchange_ticker", "resolution_reason": "Nan Ya PCB (NYPCB) listed on TWSE as 8046.TW", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ Strong AI Momentum — Nan Ya PCB's ABF Capacity Near Full Utilization\n\n- Nan Ya PCB's overall substrate utilization exceeds 90%, with ABF substrates effectively running at full capacity; Kunshan fab utilization expected to rise further this year\n\n- Plans to drive H2 growth through product mix improvement amid tight capacity; future expansion to be flexibly adjusted based on customer demand\n\n- Actively negotiating cost pass-through with customers in response to rising raw material prices (copper foil, precious metals); pricing environment for 2026 expected to improve vs. 2025\n\n- High yields on advanced ABF substrates continue to offset cost increases, sustaining margin structure improvement\n\n- Industry competitive dynamics shifting from long-term contract-driven to co-development with customers on high-complexity technologies, strengthening lock-in effects and raising barriers to entry\n\n- Nan Ya PCB maintaining and growing its market position through expanding projects including TPUs\n\n- AI/HPC demand driving substrate evolution toward higher layer counts (26–30+ layers) and larger form factors (140×140mm+), with concurrent increases in ASP and technical barriers\n\n- Glass substrate technology remains in early development stages with near-term displacement of existing materials unlikely; ABF substrate-centric structure expected to persist over the next 3–5 years\n\n- Continued earnings improvement driven by rising utilization and increasing mix of advanced products\n\n---\n\nhttps://t.co/31yDTdZDGZ", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.020408163265306145, "ret_signed_m1d": -0.020408163265306145, "alpha_spy_m1d": 0.0044288048563954074, "alpha_c_m1d": 0.009657541387262158, "bench_spy_m1d": -0.024836968121701553, "bench_c_m1d": -0.030065704652568304, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.050235478806907485, "ret_signed_p1d": 0.050235478806907485, "alpha_spy_p1d": 0.044466387635819116, "alpha_c_p1d": 0.047553533177259855, "bench_spy_p1d": 0.00576909117108837, "bench_c_p1d": 0.0026819456296476307, "ret_p1w": 0.02982731554160134, "ret_signed_p1w": 0.02982731554160134, "alpha_spy_p1w": -0.005571557744428635, "alpha_c_p1w": -0.03093914971350431, "bench_spy_p1w": 0.035398873286029975, "bench_c_p1w": 0.06076646525510565, "ret_p1m": 0.4866562009419153, "ret_signed_p1m": 0.4866562009419153, "alpha_spy_p1m": 0.4044532653027151, "alpha_c_p1m": 0.2890416937957321, "bench_spy_p1m": 0.08220293563920023, "bench_c_p1m": 0.19761450714618323, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6224, -0.6005, -0.5793, -0.5895, -0.5863, -0.5455, -0.5008, -0.4639, -0.4333, -0.4184, -0.4003, -0.4192, -0.3799, -0.3783, -0.4129, -0.3901, -0.4223, -0.3807, -0.3564, -0.376, -0.405, -0.3462, -0.3571, -0.3469, -0.3469, -0.3469, -0.3469, -0.3469, -0.3469, -0.3469, -0.3469, -0.2818, -0.2104, -0.1319, -0.1287, -0.1287, -0.1507, -0.2151, -0.2936, -0.2763, -0.263, -0.3359, -0.3281, -0.2614, -0.2645, -0.1915, -0.1115, -0.146, -0.1099, -0.1193, -0.1507, -0.1617, -0.2292, -0.1523, -0.1068, -0.1193, -0.0895, -0.1805, -0.0989, -0.1083, -0.1083, -0.1083, -0.0204, 0.0, 0.0502, 0.0471, 0.0283, 0.0141, 0.0298, 0.1319, 0.1366, 0.2496, 0.2386, 0.2606, 0.248, 0.3721, 0.4333, 0.4631, 0.438, 0.5777, 0.5777, 0.5479, 0.5651, 0.4757, 0.4867, 0.3407, 0.4254, 0.4129, 0.3815, 0.3297, 0.2339, 0.2418, 0.2527, 0.3014, 0.3689, 0.4662, 0.3878, 0.4474, 0.4207, 0.3344, 0.3312, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2048378916033954046", "ticker_idx": 0, "ts": "2026-04-26T12:29:25+00:00", "symbol": "BESI.AS", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC 将增加 SoIC 的订单，我很看好 BESI", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish BESI on expected TSMC SoIC order increase; endorses hybrid bonding monopoly thesis.", "resolved_tickers": ["BESI.AS"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "reply_other", "lang": "zh", "tweet_text": "@LinQingV TSMC 将增加 SoIC 的订单，我很看好 BESI。", "reply_text": null, "parent_text": "这两天聊了不少先进封装，耗材和设备都过了一遍。正好花旗这周更新了BESI（BE Semiconductor Industries），也来聊聊。\n\nBESI在Hybrid Bonding设备上的垄断地位没什么好争论的，目前全球唯一有量产交付能力的玩家。花旗将BESI https://t.co/OMb8AcXvMG", "quoted_text": null, "parent_author": "LinQingV", "ret_m1d": 0.010564496440272242, "ret_signed_m1d": 0.010564496440272242, "alpha_spy_m1d": 0.012284339971379343, "alpha_c_m1d": 0.01020896238206248, "bench_spy_m1d": -0.0017198435311071014, "bench_c_m1d": 0.00035553405820976103, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.05120966511387981, "ret_signed_p1d": -0.05120966511387981, "alpha_spy_p1d": -0.046343716385733646, "alpha_c_p1d": -0.021481821673284274, "bench_spy_p1d": -0.004865948728146163, "bench_c_p1d": -0.029727843440595536, "ret_p1w": -0.01491934253323468, "ret_signed_p1w": -0.01491934253323468, "alpha_spy_p1w": -0.018890463977555005, "alpha_c_p1w": -0.01596623300270361, "bench_spy_p1w": 0.003971121444320325, "bench_c_p1w": 0.0010468904694689307, "ret_p1m": 0.14516129032258074, "ret_signed_p1m": 0.14516129032258074, "alpha_spy_p1m": 0.09563454172893593, "alpha_c_p1m": -0.04422756724726873, "bench_spy_p1m": 0.04952674859364481, "bench_c_p1m": 0.18938885756984947, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.321, -0.3487, -0.3408, -0.3404, -0.3583, -0.3687, -0.3559, -0.3368, -0.3292, -0.3226, -0.3152, -0.324, -0.294, -0.2848, -0.2783, -0.2507, -0.3068, -0.2607, -0.2481, -0.2295, -0.2082, -0.2431, -0.2411, -0.2447, -0.2733, -0.2347, -0.2441, -0.3737, -0.3336, -0.3004, -0.299, -0.2978, -0.2583, -0.2667, -0.2449, -0.2291, -0.2633, -0.2848, -0.2685, -0.2623, -0.2567, -0.2571, -0.2994, -0.302, -0.2829, -0.2485, -0.2381, -0.2381, -0.2381, -0.2333, -0.1706, -0.1742, -0.1525, -0.171, -0.1265, -0.1281, -0.1165, -0.0912, -0.0904, -0.0868, -0.0692, -0.0311, 0.0106, 0.0, -0.0512, -0.0347, -0.0032, -0.0032, -0.0149, 0.0242, 0.0282, 0.027, 0.0516, 0.0536, 0.0073, 0.0411, 0.0754, 0.0556, 0.0315, 0.0278, 0.0685, 0.0903, 0.1032, 0.1371, 0.1452, 0.1145, 0.1597, 0.1468, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2051168254371274815", "ticker_idx": 0, "ts": "2026-05-04T05:13:15+00:00", "symbol": "KOSPI", "kind": "index", "geography": "Korea", "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "kept_implicit", "polarity_reversed": false, "quote": "7000가자~", "tweet_type": "prediction", "timeline": "unspecified", "summary": "Bullish on KOSPI continuing to rally toward 7000.", "resolved_tickers": ["EWY"], "resolution_method": "etf", "resolution_reason": "KOSPI index proxy ETF EWY", "bench_c": "SPY", "bench_c_reason": "country_fallback (suffix='')", "bench_c_industry": null, "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@blazingbees 7000가자~", "reply_text": null, "parent_text": "국장 코스피 장중 6,900pt 돌파! 역사적 신고가 하루만에 갱신!\n\n웅장하다! 연휴 끝나자마자 6,800pt 넘어 6,900pt까지 도달!\n\n웅장하다 시리지는 계속된다!! https://t.co/xV40tIJxX0", "quoted_text": null, "parent_author": "blazingbees", "ret_m1d": -0.00971939833415325, "ret_signed_m1d": -0.00971939833415325, "alpha_spy_m1d": -0.01339624769910741, "alpha_c_m1d": -0.01339624769910741, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.00367684936495416, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.060394920863043655, "ret_signed_p1d": 0.060394920863043655, "alpha_spy_p1d": 0.052372734978051394, "alpha_c_p1d": 0.052372734978051394, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.008022185884992261, "ret_p1w": 0.17886185231760288, "ret_signed_p1w": 0.17886185231760288, "alpha_spy_p1w": 0.14921048571020235, "alpha_c_p1w": 0.14921048571020235, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.02965136660740053, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2649, -0.2652, -0.2377, -0.2295, -0.2393, -0.2017, -0.2004, -0.1811, -0.1811, -0.2012, -0.1874, -0.1735, -0.1327, -0.1485, -0.1164, -0.0899, -0.0806, -0.0747, -0.0981, -0.191, -0.1786, -0.2314, -0.2253, -0.1816, -0.2035, -0.1879, -0.2451, -0.2411, -0.1864, -0.1783, -0.1938, -0.1759, -0.2311, -0.182, -0.213, -0.2191, -0.2665, -0.2605, -0.2883, -0.2481, -0.2284, -0.2489, -0.2489, -0.2301, -0.2225, -0.1438, -0.1485, -0.152, -0.1367, -0.1034, -0.1141, -0.0985, -0.0688, -0.0825, -0.1027, -0.0476, -0.0795, -0.0551, -0.0419, -0.0564, -0.0589, -0.0173, -0.0097, 0.0, 0.0604, 0.1124, 0.0804, 0.1627, 0.1789, 0.0911, 0.1531, 0.1646, 0.0933, 0.0765, 0.0638, 0.101, 0.1396, 0.1127, 0.1127, 0.2265, 0.2121, 0.2618, 0.2582, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "reply:2046129426178699543", "ticker_idx": 0, "ts": "2026-04-20T07:30:45+00:00", "symbol": "ABF substrates", "kind": "sector", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "ABF... 엄청난 상방이 있습니다", "tweet_type": "prediction", "timeline": "unspecified", "summary": "ABF 기판 섹터에 엄청난 상방 여력이 있다는 강한 강세 전망.", "resolved_tickers": ["4062.T", "3034.TW", "2383.TW", "3037.TW"], "resolution_method": "basket", "resolution_reason": "ABF substrate basket: Ibiden, Unimicron, EMC, Tripod", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "reply_other", "lang": "ko", "tweet_text": "@treeapple33 ABF... 엄청난 상방이 있습니다", "reply_text": null, "parent_text": "@jukan05 이러나 저러나 기판이 노나겠네요. MEC 화이팅 👀", "quoted_text": null, "parent_author": "treeapple33", "ret_m1d": -0.025029642561817528, "ret_signed_m1d": -0.025029642561817528, "alpha_spy_m1d": -0.02703331679707824, "alpha_c_m1d": -0.023671002173650396, "bench_spy_m1d": 0.0020036742352607106, "bench_c_m1d": -0.0013586403881671316, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.044312137008078456, "ret_signed_p1d": 0.044312137008078456, "alpha_spy_p1d": 0.05085908616301471, "alpha_c_p1d": 0.0434710080946078, "bench_spy_p1d": -0.006546949154936255, "bench_c_p1d": 0.0008411289134706568, "ret_p1w": 0.19822165278940973, "ret_signed_p1w": 0.19822165278940973, "alpha_spy_p1w": 0.18912072086472842, "alpha_c_p1w": 0.15933701342244422, "bench_spy_p1w": 0.00910093192468131, "bench_c_p1w": 0.03888463936696551, "ret_p1m": 0.2768258299258406, "ret_signed_p1m": 0.2768258299258406, "alpha_spy_p1m": 0.2415368430206002, "alpha_c_p1m": 0.1559665634297427, "bench_spy_p1m": 0.035288986905240405, "bench_c_p1m": 0.1208592664960979, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3362, -0.3179, -0.3055, -0.3192, -0.3048, -0.2894, -0.3099, -0.3068, -0.3261, -0.2939, -0.3154, -0.3357, -0.3451, -0.3096, -0.2929, -0.2891, -0.2739, -0.2703, -0.2667, -0.2624, -0.2522, -0.2443, -0.2568, -0.2429, -0.196, -0.1891, -0.1849, -0.1787, -0.1911, -0.2254, -0.2733, -0.2438, -0.2562, -0.2984, -0.2537, -0.2318, -0.2197, -0.1986, -0.1893, -0.1734, -0.1594, -0.1576, -0.1658, -0.2139, -0.2289, -0.1803, -0.167, -0.169, -0.1986, -0.2462, -0.1876, -0.1947, -0.1833, -0.1743, -0.1454, -0.074, -0.0637, -0.0471, -0.0705, -0.0468, -0.0358, -0.015, -0.025, 0.0, 0.0443, 0.0809, 0.07, 0.1611, 0.1982, 0.1647, 0.1584, 0.2191, 0.2163, 0.2442, 0.2423, 0.253, 0.3529, 0.3152, 0.3403, 0.3663, 0.3875, 0.3841, 0.2923, 0.2907, 0.2768, 0.2818, 0.3968, 0.4778, 0.5636, 0.5871, 0.5682, 0.5099, 0.6094, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2056146790488162706", "ticker_idx": 0, "ts": "2026-05-17T22:56:10+00:00", "symbol": "ABF substrates", "kind": "sector", "geography": "global", "sub_sector": "ABF substrates", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the supply-demand structure for high-end ABF substrates is expected to remain tight for an extended period", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish on high-end ABF substrates amid CPO/AI GPU buildout; TSMC CPO ramp and NVIDIA Vera Rubin drive 5–10x ABF consumption growth, supply-demand tight long-term.", "resolved_tickers": ["4062.T", "3034.TW", "2383.TW", "3037.TW"], "resolution_method": "basket", "resolution_reason": "ABF substrate basket: Ibiden, Unimicron, EMC, Tripod", "bench_c": "XLK", "bench_c_reason": "industry='Electronic Components'", "bench_c_industry": "Electronic Components", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF Substrates Emerge as a Key Battleground in CPO Integration\n\n- Amid explosive growth in ultra-high-speed transmission demand for AI data centers, TSMC's Co-Packaged Optics (CPO) strategy is also moving up a level.\n\n- TSMC recently disclosed that its \"COUPE on Substrate\" solution — which integrates a Compact Universal Photonic Engine (COUPE) — is on track to enter mass production in 2H 2026.\n\n- The industry reads this as more than just an optical communication upgrade. It signals that AI supply chain competition is expanding beyond advanced process and advanced packaging into the ABF substrate and CPO integration domain.\n\n- As NVIDIA's next-generation Vera Rubin platform continues to raise the integration level across AI GPUs, HBM, and ultra-high-speed networking interconnects, the importance of high-end substrates is set to grow rapidly.\n\n- In particular, semiconductor industry participants note that if CPO establishes itself as the mainstream AI server architecture going forward, NVIDIA may move to preemptively secure high-end ABF substrate capacity through long-term agreements (LTAs), prepayments, and strategic partnerships.\n\n- This would be a strategic move aimed at preventing a repeat of the past CoWoS and HBM supply shortages.\n\n- In addition, compared with conventional CPU substrates, substrates for AI GPUs and ASICs feature significantly larger area and substantially higher layer counts, expanding ABF material consumption by 5–10x.\n\n- With demand for AI GPUs, ASICs, and high-end networking chips continuing to grow, the supply-demand structure for high-end ABF substrates is expected to remain tight for an extended period.\n\n---\n\nhttps://t.co/brRfdlWTzA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.0010501307485989153, "ret_signed_m1d": 0.0010501307485989153, "alpha_spy_m1d": 0.0003461999574151109, "alpha_c_m1d": -0.009846828931565105, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.01089695968016402, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.009968828746970804, "ret_signed_p1d": -0.009968828746970804, "alpha_spy_p1d": -0.0033079693647276587, "alpha_c_p1d": -0.003545365147137547, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.006423463599833257, "ret_p1w": 0.19787770973468916, "ret_signed_p1w": 0.19787770973468916, "alpha_spy_p1w": 0.18841451335840814, "alpha_c_p1w": 0.16329409676044776, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.034583612974241396, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.4201, -0.4152, -0.4229, -0.4112, -0.3761, -0.3689, -0.3632, -0.3594, -0.369, -0.3935, -0.4298, -0.4059, -0.4149, -0.4464, -0.4123, -0.3938, -0.3828, -0.3662, -0.3605, -0.3492, -0.3373, -0.3347, -0.3413, -0.3775, -0.3918, -0.3547, -0.3431, -0.3435, -0.3657, -0.4034, -0.3581, -0.3626, -0.3557, -0.3502, -0.327, -0.2732, -0.2645, -0.2512, -0.2678, -0.2507, -0.2382, -0.2218, -0.2281, -0.2092, -0.1779, -0.1506, -0.1599, -0.0915, -0.0612, -0.0889, -0.0942, -0.0489, -0.0506, -0.027, -0.0285, -0.0189, 0.0473, 0.0218, 0.0419, 0.0583, 0.073, 0.0709, 0.0011, 0.0, -0.01, -0.0076, 0.0758, 0.1355, 0.1979, 0.217, 0.2043, 0.1593, 0.2234, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2056146790488162706", "ticker_idx": 1, "ts": "2026-05-17T22:56:10+00:00", "symbol": "2330.TW", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "TSMC's Co-Packaged Optics (CPO) strategy is also moving up a level… COUPE on Substrate solution… on track to enter mass production in 2H 2026", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish on high-end ABF substrates amid CPO/AI GPU buildout; TSMC CPO ramp and NVIDIA Vera Rubin drive 5–10x ABF consumption growth, supply-demand tight long-term.", "resolved_tickers": ["2330.TW"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF Substrates Emerge as a Key Battleground in CPO Integration\n\n- Amid explosive growth in ultra-high-speed transmission demand for AI data centers, TSMC's Co-Packaged Optics (CPO) strategy is also moving up a level.\n\n- TSMC recently disclosed that its \"COUPE on Substrate\" solution — which integrates a Compact Universal Photonic Engine (COUPE) — is on track to enter mass production in 2H 2026.\n\n- The industry reads this as more than just an optical communication upgrade. It signals that AI supply chain competition is expanding beyond advanced process and advanced packaging into the ABF substrate and CPO integration domain.\n\n- As NVIDIA's next-generation Vera Rubin platform continues to raise the integration level across AI GPUs, HBM, and ultra-high-speed networking interconnects, the importance of high-end substrates is set to grow rapidly.\n\n- In particular, semiconductor industry participants note that if CPO establishes itself as the mainstream AI server architecture going forward, NVIDIA may move to preemptively secure high-end ABF substrate capacity through long-term agreements (LTAs), prepayments, and strategic partnerships.\n\n- This would be a strategic move aimed at preventing a repeat of the past CoWoS and HBM supply shortages.\n\n- In addition, compared with conventional CPU substrates, substrates for AI GPUs and ASICs feature significantly larger area and substantially higher layer counts, expanding ABF material consumption by 5–10x.\n\n- With demand for AI GPUs, ASICs, and high-end networking chips continuing to grow, the supply-demand structure for high-end ABF substrates is expected to remain tight for an extended period.\n\n---\n\nhttps://t.co/brRfdlWTzA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.01116071428571419, "ret_signed_m1d": 0.01116071428571419, "alpha_spy_m1d": 0.010456783494530386, "alpha_c_m1d": -0.007478611573933813, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.018639325859648004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.015625, "ret_signed_p1d": -0.015625, "alpha_spy_p1d": -0.008964140617756855, "alpha_c_p1d": -0.011596965577761575, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.0040280344222384246, "ret_p1w": 0.03125, "ret_signed_p1w": 0.03125, "alpha_spy_p1w": 0.021786803623718987, "alpha_c_p1w": -0.02397197828969455, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.05522197828969455, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1479, -0.1479, -0.1479, -0.1545, -0.1256, -0.1034, -0.1123, -0.1123, -0.1212, -0.139, -0.1701, -0.1545, -0.159, -0.1946, -0.1768, -0.1367, -0.1612, -0.1701, -0.179, -0.1652, -0.1496, -0.1741, -0.1786, -0.192, -0.192, -0.1763, -0.1786, -0.1875, -0.2054, -0.2143, -0.1719, -0.192, -0.192, -0.192, -0.1696, -0.1295, -0.1272, -0.1071, -0.1116, -0.0826, -0.0714, -0.0692, -0.0938, -0.096, -0.0848, -0.0848, -0.0714, -0.0246, 0.0112, -0.0112, -0.0268, -0.0469, -0.0469, 0.0156, 0.0045, 0.0045, 0.0312, 0.0223, -0.0022, 0.0067, -0.0089, 0.0134, 0.0112, 0.0, -0.0156, -0.0246, -0.0045, 0.0067, 0.0312, 0.0134, 0.0268, 0.0246, 0.0513, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2056146790488162706", "ticker_idx": 2, "ts": "2026-05-17T22:56:10+00:00", "symbol": "NVDA", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "NVIDIA's next-generation Vera Rubin platform continues to raise the integration level across AI GPUs, HBM, and ultra-high-speed networking interconnects", "tweet_type": "analysis", "timeline": "years", "summary": "Bullish on high-end ABF substrates amid CPO/AI GPU buildout; TSMC CPO ramp and NVIDIA Vera Rubin drive 5–10x ABF consumption growth, supply-demand tight long-term.", "resolved_tickers": ["NVDA"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "▶ ABF Substrates Emerge as a Key Battleground in CPO Integration\n\n- Amid explosive growth in ultra-high-speed transmission demand for AI data centers, TSMC's Co-Packaged Optics (CPO) strategy is also moving up a level.\n\n- TSMC recently disclosed that its \"COUPE on Substrate\" solution — which integrates a Compact Universal Photonic Engine (COUPE) — is on track to enter mass production in 2H 2026.\n\n- The industry reads this as more than just an optical communication upgrade. It signals that AI supply chain competition is expanding beyond advanced process and advanced packaging into the ABF substrate and CPO integration domain.\n\n- As NVIDIA's next-generation Vera Rubin platform continues to raise the integration level across AI GPUs, HBM, and ultra-high-speed networking interconnects, the importance of high-end substrates is set to grow rapidly.\n\n- In particular, semiconductor industry participants note that if CPO establishes itself as the mainstream AI server architecture going forward, NVIDIA may move to preemptively secure high-end ABF substrate capacity through long-term agreements (LTAs), prepayments, and strategic partnerships.\n\n- This would be a strategic move aimed at preventing a repeat of the past CoWoS and HBM supply shortages.\n\n- In addition, compared with conventional CPU substrates, substrates for AI GPUs and ASICs feature significantly larger area and substantially higher layer counts, expanding ABF material consumption by 5–10x.\n\n- With demand for AI GPUs, ASICs, and high-end networking chips continuing to grow, the supply-demand structure for high-end ABF substrates is expected to remain tight for an extended period.\n\n---\n\nhttps://t.co/brRfdlWTzA", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.013494062167895615, "ret_signed_m1d": 0.013494062167895615, "alpha_spy_m1d": 0.01279013137671181, "alpha_c_m1d": -0.005145263691752389, "bench_spy_m1d": 0.0007039307911838044, "bench_c_m1d": 0.018639325859648004, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.00769164563481417, "ret_signed_p1d": -0.00769164563481417, "alpha_spy_p1d": -0.0010307862525710254, "alpha_c_p1d": -0.003663611212575746, "bench_spy_p1d": -0.006660859382243145, "bench_c_p1d": -0.0040280344222384246, "ret_p1w": -0.03144118955956243, "ret_signed_p1w": -0.03144118955956243, "alpha_spy_p1w": -0.040904385935843446, "alpha_c_p1w": -0.08666316784925698, "bench_spy_p1w": 0.009463196376281013, "bench_c_p1w": 0.05522197828969455, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.1545, -0.1549, -0.1462, -0.1385, -0.1326, -0.1204, -0.1684, -0.203, -0.1792, -0.1902, -0.1767, -0.1754, -0.2002, -0.1785, -0.1689, -0.1632, -0.1762, -0.1892, -0.1759, -0.1817, -0.1886, -0.1968, -0.2232, -0.21, -0.2119, -0.1963, -0.2298, -0.2465, -0.2571, -0.2155, -0.2095, -0.2021, -0.2021, -0.201, -0.1989, -0.181, -0.1728, -0.1515, -0.1485, -0.1161, -0.1055, -0.1078, -0.0928, -0.0911, -0.1009, -0.0892, -0.102, -0.0632, -0.0257, -0.0412, -0.0588, -0.1023, -0.1074, -0.1072, -0.1161, -0.0652, -0.0487, -0.032, -0.013, -0.0069, 0.0158, 0.0604, 0.0135, 0.0, -0.0077, 0.0052, -0.0126, -0.0314, -0.0314, -0.0336, -0.0437, -0.0363, -0.0503, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2044733024206754150", "ticker_idx": 0, "ts": "2026-04-16T11:01:56+00:00", "symbol": "TSM", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": false, "explicit_audit": "flipped_to_implicit", "polarity_reversed": false, "quote": "some idiots are selling because they think TSMC's margins peaked in Q1", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Implicitly bullish TSMC; dismisses sellers who think margins peaked in Q1 as wrong.", "resolved_tickers": ["TSM"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I heard some idiots are selling because they think TSMC’s margins peaked in Q1.", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.032337965605132135, "ret_signed_m1d": 0.032337965605132135, "alpha_spy_m1d": 0.034789253649135654, "alpha_c_m1d": 0.036295722512082906, "bench_spy_m1d": -0.002451288044003519, "bench_c_m1d": -0.003957756906950771, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.019677979293737602, "ret_signed_p1d": 0.019677979293737602, "alpha_spy_p1d": 0.007592294163513591, "alpha_c_p1d": -0.0009025310852588397, "bench_spy_p1d": 0.012085685130224011, "bench_c_p1d": 0.02058051037899644, "ret_p1w": 0.0531443435646799, "ret_signed_p1w": 0.0531443435646799, "alpha_spy_p1w": 0.0434672359593109, "alpha_c_p1w": -0.006332391344249899, "bench_spy_p1w": 0.009677107605368995, "bench_c_p1w": 0.059476734908929796, "ret_p1m": 0.11283885870726929, "ret_signed_p1m": 0.11283885870726929, "alpha_spy_p1m": 0.05937990251582281, "alpha_c_p1m": -0.11042420590987945, "bench_spy_p1m": 0.05345895619144647, "bench_c_p1m": 0.22326306461714873, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.0603, -0.1021, -0.105, -0.1016, -0.081, -0.0869, -0.0714, -0.0606, -0.0681, -0.0928, -0.0632, -0.0786, -0.106, -0.0923, -0.0426, -0.0246, -0.0068, 0.0267, 0.0102, 0.0055, 0.0055, -0.0005, -0.0058, -0.0109, 0.0169, 0.0156, 0.0587, 0.0641, 0.0341, 0.028, 0.013, -0.0309, -0.019, -0.0289, -0.0699, -0.043, -0.0474, -0.0269, -0.0759, -0.0715, -0.0663, -0.0478, -0.0654, -0.0676, -0.0939, -0.0685, -0.0553, -0.0429, -0.1025, -0.1008, -0.1289, -0.0699, -0.0602, -0.0669, -0.0669, -0.0594, -0.0496, 0.007, 0.0059, 0.02, 0.0171, 0.0455, 0.0323, 0.0, 0.0197, 0.008, 0.013, 0.0663, 0.0531, 0.1076, 0.1146, 0.0798, 0.0839, 0.09, 0.0945, 0.1053, 0.0855, 0.1545, 0.1398, 0.133, 0.1134, 0.0934, 0.1003, 0.1496, 0.1128, 0.0897, 0.0805, 0.1053, 0.1205, 0.1133, 0.1133, 0.1348, 0.1634, 0.1693, 0.1516, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050393020340765133", "ticker_idx": 0, "ts": "2026-05-02T01:52:45+00:00", "symbol": "MU", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "the supercycle is highly likely to extend from the previously expected 2026 into 2027", "tweet_type": "analysis", "timeline": "years", "summary": "AI CPU DRAM demand (300–400GB per unit) on top of HBM extends DRAM supercycle into 2027; bullish MU, Samsung, SK Hynix.", "resolved_tickers": ["MU"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI CPUs Are Devouring DRAM — Memory 'Shortage' to Last Another Year\n\nThe memory industry has posted record-breaking results driven by commodity DRAM prices that have surged more than 100%, and with the proliferation of AI-purposed central processing units (CPUs) now coming into play, forecasts suggest the shortage will extend for another year.\n\nIntel's recently unveiled \"AI CPU\" is expected to carry up to four times more commodity DRAM than previous generations. Combined with surging demand from graphics processing units (GPUs) that require high-capacity DRAM, observers expect the memory supply capacity of Samsung Electronics and SK hynix to fall short of demand.\n\nAccording to industry sources on the 2nd, CPU manufacturers are pursuing the integration of 300–400GB of DRAM into AI CPUs. This is up to four times the overwhelming scale compared to typical CPU products (96–256GB).\n\nCPUs Emerge as the 'AI Orchestrator'\n\nThe surge in high-capacity DRAM demand for AI CPUs is tied to the AI industry's pivot toward an inference-centric structure. Whereas AI inference was once limited to simple Q&A, it now serves as the \"orchestrator\" coordinating various agentic AIs.\n\nThe key in this process is \"context memory.\" For a CPU to coordinate the entire workflow by referencing the outputs of each agentic AI, it must remember the content. This makes scaling up memory — the space for retention — essential.\n\nUntil now, AI data centers have built computing infrastructure centered on GPUs equipped with High Bandwidth Memory (HBM). Leveraging the GPU's strength in training AI on vast amounts of data simultaneously, the focus has been on \"AI training.\" Server configurations accordingly followed an 8-GPU-to-1-CPU pattern. However, as the industry's center of gravity shifts toward inference, server configurations with substantially expanded CPU ratios are spreading.\n\nIn a recent earnings call, Intel executives explained: \"In AI inference infrastructure, the computing structure has shifted to a CPU-to-GPU ratio of 1 to 4, and the trend is moving further toward 1 to 1.\"\n\nAfter GPUs, CPUs Join the Memory Battle — Demand Snowballs\n\nThe competition for memory capacity has expanded from GPUs to CPUs, snowballing in scale. NVIDIA's next-generation AI chip \"Vera Rubin\" carries 288GB through 8 HBM stacks, while AMD's next-generation GPU MI400 boasts an even larger 432GB mammoth-class capacity.\n\nGoogle's recently unveiled custom chip — the 8th-generation Tensor Processing Unit (TPU 8i) — is also slated to feature 288GB of HBM. Add to this Intel's AI CPU \"Xeon\" and AMD's \"EPYC\" beginning to use up to 400GB of high-capacity DDR5, and the memory shortage is expected to persist longer.\n\nThe market temperature is already being proven in spot prices. According to Kiwoom Securities, while the price of legacy DDR4 (16GB basis) plunged 16% in a single month in April, the spot price of DDR5 (16GB basis) — used in AI CPUs — rose 2.8% over the same period, maintaining its price premium.\n\nAn industry source said: \"The current DRAM market is understood to be roughly 10 percentage points short of demand. With commodity DRAM demand surging on top of HBM, the supercycle is highly likely to extend from the previously expected 2026 into 2027.\"\n\n$MU $DRAM\n\n---\n\nhttps://t.co/XP6ENlQKbs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05939802152710816, "ret_signed_m1d": -0.05939802152710816, "alpha_spy_m1d": -0.06307487089206232, "alpha_c_m1d": -0.06537682680781298, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.11059068201928368, "ret_signed_p1d": 0.11059068201928368, "alpha_spy_p1d": 0.10256849613429142, "alpha_c_p1d": 0.07921675272234285, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.37970335718234316, "ret_signed_p1w": 0.37970335718234316, "alpha_spy_p1w": 0.35005199057494263, "alpha_c_p1w": 0.24252624666022515, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3421, -0.3361, -0.3156, -0.335, -0.3528, -0.2885, -0.2822, -0.2862, -0.2862, -0.3068, -0.2701, -0.2763, -0.2575, -0.27, -0.2752, -0.2561, -0.2794, -0.2849, -0.2844, -0.3416, -0.3051, -0.3115, -0.3579, -0.3249, -0.301, -0.274, -0.2971, -0.2611, -0.2339, -0.1994, -0.1993, -0.2296, -0.2667, -0.2988, -0.3141, -0.3374, -0.3836, -0.3806, -0.4418, -0.4139, -0.3619, -0.3647, -0.3647, -0.3447, -0.345, -0.2944, -0.2688, -0.2704, -0.26, -0.1922, -0.2086, -0.2068, -0.2106, -0.2221, -0.2204, -0.1543, -0.1643, -0.1383, -0.09, -0.1252, -0.1006, -0.1029, -0.0594, 0.0, 0.1106, 0.1564, 0.1217, 0.2955, 0.3797, 0.3298, 0.3941, 0.3462, 0.2571, 0.1823, 0.2121, 0.2698, 0.3221, 0.3028, 0.3028, 0.5541, 0.6106, 0.6021, 0.6844, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050393020340765133", "ticker_idx": 1, "ts": "2026-05-02T01:52:45+00:00", "symbol": "Samsung Electronics", "kind": "equity", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "observers expect the memory supply capacity of Samsung Electronics and SK hynix to fall short of demand", "tweet_type": "analysis", "timeline": "years", "summary": "AI CPU DRAM demand (300–400GB per unit) on top of HBM extends DRAM supercycle into 2027; bullish MU, Samsung, SK Hynix.", "resolved_tickers": ["005930.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "Samsung Electronics listed on KOSPI as 005930.KS", "bench_c": "XLK", "bench_c_reason": "industry='Consumer Electronics'", "bench_c_industry": "Consumer Electronics", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI CPUs Are Devouring DRAM — Memory 'Shortage' to Last Another Year\n\nThe memory industry has posted record-breaking results driven by commodity DRAM prices that have surged more than 100%, and with the proliferation of AI-purposed central processing units (CPUs) now coming into play, forecasts suggest the shortage will extend for another year.\n\nIntel's recently unveiled \"AI CPU\" is expected to carry up to four times more commodity DRAM than previous generations. Combined with surging demand from graphics processing units (GPUs) that require high-capacity DRAM, observers expect the memory supply capacity of Samsung Electronics and SK hynix to fall short of demand.\n\nAccording to industry sources on the 2nd, CPU manufacturers are pursuing the integration of 300–400GB of DRAM into AI CPUs. This is up to four times the overwhelming scale compared to typical CPU products (96–256GB).\n\nCPUs Emerge as the 'AI Orchestrator'\n\nThe surge in high-capacity DRAM demand for AI CPUs is tied to the AI industry's pivot toward an inference-centric structure. Whereas AI inference was once limited to simple Q&A, it now serves as the \"orchestrator\" coordinating various agentic AIs.\n\nThe key in this process is \"context memory.\" For a CPU to coordinate the entire workflow by referencing the outputs of each agentic AI, it must remember the content. This makes scaling up memory — the space for retention — essential.\n\nUntil now, AI data centers have built computing infrastructure centered on GPUs equipped with High Bandwidth Memory (HBM). Leveraging the GPU's strength in training AI on vast amounts of data simultaneously, the focus has been on \"AI training.\" Server configurations accordingly followed an 8-GPU-to-1-CPU pattern. However, as the industry's center of gravity shifts toward inference, server configurations with substantially expanded CPU ratios are spreading.\n\nIn a recent earnings call, Intel executives explained: \"In AI inference infrastructure, the computing structure has shifted to a CPU-to-GPU ratio of 1 to 4, and the trend is moving further toward 1 to 1.\"\n\nAfter GPUs, CPUs Join the Memory Battle — Demand Snowballs\n\nThe competition for memory capacity has expanded from GPUs to CPUs, snowballing in scale. NVIDIA's next-generation AI chip \"Vera Rubin\" carries 288GB through 8 HBM stacks, while AMD's next-generation GPU MI400 boasts an even larger 432GB mammoth-class capacity.\n\nGoogle's recently unveiled custom chip — the 8th-generation Tensor Processing Unit (TPU 8i) — is also slated to feature 288GB of HBM. Add to this Intel's AI CPU \"Xeon\" and AMD's \"EPYC\" beginning to use up to 400GB of high-capacity DDR5, and the memory shortage is expected to persist longer.\n\nThe market temperature is already being proven in spot prices. According to Kiwoom Securities, while the price of legacy DDR4 (16GB basis) plunged 16% in a single month in April, the spot price of DDR5 (16GB basis) — used in AI CPUs — rose 2.8% over the same period, maintaining its price premium.\n\nAn industry source said: \"The current DRAM market is understood to be roughly 10 percentage points short of demand. With commodity DRAM demand surging on top of HBM, the supercycle is highly likely to extend from the previously expected 2026 into 2027.\"\n\n$MU $DRAM\n\n---\n\nhttps://t.co/XP6ENlQKbs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.05161290322580647, "ret_signed_m1d": -0.05161290322580647, "alpha_spy_m1d": -0.05528975259076063, "alpha_c_m1d": -0.0505020860016141, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": -0.001110817224192373, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.022091957813239027, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.022091957813239027, "ret_p1w": 0.2279569892473119, "ret_signed_p1w": 0.2279569892473119, "alpha_spy_p1w": 0.19830562263991136, "alpha_c_p1w": 0.1302710803738718, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.09768590887344009, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2742, -0.3163, -0.3193, -0.2858, -0.2884, -0.2798, -0.2334, -0.2223, -0.2223, -0.2223, -0.2223, -0.1845, -0.1841, -0.1716, -0.1416, -0.1265, -0.0643, -0.0707, -0.0707, -0.1626, -0.2609, -0.1776, -0.1922, -0.2553, -0.1935, -0.1845, -0.1935, -0.2124, -0.1901, -0.1677, -0.1051, -0.1394, -0.1441, -0.2004, -0.1858, -0.1888, -0.227, -0.227, -0.2417, -0.2809, -0.1843, -0.2327, -0.1991, -0.1695, -0.1548, -0.0946, -0.1226, -0.114, -0.1355, -0.1118, -0.0925, -0.0645, -0.071, -0.0774, -0.0581, -0.0645, -0.0344, -0.0559, -0.0344, -0.0452, -0.028, -0.0516, -0.0516, 0.0, 0.0, 0.1441, 0.1677, 0.1548, 0.228, 0.2, 0.2215, 0.2731, 0.1634, 0.2086, 0.1849, 0.1871, 0.2882, 0.2581, 0.2581, 0.286, 0.3204, 0.2882, 0.3634, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2050393020340765133", "ticker_idx": 2, "ts": "2026-05-02T01:52:45+00:00", "symbol": "SK Hynix", "kind": "equity", "geography": "Korea", "sub_sector": "DRAM", "direction": "long", "conviction": "high", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "observers expect the memory supply capacity of Samsung Electronics and SK hynix to fall short of demand", "tweet_type": "analysis", "timeline": "years", "summary": "AI CPU DRAM demand (300–400GB per unit) on top of HBM extends DRAM supercycle into 2027; bullish MU, Samsung, SK Hynix.", "resolved_tickers": ["000660.KS"], "resolution_method": "exchange_ticker", "resolution_reason": "SK Hynix listed on KOSPI as 000660.KS", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "self_thread", "lang": "en", "tweet_text": "AI CPUs Are Devouring DRAM — Memory 'Shortage' to Last Another Year\n\nThe memory industry has posted record-breaking results driven by commodity DRAM prices that have surged more than 100%, and with the proliferation of AI-purposed central processing units (CPUs) now coming into play, forecasts suggest the shortage will extend for another year.\n\nIntel's recently unveiled \"AI CPU\" is expected to carry up to four times more commodity DRAM than previous generations. Combined with surging demand from graphics processing units (GPUs) that require high-capacity DRAM, observers expect the memory supply capacity of Samsung Electronics and SK hynix to fall short of demand.\n\nAccording to industry sources on the 2nd, CPU manufacturers are pursuing the integration of 300–400GB of DRAM into AI CPUs. This is up to four times the overwhelming scale compared to typical CPU products (96–256GB).\n\nCPUs Emerge as the 'AI Orchestrator'\n\nThe surge in high-capacity DRAM demand for AI CPUs is tied to the AI industry's pivot toward an inference-centric structure. Whereas AI inference was once limited to simple Q&A, it now serves as the \"orchestrator\" coordinating various agentic AIs.\n\nThe key in this process is \"context memory.\" For a CPU to coordinate the entire workflow by referencing the outputs of each agentic AI, it must remember the content. This makes scaling up memory — the space for retention — essential.\n\nUntil now, AI data centers have built computing infrastructure centered on GPUs equipped with High Bandwidth Memory (HBM). Leveraging the GPU's strength in training AI on vast amounts of data simultaneously, the focus has been on \"AI training.\" Server configurations accordingly followed an 8-GPU-to-1-CPU pattern. However, as the industry's center of gravity shifts toward inference, server configurations with substantially expanded CPU ratios are spreading.\n\nIn a recent earnings call, Intel executives explained: \"In AI inference infrastructure, the computing structure has shifted to a CPU-to-GPU ratio of 1 to 4, and the trend is moving further toward 1 to 1.\"\n\nAfter GPUs, CPUs Join the Memory Battle — Demand Snowballs\n\nThe competition for memory capacity has expanded from GPUs to CPUs, snowballing in scale. NVIDIA's next-generation AI chip \"Vera Rubin\" carries 288GB through 8 HBM stacks, while AMD's next-generation GPU MI400 boasts an even larger 432GB mammoth-class capacity.\n\nGoogle's recently unveiled custom chip — the 8th-generation Tensor Processing Unit (TPU 8i) — is also slated to feature 288GB of HBM. Add to this Intel's AI CPU \"Xeon\" and AMD's \"EPYC\" beginning to use up to 400GB of high-capacity DDR5, and the memory shortage is expected to persist longer.\n\nThe market temperature is already being proven in spot prices. According to Kiwoom Securities, while the price of legacy DDR4 (16GB basis) plunged 16% in a single month in April, the spot price of DDR5 (16GB basis) — used in AI CPUs — rose 2.8% over the same period, maintaining its price premium.\n\nAn industry source said: \"The current DRAM market is understood to be roughly 10 percentage points short of demand. With commodity DRAM demand surging on top of HBM, the supercycle is highly likely to extend from the previously expected 2026 into 2027.\"\n\n$MU $DRAM\n\n---\n\nhttps://t.co/XP6ENlQKbs", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.11126471123153359, "ret_signed_m1d": -0.11126471123153359, "alpha_spy_m1d": -0.11494156059648775, "alpha_c_m1d": -0.11724351651223841, "bench_spy_m1d": 0.00367684936495416, "bench_c_m1d": 0.005978805280704824, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.0, "ret_signed_p1d": 0.0, "alpha_spy_p1d": -0.008022185884992261, "alpha_c_p1d": -0.03137392929694083, "bench_spy_p1d": 0.008022185884992261, "bench_c_p1d": 0.03137392929694083, "ret_p1w": 0.2992397226039425, "ret_signed_p1w": 0.2992397226039425, "alpha_spy_p1w": 0.26958835599654196, "alpha_c_p1w": 0.1620626120818245, "bench_spy_p1w": 0.02965136660740053, "bench_c_p1w": 0.137177110522118, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.3792, -0.4192, -0.4212, -0.3881, -0.3957, -0.4068, -0.3874, -0.393, -0.393, -0.393, -0.393, -0.3833, -0.3454, -0.344, -0.3067, -0.2978, -0.2405, -0.2668, -0.2668, -0.3511, -0.4133, -0.3497, -0.3614, -0.4223, -0.3518, -0.34, -0.3573, -0.3711, -0.3269, -0.3296, -0.2702, -0.2999, -0.3041, -0.3552, -0.3186, -0.3124, -0.3552, -0.3552, -0.3967, -0.4423, -0.3801, -0.4264, -0.3946, -0.3877, -0.367, -0.2861, -0.3103, -0.2903, -0.2813, -0.2377, -0.2149, -0.2018, -0.2205, -0.1942, -0.1541, -0.1548, -0.1534, -0.1555, -0.1071, -0.1016, -0.1064, -0.1113, -0.1113, 0.0, 0.0, 0.1064, 0.1431, 0.1652, 0.2992, 0.2681, 0.3656, 0.3614, 0.2571, 0.2716, 0.2059, 0.2059, 0.3407, 0.3414, 0.3414, 0.4181, 0.5501, 0.5822, 0.6126, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "thread:2038929887386362165", "ticker_idx": 0, "ts": "2026-03-31T10:42:21+00:00", "symbol": "ONTO", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "Onto Innovation. It has already established a joint development program (JDP) with Samsung Electronics for non-destructive inspection equipment and has supplied tools to mass production lines.", "tweet_type": "news_reaction", "timeline": "months", "summary": "Onto Innovation named as the most advanced player supplying hybrid bonding inspection tools to Samsung's mass production lines, with broader adoption expected through year-end.", "resolved_tickers": ["ONTO"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductor Equipment & Materials'", "bench_c_industry": "Semiconductor Equipment & Materials", "unit_kind": "self_thread", "lang": "en", "tweet_text": "Samsung Electronics Introduces New Hybrid Bonding Inspection Equipment\n\nSamsung Electronics has begun adopting new inspection equipment for hybrid bonding. Onto Innovation's inspection tools are reportedly being validated on mass production lines.\n\nAccording to industry sources on the 31st, Samsung Electronics is in discussions or joint development with multiple companies possessing ultrasonic and X-ray-based non-destructive inspection (NDI) technologies. The primary objective is to detect micro-defects such as voids at hybrid bonding interfaces that are difficult to identify with conventional optical equipment. The inspection process also minimizes loss costs associated with expensive wafers and high-bandwidth memory (HBM).\n\nThe most advanced player is U.S.-based Onto Innovation. It has already established a joint development program (JDP) with Samsung Electronics for non-destructive inspection equipment and has supplied tools to mass production lines. The equipment utilizes laser-ultrasonic technology based on photoacoustic metrology. When a picosecond (one-trillionth of a second) ultra-short pulse pump laser is directed at the inspection target, the material undergoes momentary thermal expansion. This expansion generates ultrasonic waves that penetrate inward. These waves pass through optically opaque layers such as metal layers—which light cannot penetrate—and are reflected upon encountering underlying structures or voids. A probe laser then detects minute surface deformations to map the internal structure. This approach overcomes the limitations of conventional optical inspection and enables measurement of overlay alignment errors between upper and lower structures.\n\nDomestic (Korean) companies are expected to see equipment adoption or JDP signings materialize as early as Q2. Some have already received bonding wafer samples from Samsung Electronics and completed performance testing. Bonding wafers refer to the output of the hybrid bonding process. Test results are reported to have been positive. Other companies are also in ongoing discussions with Samsung Electronics, preparing non-destructive inspection equipment for production lines.\n\nAn industry source noted, \"Various departments within Samsung Electronics are in behind-the-scenes discussions with multiple equipment vendors,\" adding that \"both TSV (through-silicon via) inspection and hybrid bonding interface inspection for HBM are being considered.\"\n\nSamsung Electronics unveiled a video on hybrid bonding technology for HBM at NVIDIA's annual developer conference (GTC) 2026 this month. The industry expects Samsung to adopt hybrid bonding starting from next-generation HBM4E with 16 or more stacked layers. Last month, Samsung also presented a paper at the International Solid-State Circuits Conference (ISSCC) on fabricating 4F-square (4F²) DRAM using wafer-to-wafer (W2W) hybrid bonding. The approach applies a cell-over-periphery (COP) architecture to manufacture DRAM in a vertical channel transistor (VCT) structure—precisely stacking a periphery wafer and a cell array wafer on top of each other. Inspecting both of these next-generation memory types requires equipment capable of examining hybrid bonding interfaces.\n\nGenerally, qualification testing for back-end semiconductor equipment on production lines takes approximately six months. Considering this timeline, Samsung Electronics is expected to evaluate and compare inspection equipment performance from multiple vendors through the end of this year.\n\nHybrid bonding is an advanced packaging technology that directly connects copper (Cu) surfaces of two semiconductor dies when stacking and interconnecting chips. Unlike thermocompression (TC) bonding—primarily used in HBM manufacturing—it does not use micro bumps for chip-to-chip bonding. Compared to conventional soldering, hybrid bonding offers a larger bonding area and lower interfacial resistance, resulting in superior thermal dissipation and electrical characteristics.\n\n$ONTO\n\n---\n\nhttps://t.co/CT1pU4TimS", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": -0.07704687417686473, "ret_signed_m1d": -0.07704687417686473, "alpha_spy_m1d": -0.0488000257687069, "alpha_c_m1d": -0.022612874569597485, "bench_spy_m1d": -0.028246848408157832, "bench_c_m1d": -0.05443399960726725, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": 0.03242792050994869, "ret_signed_p1d": 0.03242792050994869, "alpha_spy_p1d": 0.02489345672705867, "alpha_c_p1d": 0.010075266726306209, "bench_spy_p1d": 0.007534463782890022, "bench_c_p1d": 0.022352653783642484, "ret_p1w": 0.06846439159645312, "ret_signed_p1w": 0.06846439159645312, "alpha_spy_p1w": 0.05481008234135065, "alpha_c_p1w": 0.025428397171124884, "bench_spy_p1w": 0.01365430925510247, "bench_c_p1w": 0.04303599442532824, "ret_p1m": 0.3865508664479873, "ret_signed_p1m": 0.3865508664479873, "alpha_spy_p1m": 0.29238475686607623, "alpha_c_p1m": 0.08352531995528412, "bench_spy_p1m": 0.09416610958191107, "bench_c_p1m": 0.3030255464927032, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.2302, -0.1912, -0.1311, -0.0868, -0.1027, -0.1204, -0.0752, -0.0685, -0.0372, -0.0139, 0.0623, 0.0741, 0.0741, 0.0541, 0.0561, 0.0328, 0.0332, 0.0081, 0.0165, 0.0421, 0.0393, -0.0147, -0.0129, -0.025, -0.0803, -0.079, 0.0167, 0.0439, 0.0616, 0.0846, 0.047, 0.0593, 0.0593, 0.0742, 0.0728, 0.0988, 0.0553, 0.0398, 0.0932, 0.0983, 0.0625, 0.0528, 0.0633, 0.0118, 0.0194, -0.0399, -0.1236, -0.0654, -0.0549, -0.0601, -0.0999, -0.0758, -0.0509, -0.0413, -0.0156, 0.0214, -0.0239, 0.023, 0.0702, 0.0779, -0.0016, -0.0041, -0.077, 0.0, 0.0324, 0.051, 0.051, 0.0785, 0.0685, 0.1727, 0.2043, 0.2539, 0.2318, 0.2756, 0.2625, 0.3027, 0.4179, 0.439, 0.4147, 0.4499, 0.4254, 0.5012, 0.4545, 0.377, 0.3866, 0.4388, 0.4284, 0.4607, 0.4873, 0.4339, 0.337, 0.3882, 0.4021, 0.3553, 0.3506, 0.3521, 0.3253, 0.2412, 0.2097, 0.2831, 0.2749, 0.2788, 0.2788, 0.337, 0.3015, 0.2621, 0.2593, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}
{"unit_id": "orig:2054171219851919759", "ticker_idx": 0, "ts": "2026-05-12T12:05:58+00:00", "symbol": "INTC", "kind": "equity", "geography": null, "sub_sector": null, "direction": "long", "conviction": "low", "explicit": true, "explicit_audit": "verified", "polarity_reversed": false, "quote": "AWS may be using EMIB for a lower-end version of Trainium through MediaTek", "tweet_type": "analysis", "timeline": "unspecified", "summary": "Intel's EMIB packaging technology being adopted by AWS Trainium and Google TPU programs via MediaTek; implied design-win tailwind for INTC.", "resolved_tickers": ["INTC"], "resolution_method": "identity", "resolution_reason": "structural marker; pre-classified as canonical yfinance ticker", "bench_c": "SMH", "bench_c_reason": "industry='Semiconductors'", "bench_c_industry": "Semiconductors", "unit_kind": "original", "lang": "en", "tweet_text": "I didn’t realize this, but it looks like AWS may be using EMIB for a lower-end version of Trainium through MediaTek.\n\nSimilar to the TPU v9 program handled by MediaTek, next-generation Trainium seems likely to use both EMIB and CoWoS.\n\n$INTC", "reply_text": null, "parent_text": null, "quoted_text": null, "parent_author": null, "ret_m1d": 0.07321119133048781, "ret_signed_m1d": 0.07321119133048781, "alpha_spy_m1d": 0.07169395281373037, "alpha_c_m1d": 0.04637823354234749, "bench_spy_m1d": 0.0015172385167574465, "bench_c_m1d": 0.02683295778814032, "ret_t0": 0.0, "ret_signed_t0": 0.0, "alpha_spy_t0": 0.0, "alpha_c_t0": 0.0, "bench_spy_t0": 0.0, "bench_c_t0": 0.0, "ret_p1d": -0.002653177126314965, "ret_signed_p1d": -0.002653177126314965, "alpha_spy_p1d": -0.008248025163095196, "alpha_c_p1d": -0.022626490217907413, "bench_spy_p1d": 0.005594848036780231, "bench_c_p1d": 0.01997331309159245, "ret_p1w": -0.0813365186050069, "ret_signed_p1w": -0.0813365186050069, "alpha_spy_p1w": -0.07530816203725343, "alpha_c_p1w": -0.05053032167432758, "bench_spy_p1w": -0.006028356567753468, "bench_c_p1w": -0.03080619693067932, "ret_p1m": null, "ret_signed_p1m": null, "alpha_spy_p1m": null, "alpha_c_p1m": null, "bench_spy_p1m": null, "bench_c_p1m": null, "ret_p3m": null, "ret_signed_p3m": null, "alpha_spy_p3m": null, "alpha_c_p3m": null, "bench_spy_p3m": null, "bench_c_p3m": null, "ret_p6m": null, "ret_signed_p6m": null, "alpha_spy_p6m": null, "alpha_c_p6m": null, "bench_spy_p6m": null, "bench_c_p6m": null, "price_path": [-0.6146, -0.6121, -0.6121, -0.6171, -0.6231, -0.63, -0.6343, -0.6383, -0.6176, -0.6113, -0.6231, -0.6218, -0.6228, -0.6426, -0.6221, -0.619, -0.64, -0.6221, -0.6121, -0.6022, -0.6248, -0.6205, -0.6206, -0.6347, -0.6266, -0.6171, -0.6363, -0.6351, -0.6347, -0.6088, -0.6344, -0.6424, -0.6585, -0.6341, -0.6018, -0.5823, -0.5823, -0.579, -0.5613, -0.5112, -0.4883, -0.4828, -0.4596, -0.4709, -0.4616, -0.4321, -0.4321, -0.4553, -0.4506, -0.4588, -0.4463, -0.3156, -0.2953, -0.2992, -0.2144, -0.2166, -0.174, -0.2059, -0.1033, -0.063, -0.0911, 0.0357, 0.0732, 0.0, -0.0027, -0.0388, -0.0982, -0.1031, -0.0813, -0.0137, -0.0175, -0.0064, -0.0064, 0.0241, 0.0096, 0.0023, -0.0492, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null]}